xfs_vnodeops.c

linux 内核源代码

	if (!pathlen)
		goto out;

	if (ip->i_df.if_flags & XFS_IFINLINE) {
		memcpy(link, ip->i_df.if_u1.if_data, pathlen);
		link[pathlen] = '\0';
	} else {
		error = xfs_readlink_bmap(ip, link);
	}

 out:
	xfs_iunlock(ip, XFS_ILOCK_SHARED);
	return error;
}

/*
 * xfs_fsync
 *
 * This is called to sync the inode and its data out to disk.
 * We need to hold the I/O lock while flushing the data, and
 * the inode lock while flushing the inode.  The inode lock CANNOT
 * be held while flushing the data, so acquire after we're done
 * with that.
 */
int
xfs_fsync(
	xfs_inode_t	*ip,
	int		flag,
	xfs_off_t	start,
	xfs_off_t	stop)
{
	xfs_trans_t	*tp;
	int		error;
	int		log_flushed = 0, changed = 1;

	vn_trace_entry(ip, __FUNCTION__, (inst_t *)__return_address);

	ASSERT(start >= 0 && stop >= -1);

	if (XFS_FORCED_SHUTDOWN(ip->i_mount))
		return XFS_ERROR(EIO);

	if (flag & FSYNC_DATA)
		filemap_fdatawait(vn_to_inode(XFS_ITOV(ip))->i_mapping);

	/*
	 * We always need to make sure that the required inode state
	 * is safe on disk.  The vnode might be clean but because
	 * of committed transactions that haven't hit the disk yet.
	 * Likewise, there could be unflushed non-transactional
	 * changes to the inode core that have to go to disk.
	 *
	 * The following code depends on one assumption:  that
	 * any transaction that changes an inode logs the core
	 * because it has to change some field in the inode core
	 * (typically nextents or nblocks).  That assumption
	 * implies that any transactions against an inode will
	 * catch any non-transactional updates.  If inode-altering
	 * transactions exist that violate this assumption, the
	 * code breaks.  Right now, it figures that if the involved
	 * update_* field is clear and the inode is unpinned, the
	 * inode is clean.  Either it's been flushed or it's been
	 * committed and the commit has hit the disk unpinning the inode.
	 * (Note that xfs_inode_item_format() called at commit clears
	 * the update_* fields.)
	 */
	xfs_ilock(ip, XFS_ILOCK_SHARED);

	/* If we are flushing data then we care about update_size
	 * being set, otherwise we care about update_core
	 */
	if ((flag & FSYNC_DATA) ?
			(ip->i_update_size == 0) :
			(ip->i_update_core == 0)) {
		/*
		 * Timestamps/size haven't changed since last inode
		 * flush or inode transaction commit.  That means
		 * either nothing got written or a transaction
		 * committed which caught the updates.	If the
		 * latter happened and the transaction hasn't
		 * hit the disk yet, the inode will be still
		 * be pinned.  If it is, force the log.
		 */

		xfs_iunlock(ip, XFS_ILOCK_SHARED);

		if (xfs_ipincount(ip)) {
			_xfs_log_force(ip->i_mount, (xfs_lsn_t)0,
				      XFS_LOG_FORCE |
				      ((flag & FSYNC_WAIT)
				       ? XFS_LOG_SYNC : 0),
				      &log_flushed);
		} else {
			/*
			 * If the inode is not pinned and nothing
			 * has changed we don't need to flush the
			 * cache.
			 */
			changed = 0;
		}
		error = 0;
	} else	{
		/*
		 * Kick off a transaction to log the inode
		 * core to get the updates.  Make it
		 * sync if FSYNC_WAIT is passed in (which
		 * is done by everybody but specfs).  The
		 * sync transaction will also force the log.
		 */
		xfs_iunlock(ip, XFS_ILOCK_SHARED);
		tp = xfs_trans_alloc(ip->i_mount, XFS_TRANS_FSYNC_TS);
		if ((error = xfs_trans_reserve(tp, 0,
				XFS_FSYNC_TS_LOG_RES(ip->i_mount),
				0, 0, 0)))  {
			xfs_trans_cancel(tp, 0);
			return error;
		}
		xfs_ilock(ip, XFS_ILOCK_EXCL);

