/* * xfs_rename */ int xfs_rename( xfs_inode_t *src_dp, struct xfs_name *src_name, xfs_inode_t *src_ip, xfs_inode_t *target_dp, struct xfs_name *target_name, xfs_inode_t *target_ip) { xfs_trans_t *tp = NULL; xfs_mount_t *mp = src_dp->i_mount; int new_parent; /* moving to a new dir */ int src_is_directory; /* src_name is a directory */ int error; xfs_bmap_free_t free_list; xfs_fsblock_t first_block; int cancel_flags; int committed; xfs_inode_t *inodes[4]; int spaceres; int num_inodes; xfs_itrace_entry(src_dp); xfs_itrace_entry(target_dp); if (DM_EVENT_ENABLED(src_dp, DM_EVENT_RENAME) || DM_EVENT_ENABLED(target_dp, DM_EVENT_RENAME)) { error = XFS_SEND_NAMESP(mp, DM_EVENT_RENAME, src_dp, DM_RIGHT_NULL, target_dp, DM_RIGHT_NULL, src_name->name, target_name->name, 0, 0, 0); if (error) return error; } /* Return through std_return after this point. */ new_parent = (src_dp != target_dp); src_is_directory = ((src_ip->i_d.di_mode & S_IFMT) == S_IFDIR); if (src_is_directory) { /* * Check for link count overflow on target_dp */ if (target_ip == NULL && new_parent && target_dp->i_d.di_nlink >= XFS_MAXLINK) { error = XFS_ERROR(EMLINK); goto std_return; } } xfs_sort_for_rename(src_dp, target_dp, src_ip, target_ip, inodes, &num_inodes); xfs_bmap_init(&free_list, &first_block); tp = xfs_trans_alloc(mp, XFS_TRANS_RENAME); cancel_flags = XFS_TRANS_RELEASE_LOG_RES; spaceres = XFS_RENAME_SPACE_RES(mp, target_name->len); error = xfs_trans_reserve(tp, spaceres, XFS_RENAME_LOG_RES(mp), 0, XFS_TRANS_PERM_LOG_RES, XFS_RENAME_LOG_COUNT); if (error == ENOSPC) { spaceres = 0; error = xfs_trans_reserve(tp, 0, XFS_RENAME_LOG_RES(mp), 0, XFS_TRANS_PERM_LOG_RES, XFS_RENAME_LOG_COUNT); } if (error) { xfs_trans_cancel(tp, 0); goto std_return; } /* * Attach the dquots to the inodes */ if ((error = XFS_QM_DQVOPRENAME(mp, inodes))) { xfs_trans_cancel(tp, cancel_flags); goto std_return; } /* * Lock all the participating inodes. Depending upon whether * the target_name exists in the target directory, and * whether the target directory is the same as the source * directory, we can lock from 2 to 4 inodes. */ xfs_lock_inodes(inodes, num_inodes, XFS_ILOCK_EXCL); /* * Join all the inodes to the transaction. From this point on, * we can rely on either trans_commit or trans_cancel to unlock * them. Note that we need to add a vnode reference to the * directories since trans_commit & trans_cancel will decrement * them when they unlock the inodes. Also, we need to be careful * not to add an inode to the transaction more than once. */ IHOLD(src_dp); xfs_trans_ijoin(tp, src_dp, XFS_ILOCK_EXCL); if (new_parent) { IHOLD(target_dp); xfs_trans_ijoin(tp, target_dp, XFS_ILOCK_EXCL); } IHOLD(src_ip); xfs_trans_ijoin(tp, src_ip, XFS_ILOCK_EXCL); if (target_ip) { IHOLD(target_ip); xfs_trans_ijoin(tp, target_ip, XFS_ILOCK_EXCL); } /* * If we are using project inheritance, we only allow renames * into our tree when the project IDs are the same; else the * tree quota mechanism would be circumvented. */ if (unlikely((target_dp->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) && (target_dp->i_d.di_projid != src_ip->i_d.di_projid))) { error = XFS_ERROR(EXDEV); goto error_return; } /* * Set up the target. */ if (target_ip == NULL) { /* * If there's no space reservation, check the entry will * fit before actually inserting it. */ error = xfs_dir_canenter(tp, target_dp, target_name, spaceres); if (error) goto error_return; /* * If target does not exist and the rename crosses * directories, adjust the target directory link count * to account for the ".." reference from the new entry. */ error = xfs_dir_createname(tp, target_dp, target_name, src_ip->i_ino, &first_block, &free_list, spaceres); if (error == ENOSPC) goto error_return; if (error) goto abort_return; xfs_ichgtime(target_dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG); if (new_parent && src_is_directory) { error = xfs_bumplink(tp, target_dp); if (error) goto abort_return; } } else { /* target_ip != NULL */ /* * If target exists and it's a directory, check that both * target and source are directories and that target can be * destroyed, or that neither is a directory. */ if ((target_ip->i_d.di_mode & S_IFMT) == S_IFDIR) { /* * Make sure target dir is empty. */ if (!(xfs_dir_isempty(target_ip)) || (target_ip->i_d.di_nlink > 2)) { error = XFS_ERROR(EEXIST); goto error_return; } } /* * Link the source inode under the target name. * If the source inode is a directory and we are moving * it across directories, its ".." entry will be * inconsistent until we replace that down below. * * In case there is already an entry with the same * name at the destination directory, remove it first. */ error = xfs_dir_replace(tp, target_dp, target_name, src_ip->i_ino, &first_block, &free_list, spaceres); if (error) goto abort_return; xfs_ichgtime(target_dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG); /* * Decrement the link count on the target since the target * dir no longer points to it. */ error = xfs_droplink(tp, target_ip); if (error) goto abort_return; if (src_is_directory) { /* * Drop the link from the old "." entry. */ error = xfs_droplink(tp, target_ip); if (error) goto abort_return; } } /* target_ip != NULL */ /* * Remove the source. */ if (new_parent && src_is_directory) { /* * Rewrite the ".." entry to point to the new * directory. */ error = xfs_dir_replace(tp, src_ip, &xfs_name_dotdot, target_dp->i_ino, &first_block, &free_list, spaceres); ASSERT(error != EEXIST); if (error) goto abort_return; } /* * We always want to hit the ctime on the source inode. * * This isn't strictly required by the standards since the source * inode isn't really being changed, but old unix file systems did * it and some incremental backup programs won't work without it. */ xfs_ichgtime(src_ip, XFS_ICHGTIME_CHG); /* * Adjust the link count on src_dp. This is necessary when * renaming a directory, either within one parent when * the target existed, or across two parent directories. */ if (src_is_directory && (new_parent || target_ip != NULL)) { /* * Decrement link count on src_directory since the * entry that's moved no longer points to it. */ error = xfs_droplink(tp, src_dp); if (error) goto abort_return; } error = xfs_dir_removename(tp, src_dp, src_name, src_ip->i_ino, &first_block, &free_list, spaceres); if (error) goto abort_return; xfs_ichgtime(src_dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG); xfs_trans_log_inode(tp, src_dp, XFS_ILOG_CORE); if (new_parent) xfs_trans_log_inode(tp, target_dp, XFS_ILOG_CORE); /* * If this is a synchronous mount, make sure that the * rename transaction goes to disk before returning to * the user. */ if (mp->m_flags & (XFS_MOUNT_WSYNC|XFS_MOUNT_DIRSYNC)) { xfs_trans_set_sync(tp); } error = xfs_bmap_finish(&tp, &free_list, &committed); if (error) { xfs_bmap_cancel(&free_list); xfs_trans_cancel(tp, (XFS_TRANS_RELEASE_LOG_RES | XFS_TRANS_ABORT)); goto std_return; } /* * trans_commit will unlock src_ip, target_ip & decrement * the vnode references. */ error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES); /* Fall through to std_return with error = 0 or errno from * xfs_trans_commit */ std_return: if (DM_EVENT_ENABLED(src_dp, DM_EVENT_POSTRENAME) || DM_EVENT_ENABLED(target_dp, DM_EVENT_POSTRENAME)) { (void) XFS_SEND_NAMESP (mp, DM_EVENT_POSTRENAME, src_dp, DM_RIGHT_NULL, target_dp, DM_RIGHT_NULL, src_name->name, target_name->name, 0, error, 0); } return error; abort_return: cancel_flags |= XFS_TRANS_ABORT; /* FALLTHROUGH */ error_return: xfs_bmap_cancel(&free_list); xfs_trans_cancel(tp, cancel_flags); goto std_return; }
