* [PATCH v6 01/11] selftest/mremap_test: Update the test to handle pagesize other than 4K
2021-05-24 9:01 [PATCH v6 00/11] Speedup mremap on ppc64 Aneesh Kumar K.V
@ 2021-05-24 9:01 ` Aneesh Kumar K.V
2021-05-24 9:01 ` [PATCH v6 02/11] selftest/mremap_test: Avoid crash with static build Aneesh Kumar K.V
` (9 subsequent siblings)
10 siblings, 0 replies; 20+ messages in thread
From: Aneesh Kumar K.V @ 2021-05-24 9:01 UTC (permalink / raw)
To: linux-mm, akpm
Cc: Aneesh Kumar K.V, Linus Torvalds, npiggin, kaleshsingh, joel,
linuxppc-dev
Instead of hardcoding 4K page size fetch it using sysconf(). For the performance
measurements test still assume 2M and 1G are hugepage sizes.
Reviewed-by: Kalesh Singh <kaleshsingh@google.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
---
tools/testing/selftests/vm/mremap_test.c | 113 ++++++++++++-----------
1 file changed, 61 insertions(+), 52 deletions(-)
diff --git a/tools/testing/selftests/vm/mremap_test.c b/tools/testing/selftests/vm/mremap_test.c
index 9c391d016922..c9a5461eb786 100644
--- a/tools/testing/selftests/vm/mremap_test.c
+++ b/tools/testing/selftests/vm/mremap_test.c
@@ -45,14 +45,15 @@ enum {
_4MB = 4ULL << 20,
_1GB = 1ULL << 30,
_2GB = 2ULL << 30,
- PTE = _4KB,
PMD = _2MB,
PUD = _1GB,
};
+#define PTE page_size
+
#define MAKE_TEST(source_align, destination_align, size, \
overlaps, should_fail, test_name) \
-{ \
+(struct test){ \
.name = test_name, \
.config = { \
.src_alignment = source_align, \
@@ -252,12 +253,17 @@ static int parse_args(int argc, char **argv, unsigned int *threshold_mb,
return 0;
}
+#define MAX_TEST 13
+#define MAX_PERF_TEST 3
int main(int argc, char **argv)
{
int failures = 0;
int i, run_perf_tests;
unsigned int threshold_mb = VALIDATION_DEFAULT_THRESHOLD;
unsigned int pattern_seed;
+ struct test test_cases[MAX_TEST];
+ struct test perf_test_cases[MAX_PERF_TEST];
+ int page_size;
time_t t;
pattern_seed = (unsigned int) time(&t);
@@ -268,56 +274,59 @@ int main(int argc, char **argv)
ksft_print_msg("Test configs:\n\tthreshold_mb=%u\n\tpattern_seed=%u\n\n",
threshold_mb, pattern_seed);
- struct test test_cases[] = {
- /* Expected mremap failures */
- MAKE_TEST(_4KB, _4KB, _4KB, OVERLAPPING, EXPECT_FAILURE,
- "mremap - Source and Destination Regions Overlapping"),
- MAKE_TEST(_4KB, _1KB, _4KB, NON_OVERLAPPING, EXPECT_FAILURE,
- "mremap - Destination Address Misaligned (1KB-aligned)"),
- MAKE_TEST(_1KB, _4KB, _4KB, NON_OVERLAPPING, EXPECT_FAILURE,
- "mremap - Source Address Misaligned (1KB-aligned)"),
-
- /* Src addr PTE aligned */
- MAKE_TEST(PTE, PTE, _8KB, NON_OVERLAPPING, EXPECT_SUCCESS,
- "8KB mremap - Source PTE-aligned, Destination PTE-aligned"),
-
- /* Src addr 1MB aligned */
- MAKE_TEST(_1MB, PTE, _2MB, NON_OVERLAPPING, EXPECT_SUCCESS,
- "2MB mremap - Source 1MB-aligned, Destination PTE-aligned"),
- MAKE_TEST(_1MB, _1MB, _2MB, NON_OVERLAPPING, EXPECT_SUCCESS,
- "2MB mremap - Source 1MB-aligned, Destination 1MB-aligned"),
-
- /* Src addr PMD aligned */
- MAKE_TEST(PMD, PTE, _4MB, NON_OVERLAPPING, EXPECT_SUCCESS,
- "4MB mremap - Source PMD-aligned, Destination PTE-aligned"),
- MAKE_TEST(PMD, _1MB, _4MB, NON_OVERLAPPING, EXPECT_SUCCESS,
- "4MB mremap - Source PMD-aligned, Destination 1MB-aligned"),
- MAKE_TEST(PMD, PMD, _4MB, NON_OVERLAPPING, EXPECT_SUCCESS,
- "4MB mremap - Source PMD-aligned, Destination PMD-aligned"),
-
- /* Src addr PUD aligned */
- MAKE_TEST(PUD, PTE, _2GB, NON_OVERLAPPING, EXPECT_SUCCESS,
- "2GB mremap - Source PUD-aligned, Destination PTE-aligned"),
- MAKE_TEST(PUD, _1MB, _2GB, NON_OVERLAPPING, EXPECT_SUCCESS,
- "2GB mremap - Source PUD-aligned, Destination 1MB-aligned"),
- MAKE_TEST(PUD, PMD, _2GB, NON_OVERLAPPING, EXPECT_SUCCESS,
- "2GB mremap - Source PUD-aligned, Destination PMD-aligned"),
- MAKE_TEST(PUD, PUD, _2GB, NON_OVERLAPPING, EXPECT_SUCCESS,
- "2GB mremap - Source PUD-aligned, Destination PUD-aligned"),
- };
-
- struct test perf_test_cases[] = {
- /*
- * mremap 1GB region - Page table level aligned time
- * comparison.
- */
- MAKE_TEST(PTE, PTE, _1GB, NON_OVERLAPPING, EXPECT_SUCCESS,
- "1GB mremap - Source PTE-aligned, Destination PTE-aligned"),
- MAKE_TEST(PMD, PMD, _1GB, NON_OVERLAPPING, EXPECT_SUCCESS,
- "1GB mremap - Source PMD-aligned, Destination PMD-aligned"),
- MAKE_TEST(PUD, PUD, _1GB, NON_OVERLAPPING, EXPECT_SUCCESS,
- "1GB mremap - Source PUD-aligned, Destination PUD-aligned"),
- };
+ page_size = sysconf(_SC_PAGESIZE);
+
+ /* Expected mremap failures */
+ test_cases[0] = MAKE_TEST(page_size, page_size, page_size,
+ OVERLAPPING, EXPECT_FAILURE,
+ "mremap - Source and Destination Regions Overlapping");
+
+ test_cases[1] = MAKE_TEST(page_size, page_size/4, page_size,
+ NON_OVERLAPPING, EXPECT_FAILURE,
+ "mremap - Destination Address Misaligned (1KB-aligned)");
+ test_cases[2] = MAKE_TEST(page_size/4, page_size, page_size,
+ NON_OVERLAPPING, EXPECT_FAILURE,
+ "mremap - Source Address Misaligned (1KB-aligned)");
+
+ /* Src addr PTE aligned */
+ test_cases[3] = MAKE_TEST(PTE, PTE, PTE * 2,
+ NON_OVERLAPPING, EXPECT_SUCCESS,
+ "8KB mremap - Source PTE-aligned, Destination PTE-aligned");
+
+ /* Src addr 1MB aligned */
+ test_cases[4] = MAKE_TEST(_1MB, PTE, _2MB, NON_OVERLAPPING, EXPECT_SUCCESS,
+ "2MB mremap - Source 1MB-aligned, Destination PTE-aligned");
+ test_cases[5] = MAKE_TEST(_1MB, _1MB, _2MB, NON_OVERLAPPING, EXPECT_SUCCESS,
+ "2MB mremap - Source 1MB-aligned, Destination 1MB-aligned");
+
+ /* Src addr PMD aligned */
+ test_cases[6] = MAKE_TEST(PMD, PTE, _4MB, NON_OVERLAPPING, EXPECT_SUCCESS,
+ "4MB mremap - Source PMD-aligned, Destination PTE-aligned");
+ test_cases[7] = MAKE_TEST(PMD, _1MB, _4MB, NON_OVERLAPPING, EXPECT_SUCCESS,
+ "4MB mremap - Source PMD-aligned, Destination 1MB-aligned");
+ test_cases[8] = MAKE_TEST(PMD, PMD, _4MB, NON_OVERLAPPING, EXPECT_SUCCESS,
+ "4MB mremap - Source PMD-aligned, Destination PMD-aligned");
+
+ /* Src addr PUD aligned */
+ test_cases[9] = MAKE_TEST(PUD, PTE, _2GB, NON_OVERLAPPING, EXPECT_SUCCESS,
+ "2GB mremap - Source PUD-aligned, Destination PTE-aligned");
+ test_cases[10] = MAKE_TEST(PUD, _1MB, _2GB, NON_OVERLAPPING, EXPECT_SUCCESS,
+ "2GB mremap - Source PUD-aligned, Destination 1MB-aligned");
+ test_cases[11] = MAKE_TEST(PUD, PMD, _2GB, NON_OVERLAPPING, EXPECT_SUCCESS,
+ "2GB mremap - Source PUD-aligned, Destination PMD-aligned");
+ test_cases[12] = MAKE_TEST(PUD, PUD, _2GB, NON_OVERLAPPING, EXPECT_SUCCESS,
+ "2GB mremap - Source PUD-aligned, Destination PUD-aligned");
+
+ perf_test_cases[0] = MAKE_TEST(page_size, page_size, _1GB, NON_OVERLAPPING, EXPECT_SUCCESS,
+ "1GB mremap - Source PTE-aligned, Destination PTE-aligned");
+ /*
+ * mremap 1GB region - Page table level aligned time
+ * comparison.
