1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * Copyright (c) 2015 Oracle. All rights reserved. 4 * 5 * Support for backward direction RPCs on RPC/RDMA (server-side). 6 */ 7 8 #include <linux/module.h> 9 #include <linux/sunrpc/svc_rdma.h> 10 #include "xprt_rdma.h" 11 12 #define RPCDBG_FACILITY RPCDBG_SVCXPRT 13 14 #undef SVCRDMA_BACKCHANNEL_DEBUG 15 16 /** 17 * svc_rdma_handle_bc_reply - Process incoming backchannel reply 18 * @xprt: controlling backchannel transport 19 * @rdma_resp: pointer to incoming transport header 20 * @rcvbuf: XDR buffer into which to decode the reply 21 * 22 * Returns: 23 * %0 if @rcvbuf is filled in, xprt_complete_rqst called, 24 * %-EAGAIN if server should call ->recvfrom again. 25 */ 26 int svc_rdma_handle_bc_reply(struct rpc_xprt *xprt, __be32 *rdma_resp, 27 struct xdr_buf *rcvbuf) 28 { 29 struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt); 30 struct kvec *dst, *src = &rcvbuf->head[0]; 31 struct rpc_rqst *req; 32 unsigned long cwnd; 33 u32 credits; 34 size_t len; 35 __be32 xid; 36 __be32 *p; 37 int ret; 38 39 p = (__be32 *)src->iov_base; 40 len = src->iov_len; 41 xid = *rdma_resp; 42 43 #ifdef SVCRDMA_BACKCHANNEL_DEBUG 44 pr_info("%s: xid=%08x, length=%zu\n", 45 __func__, be32_to_cpu(xid), len); 46 pr_info("%s: RPC/RDMA: %*ph\n", 47 __func__, (int)RPCRDMA_HDRLEN_MIN, rdma_resp); 48 pr_info("%s: RPC: %*ph\n", 49 __func__, (int)len, p); 50 #endif 51 52 ret = -EAGAIN; 53 if (src->iov_len < 24) 54 goto out_shortreply; 55 56 spin_lock(&xprt->recv_lock); 57 req = xprt_lookup_rqst(xprt, xid); 58 if (!req) 59 goto out_notfound; 60 61 dst = &req->rq_private_buf.head[0]; 62 memcpy(&req->rq_private_buf, &req->rq_rcv_buf, sizeof(struct xdr_buf)); 63 if (dst->iov_len < len) 64 goto out_unlock; 65 memcpy(dst->iov_base, p, len); 66 67 credits = be32_to_cpup(rdma_resp + 2); 68 if (credits == 0) 69 credits = 1; /* don't deadlock */ 70 else if (credits > r_xprt->rx_buf.rb_bc_max_requests) 71 credits = r_xprt->rx_buf.rb_bc_max_requests; 72 73 spin_lock_bh(&xprt->transport_lock); 74 cwnd = xprt->cwnd; 75 xprt->cwnd = credits << RPC_CWNDSHIFT; 76 if (xprt->cwnd > cwnd) 77 xprt_release_rqst_cong(req->rq_task); 78 spin_unlock_bh(&xprt->transport_lock); 79 80 81 ret = 0; 82 xprt_complete_rqst(req->rq_task, rcvbuf->len); 83 rcvbuf->len = 0; 84 85 out_unlock: 86 spin_unlock(&xprt->recv_lock); 87 out: 88 return ret; 89 90 out_shortreply: 91 dprintk("svcrdma: short bc reply: xprt=%p, len=%zu\n", 92 xprt, src->iov_len); 93 goto out; 94 95 out_notfound: 96 dprintk("svcrdma: unrecognized bc reply: xprt=%p, xid=%08x\n", 97 xprt, be32_to_cpu(xid)); 98 99 goto out_unlock; 100 } 101 102 /* Send a backwards direction RPC call. 103 * 104 * Caller holds the connection's mutex and has already marshaled 105 * the RPC/RDMA request. 106 * 107 * This is similar to svc_rdma_send_reply_msg, but takes a struct 108 * rpc_rqst instead, does not support chunks, and avoids blocking 109 * memory allocation. 110 * 111 * XXX: There is still an opportunity to block in svc_rdma_send() 112 * if there are no SQ entries to post the Send. This may occur if 113 * the adapter has a small maximum SQ depth. 