Kernel panic - not syncing: Out of memory occurred once after 8 hours
automatic test which cases are about Summary:logd memory leak UTResult
: 1/1
IssueID: 93440
Rootcause: the method "LogTimeEntry::FilterSecondPass"(called inside
"LogBuffer::flushTo") does not check the condition flag
"LogTimeEntry::mRelease" which will be set to true when "logd" begins
to kick off a client when it finding the client is slow-est.
Therefore the client cannot be kicked off and hence the log-element
list becomes longer and longer if the slowest client reading more
slower than the writer writing.
Change-Id: I91a44a393e63b3122c15ce269ee195e0eb4339e8
Solution: add the check.
Test: logd kills slow reader
LogTimeEntry's lifecycle is spread out in various locations. It
further seems incomplete as there is logic that assumes that its
associated thread can exit while the underlying LogTimeEntry remains
valid, however it doesn't appear that that is actually a supported
situation.
This change simplifies this logic to have only one valid state for a
LogTimeEntry: it must have its thread running and be present in
LastLogTimes. A LogTimeEntry will never be placed into LastLogTimes
unless its thread is running and its thread will remove its associated
LogTimeEntry from LastLogTimes before it has exited.
This admittedly breaks situations where a blocking socket gets issued
multiple commands with different pid filters, tail lines, etc,
however, I'm reasonably sure that these situations were already
broken. A check is added to close the socket in this case.
Test: multiple logcat instances work, logd.reader.per's are cleaned up
Change-Id: Ibe8651e7d530c5e9a8d6ce3150cd247982887cbe
logd suffers performance degradation when multiple blocking
readers connect to it. Each time when the writer thread log
new entries, all of the readers are notified regardless of
which log id they are watching.
In this case, only give notification to the readers who are
actually watching new entries' log id. This decreases logd
CPU consumption by skipping unnecessary LogBuffer::flushTo
loops.
Test: liblog-unit-tests, logd-unit-tests & CtsLiblogTestCases
logcat-unit-tests
Test: manual:
1.'logcat –b all' at constant heavy logging load level
2.simultaneously 'logcat –b crash' in another session,
a healthy crash buffer usually keep empty
3.logd CPU consumption doesn't increase after step 2
Change-Id: I4ffc045c9feb7a0998f7e47ae2173f8f6aa28e8a
Switch to a reader writer lock for the Element List lock. Also setup
for a reader writer lock for the Times list, but continue to use a
mutex where rdlock() and wrlock() are the same implementation for now.
This should improve general reader performance and prevent blocking of
other reader operations or exit by a single hung logd.reader.per
thread. For example, a full length logcat of an empty buffer (eg:
crash log buffer) will hold a lock while the iterator scans the entire
list.
Test: gTest liblog-unit-tests, logd-unit-tests, logcat-unit-tests
Bug: 37378309
Bug: 37483775
Change-Id: If5723ff4a978e17d828a75321e8f0ba91d4a09e0
Move lastTid array from local in LogBuffer::flushTo to per-reader
context in LogTimes::mLastTid and pass into LogBuffer::flushTo.
Replace NULL with nullptr in touched files.
Simplify LogTimeEntry::cleanSkip_Locked initialization of skipAhead
to memset, to match mLastTid memset initialization.
Test: gTest liblog-unit-tests, logd-unit-tests & logcat-unit-tests
Test: adb logcat -b all | grep chatty | grep -v identical
Bug: 36488201
Change-Id: I0c3887f220a57f80c0490be4b182657b9563aa3f
Use getRealTime() instead and leverage private liblog log_time
comparison and math functions. This saves 8 bytes off each
element in the logging database.
Test: gTest liblog-unit-tests logd-unit-tests logcat-unit-tests
Bug: 35373582
Change-Id: Ia55ef8b95cbb2a841ccb1dae9a24f314735b076a
Switch _all_ file's coding style to match to ease all future changes.
SideEffects: None
Test: compile
Bug: 35373582
Change-Id: I470cb17f64fa48f14aafc02f574e296bffe3a3f3
(cherry pick from commit 0ecdec7a09)
release_Locked() is called with a reference count and threadRunning,
the static analyzer can not tell this and estimates that a call to
delete this will occur. So let us invent a new call
release_nodelete_Locked() to ensure it is clear we will not be
arranging a delete this in the context of this code path. The
delete this will follow in the immediate codepath in this function
after threadRunning is cleared, and decRef_Locked() is called.
Change will also remove any developer FUD regarding release_Locked()
usage at this location.
