- 04 Oct, 2016 11 commits
-
-
Andi Kleen authored
Make alias matching the events parser case-insensitive. This is useful with the JSON events. perf uses lower case events, but the CPU manuals generally use upper case event names. The JSON files use lower case by default too. But if we search case insensitively then users can cut-n-paste the upper case event names. So the following works: % perf stat -e BR_INST_EXEC.TAKEN_INDIRECT_NEAR_CALL true Performance counter stats for 'true': 305 BR_INST_EXEC.TAKEN_INDIRECT_NEAR_CALL 0.000492799 seconds time elapsed Signed-off-by:Andi Kleen <ak@linux.intel.com> Acked-by:
Ingo Molnar <mingo@kernel.org> Tested-by:
Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Madhavan Srinivasan <maddy@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: linuxppc-dev@lists.ozlabs.org Link: http://lkml.kernel.org/r/1473978296-20712-17-git-send-email-sukadev@linux.vnet.ibm.comSigned-off-by:
-
Sukadev Bhattiprolu authored
This avoids the JSON PMU events parser having to know whether its aliases are for perf stat or perf record. Signed-off-by:
Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com> Acked-by:
-
Sukadev Bhattiprolu authored
Signed-off-by:
Jiri Olsa <jolsa@redhat.com> Cc: Madhavan Srinivasan <maddy@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: linuxppc-dev@lists.ozlabs.org Link: http://lkml.kernel.org/r/1473978296-20712-16-git-send-email-sukadev@linux.vnet.ibm.comSigned-off-by:
-
Andi Kleen authored
Add support to group the output of perf list by the Topic field in the JSON file. Example output: % perf list ... Cache: l1d.replacement [L1D data line replacements] l1d_pend_miss.pending [L1D miss oustandings duration in cycles] l1d_pend_miss.pending_cycles [Cycles with L1D load Misses outstanding] l2_l1d_wb_rqsts.all [Not rejected writebacks from L1D to L2 cache lines in any state] l2_l1d_wb_rqsts.hit_e [Not rejected writebacks from L1D to L2 cache lines in E state] l2_l1d_wb_rqsts.hit_m [Not rejected writebacks from L1D to L2 cache lines in M state] ... Pipeline: arith.fpu_div [Divide operations executed] arith.fpu_div_active [Cycles when divider is busy executing divide operations] baclears.any [Counts the total number when the front end is resteered, mainly when the BPU cannot provide a correct prediction and this is corrected by other branch handling mechanisms at the front end] br_inst_exec.all_branches [Speculative and retired branches] br_inst_exec.all_conditional [Speculative and retired macro-conditional branches] br_inst_exec.all_direct_jmp [Speculative and retired macro-unconditional branches excluding calls and indirects] br_inst_exec.all_direct_near_call [Speculative and retired direct near calls] br_inst_exec.all_indirect_jump_non_call_ret Signed-off-by: -
Sukadev Bhattiprolu authored
Previously we were dropping the useful longer descriptions that some events have in the event list completely. This patch makes them appear with perf list. Old perf list: baclears: baclears.all [Counts the number of baclears] vs new: perf list -v: ... baclears: baclears.all [The BACLEARS event counts the number of times the front end is resteered, mainly when the Branch Prediction Unit cannot provide a correct prediction and this is corrected by the Branch Address Calculator at the front end. The BACLEARS.ANY event counts the number of baclears for any type of branch] Signed-off-by: -
Sukadev Bhattiprolu authored
Implement support in jevents to parse long descriptions for events that may have them in the JSON files. A follow on patch will make this long description available to user through the 'perf list' command. Signed-off-by:
-
Andi Kleen authored
Add a PERF_CPUID variable to override the CPUID of the current CPU (within the current architecture). This is useful for testing, so that all event lists can be tested on a single system. Signed-off-by:
-
Andi Kleen authored
