fs/nfs/direct.c

   1 /*
   2  * linux/fs/nfs/direct.c
   3  *
   4  * Copyright (C) 2003 by Chuck Lever <cel@netapp.com>
   5  *
   6  * High-performance uncached I/O for the Linux NFS client
   7  *
   8  * There are important applications whose performance or correctness
   9  * depends on uncached access to file data.  Database clusters
  10  * (multiple copies of the same instance running on separate hosts)
  11  * implement their own cache coherency protocol that subsumes file
  12  * system cache protocols.  Applications that process datasets
  13  * considerably larger than the client's memory do not always benefit
  14  * from a local cache.  A streaming video server, for instance, has no
  15  * need to cache the contents of a file.
  16  *
  17  * When an application requests uncached I/O, all read and write requests
  18  * are made directly to the server; data stored or fetched via these
  19  * requests is not cached in the Linux page cache.  The client does not
  20  * correct unaligned requests from applications.  All requested bytes are
  21  * held on permanent storage before a direct write system call returns to
  22  * an application.
  23  *
  24  * Solaris implements an uncached I/O facility called directio() that
  25  * is used for backups and sequential I/O to very large files.  Solaris
  26  * also supports uncaching whole NFS partitions with "-o forcedirectio,"
  27  * an undocumented mount option.
  28  *
  29  * Designed by Jeff Kimmel, Chuck Lever, and Trond Myklebust, with
  30  * help from Andrew Morton.
  31  *
  32  * 18 Dec 2001  Initial implementation for 2.4  --cel
  33  * 08 Jul 2002  Version for 2.4.19, with bug fixes --trondmy
  34  * 08 Jun 2003  Port to 2.5 APIs  --cel
  35  * 31 Mar 2004  Handle direct I/O without VFS support  --cel
  36  * 15 Sep 2004  Parallel async reads  --cel
  37  * 04 May 2005  support O_DIRECT with aio  --cel
  38  *
  39  */
  40
  41 #include <linux/errno.h>
  42 #include <linux/sched.h>
  43 #include <linux/kernel.h>
  44 #include <linux/file.h>
  45 #include <linux/pagemap.h>
  46 #include <linux/kref.h>
  47 #include <linux/slab.h>
  48 #include <linux/task_io_accounting_ops.h>
  49 #include <linux/module.h>
  50
  51 #include <linux/nfs_fs.h>
  52 #include <linux/nfs_page.h>
  53 #include <linux/sunrpc/clnt.h>
  54
  55 #include <asm/uaccess.h>
  56 #include <linux/atomic.h>
  57
  58 #include "internal.h"
  59 #include "iostat.h"
  60 #include "pnfs.h"
  61
  62 #define NFSDBG_FACILITY         NFSDBG_VFS
  63
  64 static struct kmem_cache *nfs_direct_cachep;
  65
  66 /*
  67  * This represents a set of asynchronous requests that we're waiting on
  68  */
  69 struct nfs_direct_req {
  70         struct kref             kref;           /* release manager */
  71
  72         /* I/O parameters */
  73         struct nfs_open_context *ctx;           /* file open context info */
  74         struct nfs_lock_context *l_ctx;         /* Lock context info */
  75         struct kiocb *          iocb;           /* controlling i/o request */
  76         struct inode *          inode;          /* target file of i/o */
  77
  78         /* completion state */
  79         atomic_t                io_count;       /* i/os we're waiting for */
  80         spinlock_t              lock;           /* protect completion state */
  81         ssize_t                 count,          /* bytes actually processed */
  82                                 bytes_left,     /* bytes left to be sent */
  83                                 error;          /* any reported error */
  84         struct completion       completion;     /* wait for i/o completion */
  85
  86         /* commit state */
  87         struct nfs_mds_commit_info mds_cinfo;   /* Storage for cinfo */
  88         struct pnfs_ds_commit_info ds_cinfo;    /* Storage for cinfo */
  89         struct work_struct      work;
  90         int                     flags;
  91 #define NFS_ODIRECT_DO_COMMIT           (1)     /* an unstable reply was received */
  92 #define NFS_ODIRECT_RESCHED_WRITES      (2)     /* write verification failed */
  93         struct nfs_writeverf    verf;           /* unstable write verifier */
  94 };
  95
  96 static const struct nfs_pgio_completion_ops nfs_direct_write_completion_ops;
  97 static const struct nfs_commit_completion_ops nfs_direct_commit_completion_ops;
  98 static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode *inode);
  99 static void nfs_direct_write_schedule_work(struct work_struct *work);
 100
 101 static inline void get_dreq(struct nfs_direct_req *dreq)
 102 {
 103         atomic_inc(&dreq->io_count);
 104 }
 105
 106 static inline int put_dreq(struct nfs_direct_req *dreq)
 107 {
 108         return atomic_dec_and_test(&dreq->io_count);
 109 }
 110
 111 /**
 112  * nfs_direct_IO - NFS address space operation for direct I/O
 113  * @rw: direction (read or write)
 114  * @iocb: target I/O control block
 115  * @iov: array of vectors that define I/O buffer
 116  * @pos: offset in file to begin the operation
 117  * @nr_segs: size of iovec array
 118  *
 119  * The presence of this routine in the address space ops vector means
 120  * the NFS client supports direct I/O. However, for most direct IO, we
 121  * shunt off direct read and write requests before the VFS gets them,
 122  * so this method is only ever called for swap.
