module/zfs/zpl_file.c

   1 /*
   2  * CDDL HEADER START
   3  *
   4  * The contents of this file are subject to the terms of the
   5  * Common Development and Distribution License (the "License").
   6  * You may not use this file except in compliance with the License.
   7  *
   8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
   9  * or http://www.opensolaris.org/os/licensing.
  10  * See the License for the specific language governing permissions
  11  * and limitations under the License.
  12  *
  13  * When distributing Covered Code, include this CDDL HEADER in each
  14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
  15  * If applicable, add the following below this CDDL HEADER, with the
  16  * fields enclosed by brackets "[]" replaced with your own identifying
  17  * information: Portions Copyright [yyyy] [name of copyright owner]
  18  *
  19  * CDDL HEADER END
  20  */
  21 /*
  22  * Copyright (c) 2011, Lawrence Livermore National Security, LLC.
  23  */
  24
  25
  26 #include <sys/dmu_objset.h>
  27 #include <sys/zfs_vfsops.h>
  28 #include <sys/zfs_vnops.h>
  29 #include <sys/zfs_znode.h>
  30 #include <sys/zpl.h>
  31
  32
  33 static int
  34 zpl_open(struct inode *ip, struct file *filp)
  35 {
  36         cred_t *cr = CRED();
  37         int error;
  38         fstrans_cookie_t cookie;
  39
  40         error = generic_file_open(ip, filp);
  41         if (error)
  42                 return (error);
  43
  44         crhold(cr);
  45         cookie = spl_fstrans_mark();
  46         error = -zfs_open(ip, filp->f_mode, filp->f_flags, cr);
  47         spl_fstrans_unmark(cookie);
  48         crfree(cr);
  49         ASSERT3S(error, <=, 0);
  50
  51         return (error);
  52 }
  53
  54 static int
  55 zpl_release(struct inode *ip, struct file *filp)
  56 {
  57         cred_t *cr = CRED();
  58         int error;
  59         fstrans_cookie_t cookie;
  60
  61         cookie = spl_fstrans_mark();
  62         if (ITOZ(ip)->z_atime_dirty)
  63                 zfs_mark_inode_dirty(ip);
  64
  65         crhold(cr);
  66         error = -zfs_close(ip, filp->f_flags, cr);
  67         spl_fstrans_unmark(cookie);
  68         crfree(cr);
  69         ASSERT3S(error, <=, 0);
  70
  71         return (error);
  72 }
  73
  74 static int
  75 zpl_iterate(struct file *filp, struct dir_context *ctx)
  76 {
  77         struct dentry *dentry = filp->f_path.dentry;
  78         cred_t *cr = CRED();
  79         int error;
  80         fstrans_cookie_t cookie;
  81
  82         crhold(cr);
  83         cookie = spl_fstrans_mark();
  84         error = -zfs_readdir(dentry->d_inode, ctx, cr);
  85         spl_fstrans_unmark(cookie);
  86         crfree(cr);
  87         ASSERT3S(error, <=, 0);
  88
  89         return (error);
  90 }
  91
  92 #if !defined(HAVE_VFS_ITERATE)
  93 static int
  94 zpl_readdir(struct file *filp, void *dirent, filldir_t filldir)
  95 {
  96         struct dir_context ctx = DIR_CONTEXT_INIT(dirent, filldir, filp->f_pos);
  97         int error;
  98
  99         error = zpl_iterate(filp, &ctx);
 100         filp->f_pos = ctx.pos;
 101
 102         return (error);
 103 }
 104 #endif /* HAVE_VFS_ITERATE */
 105
 106 #if defined(HAVE_FSYNC_WITH_DENTRY)
 107 /*
 108  * Linux 2.6.x - 2.6.34 API,
 109  * Through 2.6.34 the nfsd kernel server would pass a NULL 'file struct *'
 110  * to the fops->fsync() hook.  For this reason, we must be careful not to
 111  * use filp unconditionally.
