Re: [PATCH v2 1/5] perf config: Introduce perf_config_set class

[Taeung Song <treeze.taeung@gmail.com>]

On 03/18/2016 08:27 AM, Namhyung Kim wrote:
> On Thu, Mar 17, 2016 at 11:10:12PM +0900, Taeung Song wrote:
>> Hi, Namhyung
>>
>> On 03/17/2016 09:31 PM, Namhyung Kim wrote:
>>> Hi Taeung,
>>>
>>> On Mon, Mar 14, 2016 at 09:16:05PM +0900, Taeung Song wrote:
>>>> This infrastructure code was designed for
>>>> upcoming features of perf-config.
>>>>
>>>> That collect config key-value pairs from user and
>>>> system config files (i.e. user wide ~/.perfconfig
>>>> and system wide $(sysconfdir)/perfconfig)
>>>> to manage perf's configs.
>>>>
>>>> Cc: Namhyung Kim <namhyung@kernel.org>
>>>> Cc: Jiri Olsa <jolsa@kernel.org>
>>>> Signed-off-by: Taeung Song <treeze.taeung@gmail.com>
>>>> ---
>>>>   tools/perf/builtin-config.c |   1 +
>>>>   tools/perf/util/config.c    | 123 ++++++++++++++++++++++++++++++++++++++++++++
>>>>   tools/perf/util/config.h    |  21 ++++++++
>>>>   3 files changed, 145 insertions(+)
>>>>   create mode 100644 tools/perf/util/config.h
>>>>
>>>> diff --git a/tools/perf/builtin-config.c b/tools/perf/builtin-config.c
>>>> index c42448e..412c725 100644
>>>> --- a/tools/perf/builtin-config.c
>>>> +++ b/tools/perf/builtin-config.c
>>>> @@ -12,6 +12,7 @@
>>>>   #include <subcmd/parse-options.h>
>>>>   #include "util/util.h"
>>>>   #include "util/debug.h"
>>>> +#include "util/config.h"
>>>>
>>>>   static bool use_system_config, use_user_config;
>>>>
>>>> diff --git a/tools/perf/util/config.c b/tools/perf/util/config.c
>>>> index 4e72763..b9660e4 100644
>>>> --- a/tools/perf/util/config.c
>>>> +++ b/tools/perf/util/config.c
>>>> @@ -13,6 +13,7 @@
>>>>   #include <subcmd/exec-cmd.h>
>>>>   #include "util/hist.h"  /* perf_hist_config */
>>>>   #include "util/llvm-utils.h"   /* perf_llvm_config */
>>>> +#include "config.h"
>>>>
>>>>   #define MAXNAME (256)
>>>>
>>>> @@ -506,6 +507,128 @@ out:
>>>>   	return ret;
>>>>   }
>>>>
>>>> +static struct perf_config_item *find_config(struct list_head *config_list,
>>>> +				       const char *section,
>>>> +				       const char *name)
>>>> +{
>>>> +	struct perf_config_item *config;
>>>> +
>>>> +	list_for_each_entry(config, config_list, list) {
>>>> +		if (!strcmp(config->section, section) &&
>>>> +		    !strcmp(config->name, name))
>>>> +			return config;
>>>> +	}
>>>
>>> Hmm.. why do you remove the section list?
>>>
>>
>> IMHO, there are several reasons
>>
>>      1) To use only one list (default config, custom config(user/system))
>>
>> 1-1) I used two list that were 'list_head sections'
>> and 'config_item default_configs[]'. So if checking
>> type of config variable, two for-loop must be needed
>> for each list. Because two structure was different i.e.
>>
>> 'sections' list mean config_section list
>> that each section contain config_element list.
>> (there wasn't telling about correct type of 'value' instead of string(char
>> *))
>>
>>      struct config_element {
>>              char *name;
>>              char *value;
>>              struct list_head list;
>>      };
>>
>>      struct config_section {
>>              char *name;
>>              struct list_head element_head;
>>              struct list_head list;
>>      };
>>
>> 'struct config_item default_configs[]' mean all default configs.
>>
>>      struct config_item {
>>              const char *section;
>>              const char *name;
>>              union {
>>                  bool b;
>>                  int i;
>>                  u32 l;
>>                  u64 ll;
>>                  float f;
>>                  double d;
>>                  const char *s;
>>              } value;
>>              enum config_type type;
>>              const char *desc;
>>      };
>>
>>
>> IMHO, I think this is a bit complex
>> and I want to simplify the perf's config list on perf-config.
>>
>>      2) A small number of perf's configs
>>
>> I think perf's configs aren't too many so I think
>> two structure for section and element aren't needed.
>
> OK.
>
>
>>
>>      3) A object for a config variable need to have enough info for itself
>>
>> This is a bit similar to 1) reason.
>> If using only 'struct config_item' for the config list,
>> it can contain section name, name, values(default, user config,
>> system config, both config), correct type, etc.
>>
>> If we do, we needn't to find detail for a config variable at other objects
>> e.g.
>> When we find correct type of a config variable,
>> we needn't to do for-loop for default_configs[] in order to know the
>> type.
>
> I'm not sure I understand you correctly, but I think this is not
> related to the two-level structure.

