[lttng-dev] [PATCH lttng-tools 1/3] Fix: disable kernel event based on name and event type
Jonathan Rajotte Julien
Jonathan.rajotte-julien at efficios.com
Mon Sep 21 18:32:43 EDT 2015
On 2015-09-21 03:00 PM, Jérémie Galarneau wrote:
> On Thu, Sep 17, 2015 at 4:37 PM, Jonathan Rajotte
> <jonathan.rajotte-julien at efficios.com> wrote:
>> The -a argument is interpreted as a zero-length event name
>> instead of '*' which is actually a valid wildcard event
>> name by itself. This simplify how a disable action is handled.
>>
>> The event type can now be passed as argument
>> and is a new criteria while disabling kernel events.
>> The default is to disable for all event type.
>>
>> UST and agent domain do not support yet disabling by event
>> type.
>>
>> e.g:
>> # Only disable kernel event of type tracepoint.
>> lttng disable -a -k --tracepoint
>>
>> # Only disable the event with name '*' and type syscall.
>> lttng disable -k '*' --syscall
>>
>> # Disable all kernel event of all type.
>> lttng disable -a -k
>>
>> Signed-off-by: Jonathan Rajotte <jonathan.rajotte-julien at efficios.com>
>> ---
>> src/bin/lttng-sessiond/cmd.c | 45 +++++-----
>> src/bin/lttng-sessiond/event.c | 65 +++++---------
>> src/bin/lttng-sessiond/event.h | 5 +-
>> src/bin/lttng/commands/disable_events.c | 98 ++++++++++++++++++----
>> src/lib/lttng-ctl/lttng-ctl.c | 4 -
>> .../ust/python-logging/test_python_logging.in | 2 +-
>> 6 files changed, 121 insertions(+), 98 deletions(-)
>>
>> diff --git a/src/bin/lttng-sessiond/cmd.c b/src/bin/lttng-sessiond/cmd.c
>> index 7878a05..5a1cc7a 100644
>> --- a/src/bin/lttng-sessiond/cmd.c
>> +++ b/src/bin/lttng-sessiond/cmd.c
>> @@ -1225,32 +1225,22 @@ int cmd_disable_event(struct ltt_session *session,
>>
>> switch (event->type) {
>> case LTTNG_EVENT_ALL:
>> - ret = event_kernel_disable_event_all(kchan);
>> - if (ret != LTTNG_OK) {
>> - goto error_unlock;
>> - }
>> - break;
>> - case LTTNG_EVENT_TRACEPOINT: /* fall-through */
>> + case LTTNG_EVENT_TRACEPOINT:
>> case LTTNG_EVENT_SYSCALL:
>> - if (!strcmp(event_name, "*")) {
>> - ret = event_kernel_disable_event_type(kchan,
>> - event->type);
>> + case LTTNG_EVENT_PROBE:
>> + case LTTNG_EVENT_FUNCTION:
>> + case LTTNG_EVENT_FUNCTION_ENTRY:/* fall-through */
>> + if (event_name[0] == '\0') {
>> + ret = event_kernel_disable_event(kchan,
>> + NULL, event->type);
>> } else {
>> ret = event_kernel_disable_event(kchan,
>> - event_name);
>> + event_name, event->type);
>> }
>> if (ret != LTTNG_OK) {
>> goto error_unlock;
>> }
>> break;
>> - case LTTNG_EVENT_PROBE:
>> - case LTTNG_EVENT_FUNCTION:
>> - case LTTNG_EVENT_FUNCTION_ENTRY:
>> - ret = event_kernel_disable_event(kchan, event_name);
>> - if (ret != LTTNG_OK) {
>> - goto error_unlock;
>> - }
>> - break;
>> default:
>> ret = LTTNG_ERR_UNK;
>> goto error_unlock;
>> @@ -1273,7 +1263,7 @@ int cmd_disable_event(struct ltt_session *session,
>>
>> /*
>> * If a non-default channel has been created in the
>> - * session, explicitely require that -c chan_name needs
>> + * session, explicitly require that -c chan_name needs
>> * to be provided.
