On Mon, 2009-12-07 at 08:27 +0100, Joerg Pareigis wrote: > > [...] > > I've built the module from the sources for the next stable kernel package, including this patch. So try using this: > > http://womble.decadent.org.uk/tmp/atl1e.ko > > > OK, thanks a lot. > Now I'm using this module. > Unfortunately this module runs not perfect yet. > At first I thought it's OK, because the output of "ls -lR /usr" is very fast, but when I start a X-Application through a > "ssh -X host" I see a lot of truncated pakets :-( > And also additional "ls -lR" tests produces truncated pakets... Perhaps there's a second bug in it (or elsewhere). The bug I could reproduce is definitely fixed now. I just re-tested that exact same module by running a netperf TCP streaming test: $ netperf -l 60 -H 10.42.43.10 -t TCP_MAERTS -- -H fe80::223:54ff:fe0b:134%2,inet6 TCP MAERTS TEST from ::0 (::) port 0 AF_INET6 to fe80::223:54ff:fe0b:134%2 (fe80::223:54ff:fe0b:134) port 0 AF_INET6 : demo Recv Send Send Socket Socket Message Elapsed Size Size Size Time Throughput bytes bytes bytes secs. 10^6bits/sec 87380 16384 16384 60.32 92.80 This is near-line-rate performance. The MAC and IPv6 stats on this end also showed no errors during this test. Ben. -- Ben Hutchings Logic doesn't apply to the real world. - Marvin Minsky
Attachment:
signature.asc
Description: This is a digitally signed message part