 112  */
 113 static int
 114 zpl_fsync(struct file *filp, struct dentry *dentry, int datasync)
 115 {
 116         cred_t *cr = CRED();
 117         int error;
 118         fstrans_cookie_t cookie;
 119
 120         crhold(cr);
 121         cookie = spl_fstrans_mark();
 122         error = -zfs_fsync(dentry->d_inode, datasync, cr);
 123         spl_fstrans_unmark(cookie);
 124         crfree(cr);
 125         ASSERT3S(error, <=, 0);
 126
 127         return (error);
 128 }
 129
 130 static int
 131 zpl_aio_fsync(struct kiocb *kiocb, int datasync)
 132 {
 133         struct file *filp = kiocb->ki_filp;
 134         return (zpl_fsync(filp, filp->f_path.dentry, datasync));
 135 }
 136 #elif defined(HAVE_FSYNC_WITHOUT_DENTRY)
 137 /*
 138  * Linux 2.6.35 - 3.0 API,
 139  * As of 2.6.35 the dentry argument to the fops->fsync() hook was deemed
 140  * redundant.  The dentry is still accessible via filp->f_path.dentry,
 141  * and we are guaranteed that filp will never be NULL.
 142  */
 143 static int
 144 zpl_fsync(struct file *filp, int datasync)
 145 {
 146         struct inode *inode = filp->f_mapping->host;
 147         cred_t *cr = CRED();
 148         int error;
 149         fstrans_cookie_t cookie;
 150
 151         crhold(cr);
 152         cookie = spl_fstrans_mark();
 153         error = -zfs_fsync(inode, datasync, cr);
 154         spl_fstrans_unmark(cookie);
 155         crfree(cr);
 156         ASSERT3S(error, <=, 0);
 157
 158         return (error);
 159 }
 160
 161 static int
 162 zpl_aio_fsync(struct kiocb *kiocb, int datasync)
 163 {
 164         return (zpl_fsync(kiocb->ki_filp, datasync));
 165 }
 166 #elif defined(HAVE_FSYNC_RANGE)
 167 /*
 168  * Linux 3.1 - 3.x API,
 169  * As of 3.1 the responsibility to call filemap_write_and_wait_range() has
 170  * been pushed down in to the .fsync() vfs hook.  Additionally, the i_mutex
 171  * lock is no longer held by the caller, for zfs we don't require the lock
 172  * to be held so we don't acquire it.
 173  */
 174 static int
 175 zpl_fsync(struct file *filp, loff_t start, loff_t end, int datasync)
 176 {
 177         struct inode *inode = filp->f_mapping->host;
 178         cred_t *cr = CRED();
 179         int error;
 180         fstrans_cookie_t cookie;
 181
 182         error = filemap_write_and_wait_range(inode->i_mapping, start, end);
 183         if (error)
 184                 return (error);
 185
 186         crhold(cr);
 187         cookie = spl_fstrans_mark();
 188         error = -zfs_fsync(inode, datasync, cr);
 189         spl_fstrans_unmark(cookie);
 190         crfree(cr);
 191         ASSERT3S(error, <=, 0);
 192
 193         return (error);
 194 }
 195
 196 static int
 197 zpl_aio_fsync(struct kiocb *kiocb, int datasync)
 198 {
 199         return (zpl_fsync(kiocb->ki_filp, kiocb->ki_pos,
 200             kiocb->ki_pos + kiocb->ki_nbytes, datasync));
 201 }
 202 #else
 203 #error "Unsupported fops->fsync() implementation"
 204 #endif
 205
 206 static inline ssize_t
 207 zpl_read_common_iovec(struct inode *ip, const struct iovec *iovp, size_t count,
 208     unsigned long nr_segs, loff_t *ppos, uio_seg_t segment,
 209     int flags, cred_t *cr)
 210 {
 211         ssize_t read;
 212         uio_t uio;
 213         int error;
 214         fstrans_cookie_t cookie;
 215
 216         uio.uio_iov = (struct iovec *)iovp;
 217         uio.uio_resid = count;
 218         uio.uio_iovcnt = nr_segs;
 219         uio.uio_loffset = *ppos;
 220         uio.uio_limit = MAXOFFSET_T;
 221         uio.uio_segflg = segment;
 222
 223         cookie = spl_fstrans_mark();
 224         error = -zfs_read(ip, &uio, flags, cr);
 225         spl_fstrans_unmark(cookie);
 226         if (error < 0)
 227                 return (error);
 228
 229         read = count - uio.uio_resid;
 230         *ppos += read;
 231         task_io_account_read(read);
 232
 233         return (read);
 234 }
 235
 236 inline ssize_t
 237 zpl_read_common(struct inode *ip, const char *buf, size_t len, loff_t *ppos,
 238     uio_seg_t segment, int flags, cred_t *cr)
 239 {
 240         struct iovec iov;
 241
 242         iov.iov_base = (void *)buf;
 243         iov.iov_len = len;
 244
 245         return (zpl_read_common_iovec(ip, &iov, len, 1, ppos, segment,
 246             flags, cr));
 247 }
 248
 249 static ssize_t
 250 zpl_read(struct file *filp, char __user *buf, size_t len, loff_t *ppos)
 251 {
 252         cred_t *cr = CRED();
 253         ssize_t read;
 254
 255         crhold(cr);
 256         read = zpl_read_common(filp->f_mapping->host, buf, len, ppos,
 257             UIO_USERSPACE, filp->f_flags, cr);
 258         crfree(cr);
 259
 260         return (read);
 261 }
 262
 263 static ssize_t
 264 zpl_aio_read(struct kiocb *kiocb, const struct iovec *iovp,
 265         unsigned long nr_segs, loff_t pos)
 266 {
 267         cred_t *cr = CRED();
 268         struct file *filp = kiocb->ki_filp;
 269         size_t count = kiocb->ki_nbytes;
 270         ssize_t read;
 271         size_t alloc_size = sizeof (struct iovec) * nr_segs;
 272         struct iovec *iov_tmp = kmem_alloc(alloc_size, KM_SLEEP);
 273         bcopy(iovp, iov_tmp, alloc_size);
 274
 275         ASSERT(iovp);
 276
 277         crhold(cr);
 278         read = zpl_read_common_iovec(filp->f_mapping->host, iov_tmp, count,
 279             nr_segs, &kiocb->ki_pos, UIO_USERSPACE, filp->f_flags, cr);
 280         crfree(cr);
 281
 282         kmem_free(iov_tmp, alloc_size);
 283
 284         return (read);
 285 }
 286
 287 static inline ssize_t
 288 zpl_write_common_iovec(struct inode *ip, const struct iovec *iovp, size_t count,
 289     unsigned long nr_segs, loff_t *ppos, uio_seg_t segment,
 290     int flags, cred_t *cr)
 291 {
 292         ssize_t wrote;
 293         uio_t uio;
 294         int error;
 295         fstrans_cookie_t cookie;
 296
 297         if (flags & O_APPEND)
 298                 *ppos = i_size_read(ip);
 299
 300         uio.uio_iov = (struct iovec *)iovp;
 301         uio.uio_resid = count;
 302         uio.uio_iovcnt = nr_segs;
 303         uio.uio_loffset = *ppos;
 304         uio.uio_limit = MAXOFFSET_T;
 305         uio.uio_segflg = segment;
 306
 307         cookie = spl_fstrans_mark();
 308         error = -zfs_write(ip, &uio, flags, cr);
 309         spl_fstrans_unmark(cookie);
 310         if (error < 0)
 311                 return (error);
 312
 313         wrote = count - uio.uio_resid;
 314         *ppos += wrote;
 315         task_io_account_write(wrote);
 316
 317         return (wrote);
 318 }
 319 inline ssize_t
 320 zpl_write_common(struct inode *ip, const char *buf, size_t len, loff_t *ppos,
 321     uio_seg_t segment, int flags, cred_t *cr)
 322 {
 323         struct iovec iov;
 324
 325         iov.iov_base = (void *)buf;
