Ideas, help? Samba and SCP crashes network on linux

[mike456]

Hello,

Summary: Weird network crashes when moving massive (300 MB+) amounts of files between Windows box -> Linux box (in direction Windows>Linux). Loose all TCP/IP sockets - can't open new windows in X, telnet/ftp/ssh sessions, can't ping anything, can't ping linux box from windows, SIGHUPing dhcpcd and ifconfig eth0 up/down do NOTHING. The box is dead in the water. File transfer hangs from linux side (just suddenly stops), windows box is fine (network okay). Experienced on both Samba (2.2.8a) and SCP (ssh 3.7.1_p2) when linux box is host/server and windows box is client. Somewhat intermittent and random, but consistently during large SMB or SCP transfers (at random # of bytes, random files, random directories...etc).

Some info-mation:
- Gentoo 2.4.20-r8 kernel
- AMD 1200 Duron, 512 MB, Asus A7V8X-X motherboard/chipset, built-in network (Realtek 10/100 onboard)
- Onboard audio
- ELSA (something) 32MB video card (TNT2).
- No overclocking on anything
- 2 HDs on same controller, 1 DVDrom on separate controller
- 10/100 Network
- DLink 704P router between internet and linux box.
- Windows box is W2K, SP3, also 10/100
- No PCI cards besides video on MB.

I don't suspect that it is a hardware problem, it seems like either (1) network - the DLink router may be doing weird things with packets -- somewhat unlikely or (2) Inherent in the kernel or (3) Samba and SSH (unlikely that it would be both). It could be a crappy network card.

Samba and SSH work fine (I got them set up properly) the only thing that is open to the outside world is port 22 (for SSH) from outside the router. All other traffic to the linux box is blocked from the outside world via the router.

I have no problem with long SSH sessions originating from the Linux box to other boxes.

Linux box is VERY stable otherwise. Uptimes of 29 days with no odd crashes are pretty usual. The only instability I've seen is this network transfer of LARGE files from a windows box.

I have no lockups when copying big files around between HDs, or zipping, taring or whatever. The lockups are network related (esp samba and ssh/scp) not controller/hard disk.

When the networking does crash on linux side, I can still do stuff on the machine (ie. reboot it). I just can't do anything network related because the networking is TOTALLY locked up, even the local loopback is toast.

Nothing interesting in metalogs critical or everything logs (I tried the whole USR1 business). The machine thinks everything is fine, but networking totally dies. Looked at smbd.log and ssh log, but also don't contain any hints.

Anyone seen this / have any ideas?

Thanks,
mike

[mike456]

Anyway,

Yeah mike. I know what your problem is. Lucky for you I've been there. :<

Turns out that the version of the Via Rhine-2 network card driver that was built into my motherboard is BUGGY. In fact, all the versions included up to gentoo-sources 2.4.20-r8 are the same buggy version.

If you don't believe me, check the kernel mailing list ... search for Via Rhine and you'll see what I mean. Everyone seems to get crashes transferring large files both to and from linux. Really Suxors!

The Solution?
I determined that the fixed Via Rhine driver was in kernel versions 2.4.22 and up. So, emerged the plain vanilla sources kernel and built.

Downside is unoptimized kernel, but Gentoo will catch up eventually with the correct via_rhine.c driver file (btw. version LK1.19 and up are OK, mine was 1.12 and causing problems... Check dmesg to see what version you are using.)

Now I just have to get my CUPS printing more than 4 pages at a time!

sigh
mike

[blight01]

Hello,

I have the same problem than mike456...The network goes down when I try to updload some big files from Windows to Linux through samba and kde on a VFAT PARTITION mounted into a directory of my root partition. I don't know why...And I use the Via Rhine driver, so It might be the cause of these networking disasters!

But what it is strange is when I download these big files directly into my root partitition, the problem seems to disappear. I saw that a vfat partition mounted into a directory of the root partition can cause some problem: https://forums.gentoo.org/viewtopic.php?t=108435&highlight=samba+problem+files

Does someone encounter the same mysterie?

Ciao.

[blight01]

Hello again,

I have just replaced my old via-rhine driver (ver 1.1.14) by a new one (ver 1.16) and I see that my problem remains...I cannot copy big files from a windows box into my linux box if these files are copied on a vfat partition mounted under a directory of the root partition.

This is just a useless message unless someone can help me...