1 /* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */
2 /*
3 * Copyright (c) 2005-2006 Network Appliance, Inc. All rights reserved.
4 *
5 * This software is available to you under a choice of one of two
6 * licenses. You may choose to be licensed under the terms of the GNU
7 * General Public License (GPL) Version 2, available from the file
8 * COPYING in the main directory of this source tree, or the BSD-type
9 * license below:
10 *
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
13 * are met:
14 *
15 * Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 *
18 * Redistributions in binary form must reproduce the above
19 * copyright notice, this list of conditions and the following
20 * disclaimer in the documentation and/or other materials provided
21 * with the distribution.
22 *
23 * Neither the name of the Network Appliance, Inc. nor the names of
24 * its contributors may be used to endorse or promote products
25 * derived from this software without specific prior written
26 * permission.
27 *
28 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
29 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
30 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
31 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
32 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
33 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
34 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
35 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
36 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
37 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
38 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
39 *
40 * Author: Tom Tucker <tom@opengridcomputing.com>
41 */
42
43 #ifndef SVC_RDMA_H
44 #define SVC_RDMA_H
45 #include <linux/llist.h>
46 #include <linux/sunrpc/xdr.h>
47 #include <linux/sunrpc/svcsock.h>
48 #include <linux/sunrpc/rpc_rdma.h>
49 #include <linux/sunrpc/rpc_rdma_cid.h>
50 #include <linux/sunrpc/svc_rdma_pcl.h>
51
52 #include <linux/percpu_counter.h>
53 #include <rdma/ib_verbs.h>
54 #include <rdma/rdma_cm.h>
55
56 /* Default and maximum inline threshold sizes */
57 enum {
58 RPCRDMA_PULLUP_THRESH = RPCRDMA_V1_DEF_INLINE_SIZE >> 1,
59 RPCRDMA_DEF_INLINE_THRESH = 4096,
60 RPCRDMA_MAX_INLINE_THRESH = 65536
61 };
62
63 /* RPC/RDMA parameters and stats */
64 extern unsigned int svcrdma_ord;
65 extern unsigned int svcrdma_max_requests;
66 extern unsigned int svcrdma_max_bc_requests;
67 extern unsigned int svcrdma_max_req_size;
68 extern struct workqueue_struct *svcrdma_wq;
69
70 extern struct percpu_counter svcrdma_stat_read;
71 extern struct percpu_counter svcrdma_stat_recv;
72 extern struct percpu_counter svcrdma_stat_sq_starve;
73 extern struct percpu_counter svcrdma_stat_write;
74
75 struct svcxprt_rdma {
76 struct svc_xprt sc_xprt; /* SVC transport structure */
77 struct rdma_cm_id *sc_cm_id; /* RDMA connection id */
78 struct list_head sc_accept_q; /* Conn. waiting accept */
79 int sc_ord; /* RDMA read limit */
80 int sc_max_send_sges;
81 bool sc_snd_w_inv; /* OK to use Send With Invalidate */
82
83 atomic_t sc_sq_avail; /* SQEs ready to be consumed */
84 unsigned int sc_sq_depth; /* Depth of SQ */
85 __be32 sc_fc_credits; /* Forward credits */
86 u32 sc_max_requests; /* Max requests */
87 u32 sc_max_bc_requests;/* Backward credits */