 123  */
 124 ssize_t nfs_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov, loff_t pos, unsigned long nr_segs)
 125 {
 126 #ifndef CONFIG_NFS_SWAP
 127         dprintk("NFS: nfs_direct_IO (%pD) off/no(%Ld/%lu) EINVAL\n",
 128                         iocb->ki_filp, (long long) pos, nr_segs);
 129
 130         return -EINVAL;
 131 #else
 132         VM_BUG_ON(iocb->ki_nbytes != PAGE_SIZE);
 133
 134         if (rw == READ || rw == KERNEL_READ)
 135                 return nfs_file_direct_read(iocb, iov, nr_segs, pos,
 136                                 rw == READ ? true : false);
 137         return nfs_file_direct_write(iocb, iov, nr_segs, pos,
 138                                 rw == WRITE ? true : false);
 139 #endif /* CONFIG_NFS_SWAP */
 140 }
 141
 142 static void nfs_direct_release_pages(struct page **pages, unsigned int npages)
 143 {
 144         unsigned int i;
 145         for (i = 0; i < npages; i++)
 146                 page_cache_release(pages[i]);
 147 }
 148
 149 void nfs_init_cinfo_from_dreq(struct nfs_commit_info *cinfo,
 150                               struct nfs_direct_req *dreq)
 151 {
 152         cinfo->lock = &dreq->lock;
 153         cinfo->mds = &dreq->mds_cinfo;
 154         cinfo->ds = &dreq->ds_cinfo;
 155         cinfo->dreq = dreq;
 156         cinfo->completion_ops = &nfs_direct_commit_completion_ops;
 157 }
 158
 159 static inline struct nfs_direct_req *nfs_direct_req_alloc(void)
 160 {
 161         struct nfs_direct_req *dreq;
 162
 163         dreq = kmem_cache_zalloc(nfs_direct_cachep, GFP_KERNEL);
 164         if (!dreq)
 165                 return NULL;
 166
 167         kref_init(&dreq->kref);
 168         kref_get(&dreq->kref);
 169         init_completion(&dreq->completion);
 170         INIT_LIST_HEAD(&dreq->mds_cinfo.list);
 171         INIT_WORK(&dreq->work, nfs_direct_write_schedule_work);
 172         spin_lock_init(&dreq->lock);
 173
 174         return dreq;
 175 }
 176
 177 static void nfs_direct_req_free(struct kref *kref)
 178 {
 179         struct nfs_direct_req *dreq = container_of(kref, struct nfs_direct_req, kref);
 180
 181         if (dreq->l_ctx != NULL)
 182                 nfs_put_lock_context(dreq->l_ctx);
 183         if (dreq->ctx != NULL)
 184                 put_nfs_open_context(dreq->ctx);
 185         kmem_cache_free(nfs_direct_cachep, dreq);
 186 }
 187
 188 static void nfs_direct_req_release(struct nfs_direct_req *dreq)
 189 {
 190         kref_put(&dreq->kref, nfs_direct_req_free);
 191 }
 192
 193 ssize_t nfs_dreq_bytes_left(struct nfs_direct_req *dreq)
 194 {
 195         return dreq->bytes_left;
 196 }
 197 EXPORT_SYMBOL_GPL(nfs_dreq_bytes_left);
 198
 199 /*
 200  * Collects and returns the final error value/byte-count.
 201  */
 202 static ssize_t nfs_direct_wait(struct nfs_direct_req *dreq)
 203 {
 204         ssize_t result = -EIOCBQUEUED;
 205
 206         /* Async requests don't wait here */
 207         if (dreq->iocb)
 208                 goto out;
 209
 210         result = wait_for_completion_killable(&dreq->completion);
 211
 212         if (!result)
 213                 result = dreq->error;
 214         if (!result)
 215                 result = dreq->count;
 216
 217 out:
 218         return (ssize_t) result;
 219 }
 220
 221 /*
 222  * Synchronous I/O uses a stack-allocated iocb.  Thus we can't trust
 223  * the iocb is still valid here if this is a synchronous request.
