fs/xfs/linux-2.6/xfs_file.c

   1 /*
   2  * Copyright (c) 2000-2005 Silicon Graphics, Inc.
   3  * All Rights Reserved.
   4  *
   5  * This program is free software; you can redistribute it and/or
   6  * modify it under the terms of the GNU General Public License as
   7  * published by the Free Software Foundation.
   8  *
   9  * This program is distributed in the hope that it would be useful,
  10  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  11  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  12  * GNU General Public License for more details.
  13  *
  14  * You should have received a copy of the GNU General Public License
  15  * along with this program; if not, write the Free Software Foundation,
  16  * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
  17  */
  18 #include "xfs.h"
  19 #include "xfs_bit.h"
  20 #include "xfs_log.h"
  21 #include "xfs_inum.h"
  22 #include "xfs_sb.h"
  23 #include "xfs_ag.h"
  24 #include "xfs_dir2.h"
  25 #include "xfs_trans.h"
  26 #include "xfs_dmapi.h"
  27 #include "xfs_mount.h"
  28 #include "xfs_bmap_btree.h"
  29 #include "xfs_alloc_btree.h"
  30 #include "xfs_ialloc_btree.h"
  31 #include "xfs_alloc.h"
  32 #include "xfs_btree.h"
  33 #include "xfs_attr_sf.h"
  34 #include "xfs_dir2_sf.h"
  35 #include "xfs_dinode.h"
  36 #include "xfs_inode.h"
  37 #include "xfs_error.h"
  38 #include "xfs_rw.h"
  39 #include "xfs_ioctl32.h"
  40
  41 #include <linux/dcache.h>
  42 #include <linux/smp_lock.h>
  43
  44 static struct vm_operations_struct xfs_file_vm_ops;
  45 #ifdef CONFIG_XFS_DMAPI
  46 static struct vm_operations_struct xfs_dmapi_file_vm_ops;
  47 #endif
  48
  49 STATIC inline ssize_t
  50 __xfs_file_read(
  51         struct kiocb            *iocb,
  52         char                    __user *buf,
  53         int                     ioflags,
  54         size_t                  count,
  55         loff_t                  pos)
  56 {
  57         struct iovec            iov = {buf, count};
  58         struct file             *file = iocb->ki_filp;
  59         bhv_vnode_t             *vp = vn_from_inode(file->f_dentry->d_inode);
  60
  61         BUG_ON(iocb->ki_pos != pos);
  62         if (unlikely(file->f_flags & O_DIRECT))
  63                 ioflags |= IO_ISDIRECT;
  64         return bhv_vop_read(vp, iocb, &iov, 1, &iocb->ki_pos, ioflags, NULL);
  65 }
  66
  67 STATIC ssize_t
  68 xfs_file_aio_read(
  69         struct kiocb            *iocb,
  70         char                    __user *buf,
  71         size_t                  count,
  72         loff_t                  pos)
  73 {
  74         return __xfs_file_read(iocb, buf, IO_ISAIO, count, pos);
  75 }
  76
  77 STATIC ssize_t
  78 xfs_file_aio_read_invis(
  79         struct kiocb            *iocb,
  80         char                    __user *buf,
  81         size_t                  count,
  82         loff_t                  pos)
  83 {
  84         return __xfs_file_read(iocb, buf, IO_ISAIO|IO_INVIS, count, pos);
  85 }
  86
  87 STATIC inline ssize_t
  88 __xfs_file_write(
  89         struct kiocb    *iocb,
  90         const char      __user *buf,
  91         int             ioflags,
  92         size_t          count,
  93         loff_t          pos)
  94 {
  95         struct iovec    iov = {(void __user *)buf, count};
  96         struct file     *file = iocb->ki_filp;
  97         struct inode    *inode = file->f_mapping->host;
  98         bhv_vnode_t     *vp = vn_from_inode(inode);
  99
 100         BUG_ON(iocb->ki_pos != pos);
 101         if (unlikely(file->f_flags & O_DIRECT))
 102                 ioflags |= IO_ISDIRECT;
