mm: split underused THPs
Subsystems (12)
v4 cover mm: split underused THPs
The current upstream default policy for THP is always. However, Meta uses madvise in production as the current THP=always policy vastly overprovisions THPs in sparsely accessed memory areas, resulting in excessive memory pressure and premature OOM killing.
Using madvise + relying on khugepaged has certain drawbacks over THP=always. Using madvise hints mean THPs aren't "transparent" and require userspace changes. Waiting for khugepaged to scan memory and collapse pages into THP can be slow and unpredictable in terms of performance (i.e. you dont know when the collapse will happen), while production environments require predictable performance. If there is enough memory available, its better for both performance and predictability to have a THP from fault time, i.e. THP=always rather than wait for khugepaged to collapse it, and deal with sparsely populated THPs when the system is running out of memory.
This patch-series is an attempt to mitigate the issue of running out of memory when THP is always enabled. During runtime whenever a THP is being faulted in or collapsed by khugepaged, the THP is added to a list. Whenever memory reclaim happens, the kernel runs the deferred_split shrinker which goes through the list and checks if the THP was underused, i.e. how many of the base 4K pages of the entire THP were zero-filled. If this number goes above a certain threshold, the shrinker will attempt to split that THP. Then at remap time, the pages that were zero-filled are mapped to the shared zeropage, hence saving memory. This method avoids the downside of wasting memory in areas where THP is sparsely filled when THP is always enabled, while still providing the upside THPs like reduced TLB misses without having to use madvise.
Meta production workloads that were CPU bound (>99% CPU utilzation) were tested with THP shrinker. The results after 2 hours are as follows:
| THP=madvise | THP=always | THP=always
| | | + shrinker series
| | | + max_ptes_none=409
-----------------------------------------------------------------------------
Performance improvement | - | +1.8% | +1.7%
(over THP=madvise) | | |-----------------------------------------------------------------------------
Memory usage | 54.6G | 58.8G (+7.7%) | 55.9G (+2.4%)
-----------------------------------------------------------------------------
max_ptes_none=409 means that any THP that has more than 409 out of 512 (80%) zero filled filled pages will be split.
To test out the patches, the below commands without the shrinker will invoke OOM killer immediately and kill stress, but will not fail with the shrinker:
echo 450 > /sys/kernel/mm/transparent_hugepage/khugepaged/max_ptes_none mkdir /sys/fs/cgroup/test
echo $$ > /sys/fs/cgroup/test/cgroup.procs
echo 20M > /sys/fs/cgroup/test/memory.max
echo 0 > /sys/fs/cgroup/test/memory.swap.max
# allocate twice memory.max for each stress worker and touch 40/512 of # each THP, i.e. vm-stride 50K. # With the shrinker, max_ptes_none of 470 and below won't invoke OOM # killer. # Without the shrinker, OOM killer is invoked immediately irrespective # of max_ptes_none value and kills stress. stress --vm 1 --vm-bytes 40M --vm-stride 50K
v3 -> v4:
- do not clear partially_mapped flag on hugeTLB folios (Yu Zhao).
- fix condition for calling deferred_folio_split in partially mapped case and count for partially mapped vm events (Barry Song).
- use non-atomic versions of set/clear partially_mapped flags (David Hildenbrand)
- use PG_partially_mapped = PG_reclaim (Matthew Wilcox)
- delete folio from lru list and folio_batch_add "new_folio" instead of folio in __split_huge_page. (Kairui Song)
- fix deadlock in deferred_split_scan by not doing folio_put while holding split_queue_lock (Hugh Dickins)
- underutilized to underused and thp_low_util_shrinker to shrink_underused (Hugh Dickins)
v2 -> v3:
- Use my_zero_pfn instead of page_to_pfn(ZERO_PAGE(..)) (Johannes)
- Use flags argument instead of bools in remove_migration_ptes (Johannes)
- Use a new flag in folio->_flags_1 instead of folio->_partially_mapped (David Hildenbrand).
- Split out the last patch of v2 into 3, one for introducing the flag, one for splitting underutilized THPs on _deferred_list and one for adding sysfs entry to disable splitting (David Hildenbrand).
v1 -> v2:
- Turn page checks and operations to folio versions in __split_huge_page. This means patches 1 and 2 from v1 are no longer needed. (David Hildenbrand)
- Map to shared zeropage in all cases if the base page is zero-filled. The uffd selftest was removed.
(David Hildenbrand).
- rename 'dirty' to 'contains_data' in try_to_map_unused_to_zeropage (Rik van Riel).
- Use unsigned long instead of uint64_t (kernel test robot).
Alexander Zhu (1):
mm: selftest to verify zero-filled pages are mapped to zeropage
Usama Arif (3):
mm: Introduce a pageflag for partially mapped folios
mm: split underused THPs
mm: add sysfs entry to disable splitting underused THPs
Yu Zhao (2):
mm: free zapped tail pages when splitting isolated thp
mm: remap unused subpages to shared zeropage when splitting isolated
thp
Documentation/admin-guide/mm/transhuge.rst | 16 ++ include/linux/huge_mm.h | 4 +- include/linux/khugepaged.h | 1 + include/linux/page-flags.h | 11 ++ include/linux/rmap.h | 7 +- include/linux/vm_event_item.h | 1 + mm/huge_memory.c | 145 ++++++++++++++++-- mm/internal.h | 4 +- mm/khugepaged.c | 3 +- mm/memcontrol.c | 3 +- mm/migrate.c | 75 +++++++-- mm/migrate_device.c | 4 +- mm/page_alloc.c | 5 +- mm/rmap.c | 5 +- mm/vmscan.c | 3 +- mm/vmstat.c | 1 + .../selftests/mm/split_huge_page_test.c | 71 +++++++++ tools/testing/selftests/mm/vm_util.c | 22 +++ tools/testing/selftests/mm/vm_util.h | 1 + 19 files changed, 346 insertions(+), 36 deletions(-)
Signature · 1 line
2.43.5
PATCH v4 1/6 mm: free zapped tail pages when splitting isolated thp
From: Yu Zhao <[email protected]>
If a tail page has only two references left, one inherited from the isolation of its head and the other from lru_add_page_tail() which we are about to drop, it means this tail page was concurrently zapped. Then we can safely free it and save page reclaim or migration the trouble of trying it.
--- mm/huge_memory.c | 28 ++++++++++++++++++++++++++++ 1 file changed, 28 insertions(+)
mm/huge_memory.c +28 -0
@@ -3059,7 +3059,9 @@ static void __split_huge_page(struct page *page, struct list_head *list,unsigned int new_nr =1 << new_order;int order =folio_order (folio);unsigned int nr =1 << order; +struct folio_batch free_folios; +folio_batch_init (&free_folios);/* complete memcg works before add pages to LRU */ split_page_memcg (head, order, new_order);
@@ -3143,6 +3145,27 @@ static void __split_huge_page(struct page *page, struct list_head *list,if (subpage == page)continue ;folio_unlock (new_folio); +/* +* If a folio has only two references left, one inherited +* from the isolation of its head and the other from +* lru_add_page_tail() which we are about to drop, it means this +* folio was concurrently zapped. Then we can safely free it +* and save page reclaim or migration the trouble of trying it. +*/ +if (list &&folio_ref_freeze (new_folio,2 )) { +VM_WARN_ON_ONCE_FOLIO (folio_test_lru (new_folio), new_folio); +VM_WARN_ON_ONCE_FOLIO (folio_test_large (new_folio), new_folio); +VM_WARN_ON_ONCE_FOLIO (folio_mapped (new_folio), new_folio); + +folio_clear_active (new_folio); +folio_clear_unevictable (new_folio); +list_del (&new_folio->lru); +if (!folio_batch_add (&free_folios, new_folio)) { +mem_cgroup_uncharge_folios (&free_folios); +free_unref_folios (&free_folios); + } +continue ; + }/* * Subpages may be freed if there wasn't any mapping
@@ -3153,6 +3176,11 @@ static void __split_huge_page(struct page *page, struct list_head *list,*/ free_page_and_swap_cache (subpage); } + +if (free_folios.nr) { +mem_cgroup_uncharge_folios (&free_folios); +free_unref_folios (&free_folios); + } }/* Racy check whether the huge page can be split */
Signature · 1 line
2.43.5
PATCH v4 2/6 mm: remap unused subpages to shared zeropage when splitting isolated thp
In reply to Usama ArifFrom: Yu Zhao <[email protected]>
Here being unused means containing only zeros and inaccessible to userspace. When splitting an isolated thp under reclaim or migration, the unused subpages can be mapped to the shared zeropage, hence saving memory. This is particularly helpful when the internal fragmentation of a thp is high, i.e. it has many untouched subpages.
This is also a prerequisite for THP low utilization shrinker which will be introduced in later patches, where underutilized THPs are split, and the zero-filled pages are freed saving memory.
--- include/linux/rmap.h | 7 ++++- mm/huge_memory.c | 8 ++--- mm/migrate.c | 72 ++++++++++++++++++++++++++++++++++++++------ mm/migrate_device.c | 4 +-- 4 files changed, 75 insertions(+), 16 deletions(-)
include/linux/rmap.h +6 -1
@@ -745,7 +745,12 @@ int folio_mkclean(struct folio *);int pfn_mkclean_range (unsigned long pfn,unsigned long nr_pages,pgoff_t pgoff,struct vm_area_struct *vma); -void remove_migration_ptes (struct folio *src,struct folio *dst,bool locked); +enum rmp_flags { +RMP_LOCKED =1 <<0 , +RMP_USE_SHARED_ZEROPAGE =1 <<1 , +}; + +void remove_migration_ptes (struct folio *src,struct folio *dst,int flags);/* * rmap_walk_control: To control rmap traversing for specific needs
mm/huge_memory.c +4 -4
@@ -2911,7 +2911,7 @@ bool unmap_huge_pmd_locked(struct vm_area_struct *vma, unsigned long addr,return false; } -static void remap_page (struct folio *folio,unsigned long nr) +static void remap_page (struct folio *folio,unsigned long nr,int flags) {int i =0 ;
@@ -2919,7 +2919,7 @@ static void remap_page(struct folio *folio, unsigned long nr)if (!folio_test_anon (folio))return ;for (;;) { -remove_migration_ptes (folio, folio, true); +remove_migration_ptes (folio, folio,RMP_LOCKED | flags); i +=folio_nr_pages (folio);if (i >= nr)break ;
@@ -3129,7 +3129,7 @@ static void __split_huge_page(struct page *page, struct list_head *list,if (nr_dropped)shmem_uncharge (folio->mapping->host, nr_dropped); -remap_page (folio, nr); +remap_page (folio, nr,PageAnon (head) ?RMP_USE_SHARED_ZEROPAGE :0 );/* * set page to its compound_head when split to non order-0 pages, so
@@ -3425,7 +3425,7 @@ int split_huge_page_to_list_to_order(struct page *page, struct list_head *list,if (mapping)xas_unlock (&xas);local_irq_enable (); -remap_page (folio,folio_nr_pages (folio)); +remap_page (folio,folio_nr_pages (folio),0 ); ret = -EAGAIN ; }
mm/migrate.c +63 -9
124 lines not on this page · raw message
mm/migrate_device.c +2 -2
@@ -424,7 +424,7 @@ static unsigned long migrate_device_unmap(unsigned long *src_pfns,continue ; folio =page_folio (page); -remove_migration_ptes (folio, folio, false); +remove_migration_ptes (folio, folio,0 ); src_pfns[i] =0 ;folio_unlock (folio);
@@ -837,7 +837,7 @@ void migrate_device_finalize(unsigned long *src_pfns, src =page_folio (page); dst =page_folio (newpage); -remove_migration_ptes (src, dst, false); +remove_migration_ptes (src, dst,0 );folio_unlock (src);if (is_zone_device_page (page))
Signature · 1 line
2.43.5
PATCH v4 3/6 mm: selftest to verify zero-filled pages are mapped to zeropage
In reply to Usama ArifFrom: Alexander Zhu <[email protected]>
When a THP is split, any subpage that is zero-filled will be mapped to the shared zeropage, hence saving memory. Add selftest to verify this by allocating zero-filled THP and comparing RssAnon before and after split.
--- .../selftests/mm/split_huge_page_test.c | 71 +++++++++++++++++++ tools/testing/selftests/mm/vm_util.c | 22 ++++++ tools/testing/selftests/mm/vm_util.h | 1 + 3 files changed, 94 insertions(+)
tools/testing/selftests/mm/split_huge_page_test.c +71 -0
@@ -84,6 +84,76 @@ static void write_debugfs(const char *fmt, ...)write_file (SPLIT_DEBUGFS , input, ret +1 ); } +static char *allocate_zero_filled_hugepage (size_t len) +{ +char *result; +size_t i; + + result =memalign (pmd_pagesize, len); +if (!result) { +printf ("Fail to allocate memory\n" ); +exit (EXIT_FAILURE ); + } + +madvise (result, len,MADV_HUGEPAGE ); + +for (i =0 ; i < len; i++) + result[i] = (char )0 ; + +return result; +} + +static void verify_rss_anon_split_huge_page_all_zeroes (char *one_page,int nr_hpages,size_t len) +{ +unsigned long rss_anon_before, rss_anon_after; +size_t i; + +if (!check_huge_anon (one_page,4 , pmd_pagesize)) { +printf ("No THP is allocated\n" ); +exit (EXIT_FAILURE ); + } + + rss_anon_before =rss_anon (); +if (!rss_anon_before) { +printf ("No RssAnon is allocated before split\n" ); +exit (EXIT_FAILURE ); + } + +/* split all THPs */ +write_debugfs (PID_FMT ,getpid (), (uint64_t )one_page, + (uint64_t )one_page + len,0 ); + +for (i =0 ; i < len; i++) +if (one_page[i] != (char )0 ) { +printf ("%ld byte corrupted\n" , i); +exit (EXIT_FAILURE ); + } + +if (!check_huge_anon (one_page,0 , pmd_pagesize)) { +printf ("Still AnonHugePages not split\n" ); +exit (EXIT_FAILURE ); + } + + rss_anon_after =rss_anon (); +if (rss_anon_after >= rss_anon_before) { +printf ("Incorrect RssAnon value. Before: %ld After: %ld\n" , + rss_anon_before, rss_anon_after); +exit (EXIT_FAILURE ); + } +} + +void split_pmd_zero_pages (void ) +{ +char *one_page; +int nr_hpages =4 ; +size_t len = nr_hpages * pmd_pagesize; + + one_page =allocate_zero_filled_hugepage (len); +verify_rss_anon_split_huge_page_all_zeroes (one_page, nr_hpages, len); +printf ("Split zero filled huge pages successful\n" ); +free (one_page); +} +void split_pmd_thp (void ) {char *one_page;
@@ -431,6 +501,7 @@ int main(int argc, char **argv) fd_size =2 * pmd_pagesize; +split_pmd_zero_pages ();split_pmd_thp ();split_pte_mapped_thp ();split_file_backed_thp ();
tools/testing/selftests/mm/vm_util.c +22 -0
@@ -12,6 +12,7 @@#define PMD_SIZE_FILE_PATH "/sys/kernel/mm/transparent_hugepage/hpage_pmd_size"#define SMAP_FILE_PATH "/proc/self/smaps" +#define STATUS_FILE_PATH "/proc/self/status"#define MAX_LINE_LENGTH 500unsigned int __page_size;
@@ -171,6 +172,27 @@ uint64_t read_pmd_pagesize(void)return strtoul (buf,NULL ,10 ); } +unsigned long rss_anon (void ) +{ +unsigned long rss_anon =0 ; +FILE *fp; +char buffer[MAX_LINE_LENGTH ]; + + fp =fopen (STATUS_FILE_PATH ,"r" ); +if (!fp) +ksft_exit_fail_msg ("%s: Failed to open file %s\n" , __func__,STATUS_FILE_PATH ); + +if (!check_for_pattern (fp,"RssAnon:" , buffer,sizeof (buffer))) + goto err_out; + +if (sscanf (buffer,"RssAnon:%10lu kB" , &rss_anon) !=1 ) +ksft_exit_fail_msg ("Reading status error\n" ); + +err_out: +fclose (fp); +return rss_anon; +} +bool __check_huge (void *addr,char *pattern,int nr_hpages,uint64_t hpage_size) {
tools/testing/selftests/mm/vm_util.h +1 -0
@@ -39,6 +39,7 @@ unsigned long pagemap_get_pfn(int fd, char *start);void clear_softdirty (void );bool check_for_pattern (FILE *fp,const char *pattern,char *buf,size_t len);uint64_t read_pmd_pagesize (void ); +uint64_t rss_anon (void );bool check_huge_anon (void *addr,int nr_hpages,uint64_t hpage_size);bool check_huge_file (void *addr,int nr_hpages,uint64_t hpage_size);bool check_huge_shmem (void *addr,int nr_hpages,uint64_t hpage_size);
Signature · 1 line
2.43.5
Re: [PATCH v4 3/6] mm: selftest to verify zero-filled pages are mapped to zeropage
Quoted 150 lines · Usama Arif
From: Alexander Zhu <[email protected]>
When a THP is split, any subpage that is zero-filled will be mapped to the shared zeropage, hence saving memory. Add selftest to verify this by allocating zero-filled THP and comparing RssAnon before and after split.
