Previously, if we failed to read a file after successfully opening it,
(e.g. because it's a directory), we would return prematurely without
notifying our caller, which would then wait forever for a response that
isn't coming.
Switch all of the adb install callsites over to checking the result,
but leave the other callsites ignoring the result, because they're
either deprecated, or don't care.
Bug: http://b/145621968
Test: mkdir foo.apk; adb install foo.apk
Change-Id: Ia82f8280b144f7881e067a10cd17f9a89019cf3f
fwrite can cross windows platform to show log on terminal,
so we use it instead of WriteFdExactly.
Change-Id: I252c15a0bc72e1dbd98b7b5ec8000ecdcde16416
Signed-off-by: Junyong Sun <sunjy516@gmail.com>
This patch introduce "service.adb.listen_addrs", a new
string type property, while keeping the old properties.
The new property added in this patch is used to listen
on UNIX domain socket "localfilesystem".
"service.adb.listen_addrs" can be used to listen on
multiple addresses, such as tcp:5555 and tcp:5556.
It is separated by ',' (comma) character.
In the process of introducing the new socket type, the
method tcp_connect is removed and combined into
socket_spec_connect.
Without specifying using the new property, adb will
try to listen on both tcp and vsock (following the
previous implementation).
Some examples of the new property value are:
- "tcp:5555"
- "vsock:5555"
- "localfilesystem:/tmp/test"
- "tcp:5555,vsock:5555"
Bug: 133378083
Test: On master-arc-dev:
adb root;
setprop service.adb.listen_addrs localfilesystem:
<path_to_socket>;
adb connect localfilesystem:<path_to_socket>;
adb -s localfilesystem:<path_to_socket> shell;
inside Chrome OS.
Test: On aosp_bluefin:
setprop service.adb.listen_addr tcp:5555;
adb connect <IP>:5555; adb shell;
Test: On aosp_bluefin:
setprop service.adb.tcp.port 5555;
adb connect <IP>:5555; adb shell;
Test: On aosp_bluefin:
setprop service.adb.listen_addrs tcp:5555,tcp:6666;
adb connect <IP>:5555; adb shell;
adb connect <IP>:6666; adb shell;
Test: On aosp_bluefin:
./adb_test;
Test: On cuttlefish:
launch_cvd;
adb -s 127.0.0.1:6520 shell;
Test: Ran host tests:
./adb_test;
./adb_integration_test_adb;
./adb_integration_test_device;
Change-Id: I8289bf0ab3305cf23ce5695ffba46845d58ef6bb
Actually log results of adb commands, so we can see why this fails.
Fix the asan-detected use of a value from a reused part of the stack.
Test: atest FastDeployTest on Linux (with an asan adb)
Change-Id: I4d2bbae62a301e16065d604a2c9918077489cafb
When the adb client starts the adb server, it waits until the server
reports that it's fully-initialized (via reply-fd) before executing
its adb client operation. This wait prevent that adb client from
talking to the server while it's initializing and becoming confused.
But if a *different* adb client connects to the server while it's
initializing, *that* client can temporarily observe unexpected state
while the server initializes itself. For example, such a client can
observe a device that's alive and connected as being offline while the
server connects to it.
The new socket activation support makes this race more apparent, since
in the socket activation configuration, there's no initial adb client
waiting for the server's all-clear indication and so even the first
client observes the partially-initialized server state.
This CL prevents the server accepting *any* client connection until
the server has fully initialized itself, preventing all clients, not
just the initial client, from observing a
partially-initialized server.
Test: test_adb.py; test_adb gtest binary
Test: [with socket activation] adb kill-server; adb devices
Change-Id: I5d399ee62436eee63340b6b8b0f64131ad17ac65
Socket activation allows adb to be run as a system daemon that starts
only as needed and supports race-free transparent restarts of the adb
server after a client issues an "adb kill-server" command.
Test: see SOCKET-ACTIVATION.txt
Change-Id: Ieabf08710ce4365e5513102f3aa578560aa7355e
Call adb_notify_device_scan_complete when we finish enumerating USB
devices. The original intent of the code appears to have been to have
adb_wait_for_device_initialization return as soon as we've finished
looking around at the system USB environment on daemon startup, but at
some point, we forgot to actually notify the init thread that we
finished scanning all the USB devices, forcing the wait on init_cv to
time out after three seconds on every daemon initialization.
After this change, the daemon starts in a few milliseconds on my Linux
machine
Test: killall adb && sleep 1 && time adb server && adb shell ls
Change-Id: I0bc1da7a597d2077dd2b591560d03798b05905b7
Before we just got "Failed to read key". After:
adb E 10-07 08:20:14 258249 258249 auth.cpp:176] Failed to read key \
from '/usr/local/google/home/enh/.android/adbkey'
94390117965240:error:0900006e:PEM routines:OPENSSL_internal:NO_START_LINE:\
external/boringssl/src/crypto/pem/pem_lib.c:622:Expecting: ANY PRIVATE KEY
Also fix the misleading "Failed to generate" message from adb_auth_init.
