[ Upstream commit 784e8adda4cdb3e2510742023729851b6c08803c ]
Currently, the 'weight' field in the perf sample has latency information
for some instructions like in memory accesses. And perf tool has 'weight'
and 'local_weight' sort keys to display the info.
But it's somewhat confusing what it shows exactly. In my understanding,
'local_weight' shows a weight in a single sample, and (global) 'weight'
shows a sum of the weights in the hist_entry.
For example:
$ perf mem record -t load dd if=/dev/zero of=/dev/null bs=4k count=1M
$ perf report --stdio -n -s +local_weight
...
#
# Overhead Samples Command Shared Object Symbol Local Weight
# ........ ....... ....... ................ ......................... ............
#
21.23% 313 dd [kernel.vmlinux] [k] lockref_get_not_zero 32
12.43% 183 dd [kernel.vmlinux] [k] lockref_get_not_zero 35
11.97% 159 dd [kernel.vmlinux] [k] lockref_get_not_zero 36
10.40% 141 dd [kernel.vmlinux] [k] lockref_put_return 32
7.63% 113 dd [kernel.vmlinux] [k] lockref_get_not_zero 33
6.37% 92 dd [kernel.vmlinux] [k] lockref_get_not_zero 34
6.15% 90 dd [kernel.vmlinux] [k] lockref_put_return 33
...
So let's look at the 'lockref_get_not_zero' symbols. The top entry
shows that 313 samples were captured with 'local_weight' 32, so the
total weight should be 313 x 32 = 10016. But it's not the case:
$ perf report --stdio -n -s +local_weight,weight -S lockref_get_not_zero
...
#
# Overhead Samples Command Shared Object Local Weight Weight
# ........ ....... ....... ................ ............ ......
#
1.36% 4 dd [kernel.vmlinux] 36 144
0.47% 4 dd [kernel.vmlinux] 37 148
0.42% 4 dd [kernel.vmlinux] 32 128
0.40% 4 dd [kernel.vmlinux] 34 136
0.35% 4 dd [kernel.vmlinux] 36 144
0.34% 4 dd [kernel.vmlinux] 35 140
0.30% 4 dd [kernel.vmlinux] 36 144
0.30% 4 dd [kernel.vmlinux] 34 136
0.30% 4 dd [kernel.vmlinux] 32 128
0.30% 4 dd [kernel.vmlinux] 32 128
...
With the 'weight' sort key, it's divided to 4 samples even with the same
info ('comm', 'dso', 'sym' and 'local_weight'). I don't think this is
what we want.
I found this because of the way it aggregates the 'weight' value. Since
it's not a period, we should not add them in the he->stat. Otherwise,
two 32 'weight' entries will create a 64 'weight' entry.
After that, new 32 'weight' samples don't have a matching entry so it'd
create a new entry and make it a 64 'weight' entry again and again.
Later, they will be merged into 128 'weight' entries during the
hists__collapse_resort() with 4 samples, multiple times like above.
Let's keep the weight and display it differently. For 'local_weight',
it can show the weight as is, and for (global) 'weight' it can display
the number multiplied by the number of samples.
With this change, I can see the expected numbers.
$ perf report --stdio -n -s +local_weight,weight -S lockref_get_not_zero
...
#
# Overhead Samples Command Shared Object Local Weight Weight
# ........ ....... ....... ................ ............ .....
#
21.23% 313 dd [kernel.vmlinux] 32 10016
12.43% 183 dd [kernel.vmlinux] 35 6405
11.97% 159 dd [kernel.vmlinux] 36 5724
7.63% 113 dd [kernel.vmlinux] 33 3729
6.37% 92 dd [kernel.vmlinux] 34 3128
4.17% 59 dd [kernel.vmlinux] 37 2183
0.08% 1 dd [kernel.vmlinux] 269 269
0.08% 1 dd [kernel.vmlinux] 38 38
Reviewed-by: Athira Jajeev <atrajeev@linux.vnet.ibm.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Athira Jajeev <atrajeev@linux.vnet.ibm.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: https://lore.kernel.org/r/20211105225617.151364-1-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit f96b4675839b66168f5a07bf964dde6c2f1c4885 ]
Use get_unaligned() instead of memcpy() to access potentially unaligned
memory, which, when accessed through a pointer, leads to undefined
behavior. get_unaligned() describes much better what is happening there
anyway even if memcpy() does the job.
In addition, since perf tool builds with -Werror, it would fire with:
util/intel-pt-decoder/../../../arch/x86/lib/insn.c: In function '__insn_get_emulate_prefix':
tools/include/../include/asm-generic/unaligned.h:10:15: error: packed attribute is unnecessary [-Werror=packed]
10 | const struct { type x; } __packed *__pptr = (typeof(__pptr))(ptr); \
because -Werror=packed would complain if the packed attribute would have
no effect on the layout of the structure.
In this case, that is intentional so disable the warning only for that
compilation unit.
That part is Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
No functional changes.
