[PATCH 00/20] mm: Preemptibility -v10

[PATCH 00/20] mm: Preemptibility -v10

[Peter Zijlstra]

Another -rc1, another posting.

Rework the existing mmu_gather infrastructure.

The direct purpose of these patches was to allow preemptible mmu_gather,
but even without that I think these patches provide an improvement to the
status quo.

The first 9 patches rework the mmu_gather infrastructure. For review purpose
I've split them into generic and per-arch patches with the last of those a
generic cleanup.

Also provided is a rollup of these patches, which is used as a commit in the
git tree referenced below.

The next patch provides generic RCU page-table freeing, and the follow up
is a patch converting s390 to use this. I've also got 4 patches from
DaveM lined up (not included in this series) that uses this to implement
gup_fast() for sparc64.

Then there is one patch that extends the generic mmu_gather batching.

After that follow the mm preemptibility patches, these make part of the mm a
lot more preemptible. It converts i_mmap_lock and anon_vma->lock to mutexes
which together with the mmu_gather rework makes mmu_gather preemptible as well.

Making i_mmap_lock a mutex also enables a clean-up of the truncate code.

This also allows for preemptible mmu_notifiers, something that XPMEM I think
wants.

Furthermore, it removes the new and universially detested unmap_mutex.

git://git.kernel.org/pub/scm/linux/kernel/git/peterz/linux-2.6-mmu_gather.git

[PATCH 00/20] mm: Preemptibility -v10

[Peter Zijlstra]

Another -rc1, another posting.

Rework the existing mmu_gather infrastructure.

The direct purpose of these patches was to allow preemptible mmu_gather,
but even without that I think these patches provide an improvement to the
status quo.

The first 9 patches rework the mmu_gather infrastructure. For review purpose
I've split them into generic and per-arch patches with the last of those a
generic cleanup.

Also provided is a rollup of these patches, which is used as a commit in the
git tree referenced below.

The next patch provides generic RCU page-table freeing, and the follow up
is a patch converting s390 to use this. I've also got 4 patches from
DaveM lined up (not included in this series) that uses this to implement
gup_fast() for sparc64.

Then there is one patch that extends the generic mmu_gather batching.

After that follow the mm preemptibility patches, these make part of the mm a
lot more preemptible. It converts i_mmap_lock and anon_vma->lock to mutexes
which together with the mmu_gather rework makes mmu_gather preemptible as well.

Making i_mmap_lock a mutex also enables a clean-up of the truncate code.

This also allows for preemptible mmu_notifiers, something that XPMEM I think
wants.

Furthermore, it removes the new and universially detested unmap_mutex.

git://git.kernel.org/pub/scm/linux/kernel/git/peterz/linux-2.6-mmu_gather.git

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 00/20] mm: Preemptibility -v10

[Peter Zijlstra]

Another -rc1, another posting.

Rework the existing mmu_gather infrastructure.

The direct purpose of these patches was to allow preemptible mmu_gather,
but even without that I think these patches provide an improvement to the
status quo.

The first 9 patches rework the mmu_gather infrastructure. For review purpose
I've split them into generic and per-arch patches with the last of those a
generic cleanup.

Also provided is a rollup of these patches, which is used as a commit in the
git tree referenced below.

The next patch provides generic RCU page-table freeing, and the follow up
is a patch converting s390 to use this. I've also got 4 patches from
DaveM lined up (not included in this series) that uses this to implement
gup_fast() for sparc64.

Then there is one patch that extends the generic mmu_gather batching.

After that follow the mm preemptibility patches, these make part of the mm a
lot more preemptible. It converts i_mmap_lock and anon_vma->lock to mutexes
which together with the mmu_gather rework makes mmu_gather preemptible as well.

Making i_mmap_lock a mutex also enables a clean-up of the truncate code.

This also allows for preemptible mmu_notifiers, something that XPMEM I think
wants.

Furthermore, it removes the new and universially detested unmap_mutex.

git://git.kernel.org/pub/scm/linux/kernel/git/peterz/linux-2.6-mmu_gather.git

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 01/20] mm: mmu_gather rework

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-mm-preemptible_mmu_gather.patch --]
[-- Type: text/plain, Size: 14037 bytes --]

Remove the first obstackle towards a fully preemptible mmu_gather.

The current scheme assumes mmu_gather is always done with preemption
disabled and uses per-cpu storage for the page batches. Change this to
try and allocate a page for batching and in case of failure, use a
small on-stack array to make some progress.

Preemptible mmu_gather is desired in general and usable once
i_mmap_lock becomes a mutex. Doing it before the mutex conversion
saves us from having to rework the code by moving the mmu_gather
bits inside the pte_lock.

Also avoid flushing the tlb batches from under the pte lock,
this is useful even without the i_mmap_lock conversion as it
significantly reduces pte lock hold times.

Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Miller <davem@davemloft.net>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Russell King <rmk@arm.linux.org.uk>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: Jeff Dike <jdike@addtoit.com>
Cc: Tony Luck <tony.luck@intel.com>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Hugh Dickins <hughd@google.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 fs/exec.c                 |   10 ++--
 include/asm-generic/tlb.h |   96 +++++++++++++++++++++++++++++++++-------------
 include/linux/mm.h        |    2 
 mm/memory.c               |   46 +++++++++++-----------
 mm/mmap.c                 |   18 ++++----
 5 files changed, 107 insertions(+), 65 deletions(-)

Index: linux-2.6/fs/exec.c
===================================================================
--- linux-2.6.orig/fs/exec.c
+++ linux-2.6/fs/exec.c
@@ -553,7 +553,7 @@ static int shift_arg_pages(struct vm_are
 	unsigned long length = old_end - old_start;
 	unsigned long new_start = old_start - shift;
 	unsigned long new_end = old_end - shift;
-	struct mmu_gather *tlb;
+	struct mmu_gather tlb;
 
 	BUG_ON(new_start > new_end);
 
@@ -579,12 +579,12 @@ static int shift_arg_pages(struct vm_are
 		return -ENOMEM;
 
 	lru_add_drain();
-	tlb = tlb_gather_mmu(mm, 0);
+	tlb_gather_mmu(&tlb, mm, 0);
 	if (new_end > old_start) {
 		/*
 		 * when the old and new regions overlap clear from new_end.
 		 */
-		free_pgd_range(tlb, new_end, old_end, new_end,
+		free_pgd_range(&tlb, new_end, old_end, new_end,
 			vma->vm_next ? vma->vm_next->vm_start : 0);
 	} else {
 		/*
@@ -593,10 +593,10 @@ static int shift_arg_pages(struct vm_are
 		 * have constraints on va-space that make this illegal (IA64) -
 		 * for the others its just a little faster.
 		 */
-		free_pgd_range(tlb, old_start, old_end, new_end,
+		free_pgd_range(&tlb, old_start, old_end, new_end,
 			vma->vm_next ? vma->vm_next->vm_start : 0);
 	}
-	tlb_finish_mmu(tlb, new_end, old_end);
+	tlb_finish_mmu(&tlb, new_end, old_end);
 
 	/*
 	 * Shrink the vma to just the new range.  Always succeeds.
Index: linux-2.6/include/asm-generic/tlb.h
===================================================================
--- linux-2.6.orig/include/asm-generic/tlb.h
+++ linux-2.6/include/asm-generic/tlb.h
@@ -5,6 +5,8 @@
  * Copyright 2001 Red Hat, Inc.
  * Based on code from mm/memory.c Copyright Linus Torvalds and others.
  *
+ * Copyright 2011 Red Hat, Inc., Peter Zijlstra <pzijlstr@redhat.com>
+ *
  * This program is free software; you can redistribute it and/or
  * modify it under the terms of the GNU General Public License
  * as published by the Free Software Foundation; either version
@@ -22,51 +24,71 @@
  * and page free order so much..
  */
 #ifdef CONFIG_SMP
-  #ifdef ARCH_FREE_PTR_NR
-    #define FREE_PTR_NR   ARCH_FREE_PTR_NR
-  #else
-    #define FREE_PTE_NR	506
-  #endif
   #define tlb_fast_mode(tlb) ((tlb)->nr == ~0U)
 #else
-  #define FREE_PTE_NR	1
   #define tlb_fast_mode(tlb) 1
 #endif
 
+/*
+ * If we can't allocate a page to make a big patch of page pointers
+ * to work on, then just handle a few from the on-stack structure.
+ */
+#define MMU_GATHER_BUNDLE	8
+
 /* struct mmu_gather is an opaque type used by the mm code for passing around
  * any data needed by arch specific code for tlb_remove_page.
  */
 struct mmu_gather {
 	struct mm_struct	*mm;
 	unsigned int		nr;	/* set to ~0U means fast mode */
+	unsigned int		max;	/* nr < max */
 	unsigned int		need_flush;/* Really unmapped some ptes? */
 	unsigned int		fullmm; /* non-zero means full mm flush */
-	struct page *		pages[FREE_PTE_NR];
+#ifdef HAVE_ARCH_MMU_GATHER
+	struct arch_mmu_gather	arch;
+#endif
+	struct page		**pages;
+	struct page		*local[MMU_GATHER_BUNDLE];
 };
 
-/* Users of the generic TLB shootdown code must declare this storage space. */
-DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
+static inline void __tlb_alloc_page(struct mmu_gather *tlb)
+{
+	unsigned long addr = __get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
+
+	if (addr) {
+		tlb->pages = (void *)addr;
+		tlb->max = PAGE_SIZE / sizeof(struct page *);
+	}
+}
 
 /* tlb_gather_mmu
- *	Return a pointer to an initialized struct mmu_gather.
+ *	Called to initialize an (on-stack) mmu_gather structure for page-table
+ *	tear-down from @mm. The @fullmm argument is used when @mm is without
+ *	users and we're going to destroy the full address space (exit/execve).
  */
-static inline struct mmu_gather *
-tlb_gather_mmu(struct mm_struct *mm, unsigned int full_mm_flush)
+static inline void
+tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, bool fullmm)
 {
-	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);
-
 	tlb->mm = mm;
 
-	/* Use fast mode if only one CPU is online */
-	tlb->nr = num_online_cpus() > 1 ? 0U : ~0U;
+	tlb->max = ARRAY_SIZE(tlb->local);
+	tlb->pages = tlb->local;
+
+	if (num_online_cpus() > 1) {
+		tlb->nr = 0;
+		__tlb_alloc_page(tlb);
+	} else /* Use fast mode if only one CPU is online */
+		tlb->nr = ~0U;
 
-	tlb->fullmm = full_mm_flush;
+	tlb->fullmm = fullmm;
 
-	return tlb;
+#ifdef HAVE_ARCH_MMU_GATHER
+	tlb->arch = ARCH_MMU_GATHER_INIT;
+#endif
 }
 
 static inline void
-tlb_flush_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
+tlb_flush_mmu(struct mmu_gather *tlb)
 {
 	if (!tlb->need_flush)
 		return;
@@ -75,6 +97,13 @@ tlb_flush_mmu(struct mmu_gather *tlb, un
 	if (!tlb_fast_mode(tlb)) {
 		free_pages_and_swap_cache(tlb->pages, tlb->nr);
 		tlb->nr = 0;
+		/*
+		 * If we are using the local on-stack array of pages for MMU
+		 * gather, try allocating an off-stack array again as we have
+		 * recently freed pages.
+		 */
+		if (tlb->pages == tlb->local)
+			__tlb_alloc_page(tlb);
 	}
 }
 
@@ -85,29 +114,42 @@ tlb_flush_mmu(struct mmu_gather *tlb, un
 static inline void
 tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
 {
-	tlb_flush_mmu(tlb, start, end);
+	tlb_flush_mmu(tlb);
 
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
 
-	put_cpu_var(mmu_gathers);
+	if (tlb->pages != tlb->local)
+		free_pages((unsigned long)tlb->pages, 0);
 }
 
-/* tlb_remove_page
+/* __tlb_remove_page
  *	Must perform the equivalent to __free_pte(pte_get_and_clear(ptep)), while
  *	handling the additional races in SMP caused by other CPUs caching valid
- *	mappings in their TLBs.
+ *	mappings in their TLBs. Returns the number of free page slots left.
+ *	When out of page slots we must call tlb_flush_mmu().
  */
-static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
 {
 	tlb->need_flush = 1;
 	if (tlb_fast_mode(tlb)) {
 		free_page_and_swap_cache(page);
-		return;
+		return 1; /* avoid calling tlb_flush_mmu() */
 	}
 	tlb->pages[tlb->nr++] = page;
-	if (tlb->nr >= FREE_PTE_NR)
-		tlb_flush_mmu(tlb, 0, 0);
+	VM_BUG_ON(tlb->nr > tlb->max);
+
+	return tlb->max - tlb->nr;
+}
+
+/* tlb_remove_page
+ *	Similar to __tlb_remove_page but will call tlb_flush_mmu() itself when
+ *	required.
+ */
+static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	if (!__tlb_remove_page(tlb, page))
+		tlb_flush_mmu(tlb);
 }
 
 /**
Index: linux-2.6/include/linux/mm.h
===================================================================
--- linux-2.6.orig/include/linux/mm.h
+++ linux-2.6/include/linux/mm.h
@@ -904,7 +904,7 @@ int zap_vma_ptes(struct vm_area_struct *
 		unsigned long size);
 unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
 		unsigned long size, struct zap_details *);
-unsigned long unmap_vmas(struct mmu_gather **tlb,
+unsigned long unmap_vmas(struct mmu_gather *tlb,
 		struct vm_area_struct *start_vma, unsigned long start_addr,
 		unsigned long end_addr, unsigned long *nr_accounted,
 		struct zap_details *);
Index: linux-2.6/mm/memory.c
===================================================================
--- linux-2.6.orig/mm/memory.c
+++ linux-2.6/mm/memory.c
@@ -912,12 +912,13 @@ static unsigned long zap_pte_range(struc
 				long *zap_work, struct zap_details *details)
 {
 	struct mm_struct *mm = tlb->mm;
+	int force_flush = 0;
 	pte_t *pte;
 	spinlock_t *ptl;
 	int rss[NR_MM_COUNTERS];
 
 	init_rss_vec(rss);
-
+again:
 	pte = pte_offset_map_lock(mm, pmd, addr, &ptl);
 	arch_enter_lazy_mmu_mode();
 	do {
@@ -974,7 +975,9 @@ static unsigned long zap_pte_range(struc
 			page_remove_rmap(page);
 			if (unlikely(page_mapcount(page) < 0))
 				print_bad_pte(vma, addr, ptent, page);
-			tlb_remove_page(tlb, page);
+			force_flush = !__tlb_remove_page(tlb, page);
+			if (force_flush)
+				break;
 			continue;
 		}
 		/*
@@ -1001,6 +1004,18 @@ static unsigned long zap_pte_range(struc
 	arch_leave_lazy_mmu_mode();
 	pte_unmap_unlock(pte - 1, ptl);
 
+	/*
+	 * mmu_gather ran out of room to batch pages, we break out of
+	 * the PTE lock to avoid doing the potential expensive TLB invalidate
+	 * and page-free while holding it.
+	 */
+	if (force_flush) {
+		force_flush = 0;
+		tlb_flush_mmu(tlb);
+		if (addr != end)
+			goto again;
+	}
+
 	return addr;
 }
 
@@ -1121,17 +1136,14 @@ static unsigned long unmap_page_range(st
  * ensure that any thus-far unmapped pages are flushed before unmap_vmas()
  * drops the lock and schedules.
  */
-unsigned long unmap_vmas(struct mmu_gather **tlbp,
+unsigned long unmap_vmas(struct mmu_gather *tlb,
 		struct vm_area_struct *vma, unsigned long start_addr,
 		unsigned long end_addr, unsigned long *nr_accounted,
 		struct zap_details *details)
 {
 	long zap_work = ZAP_BLOCK_SIZE;
-	unsigned long tlb_start = 0;	/* For tlb_finish_mmu */
-	int tlb_start_valid = 0;
 	unsigned long start = start_addr;
 	spinlock_t *i_mmap_lock = details? details->i_mmap_lock: NULL;
-	int fullmm = (*tlbp)->fullmm;
 	struct mm_struct *mm = vma->vm_mm;
 
 	mmu_notifier_invalidate_range_start(mm, start_addr, end_addr);
@@ -1152,11 +1164,6 @@ unsigned long unmap_vmas(struct mmu_gath
 			untrack_pfn_vma(vma, 0, 0);
 
 		while (start != end) {
-			if (!tlb_start_valid) {
-				tlb_start = start;
-				tlb_start_valid = 1;
-			}
-
 			if (unlikely(is_vm_hugetlb_page(vma))) {
 				/*
 				 * It is undesirable to test vma->vm_file as it
@@ -1177,7 +1184,7 @@ unsigned long unmap_vmas(struct mmu_gath
 
 				start = end;
 			} else
-				start = unmap_page_range(*tlbp, vma,
+				start = unmap_page_range(tlb, vma,
 						start, end, &zap_work, details);
 
 			if (zap_work > 0) {
@@ -1185,19 +1192,13 @@ unsigned long unmap_vmas(struct mmu_gath
 				break;
 			}
 
-			tlb_finish_mmu(*tlbp, tlb_start, start);
-
 			if (need_resched() ||
 				(i_mmap_lock && spin_needbreak(i_mmap_lock))) {
-				if (i_mmap_lock) {
-					*tlbp = NULL;
+				if (i_mmap_lock)
 					goto out;
-				}
 				cond_resched();
 			}
 
-			*tlbp = tlb_gather_mmu(vma->vm_mm, fullmm);
-			tlb_start_valid = 0;
 			zap_work = ZAP_BLOCK_SIZE;
 		}
 	}
@@ -1217,16 +1218,15 @@ unsigned long zap_page_range(struct vm_a
 		unsigned long size, struct zap_details *details)
 {
 	struct mm_struct *mm = vma->vm_mm;
-	struct mmu_gather *tlb;
+	struct mmu_gather tlb;
 	unsigned long end = address + size;
 	unsigned long nr_accounted = 0;
 
 	lru_add_drain();
-	tlb = tlb_gather_mmu(mm, 0);
+	tlb_gather_mmu(&tlb, mm, 0);
 	update_hiwater_rss(mm);
 	end = unmap_vmas(&tlb, vma, address, end, &nr_accounted, details);
-	if (tlb)
-		tlb_finish_mmu(tlb, address, end);
+	tlb_finish_mmu(&tlb, address, end);
 	return end;
 }
 
Index: linux-2.6/mm/mmap.c
===================================================================
--- linux-2.6.orig/mm/mmap.c
+++ linux-2.6/mm/mmap.c
@@ -1913,17 +1913,17 @@ static void unmap_region(struct mm_struc
 		unsigned long start, unsigned long end)
 {
 	struct vm_area_struct *next = prev? prev->vm_next: mm->mmap;
-	struct mmu_gather *tlb;
+	struct mmu_gather tlb;
 	unsigned long nr_accounted = 0;
 
 	lru_add_drain();
-	tlb = tlb_gather_mmu(mm, 0);
+	tlb_gather_mmu(&tlb, mm, 0);
 	update_hiwater_rss(mm);
 	unmap_vmas(&tlb, vma, start, end, &nr_accounted, NULL);
 	vm_unacct_memory(nr_accounted);
-	free_pgtables(tlb, vma, prev? prev->vm_end: FIRST_USER_ADDRESS,
-				 next? next->vm_start: 0);
-	tlb_finish_mmu(tlb, start, end);
+	free_pgtables(&tlb, vma, prev ? prev->vm_end : FIRST_USER_ADDRESS,
+				 next ? next->vm_start : 0);
+	tlb_finish_mmu(&tlb, start, end);
 }
 
 /*
@@ -2265,7 +2265,7 @@ EXPORT_SYMBOL(do_brk);
 /* Release all mmaps. */
 void exit_mmap(struct mm_struct *mm)
 {
-	struct mmu_gather *tlb;
+	struct mmu_gather tlb;
 	struct vm_area_struct *vma;
 	unsigned long nr_accounted = 0;
 	unsigned long end;
@@ -2290,14 +2290,14 @@ void exit_mmap(struct mm_struct *mm)
 
 	lru_add_drain();
 	flush_cache_mm(mm);
-	tlb = tlb_gather_mmu(mm, 1);
+	tlb_gather_mmu(&tlb, mm, 1);
 	/* update_hiwater_rss(mm) here? but nobody should be looking */
 	/* Use -1 here to ensure all VMAs in the mm are unmapped */
 	end = unmap_vmas(&tlb, vma, 0, -1, &nr_accounted, NULL);
 	vm_unacct_memory(nr_accounted);
 
-	free_pgtables(tlb, vma, FIRST_USER_ADDRESS, 0);
-	tlb_finish_mmu(tlb, 0, end);
+	free_pgtables(&tlb, vma, FIRST_USER_ADDRESS, 0);
+	tlb_finish_mmu(&tlb, 0, end);
 
 	/*
 	 * Walk the list again, actually closing and freeing it,

[PATCH 01/20] mm: mmu_gather rework

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-mm-preemptible_mmu_gather.patch --]
[-- Type: text/plain, Size: 14340 bytes --]

Remove the first obstackle towards a fully preemptible mmu_gather.

The current scheme assumes mmu_gather is always done with preemption
disabled and uses per-cpu storage for the page batches. Change this to
try and allocate a page for batching and in case of failure, use a
small on-stack array to make some progress.

Preemptible mmu_gather is desired in general and usable once
i_mmap_lock becomes a mutex. Doing it before the mutex conversion
saves us from having to rework the code by moving the mmu_gather
bits inside the pte_lock.

Also avoid flushing the tlb batches from under the pte lock,
this is useful even without the i_mmap_lock conversion as it
significantly reduces pte lock hold times.

Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Miller <davem@davemloft.net>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Russell King <rmk@arm.linux.org.uk>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: Jeff Dike <jdike@addtoit.com>
Cc: Tony Luck <tony.luck@intel.com>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Hugh Dickins <hughd@google.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 fs/exec.c                 |   10 ++--
 include/asm-generic/tlb.h |   96 +++++++++++++++++++++++++++++++++-------------
 include/linux/mm.h        |    2 
 mm/memory.c               |   46 +++++++++++-----------
 mm/mmap.c                 |   18 ++++----
 5 files changed, 107 insertions(+), 65 deletions(-)

Index: linux-2.6/fs/exec.c
===================================================================
--- linux-2.6.orig/fs/exec.c
+++ linux-2.6/fs/exec.c
@@ -553,7 +553,7 @@ static int shift_arg_pages(struct vm_are
 	unsigned long length = old_end - old_start;
 	unsigned long new_start = old_start - shift;
 	unsigned long new_end = old_end - shift;
-	struct mmu_gather *tlb;
+	struct mmu_gather tlb;
 
 	BUG_ON(new_start > new_end);
 
@@ -579,12 +579,12 @@ static int shift_arg_pages(struct vm_are
 		return -ENOMEM;
 
 	lru_add_drain();
-	tlb = tlb_gather_mmu(mm, 0);
+	tlb_gather_mmu(&tlb, mm, 0);
 	if (new_end > old_start) {
 		/*
 		 * when the old and new regions overlap clear from new_end.
 		 */
-		free_pgd_range(tlb, new_end, old_end, new_end,
+		free_pgd_range(&tlb, new_end, old_end, new_end,
 			vma->vm_next ? vma->vm_next->vm_start : 0);
 	} else {
 		/*
@@ -593,10 +593,10 @@ static int shift_arg_pages(struct vm_are
 		 * have constraints on va-space that make this illegal (IA64) -
 		 * for the others its just a little faster.
 		 */
-		free_pgd_range(tlb, old_start, old_end, new_end,
+		free_pgd_range(&tlb, old_start, old_end, new_end,
 			vma->vm_next ? vma->vm_next->vm_start : 0);
 	}
-	tlb_finish_mmu(tlb, new_end, old_end);
+	tlb_finish_mmu(&tlb, new_end, old_end);
 
 	/*
 	 * Shrink the vma to just the new range.  Always succeeds.
Index: linux-2.6/include/asm-generic/tlb.h
===================================================================
--- linux-2.6.orig/include/asm-generic/tlb.h
+++ linux-2.6/include/asm-generic/tlb.h
@@ -5,6 +5,8 @@
  * Copyright 2001 Red Hat, Inc.
  * Based on code from mm/memory.c Copyright Linus Torvalds and others.
  *
+ * Copyright 2011 Red Hat, Inc., Peter Zijlstra <pzijlstr@redhat.com>
+ *
  * This program is free software; you can redistribute it and/or
  * modify it under the terms of the GNU General Public License
  * as published by the Free Software Foundation; either version
@@ -22,51 +24,71 @@
  * and page free order so much..
  */
 #ifdef CONFIG_SMP
-  #ifdef ARCH_FREE_PTR_NR
-    #define FREE_PTR_NR   ARCH_FREE_PTR_NR
-  #else
-    #define FREE_PTE_NR	506
-  #endif
   #define tlb_fast_mode(tlb) ((tlb)->nr == ~0U)
 #else
-  #define FREE_PTE_NR	1
   #define tlb_fast_mode(tlb) 1
 #endif
 
+/*
+ * If we can't allocate a page to make a big patch of page pointers
+ * to work on, then just handle a few from the on-stack structure.
+ */
+#define MMU_GATHER_BUNDLE	8
+
 /* struct mmu_gather is an opaque type used by the mm code for passing around
  * any data needed by arch specific code for tlb_remove_page.
  */
 struct mmu_gather {
 	struct mm_struct	*mm;
 	unsigned int		nr;	/* set to ~0U means fast mode */
+	unsigned int		max;	/* nr < max */
 	unsigned int		need_flush;/* Really unmapped some ptes? */
 	unsigned int		fullmm; /* non-zero means full mm flush */
-	struct page *		pages[FREE_PTE_NR];
+#ifdef HAVE_ARCH_MMU_GATHER
+	struct arch_mmu_gather	arch;
+#endif
+	struct page		**pages;
+	struct page		*local[MMU_GATHER_BUNDLE];
 };
 
-/* Users of the generic TLB shootdown code must declare this storage space. */
-DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
+static inline void __tlb_alloc_page(struct mmu_gather *tlb)
+{
+	unsigned long addr = __get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
+
+	if (addr) {
+		tlb->pages = (void *)addr;
+		tlb->max = PAGE_SIZE / sizeof(struct page *);
+	}
+}
 
 /* tlb_gather_mmu
- *	Return a pointer to an initialized struct mmu_gather.
+ *	Called to initialize an (on-stack) mmu_gather structure for page-table
+ *	tear-down from @mm. The @fullmm argument is used when @mm is without
+ *	users and we're going to destroy the full address space (exit/execve).
  */
-static inline struct mmu_gather *
-tlb_gather_mmu(struct mm_struct *mm, unsigned int full_mm_flush)
+static inline void
+tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, bool fullmm)
 {
-	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);
-
 	tlb->mm = mm;
 
-	/* Use fast mode if only one CPU is online */
-	tlb->nr = num_online_cpus() > 1 ? 0U : ~0U;
+	tlb->max = ARRAY_SIZE(tlb->local);
+	tlb->pages = tlb->local;
+
+	if (num_online_cpus() > 1) {
+		tlb->nr = 0;
+		__tlb_alloc_page(tlb);
+	} else /* Use fast mode if only one CPU is online */
+		tlb->nr = ~0U;
 
-	tlb->fullmm = full_mm_flush;
+	tlb->fullmm = fullmm;
 
-	return tlb;
+#ifdef HAVE_ARCH_MMU_GATHER
+	tlb->arch = ARCH_MMU_GATHER_INIT;
+#endif
 }
 
 static inline void
-tlb_flush_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
+tlb_flush_mmu(struct mmu_gather *tlb)
 {
 	if (!tlb->need_flush)
 		return;
@@ -75,6 +97,13 @@ tlb_flush_mmu(struct mmu_gather *tlb, un
 	if (!tlb_fast_mode(tlb)) {
 		free_pages_and_swap_cache(tlb->pages, tlb->nr);
 		tlb->nr = 0;
+		/*
+		 * If we are using the local on-stack array of pages for MMU
+		 * gather, try allocating an off-stack array again as we have
+		 * recently freed pages.
+		 */
+		if (tlb->pages == tlb->local)
+			__tlb_alloc_page(tlb);
 	}
 }
 
@@ -85,29 +114,42 @@ tlb_flush_mmu(struct mmu_gather *tlb, un
 static inline void
 tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
 {
-	tlb_flush_mmu(tlb, start, end);
+	tlb_flush_mmu(tlb);
 
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
 
-	put_cpu_var(mmu_gathers);
+	if (tlb->pages != tlb->local)
+		free_pages((unsigned long)tlb->pages, 0);
 }
 
-/* tlb_remove_page
+/* __tlb_remove_page
  *	Must perform the equivalent to __free_pte(pte_get_and_clear(ptep)), while
  *	handling the additional races in SMP caused by other CPUs caching valid
- *	mappings in their TLBs.
+ *	mappings in their TLBs. Returns the number of free page slots left.
+ *	When out of page slots we must call tlb_flush_mmu().
  */
-static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
 {
 	tlb->need_flush = 1;
 	if (tlb_fast_mode(tlb)) {
 		free_page_and_swap_cache(page);
-		return;
+		return 1; /* avoid calling tlb_flush_mmu() */
 	}
 	tlb->pages[tlb->nr++] = page;
-	if (tlb->nr >= FREE_PTE_NR)
-		tlb_flush_mmu(tlb, 0, 0);
+	VM_BUG_ON(tlb->nr > tlb->max);
+
+	return tlb->max - tlb->nr;
+}
+
+/* tlb_remove_page
+ *	Similar to __tlb_remove_page but will call tlb_flush_mmu() itself when
+ *	required.
+ */
+static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	if (!__tlb_remove_page(tlb, page))
+		tlb_flush_mmu(tlb);
 }
 
 /**
Index: linux-2.6/include/linux/mm.h
===================================================================
--- linux-2.6.orig/include/linux/mm.h
+++ linux-2.6/include/linux/mm.h
@@ -904,7 +904,7 @@ int zap_vma_ptes(struct vm_area_struct *
 		unsigned long size);
 unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
 		unsigned long size, struct zap_details *);
-unsigned long unmap_vmas(struct mmu_gather **tlb,
+unsigned long unmap_vmas(struct mmu_gather *tlb,
 		struct vm_area_struct *start_vma, unsigned long start_addr,
 		unsigned long end_addr, unsigned long *nr_accounted,
 		struct zap_details *);
Index: linux-2.6/mm/memory.c
===================================================================
--- linux-2.6.orig/mm/memory.c
+++ linux-2.6/mm/memory.c
@@ -912,12 +912,13 @@ static unsigned long zap_pte_range(struc
 				long *zap_work, struct zap_details *details)
 {
 	struct mm_struct *mm = tlb->mm;
+	int force_flush = 0;
 	pte_t *pte;
 	spinlock_t *ptl;
 	int rss[NR_MM_COUNTERS];
 
 	init_rss_vec(rss);
-
+again:
 	pte = pte_offset_map_lock(mm, pmd, addr, &ptl);
 	arch_enter_lazy_mmu_mode();
 	do {
@@ -974,7 +975,9 @@ static unsigned long zap_pte_range(struc
 			page_remove_rmap(page);
 			if (unlikely(page_mapcount(page) < 0))
 				print_bad_pte(vma, addr, ptent, page);
-			tlb_remove_page(tlb, page);
+			force_flush = !__tlb_remove_page(tlb, page);
+			if (force_flush)
+				break;
 			continue;
 		}
 		/*
@@ -1001,6 +1004,18 @@ static unsigned long zap_pte_range(struc
 	arch_leave_lazy_mmu_mode();
 	pte_unmap_unlock(pte - 1, ptl);
 
+	/*
+	 * mmu_gather ran out of room to batch pages, we break out of
+	 * the PTE lock to avoid doing the potential expensive TLB invalidate
+	 * and page-free while holding it.
+	 */
+	if (force_flush) {
+		force_flush = 0;
+		tlb_flush_mmu(tlb);
+		if (addr != end)
+			goto again;
+	}
+
 	return addr;
 }
 
@@ -1121,17 +1136,14 @@ static unsigned long unmap_page_range(st
  * ensure that any thus-far unmapped pages are flushed before unmap_vmas()
  * drops the lock and schedules.
  */
-unsigned long unmap_vmas(struct mmu_gather **tlbp,
+unsigned long unmap_vmas(struct mmu_gather *tlb,
 		struct vm_area_struct *vma, unsigned long start_addr,
 		unsigned long end_addr, unsigned long *nr_accounted,
 		struct zap_details *details)
 {
 	long zap_work = ZAP_BLOCK_SIZE;
-	unsigned long tlb_start = 0;	/* For tlb_finish_mmu */
-	int tlb_start_valid = 0;
 	unsigned long start = start_addr;
 	spinlock_t *i_mmap_lock = details? details->i_mmap_lock: NULL;
-	int fullmm = (*tlbp)->fullmm;
 	struct mm_struct *mm = vma->vm_mm;
 
 	mmu_notifier_invalidate_range_start(mm, start_addr, end_addr);
@@ -1152,11 +1164,6 @@ unsigned long unmap_vmas(struct mmu_gath
 			untrack_pfn_vma(vma, 0, 0);
 
 		while (start != end) {
-			if (!tlb_start_valid) {
-				tlb_start = start;
-				tlb_start_valid = 1;
-			}
-
 			if (unlikely(is_vm_hugetlb_page(vma))) {
 				/*
 				 * It is undesirable to test vma->vm_file as it
@@ -1177,7 +1184,7 @@ unsigned long unmap_vmas(struct mmu_gath
 
 				start = end;
 			} else
-				start = unmap_page_range(*tlbp, vma,
+				start = unmap_page_range(tlb, vma,
 						start, end, &zap_work, details);
 
 			if (zap_work > 0) {
@@ -1185,19 +1192,13 @@ unsigned long unmap_vmas(struct mmu_gath
 				break;
 			}
 
-			tlb_finish_mmu(*tlbp, tlb_start, start);
-
 			if (need_resched() ||
 				(i_mmap_lock && spin_needbreak(i_mmap_lock))) {
-				if (i_mmap_lock) {
-					*tlbp = NULL;
+				if (i_mmap_lock)
 					goto out;
-				}
 				cond_resched();
 			}
 
-			*tlbp = tlb_gather_mmu(vma->vm_mm, fullmm);
-			tlb_start_valid = 0;
 			zap_work = ZAP_BLOCK_SIZE;
 		}
 	}
@@ -1217,16 +1218,15 @@ unsigned long zap_page_range(struct vm_a
 		unsigned long size, struct zap_details *details)
 {
 	struct mm_struct *mm = vma->vm_mm;
-	struct mmu_gather *tlb;
+	struct mmu_gather tlb;
 	unsigned long end = address + size;
 	unsigned long nr_accounted = 0;
 
 	lru_add_drain();
-	tlb = tlb_gather_mmu(mm, 0);
+	tlb_gather_mmu(&tlb, mm, 0);
 	update_hiwater_rss(mm);
 	end = unmap_vmas(&tlb, vma, address, end, &nr_accounted, details);
-	if (tlb)
-		tlb_finish_mmu(tlb, address, end);
+	tlb_finish_mmu(&tlb, address, end);
 	return end;
 }
 
Index: linux-2.6/mm/mmap.c
===================================================================
--- linux-2.6.orig/mm/mmap.c
+++ linux-2.6/mm/mmap.c
@@ -1913,17 +1913,17 @@ static void unmap_region(struct mm_struc
 		unsigned long start, unsigned long end)
 {
 	struct vm_area_struct *next = prev? prev->vm_next: mm->mmap;
-	struct mmu_gather *tlb;
+	struct mmu_gather tlb;
 	unsigned long nr_accounted = 0;
 
 	lru_add_drain();
-	tlb = tlb_gather_mmu(mm, 0);
+	tlb_gather_mmu(&tlb, mm, 0);
 	update_hiwater_rss(mm);
 	unmap_vmas(&tlb, vma, start, end, &nr_accounted, NULL);
 	vm_unacct_memory(nr_accounted);
-	free_pgtables(tlb, vma, prev? prev->vm_end: FIRST_USER_ADDRESS,
-				 next? next->vm_start: 0);
-	tlb_finish_mmu(tlb, start, end);
+	free_pgtables(&tlb, vma, prev ? prev->vm_end : FIRST_USER_ADDRESS,
+				 next ? next->vm_start : 0);
+	tlb_finish_mmu(&tlb, start, end);
 }
 
 /*
@@ -2265,7 +2265,7 @@ EXPORT_SYMBOL(do_brk);
 /* Release all mmaps. */
 void exit_mmap(struct mm_struct *mm)
 {
-	struct mmu_gather *tlb;
+	struct mmu_gather tlb;
 	struct vm_area_struct *vma;
 	unsigned long nr_accounted = 0;
 	unsigned long end;
@@ -2290,14 +2290,14 @@ void exit_mmap(struct mm_struct *mm)
 
 	lru_add_drain();
 	flush_cache_mm(mm);
-	tlb = tlb_gather_mmu(mm, 1);
+	tlb_gather_mmu(&tlb, mm, 1);
 	/* update_hiwater_rss(mm) here? but nobody should be looking */
 	/* Use -1 here to ensure all VMAs in the mm are unmapped */
 	end = unmap_vmas(&tlb, vma, 0, -1, &nr_accounted, NULL);
 	vm_unacct_memory(nr_accounted);
 
-	free_pgtables(tlb, vma, FIRST_USER_ADDRESS, 0);
-	tlb_finish_mmu(tlb, 0, end);
+	free_pgtables(&tlb, vma, FIRST_USER_ADDRESS, 0);
+	tlb_finish_mmu(&tlb, 0, end);
 
 	/*
 	 * Walk the list again, actually closing and freeing it,


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 01/20] mm: mmu_gather rework

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-mm-preemptible_mmu_gather.patch --]
[-- Type: text/plain, Size: 14340 bytes --]

Remove the first obstackle towards a fully preemptible mmu_gather.

The current scheme assumes mmu_gather is always done with preemption
disabled and uses per-cpu storage for the page batches. Change this to
try and allocate a page for batching and in case of failure, use a
small on-stack array to make some progress.

Preemptible mmu_gather is desired in general and usable once
i_mmap_lock becomes a mutex. Doing it before the mutex conversion
saves us from having to rework the code by moving the mmu_gather
bits inside the pte_lock.

Also avoid flushing the tlb batches from under the pte lock,
this is useful even without the i_mmap_lock conversion as it
significantly reduces pte lock hold times.

Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Miller <davem@davemloft.net>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Russell King <rmk@arm.linux.org.uk>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: Jeff Dike <jdike@addtoit.com>
Cc: Tony Luck <tony.luck@intel.com>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Hugh Dickins <hughd@google.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 fs/exec.c                 |   10 ++--
 include/asm-generic/tlb.h |   96 +++++++++++++++++++++++++++++++++-------------
 include/linux/mm.h        |    2 
 mm/memory.c               |   46 +++++++++++-----------
 mm/mmap.c                 |   18 ++++----
 5 files changed, 107 insertions(+), 65 deletions(-)

Index: linux-2.6/fs/exec.c
===================================================================
--- linux-2.6.orig/fs/exec.c
+++ linux-2.6/fs/exec.c
@@ -553,7 +553,7 @@ static int shift_arg_pages(struct vm_are
 	unsigned long length = old_end - old_start;
 	unsigned long new_start = old_start - shift;
 	unsigned long new_end = old_end - shift;
-	struct mmu_gather *tlb;
+	struct mmu_gather tlb;
 
 	BUG_ON(new_start > new_end);
 
@@ -579,12 +579,12 @@ static int shift_arg_pages(struct vm_are
 		return -ENOMEM;
 
 	lru_add_drain();
-	tlb = tlb_gather_mmu(mm, 0);
+	tlb_gather_mmu(&tlb, mm, 0);
 	if (new_end > old_start) {
 		/*
 		 * when the old and new regions overlap clear from new_end.
 		 */
-		free_pgd_range(tlb, new_end, old_end, new_end,
+		free_pgd_range(&tlb, new_end, old_end, new_end,
 			vma->vm_next ? vma->vm_next->vm_start : 0);
 	} else {
 		/*
@@ -593,10 +593,10 @@ static int shift_arg_pages(struct vm_are
 		 * have constraints on va-space that make this illegal (IA64) -
 		 * for the others its just a little faster.
 		 */
-		free_pgd_range(tlb, old_start, old_end, new_end,
+		free_pgd_range(&tlb, old_start, old_end, new_end,
 			vma->vm_next ? vma->vm_next->vm_start : 0);
 	}
-	tlb_finish_mmu(tlb, new_end, old_end);
+	tlb_finish_mmu(&tlb, new_end, old_end);
 
 	/*
 	 * Shrink the vma to just the new range.  Always succeeds.
Index: linux-2.6/include/asm-generic/tlb.h
===================================================================
--- linux-2.6.orig/include/asm-generic/tlb.h
+++ linux-2.6/include/asm-generic/tlb.h
@@ -5,6 +5,8 @@
  * Copyright 2001 Red Hat, Inc.
  * Based on code from mm/memory.c Copyright Linus Torvalds and others.
  *
+ * Copyright 2011 Red Hat, Inc., Peter Zijlstra <pzijlstr@redhat.com>
+ *
  * This program is free software; you can redistribute it and/or
  * modify it under the terms of the GNU General Public License
  * as published by the Free Software Foundation; either version
@@ -22,51 +24,71 @@
  * and page free order so much..
  */
 #ifdef CONFIG_SMP
-  #ifdef ARCH_FREE_PTR_NR
-    #define FREE_PTR_NR   ARCH_FREE_PTR_NR
-  #else
-    #define FREE_PTE_NR	506
-  #endif
   #define tlb_fast_mode(tlb) ((tlb)->nr == ~0U)
 #else
-  #define FREE_PTE_NR	1
   #define tlb_fast_mode(tlb) 1
 #endif
 
+/*
+ * If we can't allocate a page to make a big patch of page pointers
+ * to work on, then just handle a few from the on-stack structure.
+ */
+#define MMU_GATHER_BUNDLE	8
+
 /* struct mmu_gather is an opaque type used by the mm code for passing around
  * any data needed by arch specific code for tlb_remove_page.
  */
 struct mmu_gather {
 	struct mm_struct	*mm;
 	unsigned int		nr;	/* set to ~0U means fast mode */
+	unsigned int		max;	/* nr < max */
 	unsigned int		need_flush;/* Really unmapped some ptes? */
 	unsigned int		fullmm; /* non-zero means full mm flush */
-	struct page *		pages[FREE_PTE_NR];
+#ifdef HAVE_ARCH_MMU_GATHER
+	struct arch_mmu_gather	arch;
+#endif
+	struct page		**pages;
+	struct page		*local[MMU_GATHER_BUNDLE];
 };
 
-/* Users of the generic TLB shootdown code must declare this storage space. */
-DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
+static inline void __tlb_alloc_page(struct mmu_gather *tlb)
+{
+	unsigned long addr = __get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
+
+	if (addr) {
+		tlb->pages = (void *)addr;
+		tlb->max = PAGE_SIZE / sizeof(struct page *);
+	}
+}
 
 /* tlb_gather_mmu
- *	Return a pointer to an initialized struct mmu_gather.
+ *	Called to initialize an (on-stack) mmu_gather structure for page-table
+ *	tear-down from @mm. The @fullmm argument is used when @mm is without
+ *	users and we're going to destroy the full address space (exit/execve).
  */
-static inline struct mmu_gather *
-tlb_gather_mmu(struct mm_struct *mm, unsigned int full_mm_flush)
+static inline void
+tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, bool fullmm)
 {
-	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);
-
 	tlb->mm = mm;
 
-	/* Use fast mode if only one CPU is online */
-	tlb->nr = num_online_cpus() > 1 ? 0U : ~0U;
+	tlb->max = ARRAY_SIZE(tlb->local);
+	tlb->pages = tlb->local;
+
+	if (num_online_cpus() > 1) {
+		tlb->nr = 0;
+		__tlb_alloc_page(tlb);
+	} else /* Use fast mode if only one CPU is online */
+		tlb->nr = ~0U;
 
-	tlb->fullmm = full_mm_flush;
+	tlb->fullmm = fullmm;
 
-	return tlb;
+#ifdef HAVE_ARCH_MMU_GATHER
+	tlb->arch = ARCH_MMU_GATHER_INIT;
+#endif
 }
 
 static inline void
-tlb_flush_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
+tlb_flush_mmu(struct mmu_gather *tlb)
 {
 	if (!tlb->need_flush)
 		return;
@@ -75,6 +97,13 @@ tlb_flush_mmu(struct mmu_gather *tlb, un
 	if (!tlb_fast_mode(tlb)) {
 		free_pages_and_swap_cache(tlb->pages, tlb->nr);
 		tlb->nr = 0;
+		/*
+		 * If we are using the local on-stack array of pages for MMU
+		 * gather, try allocating an off-stack array again as we have
+		 * recently freed pages.
+		 */
+		if (tlb->pages == tlb->local)
+			__tlb_alloc_page(tlb);
 	}
 }
 
@@ -85,29 +114,42 @@ tlb_flush_mmu(struct mmu_gather *tlb, un
 static inline void
 tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
 {
-	tlb_flush_mmu(tlb, start, end);
+	tlb_flush_mmu(tlb);
 
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
 
-	put_cpu_var(mmu_gathers);
+	if (tlb->pages != tlb->local)
+		free_pages((unsigned long)tlb->pages, 0);
 }
 
-/* tlb_remove_page
+/* __tlb_remove_page
  *	Must perform the equivalent to __free_pte(pte_get_and_clear(ptep)), while
  *	handling the additional races in SMP caused by other CPUs caching valid
- *	mappings in their TLBs.
+ *	mappings in their TLBs. Returns the number of free page slots left.
+ *	When out of page slots we must call tlb_flush_mmu().
  */
-static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
 {
 	tlb->need_flush = 1;
 	if (tlb_fast_mode(tlb)) {
 		free_page_and_swap_cache(page);
-		return;
+		return 1; /* avoid calling tlb_flush_mmu() */
 	}
 	tlb->pages[tlb->nr++] = page;
-	if (tlb->nr >= FREE_PTE_NR)
-		tlb_flush_mmu(tlb, 0, 0);
+	VM_BUG_ON(tlb->nr > tlb->max);
+
+	return tlb->max - tlb->nr;
+}
+
+/* tlb_remove_page
+ *	Similar to __tlb_remove_page but will call tlb_flush_mmu() itself when
+ *	required.
+ */
+static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	if (!__tlb_remove_page(tlb, page))
+		tlb_flush_mmu(tlb);
 }
 
 /**
Index: linux-2.6/include/linux/mm.h
===================================================================
--- linux-2.6.orig/include/linux/mm.h
+++ linux-2.6/include/linux/mm.h
@@ -904,7 +904,7 @@ int zap_vma_ptes(struct vm_area_struct *
 		unsigned long size);
 unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
 		unsigned long size, struct zap_details *);
-unsigned long unmap_vmas(struct mmu_gather **tlb,
+unsigned long unmap_vmas(struct mmu_gather *tlb,
 		struct vm_area_struct *start_vma, unsigned long start_addr,
 		unsigned long end_addr, unsigned long *nr_accounted,
 		struct zap_details *);
Index: linux-2.6/mm/memory.c
===================================================================
--- linux-2.6.orig/mm/memory.c
+++ linux-2.6/mm/memory.c
@@ -912,12 +912,13 @@ static unsigned long zap_pte_range(struc
 				long *zap_work, struct zap_details *details)
 {
 	struct mm_struct *mm = tlb->mm;
+	int force_flush = 0;
 	pte_t *pte;
 	spinlock_t *ptl;
 	int rss[NR_MM_COUNTERS];
 
 	init_rss_vec(rss);
-
+again:
 	pte = pte_offset_map_lock(mm, pmd, addr, &ptl);
 	arch_enter_lazy_mmu_mode();
 	do {
@@ -974,7 +975,9 @@ static unsigned long zap_pte_range(struc
 			page_remove_rmap(page);
 			if (unlikely(page_mapcount(page) < 0))
 				print_bad_pte(vma, addr, ptent, page);
-			tlb_remove_page(tlb, page);
+			force_flush = !__tlb_remove_page(tlb, page);
+			if (force_flush)
+				break;
 			continue;
 		}
 		/*
@@ -1001,6 +1004,18 @@ static unsigned long zap_pte_range(struc
 	arch_leave_lazy_mmu_mode();
 	pte_unmap_unlock(pte - 1, ptl);
 
+	/*
+	 * mmu_gather ran out of room to batch pages, we break out of
+	 * the PTE lock to avoid doing the potential expensive TLB invalidate
+	 * and page-free while holding it.
+	 */
+	if (force_flush) {
+		force_flush = 0;
+		tlb_flush_mmu(tlb);
+		if (addr != end)
+			goto again;
+	}
+
 	return addr;
 }
 
@@ -1121,17 +1136,14 @@ static unsigned long unmap_page_range(st
  * ensure that any thus-far unmapped pages are flushed before unmap_vmas()
  * drops the lock and schedules.
  */
-unsigned long unmap_vmas(struct mmu_gather **tlbp,
+unsigned long unmap_vmas(struct mmu_gather *tlb,
 		struct vm_area_struct *vma, unsigned long start_addr,
 		unsigned long end_addr, unsigned long *nr_accounted,
 		struct zap_details *details)
 {
 	long zap_work = ZAP_BLOCK_SIZE;
-	unsigned long tlb_start = 0;	/* For tlb_finish_mmu */
-	int tlb_start_valid = 0;
 	unsigned long start = start_addr;
 	spinlock_t *i_mmap_lock = details? details->i_mmap_lock: NULL;
-	int fullmm = (*tlbp)->fullmm;
 	struct mm_struct *mm = vma->vm_mm;
 
 	mmu_notifier_invalidate_range_start(mm, start_addr, end_addr);
@@ -1152,11 +1164,6 @@ unsigned long unmap_vmas(struct mmu_gath
 			untrack_pfn_vma(vma, 0, 0);
 
 		while (start != end) {
-			if (!tlb_start_valid) {
-				tlb_start = start;
-				tlb_start_valid = 1;
-			}
-
 			if (unlikely(is_vm_hugetlb_page(vma))) {
 				/*
 				 * It is undesirable to test vma->vm_file as it
@@ -1177,7 +1184,7 @@ unsigned long unmap_vmas(struct mmu_gath
 
 				start = end;
 			} else
-				start = unmap_page_range(*tlbp, vma,
+				start = unmap_page_range(tlb, vma,
 						start, end, &zap_work, details);
 
 			if (zap_work > 0) {
@@ -1185,19 +1192,13 @@ unsigned long unmap_vmas(struct mmu_gath
 				break;
 			}
 
-			tlb_finish_mmu(*tlbp, tlb_start, start);
-
 			if (need_resched() ||
 				(i_mmap_lock && spin_needbreak(i_mmap_lock))) {
-				if (i_mmap_lock) {
-					*tlbp = NULL;
+				if (i_mmap_lock)
 					goto out;
-				}
 				cond_resched();
 			}
 
-			*tlbp = tlb_gather_mmu(vma->vm_mm, fullmm);
-			tlb_start_valid = 0;
 			zap_work = ZAP_BLOCK_SIZE;
 		}
 	}
@@ -1217,16 +1218,15 @@ unsigned long zap_page_range(struct vm_a
 		unsigned long size, struct zap_details *details)
 {
 	struct mm_struct *mm = vma->vm_mm;
-	struct mmu_gather *tlb;
+	struct mmu_gather tlb;
 	unsigned long end = address + size;
 	unsigned long nr_accounted = 0;
 
 	lru_add_drain();
-	tlb = tlb_gather_mmu(mm, 0);
+	tlb_gather_mmu(&tlb, mm, 0);
 	update_hiwater_rss(mm);
 	end = unmap_vmas(&tlb, vma, address, end, &nr_accounted, details);
-	if (tlb)
-		tlb_finish_mmu(tlb, address, end);
+	tlb_finish_mmu(&tlb, address, end);
 	return end;
 }
 
Index: linux-2.6/mm/mmap.c
===================================================================
--- linux-2.6.orig/mm/mmap.c
+++ linux-2.6/mm/mmap.c
@@ -1913,17 +1913,17 @@ static void unmap_region(struct mm_struc
 		unsigned long start, unsigned long end)
 {
 	struct vm_area_struct *next = prev? prev->vm_next: mm->mmap;
-	struct mmu_gather *tlb;
+	struct mmu_gather tlb;
 	unsigned long nr_accounted = 0;
 
 	lru_add_drain();
-	tlb = tlb_gather_mmu(mm, 0);
+	tlb_gather_mmu(&tlb, mm, 0);
 	update_hiwater_rss(mm);
 	unmap_vmas(&tlb, vma, start, end, &nr_accounted, NULL);
 	vm_unacct_memory(nr_accounted);
-	free_pgtables(tlb, vma, prev? prev->vm_end: FIRST_USER_ADDRESS,
-				 next? next->vm_start: 0);
-	tlb_finish_mmu(tlb, start, end);
+	free_pgtables(&tlb, vma, prev ? prev->vm_end : FIRST_USER_ADDRESS,
+				 next ? next->vm_start : 0);
+	tlb_finish_mmu(&tlb, start, end);
 }
 
 /*
@@ -2265,7 +2265,7 @@ EXPORT_SYMBOL(do_brk);
 /* Release all mmaps. */
 void exit_mmap(struct mm_struct *mm)
 {
-	struct mmu_gather *tlb;
+	struct mmu_gather tlb;
 	struct vm_area_struct *vma;
 	unsigned long nr_accounted = 0;
 	unsigned long end;
@@ -2290,14 +2290,14 @@ void exit_mmap(struct mm_struct *mm)
 
 	lru_add_drain();
 	flush_cache_mm(mm);
-	tlb = tlb_gather_mmu(mm, 1);
+	tlb_gather_mmu(&tlb, mm, 1);
 	/* update_hiwater_rss(mm) here? but nobody should be looking */
 	/* Use -1 here to ensure all VMAs in the mm are unmapped */
 	end = unmap_vmas(&tlb, vma, 0, -1, &nr_accounted, NULL);
 	vm_unacct_memory(nr_accounted);
 
-	free_pgtables(tlb, vma, FIRST_USER_ADDRESS, 0);
-	tlb_finish_mmu(tlb, 0, end);
+	free_pgtables(&tlb, vma, FIRST_USER_ADDRESS, 0);
+	tlb_finish_mmu(&tlb, 0, end);
 
 	/*
 	 * Walk the list again, actually closing and freeing it,


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 02/20] powerpc: mmu_gather rework

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-powerpc-preemptible_mmu_gather.patch --]
[-- Type: text/plain, Size: 7834 bytes --]

Fix up powerpc to the new mmu_gather stuffs.

PPC has an extra batching queue to RCU free the actual pagetable
allocations, use the ARCH extentions for that for now.

For the ppc64_tlb_batch, which tracks the vaddrs to unhash from the
hardware hash-table, keep using per-cpu arrays but flush on context
switch and use a TLF bit to track the lazy_mmu state.

Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 arch/powerpc/include/asm/pgalloc.h     |    4 ++--
 arch/powerpc/include/asm/thread_info.h |    2 ++
 arch/powerpc/include/asm/tlb.h         |   10 ++++++++++
 arch/powerpc/kernel/process.c          |   23 ++++++++++++++++++++++-
 arch/powerpc/mm/pgtable.c              |   14 ++++----------
 arch/powerpc/mm/tlb_hash32.c           |    2 +-
 arch/powerpc/mm/tlb_hash64.c           |    6 ++++--
 arch/powerpc/mm/tlb_nohash.c           |    2 +-
 8 files changed, 46 insertions(+), 17 deletions(-)

Index: linux-2.6/arch/powerpc/include/asm/tlb.h
===================================================================
--- linux-2.6.orig/arch/powerpc/include/asm/tlb.h
+++ linux-2.6/arch/powerpc/include/asm/tlb.h
@@ -28,6 +28,16 @@
 #define tlb_start_vma(tlb, vma)	do { } while (0)
 #define tlb_end_vma(tlb, vma)	do { } while (0)
 
+#define HAVE_ARCH_MMU_GATHER 1
+
+struct pte_freelist_batch;
+
+struct arch_mmu_gather {
+	struct pte_freelist_batch *batch;
+};
+
+#define ARCH_MMU_GATHER_INIT (struct arch_mmu_gather){ .batch = NULL, }
+
 extern void tlb_flush(struct mmu_gather *tlb);
 
 /* Get the generic bits... */
Index: linux-2.6/arch/powerpc/kernel/process.c
===================================================================
--- linux-2.6.orig/arch/powerpc/kernel/process.c
+++ linux-2.6/arch/powerpc/kernel/process.c
@@ -395,6 +395,9 @@ struct task_struct *__switch_to(struct t
 	struct thread_struct *new_thread, *old_thread;
 	unsigned long flags;
 	struct task_struct *last;
+#ifdef CONFIG_PPC_BOOK3S_64
+	struct ppc64_tlb_batch *batch;
+#endif
 
 #ifdef CONFIG_SMP
 	/* avoid complexity of lazy save/restore of fpu
@@ -513,7 +516,17 @@ struct task_struct *__switch_to(struct t
 		old_thread->accum_tb += (current_tb - start_tb);
 		new_thread->start_tb = current_tb;
 	}
-#endif
+#endif /* CONFIG_PPC64 */
+
+#ifdef CONFIG_PPC_BOOK3S_64
+	batch = &__get_cpu_var(ppc64_tlb_batch);
+	if (batch->active) {
+		current_thread_info()->local_flags |= _TLF_LAZY_MMU;
+		if (batch->index)
+			__flush_tlb_pending(batch);
+		batch->active = 0;
+	}
+#endif /* CONFIG_PPC_BOOK3S_64 */
 
 	local_irq_save(flags);
 
@@ -528,6 +541,14 @@ struct task_struct *__switch_to(struct t
 	hard_irq_disable();
 	last = _switch(old_thread, new_thread);
 
+#ifdef CONFIG_PPC_BOOK3S_64
+	if (current_thread_info()->local_flags & _TLF_LAZY_MMU) {
+		current_thread_info()->local_flags &= ~_TLF_LAZY_MMU;
+		batch = &__get_cpu_var(ppc64_tlb_batch);
+		batch->active = 1;
+	}
+#endif /* CONFIG_PPC_BOOK3S_64 */
+
 	local_irq_restore(flags);
 
 	return last;
Index: linux-2.6/arch/powerpc/mm/pgtable.c
===================================================================
--- linux-2.6.orig/arch/powerpc/mm/pgtable.c
+++ linux-2.6/arch/powerpc/mm/pgtable.c
@@ -33,8 +33,6 @@
 
 #include "mmu_decl.h"
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 #ifdef CONFIG_SMP
 
 /*
@@ -43,7 +41,6 @@ DEFINE_PER_CPU(struct mmu_gather, mmu_ga
  * freeing a page table page that is being walked without locks
  */
 
-static DEFINE_PER_CPU(struct pte_freelist_batch *, pte_freelist_cur);
 static unsigned long pte_freelist_forced_free;
 
 struct pte_freelist_batch
@@ -97,12 +94,10 @@ static void pte_free_submit(struct pte_f
 
 void pgtable_free_tlb(struct mmu_gather *tlb, void *table, unsigned shift)
 {
-	/* This is safe since tlb_gather_mmu has disabled preemption */
-	struct pte_freelist_batch **batchp = &__get_cpu_var(pte_freelist_cur);
+	struct pte_freelist_batch **batchp = &tlb->arch.batch;
 	unsigned long pgf;
 
-	if (atomic_read(&tlb->mm->mm_users) < 2 ||
-	    cpumask_equal(mm_cpumask(tlb->mm), cpumask_of(smp_processor_id()))){
+	if (atomic_read(&tlb->mm->mm_users) < 2) {
 		pgtable_free(table, shift);
 		return;
 	}
@@ -124,10 +119,9 @@ void pgtable_free_tlb(struct mmu_gather 
 	}
 }
 
-void pte_free_finish(void)
+void pte_free_finish(struct mmu_gather *tlb)
 {
-	/* This is safe since tlb_gather_mmu has disabled preemption */
-	struct pte_freelist_batch **batchp = &__get_cpu_var(pte_freelist_cur);
+	struct pte_freelist_batch **batchp = &tlb->arch.batch;
 
 	if (*batchp == NULL)
 		return;
Index: linux-2.6/arch/powerpc/include/asm/thread_info.h
===================================================================
--- linux-2.6.orig/arch/powerpc/include/asm/thread_info.h
+++ linux-2.6/arch/powerpc/include/asm/thread_info.h
@@ -139,10 +139,12 @@ static inline struct thread_info *curren
 #define TLF_NAPPING		0	/* idle thread enabled NAP mode */
 #define TLF_SLEEPING		1	/* suspend code enabled SLEEP mode */
 #define TLF_RESTORE_SIGMASK	2	/* Restore signal mask in do_signal */
+#define TLF_LAZY_MMU		3	/* tlb_batch is active */
 
 #define _TLF_NAPPING		(1 << TLF_NAPPING)
 #define _TLF_SLEEPING		(1 << TLF_SLEEPING)
 #define _TLF_RESTORE_SIGMASK	(1 << TLF_RESTORE_SIGMASK)
+#define _TLF_LAZY_MMU		(1 << TLF_LAZY_MMU)
 
 #ifndef __ASSEMBLY__
 #define HAVE_SET_RESTORE_SIGMASK	1
Index: linux-2.6/arch/powerpc/include/asm/pgalloc.h
===================================================================
--- linux-2.6.orig/arch/powerpc/include/asm/pgalloc.h
+++ linux-2.6/arch/powerpc/include/asm/pgalloc.h
@@ -32,13 +32,13 @@ static inline void pte_free(struct mm_st
 
 #ifdef CONFIG_SMP
 extern void pgtable_free_tlb(struct mmu_gather *tlb, void *table, unsigned shift);
-extern void pte_free_finish(void);
+extern void pte_free_finish(struct mmu_gather *tlb);
 #else /* CONFIG_SMP */
 static inline void pgtable_free_tlb(struct mmu_gather *tlb, void *table, unsigned shift)
 {
 	pgtable_free(table, shift);
 }
-static inline void pte_free_finish(void) { }
+static inline void pte_free_finish(struct mmu_gather *tlb) { }
 #endif /* !CONFIG_SMP */
 
 static inline void __pte_free_tlb(struct mmu_gather *tlb, struct page *ptepage,
Index: linux-2.6/arch/powerpc/mm/tlb_hash32.c
===================================================================
--- linux-2.6.orig/arch/powerpc/mm/tlb_hash32.c
+++ linux-2.6/arch/powerpc/mm/tlb_hash32.c
@@ -73,7 +73,7 @@ void tlb_flush(struct mmu_gather *tlb)
 	}
 
 	/* Push out batch of freed page tables */
-	pte_free_finish();
+	pte_free_finish(tlb);
 }
 
 /*
Index: linux-2.6/arch/powerpc/mm/tlb_nohash.c
===================================================================
--- linux-2.6.orig/arch/powerpc/mm/tlb_nohash.c
+++ linux-2.6/arch/powerpc/mm/tlb_nohash.c
@@ -301,7 +301,7 @@ void tlb_flush(struct mmu_gather *tlb)
 	flush_tlb_mm(tlb->mm);
 
 	/* Push out batch of freed page tables */
-	pte_free_finish();
+	pte_free_finish(tlb);
 }
 
 /*
Index: linux-2.6/arch/powerpc/mm/tlb_hash64.c
===================================================================
--- linux-2.6.orig/arch/powerpc/mm/tlb_hash64.c
+++ linux-2.6/arch/powerpc/mm/tlb_hash64.c
@@ -155,7 +155,7 @@ void __flush_tlb_pending(struct ppc64_tl
 
 void tlb_flush(struct mmu_gather *tlb)
 {
-	struct ppc64_tlb_batch *tlbbatch = &__get_cpu_var(ppc64_tlb_batch);
+	struct ppc64_tlb_batch *tlbbatch = &get_cpu_var(ppc64_tlb_batch);
 
 	/* If there's a TLB batch pending, then we must flush it because the
 	 * pages are going to be freed and we really don't want to have a CPU
@@ -164,8 +164,10 @@ void tlb_flush(struct mmu_gather *tlb)
 	if (tlbbatch->index)
 		__flush_tlb_pending(tlbbatch);
 
+	put_cpu_var(ppc64_tlb_batch);
+
 	/* Push out batch of freed page tables */
-	pte_free_finish();
+	pte_free_finish(tlb);
 }
 
 /**

[PATCH 02/20] powerpc: mmu_gather rework

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-powerpc-preemptible_mmu_gather.patch --]
[-- Type: text/plain, Size: 8137 bytes --]

Fix up powerpc to the new mmu_gather stuffs.

PPC has an extra batching queue to RCU free the actual pagetable
allocations, use the ARCH extentions for that for now.

For the ppc64_tlb_batch, which tracks the vaddrs to unhash from the
hardware hash-table, keep using per-cpu arrays but flush on context
switch and use a TLF bit to track the lazy_mmu state.

Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 arch/powerpc/include/asm/pgalloc.h     |    4 ++--
 arch/powerpc/include/asm/thread_info.h |    2 ++
 arch/powerpc/include/asm/tlb.h         |   10 ++++++++++
 arch/powerpc/kernel/process.c          |   23 ++++++++++++++++++++++-
 arch/powerpc/mm/pgtable.c              |   14 ++++----------
 arch/powerpc/mm/tlb_hash32.c           |    2 +-
 arch/powerpc/mm/tlb_hash64.c           |    6 ++++--
 arch/powerpc/mm/tlb_nohash.c           |    2 +-
 8 files changed, 46 insertions(+), 17 deletions(-)

Index: linux-2.6/arch/powerpc/include/asm/tlb.h
===================================================================
--- linux-2.6.orig/arch/powerpc/include/asm/tlb.h
+++ linux-2.6/arch/powerpc/include/asm/tlb.h
@@ -28,6 +28,16 @@
 #define tlb_start_vma(tlb, vma)	do { } while (0)
 #define tlb_end_vma(tlb, vma)	do { } while (0)
 
+#define HAVE_ARCH_MMU_GATHER 1
+
+struct pte_freelist_batch;
+
+struct arch_mmu_gather {
+	struct pte_freelist_batch *batch;
+};
+
+#define ARCH_MMU_GATHER_INIT (struct arch_mmu_gather){ .batch = NULL, }
+
 extern void tlb_flush(struct mmu_gather *tlb);
 
 /* Get the generic bits... */
Index: linux-2.6/arch/powerpc/kernel/process.c
===================================================================
--- linux-2.6.orig/arch/powerpc/kernel/process.c
+++ linux-2.6/arch/powerpc/kernel/process.c
@@ -395,6 +395,9 @@ struct task_struct *__switch_to(struct t
 	struct thread_struct *new_thread, *old_thread;
 	unsigned long flags;
 	struct task_struct *last;
+#ifdef CONFIG_PPC_BOOK3S_64
+	struct ppc64_tlb_batch *batch;
+#endif
 
 #ifdef CONFIG_SMP
 	/* avoid complexity of lazy save/restore of fpu
@@ -513,7 +516,17 @@ struct task_struct *__switch_to(struct t
 		old_thread->accum_tb += (current_tb - start_tb);
 		new_thread->start_tb = current_tb;
 	}
-#endif
+#endif /* CONFIG_PPC64 */
+
+#ifdef CONFIG_PPC_BOOK3S_64
+	batch = &__get_cpu_var(ppc64_tlb_batch);
+	if (batch->active) {
+		current_thread_info()->local_flags |= _TLF_LAZY_MMU;
+		if (batch->index)
+			__flush_tlb_pending(batch);
+		batch->active = 0;
+	}
+#endif /* CONFIG_PPC_BOOK3S_64 */
 
 	local_irq_save(flags);
 
@@ -528,6 +541,14 @@ struct task_struct *__switch_to(struct t
 	hard_irq_disable();
 	last = _switch(old_thread, new_thread);
 
+#ifdef CONFIG_PPC_BOOK3S_64
+	if (current_thread_info()->local_flags & _TLF_LAZY_MMU) {
+		current_thread_info()->local_flags &= ~_TLF_LAZY_MMU;
+		batch = &__get_cpu_var(ppc64_tlb_batch);
+		batch->active = 1;
+	}
+#endif /* CONFIG_PPC_BOOK3S_64 */
+
 	local_irq_restore(flags);
 
 	return last;
Index: linux-2.6/arch/powerpc/mm/pgtable.c
===================================================================
--- linux-2.6.orig/arch/powerpc/mm/pgtable.c
+++ linux-2.6/arch/powerpc/mm/pgtable.c
@@ -33,8 +33,6 @@
 
 #include "mmu_decl.h"
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 #ifdef CONFIG_SMP
 
 /*
@@ -43,7 +41,6 @@ DEFINE_PER_CPU(struct mmu_gather, mmu_ga
  * freeing a page table page that is being walked without locks
  */
 
-static DEFINE_PER_CPU(struct pte_freelist_batch *, pte_freelist_cur);
 static unsigned long pte_freelist_forced_free;
 
 struct pte_freelist_batch
@@ -97,12 +94,10 @@ static void pte_free_submit(struct pte_f
 
 void pgtable_free_tlb(struct mmu_gather *tlb, void *table, unsigned shift)
 {
-	/* This is safe since tlb_gather_mmu has disabled preemption */
-	struct pte_freelist_batch **batchp = &__get_cpu_var(pte_freelist_cur);
+	struct pte_freelist_batch **batchp = &tlb->arch.batch;
 	unsigned long pgf;
 
-	if (atomic_read(&tlb->mm->mm_users) < 2 ||
-	    cpumask_equal(mm_cpumask(tlb->mm), cpumask_of(smp_processor_id()))){
+	if (atomic_read(&tlb->mm->mm_users) < 2) {
 		pgtable_free(table, shift);
 		return;
 	}
@@ -124,10 +119,9 @@ void pgtable_free_tlb(struct mmu_gather 
 	}
 }
 
-void pte_free_finish(void)
+void pte_free_finish(struct mmu_gather *tlb)
 {
-	/* This is safe since tlb_gather_mmu has disabled preemption */
-	struct pte_freelist_batch **batchp = &__get_cpu_var(pte_freelist_cur);
+	struct pte_freelist_batch **batchp = &tlb->arch.batch;
 
 	if (*batchp == NULL)
 		return;
Index: linux-2.6/arch/powerpc/include/asm/thread_info.h
===================================================================
--- linux-2.6.orig/arch/powerpc/include/asm/thread_info.h
+++ linux-2.6/arch/powerpc/include/asm/thread_info.h
@@ -139,10 +139,12 @@ static inline struct thread_info *curren
 #define TLF_NAPPING		0	/* idle thread enabled NAP mode */
 #define TLF_SLEEPING		1	/* suspend code enabled SLEEP mode */
 #define TLF_RESTORE_SIGMASK	2	/* Restore signal mask in do_signal */
+#define TLF_LAZY_MMU		3	/* tlb_batch is active */
 
 #define _TLF_NAPPING		(1 << TLF_NAPPING)
 #define _TLF_SLEEPING		(1 << TLF_SLEEPING)
 #define _TLF_RESTORE_SIGMASK	(1 << TLF_RESTORE_SIGMASK)
+#define _TLF_LAZY_MMU		(1 << TLF_LAZY_MMU)
 
 #ifndef __ASSEMBLY__
 #define HAVE_SET_RESTORE_SIGMASK	1
Index: linux-2.6/arch/powerpc/include/asm/pgalloc.h
===================================================================
--- linux-2.6.orig/arch/powerpc/include/asm/pgalloc.h
+++ linux-2.6/arch/powerpc/include/asm/pgalloc.h
@@ -32,13 +32,13 @@ static inline void pte_free(struct mm_st
 
 #ifdef CONFIG_SMP
 extern void pgtable_free_tlb(struct mmu_gather *tlb, void *table, unsigned shift);
-extern void pte_free_finish(void);
+extern void pte_free_finish(struct mmu_gather *tlb);
 #else /* CONFIG_SMP */
 static inline void pgtable_free_tlb(struct mmu_gather *tlb, void *table, unsigned shift)
 {
 	pgtable_free(table, shift);
 }
-static inline void pte_free_finish(void) { }
+static inline void pte_free_finish(struct mmu_gather *tlb) { }
 #endif /* !CONFIG_SMP */
 
 static inline void __pte_free_tlb(struct mmu_gather *tlb, struct page *ptepage,
Index: linux-2.6/arch/powerpc/mm/tlb_hash32.c
===================================================================
--- linux-2.6.orig/arch/powerpc/mm/tlb_hash32.c
+++ linux-2.6/arch/powerpc/mm/tlb_hash32.c
@@ -73,7 +73,7 @@ void tlb_flush(struct mmu_gather *tlb)
 	}
 
 	/* Push out batch of freed page tables */
-	pte_free_finish();
+	pte_free_finish(tlb);
 }
 
 /*
Index: linux-2.6/arch/powerpc/mm/tlb_nohash.c
===================================================================
--- linux-2.6.orig/arch/powerpc/mm/tlb_nohash.c
+++ linux-2.6/arch/powerpc/mm/tlb_nohash.c
@@ -301,7 +301,7 @@ void tlb_flush(struct mmu_gather *tlb)
 	flush_tlb_mm(tlb->mm);
 
 	/* Push out batch of freed page tables */
-	pte_free_finish();
+	pte_free_finish(tlb);
 }
 
 /*
Index: linux-2.6/arch/powerpc/mm/tlb_hash64.c
===================================================================
--- linux-2.6.orig/arch/powerpc/mm/tlb_hash64.c
+++ linux-2.6/arch/powerpc/mm/tlb_hash64.c
@@ -155,7 +155,7 @@ void __flush_tlb_pending(struct ppc64_tl
 
 void tlb_flush(struct mmu_gather *tlb)
 {
-	struct ppc64_tlb_batch *tlbbatch = &__get_cpu_var(ppc64_tlb_batch);
+	struct ppc64_tlb_batch *tlbbatch = &get_cpu_var(ppc64_tlb_batch);
 
 	/* If there's a TLB batch pending, then we must flush it because the
 	 * pages are going to be freed and we really don't want to have a CPU
@@ -164,8 +164,10 @@ void tlb_flush(struct mmu_gather *tlb)
 	if (tlbbatch->index)
 		__flush_tlb_pending(tlbbatch);
 
+	put_cpu_var(ppc64_tlb_batch);
+
 	/* Push out batch of freed page tables */
-	pte_free_finish();
+	pte_free_finish(tlb);
 }
 
 /**


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 02/20] powerpc: mmu_gather rework

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-powerpc-preemptible_mmu_gather.patch --]
[-- Type: text/plain, Size: 8137 bytes --]

Fix up powerpc to the new mmu_gather stuffs.

PPC has an extra batching queue to RCU free the actual pagetable
allocations, use the ARCH extentions for that for now.

For the ppc64_tlb_batch, which tracks the vaddrs to unhash from the
hardware hash-table, keep using per-cpu arrays but flush on context
switch and use a TLF bit to track the lazy_mmu state.

Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 arch/powerpc/include/asm/pgalloc.h     |    4 ++--
 arch/powerpc/include/asm/thread_info.h |    2 ++
 arch/powerpc/include/asm/tlb.h         |   10 ++++++++++
 arch/powerpc/kernel/process.c          |   23 ++++++++++++++++++++++-
 arch/powerpc/mm/pgtable.c              |   14 ++++----------
 arch/powerpc/mm/tlb_hash32.c           |    2 +-
 arch/powerpc/mm/tlb_hash64.c           |    6 ++++--
 arch/powerpc/mm/tlb_nohash.c           |    2 +-
 8 files changed, 46 insertions(+), 17 deletions(-)

Index: linux-2.6/arch/powerpc/include/asm/tlb.h
===================================================================
--- linux-2.6.orig/arch/powerpc/include/asm/tlb.h
+++ linux-2.6/arch/powerpc/include/asm/tlb.h
@@ -28,6 +28,16 @@
 #define tlb_start_vma(tlb, vma)	do { } while (0)
 #define tlb_end_vma(tlb, vma)	do { } while (0)
 
+#define HAVE_ARCH_MMU_GATHER 1
+
+struct pte_freelist_batch;
+
+struct arch_mmu_gather {
+	struct pte_freelist_batch *batch;
+};
+
+#define ARCH_MMU_GATHER_INIT (struct arch_mmu_gather){ .batch = NULL, }
+
 extern void tlb_flush(struct mmu_gather *tlb);
 
 /* Get the generic bits... */
Index: linux-2.6/arch/powerpc/kernel/process.c
===================================================================
--- linux-2.6.orig/arch/powerpc/kernel/process.c
+++ linux-2.6/arch/powerpc/kernel/process.c
@@ -395,6 +395,9 @@ struct task_struct *__switch_to(struct t
 	struct thread_struct *new_thread, *old_thread;
 	unsigned long flags;
 	struct task_struct *last;
+#ifdef CONFIG_PPC_BOOK3S_64
+	struct ppc64_tlb_batch *batch;
+#endif
 
 #ifdef CONFIG_SMP
 	/* avoid complexity of lazy save/restore of fpu
@@ -513,7 +516,17 @@ struct task_struct *__switch_to(struct t
 		old_thread->accum_tb += (current_tb - start_tb);
 		new_thread->start_tb = current_tb;
 	}
-#endif
+#endif /* CONFIG_PPC64 */
+
+#ifdef CONFIG_PPC_BOOK3S_64
+	batch = &__get_cpu_var(ppc64_tlb_batch);
+	if (batch->active) {
+		current_thread_info()->local_flags |= _TLF_LAZY_MMU;
+		if (batch->index)
+			__flush_tlb_pending(batch);
+		batch->active = 0;
+	}
+#endif /* CONFIG_PPC_BOOK3S_64 */
 
 	local_irq_save(flags);
 
@@ -528,6 +541,14 @@ struct task_struct *__switch_to(struct t
 	hard_irq_disable();
 	last = _switch(old_thread, new_thread);
 
+#ifdef CONFIG_PPC_BOOK3S_64
+	if (current_thread_info()->local_flags & _TLF_LAZY_MMU) {
+		current_thread_info()->local_flags &= ~_TLF_LAZY_MMU;
+		batch = &__get_cpu_var(ppc64_tlb_batch);
+		batch->active = 1;
+	}
+#endif /* CONFIG_PPC_BOOK3S_64 */
+
 	local_irq_restore(flags);
 
 	return last;
Index: linux-2.6/arch/powerpc/mm/pgtable.c
===================================================================
--- linux-2.6.orig/arch/powerpc/mm/pgtable.c
+++ linux-2.6/arch/powerpc/mm/pgtable.c
@@ -33,8 +33,6 @@
 
 #include "mmu_decl.h"
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 #ifdef CONFIG_SMP
 
 /*
@@ -43,7 +41,6 @@ DEFINE_PER_CPU(struct mmu_gather, mmu_ga
  * freeing a page table page that is being walked without locks
  */
 
-static DEFINE_PER_CPU(struct pte_freelist_batch *, pte_freelist_cur);
 static unsigned long pte_freelist_forced_free;
 
 struct pte_freelist_batch
@@ -97,12 +94,10 @@ static void pte_free_submit(struct pte_f
 
 void pgtable_free_tlb(struct mmu_gather *tlb, void *table, unsigned shift)
 {
-	/* This is safe since tlb_gather_mmu has disabled preemption */
-	struct pte_freelist_batch **batchp = &__get_cpu_var(pte_freelist_cur);
+	struct pte_freelist_batch **batchp = &tlb->arch.batch;
 	unsigned long pgf;
 
-	if (atomic_read(&tlb->mm->mm_users) < 2 ||
-	    cpumask_equal(mm_cpumask(tlb->mm), cpumask_of(smp_processor_id()))){
+	if (atomic_read(&tlb->mm->mm_users) < 2) {
 		pgtable_free(table, shift);
 		return;
 	}
@@ -124,10 +119,9 @@ void pgtable_free_tlb(struct mmu_gather 
 	}
 }
 
-void pte_free_finish(void)
+void pte_free_finish(struct mmu_gather *tlb)
 {
-	/* This is safe since tlb_gather_mmu has disabled preemption */
-	struct pte_freelist_batch **batchp = &__get_cpu_var(pte_freelist_cur);
+	struct pte_freelist_batch **batchp = &tlb->arch.batch;
 
 	if (*batchp == NULL)
 		return;
Index: linux-2.6/arch/powerpc/include/asm/thread_info.h
===================================================================
--- linux-2.6.orig/arch/powerpc/include/asm/thread_info.h
+++ linux-2.6/arch/powerpc/include/asm/thread_info.h
@@ -139,10 +139,12 @@ static inline struct thread_info *curren
 #define TLF_NAPPING		0	/* idle thread enabled NAP mode */
 #define TLF_SLEEPING		1	/* suspend code enabled SLEEP mode */
 #define TLF_RESTORE_SIGMASK	2	/* Restore signal mask in do_signal */
+#define TLF_LAZY_MMU		3	/* tlb_batch is active */
 
 #define _TLF_NAPPING		(1 << TLF_NAPPING)
 #define _TLF_SLEEPING		(1 << TLF_SLEEPING)
 #define _TLF_RESTORE_SIGMASK	(1 << TLF_RESTORE_SIGMASK)
+#define _TLF_LAZY_MMU		(1 << TLF_LAZY_MMU)
 
 #ifndef __ASSEMBLY__
 #define HAVE_SET_RESTORE_SIGMASK	1
Index: linux-2.6/arch/powerpc/include/asm/pgalloc.h
===================================================================
--- linux-2.6.orig/arch/powerpc/include/asm/pgalloc.h
+++ linux-2.6/arch/powerpc/include/asm/pgalloc.h
@@ -32,13 +32,13 @@ static inline void pte_free(struct mm_st
 
 #ifdef CONFIG_SMP
 extern void pgtable_free_tlb(struct mmu_gather *tlb, void *table, unsigned shift);
-extern void pte_free_finish(void);
+extern void pte_free_finish(struct mmu_gather *tlb);
 #else /* CONFIG_SMP */
 static inline void pgtable_free_tlb(struct mmu_gather *tlb, void *table, unsigned shift)
 {
 	pgtable_free(table, shift);
 }
-static inline void pte_free_finish(void) { }
+static inline void pte_free_finish(struct mmu_gather *tlb) { }
 #endif /* !CONFIG_SMP */
 
 static inline void __pte_free_tlb(struct mmu_gather *tlb, struct page *ptepage,
Index: linux-2.6/arch/powerpc/mm/tlb_hash32.c
===================================================================
--- linux-2.6.orig/arch/powerpc/mm/tlb_hash32.c
+++ linux-2.6/arch/powerpc/mm/tlb_hash32.c
@@ -73,7 +73,7 @@ void tlb_flush(struct mmu_gather *tlb)
 	}
 
 	/* Push out batch of freed page tables */
-	pte_free_finish();
+	pte_free_finish(tlb);
 }
 
 /*
Index: linux-2.6/arch/powerpc/mm/tlb_nohash.c
===================================================================
--- linux-2.6.orig/arch/powerpc/mm/tlb_nohash.c
+++ linux-2.6/arch/powerpc/mm/tlb_nohash.c
@@ -301,7 +301,7 @@ void tlb_flush(struct mmu_gather *tlb)
 	flush_tlb_mm(tlb->mm);
 
 	/* Push out batch of freed page tables */
-	pte_free_finish();
+	pte_free_finish(tlb);
 }
 
 /*
Index: linux-2.6/arch/powerpc/mm/tlb_hash64.c
===================================================================
--- linux-2.6.orig/arch/powerpc/mm/tlb_hash64.c
+++ linux-2.6/arch/powerpc/mm/tlb_hash64.c
@@ -155,7 +155,7 @@ void __flush_tlb_pending(struct ppc64_tl
 
 void tlb_flush(struct mmu_gather *tlb)
 {
-	struct ppc64_tlb_batch *tlbbatch = &__get_cpu_var(ppc64_tlb_batch);
+	struct ppc64_tlb_batch *tlbbatch = &get_cpu_var(ppc64_tlb_batch);
 
 	/* If there's a TLB batch pending, then we must flush it because the
 	 * pages are going to be freed and we really don't want to have a CPU
@@ -164,8 +164,10 @@ void tlb_flush(struct mmu_gather *tlb)
 	if (tlbbatch->index)
 		__flush_tlb_pending(tlbbatch);
 
+	put_cpu_var(ppc64_tlb_batch);
+
 	/* Push out batch of freed page tables */
-	pte_free_finish();
+	pte_free_finish(tlb);
 }
 
 /**


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 03/20] sparc: mmu_gather rework

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-sparc-preemptible_mmu_gather.patch --]
[-- Type: text/plain, Size: 10608 bytes --]

Rework the sparc mmu_gather usage to conform to the new world order :-)

Sparc mmu_gather does two things:
 - tracks vaddrs to unhash
 - tracks pages to free

Split these two things like powerpc has done and keep the vaddrs
in per-cpu data structures and flush them on context switch.

The remaining bits can then use the generic mmu_gather.

Acked-by: David Miller <davem@davemloft.net>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 arch/sparc/include/asm/pgalloc_64.h  |    3 +
 arch/sparc/include/asm/pgtable_64.h  |   15 ++++-
 arch/sparc/include/asm/tlb_64.h      |   91 ++---------------------------------
 arch/sparc/include/asm/tlbflush_64.h |   12 +++-
 arch/sparc/mm/tlb.c                  |   43 +++++++++-------
 arch/sparc/mm/tsb.c                  |   15 +++--
 6 files changed, 63 insertions(+), 116 deletions(-)

Index: linux-2.6/arch/sparc/include/asm/pgalloc_64.h
===================================================================
--- linux-2.6.orig/arch/sparc/include/asm/pgalloc_64.h
+++ linux-2.6/arch/sparc/include/asm/pgalloc_64.h
@@ -78,4 +78,7 @@ static inline void check_pgt_cache(void)
 	quicklist_trim(0, NULL, 25, 16);
 }
 
+#define __pte_free_tlb(tlb, pte, addr)	pte_free((tlb)->mm, pte)
+#define __pmd_free_tlb(tlb, pmd, addr)	pmd_free((tlb)->mm, pmd)
+
 #endif /* _SPARC64_PGALLOC_H */
Index: linux-2.6/arch/sparc/include/asm/tlb_64.h
===================================================================
--- linux-2.6.orig/arch/sparc/include/asm/tlb_64.h
+++ linux-2.6/arch/sparc/include/asm/tlb_64.h
@@ -7,66 +7,11 @@
 #include <asm/tlbflush.h>
 #include <asm/mmu_context.h>
 
-#define TLB_BATCH_NR	192
-
-/*
- * For UP we don't need to worry about TLB flush
- * and page free order so much..
- */
-#ifdef CONFIG_SMP
-  #define FREE_PTE_NR	506
-  #define tlb_fast_mode(bp) ((bp)->pages_nr == ~0U)
-#else
-  #define FREE_PTE_NR	1
-  #define tlb_fast_mode(bp) 1
-#endif
-
-struct mmu_gather {
-	struct mm_struct *mm;
-	unsigned int pages_nr;
-	unsigned int need_flush;
-	unsigned int fullmm;
-	unsigned int tlb_nr;
-	unsigned long vaddrs[TLB_BATCH_NR];
-	struct page *pages[FREE_PTE_NR];
-};
-
-DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 #ifdef CONFIG_SMP
 extern void smp_flush_tlb_pending(struct mm_struct *,
 				  unsigned long, unsigned long *);
 #endif
 
-extern void __flush_tlb_pending(unsigned long, unsigned long, unsigned long *);
-extern void flush_tlb_pending(void);
-
-static inline struct mmu_gather *tlb_gather_mmu(struct mm_struct *mm, unsigned int full_mm_flush)
-{
-	struct mmu_gather *mp = &get_cpu_var(mmu_gathers);
-
-	BUG_ON(mp->tlb_nr);
-
-	mp->mm = mm;
-	mp->pages_nr = num_online_cpus() > 1 ? 0U : ~0U;
-	mp->fullmm = full_mm_flush;
-
-	return mp;
-}
-
-
-static inline void tlb_flush_mmu(struct mmu_gather *mp)
-{
-	if (!mp->fullmm)
-		flush_tlb_pending();
-	if (mp->need_flush) {
-		free_pages_and_swap_cache(mp->pages, mp->pages_nr);
-		mp->pages_nr = 0;
-		mp->need_flush = 0;
-	}
-
-}
-
 #ifdef CONFIG_SMP
 extern void smp_flush_tlb_mm(struct mm_struct *mm);
 #define do_flush_tlb_mm(mm) smp_flush_tlb_mm(mm)
@@ -74,38 +19,14 @@ extern void smp_flush_tlb_mm(struct mm_s
 #define do_flush_tlb_mm(mm) __flush_tlb_mm(CTX_HWBITS(mm->context), SECONDARY_CONTEXT)
 #endif
 
-static inline void tlb_finish_mmu(struct mmu_gather *mp, unsigned long start, unsigned long end)
-{
-	tlb_flush_mmu(mp);
-
-	if (mp->fullmm)
-		mp->fullmm = 0;
-
-	/* keep the page table cache within bounds */
-	check_pgt_cache();
-
-	put_cpu_var(mmu_gathers);
-}
-
-static inline void tlb_remove_page(struct mmu_gather *mp, struct page *page)
-{
-	if (tlb_fast_mode(mp)) {
-		free_page_and_swap_cache(page);
-		return;
-	}
-	mp->need_flush = 1;
-	mp->pages[mp->pages_nr++] = page;
-	if (mp->pages_nr >= FREE_PTE_NR)
-		tlb_flush_mmu(mp);
-}
-
-#define tlb_remove_tlb_entry(mp,ptep,addr) do { } while (0)
-#define pte_free_tlb(mp, ptepage, addr) pte_free((mp)->mm, ptepage)
-#define pmd_free_tlb(mp, pmdp, addr) pmd_free((mp)->mm, pmdp)
-#define pud_free_tlb(tlb,pudp, addr) __pud_free_tlb(tlb,pudp,addr)
+extern void __flush_tlb_pending(unsigned long, unsigned long, unsigned long *);
+extern void flush_tlb_pending(void);
 
-#define tlb_migrate_finish(mm)	do { } while (0)
 #define tlb_start_vma(tlb, vma) do { } while (0)
 #define tlb_end_vma(tlb, vma)	do { } while (0)
+#define __tlb_remove_tlb_entry(tlb, ptep, address) do { } while (0)
+#define tlb_flush(tlb)	flush_tlb_pending()
+
+#include <asm-generic/tlb.h>
 
 #endif /* _SPARC64_TLB_H */
Index: linux-2.6/arch/sparc/include/asm/tlbflush_64.h
===================================================================
--- linux-2.6.orig/arch/sparc/include/asm/tlbflush_64.h
+++ linux-2.6/arch/sparc/include/asm/tlbflush_64.h
@@ -5,9 +5,17 @@
 #include <asm/mmu_context.h>
 
 /* TSB flush operations. */
-struct mmu_gather;
+
+#define TLB_BATCH_NR	192
+
+struct tlb_batch {
+	struct mm_struct *mm;
+	unsigned long tlb_nr;
+	unsigned long vaddrs[TLB_BATCH_NR];
+};
+
 extern void flush_tsb_kernel_range(unsigned long start, unsigned long end);
-extern void flush_tsb_user(struct mmu_gather *mp);
+extern void flush_tsb_user(struct tlb_batch *tb);
 
 /* TLB flush operations. */
 
Index: linux-2.6/arch/sparc/mm/tlb.c
===================================================================
--- linux-2.6.orig/arch/sparc/mm/tlb.c
+++ linux-2.6/arch/sparc/mm/tlb.c
@@ -19,33 +19,34 @@
 
 /* Heavily inspired by the ppc64 code.  */
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
+static DEFINE_PER_CPU(struct tlb_batch, tlb_batch);
 
 void flush_tlb_pending(void)
 {
-	struct mmu_gather *mp = &get_cpu_var(mmu_gathers);
+	struct tlb_batch *tb = &get_cpu_var(tlb_batch);
 
-	if (mp->tlb_nr) {
-		flush_tsb_user(mp);
+	if (tb->tlb_nr) {
+		flush_tsb_user(tb);
 
-		if (CTX_VALID(mp->mm->context)) {
+		if (CTX_VALID(tb->mm->context)) {
 #ifdef CONFIG_SMP
-			smp_flush_tlb_pending(mp->mm, mp->tlb_nr,
-					      &mp->vaddrs[0]);
+			smp_flush_tlb_pending(tb->mm, tb->tlb_nr,
+					      &tb->vaddrs[0]);
 #else
-			__flush_tlb_pending(CTX_HWBITS(mp->mm->context),
-					    mp->tlb_nr, &mp->vaddrs[0]);
+			__flush_tlb_pending(CTX_HWBITS(tb->mm->context),
+					    tb->tlb_nr, &tb->vaddrs[0]);
 #endif
 		}
-		mp->tlb_nr = 0;
+		tb->tlb_nr = 0;
 	}
 
-	put_cpu_var(mmu_gathers);
+	put_cpu_var(tlb_batch);
 }
 
-void tlb_batch_add(struct mm_struct *mm, unsigned long vaddr, pte_t *ptep, pte_t orig)
+void tlb_batch_add(struct mm_struct *mm, unsigned long vaddr,
+		   pte_t *ptep, pte_t orig, int fullmm)
 {
-	struct mmu_gather *mp = &__get_cpu_var(mmu_gathers);
+	struct tlb_batch *tb = &get_cpu_var(tlb_batch);
 	unsigned long nr;
 
 	vaddr &= PAGE_MASK;
@@ -77,21 +78,25 @@ void tlb_batch_add(struct mm_struct *mm,
 
 no_cache_flush:
 
-	if (mp->fullmm)
+	if (fullmm) {
+		put_cpu_var(tlb_batch);
 		return;
+	}
 
-	nr = mp->tlb_nr;
+	nr = tb->tlb_nr;
 
-	if (unlikely(nr != 0 && mm != mp->mm)) {
+	if (unlikely(nr != 0 && mm != tb->mm)) {
 		flush_tlb_pending();
 		nr = 0;
 	}
 
 	if (nr == 0)
-		mp->mm = mm;
+		tb->mm = mm;
 
-	mp->vaddrs[nr] = vaddr;
-	mp->tlb_nr = ++nr;
+	tb->vaddrs[nr] = vaddr;
+	tb->tlb_nr = ++nr;
 	if (nr >= TLB_BATCH_NR)
 		flush_tlb_pending();
+
+	put_cpu_var(tlb_batch);
 }
Index: linux-2.6/arch/sparc/mm/tsb.c
===================================================================
--- linux-2.6.orig/arch/sparc/mm/tsb.c
+++ linux-2.6/arch/sparc/mm/tsb.c
@@ -47,12 +47,13 @@ void flush_tsb_kernel_range(unsigned lon
 	}
 }
 
-static void __flush_tsb_one(struct mmu_gather *mp, unsigned long hash_shift, unsigned long tsb, unsigned long nentries)
+static void __flush_tsb_one(struct tlb_batch *tb, unsigned long hash_shift,
+			    unsigned long tsb, unsigned long nentries)
 {
 	unsigned long i;
 
-	for (i = 0; i < mp->tlb_nr; i++) {
-		unsigned long v = mp->vaddrs[i];
+	for (i = 0; i < tb->tlb_nr; i++) {
+		unsigned long v = tb->vaddrs[i];
 		unsigned long tag, ent, hash;
 
 		v &= ~0x1UL;
@@ -65,9 +66,9 @@ static void __flush_tsb_one(struct mmu_g
 	}
 }
 
-void flush_tsb_user(struct mmu_gather *mp)
+void flush_tsb_user(struct tlb_batch *tb)
 {
-	struct mm_struct *mm = mp->mm;
+	struct mm_struct *mm = tb->mm;
 	unsigned long nentries, base, flags;
 
 	spin_lock_irqsave(&mm->context.lock, flags);
@@ -76,7 +77,7 @@ void flush_tsb_user(struct mmu_gather *m
 	nentries = mm->context.tsb_block[MM_TSB_BASE].tsb_nentries;
 	if (tlb_type == cheetah_plus || tlb_type == hypervisor)
 		base = __pa(base);
-	__flush_tsb_one(mp, PAGE_SHIFT, base, nentries);
+	__flush_tsb_one(tb, PAGE_SHIFT, base, nentries);
 
 #ifdef CONFIG_HUGETLB_PAGE
 	if (mm->context.tsb_block[MM_TSB_HUGE].tsb) {
@@ -84,7 +85,7 @@ void flush_tsb_user(struct mmu_gather *m
 		nentries = mm->context.tsb_block[MM_TSB_HUGE].tsb_nentries;
 		if (tlb_type == cheetah_plus || tlb_type == hypervisor)
 			base = __pa(base);
-		__flush_tsb_one(mp, HPAGE_SHIFT, base, nentries);
+		__flush_tsb_one(tb, HPAGE_SHIFT, base, nentries);
 	}
 #endif
 	spin_unlock_irqrestore(&mm->context.lock, flags);
Index: linux-2.6/arch/sparc/include/asm/pgtable_64.h
===================================================================
--- linux-2.6.orig/arch/sparc/include/asm/pgtable_64.h
+++ linux-2.6/arch/sparc/include/asm/pgtable_64.h
@@ -655,9 +655,11 @@ static inline int pte_special(pte_t pte)
 #define pte_unmap(pte)			do { } while (0)
 
 /* Actual page table PTE updates.  */
-extern void tlb_batch_add(struct mm_struct *mm, unsigned long vaddr, pte_t *ptep, pte_t orig);
+extern void tlb_batch_add(struct mm_struct *mm, unsigned long vaddr,
+			  pte_t *ptep, pte_t orig, int fullmm);
 
-static inline void set_pte_at(struct mm_struct *mm, unsigned long addr, pte_t *ptep, pte_t pte)
+static inline void __set_pte_at(struct mm_struct *mm, unsigned long addr,
+			     pte_t *ptep, pte_t pte, int fullmm)
 {
 	pte_t orig = *ptep;
 
@@ -670,12 +672,19 @@ static inline void set_pte_at(struct mm_
 	 *             and SUN4V pte layout, so this inline test is fine.
 	 */
 	if (likely(mm != &init_mm) && (pte_val(orig) & _PAGE_VALID))
-		tlb_batch_add(mm, addr, ptep, orig);
+		tlb_batch_add(mm, addr, ptep, orig, fullmm);
 }
 
+#define set_pte_at(mm,addr,ptep,pte)	\
+	__set_pte_at((mm), (addr), (ptep), (pte), 0)
+
 #define pte_clear(mm,addr,ptep)		\
 	set_pte_at((mm), (addr), (ptep), __pte(0UL))
 
+#define __HAVE_ARCH_PTE_CLEAR_NOT_PRESENT_FULL
+#define pte_clear_not_present_full(mm,addr,ptep,fullmm)	\
+	__set_pte_at((mm), (addr), (ptep), __pte(0UL), (fullmm))
+
 #ifdef DCACHE_ALIASING_POSSIBLE
 #define __HAVE_ARCH_MOVE_PTE
 #define move_pte(pte, prot, old_addr, new_addr)				\

[PATCH 03/20] sparc: mmu_gather rework

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-sparc-preemptible_mmu_gather.patch --]
[-- Type: text/plain, Size: 10911 bytes --]

Rework the sparc mmu_gather usage to conform to the new world order :-)

Sparc mmu_gather does two things:
 - tracks vaddrs to unhash
 - tracks pages to free

Split these two things like powerpc has done and keep the vaddrs
in per-cpu data structures and flush them on context switch.

The remaining bits can then use the generic mmu_gather.

Acked-by: David Miller <davem@davemloft.net>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 arch/sparc/include/asm/pgalloc_64.h  |    3 +
 arch/sparc/include/asm/pgtable_64.h  |   15 ++++-
 arch/sparc/include/asm/tlb_64.h      |   91 ++---------------------------------
 arch/sparc/include/asm/tlbflush_64.h |   12 +++-
 arch/sparc/mm/tlb.c                  |   43 +++++++++-------
 arch/sparc/mm/tsb.c                  |   15 +++--
 6 files changed, 63 insertions(+), 116 deletions(-)

Index: linux-2.6/arch/sparc/include/asm/pgalloc_64.h
===================================================================
--- linux-2.6.orig/arch/sparc/include/asm/pgalloc_64.h
+++ linux-2.6/arch/sparc/include/asm/pgalloc_64.h
@@ -78,4 +78,7 @@ static inline void check_pgt_cache(void)
 	quicklist_trim(0, NULL, 25, 16);
 }
 
+#define __pte_free_tlb(tlb, pte, addr)	pte_free((tlb)->mm, pte)
+#define __pmd_free_tlb(tlb, pmd, addr)	pmd_free((tlb)->mm, pmd)
+
 #endif /* _SPARC64_PGALLOC_H */
Index: linux-2.6/arch/sparc/include/asm/tlb_64.h
===================================================================
--- linux-2.6.orig/arch/sparc/include/asm/tlb_64.h
+++ linux-2.6/arch/sparc/include/asm/tlb_64.h
@@ -7,66 +7,11 @@
 #include <asm/tlbflush.h>
 #include <asm/mmu_context.h>
 
-#define TLB_BATCH_NR	192
-
-/*
- * For UP we don't need to worry about TLB flush
- * and page free order so much..
- */
-#ifdef CONFIG_SMP
-  #define FREE_PTE_NR	506
-  #define tlb_fast_mode(bp) ((bp)->pages_nr == ~0U)
-#else
-  #define FREE_PTE_NR	1
-  #define tlb_fast_mode(bp) 1
-#endif
-
-struct mmu_gather {
-	struct mm_struct *mm;
-	unsigned int pages_nr;
-	unsigned int need_flush;
-	unsigned int fullmm;
-	unsigned int tlb_nr;
-	unsigned long vaddrs[TLB_BATCH_NR];
-	struct page *pages[FREE_PTE_NR];
-};
-
-DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 #ifdef CONFIG_SMP
 extern void smp_flush_tlb_pending(struct mm_struct *,
 				  unsigned long, unsigned long *);
 #endif
 
-extern void __flush_tlb_pending(unsigned long, unsigned long, unsigned long *);
-extern void flush_tlb_pending(void);
-
-static inline struct mmu_gather *tlb_gather_mmu(struct mm_struct *mm, unsigned int full_mm_flush)
-{
-	struct mmu_gather *mp = &get_cpu_var(mmu_gathers);
-
-	BUG_ON(mp->tlb_nr);
-
-	mp->mm = mm;
-	mp->pages_nr = num_online_cpus() > 1 ? 0U : ~0U;
-	mp->fullmm = full_mm_flush;
-
-	return mp;
-}
-
-
-static inline void tlb_flush_mmu(struct mmu_gather *mp)
-{
-	if (!mp->fullmm)
-		flush_tlb_pending();
-	if (mp->need_flush) {
-		free_pages_and_swap_cache(mp->pages, mp->pages_nr);
-		mp->pages_nr = 0;
-		mp->need_flush = 0;
-	}
-
-}
-
 #ifdef CONFIG_SMP
 extern void smp_flush_tlb_mm(struct mm_struct *mm);
 #define do_flush_tlb_mm(mm) smp_flush_tlb_mm(mm)
@@ -74,38 +19,14 @@ extern void smp_flush_tlb_mm(struct mm_s
 #define do_flush_tlb_mm(mm) __flush_tlb_mm(CTX_HWBITS(mm->context), SECONDARY_CONTEXT)
 #endif
 
-static inline void tlb_finish_mmu(struct mmu_gather *mp, unsigned long start, unsigned long end)
-{
-	tlb_flush_mmu(mp);
-
-	if (mp->fullmm)
-		mp->fullmm = 0;
-
-	/* keep the page table cache within bounds */
-	check_pgt_cache();
-
-	put_cpu_var(mmu_gathers);
-}
-
-static inline void tlb_remove_page(struct mmu_gather *mp, struct page *page)
-{
-	if (tlb_fast_mode(mp)) {
-		free_page_and_swap_cache(page);
-		return;
-	}
-	mp->need_flush = 1;
-	mp->pages[mp->pages_nr++] = page;
-	if (mp->pages_nr >= FREE_PTE_NR)
-		tlb_flush_mmu(mp);
-}
-
-#define tlb_remove_tlb_entry(mp,ptep,addr) do { } while (0)
-#define pte_free_tlb(mp, ptepage, addr) pte_free((mp)->mm, ptepage)
-#define pmd_free_tlb(mp, pmdp, addr) pmd_free((mp)->mm, pmdp)
-#define pud_free_tlb(tlb,pudp, addr) __pud_free_tlb(tlb,pudp,addr)
+extern void __flush_tlb_pending(unsigned long, unsigned long, unsigned long *);
+extern void flush_tlb_pending(void);
 
-#define tlb_migrate_finish(mm)	do { } while (0)
 #define tlb_start_vma(tlb, vma) do { } while (0)
 #define tlb_end_vma(tlb, vma)	do { } while (0)
+#define __tlb_remove_tlb_entry(tlb, ptep, address) do { } while (0)
+#define tlb_flush(tlb)	flush_tlb_pending()
+
+#include <asm-generic/tlb.h>
 
 #endif /* _SPARC64_TLB_H */
Index: linux-2.6/arch/sparc/include/asm/tlbflush_64.h
===================================================================
--- linux-2.6.orig/arch/sparc/include/asm/tlbflush_64.h
+++ linux-2.6/arch/sparc/include/asm/tlbflush_64.h
@@ -5,9 +5,17 @@
 #include <asm/mmu_context.h>
 
 /* TSB flush operations. */
-struct mmu_gather;
+
+#define TLB_BATCH_NR	192
+
+struct tlb_batch {
+	struct mm_struct *mm;
+	unsigned long tlb_nr;
+	unsigned long vaddrs[TLB_BATCH_NR];
+};
+
 extern void flush_tsb_kernel_range(unsigned long start, unsigned long end);
-extern void flush_tsb_user(struct mmu_gather *mp);
+extern void flush_tsb_user(struct tlb_batch *tb);
 
 /* TLB flush operations. */
 
Index: linux-2.6/arch/sparc/mm/tlb.c
===================================================================
--- linux-2.6.orig/arch/sparc/mm/tlb.c
+++ linux-2.6/arch/sparc/mm/tlb.c
@@ -19,33 +19,34 @@
 
 /* Heavily inspired by the ppc64 code.  */
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
+static DEFINE_PER_CPU(struct tlb_batch, tlb_batch);
 
 void flush_tlb_pending(void)
 {
-	struct mmu_gather *mp = &get_cpu_var(mmu_gathers);
+	struct tlb_batch *tb = &get_cpu_var(tlb_batch);
 
-	if (mp->tlb_nr) {
-		flush_tsb_user(mp);
+	if (tb->tlb_nr) {
+		flush_tsb_user(tb);
 
-		if (CTX_VALID(mp->mm->context)) {
+		if (CTX_VALID(tb->mm->context)) {
 #ifdef CONFIG_SMP
-			smp_flush_tlb_pending(mp->mm, mp->tlb_nr,
-					      &mp->vaddrs[0]);
+			smp_flush_tlb_pending(tb->mm, tb->tlb_nr,
+					      &tb->vaddrs[0]);
 #else
-			__flush_tlb_pending(CTX_HWBITS(mp->mm->context),
-					    mp->tlb_nr, &mp->vaddrs[0]);
+			__flush_tlb_pending(CTX_HWBITS(tb->mm->context),
+					    tb->tlb_nr, &tb->vaddrs[0]);
 #endif
 		}
-		mp->tlb_nr = 0;
+		tb->tlb_nr = 0;
 	}
 
-	put_cpu_var(mmu_gathers);
+	put_cpu_var(tlb_batch);
 }
 
-void tlb_batch_add(struct mm_struct *mm, unsigned long vaddr, pte_t *ptep, pte_t orig)
+void tlb_batch_add(struct mm_struct *mm, unsigned long vaddr,
+		   pte_t *ptep, pte_t orig, int fullmm)
 {
-	struct mmu_gather *mp = &__get_cpu_var(mmu_gathers);
+	struct tlb_batch *tb = &get_cpu_var(tlb_batch);
 	unsigned long nr;
 
 	vaddr &= PAGE_MASK;
@@ -77,21 +78,25 @@ void tlb_batch_add(struct mm_struct *mm,
 
 no_cache_flush:
 
-	if (mp->fullmm)
+	if (fullmm) {
+		put_cpu_var(tlb_batch);
 		return;
+	}
 
-	nr = mp->tlb_nr;
+	nr = tb->tlb_nr;
 
-	if (unlikely(nr != 0 && mm != mp->mm)) {
+	if (unlikely(nr != 0 && mm != tb->mm)) {
 		flush_tlb_pending();
 		nr = 0;
 	}
 
 	if (nr == 0)
-		mp->mm = mm;
+		tb->mm = mm;
 
-	mp->vaddrs[nr] = vaddr;
-	mp->tlb_nr = ++nr;
+	tb->vaddrs[nr] = vaddr;
+	tb->tlb_nr = ++nr;
 	if (nr >= TLB_BATCH_NR)
 		flush_tlb_pending();
+
+	put_cpu_var(tlb_batch);
 }
Index: linux-2.6/arch/sparc/mm/tsb.c
===================================================================
--- linux-2.6.orig/arch/sparc/mm/tsb.c
+++ linux-2.6/arch/sparc/mm/tsb.c
@@ -47,12 +47,13 @@ void flush_tsb_kernel_range(unsigned lon
 	}
 }
 
-static void __flush_tsb_one(struct mmu_gather *mp, unsigned long hash_shift, unsigned long tsb, unsigned long nentries)
+static void __flush_tsb_one(struct tlb_batch *tb, unsigned long hash_shift,
+			    unsigned long tsb, unsigned long nentries)
 {
 	unsigned long i;
 
-	for (i = 0; i < mp->tlb_nr; i++) {
-		unsigned long v = mp->vaddrs[i];
+	for (i = 0; i < tb->tlb_nr; i++) {
+		unsigned long v = tb->vaddrs[i];
 		unsigned long tag, ent, hash;
 
 		v &= ~0x1UL;
@@ -65,9 +66,9 @@ static void __flush_tsb_one(struct mmu_g
 	}
 }
 
-void flush_tsb_user(struct mmu_gather *mp)
+void flush_tsb_user(struct tlb_batch *tb)
 {
-	struct mm_struct *mm = mp->mm;
+	struct mm_struct *mm = tb->mm;
 	unsigned long nentries, base, flags;
 
 	spin_lock_irqsave(&mm->context.lock, flags);
@@ -76,7 +77,7 @@ void flush_tsb_user(struct mmu_gather *m
 	nentries = mm->context.tsb_block[MM_TSB_BASE].tsb_nentries;
 	if (tlb_type == cheetah_plus || tlb_type == hypervisor)
 		base = __pa(base);
-	__flush_tsb_one(mp, PAGE_SHIFT, base, nentries);
+	__flush_tsb_one(tb, PAGE_SHIFT, base, nentries);
 
 #ifdef CONFIG_HUGETLB_PAGE
 	if (mm->context.tsb_block[MM_TSB_HUGE].tsb) {
@@ -84,7 +85,7 @@ void flush_tsb_user(struct mmu_gather *m
 		nentries = mm->context.tsb_block[MM_TSB_HUGE].tsb_nentries;
 		if (tlb_type == cheetah_plus || tlb_type == hypervisor)
 			base = __pa(base);
-		__flush_tsb_one(mp, HPAGE_SHIFT, base, nentries);
+		__flush_tsb_one(tb, HPAGE_SHIFT, base, nentries);
 	}
 #endif
 	spin_unlock_irqrestore(&mm->context.lock, flags);
Index: linux-2.6/arch/sparc/include/asm/pgtable_64.h
===================================================================
--- linux-2.6.orig/arch/sparc/include/asm/pgtable_64.h
+++ linux-2.6/arch/sparc/include/asm/pgtable_64.h
@@ -655,9 +655,11 @@ static inline int pte_special(pte_t pte)
 #define pte_unmap(pte)			do { } while (0)
 
 /* Actual page table PTE updates.  */
-extern void tlb_batch_add(struct mm_struct *mm, unsigned long vaddr, pte_t *ptep, pte_t orig);
+extern void tlb_batch_add(struct mm_struct *mm, unsigned long vaddr,
+			  pte_t *ptep, pte_t orig, int fullmm);
 
-static inline void set_pte_at(struct mm_struct *mm, unsigned long addr, pte_t *ptep, pte_t pte)
+static inline void __set_pte_at(struct mm_struct *mm, unsigned long addr,
+			     pte_t *ptep, pte_t pte, int fullmm)
 {
 	pte_t orig = *ptep;
 
@@ -670,12 +672,19 @@ static inline void set_pte_at(struct mm_
 	 *             and SUN4V pte layout, so this inline test is fine.
 	 */
 	if (likely(mm != &init_mm) && (pte_val(orig) & _PAGE_VALID))
-		tlb_batch_add(mm, addr, ptep, orig);
+		tlb_batch_add(mm, addr, ptep, orig, fullmm);
 }
 
+#define set_pte_at(mm,addr,ptep,pte)	\
+	__set_pte_at((mm), (addr), (ptep), (pte), 0)
+
 #define pte_clear(mm,addr,ptep)		\
 	set_pte_at((mm), (addr), (ptep), __pte(0UL))
 
+#define __HAVE_ARCH_PTE_CLEAR_NOT_PRESENT_FULL
+#define pte_clear_not_present_full(mm,addr,ptep,fullmm)	\
+	__set_pte_at((mm), (addr), (ptep), __pte(0UL), (fullmm))
+
 #ifdef DCACHE_ALIASING_POSSIBLE
 #define __HAVE_ARCH_MOVE_PTE
 #define move_pte(pte, prot, old_addr, new_addr)				\


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 03/20] sparc: mmu_gather rework

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-sparc-preemptible_mmu_gather.patch --]
[-- Type: text/plain, Size: 10911 bytes --]

Rework the sparc mmu_gather usage to conform to the new world order :-)

Sparc mmu_gather does two things:
 - tracks vaddrs to unhash
 - tracks pages to free

Split these two things like powerpc has done and keep the vaddrs
in per-cpu data structures and flush them on context switch.

