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_dir.h"
  25 #include "xfs_dir2.h"
  26 #include "xfs_trans.h"
  27 #include "xfs_dmapi.h"
  28 #include "xfs_mount.h"
  29 #include "xfs_bmap_btree.h"
  30 #include "xfs_alloc_btree.h"
  31 #include "xfs_ialloc_btree.h"
  32 #include "xfs_alloc.h"
  33 #include "xfs_btree.h"
  34 #include "xfs_attr_sf.h"
  35 #include "xfs_dir_sf.h"
  36 #include "xfs_dir2_sf.h"
  37 #include "xfs_dinode.h"
  38 #include "xfs_inode.h"
  39 #include "xfs_error.h"
  40 #include "xfs_rw.h"
  41 #include "xfs_ioctl32.h"
  42
  43 #include <linux/dcache.h>
  44 #include <linux/smp_lock.h>
  45
  46 static struct vm_operations_struct xfs_file_vm_ops;
  47 #ifdef CONFIG_XFS_DMAPI
  48 static struct vm_operations_struct xfs_dmapi_file_vm_ops;
  49 #endif
  50
  51 STATIC inline ssize_t
  52 __xfs_file_read(
  53         struct kiocb            *iocb,
  54         char                    __user *buf,
  55         int                     ioflags,
  56         size_t                  count,
  57         loff_t                  pos)
  58 {
  59         struct iovec            iov = {buf, count};
  60         struct file             *file = iocb->ki_filp;
  61         vnode_t                 *vp = vn_from_inode(file->f_dentry->d_inode);
  62         ssize_t                 rval;
  63
  64         BUG_ON(iocb->ki_pos != pos);
  65
  66         if (unlikely(file->f_flags & O_DIRECT))
  67                 ioflags |= IO_ISDIRECT;
  68         VOP_READ(vp, iocb, &iov, 1, &iocb->ki_pos, ioflags, NULL, rval);
  69         return rval;
  70 }
  71
  72 STATIC ssize_t
  73 xfs_file_aio_read(
  74         struct kiocb            *iocb,
  75         char                    __user *buf,
  76         size_t                  count,
  77         loff_t                  pos)
  78 {
  79         return __xfs_file_read(iocb, buf, IO_ISAIO, count, pos);
  80 }
  81
  82 STATIC ssize_t
  83 xfs_file_aio_read_invis(
  84         struct kiocb            *iocb,
  85         char                    __user *buf,
  86         size_t                  count,
  87         loff_t                  pos)
  88 {
  89         return __xfs_file_read(iocb, buf, IO_ISAIO|IO_INVIS, count, pos);
  90 }
  91
  92 STATIC inline ssize_t
  93 __xfs_file_write(
  94         struct kiocb    *iocb,
  95         const char      __user *buf,
  96         int             ioflags,
  97         size_t          count,
  98         loff_t          pos)
  99 {
 100         struct iovec    iov = {(void __user *)buf, count};
 101         struct file     *file = iocb->ki_filp;
 102         struct inode    *inode = file->f_mapping->host;
 103         vnode_t         *vp = vn_from_inode(inode);
 104         ssize_t         rval;
 105
 106         BUG_ON(iocb->ki_pos != pos);
 107         if (unlikely(file->f_flags & O_DIRECT))
 108                 ioflags |= IO_ISDIRECT;
 109
 110         VOP_WRITE(vp, iocb, &iov, 1, &iocb->ki_pos, ioflags, NULL, rval);
 111         return rval;
 112 }
 113
 114 STATIC ssize_t
 115 xfs_file_aio_write(
 116         struct kiocb            *iocb,
 117         const char              __user *buf,
 118         size_t                  count,
 119         loff_t                  pos)
 120 {
 121         return __xfs_file_write(iocb, buf, IO_ISAIO, count, pos);
 122 }
 123
 124 STATIC ssize_t
 125 xfs_file_aio_write_invis(
 126         struct kiocb            *iocb,
 127         const char              __user *buf,
 128         size_t                  count,
 129         loff_t                  pos)
 130 {
 131         return __xfs_file_write(iocb, buf, IO_ISAIO|IO_INVIS, count, pos);