+ */
+ perf_test_cases[1] = MAKE_TEST(PMD, PMD, _1GB, NON_OVERLAPPING, EXPECT_SUCCESS,
+ "1GB mremap - Source PMD-aligned, Destination PMD-aligned");
+ perf_test_cases[2] = MAKE_TEST(PUD, PUD, _1GB, NON_OVERLAPPING, EXPECT_SUCCESS,
+ "1GB mremap - Source PUD-aligned, Destination PUD-aligned");
run_perf_tests = (threshold_mb == VALIDATION_NO_THRESHOLD) ||
(threshold_mb * _1MB >= _1GB);
--
2.31.1
^ permalink raw reply related [flat|nested] 20+ messages in thread* [PATCH v6 02/11] selftest/mremap_test: Avoid crash with static build
2021-05-24 9:01 [PATCH v6 00/11] Speedup mremap on ppc64 Aneesh Kumar K.V
2021-05-24 9:01 ` [PATCH v6 01/11] selftest/mremap_test: Update the test to handle pagesize other than 4K Aneesh Kumar K.V
@ 2021-05-24 9:01 ` Aneesh Kumar K.V
2021-05-24 9:01 ` [PATCH v6 03/11] mm/mremap: Convert huge PUD move to separate helper Aneesh Kumar K.V
` (8 subsequent siblings)
10 siblings, 0 replies; 20+ messages in thread
From: Aneesh Kumar K.V @ 2021-05-24 9:01 UTC (permalink / raw)
To: linux-mm, akpm
Cc: Aneesh Kumar K.V, Linus Torvalds, npiggin, kaleshsingh, joel,
linuxppc-dev
With a large mmap map size, we can overlap with the text area and using
MAP_FIXED results in unmapping that area. Switch to MAP_FIXED_NOREPLACE
and handle the EEXIST error.
Reviewed-by: Kalesh Singh <kaleshsingh@google.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
---
tools/testing/selftests/vm/mremap_test.c | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)
diff --git a/tools/testing/selftests/vm/mremap_test.c b/tools/testing/selftests/vm/mremap_test.c
index c9a5461eb786..0624d1bd71b5 100644
--- a/tools/testing/selftests/vm/mremap_test.c
+++ b/tools/testing/selftests/vm/mremap_test.c
@@ -75,9 +75,10 @@ static void *get_source_mapping(struct config c)
retry:
addr += c.src_alignment;
src_addr = mmap((void *) addr, c.region_size, PROT_READ | PROT_WRITE,
- MAP_FIXED | MAP_ANONYMOUS | MAP_SHARED, -1, 0);
+ MAP_FIXED_NOREPLACE | MAP_ANONYMOUS | MAP_SHARED,
+ -1, 0);
if (src_addr == MAP_FAILED) {
- if (errno == EPERM)
+ if (errno == EPERM || errno == EEXIST)
goto retry;
goto error;
}
--
2.31.1
^ permalink raw reply related [flat|nested] 20+ messages in thread* [PATCH v6 03/11] mm/mremap: Convert huge PUD move to separate helper
2021-05-24 9:01 [PATCH v6 00/11] Speedup mremap on ppc64 Aneesh Kumar K.V
2021-05-24 9:01 ` [PATCH v6 01/11] selftest/mremap_test: Update the test to handle pagesize other than 4K Aneesh Kumar K.V
2021-05-24 9:01 ` [PATCH v6 02/11] selftest/mremap_test: Avoid crash with static build Aneesh Kumar K.V
@ 2021-05-24 9:01 ` Aneesh Kumar K.V
2021-05-24 9:01 ` [PATCH v6 04/11] mm/mremap: Use pmd/pud_poplulate to update page table entries Aneesh Kumar K.V
` (7 subsequent siblings)
10 siblings, 0 replies; 20+ messages in thread
From: Aneesh Kumar K.V @ 2021-05-24 9:01 UTC (permalink / raw)
To: linux-mm, akpm
Cc: Aneesh Kumar K.V, Linus Torvalds, npiggin, kaleshsingh, joel,
linuxppc-dev
With TRANSPARENT_HUGEPAGE_PUD enabled the kernel can find huge PUD entries.
Add a helper to move huge PUD entries on mremap().
This will be used by a later patch to optimize mremap of PUD_SIZE aligned
level 4 PTE mapped address
This also make sure we support mremap on huge PUD entries even with
CONFIG_HAVE_MOVE_PUD disabled.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
---
mm/mremap.c | 80 ++++++++++++++++++++++++++++++++++++++++++++++++-----
1 file changed, 73 insertions(+), 7 deletions(-)
diff --git a/mm/mremap.c b/mm/mremap.c
index ec8f840399ed..1d6fadbd4820 100644
--- a/mm/mremap.c
+++ b/mm/mremap.c
@@ -324,10 +324,62 @@ static inline bool move_normal_pud(struct vm_area_struct *vma,
}
#endif
+
+#ifdef CONFIG_TRANSPARENT_HUGEPAGE_PUD
+static bool move_huge_pud(struct vm_area_struct *vma, unsigned long old_addr,
+ unsigned long new_addr, pud_t *old_pud, pud_t *new_pud)
+{
+ spinlock_t *old_ptl, *new_ptl;
+ struct mm_struct *mm = vma->vm_mm;
+ pud_t pud;
+
+ /*
+ * The destination pud shouldn't be established, free_pgtables()
+ * should have released it.
+ */
+ if (WARN_ON_ONCE(!pud_none(*new_pud)))
+ return false;
+
+ /*
+ * We don't have to worry about the ordering of src and dst
+ * ptlocks because exclusive mmap_lock prevents deadlock.
+ */
+ old_ptl = pud_lock(vma->vm_mm, old_pud);
+ new_ptl = pud_lockptr(mm, new_pud);
+ if (new_ptl != old_ptl)
+ spin_lock_nested(new_ptl, SINGLE_DEPTH_NESTING);
+
+ /* Clear the pud */
+ pud = *old_pud;
+ pud_clear(old_pud);
+
+ VM_BUG_ON(!pud_none(*new_pud));
+
+ /* Set the new pud */
+ /* mark soft_ditry when we add pud level soft dirty support */
+ set_pud_at(mm, new_addr, new_pud, pud);
+ flush_pud_tlb_range(vma, old_addr, old_addr + HPAGE_PUD_SIZE);
+ if (new_ptl != old_ptl)
+ spin_unlock(new_ptl);
+ spin_unlock(old_ptl);
+
+ return true;
+}
+#else
+static bool move_huge_pud(struct vm_area_struct *vma, unsigned long old_addr,
+ unsigned long new_addr, pud_t *old_pud, pud_t *new_pud)
+{
+ WARN_ON_ONCE(1);
+ return false;
+
+}
+#endif
+
enum pgt_entry {
NORMAL_PMD,
HPAGE_PMD,
NORMAL_PUD,
+ HPAGE_PUD,
};
/*
@@ -347,6 +399,7 @@ static __always_inline unsigned long get_extent(enum pgt_entry entry,
mask = PMD_MASK;
size = PMD_SIZE;
break;
+ case HPAGE_PUD:
case NORMAL_PUD:
mask = PUD_MASK;
size = PUD_SIZE;
@@ -395,6 +448,11 @@ static bool move_pgt_entry(enum pgt_entry entry, struct vm_area_struct *vma,
move_huge_pmd(vma, old_addr, new_addr, old_entry,
new_entry);
break;
+ case HPAGE_PUD:
+ moved = move_huge_pud(vma, old_addr, new_addr, old_entry,
+ new_entry);
+ break;
+
default:
WARN_ON_ONCE(1);
break;
@@ -414,6 +472,7 @@ unsigned long move_page_tables(struct vm_area_struct *vma,
unsigned long extent, old_end;
struct mmu_notifier_range range;
pmd_t *old_pmd, *new_pmd;
+ pud_t *old_pud, *new_pud;
old_end = old_addr + len;
flush_cache_range(vma, old_addr, old_end);
@@ -429,15 +488,22 @@ unsigned long move_page_tables(struct vm_area_struct *vma,
* PUD level if possible.