 224  */
 225 static void nfs_direct_complete(struct nfs_direct_req *dreq)
 226 {
 227         if (dreq->iocb) {
 228                 long res = (long) dreq->error;
 229                 if (!res)
 230                         res = (long) dreq->count;
 231                 aio_complete(dreq->iocb, res, 0);
 232         }
 233         complete_all(&dreq->completion);
 234
 235         nfs_direct_req_release(dreq);
 236 }
 237
 238 static void nfs_direct_readpage_release(struct nfs_page *req)
 239 {
 240         dprintk("NFS: direct read done (%s/%lld %d@%lld)\n",
 241                 req->wb_context->dentry->d_inode->i_sb->s_id,
 242                 (long long)NFS_FILEID(req->wb_context->dentry->d_inode),
 243                 req->wb_bytes,
 244                 (long long)req_offset(req));
 245         nfs_release_request(req);
 246 }
 247
 248 static void nfs_direct_read_completion(struct nfs_pgio_header *hdr)
 249 {
 250         unsigned long bytes = 0;
 251         struct nfs_direct_req *dreq = hdr->dreq;
 252
 253         if (test_bit(NFS_IOHDR_REDO, &hdr->flags))
 254                 goto out_put;
 255
 256         spin_lock(&dreq->lock);
 257         if (test_bit(NFS_IOHDR_ERROR, &hdr->flags) && (hdr->good_bytes == 0))
 258                 dreq->error = hdr->error;
 259         else
 260                 dreq->count += hdr->good_bytes;
 261         spin_unlock(&dreq->lock);
 262
 263         while (!list_empty(&hdr->pages)) {
 264                 struct nfs_page *req = nfs_list_entry(hdr->pages.next);
 265                 struct page *page = req->wb_page;
 266
 267                 if (!PageCompound(page) && bytes < hdr->good_bytes)
 268                         set_page_dirty(page);
 269                 bytes += req->wb_bytes;
 270                 nfs_list_remove_request(req);
 271                 nfs_direct_readpage_release(req);
 272         }
 273 out_put:
 274         if (put_dreq(dreq))
 275                 nfs_direct_complete(dreq);
 276         hdr->release(hdr);
 277 }
 278
 279 static void nfs_read_sync_pgio_error(struct list_head *head)
 280 {
 281         struct nfs_page *req;
 282
 283         while (!list_empty(head)) {
 284                 req = nfs_list_entry(head->next);
 285                 nfs_list_remove_request(req);
 286                 nfs_release_request(req);
 287         }
 288 }
 289
 290 static void nfs_direct_pgio_init(struct nfs_pgio_header *hdr)
 291 {
 292         get_dreq(hdr->dreq);
 293 }
 294
 295 static const struct nfs_pgio_completion_ops nfs_direct_read_completion_ops = {
 296         .error_cleanup = nfs_read_sync_pgio_error,
 297         .init_hdr = nfs_direct_pgio_init,
 298         .completion = nfs_direct_read_completion,
 299 };
 300
 301 /*
 302  * For each rsize'd chunk of the user's buffer, dispatch an NFS READ
 303  * operation.  If nfs_readdata_alloc() or get_user_pages() fails,
 304  * bail and stop sending more reads.  Read length accounting is
 305  * handled automatically by nfs_direct_read_result().  Otherwise, if
 306  * no requests have been sent, just return an error.
 307  */
 308 static ssize_t nfs_direct_read_schedule_segment(struct nfs_pageio_descriptor *desc,
 309                                                 const struct iovec *iov,
 310                                                 loff_t pos, bool uio)
 311 {
 312         struct nfs_direct_req *dreq = desc->pg_dreq;
 313         struct nfs_open_context *ctx = dreq->ctx;
 314         struct inode *inode = ctx->dentry->d_inode;
 315         unsigned long user_addr = (unsigned long)iov->iov_base;
 316         size_t count = iov->iov_len;
 317         size_t rsize = NFS_SERVER(inode)->rsize;
 318         unsigned int pgbase;
 319         int result;
 320         ssize_t started = 0;
 321         struct page **pagevec = NULL;
 322         unsigned int npages;
 323
 324         do {
 325                 size_t bytes;
 326                 int i;
 327
 328                 pgbase = user_addr & ~PAGE_MASK;
 329                 bytes = min(max_t(size_t, rsize, PAGE_SIZE), count);
 330
 331                 result = -ENOMEM;
 332                 npages = nfs_page_array_len(pgbase, bytes);
 333                 if (!pagevec)
 334                         pagevec = kmalloc(npages * sizeof(struct page *),
 335                                           GFP_KERNEL);
 336                 if (!pagevec)
 337                         break;
 338                 if (uio) {
 339                         down_read(&current->mm->mmap_sem);
 340                         result = get_user_pages(current, current->mm, user_addr,
 341                                         npages, 1, 0, pagevec, NULL);
 342                         up_read(&current->mm->mmap_sem);
 343                         if (result < 0)
 344                                 break;
 345                 } else {
 346                         WARN_ON(npages != 1);
 347                         result = get_kernel_page(user_addr, 1, pagevec);
 348                         if (WARN_ON(result != 1))
 349                                 break;
 350                 }
 351
 352                 if ((unsigned)result < npages) {
 353                         bytes = result * PAGE_SIZE;
 354                         if (bytes <= pgbase) {
 355                                 nfs_direct_release_pages(pagevec, result);
 356                                 break;
 357                         }
 358                         bytes -= pgbase;