Signed-off-by Alexander Zhu <[email protected]>Acked-by Rik van Riel <[email protected]>Signed-off-by Usama Arif <[email protected]>--- .../selftests/mm/split_huge_page_test.c | 71 +++++++++++++++++++ tools/testing/selftests/mm/vm_util.c | 22 ++++++ tools/testing/selftests/mm/vm_util.h | 1 + 3 files changed, 94 insertions(+)tools/testing/selftests/mm/split_huge_page_test.c +71
-0@@ -84,6 +84,76 @@ static void write_debugfs(const char *fmt, ...)write_file (SPLIT_DEBUGFS , input, ret +1 ); } +static char *allocate_zero_filled_hugepage (size_t len) +{ +char *result; +size_t i; + + result =memalign (pmd_pagesize, len); +if (!result) { +printf ("Fail to allocate memory\n" ); +exit (EXIT_FAILURE ); + } + +madvise (result, len,MADV_HUGEPAGE ); + +for (i =0 ; i < len; i++) + result[i] = (char )0 ; + +return result; +} + +static void verify_rss_anon_split_huge_page_all_zeroes (char *one_page,int nr_hpages,size_t len) +{ +unsigned long rss_anon_before, rss_anon_after; +size_t i; + +if (!check_huge_anon (one_page,4 , pmd_pagesize)) { +printf ("No THP is allocated\n" ); +exit (EXIT_FAILURE ); + } + + rss_anon_before =rss_anon (); +if (!rss_anon_before) { +printf ("No RssAnon is allocated before split\n" ); +exit (EXIT_FAILURE ); + } + +/* split all THPs */ +write_debugfs (PID_FMT ,getpid (), (uint64_t )one_page, + (uint64_t )one_page + len,0 ); + +for (i =0 ; i < len; i++) +if (one_page[i] != (char )0 ) { +printf ("%ld byte corrupted\n" , i); +exit (EXIT_FAILURE ); + } + +if (!check_huge_anon (one_page,0 , pmd_pagesize)) { +printf ("Still AnonHugePages not split\n" ); +exit (EXIT_FAILURE ); + } + + rss_anon_after =rss_anon (); +if (rss_anon_after >= rss_anon_before) { +printf ("Incorrect RssAnon value. Before: %ld After: %ld\n" , + rss_anon_before, rss_anon_after); +exit (EXIT_FAILURE ); + } +} + +void split_pmd_zero_pages (void ) +{ +char *one_page; +int nr_hpages =4 ; +size_t len = nr_hpages * pmd_pagesize; + + one_page =allocate_zero_filled_hugepage (len); +verify_rss_anon_split_huge_page_all_zeroes (one_page, nr_hpages, len); +printf ("Split zero filled huge pages successful\n" ); +free (one_page); +} +void split_pmd_thp (void ) {char *one_page;@@ -431,6 +501,7 @@ int main(int argc, char **argv) fd_size =2 * pmd_pagesize; +split_pmd_zero_pages ();split_pmd_thp ();split_pte_mapped_thp ();split_file_backed_thp ();tools/testing/selftests/mm/vm_util.c +22
-0@@ -12,6 +12,7 @@#define PMD_SIZE_FILE_PATH "/sys/kernel/mm/transparent_hugepage/hpage_pmd_size"#define SMAP_FILE_PATH "/proc/self/smaps" +#define STATUS_FILE_PATH "/proc/self/status"#define MAX_LINE_LENGTH 500unsigned int __page_size;@@ -171,6 +172,27 @@ uint64_t read_pmd_pagesize(void)return strtoul (buf,NULL ,10 ); } +unsigned long rss_anon (void ) +{ +unsigned long rss_anon =0 ; +FILE *fp; +char buffer[MAX_LINE_LENGTH ]; + + fp =fopen (STATUS_FILE_PATH ,"r" ); +if (!fp) +ksft_exit_fail_msg ("%s: Failed to open file %s\n" , __func__,STATUS_FILE_PATH ); + +if (!check_for_pattern (fp,"RssAnon:" , buffer,sizeof (buffer))) + goto err_out; + +if (sscanf (buffer,"RssAnon:%10lu kB" , &rss_anon) !=1 ) +ksft_exit_fail_msg ("Reading status error\n" ); + +err_out: +fclose (fp); +return rss_anon; +} +bool __check_huge (void *addr,char *pattern,int nr_hpages,uint64_t hpage_size) {tools/testing/selftests/mm/vm_util.h +0
-0
tools/testing/selftests/mm/vm_util.h unsigned long pagemap_get_pfn(int fd, char *start); · line 45
Quoted 7 lines
tools/testing/selftests/mm/vm_util.h +1
-0@@ -39,6 +39,7 @@ unsigned long pagemap_get_pfn(int fd, char *start);void clear_softdirty (void );bool check_for_pattern (FILE *fp,const char *pattern,char *buf,size_t len);uint64_t read_pmd_pagesize (void ); +uint64_t rss_anon (void );bool check_huge_anon (void *addr,int nr_hpages,uint64_t hpage_size);bool check_huge_file (void *addr,int nr_hpages,uint64_t hpage_size);bool check_huge_shmem (void *addr,int nr_hpages,uint64_t hpage_size);
Need below fixlet as well
From 2476b7a46908c801ab446f41566f4af52b939ac7 Mon Sep 17 00:00:00 2001
From: Usama Arif <[email protected]>
Date: Wed, 21 Aug 2024 20:06:53 +0100
Subject: [PATCH] mm: selftest to verify zero-filled pages are mapped to
zeropage fix
change uint64_t to unsigned long for rss_anon
--- tools/testing/selftests/mm/vm_util.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
tools/testing/selftests/mm/vm_util.h +1 -1
@@ -39,7 +39,7 @@ unsigned long pagemap_get_pfn(int fd, char *start);void clear_softdirty (void );bool check_for_pattern (FILE *fp,const char *pattern,char *buf,size_t len);uint64_t read_pmd_pagesize (void ); -uint64_t rss_anon (void ); +unsigned long rss_anon (void );bool check_huge_anon (void *addr,int nr_hpages,uint64_t hpage_size);bool check_huge_file (void *addr,int nr_hpages,uint64_t hpage_size);bool check_huge_shmem (void *addr,int nr_hpages,uint64_t hpage_size);
Signature · 1 line
2.43.5
kernel test robot
build botRe: [PATCH v4 3/6] mm: selftest to verify zero-filled pages are mapped to zeropage
In reply to Usama Arifkernel test robot noticed the following build errors:
Hi Usama,
kernel test robot noticed the following build errors:
[auto build test ERROR on akpm-mm/mm-everything]
url: https://github.com/intel-lab-lkp/linux/commits/Usama-Arif/mm-free-zapped-tail-pages-when-splitting-isolated-thp/20240819-103336 base: https://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm.git mm-everything patch link: https://lore.kernel.org/r/20240819023145.2415299-4-usamaarif642%40gmail.com patch subject: [PATCH v4 3/6] mm: selftest to verify zero-filled pages are mapped to zeropage
:::::: branch date: 2 days ago
:::::: commit date: 2 days ago
compiler: gcc-12 (Debian 12.2.0-14) 12.2.0
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20240821/[email protected]/reproduce)
If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <[email protected]>
| Closes: https://lore.kernel.org/r/[email protected]/
All errors (new ones prefixed by >>):
>> vm_util.c:175:15: error: conflicting types for 'rss_anon'; have 'long unsigned int(void)'
175 | unsigned long rss_anon(void)
| ^~~~~~~~
In file included from vm_util.c:11:
vm_util.h:42:10: note: previous declaration of 'rss_anon' with type 'uint64_t(void)' {aka 'long long unsigned int(void)'}
42 | uint64_t rss_anon(void);
| ^~~~~~~~vim +175 tools/testing/selftests/mm/vm_util.c
642bc52aed9c99e tools/testing/selftests/vm/vm_util.c Muhammad Usama Anjum 2022-04-28 174
10bf5ecc7e9d437 tools/testing/selftests/mm/vm_util.c Alexander Zhu 2024-08-19 @175 unsigned long rss_anon(void)
10bf5ecc7e9d437 tools/testing/selftests/mm/vm_util.c Alexander Zhu 2024-08-19 176 {
10bf5ecc7e9d437 tools/testing/selftests/mm/vm_util.c Alexander Zhu 2024-08-19 177 unsigned long rss_anon = 0;
10bf5ecc7e9d437 tools/testing/selftests/mm/vm_util.c Alexander Zhu 2024-08-19 178 FILE *fp;
10bf5ecc7e9d437 tools/testing/selftests/mm/vm_util.c Alexander Zhu 2024-08-19 179 char buffer[MAX_LINE_LENGTH];
10bf5ecc7e9d437 tools/testing/selftests/mm/vm_util.c Alexander Zhu 2024-08-19 180
10bf5ecc7e9d437 tools/testing/selftests/mm/vm_util.c Alexander Zhu 2024-08-19 181 fp = fopen(STATUS_FILE_PATH, "r");
10bf5ecc7e9d437 tools/testing/selftests/mm/vm_util.c Alexander Zhu 2024-08-19 182 if (!fp)
10bf5ecc7e9d437 tools/testing/selftests/mm/vm_util.c Alexander Zhu 2024-08-19 183 ksft_exit_fail_msg("%s: Failed to open file %s\n", __func__, STATUS_FILE_PATH);
10bf5ecc7e9d437 tools/testing/selftests/mm/vm_util.c Alexander Zhu 2024-08-19 184
10bf5ecc7e9d437 tools/testing/selftests/mm/vm_util.c Alexander Zhu 2024-08-19 185 if (!check_for_pattern(fp, "RssAnon:", buffer, sizeof(buffer)))
10bf5ecc7e9d437 tools/testing/selftests/mm/vm_util.c Alexander Zhu 2024-08-19 186 goto err_out;
10bf5ecc7e9d437 tools/testing/selftests/mm/vm_util.c Alexander Zhu 2024-08-19 187
10bf5ecc7e9d437 tools/testing/selftests/mm/vm_util.c Alexander Zhu 2024-08-19 188 if (sscanf(buffer, "RssAnon:%10lu kB", &rss_anon) != 1)
10bf5ecc7e9d437 tools/testing/selftests/mm/vm_util.c Alexander Zhu 2024-08-19 189 ksft_exit_fail_msg("Reading status error\n");
10bf5ecc7e9d437 tools/testing/selftests/mm/vm_util.c Alexander Zhu 2024-08-19 190
10bf5ecc7e9d437 tools/testing/selftests/mm/vm_util.c Alexander Zhu 2024-08-19 191 err_out:
10bf5ecc7e9d437 tools/testing/selftests/mm/vm_util.c Alexander Zhu 2024-08-19 192 fclose(fp);
10bf5ecc7e9d437 tools/testing/selftests/mm/vm_util.c Alexander Zhu 2024-08-19 193 return rss_anon;
10bf5ecc7e9d437 tools/testing/selftests/mm/vm_util.c Alexander Zhu 2024-08-19 194 }
10bf5ecc7e9d437 tools/testing/selftests/mm/vm_util.c Alexander Zhu 2024-08-19 195 Signature · 2 lines
0-DAY CI Kernel Test Service https://github.com/intel/lkp-tests/wiki
PATCH v4 4/6 mm: Introduce a pageflag for partially mapped folios
In reply to Usama ArifCurrently folio->_deferred_list is used to keep track of partially_mapped folios that are going to be split under memory pressure. In the next patch, all THPs that are faulted in and collapsed by khugepaged are also going to be tracked using _deferred_list.
This patch introduces a pageflag to be able to distinguish between partially mapped folios and others in the deferred_list at split time in deferred_split_scan. Its needed as __folio_remove_rmap decrements _mapcount, _large_mapcount and _entire_mapcount, hence it won't be possible to distinguish between partially mapped folios and others in deferred_split_scan.
Eventhough it introduces an extra flag to track if the folio is partially mapped, there is no functional change intended with this patch and the flag is not useful in this patch itself, it will become useful in the next patch when _deferred_list has non partially mapped folios.
--- include/linux/huge_mm.h | 4 ++-- include/linux/page-flags.h | 11 +++++++++++ mm/huge_memory.c | 23 ++++++++++++++++------- mm/internal.h | 4 +++- mm/memcontrol.c | 3 ++- mm/migrate.c | 3 ++- mm/page_alloc.c | 5 +++-- mm/rmap.c | 5 +++-- mm/vmscan.c | 3 ++- 9 files changed, 44 insertions(+), 17 deletions(-)
include/linux/huge_mm.h +2 -2
@@ -321,7 +321,7 @@ static inline int split_huge_page(struct page *page) {return split_huge_page_to_list_to_order (page,NULL ,0 ); } -void deferred_split_folio (struct folio *folio); +void deferred_split_folio (struct folio *folio,bool partially_mapped);void __split_huge_pmd (struct vm_area_struct *vma,pmd_t *pmd,unsigned long address,bool freeze,struct folio *folio);
@@ -495,7 +495,7 @@ static inline int split_huge_page(struct page *page) {return 0 ; } -static inline void deferred_split_folio (struct folio *folio) {} +static inline void deferred_split_folio (struct folio *folio,bool partially_mapped) {}#define split_huge_pmd (__vma, __pmd, __address) \ do { } while (0)
include/linux/page-flags.h +11 -0
@@ -182,6 +182,7 @@ enum pageflags {/* At least one page in this folio has the hwpoison flag set */ PG_has_hwpoisoned = PG_active, PG_large_rmappable = PG_workingset,/* anon or file-backed */ + PG_partially_mapped = PG_reclaim, /* was identified to be partially mapped */ };#define PAGEFLAGS_MASK ((1UL << NR_PAGEFLAGS) - 1)
@@ -861,8 +862,18 @@ static inline void ClearPageCompound(struct page *page)ClearPageHead (page); } FOLIO_FLAG(large_rmappable, FOLIO_SECOND_PAGE) +FOLIO_TEST_FLAG (partially_mapped,FOLIO_SECOND_PAGE ) +/* +* PG_partially_mapped is protected by deferred_split split_queue_lock, +* so its safe to use non-atomic set/clear. + */ +__FOLIO_SET_FLAG(partially_mapped, FOLIO_SECOND_PAGE) +__FOLIO_CLEAR_FLAG(partially_mapped, FOLIO_SECOND_PAGE)#else FOLIO_FLAG_FALSE(large_rmappable) +FOLIO_TEST_FLAG_FALSE(partially_mapped) +__FOLIO_SET_FLAG_NOOP(partially_mapped) +__FOLIO_CLEAR_FLAG_NOOP(partially_mapped)#endif #define PG_head_mask ((1UL << PG_head))
mm/huge_memory.c +16 -7
@@ -3398,6 +3398,7 @@ int split_huge_page_to_list_to_order(struct page *page, struct list_head *list,* page_deferred_list. */ list_del_init (&folio->_deferred_list); +__folio_clear_partially_mapped (folio); }spin_unlock (&ds_queue->split_queue_lock);if (mapping) {
@@ -3454,11 +3455,13 @@ void __folio_undo_large_rmappable(struct folio *folio)if (!list_empty (&folio->_deferred_list)) { ds_queue->split_queue_len--;list_del_init (&folio->_deferred_list); +__folio_clear_partially_mapped (folio); }spin_unlock_irqrestore (&ds_queue->split_queue_lock, flags); } -void deferred_split_folio (struct folio *folio) +/* partially_mapped=false won't clear PG_partially_mapped folio flag */ +void deferred_split_folio (struct folio *folio,bool partially_mapped) {struct deferred_split *ds_queue =get_deferred_split_queue (folio);#ifdef CONFIG_MEMCG
@@ -3486,14 +3489,19 @@ void deferred_split_folio(struct folio *folio)if (folio_test_swapcache (folio))return ; -if (!list_empty (&folio->_deferred_list)) -return ; -spin_lock_irqsave (&ds_queue->split_queue_lock, flags); +if (partially_mapped) { +if (!folio_test_partially_mapped (folio)) { +__folio_set_partially_mapped (folio); +if (folio_test_pmd_mappable (folio)) +count_vm_event (THP_DEFERRED_SPLIT_PAGE ); +count_mthp_stat (folio_order (folio),MTHP_STAT_SPLIT_DEFERRED ); + } + }else { +/* partially mapped folios cannot become non-partially mapped */ +VM_WARN_ON_FOLIO (folio_test_partially_mapped (folio), folio); + }if (list_empty (&folio->_deferred_list)) { -if (folio_test_pmd_mappable (folio)) -count_vm_event (THP_DEFERRED_SPLIT_PAGE ); -count_mthp_stat (folio_order (folio),MTHP_STAT_SPLIT_DEFERRED );list_add_tail (&folio->_deferred_list, &ds_queue->split_queue); ds_queue->split_queue_len++;#ifdef CONFIG_MEMCG
@@ -3542,6 +3550,7 @@ static unsigned long deferred_split_scan(struct shrinker *shrink, }else {/* We lost race with folio_put() */ list_del_init (&folio->_deferred_list); +__folio_clear_partially_mapped (folio); ds_queue->split_queue_len--; }if (!--sc->nr_to_scan)
mm/internal.h +3 -1
@@ -662,8 +662,10 @@ static inline void prep_compound_head(struct page *page, unsigned int order)atomic_set (&folio->_entire_mapcount,-1 );atomic_set (&folio->_nr_pages_mapped,0 );atomic_set (&folio->_pincount,0 ); -if (order >1 ) +if (order >1 ) {INIT_LIST_HEAD (&folio->_deferred_list); +__folio_clear_partially_mapped (folio); + } }static inline void prep_compound_tail (struct page *head,int tail_idx)
mm/memcontrol.c +2 -1
@@ -4669,7 +4669,8 @@ static void uncharge_folio(struct folio *folio, struct uncharge_gather *ug)VM_BUG_ON_FOLIO (folio_test_lru (folio), folio);VM_BUG_ON_FOLIO (folio_order (folio) >1 && !folio_test_hugetlb (folio) && - !list_empty (&folio->_deferred_list), folio); + !list_empty (&folio->_deferred_list) && +folio_test_partially_mapped (folio), folio);/* * Nobody should be changing or seriously looking at
mm/migrate.c +2 -1
@@ -1735,7 +1735,8 @@ static int migrate_pages_batch(struct list_head *from,* use _deferred_list. */ if (nr_pages >2 && - !list_empty (&folio->_deferred_list)) { + !list_empty (&folio->_deferred_list) && +folio_test_partially_mapped (folio)) {if (!try_split_folio (folio, split_folios, mode)) { nr_failed++; stats->nr_thp_failed += is_thp;
mm/page_alloc.c +3 -2
@@ -957,8 +957,9 @@ static int free_tail_page_prepare(struct page *head_page, struct page *page)break ;case 2 :/* the second tail page: deferred_list overlaps ->mapping */ -if (unlikely (!list_empty (&folio->_deferred_list))) { -bad_page (page,"on deferred list" ); +if (unlikely (!list_empty (&folio->_deferred_list) && +folio_test_partially_mapped (folio))) { +bad_page (page,"partially mapped folio on deferred list" ); goto out; }break ;
mm/rmap.c +3 -2
@@ -1578,8 +1578,9 @@ static __always_inline void __folio_remove_rmap(struct folio *folio,* Check partially_mapped first to ensure it is a large folio. */ if (partially_mapped &&folio_test_anon (folio) && -list_empty (&folio->_deferred_list)) -deferred_split_folio (folio); + !folio_test_partially_mapped (folio)) +deferred_split_folio (folio, true); +__folio_mod_stat (folio, -nr, -nr_pmdmapped);/*
mm/vmscan.c +2 -1
@@ -1233,7 +1233,8 @@ static unsigned int shrink_folio_list(struct list_head *folio_list,* Split partially mapped folios right away. * We can free the unmapped pages without IO. */ -if (data_race (!list_empty (&folio->_deferred_list)) && +if (data_race (!list_empty (&folio->_deferred_list) && +folio_test_partially_mapped (folio)) &&split_folio_to_list (folio, folio_list)) goto activate_locked; }
Signature · 1 line
2.43.5
Barry Song
reviewerRe: [PATCH v4 4/6] mm: Introduce a pageflag for partially mapped folios
Hi Usama,
I feel it is much better now! thanks!
