providers/rxe: Add new API ibv_wr_flush()

Currently, we only implement it on rxe Signed-off-by: Li Zhijian <lizhijian@fujitsu.com>
linux-rdma · Jan 14, 2023 · 20a37f3 · 20a37f3
1 parent 16ac905
commit 20a37f3
Show file tree

Hide file tree

Showing 4 changed files with 57 additions and 2 deletions.
diff --git a/libibverbs/man/CMakeLists.txt b/libibverbs/man/CMakeLists.txt
@@ -136,4 +136,5 @@ rdma_alias_man_pages(
   ibv_wr_post.3 ibv_wr_set_sge_list.3
   ibv_wr_post.3 ibv_wr_set_ud_addr.3
   ibv_wr_post.3 ibv_wr_set_xrc_srqn.3
+  ibv_wr_post.3 ibv_wr_flush.3
   )
diff --git a/libibverbs/man/ibv_wr_post.3.md b/libibverbs/man/ibv_wr_post.3.md
@@ -16,7 +16,7 @@ ibv_wr_atomic_cmp_swp, ibv_wr_atomic_fetch_add - Post remote atomic operation wo
 
 ibv_wr_bind_mw, ibv_wr_local_inv - Post work requests for memory windows
 
-ibv_wr_rdma_read, ibv_wr_rdma_write, ibv_wr_rdma_write_imm - Post RDMA work requests
+ibv_wr_rdma_read, ibv_wr_rdma_write, ibv_wr_rdma_write_imm, ibv_wr_flush - Post RDMA work requests
 
