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
  48 #include <linux/nfs_fs.h>
  49 #include <linux/nfs_page.h>
  50 #include <linux/sunrpc/clnt.h>
  51
  52 #include <asm/system.h>
  53 #include <asm/uaccess.h>
  54 #include <asm/atomic.h>
  55
  56 #include "internal.h"
  57 #include "iostat.h"
  58
  59 #define NFSDBG_FACILITY         NFSDBG_VFS
  60
  61 static struct kmem_cache *nfs_direct_cachep;
  62
  63 /*
  64  * This represents a set of asynchronous requests that we're waiting on
  65  */
  66 struct nfs_direct_req {
  67         struct kref             kref;           /* release manager */
  68
  69         /* I/O parameters */
  70         struct nfs_open_context *ctx;           /* file open context info */
  71         struct kiocb *          iocb;           /* controlling i/o request */
  72         struct inode *          inode;          /* target file of i/o */
  73
  74         /* completion state */
  75         atomic_t                io_count;       /* i/os we're waiting for */
  76         spinlock_t              lock;           /* protect completion state */
  77         ssize_t                 count,          /* bytes actually processed */
  78                                 error;          /* any reported error */
  79         struct completion       completion;     /* wait for i/o completion */
  80
  81         /* commit state */
  82         struct list_head        rewrite_list;   /* saved nfs_write_data structs */
  83         struct nfs_write_data * commit_data;    /* special write_data for commits */
  84         int                     flags;
  85 #define NFS_ODIRECT_DO_COMMIT           (1)     /* an unstable reply was received */
  86 #define NFS_ODIRECT_RESCHED_WRITES      (2)     /* write verification failed */
  87         struct nfs_writeverf    verf;           /* unstable write verifier */
  88 };
  89
  90 static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode *inode);
  91 static const struct rpc_call_ops nfs_write_direct_ops;
  92
  93 static inline void get_dreq(struct nfs_direct_req *dreq)
  94 {
  95         atomic_inc(&dreq->io_count);
  96 }
  97
  98 static inline int put_dreq(struct nfs_direct_req *dreq)
  99 {
 100         return atomic_dec_and_test(&dreq->io_count);
 101 }
 102
 103 /**
 104  * nfs_direct_IO - NFS address space operation for direct I/O
 105  * @rw: direction (read or write)
 106  * @iocb: target I/O control block
 107  * @iov: array of vectors that define I/O buffer
 108  * @pos: offset in file to begin the operation
 109  * @nr_segs: size of iovec array
 110  *
 111  * The presence of this routine in the address space ops vector means
 112  * the NFS client supports direct I/O.  However, we shunt off direct
 113  * read and write requests before the VFS gets them, so this method
 114  * should never be called.
 115  */
 116 ssize_t nfs_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov, loff_t pos, unsigned long nr_segs)
 117 {
 118         dprintk("NFS: nfs_direct_IO (%s) off/no(%Ld/%lu) EINVAL\n",
 119                         iocb->ki_filp->f_path.dentry->d_name.name,
 120                         (long long) pos, nr_segs);
 121
 122         return -EINVAL;
 123 }
 124
 125 static void nfs_direct_dirty_pages(struct page **pages, unsigned int pgbase, size_t count)
 126 {
 127         unsigned int npages;
 128         unsigned int i;
 129
 130         if (count == 0)
 131                 return;
 132         pages += (pgbase >> PAGE_SHIFT);
 133         npages = (count + (pgbase & ~PAGE_MASK) + PAGE_SIZE - 1) >> PAGE_SHIFT;
 134         for (i = 0; i < npages; i++) {
 135                 struct page *page = pages[i];
 136                 if (!PageCompound(page))
 137                         set_page_dirty(page);
 138         }
 139 }
 140
 141 static void nfs_direct_release_pages(struct page **pages, unsigned int npages)
 142 {
 143         unsigned int i;
 144         for (i = 0; i < npages; i++)
 145                 page_cache_release(pages[i]);
 146 }
 147
 148 static inline struct nfs_direct_req *nfs_direct_req_alloc(void)
 149 {
 150         struct nfs_direct_req *dreq;
 151
 152         dreq = kmem_cache_alloc(nfs_direct_cachep, GFP_KERNEL);
 153         if (!dreq)
 154                 return NULL;
 155
 156         kref_init(&dreq->kref);
 157         kref_get(&dreq->kref);
 158         init_completion(&dreq->completion);
 159         INIT_LIST_HEAD(&dreq->rewrite_list);
 160         dreq->iocb = NULL;
 161         dreq->ctx = NULL;
 162         spin_lock_init(&dreq->lock);
 163         atomic_set(&dreq->io_count, 0);
 164         dreq->count = 0;
 165         dreq->error = 0;
 166         dreq->flags = 0;
 167
 168         return dreq;
 169 }
 170
 171 static void nfs_direct_req_free(struct kref *kref)
 172 {
 173         struct nfs_direct_req *dreq = container_of(kref, struct nfs_direct_req, kref);
 174
 175         if (dreq->ctx != NULL)
 176                 put_nfs_open_context(dreq->ctx);
 177         kmem_cache_free(nfs_direct_cachep, dreq);
 178 }
 179
 180 static void nfs_direct_req_release(struct nfs_direct_req *dreq)
 181 {
 182         kref_put(&dreq->kref, nfs_direct_req_free);
 183 }
 184
 185 /*
 186  * Collects and returns the final error value/byte-count.
