1/* 2 * Copyright (c) 2014 Open Grid Computing, Inc. All rights reserved. 3 * Copyright (c) 2005-2006 Network Appliance, Inc. All rights reserved. 4 * 5 * This software is available to you under a choice of one of two 6 * licenses. You may choose to be licensed under the terms of the GNU 7 * General Public License (GPL) Version 2, available from the file 8 * COPYING in the main directory of this source tree, or the BSD-type 9 * license below: 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 15 * Redistributions of source code must retain the above copyright 16 * notice, this list of conditions and the following disclaimer. 17 * 18 * Redistributions in binary form must reproduce the above 19 * copyright notice, this list of conditions and the following 20 * disclaimer in the documentation and/or other materials provided 21 * with the distribution. 22 * 23 * Neither the name of the Network Appliance, Inc. nor the names of 24 * its contributors may be used to endorse or promote products 25 * derived from this software without specific prior written 26 * permission. 27 * 28 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 29 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 30 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 31 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 32 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 33 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 34 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 35 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 36 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 37 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 38 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 39 * 40 * Author: Tom Tucker <tom@opengridcomputing.com> 41 */ 42 43#include <linux/sunrpc/debug.h> 44#include <linux/sunrpc/rpc_rdma.h> 45#include <linux/spinlock.h> 46#include <asm/unaligned.h> 47#include <rdma/ib_verbs.h> 48#include <rdma/rdma_cm.h> 49#include <linux/sunrpc/svc_rdma.h> 50 51#define RPCDBG_FACILITY RPCDBG_SVCXPRT 52 53static int map_xdr(struct svcxprt_rdma *xprt, 54 struct xdr_buf *xdr, 55 struct svc_rdma_req_map *vec) 56{ 57 int sge_no; 58 u32 sge_bytes; 59 u32 page_bytes; 60 u32 page_off; 61 int page_no; 62 63 if (xdr->len != 64 (xdr->head[0].iov_len + xdr->page_len + xdr->tail[0].iov_len)) { 65 pr_err("svcrdma: map_xdr: XDR buffer length error\n"); 66 return -EIO; 67 } 68 69 /* Skip the first sge, this is for the RPCRDMA header */ 70 sge_no = 1; 71 72 /* Head SGE */ 73 vec->sge[sge_no].iov_base = xdr->head[0].iov_base; 74 vec->sge[sge_no].iov_len = xdr->head[0].iov_len; 75 sge_no++; 76 77 /* pages SGE */ 78 page_no = 0; 79 page_bytes = xdr->page_len; 80 page_off = xdr->page_base; 81 while (page_bytes) { 82 vec->sge[sge_no].iov_base = 83 page_address(xdr->pages[page_no]) + page_off; 84 sge_bytes = min_t(u32, page_bytes, (PAGE_SIZE - page_off)); 85 page_bytes -= sge_bytes; 86 vec->sge[sge_no].iov_len = sge_bytes; 87 88 sge_no++; 89 page_no++; 90 page_off = 0; /* reset for next time through loop */ 91 } 92 93 /* Tail SGE */ 94 if (xdr->tail[0].iov_len) { 95 vec->sge[sge_no].iov_base = xdr->tail[0].iov_base; 96 vec->sge[sge_no].iov_len = xdr->tail[0].iov_len; 97 sge_no++; 98 } 99 100 dprintk("svcrdma: map_xdr: sge_no %d page_no %d " 101 "page_base %u page_len %u head_len %zu tail_len %zu\n", 102 sge_no, page_no, xdr->page_base, xdr->page_len, 103 xdr->head[0].iov_len, xdr->tail[0].iov_len); 