1 // SPDX-License-Identifier: GPL-2.0-only 2 /* 3 * linux/fs/nfs/direct.c 4 * 5 * Copyright (C) 2003 by Chuck Lever <cel@netapp.com> 6 * 7 * High-performance uncached I/O for the Linux NFS client 8 * 9 * There are important applications whose performance or correctness 10 * depends on uncached access to file data. Database clusters 11 * (multiple copies of the same instance running on separate hosts) 12 * implement their own cache coherency protocol that subsumes file 13 * system cache protocols. Applications that process datasets 14 * considerably larger than the client's memory do not always benefit 15 * from a local cache. A streaming video server, for instance, has no 16 * need to cache the contents of a file. 17 * 18 * When an application requests uncached I/O, all read and write requests 19 * are made directly to the server; data stored or fetched via these 20 * requests is not cached in the Linux page cache. The client does not 21 * correct unaligned requests from applications. All requested bytes are 22 * held on permanent storage before a direct write system call returns to 23 * an application. 24 * 25 * Solaris implements an uncached I/O facility called directio() that 26 * is used for backups and sequential I/O to very large files. Solaris 27 * also supports uncaching whole NFS partitions with "-o forcedirectio," 28 * an undocumented mount option. 29 * 30 * Designed by Jeff Kimmel, Chuck Lever, and Trond Myklebust, with 31 * help from Andrew Morton. 32 * 33 * 18 Dec 2001 Initial implementation for 2.4 --cel 34 * 08 Jul 2002 Version for 2.4.19, with bug fixes --trondmy 35 * 08 Jun 2003 Port to 2.5 APIs --cel 36 * 31 Mar 2004 Handle direct I/O without VFS support --cel 37 * 15 Sep 2004 Parallel async reads --cel 38 * 04 May 2005 support O_DIRECT with aio --cel 39 * 40 */ 41 42 #include <linux/errno.h> 43 #include <linux/sched.h> 44 #include <linux/kernel.h> 45 #include <linux/file.h> 46 #include <linux/pagemap.h> 47 #include <linux/kref.h> 48 #include <linux/slab.h> 49 #include <linux/task_io_accounting_ops.h> 50 #include <linux/module.h> 51 52 #include <linux/nfs_fs.h> 53 #include <linux/nfs_page.h> 54 #include <linux/sunrpc/clnt.h> 55 56 #include <linux/uaccess.h> 57 #include <linux/atomic.h> 58 59 #include "internal.h" 60 #include "iostat.h" 61 #include "pnfs.h" 62 #include "fscache.h" 63 #include "nfstrace.h" 64 65 #define NFSDBG_FACILITY NFSDBG_VFS 66 67 static struct kmem_cache *nfs_direct_cachep; 68 69 static const struct nfs_pgio_completion_ops nfs_direct_write_completion_ops; 70 static const struct nfs_commit_completion_ops nfs_direct_commit_completion_ops; 71 static void nfs_direct_write_complete(struct nfs_direct_req *dreq); 72 static void nfs_direct_write_schedule_work(struct work_struct *work); 73 74 static inline void get_dreq(struct nfs_direct_req *dreq) 75 { 76 atomic_inc(&dreq->io_count); 77 } 78 79 static inline int put_dreq(struct nfs_direct_req *dreq) 80 { 81 return atomic_dec_and_test(&dreq->io_count); 82 } 83 84 static void 85 nfs_direct_handle_truncated(struct nfs_direct_req *dreq, 86 const struct nfs_pgio_header *hdr, 87 ssize_t dreq_len) 88 { 89 if (!