Avi Kivity reported a problem where the __weak
btf__load_from_kernel_by_id() in tools/perf/util/bpf-event.c was being
used and it called btf__get_from_id() in tools/lib/bpf/btf.c that in
turn called back to btf__load_from_kernel_by_id(), resulting in an
endless loop.
Fix this by adding a feature test to check if
btf__load_from_kernel_by_id() is available when building perf with
LIBBPF_DYNAMIC=1, and if not then provide the fallback to the old
btf__get_from_id(), that doesn't call back to btf__load_from_kernel_by_id()
since at that time it didn't exist at all.
Tested on Fedora 35 where we have libbpf-devel 0.4.0 with LIBBPF_DYNAMIC
where we don't have btf__load_from_kernel_by_id() and thus its feature
test fail, not defining HAVE_LIBBPF_BTF__LOAD_FROM_KERNEL_BY_ID:
$ cat /tmp/build/perf-urgent/feature/test-libbpf-btf__load_from_kernel_by_id.make.output
test-libbpf-btf__load_from_kernel_by_id.c: In function ‘main’:
test-libbpf-btf__load_from_kernel_by_id.c:6:16: error: implicit declaration of function ‘btf__load_from_kernel_by_id’ [-Werror=implicit-function-declaration]
6 | return btf__load_from_kernel_by_id(20151128, NULL);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~
cc1: all warnings being treated as errors
$
$ nm /tmp/build/perf-urgent/perf | grep btf__load_from_kernel_by_id
00000000005ba180 T btf__load_from_kernel_by_id
$
$ objdump --disassemble=btf__load_from_kernel_by_id -S /tmp/build/perf-urgent/perf
/tmp/build/perf-urgent/perf: file format elf64-x86-64
<SNIP>
00000000005ba180 <btf__load_from_kernel_by_id>:
#include "record.h"
#include "util/synthetic-events.h"
#ifndef HAVE_LIBBPF_BTF__LOAD_FROM_KERNEL_BY_ID
struct btf *btf__load_from_kernel_by_id(__u32 id)
{
5ba180: 55 push %rbp
5ba181: 48 89 e5 mov %rsp,%rbp
5ba184: 48 83 ec 10 sub $0x10,%rsp
5ba188: 64 48 8b 04 25 28 00 mov %fs:0x28,%rax
5ba18f: 00 00
5ba191: 48 89 45 f8 mov %rax,-0x8(%rbp)
5ba195: 31 c0 xor %eax,%eax
struct btf *btf;
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wdeprecated-declarations"
int err = btf__get_from_id(id, &btf);
5ba197: 48 8d 75 f0 lea -0x10(%rbp),%rsi
5ba19b: e8 a0 57 e5 ff call 40f940 <btf__get_from_id@plt>
5ba1a0: 89 c2 mov %eax,%edx
#pragma GCC diagnostic pop
return err ? ERR_PTR(err) : btf;
5ba1a2: 48 98 cltq
5ba1a4: 85 d2 test %edx,%edx
5ba1a6: 48 0f 44 45 f0 cmove -0x10(%rbp),%rax
}
<SNIP>
Fixes: 218e7b775d ("perf bpf: Provide a weak btf__load_from_kernel_by_id() for older libbpf versions")
Reported-by: Avi Kivity <avi@scylladb.com>
Link: https://lore.kernel.org/linux-perf-users/f0add43b-3de5-20c5-22c4-70aff4af959f@scylladb.com
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/linux-perf-users/YobjjFOblY4Xvwo7@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
ptwrite is an Intel x86 instruction that writes arbitrary values into an
Intel PT trace. It is not supported on all hardware, so provide an
alternative that makes use of TNT packets to convey the payload data.
TNT packets encode Taken/Not-taken conditional branch information, so
taking branches based on the payload value will encode the value into
the TNT packet. Refer to the changes to the documentation file
perf-intel-pt.txt in this patch for an example.
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Link: https://lore.kernel.org/r/20220509152400.376613-2-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Since for cpu_core or cpu_atom, they have different topdown events
groups.
For cpu_core, --topdown equals to:
"{slots,cpu_core/topdown-retiring/,cpu_core/topdown-bad-spec/,
cpu_core/topdown-fe-bound/,cpu_core/topdown-be-bound/,
cpu_core/topdown-heavy-ops/,cpu_core/topdown-br-mispredict/,
cpu_core/topdown-fetch-lat/,cpu_core/topdown-mem-bound/}"
For cpu_atom, --topdown equals to:
"{cpu_atom/topdown-retiring/,cpu_atom/topdown-bad-spec/,
cpu_atom/topdown-fe-bound/,cpu_atom/topdown-be-bound/}"
To simplify the implementation, on hybrid, --topdown is used
together with --cputype. If without --cputype, it uses cpu_core
topdown events by default.