104 105 vec->count = sge_no; 106 return 0; 107} 108 109static dma_addr_t dma_map_xdr(struct svcxprt_rdma *xprt, 110 struct xdr_buf *xdr, 111 u32 xdr_off, size_t len, int dir) 112{ 113 struct page *page; 114 dma_addr_t dma_addr; 115 if (xdr_off < xdr->head[0].iov_len) { 116 /* This offset is in the head */ 117 xdr_off += (unsigned long)xdr->head[0].iov_base & ~PAGE_MASK; 118 page = virt_to_page(xdr->head[0].iov_base); 119 } else { 120 xdr_off -= xdr->head[0].iov_len; 121 if (xdr_off < xdr->page_len) { 122 /* This offset is in the page list */ 123 xdr_off += xdr->page_base; 124 page = xdr->pages[xdr_off >> PAGE_SHIFT]; 125 xdr_off &= ~PAGE_MASK; 126 } else { 127 /* This offset is in the tail */ 128 xdr_off -= xdr->page_len; 129 xdr_off += (unsigned long) 130 xdr->tail[0].iov_base & ~PAGE_MASK; 131 page = virt_to_page(xdr->tail[0].iov_base); 132 } 133 } 134 dma_addr = ib_dma_map_page(xprt->sc_cm_id->device, page, xdr_off, 135 min_t(size_t, PAGE_SIZE, len), dir); 136 return dma_addr; 137} 138 139/* Assumptions: 140 * - The specified write_len can be represented in sc_max_sge * PAGE_SIZE 141 */ 142static int send_write(struct svcxprt_rdma *xprt, struct svc_rqst *rqstp, 143 u32 rmr, u64 to, 144 u32 xdr_off, int write_len, 145 struct svc_rdma_req_map *vec) 146{ 147 struct ib_send_wr write_wr; 148 struct ib_sge *sge; 149 int xdr_sge_no; 150 int sge_no; 151 int sge_bytes; 152 int sge_off; 153 int bc; 154 struct svc_rdma_op_ctxt *ctxt; 155 156 if (vec->count > RPCSVC_MAXPAGES) { 157 pr_err("svcrdma: Too many pages (%lu)\n", vec->count); 158 return -EIO; 159 } 160 161 dprintk("svcrdma: RDMA_WRITE rmr=%x, to=%llx, xdr_off=%d, " 162 "write_len=%d, vec->sge=%p, vec->count=%lu\n", 163 rmr, (unsigned long long)to, xdr_off, 164 write_len, vec->sge, vec->count); 165 166 ctxt = svc_rdma_get_context(xprt); 167 ctxt->direction = DMA_TO_DEVICE; 168 sge = ctxt->sge; 169 170 /* Find the SGE associated with xdr_off */ 171 for (bc = xdr_off, xdr_sge_no = 1; bc && xdr_sge_no < vec->count; 172 xdr_sge_no++) { 173 if (vec->sge[xdr_sge_no].iov_len > bc) 174 break; 175 bc -= vec->sge[xdr_sge_no].iov_len; 176 } 177 178 sge_off = bc; 179 bc = write_len; 180 sge_no = 0; 181 182 /* Copy the remaining SGE */ 183 while (bc != 0) { 184 sge_bytes = min_t(size_t, 185 bc, vec->sge[xdr_sge_no].iov_len-sge_off); 186 sge[sge_no].length = sge_bytes; 187 sge[sge_no].addr = 188 dma_map_xdr(xprt, &rqstp->rq_res, xdr_off, 189 sge_bytes, DMA_TO_DEVICE); 190 xdr_off += sge_bytes; 191 if (ib_dma_mapping_error(xprt->sc_cm_id->device, 192 sge[sge_no].addr)) 193 goto err; 194 atomic_inc(&xprt->sc_dma_used); 195 sge[sge_no].lkey = xprt->sc_dma_lkey; 196 ctxt->count++; 197 sge_off = 0; 198 sge_no++; 199 xdr_sge_no++; 200 if (xdr_sge_no > vec->count) { 201 pr_err("svcrdma: Too many sges (%d)\n", xdr_sge_no); 202 goto err; 203 } 204 bc -= sge_bytes; 205 if (sge_no == xprt->sc_max_sge) 206 break; 207 } 208 209 /* Prepare WRITE WR */ 210 memset(&write_wr, 0, sizeof write_wr); 211 ctxt->wr_op = IB_WR_RDMA_WRITE; 212 write_wr.wr_id = (unsigned long)ctxt; 213 write_wr.sg_list = &sge[0]; 214 write_wr.num_sge = sge_no; 215 write_wr.opcode = IB_WR_RDMA_WRITE; 216 write_wr.send_flags = IB_SEND_SIGNALED; 217 write_wr.wr.rdma.rkey = rmr; 218 write_wr.wr.rdma.remote_addr = to; 219 220 /* Post It */ 221 atomic_inc(&rdma_stat_write); 222 if (svc_rdma_send(xprt, &write_wr)) 223 goto err; 224 return write_len - bc; 225 err: 