Add a --no-desc flag to 'perf list' to not print the event descriptions that were earlier added for JSON events. This may be useful to get a less crowded listing. It's still default to print descriptions as that is the more useful default for most users. Signed-off-by:
-
Andi Kleen authored
Automatically adapt the now wider and word wrapped perf list output to wider terminals. This requires querying the terminal before the auto pager takes over, and exporting this information from the pager subsystem. Signed-off-by:
Namhyung Kim <namhyung@kernel.org> Tested-by:
-
Andi Kleen authored
Add support to print alias descriptions in perf list, which are taken from the generated event files. The sorting code is changed to put the events with descriptions at the end. The descriptions are printed as possibly multiple word wrapped lines. Example output: % perf list ... arith.fpu_div [Divide operations executed] arith.fpu_div_active [Cycles when divider is busy executing divide operations] Committer notes: Further testing on a Broadwell machine (ThinkPad t450s), using these files: $ find tools/perf/pmu-events/arch/x86/ tools/perf/pmu-events/arch/x86/ tools/perf/pmu-events/arch/x86/Broadwell tools/perf/pmu-events/arch/x86/Broadwell/Cache.json tools/perf/pmu-events/arch/x86/Broadwell/Other.json tools/perf/pmu-events/arch/x86/Broadwell/Frontend.json tools/perf/pmu-events/arch/x86/Broadwell/Virtual-Memory.json tools/perf/pmu-events/arch/x86/Broadwell/Pipeline.json tools/perf/pmu-events/arch/x86/Broadwell/Floating-point.json tools/perf/pmu-events/arch/x86/Broadwell/Memory.json tools/perf/pmu-events/arch/x86/mapfile.csv $ Taken from: https://github.com/sukadev/linux/tree/json-code+data-v21/tools/perf/pmu-events/arch/x86/ to get this machinery to actually parse JSON files, generate $(OUTPUT)pmu-events/pmu-events.c, compile it and link it with perf, that will then use the table it contains, these files will be submitted right after this patchkit. [acme@jouet linux]$ perf list page_walker List of pre-defined events (to be used in -e): page_walker_loads.dtlb_l1 [Number of DTLB page walker hits in the L1+FB] page_walker_loads.dtlb_l2 [Number of DTLB page walker hits in the L2] page_walker_loads.dtlb_l3 [Number of DTLB page walker hits in the L3 + XSNP] page_walker_loads.dtlb_memory [Number of DTLB page walker hits in Memory] page_walker_loads.itlb_l1 [Number of ITLB page walker hits in the L1+FB] page_walker_loads.itlb_l2 [Number of ITLB page walker hits in the L2] page_walker_loads.itlb_l3 [Number of ITLB page walker hits in the L3 + XSNP] [acme@jouet linux]$ Signed-off-by: -
Andi Kleen authored
To work with existing mapfiles, assume that the first line in 'mapfile.csv' is a header line and skip over it. Signed-off-by:
-
- 03 Oct, 2016 9 commits
-
-
Andi Kleen authored
Implement the code to match CPU types to mapfile types for x86 based on CPUID. This extends an existing similar function, but changes it to use the x86 mapfile cpu description. This allows to resolve event lists generated by jevents. Signed-off-by:
-
Sukadev Bhattiprolu authored
Implement code that returns the generic CPU ID string for Powerpc. This will be used to identify the specific table of PMU events to parse/compare user specified events against. Signed-off-by:
-
Sukadev Bhattiprolu authored
At run time (when 'perf' is starting up), locate the specific table of PMU events that corresponds to the current CPU. Using that table, create aliases for the each of the PMU events in the CPU. The use these aliases to parse the user specified perf event. In short this would allow the user to specify events using their aliases rather than raw event codes. Based on input and some earlier patches from Andi Kleen, Jiri Olsa. Signed-off-by:
-
Andi Kleen authored
