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/zfs_vfsops.h>
  27 #include <sys/zfs_vnops.h>
  28 #include <sys/zfs_znode.h>
  29 #include <sys/zpl.h>
  30
  31
  32 static int
  33 zpl_open(struct inode *ip, struct file *filp)
  34 {
  35         cred_t *cr = CRED();
  36         int error;
  37
  38         crhold(cr);
  39         error = -zfs_open(ip, filp->f_mode, filp->f_flags, cr);
  40         crfree(cr);
  41         ASSERT3S(error, <=, 0);
  42
  43         if (error)
  44                 return (error);
  45
  46         return generic_file_open(ip, filp);
  47 }
  48
  49 static int
  50 zpl_release(struct inode *ip, struct file *filp)
  51 {
  52         cred_t *cr = CRED();
  53         int error;
  54
  55         crhold(cr);
  56         error = -zfs_close(ip, filp->f_flags, cr);
  57         crfree(cr);
  58         ASSERT3S(error, <=, 0);
  59
  60         return (error);
  61 }
  62
  63 static int
  64 zpl_readdir(struct file *filp, void *dirent, filldir_t filldir)
  65 {
  66         struct dentry *dentry = filp->f_path.dentry;
  67         cred_t *cr = CRED();
  68         int error;
  69
  70         crhold(cr);
  71         error = -zfs_readdir(dentry->d_inode, dirent, filldir,
  72             &filp->f_pos, cr);
  73         crfree(cr);
  74         ASSERT3S(error, <=, 0);
  75
  76         return (error);
  77 }
  78
  79 #if defined(HAVE_FSYNC_WITH_DENTRY)
  80 /*
  81  * Linux 2.6.x - 2.6.34 API,
  82  * Through 2.6.34 the nfsd kernel server would pass a NULL 'file struct *'
  83  * to the fops->fsync() hook.  For this reason, we must be careful not to
  84  * use filp unconditionally.
  85  */
  86 static int
  87 zpl_fsync(struct file *filp, struct dentry *dentry, int datasync)
  88 {
  89         cred_t *cr = CRED();
  90         int error;
  91
  92         crhold(cr);
  93         error = -zfs_fsync(dentry->d_inode, datasync, cr);
  94         crfree(cr);
  95         ASSERT3S(error, <=, 0);
  96
  97         return (error);
  98 }
  99
 100 #elif defined(HAVE_FSYNC_WITHOUT_DENTRY)
 101 /*
 102  * Linux 2.6.35 - 3.0 API,
 103  * As of 2.6.35 the dentry argument to the fops->fsync() hook was deemed
 104  * redundant.  The dentry is still accessible via filp->f_path.dentry,
 105  * and we are guaranteed that filp will never be NULL.
 106  */
 107 static int
 108 zpl_fsync(struct file *filp, int datasync)
 109 {
 110         struct inode *inode = filp->f_mapping->host;
 111         cred_t *cr = CRED();
 112         int error;
 113
 114         crhold(cr);
 115         error = -zfs_fsync(inode, datasync, cr);
 116         crfree(cr);
 117         ASSERT3S(error, <=, 0);
 118
 119         return (error);
 120 }
 121
 122 #elif defined(HAVE_FSYNC_RANGE)
 123 /*
 124  * Linux 3.1 - 3.x API,
 125  * As of 3.1 the responsibility to call filemap_write_and_wait_range() has
 126  * been pushed down in to the .fsync() vfs hook.  Additionally, the i_mutex
 127  * lock is no longer held by the caller, for zfs we don't require the lock
 128  * to be held so we don't acquire it.
 129  */
 130 static int
 131 zpl_fsync(struct file *filp, loff_t start, loff_t end, int datasync)
 132 {
 133         struct inode *inode = filp->f_mapping->host;
 134         cred_t *cr = CRED();
 135         int error;
 136
 137         error = filemap_write_and_wait_range(inode->i_mapping, start, end);
 138         if (error)
 139                 return (error);
 140
 141         crhold(cr);
 142         error = -zfs_fsync(inode, datasync, cr);
 143         crfree(cr);
 144         ASSERT3S(error, <=, 0);
 145
 146         return (error);
 147 }
 148 #else
 149 #error "Unsupported fops->fsync() implementation"
 150 #endif
 151
 152 ssize_t
 153 zpl_read_common(struct inode *ip, const char *buf, size_t len, loff_t pos,
 154      uio_seg_t segment, int flags, cred_t *cr)
 155 {
 156         int error;
 157         struct iovec iov;
 158         uio_t uio;
 159
 160         iov.iov_base = (void *)buf;
 161         iov.iov_len = len;
 162
 163         uio.uio_iov = &iov;
 164         uio.uio_resid = len;
 165         uio.uio_iovcnt = 1;
 166         uio.uio_loffset = pos;
 167         uio.uio_limit = MAXOFFSET_T;
 168         uio.uio_segflg = segment;
 169
 170         error = -zfs_read(ip, &uio, flags, cr);
 171         if (error < 0)
 172                 return (error);
 173
 174         return (len - uio.uio_resid);
 175 }
 176
 177 static ssize_t
 178 zpl_read(struct file *filp, char __user *buf, size_t len, loff_t *ppos)
 179 {
 180         cred_t *cr = CRED();
 181         ssize_t read;
 182
 183         crhold(cr);
 184         read = zpl_read_common(filp->f_mapping->host, buf, len, *ppos,
 185             UIO_USERSPACE, filp->f_flags, cr);
 186         crfree(cr);
 187
 188         if (read < 0)
 189                 return (read);
 190
 191         *ppos += read;
 192         return (read);
 193 }
 194
 195 ssize_t
 196 zpl_write_common(struct inode *ip, const char *buf, size_t len, loff_t pos,
 197     uio_seg_t segment, int flags, cred_t *cr)
 198 {
 199         int error;
 200         struct iovec iov;
 201         uio_t uio;
 202
 203         iov.iov_base = (void *)buf;
 204         iov.iov_len = len;
 205
 206         uio.uio_iov = &iov;
 207         uio.uio_resid = len,
 208         uio.uio_iovcnt = 1;
 209         uio.uio_loffset = pos;
 210         uio.uio_limit = MAXOFFSET_T;
 211         uio.uio_segflg = segment;
 212
 213         error = -zfs_write(ip, &uio, flags, cr);
 214         if (error < 0)
 215                 return (error);
 216
 217         return (len - uio.uio_resid);
 218 }
 219
 220 static ssize_t
 221 zpl_write(struct file *filp, const char __user *buf, size_t len, loff_t *ppos)
 222 {
 223         cred_t *cr = CRED();
 224         ssize_t wrote;
 225
 226         crhold(cr);
 227         wrote = zpl_write_common(filp->f_mapping->host, buf, len, *ppos,
 228             UIO_USERSPACE, filp->f_flags, cr);
 229         crfree(cr);
 230
 231         if (wrote < 0)
 232                 return (wrote);
 233
 234         *ppos += wrote;
 235         return (wrote);
 236 }
 237
 238 /*
 239  * It's worth taking a moment to describe how mmap is implemented
 240  * for zfs because it differs considerably from other Linux filesystems.
 241  * However, this issue is handled the same way under OpenSolaris.
 242  *
 243  * The issue is that by design zfs bypasses the Linux page cache and
 244  * leaves all caching up to the ARC.  This has been shown to work
 245  * well for the common read(2)/write(2) case.  However, mmap(2)
 246  * is problem because it relies on being tightly integrated with the
 247  * page cache.  To handle this we cache mmap'ed files twice, once in
 248  * the ARC and a second time in the page cache.  The code is careful
 249  * to keep both copies synchronized.