# ./perf stat --topdown -a sleep 1
WARNING: default to use cpu_core topdown events
Performance counter stats for 'system wide':
retiring bad speculation frontend bound backend bound heavy operations light operations branch mispredict machine clears fetch latency fetch bandwidth memory bound Core bound
4.1% 0.0% 5.1% 90.8% 2.3% 1.8% 0.0% 0.0% 4.2% 0.9% 9.9% 81.0%
1.002624229 seconds time elapsed
# ./perf stat --topdown -a --cputype atom sleep 1
Performance counter stats for 'system wide':
retiring bad speculation frontend bound backend bound
13.5% 0.1% 31.2% 55.2%
1.002366987 seconds time elapsed
Reviewed-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Xing Zhengjun <zhengjun.xing@linux.intel.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://lore.kernel.org/r/20220422065635.767648-3-zhengjun.xing@linux.intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The header TargetRegistry.h has moved in LLVM/clang 14.
Committer notes:
The problem as noticed when building in ubuntu:22.04:
90 98.61 ubuntu:22.04 : FAIL gcc version 11.2.0 (Ubuntu 11.2.0-19ubuntu1)
util/c++/clang.cpp:23:10: fatal error: llvm/Support/TargetRegistry.h: No such file or directory
23 | #include "llvm/Support/TargetRegistry.h"
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
compilation terminated.
Fixed after applying this patch.
Reported-by: Arnaldo Carvalho de Melo <acme@kernel.org>
Signed-off-by: Guilherme Amadio <amadio@gentoo.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Link: https://twitter.com/GuilhermeAmadio/status/1514970524232921088
Link: http://lore.kernel.org/lkml/Ylp0M/VYgHOxtcnF@gentoo.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
One metric such as 'Kernel_Utilization' may be from different PMUs and
consists of different events.
For core,
Kernel_Utilization = cpu_clk_unhalted.thread:k / cpu_clk_unhalted.thread
For atom,
Kernel_Utilization = cpu_clk_unhalted.core:k / cpu_clk_unhalted.core
The metric group string for core is:
'{cpu_clk_unhalted.thread/metric-id=cpu_clk_unhalted.thread:k/k,cpu_clk_unhalted.thread/metric-id=cpu_clk_unhalted.thread/}:W'
It's internally expanded to:
'{cpu_clk_unhalted.thread_p/metric-id=cpu_clk_unhalted.thread_p:k/k,cpu_clk_unhalted.thread/metric-id=cpu_clk_unhalted.thread/}:W#cpu_core'
The metric group string for atom is:
'{cpu_clk_unhalted.core/metric-id=cpu_clk_unhalted.core:k/k,cpu_clk_unhalted.core/metric-id=cpu_clk_unhalted.core/}:W'
It's internally expanded to:
'{cpu_clk_unhalted.core/metric-id=cpu_clk_unhalted.core:k/k,cpu_clk_unhalted.core/metric-id=cpu_clk_unhalted.core/}:W#cpu_atom'
That means the group "{cpu_clk_unhalted.thread:k,cpu_clk_unhalted.thread}:W"
is from cpu_core PMU and the group "{cpu_clk_unhalted.core:k,cpu_clk_unhalted.core}"
is from cpu_atom PMU. And then next, check if the events in the group are
valid on that PMU. If one event is not valid on that PMU, the associated
group would be removed internally.
In this example, cpu_clk_unhalted.thread is valid on cpu_core and
cpu_clk_unhalted.core is valid on cpu_atom. So the checks for these two
groups are passed.
Before:
# ./perf stat -M Kernel_Utilization -a sleep 1
WARNING: events in group from different hybrid PMUs!
