Re: [10.0-BETA] Reproductible panic (filesystem)

[BERTRAND Joël <joel.bertrand%systella.fr@localhost>]

Taylor R Campbell a écrit :
>> Date: Thu, 18 May 2023 22:55:04 +0200
>> From: BERTRAND Joël <joel.bertrand%systella.fr@localhost>
>>
>> 	Both wd0 and wd1 have been tested and don't have smart alerts. When I
>> restart this server, I can see a message just before system reboots. If
>> I remember, this message indicates that ccd0 is removed from system
>> before dk5 and could explain why filesystem is corrupted after next boot.
> 
> Do you have a serial console that could capture the message?  Can you
> share that output?

	Sorry, I don't have.

> If the message you mentioned is:
> 
> 	ccd0: detached
> 
> then dk5 should have been detached already via dkwedge_delall, which
> is done before the print in ccdioctl(CCDIOCCLR).

	This morning, ccd0 (and of course dk5) are "locked". All tasks stall in
uninterruptible state when they try to access to dk5:

legendre# cd /var/squid/cache
legendre# du -hs
^C^C^C^C^Cs^C^C^C

> I recently fixed a number of races and other issues in dk(4) in HEAD
> (and have some changes upcoming to fix more).  Is it reproducible with
> a kernel from HEAD?

	I haven't tested with -current. Do you have a patch against -10.0 ?

	Regards,

	JB