fs/xfs/linux-2.6/xfs_file.c

   1 /*
   2  * Copyright (c) 2000-2005 Silicon Graphics, Inc.  All Rights Reserved.
   3  *
   4  * This program is free software; you can redistribute it and/or modify it
   5  * under the terms of version 2 of the GNU General Public License as
   6  * published by the Free Software Foundation.
   7  *
   8  * This program is distributed in the hope that it would be useful, but
   9  * WITHOUT ANY WARRANTY; without even the implied warranty of
  10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
  11  *
  12  * Further, this software is distributed without any warranty that it is
  13  * free of the rightful claim of any third person regarding infringement
  14  * or the like.  Any license provided herein, whether implied or
  15  * otherwise, applies only to this software file.  Patent licenses, if
  16  * any, provided herein do not apply to combinations of this program with
  17  * other software, or any other product whatsoever.
  18  *
  19  * You should have received a copy of the GNU General Public License along
  20  * with this program; if not, write the Free Software Foundation, Inc., 59
  21  * Temple Place - Suite 330, Boston MA 02111-1307, USA.
  22  *
  23  * Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy,
  24  * Mountain View, CA  94043, or:
  25  *
  26  * http://www.sgi.com
  27  *
  28  * For further information regarding this notice, see:
  29  *
  30  * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
  31  */
  32
  33 #include "xfs.h"
  34 #include "xfs_inum.h"
  35 #include "xfs_log.h"
  36 #include "xfs_sb.h"
  37 #include "xfs_dir.h"
  38 #include "xfs_dir2.h"
  39 #include "xfs_trans.h"
  40 #include "xfs_dmapi.h"
  41 #include "xfs_mount.h"
  42 #include "xfs_bmap_btree.h"
  43 #include "xfs_alloc_btree.h"
  44 #include "xfs_ialloc_btree.h"
  45 #include "xfs_alloc.h"
  46 #include "xfs_btree.h"
  47 #include "xfs_attr_sf.h"
  48 #include "xfs_dir_sf.h"
  49 #include "xfs_dir2_sf.h"
  50 #include "xfs_dinode.h"
  51 #include "xfs_inode.h"
  52 #include "xfs_error.h"
  53 #include "xfs_rw.h"
  54 #include "xfs_ioctl32.h"
  55
  56 #include <linux/dcache.h>
  57 #include <linux/smp_lock.h>
  58
  59 static struct vm_operations_struct linvfs_file_vm_ops;
  60
  61
  62 STATIC inline ssize_t
  63 __linvfs_read(
  64         struct kiocb            *iocb,
  65         char                    __user *buf,
  66         int                     ioflags,
  67         size_t                  count,
  68         loff_t                  pos)
  69 {
  70         struct iovec            iov = {buf, count};
  71         struct file             *file = iocb->ki_filp;
  72         vnode_t                 *vp = LINVFS_GET_VP(file->f_dentry->d_inode);
  73         ssize_t                 rval;
  74
  75         BUG_ON(iocb->ki_pos != pos);
  76
  77         if (unlikely(file->f_flags & O_DIRECT))
  78                 ioflags |= IO_ISDIRECT;
  79         VOP_READ(vp, iocb, &iov, 1, &iocb->ki_pos, ioflags, NULL, rval);
  80         return rval;
  81 }
  82
  83
  84 STATIC ssize_t
  85 linvfs_aio_read(
  86         struct kiocb            *iocb,
  87         char                    __user *buf,
  88         size_t                  count,
  89         loff_t                  pos)
  90 {
  91         return __linvfs_read(iocb, buf, IO_ISAIO, count, pos);
  92 }
  93
  94 STATIC ssize_t
  95 linvfs_aio_read_invis(
  96         struct kiocb            *iocb,
  97         char                    __user *buf,
  98         size_t                  count,
  99         loff_t                  pos)
 100 {
 101         return __linvfs_read(iocb, buf, IO_ISAIO|IO_INVIS, count, pos);
 102 }
 103
 104
 105 STATIC inline ssize_t
 106 __linvfs_write(
 107         struct kiocb    *iocb,
 108         const char      __user *buf,
 109         int             ioflags,
 110         size_t          count,
 111         loff_t          pos)
 112 {
 113         struct iovec    iov = {(void __user *)buf, count};
 114         struct file     *file = iocb->ki_filp;
