- 16 Feb, 2021 4 commits
-
-
Arnaldo Carvalho de Melo authored
To get some fixes that didn't made into 5.11. Signed-off-by:Arnaldo Carvalho de Melo <acme@redhat.com>
-
Leo Yan authored
The sample structure contains the field 'data_src' which is used to tell the data operation attributions, e.g. operation type is loading or storing, cache level, it's snooping or remote accessing, etc. At the end, the 'data_src' will be parsed by perf mem/c2c tools to display human readable strings. This patch is to fill the 'data_src' field in the synthesized samples base on different types. Currently perf tool can display statistics for L1/L2/L3 caches but it doesn't support the 'last level cache'. To fit to current implementation, 'data_src' field uses L3 cache for last level cache. Before this commit, perf mem report looks like this: # Samples: 75K of event 'l1d-miss' # Total weight : 75951 # Sort order : local_weight,mem,sym,dso,symbol_daddr,dso_daddr,snoop,tlb,locked # # Overhead Samples Local Weight Memory access Symbol Shared Object Data Symbol Data Object Snoop TLB access # ........ ....... ............ ............. ...................... ............. ...................... ........... ..... .......... # 81.56% 61945 0 N/A [.] 0x00000000000009d8 serial_c [.] 0000000000000000 [unknown] N/A N/A 18.44% 14003 0 N/A [.] 0x0000000000000828 serial_c [.] 0000000000000000 [unknown] N/A N/A Now on a system with Arm SPE, addresses and access types are displayed: # Samples: 75K of event 'l1d-miss' # Total weight : 75951 # Sort order : local_weight,mem,sym,dso,symbol_daddr,dso_daddr,snoop,tlb,locked # # Overhead Samples Local Weight Memory access Symbol Shared Object Data Symbol Data Object Snoop TLB access # ........ ....... ............ ............. ...................... ............. ...................... ........... ..... .......... # 0.43% 324 0 L1 miss [.] 0x00000000000009d8 serial_c [.] 0x0000ffff80794e00 anon N/A Walker hit 0.42% 322 0 L1 miss [.] 0x00000000000009d8 serial_c [.] 0x0000ffff80794580 anon N/A Walker hit Signed-off-by:Leo Yan <leo.yan@linaro.org> Reviewed-by:
James Clark <james.clark@arm.com> Tested-by:
-
Leo Yan authored
The memory event can deliver two benefits: - The first benefit is the memory event can give out global view for memory accessing, rather than organizing events with scatter mode (e.g. uses separate event for L1 cache, last level cache, etc) which which can only display a event for single memory type, memory events include all memory accessing so it can display the data accessing cross memory levels in the same view; - The second benefit is the sample generation might introduce a big overhead and need to wait for long time for Perf reporting, we can specify itrace option '--itrace=M' to filter out other events and only output memory events, this can significantly reduce the overhead caused by generating samples. This patch is to enable memory event for Arm SPE. Signed-off-by:
-
Leo Yan authored
To properly handle memory and branch samples, this patch divides into two functions for generating samples: arm_spe__synth_mem_sample() is for synthesizing memory and TLB samples; arm_spe__synth_branch_sample() is to synthesize branch samples. Arm SPE backend decoder has passed virtual and physical address through packets, the address info is stored into the synthesize samples in the function arm_spe__synth_mem_sample(). Committer notes: Fixed this: 36 46.77 fedora:27 : FAIL clang version 5.0.2 (tags/RELEASE_502/final) util/arm-spe.c:269:34: error: missing field 'pid' initializer [-Werror,-Wmissing-field-initializers] struct perf_sample sample = { 0 }; ^ util/arm-spe.c:288:34: error: missing field 'pid' initializer [-Werror,-Wmissing-field-initializers] struct perf_sample sample = { 0 }; By using = { .ip = 0, }; Signed-off-by:
-
- 12 Feb, 2021 8 commits
-
-
Jianlin Lv authored
Perf failed to add a kretprobe event with debuginfo of vmlinux which is compiled by gcc with -fpatchable-function-entry option enabled. The same issue with kernel module. Issue: # perf probe -v 'kernel_clone%return $retval' ...... Writing event: r:probe/kernel_clone__return _text+599624 $retval Failed to write event: Invalid argument Error: Failed to add events. Reason: Invalid argument (Code: -22) # cat /sys/kernel/debug/tracing/error_log [156.75] trace_kprobe: error: Retprobe address must be an function entry Command: r:probe/kernel_clone__return _text+599624 $retval ^ # llvm-dwarfdump vmlinux |grep -A 10 -w 0x00df2c2b 0x00df2c2b: DW_TAG_subprogram DW_AT_external (true) DW_AT_name ("kernel_clone") DW_AT_decl_file ("/home/code/linux-next/kernel/fork.c") DW_AT_decl_line (2423) DW_AT_decl_column (0x07) DW_AT_prototyped (true) DW_AT_type (0x00dcd492 "pid_t") DW_AT_low_pc (0xffff800010092648) DW_AT_high_pc (0xffff800010092b9c) DW_AT_frame_base (DW_OP_call_frame_cfa) # cat /proc/kallsyms |grep kernel_clone ffff800010092640 T kernel_clone # readelf -s vmlinux |grep -i kernel_clone 183173: ffff800010092640 1372 FUNC GLOBAL DEFAULT 2 kernel_clone # objdump -d vmlinux |grep -A 10 -w \<kernel_clone\>: ffff800010092640 <kernel_clone>: ffff800010092640: d503201f nop ffff800010092644: d503201f nop ffff800010092648: d503233f paciasp ffff80001009264c: a9b87bfd stp x29, x30, [sp, #-128]! ffff800010092650: 910003fd mov x29, sp ffff800010092654: a90153f3 stp x19, x20, [sp, #16] The entry address of kernel_clone converted by debuginfo is _text+599624 (0x92648), which is consistent with the value of DW_AT_low_pc attribute. But the symbolic address of kernel_clone from /proc/kallsyms is ffff800010092640. This issue is found on arm64, -fpatchable-function-entry=2 is enabled when CONFIG_DYNAMIC_FTRACE_WITH_REGS=y; Just as objdump displayed the assembler contents of kernel_clone, GCC generate 2 NOPs at the beginning of each function. kprobe_on_func_entry detects that (_text+599624) is not the entry address of the function, which leads to the failure of adding kretprobe event. kprobe_on_func_entry ->_kprobe_addr ->kallsyms_lookup_size_offset ->arch_kprobe_on_func_entry // FALSE The cause of the issue is that the first instruction in the compile unit indicated by DW_AT_low_pc does not include NOPs. This issue exists in all gcc versions that support -fpatchable-function-entry option. I have reported it to the GCC community: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98776 Currently arm64 and PA-RISC may enable fpatchable-function-entry option. The kernel compiled with clang does not have this issue. FIX: This GCC issue only cause the registration failure of the kretprobe event which doesn't need debuginfo. So, stop using debuginfo for retprobe. map will be used to query the probe function address. Signed-off-by:Jianlin Lv <Jianlin.Lv@arm.com> Acked-by:
Masami Hiramatsu <mhiramat@kernel.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: clang-built-linux@googlegroups.com Cc: Frank Ch. Eigler <fche@redhat.com> Cc: Ian Rogers <irogers@google.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Nathan Chancellor <nathan@kernel.org> Cc: Nick Desaulniers <ndesaulniers@google.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Sumanth Korikkar <sumanthk@linux.ibm.com> Link: http://lore.kernel.org/lkml/20210210062646.2377995-1-Jianlin.Lv@arm.comSigned-off-by:
-
Nicholas Fraser authored
The first time dso__load() was called on a PE file it always returned -1 error. This caused the first call to map__find_symbol() to always fail on a PE file so the first sample from each PE file always had symbol <unknown>. Subsequent samples succeed however because the DSO is already loaded. This fixes dso__load() to return 0 when successfully loading a DSO with libbfd. Fixes: eac9a434 ("perf symbols: Try reading the symbol table with libbfd") Signed-off-by:
Nicholas Fraser <nfraser@codeweavers.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Frank Ch. Eigler <fche@redhat.com> Cc: Huw Davies <huw@codeweavers.com> Cc: Ian Rogers <irogers@google.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Kim Phillips <kim.phillips@amd.com> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Remi Bernon <rbernon@codeweavers.com> Cc: Song Liu <songliubraving@fb.com> Cc: Tommi Rantala <tommi.t.rantala@nokia.com> Cc: Ulrich Czekalla <uczekalla@codeweavers.com> Link: http://lore.kernel.org/lkml/1671b43b-09c3-1911-dbf8-7f030242fbf7@codeweavers.comSigned-off-by:
-
Nicholas Fraser authored
dso__load_bfd_symbols() attempts to load a DSO at its original path, then closes it and loads the file in the debug cache. This is incorrect. It should ignore the original file and work with only the debug cache. The original file may have changed or may not even exist, for example if the debug cache has been transferred to another machine via "perf archive". This fix makes it only load the file in the debug cache. Further notes from Nicholas: dso__load_bfd_symbols() is called in a loop from dso__load() for a variety of paths. These are generated by the various DSO_BINARY_TYPEs in the binary_type_symtab list at the top of util/symbol.c. In each case the debugfile passed to dso__load_bfd_symbols() is the path to try. One of those iterations (the first one I believe) passes the original path as the debugfile. If the file still exists at the original path, this is the one that ends up being used in case the debugcache was deleted or the PE file doesn't have a build-id. A later iteration (BUILD_ID_CACHE) passes debugfile as the file in the debugcache if it has a build-id. Even if the file was previously loaded at its original path, (if I understand correctly) this load will override it so the debugcache file ends up being used. Committer notes: So if it fails to find in the cache, it will eventually hope for the best and look at the path in the local filesystem, which in many cases is enough. At some point we need to switch from this "hope for the best" approach to one that warns the user that there is no guarantee, if no buildid is present, that just by looking at the pathname the symbolisation will work. Signed-off-by:
