perf trace: Implement syscall summary in BPF
Subsystems (2)
Namhyung Kim
maintainerPATCH v4 1/2 perf trace: Implement syscall summary in BPF
When -s/--summary option is used, it doesn't need (augmented) arguments of syscalls. Let's skip the augmentation and load another small BPF program to collect the statistics in the kernel instead of copying the data to the ring-buffer to calculate the stats in userspace. This will be much more light-weight than the existing approach and remove any lost events.
Let's add a new option --bpf-summary to control this behavior. I cannot make it default because there's no way to get e_machine in the BPF which is needed for detecting different ABIs like 32-bit compat mode.
No functional changes intended except for no more LOST events. :)
$ sudo ./perf trace -as --summary-mode=total --bpf-summary sleep 1
Summary of events:
total, 6194 events
syscall calls errors total min avg max stddev
(msec) (msec) (msec) (msec) (%)
--------------- -------- ------ -------- --------- --------- --------- ------
epoll_wait 561 0 4530.843 0.000 8.076 520.941 18.75%
futex 693 45 4317.231 0.000 6.230 500.077 21.98%
poll 300 0 1040.109 0.000 3.467 120.928 17.02%
clock_nanosleep 1 0 1000.172 1000.172 1000.172 1000.172 0.00%
ppoll 360 0 872.386 0.001 2.423 253.275 41.91%
epoll_pwait 14 0 384.349 0.001 27.453 380.002 98.79%
pselect6 14 0 108.130 7.198 7.724 8.206 0.85%
nanosleep 39 0 43.378 0.069 1.112 10.084 44.23%
...---
v4)
* fix segfault on -S (Howard)
* correct some comments (Howard)
v3)
* support -S/--with-summary option too (Howard)
* make it work only with -a/--all-cpus (Howard)
* fix stddev calculation (Howard)
* add some comments about syscall_data (Howard)
v2)
* Rebased on top of Ian's e_machine changes
* add --bpf-summary option
* support per-thread summary
* add stddev calculation (Howard)
tools/perf/Documentation/perf-trace.txt | 6 + tools/perf/Makefile.perf | 2 +- tools/perf/builtin-trace.c | 54 ++- tools/perf/util/Build | 1 + tools/perf/util/bpf-trace-summary.c | 347 ++++++++++++++++++ .../perf/util/bpf_skel/syscall_summary.bpf.c | 118 ++++++ tools/perf/util/bpf_skel/syscall_summary.h | 25 ++ tools/perf/util/trace.h | 37 ++ 8 files changed, 577 insertions(+), 13 deletions(-) create mode 100644 tools/perf/util/bpf-trace-summary.c create mode 100644 tools/perf/util/bpf_skel/syscall_summary.bpf.c create mode 100644 tools/perf/util/bpf_skel/syscall_summary.h create mode 100644 tools/perf/util/trace.h
tools/perf/Documentation/perf-trace.txt +6 -0
@@ -251,6 +251,12 @@ the thread executes on the designated CPUs. Default is to monitor all CPUs. pretty-printing serves as a fallback to hand-crafted pretty printers, as the latter can better pretty-print integer flags and struct pointers. +--bpf-summary:: + Collect system call statistics in BPF. This is only for live mode and + works well with -s/--summary option where no argument information is + required. + + PAGEFAULTS ----------
tools/perf/Makefile.perf +1 -1
@@ -1216,7 +1216,7 @@ SKELETONS += $(SKEL_OUT)/bperf_leader.skel.h $(SKEL_OUT)/bperf_follower.skel.h SKELETONS += $(SKEL_OUT)/bperf_cgroup.skel.h $(SKEL_OUT)/func_latency.skel.h SKELETONS += $(SKEL_OUT)/off_cpu.skel.h $(SKEL_OUT)/lock_contention.skel.h SKELETONS += $(SKEL_OUT)/kwork_trace.skel.h $(SKEL_OUT)/sample_filter.skel.h -SKELETONS += $(SKEL_OUT)/kwork_top.skel.h +SKELETONS += $(SKEL_OUT)/kwork_top.skel.h $(SKEL_OUT)/syscall_summary.skel.h SKELETONS += $(SKEL_OUT)/bench_uprobe.skel.h SKELETONS += $(SKEL_OUT)/augmented_raw_syscalls.skel.h
tools/perf/builtin-trace.c +42 -12
143 lines not on this page · raw message
tools/perf/util/Build +1 -0
@@ -171,6 +171,7 @@ perf-util-$(CONFIG_PERF_BPF_SKEL) += bpf_off_cpu.o perf-util-$(CONFIG_PERF_BPF_SKEL) += bpf-filter.o perf-util-$(CONFIG_PERF_BPF_SKEL) += bpf-filter-flex.o perf-util-$(CONFIG_PERF_BPF_SKEL) += bpf-filter-bison.o +perf-util-$(CONFIG_PERF_BPF_SKEL) += bpf-trace-summary.o perf-util-$(CONFIG_PERF_BPF_SKEL) += btf.o ifeq ($(CONFIG_LIBTRACEEVENT),y)
tools/perf/util/bpf-trace-summary.c new file +347 -0
347 lines not on this page · raw message
tools/perf/util/bpf_skel/syscall_summary.bpf.c new file +118 -0
118 lines not on this page · raw message
tools/perf/util/bpf_skel/syscall_summary.h new file +25 -0
@@ -0,0 +1,25 @@ +// SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause) +/* Data structures shared between BPF and tools. */ +#ifndef UTIL_BPF_SKEL_SYSCALL_SUMMARY_H +#define UTIL_BPF_SKEL_SYSCALL_SUMMARY_H + +enum syscall_aggr_mode { +SYSCALL_AGGR_THREAD , +SYSCALL_AGGR_CPU , +}; + +struct syscall_key { +int cpu_or_tid; +int nr; +}; + +struct syscall_stats { +u64 total_time; +u64 squared_sum; +u64 max_time; +u64 min_time; +u32 count; +u32 error; +}; + +#endif /* UTIL_BPF_SKEL_SYSCALL_SUMMARY_H */
tools/perf/util/trace.h new file +37 -0
@@ -0,0 +1,37 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +#ifndef UTIL_TRACE_H +#define UTIL_TRACE_H + +#include <stdio.h> /* for FILE */ + +enum trace_summary_mode { +SUMMARY__NONE =0 , +SUMMARY__BY_TOTAL , +SUMMARY__BY_THREAD , +}; + +#ifdef HAVE_BPF_SKEL + +int trace_prepare_bpf_summary (enum trace_summary_mode mode); +void trace_start_bpf_summary (void ); +void trace_end_bpf_summary (void ); +int trace_print_bpf_summary (FILE *fp); +void trace_cleanup_bpf_summary (void ); + +#else /* !HAVE_BPF_SKEL */ + +static inline int trace_prepare_bpf_summary (enum trace_summary_mode mode__maybe_unused ) +{ +return -1 ; +} +static inline void trace_start_bpf_summary (void ) {} +static inline void trace_end_bpf_summary (void ) {} +static inline int trace_print_bpf_summary (FILE *fp__maybe_unused ) +{ +return 0 ; +} +static inline void trace_cleanup_bpf_summary (void ) {} + +#endif /* HAVE_BPF_SKEL */ + +#endif /* UTIL_TRACE_H */
Signature · 1 line
2.49.0.395.g12beb8f557-goog
Namhyung Kim
maintainerPATCH v4 2/2 perf test: Add perf trace summary test
$ sudo ./perf test -vv 'trace summary' 109: perf trace summary: --- start --- test child forked, pid 3501572 testing: perf trace -s -- true testing: perf trace -S -- true testing: perf trace -s --summary-mode=thread -- true testing: perf trace -S --summary-mode=total -- true testing: perf trace -as --summary-mode=thread --no-bpf-summary -- true testing: perf trace -as --summary-mode=total --no-bpf-summary -- true testing: perf trace -as --summary-mode=thread --bpf-summary -- true testing: perf trace -as --summary-mode=total --bpf-summary -- true testing: perf trace -aS --summary-mode=total --bpf-summary -- true ---- end(0) ---- 109: perf trace summary : Ok
--- tools/perf/tests/shell/trace_summary.sh | 65 +++++++++++++++++++++++++ 1 file changed, 65 insertions(+) create mode 100755 tools/perf/tests/shell/trace_summary.sh