/* * xfs_rename */ int xfs_rename( bhv_desc_t *src_dir_bdp, bhv_vname_t *src_vname, bhv_vnode_t *target_dir_vp, bhv_vname_t *target_vname, cred_t *credp) { xfs_trans_t *tp; xfs_inode_t *src_dp, *target_dp, *src_ip, *target_ip; xfs_mount_t *mp; int new_parent; /* moving to a new dir */ int src_is_directory; /* src_name is a directory */ int error; xfs_bmap_free_t free_list; xfs_fsblock_t first_block; int cancel_flags; int committed; xfs_inode_t *inodes[4]; int target_ip_dropped = 0; /* dropped target_ip link? */ bhv_vnode_t *src_dir_vp; int spaceres; int target_link_zero = 0; int num_inodes; char *src_name = VNAME(src_vname); char *target_name = VNAME(target_vname); int src_namelen = VNAMELEN(src_vname); int target_namelen = VNAMELEN(target_vname); src_dir_vp = BHV_TO_VNODE(src_dir_bdp); vn_trace_entry(src_dir_vp, "xfs_rename", (inst_t *)__return_address); vn_trace_entry(target_dir_vp, "xfs_rename", (inst_t *)__return_address); /* * Find the XFS behavior descriptor for the target directory * vnode since it was not handed to us. */ target_dp = xfs_vtoi(target_dir_vp); if (target_dp == NULL) { return XFS_ERROR(EXDEV); } src_dp = XFS_BHVTOI(src_dir_bdp); mp = src_dp->i_mount; if (DM_EVENT_ENABLED(src_dir_vp->v_vfsp, src_dp, DM_EVENT_RENAME) || DM_EVENT_ENABLED(target_dir_vp->v_vfsp, target_dp, DM_EVENT_RENAME)) { error = XFS_SEND_NAMESP(mp, DM_EVENT_RENAME, src_dir_vp, DM_RIGHT_NULL, target_dir_vp, DM_RIGHT_NULL, src_name, target_name, 0, 0, 0); if (error) { return error; } } /* Return through std_return after this point. */ /* * Lock all the participating inodes. Depending upon whether * the target_name exists in the target directory, and * whether the target directory is the same as the source * directory, we can lock from 2 to 4 inodes. * xfs_lock_for_rename() will return ENOENT if src_name * does not exist in the source directory. */ tp = NULL; error = xfs_lock_for_rename(src_dp, target_dp, src_vname, target_vname, &src_ip, &target_ip, inodes, &num_inodes); if (error) { /* * We have nothing locked, no inode references, and * no transaction, so just get out. */ goto std_return; } ASSERT(src_ip != NULL); if ((src_ip->i_d.di_mode & S_IFMT) == S_IFDIR) { /* * Check for link count overflow on target_dp */ if (target_ip == NULL && (src_dp != target_dp) && target_dp->i_d.di_nlink >= XFS_MAXLINK) { error = XFS_ERROR(EMLINK); xfs_rename_unlock4(inodes, XFS_ILOCK_SHARED); goto rele_return; } } /* * If we are using project inheritance, we only allow renames * into our tree when the project IDs are the same; else the * tree quota mechanism would be circumvented. */ if (unlikely((target_dp->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) && (target_dp->i_d.di_projid != src_ip->i_d.di_projid))) { error = XFS_ERROR(EXDEV); xfs_rename_unlock4(inodes, XFS_ILOCK_SHARED); goto rele_return; } new_parent = (src_dp != target_dp); src_is_directory = ((src_ip->i_d.di_mode & S_IFMT) == S_IFDIR); /* * Drop the locks on our inodes so that we can start the transaction. */ xfs_rename_unlock4(inodes, XFS_ILOCK_SHARED); XFS_BMAP_INIT(&free_list, &first_block); tp = xfs_trans_alloc(mp, XFS_TRANS_RENAME); cancel_flags = XFS_TRANS_RELEASE_LOG_RES; spaceres = XFS_RENAME_SPACE_RES(mp, target_namelen); error = xfs_trans_reserve(tp, spaceres, XFS_RENAME_LOG_RES(mp), 0, XFS_TRANS_PERM_LOG_RES, XFS_RENAME_LOG_COUNT); if (error == ENOSPC) { spaceres = 0; error = xfs_trans_reserve(tp, 0, XFS_RENAME_LOG_RES(mp), 0, XFS_TRANS_PERM_LOG_RES, XFS_RENAME_LOG_COUNT); } if (error) { xfs_trans_cancel(tp, 0); goto rele_return; } /* * Attach the dquots to the inodes */ if ((error = XFS_QM_DQVOPRENAME(mp, inodes))) { xfs_trans_cancel(tp, cancel_flags); goto rele_return; } /* * Reacquire the inode locks we dropped above. */ xfs_lock_inodes(inodes, num_inodes, 0, XFS_ILOCK_EXCL); /* * Join all the inodes to the transaction. From this point on, * we can rely on either trans_commit or trans_cancel to unlock * them. Note that we need to add a vnode reference to the * directories since trans_commit & trans_cancel will decrement * them when they unlock the inodes. Also, we need to be careful * not to add an inode to the transaction more than once. */ VN_HOLD(src_dir_vp); xfs_trans_ijoin(tp, src_dp, XFS_ILOCK_EXCL); if (new_parent) { VN_HOLD(target_dir_vp); xfs_trans_ijoin(tp, target_dp, XFS_ILOCK_EXCL); } if ((src_ip != src_dp) && (src_ip != target_dp)) { xfs_trans_ijoin(tp, src_ip, XFS_ILOCK_EXCL); } if ((target_ip != NULL) && (target_ip != src_ip) && (target_ip != src_dp) && (target_ip != target_dp)) { xfs_trans_ijoin(tp, target_ip, XFS_ILOCK_EXCL); } /* * Set up the target. */ if (target_ip == NULL) { /* * If there's no space reservation, check the entry will * fit before actually inserting it. */ if (spaceres == 0 && (error = xfs_dir_canenter(tp, target_dp, target_name, target_namelen))) goto error_return; /* * If target does not exist and the rename crosses * directories, adjust the target directory link count * to account for the ".." reference from the new entry. */ error = xfs_dir_createname(tp, target_dp, target_name, target_namelen, src_ip->i_ino, &first_block, &free_list, spaceres); if (error == ENOSPC) goto error_return; if (error) goto abort_return; xfs_ichgtime(target_dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG); if (new_parent && src_is_directory) { error = xfs_bumplink(tp, target_dp); if (error) goto abort_return; } } else { /* target_ip != NULL */ /* * If target exists and it's a directory, check that both * target and source are directories and that target can be * destroyed, or that neither is a directory. */ if ((target_ip->i_d.di_mode & S_IFMT) == S_IFDIR) { /* * Make sure target dir is empty. */ if (!(xfs_dir_isempty(target_ip)) || (target_ip->i_d.di_nlink > 2)) { error = XFS_ERROR(EEXIST); goto error_return; } } /* * Link the source inode under the target name. * If the source inode is a directory and we are moving * it across directories, its ".." entry will be * inconsistent until we replace that down below. * * In case there is already an entry with the same * name at the destination directory, remove it first. */ error = xfs_dir_replace(tp, target_dp, target_name, target_namelen, src_ip->i_ino, &first_block, &free_list, spaceres); if (error) goto abort_return; xfs_ichgtime(target_dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG); /* * Decrement the link count on the target since the target * dir no longer points to it. */ error = xfs_droplink(tp, target_ip); if (error) goto abort_return; target_ip_dropped = 1; if (src_is_directory) { /* * Drop the link from the old "." entry. */ error = xfs_droplink(tp, target_ip); if (error) goto abort_return; } /* Do this test while we still hold the locks */ target_link_zero = (target_ip)->i_d.di_nlink==0; } /* target_ip != NULL */ /* * Remove the source. */ if (new_parent && src_is_directory) { /* * Rewrite the ".." entry to point to the new * directory. */ error = xfs_dir_replace(tp, src_ip, "..", 2, target_dp->i_ino, &first_block, &free_list, spaceres); ASSERT(error != EEXIST); if (error) goto abort_return; xfs_ichgtime(src_ip, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG); } else { /* * We always want to hit the ctime on the source inode. * We do it in the if clause above for the 'new_parent && * src_is_directory' case, and here we get all the other * cases. This isn't strictly required by the standards * since the source inode isn't really being changed, * but old unix file systems did it and some incremental * backup programs won't work without it. */ xfs_ichgtime(src_ip, XFS_ICHGTIME_CHG); } /* * Adjust the link count on src_dp. This is necessary when * renaming a directory, either within one parent when * the target existed, or across two parent directories. */ if (src_is_directory && (new_parent || target_ip != NULL)) { /* * Decrement link count on src_directory since the * entry that's moved no longer points to it. */ error = xfs_droplink(tp, src_dp); if (error) goto abort_return; } error = xfs_dir_removename(tp, src_dp, src_name, src_namelen, src_ip->i_ino, &first_block, &free_list, spaceres); if (error) goto abort_return; xfs_ichgtime(src_dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG); /* * Update the generation counts on all the directory inodes * that we're modifying. */ src_dp->i_gen++; xfs_trans_log_inode(tp, src_dp, XFS_ILOG_CORE); if (new_parent) { target_dp->i_gen++; xfs_trans_log_inode(tp, target_dp, XFS_ILOG_CORE); } /* * If there was a target inode, take an extra reference on * it here so that it doesn't go to xfs_inactive() from * within the commit. */ if (target_ip != NULL) { IHOLD(target_ip); } /* * If this is a synchronous mount, make sure that the * rename transaction goes to disk before returning to * the user. */ if (mp->m_flags & (XFS_MOUNT_WSYNC|XFS_MOUNT_DIRSYNC)) { xfs_trans_set_sync(tp); } /* * Take refs. for vop_link_removed calls below. No need to worry * about directory refs. because the caller holds them. * * Do holds before the xfs_bmap_finish since it might rele them down * to zero. */ if (target_ip_dropped) IHOLD(target_ip); IHOLD(src_ip); error = xfs_bmap_finish(&tp, &free_list, &committed); if (error) { xfs_bmap_cancel(&free_list); xfs_trans_cancel(tp, (XFS_TRANS_RELEASE_LOG_RES | XFS_TRANS_ABORT)); if (target_ip != NULL) { IRELE(target_ip); } if (target_ip_dropped) { IRELE(target_ip); } IRELE(src_ip); goto std_return; } /* * trans_commit will unlock src_ip, target_ip & decrement * the vnode references. */ error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES); if (target_ip != NULL) { xfs_refcache_purge_ip(target_ip); IRELE(target_ip); } /* * Let interposed file systems know about removed links. */ if (target_ip_dropped) { bhv_vop_link_removed(XFS_ITOV(target_ip), target_dir_vp, target_link_zero); IRELE(target_ip); } IRELE(src_ip); /* Fall through to std_return with error = 0 or errno from * xfs_trans_commit */ std_return: if (DM_EVENT_ENABLED(src_dir_vp->v_vfsp, src_dp, DM_EVENT_POSTRENAME) || DM_EVENT_ENABLED(target_dir_vp->v_vfsp, target_dp, DM_EVENT_POSTRENAME)) { (void) XFS_SEND_NAMESP (mp, DM_EVENT_POSTRENAME, src_dir_vp, DM_RIGHT_NULL, target_dir_vp, DM_RIGHT_NULL, src_name, target_name, 0, error, 0); } return error; abort_return: cancel_flags |= XFS_TRANS_ABORT; /* FALLTHROUGH */ error_return: xfs_bmap_cancel(&free_list); xfs_trans_cancel(tp, cancel_flags); goto std_return; rele_return: IRELE(src_ip); if (target_ip != NULL) { IRELE(target_ip); } goto std_return; }