 326         iov.iov_len = len;
 327
 328         return (zpl_write_common_iovec(ip, &iov, len, 1, ppos, segment,
 329             flags, cr));
 330 }
 331
 332 static ssize_t
 333 zpl_write(struct file *filp, const char __user *buf, size_t len, loff_t *ppos)
 334 {
 335         cred_t *cr = CRED();
 336         ssize_t wrote;
 337
 338         crhold(cr);
 339         wrote = zpl_write_common(filp->f_mapping->host, buf, len, ppos,
 340             UIO_USERSPACE, filp->f_flags, cr);
 341         crfree(cr);
 342
 343         return (wrote);
 344 }
 345
 346 static ssize_t
 347 zpl_aio_write(struct kiocb *kiocb, const struct iovec *iovp,
 348         unsigned long nr_segs, loff_t pos)
 349 {
 350         cred_t *cr = CRED();
 351         struct file *filp = kiocb->ki_filp;
 352         size_t count = kiocb->ki_nbytes;
 353         ssize_t wrote;
 354         size_t alloc_size = sizeof (struct iovec) * nr_segs;
 355         struct iovec *iov_tmp = kmem_alloc(alloc_size, KM_SLEEP);
 356         bcopy(iovp, iov_tmp, alloc_size);
 357
 358         ASSERT(iovp);
 359
 360         crhold(cr);
 361         wrote = zpl_write_common_iovec(filp->f_mapping->host, iov_tmp, count,
 362             nr_segs, &kiocb->ki_pos, UIO_USERSPACE, filp->f_flags, cr);
 363         crfree(cr);
 364
 365         kmem_free(iov_tmp, alloc_size);
 366
 367         return (wrote);
 368 }
 369
 370 static loff_t
 371 zpl_llseek(struct file *filp, loff_t offset, int whence)
 372 {
 373 #if defined(SEEK_HOLE) && defined(SEEK_DATA)
 374         fstrans_cookie_t cookie;
 375
 376         if (whence == SEEK_DATA || whence == SEEK_HOLE) {
 377                 struct inode *ip = filp->f_mapping->host;
 378                 loff_t maxbytes = ip->i_sb->s_maxbytes;
 379                 loff_t error;
 380
 381                 spl_inode_lock(ip);
 382                 cookie = spl_fstrans_mark();
 383                 error = -zfs_holey(ip, whence, &offset);
 384                 spl_fstrans_unmark(cookie);
 385                 if (error == 0)
 386                         error = lseek_execute(filp, ip, offset, maxbytes);
 387                 spl_inode_unlock(ip);
 388
 389                 return (error);
 390         }
 391 #endif /* SEEK_HOLE && SEEK_DATA */
 392
 393         return (generic_file_llseek(filp, offset, whence));
 394 }
 395
 396 /*
 397  * It's worth taking a moment to describe how mmap is implemented
 398  * for zfs because it differs considerably from other Linux filesystems.
 399  * However, this issue is handled the same way under OpenSolaris.
 400  *
 401  * The issue is that by design zfs bypasses the Linux page cache and
 402  * leaves all caching up to the ARC.  This has been shown to work
 403  * well for the common read(2)/write(2) case.  However, mmap(2)
 404  * is problem because it relies on being tightly integrated with the
 405  * page cache.  To handle this we cache mmap'ed files twice, once in
 406  * the ARC and a second time in the page cache.  The code is careful
 407  * to keep both copies synchronized.
 408  *
 409  * When a file with an mmap'ed region is written to using write(2)
 410  * both the data in the ARC and existing pages in the page cache
 411  * are updated.  For a read(2) data will be read first from the page
 412  * cache then the ARC if needed.  Neither a write(2) or read(2) will
 413  * will ever result in new pages being added to the page cache.
