| Message ID | 20220712022807.44113-1-rongwei.wang@linux.alibaba.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | [v2,1/3] mm/slub: fix the race between validate_slab and slab_free | expand |
Hi
According to all's point in PATCH v1 [1], I rewrote the first patch
"mm/slub: fix the race between validate_slab and slab_free". And now,
these changes only works when DEBUG SLUB enabled. Plus, here some
performance test can been found in [2] (Thanks Christoph's suggestion).
changelog
v1->v2:
*mm/slub: fix the race between validate_slab and slab_free
make these changes can work when debug slub enabled.
*mm/slub: improve consistency of nr_slabs count
nothing
*mm/slub: delete confusing pr_err when debugging slub
only deleting the confusing pr_err().
For convenient, showing the latest test data here (copy from [2]):
testcase used: https://github.com/netoptimizer/prototype-kernel.git
(slab_test)
Single thread testing
1. Kmalloc: Repeatedly allocate then free test
before fix
kmalloc kfree kmalloc kfree
10000 times 8 4 cycles 5 cycles 4 cycles 5 cycles
10000 times 16 3 cycles 5 cycles 3 cycles 5 cycles
10000 times 32 3 cycles 5 cycles 3 cycles 5 cycles
10000 times 64 3 cycles 5 cycles 3 cycles 5 cycles
10000 times 128 3 cycles 5 cycles 3 cycles 5 cycles
10000 times 256 14 cycles 9 cycles 6 cycles 8 cycles
10000 times 512 9 cycles 8 cycles 9 cycles 10 cycles
10000 times 1024 48 cycles 10 cycles 6 cycles 10 cycles
10000 times 2048 31 cycles 12 cycles 35 cycles 13 cycles
10000 times 4096 96 cycles 17 cycles 96 cycles 18 cycles
10000 times 8192 188 cycles 27 cycles 190 cycles 27 cycles
10000 times 16384 117 cycles 38 cycles 115 cycles 38 cycles
2. Kmalloc: alloc/free test
before fix
10000 times kmalloc(8)/kfree 3 cycles 3 cycles
10000 times kmalloc(16)/kfree 3 cycles 3 cycles
10000 times kmalloc(32)/kfree 3 cycles 3 cycles
10000 times kmalloc(64)/kfree 3 cycles 3 cycles
10000 times kmalloc(128)/kfree 3 cycles 3 cycles
10000 times kmalloc(256)/kfree 3 cycles 3 cycles
10000 times kmalloc(512)/kfree 3 cycles 3 cycles
10000 times kmalloc(1024)/kfree 3 cycles 3 cycles
10000 times kmalloc(2048)/kfree 3 cycles 3 cycles
10000 times kmalloc(4096)/kfree 3 cycles 3 cycles
10000 times kmalloc(8192)/kfree 3 cycles 3 cycles
10000 times kmalloc(16384)/kfree 33 cycles 33 cycles
Concurrent allocs
before fix
Kmalloc N*alloc N*free(8) Average=13/14 Average=14/15
Kmalloc N*alloc N*free(16) Average=13/15 Average=13/15
Kmalloc N*alloc N*free(32) Average=13/15 Average=13/15
Kmalloc N*alloc N*free(64) Average=13/15 Average=13/15
Kmalloc N*alloc N*free(128) Average=13/15 Average=13/15
Kmalloc N*alloc N*free(256) Average=137/29 Average=134/39
Kmalloc N*alloc N*free(512) Average=61/29 Average=64/28
Kmalloc N*alloc N*free(1024) Average=465/50 Average=656/55
Kmalloc N*alloc N*free(2048) Average=503/97 Average=422/97
Kmalloc N*alloc N*free(4096) Average=1592/206 Average=1624/207
Kmalloc N*(alloc free)(8) Average=3 Average=3
Kmalloc N*(alloc free)(16) Average=3 Average=3
Kmalloc N*(alloc free)(32) Average=3 Average=3
Kmalloc N*(alloc free)(64) Average=3 Average=3
Kmalloc N*(alloc free)(128) Average=3 Average=3
Kmalloc N*(alloc free)(256) Average=3 Average=3
Kmalloc N*(alloc free)(512) Average=3 Average=3
Kmalloc N*(alloc free)(1024) Average=3 Average=3
Kmalloc N*(alloc free)(2048) Average=3 Average=3
Kmalloc N*(alloc free)(4096) Average=3 Average=3
The above data seems indicate that this modification (only works when
kmem_cache_debug(s) is true) does not introduce significant performance
impact. And if you have better suggestion of testcase, please let me
know, Thanks!
