On Tue, Sep 27, 2016 at 12:11:17PM +0300, Kirill A. Shutemov wrote: > On Tue, Sep 27, 2016 at 09:54:12AM +0100, Mel Gorman wrote: > > On Tue, Sep 27, 2016 at 09:30:55AM +0200, Peter Zijlstra wrote: > > > > Also, if those bitlock ops had a different bit that showed contention, > > > > we could actually skip *all* of this, and just see that "oh, nobody is > > > > waiting on this page anyway, so there's no point in looking up those > > > > wait queues". We don't have that many "__wait_on_bit()" users, maybe > > > > we could say that the bitlocks do have to haev *two* bits: one for the > > > > lock bit itself, and one for "there is contention". > > > > > > That would be fairly simple to implement, the difficulty would be > > > actually getting a page-flag to use for this. We're running pretty low > > > in available bits :/ > > > > Simple is relative unless I drastically overcomplicated things and it > > wouldn't be the first time. 64-bit only side-steps the page flag issue > > as long as we can live with that. > > Looks like we don't ever lock slab pages. Unless I miss something. > > We can try to use PG_locked + PG_slab to indicate contation. > > I tried to boot kernel with CONFIG_SLUB + BUG_ON(PageSlab()) in > trylock/unlock_page() codepath. Works fine, but more inspection is > required. SLUB used bit_spin_lock via slab_lock instead of trylock/unlock. -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>