Sym2 driver timeouts on powerstacks

Sym2 driver timeouts on powerstacks

[Marc Dietrich]

[-- Attachment #1: Type: text/plain, Size: 702 bytes --]


Hello list,

I'm trying to get a 2.6.10 kernel to run on a Motorola Powerstack II (Utah). 
This is a PReP (powerpc) machine.

This machine runs fine on 2.4 series kernels, but fails on 2.6 due to timeouts 
in the scsi driver. The 2.4 series kernels used the (I guess now obsoleted) 
ncr53c8xx.c/sym53c8xx.c driver.
There were some PReP specific "fixed" in the old driver, which not have been 
ported to the sym2 driver. I tried to backport the "latency" and the 
"pci_command_master/parity" fixes (see the 2.4 log), but even with this I had 
no success.

I attached the full debug output of both 2.4 and 2.6 kernels.

Please CC me on answers, because I'm not subscribed to the list.

Greetings

Marc


[-- Attachment #2: serial_2.4.24-pre1.log.bz2 --]
[-- Type: application/x-bzip2, Size: 6254 bytes --]

[-- Attachment #3: serial_2.6.10.log.bz2 --]
[-- Type: application/x-bzip2, Size: 3058 bytes --]

Re: Sym2 driver timeouts on powerstacks

[Matthew Wilcox]

On Tue, Jan 04, 2005 at 01:54:49PM +0100, Marc Dietrich wrote:
> This machine runs fine on 2.4 series kernels, but fails on 2.6 due to timeouts 
> in the scsi driver. The 2.4 series kernels used the (I guess now obsoleted) 
> ncr53c8xx.c/sym53c8xx.c driver.
> There were some PReP specific "fixed" in the old driver, which not have been 
> ported to the sym2 driver. I tried to backport the "latency" and the 
> "pci_command_master/parity" fixes (see the 2.4 log), but even with this I had 
> no success.

I'm not surprised, those patches ought to have no effect.
PCI_COMMAND_MASTER is set when the driver calls pci_set_master().
PCI_COMMAND_PARITY should be set by the platform code, but afaict is
currently not set on ppc.

> I attached the full debug output of both 2.4 and 2.6 kernels.

Umm.  Not terribly helpful.  Could you delete all the spurious blank lines
next time, and it'd be much more helpful if you could boot with the option
"sym53c8xx.debug=0x1fff"

-- 
"Next the statesmen will invent cheap lies, putting the blame upon 
the nation that is attacked, and every man will be glad of those
conscience-soothing falsities, and will diligently study them, and refuse
to examine any refutations of them; and thus he will by and by convince 
himself that the war is just, and will thank God for the better sleep 
he enjoys after this process of grotesque self-deception." -- Mark Twain

Re: Sym2 driver timeouts on powerstacks

[Matthew Wilcox]

On Tue, Jan 04, 2005 at 01:54:49PM +0100, Marc Dietrich wrote:
> I'm trying to get a 2.6.10 kernel to run on a Motorola Powerstack II (Utah). 
> This is a PReP (powerpc) machine.

I just had a thought:

sym0: SCSI BUS has been reset.
scsi0 : sym-2.1.18m
elevator: using anticipatory as default io scheduler
sym0:0:0:ccb @c7d9ac00 using tag 256.
sym0: queuepos=2.  
sym0:0:0: ABORT operation started.
sym0:0:0: ABORT operation timed-out.
sym0:0:0: DEVICE RESET operation started.
sym0:0:0: DEVICE RESET operation timed-out.
sym0:0:0: BUS RESET operation started.
sym0:0:0: BUS RESET operation timed-out.
sym0:0:0: HOST RESET operation started.
sym0:0:0:ccb @c7d9ac00 freeing tag 256.
sym0: SCSI BUS has been reset.

This is sym2's typical behaviour when it receives no interrupts.
I think if you were to NFS-root this box, you'd see equal problems with
the network driver.  Try asking on the powerpc lists, see if there's a
known interrupt routing problem.

-- 
"Next the statesmen will invent cheap lies, putting the blame upon 
the nation that is attacked, and every man will be glad of those
conscience-soothing falsities, and will diligently study them, and refuse
to examine any refutations of them; and thus he will by and by convince 
himself that the war is just, and will thank God for the better sleep 
he enjoys after this process of grotesque self-deception." -- Mark Twain

Re: Sym2 driver timeouts on powerstacks

[marvin24]

Hi Matt,

in the meantime I checked 2.6.8 kernel and this one works fine. There were 
some changes in the interrupt code in in version 2.6.9 I think, so I second 
your opinion that the real problem is not the scsi driver, but some irq 
problems. Other people (on linuxppc-dev@ozlab.org) also noted this.
Unfortunatly, I have no direct access to this machine anymore to do some 
debuging, but I'm sure this will come up again in the (far) future - so be 
warned...

Anyway, thanks for your help

Marc


Am Freitag, 7. Januar 2005 15:48 schrieben Sie:
> On Tue, Jan 04, 2005 at 01:54:49PM +0100, Marc Dietrich wrote:
> > I'm trying to get a 2.6.10 kernel to run on a Motorola Powerstack II
> > (Utah). This is a PReP (powerpc) machine.
>
> I just had a thought:
>
> sym0: SCSI BUS has been reset.
> scsi0 : sym-2.1.18m
> elevator: using anticipatory as default io scheduler
> sym0:0:0:ccb @c7d9ac00 using tag 256.
> sym0: queuepos=2.
> sym0:0:0: ABORT operation started.
> sym0:0:0: ABORT operation timed-out.
> sym0:0:0: DEVICE RESET operation started.
> sym0:0:0: DEVICE RESET operation timed-out.
> sym0:0:0: BUS RESET operation started.
> sym0:0:0: BUS RESET operation timed-out.
> sym0:0:0: HOST RESET operation started.
> sym0:0:0:ccb @c7d9ac00 freeing tag 256.
> sym0: SCSI BUS has been reset.
>
> This is sym2's typical behaviour when it receives no interrupts.
> I think if you were to NFS-root this box, you'd see equal problems with
> the network driver.  Try asking on the powerpc lists, see if there's a
> known interrupt routing problem.