[tor-bugs] #4626 [Tor Directory Authority]: Very high cpu usage for gabelmoo running master

[Tor Bug Tracker & Wiki]

#4626: Very high cpu usage for gabelmoo running master
-------------------------------------+--------------------------------------
 Reporter:  Sebastian                |          Owner:                    
     Type:  defect                   |         Status:  needs_review      
 Priority:  major                    |      Milestone:  Tor: 0.2.3.x-final
Component:  Tor Directory Authority  |        Version:                    
 Keywords:                           |         Parent:                    
   Points:                           |   Actualpoints:                    
-------------------------------------+--------------------------------------

Comment(by asn):

 It seems like Sebastan's profiles hint that the bug is in `44619d5e4`. And
 the bug exists even with `bug4626_try1` cherry-picked! Fun stuff!

 A difference between the pre-44619d5e4 and the post-44619d5e4
 `tor_tls_read()`, is that in the latter case, when we first detect the
 renegotiation we return `TOR_TLS_WANTREAD`. pre-44619d5e4 we used to
 return a positive integer, since we only detected renegotiation with the
 first application data packet.

 Could calling the TLS renegotiation callback and then returning
 `TOR_TLS_WANTREAD` cause all this chaos?

 The profiles seem to indicate a huge amount of SSL reads and writes
 (`connection_handle_write`, `flush_buf_tls`, etc.), and there also seems
 to be too much `ssl3_accept()`, which probably means that we are getting
 in the SSL protocol state machine loop more than we should.

-- 
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/4626#comment:14>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online