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