vsock is a socket address family for communicating into and out of
virtual machines. Addresses have a port and CID. The CID is unique to
each virtual machine on the computer. The VM host always has CID 2.
http://man7.org/linux/man-pages/man7/vsock.7.html
Inside the android guest, the adb daemon hosts a vsock server with
VMADDR_CID_ANY, automatically using the guest CID. The adb server
can now connect to addresses of the form vsock:cid:port, where the CID
must be specified and the port defaults to 5555.
This is a significant speed improvement for ADB connections in
Cuttlefish, with 150-200 MB/s for `adb push` and 100-150 MB/s for
`adb pull`. It also allows removing some proxying steps from Cuttlefish,
simplifying the full connection path, and removes a dependency on the
unstable ivshmem protocol.
Commands tested against a Cuttlefish VM with CID 3:
adb connect vsock:3:5555
adb -s vsock:3:5555 shell
adb disconnect vsock:3:5555
Supporting "adb disconnect" and "adb -s" required modifying some of the
parts that parse addresses / serials.
push/pull trials with native adb vsock support in cuttlefish:
100m: 1 file pushed. 297.6 MB/s (104857600 bytes in 0.336s)
100m: 1 file pushed. 270.3 MB/s (104857600 bytes in 0.370s)
100m: 1 file pushed. 271.7 MB/s (104857600 bytes in 0.368s)
100m: 1 file pushed. 250.5 MB/s (104857600 bytes in 0.399s)
100m: 1 file pushed. 277.1 MB/s (104857600 bytes in 0.361s)
100m: 1 file pushed. 263.5 MB/s (104857600 bytes in 0.379s)
100m: 1 file pushed. 242.6 MB/s (104857600 bytes in 0.412s)
100m: 1 file pushed. 271.8 MB/s (104857600 bytes in 0.368s)
100m: 1 file pushed. 267.1 MB/s (104857600 bytes in 0.374s)
/data/local/tmp/100m: 1 file pulled. 212.8 MB/s (104857600 bytes in 0.470s)
/data/local/tmp/100m: 1 file pulled. 236.7 MB/s (104857600 bytes in 0.423s)
/data/local/tmp/100m: 1 file pulled. 201.2 MB/s (104857600 bytes in 0.497s)
/data/local/tmp/100m: 1 file pulled. 255.6 MB/s (104857600 bytes in 0.391s)
/data/local/tmp/100m: 1 file pulled. 199.6 MB/s (104857600 bytes in 0.501s)
/data/local/tmp/100m: 1 file pulled. 214.6 MB/s (104857600 bytes in 0.466s)
/data/local/tmp/100m: 1 file pulled. 254.2 MB/s (104857600 bytes in 0.393s)
/data/local/tmp/100m: 1 file pulled. 212.5 MB/s (104857600 bytes in 0.471s)
/data/local/tmp/100m: 1 file pulled. 218.9 MB/s (104857600 bytes in 0.457s)
/data/local/tmp/100m: 1 file pulled. 223.6 MB/s (104857600 bytes in 0.447s)
Bug: 121166534
Change-Id: I50f21fb5c9acafb8daa789df4e28c9e1bbbbf2ef
Test: adb connect/shell/disconnect
Preparatory step for testing adb on GCE on non-linux hosts: instead of
pointing them at a device (emulated or otherwise), point them at adbd
running on a linux host instead.
Test: adbd & adb connect localhost:5555; adb -e wait-for-device shell
Change-Id: Ib22d51a4fc9e6e68f71bf1b3b9b2e1b0bd844760
adb --fastdeploy creates a TemporaryFile to which the stdout for a shell
command is redirected. This happens while the file is still open (the
TemporaryFile object holds the file handle). In Linux this works, but
Windows blocks the operation with an error of "The process cannot access
the file because it is being used by another process".
This change closes the file before the shell tries to write to it.
Test: mm -j 72
Test: adb install --fastdeploy --force-agent --local-agent /mnt/raid/boat-attack-apk/boat-attack-swappy.apk
Bug: 122592986
Change-Id: Iaaaf62cda43e4714d7f979e6a690549b383a7b82
This will make it easier to add new types of transports by combining the
logic the adb server and adb daemon use to connect to things.
