tracing/filters: add TRACE_EVENT_FORMAT_NOFILTER event macro
PATCH 1/1 tracing/filters: add TRACE_EVENT_FORMAT_NOFILTER event macro
Frederic Weisbecker suggested that the trace_special event shouldn't be filterable; this patch adds a TRACE_EVENT_FORMAT_NOFILTER event macro that allows an event format to be exported without having a filter attached, and removes filtering from the trace_special event.
--- kernel/trace/trace.c | 2 -- kernel/trace/trace.h | 2 ++ kernel/trace/trace_event_types.h | 2 +- kernel/trace/trace_export.c | 33 +++++++++++++++++++++++++++++++++ 4 files changed, 36 insertions(+), 3 deletions(-)
kernel/trace/trace.c +0 -2
@@ -1064,7 +1064,6 @@ ftrace_trace_special(void *__tr, unsigned long arg1, unsigned long arg2, unsigned long arg3, int pc) { - struct ftrace_event_call *call = &event_special; struct ring_buffer_event *event;struct trace_array *tr = __tr;struct special_entry *entry;
@@ -1077,7 +1076,6 @@ ftrace_trace_special(void *__tr, entry->arg1 = arg1; entry->arg2 = arg2; entry->arg3 = arg3; -filter_check_discard (call, entry, event);trace_buffer_unlock_commit (tr, event,0 , pc); }
kernel/trace/trace.h +2 -0
@@ -909,6 +909,8 @@ do { \#undef TRACE_EVENT_FORMAT#define TRACE_EVENT_FORMAT (call, proto, args, fmt, tstruct, tpfmt) \ extern struct ftrace_event_call event_##call; +#undef TRACE_EVENT_FORMAT_NOFILTER +#define TRACE_EVENT_FORMAT_NOFILTER (call, proto, args, fmt, tstruct, tpfmt)#include "trace_event_types.h" #endif /* _LINUX_KERNEL_TRACE_H */
kernel/trace/trace_event_types.h +1 -1
@@ -57,7 +57,7 @@ TRACE_EVENT_FORMAT(context_switch, TRACE_CTX, ctx_switch_entry, ignore,TP_RAW_FMT ("%u:%u:%u ==+ %u:%u:%u [%03u]" ) ); -TRACE_EVENT_FORMAT(special, TRACE_SPECIAL, special_entry, ignore, +TRACE_EVENT_FORMAT_NOFILTER(special, TRACE_SPECIAL, special_entry, ignore, TRACE_STRUCT( TRACE_FIELD(unsigned long, arg1, arg1) TRACE_FIELD(unsigned long, arg2, arg2)
kernel/trace/trace_export.c +33 -0
@@ -65,6 +65,22 @@ ftrace_format_##call(struct trace_seq *s) \return ret; \ } +#undef TRACE_EVENT_FORMAT_NOFILTER +#define TRACE_EVENT_FORMAT_NOFILTER (call, proto, args, fmt, tstruct, \ + tpfmt) \ +static int \ +ftrace_format_##call(struct trace_seq *s) \ +{ \ + struct args field; \ + int ret; \ + \ + tstruct; \ + \ + trace_seq_printf(s, "\nprint fmt: \"%s\"\n", tpfmt); \ + \ + return ret; \ +} +#include "trace_event_types.h" #undef TRACE_ZERO_CHAR
@@ -109,6 +125,19 @@ static int ftrace_raw_init_event_##call(void) \return 0 ; \ } \ +#undef TRACE_EVENT_FORMAT_NOFILTER +#define TRACE_EVENT_FORMAT_NOFILTER (call, proto, args, fmt, tstruct, \ + tpfmt) \ + \ +struct ftrace_event_call __used \ +__attribute__((__aligned__(4))) \ +__attribute__((section("_ftrace_events"))) event_##call = { \ + .name = #call, \ + .id = proto, \ + .system = __stringify(TRACE_SYSTEM), \ + .show_format = ftrace_format_##call, \ +}; +#include "trace_event_types.h" #undef TRACE_FIELD
@@ -150,4 +179,8 @@ ftrace_define_fields_##call(void) \return ret; \ } +#undef TRACE_EVENT_FORMAT_NOFILTER +#define TRACE_EVENT_FORMAT_NOFILTER (call, proto, args, fmt, tstruct, \ + tpfmt) +#include "trace_event_types.h"
Signature · 1 line
1.5.6.3
Steven Rostedt
maintainerQuoted 5 lines · Tom Zanussi
Frederic Weisbecker suggested that the
trace_specialevent shouldn't be filterable; this patch adds aTRACE_EVENT_FORMAT_NOFILTERevent macro that allows an event format to be exported without having a filter attached, and removes filtering from thetrace_specialevent.
Frederic,
Do you remember why trace_special should not be filtered? I think earlier we use to use it for lots of special markings, but now that we have trace_printk, at least I have not used it in a long time.
Reason why I'm asking, is that I've wrote a patch that automates the format of the debugfs/tracing/events/ftrace/* files. I'm using macros like we have in include/trace/events/ to create the ftrace internal structures. This way we get rid of the manual exporting of that directory and now the formats will be automatically match the ring buffer internals.
This also adds some entries that were not ported, just because it is automated we get all of them. I'm thinking it would be more powerful to let all ftrace entries be filtered. Even the trace_printks.
What do you think?