 132 }
 133
 134 STATIC inline ssize_t
 135 __xfs_file_readv(
 136         struct file             *file,
 137         const struct iovec      *iov,
 138         int                     ioflags,
 139         unsigned long           nr_segs,
 140         loff_t                  *ppos)
 141 {
 142         struct inode    *inode = file->f_mapping->host;
 143         vnode_t         *vp = vn_from_inode(inode);
 144         struct kiocb    kiocb;
 145         ssize_t         rval;
 146
 147         init_sync_kiocb(&kiocb, file);
 148         kiocb.ki_pos = *ppos;
 149
 150         if (unlikely(file->f_flags & O_DIRECT))
 151                 ioflags |= IO_ISDIRECT;
 152         VOP_READ(vp, &kiocb, iov, nr_segs, &kiocb.ki_pos, ioflags, NULL, rval);
 153
 154         *ppos = kiocb.ki_pos;
 155         return rval;
 156 }
 157
 158 STATIC ssize_t
 159 xfs_file_readv(
 160         struct file             *file,
 161         const struct iovec      *iov,
 162         unsigned long           nr_segs,
 163         loff_t                  *ppos)
 164 {
 165         return __xfs_file_readv(file, iov, 0, nr_segs, ppos);
 166 }
 167
 168 STATIC ssize_t
 169 xfs_file_readv_invis(
 170         struct file             *file,
 171         const struct iovec      *iov,
 172         unsigned long           nr_segs,
 173         loff_t                  *ppos)
 174 {
 175         return __xfs_file_readv(file, iov, IO_INVIS, nr_segs, ppos);
 176 }
 177
 178 STATIC inline ssize_t
 179 __xfs_file_writev(
 180         struct file             *file,
 181         const struct iovec      *iov,
 182         int                     ioflags,
 183         unsigned long           nr_segs,
 184         loff_t                  *ppos)
 185 {
 186         struct inode    *inode = file->f_mapping->host;
 187         vnode_t         *vp = vn_from_inode(inode);
 188         struct kiocb    kiocb;
 189         ssize_t         rval;
 190
 191         init_sync_kiocb(&kiocb, file);
 192         kiocb.ki_pos = *ppos;
 193         if (unlikely(file->f_flags & O_DIRECT))
 194                 ioflags |= IO_ISDIRECT;
 195
 196         VOP_WRITE(vp, &kiocb, iov, nr_segs, &kiocb.ki_pos, ioflags, NULL, rval);
 197
 198         *ppos = kiocb.ki_pos;
 199         return rval;
 200 }
 201
 202 STATIC ssize_t
 203 xfs_file_writev(
 204         struct file             *file,
 205         const struct iovec      *iov,
 206         unsigned long           nr_segs,
 207         loff_t                  *ppos)
 208 {
 209         return __xfs_file_writev(file, iov, 0, nr_segs, ppos);
 210 }
 211
 212 STATIC ssize_t
 213 xfs_file_writev_invis(
 214         struct file             *file,
 215         const struct iovec      *iov,
 216         unsigned long           nr_segs,
 217         loff_t                  *ppos)
 218 {
 219         return __xfs_file_writev(file, iov, IO_INVIS, nr_segs, ppos);
 220 }
 221
 222 STATIC ssize_t
 223 xfs_file_sendfile(
 224         struct file             *filp,
 225         loff_t                  *pos,
 226         size_t                  count,
 227         read_actor_t            actor,
 228         void                    *target)
 229 {
 230         vnode_t                 *vp = vn_from_inode(filp->f_dentry->d_inode);
 231         ssize_t                 rval;
 232
 233         VOP_SENDFILE(vp, filp, pos, 0, count, actor, target, NULL, rval);
 234         return rval;
 235 }
 236
 237 STATIC ssize_t
 238 xfs_file_sendfile_invis(
 239         struct file             *filp,
 240         loff_t                  *pos,
 241         size_t                  count,
