On Wed, Nov 22, 2017 at 01:06:37PM -0800, Matthew Wilcox wrote: > From: Matthew Wilcox <mawilcox@xxxxxxxxxxxxx> > > I've lost count of the number of times I've posted the XArray before, > so time for a new numbering scheme. Here're two earlier versions, > https://lkml.org/lkml/2017/3/17/724 > https://lwn.net/Articles/715948/ (this one's more loquacious in its > description of things that are better about the radix tree API than the > XArray). > > This time around, I've gone for an approach of many small changes. > Unfortunately, that means you get 62 moderate patches instead of dozens > of big ones. Where's the API documentation that tells things like constraints about locking and lock-less lookups via RCU? e.g. I notice in the XFS patches you seem to randomly strip out rcu_read_lock/unlock() pairs that are currently around radix tree lookup operations without explanation. Without documentation describing how this stuff is supposed to work, review is somewhat difficult... Cheers, Dave. -- Dave Chinner david@xxxxxxxxxxxxx
