2012-10-31 06:02:06 +08:00
|
|
|
The struct perf_event_attr test (attr tests) support
|
|
|
|
|
====================================================
|
|
|
|
|
This testing support is embedded into perf directly and is governed
|
|
|
|
|
by the PERF_TEST_ATTR environment variable and hook inside the
|
|
|
|
|
sys_perf_event_open function.
|
|
|
|
|
|
|
|
|
|
The general idea is to store 'struct perf_event_attr' details for
|
|
|
|
|
each event created within single perf command. Each event details
|
|
|
|
|
are stored into separate text file. Once perf command is finished
|
|
|
|
|
these files are checked for values we expect for command.
|
|
|
|
|
|
|
|
|
|
The attr tests consist of following parts:
|
|
|
|
|
|
|
|
|
|
tests/attr.c
|
|
|
|
|
------------
|
|
|
|
|
This is the sys_perf_event_open hook implementation. The hook
|
|
|
|
|
is triggered when the PERF_TEST_ATTR environment variable is
|
|
|
|
|
defined. It must contain name of existing directory with access
|
|
|
|
|
and write permissions.
|
|
|
|
|
|
|
|
|
|
For each sys_perf_event_open call event details are stored in
|
|
|
|
|
separate file. Besides 'struct perf_event_attr' values we also
|
|
|
|
|
store 'fd' and 'group_fd' values to allow checking for groups.
|
|
|
|
|
|
|
|
|
|
tests/attr.py
|
|
|
|
|
-------------
|
|
|
|
|
This is the python script that does all the hard work. It reads
|
|
|
|
|
the test definition, executes it and checks results.
|
|
|
|
|
|
|
|
|
|
tests/attr/
|
|
|
|
|
-----------
|
|
|
|
|
Directory containing all attr test definitions.
|
|
|
|
|
Following tests are defined (with perf commands):
|
|
|
|
|
|
|
|
|
|
perf record kill (test-record-basic)
|
|
|
|
|
perf record -b kill (test-record-branch-any)
|
|
|
|
|
perf record -j any kill (test-record-branch-filter-any)
|
|
|
|
|
perf record -j any_call kill (test-record-branch-filter-any_call)
|
|
|
|
|
perf record -j any_ret kill (test-record-branch-filter-any_ret)
|
|
|
|
|
perf record -j hv kill (test-record-branch-filter-hv)
|
|
|
|
|
perf record -j ind_call kill (test-record-branch-filter-ind_call)
|
|
|
|
|
perf record -j k kill (test-record-branch-filter-k)
|
|
|
|
|
perf record -j u kill (test-record-branch-filter-u)
|
|
|
|
|
perf record -c 123 kill (test-record-count)
|
|
|
|
|
perf record -d kill (test-record-data)
|
|
|
|
|
perf record -F 100 kill (test-record-freq)
|
perf test: Update command line callchain attribute tests
The "struct perf_event_attr setup" entry in 'perf test' is in fact a
series of tests that will exec the tools, passing different sets of
command line arguments to then intercept the sys_perf_event_open
syscall, in user space, to check that the perf_event_attr->sample_type
and other feature request bits are setup as expected.
We recently restored the callchain requesting command line argument, -g,
to not require a parameter ("dwarf" or "fp"), instead using a default
("fp" for now) and making the long option variant, --call-chain, be the
one to be used when a different callchain collection method is
preferred.
The "struct perf_event_attr setup" test failed because we forgot to
update the tests involving callchains, not switching from, '-g dwarf' to
'--call-chain dwarf', making 'perf test' detect it:
[root@sandy ~]# perf test -v 13
13: struct perf_event_attr setup :
--- start ---
running '/home/acme/libexec/perf-core/tests/attr/test-record-basic'
running '/home/acme/libexec/perf-core/tests/attr/test-record-branch-any'
<SNIP>
running '/home/acme/libexec/perf-core/tests/attr/test-record-graph-default'
running '/home/acme/libexec/perf-core/tests/attr/test-record-graph-dwarf'
expected sample_type=12583, got 295
expected exclude_callchain_user=1, got 0
expected sample_stack_user=8192, got 0
FAILED '/home/acme/libexec/perf-core/tests/attr/test-record-graph-dwarf' - match failure
---- end ----
struct perf_event_attr setup: FAILED!
[root@sandy ~]#
Fix all of them now to use --call-chain when explicitely specifying a
method.
There is still work to do, as '-g fp', for instance, passed without
problems.
In that case 'perf test' saw no problems as the intercepted syscall got
the bits as expected, i.e. the default is 'fp', but the fact that 'fp'
may be an existing program and the specified workload would then be
passed as a parameter to it is an usability problem that needs fixing.
Next merge window tho.
Acked-by: Jiri Olsa <jolsa@redhat.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-jr3oq1k5iywnp7vvqlslzydm@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2013-10-31 20:46:59 +08:00
|
|
|
perf record -g kill (test-record-graph-default)
|
|
|
|
|
perf record --call-graph dwarf kill (test-record-graph-dwarf)
|
|
|
|
|
perf record --call-graph fp kill (test-record-graph-fp)
|
2012-10-31 06:02:06 +08:00
|
|
|
perf record --group -e cycles,instructions kill (test-record-group)
|
|
|
|
|
perf record -e '{cycles,instructions}' kill (test-record-group1)
|
|
|
|
|
perf record -D kill (test-record-no-delay)
|
|
|
|
|
perf record -i kill (test-record-no-inherit)
|
|
|
|
|
perf record -n kill (test-record-no-samples)
|
|
|
|
|
perf record -c 100 -P kill (test-record-period)
|
|
|
|
|
perf record -R kill (test-record-raw)
|
|
|
|
|
perf stat -e cycles kill (test-stat-basic)
|
|
|
|
|
perf stat kill (test-stat-default)
|
|
|
|
|
perf stat -d kill (test-stat-detailed-1)
|
|
|
|
|
perf stat -dd kill (test-stat-detailed-2)
|
|
|
|
|
perf stat -ddd kill (test-stat-detailed-3)
|
|
|
|
|
perf stat --group -e cycles,instructions kill (test-stat-group)
|
|
|
|
|
perf stat -e '{cycles,instructions}' kill (test-stat-group1)
|
|
|
|
|
perf stat -i -e cycles kill (test-stat-no-inherit)
|