(test_bit(NFS_IOHDR_ERROR, &hdr->flags) || 90 test_bit(NFS_IOHDR_EOF, &hdr->flags))) 91 return; 92 if (dreq->max_count >= dreq_len) { 93 dreq->max_count = dreq_len; 94 if (dreq->count > dreq_len) 95 dreq->count = dreq_len; 96 97 if (test_bit(NFS_IOHDR_ERROR, &hdr->flags)) 98 dreq->error = hdr->error; 99 else /* Clear outstanding error if this is EOF */ 100 dreq->error = 0; 101 } 102 } 103 104 static void 105 nfs_direct_count_bytes(struct nfs_direct_req *dreq, 106 const struct nfs_pgio_header *hdr) 107 { 108 loff_t hdr_end = hdr->io_start + hdr->good_bytes; 109 ssize_t dreq_len = 0; 110 111 if (hdr_end > dreq->io_start) 112 dreq_len = hdr_end - dreq->io_start; 113 114 nfs_direct_handle_truncated(dreq, hdr, dreq_len); 115 116 if (dreq_len > dreq->max_count) 117 dreq_len = dreq->max_count; 118 119 if (dreq->count < dreq_len) 120 dreq->count = dreq_len; 121 } 122 123 /** 124 * nfs_swap_rw - NFS address space operation for swap I/O 125 * @iocb: target I/O control block 126 * @iter: I/O buffer 127 * 128 * Perform IO to the swap-file. This is much like direct IO. 129 */ 130 int nfs_swap_rw(struct kiocb *iocb, struct iov_iter *iter) 131 { 132 ssize_t ret; 133 134 VM_BUG_ON(iov_iter_count(iter) != PAGE_SIZE); 135 136 if (iov_iter_rw(iter) == READ) 137 ret = nfs_file_direct_read(iocb, iter, true); 138 else 139 ret = nfs_file_direct_write(iocb, iter, true); 140 if (ret < 0) 141 return ret; 142 return 0; 143 } 144 145 static void nfs_direct_release_pages(struct page **pages, unsigned int npages) 146 { 147 unsigned int i; 148 for (i = 0; i < npages; i++) 149 put_page(pages[i]); 150 } 151 152 void nfs_init_cinfo_from_dreq(struct nfs_commit_info *cinfo, 153 struct nfs_direct_req *dreq) 154 { 155 cinfo->inode = dreq->inode; 156 cinfo->mds = &dreq->mds_cinfo; 157 cinfo->ds = &dreq->ds_cinfo; 158 cinfo->dreq = dreq; 159 cinfo->completion_ops = &nfs_direct_commit_completion_ops; 160 } 161 162 static inline struct nfs_direct_req *nfs_direct_req_alloc(void) 163 { 164 struct nfs_direct_req *dreq; 165 166 dreq = kmem_cache_zalloc(nfs_direct_cachep, GFP_KERNEL); 167 if (!dreq) 168 return NULL; 169 170 kref_init(&dreq->kref); 171 kref_get(&dreq->kref); 172 init_completion(&dreq->completion); 173 INIT_LIST_HEAD(&dreq->mds_cinfo.list); 174 pnfs_init_ds_commit_info(&dreq->ds_cinfo); 175 INIT_WORK(&dreq->work, nfs_direct_write_schedule_work); 176 spin_lock_init(&dreq->lock); 177 178 return dreq; 179 } 180 181 static void nfs_direct_req_free(struct kref *kref) 182 { 183 struct nfs_direct_req *dreq = container_of(kref, struct nfs_direct_req, kref); 184 185 pnfs_release_ds_info(&dreq->ds_cinfo, dreq->inode); 186 if (dreq->l_ctx != NULL) 187 nfs_put_lock_context(dreq->l_ctx); 188 if (dreq->ctx != NULL) 189 put_nfs_open_context(dreq->ctx); 190 kmem_cache_free(nfs_direct_cachep, dreq); 191 } 192 193 static void nfs_direct_req_release(struct nfs_direct_req *dreq) 194 { 195 kref_put(&dreq->kref, nfs_direct_req_free); 196 } 197 198 ssize_t nfs_dreq_bytes_left(struct nfs_direct_req *dreq) 199 { 200 return dreq->bytes_left; 201 } 202 EXPORT_SYMBOL_GPL(nfs_dreq_bytes_left); 203 204 /* 205 * Collects and returns the final error value/byte-count. 206 */ 207 static ssize_t nfs_direct_wait(struct nfs_direct_req *dreq) 208 { 209 ssize_t result = -EIOCBQUEUED; 210 211 /* Async requests don't wait here */ 212 if (dreq->iocb) 213 goto out; 214 215 result = wait_for_completion_killable(&dreq->completion); 216 217 if (!result) { 218 result = dreq->count; 219 WARN_ON_ONCE(dreq->count < 0); 220 } 221 if (!result) 222 result = dreq->error; 223 224 out: 225 return (ssize_t) result; 226 } 227 228 /* 229 * Synchronous I/O uses a stack-allocated iocb. Thus we can't trust 230 * the iocb is still valid here if this is a synchronous request. 