Bug:121166534
Test: adb connect against a cuttlefish instance over the shm proxy.
Change-Id: Ic7fc848c60a85eef968c3735838c87cb7fdaf38b
platform-tools/deployagent.jar is built incorrectly for userdebug sdk
builds (which the build machines in android-build use). In eng builds
dexpreopt is disable by default, so the file is built correctly.
This change disables dexpreopt for deployagent in all builds, for
consistency.
Bug: 122831369
Test: lunch sdk-eng
Test: make sdk
Change-Id: I370f0536d906690b153f114ba3dd589a14d222d4
This CL introduces two changes to adb install-multi-package:
- If there is at least one apex package in the list of packages, use the
--staged parameter for both the parent and the child sessions
- When the package being sent is an apex, use the --apex parameter
Bug: 118865310
Test: Printed out the resulting commands and verified that both
non-staged and staged workflow are accepted by PackageManager. Tried
scheduling install sessions for a mix of APK/APEX, only APKs, only APEX.
Change-Id: I8d1a6a7c5408fb95c10d79e38ddaf115a46f5d8b
This change adds an install-atomic command to adb that is shorthand for
creating an atomic install session and an individual session for each
APK supplied to the command.
Bug: 109941548
Test: run command with multiple APKs, observe atomic install
Change-Id: I2817a1ed2d312925d9c7bd621e6c82670a6275fd
In setup_mdns_thread mdns_registered should only be set to true
if DNSServiceRegister was successful.
Test: N/A
Change-Id: I8a2798cced91b05b7d8bf5f0df4d16719c581dee
Signed-off-by: Gustav Svensson <gustav.svensson@volvo.corp-partner.google.com>
Once launched, abb will listen for incoming Binder cli requests.
Executing in-process provides 6x latency improvement (125ms vs 25ms on
PixelXL) for commands like 'package path'
Intended usage by Android Studio for fast deployment and patching of APKs.
Test: manual
BUG: 111621042
Change-Id: Ica84eb2ec9628efa441ecd627b119f3361feaf9f
Rename it to something more appropriate, while we're at it.
Test: test_adb.py
Test: test_device.py
Test: $ANDROID_HOST_OUT/nativetest64/adb_test/adb_test
Test: adb shell /data/nativetest64/adbd_test/adbd_test
Change-Id: I0f95d348eeacb45a810696d748c8340d2068f666
Before overlayfs, we supported deduplicated filesystems by undoing
deduplication in recovery. This required an extra reboot cycle, so we
changed "adb remount" to disable verity and boot to recovery in one
command.
After overlayfs, adb remount is still trying to undo deduplication,
which leads to very confusing messages. This patch makes things a bit
clearer. "adb remount" will disable verity, which installs overlayfs.
"adb remount -R" will do the same except automatically reboot.
Bug: N/A
Test: adb remount on dynamic partitions device
Change-Id: Id72f6b9e2297c2f4d5722d5679f6264fe660e631
--fastdeploy does not require -r anymore, and reverts to a normal install
if the application is not already on the device.
Bug: 120828611
Test: mm -j72
Test: adb install --fastdeploy --force-agent --local-agent /mnt/raid/boat-attack-apk/boat-attack-swappy.apk
Change-Id: Ice2a71493a34ee7d0debabcce6a9aebb0af79e62
Add some services that skip the service fd to see how much of a benefit
it'll be to eliminate it.
Test: adb raw source:$((300 * 1024 * 1024)) | pv > /dev/null
Test: dd if=/dev/zero bs=1M count=100 | pv | adb raw sink:$((100 * 1024 * 1024))
Change-Id: I042f25f85b16ae9869cb1f1e306d8671b024ed97
We were dynamic_casting to UsbConnection to check for USB connections,
but the actual type was a BlockingConnectionAdapter wrapping a
UsbConnection, with the result that unplugging an inaccessible (due to
permissions) device on Linux wouldn't make the device go away.
