Re: Saving kernel crash dumps

On Tue, 30 Sep 2025 05:13:52 +1000
matthew green <[email protected]> wrote:

> by default, the 'b' partition of the boot device will be used, if
> it is large enough and of type 'swap'.  can you try that setup?
> 
> the crash might be happening too early.  in which case, maybe you
> can get kernel gdb over serial to work.  options KGDB -- it might
> not work right now, i haven't tried it for a long time and not
> really sure if i did it with sparc64 vs sparc...
> 
> 
> .mrg.

I created a swap partition greater than the size of system RAM,
performed sync after panic, but not sure if anything was written to
swap. It took about 5 seconds to sync and there is no way that
64 GiB of RAM image was written to USB stick in such a short amount
of time, unless it did not use all the RAM at this stage.

When I insert this USB stick into sun4u system, I don't know how to
force savecore to look at /dev/sd1b and extract the core file, as it
complains it cannot find the device.

# vi /etc/fstab
/dev/sd1b    none      swap    sw,dp            0 0

# swapctl -A
swapctl: setting dump device to /dev/sd1b

# mount /dev/sd1a /mnt
# savecore -n -v -N /mnt/netbsd
dumplo = 13191073792 (25763816 * 512)
savecore: can't find device 0/124


I have these partitions on USB stick:

# disklabel /dev/rsd1
3 partitions:
#        size    offset     fstype [fsize bsize cpg/sgs]
 a:  16777216         0     4.2BSD      0     0     0  # (Cyl.      0 -   8191)
 b: 233308160  16777216       swap                     # (Cyl.   8192 - 122111)
 c: 250085376         0     unused      0     0        # (Cyl.      0 - 122111)