metrics_daemon should never get a stale answer on whether or not the
metrics are enabled.
This is important as metrics_daemon will be the "source of truth" for
other components.
BUG: 24386281
TEST: unit tests.
Change-Id: I573568abe5d1b840683cede2fdf32cdae028a81a
"Out of date" is only probably true. You might equally well have an older
client talking to a newer server. So tell the truth and include the actual
version numbers.
Change-Id: I821de88f5baf65bf2623363129c60c496b407bff
Discovered that we had a few libc hotspots. Adjust code to generally
reduce or nullify the number of calls to malloc, free, strlen,
strcmp, strncmp, memcmp & strncasecmp. Total gain looks to be about
3% of logd's processing time. malloc still contributes to 3%, but all
others are now total 0.5%.
Bug: 23685592
Change-Id: Ife721121667969260cdb8b055524ae90f5911278
Sending the SEND/DATA/DONE packets all in one write makes small file and
symbolic link transfer orders of magnitude faster. Transferring all the
symbolic links in /system/bin or all the cacerts in /system/etc goes down
from about 6s in each case to less than 0.5s.
Change-Id: Ia1706cef25535d0b685d3744241a0557a52b22ff
strcmp was 1/10 #2 behind find_property in __android_log_level(),
now virtually eliminated from performance profile.
Bug: 23685592
Change-Id: I3978886193af77e489c6d1728d6a26b7f53f8f2f
Instead of reporting the metrics over both a long and a short period,
collect and report only over a short period. This makes the code simpler
and the metrics easier to understand.
Also move the collection out of metrics_daemon and into a separate
collector to make it simpler to understand.
BUG: 24464945
Change-Id: I17e52536aaa75321a5e34f42ed488545c2c3efde
FileMaps should be movable, thereby not requiring them to be only used
with a unique_ptr as they currently are.
Change-Id: I0fb8013bf398a2ced5420d85ba888c2a7fc5a496
* fdevent_{set,add,del}()
* CHECK() that the fdevent is FDE_ACTIVE, otherwise the caller would
be waiting for events that will never arrive.
* Remove ~ from "fde->events &= ~events" to keep desired bits instead
of turning off desired bits.
* fdevent_call_fdfunc()
* CHECK(FDE_PENDING) since it should always be true if the fdevent was
on the pending list, or if fdevent_subproc_event_func() is faking
things. The goal is to try to avoid losing events.
Change-Id: I979c2fffa0b3d6b635488cde11dc544691193018
Signed-off-by: Spencer Low <CompareAndSwap@gmail.com>
Read in the ro.build.id property for version, and
ro.product.product_id for the Product ID.
Bug: 22874192
Change-Id: I9a3fbf375d49d04fc7bf6700c5987cb9e435c318
ddmlib does not use the ADB client, but instead connects directly to
the adb server. This breaks some of the assumptions I previously made
when enabling the shell protocol.
To fix this, the adb server now defaults to no protocol for the
standalone command, and the shell protocol must be explicitly requested
by the client. For example:
shell:echo foo -- no shell protocol
shell,v2:echo foo -- shell protocol
As long as I was touching the shell service arguments I also changed
them to no longer duplicate the command-line arguments. This allows
more flexibility to change the adb client CLI if necessary and makes
the code more readable.
Bug: http://b/24148636
Change-Id: I28d5ae578cf18cbe79347dc89cea1750ff4571a8
It is possible that the adb server on host has many sockets in
CLOSE_WAIT state. To prevent socket leak, always enable POLLRDHUP
in fdevent.cpp to detect sockets in CLOSE_WAIT state.
Update LocalSocketTest unit tests:
Change half_close_with_packet to read_from_closing_socket, as reading
from a SHUT_WR socket is not needed in adb.
Change close_with_no_events_installed to close_socket_in_CLOSE_WAIT_state,
as the latter is more close to the real situation in use.
Bug: 23314034
Change-Id: Ice4f4036624e5584eab6ba5848e7f169c92f037f
* changes:
logd: klogd deal with nuls in dmesg
logd: log_strtok_r deal with nuls
liblog: logprint: printable nul
logd: klogd: sniff for time correction on Mediatek
Clang 3.8 warns against this now, and it is tripping up part of a test.
We suppress the warning, since we are trying to actually test that this
behavior is properly handled.
Change-Id: I8705900572e1a6704bbdc001fece3e2d16c7077c
Switch to using string and length in all transactions, treating
trailing nuls the same as spaces.
ToDo: change dumpstate (bugreport) to use logcat -b printable _regardless_
Bug: 23517551
Change-Id: I42162365e6bf8ed79d356e7b689a673902116fdb
Rename to log_strntok_r and change from dealing with strings
to dealing with a string and an associated length.
Bug: 23517551
Change-Id: Ia72f1305a53f55eeef9861ac378fb8205fd2378e
Need some more flexibility when parsing kernel messages
cluttered with extra fluff. This is the minimal relaxation
we can do to the rules to ensure that we work on all
possible devices and kernels when sniffing for time
correction information.
We want to minimize any future maintenance, keep in mind
klogd is a "userdebug" or "eng" feature and is disabled
in "user" builds. Manage expectations.
Bug: 23517551
Change-Id: I026d074e14fb2550e728683e85a973bd87e78a9c
Instead of using socket(..., 0), pass IPPROTO_TCP or IPPROTO_UDP. Using
zero wasn't buying us anything and was different than popular apps like
Chrome. We should stick to what everyone else does so that we don't go
down different code-paths and potentially hit Winsock service provider
issues that everyone else is (accidentally) avoiding.
Also CHECK() if send() returns an erroneous value as described in the
Chromium source.
Also add comment about socket buffer sizing and Windows Vista.
Change-Id: I63db8f6de352fe1e9525cbc9cfc040eb02a4b9cd
Signed-off-by: Spencer Low <CompareAndSwap@gmail.com>
