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         const struct iovec      *iov,
  53         unsigned long           nr_segs,
  54         int                     ioflags,
  55         loff_t                  pos)
  56 {
  57         struct file             *file = iocb->ki_filp;
  58         bhv_vnode_t             *vp = vn_from_inode(file->f_path.dentry->d_inode);
  59
  60         BUG_ON(iocb->ki_pos != pos);
  61         if (unlikely(file->f_flags & O_DIRECT))
  62                 ioflags |= IO_ISDIRECT;
  63         return bhv_vop_read(vp, iocb, iov, nr_segs, &iocb->ki_pos,
  64                                 ioflags, NULL);
  65 }
  66
  67 STATIC ssize_t
  68 xfs_file_aio_read(
  69         struct kiocb            *iocb,
  70         const struct iovec      *iov,
  71         unsigned long           nr_segs,
  72         loff_t                  pos)
  73 {
  74         return __xfs_file_read(iocb, iov, nr_segs, IO_ISAIO, pos);
  75 }
  76
  77 STATIC ssize_t
  78 xfs_file_aio_read_invis(
  79         struct kiocb            *iocb,
  80         const struct iovec      *iov,
  81         unsigned long           nr_segs,
  82         loff_t                  pos)
  83 {
  84         return __xfs_file_read(iocb, iov, nr_segs, IO_ISAIO|IO_INVIS, pos);
  85 }
  86
  87 STATIC_INLINE ssize_t
  88 __xfs_file_write(
  89         struct kiocb            *iocb,
  90         const struct iovec      *iov,
  91         unsigned long           nr_segs,
  92         int                     ioflags,
  93         loff_t                  pos)
  94 {
  95         struct file     *file = iocb->ki_filp;
  96         struct inode    *inode = file->f_mapping->host;
  97         bhv_vnode_t     *vp = vn_from_inode(inode);
  98
  99         BUG_ON(iocb->ki_pos != pos);
 100         if (unlikely(file->f_flags & O_DIRECT))
 101                 ioflags |= IO_ISDIRECT;
 102         return bhv_vop_write(vp, iocb, iov, nr_segs, &iocb->ki_pos,
 103                                 ioflags, NULL);
 104 }
 105
 106 STATIC ssize_t
 107 xfs_file_aio_write(
 108         struct kiocb            *iocb,
 109         const struct iovec      *iov,
 110         unsigned long           nr_segs,
 111         loff_t                  pos)
 112 {
 113         return __xfs_file_write(iocb, iov, nr_segs, IO_ISAIO, pos);
 114 }
 115
 116 STATIC ssize_t
 117 xfs_file_aio_write_invis(
 118         struct kiocb            *iocb,
 119         const struct iovec      *iov,
 120         unsigned long           nr_segs,
 121         loff_t                  pos)
 122 {
 123         return __xfs_file_write(iocb, iov, nr_segs, IO_ISAIO|IO_INVIS, pos);
 124 }
 125
 126 STATIC ssize_t
 127 xfs_file_splice_read(
 128         struct file             *infilp,
 129         loff_t                  *ppos,
 130         struct pipe_inode_info  *pipe,
 131         size_t                  len,
 132         unsigned int            flags)
 133 {
 134         return bhv_vop_splice_read(vn_from_inode(infilp->f_path.dentry->d_inode),
 135                                    infilp, ppos, pipe, len, flags, 0, NULL);
 136 }
 137
 138 STATIC ssize_t
 139 xfs_file_splice_read_invis(
 140         struct file             *infilp,
 141         loff_t                  *ppos,
 142         struct pipe_inode_info  *pipe,
 143         size_t                  len,
 144         unsigned int            flags)
 145 {
 146         return bhv_vop_splice_read(vn_from_inode(infilp->f_path.dentry->d_inode),
 147                                    infilp, ppos, pipe, len, flags, IO_INVIS,
 148                                    NULL);
 149 }
 150
 151 STATIC ssize_t
 152 xfs_file_splice_write(
 153         struct pipe_inode_info  *pipe,
 154         struct file             *outfilp,
 155         loff_t                  *ppos,
 156         size_t                  len,
 157         unsigned int            flags)
 158 {
 159         return bhv_vop_splice_write(vn_from_inode(outfilp->f_path.dentry->d_inode),
 160                                     pipe, outfilp, ppos, len, flags, 0, NULL);
 161 }
 162
 163 STATIC ssize_t
 164 xfs_file_splice_write_invis(