tools/perf/tests/shell/trace_summary.sh new file +65 -0
@@ -0,0 +1,65 @@ +#!/bin/sh +# perf trace summary +# SPDX-License-Identifier: GPL-2.0 + +# Check that perf trace works with various summary mode + +# shellcheck source=lib/probe.sh +. "$(dirname $0)"/lib/probe.sh + +skip_if_no_perf_trace || exit 2 +[ "$(id -u)" = 0 ] || exit 2 + +OUTPUT=$(mktemp /tmp/perf_trace_test.XXXXX) + +test_perf_trace() { + args=$1 + workload="true" + search="^\s*(open|read|close).*[0-9]+%$" + + echo "testing: perf trace ${args} -- ${workload}" + perf trace ${args} -- ${workload} >${OUTPUT} 2>&1 + if [ $? -ne 0 ]; then + echo "Error: perf trace ${args} failed unexpectedly" + cat ${OUTPUT} + rm -f ${OUTPUT} + exit 1 + fi + + count=$(grep -E -c -m 3 "${search}" ${OUTPUT}) + if [ "${count}" != "3" ]; then + echo "Error: cannot find enough pattern ${search} in the output" + cat ${OUTPUT} + rm -f ${OUTPUT} + exit 1 + fi +} + +# summary only for a process +test_perf_trace "-s" + +# normal output with summary at the end +test_perf_trace "-S" + +# summary only with an explicit summary mode +test_perf_trace "-s --summary-mode=thread" + +# summary with normal output - total summary mode +test_perf_trace "-S --summary-mode=total" + +# summary only for system wide - per-thread summary +test_perf_trace "-as --summary-mode=thread --no-bpf-summary" + +# summary only for system wide - total summary mode +test_perf_trace "-as --summary-mode=total --no-bpf-summary" + +# summary only for system wide - per-thread summary with BPF +test_perf_trace "-as --summary-mode=thread --bpf-summary" + +# summary only for system wide - total summary mode with BPF +test_perf_trace "-as --summary-mode=total --bpf-summary" + +# summary with normal output for system wide - total summary mode with BPF +test_perf_trace "-aS --summary-mode=total --bpf-summary" + +rm -f ${OUTPUT}
Signature · 1 line
2.49.0.395.g12beb8f557-goog
Re: [PATCH v4 2/2] perf test: Add perf trace summary test
Hello,
Quoted 96 lines · Namhyung Kim
$ sudo ./perf test -vv 'trace summary' 109: perf trace summary: --- start --- test child forked, pid 3501572 testing: perf trace -s -- true testing: perf trace -S -- true testing: perf trace -s --summary-mode=thread -- true testing: perf trace -S --summary-mode=total -- true testing: perf trace -as --summary-mode=thread --no-bpf-summary -- true testing: perf trace -as --summary-mode=total --no-bpf-summary -- true testing: perf trace -as --summary-mode=thread --bpf-summary -- true testing: perf trace -as --summary-mode=total --bpf-summary -- true testing: perf trace -aS --summary-mode=total --bpf-summary -- true ---- end(0) ---- 109: perf trace summary : OkCc Howard Chu <[email protected]>Signed-off-by Namhyung Kim <[email protected]>--- tools/perf/tests/shell/trace_summary.sh | 65 +++++++++++++++++++++++++ 1 file changed, 65 insertions(+) create mode 100755 tools/perf/tests/shell/trace_summary.shtools/perf/tests/shell/trace_summary.sh new file +65
-1@@ -0,0 +1,65 @@ +#!/bin/sh +# perf trace summary +# SPDX-License-Identifier: GPL-2.0 + +# Check that perf trace works with various summary mode + +# shellcheck source=lib/probe.sh +. "$(dirname $0)"/lib/probe.sh + +skip_if_no_perf_trace || exit 2 +[ "$(id -u)" = 0 ] || exit 2 + +OUTPUT=$(mktemp /tmp/perf_trace_test.XXXXX) + +test_perf_trace() { + args=$1 + workload="true" + search="^\s*(open|read|close).*[0-9]+%$" + + echo "testing: perf trace ${args} -- ${workload}" + perf trace ${args} -- ${workload} >${OUTPUT} 2>&1 + if [ $? -ne 0 ]; then + echo "Error: perf trace ${args} failed unexpectedly" + cat ${OUTPUT} + rm -f ${OUTPUT} + exit 1 + fi + + count=$(grep -E -c -m 3 "${search}" ${OUTPUT}) + if [ "${count}" != "3" ]; then + echo "Error: cannot find enough pattern ${search} in the output" + cat ${OUTPUT} + rm -f ${OUTPUT} + exit 1 + fi +} + +# summary only for a process +test_perf_trace "-s" + +# normal output with summary at the end +test_perf_trace "-S" + +# summary only with an explicit summary mode +test_perf_trace "-s --summary-mode=thread" + +# summary with normal output - total summary mode +test_perf_trace "-S --summary-mode=total" + +# summary only for system wide - per-thread summary +test_perf_trace "-as --summary-mode=thread --no-bpf-summary" + +# summary only for system wide - total summary mode +test_perf_trace "-as --summary-mode=total --no-bpf-summary" + +# summary only for system wide - per-thread summary with BPF +test_perf_trace "-as --summary-mode=thread --bpf-summary" + +# summary only for system wide - total summary mode with BPF +test_perf_trace "-as --summary-mode=total --bpf-summary" + +# summary with normal output for system wide - total summary mode with BPF +test_perf_trace "-aS --summary-mode=total --bpf-summary" + +rm -f ${OUTPUT} --
Quoted 2 lines
2.49.0.395.g12beb8f557-goog
Didn't quite get the combinatorial logic but it sure covers a lot :)
Thanks,
Howard
Hello Namhyung,
Quoted 33 lines · Namhyung Kim
When -s/--summary option is used, it doesn't need (augmented) arguments of syscalls. Let's skip the augmentation and load another small BPF program to collect the statistics in the kernel instead of copying the data to the ring-buffer to calculate the stats in userspace. This will be much more light-weight than the existing approach and remove any lost events.
Let's add a new option --bpf-summary to control this behavior. I cannot make it default because there's no way to get
e_machinein the BPF which is needed for detecting different ABIs like 32-bit compat mode.No functional changes intended except for no more LOST events. :)
$ sudo ./perf trace -as --summary-mode=total --bpf-summary sleep 1Summary of events: total, 6194 events syscall calls errors total min avg max stddev (msec) (msec) (msec) (msec) (%) --------------- -------- ------ -------- --------- --------- --------- ------ epoll_wait 561 0 4530.843 0.000 8.076 520.941 18.75% futex 693 45 4317.231 0.000 6.230 500.077 21.98% poll 300 0 1040.109 0.000 3.467 120.928 17.02% clock_nanosleep 1 0 1000.172 1000.172 1000.172 1000.172 0.00% ppoll 360 0 872.386 0.001 2.423 253.275 41.91% epoll_pwait 14 0 384.349 0.001 27.453 380.002 98.79% pselect6 14 0 108.130 7.198 7.724 8.206 0.85% nanosleep 39 0 43.378 0.069 1.112 10.084 44.23% ...
Quoted 6 lines
Cc Howard Chu <[email protected]>Signed-off-by Namhyung Kim <[email protected]>---
v4)
* fix segfault on -S (Howard)
* correct some comments (Howard)
+ if (!hashmap__find(hash, map_key->nr, &data)) {
I think you should mention the hashmap's map_key->nr update, as this change is actually important for the feature.
