[PVE-User] Kernel panic with 2.6.32-6-pve

[Dietmar Maurer]

It would be great if someone finds a reproducible test case?

Also, please report a bug at bugzilla.openvz.org.

> -----Original Message-----
> From: Shane Hathaway [mailto:shane at hathawaymix.org]
> Sent: Montag, 17. Oktober 2011 20:57
> To: Dietmar Maurer
> Cc: Koichi MATSUMOTO; pve-user at pve.proxmox.com
> Subject: Re: [PVE-User] Kernel panic with 2.6.32-6-pve
> 
> On 10/16/2011 06:45 AM, Dietmar Maurer wrote:
> >> Unfortunately this kernel also panicked 12 hours after booting.
> >> Should I try 2.0...?
> >
> > No, that is the same kernel.
> 
> FWIW, it turns out kernel 2.6.32-6-pve still isn't working for me either.  I
> determined that the panic happens while iptables is trying to reject a packet.  It
> only happens sometimes, but when it does, the panic is immediate.  When it
> sends the corresponding ICMP message, apparently some of the ICMP packet is
> corrupted even before it leaves the kernel and the panic occurs promptly
> thereafter.  The trace (including the iptables log entry) looks like this:
> 
> DROP icmp IN= OUT=vmbr1 SRC=10.9.0.6 DST=159.96.198.227 LEN=197
> TOS=0x00
> PREC=0xC0 TTL=64 ID=54238 PROTO=ICMP TYPE=192 CODE=228 Kernel panic -
> not syncing: stack-protector: Kernel stack is corrupted
> in: ffffffff8149c84c
> 
> Pid: 0, comm: swapper Not tainted 2.6.32-6-pve #1 Call Trace:
>   <IRQ>  [<ffffffff814f1e66>] ? panic+0xa5/0x189
>   [<ffffffff8149c84c>] ? icmp_send+0x37c/0x790
>   [<ffffffff8147154f>] ? __ip_make_skb+0x2df/0x3c0
>   [<ffffffff81471fd6>] ? ip_local_out+0x16/0x40
>   [<ffffffff8147201b>] ? ip_send_skb+0x1b/0x90
>   [<ffffffff814720bb>] ? ip_push_pending_frames+0x2b/0x40
>   [<ffffffff8149c402>] ? icmp_push_reply+0xf2/0x120
>   [<ffffffff81069adb>] ? __stack_chk_fail+0x1b/0x30
>   [<ffffffff8149c84c>] ? icmp_send+0x37c/0x790
>   [<ffffffff8106af46>] ? vprintk+0x36/0x50
>   [<ffffffffa02003de>] ? ipt_do_table+0x3de/0x654 [ip_tables]
> 
> Notice ICMP packet: type=192, code=228.  I assume that's corruption.
> It's different every time: I've also seen type=110, code=46.
> 
> I've switched to 2.6.18, which is working fine, but I am willing to test more
> versions of 2.6.32 if you would like me to.  It's the least I can do for you guys.
> 
> Shane