trace-cmd: Improve multi-instance support
v3 cover trace-cmd: Improve multi-instance support
NOTE: The previous two versions of this patch set were distributed off-list.
v3 changes:
* Improved clarification of options in documentation of trace-cmd reset and added examples.
v2 changes:
* Rebased this patch series on Steve's simplification of first_instance logic.
Improve the completeness and consistency of support for multiple buffer instances in the extract, snapshot, stat, start, stop, reset, and restart sub-commands.
For the commands that lacked it, added support for -B and -t options, consistent with the existing behavior of stop, reset, and restart. That is, including a -B instance-name option will cause these commands to exclude the top buffer instance unless -t is also provided.
Exception: trace-cmd start retains its existing behavior regarding the top instance, consistent with trace-cmd record.
For all of these sub-commands except start, add a new -a option, which causes it to operate on all existing instances. As with -B, -a will exclude the top instance unless -t is also given.
Updated documentation accordingly.
Howard Cochran (5):
trace-cmd extract: Add -B and -t for multiple instances trace-cmd extract: Add -a option for all instances trace-cmd stop, reset, restart: Add -a for all instances trace-cmd snapshot: Fully support multiple instances trace-cmd stat: Support -a for all buffers. Update docs.
Documentation/trace-cmd-extract.1.txt | 16 ++++ Documentation/trace-cmd-reset.1.txt | 57 ++++++++++-- Documentation/trace-cmd-snapshot.1.txt | 16 +++- Documentation/trace-cmd-stat.1.txt | 19 ++++ Documentation/trace-cmd-stop.1.txt | 13 ++- trace-cmd.c | 2 +- trace-local.h | 6 +- trace-record.c | 160 ++++++++++++++++++++++++++++----- trace-snapshot.c | 90 ++++++++++--------- trace-stat.c | 15 +++- trace-usage.c | 42 ++++++--- 11 files changed, 341 insertions(+), 95 deletions(-)
Signature · 1 line
1.9.1
PATCH v3 2/5 trace-cmd extract: Add -a option for all instances
This option will extract all instances that currently exist in the system, including the top instance. This differs from the meaning of -a for record and stream (enable all events), which would have no purpose for extract. Such difference in meaning already exists for -s, so this seemed reasonable.
Had to fix a bug in create_instance(), in which it ignored its parameter.
Signed-off-by Howard Cochran <[email protected]>
--- Documentation/trace-cmd-extract.1.txt | 16 +++++++ trace-local.h | 2 +- trace-record.c | 81 +++++++++++++++++++++++++++++++++-- trace-usage.c | 5 ++- 4 files changed, 98 insertions(+), 6 deletions(-)
Documentation/trace-cmd-extract.1.txt +16 -0
@@ -52,6 +52,22 @@ OPTIONS does cause the extract routine to disable all tracing. That is, the end of the extract will perform something similar to trace-cmd-reset(1). +*-B* 'buffer-name':: + If the kernel supports multiple buffers, this will extract the trace for + only the given buffer. It does not affect any other buffer. This may be + used multiple times to specify different buffers. When this option is + used, the top level instance will not be extracted unless *-t* is given. + +*-a*:: + Extract all existing buffer instances. When this option is used, the + top level instance will not be extracted unless *-t* is given. + +*-t*:: + Extracts the top level instance buffer. Without the *-B* or *-a* option + this is the same as the default. But if *-B* or *-a* is used, this is + required if the top level instance buffer should also be extracted. + + SEE ALSO -------- trace-cmd(1), trace-cmd-record(1), trace-cmd-report(1), trace-cmd-start(1),
trace-local.h +1 -1
@@ -176,7 +176,7 @@ extern struct buffer_instance *first_instance;#define for_all_instances (i) for (i = first_instance; i; \ i = i == &top_instance ? buffer_instances : (i)->next) -struct buffer_instance *create_instance (char *name); +struct buffer_instance *create_instance (const char *name);void add_instance (struct buffer_instance *instance);char *get_instance_file (struct buffer_instance *instance,const char *file);
trace-record.c +78 -3
103 lines not on this page · raw message
trace-usage.c +3 -2
@@ -62,11 +62,12 @@ static struct usage_help usage_help[] = { {"extract" ,"extract a trace from the kernel" , -" %s extract [-p plugin][-O option][-o file][-B buf][-s][-t]\n" +" %s extract [-p plugin][-O option][-o file][-B buf][-s][-a][-t]\n" " Uses similar options as record, but only reads an existing trace.\n" " -s : extract the snapshot instead of the main buffer\n" " -B : extract a given buffer (more than one may be specified)\n" -" -t : include the top level buffer (useful with -B)\n" +" -a : extract all buffers (except top one)\n" +" -t : extract the top level buffer (useful with -B and -a)\n" }, { "stop",
Signature · 1 line
1.9.1
PATCH v3 1/5 trace-cmd extract: Add -B and -t for multiple instances
In reply to Howard CochranAdd -B option to trace-cmd extract so that it can extract listed instances. As with start, stop, and reset, the top level instance is not included when any -B are given. However, -t makes it also include the top instance. As with other commands, if neither option is given, only operate on the top instance.
