mm: show proportional swap share of the mapping
Subsystems (5)
PATCH v2 1/1 mm: show proportional swap share of the mapping
We want to know per-process workingset size for smart memory management on userland and we use swap(ex, zram) heavily to maximize memory efficiency so workingset includes swap as well as RSS.
On such system, if there are lots of shared anonymous pages, it's really hard to figure out exactly how many each process consumes memory(ie, rss + wap) if the system has lots of shared anonymous memory(e.g, android).
This patch introduces SwapPss field on /proc/<pid>/smaps so we can get more exact workingset size per process.
Bongkyu tested it. Result is below.
1. 50M used swap
SwapTotal: 461976 kB
SwapFree: 411192 kB
$ adb shell cat /proc/*/smaps | grep "SwapPss:" | awk '{sum += $2} END {print sum}';
48236
$ adb shell cat /proc/*/smaps | grep "Swap:" | awk '{sum += $2} END {print sum}';
1411842. 240M used swap
SwapTotal: 461976 kB
SwapFree: 216808 kB
$ adb shell cat /proc/*/smaps | grep "SwapPss:" | awk '{sum += $2} END {print sum}';
230315
$ adb shell cat /proc/*/smaps | grep "Swap:" | awk '{sum += $2} END {print sum}';
1387744* from v1
* add more description - Andrew
* swp_swacount fix on !CONFIG_SWP - Sergey
* add what PSS is to proc.txt - Andrew
* Bring quote from lwn.net - Corbet
* http://lwn.net/Articles/230975/
--- Documentation/filesystems/proc.txt | 18 +++++++++++----- fs/proc/task_mmu.c | 18 ++++++++++++++-- include/linux/swap.h | 6 ++++++ mm/swapfile.c | 42 ++++++++++++++++++++++++++++++++++++++ 4 files changed, 77 insertions(+), 7 deletions(-)
Documentation/filesystems/proc.txt +13 -5
@@ -423,6 +423,7 @@ Private_Dirty: 0 kB Referenced: 892 kB Anonymous: 0 kB Swap: 0 kB +SwapPss: 0 kB KernelPageSize: 4 kB MMUPageSize: 4 kB Locked: 374 kB
@@ -432,16 +433,23 @@ the first of these lines shows the same information as is displayed for the mapping in /proc/PID/maps. The remaining lines show the size of the mapping (size), the amount of the mapping that is currently resident in RAM (RSS), the process' proportional share of this mapping (PSS), the number of clean and -dirty private pages in the mapping. Note that even a page which is part of a -MAP_SHARED mapping, but has only a single pte mapped, i.e. is currently used -by only one process, is accounted as private and not as shared. "Referenced" -indicates the amount of memory currently marked as referenced or accessed. +dirty private pages in the mapping. + +The "proportional set size" (PSS) of a process is the count of pages it has +in memory, where each page is divided by the number of processes sharing it. +So if a process has 1000 pages all to itself, and 1000 shared with one other +process, its PSS will be 1500. +Note that even a page which is part of a MAP_SHARED mapping, but has only +a single pte mapped, i.e. is currently used by only one process, is accounted +as private and not as shared. +"Referenced" indicates the amount of memory currently marked as referenced or +accessed. "Anonymous" shows the amount of memory that does not belong to any file. Even a mapping associated with a file may contain anonymous pages: when MAP_PRIVATE and a page is modified, the file page is replaced by a private anonymous copy. "Swap" shows how much would-be-anonymous memory is also used, but out on swap. - +"SwapPss" shows proportional swap share of this mapping. "VmFlags" field deserves a separate description. This member represents the kernel flags associated with the particular virtual memory area in two letter encoded manner. The codes are the following:
