| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <0102019acc781e84-04b7506a-b3d0-4290-a2dd-5c8e6a8a1f5b-000000@eu-west-1.amazonses.com>
Date: Fri, 28 Nov 2025 21:56:55 +0000
From: Derek Foreman <derek.foreman@...labora.com>
To: Namhyung Kim <namhyung@...nel.org>
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>, Ian Rogers <irogers@...gle.com>,
Adrian Hunter <adrian.hunter@...el.com>,
James Clark <james.clark@...aro.org>, kernel@...labora.com,
linux-perf-users@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] perf data: Allow filtering conversion by start and end
time
On 11/27/25 3:34 PM, Namhyung Kim wrote:
> Hello,
>
> On Mon, Nov 24, 2025 at 03:23:00PM -0600, Derek Foreman wrote:
>> This adds a feature to allow restricting the range of converted samples
>> with start and end times specified in nanoseconds.
>>
>> Use it like this:
>>
>> $ perf data convert --to-json out.json -s 12345 -e 13340
>>
>> Start, end, or both may be omitted.
> Looks useful, thanks!
Thanks for the review!
>> Signed-off-by: Derek Foreman <derek.foreman@...labora.com>
>> ---
>> tools/perf/Documentation/perf-data.txt | 6 ++++++
>> tools/perf/builtin-data.c | 4 ++++
>> tools/perf/util/data-convert-bt.c | 18 ++++++++++++++++++
>> tools/perf/util/data-convert-json.c | 18 ++++++++++++++++++
>> tools/perf/util/data-convert.h | 3 +++
>> 5 files changed, 49 insertions(+)
>>
>> diff --git a/tools/perf/Documentation/perf-data.txt b/tools/perf/Documentation/perf-data.txt
>> index 417bf17e265c..db0de1f354a3 100644
>> --- a/tools/perf/Documentation/perf-data.txt
>> +++ b/tools/perf/Documentation/perf-data.txt
>> @@ -40,6 +40,12 @@ OPTIONS for 'convert'
>> --force::
>> Don't complain, do it.
>>
>> +-s::
> Please add this as well.
>
> --start=<TIME>::
>
>> + Start time, in nanoseconds. Samples before this time will not be converted.
> It'd be nice if you mention that users can see the timestamp from perf
> script or so.
After looking at perf script, I realized I'd reinvented the wheel here,
so in v2 I've used `--time` and the util functions that already exist
for parsing and filtering time strings.
I mostly copy and pasted the --time help from another file. I think
asciidoc has a way to templatize this with an include file (it's in at
least 3 files now), but I wasn't sure if I should try to do that.
>
>> +
>> +-e::
> --end=<TIME>::
>
>> + End time, in nanoseconds. Samples after this time will not be converted.
>> +
>> -v::
>> --verbose::
>> Be more verbose (show counter open errors, etc).
>> diff --git a/tools/perf/builtin-data.c b/tools/perf/builtin-data.c
>> index ce51cbf6dc97..da813c62e284 100644
>> --- a/tools/perf/builtin-data.c
>> +++ b/tools/perf/builtin-data.c
>> @@ -33,6 +33,8 @@ const char *to_ctf;
>> struct perf_data_convert_opts opts = {
>> .force = false,
>> .all = false,
>> + .range_start = 0,
>> + .range_end = 0,
>> };
>>
>> const struct option data_options[] = {
>> @@ -45,6 +47,8 @@ const struct option data_options[] = {
>> #endif
>> OPT_BOOLEAN('f', "force", &opts.force, "don't complain, do it"),
>> OPT_BOOLEAN(0, "all", &opts.all, "Convert all events"),
>> + OPT_U64('s', "start", &opts.range_start, "Earliest timestamp to convert"),
>> + OPT_U64('e', "end", &opts.range_end, "Latest timestamp to convert"),
>> OPT_END()
>> };
>>
>> diff --git a/tools/perf/util/data-convert-bt.c b/tools/perf/util/data-convert-bt.c
>> index 3d2e437e1354..6819f11fe900 100644
>> --- a/tools/perf/util/data-convert-bt.c
>> +++ b/tools/perf/util/data-convert-bt.c
>> @@ -91,9 +91,13 @@ struct convert {
>> struct perf_tool tool;
>> struct ctf_writer writer;
>>
>> + u64 start;
>> + u64 end;
>> +
>> u64 events_size;
>> u64 events_count;
>> u64 non_sample_count;
>> + u64 skipped;
>>
>> /* Ordered events configured queue size. */
>> u64 queue_size;
>> @@ -811,6 +815,12 @@ static int process_sample_event(const struct perf_tool *tool,
>> if (WARN_ONCE(!priv, "Failed to setup all events.\n"))
>> return 0;
>>
>> + if (sample->time < c->start ||
>> + (c->end && sample->time > c->end)) {
>> + ++c->skipped;
>> + return 0;
>> + }
>> +
>> event_class = priv->event_class;
>>
>> /* update stats */
>> @@ -1626,6 +1636,9 @@ int bt_convert__perf2ctf(const char *input, const char *path,
>> c.tool.namespaces = perf_event__process_namespaces;
>> c.tool.ordering_requires_timestamps = true;
>>
>> + c.start = opts->range_start;
>> + c.end = opts->range_end;
>> +
>> if (opts->all) {
>> c.tool.comm = process_comm_event;
>> c.tool.exit = process_exit_event;
>> @@ -1677,6 +1690,11 @@ int bt_convert__perf2ctf(const char *input, const char *path,
>> "[ perf data convert: Converted '%s' into CTF data '%s' ]\n",
>> data.path, path);
>>
>> + if (c.skipped)
>> + fprintf(stderr,
>> + "[ perf data convert: Skipped %" PRIu64 " samples",
>> + c.skipped);
> This is a strange style to have the literal in a different line. It'd
> be nice if you can fix others too - probably in a separate commit.
