[lkml]   [2010]   [Aug]   [26]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: [PATCH] mm: fix hang on anon_vma->root->lock
Hi Hugh,

On Wed, Aug 25, 2010 at 11:12:54PM -0700, Hugh Dickins wrote:
> After several hours, kbuild tests hang with anon_vma_prepare() spinning on
> a newly allocated anon_vma's lock - on a box with CONFIG_TREE_PREEMPT_RCU=y
> (which makes this very much more likely, but it could happen without).
> The ever-subtle page_lock_anon_vma() now needs a further twist: since
> anon_vma_prepare() and anon_vma_fork() are liable to change the ->root
> of a reused anon_vma structure at any moment, page_lock_anon_vma()
> needs to check page_mapped() again before succeeding, otherwise
> page_unlock_anon_vma() might address a different root->lock.

I don't get it, the anon_vma can be freed and reused only after we run
rcu_read_unlock(). And the anon_vma->root can't change unless the
anon_vma is freed and reused. Last but not the least by the time
page->mapping points to "anon_vma" the "anon_vma->root" is already
initialized and stable.

The page_mapped test is only relevant against the rcu_read_lock, not
the spin_lock, so how it can make a difference to run it twice inside
the same rcu_read_lock protected critical section? The first one still
is valid also after the anon_vma_lock() returns, it's not like that
anon_vma_lock drops the rcu_read_lock internally.

Furthermore no need of ACCESS_ONCE on the anon_vma->root because it
can't change from under us as the anon_vma can't be freed from under
us until rcu_read_unlock returns (after we verified the first time
that page_mapped is true under the rcu_read_lock, which we already do
before trying to take the anon_vma_lock).

 \ /
  Last update: 2010-08-27 01:53    [W:0.047 / U:2.592 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site