networking-related oopses on 2.6.9

[Please Cc me, I'm not subscribed to either list.]

In the last few days, after about two months it was running, one of my
servers (an IBM B50) running 2.6.9 vanilla started oopsing.
A twin server running the same kernel had no problems, but it's less
loaded.
Now I switched it back to 2.6.6, which never showed problems (but I need
something newer, because SSM multicast is broken in this release).

I have been able to capture the latest two oopses from the serial
console, is this a known problem?

KERNEL: assertion (!skb_queue_empty(&sk->sk_write_queue)) failed at net/ipv4/tc
Oops: kernel access of bad area, sig: 11 [#1]
NIP: C01B3BBC LR: C01B3B8C SP: D01D7DF0 REGS: d01d7d40 TRAP: 0300    Not tainted
MSR: 00009032 EE: 1 PR: 0 FP: 0 ME: 1 IR/DR: 11
DAR: 00000048, DSISR: 40000000
TASK = cf5439d0[9999] 'bryar.cgi' THREAD: d01d6000Last syscall: 6
GPR00: 00000000 D01D7DF0 CF5439D0 000004B8 00000500 00000308 00000000 C034A250
GPR08: 000004B8 FFFFFFF5 00000000 00000000 34D0744B 1001B8D4 00000000 00000000
GPR16: 00000000 00000080 102E1FE8 00000000 102E1FDC 102E1FF0 00000000 10003078
GPR24: C0360000 C0340000 C0360000 000004B8 C0310000 CC4371E0 CC4373B4 00000000
NIP [c01b3bbc] tcp_retransmit_skb+0x54/0x39c
LR [c01b3b8c] tcp_retransmit_skb+0x24/0x39c
Call trace:
 [c01b61f0] tcp_retransmit_timer+0x134/0x47c
 [c01b65fc] tcp_write_timer+0xc4/0xec
 [c002396c] run_timer_softirq+0xf8/0x1ac
 [c001f004] __do_softirq+0xdc/0xec
 [c001f06c] do_softirq+0x58/0x5c
 [c00083bc] timer_interrupt+0x1fc/0x22c
 [c0006660] ret_from_except+0x0/0x14
Kernel panic - not syncing: Aiee, killing interrupt handler!


Oops: kernel access of bad area, sig: 11 [#1]
NIP: C01AB880 LR: C01AC6D0 SP: C030FBE0 REGS: c030fb30 TRAP: 0300    Not tainted
MSR: 00009032 EE: 1 PR: 0 FP: 0 ME: 1 IR/DR: 11
DAR: 00000050, DSISR: 40000000
TASK = c02973e0[0] 'swapper' THREAD: c030e000Last syscall: 120
GPR00: 00000000 C030FBE0 C02973E0 00000001 D21EEB34 00000003 00000001 0000021D
GPR08: 00000000 00000000 001AA19A 00000000 2401C088 00000000 00000000 00000000
GPR16: 00000000 00000000 00000000 00000000 C0340000 00000004 DFD37648 00000006
GPR24: C0310000 00000003 00000000 00000000 7ABF8951 D21EE960 D21EE960 D21EEB34
NIP [c01ab880] tcp_time_to_recover+0x84/0x1f4
LR [c01ac6d0] tcp_fastretrans_alert+0x1a4/0x7ac
Call trace:
 [c01ac6d0] tcp_fastretrans_alert+0x1a4/0x7ac
 [c01adc90] tcp_ack+0x178/0x560
 [c01b0a34] tcp_rcv_established+0x410/0x824
 [c020cf18] tcp_v6_do_rcv+0x2a0/0x320
 [c020d5d0] tcp_v6_rcv+0x638/0x854
 [c01ec188] ip6_input+0x1dc/0x384
 [c01ebe98] ipv6_rcv+0x1a8/0x2bc
 [c01882cc] netif_receive_skb+0x178/0x250
 [c018844c] process_backlog+0xa8/0x188
 [c01885c4] net_rx_action+0x98/0x178
 [c001f004] __do_softirq+0xdc/0xec
 [c001f06c] do_softirq+0x58/0x5c
 [c0007e70] do_IRQ+0xc8/0xcc
 [c0006660] ret_from_except+0x0/0x14
 [c0008140] default_idle+0x14/0x5c
Kernel panic - not syncing: Aiee, killing interrupt handler!

-- 
ciao,
Marco

Marco d'Itri [off-list ref] wrote:

I have been able to capture the latest two oopses from the serial
console, is this a known problem?

KERNEL: assertion (!skb_queue_empty(&sk->sk_write_queue)) failed at net/ipv4/tc

Yes these are known problems which should be fixed in 2.6.10.  However,
there is at least one other similar issue in 2.6.10 so you'll either
want to stay with 2.6.6 or go up to the latest bk snapshot.
-- 
Visit Openswan at http://www.openswan.org/
Email: Herbert Xu ~{PmV>HI~} [off-list ref]
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt