[patch 53/54] powerpc/book3s64/mm: update flush_tlb_range to flush page walk cache

[Andrew Morton <akpm@linux-foundation.org>]

From: "Aneesh Kumar K.V" <aneesh.kumar@linux.ibm.com>
Subject: powerpc/book3s64/mm: update flush_tlb_range to flush page walk cache

flush_tlb_range is special in that we don't specify the page size used for
the translation.  Hence when flushing TLB we flush the translation cache
for all possible page sizes.  The kernel also uses the same interface when
moving page tables around.  Such a move requires us to flush the page walk
cache.

Instead of adding another interface to force page walk cache flush, update
flush_tlb_range to flush page walk cache if the range flushed is more than
the PMD range.  A page table move will always involve an invalidate range
more than PMD_SIZE.

Running microbenchmark with mprotect and parallel memory access didn't
show any observable performance impact.

Link: https://lkml.kernel.org/r/20210616045735.374532-3-aneesh.kumar@linux.ibm.com
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Cc: Christophe Leroy <christophe.leroy@csgroup.eu>
Cc: Hugh Dickins <hughd@google.com>
Cc: Joel Fernandes <joel@joelfernandes.org>
Cc: Kalesh Singh <kaleshsingh@google.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Nicholas Piggin <npiggin@gmail.com>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
---

 arch/powerpc/include/asm/book3s/64/tlbflush-radix.h |    2 
 arch/powerpc/mm/book3s64/radix_hugetlbpage.c        |    8 +
 arch/powerpc/mm/book3s64/radix_tlb.c                |   44 ++++++----
 3 files changed, 36 insertions(+), 18 deletions(-)

--- a/arch/powerpc/include/asm/book3s/64/tlbflush-radix.h~powerpc-book3s64-mm-update-flush_tlb_range-to-flush-page-walk-cache
+++ a/arch/powerpc/include/asm/book3s/64/tlbflush-radix.h
@@ -64,6 +64,8 @@ extern void radix__flush_hugetlb_tlb_ran
 					   unsigned long start, unsigned long end);
 extern void radix__flush_tlb_range_psize(struct mm_struct *mm, unsigned long start,
 					 unsigned long end, int psize);
+void radix__flush_tlb_pwc_range_psize(struct mm_struct *mm, unsigned long start,
+				      unsigned long end, int psize);
 extern void radix__flush_pmd_tlb_range(struct vm_area_struct *vma,
 				       unsigned long start, unsigned long end);
 extern void radix__flush_tlb_range(struct vm_area_struct *vma, unsigned long start,
--- a/arch/powerpc/mm/book3s64/radix_hugetlbpage.c~powerpc-book3s64-mm-update-flush_tlb_range-to-flush-page-walk-cache
+++ a/arch/powerpc/mm/book3s64/radix_hugetlbpage.c
@@ -32,7 +32,13 @@ void radix__flush_hugetlb_tlb_range(stru
 	struct hstate *hstate = hstate_file(vma->vm_file);
 
 	psize = hstate_get_psize(hstate);
-	radix__flush_tlb_range_psize(vma->vm_mm, start, end, psize);
+	/*
+	 * Flush PWC even if we get PUD_SIZE hugetlb invalidate to keep this simpler.
+	 */
+	if (end - start >= PUD_SIZE)
+		radix__flush_tlb_pwc_range_psize(vma->vm_mm, start, end, psize);
+	else
+		radix__flush_tlb_range_psize(vma->vm_mm, start, end, psize);
 }
 
 /*
--- a/arch/powerpc/mm/book3s64/radix_tlb.c~powerpc-book3s64-mm-update-flush_tlb_range-to-flush-page-walk-cache
+++ a/arch/powerpc/mm/book3s64/radix_tlb.c
@@ -1111,14 +1111,13 @@ static unsigned long tlb_local_single_pa
 
 static inline void __radix__flush_tlb_range(struct mm_struct *mm,
 					    unsigned long start, unsigned long end)
-
 {
 	unsigned long pid;
 	unsigned int page_shift = mmu_psize_defs[mmu_virtual_psize].shift;
 	unsigned long page_size = 1UL << page_shift;
 	unsigned long nr_pages = (end - start) >> page_shift;
 	bool fullmm = (end == TLB_FLUSH_ALL);
-	bool flush_pid;
+	bool flush_pid, flush_pwc = false;
 	enum tlb_flush_type type;
 
 	pid = mm->context.id;
@@ -1137,8 +1136,16 @@ static inline void __radix__flush_tlb_ra
 		flush_pid = nr_pages > tlb_single_page_flush_ceiling;
 	else
 		flush_pid = nr_pages > tlb_local_single_page_flush_ceiling;
+	/*
+	 * full pid flush already does the PWC flush. if it is not full pid
+	 * flush check the range is more than PMD and force a pwc flush
+	 * mremap() depends on this behaviour.
+	 */
+	if (!flush_pid && (end - start) >= PMD_SIZE)
+		flush_pwc = true;
 
 	if (!mmu_has_feature(MMU_FTR_GTSE) && type == FLUSH_TYPE_GLOBAL) {
+		unsigned long type = H_RPTI_TYPE_TLB;
 		unsigned long tgt = H_RPTI_TARGET_CMMU;
 		unsigned long pg_sizes = psize_to_rpti_pgsize(mmu_virtual_psize);
 
