nvme-5.12 is crashing with fabrics discover

[Chaitanya Kulkarni]

On 2/11/21 2:36 AM, Sagi Grimberg wrote:
> Hey,
>
> Started seeing nvme discovery crashing since commit:
> a7c7f7b2b641 ("nvme: use bio_set_dev to assign ->bi_bdev")
>
> Not yet sure what's going on, but reverting this commit
> makes the issue go away.
>
> Trace:
> --
> [   86.901818] nvme nvme0: new ctrl: NQN 
> "nqn.2014-08.org.nvmexpress.discovery", addr 192.168.123.1:8009
> [   86.902223] BUG: kernel NULL pointer dereference, address: 
> 0000000000000368
> [   86.903345] #PF: supervisor read access in kernel mode
> [   86.904949] #PF: error_code(0x0000) - not-present page
> [   86.906202] PGD 0 P4D 0
> [   86.907020] Oops: 0000 [#1] SMP PTI
> [   86.908025] CPU: 3 PID: 1470 Comm: nvme Tainted: G           OE 
> 5.11.0-rc5+ #35
> [   86.909333] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 
> 1.14.0-1 04/01/2014
> [   86.910924] RIP: 0010:bio_associate_blkg_from_css+0x45/0x140
> [   86.912854] Code: 38 f6 c2 03 0f 85 de 00 00 00 65 48 ff 0a e8 12 52 
> cb ff 4d 85 e4 74 4d 49 83 bc 24 c0 00 00 00 00 74 42 48 8b 45 08 4c 89 
> e7 <48> 8b 80 68 03 00 00 48 8b 70 50 e8 0b fe ff ff 48 89 c3 48 85 c0
> [   86.915683] RSP: 0018:ffffbe6c409d3da0 EFLAGS: 00010282
> [   86.917316] RAX: 0000000000000000 RBX: 0000000000000000 RCX: 
> 0000000000000010
> [   86.919128] RDX: ffff9e82f3b60000 RSI: ffff9e82f0fdea00 RDI: 
> ffff9e82f0fdea00
> [   86.919135] RBP: ffff9e82c2769000 R08: 0000000000000001 R09: 
> 00000000409d3d00
> [   86.919137] R10: ffff9e82c2769000 R11: ffff9e82c2769078 R12: 
> ffff9e82f0fdea00
> [   86.919139] R13: 0000000000000000 R14: ffff9e82c2769000 R15: 
> 0000000000000000
> [   86.919143] FS:  00007fbecaea0780(0000) GS:ffff9e833dd80000(0000) 
> knlGS:0000000000000000
> [   86.919146] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> [   86.919149] CR2: 0000000000000368 CR3: 00000000034f6004 CR4: 
> 0000000000370ee0
> [   86.919162] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 
> 0000000000000000
> [   86.919164] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 
> 0000000000000400
> [   86.919169] Call Trace:
> [   86.919243]  bio_associate_blkg+0x20/0x70
> [   86.923268]  nvme_submit_user_cmd+0xd9/0x260 [nvme_core]
> [   86.923391]  nvme_user_cmd+0x11f/0x190 [nvme_core]
> --
>
>
I think fix for this is present in the linux-block/for-next tree, can
you see if that fixes
your problem ?

commit 59c157433fbc6a7f63f3d708ca2966d0f56bcb7c
Author: Chaitanya Kulkarni <chaitanya.kulkarni at wdc.com>
Date:   Tue Jan 26 21:37:38 2021 -0800

    nvme-core: check bdev value for NULL