PMU name is printed repeatedly for interval print, for example:
perf stat --no-merge -e 'unc_m_clockticks' -a -I 1000
# time counts unit events
1.001053069 243,702,144 unc_m_clockticks [uncore_imc_4]
1.001053069 244,268,304 unc_m_clockticks [uncore_imc_2]
1.001053069 244,427,386 unc_m_clockticks [uncore_imc_0]
1.001053069 244,583,760 unc_m_clockticks [uncore_imc_5]
1.001053069 244,738,971 unc_m_clockticks [uncore_imc_3]
1.001053069 244,880,309 unc_m_clockticks [uncore_imc_1]
2.002024821 240,818,200 unc_m_clockticks [uncore_imc_4] [uncore_imc_4]
2.002024821 240,767,812 unc_m_clockticks [uncore_imc_2] [uncore_imc_2]
2.002024821 240,764,215 unc_m_clockticks [uncore_imc_0] [uncore_imc_0]
2.002024821 240,759,504 unc_m_clockticks [uncore_imc_5] [uncore_imc_5]
2.002024821 240,755,992 unc_m_clockticks [uncore_imc_3] [uncore_imc_3]
2.002024821 240,750,403 unc_m_clockticks [uncore_imc_1] [uncore_imc_1]
For each print, the PMU name is unconditionally appended to the
counter->name.
Need to check the counter->name first. If the PMU name is already
appended, do nothing.
Committer notes:
Add and use perf_evsel->uniquified_name bool instead of doing the more
expensive strstr(event->name, pmu->name).
Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Agustin Vega-Frias <agustinv@codeaurora.org>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Ganapatrao Kulkarni <ganapatrao.kulkarni@cavium.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Shaokun Zhang <zhangshaokun@hisilicon.com>
Cc: Will Deacon <will.deacon@arm.com>
Fixes: 8c5421c016 ("perf pmu: Display pmu name when printing unmerged events in stat")
Link: http://lkml.kernel.org/r/1524594014-79243-5-git-send-email-kan.liang@linux.intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Perf doesn't support mixed events from different PMUs (except software
event) in a group. The perf stat should output <not counted>/<not
supported> for all events, but it doesn't. For example,
perf stat -e '{cycles,uncore_imc_5/umask=0xF,event=0x4/,instructions}'
<not counted> cycles
<not supported> uncore_imc_5/umask=0xF,event=0x4/
1,024,300 instructions
If perf fails to open an event, it doesn't error out directly. It will
disable some features and retry, until the event is opened or all
features are disabled. The disabled features will not be re-enabled. The
group read is one of these features.
For the example as above, the IMC event and the leader event "cycles"
are from different PMUs. Opening the IMC event must fail. The group read
feature must be disabled for IMC event and the followed event
"instructions". The "instructions" event has the same PMU as the leader
"cycles". It can be opened successfully. Since the group read feature
has been disabled, the "instructions" event will be read as a single
event, which definitely has a value.
The group read fallback is still useful for the case which kernel
doesn't support group read. It is good enough to be handled only by the
leader.
For the fallback request from members, it must be caused by an error.
The fallback only breaks the semantics of group. Limit the group read
fallback only for the leader.
Committer testing:
On a broadwell t450s notebook:
Before:
# perf stat -e '{cycles,unc_cbo_cache_lookup.read_i,instructions}' sleep 1
Performance counter stats for 'sleep 1':
<not counted> cycles
<not supported> unc_cbo_cache_lookup.read_i
818,206 instructions
1.003170887 seconds time elapsed
Some events weren't counted. Try disabling the NMI watchdog:
echo 0 > /proc/sys/kernel/nmi_watchdog
perf stat ...
echo 1 > /proc/sys/kernel/nmi_watchdog
After:
# perf stat -e '{cycles,unc_cbo_cache_lookup.read_i,instructions}' sleep 1
Performance counter stats for 'sleep 1':
<not counted> cycles
<not supported> unc_cbo_cache_lookup.read_i
<not counted> instructions
1.001380511 seconds time elapsed
Some events weren't counted. Try disabling the NMI watchdog:
echo 0 > /proc/sys/kernel/nmi_watchdog
perf stat ...
