Messages in this thread | | | Date | Thu, 16 May 2013 13:28:52 -0400 | Subject | Re: [PATCH -next] ashmem: Fix ashmem_shrink deadlock. | From | Robert Love <> |
| |
On Thu, May 16, 2013 at 1:19 PM, Andrew Morton <akpm@linux-foundation.org> wrote: > On Thu, 16 May 2013 13:08:17 -0400 Robert Love <rlove@google.com> wrote: >> This problem seems a rare proper use of mutex_trylock. > > Not really. The need for a trylock is often an indication that a > subsystem has a locking misdesign. That is indeed the case here.
It is exactly the same as PF_MEMALLOC. We've got an effectively asynchronous event (shrinking) that can occur while you are holding locks requisite to that shrinking. Given that the shrinkage is best effort, a trylock actually communicates the intent pretty well: "If possible, grab this lock and shrink."
I think the idiomatic fix is to introduce a GFP_SHMEM but that seems overkill. Lots of the GFP flags are really just preventing recursing into the shrinkage code and it seems ill-designed that we require developers to know where they might end up. But we can disagree. :)
> Well, it's not exactly a ton of work, but adding a per-ashmem_area lock > to protect ->file would rather be putting lipstick on a pig. I suppose > we can put the trylock in there and run away, but it wouldn't hurt to > drop in a big fat comment somewhere explaining that the driver should be > migrated to a per-object locking scheme.
Unfortunately I think ashmem_shrink would need to grab the per-object lock too; it needs to update the ranges. I'm sure we could re-design this but I don't think it is as easy as simply pushing the locking into the objects.
Robert
| |