The remaining bits can then use the generic mmu_gather.

Acked-by: David Miller <davem@davemloft.net>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 arch/sparc/include/asm/pgalloc_64.h  |    3 +
 arch/sparc/include/asm/pgtable_64.h  |   15 ++++-
 arch/sparc/include/asm/tlb_64.h      |   91 ++---------------------------------
 arch/sparc/include/asm/tlbflush_64.h |   12 +++-
 arch/sparc/mm/tlb.c                  |   43 +++++++++-------
 arch/sparc/mm/tsb.c                  |   15 +++--
 6 files changed, 63 insertions(+), 116 deletions(-)

Index: linux-2.6/arch/sparc/include/asm/pgalloc_64.h
===================================================================
--- linux-2.6.orig/arch/sparc/include/asm/pgalloc_64.h
+++ linux-2.6/arch/sparc/include/asm/pgalloc_64.h
@@ -78,4 +78,7 @@ static inline void check_pgt_cache(void)
 	quicklist_trim(0, NULL, 25, 16);
 }
 
+#define __pte_free_tlb(tlb, pte, addr)	pte_free((tlb)->mm, pte)
+#define __pmd_free_tlb(tlb, pmd, addr)	pmd_free((tlb)->mm, pmd)
+
 #endif /* _SPARC64_PGALLOC_H */
Index: linux-2.6/arch/sparc/include/asm/tlb_64.h
===================================================================
--- linux-2.6.orig/arch/sparc/include/asm/tlb_64.h
+++ linux-2.6/arch/sparc/include/asm/tlb_64.h
@@ -7,66 +7,11 @@
 #include <asm/tlbflush.h>
 #include <asm/mmu_context.h>
 
-#define TLB_BATCH_NR	192
-
-/*
- * For UP we don't need to worry about TLB flush
- * and page free order so much..
- */
-#ifdef CONFIG_SMP
-  #define FREE_PTE_NR	506
-  #define tlb_fast_mode(bp) ((bp)->pages_nr == ~0U)
-#else
-  #define FREE_PTE_NR	1
-  #define tlb_fast_mode(bp) 1
-#endif
-
-struct mmu_gather {
-	struct mm_struct *mm;
-	unsigned int pages_nr;
-	unsigned int need_flush;
-	unsigned int fullmm;
-	unsigned int tlb_nr;
-	unsigned long vaddrs[TLB_BATCH_NR];
-	struct page *pages[FREE_PTE_NR];
-};
-
-DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 #ifdef CONFIG_SMP
 extern void smp_flush_tlb_pending(struct mm_struct *,
 				  unsigned long, unsigned long *);
 #endif
 
-extern void __flush_tlb_pending(unsigned long, unsigned long, unsigned long *);
-extern void flush_tlb_pending(void);
-
-static inline struct mmu_gather *tlb_gather_mmu(struct mm_struct *mm, unsigned int full_mm_flush)
-{
-	struct mmu_gather *mp = &get_cpu_var(mmu_gathers);
-
-	BUG_ON(mp->tlb_nr);
-
-	mp->mm = mm;
-	mp->pages_nr = num_online_cpus() > 1 ? 0U : ~0U;
-	mp->fullmm = full_mm_flush;
-
-	return mp;
-}
-
-
-static inline void tlb_flush_mmu(struct mmu_gather *mp)
-{
-	if (!mp->fullmm)
-		flush_tlb_pending();
-	if (mp->need_flush) {
-		free_pages_and_swap_cache(mp->pages, mp->pages_nr);
-		mp->pages_nr = 0;
-		mp->need_flush = 0;
-	}
-
-}
-
 #ifdef CONFIG_SMP
 extern void smp_flush_tlb_mm(struct mm_struct *mm);
 #define do_flush_tlb_mm(mm) smp_flush_tlb_mm(mm)
@@ -74,38 +19,14 @@ extern void smp_flush_tlb_mm(struct mm_s
 #define do_flush_tlb_mm(mm) __flush_tlb_mm(CTX_HWBITS(mm->context), SECONDARY_CONTEXT)
 #endif
 
-static inline void tlb_finish_mmu(struct mmu_gather *mp, unsigned long start, unsigned long end)
-{
-	tlb_flush_mmu(mp);
-
-	if (mp->fullmm)
-		mp->fullmm = 0;
-
-	/* keep the page table cache within bounds */
-	check_pgt_cache();
-
-	put_cpu_var(mmu_gathers);
-}
-
-static inline void tlb_remove_page(struct mmu_gather *mp, struct page *page)
-{
-	if (tlb_fast_mode(mp)) {
-		free_page_and_swap_cache(page);
-		return;
-	}
-	mp->need_flush = 1;
-	mp->pages[mp->pages_nr++] = page;
-	if (mp->pages_nr >= FREE_PTE_NR)
-		tlb_flush_mmu(mp);
-}
-
-#define tlb_remove_tlb_entry(mp,ptep,addr) do { } while (0)
-#define pte_free_tlb(mp, ptepage, addr) pte_free((mp)->mm, ptepage)
-#define pmd_free_tlb(mp, pmdp, addr) pmd_free((mp)->mm, pmdp)
-#define pud_free_tlb(tlb,pudp, addr) __pud_free_tlb(tlb,pudp,addr)
+extern void __flush_tlb_pending(unsigned long, unsigned long, unsigned long *);
+extern void flush_tlb_pending(void);
 
-#define tlb_migrate_finish(mm)	do { } while (0)
 #define tlb_start_vma(tlb, vma) do { } while (0)
 #define tlb_end_vma(tlb, vma)	do { } while (0)
+#define __tlb_remove_tlb_entry(tlb, ptep, address) do { } while (0)
+#define tlb_flush(tlb)	flush_tlb_pending()
+
+#include <asm-generic/tlb.h>
 
 #endif /* _SPARC64_TLB_H */
Index: linux-2.6/arch/sparc/include/asm/tlbflush_64.h
===================================================================
--- linux-2.6.orig/arch/sparc/include/asm/tlbflush_64.h
+++ linux-2.6/arch/sparc/include/asm/tlbflush_64.h
@@ -5,9 +5,17 @@
 #include <asm/mmu_context.h>
 
 /* TSB flush operations. */
-struct mmu_gather;
+
+#define TLB_BATCH_NR	192
+
+struct tlb_batch {
+	struct mm_struct *mm;
+	unsigned long tlb_nr;
+	unsigned long vaddrs[TLB_BATCH_NR];
+};
+
 extern void flush_tsb_kernel_range(unsigned long start, unsigned long end);
-extern void flush_tsb_user(struct mmu_gather *mp);
+extern void flush_tsb_user(struct tlb_batch *tb);
 
 /* TLB flush operations. */
 
Index: linux-2.6/arch/sparc/mm/tlb.c
===================================================================
--- linux-2.6.orig/arch/sparc/mm/tlb.c
+++ linux-2.6/arch/sparc/mm/tlb.c
@@ -19,33 +19,34 @@
 
 /* Heavily inspired by the ppc64 code.  */
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
+static DEFINE_PER_CPU(struct tlb_batch, tlb_batch);
 
 void flush_tlb_pending(void)
 {
-	struct mmu_gather *mp = &get_cpu_var(mmu_gathers);
+	struct tlb_batch *tb = &get_cpu_var(tlb_batch);
 
-	if (mp->tlb_nr) {
-		flush_tsb_user(mp);
+	if (tb->tlb_nr) {
+		flush_tsb_user(tb);
 
-		if (CTX_VALID(mp->mm->context)) {
+		if (CTX_VALID(tb->mm->context)) {
 #ifdef CONFIG_SMP
-			smp_flush_tlb_pending(mp->mm, mp->tlb_nr,
-					      &mp->vaddrs[0]);
+			smp_flush_tlb_pending(tb->mm, tb->tlb_nr,
+					      &tb->vaddrs[0]);
 #else
-			__flush_tlb_pending(CTX_HWBITS(mp->mm->context),
-					    mp->tlb_nr, &mp->vaddrs[0]);
+			__flush_tlb_pending(CTX_HWBITS(tb->mm->context),
+					    tb->tlb_nr, &tb->vaddrs[0]);
 #endif
 		}
-		mp->tlb_nr = 0;
+		tb->tlb_nr = 0;
 	}
 
-	put_cpu_var(mmu_gathers);
+	put_cpu_var(tlb_batch);
 }
 
-void tlb_batch_add(struct mm_struct *mm, unsigned long vaddr, pte_t *ptep, pte_t orig)
+void tlb_batch_add(struct mm_struct *mm, unsigned long vaddr,
+		   pte_t *ptep, pte_t orig, int fullmm)
 {
-	struct mmu_gather *mp = &__get_cpu_var(mmu_gathers);
+	struct tlb_batch *tb = &get_cpu_var(tlb_batch);
 	unsigned long nr;
 
 	vaddr &= PAGE_MASK;
@@ -77,21 +78,25 @@ void tlb_batch_add(struct mm_struct *mm,
 
 no_cache_flush:
 
-	if (mp->fullmm)
+	if (fullmm) {
+		put_cpu_var(tlb_batch);
 		return;
+	}
 
-	nr = mp->tlb_nr;
+	nr = tb->tlb_nr;
 
-	if (unlikely(nr != 0 && mm != mp->mm)) {
+	if (unlikely(nr != 0 && mm != tb->mm)) {
 		flush_tlb_pending();
 		nr = 0;
 	}
 
 	if (nr == 0)
-		mp->mm = mm;
+		tb->mm = mm;
 
-	mp->vaddrs[nr] = vaddr;
-	mp->tlb_nr = ++nr;
+	tb->vaddrs[nr] = vaddr;
+	tb->tlb_nr = ++nr;
 	if (nr >= TLB_BATCH_NR)
 		flush_tlb_pending();
+
+	put_cpu_var(tlb_batch);
 }
Index: linux-2.6/arch/sparc/mm/tsb.c
===================================================================
--- linux-2.6.orig/arch/sparc/mm/tsb.c
+++ linux-2.6/arch/sparc/mm/tsb.c
@@ -47,12 +47,13 @@ void flush_tsb_kernel_range(unsigned lon
 	}
 }
 
-static void __flush_tsb_one(struct mmu_gather *mp, unsigned long hash_shift, unsigned long tsb, unsigned long nentries)
+static void __flush_tsb_one(struct tlb_batch *tb, unsigned long hash_shift,
+			    unsigned long tsb, unsigned long nentries)
 {
 	unsigned long i;
 
-	for (i = 0; i < mp->tlb_nr; i++) {
-		unsigned long v = mp->vaddrs[i];
+	for (i = 0; i < tb->tlb_nr; i++) {
+		unsigned long v = tb->vaddrs[i];
 		unsigned long tag, ent, hash;
 
 		v &= ~0x1UL;
@@ -65,9 +66,9 @@ static void __flush_tsb_one(struct mmu_g
 	}
 }
 
-void flush_tsb_user(struct mmu_gather *mp)
+void flush_tsb_user(struct tlb_batch *tb)
 {
-	struct mm_struct *mm = mp->mm;
+	struct mm_struct *mm = tb->mm;
 	unsigned long nentries, base, flags;
 
 	spin_lock_irqsave(&mm->context.lock, flags);
@@ -76,7 +77,7 @@ void flush_tsb_user(struct mmu_gather *m
 	nentries = mm->context.tsb_block[MM_TSB_BASE].tsb_nentries;
 	if (tlb_type == cheetah_plus || tlb_type == hypervisor)
 		base = __pa(base);
-	__flush_tsb_one(mp, PAGE_SHIFT, base, nentries);
+	__flush_tsb_one(tb, PAGE_SHIFT, base, nentries);
 
 #ifdef CONFIG_HUGETLB_PAGE
 	if (mm->context.tsb_block[MM_TSB_HUGE].tsb) {
@@ -84,7 +85,7 @@ void flush_tsb_user(struct mmu_gather *m
 		nentries = mm->context.tsb_block[MM_TSB_HUGE].tsb_nentries;
 		if (tlb_type == cheetah_plus || tlb_type == hypervisor)
 			base = __pa(base);
-		__flush_tsb_one(mp, HPAGE_SHIFT, base, nentries);
+		__flush_tsb_one(tb, HPAGE_SHIFT, base, nentries);
 	}
 #endif
 	spin_unlock_irqrestore(&mm->context.lock, flags);
Index: linux-2.6/arch/sparc/include/asm/pgtable_64.h
===================================================================
--- linux-2.6.orig/arch/sparc/include/asm/pgtable_64.h
+++ linux-2.6/arch/sparc/include/asm/pgtable_64.h
@@ -655,9 +655,11 @@ static inline int pte_special(pte_t pte)
 #define pte_unmap(pte)			do { } while (0)
 
 /* Actual page table PTE updates.  */
-extern void tlb_batch_add(struct mm_struct *mm, unsigned long vaddr, pte_t *ptep, pte_t orig);
+extern void tlb_batch_add(struct mm_struct *mm, unsigned long vaddr,
+			  pte_t *ptep, pte_t orig, int fullmm);
 
-static inline void set_pte_at(struct mm_struct *mm, unsigned long addr, pte_t *ptep, pte_t pte)
+static inline void __set_pte_at(struct mm_struct *mm, unsigned long addr,
+			     pte_t *ptep, pte_t pte, int fullmm)
 {
 	pte_t orig = *ptep;
 
@@ -670,12 +672,19 @@ static inline void set_pte_at(struct mm_
 	 *             and SUN4V pte layout, so this inline test is fine.
 	 */
 	if (likely(mm != &init_mm) && (pte_val(orig) & _PAGE_VALID))
-		tlb_batch_add(mm, addr, ptep, orig);
+		tlb_batch_add(mm, addr, ptep, orig, fullmm);
 }
 
+#define set_pte_at(mm,addr,ptep,pte)	\
+	__set_pte_at((mm), (addr), (ptep), (pte), 0)
+
 #define pte_clear(mm,addr,ptep)		\
 	set_pte_at((mm), (addr), (ptep), __pte(0UL))
 
+#define __HAVE_ARCH_PTE_CLEAR_NOT_PRESENT_FULL
+#define pte_clear_not_present_full(mm,addr,ptep,fullmm)	\
+	__set_pte_at((mm), (addr), (ptep), __pte(0UL), (fullmm))
+
 #ifdef DCACHE_ALIASING_POSSIBLE
 #define __HAVE_ARCH_MOVE_PTE
 #define move_pte(pte, prot, old_addr, new_addr)				\


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 04/20] s390: mmu_gather rework

[Peter Zijlstra]

[-- Attachment #1: martin_schwidefsky-s390-preemptible_mmu_gather.patch --]
[-- Type: text/plain, Size: 4183 bytes --]

Adapt the stand-alone s390 mmu_gather implementation to the new
preemptible mmu_gather interface.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <20101126145410.881573395@chello.nl>
---
 arch/s390/include/asm/tlb.h |   62 ++++++++++++++++++++++++++------------------
 1 file changed, 37 insertions(+), 25 deletions(-)

Index: linux-2.6/arch/s390/include/asm/tlb.h
===================================================================
--- linux-2.6.orig/arch/s390/include/asm/tlb.h
+++ linux-2.6/arch/s390/include/asm/tlb.h
@@ -29,65 +29,77 @@
 #include <asm/smp.h>
 #include <asm/tlbflush.h>
 
-#ifndef CONFIG_SMP
-#define TLB_NR_PTRS	1
-#else
-#define TLB_NR_PTRS	508
-#endif
-
 struct mmu_gather {
 	struct mm_struct *mm;
 	unsigned int fullmm;
 	unsigned int nr_ptes;
 	unsigned int nr_pxds;
-	void *array[TLB_NR_PTRS];
+	unsigned int max;
+	void **array;
+	void *local[8];
 };
 
-DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
-
-static inline struct mmu_gather *tlb_gather_mmu(struct mm_struct *mm,
-						unsigned int full_mm_flush)
+static inline void __tlb_alloc_page(struct mmu_gather *tlb)
 {
-	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);
+	unsigned long addr = __get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
 
+	if (addr) {
+		tlb->array = (void *) addr;
+		tlb->max = PAGE_SIZE / sizeof(void *);
+	}
+}
+
+static inline void tlb_gather_mmu(struct mmu_gather *tlb,
+				  struct mm_struct *mm,
+				  unsigned int full_mm_flush)
+{
 	tlb->mm = mm;
+	tlb->max = ARRAY_SIZE(tlb->local);
+	tlb->array = tlb->local;
 	tlb->fullmm = full_mm_flush;
-	tlb->nr_ptes = 0;
-	tlb->nr_pxds = TLB_NR_PTRS;
 	if (tlb->fullmm)
 		__tlb_flush_mm(mm);
-	return tlb;
+	else
+		__tlb_alloc_page(tlb);
+	tlb->nr_ptes = 0;
+	tlb->nr_pxds = tlb->max;
 }
 
-static inline void tlb_flush_mmu(struct mmu_gather *tlb,
-				 unsigned long start, unsigned long end)
+static inline void tlb_flush_mmu(struct mmu_gather *tlb)
 {
-	if (!tlb->fullmm && (tlb->nr_ptes > 0 || tlb->nr_pxds < TLB_NR_PTRS))
+	if (!tlb->fullmm && (tlb->nr_ptes > 0 || tlb->nr_pxds < tlb->max))
 		__tlb_flush_mm(tlb->mm);
 	while (tlb->nr_ptes > 0)
 		page_table_free_rcu(tlb->mm, tlb->array[--tlb->nr_ptes]);
-	while (tlb->nr_pxds < TLB_NR_PTRS)
+	while (tlb->nr_pxds < tlb->max)
 		crst_table_free_rcu(tlb->mm, tlb->array[tlb->nr_pxds++]);
 }
 
 static inline void tlb_finish_mmu(struct mmu_gather *tlb,
 				  unsigned long start, unsigned long end)
 {
-	tlb_flush_mmu(tlb, start, end);
+	tlb_flush_mmu(tlb);
 
 	rcu_table_freelist_finish();
 
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
 
-	put_cpu_var(mmu_gathers);
+	if (tlb->array != tlb->local)
+		free_pages((unsigned long) tlb->array, 0);
 }
 
 /*
  * Release the page cache reference for a pte removed by
- * tlb_ptep_clear_flush. In both flush modes the tlb fo a page cache page
+ * tlb_ptep_clear_flush. In both flush modes the tlb for a page cache page
  * has already been freed, so just do free_page_and_swap_cache.
  */
+static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	free_page_and_swap_cache(page);
+	return 1; /* avoid calling tlb_flush_mmu */
+}
+
 static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
 {
 	free_page_and_swap_cache(page);
@@ -103,7 +115,7 @@ static inline void pte_free_tlb(struct m
 	if (!tlb->fullmm) {
 		tlb->array[tlb->nr_ptes++] = pte;
 		if (tlb->nr_ptes >= tlb->nr_pxds)
-			tlb_flush_mmu(tlb, 0, 0);
+			tlb_flush_mmu(tlb);
 	} else
 		page_table_free(tlb->mm, (unsigned long *) pte);
 }
@@ -124,7 +136,7 @@ static inline void pmd_free_tlb(struct m
 	if (!tlb->fullmm) {
 		tlb->array[--tlb->nr_pxds] = pmd;
 		if (tlb->nr_ptes >= tlb->nr_pxds)
-			tlb_flush_mmu(tlb, 0, 0);
+			tlb_flush_mmu(tlb);
 	} else
 		crst_table_free(tlb->mm, (unsigned long *) pmd);
 #endif
@@ -146,7 +158,7 @@ static inline void pud_free_tlb(struct m
 	if (!tlb->fullmm) {
 		tlb->array[--tlb->nr_pxds] = pud;
 		if (tlb->nr_ptes >= tlb->nr_pxds)
-			tlb_flush_mmu(tlb, 0, 0);
+			tlb_flush_mmu(tlb);
 	} else
 		crst_table_free(tlb->mm, (unsigned long *) pud);
 #endif

[PATCH 04/20] s390: mmu_gather rework

[Peter Zijlstra]

[-- Attachment #1: martin_schwidefsky-s390-preemptible_mmu_gather.patch --]
[-- Type: text/plain, Size: 4486 bytes --]

Adapt the stand-alone s390 mmu_gather implementation to the new
preemptible mmu_gather interface.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <20101126145410.881573395@chello.nl>
---
 arch/s390/include/asm/tlb.h |   62 ++++++++++++++++++++++++++------------------
 1 file changed, 37 insertions(+), 25 deletions(-)

Index: linux-2.6/arch/s390/include/asm/tlb.h
===================================================================
--- linux-2.6.orig/arch/s390/include/asm/tlb.h
+++ linux-2.6/arch/s390/include/asm/tlb.h
@@ -29,65 +29,77 @@
 #include <asm/smp.h>
 #include <asm/tlbflush.h>
 
-#ifndef CONFIG_SMP
-#define TLB_NR_PTRS	1
-#else
-#define TLB_NR_PTRS	508
-#endif
-
 struct mmu_gather {
 	struct mm_struct *mm;
 	unsigned int fullmm;
 	unsigned int nr_ptes;
 	unsigned int nr_pxds;
-	void *array[TLB_NR_PTRS];
+	unsigned int max;
+	void **array;
+	void *local[8];
 };
 
-DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
-
-static inline struct mmu_gather *tlb_gather_mmu(struct mm_struct *mm,
-						unsigned int full_mm_flush)
+static inline void __tlb_alloc_page(struct mmu_gather *tlb)
 {
-	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);
+	unsigned long addr = __get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
 
+	if (addr) {
+		tlb->array = (void *) addr;
+		tlb->max = PAGE_SIZE / sizeof(void *);
+	}
+}
+
+static inline void tlb_gather_mmu(struct mmu_gather *tlb,
+				  struct mm_struct *mm,
+				  unsigned int full_mm_flush)
+{
 	tlb->mm = mm;
+	tlb->max = ARRAY_SIZE(tlb->local);
+	tlb->array = tlb->local;
 	tlb->fullmm = full_mm_flush;
-	tlb->nr_ptes = 0;
-	tlb->nr_pxds = TLB_NR_PTRS;
 	if (tlb->fullmm)
 		__tlb_flush_mm(mm);
-	return tlb;
+	else
+		__tlb_alloc_page(tlb);
+	tlb->nr_ptes = 0;
+	tlb->nr_pxds = tlb->max;
 }
 
-static inline void tlb_flush_mmu(struct mmu_gather *tlb,
-				 unsigned long start, unsigned long end)
+static inline void tlb_flush_mmu(struct mmu_gather *tlb)
 {
-	if (!tlb->fullmm && (tlb->nr_ptes > 0 || tlb->nr_pxds < TLB_NR_PTRS))
+	if (!tlb->fullmm && (tlb->nr_ptes > 0 || tlb->nr_pxds < tlb->max))
 		__tlb_flush_mm(tlb->mm);
 	while (tlb->nr_ptes > 0)
 		page_table_free_rcu(tlb->mm, tlb->array[--tlb->nr_ptes]);
-	while (tlb->nr_pxds < TLB_NR_PTRS)
+	while (tlb->nr_pxds < tlb->max)
 		crst_table_free_rcu(tlb->mm, tlb->array[tlb->nr_pxds++]);
 }
 
 static inline void tlb_finish_mmu(struct mmu_gather *tlb,
 				  unsigned long start, unsigned long end)
 {
-	tlb_flush_mmu(tlb, start, end);
+	tlb_flush_mmu(tlb);
 
 	rcu_table_freelist_finish();
 
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
 
-	put_cpu_var(mmu_gathers);
+	if (tlb->array != tlb->local)
+		free_pages((unsigned long) tlb->array, 0);
 }
 
 /*
  * Release the page cache reference for a pte removed by
- * tlb_ptep_clear_flush. In both flush modes the tlb fo a page cache page
+ * tlb_ptep_clear_flush. In both flush modes the tlb for a page cache page
  * has already been freed, so just do free_page_and_swap_cache.
  */
+static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	free_page_and_swap_cache(page);
+	return 1; /* avoid calling tlb_flush_mmu */
+}
+
 static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
 {
 	free_page_and_swap_cache(page);
@@ -103,7 +115,7 @@ static inline void pte_free_tlb(struct m
 	if (!tlb->fullmm) {
 		tlb->array[tlb->nr_ptes++] = pte;
 		if (tlb->nr_ptes >= tlb->nr_pxds)
-			tlb_flush_mmu(tlb, 0, 0);
+			tlb_flush_mmu(tlb);
 	} else
 		page_table_free(tlb->mm, (unsigned long *) pte);
 }
@@ -124,7 +136,7 @@ static inline void pmd_free_tlb(struct m
 	if (!tlb->fullmm) {
 		tlb->array[--tlb->nr_pxds] = pmd;
 		if (tlb->nr_ptes >= tlb->nr_pxds)
-			tlb_flush_mmu(tlb, 0, 0);
+			tlb_flush_mmu(tlb);
 	} else
 		crst_table_free(tlb->mm, (unsigned long *) pmd);
 #endif
@@ -146,7 +158,7 @@ static inline void pud_free_tlb(struct m
 	if (!tlb->fullmm) {
 		tlb->array[--tlb->nr_pxds] = pud;
 		if (tlb->nr_ptes >= tlb->nr_pxds)
-			tlb_flush_mmu(tlb, 0, 0);
+			tlb_flush_mmu(tlb);
 	} else
 		crst_table_free(tlb->mm, (unsigned long *) pud);
 #endif


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 04/20] s390: mmu_gather rework

[Peter Zijlstra]

[-- Attachment #1: martin_schwidefsky-s390-preemptible_mmu_gather.patch --]
[-- Type: text/plain, Size: 4486 bytes --]

Adapt the stand-alone s390 mmu_gather implementation to the new
preemptible mmu_gather interface.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <20101126145410.881573395@chello.nl>
---
 arch/s390/include/asm/tlb.h |   62 ++++++++++++++++++++++++++------------------
 1 file changed, 37 insertions(+), 25 deletions(-)

Index: linux-2.6/arch/s390/include/asm/tlb.h
===================================================================
--- linux-2.6.orig/arch/s390/include/asm/tlb.h
+++ linux-2.6/arch/s390/include/asm/tlb.h
@@ -29,65 +29,77 @@
 #include <asm/smp.h>
 #include <asm/tlbflush.h>
 
-#ifndef CONFIG_SMP
-#define TLB_NR_PTRS	1
-#else
-#define TLB_NR_PTRS	508
-#endif
-
 struct mmu_gather {
 	struct mm_struct *mm;
 	unsigned int fullmm;
 	unsigned int nr_ptes;
 	unsigned int nr_pxds;
-	void *array[TLB_NR_PTRS];
+	unsigned int max;
+	void **array;
+	void *local[8];
 };
 
-DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
-
-static inline struct mmu_gather *tlb_gather_mmu(struct mm_struct *mm,
-						unsigned int full_mm_flush)
+static inline void __tlb_alloc_page(struct mmu_gather *tlb)
 {
-	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);
+	unsigned long addr = __get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
 
+	if (addr) {
+		tlb->array = (void *) addr;
+		tlb->max = PAGE_SIZE / sizeof(void *);
+	}
+}
+
+static inline void tlb_gather_mmu(struct mmu_gather *tlb,
+				  struct mm_struct *mm,
+				  unsigned int full_mm_flush)
+{
 	tlb->mm = mm;
+	tlb->max = ARRAY_SIZE(tlb->local);
+	tlb->array = tlb->local;
 	tlb->fullmm = full_mm_flush;
-	tlb->nr_ptes = 0;
-	tlb->nr_pxds = TLB_NR_PTRS;
 	if (tlb->fullmm)
 		__tlb_flush_mm(mm);
-	return tlb;
+	else
+		__tlb_alloc_page(tlb);
+	tlb->nr_ptes = 0;
+	tlb->nr_pxds = tlb->max;
 }
 
-static inline void tlb_flush_mmu(struct mmu_gather *tlb,
-				 unsigned long start, unsigned long end)
+static inline void tlb_flush_mmu(struct mmu_gather *tlb)
 {
-	if (!tlb->fullmm && (tlb->nr_ptes > 0 || tlb->nr_pxds < TLB_NR_PTRS))
+	if (!tlb->fullmm && (tlb->nr_ptes > 0 || tlb->nr_pxds < tlb->max))
 		__tlb_flush_mm(tlb->mm);
 	while (tlb->nr_ptes > 0)
 		page_table_free_rcu(tlb->mm, tlb->array[--tlb->nr_ptes]);
-	while (tlb->nr_pxds < TLB_NR_PTRS)
+	while (tlb->nr_pxds < tlb->max)
 		crst_table_free_rcu(tlb->mm, tlb->array[tlb->nr_pxds++]);
 }
 
 static inline void tlb_finish_mmu(struct mmu_gather *tlb,
 				  unsigned long start, unsigned long end)
 {
-	tlb_flush_mmu(tlb, start, end);
+	tlb_flush_mmu(tlb);
 
 	rcu_table_freelist_finish();
 
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
 
-	put_cpu_var(mmu_gathers);
+	if (tlb->array != tlb->local)
+		free_pages((unsigned long) tlb->array, 0);
 }
 
 /*
  * Release the page cache reference for a pte removed by
- * tlb_ptep_clear_flush. In both flush modes the tlb fo a page cache page
+ * tlb_ptep_clear_flush. In both flush modes the tlb for a page cache page
  * has already been freed, so just do free_page_and_swap_cache.
  */
+static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	free_page_and_swap_cache(page);
+	return 1; /* avoid calling tlb_flush_mmu */
+}
+
 static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
 {
 	free_page_and_swap_cache(page);
@@ -103,7 +115,7 @@ static inline void pte_free_tlb(struct m
 	if (!tlb->fullmm) {
 		tlb->array[tlb->nr_ptes++] = pte;
 		if (tlb->nr_ptes >= tlb->nr_pxds)
-			tlb_flush_mmu(tlb, 0, 0);
+			tlb_flush_mmu(tlb);
 	} else
 		page_table_free(tlb->mm, (unsigned long *) pte);
 }
@@ -124,7 +136,7 @@ static inline void pmd_free_tlb(struct m
 	if (!tlb->fullmm) {
 		tlb->array[--tlb->nr_pxds] = pmd;
 		if (tlb->nr_ptes >= tlb->nr_pxds)
-			tlb_flush_mmu(tlb, 0, 0);
+			tlb_flush_mmu(tlb);
 	} else
 		crst_table_free(tlb->mm, (unsigned long *) pmd);
 #endif
@@ -146,7 +158,7 @@ static inline void pud_free_tlb(struct m
 	if (!tlb->fullmm) {
 		tlb->array[--tlb->nr_pxds] = pud;
 		if (tlb->nr_ptes >= tlb->nr_pxds)
-			tlb_flush_mmu(tlb, 0, 0);
+			tlb_flush_mmu(tlb);
 	} else
 		crst_table_free(tlb->mm, (unsigned long *) pud);
 #endif


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 05/20] arm: mmu_gather rework

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-arm-preemptible_mmu_gather.patch --]
[-- Type: text/plain, Size: 3256 bytes --]

Fix up the arm mmu_gather code to conform to the new API.

Cc: Russell King <rmk@arm.linux.org.uk>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 arch/arm/include/asm/tlb.h |   53 +++++++++++++++++++++++++++++++--------------
 1 file changed, 37 insertions(+), 16 deletions(-)

Index: linux-2.6/arch/arm/include/asm/tlb.h
===================================================================
--- linux-2.6.orig/arch/arm/include/asm/tlb.h
+++ linux-2.6/arch/arm/include/asm/tlb.h
@@ -41,12 +41,12 @@
  */
 #if defined(CONFIG_SMP) || defined(CONFIG_CPU_32v7)
 #define tlb_fast_mode(tlb)	0
-#define FREE_PTE_NR		500
 #else
 #define tlb_fast_mode(tlb)	1
-#define FREE_PTE_NR		0
 #endif
 
+#define MMU_GATHER_BUNDLE	8
+
 /*
  * TLB handling.  This allows us to remove pages from the page
  * tables, and efficiently handle the TLB issues.
@@ -58,7 +58,9 @@ struct mmu_gather {
 	unsigned long		range_start;
 	unsigned long		range_end;
 	unsigned int		nr;
-	struct page		*pages[FREE_PTE_NR];
+	unsigned int		max;
+	struct page		**pages;
+	struct page		*local[MMU_GATHER_BUNDLE];
 };
 
 DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
@@ -97,26 +99,37 @@ static inline void tlb_add_flush(struct 
 	}
 }
 
+static inline void __tlb_alloc_page(struct mmu_gather *tlb)
+{
+	unsigned long addr = __get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
+
+	if (addr) {
+		tlb->pages = (void *)addr;
+		tlb->max = PAGE_SIZE / sizeof(struct page *);
+	}
+}
+
 static inline void tlb_flush_mmu(struct mmu_gather *tlb)
 {
 	tlb_flush(tlb);
 	if (!tlb_fast_mode(tlb)) {
 		free_pages_and_swap_cache(tlb->pages, tlb->nr);
 		tlb->nr = 0;
+		if (tlb->pages == tlb->local)
+			__tlb_alloc_page(tlb);
 	}
 }
 
-static inline struct mmu_gather *
-tlb_gather_mmu(struct mm_struct *mm, unsigned int full_mm_flush)
+static inline void
+tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned int fullmm)
 {
-	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);
-
 	tlb->mm = mm;
-	tlb->fullmm = full_mm_flush;
+	tlb->fullmm = fullmm;
 	tlb->vma = NULL;
+	tlb->max = ARRAY_SIZE(tlb->local);
+	tlb->pages = tlb->local;
 	tlb->nr = 0;
-
-	return tlb;
+	__tlb_alloc_page(tlb);
 }
 
 static inline void
@@ -127,7 +140,8 @@ tlb_finish_mmu(struct mmu_gather *tlb, u
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
 
-	put_cpu_var(mmu_gathers);
+	if (tlb->pages != tlb->local)
+		free_pages((unsigned long)tlb->pages, 0);
 }
 
 /*
@@ -162,15 +176,22 @@ tlb_end_vma(struct mmu_gather *tlb, stru
 		tlb_flush(tlb);
 }
 
-static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
 {
 	if (tlb_fast_mode(tlb)) {
 		free_page_and_swap_cache(page);
-	} else {
-		tlb->pages[tlb->nr++] = page;
-		if (tlb->nr >= FREE_PTE_NR)
-			tlb_flush_mmu(tlb);
+		return 1; /* avoid calling tlb_flush_mmu */
 	}
+
+	tlb->pages[tlb->nr++] = page;
+	VM_BUG_ON(tlb->nr > tlb->max);
+	return tlb->max - tlb->nr;
+}
+
+static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	if (!__tlb_remove_page(tlb, page))
+		tlb_flush_mmu(tlb);
 }
 
 static inline void __pte_free_tlb(struct mmu_gather *tlb, pgtable_t pte,

[PATCH 05/20] arm: mmu_gather rework

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-arm-preemptible_mmu_gather.patch --]
[-- Type: text/plain, Size: 3559 bytes --]

Fix up the arm mmu_gather code to conform to the new API.

Cc: Russell King <rmk@arm.linux.org.uk>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 arch/arm/include/asm/tlb.h |   53 +++++++++++++++++++++++++++++++--------------
 1 file changed, 37 insertions(+), 16 deletions(-)

Index: linux-2.6/arch/arm/include/asm/tlb.h
===================================================================
--- linux-2.6.orig/arch/arm/include/asm/tlb.h
+++ linux-2.6/arch/arm/include/asm/tlb.h
@@ -41,12 +41,12 @@
  */
 #if defined(CONFIG_SMP) || defined(CONFIG_CPU_32v7)
 #define tlb_fast_mode(tlb)	0
-#define FREE_PTE_NR		500
 #else
 #define tlb_fast_mode(tlb)	1
-#define FREE_PTE_NR		0
 #endif
 
+#define MMU_GATHER_BUNDLE	8
+
 /*
  * TLB handling.  This allows us to remove pages from the page
  * tables, and efficiently handle the TLB issues.
@@ -58,7 +58,9 @@ struct mmu_gather {
 	unsigned long		range_start;
 	unsigned long		range_end;
 	unsigned int		nr;
-	struct page		*pages[FREE_PTE_NR];
+	unsigned int		max;
+	struct page		**pages;
+	struct page		*local[MMU_GATHER_BUNDLE];
 };
 
 DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
@@ -97,26 +99,37 @@ static inline void tlb_add_flush(struct 
 	}
 }
 
+static inline void __tlb_alloc_page(struct mmu_gather *tlb)
+{
+	unsigned long addr = __get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
+
+	if (addr) {
+		tlb->pages = (void *)addr;
+		tlb->max = PAGE_SIZE / sizeof(struct page *);
+	}
+}
+
 static inline void tlb_flush_mmu(struct mmu_gather *tlb)
 {
 	tlb_flush(tlb);
 	if (!tlb_fast_mode(tlb)) {
 		free_pages_and_swap_cache(tlb->pages, tlb->nr);
 		tlb->nr = 0;
+		if (tlb->pages == tlb->local)
+			__tlb_alloc_page(tlb);
 	}
 }
 
-static inline struct mmu_gather *
-tlb_gather_mmu(struct mm_struct *mm, unsigned int full_mm_flush)
+static inline void
+tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned int fullmm)
 {
-	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);
-
 	tlb->mm = mm;
-	tlb->fullmm = full_mm_flush;
+	tlb->fullmm = fullmm;
 	tlb->vma = NULL;
+	tlb->max = ARRAY_SIZE(tlb->local);
+	tlb->pages = tlb->local;
 	tlb->nr = 0;
-
-	return tlb;
+	__tlb_alloc_page(tlb);
 }
 
 static inline void
@@ -127,7 +140,8 @@ tlb_finish_mmu(struct mmu_gather *tlb, u
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
 
-	put_cpu_var(mmu_gathers);
+	if (tlb->pages != tlb->local)
+		free_pages((unsigned long)tlb->pages, 0);
 }
 
 /*
@@ -162,15 +176,22 @@ tlb_end_vma(struct mmu_gather *tlb, stru
 		tlb_flush(tlb);
 }
 
-static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
 {
 	if (tlb_fast_mode(tlb)) {
 		free_page_and_swap_cache(page);
-	} else {
-		tlb->pages[tlb->nr++] = page;
-		if (tlb->nr >= FREE_PTE_NR)
-			tlb_flush_mmu(tlb);
+		return 1; /* avoid calling tlb_flush_mmu */
 	}
+
+	tlb->pages[tlb->nr++] = page;
+	VM_BUG_ON(tlb->nr > tlb->max);
+	return tlb->max - tlb->nr;
+}
+
+static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	if (!__tlb_remove_page(tlb, page))
+		tlb_flush_mmu(tlb);
 }
 
 static inline void __pte_free_tlb(struct mmu_gather *tlb, pgtable_t pte,


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 05/20] arm: mmu_gather rework

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-arm-preemptible_mmu_gather.patch --]
[-- Type: text/plain, Size: 3559 bytes --]

Fix up the arm mmu_gather code to conform to the new API.

Cc: Russell King <rmk@arm.linux.org.uk>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 arch/arm/include/asm/tlb.h |   53 +++++++++++++++++++++++++++++++--------------
 1 file changed, 37 insertions(+), 16 deletions(-)

Index: linux-2.6/arch/arm/include/asm/tlb.h
===================================================================
--- linux-2.6.orig/arch/arm/include/asm/tlb.h
+++ linux-2.6/arch/arm/include/asm/tlb.h
@@ -41,12 +41,12 @@
  */
 #if defined(CONFIG_SMP) || defined(CONFIG_CPU_32v7)
 #define tlb_fast_mode(tlb)	0
-#define FREE_PTE_NR		500
 #else
 #define tlb_fast_mode(tlb)	1
-#define FREE_PTE_NR		0
 #endif
 
+#define MMU_GATHER_BUNDLE	8
+
 /*
  * TLB handling.  This allows us to remove pages from the page
  * tables, and efficiently handle the TLB issues.
@@ -58,7 +58,9 @@ struct mmu_gather {
 	unsigned long		range_start;
 	unsigned long		range_end;
 	unsigned int		nr;
-	struct page		*pages[FREE_PTE_NR];
+	unsigned int		max;
+	struct page		**pages;
+	struct page		*local[MMU_GATHER_BUNDLE];
 };
 
 DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
@@ -97,26 +99,37 @@ static inline void tlb_add_flush(struct 
 	}
 }
 
+static inline void __tlb_alloc_page(struct mmu_gather *tlb)
+{
+	unsigned long addr = __get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
+
+	if (addr) {
+		tlb->pages = (void *)addr;
+		tlb->max = PAGE_SIZE / sizeof(struct page *);
+	}
+}
+
 static inline void tlb_flush_mmu(struct mmu_gather *tlb)
 {
 	tlb_flush(tlb);
 	if (!tlb_fast_mode(tlb)) {
 		free_pages_and_swap_cache(tlb->pages, tlb->nr);
 		tlb->nr = 0;
+		if (tlb->pages == tlb->local)
+			__tlb_alloc_page(tlb);
 	}
 }
 
-static inline struct mmu_gather *
-tlb_gather_mmu(struct mm_struct *mm, unsigned int full_mm_flush)
+static inline void
+tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned int fullmm)
 {
-	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);
-
 	tlb->mm = mm;
-	tlb->fullmm = full_mm_flush;
+	tlb->fullmm = fullmm;
 	tlb->vma = NULL;
+	tlb->max = ARRAY_SIZE(tlb->local);
+	tlb->pages = tlb->local;
 	tlb->nr = 0;
-
-	return tlb;
+	__tlb_alloc_page(tlb);
 }
 
 static inline void
@@ -127,7 +140,8 @@ tlb_finish_mmu(struct mmu_gather *tlb, u
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
 
-	put_cpu_var(mmu_gathers);
+	if (tlb->pages != tlb->local)
+		free_pages((unsigned long)tlb->pages, 0);
 }
 
 /*
@@ -162,15 +176,22 @@ tlb_end_vma(struct mmu_gather *tlb, stru
 		tlb_flush(tlb);
 }
 
-static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
 {
 	if (tlb_fast_mode(tlb)) {
 		free_page_and_swap_cache(page);
-	} else {
-		tlb->pages[tlb->nr++] = page;
-		if (tlb->nr >= FREE_PTE_NR)
-			tlb_flush_mmu(tlb);
+		return 1; /* avoid calling tlb_flush_mmu */
 	}
+
+	tlb->pages[tlb->nr++] = page;
+	VM_BUG_ON(tlb->nr > tlb->max);
+	return tlb->max - tlb->nr;
+}
+
+static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	if (!__tlb_remove_page(tlb, page))
+		tlb_flush_mmu(tlb);
 }
 
 static inline void __pte_free_tlb(struct mmu_gather *tlb, pgtable_t pte,


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 06/20] sh: mmu_gather rework

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-sh-preemptible_mmu_gather.patch --]
[-- Type: text/plain, Size: 2036 bytes --]

Fix up the sh mmu_gather code to conform to the new API.

Cc: Paul Mundt <lethal@linux-sh.org>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 arch/sh/include/asm/tlb.h |   28 +++++++++++++++++-----------
 1 file changed, 17 insertions(+), 11 deletions(-)

Index: linux-2.6/arch/sh/include/asm/tlb.h
===================================================================
--- linux-2.6.orig/arch/sh/include/asm/tlb.h
+++ linux-2.6/arch/sh/include/asm/tlb.h
@@ -23,8 +23,6 @@ struct mmu_gather {
 	unsigned long		start, end;
 };
 
-DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 static inline void init_tlb_gather(struct mmu_gather *tlb)
 {
 	tlb->start = TASK_SIZE;
@@ -36,17 +34,13 @@ static inline void init_tlb_gather(struc
 	}
 }
 
-static inline struct mmu_gather *
-tlb_gather_mmu(struct mm_struct *mm, unsigned int full_mm_flush)
+static inline void
+tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned int full_mm_flush)
 {
-	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);
-
 	tlb->mm = mm;
 	tlb->fullmm = full_mm_flush;
 
 	init_tlb_gather(tlb);
-
-	return tlb;
 }
 
 static inline void
@@ -57,8 +51,6 @@ tlb_finish_mmu(struct mmu_gather *tlb, u
 
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
-
-	put_cpu_var(mmu_gathers);
 }
 
 static inline void
@@ -91,7 +83,21 @@ tlb_end_vma(struct mmu_gather *tlb, stru
 	}
 }
 
-#define tlb_remove_page(tlb,page)	free_page_and_swap_cache(page)
+static inline void tlb_flush_mmu(struct mmu_gather *tlb)
+{
+}
+
+static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	free_page_and_swap_cache(page);
+	return 1; /* avoid calling tlb_flush_mmu */
+}
+
+static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	__tlb_remove_page(tlb, page);
+}
+
 #define pte_free_tlb(tlb, ptep, addr)	pte_free((tlb)->mm, ptep)
 #define pmd_free_tlb(tlb, pmdp, addr)	pmd_free((tlb)->mm, pmdp)
 #define pud_free_tlb(tlb, pudp, addr)	pud_free((tlb)->mm, pudp)

[PATCH 06/20] sh: mmu_gather rework

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-sh-preemptible_mmu_gather.patch --]
[-- Type: text/plain, Size: 2339 bytes --]

Fix up the sh mmu_gather code to conform to the new API.

Cc: Paul Mundt <lethal@linux-sh.org>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 arch/sh/include/asm/tlb.h |   28 +++++++++++++++++-----------
 1 file changed, 17 insertions(+), 11 deletions(-)

Index: linux-2.6/arch/sh/include/asm/tlb.h
===================================================================
--- linux-2.6.orig/arch/sh/include/asm/tlb.h
+++ linux-2.6/arch/sh/include/asm/tlb.h
@@ -23,8 +23,6 @@ struct mmu_gather {
 	unsigned long		start, end;
 };
 
-DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 static inline void init_tlb_gather(struct mmu_gather *tlb)
 {
 	tlb->start = TASK_SIZE;
@@ -36,17 +34,13 @@ static inline void init_tlb_gather(struc
 	}
 }
 
-static inline struct mmu_gather *
-tlb_gather_mmu(struct mm_struct *mm, unsigned int full_mm_flush)
+static inline void
+tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned int full_mm_flush)
 {
-	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);
-
 	tlb->mm = mm;
 	tlb->fullmm = full_mm_flush;
 
 	init_tlb_gather(tlb);
-
-	return tlb;
 }
 
 static inline void
@@ -57,8 +51,6 @@ tlb_finish_mmu(struct mmu_gather *tlb, u
 
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
-
-	put_cpu_var(mmu_gathers);
 }
 
 static inline void
@@ -91,7 +83,21 @@ tlb_end_vma(struct mmu_gather *tlb, stru
 	}
 }
 
-#define tlb_remove_page(tlb,page)	free_page_and_swap_cache(page)
+static inline void tlb_flush_mmu(struct mmu_gather *tlb)
+{
+}
+
+static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	free_page_and_swap_cache(page);
+	return 1; /* avoid calling tlb_flush_mmu */
+}
+
+static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	__tlb_remove_page(tlb, page);
+}
+
 #define pte_free_tlb(tlb, ptep, addr)	pte_free((tlb)->mm, ptep)
 #define pmd_free_tlb(tlb, pmdp, addr)	pmd_free((tlb)->mm, pmdp)
 #define pud_free_tlb(tlb, pudp, addr)	pud_free((tlb)->mm, pudp)


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 06/20] sh: mmu_gather rework

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-sh-preemptible_mmu_gather.patch --]
[-- Type: text/plain, Size: 2339 bytes --]

Fix up the sh mmu_gather code to conform to the new API.

Cc: Paul Mundt <lethal@linux-sh.org>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 arch/sh/include/asm/tlb.h |   28 +++++++++++++++++-----------
 1 file changed, 17 insertions(+), 11 deletions(-)

Index: linux-2.6/arch/sh/include/asm/tlb.h
===================================================================
--- linux-2.6.orig/arch/sh/include/asm/tlb.h
+++ linux-2.6/arch/sh/include/asm/tlb.h
@@ -23,8 +23,6 @@ struct mmu_gather {
 	unsigned long		start, end;
 };
 
-DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 static inline void init_tlb_gather(struct mmu_gather *tlb)
 {
 	tlb->start = TASK_SIZE;
@@ -36,17 +34,13 @@ static inline void init_tlb_gather(struc
 	}
 }
 
-static inline struct mmu_gather *
-tlb_gather_mmu(struct mm_struct *mm, unsigned int full_mm_flush)
+static inline void
+tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned int full_mm_flush)
 {
-	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);
-
 	tlb->mm = mm;
 	tlb->fullmm = full_mm_flush;
 
 	init_tlb_gather(tlb);
-
-	return tlb;
 }
 
 static inline void
@@ -57,8 +51,6 @@ tlb_finish_mmu(struct mmu_gather *tlb, u
 
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
-
-	put_cpu_var(mmu_gathers);
 }
 
 static inline void
@@ -91,7 +83,21 @@ tlb_end_vma(struct mmu_gather *tlb, stru
 	}
 }
 
-#define tlb_remove_page(tlb,page)	free_page_and_swap_cache(page)
+static inline void tlb_flush_mmu(struct mmu_gather *tlb)
+{
+}
+
+static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	free_page_and_swap_cache(page);
+	return 1; /* avoid calling tlb_flush_mmu */
+}
+
+static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	__tlb_remove_page(tlb, page);
+}
+
 #define pte_free_tlb(tlb, ptep, addr)	pte_free((tlb)->mm, ptep)
 #define pmd_free_tlb(tlb, pmdp, addr)	pmd_free((tlb)->mm, pmdp)
 #define pud_free_tlb(tlb, pudp, addr)	pud_free((tlb)->mm, pudp)


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 07/20] ia64: mmu_gather rework

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-ia64-preemptible_mmu_gather.patch --]
[-- Type: text/plain, Size: 4375 bytes --]

Fix up the ia64 mmu_gather code to conform to the new API.

Acked-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 arch/ia64/include/asm/tlb.h |   66 ++++++++++++++++++++++++++++++--------------
 1 file changed, 46 insertions(+), 20 deletions(-)

Index: linux-2.6/arch/ia64/include/asm/tlb.h
===================================================================
--- linux-2.6.orig/arch/ia64/include/asm/tlb.h
+++ linux-2.6/arch/ia64/include/asm/tlb.h
@@ -47,21 +47,27 @@
 #include <asm/machvec.h>
 
 #ifdef CONFIG_SMP
-# define FREE_PTE_NR		2048
 # define tlb_fast_mode(tlb)	((tlb)->nr == ~0U)
 #else
-# define FREE_PTE_NR		0
 # define tlb_fast_mode(tlb)	(1)
 #endif
 
+/*
+ * If we can't allocate a page to make a big batch of page pointers
+ * to work on, then just handle a few from the on-stack structure.
+ */
+#define	IA64_GATHER_BUNDLE	8
+
 struct mmu_gather {
 	struct mm_struct	*mm;
 	unsigned int		nr;		/* == ~0U => fast mode */
+	unsigned int		max;
 	unsigned char		fullmm;		/* non-zero means full mm flush */
 	unsigned char		need_flush;	/* really unmapped some PTEs? */
 	unsigned long		start_addr;
 	unsigned long		end_addr;
-	struct page 		*pages[FREE_PTE_NR];
+	struct page		**pages;
+	struct page		*local[IA64_GATHER_BUNDLE];
 };
 
 struct ia64_tr_entry {
@@ -90,9 +96,6 @@ extern struct ia64_tr_entry *ia64_idtrs[
 #define RR_RID_MASK	0x00000000ffffff00L
 #define RR_TO_RID(val) 	((val >> 8) & 0xffffff)
 
-/* Users of the generic TLB shootdown code must declare this storage space. */
-DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /*
  * Flush the TLB for address range START to END and, if not in fast mode, release the
  * freed pages that where gathered up to this point.
@@ -147,15 +150,23 @@ ia64_tlb_flush_mmu (struct mmu_gather *t
 	}
 }
 
-/*
- * Return a pointer to an initialized struct mmu_gather.
- */
-static inline struct mmu_gather *
-tlb_gather_mmu (struct mm_struct *mm, unsigned int full_mm_flush)
+static inline void __tlb_alloc_page(struct mmu_gather *tlb)
 {
-	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);
+	unsigned long addr = __get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
 
+	if (addr) {
+		tlb->pages = (void *)addr;
+		tlb->max = PAGE_SIZE / sizeof(void *);
+	}
+}
+
+
+static inline void
+tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned int full_mm_flush)
+{
 	tlb->mm = mm;
+	tlb->max = ARRAY_SIZE(tlb->local);
+	tlb->pages = tlb->local;
 	/*
 	 * Use fast mode if only 1 CPU is online.
 	 *
@@ -172,7 +183,6 @@ tlb_gather_mmu (struct mm_struct *mm, un
 	tlb->nr = (num_online_cpus() == 1) ? ~0U : 0;
 	tlb->fullmm = full_mm_flush;
 	tlb->start_addr = ~0UL;
-	return tlb;
 }
 
 /*
@@ -180,7 +190,7 @@ tlb_gather_mmu (struct mm_struct *mm, un
  * collected.
  */
 static inline void
-tlb_finish_mmu (struct mmu_gather *tlb, unsigned long start, unsigned long end)
+tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
 {
 	/*
 	 * Note: tlb->nr may be 0 at this point, so we can't rely on tlb->start_addr and
@@ -191,7 +201,8 @@ tlb_finish_mmu (struct mmu_gather *tlb, 
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
 
-	put_cpu_var(mmu_gathers);
+	if (tlb->pages != tlb->local)
+		free_pages((unsigned long)tlb->pages, 0);
 }
 
 /*
@@ -199,18 +210,33 @@ tlb_finish_mmu (struct mmu_gather *tlb, 
  * must be delayed until after the TLB has been flushed (see comments at the beginning of
  * this file).
  */
-static inline void
-tlb_remove_page (struct mmu_gather *tlb, struct page *page)
+static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
 {
 	tlb->need_flush = 1;
 
 	if (tlb_fast_mode(tlb)) {
 		free_page_and_swap_cache(page);
-		return;
+		return 1; /* avoid calling tlb_flush_mmu */
 	}
+
+	if (!tlb->nr && tlb->pages == tlb->local)
+		__tlb_alloc_page(tlb);
+
 	tlb->pages[tlb->nr++] = page;
-	if (tlb->nr >= FREE_PTE_NR)
-		ia64_tlb_flush_mmu(tlb, tlb->start_addr, tlb->end_addr);
+	VM_BUG_ON(tlb->nr > tlb->max);
+
+	return tlb->max - tlb->nr;
+}
+
+static inline void tlb_flush_mmu(struct mmu_gather *tlb)
+{
+	ia64_tlb_flush_mmu(tlb, tlb->start_addr, tlb->end_addr);
+}
+
+static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	if (!__tlb_remove_page(tlb, page))
+		tlb_flush_mmu(tlb);
 }
 
 /*

[PATCH 07/20] ia64: mmu_gather rework

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-ia64-preemptible_mmu_gather.patch --]
[-- Type: text/plain, Size: 4373 bytes --]

Fix up the ia64 mmu_gather code to conform to the new API.

Acked-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 arch/ia64/include/asm/tlb.h |   66 ++++++++++++++++++++++++++++++--------------
 1 file changed, 46 insertions(+), 20 deletions(-)

Index: linux-2.6/arch/ia64/include/asm/tlb.h
===================================================================
--- linux-2.6.orig/arch/ia64/include/asm/tlb.h
+++ linux-2.6/arch/ia64/include/asm/tlb.h
@@ -47,21 +47,27 @@
 #include <asm/machvec.h>
 
 #ifdef CONFIG_SMP
-# define FREE_PTE_NR		2048
 # define tlb_fast_mode(tlb)	((tlb)->nr == ~0U)
 #else
-# define FREE_PTE_NR		0
 # define tlb_fast_mode(tlb)	(1)
 #endif
 
+/*
+ * If we can't allocate a page to make a big batch of page pointers
+ * to work on, then just handle a few from the on-stack structure.
+ */
+#define	IA64_GATHER_BUNDLE	8
+
 struct mmu_gather {
 	struct mm_struct	*mm;
 	unsigned int		nr;		/* == ~0U => fast mode */
+	unsigned int		max;
 	unsigned char		fullmm;		/* non-zero means full mm flush */
 	unsigned char		need_flush;	/* really unmapped some PTEs? */
 	unsigned long		start_addr;
 	unsigned long		end_addr;
-	struct page 		*pages[FREE_PTE_NR];
+	struct page		**pages;
+	struct page		*local[IA64_GATHER_BUNDLE];
 };
 
 struct ia64_tr_entry {
@@ -90,9 +96,6 @@ extern struct ia64_tr_entry *ia64_idtrs[
 #define RR_RID_MASK	0x00000000ffffff00L
 #define RR_TO_RID(val) 	((val >> 8) & 0xffffff)
 
-/* Users of the generic TLB shootdown code must declare this storage space. */
-DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /*
  * Flush the TLB for address range START to END and, if not in fast mode, release the
  * freed pages that where gathered up to this point.
@@ -147,15 +150,23 @@ ia64_tlb_flush_mmu (struct mmu_gather *t
 	}
 }
 
-/*
- * Return a pointer to an initialized struct mmu_gather.
- */
-static inline struct mmu_gather *
-tlb_gather_mmu (struct mm_struct *mm, unsigned int full_mm_flush)
+static inline void __tlb_alloc_page(struct mmu_gather *tlb)
 {
-	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);
+	unsigned long addr = __get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
 
+	if (addr) {
+		tlb->pages = (void *)addr;
+		tlb->max = PAGE_SIZE / sizeof(void *);
+	}
+}
+
+
+static inline void
+tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned int full_mm_flush)
+{
 	tlb->mm = mm;
+	tlb->max = ARRAY_SIZE(tlb->local);
+	tlb->pages = tlb->local;
 	/*
 	 * Use fast mode if only 1 CPU is online.
 	 *
@@ -172,7 +183,6 @@ tlb_gather_mmu (struct mm_struct *mm, un
 	tlb->nr = (num_online_cpus() == 1) ? ~0U : 0;
 	tlb->fullmm = full_mm_flush;
 	tlb->start_addr = ~0UL;
-	return tlb;
 }
 
 /*
@@ -180,7 +190,7 @@ tlb_gather_mmu (struct mm_struct *mm, un
  * collected.
  */
 static inline void
-tlb_finish_mmu (struct mmu_gather *tlb, unsigned long start, unsigned long end)
+tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
 {
 	/*
 	 * Note: tlb->nr may be 0 at this point, so we can't rely on tlb->start_addr and
@@ -191,7 +201,8 @@ tlb_finish_mmu (struct mmu_gather *tlb, 
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
 
-	put_cpu_var(mmu_gathers);
+	if (tlb->pages != tlb->local)
+		free_pages((unsigned long)tlb->pages, 0);
 }
 
 /*
@@ -199,18 +210,33 @@ tlb_finish_mmu (struct mmu_gather *tlb, 
  * must be delayed until after the TLB has been flushed (see comments at the beginning of
  * this file).
  */
-static inline void
-tlb_remove_page (struct mmu_gather *tlb, struct page *page)
+static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
 {
 	tlb->need_flush = 1;
 
 	if (tlb_fast_mode(tlb)) {
 		free_page_and_swap_cache(page);
-		return;
+		return 1; /* avoid calling tlb_flush_mmu */
 	}
+
+	if (!tlb->nr && tlb->pages == tlb->local)
+		__tlb_alloc_page(tlb);
+
 	tlb->pages[tlb->nr++] = page;
-	if (tlb->nr >= FREE_PTE_NR)
-		ia64_tlb_flush_mmu(tlb, tlb->start_addr, tlb->end_addr);
+	VM_BUG_ON(tlb->nr > tlb->max);
+
+	return tlb->max - tlb->nr;
+}
+
+static inline void tlb_flush_mmu(struct mmu_gather *tlb)
+{
+	ia64_tlb_flush_mmu(tlb, tlb->start_addr, tlb->end_addr);
+}
+
+static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	if (!__tlb_remove_page(tlb, page))
+		tlb_flush_mmu(tlb);
 }
 
 /*

[PATCH 07/20] ia64: mmu_gather rework

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-ia64-preemptible_mmu_gather.patch --]
[-- Type: text/plain, Size: 4678 bytes --]

Fix up the ia64 mmu_gather code to conform to the new API.

Acked-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 arch/ia64/include/asm/tlb.h |   66 ++++++++++++++++++++++++++++++--------------
 1 file changed, 46 insertions(+), 20 deletions(-)

Index: linux-2.6/arch/ia64/include/asm/tlb.h
===================================================================
--- linux-2.6.orig/arch/ia64/include/asm/tlb.h
+++ linux-2.6/arch/ia64/include/asm/tlb.h
@@ -47,21 +47,27 @@
 #include <asm/machvec.h>
 
 #ifdef CONFIG_SMP
-# define FREE_PTE_NR		2048
 # define tlb_fast_mode(tlb)	((tlb)->nr == ~0U)
 #else
-# define FREE_PTE_NR		0
 # define tlb_fast_mode(tlb)	(1)
 #endif
 
+/*
+ * If we can't allocate a page to make a big batch of page pointers
+ * to work on, then just handle a few from the on-stack structure.
+ */
+#define	IA64_GATHER_BUNDLE	8
+
 struct mmu_gather {
 	struct mm_struct	*mm;
 	unsigned int		nr;		/* == ~0U => fast mode */
+	unsigned int		max;
 	unsigned char		fullmm;		/* non-zero means full mm flush */
 	unsigned char		need_flush;	/* really unmapped some PTEs? */
 	unsigned long		start_addr;
 	unsigned long		end_addr;
-	struct page 		*pages[FREE_PTE_NR];
+	struct page		**pages;
+	struct page		*local[IA64_GATHER_BUNDLE];
 };
 
 struct ia64_tr_entry {
@@ -90,9 +96,6 @@ extern struct ia64_tr_entry *ia64_idtrs[
 #define RR_RID_MASK	0x00000000ffffff00L
 #define RR_TO_RID(val) 	((val >> 8) & 0xffffff)
 
-/* Users of the generic TLB shootdown code must declare this storage space. */
-DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /*
  * Flush the TLB for address range START to END and, if not in fast mode, release the
  * freed pages that where gathered up to this point.
@@ -147,15 +150,23 @@ ia64_tlb_flush_mmu (struct mmu_gather *t
 	}
 }
 
-/*
- * Return a pointer to an initialized struct mmu_gather.
- */
-static inline struct mmu_gather *
-tlb_gather_mmu (struct mm_struct *mm, unsigned int full_mm_flush)
+static inline void __tlb_alloc_page(struct mmu_gather *tlb)
 {
-	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);
+	unsigned long addr = __get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
 
+	if (addr) {
+		tlb->pages = (void *)addr;
+		tlb->max = PAGE_SIZE / sizeof(void *);
+	}
+}
+
+
+static inline void
+tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned int full_mm_flush)
+{
 	tlb->mm = mm;
+	tlb->max = ARRAY_SIZE(tlb->local);
+	tlb->pages = tlb->local;
 	/*
 	 * Use fast mode if only 1 CPU is online.
 	 *
@@ -172,7 +183,6 @@ tlb_gather_mmu (struct mm_struct *mm, un
 	tlb->nr = (num_online_cpus() == 1) ? ~0U : 0;
 	tlb->fullmm = full_mm_flush;
 	tlb->start_addr = ~0UL;
-	return tlb;
 }
 
 /*
@@ -180,7 +190,7 @@ tlb_gather_mmu (struct mm_struct *mm, un
  * collected.
  */
 static inline void
-tlb_finish_mmu (struct mmu_gather *tlb, unsigned long start, unsigned long end)
+tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
 {
 	/*
 	 * Note: tlb->nr may be 0 at this point, so we can't rely on tlb->start_addr and
@@ -191,7 +201,8 @@ tlb_finish_mmu (struct mmu_gather *tlb, 
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
 
-	put_cpu_var(mmu_gathers);
+	if (tlb->pages != tlb->local)
+		free_pages((unsigned long)tlb->pages, 0);
 }
 
 /*
@@ -199,18 +210,33 @@ tlb_finish_mmu (struct mmu_gather *tlb, 
  * must be delayed until after the TLB has been flushed (see comments at the beginning of
  * this file).
  */
-static inline void
-tlb_remove_page (struct mmu_gather *tlb, struct page *page)
+static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
 {
 	tlb->need_flush = 1;
 
 	if (tlb_fast_mode(tlb)) {
 		free_page_and_swap_cache(page);
-		return;
+		return 1; /* avoid calling tlb_flush_mmu */
 	}
+
+	if (!tlb->nr && tlb->pages == tlb->local)
+		__tlb_alloc_page(tlb);
+
 	tlb->pages[tlb->nr++] = page;
-	if (tlb->nr >= FREE_PTE_NR)
-		ia64_tlb_flush_mmu(tlb, tlb->start_addr, tlb->end_addr);
+	VM_BUG_ON(tlb->nr > tlb->max);
+
+	return tlb->max - tlb->nr;
+}
+
+static inline void tlb_flush_mmu(struct mmu_gather *tlb)
+{
+	ia64_tlb_flush_mmu(tlb, tlb->start_addr, tlb->end_addr);
+}
+
+static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	if (!__tlb_remove_page(tlb, page))
+		tlb_flush_mmu(tlb);
 }
 
 /*


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 08/20] um: mmu_gather rework

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-um-preemptible_mmu_gather.patch --]
[-- Type: text/plain, Size: 2573 bytes --]

Fix up the um mmu_gather code to conform to the new API.

Cc: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 arch/um/include/asm/tlb.h |   29 +++++++++++------------------
 1 file changed, 11 insertions(+), 18 deletions(-)

Index: linux-2.6/arch/um/include/asm/tlb.h
===================================================================
--- linux-2.6.orig/arch/um/include/asm/tlb.h
+++ linux-2.6/arch/um/include/asm/tlb.h
@@ -22,9 +22,6 @@ struct mmu_gather {
 	unsigned int		fullmm; /* non-zero means full mm flush */
 };
 
-/* Users of the generic TLB shootdown code must declare this storage space. */
-DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 static inline void __tlb_remove_tlb_entry(struct mmu_gather *tlb, pte_t *ptep,
 					  unsigned long address)
 {
@@ -47,27 +44,20 @@ static inline void init_tlb_gather(struc
 	}
 }
 
-/* tlb_gather_mmu
- *	Return a pointer to an initialized struct mmu_gather.
- */
-static inline struct mmu_gather *
-tlb_gather_mmu(struct mm_struct *mm, unsigned int full_mm_flush)
+static inline void
+tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned int full_mm_flush)
 {
-	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);
-
 	tlb->mm = mm;
 	tlb->fullmm = full_mm_flush;
 
 	init_tlb_gather(tlb);
-
-	return tlb;
 }
 
 extern void flush_tlb_mm_range(struct mm_struct *mm, unsigned long start,
 			       unsigned long end);
 
 static inline void
-tlb_flush_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
+tlb_flush_mmu(struct mmu_gather *tlb)
 {
 	if (!tlb->need_flush)
 		return;
@@ -83,12 +73,10 @@ tlb_flush_mmu(struct mmu_gather *tlb, un
 static inline void
 tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
 {
-	tlb_flush_mmu(tlb, start, end);
+	tlb_flush_mmu(tlb);
 
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
-
-	put_cpu_var(mmu_gathers);
 }
 
 /* tlb_remove_page
@@ -96,11 +84,16 @@ tlb_finish_mmu(struct mmu_gather *tlb, u
  *	while handling the additional races in SMP caused by other CPUs
  *	caching valid mappings in their TLBs.
  */
-static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
 {
 	tlb->need_flush = 1;
 	free_page_and_swap_cache(page);
-	return;
+	return 1; /* avoid calling tlb_flush_mmu */
+}
+
+static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	__tlb_remove_page(tlb, page);
 }
 
 /**

[PATCH 08/20] um: mmu_gather rework

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-um-preemptible_mmu_gather.patch --]
[-- Type: text/plain, Size: 2571 bytes --]

Fix up the um mmu_gather code to conform to the new API.

Cc: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 arch/um/include/asm/tlb.h |   29 +++++++++++------------------
 1 file changed, 11 insertions(+), 18 deletions(-)

Index: linux-2.6/arch/um/include/asm/tlb.h
===================================================================
--- linux-2.6.orig/arch/um/include/asm/tlb.h
+++ linux-2.6/arch/um/include/asm/tlb.h
@@ -22,9 +22,6 @@ struct mmu_gather {
 	unsigned int		fullmm; /* non-zero means full mm flush */
 };
 
-/* Users of the generic TLB shootdown code must declare this storage space. */
-DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 static inline void __tlb_remove_tlb_entry(struct mmu_gather *tlb, pte_t *ptep,
 					  unsigned long address)
 {
@@ -47,27 +44,20 @@ static inline void init_tlb_gather(struc
 	}
 }
 
-/* tlb_gather_mmu
- *	Return a pointer to an initialized struct mmu_gather.
- */
-static inline struct mmu_gather *
-tlb_gather_mmu(struct mm_struct *mm, unsigned int full_mm_flush)
+static inline void
+tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned int full_mm_flush)
 {
-	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);
-
 	tlb->mm = mm;
 	tlb->fullmm = full_mm_flush;
 
 	init_tlb_gather(tlb);
-
-	return tlb;
 }
 
 extern void flush_tlb_mm_range(struct mm_struct *mm, unsigned long start,
 			       unsigned long end);
 
 static inline void
-tlb_flush_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
+tlb_flush_mmu(struct mmu_gather *tlb)
 {
 	if (!tlb->need_flush)
 		return;
@@ -83,12 +73,10 @@ tlb_flush_mmu(struct mmu_gather *tlb, un
 static inline void
 tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
 {
-	tlb_flush_mmu(tlb, start, end);
+	tlb_flush_mmu(tlb);
 
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
-
-	put_cpu_var(mmu_gathers);
 }
 
 /* tlb_remove_page
@@ -96,11 +84,16 @@ tlb_finish_mmu(struct mmu_gather *tlb, u
  *	while handling the additional races in SMP caused by other CPUs
  *	caching valid mappings in their TLBs.
  */
-static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
 {
 	tlb->need_flush = 1;
 	free_page_and_swap_cache(page);
-	return;
+	return 1; /* avoid calling tlb_flush_mmu */
+}
+
+static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	__tlb_remove_page(tlb, page);
 }
 
 /**

[PATCH 08/20] um: mmu_gather rework

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-um-preemptible_mmu_gather.patch --]
[-- Type: text/plain, Size: 2876 bytes --]

Fix up the um mmu_gather code to conform to the new API.

Cc: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 arch/um/include/asm/tlb.h |   29 +++++++++++------------------
 1 file changed, 11 insertions(+), 18 deletions(-)

Index: linux-2.6/arch/um/include/asm/tlb.h
===================================================================
--- linux-2.6.orig/arch/um/include/asm/tlb.h
+++ linux-2.6/arch/um/include/asm/tlb.h
@@ -22,9 +22,6 @@ struct mmu_gather {
 	unsigned int		fullmm; /* non-zero means full mm flush */
 };
 
-/* Users of the generic TLB shootdown code must declare this storage space. */
-DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 static inline void __tlb_remove_tlb_entry(struct mmu_gather *tlb, pte_t *ptep,
 					  unsigned long address)
 {
@@ -47,27 +44,20 @@ static inline void init_tlb_gather(struc
 	}
 }
 
-/* tlb_gather_mmu
- *	Return a pointer to an initialized struct mmu_gather.
- */
-static inline struct mmu_gather *
-tlb_gather_mmu(struct mm_struct *mm, unsigned int full_mm_flush)
+static inline void
+tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned int full_mm_flush)
 {
-	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);
-
 	tlb->mm = mm;
 	tlb->fullmm = full_mm_flush;
 
 	init_tlb_gather(tlb);
-
-	return tlb;
 }
 
 extern void flush_tlb_mm_range(struct mm_struct *mm, unsigned long start,
 			       unsigned long end);
 
 static inline void
-tlb_flush_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
+tlb_flush_mmu(struct mmu_gather *tlb)
 {
 	if (!tlb->need_flush)
 		return;
@@ -83,12 +73,10 @@ tlb_flush_mmu(struct mmu_gather *tlb, un
 static inline void
 tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
 {
-	tlb_flush_mmu(tlb, start, end);
+	tlb_flush_mmu(tlb);
 
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
-
-	put_cpu_var(mmu_gathers);
 }
 
 /* tlb_remove_page
@@ -96,11 +84,16 @@ tlb_finish_mmu(struct mmu_gather *tlb, u
  *	while handling the additional races in SMP caused by other CPUs
  *	caching valid mappings in their TLBs.
  */
-static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
 {
 	tlb->need_flush = 1;
 	free_page_and_swap_cache(page);
-	return;
+	return 1; /* avoid calling tlb_flush_mmu */
+}
+
+static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	__tlb_remove_page(tlb, page);
 }
 
 /**


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 09/20] mm: Now that all old mmu_gather code is gone, remove the storage

[Peter Zijlstra]

[-- Attachment #1: arch-remove-mmu_gather.patch --]
[-- Type: text/plain, Size: 9315 bytes --]

XXX fold all the mmu_gather rework patches into one for submission

Reported-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 arch/alpha/mm/init.c      |    2 --
 arch/arm/mm/mmu.c         |    2 --
 arch/avr32/mm/init.c      |    2 --
 arch/cris/mm/init.c       |    2 --
 arch/frv/mm/init.c        |    2 --
 arch/ia64/mm/init.c       |    2 --
 arch/m32r/mm/init.c       |    2 --
 arch/m68k/mm/init_mm.c    |    2 --
 arch/microblaze/mm/init.c |    2 --
 arch/mips/mm/init.c       |    2 --
 arch/mn10300/mm/init.c    |    2 --
 arch/parisc/mm/init.c     |    2 --
 arch/s390/mm/pgtable.c    |    1 -
 arch/score/mm/init.c      |    2 --
 arch/sh/mm/init.c         |    1 -
 arch/sparc/mm/init_32.c   |    2 --
 arch/tile/mm/init.c       |    2 --
 arch/um/kernel/smp.c      |    3 ---
 arch/unicore32/mm/mmu.c   |    2 --
 arch/x86/mm/init.c        |    2 --
 arch/xtensa/mm/mmu.c      |    2 --
 21 files changed, 41 deletions(-)

Index: linux-2.6/arch/alpha/mm/init.c
===================================================================
--- linux-2.6.orig/arch/alpha/mm/init.c
+++ linux-2.6/arch/alpha/mm/init.c
@@ -32,8 +32,6 @@
 #include <asm/console.h>
 #include <asm/tlb.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 extern void die_if_kernel(char *,struct pt_regs *,long);
 
 static struct pcb_struct original_pcb;
Index: linux-2.6/arch/arm/mm/mmu.c
===================================================================
--- linux-2.6.orig/arch/arm/mm/mmu.c
+++ linux-2.6/arch/arm/mm/mmu.c
@@ -31,8 +31,6 @@
 
 #include "mm.h"
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /*
  * empty_zero_page is a special page that is used for
  * zero-initialized data and COW.
Index: linux-2.6/arch/avr32/mm/init.c
===================================================================
--- linux-2.6.orig/arch/avr32/mm/init.c
+++ linux-2.6/arch/avr32/mm/init.c
@@ -25,8 +25,6 @@
 #include <asm/setup.h>
 #include <asm/sections.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 pgd_t swapper_pg_dir[PTRS_PER_PGD] __page_aligned_data;
 
 struct page *empty_zero_page;
Index: linux-2.6/arch/cris/mm/init.c
===================================================================
--- linux-2.6.orig/arch/cris/mm/init.c
+++ linux-2.6/arch/cris/mm/init.c
@@ -13,8 +13,6 @@
 #include <linux/bootmem.h>
 #include <asm/tlb.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 unsigned long empty_zero_page;
 
 extern char _stext, _edata, _etext; /* From linkerscript */
Index: linux-2.6/arch/frv/mm/init.c
===================================================================
--- linux-2.6.orig/arch/frv/mm/init.c
+++ linux-2.6/arch/frv/mm/init.c
@@ -41,8 +41,6 @@
 
 #undef DEBUG
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /*
  * BAD_PAGE is the page that is used for page faults when linux
  * is out-of-memory. Older versions of linux just did a
Index: linux-2.6/arch/ia64/mm/init.c
===================================================================
--- linux-2.6.orig/arch/ia64/mm/init.c
+++ linux-2.6/arch/ia64/mm/init.c
@@ -36,8 +36,6 @@
 #include <asm/mca.h>
 #include <asm/paravirt.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 extern void ia64_tlb_init (void);
 
 unsigned long MAX_DMA_ADDRESS = PAGE_OFFSET + 0x100000000UL;
Index: linux-2.6/arch/m32r/mm/init.c
===================================================================
--- linux-2.6.orig/arch/m32r/mm/init.c
+++ linux-2.6/arch/m32r/mm/init.c
@@ -35,8 +35,6 @@ extern char __init_begin, __init_end;
 
 pgd_t swapper_pg_dir[1024];
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /*
  * Cache of MMU context last used.
  */
Index: linux-2.6/arch/microblaze/mm/init.c
===================================================================
--- linux-2.6.orig/arch/microblaze/mm/init.c
+++ linux-2.6/arch/microblaze/mm/init.c
@@ -32,8 +32,6 @@ unsigned int __page_offset;
 EXPORT_SYMBOL(__page_offset);
 
 #else
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 static int init_bootmem_done;
 #endif /* CONFIG_MMU */
 
Index: linux-2.6/arch/mips/mm/init.c
===================================================================
--- linux-2.6.orig/arch/mips/mm/init.c
+++ linux-2.6/arch/mips/mm/init.c
@@ -64,8 +64,6 @@
 
 #endif /* CONFIG_MIPS_MT_SMTC */
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /*
  * We have up to 8 empty zeroed pages so we can map one of the right colour
  * when needed.  This is necessary only on R4000 / R4400 SC and MC versions
Index: linux-2.6/arch/mn10300/mm/init.c
===================================================================
--- linux-2.6.orig/arch/mn10300/mm/init.c
+++ linux-2.6/arch/mn10300/mm/init.c
@@ -37,8 +37,6 @@
 #include <asm/tlb.h>
 #include <asm/sections.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 unsigned long highstart_pfn, highend_pfn;
 
 #ifdef CONFIG_MN10300_HAS_ATOMIC_OPS_UNIT
Index: linux-2.6/arch/parisc/mm/init.c
===================================================================
--- linux-2.6.orig/arch/parisc/mm/init.c
+++ linux-2.6/arch/parisc/mm/init.c
@@ -31,8 +31,6 @@
 #include <asm/mmzone.h>
 #include <asm/sections.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 extern int  data_start;
 
 #ifdef CONFIG_DISCONTIGMEM
Index: linux-2.6/arch/s390/mm/pgtable.c
===================================================================
--- linux-2.6.orig/arch/s390/mm/pgtable.c
+++ linux-2.6/arch/s390/mm/pgtable.c
@@ -36,7 +36,6 @@ struct rcu_table_freelist {
 	((PAGE_SIZE - sizeof(struct rcu_table_freelist)) \
 	  / sizeof(unsigned long))
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
 static DEFINE_PER_CPU(struct rcu_table_freelist *, rcu_table_freelist);
 
 static void __page_table_free(struct mm_struct *mm, unsigned long *table);
Index: linux-2.6/arch/score/mm/init.c
===================================================================
--- linux-2.6.orig/arch/score/mm/init.c
+++ linux-2.6/arch/score/mm/init.c
@@ -38,8 +38,6 @@
 #include <asm/sections.h>
 #include <asm/tlb.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 unsigned long empty_zero_page;
 EXPORT_SYMBOL_GPL(empty_zero_page);
 
Index: linux-2.6/arch/sh/mm/init.c
===================================================================
--- linux-2.6.orig/arch/sh/mm/init.c
+++ linux-2.6/arch/sh/mm/init.c
@@ -28,7 +28,6 @@
 #include <asm/cache.h>
 #include <asm/sizes.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
 pgd_t swapper_pg_dir[PTRS_PER_PGD];
 
 void __init generic_mem_init(void)
Index: linux-2.6/arch/sparc/mm/init_32.c
===================================================================
--- linux-2.6.orig/arch/sparc/mm/init_32.c
+++ linux-2.6/arch/sparc/mm/init_32.c
@@ -37,8 +37,6 @@
 #include <asm/prom.h>
 #include <asm/leon.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 unsigned long *sparc_valid_addr_bitmap;
 EXPORT_SYMBOL(sparc_valid_addr_bitmap);
 
Index: linux-2.6/arch/tile/mm/init.c
===================================================================
--- linux-2.6.orig/arch/tile/mm/init.c
+++ linux-2.6/arch/tile/mm/init.c
@@ -60,8 +60,6 @@ unsigned long VMALLOC_RESERVE = CONFIG_V
 EXPORT_SYMBOL(VMALLOC_RESERVE);
 #endif
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /* Create an L2 page table */
 static pte_t * __init alloc_pte(void)
 {
Index: linux-2.6/arch/um/kernel/smp.c
===================================================================
--- linux-2.6.orig/arch/um/kernel/smp.c
+++ linux-2.6/arch/um/kernel/smp.c
@@ -7,9 +7,6 @@
 #include "asm/pgalloc.h"
 #include "asm/tlb.h"
 
-/* For some reason, mmu_gathers are referenced when CONFIG_SMP is off. */
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 #ifdef CONFIG_SMP
 
 #include "linux/sched.h"
Index: linux-2.6/arch/x86/mm/init.c
===================================================================
--- linux-2.6.orig/arch/x86/mm/init.c
+++ linux-2.6/arch/x86/mm/init.c
@@ -16,8 +16,6 @@
 #include <asm/tlb.h>
 #include <asm/proto.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 unsigned long __initdata pgt_buf_start;
 unsigned long __meminitdata pgt_buf_end;
 unsigned long __meminitdata pgt_buf_top;
Index: linux-2.6/arch/xtensa/mm/mmu.c
===================================================================
--- linux-2.6.orig/arch/xtensa/mm/mmu.c
+++ linux-2.6/arch/xtensa/mm/mmu.c
@@ -14,8 +14,6 @@
 #include <asm/mmu_context.h>
 #include <asm/page.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 void __init paging_init(void)
 {
 	memset(swapper_pg_dir, 0, PAGE_SIZE);
Index: linux-2.6/arch/unicore32/mm/mmu.c
===================================================================
--- linux-2.6.orig/arch/unicore32/mm/mmu.c
+++ linux-2.6/arch/unicore32/mm/mmu.c
@@ -30,8 +30,6 @@
 
 #include "mm.h"
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /*
  * empty_zero_page is a special page that is used for
  * zero-initialized data and COW.
Index: linux-2.6/arch/m68k/mm/init_mm.c
===================================================================
--- linux-2.6.orig/arch/m68k/mm/init_mm.c
+++ linux-2.6/arch/m68k/mm/init_mm.c
@@ -32,8 +32,6 @@
 #include <asm/sections.h>
 #include <asm/tlb.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 pg_data_t pg_data_map[MAX_NUMNODES];
 EXPORT_SYMBOL(pg_data_map);
 

[PATCH 09/20] mm: Now that all old mmu_gather code is gone, remove the storage

[Peter Zijlstra]

[-- Attachment #1: arch-remove-mmu_gather.patch --]
[-- Type: text/plain, Size: 9618 bytes --]

XXX fold all the mmu_gather rework patches into one for submission

Reported-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 arch/alpha/mm/init.c      |    2 --
 arch/arm/mm/mmu.c         |    2 --
 arch/avr32/mm/init.c      |    2 --
 arch/cris/mm/init.c       |    2 --
 arch/frv/mm/init.c        |    2 --
 arch/ia64/mm/init.c       |    2 --
 arch/m32r/mm/init.c       |    2 --
 arch/m68k/mm/init_mm.c    |    2 --
 arch/microblaze/mm/init.c |    2 --
 arch/mips/mm/init.c       |    2 --
 arch/mn10300/mm/init.c    |    2 --
 arch/parisc/mm/init.c     |    2 --
 arch/s390/mm/pgtable.c    |    1 -
 arch/score/mm/init.c      |    2 --
 arch/sh/mm/init.c         |    1 -
 arch/sparc/mm/init_32.c   |    2 --
 arch/tile/mm/init.c       |    2 --
 arch/um/kernel/smp.c      |    3 ---
 arch/unicore32/mm/mmu.c   |    2 --
 arch/x86/mm/init.c        |    2 --
 arch/xtensa/mm/mmu.c      |    2 --
 21 files changed, 41 deletions(-)

Index: linux-2.6/arch/alpha/mm/init.c
===================================================================
--- linux-2.6.orig/arch/alpha/mm/init.c
+++ linux-2.6/arch/alpha/mm/init.c
@@ -32,8 +32,6 @@
 #include <asm/console.h>
 #include <asm/tlb.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 extern void die_if_kernel(char *,struct pt_regs *,long);
 
 static struct pcb_struct original_pcb;
Index: linux-2.6/arch/arm/mm/mmu.c
===================================================================
--- linux-2.6.orig/arch/arm/mm/mmu.c
+++ linux-2.6/arch/arm/mm/mmu.c
@@ -31,8 +31,6 @@
 
 #include "mm.h"
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /*
  * empty_zero_page is a special page that is used for
  * zero-initialized data and COW.
Index: linux-2.6/arch/avr32/mm/init.c
===================================================================
--- linux-2.6.orig/arch/avr32/mm/init.c
+++ linux-2.6/arch/avr32/mm/init.c
@@ -25,8 +25,6 @@
 #include <asm/setup.h>
 #include <asm/sections.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 pgd_t swapper_pg_dir[PTRS_PER_PGD] __page_aligned_data;
 
 struct page *empty_zero_page;
Index: linux-2.6/arch/cris/mm/init.c
===================================================================
--- linux-2.6.orig/arch/cris/mm/init.c
+++ linux-2.6/arch/cris/mm/init.c
@@ -13,8 +13,6 @@
 #include <linux/bootmem.h>
 #include <asm/tlb.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 unsigned long empty_zero_page;
 
 extern char _stext, _edata, _etext; /* From linkerscript */
Index: linux-2.6/arch/frv/mm/init.c
===================================================================
--- linux-2.6.orig/arch/frv/mm/init.c
+++ linux-2.6/arch/frv/mm/init.c
@@ -41,8 +41,6 @@
 
 #undef DEBUG
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /*
  * BAD_PAGE is the page that is used for page faults when linux
  * is out-of-memory. Older versions of linux just did a
Index: linux-2.6/arch/ia64/mm/init.c
===================================================================
--- linux-2.6.orig/arch/ia64/mm/init.c
+++ linux-2.6/arch/ia64/mm/init.c
@@ -36,8 +36,6 @@
 #include <asm/mca.h>
 #include <asm/paravirt.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 extern void ia64_tlb_init (void);
 
 unsigned long MAX_DMA_ADDRESS = PAGE_OFFSET + 0x100000000UL;
Index: linux-2.6/arch/m32r/mm/init.c
===================================================================
--- linux-2.6.orig/arch/m32r/mm/init.c
+++ linux-2.6/arch/m32r/mm/init.c
@@ -35,8 +35,6 @@ extern char __init_begin, __init_end;
 
 pgd_t swapper_pg_dir[1024];
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /*
  * Cache of MMU context last used.
  */
Index: linux-2.6/arch/microblaze/mm/init.c
===================================================================
--- linux-2.6.orig/arch/microblaze/mm/init.c
+++ linux-2.6/arch/microblaze/mm/init.c
@@ -32,8 +32,6 @@ unsigned int __page_offset;
 EXPORT_SYMBOL(__page_offset);
 
 #else
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 static int init_bootmem_done;
 #endif /* CONFIG_MMU */
 
Index: linux-2.6/arch/mips/mm/init.c
===================================================================
--- linux-2.6.orig/arch/mips/mm/init.c
+++ linux-2.6/arch/mips/mm/init.c
@@ -64,8 +64,6 @@
 
 #endif /* CONFIG_MIPS_MT_SMTC */
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /*
  * We have up to 8 empty zeroed pages so we can map one of the right colour
  * when needed.  This is necessary only on R4000 / R4400 SC and MC versions
Index: linux-2.6/arch/mn10300/mm/init.c
===================================================================
--- linux-2.6.orig/arch/mn10300/mm/init.c
+++ linux-2.6/arch/mn10300/mm/init.c
@@ -37,8 +37,6 @@
 #include <asm/tlb.h>
 #include <asm/sections.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 unsigned long highstart_pfn, highend_pfn;
 
 #ifdef CONFIG_MN10300_HAS_ATOMIC_OPS_UNIT
Index: linux-2.6/arch/parisc/mm/init.c
===================================================================
--- linux-2.6.orig/arch/parisc/mm/init.c
+++ linux-2.6/arch/parisc/mm/init.c
@@ -31,8 +31,6 @@
 #include <asm/mmzone.h>
 #include <asm/sections.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 extern int  data_start;
 
 #ifdef CONFIG_DISCONTIGMEM
Index: linux-2.6/arch/s390/mm/pgtable.c
===================================================================
--- linux-2.6.orig/arch/s390/mm/pgtable.c
+++ linux-2.6/arch/s390/mm/pgtable.c
@@ -36,7 +36,6 @@ struct rcu_table_freelist {
 	((PAGE_SIZE - sizeof(struct rcu_table_freelist)) \
 	  / sizeof(unsigned long))
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
 static DEFINE_PER_CPU(struct rcu_table_freelist *, rcu_table_freelist);
 
 static void __page_table_free(struct mm_struct *mm, unsigned long *table);
Index: linux-2.6/arch/score/mm/init.c
===================================================================
--- linux-2.6.orig/arch/score/mm/init.c
+++ linux-2.6/arch/score/mm/init.c
@@ -38,8 +38,6 @@
 #include <asm/sections.h>
 #include <asm/tlb.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 unsigned long empty_zero_page;
 EXPORT_SYMBOL_GPL(empty_zero_page);
 
Index: linux-2.6/arch/sh/mm/init.c
===================================================================
--- linux-2.6.orig/arch/sh/mm/init.c
+++ linux-2.6/arch/sh/mm/init.c
@@ -28,7 +28,6 @@
 #include <asm/cache.h>
 #include <asm/sizes.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
 pgd_t swapper_pg_dir[PTRS_PER_PGD];
 
 void __init generic_mem_init(void)
Index: linux-2.6/arch/sparc/mm/init_32.c
===================================================================
--- linux-2.6.orig/arch/sparc/mm/init_32.c
+++ linux-2.6/arch/sparc/mm/init_32.c
@@ -37,8 +37,6 @@
 #include <asm/prom.h>
 #include <asm/leon.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 unsigned long *sparc_valid_addr_bitmap;
 EXPORT_SYMBOL(sparc_valid_addr_bitmap);
 
Index: linux-2.6/arch/tile/mm/init.c
===================================================================
--- linux-2.6.orig/arch/tile/mm/init.c
+++ linux-2.6/arch/tile/mm/init.c
@@ -60,8 +60,6 @@ unsigned long VMALLOC_RESERVE = CONFIG_V
 EXPORT_SYMBOL(VMALLOC_RESERVE);
 #endif
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /* Create an L2 page table */
 static pte_t * __init alloc_pte(void)
 {
Index: linux-2.6/arch/um/kernel/smp.c
===================================================================
--- linux-2.6.orig/arch/um/kernel/smp.c
+++ linux-2.6/arch/um/kernel/smp.c
@@ -7,9 +7,6 @@
 #include "asm/pgalloc.h"
 #include "asm/tlb.h"
 
-/* For some reason, mmu_gathers are referenced when CONFIG_SMP is off. */
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 #ifdef CONFIG_SMP
 
 #include "linux/sched.h"
Index: linux-2.6/arch/x86/mm/init.c
===================================================================
--- linux-2.6.orig/arch/x86/mm/init.c
+++ linux-2.6/arch/x86/mm/init.c
@@ -16,8 +16,6 @@
 #include <asm/tlb.h>
 #include <asm/proto.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 unsigned long __initdata pgt_buf_start;
 unsigned long __meminitdata pgt_buf_end;
 unsigned long __meminitdata pgt_buf_top;
Index: linux-2.6/arch/xtensa/mm/mmu.c
===================================================================
--- linux-2.6.orig/arch/xtensa/mm/mmu.c
+++ linux-2.6/arch/xtensa/mm/mmu.c
@@ -14,8 +14,6 @@
 #include <asm/mmu_context.h>
 #include <asm/page.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 void __init paging_init(void)
 {
 	memset(swapper_pg_dir, 0, PAGE_SIZE);
Index: linux-2.6/arch/unicore32/mm/mmu.c
===================================================================
--- linux-2.6.orig/arch/unicore32/mm/mmu.c
+++ linux-2.6/arch/unicore32/mm/mmu.c
@@ -30,8 +30,6 @@
 
 #include "mm.h"
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /*
  * empty_zero_page is a special page that is used for
  * zero-initialized data and COW.
Index: linux-2.6/arch/m68k/mm/init_mm.c
===================================================================
--- linux-2.6.orig/arch/m68k/mm/init_mm.c
+++ linux-2.6/arch/m68k/mm/init_mm.c
@@ -32,8 +32,6 @@
 #include <asm/sections.h>
 #include <asm/tlb.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 pg_data_t pg_data_map[MAX_NUMNODES];
 EXPORT_SYMBOL(pg_data_map);
 


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 09/20] mm: Now that all old mmu_gather code is gone, remove the storage

[Peter Zijlstra]

[-- Attachment #1: arch-remove-mmu_gather.patch --]
[-- Type: text/plain, Size: 9618 bytes --]

XXX fold all the mmu_gather rework patches into one for submission

Reported-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 arch/alpha/mm/init.c      |    2 --
 arch/arm/mm/mmu.c         |    2 --
 arch/avr32/mm/init.c      |    2 --
 arch/cris/mm/init.c       |    2 --
 arch/frv/mm/init.c        |    2 --
 arch/ia64/mm/init.c       |    2 --
 arch/m32r/mm/init.c       |    2 --
 arch/m68k/mm/init_mm.c    |    2 --
 arch/microblaze/mm/init.c |    2 --
 arch/mips/mm/init.c       |    2 --
 arch/mn10300/mm/init.c    |    2 --
 arch/parisc/mm/init.c     |    2 --
 arch/s390/mm/pgtable.c    |    1 -
 arch/score/mm/init.c      |    2 --
 arch/sh/mm/init.c         |    1 -
 arch/sparc/mm/init_32.c   |    2 --
 arch/tile/mm/init.c       |    2 --
 arch/um/kernel/smp.c      |    3 ---
 arch/unicore32/mm/mmu.c   |    2 --
 arch/x86/mm/init.c        |    2 --
 arch/xtensa/mm/mmu.c      |    2 --
 21 files changed, 41 deletions(-)

Index: linux-2.6/arch/alpha/mm/init.c
===================================================================
--- linux-2.6.orig/arch/alpha/mm/init.c
+++ linux-2.6/arch/alpha/mm/init.c
@@ -32,8 +32,6 @@
 #include <asm/console.h>
 #include <asm/tlb.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 extern void die_if_kernel(char *,struct pt_regs *,long);
 
 static struct pcb_struct original_pcb;
Index: linux-2.6/arch/arm/mm/mmu.c
===================================================================
--- linux-2.6.orig/arch/arm/mm/mmu.c
+++ linux-2.6/arch/arm/mm/mmu.c
@@ -31,8 +31,6 @@
 
 #include "mm.h"
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /*
  * empty_zero_page is a special page that is used for
  * zero-initialized data and COW.
Index: linux-2.6/arch/avr32/mm/init.c
===================================================================
--- linux-2.6.orig/arch/avr32/mm/init.c
+++ linux-2.6/arch/avr32/mm/init.c
@@ -25,8 +25,6 @@
 #include <asm/setup.h>
 #include <asm/sections.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 pgd_t swapper_pg_dir[PTRS_PER_PGD] __page_aligned_data;
 
 struct page *empty_zero_page;
Index: linux-2.6/arch/cris/mm/init.c
===================================================================
--- linux-2.6.orig/arch/cris/mm/init.c
+++ linux-2.6/arch/cris/mm/init.c
@@ -13,8 +13,6 @@
 #include <linux/bootmem.h>
 #include <asm/tlb.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 unsigned long empty_zero_page;
 
 extern char _stext, _edata, _etext; /* From linkerscript */
Index: linux-2.6/arch/frv/mm/init.c
===================================================================
--- linux-2.6.orig/arch/frv/mm/init.c
+++ linux-2.6/arch/frv/mm/init.c
@@ -41,8 +41,6 @@
 
 #undef DEBUG
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /*
  * BAD_PAGE is the page that is used for page faults when linux
  * is out-of-memory. Older versions of linux just did a
Index: linux-2.6/arch/ia64/mm/init.c
===================================================================
--- linux-2.6.orig/arch/ia64/mm/init.c
+++ linux-2.6/arch/ia64/mm/init.c
@@ -36,8 +36,6 @@
 #include <asm/mca.h>
 #include <asm/paravirt.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 extern void ia64_tlb_init (void);
 
 unsigned long MAX_DMA_ADDRESS = PAGE_OFFSET + 0x100000000UL;
Index: linux-2.6/arch/m32r/mm/init.c
===================================================================
--- linux-2.6.orig/arch/m32r/mm/init.c
+++ linux-2.6/arch/m32r/mm/init.c
@@ -35,8 +35,6 @@ extern char __init_begin, __init_end;
 
 pgd_t swapper_pg_dir[1024];
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /*
  * Cache of MMU context last used.
  */
Index: linux-2.6/arch/microblaze/mm/init.c
===================================================================
--- linux-2.6.orig/arch/microblaze/mm/init.c
+++ linux-2.6/arch/microblaze/mm/init.c
@@ -32,8 +32,6 @@ unsigned int __page_offset;
 EXPORT_SYMBOL(__page_offset);
 
 #else
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 static int init_bootmem_done;
 #endif /* CONFIG_MMU */
 
Index: linux-2.6/arch/mips/mm/init.c
===================================================================
--- linux-2.6.orig/arch/mips/mm/init.c
+++ linux-2.6/arch/mips/mm/init.c
@@ -64,8 +64,6 @@
 
 #endif /* CONFIG_MIPS_MT_SMTC */
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /*
  * We have up to 8 empty zeroed pages so we can map one of the right colour
  * when needed.  This is necessary only on R4000 / R4400 SC and MC versions
Index: linux-2.6/arch/mn10300/mm/init.c
===================================================================
--- linux-2.6.orig/arch/mn10300/mm/init.c
+++ linux-2.6/arch/mn10300/mm/init.c
@@ -37,8 +37,6 @@
 #include <asm/tlb.h>
 #include <asm/sections.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 unsigned long highstart_pfn, highend_pfn;
 
 #ifdef CONFIG_MN10300_HAS_ATOMIC_OPS_UNIT
Index: linux-2.6/arch/parisc/mm/init.c
===================================================================
--- linux-2.6.orig/arch/parisc/mm/init.c
+++ linux-2.6/arch/parisc/mm/init.c
@@ -31,8 +31,6 @@
 #include <asm/mmzone.h>
 #include <asm/sections.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 extern int  data_start;
 
 #ifdef CONFIG_DISCONTIGMEM
Index: linux-2.6/arch/s390/mm/pgtable.c
===================================================================
--- linux-2.6.orig/arch/s390/mm/pgtable.c
+++ linux-2.6/arch/s390/mm/pgtable.c
@@ -36,7 +36,6 @@ struct rcu_table_freelist {
 	((PAGE_SIZE - sizeof(struct rcu_table_freelist)) \
 	  / sizeof(unsigned long))
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
 static DEFINE_PER_CPU(struct rcu_table_freelist *, rcu_table_freelist);
 
 static void __page_table_free(struct mm_struct *mm, unsigned long *table);
Index: linux-2.6/arch/score/mm/init.c
===================================================================
--- linux-2.6.orig/arch/score/mm/init.c
+++ linux-2.6/arch/score/mm/init.c
@@ -38,8 +38,6 @@
 #include <asm/sections.h>
 #include <asm/tlb.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 unsigned long empty_zero_page;
 EXPORT_SYMBOL_GPL(empty_zero_page);
 
Index: linux-2.6/arch/sh/mm/init.c
===================================================================
--- linux-2.6.orig/arch/sh/mm/init.c
+++ linux-2.6/arch/sh/mm/init.c
@@ -28,7 +28,6 @@
 #include <asm/cache.h>
 #include <asm/sizes.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
 pgd_t swapper_pg_dir[PTRS_PER_PGD];
 
 void __init generic_mem_init(void)
Index: linux-2.6/arch/sparc/mm/init_32.c
===================================================================
--- linux-2.6.orig/arch/sparc/mm/init_32.c
+++ linux-2.6/arch/sparc/mm/init_32.c
@@ -37,8 +37,6 @@
 #include <asm/prom.h>
 #include <asm/leon.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 unsigned long *sparc_valid_addr_bitmap;
 EXPORT_SYMBOL(sparc_valid_addr_bitmap);
 
Index: linux-2.6/arch/tile/mm/init.c
===================================================================
--- linux-2.6.orig/arch/tile/mm/init.c
+++ linux-2.6/arch/tile/mm/init.c
@@ -60,8 +60,6 @@ unsigned long VMALLOC_RESERVE = CONFIG_V
 EXPORT_SYMBOL(VMALLOC_RESERVE);
 #endif
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /* Create an L2 page table */
 static pte_t * __init alloc_pte(void)
 {
Index: linux-2.6/arch/um/kernel/smp.c
===================================================================
--- linux-2.6.orig/arch/um/kernel/smp.c
+++ linux-2.6/arch/um/kernel/smp.c
@@ -7,9 +7,6 @@
 #include "asm/pgalloc.h"
 #include "asm/tlb.h"
 
-/* For some reason, mmu_gathers are referenced when CONFIG_SMP is off. */
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 #ifdef CONFIG_SMP
 
 #include "linux/sched.h"
Index: linux-2.6/arch/x86/mm/init.c
===================================================================
--- linux-2.6.orig/arch/x86/mm/init.c
+++ linux-2.6/arch/x86/mm/init.c
@@ -16,8 +16,6 @@
 #include <asm/tlb.h>
 #include <asm/proto.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 unsigned long __initdata pgt_buf_start;
 unsigned long __meminitdata pgt_buf_end;
 unsigned long __meminitdata pgt_buf_top;
Index: linux-2.6/arch/xtensa/mm/mmu.c
===================================================================
--- linux-2.6.orig/arch/xtensa/mm/mmu.c
+++ linux-2.6/arch/xtensa/mm/mmu.c
@@ -14,8 +14,6 @@
 #include <asm/mmu_context.h>
 #include <asm/page.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 void __init paging_init(void)
 {
 	memset(swapper_pg_dir, 0, PAGE_SIZE);
Index: linux-2.6/arch/unicore32/mm/mmu.c
===================================================================
--- linux-2.6.orig/arch/unicore32/mm/mmu.c
+++ linux-2.6/arch/unicore32/mm/mmu.c
@@ -30,8 +30,6 @@
 
 #include "mm.h"
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /*
  * empty_zero_page is a special page that is used for
  * zero-initialized data and COW.
Index: linux-2.6/arch/m68k/mm/init_mm.c
===================================================================
--- linux-2.6.orig/arch/m68k/mm/init_mm.c
+++ linux-2.6/arch/m68k/mm/init_mm.c
@@ -32,8 +32,6 @@
 #include <asm/sections.h>
 #include <asm/tlb.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 pg_data_t pg_data_map[MAX_NUMNODES];
 EXPORT_SYMBOL(pg_data_map);
 


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 10/20] mm, powerpc: Move the RCU page-table freeing into generic code

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-mm_powerpc-move_the_rcu_page-table_freeing_into.patch --]
[-- Type: text/plain, Size: 12603 bytes --]

In case other architectures require RCU freed page-tables to implement
gup_fast() and software filled hashes and similar things, provide the
means to do so by moving the logic into generic code.