*/
extent = get_extent(NORMAL_PUD, old_addr, old_end, new_addr);
- if (IS_ENABLED(CONFIG_HAVE_MOVE_PUD) && extent == PUD_SIZE) {
- pud_t *old_pud, *new_pud;
- old_pud = get_old_pud(vma->vm_mm, old_addr);
- if (!old_pud)
+ old_pud = get_old_pud(vma->vm_mm, old_addr);
+ if (!old_pud)
+ continue;
+ new_pud = alloc_new_pud(vma->vm_mm, vma, new_addr);
+ if (!new_pud)
+ break;
+ if (pud_trans_huge(*old_pud) || pud_devmap(*old_pud)) {
+ if (extent == HPAGE_PUD_SIZE) {
+ move_pgt_entry(HPAGE_PUD, vma, old_addr, new_addr,
+ old_pud, new_pud, need_rmap_locks);
+ /* We ignore and continue on error? */
continue;
- new_pud = alloc_new_pud(vma->vm_mm, vma, new_addr);
- if (!new_pud)
- break;
+ }
+ } else if (IS_ENABLED(CONFIG_HAVE_MOVE_PUD) && extent == PUD_SIZE) {
+
if (move_pgt_entry(NORMAL_PUD, vma, old_addr, new_addr,
old_pud, new_pud, need_rmap_locks))
continue;
--
2.31.1
^ permalink raw reply related [flat|nested] 20+ messages in thread* [PATCH v6 04/11] mm/mremap: Use pmd/pud_poplulate to update page table entries
2021-05-24 9:01 [PATCH v6 00/11] Speedup mremap on ppc64 Aneesh Kumar K.V
` (2 preceding siblings ...)
2021-05-24 9:01 ` [PATCH v6 03/11] mm/mremap: Convert huge PUD move to separate helper Aneesh Kumar K.V
@ 2021-05-24 9:01 ` Aneesh Kumar K.V
2021-05-24 9:01 ` [PATCH v6 05/11] powerpc/mm/book3s64: Fix possible build error Aneesh Kumar K.V
` (6 subsequent siblings)
10 siblings, 0 replies; 20+ messages in thread
From: Aneesh Kumar K.V @ 2021-05-24 9:01 UTC (permalink / raw)
To: linux-mm, akpm
Cc: Aneesh Kumar K.V, Linus Torvalds, npiggin, kaleshsingh, joel,
linuxppc-dev
pmd/pud_populate is the right interface to be used to set the respective
page table entries. Some architectures like ppc64 do assume that set_pmd/pud_at
can only be used to set a hugepage PTE. Since we are not setting up a hugepage
PTE here, use the pmd/pud_populate interface.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
---
mm/mremap.c | 7 +++----
1 file changed, 3 insertions(+), 4 deletions(-)
diff --git a/mm/mremap.c b/mm/mremap.c
index 1d6fadbd4820..7372c8c0cf26 100644
--- a/mm/mremap.c
+++ b/mm/mremap.c
@@ -26,6 +26,7 @@
#include <asm/cacheflush.h>
#include <asm/tlbflush.h>
+#include <asm/pgalloc.h>
#include "internal.h"
@@ -257,9 +258,8 @@ static bool move_normal_pmd(struct vm_area_struct *vma, unsigned long old_addr,
pmd_clear(old_pmd);
VM_BUG_ON(!pmd_none(*new_pmd));
+ pmd_populate(mm, new_pmd, pmd_pgtable(pmd));
- /* Set the new pmd */
- set_pmd_at(mm, new_addr, new_pmd, pmd);
flush_tlb_range(vma, old_addr, old_addr + PMD_SIZE);
if (new_ptl != old_ptl)
spin_unlock(new_ptl);
@@ -306,8 +306,7 @@ static bool move_normal_pud(struct vm_area_struct *vma, unsigned long old_addr,
VM_BUG_ON(!pud_none(*new_pud));
- /* Set the new pud */
- set_pud_at(mm, new_addr, new_pud, pud);
+ pud_populate(mm, new_pud, (pmd_t *)pud_page_vaddr(pud));
flush_tlb_range(vma, old_addr, old_addr + PUD_SIZE);
if (new_ptl != old_ptl)
spin_unlock(new_ptl);
--
2.31.1
^ permalink raw reply related [flat|nested] 20+ messages in thread* [PATCH v6 05/11] powerpc/mm/book3s64: Fix possible build error
2021-05-24 9:01 [PATCH v6 00/11] Speedup mremap on ppc64 Aneesh Kumar K.V
` (3 preceding siblings ...)
2021-05-24 9:01 ` [PATCH v6 04/11] mm/mremap: Use pmd/pud_poplulate to update page table entries Aneesh Kumar K.V
@ 2021-05-24 9:01 ` Aneesh Kumar K.V
2021-05-24 9:01 ` [PATCH v6 06/11] powerpc/mm/book3s64: Update tlb flush routines to take a page walk cache flush argument Aneesh Kumar K.V
` (5 subsequent siblings)
10 siblings, 0 replies; 20+ messages in thread
From: Aneesh Kumar K.V @ 2021-05-24 9:01 UTC (permalink / raw)
To: linux-mm, akpm
Cc: Aneesh Kumar K.V, Linus Torvalds, npiggin, kaleshsingh, joel,
linuxppc-dev
Update _tlbiel_pid() such that we can avoid build errors like below when
using this function in other places.
arch/powerpc/mm/book3s64/radix_tlb.c: In function ‘__radix__flush_tlb_range_psize’:
arch/powerpc/mm/book3s64/radix_tlb.c:114:2: warning: ‘asm’ operand 3 probably does not match constraints
114 | asm volatile(PPC_TLBIEL(%0, %4, %3, %2, %1)
| ^~~
arch/powerpc/mm/book3s64/radix_tlb.c:114:2: error: impossible constraint in ‘asm’
make[4]: *** [scripts/Makefile.build:271: arch/powerpc/mm/book3s64/radix_tlb.o] Error 1
m
With this fix, we can also drop the __always_inline in __radix_flush_tlb_range_psize
which was added by commit e12d6d7d46a6 ("powerpc/mm/radix: mark __radix__flush_tlb_range_psize() as __always_inline")
Reviewed-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Acked-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
---
arch/powerpc/mm/book3s64/radix_tlb.c | 26 +++++++++++++++++---------
1 file changed, 17 insertions(+), 9 deletions(-)
diff --git a/arch/powerpc/mm/book3s64/radix_tlb.c b/arch/powerpc/mm/book3s64/radix_tlb.c
index 409e61210789..817a02ef6032 100644
--- a/arch/powerpc/mm/book3s64/radix_tlb.c
+++ b/arch/powerpc/mm/book3s64/radix_tlb.c
@@ -291,22 +291,30 @@ static inline void fixup_tlbie_lpid(unsigned long lpid)
/*
* We use 128 set in radix mode and 256 set in hpt mode.
*/
-static __always_inline void _tlbiel_pid(unsigned long pid, unsigned long ric)
+static inline void _tlbiel_pid(unsigned long pid, unsigned long ric)
{
int set;
asm volatile("ptesync": : :"memory");
- /*
- * Flush the first set of the TLB, and if we're doing a RIC_FLUSH_ALL,
- * also flush the entire Page Walk Cache.
