The panic should be fixed in sys/dev/hpc/hpcfb.c rev 1.62: http://mail-index.netbsd.org/source-changes/2021/06/05/msg130014.html However, userland is quite buggy unfortunately :(. sh reports random signals when children exit, e.g., ---- # cc cc: fatal error: no input files compilation terminated. [1] Window size changed (core dumped) cc ---- (core was not dumped actually...) If this is due to cache management for VR41xx as pointed out by Nick, HPCMIPS_L1CACHE_DISABLE option may be useful for debug. However, this option has been left broken ~ 20 years... Thanks, rin On 2021/06/01 0:11, David Brownlee wrote:
Could you try booting http://sync.absd.org/hpcmips/netbsd-2021-05-31-extra-output.gz It should still panic with the double console attachment, but hopefully may give more output (some DBGX lines) as it does David On Sun, 30 May 2021 at 13:31, Alexander Bochmann <ab%lists.gxis.de@localhost> wrote:...on Fri, May 28, 2021 at 11:38:44PM +0200, Alexander Bochmann wrote: > ...on Thu, May 27, 2021 at 10:04:37PM +0200, Alexander Bochmann wrote: > > I think I'll quickly try to build a kernel with NBIVIDEO, just to see > > what happens. > Heh no, that's not it. After some trial and error with kernel build options, it turns out that the console doesn't seem to get initialized at all when bivideo is excluded (and having pseudo-device biconsdev is required too). bicons doesn't seem to be used elsewhere besides hpc machines, so if it's involved with the problem, it wouldn't show up anywhere else - but on the other hand I can't see it calling *_cnattach, at least not in an obvious way, so I'm most probably looking in the wrong direction with this. I also built a HEAD kernel that drops anything that's not used by the Z50 hardware, and the panic still occurs. Alex.