On Friday 30 May 2003 17:29, Daniel Jarboe wrote:
> May 30 11:13:24 linprox kernel: end_request: I/O error, dev 5e:01
> (dasd), sector
> 3429024
>
>
> The device corresponds to /dev/dasda1 which is dasd=8666(ro) in zipl's
> parmfile and ro in fstab. It seems to happen with squid activity. I
> have to wonder if it's more than something trying to write to the R/O
> portions of the filesystem because if I do a touch then I get an error
> in my shell "touch: creating `blah': Read-only file system" but none of
> these I/O error messages. This is with basevol+guestvol disgussed in
> Large Scale Linux sg246824. It's interesting that they are different
> sectors, but fairly close together.
This is a known problem in the linux memory management code that only
shows up on s390. You can probably find a detailed explanation in
the mailing list archives. The good news is that the problem is completely
harmless, except for filling up the system logger.
Newer 2.4 kernels (e.g. the one from SLES 8) have a workaround that make
it less likely to show up. A correct fix is more intrusive and is currently
only in the latest 2.5.70-mm* kernels (don't run those on a production
system).
Arnd <><
