LXR linux/include/linux/sunrpc/svc

   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 */
  52extern unsigned int svcrdma_ord;
  53extern unsigned int svcrdma_max_requests;
  54extern unsigned int svcrdma_max_bc_requests;
  55extern unsigned int svcrdma_max_req_size;
  56
  57extern atomic_t rdma_stat_recv;
  58extern atomic_t rdma_stat_read;
  59extern atomic_t rdma_stat_write;
  60extern atomic_t rdma_stat_sq_starve;
  61extern atomic_t rdma_stat_rq_starve;
  62extern atomic_t rdma_stat_rq_poll;
  63extern atomic_t rdma_stat_rq_prod;
  64extern atomic_t rdma_stat_sq_poll;
  65extern atomic_t rdma_stat_sq_prod;
  66
  67/*
  68 * Contexts are built when an RDMA request is created and are a
  69 * record of the resources that can be recovered when the request
  70 * completes.
  71 */
  72struct svc_rdma_op_ctxt {
  73        struct list_head free;
  74        struct svc_rdma_op_ctxt *read_hdr;
  75        struct svc_rdma_fastreg_mr *frmr;
  76        int hdr_count;
  77        struct xdr_buf arg;
  78        struct ib_cqe cqe;
  79        struct ib_cqe reg_cqe;
  80        struct ib_cqe inv_cqe;
  81        struct list_head dto_q;
  82        u32 byte_len;
  83        u32 position;
  84        struct svcxprt_rdma *xprt;
  85        unsigned long flags;
  86        enum dma_data_direction direction;
  87        int count;
  88        unsigned int mapped_sges;
  89        struct ib_sge sge[RPCSVC_MAXPAGES];
  90        struct page *pages[RPCSVC_MAXPAGES];
  91};
  92
  93/*
  94 * NFS_ requests are mapped on the client side by the chunk lists in
  95 * the RPCRDMA header. During the fetching of the RPC from the client
  96 * and the writing of the reply to the client, the memory in the
  97 * client and the memory in the server must be mapped as contiguous
  98 * vaddr/len for access by the hardware. These data strucures keep
  99 * these mappings.
 100 *
 101 * For an RDMA_WRITE, the 'sge' maps the RPC REPLY. For RDMA_READ, the
 102 * 'sge' in the svc_rdma_req_map maps the server side RPC reply and the
 103 * 'ch' field maps the read-list of the RPCRDMA header to the 'sge'
 104 * mapping of the reply.
 105 */
 106struct svc_rdma_chunk_sge {
 107        int start;              /* sge no for this chunk */
 108        int count;              /* sge count for this chunk */
 109};
 110struct svc_rdma_fastreg_mr {
 111        struct ib_mr *mr;
 112        struct scatterlist *sg;
 113        int sg_nents;
 114        unsigned long access_flags;
 115        enum dma_data_direction direction;
 116        struct list_head frmr_list;
 117};
 118struct svc_rdma_req_map {
 119        struct list_head free;
 120        unsigned long count;
 121        union {
 122                struct kvec sge[RPCSVC_MAXPAGES];
 123                struct svc_rdma_chunk_sge ch[RPCSVC_MAXPAGES];
 124                unsigned long lkey[RPCSVC_MAXPAGES];
 125        };
 126};
 127#define RDMACTXT_F_LAST_CTXT    2
 128
 129#define SVCRDMA_DEVCAP_FAST_REG         1       /* fast mr registration */