226 svc_rdma_unmap_dma(ctxt); 227 svc_rdma_put_context(ctxt, 0); 228 /* Fatal error, close transport */ 229 return -EIO; 230} 231 232static int send_write_chunks(struct svcxprt_rdma *xprt, 233 struct rpcrdma_msg *rdma_argp, 234 struct rpcrdma_msg *rdma_resp, 235 struct svc_rqst *rqstp, 236 struct svc_rdma_req_map *vec) 237{ 238 u32 xfer_len = rqstp->rq_res.page_len + rqstp->rq_res.tail[0].iov_len; 239 int write_len; 240 u32 xdr_off; 241 int chunk_off; 242 int chunk_no; 243 struct rpcrdma_write_array *arg_ary; 244 struct rpcrdma_write_array *res_ary; 245 int ret; 246 247 arg_ary = svc_rdma_get_write_array(rdma_argp); 248 if (!arg_ary) 249 return 0; 250 res_ary = (struct rpcrdma_write_array *) 251 &rdma_resp->rm_body.rm_chunks[1]; 252 253 /* Write chunks start at the pagelist */ 254 for (xdr_off = rqstp->rq_res.head[0].iov_len, chunk_no = 0; 255 xfer_len && chunk_no < arg_ary->wc_nchunks; 256 chunk_no++) { 257 struct rpcrdma_segment *arg_ch; 258 u64 rs_offset; 259 260 arg_ch = &arg_ary->wc_array[chunk_no].wc_target; 261 write_len = min(xfer_len, ntohl(arg_ch->rs_length)); 262 263 /* Prepare the response chunk given the length actually 264 * written */ 265 xdr_decode_hyper((__be32 *)&arg_ch->rs_offset, &rs_offset); 266 svc_rdma_xdr_encode_array_chunk(res_ary, chunk_no, 267 arg_ch->rs_handle, 268 arg_ch->rs_offset, 269 write_len); 270 chunk_off = 0; 271 while (write_len) { 272 ret = send_write(xprt, rqstp, 273 ntohl(arg_ch->rs_handle), 274 rs_offset + chunk_off, 275 xdr_off, 276 write_len, 277 vec); 278 if (ret <= 0) { 279 dprintk("svcrdma: RDMA_WRITE failed, ret=%d\n", 280 ret); 281 return -EIO; 282 } 283 chunk_off += ret; 284 xdr_off += ret; 285 xfer_len -= ret; 286 write_len -= ret; 287 } 288 } 289 /* Update the req with the number of chunks actually used */ 290 svc_rdma_xdr_encode_write_list(rdma_resp, chunk_no); 291 292 return rqstp->rq_res.page_len + rqstp->rq_res.tail[0].iov_len; 293} 294 295static int send_reply_chunks(struct svcxprt_rdma *xprt, 296 struct rpcrdma_msg *rdma_argp, 297 struct rpcrdma_msg *rdma_resp, 298 struct svc_rqst *rqstp, 299 struct svc_rdma_req_map *vec) 300{ 301 u32 xfer_len = rqstp->rq_res.len; 302 int write_len; 303 u32 xdr_off; 304 int chunk_no; 305 int chunk_off; 306 int nchunks; 307 struct rpcrdma_segment *ch; 308 struct rpcrdma_write_array *arg_ary; 309 struct rpcrdma_write_array *res_ary; 310 int ret; 311 312 arg_ary = svc_rdma_get_reply_array(rdma_argp); 313 if (!arg_ary) 314 return 0; 315 /* XXX: need to fix when reply lists occur with read-list and or 316 * write-list */ 317 res_ary = (struct rpcrdma_write_array *) 318 &rdma_resp->rm_body.rm_chunks[2]; 319 320 /* xdr offset starts at RPC message */ 321 nchunks = ntohl(arg_ary->wc_nchunks); 322 for (xdr_off = 0, chunk_no = 0; 323 xfer_len && chunk_no < nchunks; 324 chunk_no++) { 325 u64 rs_offset; 326 ch = &arg_ary->wc_array[chunk_no].wc_target; 327 write_len = min(xfer_len, htonl(ch->rs_length)); 328 329 /* Prepare the reply chunk given the length actually 330 * written */ 331 xdr_decode_hyper((__be32 *)&ch->rs_offset, &rs_offset); 332 svc_rdma_xdr_encode_array_chunk(res_ary, chunk_no, 333 ch->rs_handle, ch->rs_offset, 334 write_len); 335 chunk_off = 0; 336 while (write_len) { 337 ret = send_write(xprt, rqstp, 338 ntohl(ch->rs_handle), 339 rs_offset + chunk_off, 340 xdr_off, 341 write_len, 342 vec); 343 if (ret <= 0) { 344 dprintk("svcrdma: RDMA_WRITE failed, ret=%d\n", 345 ret); 346 