perf ftrace: Add -n/--use-nsec option for latency
Namhyung Kim
maintainerPATCH 1/2 perf ftrace: Add -n/--use-nsec option for latency
Sometimes we want to see nano-second granularity.
$ sudo perf ftrace latency -T dput -a sleep 1
# DURATION | COUNT | GRAPH |
0 - 1 us | 2098375 | ############################# |
1 - 2 us | 61 | |
2 - 4 us | 33 | |
4 - 8 us | 13 | |
8 - 16 us | 124 | |
16 - 32 us | 123 | |
32 - 64 us | 1 | |
64 - 128 us | 0 | |
128 - 256 us | 1 | |
256 - 512 us | 0 | |
512 - 1024 us | 0 | |
1 - 2 ms | 0 | |
2 - 4 ms | 0 | |
4 - 8 ms | 0 | |
8 - 16 ms | 0 | |
16 - 32 ms | 0 | |
32 - 64 ms | 0 | |
64 - 128 ms | 0 | |
128 - 256 ms | 0 | |
256 - 512 ms | 0 | |
512 - 1024 ms | 0 | |
1 - ... s | 0 | | $ sudo perf ftrace latency -T dput -a -n sleep 1
# DURATION | COUNT | GRAPH |
0 - 1 us | 0 | |
1 - 2 ns | 0 | |
2 - 4 ns | 0 | |
4 - 8 ns | 0 | |
8 - 16 ns | 0 | |
16 - 32 ns | 0 | |
32 - 64 ns | 0 | |
64 - 128 ns | 1163434 | ############## |
128 - 256 ns | 914102 | ############# |
256 - 512 ns | 884 | |
512 - 1024 ns | 613 | |
1 - 2 us | 31 | |
2 - 4 us | 17 | |
4 - 8 us | 7 | |
8 - 16 us | 123 | |
16 - 32 us | 83 | |
32 - 64 us | 0 | |
64 - 128 us | 0 | |
128 - 256 us | 0 | |
256 - 512 us | 0 | |
512 - 1024 us | 0 | |
1 - ... ms | 0 | |Cc Changbin Du <[email protected]>
Signed-off-by Namhyung Kim <[email protected]>
--- tools/perf/builtin-ftrace.c | 24 +++++++++++++-------- tools/perf/util/bpf_ftrace.c | 2 ++ tools/perf/util/bpf_skel/func_latency.bpf.c | 6 ++++-- tools/perf/util/ftrace.h | 1 + 4 files changed, 22 insertions(+), 11 deletions(-)
tools/perf/builtin-ftrace.c +15 -9
@@ -680,7 +680,8 @@ static int __cmd_ftrace(struct perf_ftrace *ftrace)return (done && !workload_exec_errno) ?0 :-1 ; } -static void make_histogram (int buckets[],char *buf,size_t len,char *linebuf) +static void make_histogram (int buckets[],char *buf,size_t len,char *linebuf, +bool use_nsec) {char *p, *q;char *unit;
@@ -727,6 +728,9 @@ static void make_histogram(int buckets[], char *buf, size_t len, char *linebuf)if (!unit ||strncmp (unit," us" ,3 )) goto next; +if (use_nsec) + num *=1000 ; + i =log2 (num);if (i <0 ) i =0 ;
@@ -744,7 +748,7 @@ static void make_histogram(int buckets[], char *buf, size_t len, char *linebuf)strcat (linebuf, p); } -static void display_histogram (int buckets[]) +static void display_histogram (int buckets[],bool use_nsec) {int i;int total =0 ;
@@ -770,12 +774,12 @@ static void display_histogram(int buckets[])for (i =1 ; i <NUM_BUCKET -1 ; i++) {int start = (1 << (i -1 ));int stop =1 << i; -const char *unit ="us" ; +const char *unit = use_nsec ?"ns" :"us" ;if (start >=1024 ) { start >>=10 ; stop >>=10 ; - unit ="ms" ; + unit = use_nsec ?"us" :"ms" ; } bar_len = buckets[i] * bar_total / total; printf(" %4d - %-4d %s | %10d | %.*s%*s |\n",