Fixes: 5ba1071f75 ("x86/insn, tools/x86: Fix undefined behavior due to potential unaligned accesses")
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Borislav Petkov <bp@suse.de>
Acked-by: Masami Hiramatsu <mhiramat@kernel.org>
Tested-by: Stephen Rothwell <sfr@canb.auug.org.au>
Link: https://lkml.kernel.org/r/YVSsIkj9Z29TyUjE@zn.tnic
Signed-off-by: Sasha Levin <sashal@kernel.org>
-F weight in perf script is broken.
# ./perf mem record
# ./perf script -F weight
Samples for 'dummy:HG' event do not have WEIGHT attribute set. Cannot
print 'weight' field.
The sample type, PERF_SAMPLE_WEIGHT_STRUCT, is an alternative of the
PERF_SAMPLE_WEIGHT sample type. They share the same space, weight. The
lower 32 bits are exactly the same for both sample type. The higher 32
bits may be different for different architecture. For a new kernel on
x86, the PERF_SAMPLE_WEIGHT_STRUCT is used. For an old kernel or other
ARCHs, the PERF_SAMPLE_WEIGHT is used.
With -F weight, current perf script will only check the input string
"weight" with the PERF_SAMPLE_WEIGHT sample type. Because the commit
ea8d0ed6ea ("perf tools: Support PERF_SAMPLE_WEIGHT_STRUCT") didn't
update the PERF_SAMPLE_WEIGHT_STRUCT sample type for perf script. For a
new kernel on x86, the check fails.
Use PERF_SAMPLE_WEIGHT_TYPE, which supports both sample types, to
replace PERF_SAMPLE_WEIGHT
Fixes: ea8d0ed6ea ("perf tools: Support PERF_SAMPLE_WEIGHT_STRUCT")
Reported-by: Joe Mario <jmario@redhat.com>
Reviewed-by: Kajol Jain <kjain@linux.ibm.com>
Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
Tested-by: Jiri Olsa <jolsa@redhat.com>
Tested-by: Joe Mario <jmario@redhat.com>
Acked-by: Jiri Olsa <jolsa@redhat.com>
Acked-by: Joe Mario <jmario@redhat.com>
Cc: Andi Kleen <ak@linux.intel.com>
Link: https://lore.kernel.org/r/1632929894-102778-1-git-send-email-kan.liang@linux.intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Got following build fail on powerpc:
CC arch/powerpc/util/skip-callchain-idx.o
In function ‘check_return_reg’,
inlined from ‘check_return_addr’ at arch/powerpc/util/skip-callchain-idx.c:213:7,
inlined from ‘arch_skip_callchain_idx’ at arch/powerpc/util/skip-callchain-idx.c:265:7:
arch/powerpc/util/skip-callchain-idx.c:54:18: error: ‘dwarf_frame_register’ accessing 96 bytes \
in a region of size 64 [-Werror=stringop-overflow=]
54 | result = dwarf_frame_register(frame, ra_regno, ops_mem, &ops, &nops);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
arch/powerpc/util/skip-callchain-idx.c: In function ‘arch_skip_callchain_idx’:
arch/powerpc/util/skip-callchain-idx.c:54:18: note: referencing argument 3 of type ‘Dwarf_Op *’
In file included from /usr/include/elfutils/libdwfl.h:32,
from arch/powerpc/util/skip-callchain-idx.c:10:
/usr/include/elfutils/libdw.h:1069:12: note: in a call to function ‘dwarf_frame_register’
1069 | extern int dwarf_frame_register (Dwarf_Frame *frame, int regno,
| ^~~~~~~~~~~~~~~~~~~~
cc1: all warnings being treated as errors
The dwarf_frame_register args changed with [1],
Updating ops_mem accordingly.
[1] https://sourceware.org/git/?p=elfutils.git;a=commit;h=5621fe5443da23112170235dd5cac161e5c75e65
Reviewed-by: Kajol Jain <kjain@linux.ibm.com>
Signed-off-by: Jiri Olsa <jolsa@redhat.com>
Acked-by: Mark Wieelard <mjw@redhat.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
Link: https://lore.kernel.org/r/20210928195253.1267023-1-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
When perf.data is not written cleanly, we would like to process existing
data as much as possible (please see f_header.data.size == 0 condition
in perf_session__read_header). However, perf.data with partial data may
crash perf. Specifically, we see crash in 'perf script' for NULL
session->header.env.arch.
Fix this by checking session->header.env.arch before using it to determine
native_arch. Also split the if condition so it is easier to read.
Committer notes:
If it is a pipe, we already assume is a native arch, so no need to check
session->header.env.arch.
Signed-off-by: Song Liu <songliubraving@fb.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: kernel-team@fb.com
Cc: stable@vger.kernel.org
Link: http://lore.kernel.org/lkml/20211004053238.514936-1-songliubraving@fb.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The Topdown metrics events were added as 'perf stat' default events
since commit 42641d6f4d ("perf stat: Add Topdown metrics events as
default events").