>> */
>> if (usess->has_non_default_channel && channel_name[0] == '\0') {
>> @@ -1290,10 +1280,12 @@ int cmd_disable_event(struct ltt_session *session,
>>
>> switch (event->type) {
>> case LTTNG_EVENT_ALL:
>> - if (strlen(event->name) == 1 &&
>> - !strncmp(event->name, "*", 1)) {
>> - ret = event_ust_disable_all_tracepoints(usess,
>> - uchan);
>> + /*
>> + * An empty event name means that everything
>> + * should be disabled.
>> + */
>> + if (event->name[0] == '\0') {
>> + ret = event_ust_disable_all_tracepoints(usess, uchan);
>> } else {
>> ret = event_ust_disable_tracepoint(usess, uchan,
>> event_name);
> Please move this hunk and the following one to separate patches as
> they fix the same problem for user space and agent domains. This
> commit's title implies that it only addresses the issue in the kernel
> domain.
>
>> @@ -1333,8 +1325,11 @@ int cmd_disable_event(struct ltt_session *session,
>> ret = -LTTNG_ERR_UST_EVENT_NOT_FOUND;
>> goto error_unlock;
>> }
>> - /* The wild card * means that everything should be disabled. */
>> - if (strncmp(event->name, "*", 1) == 0 && strlen(event->name) == 1) {
>> + /*
>> + * An empty event name means that everything
>> + * should be disabled.
>> + */
>> + if (event->name[0] == '\0') {
>> ret = event_agent_disable_all(usess, agt);
>> } else {
>> ret = event_agent_disable(usess, agt, event_name);
>> diff --git a/src/bin/lttng-sessiond/event.c b/src/bin/lttng-sessiond/event.c
>> index 13f09a5..a53d4e9 100644
>> --- a/src/bin/lttng-sessiond/event.c
>> +++ b/src/bin/lttng-sessiond/event.c
>> @@ -61,45 +61,15 @@ static void add_unique_ust_event(struct lttng_ht *ht,
>> }
>>
>> /*
>> - * Disable kernel tracepoint event for a channel from the kernel session.
>> + * Disable kernel tracepoint events for a channel from the kernel session of
>> + * a specified event_name and event type.
>> + * On type LTTNG_EVENT_ALL all events with event_name are disabled.
>> + * If event_name if NULL all events of the specified type are disabled.
> if NULL -> is NULL
>
>> */
>> int event_kernel_disable_event(struct ltt_kernel_channel *kchan,
>> - char *event_name)
>> -{
>> - int ret;
>> - struct ltt_kernel_event *kevent;
>> -
>> - assert(kchan);
>> -
>> - kevent = trace_kernel_get_event_by_name(event_name, kchan,
>> - LTTNG_EVENT_ALL);
>> - if (kevent == NULL) {
>> - ret = LTTNG_ERR_NO_EVENT;
>> - goto error;
>> - }
>> -
>> - ret = kernel_disable_event(kevent);
>> - if (ret < 0) {
>> - ret = LTTNG_ERR_KERN_DISABLE_FAIL;
>> - goto error;
>> - }
>> -
>> - DBG("Kernel event %s disable for channel %s.",
>> - kevent->event->name, kchan->channel->name);
>> -
>> - ret = LTTNG_OK;
>> -
>> -error:
>> - return ret;
>> -}
>> -
>> -/*
>> - * Disable kernel tracepoint events for a channel from the kernel session.
>> - */
>> -int event_kernel_disable_event_type(struct ltt_kernel_channel *kchan,
>> - enum lttng_event_type type)
>> + char *event_name, enum lttng_event_type type)
>> {
>> - int ret;
>> + int ret, error = 0, found = 0;
>> struct ltt_kernel_event *kevent;
>>
>> assert(kchan);
>> @@ -108,22 +78,25 @@ int event_kernel_disable_event_type(struct ltt_kernel_channel *kchan,
>> cds_list_for_each_entry(kevent, &kchan->events_list.head, list) {
>> if (type != LTTNG_EVENT_ALL && kevent->type != type)
>> continue;
>> + if (event_name != NULL && strcmp(event_name, kevent->event->name)) {
>> + continue;
>> + }
>> + found++;
>> ret = kernel_disable_event(kevent);
>> if (ret < 0) {
>> - /* We continue disabling the rest */
>> + error = 1;
>> continue;
>> }
>> }
>> - ret = LTTNG_OK;
>> - return ret;
>> -}
>> + DBG("Disable kernel event: found %d events with name: %s and type: %d",
>> + found, event_name, type);
> event_name could be NULL here.
