Hi James,
Thansk for the info. I will try some experiments and see if I can work
out where the scsi commands/responses are getting lost.
I do not beleave the issue with the ESP is a hardware problem. The PROM
and SunOS-4.1.1 have no problem finding/accessing the disk on the ESP and
I did not observe any issues prior to testing Linux v2.6.22. The V1 ESP
SCSI driver never failed to find the ESP disk (on kernels that got that
far) but I have not done as much testing with the V2 ESP driver as memory
problems post v2.6.21 needed to be fixed to get things up and running.
Regards
Mark Fortescue.
On Fri, 20 Jul 2007, James Bottomley wrote:
On Fri, 2007-07-20 at 14:26 +0100, Mark Fortescue wrote:
I have been having intermittant issues with the QLGC,isp driver
(qlogicpti.c) not finding any attached devices on my sun4c. Having now
moved on to Linux-2.6.22 this problem is now also affecting the ESP SCSI
driver and is nolonger intermittant.
My test setup uses an NFS root so it is currently independent of the SCSI
drivers/subsystem. I am hoping to change this in the near future so that
my sun4c boots into linux by default.
What kernel options can I use to try to work out why the drivers are
ignoring disks (there are no other types of attached device at present).
Are any special options needed to force the scsi subsystem to spin the
disk up and wait untill they become ready (The disk [target 0:0:3] on the
ESP auto-spins down if it is idle for too long and now that booting takes
longer ...) ?
nothing in the driver ... that's a firmware issue. What all SPI drivers
do is address the device and wait 250ms for a response (as per the
standard). If nothing comes back, they assume there's nothing on the
bus. You might be able to increase the no connect timeout, but it looks
to be a firmware property of both qlogicpti and esp (and if you do this,
scanning will take longer). 250ms is a huge amount of time in SPI, so
I'd suspect there was some type of bus or device problem.
Is there a way to get a NON-moduler scsi driver to re-scan its SCSI bus
from shell prompt (writing to a sysfs/procfs file)?
/sys/class/scsi_host/host<n>/scan
From my point of view, it looks like the command timeout is very much too
short. Is there a simple way of testing this theory on the ESP driver?
See above, but really 250ms should be more than adequate.
James
-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