 187  */
 188 static ssize_t nfs_direct_wait(struct nfs_direct_req *dreq)
 189 {
 190         ssize_t result = -EIOCBQUEUED;
 191
 192         /* Async requests don't wait here */
 193         if (dreq->iocb)
 194                 goto out;
 195
 196         result = wait_for_completion_killable(&dreq->completion);
 197
 198         if (!result)
 199                 result = dreq->error;
 200         if (!result)
 201                 result = dreq->count;
 202
 203 out:
 204         return (ssize_t) result;
 205 }
 206
 207 /*
 208  * Synchronous I/O uses a stack-allocated iocb.  Thus we can't trust
 209  * the iocb is still valid here if this is a synchronous request.
 210  */
 211 static void nfs_direct_complete(struct nfs_direct_req *dreq)
 212 {
 213         if (dreq->iocb) {
 214                 long res = (long) dreq->error;
 215                 if (!res)
 216                         res = (long) dreq->count;
 217                 aio_complete(dreq->iocb, res, 0);
 218         }
 219         complete_all(&dreq->completion);
 220
 221         nfs_direct_req_release(dreq);
 222 }
 223
 224 /*
 225  * We must hold a reference to all the pages in this direct read request
 226  * until the RPCs complete.  This could be long *after* we are woken up in
 227  * nfs_direct_wait (for instance, if someone hits ^C on a slow server).
 228  */
 229 static void nfs_direct_read_result(struct rpc_task *task, void *calldata)
 230 {
 231         struct nfs_read_data *data = calldata;
 232
 233         nfs_readpage_result(task, data);
 234 }
 235
 236 static void nfs_direct_read_release(void *calldata)
 237 {
 238
 239         struct nfs_read_data *data = calldata;
 240         struct nfs_direct_req *dreq = (struct nfs_direct_req *) data->req;
 241         int status = data->task.tk_status;
 242
 243         spin_lock(&dreq->lock);
 244         if (unlikely(status < 0)) {
 245                 dreq->error = status;
 246                 spin_unlock(&dreq->lock);
 247         } else {
 248                 dreq->count += data->res.count;
 249                 spin_unlock(&dreq->lock);
 250                 nfs_direct_dirty_pages(data->pagevec,
 251                                 data->args.pgbase,
 252                                 data->res.count);
 253         }
 254         nfs_direct_release_pages(data->pagevec, data->npages);
 255
 256         if (put_dreq(dreq))
 257                 nfs_direct_complete(dreq);
 258         nfs_readdata_release(calldata);
 259 }
 260
 261 static const struct rpc_call_ops nfs_read_direct_ops = {
 262         .rpc_call_done = nfs_direct_read_result,
 263         .rpc_release = nfs_direct_read_release,
 264 };
 265
 266 /*
 267  * For each rsize'd chunk of the user's buffer, dispatch an NFS READ
 268  * operation.  If nfs_readdata_alloc() or get_user_pages() fails,
 269  * bail and stop sending more reads.  Read length accounting is
 270  * handled automatically by nfs_direct_read_result().  Otherwise, if
 271  * no requests have been sent, just return an error.