231 */ 232 static void nfs_direct_complete(struct nfs_direct_req *dreq) 233 { 234 struct inode *inode = dreq->inode; 235 236 inode_dio_end(inode); 237 238 if (dreq->iocb) { 239 long res = (long) dreq->error; 240 if (dreq->count != 0) { 241 res = (long) dreq->count; 242 WARN_ON_ONCE(dreq->count < 0); 243 } 244 dreq->iocb->ki_complete(dreq->iocb, res); 245 } 246 247 complete(&dreq->completion); 248 249 nfs_direct_req_release(dreq); 250 } 251 252 static void nfs_direct_read_completion(struct nfs_pgio_header *hdr) 253 { 254 unsigned long bytes = 0; 255 struct nfs_direct_req *dreq = hdr->dreq; 256 257 spin_lock(&dreq->lock); 258 if (test_bit(NFS_IOHDR_REDO, &hdr->flags)) { 259 spin_unlock(&dreq->lock); 260 goto out_put; 261 } 262 263 nfs_direct_count_bytes(dreq, hdr); 264 spin_unlock(&dreq->lock); 265 266 while (!list_empty(&hdr->pages)) { 267 struct nfs_page *req = nfs_list_entry(hdr->pages.next); 268 struct page *page = req->wb_page; 269 270 if (!PageCompound(page) && bytes < hdr->good_bytes && 271 (dreq->flags == NFS_ODIRECT_SHOULD_DIRTY)) 272 set_page_dirty(page); 273 bytes += req->wb_bytes; 274 nfs_list_remove_request(req); 275 nfs_release_request(req); 276 } 277 out_put: 278 if (put_dreq(dreq)) 279 nfs_direct_complete(dreq); 280 hdr->release(hdr); 281 } 282 283 static void nfs_read_sync_pgio_error(struct list_head *head, int error) 284 { 285 struct nfs_page *req; 286 287 while (!list_empty(head)) { 288 req = nfs_list_entry(head->next); 289 nfs_list_remove_request(req); 290 nfs_release_request(req); 291 } 292 } 293 294 static void nfs_direct_pgio_init(struct nfs_pgio_header *hdr) 295 { 296 get_dreq(hdr->dreq); 297 } 298 299 static const struct nfs_pgio_completion_ops nfs_direct_read_completion_ops = { 300 .error_cleanup = nfs_read_sync_pgio_error, 301 .init_hdr = nfs_direct_pgio_init, 302 .completion = nfs_direct_read_completion, 303 }; 304 305 /* 306 * For each rsize'd chunk of the user's buffer, dispatch an NFS READ 307 * operation. If nfs_readdata_alloc() or get_user_pages() fails, 308 * bail and stop sending more reads. Read length accounting is 309 * handled automatically by nfs_direct_read_result(). Otherwise, if 310 * no requests have been sent, just return an error. 311 */ 312 313 static ssize_t nfs_direct_read_schedule_iovec(struct nfs_direct_req *dreq, 314 struct iov_iter *iter, 315 loff_t pos) 316 { 317 struct nfs_pageio_descriptor desc; 318 struct inode *inode = dreq->inode; 319 ssize_t result = -EINVAL; 320 size_t requested_bytes = 0; 321 size_t rsize = max_t(size_t, NFS_SERVER(inode)->rsize, PAGE_SIZE); 322 323 nfs_pageio_init_read(&desc, dreq->inode, false, 324 &nfs_direct_read_completion_ops); 325 get_dreq(dreq); 326 desc.pg_dreq = dreq; 327 inode_dio_begin(inode); 328 329 while (iov_iter_count(iter)) { 330 struct page **pagevec; 331 size_t bytes; 332 size_t pgbase; 333 unsigned npages, i; 334 335 result = iov_iter_get_pages_alloc2(iter, &pagevec, 336 rsize, &pgbase); 337 if (result < 0) 338 break; 339 340 bytes = result; 341 npages = (result + pgbase + PAGE_SIZE - 1) / PAGE_SIZE; 342 for (i = 0; i < npages; i++) { 343 struct nfs_page *req; 344 unsigned int req_len = min_t(size_t, bytes, PAGE_SIZE - pgbase); 345 /* XXX do we need to do the eof zeroing found in async_filler? */ 346 req = nfs_create_request(dreq->ctx, pagevec[i], 347 pgbase, req_len); 348 if (IS_ERR(req)) { 349 result = PTR_ERR(req); 350 break; 351 } 352 req->wb_index = pos >> PAGE_SHIFT; 353 req->wb_offset = pos & ~PAGE_MASK; 354 if (!nfs_pageio_add_request(&desc, req)) { 355 result = desc.pg_error; 356 nfs_release_request(req); 357 break; 358 } 359 pgbase = 0; 360 bytes -= req_len; 361 requested_bytes += req_len; 362 pos += req_len; 363 dreq->bytes_left -= req_len; 364 } 365 nfs_direct_release_pages(pagevec, npages); 366 kvfree(pagevec); 367 if (result < 0) 368 break; 369 } 370 371 nfs_pageio_complete(&desc); 372 373 /* 374 * If no bytes were started, return the error, and let the 375 * generic layer handle the completion. 