Shave off another 4.7%: remove a linear search when figuring out which circuits wanted us to open a given OR connection.


svn:r5489

svn:r5488

svn:r5487

Ive spent about 10 hours benchmarking permutations on our counter-mode implementation.  This is the one that flies. (Avoid OpenSSL; optimizze rijndael calling convention to avoid needing to marshall and unmarshall counter.) This should speed up cell encryption by about 27%.


svn:r5486

svn:r5485

assert_buf_ok was showing up in the profile; remove it from critical-path functions, since it hasnt triggered for ages.


svn:r5479

svn:r5478

svn:r5477

to one that does.


svn:r5475

so they weren't showing up.


svn:r5474

them already in do_hup() by marking our log files as temporary and then
rotating them in options_init_from_torrc().


svn:r5473

svn:r5472

svn:r5471

Move connection_or_remove_from_identity_map() to connection_unlink, but dont remove the other; just make it warn.


svn:r5470

Rename connection_get_by_identity_digest to reflect that it is OR-only.  Make it use a hashtable instead of a linear search.


svn:r5469

svn:r5468

svn:r5467

Fix a bug weasel found: when we are an authority and somebody requests "all" statuses, cconsider regenerating our own.


svn:r5466

Make arguments to circuit_set_circid_orconn match the connection that is closing in circuit_about_to_close_connection. This accounts for our mysterious "arbitrary ram gets decremented!" bug.


svn:r5465

svn:r5464

conn over and over. that's really bad for memory.


svn:r5463

(nick, do i have my libevent strerror dance moves correct?)


svn:r5462

svn:r5461

Weasel says circuit_get_by_conn is his main timesink.  Most of its
users were just checking whether OR conns had circuits, so add a
circuit count to OR conns, and check that. One was
circuit_about_to_close_conn, which was doing an O(n^2) series of calls
to get all circs on an OR conn, so make an O(n) function for that.
Finally, circuit_get_by_edge_conn was using it as a sanity test that
has been around for a while but never found any actualy insanity, so
kill that.

circuit_get_by_conn is finally dead, which is good, since it was never
sane to begin with.


svn:r5460

svn:r5459

svn:r5458

separately. It's important to keep them separate because internal
circuits have their last hops picked like middle hops, rather than like
exit hops. So exiting on them will break the user's expectations.

- Stop cannibalizing internal circuits for general exits, and stop
  cannibalizing exit circuits for rendezvous stuff.

- Don't let new exit streams attach to internal circuits.

- When deciding if we have enough circuits for internal and for exit,
  don't count the wrong ones.

- Treat predicted resolves as predicted port 80 exits.


svn:r5457

for non-OR conns. this should save a bit of time.


svn:r5456

This is not a real fix. I didn't look at the rest of the code.
Nick?


svn:r5455

svn:r5454

(intended to be cannibalized later for rendezvous and introduction
circuits), we were picking them so that they had useful exit nodes. There
was no need for this, and it actually aids some statistical attacks.


svn:r5453

svn:r5452

svn:r5451

svn:r5450

svn:r5449

svn:r5448

svn:r5447

svn:r5446

Make dropping very old descriptors faster (remove old correctness check). Track bytes dropped that are still in our store or journal, and rebuild when it gets very high.


svn:r5445

svn:r5444