[1]
https://lore.kernel.org/linux-mm/alpine.DEB.2.22.394.2206081417370.465021@gentwo.de/T/#m2832b1983a229183aabfd6eb71a2eb39ecd0d08a
[2]
https://lore.kernel.org/linux-mm/alpine.DEB.2.22.394.2206081417370.465021@gentwo.de/T/#m75f1f32ad590fb13ac9e771030fafd15c7db8cb1
Thanks for your time!
On 7/12/22 10:28 AM, Rongwei Wang wrote:
> In use cases where allocating and freeing slab frequently, some
> error messages, such as "Left Redzone overwritten", "First byte
> 0xbb instead of 0xcc" would be printed when validating slabs.
> That's because an object has been filled with SLAB_RED_INACTIVE,
> but has not been added to slab's freelist. And between these
> two states, the behaviour of validating slab is likely to occur.
>
> Actually, it doesn't mean the slab can not work stably. But, these
> confusing messages will disturb slab debugging more or less.
>
> Signed-off-by: Rongwei Wang <rongwei.wang@linux.alibaba.com>
> ---
> mm/slub.c | 43 +++++++++++++++++++++++++------------------
> 1 file changed, 25 insertions(+), 18 deletions(-)
>
> diff --git a/mm/slub.c b/mm/slub.c
> index b1281b8654bd..e950d8df8380 100644
> --- a/mm/slub.c
> +++ b/mm/slub.c
> @@ -1391,18 +1391,16 @@ static noinline int free_debug_processing(
> void *head, void *tail, int bulk_cnt,
> unsigned long addr)
> {
> - struct kmem_cache_node *n = get_node(s, slab_nid(slab));
> void *object = head;
> int cnt = 0;
> - unsigned long flags, flags2;
> + unsigned long flags;
> int ret = 0;
> depot_stack_handle_t handle = 0;
>
> if (s->flags & SLAB_STORE_USER)
> handle = set_track_prepare();
>
> - spin_lock_irqsave(&n->list_lock, flags);
> - slab_lock(slab, &flags2);
> + slab_lock(slab, &flags);
>
> if (s->flags & SLAB_CONSISTENCY_CHECKS) {
> if (!check_slab(s, slab))
> @@ -1435,8 +1433,7 @@ static noinline int free_debug_processing(
> slab_err(s, slab, "Bulk freelist count(%d) invalid(%d)\n",
> bulk_cnt, cnt);
>
> - slab_unlock(slab, &flags2);
> - spin_unlock_irqrestore(&n->list_lock, flags);
> + slab_unlock(slab, &flags);
> if (!ret)
> slab_fix(s, "Object at 0x%p not freed", object);
> return ret;
> @@ -3330,7 +3327,7 @@ static void __slab_free(struct kmem_cache *s, struct slab *slab,
>
> {
> void *prior;
> - int was_frozen;
> + int was_frozen, to_take_off = 0;
> struct slab new;
> unsigned long counters;
> struct kmem_cache_node *n = NULL;
> @@ -3341,14 +3338,23 @@ static void __slab_free(struct kmem_cache *s, struct slab *slab,
> if (kfence_free(head))
> return;
>
> - if (kmem_cache_debug(s) &&
> - !free_debug_processing(s, slab, head, tail, cnt, addr))
> - return;
> + n = get_node(s, slab_nid(slab));
> + if (kmem_cache_debug(s)) {
> + int ret;
>
> - do {
> - if (unlikely(n)) {
> + spin_lock_irqsave(&n->list_lock, flags);
> + ret = free_debug_processing(s, slab, head, tail, cnt, addr);
> + if (!ret) {
> spin_unlock_irqrestore(&n->list_lock, flags);
> - n = NULL;
> + return;
> + }
> + }
> +
> + do {
> + if (unlikely(to_take_off)) {
> + if (!kmem_cache_debug(s))
> + spin_unlock_irqrestore(&n->list_lock, flags);
> + to_take_off = 0;
> }
> prior = slab->freelist;
> counters = slab->counters;
> @@ -3369,8 +3375,6 @@ static void __slab_free(struct kmem_cache *s, struct slab *slab,
> new.frozen = 1;
>
> } else { /* Needs to be taken off a list */
> -
> - n = get_node(s, slab_nid(slab));
> /*
> * Speculatively acquire the list_lock.
