john reynolds wrote: > Michael, > I went back to gao11 and ran it overnight with the same > test, Web server refreshing every 1 sec and both ends > pinging each other. It did not lock up yet. gao66 would > have locked up by now. The only change I made in the > gao11 and gao66 eth drivers was I used the cirrus SPI > to get the hardware mac addr. I don't think this would > have caused this problem. Please make sure that you can reproduce the problem with an unmodified gao66. When there is a bug to track down, the least thing one would want is an unnecessary variable like this. Tell me if there is something that prevents you from testing with an unmodified kernel, we need to think about another possibility then. I don't know if it's really necessary for your application to set a certain MAC (instead of doing with a default one, be it static or random) but you can can set it from userspace with: ifconfig eth0 down # eth0 needs to be down to set the mac ifconfig eth0 hw ether 00:11:22:33:44:55 # set the mac Dont forget to bring eth0 up again. > May I can try gao66 with just > pinging. Would be great if we could trim it down to something that I can reproduce, but we should not spend too much time, in case that turns out to be hard / impossible. > I'll try to dig into this but I'm not sure I know what I'm doing :) The first thing I'd like to know is what gaoXX version introduced the bug. If you can spot the problem "using your eyes only" you can skip the next lines and send me a patch. Else we've got two options: a) use git-bisect to do a binary search for that version and review the changes once we've found it. If you're a bit familiar with git, skimming through [1] will get you started in no time. b) If you don't want to use git, we'll have to do the same thing manually[2]. Please tell me whether you'd like to continue with a) or b). [1] http://www.kernel.org/pub/software/scm/git/docs/git-bisect.html [2] linux-2.6.14/Documentation/BUG-HUNTING