Message-ID: <5767BEC0.5030806@gmail.com>
Date:	Mon, 20 Jun 2016 19:00:32 +0900
From:	Taeung Song <treeze.taeung@...il.com>
To:	Arnaldo Carvalho de Melo <arnaldo.melo@...il.com>
Cc:	linux-kernel@...r.kernel.org, Jiri Olsa <jolsa@...nel.org>,
	Namhyung Kim <namhyung@...nel.org>,
	Ingo Molnar <mingo@...nel.org>,
	Peter Zijlstra <peterz@...radead.org>,
	Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
	Masami Hiramatsu <mhiramat@...nel.org>,
	Wang Nan <wangnan0@...wei.com>, Jiri Olsa <jolsa@...hat.com>,
	Ingo Molnar <mingo@...hat.com>
Subject: Re: [PATCH v8 3/5] perf config: Reimplement perf_config() using
 perf_config_set__it to

Hi, Arnaldo :-D

I also discussed this patchset with Namhyung at offline gathering.
I sent v9 patchset that contains what Namhyung and you advice me to do.
If having your spare time, please review the v9 patchset.
If you do, I'd appreciate it :-)

Thanks,
Taeung

On 06/12/2016 05:57 PM, Taeung Song wrote:
> Hi, Arnaldo
>
> What do you think about setting all actual config variables
> before a sub-command start at perf.c ?
>
> (in order to free the config set that contains all configs
> by perf_config_set__delete() after using it was finished where it was
> needed)
>
> i.e. initialize all actual config variables at perf.c
> instead of at each source file as below.
> (before running a particular sub-command at run_builitin())
>
> util/alias.c
> 22:    perf_config(alias_lookup_cb, NULL);
>
> util/intel-pt.c
> 330:    perf_config(intel_pt_config_div, &d);
> 1993:static int intel_pt_perf_config(const char *var, const char *value,
> void *data)
> 2047:    perf_config(intel_pt_perf_config, pt);
>
> util/data-convert-bt.c
> 1302:    perf_config(convert__config, &c);
>
> util/help-unknown-cmd.c
> 62:    perf_config(perf_unknown_cmd_config, NULL);
>
> builtin-help.c
> 454:    perf_config(perf_help_config, &help_format);
>
> perf.c
> 94:    perf_config(pager_command_config, &c);
> 117:    perf_config(browser_command_config, &c);
> 561:    perf_config(perf_default_config, NULL);
>
> builtin-record.c
> 1432:    perf_config(perf_record_config, rec);
>
> ui/browsers/annotate.c
> 1163:    perf_config(annotate__config, NULL);
>
> ui/browser.c
> 743:    perf_config(ui_browser__color_config, NULL);
>
> builtin-report.c
> 829:    perf_config(report__config, &report);
>
> builtin-kmem.c
> 1899:    perf_config(kmem_config, NULL);
>
> builtin-top.c
> 1231:    perf_config(perf_top_config, &top);
>
>
> If we do, we could set all actual config variables
> before a sub-command work. And the config set that is made
> by perf_config_set__new() would be reused
> when we initialize all actual config variables
> and then if using is is finished, we could free the config set.
>
>
> Thanks,
> Taeung
>
>
> On 06/10/2016 07:58 PM, Taeung Song wrote:
>>
>>
>> On 06/09/2016 10:34 PM, Arnaldo Carvalho de Melo wrote:
>>> Em Wed, Jun 08, 2016 at 09:36:51PM +0900, Taeung Song escreveu:
>>>> Many sub-commands use perf_config() so
>>>> everytime perf_config() is called, perf_config() always read config
>>>> files.
>>>> (i.e. user config '~/.perfconfig' and system config
>>>> '$(sysconfdir)/perfconfig')
>>>
>>> And this is not always a bad thing, think about being in 'mutt' and
>>> adding an entry to ~/.mail_aliases, then going to compose a message,
>>> would be good that the just added entry to ~/.mail_aliases be considered
>>> when adding recipients to your messages, right?
