tjh.dev
Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
1/*
2 * Copyright (c) 2005-2006 Network Appliance, Inc. All rights reserved.
3 *
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
8 * license below:
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 *
14 * Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 *
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.
21 *
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
25 * permission.
26 *
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.
38 *
39 * Author: Tom Tucker <tom@opengridcomputing.com>
40 */
41
42#ifndef SVC_RDMA_H
43#define SVC_RDMA_H
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>
49#define SVCRDMA_DEBUG
50
51/* Default and maximum inline threshold sizes */
52enum {
53 RPCRDMA_DEF_INLINE_THRESH = 4096,
54 RPCRDMA_MAX_INLINE_THRESH = 65536
55};
56
57/* RPC/RDMA parameters and stats */
58extern unsigned int svcrdma_ord;
59extern unsigned int svcrdma_max_requests;
60extern unsigned int svcrdma_max_bc_requests;
61extern unsigned int svcrdma_max_req_size;
62
63extern atomic_t rdma_stat_recv;
64extern atomic_t rdma_stat_read;
65extern atomic_t rdma_stat_write;
66extern atomic_t rdma_stat_sq_starve;
67extern atomic_t rdma_stat_rq_starve;
68extern atomic_t rdma_stat_rq_poll;
69extern atomic_t rdma_stat_rq_prod;
70extern atomic_t rdma_stat_sq_poll;
71extern atomic_t rdma_stat_sq_prod;
72
73/*
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
76 * completes.
77 */
78struct svc_rdma_op_ctxt {
79 struct list_head list;
80 struct svc_rdma_op_ctxt *read_hdr;
81 struct svc_rdma_fastreg_mr *frmr;
82 int hdr_count;
83 struct xdr_buf arg;
84 struct ib_cqe cqe;
85 struct ib_cqe reg_cqe;
86 struct ib_cqe inv_cqe;
87 u32 byte_len;
88 u32 position;
89 struct svcxprt_rdma *xprt;
90 unsigned long flags;
91 enum dma_data_direction direction;
92 int count;
93 unsigned int mapped_sges;
94 struct ib_send_wr send_wr;
95 struct ib_sge sge[1 + RPCRDMA_MAX_INLINE_THRESH / PAGE_SIZE];
96 struct page *pages[RPCSVC_MAXPAGES];
97};
98
99struct svc_rdma_fastreg_mr {
100 struct ib_mr *mr;
101 struct scatterlist *sg;
102 int sg_nents;
103 unsigned long access_flags;
104 enum dma_data_direction direction;
105 struct list_head frmr_list;
106};
107
108#define RDMACTXT_F_LAST_CTXT 2
109
110#define SVCRDMA_DEVCAP_FAST_REG 1 /* fast mr registration */
111#define SVCRDMA_DEVCAP_READ_W_INV 2 /* read w/ invalidate */
112
113struct svcxprt_rdma {
114 struct svc_xprt sc_xprt; /* SVC transport structure */
115 struct rdma_cm_id *sc_cm_id; /* RDMA connection id */
116 struct list_head sc_accept_q; /* Conn. waiting accept */
117 int sc_ord; /* RDMA read limit */
118 int sc_max_sge;
119 int sc_max_sge_rd; /* max sge for read target */
120 bool sc_snd_w_inv; /* OK to use Send With Invalidate */
121
122 atomic_t sc_sq_avail; /* SQEs ready to be consumed */
123 unsigned int sc_sq_depth; /* Depth of SQ */
124 unsigned int sc_rq_depth; /* Depth of RQ */
125 __be32 sc_fc_credits; /* Forward credits */
126 u32 sc_max_requests; /* Max requests */
127 u32 sc_max_bc_requests;/* Backward credits */
128 int sc_max_req_size; /* Size of each RQ WR buf */
129 u8 sc_port_num;
130
131 struct ib_pd *sc_pd;
132
133 spinlock_t sc_ctxt_lock;
134 struct list_head sc_ctxts;
135 int sc_ctxt_used;
136 spinlock_t sc_rw_ctxt_lock;
137 struct list_head sc_rw_ctxts;
138
139 struct list_head sc_rq_dto_q;
140 spinlock_t sc_rq_dto_lock;
141 struct ib_qp *sc_qp;
142 struct ib_cq *sc_rq_cq;
143 struct ib_cq *sc_sq_cq;
144 int (*sc_reader)(struct svcxprt_rdma *,
145 struct svc_rqst *,
146 struct svc_rdma_op_ctxt *,
147 int *, u32 *, u32, u32, u64, bool);
148 u32 sc_dev_caps; /* distilled device caps */
149 unsigned int sc_frmr_pg_list_len;
150 struct list_head sc_frmr_q;
151 spinlock_t sc_frmr_q_lock;
152
153 spinlock_t sc_lock; /* transport lock */
154
155 wait_queue_head_t sc_send_wait; /* SQ exhaustion waitlist */
156 unsigned long sc_flags;
157 struct list_head sc_read_complete_q;
158 struct work_struct sc_work;
159};
160/* sc_flags */
161#define RDMAXPRT_CONN_PENDING 3
162
163#define RPCRDMA_LISTEN_BACKLOG 10
164/* The default ORD value is based on two outstanding full-size writes with a
165 * page size of 4k, or 32k * 2 ops / 4k = 16 outstanding RDMA_READ. */
166#define RPCRDMA_ORD (64/4)
167#define RPCRDMA_MAX_REQUESTS 32
168
169/* Typical ULP usage of BC requests is NFSv4.1 backchannel. Our
170 * current NFSv4.1 implementation supports one backchannel slot.