Quoted 98 lines · Usama Arif
Currently
folio->_deferred_listis used to keep track ofpartially_mappedfolios that are going to be split under memory pressure. In the next patch, all THPs that are faulted in and collapsed by khugepaged are also going to be tracked using_deferred_list.This patch introduces a pageflag to be able to distinguish between partially mapped folios and others in the
deferred_listat split time indeferred_split_scan. Its needed as__folio_remove_rmapdecrements_mapcount,_large_mapcountand_entire_mapcount, hence it won't be possible to distinguish between partially mapped folios and others indeferred_split_scan.Eventhough it introduces an extra flag to track if the folio is partially mapped, there is no functional change intended with this patch and the flag is not useful in this patch itself, it will become useful in the next patch when
_deferred_listhas non partially mapped folios.Signed-off-by Usama Arif <[email protected]>--- include/linux/huge_mm.h | 4 ++-- include/linux/page-flags.h | 11 +++++++++++ mm/huge_memory.c | 23 ++++++++++++++++------- mm/internal.h | 4 +++- mm/memcontrol.c | 3 ++- mm/migrate.c | 3 ++- mm/page_alloc.c | 5 +++-- mm/rmap.c | 5 +++-- mm/vmscan.c | 3 ++- 9 files changed, 44 insertions(+), 17 deletions(-)include/linux/huge_mm.h +2
-2@@ -321,7 +321,7 @@ static inline int split_huge_page(struct page *page) {return split_huge_page_to_list_to_order (page,NULL ,0 ); } -void deferred_split_folio (struct folio *folio); +void deferred_split_folio (struct folio *folio,bool partially_mapped);void __split_huge_pmd (struct vm_area_struct *vma,pmd_t *pmd,unsigned long address,bool freeze,struct folio *folio);@@ -495,7 +495,7 @@ static inline int split_huge_page(struct page *page) {return 0 ; } -static inline void deferred_split_folio (struct folio *folio) {} +static inline void deferred_split_folio (struct folio *folio,bool partially_mapped) {}#define split_huge_pmd (__vma, __pmd, __address) \ do { } while (0)include/linux/page-flags.h +11
-0@@ -182,6 +182,7 @@ enum pageflags {/* At least one page in this folio has the hwpoison flag set */ PG_has_hwpoisoned = PG_active, PG_large_rmappable = PG_workingset,/* anon or file-backed */ + PG_partially_mapped = PG_reclaim, /* was identified to be partially mapped */ };#define PAGEFLAGS_MASK ((1UL << NR_PAGEFLAGS) - 1)@@ -861,8 +862,18 @@ static inline void ClearPageCompound(struct page *page)ClearPageHead (page); } FOLIO_FLAG(large_rmappable, FOLIO_SECOND_PAGE) +FOLIO_TEST_FLAG (partially_mapped,FOLIO_SECOND_PAGE ) +/* +* PG_partially_mapped is protected by deferred_split split_queue_lock, +* so its safe to use non-atomic set/clear. + */ +__FOLIO_SET_FLAG(partially_mapped, FOLIO_SECOND_PAGE) +__FOLIO_CLEAR_FLAG(partially_mapped, FOLIO_SECOND_PAGE)#else FOLIO_FLAG_FALSE(large_rmappable) +FOLIO_TEST_FLAG_FALSE(partially_mapped) +__FOLIO_SET_FLAG_NOOP(partially_mapped) +__FOLIO_CLEAR_FLAG_NOOP(partially_mapped)#endif #define PG_head_mask ((1UL << PG_head))mm/huge_memory.c +1
-0@@ -3398,6 +3398,7 @@ int split_huge_page_to_list_to_order(struct page *page, struct list_head *list,* page_deferred_list. */ list_del_init (&folio->_deferred_list); +__folio_clear_partially_mapped (folio); }spin_unlock (&ds_queue->split_queue_lock);if (mapping) {
mm/huge_memory.c void __folio_undo_large_rmappable(struct folio *folio) · line 3458
Quoted 4 lines
mm/huge_memory.c +1
-0@@ -3454,11 +3455,13 @@ void __folio_undo_large_rmappable(struct folio *folio)if (!list_empty (&folio->_deferred_list)) { ds_queue->split_queue_len--;list_del_init (&folio->_deferred_list); +__folio_clear_partially_mapped (folio);
is it possible to make things clearer by
if (folio_clear_partially_mapped)
__folio_clear_partially_mapped(folio);
While writing without conditions isn't necessarily wrong, adding a condition will improve the readability of the code and enhance the clarity of my mTHP counters series. also help decrease smp cache sync if we can avoid unnecessary writing?
Quoted 37 lines
mm/huge_memory.c +13
-7}spin_unlock_irqrestore (&ds_queue->split_queue_lock, flags); } -void deferred_split_folio (struct folio *folio) +/* partially_mapped=false won't clear PG_partially_mapped folio flag */ +void deferred_split_folio (struct folio *folio,bool partially_mapped) {struct deferred_split *ds_queue =get_deferred_split_queue (folio);#ifdef CONFIG_MEMCG @@ -3486,14 +3489,19 @@ void deferred_split_folio(struct folio *folio)if (folio_test_swapcache (folio))return ; -if (!list_empty (&folio->_deferred_list)) -return ; -spin_lock_irqsave (&ds_queue->split_queue_lock, flags); +if (partially_mapped) { +if (!folio_test_partially_mapped (folio)) { +__folio_set_partially_mapped (folio); +if (folio_test_pmd_mappable (folio)) +count_vm_event (THP_DEFERRED_SPLIT_PAGE ); +count_mthp_stat (folio_order (folio),MTHP_STAT_SPLIT_DEFERRED ); + } + }else { +/* partially mapped folios cannot become non-partially mapped */ +VM_WARN_ON_FOLIO (folio_test_partially_mapped (folio), folio); + }if (list_empty (&folio->_deferred_list)) { -if (folio_test_pmd_mappable (folio)) -count_vm_event (THP_DEFERRED_SPLIT_PAGE ); -count_mthp_stat (folio_order (folio),MTHP_STAT_SPLIT_DEFERRED );list_add_tail (&folio->_deferred_list, &ds_queue->split_queue); ds_queue->split_queue_len++;#ifdef CONFIG_MEMCG
mm/huge_memory.c static unsigned long deferred_split_scan(struct shrinker *shrink, · line 3553
Quoted 4 lines
mm/huge_memory.c +1
-0@@ -3542,6 +3550,7 @@ static unsigned long deferred_split_scan(struct shrinker *shrink, }else {/* We lost race with folio_put() */ list_del_init (&folio->_deferred_list); +__folio_clear_partially_mapped (folio);
as above? Do we also need if(test) for split_huge_page_to_list_to_order()?
Quoted 8 lines
mm/huge_memory.c +0
-0ds_queue->split_queue_len--; }if (!--sc->nr_to_scan)mm/internal.h +0
-0
mm/internal.h static inline void prep_compound_head(struct page *page, unsigned int order) · line 667
Quoted 7 lines
mm/internal.h +2
-1@@ -662,8 +662,10 @@ static inline void prep_compound_head(struct page *page, unsigned int order)atomic_set (&folio->_entire_mapcount,-1 );atomic_set (&folio->_nr_pages_mapped,0 );atomic_set (&folio->_pincount,0 ); -if (order >1 ) +if (order >1 ) {INIT_LIST_HEAD (&folio->_deferred_list); +__folio_clear_partially_mapped (folio);
if partially_mapped is true for a new folio, does it mean we already have a bug somewhere?
How is it possible for a new folio to be partially mapped?
Quoted 79 lines
mm/internal.h +1
-0+ } }static inline void prep_compound_tail (struct page *head,int tail_idx)mm/memcontrol.c +2
-1@@ -4669,7 +4669,8 @@ static void uncharge_folio(struct folio *folio, struct uncharge_gather *ug)VM_BUG_ON_FOLIO (folio_test_lru (folio), folio);VM_BUG_ON_FOLIO (folio_order (folio) >1 && !folio_test_hugetlb (folio) && - !list_empty (&folio->_deferred_list), folio); + !list_empty (&folio->_deferred_list) && +folio_test_partially_mapped (folio), folio);/* * Nobody should be changing or seriously looking at mm/migrate.c +2
-1@@ -1735,7 +1735,8 @@ static int migrate_pages_batch(struct list_head *from,* use _deferred_list. */ if (nr_pages >2 && - !list_empty (&folio->_deferred_list)) { + !list_empty (&folio->_deferred_list) && +folio_test_partially_mapped (folio)) {if (!try_split_folio (folio, split_folios, mode)) { nr_failed++; stats->nr_thp_failed += is_thp;mm/page_alloc.c +3
-2@@ -957,8 +957,9 @@ static int free_tail_page_prepare(struct page *head_page, struct page *page)break ;case 2 :/* the second tail page: deferred_list overlaps ->mapping */ -if (unlikely (!list_empty (&folio->_deferred_list))) { -bad_page (page,"on deferred list" ); +if (unlikely (!list_empty (&folio->_deferred_list) && +folio_test_partially_mapped (folio))) { +bad_page (page,"partially mapped folio on deferred list" ); goto out; }break ;mm/rmap.c +3
-2@@ -1578,8 +1578,9 @@ static __always_inline void __folio_remove_rmap(struct folio *folio,* Check partially_mapped first to ensure it is a large folio. */ if (partially_mapped &&folio_test_anon (folio) && -list_empty (&folio->_deferred_list)) -deferred_split_folio (folio); + !folio_test_partially_mapped (folio)) +deferred_split_folio (folio, true); +__folio_mod_stat (folio, -nr, -nr_pmdmapped);/* mm/vmscan.c +2
-2@@ -1233,7 +1233,8 @@ static unsigned int shrink_folio_list(struct list_head *folio_list,* Split partially mapped folios right away. * We can free the unmapped pages without IO. */ -if (data_race (!list_empty (&folio->_deferred_list)) && +if (data_race (!list_empty (&folio->_deferred_list) && +folio_test_partially_mapped (folio)) &&split_folio_to_list (folio, folio_list)) goto activate_locked; } --
Quoted 2 lines
2.43.5
Thanks
Barry
Barry Song
reviewerRe: [PATCH v4 4/6] mm: Introduce a pageflag for partially mapped folios
Quoted 111 lines · Barry Song
Hi Usama,
I feel it is much better now! thanks!
Quoted 98 lines · Usama Arif
Currently
folio->_deferred_listis used to keep track ofpartially_mappedfolios that are going to be split under memory pressure. In the next patch, all THPs that are faulted in and collapsed by khugepaged are also going to be tracked using_deferred_list.This patch introduces a pageflag to be able to distinguish between partially mapped folios and others in the
deferred_listat split time indeferred_split_scan. Its needed as__folio_remove_rmapdecrements_mapcount,_large_mapcountand_entire_mapcount, hence it won't be possible to distinguish between partially mapped folios and others indeferred_split_scan.Eventhough it introduces an extra flag to track if the folio is partially mapped, there is no functional change intended with this patch and the flag is not useful in this patch itself, it will become useful in the next patch when
_deferred_listhas non partially mapped folios.Signed-off-by Usama Arif <[email protected]>--- include/linux/huge_mm.h | 4 ++-- include/linux/page-flags.h | 11 +++++++++++ mm/huge_memory.c | 23 ++++++++++++++++------- mm/internal.h | 4 +++- mm/memcontrol.c | 3 ++- mm/migrate.c | 3 ++- mm/page_alloc.c | 5 +++-- mm/rmap.c | 5 +++-- mm/vmscan.c | 3 ++- 9 files changed, 44 insertions(+), 17 deletions(-)include/linux/huge_mm.h +2
-2@@ -321,7 +321,7 @@ static inline int split_huge_page(struct page *page) {return split_huge_page_to_list_to_order (page,NULL ,0 ); } -void deferred_split_folio (struct folio *folio); +void deferred_split_folio (struct folio *folio,bool partially_mapped);void __split_huge_pmd (struct vm_area_struct *vma,pmd_t *pmd,unsigned long address,bool freeze,struct folio *folio);@@ -495,7 +495,7 @@ static inline int split_huge_page(struct page *page) {return 0 ; } -static inline void deferred_split_folio (struct folio *folio) {} +static inline void deferred_split_folio (struct folio *folio,bool partially_mapped) {}#define split_huge_pmd (__vma, __pmd, __address) \ do { } while (0)include/linux/page-flags.h +11
-0@@ -182,6 +182,7 @@ enum pageflags {/* At least one page in this folio has the hwpoison flag set */ PG_has_hwpoisoned = PG_active, PG_large_rmappable = PG_workingset,/* anon or file-backed */ + PG_partially_mapped = PG_reclaim, /* was identified to be partially mapped */ };#define PAGEFLAGS_MASK ((1UL << NR_PAGEFLAGS) - 1)@@ -861,8 +862,18 @@ static inline void ClearPageCompound(struct page *page)ClearPageHead (page); } FOLIO_FLAG(large_rmappable, FOLIO_SECOND_PAGE) +FOLIO_TEST_FLAG (partially_mapped,FOLIO_SECOND_PAGE ) +/* +* PG_partially_mapped is protected by deferred_split split_queue_lock, +* so its safe to use non-atomic set/clear. + */ +__FOLIO_SET_FLAG(partially_mapped, FOLIO_SECOND_PAGE) +__FOLIO_CLEAR_FLAG(partially_mapped, FOLIO_SECOND_PAGE)#else FOLIO_FLAG_FALSE(large_rmappable) +FOLIO_TEST_FLAG_FALSE(partially_mapped) +__FOLIO_SET_FLAG_NOOP(partially_mapped) +__FOLIO_CLEAR_FLAG_NOOP(partially_mapped)#endif #define PG_head_mask ((1UL << PG_head))mm/huge_memory.c +1
-0@@ -3398,6 +3398,7 @@ int split_huge_page_to_list_to_order(struct page *page, struct list_head *list,* page_deferred_list. */ list_del_init (&folio->_deferred_list); +__folio_clear_partially_mapped (folio); }spin_unlock (&ds_queue->split_queue_lock);if (mapping) {mm/huge_memory.c void __folio_undo_large_rmappable(struct folio *folio) · line 3458
Quoted 4 lines
mm/huge_memory.c +1
-0@@ -3454,11 +3455,13 @@ void __folio_undo_large_rmappable(struct folio *folio)if (!list_empty (&folio->_deferred_list)) { ds_queue->split_queue_len--;list_del_init (&folio->_deferred_list); +__folio_clear_partially_mapped (folio);is it possible to make things clearer by
Quoted 2 lines
if (
folio_clear_partially_mapped)__folio_clear_partially_mapped(folio);
sorry for typo,
if (folio_test_partially_mapped)
__folio_clear_partially_mapped(folio);
Quoted 155 lines
While writing without conditions isn't necessarily wrong, adding a condition will improve the readability of the code and enhance the clarity of my mTHP counters series. also help decrease smp cache sync if we can avoid unnecessary writing?