- */
- __tlbiel_pid(pid, 0, ric);
+ switch (ric) {
+ case RIC_FLUSH_PWC:
- /* For PWC, only one flush is needed */
- if (ric == RIC_FLUSH_PWC) {
+ /* For PWC, only one flush is needed */
+ __tlbiel_pid(pid, 0, RIC_FLUSH_PWC);
ppc_after_tlbiel_barrier();
return;
+ case RIC_FLUSH_TLB:
+ __tlbiel_pid(pid, 0, RIC_FLUSH_TLB);
+ break;
+ case RIC_FLUSH_ALL:
+ default:
+ /*
+ * Flush the first set of the TLB, and if
+ * we're doing a RIC_FLUSH_ALL, also flush
+ * the entire Page Walk Cache.
+ */
+ __tlbiel_pid(pid, 0, RIC_FLUSH_ALL);
}
if (!cpu_has_feature(CPU_FTR_ARCH_31)) {
@@ -1176,7 +1184,7 @@ void radix__tlb_flush(struct mmu_gather *tlb)
}
}
-static __always_inline void __radix__flush_tlb_range_psize(struct mm_struct *mm,
+static void __radix__flush_tlb_range_psize(struct mm_struct *mm,
unsigned long start, unsigned long end,
int psize, bool also_pwc)
{
--
2.31.1
^ permalink raw reply related [flat|nested] 20+ messages in thread* [PATCH v6 06/11] powerpc/mm/book3s64: Update tlb flush routines to take a page walk cache flush argument
2021-05-24 9:01 [PATCH v6 00/11] Speedup mremap on ppc64 Aneesh Kumar K.V
` (4 preceding siblings ...)
2021-05-24 9:01 ` [PATCH v6 05/11] powerpc/mm/book3s64: Fix possible build error Aneesh Kumar K.V
@ 2021-05-24 9:01 ` Aneesh Kumar K.V
2021-05-24 9:01 ` [PATCH v6 07/11] mm/mremap: Use range flush that does TLB and page walk cache flush Aneesh Kumar K.V
` (4 subsequent siblings)
10 siblings, 0 replies; 20+ messages in thread
From: Aneesh Kumar K.V @ 2021-05-24 9:01 UTC (permalink / raw)
To: linux-mm, akpm
Cc: Aneesh Kumar K.V, Linus Torvalds, npiggin, kaleshsingh, joel,
linuxppc-dev
No functional change in this patch
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
---
.../include/asm/book3s/64/tlbflush-radix.h | 19 +++++++-----
arch/powerpc/include/asm/book3s/64/tlbflush.h | 23 ++++++++++++---
arch/powerpc/mm/book3s64/radix_hugetlbpage.c | 4 +--
arch/powerpc/mm/book3s64/radix_tlb.c | 29 +++++++------------
4 files changed, 42 insertions(+), 33 deletions(-)
diff --git a/arch/powerpc/include/asm/book3s/64/tlbflush-radix.h b/arch/powerpc/include/asm/book3s/64/tlbflush-radix.h
index 8b33601cdb9d..171441a43b35 100644
--- a/arch/powerpc/include/asm/book3s/64/tlbflush-radix.h
+++ b/arch/powerpc/include/asm/book3s/64/tlbflush-radix.h
@@ -56,15 +56,18 @@ static inline void radix__flush_all_lpid_guest(unsigned int lpid)
}
#endif
-extern void radix__flush_hugetlb_tlb_range(struct vm_area_struct *vma,
- 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);
-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,
+void radix__flush_hugetlb_tlb_range(struct vm_area_struct *vma,
+ unsigned long start, unsigned long end,
+ bool flush_pwc);
+void radix__flush_pmd_tlb_range(struct vm_area_struct *vma,
+ unsigned long start, unsigned long end,
+ bool flush_pwc);
+void radix__flush_tlb_pwc_range_psize(struct mm_struct *mm, unsigned long start,
+ unsigned long end, int psize, bool flush_pwc);
+void radix__flush_tlb_range(struct vm_area_struct *vma, unsigned long start,
unsigned long end);
-extern void radix__flush_tlb_kernel_range(unsigned long start, unsigned long end);
+void radix__flush_tlb_kernel_range(unsigned long start, unsigned long end);
+
extern void radix__local_flush_tlb_mm(struct mm_struct *mm);
extern void radix__local_flush_all_mm(struct mm_struct *mm);
diff --git a/arch/powerpc/include/asm/book3s/64/tlbflush.h b/arch/powerpc/include/asm/book3s/64/tlbflush.h
index 215973b4cb26..f9f8a3a264f7 100644
--- a/arch/powerpc/include/asm/book3s/64/tlbflush.h
+++ b/arch/powerpc/include/asm/book3s/64/tlbflush.h
@@ -45,13 +45,30 @@ static inline void tlbiel_all_lpid(bool radix)
hash__tlbiel_all(TLB_INVAL_SCOPE_LPID);
}
+static inline void flush_pmd_tlb_pwc_range(struct vm_area_struct *vma,
+ unsigned long start,
+ unsigned long end,
+ bool flush_pwc)
+{
+ if (radix_enabled())
+ return radix__flush_pmd_tlb_range(vma, start, end, flush_pwc);
+ return hash__flush_tlb_range(vma, start, end);
+}
#define __HAVE_ARCH_FLUSH_PMD_TLB_RANGE
static inline void flush_pmd_tlb_range(struct vm_area_struct *vma,
unsigned long start, unsigned long end)
+{
+ return flush_pmd_tlb_pwc_range(vma, start, end, false);
+}
+
+static inline void flush_hugetlb_tlb_pwc_range(struct vm_area_struct *vma,
+ unsigned long start,
+ unsigned long end,
+ bool flush_pwc)
{
if (radix_enabled())
- return radix__flush_pmd_tlb_range(vma, start, end);
+ return radix__flush_hugetlb_tlb_range(vma, start, end, flush_pwc);
return hash__flush_tlb_range(vma, start, end);
}
@@ -60,9 +77,7 @@ static inline void flush_hugetlb_tlb_range(struct vm_area_struct *vma,
unsigned long start,
unsigned long end)
{
- if (radix_enabled())
- return radix__flush_hugetlb_tlb_range(vma, start, end);
- return hash__flush_tlb_range(vma, start, end);
+ return flush_hugetlb_tlb_pwc_range(vma, start, end, false);
}
static inline void flush_tlb_range(struct vm_area_struct *vma,
diff --git a/arch/powerpc/mm/book3s64/radix_hugetlbpage.c b/arch/powerpc/mm/book3s64/radix_hugetlbpage.c
index cb91071eef52..e62f5679b119 100644
--- a/arch/powerpc/mm/book3s64/radix_hugetlbpage.c
+++ b/arch/powerpc/mm/book3s64/radix_hugetlbpage.c
@@ -26,13 +26,13 @@ void radix__local_flush_hugetlb_page(struct vm_area_struct *vma, unsigned long v
}
void radix__flush_hugetlb_tlb_range(struct vm_area_struct *vma, unsigned long start,
- unsigned long end)
+ unsigned long end, bool flush_pwc)
{
int psize;
struct hstate *hstate = hstate_file(vma->vm_file);
psize = hstate_get_psize(hstate);
- radix__flush_tlb_range_psize(vma->vm_mm, start, end, psize);
+ radix__flush_tlb_pwc_range_psize(vma->vm_mm, start, end, psize, flush_pwc);
}
/*
diff --git a/arch/powerpc/mm/book3s64/radix_tlb.c b/arch/powerpc/mm/book3s64/radix_tlb.c
index 817a02ef6032..5a59e19f9e53 100644
--- a/arch/powerpc/mm/book3s64/radix_tlb.c
+++ b/arch/powerpc/mm/book3s64/radix_tlb.c
@@ -1090,7 +1090,7 @@ void radix__flush_tlb_range(struct vm_area_struct *vma, unsigned long start,
{
#ifdef CONFIG_HUGETLB_PAGE
if (is_vm_hugetlb_page(vma))
- return radix__flush_hugetlb_tlb_range(vma, start, end);
+ return radix__flush_hugetlb_tlb_range(vma, start, end, false);
#endif
__radix__flush_tlb_range(vma->vm_mm, start, end);
@@ -1151,9 +1151,6 @@ void radix__flush_all_lpid_guest(unsigned int lpid)
_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;
@@ -1177,10 +1174,8 @@ void radix__tlb_flush(struct mmu_gather *tlb)
else
radix__flush_all_mm(mm);
} else {
- if (!tlb->freed_tables)
- radix__flush_tlb_range_psize(mm, start, end, psize);
- else
- radix__flush_tlb_pwc_range_psize(mm, start, end, psize);
+ radix__flush_tlb_pwc_range_psize(mm, start,
+ end, psize, tlb->freed_tables);
}
}
@@ -1254,16 +1249,10 @@ static void __radix__flush_tlb_range_psize(struct mm_struct *mm,
preempt_enable();
}
-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, bool flush_pwc)
{
- 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)
-{
- __radix__flush_tlb_range_psize(mm, start, end, psize, true);
+ __radix__flush_tlb_range_psize(mm, start, end, psize, flush_pwc);
}
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
@@ -1315,9 +1304,11 @@ void radix__flush_tlb_collapsed_pmd(struct mm_struct *mm, unsigned long addr)
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
void radix__flush_pmd_tlb_range(struct vm_area_struct *vma,
- unsigned long start, unsigned long end)
+ unsigned long start, unsigned long end,
+ bool flush_pwc)
{
- radix__flush_tlb_range_psize(vma->vm_mm, start, end, MMU_PAGE_2M);
+ __radix__flush_tlb_range_psize(vma->vm_mm, start,
+ end, MMU_PAGE_2M, flush_pwc);
}
EXPORT_SYMBOL(radix__flush_pmd_tlb_range);
--
2.31.1
^ permalink raw reply related [flat|nested] 20+ messages in thread* [PATCH v6 07/11] mm/mremap: Use range flush that does TLB and page walk cache flush
2021-05-24 9:01 [PATCH v6 00/11] Speedup mremap on ppc64 Aneesh Kumar K.V
` (5 preceding siblings ...)
