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_dir.h"
  28 #include "xfs_dir2.h"
  29 #include "xfs_dmapi.h"
  30 #include "xfs_mount.h"
  31 #include "xfs_bmap_btree.h"
  32 #include "xfs_alloc_btree.h"
  33 #include "xfs_ialloc_btree.h"
  34 #include "xfs_dir_sf.h"
  35 #include "xfs_dir2_sf.h"
  36 #include "xfs_attr_sf.h"
  37 #include "xfs_dinode.h"
  38 #include "xfs_inode.h"
  39 #include "xfs_inode_item.h"
  40 #include "xfs_bmap.h"
  41 #include "xfs_btree.h"
  42 #include "xfs_ialloc.h"
  43 #include "xfs_itable.h"
  44 #include "xfs_dfrag.h"
  45 #include "xfs_error.h"
  46 #include "xfs_mac.h"
  47 #include "xfs_rw.h"
  48
  49 /*
  50  * Syssgi interface for swapext
  51  */
  52 int
  53 xfs_swapext(
  54         xfs_swapext_t   __user *sxu)
  55 {
  56         xfs_swapext_t   *sxp;
  57         xfs_inode_t     *ip=NULL, *tip=NULL;
  58         xfs_mount_t     *mp;
  59         struct file     *fp = NULL, *tfp = NULL;
  60         vnode_t         *vp, *tvp;
  61         int             error = 0;
  62
  63         sxp = kmem_alloc(sizeof(xfs_swapext_t), KM_MAYFAIL);
  64         if (!sxp) {
  65                 error = XFS_ERROR(ENOMEM);
  66                 goto error0;
  67         }
  68
  69         if (copy_from_user(sxp, sxu, sizeof(xfs_swapext_t))) {
  70                 error = XFS_ERROR(EFAULT);
  71                 goto error0;
  72         }
  73
  74         /* Pull information for the target fd */
  75         if (((fp = fget((int)sxp->sx_fdtarget)) == NULL) ||
  76             ((vp = vn_from_inode(fp->f_dentry->d_inode)) == NULL))  {
  77                 error = XFS_ERROR(EINVAL);
  78                 goto error0;
  79         }
  80
  81         ip = xfs_vtoi(vp);
  82         if (ip == NULL) {
  83                 error = XFS_ERROR(EBADF);
  84                 goto error0;
  85         }
  86
  87         if (((tfp = fget((int)sxp->sx_fdtmp)) == NULL) ||
  88             ((tvp = vn_from_inode(tfp->f_dentry->d_inode)) == NULL)) {
  89                 error = XFS_ERROR(EINVAL);
  90                 goto error0;
  91         }
  92
  93         tip = xfs_vtoi(tvp);
  94         if (tip == NULL) {
  95                 error = XFS_ERROR(EBADF);
  96                 goto error0;
  97         }
  98
  99         if (ip->i_mount != tip->i_mount) {
 100                 error =  XFS_ERROR(EINVAL);
 101                 goto error0;
 102         }
 103
 104         if (ip->i_ino == tip->i_ino) {
 105                 error =  XFS_ERROR(EINVAL);
 106                 goto error0;
 107         }
 108
 109         mp = ip->i_mount;
 110
 111         if (XFS_FORCED_SHUTDOWN(mp)) {
 112                 error =  XFS_ERROR(EIO);
 113                 goto error0;
 114         }
 115
 116         error = XFS_SWAP_EXTENTS(mp, &ip->i_iocore, &tip->i_iocore, sxp);
 117
 118  error0:
 119         if (fp != NULL)
 120                 fput(fp);
 121         if (tfp != NULL)
 122                 fput(tfp);
 123
 124         if (sxp != NULL)
 125                 kmem_free(sxp, sizeof(xfs_swapext_t));
 126
 127         return error;
 128 }
 129
 130 int
 131 xfs_swap_extents(
 132         xfs_inode_t     *ip,
 133         xfs_inode_t     *tip,
 134         xfs_swapext_t   *sxp)
 135 {
 136         xfs_mount_t     *mp;
 137         xfs_inode_t     *ips[2];
 138         xfs_trans_t     *tp;
 139         xfs_bstat_t     *sbp = &sxp->sx_stat;
 140         vnode_t         *vp, *tvp;
 141         xfs_ifork_t     *tempifp, *ifp, *tifp;
 142         int             ilf_fields, tilf_fields;
 143         static uint     lock_flags = XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL;
 144         int             error = 0;
 145         int             aforkblks = 0;
 146         int             taforkblks = 0;
 147         __uint64_t      tmp;
 148         char            locked = 0;
 149
 150         mp = ip->i_mount;
 151
 152         tempifp = kmem_alloc(sizeof(xfs_ifork_t), KM_MAYFAIL);
 153         if (!tempifp) {
 154                 error = XFS_ERROR(ENOMEM);
 155                 goto error0;
 156         }
 157
 158         sbp = &sxp->sx_stat;
 159         vp = XFS_ITOV(ip);
