[PATCH -v2 00/10] Rewrite Motorola MMU page-table layout

[PATCH -v2 00/10] Rewrite Motorola MMU page-table layout

[Peter Zijlstra]

Hi!

In order to faciliate Will's READ_ONCE() patches:

  https://lkml.kernel.org/r/20200123153341.19947-1-will@kernel.org

we need to fix m68k/motorola to not have a giant pmd_t. These patches do so and
are tested using ARAnyM/68040.

Michael tested the previous version on his Atari Falcon/68030.

Build tested for sun3/coldfire.

Please consider!

Changes since -v1:
 - fixed sun3/coldfire build issues
 - unified motorola mmu page setup
 - added enum to table allocator
 - moved pointer table allocator to motorola.c
 - converted coldfire pgtable_t
 - fixed coldfire pgd_alloc
 - fixed coldfire nocache

---
 arch/m68k/include/asm/mcf_pgalloc.h      |  31 ++---
 arch/m68k/include/asm/motorola_pgalloc.h |  74 ++++------
 arch/m68k/include/asm/motorola_pgtable.h |  36 +++--
 arch/m68k/include/asm/page.h             |  16 ++-
 arch/m68k/include/asm/pgtable_mm.h       |  10 +-
 arch/m68k/mm/init.c                      |  34 +++--
 arch/m68k/mm/kmap.c                      |  36 +++--
 arch/m68k/mm/memory.c                    | 103 --------------
 arch/m68k/mm/motorola.c                  | 228 +++++++++++++++++++++++++------
 9 files changed, 302 insertions(+), 266 deletions(-)

[PATCH -v2 01/10] m68k,mm: Remove stray nocache in ColdFire pgalloc

[Peter Zijlstra]

Since ColdFire V4e is a software TLB-miss architecture, there is no
need for page-tables to be mapped uncached. Remove this stray
nocache_page() dance, which isn't paired with a cache_page() and looks
like a copy/paste/edit fail.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
---
 arch/m68k/include/asm/mcf_pgalloc.h |    6 +-----
 1 file changed, 1 insertion(+), 5 deletions(-)

--- a/arch/m68k/include/asm/mcf_pgalloc.h
+++ b/arch/m68k/include/asm/mcf_pgalloc.h
@@ -55,12 +55,8 @@ static inline struct page *pte_alloc_one
 	}
 
 	pte = kmap(page);
-	if (pte) {
+	if (pte)
 		clear_page(pte);
-		__flush_page_to_ram(pte);
-		flush_tlb_kernel_page(pte);
-		nocache_page(pte);
-	}
 	kunmap(page);
 
 	return page;

[PATCH -v2 02/10] m68k,mm: Fix ColdFire pgd_alloc()

[Peter Zijlstra]

From: Will Deacon <will@kernel.org>

I also notice that building for m5475evb_defconfig with vanilla v5.5
triggers this scary looking warning due to a mismatch between the pgd
size and the (8k!) page size:

 | In function 'pgd_alloc.isra.111',
 |     inlined from 'mm_alloc_pgd' at kernel/fork.c:634:12,
 |     inlined from 'mm_init.isra.112' at kernel/fork.c:1043:6:
 | ./arch/m68k/include/asm/string.h:72:25: warning: '__builtin_memcpy' forming offset [4097, 8192] is out of the bounds [0, 4096] of object 'kernel_pg_dir' with type 'pgd_t[1024]' {aka 'struct <anonymous>[1024]'} [-Warray-bounds]
 |  #define memcpy(d, s, n) __builtin_memcpy(d, s, n)
 |                          ^~~~~~~~~~~~~~~~~~~~~~~~~
 | ./arch/m68k/include/asm/mcf_pgalloc.h:93:2: note: in expansion of macro 'memcpy'
 |   memcpy(new_pgd, swapper_pg_dir, PAGE_SIZE);
 |   ^~~~~~

Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
---
 arch/m68k/include/asm/mcf_pgalloc.h |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

--- a/arch/m68k/include/asm/mcf_pgalloc.h
+++ b/arch/m68k/include/asm/mcf_pgalloc.h
@@ -90,7 +90,7 @@ static inline pgd_t *pgd_alloc(struct mm
 	new_pgd = (pgd_t *)__get_free_page(GFP_DMA | __GFP_NOWARN);
 	if (!new_pgd)
 		return NULL;
-	memcpy(new_pgd, swapper_pg_dir, PAGE_SIZE);
+	memcpy(new_pgd, swapper_pg_dir, PTRS_PER_PGD * sizeof(pgd_t));
 	memset(new_pgd, 0, PAGE_OFFSET >> PGDIR_SHIFT);
 	return new_pgd;
 }

[PATCH -v2 03/10] m68k,mm: Unify Motorola MMU page setup

[Peter Zijlstra]

Seeing how there are 5 copies of this magic code, one of which is
unexplainably different, unify and document things.

Suggested-by: Will Deacon <will@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
---
 arch/m68k/include/asm/motorola_pgalloc.h |   23 ++++++++++-------------
 arch/m68k/mm/memory.c                    |    5 ++---
 arch/m68k/mm/motorola.c                  |   30 ++++++++++++++++++++++++------
 3 files changed, 36 insertions(+), 22 deletions(-)

--- a/arch/m68k/include/asm/motorola_pgalloc.h
+++ b/arch/m68k/include/asm/motorola_pgalloc.h
@@ -5,6 +5,9 @@
 #include <asm/tlb.h>
 #include <asm/tlbflush.h>
 
+extern void mmu_page_ctor(void *page);
+extern void mmu_page_dtor(void *page);
+
 extern pmd_t *get_pointer_table(void);
 extern int free_pointer_table(pmd_t *);
 
@@ -13,25 +16,21 @@ static inline pte_t *pte_alloc_one_kerne
 	pte_t *pte;
 
 	pte = (pte_t *)__get_free_page(GFP_KERNEL|__GFP_ZERO);
-	if (pte) {
-		__flush_page_to_ram(pte);
-		flush_tlb_kernel_page(pte);
-		nocache_page(pte);
-	}
+	if (pte)
+		mmu_page_ctor(pte);
 
 	return pte;
 }
 
 static inline void pte_free_kernel(struct mm_struct *mm, pte_t *pte)
 {
-	cache_page(pte);
+	mmu_page_dtor(pte);
 	free_page((unsigned long) pte);
 }
 
 static inline pgtable_t pte_alloc_one(struct mm_struct *mm)
 {
 	struct page *page;
-	pte_t *pte;
 
 	page = alloc_pages(GFP_KERNEL|__GFP_ZERO, 0);
 	if(!page)
@@ -41,18 +40,16 @@ static inline pgtable_t pte_alloc_one(st
 		return NULL;
 	}
 
-	pte = kmap(page);
-	__flush_page_to_ram(pte);
-	flush_tlb_kernel_page(pte);
-	nocache_page(pte);
+	mmu_page_ctor(kmap(page));
 	kunmap(page);
+
 	return page;
 }
 
 static inline void pte_free(struct mm_struct *mm, pgtable_t page)
 {
 	pgtable_pte_page_dtor(page);
-	cache_page(kmap(page));
+	mmu_page_dtor(kmap(page));
 	kunmap(page);
 	__free_page(page);
 }
@@ -61,7 +58,7 @@ static inline void __pte_free_tlb(struct
 				  unsigned long address)
 {
 	pgtable_pte_page_dtor(page);
-	cache_page(kmap(page));
+	mmu_page_dtor(kmap(page));
 	kunmap(page);
 	__free_page(page);
 }
--- a/arch/m68k/mm/memory.c
+++ b/arch/m68k/mm/memory.c
@@ -77,8 +77,7 @@ pmd_t *get_pointer_table (void)
 		if (!(page = (void *)get_zeroed_page(GFP_KERNEL)))
 			return NULL;
 
-		flush_tlb_kernel_page(page);
-		nocache_page(page);
+		mmu_page_ctor(page);
 
 		new = PD_PTABLE(page);
 		PD_MARKBITS(new) = 0xfe;
@@ -112,7 +111,7 @@ int free_pointer_table (pmd_t *ptable)
 	if (PD_MARKBITS(dp) == 0xff) {
 		/* all tables in page are free, free page */
 		list_del(dp);
-		cache_page((void *)page);
+		mmu_page_dtor((void *)page);
 		free_page (page);
 		return 1;
 	} else if (ptable_list.next != dp) {
--- a/arch/m68k/mm/motorola.c
+++ b/arch/m68k/mm/motorola.c
@@ -45,6 +45,28 @@ unsigned long mm_cachebits;
 EXPORT_SYMBOL(mm_cachebits);
 #endif
 
+
+/*
+ * Motorola 680x0 user's manual recommends using uncached memory for address
+ * translation tables.
+ *
+ * Seeing how the MMU can be external on (some of) these chips, that seems like
+ * a very important recommendation to follow. Provide some helpers to combat
+ * 'variation' amongst the users of this.
+ */
+
+void mmu_page_ctor(void *page)
+{
+	__flush_page_to_ram(page);
+	flush_tlb_kernel_page(page);
+	nocache_page(page);
+}
+
+void mmu_page_dtor(void *page)
+{
+	cache_page(page);
+}
+
 /* size of memory already mapped in head.S */
 extern __initdata unsigned long m68k_init_mapped_size;
 
@@ -60,9 +82,7 @@ static pte_t * __init kernel_page_table(
 		      __func__, PAGE_SIZE, PAGE_SIZE);
 
 	clear_page(ptablep);
-	__flush_page_to_ram(ptablep);
-	flush_tlb_kernel_page(ptablep);
-	nocache_page(ptablep);
+	mmu_page_ctor(ptablep);
 
 	return ptablep;
 }
@@ -106,9 +126,7 @@ static pmd_t * __init kernel_ptr_table(v
 			      __func__, PAGE_SIZE, PAGE_SIZE);
 
 		clear_page(last_pgtable);
-		__flush_page_to_ram(last_pgtable);
-		flush_tlb_kernel_page(last_pgtable);
-		nocache_page(last_pgtable);
+		mmu_page_ctor(last_pgtable);
 	}
 
 	return last_pgtable;

[PATCH -v2 04/10] m68k,mm: Move the pointer table allocator to motorola.c

[Peter Zijlstra]

Only the Motorola MMU makes use of this allocator, it is a waste of
.text to include it for Sun3/ColdFire. Also, this is going to avoid
build issues when we're going to make it more Motorola specific.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
---
 arch/m68k/mm/memory.c   |  102 ------------------------------------------------
 arch/m68k/mm/motorola.c |  102 ++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 102 insertions(+), 102 deletions(-)

--- a/arch/m68k/mm/memory.c
+++ b/arch/m68k/mm/memory.c
@@ -22,108 +22,6 @@
 #include <asm/machdep.h>
 
