Bug 8999
Summary: | Nvidia MCP51 - autoreset network card (2.6.22.5) - forcedeth | ||
---|---|---|---|
Product: | Drivers | Reporter: | Pol Hallen (polhallen) |
Component: | Network | Assignee: | platform_i386 |
Status: | REJECTED INSUFFICIENT_DATA | ||
Severity: | high | CC: | akpm, andi-bz, bugzilla77 |
Priority: | P1 | ||
Hardware: | All | ||
OS: | Linux | ||
Kernel Version: | 2.6.22.5 | Subsystem: | |
Regression: | --- | Bisected commit-id: |
Description
Pol Hallen
2007-09-09 04:00:41 UTC
So that's the third type of NIC you've used on that motherboard, yes? If so, it would appear to be some infrequently-occurring interrupt lossage. Over in http://bugzilla.kernel.org/show_bug.cgi?id=8700 Francois said "Problem is said to happen from 2.6.19-1.2895 (and supposely not in 2.6.18-1.2798)". Has anyone confrimed that any earlier kernel was OK? If so, whcih version? > So that's the third type of NIC you've used on that motherboard, yes? There's an integrated network card, and I added a pci network card (eth2) Because also with windows the integrated network card don't run. The system doesn't see it. Like the ethernet doesn't exist. > Over in http://bugzilla.kernel.org/show_bug.cgi?id=8700 Francois said > "Problem is said to happen from 2.6.19-1.2895 (and supposely not in > 2.6.18-1.2798)". Has anyone confrimed that any earlier kernel was OK? > If so, whcih version? I remember that old version's kernel (2.6.18 and later) had the problem described in 8700 bug. But with 2.6.22.5 kernel the problem is "only" autoreset network, so, I lost the transfers, I restart the network and my system run ok. Pol Might be a temporary solution if I set a traffic shape on my system, so I set (example) 2Mb/s on ethernet card? Thanks :-) Pol The driver was working OK on 2.6.17 (mandriva release so could have patchs ?) The only problem was a "Too many interrupts (6)" showing many times, but the card was not stopping. Ok :-) I'll try 2.6.17 But the problem is that I must reload ethernet interfaces after the problem :-( [...] NETDEV WATCHDOG: eth2: transmit timed out eth2: transmit timed out, tx_status 00 status e000. diagnostics: net 0cc0 media 8802 dma 00a00021 fifo 0000 Flags; bus-master 1, dirty 453299(3) current 453315(3) Transmit list 1f91f3e0 vs. df91f3e0. 0: @df91f200 length 8000004e status 0000004e 1: @df91f2a0 length 8000004e status 8000004e 2: @df91f340 length 8000004e status 8000004e 3: @df91f3e0 length 8000005c status 0000005c 4: @df91f480 length 80000048 status 00000048 5: @df91f520 length 8000005c status 0000005c 6: @df91f5c0 length 80000048 status 00000048 7: @df91f660 length 8000005c status 0000005c 8: @df91f700 length 80000048 status 00000048 9: @df91f7a0 length 80000048 status 00000048 10: @df91f840 length 800000e1 status 000000e1 11: @df91f8e0 length 80000042 status 00000042 12: @df91f980 length 80000042 status 00000042 13: @df91fa20 length 80000042 status 00000042 14: @df91fac0 length 8000004e status 0000004e 15: @df91fb60 length 8000004e status 0000004e eth2: Resetting the Tx ring pointer. [...] because my ethernet went down. I don't understand :-( I've the same problem (resetting the Tx ring pointer) also without forcedeth module! If isn't forcedeth the problem, where can I probe? I also tried with several network card (realtek, intel, etc.) but I've the same problem :-/ Is it a possible bug in amd74xx module? Can U help me please? Thanks Pol don't know, sorry. Maybe the machine has a bad power supply, and that is affecting the MAC hardware on both the nvidia and 3c59x interfaces? I'm out of myself :-) I connected to my server the ieee1394 (over ethernet) and my system is ok! ;-) I tried for many hours same tranfers! But I need the network card to run vmware (because vmware use the network to communicate with the same hard disk). When the system crash (because sometime the network reset itself - sometime the monitors power down and mouse+keyboard doesn't run) the ieee1394 (from other hosts) ping but the network interface doesn't ping :-| What I can do is reset my system :-(( What I'll do?! Thanks! (Using winXP it's all ok) Any luck with this patch ? http://git.kernel.org/?p=linux/kernel/git/stable/linux-2.6.22.y.git;a=commit;h=108c55dda1bd750b87b5fccc9cd4658222cf2dc3 PS: Pol hallen: you don't need NIC card for VMware. Just set up a bridge interface. A thing: is forcedeth driver only a network driver? Or also chipset driver? Because I turned off my integrated nvidia network card, and I using another (pci) network card. Thanks! Pol Excuse me if I insist: my problem happens also with firewire card (over tcp). I think that the problem isn't in forcedeth.c driver but into the nvidia chipset. I should change my motherboard? (how I'll buy? core duo 2) I need help because this is mine main computer. Thanks. Pol In you hardware description i saw: Ethernet controller: 3Com Corporation 3c905 100BaseTX [Boomerang] If you are using this type of network card, check my patch for increse RING BUFFER size at : http://linux.mantech.ro/3c59x.html This trick may be useful for other network cards. I run a router with 5 3COM 3c905B 100BaseTX [Cyclon] network cards for more that 1 week with those RING Buffers and works perfectly. Rarely apears some overruns packets, on a router with high load up to 80-90Mbit upload or download (~140-150Mbit total in full duplex) Pol, is this still a problem with recent kernel versions ? > is this still a problem with recent kernel versions ?
Hi,
I already sold those motherboard 1 year ago.
I'm sorry :-|
Thanks
Pol
Thanks for the update |