fs/nfs/direct.c

   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
  98         if (test_bit(NFS_IOHDR_ERROR, &hdr->flags) && !dreq->error)
  99                 dreq->error = hdr->error;
 100 }
 101
 102 static void
 103 nfs_direct_count_bytes(struct nfs_direct_req *dreq,
 104                        const struct nfs_pgio_header *hdr)
 105 {
 106         loff_t hdr_end = hdr->io_start + hdr->good_bytes;
 107         ssize_t dreq_len = 0;
 108
 109         if (hdr_end > dreq->io_start)
 110                 dreq_len = hdr_end - dreq->io_start;
 111
 112         nfs_direct_handle_truncated(dreq, hdr, dreq_len);
 113
 114         if (dreq_len > dreq->max_count)
 115                 dreq_len = dreq->max_count;
 116
 117         if (dreq->count < dreq_len)
 118                 dreq->count = dreq_len;
 119 }
 120
 121 static void nfs_direct_truncate_request(struct nfs_direct_req *dreq,
 122                                         struct nfs_page *req)
 123 {
 124         loff_t offs = req_offset(req);
 125         size_t req_start = (size_t)(offs - dreq->io_start);
 126
 127         if (req_start < dreq->max_count)
 128                 dreq->max_count = req_start;
 129         if (req_start < dreq->count)
 130                 dreq->count = req_start;
 131 }
 132
 133 /**
 134  * nfs_swap_rw - NFS address space operation for swap I/O
 135  * @iocb: target I/O control block
 136  * @iter: I/O buffer
 137  *
 138  * Perform IO to the swap-file.  This is much like direct IO.
 139  */
 140 int nfs_swap_rw(struct kiocb *iocb, struct iov_iter *iter)
 141 {
 142         ssize_t ret;
 143
 144         VM_BUG_ON(iov_iter_count(iter) != PAGE_SIZE);
 145
 146         if (iov_iter_rw(iter) == READ)
 147                 ret = nfs_file_direct_read(iocb, iter, true);
 148         else
 149                 ret = nfs_file_direct_write(iocb, iter, true);
 150         if (ret < 0)
 151                 return ret;
 152         return 0;
 153 }
 154
 155 static void nfs_direct_release_pages(struct page **pages, unsigned int npages)
 156 {
 157         unsigned int i;
 158         for (i = 0; i < npages; i++)
 159                 put_page(pages[i]);
 160 }
 161
 162 void nfs_init_cinfo_from_dreq(struct nfs_commit_info *cinfo,
 163                               struct nfs_direct_req *dreq)
 164 {
 165         cinfo->inode = dreq->inode;
 166         cinfo->mds = &dreq->mds_cinfo;
 167         cinfo->ds = &dreq->ds_cinfo;
 168         cinfo->dreq = dreq;
 169         cinfo->completion_ops = &nfs_direct_commit_completion_ops;
 170 }
 171
 172 static inline struct nfs_direct_req *nfs_direct_req_alloc(void)
 173 {
 174         struct nfs_direct_req *dreq;
 175
 176         dreq = kmem_cache_zalloc(nfs_direct_cachep, GFP_KERNEL);
 177         if (!dreq)
 178                 return NULL;
 179
 180         kref_init(&dreq->kref);
 181         kref_get(&dreq->kref);
 182         init_completion(&dreq->completion);
 183         INIT_LIST_HEAD(&dreq->mds_cinfo.list);
 184         pnfs_init_ds_commit_info(&dreq->ds_cinfo);
 185         INIT_WORK(&dreq->work, nfs_direct_write_schedule_work);
 186         spin_lock_init(&dreq->lock);
 187
 188         return dreq;
 189 }
 190
 191 static void nfs_direct_req_free(struct kref *kref)
 192 {
 193         struct nfs_direct_req *dreq = container_of(kref, struct nfs_direct_req, kref);
 194
 195         pnfs_release_ds_info(&dreq->ds_cinfo, dreq->inode);
 196         if (dreq->l_ctx != NULL)
 197                 nfs_put_lock_context(dreq->l_ctx);
 198         if (dreq->ctx != NULL)
 199                 put_nfs_open_context(dreq->ctx);
 200         kmem_cache_free(nfs_direct_cachep, dreq);
 201 }
 202
 203 static void nfs_direct_req_release(struct nfs_direct_req *dreq)
 204 {
 205         kref_put(&dreq->kref, nfs_direct_req_free);
 206 }
 207
 208 ssize_t nfs_dreq_bytes_left(struct nfs_direct_req *dreq, loff_t offset)
 209 {
 210         loff_t start = offset - dreq->io_start;
 211         return dreq->max_count - start;
 212 }
 213 EXPORT_SYMBOL_GPL(nfs_dreq_bytes_left);
 214
 215 /*
 216  * Collects and returns the final error value/byte-count.
