Re: PATCH: rewrite of invalidate_inode_pages

[Trond Myklebust <trond.myklebust@fys.uio.no>]

>>>>> " " == Juan J Quintela <quintela@fi.udc.es> writes:

     > Then you want only invalidate the non_locked pages: do you

That's right. This patch looks much more appropriate.

     > + while (count == ITERATIONS) {
     > + spin_lock(&pagecache_lock);
     > + spin_lock(&pagemap_lru_lock);
     > + head = &inode->i_mapping->pages;
     > + curr = head->next;
     > + count = 0;
     > +
     > + while ((curr != head) && (count++ < ITERATIONS)) {

Just one question: Isn't it better to do it all in 1 iteration through
the loop rather than doing it in batches of 100 pages?
You can argue that you're freeing up the spinlocks for the duration of
the loop_and_test, but is that really going to make a huge difference
to SMP performance?

Cheers,
  Trond
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux.eu.org/Linux-MM/