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 linvfs_file_vm_ops;
  47 #ifdef CONFIG_XFS_DMAPI
  48 static struct vm_operations_struct linvfs_dmapi_file_vm_ops;
  49 #endif
  50
  51 STATIC inline ssize_t
  52 __linvfs_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 = LINVFS_GET_VP(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 linvfs_aio_read(
  75         struct kiocb            *iocb,
  76         char                    __user *buf,
  77         size_t                  count,
  78         loff_t                  pos)
  79 {
  80         return __linvfs_read(iocb, buf, IO_ISAIO, count, pos);
  81 }
  82
  83 STATIC ssize_t
  84 linvfs_aio_read_invis(
  85         struct kiocb            *iocb,
  86         char                    __user *buf,
  87         size_t                  count,
  88         loff_t                  pos)
  89 {
  90         return __linvfs_read(iocb, buf, IO_ISAIO|IO_INVIS, count, pos);
  91 }
  92
  93
  94 STATIC inline ssize_t
  95 __linvfs_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 = LINVFS_GET_VP(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 linvfs_aio_write(
 119         struct kiocb            *iocb,
 120         const char              __user *buf,
 121         size_t                  count,
 122         loff_t                  pos)
 123 {
 124         return __linvfs_write(iocb, buf, IO_ISAIO, count, pos);
 125 }
 126
 127 STATIC ssize_t
 128 linvfs_aio_write_invis(
 129         struct kiocb            *iocb,
 130         const char              __user *buf,
 131         size_t                  count,
 132         loff_t                  pos)
 133 {
 134         return __linvfs_write(iocb, buf, IO_ISAIO|IO_INVIS, count, pos);
 135 }
 136
 137
 138 STATIC inline ssize_t
 139 __linvfs_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 = LINVFS_GET_VP(inode);
 148         struct kiocb    *kiocb;
 149         ssize_t         rval;
 150
 151         kiocb = kmalloc(sizeof(*kiocb), GFP_KERNEL);
 152         if (unlikely(!kiocb))
 153                 return -ENOMEM;
 154
 155         init_sync_kiocb(kiocb, file);
 156         kiocb->ki_pos = *ppos;
 157
 158         if (unlikely(file->f_flags & O_DIRECT))
 159                 ioflags |= IO_ISDIRECT;
 160         VOP_READ(vp, kiocb, iov, nr_segs, &kiocb->ki_pos, ioflags, NULL, rval);
 161
 162         *ppos = kiocb->ki_pos;
 163         kfree(kiocb);
 164         return rval;
 165 }
 166
 167 STATIC ssize_t
 168 linvfs_readv(
 169         struct file             *file,
 170         const struct iovec      *iov,
 171         unsigned long           nr_segs,
 172         loff_t                  *ppos)
 173 {
 174         return __linvfs_readv(file, iov, 0, nr_segs, ppos);
 175 }
 176
 177 STATIC ssize_t
 178 linvfs_readv_invis(
 179         struct file             *file,
 180         const struct iovec      *iov,
 181         unsigned long           nr_segs,
 182         loff_t                  *ppos)
 183 {
 184         return __linvfs_readv(file, iov, IO_INVIS, nr_segs, ppos);
 185 }
 186
 187
 188 STATIC inline ssize_t
 189 __linvfs_writev(
 190         struct file             *file,
 191         const struct iovec      *iov,
 192         int                     ioflags,
 193         unsigned long           nr_segs,
 194         loff_t                  *ppos)
 195 {
 196         struct inode    *inode = file->f_mapping->host;
 197         vnode_t         *vp = LINVFS_GET_VP(inode);
 198         struct kiocb    *kiocb;
 199         ssize_t         rval;
 200
 201         kiocb = kmalloc(sizeof(*kiocb), GFP_KERNEL);
 202         if (unlikely(!kiocb))
 203                 return -ENOMEM;
 204
 205         init_sync_kiocb(kiocb, file);
 206         kiocb->ki_pos = *ppos;
 207         if (unlikely(file->f_flags & O_DIRECT))
 208                 ioflags |= IO_ISDIRECT;
 209
 210         VOP_WRITE(vp, kiocb, iov, nr_segs, &kiocb->ki_pos, ioflags, NULL, rval);
 211
 212         *ppos = kiocb->ki_pos;
 213         kfree(kiocb);
 214         return rval;
 215 }
 216
 217
 218 STATIC ssize_t
 219 linvfs_writev(
 220         struct file             *file,
 221         const struct iovec      *iov,
 222         unsigned long           nr_segs,