 272  */
 273 static ssize_t nfs_direct_read_schedule_segment(struct nfs_direct_req *dreq,
 274                                                 const struct iovec *iov,
 275                                                 loff_t pos)
 276 {
 277         struct nfs_open_context *ctx = dreq->ctx;
 278         struct inode *inode = ctx->path.dentry->d_inode;
 279         unsigned long user_addr = (unsigned long)iov->iov_base;
 280         size_t count = iov->iov_len;
 281         size_t rsize = NFS_SERVER(inode)->rsize;
 282         struct rpc_task *task;
 283         struct rpc_message msg = {
 284                 .rpc_cred = ctx->cred,
 285         };
 286         struct rpc_task_setup task_setup_data = {
 287                 .rpc_client = NFS_CLIENT(inode),
 288                 .rpc_message = &msg,
 289                 .callback_ops = &nfs_read_direct_ops,
 290                 .workqueue = nfsiod_workqueue,
 291                 .flags = RPC_TASK_ASYNC,
 292         };
 293         unsigned int pgbase;
 294         int result;
 295         ssize_t started = 0;
 296
 297         do {
 298                 struct nfs_read_data *data;
 299                 size_t bytes;
 300
 301                 pgbase = user_addr & ~PAGE_MASK;
 302                 bytes = min(rsize,count);
 303
 304                 result = -ENOMEM;
 305                 data = nfs_readdata_alloc(nfs_page_array_len(pgbase, bytes));
 306                 if (unlikely(!data))
 307                         break;
 308
 309                 down_read(&current->mm->mmap_sem);
 310                 result = get_user_pages(current, current->mm, user_addr,
 311                                         data->npages, 1, 0, data->pagevec, NULL);
 312                 up_read(&current->mm->mmap_sem);
 313                 if (result < 0) {
 314                         nfs_readdata_release(data);
 315                         break;
 316                 }
 317                 if ((unsigned)result < data->npages) {
 318                         bytes = result * PAGE_SIZE;
 319                         if (bytes <= pgbase) {
 320                                 nfs_direct_release_pages(data->pagevec, result);
 321                                 nfs_readdata_release(data);
 322                                 break;
 323                         }
 324                         bytes -= pgbase;
 325                         data->npages = result;
 326                 }
 327
 328                 get_dreq(dreq);
 329
 330                 data->req = (struct nfs_page *) dreq;
 331                 data->inode = inode;
 332                 data->cred = msg.rpc_cred;
 333                 data->args.fh = NFS_FH(inode);
 334                 data->args.context = get_nfs_open_context(ctx);
 335                 data->args.offset = pos;
 336                 data->args.pgbase = pgbase;
 337                 data->args.pages = data->pagevec;
 338                 data->args.count = bytes;
 339                 data->res.fattr = &data->fattr;
 340                 data->res.eof = 0;
 341                 data->res.count = bytes;
 342                 msg.rpc_argp = &data->args;
 343                 msg.rpc_resp = &data->res;
 344
 345                 task_setup_data.task = &data->task;
 346                 task_setup_data.callback_data = data;
 347                 NFS_PROTO(inode)->read_setup(data, &msg);
 348
 349                 task = rpc_run_task(&task_setup_data);
 350                 if (!IS_ERR(task))
 351                         rpc_put_task(task);
 352
 353                 dprintk("NFS: %5u initiated direct read call "
 354                         "(req %s/%Ld, %zu bytes @ offset %Lu)\n",
 355                                 data->task.tk_pid,
 356                                 inode->i_sb->s_id,
 357                                 (long long)NFS_FILEID(inode),
 358                                 bytes,
 359                                 (unsigned long long)data->args.offset);
 360
 361                 started += bytes;
 362                 user_addr += bytes;
 363                 pos += bytes;
 364                 /* FIXME: Remove this unnecessary math from final patch */
 365                 pgbase += bytes;
 366                 pgbase &= ~PAGE_MASK;
 367                 BUG_ON(pgbase != (user_addr & ~PAGE_MASK));
 368
 369                 count -= bytes;
 370         } while (count != 0);
 371
 372         if (started)
