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
  47 STATIC_INLINE ssize_t
  48 __xfs_file_read(
  49         struct kiocb            *iocb,
  50         const struct iovec      *iov,
  51         unsigned long           nr_segs,
  52         int                     ioflags,
  53         loff_t                  pos)
  54 {
  55         struct file             *file = iocb->ki_filp;
  56
  57         BUG_ON(iocb->ki_pos != pos);
  58         if (unlikely(file->f_flags & O_DIRECT))
  59                 ioflags |= IO_ISDIRECT;
  60         return xfs_read(XFS_I(file->f_path.dentry->d_inode), iocb, iov,
  61                                 nr_segs, &iocb->ki_pos, ioflags);
  62 }
  63
  64 STATIC ssize_t
  65 xfs_file_aio_read(
  66         struct kiocb            *iocb,
  67         const struct iovec      *iov,
  68         unsigned long           nr_segs,
  69         loff_t                  pos)
  70 {
  71         return __xfs_file_read(iocb, iov, nr_segs, IO_ISAIO, pos);
  72 }
  73
  74 STATIC ssize_t
  75 xfs_file_aio_read_invis(
  76         struct kiocb            *iocb,
  77         const struct iovec      *iov,
  78         unsigned long           nr_segs,
  79         loff_t                  pos)
  80 {
  81         return __xfs_file_read(iocb, iov, nr_segs, IO_ISAIO|IO_INVIS, pos);
  82 }
  83
  84 STATIC_INLINE ssize_t
  85 __xfs_file_write(
  86         struct kiocb            *iocb,
  87         const struct iovec      *iov,
  88         unsigned long           nr_segs,
  89         int                     ioflags,
  90         loff_t                  pos)
  91 {
  92         struct file     *file = iocb->ki_filp;
  93
  94         BUG_ON(iocb->ki_pos != pos);
  95         if (unlikely(file->f_flags & O_DIRECT))
  96                 ioflags |= IO_ISDIRECT;
  97         return xfs_write(XFS_I(file->f_mapping->host), iocb, iov, nr_segs,
  98                                 &iocb->ki_pos, ioflags);
  99 }
 100
 101 STATIC ssize_t
 102 xfs_file_aio_write(
 103         struct kiocb            *iocb,
 104         const struct iovec      *iov,
 105         unsigned long           nr_segs,
 106         loff_t                  pos)
 107 {
 108         return __xfs_file_write(iocb, iov, nr_segs, IO_ISAIO, pos);
 109 }
 110
 111 STATIC ssize_t
 112 xfs_file_aio_write_invis(
 113         struct kiocb            *iocb,
 114         const struct iovec      *iov,
 115         unsigned long           nr_segs,
 116         loff_t                  pos)
 117 {
 118         return __xfs_file_write(iocb, iov, nr_segs, IO_ISAIO|IO_INVIS, pos);
 119 }
 120
 121 STATIC ssize_t
 122 xfs_file_splice_read(
 123         struct file             *infilp,
 124         loff_t                  *ppos,
 125         struct pipe_inode_info  *pipe,
 126         size_t                  len,
 127         unsigned int            flags)
 128 {
 129         return xfs_splice_read(XFS_I(infilp->f_path.dentry->d_inode),
 130                                    infilp, ppos, pipe, len, flags, 0);
 131 }
 132
 133 STATIC ssize_t
 134 xfs_file_splice_read_invis(
 135         struct file             *infilp,
 136         loff_t                  *ppos,
 137         struct pipe_inode_info  *pipe,
 138         size_t                  len,
 139         unsigned int            flags)
 140 {