2021-05-24 9:01 ` [PATCH v6 06/11] powerpc/mm/book3s64: Update tlb flush routines to take a page walk cache flush argument Aneesh Kumar K.V
@ 2021-05-24 9:01 ` Aneesh Kumar K.V
2021-05-24 17:02 ` Linus Torvalds
2021-05-24 9:01 ` [PATCH v6 08/11] mm/mremap: properly flush the TLB on mremap Aneesh Kumar K.V
` (3 subsequent siblings)
10 siblings, 1 reply; 20+ messages in thread
From: Aneesh Kumar K.V @ 2021-05-24 9:01 UTC (permalink / raw)
To: linux-mm, akpm
Cc: Aneesh Kumar K.V, Linus Torvalds, npiggin, kaleshsingh, joel,
linuxppc-dev
Some architectures do have the concept of page walk cache which need
to be flush when updating higher levels of page tables. A fast mremap
that involves moving page table pages instead of copying pte entries
should flush page walk cache since the old translation cache is no more
valid.
Add new helper flush_pte_tlb_pwc_range() which invalidates both TLB and
page walk cache where TLB entries are mapped with page size PAGE_SIZE.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
---
arch/powerpc/include/asm/book3s/64/tlbflush.h | 10 ++++++++++
mm/mremap.c | 14 ++++++++++++--
2 files changed, 22 insertions(+), 2 deletions(-)
diff --git a/arch/powerpc/include/asm/book3s/64/tlbflush.h b/arch/powerpc/include/asm/book3s/64/tlbflush.h
index f9f8a3a264f7..e84fee9db106 100644
--- a/arch/powerpc/include/asm/book3s/64/tlbflush.h
+++ b/arch/powerpc/include/asm/book3s/64/tlbflush.h
@@ -80,6 +80,16 @@ static inline void flush_hugetlb_tlb_range(struct vm_area_struct *vma,
return flush_hugetlb_tlb_pwc_range(vma, start, end, false);
}
+#define flush_pte_tlb_pwc_range flush_tlb_pwc_range
+static inline void flush_pte_tlb_pwc_range(struct vm_area_struct *vma,
+ unsigned long start, unsigned long end)
+{
+ if (radix_enabled())
+ return radix__flush_tlb_pwc_range_psize(vma->vm_mm, start,
+ end, mmu_virtual_psize, true);
+ return hash__flush_tlb_range(vma, start, end);
+}
+
static inline void flush_tlb_range(struct vm_area_struct *vma,
unsigned long start, unsigned long end)
{
diff --git a/mm/mremap.c b/mm/mremap.c
index 7372c8c0cf26..000a71917557 100644
--- a/mm/mremap.c
+++ b/mm/mremap.c
@@ -210,6 +210,16 @@ static void move_ptes(struct vm_area_struct *vma, pmd_t *old_pmd,
drop_rmap_locks(vma);
}
+#ifndef flush_pte_tlb_pwc_range
+#define flush_pte_tlb_pwc_range flush_pte_tlb_pwc_range
+static inline void flush_pte_tlb_pwc_range(struct vm_area_struct *vma,
+ unsigned long start,
+ unsigned long end)
+{
+ return flush_tlb_range(vma, start, end);
+}
+#endif
+
#ifdef CONFIG_HAVE_MOVE_PMD
static bool move_normal_pmd(struct vm_area_struct *vma, unsigned long old_addr,
unsigned long new_addr, pmd_t *old_pmd, pmd_t *new_pmd)
@@ -260,7 +270,7 @@ static bool move_normal_pmd(struct vm_area_struct *vma, unsigned long old_addr,
VM_BUG_ON(!pmd_none(*new_pmd));
pmd_populate(mm, new_pmd, pmd_pgtable(pmd));
- flush_tlb_range(vma, old_addr, old_addr + PMD_SIZE);
+ flush_pte_tlb_pwc_range(vma, old_addr, old_addr + PMD_SIZE);
if (new_ptl != old_ptl)
spin_unlock(new_ptl);
spin_unlock(old_ptl);
@@ -307,7 +317,7 @@ static bool move_normal_pud(struct vm_area_struct *vma, unsigned long old_addr,
VM_BUG_ON(!pud_none(*new_pud));
pud_populate(mm, new_pud, (pmd_t *)pud_page_vaddr(pud));
- flush_tlb_range(vma, old_addr, old_addr + PUD_SIZE);
+ flush_pte_tlb_pwc_range(vma, old_addr, old_addr + PUD_SIZE);
if (new_ptl != old_ptl)
spin_unlock(new_ptl);
spin_unlock(old_ptl);
--
2.31.1
^ permalink raw reply related [flat|nested] 20+ messages in thread* Re: [PATCH v6 07/11] mm/mremap: Use range flush that does TLB and page walk cache flush
2021-05-24 9:01 ` [PATCH v6 07/11] mm/mremap: Use range flush that does TLB and page walk cache flush Aneesh Kumar K.V
@ 2021-05-24 17:02 ` Linus Torvalds
2021-05-25 13:27 ` Aneesh Kumar K.V
0 siblings, 1 reply; 20+ messages in thread
From: Linus Torvalds @ 2021-05-24 17:02 UTC (permalink / raw)
To: Aneesh Kumar K.V
Cc: Nick Piggin, Linux-MM, Kalesh Singh, Joel Fernandes,
Andrew Morton, linuxppc-dev
On Sun, May 23, 2021 at 11:04 PM Aneesh Kumar K.V
<aneesh.kumar@linux.ibm.com> wrote:
>
> Add new helper flush_pte_tlb_pwc_range() which invalidates both TLB and
> page walk cache where TLB entries are mapped with page size PAGE_SIZE.
So I dislike this patch for two reasons:
(a) naming.
If the ppc people want to use crazy TLA's that have no meaning outside
of the powerpc community, that's fine. But only in powerpc code.
"pwc" makes no sense to me, or to anybody else that isn't intimately
involved in low-level powerpc stuff. I assume it's "page walk cache",
but honestly, outside of this area, PWC is mostly used for a specific
type of webcam.
So there's no way I'd accept this as-is, simply because of that.
flush_pte_tlb_pwc_range() is simply not an acceptable name. You would
have to spell it out, not use an obscure TLA.
But I think you don't even want to do that, because of
(b) is this even worth it as a public interface?
Why doesn't the powerpc radix TLB flushing code just always flush the
page table walking cache when the range is larger than a PMD?
Once you have big flush ranges like that, I don't believe it makes any
sense not to flush the walking cache too.
NOTE! This is particularly true as "flush the walking cache" isn't a
well-defined operation anyway. Which _levels_ of the walking cache?
Again, the size (and alignment) of the flush would actually tell you.
A new boolean "flush" parameter does *NOT* tell that at all.
So I think this new interface is mis-named, but I also think it's
pointless. Just DTRT automatically when somebody asks for a flush that
covers a PMD range (or a PUD range).