echo 1 > /proc/sys/kernel/nmi_watchdog
#
Reported-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Agustin Vega-Frias <agustinv@codeaurora.org>
Cc: Ganapatrao Kulkarni <ganapatrao.kulkarni@cavium.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Shaokun Zhang <zhangshaokun@hisilicon.com>
Cc: Will Deacon <will.deacon@arm.com>
Fixes: 82bf311e15 ("perf stat: Use group read for event groups")
Link: http://lkml.kernel.org/r/1524594014-79243-3-git-send-email-kan.liang@linux.intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
When counting uncore event with alias, core event is mistakenly
involved, for example:
perf stat --no-merge -e "unc_m_cas_count.all" -C0 sleep 1
Performance counter stats for 'CPU(s) 0':
0 unc_m_cas_count.all [uncore_imc_4]
0 unc_m_cas_count.all [uncore_imc_2]
0 unc_m_cas_count.all [uncore_imc_0]
153,640 unc_m_cas_count.all [cpu]
0 unc_m_cas_count.all [uncore_imc_5]
25,026 unc_m_cas_count.all [uncore_imc_3]
0 unc_m_cas_count.all [uncore_imc_1]
1.001447890 seconds time elapsed
The reason is that current implementation doesn't check PMU name of a
event when adding its alias into the alias list for core PMU. The
uncore event aliases are mistakenly added.
This bug was introduced in:
commit 14b22ae028 ("perf pmu: Add helper function is_pmu_core to
detect PMU CORE devices")
Checking the PMU name for all PMUs on X86 and other architectures except
ARM.
There is no behavior change for ARM.
Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Agustin Vega-Frias <agustinv@codeaurora.org>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Ganapatrao Kulkarni <ganapatrao.kulkarni@cavium.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Shaokun Zhang <zhangshaokun@hisilicon.com>
Cc: Will Deacon <will.deacon@arm.com>
Fixes: 14b22ae028 ("perf pmu: Add helper function is_pmu_core to detect PMU CORE devices")
Link: http://lkml.kernel.org/r/1524594014-79243-1-git-send-email-kan.liang@linux.intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Currently all the event parsing fails end up in the event_pmu rule, and
display misleading help like:
$ perf stat -e inst kill
event syntax error: 'inst'
\___ Cannot find PMU `inst'. Missing kernel support?
...
The reason is that the event_pmu is too strong and match also single
string. Changing it to force the '/' separators to be part of the rule,
and getting the proper error now:
$ perf stat -e inst kill
event syntax error: 'inst'
\___ parser error
Run 'perf list' for a list of valid events
...
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Reported-by: Ingo Molnar <mingo@kernel.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20180423090823.32309-5-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The 'perf stat' fallback for EACCES error sets the exclude_kernel
perf_event_attr and tries perf_event_open() again with it. In addition,
it also changes the name of the event to reflect that change by adding
the 'u' modifier.
But it does not take into account the '/' separator, so the event name
can end up mangled, like: (note the '/:' characters)
$ perf stat -e cpu/cpu-cycles/ kill
...
386,832 cpu/cpu-cycles/:u
Adding the code to check on the '/' separator and set the following
correct event name:
$ perf stat -e cpu/cpu-cycles/ kill
...
388,548 cpu/cpu-cycles/u
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20180423090823.32309-4-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Make the type field in pmu-events/arch/s390/mapfile.cvs more generic to
match the created cpuid string for s390.
The pattern also checks for the counter first version number and counter
second version number ([13]\.[1-5]) and the authorization field which
follows.
These numbers do not exist in the cpuid identification string when perf
commands are executed on a z/VM environment (which does not support CPU
counter measurement facility).
CPUID string for LPAR:
cpuid : IBM,3906,704,M03,3.5,002f
CPUID string for z/VM:
cpuid : IBM,2964,702,N96
This allows the removal of s390 specific cpuid compare code and uses the
common compare function with its regular expression matching algorithm.
Signed-off-by: Thomas Richter <tmricht@linux.ibm.com>
Reviewed-by: Hendrik Brueckner <brueckner@linux.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Link: http://lkml.kernel.org/r/20180423081745.3672-1-tmricht@linux.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
map_groups__fixup_end() was called to set the end addresses of kernel
and module maps. But now since machine__create_modules() sets the end
address of modules properly, the only remaining piece is the kernel map.
We can set it with adjacent module's address directly instead of calling
map_groups__fixup_end(). If there's no module after the kernel map, the
end address will be ~0ULL.
Since it also changes the start address of the kernel map, it needs to
re-insert the map to the kmaps in order to keep a correct ordering. Kim
reported that it caused problems on ARM64.
