[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <1ff39c81-f207-480a-8169-85b273631594@intel.com>
Date: Mon, 5 Feb 2024 15:10:58 +0200
From: Adrian Hunter <adrian.hunter@...el.com>
To: Changbin Du <changbin.du@...wei.com>
Cc: Peter Zijlstra <peterz@...radead.org>, Ingo Molnar <mingo@...hat.com>,
Arnaldo Carvalho de Melo <acme@...nel.org>,
Mark Rutland <mark.rutland@....com>,
Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
Jiri Olsa <jolsa@...nel.org>, Namhyung Kim <namhyung@...nel.org>,
Ian Rogers <irogers@...gle.com>, linux-kernel@...r.kernel.org,
linux-perf-users@...r.kernel.org, Andi Kleen <ak@...ux.intel.com>,
Thomas Richter <tmricht@...ux.ibm.com>, changbin.du@...il.com
Subject: Re: [PATCH v5 3/5] perf: script: add field 'disasm' to display
mnemonic instructions
On 5/02/24 14:19, Changbin Du wrote:
> On Mon, Feb 05, 2024 at 11:23:21AM +0200, Adrian Hunter wrote:
>> On 22/01/24 13:20, Changbin Du wrote:
>>> In addition to the 'insn' field, this adds a new field 'disasm' to
>>> display mnemonic instructions instead of the raw code.
>>>
>>> $ sudo perf script -F +disasm
>>> perf-exec 1443864 [006] 2275506.209848: psb: psb offs: 0 0 [unknown] ([unknown])
>>> perf-exec 1443864 [006] 2275506.209848: cbr: cbr: 41 freq: 4100 MHz (114%) 0 [unknown] ([unknown])
>>> ls 1443864 [006] 2275506.209905: 1 branches:uH: 7f216b426100 _start+0x0 (/usr/lib/x86_64-linux-gnu/ld-2.31.so) movq %rsp, %rdi
>>> ls 1443864 [006] 2275506.209908: 1 branches:uH: 7f216b426103 _start+0x3 (/usr/lib/x86_64-linux-gnu/ld-2.31.so) callq _dl_start+0x0
>>>
>>> Signed-off-by: Changbin Du <changbin.du@...wei.com>
>>>
>>> ---
>>> v2:
>>> - update Documentation.
>>> ---
>>> tools/perf/Documentation/perf-script.txt | 13 +++++++------
>>> tools/perf/builtin-script.c | 8 +++++++-
>>> 2 files changed, 14 insertions(+), 7 deletions(-)
>>>
>>> diff --git a/tools/perf/Documentation/perf-script.txt b/tools/perf/Documentation/perf-script.txt
>>> index ff9a52e44688..578fa59f51a5 100644
>>> --- a/tools/perf/Documentation/perf-script.txt
>>> +++ b/tools/perf/Documentation/perf-script.txt
>>> @@ -132,9 +132,10 @@ OPTIONS
>>> Comma separated list of fields to print. Options are:
>>> comm, tid, pid, time, cpu, event, trace, ip, sym, dso, dsoff, addr, symoff,
>>> srcline, period, iregs, uregs, brstack, brstacksym, flags, bpf-output,
>>> - brstackinsn, brstackinsnlen, brstackoff, callindent, insn, insnlen, synth,
>>> - phys_addr, metric, misc, srccode, ipc, data_page_size, code_page_size, ins_lat,
>>> - machine_pid, vcpu, cgroup, retire_lat.
>>> + brstackinsn, brstackinsnlen, brstackoff, callindent, insn, disasm,
>>> + insnlen, synth, phys_addr, metric, misc, srccode, ipc, data_page_size,
>>> + code_page_size, ins_lat, machine_pid, vcpu, cgroup, retire_lat.
>>> +
>>> Field list can be prepended with the type, trace, sw or hw,
>>> to indicate to which event type the field list applies.
>>> e.g., -F sw:comm,tid,time,ip,sym and -F trace:time,cpu,trace
>>> @@ -217,9 +218,9 @@ OPTIONS
>>> Instruction Trace decoding. For calls and returns, it will display the
>>> name of the symbol indented with spaces to reflect the stack depth.
>>>
>>> - When doing instruction trace decoding insn and insnlen give the
>>> - instruction bytes and the instruction length of the current
>>> - instruction.
>>> + When doing instruction trace decoding, insn, disasm and insnlen give the
>>> + instruction bytes, disassembled instructions (requires libcapstone support)
>>> + and the instruction length of the current instruction respectively.
>>>
>>> The synth field is used by synthesized events which may be created when
>>> Instruction Trace decoding.
>>> diff --git a/tools/perf/builtin-script.c b/tools/perf/builtin-script.c
>>> index 4817a37f16e2..4ac9670704ff 100644
>>> --- a/tools/perf/builtin-script.c
>>> +++ b/tools/perf/builtin-script.c
>>> @@ -135,6 +135,7 @@ enum perf_output_field {
>>> PERF_OUTPUT_CGROUP = 1ULL << 39,
>>> PERF_OUTPUT_RETIRE_LAT = 1ULL << 40,
>>> PERF_OUTPUT_DSOFF = 1ULL << 41,
>>> + PERF_OUTPUT_DISASM = 1ULL << 42,
>>> };
>>>
>>> struct perf_script {
>>> @@ -190,6 +191,7 @@ struct output_option {
>>> {.str = "bpf-output", .field = PERF_OUTPUT_BPF_OUTPUT},
>>> {.str = "callindent", .field = PERF_OUTPUT_CALLINDENT},
>>> {.str = "insn", .field = PERF_OUTPUT_INSN},
>>> + {.str = "disasm", .field = PERF_OUTPUT_DISASM},
>>> {.str = "insnlen", .field = PERF_OUTPUT_INSNLEN},
>>> {.str = "brstackinsn", .field = PERF_OUTPUT_BRSTACKINSN},
>>> {.str = "brstackoff", .field = PERF_OUTPUT_BRSTACKOFF},
>>> @@ -1515,6 +1517,10 @@ static int perf_sample__fprintf_insn(struct perf_sample *sample,
>>> printed += fprintf(fp, " insn: ");
>>> printed += sample__fprintf_insn_raw(sample, fp);
>>> }
>>> + if (PRINT_FIELD(DISASM) && sample->insn_len) {
>>> + printed += fprintf(fp, "\t\t");
>>
>> This is good, except if both 'insn' and 'disasm' are used together.
>> It either:
>> a) without libcapstone, prints insn bytes twice
>>
>> Probably simpler to make 'disasm' without libcapstone
>> a fatal error explaining that perf needs to be built
>> with libcapstone support for 'disasm' to work.
>>
> Instead of fatal error, I print a warning message for this. Because
> perf_sample__fprintf_insn() cannot return negtive error number.
It could be validated in advance, perhaps in parse_output_fields().
>
>> b) with libcapstone, disassembly does not line up nicely
>>
>
>
Powered by blists - more mailing lists