 165         struct pipe_inode_info  *pipe,
 166         struct file             *outfilp,
 167         loff_t                  *ppos,
 168         size_t                  len,
 169         unsigned int            flags)
 170 {
 171         return bhv_vop_splice_write(vn_from_inode(outfilp->f_path.dentry->d_inode),
 172                                     pipe, outfilp, ppos, len, flags, IO_INVIS,
 173                                     NULL);
 174 }
 175
 176 STATIC int
 177 xfs_file_open(
 178         struct inode    *inode,
 179         struct file     *filp)
 180 {
 181         if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
 182                 return -EFBIG;
 183         return -bhv_vop_open(vn_from_inode(inode), NULL);
 184 }
 185
 186 STATIC int
 187 xfs_file_release(
 188         struct inode    *inode,
 189         struct file     *filp)
 190 {
 191         bhv_vnode_t     *vp = vn_from_inode(inode);
 192
 193         if (vp)
 194                 return -bhv_vop_release(vp);
 195         return 0;
 196 }
 197
 198 STATIC int
 199 xfs_file_fsync(
 200         struct file     *filp,
 201         struct dentry   *dentry,
 202         int             datasync)
 203 {
 204         bhv_vnode_t     *vp = vn_from_inode(dentry->d_inode);
 205         int             flags = FSYNC_WAIT;
 206
 207         if (datasync)
 208                 flags |= FSYNC_DATA;
 209         if (VN_TRUNC(vp))
 210                 VUNTRUNCATE(vp);
 211         return -bhv_vop_fsync(vp, flags, NULL, (xfs_off_t)0, (xfs_off_t)-1);
 212 }
 213
 214 #ifdef CONFIG_XFS_DMAPI
 215 STATIC struct page *
 216 xfs_vm_fault(
 217         struct vm_area_struct   *vma,
 218         struct fault_data       *fdata)
 219 {
 220         struct inode    *inode = vma->vm_file->f_path.dentry->d_inode;
 221         bhv_vnode_t     *vp = vn_from_inode(inode);
 222
 223         ASSERT_ALWAYS(vp->v_vfsp->vfs_flag & VFS_DMI);
 224         if (XFS_SEND_MMAP(XFS_VFSTOM(vp->v_vfsp), vma, 0)) {
 225                 fdata->type = VM_FAULT_SIGBUS;
 226                 return NULL;
 227         }
 228         return filemap_fault(vma, fdata);
 229 }
 230 #endif /* CONFIG_XFS_DMAPI */
 231
 232 STATIC int
 233 xfs_file_readdir(
 234         struct file     *filp,
 235         void            *dirent,
 236         filldir_t       filldir)
 237 {
 238         int             error = 0;
 239         bhv_vnode_t     *vp = vn_from_inode(filp->f_path.dentry->d_inode);
 240         uio_t           uio;
 241         iovec_t         iov;
 242         int             eof = 0;
 243         caddr_t         read_buf;
 244         int             namelen, size = 0;
 245         size_t          rlen = PAGE_CACHE_SIZE;
 246         xfs_off_t       start_offset, curr_offset;
 247         xfs_dirent_t    *dbp = NULL;
 248
 249         /* Try fairly hard to get memory */
 250         do {
 251                 if ((read_buf = kmalloc(rlen, GFP_KERNEL)))
 252                         break;
 253                 rlen >>= 1;
 254         } while (rlen >= 1024);
 255
 256         if (read_buf == NULL)
 257                 return -ENOMEM;
 258
 259         uio.uio_iov = &iov;
 260         uio.uio_segflg = UIO_SYSSPACE;
 261         curr_offset = filp->f_pos;
 262         if (filp->f_pos != 0x7fffffff)
 263                 uio.uio_offset = filp->f_pos;
 264         else
 265                 uio.uio_offset = 0xffffffff;
 266
 267         while (!eof) {
 268                 uio.uio_resid = iov.iov_len = rlen;
 269                 iov.iov_base = read_buf;
 270                 uio.uio_iovcnt = 1;
 271
 272                 start_offset = uio.uio_offset;
 273
 274                 error = bhv_vop_readdir(vp, &uio, NULL, &eof);
 275                 if ((uio.uio_offset == start_offset) || error) {
 276                         size = 0;
 277                         break;
 278                 }
 279
 280                 size = rlen - uio.uio_resid;
 281                 dbp = (xfs_dirent_t *)read_buf;
 282                 while (size > 0) {
 283                         namelen = strlen(dbp->d_name);
 284
 285                         if (filldir(dirent, dbp->d_name, namelen,