376 */ 377 if (requested_bytes == 0) { 378 inode_dio_end(inode); 379 nfs_direct_req_release(dreq); 380 return result < 0 ? result : -EIO; 381 } 382 383 if (put_dreq(dreq)) 384 nfs_direct_complete(dreq); 385 return requested_bytes; 386 } 387 388 /** 389 * nfs_file_direct_read - file direct read operation for NFS files 390 * @iocb: target I/O control block 391 * @iter: vector of user buffers into which to read data 392 * @swap: flag indicating this is swap IO, not O_DIRECT IO 393 * 394 * We use this function for direct reads instead of calling 395 * generic_file_aio_read() in order to avoid gfar's check to see if 396 * the request starts before the end of the file. For that check 397 * to work, we must generate a GETATTR before each direct read, and 398 * even then there is a window between the GETATTR and the subsequent 399 * READ where the file size could change. Our preference is simply 400 * to do all reads the application wants, and the server will take 401 * care of managing the end of file boundary. 402 * 403 * This function also eliminates unnecessarily updating the file's 404 * atime locally, as the NFS server sets the file's atime, and this 405 * client must read the updated atime from the server back into its 406 * cache. 407 */ 408 ssize_t nfs_file_direct_read(struct kiocb *iocb, struct iov_iter *iter, 409 bool swap) 410 { 411 struct file *file = iocb->ki_filp; 412 struct address_space *mapping = file->f_mapping; 413 struct inode *inode = mapping->host; 414 struct nfs_direct_req *dreq; 415 struct nfs_lock_context *l_ctx; 416 ssize_t result, requested; 417 size_t count = iov_iter_count(iter); 418 nfs_add_stats(mapping->host, NFSIOS_DIRECTREADBYTES, count); 419 420 dfprintk(FILE, "NFS: direct read(%pD2, %zd@%Ld)\n", 421 file, count, (long long) iocb->ki_pos); 422 423 result = 0; 424 if (!count) 425 goto out; 426 427 task_io_account_read(count); 428 429 result = -ENOMEM; 430 dreq = nfs_direct_req_alloc(); 431 if (dreq == NULL) 432 goto out; 433 434 dreq->inode = inode; 435 dreq->bytes_left = dreq->max_count = count; 436 dreq->io_start = iocb->ki_pos; 437 dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp)); 438 l_ctx = nfs_get_lock_context(dreq->ctx); 439 if (IS_ERR(l_ctx)) { 440 result = PTR_ERR(l_ctx); 441 nfs_direct_req_release(dreq); 442 goto out_release; 443 } 444 dreq->l_ctx = l_ctx; 445 if (!is_sync_kiocb(iocb)) 446 dreq->iocb = iocb; 447 448 if (user_backed_iter(iter)) 449 dreq->flags = NFS_ODIRECT_SHOULD_DIRTY; 450 451 if (!swap) 452 nfs_start_io_direct(inode); 453 454 NFS_I(inode)->read_io += count; 455 requested = nfs_direct_read_schedule_iovec(dreq, iter, iocb->ki_pos); 456 457 if (!swap) 458 nfs_end_io_direct(inode); 459 460 if (requested > 0) { 461 result = nfs_direct_wait(dreq); 462 if (result > 0) { 463 requested -= result; 464 iocb->ki_pos += result; 465 } 466 iov_iter_revert(iter, requested); 467 } else { 468 result = requested; 469 } 470 471 out_release: 472 nfs_direct_req_release(dreq); 473 out: 474 return result; 475 } 476 477 static void 478 nfs_direct_join_group(struct list_head *list, struct inode *inode) 479 { 480 struct nfs_page *req, *next; 481 482 list_for_each_entry(req, list, wb_list) { 483 if (req->wb_head != req || req->wb_this_page == req) 484 continue; 485 for (next = req->wb_this_page; 486 next != req->wb_head; 487 next = next->wb_this_page) { 488 nfs_list_remove_request(next); 