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         sxp = kmem_alloc(sizeof(xfs_swapext_t), KM_MAYFAIL);
  59         if (!sxp) {
  60                 error = XFS_ERROR(ENOMEM);
  61                 goto out;
  62         }
  63
  64         /* Pull information for the target fd */
  65         file = fget((int)sxp->sx_fdtarget);
  66         if (!file) {
  67                 error = XFS_ERROR(EINVAL);
  68                 goto out_free_sxp;
  69         }
  70
  71         if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND)) {
  72                 error = XFS_ERROR(EBADF);
  73                 goto out_put_file;
  74         }
  75
  76         target_file = fget((int)sxp->sx_fdtmp);
  77         if (!target_file) {
  78                 error = XFS_ERROR(EINVAL);
  79                 goto out_put_file;
  80         }
  81
  82         if (!(target_file->f_mode & FMODE_WRITE) ||
  83             (target_file->f_flags & O_APPEND)) {
  84                 error = XFS_ERROR(EBADF);
  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_free_sxp:
 113         kmem_free(sxp);
 114  out:
 115         return error;
 116 }
 117
 118 int
 119 xfs_swap_extents(
 120         xfs_inode_t     *ip,
 121         xfs_inode_t     *tip,
 122         xfs_swapext_t   *sxp)
 123 {
 124         xfs_mount_t     *mp;
 125         xfs_trans_t     *tp;
 126         xfs_bstat_t     *sbp = &sxp->sx_stat;
 127         xfs_ifork_t     *tempifp, *ifp, *tifp;
 128         int             ilf_fields, tilf_fields;
 129         static uint     lock_flags = XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL;
 130         int             error = 0;
 131         int             aforkblks = 0;
 132         int             taforkblks = 0;
 133         __uint64_t      tmp;
 134         char            locked = 0;
 135
 136         mp = ip->i_mount;
 137
 138         tempifp = kmem_alloc(sizeof(xfs_ifork_t), KM_MAYFAIL);
 139         if (!tempifp) {
 140                 error = XFS_ERROR(ENOMEM);
 141                 goto error0;
 142         }
 143
 144         sbp = &sxp->sx_stat;
 145
 146         /*
 147          * we have to do two separate lock calls here to keep lockdep
 148          * happy. If we try to get all the locks in one call, lock will
 149          * report false positives when we drop the ILOCK and regain them
 150          * below.
 151          */
 152         xfs_lock_two_inodes(ip, tip, XFS_IOLOCK_EXCL);
 153         xfs_lock_two_inodes(ip, tip, XFS_ILOCK_EXCL);
 154         locked = 1;
 155
 156         /* Verify that both files have the same format */
 157         if ((ip->i_d.di_mode & S_IFMT) != (tip->i_d.di_mode & S_IFMT)) {
 158                 error = XFS_ERROR(EINVAL);
 159                 goto error0;
 160         }
 161
 162         /* Verify both files are either real-time or non-realtime */
 163         if (XFS_IS_REALTIME_INODE(ip) != XFS_IS_REALTIME_INODE(tip)) {
 164                 error = XFS_ERROR(EINVAL);
 165                 goto error0;
 166         }
 167
 168         /* Should never get a local format */
 169         if (ip->i_d.di_format == XFS_DINODE_FMT_LOCAL ||
 170             tip->i_d.di_format == XFS_DINODE_FMT_LOCAL) {
 171                 error = XFS_ERROR(EINVAL);
 172                 goto error0;
 173         }
 174
 175         if (VN_CACHED(VFS_I(tip)) != 0) {
 176                 xfs_inval_cached_trace(tip, 0, -1, 0, -1);
 177                 error = xfs_flushinval_pages(tip, 0, -1,
 178                                 FI_REMAPF_LOCKED);
 179                 if (error)
 180                         goto error0;
 181         }
 182
 183         /* Verify O_DIRECT for ftmp */
 184         if (VN_CACHED(VFS_I(tip)) != 0) {
 185                 error = XFS_ERROR(EINVAL);
 186                 goto error0;
 187         }
 188
 189         /* Verify all data are being swapped */
 190         if (sxp->sx_offset != 0 ||
 191             sxp->sx_length != ip->i_d.di_size ||
 192             sxp->sx_length != tip->i_d.di_size) {
 193                 error = XFS_ERROR(EFAULT);
 194                 goto error0;
 195         }
 196
 197         /*
 198          * If the target has extended attributes, the tmp file
 199          * must also in order to ensure the correct data fork
 200          * format.