 286                                         (loff_t) curr_offset & 0x7fffffff,
 287                                         (ino_t) dbp->d_ino,
 288                                         DT_UNKNOWN)) {
 289                                 goto done;
 290                         }
 291                         size -= dbp->d_reclen;
 292                         curr_offset = (loff_t)dbp->d_off /* & 0x7fffffff */;
 293                         dbp = (xfs_dirent_t *)((char *)dbp + dbp->d_reclen);
 294                 }
 295         }
 296 done:
 297         if (!error) {
 298                 if (size == 0)
 299                         filp->f_pos = uio.uio_offset & 0x7fffffff;
 300                 else if (dbp)
 301                         filp->f_pos = curr_offset;
 302         }
 303
 304         kfree(read_buf);
 305         return -error;
 306 }
 307
 308 STATIC int
 309 xfs_file_mmap(
 310         struct file     *filp,
 311         struct vm_area_struct *vma)
 312 {
 313         vma->vm_ops = &xfs_file_vm_ops;
 314         vma->vm_flags |= VM_CAN_INVALIDATE | VM_CAN_NONLINEAR;
 315
 316 #ifdef CONFIG_XFS_DMAPI
 317         if (vn_from_inode(filp->f_path.dentry->d_inode)->v_vfsp->vfs_flag & VFS_DMI)
 318                 vma->vm_ops = &xfs_dmapi_file_vm_ops;
 319 #endif /* CONFIG_XFS_DMAPI */
 320
 321         file_accessed(filp);
 322         return 0;
 323 }
 324
 325 STATIC long
 326 xfs_file_ioctl(
 327         struct file     *filp,
 328         unsigned int    cmd,
 329         unsigned long   p)
 330 {
 331         int             error;
 332         struct inode    *inode = filp->f_path.dentry->d_inode;
 333         bhv_vnode_t     *vp = vn_from_inode(inode);
 334
 335         error = bhv_vop_ioctl(vp, inode, filp, 0, cmd, (void __user *)p);
 336         VMODIFY(vp);
 337
 338         /* NOTE:  some of the ioctl's return positive #'s as a
 339          *        byte count indicating success, such as
 340          *        readlink_by_handle.  So we don't "sign flip"
 341          *        like most other routines.  This means true
 342          *        errors need to be returned as a negative value.
 343          */
 344         return error;
 345 }
 346
 347 STATIC long
 348 xfs_file_ioctl_invis(
 349         struct file     *filp,
 350         unsigned int    cmd,
 351         unsigned long   p)
 352 {
 353         int             error;
 354         struct inode    *inode = filp->f_path.dentry->d_inode;
 355         bhv_vnode_t     *vp = vn_from_inode(inode);
 356
 357         error = bhv_vop_ioctl(vp, inode, filp, IO_INVIS, cmd, (void __user *)p);
 358         VMODIFY(vp);
 359
 360         /* NOTE:  some of the ioctl's return positive #'s as a
 361          *        byte count indicating success, such as
 362          *        readlink_by_handle.  So we don't "sign flip"
 363          *        like most other routines.  This means true
 364          *        errors need to be returned as a negative value.
 365          */
 366         return error;
 367 }
 368
 369 #ifdef CONFIG_XFS_DMAPI
 370 #ifdef HAVE_VMOP_MPROTECT
 371 STATIC int
 372 xfs_vm_mprotect(
 373         struct vm_area_struct *vma,
 374         unsigned int    newflags)
 375 {
 376         bhv_vnode_t     *vp = vn_from_inode(vma->vm_file->f_path.dentry->d_inode);
 377         int             error = 0;
 378
 379         if (vp->v_vfsp->vfs_flag & VFS_DMI) {
 380                 if ((vma->vm_flags & VM_MAYSHARE) &&
 381                     (newflags & VM_WRITE) && !(vma->vm_flags & VM_WRITE)) {
 382                         xfs_mount_t     *mp = XFS_VFSTOM(vp->v_vfsp);
 383
 384                         error = XFS_SEND_MMAP(mp, vma, VM_WRITE);
 385                     }
 386         }
 387         return error;
 388 }
 389 #endif /* HAVE_VMOP_MPROTECT */
 390 #endif /* CONFIG_XFS_DMAPI */
 391
 392 #ifdef HAVE_FOP_OPEN_EXEC
 393 /* If the user is attempting to execute a file that is offline then
 394  * we have to trigger a DMAPI READ event before the file is marked as busy
 395  * otherwise the invisible I/O will not be able to write to the file to bring
 396  * it back online.