> * If the cmpxchg does not succeed then we may
> @@ -3379,8 +3383,10 @@ static void __slab_free(struct kmem_cache *s, struct slab *slab,
> * Otherwise the list_lock will synchronize with
> * other processors updating the list of slabs.
> */
> - spin_lock_irqsave(&n->list_lock, flags);
> + if (!kmem_cache_debug(s))
> + spin_lock_irqsave(&n->list_lock, flags);
>
> + to_take_off = 1;
> }
> }
>
> @@ -3389,8 +3395,9 @@ static void __slab_free(struct kmem_cache *s, struct slab *slab,
> head, new.counters,
> "__slab_free"));
>
> - if (likely(!n)) {
> -
> + if (likely(!to_take_off)) {
> + if (kmem_cache_debug(s))
> + spin_unlock_irqrestore(&n->list_lock, flags);
> if (likely(was_frozen)) {
> /*
> * The list lock was not taken therefore no list
On Tue, Jul 12, 2022 at 10:28:05AM +0800, Rongwei Wang wrote: > In use cases where allocating and freeing slab frequently, some > error messages, such as "Left Redzone overwritten", "First byte > 0xbb instead of 0xcc" would be printed when validating slabs. > That's because an object has been filled with SLAB_RED_INACTIVE, > but has not been added to slab's freelist. And between these > two states, the behaviour of validating slab is likely to occur. > > Actually, it doesn't mean the slab can not work stably. But, these > confusing messages will disturb slab debugging more or less. > > Signed-off-by: Rongwei Wang <rongwei.wang@linux.alibaba.com> > --- > mm/slub.c | 43 +++++++++++++++++++++++++------------------ > 1 file changed, 25 insertions(+), 18 deletions(-) > This makes the code more complex. A part of me says it may be more pleasant to split implementation allocating from caches for debugging. That would make it simpler. something like: __slab_alloc() { if (kmem_cache_debug(s)) slab_alloc_debug() else ___slab_alloc() } slab_free() { if (kmem_cache_debug(s)) slab_free_debug() else __do_slab_free() } See also: https://lore.kernel.org/lkml/faf416b9-f46c-8534-7fb7-557c046a564d@suse.cz/ > diff --git a/mm/slub.c b/mm/slub.c > index b1281b8654bd..e950d8df8380 100644 > --- a/mm/slub.c > +++ b/mm/slub.c > @@ -1391,18 +1391,16 @@ static noinline int free_debug_processing( > void *head, void *tail, int bulk_cnt, > unsigned long addr) > { > - struct kmem_cache_node *n = get_node(s, slab_nid(slab)); > void *object = head; > int cnt = 0; > - unsigned long flags, flags2; > + unsigned long flags; > int ret = 0; > depot_stack_handle_t handle = 0; > > if (s->flags & SLAB_STORE_USER) > handle = set_track_prepare(); > > - spin_lock_irqsave(&n->list_lock, flags); > - slab_lock(slab, &flags2); > + slab_lock(slab, &flags); > > if (s->flags & SLAB_CONSISTENCY_CHECKS) { > if (!check_slab(s, slab)) > @@ -1435,8 +1433,7 @@ static noinline int free_debug_processing( > slab_err(s, slab, "Bulk freelist count(%d) invalid(%d)\n", > bulk_cnt, cnt); > > - slab_unlock(slab, &flags2); > - spin_unlock_irqrestore(&n->list_lock, flags); > + slab_unlock(slab, &flags); > if (!ret) > slab_fix(s, "Object at 0x%p not freed", object); > return ret; > @@ -3330,7 +3327,7 @@ static void __slab_free(struct kmem_cache *s, struct slab *slab, > > { > void *prior; > - int was_frozen; > + int was_frozen, to_take_off = 0; > struct slab new; > unsigned long counters; > struct kmem_cache_node *n = NULL; > @@ -3341,14 +3338,23 @@ static void __slab_free(struct