When running one of the gentoo-sources-2.4.22-r[123] kernels, I am having what I think are networking problems. The most re-producable symptom is premature termination of SMTP connections. This only occurs when a large email is sent to my Gentoo server and only when the SMTP connection uses TLS(SSL). I looked at some Ethereal traces to see who was terminating the connection and it is sometimes the client and sometimes the server. The best explanation I can come up with is that the network packets are getting corrupted somewhere and so one end or the other disconnects due to SSL protocol errors (there is nothing informative in the logs). I have also seen some corruption of files on a samba share. This is harder to reproduce, so I cannot say with certainty that it is caused by the same problem, but it does appear to be consistent with the SMTP symptoms described above. I have tested various kernels with the following results... gentoo-sources-2.4.20-r5 OK gentoo-sources-2.4.20-r10 OK gentoo-sources-2.4.22-r1 Bust gentoo-sources-2.4.22-r2 Bust gentoo-sources-2.4.22-r3 Bust vanilla-sources-2.4.22 OK gs-sources-2.4.22_pre2 OK gs-sources-2.4.23_pre8-r2 OK vanilla-sources-2.4.24 OK Reproducible: Always Steps to Reproduce: 1. Boot to a gentoo-sources-2.4.22-r[123] kernel. 2. Send a large email to the server using an encrypted (TLS) SMTP connection. Actual Results: The SMTP connection is terminated by the client or server before the transfer is completed. Expected Results: The email should be delived to the server without errors. I tried different email clients, but they all show the same problem. I tried updating my MTA (exim) to the latest version, but the errors persisted. I tried other SSL operations, such as scp'ing large files, but they were fine. The only variable that seems to affect the problem is the running kernel version. I would like to try modifying the gentoo-sources patch set for gentoo-sources- 2.4.22-r3 to try and pin down which patch is causing the problems. Is there a place where I can find the individual patch sets that went into distfiles/gentoo-sources-2.4.22-r2.patch.bz2? Chris.
Can give us a little more info? nic, other hardware specs, etc.
Hi, It's an Athlon XP 1800, with FSB at 100MHz (not 133). The NIC is a PCI Intel PRO/100. Attaching dmesg output, /proc/pci & /usr/src/linux-2.4.22-gentoo-r3/.config. lsmod shows I have the usbcore, ide-scsi, scsi_mod & e100 modules loaded. As I said, if you can furnish me with access to the individual patches in gentoo-sources-2.4.22, them I'm prepared to narrow down the problem to one of them (by adding them to the working vanilla kernel until it breaks). Chris.
Created attachment 23386 [details] Kernel config for one of the broken kernels (gentoo-sources-2.4.22-r3) All the other kernels I tested used this .config file after being run through make oldconfig.
Created attachment 23387 [details] dmesg output for the affected system (when running one of the OK gs-sources kernels)
Created attachment 23389 [details] contents of /proc/pci on the affected system.
what happens if you try the eepro100 module instead of the e100 module?
Hi, I just tried the eepro100 module against 2.4.22-gentoo-r3. No difference; the problem persists. Chris.
I wonder if you could try something else for me? Try reverting thie patch: http://gentoo.brianandsara.net/kernel/gentoo-sources-2.4.22/036_fast-csum See if that fixes your issues.
I reverted 036_fast-csum from gentoo-source-2.4.22-r3 and the problem goes away. Hooray. How did you guess it would be that patch? :-)
Wasn't me, you can thank plasmaroo for that one. I'll get a -r5 out as soon as I can. Thanks for testing that for us.
-r5 is in cvs, give it a little while then sync, then update. Thanks for the report and testing.
Thanks for your prompt responses. I'll test -r5 just as soon as bug 11392 is fixed (!) Chris.
if the md5sum for /usr/portage/distfiles/gentoo-sources-2.4.22-r5.patch.bz2 isn't 7f4a97d9c29f7dfc959a7a7efb077e29 you need to rm it, emerge --sync, and retry. There was a bad patch on the mirrors for a little while.
I've tested gentoo-sources-2.4.22-r5 and it looks okay too. Many thanks, Chris.
This was resolved quite a while ago. Time for full closure.