fs/proc/task_mmu.c +16 -2
@@ -446,6 +446,7 @@ struct mem_size_stats {unsigned long anonymous_thp;unsigned long swap;u64 pss; +u64 swap_pss; }; static void smaps_account(struct mem_size_stats *mss, struct page *page,
@@ -492,9 +493,20 @@ static void smaps_pte_entry(pte_t *pte, unsigned long addr, }else if (is_swap_pte (*pte)) {swp_entry_t swpent =pte_to_swp_entry (*pte); -if (!non_swap_entry (swpent)) +if (!non_swap_entry (swpent)) { +int mapcount; + mss->swap +=PAGE_SIZE ; -else if (is_migration_entry (swpent)) + mapcount =swp_swapcount (swpent); +if (mapcount >=2 ) { +u64 pss_delta = (u64 )PAGE_SIZE <<PSS_SHIFT ; + +do_div (pss_delta, mapcount); + mss->swap_pss += pss_delta; + }else { + mss->swap_pss += (u64 )PAGE_SIZE <<PSS_SHIFT ; + } + }else if (is_migration_entry (swpent)) page =migration_entry_to_page (swpent); }
@@ -638,6 +650,7 @@ static int show_smap(struct seq_file *m, void *v, int is_pid)"Anonymous: %8lu kB\n" "AnonHugePages: %8lu kB\n" "Swap: %8lu kB\n" +"SwapPss: %8lu kB\n" "KernelPageSize: %8lu kB\n" "MMUPageSize: %8lu kB\n" "Locked: %8lu kB\n",
@@ -652,6 +665,7 @@ static int show_smap(struct seq_file *m, void *v, int is_pid) mss.anonymous >>10 , mss.anonymous_thp >>10 , mss.swap >>10 , + (unsigned long )(mss.swap_pss >> (10 +PSS_SHIFT )),vma_kernel_pagesize (vma) >>10 ,vma_mmu_pagesize (vma) >>10 , (vma->vm_flags & VM_LOCKED) ?
include/linux/swap.h +6 -0
@@ -432,6 +432,7 @@ extern unsigned int count_swap_pages(int, int);extern sector_t map_swap_page (struct page *,struct block_device **);extern sector_t swapdev_block (int ,pgoff_t );extern int page_swapcount (struct page *); +extern int swp_swapcount (swp_entry_t entry);extern struct swap_info_struct *page_swap_info (struct page *);extern int reuse_swap_page (struct page *);extern int try_to_free_swap (struct page *);
@@ -523,6 +524,11 @@ static inline int page_swapcount(struct page *page)return 0 ; } +static inline int swp_swapcount (swp_entry_t entry) +{ +return 0 ; +} +#define reuse_swap_page (page) (page_mapcount(page) == 1)static inline int try_to_free_swap (struct page *page)
mm/swapfile.c +42 -0
48 lines not on this page · raw message
Signature · 7 lines
1.9.1 -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
It seems merge windows is closed so bump up.
Quoted 223 lines · Minchan Kim
We want to know per-process workingset size for smart memory management on userland and we use
swap(ex, zram)heavily to maximize memory efficiency so workingset includes swap as well as RSS.On such system, if there are lots of shared anonymous pages, it's really hard to figure out exactly how many each process consumes
memory(ie, rss + wap)if the system has lots of shared anonymousmemory(e.g, android).This patch introduces SwapPss field on /proc/<pid>/smaps so we can get more exact workingset size per process.
Bongkyu tested it. Result is below.
1. 50M used swap
SwapTotal: 461976 kB
SwapFree: 411192 kB$ adb shell cat /proc/*/smaps | grep "SwapPss:" | awk '{sum += $2} END {print sum}'; 48236 $ adb shell cat /proc/*/smaps | grep "Swap:" | awk '{sum += $2} END {print sum}'; 1411842. 240M used swap
SwapTotal: 461976 kB
SwapFree: 216808 kB$ adb shell cat /proc/*/smaps | grep "SwapPss:" | awk '{sum += $2} END {print sum}'; 230315 $ adb shell cat /proc/*/smaps | grep "Swap:" | awk '{sum += $2} END {print sum}'; 1387744* from v1
* add more description - Andrew
*swp_swacountfix on !CONFIG_SWP - Sergey
* add what PSS is to proc.txt - Andrew
* Bring quote from lwn.net - Corbet
* http://lwn.net/Articles/230975/Cc Hugh Dickins <[email protected]>Cc Sergey Senozhatsky <[email protected]>Cc Jonathan Corbet <[email protected]>Report-and-Tested-by Bongkyu Kim <[email protected]>Signed-off-by Minchan Kim <[email protected]>--- Documentation/filesystems/proc.txt | 18 +++++++++++----- fs/proc/task_mmu.c | 18 ++++++++++++++-- include/linux/swap.h | 6 ++++++ mm/swapfile.c | 42 ++++++++++++++++++++++++++++++++++++++ 4 files changed, 77 insertions(+), 7 deletions(-)Documentation/filesystems/proc.txt +13
-5@@ -423,6 +423,7 @@ Private_Dirty: 0 kB Referenced: 892 kB Anonymous: 0 kB Swap: 0 kB +SwapPss: 0 kB KernelPageSize: 4 kB MMUPageSize: 4 kB Locked: 374 kB@@ -432,16 +433,23 @@ the first of these lines shows the same information as is displayed for the mapping in /proc/PID/maps. The remaining lines show the size of the mapping (size), the amount of the mapping that is currently resident in RAM (RSS), the process' proportional share of this mapping (PSS), the number of clean and -dirty private pages in the mapping. Note that even a page which is part of a -MAP_SHARED mapping, but has only a single pte mapped, i.e. is currently used -by only one process, is accounted as private and not as shared. "Referenced" -indicates the amount of memory currently marked as referenced or accessed. +dirty private pages in the mapping. + +The "proportional set size" (PSS) of a process is the count of pages it has +in memory, where each page is divided by the number of processes sharing it. +So if a process has 1000 pages all to itself, and 1000 shared with one other +process, its PSS will be 1500. +Note that even a page which is part of a MAP_SHARED mapping, but has only +a single pte mapped, i.e. is currently used by only one process, is accounted +as private and not as shared. +"Referenced" indicates the amount of memory currently marked as referenced or +accessed. "Anonymous" shows the amount of memory that does not belong to any file. Even a mapping associated with a file may contain anonymous pages: when MAP_PRIVATE and a page is modified, the file page is replaced by a private anonymous copy. "Swap" shows how much would-be-anonymous memory is also used, but out on swap. - +"SwapPss" shows proportional swap share of this mapping. "VmFlags" field deserves a separate description. This member represents the kernel flags associated with the particular virtual memory area in two letter encoded manner. The codes are the following:fs/proc/task_mmu.c +16
-2@@ -446,6 +446,7 @@ struct mem_size_stats {unsigned long anonymous_thp;unsigned long swap;u64 pss; +u64 swap_pss; }; static void smaps_account(struct mem_size_stats *mss, struct page *page,@@ -492,9 +493,20 @@ static void smaps_pte_entry(pte_t *pte, unsigned long addr, }else if (is_swap_pte (*pte)) {swp_entry_t swpent =pte_to_swp_entry (*pte); -if (!non_swap_entry (swpent)) +if (!non_swap_entry (swpent)) { +int mapcount; + mss->swap +=PAGE_SIZE ; -else if (is_migration_entry (swpent)) + mapcount =swp_swapcount (swpent); +if (mapcount >=2 ) { +u64 pss_delta = (u64 )PAGE_SIZE <<PSS_SHIFT ; + +do_div (pss_delta, mapcount); + mss->swap_pss += pss_delta; + }else { + mss->swap_pss += (u64 )PAGE_SIZE <<PSS_SHIFT ; + } + }else if (is_migration_entry (swpent)) page =migration_entry_to_page (swpent); }@@ -638,6 +650,7 @@ static int show_smap(struct seq_file *m, void *v, int is_pid)"Anonymous: %8lu kB\n" "AnonHugePages: %8lu kB\n" "Swap: %8lu kB\n" +"SwapPss: %8lu kB\n" "KernelPageSize: %8lu kB\n" "MMUPageSize: %8lu kB\n" "Locked: %8lu kB\n",@@ -652,6 +665,7 @@ static int show_smap(struct seq_file *m, void *v, int is_pid) mss.anonymous >>10 , mss.anonymous_thp >>10 , mss.swap >>10 , + (unsigned long )(mss.swap_pss >> (10 +PSS_SHIFT )),vma_kernel_pagesize (vma) >>10 ,vma_mmu_pagesize (vma) >>10 , (vma->vm_flags & VM_LOCKED) ?include/linux/swap.h +6