 414  *
 415  * New pages are added to the page cache only via .readpage() which
 416  * is called when the vfs needs to read a page off disk to back the
 417  * virtual memory region.  These pages may be modified without
 418  * notifying the ARC and will be written out periodically via
 419  * .writepage().  This will occur due to either a sync or the usual
 420  * page aging behavior.  Note because a read(2) of a mmap'ed file
 421  * will always check the page cache first even when the ARC is out
 422  * of date correct data will still be returned.
 423  *
 424  * While this implementation ensures correct behavior it does have
 425  * have some drawbacks.  The most obvious of which is that it
 426  * increases the required memory footprint when access mmap'ed
 427  * files.  It also adds additional complexity to the code keeping
 428  * both caches synchronized.
 429  *
 430  * Longer term it may be possible to cleanly resolve this wart by
 431  * mapping page cache pages directly on to the ARC buffers.  The
 432  * Linux address space operations are flexible enough to allow
 433  * selection of which pages back a particular index.  The trick
 434  * would be working out the details of which subsystem is in
 435  * charge, the ARC, the page cache, or both.  It may also prove
 436  * helpful to move the ARC buffers to a scatter-gather lists
 437  * rather than a vmalloc'ed region.
 438  */
 439 static int
 440 zpl_mmap(struct file *filp, struct vm_area_struct *vma)
 441 {
 442         struct inode *ip = filp->f_mapping->host;
 443         znode_t *zp = ITOZ(ip);
 444         int error;
 445         fstrans_cookie_t cookie;
 446
 447         cookie = spl_fstrans_mark();
 448         error = -zfs_map(ip, vma->vm_pgoff, (caddr_t *)vma->vm_start,
 449             (size_t)(vma->vm_end - vma->vm_start), vma->vm_flags);
 450         spl_fstrans_unmark(cookie);
 451         if (error)
 452                 return (error);
 453
 454         error = generic_file_mmap(filp, vma);
 455         if (error)
 456                 return (error);
 457
 458         mutex_enter(&zp->z_lock);
 459         zp->z_is_mapped = 1;
 460         mutex_exit(&zp->z_lock);
 461
 462         return (error);
 463 }
 464
 465 /*
 466  * Populate a page with data for the Linux page cache.  This function is
 467  * only used to support mmap(2).  There will be an identical copy of the
 468  * data in the ARC which is kept up to date via .write() and .writepage().
 469  *
 470  * Current this function relies on zpl_read_common() and the O_DIRECT
 471  * flag to read in a page.  This works but the more correct way is to
 472  * update zfs_fillpage() to be Linux friendly and use that interface.
 473  */
 474 static int
 475 zpl_readpage(struct file *filp, struct page *pp)
 476 {
 477         struct inode *ip;
 478         struct page *pl[1];
 479         int error = 0;
 480         fstrans_cookie_t cookie;
 481
 482         ASSERT(PageLocked(pp));
 483         ip = pp->mapping->host;
 484         pl[0] = pp;
 485
 486         cookie = spl_fstrans_mark();
 487         error = -zfs_getpage(ip, pl, 1);
 488         spl_fstrans_unmark(cookie);
 489
 490         if (error) {
 491                 SetPageError(pp);
 492                 ClearPageUptodate(pp);
 493         } else {
 494                 ClearPageError(pp);
 495                 SetPageUptodate(pp);
 496                 flush_dcache_page(pp);
 497         }
 498
 499         unlock_page(pp);
 500         return (error);
 501 }
 502
 503 /*
 504  * Populate a set of pages with data for the Linux page cache.  This
 505  * function will only be called for read ahead and never for demand
 506  * paging.  For simplicity, the code relies on read_cache_pages() to
 507  * correctly lock each page for IO and call zpl_readpage().