88 int sc_max_req_size; /* Size of each RQ WR buf */
89 u8 sc_port_num;
90
91 struct ib_pd *sc_pd;
92
93 spinlock_t sc_send_lock;
94 struct llist_head sc_send_ctxts;
95 spinlock_t sc_rw_ctxt_lock;
96 struct llist_head sc_rw_ctxts;
97
98 u32 sc_pending_recvs;
99 u32 sc_recv_batch;
100 struct list_head sc_rq_dto_q;
101 struct list_head sc_read_complete_q;
102 spinlock_t sc_rq_dto_lock;
103 struct ib_qp *sc_qp;
104 struct ib_cq *sc_rq_cq;
105 struct ib_cq *sc_sq_cq;
106
107 spinlock_t sc_lock; /* transport lock */
108
109 wait_queue_head_t sc_send_wait; /* SQ exhaustion waitlist */
110 unsigned long sc_flags;
111 struct work_struct sc_work;
112
113 struct llist_head sc_recv_ctxts;
114
115 atomic_t sc_completion_ids;
116 };
117 /* sc_flags */
118 #define RDMAXPRT_CONN_PENDING 3
119
svc_rdma_rqst_rdma(struct svc_rqst * rqstp)120 static inline struct svcxprt_rdma *svc_rdma_rqst_rdma(struct svc_rqst *rqstp)
121 {
122 struct svc_xprt *xprt = rqstp->rq_xprt;
123
124 return container_of(xprt, struct svcxprt_rdma, sc_xprt);
125 }
126
127 /*
128 * Default connection parameters
129 */
130 enum {
131 RPCRDMA_LISTEN_BACKLOG = 10,
132 RPCRDMA_MAX_REQUESTS = 64,
133 RPCRDMA_MAX_BC_REQUESTS = 2,
134 };
135
136 #define RPCSVC_MAXPAYLOAD_RDMA RPCSVC_MAXPAYLOAD
137
138 /**
139 * svc_rdma_send_cid_init - Initialize a Receive Queue completion ID
140 * @rdma: controlling transport
141 * @cid: completion ID to initialize
142 */
svc_rdma_recv_cid_init(struct svcxprt_rdma * rdma,struct rpc_rdma_cid * cid)143 static inline void svc_rdma_recv_cid_init(struct svcxprt_rdma *rdma,
144 struct rpc_rdma_cid *cid)
145 {
146 cid->ci_queue_id = rdma->sc_rq_cq->res.id;
147 cid->ci_completion_id = atomic_inc_return(&rdma->sc_completion_ids);
148 }
149
150 /**
151 * svc_rdma_send_cid_init - Initialize a Send Queue completion ID
152 * @rdma: controlling transport
153 * @cid: completion ID to initialize
154 */
svc_rdma_send_cid_init(struct svcxprt_rdma * rdma,struct rpc_rdma_cid * cid)155 static inline void svc_rdma_send_cid_init(struct svcxprt_rdma *rdma,
156 struct rpc_rdma_cid *cid)
157 {
158 cid->ci_queue_id = rdma->sc_sq_cq->res.id;
159 cid->ci_completion_id = atomic_inc_return(&rdma->sc_completion_ids);
160 }
161
162 /*
163 * A chunk context tracks all I/O for moving one Read or Write
164 * chunk. This is a set of rdma_rw's that handle data movement
165 * for all segments of one chunk.
166 */
167 struct svc_rdma_chunk_ctxt {
168 struct rpc_rdma_cid cc_cid;
169 struct ib_cqe cc_cqe;
170 struct list_head cc_rwctxts;
171 ktime_t cc_posttime;
172 int cc_sqecount;
173 };
174
175 struct svc_rdma_recv_ctxt {
176 struct llist_node rc_node;
177 struct list_head rc_list;
178 struct ib_recv_wr rc_recv_wr;
179 struct ib_cqe rc_cqe;
180 struct rpc_rdma_cid rc_cid;
181 struct ib_sge rc_recv_sge;
182 void *rc_recv_buf;
183 struct xdr_stream rc_stream;
184 u32 rc_byte_len;
185 u32 rc_inv_rkey;
186 __be32 rc_msgtype;
187
188 /* State for pulling a Read chunk */
189 unsigned int rc_pageoff;
190 unsigned int rc_curpage;
191 unsigned int rc_readbytes;
192 struct xdr_buf rc_saved_arg;
193 struct svc_rdma_chunk_ctxt rc_cc;
194
195 struct svc_rdma_pcl rc_call_pcl;
196
197 struct svc_rdma_pcl rc_read_pcl;
198 struct svc_rdma_chunk *rc_cur_result_payload;
199 struct svc_rdma_pcl rc_write_pcl;
200 struct svc_rdma_pcl rc_reply_pcl;
