namei.c

ocfs1.4.1 oracle分布式文件系统

	struct buffer_head *new_dir_bh = NULL;
	struct ocfs2_dir_entry *old_inode_dot_dot_de = NULL, *old_de = NULL,
		*new_de = NULL;
	struct buffer_head *new_de_bh = NULL, *old_de_bh = NULL; // bhs for above
	struct buffer_head *old_inode_de_bh = NULL; // if old_dentry is a dir,
						    // this is the 1st dirent bh
	nlink_t old_dir_nlink = old_dir->i_nlink;
	struct ocfs2_dinode *old_di;

	/* At some point it might be nice to break this function up a
	 * bit. */

	mlog_entry("(0x%p, 0x%p, 0x%p, 0x%p, from='%.*s' to='%.*s')\n",
		   old_dir, old_dentry, new_dir, new_dentry,
		   old_dentry->d_name.len, old_dentry->d_name.name,
		   new_dentry->d_name.len, new_dentry->d_name.name);

	osb = OCFS2_SB(old_dir->i_sb);

	if (new_inode) {
		if (!igrab(new_inode))
			BUG();
	}

	/* Assume a directory hierarchy thusly:
	 * a/b/c
	 * a/d
	 * a,b,c, and d are all directories.
	 *
	 * from cwd of 'a' on both nodes:
	 * node1: mv b/c d
	 * node2: mv d   b/c
	 *
	 * And that's why, just like the VFS, we need a file system
	 * rename lock. */
	if (old_dir != new_dir && S_ISDIR(old_inode->i_mode)) {
		status = ocfs2_rename_lock(osb);
		if (status < 0) {
			mlog_errno(status);
			goto bail;
		}
		rename_lock = 1;
	}

	/* if old and new are the same, this'll just do one lock. */
	status = ocfs2_double_lock(osb, &old_dir_bh, old_dir,
				   &new_dir_bh, new_dir);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}
	parents_locked = 1;

	/* make sure both dirs have bhs
	 * get an extra ref on old_dir_bh if old==new */
	if (!new_dir_bh) {
		if (old_dir_bh) {
			new_dir_bh = old_dir_bh;
			get_bh(new_dir_bh);
		} else {
			mlog(ML_ERROR, "no old_dir_bh!\n");
			status = -EIO;
			goto bail;
		}
	}

	/*
	 * Aside from allowing a meta data update, the locking here
	 * also ensures that the downconvert thread on other nodes
	 * won't have to concurrently downconvert the inode and the
	 * dentry locks.
	 */
	status = ocfs2_inode_lock(old_inode, &old_inode_bh, 1);
	if (status < 0) {
		if (status != -ENOENT)
			mlog_errno(status);
		goto bail;
	}
	old_child_locked = 1;

	status = ocfs2_remote_dentry_delete(old_dentry);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

	if (S_ISDIR(old_inode->i_mode)) {
		u64 old_inode_parent;

		status = ocfs2_find_files_on_disk("..", 2, &old_inode_parent,
						  old_inode, &old_inode_de_bh,
						  &old_inode_dot_dot_de);
		if (status) {
			status = -EIO;
			goto bail;
		}

		if (old_inode_parent != OCFS2_I(old_dir)->ip_blkno) {
			status = -EIO;
			goto bail;
		}

		if (!new_inode && new_dir != old_dir &&
		    new_dir->i_nlink >= OCFS2_LINK_MAX) {
			status = -EMLINK;
			goto bail;
		}
	}

	status = ocfs2_lookup_ino_from_name(old_dir, old_dentry->d_name.name,
					    old_dentry->d_name.len,
					    &old_de_ino);
	if (status) {
		status = -ENOENT;
		goto bail;
	}

	/*
	 *  Check for inode number is _not_ due to possible IO errors.
	 *  We might rmdir the source, keep it as pwd of some process
	 *  and merrily kill the link to whatever was created under the
	 *  same name. Goodbye sticky bit ;-<
	 */
	if (old_de_ino != OCFS2_I(old_inode)->ip_blkno) {
		status = -ENOENT;
		goto bail;
	}