 115         struct inode    *inode = file->f_mapping->host;
 116         vnode_t         *vp = LINVFS_GET_VP(inode);
 117         ssize_t         rval;
 118
 119         BUG_ON(iocb->ki_pos != pos);
 120         if (unlikely(file->f_flags & O_DIRECT))
 121                 ioflags |= IO_ISDIRECT;
 122
 123         VOP_WRITE(vp, iocb, &iov, 1, &iocb->ki_pos, ioflags, NULL, rval);
 124         return rval;
 125 }
 126
 127
 128 STATIC ssize_t
 129 linvfs_aio_write(
 130         struct kiocb            *iocb,
 131         const char              __user *buf,
 132         size_t                  count,
 133         loff_t                  pos)
 134 {
 135         return __linvfs_write(iocb, buf, IO_ISAIO, count, pos);
 136 }
 137
 138 STATIC ssize_t
 139 linvfs_aio_write_invis(
 140         struct kiocb            *iocb,
 141         const char              __user *buf,
 142         size_t                  count,
 143         loff_t                  pos)
 144 {
 145         return __linvfs_write(iocb, buf, IO_ISAIO|IO_INVIS, count, pos);
 146 }
 147
 148
 149 STATIC inline ssize_t
 150 __linvfs_readv(
 151         struct file             *file,
 152         const struct iovec      *iov,
 153         int                     ioflags,
 154         unsigned long           nr_segs,
 155         loff_t                  *ppos)
 156 {
 157         struct inode    *inode = file->f_mapping->host;
 158         vnode_t         *vp = LINVFS_GET_VP(inode);
 159         struct          kiocb kiocb;
 160         ssize_t         rval;
 161
 162         init_sync_kiocb(&kiocb, file);
 163         kiocb.ki_pos = *ppos;
 164
 165         if (unlikely(file->f_flags & O_DIRECT))
 166                 ioflags |= IO_ISDIRECT;
 167         VOP_READ(vp, &kiocb, iov, nr_segs, &kiocb.ki_pos, ioflags, NULL, rval);
 168
 169         *ppos = kiocb.ki_pos;
 170         return rval;
 171 }
 172
 173 STATIC ssize_t
 174 linvfs_readv(
 175         struct file             *file,
 176         const struct iovec      *iov,
 177         unsigned long           nr_segs,
 178         loff_t                  *ppos)
 179 {
 180         return __linvfs_readv(file, iov, 0, nr_segs, ppos);
 181 }
 182
 183 STATIC ssize_t
 184 linvfs_readv_invis(
 185         struct file             *file,
 186         const struct iovec      *iov,
 187         unsigned long           nr_segs,
 188         loff_t                  *ppos)
 189 {
 190         return __linvfs_readv(file, iov, IO_INVIS, nr_segs, ppos);
 191 }
 192
 193
 194 STATIC inline ssize_t
 195 __linvfs_writev(
 196         struct file             *file,
 197         const struct iovec      *iov,
 198         int                     ioflags,
 199         unsigned long           nr_segs,
 200         loff_t                  *ppos)
 201 {
 202         struct inode    *inode = file->f_mapping->host;
 203         vnode_t         *vp = LINVFS_GET_VP(inode);
 204         struct          kiocb kiocb;
 205         ssize_t         rval;
 206
 207         init_sync_kiocb(&kiocb, file);
 208         kiocb.ki_pos = *ppos;
 209         if (unlikely(file->f_flags & O_DIRECT))
 210                 ioflags |= IO_ISDIRECT;
 211
 212         VOP_WRITE(vp, &kiocb, iov, nr_segs, &kiocb.ki_pos, ioflags, NULL, rval);
 213
 214         *ppos = kiocb.ki_pos;
 215         return rval;
 216 }
 217
 218
 219 STATIC ssize_t
 220 linvfs_writev(
 221         struct file             *file,
 222         const struct iovec      *iov,
 223         unsigned long           nr_segs,
 224         loff_t                  *ppos)
 225 {
 226         return __linvfs_writev(file, iov, 0, nr_segs, ppos);
 227 }
 228
 229 STATIC ssize_t
 230 linvfs_writev_invis(
 231         struct file             *file,
 232         const struct iovec      *iov,
 233         unsigned long           nr_segs,
 234         loff_t                  *ppos)
 235 {
 236         return __linvfs_writev(file, iov, IO_INVIS, nr_segs, ppos);
 237 }
 238
 239 STATIC ssize_t
 240 linvfs_sendfile(
 241         struct file             *filp,
 242         loff_t                  *ppos,
