rsocket: Add support for iWarp

iWarp does not support RDMA writes with immediate data.
Instead of sending messages using immediate data, allow
the rsocket protocol to exchange messages using sends.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>

diff --git a/src/rsocket.c b/src/rsocket.c
index ca771165dec3b2c2395a11537dccc55eb5f6c98b..de15f58ba86223b80544d4cb8e57d463665c8c05 100644 (file)
--- a/src/rsocket.c
+++ b/src/rsocket.c
@@ -207,7 +207,12 @@ enum rs_state {
        rs_error           =                0x2000,
 };
 
-#define RS_OPT_SWAP_SGL 1
+#define RS_OPT_SWAP_SGL        (1 << 0)
+/*
+ * iWarp does not support RDMA write with immediate data.  For iWarp, we
+ * transfer rsocket messages as inline sends.
+ */
+#define RS_OPT_MSG_SEND        (1 << 1)
 
 union socket_addr {
        struct sockaddr         sa;
@@ -284,6 +289,7 @@ struct rsocket {
                        volatile struct rs_sge    *target_sgl;
                        struct rs_iomap   *target_iomap;
 
+                       int               rbuf_msg_index;
                        int               rbuf_bytes_avail;
                        int               rbuf_free_offset;
                        int               rbuf_offset;
@@ -635,6 +641,7 @@ static void ds_set_qp_size(struct rsocket *rs)
 
 static int rs_init_bufs(struct rsocket *rs)
 {
+       uint32_t rbuf_msg_size;
        size_t len;
 
        rs->rmsg = calloc(rs->rq_size + 1, sizeof(*rs->rmsg));
@@ -664,11 +671,14 @@ static int rs_init_bufs(struct rsocket *rs)
        if (rs->target_iomap_size)
                rs->target_iomap = (struct rs_iomap *) (rs->target_sgl + RS_SGL_SIZE);
 
-       rs->rbuf = calloc(rs->rbuf_size, sizeof(*rs->rbuf));
+       rbuf_msg_size = rs->rbuf_size;
+       if (rs->opts & RS_OPT_MSG_SEND)
+               rbuf_msg_size += rs->rq_size * RS_MSG_SIZE;
+       rs->rbuf = calloc(rbuf_msg_size, 1);
        if (!rs->rbuf)
                return ERR(ENOMEM);
 
-       rs->rmr = rdma_reg_write(rs->cm_id, rs->rbuf, rs->rbuf_size);
+       rs->rmr = rdma_reg_write(rs->cm_id, rs->rbuf, rbuf_msg_size);
        if (!rs->rmr)
                return -1;
 
@@ -685,8 +695,7 @@ static int rs_init_bufs(struct rsocket *rs)
 
 static int ds_init_bufs(struct ds_qp *qp)
 {
-       qp->rbuf = calloc(qp->rs->rbuf_size + sizeof(struct ibv_grh),
-                         sizeof(*qp->rbuf));
+       qp->rbuf = calloc(qp->rs->rbuf_size + sizeof(struct ibv_grh), 1);
        if (!qp->rbuf)
                return ERR(ENOMEM);
 
@@ -740,11 +749,27 @@ err1:
 static inline int rs_post_recv(struct rsocket *rs)
 {
        struct ibv_recv_wr wr, *bad;
+       struct ibv_sge sge;
 
        wr.wr_id = rs_recv_wr_id(0);
        wr.next = NULL;
-       wr.sg_list = NULL;
-       wr.num_sge = 0;
+       if (!(rs->opts & RS_OPT_MSG_SEND)) {
+               wr.sg_list = NULL;
+               wr.num_sge = 0;
+       } else {
+               sge.addr = (uintptr_t) rs->rbuf + rs->rbuf_size +
+                          (rs->rbuf_msg_index * RS_MSG_SIZE);
+               sge.length = sizeof(uint32_t);
+               sge.lkey = rs->rmr->lkey;
+
+               wr.sg_list = &sge;
+               wr.num_sge = 1;
+               if(++rs->rbuf_msg_index == rs->rq_size)
+                       rs->rbuf_msg_index = 0;
+               /* TODO we need wr_id to reference back to receive buffer, so we can
+                * retrieve the incoming message from the buffer
+                */
+       }
 
        return rdma_seterrno(ibv_post_recv(rs->cm_id->qp, &wr, &bad));
 }
@@ -775,6 +800,8 @@ static int rs_create_ep(struct rsocket *rs)
        int i, ret;
 
        rs_set_qp_size(rs);
+       if (rs->cm_id->context->device->transport_type == IBV_TRANSPORT_IWARP)
+               rs->opts |= RS_OPT_MSG_SEND;
        ret = rs_init_bufs(rs);
        if (ret)
                return ret;