Re: port-amd64/39283: Kernel crash on Dell Poweredge 2950

[Fredrik Carlsson <fredrik%netbsd.se@localhost>]

 <fredrik <at> netbsd.se> writes:

> 
> >>> Hi,
> >>>
> >>> The next time it crashes like this, please do "show reg". Take the
> >>> value
> >>> of
> >>> 'rsp' and do something like this:
> >>>
> >>> x/Lx 0xffffffff12345678,40
> >>>
> >>> ... adding the ,40 at the end. That will dump some of the stack. Go
> >>> through each value dumped, one at a time, until you find something
> >>> useful
> >>> giving each value to this command:
> >>>
> >>> x/I 0xffffffff12345678
> >>>
> >>> You will get a function name at some point - probably on the first
> >>> value.
> >>> It
> >>> will show us where the call to NULL is being made from.
> >>>
> >>> Thanks,
> >>> Andrew
> >>>
> >>
> >> Hi,
> >>
> >> The machine hanged sometime tonight again, here is the stack dump, hope
> >> it
> >> helps:
> >>
> 
> Hi again,
> 
> After trying the latest 5.0 and removing all debug stuff in the kernel the
> machines actually survives about a week with an MP-kernel, it seems to die
> when we start all the thread intensive stuff.
> 
> Regards
> Fredrik
> 
> 

A quick update... 

The server is now running 5.0_RC2 and after exactly 8 days the filesystem seems
to crash, usually the whole server used to panic so I suppose this is an
improvement ;-) It's of course unusable because the only way to fix it is to
power cycle the server

This was in the log:
Feb 25 03:26:45 frost /netbsd: d0(mfi0:0:0:0): unable to allocate scsipi_xfer
Feb 25 03:26:45 frost /netbsd: sd0(mfi0:0:0:0): unable to allocate scsipi_xfer

Regards
Fredrik