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>
Synthesize an attribute event and sample events for Intel PT Event Trace
events represented by CFE and EVD packets.
Committer notes:
Make 'struct perf_synth_intel_evd evd[]' evd[0] at the end of 'struct
perf_synth_intel_evt' as it is breaking the build with in many compilers
with (e.g. clang version 13.0.0 (Fedora 13.0.0-3.fc35)):
util/intel-pt.c:2213:31: error: field 'cfe' with variable sized type 'struct perf_synth_intel_evt' not at the end of a struct or class is a GNU extension [-Werror,-Wgnu-variable-sized-type-not-at-end]
struct perf_synth_intel_evt cfe;
^
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: https://lore.kernel.org/r/20220124084201.2699795-15-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Parser did not take ':' into account.
Example:
Before:
$ perf record -e intel_pt//u uname
Linux
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.026 MB perf.data ]
$ perf inject -i perf.data --vm-time-correlation="dry-run 123"
$ perf inject -i perf.data --vm-time-correlation="dry-run 123:456"
Failed to parse VM Time Correlation options
0x620 [0x98]: failed to process type: 70 [Invalid argument]
$
After:
$ perf inject -i perf.data --vm-time-correlation="dry-run 123:456"
$
Fixes: e3ff42bdeb ("perf intel-pt: Parse VM Time Correlation options and set up decoding")
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Riccardo Mancini <rickyman7@gmail.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20211215080636.149562-2-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Normally, for cycle-acccurate mode, IPC values are an exact number of
instructions and cycles. Due to the granularity of timestamps, that happens
only when a CYC packet correlates to the event.
Support the itrace 'A' option, to use instead, the number of cycles
associated with the current timestamp. This provides IPC information for
every change of timestamp, but at the expense of accuracy. Due to the
granularity of timestamps, the actual number of cycles increases even
though the cycles reported does not. The number of instructions is known,
but if IPC is reported, cycles can be too low and so IPC is too high. Note
that inaccuracy decreases as the period of sampling increases i.e. if the
number of cycles is too low by a small amount, that becomes less
significant if the number of cycles is large.
Furthermore, it can be used in conjunction with dlfilter-show-cycles.so
to provide higher granularity cycle information.
Reviewed-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: https://lore.kernel.org/r/20211027080334.365596-4-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The Intel PT decoder limits the number of unconditional branches (e.g.
jmps) decoded without consuming any trace packets. Generally, a loop
needs a conditional branch which generates a TNT packet, whereas a "ret"
instruction will generate a TIP or TNT packet. So exceeding the limit is
assumed to be a never-ending loop, which can happen if there has been a
decoding error putting the decoder at the wrong place in the code.
Up until now, the limit of 10000 has been enough but some analytic
purposes have been reported to exceed that.
Increase the limit to 100000, and make it configurable via perf config
intel-pt.max-loops. Also amend the "Never-ending loop" message to
mention the configuration entry.
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Reviewed-by: Andi Kleen <ak@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lore.kernel.org/lkml/20210701175132.3977-1-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The decoder reports the current instruction if it was decoded. In some
cases the current instruction is not decoded, in which case the instruction
bytes length must be set to zero. Ensure that is always done.
Note perf script can anyway get the instruction bytes for any samples where
they are not present.
Also note, that there is a redundant "ptq->insn_len = 0" statement which is
not removed until a subsequent patch in order to make this patch apply
cleanly to stable branches.
Example:
A machne that supports TSX is required. It will have flag "rtm". Kernel
parameter tsx=on may be required.
# for w in `cat /proc/cpuinfo | grep -m1 flags `;do echo $w | grep rtm ; done
rtm
Test program:
#include <stdio.h>
#include <immintrin.h>
int main()
{
int x = 0;
if (_xbegin() == _XBEGIN_STARTED) {
x = 1;
_xabort(1);
} else {
printf("x = %d\n", x);
}
return 0;
}
Compile with -mrtm i.e.
