| 870dd82f0a1190b1890ca3e592f68bc5bbece627 |
|
18-May-2018 |
Florian Mayer <fmayer@google.com> |
Correctly handle EINTR.
This fixes a bug that made the CPU reader loop exit when the program
gets send SIGSTOP.
Cherry pick of http://aosp/689181.
Bug: 79969427
Merged-In: I09ae42adb5b211e0eae73364dcef1b4d33762ed8
Change-Id: I0f76ba9d3aba77a607175fc1f31ec7edcfa2afe6
/external/perfetto/src/ftrace_reader/cpu_reader.h
|
| 3cbb10a30c000fcd2cd4658dd215778422c7bbb1 |
|
10-Apr-2018 |
Primiano Tucci <primiano@google.com> |
Remove warning suppressions and fix many casting bugs
This change removes most warning suppressions, in particular
from production code. In the past we inherit a bunch of -Wno-xxx
suppressions required to build gtest and libprotobuf headers.
Doing so, however, caused the suppressions to propagate back to
the translation units that were including any protobuf header
or any auto-generated .pb.h stub.
This change moves the gtest and probobuf header to be a
system include (-isystem vs -I). Doing so implicitly blacklists
any compiler warning on the headers.
This CL then re-enables warnings and deals with the fall out of
fixes that came out of this.
Bug: 77316877
Test: pefetto_unittests / perfetto_integrationtests
Change-Id: I3a01852ebf7d0b9bf19658ddf117209d129c70be
/external/perfetto/src/ftrace_reader/cpu_reader.h
|
| 727d1cb433bf06f85a69312a5c3fc1610b7cb1b6 |
|
03-Apr-2018 |
Florian Mayer <fmayer@google.com> |
Remove dead code
Bug: 76169489
Change-Id: I8b41f27ff8e41fb90ead747c9d0aa6287d11b30a
/external/perfetto/src/ftrace_reader/cpu_reader.h
|
| 808d6df6d555faf56d2ad537d5b08dfaf3e72dd9 |
|
31-Mar-2018 |
Primiano Tucci <primiano@google.com> |
Support building most targets on mac
Rationale: takes very little efforts, increases unittest-ability and
avoid having to think/#ifdef when using the watchdog.
Bug: 76169489
Change-Id: Ife14e2ec77d21b17fa48a48f0629c4026a60bf2c
/external/perfetto/src/ftrace_reader/cpu_reader.h
|
| b8e235b7ab322672c2646050fca0c6c7cf22cef4 |
|
22-Mar-2018 |
Florian Mayer <fmayer@google.com> |
Ignore traced_probe's own I/O activity.
Bug: 73947457
Change-Id: I15c24cccc20c5350196bbc10dbcd3617bea6aaae
/external/perfetto/src/ftrace_reader/cpu_reader.h
|
| 2c3ac7fd218a9de6e842a68338750447be369c76 |
|
22-Mar-2018 |
Hector Dearman <hjd@google.com> |
ftrace_reader: Adjust DCHECK on Inode collection
In theory each ftrace inode is preceded by an block device id, this
adds a DCHECK to test this.
Bug: 73873362
Change-Id: Ifb7ab9845a6ae1490e9b94da4259e51ca74cb7ce
/external/perfetto/src/ftrace_reader/cpu_reader.h
|
| cd3b6d37da21315216af2044bf4f34d57882f74c |
|
20-Mar-2018 |
Hector Dearman <hjd@google.com> |
traced_probes: Add data_source_types.h
We need a place to put types that src/ftrace_reader, src/traced/probes,
and, future producer impls (e.g. src/memory) can all depend on.
include/perfetto/traced is the pragmatic place to avoid intoducing an
extra build target.
Bug: 73625480
Change-Id: I8a053411a81d3d3d877e0a7325e89fd524aa373f
/external/perfetto/src/ftrace_reader/cpu_reader.h
|
| 8ce30879c9384dfb870d7496fdff62e2e0f5db50 |
|
19-Mar-2018 |
Anna Zappone <azappone@google.com> |
ftrace_reader: Update for uint64 block device ids
Switch previous code for plumbing uint32 device ids to translate to uint64 and add
translation strategies for uint64 device ids.
Add static_casts for inode numbers as well.