 359                         npages = result;
 360                 }
 361
 362                 for (i = 0; i < npages; i++) {
 363                         struct nfs_page *req;
 364                         unsigned int req_len = min_t(size_t, bytes, PAGE_SIZE - pgbase);
 365                         /* XXX do we need to do the eof zeroing found in async_filler? */
 366                         req = nfs_create_request(dreq->ctx, dreq->inode,
 367                                                  pagevec[i],
 368                                                  pgbase, req_len);
 369                         if (IS_ERR(req)) {
 370                                 result = PTR_ERR(req);
 371                                 break;
 372                         }
 373                         req->wb_index = pos >> PAGE_SHIFT;
 374                         req->wb_offset = pos & ~PAGE_MASK;
 375                         if (!nfs_pageio_add_request(desc, req)) {
 376                                 result = desc->pg_error;
 377                                 nfs_release_request(req);
 378                                 break;
 379                         }
 380                         pgbase = 0;
 381                         bytes -= req_len;
 382                         started += req_len;
 383                         user_addr += req_len;
 384                         pos += req_len;
 385                         count -= req_len;
 386                         dreq->bytes_left -= req_len;
 387                 }
 388                 /* The nfs_page now hold references to these pages */
 389                 nfs_direct_release_pages(pagevec, npages);
 390         } while (count != 0 && result >= 0);
 391
 392         kfree(pagevec);
 393
 394         if (started)
 395                 return started;
 396         return result < 0 ? (ssize_t) result : -EFAULT;
 397 }
 398
 399 static ssize_t nfs_direct_read_schedule_iovec(struct nfs_direct_req *dreq,
 400                                               const struct iovec *iov,
 401                                               unsigned long nr_segs,
 402                                               loff_t pos, bool uio)
 403 {
 404         struct nfs_pageio_descriptor desc;
 405         ssize_t result = -EINVAL;
 406         size_t requested_bytes = 0;
 407         unsigned long seg;
 408
 409         NFS_PROTO(dreq->inode)->read_pageio_init(&desc, dreq->inode,
 410                              &nfs_direct_read_completion_ops);
 411         get_dreq(dreq);
 412         desc.pg_dreq = dreq;
 413
 414         for (seg = 0; seg < nr_segs; seg++) {
 415                 const struct iovec *vec = &iov[seg];
 416                 result = nfs_direct_read_schedule_segment(&desc, vec, pos, uio);
 417                 if (result < 0)
 418                         break;
 419                 requested_bytes += result;
 420                 if ((size_t)result < vec->iov_len)
 421                         break;
 422                 pos += vec->iov_len;
 423         }
 424
 425         nfs_pageio_complete(&desc);
 426
 427         /*
 428          * If no bytes were started, return the error, and let the
 429          * generic layer handle the completion.
 430          */
 431         if (requested_bytes == 0) {
 432                 nfs_direct_req_release(dreq);
 433                 return result < 0 ? result : -EIO;
 434         }
 435
 436         if (put_dreq(dreq))
 437                 nfs_direct_complete(dreq);
 438         return 0;
 439 }
 440
 441 static ssize_t nfs_direct_read(struct kiocb *iocb, const struct iovec *iov,
 442                                unsigned long nr_segs, loff_t pos, bool uio)
 443 {
 444         ssize_t result = -ENOMEM;
 445         struct inode *inode = iocb->ki_filp->f_mapping->host;
 446         struct nfs_direct_req *dreq;
 447         struct nfs_lock_context *l_ctx;
 448
 449         dreq = nfs_direct_req_alloc();
 450         if (dreq == NULL)
 451                 goto out;
 452
 453         dreq->inode = inode;
 454         dreq->bytes_left = iov_length(iov, nr_segs);
 455         dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp));
 456         l_ctx = nfs_get_lock_context(dreq->ctx);
 457         if (IS_ERR(l_ctx)) {
 458                 result = PTR_ERR(l_ctx);
 459                 goto out_release;
 460         }
 461         dreq->l_ctx = l_ctx;
 462         if (!is_sync_kiocb(iocb))
 463                 dreq->iocb = iocb;
 464
 465         NFS_I(inode)->read_io += iov_length(iov, nr_segs);
 466         result = nfs_direct_read_schedule_iovec(dreq, iov, nr_segs, pos, uio);
 467         if (!result)
 468                 result = nfs_direct_wait(dreq);
 469 out_release:
 470         nfs_direct_req_release(dreq);
 471 out:
 472         return result;
 473 }
 474
 475 static void nfs_inode_dio_write_done(struct inode *inode)
 476 {
 477         nfs_zap_mapping(inode, inode->i_mapping);
 478         inode_dio_done(inode);
 479 }
 480
 481 #if IS_ENABLED(CONFIG_NFS_V3) || IS_ENABLED(CONFIG_NFS_V4)
 482 static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq)
 483 {
 484         struct nfs_pageio_descriptor desc;
 485         struct nfs_page *req, *tmp;
 486         LIST_HEAD(reqs);
 487         struct nfs_commit_info cinfo;
 488         LIST_HEAD(failed);
 489
 490         nfs_init_cinfo_from_dreq(&cinfo, dreq);
 491         pnfs_recover_commit_reqs(dreq->inode, &reqs, &cinfo);