 242         read_actor_t            actor,
 243         void                    *target)
 244 {
 245         vnode_t                 *vp = vn_from_inode(filp->f_dentry->d_inode);
 246         ssize_t                 rval;
 247
 248         VOP_SENDFILE(vp, filp, pos, IO_INVIS, count, actor, target, NULL, rval);
 249         return rval;
 250 }
 251
 252 STATIC ssize_t
 253 xfs_file_splice_read(
 254         struct file             *infilp,
 255         loff_t                  *ppos,
 256         struct pipe_inode_info  *pipe,
 257         size_t                  len,
 258         unsigned int            flags)
 259 {
 260         vnode_t                 *vp = vn_from_inode(infilp->f_dentry->d_inode);
 261         ssize_t                 rval;
 262
 263         VOP_SPLICE_READ(vp, infilp, ppos, pipe, len, flags, 0, NULL, rval);
 264         return rval;
 265 }
 266
 267 STATIC ssize_t
 268 xfs_file_splice_read_invis(
 269         struct file             *infilp,
 270         loff_t                  *ppos,
 271         struct pipe_inode_info  *pipe,
 272         size_t                  len,
 273         unsigned int            flags)
 274 {
 275         vnode_t                 *vp = vn_from_inode(infilp->f_dentry->d_inode);
 276         ssize_t                 rval;
 277
 278         VOP_SPLICE_READ(vp, infilp, ppos, pipe, len, flags, IO_INVIS, NULL, rval);
 279         return rval;
 280 }
 281
 282 STATIC ssize_t
 283 xfs_file_splice_write(
 284         struct pipe_inode_info  *pipe,
 285         struct file             *outfilp,
 286         loff_t                  *ppos,
 287         size_t                  len,
 288         unsigned int            flags)
 289 {
 290         vnode_t                 *vp = vn_from_inode(outfilp->f_dentry->d_inode);
 291         ssize_t                 rval;
 292
 293         VOP_SPLICE_WRITE(vp, pipe, outfilp, ppos, len, flags, 0, NULL, rval);
 294         return rval;
 295 }
 296
 297 STATIC ssize_t
 298 xfs_file_splice_write_invis(
 299         struct pipe_inode_info  *pipe,
 300         struct file             *outfilp,
 301         loff_t                  *ppos,
 302         size_t                  len,
 303         unsigned int            flags)
 304 {
 305         vnode_t                 *vp = vn_from_inode(outfilp->f_dentry->d_inode);
 306         ssize_t                 rval;
 307
 308         VOP_SPLICE_WRITE(vp, pipe, outfilp, ppos, len, flags, IO_INVIS, NULL, rval);
 309         return rval;
 310 }
 311
 312 STATIC int
 313 xfs_file_open(
 314         struct inode    *inode,
 315         struct file     *filp)
 316 {
 317         vnode_t         *vp = vn_from_inode(inode);
 318         int             error;
 319
 320         if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
 321                 return -EFBIG;
 322         VOP_OPEN(vp, NULL, error);
 323         return -error;
 324 }
 325
 326 STATIC int
 327 xfs_file_release(
 328         struct inode    *inode,
 329         struct file     *filp)
 330 {
 331         vnode_t         *vp = vn_from_inode(inode);
 332         int             error = 0;
 333
 334         if (vp)
 335                 VOP_RELEASE(vp, error);
 336         return -error;
 337 }
 338
 339 STATIC int
 340 xfs_file_fsync(
 341         struct file     *filp,
 342         struct dentry   *dentry,
 343         int             datasync)
 344 {
 345         struct inode    *inode = dentry->d_inode;
 346         vnode_t         *vp = vn_from_inode(inode);
 347         int             error;
 348         int             flags = FSYNC_WAIT;
 349
 350         if (datasync)
 351                 flags |= FSYNC_DATA;
