Re: [PATCH] block: aoe: Fix kernel crash due to atomic sleep when exiting

From: Jens Axboe
Date: Thu Aug 08 2019 - 09:29:21 EST


On 8/7/19 8:09 PM, zhe.he@xxxxxxxxxxxxx wrote:
> From: He Zhe <zhe.he@xxxxxxxxxxxxx>
>
> Since commit 3582dd291788 ("aoe: convert aoeblk to blk-mq"), aoedev_downdev
> has had the possibility of sleeping and causing the following crash.
>
> BUG: scheduling while atomic: rmmod/2242/0x00000003
> Modules linked in: aoe
> Preemption disabled at:
> [<ffffffffc01d95e5>] flush+0x95/0x4a0 [aoe]
> CPU: 7 PID: 2242 Comm: rmmod Tainted: G I 5.2.3 #1
> Hardware name: Intel Corporation S5520HC/S5520HC, BIOS S5500.86B.01.10.0025.030220091519 03/02/2009
> Call Trace:
> dump_stack+0x4f/0x6a
> ? flush+0x95/0x4a0 [aoe]
> __schedule_bug.cold+0x44/0x54
> __schedule+0x44f/0x680
> schedule+0x44/0xd0
> blk_mq_freeze_queue_wait+0x46/0xb0
> ? wait_woken+0x80/0x80
> blk_mq_freeze_queue+0x1b/0x20
> aoedev_downdev+0x111/0x160 [aoe]
> flush+0xff/0x4a0 [aoe]
> aoedev_exit+0x23/0x30 [aoe]
> aoe_exit+0x35/0x948 [aoe]
> __se_sys_delete_module+0x183/0x210
> __x64_sys_delete_module+0x16/0x20
> do_syscall_64+0x4d/0x130
> entry_SYSCALL_64_after_hwframe+0x44/0xa9
> RIP: 0033:0x7f24e0043b07
> Code: 73 01 c3 48 8b 0d 89 73 0b 00 f7 d8 64 89 01 48 83 c8 ff c3 66 2e 0f
> 1f 84 00 00 00 00 00 0f 1f 44 00 00 b8 b0 00 00 00 0f 05 <48> 3d 01 f0 ff
> ff 73 01 c3 48 8b 0d 59 73 0b 00 f7 d8 64 89 01 48
> RSP: 002b:00007ffe18f7f1e8 EFLAGS: 00000206 ORIG_RAX: 00000000000000b0
> RAX: ffffffffffffffda RBX: 0000000000000000 RCX: 00007f24e0043b07
> RDX: 000000000000000a RSI: 0000000000000800 RDI: 0000555c3ecf87c8
> RBP: 00007ffe18f7f1f0 R08: 0000000000000000 R09: 0000000000000000
> R10: 00007f24e00b4ac0 R11: 0000000000000206 R12: 00007ffe18f7f238
> R13: 00007ffe18f7f410 R14: 00007ffe18f80e73 R15: 0000555c3ecf8760
>
> This patch, handling in the same way of pass two, unlocks the locks and
> restart pass one after aoedev_downdev is done.

Applied, thanks.

--
Jens Axboe