Quoted 32 lines
v3)
* support -S/--with-summary option too (Howard)
* make it work only with -a/--all-cpus (Howard)
* fix stddev calculation (Howard)
* add some comments aboutsyscall_data(Howard)v2)
* Rebased on top of Ian'se_machinechanges
* add --bpf-summary option
* support per-thread summary
* add stddev calculation (Howard)tools/perf/Documentation/perf-trace.txt | 6 + tools/perf/Makefile.perf | 2 +- tools/perf/builtin-trace.c | 54 ++- tools/perf/util/Build | 1 + tools/perf/util/bpf-trace-summary.c | 347 ++++++++++++++++++ .../perf/util/bpf_skel/syscall_summary.bpf.c | 118 ++++++ tools/perf/util/bpf_skel/syscall_summary.h | 25 ++ tools/perf/util/trace.h | 37 ++ 8 files changed, 577 insertions(+), 13 deletions(-) create mode 100644 tools/perf/util/bpf-trace-summary.c create mode 100644 tools/perf/util/bpf_skel/syscall_summary.bpf.c create mode 100644 tools/perf/util/bpf_skel/syscall_summary.h create mode 100644 tools/perf/util/trace.htools/perf/Documentation/perf-trace.txt +0
-0
tools/perf/Documentation/perf-trace.txt the thread executes on the designated CPUs. Default is to monitor all CPUs. · line 257
Quoted 7 lines
tools/perf/Documentation/perf-trace.txt +4
-0@@ -251,6 +251,12 @@ the thread executes on the designated CPUs. Default is to monitor all CPUs. pretty-printing serves as a fallback to hand-crafted pretty printers, as the latter can better pretty-print integer flags and struct pointers. +--bpf-summary:: + Collect system call statistics in BPF. This is only for live mode and + works well with -s/--summary option where no argument information is + required.
It works with -S as well, doesn't it?
Anyway, I don't mind adding these details later on, so
Quoted 83 lines · Howard Chu
Hello Namhyung,
Quoted 33 lines · Namhyung Kim
When -s/--summary option is used, it doesn't need (augmented) arguments of syscalls. Let's skip the augmentation and load another small BPF program to collect the statistics in the kernel instead of copying the data to the ring-buffer to calculate the stats in userspace. This will be much more light-weight than the existing approach and remove any lost events.
Let's add a new option --bpf-summary to control this behavior. I cannot make it default because there's no way to get
e_machinein the BPF which is needed for detecting different ABIs like 32-bit compat mode.No functional changes intended except for no more LOST events. :)
$ sudo ./perf trace -as --summary-mode=total --bpf-summary sleep 1Summary of events: total, 6194 events syscall calls errors total min avg max stddev (msec) (msec) (msec) (msec) (%) --------------- -------- ------ -------- --------- --------- --------- ------ epoll_wait 561 0 4530.843 0.000 8.076 520.941 18.75% futex 693 45 4317.231 0.000 6.230 500.077 21.98% poll 300 0 1040.109 0.000 3.467 120.928 17.02% clock_nanosleep 1 0 1000.172 1000.172 1000.172 1000.172 0.00% ppoll 360 0 872.386 0.001 2.423 253.275 41.91% epoll_pwait 14 0 384.349 0.001 27.453 380.002 98.79% pselect6 14 0 108.130 7.198 7.724 8.206 0.85% nanosleep 39 0 43.378 0.069 1.112 10.084 44.23% ...Quoted 6 lines
Cc Howard Chu <[email protected]>Signed-off-by Namhyung Kim <[email protected]>---
v4)
* fix segfault on -S (Howard)
* correct some comments (Howard)+ if (!hashmap__find(hash,
map_key->nr, &data)) {I think you should mention the hashmap's
map_key->nrupdate, as this change is actually important for the feature.Quoted 35 lines
v3)
* support -S/--with-summary option too (Howard)
* make it work only with -a/--all-cpus (Howard)
* fix stddev calculation (Howard)
* add some comments aboutsyscall_data(Howard)v2)
* Rebased on top of Ian'se_machinechanges
* add --bpf-summary option
* support per-thread summary
* add stddev calculation (Howard)tools/perf/Documentation/perf-trace.txt | 6 + tools/perf/Makefile.perf | 2 +- tools/perf/builtin-trace.c | 54 ++- tools/perf/util/Build | 1 + tools/perf/util/bpf-trace-summary.c | 347 ++++++++++++++++++ .../perf/util/bpf_skel/syscall_summary.bpf.c | 118 ++++++ tools/perf/util/bpf_skel/syscall_summary.h | 25 ++ tools/perf/util/trace.h | 37 ++ 8 files changed, 577 insertions(+), 13 deletions(-) create mode 100644 tools/perf/util/bpf-trace-summary.c create mode 100644 tools/perf/util/bpf_skel/syscall_summary.bpf.c create mode 100644 tools/perf/util/bpf_skel/syscall_summary.h create mode 100644 tools/perf/util/trace.htools/perf/Documentation/perf-trace.txt +0
-0@@ -251,6 +251,12 @@ the thread executes on the designated CPUs. Default is to monitor all CPUs. pretty-printing serves as a fallback to hand-crafted pretty printers, as the latter can better pretty-print integer flags and struct pointers.
Quoted 4 lines
tools/perf/Documentation/perf-trace.txt the thread executes on the designated CPUs. Default is to monitor all CPUs. · line 257
Quoted 4 lines
tools/perf/Documentation/perf-trace.txt +4
-0+--bpf-summary:: + Collect system call statistics in BPF. This is only for live mode and + works well with -s/--summary option where no argument information is + required.
root@number:~#>
Quoted 1 line
It works with -S as well, doesn't it?
Yes, I tested it:
root@number:~# perf trace -aS --summary-mode=total --bpf-summary sleep 0.000000001
0.011 ( 0.008 ms): :146484/146484 execve(filename: "/home/acme/libexec/perf-core/sleep", argv: 0x7ffcdf2108f0, envp: 0x37fabf70) = -1 ENOENT (No such file or directory)
0.021 ( 0.002 ms): :146484/146484 execve(filename: "/root/.local/bin/sleep", argv: 0x7ffcdf2108f0, envp: 0x37fabf70) = -1 ENOENT (No such file or directory)
0.024 ( 0.002 ms): :146484/146484 execve(filename: "/root/bin/sleep", argv: 0x7ffcdf2108f0, envp: 0x37fabf70) = -1 ENOENT (No such file or directory)
0.026 ( 0.002 ms): :146484/146484 execve(filename: "/usr/local/sbin/sleep", argv: 0x7ffcdf2108f0, envp: 0x37fabf70) = -1 ENOENT (No such file or directory)
0.029 ( 0.001 ms): :146484/146484 execve(filename: "/usr/local/bin/sleep", argv: 0x7ffcdf2108f0, envp: 0x37fabf70) = -1 ENOENT (No such file or directory)
? ( ): sudo/115804 ... [continued]: ppoll()) = 1
0.032 ( ): :146484/146484 execve(filename: "/usr/sbin/sleep", argv: 0x7ffcdf2108f0, envp: 0x37fabf70) ...
