From Bugzilla Helper: User-Agent: Mozilla/4.78 [en] (X11; U; Linux 2.4.7-10 i686) Description of problem: Every 4-8 hours ethernet connectivity is lost and never spontaneously recovers afterwards. Transmit errors appear in /var/log/messages Version-Release number of selected component (if applicable): How reproducible: Always Steps to Reproduce: 1.Boot up 2.Wait 4-8 hours and tranmit errors appear 3. Reboot required Actual Results: Ethernet connectivity lost. Excerpt from /var/log/message showing problem. Message below repeat with changing "Status register values" Full log available on request. Nov 8 04:09:24 norm su(pam_unix)[27362]: session closed for user news Nov 8 04:47:35 norm sysctl: net.ipv4.ip_forward = 0 Nov 8 04:47:35 norm sysctl: net.ipv4.conf.default.rp_filter = 1 Nov 8 04:47:35 norm sysctl: kernel.sysrq = 0 Nov 8 04:47:35 norm network: Setting network parameters: succeeded Nov 8 04:47:36 norm network: Bringing up interface lo: succeeded Nov 8 04:47:36 norm kernel: PCI: Found IRQ 11 for device 00:0d.0 Nov 8 04:47:36 norm kernel: 3c59x: Donald Becker and others. www.scyld.com/network/vortex.html Nov 8 04:47:36 norm kernel: 00:0d.0: 3Com PCI 3c905 Boomerang 100baseTx at 0x1040. Vers LK1.1.16 Nov 8 04:47:36 norm kernel: eth0: Transmit error, Tx status register d0. Nov 8 04:47:36 norm kernel: Flags; bus-master 1, dirty 1(1) current 1(1) Nov 8 04:47:36 norm kernel: Transmit list 00000000 vs. c61a9240. Nov 8 04:47:36 norm kernel: 0: @c61a9200 length 8000002a status 8000002a Nov 8 04:47:36 norm kernel: 1: @c61a9240 length 00000000 status 00000000 Nov 8 04:47:36 norm kernel: 2: @c61a9280 length 00000000 status 00000000 Nov 8 04:47:36 norm kernel: 3: @c61a92c0 length 00000000 status 00000000 Nov 8 04:47:36 norm kernel: 4: @c61a9300 length 00000000 status 00000000 Nov 8 04:47:36 norm kernel: 5: @c61a9340 length 00000000 status 00000000 Nov 8 04:47:36 norm kernel: 6: @c61a9380 length 00000000 status 00000000 Nov 8 04:47:36 norm kernel: 7: @c61a93c0 length 00000000 status 00000000 Nov 8 04:47:36 norm kernel: 8: @c61a9400 length 00000000 status 00000000 Nov 8 04:47:36 norm kernel: 9: @c61a9440 length 00000000 status 00000000 Nov 8 04:47:36 norm kernel: 10: @c61a9480 length 00000000 status 00000000 Nov 8 04:47:36 norm kernel: 11: @c61a94c0 length 00000000 status 00000000 Nov 8 04:47:36 norm kernel: 12: @c61a9500 length 00000000 status 00000000 Nov 8 04:47:36 norm kernel: 13: @c61a9540 length 00000000 status 00000000 Nov 8 04:47:36 norm kernel: 14: @c61a9580 length 00000000 status 00000000 Nov 8 04:47:36 norm kernel: 15: @c61a95c0 length 00000000 status 00000000 Nov 8 04:47:38 norm network: Bringing up interface eth0: succeeded Nov 8 04:47:39 norm kernel: eth0: Host error, FIFO diagnostic register 2000. Nov 8 04:47:39 norm kernel: eth0: PCI bus error, bus status 00a00021 Nov 8 04:47:40 norm kernel: eth0: Transmit error, Tx status register d0. Nov 8 04:47:40 norm kernel: Flags; bus-master 1, dirty 1(1) current 1(1) Nov 8 04:47:40 norm kernel: Transmit list 00000000 vs. c61a9240. Nov 8 04:47:40 norm kernel: 0: @c61a9200 length 8000002a status 8000002a Nov 8 04:47:40 norm kernel: 1: @c61a9240 length 8000002a status 0000002a Nov 8 04:47:40 norm kernel: 2: @c61a9280 length 8000002a status 0000002a Nov 8 04:47:40 norm kernel: 3: @c61a92c0 length 8000002a status 8000002a Nov 8 04:47:40 norm kernel: 4: @c61a9300 length 00000000 status 00000000 Nov 8 04:47:40 norm kernel: 5: @c61a9340 length 00000000 status 00000000 Nov 8 04:47:40 norm kernel: 6: @c61a9380 length 00000000 status 00000000 Additional info: Connectivity maintained just like my RH6.0 installation which had been up for 30days before installation of RH 7.2 RH 7.2 was installed on a separate partition so RH 6 boot still possible via boot floppy.
Clarifications. System is a Compaq DeskPro. System ran RH6.1 for months without reboots until last week when RH 7.2 was added. Problem occurs when booting in RH7.2 and has occured consistently following first and second installation of RH7.2 System was not upgraded. RH7.2 was installed on a 2nd HD. RH6.1 still available for boot via floppy.
I had a similar problem after updating from 6.x to 7.2. The system had two 3c59x based cards and one of the cards was not working after a while. When swapping the two cards, the other interface was shutting down after a while. I decided to change the card instead of investigating the problem further as I needed to get the machine back up right away. You may want to see if the apm service is running as it was also showing up in the log at about the same time and I believe it may be the cause of the problem, restarting the network services.
Workaround: The problem can be fixed by shutting down Advanced Power Management (APM). There may be a better workaround than this involving setting NET_RESTART="no" NETFS_RESTART="no" in /etc/sysconfig/apmd to prevent only the network from being shut down but I have not yet tried this.
I have the same problem on a standard-non-apm-system with the stock RH-Kernel (2.4.7) - after a while, the 3c905 ("Votex/Boomerang/Tornado et. al") simply drops the link. I've had this problem also on a 2.4.17 custom compiled. Seems that my custom 2.4.18 works fine with those cards.
Can this be verified as fixed in the current 8.0, or stock kernel.org 3c59x driver?
No response after multiple months, I'm going to close this bug. Please file a new bug if this continues to occur in 8.0 release, 8.1 beta, or the latest errata kernel for 7.0/7.1/7.2.