Reported-by: Kim Phillips <kim.phillips@arm.com>
Tested-by: Kim Phillips <kim.phillips@arm.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: kernel-team@lge.com
Link: http://lkml.kernel.org/r/20180419235915.GA19067@sejong
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
When perf record encounters an error setting up an event it suggests
to enable CONFIG_PERF_EVENTS. This is misleading because:
- Usually it is enabled (it is really hard to disable on x86)
- The problem is usually somewhere else, e.g. the CPU is not supported
or an invalid configuration has been used.
Remove the misleading suggestion.
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Link: http://lkml.kernel.org/r/20180406203812.3087-4-andi@firstfloor.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
'perf list' with flags -d and -v print a description (-d) or a very
verbose explanation (-v) of CPU specific counter events. These
descriptions are provided with the json files in directory
pmu-events/arch/s390/*.json.
Display of these descriptions on s390 requires the corresponding json
files.
On s390 this does not work because function is_pmu_core() does not
detect the s390 directory name where the CPU specific events are listed.
On x86 it is:
/sys/bus/event_source/devices/cpu
whereas on s390 it is:
/sys/bus/event_source/devices/cpum_cf
/sys/bus/event_source/devices/cpum_sf
Fix this by adding s390 directory name testing to function
is_pmu_core(). This is the same approach as taken for the ARM platform.
Output before:
[root@s35lp76 perf]# ./perf list -d pmu
List of pre-defined events (to be used in -e):
cpum_cf/AES_BLOCKED_CYCLES/ [Kernel PMU event]
cpum_cf/AES_BLOCKED_FUNCTIONS/ [Kernel PMU event]
cpum_cf/AES_CYCLES/ [Kernel PMU event]
cpum_cf/AES_FUNCTIONS/ [Kernel PMU event]
....
cpum_cf/TX_NC_TEND/ [Kernel PMU event]
cpum_cf/VX_BCD_EXECUTION_SLOTS/ [Kernel PMU event]
cpum_sf/SF_CYCLES_BASIC/ [Kernel PMU event]
Output after:
[root@s35lp76 perf]# ./perf list -d pmu
List of pre-defined events (to be used in -e):
cpum_cf/AES_BLOCKED_CYCLES/ [Kernel PMU event]
cpum_cf/AES_BLOCKED_FUNCTIONS/ [Kernel PMU event]
cpum_cf/AES_CYCLES/ [Kernel PMU event]
cpum_cf/AES_FUNCTIONS/ [Kernel PMU event]
....
cpum_cf/TX_NC_TEND/ [Kernel PMU event]
cpum_cf/VX_BCD_EXECUTION_SLOTS/ [Kernel PMU event]
cpum_sf/SF_CYCLES_BASIC/ [Kernel PMU event]
3906:
bcd_dfp_execution_slots
[BCD DFP Execution Slots]
decimal_instructions
[Decimal Instructions]
dtlb2_gpage_writes
[DTLB2 GPAGE Writes]
dtlb2_hpage_writes
[DTLB2 HPAGE Writes]
dtlb2_misses
[DTLB2 Misses]
dtlb2_writes
[DTLB2 Writes]
itlb2_misses
[ITLB2 Misses]
itlb2_writes
[ITLB2 Writes]
l1c_tlb2_misses
[L1C TLB2 Misses]
.....
cfvn 3:
cpu_cycles
[CPU Cycles]
instructions
[Instructions]
l1d_dir_writes
[L1D Directory Writes]
l1d_penalty_cycles
[L1D Penalty Cycles]
l1i_dir_writes
[L1I Directory Writes]
l1i_penalty_cycles
[L1I Penalty Cycles]
problem_state_cpu_cycles
[Problem State CPU Cycles]
problem_state_instructions
[Problem State Instructions]
....
csvn generic:
aes_blocked_cycles
[AES Blocked Cycles]
aes_blocked_functions
[AES Blocked Functions]
aes_cycles
[AES Cycles]
aes_functions
[AES Functions]
dea_blocked_cycles
[DEA Blocked Cycles]
dea_blocked_functions
[DEA Blocked Functions]
....
Signed-off-by: Thomas Richter <tmricht@linux.vnet.ibm.com>
Reviewed-by: Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Link: http://lkml.kernel.org/r/20180416132314.33249-1-tmricht@linux.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
In the TUI the 's' hotkey can be used to switch to another perf.data
file in the current directory, but that got broken in Fixes:
b01141f4f5 ("perf annotate: Initialize the priv are in symbol__new()"),
that would show this once another file was chosen:
┌─Fatal Error─────────────────────────────────────┐
│Annotation needs to be init before symbol__init()│
│ │
│ │
│Press any key... │
└─────────────────────────────────────────────────┘
Fix it by just silently bailing out if symbol__annotation_init() was already
called, just like is done with symbol__init(), i.e. they are done just once at
session start, not when switching to a new perf.data file.