-0@@ -432,6 +432,7 @@ extern unsigned int count_swap_pages(int, int);extern sector_t map_swap_page (struct page *,struct block_device **);extern sector_t swapdev_block (int ,pgoff_t );extern int page_swapcount (struct page *); +extern int swp_swapcount (swp_entry_t entry);extern struct swap_info_struct *page_swap_info (struct page *);extern int reuse_swap_page (struct page *);extern int try_to_free_swap (struct page *);@@ -523,6 +524,11 @@ static inline int page_swapcount(struct page *page)return 0 ; } +static inline int swp_swapcount (swp_entry_t entry) +{ +return 0 ; +} +#define reuse_swap_page (page) (page_mapcount(page) == 1)static inline int try_to_free_swap (struct page *page)mm/swapfile.c +42
-0@@ -875,6 +875,48 @@ int page_swapcount(struct page *page) }/* +* How many references to @entry are currently swapped out? +* This considers COUNT_CONTINUED so it returns exact answer. +*/ +int swp_swapcount (swp_entry_t entry) +{ +int count, tmp_count, n; +struct swap_info_struct *p; +struct page *page; +pgoff_t offset; +unsigned char *map; + + p =swap_info_get (entry); +if (!p) +return 0 ; + + count =swap_count (p->swap_map[swp_offset (entry)]); +if (!(count &COUNT_CONTINUED )) + goto out; + + count &= ~COUNT_CONTINUED ; + n =SWAP_MAP_MAX +1 ; + + offset =swp_offset (entry); + page =vmalloc_to_page (p->swap_map + offset); + offset &= ~PAGE_MASK ; +VM_BUG_ON (page_private (page) !=SWP_CONTINUED ); + +do { + page = list_entry(page->lru.next, struct page, lru); + map =kmap_atomic (page) + offset; + tmp_count = *map; +kunmap_atomic (map); + + count += (tmp_count & ~COUNT_CONTINUED ) * n; + n *= (SWAP_CONT_MAX +1 ); + }while (tmp_count &COUNT_CONTINUED ); +out: +spin_unlock (&p->lock); +return count; +} + +/* * We can write to an anon page without COW if there are no other references * to it. And as a side-effect, free up its swap: because the old content * on disk will never be read, and seeking back there to write new content Signature · 1 line
1.9.1
Signature · 8 lines
Kind regards, Minchan Kim -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Andrew Morton
maintainerQuoted 33 lines · On Mon, 15 Jun 2015 22:06:54 +0900 Minchan Kim <[email protected]> wrote:
We want to know per-process workingset size for smart memory management on userland and we use
swap(ex, zram)heavily to maximize memory efficiency so workingset includes swap as well as RSS.On such system, if there are lots of shared anonymous pages, it's really hard to figure out exactly how many each process consumes
memory(ie, rss + wap)if the system has lots of shared anonymousmemory(e.g, android).This patch introduces SwapPss field on /proc/<pid>/smaps so we can get more exact workingset size per process.
...
(quoted diff) +19
-0+int swp_swapcount(swp_entry_t entry) +{ + int count, tmp_count, n; + struct swap_info_struct *p; + struct page *page; + pgoff_t offset; + unsigned char *map; + + p = swap_info_get(entry); + if (!p) + return 0; + + count = swap_count(p->swap_map[swp_offset(entry)]); + if (!(count & COUNT_CONTINUED)) + goto out; + + count &= ~COUNT_CONTINUED; + n = SWAP_MAP_MAX + 1; +
Review
Quoted 10 lines
(quoted diff) +10
-0+ offset = swp_offset(entry); + page = vmalloc_to_page(p->swap_map + offset); + offset &= ~PAGE_MASK; + VM_BUG_ON(page_private(page) != SWP_CONTINUED); + + do { + page = list_entry(page->lru.next, struct page, lru); + map = kmap_atomic(page) + offset; + tmp_count = *map; + kunmap_atomic(map);
A little thing: I've never liked the way that kunmap_atomic() accepts any address within the page. It's weird, and it makes the reviewer have to scramble around to make sure the offset can never be >= PAGE_SIZE.