-
Leo Yan authored
This patch is to store operation type in packet structure. Signed-off-by:
-
Leo Yan authored
This patch is to store virtual and physical memory addresses in packet, which will be used for memory samples. Signed-off-by:
-
Leo Yan authored
This patch is to enable sample type PERF_SAMPLE_DATA_SRC for Arm SPE in the perf data, when output the tracing data, it tells tools that it contains data source in the memory event. Signed-off-by:
-
Ian Rogers authored
Minor cleanup. Signed-off-by:
Ian Rogers <irogers@google.com> Acked-by:
Jiri Olsa <jolsa@redhat.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Stephane Eranian <eranian@google.com> Link: http://lore.kernel.org/lkml/20210211183914.4093187-1-irogers@google.comSigned-off-by:
-
Ian Rogers authored
Migrated to libperf in: 4b247fa7 ("libperf: Adopt xyarray class from perf") Signed-off-by:
-
- 11 Feb, 2021 24 commits
-
-
Dmitry Safonov authored
GCC (GCC) 8.4.0 20200304 fails to build perf with: : util/symbol.c: In function 'dso__load_bfd_symbols': : util/symbol.c:1626:16: error: comparison of integer expressions of different signednes : for (i = 0; i < symbols_count; ++i) { : ^ : util/symbol.c:1632:16: error: comparison of integer expressions of different signednes : while (i + 1 < symbols_count && : ^ : util/symbol.c:1637:13: error: comparison of integer expressions of different signednes : if (i + 1 < symbols_count && : ^ : cc1: all warnings being treated as errors It's unlikely that the symtable will be that big, but the fix is an oneliner and as perf has CORE_CFLAGS += -Wextra, which makes build to fail together with CORE_CFLAGS += -Werror Fixes: eac9a434 ("perf symbols: Try reading the symbol table with libbfd") Signed-off-by:Dmitry Safonov <dima@arista.com> Acked-by:
Namhyung Kim <namhyung@kernel.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Dmitry Safonov <0x7f454c46@gmail.com> Cc: Jacek Caban <jacek@codeweavers.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Remi Bernon <rbernon@codeweavers.com> Link: http://lore.kernel.org/lkml/20210209145148.178702-1-dima@arista.comSigned-off-by:
-
Martin Liška authored
Considering the following testcase: int foo(int a, int b) { for (unsigned i = 0; i < 1000000000; i++) a += b; return a; } int main() { foo (3, 4); return 0; } 'perf annotate' displays: 86.52 │40055e: → ja 40056c <foo(int, int)+0x26> 13.37 │400560: mov -0x18(%rbp),%eax │400563: add %eax,-0x14(%rbp) │400566: addl $0x1,-0x4(%rbp) 0.11 │40056a: → jmp 400557 <foo(int, int)+0x11> │40056c: mov -0x14(%rbp),%eax │40056f: pop %rbp and the 'ja 40056c' does not link to the location in the function. It's caused by fact that comma is wrongly parsed, it's part of function signature. With my patch I see: 86.52 │ ┌──ja 26 13.37 │ │ mov -0x18(%rbp),%eax │ │ add %eax,-0x14(%rbp) │ │ addl $0x1,-0x4(%rbp) 0.11 │ │↑ jmp 11 │26:└─→mov -0x14(%rbp),%eax and 'o' output prints: 86.52 │4005┌── ↓ ja 40056c <foo(int, int)+0x26> 13.37 │4005│0: mov -0x18(%rbp),%eax │4005│3: add %eax,-0x14(%rbp) │4005│6: addl $0x1,-0x4(%rbp) 0.11 │4005│a: ↑ jmp 400557 <foo(int, int)+0x11> │4005└─→ mov -0x14(%rbp),%eax On the contrary, compiling the very same file with gcc -x c, the parsing is fine because function arguments are not displayed: jmp 400543 <foo+0x1d> Committer testing: Before: $ cat cpp_args_annotate.c int foo(int a, int b) { for (unsigned i = 0; i < 1000000000; i++) a += b; return a; } int main() { foo (3, 4); return 0; } $ gcc --version |& head -1 gcc (GCC) 10.2.1 20201125 (Red Hat 10.2.1-9) $ gcc -g cpp_args_annotate.c -o cpp_args_annotate $ perf record ./cpp_args_annotate [ perf record: Woken up 2 times to write data ] [ perf record: Captured and wrote 0.275 MB perf.data (7188 samples) ] $ perf annotate --stdio2 foo Samples: 7K of event 'cycles:u', 4000 Hz, Event count (approx.): 7468429289, [percent: local period] foo() /home/acme/c/cpp_args_annotate Percent 0000000000401106 <foo>: foo(): int foo(int a, int b) { push %rbp mov %rsp,%rbp mov %edi,-0x14(%rbp) mov %esi,-0x18(%rbp) for (unsigned i = 0; i < 1000000000; i++) movl $0x0,-0x4(%rbp) ↓ jmp 1d a += b; 13.45 13: mov -0x18(%rbp),%eax add %eax,-0x14(%rbp) for (unsigned i = 0; i < 1000000000; i++) addl $0x1,-0x4(%rbp) 0.09 1d: cmpl $0x3b9ac9ff,-0x4(%rbp) 86.46 ↑ jbe 13 return a; mov -0x14(%rbp),%eax } pop %rbp ← retq $ I.e. works for C, now lets switch to C++: $ g++ -g cpp_args_annotate.c -o cpp_args_annotate $ perf record ./cpp_args_annotate [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.268 