SCSI timeouts with sym-2 driver

SCSI timeouts with sym-2 driver

[Jones, David G]

I'm trying to debug a SCSI problem on an embedded system with a MPC8540
CPU and LSI53C1010-66 controller. The system is running a 2.4.25 kernel
and sym-2.1.19-pre3 driver. A Seagate 300 GB drive is attached to one
channel of the controller and an Adtron SDDS-01A1 PC Card reader is
attached to the other channel.

Everything works fine until I try to use the hard drive and PC card at
the same time. Then I get SCSI timeouts for the PC card reader and
eventually I/O errors for the PCCard devices. Continously copying a file
from the PCCard reader to the hard drive results in these messages in
the kernel log :

sym0:0:0: queue depth is now 135
sym0:0:0: queue depth is now 136
sym0:0:0: queue depth is now 137
Command timed out active=4 busy=4 failed=1
Command timed out chan=0 target=0 lun=1
Command timed out active=4 busy=4 failed=2
Command timed out chan=0 target=0 lun=0
Command timed out active=4 busy=4 failed=3
Command timed out chan=0 target=0 lun=1
Command timed out active=4 busy=4 failed=4
Command timed out chan=0 target=0 lun=0
sym1:0:0: ABORT operation started.
sym1:0:0: ABORT operation timed-out.
sym1:0:0: ABORT operation started.
sym1:0:0: ABORT operation timed-out.
sym1:0:1: ABORT operation started.
sym1:0:1: ABORT operation timed-out.
sym1:0:1: ABORT operation started.
sym1:0:1: ABORT operation timed-out.
sym1:0:0: DEVICE RESET operation started.
sym1:0:0: DEVICE RESET operation timed-out.
sym1:0:1: DEVICE RESET operation started.
sym1:0:1: DEVICE RESET operation timed-out.
sym1:0:0: BUS RESET operation started.
sym1: SCSI BUS reset detected.
sym1: enabling clock multiplier
sym1: Downloading SCSI SCRIPTS.
sym1: SCSI BUS has been reset.
sym1: command processing suspended for 3 seconds
sym1:0:0: BUS RESET operation complete.
sym1: command processing resumed
sym1:0:sync msgout: 1-3-1-29-f.
sym1:0:sync msgin: 1-3-1-29-f.
sym1:0:sdtr: ofs=15 per=41 div=5 fak=1 chg=0.
sym1:0:sync msgout: 1-3-1-29-f.
sym1:0:sync msgin: 1-3-1-29-f.
sym1:0:sdtr: ofs=15 per=41 div=5 fak=1 chg=0.
sym1:0:sync msgout: 1-3-1-29-f.
sym1:0:sync msgin: 1-3-1-29-f.
sym1:0:sdtr: ofs=15 per=41 div=5 fak=1 chg=0.
 UNIT_ATTENTION ?
This architecture does not implement dump_stack()
 I/O error: dev 08:21, sector 1099392
SCSI disk has been changed. Prohibiting further I/O.
from scsi_request_fn
 I/O error: dev 08:21, sector 1099896


I/O errors never occur for the hard drive, just the PC Card reader
device. Also copying files to /dev/null on the PCCard reader and hard
drive simultaneously results in about 2-3 MB/s on the hard drive, verus
50-60 MB/s with just the hard drive alone.

Has anyone seen a similar problem ? What's the best stratgey to debug
this kind of problem ? Thanks !

Re: SCSI timeouts with sym-2 driver

[Matthew Wilcox]

On Fri, Sep 15, 2006 at 01:36:37PM -0400, Jones, David G wrote:
> Everything works fine until I try to use the hard drive and PC card at
> the same time. Then I get SCSI timeouts for the PC card reader and
> eventually I/O errors for the PCCard devices. Continously copying a file
> from the PCCard reader to the hard drive results in these messages in
> the kernel log :

Sounds like you've got both these devices on the same interrupt; is that
true?  If so, I'd urge you to examine both of the interrupt handlers; it
feels like one of the drivers is not playing well with the other when an
interrupt occurs on both at the same time.  Or maybe it's a problem in
the arch code where it doesn't call both handlers?

In any case, trying a recent 2.6 kernel may give better results.

RE: SCSI timeouts with sym-2 driver

[Jones, David G]

Matt,

Thanks for the response. Each device is on a separate interrupt. But it
does seem like an interrupt related problem, doesn't it ? 

-----Original Message-----
From: Matthew Wilcox [mailto:matthew@wil.cx] 
Sent: Friday, September 15, 2006 1:41 PM
To: Jones, David G
Cc: linux-scsi@vger.kernel.org
Subject: Re: SCSI timeouts with sym-2 driver

On Fri, Sep 15, 2006 at 01:36:37PM -0400, Jones, David G wrote:
> Everything works fine until I try to use the hard drive and PC card at

> the same time. Then I get SCSI timeouts for the PC card reader and 
> eventually I/O errors for the PCCard devices. Continously copying a 
> file from the PCCard reader to the hard drive results in these 
> messages in the kernel log :

Sounds like you've got both these devices on the same interrupt; is that
true?  If so, I'd urge you to examine both of the interrupt handlers; it
feels like one of the drivers is not playing well with the other when an
interrupt occurs on both at the same time.  Or maybe it's a problem in
the arch code where it doesn't call both handlers?

In any case, trying a recent 2.6 kernel may give better results.