On Mon, 8 Jun 2009, Christiano Farina Haesbaert wrote: > Hmm, I didn't know we had a soft limit, I thought you had to recompile > the kernel and tune max files, I agree, I'll add support for expanding > the limit. Is there any sane limit ? Or I can push as far as it goes ?
Does your diff preallocate all the fds? If not, then you should just let it support upto RLIMIT_NOFILE. > I didn't pay much attention to the kvars monitoring yet, are those > values per process in the kernel ? No, there is a global list of TCBs. See the kfind_tcb() function to see how I grovel through kmem to find them. You will have to modify that function to match your list of connections against the live TCBs to extract the values. Be warned, this isn't particularly pretty :) > Another feature I was thinking would be to dump the output in one file > per host. What do you guys think ? Besides correcting fd limit, udp > support, any other ideas ? Whatever the network hackers think is useful :) -d
