Changelog
---------
v2 --> v3:
* Made orig_pte and orig_pmd a union per Mel (patch 1/7).
* Renamed pmd and oldpmd in patch 3/7 per Huang Ying.
* Used total_mapcount() instead of page_mapcount() in patch 6/7.
* Collected ack tags from Mel.
* Rebased to linux-next-20210513.
v1 --> v2:
* Adopted the suggestion from Gerald Schaefer to skip huge PMD for S390
for now.
* Used PageTransHuge to distinguish base page or THP instead of a new
parameter for migrate_misplaced_page() per Huang Ying.
* Restored PMD lazily to avoid unnecessary TLB shootdown per Huang Ying.
* Skipped shared THP.
* Updated counters correctly.
* Rebased to linux-next (next-20210412).
When the THP NUMA fault support was added THP migration was not supported yet.
So the ad hoc THP migration was implemented in NUMA fault handling. Since v4.14
THP migration has been supported so it doesn't make too much sense to still keep
another THP migration implementation rather than using the generic migration
code. It is definitely a maintenance burden to keep two THP migration
implementation for different code paths and it is more error prone. Using the
generic THP migration implementation allows us remove the duplicate code and
some hacks needed by the old ad hoc implementation.
A quick grep shows x86_64, PowerPC (book3s), ARM64 ans S390 support both THP
and NUMA balancing. The most of them support THP migration except for S390.
Zi Yan tried to add THP migration support for S390 before but it was not
accepted due to the design of S390 PMD. For the discussion, please see:
https://lkml.org/lkml/2018/4/27/953.
Per the discussion with Gerald Schaefer in v1 it is acceptible to skip huge
PMD for S390 for now.
I saw there were some hacks about gup from git history, but I didn't figure out
if they have been removed or not since I just found FOLL_NUMA code in the current
gup implementation and they seems useful.
Patch #1 ~ #2 are preparation patches.
Patch #3 is the real meat.
Patch #4 ~ #6 keep consistent counters and behaviors with before.
Patch #7 skips change huge PMD to prot_none if thp migration is not supported.
Test
----
Did some tests to measure the latency of do_huge_pmd_numa_page.
The test VM has 80 vcpus and 64G memory. The test would create 2
processes to consume 128G memory together which would incur memory
pressure to cause THP splits. And it also creates 80 processes to hog
cpu, and the memory consumer processes are bound to different nodes
periodically in order to increase NUMA faults.
The below test script is used:
echo 3 > /proc/sys/vm/drop_caches
# Run stress-ng for 24 hours
./stress-ng/stress-ng --vm 2 --vm-bytes 64G --timeout 24h &
PID=$!
./stress-ng/stress-ng --cpu $NR_CPUS --timeout 24h &
# Wait for vm stressors forked
sleep 5
PID_1=`pgrep -P $PID | awk 'NR == 1'`
PID_2=`pgrep -P $PID | awk 'NR == 2'`
JOB1=`pgrep -P $PID_1`
JOB2=`pgrep -P $PID_2`
# Bind load jobs to different nodes periodically to force generate
# cross node memory access
while [ -d "/proc/$PID" ]
do
taskset -apc 8 $JOB1
taskset -apc 8 $JOB2
sleep 300
taskset -apc 58 $JOB1
taskset -apc 58 $JOB2
sleep 300
done
With the above test the histogram of latency of do_huge_pmd_numa_page
is as shown below. Since the number of do_huge_pmd_numa_page varies
drastically for each run (should be due to scheduler), so I converted
the raw number to percentage.
patched base
@us[stress-ng]:
[0] 3.57% 0.16%
[1] 55.68% 18.36%
[2, 4) 10.46% 40.44%
[4, 8) 7.26% 17.82%
[8, 16) 21.12% 13.41%
[16, 32) 1.06% 4.27%
[32, 64) 0.56% 4.07%
[64, 128) 0.16% 0.35%
[128, 256) < 0.1% < 0.1%
[256, 512) < 0.1% < 0.1%
[512, 1K) < 0.1% < 0.1%
[1K, 2K) < 0.1% < 0.1%
[2K, 4K) < 0.1% < 0.1%
[4K, 8K) < 0.1% < 0.1%
[8K, 16K) < 0.1% < 0.1%
[16K, 32K) < 0.1% < 0.1%
[32K, 64K) < 0.1% < 0.1%
Per the result, patched kernel is even slightly better than the base
kernel. I think this is because the lock contention against THP split
is less than base kernel due to the refactor.