0.146 ( 0.002 ms): sudo/115804 rt_sigaction(sig: TTIN, act: (struct sigaction){.sa_handler = (__sighandler_t)0x557bdbdec4c0,.sa_flags = (long unsigned int)67108864,.sa_restorer = (__sigrestore_t)0x7f50c6627bf0,}, oact: 0x7ffff79f7d80, sigsetsize: 8) = 0
0.150 ( 0.003 ms): sudo/115804 read(fd: 9</dev/ptmx>, buf: 0x557be6008260, count: 65536) = 297
0.155 ( 0.001 ms): sudo/115804 rt_sigaction(sig: TTIN, act: (struct sigaction){.sa_handler = (__sighandler_t)0x1,.sa_flags = (long unsigned int)335544320,.sa_restorer = (__sigrestore_t)0x7f50c6627bf0,}, sigsetsize: 8) = 0
0.158 ( 0.001 ms): sudo/115804 rt_sigprocmask(nset: 0x557bdbe1a6a0, oset: 0x7ffff79f7d70, sigsetsize: 8) = 0
0.162 ( 0.001 ms): sudo/115804 rt_sigprocmask(how: SETMASK, nset: 0x7ffff79f7d70, sigsetsize: 8) = 0
0.165 ( 0.002 ms): sudo/115804 ppoll(ufds: 0x557be5f955b0, nfds: 5, sigsetsize: 8) = 2
0.169 ( 0.001 ms): sudo/115804 rt_sigaction(sig: TTIN, act: (struct sigaction){.sa_handler = (__sighandler_t)0x557bdbdec4c0,.sa_flags = (long unsigned int)67108864,.sa_restorer = (__sigrestore_t)0x7f50c6627bf0,}, oact: 0x7ffff79f7d80, sigsetsize: 8) = 0
0.171 ( 0.002 ms): sudo/115804 read(fd: 9</dev/ptmx>, buf: 0x557be6008389, count: 65239) = 502
0.175 ( 0.001 ms): sudo/115804 rt_sigaction(sig: TTIN, act: (struct sigaction){.sa_handler = (__sighandler_t)0x1,.sa_flags = (long unsigned int)335544320,.sa_restorer = (__sigrestore_t)0x7f50c6627bf0,}, sigsetsize: 8) = 0
0.177 ( 0.001 ms): sudo/115804 rt_sigprocmask(nset: 0x557bdbe1a6a0, oset: 0x7ffff79f7d70, sigsetsize: 8) = 0
0.179 ( 0.001 ms): sudo/115804 rt_sigprocmask(how: SETMASK, nset: 0x7ffff79f7d70, sigsetsize: 8) = 0
0.181 ( 0.001 ms): sudo/115804 rt_sigaction(sig: TTOU, act: (struct sigaction){.sa_handler = (__sighandler_t)0x557bdbdec4d0,.sa_flags = (long unsigned int)67108864,.sa_restorer = (__sigrestore_t)0x7f50c6627bf0,}, oact: 0x7ffff79f7d80, sigsetsize: 8) = 0
0.183 ( 0.004 ms): sudo/115804 write(fd: 8</dev/tty>, buf: ? ( ): :146484/, count: 799) = 799
0.189 ( 0.001 ms): sudo/115804 rt_sigaction(sig: TTOU, act: (struct sigaction){.sa_handler = (__sighandler_t)0x1,.sa_flags = (long unsigned int)335544320,.sa_restorer = (__sigrestore_t)0x7f50c6627bf0,}, sigsetsize: 8) = 0
0.193 ( 0.002 ms): sudo/115804 ppoll(ufds: 0x557be5f955b0, nfds: 4, sigsetsize: 8) = 1
0.196 ( 0.001 ms): sudo/115804 rt_sigaction(sig: TTIN, act: (struct sigaction){.sa_handler = (__sighandler_t)0x557bdbdec4c0,.sa_flags = (long unsigned int)67108864,.sa_restorer = (__sigrestore_t)0x7f50c6627bf0,}, oact: 0x7ffff79f7d80, sigsetsize: 8) = 0
0.199 ( 0.002 ms): sudo/115804 read(fd: 9</dev/ptmx>, buf: 0x557be6008260, count: 65536) = 379
0.201 ( 0.001 ms): sudo/115804 rt_sigaction(sig: TTIN, act: (struct sigaction){.sa_handler = (__sighandler_t)0x1,.sa_flags = (long unsigned int)335544320,.sa_restorer = (__sigrestore_t)0x7f50c6627bf0,}, sigsetsize: 8) = 0
0.203 ( 0.001 ms): sudo/115804 rt_sigprocmask(nset: 0x557bdbe1a6a0, oset: 0x7ffff79f7d70, sigsetsize: 8) = 0
0.205 ( 0.001 ms): sudo/115804 rt_sigprocmask(how: SETMASK, nset: 0x7ffff79f7d70, sigsetsize: 8) = 0
0.206 ( 0.002 ms): sudo/115804 ppoll(ufds: 0x557be5f955b0, nfds: 5, sigsetsize: 8) = 1
0.209 ( 0.001 ms): sudo/115804 rt_sigaction(sig: TTOU, act: (struct sigaction){.sa_handler = (__sighandler_t)0x557bdbdec4d0,.sa_flags = (long unsigned int)67108864,.sa_restorer = (__sigrestore_t)0x7f50c6627bf0,}, oact: 0x7ffff79f7d80, sigsetsize: 8) = 0
0.211 ( 0.002 ms): sudo/115804 write(fd: 8</dev/tty>, buf: ): :146484/146484 execve(filenam, count: 379) = 379
0.213 ( 0.001 ms): sudo/115804 rt_sigaction(sig: TTOU, act: (struct sigaction){.sa_handler = (__sighandler_t)0x1,.sa_flags = (long unsigned int)335544320,.sa_restorer = (__sigrestore_t)0x7f50c6627bf0,}, sigsetsize: 8) = 0
? ( ): ptyxis/3622 ... [continued]: ppoll()) = 1
0.215 ( ): sudo/115804 ppoll(ufds: 0x557be5f955b0, nfds: 4, sigsetsize: 8) ...
0.196 ( 0.002 ms): ptyxis/3622 write(fd: 4<anon_inode:[eventfd]>, buf: \1\0\0\0\0\0\0\0, count: 8) = 8
0.206 ( 0.003 ms): ptyxis/3622 read(fd: 41</dev/ptmx>, buf: 0x5586a84ee428, count: 8136) = 800
0.209 ( 0.001 ms): ptyxis/3622 read(fd: 41</dev/ptmx>, buf: 0x5586a84ee747, count: 7337) = -1 EAGAIN (Resource temporarily unavailable)
0.221 ( 0.001 ms): ptyxis/3622 write(fd: 4<anon_inode:[eventfd]>, buf: \1\0\0\0\0\0\0\0, count: 8) = 8
0.224 ( 0.002 ms): ptyxis/3622 ppoll(ufds: 0x5586a7e8f120, nfds: 10, tsp: 0x7ffdd1fbb470, sigsetsize: 8) = 2
0.227 ( 0.001 ms): ptyxis/3622 read(fd: 4<anon_inode:[eventfd]>, buf: 0x7ffdd1fbb3a0, count: 8) = 8
0.229 ( 0.001 ms): ptyxis/3622 write(fd: 4<anon_inode:[eventfd]>, buf: \1\0\0\0\0\0\0\0, count: 8) = 8
0.231 ( 0.001 ms): ptyxis/3622 read(fd: 41</dev/ptmx>, buf: 0x5586a84ee747, count: 7337) = 380
0.233 ( 0.001 ms): ptyxis/3622 read(fd: 41</dev/ptmx>, buf: 0x5586a84ee8c2, count: 6958) = -1 EAGAIN (Resource temporarily unavailable)
0.234 ( 0.001 ms): ptyxis/3622 write(fd: 4<anon_inode:[eventfd]>, buf: \1\0\0\0\0\0\0\0, count: 8) = 8
0.236 ( 0.001 ms): ptyxis/3622 ppoll(ufds: 0x5586a7e8f120, nfds: 10, tsp: 0x7ffdd1fbb470, sigsetsize: 8) = 1
0.238 ( 0.001 ms): ptyxis/3622 read(fd: 4<anon_inode:[eventfd]>, buf: 0x7ffdd1fbb3a0, count: 8) = 8
? ( ): mdns_service/5565 ... [continued]: recvfrom()) = -1 EAGAIN (Resource temporarily unavailable)
0.241 ( ): ptyxis/3622 ppoll(ufds: 0x5586a7e8f120, nfds: 10, tsp: 0x7ffdd1fbb470, sigsetsize: 8) ...
0.032 ( 0.627 ms): sleep/146484 ... [continued]: execve()) = 0
1.059 ( ): mdns_service/5565 recvfrom(fd: 292<socket:[81029]>, ubuf: 0x7f6cfdc454c0, size: 9000, addr: 0x7f6cfdc47b00, addr_len: 0x7f6cfdc47a00) ...