		/*
		 * Note - it's possible that we might have pushed
		 * ourselves out of the way during trans_reserve
		 * which would flush the inode.	 But there's no
		 * guarantee that the inode buffer has actually
		 * gone out yet (it's delwri).	Plus the buffer
		 * could be pinned anyway if it's part of an
		 * inode in another recent transaction.	 So we
		 * play it safe and fire off the transaction anyway.
		 */
		xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
		xfs_trans_ihold(tp, ip);
		xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
		if (flag & FSYNC_WAIT)
			xfs_trans_set_sync(tp);
		error = _xfs_trans_commit(tp, 0, &log_flushed);

		xfs_iunlock(ip, XFS_ILOCK_EXCL);
	}

	if ((ip->i_mount->m_flags & XFS_MOUNT_BARRIER) && changed) {
		/*
		 * If the log write didn't issue an ordered tag we need
		 * to flush the disk cache for the data device now.
		 */
		if (!log_flushed)
			xfs_blkdev_issue_flush(ip->i_mount->m_ddev_targp);

		/*
		 * If this inode is on the RT dev we need to flush that
		 * cache as well.
		 */
		if (ip->i_d.di_flags & XFS_DIFLAG_REALTIME)
			xfs_blkdev_issue_flush(ip->i_mount->m_rtdev_targp);
	}

	return error;
}

/*
 * This is called by xfs_inactive to free any blocks beyond eof
 * when the link count isn't zero and by xfs_dm_punch_hole() when
 * punching a hole to EOF.
 */
int
xfs_free_eofblocks(
	xfs_mount_t	*mp,
	xfs_inode_t	*ip,
	int		flags)
{
	xfs_trans_t	*tp;
	int		error;
	xfs_fileoff_t	end_fsb;
	xfs_fileoff_t	last_fsb;
	xfs_filblks_t	map_len;
	int		nimaps;
	xfs_bmbt_irec_t	imap;
	int		use_iolock = (flags & XFS_FREE_EOF_LOCK);

	/*
	 * Figure out if there are any blocks beyond the end
	 * of the file.  If not, then there is nothing to do.
	 */
	end_fsb = XFS_B_TO_FSB(mp, ((xfs_ufsize_t)ip->i_size));
	last_fsb = XFS_B_TO_FSB(mp, (xfs_ufsize_t)XFS_MAXIOFFSET(mp));
	map_len = last_fsb - end_fsb;
	if (map_len <= 0)
		return 0;

	nimaps = 1;
	xfs_ilock(ip, XFS_ILOCK_SHARED);
	error = XFS_BMAPI(mp, NULL, &ip->i_iocore, end_fsb, map_len, 0,
			  NULL, 0, &imap, &nimaps, NULL, NULL);
	xfs_iunlock(ip, XFS_ILOCK_SHARED);

	if (!error && (nimaps != 0) &&
	    (imap.br_startblock != HOLESTARTBLOCK ||
	     ip->i_delayed_blks)) {
		/*
		 * Attach the dquots to the inode up front.
		 */
		if ((error = XFS_QM_DQATTACH(mp, ip, 0)))
			return error;

		/*
		 * There are blocks after the end of file.
		 * Free them up now by truncating the file to
		 * its current size.
		 */
		tp = xfs_trans_alloc(mp, XFS_TRANS_INACTIVE);

		/*
		 * Do the xfs_itruncate_start() call before
		 * reserving any log space because
		 * itruncate_start will call into the buffer
		 * cache and we can't
		 * do that within a transaction.
		 */
		if (use_iolock)
			xfs_ilock(ip, XFS_IOLOCK_EXCL);
		error = xfs_itruncate_start(ip, XFS_ITRUNC_DEFINITE,
				    ip->i_size);
		if (error) {
			xfs_trans_cancel(tp, 0);
			if (use_iolock)
				xfs_iunlock(ip, XFS_IOLOCK_EXCL);
			return error;
		}

		error = xfs_trans_reserve(tp, 0,
					  XFS_ITRUNCATE_LOG_RES(mp),
					  0, XFS_TRANS_PERM_LOG_RES,
					  XFS_ITRUNCATE_LOG_COUNT);
		if (error) {
			ASSERT(XFS_FORCED_SHUTDOWN(mp));
			xfs_trans_cancel(tp, 0);
			xfs_iunlock(ip, XFS_IOLOCK_EXCL);
			return error;
		}

		xfs_ilock(ip, XFS_ILOCK_EXCL);
		xfs_trans_ijoin(tp, ip,
				XFS_IOLOCK_EXCL |
				XFS_ILOCK_EXCL);
		xfs_trans_ihold(tp, ip);

		error = xfs_itruncate_finish(&tp, ip,
					     ip->i_size,
					     XFS_DATA_FORK,
					     0);
		/*
		 * If we get an error at this point we
		 * simply don't bother truncating the file.
		 */
		if (error) {
			xfs_trans_cancel(tp,
					 (XFS_TRANS_RELEASE_LOG_RES |
					  XFS_TRANS_ABORT));
		} else {
			error = xfs_trans_commit(tp,
						XFS_TRANS_RELEASE_LOG_RES);
		}
		xfs_iunlock(ip, (use_iolock ? (XFS_IOLOCK_EXCL|XFS_ILOCK_EXCL)
					    : XFS_ILOCK_EXCL));
	}
	return error;
}

