r8169 puking

oogle · 11-07-2004, 01:04 AM

r8169 driver loads and works for awhile, After a long transfer amount, the nic dies ala:

r8169 eth1 PCI error (status: 0x8000). Device disabled.

nothing else pops in through dmesg, it just stops the device, and stops it forever. reloading the module is no good.

Maartenc · 11-24-2004, 02:08 PM

I also have a problem with the r8169 nic. My Debian machine dies after I transferred a few gigs. I cannot find an error message in the logs and there's no message on the screen. The whole machine just freezes. I have the problem with the 2.4 and latest 2.6 kernel.

Maybe the same problem?

Ueimor · 11-24-2004, 04:39 PM

The issue is most probably different. There has been a few reports of PCI errors
since the driver notice these. On x86, I'd be tempted to simply ignore it (assuming
the host system is not overclocked nor otherwise flaky).

If such is the case, can you simply comment out the 'status & SYSErr' test in
rtl8169_interrupt() and see if things perform better ?

If they don't, the kernel/compiler versions + complete dmesg after boot + lspci -vx
will be welcome (not that I'll necessarily have a fix...).

--
Ueimor

Sjorrit · 11-24-2004, 05:06 PM

Download and compile the latest driver from
http://www.realtek.com.tw/downloads/...L8169S-32%2f64

This solved the problem for my 8110s (uses the same driver as the 8169) with IPv4. No succes though on IPv6 as I posted in in
http://www.linuxquestions.org/questi...hreadid=258394

Greets,

Jorrit

Ueimor · 11-24-2004, 05:57 PM

rtl8169_tx_interrupt() in Realtek's driver is _very_ wrong. It may take some time
for you to notice but the symptoms will be pretty clear

)

If you can reproduce the issue with a recent 2.4.x kernel without any binary only
module loaded, I'll at least update the backport of the r8169 driver. OTOH, I have
no interest trying to fix bugs in the r8169 driver while some piece of code does
interesting things in the background.

--
Ueimor