Re: [RFC] Basing timeouts in NSE on host.times.timeout

[nnposter () users sourceforge net]

I am running into issues where http.post() calls are timing out. I have
traced the root cause to the newly implemented stdnse.get_timeout().
Specifically:

max_timeout == 8000
min_timeout == 1012 (artificially made unique for tracing purposes)
host.times.timeout == 0.1

The returned value (1012ms) causes:

NSOCK INFO [0.5100s] nsi_new2(): nsi_new (IOD #2)
NSOCK INFO [0.5130s] nsock_connect_ssl(): SSL connection requested to 10.x.x.x:443/tcp (IOD #2) EID 41
NSOCK INFO [0.6440s] nsock_trace_handler_callback(): Callback: SSL-CONNECT SUCCESS for EID 41 [10.x.x.x:443]
NSOCK INFO [0.6450s] nsock_trace_handler_callback(): Callback: WRITE SUCCESS for EID 51 [10.x.x.x:443]
NSOCK INFO [0.6450s] nsock_read(): Read request from IOD #2 [10.x.x.x:443] (timeout: 1012ms) EID 58
NSOCK INFO [1.6570s] nsock_trace_handler_callback(): Callback: READ TIMEOUT for EID 58 [10.x.x.x:443]
NSOCK INFO [1.6570s] nsi_delete(): nsi_delete (IOD #2)

By bumping up min_timeout in stdnse.get_timeout() I was able to observe
that this particular read requires about 2200ms.

It makes me wonder whether it is appropriate to derive read timeouts
from RTTs.


Cheers,
nnposter
_______________________________________________
Sent through the dev mailing list
http://nmap.org/mailman/listinfo/dev
Archived at http://seclists.org/nmap-dev/