 508  */
 509 static int
 510 zpl_readpages(struct file *filp, struct address_space *mapping,
 511         struct list_head *pages, unsigned nr_pages)
 512 {
 513         return (read_cache_pages(mapping, pages,
 514             (filler_t *)zpl_readpage, filp));
 515 }
 516
 517 int
 518 zpl_putpage(struct page *pp, struct writeback_control *wbc, void *data)
 519 {
 520         struct address_space *mapping = data;
 521         fstrans_cookie_t cookie;
 522
 523         ASSERT(PageLocked(pp));
 524         ASSERT(!PageWriteback(pp));
 525
 526         cookie = spl_fstrans_mark();
 527         (void) zfs_putpage(mapping->host, pp, wbc);
 528         spl_fstrans_unmark(cookie);
 529
 530         return (0);
 531 }
 532
 533 static int
 534 zpl_writepages(struct address_space *mapping, struct writeback_control *wbc)
 535 {
 536         znode_t         *zp = ITOZ(mapping->host);
 537         zfs_sb_t        *zsb = ITOZSB(mapping->host);
 538         enum writeback_sync_modes sync_mode;
 539         int result;
 540
 541         ZFS_ENTER(zsb);
 542         if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS)
 543                 wbc->sync_mode = WB_SYNC_ALL;
 544         ZFS_EXIT(zsb);
 545         sync_mode = wbc->sync_mode;
 546
 547         /*
 548          * We don't want to run write_cache_pages() in SYNC mode here, because
 549          * that would make putpage() wait for a single page to be committed to
 550          * disk every single time, resulting in atrocious performance. Instead
 551          * we run it once in non-SYNC mode so that the ZIL gets all the data,
 552          * and then we commit it all in one go.
 553          */
 554         wbc->sync_mode = WB_SYNC_NONE;
 555         result = write_cache_pages(mapping, wbc, zpl_putpage, mapping);
 556         if (sync_mode != wbc->sync_mode) {
 557                 ZFS_ENTER(zsb);
 558                 ZFS_VERIFY_ZP(zp);
 559                 if (zsb->z_log != NULL)
 560                         zil_commit(zsb->z_log, zp->z_id);
 561                 ZFS_EXIT(zsb);
 562
 563                 /*
 564                  * We need to call write_cache_pages() again (we can't just
 565                  * return after the commit) because the previous call in
 566                  * non-SYNC mode does not guarantee that we got all the dirty
 567                  * pages (see the implementation of write_cache_pages() for
 568                  * details). That being said, this is a no-op in most cases.
 569                  */
 570                 wbc->sync_mode = sync_mode;
 571                 result = write_cache_pages(mapping, wbc, zpl_putpage, mapping);
 572         }
 573         return (result);
 574 }
 575
 576 /*
 577  * Write out dirty pages to the ARC, this function is only required to
 578  * support mmap(2).  Mapped pages may be dirtied by memory operations
 579  * which never call .write().  These dirty pages are kept in sync with
 580  * the ARC buffers via this hook.
 581  */
 582 static int
 583 zpl_writepage(struct page *pp, struct writeback_control *wbc)
 584 {
 585         if (ITOZSB(pp->mapping->host)->z_os->os_sync == ZFS_SYNC_ALWAYS)
 586                 wbc->sync_mode = WB_SYNC_ALL;
 587
 588         return (zpl_putpage(pp, wbc, pp->mapping));
 589 }
 590
 591 /*
 592  * The only flag combination which matches the behavior of zfs_space()
 593  * is FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE.  The FALLOC_FL_PUNCH_HOLE
 594  * flag was introduced in the 2.6.38 kernel.
