1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * Copyright (c) 2015-2018 Oracle. All rights reserved. 4 * 5 * Support for backward direction RPCs on RPC/RDMA (server-side). 6 */ 7 8 #include <linux/sunrpc/svc_rdma.h> 9 10 #include "xprt_rdma.h" 11 #include <trace/events/rpcrdma.h> 12 13 /** 14 * svc_rdma_handle_bc_reply - Process incoming backchannel Reply 15 * @rqstp: resources for handling the Reply 16 * @rctxt: Received message 17 * 18 */ 19 void svc_rdma_handle_bc_reply(struct svc_rqst *rqstp, 20 struct svc_rdma_recv_ctxt *rctxt) 21 { 22 struct svc_xprt *sxprt = rqstp->rq_xprt; 23 struct rpc_xprt *xprt = sxprt->xpt_bc_xprt; 24 struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt); 25 struct xdr_buf *rcvbuf = &rqstp->rq_arg; 26 struct kvec *dst, *src = &rcvbuf->head[0]; 27 __be32 *rdma_resp = rctxt->rc_recv_buf; 28 struct rpc_rqst *req; 29 u32 credits; 30 31 spin_lock(&xprt->queue_lock); 32 req = xprt_lookup_rqst(xprt, *rdma_resp); 33 if (!req) 34 goto out_unlock; 35 36 dst = &req->rq_private_buf.head[0]; 37 memcpy(&req->rq_private_buf, &req->rq_rcv_buf, sizeof(struct xdr_buf)); 38 if (dst->iov_len < src->iov_len) 39 goto out_unlock; 40 memcpy(dst->iov_base, src->iov_base, src->iov_len); 41 xprt_pin_rqst(req); 42 spin_unlock(&xprt->queue_lock); 43 44 credits = be32_to_cpup(rdma_resp + 2); 45 if (credits == 0) 46 credits = 1; /* don't deadlock */ 47 else if (credits > r_xprt->rx_buf.rb_bc_max_requests) 48 credits = r_xprt->rx_buf.rb_bc_max_requests; 49 spin_lock(&xprt->transport_lock); 50 xprt->cwnd = credits << RPC_CWNDSHIFT; 51 spin_unlock(&xprt->transport_lock); 52 53 spin_lock(&xprt->queue_lock); 54 xprt_complete_rqst(req->rq_task, rcvbuf->len); 55 xprt_unpin_rqst(req); 56 rcvbuf->len = 0; 57 58 out_unlock: 59 spin_unlock(&xprt->queue_lock); 60 } 61 62 /* Send a backwards direction RPC call. 63 * 64 * Caller holds the connection's mutex and has already marshaled 65 * the RPC/RDMA request. 66 * 67 * This is similar to svc_rdma_send_reply_msg, but takes a struct 68 * rpc_rqst instead, does not support chunks, and avoids blocking 69 * memory allocation. 70 * 71 * XXX: There is still an opportunity to block in svc_rdma_send() 72 * if there are no SQ entries to post the Send. This may occur if 73 * the adapter has a small maximum SQ depth. 74 */ 75 static int svc_rdma_bc_sendto(struct svcxprt_rdma *rdma, 76 struct rpc_rqst *rqst, 77 struct svc_rdma_send_ctxt *sctxt) 78 { 79 struct svc_rdma_recv_ctxt *rctxt; 80 int ret; 81 82 rctxt = svc_rdma_recv_ctxt_get(rdma); 83 if (!rctxt) 84 return -EIO; 85 86 ret = svc_rdma_map_reply_msg(rdma, sctxt, rctxt, &rqst->rq_snd_buf); 87 svc_rdma_recv_ctxt_put(rdma, rctxt); 88 if (ret < 0) 89 return -EIO; 90 91 /* Bump page refcnt so Send completion doesn't release 92 * the rq_buffer before all retransmits are complete. 93 */ 94 get_page(virt_to_page(rqst->rq_buffer)); 95 sctxt->sc_send_wr.opcode = IB_WR_SEND; 96 return svc_rdma_send(rdma, sctxt); 97 } 98 99 /* Server-side transport endpoint wants a whole page for its send 100 * buffer. The client RPC code constructs the RPC header in this 101 * buffer before it invokes ->send_request. 102 */ 103 static int 104 xprt_rdma_bc_allocate(struct rpc_task *task) 105 { 106 struct rpc_rqst *rqst = task->tk_rqstp; 107 size_t size = rqst->rq_callsize; 108 struct page *page; 109 110 if (size > PAGE_SIZE) { 111 WARN_ONCE(1, "svcrdma: large bc buffer request (size %zu)\n", 112 size); 113 return -EINVAL; 114 } 115 116 page = alloc_page(RPCRDMA_DEF_GFP); 117 if (!page) 118 return -ENOMEM; 119 rqst->rq_buffer = page_address(page); 120 121 rqst->rq_rbuffer = kmalloc(rqst->rq_rcvsize, RPCRDMA_DEF_GFP); 122 if (!rqst->rq_rbuffer) { 123 put_page(page); 124 return -ENOMEM; 125 } 126 return 0; 127 } 128 129 static void 130 xprt_rdma_bc_free(struct rpc_task *task) 131 { 132 struct rpc_rqst *rqst = task->tk_rqstp; 133 134 put_page(virt_to_page(rqst->rq_buffer)); 135 kfree(rqst->rq_rbuffer); 136 } 137 138 static int 139 rpcrdma_bc_send_request(struct svcxprt_rdma *rdma, struct rpc_rqst *rqst) 140 { 141 struct rpc_xprt *xprt = rqst->rq_xprt; 142 struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt); 143 struct svc_rdma_send_ctxt *ctxt; 144 __be32 *p; 145 int rc; 146 147 ctxt = svc_rdma_send_ctxt_get(rdma); 148 if (!ctxt) 149 goto drop_connection; 150 151 p = xdr_reserve_space(&ctxt->sc_stream, RPCRDMA_HDRLEN_MIN); 152 if (!p) 153 goto put_ctxt; 154 *p++ = rqst->rq_xid; 155 *p++ = rpcrdma_version; 156 *p++ = cpu_to_be32(r_xprt->rx_buf.rb_bc_max_requests); 157 *p++ = rdma_msg; 158 *p++ = xdr_zero; 159 *p++ = xdr_zero; 160 *p = xdr_zero; 161 162 rqst->rq_xtime = ktime_get(); 163 rc = svc_rdma_bc_sendto(rdma, rqst, ctxt); 164 if (rc) 165 goto put_ctxt; 166 return 0; 167 168 put_ctxt: 169 svc_rdma_send_ctxt_put(rdma, ctxt); 170 171 drop_connection: 172 return -ENOTCONN; 173 } 174 175 /** 176 * xprt_rdma_bc_send_request - Send a reverse-direction Call 177 * @rqst: rpc_rqst containing Call message to be sent 178 * 179 * Return values: 180 * %0 if the message was sent successfully 181 * %ENOTCONN if the message was not sent 182 */ 183 static int xprt_rdma_bc_send_request(struct rpc_rqst *rqst) 184 { 185 struct svc_xprt *sxprt = rqst->rq_xprt->bc_xprt; 186 struct svcxprt_rdma *rdma = 187 container_of(sxprt, struct svcxprt_rdma, sc_xprt); 188 int ret; 189 190 if (test_bit(XPT_DEAD, &sxprt->xpt_flags)) 191 return -ENOTCONN; 192 193 ret = rpcrdma_bc_send_request(rdma, rqst); 194 if (ret == -ENOTCONN) 195 svc_close_xprt(sxprt); 196 return ret; 197 } 198 199 static void 200 xprt_rdma_bc_close(struct rpc_xprt *xprt) 201 { 202 xprt_disconnect_done(xprt); 203 xprt->cwnd = RPC_CWNDSHIFT; 204 } 205 206 static void 207 xprt_rdma_bc_put(struct rpc_xprt *xprt) 208 { 209 xprt_rdma_free_addresses(xprt); 210 xprt_free(xprt); 211 } 212 213 static const struct rpc_xprt_ops xprt_rdma_bc_procs = { 214 .reserve_xprt = xprt_reserve_xprt_cong, 215 .release_xprt = xprt_release_xprt_cong, 216 .alloc_slot = xprt_alloc_slot, 217 .free_slot = xprt_free_slot, 218 .release_request = xprt_release_rqst_cong, 219 .buf_alloc = xprt_rdma_bc_allocate, 220 .buf_free = xprt_rdma_bc_free, 221 .send_request = xprt_rdma_bc_send_request, 222 .wait_for_reply_request = xprt_wait_for_reply_request_def, 223 .close = xprt_rdma_bc_close, 224 .destroy = xprt_rdma_bc_put, 225 .print_stats = xprt_rdma_print_stats 226 }; 227 228 static const struct rpc_timeout xprt_rdma_bc_timeout = { 229 .to_initval = 60 * HZ, 230 .to_maxval = 60 * HZ, 231 }; 232 233 /* It shouldn't matter if the number of backchannel session slots 234 * doesn't match the number of RPC/RDMA credits. That just means 235 * one or the other will have extra slots that aren't used. 236 */ 237 static struct rpc_xprt * 238 xprt_setup_rdma_bc(struct xprt_create *args) 239 { 240 struct rpc_xprt *xprt; 241 struct rpcrdma_xprt *new_xprt; 242 243 if (args->addrlen > sizeof(xprt->addr)) 244 return ERR_PTR(-EBADF); 245 246 xprt = xprt_alloc(args->net, sizeof(*new_xprt), 247 RPCRDMA_MAX_BC_REQUESTS, 248 RPCRDMA_MAX_BC_REQUESTS); 249 if (!xprt) 250 return ERR_PTR(-ENOMEM); 251 252 xprt->timeout = &xprt_rdma_bc_timeout; 253 xprt_set_bound(xprt); 254 xprt_set_connected(xprt); 255 xprt->bind_timeout = RPCRDMA_BIND_TO; 256 xprt->reestablish_timeout = RPCRDMA_INIT_REEST_TO; 257 xprt->idle_timeout = RPCRDMA_IDLE_DISC_TO; 258 259 xprt->prot = XPRT_TRANSPORT_BC_RDMA; 260 xprt->ops = &xprt_rdma_bc_procs; 261 262 memcpy(&xprt->addr, args->dstaddr, args->addrlen); 263 xprt->addrlen = args->addrlen; 264 xprt_rdma_format_addresses(xprt, (struct sockaddr *)&xprt->addr); 265 xprt->resvport = 0; 266 267 xprt->max_payload = xprt_rdma_max_inline_read; 268 269 new_xprt = rpcx_to_rdmax(xprt); 270 new_xprt->rx_buf.rb_bc_max_requests = xprt->max_reqs; 271 272 xprt_get(xprt); 273 args->bc_xprt->xpt_bc_xprt = xprt; 274 xprt->bc_xprt = args->bc_xprt; 275 276 /* Final put for backchannel xprt is in __svc_rdma_free */ 277 xprt_get(xprt); 278 return xprt; 279 } 280 281 struct xprt_class xprt_rdma_bc = { 282 .list = LIST_HEAD_INIT(xprt_rdma_bc.list), 283 .name = "rdma backchannel", 284 .owner = THIS_MODULE, 285 .ident = XPRT_TRANSPORT_BC_RDMA, 286 .setup = xprt_setup_rdma_bc, 287 }; 288