 160         tvp = XFS_ITOV(tip);
 161
 162         /* Lock in i_ino order */
 163         if (ip->i_ino < tip->i_ino) {
 164                 ips[0] = ip;
 165                 ips[1] = tip;
 166         } else {
 167                 ips[0] = tip;
 168                 ips[1] = ip;
 169         }
 170
 171         xfs_lock_inodes(ips, 2, 0, lock_flags);
 172         locked = 1;
 173
 174         /* Check permissions */
 175         error = xfs_iaccess(ip, S_IWUSR, NULL);
 176         if (error)
 177                 goto error0;
 178
 179         error = xfs_iaccess(tip, S_IWUSR, NULL);
 180         if (error)
 181                 goto error0;
 182
 183         /* Verify that both files have the same format */
 184         if ((ip->i_d.di_mode & S_IFMT) != (tip->i_d.di_mode & S_IFMT)) {
 185                 error = XFS_ERROR(EINVAL);
 186                 goto error0;
 187         }
 188
 189         /* Verify both files are either real-time or non-realtime */
 190         if ((ip->i_d.di_flags & XFS_DIFLAG_REALTIME) !=
 191             (tip->i_d.di_flags & XFS_DIFLAG_REALTIME)) {
 192                 error = XFS_ERROR(EINVAL);
 193                 goto error0;
 194         }
 195
 196         /* Should never get a local format */
 197         if (ip->i_d.di_format == XFS_DINODE_FMT_LOCAL ||
 198             tip->i_d.di_format == XFS_DINODE_FMT_LOCAL) {
 199                 error = XFS_ERROR(EINVAL);
 200                 goto error0;
 201         }
 202
 203         if (VN_CACHED(tvp) != 0) {
 204                 xfs_inval_cached_trace(&tip->i_iocore, 0, -1, 0, -1);
 205                 VOP_FLUSHINVAL_PAGES(tvp, 0, -1, FI_REMAPF_LOCKED);
 206         }
 207
 208         /* Verify O_DIRECT for ftmp */
 209         if (VN_CACHED(tvp) != 0) {
 210                 error = XFS_ERROR(EINVAL);
 211                 goto error0;
 212         }
 213
 214         /* Verify all data are being swapped */
 215         if (sxp->sx_offset != 0 ||
 216             sxp->sx_length != ip->i_d.di_size ||
 217             sxp->sx_length != tip->i_d.di_size) {
 218                 error = XFS_ERROR(EFAULT);
 219                 goto error0;
 220         }
 221
 222         /*
 223          * If the target has extended attributes, the tmp file
 224          * must also in order to ensure the correct data fork
 225          * format.
 226          */
 227         if ( XFS_IFORK_Q(ip) != XFS_IFORK_Q(tip) ) {
 228                 error = XFS_ERROR(EINVAL);
 229                 goto error0;
 230         }
 231
 232         /*
 233          * Compare the current change & modify times with that
 234          * passed in.  If they differ, we abort this swap.
 235          * This is the mechanism used to ensure the calling
 236          * process that the file was not changed out from
 237          * under it.
 238          */
 239         if ((sbp->bs_ctime.tv_sec != ip->i_d.di_ctime.t_sec) ||
 240             (sbp->bs_ctime.tv_nsec != ip->i_d.di_ctime.t_nsec) ||
 241             (sbp->bs_mtime.tv_sec != ip->i_d.di_mtime.t_sec) ||
 242             (sbp->bs_mtime.tv_nsec != ip->i_d.di_mtime.t_nsec)) {
 243                 error = XFS_ERROR(EBUSY);
 244                 goto error0;
 245         }
 246
 247         /* We need to fail if the file is memory mapped.  Once we have tossed
 248          * all existing pages, the page fault will have no option
 249          * but to go to the filesystem for pages. By making the page fault call
 250          * VOP_READ (or write in the case of autogrow) they block on the iolock
 251          * until we have switched the extents.
 252          */
 253         if (VN_MAPPED(vp)) {
 254                 error = XFS_ERROR(EBUSY);
 255                 goto error0;
 256         }
 257
 258         xfs_iunlock(ip, XFS_ILOCK_EXCL);
 259         xfs_iunlock(tip, XFS_ILOCK_EXCL);
 260
 261         /*
 262          * There is a race condition here since we gave up the
 263          * ilock.  However, the data fork will not change since
 264          * we have the iolock (locked for truncation too) so we
 265          * are safe.  We don't really care if non-io related
 266          * fields change.