Linus
^ permalink raw reply [flat|nested] 20+ messages in thread* Re: [PATCH v6 07/11] mm/mremap: Use range flush that does TLB and page walk cache flush
2021-05-24 17:02 ` Linus Torvalds
@ 2021-05-25 13:27 ` Aneesh Kumar K.V
2021-05-25 17:08 ` Linus Torvalds
0 siblings, 1 reply; 20+ messages in thread
From: Aneesh Kumar K.V @ 2021-05-25 13:27 UTC (permalink / raw)
To: Linus Torvalds
Cc: Nick Piggin, Linux-MM, Kalesh Singh, Joel Fernandes,
Andrew Morton, linuxppc-dev
Linus Torvalds <torvalds@linux-foundation.org> writes:
> On Sun, May 23, 2021 at 11:04 PM Aneesh Kumar K.V
> <aneesh.kumar@linux.ibm.com> wrote:
>>
>> Add new helper flush_pte_tlb_pwc_range() which invalidates both TLB and
>> page walk cache where TLB entries are mapped with page size PAGE_SIZE.
>
> So I dislike this patch for two reasons:
>
> (a) naming.
>
> If the ppc people want to use crazy TLA's that have no meaning outside
> of the powerpc community, that's fine. But only in powerpc code.
>
> "pwc" makes no sense to me, or to anybody else that isn't intimately
> involved in low-level powerpc stuff. I assume it's "page walk cache",
> but honestly, outside of this area, PWC is mostly used for a specific
> type of webcam.
>
> So there's no way I'd accept this as-is, simply because of that.
> flush_pte_tlb_pwc_range() is simply not an acceptable name. You would
> have to spell it out, not use an obscure TLA.
>
> But I think you don't even want to do that, because of
How about flush_tlb_and_page_table_cache() ?
>
> (b) is this even worth it as a public interface?
>
> Why doesn't the powerpc radix TLB flushing code just always flush the
> page table walking cache when the range is larger than a PMD?
>
> Once you have big flush ranges like that, I don't believe it makes any
> sense not to flush the walking cache too.
But such a large range invalidate doesn't imply we are freeing page
tables. Hence forcing a page table cache flush for large range
invalidate can have performance impact. ppc64 don't do a range page
table cache invalidate. Hence we will have to flush the full page table
cache.
>
> NOTE! This is particularly true as "flush the walking cache" isn't a
> well-defined operation anyway. Which _levels_ of the walking cache?
> Again, the size (and alignment) of the flush would actually tell you.
> A new boolean "flush" parameter does *NOT* tell that at all.
>
> So I think this new interface is mis-named, but I also think it's
> pointless. Just DTRT automatically when somebody asks for a flush that
> covers a PMD range (or a PUD range).
>
> Linus
-aneesh
^ permalink raw reply [flat|nested] 20+ messages in thread
* Re: [PATCH v6 07/11] mm/mremap: Use range flush that does TLB and page walk cache flush
2021-05-25 13:27 ` Aneesh Kumar K.V
@ 2021-05-25 17:08 ` Linus Torvalds
0 siblings, 0 replies; 20+ messages in thread
From: Linus Torvalds @ 2021-05-25 17:08 UTC (permalink / raw)
To: Aneesh Kumar K.V
Cc: Nick Piggin, Linux-MM, Kalesh Singh, Joel Fernandes,
Andrew Morton, linuxppc-dev
On Tue, May 25, 2021 at 3:28 AM Aneesh Kumar K.V
<aneesh.kumar@linux.ibm.com> wrote:
>
> How about flush_tlb_and_page_table_cache() ?
Honestly, I'd prefer it to be a separate function.
So keep the existing
flush_tlb()
as-is, and add a
flush_tlb_walking_cache()
and document that any architecture that flushes the walker caches as
part of the regular tlb flush can just make that a no-op.
Would that work for powerpc?
But:
> > (b) is this even worth it as a public interface?
>
> But such a large range invalidate doesn't imply we are freeing page
> tables.
No. But it's what everybody else (ie x86 and ARM) does, and if you're
flushing megabytes of TLB's, what's the downside of flushing a few TLB
walker cache entries?
You already do that for internal powerpc errata anyway (ie
"mm_needs_flush_escalation()"), so I'm saying that you might as well
treat the page walker cache as a powerpc-internal implementation
thing.
Put another way: can you even _measure_ the difference between "just
make powerpc look like everybody else" and "add a new explicit page
table walker cache flush function interface"?
Now, from a quick look at the powerpc code, it looks like powerpc is a
bit mis-architected, and when you flush the walker cache, you flush
everything for that ASID. x86 and arm only flush the parts affected by
the TLB flush range (now, admittedly, that's what they do
_architecturally_ - for all I know the actual hardware just always
flushes all walker caches when you flush any TLB and so maybe they act
exactly like the powrpc RIC_FLUSH_PWC in practice).
So maybe it's measurable. But I kind of doubt it, and I'd like to know
that you've actually done some testing to see that "yes, this matters,
I can't just say 'if flushing more than a pmd, just flush the walker
cache too'".
Hmm?
Linus
^ permalink raw reply [flat|nested] 20+ messages in thread
* [PATCH v6 08/11] mm/mremap: properly flush the TLB on mremap.
2021-05-24 9:01 [PATCH v6 00/11] Speedup mremap on ppc64 Aneesh Kumar K.V
` (6 preceding siblings ...)
2021-05-24 9:01 ` [PATCH v6 07/11] mm/mremap: Use range flush that does TLB and page walk cache flush Aneesh Kumar K.V
@ 2021-05-24 9:01 ` Aneesh Kumar K.V
2021-05-24 9:01 ` [PATCH v6 09/11] mm/mremap: Fix race between mremap and pageout Aneesh Kumar K.V
` (2 subsequent siblings)
10 siblings, 0 replies; 20+ messages in thread
From: Aneesh Kumar K.V @ 2021-05-24 9:01 UTC (permalink / raw)
To: linux-mm, akpm
Cc: Aneesh Kumar K.V, Linus Torvalds, npiggin, kaleshsingh, joel,
linuxppc-dev
As explained in
commit eb66ae030829 ("mremap: properly flush TLB before releasing the page")
mremap is special in that it doesn't take ownership of the page. The
optimized version for PUD/PMD aligned mremap also doesn't hold the ptl lock.
Hence flush the TLB before we update the new page table location. This ensures
the kernel invalidates the older translation cache before it can free the page via
the newly inserted translation.
Fixes: c49dd3401802 ("mm: speedup mremap on 1GB or larger regions")
Fixes: 2c91bd4a4e2e ("mm: speed up mremap by 20x on large regions")
Link: https://lore.kernel.org/linux-mm/CAHk-=wjq8thag3uNv-2MMu75OgX5ybMon7gZDUHYwzeTwcZHoA@mail.gmail.com
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
---
mm/mremap.c | 11 ++++++++---
1 file changed, 8 insertions(+), 3 deletions(-)
diff --git a/mm/mremap.c b/mm/mremap.c
index 000a71917557..8967a3707332 100644
--- a/mm/mremap.c
+++ b/mm/mremap.c
@@ -266,11 +266,13 @@ static bool move_normal_pmd(struct vm_area_struct *vma, unsigned long old_addr,
/* Clear the pmd */
pmd = *old_pmd;
pmd_clear(old_pmd);
-
+ /*
+ * flush the TLB before we move the page table entries.
+ */
+ flush_pte_tlb_pwc_range(vma, old_addr, old_addr + PMD_SIZE);
VM_BUG_ON(!pmd_none(*new_pmd));
pmd_populate(mm, new_pmd, pmd_pgtable(pmd));
- flush_pte_tlb_pwc_range(vma, old_addr, old_addr + PMD_SIZE);
if (new_ptl != old_ptl)
spin_unlock(new_ptl);
spin_unlock(old_ptl);
@@ -313,11 +315,14 @@ static bool move_normal_pud(struct vm_area_struct *vma, unsigned long old_addr,
/* Clear the pud */
pud = *old_pud;
pud_clear(old_pud);
+ /*
+ * flush the TLB before we move the page table entries.
+ */
+ flush_pte_tlb_pwc_range(vma, old_addr, old_addr + PUD_SIZE);
VM_BUG_ON(!pud_none(*new_pud));
pud_populate(mm, new_pud, (pmd_t *)pud_page_vaddr(pud));
- flush_pte_tlb_pwc_range(vma, old_addr, old_addr + PUD_SIZE);
if (new_ptl != old_ptl)
spin_unlock(new_ptl);
spin_unlock(old_ptl);
--
2.31.1
^ permalink raw reply related [flat|nested] 20+ messages in thread* [PATCH v6 09/11] mm/mremap: Fix race between mremap and pageout
2021-05-24 9:01 [PATCH v6 00/11] Speedup mremap on ppc64 Aneesh Kumar K.V
` (7 preceding siblings ...)