 373                 return started;
 374         return result < 0 ? (ssize_t) result : -EFAULT;
 375 }
 376
 377 static ssize_t nfs_direct_read_schedule_iovec(struct nfs_direct_req *dreq,
 378                                               const struct iovec *iov,
 379                                               unsigned long nr_segs,
 380                                               loff_t pos)
 381 {
 382         ssize_t result = -EINVAL;
 383         size_t requested_bytes = 0;
 384         unsigned long seg;
 385
 386         get_dreq(dreq);
 387
 388         for (seg = 0; seg < nr_segs; seg++) {
 389                 const struct iovec *vec = &iov[seg];
 390                 result = nfs_direct_read_schedule_segment(dreq, vec, pos);
 391                 if (result < 0)
 392                         break;
 393                 requested_bytes += result;
 394                 if ((size_t)result < vec->iov_len)
 395                         break;
 396                 pos += vec->iov_len;
 397         }
 398
 399         if (put_dreq(dreq))
 400                 nfs_direct_complete(dreq);
 401
 402         if (requested_bytes != 0)
 403                 return 0;
 404
 405         if (result < 0)
 406                 return result;
 407         return -EIO;
 408 }
 409
 410 static ssize_t nfs_direct_read(struct kiocb *iocb, const struct iovec *iov,
 411                                unsigned long nr_segs, loff_t pos)
 412 {
 413         ssize_t result = 0;
 414         struct inode *inode = iocb->ki_filp->f_mapping->host;
 415         struct nfs_direct_req *dreq;
 416
 417         dreq = nfs_direct_req_alloc();
 418         if (!dreq)
 419                 return -ENOMEM;
 420
 421         dreq->inode = inode;
 422         dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp));
 423         if (!is_sync_kiocb(iocb))
 424                 dreq->iocb = iocb;
 425
 426         result = nfs_direct_read_schedule_iovec(dreq, iov, nr_segs, pos);
 427         if (!result)
 428                 result = nfs_direct_wait(dreq);
 429         nfs_direct_req_release(dreq);
 430
 431         return result;
 432 }
 433
 434 static void nfs_direct_free_writedata(struct nfs_direct_req *dreq)
 435 {
 436         while (!list_empty(&dreq->rewrite_list)) {
 437                 struct nfs_write_data *data = list_entry(dreq->rewrite_list.next, struct nfs_write_data, pages);
 438                 list_del(&data->pages);
 439                 nfs_direct_release_pages(data->pagevec, data->npages);
 440                 nfs_writedata_release(data);
 441         }
 442 }
 443
 444 #if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
 445 static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq)
 446 {
 447         struct inode *inode = dreq->inode;
 448         struct list_head *p;
 449         struct nfs_write_data *data;
 450         struct rpc_task *task;
 451         struct rpc_message msg = {
 452                 .rpc_cred = dreq->ctx->cred,
 453         };
 454         struct rpc_task_setup task_setup_data = {
 455                 .rpc_client = NFS_CLIENT(inode),
 456                 .callback_ops = &nfs_write_direct_ops,
 457                 .workqueue = nfsiod_workqueue,
 458                 .flags = RPC_TASK_ASYNC,
 459         };
 460
 461         dreq->count = 0;
 462         get_dreq(dreq);
 463
 464         list_for_each(p, &dreq->rewrite_list) {
 465                 data = list_entry(p, struct nfs_write_data, pages);
 466
 467                 get_dreq(dreq);
 468
 469                 /* Use stable writes */
 470                 data->args.stable = NFS_FILE_SYNC;
 471
 472                 /*
 473                  * Reset data->res.
 474                  */
 475                 nfs_fattr_init(&data->fattr);
 476                 data->res.count = data->args.count;
 477                 memset(&data->verf, 0, sizeof(data->verf));
 478
 479                 /*
 480                  * Reuse data->task; data->args should not have changed
 481                  * since the original request was sent.
 482                  */
 483                 task_setup_data.task = &data->task;
 484                 task_setup_data.callback_data = data;
 485                 msg.rpc_argp = &data->args;
 486                 msg.rpc_resp = &data->res;
 487                 NFS_PROTO(inode)->write_setup(data, &msg);
 488
 489                 /*
 490                  * We're called via an RPC callback, so BKL is already held.