 267          */
 268
 269         VOP_TOSS_PAGES(vp, 0, -1, FI_REMAPF);
 270
 271         tp = xfs_trans_alloc(mp, XFS_TRANS_SWAPEXT);
 272         if ((error = xfs_trans_reserve(tp, 0,
 273                                      XFS_ICHANGE_LOG_RES(mp), 0,
 274                                      0, 0))) {
 275                 xfs_iunlock(ip,  XFS_IOLOCK_EXCL);
 276                 xfs_iunlock(tip, XFS_IOLOCK_EXCL);
 277                 xfs_trans_cancel(tp, 0);
 278                 locked = 0;
 279                 goto error0;
 280         }
 281         xfs_lock_inodes(ips, 2, 0, XFS_ILOCK_EXCL);
 282
 283         /*
 284          * Count the number of extended attribute blocks
 285          */
 286         if ( ((XFS_IFORK_Q(ip) != 0) && (ip->i_d.di_anextents > 0)) &&
 287              (ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
 288                 error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK, &aforkblks);
 289                 if (error) {
 290                         xfs_trans_cancel(tp, 0);
 291                         goto error0;
 292                 }
 293         }
 294         if ( ((XFS_IFORK_Q(tip) != 0) && (tip->i_d.di_anextents > 0)) &&
 295              (tip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
 296                 error = xfs_bmap_count_blocks(tp, tip, XFS_ATTR_FORK,
 297                         &taforkblks);
 298                 if (error) {
 299                         xfs_trans_cancel(tp, 0);
 300                         goto error0;
 301                 }
 302         }
 303
 304         /*
 305          * Swap the data forks of the inodes
 306          */
 307         ifp = &ip->i_df;
 308         tifp = &tip->i_df;
 309         *tempifp = *ifp;        /* struct copy */
 310         *ifp = *tifp;           /* struct copy */
 311         *tifp = *tempifp;       /* struct copy */
 312
 313         /*
 314          * Fix the on-disk inode values
 315          */
 316         tmp = (__uint64_t)ip->i_d.di_nblocks;
 317         ip->i_d.di_nblocks = tip->i_d.di_nblocks - taforkblks + aforkblks;
 318         tip->i_d.di_nblocks = tmp + taforkblks - aforkblks;
 319
 320         tmp = (__uint64_t) ip->i_d.di_nextents;
 321         ip->i_d.di_nextents = tip->i_d.di_nextents;
 322         tip->i_d.di_nextents = tmp;
 323
 324         tmp = (__uint64_t) ip->i_d.di_format;
 325         ip->i_d.di_format = tip->i_d.di_format;
 326         tip->i_d.di_format = tmp;
 327
 328         ilf_fields = XFS_ILOG_CORE;
 329
 330         switch(ip->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 (ip->i_d.di_nextents <= XFS_INLINE_EXTS) {
 337                         ifp->if_u1.if_extents =
 338                                 ifp->if_u2.if_inline_ext;
 339                 }
 340                 ilf_fields |= XFS_ILOG_DEXT;
 341                 break;
 342         case XFS_DINODE_FMT_BTREE:
 343                 ilf_fields |= XFS_ILOG_DBROOT;
 344                 break;
 345         }
 346
 347         tilf_fields = XFS_ILOG_CORE;
 348
 349         switch(tip->i_d.di_format) {
 350         case XFS_DINODE_FMT_EXTENTS:
 351                 /* If the extents fit in the inode, fix the
 352                  * pointer.  Otherwise it's already NULL or
 353                  * pointing to the extent.
 354                  */
 355                 if (tip->i_d.di_nextents <= XFS_INLINE_EXTS) {
 356                         tifp->if_u1.if_extents =
 357                                 tifp->if_u2.if_inline_ext;
 358                 }
 359                 tilf_fields |= XFS_ILOG_DEXT;
 360                 break;
 361         case XFS_DINODE_FMT_BTREE:
 362                 tilf_fields |= XFS_ILOG_DBROOT;
 363                 break;
 364         }
 365
 366         /*
 367          * Increment vnode ref counts since xfs_trans_commit &
 368          * xfs_trans_cancel will both unlock the inodes and
 369          * decrement the associated ref counts.
 370          */
 371         VN_HOLD(vp);
 372         VN_HOLD(tvp);
 373
 374         xfs_trans_ijoin(tp, ip, lock_flags);
 375         xfs_trans_ijoin(tp, tip, lock_flags);
 376
 377         xfs_trans_log_inode(tp, ip,  ilf_fields);
 378         xfs_trans_log_inode(tp, tip, tilf_fields);
 379
 380         /*
 381          * If this is a synchronous mount, make sure that the
 382          * transaction goes to disk before returning to the user.
 383          */
 384         if (mp->m_flags & XFS_MOUNT_WSYNC) {
 385                 xfs_trans_set_sync(tp);
 386         }
 387
 388         error = xfs_trans_commit(tp, XFS_TRANS_SWAPEXT, NULL);
 389         locked = 0;
 390
 391  error0:
 392         if (locked) {
 393                 xfs_iunlock(ip,  lock_flags);
 394                 xfs_iunlock(tip, lock_flags);
 395         }
 396         if (tempifp != NULL)
 397                 kmem_free(tempifp, sizeof(xfs_ifork_t));
 398         return error;
 399 }