[BUGFIX PATCH v3 0/4] selftests/ftrace: Fix ftracetest testcases for non-function tracer case
Hi,
Here is the 3rd version of patches to fix some issues which happens on the kernel with CONFIG_FUNCTION_TRACER=n or CONFIG_DYNAMIC_FTRACE=n.
In this version and v2, I updated the descriptions of the first 2 patches according to Steve's comment, added Steve's Reviewed-by to the 3rd patch, and added the 4th patch which was newly found.
Thank you,
---
Masami Hiramatsu (4): selftests/ftrace: Fix to check the existence of set_ftrace_filter selftests/ftrace: Fix ftrace test cases to check unsupported selftests/ftrace: Do not to use absolute debugfs path selftests/ftrace: Fix multiple kprobe testcase
.../ftrace/test.d/ftrace/func-filter-stacktrace.tc | 2 ++ .../selftests/ftrace/test.d/ftrace/func_cpumask.tc | 5 +++++ tools/testing/selftests/ftrace/test.d/functions | 4 +++- .../ftrace/test.d/kprobe/multiple_kprobes.tc | 6 +++--- .../inter-event/trigger-action-hist-xfail.tc | 4 ++-- .../inter-event/trigger-onchange-action-hist.tc | 2 +- .../inter-event/trigger-snapshot-action-hist.tc | 4 ++-- 7 files changed, 18 insertions(+), 9 deletions(-)
-- Masami Hiramatsu (Linaro) mhiramat@kernel.org
If we run ftracetest on the kernel with CONFIG_DYNAMIC_FTRACE=n, there is no set_ftrace_filter and all test cases are failed, because reset_ftrace_filter returns an error. Let's check whether set_ftrace_filter exists and remove redundant set_ftrace_filter from initialize_ftrace().
Signed-off-by: Masami Hiramatsu mhiramat@kernel.org --- tools/testing/selftests/ftrace/test.d/functions | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/tools/testing/selftests/ftrace/test.d/functions b/tools/testing/selftests/ftrace/test.d/functions index 86986c4bba54..19d288cdf336 100644 --- a/tools/testing/selftests/ftrace/test.d/functions +++ b/tools/testing/selftests/ftrace/test.d/functions @@ -46,6 +46,9 @@ reset_events_filter() { # reset all current setting filters }
reset_ftrace_filter() { # reset all triggers in set_ftrace_filter + if [ ! -f set_ftrace_filter ]; then + return 0 + fi echo > set_ftrace_filter grep -v '^#' set_ftrace_filter | while read t; do tr=`echo $t | cut -d: -f2` @@ -93,7 +96,6 @@ initialize_ftrace() { # Reset ftrace to initial-state disable_events [ -f set_event_pid ] && echo > set_event_pid [ -f set_ftrace_pid ] && echo > set_ftrace_pid - [ -f set_ftrace_filter ] && echo | tee set_ftrace_* [ -f set_graph_function ] && echo | tee set_graph_* [ -f stack_trace_filter ] && echo > stack_trace_filter [ -f kprobe_events ] && echo > kprobe_events
On Mon, 25 Nov 2019 15:57:00 +0900 Masami Hiramatsu mhiramat@kernel.org wrote:
If we run ftracetest on the kernel with CONFIG_DYNAMIC_FTRACE=n, there is no set_ftrace_filter and all test cases are failed, because reset_ftrace_filter returns an error. Let's check whether set_ftrace_filter exists and remove redundant set_ftrace_filter from initialize_ftrace().
Signed-off-by: Masami Hiramatsu mhiramat@kernel.org
tools/testing/selftests/ftrace/test.d/functions | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/tools/testing/selftests/ftrace/test.d/functions b/tools/testing/selftests/ftrace/test.d/functions index 86986c4bba54..19d288cdf336 100644 --- a/tools/testing/selftests/ftrace/test.d/functions +++ b/tools/testing/selftests/ftrace/test.d/functions @@ -46,6 +46,9 @@ reset_events_filter() { # reset all current setting filters } reset_ftrace_filter() { # reset all triggers in set_ftrace_filter
- if [ ! -f set_ftrace_filter ]; then
return 0- fi echo > set_ftrace_filter grep -v '^#' set_ftrace_filter | while read t; do tr=`echo $t | cut -d: -f2`
@@ -93,7 +96,6 @@ initialize_ftrace() { # Reset ftrace to initial-state disable_events [ -f set_event_pid ] && echo > set_event_pid [ -f set_ftrace_pid ] && echo > set_ftrace_pid
- [ -f set_ftrace_filter ] && echo | tee set_ftrace_*
The above should be changed to:
[ -f set_ftrace_notrace ] && echo > set_ftrace_notrace
-- Steve
[ -f set_graph_function ] && echo | tee set_graph_* [ -f stack_trace_filter ] && echo > stack_trace_filter [ -f kprobe_events ] && echo > kprobe_events
On Mon, 25 Nov 2019 09:44:45 -0500 Steven Rostedt rostedt@goodmis.org wrote:
On Mon, 25 Nov 2019 15:57:00 +0900 Masami Hiramatsu mhiramat@kernel.org wrote:
If we run ftracetest on the kernel with CONFIG_DYNAMIC_FTRACE=n, there is no set_ftrace_filter and all test cases are failed, because reset_ftrace_filter returns an error. Let's check whether set_ftrace_filter exists and remove redundant set_ftrace_filter from initialize_ftrace().