Requested-by: David Miller <davem@davemloft.net>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 arch/Kconfig                       |    3 +
 arch/powerpc/Kconfig               |    1 
 arch/powerpc/include/asm/pgalloc.h |   21 ++++++-
 arch/powerpc/include/asm/tlb.h     |   10 ---
 arch/powerpc/mm/pgtable.c          |   98 -------------------------------------
 arch/powerpc/mm/tlb_hash32.c       |    3 -
 arch/powerpc/mm/tlb_hash64.c       |    3 -
 arch/powerpc/mm/tlb_nohash.c       |    3 -
 include/asm-generic/tlb.h          |   56 +++++++++++++++++++--
 mm/memory.c                        |   77 +++++++++++++++++++++++++++++
 10 files changed, 150 insertions(+), 125 deletions(-)

Index: linux-2.6/arch/powerpc/include/asm/pgalloc.h
===================================================================
--- linux-2.6.orig/arch/powerpc/include/asm/pgalloc.h
+++ linux-2.6/arch/powerpc/include/asm/pgalloc.h
@@ -31,14 +31,29 @@ static inline void pte_free(struct mm_st
 #endif
 
 #ifdef CONFIG_SMP
-extern void pgtable_free_tlb(struct mmu_gather *tlb, void *table, unsigned shift);
-extern void pte_free_finish(struct mmu_gather *tlb);
+struct mmu_gather;
+extern void tlb_remove_table(struct mmu_gather *, void *);
+
+static inline void pgtable_free_tlb(struct mmu_gather *tlb, void *table, int shift)
+{
+	unsigned long pgf = (unsigned long)table;
+	BUG_ON(shift > MAX_PGTABLE_INDEX_SIZE);
+	pgf |= shift;
+	tlb_remove_table(tlb, (void *)pgf);
+}
+
+static inline void __tlb_remove_table(void *_table)
+{
+	void *table = (void *)((unsigned long)_table & ~MAX_PGTABLE_INDEX_SIZE);
+	unsigned shift = (unsigned long)_table & MAX_PGTABLE_INDEX_SIZE;
+
+	pgtable_free(table, shift);
+}
 #else /* CONFIG_SMP */
 static inline void pgtable_free_tlb(struct mmu_gather *tlb, void *table, unsigned shift)
 {
 	pgtable_free(table, shift);
 }
-static inline void pte_free_finish(struct mmu_gather *tlb) { }
 #endif /* !CONFIG_SMP */
 
 static inline void __pte_free_tlb(struct mmu_gather *tlb, struct page *ptepage,
Index: linux-2.6/arch/powerpc/include/asm/tlb.h
===================================================================
--- linux-2.6.orig/arch/powerpc/include/asm/tlb.h
+++ linux-2.6/arch/powerpc/include/asm/tlb.h
@@ -28,16 +28,6 @@
 #define tlb_start_vma(tlb, vma)	do { } while (0)
 #define tlb_end_vma(tlb, vma)	do { } while (0)
 
-#define HAVE_ARCH_MMU_GATHER 1
-
-struct pte_freelist_batch;
-
-struct arch_mmu_gather {
-	struct pte_freelist_batch *batch;
-};
-
-#define ARCH_MMU_GATHER_INIT (struct arch_mmu_gather){ .batch = NULL, }
-
 extern void tlb_flush(struct mmu_gather *tlb);
 
 /* Get the generic bits... */
Index: linux-2.6/arch/powerpc/mm/pgtable.c
===================================================================
--- linux-2.6.orig/arch/powerpc/mm/pgtable.c
+++ linux-2.6/arch/powerpc/mm/pgtable.c
@@ -33,104 +33,6 @@
 
 #include "mmu_decl.h"
 
-#ifdef CONFIG_SMP
-
-/*
- * Handle batching of page table freeing on SMP. Page tables are
- * queued up and send to be freed later by RCU in order to avoid
- * freeing a page table page that is being walked without locks
- */
-
-static unsigned long pte_freelist_forced_free;
-
-struct pte_freelist_batch
-{
-	struct rcu_head	rcu;
-	unsigned int	index;
-	unsigned long	tables[0];
-};
-
-#define PTE_FREELIST_SIZE \
-	((PAGE_SIZE - sizeof(struct pte_freelist_batch)) \
-	  / sizeof(unsigned long))
-
-static void pte_free_smp_sync(void *arg)
-{
-	/* Do nothing, just ensure we sync with all CPUs */
-}
-
-/* This is only called when we are critically out of memory
- * (and fail to get a page in pte_free_tlb).
- */
-static void pgtable_free_now(void *table, unsigned shift)
-{
-	pte_freelist_forced_free++;
-
-	smp_call_function(pte_free_smp_sync, NULL, 1);
-
-	pgtable_free(table, shift);
-}
-
-static void pte_free_rcu_callback(struct rcu_head *head)
-{
-	struct pte_freelist_batch *batch =
-		container_of(head, struct pte_freelist_batch, rcu);
-	unsigned int i;
-
-	for (i = 0; i < batch->index; i++) {
-		void *table = (void *)(batch->tables[i] & ~MAX_PGTABLE_INDEX_SIZE);
-		unsigned shift = batch->tables[i] & MAX_PGTABLE_INDEX_SIZE;
-
-		pgtable_free(table, shift);
-	}
-
-	free_page((unsigned long)batch);
-}
-
-static void pte_free_submit(struct pte_freelist_batch *batch)
-{
-	call_rcu_sched(&batch->rcu, pte_free_rcu_callback);
-}
-
-void pgtable_free_tlb(struct mmu_gather *tlb, void *table, unsigned shift)
-{
-	struct pte_freelist_batch **batchp = &tlb->arch.batch;
-	unsigned long pgf;
-
-	if (atomic_read(&tlb->mm->mm_users) < 2) {
-		pgtable_free(table, shift);
-		return;
-	}
-
-	if (*batchp == NULL) {
-		*batchp = (struct pte_freelist_batch *)__get_free_page(GFP_ATOMIC);
-		if (*batchp == NULL) {
-			pgtable_free_now(table, shift);
-			return;
-		}
-		(*batchp)->index = 0;
-	}
-	BUG_ON(shift > MAX_PGTABLE_INDEX_SIZE);
-	pgf = (unsigned long)table | shift;
-	(*batchp)->tables[(*batchp)->index++] = pgf;
-	if ((*batchp)->index == PTE_FREELIST_SIZE) {
-		pte_free_submit(*batchp);
-		*batchp = NULL;
-	}
-}
-
-void pte_free_finish(struct mmu_gather *tlb)
-{
-	struct pte_freelist_batch **batchp = &tlb->arch.batch;
-
-	if (*batchp == NULL)
-		return;
-	pte_free_submit(*batchp);
-	*batchp = NULL;
-}
-
-#endif /* CONFIG_SMP */
-
 static inline int is_exec_fault(void)
 {
 	return current->thread.regs && TRAP(current->thread.regs) == 0x400;
Index: linux-2.6/arch/powerpc/mm/tlb_hash32.c
===================================================================
--- linux-2.6.orig/arch/powerpc/mm/tlb_hash32.c
+++ linux-2.6/arch/powerpc/mm/tlb_hash32.c
@@ -71,9 +71,6 @@ void tlb_flush(struct mmu_gather *tlb)
 		 */
 		_tlbia();
 	}
-
-	/* Push out batch of freed page tables */
-	pte_free_finish(tlb);
 }
 
 /*
Index: linux-2.6/arch/powerpc/mm/tlb_hash64.c
===================================================================
--- linux-2.6.orig/arch/powerpc/mm/tlb_hash64.c
+++ linux-2.6/arch/powerpc/mm/tlb_hash64.c
@@ -165,9 +165,6 @@ void tlb_flush(struct mmu_gather *tlb)
 		__flush_tlb_pending(tlbbatch);
 
 	put_cpu_var(ppc64_tlb_batch);
-
-	/* Push out batch of freed page tables */
-	pte_free_finish(tlb);
 }
 
 /**
Index: linux-2.6/arch/powerpc/mm/tlb_nohash.c
===================================================================
--- linux-2.6.orig/arch/powerpc/mm/tlb_nohash.c
+++ linux-2.6/arch/powerpc/mm/tlb_nohash.c
@@ -299,9 +299,6 @@ EXPORT_SYMBOL(flush_tlb_range);
 void tlb_flush(struct mmu_gather *tlb)
 {
 	flush_tlb_mm(tlb->mm);
-
-	/* Push out batch of freed page tables */
-	pte_free_finish(tlb);
 }
 
 /*
Index: linux-2.6/include/asm-generic/tlb.h
===================================================================
--- linux-2.6.orig/include/asm-generic/tlb.h
+++ linux-2.6/include/asm-generic/tlb.h
@@ -29,6 +29,49 @@
   #define tlb_fast_mode(tlb) 1
 #endif
 
+#ifdef CONFIG_HAVE_RCU_TABLE_FREE
+/*
+ * Semi RCU freeing of the page directories.
+ *
+ * This is needed by some architectures to implement software pagetable walkers.
+ *
+ * gup_fast() and other software pagetable walkers do a lockless page-table
+ * walk and therefore needs some synchronization with the freeing of the page
+ * directories. The chosen means to accomplish that is by disabling IRQs over
+ * the walk.
+ *
+ * Architectures that use IPIs to flush TLBs will then automagically DTRT,
+ * since we unlink the page, flush TLBs, free the page. Since the disabling of
+ * IRQs delays the completion of the TLB flush we can never observe an already
+ * freed page.
+ *
+ * Architectures that do not have this (PPC) need to delay the freeing by some
+ * other means, this is that means.
+ *
+ * What we do is batch the freed directory pages (tables) and RCU free them.
+ * We use the sched RCU variant, as that guarantees that IRQ/preempt disabling
+ * holds off grace periods.
+ *
+ * However, in order to batch these pages we need to allocate storage, this
+ * allocation is deep inside the MM code and can thus easily fail on memory
+ * pressure. To guarantee progress we fall back to single table freeing, see
+ * the implementation of tlb_remove_table_one().
+ *
+ */
+struct mmu_table_batch {
+	struct rcu_head		rcu;
+	unsigned int		nr;
+	void			*tables[0];
+};
+
+#define MAX_TABLE_BATCH		\
+	((PAGE_SIZE - sizeof(struct mmu_table_batch)) / sizeof(void *))
+
+extern void tlb_table_flush(struct mmu_gather *tlb);
+extern void tlb_remove_table(struct mmu_gather *tlb, void *table);
+
+#endif
+
 /*
  * If we can't allocate a page to make a big patch of page pointers
  * to work on, then just handle a few from the on-stack structure.
@@ -40,13 +83,13 @@
  */
 struct mmu_gather {
 	struct mm_struct	*mm;
+#ifdef CONFIG_HAVE_RCU_TABLE_FREE
+	struct mmu_table_batch	*batch;
+#endif
 	unsigned int		nr;	/* set to ~0U means fast mode */
 	unsigned int		max;	/* nr < max */
 	unsigned int		need_flush;/* Really unmapped some ptes? */
 	unsigned int		fullmm; /* non-zero means full mm flush */
-#ifdef HAVE_ARCH_MMU_GATHER
-	struct arch_mmu_gather	arch;
-#endif
 	struct page		**pages;
 	struct page		*local[MMU_GATHER_BUNDLE];
 };
@@ -82,8 +125,8 @@ tlb_gather_mmu(struct mmu_gather *tlb, s
 
 	tlb->fullmm = fullmm;
 
-#ifdef HAVE_ARCH_MMU_GATHER
-	tlb->arch = ARCH_MMU_GATHER_INIT;
+#ifdef CONFIG_HAVE_RCU_TABLE_FREE
+	tlb->batch = NULL;
 #endif
 }
 
@@ -94,6 +137,9 @@ tlb_flush_mmu(struct mmu_gather *tlb)
 		return;
 	tlb->need_flush = 0;
 	tlb_flush(tlb);
+#ifdef CONFIG_HAVE_RCU_TABLE_FREE
+	tlb_table_flush(tlb);
+#endif
 	if (!tlb_fast_mode(tlb)) {
 		free_pages_and_swap_cache(tlb->pages, tlb->nr);
 		tlb->nr = 0;
Index: linux-2.6/mm/memory.c
===================================================================
--- linux-2.6.orig/mm/memory.c
+++ linux-2.6/mm/memory.c
@@ -193,6 +193,83 @@ static void check_sync_rss_stat(struct t
 
 #endif
 
+#ifdef CONFIG_HAVE_RCU_TABLE_FREE
+
+/*
+ * See the comment near struct mmu_table_batch.
+ */
+
+static void tlb_remove_table_smp_sync(void *arg)
+{
+	/* Simply deliver the interrupt */
+}
+
+static void tlb_remove_table_one(void *table)
+{
+	/*
+	 * This isn't an RCU grace period and hence the page-tables cannot be
+	 * assumed to be actually RCU-freed.
+	 *
+	 * It is however sufficient for software page-table walkers that rely on
+	 * IRQ disabling. See the comment near struct mmu_table_batch.
+	 */
+	smp_call_function(tlb_remove_table_smp_sync, NULL, 1);
+	__tlb_remove_table(table);
+}
+
+static void tlb_remove_table_rcu(struct rcu_head *head)
+{
+	struct mmu_table_batch *batch;
+	int i;
+
+	batch = container_of(head, struct mmu_table_batch, rcu);
+
+	for (i = 0; i < batch->nr; i++)
+		__tlb_remove_table(batch->tables[i]);
+
+	free_page((unsigned long)batch);
+}
+
+void tlb_table_flush(struct mmu_gather *tlb)
+{
+	struct mmu_table_batch **batch = &tlb->batch;
+
+	if (*batch) {
+		call_rcu_sched(&(*batch)->rcu, tlb_remove_table_rcu);
+		*batch = NULL;
+	}
+}
+
+void tlb_remove_table(struct mmu_gather *tlb, void *table)
+{
+	struct mmu_table_batch **batch = &tlb->batch;
+
+	tlb->need_flush = 1;
+
+	/*
+	 * When there's less then two users of this mm there cannot be a
+	 * concurrent page-table walk.
+	 */
+	if (atomic_read(&tlb->mm->mm_users) < 2) {
+		__tlb_remove_table(table);
+		return;
+	}
+
+	if (*batch == NULL) {
+		*batch = (struct mmu_table_batch *)__get_free_page(GFP_NOWAIT | __GFP_NOWARN);
+		if (*batch == NULL) {
+			tlb_remove_table_one(table);
+			return;
+		}
+		(*batch)->nr = 0;
+	}
+	(*batch)->tables[(*batch)->nr++] = table;
+	if ((*batch)->nr == MAX_TABLE_BATCH)
+		tlb_table_flush(tlb);
+}
+
+#endif
+
 /*
  * If a p?d_bad entry is found while walking page tables, report
  * the error, before resetting entry to p?d_none.  Usually (but
Index: linux-2.6/arch/Kconfig
===================================================================
--- linux-2.6.orig/arch/Kconfig
+++ linux-2.6/arch/Kconfig
@@ -178,4 +178,7 @@ config HAVE_ARCH_JUMP_LABEL
 config HAVE_ARCH_MUTEX_CPU_RELAX
 	bool
 
+config HAVE_RCU_TABLE_FREE
+	bool
+
 source "kernel/gcov/Kconfig"
Index: linux-2.6/arch/powerpc/Kconfig
===================================================================
--- linux-2.6.orig/arch/powerpc/Kconfig
+++ linux-2.6/arch/powerpc/Kconfig
@@ -140,6 +140,7 @@ config PPC
 	select IRQ_PER_CPU
 	select GENERIC_IRQ_SHOW
 	select GENERIC_IRQ_SHOW_LEVEL
+	select HAVE_RCU_TABLE_FREE if SMP
 
 config EARLY_PRINTK
 	bool

[PATCH 10/20] mm, powerpc: Move the RCU page-table freeing into generic code

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-mm_powerpc-move_the_rcu_page-table_freeing_into.patch --]
[-- Type: text/plain, Size: 12906 bytes --]

In case other architectures require RCU freed page-tables to implement
gup_fast() and software filled hashes and similar things, provide the
means to do so by moving the logic into generic code.

Requested-by: David Miller <davem@davemloft.net>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 arch/Kconfig                       |    3 +
 arch/powerpc/Kconfig               |    1 
 arch/powerpc/include/asm/pgalloc.h |   21 ++++++-
 arch/powerpc/include/asm/tlb.h     |   10 ---
 arch/powerpc/mm/pgtable.c          |   98 -------------------------------------
 arch/powerpc/mm/tlb_hash32.c       |    3 -
 arch/powerpc/mm/tlb_hash64.c       |    3 -
 arch/powerpc/mm/tlb_nohash.c       |    3 -
 include/asm-generic/tlb.h          |   56 +++++++++++++++++++--
 mm/memory.c                        |   77 +++++++++++++++++++++++++++++
 10 files changed, 150 insertions(+), 125 deletions(-)

Index: linux-2.6/arch/powerpc/include/asm/pgalloc.h
===================================================================
--- linux-2.6.orig/arch/powerpc/include/asm/pgalloc.h
+++ linux-2.6/arch/powerpc/include/asm/pgalloc.h
@@ -31,14 +31,29 @@ static inline void pte_free(struct mm_st
 #endif
 
 #ifdef CONFIG_SMP
-extern void pgtable_free_tlb(struct mmu_gather *tlb, void *table, unsigned shift);
-extern void pte_free_finish(struct mmu_gather *tlb);
+struct mmu_gather;
+extern void tlb_remove_table(struct mmu_gather *, void *);
+
+static inline void pgtable_free_tlb(struct mmu_gather *tlb, void *table, int shift)
+{
+	unsigned long pgf = (unsigned long)table;
+	BUG_ON(shift > MAX_PGTABLE_INDEX_SIZE);
+	pgf |= shift;
+	tlb_remove_table(tlb, (void *)pgf);
+}
+
+static inline void __tlb_remove_table(void *_table)
+{
+	void *table = (void *)((unsigned long)_table & ~MAX_PGTABLE_INDEX_SIZE);
+	unsigned shift = (unsigned long)_table & MAX_PGTABLE_INDEX_SIZE;
+
+	pgtable_free(table, shift);
+}
 #else /* CONFIG_SMP */
 static inline void pgtable_free_tlb(struct mmu_gather *tlb, void *table, unsigned shift)
 {
 	pgtable_free(table, shift);
 }
-static inline void pte_free_finish(struct mmu_gather *tlb) { }
 #endif /* !CONFIG_SMP */
 
 static inline void __pte_free_tlb(struct mmu_gather *tlb, struct page *ptepage,
Index: linux-2.6/arch/powerpc/include/asm/tlb.h
===================================================================
--- linux-2.6.orig/arch/powerpc/include/asm/tlb.h
+++ linux-2.6/arch/powerpc/include/asm/tlb.h
@@ -28,16 +28,6 @@
 #define tlb_start_vma(tlb, vma)	do { } while (0)
 #define tlb_end_vma(tlb, vma)	do { } while (0)
 
-#define HAVE_ARCH_MMU_GATHER 1
-
-struct pte_freelist_batch;
-
-struct arch_mmu_gather {
-	struct pte_freelist_batch *batch;
-};
-
-#define ARCH_MMU_GATHER_INIT (struct arch_mmu_gather){ .batch = NULL, }
-
 extern void tlb_flush(struct mmu_gather *tlb);
 
 /* Get the generic bits... */
Index: linux-2.6/arch/powerpc/mm/pgtable.c
===================================================================
--- linux-2.6.orig/arch/powerpc/mm/pgtable.c
+++ linux-2.6/arch/powerpc/mm/pgtable.c
@@ -33,104 +33,6 @@
 
 #include "mmu_decl.h"
 
-#ifdef CONFIG_SMP
-
-/*
- * Handle batching of page table freeing on SMP. Page tables are
- * queued up and send to be freed later by RCU in order to avoid
- * freeing a page table page that is being walked without locks
- */
-
-static unsigned long pte_freelist_forced_free;
-
-struct pte_freelist_batch
-{
-	struct rcu_head	rcu;
-	unsigned int	index;
-	unsigned long	tables[0];
-};
-
-#define PTE_FREELIST_SIZE \
-	((PAGE_SIZE - sizeof(struct pte_freelist_batch)) \
-	  / sizeof(unsigned long))
-
-static void pte_free_smp_sync(void *arg)
-{
-	/* Do nothing, just ensure we sync with all CPUs */
-}
-
-/* This is only called when we are critically out of memory
- * (and fail to get a page in pte_free_tlb).
- */
-static void pgtable_free_now(void *table, unsigned shift)
-{
-	pte_freelist_forced_free++;
-
-	smp_call_function(pte_free_smp_sync, NULL, 1);
-
-	pgtable_free(table, shift);
-}
-
-static void pte_free_rcu_callback(struct rcu_head *head)
-{
-	struct pte_freelist_batch *batch =
-		container_of(head, struct pte_freelist_batch, rcu);
-	unsigned int i;
-
-	for (i = 0; i < batch->index; i++) {
-		void *table = (void *)(batch->tables[i] & ~MAX_PGTABLE_INDEX_SIZE);
-		unsigned shift = batch->tables[i] & MAX_PGTABLE_INDEX_SIZE;
-
-		pgtable_free(table, shift);
-	}
-
-	free_page((unsigned long)batch);
-}
-
-static void pte_free_submit(struct pte_freelist_batch *batch)
-{
-	call_rcu_sched(&batch->rcu, pte_free_rcu_callback);
-}
-
-void pgtable_free_tlb(struct mmu_gather *tlb, void *table, unsigned shift)
-{
-	struct pte_freelist_batch **batchp = &tlb->arch.batch;
-	unsigned long pgf;
-
-	if (atomic_read(&tlb->mm->mm_users) < 2) {
-		pgtable_free(table, shift);
-		return;
-	}
-
-	if (*batchp == NULL) {
-		*batchp = (struct pte_freelist_batch *)__get_free_page(GFP_ATOMIC);
-		if (*batchp == NULL) {
-			pgtable_free_now(table, shift);
-			return;
-		}
-		(*batchp)->index = 0;
-	}
-	BUG_ON(shift > MAX_PGTABLE_INDEX_SIZE);
-	pgf = (unsigned long)table | shift;
-	(*batchp)->tables[(*batchp)->index++] = pgf;
-	if ((*batchp)->index == PTE_FREELIST_SIZE) {
-		pte_free_submit(*batchp);
-		*batchp = NULL;
-	}
-}
-
-void pte_free_finish(struct mmu_gather *tlb)
-{
-	struct pte_freelist_batch **batchp = &tlb->arch.batch;
-
-	if (*batchp == NULL)
-		return;
-	pte_free_submit(*batchp);
-	*batchp = NULL;
-}
-
-#endif /* CONFIG_SMP */
-
 static inline int is_exec_fault(void)
 {
 	return current->thread.regs && TRAP(current->thread.regs) == 0x400;
Index: linux-2.6/arch/powerpc/mm/tlb_hash32.c
===================================================================
--- linux-2.6.orig/arch/powerpc/mm/tlb_hash32.c
+++ linux-2.6/arch/powerpc/mm/tlb_hash32.c
@@ -71,9 +71,6 @@ void tlb_flush(struct mmu_gather *tlb)
 		 */
 		_tlbia();
 	}
-
-	/* Push out batch of freed page tables */
-	pte_free_finish(tlb);
 }
 
 /*
Index: linux-2.6/arch/powerpc/mm/tlb_hash64.c
===================================================================
--- linux-2.6.orig/arch/powerpc/mm/tlb_hash64.c
+++ linux-2.6/arch/powerpc/mm/tlb_hash64.c
@@ -165,9 +165,6 @@ void tlb_flush(struct mmu_gather *tlb)
 		__flush_tlb_pending(tlbbatch);
 
 	put_cpu_var(ppc64_tlb_batch);
-
-	/* Push out batch of freed page tables */
-	pte_free_finish(tlb);
 }
 
 /**
Index: linux-2.6/arch/powerpc/mm/tlb_nohash.c
===================================================================
--- linux-2.6.orig/arch/powerpc/mm/tlb_nohash.c
+++ linux-2.6/arch/powerpc/mm/tlb_nohash.c
@@ -299,9 +299,6 @@ EXPORT_SYMBOL(flush_tlb_range);
 void tlb_flush(struct mmu_gather *tlb)
 {
 	flush_tlb_mm(tlb->mm);
-
-	/* Push out batch of freed page tables */
-	pte_free_finish(tlb);
 }
 
 /*
Index: linux-2.6/include/asm-generic/tlb.h
===================================================================
--- linux-2.6.orig/include/asm-generic/tlb.h
+++ linux-2.6/include/asm-generic/tlb.h
@@ -29,6 +29,49 @@
   #define tlb_fast_mode(tlb) 1
 #endif
 
+#ifdef CONFIG_HAVE_RCU_TABLE_FREE
+/*
+ * Semi RCU freeing of the page directories.
+ *
+ * This is needed by some architectures to implement software pagetable walkers.
+ *
+ * gup_fast() and other software pagetable walkers do a lockless page-table
+ * walk and therefore needs some synchronization with the freeing of the page
+ * directories. The chosen means to accomplish that is by disabling IRQs over
+ * the walk.
+ *
+ * Architectures that use IPIs to flush TLBs will then automagically DTRT,
+ * since we unlink the page, flush TLBs, free the page. Since the disabling of
+ * IRQs delays the completion of the TLB flush we can never observe an already
+ * freed page.
+ *
+ * Architectures that do not have this (PPC) need to delay the freeing by some
+ * other means, this is that means.
+ *
+ * What we do is batch the freed directory pages (tables) and RCU free them.
+ * We use the sched RCU variant, as that guarantees that IRQ/preempt disabling
+ * holds off grace periods.
+ *
+ * However, in order to batch these pages we need to allocate storage, this
+ * allocation is deep inside the MM code and can thus easily fail on memory
+ * pressure. To guarantee progress we fall back to single table freeing, see
+ * the implementation of tlb_remove_table_one().
+ *
+ */
+struct mmu_table_batch {
+	struct rcu_head		rcu;
+	unsigned int		nr;
+	void			*tables[0];
+};
+
+#define MAX_TABLE_BATCH		\
+	((PAGE_SIZE - sizeof(struct mmu_table_batch)) / sizeof(void *))
+
+extern void tlb_table_flush(struct mmu_gather *tlb);
+extern void tlb_remove_table(struct mmu_gather *tlb, void *table);
+
+#endif
+
 /*
  * If we can't allocate a page to make a big patch of page pointers
  * to work on, then just handle a few from the on-stack structure.
@@ -40,13 +83,13 @@
  */
 struct mmu_gather {
 	struct mm_struct	*mm;
+#ifdef CONFIG_HAVE_RCU_TABLE_FREE
+	struct mmu_table_batch	*batch;
+#endif
 	unsigned int		nr;	/* set to ~0U means fast mode */
 	unsigned int		max;	/* nr < max */
 	unsigned int		need_flush;/* Really unmapped some ptes? */
 	unsigned int		fullmm; /* non-zero means full mm flush */
-#ifdef HAVE_ARCH_MMU_GATHER
-	struct arch_mmu_gather	arch;
-#endif
 	struct page		**pages;
 	struct page		*local[MMU_GATHER_BUNDLE];
 };
@@ -82,8 +125,8 @@ tlb_gather_mmu(struct mmu_gather *tlb, s
 
 	tlb->fullmm = fullmm;
 
-#ifdef HAVE_ARCH_MMU_GATHER
-	tlb->arch = ARCH_MMU_GATHER_INIT;
+#ifdef CONFIG_HAVE_RCU_TABLE_FREE
+	tlb->batch = NULL;
 #endif
 }
 
@@ -94,6 +137,9 @@ tlb_flush_mmu(struct mmu_gather *tlb)
 		return;
 	tlb->need_flush = 0;
 	tlb_flush(tlb);
+#ifdef CONFIG_HAVE_RCU_TABLE_FREE
+	tlb_table_flush(tlb);
+#endif
 	if (!tlb_fast_mode(tlb)) {
 		free_pages_and_swap_cache(tlb->pages, tlb->nr);
 		tlb->nr = 0;
Index: linux-2.6/mm/memory.c
===================================================================
--- linux-2.6.orig/mm/memory.c
+++ linux-2.6/mm/memory.c
@@ -193,6 +193,83 @@ static void check_sync_rss_stat(struct t
 
 #endif
 
+#ifdef CONFIG_HAVE_RCU_TABLE_FREE
+
+/*
+ * See the comment near struct mmu_table_batch.
+ */
+
+static void tlb_remove_table_smp_sync(void *arg)
+{
+	/* Simply deliver the interrupt */
+}
+
+static void tlb_remove_table_one(void *table)
+{
+	/*
+	 * This isn't an RCU grace period and hence the page-tables cannot be
+	 * assumed to be actually RCU-freed.
+	 *
+	 * It is however sufficient for software page-table walkers that rely on
+	 * IRQ disabling. See the comment near struct mmu_table_batch.
+	 */
+	smp_call_function(tlb_remove_table_smp_sync, NULL, 1);
+	__tlb_remove_table(table);
+}
+
+static void tlb_remove_table_rcu(struct rcu_head *head)
+{
+	struct mmu_table_batch *batch;
+	int i;
+
+	batch = container_of(head, struct mmu_table_batch, rcu);
+
+	for (i = 0; i < batch->nr; i++)
+		__tlb_remove_table(batch->tables[i]);
+
+	free_page((unsigned long)batch);
+}
+
+void tlb_table_flush(struct mmu_gather *tlb)
+{
+	struct mmu_table_batch **batch = &tlb->batch;
+
+	if (*batch) {
+		call_rcu_sched(&(*batch)->rcu, tlb_remove_table_rcu);
+		*batch = NULL;
+	}
+}
+
+void tlb_remove_table(struct mmu_gather *tlb, void *table)
+{
+	struct mmu_table_batch **batch = &tlb->batch;
+
+	tlb->need_flush = 1;
+
+	/*
+	 * When there's less then two users of this mm there cannot be a
+	 * concurrent page-table walk.
+	 */
+	if (atomic_read(&tlb->mm->mm_users) < 2) {
+		__tlb_remove_table(table);
+		return;
+	}
+
+	if (*batch == NULL) {
+		*batch = (struct mmu_table_batch *)__get_free_page(GFP_NOWAIT | __GFP_NOWARN);
+		if (*batch == NULL) {
+			tlb_remove_table_one(table);
+			return;
+		}
+		(*batch)->nr = 0;
+	}
+	(*batch)->tables[(*batch)->nr++] = table;
+	if ((*batch)->nr == MAX_TABLE_BATCH)
+		tlb_table_flush(tlb);
+}
+
+#endif
+
 /*
  * If a p?d_bad entry is found while walking page tables, report
  * the error, before resetting entry to p?d_none.  Usually (but
Index: linux-2.6/arch/Kconfig
===================================================================
--- linux-2.6.orig/arch/Kconfig
+++ linux-2.6/arch/Kconfig
@@ -178,4 +178,7 @@ config HAVE_ARCH_JUMP_LABEL
 config HAVE_ARCH_MUTEX_CPU_RELAX
 	bool
 
+config HAVE_RCU_TABLE_FREE
+	bool
+
 source "kernel/gcov/Kconfig"
Index: linux-2.6/arch/powerpc/Kconfig
===================================================================
--- linux-2.6.orig/arch/powerpc/Kconfig
+++ linux-2.6/arch/powerpc/Kconfig
@@ -140,6 +140,7 @@ config PPC
 	select IRQ_PER_CPU
 	select GENERIC_IRQ_SHOW
 	select GENERIC_IRQ_SHOW_LEVEL
+	select HAVE_RCU_TABLE_FREE if SMP
 
 config EARLY_PRINTK
 	bool


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 10/20] mm, powerpc: Move the RCU page-table freeing into generic code

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-mm_powerpc-move_the_rcu_page-table_freeing_into.patch --]
[-- Type: text/plain, Size: 12906 bytes --]

In case other architectures require RCU freed page-tables to implement
gup_fast() and software filled hashes and similar things, provide the
means to do so by moving the logic into generic code.

Requested-by: David Miller <davem@davemloft.net>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 arch/Kconfig                       |    3 +
 arch/powerpc/Kconfig               |    1 
 arch/powerpc/include/asm/pgalloc.h |   21 ++++++-
 arch/powerpc/include/asm/tlb.h     |   10 ---
 arch/powerpc/mm/pgtable.c          |   98 -------------------------------------
 arch/powerpc/mm/tlb_hash32.c       |    3 -
 arch/powerpc/mm/tlb_hash64.c       |    3 -
 arch/powerpc/mm/tlb_nohash.c       |    3 -
 include/asm-generic/tlb.h          |   56 +++++++++++++++++++--
 mm/memory.c                        |   77 +++++++++++++++++++++++++++++
 10 files changed, 150 insertions(+), 125 deletions(-)

Index: linux-2.6/arch/powerpc/include/asm/pgalloc.h
===================================================================
--- linux-2.6.orig/arch/powerpc/include/asm/pgalloc.h
+++ linux-2.6/arch/powerpc/include/asm/pgalloc.h
@@ -31,14 +31,29 @@ static inline void pte_free(struct mm_st
 #endif
 
 #ifdef CONFIG_SMP
-extern void pgtable_free_tlb(struct mmu_gather *tlb, void *table, unsigned shift);
-extern void pte_free_finish(struct mmu_gather *tlb);
+struct mmu_gather;
+extern void tlb_remove_table(struct mmu_gather *, void *);
+
+static inline void pgtable_free_tlb(struct mmu_gather *tlb, void *table, int shift)
+{
+	unsigned long pgf = (unsigned long)table;
+	BUG_ON(shift > MAX_PGTABLE_INDEX_SIZE);
+	pgf |= shift;
+	tlb_remove_table(tlb, (void *)pgf);
+}
+
+static inline void __tlb_remove_table(void *_table)
+{
+	void *table = (void *)((unsigned long)_table & ~MAX_PGTABLE_INDEX_SIZE);
+	unsigned shift = (unsigned long)_table & MAX_PGTABLE_INDEX_SIZE;
+
+	pgtable_free(table, shift);
+}
 #else /* CONFIG_SMP */
 static inline void pgtable_free_tlb(struct mmu_gather *tlb, void *table, unsigned shift)
 {
 	pgtable_free(table, shift);
 }
-static inline void pte_free_finish(struct mmu_gather *tlb) { }
 #endif /* !CONFIG_SMP */
 
 static inline void __pte_free_tlb(struct mmu_gather *tlb, struct page *ptepage,
Index: linux-2.6/arch/powerpc/include/asm/tlb.h
===================================================================
--- linux-2.6.orig/arch/powerpc/include/asm/tlb.h
+++ linux-2.6/arch/powerpc/include/asm/tlb.h
@@ -28,16 +28,6 @@
 #define tlb_start_vma(tlb, vma)	do { } while (0)
 #define tlb_end_vma(tlb, vma)	do { } while (0)
 
-#define HAVE_ARCH_MMU_GATHER 1
-
-struct pte_freelist_batch;
-
-struct arch_mmu_gather {
-	struct pte_freelist_batch *batch;
-};
-
-#define ARCH_MMU_GATHER_INIT (struct arch_mmu_gather){ .batch = NULL, }
-
 extern void tlb_flush(struct mmu_gather *tlb);
 
 /* Get the generic bits... */
Index: linux-2.6/arch/powerpc/mm/pgtable.c
===================================================================
--- linux-2.6.orig/arch/powerpc/mm/pgtable.c
+++ linux-2.6/arch/powerpc/mm/pgtable.c
@@ -33,104 +33,6 @@
 
 #include "mmu_decl.h"
 
-#ifdef CONFIG_SMP
-
-/*
- * Handle batching of page table freeing on SMP. Page tables are
- * queued up and send to be freed later by RCU in order to avoid
- * freeing a page table page that is being walked without locks
- */
-
-static unsigned long pte_freelist_forced_free;
-
-struct pte_freelist_batch
-{
-	struct rcu_head	rcu;
-	unsigned int	index;
-	unsigned long	tables[0];
-};
-
-#define PTE_FREELIST_SIZE \
-	((PAGE_SIZE - sizeof(struct pte_freelist_batch)) \
-	  / sizeof(unsigned long))
-
-static void pte_free_smp_sync(void *arg)
-{
-	/* Do nothing, just ensure we sync with all CPUs */
-}
-
-/* This is only called when we are critically out of memory
- * (and fail to get a page in pte_free_tlb).
- */
-static void pgtable_free_now(void *table, unsigned shift)
-{
-	pte_freelist_forced_free++;
-
-	smp_call_function(pte_free_smp_sync, NULL, 1);
-
-	pgtable_free(table, shift);
-}
-
-static void pte_free_rcu_callback(struct rcu_head *head)
-{
-	struct pte_freelist_batch *batch =
-		container_of(head, struct pte_freelist_batch, rcu);
-	unsigned int i;
-
-	for (i = 0; i < batch->index; i++) {
-		void *table = (void *)(batch->tables[i] & ~MAX_PGTABLE_INDEX_SIZE);
-		unsigned shift = batch->tables[i] & MAX_PGTABLE_INDEX_SIZE;
-
-		pgtable_free(table, shift);
-	}
-
-	free_page((unsigned long)batch);
-}
-
-static void pte_free_submit(struct pte_freelist_batch *batch)
-{
-	call_rcu_sched(&batch->rcu, pte_free_rcu_callback);
-}
-
-void pgtable_free_tlb(struct mmu_gather *tlb, void *table, unsigned shift)
-{
-	struct pte_freelist_batch **batchp = &tlb->arch.batch;
-	unsigned long pgf;
-
-	if (atomic_read(&tlb->mm->mm_users) < 2) {
-		pgtable_free(table, shift);
-		return;
-	}
-
-	if (*batchp == NULL) {
-		*batchp = (struct pte_freelist_batch *)__get_free_page(GFP_ATOMIC);
-		if (*batchp == NULL) {
-			pgtable_free_now(table, shift);
-			return;
-		}
-		(*batchp)->index = 0;
-	}
-	BUG_ON(shift > MAX_PGTABLE_INDEX_SIZE);
-	pgf = (unsigned long)table | shift;
-	(*batchp)->tables[(*batchp)->index++] = pgf;
-	if ((*batchp)->index == PTE_FREELIST_SIZE) {
-		pte_free_submit(*batchp);
-		*batchp = NULL;
-	}
-}
-
-void pte_free_finish(struct mmu_gather *tlb)
-{
-	struct pte_freelist_batch **batchp = &tlb->arch.batch;
-
-	if (*batchp == NULL)
-		return;
-	pte_free_submit(*batchp);
-	*batchp = NULL;
-}
-
-#endif /* CONFIG_SMP */
-
 static inline int is_exec_fault(void)
 {
 	return current->thread.regs && TRAP(current->thread.regs) == 0x400;
Index: linux-2.6/arch/powerpc/mm/tlb_hash32.c
===================================================================
--- linux-2.6.orig/arch/powerpc/mm/tlb_hash32.c
+++ linux-2.6/arch/powerpc/mm/tlb_hash32.c
@@ -71,9 +71,6 @@ void tlb_flush(struct mmu_gather *tlb)
 		 */
 		_tlbia();
 	}
-
-	/* Push out batch of freed page tables */
-	pte_free_finish(tlb);
 }
 
 /*
Index: linux-2.6/arch/powerpc/mm/tlb_hash64.c
===================================================================
--- linux-2.6.orig/arch/powerpc/mm/tlb_hash64.c
+++ linux-2.6/arch/powerpc/mm/tlb_hash64.c
@@ -165,9 +165,6 @@ void tlb_flush(struct mmu_gather *tlb)
 		__flush_tlb_pending(tlbbatch);
 
 	put_cpu_var(ppc64_tlb_batch);
-
-	/* Push out batch of freed page tables */
-	pte_free_finish(tlb);
 }
 
 /**
Index: linux-2.6/arch/powerpc/mm/tlb_nohash.c
===================================================================
--- linux-2.6.orig/arch/powerpc/mm/tlb_nohash.c
+++ linux-2.6/arch/powerpc/mm/tlb_nohash.c
@@ -299,9 +299,6 @@ EXPORT_SYMBOL(flush_tlb_range);
 void tlb_flush(struct mmu_gather *tlb)
 {
 	flush_tlb_mm(tlb->mm);
-
-	/* Push out batch of freed page tables */
-	pte_free_finish(tlb);
 }
 
 /*
Index: linux-2.6/include/asm-generic/tlb.h
===================================================================
--- linux-2.6.orig/include/asm-generic/tlb.h
+++ linux-2.6/include/asm-generic/tlb.h
@@ -29,6 +29,49 @@
   #define tlb_fast_mode(tlb) 1
 #endif
 
+#ifdef CONFIG_HAVE_RCU_TABLE_FREE
+/*
+ * Semi RCU freeing of the page directories.
+ *
+ * This is needed by some architectures to implement software pagetable walkers.
+ *
+ * gup_fast() and other software pagetable walkers do a lockless page-table
+ * walk and therefore needs some synchronization with the freeing of the page
+ * directories. The chosen means to accomplish that is by disabling IRQs over
+ * the walk.
+ *
+ * Architectures that use IPIs to flush TLBs will then automagically DTRT,
+ * since we unlink the page, flush TLBs, free the page. Since the disabling of
+ * IRQs delays the completion of the TLB flush we can never observe an already
+ * freed page.
+ *
+ * Architectures that do not have this (PPC) need to delay the freeing by some
+ * other means, this is that means.
+ *
+ * What we do is batch the freed directory pages (tables) and RCU free them.
+ * We use the sched RCU variant, as that guarantees that IRQ/preempt disabling
+ * holds off grace periods.
+ *
+ * However, in order to batch these pages we need to allocate storage, this
+ * allocation is deep inside the MM code and can thus easily fail on memory
+ * pressure. To guarantee progress we fall back to single table freeing, see
+ * the implementation of tlb_remove_table_one().
+ *
+ */
+struct mmu_table_batch {
+	struct rcu_head		rcu;
+	unsigned int		nr;
+	void			*tables[0];
+};
+
+#define MAX_TABLE_BATCH		\
+	((PAGE_SIZE - sizeof(struct mmu_table_batch)) / sizeof(void *))
+
+extern void tlb_table_flush(struct mmu_gather *tlb);
+extern void tlb_remove_table(struct mmu_gather *tlb, void *table);
+
+#endif
+
 /*
  * If we can't allocate a page to make a big patch of page pointers
  * to work on, then just handle a few from the on-stack structure.
@@ -40,13 +83,13 @@
  */
 struct mmu_gather {
 	struct mm_struct	*mm;
+#ifdef CONFIG_HAVE_RCU_TABLE_FREE
+	struct mmu_table_batch	*batch;
+#endif
 	unsigned int		nr;	/* set to ~0U means fast mode */
 	unsigned int		max;	/* nr < max */
 	unsigned int		need_flush;/* Really unmapped some ptes? */
 	unsigned int		fullmm; /* non-zero means full mm flush */
-#ifdef HAVE_ARCH_MMU_GATHER
-	struct arch_mmu_gather	arch;
-#endif
 	struct page		**pages;
 	struct page		*local[MMU_GATHER_BUNDLE];
 };
@@ -82,8 +125,8 @@ tlb_gather_mmu(struct mmu_gather *tlb, s
 
 	tlb->fullmm = fullmm;
 
-#ifdef HAVE_ARCH_MMU_GATHER
-	tlb->arch = ARCH_MMU_GATHER_INIT;
+#ifdef CONFIG_HAVE_RCU_TABLE_FREE
+	tlb->batch = NULL;
 #endif
 }
 
@@ -94,6 +137,9 @@ tlb_flush_mmu(struct mmu_gather *tlb)
 		return;
 	tlb->need_flush = 0;
 	tlb_flush(tlb);
+#ifdef CONFIG_HAVE_RCU_TABLE_FREE
+	tlb_table_flush(tlb);
+#endif
 	if (!tlb_fast_mode(tlb)) {
 		free_pages_and_swap_cache(tlb->pages, tlb->nr);
 		tlb->nr = 0;
Index: linux-2.6/mm/memory.c
===================================================================
--- linux-2.6.orig/mm/memory.c
+++ linux-2.6/mm/memory.c
@@ -193,6 +193,83 @@ static void check_sync_rss_stat(struct t
 
 #endif
 
+#ifdef CONFIG_HAVE_RCU_TABLE_FREE
+
+/*
+ * See the comment near struct mmu_table_batch.
+ */
+
+static void tlb_remove_table_smp_sync(void *arg)
+{
+	/* Simply deliver the interrupt */
+}
+
+static void tlb_remove_table_one(void *table)
+{
+	/*
+	 * This isn't an RCU grace period and hence the page-tables cannot be
+	 * assumed to be actually RCU-freed.
+	 *
+	 * It is however sufficient for software page-table walkers that rely on
+	 * IRQ disabling. See the comment near struct mmu_table_batch.
+	 */
+	smp_call_function(tlb_remove_table_smp_sync, NULL, 1);
+	__tlb_remove_table(table);
+}
+
+static void tlb_remove_table_rcu(struct rcu_head *head)
+{
+	struct mmu_table_batch *batch;
+	int i;
+
+	batch = container_of(head, struct mmu_table_batch, rcu);
+
+	for (i = 0; i < batch->nr; i++)
+		__tlb_remove_table(batch->tables[i]);
+
+	free_page((unsigned long)batch);
+}
+
+void tlb_table_flush(struct mmu_gather *tlb)
+{
+	struct mmu_table_batch **batch = &tlb->batch;
+
+	if (*batch) {
+		call_rcu_sched(&(*batch)->rcu, tlb_remove_table_rcu);
+		*batch = NULL;
+	}
+}
+
+void tlb_remove_table(struct mmu_gather *tlb, void *table)
+{
+	struct mmu_table_batch **batch = &tlb->batch;
+
+	tlb->need_flush = 1;
+
+	/*
+	 * When there's less then two users of this mm there cannot be a
+	 * concurrent page-table walk.
+	 */
+	if (atomic_read(&tlb->mm->mm_users) < 2) {
+		__tlb_remove_table(table);
+		return;
+	}
+
+	if (*batch == NULL) {
+		*batch = (struct mmu_table_batch *)__get_free_page(GFP_NOWAIT | __GFP_NOWARN);
+		if (*batch == NULL) {
+			tlb_remove_table_one(table);
+			return;
+		}
+		(*batch)->nr = 0;
+	}
+	(*batch)->tables[(*batch)->nr++] = table;
+	if ((*batch)->nr == MAX_TABLE_BATCH)
+		tlb_table_flush(tlb);
+}
+
+#endif
+
 /*
  * If a p?d_bad entry is found while walking page tables, report
  * the error, before resetting entry to p?d_none.  Usually (but
Index: linux-2.6/arch/Kconfig
===================================================================
--- linux-2.6.orig/arch/Kconfig
+++ linux-2.6/arch/Kconfig
@@ -178,4 +178,7 @@ config HAVE_ARCH_JUMP_LABEL
 config HAVE_ARCH_MUTEX_CPU_RELAX
 	bool
 
+config HAVE_RCU_TABLE_FREE
+	bool
+
 source "kernel/gcov/Kconfig"
Index: linux-2.6/arch/powerpc/Kconfig
===================================================================
--- linux-2.6.orig/arch/powerpc/Kconfig
+++ linux-2.6/arch/powerpc/Kconfig
@@ -140,6 +140,7 @@ config PPC
 	select IRQ_PER_CPU
 	select GENERIC_IRQ_SHOW
 	select GENERIC_IRQ_SHOW_LEVEL
+	select HAVE_RCU_TABLE_FREE if SMP
 
 config EARLY_PRINTK
 	bool


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 11/20] s390: use generic RCP page-table freeing

[Peter Zijlstra]

[-- Attachment #1: martin_schwidefsky-s390-use_generic_rcp_page-table_freeing.patch --]
[-- Type: text/plain, Size: 16209 bytes --]

From: Martin Schwidefsky <schwidefsky@de.ibm.com>

Now that we have a generic implementation for RCU based page table
freeing, use it for s390 as well. It saves a couple of lines.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <20110215193717.30c2bb0a@mschwide.boeblingen.de.ibm.com>
---
 arch/s390/Kconfig               |    1 
 arch/s390/include/asm/pgalloc.h |   19 +--
 arch/s390/include/asm/tlb.h     |   92 ++++++++-----------
 arch/s390/mm/pgtable.c          |  192 +++++-----------------------------------
 4 files changed, 76 insertions(+), 228 deletions(-)

Index: linux-2.6/arch/s390/Kconfig
===================================================================
--- linux-2.6.orig/arch/s390/Kconfig
+++ linux-2.6/arch/s390/Kconfig
@@ -88,6 +88,7 @@ config S390
 	select HAVE_KERNEL_XZ
 	select HAVE_GET_USER_PAGES_FAST
 	select HAVE_ARCH_MUTEX_CPU_RELAX
+	select HAVE_RCU_TABLE_FREE
 	select ARCH_INLINE_SPIN_TRYLOCK
 	select ARCH_INLINE_SPIN_TRYLOCK_BH
 	select ARCH_INLINE_SPIN_LOCK
Index: linux-2.6/arch/s390/include/asm/pgalloc.h
===================================================================
--- linux-2.6.orig/arch/s390/include/asm/pgalloc.h
+++ linux-2.6/arch/s390/include/asm/pgalloc.h
@@ -20,12 +20,11 @@
 #define check_pgt_cache()	do {} while (0)
 
 unsigned long *crst_table_alloc(struct mm_struct *, int);
-void crst_table_free(struct mm_struct *, unsigned long *);
-void crst_table_free_rcu(struct mm_struct *, unsigned long *);
+void crst_table_free(unsigned long *);
 
 unsigned long *page_table_alloc(struct mm_struct *);
-void page_table_free(struct mm_struct *, unsigned long *);
-void page_table_free_rcu(struct mm_struct *, unsigned long *);
+void page_table_free(unsigned long *);
+
 void disable_noexec(struct mm_struct *, struct task_struct *);
 
 static inline void clear_table(unsigned long *s, unsigned long val, size_t n)
@@ -95,7 +94,7 @@ static inline pud_t *pud_alloc_one(struc
 		crst_table_init(table, _REGION3_ENTRY_EMPTY);
 	return (pud_t *) table;
 }
-#define pud_free(mm, pud) crst_table_free(mm, (unsigned long *) pud)
+#define pud_free(mm, pud) crst_table_free((unsigned long *) pud)
 
 static inline pmd_t *pmd_alloc_one(struct mm_struct *mm, unsigned long vmaddr)
 {
@@ -104,7 +103,7 @@ static inline pmd_t *pmd_alloc_one(struc
 		crst_table_init(table, _SEGMENT_ENTRY_EMPTY);
 	return (pmd_t *) table;
 }
-#define pmd_free(mm, pmd) crst_table_free(mm, (unsigned long *) pmd)
+#define pmd_free(mm, pmd) crst_table_free((unsigned long *) pmd)
 
 static inline void pgd_populate_kernel(struct mm_struct *mm,
 				       pgd_t *pgd, pud_t *pud)
@@ -148,7 +147,7 @@ static inline pgd_t *pgd_alloc(struct mm
 	return (pgd_t *)
 		crst_table_alloc(mm, user_mode == SECONDARY_SPACE_MODE);
 }
-#define pgd_free(mm, pgd) crst_table_free(mm, (unsigned long *) pgd)
+#define pgd_free(mm, pgd) crst_table_free((unsigned long *) pgd)
 
 static inline void pmd_populate_kernel(struct mm_struct *mm,
 				       pmd_t *pmd, pte_t *pte)
@@ -175,9 +174,7 @@ static inline void pmd_populate(struct m
 #define pte_alloc_one_kernel(mm, vmaddr) ((pte_t *) page_table_alloc(mm))
 #define pte_alloc_one(mm, vmaddr) ((pte_t *) page_table_alloc(mm))
 
-#define pte_free_kernel(mm, pte) page_table_free(mm, (unsigned long *) pte)
-#define pte_free(mm, pte) page_table_free(mm, (unsigned long *) pte)
-
-extern void rcu_table_freelist_finish(void);
+#define pte_free_kernel(mm, pte) page_table_free((unsigned long *) pte)
+#define pte_free(mm, pte) page_table_free((unsigned long *) pte)
 
 #endif /* _S390_PGALLOC_H */
Index: linux-2.6/arch/s390/include/asm/tlb.h
===================================================================
--- linux-2.6.orig/arch/s390/include/asm/tlb.h
+++ linux-2.6/arch/s390/include/asm/tlb.h
@@ -29,50 +29,45 @@
 #include <asm/smp.h>
 #include <asm/tlbflush.h>
 
+struct mmu_table_batch {
+	struct rcu_head rcu;
+	unsigned int nr;
+	void *tables[0];
+};
+
+#define MAX_TABLE_BATCH \
+	((PAGE_SIZE - sizeof(struct mmu_table_batch)) / sizeof(void *))
+
+void tlb_table_flush(struct mmu_gather *);
+void tlb_remove_table(struct mmu_gather *, void *);
+
 struct mmu_gather {
 	struct mm_struct *mm;
 	unsigned int fullmm;
-	unsigned int nr_ptes;
-	unsigned int nr_pxds;
-	unsigned int max;
-	void **array;
-	void *local[8];
+	struct mmu_table_batch *batch;
+	/* need_flush is used only for page tables */
+	unsigned int need_flush : 1;
 };
 
-static inline void __tlb_alloc_page(struct mmu_gather *tlb)
-{
-	unsigned long addr = __get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
-
-	if (addr) {
-		tlb->array = (void *) addr;
-		tlb->max = PAGE_SIZE / sizeof(void *);
-	}
-}
-
 static inline void tlb_gather_mmu(struct mmu_gather *tlb,
 				  struct mm_struct *mm,
 				  unsigned int full_mm_flush)
 {
 	tlb->mm = mm;
-	tlb->max = ARRAY_SIZE(tlb->local);
-	tlb->array = tlb->local;
 	tlb->fullmm = full_mm_flush;
 	if (tlb->fullmm)
 		__tlb_flush_mm(mm);
-	else
-		__tlb_alloc_page(tlb);
-	tlb->nr_ptes = 0;
-	tlb->nr_pxds = tlb->max;
+	tlb->batch = NULL;
+	tlb->need_flush = 0;
 }
 
 static inline void tlb_flush_mmu(struct mmu_gather *tlb)
 {
-	if (!tlb->fullmm && (tlb->nr_ptes > 0 || tlb->nr_pxds < tlb->max))
+	if (tlb->need_flush) {
+		tlb->need_flush = 0;
 		__tlb_flush_mm(tlb->mm);
-	while (tlb->nr_ptes > 0)
-		page_table_free_rcu(tlb->mm, tlb->array[--tlb->nr_ptes]);
-	while (tlb->nr_pxds < tlb->max)
-		crst_table_free_rcu(tlb->mm, tlb->array[tlb->nr_pxds++]);
+		tlb_table_flush(tlb);
+	}
 }
 
 static inline void tlb_finish_mmu(struct mmu_gather *tlb,
@@ -80,13 +75,8 @@ static inline void tlb_finish_mmu(struct
 {
 	tlb_flush_mmu(tlb);
 
-	rcu_table_freelist_finish();
-
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
-
-	if (tlb->array != tlb->local)
-		free_pages((unsigned long) tlb->array, 0);
 }
 
 /*
@@ -112,12 +102,11 @@ static inline void tlb_remove_page(struc
 static inline void pte_free_tlb(struct mmu_gather *tlb, pgtable_t pte,
 				unsigned long address)
 {
-	if (!tlb->fullmm) {
-		tlb->array[tlb->nr_ptes++] = pte;
-		if (tlb->nr_ptes >= tlb->nr_pxds)
-			tlb_flush_mmu(tlb);
-	} else
-		page_table_free(tlb->mm, (unsigned long *) pte);
+	if (!tlb->fullmm)
+		/* Use LSB to distinguish crst table vs. page table */
+		tlb_remove_table(tlb, (void *) pte + 1);
+	else
+		page_table_free((unsigned long *) pte);
 }
 
 /*
@@ -133,12 +122,10 @@ static inline void pmd_free_tlb(struct m
 #ifdef __s390x__
 	if (tlb->mm->context.asce_limit <= (1UL << 31))
 		return;
-	if (!tlb->fullmm) {
-		tlb->array[--tlb->nr_pxds] = pmd;
-		if (tlb->nr_ptes >= tlb->nr_pxds)
-			tlb_flush_mmu(tlb);
-	} else
-		crst_table_free(tlb->mm, (unsigned long *) pmd);
+	if (!tlb->fullmm)
+		tlb_remove_table(tlb, pmd);
+	else
+		crst_table_free((unsigned long *) pmd);
 #endif
 }
 
@@ -155,15 +142,22 @@ static inline void pud_free_tlb(struct m
 #ifdef __s390x__
 	if (tlb->mm->context.asce_limit <= (1UL << 42))
 		return;
-	if (!tlb->fullmm) {
-		tlb->array[--tlb->nr_pxds] = pud;
-		if (tlb->nr_ptes >= tlb->nr_pxds)
-			tlb_flush_mmu(tlb);
-	} else
-		crst_table_free(tlb->mm, (unsigned long *) pud);
+	if (!tlb->fullmm)
+		tlb_remove_table(tlb, pud);
+	else
+		crst_table_free((unsigned long *) pud);
 #endif
 }
 
+static inline void __tlb_remove_table(void *table)
+{
+	/* Use LSB to distinguish crst table vs. page table */
+	if ((unsigned long) table & 1)
+		page_table_free(table - 1);
+	else
+		crst_table_free(table);
+}
+
 #define tlb_start_vma(tlb, vma)			do { } while (0)
 #define tlb_end_vma(tlb, vma)			do { } while (0)
 #define tlb_remove_tlb_entry(tlb, ptep, addr)	do { } while (0)
Index: linux-2.6/arch/s390/mm/pgtable.c
===================================================================
--- linux-2.6.orig/arch/s390/mm/pgtable.c
+++ linux-2.6/arch/s390/mm/pgtable.c
@@ -24,92 +24,17 @@
 #include <asm/tlbflush.h>
 #include <asm/mmu_context.h>
 
-struct rcu_table_freelist {
-	struct rcu_head rcu;
-	struct mm_struct *mm;
-	unsigned int pgt_index;
-	unsigned int crst_index;
-	unsigned long *table[0];
-};
-
-#define RCU_FREELIST_SIZE \
-	((PAGE_SIZE - sizeof(struct rcu_table_freelist)) \
-	  / sizeof(unsigned long))
-
-static DEFINE_PER_CPU(struct rcu_table_freelist *, rcu_table_freelist);
-
-static void __page_table_free(struct mm_struct *mm, unsigned long *table);
-static void __crst_table_free(struct mm_struct *mm, unsigned long *table);
-
-static struct rcu_table_freelist *rcu_table_freelist_get(struct mm_struct *mm)
-{
-	struct rcu_table_freelist **batchp = &__get_cpu_var(rcu_table_freelist);
-	struct rcu_table_freelist *batch = *batchp;
-
-	if (batch)
-		return batch;
-	batch = (struct rcu_table_freelist *) __get_free_page(GFP_ATOMIC);
-	if (batch) {
-		batch->mm = mm;
-		batch->pgt_index = 0;
-		batch->crst_index = RCU_FREELIST_SIZE;
-		*batchp = batch;
-	}
-	return batch;
-}
-
-static void rcu_table_freelist_callback(struct rcu_head *head)
-{
-	struct rcu_table_freelist *batch =
-		container_of(head, struct rcu_table_freelist, rcu);
-
-	while (batch->pgt_index > 0)
-		__page_table_free(batch->mm, batch->table[--batch->pgt_index]);
-	while (batch->crst_index < RCU_FREELIST_SIZE)
-		__crst_table_free(batch->mm, batch->table[batch->crst_index++]);
-	free_page((unsigned long) batch);
-}
-
-void rcu_table_freelist_finish(void)
-{
-	struct rcu_table_freelist *batch = __get_cpu_var(rcu_table_freelist);
-
-	if (!batch)
-		return;
-	call_rcu(&batch->rcu, rcu_table_freelist_callback);
-	__get_cpu_var(rcu_table_freelist) = NULL;
-}
-
-static void smp_sync(void *arg)
-{
-}
 
 #ifndef CONFIG_64BIT
 #define ALLOC_ORDER	1
 #define TABLES_PER_PAGE	4
 #define FRAG_MASK	15UL
 #define SECOND_HALVES	10UL
-
-void clear_table_pgstes(unsigned long *table)
-{
-	clear_table(table, _PAGE_TYPE_EMPTY, PAGE_SIZE/4);
-	memset(table + 256, 0, PAGE_SIZE/4);
-	clear_table(table + 512, _PAGE_TYPE_EMPTY, PAGE_SIZE/4);
-	memset(table + 768, 0, PAGE_SIZE/4);
-}
-
 #else
 #define ALLOC_ORDER	2
 #define TABLES_PER_PAGE	2
 #define FRAG_MASK	3UL
 #define SECOND_HALVES	2UL
-
-void clear_table_pgstes(unsigned long *table)
-{
-	clear_table(table, _PAGE_TYPE_EMPTY, PAGE_SIZE/2);
-	memset(table + 256, 0, PAGE_SIZE/2);
-}
-
 #endif
 
 unsigned long VMALLOC_START = VMALLOC_END - VMALLOC_SIZE;
@@ -138,6 +63,7 @@ unsigned long *crst_table_alloc(struct m
 			return NULL;
 		}
 		page->index = page_to_phys(shadow);
+		page->private = (unsigned long) mm;
 	}
 	spin_lock_bh(&mm->context.list_lock);
 	list_add(&page->lru, &mm->context.crst_list);
@@ -145,47 +71,19 @@ unsigned long *crst_table_alloc(struct m
 	return (unsigned long *) page_to_phys(page);
 }
 
-static void __crst_table_free(struct mm_struct *mm, unsigned long *table)
-{
-	unsigned long *shadow = get_shadow_table(table);
-
-	if (shadow)
-		free_pages((unsigned long) shadow, ALLOC_ORDER);
-	free_pages((unsigned long) table, ALLOC_ORDER);
-}
-
-void crst_table_free(struct mm_struct *mm, unsigned long *table)
-{
-	struct page *page = virt_to_page(table);
-
-	spin_lock_bh(&mm->context.list_lock);
-	list_del(&page->lru);
-	spin_unlock_bh(&mm->context.list_lock);
-	__crst_table_free(mm, table);
-}
-
-void crst_table_free_rcu(struct mm_struct *mm, unsigned long *table)
+void crst_table_free(unsigned long *table)
 {
-	struct rcu_table_freelist *batch;
 	struct page *page = virt_to_page(table);
+	struct mm_struct *mm = (struct mm_struct *) page->private;
+	unsigned long *shadow = get_shadow_table(table);
 
 	spin_lock_bh(&mm->context.list_lock);
 	list_del(&page->lru);
+	page->private = 0;
 	spin_unlock_bh(&mm->context.list_lock);
-	if (atomic_read(&mm->mm_users) < 2 &&
-	    cpumask_equal(mm_cpumask(mm), cpumask_of(smp_processor_id()))) {
-		__crst_table_free(mm, table);
-		return;
-	}
-	batch = rcu_table_freelist_get(mm);
-	if (!batch) {
-		smp_call_function(smp_sync, NULL, 1);
-		__crst_table_free(mm, table);
-		return;
-	}
-	batch->table[--batch->crst_index] = table;
-	if (batch->pgt_index >= batch->crst_index)
-		rcu_table_freelist_finish();
+	if (shadow)
+		free_pages((unsigned long) shadow, ALLOC_ORDER);
+	free_pages((unsigned long) table, ALLOC_ORDER);
 }
 
 #ifdef CONFIG_64BIT
@@ -223,7 +121,7 @@ int crst_table_upgrade(struct mm_struct 
 	}
 	spin_unlock_bh(&mm->page_table_lock);
 	if (table)
-		crst_table_free(mm, table);
+		crst_table_free(table);
 	if (mm->context.asce_limit < limit)
 		goto repeat;
 	update_mm(mm, current);
@@ -257,7 +155,7 @@ void crst_table_downgrade(struct mm_stru
 		}
 		mm->pgd = (pgd_t *) (pgd_val(*pgd) & _REGION_ENTRY_ORIGIN);
 		mm->task_size = mm->context.asce_limit;
-		crst_table_free(mm, (unsigned long *) pgd);
+		crst_table_free((unsigned long *) pgd);
 	}
 	update_mm(mm, current);
 }
@@ -288,11 +186,7 @@ unsigned long *page_table_alloc(struct m
 			return NULL;
 		pgtable_page_ctor(page);
 		page->flags &= ~FRAG_MASK;
-		table = (unsigned long *) page_to_phys(page);
-		if (mm->context.has_pgste)
-			clear_table_pgstes(table);
-		else
-			clear_table(table, _PAGE_TYPE_EMPTY, PAGE_SIZE);
+		page->private = (unsigned long) mm;
 		spin_lock_bh(&mm->context.list_lock);
 		list_add(&page->lru, &mm->context.pgtable_list);
 	}
@@ -305,42 +199,34 @@ unsigned long *page_table_alloc(struct m
 	if ((page->flags & FRAG_MASK) == ((1UL << TABLES_PER_PAGE) - 1))
 		list_move_tail(&page->lru, &mm->context.pgtable_list);
 	spin_unlock_bh(&mm->context.list_lock);
+	clear_table(table, _PAGE_TYPE_EMPTY, PTRS_PER_PTE * sizeof(long));
+	if (mm->context.noexec)
+		clear_table(table + 256, _PAGE_TYPE_EMPTY,
+			    PTRS_PER_PTE * sizeof(long));
+	else if (mm->context.has_pgste)
+		clear_table(table + 256, 0, PTRS_PER_PTE * sizeof(long));
 	return table;
 }
 
-static void __page_table_free(struct mm_struct *mm, unsigned long *table)
+void page_table_free(unsigned long *table)
 {
-	struct page *page;
-	unsigned long bits;
-
-	bits = ((unsigned long) table) & 15;
-	table = (unsigned long *)(((unsigned long) table) ^ bits);
-	page = pfn_to_page(__pa(table) >> PAGE_SHIFT);
-	page->flags ^= bits;
-	if (!(page->flags & FRAG_MASK)) {
-		pgtable_page_dtor(page);
-		__free_page(page);
-	}
-}
-
-void page_table_free(struct mm_struct *mm, unsigned long *table)
-{
-	struct page *page;
+	struct page *page = virt_to_page(table);
+	struct mm_struct *mm = (struct mm_struct *) page->private;
 	unsigned long bits;
 
 	bits = (mm->context.noexec || mm->context.has_pgste) ? 3UL : 1UL;
 	bits <<= (__pa(table) & (PAGE_SIZE - 1)) / 256 / sizeof(unsigned long);
-	page = pfn_to_page(__pa(table) >> PAGE_SHIFT);
 	spin_lock_bh(&mm->context.list_lock);
 	page->flags ^= bits;
 	if (page->flags & FRAG_MASK) {
 		/* Page now has some free pgtable fragments. */
-		if (!list_empty(&page->lru))
-			list_move(&page->lru, &mm->context.pgtable_list);
+		list_move(&page->lru, &mm->context.pgtable_list);
 		page = NULL;
-	} else
+	} else {
 		/* All fragments of the 4K page have been freed. */
 		list_del(&page->lru);
+		page->private = 0;
+	}
 	spin_unlock_bh(&mm->context.list_lock);
 	if (page) {
 		pgtable_page_dtor(page);
@@ -348,36 +234,6 @@ void page_table_free(struct mm_struct *m
 	}
 }
 
-void page_table_free_rcu(struct mm_struct *mm, unsigned long *table)
-{
-	struct rcu_table_freelist *batch;
-	struct page *page;
-	unsigned long bits;
-
-	if (atomic_read(&mm->mm_users) < 2 &&
-	    cpumask_equal(mm_cpumask(mm), cpumask_of(smp_processor_id()))) {
-		page_table_free(mm, table);
-		return;
-	}
-	batch = rcu_table_freelist_get(mm);
-	if (!batch) {
-		smp_call_function(smp_sync, NULL, 1);
-		page_table_free(mm, table);
-		return;
-	}
-	bits = (mm->context.noexec || mm->context.has_pgste) ? 3UL : 1UL;
-	bits <<= (__pa(table) & (PAGE_SIZE - 1)) / 256 / sizeof(unsigned long);
-	page = pfn_to_page(__pa(table) >> PAGE_SHIFT);
-	spin_lock_bh(&mm->context.list_lock);
-	/* Delayed freeing with rcu prevents reuse of pgtable fragments */
-	list_del_init(&page->lru);
-	spin_unlock_bh(&mm->context.list_lock);
-	table = (unsigned long *)(((unsigned long) table) | bits);
-	batch->table[batch->pgt_index++] = table;
-	if (batch->pgt_index >= batch->crst_index)
-		rcu_table_freelist_finish();
-}
-
 void disable_noexec(struct mm_struct *mm, struct task_struct *tsk)
 {
 	struct page *page;

[PATCH 11/20] s390: use generic RCP page-table freeing

[Peter Zijlstra]

[-- Attachment #1: martin_schwidefsky-s390-use_generic_rcp_page-table_freeing.patch --]
[-- Type: text/plain, Size: 16512 bytes --]

From: Martin Schwidefsky <schwidefsky@de.ibm.com>

Now that we have a generic implementation for RCU based page table
freeing, use it for s390 as well. It saves a couple of lines.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <20110215193717.30c2bb0a@mschwide.boeblingen.de.ibm.com>
---
 arch/s390/Kconfig               |    1 
 arch/s390/include/asm/pgalloc.h |   19 +--
 arch/s390/include/asm/tlb.h     |   92 ++++++++-----------
 arch/s390/mm/pgtable.c          |  192 +++++-----------------------------------
 4 files changed, 76 insertions(+), 228 deletions(-)

Index: linux-2.6/arch/s390/Kconfig
===================================================================
--- linux-2.6.orig/arch/s390/Kconfig
+++ linux-2.6/arch/s390/Kconfig
@@ -88,6 +88,7 @@ config S390
 	select HAVE_KERNEL_XZ
 	select HAVE_GET_USER_PAGES_FAST
 	select HAVE_ARCH_MUTEX_CPU_RELAX
+	select HAVE_RCU_TABLE_FREE
 	select ARCH_INLINE_SPIN_TRYLOCK
 	select ARCH_INLINE_SPIN_TRYLOCK_BH
 	select ARCH_INLINE_SPIN_LOCK
Index: linux-2.6/arch/s390/include/asm/pgalloc.h
===================================================================
--- linux-2.6.orig/arch/s390/include/asm/pgalloc.h
+++ linux-2.6/arch/s390/include/asm/pgalloc.h
@@ -20,12 +20,11 @@
 #define check_pgt_cache()	do {} while (0)
 
 unsigned long *crst_table_alloc(struct mm_struct *, int);
-void crst_table_free(struct mm_struct *, unsigned long *);
-void crst_table_free_rcu(struct mm_struct *, unsigned long *);
+void crst_table_free(unsigned long *);
 
 unsigned long *page_table_alloc(struct mm_struct *);
-void page_table_free(struct mm_struct *, unsigned long *);
-void page_table_free_rcu(struct mm_struct *, unsigned long *);
+void page_table_free(unsigned long *);
+
 void disable_noexec(struct mm_struct *, struct task_struct *);
 
 static inline void clear_table(unsigned long *s, unsigned long val, size_t n)
@@ -95,7 +94,7 @@ static inline pud_t *pud_alloc_one(struc
 		crst_table_init(table, _REGION3_ENTRY_EMPTY);
 	return (pud_t *) table;
 }
-#define pud_free(mm, pud) crst_table_free(mm, (unsigned long *) pud)
+#define pud_free(mm, pud) crst_table_free((unsigned long *) pud)
 
 static inline pmd_t *pmd_alloc_one(struct mm_struct *mm, unsigned long vmaddr)
 {
@@ -104,7 +103,7 @@ static inline pmd_t *pmd_alloc_one(struc
 		crst_table_init(table, _SEGMENT_ENTRY_EMPTY);
 	return (pmd_t *) table;
 }
-#define pmd_free(mm, pmd) crst_table_free(mm, (unsigned long *) pmd)
+#define pmd_free(mm, pmd) crst_table_free((unsigned long *) pmd)
 
 static inline void pgd_populate_kernel(struct mm_struct *mm,
 				       pgd_t *pgd, pud_t *pud)
@@ -148,7 +147,7 @@ static inline pgd_t *pgd_alloc(struct mm
 	return (pgd_t *)
 		crst_table_alloc(mm, user_mode == SECONDARY_SPACE_MODE);
 }
-#define pgd_free(mm, pgd) crst_table_free(mm, (unsigned long *) pgd)
+#define pgd_free(mm, pgd) crst_table_free((unsigned long *) pgd)
 
 static inline void pmd_populate_kernel(struct mm_struct *mm,
 				       pmd_t *pmd, pte_t *pte)
@@ -175,9 +174,7 @@ static inline void pmd_populate(struct m
 #define pte_alloc_one_kernel(mm, vmaddr) ((pte_t *) page_table_alloc(mm))
 #define pte_alloc_one(mm, vmaddr) ((pte_t *) page_table_alloc(mm))
 
-#define pte_free_kernel(mm, pte) page_table_free(mm, (unsigned long *) pte)
-#define pte_free(mm, pte) page_table_free(mm, (unsigned long *) pte)
-
-extern void rcu_table_freelist_finish(void);
+#define pte_free_kernel(mm, pte) page_table_free((unsigned long *) pte)
+#define pte_free(mm, pte) page_table_free((unsigned long *) pte)
 
 #endif /* _S390_PGALLOC_H */
Index: linux-2.6/arch/s390/include/asm/tlb.h
===================================================================
--- linux-2.6.orig/arch/s390/include/asm/tlb.h
+++ linux-2.6/arch/s390/include/asm/tlb.h
@@ -29,50 +29,45 @@
 #include <asm/smp.h>
 #include <asm/tlbflush.h>
 
+struct mmu_table_batch {
+	struct rcu_head rcu;
+	unsigned int nr;
+	void *tables[0];
+};
+
+#define MAX_TABLE_BATCH \
+	((PAGE_SIZE - sizeof(struct mmu_table_batch)) / sizeof(void *))
+
+void tlb_table_flush(struct mmu_gather *);
+void tlb_remove_table(struct mmu_gather *, void *);
+
 struct mmu_gather {
 	struct mm_struct *mm;
 	unsigned int fullmm;
-	unsigned int nr_ptes;
-	unsigned int nr_pxds;
-	unsigned int max;
-	void **array;
-	void *local[8];
+	struct mmu_table_batch *batch;
+	/* need_flush is used only for page tables */
+	unsigned int need_flush : 1;
 };
 
-static inline void __tlb_alloc_page(struct mmu_gather *tlb)
-{
-	unsigned long addr = __get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
-
-	if (addr) {
-		tlb->array = (void *) addr;
-		tlb->max = PAGE_SIZE / sizeof(void *);
-	}
-}
-
 static inline void tlb_gather_mmu(struct mmu_gather *tlb,
 				  struct mm_struct *mm,
 				  unsigned int full_mm_flush)
 {
 	tlb->mm = mm;
-	tlb->max = ARRAY_SIZE(tlb->local);
-	tlb->array = tlb->local;
 	tlb->fullmm = full_mm_flush;
 	if (tlb->fullmm)
 		__tlb_flush_mm(mm);
-	else
-		__tlb_alloc_page(tlb);
-	tlb->nr_ptes = 0;
-	tlb->nr_pxds = tlb->max;
+	tlb->batch = NULL;
+	tlb->need_flush = 0;
 }
 
 static inline void tlb_flush_mmu(struct mmu_gather *tlb)
 {
-	if (!tlb->fullmm && (tlb->nr_ptes > 0 || tlb->nr_pxds < tlb->max))
+	if (tlb->need_flush) {
+		tlb->need_flush = 0;
 		__tlb_flush_mm(tlb->mm);
-	while (tlb->nr_ptes > 0)
-		page_table_free_rcu(tlb->mm, tlb->array[--tlb->nr_ptes]);
-	while (tlb->nr_pxds < tlb->max)
-		crst_table_free_rcu(tlb->mm, tlb->array[tlb->nr_pxds++]);
+		tlb_table_flush(tlb);
+	}
 }
 
 static inline void tlb_finish_mmu(struct mmu_gather *tlb,
@@ -80,13 +75,8 @@ static inline void tlb_finish_mmu(struct
 {
 	tlb_flush_mmu(tlb);
 
-	rcu_table_freelist_finish();
-
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
-
-	if (tlb->array != tlb->local)
-		free_pages((unsigned long) tlb->array, 0);
 }
 
 /*
@@ -112,12 +102,11 @@ static inline void tlb_remove_page(struc
 static inline void pte_free_tlb(struct mmu_gather *tlb, pgtable_t pte,
 				unsigned long address)
 {
-	if (!tlb->fullmm) {
-		tlb->array[tlb->nr_ptes++] = pte;
-		if (tlb->nr_ptes >= tlb->nr_pxds)
-			tlb_flush_mmu(tlb);
-	} else
-		page_table_free(tlb->mm, (unsigned long *) pte);
+	if (!tlb->fullmm)
+		/* Use LSB to distinguish crst table vs. page table */
+		tlb_remove_table(tlb, (void *) pte + 1);
+	else
+		page_table_free((unsigned long *) pte);
 }
 
 /*
@@ -133,12 +122,10 @@ static inline void pmd_free_tlb(struct m
 #ifdef __s390x__
 	if (tlb->mm->context.asce_limit <= (1UL << 31))
 		return;
-	if (!tlb->fullmm) {
-		tlb->array[--tlb->nr_pxds] = pmd;
-		if (tlb->nr_ptes >= tlb->nr_pxds)
-			tlb_flush_mmu(tlb);
-	} else
-		crst_table_free(tlb->mm, (unsigned long *) pmd);
+	if (!tlb->fullmm)
+		tlb_remove_table(tlb, pmd);
+	else
+		crst_table_free((unsigned long *) pmd);
 #endif
 }
 
@@ -155,15 +142,22 @@ static inline void pud_free_tlb(struct m
 #ifdef __s390x__
 	if (tlb->mm->context.asce_limit <= (1UL << 42))
 		return;
-	if (!tlb->fullmm) {
-		tlb->array[--tlb->nr_pxds] = pud;
-		if (tlb->nr_ptes >= tlb->nr_pxds)
-			tlb_flush_mmu(tlb);
-	} else
-		crst_table_free(tlb->mm, (unsigned long *) pud);
+	if (!tlb->fullmm)
+		tlb_remove_table(tlb, pud);
+	else
+		crst_table_free((unsigned long *) pud);
 #endif
 }
 
+static inline void __tlb_remove_table(void *table)
+{
+	/* Use LSB to distinguish crst table vs. page table */
+	if ((unsigned long) table & 1)
+		page_table_free(table - 1);
+	else
+		crst_table_free(table);
+}
+
 #define tlb_start_vma(tlb, vma)			do { } while (0)
 #define tlb_end_vma(tlb, vma)			do { } while (0)
 #define tlb_remove_tlb_entry(tlb, ptep, addr)	do { } while (0)
Index: linux-2.6/arch/s390/mm/pgtable.c
===================================================================
--- linux-2.6.orig/arch/s390/mm/pgtable.c
+++ linux-2.6/arch/s390/mm/pgtable.c
@@ -24,92 +24,17 @@
 #include <asm/tlbflush.h>
 #include <asm/mmu_context.h>
 
-struct rcu_table_freelist {
-	struct rcu_head rcu;
-	struct mm_struct *mm;
-	unsigned int pgt_index;
-	unsigned int crst_index;
-	unsigned long *table[0];
-};
-
-#define RCU_FREELIST_SIZE \
-	((PAGE_SIZE - sizeof(struct rcu_table_freelist)) \
-	  / sizeof(unsigned long))
-
-static DEFINE_PER_CPU(struct rcu_table_freelist *, rcu_table_freelist);
-
-static void __page_table_free(struct mm_struct *mm, unsigned long *table);
-static void __crst_table_free(struct mm_struct *mm, unsigned long *table);
-
-static struct rcu_table_freelist *rcu_table_freelist_get(struct mm_struct *mm)
-{
-	struct rcu_table_freelist **batchp = &__get_cpu_var(rcu_table_freelist);
-	struct rcu_table_freelist *batch = *batchp;
-
-	if (batch)
-		return batch;
-	batch = (struct rcu_table_freelist *) __get_free_page(GFP_ATOMIC);
-	if (batch) {
-		batch->mm = mm;
-		batch->pgt_index = 0;
-		batch->crst_index = RCU_FREELIST_SIZE;
-		*batchp = batch;
-	}
-	return batch;
-}
-
-static void rcu_table_freelist_callback(struct rcu_head *head)
-{
-	struct rcu_table_freelist *batch =
-		container_of(head, struct rcu_table_freelist, rcu);
-
-	while (batch->pgt_index > 0)
-		__page_table_free(batch->mm, batch->table[--batch->pgt_index]);
-	while (batch->crst_index < RCU_FREELIST_SIZE)
-		__crst_table_free(batch->mm, batch->table[batch->crst_index++]);
-	free_page((unsigned long) batch);
-}
-
-void rcu_table_freelist_finish(void)
-{
-	struct rcu_table_freelist *batch = __get_cpu_var(rcu_table_freelist);
-
-	if (!batch)
-		return;
-	call_rcu(&batch->rcu, rcu_table_freelist_callback);
-	__get_cpu_var(rcu_table_freelist) = NULL;
-}
-
-static void smp_sync(void *arg)
-{
-}
 
 #ifndef CONFIG_64BIT
 #define ALLOC_ORDER	1
 #define TABLES_PER_PAGE	4
 #define FRAG_MASK	15UL
 #define SECOND_HALVES	10UL
-
-void clear_table_pgstes(unsigned long *table)
-{
-	clear_table(table, _PAGE_TYPE_EMPTY, PAGE_SIZE/4);
-	memset(table + 256, 0, PAGE_SIZE/4);
-	clear_table(table + 512, _PAGE_TYPE_EMPTY, PAGE_SIZE/4);
-	memset(table + 768, 0, PAGE_SIZE/4);
-}
-
 #else
 #define ALLOC_ORDER	2
 #define TABLES_PER_PAGE	2
 #define FRAG_MASK	3UL
 #define SECOND_HALVES	2UL
-
-void clear_table_pgstes(unsigned long *table)
-{
-	clear_table(table, _PAGE_TYPE_EMPTY, PAGE_SIZE/2);
-	memset(table + 256, 0, PAGE_SIZE/2);
-}
-
 #endif
 
 unsigned long VMALLOC_START = VMALLOC_END - VMALLOC_SIZE;
@@ -138,6 +63,7 @@ unsigned long *crst_table_alloc(struct m
 			return NULL;
 		}
 		page->index = page_to_phys(shadow);
+		page->private = (unsigned long) mm;
 	}
 	spin_lock_bh(&mm->context.list_lock);
 	list_add(&page->lru, &mm->context.crst_list);
@@ -145,47 +71,19 @@ unsigned long *crst_table_alloc(struct m
 	return (unsigned long *) page_to_phys(page);
 }
 
-static void __crst_table_free(struct mm_struct *mm, unsigned long *table)
-{
-	unsigned long *shadow = get_shadow_table(table);
-
-	if (shadow)
-		free_pages((unsigned long) shadow, ALLOC_ORDER);
-	free_pages((unsigned long) table, ALLOC_ORDER);
-}
-
-void crst_table_free(struct mm_struct *mm, unsigned long *table)
-{
-	struct page *page = virt_to_page(table);
-
-	spin_lock_bh(&mm->context.list_lock);
-	list_del(&page->lru);
-	spin_unlock_bh(&mm->context.list_lock);
-	__crst_table_free(mm, table);
-}
-
-void crst_table_free_rcu(struct mm_struct *mm, unsigned long *table)
+void crst_table_free(unsigned long *table)
 {
-	struct rcu_table_freelist *batch;
 	struct page *page = virt_to_page(table);
+	struct mm_struct *mm = (struct mm_struct *) page->private;
+	unsigned long *shadow = get_shadow_table(table);
 
 	spin_lock_bh(&mm->context.list_lock);
 	list_del(&page->lru);
+	page->private = 0;
 	spin_unlock_bh(&mm->context.list_lock);
-	if (atomic_read(&mm->mm_users) < 2 &&
-	    cpumask_equal(mm_cpumask(mm), cpumask_of(smp_processor_id()))) {
-		__crst_table_free(mm, table);
-		return;
-	}
-	batch = rcu_table_freelist_get(mm);
-	if (!batch) {
-		smp_call_function(smp_sync, NULL, 1);
-		__crst_table_free(mm, table);
-		return;
-	}
-	batch->table[--batch->crst_index] = table;
-	if (batch->pgt_index >= batch->crst_index)
-		rcu_table_freelist_finish();
+	if (shadow)
+		free_pages((unsigned long) shadow, ALLOC_ORDER);
+	free_pages((unsigned long) table, ALLOC_ORDER);
 }
 
 #ifdef CONFIG_64BIT
@@ -223,7 +121,7 @@ int crst_table_upgrade(struct mm_struct 
 	}
 	spin_unlock_bh(&mm->page_table_lock);
 	if (table)
-		crst_table_free(mm, table);
+		crst_table_free(table);
 	if (mm->context.asce_limit < limit)
 		goto repeat;
 	update_mm(mm, current);
@@ -257,7 +155,7 @@ void crst_table_downgrade(struct mm_stru
 		}
 		mm->pgd = (pgd_t *) (pgd_val(*pgd) & _REGION_ENTRY_ORIGIN);
 		mm->task_size = mm->context.asce_limit;
-		crst_table_free(mm, (unsigned long *) pgd);
+		crst_table_free((unsigned long *) pgd);
 	}
 	update_mm(mm, current);
 }
@@ -288,11 +186,7 @@ unsigned long *page_table_alloc(struct m
 			return NULL;
 		pgtable_page_ctor(page);
 		page->flags &= ~FRAG_MASK;
-		table = (unsigned long *) page_to_phys(page);
-		if (mm->context.has_pgste)
-			clear_table_pgstes(table);
-		else
-			clear_table(table, _PAGE_TYPE_EMPTY, PAGE_SIZE);
+		page->private = (unsigned long) mm;
 		spin_lock_bh(&mm->context.list_lock);
 		list_add(&page->lru, &mm->context.pgtable_list);
 	}
@@ -305,42 +199,34 @@ unsigned long *page_table_alloc(struct m
 	if ((page->flags & FRAG_MASK) == ((1UL << TABLES_PER_PAGE) - 1))
 		list_move_tail(&page->lru, &mm->context.pgtable_list);
 	spin_unlock_bh(&mm->context.list_lock);
+	clear_table(table, _PAGE_TYPE_EMPTY, PTRS_PER_PTE * sizeof(long));
+	if (mm->context.noexec)
+		clear_table(table + 256, _PAGE_TYPE_EMPTY,
+			    PTRS_PER_PTE * sizeof(long));
+	else if (mm->context.has_pgste)
+		clear_table(table + 256, 0, PTRS_PER_PTE * sizeof(long));
 	return table;
 }
 
-static void __page_table_free(struct mm_struct *mm, unsigned long *table)
+void page_table_free(unsigned long *table)
 {
-	struct page *page;
-	unsigned long bits;
-
-	bits = ((unsigned long) table) & 15;
-	table = (unsigned long *)(((unsigned long) table) ^ bits);
-	page = pfn_to_page(__pa(table) >> PAGE_SHIFT);
-	page->flags ^= bits;
-	if (!(page->flags & FRAG_MASK)) {
-		pgtable_page_dtor(page);
-		__free_page(page);
-	}
-}
-
-void page_table_free(struct mm_struct *mm, unsigned long *table)
-{
-	struct page *page;
+	struct page *page = virt_to_page(table);
+	struct mm_struct *mm = (struct mm_struct *) page->private;
 	unsigned long bits;
 
 	bits = (mm->context.noexec || mm->context.has_pgste) ? 3UL : 1UL;
 	bits <<= (__pa(table) & (PAGE_SIZE - 1)) / 256 / sizeof(unsigned long);
-	page = pfn_to_page(__pa(table) >> PAGE_SHIFT);
 	spin_lock_bh(&mm->context.list_lock);
 	page->flags ^= bits;
 	if (page->flags & FRAG_MASK) {
 		/* Page now has some free pgtable fragments. */
-		if (!list_empty(&page->lru))
-			list_move(&page->lru, &mm->context.pgtable_list);
+		list_move(&page->lru, &mm->context.pgtable_list);
 		page = NULL;
-	} else
+	} else {
 		/* All fragments of the 4K page have been freed. */
 		list_del(&page->lru);
+		page->private = 0;
+	}
 	spin_unlock_bh(&mm->context.list_lock);
 	if (page) {
 		pgtable_page_dtor(page);
@@ -348,36 +234,6 @@ void page_table_free(struct mm_struct *m
 	}
 }
 
-void page_table_free_rcu(struct mm_struct *mm, unsigned long *table)
-{
-	struct rcu_table_freelist *batch;
-	struct page *page;
-	unsigned long bits;
-
-	if (atomic_read(&mm->mm_users) < 2 &&
-	    cpumask_equal(mm_cpumask(mm), cpumask_of(smp_processor_id()))) {
-		page_table_free(mm, table);
-		return;
-	}
-	batch = rcu_table_freelist_get(mm);
-	if (!batch) {
-		smp_call_function(smp_sync, NULL, 1);
-		page_table_free(mm, table);
-		return;
-	}
-	bits = (mm->context.noexec || mm->context.has_pgste) ? 3UL : 1UL;
-	bits <<= (__pa(table) & (PAGE_SIZE - 1)) / 256 / sizeof(unsigned long);
-	page = pfn_to_page(__pa(table) >> PAGE_SHIFT);
-	spin_lock_bh(&mm->context.list_lock);
-	/* Delayed freeing with rcu prevents reuse of pgtable fragments */
-	list_del_init(&page->lru);
-	spin_unlock_bh(&mm->context.list_lock);
-	table = (unsigned long *)(((unsigned long) table) | bits);
-	batch->table[batch->pgt_index++] = table;
-	if (batch->pgt_index >= batch->crst_index)
-		rcu_table_freelist_finish();
-}
-
 void disable_noexec(struct mm_struct *mm, struct task_struct *tsk)
 {
 	struct page *page;


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 11/20] s390: use generic RCP page-table freeing

[Peter Zijlstra]

[-- Attachment #1: martin_schwidefsky-s390-use_generic_rcp_page-table_freeing.patch --]
[-- Type: text/plain, Size: 16512 bytes --]

From: Martin Schwidefsky <schwidefsky@de.ibm.com>

Now that we have a generic implementation for RCU based page table
freeing, use it for s390 as well. It saves a couple of lines.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <20110215193717.30c2bb0a@mschwide.boeblingen.de.ibm.com>
---
 arch/s390/Kconfig               |    1 
 arch/s390/include/asm/pgalloc.h |   19 +--
 arch/s390/include/asm/tlb.h     |   92 ++++++++-----------
 arch/s390/mm/pgtable.c          |  192 +++++-----------------------------------
 4 files changed, 76 insertions(+), 228 deletions(-)

Index: linux-2.6/arch/s390/Kconfig
===================================================================
--- linux-2.6.orig/arch/s390/Kconfig
+++ linux-2.6/arch/s390/Kconfig
@@ -88,6 +88,7 @@ config S390
 	select HAVE_KERNEL_XZ
 	select HAVE_GET_USER_PAGES_FAST
 	select HAVE_ARCH_MUTEX_CPU_RELAX
+	select HAVE_RCU_TABLE_FREE
 	select ARCH_INLINE_SPIN_TRYLOCK
 	select ARCH_INLINE_SPIN_TRYLOCK_BH
 	select ARCH_INLINE_SPIN_LOCK
Index: linux-2.6/arch/s390/include/asm/pgalloc.h
===================================================================
--- linux-2.6.orig/arch/s390/include/asm/pgalloc.h
+++ linux-2.6/arch/s390/include/asm/pgalloc.h
@@ -20,12 +20,11 @@
 #define check_pgt_cache()	do {} while (0)
 
 unsigned long *crst_table_alloc(struct mm_struct *, int);
-void crst_table_free(struct mm_struct *, unsigned long *);
-void crst_table_free_rcu(struct mm_struct *, unsigned long *);
+void crst_table_free(unsigned long *);
 
 unsigned long *page_table_alloc(struct mm_struct *);
-void page_table_free(struct mm_struct *, unsigned long *);
-void page_table_free_rcu(struct mm_struct *, unsigned long *);
+void page_table_free(unsigned long *);
+
 void disable_noexec(struct mm_struct *, struct task_struct *);
 
 static inline void clear_table(unsigned long *s, unsigned long val, size_t n)
@@ -95,7 +94,7 @@ static inline pud_t *pud_alloc_one(struc
 		crst_table_init(table, _REGION3_ENTRY_EMPTY);
 	return (pud_t *) table;
 }
-#define pud_free(mm, pud) crst_table_free(mm, (unsigned long *) pud)
+#define pud_free(mm, pud) crst_table_free((unsigned long *) pud)
 
 static inline pmd_t *pmd_alloc_one(struct mm_struct *mm, unsigned long vmaddr)
 {
@@ -104,7 +103,7 @@ static inline pmd_t *pmd_alloc_one(struc
 		crst_table_init(table, _SEGMENT_ENTRY_EMPTY);
 	return (pmd_t *) table;
 }
-#define pmd_free(mm, pmd) crst_table_free(mm, (unsigned long *) pmd)
+#define pmd_free(mm, pmd) crst_table_free((unsigned long *) pmd)
 
 static inline void pgd_populate_kernel(struct mm_struct *mm,
 				       pgd_t *pgd, pud_t *pud)
@@ -148,7 +147,7 @@ static inline pgd_t *pgd_alloc(struct mm
 	return (pgd_t *)
 		crst_table_alloc(mm, user_mode == SECONDARY_SPACE_MODE);
 }
-#define pgd_free(mm, pgd) crst_table_free(mm, (unsigned long *) pgd)
+#define pgd_free(mm, pgd) crst_table_free((unsigned long *) pgd)
 
 static inline void pmd_populate_kernel(struct mm_struct *mm,
 				       pmd_t *pmd, pte_t *pte)
@@ -175,9 +174,7 @@ static inline void pmd_populate(struct m
 #define pte_alloc_one_kernel(mm, vmaddr) ((pte_t *) page_table_alloc(mm))
 #define pte_alloc_one(mm, vmaddr) ((pte_t *) page_table_alloc(mm))
 
-#define pte_free_kernel(mm, pte) page_table_free(mm, (unsigned long *) pte)
-#define pte_free(mm, pte) page_table_free(mm, (unsigned long *) pte)
-
-extern void rcu_table_freelist_finish(void);
+#define pte_free_kernel(mm, pte) page_table_free((unsigned long *) pte)
+#define pte_free(mm, pte) page_table_free((unsigned long *) pte)
 
 #endif /* _S390_PGALLOC_H */
Index: linux-2.6/arch/s390/include/asm/tlb.h
===================================================================
--- linux-2.6.orig/arch/s390/include/asm/tlb.h
+++ linux-2.6/arch/s390/include/asm/tlb.h
@@ -29,50 +29,45 @@
 #include <asm/smp.h>
 #include <asm/tlbflush.h>
 
+struct mmu_table_batch {
+	struct rcu_head rcu;
+	unsigned int nr;
+	void *tables[0];
+};
+
+#define MAX_TABLE_BATCH \
+	((PAGE_SIZE - sizeof(struct mmu_table_batch)) / sizeof(void *))
+
+void tlb_table_flush(struct mmu_gather *);
+void tlb_remove_table(struct mmu_gather *, void *);
+
 struct mmu_gather {
 	struct mm_struct *mm;
 	unsigned int fullmm;
-	unsigned int nr_ptes;
-	unsigned int nr_pxds;
-	unsigned int max;
-	void **array;
-	void *local[8];
+	struct mmu_table_batch *batch;
+	/* need_flush is used only for page tables */
+	unsigned int need_flush : 1;
 };
 
-static inline void __tlb_alloc_page(struct mmu_gather *tlb)
-{
-	unsigned long addr = __get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
-
-	if (addr) {
-		tlb->array = (void *) addr;
-		tlb->max = PAGE_SIZE / sizeof(void *);
-	}
-}
-
 static inline void tlb_gather_mmu(struct mmu_gather *tlb,
 				  struct mm_struct *mm,
 				  unsigned int full_mm_flush)
 {
 	tlb->mm = mm;
-	tlb->max = ARRAY_SIZE(tlb->local);
-	tlb->array = tlb->local;
 	tlb->fullmm = full_mm_flush;
 	if (tlb->fullmm)
 		__tlb_flush_mm(mm);
-	else
-		__tlb_alloc_page(tlb);
-	tlb->nr_ptes = 0;
-	tlb->nr_pxds = tlb->max;
+	tlb->batch = NULL;
+	tlb->need_flush = 0;
 }
 
 static inline void tlb_flush_mmu(struct mmu_gather *tlb)
 {
-	if (!tlb->fullmm && (tlb->nr_ptes > 0 || tlb->nr_pxds < tlb->max))
+	if (tlb->need_flush) {
+		tlb->need_flush = 0;
 		__tlb_flush_mm(tlb->mm);
-	while (tlb->nr_ptes > 0)
-		page_table_free_rcu(tlb->mm, tlb->array[--tlb->nr_ptes]);
-	while (tlb->nr_pxds < tlb->max)
-		crst_table_free_rcu(tlb->mm, tlb->array[tlb->nr_pxds++]);
+		tlb_table_flush(tlb);
+	}
 }
 
 static inline void tlb_finish_mmu(struct mmu_gather *tlb,
@@ -80,13 +75,8 @@ static inline void tlb_finish_mmu(struct
 {
 	tlb_flush_mmu(tlb);
 
-	rcu_table_freelist_finish();
-
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
-
-	if (tlb->array != tlb->local)
-		free_pages((unsigned long) tlb->array, 0);
 }
 
 /*
@@ -112,12 +102,11 @@ static inline void tlb_remove_page(struc
 static inline void pte_free_tlb(struct mmu_gather *tlb, pgtable_t pte,
 				unsigned long address)
 {
-	if (!tlb->fullmm) {
-		tlb->array[tlb->nr_ptes++] = pte;
-		if (tlb->nr_ptes >= tlb->nr_pxds)
-			tlb_flush_mmu(tlb);
-	} else
-		page_table_free(tlb->mm, (unsigned long *) pte);
+	if (!tlb->fullmm)
+		/* Use LSB to distinguish crst table vs. page table */
+		tlb_remove_table(tlb, (void *) pte + 1);
+	else
+		page_table_free((unsigned long *) pte);
 }
 
 /*
@@ -133,12 +122,10 @@ static inline void pmd_free_tlb(struct m
 #ifdef __s390x__
 	if (tlb->mm->context.asce_limit <= (1UL << 31))
 		return;
-	if (!tlb->fullmm) {
-		tlb->array[--tlb->nr_pxds] = pmd;
-		if (tlb->nr_ptes >= tlb->nr_pxds)
-			tlb_flush_mmu(tlb);
-	} else
-		crst_table_free(tlb->mm, (unsigned long *) pmd);
+	if (!tlb->fullmm)
+		tlb_remove_table(tlb, pmd);
+	else
+		crst_table_free((unsigned long *) pmd);
 #endif
 }
 
@@ -155,15 +142,22 @@ static inline void pud_free_tlb(struct m
 #ifdef __s390x__
 	if (tlb->mm->context.asce_limit <= (1UL << 42))
 		return;
-	if (!tlb->fullmm) {
-		tlb->array[--tlb->nr_pxds] = pud;
-		if (tlb->nr_ptes >= tlb->nr_pxds)
-			tlb_flush_mmu(tlb);
-	} else
-		crst_table_free(tlb->mm, (unsigned long *) pud);
+	if (!tlb->fullmm)
+		tlb_remove_table(tlb, pud);
+	else
+		crst_table_free((unsigned long *) pud);
 #endif
 }
 
+static inline void __tlb_remove_table(void *table)
+{
+	/* Use LSB to distinguish crst table vs. page table */
+	if ((unsigned long) table & 1)
+		page_table_free(table - 1);
+	else
+		crst_table_free(table);
+}
+
 #define tlb_start_vma(tlb, vma)			do { } while (0)
 #define tlb_end_vma(tlb, vma)			do { } while (0)
 #define tlb_remove_tlb_entry(tlb, ptep, addr)	do { } while (0)
Index: linux-2.6/arch/s390/mm/pgtable.c
===================================================================
--- linux-2.6.orig/arch/s390/mm/pgtable.c
+++ linux-2.6/arch/s390/mm/pgtable.c
@@ -24,92 +24,17 @@
 #include <asm/tlbflush.h>
 #include <asm/mmu_context.h>
 
-struct rcu_table_freelist {
-	struct rcu_head rcu;
-	struct mm_struct *mm;
-	unsigned int pgt_index;
-	unsigned int crst_index;
-	unsigned long *table[0];
-};
-
-#define RCU_FREELIST_SIZE \
-	((PAGE_SIZE - sizeof(struct rcu_table_freelist)) \
-	  / sizeof(unsigned long))
-
-static DEFINE_PER_CPU(struct rcu_table_freelist *, rcu_table_freelist);
-
-static void __page_table_free(struct mm_struct *mm, unsigned long *table);
-static void __crst_table_free(struct mm_struct *mm, unsigned long *table);
-
-static struct rcu_table_freelist *rcu_table_freelist_get(struct mm_struct *mm)
-{
-	struct rcu_table_freelist **batchp = &__get_cpu_var(rcu_table_freelist);
-	struct rcu_table_freelist *batch = *batchp;
-
-	if (batch)
-		return batch;
-	batch = (struct rcu_table_freelist *) __get_free_page(GFP_ATOMIC);
-	if (batch) {
-		batch->mm = mm;
-		batch->pgt_index = 0;
-		batch->crst_index = RCU_FREELIST_SIZE;
-		*batchp = batch;
-	}
-	return batch;
-}
-
-static void rcu_table_freelist_callback(struct rcu_head *head)
-{
-	struct rcu_table_freelist *batch =
-		container_of(head, struct rcu_table_freelist, rcu);
-
-	while (batch->pgt_index > 0)
-		__page_table_free(batch->mm, batch->table[--batch->pgt_index]);
-	while (batch->crst_index < RCU_FREELIST_SIZE)
-		__crst_table_free(batch->mm, batch->table[batch->crst_index++]);
-	free_page((unsigned long) batch);
-}
-
-void rcu_table_freelist_finish(void)
-{
-	struct rcu_table_freelist *batch = __get_cpu_var(rcu_table_freelist);
-
-	if (!batch)
-		return;
-	call_rcu(&batch->rcu, rcu_table_freelist_callback);
-	__get_cpu_var(rcu_table_freelist) = NULL;
-}
-
-static void smp_sync(void *arg)
-{
-}
 
 #ifndef CONFIG_64BIT
 #define ALLOC_ORDER	1
 #define TABLES_PER_PAGE	4
 #define FRAG_MASK	15UL
 #define SECOND_HALVES	10UL
-
-void clear_table_pgstes(unsigned long *table)
-{
-	clear_table(table, _PAGE_TYPE_EMPTY, PAGE_SIZE/4);
-	memset(table + 256, 0, PAGE_SIZE/4);
-	clear_table(table + 512, _PAGE_TYPE_EMPTY, PAGE_SIZE/4);
-	memset(table + 768, 0, PAGE_SIZE/4);
-}
-
 #else
 #define ALLOC_ORDER	2
 #define TABLES_PER_PAGE	2
 #define FRAG_MASK	3UL
 #define SECOND_HALVES	2UL
-
-void clear_table_pgstes(unsigned long *table)
-{
-	clear_table(table, _PAGE_TYPE_EMPTY, PAGE_SIZE/2);
-	memset(table + 256, 0, PAGE_SIZE/2);
-}
-
 #endif
 
 unsigned long VMALLOC_START = VMALLOC_END - VMALLOC_SIZE;
@@ -138,6 +63,7 @@ unsigned long *crst_table_alloc(struct m
 			return NULL;
 		}
 		page->index = page_to_phys(shadow);
+		page->private = (unsigned long) mm;
 	}
 	spin_lock_bh(&mm->context.list_lock);
 	list_add(&page->lru, &mm->context.crst_list);
@@ -145,47 +71,19 @@ unsigned long *crst_table_alloc(struct m
 	return (unsigned long *) page_to_phys(page);
 }
 
-static void __crst_table_free(struct mm_struct *mm, unsigned long *table)
-{
-	unsigned long *shadow = get_shadow_table(table);
-
-	if (shadow)
-		free_pages((unsigned long) shadow, ALLOC_ORDER);
-	free_pages((unsigned long) table, ALLOC_ORDER);
-}
-
-void crst_table_free(struct mm_struct *mm, unsigned long *table)
-{
-	struct page *page = virt_to_page(table);
-
-	spin_lock_bh(&mm->context.list_lock);
-	list_del(&page->lru);
-	spin_unlock_bh(&mm->context.list_lock);
-	__crst_table_free(mm, table);
-}
-
-void crst_table_free_rcu(struct mm_struct *mm, unsigned long *table)
+void crst_table_free(unsigned long *table)
 {
-	struct rcu_table_freelist *batch;
 	struct page *page = virt_to_page(table);
+	struct mm_struct *mm = (struct mm_struct *) page->private;
+	unsigned long *shadow = get_shadow_table(table);
 
 	spin_lock_bh(&mm->context.list_lock);
 	list_del(&page->lru);
+	page->private = 0;
 	spin_unlock_bh(&mm->context.list_lock);
-	if (atomic_read(&mm->mm_users) < 2 &&
-	    cpumask_equal(mm_cpumask(mm), cpumask_of(smp_processor_id()))) {
-		__crst_table_free(mm, table);
-		return;
-	}
-	batch = rcu_table_freelist_get(mm);
-	if (!batch) {
-		smp_call_function(smp_sync, NULL, 1);
-		__crst_table_free(mm, table);
-		return;
-	}
-	batch->table[--batch->crst_index] = table;
-	if (batch->pgt_index >= batch->crst_index)
-		rcu_table_freelist_finish();
+	if (shadow)
+		free_pages((unsigned long) shadow, ALLOC_ORDER);
+	free_pages((unsigned long) table, ALLOC_ORDER);
 }
 
 #ifdef CONFIG_64BIT
@@ -223,7 +121,7 @@ int crst_table_upgrade(struct mm_struct 
 	}
 	spin_unlock_bh(&mm->page_table_lock);
 	if (table)
-		crst_table_free(mm, table);
+		crst_table_free(table);
 	if (mm->context.asce_limit < limit)
 		goto repeat;
 	update_mm(mm, current);
@@ -257,7 +155,7 @@ void crst_table_downgrade(struct mm_stru
 		}
 		mm->pgd = (pgd_t *) (pgd_val(*pgd) & _REGION_ENTRY_ORIGIN);
 		mm->task_size = mm->context.asce_limit;
-		crst_table_free(mm, (unsigned long *) pgd);
+		crst_table_free((unsigned long *) pgd);
 	}
 	update_mm(mm, current);
 }
@@ -288,11 +186,7 @@ unsigned long *page_table_alloc(struct m
 			return NULL;
 		pgtable_page_ctor(page);
 		page->flags &= ~FRAG_MASK;
-		table = (unsigned long *) page_to_phys(page);
-		if (mm->context.has_pgste)
-			clear_table_pgstes(table);
-		else
-			clear_table(table, _PAGE_TYPE_EMPTY, PAGE_SIZE);
+		page->private = (unsigned long) mm;
 		spin_lock_bh(&mm->context.list_lock);
 		list_add(&page->lru, &mm->context.pgtable_list);
 	}
@@ -305,42 +199,34 @@ unsigned long *page_table_alloc(struct m
 	if ((page->flags & FRAG_MASK) == ((1UL << TABLES_PER_PAGE) - 1))
 		list_move_tail(&page->lru, &mm->context.pgtable_list);
 	spin_unlock_bh(&mm->context.list_lock);
+	clear_table(table, _PAGE_TYPE_EMPTY, PTRS_PER_PTE * sizeof(long));
+	if (mm->context.noexec)
+		clear_table(table + 256, _PAGE_TYPE_EMPTY,
+			    PTRS_PER_PTE * sizeof(long));
+	else if (mm->context.has_pgste)
+		clear_table(table + 256, 0, PTRS_PER_PTE * sizeof(long));
 	return table;
 }
 
-static void __page_table_free(struct mm_struct *mm, unsigned long *table)
+void page_table_free(unsigned long *table)
 {
-	struct page *page;
-	unsigned long bits;
-
-	bits = ((unsigned long) table) & 15;
-	table = (unsigned long *)(((unsigned long) table) ^ bits);
-	page = pfn_to_page(__pa(table) >> PAGE_SHIFT);
-	page->flags ^= bits;
-	if (!(page->flags & FRAG_MASK)) {
-		pgtable_page_dtor(page);
-		__free_page(page);
-	}
-}
-
-void page_table_free(struct mm_struct *mm, unsigned long *table)
-{
-	struct page *page;
+	struct page *page = virt_to_page(table);
+	struct mm_struct *mm = (struct mm_struct *) page->private;
 	unsigned long bits;
 
 	bits = (mm->context.noexec || mm->context.has_pgste) ? 3UL : 1UL;
 	bits <<= (__pa(table) & (PAGE_SIZE - 1)) / 256 / sizeof(unsigned long);
-	page = pfn_to_page(__pa(table) >> PAGE_SHIFT);
 	spin_lock_bh(&mm->context.list_lock);
 	page->flags ^= bits;
 	if (page->flags & FRAG_MASK) {
 		/* Page now has some free pgtable fragments. */
-		if (!list_empty(&page->lru))
-			list_move(&page->lru, &mm->context.pgtable_list);
+		list_move(&page->lru, &mm->context.pgtable_list);
 		page = NULL;
-	} else
+	} else {
 		/* All fragments of the 4K page have been freed. */
 		list_del(&page->lru);
+		page->private = 0;
+	}
 	spin_unlock_bh(&mm->context.list_lock);
 	if (page) {
 		pgtable_page_dtor(page);
@@ -348,36 +234,6 @@ void page_table_free(struct mm_struct *m
 	}
 }
 
-void page_table_free_rcu(struct mm_struct *mm, unsigned long *table)
-{
-	struct rcu_table_freelist *batch;
-	struct page *page;
-	unsigned long bits;
-
-	if (atomic_read(&mm->mm_users) < 2 &&
-	    cpumask_equal(mm_cpumask(mm), cpumask_of(smp_processor_id()))) {
-		page_table_free(mm, table);
-		return;
-	}
-	batch = rcu_table_freelist_get(mm);
-	if (!batch) {
-		smp_call_function(smp_sync, NULL, 1);
-		page_table_free(mm, table);
-		return;
-	}
-	bits = (mm->context.noexec || mm->context.has_pgste) ? 3UL : 1UL;
-	bits <<= (__pa(table) & (PAGE_SIZE - 1)) / 256 / sizeof(unsigned long);
-	page = pfn_to_page(__pa(table) >> PAGE_SHIFT);
-	spin_lock_bh(&mm->context.list_lock);
-	/* Delayed freeing with rcu prevents reuse of pgtable fragments */
-	list_del_init(&page->lru);
-	spin_unlock_bh(&mm->context.list_lock);
-	table = (unsigned long *)(((unsigned long) table) | bits);
-	batch->table[batch->pgt_index++] = table;
-	if (batch->pgt_index >= batch->crst_index)
-		rcu_table_freelist_finish();
-}
-
 void disable_noexec(struct mm_struct *mm, struct task_struct *tsk)
 {
 	struct page *page;


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 12/20] mm: Extended batches for generic mmu_gather

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-mm-extended_batches_for_generic_mmu_gather.patch --]
[-- Type: text/plain, Size: 5832 bytes --]

Instead of using a single batch (the small on-stack, or an allocated
page), try and extend the batch every time it runs out and only flush
once either the extend fails or we're done.