kmem_cache *s, struct slab *slab, > if (kfence_free(head)) > return; > > - if (kmem_cache_debug(s) && > - !free_debug_processing(s, slab, head, tail, cnt, addr)) > - return; > + n = get_node(s, slab_nid(slab)); > + if (kmem_cache_debug(s)) { > + int ret; > > - do { > - if (unlikely(n)) { > + spin_lock_irqsave(&n->list_lock, flags); > + ret = free_debug_processing(s, slab, head, tail, cnt, addr); > + if (!ret) { > spin_unlock_irqrestore(&n->list_lock, flags); > - n = NULL; > + return; > + } > + } > + > + do { > + if (unlikely(to_take_off)) { > + if (!kmem_cache_debug(s)) > + spin_unlock_irqrestore(&n->list_lock, flags); > + to_take_off = 0; > } > prior = slab->freelist; > counters = slab->counters; > @@ -3369,8 +3375,6 @@ static void __slab_free(struct kmem_cache *s, struct slab *slab, > new.frozen = 1; > > } else { /* Needs to be taken off a list */ > - > - n = get_node(s, slab_nid(slab)); > /* > * Speculatively acquire the list_lock. > * If the cmpxchg does not succeed then we may > @@ -3379,8 +3383,10 @@ static void __slab_free(struct kmem_cache *s, struct slab *slab, > * Otherwise the list_lock will synchronize with > * other processors updating the list of slabs. > */ > - spin_lock_irqsave(&n->list_lock, flags); > + if (!kmem_cache_debug(s)) > + spin_lock_irqsave(&n->list_lock, flags); > > + to_take_off = 1; > } > } > > @@ -3389,8 +3395,9 @@ static void __slab_free(struct kmem_cache *s, struct slab *slab, > head, new.counters, > "__slab_free")); > > - if (likely(!n)) { > - > + if (likely(!to_take_off)) { > + if (kmem_cache_debug(s)) > + spin_unlock_irqrestore(&n->list_lock, flags); > if (likely(was_frozen)) { > /* > * The list lock was not taken therefore no list > -- > 2.27.0 >
On 7/13/22 6:22 PM, Hyeonggon Yoo wrote: > On Tue, Jul 12, 2022 at 10:28:05AM +0800, Rongwei Wang wrote: >> In use cases where allocating and freeing slab frequently, some >> error messages, such as "Left Redzone overwritten", "First byte >> 0xbb instead of 0xcc" would be printed when validating slabs. >> That's because an object has been filled with SLAB_RED_INACTIVE, >> but has not been added to slab's freelist. And between these >> two states, the behaviour of validating slab is likely to occur. >> >> Actually, it doesn't mean the slab can not work stably. But, these >> confusing messages will disturb slab debugging more or less. >> >> Signed-off-by: Rongwei Wang <rongwei.wang@linux.alibaba.com> >> --- >> mm/slub.c | 43 +++++++++++++++++++++++++------------------ >> 1 file changed, 25 insertions(+), 18 deletions(-) >> > > This makes the code more complex. > > A part of me says it may be more pleasant to split implementation > allocating from caches for debugging. That would make it simpler. > > something like: > > __slab_alloc() { > if (kmem_cache_debug(s)) > slab_alloc_debug() > else > ___slab_alloc() > } > > slab_free() { > if (kmem_cache_debug(s)) > slab_free_debug() > else > __do_slab_free() > } Oh, I also have same idea, but not sure whether it is accepted because of it needs more changes than now. Since you agree with this way, I can rewrite this code. Thanks. > > See also: > https://lore.kernel.org/lkml/faf416b9-f46c-8534-7fb7-557c046a564d@suse.cz/ Thanks, it seems that I had missed it. > >> diff --git a/mm/slub.c b/mm/slub.c >> index b1281b8654bd..e950d8df8380 