2 * Copyright (c) 2005-2006 Network Appliance, Inc. All rights reserved.
4 * This software is available to you under a choice of one of two
5 * licenses. You may choose to be licensed under the terms of the GNU
6 * General Public License (GPL) Version 2, available from the file
7 * COPYING in the main directory of this source tree, or the BSD-type
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
14 * Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
17 * Redistributions in binary form must reproduce the above
18 * copyright notice, this list of conditions and the following
19 * disclaimer in the documentation and/or other materials provided
20 * with the distribution.
22 * Neither the name of the Network Appliance, Inc. nor the names of
23 * its contributors may be used to endorse or promote products
24 * derived from this software without specific prior written
27 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
28 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
29 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
30 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
31 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
32 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
33 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
34 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
35 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
36 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
37 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
39 * Author: Tom Tucker <tom@opengridcomputing.com>
44 #include <linux/sunrpc/xdr.h>
45 #include <linux/sunrpc/svcsock.h>
46 #include <linux/sunrpc/rpc_rdma.h>
47 #include <rdma/ib_verbs.h>
48 #include <rdma/rdma_cm.h>
51 /* Default and maximum inline threshold sizes */
53 RPCRDMA_DEF_INLINE_THRESH = 4096,
54 RPCRDMA_MAX_INLINE_THRESH = 65536
57 /* RPC/RDMA parameters and stats */
58 extern unsigned int svcrdma_ord;
59 extern unsigned int svcrdma_max_requests;
60 extern unsigned int svcrdma_max_bc_requests;
61 extern unsigned int svcrdma_max_req_size;
63 extern atomic_t rdma_stat_recv;
64 extern atomic_t rdma_stat_read;
65 extern atomic_t rdma_stat_write;
66 extern atomic_t rdma_stat_sq_starve;
67 extern atomic_t rdma_stat_rq_starve;
68 extern atomic_t rdma_stat_rq_poll;
69 extern atomic_t rdma_stat_rq_prod;
70 extern atomic_t rdma_stat_sq_poll;
71 extern atomic_t rdma_stat_sq_prod;
74 * Contexts are built when an RDMA request is created and are a
75 * record of the resources that can be recovered when the request
78 struct svc_rdma_op_ctxt {
79 struct list_head list;
80 struct svc_rdma_op_ctxt *read_hdr;
81 struct svc_rdma_fastreg_mr *frmr;
86 struct svcxprt_rdma *xprt;
88 enum dma_data_direction direction;
90 unsigned int mapped_sges;
91 struct ib_send_wr send_wr;
92 struct ib_sge sge[1 + RPCRDMA_MAX_INLINE_THRESH / PAGE_SIZE];
93 struct page *pages[RPCSVC_MAXPAGES];
96 struct svc_rdma_fastreg_mr {
98 struct scatterlist *sg;
100 unsigned long access_flags;
101 enum dma_data_direction direction;
102 struct list_head frmr_list;
105 #define RDMACTXT_F_LAST_CTXT 2
107 #define SVCRDMA_DEVCAP_FAST_REG 1 /* fast mr registration */
108 #define SVCRDMA_DEVCAP_READ_W_INV 2 /* read w/ invalidate */
110 struct svcxprt_rdma {
111 struct svc_xprt sc_xprt; /* SVC transport structure */
112 struct rdma_cm_id *sc_cm_id; /* RDMA connection id */
113 struct list_head sc_accept_q; /* Conn. waiting accept */
114 int sc_ord; /* RDMA read limit */
116 bool sc_snd_w_inv; /* OK to use Send With Invalidate */
118 atomic_t sc_sq_avail; /* SQEs ready to be consumed */
119 unsigned int sc_sq_depth; /* Depth of SQ */
120 unsigned int sc_rq_depth; /* Depth of RQ */
121 __be32 sc_fc_credits; /* Forward credits */
122 u32 sc_max_requests; /* Max requests */
123 u32 sc_max_bc_requests;/* Backward credits */
124 int sc_max_req_size; /* Size of each RQ WR buf */
129 spinlock_t sc_ctxt_lock;
130 struct list_head sc_ctxts;
132 spinlock_t sc_rw_ctxt_lock;
133 struct list_head sc_rw_ctxts;
135 struct list_head sc_rq_dto_q;
136 spinlock_t sc_rq_dto_lock;
138 struct ib_cq *sc_rq_cq;
139 struct ib_cq *sc_sq_cq;
140 u32 sc_dev_caps; /* distilled device caps */
141 unsigned int sc_frmr_pg_list_len;
142 struct list_head sc_frmr_q;
143 spinlock_t sc_frmr_q_lock;
145 spinlock_t sc_lock; /* transport lock */
147 wait_queue_head_t sc_send_wait; /* SQ exhaustion waitlist */
148 unsigned long sc_flags;
149 struct list_head sc_read_complete_q;
150 struct work_struct sc_work;
153 #define RDMAXPRT_CONN_PENDING 3
155 #define RPCRDMA_LISTEN_BACKLOG 10
156 /* The default ORD value is based on two outstanding full-size writes with a
157 * page size of 4k, or 32k * 2 ops / 4k = 16 outstanding RDMA_READ. */
158 #define RPCRDMA_ORD (64/4)
159 #define RPCRDMA_MAX_REQUESTS 32
161 /* Typical ULP usage of BC requests is NFSv4.1 backchannel. Our
162 * current NFSv4.1 implementation supports one backchannel slot.