However, the perf attr tests were not updated
accordingly.
The perf attr test fails on the platform which supports Topdown metrics.
# perf test 17
17: Setup struct perf_event_attr :FAILED!
Add Topdown metrics events into perf attr test cases. Make them optional
since they are only available on newer platforms.
Fixes: 42641d6f4d ("perf stat: Add Topdown metrics events as default events")
Reported-by: kernel test robot <oliver.sang@intel.com>
Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lore.kernel.org/lkml/1633031566-176517-1-git-send-email-kan.liang@linux.intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
If the 'perf iostat' user specifies two or more iio_root_ports and also
specifies the cpu(s) by -C which is not *connected to all* the above iio
ports, the iostat_print_metric() will run into trouble:
For example:
$ perf iostat list
S0-uncore_iio_0<0000:16>
S1-uncore_iio_0<0000:97> # <--- CPU 1 is located in the socket S0
$ perf iostat 0000:16,0000:97 -C 1 -- ls
port Inbound Read(MB) Inbound Write(MB) Outbound Read(MB) Outbound
Write(MB) ../perf-iostat: line 12: 104418 Segmentation fault
(core dumped) perf stat --iostat$DELIMITER$*
The core-dump stack says, in the above corner case, the returned
(struct perf_counts_values *) count will be NULL, and the caller
iostat_print_metric() apparently doesn't not handle this case.
433 struct perf_counts_values *count = perf_counts(evsel->counts, die, 0);
434
435 if (count->run && count->ena) {
(gdb) p count
$1 = (struct perf_counts_values *) 0x0
The deeper reason is that there are actually no statistics from the user
specified pair "iostat 0000:X, -C (disconnected) Y ", but let's fix it with
minimum cost by adding a NULL check in the user space.
Fixes: f9ed693e8b ("perf stat: Enable iostat mode for x86 platforms")
Signed-off-by: Like Xu <likexu@tencent.com>
Cc: Alexander Antonov <alexander.antonov@linux.intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.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: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lore.kernel.org/lkml/20210927081115.39568-2-likexu@tencent.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
An iostate use case like "perf iostat 0000:16,0000:97 -- ls" should be
implemented to work in system-wide mode to ensure that the output from
print_header() is consistent with the user documentation perf-iostat.txt,
rather than incorrectly assuming that the kernel does not support it:
Error:
The sys_perf_event_open() syscall returned with 22 (Invalid argument) \
for event (uncore_iio_0/event=0x83,umask=0x04,ch_mask=0xF,fc_mask=0x07/).
/bin/dmesg | grep -i perf may provide additional information.
This error is easily fixed by assigning system-wide mode by default
for IOSTAT_RUN only when the target cpu_list is unspecified.
Fixes: f07952b179 ("perf stat: Basic support for iostat in perf")
Signed-off-by: Like Xu <likexu@tencent.com>
Cc: Alexander Antonov <alexander.antonov@linux.intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.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: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lore.kernel.org/lkml/20210927081115.39568-1-likexu@tencent.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
If there is no configuration file at first, the user can write any pair
of "key.subkey=value" to the newly created configuration file, while
value validation against a valid configurable key is *deferred* until
the next execution or the implied execution of "perf config ... ".
For example:
$ rm ~/.perfconfig
$ perf config call-graph.dump-size=65529
$ cat ~/.perfconfig
# this file is auto-generated.
[call-graph]
dump-size = 65529
$ perf config call-graph.dump-size=2048
callchain: Incorrect stack dump size (max 65528): 65529
Error: wrong config key-value pair call-graph.dump-size=65529
The user might expect that the second value 2048 is valid and can be
updated to the configuration file, but the error message is very
confusing because the first value 65529 is not reported as an error
during the last configuration.
It is recommended not to change the current behavior of delayed
validation (as more effort is needed), but to refine the original error
message to *clearly indicate* that the cause of the error is the
configuration file.
Signed-off-by: Like Xu <likexu@tencent.com>
Acked-by: 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: Peter Zijlstra <peterz@infradead.org>
Link: http://lore.kernel.org/lkml/20210924115817.58689-1-likexu@tencent.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
This test occasionally fails on aarch64 when a sample is taken in
free@plt and it fails with "Bytes read differ from those read by
objdump".
This is because that symbol is near a section boundary in the elf file.
Despite the -z option to always output zeros, objdump uses
bfd_map_over_sections() to iterate through the elf file so it doesn't
see outside of the sections where these zeros are and can't print them.
For example this boundary proceeds free@plt in libc with a gap of 48
bytes between .plt and .text:
objdump -d -z --start-address=0x23cc8 --stop-address=0x23d08 libc-2.30.so
libc-2.30.so: file format elf64-littleaarch64
Disassembly of section .plt:
0000000000023cc8 <*ABS*+0x7fd00@plt+0x8>:
23cc8: 91018210 add x16, x16, #0x60
23ccc: d61f0220 br x17
Disassembly of section .text:
0000000000023d00 <abort@@GLIBC_2.17-0x98>:
23d00: a9bf7bfd stp x29, x30, [sp, #-16]!
