Messages in this thread Patch in this message | | | Date | Wed, 21 Jan 2015 14:21:07 -0800 | From | Andrew Morton <> | Subject | Re: mmotm: mm-slub-optimize-alloc-free-fastpath-by-removing-preemption-on-off.patch is causing preemptible splats |
| |
On Wed, 21 Jan 2015 15:11:38 +0100 Michal Hocko <mhocko@suse.cz> wrote:
> On Wed 21-01-15 15:06:03, Krzysztof Koz__owski wrote: > [...] > > Same here :) [1] . So actually only ARM seems affected (both armv7 and > > armv8) because it is the only one which uses smp_processor_id() in > > my_cpu_offset. > > This was on x86_64 with CONFIG_DEBUG_PREEMPT so it is not only ARM > specific. >
Hopefully mm-slub-optimize-alloc-free-fastpath-by-removing-preemption-on-off-v3.patch will fix this.
The most recent -mmotm was a bit of a trainwreck. I'm scrambling to get the holes plugged so I can get another mmotm out today.
From: Joonsoo Kim <iamjoonsoo.kim@lge.com> Subject: mm/slub: optimize alloc/free fastpath by removing preemption on/off
Change from v2: - use raw_cpu_ptr() rather than this_cpu_ptr() to avoid warning from preemption debug check since this is intended behaviour - fix typo alogorithm -> algorithm
Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com> Acked-by: Christoph Lameter <cl@linux.com> Acked-by: Jesper Dangaard Brouer <brouer@redhat.com> Tested-by: Jesper Dangaard Brouer <brouer@redhat.com> Cc: Pekka Enberg <penberg@kernel.org> Cc: David Rientjes <rientjes@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> ---
mm/slub.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
diff -puN mm/slub.c~mm-slub-optimize-alloc-free-fastpath-by-removing-preemption-on-off-v3 mm/slub.c --- a/mm/slub.c~mm-slub-optimize-alloc-free-fastpath-by-removing-preemption-on-off-v3 +++ a/mm/slub.c @@ -2404,11 +2404,11 @@ 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)); /* - * Irqless object alloc/free alogorithm used here depends on sequence + * Irqless object alloc/free algorithm used here depends on sequence * of fetching cpu_slab's data. tid should be fetched before anything * on c to guarantee that object and page associated with previous tid * won't be used with current tid. If we fetch tid first, object and @@ -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() */ _
| |