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 53/* 54 * Replace the pages in the rq_argpages array with the pages from the SGE in 55 * the RDMA_RECV completion. The SGL should contain full pages up until the 56 * last one. 57 */ 58static void rdma_build_arg_xdr(struct svc_rqst *rqstp, 59 struct svc_rdma_op_ctxt *ctxt, 60 u32 byte_count) 61{ 62 struct rpcrdma_msg *rmsgp; 63 struct page *page; 64 u32 bc; 65 int sge_no; 66 67 /* Swap the page in the SGE with the page in argpages */ 68 page = ctxt->pages[0]; 69 put_page(rqstp->rq_pages[0]); 70 rqstp->rq_pages[0] = page; 71 72 /* Set up the XDR head */ 73 rqstp->rq_arg.head[0].iov_base = page_address(page); 74 rqstp->rq_arg.head[0].iov_len = 75 min_t(size_t, byte_count, ctxt->sge[0].length); 76 rqstp->rq_arg.len = byte_count; 77 rqstp->rq_arg.buflen = byte_count; 78 79 /* Compute bytes past head in the SGL */ 80 bc = byte_count - rqstp->rq_arg.head[0].iov_len; 81 82 /* If data remains, store it in the pagelist */ 83 rqstp->rq_arg.page_len = bc; 84 rqstp->rq_arg.page_base = 0; 85 86 /* RDMA_NOMSG: RDMA READ data should land just after RDMA RECV data */ 87 rmsgp = (struct rpcrdma_msg *)rqstp->rq_arg.head[0].iov_base; 88 if (be32_to_cpu(rmsgp->rm_type) == RDMA_NOMSG) 89 rqstp->rq_arg.pages = &rqstp->rq_pages[0]; 90 else 91 rqstp->rq_arg.pages = &rqstp->rq_pages[1]; 92 93 sge_no = 1; 94 while (bc && sge_no < ctxt->count) { 95 page = ctxt->pages[sge_no]; 96 put_page(rqstp->rq_pages[sge_no]); 97 rqstp->rq_pages[sge_no] = page; 98 bc -= min_t(u32, bc, ctxt->sge[sge_no].length); 99 rqstp->rq_arg.buflen += ctxt->sge[sge_no].length; 100 sge_no++; 101 } 102 rqstp->rq_respages = &rqstp->rq_pages[sge_no]; 103 rqstp->rq_next_page = rqstp->rq_respages + 1; 104 105 /* If not all pages were used from the SGL, free the remaining ones */ 106 bc = sge_no; 107 while (sge_no < ctxt->count) { 108 page = ctxt->pages[sge_no++]; 109 put_page(page); 110 } 111 ctxt->count = bc; 112 113 /* Set up tail */ 114 rqstp->rq_arg.tail[0].iov_base = NULL; 115 rqstp->rq_arg.tail[0].iov_len = 0; 116} 117 118static int rdma_read_max_sge(struct svcxprt_rdma *xprt, int sge_count) 119{ 120 if (rdma_node_get_transport(xprt->sc_cm_id->device->node_type) == 121 RDMA_TRANSPORT_IWARP) 122 return 1; 123 else 124 return min_t(int, sge_count, xprt->sc_max_sge); 125} 126 127/* Issue an RDMA_READ using the local lkey to map the data sink */ 128int rdma_read_chunk_lcl(struct svcxprt_rdma *xprt, 129 struct svc_rqst *rqstp, 130 struct svc_rdma_op_ctxt *head, 131 int *page_no, 132 u32 *page_offset, 133 u32 rs_handle, 134 u32 rs_length, 135 u64 rs_offset, 136 bool last) 137{ 138 struct ib_send_wr read_wr; 139 int pages_needed = PAGE_ALIGN(*page_offset + rs_length) >> PAGE_SHIFT; 140 struct svc_rdma_op_ctxt *ctxt = svc_rdma_get_context(xprt); 141 int ret, read, pno; 142 u32 pg_off = *page_offset; 143 u32 pg_no = *page_no; 144 145 ctxt->direction = DMA_FROM_DEVICE; 146 ctxt->read_hdr = head; 147 pages_needed = 148 min_t(int, pages_needed, rdma_read_max_sge(xprt, pages_needed)); 149 read = min_t(int, (pages_needed << PAGE_SHIFT) - *page_offset, 150 rs_length); 151 152 for (pno = 0; pno < pages_needed; pno++) { 153 int len = min_t(int, rs_length, PAGE_SIZE - pg_off); 154 155 head->arg.pages[pg_no] = rqstp->rq_arg.pages[pg_no]; 156 head->arg.page_len += len; 157 head->arg.len += len; 158 if (!pg_off) 159 head->count++; 160 rqstp->rq_respages = &rqstp->rq_arg.pages[pg_no+1]; 161 rqstp->rq_next_page = rqstp->rq_respages + 1; 162 ctxt->sge[pno].addr = 163 ib_dma_map_page(xprt->sc_cm_id->device, 164 head->arg.pages[pg_no], pg_off, 165 PAGE_SIZE - pg_off, 166 DMA_FROM_DEVICE); 167 ret = ib_dma_mapping_error(xprt->sc_cm_id->device, 168 ctxt->sge[pno].addr); 169 if (ret) 170 goto err; 171 atomic_inc(&xprt->sc_dma_used); 172 173 /* The lkey here is either a local dma lkey or a dma_mr lkey */ 174 ctxt->sge[pno].lkey = xprt->sc_dma_lkey; 175 ctxt->sge[pno].length = len; 176 ctxt->count++; 177 178 /* adjust offset and wrap to next page if needed */ 179 pg_off += len; 180 if (pg_off == PAGE_SIZE) { 181 pg_off = 0; 182 pg_no++; 183 } 184 rs_length -= len; 185 } 186 187 if (last && rs_length == 0) 188 set_bit(RDMACTXT_F_LAST_CTXT, &ctxt->flags); 189 else 190 clear_bit(RDMACTXT_F_LAST_CTXT, &ctxt->flags); 191 192 memset(&read_wr, 0, sizeof(read_wr)); 193 read_wr.wr_id = (unsigned long)ctxt; 194 read_wr.opcode = IB_WR_RDMA_READ; 195 ctxt->wr_op = read_wr.opcode; 196 read_wr.send_flags = IB_SEND_SIGNALED; 197 read_wr.wr.rdma.rkey = rs_handle; 198 read_wr.wr.rdma.remote_addr = rs_offset; 199 read_wr.sg_list = ctxt->sge; 200 read_wr.num_sge = pages_needed; 201 202 ret = svc_rdma_send(xprt, &read_wr); 203 if (ret) { 204 pr_err("svcrdma: Error %d posting RDMA_READ\n", ret); 205 set_bit(XPT_CLOSE, &xprt->sc_xprt.xpt_flags); 206 goto err; 207 } 208 209 /* return current location in page array */ 210 *page_no = pg_no; 211 *page_offset = pg_off; 212 ret = read; 213 atomic_inc(&rdma_stat_read); 214 return ret; 215 err: 216 svc_rdma_unmap_dma(ctxt); 217 svc_rdma_put_context(ctxt, 0); 218 return ret; 219} 220 221/* Issue an RDMA_READ using an FRMR to map the data sink */ 222int rdma_read_chunk_frmr(struct svcxprt_rdma *xprt, 223 struct svc_rqst *rqstp, 224 struct svc_rdma_op_ctxt *head, 225 int *page_no, 226 u32 *page_offset, 227 u32 rs_handle, 228 u32 rs_length, 229 u64 rs_offset, 230 bool last) 231{ 232 struct ib_send_wr read_wr; 233 struct ib_send_wr inv_wr; 234 struct ib_send_wr fastreg_wr; 235 u8 key; 236 int pages_needed = PAGE_ALIGN(*page_offset + rs_length) >> PAGE_SHIFT; 237 struct svc_rdma_op_ctxt *ctxt = svc_rdma_get_context(xprt); 238 struct svc_rdma_fastreg_mr *frmr = svc_rdma_get_frmr(xprt); 239 int ret, read, pno; 240 u32 pg_off = *page_offset; 241 u32 pg_no = *page_no; 242 243 if (IS_ERR(frmr)) 244 return -ENOMEM; 245 246 ctxt->direction = DMA_FROM_DEVICE; 247 ctxt->frmr = frmr; 248 pages_needed = min_t(int, pages_needed, xprt->sc_frmr_pg_list_len); 249 read = min_t(int, (pages_needed << PAGE_SHIFT) - *page_offset, 250 rs_length); 251 252 frmr->kva = page_address(rqstp->rq_arg.pages[pg_no]); 253 frmr->direction = DMA_FROM_DEVICE; 254 frmr->access_flags = (IB_ACCESS_LOCAL_WRITE|IB_ACCESS_REMOTE_WRITE); 255 frmr->map_len = pages_needed << PAGE_SHIFT; 256 frmr->page_list_len = pages_needed; 