 
-/* ++andreas: {get,free}_pointer_table rewritten to use unused fields from
-   struct page instead of separately kmalloced struct.  Stolen from
-   arch/sparc/mm/srmmu.c ... */
-
-typedef struct list_head ptable_desc;
-static LIST_HEAD(ptable_list);
-
-#define PD_PTABLE(page) ((ptable_desc *)&(virt_to_page(page)->lru))
-#define PD_PAGE(ptable) (list_entry(ptable, struct page, lru))
-#define PD_MARKBITS(dp) (*(unsigned char *)&PD_PAGE(dp)->index)
-
-#define PTABLE_SIZE (PTRS_PER_PMD * sizeof(pmd_t))
-
-void __init init_pointer_table(unsigned long ptable)
-{
-	ptable_desc *dp;
-	unsigned long page = ptable & PAGE_MASK;
-	unsigned char mask = 1 << ((ptable - page)/PTABLE_SIZE);
-
-	dp = PD_PTABLE(page);
-	if (!(PD_MARKBITS(dp) & mask)) {
-		PD_MARKBITS(dp) = 0xff;
-		list_add(dp, &ptable_list);
-	}
-
-	PD_MARKBITS(dp) &= ~mask;
-	pr_debug("init_pointer_table: %lx, %x\n", ptable, PD_MARKBITS(dp));
-
-	/* unreserve the page so it's possible to free that page */
-	__ClearPageReserved(PD_PAGE(dp));
-	init_page_count(PD_PAGE(dp));
-
-	return;
-}
-
-pmd_t *get_pointer_table (void)
-{
-	ptable_desc *dp = ptable_list.next;
-	unsigned char mask = PD_MARKBITS (dp);
-	unsigned char tmp;
-	unsigned int off;
-
-	/*
-	 * For a pointer table for a user process address space, a
-	 * table is taken from a page allocated for the purpose.  Each
-	 * page can hold 8 pointer tables.  The page is remapped in
-	 * virtual address space to be noncacheable.
-	 */
-	if (mask == 0) {
-		void *page;
-		ptable_desc *new;
-
-		if (!(page = (void *)get_zeroed_page(GFP_KERNEL)))
-			return NULL;
-
-		mmu_page_ctor(page);
-
-		new = PD_PTABLE(page);
-		PD_MARKBITS(new) = 0xfe;
-		list_add_tail(new, dp);
-
-		return (pmd_t *)page;
-	}
-
-	for (tmp = 1, off = 0; (mask & tmp) == 0; tmp <<= 1, off += PTABLE_SIZE)
-		;
-	PD_MARKBITS(dp) = mask & ~tmp;
-	if (!PD_MARKBITS(dp)) {
-		/* move to end of list */
-		list_move_tail(dp, &ptable_list);
-	}
-	return (pmd_t *) (page_address(PD_PAGE(dp)) + off);
-}
-
-int free_pointer_table (pmd_t *ptable)
-{
-	ptable_desc *dp;
-	unsigned long page = (unsigned long)ptable & PAGE_MASK;
-	unsigned char mask = 1 << (((unsigned long)ptable - page)/PTABLE_SIZE);
-
-	dp = PD_PTABLE(page);
-	if (PD_MARKBITS (dp) & mask)
-		panic ("table already free!");
-
-	PD_MARKBITS (dp) |= mask;
-
-	if (PD_MARKBITS(dp) == 0xff) {
-		/* all tables in page are free, free page */
-		list_del(dp);
-		mmu_page_dtor((void *)page);
-		free_page (page);
-		return 1;
-	} else if (ptable_list.next != dp) {
-		/*
-		 * move this descriptor to the front of the list, since
-		 * it has one or more free tables.
-		 */
-		list_move(dp, &ptable_list);
-	}
-	return 0;
-}
-
 /* invalidate page in both caches */
 static inline void clear040(unsigned long paddr)
 {
--- a/arch/m68k/mm/motorola.c
+++ b/arch/m68k/mm/motorola.c
@@ -67,6 +67,108 @@ void mmu_page_dtor(void *page)
 	cache_page(page);
 }
 
+/* ++andreas: {get,free}_pointer_table rewritten to use unused fields from
+   struct page instead of separately kmalloced struct.  Stolen from
+   arch/sparc/mm/srmmu.c ... */
+
+typedef struct list_head ptable_desc;
+static LIST_HEAD(ptable_list);
+
+#define PD_PTABLE(page) ((ptable_desc *)&(virt_to_page(page)->lru))
+#define PD_PAGE(ptable) (list_entry(ptable, struct page, lru))
+#define PD_MARKBITS(dp) (*(unsigned char *)&PD_PAGE(dp)->index)
+
+#define PTABLE_SIZE (PTRS_PER_PMD * sizeof(pmd_t))
+
+void __init init_pointer_table(unsigned long ptable)
+{
+	ptable_desc *dp;
+	unsigned long page = ptable & PAGE_MASK;
+	unsigned char mask = 1 << ((ptable - page)/PTABLE_SIZE);
+
+	dp = PD_PTABLE(page);
+	if (!(PD_MARKBITS(dp) & mask)) {
+		PD_MARKBITS(dp) = 0xff;
+		list_add(dp, &ptable_list);
+	}
+
+	PD_MARKBITS(dp) &= ~mask;
+	pr_debug("init_pointer_table: %lx, %x\n", ptable, PD_MARKBITS(dp));
+
+	/* unreserve the page so it's possible to free that page */
+	__ClearPageReserved(PD_PAGE(dp));
+	init_page_count(PD_PAGE(dp));
+
+	return;
+}
+
+pmd_t *get_pointer_table (void)
+{
+	ptable_desc *dp = ptable_list.next;
+	unsigned char mask = PD_MARKBITS (dp);
+	unsigned char tmp;
+	unsigned int off;
+
+	/*
+	 * For a pointer table for a user process address space, a
+	 * table is taken from a page allocated for the purpose.  Each
+	 * page can hold 8 pointer tables.  The page is remapped in
+	 * virtual address space to be noncacheable.
+	 */
+	if (mask == 0) {
+		void *page;
+		ptable_desc *new;
+
+		if (!(page = (void *)get_zeroed_page(GFP_KERNEL)))
+			return NULL;
+
+		mmu_page_ctor(page);
+
+		new = PD_PTABLE(page);
+		PD_MARKBITS(new) = 0xfe;
+		list_add_tail(new, dp);
+
+		return (pmd_t *)page;
+	}
+
+	for (tmp = 1, off = 0; (mask & tmp) == 0; tmp <<= 1, off += PTABLE_SIZE)
+		;
+	PD_MARKBITS(dp) = mask & ~tmp;
+	if (!PD_MARKBITS(dp)) {
+		/* move to end of list */
+		list_move_tail(dp, &ptable_list);
+	}
+	return (pmd_t *) (page_address(PD_PAGE(dp)) + off);
+}
+
+int free_pointer_table (pmd_t *ptable)
+{
+	ptable_desc *dp;
+	unsigned long page = (unsigned long)ptable & PAGE_MASK;
+	unsigned char mask = 1 << (((unsigned long)ptable - page)/PTABLE_SIZE);
+
+	dp = PD_PTABLE(page);
+	if (PD_MARKBITS (dp) & mask)
+		panic ("table already free!");
+
+	PD_MARKBITS (dp) |= mask;
+
+	if (PD_MARKBITS(dp) == 0xff) {
+		/* all tables in page are free, free page */
+		list_del(dp);
+		mmu_page_dtor((void *)page);
+		free_page (page);
+		return 1;
+	} else if (ptable_list.next != dp) {
+		/*
+		 * move this descriptor to the front of the list, since
+		 * it has one or more free tables.
+		 */
+		list_move(dp, &ptable_list);
+	}
+	return 0;
+}
+
 /* size of memory already mapped in head.S */
 extern __initdata unsigned long m68k_init_mapped_size;
 

[PATCH -v2 05/10] m68k,mm: Restructure Motorola MMU page-table layout

[Peter Zijlstra]

The Motorola 68xxx MMUs, 040 (and later) have a fixed 7,7,{5,6}
page-table setup, where the last depends on the page-size selected (8k
vs 4k resp.), and head.S selects 4K pages. For 030 (and earlier) we
explicitly program 7,7,6 and 4K pages in %tc.

However, the current code implements this mightily weird. What it does
is group 16 of those (6 bit) pte tables into one 4k page to not waste
space. The down-side is that that forces pmd_t to be a 16-tuple
pointing to consecutive pte tables.

This breaks the generic code which assumes READ_ONCE(*pmd) will be
word sized.

Therefore implement a straight forward 7,7,6 3 level page-table setup,
with the addition (for 020/030) of (partial) large-page support. For
now this increases the memory footprint for pte-tables 15 fold.

Tested with ARAnyM/68040 emulation.

Suggested-by: Will Deacon <will@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
---
 arch/m68k/include/asm/motorola_pgtable.h |   15 +-----------
 arch/m68k/include/asm/page.h             |    6 ++---
 arch/m68k/include/asm/pgtable_mm.h       |   10 ++++----
 arch/m68k/mm/kmap.c                      |   36 +++++++++++++------------------
 arch/m68k/mm/motorola.c                  |   28 +++++++++++-------------
 5 files changed, 39 insertions(+), 56 deletions(-)

--- a/arch/m68k/include/asm/motorola_pgtable.h
+++ b/arch/m68k/include/asm/motorola_pgtable.h
@@ -108,13 +108,7 @@ static inline pte_t pte_modify(pte_t pte
 
 static inline void pmd_set(pmd_t *pmdp, pte_t *ptep)
 {
-	unsigned long ptbl = virt_to_phys(ptep) | _PAGE_TABLE | _PAGE_ACCESSED;
-	unsigned long *ptr = pmdp->pmd;
-	short i = 16;
-	while (--i >= 0) {
-		*ptr++ = ptbl;
-		ptbl += (sizeof(pte_t)*PTRS_PER_PTE/16);
-	}
+	pmd_val(*pmdp) = virt_to_phys(ptep) | _PAGE_TABLE | _PAGE_ACCESSED;
 }
 
 static inline void pud_set(pud_t *pudp, pmd_t *pmdp)
@@ -138,12 +132,7 @@ static inline void pud_set(pud_t *pudp,
 #define pmd_none(pmd)		(!pmd_val(pmd))
 #define pmd_bad(pmd)		((pmd_val(pmd) & _DESCTYPE_MASK) != _PAGE_TABLE)
 #define pmd_present(pmd)	(pmd_val(pmd) & _PAGE_TABLE)
-#define pmd_clear(pmdp) ({			\
-	unsigned long *__ptr = pmdp->pmd;	\
-	short __i = 16;				\
-	while (--__i >= 0)			\
-		*__ptr++ = 0;			\
-})
+#define pmd_clear(pmdp)		({ pmd_val(*pmdp) = 0; })
 #define pmd_page(pmd)		virt_to_page(__va(pmd_val(pmd)))
 
 
--- a/arch/m68k/include/asm/page.h
+++ b/arch/m68k/include/asm/page.h
@@ -22,9 +22,9 @@
  * These are used to make use of C type-checking..
  */
 #if !defined(CONFIG_MMU) || CONFIG_PGTABLE_LEVELS == 3
-typedef struct { unsigned long pmd[16]; } pmd_t;
-#define pmd_val(x)	((&x)->pmd[0])
-#define __pmd(x)	((pmd_t) { { (x) }, })
+typedef struct { unsigned long pmd; } pmd_t;
+#define pmd_val(x)	((&x)->pmd)
+#define __pmd(x)	((pmd_t) { (x) } )
 #endif
 
 typedef struct { unsigned long pte; } pte_t;
--- a/arch/m68k/include/asm/pgtable_mm.h
+++ b/arch/m68k/include/asm/pgtable_mm.h
@@ -36,7 +36,7 @@
 
 /* PMD_SHIFT determines the size of the area a second-level page table can map */
 #if CONFIG_PGTABLE_LEVELS == 3
-#define PMD_SHIFT	22
+#define PMD_SHIFT	18
 #endif
 #define PMD_SIZE	(1UL << PMD_SHIFT)
 #define PMD_MASK	(~(PMD_SIZE-1))
@@ -67,8 +67,8 @@
 #define PTRS_PER_PMD	1
 #define PTRS_PER_PGD	1024
 #else
-#define PTRS_PER_PTE	1024
-#define PTRS_PER_PMD	8
+#define PTRS_PER_PTE	64
+#define PTRS_PER_PMD	128
 #define PTRS_PER_PGD	128
 #endif
 #define USER_PTRS_PER_PGD	(TASK_SIZE/PGDIR_SIZE)
@@ -76,8 +76,8 @@
 
 /* Virtual address region for use by kernel_map() */
 #ifdef CONFIG_SUN3
-#define KMAP_START     0x0DC00000
-#define KMAP_END       0x0E000000
+#define KMAP_START	0x0dc00000
+#define KMAP_END	0x0e000000
 #elif defined(CONFIG_COLDFIRE)
 #define KMAP_START	0xe0000000
 #define KMAP_END	0xf0000000
--- a/arch/m68k/mm/kmap.c
+++ b/arch/m68k/mm/kmap.c
@@ -24,8 +24,6 @@
 