To exclude the affect from THP split, I also did test w/o memory
pressure. No obvious regression is spotted. The below is the test
result *w/o* memory pressure.
patched base
@us[stress-ng]:
[0] 7.97% 18.4%
[1] 69.63% 58.24%
[2, 4) 4.18% 2.63%
[4, 8) 0.22% 0.17%
[8, 16) 1.03% 0.92%
[16, 32) 0.14% < 0.1%
[32, 64) < 0.1% < 0.1%
[64, 128) < 0.1% < 0.1%
[128, 256) < 0.1% < 0.1%
[256, 512) 0.45% 1.19%
[512, 1K) 15.45% 17.27%
[1K, 2K) < 0.1% < 0.1%
[2K, 4K) < 0.1% < 0.1%
[4K, 8K) < 0.1% < 0.1%
[8K, 16K) 0.86% 0.88%
[16K, 32K) < 0.1% 0.15%
[32K, 64K) < 0.1% < 0.1%
[64K, 128K) < 0.1% < 0.1%
[128K, 256K) < 0.1% < 0.1%
The series also survived a series of tests that exercise NUMA balancing
migrations by Mel.
Yang Shi (7):
mm: memory: add orig_pmd to struct vm_fault
mm: memory: make numa_migrate_prep() non-static
mm: thp: refactor NUMA fault handling
mm: migrate: account THP NUMA migration counters correctly
mm: migrate: don't split THP for misplaced NUMA page
mm: migrate: check mapcount for THP instead of ref count
mm: thp: skip make PMD PROT_NONE if THP migration is not supported
include/linux/huge_mm.h | 9 ++---
include/linux/migrate.h | 23 -----------
include/linux/mm.h | 3 ++
mm/huge_memory.c | 156 +++++++++++++++++++++++++-----------------------------------------------
mm/internal.h | 21 ++--------
mm/memory.c | 31 +++++++--------
mm/migrate.c | 204 +++++++++++++++++++++--------------------------------------------------------------------------
7 files changed, 123 insertions(+), 324 deletions(-)
The generic migration path will check refcount, so no need check refcount here.
But the old code actually prevents from migrating shared THP (mapped by multiple
processes), so bail out early if mapcount is > 1 to keep the behavior.
Signed-off-by: Yang Shi <[email protected]>
---
mm/migrate.c | 16 ++++------------
1 file changed, 4 insertions(+), 12 deletions(-)
diff --git a/mm/migrate.c b/mm/migrate.c
index 035daaec1ca0..77215f28502b 100644
--- a/mm/migrate.c
+++ b/mm/migrate.c
@@ -2070,6 +2070,10 @@ static int numamigrate_isolate_page(pg_data_t *pgdat, struct page *page)
VM_BUG_ON_PAGE(compound_order(page) && !PageTransHuge(page), page);
+ /* Do not migrate THP mapped by multiple processes */
+ if (PageTransHuge(page) && total_mapcount(page) > 1)
+ return 0;
+
/* Avoid migrating to a node that is nearly full */
if (!migrate_balanced_pgdat(pgdat, compound_nr(page)))
return 0;
@@ -2077,18 +2081,6 @@ static int numamigrate_isolate_page(pg_data_t *pgdat, struct page *page)
if (isolate_lru_page(page))
return 0;
- /*
- * migrate_misplaced_transhuge_page() skips page migration's usual
- * check on page_count(), so we must do it here, now that the page
- * has been isolated: a GUP pin, or any other pin, prevents migration.
- * The expected page count is 3: 1 for page's mapcount and 1 for the
- * caller's pin and 1 for the reference taken by isolate_lru_page().