 352         VOP_FSYNC(vp, flags, NULL, (xfs_off_t)0, (xfs_off_t)-1, error);
 353         return -error;
 354 }
 355
 356 #ifdef CONFIG_XFS_DMAPI
 357 STATIC struct page *
 358 xfs_vm_nopage(
 359         struct vm_area_struct   *area,
 360         unsigned long           address,
 361         int                     *type)
 362 {
 363         struct inode    *inode = area->vm_file->f_dentry->d_inode;
 364         vnode_t         *vp = vn_from_inode(inode);
 365         xfs_mount_t     *mp = XFS_VFSTOM(vp->v_vfsp);
 366         int             error;
 367
 368         ASSERT_ALWAYS(vp->v_vfsp->vfs_flag & VFS_DMI);
 369
 370         error = XFS_SEND_MMAP(mp, area, 0);
 371         if (error)
 372                 return NULL;
 373
 374         return filemap_nopage(area, address, type);
 375 }
 376 #endif /* CONFIG_XFS_DMAPI */
 377
 378 STATIC int
 379 xfs_file_readdir(
 380         struct file     *filp,
 381         void            *dirent,
 382         filldir_t       filldir)
 383 {
 384         int             error = 0;
 385         vnode_t         *vp = vn_from_inode(filp->f_dentry->d_inode);
 386         uio_t           uio;
 387         iovec_t         iov;
 388         int             eof = 0;
 389         caddr_t         read_buf;
 390         int             namelen, size = 0;
 391         size_t          rlen = PAGE_CACHE_SIZE;
 392         xfs_off_t       start_offset, curr_offset;
 393         xfs_dirent_t    *dbp = NULL;
 394
 395         /* Try fairly hard to get memory */
 396         do {
 397                 if ((read_buf = (caddr_t)kmalloc(rlen, GFP_KERNEL)))
 398                         break;
 399                 rlen >>= 1;
 400         } while (rlen >= 1024);
 401
 402         if (read_buf == NULL)
 403                 return -ENOMEM;
 404
 405         uio.uio_iov = &iov;
 406         uio.uio_segflg = UIO_SYSSPACE;
 407         curr_offset = filp->f_pos;
 408         if (filp->f_pos != 0x7fffffff)
 409                 uio.uio_offset = filp->f_pos;
 410         else
 411                 uio.uio_offset = 0xffffffff;
 412
 413         while (!eof) {
 414                 uio.uio_resid = iov.iov_len = rlen;
 415                 iov.iov_base = read_buf;
 416                 uio.uio_iovcnt = 1;
 417
 418                 start_offset = uio.uio_offset;
 419
 420                 VOP_READDIR(vp, &uio, NULL, &eof, error);
 421                 if ((uio.uio_offset == start_offset) || error) {
 422                         size = 0;
 423                         break;
 424                 }
 425
 426                 size = rlen - uio.uio_resid;
 427                 dbp = (xfs_dirent_t *)read_buf;
 428                 while (size > 0) {
 429                         namelen = strlen(dbp->d_name);
 430
 431                         if (filldir(dirent, dbp->d_name, namelen,
 432                                         (loff_t) curr_offset & 0x7fffffff,
 433                                         (ino_t) dbp->d_ino,
 434                                         DT_UNKNOWN)) {
 435                                 goto done;
 436                         }
 437                         size -= dbp->d_reclen;
 438                         curr_offset = (loff_t)dbp->d_off /* & 0x7fffffff */;
 439                         dbp = (xfs_dirent_t *)((char *)dbp + dbp->d_reclen);
 440                 }
 441         }
 442 done:
 443         if (!error) {
 444                 if (size == 0)
 445                         filp->f_pos = uio.uio_offset & 0x7fffffff;
 446                 else if (dbp)
 447                         filp->f_pos = curr_offset;
