Re: [PATCH 05/13] ftrace: Call trace_parser_clear() properly

[Li Zefan <lizf@cn.fujitsu.com>]

Frederic Weisbecker wrote:
> On Tue, Dec 08, 2009 at 11:15:30AM +0800, Li Zefan wrote:
>> I found a weird behavior:
>>
>>   # echo 'fuse:*' > set_ftrace_filter
>>   bash: echo: write error: Invalid argument
>>   # cat set_ftrace_filter
>>   fuse_dev_fasync
>>   fuse_dev_poll
>>   fuse_copy_do
>>
>> We should call trace_parser_clear() no matter ftrace_process_regex()
>> returns 0 or -errno.
>>
>> Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
>> ---
>>  kernel/trace/ftrace.c |    3 +--
>>  1 files changed, 1 insertions(+), 2 deletions(-)
>>
>> diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c
>> index 08a3fb5..ff8aecd 100644
>> --- a/kernel/trace/ftrace.c
>> +++ b/kernel/trace/ftrace.c
>> @@ -2208,10 +2208,9 @@ ftrace_regex_write(struct file *file, const char __user *ubuf,
>>  	    !trace_parser_cont(parser)) {
>>  		ret = ftrace_process_regex(parser->buffer,
>>  					   parser->idx, enable);
>> +		trace_parser_clear(parser);
>>  		if (ret)
>>  			goto out_unlock;
>> -
>> -		trace_parser_clear(parser);
>>  	}
> 
> 
> I'm missing something. How that can happen. Anytime we reopen
> the file, the parser is re-allocated.
> 

It happened at file closing..

static int
ftrace_regex_release(struct inode *inode, struct file *file, int enable)
{
	...
        parser = &iter->parser;
        if (trace_parser_loaded(parser)) {
                parser->buffer[parser->idx] = 0;
		/* here ! */
                ftrace_match_records(parser->buffer, parser->idx, enable);
        }
	...
}

> I guess that happens if you open in rw mode? But not using the
> example in the changelog?

I've confirmed that example I was using can reveal this bug.