Signed-off-by: Masami Hiramatsu mhiramat@kernel.org
tools/testing/selftests/ftrace/test.d/functions | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/tools/testing/selftests/ftrace/test.d/functions b/tools/testing/selftests/ftrace/test.d/functions index 86986c4bba54..19d288cdf336 100644 --- a/tools/testing/selftests/ftrace/test.d/functions +++ b/tools/testing/selftests/ftrace/test.d/functions @@ -46,6 +46,9 @@ reset_events_filter() { # reset all current setting filters } reset_ftrace_filter() { # reset all triggers in set_ftrace_filter
- if [ ! -f set_ftrace_filter ]; then
return 0- fi echo > set_ftrace_filter grep -v '^#' set_ftrace_filter | while read t; do tr=`echo $t | cut -d: -f2`
@@ -93,7 +96,6 @@ initialize_ftrace() { # Reset ftrace to initial-state disable_events [ -f set_event_pid ] && echo > set_event_pid [ -f set_ftrace_pid ] && echo > set_ftrace_pid
- [ -f set_ftrace_filter ] && echo | tee set_ftrace_*
The above should be changed to:
[ -f set_ftrace_notrace ] && echo > set_ftrace_notrace
Ah, good point! I think that should be done by another patch, since it will improve ftracetest.
Thank you,
-- Steve
[ -f set_graph_function ] && echo | tee set_graph_* [ -f stack_trace_filter ] && echo > stack_trace_filter [ -f kprobe_events ] && echo > kprobe_events
On Tue, 26 Nov 2019 09:13:45 +0900 Masami Hiramatsu mhiramat@kernel.org wrote:
@@ -93,7 +96,6 @@ initialize_ftrace() { # Reset ftrace to initial-state disable_events [ -f set_event_pid ] && echo > set_event_pid [ -f set_ftrace_pid ] && echo > set_ftrace_pid
- [ -f set_ftrace_filter ] && echo | tee set_ftrace_*
The above should be changed to:
[ -f set_ftrace_notrace ] && echo > set_ftrace_notrace
Ah, good point! I think that should be done by another patch, since it will improve ftracetest.
No, it belongs in this patch, because you are removing:
[ -f set_ftrace_filter ] && echo | tee set_ftrace_*
which is equivalent to:
[ -f set_ftrace_filter ] && echo > set_ftrace_filter [ -f set_ftrace_filter ] && echo > set_ftrace_notrace
as the "tee set_ftrace_*" covers both.
Without this change, this patch removes the update to set_ftrace_notrace.
-- Steve
On Mon, 25 Nov 2019 20:54:29 -0500 Steven Rostedt rostedt@goodmis.org wrote:
On Tue, 26 Nov 2019 09:13:45 +0900 Masami Hiramatsu mhiramat@kernel.org wrote:
@@ -93,7 +96,6 @@ initialize_ftrace() { # Reset ftrace to initial-state disable_events [ -f set_event_pid ] && echo > set_event_pid [ -f set_ftrace_pid ] && echo > set_ftrace_pid
- [ -f set_ftrace_filter ] && echo | tee set_ftrace_*
The above should be changed to:
[ -f set_ftrace_notrace ] && echo > set_ftrace_notrace
Ah, good point! I think that should be done by another patch, since it will improve ftracetest.
No, it belongs in this patch, because you are removing:
[ -f set_ftrace_filter ] && echo | tee set_ftrace_*
which is equivalent to:
[ -f set_ftrace_filter ] && echo > set_ftrace_filter [ -f set_ftrace_filter ] && echo > set_ftrace_notrace
as the "tee set_ftrace_*" covers both.
Ah, I see.
Without this change, this patch removes the update to set_ftrace_notrace.
OK, I'll update it.
Thank you!
-- Steve
Since dynamic function tracer can be disabled, set_ftrace_filter can be disappeared. Test cases which depends on it, must check whether the set_ftrace_filter exists or not before testing and if not, return as unsupported.