What you said is right.
I just thought using two lists (list_head sections, default_configs[])
was complex..
At this patchset, I only focused on simplifying the config list on 
perf-config..

As you said, it hasn't problems to use the two-level structure i.e.

     struct config_element
     struct config_section (that has multiple elements)

So, what about this structures ?
(you may already think about it, but..)

     struct config_element {
             char *name;
             char *value; /*from the two config file (user/system)*/
             union {
                 bool b;
                 int i;
                 u32 l;
                 u64 ll;
                 float f;
                 double d;
                 const char *s;
             } default_value; /* perf's default config value */
             enum config_type type;
             struct list_head list;
     };

     struct config_section {
             char *name;
             struct list_head element_head;
             struct list_head list;
     };

Because I want to use only one list (default_configs + sections)
for more concise code than old code.

>
>>
>>
>> I think this is better than old two structure.
>>
>>>> +
>>>> +	return NULL;
>>>> +}
>>>> +
>>>> +static struct perf_config_item *add_config(struct list_head *config_list,
>>>> +					   const char *section,
>>>> +					   const char *name)
>>>> +{
>>>> +	struct perf_config_item *config = zalloc(sizeof(*config));
>>>> +
>>>> +	if (!config)
>>>> +		return NULL;
>>>> +
>>>> +	config->section = strdup(section);
>>>> +	if (!section)
>>>> +		goto out_err;
>>>> +
>>>> +	config->name = strdup(name);
>>>> +	if (!name) {
>>>> +		free((char *)config->section);
>>>> +		goto out_err;
>>>> +	}
>>>> +
>>>> +	list_add_tail(&config->list, config_list);
>>>> +	return config;
>>>> +
>>>> +out_err:
>>>> +	free(config);
>>>> +	pr_err("%s: strdup failed\n", __func__);
>>>> +	return NULL;
>>>> +}
>>>> +
>>>> +static int set_value(struct perf_config_item *config, const char *value)
>>>> +{
>>>> +	char *val = strdup(value);
>>>> +
>>>> +	if (!val)
>>>> +		return -1;
>>>> +	config->value = val;
>>>
>>> It seems to overwrite old value..
>>>
>>
>> Yes, I know it.
>> If don't using '--user' or '--system',
>> there isn't exclusive config file path
>> then have to read both config files.
>>
>> But because user config file has a high order of priority,
>> if two config file has same variable, old value(for system config)
>> must be overwrote by new value(for user config).
>
> But shouldn't it free the old value before overwriting?
>

Sorry, I missed free() out.
I'll fix it.

Thanks,
Taeung

>
>>
>>>
>>>
>>>> +
>>>> +	return 0;
>>>> +}