On Fri 03-06-16 17:10:01, Andrea Arcangeli wrote: > Hello Michal, > > CC'ed Hugh, > > On Fri, Jun 03, 2016 at 04:46:00PM +0200, Michal Hocko wrote: > > What do you think about the external dependencies mentioned above. Do > > you think this is a sufficient argument wrt. occasional higher > > latencies? > > It's a tradeoff and both latencies would be short and uncommon so it's > hard to tell. > > There's also mmput_async for paths that may care about mmput > latencies. Exit itself cannot use it, it's mostly for people taking > the mm_users pin that may not want to wait for mmput to run. It also > shouldn't happen that often, it's a slow path. > > The whole model inherited from KSM is to deliberately depend only on > the mmap_sem + test_exit + mm_count, and never on mm_users, which to > me in principle doesn't sound bad.
I do agree that this model is quite clever (albeit convoluted). It just assumes that all other mmap_sem users are behaving the same. Now most in-kernel users will do get_task_mm() and then lock mmap_sem, but I haven't checked all of them and it is quite possible that some of those would like to optimize in a similar way and only increment mm_count. I might be too pessimistic about the out of mm code but I would feel much better if the exit path didn't depend on them. Anyway, if the current model sounds better I will definitely not insist on my patch. It is more of an idea for simplification than a fix for anything I have seen happening in the real life. -- Michal Hocko SUSE Labs
