| Message ID | 20210507150553.208763-4-peterx@redhat.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | mm/gup: Fix pin page write cache bouncing on has_pinned | expand |
On 5/7/21 8:05 AM, Peter Xu wrote: > From: Andrea Arcangeli <aarcange@redhat.com> > > has_pinned 32bit can be packed in the MMF_HAS_PINNED bit as a noop > cleanup. > > Any atomic_inc/dec to the mm cacheline shared by all threads in > pin-fast would reintroduce a loss of SMP scalability to pin-fast, so > there's no future potential usefulness to keep an atomic in the mm for > this. > > set_bit(MMF_HAS_PINNED) will be theoretically a bit slower than > WRITE_ONCE (atomic_set is equivalent to WRITE_ONCE), but the set_bit > (just like atomic_set after this commit) has to be still issued only > once per "mm", so the difference between the two will be lost in the > noise. > > will-it-scale "mmap2" shows no change in performance with enterprise > config as expected. > > will-it-scale "pin_fast" retains the > 4000% SMP scalability > performance improvement against upstream as expected. > > This is a noop as far as overall performance and SMP scalability are > concerned. > > Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> > [peterx: Fix build for task_mmu.c, introduce mm_set_has_pinned_flag, fix > comment here and there] > Signed-off-by: Peter Xu <peterx@redhat.com> > --- > fs/proc/task_mmu.c | 2 +- > include/linux/mm.h | 2 +- > include/linux/mm_types.h | 10 ---------- > include/linux/sched/coredump.h | 8 ++++++++ > kernel/fork.c | 1 - > mm/gup.c | 19 +++++++++++++++---- > 6 files changed, 25 insertions(+), 17 deletions(-) > Looks good. Reviewed-by: John Hubbard <jhubbard@nvidia.com> thanks,
Hi Peter, Andrea, On Fri, May 7, 2021 at 7:26 PM Peter Xu <peterx@redhat.com> wrote: > From: Andrea Arcangeli <aarcange@redhat.com> > > has_pinned 32bit can be packed in the MMF_HAS_PINNED bit as a noop > cleanup. > > Any atomic_inc/dec to the mm cacheline shared by all threads in > pin-fast would reintroduce a loss of SMP scalability to pin-fast, so > there's no future potential usefulness to keep an atomic in the mm for > this. > > set_bit(MMF_HAS_PINNED) will be theoretically a bit slower than > WRITE_ONCE (atomic_set is equivalent to WRITE_ONCE), but the set_bit > (just like atomic_set after this commit) has to be still issued only > once per "mm", so the difference between the two will be lost in the > noise. > > will-it-scale "mmap2" shows no change in performance with enterprise > config as expected. > > will-it-scale "pin_fast" retains the > 4000% SMP scalability > performance improvement against upstream as expected. > > This is a noop as far as overall performance and SMP scalability are > concerned. > > Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> > [peterx: Fix build for task_mmu.c, introduce mm_set_has_pinned_flag, fix > comment here and there] > Signed-off-by: Peter Xu <peterx@redhat.com> Thanks for your patch, which is now in linux-next. > diff --git a/mm/gup.c b/mm/gup.c > index 9933bc5c2eff2..bb130723a6717 100644 > --- a/mm/gup.c > +++ b/mm/gup.c > @@ -1270,6 +1270,17 @@ int fixup_user_fault(struct mm_struct *mm, > } > EXPORT_SYMBOL_GPL(fixup_user_fault); > > +/* > + * Set the MMF_HAS_PINNED if not set yet; after set it'll be there for the mm's > + * lifecycle. Avoid setting the bit unless necessary, or it might cause write > + * cache bouncing on large SMP machines for concurrent pinned