 243         size_t                  count,
 244         read_actor_t            actor,
 245         void                    *target)
 246 {
 247         vnode_t                 *vp = LINVFS_GET_VP(filp->f_dentry->d_inode);
 248         ssize_t                 rval;
 249
 250         VOP_SENDFILE(vp, filp, ppos, 0, count, actor, target, NULL, rval);
 251         return rval;
 252 }
 253
 254
 255 STATIC int
 256 linvfs_open(
 257         struct inode    *inode,
 258         struct file     *filp)
 259 {
 260         vnode_t         *vp = LINVFS_GET_VP(inode);
 261         int             error;
 262
 263         if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
 264                 return -EFBIG;
 265
 266         ASSERT(vp);
 267         VOP_OPEN(vp, NULL, error);
 268         return -error;
 269 }
 270
 271
 272 STATIC int
 273 linvfs_release(
 274         struct inode    *inode,
 275         struct file     *filp)
 276 {
 277         vnode_t         *vp = LINVFS_GET_VP(inode);
 278         int             error = 0;
 279
 280         if (vp)
 281                 VOP_RELEASE(vp, error);
 282         return -error;
 283 }
 284
 285
 286 STATIC int
 287 linvfs_fsync(
 288         struct file     *filp,
 289         struct dentry   *dentry,
 290         int             datasync)
 291 {
 292         struct inode    *inode = dentry->d_inode;
 293         vnode_t         *vp = LINVFS_GET_VP(inode);
 294         int             error;
 295         int             flags = FSYNC_WAIT;
 296
 297         if (datasync)
 298                 flags |= FSYNC_DATA;
 299
 300         ASSERT(vp);
 301         VOP_FSYNC(vp, flags, NULL, (xfs_off_t)0, (xfs_off_t)-1, error);
 302         return -error;
 303 }
 304
 305 /*
 306  * linvfs_readdir maps to VOP_READDIR().
 307  * We need to build a uio, cred, ...
 308  */
 309
 310 #define nextdp(dp)      ((struct xfs_dirent *)((char *)(dp) + (dp)->d_reclen))
 311
 312 STATIC int
 313 linvfs_readdir(
 314         struct file     *filp,
 315         void            *dirent,
 316         filldir_t       filldir)
 317 {
 318         int             error = 0;
 319         vnode_t         *vp;
 320         uio_t           uio;
 321         iovec_t         iov;
 322         int             eof = 0;
 323         caddr_t         read_buf;
 324         int             namelen, size = 0;
 325         size_t          rlen = PAGE_CACHE_SIZE;
 326         xfs_off_t       start_offset, curr_offset;
 327         xfs_dirent_t    *dbp = NULL;
 328
 329         vp = LINVFS_GET_VP(filp->f_dentry->d_inode);
 330         ASSERT(vp);
 331
 332         /* Try fairly hard to get memory */
 333         do {
 334                 if ((read_buf = (caddr_t)kmalloc(rlen, GFP_KERNEL)))
 335                         break;
 336                 rlen >>= 1;
 337         } while (rlen >= 1024);
 338
 339         if (read_buf == NULL)
 340                 return -ENOMEM;
 341
 342         uio.uio_iov = &iov;
 343         uio.uio_segflg = UIO_SYSSPACE;
 344         curr_offset = filp->f_pos;
 345         if (filp->f_pos != 0x7fffffff)
 346                 uio.uio_offset = filp->f_pos;
 347         else
 348                 uio.uio_offset = 0xffffffff;
 349
 350         while (!eof) {
 351                 uio.uio_resid = iov.iov_len = rlen;
 352                 iov.iov_base = read_buf;
 353                 uio.uio_iovcnt = 1;
 354
 355                 start_offset = uio.uio_offset;
 356
 357                 VOP_READDIR(vp, &uio, NULL, &eof, error);
 358                 if ((uio.uio_offset == start_offset) || error) {
 359                         size = 0;
 360                         break;
 361                 }
 362
 363                 size = rlen - uio.uio_resid;
 364                 dbp = (xfs_dirent_t *)read_buf;
 365                 while (size > 0) {
 366                         namelen = strlen(dbp->d_name);
 367
 368                         if (filldir(dirent, dbp->d_name, namelen,
 369                                         (loff_t) curr_offset & 0x7fffffff,
