Longstanding APIC/NE2K bug

DISCUSSION [email protected] 3 participants 3 messages Jun 17, 2001 – Jun 18, 2001

Share & tools

[email protected]

author

2001-06-17 04:07 UTC #0 lore raw

There has been a bug in the 2.4.x series of kernels for a long time (at least -pre9) concerning SMP and ne2k-pci.

Maciej W. Rozycki posted a patch back during 2.4.0 that fixed this problem "[patch] 2.4.0, 2.4.0-ac12: APIC lock-ups" in late January. I've been trying new kernels regularly since, and the patch doesn't seem to have made it in (tested 2.4.2, .3, .4 and .5). Falling back on my patched 2.4.0 works fine.

Symptoms: Network driver locks up. Repeated messages of "ETH0: Transmit timeout" occurs. Unloading and reloading network drivers does not help, reboot is required. Usually only triggered by heavy network traffic (300-400 megs at 700k or so usually does it).

Robert Cicconetti

Jens Gecius

2001-06-17 06:25 UTC #1 lore raw

Quoted 9 lines · [email protected] writes:

There has been a bug in the 2.4.x series of kernels for a long time (at least -pre9) concerning SMP and ne2k-pci.
Maciej W. Rozycki posted a patch back during 2.4.0 that fixed this problem "[patch] 2.4.0, 2.4.0-ac12: APIC lock-ups" in late January. I've been trying new kernels regularly since, and the patch doesn't seem to have made it in (tested 2.4.2, .3, .4 and .5). Falling back on my patched 2.4.0 works fine.

Quoted 4 lines

Symptoms: Network driver locks up. Repeated messages of "ETH0: Transmit timeout" occurs. Unloading and reloading network drivers does not help, reboot is required. Usually only triggered by heavy network traffic (300-400 megs at 700k or so usually does it).

This fits exactly my problems I mentioned a couple weeks ago. Same question here. Therefore my question: can we expect to see this patch implemented? If not, any other suggestions?

Signature · 2 lines

Tschoe,                    Get my gpg-public-key here
 Jens                     http://gecius.de/gpg-key.txt

Raphael Manfredi

2001-06-18 06:09 UTC #2 lore raw

In reply to [email protected]

Quoting [email protected] from ml.linux.kernel:
:There has been a bug in the 2.4.x series of kernels for a long time (at :least -pre9) concerning SMP and ne2k-pci.
:
:Symptoms: Network driver locks up. Repeated messages of "ETH0: Transmit :timeout" occurs. Unloading and reloading network drivers does not help, :reboot is required. Usually only triggered by heavy network traffic :(300-400 megs at 700k or so usually does it).

I concur. This happened to me tonight with:

	Linux nice 2.4.4-ac9 #1 SMP Wed May 16 16:17:59 MEST 2001 i686 unknown

Here's the syslogs, flushed to the remote loghost machine somehow, before the network became dead (closed connections, at the end of the logs).

NB: It all started with an APIC error (unexpected IRQ trap, meaning an error that passed through checksum controls).

Raphael

The logs:

Jun 18 04:36:23 nice kernel: unexpected IRQ trap at vector 7d 
Jun 18 04:36:35 nice kernel: NETDEV WATCHDOG: eth0: transmit timed out 
Jun 18 04:36:35 nice kernel: eth0: Tx queue start entry 11638512  dirty entry 11638508. 
Jun 18 04:36:47 nice kernel: eth0:  Tx descriptor 0 is 00002000. (queue head) 
Jun 18 04:36:47 nice kernel: eth0:  Tx descriptor 1 is 00002000. 
Jun 18 04:36:47 nice kernel: eth0:  Tx descriptor 2 is 00002000. 
Jun 18 04:36:47 nice kernel: eth0:  Tx descriptor 3 is 00002000. 
Jun 18 04:36:59 nice kernel: eth0: Setting 100mbps full-duplex based on auto-negotiated partner ability 45e1. 
Jun 18 04:37:11 nice kernel: NETDEV WATCHDOG: eth0: transmit timed out 
Jun 18 04:37:23 nice kernel: eth0:  Tx descriptor 0 is 00002000. (queue head) 
Jun 18 04:37:23 nice kernel: eth0:  Tx descriptor 1 is 00002000. 
Jun 18 04:37:23 nice kernel: eth0:  Tx descriptor 2 is 00002000. 
Jun 18 04:37:23 nice kernel: eth0: Tx queue start entry 4  dirty entry 0. 
Jun 18 04:37:35 nice kernel: eth0:  Tx descriptor 3 is 00002000. 
Jun 18 04:37:35 nice kernel: eth0: Setting 100mbps full-duplex based on auto-negotiated partner ability 45e1. 
Jun 18 04:38:11 nice kernel: NETDEV WATCHDOG: eth0: transmit timed out 
Jun 18 04:38:25 nice kernel: eth0:  Tx descriptor 0 is 00002000. (queue head) 
Jun 18 04:38:25 nice kernel: eth0:  Tx descriptor 1 is 00002000. 
Jun 18 04:38:25 nice kernel: eth0:  Tx descriptor 2 is 00002000. 
Jun 18 04:38:25 nice kernel: eth0: Tx queue start entry 4  dirty entry 0. 
Jun 18 04:38:35 nice kernel: eth0:  Tx descriptor 3 is 00002000. 
Jun 18 04:38:35 nice kernel: eth0: Setting 100mbps full-duplex based on auto-negotiated partner ability 45e1. 
Jun 18 04:38:47 nice kernel: nfs: server lyon not responding, still trying 
Jun 18 04:38:59 nice kernel: NETDEV WATCHDOG: eth0: transmit timed out 
Jun 18 04:38:59 nice kernel: eth0: Tx queue start entry 4  dirty entry 0. 
Jun 18 04:39:11 nice kernel: eth0:  Tx descriptor 0 is 00002000. (queue head) 
Jun 18 04:39:11 nice kernel: eth0:  Tx descriptor 1 is 00002000. 
Jun 18 04:39:11 nice kernel: eth0:  Tx descriptor 2 is 00002000. 
Jun 18 04:39:11 nice kernel: eth0:  Tx descriptor 3 is 00002000. 
Jun 18 04:39:23 nice kernel: eth0: Setting 100mbps full-duplex based on auto-negotiated partner ability 45e1. 
Jun 18 04:39:35 nice kernel: NETDEV WATCHDOG: eth0: transmit timed out 
Jun 18 04:39:35 nice kernel: eth0:  Tx descriptor 0 is 00002000. (queue head) 
Jun 18 04:39:35 nice kernel: eth0: Tx queue start entry 4  dirty entry 0. 
Jun 18 04:39:47 nice kernel: eth0:  Tx descriptor 1 is 00002000. 
Jun 18 04:39:47 nice kernel: eth0:  Tx descriptor 2 is 00002000. 
Jun 18 04:39:47 nice kernel: eth0:  Tx descriptor 3 is 00002000. 
Jun 18 04:39:59 nice kernel: eth0: Setting 100mbps full-duplex based on auto-negotiated partner ability 45e1. 
Jun 18 04:52:08 paris sshd[12120]: fatal: Read error from remote host: Connection timed out 
Jun 18 04:52:10 lyon sshd[800]: fatal: Read error from remote host: Connection timed out
Jun 18 04:54:03 lyon smbd[10932]: [2001/06/18 04:54:03, 0] lib/util_sock.c:read_socket_data(477) 
Jun 18 04:56:31 lyon sshd[997]: fatal: Read error from remote host: Connection timed out