From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; fr-FR; rv:1.7.12) Gecko/20050920 Firefox/1.0.7 Description of problem: Starting with kernel 2.6.14, my Ethernet stops working as soon as the traffic picks up on that Ethernet (ex. running a "vncviewer" across my LAN, it stops right during the initial transfer of the desktop). Version-Release number of selected component (if applicable): kernel-2.6.14-1.1637_FC4 How reproducible: Always Steps to Reproduce: 1. boot with kernel 2.6.14 2. start some heavy traffic application 3. within one second I get the error messages reproduced below in syslog Actual Results: Nov 12 16:38:37 prt3 kernel: NETDEV WATCHDOG: eth0: transmit timed out Nov 12 16:38:37 prt3 kernel: eth0: transmit timed out, tx_status 00 status 8000. Nov 12 16:38:37 prt3 kernel: diagnostics: net 0cc6 media 8880 dma 000000a0 fifo 0000 Nov 12 16:38:37 prt3 kernel: Flags; bus-master 1, dirty 765(13) current 781(13) Nov 12 16:38:37 prt3 kernel: Transmit list 01bf2a20 vs. c1bf2a20. Nov 12 16:38:37 prt3 kernel: 0: @c1bf2200 length 00000042 status 000005ea Nov 12 16:38:37 prt3 kernel: 1: @c1bf22a0 length 00000042 status 000005ea Nov 12 16:38:37 prt3 kernel: 2: @c1bf2340 length 00000042 status 000005ea Nov 12 16:38:37 prt3 kernel: 3: @c1bf23e0 length 00000042 status 000005ea Nov 12 16:38:37 prt3 kernel: 4: @c1bf2480 length 00000042 status 000005ea Nov 12 16:38:37 prt3 kernel: 5: @c1bf2520 length 00000042 status 000005ea Nov 12 16:38:37 prt3 kernel: 6: @c1bf25c0 length 00000042 status 000005ea Nov 12 16:38:37 prt3 kernel: 7: @c1bf2660 length 00000042 status 00000b92 Nov 12 16:38:37 prt3 kernel: 8: @c1bf2700 length 00000042 status 00000b92 Nov 12 16:38:37 prt3 kernel: 9: @c1bf27a0 length 80000042 status 00000042 Nov 12 16:38:37 prt3 kernel: 10: @c1bf2840 length 80000042 status 00000042 Nov 12 16:38:37 prt3 kernel: 11: @c1bf28e0 length 80000042 status 80000042 Nov 12 16:38:37 prt3 kernel: 12: @c1bf2980 length 80000042 status 80000042 Nov 12 16:38:37 prt3 kernel: 13: @c1bf2a20 length 00000042 status 00000b92 Nov 12 16:38:37 prt3 kernel: 14: @c1bf2ac0 length 00000042 status 00000b92 Nov 12 16:38:37 prt3 kernel: 15: @c1bf2b60 length 00000042 status 000005ea Expected Results: Traffic should flow silently, as is the case with older kernel (currently running 2.6.13-1.1532_FC4). Additional info: The interface giving trouble is member of a bridge: bridge name bridge id STP enabled interfaces br0 8000.00b0d014a714 no eth0 tap0 Here's some info about the hardware itself: 02:0c.0 Ethernet controller: 3Com Corporation 3c905C-TX/TX-M [Tornado] (rev 78) Subsystem: Dell: Unknown device 00c7 Flags: bus master, medium devsel, latency 64, IRQ 3 I/O ports at ec00 [size=128] Memory at ff2fec00 (32-bit, non-prefetchable) [size=128] Expansion ROM at 44000000 [disabled] [size=128K] Capabilities: <available only to root>
Just gave a try to the newly released 2.6.14-1.1644_FC4 and found the same problem. I must clarify something from my initial report: the "vncviewer" that causes this problem is running on a remote machine connecting to the "vncserver" that's running on the machine where the failure occurs. In other words, it suggest that the problem occurs when a heavy burst of traffic is beeing _transmitted_. Still no problem with 2.6.13-1.1532_FC4. Regards.
The fedora-netdev kernels are available here: http://people.redhat.com/linville/kernels/fedora-netdev/ Please give those a try and post the results here...thanks!
Hi John, I have just installed FC4.netdev.3 (kernel-2.6.14-1.1644_FC4.netdev.3) and it resolves this issue. In order to compile the VMware modules, I have also installed the matching kernel-devel. Except for a minor issue (kernel built with gcc 4.0.2 while I only have 4.0.1 so far) everything works fine. (The gcc issue did not impair the process.) Of course, I have only tested superficially so far. Should anything else be adversely affected by this release, I will let you know. Thanks.
I'm going to close this as fixed upstream. I can't be sure when it will be in the official Fedora kernels, but it should remain in Fedora-netdev until then.
I've just found this issue is resolved in mainstream Fedora kernel 2.6.14-1.1653_FC4. I have changed the status accordingly.