23d04: 910003fd mov x29, sp
Taking a sample in free@plt is very rare because it is so small, but the
test can be forced to fail almost every time on any platform by linking
the test with a shared library that has a single empty function and
calling it in a loop.
The fix is to zero the buffers so that when there is a jump in the
addresses output by objdump, zeros are already filled in between.
Signed-off-by: James Clark <james.clark@arm.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: http://lore.kernel.org/lkml/20210906152238.3415467-1-james.clark@arm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Perf code re-implements libbpf's btf__load_from_kernel_by_id() API as
a weak function, presumably to dynamically link against old version of
libbpf shared library. Unfortunately this causes compilation warning
when perf is compiled against libbpf v0.6+.
For now, just ignore deprecation warning, but there might be a better
solution, depending on perf's needs.
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: kernel-team@fb.com
LPU-Reference: 20210914170004.4185659-1-andrii@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
It's later supposed to be either a correct address or NULL. Without the
initialization, it may contain an undefined value which results in the
following segmentation fault:
# perf top --sort comm -g --ignore-callees=do_idle
terminates with:
#0 0x00007ffff56b7685 in __strlen_avx2 () from /lib64/libc.so.6
#1 0x00007ffff55e3802 in strdup () from /lib64/libc.so.6
#2 0x00005555558cb139 in hist_entry__init (callchain_size=<optimized out>, sample_self=true, template=0x7fffde7fb110, he=0x7fffd801c250) at util/hist.c:489
#3 hist_entry__new (template=template@entry=0x7fffde7fb110, sample_self=sample_self@entry=true) at util/hist.c:564
#4 0x00005555558cb4ba in hists__findnew_entry (hists=hists@entry=0x5555561d9e38, entry=entry@entry=0x7fffde7fb110, al=al@entry=0x7fffde7fb420,
sample_self=sample_self@entry=true) at util/hist.c:657
#5 0x00005555558cba1b in __hists__add_entry (hists=hists@entry=0x5555561d9e38, al=0x7fffde7fb420, sym_parent=<optimized out>, bi=bi@entry=0x0, mi=mi@entry=0x0,
sample=sample@entry=0x7fffde7fb4b0, sample_self=true, ops=0x0, block_info=0x0) at util/hist.c:288
#6 0x00005555558cbb70 in hists__add_entry (sample_self=true, sample=0x7fffde7fb4b0, mi=0x0, bi=0x0, sym_parent=<optimized out>, al=<optimized out>, hists=0x5555561d9e38)
at util/hist.c:1056
#7 iter_add_single_cumulative_entry (iter=0x7fffde7fb460, al=<optimized out>) at util/hist.c:1056
#8 0x00005555558cc8a4 in hist_entry_iter__add (iter=iter@entry=0x7fffde7fb460, al=al@entry=0x7fffde7fb420, max_stack_depth=<optimized out>, arg=arg@entry=0x7fffffff7db0)
at util/hist.c:1231
#9 0x00005555557cdc9a in perf_event__process_sample (machine=<optimized out>, sample=0x7fffde7fb4b0, evsel=<optimized out>, event=<optimized out>, tool=0x7fffffff7db0)
at builtin-top.c:842
#10 deliver_event (qe=<optimized out>, qevent=<optimized out>) at builtin-top.c:1202
#11 0x00005555558a9318 in do_flush (show_progress=false, oe=0x7fffffff80e0) at util/ordered-events.c:244
#12 __ordered_events__flush (oe=oe@entry=0x7fffffff80e0, how=how@entry=OE_FLUSH__TOP, timestamp=timestamp@entry=0) at util/ordered-events.c:323
#13 0x00005555558a9789 in __ordered_events__flush (timestamp=<optimized out>, how=<optimized out>, oe=<optimized out>) at util/ordered-events.c:339
#14 ordered_events__flush (how=OE_FLUSH__TOP, oe=0x7fffffff80e0) at util/ordered-events.c:341
#15 ordered_events__flush (oe=oe@entry=0x7fffffff80e0, how=how@entry=OE_FLUSH__TOP) at util/ordered-events.c:339
#16 0x00005555557cd631 in process_thread (arg=0x7fffffff7db0) at builtin-top.c:1114
#17 0x00007ffff7bb817a in start_thread () from /lib64/libpthread.so.0
#18 0x00007ffff5656dc3 in clone () from /lib64/libc.so.6
If you look at the frame #2, the code is:
488 if (he->srcline) {
489 he->srcline = strdup(he->srcline);
490 if (he->srcline == NULL)
491 goto err_rawdata;
492 }
If he->srcline is not NULL (it is not NULL if it is uninitialized rubbish),
it gets strdupped and strdupping a rubbish random string causes the problem.
Also, if you look at the commit 1fb7d06a50, it adds the srcline property
into the struct, but not initializing it everywhere needed.
