Although ever present, an increased regression introduced with
commit b6bee33182 (liblog: logd:
support logd.timestamp = monotonic).
A signal handler can interrupt in locked context, if log is written
in the signal handler, we are in deadlock. To reduce the contention
and chances for this problem separate out timestamp lock from is
loggable lock to reduce contention situations. Provide a best-guess
response if lock would fail in timestamp path.
Use a common lock() inline within each module, with a comment speaking
to the issues surrounding calling a function that has a mutex within
a signal handler.
ToDo: Hold off signals temporarily in mainline, restart when unblock.
Can not use pthread_sigmask(SIG_BLOCK,,) as it breaks AtCmd.
Signed-off-by: Mark Salyzyn <salyzyn@google.com>
Bug: 25563384
Change-Id: I47e2c87c988c3e359eb9eef129c6a3a08e9eedef
Although ever present, an increased regression introduced with
commit b6bee33182 (liblog: logd:
support logd.timestamp = monotonic).
A signal handler can interrupt in locked context, if log is written
in the signal handler, we are in deadlock. Block signals while we
are locked. Separate out timestamp lock from is loggable lock to
reduce contention situations. Provide a best-guess response if
lock would fail in timestamp path.
Bug: 25563384
Change-Id: I6dccd6b99ebace1c473c03a785a35c63ed5c6a8a
if ro.logd.timestamp or persist.logd.timestamp are set to the value
monotonic then liblog writer, liblog printing and logd all switch to
recording/printing monotonic time rather than realtime. If reinit
detects a change for presist.logd.timestamp, correct the older entry
timestamps in place.
ToDo: A corner case condition where new log entries in monotonic time
occur before logd reinit detects persist.logd.timestamp, there
will be a few out-of-order entries, but with accurate
timestamps. This problem does not happen for ro.logd.timestamp
as it is set before logd starts.
NB: This offers a nano second time accuracy on all log entries
that may be more suitable for merging with other system
activities, such as systrace, that also use monotonic time. This
feature is for debugging.
Bug: 23668800
Change-Id: Iee6dab7140061b1a6627254921411f61b01aa5c2
__android_log_write() was just a special form of
__android_log_buf_write that used a default log ID and set the abort
message for fatal messages. Presumably the latter was intended to be
set for __android_log_buf_write as well.
Change-Id: I51cff7561a2754676b2088d95fa4f4505ac3c3c2
__write_to_log_null never retries initialization and only
made sense for the kernel logger. The user space logger
can come and go.
Bug: 19732485
Change-Id: Iac34ea1c52ec82db4ee0c2c73ba0950ace4d4dec
A regression after "liblog: Instrument logging of logd write drops"
where an empty payload would result in an insufficient iovec to
report the write drop.
Change-Id: Iffabcfbb0680898d7a42004700e638e9d940ff5f
If you rely on __builtin_trap, it's likely to use an illegal instruction,
which is a misleading way to abort. If we just call abort, it's more
immediately obvious that we've aborted.
Bug: 19644330
Change-Id: I63a962e4748aec7b019ea94b007593e478a3b61a
Fix host/sdk builds:
- Drop logprint from list of host products
- Drop <endian.h> for FAKE_LOG_DEVICE
Change-Id: I8aa854413ff6d809f0b04987cf913eb228e4213c
- If logger system is prostrated, send an event message with the
liblog tag from the associated UID and PID with a count of
dropped messages once logging is resumed.
- Added to the README a description of the error return values.
- Describe in the README the appropriate mitigations for dropped
messages.
- If the caller sees this message, then
/proc/sys/net/unix/max_dgram_qlen is likely too small
Change-Id: Iaf387b9e5e1b6aa93bebc7481f9e8353732e3229
- Enable build for 32 and 64 bit benchmark executables
- Fix some cosmetics issues in logd_write.c
Change-Id: I544446e5116607d7fec89171135f6e1eff6aebd8
Windows PE format doesn't have the same support for weak symbols
as ELF does.
The symbol android_log_id_to_name was strangely omitted from the
resulting Windows executable and the callsite where it was supposed
to be called was relaced with a different symbol, which caused a crash
in AAPT on Windows.
This change works due to a careful set of #ifdefs that ensure only one
definition of android_log_id_to_name is defined, but there are other
uses of __attribute__((weak)) which should be inspected.
Change-Id: I3b58076e31d9b28c5143773a170e9ffda9fa3815
Since android_set_abort_message() is quite helpful to the platform, it's
becoming promoted to a real header with a non-private name.
Bug: 17059126
Change-Id: I3ed12b66eb07f3a6a08366d5eee147809d9a3ce9
(cherry picked from commit 4047075da6)
Socket file descriptors remain open across exec unless
SOCK_CLOEXEC is set. Enable this option, to avoid leaking
file descriptors.