-- Steve
Quoted 111 lines
Signed-off-by Tom Zanussi <[email protected]>--- kernel/trace/trace.c | 2 -- kernel/trace/trace.h | 2 ++ kernel/trace/trace_event_types.h | 2 +- kernel/trace/trace_export.c | 33 +++++++++++++++++++++++++++++++++ 4 files changed, 36 insertions(+), 3 deletions(-)kernel/trace/trace.c +0
-2@@ -1064,7 +1064,6 @@ ftrace_trace_special(void *__tr, unsigned long arg1, unsigned long arg2, unsigned long arg3, int pc) { - struct ftrace_event_call *call = &event_special; struct ring_buffer_event *event;struct trace_array *tr = __tr;struct special_entry *entry;@@ -1077,7 +1076,6 @@ ftrace_trace_special(void *__tr, entry->arg1 = arg1; entry->arg2 = arg2; entry->arg3 = arg3; -filter_check_discard (call, entry, event);trace_buffer_unlock_commit (tr, event,0 , pc); }kernel/trace/trace.h +2
-0@@ -909,6 +909,8 @@ do { \#undef TRACE_EVENT_FORMAT#define TRACE_EVENT_FORMAT (call, proto, args, fmt, tstruct, tpfmt) \ extern struct ftrace_event_call event_##call; +#undef TRACE_EVENT_FORMAT_NOFILTER +#define TRACE_EVENT_FORMAT_NOFILTER (call, proto, args, fmt, tstruct, tpfmt)#include "trace_event_types.h" #endif /* _LINUX_KERNEL_TRACE_H */ kernel/trace/trace_event_types.h +1
-1@@ -57,7 +57,7 @@ TRACE_EVENT_FORMAT(context_switch, TRACE_CTX, ctx_switch_entry, ignore,TP_RAW_FMT ("%u:%u:%u ==+ %u:%u:%u [%03u]" ) ); -TRACE_EVENT_FORMAT(special, TRACE_SPECIAL, special_entry, ignore, +TRACE_EVENT_FORMAT_NOFILTER(special, TRACE_SPECIAL, special_entry, ignore, TRACE_STRUCT( TRACE_FIELD(unsigned long, arg1, arg1) TRACE_FIELD(unsigned long, arg2, arg2)kernel/trace/trace_export.c +33
-0@@ -65,6 +65,22 @@ ftrace_format_##call(struct trace_seq *s) \return ret; \ } +#undef TRACE_EVENT_FORMAT_NOFILTER +#define TRACE_EVENT_FORMAT_NOFILTER (call, proto, args, fmt, tstruct, \ + tpfmt) \ +static int \ +ftrace_format_##call(struct trace_seq *s) \ +{ \ + struct args field; \ + int ret; \ + \ + tstruct; \ + \ + trace_seq_printf(s, "\nprint fmt: \"%s\"\n", tpfmt); \ + \ + return ret; \ +} +#include "trace_event_types.h" #undef TRACE_ZERO_CHAR@@ -109,6 +125,19 @@ static int ftrace_raw_init_event_##call(void) \return 0 ; \ } \ +#undef TRACE_EVENT_FORMAT_NOFILTER +#define TRACE_EVENT_FORMAT_NOFILTER (call, proto, args, fmt, tstruct, \ + tpfmt) \ + \ +struct ftrace_event_call __used \ +__attribute__((__aligned__(4))) \ +__attribute__((section("_ftrace_events"))) event_##call = { \ + .name = #call, \ + .id = proto, \ + .system = __stringify(TRACE_SYSTEM), \ + .show_format = ftrace_format_##call, \ +}; +#include "trace_event_types.h" #undef TRACE_FIELD@@ -150,4 +179,8 @@ ftrace_define_fields_##call(void) \return ret; \ } +#undef TRACE_EVENT_FORMAT_NOFILTER +#define TRACE_EVENT_FORMAT_NOFILTER (call, proto, args, fmt, tstruct, \ + tpfmt) +#include "trace_event_types.h"
Quoted 20 lines · Steven Rostedt
Quoted 5 lines · Tom Zanussi
Frederic Weisbecker suggested that the
trace_specialevent shouldn't be filterable; this patch adds aTRACE_EVENT_FORMAT_NOFILTERevent macro that allows an event format to be exported without having a filter attached, and removes filtering from thetrace_specialevent.Frederic,
Do you remember why
trace_specialshould not be filtered? I think earlier we use to use it for lots of special markings, but now that we havetrace_printk, at least I have not used it in a long time.Reason why I'm asking, is that I've wrote a patch that automates the format of the debugfs/tracing/events/ftrace/* files. I'm using macros like we have in
include/trace/events/to create the ftrace internal structures. This way we get rid of the manual exporting of that directory and now the formats will be automatically match the ring buffer internals.
Quoted 7 lines
This also adds some entries that were not ported, just because it is automated we get all of them. I'm thinking it would be more powerful to let all ftrace entries be filtered. Even the
trace_printks.What do you think?
-- Steve
It's not that it would harm but it would be meaningless. The tracer that uses the special entry is the only one that can give sense to it.
Sysprof is the only user IIRC.
If I'm not wrong, the first field gives the sense of the other fields:
0 -> pid
1 -> backtrace node
2 -> backtrace node (userland)
3 -> backtrace overflow
Well, actually while thinking more about it, we may want to
filter with "arg0 == 1 && arg1 == addr_to_filter_in_a_callchain" or things like that.
So I guess actually that could be useful.
I did not think about it at that time...