Committer notes:
Now I see, when using --ignore-callees=do_idle we end up here at line
2189 in add_callchain_ip():
2181 if (al.sym != NULL) {
2182 if (perf_hpp_list.parent && !*parent &&
2183 symbol__match_regex(al.sym, &parent_regex))
2184 *parent = al.sym;
2185 else if (have_ignore_callees && root_al &&
2186 symbol__match_regex(al.sym, &ignore_callees_regex)) {
2187 /* Treat this symbol as the root,
2188 forgetting its callees. */
2189 *root_al = al;
2190 callchain_cursor_reset(cursor);
2191 }
2192 }
And the al that doesn't have the ->srcline field initialized will be
copied to the root_al, so then, back to:
1211 int hist_entry_iter__add(struct hist_entry_iter *iter, struct addr_location *al,
1212 int max_stack_depth, void *arg)
1213 {
1214 int err, err2;
1215 struct map *alm = NULL;
1216
1217 if (al)
1218 alm = map__get(al->map);
1219
1220 err = sample__resolve_callchain(iter->sample, &callchain_cursor, &iter->parent,
1221 iter->evsel, al, max_stack_depth);
1222 if (err) {
1223 map__put(alm);
1224 return err;
1225 }
1226
1227 err = iter->ops->prepare_entry(iter, al);
1228 if (err)
1229 goto out;
1230
1231 err = iter->ops->add_single_entry(iter, al);
1232 if (err)
1233 goto out;
1234
That al at line 1221 is what hist_entry_iter__add() (called from
sample__resolve_callchain()) saw as 'root_al', and then:
iter->ops->add_single_entry(iter, al);
will go on with al->srcline with a bogus value, I'll add the above
sequence to the cset and apply, thanks!
Signed-off-by: Michael Petlan <mpetlan@redhat.com>
CC: Milian Wolff <milian.wolff@kdab.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Fixes: 1fb7d06a50 ("perf report Use srcline from callchain for hist entries")
Link: https //lore.kernel.org/r/20210719145332.29747-1-mpetlan@redhat.com
Reported-by: Juri Lelli <jlelli@redhat.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
set_print_ip_opts() was not being called when type != attr->type
because there is not a one-to-one relationship between output types
and attr->type. That resulted in ip not printing.
The attr_type() function is removed, and the match of attr->type to
output type is corrected.
Example on ADL using taskset to select an atom cpu:
# perf record -e cpu_atom/cpu-cycles/ taskset 0x1000 uname
Linux
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.003 MB perf.data (7 samples) ]
Before:
# perf script | head
taskset 428 [-01] 10394.179041: 1 cpu_atom/cpu-cycles/:
taskset 428 [-01] 10394.179043: 1 cpu_atom/cpu-cycles/:
taskset 428 [-01] 10394.179044: 11 cpu_atom/cpu-cycles/:
taskset 428 [-01] 10394.179045: 407 cpu_atom/cpu-cycles/:
taskset 428 [-01] 10394.179046: 16789 cpu_atom/cpu-cycles/:
taskset 428 [-01] 10394.179052: 676300 cpu_atom/cpu-cycles/:
uname 428 [-01] 10394.179278: 4079859 cpu_atom/cpu-cycles/:
After:
# perf script | head
taskset 428 10394.179041: 1 cpu_atom/cpu-cycles/: ffffffff95a0bb97 __intel_pmu_enable_all.constprop.48+0x47 ([kernel.kallsyms])
taskset 428 10394.179043: 1 cpu_atom/cpu-cycles/: ffffffff95a0bb97 __intel_pmu_enable_all.constprop.48+0x47 ([kernel.kallsyms])
taskset 428 10394.179044: 11 cpu_atom/cpu-cycles/: ffffffff95a0bb97 __intel_pmu_enable_all.constprop.48+0x47 ([kernel.kallsyms])
taskset 428 10394.179045: 407 cpu_atom/cpu-cycles/: ffffffff95a0bb97 __intel_pmu_enable_all.constprop.48+0x47 ([kernel.kallsyms])
taskset 428 10394.179046: 16789 cpu_atom/cpu-cycles/: ffffffff95a0bb97 __intel_pmu_enable_all.constprop.48+0x47 ([kernel.kallsyms])
taskset 428 10394.179052: 676300 cpu_atom/cpu-cycles/: 7f829ef73800 cfree+0x0 (/lib/libc-2.32.so)
uname 428 10394.179278: 4079859 cpu_atom/cpu-cycles/: ffffffff95bae912 vma_interval_tree_remove+0x1f2 ([kernel.kallsyms])
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Reviewed-by: Kan Liang <kan.liang@linux.intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lore.kernel.org/lkml/20210911133053.15682-1-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Some x86 microarchitectures fuse a subset of cmp/test/ALU instructions
with branch instructions, and thus perf annotate highlight such valid
pairs as fused.
