nihilist/tor
1
0
Fork 0
mirror of https://gitlab.torproject.org/tpo/core/tor.git synced 2024-11-27 22:03:31 +01:00
Code Issues Packages Projects Releases Wiki Activity
38,263 Commits 53 Branches 630 Tags 128 MiB
7aa496a2e0
Commit Graph

38263 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
Mike Perry
a956b20c1d Add a delta parameter to Vegas. 
This allows us to cap the queue use during steady state. In shadow, this
reduced instances of long circuit queues at relays.
 2022-02-22 19:28:35 +00:00
Mike Perry
3a3f111b23 Exit slow start at the gamma threshold 
This improves performance and fairness.
 2022-02-22 19:28:35 +00:00
Mike Perry
1960bf09f3 Increment by at least 2 sendme_incs in slow start 2022-02-22 19:28:35 +00:00
Mike Perry
d62f9c9d00 Only apply more frequent cwnd updates after slow start. 2022-02-22 19:28:35 +00:00
Mike Perry
43f6f3fd3a Make N_EWMA a percentage of CWND update rate, capped by a max value. 
This proved better in Shadow sims than just a multiple of CWND.
 2022-02-22 19:28:35 +00:00
Mike Perry
3079e2cacc Relocate two parameters from circuit scope to global scope. 
Changes in these will not affect in-progress transfers.
 2022-02-22 19:28:35 +00:00
Mike Perry
4f3a0e39cf Guard against 0 time delta in BDP calc. 
This can only happen in Shadow, but it will cause issues there.
 2022-02-22 19:28:35 +00:00
Mike Perry
8052d0c2c0 Increase RTT ratio used to detect monotime jumps/stalls. 
In Shadow, we saw RTT jumps as high as 1000 naturally.

So let's set this to 5000, to give us some breathing room.
 2022-02-22 19:28:35 +00:00
Mike Perry
338d00ba92 Fix NULL pointer deref in logs 2022-02-22 19:28:35 +00:00
Mike Perry
86f81abe30 Properly compute the number or recv cells from deliver_window 
Without this conversion, there is an implict 1000-recv_cells, which causes
the mod to fail if it is not a factor of 1000.
 2022-02-22 19:28:35 +00:00
Mike Perry
b2553bfba2 Use path type hint for Vegas queue parameters. 
These parameters will vary depending on path length, especially for onions.
 2022-02-22 19:28:35 +00:00
Mike Perry
0a6cde8756 Set new defaults for congestion control parameters. 
Defaults determined from Shadow experimentation.

More parameter functionality changes to follow.
 2022-02-22 19:28:35 +00:00
David Goulet
d4cf3fadec cc: Change edge_get_ccontrol() to look at both cpath and on_circuit 
Signed-off-by: David Goulet <dgoulet@torproject.org>
 2022-02-22 19:28:35 +00:00
David Goulet
27d948dab8 hs: Fix tests for congestion control 2022-02-22 19:28:35 +00:00
David Goulet
6f45ad3771 hs: Setup congestion control on service rends using intro data 
Signed-off-by: David Goulet <dgoulet@torproject.org>
 2022-02-22 19:28:35 +00:00
David Goulet
729dd14fde hs: Decode and cache the INTRODUCE cell congestion control extension 
Signed-off-by: David Goulet <dgoulet@torproject.org>
 2022-02-22 19:28:35 +00:00
David Goulet
38e9d9b465 hs: Build INTRODUCE extension in the encrypted section 
Signed-off-by: David Goulet <dgoulet@torproject.org>
 2022-02-22 19:28:35 +00:00
David Goulet
c79df44d22 hs: Setup congestion control on client rends 
Signed-off-by: David Goulet <dgoulet@torproject.org>
 2022-02-22 19:28:35 +00:00
David Goulet
0eaf0e8a31 hs: Republish onion descriptor on sendme_inc change 
Republishing is necessary to ensure that clients connect using the correct
sendme_inc upon any change. Additionally, introduction points must be
re-chosen, so that cached descriptors with old values are not usable.

We do not expect to change sendme_inc, unless cell size or TLS record size
changes, so this should be rare.

Signed-off-by: David Goulet <dgoulet@torproject.org>
 2022-02-22 19:28:34 +00:00
David Goulet
89f5eeefb8 hs: Decode flow-control line 
This puts the flow control version (unparsed) in the descriptor. The
client doesn't use it yet.

Signed-off-by: David Goulet <dgoulet@torproject.org>
 2022-02-22 19:28:34 +00:00
David Goulet
95639f35ae hs: Encode flow control value in the descriptor 
This simply adds the "flow-control" line, as detailed in prop324, to the
descriptor. No decoding is done at this commit.

Part of #40506
 2022-02-22 19:28:34 +00:00
David Goulet
02f4e7b42e cc: Export sendme_inc validation into public function 
This is needed for client validation of server descriptor value,
before launching a rend/intro.
 2022-02-22 19:28:34 +00:00
David Goulet
bbf160d311 cc: Use trunnel extension for ntorv3 circ parameters 
Signed-off-by: David Goulet <dgoulet@torproject.org>
 2022-02-22 19:28:34 +00:00
David Goulet
b5439d6bd0 trunnel: Make hs/cell_common.trunnel generic 
Move it to extension.trunnel instead so that extension ABI construction
can be used in other parts of tor than just HS cells.