Test: manual
Change-Id: Icb4acea5fd3c3baa9691698686213e122e898e4a
Certain error conditions were getting lost because adb was wrongly
reading DeployAgent's return code as always 0.
Test: mm -j 72
Test: adb install -r --fastdeploy --force-agent --local-agent /mnt/raid/boat-attack-apk/boat-attack-swappy.apk
Bug: 120197330
Change-Id: If835fd6ca2051be8e5ff6c957e08b1e458053989
If we have multiple keys available for authentication (ADB_VENDOR_KEYS
+ the one in ~/.android), we will still have keys in our list of
avilable keys after we've successfully connected. A subsequent
reconnection will start authorizing using the list of keys after the
key that actually worked, resulting in that session being unauthorized
until another reconnection happens. Clear the key list before
reconnecting to fix this. (We could do this after successfully
connecting, but we need to do this before reconnecting anyway, because
our connection could have died during authorization.)
Bug: http://b/117267347
Test: `adb connect foo; adb -s foo reconnect device` with ADB_VENDOR_KEYS
Change-Id: Ieb7dcc28e333c89ae0d75f97e89bcd1b571cb299
This reverts commit 1272e7a592.
Reason for revert: All notice fix changes have been merged and running peacefully for a sustained period. Now this can be reverted for good.
Change-Id: I74e4e291a504fcb9a14ae50e916e494c7359db34
The --fastdeploy switch caused errors when CRC collisions were present in the input apk and/or
an apk with a similar package name to the input apk was already installed on the device.
Test: mm -j 64
Test: adb install -r --fastdeploy --force-agent --local-agent /mnt/raid/boat-attack-apk/boat-attack-swappy.apk
Bug: 119934862
Change-Id: Ibfe0cec38bdbb7371803fc2f73b0ec1697cef624
Removes a few more Android.mk files.
Test: check for adb, mkbootfs, and fastboot in the build artifacts
Change-Id: Ie4e50a363a734d0b9207f0d0098c54719f038e76
An adbkey/adbkey.pub pair that doesn't match up results in a
hard-to-diagnose scenario where "Always allow from this computer"
doesn't work. The private key contains all of the information that's
in the public key, so just extract the public key out of the private
key instead of storing them separately.
Bug: http://b/119634232
Test: rm ~/.android/adbkey.pub; adb kill-server; adb shell true
Test: rm ~/.android/adbkey*; adb kill-server; adb shell true
Change-Id: I0ae9033dbcd119c12cfb2b3977f1f1954ac800c1
Large opportunistic writes would perform a write without updating
writable_ or waking up the polling thread, which resulted in the worker
thread never polling with POLLOUT.
Test: adb_benchmark
Change-Id: Ifed3b97a4b647b539dcd2df858572fa7da9a22d0
Iccfe3bd4fe45a0319bd9f23b8cbff4c7070c9f4d changed Block from using
malloc to std::make_unique, which does the equivalent of
`new char[size]()`, which value initializes the array members to 0.
Switch to `reset(new char[size])` to avoid this costly initialization.
Test: adb_benchmark
Change-Id: I09aacb949a7bd4a946ce35a8ee65d1f451577b72
It's possible to build against an SDK that's newer than the OS that
we're actually running via Xcode update, and Apple increments the
unversioned IOKit interface IDs to the newest version on every release,
which leads to mysterious failures to acquire an interface.
Pin the interface versions to IOUSBFamily 5.0.0, which shipped on OS X
10.7.3.
Bug: http://b/119264733
Test: ./test_device.py
Change-Id: Id26760bc62c89a1f7ef67511b21f9d9252ab69f3
Use Windows 10's improved virtual terminal emulation to handle more
terminal sequences, to fix console resize artifacts, and to fix issues
with the last column.
`top' and `more' work well now.
Use a sensible default TERM value of xterm-256color.
Test: manual on Windows 10
Bug: https://issuetracker.google.com/issues/37081235
Change-Id: If6612f3bb7db1ecaaa4b7930c9e54075ba7abb3f
Signed-off-by: Spencer Low <CompareAndSwap@gmail.com>
Regression introduced commit bb1efbd1c6
By not first checking if overlayfs mount on /system is in /proc/mounts
remount fails to discover the override and take the selected action.