257 258 for (pno = 0; pno < pages_needed; pno++) { 259 int len = min_t(int, rs_length, PAGE_SIZE - pg_off); 260 261 head->arg.pages[pg_no] = rqstp->rq_arg.pages[pg_no]; 262 head->arg.page_len += len; 263 head->arg.len += len; 264 if (!pg_off) 265 head->count++; 266 rqstp->rq_respages = &rqstp->rq_arg.pages[pg_no+1]; 267 rqstp->rq_next_page = rqstp->rq_respages + 1; 268 frmr->page_list->page_list[pno] = 269 ib_dma_map_page(xprt->sc_cm_id->device, 270 head->arg.pages[pg_no], 0, 271 PAGE_SIZE, DMA_FROM_DEVICE); 272 ret = ib_dma_mapping_error(xprt->sc_cm_id->device, 273 frmr->page_list->page_list[pno]); 274 if (ret) 275 goto err; 276 atomic_inc(&xprt->sc_dma_used); 277 278 /* adjust offset and wrap to next page if needed */ 279 pg_off += len; 280 if (pg_off == PAGE_SIZE) { 281 pg_off = 0; 282 pg_no++; 283 } 284 rs_length -= len; 285 } 286 287 if (last && rs_length == 0) 288 set_bit(RDMACTXT_F_LAST_CTXT, &ctxt->flags); 289 else 290 clear_bit(RDMACTXT_F_LAST_CTXT, &ctxt->flags); 291 292 /* Bump the key */ 293 key = (u8)(frmr->mr->lkey & 0x000000FF); 294 ib_update_fast_reg_key(frmr->mr, ++key); 295 296 ctxt->sge[0].addr = (unsigned long)frmr->kva + *page_offset; 297 ctxt->sge[0].lkey = frmr->mr->lkey; 298 ctxt->sge[0].length = read; 299 ctxt->count = 1; 300 ctxt->read_hdr = head; 301 302 /* Prepare FASTREG WR */ 303 memset(&fastreg_wr, 0, sizeof(fastreg_wr)); 304 fastreg_wr.opcode = IB_WR_FAST_REG_MR; 305 fastreg_wr.send_flags = IB_SEND_SIGNALED; 306 fastreg_wr.wr.fast_reg.iova_start = (unsigned long)frmr->kva; 307 fastreg_wr.wr.fast_reg.page_list = frmr->page_list; 308 fastreg_wr.wr.fast_reg.page_list_len = frmr->page_list_len; 309 fastreg_wr.wr.fast_reg.page_shift = PAGE_SHIFT; 310 fastreg_wr.wr.fast_reg.length = frmr->map_len; 311 fastreg_wr.wr.fast_reg.access_flags = frmr->access_flags; 312 fastreg_wr.wr.fast_reg.rkey = frmr->mr->lkey; 313 fastreg_wr.next = &read_wr; 314 315 /* Prepare RDMA_READ */ 316 memset(&read_wr, 0, sizeof(read_wr)); 317 read_wr.send_flags = IB_SEND_SIGNALED; 318 read_wr.wr.rdma.rkey = rs_handle; 319 read_wr.wr.rdma.remote_addr = rs_offset; 320 read_wr.sg_list = ctxt->sge; 321 read_wr.num_sge = 1; 322 if (xprt->sc_dev_caps & SVCRDMA_DEVCAP_READ_W_INV) { 323 read_wr.opcode = IB_WR_RDMA_READ_WITH_INV; 324 read_wr.wr_id = (unsigned long)ctxt; 325 read_wr.ex.invalidate_rkey = ctxt->frmr->mr->lkey; 326 } else { 327 read_wr.opcode = IB_WR_RDMA_READ; 328 read_wr.next = &inv_wr; 329 /* Prepare invalidate */ 330 memset(&inv_wr, 0, sizeof(inv_wr)); 331 inv_wr.wr_id = (unsigned long)ctxt; 332 inv_wr.opcode = IB_WR_LOCAL_INV; 333 inv_wr.send_flags = IB_SEND_SIGNALED | IB_SEND_FENCE; 334 inv_wr.ex.invalidate_rkey = frmr->mr->lkey; 335 } 336 ctxt->wr_op = read_wr.opcode; 337 338 /* Post the chain */ 339 ret = svc_rdma_send(xprt, &fastreg_wr); 340 if (ret) { 341 pr_err("svcrdma: Error %d posting RDMA_READ\n", ret); 342 set_bit(XPT_CLOSE, &xprt->sc_xprt.xpt_flags); 343 goto err; 344 } 345 346 /* return current location in page array */ 347 *page_no = pg_no; 348 *page_offset = pg_off; 349 ret = read; 350 atomic_inc(&rdma_stat_read); 351 return ret; 352 err: 353 svc_rdma_unmap_dma(ctxt); 354 svc_rdma_put_context(ctxt, 0); 355 svc_rdma_put_frmr(xprt, frmr); 356 return ret; 357} 358 359static unsigned int 360rdma_rcl_chunk_count(struct rpcrdma_read_chunk *ch) 361{ 362 unsigned int count; 363 364 for (count = 0; ch->rc_discrim != xdr_zero; ch++) 365 count++; 366 return count; 367} 368 369/* If there was additional inline content, append it to the end of arg.pages. 