 201          */
 202         if ( XFS_IFORK_Q(ip) != XFS_IFORK_Q(tip) ) {
 203                 error = XFS_ERROR(EINVAL);
 204                 goto error0;
 205         }
 206
 207         /*
 208          * Compare the current change & modify times with that
 209          * passed in.  If they differ, we abort this swap.
 210          * This is the mechanism used to ensure the calling
 211          * process that the file was not changed out from
 212          * under it.
 213          */
 214         if ((sbp->bs_ctime.tv_sec != ip->i_d.di_ctime.t_sec) ||
 215             (sbp->bs_ctime.tv_nsec != ip->i_d.di_ctime.t_nsec) ||
 216             (sbp->bs_mtime.tv_sec != ip->i_d.di_mtime.t_sec) ||
 217             (sbp->bs_mtime.tv_nsec != ip->i_d.di_mtime.t_nsec)) {
 218                 error = XFS_ERROR(EBUSY);
 219                 goto error0;
 220         }
 221
 222         /* We need to fail if the file is memory mapped.  Once we have tossed
 223          * all existing pages, the page fault will have no option
 224          * but to go to the filesystem for pages. By making the page fault call
 225          * vop_read (or write in the case of autogrow) they block on the iolock
 226          * until we have switched the extents.
 227          */
 228         if (VN_MAPPED(VFS_I(ip))) {
 229                 error = XFS_ERROR(EBUSY);
 230                 goto error0;
 231         }
 232
 233         xfs_iunlock(ip, XFS_ILOCK_EXCL);
 234         xfs_iunlock(tip, XFS_ILOCK_EXCL);
 235
 236         /*
 237          * There is a race condition here since we gave up the
 238          * ilock.  However, the data fork will not change since
 239          * we have the iolock (locked for truncation too) so we
 240          * are safe.  We don't really care if non-io related
 241          * fields change.
 242          */
 243
 244         xfs_tosspages(ip, 0, -1, FI_REMAPF);
 245
 246         tp = xfs_trans_alloc(mp, XFS_TRANS_SWAPEXT);
 247         if ((error = xfs_trans_reserve(tp, 0,
 248                                      XFS_ICHANGE_LOG_RES(mp), 0,
 249                                      0, 0))) {
 250                 xfs_iunlock(ip,  XFS_IOLOCK_EXCL);
 251                 xfs_iunlock(tip, XFS_IOLOCK_EXCL);
 252                 xfs_trans_cancel(tp, 0);
 253                 locked = 0;
 254                 goto error0;
 255         }
 256         xfs_lock_two_inodes(ip, tip, XFS_ILOCK_EXCL);
 257
 258         /*
 259          * Count the number of extended attribute blocks
 260          */
 261         if ( ((XFS_IFORK_Q(ip) != 0) && (ip->i_d.di_anextents > 0)) &&
 262              (ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
 263                 error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK, &aforkblks);
 264                 if (error) {
 265                         xfs_trans_cancel(tp, 0);
 266                         goto error0;
 267                 }
 268         }
 269         if ( ((XFS_IFORK_Q(tip) != 0) && (tip->i_d.di_anextents > 0)) &&
 270              (tip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
 271                 error = xfs_bmap_count_blocks(tp, tip, XFS_ATTR_FORK,
 272                         &taforkblks);
 273                 if (error) {
 274                         xfs_trans_cancel(tp, 0);
 275                         goto error0;
 276                 }
