fs/xfs/xfs_dfrag.c

   1 /*
   2  * Copyright (c) 2000-2006 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_fs.h"
  20 #include "xfs_types.h"
  21 #include "xfs_bit.h"
  22 #include "xfs_log.h"
  23 #include "xfs_inum.h"
  24 #include "xfs_trans.h"
  25 #include "xfs_sb.h"
  26 #include "xfs_ag.h"
  27 #include "xfs_dir2.h"
  28 #include "xfs_dmapi.h"
  29 #include "xfs_mount.h"
  30 #include "xfs_bmap_btree.h"
  31 #include "xfs_alloc_btree.h"
  32 #include "xfs_ialloc_btree.h"
  33 #include "xfs_dir2_sf.h"
  34 #include "xfs_attr_sf.h"
  35 #include "xfs_dinode.h"
  36 #include "xfs_inode.h"
  37 #include "xfs_inode_item.h"
  38 #include "xfs_bmap.h"
  39 #include "xfs_btree.h"
  40 #include "xfs_ialloc.h"
  41 #include "xfs_itable.h"
  42 #include "xfs_dfrag.h"
  43 #include "xfs_error.h"
  44 #include "xfs_rw.h"
  45 #include "xfs_vnodeops.h"
  46
  47 /*
  48  * Syssgi interface for swapext
  49  */
  50 int
  51 xfs_swapext(
  52         xfs_swapext_t   __user *sxu)
  53 {
  54         xfs_swapext_t   *sxp;
  55         xfs_inode_t     *ip, *tip;
  56         struct file     *file, *target_file;
  57         int             error = 0;
  58
  59         sxp = kmem_alloc(sizeof(xfs_swapext_t), KM_MAYFAIL);
  60         if (!sxp) {
  61                 error = XFS_ERROR(ENOMEM);
  62                 goto out;
  63         }
  64
  65         if (copy_from_user(sxp, sxu, sizeof(xfs_swapext_t))) {
  66                 error = XFS_ERROR(EFAULT);
  67                 goto out_free_sxp;
  68         }
  69
  70         /* Pull information for the target fd */
  71         file = fget((int)sxp->sx_fdtarget);
  72         if (!file) {
  73                 error = XFS_ERROR(EINVAL);
  74                 goto out_free_sxp;
  75         }
  76
  77         if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND)) {
  78                 error = XFS_ERROR(EBADF);
  79                 goto out_put_file;
  80         }
  81
  82         target_file = fget((int)sxp->sx_fdtmp);
  83         if (!target_file) {
  84                 error = XFS_ERROR(EINVAL);
  85                 goto out_put_file;
  86         }
  87
  88         if (!(target_file->f_mode & FMODE_WRITE) ||
  89             (target_file->f_flags & O_APPEND)) {
  90                 error = XFS_ERROR(EBADF);
  91                 goto out_put_target_file;
  92         }
  93
  94         ip = XFS_I(file->f_path.dentry->d_inode);
  95         tip = XFS_I(target_file->f_path.dentry->d_inode);
  96
  97         if (ip->i_mount != tip->i_mount) {
  98                 error = XFS_ERROR(EINVAL);
  99                 goto out_put_target_file;
 100         }
 101
 102         if (ip->i_ino == tip->i_ino) {
 103                 error = XFS_ERROR(EINVAL);
 104                 goto out_put_target_file;
 105         }
 106
 107         if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
 108                 error = XFS_ERROR(EIO);
 109                 goto out_put_target_file;
 110         }
 111
 112         error = xfs_swap_extents(ip, tip, sxp);
 113
 114  out_put_target_file:
 115         fput(target_file);
 116  out_put_file:
 117         fput(file);
 118  out_free_sxp:
 119         kmem_free(sxp);
 120  out:
 121         return error;
 122 }
 123
 124 int
 125 xfs_swap_extents(
 126         xfs_inode_t     *ip,
 127         xfs_inode_t     *tip,
 128         xfs_swapext_t   *sxp)
 129 {
 130         xfs_mount_t     *mp;
 131         xfs_trans_t     *tp;
 132         xfs_bstat_t     *sbp = &sxp->sx_stat;
 133         xfs_ifork_t     *tempifp, *ifp, *tifp;
 134         int             ilf_fields, tilf_fields;
 135         static uint     lock_flags = XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL;
 136         int             error = 0;
 137         int             aforkblks = 0;
 138         int             taforkblks = 0;
 139         __uint64_t      tmp;
 140         char            locked = 0;
 141