 103         return bhv_vop_write(vp, iocb, &iov, 1, &iocb->ki_pos, ioflags, NULL);
 104 }
 105
 106 STATIC ssize_t
 107 xfs_file_aio_write(
 108         struct kiocb            *iocb,
 109         const char              __user *buf,
 110         size_t                  count,
 111         loff_t                  pos)
 112 {
 113         return __xfs_file_write(iocb, buf, IO_ISAIO, count, pos);
 114 }
 115
 116 STATIC ssize_t
 117 xfs_file_aio_write_invis(
 118         struct kiocb            *iocb,
 119         const char              __user *buf,
 120         size_t                  count,
 121         loff_t                  pos)
 122 {
 123         return __xfs_file_write(iocb, buf, IO_ISAIO|IO_INVIS, count, pos);
 124 }
 125
 126 STATIC inline ssize_t
 127 __xfs_file_readv(
 128         struct file             *file,
 129         const struct iovec      *iov,
 130         int                     ioflags,
 131         unsigned long           nr_segs,
 132         loff_t                  *ppos)
 133 {
 134         struct inode    *inode = file->f_mapping->host;
 135         bhv_vnode_t     *vp = vn_from_inode(inode);
 136         struct kiocb    kiocb;
 137         ssize_t         rval;
 138
 139         init_sync_kiocb(&kiocb, file);
 140         kiocb.ki_pos = *ppos;
 141
 142         if (unlikely(file->f_flags & O_DIRECT))
 143                 ioflags |= IO_ISDIRECT;
 144         rval = bhv_vop_read(vp, &kiocb, iov, nr_segs,
 145                                 &kiocb.ki_pos, ioflags, NULL);
 146
 147         *ppos = kiocb.ki_pos;
 148         return rval;
 149 }
 150
 151 STATIC ssize_t
 152 xfs_file_readv(
 153         struct file             *file,
 154         const struct iovec      *iov,
 155         unsigned long           nr_segs,
 156         loff_t                  *ppos)
 157 {
 158         return __xfs_file_readv(file, iov, 0, nr_segs, ppos);
 159 }
 160
 161 STATIC ssize_t
 162 xfs_file_readv_invis(
 163         struct file             *file,
 164         const struct iovec      *iov,
 165         unsigned long           nr_segs,
 166         loff_t                  *ppos)
 167 {
 168         return __xfs_file_readv(file, iov, IO_INVIS, nr_segs, ppos);
 169 }
 170
 171 STATIC inline ssize_t
 172 __xfs_file_writev(
 173         struct file             *file,
 174         const struct iovec      *iov,
 175         int                     ioflags,
 176         unsigned long           nr_segs,
 177         loff_t                  *ppos)
 178 {
 179         struct inode    *inode = file->f_mapping->host;
 180         bhv_vnode_t     *vp = vn_from_inode(inode);
 181         struct kiocb    kiocb;
 182         ssize_t         rval;
 183
 184         init_sync_kiocb(&kiocb, file);
 185         kiocb.ki_pos = *ppos;
 186         if (unlikely(file->f_flags & O_DIRECT))
 187                 ioflags |= IO_ISDIRECT;
 188
 189         rval = bhv_vop_write(vp, &kiocb, iov, nr_segs,
 190                                  &kiocb.ki_pos, ioflags, NULL);
 191
 192         *ppos = kiocb.ki_pos;
 193         return rval;
 194 }
 195
 196 STATIC ssize_t
 197 xfs_file_writev(
 198         struct file             *file,
 199         const struct iovec      *iov,
 200         unsigned long           nr_segs,
 201         loff_t                  *ppos)
 202 {
 203         return __xfs_file_writev(file, iov, 0, nr_segs, ppos);
 204 }
 205
 206 STATIC ssize_t
 207 xfs_file_writev_invis(
 208         struct file             *file,
 209         const struct iovec      *iov,
 210         unsigned long           nr_segs,
 211         loff_t                  *ppos)
