ivampiresp/Spdk
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

nvmf: align the data buffers for AIO backend

Browse Source 
Currently SPDK NVMeoF will allocate data buffer from mempool, however,
some data buffers may not block size aligned, this is okay for NVMe
backend but will fail for AIO backend. 4KiB aligned data buffers
should be okay for most AIO backends.

Change-Id: If7b9c2962d716addbd755e80e59620377f2ba488
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.gerrithub.io/389245
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
This commit is contained in:
Changpeng Liu 2017-11-28 01:34:38 -05:00 committed by Jim Harris
parent 3dd4d63940
commit 415a1485cf
1 changed files with 20 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

30
lib/nvmf/rdma.c
View File

@ -56,6 +56,13 @@
#define NVMF_DEFAULT_TX_SGE 1
#define NVMF_DEFAULT_RX_SGE 2
/* AIO backend requires block size aligned data buffers,
* extra 4KiB aligned data buffer should work for most devices.
*/
#define SHIFT_4KB 12
#define NVMF_DATA_BUFFER_ALIGNMENT (1 << SHIFT_4KB)
#define NVMF_DATA_BUFFER_MASK (NVMF_DATA_BUFFER_ALIGNMENT - 1)
enum spdk_nvmf_rdma_request_state {
/* The request is not currently in use */
RDMA_REQUEST_STATE_FREE = 0,
@ -117,7 +124,7 @@ struct spdk_nvmf_rdma_recv {
struct spdk_nvmf_rdma_request {
struct spdk_nvmf_request req;
bool data_from_pool;
void *data_from_pool;
enum spdk_nvmf_rdma_request_state state;
@ -914,14 +921,17 @@ spdk_nvmf_rdma_request_parse_sgl(struct spdk_nvmf_rdma_transport *rtransport,
}
rdma_req->req.length = sgl->keyed.length;
rdma_req->req.data = spdk_mempool_get(rtransport->data_buf_pool);
if (!rdma_req->req.data) {
rdma_req->data_from_pool = spdk_mempool_get(rtransport->data_buf_pool);
if (!rdma_req->data_from_pool) {
/* No available buffers. Queue this request up. */
SPDK_DEBUGLOG(SPDK_TRACE_RDMA, "No available large data buffers. Queueing request %p\n", rdma_req);
return 0;
}
rdma_req->data_from_pool = true;
/* AIO backend requires block size aligned data buffers,
* 4KiB aligned data buffer should work for most devices.
*/
rdma_req->req.data = (void *)((uintptr_t)(rdma_req->data_from_pool + NVMF_DATA_BUFFER_MASK)
& ~NVMF_DATA_BUFFER_MASK);
rdma_req->data.sgl[0].addr = (uintptr_t)rdma_req->req.data;
rdma_req->data.sgl[0].length = sgl->keyed.length;
rdma_req->data.sgl[0].lkey = ((struct ibv_mr *)spdk_mem_map_translate(device->map,
@ -956,7 +966,7 @@ spdk_nvmf_rdma_request_parse_sgl(struct spdk_nvmf_rdma_transport *rtransport,
}
rdma_req->req.data = rdma_req->recv->buf + offset;
rdma_req->data_from_pool = false;
rdma_req->data_from_pool = NULL;
rdma_req->req.length = sgl->unkeyed.length;
return 0;
}
@ -1045,7 +1055,7 @@ spdk_nvmf_rdma_request_process(struct spdk_nvmf_rdma_transport *rtransport,
/* If data is transferring from host to controller and the data didn't
* arrive using in capsule data, we need to do a transfer from the host.
*/
if (rdma_req->req.xfer == SPDK_NVME_DATA_HOST_TO_CONTROLLER && rdma_req->data_from_pool) {
if (rdma_req->req.xfer == SPDK_NVME_DATA_HOST_TO_CONTROLLER && rdma_req->data_from_pool != NULL) {
rdma_req->state = RDMA_REQUEST_STATE_TRANSFER_PENDING_HOST_TO_CONTROLLER;
TAILQ_INSERT_TAIL(&rqpair->pending_rdma_rw_queue, rdma_req, link);
break;
@ -1116,8 +1126,8 @@ spdk_nvmf_rdma_request_process(struct spdk_nvmf_rdma_transport *rtransport,
if (rdma_req->data_from_pool) {
/* Put the buffer back in the pool */
spdk_mempool_put(rtransport->data_buf_pool, rdma_req->req.data);
rdma_req->data_from_pool = false;
spdk_mempool_put(rtransport->data_buf_pool, rdma_req->data_from_pool);
rdma_req->data_from_pool = NULL;
}
rdma_req->req.length = 0;
rdma_req->req.data = NULL;
@ -1184,7 +1194,7 @@ spdk_nvmf_rdma_create(struct spdk_nvmf_tgt *tgt)
rtransport->data_buf_pool = spdk_mempool_create("spdk_nvmf_rdma",
rtransport->max_queue_depth * 4, /* The 4 is arbitrarily chosen. Needs to be configurable. */
rtransport->max_io_size,
rtransport->max_io_size + NVMF_DATA_BUFFER_ALIGNMENT,
SPDK_MEMPOOL_DEFAULT_CACHE_SIZE,
SPDK_ENV_SOCKET_ID_ANY);
if (!rtransport->data_buf_pool) {