 492         spin_lock(cinfo.lock);
 493         nfs_scan_commit_list(&cinfo.mds->list, &reqs, &cinfo, 0);
 494         spin_unlock(cinfo.lock);
 495
 496         dreq->count = 0;
 497         get_dreq(dreq);
 498
 499         NFS_PROTO(dreq->inode)->write_pageio_init(&desc, dreq->inode, FLUSH_STABLE,
 500                               &nfs_direct_write_completion_ops);
 501         desc.pg_dreq = dreq;
 502
 503         list_for_each_entry_safe(req, tmp, &reqs, wb_list) {
 504                 if (!nfs_pageio_add_request(&desc, req)) {
 505                         nfs_list_remove_request(req);
 506                         nfs_list_add_request(req, &failed);
 507                         spin_lock(cinfo.lock);
 508                         dreq->flags = 0;
 509                         dreq->error = -EIO;
 510                         spin_unlock(cinfo.lock);
 511                 }
 512                 nfs_release_request(req);
 513         }
 514         nfs_pageio_complete(&desc);
 515
 516         while (!list_empty(&failed)) {
 517                 req = nfs_list_entry(failed.next);
 518                 nfs_list_remove_request(req);
 519                 nfs_unlock_and_release_request(req);
 520         }
 521
 522         if (put_dreq(dreq))
 523                 nfs_direct_write_complete(dreq, dreq->inode);
 524 }
 525
 526 static void nfs_direct_commit_complete(struct nfs_commit_data *data)
 527 {
 528         struct nfs_direct_req *dreq = data->dreq;
 529         struct nfs_commit_info cinfo;
 530         struct nfs_page *req;
 531         int status = data->task.tk_status;
 532
 533         nfs_init_cinfo_from_dreq(&cinfo, dreq);
 534         if (status < 0) {
 535                 dprintk("NFS: %5u commit failed with error %d.\n",
 536                         data->task.tk_pid, status);
 537                 dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
 538         } else if (memcmp(&dreq->verf, &data->verf, sizeof(data->verf))) {
 539                 dprintk("NFS: %5u commit verify failed\n", data->task.tk_pid);
 540                 dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
 541         }
 542
 543         dprintk("NFS: %5u commit returned %d\n", data->task.tk_pid, status);
 544         while (!list_empty(&data->pages)) {
 545                 req = nfs_list_entry(data->pages.next);
 546                 nfs_list_remove_request(req);
 547                 if (dreq->flags == NFS_ODIRECT_RESCHED_WRITES) {
 548                         /* Note the rewrite will go through mds */
 549                         nfs_mark_request_commit(req, NULL, &cinfo);
 550                 } else
 551                         nfs_release_request(req);
 552                 nfs_unlock_and_release_request(req);
 553         }
 554
 555         if (atomic_dec_and_test(&cinfo.mds->rpcs_out))
 556                 nfs_direct_write_complete(dreq, data->inode);
 557 }
 558
 559 static void nfs_direct_error_cleanup(struct nfs_inode *nfsi)
 560 {
 561         /* There is no lock to clear */
 562 }
 563
 564 static const struct nfs_commit_completion_ops nfs_direct_commit_completion_ops = {
 565         .completion = nfs_direct_commit_complete,
 566         .error_cleanup = nfs_direct_error_cleanup,
 567 };
 568
 569 static void nfs_direct_commit_schedule(struct nfs_direct_req *dreq)
 570 {
 571         int res;
 572         struct nfs_commit_info cinfo;
 573         LIST_HEAD(mds_list);
 574
 575         nfs_init_cinfo_from_dreq(&cinfo, dreq);
 576         nfs_scan_commit(dreq->inode, &mds_list, &cinfo);
 577         res = nfs_generic_commit_list(dreq->inode, &mds_list, 0, &cinfo);
 578         if (res < 0) /* res == -ENOMEM */
 579                 nfs_direct_write_reschedule(dreq);
 580 }
 581
 582 static void nfs_direct_write_schedule_work(struct work_struct *work)
 583 {
 584         struct nfs_direct_req *dreq = container_of(work, struct nfs_direct_req, work);
 585         int flags = dreq->flags;
 586
 587         dreq->flags = 0;
 588         switch (flags) {
 589                 case NFS_ODIRECT_DO_COMMIT:
 590                         nfs_direct_commit_schedule(dreq);
 591                         break;
 592                 case NFS_ODIRECT_RESCHED_WRITES:
 593                         nfs_direct_write_reschedule(dreq);
 594                         break;
 595                 default:
 596                         nfs_inode_dio_write_done(dreq->inode);
 597                         nfs_direct_complete(dreq);
 598         }
 599 }
 600
 601 static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode *inode)
 602 {
 603         schedule_work(&dreq->work); /* Calls nfs_direct_write_schedule_work */
 604 }
 605
 606 #else
 607 static void nfs_direct_write_schedule_work(struct work_struct *work)
 608 {
 609 }
 610
 611 static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode *inode)
 612 {
 613         nfs_inode_dio_write_done(inode);
 614         nfs_direct_complete(dreq);
 615 }
 616 #endif
 617
 618 /*
 619  * NB: Return the value of the first error return code.  Subsequent
 620  *     errors after the first one are ignored.
 621  */
 622 /*
 623  * For each wsize'd chunk of the user's buffer, dispatch an NFS WRITE
 624  * operation.  If nfs_writedata_alloc() or get_user_pages() fails,
 625  * bail and stop sending more writes.  Write length accounting is
 626  * handled automatically by nfs_direct_write_result().  Otherwise, if
 627  * no requests have been sent, just return an error.