 250  *
 251  * When a file with an mmap'ed region is written to using write(2)
 252  * both the data in the ARC and existing pages in the page cache
 253  * are updated.  For a read(2) data will be read first from the page
 254  * cache then the ARC if needed.  Neither a write(2) or read(2) will
 255  * will ever result in new pages being added to the page cache.
 256  *
 257  * New pages are added to the page cache only via .readpage() which
 258  * is called when the vfs needs to read a page off disk to back the
 259  * virtual memory region.  These pages may be modified without
 260  * notifying the ARC and will be written out periodically via
 261  * .writepage().  This will occur due to either a sync or the usual
 262  * page aging behavior.  Note because a read(2) of a mmap'ed file
 263  * will always check the page cache first even when the ARC is out
 264  * of date correct data will still be returned.
 265  *
 266  * While this implementation ensures correct behavior it does have
 267  * have some drawbacks.  The most obvious of which is that it
 268  * increases the required memory footprint when access mmap'ed
 269  * files.  It also adds additional complexity to the code keeping
 270  * both caches synchronized.
 271  *
 272  * Longer term it may be possible to cleanly resolve this wart by
 273  * mapping page cache pages directly on to the ARC buffers.  The
 274  * Linux address space operations are flexible enough to allow
 275  * selection of which pages back a particular index.  The trick
 276  * would be working out the details of which subsystem is in
 277  * charge, the ARC, the page cache, or both.  It may also prove
 278  * helpful to move the ARC buffers to a scatter-gather lists
 279  * rather than a vmalloc'ed region.
 280  */
 281 static int
 282 zpl_mmap(struct file *filp, struct vm_area_struct *vma)
 283 {
 284         struct inode *ip = filp->f_mapping->host;
 285         znode_t *zp = ITOZ(ip);
 286         int error;
 287
 288         error = -zfs_map(ip, vma->vm_pgoff, (caddr_t *)vma->vm_start,
 289             (size_t)(vma->vm_end - vma->vm_start), vma->vm_flags);
 290         if (error)
 291                 return (error);
 292
 293         error = generic_file_mmap(filp, vma);
 294         if (error)
 295                 return (error);
 296
 297         mutex_enter(&zp->z_lock);
 298         zp->z_is_mapped = 1;
 299         mutex_exit(&zp->z_lock);
 300
 301         return (error);
 302 }
 303
 304 /*
 305  * Populate a page with data for the Linux page cache.  This function is
 306  * only used to support mmap(2).  There will be an identical copy of the
 307  * data in the ARC which is kept up to date via .write() and .writepage().
 308  *
 309  * Current this function relies on zpl_read_common() and the O_DIRECT
 310  * flag to read in a page.  This works but the more correct way is to
 311  * update zfs_fillpage() to be Linux friendly and use that interface.
 312  */
 313 static int
 314 zpl_readpage(struct file *filp, struct page *pp)
 315 {
 316         struct inode *ip;
 317         struct page *pl[1];
 318         int error = 0;
 319
 320         ASSERT(PageLocked(pp));
 321         ip = pp->mapping->host;
 322         pl[0] = pp;
 323
 324         error = -zfs_getpage(ip, pl, 1);
 325
 326         if (error) {
 327                 SetPageError(pp);
 328                 ClearPageUptodate(pp);
 329         } else {
 330                 ClearPageError(pp);
 331                 SetPageUptodate(pp);
 332                 flush_dcache_page(pp);
 333         }
 334
 335         unlock_page(pp);
 336         return error;
 337 }
 338
 339 /*
 340  * Populate a set of pages with data for the Linux page cache.  This
 341  * function will only be called for read ahead and never for demand
 342  * paging.  For simplicity, the code relies on read_cache_pages() to
 343  * correctly lock each page for IO and call zpl_readpage().