 277         }
 278
 279         /*
 280          * Swap the data forks of the inodes
 281          */
 282         ifp = &ip->i_df;
 283         tifp = &tip->i_df;
 284         *tempifp = *ifp;        /* struct copy */
 285         *ifp = *tifp;           /* struct copy */
 286         *tifp = *tempifp;       /* struct copy */
 287
 288         /*
 289          * Fix the on-disk inode values
 290          */
 291         tmp = (__uint64_t)ip->i_d.di_nblocks;
 292         ip->i_d.di_nblocks = tip->i_d.di_nblocks - taforkblks + aforkblks;
 293         tip->i_d.di_nblocks = tmp + taforkblks - aforkblks;
 294
 295         tmp = (__uint64_t) ip->i_d.di_nextents;
 296         ip->i_d.di_nextents = tip->i_d.di_nextents;
 297         tip->i_d.di_nextents = tmp;
 298
 299         tmp = (__uint64_t) ip->i_d.di_format;
 300         ip->i_d.di_format = tip->i_d.di_format;
 301         tip->i_d.di_format = tmp;
 302
 303         ilf_fields = XFS_ILOG_CORE;
 304
 305         switch(ip->i_d.di_format) {
 306         case XFS_DINODE_FMT_EXTENTS:
 307                 /* If the extents fit in the inode, fix the
 308                  * pointer.  Otherwise it's already NULL or
 309                  * pointing to the extent.
 310                  */
 311                 if (ip->i_d.di_nextents <= XFS_INLINE_EXTS) {
 312                         ifp->if_u1.if_extents =
 313                                 ifp->if_u2.if_inline_ext;
 314                 }
 315                 ilf_fields |= XFS_ILOG_DEXT;
 316                 break;
 317         case XFS_DINODE_FMT_BTREE:
 318                 ilf_fields |= XFS_ILOG_DBROOT;
 319                 break;
 320         }
 321
 322         tilf_fields = XFS_ILOG_CORE;
 323
 324         switch(tip->i_d.di_format) {
 325         case XFS_DINODE_FMT_EXTENTS:
 326                 /* If the extents fit in the inode, fix the
 327                  * pointer.  Otherwise it's already NULL or
 328                  * pointing to the extent.
 329                  */
 330                 if (tip->i_d.di_nextents <= XFS_INLINE_EXTS) {
 331                         tifp->if_u1.if_extents =
 332                                 tifp->if_u2.if_inline_ext;
 333                 }
 334                 tilf_fields |= XFS_ILOG_DEXT;
 335                 break;
 336         case XFS_DINODE_FMT_BTREE:
 337                 tilf_fields |= XFS_ILOG_DBROOT;
 338                 break;
 339         }
 340
 341
 342         IHOLD(ip);
 343         xfs_trans_ijoin(tp, ip, lock_flags);
 344
 345         IHOLD(tip);
 346         xfs_trans_ijoin(tp, tip, lock_flags);
 347
 348         xfs_trans_log_inode(tp, ip,  ilf_fields);
 349         xfs_trans_log_inode(tp, tip, tilf_fields);
 350
 351         /*
 352          * If this is a synchronous mount, make sure that the
 353          * transaction goes to disk before returning to the user.
 354          */
 355         if (mp->m_flags & XFS_MOUNT_WSYNC) {
 356                 xfs_trans_set_sync(tp);
 357         }
 358
 359         error = xfs_trans_commit(tp, XFS_TRANS_SWAPEXT);
 360         locked = 0;
 361
 362  error0:
 363         if (locked) {
 364                 xfs_iunlock(ip,  lock_flags);
 365                 xfs_iunlock(tip, lock_flags);
 366         }
 367         if (tempifp != NULL)
 368                 kmem_free(tempifp);
 369         return error;
 370 }