 ibv_wr_send, ibv_wr_send_imm, ibv_wr_send_inv - Post send work requests
 
@@ -73,6 +73,8 @@ void ibv_wr_set_sge_list(struct ibv_qp_ex *qp, size_t num_sge,
 void ibv_wr_set_ud_addr(struct ibv_qp_ex *qp, struct ibv_ah *ah,
                         uint32_t remote_qpn, uint32_t remote_qkey);
 void ibv_wr_set_xrc_srqn(struct ibv_qp_ex *qp, uint32_t remote_srqn);
+void ibv_wr_flush(struct ibv_qp_ex *qp, uint32_t rkey, uint64_t remote_addr,
+                  size_t len, uint8_t type, uint8_t level);
 ```
 
 # DESCRIPTION
@@ -145,6 +147,7 @@ ibv_qp_init_attr_ex* (see the EXAMPLE below).
 | LOCAL_INV            | ibv_wr_local_inv()        | UC, RC, XRC_SEND                 | NONE     |
 | RDMA_READ            | ibv_wr_rdma_read()        | RC, XRC_SEND                     | DATA, QP |
 | RDMA_WRITE           | ibv_wr_rdma_write()       | UC, RC, XRC_SEND                 | DATA, QP |
+| FLUSH                | ibv_wr_flush()            | RC, RD, XRC_SEND                 | DATA, QP |
 | RDMA_WRITE_WITH_IMM  | ibv_wr_rdma_write_imm()   | UC, RC, XRC_SEND                 | DATA, QP |
 | SEND                 | ibv_wr_send()             | UD, UC, RC, XRC_SEND, RAW_PACKET | DATA, QP |
 | SEND_WITH_IMM        | ibv_wr_send_imm()         | UD, UC, RC, SRC SEND             | DATA, QP |

diff --git a/libibverbs/verbs.h b/libibverbs/verbs.h
@@ -517,6 +517,7 @@ enum ibv_wc_opcode {
 	IBV_WC_BIND_MW,
 	IBV_WC_LOCAL_INV,
 	IBV_WC_TSO,
+	IBV_WC_FLUSH,
 	IBV_WC_ATOMIC_WRITE = 9,
 /*
  * Set value of IBV_WC_RECV so consumers can test if a completion is a
@@ -951,6 +952,7 @@ enum ibv_qp_create_send_ops_flags {
 	IBV_QP_EX_WITH_BIND_MW			= 1 << 8,
 	IBV_QP_EX_WITH_SEND_WITH_INV		= 1 << 9,
 	IBV_QP_EX_WITH_TSO			= 1 << 10,
+	IBV_QP_EX_WITH_FLUSH			= 1 << 11,
 	IBV_QP_EX_WITH_ATOMIC_WRITE		= 1 << 12,
 };
 
@@ -1098,6 +1100,7 @@ enum ibv_wr_opcode {
 	IBV_WR_SEND_WITH_INV,
 	IBV_WR_TSO,
 	IBV_WR_DRIVER1,
+	IBV_WR_FLUSH = 14,
 	IBV_WR_ATOMIC_WRITE = 15,
 };
 
@@ -1109,6 +1112,16 @@ enum ibv_send_flags {
 	IBV_SEND_IP_CSUM	= 1 << 4
 };
 
+enum ibv_placement_type {
+	IBV_FLUSH_GLOBAL = 1U << 0,
+	IBV_FLUSH_PERSISTENT = 1U << 1,
+};
+
+enum ibv_selectivity_level {
+	IBV_FLUSH_RANGE = 0,
+	IBV_FLUSH_MR,
+};
+
 struct ibv_data_buf {
 	void *addr;
 	size_t length;
@@ -1318,6 +1331,9 @@ struct ibv_qp_ex {
 
 	void (*wr_atomic_write)(struct ibv_qp_ex *qp, uint32_t rkey,
 				uint64_t remote_addr, const void *atomic_wr);
+	void (*wr_flush)(struct ibv_qp_ex *qp, uint32_t rkey,
+			 uint64_t remote_addr, size_t len, uint8_t type,
+			 uint8_t level);
 };
 
 struct ibv_qp_ex *ibv_qp_to_qp_ex(struct ibv_qp *qp);
@@ -1360,6 +1376,13 @@ static inline void ibv_wr_rdma_write(struct ibv_qp_ex *qp, uint32_t rkey,
 	qp->wr_rdma_write(qp, rkey, remote_addr);
 }
 
+static inline void ibv_wr_flush(struct ibv_qp_ex *qp, uint32_t rkey,
+				uint64_t remote_addr, size_t len, uint8_t type,
+				uint8_t level)
+{
+	qp->wr_flush(qp, rkey, remote_addr, len, type, level);
+}
+
 static inline void ibv_wr_rdma_write_imm(struct ibv_qp_ex *qp, uint32_t rkey,
 					 uint64_t remote_addr, __be32 imm_data)
 {

diff --git a/providers/rxe/rxe.c b/providers/rxe/rxe.c
@@ -922,6 +922,31 @@ static void wr_rdma_write(struct ibv_qp_ex *ibqp, uint32_t rkey,
 	advance_qp_cur_index(qp);
 }
 
+static void wr_flush(struct ibv_qp_ex *ibqp, uint32_t rkey,
+		     uint64_t remote_addr, size_t length,
+		     uint8_t type, uint8_t level)
+{
+	struct rxe_qp *qp = container_of(ibqp, struct rxe_qp, vqp.qp_ex);
+	struct rxe_send_wqe *wqe = addr_from_index(qp->sq.queue, qp->cur_index);
+
+	if (check_qp_queue_full(qp))
+		return;
+
+	memset(wqe, 0, sizeof(*wqe));
+
+	wqe->wr.wr_id = qp->vqp.qp_ex.wr_id;
+	wqe->wr.opcode = IBV_WR_FLUSH;
+	wqe->wr.send_flags = qp->vqp.qp_ex.wr_flags;
+	wqe->wr.wr.flush.remote_addr = remote_addr;
+	wqe->wr.wr.flush.rkey = rkey;
+	wqe->wr.wr.flush.type = type;
+	wqe->wr.wr.flush.level = level;
+	wqe->iova = remote_addr;
+	wqe->ssn = qp->ssn++;
+
+	advance_qp_cur_index(qp);
+}
+
 static void wr_atomic_write(struct ibv_qp_ex *ibqp, uint32_t rkey,
 			    uint64_t remote_addr, const void *atomic_wr)
 {
@@ -1266,7 +1291,7 @@ enum {
 		IBV_QP_EX_WITH_RDMA_READ | IBV_QP_EX_WITH_ATOMIC_CMP_AND_SWP |
 		IBV_QP_EX_WITH_ATOMIC_FETCH_AND_ADD | IBV_QP_EX_WITH_LOCAL_INV |
 		IBV_QP_EX_WITH_BIND_MW | IBV_QP_EX_WITH_SEND_WITH_INV |
-		IBV_QP_EX_WITH_ATOMIC_WRITE,
+		IBV_QP_EX_WITH_FLUSH | IBV_QP_EX_WITH_ATOMIC_WRITE,
 
 	RXE_SUP_UC_QP_SEND_OPS_FLAGS =
 		IBV_QP_EX_WITH_RDMA_WRITE | IBV_QP_EX_WITH_RDMA_WRITE_WITH_IMM |
@@ -1334,6 +1359,9 @@ static void set_qp_send_ops(struct rxe_qp *qp, uint64_t flags)
 	if (flags & IBV_QP_EX_WITH_RDMA_WRITE)
 		qp->vqp.qp_ex.wr_rdma_write = wr_rdma_write;
 
+	if (flags & IBV_QP_EX_WITH_FLUSH)
+		qp->vqp.qp_ex.wr_flush = wr_flush;
+
 	if (flags & IBV_QP_EX_WITH_RDMA_WRITE_WITH_IMM)
 		qp->vqp.qp_ex.wr_rdma_write_imm = wr_rdma_write_imm;