Requested-by: Nick Piggin <npiggin@suse.de>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 include/asm-generic/tlb.h |  129 +++++++++++++++++++++++++++++-----------------
 1 file changed, 83 insertions(+), 46 deletions(-)

Index: linux-2.6/include/asm-generic/tlb.h
===================================================================
--- linux-2.6.orig/include/asm-generic/tlb.h
+++ linux-2.6/include/asm-generic/tlb.h
@@ -19,16 +19,6 @@
 #include <asm/pgalloc.h>
 #include <asm/tlbflush.h>
 
-/*
- * For UP we don't need to worry about TLB flush
- * and page free order so much..
- */
-#ifdef CONFIG_SMP
-  #define tlb_fast_mode(tlb) ((tlb)->nr == ~0U)
-#else
-  #define tlb_fast_mode(tlb) 1
-#endif
-
 #ifdef CONFIG_HAVE_RCU_TABLE_FREE
 /*
  * Semi RCU freeing of the page directories.
@@ -78,6 +68,16 @@ extern void tlb_remove_table(struct mmu_
  */
 #define MMU_GATHER_BUNDLE	8
 
+struct mmu_gather_batch {
+	struct mmu_gather_batch	*next;
+	unsigned int		nr;
+	unsigned int		max;
+	struct page		*pages[0];
+};
+
+#define MAX_GATHER_BATCH	\
+	((PAGE_SIZE - sizeof(struct mmu_gather_batch)) / sizeof(void *))
+
 /* struct mmu_gather is an opaque type used by the mm code for passing around
  * any data needed by arch specific code for tlb_remove_page.
  */
@@ -86,22 +86,48 @@ struct mmu_gather {
 #ifdef CONFIG_HAVE_RCU_TABLE_FREE
 	struct mmu_table_batch	*batch;
 #endif
-	unsigned int		nr;	/* set to ~0U means fast mode */
-	unsigned int		max;	/* nr < max */
-	unsigned int		need_flush;/* Really unmapped some ptes? */
-	unsigned int		fullmm; /* non-zero means full mm flush */
-	struct page		**pages;
-	struct page		*local[MMU_GATHER_BUNDLE];
+	unsigned int		need_flush : 1,	/* Did free PTEs */
+				fast_mode  : 1; /* No batching   */
+
+	unsigned int		fullmm;
+
+	struct mmu_gather_batch *active;
+	struct mmu_gather_batch	local;
+	struct page		*__pages[MMU_GATHER_BUNDLE];
 };
 
-static inline void __tlb_alloc_page(struct mmu_gather *tlb)
+/*
+ * For UP we don't need to worry about TLB flush
+ * and page free order so much..
+ */
+#ifdef CONFIG_SMP
+  #define tlb_fast_mode(tlb) (tlb->fast_mode)
+#else
+  #define tlb_fast_mode(tlb) 1
+#endif
+
+static inline int tlb_next_batch(struct mmu_gather *tlb)
 {
-	unsigned long addr = __get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
+	struct mmu_gather_batch *batch;
 
-	if (addr) {
-		tlb->pages = (void *)addr;
-		tlb->max = PAGE_SIZE / sizeof(struct page *);
+	batch = tlb->active;
+	if (batch->next) {
+		tlb->active = batch->next;
+		return 1;
 	}
+
+	batch = (void *)__get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
+	if (!batch)
+		return 0;
+
+	batch->next = NULL;
+	batch->nr   = 0;
+	batch->max  = MAX_GATHER_BATCH;
+
+	tlb->active->next = batch;
+	tlb->active = batch;
+
+	return 1;
 }
 
 /* tlb_gather_mmu
@@ -114,16 +140,13 @@ tlb_gather_mmu(struct mmu_gather *tlb, s
 {
 	tlb->mm = mm;
 
-	tlb->max = ARRAY_SIZE(tlb->local);
-	tlb->pages = tlb->local;
-
-	if (num_online_cpus() > 1) {
-		tlb->nr = 0;
-		__tlb_alloc_page(tlb);
-	} else /* Use fast mode if only one CPU is online */
-		tlb->nr = ~0U;
-
-	tlb->fullmm = fullmm;
+	tlb->fullmm     = fullmm;
+	tlb->need_flush = 0;
+	tlb->fast_mode  = (num_possible_cpus() == 1);
+	tlb->local.next = NULL;
+	tlb->local.nr   = 0;
+	tlb->local.max  = ARRAY_SIZE(tlb->__pages);
+	tlb->active     = &tlb->local;
 
 #ifdef CONFIG_HAVE_RCU_TABLE_FREE
 	tlb->batch = NULL;
@@ -133,6 +156,8 @@ tlb_gather_mmu(struct mmu_gather *tlb, s
 static inline void
 tlb_flush_mmu(struct mmu_gather *tlb)
 {
+	struct mmu_gather_batch *batch;
+
 	if (!tlb->need_flush)
 		return;
 	tlb->need_flush = 0;
@@ -140,17 +165,15 @@ tlb_flush_mmu(struct mmu_gather *tlb)
 #ifdef CONFIG_HAVE_RCU_TABLE_FREE
 	tlb_table_flush(tlb);
 #endif
-	if (!tlb_fast_mode(tlb)) {
-		free_pages_and_swap_cache(tlb->pages, tlb->nr);
-		tlb->nr = 0;
-		/*
-		 * If we are using the local on-stack array of pages for MMU
-		 * gather, try allocating an off-stack array again as we have
-		 * recently freed pages.
-		 */
-		if (tlb->pages == tlb->local)
-			__tlb_alloc_page(tlb);
+
+	if (tlb_fast_mode(tlb))
+		return;
+
+	for (batch = &tlb->local; batch; batch = batch->next) {
+		free_pages_and_swap_cache(batch->pages, batch->nr);
+		batch->nr = 0;
 	}
+	tlb->active = &tlb->local;
 }
 
 /* tlb_finish_mmu
@@ -160,13 +183,18 @@ tlb_flush_mmu(struct mmu_gather *tlb)
 static inline void
 tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
 {
+	struct mmu_gather_batch *batch, *next;
+
 	tlb_flush_mmu(tlb);
 
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
 
-	if (tlb->pages != tlb->local)
-		free_pages((unsigned long)tlb->pages, 0);
+	for (batch = tlb->local.next; batch; batch = next) {
+		next = batch->next;
+		free_pages((unsigned long)batch, 0);
+	}
+	tlb->local.next = NULL;
 }
 
 /* __tlb_remove_page
@@ -177,15 +205,24 @@ tlb_finish_mmu(struct mmu_gather *tlb, u
  */
 static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
 {
+	struct mmu_gather_batch *batch;
+
 	tlb->need_flush = 1;
+
 	if (tlb_fast_mode(tlb)) {
 		free_page_and_swap_cache(page);
 		return 1; /* avoid calling tlb_flush_mmu() */
 	}
-	tlb->pages[tlb->nr++] = page;
-	VM_BUG_ON(tlb->nr > tlb->max);
 
-	return tlb->max - tlb->nr;
+	batch = tlb->active;
+	batch->pages[batch->nr++] = page;
+	VM_BUG_ON(batch->nr > batch->max);
+	if (batch->nr == batch->max) {
+		if (!tlb_next_batch(tlb))
+			return 0;
+	}
+
+	return batch->max - batch->nr;
 }
 
 /* tlb_remove_page

[PATCH 12/20] mm: Extended batches for generic mmu_gather

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-mm-extended_batches_for_generic_mmu_gather.patch --]
[-- Type: text/plain, Size: 5830 bytes --]

Instead of using a single batch (the small on-stack, or an allocated
page), try and extend the batch every time it runs out and only flush
once either the extend fails or we're done.

Requested-by: Nick Piggin <npiggin@suse.de>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 include/asm-generic/tlb.h |  129 +++++++++++++++++++++++++++++-----------------
 1 file changed, 83 insertions(+), 46 deletions(-)

Index: linux-2.6/include/asm-generic/tlb.h
===================================================================
--- linux-2.6.orig/include/asm-generic/tlb.h
+++ linux-2.6/include/asm-generic/tlb.h
@@ -19,16 +19,6 @@
 #include <asm/pgalloc.h>
 #include <asm/tlbflush.h>
 
-/*
- * For UP we don't need to worry about TLB flush
- * and page free order so much..
- */
-#ifdef CONFIG_SMP
-  #define tlb_fast_mode(tlb) ((tlb)->nr == ~0U)
-#else
-  #define tlb_fast_mode(tlb) 1
-#endif
-
 #ifdef CONFIG_HAVE_RCU_TABLE_FREE
 /*
  * Semi RCU freeing of the page directories.
@@ -78,6 +68,16 @@ extern void tlb_remove_table(struct mmu_
  */
 #define MMU_GATHER_BUNDLE	8
 
+struct mmu_gather_batch {
+	struct mmu_gather_batch	*next;
+	unsigned int		nr;
+	unsigned int		max;
+	struct page		*pages[0];
+};
+
+#define MAX_GATHER_BATCH	\
+	((PAGE_SIZE - sizeof(struct mmu_gather_batch)) / sizeof(void *))
+
 /* struct mmu_gather is an opaque type used by the mm code for passing around
  * any data needed by arch specific code for tlb_remove_page.
  */
@@ -86,22 +86,48 @@ struct mmu_gather {
 #ifdef CONFIG_HAVE_RCU_TABLE_FREE
 	struct mmu_table_batch	*batch;
 #endif
-	unsigned int		nr;	/* set to ~0U means fast mode */
-	unsigned int		max;	/* nr < max */
-	unsigned int		need_flush;/* Really unmapped some ptes? */
-	unsigned int		fullmm; /* non-zero means full mm flush */
-	struct page		**pages;
-	struct page		*local[MMU_GATHER_BUNDLE];
+	unsigned int		need_flush : 1,	/* Did free PTEs */
+				fast_mode  : 1; /* No batching   */
+
+	unsigned int		fullmm;
+
+	struct mmu_gather_batch *active;
+	struct mmu_gather_batch	local;
+	struct page		*__pages[MMU_GATHER_BUNDLE];
 };
 
-static inline void __tlb_alloc_page(struct mmu_gather *tlb)
+/*
+ * For UP we don't need to worry about TLB flush
+ * and page free order so much..
+ */
+#ifdef CONFIG_SMP
+  #define tlb_fast_mode(tlb) (tlb->fast_mode)
+#else
+  #define tlb_fast_mode(tlb) 1
+#endif
+
+static inline int tlb_next_batch(struct mmu_gather *tlb)
 {
-	unsigned long addr = __get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
+	struct mmu_gather_batch *batch;
 
-	if (addr) {
-		tlb->pages = (void *)addr;
-		tlb->max = PAGE_SIZE / sizeof(struct page *);
+	batch = tlb->active;
+	if (batch->next) {
+		tlb->active = batch->next;
+		return 1;
 	}
+
+	batch = (void *)__get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
+	if (!batch)
+		return 0;
+
+	batch->next = NULL;
+	batch->nr   = 0;
+	batch->max  = MAX_GATHER_BATCH;
+
+	tlb->active->next = batch;
+	tlb->active = batch;
+
+	return 1;
 }
 
 /* tlb_gather_mmu
@@ -114,16 +140,13 @@ tlb_gather_mmu(struct mmu_gather *tlb, s
 {
 	tlb->mm = mm;
 
-	tlb->max = ARRAY_SIZE(tlb->local);
-	tlb->pages = tlb->local;
-
-	if (num_online_cpus() > 1) {
-		tlb->nr = 0;
-		__tlb_alloc_page(tlb);
-	} else /* Use fast mode if only one CPU is online */
-		tlb->nr = ~0U;
-
-	tlb->fullmm = fullmm;
+	tlb->fullmm     = fullmm;
+	tlb->need_flush = 0;
+	tlb->fast_mode  = (num_possible_cpus() == 1);
+	tlb->local.next = NULL;
+	tlb->local.nr   = 0;
+	tlb->local.max  = ARRAY_SIZE(tlb->__pages);
+	tlb->active     = &tlb->local;
 
 #ifdef CONFIG_HAVE_RCU_TABLE_FREE
 	tlb->batch = NULL;
@@ -133,6 +156,8 @@ tlb_gather_mmu(struct mmu_gather *tlb, s
 static inline void
 tlb_flush_mmu(struct mmu_gather *tlb)
 {
+	struct mmu_gather_batch *batch;
+
 	if (!tlb->need_flush)
 		return;
 	tlb->need_flush = 0;
@@ -140,17 +165,15 @@ tlb_flush_mmu(struct mmu_gather *tlb)
 #ifdef CONFIG_HAVE_RCU_TABLE_FREE
 	tlb_table_flush(tlb);
 #endif
-	if (!tlb_fast_mode(tlb)) {
-		free_pages_and_swap_cache(tlb->pages, tlb->nr);
-		tlb->nr = 0;
-		/*
-		 * If we are using the local on-stack array of pages for MMU
-		 * gather, try allocating an off-stack array again as we have
-		 * recently freed pages.
-		 */
-		if (tlb->pages == tlb->local)
-			__tlb_alloc_page(tlb);
+
+	if (tlb_fast_mode(tlb))
+		return;
+
+	for (batch = &tlb->local; batch; batch = batch->next) {
+		free_pages_and_swap_cache(batch->pages, batch->nr);
+		batch->nr = 0;
 	}
+	tlb->active = &tlb->local;
 }
 
 /* tlb_finish_mmu
@@ -160,13 +183,18 @@ tlb_flush_mmu(struct mmu_gather *tlb)
 static inline void
 tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
 {
+	struct mmu_gather_batch *batch, *next;
+
 	tlb_flush_mmu(tlb);
 
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
 
-	if (tlb->pages != tlb->local)
-		free_pages((unsigned long)tlb->pages, 0);
+	for (batch = tlb->local.next; batch; batch = next) {
+		next = batch->next;
+		free_pages((unsigned long)batch, 0);
+	}
+	tlb->local.next = NULL;
 }
 
 /* __tlb_remove_page
@@ -177,15 +205,24 @@ tlb_finish_mmu(struct mmu_gather *tlb, u
  */
 static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
 {
+	struct mmu_gather_batch *batch;
+
 	tlb->need_flush = 1;
+
 	if (tlb_fast_mode(tlb)) {
 		free_page_and_swap_cache(page);
 		return 1; /* avoid calling tlb_flush_mmu() */
 	}
-	tlb->pages[tlb->nr++] = page;
-	VM_BUG_ON(tlb->nr > tlb->max);
 
-	return tlb->max - tlb->nr;
+	batch = tlb->active;
+	batch->pages[batch->nr++] = page;
+	VM_BUG_ON(batch->nr > batch->max);
+	if (batch->nr == batch->max) {
+		if (!tlb_next_batch(tlb))
+			return 0;
+	}
+
+	return batch->max - batch->nr;
 }
 
 /* tlb_remove_page

[PATCH 12/20] mm: Extended batches for generic mmu_gather

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-mm-extended_batches_for_generic_mmu_gather.patch --]
[-- Type: text/plain, Size: 6135 bytes --]

Instead of using a single batch (the small on-stack, or an allocated
page), try and extend the batch every time it runs out and only flush
once either the extend fails or we're done.

Requested-by: Nick Piggin <npiggin@suse.de>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 include/asm-generic/tlb.h |  129 +++++++++++++++++++++++++++++-----------------
 1 file changed, 83 insertions(+), 46 deletions(-)

Index: linux-2.6/include/asm-generic/tlb.h
===================================================================
--- linux-2.6.orig/include/asm-generic/tlb.h
+++ linux-2.6/include/asm-generic/tlb.h
@@ -19,16 +19,6 @@
 #include <asm/pgalloc.h>
 #include <asm/tlbflush.h>
 
-/*
- * For UP we don't need to worry about TLB flush
- * and page free order so much..
- */
-#ifdef CONFIG_SMP
-  #define tlb_fast_mode(tlb) ((tlb)->nr == ~0U)
-#else
-  #define tlb_fast_mode(tlb) 1
-#endif
-
 #ifdef CONFIG_HAVE_RCU_TABLE_FREE
 /*
  * Semi RCU freeing of the page directories.
@@ -78,6 +68,16 @@ extern void tlb_remove_table(struct mmu_
  */
 #define MMU_GATHER_BUNDLE	8
 
+struct mmu_gather_batch {
+	struct mmu_gather_batch	*next;
+	unsigned int		nr;
+	unsigned int		max;
+	struct page		*pages[0];
+};
+
+#define MAX_GATHER_BATCH	\
+	((PAGE_SIZE - sizeof(struct mmu_gather_batch)) / sizeof(void *))
+
 /* struct mmu_gather is an opaque type used by the mm code for passing around
  * any data needed by arch specific code for tlb_remove_page.
  */
@@ -86,22 +86,48 @@ struct mmu_gather {
 #ifdef CONFIG_HAVE_RCU_TABLE_FREE
 	struct mmu_table_batch	*batch;
 #endif
-	unsigned int		nr;	/* set to ~0U means fast mode */
-	unsigned int		max;	/* nr < max */
-	unsigned int		need_flush;/* Really unmapped some ptes? */
-	unsigned int		fullmm; /* non-zero means full mm flush */
-	struct page		**pages;
-	struct page		*local[MMU_GATHER_BUNDLE];
+	unsigned int		need_flush : 1,	/* Did free PTEs */
+				fast_mode  : 1; /* No batching   */
+
+	unsigned int		fullmm;
+
+	struct mmu_gather_batch *active;
+	struct mmu_gather_batch	local;
+	struct page		*__pages[MMU_GATHER_BUNDLE];
 };
 
-static inline void __tlb_alloc_page(struct mmu_gather *tlb)
+/*
+ * For UP we don't need to worry about TLB flush
+ * and page free order so much..
+ */
+#ifdef CONFIG_SMP
+  #define tlb_fast_mode(tlb) (tlb->fast_mode)
+#else
+  #define tlb_fast_mode(tlb) 1
+#endif
+
+static inline int tlb_next_batch(struct mmu_gather *tlb)
 {
-	unsigned long addr = __get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
+	struct mmu_gather_batch *batch;
 
-	if (addr) {
-		tlb->pages = (void *)addr;
-		tlb->max = PAGE_SIZE / sizeof(struct page *);
+	batch = tlb->active;
+	if (batch->next) {
+		tlb->active = batch->next;
+		return 1;
 	}
+
+	batch = (void *)__get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
+	if (!batch)
+		return 0;
+
+	batch->next = NULL;
+	batch->nr   = 0;
+	batch->max  = MAX_GATHER_BATCH;
+
+	tlb->active->next = batch;
+	tlb->active = batch;
+
+	return 1;
 }
 
 /* tlb_gather_mmu
@@ -114,16 +140,13 @@ tlb_gather_mmu(struct mmu_gather *tlb, s
 {
 	tlb->mm = mm;
 
-	tlb->max = ARRAY_SIZE(tlb->local);
-	tlb->pages = tlb->local;
-
-	if (num_online_cpus() > 1) {
-		tlb->nr = 0;
-		__tlb_alloc_page(tlb);
-	} else /* Use fast mode if only one CPU is online */
-		tlb->nr = ~0U;
-
-	tlb->fullmm = fullmm;
+	tlb->fullmm     = fullmm;
+	tlb->need_flush = 0;
+	tlb->fast_mode  = (num_possible_cpus() == 1);
+	tlb->local.next = NULL;
+	tlb->local.nr   = 0;
+	tlb->local.max  = ARRAY_SIZE(tlb->__pages);
+	tlb->active     = &tlb->local;
 
 #ifdef CONFIG_HAVE_RCU_TABLE_FREE
 	tlb->batch = NULL;
@@ -133,6 +156,8 @@ tlb_gather_mmu(struct mmu_gather *tlb, s
 static inline void
 tlb_flush_mmu(struct mmu_gather *tlb)
 {
+	struct mmu_gather_batch *batch;
+
 	if (!tlb->need_flush)
 		return;
 	tlb->need_flush = 0;
@@ -140,17 +165,15 @@ tlb_flush_mmu(struct mmu_gather *tlb)
 #ifdef CONFIG_HAVE_RCU_TABLE_FREE
 	tlb_table_flush(tlb);
 #endif
-	if (!tlb_fast_mode(tlb)) {
-		free_pages_and_swap_cache(tlb->pages, tlb->nr);
-		tlb->nr = 0;
-		/*
-		 * If we are using the local on-stack array of pages for MMU
-		 * gather, try allocating an off-stack array again as we have
-		 * recently freed pages.
-		 */
-		if (tlb->pages == tlb->local)
-			__tlb_alloc_page(tlb);
+
+	if (tlb_fast_mode(tlb))
+		return;
+
+	for (batch = &tlb->local; batch; batch = batch->next) {
+		free_pages_and_swap_cache(batch->pages, batch->nr);
+		batch->nr = 0;
 	}
+	tlb->active = &tlb->local;
 }
 
 /* tlb_finish_mmu
@@ -160,13 +183,18 @@ tlb_flush_mmu(struct mmu_gather *tlb)
 static inline void
 tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
 {
+	struct mmu_gather_batch *batch, *next;
+
 	tlb_flush_mmu(tlb);
 
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
 
-	if (tlb->pages != tlb->local)
-		free_pages((unsigned long)tlb->pages, 0);
+	for (batch = tlb->local.next; batch; batch = next) {
+		next = batch->next;
+		free_pages((unsigned long)batch, 0);
+	}
+	tlb->local.next = NULL;
 }
 
 /* __tlb_remove_page
@@ -177,15 +205,24 @@ tlb_finish_mmu(struct mmu_gather *tlb, u
  */
 static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
 {
+	struct mmu_gather_batch *batch;
+
 	tlb->need_flush = 1;
+
 	if (tlb_fast_mode(tlb)) {
 		free_page_and_swap_cache(page);
 		return 1; /* avoid calling tlb_flush_mmu() */
 	}
-	tlb->pages[tlb->nr++] = page;
-	VM_BUG_ON(tlb->nr > tlb->max);
 
-	return tlb->max - tlb->nr;
+	batch = tlb->active;
+	batch->pages[batch->nr++] = page;
+	VM_BUG_ON(batch->nr > batch->max);
+	if (batch->nr == batch->max) {
+		if (!tlb_next_batch(tlb))
+			return 0;
+	}
+
+	return batch->max - batch->nr;
 }
 
 /* tlb_remove_page


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 13/20] lockdep, mutex: Provide mutex_lock_nest_lock

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-lockdep_mutex-provide_mutex_lock_nest_lock.patch --]
[-- Type: text/plain, Size: 5475 bytes --]

Provide the mutex_lock_nest_lock() annotation.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 include/linux/lockdep.h |    3 +++
 include/linux/mutex.h   |    9 +++++++++
 kernel/mutex.c          |   25 +++++++++++++++++--------
 3 files changed, 29 insertions(+), 8 deletions(-)

Index: linux-2.6/include/linux/lockdep.h
===================================================================
--- linux-2.6.orig/include/linux/lockdep.h
+++ linux-2.6/include/linux/lockdep.h
@@ -495,12 +495,15 @@ static inline void print_irqtrace_events
 #ifdef CONFIG_DEBUG_LOCK_ALLOC
 # ifdef CONFIG_PROVE_LOCKING
 #  define mutex_acquire(l, s, t, i)		lock_acquire(l, s, t, 0, 2, NULL, i)
+#  define mutex_acquire_nest(l, s, t, n, i)	lock_acquire(l, s, t, 0, 2, n, i)
 # else
 #  define mutex_acquire(l, s, t, i)		lock_acquire(l, s, t, 0, 1, NULL, i)
+#  define mutex_acquire_nest(l, s, t, n, i)	lock_acquire(l, s, t, 0, 1, n, i)
 # endif
 # define mutex_release(l, n, i)			lock_release(l, n, i)
 #else
 # define mutex_acquire(l, s, t, i)		do { } while (0)
+# define mutex_acquire_nest(l, s, t, n, i)	do { } while (0)
 # define mutex_release(l, n, i)			do { } while (0)
 #endif
 
Index: linux-2.6/include/linux/mutex.h
===================================================================
--- linux-2.6.orig/include/linux/mutex.h
+++ linux-2.6/include/linux/mutex.h
@@ -132,6 +132,7 @@ static inline int mutex_is_locked(struct
  */
 #ifdef CONFIG_DEBUG_LOCK_ALLOC
 extern void mutex_lock_nested(struct mutex *lock, unsigned int subclass);
+extern void _mutex_lock_nest_lock(struct mutex *lock, struct lockdep_map *nest_lock);
 extern int __must_check mutex_lock_interruptible_nested(struct mutex *lock,
 					unsigned int subclass);
 extern int __must_check mutex_lock_killable_nested(struct mutex *lock,
@@ -140,6 +141,13 @@ extern int __must_check mutex_lock_killa
 #define mutex_lock(lock) mutex_lock_nested(lock, 0)
 #define mutex_lock_interruptible(lock) mutex_lock_interruptible_nested(lock, 0)
 #define mutex_lock_killable(lock) mutex_lock_killable_nested(lock, 0)
+
+#define mutex_lock_nest_lock(lock, nest_lock)				\
+do {									\
+	typecheck(struct lockdep_map *, &(nest_lock)->dep_map);		\
+	_mutex_lock_nest_lock(lock, &(nest_lock)->dep_map);		\
+} while (0)
+
 #else
 extern void mutex_lock(struct mutex *lock);
 extern int __must_check mutex_lock_interruptible(struct mutex *lock);
@@ -148,6 +156,7 @@ extern int __must_check mutex_lock_killa
 # define mutex_lock_nested(lock, subclass) mutex_lock(lock)
 # define mutex_lock_interruptible_nested(lock, subclass) mutex_lock_interruptible(lock)
 # define mutex_lock_killable_nested(lock, subclass) mutex_lock_killable(lock)
+# define mutex_lock_nest_lock(lock, nest_lock) mutex_lock(lock)
 #endif
 
 /*
Index: linux-2.6/kernel/mutex.c
===================================================================
--- linux-2.6.orig/kernel/mutex.c
+++ linux-2.6/kernel/mutex.c
@@ -131,14 +131,14 @@ EXPORT_SYMBOL(mutex_unlock);
  */
 static inline int __sched
 __mutex_lock_common(struct mutex *lock, long state, unsigned int subclass,
-	       	unsigned long ip)
+		    struct lockdep_map *nest_lock, unsigned long ip)
 {
 	struct task_struct *task = current;
 	struct mutex_waiter waiter;
 	unsigned long flags;
 
 	preempt_disable();
-	mutex_acquire(&lock->dep_map, subclass, 0, ip);
+	mutex_acquire_nest(&lock->dep_map, subclass, 0, nest_lock, ip);
 
 #ifdef CONFIG_MUTEX_SPIN_ON_OWNER
 	/*
@@ -276,16 +276,25 @@ void __sched
 mutex_lock_nested(struct mutex *lock, unsigned int subclass)
 {
 	might_sleep();
-	__mutex_lock_common(lock, TASK_UNINTERRUPTIBLE, subclass, _RET_IP_);
+	__mutex_lock_common(lock, TASK_UNINTERRUPTIBLE, subclass, NULL, _RET_IP_);
 }
 
 EXPORT_SYMBOL_GPL(mutex_lock_nested);
 
+void __sched
+_mutex_lock_nest_lock(struct mutex *lock, struct lockdep_map *nest)
+{
+	might_sleep();
+	__mutex_lock_common(lock, TASK_UNINTERRUPTIBLE, 0, nest, _RET_IP_);
+}
+
+EXPORT_SYMBOL_GPL(_mutex_lock_nest_lock);
+
 int __sched
 mutex_lock_killable_nested(struct mutex *lock, unsigned int subclass)
 {
 	might_sleep();
-	return __mutex_lock_common(lock, TASK_KILLABLE, subclass, _RET_IP_);
+	return __mutex_lock_common(lock, TASK_KILLABLE, subclass, NULL, _RET_IP_);
 }
 EXPORT_SYMBOL_GPL(mutex_lock_killable_nested);
 
@@ -294,7 +303,7 @@ mutex_lock_interruptible_nested(struct m
 {
 	might_sleep();
 	return __mutex_lock_common(lock, TASK_INTERRUPTIBLE,
-				   subclass, _RET_IP_);
+				   subclass, NULL, _RET_IP_);
 }
 
 EXPORT_SYMBOL_GPL(mutex_lock_interruptible_nested);
@@ -400,7 +409,7 @@ __mutex_lock_slowpath(atomic_t *lock_cou
 {
 	struct mutex *lock = container_of(lock_count, struct mutex, count);
 
-	__mutex_lock_common(lock, TASK_UNINTERRUPTIBLE, 0, _RET_IP_);
+	__mutex_lock_common(lock, TASK_UNINTERRUPTIBLE, 0, NULL, _RET_IP_);
 }
 
 static noinline int __sched
@@ -408,7 +417,7 @@ __mutex_lock_killable_slowpath(atomic_t 
 {
 	struct mutex *lock = container_of(lock_count, struct mutex, count);
 
-	return __mutex_lock_common(lock, TASK_KILLABLE, 0, _RET_IP_);
+	return __mutex_lock_common(lock, TASK_KILLABLE, 0, NULL, _RET_IP_);
 }
 
 static noinline int __sched
@@ -416,7 +425,7 @@ __mutex_lock_interruptible_slowpath(atom
 {
 	struct mutex *lock = container_of(lock_count, struct mutex, count);
 
-	return __mutex_lock_common(lock, TASK_INTERRUPTIBLE, 0, _RET_IP_);
+	return __mutex_lock_common(lock, TASK_INTERRUPTIBLE, 0, NULL, _RET_IP_);
 }
 #endif
 

[PATCH 13/20] lockdep, mutex: Provide mutex_lock_nest_lock

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-lockdep_mutex-provide_mutex_lock_nest_lock.patch --]
[-- Type: text/plain, Size: 5778 bytes --]

Provide the mutex_lock_nest_lock() annotation.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 include/linux/lockdep.h |    3 +++
 include/linux/mutex.h   |    9 +++++++++
 kernel/mutex.c          |   25 +++++++++++++++++--------
 3 files changed, 29 insertions(+), 8 deletions(-)

Index: linux-2.6/include/linux/lockdep.h
===================================================================
--- linux-2.6.orig/include/linux/lockdep.h
+++ linux-2.6/include/linux/lockdep.h
@@ -495,12 +495,15 @@ static inline void print_irqtrace_events
 #ifdef CONFIG_DEBUG_LOCK_ALLOC
 # ifdef CONFIG_PROVE_LOCKING
 #  define mutex_acquire(l, s, t, i)		lock_acquire(l, s, t, 0, 2, NULL, i)
+#  define mutex_acquire_nest(l, s, t, n, i)	lock_acquire(l, s, t, 0, 2, n, i)
 # else
 #  define mutex_acquire(l, s, t, i)		lock_acquire(l, s, t, 0, 1, NULL, i)
+#  define mutex_acquire_nest(l, s, t, n, i)	lock_acquire(l, s, t, 0, 1, n, i)
 # endif
 # define mutex_release(l, n, i)			lock_release(l, n, i)
 #else
 # define mutex_acquire(l, s, t, i)		do { } while (0)
+# define mutex_acquire_nest(l, s, t, n, i)	do { } while (0)
 # define mutex_release(l, n, i)			do { } while (0)
 #endif
 
Index: linux-2.6/include/linux/mutex.h
===================================================================
--- linux-2.6.orig/include/linux/mutex.h
+++ linux-2.6/include/linux/mutex.h
@@ -132,6 +132,7 @@ static inline int mutex_is_locked(struct
  */
 #ifdef CONFIG_DEBUG_LOCK_ALLOC
 extern void mutex_lock_nested(struct mutex *lock, unsigned int subclass);
+extern void _mutex_lock_nest_lock(struct mutex *lock, struct lockdep_map *nest_lock);
 extern int __must_check mutex_lock_interruptible_nested(struct mutex *lock,
 					unsigned int subclass);
 extern int __must_check mutex_lock_killable_nested(struct mutex *lock,
@@ -140,6 +141,13 @@ extern int __must_check mutex_lock_killa
 #define mutex_lock(lock) mutex_lock_nested(lock, 0)
 #define mutex_lock_interruptible(lock) mutex_lock_interruptible_nested(lock, 0)
 #define mutex_lock_killable(lock) mutex_lock_killable_nested(lock, 0)
+
+#define mutex_lock_nest_lock(lock, nest_lock)				\
+do {									\
+	typecheck(struct lockdep_map *, &(nest_lock)->dep_map);		\
+	_mutex_lock_nest_lock(lock, &(nest_lock)->dep_map);		\
+} while (0)
+
 #else
 extern void mutex_lock(struct mutex *lock);
 extern int __must_check mutex_lock_interruptible(struct mutex *lock);
@@ -148,6 +156,7 @@ extern int __must_check mutex_lock_killa
 # define mutex_lock_nested(lock, subclass) mutex_lock(lock)
 # define mutex_lock_interruptible_nested(lock, subclass) mutex_lock_interruptible(lock)
 # define mutex_lock_killable_nested(lock, subclass) mutex_lock_killable(lock)
+# define mutex_lock_nest_lock(lock, nest_lock) mutex_lock(lock)
 #endif
 
 /*
Index: linux-2.6/kernel/mutex.c
===================================================================
--- linux-2.6.orig/kernel/mutex.c
+++ linux-2.6/kernel/mutex.c
@@ -131,14 +131,14 @@ EXPORT_SYMBOL(mutex_unlock);
  */
 static inline int __sched
 __mutex_lock_common(struct mutex *lock, long state, unsigned int subclass,
-	       	unsigned long ip)
+		    struct lockdep_map *nest_lock, unsigned long ip)
 {
 	struct task_struct *task = current;
 	struct mutex_waiter waiter;
 	unsigned long flags;
 
 	preempt_disable();
-	mutex_acquire(&lock->dep_map, subclass, 0, ip);
+	mutex_acquire_nest(&lock->dep_map, subclass, 0, nest_lock, ip);
 
 #ifdef CONFIG_MUTEX_SPIN_ON_OWNER
 	/*
@@ -276,16 +276,25 @@ void __sched
 mutex_lock_nested(struct mutex *lock, unsigned int subclass)
 {
 	might_sleep();
-	__mutex_lock_common(lock, TASK_UNINTERRUPTIBLE, subclass, _RET_IP_);
+	__mutex_lock_common(lock, TASK_UNINTERRUPTIBLE, subclass, NULL, _RET_IP_);
 }
 
 EXPORT_SYMBOL_GPL(mutex_lock_nested);
 
+void __sched
+_mutex_lock_nest_lock(struct mutex *lock, struct lockdep_map *nest)
+{
+	might_sleep();
+	__mutex_lock_common(lock, TASK_UNINTERRUPTIBLE, 0, nest, _RET_IP_);
+}
+
+EXPORT_SYMBOL_GPL(_mutex_lock_nest_lock);
+
 int __sched
 mutex_lock_killable_nested(struct mutex *lock, unsigned int subclass)
 {
 	might_sleep();
-	return __mutex_lock_common(lock, TASK_KILLABLE, subclass, _RET_IP_);
+	return __mutex_lock_common(lock, TASK_KILLABLE, subclass, NULL, _RET_IP_);
 }
 EXPORT_SYMBOL_GPL(mutex_lock_killable_nested);
 
@@ -294,7 +303,7 @@ mutex_lock_interruptible_nested(struct m
 {
 	might_sleep();
 	return __mutex_lock_common(lock, TASK_INTERRUPTIBLE,
-				   subclass, _RET_IP_);
+				   subclass, NULL, _RET_IP_);
 }
 
 EXPORT_SYMBOL_GPL(mutex_lock_interruptible_nested);
@@ -400,7 +409,7 @@ __mutex_lock_slowpath(atomic_t *lock_cou
 {
 	struct mutex *lock = container_of(lock_count, struct mutex, count);
 
-	__mutex_lock_common(lock, TASK_UNINTERRUPTIBLE, 0, _RET_IP_);
+	__mutex_lock_common(lock, TASK_UNINTERRUPTIBLE, 0, NULL, _RET_IP_);
 }
 
 static noinline int __sched
@@ -408,7 +417,7 @@ __mutex_lock_killable_slowpath(atomic_t 
 {
 	struct mutex *lock = container_of(lock_count, struct mutex, count);
 
-	return __mutex_lock_common(lock, TASK_KILLABLE, 0, _RET_IP_);
+	return __mutex_lock_common(lock, TASK_KILLABLE, 0, NULL, _RET_IP_);
 }
 
 static noinline int __sched
@@ -416,7 +425,7 @@ __mutex_lock_interruptible_slowpath(atom
 {
 	struct mutex *lock = container_of(lock_count, struct mutex, count);
 
-	return __mutex_lock_common(lock, TASK_INTERRUPTIBLE, 0, _RET_IP_);
+	return __mutex_lock_common(lock, TASK_INTERRUPTIBLE, 0, NULL, _RET_IP_);
 }
 #endif
 


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 13/20] lockdep, mutex: Provide mutex_lock_nest_lock

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-lockdep_mutex-provide_mutex_lock_nest_lock.patch --]
[-- Type: text/plain, Size: 5778 bytes --]

Provide the mutex_lock_nest_lock() annotation.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 include/linux/lockdep.h |    3 +++
 include/linux/mutex.h   |    9 +++++++++
 kernel/mutex.c          |   25 +++++++++++++++++--------
 3 files changed, 29 insertions(+), 8 deletions(-)

Index: linux-2.6/include/linux/lockdep.h
===================================================================
--- linux-2.6.orig/include/linux/lockdep.h
+++ linux-2.6/include/linux/lockdep.h
@@ -495,12 +495,15 @@ static inline void print_irqtrace_events
 #ifdef CONFIG_DEBUG_LOCK_ALLOC
 # ifdef CONFIG_PROVE_LOCKING
 #  define mutex_acquire(l, s, t, i)		lock_acquire(l, s, t, 0, 2, NULL, i)
+#  define mutex_acquire_nest(l, s, t, n, i)	lock_acquire(l, s, t, 0, 2, n, i)
 # else
 #  define mutex_acquire(l, s, t, i)		lock_acquire(l, s, t, 0, 1, NULL, i)
+#  define mutex_acquire_nest(l, s, t, n, i)	lock_acquire(l, s, t, 0, 1, n, i)
 # endif
 # define mutex_release(l, n, i)			lock_release(l, n, i)
 #else
 # define mutex_acquire(l, s, t, i)		do { } while (0)
+# define mutex_acquire_nest(l, s, t, n, i)	do { } while (0)
 # define mutex_release(l, n, i)			do { } while (0)
 #endif
 
Index: linux-2.6/include/linux/mutex.h
===================================================================
--- linux-2.6.orig/include/linux/mutex.h
+++ linux-2.6/include/linux/mutex.h
@@ -132,6 +132,7 @@ static inline int mutex_is_locked(struct
  */
 #ifdef CONFIG_DEBUG_LOCK_ALLOC
 extern void mutex_lock_nested(struct mutex *lock, unsigned int subclass);
+extern void _mutex_lock_nest_lock(struct mutex *lock, struct lockdep_map *nest_lock);
 extern int __must_check mutex_lock_interruptible_nested(struct mutex *lock,
 					unsigned int subclass);
 extern int __must_check mutex_lock_killable_nested(struct mutex *lock,
@@ -140,6 +141,13 @@ extern int __must_check mutex_lock_killa
 #define mutex_lock(lock) mutex_lock_nested(lock, 0)
 #define mutex_lock_interruptible(lock) mutex_lock_interruptible_nested(lock, 0)
 #define mutex_lock_killable(lock) mutex_lock_killable_nested(lock, 0)
+
+#define mutex_lock_nest_lock(lock, nest_lock)				\
+do {									\
+	typecheck(struct lockdep_map *, &(nest_lock)->dep_map);		\
+	_mutex_lock_nest_lock(lock, &(nest_lock)->dep_map);		\
+} while (0)
+
 #else
 extern void mutex_lock(struct mutex *lock);
 extern int __must_check mutex_lock_interruptible(struct mutex *lock);
@@ -148,6 +156,7 @@ extern int __must_check mutex_lock_killa
 # define mutex_lock_nested(lock, subclass) mutex_lock(lock)
 # define mutex_lock_interruptible_nested(lock, subclass) mutex_lock_interruptible(lock)
 # define mutex_lock_killable_nested(lock, subclass) mutex_lock_killable(lock)
+# define mutex_lock_nest_lock(lock, nest_lock) mutex_lock(lock)
 #endif
 
 /*
Index: linux-2.6/kernel/mutex.c
===================================================================
--- linux-2.6.orig/kernel/mutex.c
+++ linux-2.6/kernel/mutex.c
@@ -131,14 +131,14 @@ EXPORT_SYMBOL(mutex_unlock);
  */
 static inline int __sched
 __mutex_lock_common(struct mutex *lock, long state, unsigned int subclass,
-	       	unsigned long ip)
+		    struct lockdep_map *nest_lock, unsigned long ip)
 {
 	struct task_struct *task = current;
 	struct mutex_waiter waiter;
 	unsigned long flags;
 
 	preempt_disable();
-	mutex_acquire(&lock->dep_map, subclass, 0, ip);
+	mutex_acquire_nest(&lock->dep_map, subclass, 0, nest_lock, ip);
 
 #ifdef CONFIG_MUTEX_SPIN_ON_OWNER
 	/*
@@ -276,16 +276,25 @@ void __sched
 mutex_lock_nested(struct mutex *lock, unsigned int subclass)
 {
 	might_sleep();
-	__mutex_lock_common(lock, TASK_UNINTERRUPTIBLE, subclass, _RET_IP_);
+	__mutex_lock_common(lock, TASK_UNINTERRUPTIBLE, subclass, NULL, _RET_IP_);
 }
 
 EXPORT_SYMBOL_GPL(mutex_lock_nested);
 
+void __sched
+_mutex_lock_nest_lock(struct mutex *lock, struct lockdep_map *nest)
+{
+	might_sleep();
+	__mutex_lock_common(lock, TASK_UNINTERRUPTIBLE, 0, nest, _RET_IP_);
+}
+
+EXPORT_SYMBOL_GPL(_mutex_lock_nest_lock);
+
 int __sched
 mutex_lock_killable_nested(struct mutex *lock, unsigned int subclass)
 {
 	might_sleep();
-	return __mutex_lock_common(lock, TASK_KILLABLE, subclass, _RET_IP_);
+	return __mutex_lock_common(lock, TASK_KILLABLE, subclass, NULL, _RET_IP_);
 }
 EXPORT_SYMBOL_GPL(mutex_lock_killable_nested);
 
@@ -294,7 +303,7 @@ mutex_lock_interruptible_nested(struct m
 {
 	might_sleep();
 	return __mutex_lock_common(lock, TASK_INTERRUPTIBLE,
-				   subclass, _RET_IP_);
+				   subclass, NULL, _RET_IP_);
 }
 
 EXPORT_SYMBOL_GPL(mutex_lock_interruptible_nested);
@@ -400,7 +409,7 @@ __mutex_lock_slowpath(atomic_t *lock_cou
 {
 	struct mutex *lock = container_of(lock_count, struct mutex, count);
 
-	__mutex_lock_common(lock, TASK_UNINTERRUPTIBLE, 0, _RET_IP_);
+	__mutex_lock_common(lock, TASK_UNINTERRUPTIBLE, 0, NULL, _RET_IP_);
 }
 
 static noinline int __sched
@@ -408,7 +417,7 @@ __mutex_lock_killable_slowpath(atomic_t 
 {
 	struct mutex *lock = container_of(lock_count, struct mutex, count);
 
-	return __mutex_lock_common(lock, TASK_KILLABLE, 0, _RET_IP_);
+	return __mutex_lock_common(lock, TASK_KILLABLE, 0, NULL, _RET_IP_);
 }
 
 static noinline int __sched
@@ -416,7 +425,7 @@ __mutex_lock_interruptible_slowpath(atom
 {
 	struct mutex *lock = container_of(lock_count, struct mutex, count);
 
-	return __mutex_lock_common(lock, TASK_INTERRUPTIBLE, 0, _RET_IP_);
+	return __mutex_lock_common(lock, TASK_INTERRUPTIBLE, 0, NULL, _RET_IP_);
 }
 #endif
 


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 14/20] mm: Remove i_mmap_lock lockbreak

[Peter Zijlstra]

[-- Attachment #1: mm-fix-zap_block_size.patch --]
[-- Type: text/plain, Size: 17318 bytes --]

Hugh says:
 "The only significant loser, I think, would be page reclaim (when
  concurrent with truncation): could spin for a long time waiting for
  the i_mmap_mutex it expects would soon be dropped? "

Counter points:
 - cpu contention makes the spin stop (need_resched())
 - zap pages should be freeing pages at a higher rate than reclaim
   ever can

I think the simplification of the truncate code is definately worth it.

Effectively reverts: 2aa15890f3c (mm: prevent concurrent
unmap_mapping_range() on the same inode) and takes out the code that
caused its problem.

Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 fs/inode.c               |    1 
 include/linux/fs.h       |    2 
 include/linux/mm.h       |    2 
 include/linux/mm_types.h |    1 
 kernel/fork.c            |    1 
 mm/memory.c              |  195 ++++++-----------------------------------------
 mm/mmap.c                |   13 ---
 mm/mremap.c              |    1 
 8 files changed, 28 insertions(+), 188 deletions(-)

Index: linux-2.6/include/linux/mm.h
===================================================================
--- linux-2.6.orig/include/linux/mm.h
+++ linux-2.6/include/linux/mm.h
@@ -878,8 +878,6 @@ struct zap_details {
 	struct address_space *check_mapping;	/* Check page->mapping if set */
 	pgoff_t	first_index;			/* Lowest page->index to unmap */
 	pgoff_t last_index;			/* Highest page->index to unmap */
-	spinlock_t *i_mmap_lock;		/* For unmap_mapping_range: */
-	unsigned long truncate_count;		/* Compare vm_truncate_count */
 };
 
 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
Index: linux-2.6/mm/memory.c
===================================================================
--- linux-2.6.orig/mm/memory.c
+++ linux-2.6/mm/memory.c
@@ -986,13 +986,13 @@ int copy_page_range(struct mm_struct *ds
 static unsigned long zap_pte_range(struct mmu_gather *tlb,
 				struct vm_area_struct *vma, pmd_t *pmd,
 				unsigned long addr, unsigned long end,
-				long *zap_work, struct zap_details *details)
+				struct zap_details *details)
 {
 	struct mm_struct *mm = tlb->mm;
 	int force_flush = 0;
-	pte_t *pte;
-	spinlock_t *ptl;
 	int rss[NR_MM_COUNTERS];
+	spinlock_t *ptl;
+	pte_t *pte;
 
 	init_rss_vec(rss);
 again:
@@ -1001,12 +1001,9 @@ static unsigned long zap_pte_range(struc
 	do {
 		pte_t ptent = *pte;
 		if (pte_none(ptent)) {
-			(*zap_work)--;
 			continue;
 		}
 
-		(*zap_work) -= PAGE_SIZE;
-
 		if (pte_present(ptent)) {
 			struct page *page;
 
@@ -1075,7 +1072,7 @@ static unsigned long zap_pte_range(struc
 				print_bad_pte(vma, addr, ptent, NULL);
 		}
 		pte_clear_not_present_full(mm, addr, pte, tlb->fullmm);
-	} while (pte++, addr += PAGE_SIZE, (addr != end && *zap_work > 0));
+	} while (pte++, addr += PAGE_SIZE, addr != end);
 
 	add_mm_rss_vec(mm, rss);
 	arch_leave_lazy_mmu_mode();
@@ -1099,7 +1096,7 @@ static unsigned long zap_pte_range(struc
 static inline unsigned long zap_pmd_range(struct mmu_gather *tlb,
 				struct vm_area_struct *vma, pud_t *pud,
 				unsigned long addr, unsigned long end,
-				long *zap_work, struct zap_details *details)
+				struct zap_details *details)
 {
 	pmd_t *pmd;
 	unsigned long next;
@@ -1111,19 +1108,15 @@ static inline unsigned long zap_pmd_rang
 			if (next-addr != HPAGE_PMD_SIZE) {
 				VM_BUG_ON(!rwsem_is_locked(&tlb->mm->mmap_sem));
 				split_huge_page_pmd(vma->vm_mm, pmd);
-			} else if (zap_huge_pmd(tlb, vma, pmd)) {
-				(*zap_work)--;
+			} else if (zap_huge_pmd(tlb, vma, pmd))
 				continue;
-			}
 			/* fall through */
 		}
-		if (pmd_none_or_clear_bad(pmd)) {
-			(*zap_work)--;
+		if (pmd_none_or_clear_bad(pmd))
 			continue;
-		}
-		next = zap_pte_range(tlb, vma, pmd, addr, next,
-						zap_work, details);
-	} while (pmd++, addr = next, (addr != end && *zap_work > 0));
+		next = zap_pte_range(tlb, vma, pmd, addr, next, details);
+		cond_resched();
+	} while (pmd++, addr = next, addr != end);
 
 	return addr;
 }
@@ -1131,7 +1124,7 @@ static inline unsigned long zap_pmd_rang
 static inline unsigned long zap_pud_range(struct mmu_gather *tlb,
 				struct vm_area_struct *vma, pgd_t *pgd,
 				unsigned long addr, unsigned long end,
-				long *zap_work, struct zap_details *details)
+				struct zap_details *details)
 {
 	pud_t *pud;
 	unsigned long next;
@@ -1139,13 +1132,10 @@ static inline unsigned long zap_pud_rang
 	pud = pud_offset(pgd, addr);
 	do {
 		next = pud_addr_end(addr, end);
-		if (pud_none_or_clear_bad(pud)) {
-			(*zap_work)--;
+		if (pud_none_or_clear_bad(pud))
 			continue;
-		}
-		next = zap_pmd_range(tlb, vma, pud, addr, next,
-						zap_work, details);
-	} while (pud++, addr = next, (addr != end && *zap_work > 0));
+		next = zap_pmd_range(tlb, vma, pud, addr, next, details);
+	} while (pud++, addr = next, addr != end);
 
 	return addr;
 }
@@ -1153,7 +1143,7 @@ static inline unsigned long zap_pud_rang
 static unsigned long unmap_page_range(struct mmu_gather *tlb,
 				struct vm_area_struct *vma,
 				unsigned long addr, unsigned long end,
-				long *zap_work, struct zap_details *details)
+				struct zap_details *details)
 {
 	pgd_t *pgd;
 	unsigned long next;
@@ -1167,13 +1157,10 @@ static unsigned long unmap_page_range(st
 	pgd = pgd_offset(vma->vm_mm, addr);
 	do {
 		next = pgd_addr_end(addr, end);
-		if (pgd_none_or_clear_bad(pgd)) {
-			(*zap_work)--;
+		if (pgd_none_or_clear_bad(pgd))
 			continue;
-		}
-		next = zap_pud_range(tlb, vma, pgd, addr, next,
-						zap_work, details);
-	} while (pgd++, addr = next, (addr != end && *zap_work > 0));
+		next = zap_pud_range(tlb, vma, pgd, addr, next, details);
+	} while (pgd++, addr = next, addr != end);
 	tlb_end_vma(tlb, vma);
 	mem_cgroup_uncharge_end();
 
@@ -1218,9 +1205,7 @@ unsigned long unmap_vmas(struct mmu_gath
 		unsigned long end_addr, unsigned long *nr_accounted,
 		struct zap_details *details)
 {
-	long zap_work = ZAP_BLOCK_SIZE;
 	unsigned long start = start_addr;
-	spinlock_t *i_mmap_lock = details? details->i_mmap_lock: NULL;
 	struct mm_struct *mm = vma->vm_mm;
 
 	mmu_notifier_invalidate_range_start(mm, start_addr, end_addr);
@@ -1253,33 +1238,15 @@ unsigned long unmap_vmas(struct mmu_gath
 				 * Since no pte has actually been setup, it is
 				 * safe to do nothing in this case.
 				 */
-				if (vma->vm_file) {
+				if (vma->vm_file)
 					unmap_hugepage_range(vma, start, end, NULL);
-					zap_work -= (end - start) /
-					pages_per_huge_page(hstate_vma(vma));
-				}
 
 				start = end;
 			} else
-				start = unmap_page_range(tlb, vma,
-						start, end, &zap_work, details);
-
-			if (zap_work > 0) {
-				BUG_ON(start != end);
-				break;
-			}
-
-			if (need_resched() ||
-				(i_mmap_lock && spin_needbreak(i_mmap_lock))) {
-				if (i_mmap_lock)
-					goto out;
-				cond_resched();
-			}
-
-			zap_work = ZAP_BLOCK_SIZE;
+				start = unmap_page_range(tlb, vma, start, end, details);
 		}
 	}
-out:
+
 	mmu_notifier_invalidate_range_end(mm, start_addr, end_addr);
 	return start;	/* which is now the end (or restart) address */
 }
@@ -2541,96 +2508,11 @@ static int do_wp_page(struct mm_struct *
 	return ret;
 }
 
-/*
- * Helper functions for unmap_mapping_range().
- *
- * __ Notes on dropping i_mmap_lock to reduce latency while unmapping __
- *
- * We have to restart searching the prio_tree whenever we drop the lock,
- * since the iterator is only valid while the lock is held, and anyway
- * a later vma might be split and reinserted earlier while lock dropped.
- *
- * The list of nonlinear vmas could be handled more efficiently, using
- * a placeholder, but handle it in the same way until a need is shown.
- * It is important to search the prio_tree before nonlinear list: a vma
- * may become nonlinear and be shifted from prio_tree to nonlinear list
- * while the lock is dropped; but never shifted from list to prio_tree.
- *
- * In order to make forward progress despite restarting the search,
- * vm_truncate_count is used to mark a vma as now dealt with, so we can
- * quickly skip it next time around.  Since the prio_tree search only
- * shows us those vmas affected by unmapping the range in question, we
- * can't efficiently keep all vmas in step with mapping->truncate_count:
- * so instead reset them all whenever it wraps back to 0 (then go to 1).
- * mapping->truncate_count and vma->vm_truncate_count are protected by
- * i_mmap_lock.
- *
- * In order to make forward progress despite repeatedly restarting some
- * large vma, note the restart_addr from unmap_vmas when it breaks out:
- * and restart from that address when we reach that vma again.  It might
- * have been split or merged, shrunk or extended, but never shifted: so
- * restart_addr remains valid so long as it remains in the vma's range.
- * unmap_mapping_range forces truncate_count to leap over page-aligned
- * values so we can save vma's restart_addr in its truncate_count field.
- */
-#define is_restart_addr(truncate_count) (!((truncate_count) & ~PAGE_MASK))
-
-static void reset_vma_truncate_counts(struct address_space *mapping)
-{
-	struct vm_area_struct *vma;
-	struct prio_tree_iter iter;
-
-	vma_prio_tree_foreach(vma, &iter, &mapping->i_mmap, 0, ULONG_MAX)
-		vma->vm_truncate_count = 0;
-	list_for_each_entry(vma, &mapping->i_mmap_nonlinear, shared.vm_set.list)
-		vma->vm_truncate_count = 0;
-}
-
-static int unmap_mapping_range_vma(struct vm_area_struct *vma,
+static void unmap_mapping_range_vma(struct vm_area_struct *vma,
 		unsigned long start_addr, unsigned long end_addr,
 		struct zap_details *details)
 {
-	unsigned long restart_addr;
-	int need_break;
-
-	/*
-	 * files that support invalidating or truncating portions of the
-	 * file from under mmaped areas must have their ->fault function
-	 * return a locked page (and set VM_FAULT_LOCKED in the return).
-	 * This provides synchronisation against concurrent unmapping here.
-	 */
-
-again:
-	restart_addr = vma->vm_truncate_count;
-	if (is_restart_addr(restart_addr) && start_addr < restart_addr) {
-		start_addr = restart_addr;
-		if (start_addr >= end_addr) {
-			/* Top of vma has been split off since last time */
-			vma->vm_truncate_count = details->truncate_count;
-			return 0;
-		}
-	}
-
-	restart_addr = zap_page_range(vma, start_addr,
-					end_addr - start_addr, details);
-	need_break = need_resched() || spin_needbreak(details->i_mmap_lock);
-
-	if (restart_addr >= end_addr) {
-		/* We have now completed this vma: mark it so */
-		vma->vm_truncate_count = details->truncate_count;
-		if (!need_break)
-			return 0;
-	} else {
-		/* Note restart_addr in vma's truncate_count field */
-		vma->vm_truncate_count = restart_addr;
-		if (!need_break)
-			goto again;
-	}
-
-	spin_unlock(details->i_mmap_lock);
-	cond_resched();
-	spin_lock(details->i_mmap_lock);
-	return -EINTR;
+	zap_page_range(vma, start_addr, end_addr - start_addr, details);
 }
 
 static inline void unmap_mapping_range_tree(struct prio_tree_root *root,
@@ -2640,12 +2522,8 @@ static inline void unmap_mapping_range_t
 	struct prio_tree_iter iter;
 	pgoff_t vba, vea, zba, zea;
 
-restart:
 	vma_prio_tree_foreach(vma, &iter, root,
 			details->first_index, details->last_index) {
-		/* Skip quickly over those we have already dealt with */
-		if (vma->vm_truncate_count == details->truncate_count)
-			continue;
 
 		vba = vma->vm_pgoff;
 		vea = vba + ((vma->vm_end - vma->vm_start) >> PAGE_SHIFT) - 1;
@@ -2657,11 +2535,10 @@ static inline void unmap_mapping_range_t
 		if (zea > vea)
 			zea = vea;
 
-		if (unmap_mapping_range_vma(vma,
+		unmap_mapping_range_vma(vma,
 			((zba - vba) << PAGE_SHIFT) + vma->vm_start,
 			((zea - vba + 1) << PAGE_SHIFT) + vma->vm_start,
-				details) < 0)
-			goto restart;
+				details);
 	}
 }
 
@@ -2676,15 +2553,9 @@ static inline void unmap_mapping_range_l
 	 * across *all* the pages in each nonlinear VMA, not just the pages
 	 * whose virtual address lies outside the file truncation point.
 	 */
-restart:
 	list_for_each_entry(vma, head, shared.vm_set.list) {
-		/* Skip quickly over those we have already dealt with */
-		if (vma->vm_truncate_count == details->truncate_count)
-			continue;
 		details->nonlinear_vma = vma;
-		if (unmap_mapping_range_vma(vma, vma->vm_start,
-					vma->vm_end, details) < 0)
-			goto restart;
+		unmap_mapping_range_vma(vma, vma->vm_start, vma->vm_end, details);
 	}
 }
 
@@ -2723,26 +2594,14 @@ void unmap_mapping_range(struct address_
 	details.last_index = hba + hlen - 1;
 	if (details.last_index < details.first_index)
 		details.last_index = ULONG_MAX;
-	details.i_mmap_lock = &mapping->i_mmap_lock;
-
-	mutex_lock(&mapping->unmap_mutex);
-	spin_lock(&mapping->i_mmap_lock);
 
-	/* Protect against endless unmapping loops */
-	mapping->truncate_count++;
-	if (unlikely(is_restart_addr(mapping->truncate_count))) {
-		if (mapping->truncate_count == 0)
-			reset_vma_truncate_counts(mapping);
-		mapping->truncate_count++;
-	}
-	details.truncate_count = mapping->truncate_count;
 
+	spin_lock(&mapping->i_mmap_lock);
 	if (unlikely(!prio_tree_empty(&mapping->i_mmap)))
 		unmap_mapping_range_tree(&mapping->i_mmap, &details);
 	if (unlikely(!list_empty(&mapping->i_mmap_nonlinear)))
 		unmap_mapping_range_list(&mapping->i_mmap_nonlinear, &details);
 	spin_unlock(&mapping->i_mmap_lock);
-	mutex_unlock(&mapping->unmap_mutex);
 }
 EXPORT_SYMBOL(unmap_mapping_range);
 
Index: linux-2.6/include/linux/fs.h
===================================================================
--- linux-2.6.orig/include/linux/fs.h
+++ linux-2.6/include/linux/fs.h
@@ -643,7 +643,6 @@ struct address_space {
 	struct prio_tree_root	i_mmap;		/* tree of private and shared mappings */
 	struct list_head	i_mmap_nonlinear;/*list VM_NONLINEAR mappings */
 	spinlock_t		i_mmap_lock;	/* protect tree, count, list */
-	unsigned int		truncate_count;	/* Cover race condition with truncate */
 	unsigned long		nrpages;	/* number of total pages */
 	pgoff_t			writeback_index;/* writeback starts here */
 	const struct address_space_operations *a_ops;	/* methods */
@@ -652,7 +651,6 @@ struct address_space {
 	spinlock_t		private_lock;	/* for use by the address_space */
 	struct list_head	private_list;	/* ditto */
 	struct address_space	*assoc_mapping;	/* ditto */
-	struct mutex		unmap_mutex;    /* to protect unmapping */
 } __attribute__((aligned(sizeof(long))));
 	/*
 	 * On most architectures that alignment is already the case; but
Index: linux-2.6/include/linux/mm_types.h
===================================================================
--- linux-2.6.orig/include/linux/mm_types.h
+++ linux-2.6/include/linux/mm_types.h
@@ -175,7 +175,6 @@ struct vm_area_struct {
 					   units, *not* PAGE_CACHE_SIZE */
 	struct file * vm_file;		/* File we map to (can be NULL). */
 	void * vm_private_data;		/* was vm_pte (shared mem) */
-	unsigned long vm_truncate_count;/* truncate_count or restart_addr */
 
 #ifndef CONFIG_MMU
 	struct vm_region *vm_region;	/* NOMMU mapping region */
Index: linux-2.6/kernel/fork.c
===================================================================
--- linux-2.6.orig/kernel/fork.c
+++ linux-2.6/kernel/fork.c
@@ -379,7 +379,6 @@ static int dup_mmap(struct mm_struct *mm
 			spin_lock(&mapping->i_mmap_lock);
 			if (tmp->vm_flags & VM_SHARED)
 				mapping->i_mmap_writable++;
-			tmp->vm_truncate_count = mpnt->vm_truncate_count;
 			flush_dcache_mmap_lock(mapping);
 			/* insert tmp into the share list, just after mpnt */
 			vma_prio_tree_add(tmp, mpnt);
Index: linux-2.6/mm/mmap.c
===================================================================
--- linux-2.6.orig/mm/mmap.c
+++ linux-2.6/mm/mmap.c
@@ -464,10 +464,8 @@ static void vma_link(struct mm_struct *m
 	if (vma->vm_file)
 		mapping = vma->vm_file->f_mapping;
 
-	if (mapping) {
+	if (mapping)
 		spin_lock(&mapping->i_mmap_lock);
-		vma->vm_truncate_count = mapping->truncate_count;
-	}
 
 	__vma_link(mm, vma, prev, rb_link, rb_parent);
 	__vma_link_file(vma);
@@ -577,16 +575,7 @@ again:			remove_next = 1 + (end > next->
 		if (!(vma->vm_flags & VM_NONLINEAR))
 			root = &mapping->i_mmap;
 		spin_lock(&mapping->i_mmap_lock);
-		if (importer &&
-		    vma->vm_truncate_count != next->vm_truncate_count) {
-			/*
-			 * unmap_mapping_range might be in progress:
-			 * ensure that the expanding vma is rescanned.
-			 */
-			importer->vm_truncate_count = 0;
-		}
 		if (insert) {
-			insert->vm_truncate_count = vma->vm_truncate_count;
 			/*
 			 * Put into prio_tree now, so instantiated pages
 			 * are visible to arm/parisc __flush_dcache_page
Index: linux-2.6/mm/mremap.c
===================================================================
--- linux-2.6.orig/mm/mremap.c
+++ linux-2.6/mm/mremap.c
@@ -94,7 +94,6 @@ static void move_ptes(struct vm_area_str
 		 */
 		mapping = vma->vm_file->f_mapping;
 		spin_lock(&mapping->i_mmap_lock);
-		new_vma->vm_truncate_count = 0;
 	}
 
 	/*
Index: linux-2.6/fs/inode.c
===================================================================
--- linux-2.6.orig/fs/inode.c
+++ linux-2.6/fs/inode.c
@@ -305,7 +305,6 @@ void address_space_init_once(struct addr
 	spin_lock_init(&mapping->private_lock);
 	INIT_RAW_PRIO_TREE_ROOT(&mapping->i_mmap);
 	INIT_LIST_HEAD(&mapping->i_mmap_nonlinear);
-	mutex_init(&mapping->unmap_mutex);
 }
 EXPORT_SYMBOL(address_space_init_once);
 

[PATCH 14/20] mm: Remove i_mmap_lock lockbreak

[Peter Zijlstra]

[-- Attachment #1: mm-fix-zap_block_size.patch --]
[-- Type: text/plain, Size: 17621 bytes --]

Hugh says:
 "The only significant loser, I think, would be page reclaim (when
  concurrent with truncation): could spin for a long time waiting for
  the i_mmap_mutex it expects would soon be dropped? "

Counter points:
 - cpu contention makes the spin stop (need_resched())
 - zap pages should be freeing pages at a higher rate than reclaim
   ever can

I think the simplification of the truncate code is definately worth it.

Effectively reverts: 2aa15890f3c (mm: prevent concurrent
unmap_mapping_range() on the same inode) and takes out the code that
caused its problem.

Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 fs/inode.c               |    1 
 include/linux/fs.h       |    2 
 include/linux/mm.h       |    2 
 include/linux/mm_types.h |    1 
 kernel/fork.c            |    1 
 mm/memory.c              |  195 ++++++-----------------------------------------
 mm/mmap.c                |   13 ---
 mm/mremap.c              |    1 
 8 files changed, 28 insertions(+), 188 deletions(-)

Index: linux-2.6/include/linux/mm.h
===================================================================
--- linux-2.6.orig/include/linux/mm.h
+++ linux-2.6/include/linux/mm.h
@@ -878,8 +878,6 @@ struct zap_details {
 	struct address_space *check_mapping;	/* Check page->mapping if set */
 	pgoff_t	first_index;			/* Lowest page->index to unmap */
 	pgoff_t last_index;			/* Highest page->index to unmap */
-	spinlock_t *i_mmap_lock;		/* For unmap_mapping_range: */
-	unsigned long truncate_count;		/* Compare vm_truncate_count */
 };
 
 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
Index: linux-2.6/mm/memory.c
===================================================================
--- linux-2.6.orig/mm/memory.c
+++ linux-2.6/mm/memory.c
@@ -986,13 +986,13 @@ int copy_page_range(struct mm_struct *ds
 static unsigned long zap_pte_range(struct mmu_gather *tlb,
 				struct vm_area_struct *vma, pmd_t *pmd,
 				unsigned long addr, unsigned long end,
-				long *zap_work, struct zap_details *details)
+				struct zap_details *details)
 {
 	struct mm_struct *mm = tlb->mm;
 	int force_flush = 0;
-	pte_t *pte;
-	spinlock_t *ptl;
 	int rss[NR_MM_COUNTERS];
+	spinlock_t *ptl;
+	pte_t *pte;
 
 	init_rss_vec(rss);
 again:
@@ -1001,12 +1001,9 @@ static unsigned long zap_pte_range(struc
 	do {
 		pte_t ptent = *pte;
 		if (pte_none(ptent)) {
-			(*zap_work)--;
 			continue;
 		}
 
-		(*zap_work) -= PAGE_SIZE;
-
 		if (pte_present(ptent)) {
 			struct page *page;
 
@@ -1075,7 +1072,7 @@ static unsigned long zap_pte_range(struc
 				print_bad_pte(vma, addr, ptent, NULL);
 		}
 		pte_clear_not_present_full(mm, addr, pte, tlb->fullmm);
-	} while (pte++, addr += PAGE_SIZE, (addr != end && *zap_work > 0));
+	} while (pte++, addr += PAGE_SIZE, addr != end);
 
 	add_mm_rss_vec(mm, rss);
 	arch_leave_lazy_mmu_mode();
@@ -1099,7 +1096,7 @@ static unsigned long zap_pte_range(struc
 static inline unsigned long zap_pmd_range(struct mmu_gather *tlb,
 				struct vm_area_struct *vma, pud_t *pud,
 				unsigned long addr, unsigned long end,
-				long *zap_work, struct zap_details *details)
+				struct zap_details *details)
 {
 	pmd_t *pmd;
 	unsigned long next;
@@ -1111,19 +1108,15 @@ static inline unsigned long zap_pmd_rang
 			if (next-addr != HPAGE_PMD_SIZE) {
 				VM_BUG_ON(!rwsem_is_locked(&tlb->mm->mmap_sem));
 				split_huge_page_pmd(vma->vm_mm, pmd);
-			} else if (zap_huge_pmd(tlb, vma, pmd)) {
-				(*zap_work)--;
+			} else if (zap_huge_pmd(tlb, vma, pmd))
 				continue;
-			}
 			/* fall through */
 		}
-		if (pmd_none_or_clear_bad(pmd)) {
-			(*zap_work)--;
+		if (pmd_none_or_clear_bad(pmd))
 			continue;
-		}
-		next = zap_pte_range(tlb, vma, pmd, addr, next,
-						zap_work, details);
-	} while (pmd++, addr = next, (addr != end && *zap_work > 0));
+		next = zap_pte_range(tlb, vma, pmd, addr, next, details);
+		cond_resched();
+	} while (pmd++, addr = next, addr != end);
 
 	return addr;
 }
@@ -1131,7 +1124,7 @@ static inline unsigned long zap_pmd_rang
 static inline unsigned long zap_pud_range(struct mmu_gather *tlb,
 				struct vm_area_struct *vma, pgd_t *pgd,
 				unsigned long addr, unsigned long end,
-				long *zap_work, struct zap_details *details)
+				struct zap_details *details)
 {
 	pud_t *pud;
 	unsigned long next;
@@ -1139,13 +1132,10 @@ static inline unsigned long zap_pud_rang
 	pud = pud_offset(pgd, addr);
 	do {
 		next = pud_addr_end(addr, end);
-		if (pud_none_or_clear_bad(pud)) {
-			(*zap_work)--;
+		if (pud_none_or_clear_bad(pud))
 			continue;
-		}
-		next = zap_pmd_range(tlb, vma, pud, addr, next,
-						zap_work, details);
-	} while (pud++, addr = next, (addr != end && *zap_work > 0));
+		next = zap_pmd_range(tlb, vma, pud, addr, next, details);
+	} while (pud++, addr = next, addr != end);
 
 	return addr;
 }
@@ -1153,7 +1143,7 @@ static inline unsigned long zap_pud_rang
 static unsigned long unmap_page_range(struct mmu_gather *tlb,
 				struct vm_area_struct *vma,
 				unsigned long addr, unsigned long end,
-				long *zap_work, struct zap_details *details)
+				struct zap_details *details)
 {
 	pgd_t *pgd;
 	unsigned long next;
@@ -1167,13 +1157,10 @@ static unsigned long unmap_page_range(st
 	pgd = pgd_offset(vma->vm_mm, addr);
 	do {
 		next = pgd_addr_end(addr, end);
-		if (pgd_none_or_clear_bad(pgd)) {
-			(*zap_work)--;
+		if (pgd_none_or_clear_bad(pgd))
 			continue;
-		}
-		next = zap_pud_range(tlb, vma, pgd, addr, next,
-						zap_work, details);
-	} while (pgd++, addr = next, (addr != end && *zap_work > 0));
+		next = zap_pud_range(tlb, vma, pgd, addr, next, details);
+	} while (pgd++, addr = next, addr != end);
 	tlb_end_vma(tlb, vma);
 	mem_cgroup_uncharge_end();
 
@@ -1218,9 +1205,7 @@ unsigned long unmap_vmas(struct mmu_gath
 		unsigned long end_addr, unsigned long *nr_accounted,
 		struct zap_details *details)
 {
-	long zap_work = ZAP_BLOCK_SIZE;
 	unsigned long start = start_addr;
-	spinlock_t *i_mmap_lock = details? details->i_mmap_lock: NULL;
 	struct mm_struct *mm = vma->vm_mm;
 
 	mmu_notifier_invalidate_range_start(mm, start_addr, end_addr);
@@ -1253,33 +1238,15 @@ unsigned long unmap_vmas(struct mmu_gath
 				 * Since no pte has actually been setup, it is
 				 * safe to do nothing in this case.
 				 */
-				if (vma->vm_file) {
+				if (vma->vm_file)
 					unmap_hugepage_range(vma, start, end, NULL);
-					zap_work -= (end - start) /
-					pages_per_huge_page(hstate_vma(vma));
-				}
 
 				start = end;
 			} else
-				start = unmap_page_range(tlb, vma,
-						start, end, &zap_work, details);
-
-			if (zap_work > 0) {
-				BUG_ON(start != end);
-				break;
-			}
-
-			if (need_resched() ||
-				(i_mmap_lock && spin_needbreak(i_mmap_lock))) {
-				if (i_mmap_lock)
-					goto out;
-				cond_resched();
-			}
-
-			zap_work = ZAP_BLOCK_SIZE;
+				start = unmap_page_range(tlb, vma, start, end, details);
 		}
 	}
-out:
+
 	mmu_notifier_invalidate_range_end(mm, start_addr, end_addr);
 	return start;	/* which is now the end (or restart) address */
 }
@@ -2541,96 +2508,11 @@ static int do_wp_page(struct mm_struct *
 	return ret;
 }
 
-/*
- * Helper functions for unmap_mapping_range().
- *
- * __ Notes on dropping i_mmap_lock to reduce latency while unmapping __
- *
- * We have to restart searching the prio_tree whenever we drop the lock,
- * since the iterator is only valid while the lock is held, and anyway
- * a later vma might be split and reinserted earlier while lock dropped.
- *
- * The list of nonlinear vmas could be handled more efficiently, using
- * a placeholder, but handle it in the same way until a need is shown.
- * It is important to search the prio_tree before nonlinear list: a vma
- * may become nonlinear and be shifted from prio_tree to nonlinear list
- * while the lock is dropped; but never shifted from list to prio_tree.
- *
- * In order to make forward progress despite restarting the search,
- * vm_truncate_count is used to mark a vma as now dealt with, so we can
- * quickly skip it next time around.  Since the prio_tree search only
- * shows us those vmas affected by unmapping the range in question, we
- * can't efficiently keep all vmas in step with mapping->truncate_count:
- * so instead reset them all whenever it wraps back to 0 (then go to 1).
- * mapping->truncate_count and vma->vm_truncate_count are protected by
- * i_mmap_lock.
- *
- * In order to make forward progress despite repeatedly restarting some
- * large vma, note the restart_addr from unmap_vmas when it breaks out:
- * and restart from that address when we reach that vma again.  It might
- * have been split or merged, shrunk or extended, but never shifted: so
- * restart_addr remains valid so long as it remains in the vma's range.
- * unmap_mapping_range forces truncate_count to leap over page-aligned
- * values so we can save vma's restart_addr in its truncate_count field.
- */
-#define is_restart_addr(truncate_count) (!((truncate_count) & ~PAGE_MASK))
-
-static void reset_vma_truncate_counts(struct address_space *mapping)
-{
-	struct vm_area_struct *vma;
-	struct prio_tree_iter iter;
-
-	vma_prio_tree_foreach(vma, &iter, &mapping->i_mmap, 0, ULONG_MAX)
-		vma->vm_truncate_count = 0;
-	list_for_each_entry(vma, &mapping->i_mmap_nonlinear, shared.vm_set.list)
-		vma->vm_truncate_count = 0;
-}
-
-static int unmap_mapping_range_vma(struct vm_area_struct *vma,
+static void unmap_mapping_range_vma(struct vm_area_struct *vma,
 		unsigned long start_addr, unsigned long end_addr,
 		struct zap_details *details)
 {
-	unsigned long restart_addr;
-	int need_break;
-
-	/*
-	 * files that support invalidating or truncating portions of the
-	 * file from under mmaped areas must have their ->fault function
-	 * return a locked page (and set VM_FAULT_LOCKED in the return).
-	 * This provides synchronisation against concurrent unmapping here.
-	 */
-
-again:
-	restart_addr = vma->vm_truncate_count;
-	if (is_restart_addr(restart_addr) && start_addr < restart_addr) {
-		start_addr = restart_addr;
-		if (start_addr >= end_addr) {
-			/* Top of vma has been split off since last time */
-			vma->vm_truncate_count = details->truncate_count;
-			return 0;
-		}
-	}
-
-	restart_addr = zap_page_range(vma, start_addr,
-					end_addr - start_addr, details);
-	need_break = need_resched() || spin_needbreak(details->i_mmap_lock);
-
-	if (restart_addr >= end_addr) {
-		/* We have now completed this vma: mark it so */
-		vma->vm_truncate_count = details->truncate_count;
-		if (!need_break)
-			return 0;
-	} else {
-		/* Note restart_addr in vma's truncate_count field */
-		vma->vm_truncate_count = restart_addr;
-		if (!need_break)
-			goto again;
-	}
-
-	spin_unlock(details->i_mmap_lock);
-	cond_resched();
-	spin_lock(details->i_mmap_lock);
-	return -EINTR;
+	zap_page_range(vma, start_addr, end_addr - start_addr, details);
 }
 
 static inline void unmap_mapping_range_tree(struct prio_tree_root *root,
@@ -2640,12 +2522,8 @@ static inline void unmap_mapping_range_t
 	struct prio_tree_iter iter;
 	pgoff_t vba, vea, zba, zea;
 
-restart:
 	vma_prio_tree_foreach(vma, &iter, root,
 			details->first_index, details->last_index) {
-		/* Skip quickly over those we have already dealt with */
-		if (vma->vm_truncate_count == details->truncate_count)
-			continue;
 
 		vba = vma->vm_pgoff;
 		vea = vba + ((vma->vm_end - vma->vm_start) >> PAGE_SHIFT) - 1;
@@ -2657,11 +2535,10 @@ static inline void unmap_mapping_range_t
 		if (zea > vea)
 			zea = vea;
 
-		if (unmap_mapping_range_vma(vma,
+		unmap_mapping_range_vma(vma,
 			((zba - vba) << PAGE_SHIFT) + vma->vm_start,
 			((zea - vba + 1) << PAGE_SHIFT) + vma->vm_start,
-				details) < 0)
-			goto restart;
+				details);
 	}
 }
 
@@ -2676,15 +2553,9 @@ static inline void unmap_mapping_range_l
 	 * across *all* the pages in each nonlinear VMA, not just the pages
 	 * whose virtual address lies outside the file truncation point.
 	 */
-restart:
 	list_for_each_entry(vma, head, shared.vm_set.list) {
-		/* Skip quickly over those we have already dealt with */
-		if (vma->vm_truncate_count == details->truncate_count)
-			continue;
 		details->nonlinear_vma = vma;
-		if (unmap_mapping_range_vma(vma, vma->vm_start,
-					vma->vm_end, details) < 0)
-			goto restart;
+		unmap_mapping_range_vma(vma, vma->vm_start, vma->vm_end, details);
 	}
 }
 
@@ -2723,26 +2594,14 @@ void unmap_mapping_range(struct address_
 	details.last_index = hba + hlen - 1;
 	if (details.last_index < details.first_index)
 		details.last_index = ULONG_MAX;
-	details.i_mmap_lock = &mapping->i_mmap_lock;
-
-	mutex_lock(&mapping->unmap_mutex);
-	spin_lock(&mapping->i_mmap_lock);
 
-	/* Protect against endless unmapping loops */
-	mapping->truncate_count++;
-	if (unlikely(is_restart_addr(mapping->truncate_count))) {
-		if (mapping->truncate_count == 0)
-			reset_vma_truncate_counts(mapping);
-		mapping->truncate_count++;
-	}
-	details.truncate_count = mapping->truncate_count;
 
+	spin_lock(&mapping->i_mmap_lock);
 	if (unlikely(!prio_tree_empty(&mapping->i_mmap)))
 		unmap_mapping_range_tree(&mapping->i_mmap, &details);
 	if (unlikely(!list_empty(&mapping->i_mmap_nonlinear)))
 		unmap_mapping_range_list(&mapping->i_mmap_nonlinear, &details);
 	spin_unlock(&mapping->i_mmap_lock);
-	mutex_unlock(&mapping->unmap_mutex);
 }
 EXPORT_SYMBOL(unmap_mapping_range);
 
Index: linux-2.6/include/linux/fs.h
===================================================================
--- linux-2.6.orig/include/linux/fs.h
+++ linux-2.6/include/linux/fs.h
@@ -643,7 +643,6 @@ struct address_space {
 	struct prio_tree_root	i_mmap;		/* tree of private and shared mappings */
 	struct list_head	i_mmap_nonlinear;/*list VM_NONLINEAR mappings */
 	spinlock_t		i_mmap_lock;	/* protect tree, count, list */
-	unsigned int		truncate_count;	/* Cover race condition with truncate */
 	unsigned long		nrpages;	/* number of total pages */
 	pgoff_t			writeback_index;/* writeback starts here */
 	const struct address_space_operations *a_ops;	/* methods */
@@ -652,7 +651,6 @@ struct address_space {
 	spinlock_t		private_lock;	/* for use by the address_space */
 	struct list_head	private_list;	/* ditto */
 	struct address_space	*assoc_mapping;	/* ditto */
-	struct mutex		unmap_mutex;    /* to protect unmapping */
 } __attribute__((aligned(sizeof(long))));
 	/*
 	 * On most architectures that alignment is already the case; but
Index: linux-2.6/include/linux/mm_types.h
===================================================================
--- linux-2.6.orig/include/linux/mm_types.h
+++ linux-2.6/include/linux/mm_types.h
@@ -175,7 +175,6 @@ struct vm_area_struct {
 					   units, *not* PAGE_CACHE_SIZE */
 	struct file * vm_file;		/* File we map to (can be NULL). */
 	void * vm_private_data;		/* was vm_pte (shared mem) */
-	unsigned long vm_truncate_count;/* truncate_count or restart_addr */
 
 #ifndef CONFIG_MMU
 	struct vm_region *vm_region;	/* NOMMU mapping region */
Index: linux-2.6/kernel/fork.c
===================================================================
--- linux-2.6.orig/kernel/fork.c
+++ linux-2.6/kernel/fork.c
@@ -379,7 +379,6 @@ static int dup_mmap(struct mm_struct *mm
 			spin_lock(&mapping->i_mmap_lock);
 			if (tmp->vm_flags & VM_SHARED)
 				mapping->i_mmap_writable++;
-			tmp->vm_truncate_count = mpnt->vm_truncate_count;
 			flush_dcache_mmap_lock(mapping);
 			/* insert tmp into the share list, just after mpnt */
 			vma_prio_tree_add(tmp, mpnt);
Index: linux-2.6/mm/mmap.c
===================================================================
--- linux-2.6.orig/mm/mmap.c
+++ linux-2.6/mm/mmap.c
@@ -464,10 +464,8 @@ static void vma_link(struct mm_struct *m
 	if (vma->vm_file)
 		mapping = vma->vm_file->f_mapping;
 
-	if (mapping) {
+	if (mapping)
 		spin_lock(&mapping->i_mmap_lock);
-		vma->vm_truncate_count = mapping->truncate_count;
-	}
 
 	__vma_link(mm, vma, prev, rb_link, rb_parent);
 	__vma_link_file(vma);
@@ -577,16 +575,7 @@ again:			remove_next = 1 + (end > next->
 		if (!(vma->vm_flags & VM_NONLINEAR))
 			root = &mapping->i_mmap;
 		spin_lock(&mapping->i_mmap_lock);
-		if (importer &&
-		    vma->vm_truncate_count != next->vm_truncate_count) {
-			/*
-			 * unmap_mapping_range might be in progress:
-			 * ensure that the expanding vma is rescanned.
-			 */
-			importer->vm_truncate_count = 0;
-		}
 		if (insert) {
-			insert->vm_truncate_count = vma->vm_truncate_count;
 			/*
 			 * Put into prio_tree now, so instantiated pages
 			 * are visible to arm/parisc __flush_dcache_page
Index: linux-2.6/mm/mremap.c
===================================================================
--- linux-2.6.orig/mm/mremap.c
+++ linux-2.6/mm/mremap.c
@@ -94,7 +94,6 @@ static void move_ptes(struct vm_area_str
 		 */
 		mapping = vma->vm_file->f_mapping;
 		spin_lock(&mapping->i_mmap_lock);
-		new_vma->vm_truncate_count = 0;
 	}
 
 	/*
Index: linux-2.6/fs/inode.c
===================================================================
--- linux-2.6.orig/fs/inode.c
+++ linux-2.6/fs/inode.c
@@ -305,7 +305,6 @@ void address_space_init_once(struct addr
 	spin_lock_init(&mapping->private_lock);
 	INIT_RAW_PRIO_TREE_ROOT(&mapping->i_mmap);
 	INIT_LIST_HEAD(&mapping->i_mmap_nonlinear);
-	mutex_init(&mapping->unmap_mutex);
 }
 EXPORT_SYMBOL(address_space_init_once);
 


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 14/20] mm: Remove i_mmap_lock lockbreak

[Peter Zijlstra]

[-- Attachment #1: mm-fix-zap_block_size.patch --]
[-- Type: text/plain, Size: 17621 bytes --]

Hugh says:
 "The only significant loser, I think, would be page reclaim (when
  concurrent with truncation): could spin for a long time waiting for
  the i_mmap_mutex it expects would soon be dropped? "

Counter points:
 - cpu contention makes the spin stop (need_resched())
 - zap pages should be freeing pages at a higher rate than reclaim
   ever can

I think the simplification of the truncate code is definately worth it.

Effectively reverts: 2aa15890f3c (mm: prevent concurrent
unmap_mapping_range() on the same inode) and takes out the code that
caused its problem.

Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 fs/inode.c               |    1 
 include/linux/fs.h       |    2 
 include/linux/mm.h       |    2 
 include/linux/mm_types.h |    1 
 kernel/fork.c            |    1 
 mm/memory.c              |  195 ++++++-----------------------------------------
 mm/mmap.c                |   13 ---
 mm/mremap.c              |    1 
 8 files changed, 28 insertions(+), 188 deletions(-)

Index: linux-2.6/include/linux/mm.h
===================================================================
--- linux-2.6.orig/include/linux/mm.h
+++ linux-2.6/include/linux/mm.h
@@ -878,8 +878,6 @@ struct zap_details {
 	struct address_space *check_mapping;	/* Check page->mapping if set */
 	pgoff_t	first_index;			/* Lowest page->index to unmap */
 	pgoff_t last_index;			/* Highest page->index to unmap */
-	spinlock_t *i_mmap_lock;		/* For unmap_mapping_range: */
-	unsigned long truncate_count;		/* Compare vm_truncate_count */
 };
 
 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
Index: linux-2.6/mm/memory.c
===================================================================
--- linux-2.6.orig/mm/memory.c
+++ linux-2.6/mm/memory.c
@@ -986,13 +986,13 @@ int copy_page_range(struct mm_struct *ds
 static unsigned long zap_pte_range(struct mmu_gather *tlb,
 				struct vm_area_struct *vma, pmd_t *pmd,
 				unsigned long addr, unsigned long end,
-				long *zap_work, struct zap_details *details)
+				struct zap_details *details)
 {
 	struct mm_struct *mm = tlb->mm;
 	int force_flush = 0;
-	pte_t *pte;
-	spinlock_t *ptl;
 	int rss[NR_MM_COUNTERS];
+	spinlock_t *ptl;
+	pte_t *pte;
 
 	init_rss_vec(rss);
 again:
@@ -1001,12 +1001,9 @@ static unsigned long zap_pte_range(struc
 	do {
 		pte_t ptent = *pte;
 		if (pte_none(ptent)) {
-			(*zap_work)--;
 			continue;
 		}
 
-		(*zap_work) -= PAGE_SIZE;
-
 		if (pte_present(ptent)) {
 			struct page *page;
 
@@ -1075,7 +1072,7 @@ static unsigned long zap_pte_range(struc
 				print_bad_pte(vma, addr, ptent, NULL);
 		}
 		pte_clear_not_present_full(mm, addr, pte, tlb->fullmm);
-	} while (pte++, addr += PAGE_SIZE, (addr != end && *zap_work > 0));
+	} while (pte++, addr += PAGE_SIZE, addr != end);
 
 	add_mm_rss_vec(mm, rss);
 	arch_leave_lazy_mmu_mode();
@@ -1099,7 +1096,7 @@ static unsigned long zap_pte_range(struc
 static inline unsigned long zap_pmd_range(struct mmu_gather *tlb,
 				struct vm_area_struct *vma, pud_t *pud,
 				unsigned long addr, unsigned long end,
-				long *zap_work, struct zap_details *details)
+				struct zap_details *details)
 {
 	pmd_t *pmd;
 	unsigned long next;
@@ -1111,19 +1108,15 @@ static inline unsigned long zap_pmd_rang
 			if (next-addr != HPAGE_PMD_SIZE) {
 				VM_BUG_ON(!rwsem_is_locked(&tlb->mm->mmap_sem));
 				split_huge_page_pmd(vma->vm_mm, pmd);
-			} else if (zap_huge_pmd(tlb, vma, pmd)) {
-				(*zap_work)--;
+			} else if (zap_huge_pmd(tlb, vma, pmd))
 				continue;
-			}
 			/* fall through */
 		}
-		if (pmd_none_or_clear_bad(pmd)) {
-			(*zap_work)--;
+		if (pmd_none_or_clear_bad(pmd))
 			continue;
-		}
-		next = zap_pte_range(tlb, vma, pmd, addr, next,
-						zap_work, details);
-	} while (pmd++, addr = next, (addr != end && *zap_work > 0));
+		next = zap_pte_range(tlb, vma, pmd, addr, next, details);
+		cond_resched();
+	} while (pmd++, addr = next, addr != end);
 
 	return addr;
 }
@@ -1131,7 +1124,7 @@ static inline unsigned long zap_pmd_rang
 static inline unsigned long zap_pud_range(struct mmu_gather *tlb,
 				struct vm_area_struct *vma, pgd_t *pgd,
 				unsigned long addr, unsigned long end,
-				long *zap_work, struct zap_details *details)
+				struct zap_details *details)
 {
 	pud_t *pud;
 	unsigned long next;
@@ -1139,13 +1132,10 @@ static inline unsigned long zap_pud_rang
 	pud = pud_offset(pgd, addr);
 	do {
 		next = pud_addr_end(addr, end);
-		if (pud_none_or_clear_bad(pud)) {
-			(*zap_work)--;
+		if (pud_none_or_clear_bad(pud))
 			continue;
-		}
-		next = zap_pmd_range(tlb, vma, pud, addr, next,
-						zap_work, details);
-	} while (pud++, addr = next, (addr != end && *zap_work > 0));
+		next = zap_pmd_range(tlb, vma, pud, addr, next, details);
+	} while (pud++, addr = next, addr != end);
 
 	return addr;
 }
@@ -1153,7 +1143,7 @@ static inline unsigned long zap_pud_rang
 static unsigned long unmap_page_range(struct mmu_gather *tlb,
 				struct vm_area_struct *vma,
 				unsigned long addr, unsigned long end,
-				long *zap_work, struct zap_details *details)
+				struct zap_details *details)
 {
 	pgd_t *pgd;
 	unsigned long next;
@@ -1167,13 +1157,10 @@ static unsigned long unmap_page_range(st
 	pgd = pgd_offset(vma->vm_mm, addr);
 	do {
 		next = pgd_addr_end(addr, end);
-		if (pgd_none_or_clear_bad(pgd)) {
-			(*zap_work)--;
+		if (pgd_none_or_clear_bad(pgd))
 			continue;
-		}
-		next = zap_pud_range(tlb, vma, pgd, addr, next,
-						zap_work, details);
-	} while (pgd++, addr = next, (addr != end && *zap_work > 0));
+		next = zap_pud_range(tlb, vma, pgd, addr, next, details);
+	} while (pgd++, addr = next, addr != end);
 	tlb_end_vma(tlb, vma);
 	mem_cgroup_uncharge_end();
 
@@ -1218,9 +1205,7 @@ unsigned long unmap_vmas(struct mmu_gath
 		unsigned long end_addr, unsigned long *nr_accounted,
 		struct zap_details *details)
 {
-	long zap_work = ZAP_BLOCK_SIZE;
 	unsigned long start = start_addr;
-	spinlock_t *i_mmap_lock = details? details->i_mmap_lock: NULL;
 	struct mm_struct *mm = vma->vm_mm;
 
 	mmu_notifier_invalidate_range_start(mm, start_addr, end_addr);
@@ -1253,33 +1238,15 @@ unsigned long unmap_vmas(struct mmu_gath
 				 * Since no pte has actually been setup, it is
 				 * safe to do nothing in this case.
 				 */
-				if (vma->vm_file) {
+				if (vma->vm_file)
 					unmap_hugepage_range(vma, start, end, NULL);
-					zap_work -= (end - start) /
-					pages_per_huge_page(hstate_vma(vma));
-				}
 
 				start = end;
 			} else
-				start = unmap_page_range(tlb, vma,
-						start, end, &zap_work, details);
-
-			if (zap_work > 0) {
-				BUG_ON(start != end);
-				break;
-			}
-
-			if (need_resched() ||
-				(i_mmap_lock && spin_needbreak(i_mmap_lock))) {
-				if (i_mmap_lock)
-					goto out;
-				cond_resched();
-			}
-
-			zap_work = ZAP_BLOCK_SIZE;
+				start = unmap_page_range(tlb, vma, start, end, details);
 		}
 	}
-out:
+
 	mmu_notifier_invalidate_range_end(mm, start_addr, end_addr);
 	return start;	/* which is now the end (or restart) address */
 }
@@ -2541,96 +2508,11 @@ static int do_wp_page(struct mm_struct *
 	return ret;
 }
 
-/*
- * Helper functions for unmap_mapping_range().
- *
- * __ Notes on dropping i_mmap_lock to reduce latency while unmapping __
- *
- * We have to restart searching the prio_tree whenever we drop the lock,
- * since the iterator is only valid while the lock is held, and anyway
- * a later vma might be split and reinserted earlier while lock dropped.
- *
- * The list of nonlinear vmas could be handled more efficiently, using
- * a placeholder, but handle it in the same way until a need is shown.
- * It is important to search the prio_tree before nonlinear list: a vma
- * may become nonlinear and be shifted from prio_tree to nonlinear list
- * while the lock is dropped; but never shifted from list to prio_tree.
- *
- * In order to make forward progress despite restarting the search,
- * vm_truncate_count is used to mark a vma as now dealt with, so we can
- * quickly skip it next time around.  Since the prio_tree search only
- * shows us those vmas affected by unmapping the range in question, we
- * can't efficiently keep all vmas in step with mapping->truncate_count:
- * so instead reset them all whenever it wraps back to 0 (then go to 1).
- * mapping->truncate_count and vma->vm_truncate_count are protected by
- * i_mmap_lock.
- *
- * In order to make forward progress despite repeatedly restarting some
- * large vma, note the restart_addr from unmap_vmas when it breaks out:
- * and restart from that address when we reach that vma again.  It might
- * have been split or merged, shrunk or extended, but never shifted: so
- * restart_addr remains valid so long as it remains in the vma's range.
- * unmap_mapping_range forces truncate_count to leap over page-aligned
- * values so we can save vma's restart_addr in its truncate_count field.
- */
-#define is_restart_addr(truncate_count) (!((truncate_count) & ~PAGE_MASK))
-
-static void reset_vma_truncate_counts(struct address_space *mapping)
-{
-	struct vm_area_struct *vma;
-	struct prio_tree_iter iter;
-
-	vma_prio_tree_foreach(vma, &iter, &mapping->i_mmap, 0, ULONG_MAX)
-		vma->vm_truncate_count = 0;
-	list_for_each_entry(vma, &mapping->i_mmap_nonlinear, shared.vm_set.list)
-		vma->vm_truncate_count = 0;
-}
-
-static int unmap_mapping_range_vma(struct vm_area_struct *vma,
+static void unmap_mapping_range_vma(struct vm_area_struct *vma,
 		unsigned long start_addr, unsigned long end_addr,
 		struct zap_details *details)
 {
-	unsigned long restart_addr;
-	int need_break;
-
-	/*
-	 * files that support invalidating or truncating portions of the
-	 * file from under mmaped areas must have their ->fault function
-	 * return a locked page (and set VM_FAULT_LOCKED in the return).
-	 * This provides synchronisation against concurrent unmapping here.
-	 */
-
-again:
-	restart_addr = vma->vm_truncate_count;
-	if (is_restart_addr(restart_addr) && start_addr < restart_addr) {
-		start_addr = restart_addr;
-		if (start_addr >= end_addr) {
-			/* Top of vma has been split off since last time */
-			vma->vm_truncate_count = details->truncate_count;
-			return 0;
-		}
-	}
-
-	restart_addr = zap_page_range(vma, start_addr,
-					end_addr - start_addr, details);
-	need_break = need_resched() || spin_needbreak(details->i_mmap_lock);
-
-	if (restart_addr >= end_addr) {
-		/* We have now completed this vma: mark it so */
-		vma->vm_truncate_count = details->truncate_count;
-		if (!need_break)
-			return 0;
-	} else {
-		/* Note restart_addr in vma's truncate_count field */
-		vma->vm_truncate_count = restart_addr;
-		if (!need_break)
-			goto again;
-	}
-
-	spin_unlock(details->i_mmap_lock);
-	cond_resched();
-	spin_lock(details->i_mmap_lock);
-	return -EINTR;
+	zap_page_range(vma, start_addr, end_addr - start_addr, details);
 }
 
 static inline void unmap_mapping_range_tree(struct prio_tree_root *root,
@@ -2640,12 +2522,8 @@ static inline void unmap_mapping_range_t
 	struct prio_tree_iter iter;
 	pgoff_t vba, vea, zba, zea;
 
-restart:
 	vma_prio_tree_foreach(vma, &iter, root,
 			details->first_index, details->last_index) {
-		/* Skip quickly over those we have already dealt with */
-		if (vma->vm_truncate_count == details->truncate_count)
-			continue;
 
 		vba = vma->vm_pgoff;
 		vea = vba + ((vma->vm_end - vma->vm_start) >> PAGE_SHIFT) - 1;
@@ -2657,11 +2535,10 @@ static inline void unmap_mapping_range_t
 		if (zea > vea)
 			zea = vea;
 
-		if (unmap_mapping_range_vma(vma,
+		unmap_mapping_range_vma(vma,
 			((zba - vba) << PAGE_SHIFT) + vma->vm_start,
 			((zea - vba + 1) << PAGE_SHIFT) + vma->vm_start,
-				details) < 0)
-			goto restart;
+				details);
 	}
 }
 
@@ -2676,15 +2553,9 @@ static inline void unmap_mapping_range_l
 	 * across *all* the pages in each nonlinear VMA, not just the pages
 	 * whose virtual address lies outside the file truncation point.
 	 */
-restart:
 	list_for_each_entry(vma, head, shared.vm_set.list) {
-		/* Skip quickly over those we have already dealt with */
-		if (vma->vm_truncate_count == details->truncate_count)
-			continue;
 		details->nonlinear_vma = vma;
-		if (unmap_mapping_range_vma(vma, vma->vm_start,
-					vma->vm_end, details) < 0)
-			goto restart;
+		unmap_mapping_range_vma(vma, vma->vm_start, vma->vm_end, details);
 	}
 }
 
@@ -2723,26 +2594,14 @@ void unmap_mapping_range(struct address_
 	details.last_index = hba + hlen - 1;
 	if (details.last_index < details.first_index)
 		details.last_index = ULONG_MAX;
-	details.i_mmap_lock = &mapping->i_mmap_lock;
-
-	mutex_lock(&mapping->unmap_mutex);
-	spin_lock(&mapping->i_mmap_lock);
 
-	/* Protect against endless unmapping loops */
-	mapping->truncate_count++;
-	if (unlikely(is_restart_addr(mapping->truncate_count))) {
-		if (mapping->truncate_count == 0)
-			reset_vma_truncate_counts(mapping);
-		mapping->truncate_count++;
-	}
-	details.truncate_count = mapping->truncate_count;
 
+	spin_lock(&mapping->i_mmap_lock);
 	if (unlikely(!prio_tree_empty(&mapping->i_mmap)))
 		unmap_mapping_range_tree(&mapping->i_mmap, &details);
 	if (unlikely(!list_empty(&mapping->i_mmap_nonlinear)))
 		unmap_mapping_range_list(&mapping->i_mmap_nonlinear, &details);
 	spin_unlock(&mapping->i_mmap_lock);
-	mutex_unlock(&mapping->unmap_mutex);
 }
 EXPORT_SYMBOL(unmap_mapping_range);
 
Index: linux-2.6/include/linux/fs.h
===================================================================
--- linux-2.6.orig/include/linux/fs.h
+++ linux-2.6/include/linux/fs.h
@@ -643,7 +643,6 @@ struct address_space {
 	struct prio_tree_root	i_mmap;		/* tree of private and shared mappings */
 	struct list_head	i_mmap_nonlinear;/*list VM_NONLINEAR mappings */
 	spinlock_t		i_mmap_lock;	/* protect tree, count, list */
-	unsigned int		truncate_count;	/* Cover race condition with truncate */
 	unsigned long		nrpages;	/* number of total pages */
 	pgoff_t			writeback_index;/* writeback starts here */
 	const struct address_space_operations *a_ops;	/* methods */
@@ -652,7 +651,6 @@ struct address_space {
 	spinlock_t		private_lock;	/* for use by the address_space */
 	struct list_head	private_list;	/* ditto */
 	struct address_space	*assoc_mapping;	/* ditto */
-	struct mutex		unmap_mutex;    /* to protect unmapping */
 } __attribute__((aligned(sizeof(long))));
 	/*
 	 * On most architectures that alignment is already the case; but
Index: linux-2.6/include/linux/mm_types.h
===================================================================
--- linux-2.6.orig/include/linux/mm_types.h
+++ linux-2.6/include/linux/mm_types.h
@@ -175,7 +175,6 @@ struct vm_area_struct {
 					   units, *not* PAGE_CACHE_SIZE */
 	struct file * vm_file;		/* File we map to (can be NULL). */
 	void * vm_private_data;		/* was vm_pte (shared mem) */
-	unsigned long vm_truncate_count;/* truncate_count or restart_addr */
 
 #ifndef CONFIG_MMU
 	struct vm_region *vm_region;	/* NOMMU mapping region */
Index: linux-2.6/kernel/fork.c
===================================================================
--- linux-2.6.orig/kernel/fork.c
+++ linux-2.6/kernel/fork.c
@@ -379,7 +379,6 @@ static int dup_mmap(struct mm_struct *mm
 			spin_lock(&mapping->i_mmap_lock);
 			if (tmp->vm_flags & VM_SHARED)
 				mapping->i_mmap_writable++;
-			tmp->vm_truncate_count = mpnt->vm_truncate_count;
 			flush_dcache_mmap_lock(mapping);
 			/* insert tmp into the share list, just after mpnt */
 			vma_prio_tree_add(tmp, mpnt);
Index: linux-2.6/mm/mmap.c
===================================================================
--- linux-2.6.orig/mm/mmap.c
+++ linux-2.6/mm/mmap.c
@@ -464,10 +464,8 @@ static void vma_link(struct mm_struct *m
 	if (vma->vm_file)
 		mapping = vma->vm_file->f_mapping;
 
-	if (mapping) {
+	if (mapping)
 		spin_lock(&mapping->i_mmap_lock);
-		vma->vm_truncate_count = mapping->truncate_count;
-	}
 
 	__vma_link(mm, vma, prev, rb_link, rb_parent);
 	__vma_link_file(vma);
@@ -577,16 +575,7 @@ again:			remove_next = 1 + (end > next->
 		if (!(vma->vm_flags & VM_NONLINEAR))
 			root = &mapping->i_mmap;
 		spin_lock(&mapping->i_mmap_lock);
-		if (importer &&
-		    vma->vm_truncate_count != next->vm_truncate_count) {
-			/*
-			 * unmap_mapping_range might be in progress:
-			 * ensure that the expanding vma is rescanned.
-			 */
-			importer->vm_truncate_count = 0;
-		}
 		if (insert) {
-			insert->vm_truncate_count = vma->vm_truncate_count;
 			/*
 			 * Put into prio_tree now, so instantiated pages
 			 * are visible to arm/parisc __flush_dcache_page
Index: linux-2.6/mm/mremap.c
===================================================================
--- linux-2.6.orig/mm/mremap.c
+++ linux-2.6/mm/mremap.c
@@ -94,7 +94,6 @@ static void move_ptes(struct vm_area_str
 		 */
 		mapping = vma->vm_file->f_mapping;
 		spin_lock(&mapping->i_mmap_lock);
-		new_vma->vm_truncate_count = 0;
 	}
 
 	/*
Index: linux-2.6/fs/inode.c
===================================================================
--- linux-2.6.orig/fs/inode.c
+++ linux-2.6/fs/inode.c
@@ -305,7 +305,6 @@ void address_space_init_once(struct addr
 	spin_lock_init(&mapping->private_lock);
 	INIT_RAW_PRIO_TREE_ROOT(&mapping->i_mmap);
 	INIT_LIST_HEAD(&mapping->i_mmap_nonlinear);
-	mutex_init(&mapping->unmap_mutex);
 }
 EXPORT_SYMBOL(address_space_init_once);
 


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 15/20] mm: Convert i_mmap_lock to a mutex

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-mm-convert_i_mmap_lock_to_mutexes.patch --]
[-- Type: text/plain, Size: 21598 bytes --]

Straight fwd conversion of i_mmap_lock to a mutex

Acked-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 Documentation/lockstat.txt   |    2 +-
 Documentation/vm/locking     |    2 +-
 arch/x86/mm/hugetlbpage.c    |    4 ++--
 fs/hugetlbfs/inode.c         |    4 ++--
 fs/inode.c                   |    2 +-
 include/linux/fs.h           |    2 +-
 include/linux/mmu_notifier.h |    2 +-
 kernel/fork.c                |    4 ++--
 mm/filemap.c                 |   10 +++++-----
 mm/filemap_xip.c             |    4 ++--
 mm/fremap.c                  |    4 ++--
 mm/hugetlb.c                 |   14 +++++++-------
 mm/memory-failure.c          |    4 ++--
 mm/memory.c                  |    4 ++--
 mm/mmap.c                    |   22 +++++++++++-----------
 mm/mremap.c                  |    4 ++--
 mm/rmap.c                    |   28 ++++++++++++++--------------
 17 files changed, 58 insertions(+), 58 deletions(-)

Index: linux-2.6/arch/x86/mm/hugetlbpage.c
===================================================================
--- linux-2.6.orig/arch/x86/mm/hugetlbpage.c
+++ linux-2.6/arch/x86/mm/hugetlbpage.c
@@ -72,7 +72,7 @@ static void huge_pmd_share(struct mm_str
 	if (!vma_shareable(vma, addr))
 		return;
 
-	spin_lock(&mapping->i_mmap_lock);
+	mutex_lock(&mapping->i_mmap_mutex);
 	vma_prio_tree_foreach(svma, &iter, &mapping->i_mmap, idx, idx) {
 		if (svma == vma)
 			continue;
@@ -97,7 +97,7 @@ static void huge_pmd_share(struct mm_str
 		put_page(virt_to_page(spte));
 	spin_unlock(&mm->page_table_lock);
 out:
-	spin_unlock(&mapping->i_mmap_lock);
+	mutex_unlock(&mapping->i_mmap_mutex);
 }
 
 /*
Index: linux-2.6/fs/hugetlbfs/inode.c
===================================================================
--- linux-2.6.orig/fs/hugetlbfs/inode.c
+++ linux-2.6/fs/hugetlbfs/inode.c
@@ -412,10 +412,10 @@ static int hugetlb_vmtruncate(struct ino
 	pgoff = offset >> PAGE_SHIFT;
 
 	i_size_write(inode, offset);
-	spin_lock(&mapping->i_mmap_lock);
+	mutex_lock(&mapping->i_mmap_mutex);
 	if (!prio_tree_empty(&mapping->i_mmap))
 		hugetlb_vmtruncate_list(&mapping->i_mmap, pgoff);
-	spin_unlock(&mapping->i_mmap_lock);
+	mutex_unlock(&mapping->i_mmap_mutex);
 	truncate_hugepages(inode, offset);
 	return 0;
 }
Index: linux-2.6/fs/inode.c
===================================================================
--- linux-2.6.orig/fs/inode.c
+++ linux-2.6/fs/inode.c
@@ -318,7 +318,7 @@ void address_space_init_once(struct addr
 	memset(mapping, 0, sizeof(*mapping));
 	INIT_RADIX_TREE(&mapping->page_tree, GFP_ATOMIC);
 	spin_lock_init(&mapping->tree_lock);
-	spin_lock_init(&mapping->i_mmap_lock);
+	mutex_init(&mapping->i_mmap_mutex);
 	INIT_LIST_HEAD(&mapping->private_list);
 	spin_lock_init(&mapping->private_lock);
 	INIT_RAW_PRIO_TREE_ROOT(&mapping->i_mmap);
Index: linux-2.6/include/linux/fs.h
===================================================================
--- linux-2.6.orig/include/linux/fs.h
+++ linux-2.6/include/linux/fs.h
@@ -633,7 +633,7 @@ struct address_space {
 	unsigned int		i_mmap_writable;/* count VM_SHARED mappings */
 	struct prio_tree_root	i_mmap;		/* tree of private and shared mappings */
 	struct list_head	i_mmap_nonlinear;/*list VM_NONLINEAR mappings */
-	spinlock_t		i_mmap_lock;	/* protect tree, count, list */
+	struct mutex		i_mmap_mutex;	/* protect tree, count, list */
 	unsigned long		nrpages;	/* number of total pages */
 	pgoff_t			writeback_index;/* writeback starts here */
 	const struct address_space_operations *a_ops;	/* methods */
Index: linux-2.6/kernel/fork.c
===================================================================
--- linux-2.6.orig/kernel/fork.c
+++ linux-2.6/kernel/fork.c
@@ -383,14 +383,14 @@ static int dup_mmap(struct mm_struct *mm
 			get_file(file);
 			if (tmp->vm_flags & VM_DENYWRITE)
 				atomic_dec(&inode->i_writecount);
-			spin_lock(&mapping->i_mmap_lock);
+			mutex_lock(&mapping->i_mmap_mutex);
 			if (tmp->vm_flags & VM_SHARED)
 				mapping->i_mmap_writable++;
 			flush_dcache_mmap_lock(mapping);
 			/* insert tmp into the share list, just after mpnt */
 			vma_prio_tree_add(tmp, mpnt);
 			flush_dcache_mmap_unlock(mapping);
-			spin_unlock(&mapping->i_mmap_lock);
+			mutex_unlock(&mapping->i_mmap_mutex);
 		}
 
 		/*
Index: linux-2.6/mm/filemap_xip.c
===================================================================
--- linux-2.6.orig/mm/filemap_xip.c
+++ linux-2.6/mm/filemap_xip.c
@@ -183,7 +183,7 @@ __xip_unmap (struct address_space * mapp
 		return;
 
 retry:
-	spin_lock(&mapping->i_mmap_lock);
+	mutex_lock(&mapping->i_mmap_mutex);
 	vma_prio_tree_foreach(vma, &iter, &mapping->i_mmap, pgoff, pgoff) {
 		mm = vma->vm_mm;
 		address = vma->vm_start +
@@ -201,7 +201,7 @@ __xip_unmap (struct address_space * mapp
 			page_cache_release(page);
 		}
 	}
-	spin_unlock(&mapping->i_mmap_lock);
+	mutex_unlock(&mapping->i_mmap_mutex);
 
 	if (locked) {
 		mutex_unlock(&xip_sparse_mutex);
Index: linux-2.6/mm/fremap.c
===================================================================
--- linux-2.6.orig/mm/fremap.c
+++ linux-2.6/mm/fremap.c
@@ -211,13 +211,13 @@ SYSCALL_DEFINE5(remap_file_pages, unsign
 			}
 			goto out;
 		}
-		spin_lock(&mapping->i_mmap_lock);
+		mutex_lock(&mapping->i_mmap_mutex);
 		flush_dcache_mmap_lock(mapping);
 		vma->vm_flags |= VM_NONLINEAR;
 		vma_prio_tree_remove(vma, &mapping->i_mmap);
 		vma_nonlinear_insert(vma, &mapping->i_mmap_nonlinear);
 		flush_dcache_mmap_unlock(mapping);
-		spin_unlock(&mapping->i_mmap_lock);
+		mutex_unlock(&mapping->i_mmap_mutex);
 	}
 
 	if (vma->vm_flags & VM_LOCKED) {
Index: linux-2.6/mm/hugetlb.c
===================================================================
--- linux-2.6.orig/mm/hugetlb.c
+++ linux-2.6/mm/hugetlb.c
@@ -2205,7 +2205,7 @@ void __unmap_hugepage_range(struct vm_ar
 	unsigned long sz = huge_page_size(h);
 
 	/*
-	 * A page gathering list, protected by per file i_mmap_lock. The
+	 * A page gathering list, protected by per file i_mmap_mutex. The
 	 * lock is used to avoid list corruption from multiple unmapping
 	 * of the same page since we are using page->lru.
 	 */
@@ -2274,9 +2274,9 @@ void __unmap_hugepage_range(struct vm_ar
 void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
 			  unsigned long end, struct page *ref_page)
 {
-	spin_lock(&vma->vm_file->f_mapping->i_mmap_lock);
+	mutex_lock(&vma->vm_file->f_mapping->i_mmap_mutex);
 	__unmap_hugepage_range(vma, start, end, ref_page);
-	spin_unlock(&vma->vm_file->f_mapping->i_mmap_lock);
+	mutex_unlock(&vma->vm_file->f_mapping->i_mmap_mutex);
 }
 
 /*
@@ -2308,7 +2308,7 @@ static int unmap_ref_private(struct mm_s
 	 * this mapping should be shared between all the VMAs,
 	 * __unmap_hugepage_range() is called as the lock is already held
 	 */
-	spin_lock(&mapping->i_mmap_lock);
+	mutex_lock(&mapping->i_mmap_mutex);
 	vma_prio_tree_foreach(iter_vma, &iter, &mapping->i_mmap, pgoff, pgoff) {
 		/* Do not unmap the current VMA */
 		if (iter_vma == vma)
@@ -2326,7 +2326,7 @@ static int unmap_ref_private(struct mm_s
 				address, address + huge_page_size(h),
 				page);
 	}
-	spin_unlock(&mapping->i_mmap_lock);
+	mutex_unlock(&mapping->i_mmap_mutex);
 
 	return 1;
 }
@@ -2810,7 +2810,7 @@ void hugetlb_change_protection(struct vm
 	BUG_ON(address >= end);
 	flush_cache_range(vma, address, end);
 
-	spin_lock(&vma->vm_file->f_mapping->i_mmap_lock);
+	mutex_lock(&vma->vm_file->f_mapping->i_mmap_mutex);
 	spin_lock(&mm->page_table_lock);
 	for (; address < end; address += huge_page_size(h)) {
 		ptep = huge_pte_offset(mm, address);
@@ -2825,7 +2825,7 @@ void hugetlb_change_protection(struct vm
 		}
 	}
 	spin_unlock(&mm->page_table_lock);
-	spin_unlock(&vma->vm_file->f_mapping->i_mmap_lock);
+	mutex_unlock(&vma->vm_file->f_mapping->i_mmap_mutex);
 
 	flush_tlb_range(vma, start, end);
 }
Index: linux-2.6/mm/memory-failure.c
===================================================================
--- linux-2.6.orig/mm/memory-failure.c
+++ linux-2.6/mm/memory-failure.c
@@ -429,7 +429,7 @@ static void collect_procs_file(struct pa
 	 */
 
 	read_lock(&tasklist_lock);
-	spin_lock(&mapping->i_mmap_lock);
+	mutex_lock(&mapping->i_mmap_mutex);
 	for_each_process(tsk) {
 		pgoff_t pgoff = page->index << (PAGE_CACHE_SHIFT - PAGE_SHIFT);
 
@@ -449,7 +449,7 @@ static void collect_procs_file(struct pa
 				add_to_kill(tsk, page, vma, to_kill, tkc);
 		}
 	}
-	spin_unlock(&mapping->i_mmap_lock);
+	mutex_unlock(&mapping->i_mmap_mutex);
 	read_unlock(&tasklist_lock);
 }
 
Index: linux-2.6/mm/memory.c
===================================================================
--- linux-2.6.orig/mm/memory.c
+++ linux-2.6/mm/memory.c
@@ -2660,12 +2660,12 @@ void unmap_mapping_range(struct address_
 		details.last_index = ULONG_MAX;
 
 
-	spin_lock(&mapping->i_mmap_lock);
+	mutex_lock(&mapping->i_mmap_mutex);
 	if (unlikely(!prio_tree_empty(&mapping->i_mmap)))
 		unmap_mapping_range_tree(&mapping->i_mmap, &details);
 	if (unlikely(!list_empty(&mapping->i_mmap_nonlinear)))
 		unmap_mapping_range_list(&mapping->i_mmap_nonlinear, &details);
-	spin_unlock(&mapping->i_mmap_lock);
+	mutex_unlock(&mapping->i_mmap_mutex);
 }
 EXPORT_SYMBOL(unmap_mapping_range);
 
Index: linux-2.6/mm/mmap.c
===================================================================
--- linux-2.6.orig/mm/mmap.c
+++ linux-2.6/mm/mmap.c
@@ -190,7 +190,7 @@ int __vm_enough_memory(struct mm_struct 
 }
 
 /*
- * Requires inode->i_mapping->i_mmap_lock
+ * Requires inode->i_mapping->i_mmap_mutex
  */
 static void __remove_shared_vm_struct(struct vm_area_struct *vma,
 		struct file *file, struct address_space *mapping)
@@ -218,9 +218,9 @@ void unlink_file_vma(struct vm_area_stru
 
 	if (file) {
 		struct address_space *mapping = file->f_mapping;
-		spin_lock(&mapping->i_mmap_lock);
+		mutex_lock(&mapping->i_mmap_mutex);
 		__remove_shared_vm_struct(vma, file, mapping);
-		spin_unlock(&mapping->i_mmap_lock);
+		mutex_unlock(&mapping->i_mmap_mutex);
 	}
 }
 
@@ -465,13 +465,13 @@ static void vma_link(struct mm_struct *m
 		mapping = vma->vm_file->f_mapping;
 
 	if (mapping)
-		spin_lock(&mapping->i_mmap_lock);
+		mutex_lock(&mapping->i_mmap_mutex);
 
 	__vma_link(mm, vma, prev, rb_link, rb_parent);
 	__vma_link_file(vma);
 
 	if (mapping)
-		spin_unlock(&mapping->i_mmap_lock);
+		mutex_unlock(&mapping->i_mmap_mutex);
 
 	mm->map_count++;
 	validate_mm(mm);
@@ -574,7 +574,7 @@ again:			remove_next = 1 + (end > next->
 		mapping = file->f_mapping;
 		if (!(vma->vm_flags & VM_NONLINEAR))
 			root = &mapping->i_mmap;
-		spin_lock(&mapping->i_mmap_lock);
+		mutex_lock(&mapping->i_mmap_mutex);
 		if (insert) {
 			/*
 			 * Put into prio_tree now, so instantiated pages
@@ -641,7 +641,7 @@ again:			remove_next = 1 + (end > next->
 	if (anon_vma)
 		anon_vma_unlock(anon_vma);
 	if (mapping)
-		spin_unlock(&mapping->i_mmap_lock);
+		mutex_unlock(&mapping->i_mmap_mutex);
 
 	if (remove_next) {
 		if (file) {
@@ -2300,7 +2300,7 @@ void exit_mmap(struct mm_struct *mm)
 
 /* Insert vm structure into process list sorted by address
  * and into the inode's i_mmap tree.  If vm_file is non-NULL
- * then i_mmap_lock is taken here.
+ * then i_mmap_mutex is taken here.
  */
 int insert_vm_struct(struct mm_struct * mm, struct vm_area_struct * vma)
 {
@@ -2542,7 +2542,7 @@ static void vm_lock_mapping(struct mm_st
 		 */
 		if (test_and_set_bit(AS_MM_ALL_LOCKS, &mapping->flags))
 			BUG();
-		spin_lock_nest_lock(&mapping->i_mmap_lock, &mm->mmap_sem);
+		mutex_lock_nest_lock(&mapping->i_mmap_mutex, &mm->mmap_sem);
 	}
 }
 
@@ -2569,7 +2569,7 @@ static void vm_lock_mapping(struct mm_st
  * vma in this mm is backed by the same anon_vma or address_space.
  *
  * We can take all the locks in random order because the VM code
- * taking i_mmap_lock or anon_vma->lock outside the mmap_sem never
+ * taking i_mmap_mutex or anon_vma->lock outside the mmap_sem never
  * takes more than one of them in a row. Secondly we're protected
  * against a concurrent mm_take_all_locks() by the mm_all_locks_mutex.
  *
@@ -2641,7 +2641,7 @@ static void vm_unlock_mapping(struct add
 		 * AS_MM_ALL_LOCKS can't change to 0 from under us
 		 * because we hold the mm_all_locks_mutex.
 		 */
-		spin_unlock(&mapping->i_mmap_lock);
+		mutex_unlock(&mapping->i_mmap_mutex);
 		if (!test_and_clear_bit(AS_MM_ALL_LOCKS,
 					&mapping->flags))
 			BUG();
Index: linux-2.6/mm/mremap.c
===================================================================
--- linux-2.6.orig/mm/mremap.c
+++ linux-2.6/mm/mremap.c
@@ -93,7 +93,7 @@ static void move_ptes(struct vm_area_str
 		 * and we propagate stale pages into the dst afterward.
 		 */
 		mapping = vma->vm_file->f_mapping;
-		spin_lock(&mapping->i_mmap_lock);
+		mutex_lock(&mapping->i_mmap_mutex);
 	}
 
 	/*
@@ -122,7 +122,7 @@ static void move_ptes(struct vm_area_str
 	pte_unmap(new_pte - 1);
 	pte_unmap_unlock(old_pte - 1, old_ptl);
 	if (mapping)
-		spin_unlock(&mapping->i_mmap_lock);
+		mutex_unlock(&mapping->i_mmap_mutex);
 	mmu_notifier_invalidate_range_end(vma->vm_mm, old_start, old_end);
 }
 
Index: linux-2.6/mm/rmap.c
===================================================================
--- linux-2.6.orig/mm/rmap.c
+++ linux-2.6/mm/rmap.c
@@ -24,7 +24,7 @@
  *   inode->i_alloc_sem (vmtruncate_range)
  *   mm->mmap_sem
  *     page->flags PG_locked (lock_page)
- *       mapping->i_mmap_lock
+ *       mapping->i_mmap_mutex
  *         anon_vma->lock
  *           mm->page_table_lock or pte_lock
  *             zone->lru_lock (in mark_page_accessed, isolate_lru_page)
@@ -646,14 +646,14 @@ static int page_referenced_file(struct p
 	 * The page lock not only makes sure that page->mapping cannot
 	 * suddenly be NULLified by truncation, it makes sure that the
 	 * structure at mapping cannot be freed and reused yet,
-	 * so we can safely take mapping->i_mmap_lock.
+	 * so we can safely take mapping->i_mmap_mutex.
 	 */
 	BUG_ON(!PageLocked(page));
 
-	spin_lock(&mapping->i_mmap_lock);
+	mutex_lock(&mapping->i_mmap_mutex);
 
 	/*
-	 * i_mmap_lock does not stabilize mapcount at all, but mapcount
+	 * i_mmap_mutex does not stabilize mapcount at all, but mapcount
 	 * is more likely to be accurate if we note it after spinning.
 	 */
 	mapcount = page_mapcount(page);
@@ -675,7 +675,7 @@ static int page_referenced_file(struct p
 			break;
 	}
 
-	spin_unlock(&mapping->i_mmap_lock);
+	mutex_unlock(&mapping->i_mmap_mutex);
 	return referenced;
 }
 
@@ -762,7 +762,7 @@ static int page_mkclean_file(struct addr
 
 	BUG_ON(PageAnon(page));
 
-	spin_lock(&mapping->i_mmap_lock);
+	mutex_lock(&mapping->i_mmap_mutex);
 	vma_prio_tree_foreach(vma, &iter, &mapping->i_mmap, pgoff, pgoff) {
 		if (vma->vm_flags & VM_SHARED) {
 			unsigned long address = vma_address(page, vma);
@@ -771,7 +771,7 @@ static int page_mkclean_file(struct addr
 			ret += page_mkclean_one(page, vma, address);
 		}
 	}
-	spin_unlock(&mapping->i_mmap_lock);
+	mutex_unlock(&mapping->i_mmap_mutex);
 	return ret;
 }
 
@@ -1122,7 +1122,7 @@ int try_to_unmap_one(struct page *page, 
 	/*
 	 * We need mmap_sem locking, Otherwise VM_LOCKED check makes
 	 * unstable result and race. Plus, We can't wait here because
-	 * we now hold anon_vma->lock or mapping->i_mmap_lock.
+	 * we now hold anon_vma->lock or mapping->i_mmap_mutex.
 	 * if trylock failed, the page remain in evictable lru and later
 	 * vmscan could retry to move the page to unevictable lru if the
 	 * page is actually mlocked.
@@ -1348,7 +1348,7 @@ static int try_to_unmap_file(struct page
 	unsigned long max_nl_size = 0;
 	unsigned int mapcount;
 
-	spin_lock(&mapping->i_mmap_lock);
+	mutex_lock(&mapping->i_mmap_mutex);
 	vma_prio_tree_foreach(vma, &iter, &mapping->i_mmap, pgoff, pgoff) {
 		unsigned long address = vma_address(page, vma);
 		if (address == -EFAULT)
@@ -1394,7 +1394,7 @@ static int try_to_unmap_file(struct page
 	mapcount = page_mapcount(page);
 	if (!mapcount)
 		goto out;
-	cond_resched_lock(&mapping->i_mmap_lock);
+	cond_resched();
 
 	max_nl_size = (max_nl_size + CLUSTER_SIZE - 1) & CLUSTER_MASK;
 	if (max_nl_cursor == 0)
@@ -1416,7 +1416,7 @@ static int try_to_unmap_file(struct page
 			}
 			vma->vm_private_data = (void *) max_nl_cursor;
 		}
-		cond_resched_lock(&mapping->i_mmap_lock);
+		cond_resched();
 		max_nl_cursor += CLUSTER_SIZE;
 	} while (max_nl_cursor <= max_nl_size);
 
@@ -1428,7 +1428,7 @@ static int try_to_unmap_file(struct page
 	list_for_each_entry(vma, &mapping->i_mmap_nonlinear, shared.vm_set.list)
 		vma->vm_private_data = NULL;
 out:
-	spin_unlock(&mapping->i_mmap_lock);
+	mutex_unlock(&mapping->i_mmap_mutex);
 	return ret;
 }
 
@@ -1547,7 +1547,7 @@ static int rmap_walk_file(struct page *p
 
 	if (!mapping)
 		return ret;
-	spin_lock(&mapping->i_mmap_lock);
+	mutex_lock(&mapping->i_mmap_mutex);
 	vma_prio_tree_foreach(vma, &iter, &mapping->i_mmap, pgoff, pgoff) {
 		unsigned long address = vma_address(page, vma);
 		if (address == -EFAULT)
@@ -1561,7 +1561,7 @@ static int rmap_walk_file(struct page *p
 	 * never contain migration ptes.  Decide what to do about this
 	 * limitation to linear when we need rmap_walk() on nonlinear.
 	 */
-	spin_unlock(&mapping->i_mmap_lock);
+	mutex_unlock(&mapping->i_mmap_mutex);
 	return ret;
 }
 
Index: linux-2.6/Documentation/lockstat.txt
===================================================================
--- linux-2.6.orig/Documentation/lockstat.txt
+++ linux-2.6/Documentation/lockstat.txt
@@ -136,7 +136,7 @@ The integer part of the time values is i
                              dcache_lock:          1037           1161           0.38          45.32         774.51           6611         243371           0.15         306.48       77387.24
                          &inode->i_mutex:           161            286 18446744073709       62882.54     1244614.55           3653          20598 18446744073709       62318.60     1693822.74
                          &zone->lru_lock:            94             94           0.53           7.33          92.10           4366          32690           0.29          59.81       16350.06
-              &inode->i_data.i_mmap_lock:            79             79           0.40           3.77          53.03          11779          87755           0.28         116.93       29898.44
+              &inode->i_data.i_mmap_mutex:            79             79           0.40           3.77          53.03          11779          87755           0.28         116.93       29898.44
                         &q->__queue_lock:            48             50           0.52          31.62          86.31            774          13131           0.17         113.08       12277.52
                         &rq->rq_lock_key:            43             47           0.74          68.50         170.63           3706          33929           0.22         107.99       17460.62
                       &rq->rq_lock_key#2:            39             46           0.75           6.68          49.03           2979          32292           0.17         125.17       17137.63
Index: linux-2.6/Documentation/vm/locking
===================================================================
--- linux-2.6.orig/Documentation/vm/locking
+++ linux-2.6/Documentation/vm/locking
@@ -66,7 +66,7 @@ in some cases it is not really needed. E
 expand_stack(), it is hard to come up with a destructive scenario without 
 having the vmlist protection in this case.
 
-The page_table_lock nests with the inode i_mmap_lock and the kmem cache
+The page_table_lock nests with the inode i_mmap_mutex and the kmem cache
 c_spinlock spinlocks.  This is okay, since the kmem code asks for pages after
 dropping c_spinlock.  The page_table_lock also nests with pagecache_lock and
 pagemap_lru_lock spinlocks, and no code asks for memory with these locks
Index: linux-2.6/include/linux/mmu_notifier.h
===================================================================
--- linux-2.6.orig/include/linux/mmu_notifier.h
+++ linux-2.6/include/linux/mmu_notifier.h
@@ -150,7 +150,7 @@ struct mmu_notifier_ops {
  * Therefore notifier chains can only be traversed when either
  *
  * 1. mmap_sem is held.
- * 2. One of the reverse map locks is held (i_mmap_lock or anon_vma->lock).
+ * 2. One of the reverse map locks is held (i_mmap_mutex or anon_vma->lock).
  * 3. No other concurrent thread can access the list (release)
  */
 struct mmu_notifier {
Index: linux-2.6/mm/filemap.c
===================================================================
--- linux-2.6.orig/mm/filemap.c
+++ linux-2.6/mm/filemap.c
@@ -58,16 +58,16 @@
 /*
  * Lock ordering:
  *
- *  ->i_mmap_lock		(truncate_pagecache)
+ *  ->i_mmap_mutex		(truncate_pagecache)
  *    ->private_lock		(__free_pte->__set_page_dirty_buffers)
  *      ->swap_lock		(exclusive_swap_page, others)
  *        ->mapping->tree_lock
  *
  *  ->i_mutex
- *    ->i_mmap_lock		(truncate->unmap_mapping_range)
+ *    ->i_mmap_mutex		(truncate->unmap_mapping_range)
  *
  *  ->mmap_sem
- *    ->i_mmap_lock
+ *    ->i_mmap_mutex
  *      ->page_table_lock or pte_lock	(various, mainly in memory.c)
  *        ->mapping->tree_lock	(arch-dependent flush_dcache_mmap_lock)
  *
@@ -84,7 +84,7 @@
  *    sb_lock			(fs/fs-writeback.c)
  *    ->mapping->tree_lock	(__sync_single_inode)
  *
- *  ->i_mmap_lock
+ *  ->i_mmap_mutex
  *    ->anon_vma.lock		(vma_adjust)
  *
  *  ->anon_vma.lock
@@ -106,7 +106,7 @@
  *
  *  (code doesn't rely on that order, so you could switch it around)
  *  ->tasklist_lock             (memory_failure, collect_procs_ao)
- *    ->i_mmap_lock
+ *    ->i_mmap_mutex
  */
 
 /*

[PATCH 15/20] mm: Convert i_mmap_lock to a mutex

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-mm-convert_i_mmap_lock_to_mutexes.patch --]
[-- Type: text/plain, Size: 21901 bytes --]

Straight fwd conversion of i_mmap_lock to a mutex

Acked-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 Documentation/lockstat.txt   |    2 +-
 Documentation/vm/locking     |    2 +-
 arch/x86/mm/hugetlbpage.c    |    4 ++--
 fs/hugetlbfs/inode.c         |    4 ++--
 fs/inode.c                   |    2 +-
 include/linux/fs.h           |    2 +-
 include/linux/mmu_notifier.h |    2 +-
 kernel/fork.c                |    4 ++--
 mm/filemap.c                 |   10 +++++-----
 mm/filemap_xip.c             |    4 ++--
 mm/fremap.c                  |    4 ++--
 mm/hugetlb.c                 |   14 +++++++-------
 mm/memory-failure.c          |    4 ++--
 mm/memory.c                  |    4 ++--
 mm/mmap.c                    |   22 +++++++++++-----------
 mm/mremap.c                  |    4 ++--
 mm/rmap.c                    |   28 ++++++++++++++--------------
 17 files changed, 58 insertions(+), 58 deletions(-)

Index: linux-2.6/arch/x86/mm/hugetlbpage.c
===================================================================
--- linux-2.6.orig/arch/x86/mm/hugetlbpage.c
+++ linux-2.6/arch/x86/mm/hugetlbpage.c
@@ -72,7 +72,7 @@ static void huge_pmd_share(struct mm_str
 	if (!vma_shareable(vma, addr))
 		return;
 
-	spin_lock(&mapping->i_mmap_lock);
+	mutex_lock(&mapping->i_mmap_mutex);
 	vma_prio_tree_foreach(svma, &iter, &mapping->i_mmap, idx, idx) {
 		if (svma == vma)
 			continue;
@@ -97,7 +97,7 @@ static void huge_pmd_share(struct mm_str
 		put_page(virt_to_page(spte));
 	spin_unlock(&mm->page_table_lock);
 out:
-	spin_unlock(&mapping->i_mmap_lock);
+	mutex_unlock(&mapping->i_mmap_mutex);
 }
 
 /*
Index: linux-2.6/fs/hugetlbfs/inode.c
===================================================================
--- linux-2.6.orig/fs/hugetlbfs/inode.c
+++ linux-2.6/fs/hugetlbfs/inode.c
@@ -412,10 +412,10 @@ static int hugetlb_vmtruncate(struct ino
 	pgoff = offset >> PAGE_SHIFT;
 
 	i_size_write(inode, offset);
-	spin_lock(&mapping->i_mmap_lock);
+	mutex_lock(&mapping->i_mmap_mutex);
 	if (!prio_tree_empty(&mapping->i_mmap))
 		hugetlb_vmtruncate_list(&mapping->i_mmap, pgoff);
-	spin_unlock(&mapping->i_mmap_lock);
+	mutex_unlock(&mapping->i_mmap_mutex);
 	truncate_hugepages(inode, offset);
 	return 0;
 }
Index: linux-2.6/fs/inode.c
===================================================================
--- linux-2.6.orig/fs/inode.c
+++ linux-2.6/fs/inode.c
@@ -318,7 +318,7 @@ void address_space_init_once(struct addr
 	memset(mapping, 0, sizeof(*mapping));
 	INIT_RADIX_TREE(&mapping->page_tree, GFP_ATOMIC);
 	spin_lock_init(&mapping->tree_lock);
-	spin_lock_init(&mapping->i_mmap_lock);
+	mutex_init(&mapping->i_mmap_mutex);
 	INIT_LIST_HEAD(&mapping->private_list);
 	spin_lock_init(&mapping->private_lock);
 	INIT_RAW_PRIO_TREE_ROOT(&mapping->i_mmap);
Index: linux-2.6/include/linux/fs.h
===================================================================
--- linux-2.6.orig/include/linux/fs.h
+++ linux-2.6/include/linux/fs.h
@@ -633,7 +633,7 @@ struct address_space {
 	unsigned int		i_mmap_writable;/* count VM_SHARED mappings */
 	struct prio_tree_root	i_mmap;		/* tree of private and shared mappings */
 	struct list_head	i_mmap_nonlinear;/*list VM_NONLINEAR mappings */
-	spinlock_t		i_mmap_lock;	/* protect tree, count, list */
+	struct mutex		i_mmap_mutex;	/* protect tree, count, list */
 	unsigned long		nrpages;	/* number of total pages */
 	pgoff_t			writeback_index;/* writeback starts here */
 	const struct address_space_operations *a_ops;	/* methods */
Index: linux-2.6/kernel/fork.c
===================================================================
--- linux-2.6.orig/kernel/fork.c
+++ linux-2.6/kernel/fork.c
@@ -383,14 +383,14 @@ static int dup_mmap(struct mm_struct *mm
 			get_file(file);
 			if (tmp->vm_flags & VM_DENYWRITE)
 				atomic_dec(&inode->i_writecount);
-			spin_lock(&mapping->i_mmap_lock);
+			mutex_lock(&mapping->i_mmap_mutex);
 			if (tmp->vm_flags & VM_SHARED)
 				mapping->i_mmap_writable++;
 			flush_dcache_mmap_lock(mapping);
 			/* insert tmp into the share list, just after mpnt */
 			vma_prio_tree_add(tmp, mpnt);
 			flush_dcache_mmap_unlock(mapping);
-			spin_unlock(&mapping->i_mmap_lock);
+			mutex_unlock(&mapping->i_mmap_mutex);
 		}
 
 		/*
Index: linux-2.6/mm/filemap_xip.c
===================================================================
--- linux-2.6.orig/mm/filemap_xip.c
+++ linux-2.6/mm/filemap_xip.c
@@ -183,7 +183,7 @@ __xip_unmap (struct address_space * mapp
 		return;
 
 retry:
-	spin_lock(&mapping->i_mmap_lock);
+	mutex_lock(&mapping->i_mmap_mutex);
 	vma_prio_tree_foreach(vma, &iter, &mapping->i_mmap, pgoff, pgoff) {
 		mm = vma->vm_mm;
 		address = vma->vm_start +
@@ -201,7 +201,7 @@ __xip_unmap (struct address_space * mapp
 			page_cache_release(page);
 		}
 	}
-	spin_unlock(&mapping->i_mmap_lock);
+	mutex_unlock(&mapping->i_mmap_mutex);
 
 	if (locked) {
 		mutex_unlock(&xip_sparse_mutex);
Index: linux-2.6/mm/fremap.c
===================================================================
--- linux-2.6.orig/mm/fremap.c
+++ linux-2.6/mm/fremap.c
@@ -211,13 +211,13 @@ SYSCALL_DEFINE5(remap_file_pages, unsign
 			}
 			goto out;
 		}
-		spin_lock(&mapping->i_mmap_lock);
+		mutex_lock(&mapping->i_mmap_mutex);
 		flush_dcache_mmap_lock(mapping);
 		vma->vm_flags |= VM_NONLINEAR;
 		vma_prio_tree_remove(vma, &mapping->i_mmap);
 		vma_nonlinear_insert(vma, &mapping->i_mmap_nonlinear);
 		flush_dcache_mmap_unlock(mapping);
-		spin_unlock(&mapping->i_mmap_lock);
+		mutex_unlock(&mapping->i_mmap_mutex);
 	}
 
 	if (vma->vm_flags & VM_LOCKED) {
Index: linux-2.6/mm/hugetlb.c
===================================================================
--- linux-2.6.orig/mm/hugetlb.c
+++ linux-2.6/mm/hugetlb.c
@@ -2205,7 +2205,7 @@ void __unmap_hugepage_range(struct vm_ar
 	unsigned long sz = huge_page_size(h);
 
 	/*
-	 * A page gathering list, protected by per file i_mmap_lock. The
+	 * A page gathering list, protected by per file i_mmap_mutex. The
 	 * lock is used to avoid list corruption from multiple unmapping
 	 * of the same page since we are using page->lru.
 	 */
@@ -2274,9 +2274,9 @@ void __unmap_hugepage_range(struct vm_ar
 void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
 			  unsigned long end, struct page *ref_page)
 {
-	spin_lock(&vma->vm_file->f_mapping->i_mmap_lock);
+	mutex_lock(&vma->vm_file->f_mapping->i_mmap_mutex);
 	__unmap_hugepage_range(vma, start, end, ref_page);
-	spin_unlock(&vma->vm_file->f_mapping->i_mmap_lock);
+	mutex_unlock(&vma->vm_file->f_mapping->i_mmap_mutex);
 }
 
 /*
@@ -2308,7 +2308,7 @@ static int unmap_ref_private(struct mm_s
 	 * this mapping should be shared between all the VMAs,
 	 * __unmap_hugepage_range() is called as the lock is already held
 	 */
-	spin_lock(&mapping->i_mmap_lock);
+	mutex_lock(&mapping->i_mmap_mutex);
 	vma_prio_tree_foreach(iter_vma, &iter, &mapping->i_mmap, pgoff, pgoff) {
 		/* Do not unmap the current VMA */
 		if (iter_vma == vma)
@@ -2326,7 +2326,7 @@ static int unmap_ref_private(struct mm_s
 				address, address + huge_page_size(h),
 				page);
 	}
-	spin_unlock(&mapping->i_mmap_lock);
+	mutex_unlock(&mapping->i_mmap_mutex);
 
 	return 1;
 }
@@ -2810,7 +2810,7 @@ void hugetlb_change_protection(struct vm
 	BUG_ON(address >= end);
 	flush_cache_range(vma, address, end);
 
-	spin_lock(&vma->vm_file->f_mapping->i_mmap_lock);
+	mutex_lock(&vma->vm_file->f_mapping->i_mmap_mutex);
 	spin_lock(&mm->page_table_lock);
 	for (; address < end; address += huge_page_size(h)) {
 		ptep = huge_pte_offset(mm, address);
@@ -2825,7 +2825,7 @@ void hugetlb_change_protection(struct vm
 		}
 	}
 	spin_unlock(&mm->page_table_lock);
-	spin_unlock(&vma->vm_file->f_mapping->i_mmap_lock);
+	mutex_unlock(&vma->vm_file->f_mapping->i_mmap_mutex);
 
 	flush_tlb_range(vma, start, end);
 }
Index: linux-2.6/mm/memory-failure.c
===================================================================
--- linux-2.6.orig/mm/memory-failure.c
+++ linux-2.6/mm/memory-failure.c
@@ -429,7 +429,7 @@ static void collect_procs_file(struct pa
 	 */
 
 	read_lock(&tasklist_lock);
-	spin_lock(&mapping->i_mmap_lock);
+	mutex_lock(&mapping->i_mmap_mutex);
 	for_each_process(tsk) {
 		pgoff_t pgoff = page->index << (PAGE_CACHE_SHIFT - PAGE_SHIFT);
 
@@ -449,7 +449,7 @@ static void collect_procs_file(struct pa
 				add_to_kill(tsk, page, vma, to_kill, tkc);
 		}
 	}
-	spin_unlock(&mapping->i_mmap_lock);
+	mutex_unlock(&mapping->i_mmap_mutex);
 	read_unlock(&tasklist_lock);
 }
 
Index: linux-2.6/mm/memory.c
===================================================================
--- linux-2.6.orig/mm/memory.c
+++ linux-2.6/mm/memory.c
@@ -2660,12 +2660,12 @@ void unmap_mapping_range(struct address_
 		details.last_index = ULONG_MAX;
 
 
-	spin_lock(&mapping->i_mmap_lock);
+	mutex_lock(&mapping->i_mmap_mutex);
 	if (unlikely(!prio_tree_empty(&mapping->i_mmap)))
 		unmap_mapping_range_tree(&mapping->i_mmap, &details);
 	if (unlikely(!list_empty(&mapping->i_mmap_nonlinear)))
 		unmap_mapping_range_list(&mapping->i_mmap_nonlinear, &details);
-	spin_unlock(&mapping->i_mmap_lock);
+	mutex_unlock(&mapping->i_mmap_mutex);
 }
 EXPORT_SYMBOL(unmap_mapping_range);
 
Index: linux-2.6/mm/mmap.c
===================================================================
--- linux-2.6.orig/mm/mmap.c
+++ linux-2.6/mm/mmap.c
@@ -190,7 +190,7 @@ int __vm_enough_memory(struct mm_struct 
 }
 
 /*
- * Requires inode->i_mapping->i_mmap_lock
+ * Requires inode->i_mapping->i_mmap_mutex
  */
 static void __remove_shared_vm_struct(struct vm_area_struct *vma,
 		struct file *file, struct address_space *mapping)
@@ -218,9 +218,9 @@ void unlink_file_vma(struct vm_area_stru
 
 	if (file) {
 		struct address_space *mapping = file->f_mapping;
-		spin_lock(&mapping->i_mmap_lock);
+		mutex_lock(&mapping->i_mmap_mutex);
 		__remove_shared_vm_struct(vma, file, mapping);
-		spin_unlock(&mapping->i_mmap_lock);
+		mutex_unlock(&mapping->i_mmap_mutex);
 	}
 }
 
@@ -465,13 +465,13 @@ static void vma_link(struct mm_struct *m
 		mapping = vma->vm_file->f_mapping;
 
 	if (mapping)
-		spin_lock(&mapping->i_mmap_lock);
+		mutex_lock(&mapping->i_mmap_mutex);
 
 	__vma_link(mm, vma, prev, rb_link, rb_parent);
 	__vma_link_file(vma);
 
 	if (mapping)
-		spin_unlock(&mapping->i_mmap_lock);
+		mutex_unlock(&mapping->i_mmap_mutex);
 
 	mm->map_count++;
 	validate_mm(mm);
@@ -574,7 +574,7 @@ again:			remove_next = 1 + (end > next->
 		mapping = file->f_mapping;
 		if (!(vma->vm_flags & VM_NONLINEAR))
 			root = &mapping->i_mmap;
-		spin_lock(&mapping->i_mmap_lock);
+		mutex_lock(&mapping->i_mmap_mutex);
 		if (insert) {
 			/*
 			 * Put into prio_tree now, so instantiated pages
@@ -641,7 +641,7 @@ again:			remove_next = 1 + (end > next->
 	if (anon_vma)
 		anon_vma_unlock(anon_vma);
 	if (mapping)
-		spin_unlock(&mapping->i_mmap_lock);
+		mutex_unlock(&mapping->i_mmap_mutex);
 
 	if (remove_next) {
 		if (file) {
@@ -2300,7 +2300,7 @@ void exit_mmap(struct mm_struct *mm)
 
 /* Insert vm structure into process list sorted by address
  * and into the inode's i_mmap tree.  If vm_file is non-NULL
- * then i_mmap_lock is taken here.
+ * then i_mmap_mutex is taken here.
  */
 int insert_vm_struct(struct mm_struct * mm, struct vm_area_struct * vma)
 {
@@ -2542,7 +2542,7 @@ static void vm_lock_mapping(struct mm_st
 		 */
 		if (test_and_set_bit(AS_MM_ALL_LOCKS, &mapping->flags))
 			BUG();
-		spin_lock_nest_lock(&mapping->i_mmap_lock, &mm->mmap_sem);
+		mutex_lock_nest_lock(&mapping->i_mmap_mutex, &mm->mmap_sem);
 	}
 }
 
@@ -2569,7 +2569,7 @@ static void vm_lock_mapping(struct mm_st
  * vma in this mm is backed by the same anon_vma or address_space.
  *
  * We can take all the locks in random order because the VM code
- * taking i_mmap_lock or anon_vma->lock outside the mmap_sem never
+ * taking i_mmap_mutex or anon_vma->lock outside the mmap_sem never
  * takes more than one of them in a row. Secondly we're protected
  * against a concurrent mm_take_all_locks() by the mm_all_locks_mutex.
  *
@@ -2641,7 +2641,7 @@ static void vm_unlock_mapping(struct add
 		 * AS_MM_ALL_LOCKS can't change to 0 from under us
 		 * because we hold the mm_all_locks_mutex.
 		 */
-		spin_unlock(&mapping->i_mmap_lock);
+		mutex_unlock(&mapping->i_mmap_mutex);
 		if (!test_and_clear_bit(AS_MM_ALL_LOCKS,
 					&mapping->flags))
 			BUG();
Index: linux-2.6/mm/mremap.c
===================================================================
--- linux-2.6.orig/mm/mremap.c
+++ linux-2.6/mm/mremap.c
@@ -93,7 +93,7 @@ static void move_ptes(struct vm_area_str
 		 * and we propagate stale pages into the dst afterward.
 		 */
 		mapping = vma->vm_file->f_mapping;
-		spin_lock(&mapping->i_mmap_lock);
+		mutex_lock(&mapping->i_mmap_mutex);
 	}
 
 	/*
@@ -122,7 +122,7 @@ static void move_ptes(struct vm_area_str
 	pte_unmap(new_pte - 1);
 	pte_unmap_unlock(old_pte - 1, old_ptl);
 	if (mapping)
-		spin_unlock(&mapping->i_mmap_lock);
+		mutex_unlock(&mapping->i_mmap_mutex);
 	mmu_notifier_invalidate_range_end(vma->vm_mm, old_start, old_end);
 }
 
Index: linux-2.6/mm/rmap.c
===================================================================
--- linux-2.6.orig/mm/rmap.c
+++ linux-2.6/mm/rmap.c
@@ -24,7 +24,7 @@
  *   inode->i_alloc_sem (vmtruncate_range)
  *   mm->mmap_sem
  *     page->flags PG_locked (lock_page)
- *       mapping->i_mmap_lock
+ *       mapping->i_mmap_mutex
  *         anon_vma->lock
  *           mm->page_table_lock or pte_lock
  *             zone->lru_lock (in mark_page_accessed, isolate_lru_page)
@@ -646,14 +646,14 @@ static int page_referenced_file(struct p
 	 * The page lock not only makes sure that page->mapping cannot
 	 * suddenly be NULLified by truncation, it makes sure that the
 	 * structure at mapping cannot be freed and reused yet,
-	 * so we can safely take mapping->i_mmap_lock.
+	 * so we can safely take mapping->i_mmap_mutex.
 	 */
 	BUG_ON(!PageLocked(page));
 
-	spin_lock(&mapping->i_mmap_lock);
+	mutex_lock(&mapping->i_mmap_mutex);
 
 	/*
-	 * i_mmap_lock does not stabilize mapcount at all, but mapcount
+	 * i_mmap_mutex does not stabilize mapcount at all, but mapcount
 	 * is more likely to be accurate if we note it after spinning.
 	 */
 	mapcount = page_mapcount(page);
@@ -675,7 +675,7 @@ static int page_referenced_file(struct p
 			break;
 	}
 
-	spin_unlock(&mapping->i_mmap_lock);
+	mutex_unlock(&mapping->i_mmap_mutex);
 	return referenced;
 }
 
@@ -762,7 +762,7 @@ static int page_mkclean_file(struct addr
 
 	BUG_ON(PageAnon(page));
 
-	spin_lock(&mapping->i_mmap_lock);
+	mutex_lock(&mapping->i_mmap_mutex);
 	vma_prio_tree_foreach(vma, &iter, &mapping->i_mmap, pgoff, pgoff) {
 		if (vma->vm_flags & VM_SHARED) {
 			unsigned long address = vma_address(page, vma);
@@ -771,7 +771,7 @@ static int page_mkclean_file(struct addr
 			ret += page_mkclean_one(page, vma, address);
 		}
 	}
-	spin_unlock(&mapping->i_mmap_lock);
+	mutex_unlock(&mapping->i_mmap_mutex);
 	return ret;
 }
 
@@ -1122,7 +1122,7 @@ int try_to_unmap_one(struct page *page, 
 	/*
 	 * We need mmap_sem locking, Otherwise VM_LOCKED check makes
 	 * unstable result and race. Plus, We can't wait here because
-	 * we now hold anon_vma->lock or mapping->i_mmap_lock.
+	 * we now hold anon_vma->lock or mapping->i_mmap_mutex.
 	 * if trylock failed, the page remain in evictable lru and later
 	 * vmscan could retry to move the page to unevictable lru if the
 	 * page is actually mlocked.
@@ -1348,7 +1348,7 @@ static int try_to_unmap_file(struct page
 	unsigned long max_nl_size = 0;
 	unsigned int mapcount;
 
-	spin_lock(&mapping->i_mmap_lock);
+	mutex_lock(&mapping->i_mmap_mutex);
 	vma_prio_tree_foreach(vma, &iter, &mapping->i_mmap, pgoff, pgoff) {
 		unsigned long address = vma_address(page, vma);
 		if (address == -EFAULT)
@@ -1394,7 +1394,7 @@ static int try_to_unmap_file(struct page
 	mapcount = page_mapcount(page);
 	if (!mapcount)
 		goto out;
-	cond_resched_lock(&mapping->i_mmap_lock);
+	cond_resched();
 
 	max_nl_size = (max_nl_size + CLUSTER_SIZE - 1) & CLUSTER_MASK;
 	if (max_nl_cursor == 0)
@@ -1416,7 +1416,7 @@ static int try_to_unmap_file(struct page
 			}
 			vma->vm_private_data = (void *) max_nl_cursor;
 		}
-		cond_resched_lock(&mapping->i_mmap_lock);
+		cond_resched();
 		max_nl_cursor += CLUSTER_SIZE;
 	} while (max_nl_cursor <= max_nl_size);
 
@@ -1428,7 +1428,7 @@ static int try_to_unmap_file(struct page
 	list_for_each_entry(vma, &mapping->i_mmap_nonlinear, shared.vm_set.list)
 		vma->vm_private_data = NULL;
 out:
-	spin_unlock(&mapping->i_mmap_lock);
+	mutex_unlock(&mapping->i_mmap_mutex);
 	return ret;
 }
 
@@ -1547,7 +1547,7 @@ static int rmap_walk_file(struct page *p
 
 	if (!mapping)
 		return ret;
-	spin_lock(&mapping->i_mmap_lock);
+	mutex_lock(&mapping->i_mmap_mutex);
 	vma_prio_tree_foreach(vma, &iter, &mapping->i_mmap, pgoff, pgoff) {
 		unsigned long address = vma_address(page, vma);
 		if (address == -EFAULT)
@@ -1561,7 +1561,7 @@ static int rmap_walk_file(struct page *p
 	 * never contain migration ptes.  Decide what to do about this
 	 * limitation to linear when we need rmap_walk() on nonlinear.
 	 */
-	spin_unlock(&mapping->i_mmap_lock);
+	mutex_unlock(&mapping->i_mmap_mutex);
 	return ret;
 }
 
Index: linux-2.6/Documentation/lockstat.txt
===================================================================
--- linux-2.6.orig/Documentation/lockstat.txt
+++ linux-2.6/Documentation/lockstat.txt
@@ -136,7 +136,7 @@ The integer part of the time values is i
                              dcache_lock:          1037           1161           0.38          45.32         774.51           6611         243371           0.15         306.48       77387.24
                          &inode->i_mutex:           161            286 18446744073709       62882.54     1244614.55           3653          20598 18446744073709       62318.60     1693822.74
                          &zone->lru_lock:            94             94           0.53           7.33          92.10           4366          32690           0.29          59.81       16350.06
-              &inode->i_data.i_mmap_lock:            79             79           0.40           3.77          53.03          11779          87755           0.28         116.93       29898.44
+              &inode->i_data.i_mmap_mutex:            79             79           0.40           3.77          53.03          11779          87755           0.28         116.93       29898.44
                         &q->__queue_lock:            48             50           0.52          31.62          86.31            774          13131           0.17         113.08       12277.52
                         &rq->rq_lock_key:            43             47           0.74          68.50         170.63           3706          33929           0.22         107.99       17460.62
                       &rq->rq_lock_key#2:            39             46           0.75           6.68          49.03           2979          32292           0.17         125.17       17137.63
Index: linux-2.6/Documentation/vm/locking
===================================================================
--- linux-2.6.orig/Documentation/vm/locking
+++ linux-2.6/Documentation/vm/locking
@@ -66,7 +66,7 @@ in some cases it is not really needed. E
 expand_stack(), it is hard to come up with a destructive scenario without 
 having the vmlist protection in this case.
 
-The page_table_lock nests with the inode i_mmap_lock and the kmem cache
+The page_table_lock nests with the inode i_mmap_mutex and the kmem cache
 c_spinlock spinlocks.  This is okay, since the kmem code asks for pages after
 dropping c_spinlock.  The page_table_lock also nests with pagecache_lock and
 pagemap_lru_lock spinlocks, and no code asks for memory with these locks
Index: linux-2.6/include/linux/mmu_notifier.h
===================================================================
--- linux-2.6.orig/include/linux/mmu_notifier.h
+++ linux-2.6/include/linux/mmu_notifier.h
@@ -150,7 +150,7 @@ struct mmu_notifier_ops {
  * Therefore notifier chains can only be traversed when either
  *
  * 1. mmap_sem is held.
- * 2. One of the reverse map locks is held (i_mmap_lock or anon_vma->lock).
+ * 2. One of the reverse map locks is held (i_mmap_mutex or anon_vma->lock).
  * 3. No other concurrent thread can access the list (release)
  */
 struct mmu_notifier {
Index: linux-2.6/mm/filemap.c
===================================================================
--- linux-2.6.orig/mm/filemap.c
+++ linux-2.6/mm/filemap.c
@@ -58,16 +58,16 @@
 /*
  * Lock ordering:
  *
- *  ->i_mmap_lock		(truncate_pagecache)
+ *  ->i_mmap_mutex		(truncate_pagecache)
  *    ->private_lock		(__free_pte->__set_page_dirty_buffers)
  *      ->swap_lock		(exclusive_swap_page, others)
  *        ->mapping->tree_lock
  *
  *  ->i_mutex
- *    ->i_mmap_lock		(truncate->unmap_mapping_range)
+ *    ->i_mmap_mutex		(truncate->unmap_mapping_range)
  *
  *  ->mmap_sem
- *    ->i_mmap_lock
+ *    ->i_mmap_mutex
  *      ->page_table_lock or pte_lock	(various, mainly in memory.c)
  *        ->mapping->tree_lock	(arch-dependent flush_dcache_mmap_lock)
  *
@@ -84,7 +84,7 @@
  *    sb_lock			(fs/fs-writeback.c)
  *    ->mapping->tree_lock	(__sync_single_inode)
  *
- *  ->i_mmap_lock
+ *  ->i_mmap_mutex
  *    ->anon_vma.lock		(vma_adjust)
  *
  *  ->anon_vma.lock
@@ -106,7 +106,7 @@
  *
  *  (code doesn't rely on that order, so you could switch it around)
  *  ->tasklist_lock             (memory_failure, collect_procs_ao)
- *    ->i_mmap_lock
+ *    ->i_mmap_mutex
  */
 
 /*


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 15/20] mm: Convert i_mmap_lock to a mutex

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-mm-convert_i_mmap_lock_to_mutexes.patch --]
[-- Type: text/plain, Size: 21901 bytes --]

Straight fwd conversion of i_mmap_lock to a mutex

Acked-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 Documentation/lockstat.txt   |    2 +-
 Documentation/vm/locking     |    2 +-
 arch/x86/mm/hugetlbpage.c    |    4 ++--
 fs/hugetlbfs/inode.c         |    4 ++--
 fs/inode.c                   |    2 +-
 include/linux/fs.h           |    2 +-
 include/linux/mmu_notifier.h |    2 +-
 kernel/fork.c                |    4 ++--
 mm/filemap.c                 |   10 +++++-----
 mm/filemap_xip.c             |    4 ++--
 mm/fremap.c                  |    4 ++--
 mm/hugetlb.c                 |   14 +++++++-------
 mm/memory-failure.c          |    4 ++--
 mm/memory.c                  |    4 ++--
 mm/mmap.c                    |   22 +++++++++++-----------
 mm/mremap.c                  |    4 ++--
 mm/rmap.c                    |   28 ++++++++++++++--------------
 17 files changed, 58 insertions(+), 58 deletions(-)

Index: linux-2.6/arch/x86/mm/hugetlbpage.c
===================================================================
--- linux-2.6.orig/arch/x86/mm/hugetlbpage.c
+++ linux-2.6/arch/x86/mm/hugetlbpage.c
@@ -72,7 +72,7 @@ static void huge_pmd_share(struct mm_str
 	if (!vma_shareable(vma, addr))
 		return;
 
-	spin_lock(&mapping->i_mmap_lock);
+	mutex_lock(&mapping->i_mmap_mutex);
 	vma_prio_tree_foreach(svma, &iter, &mapping->i_mmap, idx, idx) {
 		if (svma == vma)
 			continue;
@@ -97,7 +97,7 @@ static void huge_pmd_share(struct mm_str
 		put_page(virt_to_page(spte));
 	spin_unlock(&mm->page_table_lock);
 out:
-	spin_unlock(&mapping->i_mmap_lock);
+	mutex_unlock(&mapping->i_mmap_mutex);
 }
 
 /*
Index: linux-2.6/fs/hugetlbfs/inode.c
===================================================================
--- linux-2.6.orig/fs/hugetlbfs/inode.c
+++ linux-2.6/fs/hugetlbfs/inode.c
@@ -412,10 +412,10 @@ static int hugetlb_vmtruncate(struct ino
 	pgoff = offset >> PAGE_SHIFT;
 
 	i_size_write(inode, offset);
-	spin_lock(&mapping->i_mmap_lock);
+	mutex_lock(&mapping->i_mmap_mutex);
 	if (!prio_tree_empty(&mapping->i_mmap))
 		hugetlb_vmtruncate_list(&mapping->i_mmap, pgoff);
-	spin_unlock(&mapping->i_mmap_lock);
+	mutex_unlock(&mapping->i_mmap_mutex);
 	truncate_hugepages(inode, offset);
 	return 0;
 }
Index: linux-2.6/fs/inode.c
===================================================================
--- linux-2.6.orig/fs/inode.c
+++ linux-2.6/fs/inode.c
@@ -318,7 +318,7 @@ void address_space_init_once(struct addr
 	memset(mapping, 0, sizeof(*mapping));
 	INIT_RADIX_TREE(&mapping->page_tree, GFP_ATOMIC);
 	spin_lock_init(&mapping->tree_lock);
-	spin_lock_init(&mapping->i_mmap_lock);
+	mutex_init(&mapping->i_mmap_mutex);
 	INIT_LIST_HEAD(&mapping->private_list);
 	spin_lock_init(&mapping->private_lock);
 	INIT_RAW_PRIO_TREE_ROOT(&mapping->i_mmap);
Index: linux-2.6/include/linux/fs.h
===================================================================
--- linux-2.6.orig/include/linux/fs.h
+++ linux-2.6/include/linux/fs.h
@@ -633,7 +633,7 @@ struct address_space {
 	unsigned int		i_mmap_writable;/* count VM_SHARED mappings */
 	struct prio_tree_root	i_mmap;		/* tree of private and shared mappings */
 	struct list_head	i_mmap_nonlinear;/*list VM_NONLINEAR mappings */
-	spinlock_t		i_mmap_lock;	/* protect tree, count, list */
+	struct mutex		i_mmap_mutex;	/* protect tree, count, list */
 	unsigned long		nrpages;	/* number of total pages */
 	pgoff_t			writeback_index;/* writeback starts here */
 	const struct address_space_operations *a_ops;	/* methods */
Index: linux-2.6/kernel/fork.c
===================================================================
--- linux-2.6.orig/kernel/fork.c
+++ linux-2.6/kernel/fork.c
@@ -383,14 +383,14 @@ static int dup_mmap(struct mm_struct *mm
 			get_file(file);
 			if (tmp->vm_flags & VM_DENYWRITE)
 				atomic_dec(&inode->i_writecount);
-			spin_lock(&mapping->i_mmap_lock);
+			mutex_lock(&mapping->i_mmap_mutex);
 			if (tmp->vm_flags & VM_SHARED)
 				mapping->i_mmap_writable++;
 			flush_dcache_mmap_lock(mapping);
 			/* insert tmp into the share list, just after mpnt */
 			vma_prio_tree_add(tmp, mpnt);
 			flush_dcache_mmap_unlock(mapping);
-			spin_unlock(&mapping->i_mmap_lock);
+			mutex_unlock(&mapping->i_mmap_mutex);
 		}
 
 		/*
Index: linux-2.6/mm/filemap_xip.c
===================================================================
--- linux-2.6.orig/mm/filemap_xip.c
+++ linux-2.6/mm/filemap_xip.c
@@ -183,7 +183,7 @@ __xip_unmap (struct address_space * mapp
 		return;
 
 retry:
-	spin_lock(&mapping->i_mmap_lock);
+	mutex_lock(&mapping->i_mmap_mutex);
 	vma_prio_tree_foreach(vma, &iter, &mapping->i_mmap, pgoff, pgoff) {
 		mm = vma->vm_mm;
 		address = vma->vm_start +
@@ -201,7 +201,7 @@ __xip_unmap (struct address_space * mapp
 			page_cache_release(page);
 		}
 	}
-	spin_unlock(&mapping->i_mmap_lock);
+	mutex_unlock(&mapping->i_mmap_mutex);
 
 	if (locked) {
 		mutex_unlock(&xip_sparse_mutex);
Index: linux-2.6/mm/fremap.c
===================================================================
--- linux-2.6.orig/mm/fremap.c
+++ linux-2.6/mm/fremap.c
@@ -211,13 +211,13 @@ SYSCALL_DEFINE5(remap_file_pages, unsign
 			}
 			goto out;
 		}
-		spin_lock(&mapping->i_mmap_lock);
+		mutex_lock(&mapping->i_mmap_mutex);
 		flush_dcache_mmap_lock(mapping);
 		vma->vm_flags |= VM_NONLINEAR;
 		vma_prio_tree_remove(vma, &mapping->i_mmap);
 		vma_nonlinear_insert(vma, &mapping->i_mmap_nonlinear);
 		flush_dcache_mmap_unlock(mapping);
-		spin_unlock(&mapping->i_mmap_lock);
+		mutex_unlock(&mapping->i_mmap_mutex);
 	}
 
 	if (vma->vm_flags & VM_LOCKED) {
Index: linux-2.6/mm/hugetlb.c
===================================================================
--- linux-2.6.orig/mm/hugetlb.c
+++ linux-2.6/mm/hugetlb.c
@@ -2205,7 +2205,7 @@ void __unmap_hugepage_range(struct vm_ar
 	unsigned long sz = huge_page_size(h);
 
 	/*
-	 * A page gathering list, protected by per file i_mmap_lock. The
+	 * A page gathering list, protected by per file i_mmap_mutex. The
 	 * lock is used to avoid list corruption from multiple unmapping
 	 * of the same page since we are using page->lru.
 	 */
@@ -2274,9 +2274,9 @@ void __unmap_hugepage_range(struct vm_ar
 void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
 			  unsigned long end, struct page *ref_page)
 {
-	spin_lock(&vma->vm_file->f_mapping->i_mmap_lock);
+	mutex_lock(&vma->vm_file->f_mapping->i_mmap_mutex);
 	__unmap_hugepage_range(vma, start, end, ref_page);
-	spin_unlock(&vma->vm_file->f_mapping->i_mmap_lock);
+	mutex_unlock(&vma->vm_file->f_mapping->i_mmap_mutex);
 }
 
 /*
@@ -2308,7 +2308,7 @@ static int unmap_ref_private(struct mm_s
 	 * this mapping should be shared between all the VMAs,
 	 * __unmap_hugepage_range() is called as the lock is already held
 	 */
-	spin_lock(&mapping->i_mmap_lock);
+	mutex_lock(&mapping->i_mmap_mutex);
 	vma_prio_tree_foreach(iter_vma, &iter, &mapping->i_mmap, pgoff, pgoff) {
 		/* Do not unmap the current VMA */
 		if (iter_vma == vma)
@@ -2326,7 +2326,7 @@ static int unmap_ref_private(struct mm_s
 				address, address + huge_page_size(h),
 				page);
 	}
-	spin_unlock(&mapping->i_mmap_lock);
+	mutex_unlock(&mapping->i_mmap_mutex);
 
 	return 1;
 }
@@ -2810,7 +2810,7 @@ void hugetlb_change_protection(struct vm
 	BUG_ON(address >= end);
 	flush_cache_range(vma, address, end);
 
-	spin_lock(&vma->vm_file->f_mapping->i_mmap_lock);
+	mutex_lock(&vma->vm_file->f_mapping->i_mmap_mutex);
 	spin_lock(&mm->page_table_lock);
 	for (; address < end; address += huge_page_size(h)) {
 		ptep = huge_pte_offset(mm, address);
@@ -2825,7 +2825,7 @@ void hugetlb_change_protection(struct vm
 		}
 	}
 	spin_unlock(&mm->page_table_lock);
-	spin_unlock(&vma->vm_file->f_mapping->i_mmap_lock);
+	mutex_unlock(&vma->vm_file->f_mapping->i_mmap_mutex);
 
 	flush_tlb_range(vma, start, end);
 }
Index: linux-2.6/mm/memory-failure.c
===================================================================
--- linux-2.6.orig/mm/memory-failure.c
+++ linux-2.6/mm/memory-failure.c
@@ -429,7 +429,7 @@ static void collect_procs_file(struct pa
 	 */
 
 	read_lock(&tasklist_lock);
-	spin_lock(&mapping->i_mmap_lock);
+	mutex_lock(&mapping->i_mmap_mutex);
 	for_each_process(tsk) {
 		pgoff_t pgoff = page->index << (PAGE_CACHE_SHIFT - PAGE_SHIFT);
 
@@ -449,7 +449,7 @@ static void collect_procs_file(struct pa
 				add_to_kill(tsk, page, vma, to_kill, tkc);
 		}
 	}
-	spin_unlock(&mapping->i_mmap_lock);
+	mutex_unlock(&mapping->i_mmap_mutex);
 	read_unlock(&tasklist_lock);
 }
 
Index: linux-2.6/mm/memory.c
===================================================================
--- linux-2.6.orig/mm/memory.c
+++ linux-2.6/mm/memory.c
@@ -2660,12 +2660,12 @@ void unmap_mapping_range(struct address_
 		details.last_index = ULONG_MAX;
 
 
-	spin_lock(&mapping->i_mmap_lock);
+	mutex_lock(&mapping->i_mmap_mutex);
 	if (unlikely(!prio_tree_empty(&mapping->i_mmap)))
 		unmap_mapping_range_tree(&mapping->i_mmap, &details);
 	if (unlikely(!list_empty(&mapping->i_mmap_nonlinear)))
 		unmap_mapping_range_list(&mapping->i_mmap_nonlinear, &details);
-	spin_unlock(&mapping->i_mmap_lock);
+	mutex_unlock(&mapping->i_mmap_mutex);
 }
 EXPORT_SYMBOL(unmap_mapping_range);
 
Index: linux-2.6/mm/mmap.c
===================================================================
--- linux-2.6.orig/mm/mmap.c
+++ linux-2.6/mm/mmap.c
@@ -190,7 +190,7 @@ int __vm_enough_memory(struct mm_struct 
 }
 
 /*
- * Requires inode->i_mapping->i_mmap_lock
+ * Requires inode->i_mapping->i_mmap_mutex
  */
 static void __remove_shared_vm_struct(struct vm_area_struct *vma,
 		struct file *file, struct address_space *mapping)
@@ -218,9 +218,9 @@ void unlink_file_vma(struct vm_area_stru
 
 	if (file) {
 		struct address_space *mapping = file->f_mapping;
-		spin_lock(&mapping->i_mmap_lock);
+		mutex_lock(&mapping->i_mmap_mutex);
 		__remove_shared_vm_struct(vma, file, mapping);
-		spin_unlock(&mapping->i_mmap_lock);
+		mutex_unlock(&mapping->i_mmap_mutex);
 	}
 }
 
@@ -465,13 +465,13 @@ static void vma_link(struct mm_struct *m
 		mapping = vma->vm_file->f_mapping;
 
 	if (mapping)
-		spin_lock(&mapping->i_mmap_lock);
+		mutex_lock(&mapping->i_mmap_mutex);
 
 	__vma_link(mm, vma, prev, rb_link, rb_parent);
 	__vma_link_file(vma);
 
 	if (mapping)
-		spin_unlock(&mapping->i_mmap_lock);
+		mutex_unlock(&mapping->i_mmap_mutex);
 
 	mm->map_count++;
 	validate_mm(mm);
@@ -574,7 +574,7 @@ again:			remove_next = 1 + (end > next->
 		mapping = file->f_mapping;
 		if (!(vma->vm_flags & VM_NONLINEAR))
 			root = &mapping->i_mmap;
-		spin_lock(&mapping->i_mmap_lock);
+		mutex_lock(&mapping->i_mmap_mutex);
 		if (insert) {
 			/*
 			 * Put into prio_tree now, so instantiated pages
@@ -641,7 +641,7 @@ again:			remove_next = 1 + (end > next->
 	if (anon_vma)
 		anon_vma_unlock(anon_vma);
 	if (mapping)
-		spin_unlock(&mapping->i_mmap_lock);
+		mutex_unlock(&mapping->i_mmap_mutex);
 
 	if (remove_next) {
 		if (file) {
@@ -2300,7 +2300,7 @@ void exit_mmap(struct mm_struct *mm)
 
 /* Insert vm structure into process list sorted by address
  * and into the inode's i_mmap tree.  If vm_file is non-NULL
- * then i_mmap_lock is taken here.
+ * then i_mmap_mutex is taken here.
  */
 int insert_vm_struct(struct mm_struct * mm, struct vm_area_struct * vma)
 {
@@ -2542,7 +2542,7 @@ static void vm_lock_mapping(struct mm_st
 		 */
 		if (test_and_set_bit(AS_MM_ALL_LOCKS, &mapping->flags))
 			BUG();
-		spin_lock_nest_lock(&mapping->i_mmap_lock, &mm->mmap_sem);
+		mutex_lock_nest_lock(&mapping->i_mmap_mutex, &mm->mmap_sem);
 	}
 }
 
@@ -2569,7 +2569,7 @@ static void vm_lock_mapping(struct mm_st
  * vma in this mm is backed by the same anon_vma or address_space.
  *
  * We can take all the locks in random order because the VM code
- * taking i_mmap_lock or anon_vma->lock outside the mmap_sem never
+ * taking i_mmap_mutex or anon_vma->lock outside the mmap_sem never
  * takes more than one of them in a row. Secondly we're protected
  * against a concurrent mm_take_all_locks() by the mm_all_locks_mutex.
  *
@@ -2641,7 +2641,7 @@ static void vm_unlock_mapping(struct add
 		 * AS_MM_ALL_LOCKS can't change to 0 from under us
 		 * because we hold the mm_all_locks_mutex.
 		 */
-		spin_unlock(&mapping->i_mmap_lock);
+		mutex_unlock(&mapping->i_mmap_mutex);
 		if (!test_and_clear_bit(AS_MM_ALL_LOCKS,
 					&mapping->flags))
 			BUG();
Index: linux-2.6/mm/mremap.c
===================================================================
--- linux-2.6.orig/mm/mremap.c
+++ linux-2.6/mm/mremap.c
@@ -93,7 +93,7 @@ static void move_ptes(struct vm_area_str
 		 * and we propagate stale pages into the dst afterward.
 		 */
 		mapping = vma->vm_file->f_mapping;
-		spin_lock(&mapping->i_mmap_lock);
+		mutex_lock(&mapping->i_mmap_mutex);
 	}
 
 	/*
@@ -122,7 +122,7 @@ static void move_ptes(struct vm_area_str
 	pte_unmap(new_pte - 1);
 	pte_unmap_unlock(old_pte - 1, old_ptl);
 	if (mapping)
-		spin_unlock(&mapping->i_mmap_lock);
+		mutex_unlock(&mapping->i_mmap_mutex);
 	mmu_notifier_invalidate_range_end(vma->vm_mm, old_start, old_end);
 }
 
Index: linux-2.6/mm/rmap.c
===================================================================
--- linux-2.6.orig/mm/rmap.c
+++ linux-2.6/mm/rmap.c
@@ -24,7 +24,7 @@
  *   inode->i_alloc_sem (vmtruncate_range)
  *   mm->mmap_sem
  *     page->flags PG_locked (lock_page)
- *       mapping->i_mmap_lock
+ *       mapping->i_mmap_mutex
  *         anon_vma->lock
  *           mm->page_table_lock or pte_lock
  *             zone->lru_lock (in mark_page_accessed, isolate_lru_page)
@@ -646,14 +646,14 @@ static int page_referenced_file(struct p
 	 * The page lock not only makes sure that page->mapping cannot
 	 * suddenly be NULLified by truncation, it makes sure that the
 	 * structure at mapping cannot be freed and reused yet,
-	 * so we can safely take mapping->i_mmap_lock.
+	 * so we can safely take mapping->i_mmap_mutex.
 	 */
 	BUG_ON(!PageLocked(page));
 
-	spin_lock(&mapping->i_mmap_lock);
+	mutex_lock(&mapping->i_mmap_mutex);
 
 	/*
-	 * i_mmap_lock does not stabilize mapcount at all, but mapcount
+	 * i_mmap_mutex does not stabilize mapcount at all, but mapcount
 	 * is more likely to be accurate if we note it after spinning.
 	 */
 	mapcount = page_mapcount(page);
@@ -675,7 +675,7 @@ static int page_referenced_file(struct p
 			break;
 	}
 
-	spin_unlock(&mapping->i_mmap_lock);
+	mutex_unlock(&mapping->i_mmap_mutex);
 	return referenced;
 }
 
@@ -762,7 +762,7 @@ static int page_mkclean_file(struct addr
 
 	BUG_ON(PageAnon(page));
 
-	spin_lock(&mapping->i_mmap_lock);
+	mutex_lock(&mapping->i_mmap_mutex);
 	vma_prio_tree_foreach(vma, &iter, &mapping->i_mmap, pgoff, pgoff) {
 		if (vma->vm_flags & VM_SHARED) {
 			unsigned long address = vma_address(page, vma);
@@ -771,7 +771,7 @@ static int page_mkclean_file(struct addr
 			ret += page_mkclean_one(page, vma, address);
 		}
 	}
-	spin_unlock(&mapping->i_mmap_lock);
+	mutex_unlock(&mapping->i_mmap_mutex);
 	return ret;
 }
 
@@ -1122,7 +1122,7 @@ int try_to_unmap_one(struct page *page, 
 	/*
 	 * We need mmap_sem locking, Otherwise VM_LOCKED check makes
 	 * unstable result and race. Plus, We can't wait here because
-	 * we now hold anon_vma->lock or mapping->i_mmap_lock.
+	 * we now hold anon_vma->lock or mapping->i_mmap_mutex.
 	 * if trylock failed, the page remain in evictable lru and later
 	 * vmscan could retry to move the page to unevictable lru if the
 	 * page is actually mlocked.
@@ -1348,7 +1348,7 @@ static int try_to_unmap_file(struct page
 	unsigned long max_nl_size = 0;
 	unsigned int mapcount;
 
-	spin_lock(&mapping->i_mmap_lock);
+	mutex_lock(&mapping->i_mmap_mutex);
 	vma_prio_tree_foreach(vma, &iter, &mapping->i_mmap, pgoff, pgoff) {
 		unsigned long address = vma_address(page, vma);
 		if (address == -EFAULT)
@@ -1394,7 +1394,7 @@ static int try_to_unmap_file(struct page
 	mapcount = page_mapcount(page);
 	if (!mapcount)
 		goto out;
-	cond_resched_lock(&mapping->i_mmap_lock);
+	cond_resched();
 
 	max_nl_size = (max_nl_size + CLUSTER_SIZE - 1) & CLUSTER_MASK;
 	if (max_nl_cursor == 0)
@@ -1416,7 +1416,7 @@ static int try_to_unmap_file(struct page
 			}
 			vma->vm_private_data = (void *) max_nl_cursor;
 		}
-		cond_resched_lock(&mapping->i_mmap_lock);
+		cond_resched();
 		max_nl_cursor += CLUSTER_SIZE;
 	} while (max_nl_cursor <= max_nl_size);
 
@@ -1428,7 +1428,7 @@ static int try_to_unmap_file(struct page
 	list_for_each_entry(vma, &mapping->i_mmap_nonlinear, shared.vm_set.list)
 		vma->vm_private_data = NULL;
 out:
-	spin_unlock(&mapping->i_mmap_lock);
+	mutex_unlock(&mapping->i_mmap_mutex);
 	return ret;
 }
 
@@ -1547,7 +1547,7 @@ static int rmap_walk_file(struct page *p
 
 	if (!mapping)
 		return ret;
-	spin_lock(&mapping->i_mmap_lock);
+	mutex_lock(&mapping->i_mmap_mutex);
 	vma_prio_tree_foreach(vma, &iter, &mapping->i_mmap, pgoff, pgoff) {
 		unsigned long address = vma_address(page, vma);
 		if (address == -EFAULT)
@@ -1561,7 +1561,7 @@ static int rmap_walk_file(struct page *p
 	 * never contain migration ptes.  Decide what to do about this
 	 * limitation to linear when we need rmap_walk() on nonlinear.
 	 */
-	spin_unlock(&mapping->i_mmap_lock);
+	mutex_unlock(&mapping->i_mmap_mutex);
 	return ret;
 }
 
Index: linux-2.6/Documentation/lockstat.txt
===================================================================
--- linux-2.6.orig/Documentation/lockstat.txt
+++ linux-2.6/Documentation/lockstat.txt
@@ -136,7 +136,7 @@ The integer part of the time values is i
                              dcache_lock:          1037           1161           0.38          45.32         774.51           6611         243371           0.15         306.48       77387.24
                          &inode->i_mutex:           161            286 18446744073709       62882.54     1244614.55           3653          20598 18446744073709       62318.60     1693822.74
                          &zone->lru_lock:            94             94           0.53           7.33          92.10           4366          32690           0.29          59.81       16350.06
-              &inode->i_data.i_mmap_lock:            79             79           0.40           3.77          53.03          11779          87755           0.28         116.93       29898.44
+              &inode->i_data.i_mmap_mutex:            79             79           0.40           3.77          53.03          11779          87755           0.28         116.93       29898.44
                         &q->__queue_lock:            48             50           0.52          31.62          86.31            774          13131           0.17         113.08       12277.52
                         &rq->rq_lock_key:            43             47           0.74          68.50         170.63           3706          33929           0.22         107.99       17460.62
                       &rq->rq_lock_key#2:            39             46           0.75           6.68          49.03           2979          32292           0.17         125.17       17137.63
Index: linux-2.6/Documentation/vm/locking
===================================================================
--- linux-2.6.orig/Documentation/vm/locking
+++ linux-2.6/Documentation/vm/locking
@@ -66,7 +66,7 @@ in some cases it is not really needed. E
 expand_stack(), it is hard to come up with a destructive scenario without 
 having the vmlist protection in this case.
 
-The page_table_lock nests with the inode i_mmap_lock and the kmem cache
+The page_table_lock nests with the inode i_mmap_mutex and the kmem cache
 c_spinlock spinlocks.  This is okay, since the kmem code asks for pages after
 dropping c_spinlock.  The page_table_lock also nests with pagecache_lock and
 pagemap_lru_lock spinlocks, and no code asks for memory with these locks
Index: linux-2.6/include/linux/mmu_notifier.h
===================================================================
--- linux-2.6.orig/include/linux/mmu_notifier.h
+++ linux-2.6/include/linux/mmu_notifier.h
@@ -150,7 +150,7 @@ struct mmu_notifier_ops {
  * Therefore notifier chains can only be traversed when either
  *
  * 1. mmap_sem is held.
- * 2. One of the reverse map locks is held (i_mmap_lock or anon_vma->lock).
+ * 2. One of the reverse map locks is held (i_mmap_mutex or anon_vma->lock).
  * 3. No other concurrent thread can access the list (release)
  */
 struct mmu_notifier {
Index: linux-2.6/mm/filemap.c
===================================================================
--- linux-2.6.orig/mm/filemap.c
+++ linux-2.6/mm/filemap.c
@@ -58,16 +58,16 @@
 /*
  * Lock ordering:
  *
- *  ->i_mmap_lock		(truncate_pagecache)
+ *  ->i_mmap_mutex		(truncate_pagecache)
  *    ->private_lock		(__free_pte->__set_page_dirty_buffers)
  *      ->swap_lock		(exclusive_swap_page, others)
  *        ->mapping->tree_lock
  *
  *  ->i_mutex
- *    ->i_mmap_lock		(truncate->unmap_mapping_range)
+ *    ->i_mmap_mutex		(truncate->unmap_mapping_range)
  *
  *  ->mmap_sem
- *    ->i_mmap_lock
+ *    ->i_mmap_mutex
  *      ->page_table_lock or pte_lock	(various, mainly in memory.c)
  *        ->mapping->tree_lock	(arch-dependent flush_dcache_mmap_lock)
  *
@@ -84,7 +84,7 @@
  *    sb_lock			(fs/fs-writeback.c)
  *    ->mapping->tree_lock	(__sync_single_inode)
  *
- *  ->i_mmap_lock
+ *  ->i_mmap_mutex
  *    ->anon_vma.lock		(vma_adjust)
  *
  *  ->anon_vma.lock
@@ -106,7 +106,7 @@
  *
  *  (code doesn't rely on that order, so you could switch it around)
  *  ->tasklist_lock             (memory_failure, collect_procs_ao)
- *    ->i_mmap_lock
+ *    ->i_mmap_mutex
  */
 
 /*


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 16/20] mm: Revert page_lock_anon_vma() lock annotation

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-mm-revert_page_lock_anon_vma_lock_annotation.patch --]
[-- Type: text/plain, Size: 1937 bytes --]

Its beyond ugly and gets in the way.

Cc: Namhyung Kim <namhyung@gmail.com>
Acked-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 include/linux/rmap.h |   15 +--------------
 mm/rmap.c            |    4 +---
 2 files changed, 2 insertions(+), 17 deletions(-)

Index: linux-2.6/include/linux/rmap.h
===================================================================
--- linux-2.6.orig/include/linux/rmap.h
+++ linux-2.6/include/linux/rmap.h
@@ -243,20 +243,7 @@ int try_to_munlock(struct page *);
 /*
  * Called by memory-failure.c to kill processes.
  */
-struct anon_vma *__page_lock_anon_vma(struct page *page);
-
-static inline struct anon_vma *page_lock_anon_vma(struct page *page)
-{
-	struct anon_vma *anon_vma;
-
-	__cond_lock(RCU, anon_vma = __page_lock_anon_vma(page));
-
-	/* (void) is needed to make gcc happy */
-	(void) __cond_lock(&anon_vma->root->lock, anon_vma);
-
-	return anon_vma;
-}
-
+struct anon_vma *page_lock_anon_vma(struct page *page);
 void page_unlock_anon_vma(struct anon_vma *anon_vma);
 int page_mapped_in_vma(struct page *page, struct vm_area_struct *vma);
 
Index: linux-2.6/mm/rmap.c
===================================================================
--- linux-2.6.orig/mm/rmap.c
+++ linux-2.6/mm/rmap.c
@@ -318,7 +318,7 @@ void __init anon_vma_init(void)
  * Getting a lock on a stable anon_vma from a page off the LRU is
  * tricky: page_lock_anon_vma rely on RCU to guard against the races.
  */
-struct anon_vma *__page_lock_anon_vma(struct page *page)
+struct anon_vma *page_lock_anon_vma(struct page *page)
 {
 	struct anon_vma *anon_vma, *root_anon_vma;
 	unsigned long anon_mapping;
@@ -352,8 +352,6 @@ struct anon_vma *__page_lock_anon_vma(st
 }
 
 void page_unlock_anon_vma(struct anon_vma *anon_vma)
-	__releases(&anon_vma->root->lock)
-	__releases(RCU)
 {
 	anon_vma_unlock(anon_vma);
 	rcu_read_unlock();

[PATCH 16/20] mm: Revert page_lock_anon_vma() lock annotation

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-mm-revert_page_lock_anon_vma_lock_annotation.patch --]
[-- Type: text/plain, Size: 2240 bytes --]

Its beyond ugly and gets in the way.

Cc: Namhyung Kim <namhyung@gmail.com>
Acked-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 include/linux/rmap.h |   15 +--------------
 mm/rmap.c            |    4 +---
 2 files changed, 2 insertions(+), 17 deletions(-)

Index: linux-2.6/include/linux/rmap.h
===================================================================
--- linux-2.6.orig/include/linux/rmap.h
+++ linux-2.6/include/linux/rmap.h
@@ -243,20 +243,7 @@ int try_to_munlock(struct page *);
 /*
  * Called by memory-failure.c to kill processes.
  */
-struct anon_vma *__page_lock_anon_vma(struct page *page);
-
-static inline struct anon_vma *page_lock_anon_vma(struct page *page)
-{
-	struct anon_vma *anon_vma;
-
-	__cond_lock(RCU, anon_vma = __page_lock_anon_vma(page));
-
-	/* (void) is needed to make gcc happy */
-	(void) __cond_lock(&anon_vma->root->lock, anon_vma);
-
-	return anon_vma;
-}
-
+struct anon_vma *page_lock_anon_vma(struct page *page);
 void page_unlock_anon_vma(struct anon_vma *anon_vma);
 int page_mapped_in_vma(struct page *page, struct vm_area_struct *vma);
 
Index: linux-2.6/mm/rmap.c
===================================================================
--- linux-2.6.orig/mm/rmap.c
+++ linux-2.6/mm/rmap.c
@@ -318,7 +318,7 @@ void __init anon_vma_init(void)
  * Getting a lock on a stable anon_vma from a page off the LRU is
  * tricky: page_lock_anon_vma rely on RCU to guard against the races.
  */
-struct anon_vma *__page_lock_anon_vma(struct page *page)
+struct anon_vma *page_lock_anon_vma(struct page *page)
 {
 	struct anon_vma *anon_vma, *root_anon_vma;
 	unsigned long anon_mapping;
@@ -352,8 +352,6 @@ struct anon_vma *__page_lock_anon_vma(st
 }
 
 void page_unlock_anon_vma(struct anon_vma *anon_vma)
-	__releases(&anon_vma->root->lock)
-	__releases(RCU)
 {
 	anon_vma_unlock(anon_vma);
 	rcu_read_unlock();


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 16/20] mm: Revert page_lock_anon_vma() lock annotation

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-mm-revert_page_lock_anon_vma_lock_annotation.patch --]
[-- Type: text/plain, Size: 2240 bytes --]

Its beyond ugly and gets in the way.

Cc: Namhyung Kim <namhyung@gmail.com>
Acked-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 include/linux/rmap.h |   15 +--------------
 mm/rmap.c            |    4 +---
 2 files changed, 2 insertions(+), 17 deletions(-)

Index: linux-2.6/include/linux/rmap.h
===================================================================
--- linux-2.6.orig/include/linux/rmap.h
+++ linux-2.6/include/linux/rmap.h
@@ -243,20 +243,7 @@ int try_to_munlock(struct page *);
 /*
  * Called by memory-failure.c to kill processes.
  */
-struct anon_vma *__page_lock_anon_vma(struct page *page);
-
-static inline struct anon_vma *page_lock_anon_vma(struct page *page)
-{
-	struct anon_vma *anon_vma;
-
-	__cond_lock(RCU, anon_vma = __page_lock_anon_vma(page));
-
-	/* (void) is needed to make gcc happy */
-	(void) __cond_lock(&anon_vma->root->lock, anon_vma);
-
-	return anon_vma;
-}
-
+struct anon_vma *page_lock_anon_vma(struct page *page);
 void page_unlock_anon_vma(struct anon_vma *anon_vma);
 int page_mapped_in_vma(struct page *page, struct vm_area_struct *vma);
 
Index: linux-2.6/mm/rmap.c
===================================================================
--- linux-2.6.orig/mm/rmap.c
+++ linux-2.6/mm/rmap.c
@@ -318,7 +318,7 @@ void __init anon_vma_init(void)
  * Getting a lock on a stable anon_vma from a page off the LRU is
  * tricky: page_lock_anon_vma rely on RCU to guard against the races.
  */
-struct anon_vma *__page_lock_anon_vma(struct page *page)
+struct anon_vma *page_lock_anon_vma(struct page *page)
 {
 	struct anon_vma *anon_vma, *root_anon_vma;
 	unsigned long anon_mapping;
@@ -352,8 +352,6 @@ struct anon_vma *__page_lock_anon_vma(st
 }
 
 void page_unlock_anon_vma(struct anon_vma *anon_vma)
-	__releases(&anon_vma->root->lock)
-	__releases(RCU)
 {
 	anon_vma_unlock(anon_vma);
 	rcu_read_unlock();


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 17/20] mm: Improve page_lock_anon_vma() comment

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-mm-improve_page_lock_anon_vma_comment.patch --]
[-- Type: text/plain, Size: 1718 bytes --]

A slightly more verbose comment to go along with the trickery in
page_lock_anon_vma().

Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 mm/rmap.c |   18 ++++++++++++++++--
 1 file changed, 16 insertions(+), 2 deletions(-)

Index: linux-2.6/mm/rmap.c
===================================================================
--- linux-2.6.orig/mm/rmap.c
+++ linux-2.6/mm/rmap.c
@@ -315,8 +315,22 @@ void __init anon_vma_init(void)
 }
 
 /*
- * Getting a lock on a stable anon_vma from a page off the LRU is
- * tricky: page_lock_anon_vma rely on RCU to guard against the races.
+ * Getting a lock on a stable anon_vma from a page off the LRU is tricky!
+ *
+ * Since there is no serialization what so ever against page_remove_rmap()
+ * the best this function can do is return a locked anon_vma that might
+ * have been relevant to this page.
+ *
+ * The page might have been remapped to a different anon_vma or the anon_vma
+ * returned may already be freed (and even reused).
+ *
+ * All users of this function must be very careful when walking the anon_vma
+ * chain and verify that the page in question is indeed mapped in it
+ * [ something equivalent to page_mapped_in_vma() ].
+ *
+ * Since anon_vma's slab is DESTROY_BY_RCU and we know from page_remove_rmap()
+ * that the anon_vma pointer from page->mapping is valid if there is a
+ * mapcount, we can dereference the anon_vma after observing those.
  */
 struct anon_vma *page_lock_anon_vma(struct page *page)
 {

[PATCH 17/20] mm: Improve page_lock_anon_vma() comment

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-mm-improve_page_lock_anon_vma_comment.patch --]
[-- Type: text/plain, Size: 2021 bytes --]

A slightly more verbose comment to go along with the trickery in
page_lock_anon_vma().

Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 mm/rmap.c |   18 ++++++++++++++++--
 1 file changed, 16 insertions(+), 2 deletions(-)

Index: linux-2.6/mm/rmap.c
===================================================================
--- linux-2.6.orig/mm/rmap.c
+++ linux-2.6/mm/rmap.c
@@ -315,8 +315,22 @@ void __init anon_vma_init(void)
 }
 
 /*
- * Getting a lock on a stable anon_vma from a page off the LRU is
- * tricky: page_lock_anon_vma rely on RCU to guard against the races.
+ * Getting a lock on a stable anon_vma from a page off the LRU is tricky!
+ *
+ * Since there is no serialization what so ever against page_remove_rmap()
+ * the best this function can do is return a locked anon_vma that might
+ * have been relevant to this page.
+ *
+ * The page might have been remapped to a different anon_vma or the anon_vma
+ * returned may already be freed (and even reused).
+ *
+ * All users of this function must be very careful when walking the anon_vma
+ * chain and verify that the page in question is indeed mapped in it
+ * [ something equivalent to page_mapped_in_vma() ].
+ *
+ * Since anon_vma's slab is DESTROY_BY_RCU and we know from page_remove_rmap()
+ * that the anon_vma pointer from page->mapping is valid if there is a
+ * mapcount, we can dereference the anon_vma after observing those.
  */
 struct anon_vma *page_lock_anon_vma(struct page *page)
 {


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 17/20] mm: Improve page_lock_anon_vma() comment

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-mm-improve_page_lock_anon_vma_comment.patch --]
[-- Type: text/plain, Size: 2021 bytes --]

A slightly more verbose comment to go along with the trickery in
page_lock_anon_vma().

Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 mm/rmap.c |   18 ++++++++++++++++--
 1 file changed, 16 insertions(+), 2 deletions(-)

Index: linux-2.6/mm/rmap.c
===================================================================
--- linux-2.6.orig/mm/rmap.c
+++ linux-2.6/mm/rmap.c
@@ -315,8 +315,22 @@ void __init anon_vma_init(void)
 }
 
 /*
- * Getting a lock on a stable anon_vma from a page off the LRU is
- * tricky: page_lock_anon_vma rely on RCU to guard against the races.
+ * Getting a lock on a stable anon_vma from a page off the LRU is tricky!
+ *
+ * Since there is no serialization what so ever against page_remove_rmap()
+ * the best this function can do is return a locked anon_vma that might
+ * have been relevant to this page.
+ *
+ * The page might have been remapped to a different anon_vma or the anon_vma
+ * returned may already be freed (and even reused).
+ *
+ * All users of this function must be very careful when walking the anon_vma
+ * chain and verify that the page in question is indeed mapped in it
+ * [ something equivalent to page_mapped_in_vma() ].
+ *
+ * Since anon_vma's slab is DESTROY_BY_RCU and we know from page_remove_rmap()
+ * that the anon_vma pointer from page->mapping is valid if there is a
+ * mapcount, we can dereference the anon_vma after observing those.
  */
 struct anon_vma *page_lock_anon_vma(struct page *page)
 {


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 18/20] mm: Use refcounts for page_lock_anon_vma()

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-mm-use_refcounts_for_page_lock_anon_vma.patch --]
[-- Type: text/plain, Size: 4032 bytes --]

Convert page_lock_anon_vma() over to use refcounts. This is done to
prepare for the conversion of anon_vma from spinlock to mutex.

Sadly this inceases the cost of page_lock_anon_vma() from one to two
atomics, a follow up patch addresses this, lets keep that simple for
now.

Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Acked-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 mm/migrate.c |   17 ++++-------------
 mm/rmap.c    |   42 +++++++++++++++++++++++++++---------------
 2 files changed, 31 insertions(+), 28 deletions(-)

Index: linux-2.6/mm/rmap.c
===================================================================
--- linux-2.6.orig/mm/rmap.c
+++ linux-2.6/mm/rmap.c
@@ -336,9 +336,9 @@ void __init anon_vma_init(void)
  * that the anon_vma pointer from page->mapping is valid if there is a
  * mapcount, we can dereference the anon_vma after observing those.
  */
-struct anon_vma *page_lock_anon_vma(struct page *page)
+struct anon_vma *page_get_anon_vma(struct page *page)
 {
-	struct anon_vma *anon_vma, *root_anon_vma;
+	struct anon_vma *anon_vma = NULL;
 	unsigned long anon_mapping;
 
 	rcu_read_lock();
@@ -349,30 +349,42 @@ struct anon_vma *page_lock_anon_vma(stru
 		goto out;
 
 	anon_vma = (struct anon_vma *) (anon_mapping - PAGE_MAPPING_ANON);
-	root_anon_vma = ACCESS_ONCE(anon_vma->root);
-	spin_lock(&root_anon_vma->lock);
+	if (!atomic_inc_not_zero(&anon_vma->refcount)) {
+		anon_vma = NULL;
+		goto out;
+	}
 
 	/*
 	 * If this page is still mapped, then its anon_vma cannot have been
-	 * freed.  But if it has been unmapped, we have no security against
-	 * the anon_vma structure being freed and reused (for another anon_vma:
-	 * SLAB_DESTROY_BY_RCU guarantees that - so the spin_lock above cannot
-	 * corrupt): with anon_vma_prepare() or anon_vma_fork() redirecting
-	 * anon_vma->root before page_unlock_anon_vma() is called to unlock.
+	 * freed.  But if it has been unmapped, we have no security against the
+	 * anon_vma structure being freed and reused (for another anon_vma:
+	 * SLAB_DESTROY_BY_RCU guarantees that - so the atomic_inc_not_zero()
+	 * above cannot corrupt).
 	 */
-	if (page_mapped(page))
-		return anon_vma;
-
-	spin_unlock(&root_anon_vma->lock);
+	if (!page_mapped(page)) {
+		put_anon_vma(anon_vma);
+		anon_vma = NULL;
+	}
 out:
 	rcu_read_unlock();
-	return NULL;
+
+	return anon_vma;
+}
+
+struct anon_vma *page_lock_anon_vma(struct page *page)
+{
+	struct anon_vma *anon_vma = page_get_anon_vma(page);
+
+	if (anon_vma)
+		anon_vma_lock(anon_vma);
+
+	return anon_vma;
 }
 
 void page_unlock_anon_vma(struct anon_vma *anon_vma)
 {
 	anon_vma_unlock(anon_vma);
-	rcu_read_unlock();
+	put_anon_vma(anon_vma);
 }
 
 /*
Index: linux-2.6/mm/migrate.c
===================================================================
--- linux-2.6.orig/mm/migrate.c
+++ linux-2.6/mm/migrate.c
@@ -703,15 +703,11 @@ static int unmap_and_move(new_page_t get
 		 * Only page_lock_anon_vma() understands the subtleties of
 		 * getting a hold on an anon_vma from outside one of its mms.
 		 */
-		anon_vma = page_lock_anon_vma(page);
+		anon_vma = page_get_anon_vma(page);
 		if (anon_vma) {
 			/*
-			 * Take a reference count on the anon_vma if the
-			 * page is mapped so that it is guaranteed to
-			 * exist when the page is remapped later
+			 * Anon page
 			 */
-			get_anon_vma(anon_vma);
-			page_unlock_anon_vma(anon_vma);
 		} else if (PageSwapCache(page)) {
 			/*
 			 * We cannot be sure that the anon_vma of an unmapped
@@ -840,13 +836,8 @@ static int unmap_and_move_huge_page(new_
 		lock_page(hpage);
 	}
 
-	if (PageAnon(hpage)) {
-		anon_vma = page_lock_anon_vma(hpage);
-		if (anon_vma) {
-			get_anon_vma(anon_vma);
-			page_unlock_anon_vma(anon_vma);
-		}
-	}
+	if (PageAnon(hpage))
+		anon_vma = page_get_anon_vma(hpage);
 
 	try_to_unmap(hpage, TTU_MIGRATION|TTU_IGNORE_MLOCK|TTU_IGNORE_ACCESS);
 

[PATCH 18/20] mm: Use refcounts for page_lock_anon_vma()

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-mm-use_refcounts_for_page_lock_anon_vma.patch --]
[-- Type: text/plain, Size: 4335 bytes --]

Convert page_lock_anon_vma() over to use refcounts. This is done to
prepare for the conversion of anon_vma from spinlock to mutex.

Sadly this inceases the cost of page_lock_anon_vma() from one to two
atomics, a follow up patch addresses this, lets keep that simple for
now.

Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Acked-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 mm/migrate.c |   17 ++++-------------
 mm/rmap.c    |   42 +++++++++++++++++++++++++++---------------
 2 files changed, 31 insertions(+), 28 deletions(-)

Index: linux-2.6/mm/rmap.c
===================================================================
--- linux-2.6.orig/mm/rmap.c
+++ linux-2.6/mm/rmap.c
@@ -336,9 +336,9 @@ void __init anon_vma_init(void)
  * that the anon_vma pointer from page->mapping is valid if there is a
  * mapcount, we can dereference the anon_vma after observing those.
  */
-struct anon_vma *page_lock_anon_vma(struct page *page)
+struct anon_vma *page_get_anon_vma(struct page *page)
 {
-	struct anon_vma *anon_vma, *root_anon_vma;
+	struct anon_vma *anon_vma = NULL;
 	unsigned long anon_mapping;
 
 	rcu_read_lock();
@@ -349,30 +349,42 @@ struct anon_vma *page_lock_anon_vma(stru
 		goto out;
 
 	anon_vma = (struct anon_vma *) (anon_mapping - PAGE_MAPPING_ANON);
-	root_anon_vma = ACCESS_ONCE(anon_vma->root);
-	spin_lock(&root_anon_vma->lock);
+	if (!atomic_inc_not_zero(&anon_vma->refcount)) {
+		anon_vma = NULL;
+		goto out;
+	}
 
 	/*
 	 * If this page is still mapped, then its anon_vma cannot have been
-	 * freed.  But if it has been unmapped, we have no security against
-	 * the anon_vma structure being freed and reused (for another anon_vma:
-	 * SLAB_DESTROY_BY_RCU guarantees that - so the spin_lock above cannot
-	 * corrupt): with anon_vma_prepare() or anon_vma_fork() redirecting
-	 * anon_vma->root before page_unlock_anon_vma() is called to unlock.
+	 * freed.  But if it has been unmapped, we have no security against the
+	 * anon_vma structure being freed and reused (for another anon_vma:
+	 * SLAB_DESTROY_BY_RCU guarantees that - so the atomic_inc_not_zero()
+	 * above cannot corrupt).
 	 */
-	if (page_mapped(page))
-		return anon_vma;
-
-	spin_unlock(&root_anon_vma->lock);
+	if (!page_mapped(page)) {
+		put_anon_vma(anon_vma);
+		anon_vma = NULL;
+	}
 out:
 	rcu_read_unlock();
-	return NULL;
+
+	return anon_vma;
+}
+
+struct anon_vma *page_lock_anon_vma(struct page *page)
+{
+	struct anon_vma *anon_vma = page_get_anon_vma(page);
+
+	if (anon_vma)
+		anon_vma_lock(anon_vma);
+
+	return anon_vma;
 }
 
 void page_unlock_anon_vma(struct anon_vma *anon_vma)
 {
 	anon_vma_unlock(anon_vma);
-	rcu_read_unlock();
+	put_anon_vma(anon_vma);
 }
 
 /*
Index: linux-2.6/mm/migrate.c
===================================================================
--- linux-2.6.orig/mm/migrate.c
+++ linux-2.6/mm/migrate.c
@@ -703,15 +703,11 @@ static int unmap_and_move(new_page_t get
 		 * Only page_lock_anon_vma() understands the subtleties of
 		 * getting a hold on an anon_vma from outside one of its mms.
 		 */
-		anon_vma = page_lock_anon_vma(page);
+		anon_vma = page_get_anon_vma(page);
 		if (anon_vma) {
 			/*
-			 * Take a reference count on the anon_vma if the
-			 * page is mapped so that it is guaranteed to
-			 * exist when the page is remapped later
+			 * Anon page
 			 */
-			get_anon_vma(anon_vma);
-			page_unlock_anon_vma(anon_vma);
 		} else if (PageSwapCache(page)) {
 			/*
 			 * We cannot be sure that the anon_vma of an unmapped
@@ -840,13 +836,8 @@ static int unmap_and_move_huge_page(new_
 		lock_page(hpage);
 	}
 
-	if (PageAnon(hpage)) {
-		anon_vma = page_lock_anon_vma(hpage);
-		if (anon_vma) {
-			get_anon_vma(anon_vma);
-			page_unlock_anon_vma(anon_vma);
-		}
-	}
+	if (PageAnon(hpage))
+		anon_vma = page_get_anon_vma(hpage);
 
 	try_to_unmap(hpage, TTU_MIGRATION|TTU_IGNORE_MLOCK|TTU_IGNORE_ACCESS);
 


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 18/20] mm: Use refcounts for page_lock_anon_vma()

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-mm-use_refcounts_for_page_lock_anon_vma.patch --]
[-- Type: text/plain, Size: 4335 bytes --]

Convert page_lock_anon_vma() over to use refcounts. This is done to
prepare for the conversion of anon_vma from spinlock to mutex.

