Relanding because the broken chromium test has been fixed:
https://chromium-review.googlesource.com/582196
This CL moves the responsibility for restricting the number of IPv6
interfaces used for ICE to BasicPortAllocator. This is the right place
to do it in the first place; it's where all the rest of the filtering
occurs. And NetworkManager shouldn't need to know about ICE limitations;
only the ICE classes should.
Part of the reason I'm doing this is that I want to add a
"max_ipv6_networks" API to RTCConfiguration, so that applications can
override the default easily (see linked bug). But that means that
PeerConnection would need to be able to call "set_max_ipv6_networks" on
the underlying object that does the filtering, and that method isn't
available on the "NetworkManager" base class. So rather than adding
another method to a place it doesn't belong, I'm moving it to the place
it does belong.
In the process, I noticed that "CompareNetworks" is inconsistent with
"SortNetworks"; the former orders interfaces alphabetically, and the
latter reverse-alphabetically. I believe this was unintentional, and
results in undesirable behavior (like "eth1" being preferred over
"eth0"), so I'm fixing it and adding a test.
BUG=webrtc:7703
Review-Url: https://codereview.webrtc.org/2983213002
Cr-Original-Commit-Position: refs/heads/master@{#19112}
Committed: ad9561404c
Review-Url: https://codereview.webrtc.org/2983213002
Cr-Commit-Position: refs/heads/master@{#19159}
Reason for revert:
Breaks IpcNetworkManagerTest.TestMergeNetworkList, because it has built-in assumptions about network ordering that it shouldn't have. Will reland after fixing that test.
Original issue's description:
> Move "max IPv6 networks" logic to BasicPortAllocator, and fix sorting.
>
> This CL moves the responsibility for restricting the number of IPv6
> interfaces used for ICE to BasicPortAllocator. This is the right place
> to do it in the first place; it's where all the rest of the filtering
> occurs. And NetworkManager shouldn't need to know about ICE limitations;
> only the ICE classes should.
>
> Part of the reason I'm doing this is that I want to add a
> "max_ipv6_networks" API to RTCConfiguration, so that applications can
> override the default easily (see linked bug). But that means that
> PeerConnection would need to be able to call "set_max_ipv6_networks" on
> the underlying object that does the filtering, and that method isn't
> available on the "NetworkManager" base class. So rather than adding
> another method to a place it doesn't belong, I'm moving it to the place
> it does belong.
>
> In the process, I noticed that "CompareNetworks" is inconsistent with
> "SortNetworks"; the former orders interfaces alphabetically, and the
> latter reverse-alphabetically. I believe this was unintentional, and
> results in undesirable behavior (like "eth1" being preferred over
> "eth0"), so I'm fixing it and adding a test.
>
> BUG=webrtc:7703
>
> Review-Url: https://codereview.webrtc.org/2983213002
> Cr-Commit-Position: refs/heads/master@{#19112}
> Committed: ad9561404cTBR=zhihuang@webrtc.org,pthatcher@webrtc.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=webrtc:7703
Review-Url: https://codereview.webrtc.org/2984853002
Cr-Commit-Position: refs/heads/master@{#19114}
This CL moves the responsibility for restricting the number of IPv6
interfaces used for ICE to BasicPortAllocator. This is the right place
to do it in the first place; it's where all the rest of the filtering
occurs. And NetworkManager shouldn't need to know about ICE limitations;
only the ICE classes should.
Part of the reason I'm doing this is that I want to add a
"max_ipv6_networks" API to RTCConfiguration, so that applications can
override the default easily (see linked bug). But that means that
PeerConnection would need to be able to call "set_max_ipv6_networks" on
the underlying object that does the filtering, and that method isn't
available on the "NetworkManager" base class. So rather than adding
another method to a place it doesn't belong, I'm moving it to the place
it does belong.
In the process, I noticed that "CompareNetworks" is inconsistent with
"SortNetworks"; the former orders interfaces alphabetically, and the
latter reverse-alphabetically. I believe this was unintentional, and
results in undesirable behavior (like "eth1" being preferred over
"eth0"), so I'm fixing it and adding a test.
BUG=webrtc:7703
Review-Url: https://codereview.webrtc.org/2983213002
Cr-Commit-Position: refs/heads/master@{#19112}
I used a command like this to update the paths:
perl -pi -e "s/webrtc\/base/webrtc\/rtc_base/g" `find webrtc/rtc_base -name "*.cc" -o -name "*.h"`
BUG=webrtc:7634
NOPRESUBMIT=True # cpplint errors that aren't caused by this CL.
Review-Url: https://codereview.webrtc.org/2969623003
Cr-Commit-Position: refs/heads/master@{#18870}
The only thing the physical socket server was used for was
"Wait"/"WakeUp", but it could be replaced by a simple rtc::Event.
