On 2/3/2021 3:06 PM, Chuck Lever wrote:
Changes since v2:
- Another minor optimization in rpcrdma_convert_kvec()
- Some patch description clarifications
- Add Reviewed-by (thanks Tom!)
Changes since v1:
- Respond to review comments
- Split "Remove FMR support" into three patches for clarity
- Fix implicit chunk roundup
- Improve Receive completion tracepoints
---
Chuck Lever (6):
xprtrdma: Remove FMR support in rpcrdma_convert_iovs()
xprtrdma: Simplify rpcrdma_convert_kvec() and frwr_map()
xprtrdma: Refactor invocations of offset_in_page()
rpcrdma: Fix comments about reverse-direction operation
xprtrdma: Pad optimization, revisited
rpcrdma: Capture bytes received in Receive completion tracepoints
include/trace/events/rpcrdma.h | 50 +++++++++++++++++++++-
net/sunrpc/xprtrdma/backchannel.c | 4 +-
net/sunrpc/xprtrdma/frwr_ops.c | 12 ++----
net/sunrpc/xprtrdma/rpc_rdma.c | 17 +++-----
While reviewing the changes in rpc_rdma.c, I noticed a related minor
nit, which might be worth cleaning up for clarity.
Toward the end of rpcrdma_convert_iovs, there is no longer any
need to capture the returned value of rpcrdma_convert_kvec:
if (xdrbuf->tail[0].iov_len)
seg = rpcrdma_convert_kvec(&xdrbuf->tail[0], seg, &n);
^^^^^^ --> (void)?
out:
if (unlikely(n > RPCRDMA_MAX_SEGS))
return -EIO;
return n;
The two "goto out" statements just above it are getting kinda ugly
too, but...
Tom.
net/sunrpc/xprtrdma/svc_rdma_backchannel.c | 4 +-
net/sunrpc/xprtrdma/xprt_rdma.h | 15 ++++---
6 files changed, 68 insertions(+), 34 deletions(-)
--
Chuck Lever
