Tom Tucker | d21b05f | 2007-12-12 16:13:17 -0600 | [diff] [blame] | 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 | /* RPC/RDMA parameters and stats */ |
| 52 | extern unsigned int svcrdma_ord; |
| 53 | extern unsigned int svcrdma_max_requests; |
| 54 | extern unsigned int svcrdma_max_req_size; |
| 55 | |
| 56 | extern atomic_t rdma_stat_recv; |
| 57 | extern atomic_t rdma_stat_read; |
| 58 | extern atomic_t rdma_stat_write; |
| 59 | extern atomic_t rdma_stat_sq_starve; |
| 60 | extern atomic_t rdma_stat_rq_starve; |
| 61 | extern atomic_t rdma_stat_rq_poll; |
| 62 | extern atomic_t rdma_stat_rq_prod; |
| 63 | extern atomic_t rdma_stat_sq_poll; |
| 64 | extern atomic_t rdma_stat_sq_prod; |
| 65 | |
Tom Tucker | d21b05f | 2007-12-12 16:13:17 -0600 | [diff] [blame] | 66 | /* |
| 67 | * Contexts are built when an RDMA request is created and are a |
| 68 | * record of the resources that can be recovered when the request |
| 69 | * completes. |
| 70 | */ |
| 71 | struct svc_rdma_op_ctxt { |
Tom Tucker | 02e7452 | 2008-04-30 19:50:56 -0500 | [diff] [blame] | 72 | struct svc_rdma_op_ctxt *read_hdr; |
Tom Tucker | 0d3ebb9 | 2008-09-30 13:06:13 -0500 | [diff] [blame] | 73 | struct svc_rdma_fastreg_mr *frmr; |
Tom Tucker | f820c57 | 2008-05-27 17:03:14 -0500 | [diff] [blame] | 74 | int hdr_count; |
Tom Tucker | d21b05f | 2007-12-12 16:13:17 -0600 | [diff] [blame] | 75 | struct xdr_buf arg; |
| 76 | struct list_head dto_q; |
| 77 | enum ib_wr_opcode wr_op; |
| 78 | enum ib_wc_status wc_status; |
| 79 | u32 byte_len; |
Chuck Lever | 0b056c2 | 2015-01-13 11:03:37 -0500 | [diff] [blame] | 80 | u32 position; |
Tom Tucker | d21b05f | 2007-12-12 16:13:17 -0600 | [diff] [blame] | 81 | struct svcxprt_rdma *xprt; |
| 82 | unsigned long flags; |
| 83 | enum dma_data_direction direction; |
| 84 | int count; |
| 85 | struct ib_sge sge[RPCSVC_MAXPAGES]; |
| 86 | struct page *pages[RPCSVC_MAXPAGES]; |
| 87 | }; |
| 88 | |
Tom Tucker | ab96ddd | 2008-05-28 13:54:04 -0500 | [diff] [blame] | 89 | /* |
| 90 | * NFS_ requests are mapped on the client side by the chunk lists in |
| 91 | * the RPCRDMA header. During the fetching of the RPC from the client |
| 92 | * and the writing of the reply to the client, the memory in the |
| 93 | * client and the memory in the server must be mapped as contiguous |
| 94 | * vaddr/len for access by the hardware. These data strucures keep |
| 95 | * these mappings. |
| 96 | * |
| 97 | * For an RDMA_WRITE, the 'sge' maps the RPC REPLY. For RDMA_READ, the |
| 98 | * 'sge' in the svc_rdma_req_map maps the server side RPC reply and the |
| 99 | * 'ch' field maps the read-list of the RPCRDMA header to the 'sge' |
| 100 | * mapping of the reply. |
| 101 | */ |
| 102 | struct svc_rdma_chunk_sge { |
| 103 | int start; /* sge no for this chunk */ |
| 104 | int count; /* sge count for this chunk */ |
| 105 | }; |