 212 {
 213         return __xfs_file_writev(file, iov, IO_INVIS, nr_segs, ppos);
 214 }
 215
 216 STATIC ssize_t
 217 xfs_file_sendfile(
 218         struct file             *filp,
 219         loff_t                  *pos,
 220         size_t                  count,
 221         read_actor_t            actor,
 222         void                    *target)
 223 {
 224         return bhv_vop_sendfile(vn_from_inode(filp->f_dentry->d_inode),
 225                                 filp, pos, 0, count, actor, target, NULL);
 226 }
 227
 228 STATIC ssize_t
 229 xfs_file_sendfile_invis(
 230         struct file             *filp,
 231         loff_t                  *pos,
 232         size_t                  count,
 233         read_actor_t            actor,
 234         void                    *target)
 235 {
 236         return bhv_vop_sendfile(vn_from_inode(filp->f_dentry->d_inode),
 237                                 filp, pos, IO_INVIS, count, actor, target, NULL);
 238 }
 239
 240 STATIC ssize_t
 241 xfs_file_splice_read(
 242         struct file             *infilp,
 243         loff_t                  *ppos,
 244         struct pipe_inode_info  *pipe,
 245         size_t                  len,
 246         unsigned int            flags)
 247 {
 248         return bhv_vop_splice_read(vn_from_inode(infilp->f_dentry->d_inode),
 249                                    infilp, ppos, pipe, len, flags, 0, NULL);
 250 }
 251
 252 STATIC ssize_t
 253 xfs_file_splice_read_invis(
 254         struct file             *infilp,
 255         loff_t                  *ppos,
 256         struct pipe_inode_info  *pipe,
 257         size_t                  len,
 258         unsigned int            flags)
 259 {
 260         return bhv_vop_splice_read(vn_from_inode(infilp->f_dentry->d_inode),
 261                                    infilp, ppos, pipe, len, flags, IO_INVIS,
 262                                    NULL);
 263 }
 264
 265 STATIC ssize_t
 266 xfs_file_splice_write(
 267         struct pipe_inode_info  *pipe,
 268         struct file             *outfilp,
 269         loff_t                  *ppos,
 270         size_t                  len,
 271         unsigned int            flags)
 272 {
 273         return bhv_vop_splice_write(vn_from_inode(outfilp->f_dentry->d_inode),
 274                                     pipe, outfilp, ppos, len, flags, 0, NULL);
 275 }
 276
 277 STATIC ssize_t
 278 xfs_file_splice_write_invis(
 279         struct pipe_inode_info  *pipe,
 280         struct file             *outfilp,
 281         loff_t                  *ppos,
 282         size_t                  len,
 283         unsigned int            flags)
 284 {
 285         return bhv_vop_splice_write(vn_from_inode(outfilp->f_dentry->d_inode),
 286                                     pipe, outfilp, ppos, len, flags, IO_INVIS,
 287                                     NULL);
 288 }
 289
 290 STATIC int
 291 xfs_file_open(
 292         struct inode    *inode,
 293         struct file     *filp)
 294 {
 295         if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
 296                 return -EFBIG;
 297         return -bhv_vop_open(vn_from_inode(inode), NULL);
 298 }
 299
 300 STATIC int
 301 xfs_file_close(
 302         struct file     *filp)
 303 {
 304         return -bhv_vop_close(vn_from_inode(filp->f_dentry->d_inode), 0,
 305                                 file_count(filp) > 1 ? L_FALSE : L_TRUE, NULL);
 306 }
 307
 308 STATIC int
 309 xfs_file_release(
 310         struct inode    *inode,
 311         struct file     *filp)
 312 {
 313         bhv_vnode_t     *vp = vn_from_inode(inode);