 397  */
 398 STATIC int
 399 xfs_file_open_exec(
 400         struct inode    *inode)
 401 {
 402         bhv_vnode_t     *vp = vn_from_inode(inode);
 403
 404         if (unlikely(vp->v_vfsp->vfs_flag & VFS_DMI)) {
 405                 xfs_mount_t     *mp = XFS_VFSTOM(vp->v_vfsp);
 406                 xfs_inode_t     *ip = xfs_vtoi(vp);
 407
 408                 if (!ip)
 409                         return -EINVAL;
 410                 if (DM_EVENT_ENABLED(vp->v_vfsp, ip, DM_EVENT_READ))
 411                         return -XFS_SEND_DATA(mp, DM_EVENT_READ, vp,
 412                                                0, 0, 0, NULL);
 413         }
 414         return 0;
 415 }
 416 #endif /* HAVE_FOP_OPEN_EXEC */
 417
 418 const struct file_operations xfs_file_operations = {
 419         .llseek         = generic_file_llseek,
 420         .read           = do_sync_read,
 421         .write          = do_sync_write,
 422         .aio_read       = xfs_file_aio_read,
 423         .aio_write      = xfs_file_aio_write,
 424         .splice_read    = xfs_file_splice_read,
 425         .splice_write   = xfs_file_splice_write,
 426         .unlocked_ioctl = xfs_file_ioctl,
 427 #ifdef CONFIG_COMPAT
 428         .compat_ioctl   = xfs_file_compat_ioctl,
 429 #endif
 430         .mmap           = xfs_file_mmap,
 431         .open           = xfs_file_open,
 432         .release        = xfs_file_release,
 433         .fsync          = xfs_file_fsync,
 434 #ifdef HAVE_FOP_OPEN_EXEC
 435         .open_exec      = xfs_file_open_exec,
 436 #endif
 437 };
 438
 439 const struct file_operations xfs_invis_file_operations = {
 440         .llseek         = generic_file_llseek,
 441         .read           = do_sync_read,
 442         .write          = do_sync_write,
 443         .aio_read       = xfs_file_aio_read_invis,
 444         .aio_write      = xfs_file_aio_write_invis,
 445         .splice_read    = xfs_file_splice_read_invis,
 446         .splice_write   = xfs_file_splice_write_invis,
 447         .unlocked_ioctl = xfs_file_ioctl_invis,
 448 #ifdef CONFIG_COMPAT
 449         .compat_ioctl   = xfs_file_compat_invis_ioctl,
 450 #endif
 451         .mmap           = xfs_file_mmap,
 452         .open           = xfs_file_open,
 453         .release        = xfs_file_release,
 454         .fsync          = xfs_file_fsync,
 455 };
 456
 457
 458 const struct file_operations xfs_dir_file_operations = {
 459         .read           = generic_read_dir,
 460         .readdir        = xfs_file_readdir,
 461         .unlocked_ioctl = xfs_file_ioctl,
 462 #ifdef CONFIG_COMPAT
 463         .compat_ioctl   = xfs_file_compat_ioctl,
 464 #endif
 465         .fsync          = xfs_file_fsync,
 466 };
 467
 468 static struct vm_operations_struct xfs_file_vm_ops = {
 469         .fault          = filemap_fault,
 470 };
 471
 472 #ifdef CONFIG_XFS_DMAPI
 473 static struct vm_operations_struct xfs_dmapi_file_vm_ops = {
 474         .fault          = xfs_vm_fault,
 475 #ifdef HAVE_VMOP_MPROTECT
 476         .mprotect       = xfs_vm_mprotect,
 477 #endif
 478 };
 479 #endif /* CONFIG_XFS_DMAPI */