Re: mmotm: mm-slub-optimize-alloc-free-fastpath-by-removing-preemption-on-off.patch is causing preemptible splats

[Andrew Morton]

On Wed, 21 Jan 2015 15:11:38 +0100 Michal Hocko <mhocko@xxxxxxx> 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@xxxxxxx>
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@xxxxxxx>
Acked-by: Christoph Lameter <cl@xxxxxxxxx>
Acked-by: Jesper Dangaard Brouer <brouer@xxxxxxxxxx>
Tested-by: Jesper Dangaard Brouer <brouer@xxxxxxxxxx>
Cc: Pekka Enberg <penberg@xxxxxxxxxx>
Cc: David Rientjes <rientjes@xxxxxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---

 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() */
_

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/