100644 >> --- a/mm/slub.c >> +++ b/mm/slub.c >> @@ -1391,18 +1391,16 @@ static noinline int free_debug_processing( >> void *head, void *tail, int bulk_cnt, >> unsigned long addr) >> { >> - struct kmem_cache_node *n = get_node(s, slab_nid(slab)); >> void *object = head; >> int cnt = 0; >> - unsigned long flags, flags2; >> + unsigned long flags; >> int ret = 0; >> depot_stack_handle_t handle = 0; >> >> if (s->flags & SLAB_STORE_USER) >> handle = set_track_prepare(); >> >> - spin_lock_irqsave(&n->list_lock, flags); >> - slab_lock(slab, &flags2); >> + slab_lock(slab, &flags); >> >> if (s->flags & SLAB_CONSISTENCY_CHECKS) { >> if (!check_slab(s, slab)) >> @@ -1435,8 +1433,7 @@ static noinline int free_debug_processing( >> slab_err(s, slab, "Bulk freelist count(%d) invalid(%d)\n", >> bulk_cnt, cnt); >> >> - slab_unlock(slab, &flags2); >> - spin_unlock_irqrestore(&n->list_lock, flags); >> + slab_unlock(slab, &flags); >> if (!ret) >> slab_fix(s, "Object at 0x%p not freed", object); >> return ret; >> @@ -3330,7 +3327,7 @@ static void __slab_free(struct kmem_cache *s, struct slab *slab, >> >> { >> void *prior; >> - int was_frozen; >> + int was_frozen, to_take_off = 0; >> struct slab new; >> unsigned long counters; >> struct kmem_cache_node *n = NULL; >> @@ -3341,14 +3338,23 @@ static void __slab_free(struct kmem_cache *s, struct slab *slab, >> if (kfence_free(head)) >> return; >> >> - if (kmem_cache_debug(s) && >> - !free_debug_processing(s, slab, head, tail, cnt, addr)) >> - return; >> + n = get_node(s, slab_nid(slab)); >> + if (kmem_cache_debug(s)) { >> + int ret; >> >> - do { >> - if (unlikely(n)) { >> + spin_lock_irqsave(&n->list_lock, flags); >> + ret = free_debug_processing(s, slab, head, tail, cnt, addr); >> + if (!ret) { >> spin_unlock_irqrestore(&n->list_lock, flags); >> - n = NULL; >> + return; >> + } >> + } >> + >> + do { >> + if (unlikely(to_take_off)) { >> + if (!kmem_cache_debug(s)) >> + spin_unlock_irqrestore(&n->list_lock, flags); >> + to_take_off = 0; >> } >> prior = slab->freelist; >> counters = slab->counters; >> @@ -3369,8 +3375,6 @@ static void __slab_free(struct kmem_cache *s, struct slab *slab, >> new.frozen = 1; >> >> } else { /* Needs to be taken off a list */ >> - >> - n = get_node(s, slab_nid(slab)); >> /* >> * Speculatively acquire the list_lock. >> * If the cmpxchg does not succeed then we may >> @@ -3379,8 +3383,10 @@ static void __slab_free(struct kmem_cache *s, struct slab *slab, >> * Otherwise the list_lock will synchronize with >> * other processors updating the list of slabs. >> */ >> - spin_lock_irqsave(&n->list_lock, flags); >> + if (!kmem_cache_debug(s)) >> + spin_lock_irqsave(&n->list_lock, flags); >> >> + to_take_off = 1; >> } >> } >> >> @@ -3389,8 +3395,9 @@ static void __slab_free(struct kmem_cache *s, struct slab *slab, >> head, new.counters, >> "__slab_free")); >> >> - if (likely(!n)) { >> - >> + if (likely(!to_take_off)) { >> + if (kmem_cache_debug(s)) >> + spin_unlock_irqrestore(&n->list_lock, flags); >> if (likely(was_frozen)) { >> /* >> * The list lock was not taken therefore no list >> -- >> 2.27.0 >>