Bug: 73873362
Bug: 73625480
Change-Id: Id76ef4759739e78994c7b5bd2a9f85dc7d0890e3
/external/perfetto/src/ftrace_reader/cpu_reader.h
|
| ba3eeaeb6068b9426d891041b0ab1c827bd63d02 |
|
12-Mar-2018 |
Florian Mayer <fmayer@google.com> |
Make relative includes absolute.
Change-Id: I9d3eeb67d6b39b05dd85d11746497791ac2665f9
/external/perfetto/src/ftrace_reader/cpu_reader.h
|
| 4ea73c08d0f22519a9b959adc9c41b351f494da6 |
|
09-Mar-2018 |
Anna Zappone <azappone@google.com> |
ftrace_reader: Plumb block_device_ids via metadata
Add the block device id with each inode number to metadata.
Use the stored device id in probes_producer to ensure duplicate
inode numbers are matched correctly.
All our current format files have had dev id and then inode immediately after.
When an inode number is seen, get the most recent dev id and add the pair
to metadata.
Bug: 73873362
Bug: 73625480
Change-Id: I1eeae21a840bbfa1be5cda95bd7a0010706c8e31
/external/perfetto/src/ftrace_reader/cpu_reader.h
|
| 22e4b396fcfd44ef7f70c5f2ecb8f665eee97742 |
|
08-Mar-2018 |
Florian Mayer <fmayer@google.com> |
Run clang-tidy.
Change-Id: I4b173963d14a627daf319427ba2752fca482ada4
/external/perfetto/src/ftrace_reader/cpu_reader.h
|
| f19e7c57e810e39f28f25921a6e7f7630ec3c7b1 |
|
06-Mar-2018 |
Hector Dearman <hjd@google.com> |
ftrace_reader: Plumb pids via metadata
Change-Id: I6731ee57d1b20d86700dce97653d0425d51c00c8
/external/perfetto/src/ftrace_reader/cpu_reader.h
|
| c8488034fcc8ff2fc09c4991c2b997f4702ee1cc |
|
02-Mar-2018 |
Hector Dearman <hjd@google.com> |
ftrace_reader: Plumb metadata from CpuReader
We want to neatly communicate some information (such as
inode and pids) seen in the raw ftrace buffer to
ProbesProducer where it can be used to gather aditional
information on the inode/process.
This CL plumbs an additional argument (FtraceMetadata)
into the OnBundleComplete callback. FtraceMetadata is a
container for 'out-of-band' information about the
events in the bundle. For now this is just the number
of overwritten events and the set of observed inodes.
On the ProbesProducer side we wire this argument up to a
log statement.
Change-Id: I05d38c2db7f7f75a320f86de24182cc2bc3e6821
/external/perfetto/src/ftrace_reader/cpu_reader.h
|
| 1f14b621554e89efe0294dbdd83d7e6ac823790c |
|
27-Feb-2018 |
Hector Dearman <hjd@google.com> |
ftrace_reader: Plumb overwrite from page header
The ftrace page header includes a field named "overwrite" which is in some way
related to the number of events overwritten. This plumbs that value to the
FtraceProtoBundle proto. Specifically we add a new field 'overwrite' which is the
total of all overwrite fields from the headers of all kernel ftrace pages we parsed
into this FtraceEventBundle. The value is zero if no overwriting occurred, a number
larger than zero if some overwriting occurred.
While we are here we also:
- Set cpu only once per ftrace_bundle.
- Remove cpu parameter from ParsePage.
- Add metadata out parameter to ParsePage for overwrite, inodes, pids of interest etc.
Bug: 72045210
Change-Id: I985ff880c88101317a2120b43d420fcd8272b113
/external/perfetto/src/ftrace_reader/cpu_reader.h
|
| 5a407c12d527440998c4c22ab490bfe6ab5a178a |
|
22-Feb-2018 |
Anna Zappone <azappone@google.com> |
ftrace_reader: Add initial inode to filename code
Add inode to filename mapping by walking through filesystem and saving filenames that match inode numbers from events.