Specifically, we'll use it in the ntorv3 data payload and make a
congestion control parameter extension using that binary structure.

Only rename. No code behavior changes.

Signed-off-by: David Goulet <dgoulet@torproject.org>
 2022-02-22 19:28:34 +00:00
David Goulet
dd938e58d3 protover: Add function to get the value of a single type 
We can now query the protover subsystem to get the current value we
support for a specific protover type.

This will be useful for prop324 onion service part which puts in the
FlowCtrl value in the service descriptor.

No behavior change.

Signed-off-by: David Goulet <dgoulet@torproject.org>
 2022-02-22 19:28:34 +00:00
Mike Perry
6b2086773c Add test for congestion control negotiation logic. 2022-02-22 19:28:34 +00:00
Mike Perry
f3283266c2 Add test for TAP vs ntor2+ntor3 onion queue ordering 2022-02-22 19:28:34 +00:00
Mike Perry
b2789ae72a Add hidden torrc option to always try CC negotiation. 
This is for public network testing and for sbws. Should not otherwise be used,
hence it is an undocumented __option.

The option deliberately does not allow force-disabling congestion control, as
this is bad for queueing and fairness.
 2022-02-22 19:28:34 +00:00
Mike Perry
a0eeadfba2 Handle other places that use onion handshake type values 
We want ntor and ntorv3 to use the same queues and stats.
 2022-02-22 19:28:34 +00:00
Mike Perry
812590f8aa Extend info argument updates for non-ntorv3 cases 2022-02-22 19:28:34 +00:00
Mike Perry
a07e008616 Congestion control usage of negotiated params 2022-02-22 19:28:34 +00:00
Mike Perry
1b1c261080 Protover flag handling for congestion control negotiation 2022-02-22 19:28:34 +00:00
Mike Perry
76bdadce12 Hook up client usage of congestion control negotiation 2022-02-22 19:28:34 +00:00
Mike Perry
b6d5fbba7d Implement congestion control parameter negotiation 2022-02-22 19:28:34 +00:00
Mike Perry
095224cdfa Add circuit param payload definitions to trunnel 2022-02-22 19:28:33 +00:00
Nick Mathewson
93318ba5e2 Convert TODO into TODO-324 for better visibility. 2022-02-22 19:28:33 +00:00
Nick Mathewson
662b9c1c0d Add stub argument for constructing congestion_control_t 2022-02-22 19:28:33 +00:00
Nick Mathewson
a511718a30 Negotiated circuit parameters must pass from worker. 2022-02-22 19:28:33 +00:00
Nick Mathewson
baaabb503c Use protover to signal support for ntor3 + congestion control. 2022-02-22 19:28:33 +00:00
Nick Mathewson
bd2e9a4409 Implement core of ntor3 negotiation. 
There are a lot of TODOs about what to send, whom to send it to, and
etc.
 2022-02-22 19:28:33 +00:00
Nick Mathewson
244444e8b1 Add an exported struct to onion handshakes for circuits params 
THis will eventually hold the congestion control parameters that we
negotiated, plus whatever else is relevant.
 2022-02-22 19:28:33 +00:00
Nick Mathewson
358ce9a19d Add a size argument for the buffer on onion handshake functions 2022-02-22 19:28:33 +00:00
David Goulet
d057f45c6c Merge branch 'tor-gitlab/mr/531' 2022-02-16 13:59:37 -05:00
David Goulet
7eb5437878 dirauth: Reject EOL 0.3.5.x relays 
Closes #40559

Signed-off-by: David Goulet <dgoulet@torproject.org>
 2022-02-16 13:59:30 -05:00
Jérôme Charaoui
6ef6d36296 Add CI triggers for Debian package builds 
This enables on-demand debian packaging CI builds on maintenance
branches, to replace (often unnecessary) daily scheduled builds.
 2022-02-15 17:14:34 -05:00
David Goulet
ea48ddc955 dns: Do not trigger overload general on DNS timeout 
This was missed in #40527 when the DNS timeout overload general signal
was removed.

Closes #40564

Signed-off-by: David Goulet <dgoulet@torproject.org>
 2022-02-14 13:20:53 -05:00
David Goulet
52d1c1fd31 Merge branch 'tor-gitlab/mr/518' 2022-02-09 10:36:26 -05:00
David Goulet
763d72238d kist: Don't try to flush empty outbuf 
It is possible that a scheduled channel ended up with 0 bytes in its
outbuf after the scheduling loop and having an outbuf table entry
indicating that we need to flush bytes on the wire after the loop.

This lead to attempt to write 0 bytes up to the TLS layer that would
prevent such action.

All in all, this fixes wasted CPU cycles on attempting to flush nothing.

Fixes #40548

Signed-off-by: David Goulet <dgoulet@torproject.org>
 2022-02-08 09:31:17 -05:00
David Goulet
9bf4a9e18a nss: Don't write empty payload on the wire 
Part of #40548

Signed-off-by: David Goulet <dgoulet@torproject.org>
 2022-02-08 09:31:13 -05:00
Nick Mathewson
9e86491cf8 Add a changes file for exit notice redesign (40529) 2022-02-07 11:37:50 -05:00