When annotated with source, perf uses struct disasm_line to contain
either source or instruction line from objdump output. Usually, a C
statement generates multiple instructions which include such
cmp/test/ALU + branch instruction pairs. But in case of assembly
function, each individual assembly source line generate one
instruction.
The 'perf annotate' instruction fusion logic assumes the previous
disasm_line as the previous instruction line, which is wrong because,
for assembly function, previous disasm_line contains source line. And
thus perf fails to highlight valid fused instruction pairs for assembly
functions.
Fix it by searching backward until we find an instruction line and
consider that disasm_line as fused with current branch instruction.
Before:
│ cmpq %rcx, RIP+8(%rsp)
0.00 │ cmp %rcx,0x88(%rsp)
│ je .Lerror_bad_iret <--- Source line
0.14 │ ┌──je b4 <--- Instruction line
│ │movl %ecx, %eax
After:
│ cmpq %rcx, RIP+8(%rsp)
0.00 │ ┌──cmp %rcx,0x88(%rsp)
│ │je .Lerror_bad_iret
0.14 │ ├──je b4
│ │movl %ecx, %eax
Reviewed-by: Jin Yao <yao.jin@linux.intel.com>
Signed-off-by: Ravi Bangoria <ravi.bangoria@amd.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.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>
Link: https //lore.kernel.org/r/20210911043854.8373-1-ravi.bangoria@amd.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Currently perf saves a build-id with size but old versions assumes the
size of 20. In case the build-id is less than 20 (like for MD5), it'd
fill the rest with 0s.
I saw a problem when old version of perf record saved a binary in the
build-id cache and new version of perf reads the data. The symbols
should be read from the build-id cache (as the path no longer has the
same binary) but it failed due to mismatch in the build-id.
symsrc__init: build id mismatch for /home/namhyung/.debug/.build-id/53/e4c2f42a4c61a2d632d92a72afa08f00000000/elf.
The build-id event in the data has 20 byte build-ids, but it saw a
different size (16) when it reads the build-id of the elf file in the
build-id cache.
$ readelf -n ~/.debug/.build-id/53/e4c2f42a4c61a2d632d92a72afa08f00000000/elf
Displaying notes found in: .note.gnu.build-id
Owner Data size Description
GNU 0x00000010 NT_GNU_BUILD_ID (unique build ID bitstring)
Build ID: 53e4c2f42a4c61a2d632d92a72afa08f
Let's fix this by allowing trailing zeros if the size is different.
Fixes: 39be8d0115 ("perf tools: Pass build_id object to dso__build_id_equal()")
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Acked-by: Jiri Olsa <jolsa@redhat.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lore.kernel.org/lkml/20210910224630.1084877-1-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The btf__get_from_id() function was deprecated in favour of
btf__load_from_kernel_by_id(), but it is still avaiable, so use it to
provide a weak function btf__load_from_kernel_by_id() for older libbpf
when building perf with LIBBPF_DYNAMIC=1, i.e. using the system's libbpf
package.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Previously the regext expected MOVE_MOUNT_[FT]_*, but in the next patch
a flag that doesn't match that expression will be added, MOVE_MOUNT_SET_GROUP
To make this more future proof, take advantage of the fact that the only
one we don't need to cover is MOVE_MOUNT__MASK and use MOVE_MOUNT_[^_]+_*_.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Perf records IBS (Instruction Based Sampling) extra sample data when
'perf record --raw-samples' is used with an IBS-compatible event, on a
machine that supports IBS. IBS support is indicated in
CPUID_Fn80000001_ECX bit #10.
Up until now, users have been able to see the extra sample data solely
in raw hex format using 'perf report --dump-raw-trace'. From there,
users could decode the data either manually, or by using an external
script.
Enable the built-in 'perf report --dump-raw-trace' to do the decoding of
the extra sample data bits, so manual or external script decoding isn't
necessary.