 628  */
 629 static ssize_t nfs_direct_write_schedule_segment(struct nfs_pageio_descriptor *desc,
 630                                                  const struct iovec *iov,
 631                                                  loff_t pos, bool uio)
 632 {
 633         struct nfs_direct_req *dreq = desc->pg_dreq;
 634         struct nfs_open_context *ctx = dreq->ctx;
 635         struct inode *inode = ctx->dentry->d_inode;
 636         unsigned long user_addr = (unsigned long)iov->iov_base;
 637         size_t count = iov->iov_len;
 638         size_t wsize = NFS_SERVER(inode)->wsize;
 639         unsigned int pgbase;
 640         int result;
 641         ssize_t started = 0;
 642         struct page **pagevec = NULL;
 643         unsigned int npages;
 644
 645         do {
 646                 size_t bytes;
 647                 int i;
 648
 649                 pgbase = user_addr & ~PAGE_MASK;
 650                 bytes = min(max_t(size_t, wsize, PAGE_SIZE), count);
 651
 652                 result = -ENOMEM;
 653                 npages = nfs_page_array_len(pgbase, bytes);
 654                 if (!pagevec)
 655                         pagevec = kmalloc(npages * sizeof(struct page *), GFP_KERNEL);
 656                 if (!pagevec)
 657                         break;
 658
 659                 if (uio) {
 660                         down_read(&current->mm->mmap_sem);
 661                         result = get_user_pages(current, current->mm, user_addr,
 662                                                 npages, 0, 0, pagevec, NULL);
 663                         up_read(&current->mm->mmap_sem);
 664                         if (result < 0)
 665                                 break;
 666                 } else {
 667                         WARN_ON(npages != 1);
 668                         result = get_kernel_page(user_addr, 0, pagevec);
 669                         if (WARN_ON(result != 1))
 670                                 break;
 671                 }
 672
 673                 if ((unsigned)result < npages) {
 674                         bytes = result * PAGE_SIZE;
 675                         if (bytes <= pgbase) {
 676                                 nfs_direct_release_pages(pagevec, result);
 677                                 break;
 678                         }
 679                         bytes -= pgbase;
 680                         npages = result;
 681                 }
 682
 683                 for (i = 0; i < npages; i++) {
 684                         struct nfs_page *req;
 685                         unsigned int req_len = min_t(size_t, bytes, PAGE_SIZE - pgbase);
 686
 687                         req = nfs_create_request(dreq->ctx, dreq->inode,
 688                                                  pagevec[i],
 689                                                  pgbase, req_len);
 690                         if (IS_ERR(req)) {
 691                                 result = PTR_ERR(req);
 692                                 break;
 693                         }
 694                         nfs_lock_request(req);
 695                         req->wb_index = pos >> PAGE_SHIFT;
 696                         req->wb_offset = pos & ~PAGE_MASK;
 697                         if (!nfs_pageio_add_request(desc, req)) {
 698                                 result = desc->pg_error;
 699                                 nfs_unlock_and_release_request(req);
 700                                 break;
 701                         }
 702                         pgbase = 0;
 703                         bytes -= req_len;
 704                         started += req_len;
 705                         user_addr += req_len;
 706                         pos += req_len;
 707                         count -= req_len;
 708                         dreq->bytes_left -= req_len;
 709                 }
 710                 /* The nfs_page now hold references to these pages */
 711                 nfs_direct_release_pages(pagevec, npages);
 712         } while (count != 0 && result >= 0);
 713
 714         kfree(pagevec);
 715
 716         if (started)
 717                 return started;
 718         return result < 0 ? (ssize_t) result : -EFAULT;
 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         int bit = -1;
 726         struct nfs_page *req = nfs_list_entry(hdr->pages.next);
 727
 728         if (test_bit(NFS_IOHDR_REDO, &hdr->flags))
 729                 goto out_put;
 730
 731         nfs_init_cinfo_from_dreq(&cinfo, dreq);
 732
 733         spin_lock(&dreq->lock);
 734
 735         if (test_bit(NFS_IOHDR_ERROR, &hdr->flags)) {
 736                 dreq->flags = 0;
 737                 dreq->error = hdr->error;
 738         }
 739         if (dreq->error != 0)
 740                 bit = NFS_IOHDR_ERROR;
 741         else {
 742                 dreq->count += hdr->good_bytes;
 743                 if (test_bit(NFS_IOHDR_NEED_RESCHED, &hdr->flags)) {
 744                         dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
 745                         bit = NFS_IOHDR_NEED_RESCHED;
 746                 } else if (test_bit(NFS_IOHDR_NEED_COMMIT, &hdr->flags)) {
 747                         if (dreq->flags == NFS_ODIRECT_RESCHED_WRITES)
 748                                 bit = NFS_IOHDR_NEED_RESCHED;
 749                         else if (dreq->flags == 0) {
 750                                 memcpy(&dreq->verf, hdr->verf,
 751                                        sizeof(dreq->verf));
 752                                 bit = NFS_IOHDR_NEED_COMMIT;
 753                                 dreq->flags = NFS_ODIRECT_DO_COMMIT;
 754                         } else if (dreq->flags == NFS_ODIRECT_DO_COMMIT) {