Tom Tucker | 0d3ebb9 | 2008-09-30 13:06:13 -0500 | [diff] [blame] | 106 | struct svc_rdma_fastreg_mr { |
| 107 | struct ib_mr *mr; |
| 108 | void *kva; |
| 109 | struct ib_fast_reg_page_list *page_list; |
| 110 | int page_list_len; |
| 111 | unsigned long access_flags; |
| 112 | unsigned long map_len; |
| 113 | enum dma_data_direction direction; |
| 114 | struct list_head frmr_list; |
| 115 | }; |
Tom Tucker | ab96ddd | 2008-05-28 13:54:04 -0500 | [diff] [blame] | 116 | struct svc_rdma_req_map { |
| 117 | unsigned long count; |
| 118 | union { |
| 119 | struct kvec sge[RPCSVC_MAXPAGES]; |
| 120 | struct svc_rdma_chunk_sge ch[RPCSVC_MAXPAGES]; |
Steve Wise | 0bf4828 | 2014-05-28 15:12:01 -0500 | [diff] [blame] | 121 | unsigned long lkey[RPCSVC_MAXPAGES]; |
Tom Tucker | ab96ddd | 2008-05-28 13:54:04 -0500 | [diff] [blame] | 122 | }; |
| 123 | }; |
Tom Tucker | d21b05f | 2007-12-12 16:13:17 -0600 | [diff] [blame] | 124 | #define RDMACTXT_F_LAST_CTXT 2 |
| 125 | |
Tom Tucker | 0d3ebb9 | 2008-09-30 13:06:13 -0500 | [diff] [blame] | 126 | #define SVCRDMA_DEVCAP_FAST_REG 1 /* fast mr registration */ |
| 127 | #define SVCRDMA_DEVCAP_READ_W_INV 2 /* read w/ invalidate */ |
| 128 | |
Tom Tucker | d21b05f | 2007-12-12 16:13:17 -0600 | [diff] [blame] | 129 | struct svcxprt_rdma { |
| 130 | struct svc_xprt sc_xprt; /* SVC transport structure */ |
| 131 | struct rdma_cm_id *sc_cm_id; /* RDMA connection id */ |
| 132 | struct list_head sc_accept_q; /* Conn. waiting accept */ |
| 133 | int sc_ord; /* RDMA read limit */ |
Tom Tucker | d21b05f | 2007-12-12 16:13:17 -0600 | [diff] [blame] | 134 | int sc_max_sge; |
Steve Wise | bc3fe2e | 2015-07-27 18:10:12 -0500 | [diff] [blame] | 135 | int sc_max_sge_rd; /* max sge for read target */ |
Tom Tucker | d21b05f | 2007-12-12 16:13:17 -0600 | [diff] [blame] | 136 | |
| 137 | int sc_sq_depth; /* Depth of SQ */ |
| 138 | atomic_t sc_sq_count; /* Number of SQ WR on queue */ |
| 139 | |
| 140 | int sc_max_requests; /* Depth of RQ */ |
| 141 | int sc_max_req_size; /* Size of each RQ WR buf */ |
| 142 | |
| 143 | struct ib_pd *sc_pd; |
| 144 | |
Tom Tucker | 87295b6 | 2008-05-28 13:17:44 -0500 | [diff] [blame] | 145 | atomic_t sc_dma_used; |
Tom Tucker | 8740767 | 2008-04-30 20:44:39 -0500 | [diff] [blame] | 146 | atomic_t sc_ctxt_used; |
Tom Tucker | d21b05f | 2007-12-12 16:13:17 -0600 | [diff] [blame] | 147 | struct list_head sc_rq_dto_q; |
| 148 | spinlock_t sc_rq_dto_lock; |
| 149 | struct ib_qp *sc_qp; |
| 150 | struct ib_cq *sc_rq_cq; |
| 151 | struct ib_cq *sc_sq_cq; |
| 152 | struct ib_mr *sc_phys_mr; /* MR for server memory */ |
Chuck Lever | e545241 | 2015-01-13 11:03:20 -0500 | [diff] [blame] | 153 | int (*sc_reader)(struct svcxprt_rdma *, |
| 154 | struct svc_rqst *, |
| 155 | struct svc_rdma_op_ctxt *, |
| 156 | int *, u32 *, u32, u32, u64, bool); |
Tom Tucker | 0d3ebb9 | 2008-09-30 13:06:13 -0500 | [diff] [blame] | 157 | u32 sc_dev_caps; /* distilled device caps */ |