0.676 ( 0.001 ms): sleep/146484 brk() = 0x56443e2d4000
0.689 ( 0.002 ms): sleep/146484 mmap(len: 8192, prot: READ|WRITE, flags: PRIVATE|ANONYMOUS) = 0x7fe66d41d000
0.693 ( 0.002 ms): sleep/146484 access(filename: "/etc/ld.so.preload", mode: R) = -1 ENOENT (No such file or directory)
0.698 ( 0.002 ms): sleep/146484 openat(dfd: CWD, filename: "/etc/ld.so.cache", flags: RDONLY|CLOEXEC) = 3
0.701 ( 0.001 ms): sleep/146484 fstat(fd: 3, statbuf: 0x7ffefb498350) = 0
0.704 ( 0.003 ms): sleep/146484 mmap(len: 76091, prot: READ, flags: PRIVATE, fd: 3) = 0x7fe66d40a000
0.708 ( 0.001 ms): sleep/146484 close(fd: 3) = 0
0.712 ( 0.002 ms): sleep/146484 openat(dfd: CWD, filename: "/lib64/libc.so.6", flags: RDONLY|CLOEXEC) = 3
0.715 ( 0.001 ms): sleep/146484 read(fd: 3, buf: 0x7ffefb4984b8, count: 832) = 832
0.717 ( 0.001 ms): sleep/146484 pread64(fd: 3, buf: 0x7ffefb4980a0, count: 784, pos: 64) = 784
0.719 ( 0.001 ms): sleep/146484 fstat(fd: 3, statbuf: 0x7ffefb498340) = 0
0.722 ( 0.001 ms): sleep/146484 pread64(fd: 3, buf: 0x7ffefb497f80, count: 784, pos: 64) = 784
0.723 ( 0.003 ms): sleep/146484 mmap(len: 2038872, prot: READ|EXEC, flags: PRIVATE|DENYWRITE, fd: 3) = 0x7fe66d218000
0.727 ( 0.004 ms): sleep/146484 mmap(addr: 0x7fe66d387000, len: 479232, prot: READ, flags: PRIVATE|FIXED|DENYWRITE, fd: 3, off: 0x16f000) = 0x7fe66d387000
0.733 ( 0.003 ms): sleep/146484 mmap(addr: 0x7fe66d3fc000, len: 24576, prot: READ|WRITE, flags: PRIVATE|FIXED|DENYWRITE, fd: 3, off: 0x1e3000) = 0x7fe66d3fc000
0.737 ( 0.002 ms): sleep/146484 mmap(addr: 0x7fe66d402000, len: 31832, prot: READ|WRITE, flags: PRIVATE|FIXED|ANONYMOUS) = 0x7fe66d402000
0.743 ( 0.001 ms): sleep/146484 close(fd: 3) = 0
0.748 ( 0.002 ms): sleep/146484 mmap(len: 12288, prot: READ|WRITE, flags: PRIVATE|ANONYMOUS) = 0x7fe66d215000
0.753 ( 0.001 ms): sleep/146484 arch_prctl(option: SET_FS, arg2: 0x7fe66d215740) = 0
0.754 ( 0.001 ms): sleep/146484 set_tid_address(tidptr: 0x7fe66d215a10) = 146484 (sleep)
0.757 ( 0.001 ms): sleep/146484 set_robust_list(head: (struct robust_list_head){.list = (struct robust_list){.next = (struct robust_list *)0x7fe66d215a20,},.futex_offset = (long int)-32,}, len: 24) = 0
0.759 ( 0.001 ms): sleep/146484 rseq(rseq: (struct rseq){.cpu_id = (__u32)4294967295,}, rseq_len: 32, sig: 1392848979) = 0
0.780 ( 0.003 ms): sleep/146484 mprotect(start: 0x7fe66d3fc000, len: 16384, prot: READ) = 0
0.788 ( 0.002 ms): sleep/146484 mprotect(start: 0x564438854000, len: 4096, prot: READ) = 0
0.792 ( 0.002 ms): sleep/146484 mprotect(start: 0x7fe66d459000, len: 8192, prot: READ) = 0
0.798 ( 0.001 ms): sleep/146484 prlimit64(resource: STACK, old_rlim: 0x7ffefb498e90) = 0
0.807 ( 0.003 ms): sleep/146484 munmap(addr: 0x7fe66d40a000, len: 76091) = 0
0.817 ( 0.001 ms): sleep/146484 getrandom(ubuf: 0x7fe66d407218, len: 8, flags: NONBLOCK) = 8
0.819 ( 0.001 ms): sleep/146484 brk() = 0x56443e2d4000
0.821 ( 0.003 ms): sleep/146484 brk(brk: 0x56443e2f5000) = 0x56443e2f5000
0.827 ( 0.032 ms): sleep/146484 openat(dfd: CWD, filename: "", flags: RDONLY|CLOEXEC) = 3
0.860 ( 0.001 ms): sleep/146484 fstat(fd: 3, statbuf: 0x7fe66d401800) = 0
0.862 ( 0.002 ms): sleep/146484 mmap(len: 233242544, prot: READ, flags: PRIVATE, fd: 3) = 0x7fe65f200000
0.867 ( 0.001 ms): sleep/146484 close(fd: 3) = 0
0.888 ( 0.003 ms): sleep/146484 openat(dfd: CWD, filename: "/usr/share/locale/locale.alias", flags: RDONLY|CLOEXEC) = 3
0.892 ( 0.001 ms): sleep/146484 fstat(fd: 3, statbuf: 0x7ffefb498a70) = 0
0.895 ( 0.002 ms): sleep/146484 read(fd: 3, buf: 0x56443e2d5680, count: 4096) = 2998
0.901 ( 0.001 ms): sleep/146484 read(fd: 3, buf: 0x56443e2d5680, count: 4096) = 0
0.903 ( 0.001 ms): sleep/146484 close(fd: 3) = 0
0.909 ( 0.002 ms): sleep/146484 openat(dfd: CWD, filename: "/usr/share/locale/en_US.UTF-8/LC_MESSAGES/coreutils.mo") = -1 ENOENT (No such file or directory)
0.912 ( 0.001 ms): sleep/146484 openat(dfd: CWD, filename: "/usr/share/locale/en_US.utf8/LC_MESSAGES/coreutils.mo") = -1 ENOENT (No such file or directory)
0.914 ( 0.002 ms): sleep/146484 openat(dfd: CWD, filename: "/usr/share/locale/en_US/LC_MESSAGES/coreutils.mo") = -1 ENOENT (No such file or directory)
0.916 ( 0.001 ms): sleep/146484 openat(dfd: CWD, filename: "/usr/share/locale/en.UTF-8/LC_MESSAGES/coreutils.mo") = -1 ENOENT (No such file or directory)
0.918 ( 0.001 ms): sleep/146484 openat(dfd: CWD, filename: "/usr/share/locale/en.utf8/LC_MESSAGES/coreutils.mo") = -1 ENOENT (No such file or directory)
0.920 ( 0.002 ms): sleep/146484 openat(dfd: CWD, filename: "/usr/share/locale/en/LC_MESSAGES/coreutils.mo") = -1 ENOENT (No such file or directory)
0.930 ( 0.055 ms): sleep/146484 clock_nanosleep(rqtp: { .tv_sec: 0, .tv_nsec: 1 }, rmtp: 0x7ffefb4990f0) = 0
0.987 ( 0.001 ms): sleep/146484 close(fd: 1) = 0
0.989 ( 0.001 ms): sleep/146484 close(fd: 2) = 0
0.992 ( ): sleep/146484 exit_group() = ?Summary of events:
total, 3096 events
syscall calls errors total min avg max stddev
(msec) (msec) (msec) (msec) (%)
--------------- -------- ------ -------- --------- --------- --------- ------
ppoll 317 0 47.372 0.000 0.149 3.804 17.80% recvfrom 8 8 18.000 1.986 2.250 2.996 7.19% sched_setaffinity 66 0 0.743 0.001 0.011 0.021 6.16% execve 6 5 0.644 0.001 0.107 0.630 97.28% write 1268 0 0.548 0.000 0.000 0.005 2.30% read 390 75 0.158 0.000 0.000 0.012 9.19% ioctl 138 1 0.119 0.000 0.001 0.011 14.79% newfstatat 28 17 0.079 0.001 0.003 0.025 33.91% rt_sigaction 446 0 0.077 0.000 0.000 0.002 4.78% futex 20 1 0.077 0.000 0.004 0.037 51.43% openat 13 6 0.057 0.001 0.004 0.032 51.97% clock_nanosleep 1 0 0.055 0.055 0.055 0.055 0.00% rt_sigprocmask 290 0 0.047 0.000 0.000 0.002 5.61% mmap 8 0 0.021 0.002 0.003 0.004 12.60% poll 4 0 0.015 0.000 0.004 0.014 92.77% readlink 5 0 0.014 0.001 0.003 0.005 28.80% close 15 0 0.009 0.000 0.001 0.001 12.52% pread64 10 0 0.009 0.000 0.001 0.003 26.77% recvmsg 17 13 0.008 0.000 0.000 0.002 23.80% mprotect 3 0 0.006 0.002 0.002 0.003 15.31% sendmsg 5 0 0.006 0.001 0.001 0.002 21.98% fstat 6 0 0.005 0.000 0.001 0.001 23.30% brk 3 0 0.004 0.001 0.001 0.003 39.48% munmap 1 0 0.003 0.003 0.003 0.003 0.00% access 1 1 0.002 0.002 0.002 0.002 0.00% timerfd_settime 5 0 0.002 0.000 0.000 0.000 11.96% eventfd2 1 0 0.002 0.002 0.002 0.002 0.00% sched_getaffinity 2 0 0.001 0.001 0.001 0.001 0.96% getrandom 1 0 0.001 0.001 0.001 0.001 0.00% rt_sigreturn 1 0 0.001 0.001 0.001 0.001 0.00% prlimit64 1 0 0.001 0.001 0.001 0.001 0.00% set_tid_address 1 0 0.001 0.001 0.001 0.001 0.00% getpid 6 0 0.001 0.000 0.000 0.000 14.19% arch_prctl 1 0 0.001 0.001 0.001 0.001 0.00% set_robust_list 1 0 0.001 0.001 0.001 0.001 0.00% rseq 1 0 0.001 0.001 0.001 0.001 0.00% fcntl 3 0 0.001 0.000 0.000 0.000 20.48% epoll_wait 2 0 0.001 0.000 0.000 0.000 38.12% uname 1 0 0.000 0.000 0.000 0.000 0.00%
Quoted 3 lines
Anyway, I don't mind adding these details later on, so
Reviewed-by Howard Chu <[email protected]>
Thanks, applied to perf-tools-next,
- Arnaldo
Arnaldo Carvalho de Melo
maintainerQuoted 35 lines · Howard Chu
Hello Namhyung,
Quoted 32 lines · Namhyung Kim
When -s/--summary option is used, it doesn't need (augmented) arguments of syscalls. Let's skip the augmentation and load another small BPF program to collect the statistics in the kernel instead of copying the data to the ring-buffer to calculate the stats in userspace. This will be much more light-weight than the existing approach and remove any lost events.
