I was running a samba server and notice following problems: - reading was ok - writing a small file was ok or corrupt (various) - writing a medium (50-100mb) file was big delays and file was corrupt - writing a big file (+700mb) crash windows (:P) - Connection reset in samba log - in ifconfig RX packets got errors After trying to find from where was the problem, i have check with mii-tool to see how the card was running. For mii-tool: 10baseTX, half duplex with no link (same goes for mii-diag) on my other computer (2.6.7-gentoo-r6) result was eth0: 10baseTX, half duplex, no link. Both cards are Ethernet controller: 3Com Corporation 3c905C-TX/TX-M [Tornado] (rev 74) Both cards are connect to a 10/100 full duplex switch that report both cards running at 10/100 full duplex (witch is really strange for one card if i trust mii-tool and normal for the other) With a backup kernel the 2.4.20-gaming-r6 mii-tool was working normally but not on my 2.4.20-gaming-r14. Trying another kernel (2.4.26-gentoo-r6), same, mii-tool fail to negociate... After looking at the backup kernel parameters, i notice i have compile the 3c59x inside the 2.4.20-gaming-r6 and inside the 2.6.7-r6 on the other computer, and as module on the two others. Now finally got it working with the 2.4.26-gentoo-r6 just by compiling inside the kernel (no more as module) Now: mii-tool, mii-diag works perfect mii-tool -v eth0: negotiated 100baseTx-FD flow-control, link ok product info: vendor 00:10:18, model 23 rev 6 basic mode: autonegotiation enabled basic status: autonegotiation complete, link ok capabilities: 100baseTx-FD 100baseTx-HD 10baseT-FD 10baseT-HD advertising: 100baseTx-FD 10baseT-FD flow-control link partner: 100baseTx-FD 100baseTx-HD 10baseT-FD 10baseT-HD flow-control Samba writes now works, big files, small delays, no error on nic, no error in samba logs, files not corrupt anymore... Reproducible: Always Steps to Reproduce: 1.Compile 3c59x as module 2.insmod 3c59x 3.mii-tool to test Actual Results: mii-tool results: eth0: 10baseTX, half duplex, no link Expected Results: mii-tool results: eth0: negotiated 100baseTx-FD flow-control, link ok - I'm not sure how many kernels are affects by this problem - I'm not sure how many cards are affects by this problem (at least the 3c905C-TX) - I'm not sure if only samba could be affect by that, but it was hard to see on a file copy with nfs from linux to linux (was longer, but working).
Oh :( Adding informations: motherboard is running i820 chipset (not sure it could help). I have saw some bug reports about a bug with 3c59x with gentoo livecd, might be the same if it is compile as module within the kernel use in the livecd... I have anyway decide to open a new bug because at least mii-tool is an easy way to see it in action... sorry if it's a dup :/
Looks like an upstream bug in the kernel (although it does sound odd..). If this is still an issue, could you please test on the newest releases of gentoo-sources and vanilla-sources? If the issue is present in either, please reopen this bug. Thanks