 217  */
 218 static ssize_t nfs_direct_wait(struct nfs_direct_req *dreq)
 219 {
 220         ssize_t result = -EIOCBQUEUED;
 221
 222         /* Async requests don't wait here */
 223         if (dreq->iocb)
 224                 goto out;
 225
 226         result = wait_for_completion_killable(&dreq->completion);
 227
 228         if (!result) {
 229                 result = dreq->count;
 230                 WARN_ON_ONCE(dreq->count < 0);
 231         }
 232         if (!result)
 233                 result = dreq->error;
 234
 235 out:
 236         return (ssize_t) result;
 237 }
 238
 239 /*
 240  * Synchronous I/O uses a stack-allocated iocb.  Thus we can't trust
 241  * the iocb is still valid here if this is a synchronous request.
 242  */
 243 static void nfs_direct_complete(struct nfs_direct_req *dreq)
 244 {
 245         struct inode *inode = dreq->inode;
 246
 247         inode_dio_end(inode);
 248
 249         if (dreq->iocb) {
 250                 long res = (long) dreq->error;
 251                 if (dreq->count != 0) {
 252                         res = (long) dreq->count;
 253                         WARN_ON_ONCE(dreq->count < 0);
 254                 }
 255                 dreq->iocb->ki_complete(dreq->iocb, res);
 256         }
 257
 258         complete(&dreq->completion);
 259
 260         nfs_direct_req_release(dreq);
 261 }
 262
 263 static void nfs_direct_read_completion(struct nfs_pgio_header *hdr)
 264 {
 265         unsigned long bytes = 0;
 266         struct nfs_direct_req *dreq = hdr->dreq;
 267
 268         spin_lock(&dreq->lock);
 269         if (test_bit(NFS_IOHDR_REDO, &hdr->flags)) {
 270                 spin_unlock(&dreq->lock);
 271                 goto out_put;
 272         }
 273
 274         nfs_direct_count_bytes(dreq, hdr);
 275         spin_unlock(&dreq->lock);
 276
 277         while (!list_empty(&hdr->pages)) {
 278                 struct nfs_page *req = nfs_list_entry(hdr->pages.next);
 279                 struct page *page = req->wb_page;
 280
 281                 if (!PageCompound(page) && bytes < hdr->good_bytes &&
 282                     (dreq->flags == NFS_ODIRECT_SHOULD_DIRTY))
 283                         set_page_dirty(page);
 284                 bytes += req->wb_bytes;
 285                 nfs_list_remove_request(req);
 286                 nfs_release_request(req);
 287         }
 288 out_put:
 289         if (put_dreq(dreq))
 290                 nfs_direct_complete(dreq);
 291         hdr->release(hdr);
 292 }
 293
 294 static void nfs_read_sync_pgio_error(struct list_head *head, int error)
 295 {
 296         struct nfs_page *req;
 297
 298         while (!list_empty(head)) {
 299                 req = nfs_list_entry(head->next);
 300                 nfs_list_remove_request(req);
 301                 nfs_release_request(req);
 302         }
 303 }
 304
 305 static void nfs_direct_pgio_init(struct nfs_pgio_header *hdr)
 306 {
 307         get_dreq(hdr->dreq);
 308 }
 309
 310 static const struct nfs_pgio_completion_ops nfs_direct_read_completion_ops = {
 311         .error_cleanup = nfs_read_sync_pgio_error,
 312         .init_hdr = nfs_direct_pgio_init,
 313         .completion = nfs_direct_read_completion,
 314 };
 315
 316 /*
 317  * For each rsize'd chunk of the user's buffer, dispatch an NFS READ
 318  * operation.  If nfs_readdata_alloc() or get_user_pages() fails,
 319  * bail and stop sending more reads.  Read length accounting is
 320  * handled automatically by nfs_direct_read_result().  Otherwise, if
 321  * no requests have been sent, just return an error.