Sadly this inceases the cost of page_lock_anon_vma() from one to two
atomics, a follow up patch addresses this, lets keep that simple for
now.

Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Acked-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 mm/migrate.c |   17 ++++-------------
 mm/rmap.c    |   42 +++++++++++++++++++++++++++---------------
 2 files changed, 31 insertions(+), 28 deletions(-)

Index: linux-2.6/mm/rmap.c
===================================================================
--- linux-2.6.orig/mm/rmap.c
+++ linux-2.6/mm/rmap.c
@@ -336,9 +336,9 @@ void __init anon_vma_init(void)
  * that the anon_vma pointer from page->mapping is valid if there is a
  * mapcount, we can dereference the anon_vma after observing those.
  */
-struct anon_vma *page_lock_anon_vma(struct page *page)
+struct anon_vma *page_get_anon_vma(struct page *page)
 {
-	struct anon_vma *anon_vma, *root_anon_vma;
+	struct anon_vma *anon_vma = NULL;
 	unsigned long anon_mapping;
 
 	rcu_read_lock();
@@ -349,30 +349,42 @@ struct anon_vma *page_lock_anon_vma(stru
 		goto out;
 
 	anon_vma = (struct anon_vma *) (anon_mapping - PAGE_MAPPING_ANON);
-	root_anon_vma = ACCESS_ONCE(anon_vma->root);
-	spin_lock(&root_anon_vma->lock);
+	if (!atomic_inc_not_zero(&anon_vma->refcount)) {
+		anon_vma = NULL;
+		goto out;
+	}
 
 	/*
 	 * If this page is still mapped, then its anon_vma cannot have been
-	 * freed.  But if it has been unmapped, we have no security against
-	 * the anon_vma structure being freed and reused (for another anon_vma:
-	 * SLAB_DESTROY_BY_RCU guarantees that - so the spin_lock above cannot
-	 * corrupt): with anon_vma_prepare() or anon_vma_fork() redirecting
-	 * anon_vma->root before page_unlock_anon_vma() is called to unlock.
+	 * freed.  But if it has been unmapped, we have no security against the
+	 * anon_vma structure being freed and reused (for another anon_vma:
+	 * SLAB_DESTROY_BY_RCU guarantees that - so the atomic_inc_not_zero()
+	 * above cannot corrupt).
 	 */
-	if (page_mapped(page))
-		return anon_vma;
-
-	spin_unlock(&root_anon_vma->lock);
+	if (!page_mapped(page)) {
+		put_anon_vma(anon_vma);
+		anon_vma = NULL;
+	}
 out:
 	rcu_read_unlock();
-	return NULL;
+
+	return anon_vma;
+}
+
+struct anon_vma *page_lock_anon_vma(struct page *page)
+{
+	struct anon_vma *anon_vma = page_get_anon_vma(page);
+
+	if (anon_vma)
+		anon_vma_lock(anon_vma);
+
+	return anon_vma;
 }
 
 void page_unlock_anon_vma(struct anon_vma *anon_vma)
 {
 	anon_vma_unlock(anon_vma);
-	rcu_read_unlock();
+	put_anon_vma(anon_vma);
 }
 
 /*
Index: linux-2.6/mm/migrate.c
===================================================================
--- linux-2.6.orig/mm/migrate.c
+++ linux-2.6/mm/migrate.c
@@ -703,15 +703,11 @@ static int unmap_and_move(new_page_t get
 		 * Only page_lock_anon_vma() understands the subtleties of
 		 * getting a hold on an anon_vma from outside one of its mms.
 		 */
-		anon_vma = page_lock_anon_vma(page);
+		anon_vma = page_get_anon_vma(page);
 		if (anon_vma) {
 			/*
-			 * Take a reference count on the anon_vma if the
-			 * page is mapped so that it is guaranteed to
-			 * exist when the page is remapped later
+			 * Anon page
 			 */
-			get_anon_vma(anon_vma);
-			page_unlock_anon_vma(anon_vma);
 		} else if (PageSwapCache(page)) {
 			/*
 			 * We cannot be sure that the anon_vma of an unmapped
@@ -840,13 +836,8 @@ static int unmap_and_move_huge_page(new_
 		lock_page(hpage);
 	}
 
-	if (PageAnon(hpage)) {
-		anon_vma = page_lock_anon_vma(hpage);
-		if (anon_vma) {
-			get_anon_vma(anon_vma);
-			page_unlock_anon_vma(anon_vma);
-		}
-	}
+	if (PageAnon(hpage))
+		anon_vma = page_get_anon_vma(hpage);
 
 	try_to_unmap(hpage, TTU_MIGRATION|TTU_IGNORE_MLOCK|TTU_IGNORE_ACCESS);
 


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 19/20] mm: Convert anon_vma->lock to a mutex

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-mm-anon_vma-lock_to_mutexes.patch --]
[-- Type: text/plain, Size: 7774 bytes --]

Straight fwd conversion of anon_vma->lock to a mutex.

Acked-by: Hugh Dickins <hughd@google.com>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 include/linux/huge_mm.h      |    8 ++------
 include/linux/mmu_notifier.h |    2 +-
 include/linux/rmap.h         |   14 +++++++-------
 mm/huge_memory.c             |    4 ++--
 mm/mmap.c                    |   10 +++++-----
 mm/rmap.c                    |    8 ++++----
 6 files changed, 21 insertions(+), 25 deletions(-)

Index: linux-2.6/include/linux/rmap.h
===================================================================
--- linux-2.6.orig/include/linux/rmap.h
+++ linux-2.6/include/linux/rmap.h
@@ -7,7 +7,7 @@
 #include <linux/list.h>
 #include <linux/slab.h>
 #include <linux/mm.h>
-#include <linux/spinlock.h>
+#include <linux/mutex.h>
 #include <linux/memcontrol.h>
 
 /*
@@ -26,7 +26,7 @@
  */
 struct anon_vma {
 	struct anon_vma *root;	/* Root of this anon_vma tree */
-	spinlock_t lock;	/* Serialize access to vma list */
+	struct mutex mutex;	/* Serialize access to vma list */
 	/*
 	 * The refcount is taken on an anon_vma when there is no
 	 * guarantee that the vma of page tables will exist for
@@ -64,7 +64,7 @@ struct anon_vma_chain {
 	struct vm_area_struct *vma;
 	struct anon_vma *anon_vma;
 	struct list_head same_vma;   /* locked by mmap_sem & page_table_lock */
-	struct list_head same_anon_vma;	/* locked by anon_vma->lock */
+	struct list_head same_anon_vma;	/* locked by anon_vma->mutex */
 };
 
 #ifdef CONFIG_MMU
@@ -93,24 +93,24 @@ static inline void vma_lock_anon_vma(str
 {
 	struct anon_vma *anon_vma = vma->anon_vma;
 	if (anon_vma)
-		spin_lock(&anon_vma->root->lock);
+		mutex_lock(&anon_vma->root->mutex);
 }
 
 static inline void vma_unlock_anon_vma(struct vm_area_struct *vma)
 {
 	struct anon_vma *anon_vma = vma->anon_vma;
 	if (anon_vma)
-		spin_unlock(&anon_vma->root->lock);
+		mutex_unlock(&anon_vma->root->mutex);
 }
 
 static inline void anon_vma_lock(struct anon_vma *anon_vma)
 {
-	spin_lock(&anon_vma->root->lock);
+	mutex_lock(&anon_vma->root->mutex);
 }
 
 static inline void anon_vma_unlock(struct anon_vma *anon_vma)
 {
-	spin_unlock(&anon_vma->root->lock);
+	mutex_unlock(&anon_vma->root->mutex);
 }
 
 /*
Index: linux-2.6/mm/rmap.c
===================================================================
--- linux-2.6.orig/mm/rmap.c
+++ linux-2.6/mm/rmap.c
@@ -25,7 +25,7 @@
  *   mm->mmap_sem
  *     page->flags PG_locked (lock_page)
  *       mapping->i_mmap_mutex
- *         anon_vma->lock
+ *         anon_vma->mutex
  *           mm->page_table_lock or pte_lock
  *             zone->lru_lock (in mark_page_accessed, isolate_lru_page)
  *             swap_lock (in swap_duplicate, swap_info_get)
@@ -39,7 +39,7 @@
  *
  * (code doesn't rely on that order so it could be switched around)
  * ->tasklist_lock
- *   anon_vma->lock      (memory_failure, collect_procs_anon)
+ *   anon_vma->mutex      (memory_failure, collect_procs_anon)
  *     pte map lock
  */
 
@@ -306,7 +306,7 @@ static void anon_vma_ctor(void *data)
 {
 	struct anon_vma *anon_vma = data;
 
-	spin_lock_init(&anon_vma->lock);
+	mutex_init(&anon_vma->mutex);
 	atomic_set(&anon_vma->refcount, 0);
 	INIT_LIST_HEAD(&anon_vma->head);
 }
@@ -1129,7 +1129,7 @@ int try_to_unmap_one(struct page *page, 
 	/*
 	 * We need mmap_sem locking, Otherwise VM_LOCKED check makes
 	 * unstable result and race. Plus, We can't wait here because
-	 * we now hold anon_vma->lock or mapping->i_mmap_mutex.
+	 * we now hold anon_vma->mutex or mapping->i_mmap_mutex.
 	 * if trylock failed, the page remain in evictable lru and later
 	 * vmscan could retry to move the page to unevictable lru if the
 	 * page is actually mlocked.
Index: linux-2.6/mm/mmap.c
===================================================================
--- linux-2.6.orig/mm/mmap.c
+++ linux-2.6/mm/mmap.c
@@ -2523,15 +2523,15 @@ static void vm_lock_anon_vma(struct mm_s
 		 * The LSB of head.next can't change from under us
 		 * because we hold the mm_all_locks_mutex.
 		 */
-		spin_lock_nest_lock(&anon_vma->root->lock, &mm->mmap_sem);
+		mutex_lock_nest_lock(&anon_vma->root->mutex, &mm->mmap_sem);
 		/*
 		 * We can safely modify head.next after taking the
-		 * anon_vma->root->lock. If some other vma in this mm shares
+		 * anon_vma->root->mutex. If some other vma in this mm shares
 		 * the same anon_vma we won't take it again.
 		 *
 		 * No need of atomic instructions here, head.next
 		 * can't change from under us thanks to the
-		 * anon_vma->root->lock.
+		 * anon_vma->root->mutex.
 		 */
 		if (__test_and_set_bit(0, (unsigned long *)
 				       &anon_vma->root->head.next))
@@ -2580,7 +2580,7 @@ static void vm_lock_mapping(struct mm_st
  * vma in this mm is backed by the same anon_vma or address_space.
  *
  * We can take all the locks in random order because the VM code
- * taking i_mmap_mutex or anon_vma->lock outside the mmap_sem never
+ * taking i_mmap_mutex or anon_vma->mutex outside the mmap_sem never
  * takes more than one of them in a row. Secondly we're protected
  * against a concurrent mm_take_all_locks() by the mm_all_locks_mutex.
  *
@@ -2636,7 +2636,7 @@ static void vm_unlock_anon_vma(struct an
 		 *
 		 * No need of atomic instructions here, head.next
 		 * can't change from under us until we release the
-		 * anon_vma->root->lock.
+		 * anon_vma->root->mutex.
 		 */
 		if (!__test_and_clear_bit(0, (unsigned long *)
 					  &anon_vma->root->head.next))
Index: linux-2.6/include/linux/mmu_notifier.h
===================================================================
--- linux-2.6.orig/include/linux/mmu_notifier.h
+++ linux-2.6/include/linux/mmu_notifier.h
@@ -150,7 +150,7 @@ struct mmu_notifier_ops {
  * Therefore notifier chains can only be traversed when either
  *
  * 1. mmap_sem is held.
- * 2. One of the reverse map locks is held (i_mmap_mutex or anon_vma->lock).
+ * 2. One of the reverse map locks is held (i_mmap_mutex or anon_vma->mutex).
  * 3. No other concurrent thread can access the list (release)
  */
 struct mmu_notifier {
Index: linux-2.6/mm/huge_memory.c
===================================================================
--- linux-2.6.orig/mm/huge_memory.c
+++ linux-2.6/mm/huge_memory.c
@@ -1128,7 +1128,7 @@ static int __split_huge_page_splitting(s
 		 * We can't temporarily set the pmd to null in order
 		 * to split it, the pmd must remain marked huge at all
 		 * times or the VM won't take the pmd_trans_huge paths
-		 * and it won't wait on the anon_vma->root->lock to
+		 * and it won't wait on the anon_vma->root->mutex to
 		 * serialize against split_huge_page*.
 		 */
 		pmdp_splitting_flush_notify(vma, address, pmd);
@@ -1315,7 +1315,7 @@ static int __split_huge_page_map(struct 
 	return ret;
 }
 
-/* must be called with anon_vma->root->lock hold */
+/* must be called with anon_vma->root->mutex hold */
 static void __split_huge_page(struct page *page,
 			      struct anon_vma *anon_vma)
 {
Index: linux-2.6/include/linux/huge_mm.h
===================================================================
--- linux-2.6.orig/include/linux/huge_mm.h
+++ linux-2.6/include/linux/huge_mm.h
@@ -91,12 +91,8 @@ extern void __split_huge_page_pmd(struct
 #define wait_split_huge_page(__anon_vma, __pmd)				\
 	do {								\
 		pmd_t *____pmd = (__pmd);				\
-		spin_unlock_wait(&(__anon_vma)->root->lock);		\
-		/*							\
-		 * spin_unlock_wait() is just a loop in C and so the	\
-		 * CPU can reorder anything around it.			\
-		 */							\
-		smp_mb();						\
+		anon_vma_lock(__anon_vma);				\
+		anon_vma_unlock(__anon_vma);				\
 		BUG_ON(pmd_trans_splitting(*____pmd) ||			\
 		       pmd_trans_huge(*____pmd));			\
 	} while (0)

[PATCH 19/20] mm: Convert anon_vma->lock to a mutex

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-mm-anon_vma-lock_to_mutexes.patch --]
[-- Type: text/plain, Size: 8077 bytes --]

Straight fwd conversion of anon_vma->lock to a mutex.

Acked-by: Hugh Dickins <hughd@google.com>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 include/linux/huge_mm.h      |    8 ++------
 include/linux/mmu_notifier.h |    2 +-
 include/linux/rmap.h         |   14 +++++++-------
 mm/huge_memory.c             |    4 ++--
 mm/mmap.c                    |   10 +++++-----
 mm/rmap.c                    |    8 ++++----
 6 files changed, 21 insertions(+), 25 deletions(-)

Index: linux-2.6/include/linux/rmap.h
===================================================================
--- linux-2.6.orig/include/linux/rmap.h
+++ linux-2.6/include/linux/rmap.h
@@ -7,7 +7,7 @@
 #include <linux/list.h>
 #include <linux/slab.h>
 #include <linux/mm.h>
-#include <linux/spinlock.h>
+#include <linux/mutex.h>
 #include <linux/memcontrol.h>
 
 /*
@@ -26,7 +26,7 @@
  */
 struct anon_vma {
 	struct anon_vma *root;	/* Root of this anon_vma tree */
-	spinlock_t lock;	/* Serialize access to vma list */
+	struct mutex mutex;	/* Serialize access to vma list */
 	/*
 	 * The refcount is taken on an anon_vma when there is no
 	 * guarantee that the vma of page tables will exist for
@@ -64,7 +64,7 @@ struct anon_vma_chain {
 	struct vm_area_struct *vma;
 	struct anon_vma *anon_vma;
 	struct list_head same_vma;   /* locked by mmap_sem & page_table_lock */
-	struct list_head same_anon_vma;	/* locked by anon_vma->lock */
+	struct list_head same_anon_vma;	/* locked by anon_vma->mutex */
 };
 
 #ifdef CONFIG_MMU
@@ -93,24 +93,24 @@ static inline void vma_lock_anon_vma(str
 {
 	struct anon_vma *anon_vma = vma->anon_vma;
 	if (anon_vma)
-		spin_lock(&anon_vma->root->lock);
+		mutex_lock(&anon_vma->root->mutex);
 }
 
 static inline void vma_unlock_anon_vma(struct vm_area_struct *vma)
 {
 	struct anon_vma *anon_vma = vma->anon_vma;
 	if (anon_vma)
-		spin_unlock(&anon_vma->root->lock);
+		mutex_unlock(&anon_vma->root->mutex);
 }
 
 static inline void anon_vma_lock(struct anon_vma *anon_vma)
 {
-	spin_lock(&anon_vma->root->lock);
+	mutex_lock(&anon_vma->root->mutex);
 }
 
 static inline void anon_vma_unlock(struct anon_vma *anon_vma)
 {
-	spin_unlock(&anon_vma->root->lock);
+	mutex_unlock(&anon_vma->root->mutex);
 }
 
 /*
Index: linux-2.6/mm/rmap.c
===================================================================
--- linux-2.6.orig/mm/rmap.c
+++ linux-2.6/mm/rmap.c
@@ -25,7 +25,7 @@
  *   mm->mmap_sem
  *     page->flags PG_locked (lock_page)
  *       mapping->i_mmap_mutex
- *         anon_vma->lock
+ *         anon_vma->mutex
  *           mm->page_table_lock or pte_lock
  *             zone->lru_lock (in mark_page_accessed, isolate_lru_page)
  *             swap_lock (in swap_duplicate, swap_info_get)
@@ -39,7 +39,7 @@
  *
  * (code doesn't rely on that order so it could be switched around)
  * ->tasklist_lock
- *   anon_vma->lock      (memory_failure, collect_procs_anon)
+ *   anon_vma->mutex      (memory_failure, collect_procs_anon)
  *     pte map lock
  */
 
@@ -306,7 +306,7 @@ static void anon_vma_ctor(void *data)
 {
 	struct anon_vma *anon_vma = data;
 
-	spin_lock_init(&anon_vma->lock);
+	mutex_init(&anon_vma->mutex);
 	atomic_set(&anon_vma->refcount, 0);
 	INIT_LIST_HEAD(&anon_vma->head);
 }
@@ -1129,7 +1129,7 @@ int try_to_unmap_one(struct page *page, 
 	/*
 	 * We need mmap_sem locking, Otherwise VM_LOCKED check makes
 	 * unstable result and race. Plus, We can't wait here because
-	 * we now hold anon_vma->lock or mapping->i_mmap_mutex.
+	 * we now hold anon_vma->mutex or mapping->i_mmap_mutex.
 	 * if trylock failed, the page remain in evictable lru and later
 	 * vmscan could retry to move the page to unevictable lru if the
 	 * page is actually mlocked.
Index: linux-2.6/mm/mmap.c
===================================================================
--- linux-2.6.orig/mm/mmap.c
+++ linux-2.6/mm/mmap.c
@@ -2523,15 +2523,15 @@ static void vm_lock_anon_vma(struct mm_s
 		 * The LSB of head.next can't change from under us
 		 * because we hold the mm_all_locks_mutex.
 		 */
-		spin_lock_nest_lock(&anon_vma->root->lock, &mm->mmap_sem);
+		mutex_lock_nest_lock(&anon_vma->root->mutex, &mm->mmap_sem);
 		/*
 		 * We can safely modify head.next after taking the
-		 * anon_vma->root->lock. If some other vma in this mm shares
+		 * anon_vma->root->mutex. If some other vma in this mm shares
 		 * the same anon_vma we won't take it again.
 		 *
 		 * No need of atomic instructions here, head.next
 		 * can't change from under us thanks to the
-		 * anon_vma->root->lock.
+		 * anon_vma->root->mutex.
 		 */
 		if (__test_and_set_bit(0, (unsigned long *)
 				       &anon_vma->root->head.next))
@@ -2580,7 +2580,7 @@ static void vm_lock_mapping(struct mm_st
  * vma in this mm is backed by the same anon_vma or address_space.
  *
  * We can take all the locks in random order because the VM code
- * taking i_mmap_mutex or anon_vma->lock outside the mmap_sem never
+ * taking i_mmap_mutex or anon_vma->mutex outside the mmap_sem never
  * takes more than one of them in a row. Secondly we're protected
  * against a concurrent mm_take_all_locks() by the mm_all_locks_mutex.
  *
@@ -2636,7 +2636,7 @@ static void vm_unlock_anon_vma(struct an
 		 *
 		 * No need of atomic instructions here, head.next
 		 * can't change from under us until we release the
-		 * anon_vma->root->lock.
+		 * anon_vma->root->mutex.
 		 */
 		if (!__test_and_clear_bit(0, (unsigned long *)
 					  &anon_vma->root->head.next))
Index: linux-2.6/include/linux/mmu_notifier.h
===================================================================
--- linux-2.6.orig/include/linux/mmu_notifier.h
+++ linux-2.6/include/linux/mmu_notifier.h
@@ -150,7 +150,7 @@ struct mmu_notifier_ops {
  * Therefore notifier chains can only be traversed when either
  *
  * 1. mmap_sem is held.
- * 2. One of the reverse map locks is held (i_mmap_mutex or anon_vma->lock).
+ * 2. One of the reverse map locks is held (i_mmap_mutex or anon_vma->mutex).
  * 3. No other concurrent thread can access the list (release)
  */
 struct mmu_notifier {
Index: linux-2.6/mm/huge_memory.c
===================================================================
--- linux-2.6.orig/mm/huge_memory.c
+++ linux-2.6/mm/huge_memory.c
@@ -1128,7 +1128,7 @@ static int __split_huge_page_splitting(s
 		 * We can't temporarily set the pmd to null in order
 		 * to split it, the pmd must remain marked huge at all
 		 * times or the VM won't take the pmd_trans_huge paths
-		 * and it won't wait on the anon_vma->root->lock to
+		 * and it won't wait on the anon_vma->root->mutex to
 		 * serialize against split_huge_page*.
 		 */
 		pmdp_splitting_flush_notify(vma, address, pmd);
@@ -1315,7 +1315,7 @@ static int __split_huge_page_map(struct 
 	return ret;
 }
 
-/* must be called with anon_vma->root->lock hold */
+/* must be called with anon_vma->root->mutex hold */
 static void __split_huge_page(struct page *page,
 			      struct anon_vma *anon_vma)
 {
Index: linux-2.6/include/linux/huge_mm.h
===================================================================
--- linux-2.6.orig/include/linux/huge_mm.h
+++ linux-2.6/include/linux/huge_mm.h
@@ -91,12 +91,8 @@ extern void __split_huge_page_pmd(struct
 #define wait_split_huge_page(__anon_vma, __pmd)				\
 	do {								\
 		pmd_t *____pmd = (__pmd);				\
-		spin_unlock_wait(&(__anon_vma)->root->lock);		\
-		/*							\
-		 * spin_unlock_wait() is just a loop in C and so the	\
-		 * CPU can reorder anything around it.			\
-		 */							\
-		smp_mb();						\
+		anon_vma_lock(__anon_vma);				\
+		anon_vma_unlock(__anon_vma);				\
 		BUG_ON(pmd_trans_splitting(*____pmd) ||			\
 		       pmd_trans_huge(*____pmd));			\
 	} while (0)


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 19/20] mm: Convert anon_vma->lock to a mutex

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-mm-anon_vma-lock_to_mutexes.patch --]
[-- Type: text/plain, Size: 8077 bytes --]

Straight fwd conversion of anon_vma->lock to a mutex.

Acked-by: Hugh Dickins <hughd@google.com>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 include/linux/huge_mm.h      |    8 ++------
 include/linux/mmu_notifier.h |    2 +-
 include/linux/rmap.h         |   14 +++++++-------
 mm/huge_memory.c             |    4 ++--
 mm/mmap.c                    |   10 +++++-----
 mm/rmap.c                    |    8 ++++----
 6 files changed, 21 insertions(+), 25 deletions(-)

Index: linux-2.6/include/linux/rmap.h
===================================================================
--- linux-2.6.orig/include/linux/rmap.h
+++ linux-2.6/include/linux/rmap.h
@@ -7,7 +7,7 @@
 #include <linux/list.h>
 #include <linux/slab.h>
 #include <linux/mm.h>
-#include <linux/spinlock.h>
+#include <linux/mutex.h>
 #include <linux/memcontrol.h>
 
 /*
@@ -26,7 +26,7 @@
  */
 struct anon_vma {
 	struct anon_vma *root;	/* Root of this anon_vma tree */
-	spinlock_t lock;	/* Serialize access to vma list */
+	struct mutex mutex;	/* Serialize access to vma list */
 	/*
 	 * The refcount is taken on an anon_vma when there is no
 	 * guarantee that the vma of page tables will exist for
@@ -64,7 +64,7 @@ struct anon_vma_chain {
 	struct vm_area_struct *vma;
 	struct anon_vma *anon_vma;
 	struct list_head same_vma;   /* locked by mmap_sem & page_table_lock */
-	struct list_head same_anon_vma;	/* locked by anon_vma->lock */
+	struct list_head same_anon_vma;	/* locked by anon_vma->mutex */
 };
 
 #ifdef CONFIG_MMU
@@ -93,24 +93,24 @@ static inline void vma_lock_anon_vma(str
 {
 	struct anon_vma *anon_vma = vma->anon_vma;
 	if (anon_vma)
-		spin_lock(&anon_vma->root->lock);
+		mutex_lock(&anon_vma->root->mutex);
 }
 
 static inline void vma_unlock_anon_vma(struct vm_area_struct *vma)
 {
 	struct anon_vma *anon_vma = vma->anon_vma;
 	if (anon_vma)
-		spin_unlock(&anon_vma->root->lock);
+		mutex_unlock(&anon_vma->root->mutex);
 }
 
 static inline void anon_vma_lock(struct anon_vma *anon_vma)
 {
-	spin_lock(&anon_vma->root->lock);
+	mutex_lock(&anon_vma->root->mutex);
 }
 
 static inline void anon_vma_unlock(struct anon_vma *anon_vma)
 {
-	spin_unlock(&anon_vma->root->lock);
+	mutex_unlock(&anon_vma->root->mutex);
 }
 
 /*
Index: linux-2.6/mm/rmap.c
===================================================================
--- linux-2.6.orig/mm/rmap.c
+++ linux-2.6/mm/rmap.c
@@ -25,7 +25,7 @@
  *   mm->mmap_sem
  *     page->flags PG_locked (lock_page)
  *       mapping->i_mmap_mutex
- *         anon_vma->lock
+ *         anon_vma->mutex
  *           mm->page_table_lock or pte_lock
  *             zone->lru_lock (in mark_page_accessed, isolate_lru_page)
  *             swap_lock (in swap_duplicate, swap_info_get)
@@ -39,7 +39,7 @@
  *
  * (code doesn't rely on that order so it could be switched around)
  * ->tasklist_lock
- *   anon_vma->lock      (memory_failure, collect_procs_anon)
+ *   anon_vma->mutex      (memory_failure, collect_procs_anon)
  *     pte map lock
  */
 
@@ -306,7 +306,7 @@ static void anon_vma_ctor(void *data)
 {
 	struct anon_vma *anon_vma = data;
 
-	spin_lock_init(&anon_vma->lock);
+	mutex_init(&anon_vma->mutex);
 	atomic_set(&anon_vma->refcount, 0);
 	INIT_LIST_HEAD(&anon_vma->head);
 }
@@ -1129,7 +1129,7 @@ int try_to_unmap_one(struct page *page, 
 	/*
 	 * We need mmap_sem locking, Otherwise VM_LOCKED check makes
 	 * unstable result and race. Plus, We can't wait here because
-	 * we now hold anon_vma->lock or mapping->i_mmap_mutex.
+	 * we now hold anon_vma->mutex or mapping->i_mmap_mutex.
 	 * if trylock failed, the page remain in evictable lru and later
 	 * vmscan could retry to move the page to unevictable lru if the
 	 * page is actually mlocked.
Index: linux-2.6/mm/mmap.c
===================================================================
--- linux-2.6.orig/mm/mmap.c
+++ linux-2.6/mm/mmap.c
@@ -2523,15 +2523,15 @@ static void vm_lock_anon_vma(struct mm_s
 		 * The LSB of head.next can't change from under us
 		 * because we hold the mm_all_locks_mutex.
 		 */
-		spin_lock_nest_lock(&anon_vma->root->lock, &mm->mmap_sem);
+		mutex_lock_nest_lock(&anon_vma->root->mutex, &mm->mmap_sem);
 		/*
 		 * We can safely modify head.next after taking the
-		 * anon_vma->root->lock. If some other vma in this mm shares
+		 * anon_vma->root->mutex. If some other vma in this mm shares
 		 * the same anon_vma we won't take it again.
 		 *
 		 * No need of atomic instructions here, head.next
 		 * can't change from under us thanks to the
-		 * anon_vma->root->lock.
+		 * anon_vma->root->mutex.
 		 */
 		if (__test_and_set_bit(0, (unsigned long *)
 				       &anon_vma->root->head.next))
@@ -2580,7 +2580,7 @@ static void vm_lock_mapping(struct mm_st
  * vma in this mm is backed by the same anon_vma or address_space.
  *
  * We can take all the locks in random order because the VM code
- * taking i_mmap_mutex or anon_vma->lock outside the mmap_sem never
+ * taking i_mmap_mutex or anon_vma->mutex outside the mmap_sem never
  * takes more than one of them in a row. Secondly we're protected
  * against a concurrent mm_take_all_locks() by the mm_all_locks_mutex.
  *
@@ -2636,7 +2636,7 @@ static void vm_unlock_anon_vma(struct an
 		 *
 		 * No need of atomic instructions here, head.next
 		 * can't change from under us until we release the
-		 * anon_vma->root->lock.
+		 * anon_vma->root->mutex.
 		 */
 		if (!__test_and_clear_bit(0, (unsigned long *)
 					  &anon_vma->root->head.next))
Index: linux-2.6/include/linux/mmu_notifier.h
===================================================================
--- linux-2.6.orig/include/linux/mmu_notifier.h
+++ linux-2.6/include/linux/mmu_notifier.h
@@ -150,7 +150,7 @@ struct mmu_notifier_ops {
  * Therefore notifier chains can only be traversed when either
  *
  * 1. mmap_sem is held.
- * 2. One of the reverse map locks is held (i_mmap_mutex or anon_vma->lock).
+ * 2. One of the reverse map locks is held (i_mmap_mutex or anon_vma->mutex).
  * 3. No other concurrent thread can access the list (release)
  */
 struct mmu_notifier {
Index: linux-2.6/mm/huge_memory.c
===================================================================
--- linux-2.6.orig/mm/huge_memory.c
+++ linux-2.6/mm/huge_memory.c
@@ -1128,7 +1128,7 @@ static int __split_huge_page_splitting(s
 		 * We can't temporarily set the pmd to null in order
 		 * to split it, the pmd must remain marked huge at all
 		 * times or the VM won't take the pmd_trans_huge paths
-		 * and it won't wait on the anon_vma->root->lock to
+		 * and it won't wait on the anon_vma->root->mutex to
 		 * serialize against split_huge_page*.
 		 */
 		pmdp_splitting_flush_notify(vma, address, pmd);
@@ -1315,7 +1315,7 @@ static int __split_huge_page_map(struct 
 	return ret;
 }
 
-/* must be called with anon_vma->root->lock hold */
+/* must be called with anon_vma->root->mutex hold */
 static void __split_huge_page(struct page *page,
 			      struct anon_vma *anon_vma)
 {
Index: linux-2.6/include/linux/huge_mm.h
===================================================================
--- linux-2.6.orig/include/linux/huge_mm.h
+++ linux-2.6/include/linux/huge_mm.h
@@ -91,12 +91,8 @@ extern void __split_huge_page_pmd(struct
 #define wait_split_huge_page(__anon_vma, __pmd)				\
 	do {								\
 		pmd_t *____pmd = (__pmd);				\
-		spin_unlock_wait(&(__anon_vma)->root->lock);		\
-		/*							\
-		 * spin_unlock_wait() is just a loop in C and so the	\
-		 * CPU can reorder anything around it.			\
-		 */							\
-		smp_mb();						\
+		anon_vma_lock(__anon_vma);				\
+		anon_vma_unlock(__anon_vma);				\
 		BUG_ON(pmd_trans_splitting(*____pmd) ||			\
 		       pmd_trans_huge(*____pmd));			\
 	} while (0)


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 20/20] mm: Optimize page_lock_anon_vma() fast-path

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-mm-optimize_page_lock_anon_vma_fast-path.patch --]
[-- Type: text/plain, Size: 3485 bytes --]

Optimize the page_lock_anon_vma() fast path to be one atomic op,
instead of two.

Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 mm/rmap.c |   86 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++---
 1 file changed, 82 insertions(+), 4 deletions(-)

Index: linux-2.6/mm/rmap.c
===================================================================
--- linux-2.6.orig/mm/rmap.c
+++ linux-2.6/mm/rmap.c
@@ -85,6 +85,29 @@ static inline struct anon_vma *anon_vma_
 static inline void anon_vma_free(struct anon_vma *anon_vma)
 {
 	VM_BUG_ON(atomic_read(&anon_vma->refcount));
+
+	/*
+	 * Synchronize against page_lock_anon_vma() such that
+	 * we can safely hold the lock without the anon_vma getting
+	 * freed.
+	 *
+	 * Relies on the full mb implied by the atomic_dec_and_test() from
+	 * put_anon_vma() against the acquire barrier implied by
+	 * mutex_trylock() from page_lock_anon_vma(). This orders:
+	 *
+	 * page_lock_anon_vma()		VS	put_anon_vma()
+	 *   mutex_trylock()			  atomic_dec_and_test()
+	 *   LOCK				  MB
+	 *   atomic_read()			  mutex_is_locked()
+	 *
+	 * LOCK should suffice since the actual taking of the lock must
+	 * happen _before_ what follows.
+	 */
+	if (mutex_is_locked(&anon_vma->root->mutex)) {
+		anon_vma_lock(anon_vma);
+		anon_vma_unlock(anon_vma);
+	}
+
 	kmem_cache_free(anon_vma_cachep, anon_vma);
 }
 
@@ -371,20 +394,75 @@ struct anon_vma *page_get_anon_vma(struc
 	return anon_vma;
 }
 
+/*
+ * Similar to page_get_anon_vma() except it locks the anon_vma.
+ *
+ * Its a little more complex as it tries to keep the fast path to a single
+ * atomic op -- the trylock. If we fail the trylock, we fall back to getting a
+ * reference like with page_get_anon_vma() and then block on the mutex.
+ */
 struct anon_vma *page_lock_anon_vma(struct page *page)
 {
-	struct anon_vma *anon_vma = page_get_anon_vma(page);
+	struct anon_vma *anon_vma = NULL;
+	unsigned long anon_mapping;
 
-	if (anon_vma)
-		anon_vma_lock(anon_vma);
+	rcu_read_lock();
+	anon_mapping = (unsigned long) ACCESS_ONCE(page->mapping);
+	if ((anon_mapping & PAGE_MAPPING_FLAGS) != PAGE_MAPPING_ANON)
+		goto out;
+	if (!page_mapped(page))
+		goto out;
+
+	anon_vma = (struct anon_vma *) (anon_mapping - PAGE_MAPPING_ANON);
+	if (mutex_trylock(&anon_vma->root->mutex)) {
+		/*
+		 * If we observe a !0 refcount, then holding the lock ensures
+		 * the anon_vma will not go away, see __put_anon_vma().
+		 */
+		if (!atomic_read(&anon_vma->refcount)) {
+			anon_vma_unlock(anon_vma);
+			anon_vma = NULL;
+		}
+		goto out;
+	}
+
+	/* trylock failed, we got to sleep */
+	if (!atomic_inc_not_zero(&anon_vma->refcount)) {
+		anon_vma = NULL;
+		goto out;
+	}
 
+	if (!page_mapped(page)) {
+		put_anon_vma(anon_vma);
+		anon_vma = NULL;
+		goto out;
+	}
+
+	/* we pinned the anon_vma, its safe to sleep */
+	rcu_read_unlock();
+	anon_vma_lock(anon_vma);
+
+	if (atomic_dec_and_test(&anon_vma->refcount)) {
+		/*
+		 * Oops, we held the last refcount, release the lock
+		 * and bail -- can't simply use put_anon_vma() because
+		 * we'll deadlock on the anon_vma_lock() recursion.
+		 */
+		anon_vma_unlock(anon_vma);
+		__put_anon_vma(anon_vma);
+		anon_vma = NULL;
+	}
+
+	return anon_vma;
+
+out:
+	rcu_read_unlock();
 	return anon_vma;
 }
 
 void page_unlock_anon_vma(struct anon_vma *anon_vma)
 {
 	anon_vma_unlock(anon_vma);
-	put_anon_vma(anon_vma);
 }
 
 /*

[PATCH 20/20] mm: Optimize page_lock_anon_vma() fast-path

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-mm-optimize_page_lock_anon_vma_fast-path.patch --]
[-- Type: text/plain, Size: 3788 bytes --]

Optimize the page_lock_anon_vma() fast path to be one atomic op,
instead of two.

Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 mm/rmap.c |   86 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++---
 1 file changed, 82 insertions(+), 4 deletions(-)

Index: linux-2.6/mm/rmap.c
===================================================================
--- linux-2.6.orig/mm/rmap.c
+++ linux-2.6/mm/rmap.c
@@ -85,6 +85,29 @@ static inline struct anon_vma *anon_vma_
 static inline void anon_vma_free(struct anon_vma *anon_vma)
 {
 	VM_BUG_ON(atomic_read(&anon_vma->refcount));
+
+	/*
+	 * Synchronize against page_lock_anon_vma() such that
+	 * we can safely hold the lock without the anon_vma getting
+	 * freed.
+	 *
+	 * Relies on the full mb implied by the atomic_dec_and_test() from
+	 * put_anon_vma() against the acquire barrier implied by
+	 * mutex_trylock() from page_lock_anon_vma(). This orders:
+	 *
+	 * page_lock_anon_vma()		VS	put_anon_vma()
+	 *   mutex_trylock()			  atomic_dec_and_test()
+	 *   LOCK				  MB
+	 *   atomic_read()			  mutex_is_locked()
+	 *
+	 * LOCK should suffice since the actual taking of the lock must
+	 * happen _before_ what follows.
+	 */
+	if (mutex_is_locked(&anon_vma->root->mutex)) {
+		anon_vma_lock(anon_vma);
+		anon_vma_unlock(anon_vma);
+	}
+
 	kmem_cache_free(anon_vma_cachep, anon_vma);
 }
 
@@ -371,20 +394,75 @@ struct anon_vma *page_get_anon_vma(struc
 	return anon_vma;
 }
 
+/*
+ * Similar to page_get_anon_vma() except it locks the anon_vma.
+ *
+ * Its a little more complex as it tries to keep the fast path to a single
+ * atomic op -- the trylock. If we fail the trylock, we fall back to getting a
+ * reference like with page_get_anon_vma() and then block on the mutex.
+ */
 struct anon_vma *page_lock_anon_vma(struct page *page)
 {
-	struct anon_vma *anon_vma = page_get_anon_vma(page);
+	struct anon_vma *anon_vma = NULL;
+	unsigned long anon_mapping;
 
-	if (anon_vma)
-		anon_vma_lock(anon_vma);
+	rcu_read_lock();
+	anon_mapping = (unsigned long) ACCESS_ONCE(page->mapping);
+	if ((anon_mapping & PAGE_MAPPING_FLAGS) != PAGE_MAPPING_ANON)
+		goto out;
+	if (!page_mapped(page))
+		goto out;
+
+	anon_vma = (struct anon_vma *) (anon_mapping - PAGE_MAPPING_ANON);
+	if (mutex_trylock(&anon_vma->root->mutex)) {
+		/*
+		 * If we observe a !0 refcount, then holding the lock ensures
+		 * the anon_vma will not go away, see __put_anon_vma().
+		 */
+		if (!atomic_read(&anon_vma->refcount)) {
+			anon_vma_unlock(anon_vma);
+			anon_vma = NULL;
+		}
+		goto out;
+	}
+
+	/* trylock failed, we got to sleep */
+	if (!atomic_inc_not_zero(&anon_vma->refcount)) {
+		anon_vma = NULL;
+		goto out;
+	}
 
+	if (!page_mapped(page)) {
+		put_anon_vma(anon_vma);
+		anon_vma = NULL;
+		goto out;
+	}
+
+	/* we pinned the anon_vma, its safe to sleep */
+	rcu_read_unlock();
+	anon_vma_lock(anon_vma);
+
+	if (atomic_dec_and_test(&anon_vma->refcount)) {
+		/*
+		 * Oops, we held the last refcount, release the lock
+		 * and bail -- can't simply use put_anon_vma() because
+		 * we'll deadlock on the anon_vma_lock() recursion.
+		 */
+		anon_vma_unlock(anon_vma);
+		__put_anon_vma(anon_vma);
+		anon_vma = NULL;
+	}
+
+	return anon_vma;
+
+out:
+	rcu_read_unlock();
 	return anon_vma;
 }
 
 void page_unlock_anon_vma(struct anon_vma *anon_vma)
 {
 	anon_vma_unlock(anon_vma);
-	put_anon_vma(anon_vma);
 }
 
 /*


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 20/20] mm: Optimize page_lock_anon_vma() fast-path

[Peter Zijlstra]

[-- Attachment #1: peter_zijlstra-mm-optimize_page_lock_anon_vma_fast-path.patch --]
[-- Type: text/plain, Size: 3788 bytes --]

Optimize the page_lock_anon_vma() fast path to be one atomic op,
instead of two.

Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
---
 mm/rmap.c |   86 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++---
 1 file changed, 82 insertions(+), 4 deletions(-)

Index: linux-2.6/mm/rmap.c
===================================================================
--- linux-2.6.orig/mm/rmap.c
+++ linux-2.6/mm/rmap.c
@@ -85,6 +85,29 @@ static inline struct anon_vma *anon_vma_
 static inline void anon_vma_free(struct anon_vma *anon_vma)
 {
 	VM_BUG_ON(atomic_read(&anon_vma->refcount));
+
+	/*
+	 * Synchronize against page_lock_anon_vma() such that
+	 * we can safely hold the lock without the anon_vma getting
+	 * freed.
+	 *
+	 * Relies on the full mb implied by the atomic_dec_and_test() from
+	 * put_anon_vma() against the acquire barrier implied by
+	 * mutex_trylock() from page_lock_anon_vma(). This orders:
+	 *
+	 * page_lock_anon_vma()		VS	put_anon_vma()
+	 *   mutex_trylock()			  atomic_dec_and_test()
+	 *   LOCK				  MB
+	 *   atomic_read()			  mutex_is_locked()
+	 *
+	 * LOCK should suffice since the actual taking of the lock must
+	 * happen _before_ what follows.
+	 */
+	if (mutex_is_locked(&anon_vma->root->mutex)) {
+		anon_vma_lock(anon_vma);
+		anon_vma_unlock(anon_vma);
+	}
+
 	kmem_cache_free(anon_vma_cachep, anon_vma);
 }
 
@@ -371,20 +394,75 @@ struct anon_vma *page_get_anon_vma(struc
 	return anon_vma;
 }
 
+/*
+ * Similar to page_get_anon_vma() except it locks the anon_vma.
+ *
+ * Its a little more complex as it tries to keep the fast path to a single
+ * atomic op -- the trylock. If we fail the trylock, we fall back to getting a
+ * reference like with page_get_anon_vma() and then block on the mutex.
+ */
 struct anon_vma *page_lock_anon_vma(struct page *page)
 {
-	struct anon_vma *anon_vma = page_get_anon_vma(page);
+	struct anon_vma *anon_vma = NULL;
+	unsigned long anon_mapping;
 
-	if (anon_vma)
-		anon_vma_lock(anon_vma);
+	rcu_read_lock();
+	anon_mapping = (unsigned long) ACCESS_ONCE(page->mapping);
+	if ((anon_mapping & PAGE_MAPPING_FLAGS) != PAGE_MAPPING_ANON)
+		goto out;
+	if (!page_mapped(page))
+		goto out;
+
+	anon_vma = (struct anon_vma *) (anon_mapping - PAGE_MAPPING_ANON);
+	if (mutex_trylock(&anon_vma->root->mutex)) {
+		/*
+		 * If we observe a !0 refcount, then holding the lock ensures
+		 * the anon_vma will not go away, see __put_anon_vma().
+		 */
+		if (!atomic_read(&anon_vma->refcount)) {
+			anon_vma_unlock(anon_vma);
+			anon_vma = NULL;
+		}
+		goto out;
+	}
+
+	/* trylock failed, we got to sleep */
+	if (!atomic_inc_not_zero(&anon_vma->refcount)) {
+		anon_vma = NULL;
+		goto out;
+	}
 
+	if (!page_mapped(page)) {
+		put_anon_vma(anon_vma);
+		anon_vma = NULL;
+		goto out;
+	}
+
+	/* we pinned the anon_vma, its safe to sleep */
+	rcu_read_unlock();
+	anon_vma_lock(anon_vma);
+
+	if (atomic_dec_and_test(&anon_vma->refcount)) {
+		/*
+		 * Oops, we held the last refcount, release the lock
+		 * and bail -- can't simply use put_anon_vma() because
+		 * we'll deadlock on the anon_vma_lock() recursion.
+		 */
+		anon_vma_unlock(anon_vma);
+		__put_anon_vma(anon_vma);
+		anon_vma = NULL;
+	}
+
+	return anon_vma;
+
+out:
+	rcu_read_unlock();
 	return anon_vma;
 }
 
 void page_unlock_anon_vma(struct anon_vma *anon_vma)
 {
 	anon_vma_unlock(anon_vma);
-	put_anon_vma(anon_vma);
 }
 
 /*


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 00/20] mm: Preemptibility -v10

[Peter Zijlstra]

On Fri, 2011-04-01 at 14:12 +0200, Peter Zijlstra wrote:
> 
> Also provided is a rollup of these patches, which is used as a commit in the
> git tree referenced below. 

If only I'd actually added it ;-)

---
Subject: mm: mmu_gather rework
From: Peter Zijlstra <a.p.zijlstra@chello.nl>
Date: Fri, 26 Nov 2010 15:38:51 +0100

Remove the first obstackle towards a fully preemptible mmu_gather.

The current scheme assumes mmu_gather is always done with preemption
disabled and uses per-cpu storage for the page batches. Change this to
try and allocate a page for batching and in case of failure, use a
small on-stack array to make some progress.

Preemptible mmu_gather is desired in general and usable once
i_mmap_lock becomes a mutex. Doing it before the mutex conversion
saves us from having to rework the code by moving the mmu_gather
bits inside the pte_lock.

Also avoid flushing the tlb batches from under the pte lock,
this is useful even without the i_mmap_lock conversion as it
significantly reduces pte lock hold times.

Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: David Miller <davem@davemloft.net>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Russell King <rmk@arm.linux.org.uk>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: Jeff Dike <jdike@addtoit.com>
Acked-by: Tony Luck <tony.luck@intel.com>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Hugh Dickins <hughd@google.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
---
 arch/alpha/mm/init.c                   |    2 -
 arch/arm/include/asm/tlb.h             |   53 ++++++++++++-----
 arch/arm/mm/mmu.c                      |    2 -
 arch/avr32/mm/init.c                   |    2 -
 arch/cris/mm/init.c                    |    2 -
 arch/frv/mm/init.c                     |    2 -
 arch/ia64/include/asm/tlb.h            |   66 +++++++++++++++-------
 arch/ia64/mm/init.c                    |    2 -
 arch/m32r/mm/init.c                    |    2 -
 arch/m68k/mm/init_mm.c                 |    2 -
 arch/microblaze/mm/init.c              |    2 -
 arch/mips/mm/init.c                    |    2 -
 arch/mn10300/mm/init.c                 |    2 -
 arch/parisc/mm/init.c                  |    2 -
 arch/powerpc/include/asm/pgalloc.h     |    4 +-
 arch/powerpc/include/asm/thread_info.h |    2 +
 arch/powerpc/include/asm/tlb.h         |   10 +++
 arch/powerpc/kernel/process.c          |   23 +++++++-
 arch/powerpc/mm/pgtable.c              |   14 +---
 arch/powerpc/mm/tlb_hash32.c           |    2 +-
 arch/powerpc/mm/tlb_hash64.c           |    6 +-
 arch/powerpc/mm/tlb_nohash.c           |    2 +-
 arch/s390/include/asm/tlb.h            |   62 ++++++++++++--------
 arch/s390/mm/pgtable.c                 |    1 -
 arch/score/mm/init.c                   |    2 -
 arch/sh/include/asm/tlb.h              |   28 ++++++----
 arch/sh/mm/init.c                      |    1 -
 arch/sparc/include/asm/pgalloc_64.h    |    3 +
 arch/sparc/include/asm/pgtable_64.h    |   15 ++++-
 arch/sparc/include/asm/tlb_64.h        |   91 ++----------------------------
 arch/sparc/include/asm/tlbflush_64.h   |   12 +++-
 arch/sparc/mm/init_32.c                |    2 -
 arch/sparc/mm/tlb.c                    |   43 ++++++++------
 arch/sparc/mm/tsb.c                    |   15 +++--
 arch/tile/mm/init.c                    |    2 -
 arch/um/include/asm/tlb.h              |   29 ++++------
 arch/um/kernel/smp.c                   |    3 -
 arch/unicore32/mm/mmu.c                |    2 -
 arch/x86/mm/init.c                     |    2 -
 arch/xtensa/mm/mmu.c                   |    2 -
 fs/exec.c                              |   10 ++--
 include/asm-generic/tlb.h              |   96 +++++++++++++++++++++++---------
 include/linux/mm.h                     |    2 +-
 mm/memory.c                            |   46 ++++++++--------
 mm/mmap.c                              |   18 +++---
 45 files changed, 364 insertions(+), 329 deletions(-)

diff --git a/arch/alpha/mm/init.c b/arch/alpha/mm/init.c
index 86425ab..69d0c57 100644
--- a/arch/alpha/mm/init.c
+++ b/arch/alpha/mm/init.c
@@ -32,8 +32,6 @@
 #include <asm/console.h>
 #include <asm/tlb.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 extern void die_if_kernel(char *,struct pt_regs *,long);
 
 static struct pcb_struct original_pcb;
diff --git a/arch/arm/include/asm/tlb.h b/arch/arm/include/asm/tlb.h
index 82dfe5d..265f908 100644
--- a/arch/arm/include/asm/tlb.h
+++ b/arch/arm/include/asm/tlb.h
@@ -41,12 +41,12 @@
  */
 #if defined(CONFIG_SMP) || defined(CONFIG_CPU_32v7)
 #define tlb_fast_mode(tlb)	0
-#define FREE_PTE_NR		500
 #else
 #define tlb_fast_mode(tlb)	1
-#define FREE_PTE_NR		0
 #endif
 
+#define MMU_GATHER_BUNDLE	8
+
 /*
  * TLB handling.  This allows us to remove pages from the page
  * tables, and efficiently handle the TLB issues.
@@ -58,7 +58,9 @@ struct mmu_gather {
 	unsigned long		range_start;
 	unsigned long		range_end;
 	unsigned int		nr;
-	struct page		*pages[FREE_PTE_NR];
+	unsigned int		max;
+	struct page		**pages;
+	struct page		*local[MMU_GATHER_BUNDLE];
 };
 
 DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
@@ -97,26 +99,37 @@ static inline void tlb_add_flush(struct mmu_gather *tlb, unsigned long addr)
 	}
 }
 
+static inline void __tlb_alloc_page(struct mmu_gather *tlb)
+{
+	unsigned long addr = __get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
+
+	if (addr) {
+		tlb->pages = (void *)addr;
+		tlb->max = PAGE_SIZE / sizeof(struct page *);
+	}
+}
+
 static inline void tlb_flush_mmu(struct mmu_gather *tlb)
 {
 	tlb_flush(tlb);
 	if (!tlb_fast_mode(tlb)) {
 		free_pages_and_swap_cache(tlb->pages, tlb->nr);
 		tlb->nr = 0;
+		if (tlb->pages == tlb->local)
+			__tlb_alloc_page(tlb);
 	}
 }
 
-static inline struct mmu_gather *
-tlb_gather_mmu(struct mm_struct *mm, unsigned int full_mm_flush)
+static inline void
+tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, bool fullmm)
 {
-	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);
-
 	tlb->mm = mm;
-	tlb->fullmm = full_mm_flush;
+	tlb->fullmm = fullmm;
 	tlb->vma = NULL;
+	tlb->max = ARRAY_SIZE(tlb->local);
+	tlb->pages = tlb->local;
 	tlb->nr = 0;
-
-	return tlb;
+	__tlb_alloc_page(tlb);
 }
 
 static inline void
@@ -127,7 +140,8 @@ tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
 
-	put_cpu_var(mmu_gathers);
+	if (tlb->pages != tlb->local)
+		free_pages((unsigned long)tlb->pages, 0);
 }
 
 /*
@@ -162,15 +176,22 @@ tlb_end_vma(struct mmu_gather *tlb, struct vm_area_struct *vma)
 		tlb_flush(tlb);
 }
 
-static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
 {
 	if (tlb_fast_mode(tlb)) {
 		free_page_and_swap_cache(page);
-	} else {
-		tlb->pages[tlb->nr++] = page;
-		if (tlb->nr >= FREE_PTE_NR)
-			tlb_flush_mmu(tlb);
+		return 1; /* avoid calling tlb_flush_mmu */
 	}
+
+	tlb->pages[tlb->nr++] = page;
+	VM_BUG_ON(tlb->nr > tlb->max);
+	return tlb->max - tlb->nr;
+}
+
+static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	if (!__tlb_remove_page(tlb, page))
+		tlb_flush_mmu(tlb);
 }
 
 static inline void __pte_free_tlb(struct mmu_gather *tlb, pgtable_t pte,
diff --git a/arch/arm/mm/mmu.c b/arch/arm/mm/mmu.c
index 6cf76b3..08a9236 100644
--- a/arch/arm/mm/mmu.c
+++ b/arch/arm/mm/mmu.c
@@ -31,8 +31,6 @@
 
 #include "mm.h"
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /*
  * empty_zero_page is a special page that is used for
  * zero-initialized data and COW.
diff --git a/arch/avr32/mm/init.c b/arch/avr32/mm/init.c
index a7314d4..2798c2d 100644
--- a/arch/avr32/mm/init.c
+++ b/arch/avr32/mm/init.c
@@ -25,8 +25,6 @@
 #include <asm/setup.h>
 #include <asm/sections.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 pgd_t swapper_pg_dir[PTRS_PER_PGD] __page_aligned_data;
 
 struct page *empty_zero_page;
diff --git a/arch/cris/mm/init.c b/arch/cris/mm/init.c
index df33ab8..d72ab58 100644
--- a/arch/cris/mm/init.c
+++ b/arch/cris/mm/init.c
@@ -13,8 +13,6 @@
 #include <linux/bootmem.h>
 #include <asm/tlb.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 unsigned long empty_zero_page;
 
 extern char _stext, _edata, _etext; /* From linkerscript */
diff --git a/arch/frv/mm/init.c b/arch/frv/mm/init.c
index ed64588..fbe5f0d 100644
--- a/arch/frv/mm/init.c
+++ b/arch/frv/mm/init.c
@@ -41,8 +41,6 @@
 
 #undef DEBUG
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /*
  * BAD_PAGE is the page that is used for page faults when linux
  * is out-of-memory. Older versions of linux just did a
diff --git a/arch/ia64/include/asm/tlb.h b/arch/ia64/include/asm/tlb.h
index 23cce99..c3ffe3e 100644
--- a/arch/ia64/include/asm/tlb.h
+++ b/arch/ia64/include/asm/tlb.h
@@ -47,21 +47,27 @@
 #include <asm/machvec.h>
 
 #ifdef CONFIG_SMP
-# define FREE_PTE_NR		2048
 # define tlb_fast_mode(tlb)	((tlb)->nr == ~0U)
 #else
-# define FREE_PTE_NR		0
 # define tlb_fast_mode(tlb)	(1)
 #endif
 
+/*
+ * If we can't allocate a page to make a big batch of page pointers
+ * to work on, then just handle a few from the on-stack structure.
+ */
+#define	IA64_GATHER_BUNDLE	8
+
 struct mmu_gather {
 	struct mm_struct	*mm;
 	unsigned int		nr;		/* == ~0U => fast mode */
+	unsigned int		max;
 	unsigned char		fullmm;		/* non-zero means full mm flush */
 	unsigned char		need_flush;	/* really unmapped some PTEs? */
 	unsigned long		start_addr;
 	unsigned long		end_addr;
-	struct page 		*pages[FREE_PTE_NR];
+	struct page		**pages;
+	struct page		*local[IA64_GATHER_BUNDLE];
 };
 
 struct ia64_tr_entry {
@@ -90,9 +96,6 @@ extern struct ia64_tr_entry *ia64_idtrs[NR_CPUS];
 #define RR_RID_MASK	0x00000000ffffff00L
 #define RR_TO_RID(val) 	((val >> 8) & 0xffffff)
 
-/* Users of the generic TLB shootdown code must declare this storage space. */
-DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /*
  * Flush the TLB for address range START to END and, if not in fast mode, release the
  * freed pages that where gathered up to this point.
@@ -147,15 +150,23 @@ ia64_tlb_flush_mmu (struct mmu_gather *tlb, unsigned long start, unsigned long e
 	}
 }
 
-/*
- * Return a pointer to an initialized struct mmu_gather.
- */
-static inline struct mmu_gather *
-tlb_gather_mmu (struct mm_struct *mm, unsigned int full_mm_flush)
+static inline void __tlb_alloc_page(struct mmu_gather *tlb)
 {
-	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);
+	unsigned long addr = __get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
 
+	if (addr) {
+		tlb->pages = (void *)addr;
+		tlb->max = PAGE_SIZE / sizeof(void *);
+	}
+}
+
+
+static inline void
+tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned int full_mm_flush)
+{
 	tlb->mm = mm;
+	tlb->max = ARRAY_SIZE(tlb->local);
+	tlb->pages = tlb->local;
 	/*
 	 * Use fast mode if only 1 CPU is online.
 	 *
@@ -172,7 +183,6 @@ tlb_gather_mmu (struct mm_struct *mm, unsigned int full_mm_flush)
 	tlb->nr = (num_online_cpus() == 1) ? ~0U : 0;
 	tlb->fullmm = full_mm_flush;
 	tlb->start_addr = ~0UL;
-	return tlb;
 }
 
 /*
@@ -180,7 +190,7 @@ tlb_gather_mmu (struct mm_struct *mm, unsigned int full_mm_flush)
  * collected.
  */
 static inline void
-tlb_finish_mmu (struct mmu_gather *tlb, unsigned long start, unsigned long end)
+tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
 {
 	/*
 	 * Note: tlb->nr may be 0 at this point, so we can't rely on tlb->start_addr and
@@ -191,7 +201,8 @@ tlb_finish_mmu (struct mmu_gather *tlb, unsigned long start, unsigned long end)
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
 
-	put_cpu_var(mmu_gathers);
+	if (tlb->pages != tlb->local)
+		free_pages((unsigned long)tlb->pages, 0);
 }
 
 /*
@@ -199,18 +210,33 @@ tlb_finish_mmu (struct mmu_gather *tlb, unsigned long start, unsigned long end)
  * must be delayed until after the TLB has been flushed (see comments at the beginning of
  * this file).
  */
-static inline void
-tlb_remove_page (struct mmu_gather *tlb, struct page *page)
+static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
 {
 	tlb->need_flush = 1;
 
 	if (tlb_fast_mode(tlb)) {
 		free_page_and_swap_cache(page);
-		return;
+		return 1; /* avoid calling tlb_flush_mmu */
 	}
+
+	if (!tlb->nr && tlb->pages == tlb->local)
+		__tlb_alloc_page(tlb);
+
 	tlb->pages[tlb->nr++] = page;
-	if (tlb->nr >= FREE_PTE_NR)
-		ia64_tlb_flush_mmu(tlb, tlb->start_addr, tlb->end_addr);
+	VM_BUG_ON(tlb->nr > tlb->max);
+
+	return tlb->max - tlb->nr;
+}
+
+static inline void tlb_flush_mmu(struct mmu_gather *tlb)
+{
+	ia64_tlb_flush_mmu(tlb, tlb->start_addr, tlb->end_addr);
+}
+
+static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	if (!__tlb_remove_page(tlb, page))
+		tlb_flush_mmu(tlb);
 }
 
 /*
diff --git a/arch/ia64/mm/init.c b/arch/ia64/mm/init.c
index ed41759..00cb0e2 100644
--- a/arch/ia64/mm/init.c
+++ b/arch/ia64/mm/init.c
@@ -36,8 +36,6 @@
 #include <asm/mca.h>
 #include <asm/paravirt.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 extern void ia64_tlb_init (void);
 
 unsigned long MAX_DMA_ADDRESS = PAGE_OFFSET + 0x100000000UL;
diff --git a/arch/m32r/mm/init.c b/arch/m32r/mm/init.c
index 73e2205..78b660e 100644
--- a/arch/m32r/mm/init.c
+++ b/arch/m32r/mm/init.c
@@ -35,8 +35,6 @@ extern char __init_begin, __init_end;
 
 pgd_t swapper_pg_dir[1024];
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /*
  * Cache of MMU context last used.
  */
diff --git a/arch/m68k/mm/init_mm.c b/arch/m68k/mm/init_mm.c
index 8bc8425..9113c2f 100644
--- a/arch/m68k/mm/init_mm.c
+++ b/arch/m68k/mm/init_mm.c
@@ -32,8 +32,6 @@
 #include <asm/sections.h>
 #include <asm/tlb.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 pg_data_t pg_data_map[MAX_NUMNODES];
 EXPORT_SYMBOL(pg_data_map);
 
diff --git a/arch/microblaze/mm/init.c b/arch/microblaze/mm/init.c
index c843786..213f2d6 100644
--- a/arch/microblaze/mm/init.c
+++ b/arch/microblaze/mm/init.c
@@ -32,8 +32,6 @@ unsigned int __page_offset;
 EXPORT_SYMBOL(__page_offset);
 
 #else
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 static int init_bootmem_done;
 #endif /* CONFIG_MMU */
 
diff --git a/arch/mips/mm/init.c b/arch/mips/mm/init.c
index 279599e..1aadeb4 100644
--- a/arch/mips/mm/init.c
+++ b/arch/mips/mm/init.c
@@ -64,8 +64,6 @@
 
 #endif /* CONFIG_MIPS_MT_SMTC */
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /*
  * We have up to 8 empty zeroed pages so we can map one of the right colour
  * when needed.  This is necessary only on R4000 / R4400 SC and MC versions
diff --git a/arch/mn10300/mm/init.c b/arch/mn10300/mm/init.c
index 48907cc..1380182 100644
--- a/arch/mn10300/mm/init.c
+++ b/arch/mn10300/mm/init.c
@@ -37,8 +37,6 @@
 #include <asm/tlb.h>
 #include <asm/sections.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 unsigned long highstart_pfn, highend_pfn;
 
 #ifdef CONFIG_MN10300_HAS_ATOMIC_OPS_UNIT
diff --git a/arch/parisc/mm/init.c b/arch/parisc/mm/init.c
index b7ed8d7..102f872 100644
--- a/arch/parisc/mm/init.c
+++ b/arch/parisc/mm/init.c
@@ -31,8 +31,6 @@
 #include <asm/mmzone.h>
 #include <asm/sections.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 extern int  data_start;
 
 #ifdef CONFIG_DISCONTIGMEM
diff --git a/arch/powerpc/include/asm/pgalloc.h b/arch/powerpc/include/asm/pgalloc.h
index abe8532..df1b4cb 100644
--- a/arch/powerpc/include/asm/pgalloc.h
+++ b/arch/powerpc/include/asm/pgalloc.h
@@ -32,13 +32,13 @@ static inline void pte_free(struct mm_struct *mm, pgtable_t ptepage)
 
 #ifdef CONFIG_SMP
 extern void pgtable_free_tlb(struct mmu_gather *tlb, void *table, unsigned shift);
-extern void pte_free_finish(void);
+extern void pte_free_finish(struct mmu_gather *tlb);
 #else /* CONFIG_SMP */
 static inline void pgtable_free_tlb(struct mmu_gather *tlb, void *table, unsigned shift)
 {
 	pgtable_free(table, shift);
 }
-static inline void pte_free_finish(void) { }
+static inline void pte_free_finish(struct mmu_gather *tlb) { }
 #endif /* !CONFIG_SMP */
 
 static inline void __pte_free_tlb(struct mmu_gather *tlb, struct page *ptepage,
diff --git a/arch/powerpc/include/asm/thread_info.h b/arch/powerpc/include/asm/thread_info.h
index d8529ef..37c353e 100644
--- a/arch/powerpc/include/asm/thread_info.h
+++ b/arch/powerpc/include/asm/thread_info.h
@@ -139,10 +139,12 @@ static inline struct thread_info *current_thread_info(void)
 #define TLF_NAPPING		0	/* idle thread enabled NAP mode */
 #define TLF_SLEEPING		1	/* suspend code enabled SLEEP mode */
 #define TLF_RESTORE_SIGMASK	2	/* Restore signal mask in do_signal */
+#define TLF_LAZY_MMU		3	/* tlb_batch is active */
 
 #define _TLF_NAPPING		(1 << TLF_NAPPING)
 #define _TLF_SLEEPING		(1 << TLF_SLEEPING)
 #define _TLF_RESTORE_SIGMASK	(1 << TLF_RESTORE_SIGMASK)
+#define _TLF_LAZY_MMU		(1 << TLF_LAZY_MMU)
 
 #ifndef __ASSEMBLY__
 #define HAVE_SET_RESTORE_SIGMASK	1
diff --git a/arch/powerpc/include/asm/tlb.h b/arch/powerpc/include/asm/tlb.h
index e2b428b..8f0ed7a 100644
--- a/arch/powerpc/include/asm/tlb.h
+++ b/arch/powerpc/include/asm/tlb.h
@@ -28,6 +28,16 @@
 #define tlb_start_vma(tlb, vma)	do { } while (0)
 #define tlb_end_vma(tlb, vma)	do { } while (0)
 
+#define HAVE_ARCH_MMU_GATHER 1
+
+struct pte_freelist_batch;
+
+struct arch_mmu_gather {
+	struct pte_freelist_batch *batch;
+};
+
+#define ARCH_MMU_GATHER_INIT (struct arch_mmu_gather){ .batch = NULL, }
+
 extern void tlb_flush(struct mmu_gather *tlb);
 
 /* Get the generic bits... */
diff --git a/arch/powerpc/kernel/process.c b/arch/powerpc/kernel/process.c
index f74f355..3e37f37 100644
--- a/arch/powerpc/kernel/process.c
+++ b/arch/powerpc/kernel/process.c
@@ -395,6 +395,9 @@ struct task_struct *__switch_to(struct task_struct *prev,
 	struct thread_struct *new_thread, *old_thread;
 	unsigned long flags;
 	struct task_struct *last;
+#ifdef CONFIG_PPC_BOOK3S_64
+	struct ppc64_tlb_batch *batch;
+#endif
 
 #ifdef CONFIG_SMP
 	/* avoid complexity of lazy save/restore of fpu
@@ -513,7 +516,17 @@ struct task_struct *__switch_to(struct task_struct *prev,
 		old_thread->accum_tb += (current_tb - start_tb);
 		new_thread->start_tb = current_tb;
 	}
-#endif
+#endif /* CONFIG_PPC64 */
+
+#ifdef CONFIG_PPC_BOOK3S_64
+	batch = &__get_cpu_var(ppc64_tlb_batch);
+	if (batch->active) {
+		current_thread_info()->local_flags |= _TLF_LAZY_MMU;
+		if (batch->index)
+			__flush_tlb_pending(batch);
+		batch->active = 0;
+	}
+#endif /* CONFIG_PPC_BOOK3S_64 */
 
 	local_irq_save(flags);
 
@@ -528,6 +541,14 @@ struct task_struct *__switch_to(struct task_struct *prev,
 	hard_irq_disable();
 	last = _switch(old_thread, new_thread);
 
+#ifdef CONFIG_PPC_BOOK3S_64
+	if (current_thread_info()->local_flags & _TLF_LAZY_MMU) {
+		current_thread_info()->local_flags &= ~_TLF_LAZY_MMU;
+		batch = &__get_cpu_var(ppc64_tlb_batch);
+		batch->active = 1;
+	}
+#endif /* CONFIG_PPC_BOOK3S_64 */
+
 	local_irq_restore(flags);
 
 	return last;
diff --git a/arch/powerpc/mm/pgtable.c b/arch/powerpc/mm/pgtable.c
index 6a3997f..6e72788 100644
--- a/arch/powerpc/mm/pgtable.c
+++ b/arch/powerpc/mm/pgtable.c
@@ -33,8 +33,6 @@
 
 #include "mmu_decl.h"
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 #ifdef CONFIG_SMP
 
 /*
@@ -43,7 +41,6 @@ DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
  * freeing a page table page that is being walked without locks
  */
 
-static DEFINE_PER_CPU(struct pte_freelist_batch *, pte_freelist_cur);
 static unsigned long pte_freelist_forced_free;
 
 struct pte_freelist_batch
@@ -97,12 +94,10 @@ static void pte_free_submit(struct pte_freelist_batch *batch)
 
 void pgtable_free_tlb(struct mmu_gather *tlb, void *table, unsigned shift)
 {
-	/* This is safe since tlb_gather_mmu has disabled preemption */
-	struct pte_freelist_batch **batchp = &__get_cpu_var(pte_freelist_cur);
+	struct pte_freelist_batch **batchp = &tlb->arch.batch;
 	unsigned long pgf;
 
-	if (atomic_read(&tlb->mm->mm_users) < 2 ||
-	    cpumask_equal(mm_cpumask(tlb->mm), cpumask_of(smp_processor_id()))){
+	if (atomic_read(&tlb->mm->mm_users) < 2) {
 		pgtable_free(table, shift);
 		return;
 	}
@@ -124,10 +119,9 @@ void pgtable_free_tlb(struct mmu_gather *tlb, void *table, unsigned shift)
 	}
 }
 
-void pte_free_finish(void)
+void pte_free_finish(struct mmu_gather *tlb)
 {
-	/* This is safe since tlb_gather_mmu has disabled preemption */
-	struct pte_freelist_batch **batchp = &__get_cpu_var(pte_freelist_cur);
+	struct pte_freelist_batch **batchp = &tlb->arch.batch;
 
 	if (*batchp == NULL)
 		return;
diff --git a/arch/powerpc/mm/tlb_hash32.c b/arch/powerpc/mm/tlb_hash32.c
index 690566b..d555cdb 100644
--- a/arch/powerpc/mm/tlb_hash32.c
+++ b/arch/powerpc/mm/tlb_hash32.c
@@ -73,7 +73,7 @@ void tlb_flush(struct mmu_gather *tlb)
 	}
 
 	/* Push out batch of freed page tables */
-	pte_free_finish();
+	pte_free_finish(tlb);
 }
 
 /*
diff --git a/arch/powerpc/mm/tlb_hash64.c b/arch/powerpc/mm/tlb_hash64.c
index c14d09f..5c94ca3 100644
--- a/arch/powerpc/mm/tlb_hash64.c
+++ b/arch/powerpc/mm/tlb_hash64.c
@@ -155,7 +155,7 @@ void __flush_tlb_pending(struct ppc64_tlb_batch *batch)
 
 void tlb_flush(struct mmu_gather *tlb)
 {
-	struct ppc64_tlb_batch *tlbbatch = &__get_cpu_var(ppc64_tlb_batch);
+	struct ppc64_tlb_batch *tlbbatch = &get_cpu_var(ppc64_tlb_batch);
 
 	/* If there's a TLB batch pending, then we must flush it because the
 	 * pages are going to be freed and we really don't want to have a CPU
@@ -164,8 +164,10 @@ void tlb_flush(struct mmu_gather *tlb)
 	if (tlbbatch->index)
 		__flush_tlb_pending(tlbbatch);
 
+	put_cpu_var(ppc64_tlb_batch);
+
 	/* Push out batch of freed page tables */
-	pte_free_finish();
+	pte_free_finish(tlb);
 }
 
 /**
diff --git a/arch/powerpc/mm/tlb_nohash.c b/arch/powerpc/mm/tlb_nohash.c
index 2a030d8..8eaf67d 100644
--- a/arch/powerpc/mm/tlb_nohash.c
+++ b/arch/powerpc/mm/tlb_nohash.c
@@ -301,7 +301,7 @@ void tlb_flush(struct mmu_gather *tlb)
 	flush_tlb_mm(tlb->mm);
 
 	/* Push out batch of freed page tables */
-	pte_free_finish();
+	pte_free_finish(tlb);
 }
 
 /*
diff --git a/arch/s390/include/asm/tlb.h b/arch/s390/include/asm/tlb.h
index 9074a54..77eee54 100644
--- a/arch/s390/include/asm/tlb.h
+++ b/arch/s390/include/asm/tlb.h
@@ -29,65 +29,77 @@
 #include <asm/smp.h>
 #include <asm/tlbflush.h>
 
-#ifndef CONFIG_SMP
-#define TLB_NR_PTRS	1
-#else
-#define TLB_NR_PTRS	508
-#endif
-
 struct mmu_gather {
 	struct mm_struct *mm;
 	unsigned int fullmm;
 	unsigned int nr_ptes;
 	unsigned int nr_pxds;
-	void *array[TLB_NR_PTRS];
+	unsigned int max;
+	void **array;
+	void *local[8];
 };
 
-DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
-
-static inline struct mmu_gather *tlb_gather_mmu(struct mm_struct *mm,
-						unsigned int full_mm_flush)
+static inline void __tlb_alloc_page(struct mmu_gather *tlb)
 {
-	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);
+	unsigned long addr = __get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
 
+	if (addr) {
+		tlb->array = (void *) addr;
+		tlb->max = PAGE_SIZE / sizeof(void *);
+	}
+}
+
+static inline void tlb_gather_mmu(struct mmu_gather *tlb,
+				  struct mm_struct *mm,
+				  unsigned int full_mm_flush)
+{
 	tlb->mm = mm;
+	tlb->max = ARRAY_SIZE(tlb->local);
+	tlb->array = tlb->local;
 	tlb->fullmm = full_mm_flush;
-	tlb->nr_ptes = 0;
-	tlb->nr_pxds = TLB_NR_PTRS;
 	if (tlb->fullmm)
 		__tlb_flush_mm(mm);
-	return tlb;
+	else
+		__tlb_alloc_page(tlb);
+	tlb->nr_ptes = 0;
+	tlb->nr_pxds = tlb->max;
 }
 
-static inline void tlb_flush_mmu(struct mmu_gather *tlb,
-				 unsigned long start, unsigned long end)
+static inline void tlb_flush_mmu(struct mmu_gather *tlb)
 {
-	if (!tlb->fullmm && (tlb->nr_ptes > 0 || tlb->nr_pxds < TLB_NR_PTRS))
+	if (!tlb->fullmm && (tlb->nr_ptes > 0 || tlb->nr_pxds < tlb->max))
 		__tlb_flush_mm(tlb->mm);
 	while (tlb->nr_ptes > 0)
 		page_table_free_rcu(tlb->mm, tlb->array[--tlb->nr_ptes]);
-	while (tlb->nr_pxds < TLB_NR_PTRS)
+	while (tlb->nr_pxds < tlb->max)
 		crst_table_free_rcu(tlb->mm, tlb->array[tlb->nr_pxds++]);
 }
 
 static inline void tlb_finish_mmu(struct mmu_gather *tlb,
 				  unsigned long start, unsigned long end)
 {
-	tlb_flush_mmu(tlb, start, end);
+	tlb_flush_mmu(tlb);
 
 	rcu_table_freelist_finish();
 
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
 
-	put_cpu_var(mmu_gathers);
+	if (tlb->array != tlb->local)
+		free_pages((unsigned long) tlb->array, 0);
 }
 
 /*
  * Release the page cache reference for a pte removed by
- * tlb_ptep_clear_flush. In both flush modes the tlb fo a page cache page
+ * tlb_ptep_clear_flush. In both flush modes the tlb for a page cache page
  * has already been freed, so just do free_page_and_swap_cache.
  */
+static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	free_page_and_swap_cache(page);
+	return 1; /* avoid calling tlb_flush_mmu */
+}
+
 static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
 {
 	free_page_and_swap_cache(page);
@@ -103,7 +115,7 @@ static inline void pte_free_tlb(struct mmu_gather *tlb, pgtable_t pte,
 	if (!tlb->fullmm) {
 		tlb->array[tlb->nr_ptes++] = pte;
 		if (tlb->nr_ptes >= tlb->nr_pxds)
-			tlb_flush_mmu(tlb, 0, 0);
+			tlb_flush_mmu(tlb);
 	} else
 		page_table_free(tlb->mm, (unsigned long *) pte);
 }
@@ -124,7 +136,7 @@ static inline void pmd_free_tlb(struct mmu_gather *tlb, pmd_t *pmd,
 	if (!tlb->fullmm) {
 		tlb->array[--tlb->nr_pxds] = pmd;
 		if (tlb->nr_ptes >= tlb->nr_pxds)
-			tlb_flush_mmu(tlb, 0, 0);
+			tlb_flush_mmu(tlb);
 	} else
 		crst_table_free(tlb->mm, (unsigned long *) pmd);
 #endif
@@ -146,7 +158,7 @@ static inline void pud_free_tlb(struct mmu_gather *tlb, pud_t *pud,
 	if (!tlb->fullmm) {
 		tlb->array[--tlb->nr_pxds] = pud;
 		if (tlb->nr_ptes >= tlb->nr_pxds)
-			tlb_flush_mmu(tlb, 0, 0);
+			tlb_flush_mmu(tlb);
 	} else
 		crst_table_free(tlb->mm, (unsigned long *) pud);
 #endif
diff --git a/arch/s390/mm/pgtable.c b/arch/s390/mm/pgtable.c
index e1850c2..07fcc3f 100644
--- a/arch/s390/mm/pgtable.c
+++ b/arch/s390/mm/pgtable.c
@@ -36,7 +36,6 @@ struct rcu_table_freelist {
 	((PAGE_SIZE - sizeof(struct rcu_table_freelist)) \
 	  / sizeof(unsigned long))
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
 static DEFINE_PER_CPU(struct rcu_table_freelist *, rcu_table_freelist);
 
 static void __page_table_free(struct mm_struct *mm, unsigned long *table);
diff --git a/arch/score/mm/init.c b/arch/score/mm/init.c
index 50fdec5..cee6bce 100644
--- a/arch/score/mm/init.c
+++ b/arch/score/mm/init.c
@@ -38,8 +38,6 @@
 #include <asm/sections.h>
 #include <asm/tlb.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 unsigned long empty_zero_page;
 EXPORT_SYMBOL_GPL(empty_zero_page);
 
diff --git a/arch/sh/include/asm/tlb.h b/arch/sh/include/asm/tlb.h
index 75abb38..6c308d8 100644
--- a/arch/sh/include/asm/tlb.h
+++ b/arch/sh/include/asm/tlb.h
@@ -23,8 +23,6 @@ struct mmu_gather {
 	unsigned long		start, end;
 };
 
-DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 static inline void init_tlb_gather(struct mmu_gather *tlb)
 {
 	tlb->start = TASK_SIZE;
@@ -36,17 +34,13 @@ static inline void init_tlb_gather(struct mmu_gather *tlb)
 	}
 }
 
-static inline struct mmu_gather *
-tlb_gather_mmu(struct mm_struct *mm, unsigned int full_mm_flush)
+static inline void
+tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned int full_mm_flush)
 {
-	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);
-
 	tlb->mm = mm;
 	tlb->fullmm = full_mm_flush;
 
 	init_tlb_gather(tlb);
-
-	return tlb;
 }
 
 static inline void
@@ -57,8 +51,6 @@ tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
 
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
-
-	put_cpu_var(mmu_gathers);
 }
 
 static inline void
@@ -91,7 +83,21 @@ tlb_end_vma(struct mmu_gather *tlb, struct vm_area_struct *vma)
 	}
 }
 
-#define tlb_remove_page(tlb,page)	free_page_and_swap_cache(page)
+static inline void tlb_flush_mmu(struct mmu_gather *tlb)
+{
+}
+
+static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	free_page_and_swap_cache(page);
+	return 1; /* avoid calling tlb_flush_mmu */
+}
+
+static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	__tlb_remove_page(tlb, page);
+}
+
 #define pte_free_tlb(tlb, ptep, addr)	pte_free((tlb)->mm, ptep)
 #define pmd_free_tlb(tlb, pmdp, addr)	pmd_free((tlb)->mm, pmdp)
 #define pud_free_tlb(tlb, pudp, addr)	pud_free((tlb)->mm, pudp)
diff --git a/arch/sh/mm/init.c b/arch/sh/mm/init.c
index 0d3f912..58a93fb3 100644
--- a/arch/sh/mm/init.c
+++ b/arch/sh/mm/init.c
@@ -28,7 +28,6 @@
 #include <asm/cache.h>
 #include <asm/sizes.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
 pgd_t swapper_pg_dir[PTRS_PER_PGD];
 
 void __init generic_mem_init(void)
diff --git a/arch/sparc/include/asm/pgalloc_64.h b/arch/sparc/include/asm/pgalloc_64.h
index 5bdfa2c..4e5e087 100644
--- a/arch/sparc/include/asm/pgalloc_64.h
+++ b/arch/sparc/include/asm/pgalloc_64.h
@@ -78,4 +78,7 @@ static inline void check_pgt_cache(void)
 	quicklist_trim(0, NULL, 25, 16);
 }
 
+#define __pte_free_tlb(tlb, pte, addr)	pte_free((tlb)->mm, pte)
+#define __pmd_free_tlb(tlb, pmd, addr)	pmd_free((tlb)->mm, pmd)
+
 #endif /* _SPARC64_PGALLOC_H */
diff --git a/arch/sparc/include/asm/pgtable_64.h b/arch/sparc/include/asm/pgtable_64.h
index f8dddb7..b2e85bf 100644
--- a/arch/sparc/include/asm/pgtable_64.h
+++ b/arch/sparc/include/asm/pgtable_64.h
@@ -655,9 +655,11 @@ static inline int pte_special(pte_t pte)
 #define pte_unmap(pte)			do { } while (0)
 
 /* Actual page table PTE updates.  */
-extern void tlb_batch_add(struct mm_struct *mm, unsigned long vaddr, pte_t *ptep, pte_t orig);
+extern void tlb_batch_add(struct mm_struct *mm, unsigned long vaddr,
+			  pte_t *ptep, pte_t orig, int fullmm);
 
-static inline void set_pte_at(struct mm_struct *mm, unsigned long addr, pte_t *ptep, pte_t pte)
+static inline void __set_pte_at(struct mm_struct *mm, unsigned long addr,
+			     pte_t *ptep, pte_t pte, int fullmm)
 {
 	pte_t orig = *ptep;
 
@@ -670,12 +672,19 @@ static inline void set_pte_at(struct mm_struct *mm, unsigned long addr, pte_t *p
 	 *             and SUN4V pte layout, so this inline test is fine.
 	 */
 	if (likely(mm != &init_mm) && (pte_val(orig) & _PAGE_VALID))
-		tlb_batch_add(mm, addr, ptep, orig);
+		tlb_batch_add(mm, addr, ptep, orig, fullmm);
 }
 
+#define set_pte_at(mm,addr,ptep,pte)	\
+	__set_pte_at((mm), (addr), (ptep), (pte), 0)
+
 #define pte_clear(mm,addr,ptep)		\
 	set_pte_at((mm), (addr), (ptep), __pte(0UL))
 
+#define __HAVE_ARCH_PTE_CLEAR_NOT_PRESENT_FULL
+#define pte_clear_not_present_full(mm,addr,ptep,fullmm)	\
+	__set_pte_at((mm), (addr), (ptep), __pte(0UL), (fullmm))
+
 #ifdef DCACHE_ALIASING_POSSIBLE
 #define __HAVE_ARCH_MOVE_PTE
 #define move_pte(pte, prot, old_addr, new_addr)				\
diff --git a/arch/sparc/include/asm/tlb_64.h b/arch/sparc/include/asm/tlb_64.h
index dca406b..190e189 100644
--- a/arch/sparc/include/asm/tlb_64.h
+++ b/arch/sparc/include/asm/tlb_64.h
@@ -7,66 +7,11 @@
 #include <asm/tlbflush.h>
 #include <asm/mmu_context.h>
 
-#define TLB_BATCH_NR	192
-
-/*
- * For UP we don't need to worry about TLB flush
- * and page free order so much..
- */
-#ifdef CONFIG_SMP
-  #define FREE_PTE_NR	506
-  #define tlb_fast_mode(bp) ((bp)->pages_nr == ~0U)
-#else
-  #define FREE_PTE_NR	1
-  #define tlb_fast_mode(bp) 1
-#endif
-
-struct mmu_gather {
-	struct mm_struct *mm;
-	unsigned int pages_nr;
-	unsigned int need_flush;
-	unsigned int fullmm;
-	unsigned int tlb_nr;
-	unsigned long vaddrs[TLB_BATCH_NR];
-	struct page *pages[FREE_PTE_NR];
-};
-
-DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 #ifdef CONFIG_SMP
 extern void smp_flush_tlb_pending(struct mm_struct *,
 				  unsigned long, unsigned long *);
 #endif
 
-extern void __flush_tlb_pending(unsigned long, unsigned long, unsigned long *);
-extern void flush_tlb_pending(void);
-
-static inline struct mmu_gather *tlb_gather_mmu(struct mm_struct *mm, unsigned int full_mm_flush)
-{
-	struct mmu_gather *mp = &get_cpu_var(mmu_gathers);
-
-	BUG_ON(mp->tlb_nr);
-
-	mp->mm = mm;
-	mp->pages_nr = num_online_cpus() > 1 ? 0U : ~0U;
-	mp->fullmm = full_mm_flush;
-
-	return mp;
-}
-
-
-static inline void tlb_flush_mmu(struct mmu_gather *mp)
-{
-	if (!mp->fullmm)
-		flush_tlb_pending();
-	if (mp->need_flush) {
-		free_pages_and_swap_cache(mp->pages, mp->pages_nr);
-		mp->pages_nr = 0;
-		mp->need_flush = 0;
-	}
-
-}
-
 #ifdef CONFIG_SMP
 extern void smp_flush_tlb_mm(struct mm_struct *mm);
 #define do_flush_tlb_mm(mm) smp_flush_tlb_mm(mm)
@@ -74,38 +19,14 @@ extern void smp_flush_tlb_mm(struct mm_struct *mm);
 #define do_flush_tlb_mm(mm) __flush_tlb_mm(CTX_HWBITS(mm->context), SECONDARY_CONTEXT)
 #endif
 
-static inline void tlb_finish_mmu(struct mmu_gather *mp, unsigned long start, unsigned long end)
-{
-	tlb_flush_mmu(mp);
-
-	if (mp->fullmm)
-		mp->fullmm = 0;
-
-	/* keep the page table cache within bounds */
-	check_pgt_cache();
-
-	put_cpu_var(mmu_gathers);
-}
-
-static inline void tlb_remove_page(struct mmu_gather *mp, struct page *page)
-{
-	if (tlb_fast_mode(mp)) {
-		free_page_and_swap_cache(page);
-		return;
-	}
-	mp->need_flush = 1;
-	mp->pages[mp->pages_nr++] = page;
-	if (mp->pages_nr >= FREE_PTE_NR)
-		tlb_flush_mmu(mp);
-}
-
-#define tlb_remove_tlb_entry(mp,ptep,addr) do { } while (0)
-#define pte_free_tlb(mp, ptepage, addr) pte_free((mp)->mm, ptepage)
-#define pmd_free_tlb(mp, pmdp, addr) pmd_free((mp)->mm, pmdp)
-#define pud_free_tlb(tlb,pudp, addr) __pud_free_tlb(tlb,pudp,addr)
+extern void __flush_tlb_pending(unsigned long, unsigned long, unsigned long *);
+extern void flush_tlb_pending(void);
 
-#define tlb_migrate_finish(mm)	do { } while (0)
 #define tlb_start_vma(tlb, vma) do { } while (0)
 #define tlb_end_vma(tlb, vma)	do { } while (0)
+#define __tlb_remove_tlb_entry(tlb, ptep, address) do { } while (0)
+#define tlb_flush(tlb)	flush_tlb_pending()
+
+#include <asm-generic/tlb.h>
 
 #endif /* _SPARC64_TLB_H */
diff --git a/arch/sparc/include/asm/tlbflush_64.h b/arch/sparc/include/asm/tlbflush_64.h
index fbb675d..2ef4634 100644
--- a/arch/sparc/include/asm/tlbflush_64.h
+++ b/arch/sparc/include/asm/tlbflush_64.h
@@ -5,9 +5,17 @@
 #include <asm/mmu_context.h>
 
 /* TSB flush operations. */
-struct mmu_gather;
+
+#define TLB_BATCH_NR	192
+
+struct tlb_batch {
+	struct mm_struct *mm;
+	unsigned long tlb_nr;
+	unsigned long vaddrs[TLB_BATCH_NR];
+};
+
 extern void flush_tsb_kernel_range(unsigned long start, unsigned long end);
-extern void flush_tsb_user(struct mmu_gather *mp);
+extern void flush_tsb_user(struct tlb_batch *tb);
 
 /* TLB flush operations. */
 
diff --git a/arch/sparc/mm/init_32.c b/arch/sparc/mm/init_32.c
index 4c31e2b..a755487 100644
--- a/arch/sparc/mm/init_32.c
+++ b/arch/sparc/mm/init_32.c
@@ -37,8 +37,6 @@
 #include <asm/prom.h>
 #include <asm/leon.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 unsigned long *sparc_valid_addr_bitmap;
 EXPORT_SYMBOL(sparc_valid_addr_bitmap);
 
diff --git a/arch/sparc/mm/tlb.c b/arch/sparc/mm/tlb.c
index d8f21e2..b1f279c 100644
--- a/arch/sparc/mm/tlb.c
+++ b/arch/sparc/mm/tlb.c
@@ -19,33 +19,34 @@
 
 /* Heavily inspired by the ppc64 code.  */
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
+static DEFINE_PER_CPU(struct tlb_batch, tlb_batch);
 
 void flush_tlb_pending(void)
 {
-	struct mmu_gather *mp = &get_cpu_var(mmu_gathers);
+	struct tlb_batch *tb = &get_cpu_var(tlb_batch);
 
-	if (mp->tlb_nr) {
-		flush_tsb_user(mp);
+	if (tb->tlb_nr) {
+		flush_tsb_user(tb);
 
-		if (CTX_VALID(mp->mm->context)) {
+		if (CTX_VALID(tb->mm->context)) {
 #ifdef CONFIG_SMP
-			smp_flush_tlb_pending(mp->mm, mp->tlb_nr,
-					      &mp->vaddrs[0]);
+			smp_flush_tlb_pending(tb->mm, tb->tlb_nr,
+					      &tb->vaddrs[0]);
 #else
-			__flush_tlb_pending(CTX_HWBITS(mp->mm->context),
-					    mp->tlb_nr, &mp->vaddrs[0]);
+			__flush_tlb_pending(CTX_HWBITS(tb->mm->context),
+					    tb->tlb_nr, &tb->vaddrs[0]);
 #endif
 		}
-		mp->tlb_nr = 0;
+		tb->tlb_nr = 0;
 	}
 
-	put_cpu_var(mmu_gathers);
+	put_cpu_var(tlb_batch);
 }
 
-void tlb_batch_add(struct mm_struct *mm, unsigned long vaddr, pte_t *ptep, pte_t orig)
+void tlb_batch_add(struct mm_struct *mm, unsigned long vaddr,
+		   pte_t *ptep, pte_t orig, int fullmm)
 {
-	struct mmu_gather *mp = &__get_cpu_var(mmu_gathers);
+	struct tlb_batch *tb = &get_cpu_var(tlb_batch);
 	unsigned long nr;
 
 	vaddr &= PAGE_MASK;
@@ -77,21 +78,25 @@ void tlb_batch_add(struct mm_struct *mm, unsigned long vaddr, pte_t *ptep, pte_t
 
 no_cache_flush:
 
-	if (mp->fullmm)
+	if (fullmm) {
+		put_cpu_var(tlb_batch);
 		return;
+	}
 
-	nr = mp->tlb_nr;
+	nr = tb->tlb_nr;
 
-	if (unlikely(nr != 0 && mm != mp->mm)) {
+	if (unlikely(nr != 0 && mm != tb->mm)) {
 		flush_tlb_pending();
 		nr = 0;
 	}
 
 	if (nr == 0)
-		mp->mm = mm;
+		tb->mm = mm;
 
-	mp->vaddrs[nr] = vaddr;
-	mp->tlb_nr = ++nr;
+	tb->vaddrs[nr] = vaddr;
+	tb->tlb_nr = ++nr;
 	if (nr >= TLB_BATCH_NR)
 		flush_tlb_pending();
+
+	put_cpu_var(tlb_batch);
 }
diff --git a/arch/sparc/mm/tsb.c b/arch/sparc/mm/tsb.c
index 101d7c8..9484615 100644
--- a/arch/sparc/mm/tsb.c
+++ b/arch/sparc/mm/tsb.c
@@ -47,12 +47,13 @@ void flush_tsb_kernel_range(unsigned long start, unsigned long end)
 	}
 }
 
-static void __flush_tsb_one(struct mmu_gather *mp, unsigned long hash_shift, unsigned long tsb, unsigned long nentries)
+static void __flush_tsb_one(struct tlb_batch *tb, unsigned long hash_shift,
+			    unsigned long tsb, unsigned long nentries)
 {
 	unsigned long i;
 
-	for (i = 0; i < mp->tlb_nr; i++) {
-		unsigned long v = mp->vaddrs[i];
+	for (i = 0; i < tb->tlb_nr; i++) {
+		unsigned long v = tb->vaddrs[i];
 		unsigned long tag, ent, hash;
 
 		v &= ~0x1UL;
@@ -65,9 +66,9 @@ static void __flush_tsb_one(struct mmu_gather *mp, unsigned long hash_shift, uns
 	}
 }
 
-void flush_tsb_user(struct mmu_gather *mp)
+void flush_tsb_user(struct tlb_batch *tb)
 {
-	struct mm_struct *mm = mp->mm;
+	struct mm_struct *mm = tb->mm;
 	unsigned long nentries, base, flags;
 
 	spin_lock_irqsave(&mm->context.lock, flags);
@@ -76,7 +77,7 @@ void flush_tsb_user(struct mmu_gather *mp)
 	nentries = mm->context.tsb_block[MM_TSB_BASE].tsb_nentries;
 	if (tlb_type == cheetah_plus || tlb_type == hypervisor)
 		base = __pa(base);
-	__flush_tsb_one(mp, PAGE_SHIFT, base, nentries);
+	__flush_tsb_one(tb, PAGE_SHIFT, base, nentries);
 
 #ifdef CONFIG_HUGETLB_PAGE
 	if (mm->context.tsb_block[MM_TSB_HUGE].tsb) {
@@ -84,7 +85,7 @@ void flush_tsb_user(struct mmu_gather *mp)
 		nentries = mm->context.tsb_block[MM_TSB_HUGE].tsb_nentries;
 		if (tlb_type == cheetah_plus || tlb_type == hypervisor)
 			base = __pa(base);
-		__flush_tsb_one(mp, HPAGE_SHIFT, base, nentries);
+		__flush_tsb_one(tb, HPAGE_SHIFT, base, nentries);
 	}
 #endif
 	spin_unlock_irqrestore(&mm->context.lock, flags);
diff --git a/arch/tile/mm/init.c b/arch/tile/mm/init.c
index d6e87fd..4e10c40 100644
--- a/arch/tile/mm/init.c
+++ b/arch/tile/mm/init.c
@@ -60,8 +60,6 @@ unsigned long VMALLOC_RESERVE = CONFIG_VMALLOC_RESERVE;
 EXPORT_SYMBOL(VMALLOC_RESERVE);
 #endif
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /* Create an L2 page table */
 static pte_t * __init alloc_pte(void)
 {
diff --git a/arch/um/include/asm/tlb.h b/arch/um/include/asm/tlb.h
index 660caed..4febacd 100644
--- a/arch/um/include/asm/tlb.h
+++ b/arch/um/include/asm/tlb.h
@@ -22,9 +22,6 @@ struct mmu_gather {
 	unsigned int		fullmm; /* non-zero means full mm flush */
 };
 
-/* Users of the generic TLB shootdown code must declare this storage space. */
-DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 static inline void __tlb_remove_tlb_entry(struct mmu_gather *tlb, pte_t *ptep,
 					  unsigned long address)
 {
@@ -47,27 +44,20 @@ static inline void init_tlb_gather(struct mmu_gather *tlb)
 	}
 }
 
-/* tlb_gather_mmu
- *	Return a pointer to an initialized struct mmu_gather.
- */
-static inline struct mmu_gather *
-tlb_gather_mmu(struct mm_struct *mm, unsigned int full_mm_flush)
+static inline void
+tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned int full_mm_flush)
 {
-	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);
-
 	tlb->mm = mm;
 	tlb->fullmm = full_mm_flush;
 
 	init_tlb_gather(tlb);
-
-	return tlb;
 }
 
 extern void flush_tlb_mm_range(struct mm_struct *mm, unsigned long start,
 			       unsigned long end);
 
 static inline void
-tlb_flush_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
+tlb_flush_mmu(struct mmu_gather *tlb)
 {
 	if (!tlb->need_flush)
 		return;
@@ -83,12 +73,10 @@ tlb_flush_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
 static inline void
 tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
 {
-	tlb_flush_mmu(tlb, start, end);
+	tlb_flush_mmu(tlb);
 
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
-
-	put_cpu_var(mmu_gathers);
 }
 
 /* tlb_remove_page
@@ -96,11 +84,16 @@ tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
  *	while handling the additional races in SMP caused by other CPUs
  *	caching valid mappings in their TLBs.
  */
-static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
 {
 	tlb->need_flush = 1;
 	free_page_and_swap_cache(page);
-	return;
+	return 1; /* avoid calling tlb_flush_mmu */
+}
+
+static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	__tlb_remove_page(tlb, page);
 }
 
 /**
diff --git a/arch/um/kernel/smp.c b/arch/um/kernel/smp.c
index 106bf27..d9011e0 100644
--- a/arch/um/kernel/smp.c
+++ b/arch/um/kernel/smp.c
@@ -7,9 +7,6 @@
 #include "asm/pgalloc.h"
 #include "asm/tlb.h"
 
-/* For some reason, mmu_gathers are referenced when CONFIG_SMP is off. */
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 #ifdef CONFIG_SMP
 
 #include "linux/sched.h"
diff --git a/arch/unicore32/mm/mmu.c b/arch/unicore32/mm/mmu.c
index 7bf3d58..c5b2b65 100644
--- a/arch/unicore32/mm/mmu.c
+++ b/arch/unicore32/mm/mmu.c
@@ -30,8 +30,6 @@
 
 #include "mm.h"
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /*
  * empty_zero_page is a special page that is used for
  * zero-initialized data and COW.
diff --git a/arch/x86/mm/init.c b/arch/x86/mm/init.c
index 286d289..cda082e 100644
--- a/arch/x86/mm/init.c
+++ b/arch/x86/mm/init.c
@@ -16,8 +16,6 @@
 #include <asm/tlb.h>
 #include <asm/proto.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 unsigned long __initdata pgt_buf_start;
 unsigned long __meminitdata pgt_buf_end;
 unsigned long __meminitdata pgt_buf_top;
diff --git a/arch/xtensa/mm/mmu.c b/arch/xtensa/mm/mmu.c
index 4bb91a9..ca81654 100644
--- a/arch/xtensa/mm/mmu.c
+++ b/arch/xtensa/mm/mmu.c
@@ -14,8 +14,6 @@
 #include <asm/mmu_context.h>
 #include <asm/page.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 void __init paging_init(void)
 {
 	memset(swapper_pg_dir, 0, PAGE_SIZE);
diff --git a/fs/exec.c b/fs/exec.c
index 5e62d26..14e623e 100644
--- a/fs/exec.c
+++ b/fs/exec.c
@@ -553,7 +553,7 @@ static int shift_arg_pages(struct vm_area_struct *vma, unsigned long shift)
 	unsigned long length = old_end - old_start;
 	unsigned long new_start = old_start - shift;
 	unsigned long new_end = old_end - shift;
-	struct mmu_gather *tlb;
+	struct mmu_gather tlb;
 
 	BUG_ON(new_start > new_end);
 
@@ -579,12 +579,12 @@ static int shift_arg_pages(struct vm_area_struct *vma, unsigned long shift)
 		return -ENOMEM;
 
 	lru_add_drain();
-	tlb = tlb_gather_mmu(mm, 0);
+	tlb_gather_mmu(&tlb, mm, 0);
 	if (new_end > old_start) {
 		/*
 		 * when the old and new regions overlap clear from new_end.
 		 */
-		free_pgd_range(tlb, new_end, old_end, new_end,
+		free_pgd_range(&tlb, new_end, old_end, new_end,
 			vma->vm_next ? vma->vm_next->vm_start : 0);
 	} else {
 		/*
@@ -593,10 +593,10 @@ static int shift_arg_pages(struct vm_area_struct *vma, unsigned long shift)
 		 * have constraints on va-space that make this illegal (IA64) -
 		 * for the others its just a little faster.
 		 */
-		free_pgd_range(tlb, old_start, old_end, new_end,
+		free_pgd_range(&tlb, old_start, old_end, new_end,
 			vma->vm_next ? vma->vm_next->vm_start : 0);
 	}
-	tlb_finish_mmu(tlb, new_end, old_end);
+	tlb_finish_mmu(&tlb, new_end, old_end);
 
 	/*
 	 * Shrink the vma to just the new range.  Always succeeds.
diff --git a/include/asm-generic/tlb.h b/include/asm-generic/tlb.h
index e43f976..67f21e2 100644
--- a/include/asm-generic/tlb.h
+++ b/include/asm-generic/tlb.h
@@ -5,6 +5,8 @@
  * Copyright 2001 Red Hat, Inc.
  * Based on code from mm/memory.c Copyright Linus Torvalds and others.
  *
+ * Copyright 2011 Red Hat, Inc., Peter Zijlstra <pzijlstr@redhat.com>
+ *
  * This program is free software; you can redistribute it and/or
  * modify it under the terms of the GNU General Public License
  * as published by the Free Software Foundation; either version
@@ -22,51 +24,71 @@
  * and page free order so much..
  */
 #ifdef CONFIG_SMP
-  #ifdef ARCH_FREE_PTR_NR
-    #define FREE_PTR_NR   ARCH_FREE_PTR_NR
-  #else
-    #define FREE_PTE_NR	506
-  #endif
   #define tlb_fast_mode(tlb) ((tlb)->nr == ~0U)
 #else
-  #define FREE_PTE_NR	1
   #define tlb_fast_mode(tlb) 1
 #endif
 
+/*
+ * If we can't allocate a page to make a big patch of page pointers
+ * to work on, then just handle a few from the on-stack structure.
+ */
+#define MMU_GATHER_BUNDLE	8
+
 /* struct mmu_gather is an opaque type used by the mm code for passing around
  * any data needed by arch specific code for tlb_remove_page.
  */
 struct mmu_gather {
 	struct mm_struct	*mm;
 	unsigned int		nr;	/* set to ~0U means fast mode */
+	unsigned int		max;	/* nr < max */
 	unsigned int		need_flush;/* Really unmapped some ptes? */
 	unsigned int		fullmm; /* non-zero means full mm flush */
-	struct page *		pages[FREE_PTE_NR];
+#ifdef HAVE_ARCH_MMU_GATHER
+	struct arch_mmu_gather	arch;
+#endif
+	struct page		**pages;
+	struct page		*local[MMU_GATHER_BUNDLE];
 };
 
-/* Users of the generic TLB shootdown code must declare this storage space. */
-DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
+static inline void __tlb_alloc_page(struct mmu_gather *tlb)
+{
+	unsigned long addr = __get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
+
+	if (addr) {
+		tlb->pages = (void *)addr;
+		tlb->max = PAGE_SIZE / sizeof(struct page *);
+	}
+}
 
 /* tlb_gather_mmu
- *	Return a pointer to an initialized struct mmu_gather.
+ *	Called to initialize an (on-stack) mmu_gather structure for page-table
+ *	tear-down from @mm. The @fullmm argument is used when @mm is without
+ *	users and we're going to destroy the full address space (exit/execve).
  */
-static inline struct mmu_gather *
-tlb_gather_mmu(struct mm_struct *mm, unsigned int full_mm_flush)
+static inline void
+tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned int fullmm)
 {
-	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);
-
 	tlb->mm = mm;
 
-	/* Use fast mode if only one CPU is online */
-	tlb->nr = num_online_cpus() > 1 ? 0U : ~0U;
+	tlb->max = ARRAY_SIZE(tlb->local);
+	tlb->pages = tlb->local;
+
+	if (num_online_cpus() > 1) {
+		tlb->nr = 0;
+		__tlb_alloc_page(tlb);
+	} else /* Use fast mode if only one CPU is online */
+		tlb->nr = ~0U;
 
-	tlb->fullmm = full_mm_flush;
+	tlb->fullmm = fullmm;
 
-	return tlb;
+#ifdef HAVE_ARCH_MMU_GATHER
+	tlb->arch = ARCH_MMU_GATHER_INIT;
+#endif
 }
 
 static inline void
-tlb_flush_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
+tlb_flush_mmu(struct mmu_gather *tlb)
 {
 	if (!tlb->need_flush)
 		return;
@@ -75,6 +97,13 @@ tlb_flush_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
 	if (!tlb_fast_mode(tlb)) {
 		free_pages_and_swap_cache(tlb->pages, tlb->nr);
 		tlb->nr = 0;
+		/*
+		 * If we are using the local on-stack array of pages for MMU
+		 * gather, try allocating an off-stack array again as we have
+		 * recently freed pages.
+		 */
+		if (tlb->pages == tlb->local)
+			__tlb_alloc_page(tlb);
 	}
 }
 
@@ -85,29 +114,42 @@ tlb_flush_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
 static inline void
 tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
 {
-	tlb_flush_mmu(tlb, start, end);
+	tlb_flush_mmu(tlb);
 
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
 
-	put_cpu_var(mmu_gathers);
+	if (tlb->pages != tlb->local)
+		free_pages((unsigned long)tlb->pages, 0);
 }
 
-/* tlb_remove_page
+/* __tlb_remove_page
  *	Must perform the equivalent to __free_pte(pte_get_and_clear(ptep)), while
  *	handling the additional races in SMP caused by other CPUs caching valid
- *	mappings in their TLBs.
+ *	mappings in their TLBs. Returns the number of free page slots left.
+ *	When out of page slots we must call tlb_flush_mmu().
  */
-static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
 {
 	tlb->need_flush = 1;
 	if (tlb_fast_mode(tlb)) {
 		free_page_and_swap_cache(page);
-		return;
+		return 1; /* avoid calling tlb_flush_mmu() */
 	}
 	tlb->pages[tlb->nr++] = page;
-	if (tlb->nr >= FREE_PTE_NR)
-		tlb_flush_mmu(tlb, 0, 0);
+	VM_BUG_ON(tlb->nr > tlb->max);
+
+	return tlb->max - tlb->nr;
+}
+
+/* tlb_remove_page
+ *	Similar to __tlb_remove_page but will call tlb_flush_mmu() itself when
+ *	required.
+ */
+static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	if (!__tlb_remove_page(tlb, page))
+		tlb_flush_mmu(tlb);
 }
 
 /**
diff --git a/include/linux/mm.h b/include/linux/mm.h
index 7606d7d..2bf8bf1 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -904,7 +904,7 @@ int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
 		unsigned long size);
 unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
 		unsigned long size, struct zap_details *);
-unsigned long unmap_vmas(struct mmu_gather **tlb,
+unsigned long unmap_vmas(struct mmu_gather *tlb,
 		struct vm_area_struct *start_vma, unsigned long start_addr,
 		unsigned long end_addr, unsigned long *nr_accounted,
 		struct zap_details *);
diff --git a/mm/memory.c b/mm/memory.c
index 9da8cab..e6eddc4 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -912,12 +912,13 @@ static unsigned long zap_pte_range(struct mmu_gather *tlb,
 				long *zap_work, struct zap_details *details)
 {
 	struct mm_struct *mm = tlb->mm;
+	int force_flush = 0;
 	pte_t *pte;
 	spinlock_t *ptl;
 	int rss[NR_MM_COUNTERS];
 
 	init_rss_vec(rss);
-
+again:
 	pte = pte_offset_map_lock(mm, pmd, addr, &ptl);
 	arch_enter_lazy_mmu_mode();
 	do {
@@ -974,7 +975,9 @@ static unsigned long zap_pte_range(struct mmu_gather *tlb,
 			page_remove_rmap(page);
 			if (unlikely(page_mapcount(page) < 0))
 				print_bad_pte(vma, addr, ptent, page);
-			tlb_remove_page(tlb, page);
+			force_flush = !__tlb_remove_page(tlb, page);
+			if (force_flush)
+				break;
 			continue;
 		}
 		/*
@@ -1001,6 +1004,18 @@ static unsigned long zap_pte_range(struct mmu_gather *tlb,
 	arch_leave_lazy_mmu_mode();
 	pte_unmap_unlock(pte - 1, ptl);
 
+	/*
+	 * mmu_gather ran out of room to batch pages, we break out of
+	 * the PTE lock to avoid doing the potential expensive TLB invalidate
+	 * and page-free while holding it.
+	 */
+	if (force_flush) {
+		force_flush = 0;
+		tlb_flush_mmu(tlb);
+		if (addr != end)
+			goto again;
+	}
+
 	return addr;
 }
 
@@ -1121,17 +1136,14 @@ static unsigned long unmap_page_range(struct mmu_gather *tlb,
  * ensure that any thus-far unmapped pages are flushed before unmap_vmas()
  * drops the lock and schedules.
  */
-unsigned long unmap_vmas(struct mmu_gather **tlbp,
+unsigned long unmap_vmas(struct mmu_gather *tlb,
 		struct vm_area_struct *vma, unsigned long start_addr,
 		unsigned long end_addr, unsigned long *nr_accounted,
 		struct zap_details *details)
 {
 	long zap_work = ZAP_BLOCK_SIZE;
-	unsigned long tlb_start = 0;	/* For tlb_finish_mmu */
-	int tlb_start_valid = 0;
 	unsigned long start = start_addr;
 	spinlock_t *i_mmap_lock = details? details->i_mmap_lock: NULL;
-	int fullmm = (*tlbp)->fullmm;
 	struct mm_struct *mm = vma->vm_mm;
 
 	mmu_notifier_invalidate_range_start(mm, start_addr, end_addr);
@@ -1152,11 +1164,6 @@ unsigned long unmap_vmas(struct mmu_gather **tlbp,
 			untrack_pfn_vma(vma, 0, 0);
 
 		while (start != end) {
-			if (!tlb_start_valid) {
-				tlb_start = start;
-				tlb_start_valid = 1;
-			}
-
 			if (unlikely(is_vm_hugetlb_page(vma))) {
 				/*
 				 * It is undesirable to test vma->vm_file as it
@@ -1177,7 +1184,7 @@ unsigned long unmap_vmas(struct mmu_gather **tlbp,
 
 				start = end;
 			} else
-				start = unmap_page_range(*tlbp, vma,
+				start = unmap_page_range(tlb, vma,
 						start, end, &zap_work, details);
 
 			if (zap_work > 0) {
@@ -1185,19 +1192,13 @@ unsigned long unmap_vmas(struct mmu_gather **tlbp,
 				break;
 			}
 
-			tlb_finish_mmu(*tlbp, tlb_start, start);
-
 			if (need_resched() ||
 				(i_mmap_lock && spin_needbreak(i_mmap_lock))) {
-				if (i_mmap_lock) {
-					*tlbp = NULL;
+				if (i_mmap_lock)
 					goto out;
-				}
 				cond_resched();
 			}
 
-			*tlbp = tlb_gather_mmu(vma->vm_mm, fullmm);
-			tlb_start_valid = 0;
 			zap_work = ZAP_BLOCK_SIZE;
 		}
 	}
@@ -1217,16 +1218,15 @@ unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
 		unsigned long size, struct zap_details *details)
 {
 	struct mm_struct *mm = vma->vm_mm;
-	struct mmu_gather *tlb;
+	struct mmu_gather tlb;
 	unsigned long end = address + size;
 	unsigned long nr_accounted = 0;
 
 	lru_add_drain();
-	tlb = tlb_gather_mmu(mm, 0);
+	tlb_gather_mmu(&tlb, mm, 0);
 	update_hiwater_rss(mm);
 	end = unmap_vmas(&tlb, vma, address, end, &nr_accounted, details);
-	if (tlb)
-		tlb_finish_mmu(tlb, address, end);
+	tlb_finish_mmu(&tlb, address, end);
 	return end;
 }
 
diff --git a/mm/mmap.c b/mm/mmap.c
index 2ec8eb5..f8cbc86 100644
--- a/mm/mmap.c
+++ b/mm/mmap.c
@@ -1913,17 +1913,17 @@ static void unmap_region(struct mm_struct *mm,
 		unsigned long start, unsigned long end)
 {
 	struct vm_area_struct *next = prev? prev->vm_next: mm->mmap;
-	struct mmu_gather *tlb;
+	struct mmu_gather tlb;
 	unsigned long nr_accounted = 0;
 
 	lru_add_drain();
-	tlb = tlb_gather_mmu(mm, 0);
+	tlb_gather_mmu(&tlb, mm, 0);
 	update_hiwater_rss(mm);
 	unmap_vmas(&tlb, vma, start, end, &nr_accounted, NULL);
 	vm_unacct_memory(nr_accounted);
-	free_pgtables(tlb, vma, prev? prev->vm_end: FIRST_USER_ADDRESS,
-				 next? next->vm_start: 0);
-	tlb_finish_mmu(tlb, start, end);
+	free_pgtables(&tlb, vma, prev ? prev->vm_end : FIRST_USER_ADDRESS,
+				 next ? next->vm_start : 0);
+	tlb_finish_mmu(&tlb, start, end);
 }
 
 /*
@@ -2265,7 +2265,7 @@ EXPORT_SYMBOL(do_brk);
 /* Release all mmaps. */
 void exit_mmap(struct mm_struct *mm)
 {
-	struct mmu_gather *tlb;
+	struct mmu_gather tlb;
 	struct vm_area_struct *vma;
 	unsigned long nr_accounted = 0;
 	unsigned long end;
@@ -2290,14 +2290,14 @@ void exit_mmap(struct mm_struct *mm)
 
 	lru_add_drain();
 	flush_cache_mm(mm);
-	tlb = tlb_gather_mmu(mm, 1);
+	tlb_gather_mmu(&tlb, mm, 1);
 	/* update_hiwater_rss(mm) here? but nobody should be looking */
 	/* Use -1 here to ensure all VMAs in the mm are unmapped */
 	end = unmap_vmas(&tlb, vma, 0, -1, &nr_accounted, NULL);
 	vm_unacct_memory(nr_accounted);
 
-	free_pgtables(tlb, vma, FIRST_USER_ADDRESS, 0);
-	tlb_finish_mmu(tlb, 0, end);
+	free_pgtables(&tlb, vma, FIRST_USER_ADDRESS, 0);
+	tlb_finish_mmu(&tlb, 0, end);
 
 	/*
 	 * Walk the list again, actually closing and freeing it,

Re: [PATCH 00/20] mm: Preemptibility -v10

[Peter Zijlstra]

On Fri, 2011-04-01 at 14:12 +0200, Peter Zijlstra wrote:
> 
> Also provided is a rollup of these patches, which is used as a commit in the
> git tree referenced below. 