Also, if the function tracer itself is disabled, we can not set "function" to current_tracer. Test cases must check it before testing, and return as unsupported.
Signed-off-by: Masami Hiramatsu mhiramat@kernel.org --- .../ftrace/test.d/ftrace/func-filter-stacktrace.tc | 2 ++ .../selftests/ftrace/test.d/ftrace/func_cpumask.tc | 5 +++++ 2 files changed, 7 insertions(+)
diff --git a/tools/testing/selftests/ftrace/test.d/ftrace/func-filter-stacktrace.tc b/tools/testing/selftests/ftrace/test.d/ftrace/func-filter-stacktrace.tc index 36fb59f886ea..1a52f2883fe0 100644 --- a/tools/testing/selftests/ftrace/test.d/ftrace/func-filter-stacktrace.tc +++ b/tools/testing/selftests/ftrace/test.d/ftrace/func-filter-stacktrace.tc @@ -3,6 +3,8 @@ # description: ftrace - stacktrace filter command # flags: instance
+[ ! -f set_ftrace_filter ] && exit_unsupported + echo _do_fork:stacktrace >> set_ftrace_filter
grep -q "_do_fork:stacktrace:unlimited" set_ftrace_filter diff --git a/tools/testing/selftests/ftrace/test.d/ftrace/func_cpumask.tc b/tools/testing/selftests/ftrace/test.d/ftrace/func_cpumask.tc index 86a1f07ef2ca..7757b549f0b6 100644 --- a/tools/testing/selftests/ftrace/test.d/ftrace/func_cpumask.tc +++ b/tools/testing/selftests/ftrace/test.d/ftrace/func_cpumask.tc @@ -15,6 +15,11 @@ if [ $NP -eq 1 ] ;then exit_unresolved fi
+if ! grep -q function available_tracers ; then + echo "Function trace is not enabled" + exit_unsupported +fi + ORIG_CPUMASK=`cat tracing_cpumask`
do_reset() {
On Mon, 25 Nov 2019 15:57:09 +0900 Masami Hiramatsu mhiramat@kernel.org wrote:
Since dynamic function tracer can be disabled, set_ftrace_filter can be disappeared. Test cases which depends on it, must check whether the set_ftrace_filter exists or not before testing and if not, return as unsupported.
Also, if the function tracer itself is disabled, we can not set "function" to current_tracer. Test cases must check it before testing, and return as unsupported.
Signed-off-by: Masami Hiramatsu mhiramat@kernel.org
Reviewed-by: Steven Rostedt (VMware) rostedt@goodmis.org
-- Steve
Use relative path to trigger file instead of absolute debugfs path, because if the user uses tracefs instead of debugfs, it can be mounted at /sys/kernel/tracing. Anyway, since the ftracetest is designed to be run at the tracing directory, user doesn't need to use absolute path.
Signed-off-by: Masami Hiramatsu mhiramat@kernel.org Reviewed-by: Steven Rostedt (VMware) rostedt@goodmis.org --- .../inter-event/trigger-action-hist-xfail.tc | 4 ++-- .../inter-event/trigger-onchange-action-hist.tc | 2 +- .../inter-event/trigger-snapshot-action-hist.tc | 4 ++-- 3 files changed, 5 insertions(+), 5 deletions(-)
diff --git a/tools/testing/selftests/ftrace/test.d/trigger/inter-event/trigger-action-hist-xfail.tc b/tools/testing/selftests/ftrace/test.d/trigger/inter-event/trigger-action-hist-xfail.tc index 1221240f8cf6..3f2aee115f6e 100644 --- a/tools/testing/selftests/ftrace/test.d/trigger/inter-event/trigger-action-hist-xfail.tc +++ b/tools/testing/selftests/ftrace/test.d/trigger/inter-event/trigger-action-hist-xfail.tc @@ -21,10 +21,10 @@ grep -q "snapshot()" README || exit_unsupported # version issue
echo "Test expected snapshot action failure"
-echo 'hist:keys=comm:onmatch(sched.sched_wakeup).snapshot()' >> /sys/kernel/debug/tracing/events/sched/sched_waking/trigger && exit_fail +echo 'hist:keys=comm:onmatch(sched.sched_wakeup).snapshot()' >> events/sched/sched_waking/trigger && exit_fail
echo "Test expected save action failure"
-echo 'hist:keys=comm:onmatch(sched.sched_wakeup).save(comm,prio)' >> /sys/kernel/debug/tracing/events/sched/sched_waking/trigger && exit_fail +echo 'hist:keys=comm:onmatch(sched.sched_wakeup).save(comm,prio)' >> events/sched/sched_waking/trigger && exit_fail