 344  */
 345 static int
 346 zpl_readpages(struct file *filp, struct address_space *mapping,
 347         struct list_head *pages, unsigned nr_pages)
 348 {
 349         return (read_cache_pages(mapping, pages,
 350             (filler_t *)zpl_readpage, filp));
 351 }
 352
 353 int
 354 zpl_putpage(struct page *pp, struct writeback_control *wbc, void *data)
 355 {
 356         struct address_space *mapping = data;
 357
 358         ASSERT(PageLocked(pp));
 359         ASSERT(!PageWriteback(pp));
 360
 361         (void) zfs_putpage(mapping->host, pp, wbc);
 362
 363         return (0);
 364 }
 365
 366 static int
 367 zpl_writepages(struct address_space *mapping, struct writeback_control *wbc)
 368 {
 369         return write_cache_pages(mapping, wbc, zpl_putpage, mapping);
 370 }
 371
 372 /*
 373  * Write out dirty pages to the ARC, this function is only required to
 374  * support mmap(2).  Mapped pages may be dirtied by memory operations
 375  * which never call .write().  These dirty pages are kept in sync with
 376  * the ARC buffers via this hook.
 377  */
 378 static int
 379 zpl_writepage(struct page *pp, struct writeback_control *wbc)
 380 {
 381         return zpl_putpage(pp, wbc, pp->mapping);
 382 }
 383
 384 /*
 385  * The only flag combination which matches the behavior of zfs_space()
 386  * is FALLOC_FL_PUNCH_HOLE.  This flag was introduced in the 2.6.38 kernel.
 387  */
 388 long
 389 zpl_fallocate_common(struct inode *ip, int mode, loff_t offset, loff_t len)
 390 {
 391         cred_t *cr = CRED();
 392         int error = -EOPNOTSUPP;
 393
 394         if (mode & FALLOC_FL_KEEP_SIZE)
 395                 return (-EOPNOTSUPP);
 396
 397         crhold(cr);
 398
 399 #ifdef FALLOC_FL_PUNCH_HOLE
 400         if (mode & FALLOC_FL_PUNCH_HOLE) {
 401                 flock64_t bf;
 402
 403                 bf.l_type = F_WRLCK;
 404                 bf.l_whence = 0;
 405                 bf.l_start = offset;
 406                 bf.l_len = len;
 407                 bf.l_pid = 0;
 408
 409                 error = -zfs_space(ip, F_FREESP, &bf, FWRITE, offset, cr);
 410         }
 411 #endif /* FALLOC_FL_PUNCH_HOLE */
 412
 413         crfree(cr);
 414
 415         ASSERT3S(error, <=, 0);
 416         return (error);
 417 }
 418
 419 #ifdef HAVE_FILE_FALLOCATE
 420 static long
 421 zpl_fallocate(struct file *filp, int mode, loff_t offset, loff_t len)
 422 {
 423         return zpl_fallocate_common(filp->f_path.dentry->d_inode,
 424             mode, offset, len);
 425 }
 426 #endif /* HAVE_FILE_FALLOCATE */
 427
 428 const struct address_space_operations zpl_address_space_operations = {
 429         .readpages      = zpl_readpages,
 430         .readpage       = zpl_readpage,
 431         .writepage      = zpl_writepage,
 432         .writepages     = zpl_writepages,
 433 };
 434
 435 const struct file_operations zpl_file_operations = {
 436         .open           = zpl_open,
 437         .release        = zpl_release,
 438         .llseek         = generic_file_llseek,
 439         .read           = zpl_read,
 440         .write          = zpl_write,
 441         .readdir        = zpl_readdir,
 442         .mmap           = zpl_mmap,
 443         .fsync          = zpl_fsync,
 444 #ifdef HAVE_FILE_FALLOCATE
 445         .fallocate      = zpl_fallocate,
 446 #endif /* HAVE_FILE_FALLOCATE */
 447 };
 448
 449 const struct file_operations zpl_dir_file_operations = {
 450         .llseek         = generic_file_llseek,
 451         .read           = generic_read_dir,
 452         .readdir        = zpl_readdir,
 453         .fsync          = zpl_fsync,
 454 };