Quoted 37 lines
mm/huge_memory.c +13
-7}spin_unlock_irqrestore (&ds_queue->split_queue_lock, flags); } -void deferred_split_folio (struct folio *folio) +/* partially_mapped=false won't clear PG_partially_mapped folio flag */ +void deferred_split_folio (struct folio *folio,bool partially_mapped) {struct deferred_split *ds_queue =get_deferred_split_queue (folio);#ifdef CONFIG_MEMCG @@ -3486,14 +3489,19 @@ void deferred_split_folio(struct folio *folio)if (folio_test_swapcache (folio))return ; -if (!list_empty (&folio->_deferred_list)) -return ; -spin_lock_irqsave (&ds_queue->split_queue_lock, flags); +if (partially_mapped) { +if (!folio_test_partially_mapped (folio)) { +__folio_set_partially_mapped (folio); +if (folio_test_pmd_mappable (folio)) +count_vm_event (THP_DEFERRED_SPLIT_PAGE ); +count_mthp_stat (folio_order (folio),MTHP_STAT_SPLIT_DEFERRED ); + } + }else { +/* partially mapped folios cannot become non-partially mapped */ +VM_WARN_ON_FOLIO (folio_test_partially_mapped (folio), folio); + }if (list_empty (&folio->_deferred_list)) { -if (folio_test_pmd_mappable (folio)) -count_vm_event (THP_DEFERRED_SPLIT_PAGE ); -count_mthp_stat (folio_order (folio),MTHP_STAT_SPLIT_DEFERRED );list_add_tail (&folio->_deferred_list, &ds_queue->split_queue); ds_queue->split_queue_len++;#ifdef CONFIG_MEMCG mm/huge_memory.c static unsigned long deferred_split_scan(struct shrinker *shrink, · line 3553
Quoted 4 lines
mm/huge_memory.c +1
-0@@ -3542,6 +3550,7 @@ static unsigned long deferred_split_scan(struct shrinker *shrink, }else {/* We lost race with folio_put() */ list_del_init (&folio->_deferred_list); +__folio_clear_partially_mapped (folio);as above? Do we also need if(test) for
split_huge_page_to_list_to_order()?Quoted 8 lines
mm/huge_memory.c +0
-0ds_queue->split_queue_len--; }if (!--sc->nr_to_scan)mm/internal.h +0
-0mm/internal.h static inline void prep_compound_head(struct page *page, unsigned int order) · line 667
Quoted 7 lines
mm/internal.h +2
-1@@ -662,8 +662,10 @@ static inline void prep_compound_head(struct page *page, unsigned int order)atomic_set (&folio->_entire_mapcount,-1 );atomic_set (&folio->_nr_pages_mapped,0 );atomic_set (&folio->_pincount,0 ); -if (order >1 ) +if (order >1 ) {INIT_LIST_HEAD (&folio->_deferred_list); +__folio_clear_partially_mapped (folio);if
partially_mappedis true for a new folio, does it mean we already have a bug somewhere?How is it possible for a new folio to be partially mapped?
Quoted 79 lines
mm/internal.h +1
-0+ } }static inline void prep_compound_tail (struct page *head,int tail_idx)mm/memcontrol.c +2
-1@@ -4669,7 +4669,8 @@ static void uncharge_folio(struct folio *folio, struct uncharge_gather *ug)VM_BUG_ON_FOLIO (folio_test_lru (folio), folio);VM_BUG_ON_FOLIO (folio_order (folio) >1 && !folio_test_hugetlb (folio) && - !list_empty (&folio->_deferred_list), folio); + !list_empty (&folio->_deferred_list) && +folio_test_partially_mapped (folio), folio);/* * Nobody should be changing or seriously looking at mm/migrate.c +2
-1@@ -1735,7 +1735,8 @@ static int migrate_pages_batch(struct list_head *from,* use _deferred_list. */ if (nr_pages >2 && - !list_empty (&folio->_deferred_list)) { + !list_empty (&folio->_deferred_list) && +folio_test_partially_mapped (folio)) {if (!try_split_folio (folio, split_folios, mode)) { nr_failed++; stats->nr_thp_failed += is_thp;mm/page_alloc.c +3
-2@@ -957,8 +957,9 @@ static int free_tail_page_prepare(struct page *head_page, struct page *page)break ;case 2 :/* the second tail page: deferred_list overlaps ->mapping */ -if (unlikely (!list_empty (&folio->_deferred_list))) { -bad_page (page,"on deferred list" ); +if (unlikely (!list_empty (&folio->_deferred_list) && +folio_test_partially_mapped (folio))) { +bad_page (page,"partially mapped folio on deferred list" ); goto out; }break ;mm/rmap.c +3
-2@@ -1578,8 +1578,9 @@ static __always_inline void __folio_remove_rmap(struct folio *folio,* Check partially_mapped first to ensure it is a large folio. */ if (partially_mapped &&folio_test_anon (folio) && -list_empty (&folio->_deferred_list)) -deferred_split_folio (folio); + !folio_test_partially_mapped (folio)) +deferred_split_folio (folio, true); +__folio_mod_stat (folio, -nr, -nr_pmdmapped);/* mm/vmscan.c +2
-2@@ -1233,7 +1233,8 @@ static unsigned int shrink_folio_list(struct list_head *folio_list,* Split partially mapped folios right away. * We can free the unmapped pages without IO. */ -if (data_race (!list_empty (&folio->_deferred_list)) && +if (data_race (!list_empty (&folio->_deferred_list) && +folio_test_partially_mapped (folio)) &&split_folio_to_list (folio, folio_list)) goto activate_locked; } --Quoted 2 lines
2.43.5
Thanks
Barry
Re: [PATCH v4 4/6] mm: Introduce a pageflag for partially mapped folios
In reply to Barry SongQuoted 110 lines · Barry Song
Hi Usama,
I feel it is much better now! thanks!
Quoted 98 lines · Usama Arif
Currently
folio->_deferred_listis used to keep track ofpartially_mappedfolios that are going to be split under memory pressure. In the next patch, all THPs that are faulted in and collapsed by khugepaged are also going to be tracked using_deferred_list.This patch introduces a pageflag to be able to distinguish between partially mapped folios and others in the
deferred_listat split time indeferred_split_scan. Its needed as__folio_remove_rmapdecrements_mapcount,_large_mapcountand_entire_mapcount, hence it won't be possible to distinguish between partially mapped folios and others indeferred_split_scan.Eventhough it introduces an extra flag to track if the folio is partially mapped, there is no functional change intended with this patch and the flag is not useful in this patch itself, it will become useful in the next patch when
_deferred_listhas non partially mapped folios.Signed-off-by Usama Arif <[email protected]>--- include/linux/huge_mm.h | 4 ++-- include/linux/page-flags.h | 11 +++++++++++ mm/huge_memory.c | 23 ++++++++++++++++------- mm/internal.h | 4 +++- mm/memcontrol.c | 3 ++- mm/migrate.c | 3 ++- mm/page_alloc.c | 5 +++-- mm/rmap.c | 5 +++-- mm/vmscan.c | 3 ++- 9 files changed, 44 insertions(+), 17 deletions(-)include/linux/huge_mm.h +2
-2@@ -321,7 +321,7 @@ static inline int split_huge_page(struct page *page) {return split_huge_page_to_list_to_order (page,NULL ,0 ); } -void deferred_split_folio (struct folio *folio); +void deferred_split_folio (struct folio *folio,bool partially_mapped);void __split_huge_pmd (struct vm_area_struct *vma,pmd_t *pmd,unsigned long address,bool freeze,struct folio *folio);@@ -495,7 +495,7 @@ static inline int split_huge_page(struct page *page) {return 0 ; } -static inline void deferred_split_folio (struct folio *folio) {} +static inline void deferred_split_folio (struct folio *folio,bool partially_mapped) {}#define split_huge_pmd (__vma, __pmd, __address) \ do { } while (0)include/linux/page-flags.h +11
-0@@ -182,6 +182,7 @@ enum pageflags {/* At least one page in this folio has the hwpoison flag set */ PG_has_hwpoisoned = PG_active, PG_large_rmappable = PG_workingset,/* anon or file-backed */ + PG_partially_mapped = PG_reclaim, /* was identified to be partially mapped */ };#define PAGEFLAGS_MASK ((1UL << NR_PAGEFLAGS) - 1)@@ -861,8 +862,18 @@ static inline void ClearPageCompound(struct page *page)ClearPageHead (page); } FOLIO_FLAG(large_rmappable, FOLIO_SECOND_PAGE) +FOLIO_TEST_FLAG (partially_mapped,FOLIO_SECOND_PAGE ) +/* +* PG_partially_mapped is protected by deferred_split split_queue_lock, +* so its safe to use non-atomic set/clear. + */ +__FOLIO_SET_FLAG(partially_mapped, FOLIO_SECOND_PAGE) +__FOLIO_CLEAR_FLAG(partially_mapped, FOLIO_SECOND_PAGE)#else FOLIO_FLAG_FALSE(large_rmappable) +FOLIO_TEST_FLAG_FALSE(partially_mapped) +__FOLIO_SET_FLAG_NOOP(partially_mapped) +__FOLIO_CLEAR_FLAG_NOOP(partially_mapped)#endif #define PG_head_mask ((1UL << PG_head))mm/huge_memory.c +1
-0@@ -3398,6 +3398,7 @@ int split_huge_page_to_list_to_order(struct page *page, struct list_head *list,* page_deferred_list. */ list_del_init (&folio->_deferred_list); +__folio_clear_partially_mapped (folio); }spin_unlock (&ds_queue->split_queue_lock);if (mapping) {mm/huge_memory.c void __folio_undo_large_rmappable(struct folio *folio) · line 3458
Quoted 4 lines
mm/huge_memory.c +1
-0@@ -3454,11 +3455,13 @@ void __folio_undo_large_rmappable(struct folio *folio)if (!list_empty (&folio->_deferred_list)) { ds_queue->split_queue_len--;list_del_init (&folio->_deferred_list); +__folio_clear_partially_mapped (folio);is it possible to make things clearer by
Quoted 8 lines
if (
folio_clear_partially_mapped)__folio_clear_partially_mapped(folio);While writing without conditions isn't necessarily wrong, adding a condition will improve the readability of the code and enhance the clarity of my mTHP counters series. also help decrease smp cache sync if we can avoid unnecessary writing?
Do you mean if(folio_test_partially_mapped(folio))?
I don't like this idea. I think it makes the readability worse? If I was looking at if (test) -> clear for the first time, I would become confused why its being tested if its going to be clear at the end anyways?
Quoted 63 lines
Quoted 37 lines
mm/huge_memory.c +13
-7}spin_unlock_irqrestore (&ds_queue->split_queue_lock, flags); } -void deferred_split_folio (struct folio *folio) +/* partially_mapped=false won't clear PG_partially_mapped folio flag */ +void deferred_split_folio (struct folio *folio,bool partially_mapped) {struct deferred_split *ds_queue =get_deferred_split_queue (folio);#ifdef CONFIG_MEMCG @@ -3486,14 +3489,19 @@ void deferred_split_folio(struct folio *folio)if (folio_test_swapcache (folio))return ; -if (!list_empty (&folio->_deferred_list)) -return ; -spin_lock_irqsave (&ds_queue->split_queue_lock, flags); +if (partially_mapped) { +if (!folio_test_partially_mapped (folio)) { +__folio_set_partially_mapped (folio); +if (folio_test_pmd_mappable (folio)) +count_vm_event (THP_DEFERRED_SPLIT_PAGE ); +count_mthp_stat (folio_order (folio),MTHP_STAT_SPLIT_DEFERRED ); + } + }else { +/* partially mapped folios cannot become non-partially mapped */ +VM_WARN_ON_FOLIO (folio_test_partially_mapped (folio), folio); + }if (list_empty (&folio->_deferred_list)) { -if (folio_test_pmd_mappable (folio)) -count_vm_event (THP_DEFERRED_SPLIT_PAGE ); -count_mthp_stat (folio_order (folio),MTHP_STAT_SPLIT_DEFERRED );list_add_tail (&folio->_deferred_list, &ds_queue->split_queue); ds_queue->split_queue_len++;#ifdef CONFIG_MEMCG mm/huge_memory.c static unsigned long deferred_split_scan(struct shrinker *shrink, · line 3553
Quoted 4 lines
mm/huge_memory.c +1
-0@@ -3542,6 +3550,7 @@ static unsigned long deferred_split_scan(struct shrinker *shrink, }else {/* We lost race with folio_put() */ list_del_init (&folio->_deferred_list); +__folio_clear_partially_mapped (folio);as above? Do we also need if(test) for
split_huge_page_to_list_to_order()?Quoted 8 lines
mm/huge_memory.c +0
-0ds_queue->split_queue_len--; }if (!--sc->nr_to_scan)mm/internal.h +0
-0mm/internal.h static inline void prep_compound_head(struct page *page, unsigned int order) · line 667
Quoted 7 lines
mm/internal.h +2
-1@@ -662,8 +662,10 @@ static inline void prep_compound_head(struct page *page, unsigned int order)atomic_set (&folio->_entire_mapcount,-1 );atomic_set (&folio->_nr_pages_mapped,0 );atomic_set (&folio->_pincount,0 ); -if (order >1 ) +if (order >1 ) {INIT_LIST_HEAD (&folio->_deferred_list); +__folio_clear_partially_mapped (folio);if
partially_mappedis true for a new folio, does it mean we already have a bug somewhere?
Quoted 2 lines
How is it possible for a new folio to be partially mapped?
Its not, I did it because I wanted to make it explicit that the folio is being initialized, similar to how before this INIT_LIST_HEAD(&folio->_deferred_list) is done here.
There is no functional issue in removing it here, because I believe the flag is initialized to false from start.
Quoted 84 lines
Quoted 79 lines
mm/internal.h +1
-0+ } }static inline void prep_compound_tail (struct page *head,int tail_idx)mm/memcontrol.c +2
-1@@ -4669,7 +4669,8 @@ static void uncharge_folio(struct folio *folio, struct uncharge_gather *ug)VM_BUG_ON_FOLIO (folio_test_lru (folio), folio);VM_BUG_ON_FOLIO (folio_order (folio) >1 && !folio_test_hugetlb (folio) && - !list_empty (&folio->_deferred_list), folio); + !list_empty (&folio->_deferred_list) && +folio_test_partially_mapped (folio), folio);/* * Nobody should be changing or seriously looking at mm/migrate.c +2
-1@@ -1735,7 +1735,8 @@ static int migrate_pages_batch(struct list_head *from,* use _deferred_list. */ if (nr_pages >2 && - !list_empty (&folio->_deferred_list)) { + !list_empty (&folio->_deferred_list) && +folio_test_partially_mapped (folio)) {if (!try_split_folio (folio, split_folios, mode)) { nr_failed++; stats->nr_thp_failed += is_thp;mm/page_alloc.c +3
-2@@ -957,8 +957,9 @@ static int free_tail_page_prepare(struct page *head_page, struct page *page)break ;case 2 :/* the second tail page: deferred_list overlaps ->mapping */ -if (unlikely (!list_empty (&folio->_deferred_list))) { -bad_page (page,"on deferred list" ); +if (unlikely (!list_empty (&folio->_deferred_list) && +folio_test_partially_mapped (folio))) { +bad_page (page,"partially mapped folio on deferred list" ); goto out; }break ;mm/rmap.c +3
-2@@ -1578,8 +1578,9 @@ static __always_inline void __folio_remove_rmap(struct folio *folio,* Check partially_mapped first to ensure it is a large folio. */ if (partially_mapped &&folio_test_anon (folio) && -list_empty (&folio->_deferred_list)) -deferred_split_folio (folio); + !folio_test_partially_mapped (folio)) +deferred_split_folio (folio, true); +__folio_mod_stat (folio, -nr, -nr_pmdmapped);/* mm/vmscan.c +2
-2@@ -1233,7 +1233,8 @@ static unsigned int shrink_folio_list(struct list_head *folio_list,* Split partially mapped folios right away. * We can free the unmapped pages without IO. */ -if (data_race (!list_empty (&folio->_deferred_list)) && +if (data_race (!list_empty (&folio->_deferred_list) && +folio_test_partially_mapped (folio)) &&split_folio_to_list (folio, folio_list)) goto activate_locked; } --Quoted 2 lines
2.43.5
Thanks
Barry
Barry Song
reviewerRe: [PATCH v4 4/6] mm: Introduce a pageflag for partially mapped folios
Quoted 123 lines · Usama Arif
Quoted 110 lines · Barry Song
Hi Usama,
I feel it is much better now! thanks!
