Previously, chatty had logic that would skip a certain number of log
entries to satify pruning, but otherwise keep the reader connected.
This was a best-effort attempt at helping pruning and had additional
logic that handled further disruptions, if logd's memory was 2x the
allotted memory.
The new logic has two components:
1) memcpy() each individual log message in FlushTo() such that there
are no references to the underlying log data without a lock held.
Note, that this memcpy is completely negligible for performance.
2) In Prune(), immediately delete all log chunks required to reduce
memory to the allotted amount, which is now safe given 1). If readers
will lose logs, continue to print a warning.
This additionally makes the Clear() logic deterministic. It was
previously best effort in chatty, but will immediately and always
clear all logs for SerializedLogBuffer.
Bug: 163617910
Test: logging unit tests
Test: Prune() immediately frees buffers during high log pressure
Test: Clear() immediately frees buffers during high log pressure
Change-Id: I40fe9b791312af3dc256b166e5c34425f4ca51ac
See the comment in the code for more details. 'wrapped' readers are
uncommon and error prone, and this change makes them more reliable.
Its side effect is that wrapped readers will wake more often, but
they'll still be batched to a large degree.
Bug: 163617910
Test: logging unit tests
Test: logcat --wrap does the right thing
Change-Id: I4b6f8331ff7854787c97f821b2a5bf8d7da321c6
With a couple recent changes to APIs taking a const std::string&,
instead of a const char *, we switch to directly passing in a
reference to our string.
Test: TreeHugger
Change-Id: Id9f832dd5e3def28cfeaa0d3cda7da323332b16c
This reverts commit 4af349b072.
Reason for revert: post submit test logs too much fails after this
Bug: 163611679
Change-Id: If9939f3c9864dad67536d0c5481f662b9d07c58e
This function has been marked deprecated for quite some time, has no
users, and isn't in liblog.map.txt.
Test: build
Change-Id: Ic3bf47c780b18a5d4f1e997b9265be76049fb824
Merging sparse chunk can make sparse map block bigger than 4GiB,
that can't be covered by unsigned integer type. Fix this by
changing unsigned int to uint64_t type.
Test: sparse build
Bug: 162808120
Change-Id: Id4d3f88f9d531c25c3937c99b2c81efb915605ee
Signed-off-by: Hyeongseok Kim <hyeongseok@gmail.com>
Cc: hyeongseok.kim <hyeongseok.kim@lge.com>
liblog uses a non-blocking socket for sending messages to logd as a
performance/reliability trade-off, favoring not blocking processes
when they log over reliably tracking all logs.
This change asserts that the above is the wrong trade-off: that log
relability is more important than slight delays when logging. Further
points to consider:
1) The new logd implementation without chatty has lower latency and
less variance in its ::Log() function.
2) liblog also writes to /dev/pmsg0 and this call is already blocking.
3) Logging is already expensive and excess spam must be eliminated
regardless.
Bug: 151654749
Bug: 155922578
Bug: 160314220
Test: No messages are dropped by this socket
Change-Id: I163b7d51aa73a10978850f1868cb9fc4bf9ead1b
Open README.compression.md in a markdown viewer for details.
Test: vscode and gitiles render this correctly
Change-Id: Iaa0093dc565b5c6bb016c9e97248375f0906d0c8
The name "pre_gki_level" is causing some confusion because not all
devices launching with Android R are subject to the GKI requirement.
(See b/161563110#comment11.) E.g., devices that use a 4.14-based kernel
are exempt from GKI. However, the encryption requirements still apply.
Just use __ANDROID_API_Q__ directly instead.
No change in behavior.
Change-Id: I4242745ccc9f88e084a1adfab5796daa9bc59b6e
Convert rules to build charger resources to Soong
assuming BOARD_CHARGER_NO_UI is not set. As of now,
no devices in the source tree sets this variable. Also,
these images should be installed to the system image
regardless of this variable.
This change moves battery_fail and battery_scale to
/system/etc/res/images/charger.
Test: build and inspect system/etc/res/images/charger
Fixes: 143907179
Change-Id: Idccf5631371c7520cbfcfa6e92d1e88d0f3f95e7
When charger resources are missing from /product, before this change,
images are loaded from /res/images/charger. After this change, charger
first checks if the resource exists at that location, and if not, use
resource under /system/etc/res/images/charger.
On devices that uses resources from /product, this change has no effect.
On devices that uses resources from /res,
- if an old image is mounted at /, charger continues to load resources
from /res.
- if a new image is mounted at /, that is, battery_fail and
battery_scale may not exist under /res, charger loads them from /system.
There are no font files installed by modules defined here, so any
device-specific font files stay at /res.
Also create virtual function wrappers over minui APIs to be mocked
later.
See next change on moving battery_scale and battery_fail images to
/system.
Test: build and inspect system/etc/res/images/charger
Bug: 143907179
Change-Id: I4895b9602274fa4a1ebb4d7d3b842622b81b7b51
Now that libsysutils sockets do not ignore SIGPIPE, logd itself must
ignore it and we therefore add a test that it is ignored properly.
Test: this test passes
Test: this test fails if logd doesn't ignore SIGPIPE
Change-Id: I65474967f40291a7abd11cfd279c8cde2d2bca14
Consolidate command handling for commands that take a buffer ID as
their first argument. Further use ParseUint instead of atoi(), since
atoi("bad string") returns `0` is the same as LOG_ID_MAIN.
Test: these commands are handled property
Test: `echo -n 'getLogSize ignoreme!\0' | nc -U /dev/socket/logd`
returns an error
Change-Id: I3e84a6157e16f4c110681da453e1ae1a920da7c4
To support "override" services, we need to scan partitions from least
speicific to most specific.
Bug: 163021585
Test: m
Change-Id: I26a6de4f7fb571c60038e803137a4b1c237792fd
On non-VAB devices, fake super image creation may fail because it
may not support f2fs file pinning.
Also changes global setup to a testing::Environment object because it is
the recommended way in gtest's guide. SnapshotTestEnvironement::TearDown
is automatically executed before RUN_ALL_TEST returns.
Test: vts_libsnapshot_test
Test: make IsVirtualAbEnabled() return false, then all tests are
skipped.
Bug: 162557082
Change-Id: I1382f9a4ddec146fa2d8cfb21ac66ca22a2e110f
When checking IsUnreliablePinningAllowed, the existing code calls
IsTestDir on a value starts with "/data", but IsTestDir expects
metadata_dir_. Fix it.
Also, make subdir checks more robust and add test cases for IsSubdir.
Fixes: 162557082
Test: vts_libsnapshot_test
Test: fiemap_image_test
Change-Id: I495cfae3da11d1e0800b8abf520df10dd9a29dce
liblog exposed __android_logger_get_buffer_size() which queries log
size properties, but that should not be a generic library function.
Logd should be the only process that queries these properties and
other processes should query the actual used log sizes from logd via
the android_logger_get_log_*_size() functions.
Also, always use 1MB for log buffer tests, instead of just on host and
various other clean up related to improper types.
Test: log buffers are correctly sized
Merged-In: I9b7c86bf58e569618737afe9097cf1c4d4e61d95
Change-Id: I9b7c86bf58e569618737afe9097cf1c4d4e61d95
Elliott Hughes