netdev
[Top] [All Lists]

Re: [Fwd: Re: possible bug x86 2.4.2 SMP in IP receive stack]

To: Andrew Morton <andrewm@xxxxxxxxxx>
Subject: Re: [Fwd: Re: possible bug x86 2.4.2 SMP in IP receive stack]
From: Pekka Pietikainen <pp@xxxxxxxxxxxxxx>
Date: Wed, 7 Mar 2001 15:23:48 +0200
Cc: Pekka Pietikainen <pp@xxxxxxxxxxxxxx>, Bob Felderman <feldy@xxxxxxxx>, kuznet@xxxxxxxxxxxxx, netdev@xxxxxxxxxxx
In-reply-to: <3AA60A35.F7C70C1@uow.edu.au>
References: <200103061854.KAA03475@myri.com>, <200103061854.KAA03475@myri.com> <20010306211530.A14203@netppl.fi> <3AA60A35.F7C70C1@uow.edu.au>
Sender: owner-netdev@xxxxxxxxxxx
On Wed, Mar 07, 2001 at 09:15:17PM +1100, Andrew Morton wrote:
> So you're saying that you can reproduce this crash with acenic?
> 
> And that it's specific to jumbo frames, SMP and the /proc settings
> which Bob is using?
> 
> Bob is using a non-zerocopy kernel.  What were you using?
Here's the kdb report from 2.4.2-ac13 (or rather just stacktrace
since I don't have a serial console setup :) ).
Jumbo frames seem to be necessary to reproduce the problem (or at least 
they make it a lot more probable). I haven't tried a non-SMP kernel
yet.

Unable to handle kernel paging request at 0x5a5a5a72

__free_pages+0x3
skb_release_data+0x4e
kfree_skbmem+0xe
__kfree_skb+0xe3
ip_frag_destroy+0x73
ip_degrag+0x197
ip_local_deliver+0x20
ret_from_intr

Looks like refcnt might not be increased properly somewhere, I'll
try to hunt down the problem a bit more...
-- 
Pekka Pietikainen

<Prev in Thread] Current Thread [Next in Thread>