2021-05-24 9:01 ` [PATCH v6 08/11] mm/mremap: properly flush the TLB on mremap Aneesh Kumar K.V
@ 2021-05-24 9:01 ` Aneesh Kumar K.V
2021-05-24 13:38 ` [PATCH v6 updated 9/11] " Aneesh Kumar K.V
2021-05-24 9:01 ` [PATCH v6 10/11] mm/mremap: Allow arch runtime override Aneesh Kumar K.V
2021-05-24 9:01 ` [PATCH v6 11/11] powerpc/mm: Enable HAVE_MOVE_PMD support Aneesh Kumar K.V
10 siblings, 1 reply; 20+ messages in thread
From: Aneesh Kumar K.V @ 2021-05-24 9:01 UTC (permalink / raw)
To: linux-mm, akpm
Cc: Aneesh Kumar K.V, Linus Torvalds, npiggin, kaleshsingh, joel,
linuxppc-dev
CPU 1 CPU 2 CPU 3
mremap(old_addr, new_addr) page_shrinker/try_to_unmap_one
addr = old_addr
lock(pte_ptl)
lock(pmd_ptl)
pmd = *old_pmd
pmd_clear(old_pmd)
flush_tlb_range(old_addr)
*new_pmd = pmd
*new_addr = 10; and fills
TLB with new addr
and old pfn
unlock(pmd_ptl)
ptep_get_and_clear()
flush_tlb_range(old_addr)
old pfn is free.
Stale TLB entry
Avoid the above race with MOVE_PMD by holding pte ptl in mremap and waiting for
parallel pagetable walk to finish operating on pte before updating new_pmd
With MOVE_PUD only enable MOVE_PUD only if USE_SPLIT_PTE_PTLOCKS is disabled.
In this case both pte ptl and pud ptl points to mm->page_table_lock.
Fixes: c49dd3401802 ("mm: speedup mremap on 1GB or larger regions")
Fixes: 2c91bd4a4e2e ("mm: speed up mremap by 20x on large regions")
Link: https://lore.kernel.org/linux-mm/CAHk-=wgXVR04eBNtxQfevontWnP6FDm+oj5vauQXP3S-huwbPw@mail.gmail.com
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
---
mm/mremap.c | 24 +++++++++++++++++++++++-
1 file changed, 23 insertions(+), 1 deletion(-)
diff --git a/mm/mremap.c b/mm/mremap.c
index 8967a3707332..e70b8e3b9568 100644
--- a/mm/mremap.c
+++ b/mm/mremap.c
@@ -224,7 +224,7 @@ static inline void flush_pte_tlb_pwc_range(struct vm_area_struct *vma,
static bool move_normal_pmd(struct vm_area_struct *vma, unsigned long old_addr,
unsigned long new_addr, pmd_t *old_pmd, pmd_t *new_pmd)
{
- spinlock_t *old_ptl, *new_ptl;
+ spinlock_t *pte_ptl, *old_ptl, *new_ptl;
struct mm_struct *mm = vma->vm_mm;
pmd_t pmd;
@@ -254,6 +254,7 @@ static bool move_normal_pmd(struct vm_area_struct *vma, unsigned long old_addr,
if (WARN_ON_ONCE(!pmd_none(*new_pmd)))
return false;
+
/*
* We don't have to worry about the ordering of src and dst
* ptlocks because exclusive mmap_lock prevents deadlock.
@@ -263,6 +264,14 @@ static bool move_normal_pmd(struct vm_area_struct *vma, unsigned long old_addr,
if (new_ptl != old_ptl)
spin_lock_nested(new_ptl, SINGLE_DEPTH_NESTING);
+ if (pmd_none(*old_pmd))
+ goto unlock_out;
+
+ /*
+ * Take the ptl here so that we wait for parallel page table walk
+ * and operations (eg: pageout)using old addr to finish.
+ */
+ pte_ptl = pte_lockptr(mm, old_pmd);
/* Clear the pmd */
pmd = *old_pmd;
pmd_clear(old_pmd);
@@ -270,9 +279,14 @@ static bool move_normal_pmd(struct vm_area_struct *vma, unsigned long old_addr,
* flush the TLB before we move the page table entries.
*/
flush_pte_tlb_pwc_range(vma, old_addr, old_addr + PMD_SIZE);
+
+ spin_lock(pte_ptl);
+
VM_BUG_ON(!pmd_none(*new_pmd));
pmd_populate(mm, new_pmd, pmd_pgtable(pmd));
+ spin_unlock(pte_ptl);
+unlock_out:
if (new_ptl != old_ptl)
spin_unlock(new_ptl);
spin_unlock(old_ptl);
@@ -296,6 +310,14 @@ static bool move_normal_pud(struct vm_area_struct *vma, unsigned long old_addr,
struct mm_struct *mm = vma->vm_mm;
pud_t pud;
+ /*
+ * Disable MOVE_PUD until we get the pageout done with all
+ * higher level page table locks held. With SPLIT_PTE_PTLOCKS
+ * we use mm->page_table_lock for both pte ptl and pud ptl
+ */
+ if (USE_SPLIT_PTE_PTLOCKS)
+ return false;
+
/*
* The destination pud shouldn't be established, free_pgtables()
* should have released it.
--
2.31.1
^ permalink raw reply related [flat|nested] 20+ messages in thread* [PATCH v6 updated 9/11] mm/mremap: Fix race between mremap and pageout
2021-05-24 9:01 ` [PATCH v6 09/11] mm/mremap: Fix race between mremap and pageout Aneesh Kumar K.V
@ 2021-05-24 13:38 ` Aneesh Kumar K.V
0 siblings, 0 replies; 20+ messages in thread
From: Aneesh Kumar K.V @ 2021-05-24 13:38 UTC (permalink / raw)
To: linux-mm, akpm
Cc: Aneesh Kumar K.V, Linus Torvalds, npiggin, kaleshsingh, joel,
linuxppc-dev
CPU 1 CPU 2 CPU 3
mremap(old_addr, new_addr) page_shrinker/try_to_unmap_one
addr = old_addr
lock(pte_ptl)
lock(pmd_ptl)
pmd = *old_pmd
pmd_clear(old_pmd)
flush_tlb_range(old_addr)
*new_pmd = pmd
*new_addr = 10; and fills
TLB with new addr
and old pfn
unlock(pmd_ptl)
ptep_get_and_clear()
flush_tlb_range(old_addr)
old pfn is free.
Stale TLB entry
Avoid the above race with MOVE_PMD by holding pte ptl in mremap and waiting for
parallel pagetable walk to finish operating on pte before updating new_pmd
With MOVE_PUD only enable MOVE_PUD only if USE_SPLIT_PTE_PTLOCKS is disabled.
In this case both pte ptl and pud ptl points to mm->page_table_lock.
Fixes: c49dd3401802 ("mm: speedup mremap on 1GB or larger regions")
Fixes: 2c91bd4a4e2e ("mm: speed up mremap by 20x on large regions")
Link: https://lore.kernel.org/linux-mm/CAHk-=wgXVR04eBNtxQfevontWnP6FDm+oj5vauQXP3S-huwbPw@mail.gmail.com
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
---
Change:
* Check for split PTL before taking pte ptl lock.
mm/mremap.c | 26 +++++++++++++++++++++++++-
1 file changed, 25 insertions(+), 1 deletion(-)
diff --git a/mm/mremap.c b/mm/mremap.c
index 8967a3707332..2fa3e0cb6176 100644
--- a/mm/mremap.c
+++ b/mm/mremap.c
@@ -224,7 +224,7 @@ static inline void flush_pte_tlb_pwc_range(struct vm_area_struct *vma,
static bool move_normal_pmd(struct vm_area_struct *vma, unsigned long old_addr,
unsigned long new_addr, pmd_t *old_pmd, pmd_t *new_pmd)
{
- spinlock_t *old_ptl, *new_ptl;
+ spinlock_t *pte_ptl, *old_ptl, *new_ptl;
struct mm_struct *mm = vma->vm_mm;
pmd_t pmd;
@@ -254,6 +254,7 @@ static bool move_normal_pmd(struct vm_area_struct *vma, unsigned long old_addr,
if (WARN_ON_ONCE(!pmd_none(*new_pmd)))
return false;
+
/*
* We don't have to worry about the ordering of src and dst
* ptlocks because exclusive mmap_lock prevents deadlock.
