> On Aug 15, 2017, at 12:13 PM, Trond Myklebust <trondmy@xxxxxxxxxxxxxxx> wrote: > > On Tue, 2017-08-15 at 11:05 -0400, Chuck Lever wrote: >> Reviewed-by: Chuck Lever <chuck.lever@xxxxxxxxxx> > > Thanks! I agree with your proposal for how to apply this to the RDMA > code. I'm hoping you will be able to look into that and check that the > removal of transport lock protection won't end up causing corruption of > the RDMA statistics counters, etc. The client-side RDMA statistics counters are currently not protected, in general (although yes, some of them happen to be manipulated only behind locks). I'm more worried about how to guarantee that MR invalidation occurs exactly once as an RPC completes. My plan is to wait until xprt_pin_rqst lands in mainline, then add xprtrdma usage of that API in the following kernel release. Let me know if you'd like something to go in _with_ xprt_pin_rqst instead, and we can try to co-ordinate. -- Chuck Lever -- To unsubscribe from this list: send the line "unsubscribe linux-nfs" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