gcc -Wall -Wextra -mrtm xabort.c -o xabort
Record:
perf record -e intel_pt/cyc/u --filter 'filter main @ ./xabort' ./xabort
Before:
# perf script --itrace=xe -F+flags,+insn,-period --xed --ns
xabort 1478 [007] 92161.431348581: transactions: x 400b81 main+0x14 (/root/xabort) mov $0xffffffff, %eax
xabort 1478 [007] 92161.431348624: transactions: tx abrt 400b93 main+0x26 (/root/xabort) mov $0xffffffff, %eax
After:
# perf script --itrace=xe -F+flags,+insn,-period --xed --ns
xabort 1478 [007] 92161.431348581: transactions: x 400b81 main+0x14 (/root/xabort) xbegin 0x6
xabort 1478 [007] 92161.431348624: transactions: tx abrt 400b93 main+0x26 (/root/xabort) xabort $0x1
Fixes: faaa87680b ("perf intel-pt/bts: Report instruction bytes and length in sample")
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: stable@vger.kernel.org
Link: http://lore.kernel.org/lkml/20210519074515.9262-3-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Even when VMX TSC Offset is not changing (during perf record), different
virtual machines can have different TSC Offsets. There is a Virtual Machine
Control Structure (VMCS) for each virtual CPU, the address of which is
reported to Intel PT in the VMCS packet. We do not know which VMCS belongs
to which virtual machine, so use a tree to keep track of VMCS information.
Then the decoder will be able to use the current VMCS value to look up the
current TSC Offset.
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Reviewed-by: Andi Kleen <ak@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: https://lore.kernel.org/r/20210430070309.17624-9-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The instruction latency information can be recorded on some platforms,
e.g., the Intel Sapphire Rapids server. With both memory latency
(weight) and the new instruction latency information, users can easily
locate the expensive load instructions, and also understand the time
spent in different stages. The users can optimize their applications in
different pipeline stages.
The 'weight' field is shared among different architectures. Reusing the
'weight' field may impacts other architectures. Add a new field to store
the instruction latency.
Like the 'weight' support, introduce a 'ins_lat' for the global
instruction latency, and a 'local_ins_lat' for the local instruction
latency version.
Add new sort functions, INSTR Latency and Local INSTR Latency,
accordingly.
Add local_ins_lat to the default_mem_sort_order[].
Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Madhavan Srinivasan <maddy@linux.vnet.ibm.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/1612296553-21962-7-git-send-email-kan.liang@linux.intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The new sample type, PERF_SAMPLE_WEIGHT_STRUCT, is an alternative of the
PERF_SAMPLE_WEIGHT sample type. Users can apply either the
PERF_SAMPLE_WEIGHT sample type or the PERF_SAMPLE_WEIGHT_STRUCT sample
type to retrieve the sample weight, but they cannot apply both sample
types simultaneously.
The new sample type shares the same space as the PERF_SAMPLE_WEIGHT
sample type. The lower 32 bits are exactly the same for both sample
type. The higher 32 bits may be different for different architecture.
Add arch specific arch_evsel__set_sample_weight() to set the new sample
type for X86. Only store the lower 32 bits for the sample->weight if the
new sample type is applied. In practice, no memory access could last
than 4G cycles. No data will be lost.
If the kernel doesn't support the new sample type. Fall back to the
PERF_SAMPLE_WEIGHT sample type.
There is no impact for other architectures.
Committer notes:
Fixup related to PERF_SAMPLE_CODE_PAGE_SIZE, present in acme/perf/core
but not upstream yet.
Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Madhavan Srinivasan <maddy@linux.vnet.ibm.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/1612296553-21962-6-git-send-email-kan.liang@linux.intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Use the new itrace 'q' option to add support for a mode of decoding that
ignores TNT, does not walk object code, but gets the ip from FUP and TIP
packets.
Example:
$ perf record -e intel_pt//u grep -rI pudding drivers
[ perf record: Woken up 52 times to write data ]
[ perf record: Captured and wrote 57.870 MB perf.data ]
$ time perf script --itrace=bi | wc -l
58948289
real 1m23.863s
user 1m23.251s
sys 0m7.452s
$ time perf script --itrace=biq | wc -l
3385694
real 0m4.453s
user 0m4.455s
sys 0m0.328s
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Reviewed-by: Andi Kleen <ak@linux.intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lore.kernel.org/lkml/20200710151104.15137-12-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The itrace "e" option may be followed by flags which affect what errors
will or will not be reported. Each flag must be preceded by either '+' or '-'.