/* * The following routine will acquire the locks required for a rename * operation. The code understands the semantics of renames and will * validate that name1 exists under dp1 & that name2 may or may not * exist under dp2. * * We are renaming dp1/name1 to dp2/name2. * * Return ENOENT if dp1 does not exist, other lookup errors, or 0 for success. */ STATIC int xfs_lock_for_rename( xfs_inode_t *dp1, /* old (source) directory inode */ xfs_inode_t *dp2, /* new (target) directory inode */ bhv_vname_t *vname1,/* old entry name */ bhv_vname_t *vname2,/* new entry name */ xfs_inode_t **ipp1, /* inode of old entry */ xfs_inode_t **ipp2, /* inode of new entry, if it already exists, NULL otherwise. */ xfs_inode_t **i_tab,/* array of inode returned, sorted */ int *num_inodes) /* number of inodes in array */ { xfs_inode_t *ip1, *ip2, *temp; xfs_ino_t inum1, inum2; int error; int i, j; uint lock_mode; int diff_dirs = (dp1 != dp2); ip2 = NULL; /* * First, find out the current inums of the entries so that we * can determine the initial locking order. We'll have to * sanity check stuff after all the locks have been acquired * to see if we still have the right inodes, directories, etc. */ lock_mode = xfs_ilock_map_shared(dp1); error = xfs_get_dir_entry(vname1, &ip1); if (error) { xfs_iunlock_map_shared(dp1, lock_mode); return error; } inum1 = ip1->i_ino; ASSERT(ip1); ITRACE(ip1); /* * Unlock dp1 and lock dp2 if they are different. */ if (diff_dirs) { xfs_iunlock_map_shared(dp1, lock_mode); lock_mode = xfs_ilock_map_shared(dp2); } error = xfs_dir_lookup_int(XFS_ITOBHV(dp2), lock_mode, vname2, &inum2, &ip2); if (error == ENOENT) { /* target does not need to exist. */ inum2 = 0; } else if (error) { /* * If dp2 and dp1 are the same, the next line unlocks dp1. * Got it? */ xfs_iunlock_map_shared(dp2, lock_mode); IRELE (ip1); return error; } else { ITRACE(ip2); } /* * i_tab contains a list of pointers to inodes. We initialize * the table here & we'll sort it. We will then use it to * order the acquisition of the inode locks. * * Note that the table may contain duplicates. e.g., dp1 == dp2. */ i_tab[0] = dp1; i_tab[1] = dp2; i_tab[2] = ip1; if (inum2 == 0) { *num_inodes = 3; i_tab[3] = NULL; } else { *num_inodes = 4; i_tab[3] = ip2; } /* * Sort the elements via bubble sort. (Remember, there are at * most 4 elements to sort, so this is adequate.) */ for (i=0; i < *num_inodes; i++) { for (j=1; j < *num_inodes; j++) { if (i_tab[j]->i_ino < i_tab[j-1]->i_ino) { temp = i_tab[j]; i_tab[j] = i_tab[j-1]; i_tab[j-1] = temp; } } } /* * We have dp2 locked. If it isn't first, unlock it. * If it is first, tell xfs_lock_inodes so it can skip it * when locking. if dp1 == dp2, xfs_lock_inodes will skip both * since they are equal. xfs_lock_inodes needs all these inodes * so that it can unlock and retry if there might be a dead-lock * potential with the log. */ if (i_tab[0] == dp2 && lock_mode == XFS_ILOCK_SHARED) { #ifdef DEBUG xfs_rename_skip++; #endif xfs_lock_inodes(i_tab, *num_inodes, 1, XFS_ILOCK_SHARED); } else { #ifdef DEBUG xfs_rename_nskip++; #endif xfs_iunlock_map_shared(dp2, lock_mode); xfs_lock_inodes(i_tab, *num_inodes, 0, XFS_ILOCK_SHARED); } /* * Set the return value. Null out any unused entries in i_tab. */ *ipp1 = *ipp2 = NULL; for (i=0; i < *num_inodes; i++) { if (i_tab[i]->i_ino == inum1) { *ipp1 = i_tab[i]; } if (i_tab[i]->i_ino == inum2) { *ipp2 = i_tab[i]; } } for (;i < 4; i++) { i_tab[i] = NULL; } return 0; }