 #undef DEBUG
 
-#define PTRTREESIZE	(256*1024)
-
 /*
  * For 040/060 we can use the virtual memory area like other architectures,
  * but for 020/030 we want to use early termination page descriptors and we
@@ -50,7 +48,7 @@ static inline void free_io_area(void *ad
 
 #else
 
-#define IO_SIZE		(256*1024)
+#define IO_SIZE		PMD_SIZE
 
 static struct vm_struct *iolist;
 
@@ -81,14 +79,13 @@ static void __free_io_area(void *addr, u
 
 #if CONFIG_PGTABLE_LEVELS == 3
 		if (CPU_IS_020_OR_030) {
-			int pmd_off = (virtaddr/PTRTREESIZE) & 15;
-			int pmd_type = pmd_dir->pmd[pmd_off] & _DESCTYPE_MASK;
+			int pmd_type = pmd_val(*pmd_dir) & _DESCTYPE_MASK;
 
 			if (pmd_type == _PAGE_PRESENT) {
-				pmd_dir->pmd[pmd_off] = 0;
-				virtaddr += PTRTREESIZE;
-				size -= PTRTREESIZE;
-				continue;
+				pmd_clear(pmd_dir);
+				virtaddr += PMD_SIZE;
+				size -= PMD_SIZE;
+
 			} else if (pmd_type == 0)
 				continue;
 		}
@@ -249,7 +246,7 @@ void __iomem *__ioremap(unsigned long ph
 
 	while ((long)size > 0) {
 #ifdef DEBUG
-		if (!(virtaddr & (PTRTREESIZE-1)))
+		if (!(virtaddr & (PMD_SIZE-1)))
 			printk ("\npa=%#lx va=%#lx ", physaddr, virtaddr);
 #endif
 		pgd_dir = pgd_offset_k(virtaddr);
@@ -263,10 +260,10 @@ void __iomem *__ioremap(unsigned long ph
 
 #if CONFIG_PGTABLE_LEVELS == 3
 		if (CPU_IS_020_OR_030) {
-			pmd_dir->pmd[(virtaddr/PTRTREESIZE) & 15] = physaddr;
-			physaddr += PTRTREESIZE;
-			virtaddr += PTRTREESIZE;
-			size -= PTRTREESIZE;
+			pmd_val(*pmd_dir) = physaddr;
+			physaddr += PMD_SIZE;
+			virtaddr += PMD_SIZE;
+			size -= PMD_SIZE;
 		} else
 #endif
 		{
@@ -367,13 +364,12 @@ void kernel_set_cachemode(void *addr, un
 
 #if CONFIG_PGTABLE_LEVELS == 3
 		if (CPU_IS_020_OR_030) {
-			int pmd_off = (virtaddr/PTRTREESIZE) & 15;
+			unsigned long pmd = pmd_val(*pmd_dir);
 
-			if ((pmd_dir->pmd[pmd_off] & _DESCTYPE_MASK) == _PAGE_PRESENT) {
-				pmd_dir->pmd[pmd_off] = (pmd_dir->pmd[pmd_off] &
-							 _CACHEMASK040) | cmode;
-				virtaddr += PTRTREESIZE;
-				size -= PTRTREESIZE;
+			if ((pmd & _DESCTYPE_MASK) == _PAGE_PRESENT) {
+				*pmd_dir = __pmd((pmd & _CACHEMASK040) | cmode);
+				virtaddr += PMD_SIZE;
+				size -= PMD_SIZE;
 				continue;
 			}
 		}
--- a/arch/m68k/mm/motorola.c
+++ b/arch/m68k/mm/motorola.c
@@ -236,8 +236,6 @@ static pmd_t * __init kernel_ptr_table(v
 
 static void __init map_node(int node)
 {
-#define PTRTREESIZE (256*1024)
-#define ROOTTREESIZE (32*1024*1024)
 	unsigned long physaddr, virtaddr, size;
 	pgd_t *pgd_dir;
 	p4d_t *p4d_dir;
@@ -255,21 +253,21 @@ static void __init map_node(int node)
 
 	while (size > 0) {
 #ifdef DEBUG
-		if (!(virtaddr & (PTRTREESIZE-1)))
+		if (!(virtaddr & (PMD_SIZE-1)))
 			printk ("\npa=%#lx va=%#lx ", physaddr & PAGE_MASK,
 				virtaddr);
 #endif
 		pgd_dir = pgd_offset_k(virtaddr);
 		if (virtaddr && CPU_IS_020_OR_030) {
-			if (!(virtaddr & (ROOTTREESIZE-1)) &&
-			    size >= ROOTTREESIZE) {
+			if (!(virtaddr & (PGDIR_SIZE-1)) &&
+			    size >= PGDIR_SIZE) {
 #ifdef DEBUG
 				printk ("[very early term]");
 #endif
 				pgd_val(*pgd_dir) = physaddr;
-				size -= ROOTTREESIZE;
-				virtaddr += ROOTTREESIZE;
-				physaddr += ROOTTREESIZE;
+				size -= PGDIR_SIZE;
+				virtaddr += PGDIR_SIZE;
+				physaddr += PGDIR_SIZE;
 				continue;
 			}
 		}
@@ -289,8 +287,8 @@ static void __init map_node(int node)
 #ifdef DEBUG
 				printk ("[early term]");
 #endif
-				pmd_dir->pmd[(virtaddr/PTRTREESIZE) & 15] = physaddr;
-				physaddr += PTRTREESIZE;
+				pmd_val(*pmd_dir) = physaddr;
+				physaddr += PMD_SIZE;
 			} else {
 				int i;
 #ifdef DEBUG
@@ -298,15 +296,15 @@ static void __init map_node(int node)
 #endif
 				zero_pgtable = kernel_ptr_table();
 				pte_dir = (pte_t *)zero_pgtable;
-				pmd_dir->pmd[0] = virt_to_phys(pte_dir) |
-					_PAGE_TABLE | _PAGE_ACCESSED;
+				pmd_set(pmd_dir, pte_dir);
+
 				pte_val(*pte_dir++) = 0;
 				physaddr += PAGE_SIZE;
-				for (i = 1; i < 64; physaddr += PAGE_SIZE, i++)
+				for (i = 1; i < PTRS_PER_PTE; physaddr += PAGE_SIZE, i++)
 					pte_val(*pte_dir++) = physaddr;
 			}
-			size -= PTRTREESIZE;
-			virtaddr += PTRTREESIZE;
+			size -= PMD_SIZE;
+			virtaddr += PMD_SIZE;
 		} else {
 			if (!pmd_present(*pmd_dir)) {
 #ifdef DEBUG

[PATCH -v2 06/10] m68k,mm: Improve kernel_page_table()

[Peter Zijlstra]

With the PTE-tables now only being 256 bytes, allocating a full page
for them is a giant waste. Start by improving the boot time allocator
such that init_mm initialization will at least have optimal memory
density.

Much thanks to Will Deacon in help with debugging and ferreting out
lost information on these dusty MMUs.

Notes:

 - _TABLE_MASK is reduced to account for the shorter (256 byte)
   alignment of pte-tables, per the manual, table entries should only
   ever have state in the low 4 bits (Used,WrProt,Desc1,Desc0) so it is
   still longer than strictly required. (Thanks Will!!!)

 - Also use kernel_page_table() for the 020/030 zero_pgtable case and
   consequently remove the zero_pgtable init hack (will fix up later).

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
---
 arch/m68k/include/asm/motorola_pgtable.h |   13 +++++++
 arch/m68k/mm/init.c                      |    5 ---
 arch/m68k/mm/motorola.c                  |   51 +++++++++++++++++--------------
 3 files changed, 41 insertions(+), 28 deletions(-)

--- a/arch/m68k/include/asm/motorola_pgtable.h
+++ b/arch/m68k/include/asm/motorola_pgtable.h
@@ -23,7 +23,18 @@
 #define _DESCTYPE_MASK	0x003
 
 #define _CACHEMASK040	(~0x060)
-#define _TABLE_MASK	(0xfffffe00)
+
+/*
+ * Currently set to the minimum alignment of table pointers (256 bytes).
+ * The hardware only uses the low 4 bits for state:
+ *
+ *    3 - Used
+ *    2 - Write Protected
+ *  0,1 - Desciptor Type
+ *
+ * and has the rest of the bits reserved.
+ */
+#define _TABLE_MASK	(0xffffff00)
 
 #define _PAGE_TABLE	(_PAGE_SHORT)
 #define _PAGE_CHG_MASK  (PAGE_MASK | _PAGE_ACCESSED | _PAGE_DIRTY | _PAGE_NOCACHE)
--- a/arch/m68k/mm/init.c
+++ b/arch/m68k/mm/init.c
@@ -42,7 +42,6 @@ EXPORT_SYMBOL(empty_zero_page);
 
 #if !defined(CONFIG_SUN3) && !defined(CONFIG_COLDFIRE)
 extern void init_pointer_table(unsigned long ptable);
-extern pmd_t *zero_pgtable;
 #endif
 
 #ifdef CONFIG_MMU
@@ -135,10 +134,6 @@ static inline void init_pointer_tables(v
 		if (pud_present(*pud))
 			init_pointer_table(pgd_page_vaddr(kernel_pg_dir[i]));
 	}
-
-	/* insert also pointer table that we used to unmap the zero page */
-	if (zero_pgtable)
-		init_pointer_table((unsigned long)zero_pgtable);
 #endif
 }
 
--- a/arch/m68k/mm/motorola.c
+++ b/arch/m68k/mm/motorola.c
@@ -72,27 +72,35 @@ extern __initdata unsigned long m68k_ini
 
 extern unsigned long availmem;
 
+static pte_t *last_pte_table __initdata = NULL;
+
 static pte_t * __init kernel_page_table(void)
 {
-	pte_t *ptablep;
+	pte_t *pte_table = last_pte_table;
+
+	if (((unsigned long)last_pte_table & ~PAGE_MASK) == 0) {
+		pte_table = (pte_t *)memblock_alloc_low(PAGE_SIZE, PAGE_SIZE);
+		if (!pte_table) {
+			panic("%s: Failed to allocate %lu bytes align=%lx\n",
+					__func__, PAGE_SIZE, PAGE_SIZE);
+		}
+
+		clear_page(pte_table);
+		mmu_page_ctor(pte_table);
 
-	ptablep = (pte_t *)memblock_alloc_low(PAGE_SIZE, PAGE_SIZE);
-	if (!ptablep)
-		panic("%s: Failed to allocate %lu bytes align=%lx\n",
-		      __func__, PAGE_SIZE, PAGE_SIZE);
+		last_pte_table = pte_table;
+	}
 
-	clear_page(ptablep);
-	mmu_page_ctor(ptablep);
+	last_pte_table += PTRS_PER_PTE;
 
-	return ptablep;
+	return pte_table;
 }
 
-static pmd_t *last_pgtable __initdata = NULL;
-pmd_t *zero_pgtable __initdata = NULL;
+static pmd_t *last_pmd_table __initdata = NULL;
 
 static pmd_t * __init kernel_ptr_table(void)
 {
-	if (!last_pgtable) {
+	if (!last_pmd_table) {
 		unsigned long pmd, last;
 		int i;
 
@@ -111,25 +119,25 @@ static pmd_t * __init kernel_ptr_table(v
 				last = pmd;
 		}
 
-		last_pgtable = (pmd_t *)last;
+		last_pmd_table = (pmd_t *)last;
 #ifdef DEBUG
-		printk("kernel_ptr_init: %p\n", last_pgtable);
+		printk("kernel_ptr_init: %p\n", last_pmd_table);
 #endif
 	}
 
-	last_pgtable += PTRS_PER_PMD;
-	if (((unsigned long)last_pgtable & ~PAGE_MASK) == 0) {
-		last_pgtable = (pmd_t *)memblock_alloc_low(PAGE_SIZE,
+	last_pmd_table += PTRS_PER_PMD;
+	if (((unsigned long)last_pmd_table & ~PAGE_MASK) == 0) {
+		last_pmd_table = (pmd_t *)memblock_alloc_low(PAGE_SIZE,
 							   PAGE_SIZE);
-		if (!last_pgtable)
+		if (!last_pmd_table)
 			panic("%s: Failed to allocate %lu bytes align=%lx\n",
 			      __func__, PAGE_SIZE, PAGE_SIZE);
 
-		clear_page(last_pgtable);
-		mmu_page_ctor(last_pgtable);
+		clear_page(last_pmd_table);
+		mmu_page_ctor(last_pmd_table);
 	}
 
-	return last_pgtable;
+	return last_pmd_table;
 }
 
 static void __init map_node(int node)
@@ -192,8 +200,7 @@ static void __init map_node(int node)
 #ifdef DEBUG
 				printk ("[zero map]");
 #endif
-				zero_pgtable = kernel_ptr_table();
-				pte_dir = (pte_t *)zero_pgtable;
+				pte_dir = kernel_page_table();
 				pmd_set(pmd_dir, pte_dir);
 
 				pte_val(*pte_dir++) = 0;

[PATCH -v2 07/10] m68k,mm: Use table allocator for pgtables

[Peter Zijlstra]

With the new page-table layout, using full (4k) pages for (256 byte)
pte-tables is immensely wastefull. Move the pte-tables over to the
same allocator already used for the (512 byte) higher level tables
(pgd/pmd).