So, removing this dependency makes things less confusing; the fact that
VirtualSocketServer takes a PhysicalSocketServer may lead someone to
think it uses real sockets internally, when it doesn't.
BUG=None
Review-Url: https://codereview.webrtc.org/2883313003
Cr-Commit-Position: refs/heads/master@{#18172}
For every failing test, the solution was either to do a "has IPv4" check
before the test is run, or avoid depending on real network interfaces
altogether.
This specifically fixes rtc_unittests, peerconnection_unittests, and
webrtc_nonparallel_tests.
BUG=None
Review-Url: https://codereview.webrtc.org/2881973002
Cr-Commit-Position: refs/heads/master@{#18155}
Instead, make the pointer to the associated socket server a
construction time const, and delete its lock.
Introduces a helper class AutoSocketServerThread for code
(mainly tests) which need a socket server associated with
the current thread.
BUG=webrtc:7501
Review-Url: https://codereview.webrtc.org/2828223002
Cr-Commit-Position: refs/heads/master@{#18047}
Add an attribute to the RTCConfiguration which can be used by specific
mobile devices so that the IPv6 ICE candidates on WiFi will not be collected.
BUG=b/35725283
Review-Url: https://codereview.webrtc.org/2731813002
Cr-Commit-Position: refs/heads/master@{#17100}
This isn't as good as being able to enumerate all networks, but it's better
than doing nothing; it still will provide STUN/TURN candidates for the default
route if one exists.
BUG=webrtc:6932
Review-Url: https://codereview.webrtc.org/2599673003
Cr-Commit-Position: refs/heads/master@{#15766}
Previously when a Turn port is pruned, if its candidate has been sent to the remote side, the remote side will keep the candidate and use that to create connections.
We now signal the remote side to remove the candidates so that at least no new connection will be created using the removed candidates.
Also updated the virtual socket server to better support our test cases.
1. Allow the virtual socket server to set transit delay for packets sent from a given IP address.
2. Ensure the ordered packet delivery for each socket (Previously the delivery order is enforced on the whole test case, so if a udp packet gets delayed based on its IP address, all TCP packets sent after the UDP packet will be delayed at least until the UDP packet is received).
BUG=webrtc:6380
R=deadbeef@webrtc.org, pthatcher@webrtc.org, skvlad@webrtc.org
Review URL: https://codereview.webrtc.org/2261523004 .
Cr-Commit-Position: refs/heads/master@{#14297}
I found that, depending on when it's called, ClearGettingPorts may or
may not signal CandidatesAllocationDone, and may or may not continue
to gather more ports/candidates.
I'm fixing this inconsistency by having it always signal
CandidatesAllocationDone (if needed), and always stop gathering until
the next network change event. This makes it equivalent to
StopGettingPorts, except that it allows gathering to be restarted if
a network change occurs.
I also found that P2PTransportChannel was signaling "gathering
complete" even when continual gathering was enabled. This wasn't caught
by the unit tests due to the inconsistency of ClearGettingPorts as
described above.
Review-Url: https://codereview.webrtc.org/2124283003
Cr-Commit-Position: refs/heads/master@{#13908}
It was possible that "A < B" and "B < A" both evaluated to true.
This manifested as an assert on Windows, and a memory leak on Linux.
Note that the concept of "less than" is meaningless for this object.
The operator is only needed so the object can be used as a key in an
std::map.
BUG=webrtc:6068
R=honghaiz@webrtc.org, kjellander@webrtc.org, skvlad@webrtc.org
Review URL: https://codereview.webrtc.org/2187913002 .
Cr-Commit-Position: refs/heads/master@{#13598}
If a port is not used by any channel and if it has no connection for 30
seconds, it will be removed.
Note, as long as a port is used by a transport channel, it will be kept
even if it does not have any connection. This will be beneficial to
continual gathering because new connections can be created in the future
when network changes.
BUG=
R=pthatcher@webrtc.org, zhihuang@webrtc.org
Review URL: https://codereview.webrtc.org/2171183002 .
Cr-Commit-Position: refs/heads/master@{#13567}
When the flag prune_turn_ports is set, When a high-priority turn port becomes available, it will prune low-priority ones. The pruned port will not be used for creating connections locally and its candidates will not be sent over to the remove side (unless they have been sent before being pruned).
This effectively reduces the number of TURN candidates and connections created by TURN ports.
BUG=
R=deadbeef@webrtc.org, pthatcher@webrtc.org
Review URL: https://codereview.webrtc.org/2093623004 .
Committed: https://crrev.com/17aac053f585e892114974d2eb248e05ad37f973
Cr-Original-Commit-Position: refs/heads/master@{#13335}
Cr-Commit-Position: refs/heads/master@{#13354}
Reason for revert:
Breaks Win32/Win64 Debug bots in client.webrtc waterfall
Original issue's description:
> Add config to prune low-priority TURN ports for creating connections
> When the flag prune_turn_ports is set, When a high-priority turn port becomes available, it will prune low-priority ones. The pruned port will not be used for creating connections locally and its candidates will not be sent over to the remove side (unless they have been sent before being pruned).
