mm/slub: suppress BUG messages for kmem_cache_alloc/kmem_cache_free
Subsystems (2)
PATCH 1/1 mm/slub: suppress BUG messages for kmem_cache_alloc/kmem_cache_free
After commit d2dc80750ee "mm/slub: optimize alloc/free fastpath by removing preemption on/off" and if CONFIG_DEBUG_PREEMPT is set, then huge amount of BUG messages like these happen:
BUG: using smp_processor_id() in preemptible [00000000] code: kjournald/171 caller is kmem_cache_alloc+0x41/0x132
and
BUG: using smp_processor_id() in preemptible [00000000] code: kdevtmpfs/12 caller is kmem_cache_free+0x5d/0x109
They are caused by this_cpu_ptr() that checks state of preemption. Because preemption checks are not necessary anymore in this code, then they are replaced with a raw_cpu_ptr() that does not check state of preemption.
---
These "BUGs" appear for the first time in next-20150119.
[ 0.947906] BUG: using smp_processor_id() in preemptible [00000000] code: kdevtmpfs/12 [ 0.952925] caller is kmem_cache_free+0x5d/0x109 [ 0.952931] CPU: 0 PID: 12 Comm: kdevtmpfs Tainted: G E 3.19.0-rc5-next-20150121-150119- #1 [ 0.952933] Hardware name: Bochs Bochs, BIOS Bochs 01/01/2011 [ 0.952936] 0000000000000000 ffff88000f8de1a0 ffffffff813d08ca 0000000000000000 [ 0.952940] ffffffff812036d7 0000000000000071 ffff88000c252000 ffff88000f801600 [ 0.952944] ffffea0000309400 ffffffff8113cf97 ffffffff811248e8 0000000000000000 [ 0.952948] Call Trace: [ 0.952958] [<ffffffff813d08ca>] ? dump_stack+0x4a/0x74 [ 0.952963] [<ffffffff812036d7>] ? check_preemption_disabled+0xd3/0xe4 [ 0.952968] [<ffffffff8113cf97>] ? do_path_lookup+0x47/0x52 [ 0.952971] [<ffffffff811248e8>] ? kmem_cache_free+0x5d/0x109 [ 0.952974] [<ffffffff8113cf97>] ? do_path_lookup+0x47/0x52 [ 0.952977] [<ffffffff8113cfca>] ? kern_path_create+0x28/0x117 [ 0.952983] [<ffffffff8105d38a>] ? preempt_count_sub+0xab/0xca [ 0.952987] [<ffffffff81061e5a>] ? __dequeue_entity+0x1e/0x32 [ 0.952993] [<ffffffff812d0422>] ? handle_create.isra.2+0x37/0x1b9 [ 0.952999] [<ffffffff810c8831>] ? trace_preempt_on+0xe/0x2f [ 0.953003] [<ffffffff8105d38a>] ? preempt_count_sub+0xab/0xca [ 0.953007] [<ffffffff813d1e8f>] ? __schedule+0x467/0x550 [ 0.953011] [<ffffffff810c8831>] ? trace_preempt_on+0xe/0x2f [ 0.953014] [<ffffffff812d05a4>] ? handle_create.isra.2+0x1b9/0x1b9 [ 0.953017] [<ffffffff812d068a>] ? devtmpfsd+0xe6/0x13b [ 0.953021] [<ffffffff81055ba1>] ? kthread+0x9e/0xa6 [ 0.953025] [<ffffffff81055b03>] ? __kthread_parkme+0x5c/0x5c [ 0.953030] [<ffffffff813d4c6c>] ? ret_from_fork+0x7c/0xb0 [ 0.953033] [<ffffffff81055b03>] ? __kthread_parkme+0x5c/0x5c
and
[ 0.942593] BUG: using smp_processor_id() in preemptible [00000000] code: kdevtmpfs/12 [ 0.945143] caller is kmem_cache_alloc+0x41/0x132 [ 0.945149] CPU: 0 PID: 12 Comm: kdevtmpfs Tainted: G E 3.19.0-rc5-next-20150121-150119- #1 [ 0.945151] Hardware name: Bochs Bochs, BIOS Bochs 01/01/2011 [ 0.945153] 0000000000000000 ffff88000f8de1a0 ffffffff813d08ca 0000000000000000 [ 0.945158] ffffffff812036d7 ffffffffffffffef 0000000000015f40 0000000000000010 [ 0.945162] 00000000000000d0 ffff88000f801600 ffffffff81124b43 ffffffff8114508f [ 0.945166] Call Trace: [ 0.945178] [<ffffffff813d08ca>] ? dump_stack+0x4a/0x74 [ 0.945184] [<ffffffff812036d7>] ? check_preemption_disabled+0xd3/0xe4 [ 0.945187] [<ffffffff81124b43>] ? kmem_cache_alloc+0x41/0x132 [ 0.945206] [<ffffffff8114508f>] ? inode_init_always+0xfc/0x19c [ 0.945211] [<ffffffff8113ce4c>] ? getname_kernel+0x29/0xe9 [ 0.945215] [<ffffffff8113ce4c>] ? getname_kernel+0x29/0xe9 [ 0.945219] [<ffffffff8113cf6d>] ? do_path_lookup+0x1d/0x52 [ 0.945223] [<ffffffff8113cfca>] ? kern_path_create+0x28/0x117 [ 