diff options
| author | Chuck Lever <chuck.lever@oracle.com> | 2020-03-03 21:28:14 +0300 |
|---|---|---|
| committer | Chuck Lever <chuck.lever@oracle.com> | 2020-03-16 19:04:33 +0300 |
| commit | 0dabe948f28274e7956a625a24f205016b810693 (patch) | |
| tree | 520f4aa1ad1e6973d3e5d0914927988134548644 /include | |
| parent | aee4b74a3f273b54d136132fedf575ec464f4134 (diff) | |
| download | linux-0dabe948f28274e7956a625a24f205016b810693.tar.xz | |
svcrdma: Avoid DMA mapping small RPC Replies
On some platforms, DMA mapping part of a page is more costly than
copying bytes. Indeed, not involving the I/O MMU can help the
RPC/RDMA transport scale better for tiny I/Os across more RDMA
devices. This is because interaction with the I/O MMU is eliminated
for each of these small I/Os. Without the explicit unmapping, the
NIC no longer needs to do a costly internal TLB shoot down for
buffers that are just a handful of bytes.
Since pull-up is now a more a frequent operation, I've introduced a
trace point in the pull-up path. It can be used for debugging or
user-space tools that count pull-up frequency.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Diffstat (limited to 'include')
| -rw-r--r-- | include/linux/sunrpc/svc_rdma.h | 1 | ||||
| -rw-r--r-- | include/trace/events/rpcrdma.h | 18 |
2 files changed, 19 insertions, 0 deletions
diff --git a/include/linux/sunrpc/svc_rdma.h b/include/linux/sunrpc/svc_rdma.h index a3fa5b4fa2e4..78fe2ac6dc6c 100644 --- a/include/linux/sunrpc/svc_rdma.h +++ b/include/linux/sunrpc/svc_rdma.h @@ -52,6 +52,7 @@ /* Default and maximum inline threshold sizes */ enum { + RPCRDMA_PULLUP_THRESH = RPCRDMA_V1_DEF_INLINE_SIZE >> 1, RPCRDMA_DEF_INLINE_THRESH = 4096, RPCRDMA_MAX_INLINE_THRESH = 65536 }; diff --git a/include/trace/events/rpcrdma.h b/include/trace/events/rpcrdma.h index 74b68547eefb..9238d233f8cf 100644 --- a/include/trace/events/rpcrdma.h +++ b/include/trace/events/rpcrdma.h @@ -1639,6 +1639,24 @@ TRACE_EVENT(svcrdma_dma_map_rwctx, ) ); +TRACE_EVENT(svcrdma_send_pullup, + TP_PROTO( + unsigned int len + ), + + TP_ARGS(len), + + TP_STRUCT__entry( + __field(unsigned int, len) + ), + + TP_fast_assign( + __entry->len = len; + ), + + TP_printk("len=%u", __entry->len) +); + TRACE_EVENT(svcrdma_send_failed, TP_PROTO( const struct svc_rqst *rqst, |