On Fri, Aug 27, 2010 at 12:29 PM, Christoph Lameter <cl@xxxxxxxxx> wrote: > On Fri, 27 Aug 2010, Hugh Dickins wrote: > >> Eh? My solution was a second page_mapped(page) test i.e. testing an atomic. > > Argh. Right. Looked like a global to me. Did not see the earlier local > def. > > If you still use a pointer then what does insure that the root > pointer was not changed after the ACCESS_ONCE? The free semantics > of an anon_vma? Nothing ensures that the root pointer was not changed after the ACCESS_ONCE, that's exactly why we use ACCESS_ONCE there: once we've got the lock and realize that what we've locked may not be what we wanted (or may change from what we were wanting at any moment, the page no longer being mapped there - but in that case we no longer want it), we have to be sure to unlock the one we locked, rather than the one which anon_vma->root might subsequently point to. (Umm, maybe I'm not the clearest of explainers, sorry! If you get my point, fine; if it's gibberish to you, please ask me to try again.) > > Since there is no lock taken before the mapped check none of the > earlier reads from the anon vma structure nor the page mapped check > necessarily reflect a single state of the anon_vma. There's no lock (other than RCU's read "lock") taken before the original mapped check, and that's important, otherwise our attempt to lock might actually spinon or corrupt something that was long ago an anon_vma. But we do take the anon_vma->root->lock before the second mapped check which I added. If the page is still mapped at the point of that second check, then we know that we got the right anon_vma, that the page might be mapped in it, and anon_vma->root is not going to change underneath us before the page_unlock_anon_vma(). (The page may get unmapped at any time, the lock does not protect against that; but if it's still mapped once we hold the lock, free_pgtables() cannot free the anon_vma until we're done.) Hugh -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxxx For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href