The patch titled mm: avoid false sharing of mm_counter has been added to the -mm tree. Its filename is mm-avoid-false-sharing-of-mm_counter.patch Before you just go and hit "reply", please: a) Consider who else should be cc'ed b) Prefer to cc a suitable mailing list as well c) Ideally: find the original patch on the mailing list and do a reply-to-all to that, adding suitable additional cc's *** Remember to use Documentation/SubmitChecklist when testing your code *** See http://userweb.kernel.org/~akpm/stuff/added-to-mm.txt to find out what to do about this The current -mm tree may be found at http://userweb.kernel.org/~akpm/mmotm/ ------------------------------------------------------ Subject: mm: avoid false sharing of mm_counter From: KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx> Considering the nature of per mm stats, it's the shared object among threads and can be a cache-miss point in the page fault path. This patch adds per-thread cache for mm_counter. RSS value will be counted into a struct in task_struct and synchronized with mm's one at events. Now, in this patch, the event is the number of calls to handle_mm_fault. Per-thread value is added to mm at each 64 calls. rough estimation with small benchmark on parallel thread (2threads) shows [before] 4.5 cache-miss/faults [after] 4.0 cache-miss/faults Anyway, the most contended object is mmap_sem if the number of threads grows. Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx> Cc: Minchan Kim <minchan.kim@xxxxxxxxx> Cc: Christoph Lameter <cl@xxxxxxxxxxxxxxxxxxxx> Cc: Lee Schermerhorn <lee.schermerhorn@xxxxxx> Cc: David Rientjes <rientjes@xxxxxxxxxx> Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> --- Documentation/filesystems/proc.txt | 6 + fs/exec.c | 1 include/linux/mm.h | 8 -- include/linux/mm_types.h | 6 + include/linux/sched.h | 4 - kernel/exit.c | 3 mm/memory.c | 94 ++++++++++++++++++++++++--- 7 files changed, 107 insertions(+), 15 deletions(-) diff -puN Documentation/filesystems/proc.txt~mm-avoid-false-sharing-of-mm_counter Documentation/filesystems/proc.txt --- a/Documentation/filesystems/proc.txt~mm-avoid-false-sharing-of-mm_counter +++ a/Documentation/filesystems/proc.txt @@ -189,6 +189,12 @@ memory usage. Its seven fields are expla contains details information about the process itself. Its fields are explained in Table 1-4. +(for SMP CONFIG users) +For making accounting scalable, RSS related information are handled in +asynchronous manner and the vaule may not be very precise. To see a precise +snapshot of a moment, you can see /proc/<pid>/smaps file and scan page table. +It's slow but very precise. + Table 1-2: Contents of the statm files (as of 2.6.30-rc7) .............................................................................. Field Content diff -puN fs/exec.c~mm-avoid-false-sharing-of-mm_counter fs/exec.c --- a/fs/exec.c~mm-avoid-false-sharing-of-mm_counter +++ a/fs/exec.c @@ -702,6 +702,7 @@ static int exec_mmap(struct mm_struct *m /* Notify parent that we're no longer interested in the old VM */ tsk = current; old_mm = current->mm; + sync_mm_rss(tsk, old_mm); mm_release(tsk, old_mm); if (old_mm) { diff -puN include/linux/mm.h~mm-avoid-false-sharing-of-mm_counter include/linux/mm.h --- a/include/linux/mm.h~mm-avoid-false-sharing-of-mm_counter +++ a/include/linux/mm.h @@ -870,7 +870,7 @@ int __get_user_pages_fast(unsigned long /* * per-process(per-mm_struct) statistics. */ -#if USE_SPLIT_PTLOCKS +#if defined(SPLIT_RSS_COUNTING) /* * The mm counters are not protected by its page_table_lock, * so must