164 #define RPCRDMA_MAX_BC_REQUESTS 2
166 #define RPCSVC_MAXPAYLOAD_RDMA RPCSVC_MAXPAYLOAD
168 /* Track DMA maps for this transport and context */
169 static inline void svc_rdma_count_mappings(struct svcxprt_rdma *rdma,
170 struct svc_rdma_op_ctxt *ctxt)
175 /* svc_rdma_backchannel.c */
176 extern int svc_rdma_handle_bc_reply(struct rpc_xprt *xprt,
178 struct xdr_buf *rcvbuf);
180 /* svc_rdma_recvfrom.c */
181 extern int svc_rdma_recvfrom(struct svc_rqst *);
184 extern void svc_rdma_destroy_rw_ctxts(struct svcxprt_rdma *rdma);
185 extern int svc_rdma_recv_read_chunk(struct svcxprt_rdma *rdma,
186 struct svc_rqst *rqstp,
187 struct svc_rdma_op_ctxt *head, __be32 *p);
188 extern int svc_rdma_send_write_chunk(struct svcxprt_rdma *rdma,
189 __be32 *wr_ch, struct xdr_buf *xdr);
190 extern int svc_rdma_send_reply_chunk(struct svcxprt_rdma *rdma,
191 __be32 *rp_ch, bool writelist,
192 struct xdr_buf *xdr);
194 /* svc_rdma_sendto.c */
195 extern int svc_rdma_map_reply_hdr(struct svcxprt_rdma *rdma,
196 struct svc_rdma_op_ctxt *ctxt,
197 __be32 *rdma_resp, unsigned int len);
198 extern int svc_rdma_post_send_wr(struct svcxprt_rdma *rdma,
199 struct svc_rdma_op_ctxt *ctxt,
200 int num_sge, u32 inv_rkey);
201 extern int svc_rdma_sendto(struct svc_rqst *);
203 /* svc_rdma_transport.c */
204 extern void svc_rdma_wc_send(struct ib_cq *, struct ib_wc *);
205 extern void svc_rdma_wc_reg(struct ib_cq *, struct ib_wc *);
206 extern void svc_rdma_wc_read(struct ib_cq *, struct ib_wc *);
207 extern void svc_rdma_wc_inv(struct ib_cq *, struct ib_wc *);
208 extern int svc_rdma_send(struct svcxprt_rdma *, struct ib_send_wr *);
209 extern int svc_rdma_post_recv(struct svcxprt_rdma *, gfp_t);
210 extern int svc_rdma_repost_recv(struct svcxprt_rdma *, gfp_t);
211 extern int svc_rdma_create_listen(struct svc_serv *, int, struct sockaddr *);
212 extern struct svc_rdma_op_ctxt *svc_rdma_get_context(struct svcxprt_rdma *);
213 extern void svc_rdma_put_context(struct svc_rdma_op_ctxt *, int);
214 extern void svc_rdma_unmap_dma(struct svc_rdma_op_ctxt *ctxt);
215 extern struct svc_rdma_fastreg_mr *svc_rdma_get_frmr(struct svcxprt_rdma *);
216 extern void svc_rdma_put_frmr(struct svcxprt_rdma *,
217 struct svc_rdma_fastreg_mr *);
218 extern void svc_sq_reap(struct svcxprt_rdma *);
219 extern void svc_rq_reap(struct svcxprt_rdma *);
220 extern void svc_rdma_prep_reply_hdr(struct svc_rqst *);
222 extern struct svc_xprt_class svc_rdma_class;
223 #ifdef CONFIG_SUNRPC_BACKCHANNEL
224 extern struct svc_xprt_class svc_rdma_bc_class;
228 extern struct workqueue_struct *svc_rdma_wq;
229 extern int svc_rdma_init(void);
230 extern void svc_rdma_cleanup(void);