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 xdr_buf arg;
81 struct ib_cqe cqe;
82 u32 byte_len;
83 struct svcxprt_rdma *xprt;
84 enum dma_data_direction direction;
85 int count;
86 unsigned int mapped_sges;
87 int hdr_count;
88 struct ib_send_wr send_wr;
89 struct ib_sge sge[1 + RPCRDMA_MAX_INLINE_THRESH / PAGE_SIZE];
90 struct page *pages[RPCSVC_MAXPAGES];
91};
92
93struct svcxprt_rdma {
94 struct svc_xprt sc_xprt; /* SVC transport structure */
95 struct rdma_cm_id *sc_cm_id; /* RDMA connection id */
96 struct list_head sc_accept_q; /* Conn. waiting accept */
97 int sc_ord; /* RDMA read limit */
98 int sc_max_sge;
99 bool sc_snd_w_inv; /* OK to use Send With Invalidate */
100
101 atomic_t sc_sq_avail; /* SQEs ready to be consumed */
102 unsigned int sc_sq_depth; /* Depth of SQ */
103 unsigned int sc_rq_depth; /* Depth of RQ */
104 __be32 sc_fc_credits; /* Forward credits */
105 u32 sc_max_requests; /* Max requests */
106 u32 sc_max_bc_requests;/* Backward credits */
107 int sc_max_req_size; /* Size of each RQ WR buf */
108 u8 sc_port_num;
109
110 struct ib_pd *sc_pd;
111
112 spinlock_t sc_ctxt_lock;
113 struct list_head sc_ctxts;
114 int sc_ctxt_used;
115 spinlock_t sc_rw_ctxt_lock;
116 struct list_head sc_rw_ctxts;
117
118 struct list_head sc_rq_dto_q;
119 spinlock_t sc_rq_dto_lock;
120 struct ib_qp *sc_qp;
121 struct ib_cq *sc_rq_cq;
122 struct ib_cq *sc_sq_cq;
123
124 spinlock_t sc_lock; /* transport lock */
125
126 wait_queue_head_t sc_send_wait; /* SQ exhaustion waitlist */
127 unsigned long sc_flags;
128 struct list_head sc_read_complete_q;
129 struct work_struct sc_work;
130};
131/* sc_flags */
132#define RDMAXPRT_CONN_PENDING 3
133
134#define RPCRDMA_LISTEN_BACKLOG 10
135/* The default ORD value is based on two outstanding full-size writes with a
136 * page size of 4k, or 32k * 2 ops / 4k = 16 outstanding RDMA_READ. */
137#define RPCRDMA_ORD (64/4)
138#define RPCRDMA_MAX_REQUESTS 32
139
140/* Typical ULP usage of BC requests is NFSv4.1 backchannel. Our
141 * current NFSv4.1 implementation supports one backchannel slot.
142 */
143#define RPCRDMA_MAX_BC_REQUESTS 2
144
145#define RPCSVC_MAXPAYLOAD_RDMA RPCSVC_MAXPAYLOAD
146
147/* Track DMA maps for this transport and context */
148static inline void svc_rdma_count_mappings(struct svcxprt_rdma *rdma,
149 struct svc_rdma_op_ctxt *ctxt)
150{
151 ctxt->mapped_sges++;
152}
153
154/* svc_rdma_backchannel.c */
155extern int svc_rdma_handle_bc_reply(struct rpc_xprt *xprt,
156 __be32 *rdma_resp,
157 struct xdr_buf *rcvbuf);
158
159/* svc_rdma_recvfrom.c */
160extern int svc_rdma_recvfrom(struct svc_rqst *);
161
162/* svc_rdma_rw.c */
163extern void svc_rdma_destroy_rw_ctxts(struct svcxprt_rdma *rdma);
164extern int svc_rdma_recv_read_chunk(struct svcxprt_rdma *rdma,
165 struct svc_rqst *rqstp,
166 struct svc_rdma_op_ctxt *head, __be32 *p);
167extern int svc_rdma_send_write_chunk(struct svcxprt_rdma *rdma,
168 __be32 *wr_ch, struct xdr_buf *xdr);
169extern int svc_rdma_send_reply_chunk(struct svcxprt_rdma *rdma,
170 __be32 *rp_ch, bool writelist,
171 struct xdr_buf *xdr);
172
173/* svc_rdma_sendto.c */
174extern int svc_rdma_map_reply_hdr(struct svcxprt_rdma *rdma,
175 struct svc_rdma_op_ctxt *ctxt,
176 __be32 *rdma_resp, unsigned int len);
177extern int svc_rdma_post_send_wr(struct svcxprt_rdma *rdma,
178 struct svc_rdma_op_ctxt *ctxt,
179 int num_sge, u32 inv_rkey);
180extern int svc_rdma_sendto(struct svc_rqst *);
181
182/* svc_rdma_transport.c */
183extern void svc_rdma_wc_send(struct ib_cq *, struct ib_wc *);
184extern void svc_rdma_wc_reg(struct ib_cq *, struct ib_wc *);
185extern void svc_rdma_wc_read(struct ib_cq *, struct ib_wc *);
186extern void svc_rdma_wc_inv(struct ib_cq *, struct ib_wc *);
187extern int svc_rdma_send(struct svcxprt_rdma *, struct ib_send_wr *);
188extern int svc_rdma_create_listen(struct svc_serv *, int, struct sockaddr *);
189extern struct svc_rdma_op_ctxt *svc_rdma_get_context(struct svcxprt_rdma *);
190extern void svc_rdma_put_context(struct svc_rdma_op_ctxt *, int);
191extern void svc_rdma_unmap_dma(struct svc_rdma_op_ctxt *ctxt);
192extern void svc_sq_reap(struct svcxprt_rdma *);
193extern void svc_rq_reap(struct svcxprt_rdma *);
194extern void svc_rdma_prep_reply_hdr(struct svc_rqst *);
195
196extern struct svc_xprt_class svc_rdma_class;
197#ifdef CONFIG_SUNRPC_BACKCHANNEL
198extern struct svc_xprt_class svc_rdma_bc_class;
199#endif
200
201/* svc_rdma.c */
202extern struct workqueue_struct *svc_rdma_wq;
203extern int svc_rdma_init(void);
204extern void svc_rdma_cleanup(void);
205
206#endif