/*
 * Free a symlink that has blocks associated with it.
 */
STATIC int
xfs_inactive_symlink_rmt(
	xfs_inode_t	*ip,
	xfs_trans_t	**tpp)
{
	xfs_buf_t	*bp;
	int		committed;
	int		done;
	int		error;
	xfs_fsblock_t	first_block;
	xfs_bmap_free_t	free_list;
	int		i;
	xfs_mount_t	*mp;
	xfs_bmbt_irec_t	mval[SYMLINK_MAPS];
	int		nmaps;
	xfs_trans_t	*ntp;
	int		size;
	xfs_trans_t	*tp;

	tp = *tpp;
	mp = ip->i_mount;
	ASSERT(ip->i_d.di_size > XFS_IFORK_DSIZE(ip));
	/*
	 * We're freeing a symlink that has some
	 * blocks allocated to it.  Free the
	 * blocks here.  We know that we've got
	 * either 1 or 2 extents and that we can
	 * free them all in one bunmapi call.
	 */
	ASSERT(ip->i_d.di_nextents > 0 && ip->i_d.di_nextents <= 2);
	if ((error = xfs_trans_reserve(tp, 0, XFS_ITRUNCATE_LOG_RES(mp), 0,
			XFS_TRANS_PERM_LOG_RES, XFS_ITRUNCATE_LOG_COUNT))) {
		ASSERT(XFS_FORCED_SHUTDOWN(mp));
		xfs_trans_cancel(tp, 0);
		*tpp = NULL;
		return error;
	}
	/*
	 * Lock the inode, fix the size, and join it to the transaction.
	 * Hold it so in the normal path, we still have it locked for
	 * the second transaction.  In the error paths we need it
	 * held so the cancel won't rele it, see below.
	 */
	xfs_ilock(ip, XFS_IOLOCK_EXCL | XFS_ILOCK_EXCL);
	size = (int)ip->i_d.di_size;
	ip->i_d.di_size = 0;
	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
	xfs_trans_ihold(tp, ip);
	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
	/*
	 * Find the block(s) so we can inval and unmap them.
	 */
	done = 0;
	XFS_BMAP_INIT(&free_list, &first_block);
	nmaps = ARRAY_SIZE(mval);
	if ((error = xfs_bmapi(tp, ip, 0, XFS_B_TO_FSB(mp, size),
			XFS_BMAPI_METADATA, &first_block, 0, mval, &nmaps,
			&free_list, NULL)))
		goto error0;
	/*
	 * Invalidate the block(s).
	 */
	for (i = 0; i < nmaps; i++) {
		bp = xfs_trans_get_buf(tp, mp->m_ddev_targp,
			XFS_FSB_TO_DADDR(mp, mval[i].br_startblock),
			XFS_FSB_TO_BB(mp, mval[i].br_blockcount), 0);
		xfs_trans_binval(tp, bp);
	}
	/*
	 * Unmap the dead block(s) to the free_list.
	 */
	if ((error = xfs_bunmapi(tp, ip, 0, size, XFS_BMAPI_METADATA, nmaps,
			&first_block, &free_list, NULL, &done)))
		goto error1;
	ASSERT(done);
	/*
	 * Commit the first transaction.  This logs the EFI and the inode.
	 */
	if ((error = xfs_bmap_finish(&tp, &free_list, &committed)))
		goto error1;
	/*
	 * The transaction must have been committed, since there were
	 * actually extents freed by xfs_bunmapi.  See xfs_bmap_finish.
	 * The new tp has the extent freeing and EFDs.
	 */
	ASSERT(committed);
	/*
	 * The first xact was committed, so add the inode to the new one.
	 * Mark it dirty so it will be logged and moved forward in the log as
	 * part of every commit.
	 */
	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
	xfs_trans_ihold(tp, ip);
	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
	/*
	 * Get a new, empty transaction to return to our caller.
	 */
	ntp = xfs_trans_dup(tp);
	/*
	 * Commit the transaction containing extent freeing and EFDs.
	 * If we get an error on the commit here or on the reserve below,
	 * we need to unlock the inode since the new transaction doesn't
	 * have the inode attached.
	 */
	error = xfs_trans_commit(tp, 0);
	tp = ntp;
	if (error) {
		ASSERT(XFS_FORCED_SHUTDOWN(mp));
		goto error0;
	}
	/*
	 * Remove the memory for extent descriptions (just bookkeeping).
	 */
	if (ip->i_df.if_bytes)
		xfs_idata_realloc(ip, -ip->i_df.if_bytes, XFS_DATA_FORK);
	ASSERT(ip->i_df.if_bytes == 0);
	/*
	 * Put an itruncate log reservation in the new transaction
	 * for our caller.
	 */
	if ((error = xfs_trans_reserve(tp, 0, XFS_ITRUNCATE_LOG_RES(mp), 0,
			XFS_TRANS_PERM_LOG_RES, XFS_ITRUNCATE_LOG_COUNT))) {
		ASSERT(XFS_FORCED_SHUTDOWN(mp));
		goto error0;
	}
	/*
	 * Return with the inode locked but not joined to the transaction.
	 */
	*tpp = tp;
	return 0;