Quoted 98 lines · Usama Arif
Currently
folio->_deferred_listis used to keep track ofpartially_mappedfolios that are going to be split under memory pressure. In the next patch, all THPs that are faulted in and collapsed by khugepaged are also going to be tracked using_deferred_list.This patch introduces a pageflag to be able to distinguish between partially mapped folios and others in the
deferred_listat split time indeferred_split_scan. Its needed as__folio_remove_rmapdecrements_mapcount,_large_mapcountand_entire_mapcount, hence it won't be possible to distinguish between partially mapped folios and others indeferred_split_scan.Eventhough it introduces an extra flag to track if the folio is partially mapped, there is no functional change intended with this patch and the flag is not useful in this patch itself, it will become useful in the next patch when
_deferred_listhas non partially mapped folios.Signed-off-by Usama Arif <[email protected]>--- include/linux/huge_mm.h | 4 ++-- include/linux/page-flags.h | 11 +++++++++++ mm/huge_memory.c | 23 ++++++++++++++++------- mm/internal.h | 4 +++- mm/memcontrol.c | 3 ++- mm/migrate.c | 3 ++- mm/page_alloc.c | 5 +++-- mm/rmap.c | 5 +++-- mm/vmscan.c | 3 ++- 9 files changed, 44 insertions(+), 17 deletions(-)include/linux/huge_mm.h +2
-2@@ -321,7 +321,7 @@ static inline int split_huge_page(struct page *page) {return split_huge_page_to_list_to_order (page,NULL ,0 ); } -void deferred_split_folio (struct folio *folio); +void deferred_split_folio (struct folio *folio,bool partially_mapped);void __split_huge_pmd (struct vm_area_struct *vma,pmd_t *pmd,unsigned long address,bool freeze,struct folio *folio);@@ -495,7 +495,7 @@ static inline int split_huge_page(struct page *page) {return 0 ; } -static inline void deferred_split_folio (struct folio *folio) {} +static inline void deferred_split_folio (struct folio *folio,bool partially_mapped) {}#define split_huge_pmd (__vma, __pmd, __address) \ do { } while (0)include/linux/page-flags.h +11
-0@@ -182,6 +182,7 @@ enum pageflags {/* At least one page in this folio has the hwpoison flag set */ PG_has_hwpoisoned = PG_active, PG_large_rmappable = PG_workingset,/* anon or file-backed */ + PG_partially_mapped = PG_reclaim, /* was identified to be partially mapped */ };#define PAGEFLAGS_MASK ((1UL << NR_PAGEFLAGS) - 1)@@ -861,8 +862,18 @@ static inline void ClearPageCompound(struct page *page)ClearPageHead (page); } FOLIO_FLAG(large_rmappable, FOLIO_SECOND_PAGE) +FOLIO_TEST_FLAG (partially_mapped,FOLIO_SECOND_PAGE ) +/* +* PG_partially_mapped is protected by deferred_split split_queue_lock, +* so its safe to use non-atomic set/clear. + */ +__FOLIO_SET_FLAG(partially_mapped, FOLIO_SECOND_PAGE) +__FOLIO_CLEAR_FLAG(partially_mapped, FOLIO_SECOND_PAGE)#else FOLIO_FLAG_FALSE(large_rmappable) +FOLIO_TEST_FLAG_FALSE(partially_mapped) +__FOLIO_SET_FLAG_NOOP(partially_mapped) +__FOLIO_CLEAR_FLAG_NOOP(partially_mapped)#endif #define PG_head_mask ((1UL << PG_head))mm/huge_memory.c +1
-0@@ -3398,6 +3398,7 @@ int split_huge_page_to_list_to_order(struct page *page, struct list_head *list,* page_deferred_list. */ list_del_init (&folio->_deferred_list); +__folio_clear_partially_mapped (folio); }spin_unlock (&ds_queue->split_queue_lock);if (mapping) {mm/huge_memory.c void __folio_undo_large_rmappable(struct folio *folio) · line 3458
Quoted 4 lines
mm/huge_memory.c +1
-0@@ -3454,11 +3455,13 @@ void __folio_undo_large_rmappable(struct folio *folio)if (!list_empty (&folio->_deferred_list)) { ds_queue->split_queue_len--;list_del_init (&folio->_deferred_list); +__folio_clear_partially_mapped (folio);is it possible to make things clearer by
Quoted 8 lines
if (
folio_clear_partially_mapped)__folio_clear_partially_mapped(folio);While writing without conditions isn't necessarily wrong, adding a condition will improve the readability of the code and enhance the clarity of my mTHP counters series. also help decrease smp cache sync if we can avoid unnecessary writing?
Quoted 3 lines
Do you mean if(folio_test_partially_mapped(folio))?
I don't like this idea. I think it makes the readability worse? If I was looking at if (test) -> clear for the first time, I would become confused why its being tested if its going to be clear at the end anyways?
In the pmd-order case, the majority of folios are not partially mapped. Unconditional writes will trigger cache synchronization across all CPUs (related to the MESI protocol), making them more costly. By using conditional writes, such as "if(test) write," we can avoid most unnecessary writes, which is much more efficient. Additionally, we only need to manage nr_split_deferred when the condition is met. We are carefully evaluating all scenarios to determine if modifications to the partially_mapped flag are necessary.
Quoted 155 lines
Quoted 65 lines
Quoted 39 lines
mm/huge_memory.c +13
-7}spin_unlock_irqrestore (&ds_queue->split_queue_lock, flags); } -void deferred_split_folio (struct folio *folio) +/* partially_mapped=false won't clear PG_partially_mapped folio flag */ +void deferred_split_folio (struct folio *folio,bool partially_mapped) {struct deferred_split *ds_queue =get_deferred_split_queue (folio);#ifdef CONFIG_MEMCG @@ -3486,14 +3489,19 @@ void deferred_split_folio(struct folio *folio)if (folio_test_swapcache (folio))return ; -if (!list_empty (&folio->_deferred_list)) -return ; -spin_lock_irqsave (&ds_queue->split_queue_lock, flags); +if (partially_mapped) { +if (!folio_test_partially_mapped (folio)) { +__folio_set_partially_mapped (folio); +if (folio_test_pmd_mappable (folio)) +count_vm_event (THP_DEFERRED_SPLIT_PAGE ); +count_mthp_stat (folio_order (folio),MTHP_STAT_SPLIT_DEFERRED ); + } + }else { +/* partially mapped folios cannot become non-partially mapped */ +VM_WARN_ON_FOLIO (folio_test_partially_mapped (folio), folio); + }if (list_empty (&folio->_deferred_list)) { -if (folio_test_pmd_mappable (folio)) -count_vm_event (THP_DEFERRED_SPLIT_PAGE ); -count_mthp_stat (folio_order (folio),MTHP_STAT_SPLIT_DEFERRED );list_add_tail (&folio->_deferred_list, &ds_queue->split_queue); ds_queue->split_queue_len++;#ifdef CONFIG_MEMCG mm/huge_memory.c static unsigned long deferred_split_scan(struct shrinker *shrink, · line 3553
Quoted 4 lines
mm/huge_memory.c +1
-0@@ -3542,6 +3550,7 @@ static unsigned long deferred_split_scan(struct shrinker *shrink, }else {/* We lost race with folio_put() */ list_del_init (&folio->_deferred_list); +__folio_clear_partially_mapped (folio);as above? Do we also need if(test) for
split_huge_page_to_list_to_order()?Quoted 8 lines
mm/huge_memory.c +0
-0ds_queue->split_queue_len--; }if (!--sc->nr_to_scan)mm/internal.h +0
-0mm/internal.h static inline void prep_compound_head(struct page *page, unsigned int order) · line 667
Quoted 7 lines
mm/internal.h +2
-1@@ -662,8 +662,10 @@ static inline void prep_compound_head(struct page *page, unsigned int order)atomic_set (&folio->_entire_mapcount,-1 );atomic_set (&folio->_nr_pages_mapped,0 );atomic_set (&folio->_pincount,0 ); -if (order >1 ) +if (order >1 ) {INIT_LIST_HEAD (&folio->_deferred_list); +__folio_clear_partially_mapped (folio);if
partially_mappedis true for a new folio, does it mean we already have a bug somewhere?Quoted 2 lines
How is it possible for a new folio to be partially mapped?
Its not, I did it because I wanted to make it explicit that the folio is being initialized, similar to how before this
INIT_LIST_HEAD(&folio->_deferred_list)is done here.There is no functional issue in removing it here, because I believe the flag is initialized to false from start.
Quoted 84 lines
Quoted 79 lines
mm/internal.h +1
-0+ } }static inline void prep_compound_tail (struct page *head,int tail_idx)mm/memcontrol.c +2
-1@@ -4669,7 +4669,8 @@ static void uncharge_folio(struct folio *folio, struct uncharge_gather *ug)VM_BUG_ON_FOLIO (folio_test_lru (folio), folio);VM_BUG_ON_FOLIO (folio_order (folio) >1 && !folio_test_hugetlb (folio) && - !list_empty (&folio->_deferred_list), folio); + !list_empty (&folio->_deferred_list) && +folio_test_partially_mapped (folio), folio);/* * Nobody should be changing or seriously looking at mm/migrate.c +2
-1@@ -1735,7 +1735,8 @@ static int migrate_pages_batch(struct list_head *from,* use _deferred_list. */ if (nr_pages >2 && - !list_empty (&folio->_deferred_list)) { + !list_empty (&folio->_deferred_list) && +folio_test_partially_mapped (folio)) {if (!try_split_folio (folio, split_folios, mode)) { nr_failed++; stats->nr_thp_failed += is_thp;mm/page_alloc.c +3
-2@@ -957,8 +957,9 @@ static int free_tail_page_prepare(struct page *head_page, struct page *page)break ;case 2 :/* the second tail page: deferred_list overlaps ->mapping */ -if (unlikely (!list_empty (&folio->_deferred_list))) { -bad_page (page,"on deferred list" ); +if (unlikely (!list_empty (&folio->_deferred_list) && +folio_test_partially_mapped (folio))) { +bad_page (page,"partially mapped folio on deferred list" ); goto out; }break ;mm/rmap.c +3
-2@@ -1578,8 +1578,9 @@ static __always_inline void __folio_remove_rmap(struct folio *folio,* Check partially_mapped first to ensure it is a large folio. */ if (partially_mapped &&folio_test_anon (folio) && -list_empty (&folio->_deferred_list)) -deferred_split_folio (folio); + !folio_test_partially_mapped (folio)) +deferred_split_folio (folio, true); +__folio_mod_stat (folio, -nr, -nr_pmdmapped);/* mm/vmscan.c +2
-2@@ -1233,7 +1233,8 @@ static unsigned int shrink_folio_list(struct list_head *folio_list,* Split partially mapped folios right away. * We can free the unmapped pages without IO. */ -if (data_race (!list_empty (&folio->_deferred_list)) && +if (data_race (!list_empty (&folio->_deferred_list) && +folio_test_partially_mapped (folio)) &&split_folio_to_list (folio, folio_list)) goto activate_locked; } --Quoted 2 lines
2.43.5
Thanks
Barry
Re: [PATCH v4 4/6] mm: Introduce a pageflag for partially mapped folios
Quoted 128 lines · Barry Song
Quoted 123 lines · Usama Arif
Quoted 110 lines · Barry Song
Hi Usama,
I feel it is much better now! thanks!
Quoted 98 lines · Usama Arif
Currently
folio->_deferred_listis used to keep track ofpartially_mappedfolios that are going to be split under memory pressure. In the next patch, all THPs that are faulted in and collapsed by khugepaged are also going to be tracked using_deferred_list.This patch introduces a pageflag to be able to distinguish between partially mapped folios and others in the
deferred_listat split time indeferred_split_scan. Its needed as__folio_remove_rmapdecrements_mapcount,_large_mapcountand_entire_mapcount, hence it won't be possible to distinguish between partially mapped folios and others indeferred_split_scan.Eventhough it introduces an extra flag to track if the folio is partially mapped, there is no functional change intended with this patch and the flag is not useful in this patch itself, it will become useful in the next patch when
_deferred_listhas non partially mapped folios.Signed-off-by Usama Arif <[email protected]>--- include/linux/huge_mm.h | 4 ++-- include/linux/page-flags.h | 11 +++++++++++ mm/huge_memory.c | 23 ++++++++++++++++------- mm/internal.h | 4 +++- mm/memcontrol.c | 3 ++- mm/migrate.c | 3 ++- mm/page_alloc.c | 5 +++-- mm/rmap.c | 5 +++-- mm/vmscan.c | 3 ++- 9 files changed, 44 insertions(+), 17 deletions(-)include/linux/huge_mm.h +2
-2@@ -321,7 +321,7 @@ static inline int split_huge_page(struct page *page) {return split_huge_page_to_list_to_order (page,NULL ,0 ); } -void deferred_split_folio (struct folio *folio); +void deferred_split_folio (struct folio *folio,bool partially_mapped);void __split_huge_pmd (struct vm_area_struct *vma,pmd_t *pmd,unsigned long address,bool freeze,struct folio *folio);@@ -495,7 +495,7 @@ static inline int split_huge_page(struct page *page) {return 0 ; } -static inline void deferred_split_folio (struct folio *folio) {} +static inline void deferred_split_folio (struct folio *folio,bool partially_mapped) {}#define split_huge_pmd (__vma, __pmd, __address) \ do { } while (0)include/linux/page-flags.h +11
-0@@ -182,6 +182,7 @@ enum pageflags {/* At least one page in this folio has the hwpoison flag set */ PG_has_hwpoisoned = PG_active, PG_large_rmappable = PG_workingset,/* anon or file-backed */ + PG_partially_mapped = PG_reclaim, /* was identified to be partially mapped */ };#define PAGEFLAGS_MASK ((1UL << NR_PAGEFLAGS) - 1)@@ -861,8 +862,18 @@ static inline void ClearPageCompound(struct page *page)ClearPageHead (page); } FOLIO_FLAG(large_rmappable, FOLIO_SECOND_PAGE) +FOLIO_TEST_FLAG (partially_mapped,FOLIO_SECOND_PAGE ) +/* +* PG_partially_mapped is protected by deferred_split split_queue_lock, +* so its safe to use non-atomic set/clear. + */ +__FOLIO_SET_FLAG(partially_mapped, FOLIO_SECOND_PAGE) +__FOLIO_CLEAR_FLAG(partially_mapped, FOLIO_SECOND_PAGE)#else FOLIO_FLAG_FALSE(large_rmappable) +FOLIO_TEST_FLAG_FALSE(partially_mapped) +__FOLIO_SET_FLAG_NOOP(partially_mapped) +__FOLIO_CLEAR_FLAG_NOOP(partially_mapped)#endif #define PG_head_mask ((1UL << PG_head))mm/huge_memory.c +1
-0@@ -3398,6 +3398,7 @@ int split_huge_page_to_list_to_order(struct page *page, struct list_head *list,* page_deferred_list. */ list_del_init (&folio->_deferred_list); +__folio_clear_partially_mapped (folio); }spin_unlock (&ds_queue->split_queue_lock);if (mapping) {mm/huge_memory.c void __folio_undo_large_rmappable(struct folio *folio) · line 3458
Quoted 4 lines
mm/huge_memory.c +1
-0@@ -3454,11 +3455,13 @@ void __folio_undo_large_rmappable(struct folio *folio)if (!list_empty (&folio->_deferred_list)) { ds_queue->split_queue_len--;list_del_init (&folio->_deferred_list); +__folio_clear_partially_mapped (folio);is it possible to make things clearer by
Quoted 8 lines
if (
folio_clear_partially_mapped)__folio_clear_partially_mapped(folio);While writing without conditions isn't necessarily wrong, adding a condition will improve the readability of the code and enhance the clarity of my mTHP counters series. also help decrease smp cache sync if we can avoid unnecessary writing?
Quoted 3 lines
Do you mean if(folio_test_partially_mapped(folio))?
I don't like this idea. I think it makes the readability worse? If I was looking at if (test) -> clear for the first time, I would become confused why its being tested if its going to be clear at the end anyways?
Quoted 9 lines
In the pmd-order case, the majority of folios are not partially mapped. Unconditional writes will trigger cache synchronization across all CPUs (related to the MESI protocol), making them more costly. By using conditional writes, such as "if(test) write," we can avoid most unnecessary writes, which is much more efficient. Additionally, we only need to manage
nr_split_deferredwhen the condition is met. We are carefully evaluating all scenarios to determine if modifications to thepartially_mappedflag are necessary.
Hmm okay, as you said its needed for nr_split_deferred anyways. Something like below is ok to fold in?
commit 4ae9e2067346effd902b342296987b97dee29018 (HEAD)
Author: Usama Arif <[email protected]> Date: Mon Aug 19 21:07:16 2024 +0100
mm: Introduce a pageflag for partially mapped folios fix
Test partially_mapped flag before clearing it. This should avoid unnecessary writes and will be needed in the nr_split_deferred series.
mm/huge_memory.c +4 -2
@@ -3479,7 +3479,8 @@ void __folio_undo_large_rmappable(struct folio *folio)if (!list_empty (&folio->_deferred_list)) { ds_queue->split_queue_len--;list_del_init (&folio->_deferred_list); -__folio_clear_partially_mapped (folio); +if (folio_test_partially_mapped (folio)) +__folio_clear_partially_mapped (folio); }spin_unlock_irqrestore (&ds_queue->split_queue_lock, flags); }
@@ -3610,7 +3611,8 @@ static unsigned long deferred_split_scan(struct shrinker *shrink, }else {/* We lost race with folio_put() */ list_del_init (&folio->_deferred_list); -__folio_clear_partially_mapped (folio); +if (folio_test_partially_mapped (folio)) +__folio_clear_partially_mapped (folio); ds_queue->split_queue_len--; }if (!--sc->nr_to_scan)
Barry Song
reviewerRe: [PATCH v4 4/6] mm: Introduce a pageflag for partially mapped folios
Quoted 173 lines · Usama Arif
Quoted 128 lines · Barry Song
Quoted 123 lines · Usama Arif
Quoted 110 lines · Barry Song
Hi Usama,
I feel it is much better now! thanks!
