This patch is a follow up to https://webrtc-review.googlesource.com/c/src/+/260943
which made it possible to destroy a Port before a Connection (on that
port).
This patch "reverses" this, by adding RTC_DHECK, and fixes the Port
destructor to release the Connections before invalidating the WeakPtr<>.
Currently there are no known occurrences where a Connection is destroyed after it's Port is. But prior to the change in Port destructor, a bunch unit tests failed on the newly added DCHECKs.
In addition:
a) modify StunReqquestManager to remove entry from hash before calling callback. This makes it possible for callback to modify (clear) hash.
b) clear pending requests when disconnecting from port, "should not be needed", depends on a)
c) add a getter for pending_delete()
Bug: webrtc:13892, webrtc:13865
Change-Id: I5d18f2db8d93b7cc25d18bd620063589ee9257c9
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/322861
Reviewed-by: Per Kjellander <perkj@webrtc.org>
Commit-Queue: Jonas Oreland <jonaso@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#40907}
This patch hooks up the StunDictionary to Connection
and P2PTransportChannel.
Bug: webrtc:15392
Change-Id: Ibeea4d8706ebd42f2353d9d300631c02bf0d484d
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/315100
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Commit-Queue: Jonas Oreland <jonaso@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#40519}
Add an override to TurnPort for SendBindingErrorResponse to check
if a matching connection object exists before continuing. This is
needed to match with the check in `TurnPort::DispatchPacket` whereby
we forward calls to `Port` when no matching connections exist.
Bug: chromium:1395625
Change-Id: Idf1f898c2a93de6bc2832268db1cadd52cae23a9
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/287223
Reviewed-by: Sameer Vijaykar <samvi@google.com>
Commit-Queue: Tomas Gunnarsson <tommi@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#38871}
This reverts commit 869c87a2b9d9f4194d77dd30dc4175a2ecf28a74.
Reason for revert: Re-landing
Original change's description:
> Revert "Make deletion of Connection objects more deterministic."
>
> This reverts commit 942cac2e9e6a205fd673dc003a051cfb3867f2e3.
>
> Reason for revert: Reverting while downstream updates are made.
>
> Original change's description:
> > Make deletion of Connection objects more deterministic.
> >
> > This changes most deletion paths of Connection objects to go through
> > the owner class of the Connection instances, Port.
> >
> > In situations where Connection objects still need to be deleted
> > asynchronously, `async = true` can be passed to
> > `Port::DestroyConnection` and get the same behavior as
> > `Connection::Destroy` formerly gave.
> >
> > The `Destroy()` method still exists for downstream compatibility, but
> > instead of deleting connection objects asynchronously, the deletion
> > now happens synchronously via the Port class.
> >
> > Bug: webrtc:13892, webrtc:13865
> > Change-Id: I07edb7bb5e5d93b33542581b4b09def548de9e12
> > Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/259826
> > Reviewed-by: Danil Chapovalov <danilchap@webrtc.org>
> > Commit-Queue: Tomas Gunnarsson <tommi@webrtc.org>
> > Cr-Commit-Position: refs/heads/main@{#36676}
>
> Bug: webrtc:13892, webrtc:13865
> Change-Id: I37a15692c8201716402ba5c10f249e4d3754ce4c
> Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/260862
> Bot-Commit: rubber-stamper@appspot.gserviceaccount.com <rubber-stamper@appspot.gserviceaccount.com>
> Commit-Queue: Tomas Gunnarsson <tommi@webrtc.org>
> Cr-Commit-Position: refs/heads/main@{#36736}
Bug: webrtc:13892, webrtc:13865
Change-Id: I29da6c8899d8550c26ccecbbd0fe5f5556c80212
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/260943
Reviewed-by: Danil Chapovalov <danilchap@webrtc.org>
Commit-Queue: Tomas Gunnarsson <tommi@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#37171}
This moves the construction of StunMessage instances for
ConnectionRequest, outside of the Prepare() method.
Following this, removing Construct()+Prepare() is relatively
straight forward.
Bug: none
Change-Id: Ibcf0510cef30a6e648005b43602c7ae1fb06729e
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/264558
Reviewed-by: Niels Moller <nisse@webrtc.org>
Reviewed-by: Jonas Oreland <jonaso@webrtc.org>
Commit-Queue: Tomas Gunnarsson <tommi@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#37122}
* Switch to StunRequest where ConnectionRequest isn't needed.
* Move ConnectionRequest into Connection as a subclass
* Move declaration+implementation into the .cc file.
Bug: none
Change-Id: I1591a91d8e13a0db9f89cfbba0fbc9478a55b789
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/264150
Reviewed-by: Niels Moller <nisse@webrtc.org>
Commit-Queue: Tomas Gunnarsson <tommi@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#37036}
This is to avoid using an index into a vector that's owned by Port.