Let's add a new option --bpf-summary to control this behavior. I cannot make it default because there's no way to get
e_machinein the BPF which is needed for detecting different ABIs like 32-bit compat mode.No functional changes intended except for no more LOST events. :)
$ sudo ./perf trace -as --summary-mode=total --bpf-summary sleep 1Summary of events: total, 6194 events syscall calls errors total min avg max stddev (msec) (msec) (msec) (msec) (%) --------------- -------- ------ -------- --------- --------- --------- ------ epoll_wait 561 0 4530.843 0.000 8.076 520.941 18.75% futex 693 45 4317.231 0.000 6.230 500.077 21.98% poll 300 0 1040.109 0.000 3.467 120.928 17.02% clock_nanosleep 1 0 1000.172 1000.172 1000.172 1000.172 0.00% ppoll 360 0 872.386 0.001 2.423 253.275 41.91% epoll_pwait 14 0 384.349 0.001 27.453 380.002 98.79% pselect6 14 0 108.130 7.198 7.724 8.206 0.85% nanosleep 39 0 43.378 0.069 1.112 10.084 44.23% ...
I added the following to align sched_[gs]etaffinity,
Thanks,
- Arnaldo
tools/perf/util/bpf-trace-summary.c +2 -2
@@ -139,9 +139,9 @@ static int print_common_stats(struct syscall_data *data, FILE *fp)/* TODO: support other ABIs */ name =syscalltbl__name (EM_HOST , node->syscall_nr);if (name) - printed +=fprintf (fp," %-15s" , name); + printed +=fprintf (fp," %-17s" , name);else - printed +=fprintf (fp," syscall:%-7d" , node->syscall_nr); + printed +=fprintf (fp," syscall:%-9d" , node->syscall_nr); printed += fprintf(fp, " %8u %6u %9.3f %9.3f %9.3f %9.3f %9.2f%%\n", stat->count, stat->error, total, min, avg, max,
Namhyung Kim
maintainerHi Arnaldo,
Quoted 37 lines · Arnaldo Carvalho de Melo
Quoted 35 lines · Howard Chu
Hello Namhyung,
Quoted 32 lines · Namhyung Kim
When -s/--summary option is used, it doesn't need (augmented) arguments of syscalls. Let's skip the augmentation and load another small BPF program to collect the statistics in the kernel instead of copying the data to the ring-buffer to calculate the stats in userspace. This will be much more light-weight than the existing approach and remove any lost events.
Let's add a new option --bpf-summary to control this behavior. I cannot make it default because there's no way to get
e_machinein the BPF which is needed for detecting different ABIs like 32-bit compat mode.No functional changes intended except for no more LOST events. :)
$ sudo ./perf trace -as --summary-mode=total --bpf-summary sleep 1Summary of events: total, 6194 events syscall calls errors total min avg max stddev (msec) (msec) (msec) (msec) (%) --------------- -------- ------ -------- --------- --------- --------- ------ epoll_wait 561 0 4530.843 0.000 8.076 520.941 18.75% futex 693 45 4317.231 0.000 6.230 500.077 21.98% poll 300 0 1040.109 0.000 3.467 120.928 17.02% clock_nanosleep 1 0 1000.172 1000.172 1000.172 1000.172 0.00% ppoll 360 0 872.386 0.001 2.423 253.275 41.91% epoll_pwait 14 0 384.349 0.001 27.453 380.002 98.79% pselect6 14 0 108.130 7.198 7.724 8.206 0.85% nanosleep 39 0 43.378 0.069 1.112 10.084 44.23% ...
Quoted 1 line
I added the following to align sched_[gs]etaffinity,
Thanks for processing the patch and updating this. But I'm afraid there are more syscalls with longer names and this is not the only place to print the syscall names. Also I think we need to update length of the time fields. So I prefer handling them in a separate patch later.
Thanks,
Namhyung
Quoted 17 lines
tools/perf/util/bpf-trace-summary.c +2
-2@@ -139,9 +139,9 @@ static int print_common_stats(struct syscall_data *data, FILE *fp)/* TODO: support other ABIs */ name =syscalltbl__name (EM_HOST , node->syscall_nr);if (name) - printed +=fprintf (fp," %-15s" , name); + printed +=fprintf (fp," %-17s" , name);else - printed +=fprintf (fp," syscall:%-7d" , node->syscall_nr); + printed +=fprintf (fp," syscall:%-9d" , node->syscall_nr); printed += fprintf(fp, " %8u %6u %9.3f %9.3f %9.3f %9.3f %9.2f%%\n", stat->count, stat->error, total, min, avg, max,
Arnaldo Carvalho de Melo
maintainerQuoted 18 lines · Namhyung Kim
Hi Arnaldo,
Quoted 15 lines · Arnaldo Carvalho de Melo
Quoted 13 lines · Howard Chu
Quoted 12 lines · Namhyung Kim
syscall calls errors total min avg max stddev (msec) (msec) (msec) (msec) (%) --------------- -------- ------ -------- --------- --------- --------- ------ epoll_wait 561 0 4530.843 0.000 8.076 520.941 18.75% futex 693 45 4317.231 0.000 6.230 500.077 21.98% poll 300 0 1040.109 0.000 3.467 120.928 17.02% clock_nanosleep 1 0 1000.172 1000.172 1000.172 1000.172 0.00% ppoll 360 0 872.386 0.001 2.423 253.275 41.91% epoll_pwait 14 0 384.349 0.001 27.453 380.002 98.79% pselect6 14 0 108.130 7.198 7.724 8.206 0.85% nanosleep 39 0 43.378 0.069 1.112 10.084 44.23% ...
Quoted 6 lines
Quoted 1 line
I added the following to align sched_[gs]etaffinity,
Thanks for processing the patch and updating this. But I'm afraid there are more syscalls with longer names and this is not the only place to print the syscall names. Also I think we need to update length of the time fields. So I prefer handling them in a separate patch later.
Fair enough, I'm leaving the patch as-is.
