Extend coverage of "trace-cmd reset" command
v5 cover Extend coverage of "trace-cmd reset" command
The "trace-cmd reset" command is supposed to set the ftrace state to default. However, some ftrace configurations are not reseted: synthetic events and error_log file. The command is extended to clean these.
Two new tracefs APIs were also implemented, to check if given ftrace file / directory exists:
bool tracefs_file_exist(struct tracefs_instance *instance, char *name); bool tracefs_dir_exist(struct tracefs_instance *instance, char *name);
[
v5 changes:
Removed accepted patches from the set
Address Steven's comments
v4 chanegs:
Implemented new tracefs APIs to check if given ftrace file / directory exists. Use the new APIs to check if synthetic_events and error_log files exist, before accessing them.
v3 changes:
Remove the warning in case there are no configured synthetic events.
v2 changes:
Removed these patches from the "SQL-like syntax for ftrace histograms configuration" patch set, as they are not directly related to it.
]
Tzvetomir Stoyanov (VMware) (3):
trace-cmd: Add new tracefs APIs for checking if a ftrace file /
directory exists
trace-cmd: Unit tests for new libtracefs APIs
trace-cmd: Clear error log on reset subcommand
include/tracefs/tracefs.h | 3 +++ lib/tracefs/tracefs-instance.c | 46 +++++++++++++++++++++++++++++++++- tracecmd/trace-record.c | 23 +++++++++++++++++ utest/tracefs-utest.c | 16 ++++++++++++ 4 files changed, 87 insertions(+), 1 deletion(-)
Signature · 1 line
2.25.1
PATCH v5 1/3 trace-cmd: Add new tracefs APIs for checking if a ftrace file / directory exists
Some ftrace files and directories are optional, depending on specific kernel configuration or version. It is a good practice to check if the file / directory exist, before trying to access it.
There are a lot of places in trace-cmd implementation with such checks. The new libtracefs APIs can be used for this, they are ftarce instance aware:
bool tracefs_file_exist(struct tracefs_instance *instance, char *name); bool tracefs_dir_exist(struct tracefs_instance *instance, char *name);
--- include/tracefs/tracefs.h | 3 +++ lib/tracefs/tracefs-instance.c | 46 +++++++++++++++++++++++++++++++++- 2 files changed, 48 insertions(+), 1 deletion(-)
include/tracefs/tracefs.h +3 -0
@@ -33,6 +33,9 @@ int tracefs_instance_file_write(struct tracefs_instance *instance,char *tracefs_instance_file_read (struct tracefs_instance *instance,char *file,int *psize); +bool tracefs_file_exist (struct tracefs_instance *instance,char *name); +bool tracefs_dir_exist (struct tracefs_instance *instance,char *name); +/* events */ void tracefs_list_free (char **list);char **tracefs_event_systems (const char *tracing_dir);
lib/tracefs/tracefs-instance.c +45 -1
55 lines not on this page · raw message
Signature · 1 line
2.25.1
PATCH v5 2/3 trace-cmd: Unit tests for new libtracefs APIs
In reply to Tzvetomir Stoyanov (VMware)Unit tests for tracefs APIs, checking for ftrace file / directory existence:
bool tracefs_file_exist(struct tracefs_instance *instance, char *name); bool tracefs_dir_exist(struct tracefs_instance *instance, char *name);
--- utest/tracefs-utest.c | 16 ++++++++++++++++ 1 file changed, 16 insertions(+)
utest/tracefs-utest.c +16 -0
@@ -175,6 +175,7 @@ static void test_instance_file_read(struct tracefs_instance *inst, char *fname)#define ALL_TRACERS "available_tracers"#define CUR_TRACER "current_tracer" +#define PER_CPU "per_cpu"static void test_instance_file (void ) {struct tracefs_instance *instance =NULL ;
@@ -251,6 +252,21 @@ static void test_instance_file(void)tracefs_put_tracing_file (inst_file);free (fname); +CU_TEST (tracefs_file_exist (NULL , (char *)name) == false); +CU_TEST (tracefs_dir_exist (NULL , (char *)name) == false); +CU_TEST (tracefs_file_exist (instance, (char *)name) == false); +CU_TEST (tracefs_dir_exist (instance, (char *)name) == false); + +CU_TEST (tracefs_file_exist (NULL ,CUR_TRACER ) == true); +CU_TEST (tracefs_dir_exist (NULL ,CUR_TRACER ) == false); +CU_TEST (tracefs_file_exist (instance,CUR_TRACER ) == true); +CU_TEST (tracefs_dir_exist (instance,CUR_TRACER ) == false); + +CU_TEST (tracefs_file_exist (NULL ,PER_CPU ) == false); +CU_TEST (tracefs_dir_exist (NULL ,PER_CPU ) == true); +CU_TEST (tracefs_file_exist (instance,PER_CPU ) == false); +CU_TEST (tracefs_dir_exist (instance,PER_CPU ) == true); +CU_TEST (tracefs_instance_destroy (NULL ) !=0 );CU_TEST (tracefs_instance_destroy (instance) ==0 );CU_TEST (tracefs_instance_destroy (instance) !=0 );
Signature · 1 line
2.25.1
PATCH v5 3/3 trace-cmd: Clear error log on reset subcommand
In reply to Tzvetomir Stoyanov (VMware)The "trace-cmd reset" command is supposed to set the ftrace state to default. However, the ftrace error logs are not reseted.
A logic is added to delete the content of "error_log" file when "trace-cmd reset" is executed.
--- tracecmd/trace-record.c | 23 +++++++++++++++++++++++ 1 file changed, 23 insertions(+)
tracecmd/trace-record.c +23 -0
@@ -4820,6 +4820,28 @@ static void clear_triggers(void)clear_instance_triggers (instance); } +static void clear_instance_error_log (struct buffer_instance *instance) +{ +char *file; + +if (!tracefs_file_exist (instance->tracefs,"error_log" )) +return ; + + file =tracefs_instance_get_file (instance->tracefs,"error_log" ); +if (!file) +return ; +write_file (file," " ); +tracefs_put_tracing_file (file); +} + +static void clear_error_log (void ) +{ +struct buffer_instance *instance; + +for_all_instances (instance) +clear_instance_error_log (instance); +} +static void clear_all_synth_events (void ) {char sevent[BUFSIZ ];
@@ -5564,6 +5586,7 @@ void trace_reset(int argc, char **argv)clear_filters ();clear_triggers ();clear_all_synth_events (); +clear_error_log ();/* set clock to "local" */ reset_clock ();reset_event_pid ();
Signature · 1 line
2.25.1