>
> And this line could come after the next, but it's up to you.
I think it's better your way, yes.
> Also it'd be recommended to have a pair of parentheses for multi-line
> bodies even it's a single statement.
I've added a second patch for v2. One of the lines exceeded 100 chars,
which bothered checkpatch.pl, so I left the literal on the second line
for now.
Thanks,
Derek
>> +
>> fprintf(stderr,
>> "[ perf data convert: Converted and wrote %.3f MB (%" PRIu64 " samples",
>> (double) c.events_size / 1024.0 / 1024.0,
>> diff --git a/tools/perf/util/data-convert-json.c b/tools/perf/util/data-convert-json.c
>> index 9dc1e184cf3c..dcbba34d6b08 100644
>> --- a/tools/perf/util/data-convert-json.c
>> +++ b/tools/perf/util/data-convert-json.c
>> @@ -35,7 +35,11 @@ struct convert_json {
>> struct perf_tool tool;
>> FILE *out;
>> bool first;
>> + u64 start;
>> + u64 end;
>> +
>> u64 events_count;
>> + u64 skipped;
>> };
>>
>> // Outputs a JSON-encoded string surrounded by quotes with characters escaped.
>> @@ -165,6 +169,12 @@ static int process_sample_event(const struct perf_tool *tool,
>> return -1;
>> }
>>
>> + if (sample->time < c->start ||
>> + (c->end && sample->time > c->end)) {
>> + ++c->skipped;
>> + return 0;
>> + }
>> +
>> ++c->events_count;
>>
>> if (c->first)
>> @@ -320,6 +330,9 @@ int bt_convert__perf2json(const char *input_name, const char *output_name,
>> struct convert_json c = {
>> .first = true,
>> .events_count = 0,
>> + .start = opts->range_start,
>> + .end = opts->range_end,
>> + .skipped = 0,
>> };
>> struct perf_data data = {
>> .mode = PERF_DATA_MODE_READ,
>> @@ -407,6 +420,11 @@ int bt_convert__perf2json(const char *input_name, const char *output_name,
>> "[ perf data convert: Converted '%s' into JSON data '%s' ]\n",
>> data.path, output_name);
>>
>> + if (c.skipped)
>> + fprintf(stderr,
>> + "[ perf data convert: Skipped %" PRIu64 " samples.\n",
>> + c.skipped);
> Ditto.
>
> Thanks,
> Namhyung
>
>> +
>> fprintf(stderr,
>> "[ perf data convert: Converted and wrote %.3f MB (%" PRIu64 " samples) ]\n",
>> (ftell(c.out)) / 1024.0 / 1024.0, c.events_count);
>> diff --git a/tools/perf/util/data-convert.h b/tools/perf/util/data-convert.h
>> index 1b4c5f598415..5d055e1c31c8 100644
>> --- a/tools/perf/util/data-convert.h
>> +++ b/tools/perf/util/data-convert.h
>> @@ -3,11 +3,14 @@
>> #define __DATA_CONVERT_H
>>
>> #include <stdbool.h>
>> +#include <linux/types.h>
>>
>> struct perf_data_convert_opts {
>> bool force;
>> bool all;
>> bool tod;
>> + u64 range_start;
>> + u64 range_end;
>> };
>>
>> #ifdef HAVE_LIBBABELTRACE_SUPPORT
>> --
>> 2.47.3
>>
Powered by blists - more mailing lists