114 */ 115 static int svc_rdma_bc_sendto(struct svcxprt_rdma *rdma, 116 struct rpc_rqst *rqst) 117 { 118 struct svc_rdma_op_ctxt *ctxt; 119 int ret; 120 121 ctxt = svc_rdma_get_context(rdma); 122 123 /* rpcrdma_bc_send_request builds the transport header and 124 * the backchannel RPC message in the same buffer. Thus only 125 * one SGE is needed to send both. 126 */ 127 ret = svc_rdma_map_reply_hdr(rdma, ctxt, rqst->rq_buffer, 128 rqst->rq_snd_buf.len); 129 if (ret < 0) 130 goto out_err; 131 132 ret = svc_rdma_repost_recv(rdma, GFP_NOIO); 133 if (ret) 134 goto out_err; 135 136 ret = svc_rdma_post_send_wr(rdma, ctxt, 1, 0); 137 if (ret) 138 goto out_unmap; 139 140 out_err: 141 dprintk("svcrdma: %s returns %d\n", __func__, ret); 142 return ret; 143 144 out_unmap: 145 svc_rdma_unmap_dma(ctxt); 146 svc_rdma_put_context(ctxt, 1); 147 ret = -EIO; 148 goto out_err; 149 } 150 151 /* Server-side transport endpoint wants a whole page for its send 152 * buffer. The client RPC code constructs the RPC header in this 153 * buffer before it invokes ->send_request. 154 */ 155 static int 156 xprt_rdma_bc_allocate(struct rpc_task *task) 157 { 158 struct rpc_rqst *rqst = task->tk_rqstp; 159 size_t size = rqst->rq_callsize; 160 struct page *page; 161 162 if (size > PAGE_SIZE) { 163 WARN_ONCE(1, "svcrdma: large bc buffer request (size %zu)\n", 164 size); 165 return -EINVAL; 166 } 167 168 /* svc_rdma_sendto releases this page */ 169 page = alloc_page(RPCRDMA_DEF_GFP); 170 if (!page) 171 return -ENOMEM; 172 rqst->rq_buffer = page_address(page); 173 174 rqst->rq_rbuffer = kmalloc(rqst->rq_rcvsize, RPCRDMA_DEF_GFP); 175 if (!rqst->rq_rbuffer) { 176 put_page(page); 177 return -ENOMEM; 178 } 179 return 0; 180 } 181 182 static void 183 xprt_rdma_bc_free(struct rpc_task *task) 184 { 185 struct rpc_rqst *rqst = task->tk_rqstp; 186 187 kfree(rqst->rq_rbuffer); 188 } 189 190 static int 191 rpcrdma_bc_send_request(struct svcxprt_rdma *rdma, struct rpc_rqst *rqst) 192 { 193 struct rpc_xprt *xprt = rqst->rq_xprt; 194 struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt); 195 __be32 *p; 196 int rc; 197 198 /* Space in the send buffer for an RPC/RDMA header is reserved 199 * via xprt->tsh_size. 200 */ 201 p = rqst->rq_buffer; 202 *p++ = rqst->rq_xid; 203 *p++ = rpcrdma_version; 204 *p++ = cpu_to_be32(r_xprt->rx_buf.rb_bc_max_requests); 205 *p++ = rdma_msg; 206 *p++ = xdr_zero; 207 *p++ = xdr_zero; 208 *p = xdr_zero; 209 210 #ifdef SVCRDMA_BACKCHANNEL_DEBUG 211 pr_info("%s: %*ph\n", __func__, 64, rqst->rq_buffer); 212 #endif 213 214 rc = svc_rdma_bc_sendto(rdma, rqst); 215 if (rc) 216 goto drop_connection; 217 return rc; 218 219 drop_connection: 220 dprintk("svcrdma: failed to send bc call\n"); 221 xprt_disconnect_done(xprt); 222 return -ENOTCONN; 223 } 224 225 /* Send an RPC call on the passive end of a transport 226 * connection. 227 */ 228 static int 229 xprt_rdma_bc_send_request(struct rpc_task *task) 230 { 231 struct rpc_rqst *rqst = task->tk_rqstp; 232 struct svc_xprt *sxprt = rqst->rq_xprt->bc_xprt; 233 struct svcxprt_rdma *rdma; 234 int ret; 235 236 dprintk("svcrdma: sending bc call with xid: %08x\n", 237 be32_to_cpu(rqst->rq_xid)); 238 239 if (!mutex_trylock(&sxprt->xpt_mutex)) { 240 rpc_sleep_on(&sxprt->xpt_bc_pending, task, NULL); 241 if (!mutex_trylock(&sxprt->xpt_mutex)) 242 return -EAGAIN; 243 rpc_wake_up_queued_task(&sxprt->xpt_bc_pending, task); 244 } 245 246 ret = -ENOTCONN; 247 rdma = container_of(sxprt, struct svcxprt_rdma, sc_xprt); 248 