WARNING: grouped events cpus do not match, disabling group:
anon group { CPU_CLK_UNHALTED.THREAD_P:k, CPU_CLK_UNHALTED.THREAD_P:k, CPU_CLK_UNHALTED.THREAD, CPU_CLK_UNHALTED.THREAD }
Performance counter stats for 'system wide':
17,639,501 cpu_atom/CPU_CLK_UNHALTED.CORE/ # 1.00 Kernel_Utilization
17,578,757 cpu_atom/CPU_CLK_UNHALTED.CORE:k/
1,005,350,226 ns duration_time
43,012,352 cpu_core/CPU_CLK_UNHALTED.THREAD_P:k/ # 0.99 Kernel_Utilization
17,608,010 cpu_atom/CPU_CLK_UNHALTED.THREAD_P:k/
43,608,755 cpu_core/CPU_CLK_UNHALTED.THREAD/
17,630,838 cpu_atom/CPU_CLK_UNHALTED.THREAD/
1,005,350,226 ns duration_time
1.005350226 seconds time elapsed
After:
# ./perf stat -M Kernel_Utilization -a sleep 1
Performance counter stats for 'system wide':
17,981,895 CPU_CLK_UNHALTED.CORE [cpu_atom] # 1.00 Kernel_Utilization
17,925,405 CPU_CLK_UNHALTED.CORE:k [cpu_atom]
1,004,811,366 ns duration_time
41,246,425 CPU_CLK_UNHALTED.THREAD_P:k [cpu_core] # 0.99 Kernel_Utilization
41,819,129 CPU_CLK_UNHALTED.THREAD [cpu_core]
1,004,811,366 ns duration_time
1.004811366 seconds time elapsed
Reviewed-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Xing Zhengjun <zhengjun.xing@linux.intel.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://lore.kernel.org/r/20220422065635.767648-1-zhengjun.xing@linux.intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
It bothered me that during benchmarking using 'perf stat' (to collect
for example CPU cache events) I could not simultaneously retrieve the
times spend in user or kernel mode in a machine readable format.
When running 'perf stat' the output for humans contains the times
reported by rusage and wait4.
$ perf stat -e cache-misses:u -- true
Performance counter stats for 'true':
4,206 cache-misses:u
0.001113619 seconds time elapsed
0.001175000 seconds user
0.000000000 seconds sys
But 'perf stat's machine-readable format does not provide this information.
$ perf stat -x, -e cache-misses:u -- true
4282,,cache-misses:u,492859,100.00,,
I found no way to retrieve this information using the available events
while using machine-readable output.
This patch adds two new tool internal events 'user_time' and
'system_time', similarly to the already present 'duration_time' event.
Both events use the already collected rusage information obtained by
wait4 and tracked in the global ru_stats.
Examples presenting cache-misses and rusage information in both human
and machine-readable form:
$ perf stat -e duration_time,user_time,system_time,cache-misses -- grep -q -r duration_time .
Performance counter stats for 'grep -q -r duration_time .':
67,422,542 ns duration_time:u
50,517,000 ns user_time:u
16,839,000 ns system_time:u
30,937 cache-misses:u
0.067422542 seconds time elapsed
0.050517000 seconds user
0.016839000 seconds sys
$ perf stat -x, -e duration_time,user_time,system_time,cache-misses -- grep -q -r duration_time .
72134524,ns,duration_time:u,72134524,100.00,,
65225000,ns,user_time:u,65225000,100.00,,
6865000,ns,system_time:u,6865000,100.00,,
38705,,cache-misses:u,71189328,100.00,,
Signed-off-by: Florian Fischer <florian.fischer@muhq.space>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Xing Zhengjun <zhengjun.xing@linux.intel.com>
Link: https://lore.kernel.org/r/20220420102354.468173-3-florian.fischer@muhq.space
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Perf numa bench test fails with error:
Testcase:
./perf bench numa mem -p 2 -t 1 -P 1024 -C 0,8 -M 1,0 -s 20 -zZq --thp 1 --no-data_rand_walk
Failure snippet:
<<>>
Running 'numa/mem' benchmark:
# Running main, "perf bench numa numa-mem -p 2 -t 1 -P 1024 -C 0,8 -M 1,0 -s 20 -zZq --thp 1 --no-data_rand_walk"
perf: bench/numa.c:333: bind_to_cpumask: Assertion `!(ret)' failed.
<<>>
The Testcases uses CPU's 0 and 8. In function "parse_setup_cpu_list",
There is check to see if cpu number is greater than max cpu's possible
in the system ie via "if (bind_cpu_0 >= g->p.nr_cpus || bind_cpu_1 >=
g->p.nr_cpus) {".
But it could happen that system has say 48 CPU's, but only number of
online CPU's is 0-7. Other CPU's are offlined. Since "g->p.nr_cpus" is
48, so function will go ahead and set bit for CPU 8 also in cpumask (
td->bind_cpumask).
bind_to_cpumask function is called to set affinity using
sched_setaffinity and the cpumask. Since the CPU8 is not present, set
affinity will fail here with EINVAL.
Fix this issue by adding a check to make sure that, CPU's provided in
the input argument values are online before proceeding further and skip
the test. For this, include new helper function "is_cpu_online" in
"tools/perf/util/header.c".