 755                                 if (memcmp(&dreq->verf, hdr->verf, sizeof(dreq->verf))) {
 756                                         dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
 757                                         bit = NFS_IOHDR_NEED_RESCHED;
 758                                 } else
 759                                         bit = NFS_IOHDR_NEED_COMMIT;
 760                         }
 761                 }
 762         }
 763         spin_unlock(&dreq->lock);
 764
 765         while (!list_empty(&hdr->pages)) {
 766                 req = nfs_list_entry(hdr->pages.next);
 767                 nfs_list_remove_request(req);
 768                 switch (bit) {
 769                 case NFS_IOHDR_NEED_RESCHED:
 770                 case NFS_IOHDR_NEED_COMMIT:
 771                         kref_get(&req->wb_kref);
 772                         nfs_mark_request_commit(req, hdr->lseg, &cinfo);
 773                 }
 774                 nfs_unlock_and_release_request(req);
 775         }
 776
 777 out_put:
 778         if (put_dreq(dreq))
 779                 nfs_direct_write_complete(dreq, hdr->inode);
 780         hdr->release(hdr);
 781 }
 782
 783 static void nfs_write_sync_pgio_error(struct list_head *head)
 784 {
 785         struct nfs_page *req;
 786
 787         while (!list_empty(head)) {
 788                 req = nfs_list_entry(head->next);
 789                 nfs_list_remove_request(req);
 790                 nfs_unlock_and_release_request(req);
 791         }
 792 }
 793
 794 static const struct nfs_pgio_completion_ops nfs_direct_write_completion_ops = {
 795         .error_cleanup = nfs_write_sync_pgio_error,
 796         .init_hdr = nfs_direct_pgio_init,
 797         .completion = nfs_direct_write_completion,
 798 };
 799
 800 static ssize_t nfs_direct_write_schedule_iovec(struct nfs_direct_req *dreq,
 801                                                const struct iovec *iov,
 802                                                unsigned long nr_segs,
 803                                                loff_t pos, bool uio)
 804 {
 805         struct nfs_pageio_descriptor desc;
 806         struct inode *inode = dreq->inode;
 807         ssize_t result = 0;
 808         size_t requested_bytes = 0;
 809         unsigned long seg;
 810
 811         NFS_PROTO(inode)->write_pageio_init(&desc, inode, FLUSH_COND_STABLE,
 812                               &nfs_direct_write_completion_ops);
 813         desc.pg_dreq = dreq;
 814         get_dreq(dreq);
 815         atomic_inc(&inode->i_dio_count);
 816
 817         NFS_I(dreq->inode)->write_io += iov_length(iov, nr_segs);
 818         for (seg = 0; seg < nr_segs; seg++) {
 819                 const struct iovec *vec = &iov[seg];
 820                 result = nfs_direct_write_schedule_segment(&desc, vec, pos, uio);
 821                 if (result < 0)
 822                         break;
 823                 requested_bytes += result;
 824                 if ((size_t)result < vec->iov_len)
 825                         break;
 826                 pos += vec->iov_len;
 827         }
 828         nfs_pageio_complete(&desc);
 829
 830         /*
 831          * If no bytes were started, return the error, and let the
 832          * generic layer handle the completion.
 833          */
 834         if (requested_bytes == 0) {
 835                 inode_dio_done(inode);
 836                 nfs_direct_req_release(dreq);
 837                 return result < 0 ? result : -EIO;
 838         }
 839
 840         if (put_dreq(dreq))
 841                 nfs_direct_write_complete(dreq, dreq->inode);
 842         return 0;
 843 }
 844
 845 static ssize_t nfs_direct_write(struct kiocb *iocb, const struct iovec *iov,
 846                                 unsigned long nr_segs, loff_t pos,
 847                                 size_t count, bool uio)
 848 {
 849         ssize_t result = -ENOMEM;
 850         struct inode *inode = iocb->ki_filp->f_mapping->host;
 851         struct nfs_direct_req *dreq;
 852         struct nfs_lock_context *l_ctx;
 853
 854         dreq = nfs_direct_req_alloc();
 855         if (!dreq)
 856                 goto out;
 857
 858         dreq->inode = inode;
 859         dreq->bytes_left = count;
 860         dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp));
 861         l_ctx = nfs_get_lock_context(dreq->ctx);
 862         if (IS_ERR(l_ctx)) {
 863                 result = PTR_ERR(l_ctx);
 864                 goto out_release;
 865         }
 866         dreq->l_ctx = l_ctx;
 867         if (!is_sync_kiocb(iocb))
 868                 dreq->iocb = iocb;
 869
 870         result = nfs_direct_write_schedule_iovec(dreq, iov, nr_segs, pos, uio);
 871         if (!result)
 872                 result = nfs_direct_wait(dreq);
 873 out_release:
 874         nfs_direct_req_release(dreq);
 875 out:
 876         return result;
 877 }
 878
 879 /**
 880  * nfs_file_direct_read - file direct read operation for NFS files
 881  * @iocb: target I/O control block
 882  * @iov: vector of user buffers into which to read data
 883  * @nr_segs: size of iov vector
 884  * @pos: byte offset in file where reading starts
 885  *
 886  * We use this function for direct reads instead of calling
 887  * generic_file_aio_read() in order to avoid gfar's check to see if
 888  * the request starts before the end of the file.  For that check
 889  * to work, we must generate a GETATTR before each direct read, and
 890  * even then there is a window between the GETATTR and the subsequent
 891  * READ where the file size could change.  Our preference is simply
 892  * to do all reads the application wants, and the server will take
 893  * care of managing the end of file boundary.