- Arnaldo
Arnaldo Carvalho de Melo
maintainerQuoted 17 lines · Arnaldo Carvalho de Melo
Quoted 15 lines · Namhyung Kim
Quoted 14 lines · Arnaldo Carvalho de Melo
Quoted 13 lines · Howard Chu
Quoted 12 lines · Namhyung Kim
syscall calls errors total min avg max stddev --------------- -------- ------ -------- --------- --------- --------- ------ epoll_wait 561 0 4530.843 0.000 8.076 520.941 18.75% futex 693 45 4317.231 0.000 6.230 500.077 21.98% poll 300 0 1040.109 0.000 3.467 120.928 17.02% clock_nanosleep 1 0 1000.172 1000.172 1000.172 1000.172 0.00% ppoll 360 0 872.386 0.001 2.423 253.275 41.91% epoll_pwait 14 0 384.349 0.001 27.453 380.002 98.79% pselect6 14 0 108.130 7.198 7.724 8.206 0.85% nanosleep 39 0 43.378 0.069 1.112 10.084 44.23% ...
Quoted 8 lines
Quoted 6 lines
Quoted 1 line
I added the following to align sched_[gs]etaffinity,
Thanks for processing the patch and updating this. But I'm afraid there are more syscalls with longer names and this is not the only place to print the syscall names. Also I think we need to update length of the time fields. So I prefer handling them in a separate patch later.
Fair enough, I'm leaving the patch as-is.
But, still have to look at this:
toolsbuilder@five:~$ time dm
1 114.52 almalinux:8 : Ok gcc (GCC) 8.5.0 20210514 (Red Hat 8.5.0-26) , clang version 18.1.8 (Red Hat 18.1.8-1.module_el8.10.0+3903+ca21d481) flex 2.6.1
2 111.09 almalinux:9 : Ok gcc (GCC) 11.5.0 20240719 (Red Hat 11.5.0-5) , clang version 18.1.8 (AlmaLinux OS Foundation 18.1.8-3.el9) flex 2.6.4
3: almalinux:9-i386WARNING: image platform (linux/386) does not match the expected platform (linux/amd64)
WARNING: image platform (linux/386) does not match the expected platform (linux/amd64)
132.71 almalinux:9-i386 : Ok gcc (GCC) 11.4.1 20231218 (Red Hat 11.4.1-3) , clang version 17.0.6 (AlmaLinux OS Foundation 17.0.6-5.el9) flex 2.6.4
4 21.54 alpine:3.16 : FAIL gcc version 11.2.1 20220219 (Alpine 11.2.1_git20220219)
bpf-trace-summary.c:(.text+0xf0760): undefined reference to `syscalltbl__name'
collect2: error: ld returned 1 exit status
5 16.50 alpine:3.17 : FAIL gcc version 12.2.1 20220924 (Alpine 12.2.1_git20220924-r4)
bpf-trace-summary.c:(.text+0xf2020): undefined reference to `syscalltbl__name'
collect2: error: ld returned 1 exit statusMore info:
perf-6.15.0-rc2/HEAD
perf-6.15.0-rc2/PERF-VERSION-FILE
BUILD_TARBALL_HEAD=24c0c35d4640052c61ed539a777bd3bd60d62bbf
Using built-in specs.
COLLECT_GCC=gcc
COLLECT_LTO_WRAPPER=/usr/libexec/gcc/x86_64-alpine-linux-musl/12.2.1/lto-wrapper
Target: x86_64-alpine-linux-musl
Configured with: /home/buildozer/aports/main/gcc/src/gcc-12-20220924/configure --prefix=/usr --mandir=/usr/share/man --infodir=/usr/share/info --build=x86_64-alpine-linux-musl --host=x86_64-alpine-linux-musl --target=x86_64-alpine-linux-musl --enable-checking=release --disable-fixed-point --disable-libstdcxx-pch --disable-multilib --disable-nls --disable-werror --disable-symvers --enable-__cxa_atexit --enable-default-pie --enable-default-ssp --enable-languages=c,c++,d,objc,go,fortran,ada --disable-libssp --disable-libsanitizer --enable-shared --enable-threads --enable-tls --with-bugurl=https://gitlab.alpinelinux.org/alpine/aports/-/issues --with-system-zlib --with-linker-hash-style=gnu --with-pkgversion='Alpine 12.2.1_git20220924-r4' Thread model: posix
Supported LTO compression algorithms: zlib
gcc version 12.2.1 20220924 (Alpine 12.2.1_git20220924-r4)
+ make 'NO_LIBTRACEEVENT=1' 'ARCH=' 'CROSS_COMPILE=' 'EXTRA_CFLAGS=' -C tools/perf 'O=/tmp/build/perf'
make: Entering directory '/git/perf-6.15.0-rc2/tools/perf'
BUILD: Doing 'make -j28' parallel build
Warning: Skipped check-headers due to missing ../../include
Makefile.config:563: No elfutils/debuginfod.h found, no debuginfo server support, please install libdebuginfod-dev/elfutils-debuginfod-client-devel or equivalent
Makefile.config:605: No sys/sdt.h found, no SDT events are defined, please install systemtap-sdt-devel or systemtap-sdt-dev
Makefile.config:1085: No libbabeltrace found, disables 'perf data' CTF format support, please install libbabeltrace-dev[el]/libbabeltrace-ctf-dev
Makefile.config:1128: No alternatives command found, you need to set JDIR= to point to the root of your Java directory
Makefile.config:1159: libpfm4 not found, disables libpfm4 support. Please install libpfm4-dev
Auto-detecting system features:
... libdw: [ on ]
... glibc: [ OFF ]
... libelf: [ on ] ... libnuma: [ on ] ... numa_num_possible_cpus: [ on ] ... libperl: [ on ] ... libpython: [ on ] ... libcrypto: [ on ] ... libcapstone: [ on ] ... llvm-perf: [ on ] ... zlib: [ on ] ... lzma: [ on ] ... get_cpuid: [ on ] ... bpf: [ on ] ... libaio: [ on ] ... libzstd: [ on ]
PERF_VERSION = 6.15.rc2.g24c0c35d4640
GEN /tmp/build/perf/common-cmds.h GEN /tmp/build/perf/arch/arm64/include/generated/asm/sysreg-defs.h GEN perf-archive GEN perf-iostat <SNIP> CC /tmp/build/perf/util/bpf-filter-flex.o LD /tmp/build/perf/util/perf-util-in.o LD /tmp/build/perf/perf-util-in.o AR /tmp/build/perf/libperf-util.a CC /tmp/build/perf/pmu-events/pmu-events.o LD /tmp/build/perf/pmu-events/pmu-events-in.o AR /tmp/build/perf/libpmu-events.a LINK /tmp/build/perf/perf GEN /tmp/build/perf/python/perf.cpython-310-x86_64-linux-gnu.so
/usr/lib/gcc/x86_64-alpine-linux-musl/12.2.1/../../../../x86_64-alpine-linux-musl/bin/ld: /tmp/build/perf/libperf-util.a(perf-util-in.o): in function `print_common_stats':
bpf-trace-summary.c:(.text+0xf2020): undefined reference to `syscalltbl__name'
collect2: error: ld returned 1 exit status
make[2]: *** [Makefile.perf:804: /tmp/build/perf/perf] Error 1
make[2]: *** Waiting for unfinished jobs....
make[1]: *** [Makefile.perf:290: sub-make] Error 2
make: *** [Makefile:76: all] Error 2
make: Leaving directory '/git/perf-6.15.0-rc2/tools/perf'
+ exit 1
toolsbuilder@five:~$
I'll take a look tomorrow.
- Arnaldo
Namhyung Kim
maintainerHi Arnaldo,
Quoted 109 lines · Arnaldo Carvalho de Melo
Quoted 17 lines · Arnaldo Carvalho de Melo
Quoted 15 lines · Namhyung Kim
Quoted 14 lines · Arnaldo Carvalho de Melo
Quoted 13 lines · Howard Chu
Quoted 12 lines · Namhyung Kim
syscall calls errors total min avg max stddev --------------- -------- ------ -------- --------- --------- --------- ------ epoll_wait 561 0 4530.843 0.000 8.076 520.941 18.75% futex 693 45 4317.231 0.000 6.230 500.077 21.98% poll 300 0 1040.109 0.000 3.467 120.928 17.02% clock_nanosleep 1 0 1000.172 1000.172 1000.172 1000.172 0.00% ppoll 360 0 872.386 0.001 2.423 253.275 41.91% epoll_pwait 14 0 384.349 0.001 27.453 380.002 98.79% pselect6 14 0 108.130 7.198 7.724 8.206 0.85% nanosleep 39 0 43.378 0.069 1.112 10.084 44.23% ...Quoted 8 lines
Quoted 6 lines
Quoted 1 line
I added the following to align sched_[gs]etaffinity,
Thanks for processing the patch and updating this. But I'm afraid there are more syscalls with longer names and this is not the only place to print the syscall names. Also I think we need to update length of the time fields. So I prefer handling them in a separate patch later.