We can easily avoid doing it here:
mm/swapfile.c renamed from mm/swapfile.c~mm-show-proportional-swap-share-of-the-mapping-fix +2 -2
@@ -904,8 +904,8 @@ int swp_swapcount(swp_entry_t entry) do { page = list_entry(page->lru.next, struct page, lru); - map = kmap_atomic(page) + offset; - tmp_count = *map; + map = kmap_atomic(page); + tmp_count = map[offset]; kunmap_atomic(map); count += (tmp_count & ~COUNT_CONTINUED) * n;
Review
Quoted 10 lines
(quoted diff) +7
-0+ count += (tmp_count & ~COUNT_CONTINUED) * n; + n *= (SWAP_CONT_MAX + 1); + } while (tmp_count & COUNT_CONTINUED); +out: + spin_unlock(&p->lock); + return count; +}
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Quoted 42 lines · Andrew Morton
Quoted 40 lines · On Mon, 15 Jun 2015 22:06:54 +0900 Minchan Kim <[email protected]> wrote:
We want to know per-process workingset size for smart memory management on userland and we use
swap(ex, zram)heavily to maximize memory efficiency so workingset includes swap as well as RSS.On such system, if there are lots of shared anonymous pages, it's really hard to figure out exactly how many each process consumes
memory(ie, rss + wap)if the system has lots of shared anonymousmemory(e.g, android).This patch introduces SwapPss field on /proc/<pid>/smaps so we can get more exact workingset size per process.
...
(quoted diff) +26
-0+int swp_swapcount(swp_entry_t entry) +{ + int count, tmp_count, n; + struct swap_info_struct *p; + struct page *page; + pgoff_t offset; + unsigned char *map; + + p = swap_info_get(entry); + if (!p) + return 0; + + count = swap_count(p->swap_map[swp_offset(entry)]); + if (!(count & COUNT_CONTINUED)) + goto out; + + count &= ~COUNT_CONTINUED; + n = SWAP_MAP_MAX + 1; + + offset = swp_offset(entry); + page = vmalloc_to_page(p->swap_map + offset); + offset &= ~PAGE_MASK; + VM_BUG_ON(page_private(page) != SWP_CONTINUED); + + do { + page = list_entry(page->lru.next, struct page, lru);
Quoted 8 lines
Review
Quoted 3 lines
(quoted diff) +3
-0+ map = kmap_atomic(page) + offset; + tmp_count = *map; + kunmap_atomic(map);A little thing: I've never liked the way that
kunmap_atomic()accepts any address within the page. It's weird, and it makes the reviewer have to scramble around to make sure the offset can never be >=PAGE_SIZE.
Very ture. I was bitten by that.
Thanks for the clean up.
Quoted 17 lines
We can easily avoid doing it here:
mm/swapfile.c renamed from mm/swapfile.c~mm-show-proportional-swap-share-of-the-mapping-fix +2
-2@@ -904,8 +904,8 @@ int swp_swapcount(swp_entry_t entry) do { page = list_entry(page->lru.next, struct page, lru); - map = kmap_atomic(page) + offset; - tmp_count = *map; + map = kmap_atomic(page); + tmp_count = map[offset]; kunmap_atomic(map); count += (tmp_count & ~COUNT_CONTINUED) * n;
Quoted 8 lines
Review
Quoted 8 lines
(quoted diff) +7
-0+ count += (tmp_count & ~COUNT_CONTINUED) * n; + n *= (SWAP_CONT_MAX + 1); + } while (tmp_count & COUNT_CONTINUED); +out: + spin_unlock(&p->lock); + return count; +}
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Quoted 14 lines · Minchan Kim
We want to know per-process workingset size for smart memory management on userland and we use
swap(ex, zram)heavily to maximize memory efficiency so workingset includes swap as well as RSS.On such system, if there are lots of shared anonymous pages, it's really hard to figure out exactly how many each process consumes
memory(ie, rss + wap)if the system has lots of shared anonymousmemory(e.g, android).This patch introduces SwapPss field on /proc/<pid>/smaps so we can get more exact workingset size per process.
