> PIO request cues.......as they take longer to service........is is
> possible that we are slipping out of the rq-> order and loading the
> do_ide_request with repeated calls more often than the acknowledged and
> cleared?
What about filling everything with printk-messages, stopping klogd
and adding the "dump kernel ring buffer to floppy disk"-patch (it switches
to real mode, so it schould be able to store the ring buffer)?

We should get an accurate descriptions what which CPU was doing just
before the crash,

        Manfred

Reply via email to