Bug: none
Change-Id: Ifc67fcc24bcb04e55c7b963de6d29bb9541c1495
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/263643
Commit-Queue: Tomas Gunnarsson <tommi@webrtc.org>
Reviewed-by: Danil Chapovalov <danilchap@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#37015}
This reverts commit 942cac2e9e6a205fd673dc003a051cfb3867f2e3.
Reason for revert: Reverting while downstream updates are made.
Original change's description:
> Make deletion of Connection objects more deterministic.
>
> This changes most deletion paths of Connection objects to go through
> the owner class of the Connection instances, Port.
>
> In situations where Connection objects still need to be deleted
> asynchronously, `async = true` can be passed to
> `Port::DestroyConnection` and get the same behavior as
> `Connection::Destroy` formerly gave.
>
> The `Destroy()` method still exists for downstream compatibility, but
> instead of deleting connection objects asynchronously, the deletion
> now happens synchronously via the Port class.
>
> Bug: webrtc:13892, webrtc:13865
> Change-Id: I07edb7bb5e5d93b33542581b4b09def548de9e12
> Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/259826
> Reviewed-by: Danil Chapovalov <danilchap@webrtc.org>
> Commit-Queue: Tomas Gunnarsson <tommi@webrtc.org>
> Cr-Commit-Position: refs/heads/main@{#36676}
Bug: webrtc:13892, webrtc:13865
Change-Id: I37a15692c8201716402ba5c10f249e4d3754ce4c
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/260862
Bot-Commit: rubber-stamper@appspot.gserviceaccount.com <rubber-stamper@appspot.gserviceaccount.com>
Commit-Queue: Tomas Gunnarsson <tommi@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#36736}
...and add to TCPConnection as it's still needed there.
Bug: webrtc:11943
Change-Id: Id66b890677d7e0b03a4a700414f574abd6e3af58
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/260320
Commit-Queue: Tomas Gunnarsson <tommi@webrtc.org>
Auto-Submit: Tomas Gunnarsson <tommi@webrtc.org>
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#36689}
This changes most deletion paths of Connection objects to go through
the owner class of the Connection instances, Port.
In situations where Connection objects still need to be deleted
asynchronously, `async = true` can be passed to
`Port::DestroyConnection` and get the same behavior as
`Connection::Destroy` formerly gave.
The `Destroy()` method still exists for downstream compatibility, but
instead of deleting connection objects asynchronously, the deletion
now happens synchronously via the Port class.
Bug: webrtc:13892, webrtc:13865
Change-Id: I07edb7bb5e5d93b33542581b4b09def548de9e12
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/259826
Reviewed-by: Danil Chapovalov <danilchap@webrtc.org>
Commit-Queue: Tomas Gunnarsson <tommi@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#36676}
`pending_delete_` was being used to protect from referencing a
potentially bad `port_` pointer. We now use a WeakPtr for the port
reference, which we clear inside of the Destroy() method. This means
we don't need both flags.
Bug: webrtc:13892
Change-Id: I9427829444486e97d30752893ba2a30b153a70e5
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/258980
Reviewed-by: Danil Chapovalov <danilchap@webrtc.org>
Commit-Queue: Tomas Gunnarsson <tommi@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#36554}
This is just to reduce confusion since StunMessage and StunRequest
instances are frequently used together and message objects are often
configured from within request objects (which makes the name confusing).
Bug: none
Change-Id: I8bf5e774a5149239dd3023817614d411633bf583
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/258484
Reviewed-by: Danil Chapovalov <danilchap@webrtc.org>
Auto-Submit: Tomas Gunnarsson <tommi@webrtc.org>
Commit-Queue: Tomas Gunnarsson <tommi@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#36533}
* Make StunRequest::manager_ a reference, inject ref at ctor time.
* Make other member variables private.
* Mark methods that are only used for testing with "ForTest"
* Add RTC_GUARDED_BY for member variables and thread checks.
* Remove/reduce 'friend'-ness between classes.
* Use std::unique_ptr for owned and passed message pointers.
* Rename `requests_` to `request_manager_` (type: StunRequestManager)
Bug: webrtc:13892
Change-Id: I3a5d511b3c2645bb6813352d39e9fefe422dd1de
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/258620
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Commit-Queue: Tomas Gunnarsson <tommi@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#36529}
This is to aid with catching issues whereby a connection object might
have a bad reference back to a port object, e.g. inside of an async
callback.
Bug: webrtc:13892
Change-Id: I56503fedc2865919713b10f236ce023554c68ded
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/257164
Reviewed-by: Niels Moller <nisse@webrtc.org>
Commit-Queue: Tomas Gunnarsson <tommi@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#36394}
Following [1], add many more checks for safe access to member variables.