 314
 315         if (vp)
 316                 return -bhv_vop_release(vp);
 317         return 0;
 318 }
 319
 320 STATIC int
 321 xfs_file_fsync(
 322         struct file     *filp,
 323         struct dentry   *dentry,
 324         int             datasync)
 325 {
 326         bhv_vnode_t     *vp = vn_from_inode(dentry->d_inode);
 327         int             flags = FSYNC_WAIT;
 328
 329         if (datasync)
 330                 flags |= FSYNC_DATA;
 331         if (VN_TRUNC(vp))
 332                 VUNTRUNCATE(vp);
 333         return -bhv_vop_fsync(vp, flags, NULL, (xfs_off_t)0, (xfs_off_t)-1);
 334 }
 335
 336 #ifdef CONFIG_XFS_DMAPI
 337 STATIC struct page *
 338 xfs_vm_nopage(
 339         struct vm_area_struct   *area,
 340         unsigned long           address,
 341         int                     *type)
 342 {
 343         struct inode    *inode = area->vm_file->f_dentry->d_inode;
 344         bhv_vnode_t     *vp = vn_from_inode(inode);
 345
 346         ASSERT_ALWAYS(vp->v_vfsp->vfs_flag & VFS_DMI);
 347         if (XFS_SEND_MMAP(XFS_VFSTOM(vp->v_vfsp), area, 0))
 348                 return NULL;
 349         return filemap_nopage(area, address, type);
 350 }
 351 #endif /* CONFIG_XFS_DMAPI */
 352
 353 STATIC int
 354 xfs_file_readdir(
 355         struct file     *filp,
 356         void            *dirent,
 357         filldir_t       filldir)
 358 {
 359         int             error = 0;
 360         bhv_vnode_t     *vp = vn_from_inode(filp->f_dentry->d_inode);
 361         uio_t           uio;
 362         iovec_t         iov;
 363         int             eof = 0;
 364         caddr_t         read_buf;
 365         int             namelen, size = 0;
 366         size_t          rlen = PAGE_CACHE_SIZE;
 367         xfs_off_t       start_offset, curr_offset;
 368         xfs_dirent_t    *dbp = NULL;
 369
 370         /* Try fairly hard to get memory */
 371         do {
 372                 if ((read_buf = (caddr_t)kmalloc(rlen, GFP_KERNEL)))
 373                         break;
 374                 rlen >>= 1;
 375         } while (rlen >= 1024);
 376
 377         if (read_buf == NULL)
 378                 return -ENOMEM;
 379
 380         uio.uio_iov = &iov;
 381         uio.uio_segflg = UIO_SYSSPACE;
 382         curr_offset = filp->f_pos;
 383         if (filp->f_pos != 0x7fffffff)
 384                 uio.uio_offset = filp->f_pos;
 385         else
 386                 uio.uio_offset = 0xffffffff;
 387
 388         while (!eof) {
 389                 uio.uio_resid = iov.iov_len = rlen;
 390                 iov.iov_base = read_buf;
 391                 uio.uio_iovcnt = 1;
 392
 393                 start_offset = uio.uio_offset;
 394
 395                 error = bhv_vop_readdir(vp, &uio, NULL, &eof);
 396                 if ((uio.uio_offset == start_offset) || error) {
 397                         size = 0;
 398                         break;
 399                 }
 400
 401                 size = rlen - uio.uio_resid;
 402                 dbp = (xfs_dirent_t *)read_buf;
 403                 while (size > 0) {
 404                         namelen = strlen(dbp->d_name);
 405
 406                         if (filldir(dirent, dbp->d_name, namelen,
 407                                         (loff_t) curr_offset & 0x7fffffff,
 408                                         (ino_t) dbp->d_ino,
 409                                         DT_UNKNOWN)) {
 410                                 goto done;
 411                         }
 412                         size -= dbp->d_reclen;
 413                         curr_offset = (loff_t)dbp->d_off /* & 0x7fffffff */;