>>>
>>> In the same fashion, 'perf report', 'perf annotate', 'perf top' are long
>>> running utilities that have operations that could get changes to config
>>> files without having to restart them, i.e. do annotation changes in your
>>> ~/.perfconfig and then see them reflected next time you hit 'A´ to
>>> annotate a function in 'perf top' or 'perf report'.
>>>
>>
>> Currently, this suggestion isn't already contained among features of
>> perf. right?
>>
>> I understood that you mean while perf process is already running
>> if user change a config file, the changed config can be reflected in
>> 'perf top', 'perf report' or etc without restarting perf process like
>> mutt.
>> Is it right ?
>>
>>> So, I think that if tools want ammortize the cost of reading config
>>> files, they should create an instance of the relevant object
>>> (perf_config_set?) use it and then delete it, but not keep one around
>>> for a long time.
>>>
>>
>> I got it.
>>
>> How about the two ideas that I have in mind ?
>>
>> 1) If we eliminate the config set object(perf_config_set) after using it
>> because we don't need to keep it until perf process is done,
>>
>> we could bring many codes using perf_config() at one spot of perf.c ?
>> (because it is hard to decide a point of time we destroy the config set.)
>> For example,
>>
>> (This code isn't executable)
>>
>> diff --git a/tools/perf/perf.c b/tools/perf/perf.c
>> index 15982ce..6a56985 100644
>> --- a/tools/perf/perf.c
>> +++ b/tools/perf/perf.c
>> @@ -77,6 +77,23 @@ struct pager_config {
>>           int val;
>>   };
>>
>> +static void perf_default_config_init(void)
>> +{
>> +        default_colors_config_init();
>> +        default_annoate_config_init();
>> +        default_report_config_init();
>> +        ...
>> +}
>> +
>> +static void perf_config_init(void)
>> +{
>> +        perf_default_config_init();
>> +        colors_config_init();
>> +        annotate_config_init();
>> +        report_config_init();
>> +        ...
>> +}
>> +
>>   static int pager_command_config(const char *var, const char *value,
>> void *data)
>>   {
>>           struct pager_config *c = data;
>> @@ -558,7 +575,7 @@ int main(int argc, const char **argv)
>>
>>           srandom(time(NULL));
>>
>> -        perf_config(perf_default_config, NULL);
>> +        perf_config_init();
>>           set_buildid_dir(NULL);
>>
>>           /* get debugfs/tracefs mount point from /proc/mounts */
>> diff --git a/tools/perf/ui/browser.c b/tools/perf/ui/browser.c
>> index af68a9d..4b827c2 100644
>> --- a/tools/perf/ui/browser.c
>> +++ b/tools/perf/ui/browser.c
>> @@ -740,8 +740,6 @@ void ui_browser__init(void)
>>   {
>>           int i = 0;
>>
>> -        perf_config(ui_browser__color_config, NULL);
>> -
>>           while (ui_browser__colorsets[i].name) {
>>                   struct ui_browser_colorset *c =
>> &ui_browser__colorsets[i++];
>>                   sltt_set_color(c->colorset, c->name, c->fg, c->bg);
>>
>> ... (omitted)
>>
>>
>> Many sub-commands call perf_config() at builtin-report.c, ui/browser.c
>> and etc. So I think it is ambiguous to decide where
>> perf_config_set__delete()
>> will be called instead of at run_builtin().
>>
>> (If the config set's life cycle is the same as perf's ,
>> I would call perf_config_set__delete() at run_builtin()
>> because a sub-command is done)
>>
>>
>> 2) If having the config set for perf's life,
>>
>> we would add configuration features to TUI like TUI configuring options
>> for linux kernel compiling (e.g. make menuconfig) ?
>>
>> Of course, we don't need to have the same function as make menuconfig
>> but IMHO, we could add similar way to menuconfig while perf is running.
>>
>>
>> How do you feel about these ?
>>
>>
>> Thanks, :)
>> Taeung
>>
>>>
>>>> But we need to use the config set that already contains all config
>>>> key-value pairs to avoid this repetitive work reading the config files
>>>> in perf_config(). (the config set mean a global variable 'config_set')
>>>>
>>>> In other words, if new perf_config() is called,
>>>> only first time 'config_set' is initialized collecting all configs
>>>> from the config files and it work with perf_config_set__iter().
