Messages in this thread | | | Date | Fri, 30 Aug 2013 08:28:33 -0700 | Subject | Re: [PATCH v7 1/4] spinlock: A new lockref structure for lockless update of refcount | From | Linus Torvalds <> |
| |
On Fri, Aug 30, 2013 at 12:16 AM, Ingo Molnar <mingo@kernel.org> wrote: > > * Linus Torvalds <torvalds@linux-foundation.org> wrote: > >> > BTW. Do you have your test case at hand ? >> >> My test-case is a joke. It's explicitly *trying* to get as much >> contention as possible on a dentry, by just starting up a lot of threads >> that look up one single pathname (the same one for everybody). It >> defaults to using /tmp for this, but you can specify the filename. > > Waiman's tests seemed to use sufficiently generic and varied workloads > (AIM7) and they showed pretty nice unconditional improvements with his > variant of this scheme, so I think testing with your simple testcase that > intentionally magnifies the scalability issue is 100% legit and may in > fact help tune the changes more accurately, because it has less inherent > noise.
Yes. However, what I am (not very) worried about is that people will hit some particular codepath that ends up having bad behavior.
I think I covered all the normal hotpaths in pathname lookup, which is why I'm not *that* worried, but it's still the case that my silly test-case is very limited. It's limited for a good *reason* (to try to show the worst-case scalability problem), but it's limited.
Linus
| |