 141         return xfs_splice_read(XFS_I(infilp->f_path.dentry->d_inode),
 142                                    infilp, ppos, pipe, len, flags, IO_INVIS);
 143 }
 144
 145 STATIC ssize_t
 146 xfs_file_splice_write(
 147         struct pipe_inode_info  *pipe,
 148         struct file             *outfilp,
 149         loff_t                  *ppos,
 150         size_t                  len,
 151         unsigned int            flags)
 152 {
 153         return xfs_splice_write(XFS_I(outfilp->f_path.dentry->d_inode),
 154                                     pipe, outfilp, ppos, len, flags, 0);
 155 }
 156
 157 STATIC ssize_t
 158 xfs_file_splice_write_invis(
 159         struct pipe_inode_info  *pipe,
 160         struct file             *outfilp,
 161         loff_t                  *ppos,
 162         size_t                  len,
 163         unsigned int            flags)
 164 {
 165         return xfs_splice_write(XFS_I(outfilp->f_path.dentry->d_inode),
 166                                     pipe, outfilp, ppos, len, flags, IO_INVIS);
 167 }
 168
 169 STATIC int
 170 xfs_file_open(
 171         struct inode    *inode,
 172         struct file     *filp)
 173 {
 174         if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
 175                 return -EFBIG;
 176         return -xfs_open(XFS_I(inode));
 177 }
 178
 179 STATIC int
 180 xfs_file_release(
 181         struct inode    *inode,
 182         struct file     *filp)
 183 {
 184         return -xfs_release(XFS_I(inode));
 185 }
 186
 187 STATIC int
 188 xfs_file_fsync(
 189         struct file     *filp,
 190         struct dentry   *dentry,
 191         int             datasync)
 192 {
 193         int             flags = FSYNC_WAIT;
 194
 195         if (datasync)
 196                 flags |= FSYNC_DATA;
 197         xfs_iflags_clear(XFS_I(dentry->d_inode), XFS_ITRUNCATED);
 198         return -xfs_fsync(XFS_I(dentry->d_inode), flags,
 199                         (xfs_off_t)0, (xfs_off_t)-1);
 200 }
 201
 202 /*
 203  * Unfortunately we can't just use the clean and simple readdir implementation
 204  * below, because nfs might call back into ->lookup from the filldir callback
 205  * and that will deadlock the low-level btree code.
 206  *
 207  * Hopefully we'll find a better workaround that allows to use the optimal
 208  * version at least for local readdirs for 2.6.25.
 209  */
 210 #if 0
 211 STATIC int
 212 xfs_file_readdir(
 213         struct file     *filp,
 214         void            *dirent,
 215         filldir_t       filldir)
 216 {
 217         struct inode    *inode = filp->f_path.dentry->d_inode;
 218         xfs_inode_t     *ip = XFS_I(inode);
 219         int             error;
 220         size_t          bufsize;
 221
 222         /*
 223          * The Linux API doesn't pass down the total size of the buffer
 224          * we read into down to the filesystem.  With the filldir concept
 225          * it's not needed for correct information, but the XFS dir2 leaf
 226          * code wants an estimate of the buffer size to calculate it's
 227          * readahead window and size the buffers used for mapping to
 228          * physical blocks.
 229          *
 230          * Try to give it an estimate that's good enough, maybe at some
 231          * point we can change the ->readdir prototype to include the
 232          * buffer size.
