Re: [PATCH v8 1/5] mm: introduce a common interface for balloon pages mobility

[Rafael Aquini <aquini@redhat.com>]

On Tue, Aug 21, 2012 at 10:30:31PM +0300, Michael S. Tsirkin wrote:
> On Tue, Aug 21, 2012 at 04:23:58PM -0300, Rafael Aquini wrote:
> > On Tue, Aug 21, 2012 at 10:13:30PM +0300, Michael S. Tsirkin wrote:
> > > > 
> > > > I believe rcu_dereference_protected() is what I want/need here, since this code
> > > > is always called for pages which we hold locked (PG_locked bit).
> > > 
> > > It would only help if we locked the page while updating the mapping,
> > > as far as I can see we don't.
> > >
> > 
> > But we can do it. In fact, by doing it (locking the page) we can easily avoid
> > the nasty race balloon_isolate_page / leak_balloon, in a much simpler way, IMHO.
> 
> Absolutely. Further, we should look hard at whether most RCU uses
> in this patchset can be replaced with page lock.
>

Yeah, In fact, by testing/grabbing the page lock at leak_balloon() even the
module unload X migration / putback race seems to fade away, since migration
code holds the page locked all the way.

And that seems a quite easy task to be accomplished:

....
@@ -169,21 +197,61 @@ static void leak_balloon(struct virtio_balloon *vb, size_t
num)
        /* We can only do one array worth at a time. */
        num = min(num, ARRAY_SIZE(vb->pfns));

+       mutex_lock(&vb->balloon_lock);
        for (vb->num_pfns = 0; vb->num_pfns < num;
             vb->num_pfns += VIRTIO_BALLOON_PAGES_PER_PAGE) {
+               spin_lock(&vb->pages_lock);
+               /*
+                * 'virtballoon_isolatepage()' can drain vb->pages list
+                * making us to stumble across a _temporarily_ empty list.
+                *
+                * Release the spinlock and resume from here in order to
+                * give page migration a shot to refill vb->pages list.
+                */
+               if (unlikely(list_empty(&vb->pages))) {
+                       spin_unlock(&vb->pages_lock);
+                       break;
+               }
+
                page = list_first_entry(&vb->pages, struct page, lru);
+
+               /*
+                * Grab the page lock to avoid racing against threads isolating
+                * pages from vb->pages list (it's done under page lock).
+                *
+                * Failing to grab the page lock here means this page has been
+                * selected for isolation already.
+                */
+               if (!trylock_page(page)) {
+                       spin_unlock(&vb->pages_lock);
+                       break;
+               }
+
+               clear_balloon_mapping(page);
                list_del(&page->lru);
                set_page_pfns(vb->pfns + vb->num_pfns, page);
                vb->num_pages -= VIRTIO_BALLOON_PAGES_PER_PAGE;
+               unlock_page(page);
+               spin_unlock(&vb->pages_lock);
        }

.....

--
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-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>