Re: [Bugme-new] [Bug 5378] New: aic7xxx deadlock/freeze on Adaptec AIC-7899P

Re: [Bugme-new] [Bug 5378] New: aic7xxx deadlock/freeze on Adaptec AIC-7899P

[Andrew Morton]

bugme-daemon@kernel-bugs.osdl.org wrote:
>
> http://bugzilla.kernel.org/show_bug.cgi?id=5378
> 
>            Summary: aic7xxx deadlock/freeze on Adaptec AIC-7899P
>     Kernel Version: 2.6.13.3
>             Status: NEW
>           Severity: high
>              Owner: andmike@us.ibm.com
>          Submitter: szpajder@staszic.waw.pl
> 
> 
> The following messages appeared in dmesg:
> 
> scsi0:0:1:0: Attempting to queue an ABORT message
> CDB: 0x28 0x0 0x0 0xab 0x8b 0x99 0x0 0x0 0x8 0x0
> scsi0: At time of recovery, card was not paused
> >>>>>>>>>>>>>>>>>> Dump Card State Begins <<<<<<<<<<<<<<<<<
> scsi0: Dumping Card State while idle, at SEQADDR 0x9
> Card was paused
> [...] (entire dump attached)
> 
> At the time of these errors, load average exceeded 30. After issuing SCSI RESET,
> the system went back to normal. The problem reappeared several hours later -
> load average reached 140 and all the tasks hung waiting for I/O. I was waiting
> for SCSI RESET, which did not occur this time - after about 3 minutes I had to
> reboot with sysrq.
> 
> The problem ocurred about a day after upgrading from 2.6.12.4 (which was running
> fine for over 50 days) to 2.6.13.3. Hardware: Intel SDS2 mainboard with Adaptec
> AIC-7899P SCSI onboard, 4 x Seagate ST336753LW, software RAID-5, configs, lspci,
> etc - attached. The main difference between startup dmesgs is that all hard
> drives were set up as asynchronous during bootup - this didn't occur under
> 2.6.12.4. So I went back to 2.6.12.4 for now, it seems to be ok.

ISTR that there have been several reports of this regression.

What could have caused this?

Re: [Bugme-new] [Bug 5378] New: aic7xxx deadlock/freeze on Adaptec AIC-7899P

[James Bottomley]

On Mon, 2005-10-10 at 18:41 -0700, Andrew Morton wrote:
> ISTR that there have been several reports of this regression.
> 
> What could have caused this?

Well ... the prior bug reports with this are in aic79xx, and there there
were no significant code changes between the working and the non working
versions.  The aic7xxx driver has been fairly significantly changed but
all in the area of setup and initialisation.

This one looks like a sequencer error, possibly induced by a flakey bus.
Apparently it actually managed to recover once, which is surprising for
the aic driver, but then it died a second time around.  There actually
was a sequencer change (backport from aic latest driver) that could be
responsible.  However, it's in both 2.6.12 and 2.6.13.

James