Quoted 98 lines · Usama Arif
Currently
folio->_deferred_listis used to keep track ofpartially_mappedfolios that are going to be split under memory pressure. In the next patch, all THPs that are faulted in and collapsed by khugepaged are also going to be tracked using_deferred_list.This patch introduces a pageflag to be able to distinguish between partially mapped folios and others in the
deferred_listat split time indeferred_split_scan. Its needed as__folio_remove_rmapdecrements_mapcount,_large_mapcountand_entire_mapcount, hence it won't be possible to distinguish between partially mapped folios and others indeferred_split_scan.Eventhough it introduces an extra flag to track if the folio is partially mapped, there is no functional change intended with this patch and the flag is not useful in this patch itself, it will become useful in the next patch when
_deferred_listhas non partially mapped folios.Signed-off-by Usama Arif <[email protected]>--- include/linux/huge_mm.h | 4 ++-- include/linux/page-flags.h | 11 +++++++++++ mm/huge_memory.c | 23 ++++++++++++++++------- mm/internal.h | 4 +++- mm/memcontrol.c | 3 ++- mm/migrate.c | 3 ++- mm/page_alloc.c | 5 +++-- mm/rmap.c | 5 +++-- mm/vmscan.c | 3 ++- 9 files changed, 44 insertions(+), 17 deletions(-)include/linux/huge_mm.h +2
-2@@ -321,7 +321,7 @@ static inline int split_huge_page(struct page *page) {return split_huge_page_to_list_to_order (page,NULL ,0 ); } -void deferred_split_folio (struct folio *folio); +void deferred_split_folio (struct folio *folio,bool partially_mapped);void __split_huge_pmd (struct vm_area_struct *vma,pmd_t *pmd,unsigned long address,bool freeze,struct folio *folio);@@ -495,7 +495,7 @@ static inline int split_huge_page(struct page *page) {return 0 ; } -static inline void deferred_split_folio (struct folio *folio) {} +static inline void deferred_split_folio (struct folio *folio,bool partially_mapped) {}#define split_huge_pmd (__vma, __pmd, __address) \ do { } while (0)include/linux/page-flags.h +11
-0@@ -182,6 +182,7 @@ enum pageflags {/* At least one page in this folio has the hwpoison flag set */ PG_has_hwpoisoned = PG_active, PG_large_rmappable = PG_workingset,/* anon or file-backed */ + PG_partially_mapped = PG_reclaim, /* was identified to be partially mapped */ };#define PAGEFLAGS_MASK ((1UL << NR_PAGEFLAGS) - 1)@@ -861,8 +862,18 @@ static inline void ClearPageCompound(struct page *page)ClearPageHead (page); } FOLIO_FLAG(large_rmappable, FOLIO_SECOND_PAGE) +FOLIO_TEST_FLAG (partially_mapped,FOLIO_SECOND_PAGE ) +/* +* PG_partially_mapped is protected by deferred_split split_queue_lock, +* so its safe to use non-atomic set/clear. + */ +__FOLIO_SET_FLAG(partially_mapped, FOLIO_SECOND_PAGE) +__FOLIO_CLEAR_FLAG(partially_mapped, FOLIO_SECOND_PAGE)#else FOLIO_FLAG_FALSE(large_rmappable) +FOLIO_TEST_FLAG_FALSE(partially_mapped) +__FOLIO_SET_FLAG_NOOP(partially_mapped) +__FOLIO_CLEAR_FLAG_NOOP(partially_mapped)#endif #define PG_head_mask ((1UL << PG_head))mm/huge_memory.c +1
-0@@ -3398,6 +3398,7 @@ int split_huge_page_to_list_to_order(struct page *page, struct list_head *list,* page_deferred_list. */ list_del_init (&folio->_deferred_list); +__folio_clear_partially_mapped (folio); }spin_unlock (&ds_queue->split_queue_lock);if (mapping) {mm/huge_memory.c void __folio_undo_large_rmappable(struct folio *folio) · line 3458
Quoted 4 lines
mm/huge_memory.c +1
-0@@ -3454,11 +3455,13 @@ void __folio_undo_large_rmappable(struct folio *folio)if (!list_empty (&folio->_deferred_list)) { ds_queue->split_queue_len--;list_del_init (&folio->_deferred_list); +__folio_clear_partially_mapped (folio);is it possible to make things clearer by
Quoted 8 lines
if (
folio_clear_partially_mapped)__folio_clear_partially_mapped(folio);While writing without conditions isn't necessarily wrong, adding a condition will improve the readability of the code and enhance the clarity of my mTHP counters series. also help decrease smp cache sync if we can avoid unnecessary writing?
Quoted 3 lines
Do you mean if(folio_test_partially_mapped(folio))?
I don't like this idea. I think it makes the readability worse? If I was looking at if (test) -> clear for the first time, I would become confused why its being tested if its going to be clear at the end anyways?
Quoted 9 lines
In the pmd-order case, the majority of folios are not partially mapped. Unconditional writes will trigger cache synchronization across all CPUs (related to the MESI protocol), making them more costly. By using conditional writes, such as "if(test) write," we can avoid most unnecessary writes, which is much more efficient. Additionally, we only need to manage
nr_split_deferredwhen the condition is met. We are carefully evaluating all scenarios to determine if modifications to thepartially_mappedflag are necessary.Hmm okay, as you said its needed for
nr_split_deferredanyways. Something like below is ok to fold in?commit
4ae9e2067346effd902b342296987b97dee29018(HEAD)Author: Usama Arif <[email protected]> Date: Mon Aug 19 21:07:16 2024 +0100mm: Introduce a pageflag for partially mapped folios fix
Test
partially_mappedflag before clearing it. This should avoid unnecessary writes and will be needed in thenr_split_deferredseries.Signed-off-by Usama Arif <[email protected]>mm/huge_memory.c +2
-1@@ -3479,7 +3479,8 @@ void __folio_undo_large_rmappable(struct folio *folio)if (!list_empty (&folio->_deferred_list)) { ds_queue->split_queue_len--;list_del_init (&folio->_deferred_list); -__folio_clear_partially_mapped (folio); +if (folio_test_partially_mapped (folio)) +__folio_clear_partially_mapped (folio); }spin_unlock_irqrestore (&ds_queue->split_queue_lock, flags); }
mm/huge_memory.c static unsigned long deferred_split_scan(struct shrinker *shrink, · line 3618
Quoted 9 lines
mm/huge_memory.c +2
-1@@ -3610,7 +3611,8 @@ static unsigned long deferred_split_scan(struct shrinker *shrink, }else {/* We lost race with folio_put() */ list_del_init (&folio->_deferred_list); -__folio_clear_partially_mapped (folio); +if (folio_test_partially_mapped (folio)) +__folio_clear_partially_mapped (folio); ds_queue->split_queue_len--; }if (!--sc->nr_to_scan)
Do we also need if (folio_test_partially_mapped(folio)) in split_huge_page_to_list_to_order()?
I recall that in Yu Zhao's TAO, there’s a chance of splitting (shattering) non-partially-mapped folios. To be future-proof, we might want to handle both cases equally.
By the way, we might not need to clear the flag for a new folio. This differs from the init_list, which is necessary. If a new folio has the partially_mapped flag, it indicates that we failed to clear it when freeing the folio to the buddy system, which is a bug we need to fix in the free path.
Thanks
Barry
Barry Song
reviewerRe: [PATCH v4 4/6] mm: Introduce a pageflag for partially mapped folios
Quoted 188 lines · Barry Song
Quoted 173 lines · Usama Arif
Quoted 128 lines · Barry Song
Quoted 123 lines · Usama Arif
Quoted 110 lines · Barry Song
Hi Usama,
I feel it is much better now! thanks!
Quoted 98 lines · Usama Arif
Currently
folio->_deferred_listis used to keep track ofpartially_mappedfolios that are going to be split under memory pressure. In the next patch, all THPs that are faulted in and collapsed by khugepaged are also going to be tracked using_deferred_list.This patch introduces a pageflag to be able to distinguish between partially mapped folios and others in the
deferred_listat split time indeferred_split_scan. Its needed as__folio_remove_rmapdecrements_mapcount,_large_mapcountand_entire_mapcount, hence it won't be possible to distinguish between partially mapped folios and others indeferred_split_scan.Eventhough it introduces an extra flag to track if the folio is partially mapped, there is no functional change intended with this patch and the flag is not useful in this patch itself, it will become useful in the next patch when
_deferred_listhas non partially mapped folios.Signed-off-by Usama Arif <[email protected]>--- include/linux/huge_mm.h | 4 ++-- include/linux/page-flags.h | 11 +++++++++++ mm/huge_memory.c | 23 ++++++++++++++++------- mm/internal.h | 4 +++- mm/memcontrol.c | 3 ++- mm/migrate.c | 3 ++- mm/page_alloc.c | 5 +++-- mm/rmap.c | 5 +++-- mm/vmscan.c | 3 ++- 9 files changed, 44 insertions(+), 17 deletions(-)include/linux/huge_mm.h +2
-2@@ -321,7 +321,7 @@ static inline int split_huge_page(struct page *page) {return split_huge_page_to_list_to_order (page,NULL ,0 ); } -void deferred_split_folio (struct folio *folio); +void deferred_split_folio (struct folio *folio,bool partially_mapped);void __split_huge_pmd (struct vm_area_struct *vma,pmd_t *pmd,unsigned long address,bool freeze,struct folio *folio);@@ -495,7 +495,7 @@ static inline int split_huge_page(struct page *page) {return 0 ; } -static inline void deferred_split_folio (struct folio *folio) {} +static inline void deferred_split_folio (struct folio *folio,bool partially_mapped) {}#define split_huge_pmd (__vma, __pmd, __address) \ do { } while (0)include/linux/page-flags.h +11
-0@@ -182,6 +182,7 @@ enum pageflags {/* At least one page in this folio has the hwpoison flag set */ PG_has_hwpoisoned = PG_active, PG_large_rmappable = PG_workingset,/* anon or file-backed */ + PG_partially_mapped = PG_reclaim, /* was identified to be partially mapped */ };#define PAGEFLAGS_MASK ((1UL << NR_PAGEFLAGS) - 1)@@ -861,8 +862,18 @@ static inline void ClearPageCompound(struct page *page)ClearPageHead (page); } FOLIO_FLAG(large_rmappable, FOLIO_SECOND_PAGE) +FOLIO_TEST_FLAG (partially_mapped,FOLIO_SECOND_PAGE ) +/* +* PG_partially_mapped is protected by deferred_split split_queue_lock, +* so its safe to use non-atomic set/clear. + */ +__FOLIO_SET_FLAG(partially_mapped, FOLIO_SECOND_PAGE) +__FOLIO_CLEAR_FLAG(partially_mapped, FOLIO_SECOND_PAGE)#else FOLIO_FLAG_FALSE(large_rmappable) +FOLIO_TEST_FLAG_FALSE(partially_mapped) +__FOLIO_SET_FLAG_NOOP(partially_mapped) +__FOLIO_CLEAR_FLAG_NOOP(partially_mapped)#endif #define PG_head_mask ((1UL << PG_head))mm/huge_memory.c +1
-0@@ -3398,6 +3398,7 @@ int split_huge_page_to_list_to_order(struct page *page, struct list_head *list,* page_deferred_list. */ list_del_init (&folio->_deferred_list); +__folio_clear_partially_mapped (folio); }spin_unlock (&ds_queue->split_queue_lock);if (mapping) {mm/huge_memory.c void __folio_undo_large_rmappable(struct folio *folio) · line 3458
Quoted 4 lines
mm/huge_memory.c +1
-0@@ -3454,11 +3455,13 @@ void __folio_undo_large_rmappable(struct folio *folio)if (!list_empty (&folio->_deferred_list)) { ds_queue->split_queue_len--;list_del_init (&folio->_deferred_list); +__folio_clear_partially_mapped (folio);is it possible to make things clearer by
Quoted 8 lines
if (
folio_clear_partially_mapped)__folio_clear_partially_mapped(folio);While writing without conditions isn't necessarily wrong, adding a condition will improve the readability of the code and enhance the clarity of my mTHP counters series. also help decrease smp cache sync if we can avoid unnecessary writing?
Quoted 3 lines
Do you mean if(folio_test_partially_mapped(folio))?
I don't like this idea. I think it makes the readability worse? If I was looking at if (test) -> clear for the first time, I would become confused why its being tested if its going to be clear at the end anyways?
Quoted 9 lines
In the pmd-order case, the majority of folios are not partially mapped. Unconditional writes will trigger cache synchronization across all CPUs (related to the MESI protocol), making them more costly. By using conditional writes, such as "if(test) write," we can avoid most unnecessary writes, which is much more efficient. Additionally, we only need to manage
nr_split_deferredwhen the condition is met. We are carefully evaluating all scenarios to determine if modifications to thepartially_mappedflag are necessary.Hmm okay, as you said its needed for
nr_split_deferredanyways. Something like below is ok to fold in?commit
4ae9e2067346effd902b342296987b97dee29018(HEAD)Author: Usama Arif <[email protected]> Date: Mon Aug 19 21:07:16 2024 +0100mm: Introduce a pageflag for partially mapped folios fix
Test
partially_mappedflag before clearing it. This should avoid unnecessary writes and will be needed in thenr_split_deferredseries.Signed-off-by Usama Arif <[email protected]>mm/huge_memory.c +2
-1@@ -3479,7 +3479,8 @@ void __folio_undo_large_rmappable(struct folio *folio)if (!list_empty (&folio->_deferred_list)) { ds_queue->split_queue_len--;list_del_init (&folio->_deferred_list); -__folio_clear_partially_mapped (folio); +if (folio_test_partially_mapped (folio)) +__folio_clear_partially_mapped (folio); }spin_unlock_irqrestore (&ds_queue->split_queue_lock, flags); }mm/huge_memory.c static unsigned long deferred_split_scan(struct shrinker *shrink, · line 3618
Quoted 9 lines
mm/huge_memory.c +2
-1@@ -3610,7 +3611,8 @@ static unsigned long deferred_split_scan(struct shrinker *shrink, }else {/* We lost race with folio_put() */ list_del_init (&folio->_deferred_list); -__folio_clear_partially_mapped (folio); +if (folio_test_partially_mapped (folio)) +__folio_clear_partially_mapped (folio); ds_queue->split_queue_len--; }if (!--sc->nr_to_scan)Do we also need if (
folio_test_partially_mapped(folio)) insplit_huge_page_to_list_to_order()?
Quoted 3 lines
I recall that in Yu Zhao's TAO, there’s a chance of splitting (shattering) non-partially-mapped folios. To be future-proof, we might want to handle both cases equally.
we recall we also have a real case which can split entirely_mapped folio:
mm: huge_memory: enable debugfs to split huge pages to any order
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=fc4d182316bd5309b4066fd9ef21529ea397a7d4
Quoted 8 lines
By the way, we might not need to clear the flag for a new folio. This differs from the
init_list, which is necessary. If a new folio has thepartially_mappedflag, it indicates that we failed to clear it when freeing the folio to the buddy system, which is a bug we need to fix in the free path.Thanks
Barry
Re: [PATCH v4 4/6] mm: Introduce a pageflag for partially mapped folios
Quoted 193 lines · Barry Song
Quoted 188 lines · Barry Song
Quoted 173 lines · Usama Arif
Quoted 128 lines · Barry Song
Quoted 123 lines · Usama Arif
Quoted 110 lines · Barry Song
Hi Usama,
I feel it is much better now! thanks!
Quoted 98 lines · Usama Arif
Currently
folio->_deferred_listis used to keep track ofpartially_mappedfolios that are going to be split under memory pressure. In the next patch, all THPs that are faulted in and collapsed by khugepaged are also going to be tracked using_deferred_list.This patch introduces a pageflag to be able to distinguish between partially mapped folios and others in the
deferred_listat split time indeferred_split_scan. Its needed as__folio_remove_rmapdecrements_mapcount,_large_mapcountand_entire_mapcount, hence it won't be possible to distinguish between partially mapped folios and others indeferred_split_scan.Eventhough it introduces an extra flag to track if the folio is partially mapped, there is no functional change intended with this patch and the flag is not useful in this patch itself, it will become useful in the next patch when
_deferred_listhas non partially mapped folios.Signed-off-by Usama Arif <[email protected]>--- include/linux/huge_mm.h | 4 ++-- include/linux/page-flags.h | 11 +++++++++++ mm/huge_memory.c | 23 ++++++++++++++++------- mm/internal.h | 4 +++- mm/memcontrol.c | 3 ++- mm/migrate.c | 3 ++- mm/page_alloc.c | 5 +++-- mm/rmap.c | 5 +++-- mm/vmscan.c | 3 ++- 9 files changed, 44 insertions(+), 17 deletions(-)include/linux/huge_mm.h +2
-2@@ -321,7 +321,7 @@ static inline int split_huge_page(struct page *page) {return split_huge_page_to_list_to_order (page,NULL ,0 ); } -void deferred_split_folio (struct folio *folio); +void deferred_split_folio (struct folio *folio,bool partially_mapped);void __split_huge_pmd (struct vm_area_struct *vma,pmd_t *pmd,unsigned long address,bool freeze,struct folio *folio);@@ -495,7 +495,7 @@ static inline int split_huge_page(struct page *page) {return 0 ; } -static inline void deferred_split_folio (struct folio *folio) {} +static inline void deferred_split_folio (struct folio *folio,bool partially_mapped) {}#define split_huge_pmd (__vma, __pmd, __address) \ do { } while (0)include/linux/page-flags.h +11
-0@@ -182,6 +182,7 @@ enum pageflags {/* At least one page in this folio has the hwpoison flag set */ PG_has_hwpoisoned = PG_active, PG_large_rmappable = PG_workingset,/* anon or file-backed */ + PG_partially_mapped = PG_reclaim, /* was identified to be partially mapped */ };#define PAGEFLAGS_MASK ((1UL << NR_PAGEFLAGS) - 1)@@ -861,8 +862,18 @@ static inline void ClearPageCompound(struct page *page)ClearPageHead (page); } FOLIO_FLAG(large_rmappable, FOLIO_SECOND_PAGE) +FOLIO_TEST_FLAG (partially_mapped,FOLIO_SECOND_PAGE ) +/* +* PG_partially_mapped is protected by deferred_split split_queue_lock, +* so its safe to use non-atomic set/clear. + */ +__FOLIO_SET_FLAG(partially_mapped, FOLIO_SECOND_PAGE) +__FOLIO_CLEAR_FLAG(partially_mapped, FOLIO_SECOND_PAGE)#else FOLIO_FLAG_FALSE(large_rmappable) +FOLIO_TEST_FLAG_FALSE(partially_mapped) +__FOLIO_SET_FLAG_NOOP(partially_mapped) +__FOLIO_CLEAR_FLAG_NOOP(partially_mapped)#endif #define PG_head_mask ((1UL << PG_head))mm/huge_memory.c +1
-0@@ -3398,6 +3398,7 @@ int split_huge_page_to_list_to_order(struct page *page, struct list_head *list,* page_deferred_list. */ list_del_init (&folio->_deferred_list); +__folio_clear_partially_mapped (folio); }spin_unlock (&ds_queue->split_queue_lock);if (mapping) {mm/huge_memory.c void __folio_undo_large_rmappable(struct folio *folio) · line 3458
Quoted 4 lines
mm/huge_memory.c +1
-0@@ -3454,11 +3455,13 @@ void __folio_undo_large_rmappable(struct folio *folio)if (!list_empty (&folio->_deferred_list)) { ds_queue->split_queue_len--;list_del_init (&folio->_deferred_list); +__folio_clear_partially_mapped (folio);is it possible to make things clearer by
Quoted 8 lines
if (
folio_clear_partially_mapped)__folio_clear_partially_mapped(folio);While writing without conditions isn't necessarily wrong, adding a condition will improve the readability of the code and enhance the clarity of my mTHP counters series. also help decrease smp cache sync if we can avoid unnecessary writing?
