The following reply was made to PR kern/46217; it has been noted by GNATS.
From: Manuel Bouyer <bouyer%antioche.eu.org@localhost>
To: gnats-bugs%NetBSD.org@localhost
Cc: kern-bug-people%NetBSD.org@localhost, gnats-admin%NetBSD.org@localhost,
netbsd-bugs%NetBSD.org@localhost
Subject: Re: kern/46217: NetBSD 6.0_BETA hangs with CARP enabled
Date: Sat, 24 Mar 2012 15:08:56 +0100
On Sat, Mar 17, 2012 at 07:15:01PM +0000, root%forglen.anduin.org.uk@localhost
wrote:
> System: NetBSD forglen.anduin.org.uk 6.0_BETA NetBSD 6.0_BETA (GENERIC) #0:
> Wed Mar 14 12:37:03 GMT 2012
> root%forglen.anduin.org.uk@localhost:/usr/obj/sys/arch/i386/compile/GENERIC
> i386
> Architecture: i386
> Machine: i386
> >Description:
> Compile and install a kernel with CARP enabled. Configure a
> carp interface and reboot machine. Note that the machine locks up during
> startup. Cannot be pinged on any interface.
> >How-To-Repeat:
> Edit GENERIC config to uncomment CARP. Compile. Create a carp
> interface e.g:
>
> cat /etc/ifconfig.carp0
> create
> vhid 1 192.168.0.220 netmask 255.255.255.0
>
> Boot system:
>
> [...]
> Updating fontconfig cache:
>
> <<<<break sent to interrupt hang>>>>
>
> fatal breakpoint trap in supervisor mode
> trap type 1 code 0 eip c026e8a4 cs 8 eflags 282 cr2 bb80b000 ilevel 8
> Stopped in pid 0.2 (system) at netbsd:breakpoint+0x4: popl %ebp
> db{0}> bt
> breakpoint(c0c44f80,3f8,5,c0c45940,0,c0cd2480,d9876f6c,c056dbce,228,0) at
> netbsd
> :breakpoint+0x4
> comintr(c20e3280,d99ccc18,ffe7f7ef,bdfffffe,ffffffff,77ffffff,fcfefd7f,ff7fdffc,
> fefbfff7,ffffff7f) at netbsd:comintr+0x599
> --- switch to interrupt stack ---
> Xintr_ioapic_edge10() at netbsd:Xintr_ioapic_edge10+0xb5
> --- interrupt ---
> x86_stihlt(1,1,7,c20f6d20,c0c45940,c24fa7e0,c20f6d20,c054bf9d,c2405da8,0) at
> net
> bsd:x86_stihlt+0x5
> acpicpu_cstate_idle_enter(c2405da8,0,d99cccec,c055a64e,c0c45940,c20f4d40,c0c4594
> 0,d99cccec,c20f6d20,c0c45940) at netbsd:acpicpu_cstate_idle_enter+0x42
> acpicpu_cstate_idle(c20f6d20,0,c05466a2,c20f6d20,33ce33d4,cc23cc73,33cc33cc,cc33
> cc23,37cc33cc,cc33cc33) at netbsd:acpicpu_cstate_idle+0xa8
> idle_loop(c20f6d20,e89000,e92000,0,c0100307,0,0,0,0,0) at
> netbsd:idle_loop+0x17c
>
> db{0}> ps
> PID LID S CPU FLAGS STRUCT LWP * NAME WAIT
> 217 1 3 0 80 c24fa7e0 fc-cache nanoslp
Some debugging shows that softclock (and so scheduling) is not
working any more.
This is caused by ip_carp.c 1.42 -> 1.44; you actually can't sleep in
the softclock thread because it will never wake up. The attached patch
(reverting ip_carp.c 1.42 -> 1.44) fixes the problem for me.
Antti, what is exactly the ABI issue with DELAY() you mention in
the commit message ? kern/40505 doesn't give lots of details either.
--
Manuel Bouyer <bouyer%antioche.eu.org@localhost>
NetBSD: 26 ans d'experience feront toujours la difference
--