Test: adb-remount-test.sh
Bug: 109821005
Change-Id: I1f0bae22caae1bca84fc1e33b508b54d98f4ed5c
This reverts commit d3bf1a8d7e.
Reason for revert: Now Soong can handle NOTICE files for static libraries.
Bug: 36073965
Test: mma + visual check
Change-Id: I52ebd3518f4d15800f878a6401df2c3a2d6713eb
The version bump we did wasn't actually needed, since adbd doesn't
check for either feature flag we added. Revert the change and clarify
the comment suggesting the version bump when adding features.
Test: ./test_adb.py
Test: ./test_device.py
Change-Id: I92e7f392bcb36b3bf9a236f2d31ba5133de8d72a
This patch addresses two issues with adb remount.
First, the blk_device field in fstab_rec will be incorrect if referring
to a device-mapper node (fs_mgr_update_logical_partition must be
called).
Second, devices that are not system-as-root now move their mount point
and chroot to "/" during first-stage init. However, the fstab entry will
have "/system" as its mount point. The new logic now reflects this.
Bug: N/A
Test: adb remount on hikey or device with dynamic partitions
Change-Id: I48b379db49d2def555ea49888c4994fbcf501672
sysconf(_SC_GETPW_R_SIZE_MAX) may return −1 if there is no hard limit
on the the buffer size. Some libc implementations such as musl don't
define this limit and will return -1.
Use a default buffer size to handle this case.
Change-Id: I997b13a2c2dca00574e049a259135e61c8ed8e03
Signed-off-by: Robert Yang <decatf@gmail.com>
This should fix the bloody Mac build, which doesn't have <error.h>. Since
we weren't entirely happy with error(3) anyway, switch to the toybox
style of error_exit and perror_exit, which are slightly briefer and quite
a bit more intention-revealing.
Bug: N/A
Test: builds
Change-Id: Ic8e411906c363af51657da5ce947b25a0b6bb1f3
* changes:
adbd: implement a nonblocking USB Connection.
adbd: rename daemon/usb.cpp to daemon/usb_legacy.cpp.
adbd: open functionfs ep0 as O_RDWR.
adbd: extract functionfs fd creation.
Implement a Connection that implements a nonblocking interface to
functionfs, to replace the existing implementation that uses two
threads that loop and call read and write respectively. The existing
implementation is vulnerable to a race condition that can occur when a
connection is terminated, where one thread can notice failure and
complete reinitialization of the USB endpoints before the other thread
noticed anything went wrong, resulting in either the first packet
coming from the other end disappearing in to the void, or the other end
getting a packet of garbage.
As a side benefit, this improves performance on walleye from:
push 100MiB: 10 runs: median 49.48 MiB/s, mean 50.00 MiB/s, stddev: 2.77 MiB/s
pull 100MiB: 10 runs: median 75.82 MiB/s, mean 76.18 MiB/s, stddev: 6.60 MiB/s
to:
push 100MiB: 10 runs: median 73.90 MiB/s, mean 73.51 MiB/s, stddev: 5.26 MiB/s
pull 100MiB: 10 runs: median 105.90 MiB/s, mean 107.19 MiB/s, stddev: 6.10 MiB/s
Test: python test_device.py
Change-Id: I9b77c1057965edfef739ed9736e5d76613adf60a
Previously, we were relying on tests run previously to set the main
thread ID, which fails when the tests are sharded.
Test: for i in `seq 0 4`; do GTEST_SHARD_INDEX=$i GTEST_TOTAL_SHARDS=5 /data/nativetest64/adbd_test/adbd_test; done
Change-Id: I15ce70b4598b6ffcf3f5467d16c917f23f668daf
Let's use LOG(FATAL)/PLOG(FATAL) for actual fatal stuff.
Add a Windows error(3) and move folks who didn't really mean "abort"
fatal over to it. Also get rid of syntax_error which wasn't adding a
lot of value, and most of the places it was adding "usage: " didn't seem
entirely appropriate anyway.