 223         loff_t                  *ppos)
 224 {
 225         return __linvfs_writev(file, iov, 0, nr_segs, ppos);
 226 }
 227
 228 STATIC ssize_t
 229 linvfs_writev_invis(
 230         struct file             *file,
 231         const struct iovec      *iov,
 232         unsigned long           nr_segs,
 233         loff_t                  *ppos)
 234 {
 235         return __linvfs_writev(file, iov, IO_INVIS, nr_segs, ppos);
 236 }
 237
 238 STATIC ssize_t
 239 linvfs_sendfile(
 240         struct file             *filp,
 241         loff_t                  *ppos,
 242         size_t                  count,
 243         read_actor_t            actor,
 244         void                    *target)
 245 {
 246         vnode_t                 *vp = LINVFS_GET_VP(filp->f_dentry->d_inode);
 247         ssize_t                 rval;
 248
 249         VOP_SENDFILE(vp, filp, ppos, 0, count, actor, target, NULL, rval);
 250         return rval;
 251 }
 252
 253
 254 STATIC int
 255 linvfs_open(
 256         struct inode    *inode,
 257         struct file     *filp)
 258 {
 259         vnode_t         *vp = LINVFS_GET_VP(inode);
 260         int             error;
 261
 262         if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
 263                 return -EFBIG;
 264
 265         ASSERT(vp);
 266         VOP_OPEN(vp, NULL, error);
 267         return -error;
 268 }
 269
 270
 271 STATIC int
 272 linvfs_release(
 273         struct inode    *inode,
 274         struct file     *filp)
 275 {
 276         vnode_t         *vp = LINVFS_GET_VP(inode);
 277         int             error = 0;
 278
 279         if (vp)
 280                 VOP_RELEASE(vp, error);
 281         return -error;
 282 }
 283
 284
 285 STATIC int
 286 linvfs_fsync(
 287         struct file     *filp,
 288         struct dentry   *dentry,
 289         int             datasync)
 290 {
 291         struct inode    *inode = dentry->d_inode;
 292         vnode_t         *vp = LINVFS_GET_VP(inode);
 293         int             error;
 294         int             flags = FSYNC_WAIT;
 295
 296         if (datasync)
 297                 flags |= FSYNC_DATA;
 298
 299         ASSERT(vp);
 300         VOP_FSYNC(vp, flags, NULL, (xfs_off_t)0, (xfs_off_t)-1, error);
 301         return -error;
 302 }
 303
 304 /*
 305  * linvfs_readdir maps to VOP_READDIR().
 306  * We need to build a uio, cred, ...
 307  */
 308
 309 #define nextdp(dp)      ((struct xfs_dirent *)((char *)(dp) + (dp)->d_reclen))
 310
 311 #ifdef CONFIG_XFS_DMAPI
 312
 313 STATIC struct page *
 314 linvfs_filemap_nopage(
 315         struct vm_area_struct   *area,
 316         unsigned long           address,
 317         int                     *type)
 318 {
 319         struct inode    *inode = area->vm_file->f_dentry->d_inode;
 320         vnode_t         *vp = LINVFS_GET_VP(inode);
 321         xfs_mount_t     *mp = XFS_VFSTOM(vp->v_vfsp);
 322         int             error;
 323
 324         ASSERT_ALWAYS(vp->v_vfsp->vfs_flag & VFS_DMI);
 325
 326         error = XFS_SEND_MMAP(mp, area, 0);
 327         if (error)
 328                 return NULL;
 329
 330         return filemap_nopage(area, address, type);
 331 }
 332
 333 #endif /* CONFIG_XFS_DMAPI */
 334
 335
 336 STATIC int
 337 linvfs_readdir(
 338         struct file     *filp,
 339         void            *dirent,
 340         filldir_t       filldir)
 341 {
 342         int             error = 0;
 343         vnode_t         *vp;
 344         uio_t           uio;
 345         iovec_t         iov;
 346         int             eof = 0;
 347         caddr_t         read_buf;
 348         int             namelen, size = 0;
 349         size_t          rlen = PAGE_CACHE_SIZE;
 350         xfs_off_t       start_offset, curr_offset;
 351         xfs_dirent_t    *dbp = NULL;
 352
 353         vp = LINVFS_GET_VP(filp->f_dentry->d_inode);
 354         ASSERT(vp);
 355
 356         /* Try fairly hard to get memory */
 357         do {
 358                 if ((read_buf = (caddr_t)kmalloc(rlen, GFP_KERNEL)))
 359                         break;
 360                 rlen >>= 1;
 361         } while (rlen >= 1024);
 362
 363         if (read_buf == NULL)
 364                 return -ENOMEM;
 365
 366         uio.uio_iov = &iov;
