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 #include "xfs_vnodeops.h"
  41
  42 #include <linux/dcache.h>
  43 #include <linux/smp_lock.h>
  44
  45 static struct vm_operations_struct xfs_file_vm_ops;
  46 #ifdef CONFIG_XFS_DMAPI
  47 static struct vm_operations_struct xfs_dmapi_file_vm_ops;
  48 #endif
  49
  50 STATIC_INLINE ssize_t
  51 __xfs_file_read(
  52         struct kiocb            *iocb,
  53         const struct iovec      *iov,
  54         unsigned long           nr_segs,
  55         int                     ioflags,
  56         loff_t                  pos)
  57 {
  58         struct file             *file = iocb->ki_filp;
  59
  60         BUG_ON(iocb->ki_pos != pos);
  61         if (unlikely(file->f_flags & O_DIRECT))
  62                 ioflags |= IO_ISDIRECT;
  63         return xfs_read(XFS_I(file->f_path.dentry->d_inode), iocb, iov,
  64                                 nr_segs, &iocb->ki_pos, ioflags);
  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
  97         BUG_ON(iocb->ki_pos != pos);
  98         if (unlikely(file->f_flags & O_DIRECT))
  99                 ioflags |= IO_ISDIRECT;
 100         return xfs_write(XFS_I(file->f_mapping->host), iocb, iov, nr_segs,
 101                                 &iocb->ki_pos, ioflags);
 102 }
 103
 104 STATIC ssize_t
 105 xfs_file_aio_write(
 106         struct kiocb            *iocb,
 107         const struct iovec      *iov,
 108         unsigned long           nr_segs,
 109         loff_t                  pos)
 110 {
 111         return __xfs_file_write(iocb, iov, nr_segs, IO_ISAIO, pos);
 112 }
 113
 114 STATIC ssize_t
 115 xfs_file_aio_write_invis(
 116         struct kiocb            *iocb,
 117         const struct iovec      *iov,
 118         unsigned long           nr_segs,
 119         loff_t                  pos)
 120 {
 121         return __xfs_file_write(iocb, iov, nr_segs, IO_ISAIO|IO_INVIS, pos);
 122 }
 123
 124 STATIC ssize_t
 125 xfs_file_splice_read(
 126         struct file             *infilp,
 127         loff_t                  *ppos,
 128         struct pipe_inode_info  *pipe,
 129         size_t                  len,
 130         unsigned int            flags)
 131 {
 132         return xfs_splice_read(XFS_I(infilp->f_path.dentry->d_inode),
 133                                    infilp, ppos, pipe, len, flags, 0);
 134 }
 135
 136 STATIC ssize_t
 137 xfs_file_splice_read_invis(
 138         struct file             *infilp,
 139         loff_t                  *ppos,
 140         struct pipe_inode_info  *pipe,
 141         size_t                  len,
 142         unsigned int            flags)
 143 {
 144         return xfs_splice_read(XFS_I(infilp->f_path.dentry->d_inode),
 145                                    infilp, ppos, pipe, len, flags, IO_INVIS);
 146 }
 147
 148 STATIC ssize_t
 149 xfs_file_splice_write(
 150         struct pipe_inode_info  *pipe,
 151         struct file             *outfilp,
 152         loff_t                  *ppos,
 153         size_t                  len,
 154         unsigned int            flags)
 155 {
 156         return xfs_splice_write(XFS_I(outfilp->f_path.dentry->d_inode),
 157                                     pipe, outfilp, ppos, len, flags, 0);
 158 }
 159
 160 STATIC ssize_t
 161 xfs_file_splice_write_invis(
 162         struct pipe_inode_info  *pipe,
 163         struct file             *outfilp,
 164         loff_t                  *ppos,
 165         size_t                  len,
 166         unsigned int            flags)
 167 {
 168         return xfs_splice_write(XFS_I(outfilp->f_path.dentry->d_inode),
 169                                     pipe, outfilp, ppos, len, flags, IO_INVIS);
 170 }
 171
 172 STATIC int
 173 xfs_file_open(
 174         struct inode    *inode,
 175         struct file     *filp)
 176 {
 177         if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
 178                 return -EFBIG;
 179         return -xfs_open(XFS_I(inode));
 180 }
 181
 182 STATIC int
 183 xfs_file_release(
 184         struct inode    *inode,
 185         struct file     *filp)
 186 {
 187         return -xfs_release(XFS_I(inode));
 188 }
 189
 190 STATIC int
 191 xfs_file_fsync(
 192         struct file     *filp,
 193         struct dentry   *dentry,