>
> event_name ? event_name : "NULL"
Good catch !
>> -/*
>> - * Disable all kernel event for a channel from the kernel session.
>> - */
>> -int event_kernel_disable_event_all(struct ltt_kernel_channel *kchan)
>> -{
>> - return event_kernel_disable_event_type(kchan, LTTNG_EVENT_ALL);
>> + if (!found && event_name != NULL) {
>> + error = 1;
>> + }
>> +
>> + ret = error ? LTTNG_ERR_KERN_DISABLE_FAIL : LTTNG_OK;
>> + return ret;
> This should distinguish between:
>
> 1) event found, but we failed to disable it (LTTNG_ERR_KERN_DISABLE_FAIL)
> 2) event_name != NULL && !found (LTTNG_ERR_NO_EVENT, which it returned before)
You're right.
>
>> }
>>
>> /*
>> diff --git a/src/bin/lttng-sessiond/event.h b/src/bin/lttng-sessiond/event.h
>> index 7c5231d..45dd1fc 100644
>> --- a/src/bin/lttng-sessiond/event.h
>> +++ b/src/bin/lttng-sessiond/event.h
>> @@ -23,10 +23,7 @@
>> struct agent;
>>
>> int event_kernel_disable_event(struct ltt_kernel_channel *kchan,
>> - char *event_name);
>> -int event_kernel_disable_event_type(struct ltt_kernel_channel *kchan,
>> - enum lttng_event_type type);
>> -int event_kernel_disable_event_all(struct ltt_kernel_channel *kchan);
>> + char *event_name, enum lttng_event_type event_type);
>>
>> int event_kernel_enable_event(struct ltt_kernel_channel *kchan,
>> struct lttng_event *event, char *filter_expression,
>> diff --git a/src/bin/lttng/commands/disable_events.c b/src/bin/lttng/commands/disable_events.c
>> index c6ec85f..bcbc78d 100644
>> --- a/src/bin/lttng/commands/disable_events.c
>> +++ b/src/bin/lttng/commands/disable_events.c
>> @@ -44,7 +44,12 @@ static int opt_event_type;
>> enum {
>> OPT_HELP = 1,
>> OPT_USERSPACE,
>> - OPT_SYSCALL,
>> + OPT_TYPE_SYSCALL,
>> + OPT_TYPE_TRACEPOINT,
>> + OPT_TYPE_PROBE,
>> + OPT_TYPE_FUNCTION,
>> + OPT_TYPE_FUNCTION_ENTRY,
>> + OPT_TYPE_ALL,
>> OPT_LIST_OPTIONS,
>> };
>>
>> @@ -61,8 +66,24 @@ static struct poptOption long_options[] = {
>> {"log4j", 'l', POPT_ARG_VAL, &opt_log4j, 1, 0, 0},
>> {"python", 'p', POPT_ARG_VAL, &opt_python, 1, 0, 0},
>> {"kernel", 'k', POPT_ARG_VAL, &opt_kernel, 1, 0, 0},
>> - {"syscall", 0, POPT_ARG_NONE, 0, OPT_SYSCALL, 0, 0},
>> + {"syscall", 0, POPT_ARG_NONE, 0, OPT_TYPE_SYSCALL, 0, 0},
>> + {"probe", 0, POPT_ARG_NONE, 0, OPT_TYPE_PROBE, 0, 0},
>> + {"tracepoint", 0, POPT_ARG_NONE, 0, OPT_TYPE_TRACEPOINT, 0, 0},
>> + {"function", 0, POPT_ARG_NONE, 0, OPT_TYPE_FUNCTION, 0, 0},
>> + {"all", 0, POPT_ARG_NONE, 0, OPT_TYPE_ALL, 0, 0},
>> +#if 0
>> + /*
>> + * Function entry exits as a lttng_event_type but is never used on
>> + * enable
>> + */
>> + {"function-entry", 0, POPT_ARG_NONE, 0, OPT_TYPE_FUNCTION_ENTRY, 0, 0},
>> +
>> + /* Not implemented yet */
>> + {"userspace", 'u', POPT_ARG_STRING | POPT_ARGFLAG_OPTIONAL, &opt_cmd_name, OPT_USERSPACE, 0, 0},
>> + {"pid", 'p', POPT_ARG_INT, &opt_pid, 0, 0, 0},
>> +#else
>> {"userspace", 'u', POPT_ARG_NONE, 0, OPT_USERSPACE, 0, 0},
>> +#endif
> I have removed #ifdef-ed out code from disable-event in
>
> commit 3fd691fc09199be5a52009a391b22379477aa6e7
> Author: Jérémie Galarneau <jeremie.galarneau at efficios.com>
> Date: Tue Sep 15 13:00:40 2015 -0400
>
> Remove dead code from disable-event command
>
> Signed-off-by: Jérémie Galarneau <jeremie.galarneau at efficios.com>
>
> function-entry instrumentation is simply not supported, so let's
> mention it in the switch case that handles the instrumentation type,
> but not add compiled-out code.