@@ -263,6 +264,10 @@ static bool move_normal_pmd(struct vm_area_struct *vma, unsigned long old_addr,
if (new_ptl != old_ptl)
spin_lock_nested(new_ptl, SINGLE_DEPTH_NESTING);
+ if (pmd_none(*old_pmd))
+ goto unlock_out;
+
+ pte_ptl = pte_lockptr(mm, old_pmd);
/* Clear the pmd */
pmd = *old_pmd;
pmd_clear(old_pmd);
@@ -270,9 +275,20 @@ static bool move_normal_pmd(struct vm_area_struct *vma, unsigned long old_addr,
* flush the TLB before we move the page table entries.
*/
flush_pte_tlb_pwc_range(vma, old_addr, old_addr + PMD_SIZE);
+
+ /*
+ * Take the ptl here so that we wait for parallel page table walk
+ * and operations (eg: pageout)using old addr to finish.
+ */
+ if (USE_SPLIT_PTE_PTLOCKS)
+ spin_lock(pte_ptl);
+
VM_BUG_ON(!pmd_none(*new_pmd));
pmd_populate(mm, new_pmd, pmd_pgtable(pmd));
+ if (USE_SPLIT_PTE_PTLOCKS)
+ spin_unlock(pte_ptl);
+unlock_out:
if (new_ptl != old_ptl)
spin_unlock(new_ptl);
spin_unlock(old_ptl);
@@ -296,6 +312,14 @@ static bool move_normal_pud(struct vm_area_struct *vma, unsigned long old_addr,
struct mm_struct *mm = vma->vm_mm;
pud_t pud;
+ /*
+ * Disable MOVE_PUD until we get the pageout done with all
+ * higher level page table locks held. With SPLIT_PTE_PTLOCKS
+ * we use mm->page_table_lock for both pte ptl and pud ptl
+ */
+ if (USE_SPLIT_PTE_PTLOCKS)
+ return false;
+
/*
* The destination pud shouldn't be established, free_pgtables()
* should have released it.
--
2.31.1
^ permalink raw reply related [flat|nested] 20+ messages in thread
* [PATCH v6 10/11] mm/mremap: Allow arch runtime override
2021-05-24 9:01 [PATCH v6 00/11] Speedup mremap on ppc64 Aneesh Kumar K.V
` (8 preceding siblings ...)
2021-05-24 9:01 ` [PATCH v6 09/11] mm/mremap: Fix race between mremap and pageout Aneesh Kumar K.V
@ 2021-05-24 9:01 ` Aneesh Kumar K.V
2021-05-24 9:01 ` [PATCH v6 11/11] powerpc/mm: Enable HAVE_MOVE_PMD support Aneesh Kumar K.V
10 siblings, 0 replies; 20+ messages in thread
From: Aneesh Kumar K.V @ 2021-05-24 9:01 UTC (permalink / raw)
To: linux-mm, akpm
Cc: Aneesh Kumar K.V, Linus Torvalds, npiggin, kaleshsingh, joel,
linuxppc-dev
Architectures like ppc64 support faster mremap only with radix
translation. Hence allow a runtime check w.r.t support for fast mremap.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
---
arch/powerpc/include/asm/tlb.h | 6 ++++++
mm/mremap.c | 15 ++++++++++++++-
2 files changed, 20 insertions(+), 1 deletion(-)
diff --git a/arch/powerpc/include/asm/tlb.h b/arch/powerpc/include/asm/tlb.h
index 160422a439aa..09a9ae5f3656 100644
--- a/arch/powerpc/include/asm/tlb.h
+++ b/arch/powerpc/include/asm/tlb.h
@@ -83,5 +83,11 @@ static inline int mm_is_thread_local(struct mm_struct *mm)
}
#endif
+#define arch_supports_page_table_move arch_supports_page_table_move
+static inline bool arch_supports_page_table_move(void)
+{
+ return radix_enabled();
+}
+
#endif /* __KERNEL__ */
#endif /* __ASM_POWERPC_TLB_H */
diff --git a/mm/mremap.c b/mm/mremap.c
index e70b8e3b9568..42485d19c490 100644
--- a/mm/mremap.c
+++ b/mm/mremap.c
@@ -25,7 +25,7 @@
#include <linux/userfaultfd_k.h>
#include <asm/cacheflush.h>
-#include <asm/tlbflush.h>
+#include <asm/tlb.h>
#include <asm/pgalloc.h>
#include "internal.h"
@@ -220,6 +220,15 @@ static inline void flush_pte_tlb_pwc_range(struct vm_area_struct *vma,
}
#endif
+#ifndef arch_supports_page_table_move
+#define arch_supports_page_table_move arch_supports_page_table_move
+static inline bool arch_supports_page_table_move(void)
+{
+ return IS_ENABLED(CONFIG_HAVE_MOVE_PMD) ||
+ IS_ENABLED(CONFIG_HAVE_MOVE_PUD);
+}
+#endif
+
#ifdef CONFIG_HAVE_MOVE_PMD
static bool move_normal_pmd(struct vm_area_struct *vma, unsigned long old_addr,
unsigned long new_addr, pmd_t *old_pmd, pmd_t *new_pmd)
@@ -228,6 +237,8 @@ static bool move_normal_pmd(struct vm_area_struct *vma, unsigned long old_addr,
struct mm_struct *mm = vma->vm_mm;
pmd_t pmd;
+ if (!arch_supports_page_table_move())
+ return false;
/*
* The destination pmd shouldn't be established, free_pgtables()
* should have released it.
@@ -318,6 +329,8 @@ static bool move_normal_pud(struct vm_area_struct *vma, unsigned long old_addr,
if (USE_SPLIT_PTE_PTLOCKS)
return false;
+ if (!arch_supports_page_table_move())
+ return false;
/*
* The destination pud shouldn't be established, free_pgtables()
* should have released it.
--
2.31.1
^ permalink raw reply related [flat|nested] 20+ messages in thread* [PATCH v6 11/11] powerpc/mm: Enable HAVE_MOVE_PMD support
2021-05-24 9:01 [PATCH v6 00/11] Speedup mremap on ppc64 Aneesh Kumar K.V
` (9 preceding siblings ...)
2021-05-24 9:01 ` [PATCH v6 10/11] mm/mremap: Allow arch runtime override Aneesh Kumar K.V
@ 2021-05-24 9:01 ` Aneesh Kumar K.V
10 siblings, 0 replies; 20+ messages in thread
From: Aneesh Kumar K.V @ 2021-05-24 9:01 UTC (permalink / raw)
To: linux-mm, akpm
Cc: Aneesh Kumar K.V, Linus Torvalds, npiggin, kaleshsingh, joel,
linuxppc-dev
mremap HAVE_MOVE_PMD/PUD optimization time comparison for 1GB region:
1GB mremap - Source PTE-aligned, Destination PTE-aligned
mremap time: 1122062ns
1GB mremap - Source PMD-aligned, Destination PMD-aligned
mremap time: 522062ns
1GB mremap - Source PUD-aligned, Destination PUD-aligned
mremap time: 523345ns
There is not much impact with HAVE_MOVE_PUD because the kernel by default
enables USE_SPLIT_PTE_PTLOCKS. That results in the optimized pud move being
disabled.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
---
arch/powerpc/platforms/Kconfig.cputype | 2 ++
1 file changed, 2 insertions(+)
diff --git a/arch/powerpc/platforms/Kconfig.cputype b/arch/powerpc/platforms/Kconfig.cputype
index 3ce907523b1e..2e666e569fdf 100644
--- a/arch/powerpc/platforms/Kconfig.cputype
+++ b/arch/powerpc/platforms/Kconfig.cputype
@@ -97,6 +97,8 @@ config PPC_BOOK3S_64
select PPC_HAVE_PMU_SUPPORT
select SYS_SUPPORTS_HUGETLBFS
select HAVE_ARCH_TRANSPARENT_HUGEPAGE
+ select HAVE_MOVE_PMD
+ select HAVE_MOVE_PUD
select ARCH_ENABLE_THP_MIGRATION if TRANSPARENT_HUGEPAGE
select ARCH_SUPPORTS_NUMA_BALANCING
select IRQ_WORK
--
2.31.1
^ permalink raw reply related [flat|nested] 20+ messages in thread