If only I'd actually added it ;-)

---
Subject: mm: mmu_gather rework
From: Peter Zijlstra <a.p.zijlstra@chello.nl>
Date: Fri, 26 Nov 2010 15:38:51 +0100

Remove the first obstackle towards a fully preemptible mmu_gather.

The current scheme assumes mmu_gather is always done with preemption
disabled and uses per-cpu storage for the page batches. Change this to
try and allocate a page for batching and in case of failure, use a
small on-stack array to make some progress.

Preemptible mmu_gather is desired in general and usable once
i_mmap_lock becomes a mutex. Doing it before the mutex conversion
saves us from having to rework the code by moving the mmu_gather
bits inside the pte_lock.

Also avoid flushing the tlb batches from under the pte lock,
this is useful even without the i_mmap_lock conversion as it
significantly reduces pte lock hold times.

Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: David Miller <davem@davemloft.net>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Russell King <rmk@arm.linux.org.uk>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: Jeff Dike <jdike@addtoit.com>
Acked-by: Tony Luck <tony.luck@intel.com>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Hugh Dickins <hughd@google.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
---
 arch/alpha/mm/init.c                   |    2 -
 arch/arm/include/asm/tlb.h             |   53 ++++++++++++-----
 arch/arm/mm/mmu.c                      |    2 -
 arch/avr32/mm/init.c                   |    2 -
 arch/cris/mm/init.c                    |    2 -
 arch/frv/mm/init.c                     |    2 -
 arch/ia64/include/asm/tlb.h            |   66 +++++++++++++++-------
 arch/ia64/mm/init.c                    |    2 -
 arch/m32r/mm/init.c                    |    2 -
 arch/m68k/mm/init_mm.c                 |    2 -
 arch/microblaze/mm/init.c              |    2 -
 arch/mips/mm/init.c                    |    2 -
 arch/mn10300/mm/init.c                 |    2 -
 arch/parisc/mm/init.c                  |    2 -
 arch/powerpc/include/asm/pgalloc.h     |    4 +-
 arch/powerpc/include/asm/thread_info.h |    2 +
 arch/powerpc/include/asm/tlb.h         |   10 +++
 arch/powerpc/kernel/process.c          |   23 +++++++-
 arch/powerpc/mm/pgtable.c              |   14 +---
 arch/powerpc/mm/tlb_hash32.c           |    2 +-
 arch/powerpc/mm/tlb_hash64.c           |    6 +-
 arch/powerpc/mm/tlb_nohash.c           |    2 +-
 arch/s390/include/asm/tlb.h            |   62 ++++++++++++--------
 arch/s390/mm/pgtable.c                 |    1 -
 arch/score/mm/init.c                   |    2 -
 arch/sh/include/asm/tlb.h              |   28 ++++++----
 arch/sh/mm/init.c                      |    1 -
 arch/sparc/include/asm/pgalloc_64.h    |    3 +
 arch/sparc/include/asm/pgtable_64.h    |   15 ++++-
 arch/sparc/include/asm/tlb_64.h        |   91 ++----------------------------
 arch/sparc/include/asm/tlbflush_64.h   |   12 +++-
 arch/sparc/mm/init_32.c                |    2 -
 arch/sparc/mm/tlb.c                    |   43 ++++++++------
 arch/sparc/mm/tsb.c                    |   15 +++--
 arch/tile/mm/init.c                    |    2 -
 arch/um/include/asm/tlb.h              |   29 ++++------
 arch/um/kernel/smp.c                   |    3 -
 arch/unicore32/mm/mmu.c                |    2 -
 arch/x86/mm/init.c                     |    2 -
 arch/xtensa/mm/mmu.c                   |    2 -
 fs/exec.c                              |   10 ++--
 include/asm-generic/tlb.h              |   96 +++++++++++++++++++++++---------
 include/linux/mm.h                     |    2 +-
 mm/memory.c                            |   46 ++++++++--------
 mm/mmap.c                              |   18 +++---
 45 files changed, 364 insertions(+), 329 deletions(-)

diff --git a/arch/alpha/mm/init.c b/arch/alpha/mm/init.c
index 86425ab..69d0c57 100644
--- a/arch/alpha/mm/init.c
+++ b/arch/alpha/mm/init.c
@@ -32,8 +32,6 @@
 #include <asm/console.h>
 #include <asm/tlb.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 extern void die_if_kernel(char *,struct pt_regs *,long);
 
 static struct pcb_struct original_pcb;
diff --git a/arch/arm/include/asm/tlb.h b/arch/arm/include/asm/tlb.h
index 82dfe5d..265f908 100644
--- a/arch/arm/include/asm/tlb.h
+++ b/arch/arm/include/asm/tlb.h
@@ -41,12 +41,12 @@
  */
 #if defined(CONFIG_SMP) || defined(CONFIG_CPU_32v7)
 #define tlb_fast_mode(tlb)	0
-#define FREE_PTE_NR		500
 #else
 #define tlb_fast_mode(tlb)	1
-#define FREE_PTE_NR		0
 #endif
 
+#define MMU_GATHER_BUNDLE	8
+
 /*
  * TLB handling.  This allows us to remove pages from the page
  * tables, and efficiently handle the TLB issues.
@@ -58,7 +58,9 @@ struct mmu_gather {
 	unsigned long		range_start;
 	unsigned long		range_end;
 	unsigned int		nr;
-	struct page		*pages[FREE_PTE_NR];
+	unsigned int		max;
+	struct page		**pages;
+	struct page		*local[MMU_GATHER_BUNDLE];
 };
 
 DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
@@ -97,26 +99,37 @@ static inline void tlb_add_flush(struct mmu_gather *tlb, unsigned long addr)
 	}
 }
 
+static inline void __tlb_alloc_page(struct mmu_gather *tlb)
+{
+	unsigned long addr = __get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
+
+	if (addr) {
+		tlb->pages = (void *)addr;
+		tlb->max = PAGE_SIZE / sizeof(struct page *);
+	}
+}
+
 static inline void tlb_flush_mmu(struct mmu_gather *tlb)
 {
 	tlb_flush(tlb);
 	if (!tlb_fast_mode(tlb)) {
 		free_pages_and_swap_cache(tlb->pages, tlb->nr);
 		tlb->nr = 0;
+		if (tlb->pages == tlb->local)
+			__tlb_alloc_page(tlb);
 	}
 }
 
-static inline struct mmu_gather *
-tlb_gather_mmu(struct mm_struct *mm, unsigned int full_mm_flush)
+static inline void
+tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, bool fullmm)
 {
-	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);
-
 	tlb->mm = mm;
-	tlb->fullmm = full_mm_flush;
+	tlb->fullmm = fullmm;
 	tlb->vma = NULL;
+	tlb->max = ARRAY_SIZE(tlb->local);
+	tlb->pages = tlb->local;
 	tlb->nr = 0;
-
-	return tlb;
+	__tlb_alloc_page(tlb);
 }
 
 static inline void
@@ -127,7 +140,8 @@ tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
 
-	put_cpu_var(mmu_gathers);
+	if (tlb->pages != tlb->local)
+		free_pages((unsigned long)tlb->pages, 0);
 }
 
 /*
@@ -162,15 +176,22 @@ tlb_end_vma(struct mmu_gather *tlb, struct vm_area_struct *vma)
 		tlb_flush(tlb);
 }
 
-static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
 {
 	if (tlb_fast_mode(tlb)) {
 		free_page_and_swap_cache(page);
-	} else {
-		tlb->pages[tlb->nr++] = page;
-		if (tlb->nr >= FREE_PTE_NR)
-			tlb_flush_mmu(tlb);
+		return 1; /* avoid calling tlb_flush_mmu */
 	}
+
+	tlb->pages[tlb->nr++] = page;
+	VM_BUG_ON(tlb->nr > tlb->max);
+	return tlb->max - tlb->nr;
+}
+
+static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	if (!__tlb_remove_page(tlb, page))
+		tlb_flush_mmu(tlb);
 }
 
 static inline void __pte_free_tlb(struct mmu_gather *tlb, pgtable_t pte,
diff --git a/arch/arm/mm/mmu.c b/arch/arm/mm/mmu.c
index 6cf76b3..08a9236 100644
--- a/arch/arm/mm/mmu.c
+++ b/arch/arm/mm/mmu.c
@@ -31,8 +31,6 @@
 
 #include "mm.h"
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /*
  * empty_zero_page is a special page that is used for
  * zero-initialized data and COW.
diff --git a/arch/avr32/mm/init.c b/arch/avr32/mm/init.c
index a7314d4..2798c2d 100644
--- a/arch/avr32/mm/init.c
+++ b/arch/avr32/mm/init.c
@@ -25,8 +25,6 @@
 #include <asm/setup.h>
 #include <asm/sections.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 pgd_t swapper_pg_dir[PTRS_PER_PGD] __page_aligned_data;
 
 struct page *empty_zero_page;
diff --git a/arch/cris/mm/init.c b/arch/cris/mm/init.c
index df33ab8..d72ab58 100644
--- a/arch/cris/mm/init.c
+++ b/arch/cris/mm/init.c
@@ -13,8 +13,6 @@
 #include <linux/bootmem.h>
 #include <asm/tlb.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 unsigned long empty_zero_page;
 
 extern char _stext, _edata, _etext; /* From linkerscript */
diff --git a/arch/frv/mm/init.c b/arch/frv/mm/init.c
index ed64588..fbe5f0d 100644
--- a/arch/frv/mm/init.c
+++ b/arch/frv/mm/init.c
@@ -41,8 +41,6 @@
 
 #undef DEBUG
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /*
  * BAD_PAGE is the page that is used for page faults when linux
  * is out-of-memory. Older versions of linux just did a
diff --git a/arch/ia64/include/asm/tlb.h b/arch/ia64/include/asm/tlb.h
index 23cce99..c3ffe3e 100644
--- a/arch/ia64/include/asm/tlb.h
+++ b/arch/ia64/include/asm/tlb.h
@@ -47,21 +47,27 @@
 #include <asm/machvec.h>
 
 #ifdef CONFIG_SMP
-# define FREE_PTE_NR		2048
 # define tlb_fast_mode(tlb)	((tlb)->nr == ~0U)
 #else
-# define FREE_PTE_NR		0
 # define tlb_fast_mode(tlb)	(1)
 #endif
 
+/*
+ * If we can't allocate a page to make a big batch of page pointers
+ * to work on, then just handle a few from the on-stack structure.
+ */
+#define	IA64_GATHER_BUNDLE	8
+
 struct mmu_gather {
 	struct mm_struct	*mm;
 	unsigned int		nr;		/* == ~0U => fast mode */
+	unsigned int		max;
 	unsigned char		fullmm;		/* non-zero means full mm flush */
 	unsigned char		need_flush;	/* really unmapped some PTEs? */
 	unsigned long		start_addr;
 	unsigned long		end_addr;
-	struct page 		*pages[FREE_PTE_NR];
+	struct page		**pages;
+	struct page		*local[IA64_GATHER_BUNDLE];
 };
 
 struct ia64_tr_entry {
@@ -90,9 +96,6 @@ extern struct ia64_tr_entry *ia64_idtrs[NR_CPUS];
 #define RR_RID_MASK	0x00000000ffffff00L
 #define RR_TO_RID(val) 	((val >> 8) & 0xffffff)
 
-/* Users of the generic TLB shootdown code must declare this storage space. */
-DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /*
  * Flush the TLB for address range START to END and, if not in fast mode, release the
  * freed pages that where gathered up to this point.
@@ -147,15 +150,23 @@ ia64_tlb_flush_mmu (struct mmu_gather *tlb, unsigned long start, unsigned long e
 	}
 }
 
-/*
- * Return a pointer to an initialized struct mmu_gather.
- */
-static inline struct mmu_gather *
-tlb_gather_mmu (struct mm_struct *mm, unsigned int full_mm_flush)
+static inline void __tlb_alloc_page(struct mmu_gather *tlb)
 {
-	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);
+	unsigned long addr = __get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
 
+	if (addr) {
+		tlb->pages = (void *)addr;
+		tlb->max = PAGE_SIZE / sizeof(void *);
+	}
+}
+
+
+static inline void
+tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned int full_mm_flush)
+{
 	tlb->mm = mm;
+	tlb->max = ARRAY_SIZE(tlb->local);
+	tlb->pages = tlb->local;
 	/*
 	 * Use fast mode if only 1 CPU is online.
 	 *
@@ -172,7 +183,6 @@ tlb_gather_mmu (struct mm_struct *mm, unsigned int full_mm_flush)
 	tlb->nr = (num_online_cpus() == 1) ? ~0U : 0;
 	tlb->fullmm = full_mm_flush;
 	tlb->start_addr = ~0UL;
-	return tlb;
 }
 
 /*
@@ -180,7 +190,7 @@ tlb_gather_mmu (struct mm_struct *mm, unsigned int full_mm_flush)
  * collected.
  */
 static inline void
-tlb_finish_mmu (struct mmu_gather *tlb, unsigned long start, unsigned long end)
+tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
 {
 	/*
 	 * Note: tlb->nr may be 0 at this point, so we can't rely on tlb->start_addr and
@@ -191,7 +201,8 @@ tlb_finish_mmu (struct mmu_gather *tlb, unsigned long start, unsigned long end)
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
 
-	put_cpu_var(mmu_gathers);
+	if (tlb->pages != tlb->local)
+		free_pages((unsigned long)tlb->pages, 0);
 }
 
 /*
@@ -199,18 +210,33 @@ tlb_finish_mmu (struct mmu_gather *tlb, unsigned long start, unsigned long end)
  * must be delayed until after the TLB has been flushed (see comments at the beginning of
  * this file).
  */
-static inline void
-tlb_remove_page (struct mmu_gather *tlb, struct page *page)
+static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
 {
 	tlb->need_flush = 1;
 
 	if (tlb_fast_mode(tlb)) {
 		free_page_and_swap_cache(page);
-		return;
+		return 1; /* avoid calling tlb_flush_mmu */
 	}
+
+	if (!tlb->nr && tlb->pages == tlb->local)
+		__tlb_alloc_page(tlb);
+
 	tlb->pages[tlb->nr++] = page;
-	if (tlb->nr >= FREE_PTE_NR)
-		ia64_tlb_flush_mmu(tlb, tlb->start_addr, tlb->end_addr);
+	VM_BUG_ON(tlb->nr > tlb->max);
+
+	return tlb->max - tlb->nr;
+}
+
+static inline void tlb_flush_mmu(struct mmu_gather *tlb)
+{
+	ia64_tlb_flush_mmu(tlb, tlb->start_addr, tlb->end_addr);
+}
+
+static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	if (!__tlb_remove_page(tlb, page))
+		tlb_flush_mmu(tlb);
 }
 
 /*
diff --git a/arch/ia64/mm/init.c b/arch/ia64/mm/init.c
index ed41759..00cb0e2 100644
--- a/arch/ia64/mm/init.c
+++ b/arch/ia64/mm/init.c
@@ -36,8 +36,6 @@
 #include <asm/mca.h>
 #include <asm/paravirt.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 extern void ia64_tlb_init (void);
 
 unsigned long MAX_DMA_ADDRESS = PAGE_OFFSET + 0x100000000UL;
diff --git a/arch/m32r/mm/init.c b/arch/m32r/mm/init.c
index 73e2205..78b660e 100644
--- a/arch/m32r/mm/init.c
+++ b/arch/m32r/mm/init.c
@@ -35,8 +35,6 @@ extern char __init_begin, __init_end;
 
 pgd_t swapper_pg_dir[1024];
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /*
  * Cache of MMU context last used.
  */
diff --git a/arch/m68k/mm/init_mm.c b/arch/m68k/mm/init_mm.c
index 8bc8425..9113c2f 100644
--- a/arch/m68k/mm/init_mm.c
+++ b/arch/m68k/mm/init_mm.c
@@ -32,8 +32,6 @@
 #include <asm/sections.h>
 #include <asm/tlb.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 pg_data_t pg_data_map[MAX_NUMNODES];
 EXPORT_SYMBOL(pg_data_map);
 
diff --git a/arch/microblaze/mm/init.c b/arch/microblaze/mm/init.c
index c843786..213f2d6 100644
--- a/arch/microblaze/mm/init.c
+++ b/arch/microblaze/mm/init.c
@@ -32,8 +32,6 @@ unsigned int __page_offset;
 EXPORT_SYMBOL(__page_offset);
 
 #else
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 static int init_bootmem_done;
 #endif /* CONFIG_MMU */
 
diff --git a/arch/mips/mm/init.c b/arch/mips/mm/init.c
index 279599e..1aadeb4 100644
--- a/arch/mips/mm/init.c
+++ b/arch/mips/mm/init.c
@@ -64,8 +64,6 @@
 
 #endif /* CONFIG_MIPS_MT_SMTC */
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /*
  * We have up to 8 empty zeroed pages so we can map one of the right colour
  * when needed.  This is necessary only on R4000 / R4400 SC and MC versions
diff --git a/arch/mn10300/mm/init.c b/arch/mn10300/mm/init.c
index 48907cc..1380182 100644
--- a/arch/mn10300/mm/init.c
+++ b/arch/mn10300/mm/init.c
@@ -37,8 +37,6 @@
 #include <asm/tlb.h>
 #include <asm/sections.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 unsigned long highstart_pfn, highend_pfn;
 
 #ifdef CONFIG_MN10300_HAS_ATOMIC_OPS_UNIT
diff --git a/arch/parisc/mm/init.c b/arch/parisc/mm/init.c
index b7ed8d7..102f872 100644
--- a/arch/parisc/mm/init.c
+++ b/arch/parisc/mm/init.c
@@ -31,8 +31,6 @@
 #include <asm/mmzone.h>
 #include <asm/sections.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 extern int  data_start;
 
 #ifdef CONFIG_DISCONTIGMEM
diff --git a/arch/powerpc/include/asm/pgalloc.h b/arch/powerpc/include/asm/pgalloc.h
index abe8532..df1b4cb 100644
--- a/arch/powerpc/include/asm/pgalloc.h
+++ b/arch/powerpc/include/asm/pgalloc.h
@@ -32,13 +32,13 @@ static inline void pte_free(struct mm_struct *mm, pgtable_t ptepage)
 
 #ifdef CONFIG_SMP
 extern void pgtable_free_tlb(struct mmu_gather *tlb, void *table, unsigned shift);
-extern void pte_free_finish(void);
+extern void pte_free_finish(struct mmu_gather *tlb);
 #else /* CONFIG_SMP */
 static inline void pgtable_free_tlb(struct mmu_gather *tlb, void *table, unsigned shift)
 {
 	pgtable_free(table, shift);
 }
-static inline void pte_free_finish(void) { }
+static inline void pte_free_finish(struct mmu_gather *tlb) { }
 #endif /* !CONFIG_SMP */
 
 static inline void __pte_free_tlb(struct mmu_gather *tlb, struct page *ptepage,
diff --git a/arch/powerpc/include/asm/thread_info.h b/arch/powerpc/include/asm/thread_info.h
index d8529ef..37c353e 100644
--- a/arch/powerpc/include/asm/thread_info.h
+++ b/arch/powerpc/include/asm/thread_info.h
@@ -139,10 +139,12 @@ static inline struct thread_info *current_thread_info(void)
 #define TLF_NAPPING		0	/* idle thread enabled NAP mode */
 #define TLF_SLEEPING		1	/* suspend code enabled SLEEP mode */
 #define TLF_RESTORE_SIGMASK	2	/* Restore signal mask in do_signal */
+#define TLF_LAZY_MMU		3	/* tlb_batch is active */
 
 #define _TLF_NAPPING		(1 << TLF_NAPPING)
 #define _TLF_SLEEPING		(1 << TLF_SLEEPING)
 #define _TLF_RESTORE_SIGMASK	(1 << TLF_RESTORE_SIGMASK)
+#define _TLF_LAZY_MMU		(1 << TLF_LAZY_MMU)
 
 #ifndef __ASSEMBLY__
 #define HAVE_SET_RESTORE_SIGMASK	1
diff --git a/arch/powerpc/include/asm/tlb.h b/arch/powerpc/include/asm/tlb.h
index e2b428b..8f0ed7a 100644
--- a/arch/powerpc/include/asm/tlb.h
+++ b/arch/powerpc/include/asm/tlb.h
@@ -28,6 +28,16 @@
 #define tlb_start_vma(tlb, vma)	do { } while (0)
 #define tlb_end_vma(tlb, vma)	do { } while (0)
 
+#define HAVE_ARCH_MMU_GATHER 1
+
+struct pte_freelist_batch;
+
+struct arch_mmu_gather {
+	struct pte_freelist_batch *batch;
+};
+
+#define ARCH_MMU_GATHER_INIT (struct arch_mmu_gather){ .batch = NULL, }
+
 extern void tlb_flush(struct mmu_gather *tlb);
 
 /* Get the generic bits... */
diff --git a/arch/powerpc/kernel/process.c b/arch/powerpc/kernel/process.c
index f74f355..3e37f37 100644
--- a/arch/powerpc/kernel/process.c
+++ b/arch/powerpc/kernel/process.c
@@ -395,6 +395,9 @@ struct task_struct *__switch_to(struct task_struct *prev,
 	struct thread_struct *new_thread, *old_thread;
 	unsigned long flags;
 	struct task_struct *last;
+#ifdef CONFIG_PPC_BOOK3S_64
+	struct ppc64_tlb_batch *batch;
+#endif
 
 #ifdef CONFIG_SMP
 	/* avoid complexity of lazy save/restore of fpu
@@ -513,7 +516,17 @@ struct task_struct *__switch_to(struct task_struct *prev,
 		old_thread->accum_tb += (current_tb - start_tb);
 		new_thread->start_tb = current_tb;
 	}
-#endif
+#endif /* CONFIG_PPC64 */
+
+#ifdef CONFIG_PPC_BOOK3S_64
+	batch = &__get_cpu_var(ppc64_tlb_batch);
+	if (batch->active) {
+		current_thread_info()->local_flags |= _TLF_LAZY_MMU;
+		if (batch->index)
+			__flush_tlb_pending(batch);
+		batch->active = 0;
+	}
+#endif /* CONFIG_PPC_BOOK3S_64 */
 
 	local_irq_save(flags);
 
@@ -528,6 +541,14 @@ struct task_struct *__switch_to(struct task_struct *prev,
 	hard_irq_disable();
 	last = _switch(old_thread, new_thread);
 
+#ifdef CONFIG_PPC_BOOK3S_64
+	if (current_thread_info()->local_flags & _TLF_LAZY_MMU) {
+		current_thread_info()->local_flags &= ~_TLF_LAZY_MMU;
+		batch = &__get_cpu_var(ppc64_tlb_batch);
+		batch->active = 1;
+	}
+#endif /* CONFIG_PPC_BOOK3S_64 */
+
 	local_irq_restore(flags);
 
 	return last;
diff --git a/arch/powerpc/mm/pgtable.c b/arch/powerpc/mm/pgtable.c
index 6a3997f..6e72788 100644
--- a/arch/powerpc/mm/pgtable.c
+++ b/arch/powerpc/mm/pgtable.c
@@ -33,8 +33,6 @@
 
 #include "mmu_decl.h"
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 #ifdef CONFIG_SMP
 
 /*
@@ -43,7 +41,6 @@ DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
  * freeing a page table page that is being walked without locks
  */
 
-static DEFINE_PER_CPU(struct pte_freelist_batch *, pte_freelist_cur);
 static unsigned long pte_freelist_forced_free;
 
 struct pte_freelist_batch
@@ -97,12 +94,10 @@ static void pte_free_submit(struct pte_freelist_batch *batch)
 
 void pgtable_free_tlb(struct mmu_gather *tlb, void *table, unsigned shift)
 {
-	/* This is safe since tlb_gather_mmu has disabled preemption */
-	struct pte_freelist_batch **batchp = &__get_cpu_var(pte_freelist_cur);
+	struct pte_freelist_batch **batchp = &tlb->arch.batch;
 	unsigned long pgf;
 
-	if (atomic_read(&tlb->mm->mm_users) < 2 ||
-	    cpumask_equal(mm_cpumask(tlb->mm), cpumask_of(smp_processor_id()))){
+	if (atomic_read(&tlb->mm->mm_users) < 2) {
 		pgtable_free(table, shift);
 		return;
 	}
@@ -124,10 +119,9 @@ void pgtable_free_tlb(struct mmu_gather *tlb, void *table, unsigned shift)
 	}
 }
 
-void pte_free_finish(void)
+void pte_free_finish(struct mmu_gather *tlb)
 {
-	/* This is safe since tlb_gather_mmu has disabled preemption */
-	struct pte_freelist_batch **batchp = &__get_cpu_var(pte_freelist_cur);
+	struct pte_freelist_batch **batchp = &tlb->arch.batch;
 
 	if (*batchp == NULL)
 		return;
diff --git a/arch/powerpc/mm/tlb_hash32.c b/arch/powerpc/mm/tlb_hash32.c
index 690566b..d555cdb 100644
--- a/arch/powerpc/mm/tlb_hash32.c
+++ b/arch/powerpc/mm/tlb_hash32.c
@@ -73,7 +73,7 @@ void tlb_flush(struct mmu_gather *tlb)
 	}
 
 	/* Push out batch of freed page tables */
-	pte_free_finish();
+	pte_free_finish(tlb);
 }
 
 /*
diff --git a/arch/powerpc/mm/tlb_hash64.c b/arch/powerpc/mm/tlb_hash64.c
index c14d09f..5c94ca3 100644
--- a/arch/powerpc/mm/tlb_hash64.c
+++ b/arch/powerpc/mm/tlb_hash64.c
@@ -155,7 +155,7 @@ void __flush_tlb_pending(struct ppc64_tlb_batch *batch)
 
 void tlb_flush(struct mmu_gather *tlb)
 {
-	struct ppc64_tlb_batch *tlbbatch = &__get_cpu_var(ppc64_tlb_batch);
+	struct ppc64_tlb_batch *tlbbatch = &get_cpu_var(ppc64_tlb_batch);
 
 	/* If there's a TLB batch pending, then we must flush it because the
 	 * pages are going to be freed and we really don't want to have a CPU
@@ -164,8 +164,10 @@ void tlb_flush(struct mmu_gather *tlb)
 	if (tlbbatch->index)
 		__flush_tlb_pending(tlbbatch);
 
+	put_cpu_var(ppc64_tlb_batch);
+
 	/* Push out batch of freed page tables */
-	pte_free_finish();
+	pte_free_finish(tlb);
 }
 
 /**
diff --git a/arch/powerpc/mm/tlb_nohash.c b/arch/powerpc/mm/tlb_nohash.c
index 2a030d8..8eaf67d 100644
--- a/arch/powerpc/mm/tlb_nohash.c
+++ b/arch/powerpc/mm/tlb_nohash.c
@@ -301,7 +301,7 @@ void tlb_flush(struct mmu_gather *tlb)
 	flush_tlb_mm(tlb->mm);
 
 	/* Push out batch of freed page tables */
-	pte_free_finish();
+	pte_free_finish(tlb);
 }
 
 /*
diff --git a/arch/s390/include/asm/tlb.h b/arch/s390/include/asm/tlb.h
index 9074a54..77eee54 100644
--- a/arch/s390/include/asm/tlb.h
+++ b/arch/s390/include/asm/tlb.h
@@ -29,65 +29,77 @@
 #include <asm/smp.h>
 #include <asm/tlbflush.h>
 
-#ifndef CONFIG_SMP
-#define TLB_NR_PTRS	1
-#else
-#define TLB_NR_PTRS	508
-#endif
-
 struct mmu_gather {
 	struct mm_struct *mm;
 	unsigned int fullmm;
 	unsigned int nr_ptes;
 	unsigned int nr_pxds;
-	void *array[TLB_NR_PTRS];
+	unsigned int max;
+	void **array;
+	void *local[8];
 };
 
-DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
-
-static inline struct mmu_gather *tlb_gather_mmu(struct mm_struct *mm,
-						unsigned int full_mm_flush)
+static inline void __tlb_alloc_page(struct mmu_gather *tlb)
 {
-	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);
+	unsigned long addr = __get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
 
+	if (addr) {
+		tlb->array = (void *) addr;
+		tlb->max = PAGE_SIZE / sizeof(void *);
+	}
+}
+
+static inline void tlb_gather_mmu(struct mmu_gather *tlb,
+				  struct mm_struct *mm,
+				  unsigned int full_mm_flush)
+{
 	tlb->mm = mm;
+	tlb->max = ARRAY_SIZE(tlb->local);
+	tlb->array = tlb->local;
 	tlb->fullmm = full_mm_flush;
-	tlb->nr_ptes = 0;
-	tlb->nr_pxds = TLB_NR_PTRS;
 	if (tlb->fullmm)
 		__tlb_flush_mm(mm);
-	return tlb;
+	else
+		__tlb_alloc_page(tlb);
+	tlb->nr_ptes = 0;
+	tlb->nr_pxds = tlb->max;
 }
 
-static inline void tlb_flush_mmu(struct mmu_gather *tlb,
-				 unsigned long start, unsigned long end)
+static inline void tlb_flush_mmu(struct mmu_gather *tlb)
 {
-	if (!tlb->fullmm && (tlb->nr_ptes > 0 || tlb->nr_pxds < TLB_NR_PTRS))
+	if (!tlb->fullmm && (tlb->nr_ptes > 0 || tlb->nr_pxds < tlb->max))
 		__tlb_flush_mm(tlb->mm);
 	while (tlb->nr_ptes > 0)
 		page_table_free_rcu(tlb->mm, tlb->array[--tlb->nr_ptes]);
-	while (tlb->nr_pxds < TLB_NR_PTRS)
+	while (tlb->nr_pxds < tlb->max)
 		crst_table_free_rcu(tlb->mm, tlb->array[tlb->nr_pxds++]);
 }
 
 static inline void tlb_finish_mmu(struct mmu_gather *tlb,
 				  unsigned long start, unsigned long end)
 {
-	tlb_flush_mmu(tlb, start, end);
+	tlb_flush_mmu(tlb);
 
 	rcu_table_freelist_finish();
 
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
 
-	put_cpu_var(mmu_gathers);
+	if (tlb->array != tlb->local)
+		free_pages((unsigned long) tlb->array, 0);
 }
 
 /*
  * Release the page cache reference for a pte removed by
- * tlb_ptep_clear_flush. In both flush modes the tlb fo a page cache page
+ * tlb_ptep_clear_flush. In both flush modes the tlb for a page cache page
  * has already been freed, so just do free_page_and_swap_cache.
  */
+static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	free_page_and_swap_cache(page);
+	return 1; /* avoid calling tlb_flush_mmu */
+}
+
 static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
 {
 	free_page_and_swap_cache(page);
@@ -103,7 +115,7 @@ static inline void pte_free_tlb(struct mmu_gather *tlb, pgtable_t pte,
 	if (!tlb->fullmm) {
 		tlb->array[tlb->nr_ptes++] = pte;
 		if (tlb->nr_ptes >= tlb->nr_pxds)
-			tlb_flush_mmu(tlb, 0, 0);
+			tlb_flush_mmu(tlb);
 	} else
 		page_table_free(tlb->mm, (unsigned long *) pte);
 }
@@ -124,7 +136,7 @@ static inline void pmd_free_tlb(struct mmu_gather *tlb, pmd_t *pmd,
 	if (!tlb->fullmm) {
 		tlb->array[--tlb->nr_pxds] = pmd;
 		if (tlb->nr_ptes >= tlb->nr_pxds)
-			tlb_flush_mmu(tlb, 0, 0);
+			tlb_flush_mmu(tlb);
 	} else
 		crst_table_free(tlb->mm, (unsigned long *) pmd);
 #endif
@@ -146,7 +158,7 @@ static inline void pud_free_tlb(struct mmu_gather *tlb, pud_t *pud,
 	if (!tlb->fullmm) {
 		tlb->array[--tlb->nr_pxds] = pud;
 		if (tlb->nr_ptes >= tlb->nr_pxds)
-			tlb_flush_mmu(tlb, 0, 0);
+			tlb_flush_mmu(tlb);
 	} else
 		crst_table_free(tlb->mm, (unsigned long *) pud);
 #endif
diff --git a/arch/s390/mm/pgtable.c b/arch/s390/mm/pgtable.c
index e1850c2..07fcc3f 100644
--- a/arch/s390/mm/pgtable.c
+++ b/arch/s390/mm/pgtable.c
@@ -36,7 +36,6 @@ struct rcu_table_freelist {
 	((PAGE_SIZE - sizeof(struct rcu_table_freelist)) \
 	  / sizeof(unsigned long))
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
 static DEFINE_PER_CPU(struct rcu_table_freelist *, rcu_table_freelist);
 
 static void __page_table_free(struct mm_struct *mm, unsigned long *table);
diff --git a/arch/score/mm/init.c b/arch/score/mm/init.c
index 50fdec5..cee6bce 100644
--- a/arch/score/mm/init.c
+++ b/arch/score/mm/init.c
@@ -38,8 +38,6 @@
 #include <asm/sections.h>
 #include <asm/tlb.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 unsigned long empty_zero_page;
 EXPORT_SYMBOL_GPL(empty_zero_page);
 
diff --git a/arch/sh/include/asm/tlb.h b/arch/sh/include/asm/tlb.h
index 75abb38..6c308d8 100644
--- a/arch/sh/include/asm/tlb.h
+++ b/arch/sh/include/asm/tlb.h
@@ -23,8 +23,6 @@ struct mmu_gather {
 	unsigned long		start, end;
 };
 
-DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 static inline void init_tlb_gather(struct mmu_gather *tlb)
 {
 	tlb->start = TASK_SIZE;
@@ -36,17 +34,13 @@ static inline void init_tlb_gather(struct mmu_gather *tlb)
 	}
 }
 
-static inline struct mmu_gather *
-tlb_gather_mmu(struct mm_struct *mm, unsigned int full_mm_flush)
+static inline void
+tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned int full_mm_flush)
 {
-	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);
-
 	tlb->mm = mm;
 	tlb->fullmm = full_mm_flush;
 
 	init_tlb_gather(tlb);
-
-	return tlb;
 }
 
 static inline void
@@ -57,8 +51,6 @@ tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
 
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
-
-	put_cpu_var(mmu_gathers);
 }
 
 static inline void
@@ -91,7 +83,21 @@ tlb_end_vma(struct mmu_gather *tlb, struct vm_area_struct *vma)
 	}
 }
 
-#define tlb_remove_page(tlb,page)	free_page_and_swap_cache(page)
+static inline void tlb_flush_mmu(struct mmu_gather *tlb)
+{
+}
+
+static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	free_page_and_swap_cache(page);
+	return 1; /* avoid calling tlb_flush_mmu */
+}
+
+static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	__tlb_remove_page(tlb, page);
+}
+
 #define pte_free_tlb(tlb, ptep, addr)	pte_free((tlb)->mm, ptep)
 #define pmd_free_tlb(tlb, pmdp, addr)	pmd_free((tlb)->mm, pmdp)
 #define pud_free_tlb(tlb, pudp, addr)	pud_free((tlb)->mm, pudp)
diff --git a/arch/sh/mm/init.c b/arch/sh/mm/init.c
index 0d3f912..58a93fb3 100644
--- a/arch/sh/mm/init.c
+++ b/arch/sh/mm/init.c
@@ -28,7 +28,6 @@
 #include <asm/cache.h>
 #include <asm/sizes.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
 pgd_t swapper_pg_dir[PTRS_PER_PGD];
 
 void __init generic_mem_init(void)
diff --git a/arch/sparc/include/asm/pgalloc_64.h b/arch/sparc/include/asm/pgalloc_64.h
index 5bdfa2c..4e5e087 100644
--- a/arch/sparc/include/asm/pgalloc_64.h
+++ b/arch/sparc/include/asm/pgalloc_64.h
@@ -78,4 +78,7 @@ static inline void check_pgt_cache(void)
 	quicklist_trim(0, NULL, 25, 16);
 }
 
+#define __pte_free_tlb(tlb, pte, addr)	pte_free((tlb)->mm, pte)
+#define __pmd_free_tlb(tlb, pmd, addr)	pmd_free((tlb)->mm, pmd)
+
 #endif /* _SPARC64_PGALLOC_H */
diff --git a/arch/sparc/include/asm/pgtable_64.h b/arch/sparc/include/asm/pgtable_64.h
index f8dddb7..b2e85bf 100644
--- a/arch/sparc/include/asm/pgtable_64.h
+++ b/arch/sparc/include/asm/pgtable_64.h
@@ -655,9 +655,11 @@ static inline int pte_special(pte_t pte)
 #define pte_unmap(pte)			do { } while (0)
 
 /* Actual page table PTE updates.  */
-extern void tlb_batch_add(struct mm_struct *mm, unsigned long vaddr, pte_t *ptep, pte_t orig);
+extern void tlb_batch_add(struct mm_struct *mm, unsigned long vaddr,
+			  pte_t *ptep, pte_t orig, int fullmm);
 
-static inline void set_pte_at(struct mm_struct *mm, unsigned long addr, pte_t *ptep, pte_t pte)
+static inline void __set_pte_at(struct mm_struct *mm, unsigned long addr,
+			     pte_t *ptep, pte_t pte, int fullmm)
 {
 	pte_t orig = *ptep;
 
@@ -670,12 +672,19 @@ static inline void set_pte_at(struct mm_struct *mm, unsigned long addr, pte_t *p
 	 *             and SUN4V pte layout, so this inline test is fine.
 	 */
 	if (likely(mm != &init_mm) && (pte_val(orig) & _PAGE_VALID))
-		tlb_batch_add(mm, addr, ptep, orig);
+		tlb_batch_add(mm, addr, ptep, orig, fullmm);
 }
 
+#define set_pte_at(mm,addr,ptep,pte)	\
+	__set_pte_at((mm), (addr), (ptep), (pte), 0)
+
 #define pte_clear(mm,addr,ptep)		\
 	set_pte_at((mm), (addr), (ptep), __pte(0UL))
 
+#define __HAVE_ARCH_PTE_CLEAR_NOT_PRESENT_FULL
+#define pte_clear_not_present_full(mm,addr,ptep,fullmm)	\
+	__set_pte_at((mm), (addr), (ptep), __pte(0UL), (fullmm))
+
 #ifdef DCACHE_ALIASING_POSSIBLE
 #define __HAVE_ARCH_MOVE_PTE
 #define move_pte(pte, prot, old_addr, new_addr)				\
diff --git a/arch/sparc/include/asm/tlb_64.h b/arch/sparc/include/asm/tlb_64.h
index dca406b..190e189 100644
--- a/arch/sparc/include/asm/tlb_64.h
+++ b/arch/sparc/include/asm/tlb_64.h
@@ -7,66 +7,11 @@
 #include <asm/tlbflush.h>
 #include <asm/mmu_context.h>
 
-#define TLB_BATCH_NR	192
-
-/*
- * For UP we don't need to worry about TLB flush
- * and page free order so much..
- */
-#ifdef CONFIG_SMP
-  #define FREE_PTE_NR	506
-  #define tlb_fast_mode(bp) ((bp)->pages_nr == ~0U)
-#else
-  #define FREE_PTE_NR	1
-  #define tlb_fast_mode(bp) 1
-#endif
-
-struct mmu_gather {
-	struct mm_struct *mm;
-	unsigned int pages_nr;
-	unsigned int need_flush;
-	unsigned int fullmm;
-	unsigned int tlb_nr;
-	unsigned long vaddrs[TLB_BATCH_NR];
-	struct page *pages[FREE_PTE_NR];
-};
-
-DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 #ifdef CONFIG_SMP
 extern void smp_flush_tlb_pending(struct mm_struct *,
 				  unsigned long, unsigned long *);
 #endif
 
-extern void __flush_tlb_pending(unsigned long, unsigned long, unsigned long *);
-extern void flush_tlb_pending(void);
-
-static inline struct mmu_gather *tlb_gather_mmu(struct mm_struct *mm, unsigned int full_mm_flush)
-{
-	struct mmu_gather *mp = &get_cpu_var(mmu_gathers);
-
-	BUG_ON(mp->tlb_nr);
-
-	mp->mm = mm;
-	mp->pages_nr = num_online_cpus() > 1 ? 0U : ~0U;
-	mp->fullmm = full_mm_flush;
-
-	return mp;
-}
-
-
-static inline void tlb_flush_mmu(struct mmu_gather *mp)
-{
-	if (!mp->fullmm)
-		flush_tlb_pending();
-	if (mp->need_flush) {
-		free_pages_and_swap_cache(mp->pages, mp->pages_nr);
-		mp->pages_nr = 0;
-		mp->need_flush = 0;
-	}
-
-}
-
 #ifdef CONFIG_SMP
 extern void smp_flush_tlb_mm(struct mm_struct *mm);
 #define do_flush_tlb_mm(mm) smp_flush_tlb_mm(mm)
@@ -74,38 +19,14 @@ extern void smp_flush_tlb_mm(struct mm_struct *mm);
 #define do_flush_tlb_mm(mm) __flush_tlb_mm(CTX_HWBITS(mm->context), SECONDARY_CONTEXT)
 #endif
 
-static inline void tlb_finish_mmu(struct mmu_gather *mp, unsigned long start, unsigned long end)
-{
-	tlb_flush_mmu(mp);
-
-	if (mp->fullmm)
-		mp->fullmm = 0;
-
-	/* keep the page table cache within bounds */
-	check_pgt_cache();
-
-	put_cpu_var(mmu_gathers);
-}
-
-static inline void tlb_remove_page(struct mmu_gather *mp, struct page *page)
-{
-	if (tlb_fast_mode(mp)) {
-		free_page_and_swap_cache(page);
-		return;
-	}
-	mp->need_flush = 1;
-	mp->pages[mp->pages_nr++] = page;
-	if (mp->pages_nr >= FREE_PTE_NR)
-		tlb_flush_mmu(mp);
-}
-
-#define tlb_remove_tlb_entry(mp,ptep,addr) do { } while (0)
-#define pte_free_tlb(mp, ptepage, addr) pte_free((mp)->mm, ptepage)
-#define pmd_free_tlb(mp, pmdp, addr) pmd_free((mp)->mm, pmdp)
-#define pud_free_tlb(tlb,pudp, addr) __pud_free_tlb(tlb,pudp,addr)
+extern void __flush_tlb_pending(unsigned long, unsigned long, unsigned long *);
+extern void flush_tlb_pending(void);
 
-#define tlb_migrate_finish(mm)	do { } while (0)
 #define tlb_start_vma(tlb, vma) do { } while (0)
 #define tlb_end_vma(tlb, vma)	do { } while (0)
+#define __tlb_remove_tlb_entry(tlb, ptep, address) do { } while (0)
+#define tlb_flush(tlb)	flush_tlb_pending()
+
+#include <asm-generic/tlb.h>
 
 #endif /* _SPARC64_TLB_H */
diff --git a/arch/sparc/include/asm/tlbflush_64.h b/arch/sparc/include/asm/tlbflush_64.h
index fbb675d..2ef4634 100644
--- a/arch/sparc/include/asm/tlbflush_64.h
+++ b/arch/sparc/include/asm/tlbflush_64.h
@@ -5,9 +5,17 @@
 #include <asm/mmu_context.h>
 
 /* TSB flush operations. */
-struct mmu_gather;
+
+#define TLB_BATCH_NR	192
+
+struct tlb_batch {
+	struct mm_struct *mm;
+	unsigned long tlb_nr;
+	unsigned long vaddrs[TLB_BATCH_NR];
+};
+
 extern void flush_tsb_kernel_range(unsigned long start, unsigned long end);
-extern void flush_tsb_user(struct mmu_gather *mp);
+extern void flush_tsb_user(struct tlb_batch *tb);
 
 /* TLB flush operations. */
 
diff --git a/arch/sparc/mm/init_32.c b/arch/sparc/mm/init_32.c
index 4c31e2b..a755487 100644
--- a/arch/sparc/mm/init_32.c
+++ b/arch/sparc/mm/init_32.c
@@ -37,8 +37,6 @@
 #include <asm/prom.h>
 #include <asm/leon.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 unsigned long *sparc_valid_addr_bitmap;
 EXPORT_SYMBOL(sparc_valid_addr_bitmap);
 
diff --git a/arch/sparc/mm/tlb.c b/arch/sparc/mm/tlb.c
index d8f21e2..b1f279c 100644
--- a/arch/sparc/mm/tlb.c
+++ b/arch/sparc/mm/tlb.c
@@ -19,33 +19,34 @@
 
 /* Heavily inspired by the ppc64 code.  */
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
+static DEFINE_PER_CPU(struct tlb_batch, tlb_batch);
 
 void flush_tlb_pending(void)
 {
-	struct mmu_gather *mp = &get_cpu_var(mmu_gathers);
+	struct tlb_batch *tb = &get_cpu_var(tlb_batch);
 
-	if (mp->tlb_nr) {
-		flush_tsb_user(mp);
+	if (tb->tlb_nr) {
+		flush_tsb_user(tb);
 
-		if (CTX_VALID(mp->mm->context)) {
+		if (CTX_VALID(tb->mm->context)) {
 #ifdef CONFIG_SMP
-			smp_flush_tlb_pending(mp->mm, mp->tlb_nr,
-					      &mp->vaddrs[0]);
+			smp_flush_tlb_pending(tb->mm, tb->tlb_nr,
+					      &tb->vaddrs[0]);
 #else
-			__flush_tlb_pending(CTX_HWBITS(mp->mm->context),
-					    mp->tlb_nr, &mp->vaddrs[0]);
+			__flush_tlb_pending(CTX_HWBITS(tb->mm->context),
+					    tb->tlb_nr, &tb->vaddrs[0]);
 #endif
 		}
-		mp->tlb_nr = 0;
+		tb->tlb_nr = 0;
 	}
 
-	put_cpu_var(mmu_gathers);
+	put_cpu_var(tlb_batch);
 }
 
-void tlb_batch_add(struct mm_struct *mm, unsigned long vaddr, pte_t *ptep, pte_t orig)
+void tlb_batch_add(struct mm_struct *mm, unsigned long vaddr,
+		   pte_t *ptep, pte_t orig, int fullmm)
 {
-	struct mmu_gather *mp = &__get_cpu_var(mmu_gathers);
+	struct tlb_batch *tb = &get_cpu_var(tlb_batch);
 	unsigned long nr;
 
 	vaddr &= PAGE_MASK;
@@ -77,21 +78,25 @@ void tlb_batch_add(struct mm_struct *mm, unsigned long vaddr, pte_t *ptep, pte_t
 
 no_cache_flush:
 
-	if (mp->fullmm)
+	if (fullmm) {
+		put_cpu_var(tlb_batch);
 		return;
+	}
 
-	nr = mp->tlb_nr;
+	nr = tb->tlb_nr;
 
-	if (unlikely(nr != 0 && mm != mp->mm)) {
+	if (unlikely(nr != 0 && mm != tb->mm)) {
 		flush_tlb_pending();
 		nr = 0;
 	}
 
 	if (nr == 0)
-		mp->mm = mm;
+		tb->mm = mm;
 
-	mp->vaddrs[nr] = vaddr;
-	mp->tlb_nr = ++nr;
+	tb->vaddrs[nr] = vaddr;
+	tb->tlb_nr = ++nr;
 	if (nr >= TLB_BATCH_NR)
 		flush_tlb_pending();
+
+	put_cpu_var(tlb_batch);
 }
diff --git a/arch/sparc/mm/tsb.c b/arch/sparc/mm/tsb.c
index 101d7c8..9484615 100644
--- a/arch/sparc/mm/tsb.c
+++ b/arch/sparc/mm/tsb.c
@@ -47,12 +47,13 @@ void flush_tsb_kernel_range(unsigned long start, unsigned long end)
 	}
 }
 
-static void __flush_tsb_one(struct mmu_gather *mp, unsigned long hash_shift, unsigned long tsb, unsigned long nentries)
+static void __flush_tsb_one(struct tlb_batch *tb, unsigned long hash_shift,
+			    unsigned long tsb, unsigned long nentries)
 {
 	unsigned long i;
 
-	for (i = 0; i < mp->tlb_nr; i++) {
-		unsigned long v = mp->vaddrs[i];
+	for (i = 0; i < tb->tlb_nr; i++) {
+		unsigned long v = tb->vaddrs[i];
 		unsigned long tag, ent, hash;
 
 		v &= ~0x1UL;
@@ -65,9 +66,9 @@ static void __flush_tsb_one(struct mmu_gather *mp, unsigned long hash_shift, uns
 	}
 }
 
-void flush_tsb_user(struct mmu_gather *mp)
+void flush_tsb_user(struct tlb_batch *tb)
 {
-	struct mm_struct *mm = mp->mm;
+	struct mm_struct *mm = tb->mm;
 	unsigned long nentries, base, flags;
 
 	spin_lock_irqsave(&mm->context.lock, flags);
@@ -76,7 +77,7 @@ void flush_tsb_user(struct mmu_gather *mp)
 	nentries = mm->context.tsb_block[MM_TSB_BASE].tsb_nentries;
 	if (tlb_type == cheetah_plus || tlb_type == hypervisor)
 		base = __pa(base);
-	__flush_tsb_one(mp, PAGE_SHIFT, base, nentries);
+	__flush_tsb_one(tb, PAGE_SHIFT, base, nentries);
 
 #ifdef CONFIG_HUGETLB_PAGE
 	if (mm->context.tsb_block[MM_TSB_HUGE].tsb) {
@@ -84,7 +85,7 @@ void flush_tsb_user(struct mmu_gather *mp)
 		nentries = mm->context.tsb_block[MM_TSB_HUGE].tsb_nentries;
 		if (tlb_type == cheetah_plus || tlb_type == hypervisor)
 			base = __pa(base);
-		__flush_tsb_one(mp, HPAGE_SHIFT, base, nentries);
+		__flush_tsb_one(tb, HPAGE_SHIFT, base, nentries);
 	}
 #endif
 	spin_unlock_irqrestore(&mm->context.lock, flags);
diff --git a/arch/tile/mm/init.c b/arch/tile/mm/init.c
index d6e87fd..4e10c40 100644
--- a/arch/tile/mm/init.c
+++ b/arch/tile/mm/init.c
@@ -60,8 +60,6 @@ unsigned long VMALLOC_RESERVE = CONFIG_VMALLOC_RESERVE;
 EXPORT_SYMBOL(VMALLOC_RESERVE);
 #endif
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /* Create an L2 page table */
 static pte_t * __init alloc_pte(void)
 {
diff --git a/arch/um/include/asm/tlb.h b/arch/um/include/asm/tlb.h
index 660caed..4febacd 100644
--- a/arch/um/include/asm/tlb.h
+++ b/arch/um/include/asm/tlb.h
@@ -22,9 +22,6 @@ struct mmu_gather {
 	unsigned int		fullmm; /* non-zero means full mm flush */
 };
 
-/* Users of the generic TLB shootdown code must declare this storage space. */
-DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 static inline void __tlb_remove_tlb_entry(struct mmu_gather *tlb, pte_t *ptep,
 					  unsigned long address)
 {
@@ -47,27 +44,20 @@ static inline void init_tlb_gather(struct mmu_gather *tlb)
 	}
 }
 
-/* tlb_gather_mmu
- *	Return a pointer to an initialized struct mmu_gather.
- */
-static inline struct mmu_gather *
-tlb_gather_mmu(struct mm_struct *mm, unsigned int full_mm_flush)
+static inline void
+tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned int full_mm_flush)
 {
-	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);
-
 	tlb->mm = mm;
 	tlb->fullmm = full_mm_flush;
 
 	init_tlb_gather(tlb);
-
-	return tlb;
 }
 
 extern void flush_tlb_mm_range(struct mm_struct *mm, unsigned long start,
 			       unsigned long end);
 
 static inline void
-tlb_flush_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
+tlb_flush_mmu(struct mmu_gather *tlb)
 {
 	if (!tlb->need_flush)
 		return;
@@ -83,12 +73,10 @@ tlb_flush_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
 static inline void
 tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
 {
-	tlb_flush_mmu(tlb, start, end);
+	tlb_flush_mmu(tlb);
 
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
-
-	put_cpu_var(mmu_gathers);
 }
 
 /* tlb_remove_page
@@ -96,11 +84,16 @@ tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
  *	while handling the additional races in SMP caused by other CPUs
  *	caching valid mappings in their TLBs.
  */
-static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
 {
 	tlb->need_flush = 1;
 	free_page_and_swap_cache(page);
-	return;
+	return 1; /* avoid calling tlb_flush_mmu */
+}
+
+static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	__tlb_remove_page(tlb, page);
 }
 
 /**
diff --git a/arch/um/kernel/smp.c b/arch/um/kernel/smp.c
index 106bf27..d9011e0 100644
--- a/arch/um/kernel/smp.c
+++ b/arch/um/kernel/smp.c
@@ -7,9 +7,6 @@
 #include "asm/pgalloc.h"
 #include "asm/tlb.h"
 
-/* For some reason, mmu_gathers are referenced when CONFIG_SMP is off. */
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 #ifdef CONFIG_SMP
 
 #include "linux/sched.h"
diff --git a/arch/unicore32/mm/mmu.c b/arch/unicore32/mm/mmu.c
index 7bf3d58..c5b2b65 100644
--- a/arch/unicore32/mm/mmu.c
+++ b/arch/unicore32/mm/mmu.c
@@ -30,8 +30,6 @@
 
 #include "mm.h"
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 /*
  * empty_zero_page is a special page that is used for
  * zero-initialized data and COW.
diff --git a/arch/x86/mm/init.c b/arch/x86/mm/init.c
index 286d289..cda082e 100644
--- a/arch/x86/mm/init.c
+++ b/arch/x86/mm/init.c
@@ -16,8 +16,6 @@
 #include <asm/tlb.h>
 #include <asm/proto.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 unsigned long __initdata pgt_buf_start;
 unsigned long __meminitdata pgt_buf_end;
 unsigned long __meminitdata pgt_buf_top;
diff --git a/arch/xtensa/mm/mmu.c b/arch/xtensa/mm/mmu.c
index 4bb91a9..ca81654 100644
--- a/arch/xtensa/mm/mmu.c
+++ b/arch/xtensa/mm/mmu.c
@@ -14,8 +14,6 @@
 #include <asm/mmu_context.h>
 #include <asm/page.h>
 
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
 void __init paging_init(void)
 {
 	memset(swapper_pg_dir, 0, PAGE_SIZE);
diff --git a/fs/exec.c b/fs/exec.c
index 5e62d26..14e623e 100644
--- a/fs/exec.c
+++ b/fs/exec.c
@@ -553,7 +553,7 @@ static int shift_arg_pages(struct vm_area_struct *vma, unsigned long shift)
 	unsigned long length = old_end - old_start;
 	unsigned long new_start = old_start - shift;
 	unsigned long new_end = old_end - shift;
-	struct mmu_gather *tlb;
+	struct mmu_gather tlb;
 
 	BUG_ON(new_start > new_end);
 
@@ -579,12 +579,12 @@ static int shift_arg_pages(struct vm_area_struct *vma, unsigned long shift)
 		return -ENOMEM;
 
 	lru_add_drain();
-	tlb = tlb_gather_mmu(mm, 0);
+	tlb_gather_mmu(&tlb, mm, 0);
 	if (new_end > old_start) {
 		/*
 		 * when the old and new regions overlap clear from new_end.
 		 */
-		free_pgd_range(tlb, new_end, old_end, new_end,
+		free_pgd_range(&tlb, new_end, old_end, new_end,
 			vma->vm_next ? vma->vm_next->vm_start : 0);
 	} else {
 		/*
@@ -593,10 +593,10 @@ static int shift_arg_pages(struct vm_area_struct *vma, unsigned long shift)
 		 * have constraints on va-space that make this illegal (IA64) -
 		 * for the others its just a little faster.
 		 */
-		free_pgd_range(tlb, old_start, old_end, new_end,
+		free_pgd_range(&tlb, old_start, old_end, new_end,
 			vma->vm_next ? vma->vm_next->vm_start : 0);
 	}
-	tlb_finish_mmu(tlb, new_end, old_end);
+	tlb_finish_mmu(&tlb, new_end, old_end);
 
 	/*
 	 * Shrink the vma to just the new range.  Always succeeds.
diff --git a/include/asm-generic/tlb.h b/include/asm-generic/tlb.h
index e43f976..67f21e2 100644
--- a/include/asm-generic/tlb.h
+++ b/include/asm-generic/tlb.h
@@ -5,6 +5,8 @@
  * Copyright 2001 Red Hat, Inc.
  * Based on code from mm/memory.c Copyright Linus Torvalds and others.
  *
+ * Copyright 2011 Red Hat, Inc., Peter Zijlstra <pzijlstr@redhat.com>
+ *
  * This program is free software; you can redistribute it and/or
  * modify it under the terms of the GNU General Public License
  * as published by the Free Software Foundation; either version
@@ -22,51 +24,71 @@
  * and page free order so much..
  */
 #ifdef CONFIG_SMP
-  #ifdef ARCH_FREE_PTR_NR
-    #define FREE_PTR_NR   ARCH_FREE_PTR_NR
-  #else
-    #define FREE_PTE_NR	506
-  #endif
   #define tlb_fast_mode(tlb) ((tlb)->nr == ~0U)
 #else
-  #define FREE_PTE_NR	1
   #define tlb_fast_mode(tlb) 1
 #endif
 
+/*
+ * If we can't allocate a page to make a big patch of page pointers
+ * to work on, then just handle a few from the on-stack structure.
+ */
+#define MMU_GATHER_BUNDLE	8
+
 /* struct mmu_gather is an opaque type used by the mm code for passing around
  * any data needed by arch specific code for tlb_remove_page.
  */
 struct mmu_gather {
 	struct mm_struct	*mm;
 	unsigned int		nr;	/* set to ~0U means fast mode */
+	unsigned int		max;	/* nr < max */
 	unsigned int		need_flush;/* Really unmapped some ptes? */
 	unsigned int		fullmm; /* non-zero means full mm flush */
-	struct page *		pages[FREE_PTE_NR];
+#ifdef HAVE_ARCH_MMU_GATHER
+	struct arch_mmu_gather	arch;
+#endif
+	struct page		**pages;
+	struct page		*local[MMU_GATHER_BUNDLE];
 };
 
-/* Users of the generic TLB shootdown code must declare this storage space. */
-DECLARE_PER_CPU(struct mmu_gather, mmu_gathers);
+static inline void __tlb_alloc_page(struct mmu_gather *tlb)
+{
+	unsigned long addr = __get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
+
+	if (addr) {
+		tlb->pages = (void *)addr;
+		tlb->max = PAGE_SIZE / sizeof(struct page *);
+	}
+}
 
 /* tlb_gather_mmu
- *	Return a pointer to an initialized struct mmu_gather.
+ *	Called to initialize an (on-stack) mmu_gather structure for page-table
+ *	tear-down from @mm. The @fullmm argument is used when @mm is without
+ *	users and we're going to destroy the full address space (exit/execve).
  */
-static inline struct mmu_gather *
-tlb_gather_mmu(struct mm_struct *mm, unsigned int full_mm_flush)
+static inline void
+tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned int fullmm)
 {
-	struct mmu_gather *tlb = &get_cpu_var(mmu_gathers);
-
 	tlb->mm = mm;
 
-	/* Use fast mode if only one CPU is online */
-	tlb->nr = num_online_cpus() > 1 ? 0U : ~0U;
+	tlb->max = ARRAY_SIZE(tlb->local);
+	tlb->pages = tlb->local;
+
+	if (num_online_cpus() > 1) {
+		tlb->nr = 0;
+		__tlb_alloc_page(tlb);
+	} else /* Use fast mode if only one CPU is online */
+		tlb->nr = ~0U;
 
-	tlb->fullmm = full_mm_flush;
+	tlb->fullmm = fullmm;
 
-	return tlb;
+#ifdef HAVE_ARCH_MMU_GATHER
+	tlb->arch = ARCH_MMU_GATHER_INIT;
+#endif
 }
 
 static inline void
-tlb_flush_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
+tlb_flush_mmu(struct mmu_gather *tlb)
 {
 	if (!tlb->need_flush)
 		return;
@@ -75,6 +97,13 @@ tlb_flush_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
 	if (!tlb_fast_mode(tlb)) {
 		free_pages_and_swap_cache(tlb->pages, tlb->nr);
 		tlb->nr = 0;
+		/*
+		 * If we are using the local on-stack array of pages for MMU
+		 * gather, try allocating an off-stack array again as we have
+		 * recently freed pages.
+		 */
+		if (tlb->pages == tlb->local)
+			__tlb_alloc_page(tlb);
 	}
 }
 
@@ -85,29 +114,42 @@ tlb_flush_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
 static inline void
 tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
 {
-	tlb_flush_mmu(tlb, start, end);
+	tlb_flush_mmu(tlb);
 
 	/* keep the page table cache within bounds */
 	check_pgt_cache();
 
-	put_cpu_var(mmu_gathers);
+	if (tlb->pages != tlb->local)
+		free_pages((unsigned long)tlb->pages, 0);
 }
 
-/* tlb_remove_page
+/* __tlb_remove_page
  *	Must perform the equivalent to __free_pte(pte_get_and_clear(ptep)), while
  *	handling the additional races in SMP caused by other CPUs caching valid
- *	mappings in their TLBs.
+ *	mappings in their TLBs. Returns the number of free page slots left.
+ *	When out of page slots we must call tlb_flush_mmu().
  */
-static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
 {
 	tlb->need_flush = 1;
 	if (tlb_fast_mode(tlb)) {
 		free_page_and_swap_cache(page);
-		return;
+		return 1; /* avoid calling tlb_flush_mmu() */
 	}
 	tlb->pages[tlb->nr++] = page;
-	if (tlb->nr >= FREE_PTE_NR)
-		tlb_flush_mmu(tlb, 0, 0);
+	VM_BUG_ON(tlb->nr > tlb->max);
+
+	return tlb->max - tlb->nr;
+}
+
+/* tlb_remove_page
+ *	Similar to __tlb_remove_page but will call tlb_flush_mmu() itself when
+ *	required.
+ */
+static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	if (!__tlb_remove_page(tlb, page))
+		tlb_flush_mmu(tlb);
 }
 
 /**
diff --git a/include/linux/mm.h b/include/linux/mm.h
index 7606d7d..2bf8bf1 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -904,7 +904,7 @@ int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
 		unsigned long size);
 unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
 		unsigned long size, struct zap_details *);
-unsigned long unmap_vmas(struct mmu_gather **tlb,
+unsigned long unmap_vmas(struct mmu_gather *tlb,
 		struct vm_area_struct *start_vma, unsigned long start_addr,
 		unsigned long end_addr, unsigned long *nr_accounted,
 		struct zap_details *);
diff --git a/mm/memory.c b/mm/memory.c
index 9da8cab..e6eddc4 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -912,12 +912,13 @@ static unsigned long zap_pte_range(struct mmu_gather *tlb,
 				long *zap_work, struct zap_details *details)
 {
 	struct mm_struct *mm = tlb->mm;
+	int force_flush = 0;
 	pte_t *pte;
 	spinlock_t *ptl;
 	int rss[NR_MM_COUNTERS];
 
 	init_rss_vec(rss);
-
+again:
 	pte = pte_offset_map_lock(mm, pmd, addr, &ptl);
 	arch_enter_lazy_mmu_mode();
 	do {
@@ -974,7 +975,9 @@ static unsigned long zap_pte_range(struct mmu_gather *tlb,
 			page_remove_rmap(page);
 			if (unlikely(page_mapcount(page) < 0))
 				print_bad_pte(vma, addr, ptent, page);
-			tlb_remove_page(tlb, page);
+			force_flush = !__tlb_remove_page(tlb, page);
+			if (force_flush)
+				break;
 			continue;
 		}
 		/*
@@ -1001,6 +1004,18 @@ static unsigned long zap_pte_range(struct mmu_gather *tlb,
 	arch_leave_lazy_mmu_mode();
 	pte_unmap_unlock(pte - 1, ptl);
 
+	/*
+	 * mmu_gather ran out of room to batch pages, we break out of
+	 * the PTE lock to avoid doing the potential expensive TLB invalidate
+	 * and page-free while holding it.
+	 */
+	if (force_flush) {
+		force_flush = 0;
+		tlb_flush_mmu(tlb);
+		if (addr != end)
+			goto again;
+	}
+
 	return addr;
 }
 
@@ -1121,17 +1136,14 @@ static unsigned long unmap_page_range(struct mmu_gather *tlb,
  * ensure that any thus-far unmapped pages are flushed before unmap_vmas()
  * drops the lock and schedules.
  */
-unsigned long unmap_vmas(struct mmu_gather **tlbp,
+unsigned long unmap_vmas(struct mmu_gather *tlb,
 		struct vm_area_struct *vma, unsigned long start_addr,
 		unsigned long end_addr, unsigned long *nr_accounted,
 		struct zap_details *details)
 {
 	long zap_work = ZAP_BLOCK_SIZE;
-	unsigned long tlb_start = 0;	/* For tlb_finish_mmu */
-	int tlb_start_valid = 0;
 	unsigned long start = start_addr;
 	spinlock_t *i_mmap_lock = details? details->i_mmap_lock: NULL;
-	int fullmm = (*tlbp)->fullmm;
 	struct mm_struct *mm = vma->vm_mm;
 
 	mmu_notifier_invalidate_range_start(mm, start_addr, end_addr);
@@ -1152,11 +1164,6 @@ unsigned long unmap_vmas(struct mmu_gather **tlbp,
 			untrack_pfn_vma(vma, 0, 0);
 
 		while (start != end) {
-			if (!tlb_start_valid) {
-				tlb_start = start;
-				tlb_start_valid = 1;
-			}
-
 			if (unlikely(is_vm_hugetlb_page(vma))) {
 				/*
 				 * It is undesirable to test vma->vm_file as it
@@ -1177,7 +1184,7 @@ unsigned long unmap_vmas(struct mmu_gather **tlbp,
 
 				start = end;
 			} else
-				start = unmap_page_range(*tlbp, vma,
+				start = unmap_page_range(tlb, vma,
 						start, end, &zap_work, details);
 
 			if (zap_work > 0) {
@@ -1185,19 +1192,13 @@ unsigned long unmap_vmas(struct mmu_gather **tlbp,
 				break;
 			}
 
-			tlb_finish_mmu(*tlbp, tlb_start, start);
-
 			if (need_resched() ||
 				(i_mmap_lock && spin_needbreak(i_mmap_lock))) {
-				if (i_mmap_lock) {
-					*tlbp = NULL;
+				if (i_mmap_lock)
 					goto out;
-				}
 				cond_resched();
 			}
 
-			*tlbp = tlb_gather_mmu(vma->vm_mm, fullmm);
-			tlb_start_valid = 0;
 			zap_work = ZAP_BLOCK_SIZE;
 		}
 	}
@@ -1217,16 +1218,15 @@ unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
 		unsigned long size, struct zap_details *details)
 {
 	struct mm_struct *mm = vma->vm_mm;
-	struct mmu_gather *tlb;
+	struct mmu_gather tlb;
 	unsigned long end = address + size;
 	unsigned long nr_accounted = 0;
 
 	lru_add_drain();
-	tlb = tlb_gather_mmu(mm, 0);
+	tlb_gather_mmu(&tlb, mm, 0);
 	update_hiwater_rss(mm);
 	end = unmap_vmas(&tlb, vma, address, end, &nr_accounted, details);
-	if (tlb)
-		tlb_finish_mmu(tlb, address, end);
+	tlb_finish_mmu(&tlb, address, end);
 	return end;
 }
 
diff --git a/mm/mmap.c b/mm/mmap.c
index 2ec8eb5..f8cbc86 100644
--- a/mm/mmap.c
+++ b/mm/mmap.c
@@ -1913,17 +1913,17 @@ static void unmap_region(struct mm_struct *mm,
 		unsigned long start, unsigned long end)
 {
 	struct vm_area_struct *next = prev? prev->vm_next: mm->mmap;
-	struct mmu_gather *tlb;
+	struct mmu_gather tlb;
 	unsigned long nr_accounted = 0;
 
 	lru_add_drain();
-	tlb = tlb_gather_mmu(mm, 0);
+	tlb_gather_mmu(&tlb, mm, 0);
 	update_hiwater_rss(mm);
 	unmap_vmas(&tlb, vma, start, end, &nr_accounted, NULL);
 	vm_unacct_memory(nr_accounted);
-	free_pgtables(tlb, vma, prev? prev->vm_end: FIRST_USER_ADDRESS,
-				 next? next->vm_start: 0);
-	tlb_finish_mmu(tlb, start, end);
+	free_pgtables(&tlb, vma, prev ? prev->vm_end : FIRST_USER_ADDRESS,
+				 next ? next->vm_start : 0);
+	tlb_finish_mmu(&tlb, start, end);
 }
 
 /*
@@ -2265,7 +2265,7 @@ EXPORT_SYMBOL(do_brk);
 /* Release all mmaps. */
 void exit_mmap(struct mm_struct *mm)
 {
-	struct mmu_gather *tlb;
+	struct mmu_gather tlb;
 	struct vm_area_struct *vma;
 	unsigned long nr_accounted = 0;
 	unsigned long end;
@@ -2290,14 +2290,14 @@ void exit_mmap(struct mm_struct *mm)
 
 	lru_add_drain();
 	flush_cache_mm(mm);
-	tlb = tlb_gather_mmu(mm, 1);
+	tlb_gather_mmu(&tlb, mm, 1);
 	/* update_hiwater_rss(mm) here? but nobody should be looking */
 	/* Use -1 here to ensure all VMAs in the mm are unmapped */
 	end = unmap_vmas(&tlb, vma, 0, -1, &nr_accounted, NULL);
 	vm_unacct_memory(nr_accounted);
 
-	free_pgtables(tlb, vma, FIRST_USER_ADDRESS, 0);
-	tlb_finish_mmu(tlb, 0, end);
+	free_pgtables(&tlb, vma, FIRST_USER_ADDRESS, 0);
+	tlb_finish_mmu(&tlb, 0, end);
 
 	/*
 	 * Walk the list again, actually closing and freeing it,

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 01/20] mm: mmu_gather rework

[Andrew Morton]

On Fri, 01 Apr 2011 14:12:59 +0200
Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:

> Remove the first obstackle towards a fully preemptible mmu_gather.
> 
> The current scheme assumes mmu_gather is always done with preemption
> disabled and uses per-cpu storage for the page batches. Change this to
> try and allocate a page for batching and in case of failure, use a
> small on-stack array to make some progress.
> 
> Preemptible mmu_gather is desired in general and usable once
> i_mmap_lock becomes a mutex. Doing it before the mutex conversion
> saves us from having to rework the code by moving the mmu_gather
> bits inside the pte_lock.
> 
> Also avoid flushing the tlb batches from under the pte lock,
> this is useful even without the i_mmap_lock conversion as it
> significantly reduces pte lock hold times.

There doesn't seem much point in reviewing this closely, as a lot of it
gets tossed away later in the series..

>  		free_pages_and_swap_cache(tlb->pages, tlb->nr);

It seems inappropriate that this code uses
free_page[s]_and_swap_cache().  It should go direct to put_page() and
release_pages()?  Please review this code's implicit decision to pass
"cold==0" into release_pages().

> -static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
> +static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)

I wonder if all the inlining which remains in this code is needed and
desirable.

Re: [PATCH 01/20] mm: mmu_gather rework

[Andrew Morton]

On Fri, 01 Apr 2011 14:12:59 +0200
Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:

> Remove the first obstackle towards a fully preemptible mmu_gather.
> 
> The current scheme assumes mmu_gather is always done with preemption
> disabled and uses per-cpu storage for the page batches. Change this to
> try and allocate a page for batching and in case of failure, use a
> small on-stack array to make some progress.
> 
> Preemptible mmu_gather is desired in general and usable once
> i_mmap_lock becomes a mutex. Doing it before the mutex conversion
> saves us from having to rework the code by moving the mmu_gather
> bits inside the pte_lock.
> 
> Also avoid flushing the tlb batches from under the pte lock,
> this is useful even without the i_mmap_lock conversion as it
> significantly reduces pte lock hold times.

There doesn't seem much point in reviewing this closely, as a lot of it
gets tossed away later in the series..

>  		free_pages_and_swap_cache(tlb->pages, tlb->nr);

It seems inappropriate that this code uses
free_page[s]_and_swap_cache().  It should go direct to put_page() and
release_pages()?  Please review this code's implicit decision to pass
"cold==0" into release_pages().

> -static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
> +static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)

I wonder if all the inlining which remains in this code is needed and
desirable.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 01/20] mm: mmu_gather rework

[Andrew Morton]

On Fri, 01 Apr 2011 14:12:59 +0200
Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:

> Remove the first obstackle towards a fully preemptible mmu_gather.
> 
> The current scheme assumes mmu_gather is always done with preemption
> disabled and uses per-cpu storage for the page batches. Change this to
> try and allocate a page for batching and in case of failure, use a
> small on-stack array to make some progress.
> 
> Preemptible mmu_gather is desired in general and usable once
> i_mmap_lock becomes a mutex. Doing it before the mutex conversion
> saves us from having to rework the code by moving the mmu_gather
> bits inside the pte_lock.
> 
> Also avoid flushing the tlb batches from under the pte lock,
> this is useful even without the i_mmap_lock conversion as it
> significantly reduces pte lock hold times.

There doesn't seem much point in reviewing this closely, as a lot of it
gets tossed away later in the series..

>  		free_pages_and_swap_cache(tlb->pages, tlb->nr);

It seems inappropriate that this code uses
free_page[s]_and_swap_cache().  It should go direct to put_page() and
release_pages()?  Please review this code's implicit decision to pass
"cold==0" into release_pages().

> -static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
> +static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)

I wonder if all the inlining which remains in this code is needed and
desirable.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 12/20] mm: Extended batches for generic mmu_gather

[Andrew Morton]

On Fri, 01 Apr 2011 14:13:10 +0200
Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:

> Instead of using a single batch (the small on-stack, or an allocated
> page), try and extend the batch every time it runs out and only flush
> once either the extend fails or we're done.

why?

>
> ...
>
> @@ -86,22 +86,48 @@ struct mmu_gather {
>  #ifdef CONFIG_HAVE_RCU_TABLE_FREE
>  	struct mmu_table_batch	*batch;
>  #endif
> -	unsigned int		nr;	/* set to ~0U means fast mode */
> -	unsigned int		max;	/* nr < max */
> -	unsigned int		need_flush;/* Really unmapped some ptes? */
> -	unsigned int		fullmm; /* non-zero means full mm flush */
> -	struct page		**pages;
> -	struct page		*local[MMU_GATHER_BUNDLE];
> +	unsigned int		need_flush : 1,	/* Did free PTEs */
> +				fast_mode  : 1; /* No batching   */

mmu_gather.fast_mode gets modified in several places apparently without
locking to protect itself.  I don't think that these modifications will
accidentally trash need_flush, mainly by luck.

Please review the concurrency issues here and document them clearly.

> +	unsigned int		fullmm;
> +
> +	struct mmu_gather_batch *active;
> +	struct mmu_gather_batch	local;
> +	struct page		*__pages[MMU_GATHER_BUNDLE];
>  };
>  
> -static inline void __tlb_alloc_page(struct mmu_gather *tlb)
> +/*
> + * For UP we don't need to worry about TLB flush
> + * and page free order so much..
> + */
> +#ifdef CONFIG_SMP
> +  #define tlb_fast_mode(tlb) (tlb->fast_mode)
> +#else
> +  #define tlb_fast_mode(tlb) 1
> +#endif

Mutter.

Could have been written in C.

Will cause a compile error with, for example, tlb_fast_mode(tlb + 1).

> +static inline int tlb_next_batch(struct mmu_gather *tlb)
>  {
> -	unsigned long addr = __get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
> +	struct mmu_gather_batch *batch;
>  
> -	if (addr) {
> -		tlb->pages = (void *)addr;
> -		tlb->max = PAGE_SIZE / sizeof(struct page *);
> +	batch = tlb->active;
> +	if (batch->next) {
> +		tlb->active = batch->next;
> +		return 1;
>  	}
> +
> +	batch = (void *)__get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);

A comment explaining the gfp_t decision would be useful.

> +	if (!batch)
> +		return 0;
> +
> +	batch->next = NULL;
> +	batch->nr   = 0;
> +	batch->max  = MAX_GATHER_BATCH;
> +
> +	tlb->active->next = batch;
> +	tlb->active = batch;
> +
> +	return 1;
>  }
>  
>  /* tlb_gather_mmu
> @@ -114,16 +140,13 @@ tlb_gather_mmu(struct mmu_gather *tlb, s
>  {
>  	tlb->mm = mm;
>  
> -	tlb->max = ARRAY_SIZE(tlb->local);
> -	tlb->pages = tlb->local;
> -
> -	if (num_online_cpus() > 1) {
> -		tlb->nr = 0;
> -		__tlb_alloc_page(tlb);
> -	} else /* Use fast mode if only one CPU is online */
> -		tlb->nr = ~0U;
> -
> -	tlb->fullmm = fullmm;
> +	tlb->fullmm     = fullmm;
> +	tlb->need_flush = 0;
> +	tlb->fast_mode  = (num_possible_cpus() == 1);

The changelog didn't tell us why we switched from num_online_cpus() to
num_possible_cpus().

> +	tlb->local.next = NULL;
> +	tlb->local.nr   = 0;
> +	tlb->local.max  = ARRAY_SIZE(tlb->__pages);
> +	tlb->active     = &tlb->local;
>  
>  #ifdef CONFIG_HAVE_RCU_TABLE_FREE
>  	tlb->batch = NULL;
>
> ...
>
> @@ -177,15 +205,24 @@ tlb_finish_mmu(struct mmu_gather *tlb, u
>   */
>  static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
>  {
> +	struct mmu_gather_batch *batch;
> +
>  	tlb->need_flush = 1;
> +
>  	if (tlb_fast_mode(tlb)) {
>  		free_page_and_swap_cache(page);
>  		return 1; /* avoid calling tlb_flush_mmu() */
>  	}
> -	tlb->pages[tlb->nr++] = page;
> -	VM_BUG_ON(tlb->nr > tlb->max);
>  
> -	return tlb->max - tlb->nr;
> +	batch = tlb->active;
> +	batch->pages[batch->nr++] = page;
> +	VM_BUG_ON(batch->nr > batch->max);
> +	if (batch->nr == batch->max) {
> +		if (!tlb_next_batch(tlb))
> +			return 0;
> +	}

Moving the VM_BUG_ON() down to after the if() would save a few cycles.

> +	return batch->max - batch->nr;
>  }
>  
>  /* tlb_remove_page
> 

Re: [PATCH 12/20] mm: Extended batches for generic mmu_gather

[Andrew Morton]

On Fri, 01 Apr 2011 14:13:10 +0200
Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:

> Instead of using a single batch (the small on-stack, or an allocated
> page), try and extend the batch every time it runs out and only flush
> once either the extend fails or we're done.

why?

>
> ...
>
> @@ -86,22 +86,48 @@ struct mmu_gather {
>  #ifdef CONFIG_HAVE_RCU_TABLE_FREE
>  	struct mmu_table_batch	*batch;
>  #endif
> -	unsigned int		nr;	/* set to ~0U means fast mode */
> -	unsigned int		max;	/* nr < max */
> -	unsigned int		need_flush;/* Really unmapped some ptes? */
> -	unsigned int		fullmm; /* non-zero means full mm flush */
> -	struct page		**pages;
> -	struct page		*local[MMU_GATHER_BUNDLE];
> +	unsigned int		need_flush : 1,	/* Did free PTEs */
> +				fast_mode  : 1; /* No batching   */

mmu_gather.fast_mode gets modified in several places apparently without
locking to protect itself.  I don't think that these modifications will
accidentally trash need_flush, mainly by luck.

Please review the concurrency issues here and document them clearly.

> +	unsigned int		fullmm;
> +
> +	struct mmu_gather_batch *active;
> +	struct mmu_gather_batch	local;
> +	struct page		*__pages[MMU_GATHER_BUNDLE];
>  };
>  
> -static inline void __tlb_alloc_page(struct mmu_gather *tlb)
> +/*
> + * For UP we don't need to worry about TLB flush
> + * and page free order so much..
> + */
> +#ifdef CONFIG_SMP
> +  #define tlb_fast_mode(tlb) (tlb->fast_mode)
> +#else
> +  #define tlb_fast_mode(tlb) 1
> +#endif

Mutter.

Could have been written in C.

Will cause a compile error with, for example, tlb_fast_mode(tlb + 1).

> +static inline int tlb_next_batch(struct mmu_gather *tlb)
>  {
> -	unsigned long addr = __get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
> +	struct mmu_gather_batch *batch;
>  
> -	if (addr) {
> -		tlb->pages = (void *)addr;
> -		tlb->max = PAGE_SIZE / sizeof(struct page *);
> +	batch = tlb->active;
> +	if (batch->next) {
> +		tlb->active = batch->next;
> +		return 1;
>  	}
> +
> +	batch = (void *)__get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);

A comment explaining the gfp_t decision would be useful.

> +	if (!batch)
> +		return 0;
> +
> +	batch->next = NULL;
> +	batch->nr   = 0;
> +	batch->max  = MAX_GATHER_BATCH;
> +
> +	tlb->active->next = batch;
> +	tlb->active = batch;
> +
> +	return 1;
>  }
>  
>  /* tlb_gather_mmu
> @@ -114,16 +140,13 @@ tlb_gather_mmu(struct mmu_gather *tlb, s
>  {
>  	tlb->mm = mm;
>  
> -	tlb->max = ARRAY_SIZE(tlb->local);
> -	tlb->pages = tlb->local;
> -
> -	if (num_online_cpus() > 1) {
> -		tlb->nr = 0;
> -		__tlb_alloc_page(tlb);
> -	} else /* Use fast mode if only one CPU is online */
> -		tlb->nr = ~0U;
> -
> -	tlb->fullmm = fullmm;
> +	tlb->fullmm     = fullmm;
> +	tlb->need_flush = 0;
> +	tlb->fast_mode  = (num_possible_cpus() == 1);

The changelog didn't tell us why we switched from num_online_cpus() to
num_possible_cpus().

> +	tlb->local.next = NULL;
> +	tlb->local.nr   = 0;
> +	tlb->local.max  = ARRAY_SIZE(tlb->__pages);
> +	tlb->active     = &tlb->local;
>  
>  #ifdef CONFIG_HAVE_RCU_TABLE_FREE
>  	tlb->batch = NULL;
>
> ...
>
> @@ -177,15 +205,24 @@ tlb_finish_mmu(struct mmu_gather *tlb, u
>   */
>  static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
>  {
> +	struct mmu_gather_batch *batch;
> +
>  	tlb->need_flush = 1;
> +
>  	if (tlb_fast_mode(tlb)) {
>  		free_page_and_swap_cache(page);
>  		return 1; /* avoid calling tlb_flush_mmu() */
>  	}
> -	tlb->pages[tlb->nr++] = page;
> -	VM_BUG_ON(tlb->nr > tlb->max);
>  
> -	return tlb->max - tlb->nr;
> +	batch = tlb->active;
> +	batch->pages[batch->nr++] = page;
> +	VM_BUG_ON(batch->nr > batch->max);
> +	if (batch->nr == batch->max) {
> +		if (!tlb_next_batch(tlb))
> +			return 0;
> +	}

Moving the VM_BUG_ON() down to after the if() would save a few cycles.

> +	return batch->max - batch->nr;
>  }
>  
>  /* tlb_remove_page
> 

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 13/20] lockdep, mutex: Provide mutex_lock_nest_lock

[Andrew Morton]

On Fri, 01 Apr 2011 14:13:11 +0200
Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:

> Provide the mutex_lock_nest_lock() annotation.

why?

Neither the changelog nor the code provide any documentation for this addition to
the lokdep API.

Re: [PATCH 13/20] lockdep, mutex: Provide mutex_lock_nest_lock

[Andrew Morton]

On Fri, 01 Apr 2011 14:13:11 +0200
Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:

> Provide the mutex_lock_nest_lock() annotation.

why?

Neither the changelog nor the code provide any documentation for this addition to
the lokdep API.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 14/20] mm: Remove i_mmap_lock lockbreak

[Andrew Morton]

On Fri, 01 Apr 2011 14:13:12 +0200
Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:

> Hugh says:
>  "The only significant loser, I think, would be page reclaim (when
>   concurrent with truncation): could spin for a long time waiting for
>   the i_mmap_mutex it expects would soon be dropped? "
> 
> Counter points:
>  - cpu contention makes the spin stop (need_resched())
>  - zap pages should be freeing pages at a higher rate than reclaim
>    ever can
> 
> I think the simplification of the truncate code is definately worth it.

Well, we don't need to guess.  These things are testable!

Re: [PATCH 14/20] mm: Remove i_mmap_lock lockbreak

[Andrew Morton]

On Fri, 01 Apr 2011 14:13:12 +0200
Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:

> Hugh says:
>  "The only significant loser, I think, would be page reclaim (when
>   concurrent with truncation): could spin for a long time waiting for
>   the i_mmap_mutex it expects would soon be dropped? "
> 
> Counter points:
>  - cpu contention makes the spin stop (need_resched())
>  - zap pages should be freeing pages at a higher rate than reclaim
>    ever can
> 
> I think the simplification of the truncate code is definately worth it.

Well, we don't need to guess.  These things are testable!

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 15/20] mm: Convert i_mmap_lock to a mutex

[Andrew Morton]

On Fri, 01 Apr 2011 14:13:13 +0200
Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:

> Straight fwd conversion of i_mmap_lock to a mutex

What effect does this have on kernel performance?

Workloads which take the lock at high frequency from multiple threads
should be tested.

Re: [PATCH 15/20] mm: Convert i_mmap_lock to a mutex

[Andrew Morton]

On Fri, 01 Apr 2011 14:13:13 +0200
Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:

> Straight fwd conversion of i_mmap_lock to a mutex

What effect does this have on kernel performance?

Workloads which take the lock at high frequency from multiple threads
should be tested.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 19/20] mm: Convert anon_vma->lock to a mutex

[Andrew Morton]

On Fri, 01 Apr 2011 14:13:17 +0200
Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:

> Straight fwd conversion of anon_vma->lock to a mutex.

What workloads do we expect might be adversely affected by this? 
Were such workloads tested?  With what results?

Re: [PATCH 19/20] mm: Convert anon_vma->lock to a mutex

[Andrew Morton]

On Fri, 01 Apr 2011 14:13:17 +0200
Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:

> Straight fwd conversion of anon_vma->lock to a mutex.

What workloads do we expect might be adversely affected by this? 
Were such workloads tested?  With what results?



--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 20/20] mm: Optimize page_lock_anon_vma() fast-path

[Andrew Morton]

On Fri, 01 Apr 2011 14:13:18 +0200
Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:

> Optimize the page_lock_anon_vma() fast path to be one atomic op,
> instead of two.
> 
> Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
> Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
> LKML-Reference: <new-submission>
> ---
>  mm/rmap.c |   86 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++---
>  1 file changed, 82 insertions(+), 4 deletions(-)
> 
> Index: linux-2.6/mm/rmap.c
> ===================================================================
> --- linux-2.6.orig/mm/rmap.c
> +++ linux-2.6/mm/rmap.c
> @@ -85,6 +85,29 @@ static inline struct anon_vma *anon_vma_
>  static inline void anon_vma_free(struct anon_vma *anon_vma)
>  {
>  	VM_BUG_ON(atomic_read(&anon_vma->refcount));
> +
> +	/*
> +	 * Synchronize against page_lock_anon_vma() such that
> +	 * we can safely hold the lock without the anon_vma getting
> +	 * freed.
> +	 *
> +	 * Relies on the full mb implied by the atomic_dec_and_test() from
> +	 * put_anon_vma() against the acquire barrier implied by
> +	 * mutex_trylock() from page_lock_anon_vma(). This orders:
> +	 *
> +	 * page_lock_anon_vma()		VS	put_anon_vma()
> +	 *   mutex_trylock()			  atomic_dec_and_test()
> +	 *   LOCK				  MB
> +	 *   atomic_read()			  mutex_is_locked()
> +	 *
> +	 * LOCK should suffice since the actual taking of the lock must
> +	 * happen _before_ what follows.
> +	 */
> +	if (mutex_is_locked(&anon_vma->root->mutex)) {
> +		anon_vma_lock(anon_vma);
> +		anon_vma_unlock(anon_vma);
> +	}
> +
>  	kmem_cache_free(anon_vma_cachep, anon_vma);
>  }

Did we need to include all this stuff in uniprocessor builds?

It would be neater to add a new anon_vma_is_locked().

This code is too tricksy to deserve life :(

> @@ -371,20 +394,75 @@ struct anon_vma *page_get_anon_vma(struc
>  	return anon_vma;
>  }
>  
> +/*
> + * Similar to page_get_anon_vma() except it locks the anon_vma.
> + *
> + * Its a little more complex as it tries to keep the fast path to a single
> + * atomic op -- the trylock. If we fail the trylock, we fall back to getting a
> + * reference like with page_get_anon_vma() and then block on the mutex.
> + */
>  struct anon_vma *page_lock_anon_vma(struct page *page)
>  {
> -	struct anon_vma *anon_vma = page_get_anon_vma(page);
> +	struct anon_vma *anon_vma = NULL;
> +	unsigned long anon_mapping;
>  
> -	if (anon_vma)
> -		anon_vma_lock(anon_vma);
> +	rcu_read_lock();
> +	anon_mapping = (unsigned long) ACCESS_ONCE(page->mapping);
> +	if ((anon_mapping & PAGE_MAPPING_FLAGS) != PAGE_MAPPING_ANON)
> +		goto out;

Why?  Needs a comment.

> +	if (!page_mapped(page))
> +		goto out;

Why?  How can this come about? Needs a comment.

> +
> +	anon_vma = (struct anon_vma *) (anon_mapping - PAGE_MAPPING_ANON);
> +	if (mutex_trylock(&anon_vma->root->mutex)) {

anon_vma_trylock()?

Or just remove all the wrapper functions and open-code all the locking.
These tricks all seem pretty tied-up with the mutex implementation
anyway.

> +		/*
> +		 * If we observe a !0 refcount, then holding the lock ensures
> +		 * the anon_vma will not go away, see __put_anon_vma().
> +		 */
> +		if (!atomic_read(&anon_vma->refcount)) {
> +			anon_vma_unlock(anon_vma);
> +			anon_vma = NULL;
> +		}
> +		goto out;
> +	}
> +
> +	/* trylock failed, we got to sleep */
> +	if (!atomic_inc_not_zero(&anon_vma->refcount)) {
> +		anon_vma = NULL;
> +		goto out;
> +	}
>  
> +	if (!page_mapped(page)) {
> +		put_anon_vma(anon_vma);
> +		anon_vma = NULL;
> +		goto out;
> +	}

Also quite opaque, needs decent commentary.

I'd have expected this test to occur after the lock was acquired.

> +	/* we pinned the anon_vma, its safe to sleep */
> +	rcu_read_unlock();
> +	anon_vma_lock(anon_vma);
> +
> +	if (atomic_dec_and_test(&anon_vma->refcount)) {
> +		/*
> +		 * Oops, we held the last refcount, release the lock
> +		 * and bail -- can't simply use put_anon_vma() because
> +		 * we'll deadlock on the anon_vma_lock() recursion.
> +		 */
> +		anon_vma_unlock(anon_vma);
> +		__put_anon_vma(anon_vma);
> +		anon_vma = NULL;
> +	}
> +
> +	return anon_vma;
> +
> +out:
> +	rcu_read_unlock();
>  	return anon_vma;
>  }
>  
>  void page_unlock_anon_vma(struct anon_vma *anon_vma)
>  {
>  	anon_vma_unlock(anon_vma);
> -	put_anon_vma(anon_vma);
>  }

Geeze, I hope this patch is worth it :( :(

Re: [PATCH 20/20] mm: Optimize page_lock_anon_vma() fast-path

[Andrew Morton]

On Fri, 01 Apr 2011 14:13:18 +0200
Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:

> Optimize the page_lock_anon_vma() fast path to be one atomic op,
> instead of two.
> 
> Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
> Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
> LKML-Reference: <new-submission>
> ---
>  mm/rmap.c |   86 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++---
>  1 file changed, 82 insertions(+), 4 deletions(-)
> 
> Index: linux-2.6/mm/rmap.c
> ===================================================================
> --- linux-2.6.orig/mm/rmap.c
> +++ linux-2.6/mm/rmap.c
> @@ -85,6 +85,29 @@ static inline struct anon_vma *anon_vma_
>  static inline void anon_vma_free(struct anon_vma *anon_vma)
>  {
>  	VM_BUG_ON(atomic_read(&anon_vma->refcount));
> +
> +	/*
> +	 * Synchronize against page_lock_anon_vma() such that
> +	 * we can safely hold the lock without the anon_vma getting
> +	 * freed.
> +	 *
> +	 * Relies on the full mb implied by the atomic_dec_and_test() from
> +	 * put_anon_vma() against the acquire barrier implied by
> +	 * mutex_trylock() from page_lock_anon_vma(). This orders:
> +	 *
> +	 * page_lock_anon_vma()		VS	put_anon_vma()
> +	 *   mutex_trylock()			  atomic_dec_and_test()
> +	 *   LOCK				  MB
> +	 *   atomic_read()			  mutex_is_locked()
> +	 *
> +	 * LOCK should suffice since the actual taking of the lock must
> +	 * happen _before_ what follows.
> +	 */
> +	if (mutex_is_locked(&anon_vma->root->mutex)) {
> +		anon_vma_lock(anon_vma);
> +		anon_vma_unlock(anon_vma);
> +	}
> +
>  	kmem_cache_free(anon_vma_cachep, anon_vma);
>  }

Did we need to include all this stuff in uniprocessor builds?

It would be neater to add a new anon_vma_is_locked().

This code is too tricksy to deserve life :(

> @@ -371,20 +394,75 @@ struct anon_vma *page_get_anon_vma(struc
>  	return anon_vma;
>  }
>  
> +/*
> + * Similar to page_get_anon_vma() except it locks the anon_vma.
> + *
> + * Its a little more complex as it tries to keep the fast path to a single
> + * atomic op -- the trylock. If we fail the trylock, we fall back to getting a
> + * reference like with page_get_anon_vma() and then block on the mutex.
> + */
>  struct anon_vma *page_lock_anon_vma(struct page *page)
>  {
> -	struct anon_vma *anon_vma = page_get_anon_vma(page);
> +	struct anon_vma *anon_vma = NULL;
> +	unsigned long anon_mapping;
>  
> -	if (anon_vma)
> -		anon_vma_lock(anon_vma);
> +	rcu_read_lock();
> +	anon_mapping = (unsigned long) ACCESS_ONCE(page->mapping);
> +	if ((anon_mapping & PAGE_MAPPING_FLAGS) != PAGE_MAPPING_ANON)
> +		goto out;

Why?  Needs a comment.

> +	if (!page_mapped(page))
> +		goto out;

Why?  How can this come about? Needs a comment.

> +
> +	anon_vma = (struct anon_vma *) (anon_mapping - PAGE_MAPPING_ANON);
> +	if (mutex_trylock(&anon_vma->root->mutex)) {

anon_vma_trylock()?

Or just remove all the wrapper functions and open-code all the locking.
These tricks all seem pretty tied-up with the mutex implementation
anyway.

> +		/*
> +		 * If we observe a !0 refcount, then holding the lock ensures
> +		 * the anon_vma will not go away, see __put_anon_vma().
> +		 */
> +		if (!atomic_read(&anon_vma->refcount)) {
> +			anon_vma_unlock(anon_vma);
> +			anon_vma = NULL;
> +		}
> +		goto out;
> +	}
> +
> +	/* trylock failed, we got to sleep */
> +	if (!atomic_inc_not_zero(&anon_vma->refcount)) {
> +		anon_vma = NULL;
> +		goto out;
> +	}
>  
> +	if (!page_mapped(page)) {
> +		put_anon_vma(anon_vma);
> +		anon_vma = NULL;
> +		goto out;
> +	}

Also quite opaque, needs decent commentary.

I'd have expected this test to occur after the lock was acquired.

> +	/* we pinned the anon_vma, its safe to sleep */
> +	rcu_read_unlock();
> +	anon_vma_lock(anon_vma);
> +
> +	if (atomic_dec_and_test(&anon_vma->refcount)) {
> +		/*
> +		 * Oops, we held the last refcount, release the lock
> +		 * and bail -- can't simply use put_anon_vma() because
> +		 * we'll deadlock on the anon_vma_lock() recursion.
> +		 */
> +		anon_vma_unlock(anon_vma);
> +		__put_anon_vma(anon_vma);
> +		anon_vma = NULL;
> +	}
> +
> +	return anon_vma;
> +
> +out:
> +	rcu_read_unlock();
>  	return anon_vma;
>  }
>  
>  void page_unlock_anon_vma(struct anon_vma *anon_vma)
>  {
>  	anon_vma_unlock(anon_vma);
> -	put_anon_vma(anon_vma);
>  }

Geeze, I hope this patch is worth it :( :(

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 01/20] mm: mmu_gather rework

[Peter Zijlstra]

On Tue, 2011-04-19 at 13:06 -0700, Andrew Morton wrote:
> On Fri, 01 Apr 2011 14:12:59 +0200
> Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:
> 
> > Remove the first obstackle towards a fully preemptible mmu_gather.
> > 
> > The current scheme assumes mmu_gather is always done with preemption
> > disabled and uses per-cpu storage for the page batches. Change this to
> > try and allocate a page for batching and in case of failure, use a
> > small on-stack array to make some progress.
> > 
> > Preemptible mmu_gather is desired in general and usable once
> > i_mmap_lock becomes a mutex. Doing it before the mutex conversion
> > saves us from having to rework the code by moving the mmu_gather
> > bits inside the pte_lock.
> > 
> > Also avoid flushing the tlb batches from under the pte lock,
> > this is useful even without the i_mmap_lock conversion as it
> > significantly reduces pte lock hold times.
> 
> There doesn't seem much point in reviewing this closely, as a lot of it
> gets tossed away later in the series..