Bug: 73625480
Change-Id: I2decec1f18eefee45f3b8d624c69263d2750b59d
/external/perfetto/src/ftrace_reader/cpu_reader.h
|
| e1e56b63c369ffe22dc30fb2eeaac1556266d3ec |
|
21-Feb-2018 |
Hector Dearman <hjd@google.com> |
protozero: Remove protozero from the name of files
Second part of:
https://android-review.googlesource.com/c/platform/external/perfetto/+/606697
Script:
mv ./src/protozero/protozero_message_unittest.cc ./src/protozero/message_unittest.cc
mv ./src/protozero/protozero_message.cc ./src/protozero/message.cc
mv ./src/protozero/protozero_message_handle.cc ./src/protozero/message_handle.cc
mv ./src/protozero/protozero_message_handle_unittest.cc ./src/protozero/message_handle_unittest.cc
mv ./include/perfetto/protozero/protozero_message.h ./include/perfetto/protozero/message.h
mv ./include/perfetto/protozero/protozero_message_handle.h ./include/perfetto/protozero/message_handle.h
find . -type f -name '*.h' | grep -v buildtools | grep -v out | while read f; do sed -i 's/protozero_message\.h/message\.h/g' $f; done
find . -type f -name '*.h' | grep -v buildtools | grep -v out | while read f; do sed -i 's/protozero_message_handle\.h/message_handle\.h/g' $f; done
find . -type f -name '*.cc' | grep -v buildtools | grep -v out | while read f; do sed -i 's/protozero_message\.h/message\.h/g' $f; done
find . -type f -name '*.cc' | grep -v buildtools | grep -v out | while read f; do sed -i 's/protozero_message_handle\.h/message_handle\.h/g' $f; done
find . -type f -name '*.gn' | grep -v buildtools | grep -v out | while read f; do sed -i 's/protozero_message/message/g' $f; done
find . -type f -name '*.gn' | grep -v buildtools | grep -v out | while read f; do sed -i 's/protozero_message_handle/message_handle/g' $f; done
Bug: 70386249
Change-Id: Ifdeda80db539070ec2f07466de32f855a7cc85cc
/external/perfetto/src/ftrace_reader/cpu_reader.h
|
| aaa4c1997f110dee086ee93b83bfc7ac11bfc102 |
|
19-Feb-2018 |
Hector Dearman <hjd@google.com> |
Rename protozero::ProtozeroX -> protozero::X
Commands:
find . -type f -name '*.cc' | grep -v buildtools | grep -v out | while read f; do sed -i 's/ProtoZeroMessage/Message/g' $f; done
find . -type f -name '*.h' | grep -v buildtools | grep -v out | while read f; do sed -i 's/ProtoZeroMessage/Message/g' $f; done
find . -type f -name '*.cc' | grep -v buildtools | grep -v out | while read f; do sed -i 's/ProtoZeroMessageHandle/MessageHandle/g' $f; done
find . -type f -name '*.h' | grep -v buildtools | grep -v out | while read f; do sed -i 's/ProtoZeroMessageHandle/MessageHandle/g' $f; done
Bug: 70386249
Change-Id: Iec8ef0e67a522a7b8988ce788bc63ba2775422fc
/external/perfetto/src/ftrace_reader/cpu_reader.h
|
| 74b1f2ac0ad97dd0470d17cc2d220875f739f3e1 |
|
16-Feb-2018 |
Anna Zappone <azappone@google.com> |
Save inode numbers in cpu_reader to get inode to filename mappings
Change-Id: Ic6b38010871fb1b6f126f0176105528dbc939a3f
/external/perfetto/src/ftrace_reader/cpu_reader.h
|
| 247110bb9e52876f0412ece815c81ca055a3a898 |
|
14-Feb-2018 |
Sami Kyostila <skyostil@google.com> |
Improve ftrace scheduling
This patch implements the following ftrace scheduling algorithm:
main thread [drain] [unblock]
/: |
post .-----' : |
/ : v
worker #0 [splice ...] [wakeup] [block ............] [splice]
:
worker #1 [splice ...] [wakeup] [block ........] [splice]
:
worker #2 [splice ..........................................]
:
:
drain period (100ms)
In other words, the splice(2) system call is used to move data from
the raw kernel ftrace pipe into an intermediate pipe at a page
granularity. This call allows every per-cpu worker to sleep until there
is at least one page of data available.
When a worker wakes up, it will attempt to move as many pages as
possible to its staging pipe (up to 64K, depending on the
system's pipe buffer size) in a non-blocking way. After this, it
will notify the main thread that data is available. This notification
will block the calling worker until the main thread has drained the
data.
When at least one worker has woken up, we schedule a drain operation
on the main thread for the next drain period (every 100ms by default).
The drain operation parses ftrace data from the staging pipes of
every worker having pending data. After this, each waiting worker is
allowed to issue another call to splice(), restarting the cycle.
