[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Thread Index]

[tor-bugs] #9917 [Firefox Patch Issues]: Linux Firefox ESR 17.0.9 appers to spin-wait rather than block-wait on SOCKS



#9917: Linux Firefox ESR 17.0.9 appers to spin-wait rather than block-wait on
SOCKS
----------------------------------+---------------------------
 Reporter:  starlight             |          Owner:  mikeperry
     Type:  defect                |         Status:  new
 Priority:  normal                |      Milestone:
Component:  Firefox Patch Issues  |        Version:
 Keywords:                        |  Actual Points:
Parent ID:                        |         Points:
----------------------------------+---------------------------
 Per the summary, Linux Firefox 17.0.9 as
 include in TBB x86_64 2.3.25-13 appears to
 poll-wait rather than block-wait on
 SOCKS requests.  Observed this when
 attempting to access an unresponsive
 site, in this case core.onion.

 The spin caused the VM to largely lock-up,
 badly enough that the kernel reported
 120 second non-responsive "hung-tasks"
 on several threads including 'tor'.
 However this particular aspect may be
 largely a result of running in a single
 VCPU KVM instance.

 See Firefox using all CPU in 'top' when
 this happens.  Behaves a little better
 when 'renice -n10' is used to lower FF's
 priority.

 Seems to me the behavior should be
 to block-wait and that other Firefox
 windows should remain responsive while
 one window is blocked waiting for a reply
 that will never arrive.

 Have set 'tor' threads to run realtime
 priority, RR class at -27 per 'top', which
 might affect the behavior here, though it
 really shouldn't matter.

-- 
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/9917>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online
_______________________________________________
tor-bugs mailing list
tor-bugs@xxxxxxxxxxxxxxxxxxxx
https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-bugs