 491                  */
 492                 task = rpc_run_task(&task_setup_data);
 493                 if (!IS_ERR(task))
 494                         rpc_put_task(task);
 495
 496                 dprintk("NFS: %5u rescheduled direct write call (req %s/%Ld, %u bytes @ offset %Lu)\n",
 497                                 data->task.tk_pid,
 498                                 inode->i_sb->s_id,
 499                                 (long long)NFS_FILEID(inode),
 500                                 data->args.count,
 501                                 (unsigned long long)data->args.offset);
 502         }
 503
 504         if (put_dreq(dreq))
 505                 nfs_direct_write_complete(dreq, inode);
 506 }
 507
 508 static void nfs_direct_commit_result(struct rpc_task *task, void *calldata)
 509 {
 510         struct nfs_write_data *data = calldata;
 511
 512         /* Call the NFS version-specific code */
 513         NFS_PROTO(data->inode)->commit_done(task, data);
 514 }
 515
 516 static void nfs_direct_commit_release(void *calldata)
 517 {
 518         struct nfs_write_data *data = calldata;
 519         struct nfs_direct_req *dreq = (struct nfs_direct_req *) data->req;
 520         int status = data->task.tk_status;
 521
 522         if (status < 0) {
 523                 dprintk("NFS: %5u commit failed with error %d.\n",
 524                                 data->task.tk_pid, status);
 525                 dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
 526         } else if (memcmp(&dreq->verf, &data->verf, sizeof(data->verf))) {
 527                 dprintk("NFS: %5u commit verify failed\n", data->task.tk_pid);
 528                 dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
 529         }
 530
 531         dprintk("NFS: %5u commit returned %d\n", data->task.tk_pid, status);
 532         nfs_direct_write_complete(dreq, data->inode);
 533         nfs_commitdata_release(calldata);
 534 }
 535
 536 static const struct rpc_call_ops nfs_commit_direct_ops = {
 537         .rpc_call_done = nfs_direct_commit_result,
 538         .rpc_release = nfs_direct_commit_release,
 539 };
 540
 541 static void nfs_direct_commit_schedule(struct nfs_direct_req *dreq)
 542 {
 543         struct nfs_write_data *data = dreq->commit_data;
 544         struct rpc_task *task;
 545         struct rpc_message msg = {
 546                 .rpc_argp = &data->args,
 547                 .rpc_resp = &data->res,
 548                 .rpc_cred = dreq->ctx->cred,
 549         };
 550         struct rpc_task_setup task_setup_data = {
 551                 .task = &data->task,
 552                 .rpc_client = NFS_CLIENT(dreq->inode),
 553                 .rpc_message = &msg,
 554                 .callback_ops = &nfs_commit_direct_ops,
 555                 .callback_data = data,
 556                 .workqueue = nfsiod_workqueue,
 557                 .flags = RPC_TASK_ASYNC,
 558         };
 559
 560         data->inode = dreq->inode;
 561         data->cred = msg.rpc_cred;
 562
 563         data->args.fh = NFS_FH(data->inode);
 564         data->args.offset = 0;
 565         data->args.count = 0;
 566         data->args.context = get_nfs_open_context(dreq->ctx);
 567         data->res.count = 0;
 568         data->res.fattr = &data->fattr;
 569         data->res.verf = &data->verf;
 570
 571         NFS_PROTO(data->inode)->commit_setup(data, &msg);
 572
 573         /* Note: task.tk_ops->rpc_release will free dreq->commit_data */
 574         dreq->commit_data = NULL;
 575
 576         dprintk("NFS: %5u initiated commit call\n", data->task.tk_pid);
 577
 578         task = rpc_run_task(&task_setup_data);
 579         if (!IS_ERR(task))
 580                 rpc_put_task(task);
 581 }
 582
 583 static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode *inode)
 584 {
 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                         if (dreq->commit_data != NULL)
 597                                 nfs_commit_free(dreq->commit_data);
 598                         nfs_direct_free_writedata(dreq);
 599                         nfs_zap_mapping(inode, inode->i_mapping);
 600                         nfs_direct_complete(dreq);
 601         }
 602 }
 603
 604 static void nfs_alloc_commit_data(struct nfs_direct_req *dreq)
 605 {
 606         dreq->commit_data = nfs_commitdata_alloc();
 607         if (dreq->commit_data != NULL)
 608                 dreq->commit_data->req = (struct nfs_page *) dreq;
 609 }
 610 #else
 611 static inline void nfs_alloc_commit_data(struct nfs_direct_req *dreq)
 612 {
 613         dreq->commit_data = NULL;
 614 }
 615
 616 static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode *inode)
 617 {
 618         nfs_direct_free_writedata(dreq);
 619         nfs_zap_mapping(inode, inode->i_mapping);
 620         nfs_direct_complete(dreq);
 621 }
 622 #endif
 623
 624 static void nfs_direct_write_result(struct rpc_task *task, void *calldata)
 625 {
 626         struct nfs_write_data *data = calldata;
 627
 628         if (nfs_writeback_done(task, data) != 0)
 629                 return;
 630 }
 631
 632 /*
 633  * NB: Return the value of the first error return code.  Subsequent
 634  *     errors after the first one are ignored.