 448         }
 449
 450         kfree(read_buf);
 451         return -error;
 452 }
 453
 454 STATIC int
 455 xfs_file_mmap(
 456         struct file     *filp,
 457         struct vm_area_struct *vma)
 458 {
 459         struct inode    *ip = filp->f_dentry->d_inode;
 460         vnode_t         *vp = vn_from_inode(ip);
 461         vattr_t         vattr;
 462         int             error;
 463
 464         vma->vm_ops = &xfs_file_vm_ops;
 465
 466 #ifdef CONFIG_XFS_DMAPI
 467         if (vp->v_vfsp->vfs_flag & VFS_DMI) {
 468                 vma->vm_ops = &xfs_dmapi_file_vm_ops;
 469         }
 470 #endif /* CONFIG_XFS_DMAPI */
 471
 472         vattr.va_mask = XFS_AT_UPDATIME;
 473         VOP_SETATTR(vp, &vattr, XFS_AT_UPDATIME, NULL, error);
 474         if (likely(!error))
 475                 __vn_revalidate(vp, &vattr);    /* update flags */
 476         return 0;
 477 }
 478
 479
 480 STATIC long
 481 xfs_file_ioctl(
 482         struct file     *filp,
 483         unsigned int    cmd,
 484         unsigned long   arg)
 485 {
 486         int             error;
 487         struct inode    *inode = filp->f_dentry->d_inode;
 488         vnode_t         *vp = vn_from_inode(inode);
 489
 490         VOP_IOCTL(vp, inode, filp, 0, cmd, (void __user *)arg, error);
 491         VMODIFY(vp);
 492
 493         /* NOTE:  some of the ioctl's return positive #'s as a
 494          *        byte count indicating success, such as
 495          *        readlink_by_handle.  So we don't "sign flip"
 496          *        like most other routines.  This means true
 497          *        errors need to be returned as a negative value.
 498          */
 499         return error;
 500 }
 501
 502 STATIC long
 503 xfs_file_ioctl_invis(
 504         struct file     *filp,
 505         unsigned int    cmd,
 506         unsigned long   arg)
 507 {
 508         struct inode    *inode = filp->f_dentry->d_inode;
 509         vnode_t         *vp = vn_from_inode(inode);
 510         int             error;
 511
 512         VOP_IOCTL(vp, inode, filp, IO_INVIS, cmd, (void __user *)arg, error);
 513         VMODIFY(vp);
 514
 515         /* NOTE:  some of the ioctl's return positive #'s as a
 516          *        byte count indicating success, such as
 517          *        readlink_by_handle.  So we don't "sign flip"
 518          *        like most other routines.  This means true
 519          *        errors need to be returned as a negative value.
 520          */
 521         return error;
 522 }
 523
 524 #ifdef CONFIG_XFS_DMAPI
 525 #ifdef HAVE_VMOP_MPROTECT
 526 STATIC int
 527 xfs_vm_mprotect(
 528         struct vm_area_struct *vma,
 529         unsigned int    newflags)
 530 {
 531         vnode_t         *vp = vn_from_inode(vma->vm_file->f_dentry->d_inode);
 532         int             error = 0;
 533
 534         if (vp->v_vfsp->vfs_flag & VFS_DMI) {
 535                 if ((vma->vm_flags & VM_MAYSHARE) &&
 536                     (newflags & VM_WRITE) && !(vma->vm_flags & VM_WRITE)) {
 537                         xfs_mount_t     *mp = XFS_VFSTOM(vp->v_vfsp);
 538
 539                         error = XFS_SEND_MMAP(mp, vma, VM_WRITE);
 540                     }
 541         }
 542         return error;
 543 }
 544 #endif /* HAVE_VMOP_MPROTECT */
 545 #endif /* CONFIG_XFS_DMAPI */
 546
 547 #ifdef HAVE_FOP_OPEN_EXEC
 548 /* If the user is attempting to execute a file that is offline then
 549  * we have to trigger a DMAPI READ event before the file is marked as busy
 550  * otherwise the invisible I/O will not be able to write to the file to bring
 551  * it back online.