gups. > + */ > +static inline void mm_set_has_pinned_flag(unsigned long *mm_flags) > +{ > + if (!test_bit(MMF_HAS_PINNED, mm_flags)) > + set_bit(MMF_HAS_PINNED, mm_flags); > +} > + > /* > * Please note that this function, unlike __get_user_pages will not > * return 0 for nr_pages > 0 without FOLL_NOWAIT > @@ -1292,8 +1303,8 @@ static __always_inline long __get_user_pages_locked(struct mm_struct *mm, > BUG_ON(*locked != 1); > } > > - if ((flags & FOLL_PIN) && !atomic_read(&mm->has_pinned)) > - atomic_set(&mm->has_pinned, 1); > + if (flags & FOLL_PIN) > + mm_set_has_pinned_flag(&mm->flags); > > /* > * FOLL_PIN and FOLL_GET are mutually exclusive. Traditional behavior > @@ -2617,8 +2628,8 @@ static int internal_get_user_pages_fast(unsigned long start, > FOLL_FAST_ONLY))) > return -EINVAL; > > - if ((gup_flags & FOLL_PIN) && !atomic_read(¤t->mm->has_pinned)) > - atomic_set(¤t->mm->has_pinned, 1); > + if (gup_flags & FOLL_PIN) > + mm_set_has_pinned_flag(¤t->mm->flags); noreply@ellerman.id.au reports: FAILED linux-next/m5272c3_defconfig/m68k-gcc8 Wed May 12, 19:30 http://kisskb.ellerman.id.au/kisskb/buildresult/14543658/ Commit: Add linux-next specific files for 20210512 ec85c95b0c90a17413901b018e8ade7b9eae7cad Compiler: m68k-linux-gcc (GCC) 8.1.0 / GNU ld (GNU Binutils) 2.30 mm/gup.c:2698:3: error: implicit declaration of function 'mm_set_has_pinned_flag'; did you mean 'set_tsk_thread_flag'? [-Werror=implicit-function-declaration] It's definition is inside the #ifdef CONFIG_MMU section, but the last user isn't. Gr{oetje,eeting}s, Geert
On Fri, 7 May 2021 at 20:36, Peter Xu <peterx@redhat.com> wrote: > > From: Andrea Arcangeli <aarcange@redhat.com> > > has_pinned 32bit can be packed in the MMF_HAS_PINNED bit as a noop > cleanup. > > Any atomic_inc/dec to the mm cacheline shared by all threads in > pin-fast would reintroduce a loss of SMP scalability to pin-fast, so > there's no future potential usefulness to keep an atomic in the mm for > this. > > set_bit(MMF_HAS_PINNED) will be theoretically a bit slower than > WRITE_ONCE (atomic_set is equivalent to WRITE_ONCE), but the set_bit > (just like atomic_set after this commit) has to be still issued only > once per "mm", so the difference between the two will be lost in the > noise. > > will-it-scale "mmap2" shows no change in performance with enterprise > config as expected. > > will-it-scale "pin_fast" retains the > 4000% SMP scalability > performance improvement against upstream as expected. > > This is a noop as far as overall performance and SMP scalability are > concerned. > > Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> > [peterx: Fix build for task_mmu.c, introduce mm_set_has_pinned_flag, fix > comment here and there] > Signed-off-by: Peter Xu <peterx@redhat.com> > --- > fs/proc/task_mmu.c | 2 +- > include/linux/mm.h | 2 +- > include/linux/mm_types.h | 10 ---------- > include/linux/sched/coredump.h | 8 ++++++++ > kernel/fork.c | 1 - > mm/gup.c | 19 +++++++++++++++---- > 6 files changed, 25 insertions(+), 17 deletions(-) > > diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c > index 4c95cc57a66a8..6144571942db9 100644 > --- a/fs/proc/task_mmu.c > +++ b/fs/proc/task_mmu.c > @@ -1049,7 +1049,7 @@ static inline bool pte_is_pinned(struct vm_area_struct *vma, unsigned long addr, > return false; > if (!is_cow_mapping(vma->vm_flags)) > return false; > - if (likely(!atomic_read(&vma->vm_mm->has_pinned))) > + if (likely(!test_bit(MMF_HAS_PINNED, &vma->vm_mm->flags))) > return false; > page = vm_normal_page(vma, addr, pte); > if (!page) > diff --git a/include/linux/mm.h b/include/linux/mm.h > index d6790ab0cf575..94dc84f6d8658 100644 > --- a/include/linux/mm.h > +++ b/include/linux/mm.h > @@ -1331,7 +1331,7 @@ static inline bool page_needs_cow_for_dma(struct vm_area_struct *vma, > if (!is_cow_mapping(vma->vm_flags)) > return false; > > - if (!atomic_read(&vma->vm_mm->has_pinned)) > + if (!test_bit(MMF_HAS_PINNED, &vma->vm_mm->flags)) > return false; > > return page_maybe_dma_pinned(page); > diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h > index 6613b26a88946..15d79858fadbd 100644 > --- a/include/linux/mm_types.h > +++ b/include/linux/mm_types.h > @@ -435,16 +435,6 @@ struct mm_struct { > */ > atomic_t mm_count; > > - /** > - * @has_pinned: Whether this mm has pinned any pages. This can > - * be either replaced in the future by @pinned_vm when it > - * becomes stable, or grow into a counter on its own. We're > - * aggresive on this bit now - even if the pinned pages were > - * unpinned later on, we'll still keep this bit set for the > - * lifecycle of this mm just for simplicity. > - */ > - atomic_t has_pinned; > - > /** > * @write_protect_seq: Locked when any thread is write > * protecting pages mapped by this mm to enforce a later COW, > diff --git a/include/linux/sched/coredump.h b/include/linux/sched/coredump.h > index dfd82eab29025..4d9e3a6568758 100644 > --- a/include/linux/sched/coredump.h > +++ b/include/linux/sched/coredump.h > @@ -73,6 +73,14 @@ static inline int get_dumpable(struct mm_struct *mm) > #define MMF_OOM_VICTIM 25 /* mm is the oom victim */ > #define MMF_OOM_REAP_QUEUED 26 /* mm was queued for oom_reaper */ > #define MMF_MULTIPROCESS 27 /* mm is shared between processes */ > +/* > + * MMF_HAS_PINNED: Whether this mm has pinned any pages. This can be either > + * replaced in the future by mm.pinned_vm when it becomes stable, or grow into > + * a counter on its own. We're aggresive on this bit for now: even if the > + * pinned pages were unpinned later on, we'll still keep this bit set for the > + * lifecycle of this mm, just for simplicity. > + */ > +#define MMF_HAS_PINNED 28 /* FOLL_PIN has run, never cleared */ > #define MMF_DISABLE_THP_MASK (1 << MMF_DISABLE_THP) > > #define MMF_INIT_MASK (MMF_DUMPABLE_MASK | MMF_DUMP_FILTER_MASK |\ > diff --git a/kernel/fork.c b/kernel/fork.c > index 502dc046fbc62..a71e73707ef59 100644 > --- a/kernel/fork.c > +++ b/kernel/fork.c > @@ -1026,7 +1026,6 @@ static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p, > mm_pgtables_bytes_init(mm); > mm->map_count = 0; > mm->locked_vm = 0; > - atomic_set(&mm->has_pinned, 0); > atomic64_set(&mm->pinned_vm, 0); > memset(&mm->rss_stat, 0, sizeof(mm->rss_stat)); > spin_lock_init(&mm->page_table_lock); > diff --git a/mm/gup.c b/mm/gup.c > index 9933bc5c2eff2..bb130723a6717 100644 > --- a/mm/gup.c > +++ b/mm/gup.c > @@ -1270,6 +1270,17 @@ int fixup_user_fault(struct mm_struct *mm, > } > EXPORT_SYMBOL_GPL(fixup_user_fault); > > +/* > + * Set the MMF_HAS_PINNED if not set yet; after set it'll be there for the mm's > + * lifecycle. Avoid setting the bit unless necessary, or it might cause write > + * cache bouncing on large SMP machines for concurrent pinned gups. > + */ > +static inline void mm_set_has_pinned_flag(unsigned long *mm_flags) > +{ > + if (!test_bit(MMF_HAS_PINNED, mm_flags)) > + set_bit(MMF_HAS_PINNED, mm_flags); > +} > + > /* > * Please note that this function, unlike __get_user_pages will not > * return 0 for nr_pages > 0 without FOLL_NOWAIT > @@ -1292,8 +1303,8 @@ static __always_inline long __get_user_pages_locked(struct mm_struct *mm, > BUG_ON(*locked != 1); > } > > - if ((flags & FOLL_PIN) && !atomic_read(&mm->has_pinned)) > - atomic_set(&mm->has_pinned, 1); > + if (flags & FOLL_PIN) > + mm_set_has_pinned_flag(&mm->flags); > > /* > * FOLL_PIN and FOLL_GET are mutually exclusive. Traditional behavior > @@ -2617,8 +2628,8 @@ static int internal_get_user_pages_fast(unsigned long start, > FOLL_FAST_ONLY))) > return -EINVAL; > > - if ((gup_flags & FOLL_PIN) && !atomic_read(¤t->mm->has_pinned)) > - atomic_set(¤t->mm->has_pinned, 1); > + if (gup_flags & FOLL_PIN) > + mm_set_has_pinned_flag(¤t->mm->flags); Linux next tag next-20210512 builds failed on arm, riscv, mips and sh for the tinyconfig and allnoconfig due this patch. arm, mips, riscv and sh (tinyconfig) with gcc-8 arm, mips, riscv and sh (allnoconfig) with gcc-8 arm, mips, riscv and sh (tinyconfig) with gcc-9 arm, mips, riscv and sh (allnoconfig) with gcc-9 arm, mips, riscv and sh (tinyconfig) with gcc-10 arm, mips, riscv and sh (allnoconfig) with gcc-10 mm/gup.c: In function 'internal_get_user_pages_fast': mm/gup.c:2698:3: error: implicit declaration of function 'mm_set_has_pinned_flag' [-Werror=implicit-function-declaration] 2698 | mm_set_has_pinned_flag(¤t->mm->flags); | ^~~~~~~~~~~~~~~~~~~~~~ cc1: some warnings being treated as errors make[2]: *** [/builds/linux/scripts/Makefile.build:273: mm/gup.o] Error 1 Reported-by: Naresh Kamboju <naresh.kamboju@linaro.org> #regzb introduced: 354a2e3604e2 ("mm: gup: pack has_pinned in MMF_HAS_PINNED") Build url: https://gitlab.com/Linaro/lkft/mirrors/next/linux-next/-/jobs/1255567072#L315 -- Linaro LKFT https://lkft.linaro.org
On Wed, May 12, 2021 at 11:49:05AM +0200, Geert Uytterhoeven wrote: > Hi Peter, Andrea, Hi, Geert, Naresh, (Adding Naresh too since Naresh reported the same issue at the meantime) > > On Fri, May 7, 2021 at 7:26 PM Peter Xu <peterx@redhat.com> wrote: > > From: Andrea Arcangeli <aarcange@redhat.com> > > > > has_pinned 32bit can be packed in the MMF_HAS_PINNED bit as a noop > > cleanup. > > > > Any atomic_inc/dec to the mm cacheline shared by all threads in > > pin-fast would reintroduce a loss of SMP scalability to pin-fast, so > > there's no future potential usefulness to keep an atomic in the mm for > > this. > > > > set_bit(MMF_HAS_PINNED) will be theoretically a bit slower than > > WRITE_ONCE (atomic_set is equivalent to WRITE_ONCE), but the set_bit > > (just like atomic_set after this commit) has to be still issued only > > once per "mm", so the difference between the two will be lost in the > > noise. > > > > will-it-scale "mmap2" shows no change in performance with enterprise > > config as expected. > > > > will-it-scale "pin_fast" retains the > 4000% SMP scalability > > performance improvement against upstream as expected. > > > > This is a noop as far as overall performance and SMP scalability are > > concerned. > > > > Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> > > [peterx: Fix build for task_mmu.c, introduce mm_set_has_pinned_flag, fix > > comment here and there] > > Signed-off-by: Peter Xu <peterx@redhat.com> > > Thanks for your patch, which is now in linux-next. > > > diff --git