 635  */
 636 static void nfs_direct_write_release(void *calldata)
 637 {
 638         struct nfs_write_data *data = calldata;
 639         struct nfs_direct_req *dreq = (struct nfs_direct_req *) data->req;
 640         int status = data->task.tk_status;
 641
 642         spin_lock(&dreq->lock);
 643
 644         if (unlikely(status < 0)) {
 645                 /* An error has occurred, so we should not commit */
 646                 dreq->flags = 0;
 647                 dreq->error = status;
 648         }
 649         if (unlikely(dreq->error != 0))
 650                 goto out_unlock;
 651
 652         dreq->count += data->res.count;
 653
 654         if (data->res.verf->committed != NFS_FILE_SYNC) {
 655                 switch (dreq->flags) {
 656                         case 0:
 657                                 memcpy(&dreq->verf, &data->verf, sizeof(dreq->verf));
 658                                 dreq->flags = NFS_ODIRECT_DO_COMMIT;
 659                                 break;
 660                         case NFS_ODIRECT_DO_COMMIT:
 661                                 if (memcmp(&dreq->verf, &data->verf, sizeof(dreq->verf))) {
 662                                         dprintk("NFS: %5u write verify failed\n", data->task.tk_pid);
 663                                         dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
 664                                 }
 665                 }
 666         }
 667 out_unlock:
 668         spin_unlock(&dreq->lock);
 669
 670         if (put_dreq(dreq))
 671                 nfs_direct_write_complete(dreq, data->inode);
 672 }
 673
 674 static const struct rpc_call_ops nfs_write_direct_ops = {
 675         .rpc_call_done = nfs_direct_write_result,
 676         .rpc_release = nfs_direct_write_release,
 677 };
 678
 679 /*
 680  * For each wsize'd chunk of the user's buffer, dispatch an NFS WRITE
 681  * operation.  If nfs_writedata_alloc() or get_user_pages() fails,
 682  * bail and stop sending more writes.  Write length accounting is
 683  * handled automatically by nfs_direct_write_result().  Otherwise, if
 684  * no requests have been sent, just return an error.
 685  */
 686 static ssize_t nfs_direct_write_schedule_segment(struct nfs_direct_req *dreq,
 687                                                  const struct iovec *iov,
 688                                                  loff_t pos, int sync)
 689 {
 690         struct nfs_open_context *ctx = dreq->ctx;
 691         struct inode *inode = ctx->path.dentry->d_inode;
 692         unsigned long user_addr = (unsigned long)iov->iov_base;
 693         size_t count = iov->iov_len;
 694         struct rpc_task *task;
 695         struct rpc_message msg = {
 696                 .rpc_cred = ctx->cred,
 697         };
 698         struct rpc_task_setup task_setup_data = {
 699                 .rpc_client = NFS_CLIENT(inode),
 700                 .rpc_message = &msg,
 701                 .callback_ops = &nfs_write_direct_ops,
 702                 .workqueue = nfsiod_workqueue,
 703                 .flags = RPC_TASK_ASYNC,
 704         };
 705         size_t wsize = NFS_SERVER(inode)->wsize;
 706         unsigned int pgbase;
 707         int result;
 708         ssize_t started = 0;
 709
 710         do {
 711                 struct nfs_write_data *data;
 712                 size_t bytes;
 713
 714                 pgbase = user_addr & ~PAGE_MASK;
 715                 bytes = min(wsize,count);
 716
 717                 result = -ENOMEM;
 718                 data = nfs_writedata_alloc(nfs_page_array_len(pgbase, bytes));
 719                 if (unlikely(!data))
 720                         break;
 721
 722                 down_read(&current->mm->mmap_sem);
 723                 result = get_user_pages(current, current->mm, user_addr,
 724                                         data->npages, 0, 0, data->pagevec, NULL);
 725                 up_read(&current->mm->mmap_sem);
 726                 if (result < 0) {
 727                         nfs_writedata_release(data);
 728                         break;
 729                 }
 730                 if ((unsigned)result < data->npages) {
 731                         bytes = result * PAGE_SIZE;
 732                         if (bytes <= pgbase) {
 733                                 nfs_direct_release_pages(data->pagevec, result);
 734                                 nfs_writedata_release(data);
 735                                 break;
 736                         }
 737                         bytes -= pgbase;
 738                         data->npages = result;
 739                 }
 740
 741                 get_dreq(dreq);
 742
 743                 list_move_tail(&data->pages, &dreq->rewrite_list);
 744
 745                 data->req = (struct nfs_page *) dreq;
 746                 data->inode = inode;
 747                 data->cred = msg.rpc_cred;
 748                 data->args.fh = NFS_FH(inode);
 749                 data->args.context = get_nfs_open_context(ctx);
 750                 data->args.offset = pos;
 751                 data->args.pgbase = pgbase;
 752                 data->args.pages = data->pagevec;
