Re: [PATCH][RFC] splice support

[Andrew Morton <akpm@osdl.org>]

Jens Axboe <axboe@suse.de> wrote:
>
> > find_get_pages() does "find me the next N pages above `index' which are
>  > presently in pagecache'.  So it can return an array of page*'s which do not
>  > represent contiguous pages in the file - there can be holes in there.
>  > 
>  > IOW: pages[n]->index !necessarily= pages[n+1]->index-1
>  > 
>  > Maybe the code handles that by making sure that all the pages in the range
>  > are already in pagecache - I didn't check.  But that would take some heroic
>  > locking.
> 
>  It doesn't, I'm assuming that find_get_pages() returns consequtive pages
>  atm. Would seem like the sane interface :-)

Yeah, sorry.  It's a "gather what's presently there" thing.  For writeback.

Nick has some gang-lookup-slots code.  So instead of populating an array of
page*'s you can populate an array of (effectively) page**'s.  Then one
could walk that.   All while holding ->tree_lock.    This doesn't help ;)

Or you could walk the pages[] array until you hit an ->index which doesn't
match and then toss the rest away.  That's a bit of extra work, but in the
common case all the pages will be good.  Perhaps.

>  We continue doing find_or_create_page() on the remaining, but using 'i'
>  as the 'index' addition. So if we had non-conseq pages, we'd be screwed.

Yup.

Probably the simplest for now is an open-coded find_get_page() loop.  Later
on we should optimise that into a find_get_contig_pages() which only takes
tree_lock a single time.

Doing it with a new radix_tree_gang_lookup_contig_name_me_longer() would be
relatively straightforward too.  It would bale out as soon as it hit a
not-present slot.