be incremented atomically. @@ -880,10 +880,7 @@ static inline void set_mm_counter(struct atomic_long_set(&mm->rss_stat.count[member], value); } -static inline unsigned long get_mm_counter(struct mm_struct *mm, int member) -{ - return (unsigned long)atomic_long_read(&mm->rss_stat.count[member]); -} +unsigned long get_mm_counter(struct mm_struct *mm, int member); static inline void add_mm_counter(struct mm_struct *mm, int member, long value) { @@ -971,6 +968,7 @@ static inline void setmax_mm_hiwater_rss *maxrss = hiwater_rss; } +void sync_mm_rss(struct task_struct *task, struct mm_struct *mm); /* * A callback you can register to apply pressure to ageable caches. diff -puN include/linux/mm_types.h~mm-avoid-false-sharing-of-mm_counter include/linux/mm_types.h --- a/include/linux/mm_types.h~mm-avoid-false-sharing-of-mm_counter +++ a/include/linux/mm_types.h @@ -200,9 +200,15 @@ enum { }; #if USE_SPLIT_PTLOCKS +#define SPLIT_RSS_COUNTING struct mm_rss_stat { atomic_long_t count[NR_MM_COUNTERS]; }; +/* per-thread cached information, */ +struct task_rss_stat { + int events; /* for synchronization threshold */ + int count[NR_MM_COUNTERS]; +}; #else /* !USE_SPLIT_PTLOCKS */ struct mm_rss_stat { unsigned long count[NR_MM_COUNTERS]; diff -puN include/linux/sched.h~mm-avoid-false-sharing-of-mm_counter include/linux/sched.h --- a/include/linux/sched.h~mm-avoid-false-sharing-of-mm_counter +++ a/include/linux/sched.h @@ -1223,7 +1223,9 @@ struct task_struct { struct plist_node pushable_tasks; struct mm_struct *mm, *active_mm; - +#if defined(SPLIT_RSS_COUNTING) + struct task_rss_stat rss_stat; +#endif /* task state */ int exit_state; int exit_code, exit_signal; diff -puN kernel/exit.c~mm-avoid-false-sharing-of-mm_counter kernel/exit.c --- a/kernel/exit.c~mm-avoid-false-sharing-of-mm_counter +++ a/kernel/exit.c @@ -944,7 +944,8 @@ NORET_TYPE void do_exit(long code) preempt_count()); acct_update_integrals(tsk); - + /* sync mm's RSS info before statistics gathering */ + sync_mm_rss(tsk, tsk->mm); group_dead = atomic_dec_and_test(&tsk->signal->live); if (group_dead) { hrtimer_cancel(&tsk->signal->real_timer); diff -puN mm/memory.c~mm-avoid-false-sharing-of-mm_counter mm/memory.c --- a/mm/memory.c~mm-avoid-false-sharing-of-mm_counter +++ a/mm/memory.c @@ -122,6 +122,79 @@ static int __init init_zero_pfn(void) core_initcall(init_zero_pfn); +#if defined(SPLIT_RSS_COUNTING) + +void __sync_task_rss_stat(struct task_struct *task, struct mm_struct *mm) +{ + int i; + + for (i = 0; i < NR_MM_COUNTERS; i++) { + if (task->rss_stat.count[i]) { + add_mm_counter(mm, i, task->rss_stat.count[i]); + task->rss_stat.count[i] = 0; + } + } + task->rss_stat.events = 0; +} + +static void add_mm_counter_fast(struct mm_struct *mm, int member, int val) +{ + struct task_struct *task = current; + + if (likely(task->mm == mm)) + task->rss_stat.count[member] += val; + else + add_mm_counter(mm, member, val); +} +#define inc_mm_counter_fast(mm, member) add_mm_counter_fast(mm, member,1) +#define dec_mm_counter_fast(mm, member) add_mm_counter_fast(mm, member,-1) + +/* sync counter once per 64 page faults */ +#define TASK_RSS_EVENTS_THRESH (64) +static void check_sync_rss_stat(struct task_struct *task) +{ + if (unlikely(task != current)) + return; + if (unlikely(task->rss_stat.events++ > TASK_RSS_EVENTS_THRESH)) + __sync_task_rss_stat(task, task->mm); +} + +unsigned long get_mm_counter(struct mm_struct *mm, int member) +{ + long val = 0; + + /* + * Don't use task->mm here...for avoiding to use task_get_mm().. + * The caller must guarantee task->mm is