This reduces the pte-table waste from 15x to 2x.

Due to no longer being bound to 16 consecutive tables, this might
actually already be more efficient than the old code for sparse
tables.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
---
 arch/m68k/include/asm/motorola_pgalloc.h |   44 ++++++-------------------------
 arch/m68k/include/asm/motorola_pgtable.h |    8 ++++-
 arch/m68k/include/asm/page.h             |    5 +++
 3 files changed, 21 insertions(+), 36 deletions(-)

--- a/arch/m68k/include/asm/motorola_pgalloc.h
+++ b/arch/m68k/include/asm/motorola_pgalloc.h
@@ -13,54 +13,28 @@ extern int free_pointer_table(pmd_t *);
 
 static inline pte_t *pte_alloc_one_kernel(struct mm_struct *mm)
 {
-	pte_t *pte;
-
-	pte = (pte_t *)__get_free_page(GFP_KERNEL|__GFP_ZERO);
-	if (pte)
-		mmu_page_ctor(pte);
-
-	return pte;
+	return (pte_t *)get_pointer_table();
 }
 
 static inline void pte_free_kernel(struct mm_struct *mm, pte_t *pte)
 {
-	mmu_page_dtor(pte);
-	free_page((unsigned long) pte);
+	free_pointer_table((void *)pte);
 }
 
 static inline pgtable_t pte_alloc_one(struct mm_struct *mm)
 {
-	struct page *page;
-
-	page = alloc_pages(GFP_KERNEL|__GFP_ZERO, 0);
-	if(!page)
-		return NULL;
-	if (!pgtable_pte_page_ctor(page)) {
-		__free_page(page);
-		return NULL;
-	}
-
-	mmu_page_ctor(kmap(page));
-	kunmap(page);
-
-	return page;
+	return (pte_t *)get_pointer_table();
 }
 
-static inline void pte_free(struct mm_struct *mm, pgtable_t page)
+static inline void pte_free(struct mm_struct *mm, pgtable_t pgtable)
 {
-	pgtable_pte_page_dtor(page);
-	mmu_page_dtor(kmap(page));
-	kunmap(page);
-	__free_page(page);
+	free_pointer_table((void *)pgtable);
 }
 
-static inline void __pte_free_tlb(struct mmu_gather *tlb, pgtable_t page,
+static inline void __pte_free_tlb(struct mmu_gather *tlb, pgtable_t pgtable,
 				  unsigned long address)
 {
-	pgtable_pte_page_dtor(page);
-	mmu_page_dtor(kmap(page));
-	kunmap(page);
-	__free_page(page);
+	free_pointer_table((void *)pgtable);
 }
 
 
@@ -99,9 +73,9 @@ static inline void pmd_populate_kernel(s
 
 static inline void pmd_populate(struct mm_struct *mm, pmd_t *pmd, pgtable_t page)
 {
-	pmd_set(pmd, page_address(page));
+	pmd_set(pmd, page);
 }
-#define pmd_pgtable(pmd) pmd_page(pmd)
+#define pmd_pgtable(pmd) ((pgtable_t)__pmd_page(pmd))
 
 static inline void pud_populate(struct mm_struct *mm, pud_t *pud, pmd_t *pmd)
 {
--- a/arch/m68k/include/asm/motorola_pgtable.h
+++ b/arch/m68k/include/asm/motorola_pgtable.h
@@ -144,7 +144,13 @@ static inline void pud_set(pud_t *pudp,
 #define pmd_bad(pmd)		((pmd_val(pmd) & _DESCTYPE_MASK) != _PAGE_TABLE)
 #define pmd_present(pmd)	(pmd_val(pmd) & _PAGE_TABLE)
 #define pmd_clear(pmdp)		({ pmd_val(*pmdp) = 0; })
-#define pmd_page(pmd)		virt_to_page(__va(pmd_val(pmd)))
+
+/*
+ * m68k does not have huge pages (020/030 actually could), but generic code
+ * expects pmd_page() to exists, only to then DCE it all. Provide a dummy to
+ * make the compiler happy.
+ */
+#define pmd_page(pmd)		NULL
 
 
 #define pud_none(pud)		(!pud_val(pud))
--- a/arch/m68k/include/asm/page.h
+++ b/arch/m68k/include/asm/page.h
@@ -30,7 +30,12 @@ typedef struct { unsigned long pmd; } pm
 typedef struct { unsigned long pte; } pte_t;
 typedef struct { unsigned long pgd; } pgd_t;
 typedef struct { unsigned long pgprot; } pgprot_t;
+
+#if defined(CONFIG_SUN3) || defined(CONFIG_COLDFIRE)
 typedef struct page *pgtable_t;
+#else
+typedef pte_t *pgtable_t;
+#endif
 
 #define pte_val(x)	((x).pte)
 #define pgd_val(x)	((x).pgd)

[PATCH -v2 08/10] m68k,mm: Extend table allocator for multiple sizes

[Peter Zijlstra]

In addition to the PGD/PMD table size (128*4) add a PTE table size
(64*4) to the table allocator. This completely removes the pte-table
overhead compared to the old code, even for dense tables.

Notes:

 - the allocator gained a list_empty() check to deal with there not
   being any pages at all.

 - the free mask is extended to cover more than the 8 bits required
   for the (512 byte) PGD/PMD tables.

 - NR_PAGETABLE accounting is restored.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
---
 arch/m68k/include/asm/motorola_pgalloc.h |   31 +++++++++-----
 arch/m68k/mm/init.c                      |   16 +++----
 arch/m68k/mm/motorola.c                  |   65 ++++++++++++++++++++-----------
 3 files changed, 70 insertions(+), 42 deletions(-)

--- a/arch/m68k/include/asm/motorola_pgalloc.h
+++ b/arch/m68k/include/asm/motorola_pgalloc.h
@@ -8,61 +8,68 @@
 extern void mmu_page_ctor(void *page);
 extern void mmu_page_dtor(void *page);
 
-extern pmd_t *get_pointer_table(void);
-extern int free_pointer_table(pmd_t *);
+enum m68k_table_types {
+	TABLE_PGD = 0,
+	TABLE_PMD = 0, /* same size as PGD */
+	TABLE_PTE = 1,
+};
+
+extern void init_pointer_table(void *table, int type);
+extern void *get_pointer_table(int type);
+extern int free_pointer_table(void *table, int type);
 
 static inline pte_t *pte_alloc_one_kernel(struct mm_struct *mm)
 {
-	return (pte_t *)get_pointer_table();
+	return get_pointer_table(TABLE_PTE);
 }
 
 static inline void pte_free_kernel(struct mm_struct *mm, pte_t *pte)
 {
-	free_pointer_table((void *)pte);
+	free_pointer_table(pte, TABLE_PTE);
 }
 
 static inline pgtable_t pte_alloc_one(struct mm_struct *mm)
 {
-	return (pte_t *)get_pointer_table();
+	return get_pointer_table(TABLE_PTE);
 }
 
 static inline void pte_free(struct mm_struct *mm, pgtable_t pgtable)
 {
-	free_pointer_table((void *)pgtable);
+	free_pointer_table(pgtable, TABLE_PTE);
 }
 
 static inline void __pte_free_tlb(struct mmu_gather *tlb, pgtable_t pgtable,
 				  unsigned long address)
 {
-	free_pointer_table((void *)pgtable);
+	free_pointer_table(pgtable, TABLE_PTE);
 }
 
 
 static inline pmd_t *pmd_alloc_one(struct mm_struct *mm, unsigned long address)
 {
-	return get_pointer_table();
+	return get_pointer_table(TABLE_PMD);
 }
 
 static inline int pmd_free(struct mm_struct *mm, pmd_t *pmd)
 {
-	return free_pointer_table(pmd);
+	return free_pointer_table(pmd, TABLE_PMD);
 }
 
 static inline int __pmd_free_tlb(struct mmu_gather *tlb, pmd_t *pmd,
 				 unsigned long address)
 {
-	return free_pointer_table(pmd);
+	return free_pointer_table(pmd, TABLE_PMD);
 }
 
 
 static inline void pgd_free(struct mm_struct *mm, pgd_t *pgd)
 {
-	pmd_free(mm, (pmd_t *)pgd);
+	free_pointer_table(pgd, TABLE_PGD);
 }
 
 static inline pgd_t *pgd_alloc(struct mm_struct *mm)
 {
-	return (pgd_t *)get_pointer_table();
+	return get_pointer_table(TABLE_PGD);
 }
 
 
--- a/arch/m68k/mm/init.c
+++ b/arch/m68k/mm/init.c
@@ -40,10 +40,6 @@
 void *empty_zero_page;
 EXPORT_SYMBOL(empty_zero_page);
 
-#if !defined(CONFIG_SUN3) && !defined(CONFIG_COLDFIRE)
-extern void init_pointer_table(unsigned long ptable);
-#endif
-
 #ifdef CONFIG_MMU
 
 pg_data_t pg_data_map[MAX_NUMNODES];
@@ -127,12 +123,16 @@ static inline void init_pointer_tables(v
 	int i;
 
 	/* insert pointer tables allocated so far into the tablelist */
-	init_pointer_table((unsigned long)kernel_pg_dir);
+	init_pointer_table(kernel_pg_dir, TABLE_PGD);
 	for (i = 0; i < PTRS_PER_PGD; i++) {
-		pud_t *pud = (pud_t *)(&kernel_pg_dir[i]);
+		pud_t *pud = (pud_t *)&kernel_pg_dir[i];
+		pmd_t *pmd_dir;
+
+		if (!pud_present(*pud))
+			continue;
 
-		if (pud_present(*pud))
-			init_pointer_table(pgd_page_vaddr(kernel_pg_dir[i]));
+		pmd_dir = (pmd_t *)pgd_page_vaddr(kernel_pg_dir[i]);
+		init_pointer_table(pmd_dir, TABLE_PMD);
 	}
 #endif
 }
--- a/arch/m68k/mm/motorola.c
+++ b/arch/m68k/mm/motorola.c
@@ -72,24 +72,35 @@ void mmu_page_dtor(void *page)
    arch/sparc/mm/srmmu.c ... */
 
 typedef struct list_head ptable_desc;
-static LIST_HEAD(ptable_list);
+
+static struct list_head ptable_list[2] = {
+	LIST_HEAD_INIT(ptable_list[0]),
+	LIST_HEAD_INIT(ptable_list[1]),
+};
 