 130#define SVCRDMA_DEVCAP_READ_W_INV       2       /* read w/ invalidate */
 131
 132struct svcxprt_rdma {
 133        struct svc_xprt      sc_xprt;           /* SVC transport structure */
 134        struct rdma_cm_id    *sc_cm_id;         /* RDMA connection id */
 135        struct list_head     sc_accept_q;       /* Conn. waiting accept */
 136        int                  sc_ord;            /* RDMA read limit */
 137        int                  sc_max_sge;
 138        int                  sc_max_sge_rd;     /* max sge for read target */
 139        bool                 sc_snd_w_inv;      /* OK to use Send With Invalidate */
 140
 141        atomic_t             sc_sq_avail;       /* SQEs ready to be consumed */
 142        unsigned int         sc_sq_depth;       /* Depth of SQ */
 143        unsigned int         sc_rq_depth;       /* Depth of RQ */
 144        u32                  sc_max_requests;   /* Forward credits */
 145        u32                  sc_max_bc_requests;/* Backward credits */
 146        int                  sc_max_req_size;   /* Size of each RQ WR buf */
 147
 148        struct ib_pd         *sc_pd;
 149
 150        spinlock_t           sc_ctxt_lock;
 151        struct list_head     sc_ctxts;
 152        int                  sc_ctxt_used;
 153        spinlock_t           sc_map_lock;
 154        struct list_head     sc_maps;
 155
 156        struct list_head     sc_rq_dto_q;
 157        spinlock_t           sc_rq_dto_lock;
 158        struct ib_qp         *sc_qp;
 159        struct ib_cq         *sc_rq_cq;
 160        struct ib_cq         *sc_sq_cq;
 161        int                  (*sc_reader)(struct svcxprt_rdma *,
 162                                          struct svc_rqst *,
 163                                          struct svc_rdma_op_ctxt *,
 164                                          int *, u32 *, u32, u32, u64, bool);
 165        u32                  sc_dev_caps;       /* distilled device caps */
 166        unsigned int         sc_frmr_pg_list_len;
 167        struct list_head     sc_frmr_q;
 168        spinlock_t           sc_frmr_q_lock;
 169
 170        spinlock_t           sc_lock;           /* transport lock */
 171
 172        wait_queue_head_t    sc_send_wait;      /* SQ exhaustion waitlist */
 173        unsigned long        sc_flags;
 174        struct list_head     sc_dto_q;          /* DTO tasklet I/O pending Q */
 175        struct list_head     sc_read_complete_q;
 176        struct work_struct   sc_work;
 177};
 178/* sc_flags */
 179#define RDMAXPRT_CONN_PENDING   3
 180
 181#define RPCRDMA_LISTEN_BACKLOG  10
 182/* The default ORD value is based on two outstanding full-size writes with a
 183 * page size of 4k, or 32k * 2 ops / 4k = 16 outstanding RDMA_READ.  */
 184#define RPCRDMA_ORD             (64/4)
 185#define RPCRDMA_SQ_DEPTH_MULT   8
 186#define RPCRDMA_MAX_REQUESTS    32
 187#define RPCRDMA_MAX_REQ_SIZE    4096
 188
 189/* Typical ULP usage of BC requests is NFSv4.1 backchannel. Our
 190 * current NFSv4.1 implementation supports one backchannel slot.