>
> This effectively reduces the number of TURN candidates and connections created by TURN ports.
>
> BUG=
> R=deadbeef@webrtc.org, pthatcher@webrtc.org
>
> Committed: https://crrev.com/17aac053f585e892114974d2eb248e05ad37f973
> Cr-Commit-Position: refs/heads/master@{#13335}
TBR=pthatcher@webrtc.org,deadbeef@webrtc.org,honghaiz@webrtc.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=
Review-Url: https://codereview.webrtc.org/2111663003
Cr-Commit-Position: refs/heads/master@{#13342}
When the flag prune_turn_ports is set, When a high-priority turn port becomes available, it will prune low-priority ones. The pruned port will not be used for creating connections locally and its candidates will not be sent over to the remove side (unless they have been sent before being pruned).
This effectively reduces the number of TURN candidates and connections created by TURN ports.
BUG=
R=deadbeef@webrtc.org, pthatcher@webrtc.org
Review URL: https://codereview.webrtc.org/2093623004 .
Cr-Commit-Position: refs/heads/master@{#13335}
This allows webrtc to not gather on cellular networks if wifi or
other low cost networks are present.
BUG=
Review-Url: https://codereview.webrtc.org/1987833002
Cr-Commit-Position: refs/heads/master@{#12979}
The test doesn't really care about the order; the fact that it relied
on a specific order was just an implementation detail.
However, this made the test flaky since race conditions sometimes
determine the order. It also made it frustrating to add new tests.
BUG=webrtc:5930
R=pthatcher@webrtc.org
Review URL: https://codereview.webrtc.org/2006263004 .
Cr-Commit-Position: refs/heads/master@{#12936}
This is the minumum allowed size, and will allow STUN pings to be smaller.
The unit tests on the the Gturn are also modified. A username with length of 16 bytes will be generated for Gturn only.
Review-Url: https://codereview.webrtc.org/1848083002
Cr-Commit-Position: refs/heads/master@{#12876}
According to JSEP, the candidate filter does not affect pooled
candidates because they can be filtered once they're ready to be
surfaced to the application.
So, pooled port allocator sessions will use a filter of CF_ALL, with a
new filter applied when the session is taken by a P2PTransportChannel.
When the filter is applied:
* Some candidates may no longer be returned by ReadyCandidates()
* Some candidates may no longer have a "related address" (for privacy)
* Some ports may no longer be returned by ReadyPorts()
To simplify this, the candidate filtering logic is now moved up from
the Ports to the BasicPortAllocator, with some helper methods to perform
the filtering and stripping out of data.
R=honghaiz@webrtc.org, pthatcher@webrtc.org
Review URL: https://codereview.webrtc.org/1998813002 .
Cr-Commit-Position: refs/heads/master@{#12856}
The STUN timeout is 9500ms, and the tests are waiting for 10000ms.
The 500ms margin of error is not enough for some bots (such as UBSan).
R=pthatcher@webrtc.org
Review URL: https://codereview.webrtc.org/2001093003 .
Cr-Commit-Position: refs/heads/master@{#12854}
Depends on this CL in order to work in Chromium:
https://codereview.chromium.org/1976673002/
It works by creating pooled PortAllocatorSessions which can be picked up
by a P2PTransportChannel when needed (after a local description is set).
This can optimize candidate gathering time when there is some time between
creating a PeerConnection and setting a local description.
R=pthatcher@webrtc.org
Committed: 48e9d05f51
Review URL: https://codereview.webrtc.org/1956453003 .
Cr-Commit-Position: refs/heads/master@{#12729}
Reason for revert:
Breaks remoting_unittests. They defined their own operator== which conflicts with this one.
I'll remove the operator== in a roll CL. But until it's approved, I'm reverting this so the FYI bots will pass.
Original issue's description:
> Implement RTCConfiguration.iceCandidatePoolSize.
>
> It works by creating pooled PortAllocatorSessions which can be picked up
> by a P2PTransportChannel when needed (after a local description is set).
>
> This can optimize candidate gathering time when there is some time between
> creating a PeerConnection and setting a local description.
>
> R=pthatcher@webrtc.org
>
> Committed: 48e9d05f51TBR=pthatcher@webrtc.org,honghaiz@webrtc.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Review-Url: https://codereview.webrtc.org/1972043004
Cr-Commit-Position: refs/heads/master@{#12709}
It works by creating pooled PortAllocatorSessions which can be picked up
by a P2PTransportChannel when needed (after a local description is set).
This can optimize candidate gathering time when there is some time between
creating a PeerConnection and setting a local description.
R=pthatcher@webrtc.org
Review URL: https://codereview.webrtc.org/1956453003 .
Cr-Commit-Position: refs/heads/master@{#12708}