 322  */
 323
 324 static ssize_t nfs_direct_read_schedule_iovec(struct nfs_direct_req *dreq,
 325                                               struct iov_iter *iter,
 326                                               loff_t pos)
 327 {
 328         struct nfs_pageio_descriptor desc;
 329         struct inode *inode = dreq->inode;
 330         ssize_t result = -EINVAL;
 331         size_t requested_bytes = 0;
 332         size_t rsize = max_t(size_t, NFS_SERVER(inode)->rsize, PAGE_SIZE);
 333
 334         nfs_pageio_init_read(&desc, dreq->inode, false,
 335                              &nfs_direct_read_completion_ops);
 336         get_dreq(dreq);
 337         desc.pg_dreq = dreq;
 338         inode_dio_begin(inode);
 339
 340         while (iov_iter_count(iter)) {
 341                 struct page **pagevec;
 342                 size_t bytes;
 343                 size_t pgbase;
 344                 unsigned npages, i;
 345
 346                 result = iov_iter_get_pages_alloc2(iter, &pagevec,
 347                                                   rsize, &pgbase);
 348                 if (result < 0)
 349                         break;
 350
 351                 bytes = result;
 352                 npages = (result + pgbase + PAGE_SIZE - 1) / PAGE_SIZE;
 353                 for (i = 0; i < npages; i++) {
 354                         struct nfs_page *req;
 355                         unsigned int req_len = min_t(size_t, bytes, PAGE_SIZE - pgbase);
 356                         /* XXX do we need to do the eof zeroing found in async_filler? */
 357                         req = nfs_page_create_from_page(dreq->ctx, pagevec[i],
 358                                                         pgbase, pos, req_len);
 359                         if (IS_ERR(req)) {
 360                                 result = PTR_ERR(req);
 361                                 break;
 362                         }
 363                         if (!nfs_pageio_add_request(&desc, req)) {
 364                                 result = desc.pg_error;
 365                                 nfs_release_request(req);
 366                                 break;
 367                         }
 368                         pgbase = 0;
 369                         bytes -= req_len;
 370                         requested_bytes += req_len;
 371                         pos += req_len;
 372                         dreq->bytes_left -= req_len;
 373                 }
 374                 nfs_direct_release_pages(pagevec, npages);
 375                 kvfree(pagevec);
 376                 if (result < 0)
 377                         break;
 378         }
 379
 380         nfs_pageio_complete(&desc);
 381
 382         /*
 383          * If no bytes were started, return the error, and let the
 384          * generic layer handle the completion.
 385          */
 386         if (requested_bytes == 0) {
 387                 inode_dio_end(inode);
 388                 nfs_direct_req_release(dreq);
 389                 return result < 0 ? result : -EIO;
 390         }
 391
 392         if (put_dreq(dreq))
 393                 nfs_direct_complete(dreq);
 394         return requested_bytes;
 395 }
 396
 397 /**
 398  * nfs_file_direct_read - file direct read operation for NFS files
 399  * @iocb: target I/O control block
 400  * @iter: vector of user buffers into which to read data
 401  * @swap: flag indicating this is swap IO, not O_DIRECT IO
 402  *
 403  * We use this function for direct reads instead of calling
 404  * generic_file_aio_read() in order to avoid gfar's check to see if
 405  * the request starts before the end of the file.  For that check
 406  * to work, we must generate a GETATTR before each direct read, and
 407  * even then there is a window between the GETATTR and the subsequent
 408  * READ where the file size could change.  Our preference is simply
 409  * to do all reads the application wants, and the server will take
 410  * care of managing the end of file boundary.
 411  *
 412  * This function also eliminates unnecessarily updating the file's
 413  * atime locally, as the NFS server sets the file's atime, and this
 414  * client must read the updated atime from the server back into its
 415  * cache.
 416  */
 417 ssize_t nfs_file_direct_read(struct kiocb *iocb, struct iov_iter *iter,
 418                              bool swap)
 419 {
 420         struct file *file = iocb->ki_filp;
 421         struct address_space *mapping = file->f_mapping;
 422         struct inode *inode = mapping->host;
 423         struct nfs_direct_req *dreq;
 424         struct nfs_lock_context *l_ctx;
 425         ssize_t result, requested;
 426         size_t count = iov_iter_count(iter);
 427         nfs_add_stats(mapping->host, NFSIOS_DIRECTREADBYTES, count);
 428
 429         dfprintk(FILE, "NFS: direct read(%pD2, %zd@%Ld)\n",
 430                 file, count, (long long) iocb->ki_pos);
 431
 432         result = 0;
 433         if (!count)
 434                 goto out;
