Bug#642911: TX watchdog timeout on RTL8168B

[booster@wolke7.net]

Jonathan Nieder wrote:

Hi again,

Jonathan Nieder wrote:

  

Something very similar to the fix you tested plus a couple more r8169
fixes were applied upstream:

 - v3.2-rc5~38^2~5 (r8169: Rx FIFO overflow fixes, 2011-12-04), which
   is the fix you helped with
 - v3.2-rc5~38^2~4 (r8169: fix Rx index race between FIFO overflow
   recovery and NAPI handler, 2011-12-04)
 - v3.2-rc7~12^2~8 (r8169: fix Config2 MSIEnable bit setting,
   2011-12-15)
    

Thanks again for making it happen.

  

Am I correct in guessing that v3.2-rc7 from experimental works ok?
    

What kernel do you use these days?  Have you tried 3.2.y from
sid/experimental?  Would you be interested in pursuing a fix in
2.6.32.y or 3.0.y?

Jonathan

  

Hi Jonathan,

thanks for the new info about the fix and sorry for the delay in testing.

Today I updated to the latest sid kernel (3.2.0-1-686-pae #1 SMP Fri Feb 17 06:27:21 UTC 2012)
The first tests I've done so far by transfering ~30 GB with samba showed no more error messages in
the logs and the eth0 interface is stable. I'll keep the system up and running to see if all works OK.

If there are any backports of the fix to the 3.0 and 2.6 kernels I can test them as well if you like.

The only thing that's currently not working is the green transfer LED when the connection is to my GBit
switch. If the connection is to an 100MBit switch the LED is working as expected.

Gerd