370 * Tail copy has to be done after the reader function has determined how many 371 * pages are needed for RDMA READ. 372 */ 373static int 374rdma_copy_tail(struct svc_rqst *rqstp, struct svc_rdma_op_ctxt *head, 375 u32 position, u32 byte_count, u32 page_offset, int page_no) 376{ 377 char *srcp, *destp; 378 int ret; 379 380 ret = 0; 381 srcp = head->arg.head[0].iov_base + position; 382 byte_count = head->arg.head[0].iov_len - position; 383 if (byte_count > PAGE_SIZE) { 384 dprintk("svcrdma: large tail unsupported\n"); 385 return 0; 386 } 387 388 /* Fit as much of the tail on the current page as possible */ 389 if (page_offset != PAGE_SIZE) { 390 destp = page_address(rqstp->rq_arg.pages[page_no]); 391 destp += page_offset; 392 while (byte_count--) { 393 *destp++ = *srcp++; 394 page_offset++; 395 if (page_offset == PAGE_SIZE && byte_count) 396 goto more; 397 } 398 goto done; 399 } 400 401more: 402 /* Fit the rest on the next page */ 403 page_no++; 404 destp = page_address(rqstp->rq_arg.pages[page_no]); 405 while (byte_count--) 406 *destp++ = *srcp++; 407 408 rqstp->rq_respages = &rqstp->rq_arg.pages[page_no+1]; 409 rqstp->rq_next_page = rqstp->rq_respages + 1; 410 411done: 412 byte_count = head->arg.head[0].iov_len - position; 413 head->arg.page_len += byte_count; 414 head->arg.len += byte_count; 415 head->arg.buflen += byte_count; 416 return 1; 417} 418 419static int rdma_read_chunks(struct svcxprt_rdma *xprt, 420 struct rpcrdma_msg *rmsgp, 421 struct svc_rqst *rqstp, 422 struct svc_rdma_op_ctxt *head) 423{ 424 int page_no, ret; 425 struct rpcrdma_read_chunk *ch; 426 u32 handle, page_offset, byte_count; 427 u32 position; 428 u64 rs_offset; 429 bool last; 430 431 /* If no read list is present, return 0 */ 432 ch = svc_rdma_get_read_chunk(rmsgp); 433 if (!ch) 434 return 0; 435 436 if (rdma_rcl_chunk_count(ch) > RPCSVC_MAXPAGES) 437 return -EINVAL; 438 439 /* The request is completed when the RDMA_READs complete. The 440 * head context keeps all the pages that comprise the 441 * request. 442 */ 443 head->arg.head[0] = rqstp->rq_arg.head[0]; 444 head->arg.tail[0] = rqstp->rq_arg.tail[0]; 445 head->hdr_count = head->count; 446 head->arg.page_base = 0; 447 head->arg.page_len = 0; 448 head->arg.len = rqstp->rq_arg.len; 449 head->arg.buflen = rqstp->rq_arg.buflen; 450 451 ch = (struct rpcrdma_read_chunk *)&rmsgp->rm_body.rm_chunks[0]; 452 position = be32_to_cpu(ch->rc_position); 453 454 /* RDMA_NOMSG: RDMA READ data should land just after RDMA RECV data */ 455 if (position == 0) { 456 head->arg.pages = &head->pages[0]; 457 page_offset = head->byte_len; 458 } else { 459 head->arg.pages = &head->pages[head->count]; 460 page_offset = 0; 461 } 462 463 ret = 0; 464 page_no = 0; 465 for (; ch->rc_discrim != xdr_zero; ch++) { 