This change is effectively a no-op, but landed separately from the
earlier change that's smaller but contains a fundamental assumption
gleaned from the implementation (and its use).
[1]: https://webrtc-review.googlesource.com/c/src/+/249942
Bug: webrtc:11988
Change-Id: I1568e2160c9faa6993c5b68044312f83d00e4815
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/249943
Auto-Submit: Tomas Gunnarsson <tommi@webrtc.org>
Reviewed-by: Niels Moller <nisse@webrtc.org>
Commit-Queue: Tomas Gunnarsson <tommi@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#35850}
Make sure that instances are always created+deleted on the
network thread.
Bug: webrtc:11988
Change-Id: I4fb5dd5bd14768d89ca78b348988a797fcdd130a
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/249942
Reviewed-by: Niels Moller <nisse@webrtc.org>
Auto-Submit: Tomas Gunnarsson <tommi@webrtc.org>
Commit-Queue: Tomas Gunnarsson <tommi@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#35842}
Add a timestamp for last data sent in Connection.
Move calling of rtc::TimeMillis() to Connection and remove it from RateTracker::AddSamples.
This timestamp will be used to further improve fail over logic.
BUG=None
Change-Id: I4cbc7693a0e081277590b9cb13264dc2a998202e
No-Try: True
Change-Id: I4cbc7693a0e081277590b9cb13264dc2a998202e
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/197421
Commit-Queue: Jonas Oreland <jonaso@webrtc.org>
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#32831}
The class already clears the thread that's used in its dtor
and consistently uses the same thread.
Bug: webrtc:11908
Change-Id: I5ea8d00c2e59bf46c5b369be5b23cf1d8e1875c4
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/184060
Reviewed-by: Henrik Boström <hbos@webrtc.org>
Commit-Queue: Tommi <tommi@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#32097}
As documented in webrtc:11908 this cleanup is fairly invasive and
when a part of a frequently executed code path, can be quite costly
in terms of performance overhead. This is currently the case with
synchronous calls between threads (Thread) as well with our proxy
api classes.
With this CL, all code in WebRTC should now either be using MessageHandlerAutoCleanup
or calling MessageHandler(false) explicitly.
Next steps will be to update external code to either depend on the
AutoCleanup variant, or call MessageHandler(false).
Changing the proxy classes to use TaskQueue set of concepts instead of
MessageHandler. This avoids the perf overhead related to the cleanup
above as well as incompatibility with the thread policy checks in
Thread that some current external users of the proxies would otherwise
run into (if we were to use Thread::Send() for synchronous call).
Following this we'll move the cleanup step into the AutoCleanup class
and an RTC_DCHECK that all calls to the MessageHandler are setting
the flag to false, before eventually removing the flag and make
MessageHandler pure virtual.
Bug: webrtc:11908
Change-Id: Idf4ff9bcc8438cb8c583777e282005e0bc511c8f
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/183442
Reviewed-by: Artem Titov <titovartem@webrtc.org>
Commit-Queue: Tommi <tommi@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#32049}
This patch adds a new ForgetLearnedState() method on a Connection.
The method, puts the connection into a state similar to
when it was just created.
- write_state = STATE_WRITE_INIT
- receving = false
- throw away all pending request
- reset RttEstimate
All other state is kept unchanged.
Note: It does not trigger SignalStateChange
A subsequent patch will expose the method to the IceController.
BUG: webrtc:11463
Change-Id: I055e8cd067e1bc4fd5ad64dd10f458554dbc87e3
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/171805
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Commit-Queue: Jonas Oreland <jonaso@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#30916}
This patch introduces a new type of STUN ping,
GOOG_PING_REQUEST/RESPONSE which is similar
to a STUN_BINDING but does not transmit any values.
The Connection class automatically sends these if
no STUN attributes has changed since last call to Connection::Ping()
if the remote peer has signaled that it supports it.
BUG=webrtc:11100
Change-Id: Ib1b590f0b90ca6cb56f2eb07cd62f976e246bc8c
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/159961
Commit-Queue: Jonas Oreland <jonaso@webrtc.org>
Reviewed-by: Taylor <deadbeef@webrtc.org>
Reviewed-by: Björn Terelius <terelius@webrtc.org>
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#30062}
This patch moves the SendBindingResponse from Port
to Connection. This is a behavioural NOP, and I don't
understand why it was in Port in the firs place!
Found when working on GOOG_PING.
BUG=webrtc:11100
Change-Id: I0466c5381f08ec4926ca3380e6914f0bc0dfcf63
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/161081
Commit-Queue: Jonas Oreland <jonaso@webrtc.org>
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29963}
This patch estimates the connection RTT using
EventBasedExponentialMovingAverage. The half time is
set to 500 but can be modified using field trials.
