Re: [PATCH 1/4] perf tools: remove sscanf extension %as

[Masami Hiramatsu]

(2012/09/14 10:54), Namhyung Kim wrote:
> Hi Irina,
> 
> (Adding Masami to Cc)

Thanks Irina and Namhyung :)

> On Fri, 14 Sep 2012 01:07:40 +0300, Irina Tirdea wrote:
>> From: Irina Tirdea <irina.tirdea@xxxxxxxxx>
>>
>> perf uses sscanf extension %as to read and allocate a
>> string in the same step. This is a non-standard extension
>> only present in new versions of glibc.
>>
>> Replacing the use of sscanf and %as with strtok_r calls
>> in order to parse a given string into its components.
>> This is needed in Android since bionic does not support
>> %as extension for sscanf.
>>
>> Signed-off-by: Irina Tirdea <irina.tirdea@xxxxxxxxx>
>> ---
>>  tools/perf/util/probe-event.c       |   25 ++++++++++++++++++-------
>>  tools/perf/util/trace-event-parse.c |   18 ++++++++----------
>>  2 files changed, 26 insertions(+), 17 deletions(-)
>>
>> diff --git a/tools/perf/util/probe-event.c b/tools/perf/util/probe-event.c
>> index 4ce04c2..685ddcf 100644
>> --- a/tools/perf/util/probe-event.c
>> +++ b/tools/perf/util/probe-event.c
>> @@ -1100,6 +1100,7 @@ static int parse_probe_trace_command(const char *cmd,
>>  	struct probe_trace_point *tp = &tev->point;
>>  	char pr;
>>  	char *p;
>> +	char *argv0_str = NULL, *fmt, *fmt1_str, *fmt2_str, *fmt3_str;
>>  	int ret, i, argc;
>>  	char **argv;
>>  
>> @@ -1116,14 +1117,19 @@ static int parse_probe_trace_command(const char *cmd,
>>  	}
>>  
>>  	/* Scan event and group name. */
>> -	ret = sscanf(argv[0], "%c:%a[^/ \t]/%a[^ \t]",
>> -		     &pr, (float *)(void *)&tev->group,
>> -		     (float *)(void *)&tev->event);
>> -	if (ret != 3) {
>> +	argv0_str = strdup(argv[0]);
> 
> It seems you need to check return value of strdup.

Agreed.

>> +	fmt1_str = strtok_r(argv0_str, ":", &fmt);
>> +	fmt2_str = strtok_r(NULL, "/", &fmt);
>> +	fmt3_str = strtok_r(NULL, " \t", &fmt);
>> +	if (fmt1_str == NULL || strlen(fmt1_str) != 1 || fmt2_str == NULL
>> +	    || fmt3_str == NULL) {
>>  		semantic_error("Failed to parse event name: %s\n", argv[0]);
>>  		ret = -EINVAL;
>>  		goto out;
>>  	}
>> +	pr = fmt1_str[0];
>> +	tev->group = strdup(fmt2_str);
>> +	tev->event = strdup(fmt3_str);
>>  	pr_debug("Group:%s Event:%s probe:%c\n", tev->group, tev->event, pr);
>>  
>>  	tp->retprobe = (pr == 'r');
>> @@ -1135,10 +1141,13 @@ static int parse_probe_trace_command(const char *cmd,
>>  		p++;
>>  	} else
>>  		p = argv[1];
>> -	ret = sscanf(p, "%a[^+]+%lu", (float *)(void *)&tp->symbol,
>> -		     &tp->offset);
>> -	if (ret == 1)
>> +	fmt1_str = strtok_r(p, "+", &fmt);
>> +	tp->symbol = strdup(fmt1_str);
> 
> Probably here too - although the original code didn't but I think it's
> needed.

right, it should be fixed.

>> +	fmt2_str = strtok_r(NULL, "", &fmt);
>> +	if (fmt2_str == NULL)
>>  		tp->offset = 0;
>> +	else
>> +		tp->offset = strtoul(fmt2_str, NULL, 10);
>>  
>>  	tev->nargs = argc - 2;
>>  	tev->args = zalloc(sizeof(struct probe_trace_arg) * tev->nargs);
>> @@ -1162,6 +1171,8 @@ static int parse_probe_trace_command(const char *cmd,
>>  	}
>>  	ret = 0;
>>  out:
>> +	if (argv0_str)
>> +		free(argv0_str);
> 
> The free() can handle a NULL pointer safely.

I don't care about it, since there are already lots of "if (ptr) free(ptr)"
style code ;)

> 
>>  	argv_free(argv);
>>  	return ret;
>>  }

Thank you!

-- 
Masami HIRAMATSU
Software Platform Research Dept. Linux Technology Center
Hitachi, Ltd., Yokohama Research Laboratory
E-mail: masami.hiramatsu.pt@xxxxxxxxxxx


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/