 435
 436         task_io_account_read(count);
 437
 438         result = -ENOMEM;
 439         dreq = nfs_direct_req_alloc();
 440         if (dreq == NULL)
 441                 goto out;
 442
 443         dreq->inode = inode;
 444         dreq->bytes_left = dreq->max_count = count;
 445         dreq->io_start = iocb->ki_pos;
 446         dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp));
 447         l_ctx = nfs_get_lock_context(dreq->ctx);
 448         if (IS_ERR(l_ctx)) {
 449                 result = PTR_ERR(l_ctx);
 450                 nfs_direct_req_release(dreq);
 451                 goto out_release;
 452         }
 453         dreq->l_ctx = l_ctx;
 454         if (!is_sync_kiocb(iocb))
 455                 dreq->iocb = iocb;
 456
 457         if (user_backed_iter(iter))
 458                 dreq->flags = NFS_ODIRECT_SHOULD_DIRTY;
 459
 460         if (!swap)
 461                 nfs_start_io_direct(inode);
 462
 463         NFS_I(inode)->read_io += count;
 464         requested = nfs_direct_read_schedule_iovec(dreq, iter, iocb->ki_pos);
 465
 466         if (!swap)
 467                 nfs_end_io_direct(inode);
 468
 469         if (requested > 0) {
 470                 result = nfs_direct_wait(dreq);
 471                 if (result > 0) {
 472                         requested -= result;
 473                         iocb->ki_pos += result;
 474                 }
 475                 iov_iter_revert(iter, requested);
 476         } else {
 477                 result = requested;
 478         }
 479
 480 out_release:
 481         nfs_direct_req_release(dreq);
 482 out:
 483         return result;
 484 }
 485
 486 static void nfs_direct_add_page_head(struct list_head *list,
 487                                      struct nfs_page *req)
 488 {
 489         struct nfs_page *head = req->wb_head;
 490
 491         if (!list_empty(&head->wb_list) || !nfs_lock_request(head))
 492                 return;
 493         if (!list_empty(&head->wb_list)) {
 494                 nfs_unlock_request(head);
 495                 return;
 496         }
 497         list_add(&head->wb_list, list);
 498         kref_get(&head->wb_kref);
 499         kref_get(&head->wb_kref);
 500 }
 501
 502 static void nfs_direct_join_group(struct list_head *list,
 503                                   struct nfs_commit_info *cinfo,
 504                                   struct inode *inode)
 505 {
 506         struct nfs_page *req, *subreq;
 507
 508         list_for_each_entry(req, list, wb_list) {
 509                 if (req->wb_head != req) {
 510                         nfs_direct_add_page_head(&req->wb_list, req);
 511                         continue;
 512                 }
 513                 subreq = req->wb_this_page;
 514                 if (subreq == req)
 515                         continue;
 516                 do {
 517                         /*
 518                          * Remove subrequests from this list before freeing
 519                          * them in the call to nfs_join_page_group().
 520                          */
 521                         if (!list_empty(&subreq->wb_list)) {
 522                                 nfs_list_remove_request(subreq);
 523                                 nfs_release_request(subreq);
 524                         }
 525                 } while ((subreq = subreq->wb_this_page) != req);
 526                 nfs_join_page_group(req, cinfo, inode);
 527         }
 528 }
 529
 530 static void
 531 nfs_direct_write_scan_commit_list(struct inode *inode,
 532                                   struct list_head *list,
 533                                   struct nfs_commit_info *cinfo)
 534 {
 535         mutex_lock(&NFS_I(cinfo->inode)->commit_mutex);
 536         pnfs_recover_commit_reqs(list, cinfo);
 537         nfs_scan_commit_list(&cinfo->mds->list, list, cinfo, 0);
 538         mutex_unlock(&NFS_I(cinfo->inode)->commit_mutex);
 539 }
 540
 541 static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq)
 542 {
 543         struct nfs_pageio_descriptor desc;
 544         struct nfs_page *req;
 545         LIST_HEAD(reqs);
 546         struct nfs_commit_info cinfo;
 547
 548         nfs_init_cinfo_from_dreq(&cinfo, dreq);
 549         nfs_direct_write_scan_commit_list(dreq->inode, &reqs, &cinfo);
 550
 551         nfs_direct_join_group(&reqs, &cinfo, dreq->inode);
 552
 553         nfs_clear_pnfs_ds_commit_verifiers(&dreq->ds_cinfo);
 554         get_dreq(dreq);
 555
 556         nfs_pageio_init_write(&desc, dreq->inode, FLUSH_STABLE, false,
 557                               &nfs_direct_write_completion_ops);
 558         desc.pg_dreq = dreq;
 559
 560         while (!list_empty(&reqs)) {
 561                 req = nfs_list_entry(reqs.next);
 562                 /* Bump the transmission count */
 563                 req->wb_nio++;
 564                 if (!nfs_pageio_add_request(&desc, req)) {
 565                         spin_lock(&dreq->lock);
