This is the mail archive of the systemtap@sourceware.org mailing list for the systemtap project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH v2 2.6.38-rc8-tip 17/20] 17: uprobes: filter chain


* Stephen Wilson <wilsons@start.ca> [2011-03-15 15:49:14]:

> 
> 
> 
> On Mon, Mar 14, 2011 at 07:07:22PM +0530, Srikar Dronamraju wrote:
> > 
> > Loops through the filters callbacks of currently registered
> > consumers to see if any consumer is interested in tracing this task.
> 
> Should this be part of the series?  It is not currently used.
> 
> >  /* Acquires uprobe->consumer_rwsem */
> > +static bool filter_chain(struct uprobe *uprobe, struct task_struct *t)
> > +{
> > +	struct uprobe_consumer *consumer;
> > +	bool ret = false;
> > +
> > +	down_read(&uprobe->consumer_rwsem);
> > +	for (consumer = uprobe->consumers; consumer;
> > +					consumer = consumer->next) {
> > +		if (!consumer->filter || consumer->filter(consumer, t)) {
> 
> The implementation does not seem to match the changelog description.
> Should this not be:
> 
>                 if (consumer->filter && consumer->filter(consumer, t))
> 
>   ?

filter is optional; if filter is present, then it means that the
tracer is interested in a specific set of processes that maps this
inode. If there is no filter; it means that it is interested in all
processes that map this filter. 

filter_chain() should return true if any consumer is interested in
tracing this task.
if there is a consumer who hasnt defined a filter then we dont need to loop thro remaining consumers.

Hence 

if (!consumer->filter || consumer->filter(consumer, t)) {
 
seems better suited to me.

-- 
Thanks and Regards
Srikar


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]