 #define PD_PTABLE(page) ((ptable_desc *)&(virt_to_page(page)->lru))
 #define PD_PAGE(ptable) (list_entry(ptable, struct page, lru))
-#define PD_MARKBITS(dp) (*(unsigned char *)&PD_PAGE(dp)->index)
+#define PD_MARKBITS(dp) (*(unsigned int *)&PD_PAGE(dp)->index)
+
+static const int ptable_shift[2] = {
+	7+2, /* PGD, PMD */
+	6+2, /* PTE */
+};
 
-#define PTABLE_SIZE (PTRS_PER_PMD * sizeof(pmd_t))
+#define ptable_size(type) (1U << ptable_shift[type])
+#define ptable_mask(type) ((1U << (PAGE_SIZE / ptable_size(type))) - 1)
 
-void __init init_pointer_table(unsigned long ptable)
+void __init init_pointer_table(void *table, int type)
 {
 	ptable_desc *dp;
+	unsigned long ptable = (unsigned long)table;
 	unsigned long page = ptable & PAGE_MASK;
-	unsigned char mask = 1 << ((ptable - page)/PTABLE_SIZE);
+	unsigned int mask = 1U << ((ptable - page)/ptable_size(type));
 
 	dp = PD_PTABLE(page);
 	if (!(PD_MARKBITS(dp) & mask)) {
-		PD_MARKBITS(dp) = 0xff;
-		list_add(dp, &ptable_list);
+		PD_MARKBITS(dp) = ptable_mask(type);
+		list_add(dp, &ptable_list[type]);
 	}
 
 	PD_MARKBITS(dp) &= ~mask;
@@ -102,12 +113,11 @@ void __init init_pointer_table(unsigned
 	return;
 }
 
-pmd_t *get_pointer_table (void)
+void *get_pointer_table(int type)
 {
-	ptable_desc *dp = ptable_list.next;
-	unsigned char mask = PD_MARKBITS (dp);
-	unsigned char tmp;
-	unsigned int off;
+	ptable_desc *dp = ptable_list[type].next;
+	unsigned int mask = list_empty(&ptable_list[type]) ? 0 : PD_MARKBITS(dp);
+	unsigned int tmp, off;
 
 	/*
 	 * For a pointer table for a user process address space, a
@@ -122,30 +132,39 @@ pmd_t *get_pointer_table (void)
 		if (!(page = (void *)get_zeroed_page(GFP_KERNEL)))
 			return NULL;
 
+		if (type == TABLE_PTE) {
+			/*
+			 * m68k doesn't have SPLIT_PTE_PTLOCKS for not having
+			 * SMP.
+			 */
+			pgtable_pte_page_ctor(virt_to_page(page));
+		}
+
 		mmu_page_ctor(page);
 
 		new = PD_PTABLE(page);
-		PD_MARKBITS(new) = 0xfe;
+		PD_MARKBITS(new) = ptable_mask(type) - 1;
 		list_add_tail(new, dp);
 
 		return (pmd_t *)page;
 	}
 
-	for (tmp = 1, off = 0; (mask & tmp) == 0; tmp <<= 1, off += PTABLE_SIZE)
+	for (tmp = 1, off = 0; (mask & tmp) == 0; tmp <<= 1, off += ptable_size(type))
 		;
 	PD_MARKBITS(dp) = mask & ~tmp;
 	if (!PD_MARKBITS(dp)) {
 		/* move to end of list */
-		list_move_tail(dp, &ptable_list);
+		list_move_tail(dp, &ptable_list[type]);
 	}
-	return (pmd_t *) (page_address(PD_PAGE(dp)) + off);
+	return page_address(PD_PAGE(dp)) + off;
 }
 
-int free_pointer_table (pmd_t *ptable)
+int free_pointer_table(void *table, int type)
 {
 	ptable_desc *dp;
-	unsigned long page = (unsigned long)ptable & PAGE_MASK;
-	unsigned char mask = 1 << (((unsigned long)ptable - page)/PTABLE_SIZE);
+	unsigned long ptable = (unsigned long)table;
+	unsigned long page = ptable & PAGE_MASK;
+	unsigned int mask = 1U << ((ptable - page)/ptable_size(type));
 
 	dp = PD_PTABLE(page);
 	if (PD_MARKBITS (dp) & mask)
@@ -153,18 +172,20 @@ int free_pointer_table (pmd_t *ptable)
 
 	PD_MARKBITS (dp) |= mask;
 
-	if (PD_MARKBITS(dp) == 0xff) {
+	if (PD_MARKBITS(dp) == ptable_mask(type)) {
 		/* all tables in page are free, free page */
 		list_del(dp);
 		mmu_page_dtor((void *)page);
+		if (type == TABLE_PTE)
+			pgtable_pte_page_dtor(virt_to_page(page));
 		free_page (page);
 		return 1;
-	} else if (ptable_list.next != dp) {
+	} else if (ptable_list[type].next != dp) {
 		/*
 		 * move this descriptor to the front of the list, since
 		 * it has one or more free tables.
 		 */
-		list_move(dp, &ptable_list);
+		list_move(dp, &ptable_list[type]);
 	}
 	return 0;
 }

Re: [PATCH -v2 08/10] m68k,mm: Extend table allocator for multiple sizes

[Geert Uytterhoeven]

Hoi Peter,

On Fri, Jan 31, 2020 at 1:56 PM Peter Zijlstra <peterz@infradead.org> wrote:
> In addition to the PGD/PMD table size (128*4) add a PTE table size
> (64*4) to the table allocator. This completely removes the pte-table
> overhead compared to the old code, even for dense tables.

Thanks for your patch!

> Notes:
>
>  - the allocator gained a list_empty() check to deal with there not
>    being any pages at all.
>
>  - the free mask is extended to cover more than the 8 bits required
>    for the (512 byte) PGD/PMD tables.

Being an mm-illiterate, I don't understand the relation between the number
of bits and the size (see below).

>  - NR_PAGETABLE accounting is restored.
>
> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>

WARNING: Missing Signed-off-by: line by nominal patch author 'Peter
Zijlstra <peterz@infradead.org>'
(in all patches)

I can fix that (the From?) up while applying.

> --- a/arch/m68k/mm/motorola.c
> +++ b/arch/m68k/mm/motorola.c
> @@ -72,24 +72,35 @@ void mmu_page_dtor(void *page)
>     arch/sparc/mm/srmmu.c ... */
>
>  typedef struct list_head ptable_desc;
> -static LIST_HEAD(ptable_list);
> +
> +static struct list_head ptable_list[2] = {
> +       LIST_HEAD_INIT(ptable_list[0]),
> +       LIST_HEAD_INIT(ptable_list[1]),
> +};
>
>  #define PD_PTABLE(page) ((ptable_desc *)&(virt_to_page(page)->lru))
>  #define PD_PAGE(ptable) (list_entry(ptable, struct page, lru))
> -#define PD_MARKBITS(dp) (*(unsigned char *)&PD_PAGE(dp)->index)
> +#define PD_MARKBITS(dp) (*(unsigned int *)&PD_PAGE(dp)->index)
> +
> +static const int ptable_shift[2] = {
> +       7+2, /* PGD, PMD */
> +       6+2, /* PTE */
> +};
>
> -#define PTABLE_SIZE (PTRS_PER_PMD * sizeof(pmd_t))
> +#define ptable_size(type) (1U << ptable_shift[type])
> +#define ptable_mask(type) ((1U << (PAGE_SIZE / ptable_size(type))) - 1)

So this is 0xff for PGD and PMD, like before, and 0xffff for PTE.
Why the latter value?

Thanks!

>
> -void __init init_pointer_table(unsigned long ptable)
> +void __init init_pointer_table(void *table, int type)
>  {
>         ptable_desc *dp;
> +       unsigned long ptable = (unsigned long)table;
>         unsigned long page = ptable & PAGE_MASK;
> -       unsigned char mask = 1 << ((ptable - page)/PTABLE_SIZE);
> +       unsigned int mask = 1U << ((ptable - page)/ptable_size(type));
>
>         dp = PD_PTABLE(page);
>         if (!(PD_MARKBITS(dp) & mask)) {
> -               PD_MARKBITS(dp) = 0xff;
> -               list_add(dp, &ptable_list);
> +               PD_MARKBITS(dp) = ptable_mask(type);
> +               list_add(dp, &ptable_list[type]);
>         }
>
>         PD_MARKBITS(dp) &= ~mask;

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: [PATCH -v2 08/10] m68k,mm: Extend table allocator for multiple sizes

[Peter Zijlstra]

On Fri, Feb 07, 2020 at 11:56:40AM +0100, Geert Uytterhoeven wrote:
> Hoi Peter,
> 
> On Fri, Jan 31, 2020 at 1:56 PM Peter Zijlstra <peterz@infradead.org> wrote:
> > In addition to the PGD/PMD table size (128*4) add a PTE table size
> > (64*4) to the table allocator. This completely removes the pte-table
> > overhead compared to the old code, even for dense tables.
> 
> Thanks for your patch!
> 
> > Notes:
> >
> >  - the allocator gained a list_empty() check to deal with there not
> >    being any pages at all.
> >
> >  - the free mask is extended to cover more than the 8 bits required
> >    for the (512 byte) PGD/PMD tables.
> 
> Being an mm-illiterate, I don't understand the relation between the number
> of bits and the size (see below).

If the table translates 7 bits of the address, it will have 1<<7 entries.

> >  - NR_PAGETABLE accounting is restored.
> >
> > Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
> 
> WARNING: Missing Signed-off-by: line by nominal patch author 'Peter
> Zijlstra <peterz@infradead.org>'
> (in all patches)
> 
> I can fix that (the From?) up while applying.

I'm not sure where that warning comes from, but if you feel it needs
fixing, sure. I normally only add the (Intel) thing to the SoB. I've so
far never had complaints about that.

> > --- a/arch/m68k/mm/motorola.c
> > +++ b/arch/m68k/mm/motorola.c
> > @@ -72,24 +72,35 @@ void mmu_page_dtor(void *page)
> >     arch/sparc/mm/srmmu.c ... */
> >
> >  typedef struct list_head ptable_desc;
> > -static LIST_HEAD(ptable_list);
> > +
> > +static struct list_head ptable_list[2] = {
> > +       LIST_HEAD_INIT(ptable_list[0]),
> > +       LIST_HEAD_INIT(ptable_list[1]),
> > +};
> >
> >  #define PD_PTABLE(page) ((ptable_desc *)&(virt_to_page(page)->lru))
> >  #define PD_PAGE(ptable) (list_entry(ptable, struct page, lru))
> > -#define PD_MARKBITS(dp) (*(unsigned char *)&PD_PAGE(dp)->index)
> > +#define PD_MARKBITS(dp) (*(unsigned int *)&PD_PAGE(dp)->index)
> > +
> > +static const int ptable_shift[2] = {
> > +       7+2, /* PGD, PMD */
> > +       6+2, /* PTE */
> > +};
> >
> > -#define PTABLE_SIZE (PTRS_PER_PMD * sizeof(pmd_t))
> > +#define ptable_size(type) (1U << ptable_shift[type])
> > +#define ptable_mask(type) ((1U << (PAGE_SIZE / ptable_size(type))) - 1)
> 
> So this is 0xff for PGD and PMD, like before, and 0xffff for PTE.
> Why the latter value?

The PGD/PMD being 7 bits are sizeof(unsigned long) << 7, or 512 bytes
big. In one 4k page, there fit 8 such entries. 0xFF is 8 bits set, one
for each of the 8 512 byte fragments.

For the PTE tables, which are 6 bit and of sizeof(unsigned long) << 6,
or 256 bytes, we can fit 16 in one 4k page, resulting in 0xFFFF.