@@ -785,8 +789,8 @@ static void display_histogram(int buckets[]) bar_len = buckets[NUM_BUCKET -1 ] * bar_total / total;printf (" %4d - %-4s %s | %10d | %.*s%*s |\n" , -1 ,"..." ," s" , buckets[NUM_BUCKET -1 ], bar_len, bar, - bar_total - bar_len,"" ); +1 ,"..." , use_nsec ?"ms" :" s" , buckets[NUM_BUCKET -1 ], + bar_len, bar, bar_total - bar_len,"" ); }
@@ -913,7 +917,7 @@ static int __cmd_latency(struct perf_ftrace *ftrace)if (n <0 )break ; -make_histogram (buckets, buf, n, line); +make_histogram (buckets, buf, n, line, ftrace->use_nsec); } }
@@ -930,12 +934,12 @@ static int __cmd_latency(struct perf_ftrace *ftrace)int n =read (trace_fd, buf,sizeof (buf) -1 );if (n <=0 )break ; -make_histogram (buckets, buf, n, line); +make_histogram (buckets, buf, n, line, ftrace->use_nsec); }read_func_latency (ftrace, buckets); -display_histogram (buckets); +display_histogram (buckets, ftrace->use_nsec); out:close (trace_fd);
@@ -1171,6 +1175,8 @@ int cmd_ftrace(int argc, const char **argv)OPT_BOOLEAN ('b' ,"use-bpf" , &ftrace.target.use_bpf,"Use BPF to measure function latency" ), #endif +OPT_BOOLEAN ('n' ,"--use-nsec" , &ftrace.use_nsec, +"Use nano-second histogram" ), OPT_PARENT(common_options), };const struct option *options = ftrace_options;
tools/perf/util/bpf_ftrace.c +2 -0
@@ -81,6 +81,8 @@ int perf_ftrace__latency_prepare_bpf(struct perf_ftrace *ftrace) } } + skel->bss->use_nsec = ftrace->use_nsec; + skel->links.func_begin =bpf_program__attach_kprobe (skel->progs.func_begin, false, func->name);if (IS_ERR (skel->links.func_begin)) {
tools/perf/util/bpf_skel/func_latency.bpf.c +4 -2
@@ -39,6 +39,7 @@ struct {int enabled =0 ;int has_cpu =0 ;int has_task =0 ; +int use_nsec =0 ; SEC("kprobe/func")int BPF_PROG (func_begin )
@@ -80,6 +81,7 @@ int BPF_PROG(func_end) {__u64 tid;__u64 *start; +__u64 cmp_base = use_nsec ?1 :1000 ;if (!enabled)return 0 ;
@@ -97,9 +99,9 @@ int BPF_PROG(func_end)if (delta <0 )return 0 ; -// calculate index using delta in usec +// calculate index using delta for (key =0 ; key < (NUM_BUCKET -1 ); key++) { -if (delta < ((1000UL ) << key)) +if (delta < (cmp_base << key))break ; }
tools/perf/util/ftrace.h +1 -0
@@ -17,6 +17,7 @@ struct perf_ftrace {struct list_head nograph_funcs;unsigned long percpu_buffer_size;bool inherit; +bool use_nsec;int graph_depth;int func_stack_trace;int func_irq_info;
Signature · 1 line
2.35.1.894.gb6a874cedc-goog
Namhyung Kim
maintainerPATCH 2/2 perf ftrace: Update documentation of ftrace command
Add description of perf ftrace latency subcommand.