	/* check if the target already exists (in which case we need
	 * to delete it */
	status = ocfs2_find_files_on_disk(new_dentry->d_name.name,
					  new_dentry->d_name.len,
					  &newfe_blkno, new_dir, &new_de_bh,
					  &new_de);
	/* The only error we allow here is -ENOENT because the new
	 * file not existing is perfectly valid. */
	if ((status < 0) && (status != -ENOENT)) {
		/* If we cannot find the file specified we should just */
		/* return the error... */
		mlog_errno(status);
		goto bail;
	}

	if (!new_de && new_inode) {
		/*
		 * Target was unlinked by another node while we were
		 * waiting to get to ocfs2_rename(). There isn't
		 * anything we can do here to help the situation, so
		 * bubble up the appropriate error.
		 */
		status = -ENOENT;
		goto bail;
	}

	/* In case we need to overwrite an existing file, we blow it
	 * away first */
	if (new_de) {
		/* VFS didn't think there existed an inode here, but
		 * someone else in the cluster must have raced our
		 * rename to create one. Today we error cleanly, in
		 * the future we should consider calling iget to build
		 * a new struct inode for this entry. */
		if (!new_inode) {
			status = -EACCES;

			mlog(0, "We found an inode for name %.*s but VFS "
			     "didn't give us one.\n", new_dentry->d_name.len,
			     new_dentry->d_name.name);
			goto bail;
		}

		if (OCFS2_I(new_inode)->ip_blkno != newfe_blkno) {
			status = -EACCES;

			mlog(0, "Inode %llu and dir %llu disagree. flags = %x\n",
			     (unsigned long long)OCFS2_I(new_inode)->ip_blkno,
			     (unsigned long long)newfe_blkno,
			     OCFS2_I(new_inode)->ip_flags);
			goto bail;
		}

		status = ocfs2_inode_lock(new_inode, &newfe_bh, 1);
		if (status < 0) {
			if (status != -ENOENT)
				mlog_errno(status);
			goto bail;
		}
		new_child_locked = 1;

		status = ocfs2_remote_dentry_delete(new_dentry);
		if (status < 0) {
			mlog_errno(status);
			goto bail;
		}

		newfe = (struct ocfs2_dinode *) newfe_bh->b_data;

		mlog(0, "aha rename over existing... new_de=%p new_blkno=%llu "
		     "newfebh=%p bhblocknr=%llu\n", new_de,
		     (unsigned long long)newfe_blkno, newfe_bh, newfe_bh ?
		     (unsigned long long)newfe_bh->b_blocknr : 0ULL);

		if (S_ISDIR(new_inode->i_mode) || (new_inode->i_nlink == 1)) {
			status = ocfs2_prepare_orphan_dir(osb, &orphan_dir,
							  new_inode,
							  orphan_name,
							  &orphan_entry_bh);
			if (status < 0) {
				mlog_errno(status);
				goto bail;
			}
		}
	} else {
		BUG_ON(new_dentry->d_parent->d_inode != new_dir);

		status = ocfs2_check_dir_for_entry(new_dir,
						   new_dentry->d_name.name,
						   new_dentry->d_name.len);
		if (status)
			goto bail;

		status = ocfs2_prepare_dir_for_insert(osb, new_dir, new_dir_bh,
						      new_dentry->d_name.name,
						      new_dentry->d_name.len,
						      &insert_entry_bh);
		if (status < 0) {
			mlog_errno(status);
			goto bail;
		}
	}

	handle = ocfs2_start_trans(osb, OCFS2_RENAME_CREDITS);
	if (IS_ERR(handle)) {
		status = PTR_ERR(handle);
		handle = NULL;
		mlog_errno(status);
		goto bail;
	}