 142         mp = ip->i_mount;
 143
 144         tempifp = kmem_alloc(sizeof(xfs_ifork_t), KM_MAYFAIL);
 145         if (!tempifp) {
 146                 error = XFS_ERROR(ENOMEM);
 147                 goto error0;
 148         }
 149
 150         sbp = &sxp->sx_stat;
 151
 152         xfs_lock_two_inodes(ip, tip, lock_flags);
 153         locked = 1;
 154
 155         /* Verify that both files have the same format */
 156         if ((ip->i_d.di_mode & S_IFMT) != (tip->i_d.di_mode & S_IFMT)) {
 157                 error = XFS_ERROR(EINVAL);
 158                 goto error0;
 159         }
 160
 161         /* Verify both files are either real-time or non-realtime */
 162         if (XFS_IS_REALTIME_INODE(ip) != XFS_IS_REALTIME_INODE(tip)) {
 163                 error = XFS_ERROR(EINVAL);
 164                 goto error0;
 165         }
 166
 167         /* Should never get a local format */
 168         if (ip->i_d.di_format == XFS_DINODE_FMT_LOCAL ||
 169             tip->i_d.di_format == XFS_DINODE_FMT_LOCAL) {
 170                 error = XFS_ERROR(EINVAL);
 171                 goto error0;
 172         }
 173
 174         if (VN_CACHED(VFS_I(tip)) != 0) {
 175                 xfs_inval_cached_trace(tip, 0, -1, 0, -1);
 176                 error = xfs_flushinval_pages(tip, 0, -1,
 177                                 FI_REMAPF_LOCKED);
 178                 if (error)
 179                         goto error0;
 180         }
 181
 182         /* Verify O_DIRECT for ftmp */
 183         if (VN_CACHED(VFS_I(tip)) != 0) {
 184                 error = XFS_ERROR(EINVAL);
 185                 goto error0;
 186         }
 187
 188         /* Verify all data are being swapped */
 189         if (sxp->sx_offset != 0 ||
 190             sxp->sx_length != ip->i_d.di_size ||
 191             sxp->sx_length != tip->i_d.di_size) {
 192                 error = XFS_ERROR(EFAULT);
 193                 goto error0;
 194         }
 195
 196         /*
 197          * If the target has extended attributes, the tmp file
 198          * must also in order to ensure the correct data fork
 199          * format.
 200          */
 201         if ( XFS_IFORK_Q(ip) != XFS_IFORK_Q(tip) ) {
 202                 error = XFS_ERROR(EINVAL);
 203                 goto error0;
 204         }
 205
 206         /*
 207          * Compare the current change & modify times with that
 208          * passed in.  If they differ, we abort this swap.
 209          * This is the mechanism used to ensure the calling
 210          * process that the file was not changed out from
 211          * under it.
 212          */
 213         if ((sbp->bs_ctime.tv_sec != ip->i_d.di_ctime.t_sec) ||
 214             (sbp->bs_ctime.tv_nsec != ip->i_d.di_ctime.t_nsec) ||
 215             (sbp->bs_mtime.tv_sec != ip->i_d.di_mtime.t_sec) ||
 216             (sbp->bs_mtime.tv_nsec != ip->i_d.di_mtime.t_nsec)) {
 217                 error = XFS_ERROR(EBUSY);
 218                 goto error0;
 219         }
 220
 221         /* We need to fail if the file is memory mapped.  Once we have tossed
 222          * all existing pages, the page fault will have no option
 223          * but to go to the filesystem for pages. By making the page fault call
 224          * vop_read (or write in the case of autogrow) they block on the iolock
 225          * until we have switched the extents.
 226          */
 227         if (VN_MAPPED(VFS_I(ip))) {
 228                 error = XFS_ERROR(EBUSY);
 229                 goto error0;
 230         }
 231
 232         xfs_iunlock(ip, XFS_ILOCK_EXCL);
 233         xfs_iunlock(tip, XFS_ILOCK_EXCL);
 234
 235         /*
 236          * There is a race condition here since we gave up the
 237          * ilock.  However, the data fork will not change since
 238          * we have the iolock (locked for truncation too) so we
 239          * are safe.  We don't really care if non-io related
 240          * fields change.