171 */
172#define RPCRDMA_MAX_BC_REQUESTS 2
173
174#define RPCSVC_MAXPAYLOAD_RDMA RPCSVC_MAXPAYLOAD
175
176/* Track DMA maps for this transport and context */
177static inline void svc_rdma_count_mappings(struct svcxprt_rdma *rdma,
178 struct svc_rdma_op_ctxt *ctxt)
179{
180 ctxt->mapped_sges++;
181}
182
183/* svc_rdma_backchannel.c */
184extern int svc_rdma_handle_bc_reply(struct rpc_xprt *xprt,
185 __be32 *rdma_resp,
186 struct xdr_buf *rcvbuf);
187
188/* svc_rdma_marshal.c */
189extern int svc_rdma_xdr_decode_req(struct xdr_buf *);
190
191/* svc_rdma_recvfrom.c */
192extern int svc_rdma_recvfrom(struct svc_rqst *);
193extern int rdma_read_chunk_lcl(struct svcxprt_rdma *, struct svc_rqst *,
194 struct svc_rdma_op_ctxt *, int *, u32 *,
195 u32, u32, u64, bool);
196extern int rdma_read_chunk_frmr(struct svcxprt_rdma *, struct svc_rqst *,
197 struct svc_rdma_op_ctxt *, int *, u32 *,
198 u32, u32, u64, bool);
199
200/* svc_rdma_rw.c */
201extern void svc_rdma_destroy_rw_ctxts(struct svcxprt_rdma *rdma);
202extern int svc_rdma_send_write_chunk(struct svcxprt_rdma *rdma,
203 __be32 *wr_ch, struct xdr_buf *xdr);
204extern int svc_rdma_send_reply_chunk(struct svcxprt_rdma *rdma,
205 __be32 *rp_ch, bool writelist,
206 struct xdr_buf *xdr);
207
208/* svc_rdma_sendto.c */
209extern int svc_rdma_map_reply_hdr(struct svcxprt_rdma *rdma,
210 struct svc_rdma_op_ctxt *ctxt,
211 __be32 *rdma_resp, unsigned int len);
212extern int svc_rdma_post_send_wr(struct svcxprt_rdma *rdma,
213 struct svc_rdma_op_ctxt *ctxt,
214 int num_sge, u32 inv_rkey);
215extern int svc_rdma_sendto(struct svc_rqst *);
216
217/* svc_rdma_transport.c */
218extern void svc_rdma_wc_send(struct ib_cq *, struct ib_wc *);
219extern void svc_rdma_wc_reg(struct ib_cq *, struct ib_wc *);
220extern void svc_rdma_wc_read(struct ib_cq *, struct ib_wc *);
221extern void svc_rdma_wc_inv(struct ib_cq *, struct ib_wc *);
222extern int svc_rdma_send(struct svcxprt_rdma *, struct ib_send_wr *);
223extern int svc_rdma_post_recv(struct svcxprt_rdma *, gfp_t);
224extern int svc_rdma_repost_recv(struct svcxprt_rdma *, gfp_t);
225extern int svc_rdma_create_listen(struct svc_serv *, int, struct sockaddr *);
226extern struct svc_rdma_op_ctxt *svc_rdma_get_context(struct svcxprt_rdma *);
227extern void svc_rdma_put_context(struct svc_rdma_op_ctxt *, int);
228extern void svc_rdma_unmap_dma(struct svc_rdma_op_ctxt *ctxt);
229extern struct svc_rdma_fastreg_mr *svc_rdma_get_frmr(struct svcxprt_rdma *);
230extern void svc_rdma_put_frmr(struct svcxprt_rdma *,
231 struct svc_rdma_fastreg_mr *);
232extern void svc_sq_reap(struct svcxprt_rdma *);
233extern void svc_rq_reap(struct svcxprt_rdma *);
234extern void svc_rdma_prep_reply_hdr(struct svc_rqst *);
235
236extern struct svc_xprt_class svc_rdma_class;
237#ifdef CONFIG_SUNRPC_BACKCHANNEL
238extern struct svc_xprt_class svc_rdma_bc_class;
239#endif
240
241/* svc_rdma.c */
242extern struct workqueue_struct *svc_rdma_wq;
243extern int svc_rdma_init(void);
244extern void svc_rdma_cleanup(void);
245
246#endif