 233          */
 234         bufsize = (size_t)min_t(loff_t, PAGE_SIZE, inode->i_size);
 235
 236         error = xfs_readdir(ip, dirent, bufsize,
 237                                 (xfs_off_t *)&filp->f_pos, filldir);
 238         if (error)
 239                 return -error;
 240         return 0;
 241 }
 242 #else
 243
 244 struct hack_dirent {
 245         u64             ino;
 246         loff_t          offset;
 247         int             namlen;
 248         unsigned int    d_type;
 249         char            name[];
 250 };
 251
 252 struct hack_callback {
 253         char            *dirent;
 254         size_t          len;
 255         size_t          used;
 256 };
 257
 258 STATIC int
 259 xfs_hack_filldir(
 260         void            *__buf,
 261         const char      *name,
 262         int             namlen,
 263         loff_t          offset,
 264         u64             ino,
 265         unsigned int    d_type)
 266 {
 267         struct hack_callback *buf = __buf;
 268         struct hack_dirent *de = (struct hack_dirent *)(buf->dirent + buf->used);
 269         unsigned int reclen;
 270
 271         reclen = ALIGN(sizeof(struct hack_dirent) + namlen, sizeof(u64));
 272         if (buf->used + reclen > buf->len)
 273                 return -EINVAL;
 274
 275         de->namlen = namlen;
 276         de->offset = offset;
 277         de->ino = ino;
 278         de->d_type = d_type;
 279         memcpy(de->name, name, namlen);
 280         buf->used += reclen;
 281         return 0;
 282 }
 283
 284 STATIC int
 285 xfs_file_readdir(
 286         struct file     *filp,
 287         void            *dirent,
 288         filldir_t       filldir)
 289 {
 290         struct inode    *inode = filp->f_path.dentry->d_inode;
 291         xfs_inode_t     *ip = XFS_I(inode);
 292         struct hack_callback buf;
 293         struct hack_dirent *de;
 294         int             error;
 295         loff_t          size;
 296         int             eof = 0;
 297         xfs_off_t       start_offset, curr_offset, offset;
 298
 299         /*
 300          * Try fairly hard to get memory
 301          */
 302         buf.len = PAGE_CACHE_SIZE;
 303         do {
 304                 buf.dirent = kmalloc(buf.len, GFP_KERNEL);
 305                 if (buf.dirent)
 306                         break;
 307                 buf.len >>= 1;
 308         } while (buf.len >= 1024);
 309
 310         if (!buf.dirent)
 311                 return -ENOMEM;
 312
 313         curr_offset = filp->f_pos;
 314         if (curr_offset == 0x7fffffff)
 315                 offset = 0xffffffff;
 316         else
 317                 offset = filp->f_pos;
 318
 319         while (!eof) {
 320                 unsigned int reclen;
 321
 322                 start_offset = offset;
 323
 324                 buf.used = 0;
 325                 error = -xfs_readdir(ip, &buf, buf.len, &offset,
 326                                      xfs_hack_filldir);
 327                 if (error || offset == start_offset) {
 328                         size = 0;
 329                         break;
 330                 }
 331
 332                 size = buf.used;
 333                 de = (struct hack_dirent *)buf.dirent;
 334                 while (size > 0) {
 335                         curr_offset = de->offset /* & 0x7fffffff */;
 336                         if (filldir(dirent, de->name, de->namlen,
 337                                         curr_offset & 0x7fffffff,
 338                                         de->ino, de->d_type)) {
 339                                 goto done;
 340                         }
 341
 342                         reclen = ALIGN(sizeof(struct hack_dirent) + de->namlen,
 343                                        sizeof(u64));
 344                         size -= reclen;
 345                         de = (struct hack_dirent *)((char *)de + reclen);
 346                 }
 347         }
 348
 349  done:
 350         if (!error) {
 351                 if (size == 0)
 352                         filp->f_pos = offset & 0x7fffffff;
 353                 else if (de)
 354                         filp->f_pos = curr_offset;
 355         }
 356
 357         kfree(buf.dirent);
 358         return error;
 359 }
 360 #endif
 361
 362 STATIC int
 363 xfs_file_mmap(
 364         struct file     *filp,
 365         struct vm_area_struct *vma)
 366 {
 367         vma->vm_ops = &xfs_file_vm_ops;
 368         vma->vm_flags |= VM_CAN_NONLINEAR;
 369
 370         file_accessed(filp);
 371         return 0;
 372 }
 373
 374 STATIC long
 375 xfs_file_ioctl(
 376         struct file     *filp,
 377         unsigned int    cmd,
 378         unsigned long   p)
 379 {
 380         int             error;
 381         struct inode    *inode = filp->f_path.dentry->d_inode;
 382
 383         error = xfs_ioctl(XFS_I(inode), filp, 0, cmd, (void __user *)p);
 384         xfs_iflags_set(XFS_I(inode), XFS_IMODIFIED);
 385
 386         /* NOTE:  some of the ioctl's return positive #'s as a
 387          *        byte count indicating success, such as
 388          *        readlink_by_handle.  So we don't "sign flip"
 389          *        like most other routines.  This means true
 390          *        errors need to be returned as a negative value.