Bug: http://b/141715453
Test: manually corrupted key; see above
Change-Id: I6732ee6b683c8548d596d7c22eeddab8ce9a3cea
- removed 2GB apk size cap,
- removed zip archive parsing on device (1.1M->236K agent size reduction),
- optimized matching entries search,
- added more robust matching entries search based on hash of CDr entry,
- reduced patch size by reusing Local File Header of matched entries,
- removed extra manifest parsing and extra agent calls,
- added device-side tests for agent,
- fix for Windows patch creation.
Test: atest adb_test fastdeploy_test FastDeployTests
Total time for 0-size patch reduction for 1.7G apk: 1m1.778s->0m36.234s.
Change-Id: I66d2cef1adf5b2be3325e355a7e72e9c99992369
dirname (on glibc, at least) preserves multiple leading slashes, and we
were looping until path != "/", which would lead to an infinite loop
when attempting to push to a path like //data/local/tmp.
Bug: http://b/141311284
Test: python -m unittest test_device.FileOperationsTest.test_push_multiple_slash_root
Change-Id: I182b3e89ef52579c716fdb525e9215f1fe822477
A regression from commit 8c2198c809
("adb: use shell for remount to forward return codes.") where the
optional argv[1] got missed for the remount command. This change
hands off _all_ the arguments if to a shell and activates some of
the extra features in the remount command.
$ adb remount --help
remount [-h] [-R] [-T fstab_file] [partition]...
-h --help this help
-R --reboot disable verity & reboot to facilitate remount
-T --fstab custom fstab file location
partition specific partition(s) (empty does all)
Remount specified partition(s) read-write, by name or mount point.
-R notwithstanding, verity must be disabled on partition(s).
$
SideEffects: adb remount [-h] [-R] [-T fstab_file] [partition]...
Test: adb-remount-test.sh
Bug: 138577868
Bug: 139283818
Bug: 139226412
Change-Id: I8223d4000ab20857e9b634e4d4a326eed530d7be
We've seen USB writes failing due to inability to allocate contiguous
chunks of memory in the kernel on devices, but it looks like the same
problem can occur on the host, as well. It's a mild performance
regression (90->80 MB/s on a blueline) to split the writes always, so
attempt the full write first, and fall back to splitting it up if that
fails with ENOMEM. Once we switch over the the asynchronous transport
API, we'll be able to submit multiple writes cheaply, like on devices,
so we won't need to retry at that point.
Bug: http://b/140985544
Test: test_device.py
Change-Id: I1517c348375b829dfff6796c4e9d394802b02d5b
1. Use bigger buffer for transfers - 64kb is the default size
for push, so let it be the same in streaming
2. Use abb when it's available for lower overhead
3. Add a posix_fadvise() on the source APK
4. Increase buffer sizes for the socketpair that's transferring
the data from adbd.
Overall this saves about 25% time for streaming installations
and makes it faster than the legacy push (at last!)
Test: manual
Change-Id: Ieb84284da2058944815e062ef6e4389b842565fa
If we're failing out, don't print any messages from the shutdown code:
it's not interesting (and actively confusing!) to talk about the
"Success" of finalizing an aborted session.
Also fail early if stat fails the first time (we have to tell
PackageManager how much data to expect, for some reason).
Also use `bool` as the boolean type and add quotes around filenames in
error messages.
`install_multi_package` looks like the same nonsense copy & pasted, but
I don't know what that is, and haven't seen any bug reports, so I'm
ignoring that for now.
Bug: http://b/133522725
Test: manual
Change-Id: Id054dff44f21f4c4cc5877dcc2735fc1eb25f9a0
A bit of a break with convention, but we do already document `-t` which
is actually a PackageManager option, not an adb option, and `--abi`
seems like it'll be seeing increased usage these next few years.
Also add a pointer to `adb shell pm help` for full details of
PackageManager options.
Bug: http://b/122487331
Test: manual
Change-Id: I93f48d202e5707b27caadb5af2a65c485c6d3407
Also add a -l to the undocumented `adb track-devices` for easier
testing.
Bug: https://issuetracker.google.com/118470090
Test: manual
Change-Id: I436f6e45b249b98a073aa147ad4ed67d4d87f9bd
The default port of 5555 was removed a while back, but the help text was
never updated, and other contexts still allow a default port.
Bug: https://issuetracker.google.com/128561172
Test: manual
Change-Id: I6e5e6edeb2243386a0301c0a6cad569189f9910e
Seems like the bug submitter thought that -l listed the partitions shown
on the line above, rather than the files. Be more explicit.
Bug: https://issuetracker.google.com/73122762
Test: treehugger
Change-Id: I6ed1027639a5b785a5992909430d739958d10ad7
