Re: kern/52598 (/proc/self/exe misbehaves if it's on a null mount)

[Frédéric Fauberteau <triaxx%NetBSD.org@localhost>]

The following reply was made to PR kern/52598; it has been noted by GNATS.

From: =?UTF-8?Q?Fr=C3=A9d=C3=A9ric_Fauberteau?= <triaxx%NetBSD.org@localhost>
To: gnats-bugs%netbsd.org@localhost
Cc: kern-bug-people%netbsd.org@localhost, netbsd-bugs%netbsd.org@localhost, coypu%sdf.org@localhost
Subject: Re: kern/52598 (/proc/self/exe misbehaves if it's on a null mount)
Date: Wed, 15 Jan 2020 10:16:34 +0100

 Le 2020-01-15 09:10, Martin Husemann a écrit :
 >  I have not followed the thread closely from the beginning, is this 
 > about
 >  ld.elf_so's ORIGIN handling? I thought we don't rely on that in pkgsrc
 >  rust anyway (past bootstrap).
 
 I though I suffered from pkg/53348 or pkg/53671 on NetBSD 8.0. I though 
 that backport the fix of kern/52598 should fix them. I need this host 
 since it build sets of custom packages for several domU running 8.0. 
 Otherwise, I would have updated it to 9.0_RC1.
 
 Unfortunately, the system just crashes during a bulk build (when 
 building lang/spidermonkey52) with the following backtrace:
 # crash -M netbsd.10.core
 Crash version 8.0, image version sys/arch/amd64/c.
 WARNING: versions differ, you may not be able to examine this image.
 System panicked: trap
 Backtrace from time of crash is available.
 crash> bt
 _KERNEL_OPT_NARCNET() at 0
 ?() at ffff800065288000
 log() at log+0x9c
 vlog() at vlog+0x51
 trap() at trap+0xa00
 --- trap (number 6) ---
 memmove() at memmove+0x10
 VOP_LOCK() at VOP_LOCK+0x13
 do_sys_symlinkat.isra.4() at do_sys_symlinkat.isra.4+0x1a5
 syscall() at syscall+0x1ec
 --- syscall (number 58) ---
 792a4923e20a:
 
 I admit I don't understand the WARNING message.