 753                 data->args.count = bytes;
 754                 data->args.stable = sync;
 755                 data->res.fattr = &data->fattr;
 756                 data->res.count = bytes;
 757                 data->res.verf = &data->verf;
 758
 759                 task_setup_data.task = &data->task;
 760                 task_setup_data.callback_data = data;
 761                 msg.rpc_argp = &data->args;
 762                 msg.rpc_resp = &data->res;
 763                 NFS_PROTO(inode)->write_setup(data, &msg);
 764
 765                 task = rpc_run_task(&task_setup_data);
 766                 if (!IS_ERR(task))
 767                         rpc_put_task(task);
 768
 769                 dprintk("NFS: %5u initiated direct write call "
 770                         "(req %s/%Ld, %zu bytes @ offset %Lu)\n",
 771                                 data->task.tk_pid,
 772                                 inode->i_sb->s_id,
 773                                 (long long)NFS_FILEID(inode),
 774                                 bytes,
 775                                 (unsigned long long)data->args.offset);
 776
 777                 started += bytes;
 778                 user_addr += bytes;
 779                 pos += bytes;
 780
 781                 /* FIXME: Remove this useless math from the final patch */
 782                 pgbase += bytes;
 783                 pgbase &= ~PAGE_MASK;
 784                 BUG_ON(pgbase != (user_addr & ~PAGE_MASK));
 785
 786                 count -= bytes;
 787         } while (count != 0);
 788
 789         if (started)
 790                 return started;
 791         return result < 0 ? (ssize_t) result : -EFAULT;
 792 }
 793
 794 static ssize_t nfs_direct_write_schedule_iovec(struct nfs_direct_req *dreq,
 795                                                const struct iovec *iov,
 796                                                unsigned long nr_segs,
 797                                                loff_t pos, int sync)
 798 {
 799         ssize_t result = 0;
 800         size_t requested_bytes = 0;
 801         unsigned long seg;
 802
 803         get_dreq(dreq);
 804
 805         for (seg = 0; seg < nr_segs; seg++) {
 806                 const struct iovec *vec = &iov[seg];
 807                 result = nfs_direct_write_schedule_segment(dreq, vec,
 808                                                            pos, sync);
 809                 if (result < 0)
 810                         break;
 811                 requested_bytes += result;
 812                 if ((size_t)result < vec->iov_len)
 813                         break;
 814                 pos += vec->iov_len;
 815         }
 816
 817         if (put_dreq(dreq))
 818                 nfs_direct_write_complete(dreq, dreq->inode);
 819
 820         if (requested_bytes != 0)
 821                 return 0;
 822
 823         if (result < 0)
 824                 return result;
 825         return -EIO;
 826 }
 827
 828 static ssize_t nfs_direct_write(struct kiocb *iocb, const struct iovec *iov,
 829                                 unsigned long nr_segs, loff_t pos,
 830                                 size_t count)
 831 {
 832         ssize_t result = 0;
 833         struct inode *inode = iocb->ki_filp->f_mapping->host;
 834         struct nfs_direct_req *dreq;
 835         size_t wsize = NFS_SERVER(inode)->wsize;
 836         int sync = NFS_UNSTABLE;
 837
 838         dreq = nfs_direct_req_alloc();
 839         if (!dreq)
 840                 return -ENOMEM;
 841         nfs_alloc_commit_data(dreq);
 842
 843         if (dreq->commit_data == NULL || count < wsize)
 844                 sync = NFS_FILE_SYNC;
 845
 846         dreq->inode = inode;
 847         dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp));
 848         if (!is_sync_kiocb(iocb))
 849                 dreq->iocb = iocb;
 850
 851         result = nfs_direct_write_schedule_iovec(dreq, iov, nr_segs, pos, sync);
 852         if (!result)
 853                 result = nfs_direct_wait(dreq);
 854         nfs_direct_req_release(dreq);
 855
 856         return result;
 857 }
 858
 859 /**
 860  * nfs_file_direct_read - file direct read operation for NFS files
 861  * @iocb: target I/O control block
 862  * @iov: vector of user buffers into which to read data
 863  * @nr_segs: size of iov vector
 864  * @pos: byte offset in file where reading starts
 865  *
 866  * We use this function for direct reads instead of calling
 867  * generic_file_aio_read() in order to avoid gfar's check to see if
 868  * the request starts before the end of the file.  For that check
 869  * to work, we must generate a GETATTR before each direct read, and
 870  * even then there is a window between the GETATTR and the subsequent
 871  * READ where the file size could change.  Our preference is simply
 872  * to do all reads the application wants, and the server will take
 873  * care of managing the end of file boundary.
 874  *
 875  * This function also eliminates unnecessarily updating the file's
 876  * atime locally, as the NFS server sets the file's atime, and this
 877  * client must read the updated atime from the server back into its
 878  * cache.