Change-Id: I9657a5d50cd936f190bc57f2c9778019cb7419ab
/external/perfetto/src/ftrace_reader/cpu_reader.h
|
| 680a28b924d212f5e05136a2ea1bdd77ea87e8de |
|
11-Jan-2018 |
Hector Dearman <hjd@google.com> |
ftrace_reader: Backport reading raw pipes on timer
- Backport the changes from the demo to read the raw pipe on a timer
rather than when there is data in the pipe.
- Fix case where if we disable/enable quickly we end up with two
periodic tasks for each CPU.
- Add tests.
Change-Id: I482c12323c78495a072c568f3b63f33acfe179ed
/external/perfetto/src/ftrace_reader/cpu_reader.h
|
| 77af1339a9fb654ed0e54c12c0448073e54d3279 |
|
15-Dec-2017 |
Hector Dearman <hjd@google.com> |
ftrace_reader: Add dynamic parser (part 4)
In part the forth we:
- Refactor common fields to use the new generic parsing method.
- Rename Char16 to FixedCString and make it work on strings of
any length.
- Extract the field merging logic.
Bug: 70328826
Change-Id: I116ac4c98d07fe7be7ba458c0524c3bab07bba87
/external/perfetto/src/ftrace_reader/cpu_reader.h
|
| 2ddcc7b1f5203ff60f71f761e33e1f7adee2aa9f |
|
13-Dec-2017 |
Hector Dearman <hjd@google.com> |
ftrace_reader: Add dynamic parser (part 3)
In the latest installment of this CL we:
- Add a new 'joint' enum with all valid combinations of ftrace and
proto types.
- Keep track of the total size of each event at parse time allowing
us to make only one bounds check per event.
- Modify cpu_reader to use these two new features.
- Add a new unit test to cover the increasingly complicated
ProtoTranslationTableUnittest::Create method.
Bug: 70328826
Change-Id: Ice4c110253ef94d2516762566c32b1e46b988803
/external/perfetto/src/ftrace_reader/cpu_reader.h
|
| cc6dcb0c65320bcbfef60507e2352ec8da53adc7 |
|
12-Dec-2017 |
Hector Dearman <hjd@google.com> |
ftrace_reader: Add dynamic parser (part 2)
Rather than handcoding a parser for each event use the
ProtoTranslationTable to dynamically read each field of the event.
Bug: 70328826
Change-Id: I154904c7e805f281706e21c0f0aef0bb2936926d
/external/perfetto/src/ftrace_reader/cpu_reader.h
|
| 6fd5e5cd38e334e3bb6e63afb116be19c9aa4a79 |
|
06-Dec-2017 |
Hector Dearman <hjd@google.com> |
Remove size argument from ParsePage
The kernel always gives us a full page when reading trace_pipe_raw.
We shouldn't pretend the code will work in the untested case where
we get less than a full page especially since that will probably
just hide a deeper problem if the kernel starts giving us less
than a page.
Change-Id: I2a8793aaadb3717c04b13fa5d9b0db19bc11ff7c
/external/perfetto/src/ftrace_reader/cpu_reader.h
|
| 5add654e12ebc80bd28c707fe25bf814987b6f38 |
|
06-Dec-2017 |
Hector Dearman <hjd@google.com> |
Improve CpuReader unittests
It's too painful and error prone to construct the binary format by hand
for tests. Instead we should record real examples then compare the result of
parsing them to golden protos. This does the first part allowing us to
include a representation of the binary data in the test file.
Also add minimal set of format files for testing with when we don't want
to load several hundred files.
Change-Id: Ic75d4bca6b12107e2d4df814eb912da259044f0c
/external/perfetto/src/ftrace_reader/cpu_reader.h
|
| 4f9b6d7b7b2ad7b86e1577a18dc6e2df91b98164 |
|
05-Dec-2017 |
Primiano Tucci <primiano@google.com> |
Pivot source files into top-level src,include and unify test target
See discussion in go/perfetto-build-files .
This is to disambiguate things like
#include "base/logging.h"
when in the chrome tree.
Also this CL unifies the test targets into two monolithic targets:
perfetto_tests and perfetto_benchmarks. This is to avoid ending
up with confusing binary names in the chrome tree (e.g.,
ipc_unittests)
Bug: 68710794
Change-Id: I1768e15b661406052b2be060d7aab0f1e7443a98
/external/perfetto/src/ftrace_reader/cpu_reader.h
|