Signed-off-by Howard Cochran <[email protected]>
--- trace-record.c | 20 +++++++++++++------- trace-usage.c | 6 ++++-- 2 files changed, 17 insertions(+), 9 deletions(-)
trace-record.c +13 -7
@@ -523,17 +523,20 @@ static int create_recorder(struct buffer_instance *instance, int cpu,static void flush_threads (void ) { +struct buffer_instance *instance;long ret;int i;if (!cpu_count)return ; -for (i =0 ; i < cpu_count; i++) { -/* Extract doesn't support sub buffers yet */ - ret =create_recorder (&top_instance, i,TRACE_TYPE_EXTRACT ,NULL ); -if (ret <0 ) -die ("error reading ring buffer" ); +for_all_instances (instance) { +for (i =0 ; i < cpu_count; i++) { +/* Extract doesn't support sub buffers yet */ + ret =create_recorder (instance, i,TRACE_TYPE_EXTRACT ,NULL ); +if (ret <0 ) +die ("error reading ring buffer" ); + } } }
@@ -3873,7 +3876,7 @@ void trace_record (int argc, char **argv) };if (extract) - opts ="+haf:Fp:co:O:sr:g:l:n:P:N:tb:ksiT" ; + opts ="+haf:Fp:co:O:sr:g:l:n:P:N:tb:B:ksiT" ;else opts ="+hae:f:Fp:cC:dDGo:O:s:r:vg:l:n:P:N:tb:R:B:ksSiTm:M:H:" ; c =getopt_long (argc-1 , argv+1 , opts, long_options, &option_index);
@@ -4061,7 +4064,10 @@ void trace_record (int argc, char **argv) instance->cpumask = optarg;break ;case 't' : - use_tcp =1 ; +if (extract) + topt =1 ;/* Extract top instance also */ +else + use_tcp =1 ;break ;case 'b' : instance->buffer_size =atoi (optarg);
trace-usage.c +4 -2
@@ -62,9 +62,11 @@ static struct usage_help usage_help[] = { {"extract" ,"extract a trace from the kernel" , -" %s extract [-p plugin][-O option][-o file][-s]\n" -" Uses same options as record, but only reads an existing trace.\n" +" %s extract [-p plugin][-O option][-o file][-B buf][-s][-t]\n" +" Uses similar options as record, but only reads an existing trace.\n" " -s : extract the snapshot instead of the main buffer\n" +" -B : extract a given buffer (more than one may be specified)\n" +" -t : include the top level buffer (useful with -B)\n" }, { "stop",
Signature · 1 line
1.9.1
PATCH v3 5/5 trace-cmd stat: Support -a for all buffers. Update docs.
In reply to Howard CochranAs for stop, restart, reset, extract, -a means show status of all buffer instances except the top one. Use -t to get the top one too.
Also added documentation for the existing -B and -t options.