In particular, we seemed to have confused fastdeploy.cpp into aborting
in most user error cases, and none of the reviewers noticed. Clearly
we'd all lost track of far too many options.
(I've also cleaned up a few random instances of fprintf(3) + exit(2).)
Bug: N/A
Test: manual
Change-Id: I3e8440848a24e30d928de9eded505916bc324786
Extract common functionality that will be useful for implementing a
nonblocking Connection for functionfs.
Test: mma
Change-Id: Iae4f3d55b2f56568d4a5138db9a2af5622eb83ad
Run some more tests for adb.
Test: tradefed.sh run template/local_min --template:map test=atest --include-filter adb_integration_test_device --log-level VERBOSE --log-level-display verbose
Bug: 112104122
Change-Id: I44b5a2aed8d86ab2dca62081d5b3b05f6e01cabd
Inefficient to always update the symlinks to overlayfs, can
also lead to
failed to copy 'xxx' to 'yyy': remote symlink failed: File exists
on older variants of overlayfs that do not effectively mark them
deleted.
Test: manual
Bug: 109821005
Change-Id: If1286d76f45ce14087cdb515fe8d2fed198fd9d8
This reverts commit 04b9ca8c31.
The original incarnation of this patch was falsely triggering when our
poll would block with no activity happening for an extended amount of
time. When the poll returned, we would immediately flag that as a spin
loop. Solve this by tracking the last time we looped, to detect this.
While we're at it, switch from CLOCK_MONOTONIC to CLOCK_BOOTTIME, for
the same reason.
Change-Id: Ibcdf69d93f7b6012142cafd72066f39494c1f84b
Test: ./test_device.py
Removed call to external aapt2 process
Replaced several layers of error handling with calls to fatal()
Changed output messages for failures to be more useful for diagnostics
Bug: 116753196
Test: mm
Test: adb install -r --fastdeploy --force-agent --local-agent ~/example_apks/example.apk
Test: adb install -r --fastdeploy --no-streaming --force-agent --local-agent ~/example_apks/example.apk
Change-Id: I6006d8aa584e789a086a31e79a41d1416e54402a
The standard (RFC 1122 - 4.2.2.13) says that if we call close on a
socket while we have pending data, a TCP RST should be sent to the
other end to notify it that we didn't read all of its data. However,
this can result in data that we've succesfully written out to be dropped
on the other end. To avoid this, instead of immediately closing a
socket, call shutdown on it instead, and then read from the file
descriptor until we hit EOF or an error before closing.
Bug: http://b/74616284
Test: ./test_adb.py
Test: ./test_device.py
Change-Id: I36f72bd14965821dc23de82774b0806b2db24f13
Currently `adb remount` won't remount symlink mount points.
In Android Generic System Image, there is a symlink
/product -> /system/product for devices with and without a physical
/product partition to work, respectively:
- Mount product partition under /system/product via
'mount /product' OR
- Keep using /product -> /system/product symlink,
when no product partition
Currently find_proc_mount() is seeking "/product" under /proc/mounts.
But the actual mount path is "/system/product" when GSI is used
on a device with product partition.
Bug: 111539442
Test: adb remount && touch /product/abc on both GSI and non-GSI
Change-Id: I8f15a67109d0a3f4ee18596ef7eb4280c5631b11
Also stop using size_t (which will be 32-bit on Win32) for offsets in
adb_sideload_host, and stop truncating large file sizes in our
"sideload-host" messages.
Bug: http://b/112003354
Test: builds
Change-Id: If5b7cc9d1fc7ff7ca6eaebd20418f7b061846203
Implement a version of fdevent_destroy that returns the file descriptor
previously owned by the fdevent, instead of closing it.
Bug: http://b/74616284
Test: treehugger
Change-Id: I6a14ac96e6b8c801bc71bfcc9094851c158510ae
It was doing a test with `[ -d foo`, without the closing square bracket.
Test: python -m unittest test_device.FileOperationsTest.test_push_empty
Change-Id: I996b98850cf916986ef969768a7235547fcc404a
We can not mount overlayfs on root / as it will intercept all file
transactions. Thus must remount /system instead.