Bongkyu tested it. Result is below.
Quoted 8 lines
1. 50M used swap
SwapTotal: 461976 kB
SwapFree: 411192 kB$ adb shell cat /proc/*/smaps | grep "SwapPss:" | awk '{sum += $2} END {print sum}'; 48236 $ adb shell cat /proc/*/smaps | grep "Swap:" | awk '{sum += $2} END {print sum}'; 141184
Hi Minchan,
I just found out about this patch. What kind of shared memory is that? Since it's android, I'm inclined to think something specific like ashmem. I'm asking because this patch won't help for more common type of shared memory. See my comment below.
Quoted 10 lines
2. 240M used swap
SwapTotal: 461976 kB
SwapFree: 216808 kB$ adb shell cat /proc/*/smaps | grep "SwapPss:" | awk '{sum += $2} END {print sum}'; 230315 $ adb shell cat /proc/*/smaps | grep "Swap:" | awk '{sum += $2} END {print sum}'; 1387744
snip
Quoted 13 lines
fs/proc/task_mmu.c +1
-0@@ -446,6 +446,7 @@ struct mem_size_stats {unsigned long anonymous_thp;unsigned long swap;u64 pss; +u64 swap_pss; }; static void smaps_account(struct mem_size_stats *mss, struct page *page,
fs/proc/task_mmu.c static void smaps_pte_entry(pte_t *pte, unsigned long addr, · line 493
Quoted 1 line
fs/proc/task_mmu.c +0
-0@@ -492,9 +493,20 @@ static void smaps_pte_entry(pte_t *pte, unsigned long addr, }else if (is_swap_pte (*pte)) {
This won't work for sysV shm, tmpfs and MAP_SHARED | MAP_ANONYMOUS mapping pages which are pte_none when paged out. They're currently not accounted at all when in swap.
Jerome
Quoted 115 lines
fs/proc/task_mmu.c +15
-2swp_entry_t swpent =pte_to_swp_entry (*pte); -if (!non_swap_entry (swpent)) +if (!non_swap_entry (swpent)) { +int mapcount; + mss->swap +=PAGE_SIZE ; -else if (is_migration_entry (swpent)) + mapcount =swp_swapcount (swpent); +if (mapcount >=2 ) { +u64 pss_delta = (u64 )PAGE_SIZE <<PSS_SHIFT ; + +do_div (pss_delta, mapcount); + mss->swap_pss += pss_delta; + }else { + mss->swap_pss += (u64 )PAGE_SIZE <<PSS_SHIFT ; + } + }else if (is_migration_entry (swpent)) page =migration_entry_to_page (swpent); }@@ -638,6 +650,7 @@ static int show_smap(struct seq_file *m, void *v, int is_pid)"Anonymous: %8lu kB\n" "AnonHugePages: %8lu kB\n" "Swap: %8lu kB\n" +"SwapPss: %8lu kB\n" "KernelPageSize: %8lu kB\n" "MMUPageSize: %8lu kB\n" "Locked: %8lu kB\n",@@ -652,6 +665,7 @@ static int show_smap(struct seq_file *m, void *v, int is_pid) mss.anonymous >>10 , mss.anonymous_thp >>10 , mss.swap >>10 , + (unsigned long )(mss.swap_pss >> (10 +PSS_SHIFT )),vma_kernel_pagesize (vma) >>10 ,vma_mmu_pagesize (vma) >>10 , (vma->vm_flags & VM_LOCKED) ?include/linux/swap.h +6
-0@@ -432,6 +432,7 @@ extern unsigned int count_swap_pages(int, int);extern sector_t map_swap_page (struct page *,struct block_device **);extern sector_t swapdev_block (int ,pgoff_t );extern int page_swapcount (struct page *); +extern int swp_swapcount (swp_entry_t entry);extern struct swap_info_struct *page_swap_info (struct page *);extern int reuse_swap_page (struct page *);extern int try_to_free_swap (struct page *);@@ -523,6 +524,11 @@ static inline int page_swapcount(struct page *page)return 0 ; } +static inline int swp_swapcount (swp_entry_t entry) +{ +return 0 ; +} +#define reuse_swap_page (page) (page_mapcount(page) == 1)static inline int try_to_free_swap (struct page *page)mm/swapfile.c +42