MB perf.data (6976 samples) ] $ perf annotate --stdio2 foo Samples: 6K of event 'cycles:u', 4000 Hz, Event count (approx.): 7380681761, [percent: local period] foo() /home/acme/c/cpp_args_annotate Percent 0000000000401106 <foo(int, int)>: foo(int, int): int foo(int a, int b) { push %rbp mov %rsp,%rbp mov %edi,-0x14(%rbp) mov %esi,-0x18(%rbp) for (unsigned i = 0; i < 1000000000; i++) movl $0x0,-0x4(%rbp) cmpl $0x3b9ac9ff,-0x4(%rbp) 86.53 → ja 40112c <foo(int, int)+0x26> a += b; 13.32 mov -0x18(%rbp),%eax 0.00 add %eax,-0x14(%rbp) for (unsigned i = 0; i < 1000000000; i++) addl $0x1,-0x4(%rbp) 0.15 → jmp 401117 <foo(int, int)+0x11> return a; mov -0x14(%rbp),%eax } pop %rbp ← retq $ Reproduced. Now with this patch: Reusing the C++ built binary, as we can see here: $ readelf -wi cpp_args_annotate | grep producer <c> DW_AT_producer : (indirect string, offset: 0x2e): GNU C++14 10.2.1 20201125 (Red Hat 10.2.1-9) -mtune=generic -march=x86-64 -g $ And furthermore: $ file cpp_args_annotate cpp_args_annotate: ELF 64-bit LSB executable, x86-64, version 1 (SYSV), dynamically linked, interpreter /lib64/ld-linux-x86-64.so.2, BuildID[sha1]=4fe3cab260204765605ec630d0dc7a7e93c361a9, for GNU/Linux 3.2.0, with debug_info, not stripped $ perf buildid-list -i cpp_args_annotate 4fe3cab260204765605ec630d0dc7a7e93c361a9 $ perf buildid-list | grep cpp_args_annotate 4fe3cab260204765605ec630d0dc7a7e93c361a9 /home/acme/c/cpp_args_annotate $ It now works: $ perf annotate --stdio2 foo Samples: 6K of event 'cycles:u', 4000 Hz, Event count (approx.): 7380681761, [percent: local period] foo() /home/acme/c/cpp_args_annotate Percent 0000000000401106 <foo(int, int)>: foo(int, int): int foo(int a, int b) { push %rbp mov %rsp,%rbp mov %edi,-0x14(%rbp) mov %esi,-0x18(%rbp) for (unsigned i = 0; i < 1000000000; i++) movl $0x0,-0x4(%rbp) 11: cmpl $0x3b9ac9ff,-0x4(%rbp) 86.53 ↓ ja 26 a += b; 13.32 mov -0x18(%rbp),%eax 0.00 add %eax,-0x14(%rbp) for (unsigned i = 0; i < 1000000000; i++) addl $0x1,-0x4(%rbp) 0.15 ↑ jmp 11 return a; 26: mov -0x14(%rbp),%eax } pop %rbp ← retq $ Signed-off-by:Martin Liška <mliska@suse.cz> Tested-by:
-
Kees Cook authored
As started by commit 05a5f51c ("Documentation: Replace lkml.org links with lore"), replace lkml.org links with lore to better use a single source that's more likely to stay available long-term. Signed-off-by:
Kees Kook <keescook@chromium.org> Acked-by:
Peter Zijlstra <peterz@infradead.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Joe Perches <joe@perches.com> Cc: Kees Kook <keescook@chromium.org> Cc: Mark Rutland <mark.rutland@arm.com> Link: http://lore.kernel.org/lkml/20210210234220.2401035-1-keescook@chromium.orgSigned-off-by:
-
Arnaldo Carvalho de Melo authored
To pick a new prctl introduced in: 36a6c843 ("entry: Use different define for selector variable in SUD") That don't result in any changes in tooling: $ tools/perf/trace/beauty/prctl_option.sh > before $ cp include/uapi/linux/prctl.h tools/include/uapi/linux/prctl.h $ tools/perf/trace/beauty/prctl_option.sh > after $ diff -u before after Just silences this perf tools build warning: Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Gabriel Krisman Bertazi <krisman@collabora.com> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by:
-
Jiri Olsa authored
Add a test for the perf daemon 'lock' command ensuring only one instance of daemon can run over one base directory. Committer testing: [root@five ~]# perf test -v daemon 76: daemon operations : --- start --- test child forked, pid 793255 test daemon list test daemon reconfig test daemon stop test daemon signal signal 12 sent to session 'test [793506]' signal 12 sent to session 'test [793506]' test daemon ping test daemon lock test child finished with 0 ---- end ---- daemon operations: Ok [root@five ~]# Signed-off-by:
Jiri Olsa <jolsa@kernel.org> Tested-by:
-
Jiri Olsa authored
Add a test for the perf daemon 'ping' command. The tests verifies the ping command gets proper answer from sessions. Committer testing: [root@five ~]# perf test daemon 76: daemon operations : Ok [root@five ~]# perf test -v daemon 76: daemon operations : --- start --- test child forked, pid 792143 test daemon list test daemon reconfig test daemon stop test daemon signal signal 12 sent to session 'test [792415]' signal 12 sent to session 'test [792415]' test daemon ping test child finished with 0 ---- end ---- daemon operations: Ok [root@five ~]# Signed-off-by:
-
Jiri Olsa authored
Add a test for the perf daemon 'signal' command. The test sends a signal to configured sessions and verifies the perf data files were generated accordingly. Committer testing: [root@five ~]# perf test daemon 76: daemon operations : Ok [root@five ~]# perf test -v daemon 76: daemon operations : --- start --- test child forked, pid 790017 test daemon list test daemon reconfig test daemon stop test daemon signal signal 12 sent to session 'test [790268]' signal 12 sent to session 'test [790268]' test child finished with 0 ---- end ---- daemon operations: Ok [root@five ~]# Signed-off-by:
-
Jiri Olsa authored
Add a test for the perf daemon 'stop' command. The test stops the daemon and verifies all the configured sessions are properly terminated. Committer testing: [root@five ~]# time perf test daemon 76: daemon operations : Ok [root@five ~]# time perf test -v daemon 76: daemon operations : --- start --- test child forked, pid 788560 test daemon list test daemon reconfig test daemon stop test child finished with 0 ---- end ---- daemon operations: Ok # Signed-off-by:
-
Jiri Olsa authored
Add a test for daemon reconfiguration. The test changes the configuration file and checks that the session is changed properly. Committer testing: [root@five ~]# perf test daemon 76: daemon operations : Ok [root@five ~]# time perf test daemon 76: daemon operations : Ok real 0m6.055s user 0m0.174s sys 0m0.147s [root@five ~]# time perf test -v daemon 76: daemon operations : --- start --- test child forked, pid 786863 test daemon list test daemon reconfig test child finished with 0 ---- end ---- daemon operations: Ok real 0m6.127s user 0m0.222s sys 0m0.165s [root@five ~]# Signed-off-by:
-
Jiri Olsa authored
Add test for basic perf daemon listing via the CSV output mode (-x option). Check that the configured sessions display expected values. Committer testing: [root@five ~]# perf test daemon 76: daemon operations : Ok [root@five ~]# [root@five ~]# perf test -v daemon 76: daemon operations : --- start --- test child forked, pid 785037 test daemon list test child finished with 0 ---- end ---- daemon operations: Ok [root@five ~]# Signed-off-by:
-
Jiri Olsa authored
Add usage examples to the man page. Signed-off-by:
-
Jiri Olsa authored
Display up time for both daemon and sessions. Example: # cat ~/.perfconfig [daemon] base=/opt/perfdata [session-cycles] run = -m 10M -e cycles --overwrite --switch-output -a [session-sched] run = -m 20M -e sched:* --overwrite --switch-output -a Starting the daemon: # perf daemon start Get the details with up time: # perf daemon -v [778315:daemon] base: /opt/perfdata output: /opt/perfdata/output lock: /opt/perfdata/lock up: 15 minutes [778316:cycles] perf record -m 20M -e cycles --overwrite --switch-output -a base: /opt/perfdata/session-cycles output: /opt/perfdata/session-cycles/output control: /opt/perfdata/session-cycles/control ack: /opt/perfdata/session-cycles/ack up: 10 minutes [778317:sched] perf record -m 20M -e sched:* --overwrite --switch-output -a base: /opt/perfdata/session-sched output: /opt/perfdata/session-sched/output control: /opt/perfdata/session-sched/control ack: /opt/perfdata/session-sched/ack up: 2 minutes Signed-off-by: -
Jiri Olsa authored
Use the 'stop' control command to stop perf record session. If that fails, fall back to current SIGTERM/SIGKILL pair. Signed-off-by:
-
Jiri Olsa authored
Add a 'ping' command to verify that the 'perf record' session is up and operational. It's used in the following patches via test code to make sure 'perf record' is ready to receive signals. Example: # cat ~/.perfconfig [daemon] base=/opt/perfdata [session-cycles] run = -m 10M -e cycles --overwrite --switch-output -a [session-sched] run = -m 20M -e sched:* --overwrite --switch-output -a Start the daemon: # perf daemon start Ping all sessions: # perf daemon ping OK cycles OK sched Ping specific session: # perf daemon ping --session sched OK sched Committer notes: Fixed up bug pointed by clang: Buggy: if (!pollfd.revents & POLLIN) Correct code: if (!(pollfd.revents & POLLIN)) clang warning: builtin-daemon.c:560:6: error: logical not is only applied to the left hand side of this bitwise operator [-Werror,-Wlogical-not-parentheses] if (!pollfd.revents & POLLIN) { ^ ~ builtin-daemon.c:560:6: note: add parentheses after the '!' to evaluate the bitwise operator first Also use designated initialized with pollfd, i.e.: struct pollfd pollfd = { .events = POLLIN, }; Instead of: struct pollfd pollfd = { 0, }; To get past: builtin-daemon.c:510:30: error: missing field 'events' initializer [-Werror,-Wmissing-field-initializers] struct pollfd pollfd = { 0, }; ^ 1 error generated. Signed-off-by: -
Jiri Olsa authored