 414                         dbp = (xfs_dirent_t *)((char *)dbp + dbp->d_reclen);
 415                 }
 416         }
 417 done:
 418         if (!error) {
 419                 if (size == 0)
 420                         filp->f_pos = uio.uio_offset & 0x7fffffff;
 421                 else if (dbp)
 422                         filp->f_pos = curr_offset;
 423         }
 424
 425         kfree(read_buf);
 426         return -error;
 427 }
 428
 429 STATIC int
 430 xfs_file_mmap(
 431         struct file     *filp,
 432         struct vm_area_struct *vma)
 433 {
 434         vma->vm_ops = &xfs_file_vm_ops;
 435
 436 #ifdef CONFIG_XFS_DMAPI
 437         if (vn_from_inode(filp->f_dentry->d_inode)->v_vfsp->vfs_flag & VFS_DMI)
 438                 vma->vm_ops = &xfs_dmapi_file_vm_ops;
 439 #endif /* CONFIG_XFS_DMAPI */
 440
 441         file_accessed(filp);
 442         return 0;
 443 }
 444
 445 STATIC long
 446 xfs_file_ioctl(
 447         struct file     *filp,
 448         unsigned int    cmd,
 449         unsigned long   p)
 450 {
 451         int             error;
 452         struct inode    *inode = filp->f_dentry->d_inode;
 453         bhv_vnode_t     *vp = vn_from_inode(inode);
 454
 455         error = bhv_vop_ioctl(vp, inode, filp, 0, cmd, (void __user *)p);
 456         VMODIFY(vp);
 457
 458         /* NOTE:  some of the ioctl's return positive #'s as a
 459          *        byte count indicating success, such as
 460          *        readlink_by_handle.  So we don't "sign flip"
 461          *        like most other routines.  This means true
 462          *        errors need to be returned as a negative value.
 463          */
 464         return error;
 465 }
 466
 467 STATIC long
 468 xfs_file_ioctl_invis(
 469         struct file     *filp,
 470         unsigned int    cmd,
 471         unsigned long   p)
 472 {
 473         int             error;
 474         struct inode    *inode = filp->f_dentry->d_inode;
 475         bhv_vnode_t     *vp = vn_from_inode(inode);
 476
 477         error = bhv_vop_ioctl(vp, inode, filp, IO_INVIS, cmd, (void __user *)p);
 478         VMODIFY(vp);
 479
 480         /* NOTE:  some of the ioctl's return positive #'s as a
 481          *        byte count indicating success, such as
 482          *        readlink_by_handle.  So we don't "sign flip"
 483          *        like most other routines.  This means true
 484          *        errors need to be returned as a negative value.
 485          */
 486         return error;
 487 }
 488
 489 #ifdef CONFIG_XFS_DMAPI
 490 #ifdef HAVE_VMOP_MPROTECT
 491 STATIC int
 492 xfs_vm_mprotect(
 493         struct vm_area_struct *vma,
 494         unsigned int    newflags)
 495 {
 496         bhv_vnode_t     *vp = vn_from_inode(vma->vm_file->f_dentry->d_inode);
 497         int             error = 0;
 498
 499         if (vp->v_vfsp->vfs_flag & VFS_DMI) {
 500                 if ((vma->vm_flags & VM_MAYSHARE) &&
 501                     (newflags & VM_WRITE) && !(vma->vm_flags & VM_WRITE)) {
 502                         xfs_mount_t     *mp = XFS_VFSTOM(vp->v_vfsp);
 503
 504                         error = XFS_SEND_MMAP(mp, vma, VM_WRITE);
 505                     }
 506         }
 507         return error;
 508 }
 509 #endif /* HAVE_VMOP_MPROTECT */
 510 #endif /* CONFIG_XFS_DMAPI */
 511
 512 #ifdef HAVE_FOP_OPEN_EXEC
 513 /* If the user is attempting to execute a file that is offline then
 514  * we have to trigger a DMAPI READ event before the file is marked as busy
 515  * otherwise the invisible I/O will not be able to write to the file to bring
 516  * it back online.