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Martin Liška <mliska@suse.cz>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Ravi Bangoria <ravi.bangoria@linux.vnet.ibm.com>
Cc: Thomas Richter <tmricht@linux.vnet.ibm.com>
Cc: Wang Nan <wangnan0@huawei.com>
Fixes: b01141f4f5 ("perf annotate: Initialize the priv are in symbol__new()")
Link: https://lkml.kernel.org/n/tip-ogppdtpzfax7y1h6gjdv5s6u@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Using perf on 4.16.0 kernel on s390 shows this warning:
failed: can't open node sysfs data
each time I run command perf record ... for example:
[root@s35lp76 perf]# ./perf record -e rB0000 -- sleep 1
[ perf record: Woken up 1 times to write data ]
failed: can't open node sysfs data
[ perf record: Captured and wrote 0.001 MB perf.data (4 samples) ]
[root@s35lp76 perf]#
It turns out commit e2091cedd5 ("perf tools: Add MEM_TOPOLOGY feature
to perf data file") tries to open directory named /sys/devices/system/node/
which does not exist on s390.
This is the call stack:
__cmd_record
+---> perf_session__write_header
+---> perf_header__adds_write
+---> do_write_feat
+---> write_mem_topology
+---> build_mem_topology
prints warning
The issue starts in do_write_feat() which unconditionally loops over all
features and now includes HEADER_MEM_TOPOLOGY and calls write_mem_topology().
Function record__init_features() at the beginning of __cmd_record() sets
all features and then turns off some of them.
Fix this by changing the warning to a level 2 debug output statement.
So it is only shown when debug level 2 or higher is set.
Signed-off-by: Thomas Richter <tmricht@linux.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Link: http://lkml.kernel.org/r/20180412133246.92801-1-tmricht@linux.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
As stated in tests/llvm-src-base.c, the name of the bpf function should
be "bpf_func__SyS_epoll_pwait" but this clang test fails as it tries to
lookup "bpf_func__SyS_epoll_wait".
Before applying patch:
55: builtin clang support :
55.1: builtin clang compile C source to IR : FAILED!
55.2: builtin clang compile C source to ELF object : Skip
After applying patch:
55: builtin clang support :
55.1: builtin clang compile C source to IR : Ok
55.2: builtin clang compile C source to ELF object : Ok
Signed-off-by: Sandipan Das <sandipan@linux.vnet.ibm.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Fixes: e67d52d411 ("perf clang: Update test case to use real BPF script")
Link: http://lkml.kernel.org/r/20180404180419.19056-3-sandipan@linux.vnet.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The clang API calls used by perf have changed in recent releases and
builds succeed with libclang-3.9 only. This introduces compatibility
with libclang-4.0 and above.
Without this patch, we will see the following compilation errors with
libclang-4.0+:
util/c++/clang.cpp: In function ‘clang::CompilerInvocation* perf::createCompilerInvocation(llvm::opt::ArgStringList, llvm::StringRef&, clang::DiagnosticsEngine&)’:
util/c++/clang.cpp:62:33: error: ‘IK_C’ was not declared in this scope
Opts.Inputs.emplace_back(Path, IK_C);
^~~~
util/c++/clang.cpp: In function ‘std::unique_ptr<llvm::Module> perf::getModuleFromSource(llvm::opt::ArgStringList, llvm::StringRef, llvm::IntrusiveRefCntPtr<clang::vfs::FileSystem>)’:
util/c++/clang.cpp:75:26: error: no matching function for call to ‘clang::CompilerInstance::setInvocation(clang::CompilerInvocation*)’
Clang.setInvocation(&*CI);
^
In file included from util/c++/clang.cpp:14:0:
/usr/include/clang/Frontend/CompilerInstance.h:231:8: note: candidate: void clang::CompilerInstance::setInvocation(std::shared_ptr<clang::CompilerInvocation>)
void setInvocation(std::shared_ptr<CompilerInvocation> Value);
^~~~~~~~~~~~~
Committer testing:
Tested on Fedora 27 after installing the clang-devel and llvm-devel
packages, versions:
# rpm -qa | egrep llvm\|clang
llvm-5.0.1-6.fc27.x86_64
clang-libs-5.0.1-5.fc27.x86_64
clang-5.0.1-5.fc27.x86_64
clang-tools-extra-5.0.1-5.fc27.x86_64
llvm-libs-5.0.1-6.fc27.x86_64
llvm-devel-5.0.1-6.fc27.x86_64
clang-devel-5.0.1-5.fc27.x86_64
#
Make sure you don't have some older version lying around in /usr/local,
etc, then:
$ make LIBCLANGLLVM=1 -C tools/perf install-bin
And in the end perf will be linked agains these libraries:
# ldd ~/bin/perf | egrep -i llvm\|clang
libclangAST.so.5 => /lib64/libclangAST.so.5 (0x00007f8bb2eb4000)
libclangBasic.so.5 => /lib64/libclangBasic.so.5 (0x00007f8bb29e3000)
libclangCodeGen.so.5 => /lib64/libclangCodeGen.so.5 (0x00007f8bb23f7000)
libclangDriver.so.5 => /lib64/libclangDriver.so.5 (0x00007f8bb2060000)
libclangFrontend.so.5 => /lib64/libclangFrontend.so.5 (0x00007f8bb1d06000)
libclangLex.so.5 => /lib64/libclangLex.so.5 (0x00007f8bb1a3e000)
libclangTooling.so.5 => /lib64/libclangTooling.so.5 (0x00007f8bb17d4000)
libclangEdit.so.5 => /lib64/libclangEdit.so.5 (0x00007f8bb15c5000)
libclangSema.so.5 => /lib64/libclangSema.so.5 (0x00007f8bb0cc9000)
libclangAnalysis.so.5 => /lib64/libclangAnalysis.so.5 (0x00007f8bb0a23000)
libclangParse.so.5 => /lib64/libclangParse.so.5 (0x00007f8bb0725000)
libclangSerialization.so.5 => /lib64/libclangSerialization.so.5 (0x00007f8bb039a000)
libLLVM-5.0.so => /lib64/libLLVM-5.0.so (0x00007f8bace98000)
libclangASTMatchers.so.5 => /lib64/../lib64/libclangASTMatchers.so.5 (0x00007f8bab735000)
libclangFormat.so.5 => /lib64/../lib64/libclangFormat.so.5 (0x00007f8bab4b2000)
libclangRewrite.so.5 => /lib64/../lib64/libclangRewrite.so.5 (0x00007f8bab2a1000)
libclangToolingCore.so.5 => /lib64/../lib64/libclangToolingCore.so.5 (0x00007f8bab08e000)
#
Signed-off-by: Sandipan Das <sandipan@linux.vnet.ibm.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Fixes: 00b86691c7 ("perf clang: Add builtin clang support ant test case")
Link: http://lkml.kernel.org/r/20180404180419.19056-2-sandipan@linux.vnet.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
To match what is shown in the main 'perf report/top' title lines, i.e.
if a group is being shown, either a real group (recorded with "-e
'{a,b,c}') or a forced group (using 'perf report --group' for a
perf.data file recorded without {}) we will show multiple columns,
one per event, but we were failing to show the group details, so, for:
# perf report --header-only | grep cmdline
# cmdline : /home/acme/bin/perf record -e {cycles,instructions,cache-misses}
# perf report --group
The first line was showing just "cycles", now it shows the correct line,
which is:
Samples: 578 of events 'anon group { cycles, instructions, cache-misses }', 4000 Hz, Event count (approx.): 487421794
syscall_return_via_sysret /lib/modules/4.16.0-rc7/build/vmlinux
0.22 2.97 0.00 │ ↓ jmp 6c
│ mov %cr3,%rdi
1.33 10.89 4.00 │ ↓ jmp 62
│ mov %rdi,%rax
<SNIP>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Fixes: 6920e2854e ("perf annotate browser: Show extra title line with event information")
Link: https://lkml.kernel.org/n/tip-i41tqh17c2dabnyzjh99r1oz@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
These types of jumps were confusing the annotate browser:
entry_SYSCALL_64 /lib/modules/4.16.0-rc5-00086-gdf09348f78dc/build/vmlinux
entry_SYSCALL_64 /lib/modules/4.16.0-rc5-00086-gdf09348f78dc/build/vmlinux
Percent│ffffffff81a00020: swapgs
<SNIP>
│ffffffff81a00128: ↓ jae ffffffff81a00139 <syscall_return_via_sysret+0x53>
<SNIP>
│ffffffff81a00155: → jmpq *0x825d2d(%rip) # ffffffff82225e88 <pv_cpu_ops+0xe8>
I.e. the syscall_return_via_sysret function is actually "inside" the
entry_SYSCALL_64 function, and the offsets in jumps like these (+0x53)
are relative to syscall_return_via_sysret, not to syscall_return_via_sysret.