 391          */
 392         return error;
 393 }
 394
 395 STATIC long
 396 xfs_file_ioctl_invis(
 397         struct file     *filp,
 398         unsigned int    cmd,
 399         unsigned long   p)
 400 {
 401         int             error;
 402         struct inode    *inode = filp->f_path.dentry->d_inode;
 403
 404         error = xfs_ioctl(XFS_I(inode), filp, IO_INVIS, cmd, (void __user *)p);
 405         xfs_iflags_set(XFS_I(inode), XFS_IMODIFIED);
 406
 407         /* NOTE:  some of the ioctl's return positive #'s as a
 408          *        byte count indicating success, such as
 409          *        readlink_by_handle.  So we don't "sign flip"
 410          *        like most other routines.  This means true
 411          *        errors need to be returned as a negative value.
 412          */
 413         return error;
 414 }
 415
 416 /*
 417  * mmap()d file has taken write protection fault and is being made
 418  * writable. We can set the page state up correctly for a writable
 419  * page, which means we can do correct delalloc accounting (ENOSPC
 420  * checking!) and unwritten extent mapping.
 421  */
 422 STATIC int
 423 xfs_vm_page_mkwrite(
 424         struct vm_area_struct   *vma,
 425         struct page             *page)
 426 {
 427         return block_page_mkwrite(vma, page, xfs_get_blocks);
 428 }
 429
 430 const struct file_operations xfs_file_operations = {
 431         .llseek         = generic_file_llseek,
 432         .read           = do_sync_read,
 433         .write          = do_sync_write,
 434         .aio_read       = xfs_file_aio_read,
 435         .aio_write      = xfs_file_aio_write,
 436         .splice_read    = xfs_file_splice_read,
 437         .splice_write   = xfs_file_splice_write,
 438         .unlocked_ioctl = xfs_file_ioctl,
 439 #ifdef CONFIG_COMPAT
 440         .compat_ioctl   = xfs_file_compat_ioctl,
 441 #endif
 442         .mmap           = xfs_file_mmap,
 443         .open           = xfs_file_open,
 444         .release        = xfs_file_release,
 445         .fsync          = xfs_file_fsync,
 446 #ifdef HAVE_FOP_OPEN_EXEC
 447         .open_exec      = xfs_file_open_exec,
 448 #endif
 449 };
 450
 451 const struct file_operations xfs_invis_file_operations = {
 452         .llseek         = generic_file_llseek,
 453         .read           = do_sync_read,
 454         .write          = do_sync_write,
 455         .aio_read       = xfs_file_aio_read_invis,
 456         .aio_write      = xfs_file_aio_write_invis,
 457         .splice_read    = xfs_file_splice_read_invis,
 458         .splice_write   = xfs_file_splice_write_invis,
 459         .unlocked_ioctl = xfs_file_ioctl_invis,
 460 #ifdef CONFIG_COMPAT
 461         .compat_ioctl   = xfs_file_compat_invis_ioctl,
 462 #endif
 463         .mmap           = xfs_file_mmap,
 464         .open           = xfs_file_open,
 465         .release        = xfs_file_release,
 466         .fsync          = xfs_file_fsync,
 467 };
 468
 469
 470 const struct file_operations xfs_dir_file_operations = {
 471         .read           = generic_read_dir,
 472         .readdir        = xfs_file_readdir,
 473         .unlocked_ioctl = xfs_file_ioctl,
 474 #ifdef CONFIG_COMPAT
 475         .compat_ioctl   = xfs_file_compat_ioctl,
 476 #endif
 477         .fsync          = xfs_file_fsync,
 478 };
 479
 480 static struct vm_operations_struct xfs_file_vm_ops = {
 481         .fault          = filemap_fault,
 482         .page_mkwrite   = xfs_vm_page_mkwrite,
 483 };