The flags supported by Intel PT are:
-o Suppress overflow errors
-l Suppress trace data lost errors
For example, for errors but not overflow or data lost errors:
--itrace=e-o-l
Suppressing those errors can be useful for testing and debugging because
they are not due to decoding.
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Reviewed-by: Andi Kleen <ak@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lore.kernel.org/lkml/20200710151104.15137-7-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Select text poke events when available and the kernel is being traced.
Process text poke events to invalidate entries in Intel PT's instruction
cache.
Example:
The example requires kernel config:
CONFIG_PROC_SYSCTL=y
CONFIG_SCHED_DEBUG=y
CONFIG_SCHEDSTATS=y
Before:
# perf record -o perf.data.before --kcore -a -e intel_pt//k -m,64M &
# cat /proc/sys/kernel/sched_schedstats
0
# echo 1 > /proc/sys/kernel/sched_schedstats
# cat /proc/sys/kernel/sched_schedstats
1
# echo 0 > /proc/sys/kernel/sched_schedstats
# cat /proc/sys/kernel/sched_schedstats
0
# kill %1
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 3.341 MB perf.data.before ]
[1]+ Terminated perf record -o perf.data.before --kcore -a -e intel_pt//k -m,64M
# perf script -i perf.data.before --itrace=e >/dev/null
Warning:
474 instruction trace errors
After:
# perf record -o perf.data.after --kcore -a -e intel_pt//k -m,64M &
# cat /proc/sys/kernel/sched_schedstats
0
# echo 1 > /proc/sys/kernel/sched_schedstats
# cat /proc/sys/kernel/sched_schedstats
1
# echo 0 > /proc/sys/kernel/sched_schedstats
# cat /proc/sys/kernel/sched_schedstats
0
# kill %1
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 2.646 MB perf.data.after ]
[1]+ Terminated perf record -o perf.data.after --kcore -a -e intel_pt//k -m,64M
# perf script -i perf.data.after --itrace=e >/dev/null
Example:
The example requires kernel config:
# CONFIG_FUNCTION_TRACER is not set
Before:
# perf record --kcore -m,64M -o t1 -a -e intel_pt//k &
# perf probe __schedule
Added new event:
probe:__schedule (on __schedule)
You can now use it in all perf tools, such as:
perf record -e probe:__schedule -aR sleep 1
# perf record -e probe:__schedule -aR sleep 1
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.026 MB perf.data (68 samples) ]
# perf probe -d probe:__schedule
Removed event: probe:__schedule
# kill %1
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 41.268 MB t1 ]
[1]+ Terminated perf record --kcore -m,64M -o t1 -a -e intel_pt//k
# perf script -i t1 --itrace=e >/dev/null
Warning:
207 instruction trace errors
After:
# perf record --kcore -m,64M -o t1 -a -e intel_pt//k &
# perf probe __schedule
Added new event:
probe:__schedule (on __schedule)
You can now use it in all perf tools, such as:
perf record -e probe:__schedule -aR sleep 1
# perf record -e probe:__schedule -aR sleep 1
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.028 MB perf.data (107 samples) ]
# perf probe -d probe:__schedule
Removed event: probe:__schedule
# kill %1
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 39.978 MB t1 ]
[1]+ Terminated perf record --kcore -m,64M -o t1 -a -e intel_pt//k
# perf script -i t1 --itrace=e >/dev/null
# perf script -i t1 --no-itrace -D | grep 'POKE\|KSYMBOL'
6 565303693547 0x291f18 [0x50]: PERF_RECORD_KSYMBOL addr ffffffffc027a000 len 4096 type 2 flags 0x0 name kprobe_insn_page
6 565303697010 0x291f68 [0x40]: PERF_RECORD_TEXT_POKE addr 0xffffffffc027a000 old len 0 new len 6