 552  */
 553 STATIC int
 554 xfs_file_open_exec(
 555         struct inode    *inode)
 556 {
 557         vnode_t         *vp = vn_from_inode(inode);
 558         xfs_mount_t     *mp = XFS_VFSTOM(vp->v_vfsp);
 559         int             error = 0;
 560         xfs_inode_t     *ip;
 561
 562         if (vp->v_vfsp->vfs_flag & VFS_DMI) {
 563                 ip = xfs_vtoi(vp);
 564                 if (!ip) {
 565                         error = -EINVAL;
 566                         goto open_exec_out;
 567                 }
 568                 if (DM_EVENT_ENABLED(vp->v_vfsp, ip, DM_EVENT_READ)) {
 569                         error = -XFS_SEND_DATA(mp, DM_EVENT_READ, vp,
 570                                                0, 0, 0, NULL);
 571                 }
 572         }
 573 open_exec_out:
 574         return error;
 575 }
 576 #endif /* HAVE_FOP_OPEN_EXEC */
 577
 578 const struct file_operations xfs_file_operations = {
 579         .llseek         = generic_file_llseek,
 580         .read           = do_sync_read,
 581         .write          = do_sync_write,
 582         .readv          = xfs_file_readv,
 583         .writev         = xfs_file_writev,
 584         .aio_read       = xfs_file_aio_read,
 585         .aio_write      = xfs_file_aio_write,
 586         .sendfile       = xfs_file_sendfile,
 587         .splice_read    = xfs_file_splice_read,
 588         .splice_write   = xfs_file_splice_write,
 589         .unlocked_ioctl = xfs_file_ioctl,
 590 #ifdef CONFIG_COMPAT
 591         .compat_ioctl   = xfs_file_compat_ioctl,
 592 #endif
 593         .mmap           = xfs_file_mmap,
 594         .open           = xfs_file_open,
 595         .release        = xfs_file_release,
 596         .fsync          = xfs_file_fsync,
 597 #ifdef HAVE_FOP_OPEN_EXEC
 598         .open_exec      = xfs_file_open_exec,
 599 #endif
 600 };
 601
 602 const struct file_operations xfs_invis_file_operations = {
 603         .llseek         = generic_file_llseek,
 604         .read           = do_sync_read,
 605         .write          = do_sync_write,
 606         .readv          = xfs_file_readv_invis,
 607         .writev         = xfs_file_writev_invis,
 608         .aio_read       = xfs_file_aio_read_invis,
 609         .aio_write      = xfs_file_aio_write_invis,
 610         .sendfile       = xfs_file_sendfile_invis,
 611         .splice_read    = xfs_file_splice_read_invis,
 612         .splice_write   = xfs_file_splice_write_invis,
 613         .unlocked_ioctl = xfs_file_ioctl_invis,
 614 #ifdef CONFIG_COMPAT
 615         .compat_ioctl   = xfs_file_compat_invis_ioctl,
 616 #endif
 617         .mmap           = xfs_file_mmap,
 618         .open           = xfs_file_open,
 619         .release        = xfs_file_release,
 620         .fsync          = xfs_file_fsync,
 621 };
 622
 623
 624 const struct file_operations xfs_dir_file_operations = {
 625         .read           = generic_read_dir,
 626         .readdir        = xfs_file_readdir,
 627         .unlocked_ioctl = xfs_file_ioctl,
 628 #ifdef CONFIG_COMPAT
 629         .compat_ioctl   = xfs_file_compat_ioctl,
 630 #endif
 631         .fsync          = xfs_file_fsync,
 632 };
 633
 634 static struct vm_operations_struct xfs_file_vm_ops = {
 635         .nopage         = filemap_nopage,
 636         .populate       = filemap_populate,
 637 };
 638
 639 #ifdef CONFIG_XFS_DMAPI
 640 static struct vm_operations_struct xfs_dmapi_file_vm_ops = {
 641         .nopage         = xfs_vm_nopage,
 642         .populate       = filemap_populate,
 643 #ifdef HAVE_VMOP_MPROTECT
 644         .mprotect       = xfs_vm_mprotect,
 645 #endif
 646 };
 647 #endif /* CONFIG_XFS_DMAPI */