 error1:
	xfs_bmap_cancel(&free_list);
 error0:
	/*
	 * Have to come here with the inode locked and either
	 * (held and in the transaction) or (not in the transaction).
	 * If the inode isn't held then cancel would iput it, but
	 * that's wrong since this is inactive and the vnode ref
	 * count is 0 already.
	 * Cancel won't do anything to the inode if held, but it still
	 * needs to be locked until the cancel is done, if it was
	 * joined to the transaction.
	 */
	xfs_trans_cancel(tp, XFS_TRANS_RELEASE_LOG_RES | XFS_TRANS_ABORT);
	xfs_iunlock(ip, XFS_IOLOCK_EXCL | XFS_ILOCK_EXCL);
	*tpp = NULL;
	return error;

}

STATIC int
xfs_inactive_symlink_local(
	xfs_inode_t	*ip,
	xfs_trans_t	**tpp)
{
	int		error;

	ASSERT(ip->i_d.di_size <= XFS_IFORK_DSIZE(ip));
	/*
	 * We're freeing a symlink which fit into
	 * the inode.  Just free the memory used
	 * to hold the old symlink.
	 */
	error = xfs_trans_reserve(*tpp, 0,
				  XFS_ITRUNCATE_LOG_RES(ip->i_mount),
				  0, XFS_TRANS_PERM_LOG_RES,
				  XFS_ITRUNCATE_LOG_COUNT);

	if (error) {
		xfs_trans_cancel(*tpp, 0);
		*tpp = NULL;
		return error;
	}
	xfs_ilock(ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);

	/*
	 * Zero length symlinks _can_ exist.
	 */
	if (ip->i_df.if_bytes > 0) {
		xfs_idata_realloc(ip,
				  -(ip->i_df.if_bytes),
				  XFS_DATA_FORK);
		ASSERT(ip->i_df.if_bytes == 0);
	}
	return 0;
}

STATIC int
xfs_inactive_attrs(
	xfs_inode_t	*ip,
	xfs_trans_t	**tpp)
{
	xfs_trans_t	*tp;
	int		error;
	xfs_mount_t	*mp;

	ASSERT(ismrlocked(&ip->i_iolock, MR_UPDATE));
	tp = *tpp;
	mp = ip->i_mount;
	ASSERT(ip->i_d.di_forkoff != 0);
	xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES);
	xfs_iunlock(ip, XFS_ILOCK_EXCL);

	error = xfs_attr_inactive(ip);
	if (error) {
		*tpp = NULL;
		xfs_iunlock(ip, XFS_IOLOCK_EXCL);
		return error; /* goto out */
	}

	tp = xfs_trans_alloc(mp, XFS_TRANS_INACTIVE);
	error = xfs_trans_reserve(tp, 0,
				  XFS_IFREE_LOG_RES(mp),
				  0, XFS_TRANS_PERM_LOG_RES,
				  XFS_INACTIVE_LOG_COUNT);
	if (error) {
		ASSERT(XFS_FORCED_SHUTDOWN(mp));
		xfs_trans_cancel(tp, 0);
		*tpp = NULL;
		xfs_iunlock(ip, XFS_IOLOCK_EXCL);
		return error;
	}

	xfs_ilock(ip, XFS_ILOCK_EXCL);
	xfs_trans_ijoin(tp, ip, XFS_IOLOCK_EXCL | XFS_ILOCK_EXCL);
	xfs_trans_ihold(tp, ip);
	xfs_idestroy_fork(ip, XFS_ATTR_FORK);

	ASSERT(ip->i_d.di_anextents == 0);