Test: manual
Bug: 109821005
Bug: 115751838
Change-Id: Idf646b8c1ca8a0d006bf64e23a76a7e909f9162a
Add an fstab argument for fs_mgr_overlayfs_mount_all so that it can
leverage the locally and timely acquired fstab entries. Affects all
callers, adb and init.
Test: manual
Bug: 109821005
Bug: 115751838
Change-Id: I96e2045d88525a6ce39bef63327a0fcf0704e9bc
Devices not supporting logical partitions do not require the command.
Bug: 78793464
Test: 'adb reboot fastboot' reboots to bootloader for a device
not supporting logical partitions.
Change-Id: I0a7b6d4335040e24efbf461e2dd2d56ab1985b7f
Bug: http://b/91353691
- libcxx has ETXTBSY for Windows
- adb/sysdeps/memory.h is no longer needed
Test: Build and test Windows binaries under Wine.
Change-Id: I9c27087d46c49cb25b391c4adae8d9e24724784d
For symmetry with test_adb.py, since they have different interpreter
requirements now.
Test: ./test_device.py
Change-Id: I02fe659a7216a619383661c8019d356f9ccfb34d
The system image is now a root image, even on devices that don't use
system-as-root. Fix remount to always remount /, to account for this.
Bug: http://b/115751838
Test: `adb remount; adb shell mount | grep " / "`
Change-Id: I2d89a795bd0bc53af33fe69c9d049bbb64d26f3b
Removed support for -f shortcut flag as this conflicts with some package manager flags
Renamed use_localagent global to match conventions
Fixed case where tmp files were created unnecessarily
Removed dead code (delete_host_file)
Cleaned up multiple layers of error handling by using fatal() as soon as error conditions occur
Fix: 113631900
Test: mm
Test: adb install -r --fastdeploy --nostreaming --force-agent --local-agent ~/example_apks/example.apk
Test: adb install -r --fastdeploy --no-streaming --force-agent --local-agent ~/example_apks/example.apk
Test: observe that fast deploy works as usual
Test: adb install -r -f --nostreaming --force-agent --local-agent ~/example_apks/example.apk
Test: adb install -r -f --no-streaming --force-agent --local-agent ~/example_apks/example.apk
Test: observe that fast deploy is no longer invoked by -f
Change-Id: Ic719df1003ac319e48b32f7f377f6f91d17f6a6f
(cherry picked from commit 0584689beaff604ceeccaf706dc368213d07b977)
deploypatchgenerator is now a dependency of adb
deployagent is now a java_binary target and uses the built in 'wrapper' parameter to bundle deployagent.sh
fastdeploy.cpp code in adb modified to reference deployagent rather than deployagent.sh (required to work with built in wrapper functionality).
removed near-redundant fastdeploy_init and TFastDeployConfig singleton as per:
https://android-review.googlesource.com/c/platform/system/core/+/740521/12/adb/client/fastdeploy.cpp#36
relocated kFastDeployMinApi to more appropriate location
Test: rm -rf $ANDROID_ROOT/out
Test: make sdk -j40
Test: find out/ -iname deploypatchgenerator
Test: observe that deploypatchgenerator.jar has been built as a dependency of the sdk target
Test: mm
Test: adb install -r -f --force-agent --local-agent ~/example_apks/example.apk
Test: adb install -r -f --no-streaming --force-agent --local-agent ~/example_apks/example.apk
Change-Id: I4e52d32f87774b44845bf6b5be0bae331a0b2324
Use a new Windows 10 API, SetThreadDescription(). Background info from a Chrome developer:
https://randomascii.wordpress.com/2015/10/26/thread-naming-in-windows-time-for-something-better/
Test: Win10, Vista, adb shell
Test: Ran windbg -pn adb.exe, used .dump /ma to create a dump, ran
`dx -g @$curprocess.Threads' on the dump and it showed the thread names.
Change-Id: I14ea7121605cb3fa45ce7b59e2ba5882a215b59f
Signed-off-by: Spencer Low <CompareAndSwap@gmail.com>
The code was passing an fd from adb_open() to android::base::ReadFdToString() which actually
takes a C-Runtime fd (on Windows), so it wasn't working.
