On Wed, 2017-08-16 at 19:40 +0200, Borislav Petkov wrote: > On Wed, Aug 16, 2017 at 05:28:50PM +0000, Kani, Toshimitsu wrote: > > Assuming this big spinlock works, yes, this addresses my concern > > that > > You mean, lengthy locked section. We can always switch to on-stack > buffers if there are issues or even to something more fancy like > genpool.
Yes, I meant the lengthy locked section. I still think that multiple mcis is a better approach for concurrency as it naturally addresses it. But as long as edac func and error handlers are tolerant with this spinlock and serialization, I am OK with that. > > I will test the patch with an SCI when I got a chance. I won't be > > able to test other notification types or race conditions, though. > > Thanks, here's the latest version with the atomic registration too. Sure, I will test with this version. Thanks, -Toshi
