Re: 2.6 kernel on pws433au
Uwe Schindler wrote:
> At 16:03 13.01.2006, Patrick Caulfield wrote:
>> Jaakko Linnosaari wrote:
>> > Patrick Caulfield wrote:
>> >>Has anyone got a recentish 2.6.x kernel to work on a pws433au box ?
>> > Yes. My PWS433au box runs fine with 2.6.15. The kernel package is now
>> > available at http://testoapina.tky.hut.fi/kernel/ if you want to take a
>> > look.
>> Thanks very much for that - much quicker than compiling a new kernel!
>> But that kernel won't boot either. I just get tons of I/O errors
>> scrolling on
>> the display.
>> Maybe the disk really is shot and only 2.6 can see the problem ? (2.4 is
>> fine). Maybe I'll try and dig out a different disk sometime.
> We do not know which type of scsi controller your machine has. If it is
> the mentioned QLogic ISP1020/1040 then the following could be the problem:
> * Linux 2.4 uses the "isp1020" module for this controller
> * Linux 2.6 had isp1020 until version 2.6.9 (in Debian until 2.6.8-1,
> later Norbert backported the modified qla1280 from 2.6.10 to debian
> stable 2.6.8-2). Since then the new driver "qla1280" is used. isp1020 is
> no longer in use because of a missing error handler and no maintainer
> that supported that module. In recent kernels I think it is completely
> So I exspect the following:
> * The old isp1020 driver has no or only minor error handling, so the
> errors do not appear on Linux 2.4
> * Linux 2.6 uses now qla1280 with correct error handling -> errors are
> detected and are displayed (I guess the errors with your disk are no
> fatal ones...)
> I have seen the follwing with my Alphastation because at first it missed
> a terminator at the outside end of the SCSI card. Linux 2.4 didn't see
> this. Linux 2.6<=2.6.8-1 showed a error message at boot about missing
> error handling in qlogicisp. When I then updated to the new kernel
> driver (about one year ago) the scsi driver showed sometimes some
> non-fatal errors. I terminated the bus and since then all is OK. Before
> I also had some serious file system corruption in EXT3 because of the
> missing error handling.
> Could it be that there is some SCSI termination problem that is not
> detected by the old qlogicisp driver in 2.4?
Yes, it is the Qlogic.
Hmm quite possible. I'll investigate. VMS seems to run fine which is odd but I
have been inside the box to move scsi bits around so it's possible that
something has been knocked out.
Thanks very much