Setup control fifos for session and add --control option to session arguments. Example: # cat ~/.perfconfig [daemon] base=/opt/perfdata [session-cycles] run = -m 10M -e cycles --overwrite --switch-output -a [session-sched] run = -m 20M -e sched:* --overwrite --switch-output -a Starting the daemon: # perf daemon start Use can list control fifos with (control and ack files): # perf daemon -v [776459:daemon] base: /opt/perfdata output: /opt/perfdata/output lock: /opt/perfdata/lock [776460:cycles] perf record -m 20M -e cycles --overwrite --switch-output -a base: /opt/perfdata/session-cycles output: /opt/perfdata/session-cycles/output control: /opt/perfdata/session-cycles/control ack: /opt/perfdata/session-cycles/ack [776461:sched] perf record -m 20M -e sched:* --overwrite --switch-output -a base: /opt/perfdata/session-sched output: /opt/perfdata/session-sched/output control: /opt/perfdata/session-sched/control ack: /opt/perfdata/session-sched/ack Signed-off-by: -
Jiri Olsa authored
Add 'lock' file under daemon base and flock it, so only one perf daemon can run on top of it. Each daemon tries to create and lock BASE/lock file, if it's successful we are sure we're the only daemon running over the BASE. Once daemon is finished, file descriptor to lock file is closed and lock is released. Example: # cat ~/.perfconfig [daemon] base=/opt/perfdata [session-cycles] run = -m 10M -e cycles --overwrite --switch-output -a [session-sched] run = -m 20M -e sched:* --overwrite --switch-output -a Starting the daemon: # perf daemon start And try once more: # perf daemon start failed: another perf daemon (pid 775594) owns /opt/perfdata will end up with an error, because there's already one running on top of /opt/perfdata. Committer notes: Provide lockf(F_TLOCK) when not available, i.e. transform: lockf(fd, F_TLOCK, 0); into: flock(fd, LOCK_EX | LOCK_NB); Which should be equivalent. Noticed when cross building to some odd Android NDK. Signed-off-by:
-
Jiri Olsa authored
Add 'perf daemon stop' command to stop daemon process and all running sessions. Example: # cat ~/.perfconfig [daemon] base=/opt/perfdata [session-cycles] run = -m 10M -e cycles --overwrite --switch-output -a [session-sched] run = -m 20M -e sched:* --overwrite --switch-output -a Start the daemon: # perf daemon start Stop the daemon # perf daemon stop Daemon is not running, nothing to connect to: # perf daemon connect error: Connection refused Signed-off-by:
-
Jiri Olsa authored
Allow the 'perf daemon' to send SIGUSR2 to all running sessions or just to a specific session. Example: # cat ~/.perfconfig [daemon] base=/opt/perfdata [session-cycles] run = -m 10M -e cycles --overwrite --switch-output -a [session-sched] run = -m 20M -e sched:* --overwrite --switch-output -a Start the daemon: # perf daemon start Send signal to all running sessions: # perf daemon signal signal 12 sent to session 'cycles [773738]' signal 12 sent to session 'sched [773739]' Or to specific one: # perf daemon signal --session sched signal 12 sent to session 'sched [773739]' And verify signals were delivered and perf.data dumped: # cat /opt/perfdata/session-cycles/output rounding mmap pages size to 32M (8192 pages) [ perf record: dump data: Woken up 1 times ] [ perf record: Dump perf.data.2021010220382490 ] # car /opt/perfdata/session-sched/output rounding mmap pages size to 32M (8192 pages) [ perf record: dump data: Woken up 1 times ] [ perf record: Dump perf.data.2021010220382489 ] [ perf record: dump data: Woken up 1 times ] [ perf record: Dump perf.data.2021010220393745 ] Signed-off-by:
-
Jiri Olsa authored
Add a 'list' command to display all running sessions. It's the default command if no other command is specified. Example: # cat ~/.perfconfig [daemon] base=/opt/perfdata [session-cycles] run = -m 10M -e cycles --overwrite --switch-output -a [session-sched] run = -m 20M -e sched:* --overwrite --switch-output -a Start the daemon: # perf daemon start List sessions: # perf daemon [771394:daemon] base: /opt/perfdata [771395:cycles] perf record -m 10M -e cycles --overwrite --switch-output -a [771396:sched] perf record -m 20M -e sched:* --overwrite --switch-output -a List sessions with more info: # perf daemon -v [771394:daemon] base: /opt/perfdata output: /opt/perfdata/output [771395:cycles] perf record -m 10M -e cycles --overwrite --switch-output -a base: /opt/perfdata/session-cycles output: /opt/perfdata/session-cycles/output [771396:sched] perf record -m 20M -e sched:* --overwrite --switch-output -a base: /opt/perfdata/session-sched output: /opt/perfdata/session-sched/output The 'output' file is perf record output for specific session. Note you have to stop all running perf processes manually at this point, stop command is coming in following patches. Committer notes: Fixup union initialization to overcome this in multiple older systems: 22 15.74 debian:8 : FAIL gcc version 4.9.2 (Debian 4.9.2-10+deb8u2) builtin-daemon.c: In function 'send_cmd_list': builtin-daemon.c:1386:2: error: missing initializer for field 'csv_sep' of 'struct <anonymous>' [-Werror=missing-field-initializers] }; ^ builtin-daemon.c:641:8: note: 'csv_sep' declared here char csv_sep; ^ cc1: all warnings being treated as errors Signed-off-by: -
Jiri Olsa authored