This is a modified version of an earlier patch by Andi Kleen. We expect architectures to create JSON files describing the performance monitoring (PMU) events that each CPU model/family of the architecture supports. Following is an example of the JSON file entry for an x86 event: [ ... { "EventCode": "0x00", "UMask": "0x01", "EventName": "INST_RETIRED.ANY", "BriefDescription": "Instructions retired from execution.", "PublicDescription": "Instructions retired from execution.", "Counter": "Fixed counter 1", "CounterHTOff": "Fixed counter 1", "SampleAfterValue": "2000003", "SampleAfterValue": "2000003", "MSRIndex": "0", "MSRValue": "0", "TakenAlone": "0", "CounterMask": "0", "Invert": "0", "AnyThread": "0", "EdgeDetect": "0", "PEBS": "0", "PRECISE_STORE": "0", "Errata": "null", "Offcore": "0" }, ... ] All the PMU events supported by a CPU model/family must be grouped into "topics" such as "Pipelining", "Floating-point", "Virtual-memory" etc. All events belonging to a topic must be placed in a separate JSON file (eg: "Pipelining.json") and all the topic JSON files for a CPU model must be in a separate directory. Eg: for the CPU model "Silvermont_core": $ ls tools/perf/pmu-events/arch/x86/Silvermont_core Floating-point.json Memory.json Other.json Pipelining.json Virtualmemory.json Finally, to allow multiple CPU models to share a single set of JSON files, architectures must provide a mapping between a model and its set of events: $ grep Silvermont tools/perf/pmu-events/arch/x86/mapfile.csv GenuineIntel-6-4D,V13,Silvermont_core,core GenuineIntel-6-4C,V13,Silvermont_core,core which maps each CPU, identified by [vendor, family, model, version, type] to a directory of JSON files. Thus two (or more) CPU models support the set of PMU events listed in the directory. tools/perf/pmu-events/arch/x86/Silvermont_core/ Given this organization of files, the program, jevents: - locates all JSON files for each CPU-model of the architecture, - parses all JSON files for the CPU-model and generates a C-style "PMU-events table" (pmu-events.c) for the model - locates a mapfile for the architecture - builds a global table, mapping each model of CPU to the corresponding PMU-events table. The 'pmu-events.c' is generated when building perf and added to libperf.a. The global table pmu_events_map[] table in this pmu-events.c will be used in perf in a follow-on patch. If the architecture does not have any JSON files or there is an error in processing them, an empty mapping file is created. This would allow the build of perf to proceed even if we are not able to provide aliases for events. The parser for JSON files allows parsing Intel style JSON event files. This allows to use an Intel event list directly with perf. The Intel event lists can be quite large and are too big to store in unswappable kernel memory. The conversion from JSON to C-style is straight forward. The parser knows (very little) Intel specific information, and can be easily extended to handle fields for other CPUs. The parser code is partially shared with an independent parsing library, which is 2-clause BSD licensed. To avoid any conflicts I marked those files as BSD licensed too. As part of perf they become GPLv2. Committer notes: Fixes: 1) Limit maxfds to 512 to avoid nftd() segfaulting on alloca() with a big rlim_max, as in docker containers - acme 2) Make jevents a hostprog, supporting cross compilation - jolsa 3) Use HOSTCC for jevents final step - acme 4) Define _GNU_SOURCE for asprintf, as we can't use CC's EXTRA_CFLAGS, that has to have --sysroot on the Android NDK 24 - acme 5) Removed $(srctree)/tools/perf/pmu-events/pmu-events.c from the 'clean' target, it is generated on $(OUTPUT)pmu-events/pmu-events.c, which is already taken care of in the original patch - acme Signed-off-by: -
Andi Kleen authored