a/mm/gup.c b/mm/gup.c > > index 9933bc5c2eff2..bb130723a6717 100644 > > --- a/mm/gup.c > > +++ b/mm/gup.c > > @@ -1270,6 +1270,17 @@ int fixup_user_fault(struct mm_struct *mm, > > } > > EXPORT_SYMBOL_GPL(fixup_user_fault); > > > > +/* > > + * Set the MMF_HAS_PINNED if not set yet; after set it'll be there for the mm's > > + * lifecycle. Avoid setting the bit unless necessary, or it might cause write > > + * cache bouncing on large SMP machines for concurrent pinned gups. > > + */ > > +static inline void mm_set_has_pinned_flag(unsigned long *mm_flags) > > +{ > > + if (!test_bit(MMF_HAS_PINNED, mm_flags)) > > + set_bit(MMF_HAS_PINNED, mm_flags); > > +} > > + > > /* > > * Please note that this function, unlike __get_user_pages will not > > * return 0 for nr_pages > 0 without FOLL_NOWAIT > > @@ -1292,8 +1303,8 @@ static __always_inline long __get_user_pages_locked(struct mm_struct *mm, > > BUG_ON(*locked != 1); > > } > > > > - if ((flags & FOLL_PIN) && !atomic_read(&mm->has_pinned)) > > - atomic_set(&mm->has_pinned, 1); > > + if (flags & FOLL_PIN) > > + mm_set_has_pinned_flag(&mm->flags); > > > > /* > > * FOLL_PIN and FOLL_GET are mutually exclusive. Traditional behavior > > @@ -2617,8 +2628,8 @@ static int internal_get_user_pages_fast(unsigned long start, > > FOLL_FAST_ONLY))) > > return -EINVAL; > > > > - if ((gup_flags & FOLL_PIN) && !atomic_read(¤t->mm->has_pinned)) > > - atomic_set(¤t->mm->has_pinned, 1); > > + if (gup_flags & FOLL_PIN) > > + mm_set_has_pinned_flag(¤t->mm->flags); > > noreply@ellerman.id.au reports: > > FAILED linux-next/m5272c3_defconfig/m68k-gcc8 Wed May 12, 19:30 > http://kisskb.ellerman.id.au/kisskb/buildresult/14543658/ > Commit: Add linux-next specific files for 20210512 > ec85c95b0c90a17413901b018e8ade7b9eae7cad > Compiler: m68k-linux-gcc (GCC) 8.1.0 / GNU ld (GNU Binutils) 2.30 > > mm/gup.c:2698:3: error: implicit declaration of function > 'mm_set_has_pinned_flag'; did you mean 'set_tsk_thread_flag'? > [-Werror=implicit-function-declaration] > > It's definition is inside the #ifdef CONFIG_MMU section, but the last > user isn't. Indeed that's wrong and I replied to the mm-commit email but not here to fix this up yesterday: https://lore.kernel.org/mm-commits/20210511220029.m6tGcxUIw%25akpm@linux-foundation.org/ I'll remember to reply to the thread next time. Sorry for that!
diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c index 4c95cc57a66a8..6144571942db9 100644 --- a/fs/proc/task_mmu.c +++ b/fs/proc/task_mmu.c @@ -1049,7 +1049,7 @@ static inline bool pte_is_pinned(struct vm_area_struct *vma, unsigned long addr, return false; if (!is_cow_mapping(vma->vm_flags)) return false; - if (likely(!atomic_read(&vma->vm_mm->has_pinned))) + if (likely(!test_bit(MMF_HAS_PINNED, &vma->vm_mm->flags))) return false; page = vm_normal_page(vma, addr, pte); if (!page) diff --git a/include/linux/mm.h b/include/linux/mm.h index d6790ab0cf575..94dc84f6d8658 100644 --- a/include/linux/mm.h +++ b/include/linux/mm.h @@ -1331,7 +1331,7 @@ static inline bool page_needs_cow_for_dma(struct vm_area_struct *vma, if (!is_cow_mapping(vma->vm_flags)) return false; - if (!atomic_read(&vma->vm_mm->has_pinned)) + if (!test_bit(MMF_HAS_PINNED, &vma->vm_mm->flags)) return false; return page_maybe_dma_pinned(page); diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h index 6613b26a88946..15d79858fadbd 100644 --- a/include/linux/mm_types.h +++ b/include/linux/mm_types.h @@ -435,16 +435,6 @@ struct mm_struct { */ atomic_t mm_count; - /** - * @has_pinned: Whether this mm has pinned any pages. This can - * be either replaced in the future by @pinned_vm when it - * becomes stable, or grow into a counter on its own. We're - * aggresive on this bit now - even if the pinned pages were - * unpinned later on, we'll still keep this bit set for the - * lifecycle of this mm just for simplicity. - */ - atomic_t has_pinned; - /** * @write_protect_seq: Locked when any thread is write * protecting pages mapped by this mm to enforce a later COW, diff --git a/include/linux/sched/coredump.h b/include/linux/sched/coredump.h index dfd82eab29025..4d9e3a6568758 100644 --- a/include/linux/sched/coredump.h +++ b/include/linux/sched/coredump.h @@ -73,6 +73,14 @@ static inline int get_dumpable(struct mm_struct *mm) #define MMF_OOM_VICTIM 25 /* mm is the oom victim */ #define MMF_OOM_REAP_QUEUED 26 /* mm was queued for oom_reaper */ #define MMF_MULTIPROCESS 27 /* mm is shared between processes */ +/* + * MMF_HAS_PINNED: Whether this mm has pinned any pages. This can be either + * replaced in the future by mm.pinned_vm when it becomes stable, or grow into + * a counter on its own. We're aggresive on this bit for now: even if the + * pinned pages were unpinned later on, we'll still keep this bit set for the + * lifecycle of this mm, just for simplicity. + */ +#define MMF_HAS_PINNED 28 /* FOLL_PIN has run, never cleared */ #define MMF_DISABLE_THP_MASK (1 << MMF_DISABLE_THP) #define MMF_INIT_MASK (MMF_DUMPABLE_MASK | MMF_DUMP_FILTER_MASK |\ diff --git a/kernel/fork.c b/kernel/fork.c index 502dc046fbc62..a71e73707ef59 100644 --- a/kernel/fork.c +++ b/kernel/fork.c @@ -1026,7 +1026,6 @@ static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p, mm_pgtables_bytes_init(mm); mm->map_count = 0; mm->locked_vm = 0; - atomic_set(&mm->has_pinned, 0); atomic64_set(&mm->pinned_vm, 0); memset(&mm->rss_stat, 0, sizeof(mm->rss_stat)); spin_lock_init(&mm->page_table_lock); diff --git a/mm/gup.c b/mm/gup.c index 9933bc5c2eff2..bb130723a6717 100644 --- a/mm/gup.c +++ b/mm/gup.c @@ -1270,6 +1270,17 @@ int fixup_user_fault(struct mm_struct *mm, } EXPORT_SYMBOL_GPL(fixup_user_fault); +/* + * Set the MMF_HAS_PINNED if not set yet; after set it'll be there for the mm's + * lifecycle. Avoid setting the bit unless necessary, or it might cause write + * cache bouncing on large SMP machines for concurrent pinned gups. + */ +static inline void mm_set_has_pinned_flag(unsigned long *mm_flags) +{ + if (!test_bit(MMF_HAS_PINNED, mm_flags)) + set_bit(MMF_HAS_PINNED, mm_flags); +} + /* * Please note that this function, unlike __get_user_pages will not * return 0 for nr_pages > 0 without FOLL_NOWAIT @@ -1292,8 +1303,8 @@ static __always_inline long __get_user_pages_locked(struct mm_struct *mm, BUG_ON(*locked != 1); } - if ((flags & FOLL_PIN) && !atomic_read(&mm->has_pinned)) - atomic_set(&mm->has_pinned, 1); + if (flags & FOLL_PIN) + mm_set_has_pinned_flag(&mm->flags); /* * FOLL_PIN and FOLL_GET are mutually exclusive. Traditional behavior @@ -2617,8 +2628,8 @@ static int internal_get_user_pages_fast(unsigned long start, FOLL_FAST_ONLY))) return -EINVAL; - if ((gup_flags & FOLL_PIN) && !atomic_read(¤t->mm->has_pinned)) - atomic_set(¤t->mm->has_pinned, 1); + if (gup_flags & FOLL_PIN) + mm_set_has_pinned_flag(¤t->mm->flags); if (!(gup_flags & FOLL_FAST_ONLY)) might_lock_read(¤t->mm->mmap_lock);