Help debugging hang with epoll_engine

[Daniel Miller <bonsaiviking () gmail com>]

List,

I think this may be the same issue I reported here: 
http://seclists.org/nmap-dev/2012/q1/435

This time, I was doing a large IPv6 scan (sudo ./nmap -6 -d2 -oA 
../top-1m-v6 -A -iL alexa-top-1m --log-errors -T4), which hung on NSE 
scanning:

> NSE Timing: About 99.24% done; ETC: 12:38 (0:00:19 remaining)
> NSE Timing: About 99.24% done; ETC: 12:39 (0:00:19 remaining)
> NSE Timing: About 99.24% done; ETC: 12:39 (0:00:20 remaining)
> NSE Timing: About 99.24% done; ETC: 12:40 (0:00:20 remaining)
> NSE Timing: About 99.24% done; ETC: 12:40 (0:00:20 remaining)
> NSE Timing: About 99.24% done; ETC: 12:41 (0:00:20 remaining)
> NSE Timing: About 99.24% done; ETC: 12:41 (0:00:21 remaining)
> NSE Timing: About 99.24% done; ETC: 12:42 (0:00:21 remaining)
This time, I was running under sudo, which meant I couldn't get a 
traceback or detailed status line due to the issue in this thread: 
http://seclists.org/nmap-dev/2011/q3/413 (confirmed in this thread: 
http://seclists.org/nmap-dev/2012/q2/44)

I was, however, able to attach with strace, which contained lots of 
repetitions of these calls:

> gettimeofday({1339090519, 54857}, NULL) = 0
> ioctl(4, TIOCGPGRP, [135333485])        = -1 ENOTTY (Inappropriate 
> ioctl for device)
> gettimeofday({1339090519, 55094}, NULL) = 0
> gettimeofday({1339090519, 55202}, NULL) = 0
> epoll_wait(7, {}, 128, 50)              = 0
> gettimeofday({1339090519, 105501}, NULL) = 0
> gettimeofday({1339090519, 105596}, NULL) = 0
> ioctl(4, TIOCGPGRP, [135333485])        = -1 ENOTTY (Inappropriate 
> ioctl for device)
> gettimeofday({1339090519, 105835}, NULL) = 0
The epoll_wait call led me to the epoll_loop function in 
nsock/src/engine_epoll.c, but the loop must be higher up, since 
epoll_wait is returning 0, and the condition of the inner loop is on its 
return value being -1.

I'll be testing this with a debug build, and try to get a full backtrace 
(NSE and C), but I thought perhaps someone could reason something out 
from this.

Dan
_______________________________________________
Sent through the nmap-dev mailing list
http://cgi.insecure.org/mailman/listinfo/nmap-dev
Archived at http://seclists.org/nmap-dev/