not invalid. + */ + val = atomic_long_read(&mm->rss_stat.count[member]); + /* + * counter is updated in asynchronous manner and may go to minus. + * But it's never be expected number for users. + */ + if (val < 0) + return 0; + return (unsigned long)val; +} + +void sync_mm_rss(struct task_struct *task, struct mm_struct *mm) +{ + __sync_task_rss_stat(task, mm); +} +#else + +#define inc_mm_counter_fast(mm, member) inc_mm_counter(mm, member) +#define dec_mm_counter_fast(mm, member) dec_mm_counter(mm, member) + +static void check_sync_rss_stat(struct task_struct *task) +{ +} + +void sync_mm_rss(struct task_struct *task, struct mm_struct *mm) +{ +} +#endif + /* * If a p?d_bad entry is found while walking page tables, report * the error, before resetting entry to p?d_none. Usually (but @@ -386,6 +459,8 @@ static inline void add_mm_rss_vec(struct { int i; + if (current->mm == mm) + sync_mm_rss(current, mm); for (i = 0; i < NR_MM_COUNTERS; i++) if (rss[i]) add_mm_counter(mm, i, rss[i]); @@ -1539,7 +1614,7 @@ static int insert_page(struct vm_area_st /* Ok, finally just insert the thing.. */ get_page(page); - inc_mm_counter(mm, MM_FILEPAGES); + inc_mm_counter_fast(mm, MM_FILEPAGES); page_add_file_rmap(page); set_pte_at(mm, addr, pte, mk_pte(page, prot)); @@ -2175,11 +2250,11 @@ gotten: if (likely(pte_same(*page_table, orig_pte))) { if (old_page) { if (!PageAnon(old_page)) { - dec_mm_counter(mm, MM_FILEPAGES); - inc_mm_counter(mm, MM_ANONPAGES); + dec_mm_counter_fast(mm, MM_FILEPAGES); + inc_mm_counter_fast(mm, MM_ANONPAGES); } } else - inc_mm_counter(mm, MM_ANONPAGES); + inc_mm_counter_fast(mm, MM_ANONPAGES); flush_cache_page(vma, address, pte_pfn(orig_pte)); entry = mk_pte(new_page, vma->vm_page_prot); entry = maybe_mkwrite(pte_mkdirty(entry), vma); @@ -2616,7 +2691,7 @@ static int do_swap_page(struct mm_struct * discarded at swap_free(). */ - inc_mm_counter(mm, MM_ANONPAGES); + inc_mm_counter_fast(mm, MM_ANONPAGES); pte = mk_pte(page, vma->vm_page_prot); if ((flags & FAULT_FLAG_WRITE) && reuse_swap_page(page)) { pte = maybe_mkwrite(pte_mkdirty(pte), vma); @@ -2700,7 +2775,7 @@ static int do_anonymous_page(struct mm_s if (!pte_none(*page_table)) goto release; - inc_mm_counter(mm, MM_ANONPAGES); + inc_mm_counter_fast(mm, MM_ANONPAGES); page_add_new_anon_rmap(page, vma, address); setpte: set_pte_at(mm, address, page_table, entry); @@ -2854,10 +2929,10 @@ static int __do_fault(struct mm_struct * if (flags & FAULT_FLAG_WRITE) entry = maybe_mkwrite(pte_mkdirty(entry), vma); if (anon) { - inc_mm_counter(mm, MM_ANONPAGES); + inc_mm_counter_fast(mm, MM_ANONPAGES); page_add_new_anon_rmap(page, vma, address); } else { - inc_mm_counter(mm, MM_FILEPAGES); + inc_mm_counter_fast(mm, MM_FILEPAGES); page_add_file_rmap(page); if (flags & FAULT_FLAG_WRITE) { dirty_page = page; @@ -3035,6 +3110,9 @@ int handle_mm_fault(struct mm_struct *mm count_vm_event(PGFAULT); + /* do counter updates before entering really critical section. */ + check_sync_rss_stat(current); + if (unlikely(is_vm_hugetlb_page(vma))) return hugetlb_fault(mm, vma, address, flags); _ Patches currently in -mm which might be from kamezawa.hiroyu@xxxxxxxxxxxxxx are mm-add-notifier-in-pageblock-isolation-for-balloon-drivers.patch powerpc-make-the-cmm-memory-hotplug-aware.patch mm-clean-up-mm_counter.patch mm-avoid-false-sharing-of-mm_counter.patch mm-count-swap-usage.patch mm-add-lowmem-detection-logic.patch mm-count-lowmem-rss.patch -- To unsubscribe from this list: send the line "unsubscribe mm-commits" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html