| 158 | u32 sc_dma_lkey; /* local dma key */ |
| 159 | unsigned int sc_frmr_pg_list_len; |
| 160 | struct list_head sc_frmr_q; |
| 161 | spinlock_t sc_frmr_q_lock; |
Tom Tucker | d21b05f | 2007-12-12 16:13:17 -0600 | [diff] [blame] | 162 | |
| 163 | spinlock_t sc_lock; /* transport lock */ |
| 164 | |
| 165 | wait_queue_head_t sc_send_wait; /* SQ exhaustion waitlist */ |
| 166 | unsigned long sc_flags; |
| 167 | struct list_head sc_dto_q; /* DTO tasklet I/O pending Q */ |
| 168 | struct list_head sc_read_complete_q; |
Tom Tucker | 8da91ea | 2008-04-30 22:00:46 -0500 | [diff] [blame] | 169 | struct work_struct sc_work; |
Tom Tucker | d21b05f | 2007-12-12 16:13:17 -0600 | [diff] [blame] | 170 | }; |
| 171 | /* sc_flags */ |
| 172 | #define RDMAXPRT_RQ_PENDING 1 |
| 173 | #define RDMAXPRT_SQ_PENDING 2 |
| 174 | #define RDMAXPRT_CONN_PENDING 3 |
| 175 | |
| 176 | #define RPCRDMA_LISTEN_BACKLOG 10 |
| 177 | /* The default ORD value is based on two outstanding full-size writes with a |
| 178 | * page size of 4k, or 32k * 2 ops / 4k = 16 outstanding RDMA_READ. */ |
| 179 | #define RPCRDMA_ORD (64/4) |
| 180 | #define RPCRDMA_SQ_DEPTH_MULT 8 |
Chuck Lever | d9bb5a4 | 2014-07-22 17:48:04 -0400 | [diff] [blame] | 181 | #define RPCRDMA_MAX_REQUESTS 32 |
Tom Tucker | d21b05f | 2007-12-12 16:13:17 -0600 | [diff] [blame] | 182 | #define RPCRDMA_MAX_REQ_SIZE 4096 |
| 183 | |
Chuck Lever | cc9a903 | 2015-08-07 16:55:46 -0400 | [diff] [blame] | 184 | #define RPCSVC_MAXPAYLOAD_RDMA RPCSVC_MAXPAYLOAD |
| 185 | |
Tom Tucker | d21b05f | 2007-12-12 16:13:17 -0600 | [diff] [blame] | 186 | /* svc_rdma_marshal.c */ |
Tom Tucker | d21b05f | 2007-12-12 16:13:17 -0600 | [diff] [blame] | 187 | extern int svc_rdma_xdr_decode_req(struct rpcrdma_msg **, struct svc_rqst *); |
Tom Tucker | d21b05f | 2007-12-12 16:13:17 -0600 | [diff] [blame] | 188 | extern int svc_rdma_xdr_encode_error(struct svcxprt_rdma *, |
| 189 | struct rpcrdma_msg *, |
Chuck Lever | 30b7e24 | 2015-06-04 11:21:10 -0400 | [diff] [blame] | 190 | enum rpcrdma_errcode, __be32 *); |
Tom Tucker | d21b05f | 2007-12-12 16:13:17 -0600 | [diff] [blame] | 191 | extern void svc_rdma_xdr_encode_write_list(struct rpcrdma_msg *, int); |
| 192 | extern void svc_rdma_xdr_encode_reply_array(struct rpcrdma_write_array *, int); |
| 193 | extern void svc_rdma_xdr_encode_array_chunk(struct rpcrdma_write_array *, int, |
Dan Carpenter | 1fa9c44 | 2012-02-21 10:28:04 +0300 | [diff] [blame] | 194 | __be32, __be64, u32); |
Tom Tucker | d21b05f | 2007-12-12 16:13:17 -0600 | [diff] [blame] | 195 | extern void svc_rdma_xdr_encode_reply_header(struct svcxprt_rdma *, |
| 196 | struct rpcrdma_msg *, |
| 197 | struct rpcrdma_msg *, |
| 198 | enum rpcrdma_proc); |
| 199 | extern int svc_rdma_xdr_get_reply_hdr_len(struct rpcrdma_msg *); |
| 200 | |
| 201 | /* svc_rdma_recvfrom.c */ |
| 202 | extern int svc_rdma_recvfrom(struct svc_rqst *); |