Or this may be some artifact in how the assembler marks the start and
end of a function and how this ends up in the ELF symtab for vmlinux,
i.e. syscall_return_via_sysret() isn't "inside" entry_SYSCALL_64, but
just right after it.
From readelf -sw vmlinux:
80267: ffffffff81a00020 315 NOTYPE GLOBAL DEFAULT 1 entry_SYSCALL_64
316: ffffffff81a000e6 0 NOTYPE LOCAL DEFAULT 1 syscall_return_via_sysret
0xffffffff81a00020 + 315 > 0xffffffff81a000e6
So instead of looking for offsets after that last '+' sign, calculate
offsets for jump target addresses that are inside the function being
disassembled from the absolute address, 0xffffffff81a00139 in this case,
subtracting from it the objdump address for the start of the function
being disassembled, entry_SYSCALL_64() in this case.
So, before this patch:
entry_SYSCALL_64 /lib/modules/4.16.0-rc5-00086-gdf09348f78dc/build/vmlinux
Percent│ pop %r10
│ pop %r9
│ pop %r8
│ pop %rax
│ pop %rsi
│ pop %rdx
│ pop %rsi
│ mov %rsp,%rdi
│ mov %gs:0x5004,%rsp
│ pushq 0x28(%rdi)
│ pushq (%rdi)
│ push %rax
│ ↑ jmp 6c
│ mov %cr3,%rdi
│ ↑ jmp 62
│ mov %rdi,%rax
│ and $0x7ff,%rdi
│ bt %rdi,%gs:0x2219a
│ ↑ jae 53
│ btr %rdi,%gs:0x2219a
│ mov %rax,%rdi
│ ↑ jmp 5b
After:
entry_SYSCALL_64 /lib/modules/4.16.0-rc5-00086-gdf09348f78dc/build/vmlinux
0.65 │ → jne swapgs_restore_regs_and_return_to_usermode
│ pop %r10
│ pop %r9
│ pop %r8
│ pop %rax
│ pop %rsi
│ pop %rdx
│ pop %rsi
│ mov %rsp,%rdi
│ mov %gs:0x5004,%rsp
│ pushq 0x28(%rdi)
│ pushq (%rdi)
│ push %rax
│ ↓ jmp 132
│ mov %cr3,%rdi
│ ┌──jmp 128
│ │ mov %rdi,%rax
│ │ and $0x7ff,%rdi
│ │ bt %rdi,%gs:0x2219a
│ │↓ jae 119
│ │ btr %rdi,%gs:0x2219a
│ │ mov %rax,%rdi
│ │↓ jmp 121
│119:│ mov %rax,%rdi
│ │ bts $0x3f,%rdi
│121:│ or $0x800,%rdi
│128:└─→or $0x1000,%rdi
│ mov %rdi,%cr3
│132: pop %rax
│ pop %rdi
│ pop %rsp
│ → jmpq *0x825d2d(%rip) # ffffffff82225e88 <pv_cpu_ops+0xe8>
With those at least navigating to the right destination, an improvement
for these cases seems to be to be to somehow mark those inner functions,
which in this case could be:
entry_SYSCALL_64 /lib/modules/4.16.0-rc5-00086-gdf09348f78dc/build/vmlinux
│syscall_return_via_sysret:
│ pop %r15
│ pop %r14
│ pop %r13
│ pop %r12
│ pop %rbp
│ pop %rbx
│ pop %rsi
│ pop %r10
│ pop %r9
│ pop %r8
│ pop %rax
│ pop %rsi
│ pop %rdx
│ pop %rsi
│ mov %rsp,%rdi
│ mov %gs:0x5004,%rsp
│ pushq 0x28(%rdi)
│ pushq (%rdi)
│ push %rax
│ ↓ jmp 132
│ mov %cr3,%rdi
│ ┌──jmp 128
│ │ mov %rdi,%rax
│ │ and $0x7ff,%rdi
│ │ bt %rdi,%gs:0x2219a
│ │↓ jae 119
│ │ btr %rdi,%gs:0x2219a
│ │ mov %rax,%rdi
│ │↓ jmp 121
│119:│ mov %rax,%rdi
│ │ bts $0x3f,%rdi
│121:│ or $0x800,%rdi
│128:└─→or $0x1000,%rdi
│ mov %rdi,%cr3
│132: pop %rax
│ pop %rdi
│ pop %rsp
│ → jmpq *0x825d2d(%rip) # ffffffff82225e88 <pv_cpu_ops+0xe8>
