Re: kernel BUG at mm/page_alloc.c:LINE!

[Mel Gorman]

On Thu, Jan 17, 2019 at 09:33:09AM +0100, Vlastimil Babka wrote:
> > Code: 4c 39 e3 77 c0 0f b6 8d 74 ff ff ff b8 01 00 00 00 48 d3 e0 e9 11 fd  
> > ff ff 48 c7 c6 a0 63 52 88 4c 89 e7 e8 6a 14 10 00 0f 0b <0f> 0b 48 c7 c6  
> > c0 64 52 88 4c 89 e7 e8 57 14 10 00 0f 0b 48 89 cf
> > RSP: 0000:ffff8880a78e6f58 EFLAGS: 00010007
> > RAX: 0000000000000000 RBX: 0000000000000000 RCX: ffff88812fffc7e0
> > RDX: 1ffff11025fff8fc RSI: 0000000000000007 RDI: ffff88812fffc7b0
> > RBP: ffff8880a78e7018 R08: ffff8880a78ce000 R09: ffffed1014f1cdf2
> > R10: ffffed1014f1cdf1 R11: 0000000000000003 R12: ffff88812fffc7b0
> > R13: 1ffff11014f1cdf2 R14: ffff88812fffc7b0 R15: ffff8880a78e6ff0
> > FS:  0000000000000000(0000) GS:ffff8880ae600000(0000) knlGS:0000000000000000
> > CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> > CR2: 0000000000438ca0 CR3: 0000000009871000 CR4: 00000000001426f0
> > DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
> > DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000600
> > Call Trace:
> >   fast_isolate_freepages mm/compaction.c:1356 [inline]
> 
> Mel's new code... but might be just a victim of e.g. bad struct page
> initialization?
> 

The error looks like compaction found a !PageBuddy on the free lists
while the zone lock was held. That seems bad no matter what. I expect
there will be a respin of the entire series relatively soon but none of
the fixes so far would be for that level of damage.

-- 
Mel Gorman
SUSE Labs