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         /*
 153          * we have to do two separate lock calls here to keep lockdep
 154          * happy. If we try to get all the locks in one call, lock will
 155          * report false positives when we drop the ILOCK and regain them
 156          * below.
 157          */
 158         xfs_lock_two_inodes(ip, tip, XFS_IOLOCK_EXCL);
 159         xfs_lock_two_inodes(ip, tip, XFS_ILOCK_EXCL);
 160         locked = 1;
 161
 162         /* Verify that both files have the same format */
 163         if ((ip->i_d.di_mode & S_IFMT) != (tip->i_d.di_mode & S_IFMT)) {
 164                 error = XFS_ERROR(EINVAL);
 165                 goto error0;
 166         }
 167
 168         /* Verify both files are either real-time or non-realtime */
 169         if (XFS_IS_REALTIME_INODE(ip) != XFS_IS_REALTIME_INODE(tip)) {
 170                 error = XFS_ERROR(EINVAL);
 171                 goto error0;
 172         }
 173
 174         /* Should never get a local format */
 175         if (ip->i_d.di_format == XFS_DINODE_FMT_LOCAL ||
 176             tip->i_d.di_format == XFS_DINODE_FMT_LOCAL) {
 177                 error = XFS_ERROR(EINVAL);
 178                 goto error0;
 179         }
 180
 181         if (VN_CACHED(VFS_I(tip)) != 0) {
 182                 xfs_inval_cached_trace(tip, 0, -1, 0, -1);
 183                 error = xfs_flushinval_pages(tip, 0, -1,
 184                                 FI_REMAPF_LOCKED);
 185                 if (error)
 186                         goto error0;
 187         }
 188
 189         /* Verify O_DIRECT for ftmp */
 190         if (VN_CACHED(VFS_I(tip)) != 0) {
 191                 error = XFS_ERROR(EINVAL);
 192                 goto error0;
 193         }
 194
 195         /* Verify all data are being swapped */
 196         if (sxp->sx_offset != 0 ||
 197             sxp->sx_length != ip->i_d.di_size ||
 198             sxp->sx_length != tip->i_d.di_size) {
 199                 error = XFS_ERROR(EFAULT);
 200                 goto error0;
 201         }
 202
 203         /*
 204          * If the target has extended attributes, the tmp file
 205          * must also in order to ensure the correct data fork
 206          * format.
 207          */
 208         if ( XFS_IFORK_Q(ip) != XFS_IFORK_Q(tip) ) {
 209                 error = XFS_ERROR(EINVAL);
 210                 goto error0;
 211         }
 212
 213         /*
 214          * Compare the current change & modify times with that
 215          * passed in.  If they differ, we abort this swap.
 216          * This is the mechanism used to ensure the calling
 217          * process that the file was not changed out from
 218          * under it.
 219          */
 220         if ((sbp->bs_ctime.tv_sec != ip->i_d.di_ctime.t_sec) ||
 221             (sbp->bs_ctime.tv_nsec != ip->i_d.di_ctime.t_nsec) ||
 222             (sbp->bs_mtime.tv_sec != ip->i_d.di_mtime.t_sec) ||
 223             (sbp->bs_mtime.tv_nsec != ip->i_d.di_mtime.t_nsec)) {
 224                 error = XFS_ERROR(EBUSY);
 225                 goto error0;
 226         }
 227
 228         /* We need to fail if the file is memory mapped.  Once we have tossed
 229          * all existing pages, the page fault will have no option
 230          * but to go to the filesystem for pages. By making the page fault call
 231          * vop_read (or write in the case of autogrow) they block on the iolock
 232          * until we have switched the extents.
 233          */
 234         if (VN_MAPPED(VFS_I(ip))) {
 235                 error = XFS_ERROR(EBUSY);
 236                 goto error0;
 237         }
 238
 239         xfs_iunlock(ip, XFS_ILOCK_EXCL);
 240         xfs_iunlock(tip, XFS_ILOCK_EXCL);
 241
 242         /*
 243          * There is a race condition here since we gave up the
 244          * ilock.  However, the data fork will not change since
 245          * we have the iolock (locked for truncation too) so we
 246          * are safe.  We don't really care if non-io related
 247          * fields change.