if (!test_bit(XPT_DEAD, &sxprt->xpt_flags)) 249 ret = rpcrdma_bc_send_request(rdma, rqst); 250 251 mutex_unlock(&sxprt->xpt_mutex); 252 253 if (ret < 0) 254 return ret; 255 return 0; 256 } 257 258 static void 259 xprt_rdma_bc_close(struct rpc_xprt *xprt) 260 { 261 dprintk("svcrdma: %s: xprt %p\n", __func__, xprt); 262 } 263 264 static void 265 xprt_rdma_bc_put(struct rpc_xprt *xprt) 266 { 267 dprintk("svcrdma: %s: xprt %p\n", __func__, xprt); 268 269 xprt_free(xprt); 270 module_put(THIS_MODULE); 271 } 272 273 static const struct rpc_xprt_ops xprt_rdma_bc_procs = { 274 .reserve_xprt = xprt_reserve_xprt_cong, 275 .release_xprt = xprt_release_xprt_cong, 276 .alloc_slot = xprt_alloc_slot, 277 .release_request = xprt_release_rqst_cong, 278 .buf_alloc = xprt_rdma_bc_allocate, 279 .buf_free = xprt_rdma_bc_free, 280 .send_request = xprt_rdma_bc_send_request, 281 .set_retrans_timeout = xprt_set_retrans_timeout_def, 282 .close = xprt_rdma_bc_close, 283 .destroy = xprt_rdma_bc_put, 284 .print_stats = xprt_rdma_print_stats 285 }; 286 287 static const struct rpc_timeout xprt_rdma_bc_timeout = { 288 .to_initval = 60 * HZ, 289 .to_maxval = 60 * HZ, 290 }; 291 292 /* It shouldn't matter if the number of backchannel session slots 293 * doesn't match the number of RPC/RDMA credits. That just means 294 * one or the other will have extra slots that aren't used. 295 */ 296 static struct rpc_xprt * 297 xprt_setup_rdma_bc(struct xprt_create *args) 298 { 299 struct rpc_xprt *xprt; 300 struct rpcrdma_xprt *new_xprt; 301 302 if (args->addrlen > sizeof(xprt->addr)) { 303 dprintk("RPC: %s: address too large\n", __func__); 304 return ERR_PTR(-EBADF); 305 } 306 307 xprt = xprt_alloc(args->net, sizeof(*new_xprt), 308 RPCRDMA_MAX_BC_REQUESTS, 309 RPCRDMA_MAX_BC_REQUESTS); 310 if (!xprt) { 311 dprintk("RPC: %s: couldn't allocate rpc_xprt\n", 312 __func__); 313 return ERR_PTR(-ENOMEM); 314 } 315 316 xprt->timeout = &xprt_rdma_bc_timeout; 317 xprt_set_bound(xprt); 318 xprt_set_connected(xprt); 319 xprt->bind_timeout = RPCRDMA_BIND_TO; 320 xprt->reestablish_timeout = RPCRDMA_INIT_REEST_TO; 321 xprt->idle_timeout = RPCRDMA_IDLE_DISC_TO; 322 323 xprt->prot = XPRT_TRANSPORT_BC_RDMA; 324 xprt->tsh_size = RPCRDMA_HDRLEN_MIN / sizeof(__be32); 325 xprt->ops = &xprt_rdma_bc_procs; 326 327 memcpy(&xprt->addr, args->dstaddr, args->addrlen); 328 xprt->addrlen = args->addrlen; 329 xprt_rdma_format_addresses(xprt, (struct sockaddr *)&xprt->addr); 330 xprt->resvport = 0; 331 332 xprt->max_payload = xprt_rdma_max_inline_read; 333 334 new_xprt = rpcx_to_rdmax(xprt); 335 new_xprt->rx_buf.rb_bc_max_requests = xprt->max_reqs; 336 337 xprt_get(xprt); 338 args->bc_xprt->xpt_bc_xprt = xprt; 339 xprt->bc_xprt = args->bc_xprt; 340 341 if (!try_module_get(THIS_MODULE)) 342 goto out_fail; 343 344 /* Final put for backchannel xprt is in __svc_rdma_free */ 345 xprt_get(xprt); 346 return xprt; 347 348 out_fail: 349 xprt_rdma_free_addresses(xprt); 350 args->bc_xprt->xpt_bc_xprt = NULL; 351 args->bc_xprt->xpt_bc_xps = NULL; 352 xprt_put(xprt); 353 xprt_free(xprt); 354 return ERR_PTR(-EINVAL); 355 } 356 357 struct xprt_class xprt_rdma_bc = { 358 .list = LIST_HEAD_INIT(xprt_rdma_bc.list), 359 .name = "rdma backchannel", 360 .owner = THIS_MODULE, 361 .ident = XPRT_TRANSPORT_BC_RDMA, 362 .setup = xprt_setup_rdma_bc, 363 }; 364