0.945233] [<ffffffff810c8831>] ? trace_preempt_on+0xe/0x2f [ 0.945239] [<ffffffff8105d38a>] ? preempt_count_sub+0xab/0xca [ 0.945244] [<ffffffff81061e5a>] ? __dequeue_entity+0x1e/0x32 [ 0.945250] [<ffffffff812d0422>] ? handle_create.isra.2+0x37/0x1b9 [ 0.945254] [<ffffffff810c8831>] ? trace_preempt_on+0xe/0x2f [ 0.945257] [<ffffffff8105d38a>] ? preempt_count_sub+0xab/0xca [ 0.945262] [<ffffffff813d1e8f>] ? __schedule+0x467/0x550 [ 0.945266] [<ffffffff810c8831>] ? trace_preempt_on+0xe/0x2f [ 0.945269] [<ffffffff812d05a4>] ? handle_create.isra.2+0x1b9/0x1b9 [ 0.945273] [<ffffffff812d068a>] ? devtmpfsd+0xe6/0x13b [ 0.945278] [<ffffffff81055ba1>] ? kthread+0x9e/0xa6 [ 0.945282] [<ffffffff81055b03>] ? __kthread_parkme+0x5c/0x5c [ 0.945286] [<ffffffff813d4c6c>] ? ret_from_fork+0x7c/0xb0 [ 0.945290] [<ffffffff81055b03>] ? __kthread_parkme+0x5c/0x5c
mm/slub.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
mm/slub.c +2 -2
@@ -2404,7 +2404,7 @@ redo:*/ do { tid =this_cpu_read (s->cpu_slab->tid); - c =this_cpu_ptr (s->cpu_slab); + c =raw_cpu_ptr (s->cpu_slab); }while (IS_ENABLED (CONFIG_PREEMPT ) &&unlikely (tid != c->tid));/*
@@ -2670,7 +2670,7 @@ redo:*/ do { tid =this_cpu_read (s->cpu_slab->tid); - c =this_cpu_ptr (s->cpu_slab); + c =raw_cpu_ptr (s->cpu_slab); }while (IS_ENABLED (CONFIG_PREEMPT ) &&unlikely (tid != c->tid));/* Same with comment on barrier() in slab_alloc_node() */
Signature · 7 lines
1.7.10.4 -- 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>
David Rientjes
reviewerQuoted 14 lines · Andrey Skvortsov
mm/slub.c +1
-1@@ -2404,7 +2404,7 @@ redo:*/ do { tid =this_cpu_read (s->cpu_slab->tid); - c =this_cpu_ptr (s->cpu_slab); + c =raw_cpu_ptr (s->cpu_slab); }while (IS_ENABLED (CONFIG_PREEMPT ) &&unlikely (tid != c->tid));/*
mm/slub.c redo: · line 2676
Quoted 8 lines
mm/slub.c +1
-1@@ -2670,7 +2670,7 @@ redo:*/ do { tid =this_cpu_read (s->cpu_slab->tid); - c =this_cpu_ptr (s->cpu_slab); + c =raw_cpu_ptr (s->cpu_slab); }while (IS_ENABLED (CONFIG_PREEMPT ) &&unlikely (tid != c->tid));/* Same with comment on barrier() in slab_alloc_node() */
This should already be fixed with
http://ozlabs.org/~akpm/mmotm/broken-out/mm-slub-optimize-alloc-free-fastpath-by-removing-preemption-on-off-v3.patch
You can find the latest mmotm, which was just released, at
http://ozlabs.org/~akpm/mmotm and it should be in linux-next tomorrow.
--
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 23 lines · David Rientjes
Quoted 21 lines · Andrey Skvortsov
mm/slub.c +2
-2@@ -2404,7 +2404,7 @@ redo:*/ do { tid =this_cpu_read (s->cpu_slab->tid); - c =this_cpu_ptr (s->cpu_slab); + c =raw_cpu_ptr (s->cpu_slab); }while (IS_ENABLED (CONFIG_PREEMPT ) &&unlikely (tid != c->tid));/* @@ -2670,7 +2670,7 @@ redo:*/ do { tid =this_cpu_read (s->cpu_slab->tid); - c =this_cpu_ptr (s->cpu_slab); + c =raw_cpu_ptr (s->cpu_slab); }while (IS_ENABLED (CONFIG_PREEMPT ) &&unlikely (tid != c->tid));
Quoted 7 lines
mm/slub.c redo: · line 2676
Quoted 1 line
mm/slub.c +0
-0/* Same with comment on barrier() in slab_alloc_node() */ This should already be fixed with
http://ozlabs.org/~akpm/mmotm/broken-out/mm-slub-optimize-alloc-free-fastpath-by-removing-preemption-on-off-v3.patchYou can find the latest mmotm, which was just released, at
http://ozlabs.org/~akpm/mmotm and it should be in linux-next tomorrow.
ok. I've just looked at linux-next/master and linux-next/akpm branches and that was not fixed there. Thanks for the link. I'll look there in the future for mm-related patches posting a new one.
Signature · 5 lines
Best regards, Andrey Skvortsov Secure eMail with gnupg: See http://www.gnupg.org/ PGP Key ID: 0x57A3AEAD