[tor-bugs] #27319 [Core Tor/Tor]: remove buf_get_oldest_chunk_timestamp()

[Tor Bug Tracker & Wiki]

#27319: remove buf_get_oldest_chunk_timestamp()
-------------------------------------+----------------------------------
 Reporter:  cypherpunks3             |          Owner:  (none)
     Type:  defect                   |         Status:  new
 Priority:  Medium                   |      Milestone:  Tor: unspecified
Component:  Core Tor/Tor             |        Version:
 Severity:  Normal                   |     Resolution:
 Keywords:  datatypes, buffers, oom  |  Actual Points:
Parent ID:  #23878                   |         Points:
 Reviewer:                           |        Sponsor:
-------------------------------------+----------------------------------

Comment (by cyberpunks):

 Replying to [comment:4 nickm]:
 > instead of killing the connections that had been stalled for the
 longest, it would kill connections that had been making steady progress if
 they had never flushed fully.

 This is true, I was kinda hoping that the set of connections that had not
 stalled yet also had never been flushed fully would be miniscule to
 nonexistent in the real world. Some experimental logging might be able to
 demonstrate if that's true or not? Or, instead of recording a timestamp
 when the buffere rose from containing zero bytes to nonzero bytes, record
 when it rose above `[extremely low threshold here]`, low enough to not be
 scaleable with a parallel sniper attack.

 Or to more conservatively match current behavior with less abstraction
 burden on buf_t, there is a third option:
 * Store a smartlist of uintptr_t alongside the buf_t, add an API for
 `buf_get_num_chunks()`, and add/remove timestamps from the smartlist as
 the number of chunks increases/decreases.

--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/27319#comment:6>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online