Messages in this thread | | | Subject | Re: [PATCH -mm -next] ipc,sem: untangle RCU locking with find_alloc_undo | From | Mike Galbraith <> | Date | Fri, 05 Apr 2013 18:26:51 +0200 |
| |
On Fri, 2013-04-05 at 09:21 -0400, Rik van Riel wrote: > On 04/05/2013 12:38 AM, Mike Galbraith wrote: > > On Tue, 2013-03-26 at 16:00 -0400, Rik van Riel wrote: > > >> The ipc semaphore code has a nasty RCU locking tangle, with both > >> find_alloc_undo and semtimedop taking the rcu_read_lock(). The > >> code can be cleaned up somewhat by only taking the rcu_read_lock > >> once. > >> > >> There are no other callers to find_alloc_undo. > >> > >> This should also solve the trinity issue reported by Sasha Levin. > > > > I take it this is on top of the patchlet Sasha submitted? > > Indeed, and all the other fixes that got submitted :) > > > (I hit rcu stall banging on patch set in rt with 60 synchronized core > > executive model if I let it run long enough, fwtw) > > What are you using to trigger an rcu stall?
Running a model of a userspace task scheduler. That was a fix or so ago now though. I'll try the set again on that box.
-Mike
| |