port-vax/58068: "ifconfig qe0 mtu 512" causes kernel crash on NetBSD 10.0RC6

[als%thangorodrim.ch@localhost]

>Number:         58068
>Category:       port-vax
>Synopsis:       "ifconfig qe0 mtu 512" causes kernel crash on NetBSD 10.0RC6
>Confidential:   no
>Severity:       critical
>Priority:       high
>Responsible:    port-vax-maintainer
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Sat Mar 23 11:20:00 +0000 2024
>Originator:     Alexander Schreiber
>Release:        NetBSD 10.0_RC6 (GENERIC)
>Organization:
Not Much
>Environment:
>Description:
While investigating a SIMH VAX simulator crash (simulator segfaults with certain network traffic), I wanted to test a theory and set a silly low MTU on the ethernet interface. Then this happened:

 ifconfig qe0 mtu 512
[ 401.0241561] r0=87fd9028 r1=00000000 r2=00000000 r3=87c6a780 r4=80692f34 r5=00000000 r6=87fd9028 r7=87e91a00
[ 401.0241561] r8=00000000 r9=87e91a00 r10=8090697f r11=00000000
[ 401.0241561] ap=8b87fc7c fp=8b87fc68 sp=7ffff50c pc=80259431
[ 401.0241561] panic: SEGV in kernel mode: pc 0x80259431 addr 0
[ 401.0241561] cpu0: Begin traceback...
[ 401.0241561] panic: SEGV in kernel mode: pc 0x80259431 addr 0
[ 401.0241561] Stack traceback : 
[ 401.0241561]   Process is executing in user space.
[ 401.0241561] cpu0: End traceback...

[ 401.0241561] dump to dev 9,1 not possible

HALT instruction, PC: 80013565 (PUSHAB 802A2F6B)

I posted to port-vax asking for replication on real metal and got
confirmation that it happens there too:

http://mail-index.netbsd.org/port-vax/2024/03/23/msg004969.html

There is also a patch provided to port-vax by Matthew Green:
http://mail-index.netbsd.org/port-vax/2024/03/23/msg004973.html





>How-To-Repeat:
Boot NetBSD on a VAX (or SIMH VAX) with the qe ethernet device, try to set
MTU to 512 byte, observe crash.

>Fix:
patch provided by Matthew Green (haven't gotten around to testing it):
http://mail-index.netbsd.org/port-vax/2024/03/23/msg004973.html