port-hpcmips/57296: sysinst segmentation fault

[hackerb9%gmail.com@localhost]

>Number:         57296
>Category:       port-hpcmips
>Synopsis:       sysinst segmentation fault
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    port-hpcmips-maintainer
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Sat Mar 25 08:10:00 +0000 2023
>Originator:     b9
>Release:        9.3
>Organization:
>Environment:
Unsure as NetBSD isn't installed yet on this NEC Mobile Pro 800.
>Description:
Booted from Windows CE to NetBSD-9.3 using hpcboot.exe. Sysinst seemed to work and gave me a menu. However, network configuration failed, so I went to "e: Utility menu" and then "a: Run /bin/sh", and got some strange error messages:

uid 0, pid 6, command sysinst, on /: file system full
/: write failed, file system is full 
pid 6 (sysinst): user write of 4136@0x7dd00000 at 161428 failed: 28
[1] Segmentation fault     sysinst


After that, it did give me a root prompt ('#'), so I tried `df -a` to see what was going on and it gave me another error:

sh: Cannot set process group (Operation not permitted) at 225

df reported that / was mounted on a 4MB memory disk (md0a), so it is not too surprising it filled up. I understand that these hpcmips machines often have a very small amount of RAM, but if the memory disk is so small that it is causing problems, perhaps the size should be increased slightly, perhaps to 6 or 8MB.
>How-To-Repeat:
I tried a second time and discovered that there is a tiny amount of space available (~200 KB) on the memory disk. The problem is likely was possibly that I had started up sysinst and walked away for several hours. It worked initially, but after trying to use sysinst to setup the network and setting it up manually, I got the same issue:

uid 0, pid 170, command sysinst, on /: file system full

/:write failed, file system is full
pid 170 (sysinst): user write of 98304@@x77a000


Another issue is that many programs want to dump core by default which is not a good idea when there is so little free space. 

For example, the `ping` command always dumps core when I exit it by hitting ^C (Ctrl + C). It says:

Floating point exception (core dumped)   ping 8.8.8.8

This implies to me that likely this version of ping was not compiled with the right flags. This VR4121 machine does not have floating point arithmetic on the CPU; when it hits an FPU instruction, it is supposed to cause a fault which allows the floating point operation to be processed in the kernel. 

Similarly, trying to use 'c: Configure network' sometimes works without a hitch, but occasionally, I'll see things like:

[1] Unknown signal number (core dumped) cat /etc/resolve.conf 2>/dev/null | NetBSD/hpcmip.9.3
[1] Unknown signal number (core dumped) cat /etc/resolv.conf 2>/dev/null | while read keyword domai...
>Fix:
Not yet known. Is there a way to limit the coredump size based on how much freespace is available? 

Also, I notice that when I first had the issue, I had walked away for hours with the Sysinst Menu on the screen before interacting with it. Is it possible there's a log leak somewhere?