exit_xfail diff --git a/tools/testing/selftests/ftrace/test.d/trigger/inter-event/trigger-onchange-action-hist.tc b/tools/testing/selftests/ftrace/test.d/trigger/inter-event/trigger-onchange-action-hist.tc index 064a284e4e75..c80007aa9f86 100644 --- a/tools/testing/selftests/ftrace/test.d/trigger/inter-event/trigger-onchange-action-hist.tc +++ b/tools/testing/selftests/ftrace/test.d/trigger/inter-event/trigger-onchange-action-hist.tc @@ -16,7 +16,7 @@ grep -q "onchange(var)" README || exit_unsupported # version issue
echo "Test onchange action"
-echo 'hist:keys=comm:newprio=prio:onchange($newprio).save(comm,prio) if comm=="ping"' >> /sys/kernel/debug/tracing/events/sched/sched_waking/trigger +echo 'hist:keys=comm:newprio=prio:onchange($newprio).save(comm,prio) if comm=="ping"' >> events/sched/sched_waking/trigger
ping $LOCALHOST -c 3 nice -n 1 ping $LOCALHOST -c 3 diff --git a/tools/testing/selftests/ftrace/test.d/trigger/inter-event/trigger-snapshot-action-hist.tc b/tools/testing/selftests/ftrace/test.d/trigger/inter-event/trigger-snapshot-action-hist.tc index 18fff69fc433..f546c1b66a9b 100644 --- a/tools/testing/selftests/ftrace/test.d/trigger/inter-event/trigger-snapshot-action-hist.tc +++ b/tools/testing/selftests/ftrace/test.d/trigger/inter-event/trigger-snapshot-action-hist.tc @@ -23,9 +23,9 @@ grep -q "snapshot()" README || exit_unsupported # version issue
echo "Test snapshot action"
-echo 1 > /sys/kernel/debug/tracing/events/sched/enable +echo 1 > events/sched/enable
-echo 'hist:keys=comm:newprio=prio:onchange($newprio).save(comm,prio):onchange($newprio).snapshot() if comm=="ping"' >> /sys/kernel/debug/tracing/events/sched/sched_waking/trigger +echo 'hist:keys=comm:newprio=prio:onchange($newprio).save(comm,prio):onchange($newprio).snapshot() if comm=="ping"' >> events/sched/sched_waking/trigger
ping $LOCALHOST -c 3 nice -n 1 ping $LOCALHOST -c 3
Fix multiple kprobe event testcase to work it correctly. There are 2 bugfixes. - Since `wc -l FILE` returns not only line number but also FILE filename, following "if" statement always failed. Fix this bug by replacing it with 'cat FILE | wc -l' - Since "while do-done loop" block with pipeline becomes a subshell, $N local variable is not update outside of the loop. Fix this bug by using actual target number (256) instead of $N.
Signed-off-by: Masami Hiramatsu mhiramat@kernel.org --- .../ftrace/test.d/kprobe/multiple_kprobes.tc | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/tools/testing/selftests/ftrace/test.d/kprobe/multiple_kprobes.tc b/tools/testing/selftests/ftrace/test.d/kprobe/multiple_kprobes.tc index 5862eee91e1d..6e3dbe5f96b7 100644 --- a/tools/testing/selftests/ftrace/test.d/kprobe/multiple_kprobes.tc +++ b/tools/testing/selftests/ftrace/test.d/kprobe/multiple_kprobes.tc @@ -20,9 +20,9 @@ while read i; do test $N -eq 256 && break done
-L=`wc -l kprobe_events` -if [ $L -ne $N ]; then - echo "The number of kprobes events ($L) is not $N" +L=`cat kprobe_events | wc -l` +if [ $L -ne 256 ]; then + echo "The number of kprobes events ($L) is not 256" exit_fail fi
On Mon, 25 Nov 2019 15:57:27 +0900 Masami Hiramatsu mhiramat@kernel.org wrote:
Fix multiple kprobe event testcase to work it correctly. There are 2 bugfixes.
- Since `wc -l FILE` returns not only line number but also FILE filename, following "if" statement always failed. Fix this bug by replacing it with 'cat FILE | wc -l'
- Since "while do-done loop" block with pipeline becomes a subshell, $N local variable is not update outside of the loop. Fix this bug by using actual target number (256) instead of $N.
Signed-off-by: Masami Hiramatsu mhiramat@kernel.org
Reviewed-by: Steven Rostedt (VMware) rostedt@goodmis.org
-- Steve
linux-kselftest-mirror@lists.linaro.org
-
Masami Hiramatsu -
Steven Rostedt