return -EIO; 347 } 348 chunk_off += ret; 349 xdr_off += ret; 350 xfer_len -= ret; 351 write_len -= ret; 352 } 353 } 354 /* Update the req with the number of chunks actually used */ 355 svc_rdma_xdr_encode_reply_array(res_ary, chunk_no); 356 357 return rqstp->rq_res.len; 358} 359 360/* This function prepares the portion of the RPCRDMA message to be 361 * sent in the RDMA_SEND. This function is called after data sent via 362 * RDMA has already been transmitted. There are three cases: 363 * - The RPCRDMA header, RPC header, and payload are all sent in a 364 * single RDMA_SEND. This is the "inline" case. 365 * - The RPCRDMA header and some portion of the RPC header and data 366 * are sent via this RDMA_SEND and another portion of the data is 367 * sent via RDMA. 368 * - The RPCRDMA header [NOMSG] is sent in this RDMA_SEND and the RPC 369 * header and data are all transmitted via RDMA. 370 * In all three cases, this function prepares the RPCRDMA header in 371 * sge[0], the 'type' parameter indicates the type to place in the 372 * RPCRDMA header, and the 'byte_count' field indicates how much of 373 * the XDR to include in this RDMA_SEND. NB: The offset of the payload 374 * to send is zero in the XDR. 375 */ 376static int send_reply(struct svcxprt_rdma *rdma, 377 struct svc_rqst *rqstp, 378 struct page *page, 379 struct rpcrdma_msg *rdma_resp, 380 struct svc_rdma_op_ctxt *ctxt, 381 struct svc_rdma_req_map *vec, 382 int byte_count) 383{ 384 struct ib_send_wr send_wr; 385 u32 xdr_off; 386 int sge_no; 387 int sge_bytes; 388 int page_no; 389 int pages; 390 int ret; 391 392 /* Post a recv buffer to handle another request. */ 393 ret = svc_rdma_post_recv(rdma); 394 if (ret) { 395 printk(KERN_INFO 396 "svcrdma: could not post a receive buffer, err=%d." 397 "Closing transport %p.\n", ret, rdma); 398 set_bit(XPT_CLOSE, &rdma->sc_xprt.xpt_flags); 399 svc_rdma_put_context(ctxt, 0); 400 return -ENOTCONN; 401 } 402 403 /* Prepare the context */ 404 ctxt->pages[0] = page; 405 ctxt->count = 1; 406 407 /* Prepare the SGE for the RPCRDMA Header */ 408 ctxt->sge[0].lkey = rdma->sc_dma_lkey; 409 ctxt->sge[0].length = svc_rdma_xdr_get_reply_hdr_len(rdma_resp); 410 ctxt->sge[0].addr = 411 ib_dma_map_page(rdma->sc_cm_id->device, page, 0, 412 ctxt->sge[0].length, DMA_TO_DEVICE); 413 if (ib_dma_mapping_error(rdma->sc_cm_id->device, ctxt->sge[0].addr)) 414 goto err; 415 atomic_inc(&rdma->sc_dma_used); 416 417 ctxt->direction = DMA_TO_DEVICE; 418 419 /* Map the payload indicated by 'byte_count' */ 420 xdr_off = 0; 421 for (sge_no = 1; byte_count && sge_no < vec->count; sge_no++) { 422 sge_bytes = min_t(size_t, vec->sge[sge_no].iov_len, byte_count); 423 byte_count -= sge_bytes; 424 ctxt->sge[sge_no].addr = 425 dma_map_xdr(rdma, &rqstp->rq_res, xdr_off, 426 sge_bytes, DMA_TO_DEVICE); 427 xdr_off += sge_bytes; 428 if (ib_dma_mapping_error(rdma->sc_cm_id->device, 429 ctxt->sge[sge_no].addr)) 430 goto err; 431 atomic_inc(&rdma->sc_dma_used); 432 ctxt->sge[sge_no].lkey = rdma->sc_dma_lkey; 433 ctxt->sge[sge_no].length = sge_bytes; 434 } 435 if (byte_count != 0) { 436 pr_err("svcrdma: Could not map %d bytes\n", byte_count); 437 goto err; 438 } 439 440 /* Save all respages in the ctxt and remove them from the 441 * respages array. They are our pages until the I/O 442 * completes. 