Example usage:
$ sudo perf record -c 10000001 -a --raw-samples -e ibs_fetch/rand_en=1/,ibs_op/cnt_ctl=1/ -C 0,1 taskset -c 0,1 7za b -mmt2 | perf report --dump-raw-trace
Stdout contains IBS Fetch samples, e.g.:
ibs_fetch_ctl: 02170007ffffffff MaxCnt 1048560 Cnt 1048560 Lat 7 En 1 Val 1 Comp 1 IcMiss 0 PhyAddrValid 1 L1TlbPgSz 4KB L1TlbMiss 0 L2TlbMiss 0 RandEn 1 L2Miss 0
IbsFetchLinAd: 000056016b2ead40
IbsFetchPhysAd: 000000115cedfd40
c_ibs_ext_ctl: 0000000000000000 IbsItlbRefillLat 0
..and IBS Op samples, e.g.:
ibs_op_ctl: 0000009e009e8968 MaxCnt 10000000 En 1 Val 1 CntCtl 1=uOps CurCnt 158
IbsOpRip: 000056016b2ea73d
ibs_op_data: 00000000000b0002 CompToRetCtr 2 TagToRetCtr 11 BrnRet 0 RipInvalid 0 BrnFuse 0 Microcode 0
ibs_op_data2: 0000000000000002 CacheHitSt 0=M-state RmtNode 0 DataSrc 2=Local node cache
ibs_op_data3: 0000000000c60002 LdOp 0 StOp 1 DcL1TlbMiss 0 DcL2TlbMiss 0 DcL1TlbHit2M 0 DcL1TlbHit1G 0 DcL2TlbHit2M 0 DcMiss 0 DcMisAcc 0 DcWcMemAcc 0 DcUcMemAcc 0 DcLockedOp 0 DcMissNoMabAlloc 0 DcLinAddrValid 1 DcPhyAddrValid 1 DcL2TlbHit1G 0 L2Miss 0 SwPf 0 OpMemWidth 4 bytes OpDcMissOpenMemReqs 0 DcMissLat 0 TlbRefillLat 0
IbsDCLinAd: 00007f133c319ce0
IbsDCPhysAd: 0000000270485ce0
Committer notes:
Fixed up this:
util/amd-sample-raw.c: In function ‘evlist__amd_sample_raw’:
util/amd-sample-raw.c:125:42: error: ‘ bytes’ directive output may be truncated writing 6 bytes into a region of size between 4 and 7 [-Werror=format-truncation=]
125 | " OpMemWidth %2d bytes", 1 << (reg.op_mem_width - 1));
| ^~~~~~
In file included from /usr/include/stdio.h:866,
from util/amd-sample-raw.c:7:
/usr/include/bits/stdio2.h:71:10: note: ‘__builtin___snprintf_chk’ output between 21 and 24 bytes into a destination of size 21
71 | return __builtin___snprintf_chk (__s, __n, __USE_FORTIFY_LEVEL - 1,
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
72 | __glibc_objsize (__s), __fmt,
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
73 | __va_arg_pack ());
| ~~~~~~~~~~~~~~~~~
cc1: all warnings being treated as errors
As that %2d won't limit the number of chars to 2, just state that 2 is
the minimal width:
$ cat printf.c
#include <stdio.h>
#include <stdlib.h>
int main(int argc, char *argv[])
{
char bf[64];
int len = snprintf(bf, sizeof(bf), "%2d", atoi(argv[1]));
printf("strlen(%s): %u\n", bf, len);
return 0;
}
$ ./printf 1
strlen( 1): 2
$ ./printf 12
strlen(12): 2
$ ./printf 123
strlen(123): 3
$ ./printf 1234
strlen(1234): 4
$ ./printf 12345
strlen(12345): 5
$ ./printf 123456
strlen(123456): 6
$
And since we probably don't want that output to be truncated, just
assume the worst case, as the compiler did, and add a few more chars to
that buffer.
Also use sizeof(var) instead of sizeof(dup-of-wanted-format-string) to
avoid bugs when changing one but not the other.
I also had to change this:
-#include <asm/amd-ibs.h>
+#include "../../arch/x86/include/asm/amd-ibs.h"
To make it build on other architectures, just like intel-pt does.
Signed-off-by: Kim Phillips <kim.phillips@amd.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Joao Martins <joao.m.martins@oracle.com>
Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Robert Richter <robert.richter@amd.com>
Cc: Stephane Eranian <eranian@google.com>
Link: https //lore.kernel.org/r/20210817221509.88391-4-kim.phillips@amd.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Instead of using the file offset in the debug file.
This fixes a regression from 00a3423492 ("perf symbols: Make
dso__load_bfd_symbols() load PE files from debug cache only"), causing
incorrect symbol resolution when debug file have been stripped from
non-debug sections (in which case its .text section is empty and doesn't
have any file position).
The debug files could also be created with a different file alignment,
and have different file positions from the mmap-ed binary, or have the
section reordered.
This instead looks for the file image base, using the corresponding bfd
*ABS* symbols. As PE symbols only have 4 bytes, it also needs to keep
.text section vma high bits.
Signed-off-by: Remi Bernon <rbernon@codeweavers.com>
Fixes: 00a3423492 ("perf symbols: Make dso__load_bfd_symbols() load PE files from debug cache only")
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Nicholas Fraser <nfraser@codeweavers.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lore.kernel.org/lkml/20210909192637.4139125-1-rbernon@codeweavers.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
To pick the changes in these csets:
59ab844eed ("compat: remove some compat entry points")
dce4910396 ("mm: wire up syscall process_mrelease")
b48c7236b1 ("exit/bdflush: Remove the deprecated bdflush system call")
That add support for this new syscall in tools such as 'perf trace'.
For instance, this is now possible:
# perf trace -v -e process_mrelease
event qualifier tracepoint filter: (common_pid != 19351 && common_pid != 9112) && (id == 448)
^C#
That is the filter expression attached to the raw_syscalls:sys_{enter,exit}
tracepoints.