Signed-off-by Howard Cochran <[email protected]>
--- Documentation/trace-cmd-stat.1.txt | 19 +++++++++++++++++++ trace-stat.c | 15 +++++++++++---- trace-usage.c | 6 +++++- 3 files changed, 35 insertions(+), 5 deletions(-)
Documentation/trace-cmd-stat.1.txt +19 -0
@@ -41,6 +41,25 @@ system. The status that it shows is: *Uprobes:* Shows any uprobes that are defined for tracing. +OPTIONS +------- +*-B* 'buffer-name':: + If the kernel supports multiple buffers, this will display the status of + of the given buffer instance. This may be used multiple times to + specify different buffers. When this option is used, the status of the + top level instance will not be shown unless *-t* is given. + +*-a*:: + Display the status of all existing buffer instances. When this option + is used, the status of the top level instance will not be displayed + unless *-t* is given. + +*-t*:: + Display the status of the top level instance buffer. Without the *-B* + or *-a* option, this is the same as the default. But if *-B* or *-a* + is used, this is required if the top level instance buffer should also + be displayed. + SEE ALSO -------- trace-cmd(1), trace-cmd-record(1), trace-cmd-report(1), trace-cmd-start(1),
trace-stat.c +11 -4
@@ -865,6 +865,12 @@ static void report_traceon(struct buffer_instance *instance)static void stat_instance (struct buffer_instance *instance) { +if (instance->name) { +/* Lack of trailing newline is intentional */ +printf ("\nInstance: %s\n====================" , + instance->name); + } +report_plugin (instance);report_events (instance);report_event_filters (instance);
@@ -887,7 +893,7 @@ void trace_stat (int argc, char **argv)int c;for (;;) { - c =getopt (argc-1 , argv+1 ,"tB:" ); + c =getopt (argc-1 , argv+1 ,"hatB:" );if (c ==-1 )break ;switch (c) {
@@ -897,9 +903,9 @@ void trace_stat (int argc, char **argv)case 'B' : instance =create_instance (optarg);add_instance (instance); -/* top instance requires direct access */ -if (!topt &&is_top_instance (first_instance)) - first_instance = instance; +break ; +case 'a' : +add_all_instances ();break ;case 't' :/* Force to use top instance */
@@ -911,6 +917,7 @@ void trace_stat (int argc, char **argv)usage (argv); } } +update_first_instance (topt);for_all_instances (instance) {stat_instance (instance);
trace-usage.c +5 -1
@@ -182,7 +182,11 @@ static struct usage_help usage_help[] = { {"stat" ,"show the status of the running tracing (ftrace) system" , -" %s stat" +" %s stat [-B buf][-a][-t]\n" +" Shows the status of the tracing system" +" -B status for a given buffer (more than one may be specified)\n" +" -a status for all buffers (except top one)\n" +" -t status the top level buffer (useful with -B or -a)\n" }, { "split",
Signature · 1 line
1.9.1
PATCH v3 4/5 trace-cmd snapshot: Fully support multiple instances
In reply to Howard CochranAdd support for multiple -B instances, the same as stop, reset, restart, and extract sub-commands. -a operates on all instances except the top one -t also includes the top instance
Signed-off-by Howard Cochran <[email protected]>
--- Documentation/trace-cmd-snapshot.1.txt | 16 +++++- trace-cmd.c | 2 +- trace-local.h | 4 ++ trace-record.c | 4 +- trace-snapshot.c | 90 ++++++++++++++++++---------------- trace-usage.c | 8 ++- 6 files changed, 75 insertions(+), 49 deletions(-)
Documentation/trace-cmd-snapshot.1.txt +14 -2
@@ -41,8 +41,20 @@ OPTIONS *-B* 'buf':: If a buffer instance was created, then the *-B* option will operate on - the snapshot within the buffer. - + the snapshot within the buffer. This may be used multiple times to + specify different buffers. When this option is used, the top level + instance will be ignored unless *-t* is given. + +*-a*:: + Operate on the snapshot for all existing buffer instances. + When this option is used, the top level instance will be ignored + unless *-t* is given. + +*-t*:: + Operate on the snapshot within the top level instance buffer. Without + the *-B* or *-a* option this is the same as the default. But if *-B* + or *-a* is used, this is required if the top level instance buffer + should also be affected. SEE ALSO --------
trace-cmd.c +1 -1
@@ -72,7 +72,7 @@ void *malloc_or_die(unsigned int size)return data; } -static void dump_file_content (const char *path) +void dump_file_content (const char *path) {char buf[BUFSIZ ];ssize_t n;