Quoted 3 lines
Do you mean if(folio_test_partially_mapped(folio))?
I don't like this idea. I think it makes the readability worse? If I was looking at if (test) -> clear for the first time, I would become confused why its being tested if its going to be clear at the end anyways?
Quoted 9 lines
In the pmd-order case, the majority of folios are not partially mapped. Unconditional writes will trigger cache synchronization across all CPUs (related to the MESI protocol), making them more costly. By using conditional writes, such as "if(test) write," we can avoid most unnecessary writes, which is much more efficient. Additionally, we only need to manage
nr_split_deferredwhen the condition is met. We are carefully evaluating all scenarios to determine if modifications to thepartially_mappedflag are necessary.Hmm okay, as you said its needed for
nr_split_deferredanyways. Something like below is ok to fold in?commit
4ae9e2067346effd902b342296987b97dee29018(HEAD)Author: Usama Arif <[email protected]> Date: Mon Aug 19 21:07:16 2024 +0100mm: Introduce a pageflag for partially mapped folios fix
Test
partially_mappedflag before clearing it. This should avoid unnecessary writes and will be needed in thenr_split_deferredseries.Signed-off-by Usama Arif <[email protected]>mm/huge_memory.c +2
-1@@ -3479,7 +3479,8 @@ void __folio_undo_large_rmappable(struct folio *folio)if (!list_empty (&folio->_deferred_list)) { ds_queue->split_queue_len--;list_del_init (&folio->_deferred_list); -__folio_clear_partially_mapped (folio); +if (folio_test_partially_mapped (folio)) +__folio_clear_partially_mapped (folio); }spin_unlock_irqrestore (&ds_queue->split_queue_lock, flags); }mm/huge_memory.c static unsigned long deferred_split_scan(struct shrinker *shrink, · line 3618
Quoted 9 lines
mm/huge_memory.c +2
-1@@ -3610,7 +3611,8 @@ static unsigned long deferred_split_scan(struct shrinker *shrink, }else {/* We lost race with folio_put() */ list_del_init (&folio->_deferred_list); -__folio_clear_partially_mapped (folio); +if (folio_test_partially_mapped (folio)) +__folio_clear_partially_mapped (folio); ds_queue->split_queue_len--; }if (!--sc->nr_to_scan)Do we also need if (
folio_test_partially_mapped(folio)) insplit_huge_page_to_list_to_order()?Quoted 3 lines
I recall that in Yu Zhao's TAO, there’s a chance of splitting (shattering) non-partially-mapped folios. To be future-proof, we might want to handle both cases equally.
Quoted 14 lines
we recall we also have a real case which can split
entirely_mappedfolio:mm:
huge_memory: enable debugfs to split huge pages to any order
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=fc4d182316bd5309b4066fd9ef21529ea397a7d4Quoted 8 lines
By the way, we might not need to clear the flag for a new folio. This differs from the
init_list, which is necessary. If a new folio has thepartially_mappedflag, it indicates that we failed to clear it when freeing the folio to the buddy system, which is a bug we need to fix in the free path.Thanks
Barry
I believe the below fixlet should address all concerns:
From 95492a51b1929ea274b4e5b78fc74e7736645d58 Mon Sep 17 00:00:00 2001
From: Usama Arif <[email protected]>
Date: Mon, 19 Aug 2024 21:07:16 +0100
Subject: [PATCH] mm: Introduce a pageflag for partially mapped folios fix
Test partially_mapped flag before clearing it. This should avoid unnecessary writes and will be needed in the nr_split_deferred series.
Also no need to clear partially_mapped prepping compound head, as it should start with already being cleared.
--- include/linux/page-flags.h | 2 +- mm/huge_memory.c | 9 ++++++--- mm/internal.h | 4 +--- 3 files changed, 8 insertions(+), 7 deletions(-)
include/linux/page-flags.h +1 -1
@@ -1182,7 +1182,7 @@ static __always_inline void __ClearPageAnonExclusive(struct page *page)*/ #define PAGE_FLAGS_SECOND \ (0xffUL /* order */ | 1UL << PG_has_hwpoisoned | \ - 1UL << PG_large_rmappable) + 1UL << PG_large_rmappable | 1UL << PG_partially_mapped)#define PAGE_FLAGS_PRIVATE \ (1UL << PG_private | 1UL << PG_private_2)
mm/huge_memory.c +6 -3
@@ -3422,7 +3422,8 @@ int split_huge_page_to_list_to_order(struct page *page, struct list_head *list,* page_deferred_list. */ list_del_init (&folio->_deferred_list); -__folio_clear_partially_mapped (folio); +if (folio_test_partially_mapped (folio)) +__folio_clear_partially_mapped (folio); }spin_unlock (&ds_queue->split_queue_lock);if (mapping) {
@@ -3479,7 +3480,8 @@ void __folio_undo_large_rmappable(struct folio *folio)if (!list_empty (&folio->_deferred_list)) { ds_queue->split_queue_len--;list_del_init (&folio->_deferred_list); -__folio_clear_partially_mapped (folio); +if (folio_test_partially_mapped (folio)) +__folio_clear_partially_mapped (folio); }spin_unlock_irqrestore (&ds_queue->split_queue_lock, flags); }
@@ -3610,7 +3612,8 @@ static unsigned long deferred_split_scan(struct shrinker *shrink, }else {/* We lost race with folio_put() */ list_del_init (&folio->_deferred_list); -__folio_clear_partially_mapped (folio); +if (folio_test_partially_mapped (folio)) +__folio_clear_partially_mapped (folio); ds_queue->split_queue_len--; }if (!--sc->nr_to_scan)
mm/internal.h +1 -3
@@ -662,10 +662,8 @@ static inline void prep_compound_head(struct page *page, unsigned int order)atomic_set (&folio->_entire_mapcount,-1 );atomic_set (&folio->_nr_pages_mapped,0 );atomic_set (&folio->_pincount,0 ); -if (order >1 ) { +if (order >1 )INIT_LIST_HEAD (&folio->_deferred_list); -__folio_clear_partially_mapped (folio); - } }static inline void prep_compound_tail (struct page *head,int tail_idx)
Signature · 1 line
2.43.5
Barry Song
reviewerRe: [PATCH v4 4/6] mm: Introduce a pageflag for partially mapped folios
Quoted 209 lines · Usama Arif
Quoted 205 lines · Barry Song
Quoted 188 lines · Barry Song
Quoted 173 lines · Usama Arif
Quoted 128 lines · Barry Song
Quoted 123 lines · Usama Arif
Quoted 110 lines · Barry Song
Hi Usama,
I feel it is much better now! thanks!
Quoted 98 lines · Usama Arif
Currently
folio->_deferred_listis used to keep track ofpartially_mappedfolios that are going to be split under memory pressure. In the next patch, all THPs that are faulted in and collapsed by khugepaged are also going to be tracked using_deferred_list.This patch introduces a pageflag to be able to distinguish between partially mapped folios and others in the
deferred_listat split time indeferred_split_scan. Its needed as__folio_remove_rmapdecrements_mapcount,_large_mapcountand_entire_mapcount, hence it won't be possible to distinguish between partially mapped folios and others indeferred_split_scan.Eventhough it introduces an extra flag to track if the folio is partially mapped, there is no functional change intended with this patch and the flag is not useful in this patch itself, it will become useful in the next patch when
_deferred_listhas non partially mapped folios.Signed-off-by Usama Arif <[email protected]>--- include/linux/huge_mm.h | 4 ++-- include/linux/page-flags.h | 11 +++++++++++ mm/huge_memory.c | 23 ++++++++++++++++------- mm/internal.h | 4 +++- mm/memcontrol.c | 3 ++- mm/migrate.c | 3 ++- mm/page_alloc.c | 5 +++-- mm/rmap.c | 5 +++-- mm/vmscan.c | 3 ++- 9 files changed, 44 insertions(+), 17 deletions(-)include/linux/huge_mm.h +2
-2@@ -321,7 +321,7 @@ static inline int split_huge_page(struct page *page) {return split_huge_page_to_list_to_order (page,NULL ,0 ); } -void deferred_split_folio (struct folio *folio); +void deferred_split_folio (struct folio *folio,bool partially_mapped);void __split_huge_pmd (struct vm_area_struct *vma,pmd_t *pmd,unsigned long address,bool freeze,struct folio *folio);@@ -495,7 +495,7 @@ static inline int split_huge_page(struct page *page) {return 0 ; } -static inline void deferred_split_folio (struct folio *folio) {} +static inline void deferred_split_folio (struct folio *folio,bool partially_mapped) {}#define split_huge_pmd (__vma, __pmd, __address) \ do { } while (0)include/linux/page-flags.h +11
-0@@ -182,6 +182,7 @@ enum pageflags {/* At least one page in this folio has the hwpoison flag set */ PG_has_hwpoisoned = PG_active, PG_large_rmappable = PG_workingset,/* anon or file-backed */ + PG_partially_mapped = PG_reclaim, /* was identified to be partially mapped */ };#define PAGEFLAGS_MASK ((1UL << NR_PAGEFLAGS) - 1)@@ -861,8 +862,18 @@ static inline void ClearPageCompound(struct page *page)ClearPageHead (page); } FOLIO_FLAG(large_rmappable, FOLIO_SECOND_PAGE) +FOLIO_TEST_FLAG (partially_mapped,FOLIO_SECOND_PAGE ) +/* +* PG_partially_mapped is protected by deferred_split split_queue_lock, +* so its safe to use non-atomic set/clear. + */ +__FOLIO_SET_FLAG(partially_mapped, FOLIO_SECOND_PAGE) +__FOLIO_CLEAR_FLAG(partially_mapped, FOLIO_SECOND_PAGE)#else FOLIO_FLAG_FALSE(large_rmappable) +FOLIO_TEST_FLAG_FALSE(partially_mapped) +__FOLIO_SET_FLAG_NOOP(partially_mapped) +__FOLIO_CLEAR_FLAG_NOOP(partially_mapped)#endif #define PG_head_mask ((1UL << PG_head))mm/huge_memory.c +1
-0@@ -3398,6 +3398,7 @@ int split_huge_page_to_list_to_order(struct page *page, struct list_head *list,* page_deferred_list. */ list_del_init (&folio->_deferred_list); +__folio_clear_partially_mapped (folio); }spin_unlock (&ds_queue->split_queue_lock);if (mapping) {mm/huge_memory.c void __folio_undo_large_rmappable(struct folio *folio) · line 3458
Quoted 4 lines
mm/huge_memory.c +1
-0@@ -3454,11 +3455,13 @@ void __folio_undo_large_rmappable(struct folio *folio)if (!list_empty (&folio->_deferred_list)) { ds_queue->split_queue_len--;list_del_init (&folio->_deferred_list); +__folio_clear_partially_mapped (folio);is it possible to make things clearer by
Quoted 8 lines
if (
folio_clear_partially_mapped)__folio_clear_partially_mapped(folio);While writing without conditions isn't necessarily wrong, adding a condition will improve the readability of the code and enhance the clarity of my mTHP counters series. also help decrease smp cache sync if we can avoid unnecessary writing?
Quoted 3 lines
Do you mean if(folio_test_partially_mapped(folio))?
I don't like this idea. I think it makes the readability worse? If I was looking at if (test) -> clear for the first time, I would become confused why its being tested if its going to be clear at the end anyways?
Quoted 9 lines
In the pmd-order case, the majority of folios are not partially mapped. Unconditional writes will trigger cache synchronization across all CPUs (related to the MESI protocol), making them more costly. By using conditional writes, such as "if(test) write," we can avoid most unnecessary writes, which is much more efficient. Additionally, we only need to manage
nr_split_deferredwhen the condition is met. We are carefully evaluating all scenarios to determine if modifications to thepartially_mappedflag are necessary.Hmm okay, as you said its needed for
nr_split_deferredanyways. Something like below is ok to fold in?commit
4ae9e2067346effd902b342296987b97dee29018(HEAD)Author: Usama Arif <[email protected]> Date: Mon Aug 19 21:07:16 2024 +0100mm: Introduce a pageflag for partially mapped folios fix
Test
partially_mappedflag before clearing it. This should avoid unnecessary writes and will be needed in thenr_split_deferredseries.Signed-off-by Usama Arif <[email protected]>mm/huge_memory.c +2
-1@@ -3479,7 +3479,8 @@ void __folio_undo_large_rmappable(struct folio *folio)if (!list_empty (&folio->_deferred_list)) { ds_queue->split_queue_len--;list_del_init (&folio->_deferred_list); -__folio_clear_partially_mapped (folio); +if (folio_test_partially_mapped (folio)) +__folio_clear_partially_mapped (folio); }spin_unlock_irqrestore (&ds_queue->split_queue_lock, flags); }mm/huge_memory.c static unsigned long deferred_split_scan(struct shrinker *shrink, · line 3618
Quoted 9 lines
mm/huge_memory.c +2
-1@@ -3610,7 +3611,8 @@ static unsigned long deferred_split_scan(struct shrinker *shrink, }else {/* We lost race with folio_put() */ list_del_init (&folio->_deferred_list); -__folio_clear_partially_mapped (folio); +if (folio_test_partially_mapped (folio)) +__folio_clear_partially_mapped (folio); ds_queue->split_queue_len--; }if (!--sc->nr_to_scan)Do we also need if (
folio_test_partially_mapped(folio)) insplit_huge_page_to_list_to_order()?Quoted 3 lines
I recall that in Yu Zhao's TAO, there’s a chance of splitting (shattering) non-partially-mapped folios. To be future-proof, we might want to handle both cases equally.
we recall we also have a real case which can split
entirely_mappedfolio:mm:
huge_memory: enable debugfs to split huge pages to any order
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=fc4d182316bd5309b4066fd9ef21529ea397a7d4Quoted 6 lines
By the way, we might not need to clear the flag for a new folio. This differs from the
init_list, which is necessary. If a new folio has thepartially_mappedflag, it indicates that we failed to clear it when freeing the folio to the buddy system, which is a bug we need to fix in the free path.
Quoted 4 lines
Quoted 2 lines
Quoted 2 lines
Thanks
BarryI believe the below fixlet should address all concerns:
Hi Usama,
thanks! I can't judge if we need this partially_mapped flag. but if we need, the code
looks correct to me. I'd like to leave this to David and other experts to ack.
an alternative approach might be two lists? one for entirely_mapped, the other one
for split_deferred. also seems ugly ?
On the other hand, when we want to extend your patchset to mTHP other than PMD- order, will the only deferred_list create huge lock contention while adding or removing
folios from it?
Quoted 84 lines
From
95492a51b1929ea274b4e5b78fc74e7736645d58Mon Sep 17 00:00:00 2001
From: Usama Arif <[email protected]>
Date: Mon, 19 Aug 2024 21:07:16 +0100
Subject: [PATCH] mm: Introduce a pageflag for partially mapped folios fixTest
partially_mappedflag before clearing it. This should avoid unnecessary writes and will be needed in thenr_split_deferredseries.