489 nfs_release_request(next); 490 } 491 nfs_join_page_group(req, inode); 492 } 493 } 494 495 static void 496 nfs_direct_write_scan_commit_list(struct inode *inode, 497 struct list_head *list, 498 struct nfs_commit_info *cinfo) 499 { 500 mutex_lock(&NFS_I(cinfo->inode)->commit_mutex); 501 pnfs_recover_commit_reqs(list, cinfo); 502 nfs_scan_commit_list(&cinfo->mds->list, list, cinfo, 0); 503 mutex_unlock(&NFS_I(cinfo->inode)->commit_mutex); 504 } 505 506 static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq) 507 { 508 struct nfs_pageio_descriptor desc; 509 struct nfs_page *req, *tmp; 510 LIST_HEAD(reqs); 511 struct nfs_commit_info cinfo; 512 LIST_HEAD(failed); 513 514 nfs_init_cinfo_from_dreq(&cinfo, dreq); 515 nfs_direct_write_scan_commit_list(dreq->inode, &reqs, &cinfo); 516 517 nfs_direct_join_group(&reqs, dreq->inode); 518 519 dreq->count = 0; 520 dreq->max_count = 0; 521 list_for_each_entry(req, &reqs, wb_list) 522 dreq->max_count += req->wb_bytes; 523 nfs_clear_pnfs_ds_commit_verifiers(&dreq->ds_cinfo); 524 get_dreq(dreq); 525 526 nfs_pageio_init_write(&desc, dreq->inode, FLUSH_STABLE, false, 527 &nfs_direct_write_completion_ops); 528 desc.pg_dreq = dreq; 529 530 list_for_each_entry_safe(req, tmp, &reqs, wb_list) { 531 /* Bump the transmission count */ 532 req->wb_nio++; 533 if (!nfs_pageio_add_request(&desc, req)) { 534 nfs_list_move_request(req, &failed); 535 spin_lock(&cinfo.inode->i_lock); 536 dreq->flags = 0; 537 if (desc.pg_error < 0) 538 dreq->error = desc.pg_error; 539 else 540 dreq->error = -EIO; 541 spin_unlock(&cinfo.inode->i_lock); 542 } 543 nfs_release_request(req); 544 } 545 nfs_pageio_complete(&desc); 546 547 while (!list_empty(&failed)) { 548 req = nfs_list_entry(failed.next); 549 nfs_list_remove_request(req); 550 nfs_unlock_and_release_request(req); 551 } 552 553 if (put_dreq(dreq)) 554 nfs_direct_write_complete(dreq); 555 } 556 557 static void nfs_direct_commit_complete(struct nfs_commit_data *data) 558 { 559 const struct nfs_writeverf *verf = data->res.verf; 560 struct nfs_direct_req *dreq = data->dreq; 561 struct nfs_commit_info cinfo; 562 struct nfs_page *req; 563 int status = data->task.tk_status; 564 565 trace_nfs_direct_commit_complete(dreq); 566 567 if (status < 0) { 568 /* Errors in commit are fatal */ 569 dreq->error = status; 570 dreq->max_count = 0; 571 dreq->count = 0; 572 dreq->flags = NFS_ODIRECT_DONE; 573 } else { 574 status = dreq->error; 575 } 576 577 nfs_init_cinfo_from_dreq(&cinfo, dreq); 578 579 while (!list_empty(&data->pages)) { 580 req = nfs_list_entry(data->pages.next); 581 nfs_list_remove_request(req); 582 if (status >= 0 && !nfs_write_match_verf(verf, req)) { 583 dreq->flags = NFS_ODIRECT_RESCHED_WRITES; 584 /* 585 * Despite the reboot, the write was successful, 586 * so reset wb_nio. 587 */ 588 req->wb_nio = 0; 589 nfs_mark_request_commit(req, NULL, &cinfo, 0); 590 } else /* Error or match */ 591 nfs_release_request(req); 592 nfs_unlock_and_release_request(req); 593 } 594 595 if (nfs_commit_end(cinfo.mds)) 596 nfs_direct_write_complete(dreq); 597 } 598 599 static void nfs_direct_resched_write(struct nfs_commit_info *cinfo, 600 struct nfs_page *req) 601 { 602 struct nfs_direct_req *dreq = cinfo->dreq; 603 604 trace_nfs_direct_resched_write(dreq); 605 606 spin_lock(&dreq->lock); 607 if (dreq->flags != NFS_ODIRECT_DONE) 608 dreq->flags = NFS_ODIRECT_RESCHED_WRITES; 609 spin_unlock(&dreq->lock); 610 nfs_mark_request_commit(req, NULL, cinfo, 0); 611 } 612 613 static const struct nfs_commit_completion_ops nfs_direct_commit_completion_ops = { 614 .completion = nfs_direct_commit_complete, 615 .resched_write = nfs_direct_resched_write, 616 }; 617 618 static void nfs_direct_commit_schedule(struct nfs_direct_req *dreq) 619 { 620 int res; 621 struct nfs_commit_info cinfo; 622 LIST_HEAD(mds_list); 623 624 nfs_init_cinfo_from_dreq(&cinfo, dreq); 625 nfs_scan_commit(dreq->inode, &mds_list, &cinfo); 626 res = nfs_generic_commit_list(dreq->inode, &mds_list, 0, &cinfo); 627 if (res < 0) /* res == -ENOMEM */ 628 nfs_direct_write_reschedule(dreq); 629 } 630 631 static void nfs_direct_write_clear_reqs(struct nfs_direct_req *dreq) 632 { 633 struct nfs_commit_info cinfo; 634 struct nfs_page *req; 635 LIST_HEAD(reqs); 636 637 nfs_init_cinfo_from_dreq(&cinfo, dreq); 638 nfs_direct_write_scan_commit_list(dreq->inode, &reqs, &cinfo); 639 640 while (!list_empty(&reqs)) { 641 req = nfs_list_entry(reqs.next); 642 nfs_list_remove_request(req); 643 nfs_release_request(req); 644 nfs_unlock_and_release_request(req); 645 } 646 } 647 648 static void nfs_direct_write_schedule_work(struct work_struct *work) 649 { 650 struct nfs_direct_req *dreq = container_of(work, struct nfs_direct_req, work); 651 int flags = dreq->flags; 652 653 dreq->flags = 0; 654 switch (flags) { 655 case NFS_ODIRECT_DO_COMMIT: 656 nfs_direct_commit_schedule(dreq); 657 break; 658 case NFS_ODIRECT_RESCHED_WRITES: 659 nfs_direct_write_reschedule(dreq); 660 break; 661 default: 662 nfs_direct_write_clear_reqs(dreq); 663 nfs_zap_mapping(dreq->inode, dreq->inode->i_mapping); 664 nfs_direct_complete(dreq); 665 } 666 } 667 668 static void nfs_direct_write_complete(struct nfs_direct_req *dreq) 669 { 670 trace_nfs_direct_write_complete(dreq); 671 queue_work(nfsiod_workqueue, &dreq->work); /* Calls nfs_direct_write_schedule_work */ 672 } 673 674 static void nfs_direct_write_completion(struct nfs_pgio_header *hdr) 675 { 676 struct nfs_direct_req *dreq = hdr->dreq; 677 struct nfs_commit_info cinfo; 678 struct nfs_page *req = nfs_list_entry(hdr->pages.next); 679 int flags = NFS_ODIRECT_DONE; 680 681 trace_nfs_direct_write_completion(dreq); 682 683 nfs_init_cinfo_from_dreq(&cinfo, dreq); 684 685 spin_lock(&dreq->lock); 686 if (test_bit(NFS_IOHDR_REDO, &hdr->flags)) { 687 spin_unlock(&dreq->lock); 688 goto out_put; 689 } 690 691 nfs_direct_count_bytes(dreq, hdr); 692 if (test_bit(NFS_IOHDR_UNSTABLE_WRITES, &hdr->flags)) { 693 if (!dreq->flags) 694 dreq->flags = NFS_ODIRECT_DO_COMMIT; 695 flags = dreq->flags; 696 } 697 spin_unlock(&dreq->lock); 698 699 while (!list_empty(&hdr->pages)) { 700 701 req = nfs_list_entry(hdr->pages.next); 702 nfs_list_remove_request(req); 703 if (flags == NFS_ODIRECT_DO_COMMIT) { 704 kref_get(&req->wb_kref); 705 memcpy(&req->wb_verf, &hdr->verf.verifier, 706 sizeof(req->wb_verf)); 707 nfs_mark_request_commit(req, hdr->lseg, &cinfo, 708 hdr->ds_commit_idx); 709 } else if (flags == NFS_ODIRECT_RESCHED_WRITES) { 710 kref_get(&req->wb_kref); 711 nfs_mark_request_commit(req, NULL, &cinfo, 0); 712 } 713 nfs_unlock_and_release_request(req); 714 } 715 716 out_put: 717 if (put_dreq(dreq)) 718 nfs_direct_write_complete(dreq); 719 hdr->release(hdr); 720 } 721 722 static void nfs_write_sync_pgio_error(struct list_head *head, int error) 723 { 724 struct nfs_page *req; 725 726 while (!list_empty(head)) { 727 req = nfs_list_entry(head->next); 728 nfs_list_remove_request(req); 729 nfs_unlock_and_release_request(req); 730 } 731 } 732 733 static void nfs_direct_write_reschedule_io(struct nfs_pgio_header *hdr) 734 { 735 struct nfs_direct_req *dreq = hdr->dreq; 736 737 trace_nfs_direct_write_reschedule_io(dreq); 738 