Since "BIT(x)" definition will get included from header.h, remove
that from bench/numa.c
Reported-by: Disha Goel <disgoel@linux.vnet.ibm.com>
Signed-off-by: Athira Jajeev <atrajeev@linux.vnet.ibm.com>
Tested-by: Disha Goel <disgoel@linux.vnet.ibm.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Kajol Jain <kjain@linux.ibm.com>
Cc: Madhavan Srinivasan <maddy@linux.vnet.ibm.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Nageswara R Sastry <rnsastry@linux.ibm.com>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Cc: linuxppc-dev@lists.ozlabs.org
Link: https://lore.kernel.org/r/20220412164059.42654-2-atrajeev@linux.vnet.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Fix incorrect debug message:
Attempting to add event pmu 'intel_pt' with '' that may result in
non-fatal errors
which always appears with perf record -vv and intel_pt e.g.
perf record -vv -e intel_pt//u uname
The message is incorrect because there will never be non-fatal errors.
Suppress the message if the PMU is 'selectable' i.e. meant to be
selected directly as an event.
Fixes: 4ac22b484d ("perf parse-events: Make add PMU verbose output clearer")
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Link: http://lore.kernel.org/lkml/20220411061758.2458417-1-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Commit Fixes: b9f6fbb3b2 ("perf arm64: Inject missing frames when
using 'perf record --call-graph=fp'") intended to add a 'best effort'
DWARF unwind that improved the frame pointer stack in most scenarios.
It's expected that the unwind will fail sometimes, but this shouldn't be
reported as an error. It only works when the return address can be
determined from the contents of the link register alone.
Fix the error shown when the unwinder requires extra registers by adding
a new flag that suppresses error messages. This flag is not set in the
normal --call-graph=dwarf unwind mode so that behavior is not changed.
Fixes: b9f6fbb3b2 ("perf arm64: Inject missing frames when using 'perf record --call-graph=fp'")
Reported-by: John Garry <john.garry@huawei.com>
Signed-off-by: James Clark <james.clark@arm.com>
Tested-by: John Garry <john.garry@huawei.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexandre Truong <alexandre.truong@arm.com>
Cc: German Gomez <german.gomez@arm.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/r/20220406145651.1392529-1-james.clark@arm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Using -ffat-lto-objects in the python feature test when building with
clang-13 results in:
clang-13: error: optimization flag '-ffat-lto-objects' is not supported [-Werror,-Wignored-optimization-argument]
error: command '/usr/sbin/clang' failed with exit code 1
cp: cannot stat '/tmp/build/perf/python_ext_build/lib/perf*.so': No such file or directory
make[2]: *** [Makefile.perf:639: /tmp/build/perf/python/perf.so] Error 1
Noticed when building on a docker.io/library/archlinux:base container.
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Fangrui Song <maskray@google.com>
Cc: Florian Fainelli <f.fainelli@gmail.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: John Keeping <john@metanate.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Nathan Chancellor <nathan@kernel.org>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The clang compiler complains about some options even without a source
file being available, while others require one, so use the simple
tools/build/feature/test-hello.c file.
Then check for the "is not supported" string in its output, in addition
to the "unknown argument" already being looked for.
This was noticed when building with clang-13 where -ffat-lto-objects
isn't supported and since we were looking just for "unknown argument"
and not providing a source code to clang, was mistakenly assumed as
being available and not being filtered to set of command line options
provided to clang, leading to a build failure.
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Fangrui Song <maskray@google.com>
Cc: Florian Fainelli <f.fainelli@gmail.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: John Keeping <john@metanate.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Nathan Chancellor <nathan@kernel.org>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Sedat Dilek <sedat.dilek@gmail.com>
Link: http://lore.kernel.org/lkml/
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
To pick the changes in:
fba60b171a ("libbpf: Use IS_ERR_OR_NULL() in hashmap__free()")
That don't entail any changes in tools/perf.
This addresses this perf build warning:
Warning: Kernel ABI header at 'tools/perf/util/hashmap.h' differs from latest version at 'tools/lib/bpf/hashmap.h'
diff -u tools/perf/util/hashmap.h tools/lib/bpf/hashmap.h
Not a kernel ABI, its just that this uses the mechanism in place for
checking kernel ABI files drift.
Cc: Andrii Nakryiko <andrii@kernel.org>
Cc: Ian Rogers <irogers@google.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Mauricio Vásquez <mauricio@kinvolk.io>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: http://lore.kernel.org/lkml/YkMb2SAIai2VeuUD@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