 566                         if (dreq->error < 0) {
 567                                 desc.pg_error = dreq->error;
 568                         } else if (desc.pg_error != -EAGAIN) {
 569                                 dreq->flags = 0;
 570                                 if (!desc.pg_error)
 571                                         desc.pg_error = -EIO;
 572                                 dreq->error = desc.pg_error;
 573                         } else
 574                                 dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
 575                         spin_unlock(&dreq->lock);
 576                         break;
 577                 }
 578                 nfs_release_request(req);
 579         }
 580         nfs_pageio_complete(&desc);
 581
 582         while (!list_empty(&reqs)) {
 583                 req = nfs_list_entry(reqs.next);
 584                 nfs_list_remove_request(req);
 585                 nfs_unlock_and_release_request(req);
 586                 if (desc.pg_error == -EAGAIN) {
 587                         nfs_mark_request_commit(req, NULL, &cinfo, 0);
 588                 } else {
 589                         spin_lock(&dreq->lock);
 590                         nfs_direct_truncate_request(dreq, req);
 591                         spin_unlock(&dreq->lock);
 592                         nfs_release_request(req);
 593                 }
 594         }
 595
 596         if (put_dreq(dreq))
 597                 nfs_direct_write_complete(dreq);
 598 }
 599
 600 static void nfs_direct_commit_complete(struct nfs_commit_data *data)
 601 {
 602         const struct nfs_writeverf *verf = data->res.verf;
 603         struct nfs_direct_req *dreq = data->dreq;
 604         struct nfs_commit_info cinfo;
 605         struct nfs_page *req;
 606         int status = data->task.tk_status;
 607
 608         trace_nfs_direct_commit_complete(dreq);
 609
 610         if (status < 0) {
 611                 /* Errors in commit are fatal */
 612                 dreq->error = status;
 613                 dreq->flags = NFS_ODIRECT_DONE;
 614         } else {
 615                 status = dreq->error;
 616         }
 617
 618         nfs_init_cinfo_from_dreq(&cinfo, dreq);
 619
 620         while (!list_empty(&data->pages)) {
 621                 req = nfs_list_entry(data->pages.next);
 622                 nfs_list_remove_request(req);
 623                 if (status < 0) {
 624                         spin_lock(&dreq->lock);
 625                         nfs_direct_truncate_request(dreq, req);
 626                         spin_unlock(&dreq->lock);
 627                         nfs_release_request(req);
 628                 } else if (!nfs_write_match_verf(verf, req)) {
 629                         dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
 630                         /*
 631                          * Despite the reboot, the write was successful,
 632                          * so reset wb_nio.
 633                          */
 634                         req->wb_nio = 0;
 635                         nfs_mark_request_commit(req, NULL, &cinfo, 0);
 636                 } else
 637                         nfs_release_request(req);
 638                 nfs_unlock_and_release_request(req);
 639         }
 640
 641         if (nfs_commit_end(cinfo.mds))
 642                 nfs_direct_write_complete(dreq);
 643 }
 644
 645 static void nfs_direct_resched_write(struct nfs_commit_info *cinfo,
 646                 struct nfs_page *req)
 647 {
 648         struct nfs_direct_req *dreq = cinfo->dreq;
 649
 650         trace_nfs_direct_resched_write(dreq);
 651
 652         spin_lock(&dreq->lock);
 653         if (dreq->flags != NFS_ODIRECT_DONE)
 654                 dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
 655         spin_unlock(&dreq->lock);
 656         nfs_mark_request_commit(req, NULL, cinfo, 0);
 657 }
 658
 659 static const struct nfs_commit_completion_ops nfs_direct_commit_completion_ops = {
 660         .completion = nfs_direct_commit_complete,
 661         .resched_write = nfs_direct_resched_write,
 662 };
 663
 664 static void nfs_direct_commit_schedule(struct nfs_direct_req *dreq)
 665 {
 666         int res;
 667         struct nfs_commit_info cinfo;
 668         LIST_HEAD(mds_list);
 669
 670         nfs_init_cinfo_from_dreq(&cinfo, dreq);
 671         nfs_scan_commit(dreq->inode, &mds_list, &cinfo);
 672         res = nfs_generic_commit_list(dreq->inode, &mds_list, 0, &cinfo);
 673         if (res < 0) /* res == -ENOMEM */
 674                 nfs_direct_write_reschedule(dreq);
 675 }
 676
 677 static void nfs_direct_write_clear_reqs(struct nfs_direct_req *dreq)
 678 {
 679         struct nfs_commit_info cinfo;
 680         struct nfs_page *req;
 681         LIST_HEAD(reqs);
 682
 683         nfs_init_cinfo_from_dreq(&cinfo, dreq);
 684         nfs_direct_write_scan_commit_list(dreq->inode, &reqs, &cinfo);
 685
 686         while (!list_empty(&reqs)) {
 687                 req = nfs_list_entry(reqs.next);