Re: [PATCH -v2 08/10] m68k,mm: Extend table allocator for multiple sizes

[Geert Uytterhoeven]

Hoi Peter,

On Fri, Feb 7, 2020 at 12:34 PM Peter Zijlstra <peterz@infradead.org> wrote:
> On Fri, Feb 07, 2020 at 11:56:40AM +0100, Geert Uytterhoeven wrote:
> > On Fri, Jan 31, 2020 at 1:56 PM Peter Zijlstra <peterz@infradead.org> wrote:
> > > In addition to the PGD/PMD table size (128*4) add a PTE table size
> > > (64*4) to the table allocator. This completely removes the pte-table
> > > overhead compared to the old code, even for dense tables.
> >
> > Thanks for your patch!
> >
> > > Notes:
> > >
> > >  - the allocator gained a list_empty() check to deal with there not
> > >    being any pages at all.
> > >
> > >  - the free mask is extended to cover more than the 8 bits required
> > >    for the (512 byte) PGD/PMD tables.
> >
> > Being an mm-illiterate, I don't understand the relation between the number
> > of bits and the size (see below).
>
> If the table translates 7 bits of the address, it will have 1<<7 entries.
>
> > >  - NR_PAGETABLE accounting is restored.
> > >
> > > Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
> >
> > WARNING: Missing Signed-off-by: line by nominal patch author 'Peter
> > Zijlstra <peterz@infradead.org>'
> > (in all patches)
> >
> > I can fix that (the From?) up while applying.
>
> I'm not sure where that warning comes from, but if you feel it needs
> fixing, sure. I normally only add the (Intel) thing to the SoB. I've so
> far never had complaints about that.

Checkpatch doesn't like this.

> > > --- a/arch/m68k/mm/motorola.c
> > > +++ b/arch/m68k/mm/motorola.c
> > > @@ -72,24 +72,35 @@ void mmu_page_dtor(void *page)
> > >     arch/sparc/mm/srmmu.c ... */
> > >
> > >  typedef struct list_head ptable_desc;
> > > -static LIST_HEAD(ptable_list);
> > > +
> > > +static struct list_head ptable_list[2] = {
> > > +       LIST_HEAD_INIT(ptable_list[0]),
> > > +       LIST_HEAD_INIT(ptable_list[1]),
> > > +};
> > >
> > >  #define PD_PTABLE(page) ((ptable_desc *)&(virt_to_page(page)->lru))
> > >  #define PD_PAGE(ptable) (list_entry(ptable, struct page, lru))
> > > -#define PD_MARKBITS(dp) (*(unsigned char *)&PD_PAGE(dp)->index)
> > > +#define PD_MARKBITS(dp) (*(unsigned int *)&PD_PAGE(dp)->index)
> > > +
> > > +static const int ptable_shift[2] = {
> > > +       7+2, /* PGD, PMD */
> > > +       6+2, /* PTE */
> > > +};
> > >
> > > -#define PTABLE_SIZE (PTRS_PER_PMD * sizeof(pmd_t))
> > > +#define ptable_size(type) (1U << ptable_shift[type])
> > > +#define ptable_mask(type) ((1U << (PAGE_SIZE / ptable_size(type))) - 1)
> >
> > So this is 0xff for PGD and PMD, like before, and 0xffff for PTE.
> > Why the latter value?
>
> The PGD/PMD being 7 bits are sizeof(unsigned long) << 7, or 512 bytes
> big. In one 4k page, there fit 8 such entries. 0xFF is 8 bits set, one
> for each of the 8 512 byte fragments.
>
> For the PTE tables, which are 6 bit and of sizeof(unsigned long) << 6,
> or 256 bytes, we can fit 16 in one 4k page, resulting in 0xFFFF.

Thanks!

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Checkpatch being daft, Was: [PATCH -v2 08/10] m68k,mm: Extend table allocator for multiple sizes

[Peter Zijlstra]

On Fri, Feb 07, 2020 at 01:11:54PM +0100, Geert Uytterhoeven wrote:
> On Fri, Feb 7, 2020 at 12:34 PM Peter Zijlstra <peterz@infradead.org> wrote:
> > On Fri, Feb 07, 2020 at 11:56:40AM +0100, Geert Uytterhoeven wrote:

> > > WARNING: Missing Signed-off-by: line by nominal patch author 'Peter
> > > Zijlstra <peterz@infradead.org>'
> > > (in all patches)
> > >
> > > I can fix that (the From?) up while applying.
> >
> > I'm not sure where that warning comes from, but if you feel it needs
> > fixing, sure. I normally only add the (Intel) thing to the SoB. I've so
> > far never had complaints about that.
> 
> Checkpatch doesn't like this.

Ooh, I see, that's a relatively new warning, pretty daft if you ask me.

Now I have to rediscover how I went about teaching checkpatch to STFU ;-)

Joe, should that '$email eq $author' not ignore rfc822 comments? That
is:

	Peter Zijlstra <peterz@infradead.org>

and:

	Peter Zijlstra (Intel) <peterz@infradead.org>

are, in actual fact, the same.

Re: Checkpatch being daft, Was: [PATCH -v2 08/10] m68k,mm: Extend table allocator for multiple sizes

[Peter Zijlstra]

On Fri, Feb 07, 2020 at 01:30:35PM +0100, Peter Zijlstra wrote:
> On Fri, Feb 07, 2020 at 01:11:54PM +0100, Geert Uytterhoeven wrote:
> > On Fri, Feb 7, 2020 at 12:34 PM Peter Zijlstra <peterz@infradead.org> wrote:
> > > On Fri, Feb 07, 2020 at 11:56:40AM +0100, Geert Uytterhoeven wrote:
> 
> > > > WARNING: Missing Signed-off-by: line by nominal patch author 'Peter
> > > > Zijlstra <peterz@infradead.org>'
> > > > (in all patches)
> > > >
> > > > I can fix that (the From?) up while applying.
> > >
> > > I'm not sure where that warning comes from, but if you feel it needs
> > > fixing, sure. I normally only add the (Intel) thing to the SoB. I've so
> > > far never had complaints about that.
> > 
> > Checkpatch doesn't like this.
> 
> Ooh, I see, that's a relatively new warning, pretty daft if you ask me.
> 
> Now I have to rediscover how I went about teaching checkpatch to STFU ;-)
> 
> Joe, should that '$email eq $author' not ignore rfc822 comments? That

Argh, that's me hitting on the wrong 'nominal' in checkpatch.pl, same
difference though.

> is:
> 
> 	Peter Zijlstra <peterz@infradead.org>
> 
> and:
> 
> 	Peter Zijlstra (Intel) <peterz@infradead.org>
> 
> are, in actual fact, the same.

Re: Checkpatch being daft, Was: [PATCH -v2 08/10] m68k,mm: Extend table allocator for multiple sizes

[Joe Perches]

On Fri, 2020-02-07 at 13:33 +0100, Peter Zijlstra wrote:
> On Fri, Feb 07, 2020 at 01:30:35PM +0100, Peter Zijlstra wrote:
> > On Fri, Feb 07, 2020 at 01:11:54PM +0100, Geert Uytterhoeven wrote:
> > > On Fri, Feb 7, 2020 at 12:34 PM Peter Zijlstra <peterz@infradead.org> wrote:
> > > > On Fri, Feb 07, 2020 at 11:56:40AM +0100, Geert Uytterhoeven wrote:
> > > > > WARNING: Missing Signed-off-by: line by nominal patch author 'Peter
> > > > > Zijlstra <peterz@infradead.org>'
> > > > > (in all patches)
> > > > > 
> > > > > I can fix that (the From?) up while applying.
> > > > 
> > > > I'm not sure where that warning comes from, but if you feel it needs
> > > > fixing, sure. I normally only add the (Intel) thing to the SoB. I've so
> > > > far never had complaints about that.
> > > 
> > > Checkpatch doesn't like this.
> > 
> > Ooh, I see, that's a relatively new warning, pretty daft if you ask me.
> > 
> > Now I have to rediscover how I went about teaching checkpatch to STFU ;-)
> > 
> > Joe, should that '$email eq $author' not ignore rfc822 comments? That
> 
> Argh, that's me hitting on the wrong 'nominal' in checkpatch.pl, same
> difference though.
> 
> > is:
> > 
> > 	Peter Zijlstra <peterz@infradead.org>
> > 
> > and:
> > 
> > 	Peter Zijlstra (Intel) <peterz@infradead.org>
> > 
> > are, in actual fact, the same.

Maybe this?
---
 scripts/checkpatch.pl | 39 +++++++++++++++++++++++++++++----------
 1 file changed, 29 insertions(+), 10 deletions(-)

diff --git a/scripts/checkpatch.pl b/scripts/checkpatch.pl
index f3b8434..17637d0 100755
--- a/scripts/checkpatch.pl
+++ b/scripts/checkpatch.pl
@@ -1132,6 +1132,7 @@ sub parse_email {
 	my ($formatted_email) = @_;
 
 	my $name = "";
+	my $name_comment = "";
 	my $address = "";
 	my $comment = "";
 
@@ -1164,6 +1165,10 @@ sub parse_email {
 
 	$name = trim($name);
 	$name =~ s/^\"|\"$//g;
+	$name =~ s/(\s*\([^\)]+\))\s*//;
+	if (defined($1)) {
+		$name_comment = trim($1);
+	}
 	$address = trim($address);
 	$address =~ s/^\<|\>$//g;
 
@@ -1172,7 +1177,7 @@ sub parse_email {
 		$name = "\"$name\"";
 	}
 
-	return ($name, $address, $comment);
+	return ($name, $name_comment, $address, $comment);
 }
 
 sub format_email {
@@ -1198,6 +1203,23 @@ sub format_email {
 	return $formatted_email;
 }
 
+sub reformat_email {
+	my ($email) = @_;
+
+	my ($email_name, $name_comment, $email_address, $comment) = parse_email($email);
+	return format_email($email_name, $email_address);
+}
+
+sub same_email_addresses {
+	my ($email1, $email2) = @_;
+
+	my ($email1_name, $name1_comment, $email1_address, $comment1) = parse_email($email1);
+	my ($email2_name, $name2_comment, $email2_address, $comment2) = parse_email($email2);
+
+	return $email1_name eq $email2_name &&
+	       $email1_address eq $email2_address;
+}
+
 sub which {
 	my ($bin) = @_;
 
@@ -2618,17 +2640,16 @@ sub process {
 			$author = $1;
 			$author = encode("utf8", $author) if ($line =~ /=\?utf-8\?/i);
 			$author =~ s/"//g;
+			$author = reformat_email($author);
 		}
 
 # Check the patch for a signoff:
-		if ($line =~ /^\s*signed-off-by:/i) {
+		if ($line =~ /^\s*signed-off-by:\s*(.*)/i) {
 			$signoff++;
 			$in_commit_log = 0;
 			if ($author ne '') {
-				my $l = $line;
-				$l =~ s/"//g;
-				if ($l =~ /^\s*signed-off-by:\s*\Q$author\E/i) {
-				    $authorsignoff = 1;
+				if (same_email_addresses($1, $author)) {
+					$authorsignoff = 1;
 				}
 			}
 		}
@@ -2678,7 +2699,7 @@ sub process {
 				}
 			}
 
-			my ($email_name, $email_address, $comment) = parse_email($email);
+			my ($email_name, $name_comment, $email_address, $comment) = parse_email($email);
 			my $suggested_email = format_email(($email_name, $email_address));
 			if ($suggested_email eq "") {
 				ERROR("BAD_SIGN_OFF",
@@ -2689,9 +2710,7 @@ sub process {
 				$dequoted =~ s/" </ </;
 				# Don't force email to have quotes
 				# Allow just an angle bracketed address
-				if ("$dequoted$comment" ne $email &&
-				    "<$email_address>$comment" ne $email &&
-				    "$suggested_email$comment" ne $email) {
+				if (!same_email_addresses($email, $suggested_email)) {
 					WARN("BAD_SIGN_OFF",
 					     "email address '$email' might be better as '$suggested_email$comment'\n" . $herecurr);
 				}

Re: Checkpatch being daft, Was: [PATCH -v2 08/10] m68k,mm: Extend table allocator for multiple sizes

[Peter Zijlstra]

On Sun, Feb 09, 2020 at 10:24:15AM -0800, Joe Perches wrote:
> Maybe this?