 879  */
 880 ssize_t nfs_file_direct_read(struct kiocb *iocb, const struct iovec *iov,
 881                                 unsigned long nr_segs, loff_t pos)
 882 {
 883         ssize_t retval = -EINVAL;
 884         struct file *file = iocb->ki_filp;
 885         struct address_space *mapping = file->f_mapping;
 886         size_t count;
 887
 888         count = iov_length(iov, nr_segs);
 889         nfs_add_stats(mapping->host, NFSIOS_DIRECTREADBYTES, count);
 890
 891         dprintk("nfs: direct read(%s/%s, %zd@%Ld)\n",
 892                 file->f_path.dentry->d_parent->d_name.name,
 893                 file->f_path.dentry->d_name.name,
 894                 count, (long long) pos);
 895
 896         retval = 0;
 897         if (!count)
 898                 goto out;
 899
 900         retval = nfs_sync_mapping(mapping);
 901         if (retval)
 902                 goto out;
 903
 904         retval = nfs_direct_read(iocb, iov, nr_segs, pos);
 905         if (retval > 0)
 906                 iocb->ki_pos = pos + retval;
 907
 908 out:
 909         return retval;
 910 }
 911
 912 /**
 913  * nfs_file_direct_write - file direct write operation for NFS files
 914  * @iocb: target I/O control block
 915  * @iov: vector of user buffers from which to write data
 916  * @nr_segs: size of iov vector
 917  * @pos: byte offset in file where writing starts
 918  *
 919  * We use this function for direct writes instead of calling
 920  * generic_file_aio_write() in order to avoid taking the inode
 921  * semaphore and updating the i_size.  The NFS server will set
 922  * the new i_size and this client must read the updated size
 923  * back into its cache.  We let the server do generic write
 924  * parameter checking and report problems.
 925  *
 926  * We also avoid an unnecessary invocation of generic_osync_inode(),
 927  * as it is fairly meaningless to sync the metadata of an NFS file.
 928  *
 929  * We eliminate local atime updates, see direct read above.
 930  *
 931  * We avoid unnecessary page cache invalidations for normal cached
 932  * readers of this file.
 933  *
 934  * Note that O_APPEND is not supported for NFS direct writes, as there
 935  * is no atomic O_APPEND write facility in the NFS protocol.
 936  */
 937 ssize_t nfs_file_direct_write(struct kiocb *iocb, const struct iovec *iov,
 938                                 unsigned long nr_segs, loff_t pos)
 939 {
 940         ssize_t retval = -EINVAL;
 941         struct file *file = iocb->ki_filp;
 942         struct address_space *mapping = file->f_mapping;
 943         size_t count;
 944
 945         count = iov_length(iov, nr_segs);
 946         nfs_add_stats(mapping->host, NFSIOS_DIRECTWRITTENBYTES, count);
 947
 948         dfprintk(VFS, "nfs: direct write(%s/%s, %zd@%Ld)\n",
 949                 file->f_path.dentry->d_parent->d_name.name,
 950                 file->f_path.dentry->d_name.name,
 951                 count, (long long) pos);
 952
 953         retval = generic_write_checks(file, &pos, &count, 0);
 954         if (retval)
 955                 goto out;
 956
 957         retval = -EINVAL;
 958         if ((ssize_t) count < 0)
 959                 goto out;
 960         retval = 0;
 961         if (!count)
 962                 goto out;
 963
 964         retval = nfs_sync_mapping(mapping);
 965         if (retval)
 966                 goto out;
 967
 968         retval = nfs_direct_write(iocb, iov, nr_segs, pos, count);
 969
 970         if (retval > 0)
 971                 iocb->ki_pos = pos + retval;
 972
 973 out:
 974         return retval;
 975 }
 976
 977 /**
 978  * nfs_init_directcache - create a slab cache for nfs_direct_req structures
 979  *
 980  */
 981 int __init nfs_init_directcache(void)
 982 {
 983         nfs_direct_cachep = kmem_cache_create("nfs_direct_cache",
 984                                                 sizeof(struct nfs_direct_req),
 985                                                 0, (SLAB_RECLAIM_ACCOUNT|
 986                                                         SLAB_MEM_SPREAD),
 987                                                 NULL);
 988         if (nfs_direct_cachep == NULL)
 989                 return -ENOMEM;
 990
 991         return 0;
 992 }
 993
 994 /**
 995  * nfs_destroy_directcache - destroy the slab cache for nfs_direct_req structures
 996  *
 997  */
 998 void nfs_destroy_directcache(void)
 999 {
1000         kmem_cache_destroy(nfs_direct_cachep);
1001 }