 517  */
 518 STATIC int
 519 xfs_file_open_exec(
 520         struct inode    *inode)
 521 {
 522         bhv_vnode_t     *vp = vn_from_inode(inode);
 523
 524         if (unlikely(vp->v_vfsp->vfs_flag & VFS_DMI)) {
 525                 xfs_mount_t     *mp = XFS_VFSTOM(vp->v_vfsp);
 526                 xfs_inode_t     *ip = xfs_vtoi(vp);
 527
 528                 if (!ip)
 529                         return -EINVAL;
 530                 if (DM_EVENT_ENABLED(vp->v_vfsp, ip, DM_EVENT_READ))
 531                         return -XFS_SEND_DATA(mp, DM_EVENT_READ, vp,
 532                                                0, 0, 0, NULL);
 533         }
 534         return 0;
 535 }
 536 #endif /* HAVE_FOP_OPEN_EXEC */
 537
 538 const struct file_operations xfs_file_operations = {
 539         .llseek         = generic_file_llseek,
 540         .read           = do_sync_read,
 541         .write          = do_sync_write,
 542         .readv          = xfs_file_readv,
 543         .writev         = xfs_file_writev,
 544         .aio_read       = xfs_file_aio_read,
 545         .aio_write      = xfs_file_aio_write,
 546         .sendfile       = xfs_file_sendfile,
 547         .splice_read    = xfs_file_splice_read,
 548         .splice_write   = xfs_file_splice_write,
 549         .unlocked_ioctl = xfs_file_ioctl,
 550 #ifdef CONFIG_COMPAT
 551         .compat_ioctl   = xfs_file_compat_ioctl,
 552 #endif
 553         .mmap           = xfs_file_mmap,
 554         .open           = xfs_file_open,
 555         .flush          = xfs_file_close,
 556         .release        = xfs_file_release,
 557         .fsync          = xfs_file_fsync,
 558 #ifdef HAVE_FOP_OPEN_EXEC
 559         .open_exec      = xfs_file_open_exec,
 560 #endif
 561 };
 562
 563 const struct file_operations xfs_invis_file_operations = {
 564         .llseek         = generic_file_llseek,
 565         .read           = do_sync_read,
 566         .write          = do_sync_write,
 567         .readv          = xfs_file_readv_invis,
 568         .writev         = xfs_file_writev_invis,
 569         .aio_read       = xfs_file_aio_read_invis,
 570         .aio_write      = xfs_file_aio_write_invis,
 571         .sendfile       = xfs_file_sendfile_invis,
 572         .splice_read    = xfs_file_splice_read_invis,
 573         .splice_write   = xfs_file_splice_write_invis,
 574         .unlocked_ioctl = xfs_file_ioctl_invis,
 575 #ifdef CONFIG_COMPAT
 576         .compat_ioctl   = xfs_file_compat_invis_ioctl,
 577 #endif
 578         .mmap           = xfs_file_mmap,
 579         .open           = xfs_file_open,
 580         .flush          = xfs_file_close,
 581         .release        = xfs_file_release,
 582         .fsync          = xfs_file_fsync,
 583 };
 584
 585
 586 const struct file_operations xfs_dir_file_operations = {
 587         .read           = generic_read_dir,
 588         .readdir        = xfs_file_readdir,
 589         .unlocked_ioctl = xfs_file_ioctl,
 590 #ifdef CONFIG_COMPAT
 591         .compat_ioctl   = xfs_file_compat_ioctl,
 592 #endif
 593         .fsync          = xfs_file_fsync,
 594 };
 595
 596 static struct vm_operations_struct xfs_file_vm_ops = {
 597         .nopage         = filemap_nopage,
 598         .populate       = filemap_populate,
 599 };
 600
 601 #ifdef CONFIG_XFS_DMAPI
 602 static struct vm_operations_struct xfs_dmapi_file_vm_ops = {
 603         .nopage         = xfs_vm_nopage,
 604         .populate       = filemap_populate,
 605 #ifdef HAVE_VMOP_MPROTECT
 606         .mprotect       = xfs_vm_mprotect,
 607 #endif
 608 };
 609 #endif /* CONFIG_XFS_DMAPI */