This isn't anywhere near RFC compliant, but I do think it greatly
improves the current situation, so:

Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>

one little nit below..

> ---
>  scripts/checkpatch.pl | 39 +++++++++++++++++++++++++++++----------
>  1 file changed, 29 insertions(+), 10 deletions(-)
> 
> diff --git a/scripts/checkpatch.pl b/scripts/checkpatch.pl
> index f3b8434..17637d0 100755
> --- a/scripts/checkpatch.pl
> +++ b/scripts/checkpatch.pl
> @@ -1132,6 +1132,7 @@ sub parse_email {
>  	my ($formatted_email) = @_;
>  
>  	my $name = "";
> +	my $name_comment = "";
>  	my $address = "";
>  	my $comment = "";
>  
> @@ -1164,6 +1165,10 @@ sub parse_email {
>  
>  	$name = trim($name);
>  	$name =~ s/^\"|\"$//g;
> +	$name =~ s/(\s*\([^\)]+\))\s*//;
> +	if (defined($1)) {
> +		$name_comment = trim($1);
> +	}
>  	$address = trim($address);
>  	$address =~ s/^\<|\>$//g;
>  
> @@ -1172,7 +1177,7 @@ sub parse_email {
>  		$name = "\"$name\"";
>  	}
>  
> -	return ($name, $address, $comment);
> +	return ($name, $name_comment, $address, $comment);
>  }
>  
>  sub format_email {
> @@ -1198,6 +1203,23 @@ sub format_email {
>  	return $formatted_email;
>  }
>  
> +sub reformat_email {
> +	my ($email) = @_;
> +
> +	my ($email_name, $name_comment, $email_address, $comment) = parse_email($email);
> +	return format_email($email_name, $email_address);
> +}
> +
> +sub same_email_addresses {
> +	my ($email1, $email2) = @_;
> +
> +	my ($email1_name, $name1_comment, $email1_address, $comment1) = parse_email($email1);
> +	my ($email2_name, $name2_comment, $email2_address, $comment2) = parse_email($email2);
> +
> +	return $email1_name eq $email2_name &&
> +	       $email1_address eq $email2_address;

strictly speaking only _address needs be the same for the whole thing to
arrive at the same inbox, but I suppose that for sanity's sake, this
comparison makes sense.

> +}

Re: Checkpatch being daft, Was: [PATCH -v2 08/10] m68k,mm: Extend table allocator for multiple sizes

[Joe Perches]

On Mon, 2020-02-10 at 17:38 +0100, Peter Zijlstra wrote:
> On Sun, Feb 09, 2020 at 10:24:15AM -0800, Joe Perches wrote:
> > Maybe this?
> 
> This isn't anywhere near RFC compliant,

Nothing really is.

https://metacpan.org/pod/Email::Address

doesn't really do a perfect job either, but I suppose
it'd be possible to use it one day instead.

> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
[]
>  one little nit below..
[]
> > +sub same_email_addresses {
> > +	my ($email1, $email2) = @_;
> > +
> > +	my ($email1_name, $name1_comment, $email1_address, $comment1) = parse_email($email1);
> > +	my ($email2_name, $name2_comment, $email2_address, $comment2) = parse_email($email2);
> > +
> > +	return $email1_name eq $email2_name &&
> > +	       $email1_address eq $email2_address;
> 
> strictly speaking only _address needs be the same for the whole thing to
> arrive at the same inbox, but I suppose that for sanity's sake, this
> comparison makes sense.

I know, and I believe that's true too.

Re: Checkpatch being daft, Was: [PATCH -v2 08/10] m68k,mm: Extend table allocator for multiple sizes

[Joe Perches]

On Fri, 2020-02-07 at 13:30 +0100, Peter Zijlstra wrote:
> On Fri, Feb 07, 2020 at 01:11:54PM +0100, Geert Uytterhoeven wrote:
> > On Fri, Feb 7, 2020 at 12:34 PM Peter Zijlstra <peterz@infradead.org> wrote:
> > > On Fri, Feb 07, 2020 at 11:56:40AM +0100, Geert Uytterhoeven wrote:
> > > > WARNING: Missing Signed-off-by: line by nominal patch author 'Peter
> > > > Zijlstra <peterz@infradead.org>'
> > > > (in all patches)
> > > > 
> > > > I can fix that (the From?) up while applying.
> > > 
> > > I'm not sure where that warning comes from, but if you feel it needs
> > > fixing, sure. I normally only add the (Intel) thing to the SoB. I've so
> > > far never had complaints about that.
> > 
> > Checkpatch doesn't like this.
> 
> Ooh, I see, that's a relatively new warning, pretty daft if you ask me.
> 
> Now I have to rediscover how I went about teaching checkpatch to STFU ;-)
> 
> Joe, should that '$email eq $author' not ignore rfc822 comments? That
> is:
> 
> 	Peter Zijlstra <peterz@infradead.org>
> 
> and:
> 
> 	Peter Zijlstra (Intel) <peterz@infradead.org>
> 
> are, in actual fact, the same.

checkpatch doesn't really have an rfc822 parser and would
likely require some fairly involved changes to use one.

For instance: adding
---
 scripts/checkpatch.pl | 1 +
 1 file changed, 1 insertion(+)

diff --git a/scripts/checkpatch.pl b/scripts/checkpatch.pl
index f3b8434..959b8ef 100755
--- a/scripts/checkpatch.pl
+++ b/scripts/checkpatch.pl
@@ -1164,6 +1164,7 @@ sub parse_email {
 
 	$name = trim($name);
 	$name =~ s/^\"|\"$//g;
+	$name =~ s/\s*\([^\)]+\)\s*//;
 	$address = trim($address);
 	$address =~ s/^\<|\>$//g;
 
---

Into the parse_email function to strip rfc822 comments from
email addresses would just create a couple different warnings.

For now, how about adding --ignore=NO_AUTHOR_SIGN_OFF

[PATCH -v2 09/10] m68k,mm: Fully initialize the page-table allocator

[Peter Zijlstra]

Also iterate the PMD tables to populate the PTE table allocator. This
also fully replaces the previous zero_pgtable hack.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
---
 arch/m68k/mm/init.c |   13 ++++++++++++-
 1 file changed, 12 insertions(+), 1 deletion(-)

--- a/arch/m68k/mm/init.c
+++ b/arch/m68k/mm/init.c
@@ -120,7 +120,7 @@ void free_initmem(void)
 static inline void init_pointer_tables(void)
 {
 #if defined(CONFIG_MMU) && !defined(CONFIG_SUN3) && !defined(CONFIG_COLDFIRE)
-	int i;
+	int i, j;
 
 	/* insert pointer tables allocated so far into the tablelist */
 	init_pointer_table(kernel_pg_dir, TABLE_PGD);
@@ -133,6 +133,17 @@ static inline void init_pointer_tables(v
 
 		pmd_dir = (pmd_t *)pgd_page_vaddr(kernel_pg_dir[i]);
 		init_pointer_table(pmd_dir, TABLE_PMD);
+
+		for (j = 0; j < PTRS_PER_PMD; j++) {
+			pmd_t *pmd = &pmd_dir[j];
+			pte_t *pte_dir;
+
+			if (!pmd_present(*pmd))
+				continue;
+
+			pte_dir = (pte_t *)__pmd_page(*pmd);
+			init_pointer_table(pte_dir, TABLE_PTE);
+		}
 	}
 #endif
 }

Re: [PATCH -v2 09/10] m68k,mm: Fully initialize the page-table allocator

[Geert Uytterhoeven]

Hoi Peter,

Thanks for your patch!

On Fri, Jan 31, 2020 at 1:56 PM Peter Zijlstra <peterz@infradead.org> wrote:
> Also iterate the PMD tables to populate the PTE table allocator. This
> also fully replaces the previous zero_pgtable hack.

As no code is being removed in this patch, does this mean this case was
broken since "[PATCH 06/10] m68k,mm: Improve kernel_page_table()"?

> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
> ---
>  arch/m68k/mm/init.c |   13 ++++++++++++-
>  1 file changed, 12 insertions(+), 1 deletion(-)
>
> --- a/arch/m68k/mm/init.c
> +++ b/arch/m68k/mm/init.c
> @@ -120,7 +120,7 @@ void free_initmem(void)
>  static inline void init_pointer_tables(void)
>  {
>  #if defined(CONFIG_MMU) && !defined(CONFIG_SUN3) && !defined(CONFIG_COLDFIRE)
> -       int i;
> +       int i, j;
>
>         /* insert pointer tables allocated so far into the tablelist */
>         init_pointer_table(kernel_pg_dir, TABLE_PGD);
> @@ -133,6 +133,17 @@ static inline void init_pointer_tables(v
>
>                 pmd_dir = (pmd_t *)pgd_page_vaddr(kernel_pg_dir[i]);
>                 init_pointer_table(pmd_dir, TABLE_PMD);
> +
> +               for (j = 0; j < PTRS_PER_PMD; j++) {
> +                       pmd_t *pmd = &pmd_dir[j];
> +                       pte_t *pte_dir;
> +
> +                       if (!pmd_present(*pmd))
> +                               continue;
> +
> +                       pte_dir = (pte_t *)__pmd_page(*pmd);
> +                       init_pointer_table(pte_dir, TABLE_PTE);
> +               }
>         }
>  #endif
>  }

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: [PATCH -v2 09/10] m68k,mm: Fully initialize the page-table allocator

[Peter Zijlstra]

On Fri, Feb 07, 2020 at 11:58:42AM +0100, Geert Uytterhoeven wrote:
> Hoi Peter,
> 
> Thanks for your patch!
> 
> On Fri, Jan 31, 2020 at 1:56 PM Peter Zijlstra <peterz@infradead.org> wrote:
> > Also iterate the PMD tables to populate the PTE table allocator. This
> > also fully replaces the previous zero_pgtable hack.
> 
> As no code is being removed in this patch, does this mean this case was
> broken since "[PATCH 06/10] m68k,mm: Improve kernel_page_table()"?

'broken' is not the right word, less optimal might qualify.

What this code does it add all the pages from the early kernel allocator
to the regular allocator, such that any left over fragments might be
used.

[PATCH -v2 10/10] m68k,mm: Change ColdFire pgtable_t

[Peter Zijlstra]

From: Will Deacon <will@kernel.org>

To match what we did to the Motorola MMU routines, change the ColdFire
pgalloc.

The result is that ColdFire and Sun3 pgalloc are actually very similar
and could conceivably be unified.

Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
---
 arch/m68k/include/asm/mcf_pgalloc.h |   25 +++++++++++++------------
 arch/m68k/include/asm/page.h        |    7 ++++++-
 2 files changed, 19 insertions(+), 13 deletions(-)

--- a/arch/m68k/include/asm/mcf_pgalloc.h
+++ b/arch/m68k/include/asm/mcf_pgalloc.h
@@ -28,21 +28,22 @@ extern inline pmd_t *pmd_alloc_kernel(pg
 	return (pmd_t *) pgd;
 }
 
-#define pmd_populate(mm, pmd, page) (pmd_val(*pmd) = \
-	(unsigned long)(page_address(page)))
+#define pmd_populate(mm, pmd, pte) (pmd_val(*pmd) = (unsigned long)(pte))
 
-#define pmd_populate_kernel(mm, pmd, pte) (pmd_val(*pmd) = (unsigned long)(pte))
+#define pmd_populate_kernel pmd_populate
 
-#define pmd_pgtable(pmd) pmd_page(pmd)
+#define pmd_pgtable(pmd) pfn_to_virt(pmd_val(pmd) >> PAGE_SHIFT)
 
-static inline void __pte_free_tlb(struct mmu_gather *tlb, pgtable_t page,
+static inline void __pte_free_tlb(struct mmu_gather *tlb, pgtable_t pgtable,
 				  unsigned long address)
 {
+	struct page *page = virt_to_page(pgtable);
+
 	pgtable_pte_page_dtor(page);
 	__free_page(page);
 }
 
-static inline struct page *pte_alloc_one(struct mm_struct *mm)
+static inline pgtable_t pte_alloc_one(struct mm_struct *mm)
 {
 	struct page *page = alloc_pages(GFP_DMA, 0);
 	pte_t *pte;
@@ -54,16 +55,16 @@ static inline struct page *pte_alloc_one
 		return NULL;
 	}
 
-	pte = kmap(page);
-	if (pte)
-		clear_page(pte);
-	kunmap(page);
+	pte = page_address(page);
+	clear_page(pte);
 
-	return page;
+	return pte;
 }
 
-static inline void pte_free(struct mm_struct *mm, struct page *page)
+static inline void pte_free(struct mm_struct *mm, pgtable_t pgtable)
 {
+	struct page *page = virt_to_page(pgtable);
+
 	pgtable_pte_page_dtor(page);
 	__free_page(page);
 }
--- a/arch/m68k/include/asm/page.h
+++ b/arch/m68k/include/asm/page.h
@@ -31,7 +31,12 @@ typedef struct { unsigned long pte; } pt
 typedef struct { unsigned long pgd; } pgd_t;
 typedef struct { unsigned long pgprot; } pgprot_t;
 
-#if defined(CONFIG_SUN3) || defined(CONFIG_COLDFIRE)
+#if defined(CONFIG_SUN3)
+/*
+ * Sun3 still uses the asm-generic/pgalloc.h code and thus needs this
+ * definition. It would be possible to unify Sun3 and ColdFire pgalloc and have
+ * all of m68k use the same type.
+ */
 typedef struct page *pgtable_t;
 #else
 typedef pte_t *pgtable_t;

Re: [PATCH -v2 00/10] Rewrite Motorola MMU page-table layout

[Greg Ungerer]

Hi Peter,

On 31/1/20 10:45 pm, Peter Zijlstra wrote:
> In order to faciliate Will's READ_ONCE() patches:
> 
>    https://lkml.kernel.org/r/20200123153341.19947-1-will@kernel.org
> 
> we need to fix m68k/motorola to not have a giant pmd_t. These patches do so and
> are tested using ARAnyM/68040.
> 
> Michael tested the previous version on his Atari Falcon/68030.
> 
> Build tested for sun3/coldfire.

Thanks for the quick turn around. Build looks good for me too with
this new series. I will test on real hardware on Monday.

Regards
Greg


> Please consider!
> 
> Changes since -v1:
>   - fixed sun3/coldfire build issues
>   - unified motorola mmu page setup
>   - added enum to table allocator
>   - moved pointer table allocator to motorola.c
>   - converted coldfire pgtable_t
>   - fixed coldfire pgd_alloc
>   - fixed coldfire nocache
> 
> ---
>   arch/m68k/include/asm/mcf_pgalloc.h      |  31 ++---
>   arch/m68k/include/asm/motorola_pgalloc.h |  74 ++++------
>   arch/m68k/include/asm/motorola_pgtable.h |  36 +++--
>   arch/m68k/include/asm/page.h             |  16 ++-
>   arch/m68k/include/asm/pgtable_mm.h       |  10 +-
>   arch/m68k/mm/init.c                      |  34 +++--
>   arch/m68k/mm/kmap.c                      |  36 +++--
>   arch/m68k/mm/memory.c                    | 103 --------------
>   arch/m68k/mm/motorola.c                  | 228 +++++++++++++++++++++++++------
>   9 files changed, 302 insertions(+), 266 deletions(-)
> 
> 

Re: [PATCH -v2 00/10] Rewrite Motorola MMU page-table layout

[Greg Ungerer]

Hi Peter,

On 31/1/20 11:19 pm, Greg Ungerer wrote:
> On 31/1/20 10:45 pm, Peter Zijlstra wrote:
>> In order to faciliate Will's READ_ONCE() patches:
>>
>>    https://lkml.kernel.org/r/20200123153341.19947-1-will@kernel.org
>>
>> we need to fix m68k/motorola to not have a giant pmd_t. These patches do so and
>> are tested using ARAnyM/68040.
>>
>> Michael tested the previous version on his Atari Falcon/68030.
>>
>> Build tested for sun3/coldfire.
> 
> Thanks for the quick turn around. Build looks good for me too with
> this new series. I will test on real hardware on Monday.

So this tests good for me on real hardware. I had a look over the
ColdFire changes and I have no problems with anything.
So for the ColdFire parts:

Tested-by: Greg Ungerer <gerg@linux-m68k.org>
Acked-by: Greg Ungerer <gerg@linux-m68k.org>

Regards
Greg


>> Please consider!
>>
>> Changes since -v1:
>>   - fixed sun3/coldfire build issues
>>   - unified motorola mmu page setup
>>   - added enum to table allocator
>>   - moved pointer table allocator to motorola.c
>>   - converted coldfire pgtable_t
>>   - fixed coldfire pgd_alloc
>>   - fixed coldfire nocache
>>
>> ---
>>   arch/m68k/include/asm/mcf_pgalloc.h      |  31 ++---
>>   arch/m68k/include/asm/motorola_pgalloc.h |  74 ++++------
>>   arch/m68k/include/asm/motorola_pgtable.h |  36 +++--
>>   arch/m68k/include/asm/page.h             |  16 ++-
>>   arch/m68k/include/asm/pgtable_mm.h       |  10 +-
>>   arch/m68k/mm/init.c                      |  34 +++--
>>   arch/m68k/mm/kmap.c                      |  36 +++--
>>   arch/m68k/mm/memory.c                    | 103 --------------
>>   arch/m68k/mm/motorola.c                  | 228 +++++++++++++++++++++++++------
>>   9 files changed, 302 insertions(+), 266 deletions(-)
>>
>>

Re: [PATCH -v2 00/10] Rewrite Motorola MMU page-table layout

[Michael Schmitz]

Peter,

this version tested OK on 030, so

Tested-by; Michael Schmitz <schmitzmic@gmail.com>

Am 01.02.2020 um 01:45 schrieb Peter Zijlstra:
> Hi!
>
> In order to faciliate Will's READ_ONCE() patches:
>
>   https://lkml.kernel.org/r/20200123153341.19947-1-will@kernel.org
>
> we need to fix m68k/motorola to not have a giant pmd_t. These patches do so and
> are tested using ARAnyM/68040.
>
> Michael tested the previous version on his Atari Falcon/68030.
>
> Build tested for sun3/coldfire.
>
> Please consider!
>
> Changes since -v1:
>  - fixed sun3/coldfire build issues
>  - unified motorola mmu page setup
>  - added enum to table allocator
>  - moved pointer table allocator to motorola.c
>  - converted coldfire pgtable_t
>  - fixed coldfire pgd_alloc
>  - fixed coldfire nocache
>
> ---
>  arch/m68k/include/asm/mcf_pgalloc.h      |  31 ++---
>  arch/m68k/include/asm/motorola_pgalloc.h |  74 ++++------
>  arch/m68k/include/asm/motorola_pgtable.h |  36 +++--
>  arch/m68k/include/asm/page.h             |  16 ++-
>  arch/m68k/include/asm/pgtable_mm.h       |  10 +-
>  arch/m68k/mm/init.c                      |  34 +++--
>  arch/m68k/mm/kmap.c                      |  36 +++--
>  arch/m68k/mm/memory.c                    | 103 --------------
>  arch/m68k/mm/motorola.c                  | 228 +++++++++++++++++++++++++------
>  9 files changed, 302 insertions(+), 266 deletions(-)
>
>

Re: [PATCH -v2 00/10] Rewrite Motorola MMU page-table layout

[Will Deacon]

On Fri, Jan 31, 2020 at 01:45:31PM +0100, Peter Zijlstra wrote:
> In order to faciliate Will's READ_ONCE() patches:
> 
>   https://lkml.kernel.org/r/20200123153341.19947-1-will@kernel.org
> 
> we need to fix m68k/motorola to not have a giant pmd_t. These patches do so and
> are tested using ARAnyM/68040.
> 
> Michael tested the previous version on his Atari Falcon/68030.
> 
> Build tested for sun3/coldfire.
> 
> Please consider!

Apart from the two written by me:

Acked-by: Will Deacon <will@kernel.org>

Cheers,

Will

Re: [PATCH -v2 00/10] Rewrite Motorola MMU page-table layout

[Geert Uytterhoeven]

Hi Peter, Will,

On Fri, Jan 31, 2020 at 1:56 PM Peter Zijlstra <peterz@infradead.org> wrote:
> In order to faciliate Will's READ_ONCE() patches:
>
>   https://lkml.kernel.org/r/20200123153341.19947-1-will@kernel.org
>
> we need to fix m68k/motorola to not have a giant pmd_t. These patches do so and
> are tested using ARAnyM/68040.
>
> Michael tested the previous version on his Atari Falcon/68030.
>
> Build tested for sun3/coldfire.
>
> Please consider!

Thanks, applied and queued for v5.7, using an immutable branch named
pgtable-layout-rewrite.

https://git.kernel.org/pub/scm/linux/kernel/git/geert/linux-m68k.git/log/?h=pgtable-layout-rewrite

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: [PATCH -v2 00/10] Rewrite Motorola MMU page-table layout

[Geert Uytterhoeven]

Hi Peter, Will

On Mon, Feb 10, 2020 at 12:16 PM Geert Uytterhoeven
<geert@linux-m68k.org> wrote:
> On Fri, Jan 31, 2020 at 1:56 PM Peter Zijlstra <peterz@infradead.org> wrote:
> > In order to faciliate Will's READ_ONCE() patches:
> >
> >   https://lkml.kernel.org/r/20200123153341.19947-1-will@kernel.org
> >
> > we need to fix m68k/motorola to not have a giant pmd_t. These patches do so and
> > are tested using ARAnyM/68040.
> >
> > Michael tested the previous version on his Atari Falcon/68030.
> >
> > Build tested for sun3/coldfire.
> >
> > Please consider!
>
> Thanks, applied and queued for v5.7, using an immutable branch named
> pgtable-layout-rewrite.
>
> https://git.kernel.org/pub/scm/linux/kernel/git/geert/linux-m68k.git/log/?h=pgtable-layout-rewrite

Any plans to use this? Looks like it's still part of linux-next through the m68k
tree only.

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: [PATCH -v2 00/10] Rewrite Motorola MMU page-table layout

[Peter Zijlstra]

On Mon, Mar 09, 2020 at 11:15:31AM +0100, Geert Uytterhoeven wrote:
> Hi Peter, Will
> 
> On Mon, Feb 10, 2020 at 12:16 PM Geert Uytterhoeven
> <geert@linux-m68k.org> wrote:
> > On Fri, Jan 31, 2020 at 1:56 PM Peter Zijlstra <peterz@infradead.org> wrote:
> > > In order to faciliate Will's READ_ONCE() patches:
> > >
> > >   https://lkml.kernel.org/r/20200123153341.19947-1-will@kernel.org
> > >
> > > we need to fix m68k/motorola to not have a giant pmd_t. These patches do so and
> > > are tested using ARAnyM/68040.
> > >
> > > Michael tested the previous version on his Atari Falcon/68030.
> > >
> > > Build tested for sun3/coldfire.
> > >
> > > Please consider!
> >
> > Thanks, applied and queued for v5.7, using an immutable branch named
> > pgtable-layout-rewrite.
> >
> > https://git.kernel.org/pub/scm/linux/kernel/git/geert/linux-m68k.git/log/?h=pgtable-layout-rewrite
> 
> Any plans to use this? Looks like it's still part of linux-next through the m68k
> tree only.

Argh, yes. So we ran into sparc32 also need fixups, and both me and Will
seem to have stalled on fixing that. I'll try and get back to that.