SideEffects: None
Bug: 27434831
Change-Id: I91b060b2dadc72cc449fa381c934afb577bee037
- limit AID_SYSTEM uid or gid to read security buffer messages
- adjust liblog tests to reflect the reality of this adjustment
To fully test all security buffer paths and modes
$ su 0,0,0 /data/nativetest/liblog-unit-tests/liblog-unit-tests --gtest_filter=liblog.__security*
$ su 1000,1000,1000 /data/nativetest/liblog-unit-tests/liblog-unit-tests --gtest_filter=liblog.__security*
$ su 2000,2000,2000 /data/nativetest/liblog-unit-tests/liblog-unit-tests --gtest_filter=liblog.__security*
ToDo: Integrate the above individually into the gTest Q/A testing
Bug: 26029733
Change-Id: Idcf5492db78fa6934ef6fb43f3ef861052675651
If a timeout is specified for the reader, then go to sleep
with the socket open. If the start time is about to get
pruned in the specified log buffers, then wakeup and dump
the logs; or wakeup on timeout, whichever comes first.
Bug: 25929746
Change-Id: I7d2421c2c5083b33747b84f74d9a560d3ba645df
- The reader region level indicates the location to protect, but once
it has been passed to the calling reader, then allow us to go
one beyond so that a clear or prune thread can remove the entry.
Bug: 23711431
Change-Id: I0f2389858dd8c83366c034360f67d7c363625b56
- Cleanup resulting from experience and feedback
- When filtering inside logd, drop any leading expire messages, they
are cluttering up leading edge of tombstones (which filter by pid)
- Increase and introduce EXPIRE_RATELIMIT from 1 to 10 seconds
- Increase EXPIRE_THRESHOLD from 4 to 10 count
- Improve the expire messages from:
logd : uid=1000(system) too chatty comm=com.google.android.phone,
expire 2800 lines
change tag to be more descriptive, and reduce accusatory tone to:
chatty : uid=1000(system) com.google.android.phone expire 2800
lines
- if the UID name forms a prefix for comm name, then drop UID name
Change-Id: Ied7cc04c0ab3ae02167649a0b97378e44ef7b588
- switch to simpler and faster internal sequence number, drops
a syscall overhead on 32-bit platforms.
- add ability to break-out of flushTo loop with filter return -1
allowing in reduction in reader overhead.
Change-Id: Ic5cb2b9afa4d9470153971fc9197b07279e2b79d
In a scenario in which an on-line (blocking) client is running and
a clean is attempted (logcat -c), the following can be observed:
1) the on-line logger seems to freeze
2) any other clear attempt will have no effect
What is actually happening:
In this case prune function will "instruct" the oldest timeEntry
to skip a huge number (very close to ULONG_MAX) of messages, this
being the cause of 1.
Since the consumer thread will skip all the log entries, mStart
updating will also be skipped. So a new cleaning attempt will have
the same oldest entry, nothing will be done.
Fix description:
a. keep a separated skipAhead count for individual log buffers (log_id_t)
LogTimeEntry::LogTimeEntry
LogTimeEntry::FilterSecondPass
LogTimeEntry::skipAhead
LogTimeEntry::riggerSkip_Locked
b. update LogTimeEntry::mStart even if the current message is skipped
LogTimeEntry::FilterSecondPass
c. while pruning, only take into account the LogTimeEntrys that are monitoring
the log_id in question, and provide a public method of checking this.
LogTimeEntry::isWatching
LogTimeEntry::FilterFirstPass
LogTimeEntry::FilterSecondPass
d. Reset the skip cont befor the client thtread starts to sleep, at this point
we should be up to date.
LogTimeEntry::cleanSkip_Locked
LogTimeEntry::threadStart
Change-Id: I1b369dc5b02476e633e52578266a644e37e188a5
Signed-off-by: TraianX Schiau <traianx.schiau@intel.com>
- permits easier determination of logd thread at
fault in a stack trace from debuggerd.
Bug: 14275676
Change-Id: Iac2c523147e2bcce34ab7ddcecd02582c5fa7cc0
* Create a new userspace log daemon for handling logging messages.
Original-Change-Id: I75267df16359684490121e6c31cca48614d79856
Signed-off-by: Nick Kralevich <nnk@google.com>
* Merge conflicts
* rename new syslog daemon to logd to prevent confusion with bionic syslog
* replace racy getGroups call with KISS call to client->getGid()
* Timestamps are filed at logging source
* insert entries into list in timestamp order
* Added LogTimeEntry tail filtration handling
* Added region locking around LogWriter list
* separate threads for each writer
* /dev/socket/logd* permissions
Signed-off-by: Mark Salyzyn <salyzyn@google.com>
(cherry picked from commit 3e76e0a497)
Author: Nick Kralevich <nnk@google.com>
Change-Id: Ice88b1412d8f9daa7f9119b2b5aaf684a5e28098