443 */ 444 pages = rqstp->rq_next_page - rqstp->rq_respages; 445 for (page_no = 0; page_no < pages; page_no++) { 446 ctxt->pages[page_no+1] = rqstp->rq_respages[page_no]; 447 ctxt->count++; 448 rqstp->rq_respages[page_no] = NULL; 449 /* 450 * If there are more pages than SGE, terminate SGE 451 * list so that svc_rdma_unmap_dma doesn't attempt to 452 * unmap garbage. 453 */ 454 if (page_no+1 >= sge_no) 455 ctxt->sge[page_no+1].length = 0; 456 } 457 rqstp->rq_next_page = rqstp->rq_respages + 1; 458 459 /* The loop above bumps sc_dma_used for each sge. The 460 * xdr_buf.tail gets a separate sge, but resides in the 461 * same page as xdr_buf.head. Don't count it twice. 462 */ 463 if (sge_no > ctxt->count) 464 atomic_dec(&rdma->sc_dma_used); 465 466 if (sge_no > rdma->sc_max_sge) { 467 pr_err("svcrdma: Too many sges (%d)\n", sge_no); 468 goto err; 469 } 470 memset(&send_wr, 0, sizeof send_wr); 471 ctxt->wr_op = IB_WR_SEND; 472 send_wr.wr_id = (unsigned long)ctxt; 473 send_wr.sg_list = ctxt->sge; 474 send_wr.num_sge = sge_no; 475 send_wr.opcode = IB_WR_SEND; 476 send_wr.send_flags = IB_SEND_SIGNALED; 477 478 ret = svc_rdma_send(rdma, &send_wr); 479 if (ret) 480 goto err; 481 482 return 0; 483 484 err: 485 svc_rdma_unmap_dma(ctxt); 486 svc_rdma_put_context(ctxt, 1); 487 return -EIO; 488} 489 490void svc_rdma_prep_reply_hdr(struct svc_rqst *rqstp) 491{ 492} 493 494int svc_rdma_sendto(struct svc_rqst *rqstp) 495{ 496 struct svc_xprt *xprt = rqstp->rq_xprt; 497 struct svcxprt_rdma *rdma = 498 container_of(xprt, struct svcxprt_rdma, sc_xprt); 499 struct rpcrdma_msg *rdma_argp; 500 struct rpcrdma_msg *rdma_resp; 501 struct rpcrdma_write_array *reply_ary; 502 enum rpcrdma_proc reply_type; 503 int ret; 504 int inline_bytes; 505 struct page *res_page; 506 struct svc_rdma_op_ctxt *ctxt; 507 struct svc_rdma_req_map *vec; 508 509 dprintk("svcrdma: sending response for rqstp=%p\n", rqstp); 510 511 /* Get the RDMA request header. The receive logic always 512 * places this at the start of page 0. 513 */ 514 rdma_argp = page_address(rqstp->rq_pages[0]); 515 516 /* Build an req vec for the XDR */ 517 ctxt = svc_rdma_get_context(rdma); 518 ctxt->direction = DMA_TO_DEVICE; 519 vec = svc_rdma_get_req_map(); 520 ret = map_xdr(rdma, &rqstp->rq_res, vec); 521 if (ret) 522 goto err0; 523 inline_bytes = rqstp->rq_res.len; 524 525 /* Create the RDMA response header */ 526 res_page = svc_rdma_get_page(); 527 rdma_resp = page_address(res_page); 528 reply_ary = svc_rdma_get_reply_array(rdma_argp); 529 if (reply_ary) 530 reply_type = RDMA_NOMSG; 531 else 532 reply_type = RDMA_MSG; 533 svc_rdma_xdr_encode_reply_header(rdma, rdma_argp, 534 rdma_resp, reply_type); 535 536 /* Send any write-chunk data and build resp write-list */ 537 ret = send_write_chunks(rdma, rdma_argp, rdma_resp, 538 rqstp, vec); 539 if (ret < 0) { 540 printk(KERN_ERR "svcrdma: failed to send write chunks, rc=%d\n", 541 ret); 542 goto err1; 543 } 544 inline_bytes -= ret; 545 546 /* Send any reply-list data and update resp reply-list */ 547 ret = send_reply_chunks(rdma, rdma_argp, rdma_resp, 548 rqstp, vec); 549 if (ret < 0) { 550 printk(KERN_ERR "svcrdma: failed to send reply chunks, rc=%d\n", 551 ret); 552 goto err1; 553 } 554 inline_bytes -= ret; 555 556 ret = send_reply(rdma, rqstp, res_page, rdma_resp, ctxt, vec, 557 inline_bytes); 558 svc_rdma_put_req_map(vec); 559 dprintk("svcrdma: send_reply returns %d\n", ret); 560 return ret; 561 562 err1: 563 put_page(res_page); 564 err0: 565 svc_rdma_put_req_map(vec); 566 svc_rdma_put_context(ctxt, 0); 567 return ret; 568} 569