	if (new_de) {
		if (S_ISDIR(new_inode->i_mode)) {
			if (!ocfs2_empty_dir(new_inode) ||
			    new_inode->i_nlink != 2) {
				status = -ENOTEMPTY;
				goto bail;
			}
		}
		status = ocfs2_journal_access(handle, new_inode, newfe_bh,
					      OCFS2_JOURNAL_ACCESS_WRITE);
		if (status < 0) {
			mlog_errno(status);
			goto bail;
		}

		if (S_ISDIR(new_inode->i_mode) ||
		    (newfe->i_links_count == cpu_to_le16(1))){
			status = ocfs2_orphan_add(osb, handle, new_inode,
						  newfe, orphan_name,
						  orphan_entry_bh, orphan_dir);
			if (status < 0) {
				mlog_errno(status);
				goto bail;
			}
		}

		/* change the dirent to point to the correct inode */
		status = ocfs2_update_entry(new_dir, handle, new_de_bh,
					    new_de, old_inode);
		if (status < 0) {
			mlog_errno(status);
			goto bail;
		}
		new_dir->i_version++;

		if (S_ISDIR(new_inode->i_mode))
			newfe->i_links_count = 0;
		else
			le16_add_cpu(&newfe->i_links_count, -1);

		status = ocfs2_journal_dirty(handle, newfe_bh);
		if (status < 0) {
			mlog_errno(status);
			goto bail;
		}
	} else {
		/* if the name was not found in new_dir, add it now */
		status = ocfs2_add_entry(handle, new_dentry, old_inode,
					 OCFS2_I(old_inode)->ip_blkno,
					 new_dir_bh, insert_entry_bh);
	}

	old_inode->i_ctime = CURRENT_TIME;
	mark_inode_dirty(old_inode);

	status = ocfs2_journal_access(handle, old_inode, old_inode_bh,
				      OCFS2_JOURNAL_ACCESS_WRITE);
	if (status >= 0) {
		old_di = (struct ocfs2_dinode *) old_inode_bh->b_data;

		old_di->i_ctime = cpu_to_le64(old_inode->i_ctime.tv_sec);
		old_di->i_ctime_nsec = cpu_to_le32(old_inode->i_ctime.tv_nsec);

		status = ocfs2_journal_dirty(handle, old_inode_bh);
		if (status < 0)
			mlog_errno(status);
	} else
		mlog_errno(status);

	/*
	 * Now that the name has been added to new_dir, remove the old name.
	 *
	 * We don't keep any directory entry context around until now
	 * because the insert might have changed the type of directory
	 * we're dealing with.
	 */
	old_de_bh = ocfs2_find_entry(old_dentry->d_name.name,
				     old_dentry->d_name.len,
				     old_dir, &old_de);
	if (!old_de_bh) {
		status = -EIO;
		goto bail;
	}

	status = ocfs2_delete_entry(handle, old_dir, old_de, old_de_bh);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

	if (new_inode) {
		new_inode->i_nlink--;
		new_inode->i_ctime = CURRENT_TIME;
	}
	old_dir->i_ctime = old_dir->i_mtime = CURRENT_TIME;
	if (old_inode_de_bh) {
		status = ocfs2_update_entry(old_inode, handle, old_inode_de_bh,
					    old_inode_dot_dot_de, new_dir);
		old_dir->i_nlink--;
		if (new_inode) {
			new_inode->i_nlink--;
		} else {
			inc_nlink(new_dir);
			mark_inode_dirty(new_dir);
		}
	}
	mark_inode_dirty(old_dir);
	ocfs2_mark_inode_dirty(handle, old_dir, old_dir_bh);
	if (new_inode) {
		mark_inode_dirty(new_inode);
		ocfs2_mark_inode_dirty(handle, new_inode, newfe_bh);
	}

	if (old_dir != new_dir) {
		/* Keep the same times on both directories.*/
		new_dir->i_ctime = new_dir->i_mtime = old_dir->i_ctime;

