This change exempts the emulator connections from
WaitableConnection.WaitForConnection(). This is because emulator
connections are a) more reliable and b) handled a bit differently than
normal TCP connections.
Bug: 78991667
Test: emulator -showkernel ; adb shell
Change-Id: I552946198889a82d6c265f45e8c3b38f6ac9d045
This change is in preparation to allow the TCP-based transports to be
able to reconnect. This is needed because multiple threads can access
the Connection object. It used to be safe to do because one instance of
atransport would have the same Connection instance throughout its
lifetime, but now it is possible to replace the Connection instance,
which could cause threads that were attempting to Write to an
atransport* to use-after-free the Connection instance.
Bug: 74411879
Test: system/core/adb/test_adb.py
Change-Id: I4f092be11b2095088a9a9de2c0386086814d37ce
This change adds a callback that is invoked exactly once, either when
the connection is fully established (i.e. CNXN packets have been sent
and received) or the atransport object is deleted before that (because
the connection failed).
This helps in distinguishing between successful and failing connections
for TCP. Especially when there is some kind of port
forwarding/multiplexing in between (like an SSH tunnel or SSLH proxy).
Bug: 74411879
Test: adb connect chromebook:22 (which runs an sslh tunnel to adbd).
either succeeds or fails, but not fake-succeeds.
Change-Id: I7e826c6f5d4c30338a03b2d376a857ac5d05672a
Host services are attempted after handle_host_request, which means that
failing to find a transport to give to handle_forward_request shouldn't
send an error over to the other end.
Bug: http://b/78294734
Test: `adb track-devices` with multiple devices connected
Change-Id: I46c89cc1894b51d48fea7d4e629b1d57f73e3fd6
This change uses a context manager to create the fake ADB servers (and
cleanly tear them down.
Bug: 74411879
Test: python system/core/adb/test_adb.py
Change-Id: I722d2c4135259b1b0ef00a1510aa8402e87ecf72
This change returns a different value (-EALREADY) when a connection has
already been established, as opposed to a real connection failure (which
still returns -1).
Bug: 74411879
Test: Opened a socket, tried to adb connect to it,
got "failed to connect to localhost:1337"
Change-Id: Ic216ddef7f28eb43ca750f9e51d068c077d54b07
Plumb the transport that we received the adb reverse request on through
to reverse_service, instead of trying to get a unique transport on
devices that have multiple active transports (e.g. a device with USB
(even unplugged) connected via TCP).
Bug: http://b/37066218
Bug: http://b/71898863
Test: `echo foo | nc -l 12345 & adb reverse tcp:12345 tcp:12345; adb shell nc localhost 12345` on a device connected via TCP
Change-Id: Iae199ae787f2e344126bbcacca8544cfc9844a4c
clang-format some functions that I'm about to touch, but do it in a
separate commit to avoid making the actual changes unreadable.
Test: treehugger
Change-Id: I7dc5258d9ecc1c091d42311149d9e18ae483715b
Switch from using std::string as the type we use to hold our payload in
apacket to a custom reimplementation that doesn't zero initialize. This
improves bulk transfer throughput in the adb_benchmark microbenchmark
on walleye by ~20%.
Test: adb shell taskset f0 /data/benchmarktest64/adb_benchmark/adb_benchmark
Change-Id: Ibad797701eb1460c9321b0400c5b167b89b2b4d0
OS X reports maxPacketSize as wMaxPacketSize * (bMaxBurst + 1) in the
deprecated GetPipeProperties function. Use the also-deprecated
GetPipePropetiesV2 API to get bMaxBurst and figure out what
wMaxPacketSize is. (This file is going to go away eventually, so don't
bother with switching to the recommended GetPipePropertiesV3, since
it would be a substantially larger charge.)
libusb is unaffected.
Bug: http://b/77733422
Test: python test_device.py
Change-Id: I66517d699a4f39b93ba5eb7882bd8ee6c70f3672
This change splits the capability-dropping step of adbd into two. This
is more robust when ambient capabilities are being used, since minijail
cannot currently handle that case.
Bug: 77146512
Test: grep Cap /proc/`pidof adbd`/status
CapInh: 0000000000000000
CapPrm: 0000000000000000
CapEff: 0000000000000000
CapBnd: 00000000000000c0
CapAmb: 0000000000000000
Change-Id: I0476a8d80f7a2497600196932542045f3bc87537
Profiling of adb_benchmark revealed that something like half of the CPU
time was being spent in malloc and free, which was odd because the
benchmark repeatedly mallocs and frees allocations of the same size. It
turns out that our default configuration of jemalloc will purge after
every free. Configure jemalloc to not do this, for gains of over 100%
on adb_benchmark, and up to 25% (on walleye USB3) in real-life.