 367         uio.uio_segflg = UIO_SYSSPACE;
 368         curr_offset = filp->f_pos;
 369         if (filp->f_pos != 0x7fffffff)
 370                 uio.uio_offset = filp->f_pos;
 371         else
 372                 uio.uio_offset = 0xffffffff;
 373
 374         while (!eof) {
 375                 uio.uio_resid = iov.iov_len = rlen;
 376                 iov.iov_base = read_buf;
 377                 uio.uio_iovcnt = 1;
 378
 379                 start_offset = uio.uio_offset;
 380
 381                 VOP_READDIR(vp, &uio, NULL, &eof, error);
 382                 if ((uio.uio_offset == start_offset) || error) {
 383                         size = 0;
 384                         break;
 385                 }
 386
 387                 size = rlen - uio.uio_resid;
 388                 dbp = (xfs_dirent_t *)read_buf;
 389                 while (size > 0) {
 390                         namelen = strlen(dbp->d_name);
 391
 392                         if (filldir(dirent, dbp->d_name, namelen,
 393                                         (loff_t) curr_offset & 0x7fffffff,
 394                                         (ino_t) dbp->d_ino,
 395                                         DT_UNKNOWN)) {
 396                                 goto done;
 397                         }
 398                         size -= dbp->d_reclen;
 399                         curr_offset = (loff_t)dbp->d_off /* & 0x7fffffff */;
 400                         dbp = nextdp(dbp);
 401                 }
 402         }
 403 done:
 404         if (!error) {
 405                 if (size == 0)
 406                         filp->f_pos = uio.uio_offset & 0x7fffffff;
 407                 else if (dbp)
 408                         filp->f_pos = curr_offset;
 409         }
 410
 411         kfree(read_buf);
 412         return -error;
 413 }
 414
 415
 416 STATIC int
 417 linvfs_file_mmap(
 418         struct file     *filp,
 419         struct vm_area_struct *vma)
 420 {
 421         struct inode    *ip = filp->f_dentry->d_inode;
 422         vnode_t         *vp = LINVFS_GET_VP(ip);
 423         vattr_t         va = { .va_mask = XFS_AT_UPDATIME };
 424         int             error;
 425
 426         vma->vm_ops = &linvfs_file_vm_ops;
 427
 428 #ifdef CONFIG_XFS_DMAPI
 429         if (vp->v_vfsp->vfs_flag & VFS_DMI) {
 430                 vma->vm_ops = &linvfs_dmapi_file_vm_ops;
 431         }
 432 #endif /* CONFIG_XFS_DMAPI */
 433
 434         VOP_SETATTR(vp, &va, XFS_AT_UPDATIME, NULL, error);
 435         if (!error)
 436                 vn_revalidate(vp);      /* update Linux inode flags */
 437         return 0;
 438 }
 439
 440
 441 STATIC long
 442 linvfs_ioctl(
 443         struct file     *filp,
 444         unsigned int    cmd,
 445         unsigned long   arg)
 446 {
 447         int             error;
 448         struct inode    *inode = filp->f_dentry->d_inode;
 449         vnode_t         *vp = LINVFS_GET_VP(inode);
 450
 451         VOP_IOCTL(vp, inode, filp, 0, cmd, (void __user *)arg, error);
 452         VMODIFY(vp);
 453
 454         /* NOTE:  some of the ioctl's return positive #'s as a
 455          *        byte count indicating success, such as
 456          *        readlink_by_handle.  So we don't "sign flip"
 457          *        like most other routines.  This means true
 458          *        errors need to be returned as a negative value.
 459          */
 460         return error;
 461 }
 462
 463 STATIC long
 464 linvfs_ioctl_invis(
 465         struct file     *filp,
 466         unsigned int    cmd,
 467         unsigned long   arg)
 468 {
 469         int             error;
 470         struct inode    *inode = filp->f_dentry->d_inode;
 471         vnode_t         *vp = LINVFS_GET_VP(inode);
 472
 473         ASSERT(vp);
 474         VOP_IOCTL(vp, inode, filp, IO_INVIS, cmd, (void __user *)arg, error);
 475         VMODIFY(vp);
 476
 477         /* NOTE:  some of the ioctl's return positive #'s as a
 478          *        byte count indicating success, such as
 479          *        readlink_by_handle.  So we don't "sign flip"
 480          *        like most other routines.  This means true
 481          *        errors need to be returned as a negative value.