Cc Changbin Du <[email protected]>
Signed-off-by Namhyung Kim <[email protected]>
--- tools/perf/Documentation/perf-ftrace.txt | 75 ++++++++++++++++-------- 1 file changed, 52 insertions(+), 23 deletions(-)
tools/perf/Documentation/perf-ftrace.txt +52 -23
@@ -9,32 +9,24 @@ perf-ftrace - simple wrapper for kernel's ftrace functionality SYNOPSIS -------- [verse] -'perf ftrace' <command> +'perf ftrace' {trace|latency} <command> DESCRIPTION ----------- -The 'perf ftrace' command is a simple wrapper of kernel's ftrace -functionality. It only supports single thread tracing currently and -just reads trace_pipe in text and then write it to stdout. +The 'perf ftrace' command provides a collection of subcommands which use +kernel's ftrace infrastructure. -The following options apply to perf ftrace. + 'perf ftrace trace' is a simple wrapper of the ftrace. It only supports + single thread tracing currently and just reads trace_pipe in text and then + write it to stdout. -OPTIONS -------- + 'perf ftrace latency' calculates execution latency of a given function + (optionally with BPF) and display it as a histogram. --t:: ---tracer=:: - Tracer to use when neither -G nor -F option is not - specified: function_graph or function. +The following options apply to perf ftrace. --v:: ---verbose:: - Increase the verbosity level. - --F:: ---funcs:: - List available functions to trace. It accepts a pattern to - only list interested functions. +COMMON OPTIONS +-------------- -p:: --pid=::
@@ -43,10 +35,6 @@ OPTIONS --tid=:: Trace on existing thread id (comma separated list). --D:: ---delay:: - Time (ms) to wait before starting tracing after program start. - -a:: --all-cpus:: Force system-wide collection. Scripts run without a <command>
@@ -61,6 +49,28 @@ OPTIONS Ranges of CPUs are specified with -: 0-2. Default is to trace on all online CPUs. +-v:: +--verbose:: + Increase the verbosity level. + + +OPTIONS for 'perf ftrace trace' +------------------------------- + +-t:: +--tracer=:: + Tracer to use when neither -G nor -F option is not + specified: function_graph or function. + +-F:: +--funcs:: + List available functions to trace. It accepts a pattern to + only list interested functions. + +-D:: +--delay:: + Time (ms) to wait before starting tracing after program start. + -m:: --buffer-size:: Set the size of per-cpu tracing buffer, <size> is expected to
@@ -114,6 +124,25 @@ OPTIONS thresh=<n> - Setup trace duration threshold in microseconds. depth=<n> - Set max depth for function graph tracer to follow. + +OPTIONS for 'perf ftrace latency' +--------------------------------- + +-T:: +--trace-funcs=:: + Set the function name to get the histogram. Unlike perf ftrace trace, + it only allows single function to calculate the histogram. + +-b:: +--use-bpf:: + Use BPF to measure function latency instead of using the ftrace (it + uses function_graph tracer internally). + +-n:: +--use-nsec:: + Use nano-second instead of micro-second as a base unit of the histogram. + + SEE ALSO -------- linkperf:perf-record[1], linkperf:perf-trace[1]
Signature · 1 line
2.35.1.894.gb6a874cedc-goog
Em Mon, Mar 21, 2022 at 04:46:08PM -0700, Namhyung Kim escreveu:
Quoted 11 lines
Sometimes we want to see nano-second granularity.
$ sudo perf ftrace latency -T dput -a -n sleep 1 # DURATION | COUNT | GRAPH | 0 - 1 us | 0 | | 1 - 2 ns | 0 | | 2 - 4 ns | 0 | | 4 - 8 ns | 0 | | 8 - 16 ns | 0 | | 16 - 32 ns | 0 | | 32 - 64 ns | 0 | |
Quoted 8 lines
64 - 128 ns | 1163434 | ############## |
128 - 256 ns | 914102 | ############# |
256 - 512 ns | 884 | |
512 - 1024 ns | 613 | |
1 - 2 us | 31 | |
2 - 4 us | 17 | |
4 - 8 us | 7 | |
8 - 16 us | 123 | |Thanks, applied.
- Arnaldo