6 565303838278 0x291fa8 [0x50]: PERF_RECORD_KSYMBOL addr ffffffffc027c000 len 4096 type 2 flags 0x0 name kprobe_optinsn_page
6 565303848286 0x291ff8 [0xa0]: PERF_RECORD_TEXT_POKE addr 0xffffffffc027c000 old len 0 new len 106
6 565369336743 0x292af8 [0x40]: PERF_RECORD_TEXT_POKE addr 0xffffffff88ab8890 old len 5 new len 5
7 566434327704 0x217c208 [0x40]: PERF_RECORD_TEXT_POKE addr 0xffffffff88ab8890 old len 5 new len 5
6 566456313475 0x293198 [0xa0]: PERF_RECORD_TEXT_POKE addr 0xffffffffc027c000 old len 106 new len 0
6 566456314935 0x293238 [0x40]: PERF_RECORD_TEXT_POKE addr 0xffffffffc027a000 old len 6 new len 0
Example:
The example requires kernel config:
CONFIG_FUNCTION_TRACER=y
Before:
# perf record --kcore -m,64M -o t1 -a -e intel_pt//k &
# perf probe __kmalloc
Added new event:
probe:__kmalloc (on __kmalloc)
You can now use it in all perf tools, such as:
perf record -e probe:__kmalloc -aR sleep 1
# perf record -e probe:__kmalloc -aR sleep 1
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.022 MB perf.data (6 samples) ]
# perf probe -d probe:__kmalloc
Removed event: probe:__kmalloc
# kill %1
[ perf record: Woken up 2 times to write data ]
[ perf record: Captured and wrote 43.850 MB t1 ]
[1]+ Terminated perf record --kcore -m,64M -o t1 -a -e intel_pt//k
# perf script -i t1 --itrace=e >/dev/null
Warning:
8 instruction trace errors
After:
# perf record --kcore -m,64M -o t1 -a -e intel_pt//k &
# perf probe __kmalloc
Added new event:
probe:__kmalloc (on __kmalloc)
You can now use it in all perf tools, such as:
perf record -e probe:__kmalloc -aR sleep 1
# perf record -e probe:__kmalloc -aR sleep 1
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.037 MB perf.data (206 samples) ]
# perf probe -d probe:__kmalloc
Removed event: probe:__kmalloc
# kill %1
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 41.442 MB t1 ]
[1]+ Terminated perf record --kcore -m,64M -o t1 -a -e intel_pt//k
# perf script -i t1 --itrace=e >/dev/null
# perf script -i t1 --no-itrace -D | grep 'POKE\|KSYMBOL'
5 312216133258 0x8bafe0 [0x50]: PERF_RECORD_KSYMBOL addr ffffffffc0360000 len 415 type 2 flags 0x0 name ftrace_trampoline
5 312216133494 0x8bb030 [0x1d8]: PERF_RECORD_TEXT_POKE addr 0xffffffffc0360000 old len 0 new len 415
5 312216229563 0x8bb208 [0x40]: PERF_RECORD_TEXT_POKE addr 0xffffffffac6016f5 old len 5 new len 5
5 312216239063 0x8bb248 [0x40]: PERF_RECORD_TEXT_POKE addr 0xffffffffac601803 old len 5 new len 5
5 312216727230 0x8bb288 [0x40]: PERF_RECORD_TEXT_POKE addr 0xffffffffabbea190 old len 5 new len 5
5 312216739322 0x8bb2c8 [0x40]: PERF_RECORD_TEXT_POKE addr 0xffffffffac6016f5 old len 5 new len 5
5 312216748321 0x8bb308 [0x40]: PERF_RECORD_TEXT_POKE addr 0xffffffffac601803 old len 5 new len 5
7 313287163462 0x2817430 [0x40]: PERF_RECORD_TEXT_POKE addr 0xffffffffac6016f5 old len 5 new len 5
7 313287174890 0x2817470 [0x40]: PERF_RECORD_TEXT_POKE addr 0xffffffffac601803 old len 5 new len 5
7 313287818979 0x28174b0 [0x40]: PERF_RECORD_TEXT_POKE addr 0xffffffffabbea190 old len 5 new len 5
7 313287829357 0x28174f0 [0x40]: PERF_RECORD_TEXT_POKE addr 0xffffffffac6016f5 old len 5 new len 5
7 313287841246 0x2817530 [0x40]: PERF_RECORD_TEXT_POKE addr 0xffffffffac601803 old len 5 new len 5
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Steven Rostedt (VMware) <rostedt@goodmis.org>
Cc: x86@kernel.org
Link: http://lore.kernel.org/lkml/20200512121922.8997-14-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