466 if (be32_to_cpu(ch->rc_position) != position) 467 goto err; 468 469 handle = be32_to_cpu(ch->rc_target.rs_handle), 470 byte_count = be32_to_cpu(ch->rc_target.rs_length); 471 xdr_decode_hyper((__be32 *)&ch->rc_target.rs_offset, 472 &rs_offset); 473 474 while (byte_count > 0) { 475 last = (ch + 1)->rc_discrim == xdr_zero; 476 ret = xprt->sc_reader(xprt, rqstp, head, 477 &page_no, &page_offset, 478 handle, byte_count, 479 rs_offset, last); 480 if (ret < 0) 481 goto err; 482 byte_count -= ret; 483 rs_offset += ret; 484 head->arg.buflen += ret; 485 } 486 } 487 488 /* Read list may need XDR round-up (see RFC 5666, s. 3.7) */ 489 if (page_offset & 3) { 490 u32 pad = 4 - (page_offset & 3); 491 492 head->arg.page_len += pad; 493 head->arg.len += pad; 494 head->arg.buflen += pad; 495 page_offset += pad; 496 } 497 498 ret = 1; 499 if (position && position < head->arg.head[0].iov_len) 500 ret = rdma_copy_tail(rqstp, head, position, 501 byte_count, page_offset, page_no); 502 head->arg.head[0].iov_len = position; 503 head->position = position; 504 505 err: 506 /* Detach arg pages. svc_recv will replenish them */ 507 for (page_no = 0; 508 &rqstp->rq_pages[page_no] < rqstp->rq_respages; page_no++) 509 rqstp->rq_pages[page_no] = NULL; 510 511 return ret; 512} 513 514static int rdma_read_complete(struct svc_rqst *rqstp, 515 struct svc_rdma_op_ctxt *head) 516{ 517 int page_no; 518 int ret; 519 520 /* Copy RPC pages */ 521 for (page_no = 0; page_no < head->count; page_no++) { 522 put_page(rqstp->rq_pages[page_no]); 523 rqstp->rq_pages[page_no] = head->pages[page_no]; 524 } 525 526 /* Adjustments made for RDMA_NOMSG type requests */ 527 if (head->position == 0) { 528 if (head->arg.len <= head->sge[0].length) { 529 head->arg.head[0].iov_len = head->arg.len - 530 head->byte_len; 531 head->arg.page_len = 0; 532 } else { 533 head->arg.head[0].iov_len = head->sge[0].length - 534 head->byte_len; 535 head->arg.page_len = head->arg.len - 536 head->sge[0].length; 537 } 538 } 539 540 /* Point rq_arg.pages past header */ 541 rqstp->rq_arg.pages = &rqstp->rq_pages[head->hdr_count]; 542 rqstp->rq_arg.page_len = head->arg.page_len; 543 rqstp->rq_arg.page_base = head->arg.page_base; 544 545 /* rq_respages starts after the last arg page */ 546 rqstp->rq_respages = &rqstp->rq_arg.pages[page_no]; 547 rqstp->rq_next_page = rqstp->rq_respages + 1; 548 549 /* Rebuild rq_arg head and tail. */ 550 rqstp->rq_arg.head[0] = head->arg.head[0]; 551 rqstp->rq_arg.tail[0] = head->arg.tail[0]; 552 rqstp->rq_arg.len = head->arg.len; 553 rqstp->rq_arg.buflen = head->arg.buflen; 554 555 /* Free the context */ 556 svc_rdma_put_context(head, 0); 557 558 /* XXX: What should this be? */ 559 rqstp->rq_prot = IPPROTO_MAX; 560 svc_xprt_copy_addrs(rqstp, rqstp->rq_xprt); 561 562 ret = rqstp->rq_arg.head[0].iov_len 563 + rqstp->rq_arg.page_len 564 + rqstp->rq_arg.tail[0].iov_len; 565 dprintk("svcrdma: deferred read ret=%d, rq_arg.len=%u, " 566 "rq_arg.head[0].iov_base=%p, rq_arg.head[0].iov_len=%zu\n", 567 ret, rqstp->rq_arg.len, rqstp->rq_arg.head[0].iov_base, 568 rqstp->rq_arg.head[0].iov_len); 569 570 return ret; 571} 572 573/* 574 * Set up the rqstp thread context to point to the RQ buffer. If 575 * necessary, pull additional data from the client with an RDMA_READ 576 * request. 