diff --git a/mm/slub.c b/mm/slub.c index b1281b8654bd..e950d8df8380 100644 --- a/mm/slub.c +++ b/mm/slub.c @@ -1391,18 +1391,16 @@ static noinline int free_debug_processing( void *head, void *tail, int bulk_cnt, unsigned long addr) { - struct kmem_cache_node *n = get_node(s, slab_nid(slab)); void *object = head; int cnt = 0; - unsigned long flags, flags2; + unsigned long flags; int ret = 0; depot_stack_handle_t handle = 0; if (s->flags & SLAB_STORE_USER) handle = set_track_prepare(); - spin_lock_irqsave(&n->list_lock, flags); - slab_lock(slab, &flags2); + slab_lock(slab, &flags); if (s->flags & SLAB_CONSISTENCY_CHECKS) { if (!check_slab(s, slab)) @@ -1435,8 +1433,7 @@ static noinline int free_debug_processing( slab_err(s, slab, "Bulk freelist count(%d) invalid(%d)\n", bulk_cnt, cnt); - slab_unlock(slab, &flags2); - spin_unlock_irqrestore(&n->list_lock, flags); + slab_unlock(slab, &flags); if (!ret) slab_fix(s, "Object at 0x%p not freed", object); return ret; @@ -3330,7 +3327,7 @@ static void __slab_free(struct kmem_cache *s, struct slab *slab, { void *prior; - int was_frozen; + int was_frozen, to_take_off = 0; struct slab new; unsigned long counters; struct kmem_cache_node *n = NULL; @@ -3341,14 +3338,23 @@ static void __slab_free(struct kmem_cache *s, struct slab *slab, if (kfence_free(head)) return; - if (kmem_cache_debug(s) && - !free_debug_processing(s, slab, head, tail, cnt, addr)) - return; + n = get_node(s, slab_nid(slab)); + if (kmem_cache_debug(s)) { + int ret; - do { - if (unlikely(n)) { + spin_lock_irqsave(&n->list_lock, flags); + ret = free_debug_processing(s, slab, head, tail, cnt, addr); + if (!ret) { spin_unlock_irqrestore(&n->list_lock, flags); - n = NULL; + return; + } + } + + do { + if (unlikely(to_take_off)) { + if (!kmem_cache_debug(s)) + spin_unlock_irqrestore(&n->list_lock, flags); + to_take_off = 0; } prior = slab->freelist; counters = slab->counters; @@ -3369,8 +3375,6 @@ static void __slab_free(struct kmem_cache *s, struct slab *slab, new.frozen = 1; } else { /* Needs to be taken off a list */ - - n = get_node(s, slab_nid(slab)); /* * Speculatively acquire the list_lock. * If the cmpxchg does not succeed then we may @@ -3379,8 +3383,10 @@ static void __slab_free(struct kmem_cache *s, struct slab *slab, * Otherwise the list_lock will synchronize with * other processors updating the list of slabs. */ - spin_lock_irqsave(&n->list_lock, flags); + if (!kmem_cache_debug(s)) + spin_lock_irqsave(&n->list_lock, flags); + to_take_off = 1; } } @@ -3389,8 +3395,9 @@ static void __slab_free(struct kmem_cache *s, struct slab *slab, head, new.counters, "__slab_free")); - if (likely(!n)) { - + if (likely(!to_take_off)) { + if (kmem_cache_debug(s)) + spin_unlock_irqrestore(&n->list_lock, flags); if (likely(was_frozen)) { /* * The list lock was not taken therefore no list
In use cases where allocating and freeing slab frequently, some error messages, such as "Left Redzone overwritten", "First byte 0xbb instead of 0xcc" would be printed when validating slabs. That's because an object has been filled with SLAB_RED_INACTIVE, but has not been added to slab's freelist. And between these two states, the behaviour of validating slab is likely to occur. Actually, it doesn't mean the slab can not work stably. But, these confusing messages will disturb slab debugging more or less. Signed-off-by: Rongwei Wang <rongwei.wang@linux.alibaba.com> --- mm/slub.c | 43 +++++++++++++++++++++++++------------------ 1 file changed, 25 insertions(+), 18 deletions(-)