/* * Syssgi interface for swapext */ int xfs_swapext( xfs_swapext_t __user *sxp) { xfs_swapext_t sx; xfs_inode_t *ip=NULL, *tip=NULL, *ips[2]; xfs_trans_t *tp; xfs_mount_t *mp; xfs_bstat_t *sbp; struct file *fp = NULL, *tfp = NULL; vnode_t *vp, *tvp; bhv_desc_t *bdp, *tbdp; vn_bhv_head_t *bhp, *tbhp; uint lock_flags=0; int ilf_fields, tilf_fields; int error = 0; xfs_ifork_t tempif, *ifp, *tifp; __uint64_t tmp; int aforkblks = 0; int taforkblks = 0; int locked = 0; if (copy_from_user(&sx, sxp, sizeof(sx))) return XFS_ERROR(EFAULT); /* Pull information for the target fd */ if (((fp = fget((int)sx.sx_fdtarget)) == NULL) || ((vp = LINVFS_GET_VP(fp->f_dentry->d_inode)) == NULL)) { error = XFS_ERROR(EINVAL); goto error0; } bhp = VN_BHV_HEAD(vp); bdp = vn_bhv_lookup(bhp, &xfs_vnodeops); if (bdp == NULL) { error = XFS_ERROR(EBADF); goto error0; } else { ip = XFS_BHVTOI(bdp); } if (((tfp = fget((int)sx.sx_fdtmp)) == NULL) || ((tvp = LINVFS_GET_VP(tfp->f_dentry->d_inode)) == NULL)) { error = XFS_ERROR(EINVAL); goto error0; } tbhp = VN_BHV_HEAD(tvp); tbdp = vn_bhv_lookup(tbhp, &xfs_vnodeops); if (tbdp == NULL) { error = XFS_ERROR(EBADF); goto error0; } else { tip = XFS_BHVTOI(tbdp); } if (ip->i_mount != tip->i_mount) { error = XFS_ERROR(EINVAL); goto error0; } if (ip->i_ino == tip->i_ino) { error = XFS_ERROR(EINVAL); goto error0; } mp = ip->i_mount; sbp = &sx.sx_stat; if (XFS_FORCED_SHUTDOWN(mp)) { error = XFS_ERROR(EIO); goto error0; } locked = 1; /* Lock in i_ino order */ if (ip->i_ino < tip->i_ino) { ips[0] = ip; ips[1] = tip; } else { ips[0] = tip; ips[1] = ip; } lock_flags = XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL; xfs_lock_inodes(ips, 2, 0, lock_flags); /* Check permissions */ error = xfs_iaccess(ip, S_IWUSR, NULL); if (error) goto error0; error = xfs_iaccess(tip, S_IWUSR, NULL); if (error) goto error0; /* Verify that both files have the same format */ if ((ip->i_d.di_mode & S_IFMT) != (tip->i_d.di_mode & S_IFMT)) { error = XFS_ERROR(EINVAL); goto error0; } /* Verify both files are either real-time or non-realtime */ if ((ip->i_d.di_flags & XFS_DIFLAG_REALTIME) != (tip->i_d.di_flags & XFS_DIFLAG_REALTIME)) { error = XFS_ERROR(EINVAL); goto error0; } /* Should never get a local format */ if (ip->i_d.di_format == XFS_DINODE_FMT_LOCAL || tip->i_d.di_format == XFS_DINODE_FMT_LOCAL) { error = XFS_ERROR(EINVAL); goto error0; } if (VN_CACHED(tvp) != 0) xfs_inval_cached_pages(XFS_ITOV(tip), &(tip->i_iocore), (xfs_off_t)0, 0, 0); /* Verify O_DIRECT for ftmp */ if (VN_CACHED(tvp) != 0) { error = XFS_ERROR(EINVAL); goto error0; } /* Verify all data are being swapped */ if (sx.sx_offset != 0 || sx.sx_length != ip->i_d.di_size || sx.sx_length != tip->i_d.di_size) { error = XFS_ERROR(EFAULT); goto error0; } /* * If the target has extended attributes, the tmp file * must also in order to ensure the correct data fork * format. */ if ( XFS_IFORK_Q(ip) != XFS_IFORK_Q(tip) ) { error = XFS_ERROR(EINVAL); goto error0; } /* * Compare the current change & modify times with that * passed in. If they differ, we abort this swap. * This is the mechanism used to ensure the calling * process that the file was not changed out from * under it. */ if ((sbp->bs_ctime.tv_sec != ip->i_d.di_ctime.t_sec) || (sbp->bs_ctime.tv_nsec != ip->i_d.di_ctime.t_nsec) || (sbp->bs_mtime.tv_sec != ip->i_d.di_mtime.t_sec) || (sbp->bs_mtime.tv_nsec != ip->i_d.di_mtime.t_nsec)) { error = XFS_ERROR(EBUSY); goto error0; } /* We need to fail if the file is memory mapped. Once we have tossed * all existing pages, the page fault will have no option * but to go to the filesystem for pages. By making the page fault call * VOP_READ (or write