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