The fix is to use APIs that deal with C-Runtime fds:
* unix_open()
* android::base::unique_fd
* unix_lseek() (added in this change)
I also removed an unnecessary call to GetProcessId() since we already have the process id
from the structure returned by CreateProcess().
Test: adb start-server on Win10 and Ubuntu (with a failing server)
Test: mma
Change-Id: Id6e2dd5532a02fe5d9caf96aa007a1b3434a0b59
Signed-off-by: Spencer Low <CompareAndSwap@gmail.com>
Unittest for a5b06b0ff8. Works by
simulating resume from sleep/hibernation by sending a Windows message.
Test: python3 -m unittest test_adb.PowerTest
Change-Id: I78510f30c012f68eda39764da522dbf8d03f2576
Signed-off-by: Spencer Low <CompareAndSwap@gmail.com>
Also fix adb's version number to match.
See build/soong/README.md for more information.
Test: cd system/core/fastboot; mma
Test: fastboot --version
Test: adb --version
Test: out/host/linux-x86/nativetest/fastboot_test/fastboot_test
Test: out/host/linux-x86/nativetest64/fastboot_test/fastboot_test
Change-Id: I65ea39af9183c602e84f3bc0e4a0d066a30fc464
It doesn't seem like b3c14ec693 was really
waiting for adb server death because the server would write OKAY to the
socket, causing the client to return from ReadOrderlyShutdown() before
the socket was actually closed.
The fix is to read the OKAY and then call ReadOrderlyShutdown() which
will then wait for the socket to be closed.
Test: Made the server hang in adb_server_cleanup() during kill-server
and kill-server would hang (as desired).
Test: To test resilience, tried preventing the server from sending OKAY and
adb kill-server still worked properly.
Change-Id: I4f0f11b86bbdc983670273ef5d3bb1dff7ab3697
Signed-off-by: Spencer Low <CompareAndSwap@gmail.com>
On Windows, when running adb server nodaemon and pressing Ctrl-C,
adb_server_cleanup (an atexit handler) would call kick_all_transports()
which would eventually fail a CHECK because the current thread was not
equal to the main thread. This is because Ctrl-C is implemented in
Windows by the OS creating a new thread in the process and calling the
Ctrl-C handler from there.
The CHECK fail would print out the CHECK expression and call abort()
which would record a crash in the Windows Event Log, plus would
potentially upload a crashdump to Microsoft's Watson service.
This might be a regression from d51c6df1ef.
The fix is to share more code between platforms, removing the call to
Win32 SetConsoleCtrlHandler() and just use the C Runtime's signal()
implementation which is built upon SetConsoleCtrlHandler(). The signal
handler still ends up being called from another thread, but the handler
is thread-safe enough so this seems to work.
Test: On Win10 and Vista, run adb server nodaemon and then try Ctrl-C,
Ctrl-Break and close console window.
Change-Id: I6603970616098d2b3ce68f2a3d4e5515ec859811
Signed-off-by: Spencer Low <CompareAndSwap@gmail.com>
We were accidentally returning 0 instead of the number of bytes written
whenever we wrote a USB packet that had a size that was a multiple of
the USB packet size, which resulted in the device getting kicked.
Bug: http://b/113070258
Test: python test_device.py
Change-Id: Ib3d6415545e90e1f4730afc8ad8713d10bb1534a
It was only needed when we statically linked libfec into libadbd in
past.
Test: `mmma -j system/core/adb`
Test: `m -j installclean && m -j bootimage` with aosp_taimen-userdebug.
Check the installed files in recovery image.
Change-Id: I8627d5f98568ab1d71e6a47cb1b2540a33bffdab
Make `adb disconnect` remove transports immediately, instead of on
their next reconnection cycle.
Test: adb connect unreachable:12345; adb devices; adb disconnect; adb devices
Change-Id: I35c8b57344e847575596d09216fc636be47dde64
Unit test for adb on Windows writing Unicode to an actual console
Window, as opposed to a pipe or file.