 688                 nfs_list_remove_request(req);
 689                 nfs_direct_truncate_request(dreq, req);
 690                 nfs_release_request(req);
 691                 nfs_unlock_and_release_request(req);
 692         }
 693 }
 694
 695 static void nfs_direct_write_schedule_work(struct work_struct *work)
 696 {
 697         struct nfs_direct_req *dreq = container_of(work, struct nfs_direct_req, work);
 698         int flags = dreq->flags;
 699
 700         dreq->flags = 0;
 701         switch (flags) {
 702                 case NFS_ODIRECT_DO_COMMIT:
 703                         nfs_direct_commit_schedule(dreq);
 704                         break;
 705                 case NFS_ODIRECT_RESCHED_WRITES:
 706                         nfs_direct_write_reschedule(dreq);
 707                         break;
 708                 default:
 709                         nfs_direct_write_clear_reqs(dreq);
 710                         nfs_zap_mapping(dreq->inode, dreq->inode->i_mapping);
 711                         nfs_direct_complete(dreq);
 712         }
 713 }
 714
 715 static void nfs_direct_write_complete(struct nfs_direct_req *dreq)
 716 {
 717         trace_nfs_direct_write_complete(dreq);
 718         queue_work(nfsiod_workqueue, &dreq->work); /* Calls nfs_direct_write_schedule_work */
 719 }
 720
 721 static void nfs_direct_write_completion(struct nfs_pgio_header *hdr)
 722 {
 723         struct nfs_direct_req *dreq = hdr->dreq;
 724         struct nfs_commit_info cinfo;
 725         struct nfs_page *req = nfs_list_entry(hdr->pages.next);
 726         int flags = NFS_ODIRECT_DONE;
 727
 728         trace_nfs_direct_write_completion(dreq);
 729
 730         nfs_init_cinfo_from_dreq(&cinfo, dreq);
 731
 732         spin_lock(&dreq->lock);
 733         if (test_bit(NFS_IOHDR_REDO, &hdr->flags)) {
 734                 spin_unlock(&dreq->lock);
 735                 goto out_put;
 736         }
 737
 738         nfs_direct_count_bytes(dreq, hdr);
 739         if (test_bit(NFS_IOHDR_UNSTABLE_WRITES, &hdr->flags) &&
 740             !test_bit(NFS_IOHDR_ERROR, &hdr->flags)) {
 741                 if (!dreq->flags)
 742                         dreq->flags = NFS_ODIRECT_DO_COMMIT;
 743                 flags = dreq->flags;
 744         }
 745         spin_unlock(&dreq->lock);
 746
 747         while (!list_empty(&hdr->pages)) {
 748
 749                 req = nfs_list_entry(hdr->pages.next);
 750                 nfs_list_remove_request(req);
 751                 if (flags == NFS_ODIRECT_DO_COMMIT) {
 752                         kref_get(&req->wb_kref);
 753                         memcpy(&req->wb_verf, &hdr->verf.verifier,
 754                                sizeof(req->wb_verf));
 755                         nfs_mark_request_commit(req, hdr->lseg, &cinfo,
 756                                 hdr->ds_commit_idx);
 757                 } else if (flags == NFS_ODIRECT_RESCHED_WRITES) {
 758                         kref_get(&req->wb_kref);
 759                         nfs_mark_request_commit(req, NULL, &cinfo, 0);
 760                 }
 761                 nfs_unlock_and_release_request(req);
 762         }
 763
 764 out_put:
 765         if (put_dreq(dreq))
 766                 nfs_direct_write_complete(dreq);
 767         hdr->release(hdr);
 768 }
 769
 770 static void nfs_write_sync_pgio_error(struct list_head *head, int error)
 771 {
 772         struct nfs_page *req;
 773
 774         while (!list_empty(head)) {
 775                 req = nfs_list_entry(head->next);
 776                 nfs_list_remove_request(req);
 777                 nfs_unlock_and_release_request(req);
 778         }
 779 }
 780
 781 static void nfs_direct_write_reschedule_io(struct nfs_pgio_header *hdr)
 782 {
 783         struct nfs_direct_req *dreq = hdr->dreq;
 784         struct nfs_page *req;
 785         struct nfs_commit_info cinfo;
 786
 787         trace_nfs_direct_write_reschedule_io(dreq);
 788
 789         nfs_init_cinfo_from_dreq(&cinfo, dreq);
 790         spin_lock(&dreq->lock);
 791         if (dreq->error == 0)
 792                 dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
 793         set_bit(NFS_IOHDR_REDO, &hdr->flags);
 794         spin_unlock(&dreq->lock);
 795         while (!list_empty(&hdr->pages)) {
 796                 req = nfs_list_entry(hdr->pages.next);
 797                 nfs_list_remove_request(req);
 798                 nfs_unlock_request(req);
 799                 nfs_mark_request_commit(req, NULL, &cinfo, 0);
 800         }
 801 }
 802
 803 static const struct nfs_pgio_completion_ops nfs_direct_write_completion_ops = {
 804         .error_cleanup = nfs_write_sync_pgio_error,
 805         .init_hdr = nfs_direct_pgio_init,
 806         .completion = nfs_direct_write_completion,
 807         .reschedule_io = nfs_direct_write_reschedule_io,
 808 };
 809
 810
 811 /*
 812  * NB: Return the value of the first error return code.  Subsequent
 813  *     errors after the first one are ignored.
