On Tue 09-09-25 14:41:33, Pankaj Raghav (Samsung) wrote: > On Mon, Sep 08, 2025 at 04:55:34PM +0200, Jan Kara wrote: > > > > /* > > @@ -314,6 +317,7 @@ static void do_page_cache_ra(struct readahead_control *ractl, > > loff_t isize = i_size_read(inode); > > pgoff_t end_index; /* The last page we want to read */ > > > > + trace_do_page_cache_ra(inode, index, nr_to_read, lookahead_size); > > Any reason why put a probe here instead of page_cache_ra_unbounded as > that is where the actual readahead happens? Hum, no. Originally I had it in force_page_cache_ra() but then I've decided do_page_cache_ra() is better because it captures also other places issuing non-standard readahead. But you're right that placing the tracepoint in page_cache_ra_unbounded() will achieve that as well and will be a more standard place. I'll respin the patch. Thanks for suggestion. Honza -- Jan Kara <jack@xxxxxxxx> SUSE Labs, CR