- */
- if (PageTransHuge(page) && page_count(page) != 3) {
- putback_lru_page(page);
- return 0;
- }
-
page_lru = page_is_file_lru(page);
mod_node_page_state(page_pgdat(page), NR_ISOLATED_ANON + page_lru,
thp_nr_pages(page));
--
2.26.2
The old behavior didn't split THP if migration is failed due to lack of
memory on the target node. But the THP migration does split THP, so keep
the old behavior for misplaced NUMA page migration.
Acked-by: Mel Gorman <[email protected]>
Signed-off-by: Yang Shi <[email protected]>
---
mm/migrate.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/mm/migrate.c b/mm/migrate.c
index 26f79f516100..035daaec1ca0 100644
--- a/mm/migrate.c
+++ b/mm/migrate.c
@@ -1420,6 +1420,7 @@ int migrate_pages(struct list_head *from, new_page_t get_new_page,
int swapwrite = current->flags & PF_SWAPWRITE;
int rc, nr_subpages;
LIST_HEAD(ret_pages);
+ bool nosplit = (reason == MR_NUMA_MISPLACED);
trace_mm_migrate_pages_start(mode, reason);
@@ -1491,8 +1492,9 @@ int migrate_pages(struct list_head *from, new_page_t get_new_page,
/*
* When memory is low, don't bother to try to migrate
* other pages, just exit.
+ * THP NUMA faulting doesn't split THP to retry.
*/
- if (is_thp) {
+ if (is_thp && !nosplit) {
if (!try_split_thp(page, &page2, from)) {
nr_thp_split++;
goto retry;
--
2.26.2
Now both base page and THP NUMA migration is done via migrate_misplaced_page(),
keep the counters correctly for THP.
Acked-by: Mel Gorman <[email protected]>
Signed-off-by: Yang Shi <[email protected]>
---
mm/migrate.c | 7 ++++---
1 file changed, 4 insertions(+), 3 deletions(-)
diff --git a/mm/migrate.c b/mm/migrate.c
index 4e0c2644256a..26f79f516100 100644
--- a/mm/migrate.c
+++ b/mm/migrate.c
@@ -2114,6 +2114,7 @@ int migrate_misplaced_page(struct page *page, struct vm_area_struct *vma,
LIST_HEAD(migratepages);
new_page_t *new;
bool compound;
+ unsigned int nr_pages = thp_nr_pages(page);
/*
* PTE mapped THP or HugeTLB page can't reach here so the page could
@@ -2152,13 +2153,13 @@ int migrate_misplaced_page(struct page *page, struct vm_area_struct *vma,
if (nr_remaining) {
if (!list_empty(&migratepages)) {
list_del(&page->lru);
- dec_node_page_state(page, NR_ISOLATED_ANON +
- page_is_file_lru(page));
+ mod_node_page_state(page_pgdat(page), NR_ISOLATED_ANON +
+ page_is_file_lru(page), -nr_pages);
putback_lru_page(page);
}
isolated = 0;
} else
- count_vm_numa_event(NUMA_PAGE_MIGRATE);
+ count_vm_numa_events(NUMA_PAGE_MIGRATE, nr_pages);
BUG_ON(!list_empty(&migratepages));
return isolated;
--
2.26.2
Add orig_pmd to struct vm_fault so the "orig_pmd" parameter used by huge page
fault could be removed, just like its PTE counterpart does.