 814  */
 815 /*
 816  * For each wsize'd chunk of the user's buffer, dispatch an NFS WRITE
 817  * operation.  If nfs_writedata_alloc() or get_user_pages() fails,
 818  * bail and stop sending more writes.  Write length accounting is
 819  * handled automatically by nfs_direct_write_result().  Otherwise, if
 820  * no requests have been sent, just return an error.
 821  */
 822 static ssize_t nfs_direct_write_schedule_iovec(struct nfs_direct_req *dreq,
 823                                                struct iov_iter *iter,
 824                                                loff_t pos, int ioflags)
 825 {
 826         struct nfs_pageio_descriptor desc;
 827         struct inode *inode = dreq->inode;
 828         struct nfs_commit_info cinfo;
 829         ssize_t result = 0;
 830         size_t requested_bytes = 0;
 831         size_t wsize = max_t(size_t, NFS_SERVER(inode)->wsize, PAGE_SIZE);
 832         bool defer = false;
 833
 834         trace_nfs_direct_write_schedule_iovec(dreq);
 835
 836         nfs_pageio_init_write(&desc, inode, ioflags, false,
 837                               &nfs_direct_write_completion_ops);
 838         desc.pg_dreq = dreq;
 839         get_dreq(dreq);
 840         inode_dio_begin(inode);
 841
 842         NFS_I(inode)->write_io += iov_iter_count(iter);
 843         while (iov_iter_count(iter)) {
 844                 struct page **pagevec;
 845                 size_t bytes;
 846                 size_t pgbase;
 847                 unsigned npages, i;
 848
 849                 result = iov_iter_get_pages_alloc2(iter, &pagevec,
 850                                                   wsize, &pgbase);
 851                 if (result < 0)
 852                         break;
 853
 854                 bytes = result;
 855                 npages = (result + pgbase + PAGE_SIZE - 1) / PAGE_SIZE;
 856                 for (i = 0; i < npages; i++) {
 857                         struct nfs_page *req;
 858                         unsigned int req_len = min_t(size_t, bytes, PAGE_SIZE - pgbase);
 859
 860                         req = nfs_page_create_from_page(dreq->ctx, pagevec[i],
 861                                                         pgbase, pos, req_len);
 862                         if (IS_ERR(req)) {
 863                                 result = PTR_ERR(req);
 864                                 break;
 865                         }
 866
 867                         if (desc.pg_error < 0) {
 868                                 nfs_free_request(req);
 869                                 result = desc.pg_error;
 870                                 break;
 871                         }
 872
 873                         pgbase = 0;
 874                         bytes -= req_len;
 875                         requested_bytes += req_len;
 876                         pos += req_len;
 877                         dreq->bytes_left -= req_len;
 878
 879                         if (defer) {
 880                                 nfs_mark_request_commit(req, NULL, &cinfo, 0);
 881                                 continue;
 882                         }
 883
 884                         nfs_lock_request(req);
 885                         if (nfs_pageio_add_request(&desc, req))
 886                                 continue;
 887
 888                         /* Exit on hard errors */
 889                         if (desc.pg_error < 0 && desc.pg_error != -EAGAIN) {
 890                                 result = desc.pg_error;
 891                                 nfs_unlock_and_release_request(req);
 892                                 break;
 893                         }
 894
 895                         /* If the error is soft, defer remaining requests */
 896                         nfs_init_cinfo_from_dreq(&cinfo, dreq);
 897                         spin_lock(&dreq->lock);
 898                         dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
 899                         spin_unlock(&dreq->lock);
 900                         nfs_unlock_request(req);
 901                         nfs_mark_request_commit(req, NULL, &cinfo, 0);
 902                         desc.pg_error = 0;
 903                         defer = true;
 904                 }
 905                 nfs_direct_release_pages(pagevec, npages);
 906                 kvfree(pagevec);
 907                 if (result < 0)
 908                         break;
 909         }
 910         nfs_pageio_complete(&desc);
 911
 912         /*
 913          * If no bytes were started, return the error, and let the
 914          * generic layer handle the completion.
 915          */
 916         if (requested_bytes == 0) {
 917                 inode_dio_end(inode);
 918                 nfs_direct_req_release(dreq);
 919                 return result < 0 ? result : -EIO;
 920         }
 921
 922         if (put_dreq(dreq))
 923                 nfs_direct_write_complete(dreq);
 924         return requested_bytes;
 925 }
 926
 927 /**
 928  * nfs_file_direct_write - file direct write operation for NFS files
 929  * @iocb: target I/O control block
 930  * @iter: vector of user buffers from which to write data
 931  * @swap: flag indicating this is swap IO, not O_DIRECT IO
 932  *
 933  * We use this function for direct writes instead of calling
 934  * generic_file_aio_write() in order to avoid taking the inode
 935  * semaphore and updating the i_size.  The NFS server will set
 936  * the new i_size and this client must read the updated size
 937  * back into its cache.  We let the server do generic write
 938  * parameter checking and report problems.