 248          */
 249
 250         xfs_tosspages(ip, 0, -1, FI_REMAPF);
 251
 252         tp = xfs_trans_alloc(mp, XFS_TRANS_SWAPEXT);
 253         if ((error = xfs_trans_reserve(tp, 0,
 254                                      XFS_ICHANGE_LOG_RES(mp), 0,
 255                                      0, 0))) {
 256                 xfs_iunlock(ip,  XFS_IOLOCK_EXCL);
 257                 xfs_iunlock(tip, XFS_IOLOCK_EXCL);
 258                 xfs_trans_cancel(tp, 0);
 259                 locked = 0;
 260                 goto error0;
 261         }
 262         xfs_lock_two_inodes(ip, tip, XFS_ILOCK_EXCL);
 263
 264         /*
 265          * Count the number of extended attribute blocks
 266          */
 267         if ( ((XFS_IFORK_Q(ip) != 0) && (ip->i_d.di_anextents > 0)) &&
 268              (ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
 269                 error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK, &aforkblks);
 270                 if (error) {
 271                         xfs_trans_cancel(tp, 0);
 272                         goto error0;
 273                 }
 274         }
 275         if ( ((XFS_IFORK_Q(tip) != 0) && (tip->i_d.di_anextents > 0)) &&
 276              (tip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
 277                 error = xfs_bmap_count_blocks(tp, tip, XFS_ATTR_FORK,
 278                         &taforkblks);
 279                 if (error) {
 280                         xfs_trans_cancel(tp, 0);
 281                         goto error0;
 282                 }
 283         }
 284
 285         /*
 286          * Swap the data forks of the inodes
 287          */
 288         ifp = &ip->i_df;
 289         tifp = &tip->i_df;
 290         *tempifp = *ifp;        /* struct copy */
 291         *ifp = *tifp;           /* struct copy */
 292         *tifp = *tempifp;       /* struct copy */
 293
 294         /*
 295          * Fix the on-disk inode values
 296          */
 297         tmp = (__uint64_t)ip->i_d.di_nblocks;
 298         ip->i_d.di_nblocks = tip->i_d.di_nblocks - taforkblks + aforkblks;
 299         tip->i_d.di_nblocks = tmp + taforkblks - aforkblks;
 300
 301         tmp = (__uint64_t) ip->i_d.di_nextents;
 302         ip->i_d.di_nextents = tip->i_d.di_nextents;
 303         tip->i_d.di_nextents = tmp;
 304
 305         tmp = (__uint64_t) ip->i_d.di_format;
 306         ip->i_d.di_format = tip->i_d.di_format;
 307         tip->i_d.di_format = tmp;
 308
 309         ilf_fields = XFS_ILOG_CORE;
 310
 311         switch(ip->i_d.di_format) {
 312         case XFS_DINODE_FMT_EXTENTS:
 313                 /* If the extents fit in the inode, fix the
 314                  * pointer.  Otherwise it's already NULL or
 315                  * pointing to the extent.
 316                  */
 317                 if (ip->i_d.di_nextents <= XFS_INLINE_EXTS) {
 318                         ifp->if_u1.if_extents =
 319                                 ifp->if_u2.if_inline_ext;
 320                 }
 321                 ilf_fields |= XFS_ILOG_DEXT;
 322                 break;
 323         case XFS_DINODE_FMT_BTREE:
 324                 ilf_fields |= XFS_ILOG_DBROOT;
 325                 break;
 326         }
 327
 328         tilf_fields = XFS_ILOG_CORE;
 329
 330         switch(tip->i_d.di_format) {
 331         case XFS_DINODE_FMT_EXTENTS:
 332                 /* If the extents fit in the inode, fix the
 333                  * pointer.  Otherwise it's already NULL or
 334                  * pointing to the extent.
 335                  */
 336                 if (tip->i_d.di_nextents <= XFS_INLINE_EXTS) {
 337                         tifp->if_u1.if_extents =
 338                                 tifp->if_u2.if_inline_ext;
 339                 }
 340                 tilf_fields |= XFS_ILOG_DEXT;
 341                 break;
 342         case XFS_DINODE_FMT_BTREE:
 343                 tilf_fields |= XFS_ILOG_DBROOT;
 344                 break;
 345         }
 346
 347
 348         IHOLD(ip);
 349         xfs_trans_ijoin(tp, ip, lock_flags);
 350
 351         IHOLD(tip);
 352         xfs_trans_ijoin(tp, tip, lock_flags);
 353
 354         xfs_trans_log_inode(tp, ip,  ilf_fields);
 355         xfs_trans_log_inode(tp, tip, tilf_fields);
 356
 357         /*
 358          * If this is a synchronous mount, make sure that the
 359          * transaction goes to disk before returning to the user.
 360          */
 361         if (mp->m_flags & XFS_MOUNT_WSYNC) {
 362                 xfs_trans_set_sync(tp);
 363         }
 364
 365         error = xfs_trans_commit(tp, XFS_TRANS_SWAPEXT);
 366         locked = 0;
 367
 368  error0:
 369         if (locked) {
 370                 xfs_iunlock(ip,  lock_flags);
 371                 xfs_iunlock(tip, lock_flags);
 372         }
 373         if (tempifp != NULL)
 374                 kmem_free(tempifp);
 375         return error;
 376 }