Binary package hint: dante-server
recently switched our proxy machine from i386 architecture to and AMD64 architecture, the SOCKS 5 proxy danted v1.1.18 used to work just fine on i386, but on AMD64, it crashes after starting the daemon with the following error:-
Aug 19 11:26:13 faramir danted[9982]: sockdexit(): terminating on signal 11
if I try to invoke the debugging switch to see what's going on, the process seg faults:-
danted -D -d
Syslog: Aug 19 11:47:30 (1187549250) danted[0]: (null): old: 5845040, new: 4344168
Segmentation fault
Dmesg: [38072.722771] danted[11981]: segfault at 0000000000000001 rip 00002aaaab18ffa0 rsp 00007fffffb892e8 error 4
I am running a dual quad core AMD server (8 processors), Kernel:_
Linux version 2.6.15-28-amd64-server (buildd@king) (gcc version 4.0.3 (Ubuntu 4.0.3-1ubuntu5)) #1 SMP Wed Jul 18 23:04:02 UTC 2007
After playing around a bit, I discovered that changing the number of servers to fork using the -N directive would allow me to get the danted daemon running after a fashion:-
Values of N from 1 to 4 had no effect, danted crashed out with the signal 11.
N = 5 or 6 left me with one root danted process
N=7 or 8 gave varying results, sometimes I had 3 danted process, others I had one main process and other defunct processes.
N=10 seemed to consistently give me one main danted process and 3 forked child processes.
which I'm assuming means I have been able to fork more processes than the system is trying to terminateas I increase the number of forked server processes. There's still a problem somewhere as the processes should not be being terminated.
Is this symptom still reproducible in 8.10 beta or later?