>>>> And free the config set after a sub-command work at run_builtin().
>>>>
>>>> If we do, 'config_set' can be reused wherever using perf_config()
>>>> and a feature of old perf_config() is the same as new perf_config()
>>>> work
>>>> without the repetitive work that read the config files.
>>>>
>>>> Cc: Namhyung Kim <namhyung@...nel.org>
>>>> Cc: Jiri Olsa <jolsa@...hat.com>
>>>> Cc: Wang Nan <wangnan0@...wei.com>
>>>> Cc: Peter Zijlstra <peterz@...radead.org>
>>>> Cc: Ingo Molnar <mingo@...hat.com>
>>>> Cc: Masami Hiramatsu <mhiramat@...nel.org>
>>>> Cc: Alexander Shishkin <alexander.shishkin@...ux.intel.com>
>>>> Signed-off-by: Taeung Song <treeze.taeung@...il.com>
>>>> ---
>>>>   tools/perf/builtin-config.c |  4 +++
>>>>   tools/perf/perf.c           |  1 +
>>>>   tools/perf/util/config.c    | 87
>>>> ++++++++++++++++++++++-----------------------
>>>>   tools/perf/util/config.h    |  1 +
>>>>   4 files changed, 48 insertions(+), 45 deletions(-)
>>>>
>>>> diff --git a/tools/perf/builtin-config.c b/tools/perf/builtin-config.c
>>>> index fe1b77f..cfd1036 100644
>>>> --- a/tools/perf/builtin-config.c
>>>> +++ b/tools/perf/builtin-config.c
>>>> @@ -80,6 +80,10 @@ int cmd_config(int argc, const char **argv, const
>>>> char *prefix __maybe_unused)
>>>>       else if (use_user_config)
>>>>           config_exclusive_filename = user_config;
>>>>
>>>> +    /*
>>>> +     * At only 'config' sub-command, individually use the config set
>>>> +     * because of reinitializing with options config file location.
>>>> +     */
>>>>       set = perf_config_set__new();
>>>>       if (!set) {
>>>>           ret = -1;
>>>> diff --git a/tools/perf/perf.c b/tools/perf/perf.c
>>>> index 15982ce..fe2ab7c 100644
>>>> --- a/tools/perf/perf.c> - Arnaldo
>>>> +++ b/tools/perf/perf.c
>>>> @@ -391,6 +391,7 @@ static int run_builtin(struct cmd_struct *p, int
>>>> argc, const char **argv)
>>>>
>>>>       perf_env__set_cmdline(&perf_env, argc, argv);
>>>>       status = p->fn(argc, argv, prefix);
>>>> +    perf_config_set__delete(config_set);
>>>>       exit_browser(status);
>>>>       perf_env__exit(&perf_env);
>>>>       bpf__clear();
>>>> diff --git a/tools/perf/util/config.c b/tools/perf/util/config.c
>>>> index 31e09a4..72db134 100644
>>>> --- a/tools/perf/util/config.c
>>>> +++ b/tools/perf/util/config.c
>>>> @@ -28,6 +28,7 @@ static int config_linenr;
>>>>   static int config_file_eof;
>>>>
>>>>   const char *config_exclusive_filename;
>>>> +struct perf_config_set *config_set;
>>>>
>>>>   static int get_next_char(void)
>>>>   {
>>>> @@ -478,51 +479,6 @@ static int perf_config_global(void)
>>>>       return !perf_env_bool("PERF_CONFIG_NOGLOBAL", 0);
>>>>   }
>>>>
>>>> -int perf_config(config_fn_t fn, void *data)
>>>> -{
>>>> -    int ret = -1;
>>>> -    const char *home = NULL;
>>>> -
>>>> -    /* Setting $PERF_CONFIG makes perf read _only_ the given config
>>>> file. */
>>>> -    if (config_exclusive_filename)
>>>> -        return perf_config_from_file(fn, config_exclusive_filename,
>>>> data);
>>>> -    if (perf_config_system() && !access(perf_etc_perfconfig(),
>>>> R_OK)) {
>>>> -        if (perf_config_from_file(fn, perf_etc_perfconfig(), data)
>>>> < 0)
>>>> -            goto out;
>>>> -    }