$ grep process_mrelease tools/perf/arch/x86/entry/syscalls/syscall_64.tbl
448 common process_mrelease sys_process_mrelease
$
This addresses these perf build warnings:
Warning: Kernel ABI header at 'tools/include/uapi/asm-generic/unistd.h' differs from latest version at 'include/uapi/asm-generic/unistd.h'
diff -u tools/include/uapi/asm-generic/unistd.h include/uapi/asm-generic/unistd.h
Warning: Kernel ABI header at 'tools/perf/arch/x86/entry/syscalls/syscall_64.tbl' differs from latest version at 'arch/x86/entry/syscalls/syscall_64.tbl'
diff -u tools/perf/arch/x86/entry/syscalls/syscall_64.tbl arch/x86/entry/syscalls/syscall_64.tbl
Warning: Kernel ABI header at 'tools/perf/arch/powerpc/entry/syscalls/syscall.tbl' differs from latest version at 'arch/powerpc/kernel/syscalls/syscall.tbl'
diff -u tools/perf/arch/powerpc/entry/syscalls/syscall.tbl arch/powerpc/kernel/syscalls/syscall.tbl
Warning: Kernel ABI header at 'tools/perf/arch/s390/entry/syscalls/syscall.tbl' differs from latest version at 'arch/s390/kernel/syscalls/syscall.tbl'
diff -u tools/perf/arch/s390/entry/syscalls/syscall.tbl arch/s390/kernel/syscalls/syscall.tbl
Warning: Kernel ABI header at 'tools/perf/arch/mips/entry/syscalls/syscall_n64.tbl' differs from latest version at 'arch/mips/kernel/syscalls/syscall_n64.tbl'
diff -u tools/perf/arch/mips/entry/syscalls/syscall_n64.tbl arch/mips/kernel/syscalls/syscall_n64.tbl
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Eric W. Biederman <ebiederm@xmission.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
To pick the changes in:
Fixes: d32f89da7f ("net: add accept helper not installing fd")
Fixes: bc49d8169a ("mctp: Add MCTP base")
This automagically adds support for the AF_MCTP protocol domain:
$ tools/perf/trace/beauty/socket.sh > before
$ cp include/linux/socket.h tools/perf/trace/beauty/include/linux/socket.h
$ tools/perf/trace/beauty/socket.sh > after
$ diff -u before after
--- before 2021-09-06 11:57:14.972747200 -0300
+++ after 2021-09-06 11:57:30.541920222 -0300
@@ -44,4 +44,5 @@
[42] = "QIPCRTR",
[43] = "SMC",
[44] = "XDP",
+ [45] = "MCTP",
};
$
This will allow 'perf trace' to translate 45 into "MCTP" as is done with
the other domains:
# perf trace -e socket*
0.000 chronyd/1029 socket(family: INET, type: DGRAM|CLOEXEC|NONBLOCK, protocol: IP) = 4
^C#
This addresses this perf build warning:
Warning: Kernel ABI header at 'tools/perf/trace/beauty/include/linux/socket.h' differs from latest version at 'include/linux/socket.h'
diff -u tools/perf/trace/beauty/include/linux/socket.h include/linux/socket.h
Cc: David S. Miller <davem@davemloft.net>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Jeremy Kerr <jk@codeconstruct.com.au>
Cc: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Merge more updates from Andrew Morton:
"147 patches, based on 7d2a07b769.
Subsystems affected by this patch series: mm (memory-hotplug, rmap,
ioremap, highmem, cleanups, secretmem, kfence, damon, and vmscan),
alpha, percpu, procfs, misc, core-kernel, MAINTAINERS, lib,
checkpatch, epoll, init, nilfs2, coredump, fork, pids, criu, kconfig,
selftests, ipc, and scripts"
* emailed patches from Andrew Morton <akpm@linux-foundation.org>: (94 commits)
scripts: check_extable: fix typo in user error message
mm/workingset: correct kernel-doc notations
ipc: replace costly bailout check in sysvipc_find_ipc()
selftests/memfd: remove unused variable
Kconfig.debug: drop selecting non-existing HARDLOCKUP_DETECTOR_ARCH
configs: remove the obsolete CONFIG_INPUT_POLLDEV
prctl: allow to setup brk for et_dyn executables
pid: cleanup the stale comment mentioning pidmap_init().
kernel/fork.c: unexport get_{mm,task}_exe_file
coredump: fix memleak in dump_vma_snapshot()
fs/coredump.c: log if a core dump is aborted due to changed file permissions
nilfs2: use refcount_dec_and_lock() to fix potential UAF
nilfs2: fix memory leak in nilfs_sysfs_delete_snapshot_group
nilfs2: fix memory leak in nilfs_sysfs_create_snapshot_group
nilfs2: fix memory leak in nilfs_sysfs_delete_##name##_group
nilfs2: fix memory leak in nilfs_sysfs_create_##name##_group
nilfs2: fix NULL pointer in nilfs_##name##_attr_release
nilfs2: fix memory leak in nilfs_sysfs_create_device_group
trap: cleanup trap_init()
init: move usermodehelper_enable() to populate_rootfs()
...