-0@@ -875,6 +875,48 @@ int page_swapcount(struct page *page) }/* +* How many references to @entry are currently swapped out? +* This considers COUNT_CONTINUED so it returns exact answer. +*/ +int swp_swapcount (swp_entry_t entry) +{ +int count, tmp_count, n; +struct swap_info_struct *p; +struct page *page; +pgoff_t offset; +unsigned char *map; + + p =swap_info_get (entry); +if (!p) +return 0 ; + + count =swap_count (p->swap_map[swp_offset (entry)]); +if (!(count &COUNT_CONTINUED )) + goto out; + + count &= ~COUNT_CONTINUED ; + n =SWAP_MAP_MAX +1 ; + + offset =swp_offset (entry); + page =vmalloc_to_page (p->swap_map + offset); + offset &= ~PAGE_MASK ; +VM_BUG_ON (page_private (page) !=SWP_CONTINUED ); + +do { + page = list_entry(page->lru.next, struct page, lru); + map =kmap_atomic (page) + offset; + tmp_count = *map; +kunmap_atomic (map); + + count += (tmp_count & ~COUNT_CONTINUED ) * n; + n *= (SWAP_CONT_MAX +1 ); + }while (tmp_count &COUNT_CONTINUED ); +out: +spin_unlock (&p->lock); +return count; +} + +/* * We can write to an anon page without COW if there are no other references * to it. And as a side-effect, free up its swap: because the old content * on disk will never be read, and seeking back there to write new content
Hi Jerome,
Quoted 24 lines · Jerome Marchand
Quoted 14 lines · Minchan Kim
We want to know per-process workingset size for smart memory management on userland and we use
swap(ex, zram)heavily to maximize memory efficiency so workingset includes swap as well as RSS.On such system, if there are lots of shared anonymous pages, it's really hard to figure out exactly how many each process consumes
memory(ie, rss + wap)if the system has lots of shared anonymousmemory(e.g, android).This patch introduces SwapPss field on /proc/<pid>/smaps so we can get more exact workingset size per process.
Bongkyu tested it. Result is below.
Quoted 8 lines
1. 50M used swap
SwapTotal: 461976 kB
SwapFree: 411192 kB$ adb shell cat /proc/*/smaps | grep "SwapPss:" | awk '{sum += $2} END {print sum}'; 48236 $ adb shell cat /proc/*/smaps | grep "Swap:" | awk '{sum += $2} END {print sum}'; 141184
Quoted 6 lines
Hi Minchan,
I just found out about this patch. What kind of shared memory is that? Since it's android, I'm inclined to think something specific like ashmem. I'm asking because this patch won't help for more common type of shared memory. See my comment below.
It's normal heap of parent(IOW, MAP_ANON|MAP_PRIVATE memory which is share
by child processes).
Quoted 23 lines
Quoted 11 lines
2. 240M used swap
SwapTotal: 461976 kB
SwapFree: 216808 kB$ adb shell cat /proc/*/smaps | grep "SwapPss:" | awk '{sum += $2} END {print sum}'; 230315 $ adb shell cat /proc/*/smaps | grep "Swap:" | awk '{sum += $2} END {print sum}'; 1387744snip
Quoted 11 lines
fs/proc/task_mmu.c +1
-0@@ -446,6 +446,7 @@ struct mem_size_stats {unsigned long anonymous_thp;unsigned long swap;u64 pss; +u64 swap_pss; };
Quoted 7 lines
fs/proc/task_mmu.c static void smaps_pte_entry(pte_t *pte, unsigned long addr, · line 493
Quoted 3 lines
fs/proc/task_mmu.c +0
-0static void smaps_account(struct mem_size_stats *mss, struct page *page,@@ -492,9 +493,20 @@ static void smaps_pte_entry(pte_t *pte, unsigned long addr, }else if (is_swap_pte (*pte)) {This won't work for sysV shm, tmpfs and
MAP_SHARED|MAP_ANONYMOUSmapping pages which arepte_nonewhen paged out. They're currently not accounted at all when in swap.