 191 */
 192#define RPCRDMA_MAX_BC_REQUESTS 2
 193
 194#define RPCSVC_MAXPAYLOAD_RDMA  RPCSVC_MAXPAYLOAD
 195
 196/* Track DMA maps for this transport and context */
 197static inline void svc_rdma_count_mappings(struct svcxprt_rdma *rdma,
 198                                           struct svc_rdma_op_ctxt *ctxt)
 199{
 200        ctxt->mapped_sges++;
 201}
 202
 203/* svc_rdma_backchannel.c */
 204extern int svc_rdma_handle_bc_reply(struct rpc_xprt *xprt,
 205                                    struct rpcrdma_msg *rmsgp,
 206                                    struct xdr_buf *rcvbuf);
 207
 208/* svc_rdma_marshal.c */
 209extern int svc_rdma_xdr_decode_req(struct xdr_buf *);
 210extern int svc_rdma_xdr_encode_error(struct svcxprt_rdma *,
 211                                     struct rpcrdma_msg *,
 212                                     enum rpcrdma_errcode, __be32 *);
 213extern void svc_rdma_xdr_encode_write_list(struct rpcrdma_msg *, int);
 214extern void svc_rdma_xdr_encode_reply_array(struct rpcrdma_write_array *, int);
 215extern void svc_rdma_xdr_encode_array_chunk(struct rpcrdma_write_array *, int,
 216                                            __be32, __be64, u32);
 217extern void svc_rdma_xdr_encode_reply_header(struct svcxprt_rdma *,
 218                                             struct rpcrdma_msg *,
 219                                             struct rpcrdma_msg *,
 220                                             enum rpcrdma_proc);
 221extern int svc_rdma_xdr_get_reply_hdr_len(struct rpcrdma_msg *);
 222
 223/* svc_rdma_recvfrom.c */
 224extern int svc_rdma_recvfrom(struct svc_rqst *);
 225extern int rdma_read_chunk_lcl(struct svcxprt_rdma *, struct svc_rqst *,
 226                               struct svc_rdma_op_ctxt *, int *, u32 *,
 227                               u32, u32, u64, bool);
 228extern int rdma_read_chunk_frmr(struct svcxprt_rdma *, struct svc_rqst *,
 229                                struct svc_rdma_op_ctxt *, int *, u32 *,
 230                                u32, u32, u64, bool);
 231
 232/* svc_rdma_sendto.c */
 233extern int svc_rdma_map_xdr(struct svcxprt_rdma *, struct xdr_buf *,
 234                            struct svc_rdma_req_map *, bool);
 235extern int svc_rdma_sendto(struct svc_rqst *);
 236extern void svc_rdma_send_error(struct svcxprt_rdma *, struct rpcrdma_msg *,
 237                                int);
 238
 239/* svc_rdma_transport.c */
 240extern void svc_rdma_wc_send(struct ib_cq *, struct ib_wc *);
 241extern void svc_rdma_wc_write(struct ib_cq *, struct ib_wc *);
 242extern void svc_rdma_wc_reg(struct ib_cq *, struct ib_wc *);
 243extern void svc_rdma_wc_read(struct ib_cq *, struct ib_wc *);
 244extern void svc_rdma_wc_inv(struct ib_cq *, struct ib_wc *);
 245extern int svc_rdma_send(struct svcxprt_rdma *, struct ib_send_wr *);
 246extern int svc_rdma_post_recv(struct svcxprt_rdma *, gfp_t);
 247extern int svc_rdma_repost_recv(struct svcxprt_rdma *, gfp_t);
 248extern int svc_rdma_create_listen(struct svc_serv *, int, struct sockaddr *);
 249extern struct svc_rdma_op_ctxt *svc_rdma_get_context(struct svcxprt_rdma *);
 250extern void svc_rdma_put_context(struct svc_rdma_op_ctxt *, int);
 251extern void svc_rdma_unmap_dma(struct svc_rdma_op_ctxt *ctxt);
 252extern struct svc_rdma_req_map *svc_rdma_get_req_map(struct svcxprt_rdma *);
 253extern void svc_rdma_put_req_map(struct svcxprt_rdma *,
 254                                 struct svc_rdma_req_map *);
 255extern struct svc_rdma_fastreg_mr *svc_rdma_get_frmr(struct svcxprt_rdma *);
 256extern void svc_rdma_put_frmr(struct svcxprt_rdma *,
 257                              struct svc_rdma_fastreg_mr *);
 258extern void svc_sq_reap(struct svcxprt_rdma *);
 259extern void svc_rq_reap(struct svcxprt_rdma *);
 260extern void svc_rdma_prep_reply_hdr(struct svc_rqst *);
 261
 262extern struct svc_xprt_class svc_rdma_class;
 263#ifdef CONFIG_SUNRPC_BACKCHANNEL
 264extern struct svc_xprt_class svc_rdma_bc_class;
 265#endif
 266
 267/* svc_rdma.c */
 268extern struct workqueue_struct *svc_rdma_wq;
 269extern int svc_rdma_init(void);
 270extern void svc_rdma_cleanup(void);
 271
 272#endif
 273