739 spin_lock(&dreq->lock); 740 if (dreq->error == 0) { 741 dreq->flags = NFS_ODIRECT_RESCHED_WRITES; 742 /* fake unstable write to let common nfs resend pages */ 743 hdr->verf.committed = NFS_UNSTABLE; 744 hdr->good_bytes = hdr->args.offset + hdr->args.count - 745 hdr->io_start; 746 } 747 spin_unlock(&dreq->lock); 748 } 749 750 static const struct nfs_pgio_completion_ops nfs_direct_write_completion_ops = { 751 .error_cleanup = nfs_write_sync_pgio_error, 752 .init_hdr = nfs_direct_pgio_init, 753 .completion = nfs_direct_write_completion, 754 .reschedule_io = nfs_direct_write_reschedule_io, 755 }; 756 757 758 /* 759 * NB: Return the value of the first error return code. Subsequent 760 * errors after the first one are ignored. 761 */ 762 /* 763 * For each wsize'd chunk of the user's buffer, dispatch an NFS WRITE 764 * operation. If nfs_writedata_alloc() or get_user_pages() fails, 765 * bail and stop sending more writes. Write length accounting is 766 * handled automatically by nfs_direct_write_result(). Otherwise, if 767 * no requests have been sent, just return an error. 768 */ 769 static ssize_t nfs_direct_write_schedule_iovec(struct nfs_direct_req *dreq, 770 struct iov_iter *iter, 771 loff_t pos, int ioflags) 772 { 773 struct nfs_pageio_descriptor desc; 774 struct inode *inode = dreq->inode; 775 ssize_t result = 0; 776 size_t requested_bytes = 0; 777 size_t wsize = max_t(size_t, NFS_SERVER(inode)->wsize, PAGE_SIZE); 778 779 trace_nfs_direct_write_schedule_iovec(dreq); 780 781 nfs_pageio_init_write(&desc, inode, ioflags, false, 782 &nfs_direct_write_completion_ops); 783 desc.pg_dreq = dreq; 784 get_dreq(dreq); 785 inode_dio_begin(inode); 786 787 NFS_I(inode)->write_io += iov_iter_count(iter); 788 while (iov_iter_count(iter)) { 789 struct page **pagevec; 790 size_t bytes; 791 size_t pgbase; 792 unsigned npages, i; 793 794 result = iov_iter_get_pages_alloc2(iter, &pagevec, 795 wsize, &pgbase); 796 if (result < 0) 797 break; 798 799 bytes = result; 800 npages = (result + pgbase + PAGE_SIZE - 1) / PAGE_SIZE; 801 for (i = 0; i < npages; i++) { 802 struct nfs_page *req; 803 unsigned int req_len = min_t(size_t, bytes, PAGE_SIZE - pgbase); 804 805 req = nfs_create_request(dreq->ctx, pagevec[i], 806 pgbase, req_len); 807 if (IS_ERR(req)) { 808 result = PTR_ERR(req); 809 break; 810 } 811 812 if (desc.pg_error < 0) { 813 nfs_free_request(req); 814 result = desc.pg_error; 815 break; 816 } 817 818 nfs_lock_request(req); 819 req->wb_index = pos >> PAGE_SHIFT; 820 req->wb_offset = pos & ~PAGE_MASK; 821 if (!nfs_pageio_add_request(&desc, req)) { 822 result = desc.pg_error; 823 nfs_unlock_and_release_request(req); 824 break; 825 } 826 pgbase = 0; 827 bytes -= req_len; 828 requested_bytes += req_len; 829 pos += req_len; 830 dreq->bytes_left -= req_len; 831 } 832 nfs_direct_release_pages(pagevec, npages); 833 kvfree(pagevec); 834 if (result < 0) 835 break; 836 } 837 nfs_pageio_complete(&desc); 838 839 /* 840 * If no bytes were started, return the error, and let the 841 * generic layer handle the completion. 842 */ 843 if (requested_bytes == 0) { 844 inode_dio_end(inode); 845 nfs_direct_req_release(dreq); 846 return result < 0 ? result : -EIO; 847 } 848 849 if (put_dreq(dreq)) 850 nfs_direct_write_complete(dreq); 851 return requested_bytes; 852 } 853 854 /** 855 * nfs_file_direct_write - file direct write operation for NFS files 856 * @iocb: target I/O control block 857 * @iter: vector of user buffers from which to write data 858 * @swap: flag indicating this is swap IO, not O_DIRECT IO 859 * 860 * We use this function for direct writes instead of calling 861 * generic_file_aio_write() in order to avoid taking the inode 862 * semaphore and updating the i_size. The NFS server will set 863 * the new i_size and this client must read the updated size 864 * back into its cache. We let the server do generic write 865 * parameter checking and report problems. 