Use a signalfd fd to track SIGCHLD signals as notifications for perf session termination. This way we don't need to actively check for child status, being notified if there's change. Suggested-by:
Alexei Budankov <abudankov@huawei.com> Signed-off-by:
-
Jiri Olsa authored
Add support to put the daemon process in the background. It's now enabled by default and -f option is added to keep the daemon process on the console for debugging. Signed-off-by:
-
Jiri Olsa authored
Add support to detect changes to the daemon's config file triggering a re-read of the configuration when that happens. Use a inotify file descriptor plugged into the main fdarray object for polling. Example: # cat ~/.perfconfig [daemon] base=/opt/perfdata [session-cycles] run = -m 10M -e cycles --overwrite --switch-output -a Starting the daemon: # perf daemon start Check sessions: # perf daemon [772262:daemon] base: /opt/perfdata [772263:cycles] perf record -m 10M -e cycles --overwrite --switch-output -a Change '-m 10M' to '-m 20M', and check daemon log: # tail -f /opt/perfdata/output [2021-01-02 20:31:41.234045] daemon started (pid 772262) [2021-01-02 20:31:41.235072] reconfig: ruining session [cycles:772263]: -m 10M -e cycles --overwrite --switch-output -a [2021-01-02 20:32:08.310137] reconfig: session 'cycles' killed [2021-01-02 20:32:08.310847] reconfig: ruining session [cycles:772338]: -m 20M -e cycles --overwrite --switch-output -a And the session list: # perf daemon [772262:daemon] base: /opt/perfdata [772338:cycles] perf record -m 20M -e cycles --overwrite --switch-output -a Note the changed '-m 20M' option is in place. Signed-off-by:
-
Jiri Olsa authored
Adding support to configure daemon with config file. Each client or server invocation of perf daemon needs to know the base directory, where all sessions data is stored. The base is defined with: daemon.base Base path for daemon data. All sessions data are stored under this path. The daemon allows to create record sessions. Each session is a record command spawned and monitored by perf daemon. The session is defined with: session-<NAME>.run Defines new record session for daemon. The value is record's command line without the 'record' keyword. Example: # cat ~/.perfconfig [daemon] base=/opt/perfdata [session-cycles] run = -m 10M -e cycles --overwrite --switch-output -a [session-sched] run = -m 20M -e sched:* --overwrite --switch-output -a The example above defines '/opt/perfdata' as the base directory and 2 record sessions. # perf daemon start [2021-01-28 19:47:33.454413] daemon started (pid 16015) [2021-01-28 19:47:33.455910] reconfig: ruining session [cycles:16016]: -m 10M -e cycles --overwrite --switch-output -a [2021-01-28 19:47:33.456599] reconfig: ruining session [sched:16017]: -m 20M -e sched:* --overwrite --switch-output -a # ps -ef | grep perf ... perf daemon start ... /home/jolsa/.../perf record -m 20M -e cycles --overwrite --switch-output -a ... /home/jolsa/.../perf record -m 20M -e sched:* --overwrite --switch-output -a The base directory is populated with: # find /opt/perfdata/ /opt/perfdata/ /opt/perfdata/control <- control socket /opt/perfdata/session-cycles <- data for session 'cycles': /opt/perfdata/session-cycles/output <- perf record output /opt/perfdata/session-cycles/perf.data <- perf data /opt/perfdata/session-sched <- ditto for session 'sched' /opt/perfdata/session-sched/output /opt/perfdata/session-sched/perf.data Signed-off-by: -
Jiri Olsa authored
Add support for client socket side that will be used to send commands to the daemon server socket. This patch adds only the core support, all commands using this functionality are coming in the following patches. Committer notes: Hat to patch patch it to deal with this in some systems: cc1: warnings being treated as errors builtin-daemon.c: In function 'send_cmd': MKDIR /tmp/build/perf/bench/ builtin-daemon.c:1368: error: ignoring return value of 'fwrite', declared with attribute warn_unused_result MKDIR /tmp/build/perf/tests/ make[3]: *** [/tmp/build/perf/builtin-daemon.o] Error 1 And also to not leak the 'line' buffer allocated by getline(), since you initialized line to NULL and len to zero, man page says: If *lineptr is set to NULL and *n is set 0 before the call, then getline() will allocate a buffer for storing the line. This buffer should be freed by the user program even if getline() failed. Signed-off-by:
-
- 10 Feb, 2021 4 commits
-
-
git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pmLinus Torvalds authored
Pull power management fixes from Rafael Wysocki: "Address a performance regression related to scale-invariance on x86 that may prevent turbo CPU frequencies from being used in certain workloads on systems using acpi-cpufreq as the CPU performance scaling driver and schedutil as the scaling governor" * tag 'pm-5.11-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm: cpufreq: ACPI: Update arch scale-invariance max perf ratio if CPPC is not there cpufreq: ACPI: Extend frequency tables to cover boost frequencies