>
>> {"list-options", 0, POPT_ARG_NONE, NULL, OPT_LIST_OPTIONS, NULL, NULL},
>> {0, 0, 0, 0, 0, 0, 0}
>> };
>> @@ -86,8 +107,19 @@ static void usage(FILE *ofp)
>> fprintf(ofp, " -l, --log4j Apply to Java application using LOG4j\n");
>> fprintf(ofp, " -p, --python Apply to Python application using logging\n");
>> fprintf(ofp, "\n");
>> - fprintf(ofp, "Event options:\n");
>> + fprintf(ofp, "Event type options (Only supported with kernel domain):\n");
>> + fprintf(ofp, " --all All event type (default)\n");
> type -> types
>
>> + fprintf(ofp, " --tracepoint Tracepoint event\n");
>> fprintf(ofp, " --syscall System call event\n");
>> + fprintf(ofp, " --probe Probe event\n");
>> + fprintf(ofp, " --function Function event\n");
>> +#if 0
>> + /*
>> + * Function entry exits as a lttng_event_type but is never used on
>> + * enable
>> + */
>> + fprintf(ofp, " --function-entry Function-entry event\n");
>> +#endif
> Same here; no ifdef-ed out code.
>
>> fprintf(ofp, "\n");
>> }
>>
>> @@ -103,6 +135,27 @@ const char *print_raw_channel_name(const char *name)
>> return name ? : "<default>";
>> }
>>
>> +static
>> +const char *print_event_type(const enum lttng_event_type ev_type)
>> +{
>> + switch (ev_type) {
>> + case LTTNG_EVENT_ALL:
>> + return "any";
>> + case LTTNG_EVENT_TRACEPOINT:
>> + return "tracepoint";
>> + case LTTNG_EVENT_PROBE:
>> + return "probe";
>> + case LTTNG_EVENT_FUNCTION:
>> + return "function";
>> + case LTTNG_EVENT_FUNCTION_ENTRY:
>> + return "function entry";
>> + case LTTNG_EVENT_SYSCALL:
>> + return "syscall";
>> + default:
>> + return "";
>> + }
>> +}
>> +
>> /* Mi print a partial event.
>> * enabled is 0 or 1
>> * success is 0 or 1
>> @@ -213,14 +266,8 @@ static int disable_events(char *session_name)
>> /* Set default loglevel to any/unknown */
>> event.loglevel = -1;
>>
>> - switch (opt_event_type) {
>> - case LTTNG_EVENT_SYSCALL:
>> - event.type = LTTNG_EVENT_SYSCALL;
>> - break;
>> - default:
>> - event.type = LTTNG_EVENT_ALL;
>> - break;
>> - }
>> + /* opt_event_type contain the event type to disable at this point */
>> + event.type = opt_event_type;
>>
>> if (opt_disable_all) {
>> command_ret = lttng_disable_event_ext(handle, &event, channel_name, NULL);
>> @@ -232,9 +279,9 @@ static int disable_events(char *session_name)
>> } else {
>> enabled = 0;
>> success = 1;
>> - MSG("All %s %s are disabled in channel %s",
>> + MSG("All %s events of type %s are disabled in channel %s",
>> get_domain_str(dom.type),
>> - opt_event_type == LTTNG_EVENT_SYSCALL ? "system calls" : "events",
>> + print_event_type(opt_event_type),
>> print_channel_name(channel_name));
>> }
>>
>> @@ -255,9 +302,9 @@ static int disable_events(char *session_name)
>> event.name[sizeof(event.name) - 1] = '\0';
>> command_ret = lttng_disable_event_ext(handle, &event, channel_name, NULL);
>> if (command_ret < 0) {
>> - ERR("%s %s: %s (channel %s, session %s)",
>> - opt_event_type == LTTNG_EVENT_SYSCALL ? "System call" : "Event",
