Re: "BUG: kernel NULL pointer dereference," when unmounting filesystem hitted by enospc error

[Johannes Thumshirn <jthumshirn@suse.de>]

On 24/10/2019 22:23, Peter Hjalmarsson wrote:
> Hi,
> 
> Sorry for late answer. Have been out of town for work.
> Interesting that you could not reproduce. What kind of system do you run?

I ran it up to 64 times but in a very minimal VM. Let me try to reprouce
the issue on a bare metal system.

> I have tried on a couple of system now, and they behave slightly
> different, so maybe it has something to do with which arch it runs
> what kind of crash it triggers if any?
> Also make sure that the "dd"-line in the script is not wrapped like it
> became in the mail.

Yes I've fixed that up.

> For me running trough the test in a couple of systems, I found the following:
> 
> The systems that always crashes did that during the first or the
> second run of the script (so probably no need to run for longer than
> maybe three to four times to verify).
>
> Two systems passes 5 times (32 bit arm)
> One system crashes during umount with a slightly different traceback
> but essential the same as previously reported Pine64 Rock64 rk3328:
> [503397.433500] Call trace:
> [503397.436338]  __free_pages+0x1c/0x80
> [503397.440506]  __free_raid_bio+0x84/0xf8 [btrfs]
> [503397.445713]  __remove_rbio_from_cache+0x134/0x1b8 [btrfs]
> [503397.451957]  btrfs_clear_rbio_cache.isra.0+0x5c/0x98 [btrfs]
> [503397.458473]  btrfs_free_stripe_hash_table+0x24/0x40 [btrfs]
> [503397.464891]  close_ctree+0x1b0/0x2c8 [btrfs]
> [503397.469851]  btrfs_put_super+0x20/0x30 [btrfs]
> 
> Two system crashes (one x86_64 Atom, one RPI3 arch64) during balance
> with a dmesg like:
> [10282.926420] Call Trace:
> [10282.926488]  lock_stripe_add+0x292/0x370 [btrfs]
> [10282.926560]  __raid56_parity_write+0x20/0x40 [btrfs]
> [10282.926633]  run_plug+0x131/0x150 [btrfs]
> [10282.926671]  blk_flush_plug_list+0xc2/0x110
> [10282.926708]  blk_finish_plug+0x21/0x2e
> [10282.926769]  btrfs_write_and_wait_transaction.isra.0+0x57/0xa0 [btrfs]
> [10282.926851]  btrfs_commit_transaction+0x72e/0x9a0 [btrfs]
> 
> Three system crashes (three x86_64, the one "tainted" has nvidia
> binary blobs) during umount with a dmesg like:
> [  658.646613] Call Trace:
> [  658.646675]  __free_raid_bio+0x72/0xb0 [btrfs]
> [  658.646728]  btrfs_free_stripe_hash_table+0x3d/0x70 [btrfs]
> [  658.646766]  close_ctree+0x1ea/0x2f0 [btrfs]
> [  658.646773]  generic_shutdown_super+0x6c/0x100
> [  658.646778]  kill_anon_super+0x14/0x30
> [  658.646808]  btrfs_kill_super+0x12/0xa0 [btrfs]
> 
> All full dmesg saved if you want to look at any of the other not posted below.

this could be handy, but not sure yet.


Byte,
	Johannes
-- 
Johannes Thumshirn                            SUSE Labs Filesystems
jthumshirn@suse.de                                +49 911 74053 689
SUSE Software Solutions Germany GmbH
Maxfeldstr. 5
90409 Nürnberg
Germany
(HRB 36809, AG Nürnberg)
Geschäftsführer: Felix Imendörffer
Key fingerprint = EC38 9CAB C2C4 F25D 8600 D0D0 0393 969D 2D76 0850