 194         int             datasync)
 195 {
 196         int             flags = FSYNC_WAIT;
 197
 198         if (datasync)
 199                 flags |= FSYNC_DATA;
 200         xfs_iflags_clear(XFS_I(dentry->d_inode), XFS_ITRUNCATED);
 201         return -xfs_fsync(XFS_I(dentry->d_inode), flags,
 202                         (xfs_off_t)0, (xfs_off_t)-1);
 203 }
 204
 205 #ifdef CONFIG_XFS_DMAPI
 206 STATIC int
 207 xfs_vm_fault(
 208         struct vm_area_struct   *vma,
 209         struct vm_fault *vmf)
 210 {
 211         struct inode    *inode = vma->vm_file->f_path.dentry->d_inode;
 212         bhv_vnode_t     *vp = vn_from_inode(inode);
 213
 214         ASSERT_ALWAYS(vp->v_vfsp->vfs_flag & VFS_DMI);
 215         if (XFS_SEND_MMAP(XFS_VFSTOM(vp->v_vfsp), vma, 0))
 216                 return VM_FAULT_SIGBUS;
 217         return filemap_fault(vma, vmf);
 218 }
 219 #endif /* CONFIG_XFS_DMAPI */
 220
 221 STATIC int
 222 xfs_file_readdir(
 223         struct file     *filp,
 224         void            *dirent,
 225         filldir_t       filldir)
 226 {
 227         struct inode    *inode = filp->f_path.dentry->d_inode;
 228         xfs_inode_t     *ip = XFS_I(inode);
 229         int             error;
 230         size_t          bufsize;
 231
 232         /*
 233          * The Linux API doesn't pass down the total size of the buffer
 234          * we read into down to the filesystem.  With the filldir concept
 235          * it's not needed for correct information, but the XFS dir2 leaf
 236          * code wants an estimate of the buffer size to calculate it's
 237          * readahead window and size the buffers used for mapping to
 238          * physical blocks.
 239          *
 240          * Try to give it an estimate that's good enough, maybe at some
 241          * point we can change the ->readdir prototype to include the
 242          * buffer size.
 243          */
 244         bufsize = (size_t)min_t(loff_t, PAGE_SIZE, inode->i_size);
 245
 246         error = xfs_readdir(ip, dirent, bufsize,
 247                                 (xfs_off_t *)&filp->f_pos, filldir);
 248         if (error)
 249                 return -error;
 250         return 0;
 251 }
 252
 253 STATIC int
 254 xfs_file_mmap(
 255         struct file     *filp,
 256         struct vm_area_struct *vma)
 257 {
 258         vma->vm_ops = &xfs_file_vm_ops;
 259         vma->vm_flags |= VM_CAN_NONLINEAR;
 260
 261 #ifdef CONFIG_XFS_DMAPI
 262         if (XFS_M(filp->f_path.dentry->d_inode->i_sb)->m_flags & XFS_MOUNT_DMAPI)
 263                 vma->vm_ops = &xfs_dmapi_file_vm_ops;
 264 #endif /* CONFIG_XFS_DMAPI */
 265
 266         file_accessed(filp);
 267         return 0;
 268 }
 269
 270 STATIC long
 271 xfs_file_ioctl(
 272         struct file     *filp,
 273         unsigned int    cmd,
 274         unsigned long   p)
 275 {
 276         int             error;
 277         struct inode    *inode = filp->f_path.dentry->d_inode;
 278
 279         error = xfs_ioctl(XFS_I(inode), filp, 0, cmd, (void __user *)p);
 280         xfs_iflags_set(XFS_I(inode), XFS_IMODIFIED);
 281
 282         /* NOTE:  some of the ioctl's return positive #'s as a
 283          *        byte count indicating success, such as
 284          *        readlink_by_handle.  So we don't "sign flip"
 285          *        like most other routines.  This means true
 286          *        errors need to be returned as a negative value.
 287          */
 288         return error;
 289 }
 290
 291 STATIC long
 292 xfs_file_ioctl_invis(
 293         struct file     *filp,
 294         unsigned int    cmd,
 295         unsigned long   p)
 296 {
 297         int             error;
 298         struct inode    *inode = filp->f_path.dentry->d_inode;
 299
 300         error = xfs_ioctl(XFS_I(inode), filp, IO_INVIS, cmd, (void __user *)p);
 301         xfs_iflags_set(XFS_I(inode), XFS_IMODIFIED);
 302
 303         /* NOTE:  some of the ioctl's return positive #'s as a
 304          *        byte count indicating success, such as
 305          *        readlink_by_handle.  So we don't "sign flip"
 306          *        like most other routines.  This means true
 307          *        errors need to be returned as a negative value.
