Re: [Patch 2/3] ksym_tracer: Allow bulk removal using empty or wildcard string input

[Frederic Weisbecker <fweisbec@gmail.com>]

On Sat, Jun 20, 2009 at 07:07:52PM +0530, K.Prasad wrote:
> On Sat, Jun 20, 2009 at 01:16:41AM +0200, Frederic Weisbecker wrote:
> > On Fri, Jun 19, 2009 at 10:55:12PM +0530, K.Prasad wrote:
> > > Accept an empty string or wildcard input (of the form *:---) as input for
> > > ksym_trace_filter to perform bulk removal of all monitored entries.
> > >
> > > Signed-off-by: K.Prasad <prasad@linux.vnet.ibm.com>
> > > ---
> > >  kernel/trace/trace_ksym.c |   75 +++++++++++++++++++++++++---------------------
> > >  1 file changed, 42 insertions(+), 33 deletions(-)
> > > 
> <snipped>
> > >  /*
> > >   * There can be several possible malformed requests and we attempt to capture
> > >   * all of them. We enumerate some of the rules
> > > @@ -163,7 +191,7 @@ static int parse_ksym_trace_str(char *in
> > >  	/* Check for malformed request: (2), (1) and (5) */
> > >  	if ((!input_string) ||
> > >  		(strlen(input_string) != (KSYM_TRACER_OP_LEN + 1)) ||
> > > -			(*addr == 0))
> > > +			((*addr == 0) && strncmp(*ksymname, "*", strlen("*"))))
> > 
> > 
> > I guess you can just use strcmp() here, no? As in other use
> > of this same strncmp() in this patch.
> > 
> 
> I see both functions being used in the kernel at random (what happened
> to all the buffer-overflow scare?)! Isn't an strncmp() preferred over
> the plain strcmp() (atleast it cannot be the other way round)?


I guess it depends if the string has complicated origins :)
But a plain 1 for the len would be sufficient.
Or here you can just check **ksymname != '*'

Anyway, I'm arguing about neat things that don't matter that much.

Thanks.
 
> > > @@ -251,9 +280,11 @@ static ssize_t ksym_trace_filter_write(s
> > >  	unsigned long ksym_addr = 0;
> > >  	int ret, op, changed = 0;
> > >  
> > > -	/* Ignore echo "" > ksym_trace_filter */
> > > -	if (count == 0)
> > > -		return 0;
> > > +	/* Clear all breakpoint requests if echo "" > ksym_trace_filter */
> > > +	if (count == 1) {
> > > +		ksym_trace_reset(NULL);
> > > +		return count;
> > > +	}
> > 
> > 
> > And then if I do:
> > 
> > echo -n 1 > ksym_trace_filter
> > 
> > That would also clean these breakpoints?
> > 
> > Frederic.
> >
> 
> No, I did not think about this scenario and I am changing the patch to
> handle such a case. Also, the value of 'count' cannot be '0' as such
> inputs are ignored (e.g. echo -n "" > ksym_trace_filter is simply
> ignored) and so I'm removing the check for the same.
> 
> Please find a new version of the patch sent here:
> http://lkml.org/lkml/2009/6/20/73.
> 
> Thanks,
> K.Prasad
> 
>