 370                                         (ino_t) dbp->d_ino,
 371                                         DT_UNKNOWN)) {
 372                                 goto done;
 373                         }
 374                         size -= dbp->d_reclen;
 375                         curr_offset = (loff_t)dbp->d_off /* & 0x7fffffff */;
 376                         dbp = nextdp(dbp);
 377                 }
 378         }
 379 done:
 380         if (!error) {
 381                 if (size == 0)
 382                         filp->f_pos = uio.uio_offset & 0x7fffffff;
 383                 else if (dbp)
 384                         filp->f_pos = curr_offset;
 385         }
 386
 387         kfree(read_buf);
 388         return -error;
 389 }
 390
 391
 392 STATIC int
 393 linvfs_file_mmap(
 394         struct file     *filp,
 395         struct vm_area_struct *vma)
 396 {
 397         struct inode    *ip = filp->f_dentry->d_inode;
 398         vnode_t         *vp = LINVFS_GET_VP(ip);
 399         vattr_t         va = { .va_mask = XFS_AT_UPDATIME };
 400         int             error;
 401
 402         if (vp->v_vfsp->vfs_flag & VFS_DMI) {
 403                 xfs_mount_t     *mp = XFS_VFSTOM(vp->v_vfsp);
 404
 405                 error = -XFS_SEND_MMAP(mp, vma, 0);
 406                 if (error)
 407                         return error;
 408         }
 409
 410         vma->vm_ops = &linvfs_file_vm_ops;
 411
 412         VOP_SETATTR(vp, &va, XFS_AT_UPDATIME, NULL, error);
 413         if (!error)
 414                 vn_revalidate(vp);      /* update Linux inode flags */
 415         return 0;
 416 }
 417
 418
 419 STATIC long
 420 linvfs_ioctl(
 421         struct file     *filp,
 422         unsigned int    cmd,
 423         unsigned long   arg)
 424 {
 425         int             error;
 426         struct inode *inode = filp->f_dentry->d_inode;
 427         vnode_t         *vp = LINVFS_GET_VP(inode);
 428
 429         VOP_IOCTL(vp, inode, filp, 0, cmd, (void __user *)arg, error);
 430         VMODIFY(vp);
 431
 432         /* NOTE:  some of the ioctl's return positive #'s as a
 433          *        byte count indicating success, such as
 434          *        readlink_by_handle.  So we don't "sign flip"
 435          *        like most other routines.  This means true
 436          *        errors need to be returned as a negative value.
 437          */
 438         return error;
 439 }
 440
 441 STATIC long
 442 linvfs_ioctl_invis(
 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         ASSERT(vp);
 452         VOP_IOCTL(vp, inode, filp, IO_INVIS, cmd, (void __user *)arg, error);
 453         VMODIFY(vp);
 454
 455         /* NOTE:  some of the ioctl's return positive #'s as a
 456          *        byte count indicating success, such as
 457          *        readlink_by_handle.  So we don't "sign flip"
 458          *        like most other routines.  This means true
 459          *        errors need to be returned as a negative value.
 460          */
 461         return error;
 462 }
 463
 464 #ifdef HAVE_VMOP_MPROTECT
 465 STATIC int
 466 linvfs_mprotect(
 467         struct vm_area_struct *vma,
 468         unsigned int    newflags)
 469 {
 470         vnode_t         *vp = LINVFS_GET_VP(vma->vm_file->f_dentry->d_inode);
 471         int             error = 0;
 472
 473         if (vp->v_vfsp->vfs_flag & VFS_DMI) {
 474                 if ((vma->vm_flags & VM_MAYSHARE) &&
 475                     (newflags & VM_WRITE) && !(vma->vm_flags & VM_WRITE)) {
 476                         xfs_mount_t     *mp = XFS_VFSTOM(vp->v_vfsp);
 477
 478                         error = XFS_SEND_MMAP(mp, vma, VM_WRITE);
 479                     }
 480         }
 481         return error;
 482 }
 483 #endif /* HAVE_VMOP_MPROTECT */
 484
 485 #ifdef HAVE_FOP_OPEN_EXEC
 486 /* If the user is attempting to execute a file that is offline then
 487  * we have to trigger a DMAPI READ event before the file is marked as busy
 488  * otherwise the invisible I/O will not be able to write to the file to bring
 489  * it back online.
