Re: semi-frequent 5.0_BETA crashes on SGI O2

[Andrew Doran <ad%netbsd.org@localhost>]

On Sun, Dec 21, 2008 at 04:36:02PM -0500, Rafal Boni wrote:

> panic: kernel diagnostic assertion "c->c_magic == CALLOUT_MAGIC" failed: file 
> "/
> extra/netbsd-src/netbsd-5/sys/kern/kern_timeout.c", line 424
> Stopped in pid 0.2 (system) at  netbsd:cpu_Debugger+0x4:        jr      ra
>                 bdslot: nop
> db> tr
> cpu_Debugger+4 (8dffe000,bf390000,0,0) ra 8032a0e4 sz 0
> panic+1e4 (8dffe000,805259f0,805558dc,805558a8) ra 804c6e18 sz 48
> __kernassert+48 (8dffe000,805259f0,805558dc,805558a8) ra 80311fa0 sz 32
> callout_stop+108 (8dffe000,805259f0,805558dc,805558a8) ra 8012bd98 sz 32

If you can get the address of the callout, do a 'whatis' on it. I would look
for a code path that is pushing a scsipi_xfer that has not been allocated
from scsipi_get_xs(). Some possible reasons for the crash:

- uninitialized scsipi_xfer being pushed
- scsipi_xfer being freed while in use
- scsipi_xfer being corrupted while in use (kernel memory corruption)
- tlb or cache coherency bug
- dying hardware

Andrew