Prev: I GOT $2500 FROM PAYPAL....
Next: Directory server help
From: Thomas Glanzmann on 4 Aug 2010 06:36
Hello,
9 months ago I reported that the tcp checksums on a T5520 using the
built-in e1000 network card are calculated in some cases non RFC
conform. At the same time I posted how to disable tcp offloading to
avoid the miscalculation on the broken driver:

Put ,,set ip:dohwcksum=0'' into /etc/system

Today we patched the system and after that the tcp connection was
slagish (packet lost on the same broadcast domain) so we disabled tcp
offloading again in /etc/system and rebooted. The slagish connection was
gone, but the tcp checksums are still computed non RFC 1624 conform. So
we tried the second method to turn of tcp offloading by adding the
following two lines to /kernel/drv/e1000g.conf:

tx_hcksum_enable=0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0;
lso_enable=0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0;

After a reboot the tcp checksum connection was no longer slagish and the
tcp offloading engine seems to be turned of because the checksums are
calculated RFC 1624 conform.

I didn't verify but suspect the patch 142257-05 to be responsible for no
longer be able to turn of tcp offloading using the /etc/system statement
(see above) without rendering the network connection useless.

Thomas
 | 
Pages: 1
Prev: I GOT $2500 FROM PAYPAL....
Next: Directory server help