@@ -1146,19 +1153,20 @@ static inline void __radix__flush_tlb_ra
 			pg_sizes |= psize_to_rpti_pgsize(MMU_PAGE_2M);
 		if (atomic_read(&mm->context.copros) > 0)
 			tgt |= H_RPTI_TARGET_NMMU;
-		pseries_rpt_invalidate(pid, tgt, H_RPTI_TYPE_TLB, pg_sizes,
-				       start, end);
+		if (flush_pwc)
+			type |= H_RPTI_TYPE_PWC;
+		pseries_rpt_invalidate(pid, tgt, type, pg_sizes, start, end);
 	} else if (flush_pid) {
+		/*
+		 * We are now flushing a range larger than PMD size force a RIC_FLUSH_ALL
+		 */
 		if (type == FLUSH_TYPE_LOCAL) {
-			_tlbiel_pid(pid, RIC_FLUSH_TLB);
+			_tlbiel_pid(pid, RIC_FLUSH_ALL);
 		} else {
 			if (cputlb_use_tlbie()) {
-				if (mm_needs_flush_escalation(mm))
-					_tlbie_pid(pid, RIC_FLUSH_ALL);
-				else
-					_tlbie_pid(pid, RIC_FLUSH_TLB);
+				_tlbie_pid(pid, RIC_FLUSH_ALL);
 			} else {
-				_tlbiel_pid_multicast(mm, pid, RIC_FLUSH_TLB);
+				_tlbiel_pid_multicast(mm, pid, RIC_FLUSH_ALL);
 			}
 		}
 	} else {
@@ -1174,6 +1182,9 @@ static inline void __radix__flush_tlb_ra
 
 		if (type == FLUSH_TYPE_LOCAL) {
 			asm volatile("ptesync": : :"memory");
+			if (flush_pwc)
+				/* For PWC, only one flush is needed */
+				__tlbiel_pid(pid, 0, RIC_FLUSH_PWC);
 			__tlbiel_va_range(start, end, pid, page_size, mmu_virtual_psize);
 			if (hflush)
 				__tlbiel_va_range(hstart, hend, pid,
@@ -1181,6 +1192,8 @@ static inline void __radix__flush_tlb_ra
 			ppc_after_tlbiel_barrier();
 		} else if (cputlb_use_tlbie()) {
 			asm volatile("ptesync": : :"memory");
+			if (flush_pwc)
+				__tlbie_pid(pid, RIC_FLUSH_PWC);
 			__tlbie_va_range(start, end, pid, page_size, mmu_virtual_psize);
 			if (hflush)
 				__tlbie_va_range(hstart, hend, pid,
@@ -1188,10 +1201,10 @@ static inline void __radix__flush_tlb_ra
 			asm volatile("eieio; tlbsync; ptesync": : :"memory");
 		} else {
 			_tlbiel_va_range_multicast(mm,
-					start, end, pid, page_size, mmu_virtual_psize, false);
+					start, end, pid, page_size, mmu_virtual_psize, flush_pwc);
 			if (hflush)
 				_tlbiel_va_range_multicast(mm,
-					hstart, hend, pid, PMD_SIZE, MMU_PAGE_2M, false);
+					hstart, hend, pid, PMD_SIZE, MMU_PAGE_2M, flush_pwc);
 		}
 	}
 out:
@@ -1265,9 +1278,6 @@ void radix__flush_all_lpid_guest(unsigne
 	_tlbie_lpid_guest(lpid, RIC_FLUSH_ALL);
 }
 
-static void radix__flush_tlb_pwc_range_psize(struct mm_struct *mm, unsigned long start,
-				  unsigned long end, int psize);
-
 void radix__tlb_flush(struct mmu_gather *tlb)
 {
 	int psize = 0;
@@ -1374,8 +1384,8 @@ void radix__flush_tlb_range_psize(struct
 	return __radix__flush_tlb_range_psize(mm, start, end, psize, false);
 }
 
-static void radix__flush_tlb_pwc_range_psize(struct mm_struct *mm, unsigned long start,
-				  unsigned long end, int psize)
+void radix__flush_tlb_pwc_range_psize(struct mm_struct *mm, unsigned long start,
+				      unsigned long end, int psize)
 {
 	__radix__flush_tlb_range_psize(mm, start, end, psize, true);
 }
_