kern/52198: Kernel trap 12 when debugging fork() with ald [/usr/ports/devel/ald]

[Thomas Moestl]

The following reply was made to PR kern/52198; it has been noted by GNATS.

From: Thomas Moestl <t.moestl at tu-bs.de>
To: Lolownia <lolownia at polbox.com>
Cc: FreeBSD-gnats-submit at freebsd.org
Subject: Re: kern/52198: Kernel trap 12 when debugging fork() with ald [/usr/ports/devel/ald]
Date: Wed, 14 May 2003 03:26:31 +0200

 On Wed, 2003/05/14 at 03:19:24 +0200, Thomas Moestl wrote:
 > On Wed, 2003/05/14 at 00:36:36 +0200, Lolownia wrote:
 > > 
 > > >Number:         52198
 > > >Category:       kern
 > > >Synopsis:       Kernel trap 12 when debugging fork() with ald [/usr/ports/devel/ald]
 > > [...]
 > > 	When stepping through simple assembler program with fork() syscall,
 > > 	system drops into DDB
 > > 	db> trace
 > > 	sw1b(2f,2f,2f,bfbff64c,bfbff654) at sw1b+0xa9
 > > 	fork_trampoline() at fork_trampoline
 > > 	--- trap 0xa, eip = 0x804849c, esp = 0xbfbff628, ebp = 0xbfbff644 ---
 > 
 > This is a bug in cpu_fork(); it bogusly bases the kernel eflags to be
 > set for a new process on the user eflags of the forking process (since
 > a commit which seems to originally have been intended to clear PSL_I,
 > about a year ago).
 > The patch below should fix this; it will just use a sane initial
 > value, since cloning the forking process' kernel eflags (as was done
 > before the commit mentioned above)
 
 Oops, actually, that commit introduced eflags saving and restoring in
 cpu_switch() altogether, so scratch that remarks about the previous
 behaviour. Need sleep.
 The patch should be correct, however.
 
 	- Thomas
 
 -- 
 Thomas Moestl <t.moestl at tu-bs.de>	http://www.tu-bs.de/~y0015675/
               <tmm at FreeBSD.org>		http://people.FreeBSD.org/~tmm/
 PGP fingerprint: 1C97 A604 2BD0 E492 51D0  9C0F 1FE6 4F1D 419C 776C