Also no need to clearpartially_mappedprepping compound head, as it should start with already being cleared.Signed-off-by Usama Arif <[email protected]>--- include/linux/page-flags.h | 2 +- mm/huge_memory.c | 9 ++++++--- mm/internal.h | 4 +--- 3 files changed, 8 insertions(+), 7 deletions(-)include/linux/page-flags.h +1
-1@@ -1182,7 +1182,7 @@ static __always_inline void __ClearPageAnonExclusive(struct page *page)*/ #define PAGE_FLAGS_SECOND \ (0xffUL /* order */ | 1UL << PG_has_hwpoisoned | \ - 1UL << PG_large_rmappable) + 1UL << PG_large_rmappable | 1UL << PG_partially_mapped)#define PAGE_FLAGS_PRIVATE \ (1UL << PG_private | 1UL << PG_private_2)mm/huge_memory.c +6
-3@@ -3422,7 +3422,8 @@ int split_huge_page_to_list_to_order(struct page *page, struct list_head *list,* page_deferred_list. */ list_del_init (&folio->_deferred_list); -__folio_clear_partially_mapped (folio); +if (folio_test_partially_mapped (folio)) +__folio_clear_partially_mapped (folio); }spin_unlock (&ds_queue->split_queue_lock);if (mapping) {@@ -3479,7 +3480,8 @@ void __folio_undo_large_rmappable(struct folio *folio)if (!list_empty (&folio->_deferred_list)) { ds_queue->split_queue_len--;list_del_init (&folio->_deferred_list); -__folio_clear_partially_mapped (folio); +if (folio_test_partially_mapped (folio)) +__folio_clear_partially_mapped (folio); }spin_unlock_irqrestore (&ds_queue->split_queue_lock, flags); }@@ -3610,7 +3612,8 @@ static unsigned long deferred_split_scan(struct shrinker *shrink, }else {/* We lost race with folio_put() */ list_del_init (&folio->_deferred_list); -__folio_clear_partially_mapped (folio); +if (folio_test_partially_mapped (folio)) +__folio_clear_partially_mapped (folio); ds_queue->split_queue_len--; }if (!--sc->nr_to_scan)mm/internal.h +1
-4@@ -662,10 +662,8 @@ static inline void prep_compound_head(struct page *page, unsigned int order)atomic_set (&folio->_entire_mapcount,-1 );atomic_set (&folio->_nr_pages_mapped,0 );atomic_set (&folio->_pincount,0 ); -if (order >1 ) { +if (order >1 )INIT_LIST_HEAD (&folio->_deferred_list); -__folio_clear_partially_mapped (folio); - } }static inline void prep_compound_tail (struct page *head,int tail_idx) --
Quoted 3 lines
2.43.5
Thanks
Barry
Re: [PATCH v4 4/6] mm: Introduce a pageflag for partially mapped folios
Quoted 5 lines · Barry Song
Hi Usama,
thanks! I can't judge if we need thispartially_mappedflag. but if we need, the code
looks correct to me. I'd like to leave this to David and other experts to ack.
Thanks for the reviews!
Quoted 4 lines
an alternative approach might be two lists? one for
entirely_mapped, the other one
forsplit_deferred. also seems ugly ?
That was my very first prototype! I shifted to using a bool which I sent in v1, and then a bit in _flags_1 as David suggested. I believe a bit in _flags_1 is the best way forward, as it leaves the most space in folio for future work.
Quoted 5 lines
On the other hand, when we want to extend your patchset to mTHP other than PMD- order, will the only
deferred_listcreate huge lock contention while adding or removing
folios from it?
Yes, I would imagine so. the deferred_split_queue is per memcg/node, so that helps.
Also, this work is tied to khugepaged. So would need some thought when doing it for mTHP.
I would imagine doing underused shrinker for mTHP would be less beneficial compared to doing it for 2M THP. But probably needs experimentation.
Thanks
Barry Song
reviewerRe: [PATCH v4 4/6] mm: Introduce a pageflag for partially mapped folios
Quoted 11 lines · Usama Arif
Quoted 5 lines · Barry Song
Hi Usama,
thanks! I can't judge if we need thispartially_mappedflag. but if we need, the code
looks correct to me. I'd like to leave this to David and other experts to ack.
Quoted 9 lines
Thanks for the reviews!
Quoted 4 lines
an alternative approach might be two lists? one for
entirely_mapped, the other one
forsplit_deferred. also seems ugly ?That was my very first prototype! I shifted to using a bool which I sent in v1, and then a bit in
_flags_1as David suggested. I believe a bit in_flags_1is the best way forward, as it leaves the most space in folio for future work.
Thanks for the explanation; I missed the earlier discussion.
Quoted 8 lines
Quoted 5 lines
On the other hand, when we want to extend your patchset to mTHP other than PMD- order, will the only
deferred_listcreate huge lock contention while adding or removing
folios from it?Yes, I would imagine so. the
deferred_split_queueis per memcg/node, so that helps.
Right.
Quoted 4 lines
Also, this work is tied to khugepaged. So would need some thought when doing it for mTHP.
I would imagine doing underused shrinker for mTHP would be less beneficial compared to doing it for 2M THP. But probably needs experimentation.
Agreed. I suspect this will be also useful for larger mTHPs like 1M and 512KB. While lock contention might be a concern, it shouldn't be a barrier to proceeding
with your patchset, which only supports PMD-order.
Quoted 2 lines
Thanks
Thanks
Barry
Re: [PATCH v4 4/6] mm: Introduce a pageflag for partially mapped folios
In reply to Usama Arifhi, Usama
Quoted 21 lines · Usama Arif
Quoted 5 lines · Barry Song
Hi Usama,
thanks! I can't judge if we need thispartially_mappedflag. but if we need, the code
looks correct to me. I'd like to leave this to David and other experts to ack.Thanks for the reviews!
Quoted 4 lines
an alternative approach might be two lists? one for
entirely_mapped, the other one
forsplit_deferred. also seems ugly ?That was my very first prototype! I shifted to using a bool which I sent in v1, and then a bit in
_flags_1as David suggested. I believe a bit in_flags_1is the best way forward, as it leaves the most space in folio for future work.Quoted 5 lines
On the other hand, when we want to extend your patchset to mTHP other than PMD- order, will the only
deferred_listcreate huge lock contention while adding or removing
folios from it?
Quoted 7 lines
Yes, I would imagine so. the
deferred_split_queueis per memcg/node, so that helps.Also, this work is tied to khugepaged. So would need some thought when doing it for mTHP.
I would imagine doing underused shrinker for mTHP would be less beneficial compared to doing it for 2M THP. But probably needs experimentation.
Thanks
Below is the core code snippet to support "split underused mTHP". Can we extend the khugepaged_max_ptes_none value to mthp and keep its semantics unchanged? With a small modification, Only folios with page numbers greater than khugepaged_max_ptes_none - 1 can be added to the deferred_split list and can be split. What do you think?
mm/memory.c +2 -0
@@ -4789,6 +4789,8 @@ static vm_fault_t do_anonymous_page(struct vm_fault *vmf) }folio_ref_add (folio, nr_pages -1 ); +if (nr_pages >1 && nr_pages > khugepaged_max_ptes_none -1 ) +deferred_split_folio (folio, false);add_mm_counter (vma->vm_mm,MM_ANONPAGES , nr_pages);count_mthp_stat (folio_order (folio),MTHP_STAT_ANON_FAULT_ALLOC );folio_add_new_anon_rmap (folio, vma, addr,RMAP_EXCLUSIVE );
shmem THP has the same memory expansion problem when the shmem_enabled configuration is set to always. In my opinion, it is necessary to support "split underused shmem THP", but I am not sure if there is any gap in the implementation?
Bang
Thanks
Re: [PATCH v4 4/6] mm: Introduce a pageflag for partially mapped folios
Quoted 29 lines · Bang Li
hi, Usama
Quoted 25 lines · Usama Arif
Quoted 5 lines · Barry Song
Hi Usama,
thanks! I can't judge if we need thispartially_mappedflag. but if we need, the code
looks correct to me. I'd like to leave this to David and other experts to ack.Thanks for the reviews!
Quoted 4 lines
an alternative approach might be two lists? one for
entirely_mapped, the other one
forsplit_deferred. also seems ugly ?That was my very first prototype! I shifted to using a bool which I sent in v1, and then a bit in
_flags_1as David suggested. I believe a bit in_flags_1is the best way forward, as it leaves the most space in folio for future work.Quoted 5 lines
On the other hand, when we want to extend your patchset to mTHP other than PMD- order, will the only
deferred_listcreate huge lock contention while adding or removing
folios from it?Yes, I would imagine so. the
deferred_split_queueis per memcg/node, so that helps.Also, this work is tied to khugepaged. So would need some thought when doing it for mTHP.
Quoted 9 lines
Quoted 3 lines
I would imagine doing underused shrinker for mTHP would be less beneficial compared to doing it for 2M THP. But probably needs experimentation.
Thanks
Below is the core code snippet to support "split underused mTHP". Can we extend the
khugepaged_max_ptes_nonevalue to mthp and keep its semantics unchanged? With a small modification, Only folios with page numbers greater thankhugepaged_max_ptes_none- 1 can be added to thedeferred_splitlist and can be split. What do you think?
hmm, so I believe its not as simple as that.
First mTHP support would need to be added to khugepaged. The entire khugepaged code needs to be audited for it and significantly tested. If you just look at all the instances of HPAGE_PMD_NR in khugepaged.c, you will see it will be a significant change and needs to be a series of its own.
Also, different values of max_ptes_none can have different significance for mTHP sizes. max_ptes_none of 200 does not mean anything for 512K and below, it means 78% of a 1M mTHP and 39% of a 2M THP. We might want different max_ptes_none for each mTHP if we were to do this.
The benefit of splitting underused mTHPs of lower order might not be worth the work needed to split. Someone needs to experiment with different workloads and see some improvement for these lower orders.
So probably a lot more than the below diff is needed for mTHP support!
Quoted 21 lines
mm/memory.c +2
-0@@ -4789,6 +4789,8 @@ static vm_fault_t do_anonymous_page(struct vm_fault *vmf) } folio_ref_add(folio, nr_pages - 1); + if (nr_pages > 1 && nr_pages > khugepaged_max_ptes_none - 1) + deferred_split_folio(folio, false); add_mm_counter(vma->vm_mm, MM_ANONPAGES, nr_pages); count_mthp_stat(folio_order(folio), MTHP_STAT_ANON_FAULT_ALLOC); folio_add_new_anon_rmap(folio, vma, addr, RMAP_EXCLUSIVE);shmem THP has the same memory expansion problem when the
shmem_enabledconfiguration is set to always. In my opinion, it is necessary to support "split underused shmem THP", but I am not sure if there is any gap in the implementation?Bang
Thanks
PATCH v4 5/6 mm: split underused THPs
In reply to Usama ArifThis is an attempt to mitigate the issue of running out of memory when THP is always enabled. During runtime whenever a THP is being faulted in (__do_huge_pmd_anonymous_page) or collapsed by khugepaged (collapse_huge_page), the THP is added to _deferred_list. Whenever memory reclaim happens in linux, the kernel runs the deferred_split shrinker which goes through the _deferred_list.
If the folio was partially mapped, the shrinker attempts to split it. If the folio is not partially mapped, the shrinker checks if the THP was underused, i.e. how many of the base 4K pages of the entire THP were zero-filled. If this number goes above a certain threshold (decided by /sys/kernel/mm/transparent_hugepage/khugepaged/max_ptes_none), the shrinker will attempt to split that THP. Then at remap time, the pages that were zero-filled are mapped to the shared zeropage, hence saving memory.
--- Documentation/admin-guide/mm/transhuge.rst | 6 +++ include/linux/khugepaged.h | 1 + include/linux/vm_event_item.h | 1 + mm/huge_memory.c | 60 +++++++++++++++++++++- mm/khugepaged.c | 3 +- mm/vmstat.c | 1 + 6 files changed, 69 insertions(+), 3 deletions(-)
Documentation/admin-guide/mm/transhuge.rst +6 -0
@@ -447,6 +447,12 @@ thp_deferred_split_page splitting it would free up some memory. Pages on split queue are going to be split under memory pressure. +thp_underused_split_page + is incremented when a huge page on the split queue was split + because it was underused. A THP is underused if the number of + zero pages in the THP is above a certain threshold + (/sys/kernel/mm/transparent_hugepage/khugepaged/max_ptes_none). + thp_split_pmd is incremented every time a PMD split into table of PTEs. This can happen, for instance, when application calls mprotect() or
include/linux/khugepaged.h +1 -0
@@ -4,6 +4,7 @@#include <linux/sched/coredump.h> /* MMF_VM_HUGEPAGE */ +extern unsigned int khugepaged_max_ptes_none__read_mostly ;#ifdef CONFIG_TRANSPARENT_HUGEPAGE extern struct attribute_group khugepaged_attr_group;
include/linux/vm_event_item.h +1 -0
@@ -105,6 +105,7 @@ enum vm_event_item { PGPGIN, PGPGOUT, PSWPIN, PSWPOUT,THP_SPLIT_PAGE ,THP_SPLIT_PAGE_FAILED ,THP_DEFERRED_SPLIT_PAGE , +THP_UNDERUSED_SPLIT_PAGE ,THP_SPLIT_PMD ,THP_SCAN_EXCEED_NONE_PTE , THP_SCAN_EXCEED_SWAP_PTE,
mm/huge_memory.c +58 -2
83 lines not on this page · raw message
mm/khugepaged.c +2 -1
@@ -85,7 +85,7 @@ static DECLARE_WAIT_QUEUE_HEAD(khugepaged_wait);* * Note that these are only respected if collapse was initiated by khugepaged. */ -static unsigned int khugepaged_max_ptes_none__read_mostly ; +unsigned int khugepaged_max_ptes_none__read_mostly ;static unsigned int khugepaged_max_ptes_swap__read_mostly ;static unsigned int khugepaged_max_ptes_shared__read_mostly ;
@@ -1235,6 +1235,7 @@ static int collapse_huge_page(struct mm_struct *mm, unsigned long address,pgtable_trans_huge_deposit (mm, pmd, pgtable);set_pmd_at (mm, address, pmd, _pmd);update_mmu_cache_pmd (vma, address, pmd); +deferred_split_folio (folio, false);spin_unlock (pmd_ptl); folio =NULL ;
mm/vmstat.c +1 -0
@@ -1384,6 +1384,7 @@ const char * const vmstat_text[] = {"thp_split_page" ,"thp_split_page_failed" ,"thp_deferred_split_page" , +"thp_underused_split_page" ,"thp_split_pmd" ,"thp_scan_exceed_none_pte" , "thp_scan_exceed_swap_pte",
Signature · 1 line
2.43.5
PATCH v4 6/6 mm: add sysfs entry to disable splitting underused THPs
In reply to Usama ArifIf disabled, THPs faulted in or collapsed will not be added to _deferred_list, and therefore won't be considered for splitting under memory pressure if underused.
--- Documentation/admin-guide/mm/transhuge.rst | 10 +++++++++ mm/huge_memory.c | 26 ++++++++++++++++++++++ 2 files changed, 36 insertions(+)
Documentation/admin-guide/mm/transhuge.rst +10 -0
@@ -202,6 +202,16 @@ PMD-mappable transparent hugepage:: cat /sys/kernel/mm/transparent_hugepage/hpage_pmd_size +All THPs at fault and collapse time will be added to _deferred_list, +and will therefore be split under memory presure if they are considered +"underused". A THP is underused if the number of zero-filled pages in +the THP is above max_ptes_none (see below). It is possible to disable +this behaviour by writing 0 to shrink_underused, and enable it by writing +1 to it:: + + echo 0 > /sys/kernel/mm/transparent_hugepage/shrink_underused + echo 1 > /sys/kernel/mm/transparent_hugepage/shrink_underused + khugepaged will be automatically started when PMD-sized THP is enabled (either of the per-size anon control or the top-level control are set to "always" or "madvise"), and it'll be automatically shutdown when
mm/huge_memory.c +26 -0
@@ -74,6 +74,7 @@ static unsigned long deferred_split_count(struct shrinker *shrink, struct shrink_control *sc);static unsigned long deferred_split_scan (struct shrinker *shrink,struct shrink_control *sc); +static bool split_underused_thp = true;static atomic_t huge_zero_refcount;struct folio *huge_zero_folio__read_mostly ;
@@ -439,6 +440,27 @@ static ssize_t hpage_pmd_size_show(struct kobject *kobj,static struct kobj_attribute hpage_pmd_size_attr =__ATTR_RO (hpage_pmd_size); +static ssize_t split_underused_thp_show (struct kobject *kobj, +struct kobj_attribute *attr,char *buf) +{ +return sysfs_emit (buf,"%d\n" , split_underused_thp); +} + +static ssize_t split_underused_thp_store (struct kobject *kobj, +struct kobj_attribute *attr, +const char *buf,size_t count) +{ +int err =kstrtobool (buf, &split_underused_thp); + +if (err <0 ) +return err; + +return count; +} + +static struct kobj_attribute split_underused_thp_attr =__ATTR ( + shrink_underused,0644 , split_underused_thp_show, split_underused_thp_store); +static struct attribute *hugepage_attr[] = { &enabled_attr.attr, &defrag_attr.attr,
@@ -447,6 +469,7 @@ static struct attribute *hugepage_attr[] = {#ifdef CONFIG_SHMEM &shmem_enabled_attr.attr, #endif + &split_underused_thp_attr.attr, NULL, };
@@ -3477,6 +3500,9 @@ void deferred_split_folio(struct folio *folio, bool partially_mapped)if (folio_order (folio) <=1 )return ; +if (!partially_mapped && !split_underused_thp) +return ; +/* * The try_to_unmap() in page reclaim path might reach here too, * this may cause a race condition to corrupt deferred split queue.
Signature · 1 line
2.43.5