>>>> -
>>>> -    home = getenv("HOME");
>>>> -    if (perf_config_global() && home) {
>>>> -        char *user_config = strdup(mkpath("%s/.perfconfig", home));
>>>> -        struct stat st;
>>>> -
>>>> -        if (user_config == NULL) {
>>>> -            warning("Not enough memory to process %s/.perfconfig, "
>>>> -                "ignoring it.", home);
>>>> -            goto out;
>>>> -        }
>>>> -
>>>> -        if (stat(user_config, &st) < 0)
>>>> -            goto out_free;
>>>> -
>>>> -        if (st.st_uid && (st.st_uid != geteuid())) {
>>>> -            warning("File %s not owned by current user or root, "
>>>> -                "ignoring it.", user_config);
>>>> -            goto out_free;
>>>> -        }
>>>> -
>>>> -        if (!st.st_size)
>>>> -            goto out_free;
>>>> -
>>>> -        ret = perf_config_from_file(fn, user_config, data);
>>>> -
>>>> -out_free:
>>>> -        free(user_config);
>>>> -    }
>>>> -out:
>>>> -    return ret;
>>>> -}
>>>> -
>>>>   static struct perf_config_section *find_section(struct list_head
>>>> *sections,
>>>>                           const char *section_name)
>>>>   {
>>>> @@ -706,6 +662,47 @@ struct perf_config_set *perf_config_set__new(void)
>>>>       return set;
>>>>   }
>>>>
>>>> +static int perf_config_set__iter(struct perf_config_set *set,
>>>> config_fn_t fn, void *data)
>>>> +{
>>>> +    struct perf_config_section *section;
>>>> +    struct perf_config_item *item;
>>>> +    struct list_head *sections;
>>>> +    char key[BUFSIZ];
>>>> +
>>>> +    if (set == NULL)
>>>> +        return -1;
>>>> +
>>>> +    sections = &set->sections;
>>>> +    if (list_empty(sections))
>>>> +        return -1;
>>>> +
>>>> +    list_for_each_entry(section, sections, node) {
>>>> +        list_for_each_entry(item, &section->items, node) {
>>>> +            char *value = item->value;
>>>> +
>>>> +            if (value) {
>>>> +                scnprintf(key, sizeof(key), "%s.%s",
>>>> +                      section->name, item->name);
>>>> +                if (fn(key, value, data) < 0) {
>>>> +                    pr_err("Error: wrong config key-value pair
>>>> %s=%s\n",
>>>> +                           key, value);
>>>> +                    return -1;
>>>> +                }
>>>> +            }
>>>> +        }
>>>> +    }
>>>> +
>>>> +    return 0;
>>>> +}
>>>> +
>>>> +int perf_config(config_fn_t fn, void *data)
>>>> +{
>>>> +    if (config_set == NULL)
>>>> +        config_set = perf_config_set__new();
>>>> +
>>>> +    return perf_config_set__iter(config_set, fn, data);
>>>> +}
>>>> +
>>>>   static void perf_config_item__delete(struct perf_config_item *item)
>>>>   {
>>>>       zfree(&item->name);
>>>> diff --git a/tools/perf/util/config.h b/tools/perf/util/config.h
>>>> index 35ccddb..7cc4fea 100644
>>>> --- a/tools/perf/util/config.h
>>>> +++ b/tools/perf/util/config.h
>>>> @@ -21,6 +21,7 @@ struct perf_config_set {
>>>>   };
>>>>
>>>>   extern const char *config_exclusive_filename;
>>>> +extern struct perf_config_set *config_set;
>>>>
>>>>   typedef int (*config_fn_t)(const char *, const char *, void *);
>>>>   int perf_default_config(const char *, const char *, void *);
>>>> --
>>>> 2.5.0

Powered by blists - more mailing lists