Re: [OOPS, 3.13-rc2] null ptr in dio_complete()

[Jens Axboe <axboe@kernel.dk>]

On 12/05/2013 07:22 AM, Ming Lei wrote:
> On Wed, 4 Dec 2013 12:58:38 +1100
> Dave Chinner <david@fromorbit.com> wrote:
> 
>>
>> And I just hit this from running xfs_repair which is doing
>> multithreaded direct IO directly on /dev/vdc:
>>
>> [ 1776.508599] BUG: unable to handle kernel NULL pointer dereference at 0000000000000328
>> [ 1776.510446] IP: [<ffffffff81755b6a>] blk_account_io_done+0x6a/0x180
>> [ 1776.511762] PGD 38e75b067 PUD 41987d067 PMD 0
>> [ 1776.512577] Oops: 0000 [#1] SMP
>> [ 1776.512577] Modules linked in:
>> [ 1776.512577] CPU: 0 PID: 0 Comm: swapper/0 Not tainted 3.13.0-rc2-dgc+ #75
>> [ 1776.512577] Hardware name: Bochs Bochs, BIOS Bochs 01/01/2011
>> [ 1776.512577] task: ffffffff81f37480 ti: ffffffff81f26000 task.ti: ffffffff81f26000
>> [ 1776.512577] RIP: 0010:[<ffffffff81755b6a>]  [<ffffffff81755b6a>] blk_account_io_done+0x6a/0x180
>> [ 1776.512577] RSP: 0000:ffff88011bc03da8  EFLAGS: 00010046
>> [ 1776.512577] RAX: 0000000010000000 RBX: 0000000000000000 RCX: 0000000000000000
>> [ 1776.512577] RDX: 0000000000000000 RSI: 0000000000000002 RDI: ffff8800dba59fc0
>> [ 1776.512577] RBP: ffff88011bc03db8 R08: 0000000000000000 R09: ffff88021a828928
>> [ 1776.512577] R10: 0000000000000000 R11: 0000000000000001 R12: 0000000000000000
>> [ 1776.512577] R13: 0000000000000000 R14: 0000000000002000 R15: 0000000000000000
>> [ 1776.512577] FS:  0000000000000000(0000) GS:ffff88011bc00000(0000) knlGS:0000000000000000
>> [ 1776.512577] CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
>> [ 1776.512577] CR2: 0000000000000328 CR3: 000000041a4b0000 CR4: 00000000000006f0
>> [ 1776.512577] Stack:
>> [ 1776.512577]  0000000000000000 ffff8800dba59fc0 ffff88011bc03df8 ffffffff8175e4b8
>> [ 1776.512577]  00000000000001c0 ffff8800dba59fc0 ffffe8fcfbc03a80 ffff88041aac0800
>> [ 1776.512577]  ffff88011ad53b10 0000000000000096 ffff88011bc03e08 ffffffff8175e536
>> [ 1776.512577] Call Trace:
>> [ 1776.512577]  <IRQ>
>> [ 1776.512577]  [<ffffffff8175e4b8>] blk_mq_complete_request+0xb8/0x120
>> [ 1776.512577]  [<ffffffff8175e536>] __blk_mq_end_io+0x16/0x20
>> [ 1776.512577]  [<ffffffff8175e5a8>] blk_mq_end_io+0x68/0xd0
>> [ 1776.512577]  [<ffffffff81852e47>] virtblk_done+0x67/0x110
>> [ 1776.512577]  [<ffffffff817f7925>] vring_interrupt+0x35/0x60
>> [ 1776.512577]  [<ffffffff810e48a4>] handle_irq_event_percpu+0x54/0x1e0
>> [ 1776.512577]  [<ffffffff810e4a78>] handle_irq_event+0x48/0x70
>> [ 1776.512577]  [<ffffffff810e71d7>] handle_edge_irq+0x77/0x110
>> [ 1776.512577]  [<ffffffff8104796f>] handle_irq+0xbf/0x150
>> [ 1776.512577]  [<ffffffff81aec8c2>] ? __atomic_notifier_call_chain+0x12/0x20
>> [ 1776.512577]  [<ffffffff81aec8e6>] ? atomic_notifier_call_chain+0x16/0x20
>> [ 1776.512577]  [<ffffffff81af2b8a>] do_IRQ+0x5a/0xe0
>> [ 1776.512577]  [<ffffffff81ae892d>] common_interrupt+0x6d/0x6d
>> [ 1776.512577]  <EOI>
>> [ 1776.512577]  [<ffffffff8107e626>] ? native_safe_halt+0x6/0x10
>> [ 1776.512577]  [<ffffffff8104e50f>] default_idle+0x1f/0xc0
>> [ 1776.512577]  [<ffffffff8104edb6>] arch_cpu_idle+0x26/0x30
>> [ 1776.512577]  [<ffffffff810e3d81>] cpu_startup_entry+0x81/0x240
>> [ 1776.512577]  [<ffffffff81ad4f97>] rest_init+0x77/0x80
>> [ 1776.512577]  [<ffffffff82041ee7>] start_kernel+0x3c8/0x3d5
>> [ 1776.512577]  [<ffffffff8204191e>] ? repair_env_string+0x5e/0x5e
>> [ 1776.512577]  [<ffffffff82041117>] ? early_idt_handlers+0x117/0x120
>> [ 1776.512577]  [<ffffffff820415c2>] x86_64_start_reservations+0x2a/0x2c
>> [ 1776.512577]  [<ffffffff8204170a>] x86_64_start_kernel+0x146/0x155
>> [ 1776.512577] Code: 54 41 89 c4 53 41 83 e4 01 48 2b 97 d8 00 00 00 48 8b 9f d0 00 00 00 4d 63 e4 65 44 8b 04 25 1c b0 00 00 49 8d 74 24 02 49 63 c8 <48> 8b 83 28 03
>> [ 1776.512577] RIP  [<ffffffff81755b6a>] blk_account_io_done+0x6a/0x180
>> [ 1776.512577]  RSP <ffff88011bc03da8>
>> [ 1776.512577] CR2: 0000000000000328
>> [ 1776.512577] ---[ end trace 33a5787221ff13b4 ]---
>> [ 1776.512577] Kernel panic - not syncing: Fatal exception in interrupt
>> [ 1776.512577] Shutting down cpus with NMI
>>
>> So this is looking like another virtio+blk_mq problem....
> 
> This one might be caused by using-after-'free' request because the freed
> request may be allocated to another CPU immediately, and the below patch
> may be helpful, would you mind testing it?
> 
> diff --git a/block/blk-mq.c b/block/blk-mq.c
> index 6875736..2358bdf 100644
> --- a/block/blk-mq.c
> +++ b/block/blk-mq.c
> @@ -311,12 +311,12 @@ void blk_mq_complete_request(struct request *rq, int error)
>  
>  	blk_account_io_completion(rq, bytes);
>  
> +	blk_account_io_done(rq);
> +
>  	if (rq->end_io)
>  		rq->end_io(rq, error);
>  	else
>  		blk_mq_free_request(rq);
> -
> -	blk_account_io_done(rq);
>  }
>  
>  void __blk_mq_end_io(struct request *rq, int error)

That's definitely a bug... Good catch, Ming. Applied.

-- 
Jens Axboe