See https://android-review.googlesource.com/53736 for a similar
problem with the old logging code.
Change-Id: I9e045d2291ae6680044ab86604f3ff2c55b5eaed
Implement LOG_EVENT_STRING, so that native code can create new
event log entries. This is needed to support logging SELinux denials
to the event log.
Change-Id: I6a269a832bc2f5e5da6c9dbd169ed2f901b49166
This is more general and will work for anyone's custom logging
code, as long as they use ANDROID_LOG_FATAL priority.
Change-Id: Iaf7fc0858fce04f3af407882a58ee5a827d50ddd
- API change, return -errno on failure for log writing
- ENOTCONN results in a reconnection and a retry of the write
Change-Id: Ib0a90b730edeaa7e3a3c9797e4e1b0c81b6adb0a
- structure packing
- move towards log_time from struct timespec
- extend log_time to cover differences between
log_time and struct timespec
Change-Id: I106ed0b609917306d170044054b5b32645f2a295
* Modify liblog to send all messages to the new syslog user
space daemon.
Original-Change-Id: I0ce439738cd921efb2db4c1d6a289a96bdbc8bc2
Original-Change-Id: If4eb0d09409f7e9be3eb4bb7017073dc7e931ab4
Signed-off-by: Nick Kralevich <nnk@google.com>
* Add a TARGET_USES_LOGD make flag for BoardConfig.mk to manage
whether logd is enabled for use or not.
* rename syslog to logd to avert confusion with bionic syslog
* Add fake log support back in
* prefilter for logging messages from logd
* Fill in timestamps at logging source
* update abstract log reader
* switch from using suffix for id to v3 format
* log a message when creating devices that a deprecated interface
is being utilized.
Signed-off-by: Mark Salyzyn <salyzyn@google.com>
(cherry pick from commit 099e2c1f6f706a8600c1cef74cce9066fc315480)
Change-Id: I47929a5432977a1d7235267a435cec0a7d6bd440
* changes:
libsysutils: SocketListener export release
libsysutils: Add iovec/runOnEachSocket
liblog: support struct logger_event_v2 format
liblog: update timestamp on NOTICE file
libcutils: resolve warning in iosched_policy.c
liblog: Add const pedantics
logcat: Add -T flag (-t w/o assumption of -d)
logcat: Add logcat test suite
liblog: Add cpu utilization test
liblog: Add liblog test suite
debuggerd: Support newline split in log messages
liblog: deprecate export LOGGER ioctl definitions
liblog: deprecate export of LOGGER_LOG_* defines
liblog: Add README
liblog: resolve build warning messages
liblog: high CPU usage from logcat
liblog: fix build again
liblog: drop use of sys/cdefs.h
liblog: git_master@964770 build problem
logcat: Incorporate liblog reading API
debuggerd: Incorporate liblog reading API
liblog: Interface to support abstracting log read
adb: deprecate legacy log service interface
adb: regression from Move list.c to inlines
liblog: whitespace cleanup
libcutils: bug str_parms.c:str_parms_get_float().
libcutils: UNUSED argument warnings
libsysutils: Get rid of warnings
libcutils: Move list.c to inlines on list.h
Move the liblog headers to log/ instead of cutils/ to complete
the separation of libcutils and liblog. cutils/log.h still
exists and includes log/log.h in order to support the many existing
modules that use cutils/log.h.
Change-Id: I2758c9f4aedcb809ca7ba8383d0f55041dd44345
File descriptors remain open across an exec unless FD_CLOEXEC is
set. Add O_CLOEXEC to the open() call to prevent file descriptor
leakage.
In particular, the following program will eventually run out of
file descriptors:
int main(int argc, char **argv) {
printf("===== entering main =====\n");
ALOGW("entering main");
system("ls -l /proc/self/fd/");
execv(argv[0], argv);
printf("exec failed\n");
return -1;
}
Change-Id: I5be43ab3b9f82a05f242b1f586454c50568af388
Changing __android_log_write so the tag warns users relying on the
tag to direct log output to the radio buffer to instead use Rlog or RLOG.
Change-Id: I04b7bb5e620c1ab22b9b495382a252b539947e28
Invoking LOG_ASSERT, LOG_ALWAYS_FATAL or LOG_ALWAYS_FATAL_IF variadic macros
without the printf format string arg caused compilation errors because the
variable arg list (__VA_ARGS__) was eventually passed to
__android_log_assert() func in place of a required parameter. This error
only occured in debug builds because LOG_ASSERT() is a no-op in release
builds. This change allows debug builds to succeed.
Change-Id: I7e7b7de3e501133468ce083e0e0d6e699dd59667
Signed-off-by: Chris Pearson <christopherx.c.pearson@intel.com>