This patch doesn't handle those pages because we don't have supported thoses pages. IMHO, if someone need it, it should be another patch and he can contribute it in future.
Thanks.
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Quoted 29 lines · Minchan Kim
Hi Jerome,
Quoted 26 lines · Jerome Marchand
Quoted 14 lines · Minchan Kim
We want to know per-process workingset size for smart memory management on userland and we use
swap(ex, zram)heavily to maximize memory efficiency so workingset includes swap as well as RSS.On such system, if there are lots of shared anonymous pages, it's really hard to figure out exactly how many each process consumes
memory(ie, rss + wap)if the system has lots of shared anonymousmemory(e.g, android).This patch introduces SwapPss field on /proc/<pid>/smaps so we can get more exact workingset size per process.
Bongkyu tested it. Result is below.
Quoted 8 lines
1. 50M used swap
SwapTotal: 461976 kB
SwapFree: 411192 kB$ adb shell cat /proc/*/smaps | grep "SwapPss:" | awk '{sum += $2} END {print sum}'; 48236 $ adb shell cat /proc/*/smaps | grep "Swap:" | awk '{sum += $2} END {print sum}'; 141184Hi Minchan,
Quoted 7 lines
Quoted 4 lines
I just found out about this patch. What kind of shared memory is that? Since it's android, I'm inclined to think something specific like ashmem. I'm asking because this patch won't help for more common type of shared memory. See my comment below.
It's normal heap of parent(IOW, MAP_ANON|MAP_PRIVATE memory which is share
by child processes).
Ok. I didn't imagine CoW pages would represent such a big share of swapped out pages.
Quoted 32 lines
Quoted 24 lines
Quoted 12 lines
2. 240M used swap
SwapTotal: 461976 kB
SwapFree: 216808 kB$ adb shell cat /proc/*/smaps | grep "SwapPss:" | awk '{sum += $2} END {print sum}'; 230315 $ adb shell cat /proc/*/smaps | grep "Swap:" | awk '{sum += $2} END {print sum}'; 1387744snip
Quoted 11 lines
fs/proc/task_mmu.c +1
-0@@ -446,6 +446,7 @@ struct mem_size_stats {unsigned long anonymous_thp;unsigned long swap;u64 pss; +u64 swap_pss; };Quoted 7 lines
fs/proc/task_mmu.c static void smaps_pte_entry(pte_t *pte, unsigned long addr, · line 493
Quoted 3 lines
fs/proc/task_mmu.c +0
-0static void smaps_account(struct mem_size_stats *mss, struct page *page,@@ -492,9 +493,20 @@ static void smaps_pte_entry(pte_t *pte, unsigned long addr, }else if (is_swap_pte (*pte)) {This won't work for sysV shm, tmpfs and
MAP_SHARED|MAP_ANONYMOUSmapping pages which arepte_nonewhen paged out. They're currently not accounted at all when in swap.
Quoted 3 lines
This patch doesn't handle those pages because we don't have supported thoses pages. IMHO, if someone need it, it should be another patch and he can contribute it in future.
Sure.
Quoted 3 lines
Thanks.
Quoted 7 lines · Minchan Kim
Quoted 3 lines
This won't work for sysV shm, tmpfs and
MAP_SHARED|MAP_ANONYMOUSmapping pages which arepte_nonewhen paged out. They're currently not accounted at all when in swap.This patch doesn't handle those pages because we don't have supported thoses pages. IMHO, if someone need it, it should be another patch and he can contribute it in future.
OK, time to try again...
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>