Signed-off-by: Yang Shi <[email protected]>
---
include/linux/huge_mm.h | 9 ++++-----
include/linux/mm.h | 7 ++++++-
mm/huge_memory.c | 9 ++++++---
mm/memory.c | 26 +++++++++++++-------------
4 files changed, 29 insertions(+), 22 deletions(-)
diff --git a/include/linux/huge_mm.h b/include/linux/huge_mm.h
index 60dad7c88d72..77d6d454df90 100644
--- a/include/linux/huge_mm.h
+++ b/include/linux/huge_mm.h
@@ -11,7 +11,7 @@ vm_fault_t do_huge_pmd_anonymous_page(struct vm_fault *vmf);
int copy_huge_pmd(struct mm_struct *dst_mm, struct mm_struct *src_mm,
pmd_t *dst_pmd, pmd_t *src_pmd, unsigned long addr,
struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma);
-void huge_pmd_set_accessed(struct vm_fault *vmf, pmd_t orig_pmd);
+void huge_pmd_set_accessed(struct vm_fault *vmf);
int copy_huge_pud(struct mm_struct *dst_mm, struct mm_struct *src_mm,
pud_t *dst_pud, pud_t *src_pud, unsigned long addr,
struct vm_area_struct *vma);
@@ -24,7 +24,7 @@ static inline void huge_pud_set_accessed(struct vm_fault *vmf, pud_t orig_pud)
}
#endif
-vm_fault_t do_huge_pmd_wp_page(struct vm_fault *vmf, pmd_t orig_pmd);
+vm_fault_t do_huge_pmd_wp_page(struct vm_fault *vmf);
struct page *follow_trans_huge_pmd(struct vm_area_struct *vma,
unsigned long addr, pmd_t *pmd,
unsigned int flags);
@@ -283,7 +283,7 @@ struct page *follow_devmap_pmd(struct vm_area_struct *vma, unsigned long addr,
struct page *follow_devmap_pud(struct vm_area_struct *vma, unsigned long addr,
pud_t *pud, int flags, struct dev_pagemap **pgmap);
-vm_fault_t do_huge_pmd_numa_page(struct vm_fault *vmf, pmd_t orig_pmd);
+vm_fault_t do_huge_pmd_numa_page(struct vm_fault *vmf);
extern struct page *huge_zero_page;
@@ -429,8 +429,7 @@ static inline spinlock_t *pud_trans_huge_lock(pud_t *pud,
return NULL;
}
-static inline vm_fault_t do_huge_pmd_numa_page(struct vm_fault *vmf,
- pmd_t orig_pmd)
+static inline vm_fault_t do_huge_pmd_numa_page(struct vm_fault *vmf)
{
return 0;
}
diff --git a/include/linux/mm.h b/include/linux/mm.h
index 6ce7bac74530..b3a4b405364b 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -550,7 +550,12 @@ struct vm_fault {
pud_t *pud; /* Pointer to pud entry matching
* the 'address'
*/
- pte_t orig_pte; /* Value of PTE at the time of fault */
+ union {
+ pte_t orig_pte; /* Value of PTE at the time of fault */
+ pmd_t orig_pmd; /* Value of PMD at the time of fault,
+ * used by PMD fault only.
+ */
+ };
struct page *cow_page; /* Page handler may use for COW fault */
struct page *page; /* ->fault handlers should return a
diff --git a/mm/huge_memory.c b/mm/huge_memory.c
index 19195fca1aee..692788988530 100644
--- a/mm/huge_memory.c
+++ b/mm/huge_memory.c
@@ -1245,11 +1245,12 @@ void huge_pud_set_accessed(struct vm_fault *vmf, pud_t orig_pud)
}
#endif /* CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD */
-void huge_pmd_set_accessed(struct vm_fault *vmf, pmd_t orig_pmd)
+void huge_pmd_set_accessed(struct vm_fault *vmf)
{
pmd_t entry;
unsigned long haddr;
bool write = vmf->flags & FAULT_FLAG_WRITE;