 894  *
 895  * This function also eliminates unnecessarily updating the file's
 896  * atime locally, as the NFS server sets the file's atime, and this
 897  * client must read the updated atime from the server back into its
 898  * cache.
 899  */
 900 ssize_t nfs_file_direct_read(struct kiocb *iocb, const struct iovec *iov,
 901                                 unsigned long nr_segs, loff_t pos, bool uio)
 902 {
 903         ssize_t retval = -EINVAL;
 904         struct file *file = iocb->ki_filp;
 905         struct address_space *mapping = file->f_mapping;
 906         size_t count;
 907
 908         count = iov_length(iov, nr_segs);
 909         nfs_add_stats(mapping->host, NFSIOS_DIRECTREADBYTES, count);
 910
 911         dfprintk(FILE, "NFS: direct read(%pD2, %zd@%Ld)\n",
 912                 file, count, (long long) pos);
 913
 914         retval = 0;
 915         if (!count)
 916                 goto out;
 917
 918         retval = nfs_sync_mapping(mapping);
 919         if (retval)
 920                 goto out;
 921
 922         task_io_account_read(count);
 923
 924         retval = nfs_direct_read(iocb, iov, nr_segs, pos, uio);
 925         if (retval > 0)
 926                 iocb->ki_pos = pos + retval;
 927
 928 out:
 929         return retval;
 930 }
 931
 932 /**
 933  * nfs_file_direct_write - file direct write operation for NFS files
 934  * @iocb: target I/O control block
 935  * @iov: vector of user buffers from which to write data
 936  * @nr_segs: size of iov vector
 937  * @pos: byte offset in file where writing starts
 938  *
 939  * We use this function for direct writes instead of calling
 940  * generic_file_aio_write() in order to avoid taking the inode
 941  * semaphore and updating the i_size.  The NFS server will set
 942  * the new i_size and this client must read the updated size
 943  * back into its cache.  We let the server do generic write
 944  * parameter checking and report problems.
 945  *
 946  * We eliminate local atime updates, see direct read above.
 947  *
 948  * We avoid unnecessary page cache invalidations for normal cached
 949  * readers of this file.
 950  *
 951  * Note that O_APPEND is not supported for NFS direct writes, as there
 952  * is no atomic O_APPEND write facility in the NFS protocol.
 953  */
 954 ssize_t nfs_file_direct_write(struct kiocb *iocb, const struct iovec *iov,
 955                                 unsigned long nr_segs, loff_t pos, bool uio)
 956 {
 957         ssize_t retval = -EINVAL;
 958         struct file *file = iocb->ki_filp;
 959         struct address_space *mapping = file->f_mapping;
 960         size_t count;
 961
 962         count = iov_length(iov, nr_segs);
 963         nfs_add_stats(mapping->host, NFSIOS_DIRECTWRITTENBYTES, count);
 964
 965         dfprintk(FILE, "NFS: direct write(%pD2, %zd@%Ld)\n",
 966                 file, count, (long long) pos);
 967
 968         retval = generic_write_checks(file, &pos, &count, 0);
 969         if (retval)
 970                 goto out;
 971
 972         retval = -EINVAL;
 973         if ((ssize_t) count < 0)
 974                 goto out;
 975         retval = 0;
 976         if (!count)
 977                 goto out;
 978
 979         retval = nfs_sync_mapping(mapping);
 980         if (retval)
 981                 goto out;
 982
 983         task_io_account_write(count);
 984
 985         retval = nfs_direct_write(iocb, iov, nr_segs, pos, count, uio);
 986         if (retval > 0) {
 987                 struct inode *inode = mapping->host;
 988
 989                 iocb->ki_pos = pos + retval;
 990                 spin_lock(&inode->i_lock);
 991                 if (i_size_read(inode) < iocb->ki_pos)
 992                         i_size_write(inode, iocb->ki_pos);
 993                 spin_unlock(&inode->i_lock);
 994         }
 995 out:
 996         return retval;
 997 }
 998
 999 /**
1000  * nfs_init_directcache - create a slab cache for nfs_direct_req structures
1001  *
1002  */
1003 int __init nfs_init_directcache(void)
1004 {
1005         nfs_direct_cachep = kmem_cache_create("nfs_direct_cache",
1006                                                 sizeof(struct nfs_direct_req),
1007                                                 0, (SLAB_RECLAIM_ACCOUNT|
1008                                                         SLAB_MEM_SPREAD),
1009                                                 NULL);
1010         if (nfs_direct_cachep == NULL)
1011                 return -ENOMEM;
1012
1013         return 0;
1014 }
1015
1016 /**
1017  * nfs_destroy_directcache - destroy the slab cache for nfs_direct_req structures
1018  *
1019  */
1020 void nfs_destroy_directcache(void)
1021 {
1022         kmem_cache_destroy(nfs_direct_cachep);
1023 }