Re: BUG: scheduling while atomic: cron/668/0x10c9a0c0 (was: Re: mm, page_alloc: avoid looking up the first zone in a zonelist twice)

[Mel Gorman]

On Mon, May 30, 2016 at 03:13:40PM +0200, Geert Uytterhoeven wrote:
> >     The benefit is negligible and the results are within the noise but each
> >     cycle counts.
> >
> >     Signed-off-by: Mel Gorman <mgorman@xxxxxxxxxxxxxxxxxxx>
> >     Cc: Vlastimil Babka <vbabka@xxxxxxx>
> >     Cc: Jesper Dangaard Brouer <brouer@xxxxxxxxxx>
> >     Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
> >     Signed-off-by: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx>
> 
> About one week ago, I started seeing an obscure intermittent crash during
> system shutdown on m68k/ARAnyM using atari_defconfig.
> The crash isn't 100% reproducible, but it happens during ca. 1 out of 5
> shutdowns.
> 
> I finally managed to bisect it to the above commit.
> I did verify that the parent commit didn't crash after 60 tries.
> Unfortunately I couldn't revert the offending commit on top of v4.7-rc1, due to
> conflicting changes.
> 
> Do you have any idea what's going wrong?

There isn't anything obvious from the crash log you showed but can you
try the following just in case?

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index dba8cfd0b2d6..f2c1e47adc11 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -3232,6 +3232,9 @@ __alloc_pages_slowpath(gfp_t gfp_mask, unsigned int order,
 		 * allocations are system rather than user orientated
 		 */
 		ac->zonelist = node_zonelist(numa_node_id(), gfp_mask);
+		ac->preferred_zoneref = first_zones_zonelist(ac->zonelist,
+					ac->high_zoneidx, ac->nodemask);
+		ac->classzone_idx = zonelist_zone_idx(ac->preferred_zoneref);
 		page = get_page_from_freelist(gfp_mask, order,
 						ALLOC_NO_WATERMARKS, ac);
 		if (page)