I need a JSON parser. This adds the simplest JSON parser I could find -- Serge Zaitsev's jsmn `jasmine' -- to the perf library. I merely converted it to (mostly) Linux style and added support for non 0 terminated input. The parser is quite straight forward and does not copy any data, just returns tokens with offsets into the input buffer. So it's relatively efficient and simple to use. The code is not fully checkpatch clean, but I didn't want to completely fork the upstream code. Original source: http://zserge.bitbucket.org/jsmn.html In addition I added a simple wrapper that mmaps a json file and provides some straight forward access functions. Used in follow-on patches to parse event files. Signed-off-by:
-
Jiri Olsa authored
It is used in the build process, so stop suppressing its build in tools cross builds. Signed-off-by:
Jiri Olsa <jolsa@kernel.org> Tested-by:
-
Jiri Olsa authored
In some cases, like for fixdep and shortly for jevents, we need to build a tool to run on the host that will be used in building a tool, such as perf, that is being cross compiled, so do like the kernel and provide HOSTCC, HOSTLD and HOSTAR to do that. Signed-off-by:
Andi Kleen <andi@firstfloor.org> Requested-and-Tested-by:
-
Arnaldo Carvalho de Melo authored
Experimenting a bit using cppcheck[1], a static checker brought to my attention by Colin, reducing the scope of some variables, reducing the line of source code lines in the process: $ cppcheck --enable=style tools/perf/util/thread.c Checking tools/perf/util/thread.c... [tools/perf/util/thread.c:17]: (style) The scope of the variable 'leader' can be reduced. [tools/perf/util/thread.c:133]: (style) The scope of the variable 'err' can be reduced. [tools/perf/util/thread.c:273]: (style) The scope of the variable 'err' can be reduced. Will continue later, but these are already useful, keep them. 1: https://sourceforge.net/p/cppcheck/wiki/Home/ Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Colin Ian King <colin.king@canonical.com> Cc: David Ahern <dsahern@gmail.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Wang Nan <wangnan0@huawei.com> Link: http://lkml.kernel.org/n/tip-ixws7lbycihhpmq9cc949ti6@git.kernel.orgSigned-off-by:
-
Colin Ian King authored
Static anaylsis with cppcheck[1] detected an incorrect comparison: [tools/perf/util/probe-event.c:216]: (warning) Char literal compared with pointer 'ptr2'. Did you intend to dereference it? Dereference ptr2 for the comparison to fix this. 1: https://sourceforge.net/p/cppcheck/wiki/Home/Signed-off-by:
Colin King <colin.king@canonical.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Masami Hiramatsu <mhiramat@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Ravi Bangoria <ravi.bangoria@linux.vnet.ibm.com> Cc: Wang Nan <wangnan0@huawei.com> Fixes: 35726d3a ("perf probe: Fix to cut off incompatible chars from group name") Link: http://lkml.kernel.org/r/20161003103431.18534-1-colin.king@canonical.comSigned-off-by:
-
- 29 Sep, 2016 20 commits
-
-
Ingo Molnar authored
Merge tag 'perf-core-for-mingo-20160929' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/core Pull perf/core improvements and fixes from Arnaldo Carvalho de Melo: User visible changes: --------------------- New features: - Add support for using symbols in address filters with Intel PT and ARM CoreSight (hardware assisted tracing facilities) (Adrian Hunter, Mathieu Poirier) Fixes: - Fix MMAP event synthesis for pre-existing threads when no hugetlbfs mount is in place (Adrian Hunter) - Don't ignore kernel idle symbols in 'perf script' (Adrian Hunter) - Assorted Intel PT fixes (Adrian Hunter) Improvements: - Fix handling of C++ symbols in 'perf probe' (Masami Hiramatsu) - Beautify sched_[gs]et_attr return value in 'perf trace' (Arnaldo Carvalho de Melo) Infrastructure changes: ----------------------- New features: - Add dwarf unwind 'perf test' for powerpc (Ravi Bangoria) Fixes: - Fix error paths in 'perf record' (Adrian Hunter) Documentation: - Update documentation info about quipper, a C++ parser for converting to/from perf.data/chromium profiling format (Simon Que) Build Fixes: Fix building in 32 bit platform with libbabeltrace (Wang Nan) Signed-off-by:
-
Ravi Bangoria authored
The user stack dump feature was recently added for powerpc. But there was no test case available to test it. This test works same as on other architectures by preparing a stack frame on the perf test thread and comparing each frame by unwinding it. $ ./perf test 50 50: Test dwarf unwind : Ok User stack dump for powerpc: https://lkml.org/lkml/2016/4/28/482Signed-off-by:Ravi Bangoria <ravi.bangoria@linux.vnet.ibm.com> Acked-by:
-
Masami Hiramatsu authored
Match linkage name with mangled name if exists. The linkage_name is used for storing mangled name of the object. Thus, this allows 'perf probe' to find appropriate probe point from mangled symbol as below. E.g. without this fix: ---- $ perf probe -x /usr/lib64/libstdc++.so.6 \ -D _ZNKSt15basic_fstreamXXIwSt11char_traitsIwEE7is_openEv Probe point '_ZNKSt15basic_fstreamXXIwSt11char_traitsIwEE7is_openEv' not found. Error: Failed to add events. ---- With this fix, perf probe can find the correct one. ---- $ perf probe -x /usr/lib64/libstdc++.so.6 \ -D _ZNKSt15basic_fstreamXXIwSt11char_traitsIwEE7is_openEv p:probe_libstdc/_ZNKSt15basic_fstreamXXIwSt11char_traitsIwEE7is_openEv /usr/lib64/libstdc++.so.6.0.22:0x8ca60 ---- Committer notes: After the fix, setting it for real (no -D/--definition, that amounts to a --dry-run): # perf probe -x /usr/lib64/libstdc++.so.6 _ZNKSt15basic_fstreamXXIwSt11char_traitsIwEE7is_openEv Added new event: probe_libstdc:_ZNKSt15basic_fstreamXXIwSt11char_traitsIwEE7is_openEv (on _ZNKSt15basic_fstreamXXIwSt11char_traitsIwEE7is_openEv in /usr/lib64/libstdc++.so.6.0.22) You can now use it in all perf tools, such as: perf record -e probe_libstdc:_ZNKSt15basic_fstreamXXIwSt11char_traitsIwEE7is_openEv -aR sleep 1 # perf probe -l probe_libstdc:* probe_libstdc:_ZNKSt15basic_fstreamXXIwSt11char_traitsIwEE7is_openEv (on is_open@libstdc++-v3/include/fstream in /usr/lib64/libstdc++.so.6.0.22) # Reported-by:Thomas Gleixner <tglx@linutronix.de> Signed-off-by:
Masami Hiramatsu <mhiramat@kernel.org> Tested-by:
-
Masami Hiramatsu authored
Cut off the characters which can not use for group name of uprobes when making it based on executable filename. For example, if the exec name is libstdc++.so, without this fix perf probe generates "probe_libstdc++" as the group name, but it is failed to set because '+' can not be used for group name. With this fix perf accepts only alphabet, number or '_' for group name, thus perf generates "probe_libstdc" as the group name. E.g. with this fix, you can see the event name has no "+". ---- $ ./perf probe -x /usr/lib64/libstdc++.so.6 -D is_open p:probe_libstdc/is_open /usr/lib64/libstdc++.so.6.0.22:0x8ca80 p:probe_libstdc/is_open_1 /usr/lib64/libstdc++.so.6.0.22:0x8ca70 p:probe_libstdc/is_open_2 /usr/lib64/libstdc++.so.6.0.22:0x8ca60 p:probe_libstdc/is_open_3 /usr/lib64/libstdc++.so.6.0.22:0xb0ad0 p:probe_libstdc/is_open_4 /usr/lib64/libstdc++.so.6.0.22:0xecca9 ---- Committer note: Before this fix: # perf probe -x /usr/lib64/libstdc++.so.6 is_open Failed to write event: Invalid argument Error: Failed to add events. # After the fix: # perf probe -x /usr/lib64/libstdc++.so.6 is_open Added new events: probe_libstdc:is_open (on is_open in /usr/lib64/libstdc++.so.6.0.22) probe_libstdc:is_open_1 (on is_open in /usr/lib64/libstdc++.so.6.0.22) probe_libstdc:is_open_2 (on is_open in /usr/lib64/libstdc++.so.6.0.22) probe_libstdc:is_open_3 (on is_open in /usr/lib64/libstdc++.so.6.0.22) probe_libstdc:is_open_4 (on is_open in /usr/lib64/libstdc++.so.6.0.22) You can now use it in all perf tools, such as: perf record -e probe_libstdc:is_open_4 -aR sleep 1 # perf probe -l probe_libstdc:* probe_libstdc:is_open (on is_open@libstdc++-v3/include/fstream in /usr/lib64/libstdc++.so.6.0.22) probe_libstdc:is_open_1 (on is_open@libstdc++-v3/include/fstream in /usr/lib64/libstdc++.so.6.0.22) probe_libstdc:is_open_2 (on is_open@libstdc++-v3/include/fstream in /usr/lib64/libstdc++.so.6.0.22) probe_libstdc:is_open_3 (on is_open@src/c++98/basic_file.cc in /usr/lib64/libstdc++.so.6.0.22) probe_libstdc:is_open_4 (on stdio_filebuf:5@include/ext/stdio_filebuf.h in /usr/lib64/libstdc++.so.6.0.22) # Signed-off-by: -