866 * 867 * We eliminate local atime updates, see direct read above. 868 * 869 * We avoid unnecessary page cache invalidations for normal cached 870 * readers of this file. 871 * 872 * Note that O_APPEND is not supported for NFS direct writes, as there 873 * is no atomic O_APPEND write facility in the NFS protocol. 874 */ 875 ssize_t nfs_file_direct_write(struct kiocb *iocb, struct iov_iter *iter, 876 bool swap) 877 { 878 ssize_t result, requested; 879 size_t count; 880 struct file *file = iocb->ki_filp; 881 struct address_space *mapping = file->f_mapping; 882 struct inode *inode = mapping->host; 883 struct nfs_direct_req *dreq; 884 struct nfs_lock_context *l_ctx; 885 loff_t pos, end; 886 887 dfprintk(FILE, "NFS: direct write(%pD2, %zd@%Ld)\n", 888 file, iov_iter_count(iter), (long long) iocb->ki_pos); 889 890 if (swap) 891 /* bypass generic checks */ 892 result = iov_iter_count(iter); 893 else 894 result = generic_write_checks(iocb, iter); 895 if (result <= 0) 896 return result; 897 count = result; 898 nfs_add_stats(mapping->host, NFSIOS_DIRECTWRITTENBYTES, count); 899 900 pos = iocb->ki_pos; 901 end = (pos + iov_iter_count(iter) - 1) >> PAGE_SHIFT; 902 903 task_io_account_write(count); 904 905 result = -ENOMEM; 906 dreq = nfs_direct_req_alloc(); 907 if (!dreq) 908 goto out; 909 910 dreq->inode = inode; 911 dreq->bytes_left = dreq->max_count = count; 912 dreq->io_start = pos; 913 dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp)); 914 l_ctx = nfs_get_lock_context(dreq->ctx); 915 if (IS_ERR(l_ctx)) { 916 result = PTR_ERR(l_ctx); 917 nfs_direct_req_release(dreq); 918 goto out_release; 919 } 920 dreq->l_ctx = l_ctx; 921 if (!is_sync_kiocb(iocb)) 922 dreq->iocb = iocb; 923 pnfs_init_ds_commit_info_ops(&dreq->ds_cinfo, inode); 924 925 if (swap) { 926 requested = nfs_direct_write_schedule_iovec(dreq, iter, pos, 927 FLUSH_STABLE); 928 } else { 929 nfs_start_io_direct(inode); 930 931 requested = nfs_direct_write_schedule_iovec(dreq, iter, pos, 932 FLUSH_COND_STABLE); 933 934 if (mapping->nrpages) { 935 invalidate_inode_pages2_range(mapping, 936 pos >> PAGE_SHIFT, end); 937 } 938 939 nfs_end_io_direct(inode); 940 } 941 942 if (requested > 0) { 943 result = nfs_direct_wait(dreq); 944 if (result > 0) { 945 requested -= result; 946 iocb->ki_pos = pos + result; 947 /* XXX: should check the generic_write_sync retval */ 948 generic_write_sync(iocb, result); 949 } 950 iov_iter_revert(iter, requested); 951 } else { 952 result = requested; 953 } 954 nfs_fscache_invalidate(inode, FSCACHE_INVAL_DIO_WRITE); 955 out_release: 956 nfs_direct_req_release(dreq); 957 out: 958 return result; 959 } 960 961 /** 962 * nfs_init_directcache - create a slab cache for nfs_direct_req structures 963 * 964 */ 965 int __init nfs_init_directcache(void) 966 { 967 nfs_direct_cachep = kmem_cache_create("nfs_direct_cache", 968 sizeof(struct nfs_direct_req), 969 0, (SLAB_RECLAIM_ACCOUNT| 970 SLAB_MEM_SPREAD), 971 NULL); 972 if (nfs_direct_cachep == NULL) 973 return -ENOMEM; 974 975 return 0; 976 } 977 978 /** 979 * nfs_destroy_directcache - destroy the slab cache for nfs_direct_req structures 980 * 981 */ 982 void nfs_destroy_directcache(void) 983 { 984 kmem_cache_destroy(nfs_direct_cachep); 985 } 986