Re: btrfs btree_ctree_super fault

[Chris Mason]

On 11/08/2016 09:59 AM, Dave Jones wrote:
> On Sun, Nov 06, 2016 at 11:55:39AM -0500, Dave Jones wrote:
>  > <subject changed, hopefully we're done with bio corruption for now>
>  >
>  > On Mon, Oct 31, 2016 at 01:44:55PM -0600, Chris Mason wrote:
>  >  > On Mon, Oct 31, 2016 at 12:35:16PM -0700, Linus Torvalds wrote:
>  >  > >On Mon, Oct 31, 2016 at 11:55 AM, Dave Jones <davej@xxxxxxxxxxxxxxxxx> wrote:
>  >  > >>
>  >  > >> BUG: Bad page state in process kworker/u8:12  pfn:4e0e39
>  >  > >> page:ffffea0013838e40 count:0 mapcount:0 mapping:ffff8804a20310e0 index:0x100c
>  >  > >> flags: 0x400000000000000c(referenced|uptodate)
>  >  > >> page dumped because: non-NULL mapping
>  >  > >
>  >  > >Hmm. So this seems to be btrfs-specific, right?
>  >  > >
>  >  > >I searched for all your "non-NULL mapping" cases, and they all seem to
>  >  > >have basically the same call trace, with some work thread doing
>  >  > >writeback and going through btrfs_writepages().
>  >  > >
>  >  > >Sounds like it's a race with either fallocate hole-punching or
>  >  > >truncate. I'm not seeing it, but I suspect it's btrfs, since DaveJ
>  >  > >clearly ran other filesystems too but I am not seeing this backtrace
>  >  > >for anything else.
>  >  >
>  >  > Agreed, I think this is a separate bug, almost certainly btrfs specific.
>  >  > I'll work with Dave on a better reproducer.
>  >
>  > Still refining my 'capture ftrace when trinity detects taint' feature,
>  > but in the meantime, here's a variant I don't think we've seen before:
>
> And another new one:
>
> kernel BUG at fs/btrfs/ctree.c:3172!
> invalid opcode: 0000 [#1] PREEMPT SMP DEBUG_PAGEALLOC
> CPU: 0 PID: 22702 Comm: trinity-c40 Not tainted 4.9.0-rc4-think+ #1
> task: ffff8804ffde37c0 task.stack: ffffc90002188000
> RIP: 0010:[<ffffffffa00576b9>]
>   [<ffffffffa00576b9>] btrfs_set_item_key_safe+0x179/0x190 [btrfs]
> RSP: 0000:ffffc9000218b8a8  EFLAGS: 00010246
> RAX: 0000000000000000 RBX: ffff8804fddcf348 RCX: 0000000000001000
> RDX: 0000000000000000 RSI: ffffc9000218b9ce RDI: ffffc9000218b8c7
> RBP: ffffc9000218b908 R08: 0000000000004000 R09: ffffc9000218b8c8
> R10: 0000000000000000 R11: 0000000000000001 R12: ffffc9000218b8b6
> R13: ffffc9000218b9ce R14: 0000000000000001 R15: ffff880480684a88
> FS:  00007f7c7f998b40(0000) GS:ffff880507800000(0000) knlGS:0000000000000000
> CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> CR2: 0000000000000000 CR3: 000000044f15f000 CR4: 00000000001406f0
> DR0: 00007f4ce439d000 DR1: 0000000000000000 DR2: 0000000000000000
> DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000600
> Stack:
>  ffff880501430000 d305ffffa00a2245 006c000000000002 0500000000000010
>  6c000000000002d3 0000000000001000 000000006427eebb ffff880480684a88
>  0000000000000000 ffff8804fddcf348 0000000000002000 0000000000000000
> Call Trace:
>  [<ffffffffa009cff0>] __btrfs_drop_extents+0xb00/0xe30 [btrfs]

We've been hunting this one for at least two years.  It's the white 
whale of btrfs bugs.  Josef has a semi-reliable reproducer now, but I 
think it's not the same as the pagevec based problems you reported earlier.

-chris