Masami Hiramatsu authored
Skip probes if the entry address of the target function is 0. This can happen when we're handling C++ debuginfo files. E.g. without this fix, below case still fail. ---- $ ./perf probe -x /usr/lib64/libstdc++.so.6 -vD is_open probe-definition(0): is_open symbol:is_open file:(null) line:0 offset:0 return:0 lazy:(null) 0 arguments symbol:catch file:(null) line:0 offset:0 return:0 lazy:(null) symbol:throw file:(null) line:0 offset:0 return:0 lazy:(null) symbol:rethrow file:(null) line:0 offset:0 return:0 lazy:(null) Open Debuginfo file: /usr/lib/debug/usr/lib64/libstdc++.so.6.0.22.debug Try to find probe point from debuginfo. Matched function: is_open [295df] found inline addr: 0x8ca80 Probe point found: is_open+0 found inline addr: 0x8ca70 Probe point found: is_open+0 found inline addr: 0x8ca60 Probe point found: is_open+0 Matched function: is_open [6527f] Matched function: is_open [9fe8a] Probe point found: is_open+0 Matched function: is_open [19710b] found inline addr: 0xecca9 Probe point found: stdio_filebuf+57 found inline addr: 0x0 Probe point found: swap+0 Matched function: is_open [19fc9d] Probe point found: is_open+0 Found 7 probe_trace_events. p:probe_libstdc++/is_open /usr/lib64/libstdc++.so.6.0.22:0x8ca80 p:probe_libstdc++/is_open_1 /usr/lib64/libstdc++.so.6.0.22:0x8ca70 p:probe_libstdc++/is_open_2 /usr/lib64/libstdc++.so.6.0.22:0x8ca60 p:probe_libstdc++/is_open_3 /usr/lib64/libstdc++.so.6.0.22:0xb0ad0 p:probe_libstdc++/is_open_4 /usr/lib64/libstdc++.so.6.0.22:0xecca9 Failed to synthesize probe trace event. Error: Failed to add events. Reason: Invalid argument (Code: -22) ---- This is because some instances have entry_pc == 0 (see 19710b and 19fc9d). With this fix, those are skipped. ---- $ ./perf probe -x /usr/lib64/libstdc++.so.6 -D is_open p:probe_libstdc++/is_open /usr/lib64/libstdc++.so.6.0.22:0x8ca80 p:probe_libstdc++/is_open_1 /usr/lib64/libstdc++.so.6.0.22:0x8ca70 p:probe_libstdc++/is_open_2 /usr/lib64/libstdc++.so.6.0.22:0x8ca60 p:probe_libstdc++/is_open_3 /usr/lib64/libstdc++.so.6.0.22:0xb0ad0 p:probe_libstdc++/is_open_4 /usr/lib64/libstdc++.so.6.0.22:0xecca9 ---- Signed-off-by: -
Masami Hiramatsu authored
Ignore the error when the perf probe failed to find inline function instances. This can happen when we search a method in C++ debuginfo. If there is completely no instance in target, perf probe can return an error. E.g. without this fix: ---- $ perf probe -x /usr/lib64/libstdc++.so.6 -vD showmanyc probe-definition(0): showmanyc symbol:showmanyc file:(null) line:0 offset:0 return:0 lazy:(null) 0 arguments symbol:catch file:(null) line:0 offset:0 return:0 lazy:(null) symbol:throw file:(null) line:0 offset:0 return:0 lazy:(null) symbol:rethrow file:(null) line:0 offset:0 return:0 lazy:(null) Open Debuginfo file: /usr/lib/debug/usr/lib64/libstdc++.so.6.0.22.debug Try to find probe point from debuginfo. Matched function: showmanyc An error occurred in debuginfo analysis (-2). Trying to use symbols. Failed to find symbol showmanyc in /usr/lib64/libstdc++.so.6.0.22 Error: Failed to add events. Reason: No such file or directory (Code: -2) ---- This is because one of showmanyc is defined as inline but no instance found. With this fix, it is succeeded to show as below. ---- $ perf probe -x /usr/lib64/libstdc++.so.6 -D showmanyc p:probe_libstdc++/showmanyc /usr/lib64/libstdc++.so.6.0.22:0xb0e50 p:probe_libstdc++/showmanyc_1 /usr/lib64/libstdc++.so.6.0.22:0xc7c40 p:probe_libstdc++/showmanyc_2 /usr/lib64/libstdc++.so.6.0.22:0xecfa0 p:probe_libstdc++/showmanyc_3 /usr/lib64/libstdc++.so.6.0.22:0x115fc0 p:probe_libstdc++/showmanyc_4 /usr/lib64/libstdc++.so.6.0.22:0x121a90 ---- Signed-off-by: -