That's a result of breaking patches along concept boundaries :/

> >  		free_pages_and_swap_cache(tlb->pages, tlb->nr);
> 
> It seems inappropriate that this code uses
> free_page[s]_and_swap_cache().  It should go direct to put_page() and
> release_pages()?  Please review this code's implicit decision to pass
> "cold==0" into release_pages().

Well, that isn't new with this patch, however it does look to be
correct. We're freeing user pages, those could indeed still be part of
the swapcache. Furthermore, the PAGEVEC_SIZE split in
free_pages_and_swap_cache() alone makes it worth calling that over
release_pages().

As to the cold==0, I think that too is correct since we don't actually
touch the pages themselves and we have no inkling as to their cache
state, we're simply wiping out user pages.

> > -static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
> > +static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
> 
> I wonder if all the inlining which remains in this code is needed and
> desirable.

Probably not, the big plan was to make everybody use the generic code
and then move it into mm/memory.c or so.

But I guess I can have asm-generic/tlb.h define HAVE_GENERIC_MMU_GATHER
and make the compilation in mm/memory.c conditional on that (or generate
lots of Kconfig churn).

Re: [PATCH 01/20] mm: mmu_gather rework

[Peter Zijlstra]

On Tue, 2011-04-19 at 13:06 -0700, Andrew Morton wrote:
> On Fri, 01 Apr 2011 14:12:59 +0200
> Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:
> 
> > Remove the first obstackle towards a fully preemptible mmu_gather.
> > 
> > The current scheme assumes mmu_gather is always done with preemption
> > disabled and uses per-cpu storage for the page batches. Change this to
> > try and allocate a page for batching and in case of failure, use a
> > small on-stack array to make some progress.
> > 
> > Preemptible mmu_gather is desired in general and usable once
> > i_mmap_lock becomes a mutex. Doing it before the mutex conversion
> > saves us from having to rework the code by moving the mmu_gather
> > bits inside the pte_lock.
> > 
> > Also avoid flushing the tlb batches from under the pte lock,
> > this is useful even without the i_mmap_lock conversion as it
> > significantly reduces pte lock hold times.
> 
> There doesn't seem much point in reviewing this closely, as a lot of it
> gets tossed away later in the series..

That's a result of breaking patches along concept boundaries :/

> >  		free_pages_and_swap_cache(tlb->pages, tlb->nr);
> 
> It seems inappropriate that this code uses
> free_page[s]_and_swap_cache().  It should go direct to put_page() and
> release_pages()?  Please review this code's implicit decision to pass
> "cold==0" into release_pages().

Well, that isn't new with this patch, however it does look to be
correct. We're freeing user pages, those could indeed still be part of
the swapcache. Furthermore, the PAGEVEC_SIZE split in
free_pages_and_swap_cache() alone makes it worth calling that over
release_pages().

As to the cold==0, I think that too is correct since we don't actually
touch the pages themselves and we have no inkling as to their cache
state, we're simply wiping out user pages.

> > -static inline void tlb_remove_page(struct mmu_gather *tlb, struct page *page)
> > +static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
> 
> I wonder if all the inlining which remains in this code is needed and
> desirable.

Probably not, the big plan was to make everybody use the generic code
and then move it into mm/memory.c or so.

But I guess I can have asm-generic/tlb.h define HAVE_GENERIC_MMU_GATHER
and make the compilation in mm/memory.c conditional on that (or generate
lots of Kconfig churn).

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 01/20] mm: mmu_gather rework

[Peter Zijlstra]

On Wed, 2011-04-20 at 10:47 +0200, Peter Zijlstra wrote:
> 
> But I guess I can have asm-generic/tlb.h define HAVE_GENERIC_MMU_GATHER
> and make the compilation in mm/memory.c conditional on that (or generate
> lots of Kconfig churn). 

Something like so:

---
Subject: mm: Uninline large generic tlb.h functions
From: Peter Zijlstra <a.p.zijlstra@chello.nl>
Date: Wed Apr 20 11:04:42 CEST 2011

Some of these functions have grown beyond inline sanity, move them
out-of-line.

Requested-by: Andrew Morton <akpm@linux-foundation.org>
Requested-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
---
 include/asm-generic/tlb.h |  135 ++++------------------------------------------
 mm/memory.c               |  124 +++++++++++++++++++++++++++++++++++++++++-
 2 files changed, 135 insertions(+), 124 deletions(-)

Index: linux-2.6/include/asm-generic/tlb.h
===================================================================
--- linux-2.6.orig/include/asm-generic/tlb.h
+++ linux-2.6/include/asm-generic/tlb.h
@@ -96,134 +96,25 @@ struct mmu_gather {
 	struct page		*__pages[MMU_GATHER_BUNDLE];
 };
 
-/*
- * For UP we don't need to worry about TLB flush
- * and page free order so much..
- */
-#ifdef CONFIG_SMP
-  #define tlb_fast_mode(tlb) (tlb->fast_mode)
-#else
-  #define tlb_fast_mode(tlb) 1
-#endif
+#define HAVE_GENERIC_MMU_GATHER
 
-static inline int tlb_next_batch(struct mmu_gather *tlb)
+static inline int tlb_fast_mode(struct mmu_gather *tlb)
 {
-	struct mmu_gather_batch *batch;
-
-	batch = tlb->active;
-	if (batch->next) {
-		tlb->active = batch->next;
-		return 1;
-	}
-
-	batch = (void *)__get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
-	if (!batch)
-		return 0;
-
-	batch->next = NULL;
-	batch->nr   = 0;
-	batch->max  = MAX_GATHER_BATCH;
-
-	tlb->active->next = batch;
-	tlb->active = batch;
-
+#ifdef CONFIG_SMP
+	return tlb->fast_mode;
+#else
+	/*
+	 * For UP we don't need to worry about TLB flush
+	 * and page free order so much..
+	 */
 	return 1;
-}
-
-/* tlb_gather_mmu
- *	Called to initialize an (on-stack) mmu_gather structure for page-table
- *	tear-down from @mm. The @fullmm argument is used when @mm is without
- *	users and we're going to destroy the full address space (exit/execve).
- */
-static inline void
-tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, bool fullmm)
-{
-	tlb->mm = mm;
-
-	tlb->fullmm     = fullmm;
-	tlb->need_flush = 0;
-	tlb->fast_mode  = (num_possible_cpus() == 1);
-	tlb->local.next = NULL;
-	tlb->local.nr   = 0;
-	tlb->local.max  = ARRAY_SIZE(tlb->__pages);
-	tlb->active     = &tlb->local;
-
-#ifdef CONFIG_HAVE_RCU_TABLE_FREE
-	tlb->batch = NULL;
 #endif
 }
 
-static inline void
-tlb_flush_mmu(struct mmu_gather *tlb)
-{
-	struct mmu_gather_batch *batch;
-
-	if (!tlb->need_flush)
-		return;
-	tlb->need_flush = 0;
-	tlb_flush(tlb);
-#ifdef CONFIG_HAVE_RCU_TABLE_FREE
-	tlb_table_flush(tlb);
-#endif
-
-	if (tlb_fast_mode(tlb))
-		return;
-
-	for (batch = &tlb->local; batch; batch = batch->next) {
-		free_pages_and_swap_cache(batch->pages, batch->nr);
-		batch->nr = 0;
-	}
-	tlb->active = &tlb->local;
-}
-
-/* tlb_finish_mmu
- *	Called at the end of the shootdown operation to free up any resources
- *	that were required.
- */
-static inline void
-tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
-{
-	struct mmu_gather_batch *batch, *next;
-
-	tlb_flush_mmu(tlb);
-
-	/* keep the page table cache within bounds */
-	check_pgt_cache();
-
-	for (batch = tlb->local.next; batch; batch = next) {
-		next = batch->next;
-		free_pages((unsigned long)batch, 0);
-	}
-	tlb->local.next = NULL;
-}
-
-/* __tlb_remove_page
- *	Must perform the equivalent to __free_pte(pte_get_and_clear(ptep)), while
- *	handling the additional races in SMP caused by other CPUs caching valid
- *	mappings in their TLBs. Returns the number of free page slots left.
- *	When out of page slots we must call tlb_flush_mmu().
- */
-static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
-{
-	struct mmu_gather_batch *batch;
-
-	tlb->need_flush = 1;
-
-	if (tlb_fast_mode(tlb)) {
-		free_page_and_swap_cache(page);
-		return 1; /* avoid calling tlb_flush_mmu() */
-	}
-
-	batch = tlb->active;
-	batch->pages[batch->nr++] = page;
-	VM_BUG_ON(batch->nr > batch->max);
-	if (batch->nr == batch->max) {
-		if (!tlb_next_batch(tlb))
-			return 0;
-	}
-
-	return batch->max - batch->nr;
-}
+void tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, bool fullmm);
+void tlb_flush_mmu(struct mmu_gather *tlb);
+void tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end);
+int __tlb_remove_page(struct mmu_gather *tlb, struct page *page);
 
 /* tlb_remove_page
  *	Similar to __tlb_remove_page but will call tlb_flush_mmu() itself when
Index: linux-2.6/mm/memory.c
===================================================================
--- linux-2.6.orig/mm/memory.c
+++ linux-2.6/mm/memory.c
@@ -182,7 +182,7 @@ void sync_mm_rss(struct task_struct *tas
 {
 	__sync_task_rss_stat(task, mm);
 }
-#else
+#else /* SPLIT_RSS_COUNTING */
 
 #define inc_mm_counter_fast(mm, member) inc_mm_counter(mm, member)
 #define dec_mm_counter_fast(mm, member) dec_mm_counter(mm, member)
@@ -191,7 +191,127 @@ static void check_sync_rss_stat(struct t
 {
 }
 
+#endif /* SPLIT_RSS_COUNTING */
+
+#ifdef HAVE_GENERIC_MMU_GATHER
+
+static int tlb_next_batch(struct mmu_gather *tlb)
+{
+	struct mmu_gather_batch *batch;
+
+	batch = tlb->active;
+	if (batch->next) {
+		tlb->active = batch->next;
+		return 1;
+	}
+
+	batch = (void *)__get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
+	if (!batch)
+		return 0;
+
+	batch->next = NULL;
+	batch->nr   = 0;
+	batch->max  = MAX_GATHER_BATCH;
+
+	tlb->active->next = batch;
+	tlb->active = batch;
+
+	return 1;
+}
+
+/* tlb_gather_mmu
+ *	Called to initialize an (on-stack) mmu_gather structure for page-table
+ *	tear-down from @mm. The @fullmm argument is used when @mm is without
+ *	users and we're going to destroy the full address space (exit/execve).
+ */
+void tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, bool fullmm)
+{
+	tlb->mm = mm;
+
+	tlb->fullmm     = fullmm;
+	tlb->need_flush = 0;
+	tlb->fast_mode  = (num_possible_cpus() == 1);
+	tlb->local.next = NULL;
+	tlb->local.nr   = 0;
+	tlb->local.max  = ARRAY_SIZE(tlb->__pages);
+	tlb->active     = &tlb->local;
+
+#ifdef CONFIG_HAVE_RCU_TABLE_FREE
+	tlb->batch = NULL;
 #endif
+}
+
+void tlb_flush_mmu(struct mmu_gather *tlb)
+{
+	struct mmu_gather_batch *batch;
+
+	if (!tlb->need_flush)
+		return;
+	tlb->need_flush = 0;
+	tlb_flush(tlb);
+#ifdef CONFIG_HAVE_RCU_TABLE_FREE
+	tlb_table_flush(tlb);
+#endif
+
+	if (tlb_fast_mode(tlb))
+		return;
+
+	for (batch = &tlb->local; batch; batch = batch->next) {
+		free_pages_and_swap_cache(batch->pages, batch->nr);
+		batch->nr = 0;
+	}
+	tlb->active = &tlb->local;
+}
+
+/* tlb_finish_mmu
+ *	Called at the end of the shootdown operation to free up any resources
+ *	that were required.
+ */
+void tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
+{
+	struct mmu_gather_batch *batch, *next;
+
+	tlb_flush_mmu(tlb);
+
+	/* keep the page table cache within bounds */
+	check_pgt_cache();
+
+	for (batch = tlb->local.next; batch; batch = next) {
+		next = batch->next;
+		free_pages((unsigned long)batch, 0);
+	}
+	tlb->local.next = NULL;
+}
+
+/* __tlb_remove_page
+ *	Must perform the equivalent to __free_pte(pte_get_and_clear(ptep)), while
+ *	handling the additional races in SMP caused by other CPUs caching valid
+ *	mappings in their TLBs. Returns the number of free page slots left.
+ *	When out of page slots we must call tlb_flush_mmu().
+ */
+int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	struct mmu_gather_batch *batch;
+
+	tlb->need_flush = 1;
+
+	if (tlb_fast_mode(tlb)) {
+		free_page_and_swap_cache(page);
+		return 1; /* avoid calling tlb_flush_mmu() */
+	}
+
+	batch = tlb->active;
+	batch->pages[batch->nr++] = page;
+	if (batch->nr == batch->max) {
+		if (!tlb_next_batch(tlb))
+			return 0;
+	}
+	VM_BUG_ON(batch->nr > batch->max);
+
+	return batch->max - batch->nr;
+}
+
+#endif /* HAVE_GENERIC_MMU_GATHER */
 
 #ifdef CONFIG_HAVE_RCU_TABLE_FREE
 
@@ -268,7 +388,7 @@ void tlb_remove_table(struct mmu_gather 
 		tlb_table_flush(tlb);
 }
 
-#endif
+#endif /* CONFIG_HAVE_RCU_TABLE_FREE */
 
 /*
  * If a p?d_bad entry is found while walking page tables, report

Re: [PATCH 01/20] mm: mmu_gather rework

[Peter Zijlstra]

On Wed, 2011-04-20 at 10:47 +0200, Peter Zijlstra wrote:
> 
> But I guess I can have asm-generic/tlb.h define HAVE_GENERIC_MMU_GATHER
> and make the compilation in mm/memory.c conditional on that (or generate
> lots of Kconfig churn). 

Something like so:

---
Subject: mm: Uninline large generic tlb.h functions
From: Peter Zijlstra <a.p.zijlstra@chello.nl>
Date: Wed Apr 20 11:04:42 CEST 2011

Some of these functions have grown beyond inline sanity, move them
out-of-line.

Requested-by: Andrew Morton <akpm@linux-foundation.org>
Requested-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
---
 include/asm-generic/tlb.h |  135 ++++------------------------------------------
 mm/memory.c               |  124 +++++++++++++++++++++++++++++++++++++++++-
 2 files changed, 135 insertions(+), 124 deletions(-)

Index: linux-2.6/include/asm-generic/tlb.h
===================================================================
--- linux-2.6.orig/include/asm-generic/tlb.h
+++ linux-2.6/include/asm-generic/tlb.h
@@ -96,134 +96,25 @@ struct mmu_gather {
 	struct page		*__pages[MMU_GATHER_BUNDLE];
 };
 
-/*
- * For UP we don't need to worry about TLB flush
- * and page free order so much..
- */
-#ifdef CONFIG_SMP
-  #define tlb_fast_mode(tlb) (tlb->fast_mode)
-#else
-  #define tlb_fast_mode(tlb) 1
-#endif
+#define HAVE_GENERIC_MMU_GATHER
 
-static inline int tlb_next_batch(struct mmu_gather *tlb)
+static inline int tlb_fast_mode(struct mmu_gather *tlb)
 {
-	struct mmu_gather_batch *batch;
-
-	batch = tlb->active;
-	if (batch->next) {
-		tlb->active = batch->next;
-		return 1;
-	}
-
-	batch = (void *)__get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
-	if (!batch)
-		return 0;
-
-	batch->next = NULL;
-	batch->nr   = 0;
-	batch->max  = MAX_GATHER_BATCH;
-
-	tlb->active->next = batch;
-	tlb->active = batch;
-
+#ifdef CONFIG_SMP
+	return tlb->fast_mode;
+#else
+	/*
+	 * For UP we don't need to worry about TLB flush
+	 * and page free order so much..
+	 */
 	return 1;
-}
-
-/* tlb_gather_mmu
- *	Called to initialize an (on-stack) mmu_gather structure for page-table
- *	tear-down from @mm. The @fullmm argument is used when @mm is without
- *	users and we're going to destroy the full address space (exit/execve).
- */
-static inline void
-tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, bool fullmm)
-{
-	tlb->mm = mm;
-
-	tlb->fullmm     = fullmm;
-	tlb->need_flush = 0;
-	tlb->fast_mode  = (num_possible_cpus() == 1);
-	tlb->local.next = NULL;
-	tlb->local.nr   = 0;
-	tlb->local.max  = ARRAY_SIZE(tlb->__pages);
-	tlb->active     = &tlb->local;
-
-#ifdef CONFIG_HAVE_RCU_TABLE_FREE
-	tlb->batch = NULL;
 #endif
 }
 
-static inline void
-tlb_flush_mmu(struct mmu_gather *tlb)
-{
-	struct mmu_gather_batch *batch;
-
-	if (!tlb->need_flush)
-		return;
-	tlb->need_flush = 0;
-	tlb_flush(tlb);
-#ifdef CONFIG_HAVE_RCU_TABLE_FREE
-	tlb_table_flush(tlb);
-#endif
-
-	if (tlb_fast_mode(tlb))
-		return;
-
-	for (batch = &tlb->local; batch; batch = batch->next) {
-		free_pages_and_swap_cache(batch->pages, batch->nr);
-		batch->nr = 0;
-	}
-	tlb->active = &tlb->local;
-}
-
-/* tlb_finish_mmu
- *	Called at the end of the shootdown operation to free up any resources
- *	that were required.
- */
-static inline void
-tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
-{
-	struct mmu_gather_batch *batch, *next;
-
-	tlb_flush_mmu(tlb);
-
-	/* keep the page table cache within bounds */
-	check_pgt_cache();
-
-	for (batch = tlb->local.next; batch; batch = next) {
-		next = batch->next;
-		free_pages((unsigned long)batch, 0);
-	}
-	tlb->local.next = NULL;
-}
-
-/* __tlb_remove_page
- *	Must perform the equivalent to __free_pte(pte_get_and_clear(ptep)), while
- *	handling the additional races in SMP caused by other CPUs caching valid
- *	mappings in their TLBs. Returns the number of free page slots left.
- *	When out of page slots we must call tlb_flush_mmu().
- */
-static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
-{
-	struct mmu_gather_batch *batch;
-
-	tlb->need_flush = 1;
-
-	if (tlb_fast_mode(tlb)) {
-		free_page_and_swap_cache(page);
-		return 1; /* avoid calling tlb_flush_mmu() */
-	}
-
-	batch = tlb->active;
-	batch->pages[batch->nr++] = page;
-	VM_BUG_ON(batch->nr > batch->max);
-	if (batch->nr == batch->max) {
-		if (!tlb_next_batch(tlb))
-			return 0;
-	}
-
-	return batch->max - batch->nr;
-}
+void tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, bool fullmm);
+void tlb_flush_mmu(struct mmu_gather *tlb);
+void tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end);
+int __tlb_remove_page(struct mmu_gather *tlb, struct page *page);
 
 /* tlb_remove_page
  *	Similar to __tlb_remove_page but will call tlb_flush_mmu() itself when
Index: linux-2.6/mm/memory.c
===================================================================
--- linux-2.6.orig/mm/memory.c
+++ linux-2.6/mm/memory.c
@@ -182,7 +182,7 @@ void sync_mm_rss(struct task_struct *tas
 {
 	__sync_task_rss_stat(task, mm);
 }
-#else
+#else /* SPLIT_RSS_COUNTING */
 
 #define inc_mm_counter_fast(mm, member) inc_mm_counter(mm, member)
 #define dec_mm_counter_fast(mm, member) dec_mm_counter(mm, member)
@@ -191,7 +191,127 @@ static void check_sync_rss_stat(struct t
 {
 }
 
+#endif /* SPLIT_RSS_COUNTING */
+
+#ifdef HAVE_GENERIC_MMU_GATHER
+
+static int tlb_next_batch(struct mmu_gather *tlb)
+{
+	struct mmu_gather_batch *batch;
+
+	batch = tlb->active;
+	if (batch->next) {
+		tlb->active = batch->next;
+		return 1;
+	}
+
+	batch = (void *)__get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
+	if (!batch)
+		return 0;
+
+	batch->next = NULL;
+	batch->nr   = 0;
+	batch->max  = MAX_GATHER_BATCH;
+
+	tlb->active->next = batch;
+	tlb->active = batch;
+
+	return 1;
+}
+
+/* tlb_gather_mmu
+ *	Called to initialize an (on-stack) mmu_gather structure for page-table
+ *	tear-down from @mm. The @fullmm argument is used when @mm is without
+ *	users and we're going to destroy the full address space (exit/execve).
+ */
+void tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, bool fullmm)
+{
+	tlb->mm = mm;
+
+	tlb->fullmm     = fullmm;
+	tlb->need_flush = 0;
+	tlb->fast_mode  = (num_possible_cpus() == 1);
+	tlb->local.next = NULL;
+	tlb->local.nr   = 0;
+	tlb->local.max  = ARRAY_SIZE(tlb->__pages);
+	tlb->active     = &tlb->local;
+
+#ifdef CONFIG_HAVE_RCU_TABLE_FREE
+	tlb->batch = NULL;
 #endif
+}
+
+void tlb_flush_mmu(struct mmu_gather *tlb)
+{
+	struct mmu_gather_batch *batch;
+
+	if (!tlb->need_flush)
+		return;
+	tlb->need_flush = 0;
+	tlb_flush(tlb);
+#ifdef CONFIG_HAVE_RCU_TABLE_FREE
+	tlb_table_flush(tlb);
+#endif
+
+	if (tlb_fast_mode(tlb))
+		return;
+
+	for (batch = &tlb->local; batch; batch = batch->next) {
+		free_pages_and_swap_cache(batch->pages, batch->nr);
+		batch->nr = 0;
+	}
+	tlb->active = &tlb->local;
+}
+
+/* tlb_finish_mmu
+ *	Called at the end of the shootdown operation to free up any resources
+ *	that were required.
+ */
+void tlb_finish_mmu(struct mmu_gather *tlb, unsigned long start, unsigned long end)
+{
+	struct mmu_gather_batch *batch, *next;
+
+	tlb_flush_mmu(tlb);
+
+	/* keep the page table cache within bounds */
+	check_pgt_cache();
+
+	for (batch = tlb->local.next; batch; batch = next) {
+		next = batch->next;
+		free_pages((unsigned long)batch, 0);
+	}
+	tlb->local.next = NULL;
+}
+
+/* __tlb_remove_page
+ *	Must perform the equivalent to __free_pte(pte_get_and_clear(ptep)), while
+ *	handling the additional races in SMP caused by other CPUs caching valid
+ *	mappings in their TLBs. Returns the number of free page slots left.
+ *	When out of page slots we must call tlb_flush_mmu().
+ */
+int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
+{
+	struct mmu_gather_batch *batch;
+
+	tlb->need_flush = 1;
+
+	if (tlb_fast_mode(tlb)) {
+		free_page_and_swap_cache(page);
+		return 1; /* avoid calling tlb_flush_mmu() */
+	}
+
+	batch = tlb->active;
+	batch->pages[batch->nr++] = page;
+	if (batch->nr == batch->max) {
+		if (!tlb_next_batch(tlb))
+			return 0;
+	}
+	VM_BUG_ON(batch->nr > batch->max);
+
+	return batch->max - batch->nr;
+}
+
+#endif /* HAVE_GENERIC_MMU_GATHER */
 
 #ifdef CONFIG_HAVE_RCU_TABLE_FREE
 
@@ -268,7 +388,7 @@ void tlb_remove_table(struct mmu_gather 
 		tlb_table_flush(tlb);
 }
 
-#endif
+#endif /* CONFIG_HAVE_RCU_TABLE_FREE */
 
 /*
  * If a p?d_bad entry is found while walking page tables, report

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 12/20] mm: Extended batches for generic mmu_gather

[Peter Zijlstra]

On Tue, 2011-04-19 at 13:06 -0700, Andrew Morton wrote:
> On Fri, 01 Apr 2011 14:13:10 +0200
> Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:
> 
> > Instead of using a single batch (the small on-stack, or an allocated
> > page), try and extend the batch every time it runs out and only flush
> > once either the extend fails or we're done.
> 
> why?

To avoid sending extra TLB invalidates.

> > @@ -86,22 +86,48 @@ struct mmu_gather {
> >  #ifdef CONFIG_HAVE_RCU_TABLE_FREE
> >  	struct mmu_table_batch	*batch;
> >  #endif
> > +	unsigned int		need_flush : 1,	/* Did free PTEs */
> > +				fast_mode  : 1; /* No batching   */
> 
> mmu_gather.fast_mode gets modified in several places apparently without
> locking to protect itself.  I don't think that these modifications will
> accidentally trash need_flush, mainly by luck.

The other way around I'd think.

> Please review the concurrency issues here and document them clearly.

Its an on-stack structure, there is no concurrency. /me shall add a
comment.

> > +#ifdef CONFIG_SMP
> > +  #define tlb_fast_mode(tlb) (tlb->fast_mode)
> > +#else
> > +  #define tlb_fast_mode(tlb) 1
> > +#endif
> 
> Mutter.
> 
> Could have been written in C.

Fixed in my last patch uninlining bits

> Will cause a compile error with, for example, tlb_fast_mode(tlb + 1).

Well, that'd actually be a good reason to keep the macro ;-)

> > +static inline int tlb_next_batch(struct mmu_gather *tlb)
> >  {
> > +	struct mmu_gather_batch *batch;
> >  
> > +	batch = tlb->active;
> > +	if (batch->next) {
> > +		tlb->active = batch->next;
> > +		return 1;
> >  	}
> > +
> > +	batch = (void *)__get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
> 
> A comment explaining the gfp_t decision would be useful.

Done

> > +	if (!batch)
> > +		return 0;
> > +
> > +	batch->next = NULL;
> > +	batch->nr   = 0;
> > +	batch->max  = MAX_GATHER_BATCH;
> > +
> > +	tlb->active->next = batch;
> > +	tlb->active = batch;
> > +
> > +	return 1;
> >  }
> >  
> >  /* tlb_gather_mmu
> > @@ -114,16 +140,13 @@ tlb_gather_mmu(struct mmu_gather *tlb, s
> >  {
> >  	tlb->mm = mm;
> >  
> > +	tlb->fullmm     = fullmm;
> > +	tlb->need_flush = 0;
> > +	tlb->fast_mode  = (num_possible_cpus() == 1);
> 
> The changelog didn't tell us why we switched from num_online_cpus() to
> num_possible_cpus().

And that is a very good question... somehow I remember a conversation
with BenH about this, but on second thought that might have been about
his pgtable_free_tlb() optimization (which is somewhat similar).

Let me restore that to num_online_cpus() and maybe do a later patch
removing fast_mode all together as Hugh suggested, since even UP might
have benefit from the batching due to less zone-lock activity on bulk
frees.

> > +	tlb->local.next = NULL;
> > +	tlb->local.nr   = 0;
> > +	tlb->local.max  = ARRAY_SIZE(tlb->__pages);
> > +	tlb->active     = &tlb->local;
> >  
> >  #ifdef CONFIG_HAVE_RCU_TABLE_FREE
> >  	tlb->batch = NULL;
> >
> > ...
> >
> > @@ -177,15 +205,24 @@ tlb_finish_mmu(struct mmu_gather *tlb, u
> > +	batch = tlb->active;
> > +	batch->pages[batch->nr++] = page;
> > +	VM_BUG_ON(batch->nr > batch->max);
> > +	if (batch->nr == batch->max) {
> > +		if (!tlb_next_batch(tlb))
> > +			return 0;
> > +	}
> 
> Moving the VM_BUG_ON() down to after the if() would save a few cycles.

Done.

> > +	return batch->max - batch->nr;
> >  }
> >  
> >  /* tlb_remove_page
> > 

Re: [PATCH 12/20] mm: Extended batches for generic mmu_gather

[Peter Zijlstra]

On Tue, 2011-04-19 at 13:06 -0700, Andrew Morton wrote:
> On Fri, 01 Apr 2011 14:13:10 +0200
> Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:
> 
> > Instead of using a single batch (the small on-stack, or an allocated
> > page), try and extend the batch every time it runs out and only flush
> > once either the extend fails or we're done.
> 
> why?

To avoid sending extra TLB invalidates.

> > @@ -86,22 +86,48 @@ struct mmu_gather {
> >  #ifdef CONFIG_HAVE_RCU_TABLE_FREE
> >  	struct mmu_table_batch	*batch;
> >  #endif
> > +	unsigned int		need_flush : 1,	/* Did free PTEs */
> > +				fast_mode  : 1; /* No batching   */
> 
> mmu_gather.fast_mode gets modified in several places apparently without
> locking to protect itself.  I don't think that these modifications will
> accidentally trash need_flush, mainly by luck.

The other way around I'd think.

> Please review the concurrency issues here and document them clearly.

Its an on-stack structure, there is no concurrency. /me shall add a
comment.

> > +#ifdef CONFIG_SMP
> > +  #define tlb_fast_mode(tlb) (tlb->fast_mode)
> > +#else
> > +  #define tlb_fast_mode(tlb) 1
> > +#endif
> 
> Mutter.
> 
> Could have been written in C.

Fixed in my last patch uninlining bits

> Will cause a compile error with, for example, tlb_fast_mode(tlb + 1).

Well, that'd actually be a good reason to keep the macro ;-)

> > +static inline int tlb_next_batch(struct mmu_gather *tlb)
> >  {
> > +	struct mmu_gather_batch *batch;
> >  
> > +	batch = tlb->active;
> > +	if (batch->next) {
> > +		tlb->active = batch->next;
> > +		return 1;
> >  	}
> > +
> > +	batch = (void *)__get_free_pages(GFP_NOWAIT | __GFP_NOWARN, 0);
> 
> A comment explaining the gfp_t decision would be useful.

Done

> > +	if (!batch)
> > +		return 0;
> > +
> > +	batch->next = NULL;
> > +	batch->nr   = 0;
> > +	batch->max  = MAX_GATHER_BATCH;
> > +
> > +	tlb->active->next = batch;
> > +	tlb->active = batch;
> > +
> > +	return 1;
> >  }
> >  
> >  /* tlb_gather_mmu
> > @@ -114,16 +140,13 @@ tlb_gather_mmu(struct mmu_gather *tlb, s
> >  {
> >  	tlb->mm = mm;
> >  
> > +	tlb->fullmm     = fullmm;
> > +	tlb->need_flush = 0;
> > +	tlb->fast_mode  = (num_possible_cpus() == 1);
> 
> The changelog didn't tell us why we switched from num_online_cpus() to
> num_possible_cpus().

And that is a very good question... somehow I remember a conversation
with BenH about this, but on second thought that might have been about
his pgtable_free_tlb() optimization (which is somewhat similar).

Let me restore that to num_online_cpus() and maybe do a later patch
removing fast_mode all together as Hugh suggested, since even UP might
have benefit from the batching due to less zone-lock activity on bulk
frees.

> > +	tlb->local.next = NULL;
> > +	tlb->local.nr   = 0;
> > +	tlb->local.max  = ARRAY_SIZE(tlb->__pages);
> > +	tlb->active     = &tlb->local;
> >  
> >  #ifdef CONFIG_HAVE_RCU_TABLE_FREE
> >  	tlb->batch = NULL;
> >
> > ...
> >
> > @@ -177,15 +205,24 @@ tlb_finish_mmu(struct mmu_gather *tlb, u
> > +	batch = tlb->active;
> > +	batch->pages[batch->nr++] = page;
> > +	VM_BUG_ON(batch->nr > batch->max);
> > +	if (batch->nr == batch->max) {
> > +		if (!tlb_next_batch(tlb))
> > +			return 0;
> > +	}
> 
> Moving the VM_BUG_ON() down to after the if() would save a few cycles.

Done.

> > +	return batch->max - batch->nr;
> >  }
> >  
> >  /* tlb_remove_page
> > 

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 13/20] lockdep, mutex: Provide mutex_lock_nest_lock

[Peter Zijlstra]

On Tue, 2011-04-19 at 13:06 -0700, Andrew Morton wrote:
> On Fri, 01 Apr 2011 14:13:11 +0200
> Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:
> 
> > Provide the mutex_lock_nest_lock() annotation.
> 
> why?
> 
> Neither the changelog nor the code provide any documentation for this addition to
> the lokdep API.

---
Subject: lockdep, mutex: Provide mutex_lock_nest_lock                                        
From: Peter Zijlstra <a.p.zijlstra@chello.nl>                                                
Date: Fri, 26 Nov 2010 15:39:00 +0100                                                        
                                                                                             
In order to convert i_mmap_lock to a mutex we need a mutex equivalent                        
to spin_lock_nest_lock(), thus provide the mutex_lock_nest_lock()                            
annotation.                                                                                  
                                                                                             
As with spin_lock_nest_lock(), mutex_lock_nest_lock() allows                                 
annotation of the locking pattern where an outer lock serializes the                         
acquisition order of nested locks. That is, if every time you lock                            
multiple locks A, say A1 and A2 you first acquire N, the order of                            
acquiring A1 and A2 is irrelevant.                                                           
                                                                                             
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>  
---

Re: [PATCH 13/20] lockdep, mutex: Provide mutex_lock_nest_lock

[Peter Zijlstra]

On Tue, 2011-04-19 at 13:06 -0700, Andrew Morton wrote:
> On Fri, 01 Apr 2011 14:13:11 +0200
> Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:
> 
> > Provide the mutex_lock_nest_lock() annotation.
> 
> why?
> 
> Neither the changelog nor the code provide any documentation for this addition to
> the lokdep API.

---
Subject: lockdep, mutex: Provide mutex_lock_nest_lock                                        
From: Peter Zijlstra <a.p.zijlstra@chello.nl>                                                
Date: Fri, 26 Nov 2010 15:39:00 +0100                                                        
                                                                                             
In order to convert i_mmap_lock to a mutex we need a mutex equivalent                        
to spin_lock_nest_lock(), thus provide the mutex_lock_nest_lock()                            
annotation.                                                                                  
                                                                                             
As with spin_lock_nest_lock(), mutex_lock_nest_lock() allows                                 
annotation of the locking pattern where an outer lock serializes the                         
acquisition order of nested locks. That is, if every time you lock                            
multiple locks A, say A1 and A2 you first acquire N, the order of                            
acquiring A1 and A2 is irrelevant.                                                           
                                                                                             
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>  
---

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 20/20] mm: Optimize page_lock_anon_vma() fast-path

[Peter Zijlstra]

On Tue, 2011-04-19 at 13:08 -0700, Andrew Morton wrote:
> On Fri, 01 Apr 2011 14:13:18 +0200
> Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:
> 
> > Optimize the page_lock_anon_vma() fast path to be one atomic op,
> > instead of two.
> > 
> > Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
> > Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
> > LKML-Reference: <new-submission>
> > ---
> >  mm/rmap.c |   86 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++---
> >  1 file changed, 82 insertions(+), 4 deletions(-)
> > 
> > Index: linux-2.6/mm/rmap.c
> > ===================================================================
> > --- linux-2.6.orig/mm/rmap.c
> > +++ linux-2.6/mm/rmap.c
> > @@ -85,6 +85,29 @@ static inline struct anon_vma *anon_vma_
> >  static inline void anon_vma_free(struct anon_vma *anon_vma)
> >  {
> >  	VM_BUG_ON(atomic_read(&anon_vma->refcount));
> > +
> > +	/*
> > +	 * Synchronize against page_lock_anon_vma() such that
> > +	 * we can safely hold the lock without the anon_vma getting
> > +	 * freed.
> > +	 *
> > +	 * Relies on the full mb implied by the atomic_dec_and_test() from
> > +	 * put_anon_vma() against the acquire barrier implied by
> > +	 * mutex_trylock() from page_lock_anon_vma(). This orders:
> > +	 *
> > +	 * page_lock_anon_vma()		VS	put_anon_vma()
> > +	 *   mutex_trylock()			  atomic_dec_and_test()
> > +	 *   LOCK				  MB
> > +	 *   atomic_read()			  mutex_is_locked()
> > +	 *
> > +	 * LOCK should suffice since the actual taking of the lock must
> > +	 * happen _before_ what follows.
> > +	 */
> > +	if (mutex_is_locked(&anon_vma->root->mutex)) {
> > +		anon_vma_lock(anon_vma);
> > +		anon_vma_unlock(anon_vma);
> > +	}
> > +
> >  	kmem_cache_free(anon_vma_cachep, anon_vma);
> >  }
> 
> Did we need to include all this stuff in uniprocessor builds?

For sure, even UP can schedule while holding a mutex.

> It would be neater to add a new anon_vma_is_locked().

I'd agree if there was a user outside of rmap.c, but seeing as rmap.c is
and must be aware of the whole anon_vma->root thing I don't much see the
point in extra wrappery.

> This code is too tricksy to deserve life :(

I'd mostly agree with you there, but there was a strong desire to keep
page_lock_anon_vma() a single atomic. I'll see if I can actually measure
any difference using aim7 or so, which I think is the favorite anon_vma
stress tool.

> > @@ -371,20 +394,75 @@ struct anon_vma *page_get_anon_vma(struc
> >  	return anon_vma;
> >  }
> >  
> > +/*
> > + * Similar to page_get_anon_vma() except it locks the anon_vma.
> > + *
> > + * Its a little more complex as it tries to keep the fast path to a single
> > + * atomic op -- the trylock. If we fail the trylock, we fall back to getting a
> > + * reference like with page_get_anon_vma() and then block on the mutex.
> > + */
> >  struct anon_vma *page_lock_anon_vma(struct page *page)
> >  {
> > -	struct anon_vma *anon_vma = page_get_anon_vma(page);
> > +	struct anon_vma *anon_vma = NULL;
> > +	unsigned long anon_mapping;
> >  
> > -	if (anon_vma)
> > -		anon_vma_lock(anon_vma);
> > +	rcu_read_lock();
> > +	anon_mapping = (unsigned long) ACCESS_ONCE(page->mapping);
> > +	if ((anon_mapping & PAGE_MAPPING_FLAGS) != PAGE_MAPPING_ANON)
> > +		goto out;
> 
> Why?  Needs a comment.

Uhm, why we're testing to see if there is an anon_vma at all? Or why we
need that ACCESS_ONCE()?

> > +	if (!page_mapped(page))
> > +		goto out;
> 
> Why?  How can this come about? Needs a comment.

Well, the existing comment says to look at page_get_anon_vma() and the
comment there does explain how all this is racy wrt page_remove_rmap().
Do you want more comments?

> > +
> > +	anon_vma = (struct anon_vma *) (anon_mapping - PAGE_MAPPING_ANON);
> > +	if (mutex_trylock(&anon_vma->root->mutex)) {
> 
> anon_vma_trylock()?
> 
> Or just remove all the wrapper functions and open-code all the locking.
> These tricks all seem pretty tied-up with the mutex implementation
> anyway.

Well, we cannot remove all the wrappers, anon_vma_{un,}lock() are used
outside of rmap.c and we don't want to expose the implementation of the
anon_vma locking outside of here, but like said, inside rmap.c I don't
see much reason to introduce new wrappers.

And yes, all of this is needed because of the anon_vma->lock mutex
conversion since, in general, we cannot schedule under rcu_read_lock and
therefore have to play these tricks with the reference count to bridge
the gap between rcu_read_unlock() and acquiring the lock.

> > +		/*
> > +		 * If we observe a !0 refcount, then holding the lock ensures
> > +		 * the anon_vma will not go away, see __put_anon_vma().
> > +		 */
> > +		if (!atomic_read(&anon_vma->refcount)) {
> > +			anon_vma_unlock(anon_vma);
> > +			anon_vma = NULL;
> > +		}
> > +		goto out;
> > +	}
> > +
> > +	/* trylock failed, we got to sleep */
> > +	if (!atomic_inc_not_zero(&anon_vma->refcount)) {
> > +		anon_vma = NULL;
> > +		goto out;
> > +	}
> >  
> > +	if (!page_mapped(page)) {
> > +		put_anon_vma(anon_vma);
> > +		anon_vma = NULL;
> > +		goto out;
> > +	}
> 
> Also quite opaque, needs decent commentary.
> 
> I'd have expected this test to occur after the lock was acquired.

Right, so I think we could drop that test from both here and
page_get_anon_vma() and nothing would break, its simply avoiding some
work in case we do detect the race with page_remove_rmap().

So yes, I think I'll move it down because that'll widen the scope of
this optimization.

> > +	/* we pinned the anon_vma, its safe to sleep */
> > +	rcu_read_unlock();
> > +	anon_vma_lock(anon_vma);
> > +
> > +	if (atomic_dec_and_test(&anon_vma->refcount)) {
> > +		/*
> > +		 * Oops, we held the last refcount, release the lock
> > +		 * and bail -- can't simply use put_anon_vma() because
> > +		 * we'll deadlock on the anon_vma_lock() recursion.
> > +		 */
> > +		anon_vma_unlock(anon_vma);
> > +		__put_anon_vma(anon_vma);
> > +		anon_vma = NULL;
> > +	}
> > +
> > +	return anon_vma;
> > +
> > +out:
> > +	rcu_read_unlock();
> >  	return anon_vma;
> >  }
> >  
> >  void page_unlock_anon_vma(struct anon_vma *anon_vma)
> >  {
> >  	anon_vma_unlock(anon_vma);
> > -	put_anon_vma(anon_vma);
> >  }
> 
> Geeze, I hope this patch is worth it :( :(

There is a reason this is the last patch in the series ;-)

Re: [PATCH 20/20] mm: Optimize page_lock_anon_vma() fast-path

[Peter Zijlstra]

On Tue, 2011-04-19 at 13:08 -0700, Andrew Morton wrote:
> On Fri, 01 Apr 2011 14:13:18 +0200
> Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:
> 
> > Optimize the page_lock_anon_vma() fast path to be one atomic op,
> > instead of two.
> > 
> > Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
> > Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
> > LKML-Reference: <new-submission>
> > ---
> >  mm/rmap.c |   86 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++---
> >  1 file changed, 82 insertions(+), 4 deletions(-)
> > 
> > Index: linux-2.6/mm/rmap.c
> > ===================================================================
> > --- linux-2.6.orig/mm/rmap.c
> > +++ linux-2.6/mm/rmap.c
> > @@ -85,6 +85,29 @@ static inline struct anon_vma *anon_vma_
> >  static inline void anon_vma_free(struct anon_vma *anon_vma)
> >  {
> >  	VM_BUG_ON(atomic_read(&anon_vma->refcount));
> > +
> > +	/*
> > +	 * Synchronize against page_lock_anon_vma() such that
> > +	 * we can safely hold the lock without the anon_vma getting
> > +	 * freed.
> > +	 *
> > +	 * Relies on the full mb implied by the atomic_dec_and_test() from
> > +	 * put_anon_vma() against the acquire barrier implied by
> > +	 * mutex_trylock() from page_lock_anon_vma(). This orders:
> > +	 *
> > +	 * page_lock_anon_vma()		VS	put_anon_vma()
> > +	 *   mutex_trylock()			  atomic_dec_and_test()
> > +	 *   LOCK				  MB
> > +	 *   atomic_read()			  mutex_is_locked()
> > +	 *
> > +	 * LOCK should suffice since the actual taking of the lock must
> > +	 * happen _before_ what follows.
> > +	 */
> > +	if (mutex_is_locked(&anon_vma->root->mutex)) {
> > +		anon_vma_lock(anon_vma);
> > +		anon_vma_unlock(anon_vma);
> > +	}
> > +
> >  	kmem_cache_free(anon_vma_cachep, anon_vma);
> >  }
> 
> Did we need to include all this stuff in uniprocessor builds?

For sure, even UP can schedule while holding a mutex.

> It would be neater to add a new anon_vma_is_locked().

I'd agree if there was a user outside of rmap.c, but seeing as rmap.c is
and must be aware of the whole anon_vma->root thing I don't much see the
point in extra wrappery.

> This code is too tricksy to deserve life :(

I'd mostly agree with you there, but there was a strong desire to keep
page_lock_anon_vma() a single atomic. I'll see if I can actually measure
any difference using aim7 or so, which I think is the favorite anon_vma
stress tool.

> > @@ -371,20 +394,75 @@ struct anon_vma *page_get_anon_vma(struc
> >  	return anon_vma;
> >  }
> >  
> > +/*
> > + * Similar to page_get_anon_vma() except it locks the anon_vma.
> > + *
> > + * Its a little more complex as it tries to keep the fast path to a single
> > + * atomic op -- the trylock. If we fail the trylock, we fall back to getting a
> > + * reference like with page_get_anon_vma() and then block on the mutex.
> > + */
> >  struct anon_vma *page_lock_anon_vma(struct page *page)
> >  {
> > -	struct anon_vma *anon_vma = page_get_anon_vma(page);
> > +	struct anon_vma *anon_vma = NULL;
> > +	unsigned long anon_mapping;
> >  
> > -	if (anon_vma)
> > -		anon_vma_lock(anon_vma);
> > +	rcu_read_lock();
> > +	anon_mapping = (unsigned long) ACCESS_ONCE(page->mapping);
> > +	if ((anon_mapping & PAGE_MAPPING_FLAGS) != PAGE_MAPPING_ANON)
> > +		goto out;
> 
> Why?  Needs a comment.

Uhm, why we're testing to see if there is an anon_vma at all? Or why we
need that ACCESS_ONCE()?

> > +	if (!page_mapped(page))
> > +		goto out;
> 
> Why?  How can this come about? Needs a comment.

Well, the existing comment says to look at page_get_anon_vma() and the
comment there does explain how all this is racy wrt page_remove_rmap().
Do you want more comments?

> > +
> > +	anon_vma = (struct anon_vma *) (anon_mapping - PAGE_MAPPING_ANON);
> > +	if (mutex_trylock(&anon_vma->root->mutex)) {
> 
> anon_vma_trylock()?
> 
> Or just remove all the wrapper functions and open-code all the locking.
> These tricks all seem pretty tied-up with the mutex implementation
> anyway.

Well, we cannot remove all the wrappers, anon_vma_{un,}lock() are used
outside of rmap.c and we don't want to expose the implementation of the
anon_vma locking outside of here, but like said, inside rmap.c I don't
see much reason to introduce new wrappers.

And yes, all of this is needed because of the anon_vma->lock mutex
conversion since, in general, we cannot schedule under rcu_read_lock and
therefore have to play these tricks with the reference count to bridge
the gap between rcu_read_unlock() and acquiring the lock.

> > +		/*
> > +		 * If we observe a !0 refcount, then holding the lock ensures
> > +		 * the anon_vma will not go away, see __put_anon_vma().
> > +		 */
> > +		if (!atomic_read(&anon_vma->refcount)) {
> > +			anon_vma_unlock(anon_vma);
> > +			anon_vma = NULL;
> > +		}
> > +		goto out;
> > +	}
> > +
> > +	/* trylock failed, we got to sleep */
> > +	if (!atomic_inc_not_zero(&anon_vma->refcount)) {
> > +		anon_vma = NULL;
> > +		goto out;
> > +	}
> >  
> > +	if (!page_mapped(page)) {
> > +		put_anon_vma(anon_vma);
> > +		anon_vma = NULL;
> > +		goto out;
> > +	}
> 
> Also quite opaque, needs decent commentary.
> 
> I'd have expected this test to occur after the lock was acquired.

Right, so I think we could drop that test from both here and
page_get_anon_vma() and nothing would break, its simply avoiding some
work in case we do detect the race with page_remove_rmap().

So yes, I think I'll move it down because that'll widen the scope of
this optimization.

> > +	/* we pinned the anon_vma, its safe to sleep */
> > +	rcu_read_unlock();
> > +	anon_vma_lock(anon_vma);
> > +
> > +	if (atomic_dec_and_test(&anon_vma->refcount)) {
> > +		/*
> > +		 * Oops, we held the last refcount, release the lock
> > +		 * and bail -- can't simply use put_anon_vma() because
> > +		 * we'll deadlock on the anon_vma_lock() recursion.
> > +		 */
> > +		anon_vma_unlock(anon_vma);
> > +		__put_anon_vma(anon_vma);
> > +		anon_vma = NULL;
> > +	}
> > +
> > +	return anon_vma;
> > +
> > +out:
> > +	rcu_read_unlock();
> >  	return anon_vma;
> >  }
> >  
> >  void page_unlock_anon_vma(struct anon_vma *anon_vma)
> >  {
> >  	anon_vma_unlock(anon_vma);
> > -	put_anon_vma(anon_vma);
> >  }
> 
> Geeze, I hope this patch is worth it :( :(

There is a reason this is the last patch in the series ;-)

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 20/20] mm: Optimize page_lock_anon_vma() fast-path

[Peter Zijlstra]

On Wed, 2011-04-20 at 14:38 +0200, Peter Zijlstra wrote:
> > > +   if (!page_mapped(page)) {
> > > +           put_anon_vma(anon_vma);
> > > +           anon_vma = NULL;
> > > +           goto out;
> > > +   }
> > 
> > Also quite opaque, needs decent commentary.
> > 
> > I'd have expected this test to occur after the lock was acquired.
> 
> Right, so I think we could drop that test from both here and
> page_get_anon_vma() and nothing would break, its simply avoiding some
> work in case we do detect the race with page_remove_rmap().
> 
> So yes, I think I'll move it down because that'll widen the scope of
> this optimization. 

OK, so I went trawling through the linux-mm logs and actually found why
this is needed under rcu_read_lock(), sadly I cannot seem to find a web
reference to 2006 linux-mm emails.

It was Hugh who noted that page_remove_rmap() only decrements
page->_mapcount but does not clear page->mapping, therefore we must test
page_mapped() after reading page->mapping while under the
rcu_read_lock() in order to determine if the pointer obtained is still
valid.

The comment that exists today in __page_lock_anon_vma() actually tries
to explain this

Re: [PATCH 20/20] mm: Optimize page_lock_anon_vma() fast-path

[Peter Zijlstra]

On Wed, 2011-04-20 at 14:38 +0200, Peter Zijlstra wrote:
> > > +   if (!page_mapped(page)) {
> > > +           put_anon_vma(anon_vma);
> > > +           anon_vma = NULL;
> > > +           goto out;
> > > +   }
> > 
> > Also quite opaque, needs decent commentary.
> > 
> > I'd have expected this test to occur after the lock was acquired.
> 
> Right, so I think we could drop that test from both here and
> page_get_anon_vma() and nothing would break, its simply avoiding some
> work in case we do detect the race with page_remove_rmap().
> 
> So yes, I think I'll move it down because that'll widen the scope of
> this optimization. 

OK, so I went trawling through the linux-mm logs and actually found why
this is needed under rcu_read_lock(), sadly I cannot seem to find a web
reference to 2006 linux-mm emails.

It was Hugh who noted that page_remove_rmap() only decrements
page->_mapcount but does not clear page->mapping, therefore we must test
page_mapped() after reading page->mapping while under the
rcu_read_lock() in order to determine if the pointer obtained is still
valid.

The comment that exists today in __page_lock_anon_vma() actually tries
to explain this


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 15/20] mm: Convert i_mmap_lock to a mutex

[Peter Zijlstra]

On Tue, 2011-04-19 at 13:07 -0700, Andrew Morton wrote:
> On Fri, 01 Apr 2011 14:13:13 +0200
> Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:
> 
> > Straight fwd conversion of i_mmap_lock to a mutex
> 
> What effect does this have on kernel performance?
> 
> Workloads which take the lock at high frequency from multiple threads
> should be tested.

The worst affected workload would be a lightly contended lock, uncontended
mutexes are similarly fast as uncontended spinlocks, highly contended mutexes
win hands down since we waste vastly less resources spinning, leaving
lightly contended.

The below shows a workload tailored to isolate this one lock and is tested with
various numbers of contending tasks ran on a dual socket westmere EP (2*6*2),
the first results are from an unpatched -tip kernel, the second set is with
the patches applied:


	48		24		12		6		3

real    0m33.973s	0m17.816s	0m8.762s	0m3.778s	0m1.307s
sys     13m20.203s	7m5.814s	1m44.440s	0m22.371s	0m3.751s

real    0m23.551s	0m12.265s	0m7.016s	0m3.644s	0m1.412s
sys     1m17.177s	0m41.875s	0m26.139s	0m13.872s	0m3.873s

It looks like we might have a slight loss at 3 tasks contending although I'd
have to run with larger sets in order to increase the confidence interval as
the current numbers are within each others error range.

Also, this conversion is needed to allow the anon_vma->lock conversion, which
shows an improvement over the whole range.

---

[root@westmere ~]# echo 0 > /proc/sys/kernel/nmi_watchdog 
[root@westmere ~]# for i in /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor ; do echo performance > $i; done
[root@westmere ~]# time perf stat --repeat 10 ./mmap-file 48 50000

 Performance counter stats for './mmap-file 48 50000' (10 runs):

          3.045659 task-clock-msecs         #      0.001 CPUs    ( +-   0.481% )
                48 context-switches         #      0.016 M/sec   ( +-   0.566% )
                69 CPU-migrations           #      0.023 M/sec   ( +-   1.456% )
               250 page-faults              #      0.082 M/sec   ( +-   0.040% )
         7,985,758 cycles                   #   2622.013 M/sec   ( +-   0.413% )
         3,335,330 instructions             #      0.418 IPC     ( +-   0.198% )
           614,795 branches                 #    201.859 M/sec   ( +-   0.175% )
            20,177 branch-misses            #      3.282 %       ( +-   1.299% )
           123,596 cache-references         #     40.581 M/sec   ( +-   0.693% )
            57,380 cache-misses             #     18.840 M/sec   ( +-   2.148% )

        3.380886384  seconds time elapsed   ( +-   0.159% )


real    0m33.973s
user    0m1.463s
sys     13m20.203s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-file 24 50000

 Performance counter stats for './mmap-file 24 50000' (10 runs):

          1.511615 task-clock-msecs         #      0.001 CPUs    ( +-   1.049% )
                16 context-switches         #      0.011 M/sec   ( +-   3.040% )
                23 CPU-migrations           #      0.015 M/sec   ( +-   0.437% )
               178 page-faults              #      0.118 M/sec   ( +-   0.120% )
         4,003,891 cycles                   #   2648.750 M/sec   ( +-   0.988% )
         1,767,231 instructions             #      0.441 IPC     ( +-   0.678% )
           328,415 branches                 #    217.261 M/sec   ( +-   0.707% )
             9,281 branch-misses            #      2.826 %       ( +-   1.045% )
            56,897 cache-references         #     37.640 M/sec   ( +-   0.966% )
            26,209 cache-misses             #     17.339 M/sec   ( +-   2.700% )

        1.779547040  seconds time elapsed   ( +-   0.116% )


real    0m17.816s
user    0m0.704s
sys     7m5.814s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-file 12 50000

 Performance counter stats for './mmap-file 12 50000' (10 runs):

          0.833400 task-clock-msecs         #      0.001 CPUs    ( +-   2.223% )
                 8 context-switches         #      0.010 M/sec   ( +-   4.047% )
                12 CPU-migrations           #      0.014 M/sec   ( +-   0.000% )
               142 page-faults              #      0.170 M/sec   ( +-   0.105% )
         2,218,792 cycles                   #   2662.338 M/sec   ( +-   2.025% )
         1,091,917 instructions             #      0.492 IPC     ( +-   0.558% )
           205,310 branches                 #    246.352 M/sec   ( +-   0.516% )
             6,030 branch-misses            #      2.937 %       ( +-   1.436% )
            32,260 cache-references         #     38.710 M/sec   ( +-   0.747% )
            12,526 cache-misses             #     15.030 M/sec   ( +-   2.783% )

        0.874054257  seconds time elapsed   ( +-   0.122% )


real    0m8.762s
user    0m0.283s
sys     1m44.440s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-file 6 50000

 Performance counter stats for './mmap-file 6 50000' (10 runs):

          0.525969 task-clock-msecs         #      0.001 CPUs    ( +-   1.705% )
                 3 context-switches         #      0.006 M/sec   ( +-  11.994% )
                 6 CPU-migrations           #      0.011 M/sec   ( +-   0.000% )
               124 page-faults              #      0.236 M/sec   ( +-   0.000% )
         1,413,250 cycles                   #   2686.948 M/sec   ( +-   1.138% )
           757,501 instructions             #      0.536 IPC     ( +-   0.565% )
           144,120 branches                 #    274.009 M/sec   ( +-   0.560% )
             5,098 branch-misses            #      3.538 %       ( +-   2.054% )
            22,782 cache-references         #     43.315 M/sec   ( +-   1.068% )
             6,159 cache-misses             #     11.709 M/sec   ( +-   5.052% )

        0.375598224  seconds time elapsed   ( +-   0.560% )


real    0m3.778s
user    0m0.100s
sys     0m22.371s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-file 3 50000

 Performance counter stats for './mmap-file 3 50000' (10 runs):

          0.426419 task-clock-msecs         #      0.003 CPUs    ( +-   2.223% )
                 3 context-switches         #      0.007 M/sec   ( +-   3.448% )
                 3 CPU-migrations           #      0.007 M/sec   ( +-   0.000% )
               115 page-faults              #      0.270 M/sec   ( +-   0.132% )
         1,134,258 cycles                   #   2659.963 M/sec   ( +-   1.587% )
           610,569 instructions             #      0.538 IPC     ( +-   0.289% )
           117,666 branches                 #    275.940 M/sec   ( +-   0.320% )
             4,318 branch-misses            #      3.670 %       ( +-   3.292% )
            17,922 cache-references         #     42.029 M/sec   ( +-   0.819% )
             3,905 cache-misses             #      9.157 M/sec   ( +-   7.969% )

        0.127845444  seconds time elapsed   ( +-   1.226% )


real    0m1.307s
user    0m0.158s
sys     0m3.751s



---

[root@westmere ~]# echo 0 > /proc/sys/kernel/nmi_watchdog 
[root@westmere ~]# for i in /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor ; do echo performance > $i; done
[root@westmere ~]# time perf stat --repeat 10 ./mmap-file 48 50000

 Performance counter stats for './mmap-file 48 50000' (10 runs):

          2.910655 task-clock-msecs         #      0.001 CPUs    ( +-   0.580% )
                48 context-switches         #      0.016 M/sec   ( +-   0.279% )
                52 CPU-migrations           #      0.018 M/sec   ( +-   1.225% )
               250 page-faults              #      0.086 M/sec   ( +-   0.053% )
         7,538,292 cycles                   #   2589.895 M/sec   ( +-   0.725% )
         3,285,984 instructions             #      0.436 IPC     ( +-   0.248% )
           608,888 branches                 #    209.193 M/sec   ( +-   0.256% )
            17,153 branch-misses            #      2.817 %       ( +-   0.944% )
           128,392 cache-references         #     44.111 M/sec   ( +-   0.816% )
            57,160 cache-misses             #     19.638 M/sec   ( +-   2.574% )

        2.352105837  seconds time elapsed   ( +-   0.729% )


real    0m23.551s
user    0m1.451s
sys     1m17.177s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-file 24 50000

 Performance counter stats for './mmap-file 24 50000' (10 runs):

          1.600378 task-clock-msecs         #      0.001 CPUs    ( +-   1.531% )
                24 context-switches         #      0.015 M/sec   ( +-   0.745% )
                25 CPU-migrations           #      0.015 M/sec   ( +-   0.806% )
               178 page-faults              #      0.111 M/sec   ( +-   0.118% )
         4,112,616 cycles                   #   2569.778 M/sec   ( +-   1.250% )
         1,839,878 instructions             #      0.447 IPC     ( +-   0.307% )
           343,196 branches                 #    214.447 M/sec   ( +-   0.361% )
            10,876 branch-misses            #      3.169 %       ( +-   1.514% )
            68,812 cache-references         #     42.998 M/sec   ( +-   1.175% )
            26,286 cache-misses             #     16.425 M/sec   ( +-   2.509% )

        1.224012964  seconds time elapsed   ( +-   1.200% )


real    0m12.265s
user    0m0.685s
sys     0m41.875s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-file 12 50000

 Performance counter stats for './mmap-file 12 50000' (10 runs):

          0.833480 task-clock-msecs         #      0.001 CPUs    ( +-   1.324% )
                12 context-switches         #      0.014 M/sec   ( +-   0.000% )
                12 CPU-migrations           #      0.014 M/sec   ( +-   0.000% )
               142 page-faults              #      0.170 M/sec   ( +-   0.126% )
         2,180,779 cycles                   #   2616.476 M/sec   ( +-   1.408% )
         1,127,027 instructions             #      0.517 IPC     ( +-   0.354% )
           211,726 branches                 #    254.026 M/sec   ( +-   0.327% )
             6,628 branch-misses            #      3.131 %       ( +-   1.522% )
            36,986 cache-references         #     44.376 M/sec   ( +-   1.459% )
            12,337 cache-misses             #     14.802 M/sec   ( +-   3.203% )

        0.699308658  seconds time elapsed   ( +-   0.931% )


real    0m7.016s
user    0m0.323s
sys     0m26.139s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-file 6 50000

 Performance counter stats for './mmap-file 6 50000' (10 runs):

          0.551046 task-clock-msecs         #      0.002 CPUs    ( +-   2.031% )
                 6 context-switches         #      0.011 M/sec   ( +-   3.042% )
                 6 CPU-migrations           #      0.011 M/sec   ( +-   0.000% )
               124 page-faults              #      0.226 M/sec   ( +-   0.123% )
         1,482,359 cycles                   #   2690.081 M/sec   ( +-   1.767% )
           793,407 instructions             #      0.535 IPC     ( +-   0.387% )
           150,806 branches                 #    273.673 M/sec   ( +-   0.416% )
             5,649 branch-misses            #      3.746 %       ( +-   2.961% )
            24,554 cache-references         #     44.559 M/sec   ( +-   1.985% )
             5,632 cache-misses             #     10.220 M/sec   ( +-   4.339% )

        0.362139074  seconds time elapsed   ( +-   1.599% )


real    0m3.644s
user    0m0.153s
sys     0m13.872s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-file 3 50000

 Performance counter stats for './mmap-file 3 50000' (10 runs):

          0.419324 task-clock-msecs         #      0.003 CPUs    ( +-   2.667% )
                 3 context-switches         #      0.007 M/sec   ( +-   0.000% )
                 3 CPU-migrations           #      0.007 M/sec   ( +-   0.000% )
               116 page-faults              #      0.276 M/sec   ( +-   0.264% )
         1,147,068 cycles                   #   2735.515 M/sec   ( +-   2.253% )
           623,614 instructions             #      0.544 IPC     ( +-   0.228% )
           119,060 branches                 #    283.933 M/sec   ( +-   0.207% )
             4,679 branch-misses            #      3.930 %       ( +-   2.200% )
            17,917 cache-references         #     42.729 M/sec   ( +-   1.021% )
             3,779 cache-misses             #      9.013 M/sec   ( +-   3.425% )

        0.138764629  seconds time elapsed   ( +-   0.903% )


real    0m1.412s
user    0m0.065s
sys     0m3.873s


---
#include <unistd.h>
#include <stdio.h>
#include <stdlib.h>
#include <sys/mman.h>
#include <pthread.h>
#include <errno.h>
#include <sys/types.h>
#include <sys/wait.h>
#include <fcntl.h>


unsigned long nr_loops = 1000000; /* 1M */
unsigned long page_size;
int file;

void *do_mmap_merge(void *data)
{
	int nr = (unsigned long)data;
	void *addr;
	int ret;
	int i;

	for (i = 0; i < nr_loops; i++) {
		addr = mmap(NULL, page_size, PROT_READ | PROT_WRITE, 
				MAP_SHARED, file, 0);
		if (addr == MAP_FAILED) {
			perror("thread-mmap");
			exit(-1);
		}
		ret = munmap(addr, page_size);
		if (ret) {
			perror("thread-munmap");
			exit(-1);
		}
	}

	return NULL;
}

int main(int argc, char **argv)
{
	int nr_tasks = 12;
	int i;

	if (argc > 1)
		nr_tasks = atoi(argv[1]);
	if (argc > 2)
		nr_loops = atoi(argv[2]);

	page_size = getpagesize();

	file = open("/tmp/mmap-file-test", O_RDWR | O_CREAT);

	for (i = 0; i < nr_tasks; i++) {
		if (!fork()) {
			do_mmap_merge((void *)i);
			exit(0);
		}
	}

	for (i = 0; i < nr_tasks; i++)
		wait(NULL);
}

Re: [PATCH 15/20] mm: Convert i_mmap_lock to a mutex

[Peter Zijlstra]

On Tue, 2011-04-19 at 13:07 -0700, Andrew Morton wrote:
> On Fri, 01 Apr 2011 14:13:13 +0200
> Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:
> 
> > Straight fwd conversion of i_mmap_lock to a mutex
> 
> What effect does this have on kernel performance?
> 
> Workloads which take the lock at high frequency from multiple threads
> should be tested.

The worst affected workload would be a lightly contended lock, uncontended
mutexes are similarly fast as uncontended spinlocks, highly contended mutexes
win hands down since we waste vastly less resources spinning, leaving
lightly contended.

The below shows a workload tailored to isolate this one lock and is tested with
various numbers of contending tasks ran on a dual socket westmere EP (2*6*2),
the first results are from an unpatched -tip kernel, the second set is with
the patches applied:


	48		24		12		6		3

real    0m33.973s	0m17.816s	0m8.762s	0m3.778s	0m1.307s
sys     13m20.203s	7m5.814s	1m44.440s	0m22.371s	0m3.751s

real    0m23.551s	0m12.265s	0m7.016s	0m3.644s	0m1.412s
sys     1m17.177s	0m41.875s	0m26.139s	0m13.872s	0m3.873s

It looks like we might have a slight loss at 3 tasks contending although I'd
have to run with larger sets in order to increase the confidence interval as
the current numbers are within each others error range.

Also, this conversion is needed to allow the anon_vma->lock conversion, which
shows an improvement over the whole range.

---

[root@westmere ~]# echo 0 > /proc/sys/kernel/nmi_watchdog 
[root@westmere ~]# for i in /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor ; do echo performance > $i; done
[root@westmere ~]# time perf stat --repeat 10 ./mmap-file 48 50000

 Performance counter stats for './mmap-file 48 50000' (10 runs):

          3.045659 task-clock-msecs         #      0.001 CPUs    ( +-   0.481% )
                48 context-switches         #      0.016 M/sec   ( +-   0.566% )
                69 CPU-migrations           #      0.023 M/sec   ( +-   1.456% )
               250 page-faults              #      0.082 M/sec   ( +-   0.040% )
         7,985,758 cycles                   #   2622.013 M/sec   ( +-   0.413% )
         3,335,330 instructions             #      0.418 IPC     ( +-   0.198% )
           614,795 branches                 #    201.859 M/sec   ( +-   0.175% )
            20,177 branch-misses            #      3.282 %       ( +-   1.299% )
           123,596 cache-references         #     40.581 M/sec   ( +-   0.693% )
            57,380 cache-misses             #     18.840 M/sec   ( +-   2.148% )

        3.380886384  seconds time elapsed   ( +-   0.159% )


real    0m33.973s
user    0m1.463s
sys     13m20.203s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-file 24 50000

 Performance counter stats for './mmap-file 24 50000' (10 runs):

          1.511615 task-clock-msecs         #      0.001 CPUs    ( +-   1.049% )
                16 context-switches         #      0.011 M/sec   ( +-   3.040% )
                23 CPU-migrations           #      0.015 M/sec   ( +-   0.437% )
               178 page-faults              #      0.118 M/sec   ( +-   0.120% )
         4,003,891 cycles                   #   2648.750 M/sec   ( +-   0.988% )
         1,767,231 instructions             #      0.441 IPC     ( +-   0.678% )
           328,415 branches                 #    217.261 M/sec   ( +-   0.707% )
             9,281 branch-misses            #      2.826 %       ( +-   1.045% )
            56,897 cache-references         #     37.640 M/sec   ( +-   0.966% )
            26,209 cache-misses             #     17.339 M/sec   ( +-   2.700% )

        1.779547040  seconds time elapsed   ( +-   0.116% )


real    0m17.816s
user    0m0.704s
sys     7m5.814s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-file 12 50000

 Performance counter stats for './mmap-file 12 50000' (10 runs):

          0.833400 task-clock-msecs         #      0.001 CPUs    ( +-   2.223% )
                 8 context-switches         #      0.010 M/sec   ( +-   4.047% )
                12 CPU-migrations           #      0.014 M/sec   ( +-   0.000% )
               142 page-faults              #      0.170 M/sec   ( +-   0.105% )
         2,218,792 cycles                   #   2662.338 M/sec   ( +-   2.025% )
         1,091,917 instructions             #      0.492 IPC     ( +-   0.558% )
           205,310 branches                 #    246.352 M/sec   ( +-   0.516% )
             6,030 branch-misses            #      2.937 %       ( +-   1.436% )
            32,260 cache-references         #     38.710 M/sec   ( +-   0.747% )
            12,526 cache-misses             #     15.030 M/sec   ( +-   2.783% )

        0.874054257  seconds time elapsed   ( +-   0.122% )


real    0m8.762s
user    0m0.283s
sys     1m44.440s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-file 6 50000

 Performance counter stats for './mmap-file 6 50000' (10 runs):

          0.525969 task-clock-msecs         #      0.001 CPUs    ( +-   1.705% )
                 3 context-switches         #      0.006 M/sec   ( +-  11.994% )
                 6 CPU-migrations           #      0.011 M/sec   ( +-   0.000% )
               124 page-faults              #      0.236 M/sec   ( +-   0.000% )
         1,413,250 cycles                   #   2686.948 M/sec   ( +-   1.138% )
           757,501 instructions             #      0.536 IPC     ( +-   0.565% )
           144,120 branches                 #    274.009 M/sec   ( +-   0.560% )
             5,098 branch-misses            #      3.538 %       ( +-   2.054% )
            22,782 cache-references         #     43.315 M/sec   ( +-   1.068% )
             6,159 cache-misses             #     11.709 M/sec   ( +-   5.052% )

        0.375598224  seconds time elapsed   ( +-   0.560% )


real    0m3.778s
user    0m0.100s
sys     0m22.371s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-file 3 50000

 Performance counter stats for './mmap-file 3 50000' (10 runs):

          0.426419 task-clock-msecs         #      0.003 CPUs    ( +-   2.223% )
                 3 context-switches         #      0.007 M/sec   ( +-   3.448% )
                 3 CPU-migrations           #      0.007 M/sec   ( +-   0.000% )
               115 page-faults              #      0.270 M/sec   ( +-   0.132% )
         1,134,258 cycles                   #   2659.963 M/sec   ( +-   1.587% )
           610,569 instructions             #      0.538 IPC     ( +-   0.289% )
           117,666 branches                 #    275.940 M/sec   ( +-   0.320% )
             4,318 branch-misses            #      3.670 %       ( +-   3.292% )
            17,922 cache-references         #     42.029 M/sec   ( +-   0.819% )
             3,905 cache-misses             #      9.157 M/sec   ( +-   7.969% )

        0.127845444  seconds time elapsed   ( +-   1.226% )


real    0m1.307s
user    0m0.158s
sys     0m3.751s



---

[root@westmere ~]# echo 0 > /proc/sys/kernel/nmi_watchdog 
[root@westmere ~]# for i in /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor ; do echo performance > $i; done
[root@westmere ~]# time perf stat --repeat 10 ./mmap-file 48 50000

 Performance counter stats for './mmap-file 48 50000' (10 runs):

          2.910655 task-clock-msecs         #      0.001 CPUs    ( +-   0.580% )
                48 context-switches         #      0.016 M/sec   ( +-   0.279% )
                52 CPU-migrations           #      0.018 M/sec   ( +-   1.225% )
               250 page-faults              #      0.086 M/sec   ( +-   0.053% )
         7,538,292 cycles                   #   2589.895 M/sec   ( +-   0.725% )
         3,285,984 instructions             #      0.436 IPC     ( +-   0.248% )
           608,888 branches                 #    209.193 M/sec   ( +-   0.256% )
            17,153 branch-misses            #      2.817 %       ( +-   0.944% )
           128,392 cache-references         #     44.111 M/sec   ( +-   0.816% )
            57,160 cache-misses             #     19.638 M/sec   ( +-   2.574% )

        2.352105837  seconds time elapsed   ( +-   0.729% )


real    0m23.551s
user    0m1.451s
sys     1m17.177s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-file 24 50000

 Performance counter stats for './mmap-file 24 50000' (10 runs):

          1.600378 task-clock-msecs         #      0.001 CPUs    ( +-   1.531% )
                24 context-switches         #      0.015 M/sec   ( +-   0.745% )
                25 CPU-migrations           #      0.015 M/sec   ( +-   0.806% )
               178 page-faults              #      0.111 M/sec   ( +-   0.118% )
         4,112,616 cycles                   #   2569.778 M/sec   ( +-   1.250% )
         1,839,878 instructions             #      0.447 IPC     ( +-   0.307% )
           343,196 branches                 #    214.447 M/sec   ( +-   0.361% )
            10,876 branch-misses            #      3.169 %       ( +-   1.514% )
            68,812 cache-references         #     42.998 M/sec   ( +-   1.175% )
            26,286 cache-misses             #     16.425 M/sec   ( +-   2.509% )

        1.224012964  seconds time elapsed   ( +-   1.200% )


real    0m12.265s
user    0m0.685s
sys     0m41.875s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-file 12 50000

 Performance counter stats for './mmap-file 12 50000' (10 runs):

          0.833480 task-clock-msecs         #      0.001 CPUs    ( +-   1.324% )
                12 context-switches         #      0.014 M/sec   ( +-   0.000% )
                12 CPU-migrations           #      0.014 M/sec   ( +-   0.000% )
               142 page-faults              #      0.170 M/sec   ( +-   0.126% )
         2,180,779 cycles                   #   2616.476 M/sec   ( +-   1.408% )
         1,127,027 instructions             #      0.517 IPC     ( +-   0.354% )
           211,726 branches                 #    254.026 M/sec   ( +-   0.327% )
             6,628 branch-misses            #      3.131 %       ( +-   1.522% )
            36,986 cache-references         #     44.376 M/sec   ( +-   1.459% )
            12,337 cache-misses             #     14.802 M/sec   ( +-   3.203% )

        0.699308658  seconds time elapsed   ( +-   0.931% )


real    0m7.016s
user    0m0.323s
sys     0m26.139s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-file 6 50000

 Performance counter stats for './mmap-file 6 50000' (10 runs):

          0.551046 task-clock-msecs         #      0.002 CPUs    ( +-   2.031% )
                 6 context-switches         #      0.011 M/sec   ( +-   3.042% )
                 6 CPU-migrations           #      0.011 M/sec   ( +-   0.000% )
               124 page-faults              #      0.226 M/sec   ( +-   0.123% )
         1,482,359 cycles                   #   2690.081 M/sec   ( +-   1.767% )
           793,407 instructions             #      0.535 IPC     ( +-   0.387% )
           150,806 branches                 #    273.673 M/sec   ( +-   0.416% )
             5,649 branch-misses            #      3.746 %       ( +-   2.961% )
            24,554 cache-references         #     44.559 M/sec   ( +-   1.985% )
             5,632 cache-misses             #     10.220 M/sec   ( +-   4.339% )

        0.362139074  seconds time elapsed   ( +-   1.599% )


real    0m3.644s
user    0m0.153s
sys     0m13.872s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-file 3 50000

 Performance counter stats for './mmap-file 3 50000' (10 runs):

          0.419324 task-clock-msecs         #      0.003 CPUs    ( +-   2.667% )
                 3 context-switches         #      0.007 M/sec   ( +-   0.000% )
                 3 CPU-migrations           #      0.007 M/sec   ( +-   0.000% )
               116 page-faults              #      0.276 M/sec   ( +-   0.264% )
         1,147,068 cycles                   #   2735.515 M/sec   ( +-   2.253% )
           623,614 instructions             #      0.544 IPC     ( +-   0.228% )
           119,060 branches                 #    283.933 M/sec   ( +-   0.207% )
             4,679 branch-misses            #      3.930 %       ( +-   2.200% )
            17,917 cache-references         #     42.729 M/sec   ( +-   1.021% )
             3,779 cache-misses             #      9.013 M/sec   ( +-   3.425% )

        0.138764629  seconds time elapsed   ( +-   0.903% )


real    0m1.412s
user    0m0.065s
sys     0m3.873s


---
#include <unistd.h>
#include <stdio.h>
#include <stdlib.h>
#include <sys/mman.h>
#include <pthread.h>
#include <errno.h>
#include <sys/types.h>
#include <sys/wait.h>
#include <fcntl.h>


unsigned long nr_loops = 1000000; /* 1M */
unsigned long page_size;
int file;

void *do_mmap_merge(void *data)
{
	int nr = (unsigned long)data;
	void *addr;
	int ret;
	int i;

	for (i = 0; i < nr_loops; i++) {
		addr = mmap(NULL, page_size, PROT_READ | PROT_WRITE, 
				MAP_SHARED, file, 0);
		if (addr == MAP_FAILED) {
			perror("thread-mmap");
			exit(-1);
		}
		ret = munmap(addr, page_size);
		if (ret) {
			perror("thread-munmap");
			exit(-1);
		}
	}

	return NULL;
}

int main(int argc, char **argv)
{
	int nr_tasks = 12;
	int i;

	if (argc > 1)
		nr_tasks = atoi(argv[1]);
	if (argc > 2)
		nr_loops = atoi(argv[2]);

	page_size = getpagesize();

	file = open("/tmp/mmap-file-test", O_RDWR | O_CREAT);

	for (i = 0; i < nr_tasks; i++) {
		if (!fork()) {
			do_mmap_merge((void *)i);
			exit(0);
		}
	}

	for (i = 0; i < nr_tasks; i++)
		wait(NULL);
}



--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 19/20] mm: Convert anon_vma->lock to a mutex

[Peter Zijlstra]

On Tue, 2011-04-19 at 13:07 -0700, Andrew Morton wrote:
> On Fri, 01 Apr 2011 14:13:17 +0200
> Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:
> 
> > Straight fwd conversion of anon_vma->lock to a mutex.
> 
> What workloads do we expect might be adversely affected by this? 
> Were such workloads tested?  With what results?


The worst affected workload would be a lightly contended lock, uncontended
mutexes are similarly fast as uncontended spinlocks, highly contended mutexes
win hands down since we waste vastly less resources spinning, leaving
lightly contended.

The below shows a workload tailored to isolate this one lock and is tested with
various numbers of contending tasks ran on a dual socket westmere EP (2*6*2),
the first results are from an unpatched -tip kernel, the second set is with
the patches applied:


	48		24		12		6		3

real    2m14.152s	1m7.976s	0m32.607s	0m13.593s	0m5.464s
sys     53m1.628s	27m7.502s	6m30.265s	1m22.047s	0m16.693s

real    1m14.742s	0m34.823s	0m26.320s	0m12.647s	0m4.896s
sys     4m51.984s	2m7.259s	2m11.851s	0m52.916s	0m13.610s



---

[root@westmere ~]# echo 0 > /proc/sys/kernel/nmi_watchdog 
[root@westmere ~]# for i in /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor ; do echo performance > $i; done
[root@westmere ~]# time perf stat --repeat 10 ./mmap-merge 48 50000

 Performance counter stats for './mmap-merge 48 50000' (10 runs):

          4.142580 task-clock-msecs         #      0.000 CPUs    ( +-   2.738% )
                48 context-switches         #      0.012 M/sec   ( +-   0.000% )
                65 CPU-migrations           #      0.016 M/sec   ( +-   1.357% )
               250 page-faults              #      0.060 M/sec   ( +-   0.040% )
        10,918,844 cycles                   #   2635.759 M/sec   ( +-   2.759% )
         3,697,498 instructions             #      0.339 IPC     ( +-   3.023% )
           756,868 branches                 #    182.705 M/sec   ( +-   5.291% )
            21,012 branch-misses            #      2.776 %       ( +-   1.964% )
           126,172 cache-references         #     30.457 M/sec   ( +-   1.303% )
            53,508 cache-misses             #     12.917 M/sec   ( +-   2.845% )

       13.368707820  seconds time elapsed   ( +-   0.572% )


real    2m14.152s
user    0m0.786s
sys     53m1.628s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-merge 24 50000

 Performance counter stats for './mmap-merge 24 50000' (10 runs):

          1.766450 task-clock-msecs         #      0.000 CPUs    ( +-   1.300% )
                22 context-switches         #      0.012 M/sec   ( +-   1.174% )
                23 CPU-migrations           #      0.013 M/sec   ( +-   0.000% )
               178 page-faults              #      0.101 M/sec   ( +-   0.101% )
         4,692,314 cycles                   #   2656.353 M/sec   ( +-   1.367% )
         1,911,120 instructions             #      0.407 IPC     ( +-   0.578% )
           374,376 branches                 #    211.937 M/sec   ( +-   0.588% )
            10,575 branch-misses            #      2.825 %       ( +-   1.778% )
            63,062 cache-references         #     35.700 M/sec   ( +-   1.400% )
            20,580 cache-misses             #     11.650 M/sec   ( +-   7.064% )

        6.795076582  seconds time elapsed   ( +-   0.546% )


real    1m7.976s
user    0m0.361s
sys     27m7.502s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-merge 12 50000

 Performance counter stats for './mmap-merge 12 50000' (10 runs):

          0.854767 task-clock-msecs         #      0.000 CPUs    ( +-   1.354% )
                11 context-switches         #      0.013 M/sec   ( +-   3.030% )
                12 CPU-migrations           #      0.014 M/sec   ( +-   0.000% )
               142 page-faults              #      0.166 M/sec   ( +-   0.126% )
         2,264,106 cycles                   #   2648.798 M/sec   ( +-   1.285% )
         1,138,411 instructions             #      0.503 IPC     ( +-   0.814% )
           217,651 branches                 #    254.631 M/sec   ( +-   0.740% )
             6,632 branch-misses            #      3.047 %       ( +-   1.857% )
            34,727 cache-references         #     40.628 M/sec   ( +-   1.217% )
            10,195 cache-misses             #     11.927 M/sec   ( +-   6.837% )

        3.258410151  seconds time elapsed   ( +-   0.415% )


real    0m32.607s
user    0m0.222s
sys     6m30.265s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-merge 6 50000

 Performance counter stats for './mmap-merge 6 50000' (10 runs):

          0.554700 task-clock-msecs         #      0.000 CPUs    ( +-   2.391% )
                 5 context-switches         #      0.010 M/sec   ( +-   5.660% )
                 6 CPU-migrations           #      0.011 M/sec   ( +-   0.000% )
               124 page-faults              #      0.224 M/sec   ( +-   0.107% )
         1,497,189 cycles                   #   2699.096 M/sec   ( +-   1.941% )
           794,179 instructions             #      0.530 IPC     ( +-   0.930% )
           151,675 branches                 #    273.436 M/sec   ( +-   0.903% )
             5,795 branch-misses            #      3.820 %       ( +-   2.436% )
            24,981 cache-references         #     45.036 M/sec   ( +-   1.101% )
             5,931 cache-misses             #     10.693 M/sec   ( +-   4.373% )

        1.356908302  seconds time elapsed   ( +-   0.981% )


real    0m13.593s
user    0m0.104s
sys     1m22.047s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-merge 3 50000

 Performance counter stats for './mmap-merge 3 50000' (10 runs):

          0.413204 task-clock-msecs         #      0.001 CPUs    ( +-   3.120% )
                 3 context-switches         #      0.007 M/sec   ( +-   4.762% )
                 3 CPU-migrations           #      0.007 M/sec   ( +-   0.000% )
               115 page-faults              #      0.279 M/sec   ( +-   0.203% )
         1,132,363 cycles                   #   2740.446 M/sec   ( +-   2.553% )
           609,942 instructions             #      0.539 IPC     ( +-   0.778% )
           117,542 branches                 #    284.464 M/sec   ( +-   0.762% )
             5,000 branch-misses            #      4.254 %       ( +-   2.445% )
            17,825 cache-references         #     43.138 M/sec   ( +-   1.869% )
             3,358 cache-misses             #      8.126 M/sec   ( +-   7.906% )

        0.543944086  seconds time elapsed   ( +-   0.846% )


real    0m5.464s
user    0m0.052s
sys     0m16.693s


----

[root@westmere ~]# echo 0 > /proc/sys/kernel/nmi_watchdog 
[root@westmere ~]# for i in /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor ; do echo performance > $i; done
[root@westmere ~]# time perf stat --repeat 10 ./mmap-merge 48 50000

 Performance counter stats for './mmap-merge 48 50000' (10 runs):

          3.707111 task-clock-msecs         #      0.000 CPUs    ( +-   3.961% )
                63 context-switches         #      0.017 M/sec   ( +-   8.853% )
                50 CPU-migrations           #      0.014 M/sec   ( +-   1.394% )
               250 page-faults              #      0.067 M/sec   ( +-   0.040% )
         9,564,633 cycles                   #   2580.078 M/sec   ( +-   4.053% )
         3,567,454 instructions             #      0.373 IPC     ( +-   1.245% )
           690,477 branches                 #    186.257 M/sec   ( +-   1.650% )
            21,207 branch-misses            #      3.071 %       ( +-   2.731% )
           142,269 cache-references         #     38.377 M/sec   ( +-   1.844% )
            58,300 cache-misses             #     15.727 M/sec   ( +-   1.927% )

        7.471369198  seconds time elapsed   ( +-  10.095% )


real    1m14.742s
user    0m1.825s
sys     4m51.984s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-merge 24 50000

 Performance counter stats for './mmap-merge 24 50000' (10 runs):

          1.757956 task-clock-msecs         #      0.001 CPUs    ( +-   2.702% )
                34 context-switches         #      0.019 M/sec   ( +-  10.810% )
                24 CPU-migrations           #      0.014 M/sec   ( +-   1.394% )
               178 page-faults              #      0.101 M/sec   ( +-   0.056% )
         4,554,679 cycles                   #   2590.894 M/sec   ( +-   2.721% )
         1,929,470 instructions             #      0.424 IPC     ( +-   0.882% )
           371,094 branches                 #    211.094 M/sec   ( +-   1.441% )
            11,699 branch-misses            #      3.152 %       ( +-   2.183% )
            71,996 cache-references         #     40.954 M/sec   ( +-   1.119% )
            25,416 cache-misses             #     14.458 M/sec   ( +-   3.267% )

        3.479702410  seconds time elapsed   ( +-   8.481% )


real    0m34.823s
user    0m0.820s
sys     2m7.259s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-merge 12 50000

 Performance counter stats for './mmap-merge 12 50000' (10 runs):

          0.920072 task-clock-msecs         #      0.000 CPUs    ( +-   2.413% )
                13 context-switches         #      0.014 M/sec   ( +-   3.553% )
                12 CPU-migrations           #      0.013 M/sec   ( +-   0.000% )
               142 page-faults              #      0.154 M/sec   ( +-   0.126% )
         2,453,089 cycles                   #   2666.192 M/sec   ( +-   2.294% )
         1,213,864 instructions             #      0.495 IPC     ( +-   4.309% )
           229,378 branches                 #    249.304 M/sec   ( +-   3.634% )
             7,582 branch-misses            #      3.306 %       ( +-   1.559% )
            39,421 cache-references         #     42.845 M/sec   ( +-   1.613% )
            13,022 cache-misses             #     14.153 M/sec   ( +-   2.974% )

        2.629273354  seconds time elapsed   ( +-  12.610% )


real    0m26.320s
user    0m0.355s
sys     2m11.851s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-merge 6 50000

 Performance counter stats for './mmap-merge 6 50000' (10 runs):

          0.574440 task-clock-msecs         #      0.000 CPUs    ( +-   2.073% )
                 7 context-switches         #      0.012 M/sec   ( +-   7.105% )
                 6 CPU-migrations           #      0.010 M/sec   ( +-   0.000% )
               124 page-faults              #      0.216 M/sec   ( +-   0.081% )
         1,538,323 cycles                   #   2677.954 M/sec   ( +-   2.011% )
           805,689 instructions             #      0.524 IPC     ( +-   0.777% )
           153,771 branches                 #    267.689 M/sec   ( +-   0.760% )
             5,833 branch-misses            #      3.793 %       ( +-   2.511% )
            24,123 cache-references         #     41.994 M/sec   ( +-   1.714% )
             5,552 cache-misses             #      9.666 M/sec   ( +-   6.832% )

        1.262533725  seconds time elapsed   ( +-  10.411% )


real    0m12.647s
user    0m0.148s
sys     0m52.916s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-merge 3 50000

 Performance counter stats for './mmap-merge 3 50000' (10 runs):

          0.412164 task-clock-msecs         #      0.001 CPUs    ( +-   1.961% )
                 3 context-switches         #      0.007 M/sec   ( +-   0.000% )
                 3 CPU-migrations           #      0.007 M/sec   ( +-   0.000% )
               115 page-faults              #      0.279 M/sec   ( +-   0.203% )
         1,130,757 cycles                   #   2743.460 M/sec   ( +-   1.623% )
           621,400 instructions             #      0.550 IPC     ( +-   0.266% )
           118,812 branches                 #    288.263 M/sec   ( +-   0.259% )
             4,885 branch-misses            #      4.112 %       ( +-   1.501% )
            17,666 cache-references         #     42.861 M/sec   ( +-   0.993% )
             3,588 cache-misses             #      8.706 M/sec   ( +-   5.443% )

        0.487055557  seconds time elapsed   ( +-   9.425% )


real    0m4.896s
user    0m0.080s
sys     0m13.610s



---
#include <unistd.h>
#include <stdio.h>
#include <stdlib.h>
#include <sys/mman.h>
#include <pthread.h>
#include <errno.h>
#include <sys/types.h>
#include <sys/wait.h>


unsigned long nr_loops = 1000000; /* 1M */
unsigned long page_size;
void *mmap_base;

void *do_mmap_merge(void *data)
{
	int nr = (unsigned long)data;
	void *page = mmap_base + (2 + nr * 5) * page_size;
	void *addr;
	int ret;
	int i;

	for (i = 0; i < nr_loops; i++) {
		ret = munmap(page, page_size);
		if (ret) {
			perror("thread-munmap");
			exit(-1);
		}
		addr = mmap(page, page_size, PROT_READ | PROT_WRITE, 
				MAP_PRIVATE | MAP_ANONYMOUS | MAP_FIXED, -1, 0);
		if (addr == MAP_FAILED) {
			perror("thread-mmap");
			exit(-1);
		}
	}

	return NULL;
}

int main(int argc, char **argv)
{
	int nr_tasks = 12;
	int i;

	if (argc > 1)
		nr_tasks = atoi(argv[1]);
	if (argc > 2)
		nr_loops = atoi(argv[2]);

	page_size = getpagesize();

	mmap_base = mmap(NULL, 5 * nr_tasks * page_size, PROT_READ | PROT_WRITE,
			MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
	if (mmap_base == MAP_FAILED) {
		perror("mmap");
		exit(-1);
	}

	for (i = 0; i < nr_tasks; i++) {
		if (!fork()) {
			do_mmap_merge((void *)i);
			exit(0);
		}
	}

	for (i = 0; i < nr_tasks; i++)
		wait(NULL);
}

Re: [PATCH 19/20] mm: Convert anon_vma->lock to a mutex

[Peter Zijlstra]

On Tue, 2011-04-19 at 13:07 -0700, Andrew Morton wrote:
> On Fri, 01 Apr 2011 14:13:17 +0200
> Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:
> 
> > Straight fwd conversion of anon_vma->lock to a mutex.
> 
> What workloads do we expect might be adversely affected by this? 
> Were such workloads tested?  With what results?


The worst affected workload would be a lightly contended lock, uncontended
mutexes are similarly fast as uncontended spinlocks, highly contended mutexes
win hands down since we waste vastly less resources spinning, leaving
lightly contended.

The below shows a workload tailored to isolate this one lock and is tested with
various numbers of contending tasks ran on a dual socket westmere EP (2*6*2),
the first results are from an unpatched -tip kernel, the second set is with
the patches applied:


	48		24		12		6		3

real    2m14.152s	1m7.976s	0m32.607s	0m13.593s	0m5.464s
sys     53m1.628s	27m7.502s	6m30.265s	1m22.047s	0m16.693s

real    1m14.742s	0m34.823s	0m26.320s	0m12.647s	0m4.896s
sys     4m51.984s	2m7.259s	2m11.851s	0m52.916s	0m13.610s



---

[root@westmere ~]# echo 0 > /proc/sys/kernel/nmi_watchdog 
[root@westmere ~]# for i in /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor ; do echo performance > $i; done
[root@westmere ~]# time perf stat --repeat 10 ./mmap-merge 48 50000

 Performance counter stats for './mmap-merge 48 50000' (10 runs):

          4.142580 task-clock-msecs         #      0.000 CPUs    ( +-   2.738% )
                48 context-switches         #      0.012 M/sec   ( +-   0.000% )
                65 CPU-migrations           #      0.016 M/sec   ( +-   1.357% )
               250 page-faults              #      0.060 M/sec   ( +-   0.040% )
        10,918,844 cycles                   #   2635.759 M/sec   ( +-   2.759% )
         3,697,498 instructions             #      0.339 IPC     ( +-   3.023% )
           756,868 branches                 #    182.705 M/sec   ( +-   5.291% )
            21,012 branch-misses            #      2.776 %       ( +-   1.964% )
           126,172 cache-references         #     30.457 M/sec   ( +-   1.303% )
            53,508 cache-misses             #     12.917 M/sec   ( +-   2.845% )

       13.368707820  seconds time elapsed   ( +-   0.572% )


real    2m14.152s
user    0m0.786s
sys     53m1.628s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-merge 24 50000

 Performance counter stats for './mmap-merge 24 50000' (10 runs):

          1.766450 task-clock-msecs         #      0.000 CPUs    ( +-   1.300% )
                22 context-switches         #      0.012 M/sec   ( +-   1.174% )
                23 CPU-migrations           #      0.013 M/sec   ( +-   0.000% )
               178 page-faults              #      0.101 M/sec   ( +-   0.101% )
         4,692,314 cycles                   #   2656.353 M/sec   ( +-   1.367% )
         1,911,120 instructions             #      0.407 IPC     ( +-   0.578% )
           374,376 branches                 #    211.937 M/sec   ( +-   0.588% )
            10,575 branch-misses            #      2.825 %       ( +-   1.778% )
            63,062 cache-references         #     35.700 M/sec   ( +-   1.400% )
            20,580 cache-misses             #     11.650 M/sec   ( +-   7.064% )

        6.795076582  seconds time elapsed   ( +-   0.546% )


real    1m7.976s
user    0m0.361s
sys     27m7.502s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-merge 12 50000

 Performance counter stats for './mmap-merge 12 50000' (10 runs):

          0.854767 task-clock-msecs         #      0.000 CPUs    ( +-   1.354% )
                11 context-switches         #      0.013 M/sec   ( +-   3.030% )
                12 CPU-migrations           #      0.014 M/sec   ( +-   0.000% )
               142 page-faults              #      0.166 M/sec   ( +-   0.126% )
         2,264,106 cycles                   #   2648.798 M/sec   ( +-   1.285% )
         1,138,411 instructions             #      0.503 IPC     ( +-   0.814% )
           217,651 branches                 #    254.631 M/sec   ( +-   0.740% )
             6,632 branch-misses            #      3.047 %       ( +-   1.857% )
            34,727 cache-references         #     40.628 M/sec   ( +-   1.217% )
            10,195 cache-misses             #     11.927 M/sec   ( +-   6.837% )

        3.258410151  seconds time elapsed   ( +-   0.415% )


real    0m32.607s
user    0m0.222s
sys     6m30.265s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-merge 6 50000

 Performance counter stats for './mmap-merge 6 50000' (10 runs):

          0.554700 task-clock-msecs         #      0.000 CPUs    ( +-   2.391% )
                 5 context-switches         #      0.010 M/sec   ( +-   5.660% )
                 6 CPU-migrations           #      0.011 M/sec   ( +-   0.000% )
               124 page-faults              #      0.224 M/sec   ( +-   0.107% )
         1,497,189 cycles                   #   2699.096 M/sec   ( +-   1.941% )
           794,179 instructions             #      0.530 IPC     ( +-   0.930% )
           151,675 branches                 #    273.436 M/sec   ( +-   0.903% )
             5,795 branch-misses            #      3.820 %       ( +-   2.436% )
            24,981 cache-references         #     45.036 M/sec   ( +-   1.101% )
             5,931 cache-misses             #     10.693 M/sec   ( +-   4.373% )

        1.356908302  seconds time elapsed   ( +-   0.981% )


real    0m13.593s
user    0m0.104s
sys     1m22.047s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-merge 3 50000

 Performance counter stats for './mmap-merge 3 50000' (10 runs):

          0.413204 task-clock-msecs         #      0.001 CPUs    ( +-   3.120% )
                 3 context-switches         #      0.007 M/sec   ( +-   4.762% )
                 3 CPU-migrations           #      0.007 M/sec   ( +-   0.000% )
               115 page-faults              #      0.279 M/sec   ( +-   0.203% )
         1,132,363 cycles                   #   2740.446 M/sec   ( +-   2.553% )
           609,942 instructions             #      0.539 IPC     ( +-   0.778% )
           117,542 branches                 #    284.464 M/sec   ( +-   0.762% )
             5,000 branch-misses            #      4.254 %       ( +-   2.445% )
            17,825 cache-references         #     43.138 M/sec   ( +-   1.869% )
             3,358 cache-misses             #      8.126 M/sec   ( +-   7.906% )

        0.543944086  seconds time elapsed   ( +-   0.846% )


real    0m5.464s
user    0m0.052s
sys     0m16.693s


----

[root@westmere ~]# echo 0 > /proc/sys/kernel/nmi_watchdog 
[root@westmere ~]# for i in /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor ; do echo performance > $i; done
[root@westmere ~]# time perf stat --repeat 10 ./mmap-merge 48 50000

 Performance counter stats for './mmap-merge 48 50000' (10 runs):

          3.707111 task-clock-msecs         #      0.000 CPUs    ( +-   3.961% )
                63 context-switches         #      0.017 M/sec   ( +-   8.853% )
                50 CPU-migrations           #      0.014 M/sec   ( +-   1.394% )
               250 page-faults              #      0.067 M/sec   ( +-   0.040% )
         9,564,633 cycles                   #   2580.078 M/sec   ( +-   4.053% )
         3,567,454 instructions             #      0.373 IPC     ( +-   1.245% )
           690,477 branches                 #    186.257 M/sec   ( +-   1.650% )
            21,207 branch-misses            #      3.071 %       ( +-   2.731% )
           142,269 cache-references         #     38.377 M/sec   ( +-   1.844% )
            58,300 cache-misses             #     15.727 M/sec   ( +-   1.927% )

        7.471369198  seconds time elapsed   ( +-  10.095% )


real    1m14.742s
user    0m1.825s
sys     4m51.984s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-merge 24 50000

 Performance counter stats for './mmap-merge 24 50000' (10 runs):

          1.757956 task-clock-msecs         #      0.001 CPUs    ( +-   2.702% )
                34 context-switches         #      0.019 M/sec   ( +-  10.810% )
                24 CPU-migrations           #      0.014 M/sec   ( +-   1.394% )
               178 page-faults              #      0.101 M/sec   ( +-   0.056% )
         4,554,679 cycles                   #   2590.894 M/sec   ( +-   2.721% )
         1,929,470 instructions             #      0.424 IPC     ( +-   0.882% )
           371,094 branches                 #    211.094 M/sec   ( +-   1.441% )
            11,699 branch-misses            #      3.152 %       ( +-   2.183% )
            71,996 cache-references         #     40.954 M/sec   ( +-   1.119% )
            25,416 cache-misses             #     14.458 M/sec   ( +-   3.267% )

        3.479702410  seconds time elapsed   ( +-   8.481% )


real    0m34.823s
user    0m0.820s
sys     2m7.259s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-merge 12 50000

 Performance counter stats for './mmap-merge 12 50000' (10 runs):

          0.920072 task-clock-msecs         #      0.000 CPUs    ( +-   2.413% )
                13 context-switches         #      0.014 M/sec   ( +-   3.553% )
                12 CPU-migrations           #      0.013 M/sec   ( +-   0.000% )
               142 page-faults              #      0.154 M/sec   ( +-   0.126% )
         2,453,089 cycles                   #   2666.192 M/sec   ( +-   2.294% )
         1,213,864 instructions             #      0.495 IPC     ( +-   4.309% )
           229,378 branches                 #    249.304 M/sec   ( +-   3.634% )
             7,582 branch-misses            #      3.306 %       ( +-   1.559% )
            39,421 cache-references         #     42.845 M/sec   ( +-   1.613% )
            13,022 cache-misses             #     14.153 M/sec   ( +-   2.974% )

        2.629273354  seconds time elapsed   ( +-  12.610% )


real    0m26.320s
user    0m0.355s
sys     2m11.851s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-merge 6 50000

 Performance counter stats for './mmap-merge 6 50000' (10 runs):

          0.574440 task-clock-msecs         #      0.000 CPUs    ( +-   2.073% )
                 7 context-switches         #      0.012 M/sec   ( +-   7.105% )
                 6 CPU-migrations           #      0.010 M/sec   ( +-   0.000% )
               124 page-faults              #      0.216 M/sec   ( +-   0.081% )
         1,538,323 cycles                   #   2677.954 M/sec   ( +-   2.011% )
           805,689 instructions             #      0.524 IPC     ( +-   0.777% )
           153,771 branches                 #    267.689 M/sec   ( +-   0.760% )
             5,833 branch-misses            #      3.793 %       ( +-   2.511% )
            24,123 cache-references         #     41.994 M/sec   ( +-   1.714% )
             5,552 cache-misses             #      9.666 M/sec   ( +-   6.832% )

        1.262533725  seconds time elapsed   ( +-  10.411% )


real    0m12.647s
user    0m0.148s
sys     0m52.916s
[root@westmere ~]# time perf stat --repeat 10 ./mmap-merge 3 50000

 Performance counter stats for './mmap-merge 3 50000' (10 runs):

          0.412164 task-clock-msecs         #      0.001 CPUs    ( +-   1.961% )
                 3 context-switches         #      0.007 M/sec   ( +-   0.000% )
                 3 CPU-migrations           #      0.007 M/sec   ( +-   0.000% )
               115 page-faults              #      0.279 M/sec   ( +-   0.203% )
         1,130,757 cycles                   #   2743.460 M/sec   ( +-   1.623% )
           621,400 instructions             #      0.550 IPC     ( +-   0.266% )
           118,812 branches                 #    288.263 M/sec   ( +-   0.259% )
             4,885 branch-misses            #      4.112 %       ( +-   1.501% )
            17,666 cache-references         #     42.861 M/sec   ( +-   0.993% )
             3,588 cache-misses             #      8.706 M/sec   ( +-   5.443% )

        0.487055557  seconds time elapsed   ( +-   9.425% )


real    0m4.896s
user    0m0.080s
sys     0m13.610s



---
#include <unistd.h>
#include <stdio.h>
#include <stdlib.h>
#include <sys/mman.h>
#include <pthread.h>
#include <errno.h>
#include <sys/types.h>
#include <sys/wait.h>


unsigned long nr_loops = 1000000; /* 1M */
unsigned long page_size;
void *mmap_base;

void *do_mmap_merge(void *data)
{
	int nr = (unsigned long)data;
	void *page = mmap_base + (2 + nr * 5) * page_size;
	void *addr;
	int ret;
	int i;

	for (i = 0; i < nr_loops; i++) {
		ret = munmap(page, page_size);
		if (ret) {
			perror("thread-munmap");
			exit(-1);
		}
		addr = mmap(page, page_size, PROT_READ | PROT_WRITE, 
				MAP_PRIVATE | MAP_ANONYMOUS | MAP_FIXED, -1, 0);
		if (addr == MAP_FAILED) {
			perror("thread-mmap");
			exit(-1);
		}
	}

	return NULL;
}

int main(int argc, char **argv)
{
	int nr_tasks = 12;
	int i;

	if (argc > 1)
		nr_tasks = atoi(argv[1]);
	if (argc > 2)
		nr_loops = atoi(argv[2]);

	page_size = getpagesize();

	mmap_base = mmap(NULL, 5 * nr_tasks * page_size, PROT_READ | PROT_WRITE,
			MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
	if (mmap_base == MAP_FAILED) {
		perror("mmap");
		exit(-1);
	}

	for (i = 0; i < nr_tasks; i++) {
		if (!fork()) {
			do_mmap_merge((void *)i);
			exit(0);
		}
	}

	for (i = 0; i < nr_tasks; i++)
		wait(NULL);
}


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 14/20] mm: Remove i_mmap_lock lockbreak

[Peter Zijlstra]

On Tue, 2011-04-19 at 13:07 -0700, Andrew Morton wrote:
> On Fri, 01 Apr 2011 14:13:12 +0200
> Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:
> 
> > Hugh says:
> >  "The only significant loser, I think, would be page reclaim (when
> >   concurrent with truncation): could spin for a long time waiting for
> >   the i_mmap_mutex it expects would soon be dropped? "
> > 
> > Counter points:
> >  - cpu contention makes the spin stop (need_resched())
> >  - zap pages should be freeing pages at a higher rate than reclaim
> >    ever can
> > 
> > I think the simplification of the truncate code is definately worth it.
> 
> Well, we don't need to guess.  These things are testable!

I suppose you're right, but I'm having a bit of a hard time coming up
with a sensible (reproducible) test case for the page reclaim part of
this problem set.

I'll try running 3 cyclic file scanners sized such that 2 exceed the
memory footprint of the machine and truncate the 3rd's file after
warming up.

That is, unless someone has a saner idea..

Re: [PATCH 14/20] mm: Remove i_mmap_lock lockbreak

[Peter Zijlstra]

On Tue, 2011-04-19 at 13:07 -0700, Andrew Morton wrote:
> On Fri, 01 Apr 2011 14:13:12 +0200
> Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:
> 
> > Hugh says:
> >  "The only significant loser, I think, would be page reclaim (when
> >   concurrent with truncation): could spin for a long time waiting for
> >   the i_mmap_mutex it expects would soon be dropped? "
> > 
> > Counter points:
> >  - cpu contention makes the spin stop (need_resched())
> >  - zap pages should be freeing pages at a higher rate than reclaim
> >    ever can
> > 
> > I think the simplification of the truncate code is definately worth it.
> 
> Well, we don't need to guess.  These things are testable!

I suppose you're right, but I'm having a bit of a hard time coming up
with a sensible (reproducible) test case for the page reclaim part of
this problem set.

I'll try running 3 cyclic file scanners sized such that 2 exceed the
memory footprint of the machine and truncate the 3rd's file after
warming up.

That is, unless someone has a saner idea..

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>