+ pmd_t orig_pmd = vmf->orig_pmd;
vmf->ptl = pmd_lock(vmf->vma->vm_mm, vmf->pmd);
if (unlikely(!pmd_same(*vmf->pmd, orig_pmd)))
@@ -1266,11 +1267,12 @@ void huge_pmd_set_accessed(struct vm_fault *vmf, pmd_t orig_pmd)
spin_unlock(vmf->ptl);
}
-vm_fault_t do_huge_pmd_wp_page(struct vm_fault *vmf, pmd_t orig_pmd)
+vm_fault_t do_huge_pmd_wp_page(struct vm_fault *vmf)
{
struct vm_area_struct *vma = vmf->vma;
struct page *page;
unsigned long haddr = vmf->address & HPAGE_PMD_MASK;
+ pmd_t orig_pmd = vmf->orig_pmd;
vmf->ptl = pmd_lockptr(vma->vm_mm, vmf->pmd);
VM_BUG_ON_VMA(!vma->anon_vma, vma);
@@ -1406,9 +1408,10 @@ struct page *follow_trans_huge_pmd(struct vm_area_struct *vma,
}
/* NUMA hinting page fault entry point for trans huge pmds */
-vm_fault_t do_huge_pmd_numa_page(struct vm_fault *vmf, pmd_t pmd)
+vm_fault_t do_huge_pmd_numa_page(struct vm_fault *vmf)
{
struct vm_area_struct *vma = vmf->vma;
+ pmd_t pmd = vmf->orig_pmd;
struct anon_vma *anon_vma = NULL;
struct page *page;
unsigned long haddr = vmf->address & HPAGE_PMD_MASK;
diff --git a/mm/memory.c b/mm/memory.c
index 97ca3e51cf85..f9df3ab12813 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -4252,12 +4252,12 @@ static inline vm_fault_t create_huge_pmd(struct vm_fault *vmf)
}
/* `inline' is required to avoid gcc 4.1.2 build error */
-static inline vm_fault_t wp_huge_pmd(struct vm_fault *vmf, pmd_t orig_pmd)
+static inline vm_fault_t wp_huge_pmd(struct vm_fault *vmf)
{
if (vma_is_anonymous(vmf->vma)) {
- if (userfaultfd_huge_pmd_wp(vmf->vma, orig_pmd))
+ if (userfaultfd_huge_pmd_wp(vmf->vma, vmf->orig_pmd))
return handle_userfault(vmf, VM_UFFD_WP);
- return do_huge_pmd_wp_page(vmf, orig_pmd);
+ return do_huge_pmd_wp_page(vmf);
}
if (vmf->vma->vm_ops->huge_fault) {
vm_fault_t ret = vmf->vma->vm_ops->huge_fault(vmf, PE_SIZE_PMD);
@@ -4484,26 +4484,26 @@ static vm_fault_t __handle_mm_fault(struct vm_area_struct *vma,
if (!(ret & VM_FAULT_FALLBACK))
return ret;
} else {
- pmd_t orig_pmd = *vmf.pmd;
+ vmf.orig_pmd = *vmf.pmd;
barrier();
- if (unlikely(is_swap_pmd(orig_pmd))) {
+ if (unlikely(is_swap_pmd(vmf.orig_pmd))) {
VM_BUG_ON(thp_migration_supported() &&
- !is_pmd_migration_entry(orig_pmd));
- if (is_pmd_migration_entry(orig_pmd))
+ !is_pmd_migration_entry(vmf.orig_pmd));
+ if (is_pmd_migration_entry(vmf.orig_pmd))
pmd_migration_entry_wait(mm, vmf.pmd);
return 0;
}
- if (pmd_trans_huge(orig_pmd) || pmd_devmap(orig_pmd)) {
- if (pmd_protnone(orig_pmd) && vma_is_accessible(vma))
- return do_huge_pmd_numa_page(&vmf, orig_pmd);
+ if (pmd_trans_huge(vmf.orig_pmd) || pmd_devmap(vmf.orig_pmd)) {
+ if (pmd_protnone(vmf.orig_pmd) && vma_is_accessible(vma))
+ return do_huge_pmd_numa_page(&vmf);
- if (dirty && !pmd_write(orig_pmd)) {
- ret = wp_huge_pmd(&vmf, orig_pmd);
+ if (dirty && !pmd_write(vmf.orig_pmd)) {
+ ret = wp_huge_pmd(&vmf);
if (!(ret & VM_FAULT_FALLBACK))
return ret;
} else {
- huge_pmd_set_accessed(&vmf, orig_pmd);
+ huge_pmd_set_accessed(&vmf);
return 0;
}
}
--
2.26.2