Adrian Hunter authored
Due to errata SKL014 "Intel PT TIP.PGD May Not Have Target IP Payload", the Intel PT decoder needs to match address filters against TIP.PGD packets. Parse the address filters and implement the decoder's 'pgd_ip()' callback to match the IP against the filter regions. Signed-off-by:
Adrian Hunter <adrian.hunter@intel.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Masami Hiramatsu <mhiramat@kernel.org> Cc: Mathieu Poirier <mathieu.poirier@linaro.org> Link: http://lkml.kernel.org/r/1474641528-18776-17-git-send-email-adrian.hunter@intel.comSigned-off-by:
-
Adrian Hunter authored
When address filters are used, the decoder must detect the end of a filter region (or a branch into a tracestop region) by matching Packet Generation Disabled (TIP.PGD) packets against the object code using the IP given in the packet. However, due to errata SKL014 "Intel PT TIP.PGD May Not Have Target IP Payload", that IP may not be present. Enable the decoder to handle that by adding a new callback function 'pgd_ip()' which indicates whether the IP is not traced, in which case that is the point where the trace was disabled. Signed-off-by:
-
Adrian Hunter authored
Read the address filter from the AUXTRACE_INFO event in preparation for using it to assist in decoding. Signed-off-by:
-
Adrian Hunter authored
The address filter is needed to help decode the trace, so store it in the AUXTRACE_INFO event. Signed-off-by:
-
Adrian Hunter authored
Add a helper function 'intel_pt_has()' to make it easier to determine which members the AUXTRACE_INFO event contains. Signed-off-by:
-
Adrian Hunter authored
Fix 2 places where the err variable was not being set. Signed-off-by:
-
Adrian Hunter authored
Previously the maximum non-turbo ratio was calculated from TSC assuming a 100 MHz multiplier which is correct for current hardware supporting Intel PT. However more recent kernels also now export the value, so use that in preference to the calculated value. Signed-off-by:
-
Adrian Hunter authored
Fix occasional decoder errors decoding trace data collected in snapshot mode. Snapshot mode can take successive snapshots of trace which might overlap. The decoder checks whether there is an overlap but only looks at the current and previous buffer. However buffers that do not contain synchronization (i.e. PSB) packets cannot be decoded or used for overlap checking. That means the decoder actually needs to check overlaps between the current buffer and the previous buffer that contained usable data. Make that change. Signed-off-by:
-
Adrian Hunter authored
Two SDT debug messages can occur for every DSO which is too noisy. Consequently, increase debug level of SDT messages. Signed-off-by:
-
Adrian Hunter authored
Symbols come from either the DSO or /proc/kallsyms for the kernel. Details of the functionality can be found in Documentation/perf-record.txt. Signed-off-by:
-
Adrian Hunter authored
Add a function to find the last symbol in a DSO. This will be used when parsing address filters to calculate a region that includes the entire DSO. Signed-off-by:
-
Adrian Hunter authored
Some error paths do not tidy-up. Fix that. Signed-off-by:
-
Adrian Hunter authored
In preparation for fixing the error paths, rename label 'out_symbol_exit' to be 'out' because that error path can be used irrespective of whether symbols (or anything else) has been initialized. Signed-off-by:
-
Adrian Hunter authored
Commit 608c34de ("perf symbols: Mark if a symbol is idle in the library") causes idle symbols to vanish from perf script output. That is because print functions suppress symbols marked as 'idle'. However, suppression of 'idle' functions is only used by 'perf top' and 'perf top' does not use the print functions. Consequently that functionality can simply be removed from the print functions. Signed-off-by:
-