-
git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pmLinus Torvalds authored
Pull ACPI fix from Rafael Wysocki: "Revert a problematic ACPICA commit that changed the code to attempt to update memory regions which may be read-only on some systems (Ard Biesheuvel)" * tag 'acpi-5.11-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm: Revert "ACPICA: Interpreter: fix memory leak by using existing buffer"
-
git://git.kernel.org/pub/scm/linux/kernel/git/vkoul/dmaengineLinus Torvalds authored
Pull dmaengine fixes from Vinod Koul: "Some late fixes for dmaengine: Core: - fix channel device_node deletion Driver fixes: - dw: revert of runtime pm enabling - idxd: device state fix, interrupt completion and list corruption - ti: resource leak * tag 'dmaengine-fix2-5.11' of git://git.kernel.org/pub/scm/linux/kernel/git/vkoul/dmaengine: dmaengine dw: Revert "dmaengine: dw: Enable runtime PM" dmaengine: idxd: check device state before issue command dmaengine: ti: k3-udma: Fix a resource leak in an error handling path dmaengine: move channel device_node deletion to driver dmaengine: idxd: fix misc interrupt completion dmaengine: idxd: Fix list corruption in description completion
-
git://git.kernel.org/pub/scm/linux/kernel/git/netdev/netLinus Torvalds authored
Pull networking fixes from David Miller: "Another pile of networing fixes: 1) ath9k build error fix from Arnd Bergmann 2) dma memory leak fix in mediatec driver from Lorenzo Bianconi. 3) bpf int3 kprobe fix from Alexei Starovoitov. 4) bpf stackmap integer overflow fix from Bui Quang Minh. 5) Add usb device ids for Cinterion MV31 to qmi_qwwan driver, from Christoph Schemmel. 6) Don't update deleted entry in xt_recent netfilter module, from Jazsef Kadlecsik. 7) Use after free in nftables, fix from Pablo Neira Ayuso. 8) Header checksum fix in flowtable from Sven Auhagen. 9) Validate user controlled length in qrtr code, from Sabyrzhan Tasbolatov. 10) Fix race in xen/netback, from Juergen Gross, 11) New device ID in cxgb4, from Raju Rangoju. 12) Fix ring locking in rxrpc release call, from David Howells. 13) Don't return LAPB error codes from x25_open(), from Xie He. 14) Missing error returns in gsi_channel_setup() from Alex Elder. 15) Get skb_copy_and_csum_datagram working properly with odd segment sizes, from Willem de Bruijn. 16) Missing RFS/RSS table init in enetc driver, from Vladimir Oltean. 17) Do teardown on probe failure in DSA, from Vladimir Oltean. 18) Fix compilation failures of txtimestamp selftest, from Vadim Fedorenko. 19) Limit rx per-napi gro queue size to fix latency regression, from Eric Dumazet. 20) dpaa_eth xdp fixes from Camelia Groza. 21) Missing txq mode update when switching CBS off, in stmmac driver, from Mohammad Athari Bin Ismail. 22) Failover pending logic fix in ibmvnic driver, from Sukadev Bhattiprolu. 23) Null deref fix in vmw_vsock, from Norbert Slusarek. 24) Missing verdict update in xdp paths of ena driver, from Shay Agroskin. 25) seq_file iteration fix in sctp from Neil Brown. 26) bpf 32-bit src register truncation fix on div/mod, from Daniel Borkmann. 27) Fix jmp32 pruning in bpf verifier, from Daniel Borkmann. 28) Fix locking in vsock_shutdown(), from Stefano Garzarella. 29) Various missing index bound checks in hns3 driver, from Yufeng Mo. 30) Flush ports on .phylink_mac_link_down() in dsa felix driver, from Vladimir Oltean. 31) Don't mix up stp and mrp port states in bridge layer, from Horatiu Vultur. 32) Fix locking during netif_tx_disable(), from Edwin Peer" * git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (45 commits) bpf: Fix 32 bit src register truncation on div/mod bpf: Fix verifier jmp32 pruning decision logic bpf: Fix verifier jsgt branch analysis on max bound vsock: fix locking in vsock_shutdown() net: hns3: add a check for index in hclge_get_rss_key() net: hns3: add a check for tqp_index in hclge_get_ring_chain_from_mbx() net: hns3: add a check for queue_id in hclge_reset_vf_queue() net: dsa: felix: implement port flushing on .phylink_mac_link_down switchdev: mrp: Remove SWITCHDEV_ATTR_ID_MRP_PORT_STAT bridge: mrp: Fix the usage of br_mrp_port_switchdev_set_state net: watchdog: hold device global xmit lock during tx disable netfilter: nftables: relax check for stateful expressions in set definition netfilter: conntrack: skip identical origin tuple in same zone only vsock/virtio: update credit only if socket is not closed net: fix iteration for sctp transport seq_files net: ena: Update XDP verdict upon failure net/vmw_vsock: improve locking in vsock_connect_timeout() net/vmw_vsock: fix NULL pointer dereference ibmvnic: Clear failover_pending if unable to schedule net: stmmac: set TxQ mode back to DCB after disabling CBS ...
-