 490  */
 491 STATIC int
 492 linvfs_open_exec(
 493         struct inode    *inode)
 494 {
 495         vnode_t         *vp = LINVFS_GET_VP(inode);
 496         xfs_mount_t     *mp = XFS_VFSTOM(vp->v_vfsp);
 497         int             error = 0;
 498         bhv_desc_t      *bdp;
 499         xfs_inode_t     *ip;
 500
 501         if (vp->v_vfsp->vfs_flag & VFS_DMI) {
 502                 bdp = vn_bhv_lookup(VN_BHV_HEAD(vp), &xfs_vnodeops);
 503                 if (!bdp) {
 504                         error = -EINVAL;
 505                         goto open_exec_out;
 506                 }
 507                 ip = XFS_BHVTOI(bdp);
 508                 if (DM_EVENT_ENABLED(vp->v_vfsp, ip, DM_EVENT_READ)) {
 509                         error = -XFS_SEND_DATA(mp, DM_EVENT_READ, vp,
 510                                                0, 0, 0, NULL);
 511                 }
 512         }
 513 open_exec_out:
 514         return error;
 515 }
 516 #endif /* HAVE_FOP_OPEN_EXEC */
 517
 518 struct file_operations linvfs_file_operations = {
 519         .llseek         = generic_file_llseek,
 520         .read           = do_sync_read,
 521         .write          = do_sync_write,
 522         .readv          = linvfs_readv,
 523         .writev         = linvfs_writev,
 524         .aio_read       = linvfs_aio_read,
 525         .aio_write      = linvfs_aio_write,
 526         .sendfile       = linvfs_sendfile,
 527         .unlocked_ioctl = linvfs_ioctl,
 528 #ifdef CONFIG_COMPAT
 529         .compat_ioctl   = xfs_compat_ioctl,
 530 #endif
 531         .mmap           = linvfs_file_mmap,
 532         .open           = linvfs_open,
 533         .release        = linvfs_release,
 534         .fsync          = linvfs_fsync,
 535 #ifdef HAVE_FOP_OPEN_EXEC
 536         .open_exec      = linvfs_open_exec,
 537 #endif
 538 };
 539
 540 struct file_operations linvfs_invis_file_operations = {
 541         .llseek         = generic_file_llseek,
 542         .read           = do_sync_read,
 543         .write          = do_sync_write,
 544         .readv          = linvfs_readv_invis,
 545         .writev         = linvfs_writev_invis,
 546         .aio_read       = linvfs_aio_read_invis,
 547         .aio_write      = linvfs_aio_write_invis,
 548         .sendfile       = linvfs_sendfile,
 549         .unlocked_ioctl = linvfs_ioctl_invis,
 550 #ifdef CONFIG_COMPAT
 551         .compat_ioctl   = xfs_compat_invis_ioctl,
 552 #endif
 553         .mmap           = linvfs_file_mmap,
 554         .open           = linvfs_open,
 555         .release        = linvfs_release,
 556         .fsync          = linvfs_fsync,
 557 };
 558
 559
 560 struct file_operations linvfs_dir_operations = {
 561         .read           = generic_read_dir,
 562         .readdir        = linvfs_readdir,
 563         .unlocked_ioctl = linvfs_ioctl,
 564         .fsync          = linvfs_fsync,
 565 };
 566
 567 static struct vm_operations_struct linvfs_file_vm_ops = {
 568         .nopage         = filemap_nopage,
 569         .populate       = filemap_populate,
 570 #ifdef HAVE_VMOP_MPROTECT
 571         .mprotect       = linvfs_mprotect,
 572 #endif
 573 };