Test: adb_benchmark
Change-Id: I602dd1645c3d21709c7f6a78903511ce4d576558
This reverts commit 2ef56a1174.
host_supported is broken in cc_benchmark on Mac, but we probably don't
actually too much about the host benchmarks, so just delete the line.
Bug: http://b/77585931
Test: treehugger
Change-Id: I8898b4037a3faf8e8dee2a99498a6f84e642b92b
When connecting to an address, we construct a transport first, and then
check whether we've already connected to that address. The consequent
destruction of the BlockingConnectionAdapter attempts to join threads
that haven't been started, which aborts.
Make it safe to destruct a BlockingConnectionAdapter without calling
Start on it first, to solve this.
Bug: http://b/69137547
Test: nc -l 12345 & (adb connect localhost:12345; adb connect localhost:12345)
Test: python test_adb.py
Change-Id: I6cb968a62dbac6332907e06575893d764905ee62
On platforms that implement sockets via underlying sockets (e.g. Wine),
a socket can appear to be full, and then become available for writes
again without read being called on the other end. Add a sleep after
each write to give the underlying implementation time to flush.
This doesn't help us if the buffer size is smaller than MAX_PAYLOAD,
but at least in the case of Wine, that doesn't seem to be the case.
Test: adb_test
Test: wine adb_test.exe
Change-Id: Iff1ec14bbf318b9742ec7e2fb72e34e3d6bbe6ad
Previously, each of the tests was spawning the fdevent thread manually,
in order to be able to set up listeners and such before running
fdevent_loop. Now that we have a way to run arbitrary code on the
fdevent thread, switch to having a generic fdevent thread and running
setup code via fdevent_run_on_main_thread.
Test: adb_test
Test: wine adb_test.exe
Change-Id: I517dbcbad31067b45087d9fbed67a75b75a75aec
Make it so that we handle run_on_main_thread calls after regular socket
events, so that we can use it as a way to ensure we've processed all
pending socket events.
Test: adb_test
Test: wine adb_test.exe
Change-Id: Ic215c7fed19a8e1699e759970658b3775aa08c45
Bail out immediately in our poll emulation if we have any invalid file
descriptors, instead of actually calling WSAPoll (which might block
forever).
Test: wine adb_test.exe
Change-Id: I06ccca305783ee7a1721b1585ddf73e022bd02d7
Make it easier to replace the sleep with something that actually waits
for the event loop to quiesce.
Test: adb_test
Change-Id: Iee86b058abbb607256dbfce43d8c70252b6bb0f6
Previously, processes started via `adb shell` have an oom_score_adj of
-1000, making them invisible to the oom killer. This makes running a
process that consumes all memory (e.g. by leaking in a loop) lead to
the entire rest of the system (including adbd, because of bad
heuristics in the kernel) being oom killed before getting to it.
Bug: http://b/63143027
Test: `adb shell cat /proc/self/oom_score_adj` with adb root
Change-Id: I59111134e36dc271adf4c1dd4bd4400d4fe6aee0
The switch over to soong unintentionally turned off building the
Windows tests (and there's a bug preventing them from being turned on).
Preemptively fix the build breakages that'll happen when we turn them
back on.
Test: mma with Android.bp 's/cc_test_host/cc_binary_host' hacks
Change-Id: I6582cfc61b96052537d50d8ba90533dbb66e3e1d
Match openssh's behavior with regards to login shells.
Bug: http://b/76116378
Test: `adb shell -T set -o` vs `ssh localhost -T set -o`
Test: `adb shell -tt set -o` vs `ssh localhost -tt set -o`
Test: `echo set -o | adb shell -T` vs `echo set -o | ssh localhost -T`
Test: `echo set -o | adb shell -tt` vs `echo set -o | ssh localhost -tt`
Change-Id: I6ed69b43e2ac79c947c242948b5617ece08a713e
The listen backlog seems to be more meaningful on Darwin than on Linux,
resulting in connections failing with ECONNRESET. Bump it up to the
maximum supported value to make this less likely. 128 pending
connections ought to be enough for anybody.
Bug: http://b/74616284
Test: python test_device.py
Change-Id: I5fe0205924188cf18ca1fc1204f923ab5523eeb2