trace-local.h +4 -0
@@ -47,7 +47,9 @@ struct pid_record_data {struct pevent_record *record; }; +void dump_file_content (const char *path);void show_file (const char *name); +int write_file (const char *file,const char *str,const char *type);struct tracecmd_input *read_trace_header (const char *file);int read_trace_files (void );
@@ -179,6 +181,8 @@ extern struct buffer_instance *first_instance;struct buffer_instance *create_instance (const char *name);void add_instance (struct buffer_instance *instance);char *get_instance_file (struct buffer_instance *instance,const char *file); +void add_all_instances (void ); +void update_first_instance (int topt);void show_instance_file (struct buffer_instance *instance,const char *name);int count_cpus (void );
trace-record.c +2 -2
@@ -1466,7 +1466,7 @@ static void reset_events(void)reset_events_instance (instance); } -static int write_file (const char *file,const char *str,const char *type) +int write_file (const char *file,const char *str,const char *type) {char buf[BUFSIZ ];int fd;
@@ -3778,7 +3778,7 @@ static void add_hook(struct buffer_instance *instance, const char *arg) } } -static void update_first_instance (int topt) +void update_first_instance (int topt) {if (topt || !buffer_instances) first_instance = &top_instance;
trace-snapshot.c +48 -42
128 lines not on this page · raw message
trace-usage.c +6 -2
@@ -244,13 +244,17 @@ static struct usage_help usage_help[] = { {"snapshot" ,"take snapshot of running trace" , -" %s snapshot [-s][-r][-f][-B buf][-c cpu]\n" +" %s snapshot [-s][-r][-f][-c cpu][-B buf][-a][-t]\n" " -s take a snapshot of the trace buffer\n" " -r reset current snapshot\n" " -f free the snapshot buffer\n" " without the above three options, display snapshot\n" " -c operate on the snapshot buffer for the given CPU\n" -" -B operate on the snapshot buffer for a tracing buffer instance.\n" +" -B operate on the snapshot for a given instance\n" +" (maybe be specified multiple times)\n" +" -a operate on the snapshot for all instances\n" +" -t also operate on the snapshot for top instance\n" +" (useful with -B or -a)\n" }, { "stack",
Signature · 1 line
1.9.1
PATCH v3 3/5 trace-cmd stop, reset, restart: Add -a for all instances
In reply to Howard CochranAs for extract, -a means all instances except the top one. Use -t to get the top one too.
Signed-off-by Howard Cochran <[email protected]>
--- Documentation/trace-cmd-reset.1.txt | 57 +++++++++++++++++++++++++++++++++---- Documentation/trace-cmd-stop.1.txt | 13 ++++++--- trace-record.c | 57 ++++++++++++++++++++++++++++++------- trace-usage.c | 21 ++++++++------ 4 files changed, 119 insertions(+), 29 deletions(-)
Documentation/trace-cmd-reset.1.txt +51 -6
@@ -22,6 +22,9 @@ the data in the ring buffer, and the options that were used are all lost. OPTIONS ------- +Please note that the order that options are specified on the command line is +significant. See EXAMPLES. + *-b* 'buffer_size':: When the kernel boots, the Ftrace ring buffer is of a minimal size (3 pages per CPU). The first time the tracer is used, the ring buffer size
@@ -32,6 +35,17 @@ OPTIONS trace-cmd reset -b 1 + The buffer instance affected is the one (or ones) specified by the most + recently preceding *-B*, *-t*, or *-a* option: + + When used after *-B*, resizes the buffer instance that precedes it on + the command line. + + When used after *-a*, resizes all buffer instances except the top one. + + When used after *-t* or before any *-B* or *-a*, resizes the top + instance. + *-B* 'buffer-name':: If the kernel supports multiple buffers, this will reset the trace for only the given buffer. It does not affect any other buffer. This may be