		/*
		 * This will also pick up the i_nlink change from the
		 * block above.
		 */
		ocfs2_mark_inode_dirty(handle, new_dir, new_dir_bh);
	}

	if (old_dir_nlink != old_dir->i_nlink) {
		if (!old_dir_bh) {
			mlog(ML_ERROR, "need to change nlink for old dir "
			     "%llu from %d to %d but bh is NULL!\n",
			     (unsigned long long)OCFS2_I(old_dir)->ip_blkno,
			     (int)old_dir_nlink, old_dir->i_nlink);
		} else {
			struct ocfs2_dinode *fe;
			status = ocfs2_journal_access(handle, old_dir,
						      old_dir_bh,
						      OCFS2_JOURNAL_ACCESS_WRITE);
			fe = (struct ocfs2_dinode *) old_dir_bh->b_data;
			fe->i_links_count = cpu_to_le16(old_dir->i_nlink);
			status = ocfs2_journal_dirty(handle, old_dir_bh);
		}
	}

	ocfs2_dentry_move(old_dentry, new_dentry, old_dir, new_dir);
	status = 0;
bail:
	if (rename_lock)
		ocfs2_rename_unlock(osb);

	if (handle)
		ocfs2_commit_trans(osb, handle);

	if (parents_locked)
		ocfs2_double_unlock(old_dir, new_dir);

	if (old_child_locked)
		ocfs2_inode_unlock(old_inode, 1);

	if (new_child_locked)
		ocfs2_inode_unlock(new_inode, 1);

	if (orphan_dir) {
		/* This was locked for us in ocfs2_prepare_orphan_dir() */
		ocfs2_inode_unlock(orphan_dir, 1);
		mutex_unlock(&orphan_dir->i_mutex);
		iput(orphan_dir);
	}

	if (new_inode)
		sync_mapping_buffers(old_inode->i_mapping);

	if (new_inode)
		iput(new_inode);
	if (newfe_bh)
		brelse(newfe_bh);
	if (old_inode_bh)
		brelse(old_inode_bh);
	if (old_dir_bh)
		brelse(old_dir_bh);
	if (new_dir_bh)
		brelse(new_dir_bh);
	if (new_de_bh)
		brelse(new_de_bh);
	if (old_de_bh)
		brelse(old_de_bh);
	if (old_inode_de_bh)
		brelse(old_inode_de_bh);
	if (orphan_entry_bh)
		brelse(orphan_entry_bh);
	if (insert_entry_bh)
		brelse(insert_entry_bh);

	mlog_exit(status);

	return status;
}

/*
 * we expect i_size = strlen(symname). Copy symname into the file
 * data, including the null terminator.
 */
static int ocfs2_create_symlink_data(struct ocfs2_super *osb,
				     handle_t *handle,
				     struct inode *inode,
				     const char *symname)
{
	struct buffer_head **bhs = NULL;
	const char *c;
	struct super_block *sb = osb->sb;
	u64 p_blkno, p_blocks;
	int virtual, blocks, status, i, bytes_left;

	bytes_left = i_size_read(inode) + 1;
	/* we can't trust i_blocks because we're actually going to
	 * write i_size + 1 bytes. */
	blocks = (bytes_left + sb->s_blocksize - 1) >> sb->s_blocksize_bits;

	mlog_entry("i_blocks = %llu, i_size = %llu, blocks = %d\n",
			(unsigned long long)inode->i_blocks,
			i_size_read(inode), blocks);

	/* Sanity check -- make sure we're going to fit. */
	if (bytes_left >
	    ocfs2_clusters_to_bytes(sb, OCFS2_I(inode)->ip_clusters)) {
		status = -EIO;
		mlog_errno(status);
		goto bail;
	}

	bhs = kcalloc(blocks, sizeof(struct buffer_head *), GFP_KERNEL);
	if (!bhs) {
		status = -ENOMEM;
		mlog_errno(status);
		goto bail;
	}

	status = ocfs2_extent_map_get_blocks(inode, 0, &p_blkno, &p_blocks,
					     NULL);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

	/* links can never be larger than one cluster so we know this
	 * is all going to be contiguous, but do a sanity check