Re: bin/50045: wpa_supplicant unreliable, dropping connections

[pierre-philipp braun <pbraun%nethence.com@localhost>]

The following reply was made to PR bin/50045; it has been noted by GNATS.

From: pierre-philipp braun <pbraun%nethence.com@localhost>
To: gnats-bugs%NetBSD.org@localhost
Cc: gnats-admin%netbsd.org@localhost,  netbsd-bugs%netbsd.org@localhost
Subject: Re: bin/50045: wpa_supplicant unreliable, dropping connections
Date: Tue, 14 Jul 2015 03:23:12 +0200

 >  Perhaps renewal of security associations is timing out because of delays
 >  caused by buffer bloat in the driver?  That might be worth a look.  Does
 >  latency (measured by a ping) get very high?
 
 Yes it used to happen, I remember seeing ping responses up to 1900ms 
 instead of usual ~22ms.  However, today, what I get is just a *skip* of 
 ping sequences until I break (^C) the process requesting to much 
 traffic.  For example skipping ping probes from 510 to 523:
 64 bytes from x.x.x.x: icmp_seq=509 ttl=56 time=22.900188 ms
 64 bytes from x.x.x.x: icmp_seq=524 ttl=56 time=22.311495 ms
 
 uname -a:
 NetBSD Host-001 7.99.19 NetBSD 7.99.19 (GENERIC.201507090800Z) #0: Thu 
 Jul  9 08:50:21 UTC 2015 
 builds%b45.netbsd.org@localhost:/home/builds/ab/HEAD/amd64/201507090800Z-obj/home/source/ab/HEAD/src/sys/arch/amd64/compile/GENERIC 
 amd64
 
 When I break the pkg_add process fetching the binary from a near mirror, 
 the ping probes now resume!  So the WPA supplicant seems to be rather 
 more self-healing in -current.  Are there actually changes since 7.0_RC1 
 and current 2015-07-09?  As far as I can remember this didn't happen 
 before so I can only assume it's related to my userland upgrade.  For 
 the pings being skipped instead of getting high response time, however, 
 I don't know.
 
 Thanks