 595  */
 596 #if defined(HAVE_FILE_FALLOCATE) || defined(HAVE_INODE_FALLOCATE)
 597 long
 598 zpl_fallocate_common(struct inode *ip, int mode, loff_t offset, loff_t len)
 599 {
 600         int error = -EOPNOTSUPP;
 601
 602 #if defined(FALLOC_FL_PUNCH_HOLE) && defined(FALLOC_FL_KEEP_SIZE)
 603         cred_t *cr = CRED();
 604         flock64_t bf;
 605         loff_t olen;
 606         fstrans_cookie_t cookie;
 607
 608         if (mode != (FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
 609                 return (error);
 610
 611         crhold(cr);
 612
 613         if (offset < 0 || len <= 0)
 614                 return (-EINVAL);
 615
 616         spl_inode_lock(ip);
 617         olen = i_size_read(ip);
 618
 619         if (offset > olen) {
 620                 spl_inode_unlock(ip);
 621                 return (0);
 622         }
 623         if (offset + len > olen)
 624                 len = olen - offset;
 625         bf.l_type = F_WRLCK;
 626         bf.l_whence = 0;
 627         bf.l_start = offset;
 628         bf.l_len = len;
 629         bf.l_pid = 0;
 630
 631         cookie = spl_fstrans_mark();
 632         error = -zfs_space(ip, F_FREESP, &bf, FWRITE, offset, cr);
 633         spl_fstrans_unmark(cookie);
 634         spl_inode_unlock(ip);
 635
 636         crfree(cr);
 637 #endif /* defined(FALLOC_FL_PUNCH_HOLE) && defined(FALLOC_FL_KEEP_SIZE) */
 638
 639         ASSERT3S(error, <=, 0);
 640         return (error);
 641 }
 642 #endif /* defined(HAVE_FILE_FALLOCATE) || defined(HAVE_INODE_FALLOCATE) */
 643
 644 #ifdef HAVE_FILE_FALLOCATE
 645 static long
 646 zpl_fallocate(struct file *filp, int mode, loff_t offset, loff_t len)
 647 {
 648         return zpl_fallocate_common(filp->f_path.dentry->d_inode,
 649             mode, offset, len);
 650 }
 651 #endif /* HAVE_FILE_FALLOCATE */
 652
 653 /*
 654  * Map zfs file z_pflags (xvattr_t) to linux file attributes. Only file
 655  * attributes common to both Linux and Solaris are mapped.
 656  */
 657 static int
 658 zpl_ioctl_getflags(struct file *filp, void __user *arg)
 659 {
 660         struct inode *ip = file_inode(filp);
 661         unsigned int ioctl_flags = 0;
 662         uint64_t zfs_flags = ITOZ(ip)->z_pflags;
 663         int error;
 664
 665         if (zfs_flags & ZFS_IMMUTABLE)
 666                 ioctl_flags |= FS_IMMUTABLE_FL;
 667
 668         if (zfs_flags & ZFS_APPENDONLY)
 669                 ioctl_flags |= FS_APPEND_FL;
 670
 671         if (zfs_flags & ZFS_NODUMP)
 672                 ioctl_flags |= FS_NODUMP_FL;
 673
 674         ioctl_flags &= FS_FL_USER_VISIBLE;
 675
 676         error = copy_to_user(arg, &ioctl_flags, sizeof (ioctl_flags));
 677
 678         return (error);
 679 }
 680
 681 /*
 682  * fchange() is a helper macro to detect if we have been asked to change a
 683  * flag. This is ugly, but the requirement that we do this is a consequence of
 684  * how the Linux file attribute interface was designed. Another consequence is
 685  * that concurrent modification of files suffers from a TOCTOU race. Neither
 686  * are things we can fix without modifying the kernel-userland interface, which
 687  * is outside of our jurisdiction.