 308          */
 309         return error;
 310 }
 311
 312 #ifdef CONFIG_XFS_DMAPI
 313 #ifdef HAVE_VMOP_MPROTECT
 314 STATIC int
 315 xfs_vm_mprotect(
 316         struct vm_area_struct *vma,
 317         unsigned int    newflags)
 318 {
 319         struct inode    *inode = vma->vm_file->f_path.dentry->d_inode;
 320         struct xfs_mount *mp = XFS_M(inode->i_sb);
 321         int             error = 0;
 322
 323         if (mp->m_flags & XFS_MOUNT_DMAPI) {
 324                 if ((vma->vm_flags & VM_MAYSHARE) &&
 325                     (newflags & VM_WRITE) && !(vma->vm_flags & VM_WRITE))
 326                         error = XFS_SEND_MMAP(mp, vma, VM_WRITE);
 327         }
 328         return error;
 329 }
 330 #endif /* HAVE_VMOP_MPROTECT */
 331 #endif /* CONFIG_XFS_DMAPI */
 332
 333 #ifdef HAVE_FOP_OPEN_EXEC
 334 /* If the user is attempting to execute a file that is offline then
 335  * we have to trigger a DMAPI READ event before the file is marked as busy
 336  * otherwise the invisible I/O will not be able to write to the file to bring
 337  * it back online.
 338  */
 339 STATIC int
 340 xfs_file_open_exec(
 341         struct inode    *inode)
 342 {
 343         struct xfs_mount *mp = XFS_M(inode->i_sb);
 344
 345         if (unlikely(mp->m_flags & XFS_MOUNT_DMAPI)) {
 346                 if (DM_EVENT_ENABLED(XFS_I(inode), DM_EVENT_READ)) {
 347                         bhv_vnode_t *vp = vn_from_inode(inode);
 348
 349                         return -XFS_SEND_DATA(mp, DM_EVENT_READ,
 350                                                 vp, 0, 0, 0, NULL);
 351                 }
 352         }
 353
 354         return 0;
 355 }
 356 #endif /* HAVE_FOP_OPEN_EXEC */
 357
 358 /*
 359  * mmap()d file has taken write protection fault and is being made
 360  * writable. We can set the page state up correctly for a writable
 361  * page, which means we can do correct delalloc accounting (ENOSPC
 362  * checking!) and unwritten extent mapping.
 363  */
 364 STATIC int
 365 xfs_vm_page_mkwrite(
 366         struct vm_area_struct   *vma,
 367         struct page             *page)
 368 {
 369         return block_page_mkwrite(vma, page, xfs_get_blocks);
 370 }
 371
 372 const struct file_operations xfs_file_operations = {
 373         .llseek         = generic_file_llseek,
 374         .read           = do_sync_read,
 375         .write          = do_sync_write,
 376         .aio_read       = xfs_file_aio_read,
 377         .aio_write      = xfs_file_aio_write,
 378         .splice_read    = xfs_file_splice_read,
 379         .splice_write   = xfs_file_splice_write,
 380         .unlocked_ioctl = xfs_file_ioctl,
 381 #ifdef CONFIG_COMPAT
 382         .compat_ioctl   = xfs_file_compat_ioctl,
 383 #endif
 384         .mmap           = xfs_file_mmap,
 385         .open           = xfs_file_open,
 386         .release        = xfs_file_release,
 387         .fsync          = xfs_file_fsync,
 388 #ifdef HAVE_FOP_OPEN_EXEC
 389         .open_exec      = xfs_file_open_exec,
 390 #endif
 391 };
 392
 393 const struct file_operations xfs_invis_file_operations = {
 394         .llseek         = generic_file_llseek,
 395         .read           = do_sync_read,
 396         .write          = do_sync_write,
 397         .aio_read       = xfs_file_aio_read_invis,
 398         .aio_write      = xfs_file_aio_write_invis,
 399         .splice_read    = xfs_file_splice_read_invis,
 400         .splice_write   = xfs_file_splice_write_invis,
 401         .unlocked_ioctl = xfs_file_ioctl_invis,
 402 #ifdef CONFIG_COMPAT
 403         .compat_ioctl   = xfs_file_compat_invis_ioctl,
 404 #endif
 405         .mmap           = xfs_file_mmap,
 406         .open           = xfs_file_open,
 407         .release        = xfs_file_release,
 408         .fsync          = xfs_file_fsync,
 409 };
 410
 411
 412 const struct file_operations xfs_dir_file_operations = {
 413         .read           = generic_read_dir,
 414         .readdir        = xfs_file_readdir,
 415         .unlocked_ioctl = xfs_file_ioctl,
 416 #ifdef CONFIG_COMPAT
 417         .compat_ioctl   = xfs_file_compat_ioctl,
 418 #endif
 419         .fsync          = xfs_file_fsync,
 420 };
 421
 422 static struct vm_operations_struct xfs_file_vm_ops = {
 423         .fault          = filemap_fault,
 424         .page_mkwrite   = xfs_vm_page_mkwrite,
 425 };
 426
 427 #ifdef CONFIG_XFS_DMAPI
 428 static struct vm_operations_struct xfs_dmapi_file_vm_ops = {
 429         .fault          = xfs_vm_fault,
 430         .page_mkwrite   = xfs_vm_page_mkwrite,
 431 #ifdef HAVE_VMOP_MPROTECT
 432         .mprotect       = xfs_vm_mprotect,
 433 #endif
 434 };
 435 #endif /* CONFIG_XFS_DMAPI */