This new metric is currently unused, but will
be used for exploration of of whether it can be used
instead of the existing metric.
Bug: webrtc:11140
Change-Id: I9db93e9b9eb932e3cd18935cd4ce0d90fc1cb293
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/161000
Commit-Queue: Jonas Oreland <jonaso@webrtc.org>
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29944}
This patch makes Connection::port() protected
and add explicit methods for the use cases instead
- network() - port()->Network()
- generation() - port()->generation()
This is done to easier mock a Connection.
BUG=webrtc:10647
Change-Id: I5b35477ed9f81d57cd871072874262d0a8af2d4c
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/160784
Commit-Queue: Jonas Oreland <jonaso@webrtc.org>
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29929}
We now have two downstream users of stun.h, so it appears to be
generally usable. I put this in a new dir networking/, but I'm open to
suggestions here (maybe some things in api/ should move in there).
I checked what our downstream users are actually using, and it's
cricket::ComputeStunCredentialHash
cricket::<constants>
cricket::TurnMessage
cricket::GetStunErrorResponseType
cricket::StunAttribute::CreateAddress
cricket::StunErrorCodeAttribute
cricket::StunByteStringAttribute
StunAttribute::CreateUnknownAttributes
cricket::TurnErrorType
cricket::StunMessage
I reckoned that was pretty much everything in stun.h, so I didn't
bother splitting it up. They don't use every function and constant
in there, but all _types_ of functions and constants, so for the
sake of coherence I don't think it makes sense to split it.
There's some old stuff in there like GTURN which could arguably
be split out, but it should likely go away soon anyway, so I don't
think it's worth the effort.
Steps:
1) land this
2) update downstream to point to the new header and target
3) remove p2p/base:stun_types.
Bug: webrtc:11091
Change-Id: I1f05bf06055475d25601197ec6fefb8d3b55e8e3
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/159923
Commit-Queue: Patrik Höglund <phoglund@webrtc.org>
Reviewed-by: Niels Moller <nisse@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29822}
This patch adds field trials for limiting no of STUN pings
- max_outstanding_pings
send this count of outstanding pings (pings w/o any response),
after that never send any before a reply is received.
NOTE:
1) This patch redoes https://webrtc.googlesource.com/src.git/+/0d28972d8f0659ab90cef7fd59ca54fb122b71bc
which was put into the StunRequestManager.
But that mechanism is not used for STUN pings.
2) This patch build on field-trial-parser added in https://webrtc-review.googlesource.com/c/src/+/156083
Bug: webrtc:10282
Change-Id: If2f22d2b61a28598a3aa93781c9857145576b7a1
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/156162
Commit-Queue: Jonas Oreland <jonaso@webrtc.org>
Reviewed-by: Honghai Zhang <honghaiz@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29467}
Per the latest WebRTC stats spec
(https://w3c.github.io/webrtc-stats/#dom-rtcicecandidatestats)
the address field of a peer-reflexive remote candidate should be concealed
until the same address is learnt via addIceCandidate.
This CL also refactors the sanitization-related code paths.
Bug: chromium:968161
Change-Id: I74c5da78232b2f604689867bda2937b8af827c4f
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/149381
Reviewed-by: Steve Anton <steveanton@webrtc.org>
Commit-Queue: Qingsi Wang <qingsi@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#28909}
The public API to obtain the selected candidate pair is changed to
GetSelectedCandidatePair in the ICE transport, and the returned pair
has address-sanitized candidates.
Bug: chromium:993878
Change-Id: I44f9d2385a84f9e22447108be2e57ef9e62671eb
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/149080
Reviewed-by: Steve Anton <steveanton@webrtc.org>
Commit-Queue: Qingsi Wang <qingsi@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#28869}
outgoing checks.
This change adds an experimental feature to allow an ICE agent to embed
the transaction ID of the latest connectivity check received from the
remote peer, as an auxiliary acknowledgement in additional to the check
response, in its own checks. This could facilitate the establishment of
ICE connectivity if the check process has a high RTT.
Bug: None
Change-Id: If3e6327720f13beeb14f103af3b5ffb4f9692998
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/142682
Reviewed-by: Honghai Zhang <honghaiz@webrtc.org>
Commit-Queue: Qingsi Wang <qingsi@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#28316}
This patch is a NOP and moves
- class Connection
- class ConnectionInfo
- class ProxyConnection
from port.{h/cc} to a new file called connection.{h/cc}
BUG=webrtc:10647
Change-Id: I89322d3421d272657e24a46b28ab6679fcdc9450
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/137509
Commit-Queue: Jonas Oreland <jonaso@webrtc.org>
Reviewed-by: Qingsi Wang <qingsi@webrtc.org>
Reviewed-by: Niels Moller <nisse@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#28101}