in the case of autogrow) they block on the iolock * until we have switched the extents. */ if (VN_MAPPED(vp)) { error = XFS_ERROR(EBUSY); goto error0; } xfs_iunlock(ip, XFS_ILOCK_EXCL); xfs_iunlock(tip, XFS_ILOCK_EXCL); /* * There is a race condition here since we gave up the * ilock. However, the data fork will not change since * we have the iolock (locked for truncation too) so we * are safe. We don't really care if non-io related * fields change. */ VOP_TOSS_PAGES(vp, 0, -1, FI_REMAPF); tp = xfs_trans_alloc(mp, XFS_TRANS_SWAPEXT); if ((error = xfs_trans_reserve(tp, 0, XFS_ICHANGE_LOG_RES(mp), 0, 0, 0))) { xfs_iunlock(ip, XFS_IOLOCK_EXCL); xfs_iunlock(tip, XFS_IOLOCK_EXCL); xfs_trans_cancel(tp, 0); return error; } xfs_lock_inodes(ips, 2, 0, XFS_ILOCK_EXCL); /* * Count the number of extended attribute blocks */ if ( ((XFS_IFORK_Q(ip) != 0) && (ip->i_d.di_anextents > 0)) && (ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) { error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK, &aforkblks); if (error) { xfs_iunlock(ip, lock_flags); xfs_iunlock(tip, lock_flags); xfs_trans_cancel(tp, 0); return error; } } if ( ((XFS_IFORK_Q(tip) != 0) && (tip->i_d.di_anextents > 0)) && (tip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) { error = xfs_bmap_count_blocks(tp, tip, XFS_ATTR_FORK, &taforkblks); if (error) { xfs_iunlock(ip, lock_flags); xfs_iunlock(tip, lock_flags); xfs_trans_cancel(tp, 0); return error; } } /* * Swap the data forks of the inodes */ ifp = &ip->i_df; tifp = &tip->i_df; tempif = *ifp; /* struct copy */ *ifp = *tifp; /* struct copy */ *tifp = tempif; /* struct copy */ /* * Fix the on-disk inode values */ tmp = (__uint64_t)ip->i_d.di_nblocks; ip->i_d.di_nblocks = tip->i_d.di_nblocks - taforkblks + aforkblks; tip->i_d.di_nblocks = tmp + taforkblks - aforkblks; tmp = (__uint64_t) ip->i_d.di_nextents; ip->i_d.di_nextents = tip->i_d.di_nextents; tip->i_d.di_nextents = tmp; tmp = (__uint64_t) ip->i_d.di_format; ip->i_d.di_format = tip->i_d.di_format; tip->i_d.di_format = tmp; ilf_fields = XFS_ILOG_CORE; switch(ip->i_d.di_format) { case XFS_DINODE_FMT_EXTENTS: /* If the extents fit in the inode, fix the * pointer. Otherwise it's already NULL or * pointing to the extent. */ if (ip->i_d.di_nextents <= XFS_INLINE_EXTS) { ifp->if_u1.if_extents = ifp->if_u2.if_inline_ext; } ilf_fields |= XFS_ILOG_DEXT; break; case XFS_DINODE_FMT_BTREE: ilf_fields |= XFS_ILOG_DBROOT; break; } tilf_fields = XFS_ILOG_CORE; switch(tip->i_d.di_format) { case XFS_DINODE_FMT_EXTENTS: /* If the extents fit in the inode, fix the * pointer. Otherwise it's already NULL or * pointing to the extent. */ if (tip->i_d.di_nextents <= XFS_INLINE_EXTS) { tifp->if_u1.if_extents = tifp->if_u2.if_inline_ext; } tilf_fields |= XFS_ILOG_DEXT; break; case XFS_DINODE_FMT_BTREE: tilf_fields |= XFS_ILOG_DBROOT; break; } /* * Increment vnode ref counts since xfs_trans_commit & * xfs_trans_cancel will both unlock the inodes and * decrement the associated ref counts. */ VN_HOLD(vp); VN_HOLD(tvp); xfs_trans_ijoin(tp, ip, lock_flags); xfs_trans_ijoin(tp, tip, lock_flags); xfs_trans_log_inode(tp, ip, ilf_fields); xfs_trans_log_inode(tp, tip, tilf_fields); /* * If this is a synchronous mount, make sure that the * transaction goes to disk before returning to the user. */ if (mp->m_flags & XFS_MOUNT_WSYNC) { xfs_trans_set_sync(tp); } error = xfs_trans_commit(tp, XFS_TRANS_SWAPEXT, NULL); fput(fp); fput(tfp); return error; error0: if (locked) { xfs_iunlock(ip, lock_flags); xfs_iunlock(tip, lock_flags); } if (fp != NULL) fput(fp); if (tfp != NULL) fput(tfp); return error; }