Test: Ran test on Ubuntu and Windows 10, tested version of adb with
incorrect Unicode handling and verified that test failed
Change-Id: Ibdda46d0fee83004537bcbb48a5c2fd6d3e1d593
Signed-off-by: Spencer Low <CompareAndSwap@gmail.com>
We can't remove from the middle of a priority_queue, which a followup
commit wants to do, so switch to std::set, with the side benefit of
making operator< point the right direction.
Test: mma
Test: ./test_adb.py
Change-Id: I784c1dcc91f0a9cf760e9fa1710202e37e85432b
Provide a way to selectively disable USB, automatic emulator
connection, and mDNS, to make reading through ADB_TRACE logging easier.
Test: manual
Change-Id: Ic6a06cdce14235415520cbf2d51c7f7031b7a7b3
* changes:
adb: report connection status when we're unauthorized.
adb: disable test_device.FileOperationsTest.test_pull_symlink_dir.
adb: fix test_device.FileOperationsTest.test_push_empty
adb: switch test_adb.py over to double quotes.
adb: make test_adb.py work on windows.
adb: switch test_adb.py to python3.
* changes:
init: add fs_mgr_overlayfs_mount_all to FirstStageMount
adb: add overlayfs handling for readonly system filesystems
fs_mgr: get fs_mgr_mount_all to call fs_mgr_overlayfs_mount_all
fs_mgr: add overlayfs handling for squashfs system filesystems
Previously, connecting to devices that end up as unauthorized would
wait 10 seconds before reporting failure to the user. After this
change, notification happens as soon as the adb server realizes.
Test: manual
Change-Id: If7c8d38f22da3d98b952eee6a334abc8566bb751
Switch from os.pipe to socket.socketpair for the fake adbd termination
termination signaller, that we can select on it on windows.
Test: test_adb.py on windows (2 failures on windows)
Change-Id: I37df06e465ec8be28cfb18a5c21ef30125195004
Currently, test_adb.py doesn't work on Windows, because we're selecting
on a pipe, which doesn't work on Windows. Python 3.5 adds socketpair
support to Windows, so switch over to Python 3 to make it so we can use
that.
Test: python3 test_adb.py
Change-Id: I0fbd1bca0b28324658831d5ef8ee08aefe2b0596
adb verity-on setup overlayfs directories, do not skip if verity
feature not configured for device.
adb verity-off teardown overlayfs directories, do not skip if verity
feature not configured for device.
adb remount setup overlayfs directories and mount if possible, and
do so without a reboot if possible.
Test: compile
Bug: 109821005
Bug: 110985612
Change-Id: I28c4c23fea44b886c5523f0286a523f0b0485df0
In order to run them continuously, add them to general-test
zip.
Test: make general-tests
Bug: 112104122
Change-Id: I7a1143ac638626069ed78ddd450063e9489e22e8
Previously, we were returning the result of SendOkay/SendFail in a few
places after handling a host request, which is incorrect for two
reasons. First, the return type of SendOkay/SendFail is bool, and
handle_host_request was expected to return 0 on success. Second, we
don't care if the SendOkay fails; if we got to that point, we're done
with the request, regardless of whether we succeeded to report our
result. The result of this was a spurious failure result reported after
the initial result, which was ignored by the adb client.
Test: manually straced adb server
Test: python test_adb.py
Test: python test_device.py
Change-Id: I7d45ba527e1faccbbae5b15e7a0d1557b0a84858
* changes:
adb: actually enable fdsan.
adb: move AdbCloser to its rightful place.
libdebuggerd: extract and print the fdsan table.
crash_dump: pass the address of the fdsan table.
adb was using a custom unique_fd closer that didn't have an
implementation for fdsan, which meant that none of our FDs were
actually tracked. Guard this behind ifdefs so that we only use this
on Windows, and delete our implementation of Pipe in favor of the one
in libbase while we're at it. libbase's implementation always sets
O_CLOEXEC, so fix up the instance of Pipe that doesn't expect that.
Test: mma
Test: adb start-server
Test: debuggerd `pidof adbd`
Change-Id: Ic29d641a2f93fb42384b00c51775048c8bcbe152
Cody Schuffelen