Fair enough, I'm leaving the patch as-is.
But, still have to look at this:
toolsbuilder@five:~$ time dm 1 114.52 almalinux:8 : Ok gcc (GCC) 8.5.0 20210514 (Red Hat 8.5.0-26) , clang version 18.1.8 (Red Hat 18.1.8-1.module_el8.10.0+3903+ca21d481) flex 2.6.1 2 111.09 almalinux:9 : Ok gcc (GCC) 11.5.0 20240719 (Red Hat 11.5.0-5) , clang version 18.1.8 (AlmaLinux OS Foundation 18.1.8-3.el9) flex 2.6.4 3: almalinux:9-i386WARNING: image platform (linux/386) does not match the expected platform (linux/amd64) WARNING: image platform (linux/386) does not match the expected platform (linux/amd64) 132.71 almalinux:9-i386 : Ok gcc (GCC) 11.4.1 20231218 (Red Hat 11.4.1-3) , clang version 17.0.6 (AlmaLinux OS Foundation 17.0.6-5.el9) flex 2.6.4 4 21.54 alpine:3.16 : FAIL gcc version 11.2.1 20220219 (Alpine 11.2.1_git20220219) bpf-trace-summary.c:(.text+0xf0760): undefined reference to `syscalltbl__name' collect2: error: ld returned 1 exit status 5 16.50 alpine:3.17 : FAIL gcc version 12.2.1 20220924 (Alpine 12.2.1_git20220924-r4) bpf-trace-summary.c:(.text+0xf2020): undefined reference to `syscalltbl__name' collect2: error: ld returned 1 exit statusMore info:
perf-6.15.0-rc2/HEAD
perf-6.15.0-rc2/PERF-VERSION-FILE
BUILD_TARBALL_HEAD=24c0c35d4640052c61ed539a777bd3bd60d62bbf
Using built-in specs.
COLLECT_GCC=gcc
COLLECT_LTO_WRAPPER=/usr/libexec/gcc/x86_64-alpine-linux-musl/12.2.1/lto-wrapper
Target: x86_64-alpine-linux-musl
Configured with:/home/buildozer/aports/main/gcc/src/gcc-12-20220924/configure--prefix=/usr --mandir=/usr/share/man --infodir=/usr/share/info --build=x86_64-alpine-linux-musl --host=x86_64-alpine-linux-musl --target=x86_64-alpine-linux-musl --enable-checking=release --disable-fixed-point --disable-libstdcxx-pch --disable-multilib --disable-nls --disable-werror --disable-symvers --enable-__cxa_atexit --enable-default-pie --enable-default-ssp --enable-languages=c,c++,d,objc,go,fortran,ada --disable-libssp --disable-libsanitizer --enable-shared --enable-threads --enable-tls --with-bugurl=https://gitlab.alpinelinux.org/alpine/aports/-/issues --with-system-zlib --with-linker-hash-style=gnu --with-pkgversion='Alpine 12.2.1_git20220924-r4' Thread model: posix
Supported LTO compression algorithms: zlib
gcc version 12.2.1 20220924 (Alpine 12.2.1_git20220924-r4)
+ make 'NO_LIBTRACEEVENT=1' 'ARCH=' 'CROSS_COMPILE=' 'EXTRA_CFLAGS=' -Ctools/perf'O=/tmp/build/perf'
make: Entering directory '/git/perf-6.15.0-rc2/tools/perf'
BUILD: Doing 'make -j28' parallel build
Warning: Skipped check-headers due to missing ../../include
Makefile.config:563: Noelfutils/debuginfod.hfound, no debuginfo server support, please install libdebuginfod-dev/elfutils-debuginfod-client-devel or equivalent
Makefile.config:605: Nosys/sdt.hfound, no SDT events are defined, please install systemtap-sdt-devel or systemtap-sdt-dev
Makefile.config:1085: No libbabeltrace found, disables 'perf data' CTF format support, please install libbabeltrace-dev[el]/libbabeltrace-ctf-dev
Makefile.config:1128: No alternatives command found, you need to set JDIR= to point to the root of your Java directory
Makefile.config:1159: libpfm4 not found, disables libpfm4 support. Please install libpfm4-devAuto-detecting system features:
... libdw: [ on ]
... glibc: [ OFF ]... libelf: [ on ] ... libnuma: [ on ] ... numa_num_possible_cpus: [ on ] ... libperl: [ on ] ... libpython: [ on ] ... libcrypto: [ on ] ... libcapstone: [ on ] ... llvm-perf: [ on ] ... zlib: [ on ] ... lzma: [ on ] ... get_cpuid: [ on ] ... bpf: [ on ] ... libaio: [ on ] ... libzstd: [ on ]
PERF_VERSION= 6.15.rc2.g24c0c35d4640GEN /tmp/build/perf/common-cmds.h GEN /tmp/build/perf/arch/arm64/include/generated/asm/sysreg-defs.h GEN perf-archive GEN perf-iostat <SNIP> CC /tmp/build/perf/util/bpf-filter-flex.o LD /tmp/build/perf/util/perf-util-in.o LD /tmp/build/perf/perf-util-in.o AR /tmp/build/perf/libperf-util.a CC /tmp/build/perf/pmu-events/pmu-events.o LD /tmp/build/perf/pmu-events/pmu-events-in.o AR /tmp/build/perf/libpmu-events.a LINK /tmp/build/perf/perf GEN /tmp/build/perf/python/perf.cpython-310-x86_64-linux-gnu.so
/usr/lib/gcc/x86_64-alpine-linux-musl/12.2.1/../../../../x86_64-alpine-linux-musl/bin/ld: /tmp/build/perf/libperf-util.a(perf-util-in.o): in function `print_common_stats':
bpf-trace-summary.c:(.text+0xf2020): undefined reference to `syscalltbl__name'
collect2: error: ld returned 1 exit status
make[2]: *** [Makefile.perf:804:/tmp/build/perf/perf] Error 1
make[2]: *** Waiting for unfinished jobs....
make[1]: *** [Makefile.perf:290: sub-make] Error 2
make: *** [Makefile:76: all] Error 2
make: Leaving directory '/git/perf-6.15.0-rc2/tools/perf'
+ exit 1
toolsbuilder@five:~$
Quoted 1 line
I'll take a look tomorrow.
Thanks for the report. I think it's because syscalltbl.c depends on CONFIG_TRACE but bpf-trace-summary depends on CONFIG_PERF_BPF_SKEL.
In the future, I'd like to get rid of dependency to libtraceevent in perf trace and make it possible to use BPF/BTF only.
How about this?
Thanks,
Namhyung
---8<---
tools/perf/util/Build +4 -1
@@ -173,9 +173,12 @@ perf-util-$(CONFIG_PERF_BPF_SKEL) += bpf_off_cpu.o perf-util-$(CONFIG_PERF_BPF_SKEL) += bpf-filter.o perf-util-$(CONFIG_PERF_BPF_SKEL) += bpf-filter-flex.o perf-util-$(CONFIG_PERF_BPF_SKEL) += bpf-filter-bison.o -perf-util-$(CONFIG_PERF_BPF_SKEL) += bpf-trace-summary.o perf-util-$(CONFIG_PERF_BPF_SKEL) += btf.o +ifeq ($(CONFIG_TRACE),y) + perf-util-$(CONFIG_PERF_BPF_SKEL) += bpf-trace-summary.o +endif + ifeq ($(CONFIG_LIBTRACEEVENT),y) perf-util-$(CONFIG_PERF_BPF_SKEL) += bpf_lock_contention.o endif