netdev
[Top] [All Lists]

Fw: Oops in dev_get_by_index.

To: netdev@xxxxxxxxxxx
Subject: Fw: Oops in dev_get_by_index.
From: Andrew Morton <akpm@xxxxxxxx>
Date: Tue, 22 Jun 2004 12:40:31 -0700
Cc: simon@xxxxxxxxxxxxxxxxx
Sender: netdev-bounce@xxxxxxxxxxx

Begin forwarded message:

Date: Tue, 22 Jun 2004 19:57:02 +0100
From: Simon Kelley <simon@xxxxxxxxxxxxxxxxx>
To: LKML <linux-kernel@xxxxxxxxxxxxxxx>
Cc: simon@xxxxxxxxxxxxxxxxx
Subject: Oops in dev_get_by_index.




I've received a bug-report that dnsmasq 
(http://www.thekelleys.org.uk/dnsmasq) is provoking a kernel Oops and 
I'm posting everything I know about it here in the hope that someone who 
knows the relevant code will be able to see instantly what the problem is.

The Oops looks like this:

CPU:    0
EIP:    0060:[<02237d52>]    Not tainted
EFLAGS: 00010202   (2.6.6-1.435)
EIP is at __dev_get_by_index+0x14/0x2b
eax: 10a64134   ebx: 02cabef8   ecx: 00000006   edx: 00000000
esi: 075ab000   edi: 00008910   ebp: fef735e0   esp: 02cabef0
ds: 007b   es: 007b   ss: 0068
Process dnsmasq (pid: 7524, threadinfo=02cab000 task=0e3f7270)
Stack: 02238e9a 00000000 00000000 00000000 080570eb 0977bdd0 00000006 
fef73608
00ae007e 00b44ffc fef735e0 075ab000 00008910 02266641 02239733 fef735e0
00000008 00000000 11566b80 09774436 000001da 02cabf70 0000000c 0000000c
Call Trace:
   [<02238e9a>] dev_ifname+0x30/0x66
   [<02266641>] udp_ioctl+0x0/0x6e
   [<02239733>] dev_ioctl+0x83/0x283
   [<02266641>] udp_ioctl+0x0/0x6e
   [<0226c323>] inet_ioctl+0x6e/0x73
   [<02232936>] sock_ioctl+0x26d/0x285
   [<0214f7f6>] sys_ioctl+0x1f2/0x224

Code: 0f 18 02 90 2d 34 01 00 00 39 48 34 74 08 85 d2 89 d0 75 ea

This is the kernel part of the ioctl in dnsmasq-2.8/src/forward.c which does

ifr.ifr_ifindex = if_index;
ioctl(fd, SIOCGIFNAME, &ifr)

were the fd is a UDP socket which has had recvmsg called to get the 
datagram and IP_PKTINFO auxiliary data from which the if_index is derived.

The kernel version in question is 2.6.6-1.435 from Fedora.

There's a possibility if IPv6 involvement. The kernel has IPv6 loaded 
and the network interfaces have IPv6 addresses. Under those 
circumstances dnsmasq will bind the IPv6 addresses automatically. I have 
no way of telling if the Oops was as a result of receiving an IPv6 
datagram, but it is unlikely since IPv6 is not in active use.

The Oops happens rarely: there is no known way to reproduce it on demand.

That's all the information  I have: can anybody spot a problem?

Cheers,

Simon.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

<Prev in Thread] Current Thread [Next in Thread>
  • Fw: Oops in dev_get_by_index., Andrew Morton <=