 241          */
 242
 243         xfs_tosspages(ip, 0, -1, FI_REMAPF);
 244
 245         tp = xfs_trans_alloc(mp, XFS_TRANS_SWAPEXT);
 246         if ((error = xfs_trans_reserve(tp, 0,
 247                                      XFS_ICHANGE_LOG_RES(mp), 0,
 248                                      0, 0))) {
 249                 xfs_iunlock(ip,  XFS_IOLOCK_EXCL);
 250                 xfs_iunlock(tip, XFS_IOLOCK_EXCL);
 251                 xfs_trans_cancel(tp, 0);
 252                 locked = 0;
 253                 goto error0;
 254         }
 255         xfs_lock_two_inodes(ip, tip, XFS_ILOCK_EXCL);
 256
 257         /*
 258          * Count the number of extended attribute blocks
 259          */
 260         if ( ((XFS_IFORK_Q(ip) != 0) && (ip->i_d.di_anextents > 0)) &&
 261              (ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
 262                 error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK, &aforkblks);
 263                 if (error) {
 264                         xfs_trans_cancel(tp, 0);
 265                         goto error0;
 266                 }
 267         }
 268         if ( ((XFS_IFORK_Q(tip) != 0) && (tip->i_d.di_anextents > 0)) &&
 269              (tip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
 270                 error = xfs_bmap_count_blocks(tp, tip, XFS_ATTR_FORK,
 271                         &taforkblks);
 272                 if (error) {
 273                         xfs_trans_cancel(tp, 0);
 274                         goto error0;
 275                 }
 276         }
 277
 278         /*
 279          * Swap the data forks of the inodes
 280          */
 281         ifp = &ip->i_df;
 282         tifp = &tip->i_df;
 283         *tempifp = *ifp;        /* struct copy */
 284         *ifp = *tifp;           /* struct copy */
 285         *tifp = *tempifp;       /* struct copy */
 286
 287         /*
 288          * Fix the on-disk inode values
 289          */
 290         tmp = (__uint64_t)ip->i_d.di_nblocks;
 291         ip->i_d.di_nblocks = tip->i_d.di_nblocks - taforkblks + aforkblks;
 292         tip->i_d.di_nblocks = tmp + taforkblks - aforkblks;
 293
 294         tmp = (__uint64_t) ip->i_d.di_nextents;
 295         ip->i_d.di_nextents = tip->i_d.di_nextents;
 296         tip->i_d.di_nextents = tmp;
 297
 298         tmp = (__uint64_t) ip->i_d.di_format;
 299         ip->i_d.di_format = tip->i_d.di_format;
 300         tip->i_d.di_format = tmp;
 301
 302         ilf_fields = XFS_ILOG_CORE;
 303
 304         switch(ip->i_d.di_format) {
 305         case XFS_DINODE_FMT_EXTENTS:
 306                 /* If the extents fit in the inode, fix the
 307                  * pointer.  Otherwise it's already NULL or
 308                  * pointing to the extent.
 309                  */
 310                 if (ip->i_d.di_nextents <= XFS_INLINE_EXTS) {
 311                         ifp->if_u1.if_extents =
 312                                 ifp->if_u2.if_inline_ext;
 313                 }
 314                 ilf_fields |= XFS_ILOG_DEXT;
 315                 break;
 316         case XFS_DINODE_FMT_BTREE:
 317                 ilf_fields |= XFS_ILOG_DBROOT;
 318                 break;
 319         }
 320
 321         tilf_fields = XFS_ILOG_CORE;
 322
 323         switch(tip->i_d.di_format) {
 324         case XFS_DINODE_FMT_EXTENTS:
 325                 /* If the extents fit in the inode, fix the
 326                  * pointer.  Otherwise it's already NULL or
 327                  * pointing to the extent.
 328                  */
 329                 if (tip->i_d.di_nextents <= XFS_INLINE_EXTS) {
 330                         tifp->if_u1.if_extents =
 331                                 tifp->if_u2.if_inline_ext;
 332                 }
 333                 tilf_fields |= XFS_ILOG_DEXT;
 334                 break;
 335         case XFS_DINODE_FMT_BTREE:
 336                 tilf_fields |= XFS_ILOG_DBROOT;
 337                 break;
 338         }
 339
 340
 341         IHOLD(ip);
 342         xfs_trans_ijoin(tp, ip, lock_flags);
 343
 344         IHOLD(tip);
 345         xfs_trans_ijoin(tp, tip, lock_flags);
 346
 347         xfs_trans_log_inode(tp, ip,  ilf_fields);
 348         xfs_trans_log_inode(tp, tip, tilf_fields);
 349
 350         /*
 351          * If this is a synchronous mount, make sure that the
 352          * transaction goes to disk before returning to the user.
 353          */
 354         if (mp->m_flags & XFS_MOUNT_WSYNC) {
 355                 xfs_trans_set_sync(tp);
 356         }
 357
 358         error = xfs_trans_commit(tp, XFS_TRANS_SWAPEXT);
 359         locked = 0;
 360
 361  error0:
 362         if (locked) {
 363                 xfs_iunlock(ip,  lock_flags);
 364                 xfs_iunlock(tip, lock_flags);
 365         }
 366         if (tempifp != NULL)
 367                 kmem_free(tempifp);
 368         return error;
 369 }