Chuck Lever | e545241 | 2015-01-13 11:03:20 -0500 | [diff] [blame] | 203 | extern int rdma_read_chunk_lcl(struct svcxprt_rdma *, struct svc_rqst *, |
| 204 | struct svc_rdma_op_ctxt *, int *, u32 *, |
| 205 | u32, u32, u64, bool); |
| 206 | extern int rdma_read_chunk_frmr(struct svcxprt_rdma *, struct svc_rqst *, |
| 207 | struct svc_rdma_op_ctxt *, int *, u32 *, |
| 208 | u32, u32, u64, bool); |
Tom Tucker | d21b05f | 2007-12-12 16:13:17 -0600 | [diff] [blame] | 209 | |
| 210 | /* svc_rdma_sendto.c */ |
| 211 | extern int svc_rdma_sendto(struct svc_rqst *); |
Chuck Lever | 10dc451 | 2015-07-09 16:45:28 -0400 | [diff] [blame] | 212 | extern struct rpcrdma_read_chunk * |
| 213 | svc_rdma_get_read_chunk(struct rpcrdma_msg *); |
Tom Tucker | d21b05f | 2007-12-12 16:13:17 -0600 | [diff] [blame] | 214 | |
| 215 | /* svc_rdma_transport.c */ |
| 216 | extern int svc_rdma_send(struct svcxprt_rdma *, struct ib_send_wr *); |
Tom Tucker | 008fdbc | 2008-05-07 15:47:42 -0500 | [diff] [blame] | 217 | extern void svc_rdma_send_error(struct svcxprt_rdma *, struct rpcrdma_msg *, |
| 218 | enum rpcrdma_errcode); |
Tom Tucker | d21b05f | 2007-12-12 16:13:17 -0600 | [diff] [blame] | 219 | extern int svc_rdma_post_recv(struct svcxprt_rdma *); |
| 220 | extern int svc_rdma_create_listen(struct svc_serv *, int, struct sockaddr *); |
| 221 | extern struct svc_rdma_op_ctxt *svc_rdma_get_context(struct svcxprt_rdma *); |
| 222 | extern void svc_rdma_put_context(struct svc_rdma_op_ctxt *, int); |
Tom Tucker | 146b6df | 2008-08-12 15:12:10 -0500 | [diff] [blame] | 223 | extern void svc_rdma_unmap_dma(struct svc_rdma_op_ctxt *ctxt); |
Tom Tucker | ab96ddd | 2008-05-28 13:54:04 -0500 | [diff] [blame] | 224 | extern struct svc_rdma_req_map *svc_rdma_get_req_map(void); |
| 225 | extern void svc_rdma_put_req_map(struct svc_rdma_req_map *); |
Tom Tucker | 64be8608 | 2008-10-06 14:45:18 -0500 | [diff] [blame] | 226 | extern struct svc_rdma_fastreg_mr *svc_rdma_get_frmr(struct svcxprt_rdma *); |
| 227 | extern void svc_rdma_put_frmr(struct svcxprt_rdma *, |
| 228 | struct svc_rdma_fastreg_mr *); |
Tom Tucker | d21b05f | 2007-12-12 16:13:17 -0600 | [diff] [blame] | 229 | extern void svc_sq_reap(struct svcxprt_rdma *); |
| 230 | extern void svc_rq_reap(struct svcxprt_rdma *); |
Tom Tucker | d21b05f | 2007-12-12 16:13:17 -0600 | [diff] [blame] | 231 | extern void svc_rdma_prep_reply_hdr(struct svc_rqst *); |
| 232 | |
Chuck Lever | 9468431 | 2015-10-24 17:28:16 -0400 | [diff] [blame^] | 233 | extern struct svc_xprt_class svc_rdma_class; |
| 234 | #ifdef CONFIG_SUNRPC_BACKCHANNEL |
| 235 | extern struct svc_xprt_class svc_rdma_bc_class; |
| 236 | #endif |
| 237 | |
Tom Tucker | d21b05f | 2007-12-12 16:13:17 -0600 | [diff] [blame] | 238 | /* svc_rdma.c */ |
| 239 | extern int svc_rdma_init(void); |
| 240 | extern void svc_rdma_cleanup(void); |
| 241 | |
Tom Tucker | d21b05f | 2007-12-12 16:13:17 -0600 | [diff] [blame] | 242 | #endif |