>> + ERR("%s of type %s : %s (channel %s, session %s)",
>> event_name,
>> + print_event_type(opt_event_type),
>> lttng_strerror(command_ret),
>> command_ret == -LTTNG_ERR_NEED_CHANNEL_NAME
>> ? print_raw_channel_name(channel_name)
>> @@ -271,10 +318,10 @@ static int disable_events(char *session_name)
>> */
>> enabled = 1;
>> } else {
>> - MSG("%s %s %s disabled in channel %s for session %s",
>> + MSG("%s %s of type %s disabled in channel %s for session %s",
>> get_domain_str(dom.type),
>> - opt_event_type == LTTNG_EVENT_SYSCALL ? "system call" : "event",
>> event_name,
>> + print_event_type(opt_event_type),
>> print_channel_name(channel_name),
>> session_name);
>> success = 1;
>> @@ -341,9 +388,24 @@ int cmd_disable_events(int argc, const char **argv)
>> case OPT_USERSPACE:
>> opt_userspace = 1;
>> break;
>> - case OPT_SYSCALL:
>> + case OPT_TYPE_SYSCALL:
>> opt_event_type = LTTNG_EVENT_SYSCALL;
>> break;
>> + case OPT_TYPE_TRACEPOINT:
>> + opt_event_type = LTTNG_EVENT_TRACEPOINT;
>> + break;
>> + case OPT_TYPE_PROBE:
>> + opt_event_type = LTTNG_EVENT_PROBE;
>> + break;
>> + case OPT_TYPE_FUNCTION:
>> + opt_event_type = LTTNG_EVENT_FUNCTION;
>> + break;
>> + case OPT_TYPE_FUNCTION_ENTRY:
>> + opt_event_type = LTTNG_EVENT_FUNCTION_ENTRY;
>> + break;
> We shouldn't allow this (FUNCTION and FUNCTION_ENTRY) for non-kernel
> domains as their support is not implemented.
>
>> + case OPT_TYPE_ALL:
>> + opt_event_type = LTTNG_EVENT_ALL;
>> + break;
>> case OPT_LIST_OPTIONS:
>> list_cmd_options(stdout, long_options);
>> goto end;
>> diff --git a/src/lib/lttng-ctl/lttng-ctl.c b/src/lib/lttng-ctl/lttng-ctl.c
>> index 9cbfef5..1007326 100644
>> --- a/src/lib/lttng-ctl/lttng-ctl.c
>> +++ b/src/lib/lttng-ctl/lttng-ctl.c
>> @@ -1091,10 +1091,6 @@ int lttng_disable_event_ext(struct lttng_handle *handle,
>> }
>>
>> lsm.cmd_type = LTTNG_DISABLE_EVENT;
>> - if (ev->name[0] == '\0') {
>> - /* Disable all events */
>> - lttng_ctl_copy_string(ev->name, "*", sizeof(ev->name));
>> - }
>>
>> lttng_ctl_copy_lttng_domain(&lsm.domain, &handle->domain);
>> /* FIXME: copying non-packed struct to packed struct. */
>> diff --git a/tests/regression/ust/python-logging/test_python_logging.in b/tests/regression/ust/python-logging/test_python_logging.in
>> index 8aff408..d5a9b80 100755
>> --- a/tests/regression/ust/python-logging/test_python_logging.in
>> +++ b/tests/regression/ust/python-logging/test_python_logging.in
>> @@ -351,7 +351,7 @@ function test_python_disable_all()
>> enable_python_lttng_event $SESSION_NAME $EVENT_NAME
>> enable_python_lttng_event $SESSION_NAME $EVENT_NAME2
>>
>> - disable_python_lttng_event $SESSION_NAME '*'
>> + disable_python_lttng_event $SESSION_NAME -a
> Please submit a separate patch for this.
Since I added two more commits I'll simply reply to your message with
the new patch set.
>
> Thanks!
> Jérémie
>
>> start_lttng_tracing_ok $SESSION_NAME
>>
>> --
>> 2.1.4
>>
>
>
--
Jonathan R. Julien
Efficios
More information about the lttng-dev
mailing list