@@ -39,15 +53,46 @@ OPTIONS will not be reset if this option is given (unless the *-t* option is also supplied). +*-a*:: + Reset the trace for all existing buffer instances. When this option + is used, the top level instance will not be reset unless *-t* is given. + *-d*:: - Delete the previously specified buffer instance that was specified by - the *-B* option. It is invalid to use with *-t* as the top level instance - can not be deleted. + This option deletes the instance buffer(s) specified by the most recently + preceding *-B* or *-a* option. Because the top-level instance buffer + cannot be deleted, it is invalid to use this immediatly following *-t* or + prior to any *-B* or *-a* option on the command line. *-t*:: - Resets the top level instance buffer. Without the *-B* option this is the - same as the default. But if *-B* is used, this is required if the top - level instance buffer should also be reset. + Resets the top level instance buffer. Without the *-B* or *-a* option + this is the same as the default. But if *-B* or *-a* is used, this is + required if the top level instance buffer should also be reset. + +EXAMPLES +-------- + +Reset tracing for instance-one and set its per-cpu buffer size to 4096kb. +Also deletes instance-two. The top level instance and any other instances +remain unaffected: + + trace-cmd reset -B instance-one -b 4096 -B instance-two -d + +Delete all instance buffers. Top level instance remains unaffected: + + trace-cmd reset -a -d + +Delete all instance buffers and also reset the top instance: + + trace-cmd reset -t -a -d + +Invalid. This command implies an attempt to delete the top instance: + + trace-cmd reset -a -t -d + +Reset the top instance and set its per-cpu buffer size to 1024kb. If any +instance buffers exist, they will be unaffected: + + trace-cmd reset -b 1024 SEE ALSO
Documentation/trace-cmd-stop.1.txt +9 -4
@@ -30,12 +30,17 @@ OPTIONS *-B* 'buffer-name':: If the kernel supports multiple buffers, this will stop the trace for only the given buffer. It does not affect any other buffer. This may be - used multiple times to specify different buffers. + used multiple times to specify different buffers. When this option is + used, the top level instance will not be stopped unless *-t* is given. + +*-a*:: + Stop the trace for all existing buffer instances. When this option + is used, the top level instance will not be stopped unless *-t* is given. *-t*:: - Stops the top level instance buffer. Without the *-B* option this is the - same as the default. But if *-B* is used, this is required if the top - level instance buffer should also be stopped. + Stops the top level instance buffer. Without the *-B* or *-a* option this + is the same as the default. But if *-B* or *-a* is used, this is + required if the top level instance buffer should also be stopped. SEE ALSO --------
trace-record.c +47 -10
109 lines not on this page · raw message
trace-usage.c +12 -9
@@ -72,20 +72,22 @@ static struct usage_help usage_help[] = { {"stop" ,"stop the kernel from recording trace data" , -" %s stop [-B buf [-B buf]..] [-t]\n" +" %s stop [-B buf [-B buf]..] [-a] [-t]\n" " Stops the tracer from recording more data.\n" " Used in conjunction with start\n" -" -B stop a given buffer (more than one may be specified)\n " -" -t stop the top level buffer (needed if -B is specified)\n" +" -B stop a given buffer (more than one may be specified)\n" +" -a stop all buffers (except top one)\n" +" -t stop the top level buffer (useful with -B or -a)\n" }, {"restart" ,"restart the kernel trace data recording" , -" %s restart [-B buf [-B buf]..] [-t]\n" +" %s restart [-B buf [-B buf]..] [-a] [-t]\n" " Restarts recording after a trace-cmd stop.\n" " Used in conjunction with stop\n" -" -B restart a given buffer (more than one may be specified)\n " -" -t restart the top level buffer (needed if -B is specified)\n" +" -B restart a given buffer (more than one may be specified)\n" +" -a restart all buffers (except top one)\n" +" -t restart the top level buffer (useful with -B or -a)\n" }, { "show",
@@ -114,13 +116,14 @@ static struct usage_help usage_help[] = { {"reset" ,"disable all kernel tracing and clear the trace buffers" , -" %s reset [-b size][-B buf][-d][-t]\n" +" %s reset [-b size][-B buf][-a][-d][-t]\n" " Disables the tracer (may reset trace file)\n" " Used in conjunction with start\n" " -b change the kernel buffer size (in kilobytes per CPU)\n" -" -B reset the given buffer instance (top instance ignored)\n" " -d delete the previous specified instance\n" -" -t still reset the top instance if -B option is given\n" +" -B reset the given buffer instance (may specify multiple -B)\n" +" -a reset all instances (except top one)\n" +" -t reset the top level instance (useful with -B or -a)\n" }, { "report",
Signature · 1 line
1.9.1