577 */ 578int svc_rdma_recvfrom(struct svc_rqst *rqstp) 579{ 580 struct svc_xprt *xprt = rqstp->rq_xprt; 581 struct svcxprt_rdma *rdma_xprt = 582 container_of(xprt, struct svcxprt_rdma, sc_xprt); 583 struct svc_rdma_op_ctxt *ctxt = NULL; 584 struct rpcrdma_msg *rmsgp; 585 int ret = 0; 586 int len; 587 588 dprintk("svcrdma: rqstp=%p\n", rqstp); 589 590 spin_lock_bh(&rdma_xprt->sc_rq_dto_lock); 591 if (!list_empty(&rdma_xprt->sc_read_complete_q)) { 592 ctxt = list_entry(rdma_xprt->sc_read_complete_q.next, 593 struct svc_rdma_op_ctxt, 594 dto_q); 595 list_del_init(&ctxt->dto_q); 596 spin_unlock_bh(&rdma_xprt->sc_rq_dto_lock); 597 return rdma_read_complete(rqstp, ctxt); 598 } else if (!list_empty(&rdma_xprt->sc_rq_dto_q)) { 599 ctxt = list_entry(rdma_xprt->sc_rq_dto_q.next, 600 struct svc_rdma_op_ctxt, 601 dto_q); 602 list_del_init(&ctxt->dto_q); 603 } else { 604 atomic_inc(&rdma_stat_rq_starve); 605 clear_bit(XPT_DATA, &xprt->xpt_flags); 606 ctxt = NULL; 607 } 608 spin_unlock_bh(&rdma_xprt->sc_rq_dto_lock); 609 if (!ctxt) { 610 /* This is the EAGAIN path. The svc_recv routine will 611 * return -EAGAIN, the nfsd thread will go to call into 612 * svc_recv again and we shouldn't be on the active 613 * transport list 614 */ 615 if (test_bit(XPT_CLOSE, &xprt->xpt_flags)) 616 goto close_out; 617 618 goto out; 619 } 620 dprintk("svcrdma: processing ctxt=%p on xprt=%p, rqstp=%p, status=%d\n", 621 ctxt, rdma_xprt, rqstp, ctxt->wc_status); 622 atomic_inc(&rdma_stat_recv); 623 624 /* Build up the XDR from the receive buffers. */ 625 rdma_build_arg_xdr(rqstp, ctxt, ctxt->byte_len); 626 627 /* Decode the RDMA header. */ 628 len = svc_rdma_xdr_decode_req(&rmsgp, rqstp); 629 rqstp->rq_xprt_hlen = len; 630 631 /* If the request is invalid, reply with an error */ 632 if (len < 0) { 633 if (len == -ENOSYS) 634 svc_rdma_send_error(rdma_xprt, rmsgp, ERR_VERS); 635 goto close_out; 636 } 637 638 /* Read read-list data. */ 639 ret = rdma_read_chunks(rdma_xprt, rmsgp, rqstp, ctxt); 640 if (ret > 0) { 641 /* read-list posted, defer until data received from client. */ 642 goto defer; 643 } else if (ret < 0) { 644 /* Post of read-list failed, free context. */ 645 svc_rdma_put_context(ctxt, 1); 646 return 0; 647 } 648 649 ret = rqstp->rq_arg.head[0].iov_len 650 + rqstp->rq_arg.page_len 651 + rqstp->rq_arg.tail[0].iov_len; 652 svc_rdma_put_context(ctxt, 0); 653 out: 654 dprintk("svcrdma: ret=%d, rq_arg.len=%u, " 655 "rq_arg.head[0].iov_base=%p, rq_arg.head[0].iov_len=%zd\n", 656 ret, rqstp->rq_arg.len, 657 rqstp->rq_arg.head[0].iov_base, 658 rqstp->rq_arg.head[0].iov_len); 659 rqstp->rq_prot = IPPROTO_MAX; 660 svc_xprt_copy_addrs(rqstp, xprt); 661 return ret; 662 663 close_out: 664 if (ctxt) 665 svc_rdma_put_context(ctxt, 1); 666 dprintk("svcrdma: transport %p is closing\n", xprt); 667 /* 668 * Set the close bit and enqueue it. svc_recv will see the 669 * close bit and call svc_xprt_delete 670 */ 671 set_bit(XPT_CLOSE, &xprt->xpt_flags); 672defer: 673 return 0; 674} 675