201
202 unsigned int rc_page_count;
203 struct page *rc_pages[RPCSVC_MAXPAGES];
204 };
205
206 struct svc_rdma_send_ctxt {
207 struct llist_node sc_node;
208 struct rpc_rdma_cid sc_cid;
209 struct work_struct sc_work;
210
211 struct svcxprt_rdma *sc_rdma;
212 struct ib_send_wr sc_send_wr;
213 struct ib_cqe sc_cqe;
214 struct xdr_buf sc_hdrbuf;
215 struct xdr_stream sc_stream;
216 void *sc_xprt_buf;
217 int sc_page_count;
218 int sc_cur_sge_no;
219 struct page *sc_pages[RPCSVC_MAXPAGES];
220 struct ib_sge sc_sges[];
221 };
222
223 /* svc_rdma_backchannel.c */
224 extern void svc_rdma_handle_bc_reply(struct svc_rqst *rqstp,
225 struct svc_rdma_recv_ctxt *rctxt);
226
227 /* svc_rdma_recvfrom.c */
228 extern void svc_rdma_recv_ctxts_destroy(struct svcxprt_rdma *rdma);
229 extern bool svc_rdma_post_recvs(struct svcxprt_rdma *rdma);
230 extern struct svc_rdma_recv_ctxt *
231 svc_rdma_recv_ctxt_get(struct svcxprt_rdma *rdma);
232 extern void svc_rdma_recv_ctxt_put(struct svcxprt_rdma *rdma,
233 struct svc_rdma_recv_ctxt *ctxt);
234 extern void svc_rdma_flush_recv_queues(struct svcxprt_rdma *rdma);
235 extern void svc_rdma_release_ctxt(struct svc_xprt *xprt, void *ctxt);
236 extern int svc_rdma_recvfrom(struct svc_rqst *);
237
238 /* svc_rdma_rw.c */
239 extern void svc_rdma_destroy_rw_ctxts(struct svcxprt_rdma *rdma);
240 extern void svc_rdma_cc_init(struct svcxprt_rdma *rdma,
241 struct svc_rdma_chunk_ctxt *cc);
242 extern void svc_rdma_cc_release(struct svcxprt_rdma *rdma,
243 struct svc_rdma_chunk_ctxt *cc,
244 enum dma_data_direction dir);
245 extern int svc_rdma_send_write_chunk(struct svcxprt_rdma *rdma,
246 const struct svc_rdma_chunk *chunk,
247 const struct xdr_buf *xdr);
248 extern int svc_rdma_send_reply_chunk(struct svcxprt_rdma *rdma,
249 const struct svc_rdma_recv_ctxt *rctxt,
250 const struct xdr_buf *xdr);
251 extern int svc_rdma_process_read_list(struct svcxprt_rdma *rdma,
252 struct svc_rqst *rqstp,
253 struct svc_rdma_recv_ctxt *head);
254
255 /* svc_rdma_sendto.c */
256 extern void svc_rdma_send_ctxts_destroy(struct svcxprt_rdma *rdma);
257 extern struct svc_rdma_send_ctxt *
258 svc_rdma_send_ctxt_get(struct svcxprt_rdma *rdma);
259 extern void svc_rdma_send_ctxt_put(struct svcxprt_rdma *rdma,
260 struct svc_rdma_send_ctxt *ctxt);
261 extern int svc_rdma_send(struct svcxprt_rdma *rdma,
262 struct svc_rdma_send_ctxt *ctxt);
263 extern int svc_rdma_map_reply_msg(struct svcxprt_rdma *rdma,
264 struct svc_rdma_send_ctxt *sctxt,
265 const struct svc_rdma_pcl *write_pcl,
266 const struct svc_rdma_pcl *reply_pcl,
267 const struct xdr_buf *xdr);
268 extern void svc_rdma_send_error_msg(struct svcxprt_rdma *rdma,
269 struct svc_rdma_send_ctxt *sctxt,
270 struct svc_rdma_recv_ctxt *rctxt,
271 int status);
272 extern void svc_rdma_wake_send_waiters(struct svcxprt_rdma *rdma, int avail);
273 extern int svc_rdma_sendto(struct svc_rqst *);
274 extern int svc_rdma_result_payload(struct svc_rqst *rqstp, unsigned int offset,
275 unsigned int length);
276
277 /* svc_rdma_transport.c */
278 extern struct svc_xprt_class svc_rdma_class;
279 #ifdef CONFIG_SUNRPC_BACKCHANNEL
280 extern struct svc_xprt_class svc_rdma_bc_class;
281 #endif
282
283 /* svc_rdma.c */
284 extern int svc_rdma_init(void);
285 extern void svc_rdma_cleanup(void);
286
287 #endif
288