 482          */
 483         return error;
 484 }
 485
 486 #ifdef CONFIG_XFS_DMAPI
 487 #ifdef HAVE_VMOP_MPROTECT
 488 STATIC int
 489 linvfs_mprotect(
 490         struct vm_area_struct *vma,
 491         unsigned int    newflags)
 492 {
 493         vnode_t         *vp = LINVFS_GET_VP(vma->vm_file->f_dentry->d_inode);
 494         int             error = 0;
 495
 496         if (vp->v_vfsp->vfs_flag & VFS_DMI) {
 497                 if ((vma->vm_flags & VM_MAYSHARE) &&
 498                     (newflags & VM_WRITE) && !(vma->vm_flags & VM_WRITE)) {
 499                         xfs_mount_t     *mp = XFS_VFSTOM(vp->v_vfsp);
 500
 501                         error = XFS_SEND_MMAP(mp, vma, VM_WRITE);
 502                     }
 503         }
 504         return error;
 505 }
 506 #endif /* HAVE_VMOP_MPROTECT */
 507 #endif /* CONFIG_XFS_DMAPI */
 508
 509 #ifdef HAVE_FOP_OPEN_EXEC
 510 /* If the user is attempting to execute a file that is offline then
 511  * we have to trigger a DMAPI READ event before the file is marked as busy
 512  * otherwise the invisible I/O will not be able to write to the file to bring
 513  * it back online.
 514  */
 515 STATIC int
 516 linvfs_open_exec(
 517         struct inode    *inode)
 518 {
 519         vnode_t         *vp = LINVFS_GET_VP(inode);
 520         xfs_mount_t     *mp = XFS_VFSTOM(vp->v_vfsp);
 521         int             error = 0;
 522         xfs_inode_t     *ip;
 523
 524         if (vp->v_vfsp->vfs_flag & VFS_DMI) {
 525                 ip = xfs_vtoi(vp);
 526                 if (!ip) {
 527                         error = -EINVAL;
 528                         goto open_exec_out;
 529                 }
 530                 if (DM_EVENT_ENABLED(vp->v_vfsp, ip, DM_EVENT_READ)) {
 531                         error = -XFS_SEND_DATA(mp, DM_EVENT_READ, vp,
 532                                                0, 0, 0, NULL);
 533                 }
 534         }
 535 open_exec_out:
 536         return error;
 537 }
 538 #endif /* HAVE_FOP_OPEN_EXEC */
 539
 540 struct file_operations linvfs_file_operations = {
 541         .llseek         = generic_file_llseek,
 542         .read           = do_sync_read,
 543         .write          = do_sync_write,
 544         .readv          = linvfs_readv,
 545         .writev         = linvfs_writev,
 546         .aio_read       = linvfs_aio_read,
 547         .aio_write      = linvfs_aio_write,
 548         .sendfile       = linvfs_sendfile,
 549         .unlocked_ioctl = linvfs_ioctl,
 550 #ifdef CONFIG_COMPAT
 551         .compat_ioctl   = linvfs_compat_ioctl,
 552 #endif
 553         .mmap           = linvfs_file_mmap,
 554         .open           = linvfs_open,
 555         .release        = linvfs_release,
 556         .fsync          = linvfs_fsync,
 557 #ifdef HAVE_FOP_OPEN_EXEC
 558         .open_exec      = linvfs_open_exec,
 559 #endif
 560 };
 561
 562 struct file_operations linvfs_invis_file_operations = {
 563         .llseek         = generic_file_llseek,
 564         .read           = do_sync_read,
 565         .write          = do_sync_write,
 566         .readv          = linvfs_readv_invis,
 567         .writev         = linvfs_writev_invis,
 568         .aio_read       = linvfs_aio_read_invis,
 569         .aio_write      = linvfs_aio_write_invis,
 570         .sendfile       = linvfs_sendfile,
 571         .unlocked_ioctl = linvfs_ioctl_invis,
 572 #ifdef CONFIG_COMPAT
 573         .compat_ioctl   = linvfs_compat_invis_ioctl,
 574 #endif
 575         .mmap           = linvfs_file_mmap,
 576         .open           = linvfs_open,
 577         .release        = linvfs_release,
 578         .fsync          = linvfs_fsync,
 579 };
 580
 581
 582 struct file_operations linvfs_dir_operations = {
 583         .read           = generic_read_dir,
 584         .readdir        = linvfs_readdir,
 585         .unlocked_ioctl = linvfs_ioctl,
 586 #ifdef CONFIG_COMPAT
 587         .compat_ioctl   = linvfs_compat_ioctl,
 588 #endif
 589         .fsync          = linvfs_fsync,
 590 };
 591
 592 static struct vm_operations_struct linvfs_file_vm_ops = {
 593         .nopage         = filemap_nopage,
 594         .populate       = filemap_populate,
 595 };
 596
 597 #ifdef CONFIG_XFS_DMAPI
 598 static struct vm_operations_struct linvfs_dmapi_file_vm_ops = {
 599         .nopage         = linvfs_filemap_nopage,
 600         .populate       = filemap_populate,
 601 #ifdef HAVE_VMOP_MPROTECT
 602         .mprotect       = linvfs_mprotect,
 603 #endif
 604 };
 605 #endif /* CONFIG_XFS_DMAPI */