This all gets much better viewed if one uses 'perf report --ignore-vmlinux'
forcing the usage of /proc/kcore + /proc/kallsyms, when the above
actually gets down to:
# perf report --ignore-vmlinux
## do '/64', will show the function names containing '64',
## navigate to /entry_SYSCALL_64_after_hwframe.annotation,
## press 'A' to annotate, then 'P' to print that annotation
## to a file
## From another xterm (or see on screen, this 'P' thing is for
## getting rid of those right side scroll bars/spaces):
# cat /entry_SYSCALL_64_after_hwframe.annotation
entry_SYSCALL_64_after_hwframe() /proc/kcore
Event: cycles:ppp
Percent
Disassembly of section load0:
ffffffff9aa00044 <load0>:
11.97 push %rax
4.85 push %rdi
push %rsi
2.59 push %rdx
2.27 push %rcx
0.32 pushq $0xffffffffffffffda
1.29 push %r8
xor %r8d,%r8d
1.62 push %r9
0.65 xor %r9d,%r9d
1.62 push %r10
xor %r10d,%r10d
5.50 push %r11
xor %r11d,%r11d
3.56 push %rbx
xor %ebx,%ebx
4.21 push %rbp
xor %ebp,%ebp
2.59 push %r12
0.97 xor %r12d,%r12d
3.24 push %r13
xor %r13d,%r13d
2.27 push %r14
xor %r14d,%r14d
4.21 push %r15
xor %r15d,%r15d
0.97 mov %rsp,%rdi
5.50 → callq do_syscall_64
14.56 mov 0x58(%rsp),%rcx
7.44 mov 0x80(%rsp),%r11
0.32 cmp %rcx,%r11
→ jne swapgs_restore_regs_and_return_to_usermode
0.32 shl $0x10,%rcx
0.32 sar $0x10,%rcx
3.24 cmp %rcx,%r11
→ jne swapgs_restore_regs_and_return_to_usermode
2.27 cmpq $0x33,0x88(%rsp)
1.29 → jne swapgs_restore_regs_and_return_to_usermode
mov 0x30(%rsp),%r11
8.74 cmp %r11,0x90(%rsp)
→ jne swapgs_restore_regs_and_return_to_usermode
0.32 test $0x10100,%r11
→ jne swapgs_restore_regs_and_return_to_usermode
0.32 cmpq $0x2b,0xa0(%rsp)
0.65 → jne swapgs_restore_regs_and_return_to_usermode
I.e. using kallsyms makes the function start/end be done differently
than using what is in the vmlinux ELF symtab and actually the hits
goes to entry_SYSCALL_64_after_hwframe, which is a GLOBAL() after the
start of entry_SYSCALL_64:
ENTRY(entry_SYSCALL_64)
UNWIND_HINT_EMPTY
<SNIP>
pushq $__USER_CS /* pt_regs->cs */
pushq %rcx /* pt_regs->ip */
GLOBAL(entry_SYSCALL_64_after_hwframe)
pushq %rax /* pt_regs->orig_ax */
PUSH_AND_CLEAR_REGS rax=$-ENOSYS
And it goes and ends at:
cmpq $__USER_DS, SS(%rsp) /* SS must match SYSRET */
jne swapgs_restore_regs_and_return_to_usermode
/*
* We win! This label is here just for ease of understanding
* perf profiles. Nothing jumps here.
*/
syscall_return_via_sysret:
/* rcx and r11 are already restored (see code above) */
UNWIND_HINT_EMPTY
POP_REGS pop_rdi=0 skip_r11rcx=1
So perhaps some people should really just play with '--ignore-vmlinux'
to force /proc/kcore + kallsyms.
One idea is to do both, i.e. have a vmlinux annotation and a
kcore+kallsyms one, when possible, and even show the patched location,
etc.
Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: https://lkml.kernel.org/n/tip-r11knxv8voesav31xokjiuo6@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
For instance:
entry_SYSCALL_64 /lib/modules/4.16.0-rc5-00086-gdf09348f78dc/build/vmlinux
5.50 │ → callq do_syscall_64
14.56 │ mov 0x58(%rsp),%rcx
7.44 │ mov 0x80(%rsp),%r11
0.32 │ cmp %rcx,%r11
│ → jne swapgs_restore_regs_and_return_to_usermode
0.32 │ shl $0x10,%rcx
0.32 │ sar $0x10,%rcx
3.24 │ cmp %rcx,%r11
│ → jne swapgs_restore_regs_and_return_to_usermode
2.27 │ cmpq $0x33,0x88(%rsp)
1.29 │ → jne swapgs_restore_regs_and_return_to_usermode
│ mov 0x30(%rsp),%r11
8.74 │ cmp %r11,0x90(%rsp)
│ → jne swapgs_restore_regs_and_return_to_usermode
0.32 │ test $0x10100,%r11
│ → jne swapgs_restore_regs_and_return_to_usermode
0.32 │ cmpq $0x2b,0xa0(%rsp)
0.65 │ → jne swapgs_restore_regs_and_return_to_usermode
It'll behave just like a "call" instruction, i.e. press enter or right
arrow over one such line and the browser will navigate to the annotated
disassembly of that function, which when exited, via left arrow or esc,
will come back to the calling function.
Now to support jump to an offset on a different function...
Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: https://lkml.kernel.org/n/tip-78o508mqvr8inhj63ddtw7mo@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Because they all really check if we can access data structures/visual
constructs where a "jump" instruction targets code in the same function,
i.e. things like:
__pthread_mutex_lock /usr/lib64/libpthread-2.26.so
1.95 │ mov __pthread_force_elision,%ecx
│ ┌──test %ecx,%ecx
0.07 │ ├──je 60
│ │ test $0x300,%esi
│ │↓ jne 60
│ │ or $0x100,%esi
│ │ mov %esi,0x10(%rdi)
│ 42:│ mov %esi,%edx
│ │ lea 0x16(%r8),%rsi
│ │ mov %r8,%rdi
│ │ and $0x80,%edx
│ │ add $0x8,%rsp
│ │→ jmpq __lll_lock_elision
│ │ nop
0.29 │ 60:└─→and $0x80,%esi
0.07 │ mov $0x1,%edi
0.29 │ xor %eax,%eax
2.53 │ lock cmpxchg %edi,(%r8)
And not things like that "jmpq __lll_lock_elision", that instead should behave
like a "call" instruction and "jump" to the disassembly of "___lll_lock_elision".
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: https://lkml.kernel.org/n/tip-3cwx39u3h66dfw9xjrlt7ca2@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Things like this in _cpp_lex_token (gcc's cc1 program):
cpp_named_operator2name@@Base+0xa72
Point to a place that is after the cpp_named_operator2name boundaries,
i.e. in the ELF symbol table for cc1 cpp_named_operator2name is marked
as being 32-bytes long, but it in fact is much larger than that, so we
seem to need a symbols__find() routine that looks for >= current->start
and < next_symbol->start, possibly just for C++ objects?
For now lets just make some progress by marking jumps to outside the
current function as call like.
Actual navigation will come next, with further understanding of how the
symbol searching and disassembly should be done.
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: https://lkml.kernel.org/n/tip-aiys0a0bsgm3e00hbi6fg7yy@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Just like we have in the histograms browser used as the main screen for
'perf top --tui' and 'perf report --tui', to print the current
annotation to a file with a named composed by the symbol name and the
".annotation" suffix.
Here is one example of pressing 'A' on 'perf top' to live annotate a
kernel function and then press 'P' to dump that annotation, the
resulting file:
# cat _raw_spin_lock_irqsave.annotation
_raw_spin_lock_irqsave() /proc/kcore
Event: cycles:ppp
7.14 nop
21.43 push %rbx
7.14 pushfq
pop %rax
nop
mov %rax,%rbx
cli
nop
xor %eax,%eax
mov $0x1,%edx
64.29 lock cmpxchg %edx,(%rdi)
test %eax,%eax
↓ jne 2b
mov %rbx,%rax
pop %rbx
← retq
2b: mov %eax,%esi
→ callq queued_spin_lock_slowpath
mov %rbx,%rax
pop %rbx
← retq
#
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: https://lkml.kernel.org/n/tip-zzmnrwugb5vtk7bvg0rbx150@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Kan Liang