 939  *
 940  * We eliminate local atime updates, see direct read above.
 941  *
 942  * We avoid unnecessary page cache invalidations for normal cached
 943  * readers of this file.
 944  *
 945  * Note that O_APPEND is not supported for NFS direct writes, as there
 946  * is no atomic O_APPEND write facility in the NFS protocol.
 947  */
 948 ssize_t nfs_file_direct_write(struct kiocb *iocb, struct iov_iter *iter,
 949                               bool swap)
 950 {
 951         ssize_t result, requested;
 952         size_t count;
 953         struct file *file = iocb->ki_filp;
 954         struct address_space *mapping = file->f_mapping;
 955         struct inode *inode = mapping->host;
 956         struct nfs_direct_req *dreq;
 957         struct nfs_lock_context *l_ctx;
 958         loff_t pos, end;
 959
 960         dfprintk(FILE, "NFS: direct write(%pD2, %zd@%Ld)\n",
 961                 file, iov_iter_count(iter), (long long) iocb->ki_pos);
 962
 963         if (swap)
 964                 /* bypass generic checks */
 965                 result =  iov_iter_count(iter);
 966         else
 967                 result = generic_write_checks(iocb, iter);
 968         if (result <= 0)
 969                 return result;
 970         count = result;
 971         nfs_add_stats(mapping->host, NFSIOS_DIRECTWRITTENBYTES, count);
 972
 973         pos = iocb->ki_pos;
 974         end = (pos + iov_iter_count(iter) - 1) >> PAGE_SHIFT;
 975
 976         task_io_account_write(count);
 977
 978         result = -ENOMEM;
 979         dreq = nfs_direct_req_alloc();
 980         if (!dreq)
 981                 goto out;
 982
 983         dreq->inode = inode;
 984         dreq->bytes_left = dreq->max_count = count;
 985         dreq->io_start = pos;
 986         dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp));
 987         l_ctx = nfs_get_lock_context(dreq->ctx);
 988         if (IS_ERR(l_ctx)) {
 989                 result = PTR_ERR(l_ctx);
 990                 nfs_direct_req_release(dreq);
 991                 goto out_release;
 992         }
 993         dreq->l_ctx = l_ctx;
 994         if (!is_sync_kiocb(iocb))
 995                 dreq->iocb = iocb;
 996         pnfs_init_ds_commit_info_ops(&dreq->ds_cinfo, inode);
 997
 998         if (swap) {
 999                 requested = nfs_direct_write_schedule_iovec(dreq, iter, pos,
1000                                                             FLUSH_STABLE);
1001         } else {
1002                 nfs_start_io_direct(inode);
1003
1004                 requested = nfs_direct_write_schedule_iovec(dreq, iter, pos,
1005                                                             FLUSH_COND_STABLE);
1006
1007                 if (mapping->nrpages) {
1008                         invalidate_inode_pages2_range(mapping,
1009                                                       pos >> PAGE_SHIFT, end);
1010                 }
1011
1012                 nfs_end_io_direct(inode);
1013         }
1014
1015         if (requested > 0) {
1016                 result = nfs_direct_wait(dreq);
1017                 if (result > 0) {
1018                         requested -= result;
1019                         iocb->ki_pos = pos + result;
1020                         /* XXX: should check the generic_write_sync retval */
1021                         generic_write_sync(iocb, result);
1022                 }
1023                 iov_iter_revert(iter, requested);
1024         } else {
1025                 result = requested;
1026         }
1027         nfs_fscache_invalidate(inode, FSCACHE_INVAL_DIO_WRITE);
1028 out_release:
1029         nfs_direct_req_release(dreq);
1030 out:
1031         return result;
1032 }
1033
1034 /**
1035  * nfs_init_directcache - create a slab cache for nfs_direct_req structures
1036  *
1037  */
1038 int __init nfs_init_directcache(void)
1039 {
1040         nfs_direct_cachep = kmem_cache_create("nfs_direct_cache",
1041                                                 sizeof(struct nfs_direct_req),
1042                                                 0, (SLAB_RECLAIM_ACCOUNT|
1043                                                         SLAB_MEM_SPREAD),
1044                                                 NULL);
1045         if (nfs_direct_cachep == NULL)
1046                 return -ENOMEM;
1047
1048         return 0;
1049 }
1050
1051 /**
1052  * nfs_destroy_directcache - destroy the slab cache for nfs_direct_req structures
1053  *
1054  */
1055 void nfs_destroy_directcache(void)
1056 {
1057         kmem_cache_destroy(nfs_direct_cachep);
1058 }