xprtrdma: Split rb_lock
authorChuck Lever <chuck.lever@oracle.com>
Tue, 26 May 2015 15:53:13 +0000 (11:53 -0400)
committerAnna Schumaker <Anna.Schumaker@Netapp.com>
Fri, 12 Jun 2015 17:10:37 +0000 (13:10 -0400)
commit58d1dcf5a8ebb0ce8a521286a99efdd636012bf0
tree8fcc6ca4d9a5f1234f9f73acdd0fa0bd20a5e06c
parent7e53df111beea8db2543424d07bdee2a630698c3
xprtrdma: Split rb_lock

/proc/lock_stat showed contention between rpcrdma_buffer_get/put
and the MR allocation functions during I/O intensive workloads.

Now that MRs are no longer allocated in rpcrdma_buffer_get(),
there's no reason the rb_mws list has to be managed using the
same lock as the send/receive buffers. Split that lock. The
new lock does not need to disable interrupts because buffer
get/put is never called in an interrupt context.

struct rpcrdma_buffer is re-arranged to ensure rb_mwlock and rb_mws
are always in a different cacheline than rb_lock and the buffer
pointers.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Steve Wise <swise@opengridcomputing.com>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Tested-By: Devesh Sharma <devesh.sharma@avagotech.com>
Reviewed-by: Doug Ledford <dledford@redhat.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
net/sunrpc/xprtrdma/fmr_ops.c
net/sunrpc/xprtrdma/frwr_ops.c
net/sunrpc/xprtrdma/verbs.c
net/sunrpc/xprtrdma/xprt_rdma.h