 688  */
 689
 690 #define fchange(f0, f1, b0, b1) ((((f0) & (b0)) == (b0)) != \
 691         (((b1) & (f1)) == (f1)))
 692
 693 static int
 694 zpl_ioctl_setflags(struct file *filp, void __user *arg)
 695 {
 696         struct inode    *ip = file_inode(filp);
 697         uint64_t        zfs_flags = ITOZ(ip)->z_pflags;
 698         unsigned int    ioctl_flags;
 699         cred_t          *cr = CRED();
 700         xvattr_t        xva;
 701         xoptattr_t      *xoap;
 702         int             error;
 703         fstrans_cookie_t cookie;
 704
 705         if (copy_from_user(&ioctl_flags, arg, sizeof (ioctl_flags)))
 706                 return (-EFAULT);
 707
 708         if ((ioctl_flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | FS_NODUMP_FL)))
 709                 return (-EOPNOTSUPP);
 710
 711         if ((ioctl_flags & ~(FS_FL_USER_MODIFIABLE)))
 712                 return (-EACCES);
 713
 714         if ((fchange(ioctl_flags, zfs_flags, FS_IMMUTABLE_FL, ZFS_IMMUTABLE) ||
 715             fchange(ioctl_flags, zfs_flags, FS_APPEND_FL, ZFS_APPENDONLY)) &&
 716             !capable(CAP_LINUX_IMMUTABLE))
 717                 return (-EACCES);
 718
 719         if (!zpl_inode_owner_or_capable(ip))
 720                 return (-EACCES);
 721
 722         xva_init(&xva);
 723         xoap = xva_getxoptattr(&xva);
 724
 725         XVA_SET_REQ(&xva, XAT_IMMUTABLE);
 726         if (ioctl_flags & FS_IMMUTABLE_FL)
 727                 xoap->xoa_immutable = B_TRUE;
 728
 729         XVA_SET_REQ(&xva, XAT_APPENDONLY);
 730         if (ioctl_flags & FS_APPEND_FL)
 731                 xoap->xoa_appendonly = B_TRUE;
 732
 733         XVA_SET_REQ(&xva, XAT_NODUMP);
 734         if (ioctl_flags & FS_NODUMP_FL)
 735                 xoap->xoa_nodump = B_TRUE;
 736
 737         crhold(cr);
 738         cookie = spl_fstrans_mark();
 739         error = -zfs_setattr(ip, (vattr_t *)&xva, 0, cr);
 740         spl_fstrans_unmark(cookie);
 741         crfree(cr);
 742
 743         return (error);
 744 }
 745
 746 static long
 747 zpl_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
 748 {
 749         switch (cmd) {
 750         case FS_IOC_GETFLAGS:
 751                 return (zpl_ioctl_getflags(filp, (void *)arg));
 752         case FS_IOC_SETFLAGS:
 753                 return (zpl_ioctl_setflags(filp, (void *)arg));
 754         default:
 755                 return (-ENOTTY);
 756         }
 757 }
 758
 759 #ifdef CONFIG_COMPAT
 760 static long
 761 zpl_compat_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
 762 {
 763         return (zpl_ioctl(filp, cmd, arg));
 764 }
 765 #endif /* CONFIG_COMPAT */
 766
 767
 768 const struct address_space_operations zpl_address_space_operations = {
 769         .readpages      = zpl_readpages,
 770         .readpage       = zpl_readpage,
 771         .writepage      = zpl_writepage,
 772         .writepages     = zpl_writepages,
 773 };
 774
 775 const struct file_operations zpl_file_operations = {
 776         .open           = zpl_open,
 777         .release        = zpl_release,
 778         .llseek         = zpl_llseek,
 779         .read           = zpl_read,
 780         .write          = zpl_write,
 781         .aio_read       = zpl_aio_read,
 782         .aio_write      = zpl_aio_write,
 783         .mmap           = zpl_mmap,
 784         .fsync          = zpl_fsync,
 785         .aio_fsync      = zpl_aio_fsync,
 786 #ifdef HAVE_FILE_FALLOCATE
 787         .fallocate      = zpl_fallocate,
 788 #endif /* HAVE_FILE_FALLOCATE */
 789         .unlocked_ioctl = zpl_ioctl,
 790 #ifdef CONFIG_COMPAT
 791         .compat_ioctl   = zpl_compat_ioctl,
 792 #endif
 793 };
 794
 795 const struct file_operations zpl_dir_file_operations = {
 796         .llseek         = generic_file_llseek,
 797         .read           = generic_read_dir,
 798 #ifdef HAVE_VFS_ITERATE
 799         .iterate        = zpl_iterate,
 800 #else
 801         .readdir        = zpl_readdir,
 802 #endif
 803         .fsync          = zpl_fsync,
 804         .unlocked_ioctl = zpl_ioctl,
 805 #ifdef CONFIG_COMPAT
 806         .compat_ioctl   = zpl_compat_ioctl,
 807 #endif
 808 };