xfs_lrw.c

linux-2.6.15.6

			 */
			prev_zero_fsb = NULLFILEOFF;
			prev_zero_count = 0;
			start_zero_fsb = imap.br_startoff +
					 imap.br_blockcount;
			ASSERT(start_zero_fsb <= (end_zero_fsb + 1));
			continue;
		}

		/*
		 * There are blocks in the range requested.
		 * Zero them a single write at a time.  We actually
		 * don't zero the entire range returned if it is
		 * too big and simply loop around to get the rest.
		 * That is not the most efficient thing to do, but it
		 * is simple and this path should not be exercised often.
		 */
		buf_len_fsb = XFS_FILBLKS_MIN(imap.br_blockcount,
					      mp->m_writeio_blocks << 8);
		/*
		 * Drop the inode lock while we're doing the I/O.
		 * We'll still have the iolock to protect us.
		 */
		XFS_IUNLOCK(mp, io, XFS_ILOCK_EXCL|XFS_EXTSIZE_RD);

		loff = XFS_FSB_TO_B(mp, start_zero_fsb);
		lsize = XFS_FSB_TO_B(mp, buf_len_fsb);

		error = xfs_iozero(ip, loff, lsize, end_size);

		if (error) {
			goto out_lock;
		}

		prev_zero_fsb = start_zero_fsb;
		prev_zero_count = buf_len_fsb;
		start_zero_fsb = imap.br_startoff + buf_len_fsb;
		ASSERT(start_zero_fsb <= (end_zero_fsb + 1));

		XFS_ILOCK(mp, io, XFS_ILOCK_EXCL|XFS_EXTSIZE_RD);
	}

	return 0;

out_lock:

	XFS_ILOCK(mp, io, XFS_ILOCK_EXCL|XFS_EXTSIZE_RD);
	ASSERT(error >= 0);
	return error;
}

ssize_t				/* bytes written, or (-) error */
xfs_write(
	bhv_desc_t		*bdp,
	struct kiocb		*iocb,
	const struct iovec	*iovp,
	unsigned int		nsegs,
	loff_t			*offset,
	int			ioflags,
	cred_t			*credp)
{
	struct file		*file = iocb->ki_filp;
	struct address_space	*mapping = file->f_mapping;
	struct inode		*inode = mapping->host;
	unsigned long		segs = nsegs;
	xfs_inode_t		*xip;
	xfs_mount_t		*mp;
	ssize_t			ret = 0, error = 0;
	xfs_fsize_t		isize, new_size;
	xfs_iocore_t		*io;
	vnode_t			*vp;
	unsigned long		seg;
	int			iolock;
	int			eventsent = 0;
	vrwlock_t		locktype;
	size_t			ocount = 0, count;
	loff_t			pos;
	int			need_isem = 1, need_flush = 0;

	XFS_STATS_INC(xs_write_calls);

	vp = BHV_TO_VNODE(bdp);
	xip = XFS_BHVTOI(bdp);

	for (seg = 0; seg < segs; seg++) {
		const struct iovec *iv = &iovp[seg];

		/*
		 * If any segment has a negative length, or the cumulative
		 * length ever wraps negative then return -EINVAL.
		 */
		ocount += iv->iov_len;
		if (unlikely((ssize_t)(ocount|iv->iov_len) < 0))
			return -EINVAL;
		if (access_ok(VERIFY_READ, iv->iov_base, iv->iov_len))
			continue;
		if (seg == 0)
			return -EFAULT;
		segs = seg;
		ocount -= iv->iov_len;  /* This segment is no good */
		break;
	}

	count = ocount;
	pos = *offset;

	if (count == 0)
		return 0;

	io = &xip->i_iocore;
	mp = io->io_mount;

	if (XFS_FORCED_SHUTDOWN(mp))
		return -EIO;

	fs_check_frozen(vp->v_vfsp, SB_FREEZE_WRITE);

	if (ioflags & IO_ISDIRECT) {
		xfs_buftarg_t	*target =
			(xip->i_d.di_flags & XFS_DIFLAG_REALTIME) ?
				mp->m_rtdev_targp : mp->m_ddev_targp;

		if ((pos & target->pbr_smask) || (count & target->pbr_smask))
			return XFS_ERROR(-EINVAL);

		if (!VN_CACHED(vp) && pos < i_size_read(inode))
			need_isem = 0;

		if (VN_CACHED(vp))
			need_flush = 1;
	}

relock:
	if (need_isem) {
		iolock = XFS_IOLOCK_EXCL;
		locktype = VRWLOCK_WRITE;

		down(&inode->i_sem);
	} else {
		iolock = XFS_IOLOCK_SHARED;
		locktype = VRWLOCK_WRITE_DIRECT;
	}

	xfs_ilock(xip, XFS_ILOCK_EXCL|iolock);

	isize = i_size_read(inode);

	if (file->f_flags & O_APPEND)
		*offset = isize;

start:
	error = -generic_write_checks(file, &pos, &count,
					S_ISBLK(inode->i_mode));
	if (error) {
		xfs_iunlock(xip, XFS_ILOCK_EXCL|iolock);
		goto out_unlock_isem;
	}

	new_size = pos + count;
	if (new_size > isize)
		io->io_new_size = new_size;

	if ((DM_EVENT_ENABLED(vp->v_vfsp, xip, DM_EVENT_WRITE) &&
	    !(ioflags & IO_INVIS) && !eventsent)) {
		loff_t		savedsize = pos;
		int		dmflags = FILP_DELAY_FLAG(file);

		if (need_isem)
			dmflags |= DM_FLAGS_ISEM;

		xfs_iunlock(xip, XFS_ILOCK_EXCL);
		error = XFS_SEND_DATA(xip->i_mount, DM_EVENT_WRITE, vp,
				      pos, count,
				      dmflags, &locktype);
		if (error) {
			xfs_iunlock(xip, iolock);
			goto out_unlock_isem;
		}
		xfs_ilock(xip, XFS_ILOCK_EXCL);
		eventsent = 1;

		/*
		 * The iolock was dropped and reaquired in XFS_SEND_DATA
		 * so we have to recheck the size when appending.
		 * We will only "goto start;" once, since having sent the
		 * event prevents another call to XFS_SEND_DATA, which is
		 * what allows the size to change in the first place.
		 */
		if ((file->f_flags & O_APPEND) && savedsize != isize) {
			pos = isize = xip->i_d.di_size;
			goto start;
		}
	}

	if (likely(!(ioflags & IO_INVIS))) {
		inode_update_time(inode, 1);
		xfs_ichgtime_fast(xip, inode,
				  XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
	}

	/*
	 * If the offset is beyond the size of the file, we have a couple
	 * of things to do. First, if there is already space allocated
	 * we need to either create holes or zero the disk or ...
	 *
	 * If there is a page where the previous size lands, we need
	 * to zero it out up to the new size.
	 */

	if (pos > isize) {
		error = xfs_zero_eof(BHV_TO_VNODE(bdp), io, pos,
					isize, pos + count);
		if (error) {
			xfs_iunlock(xip, XFS_ILOCK_EXCL|iolock);
			goto out_unlock_isem;
		}
	}
	xfs_iunlock(xip, XFS_ILOCK_EXCL);

	/*
	 * If we're writing the file then make sure to clear the
	 * setuid and setgid bits if the process is not being run
	 * by root.  This keeps people from modifying setuid and
	 * setgid binaries.
	 */

	if (((xip->i_d.di_mode & S_ISUID) ||
	    ((xip->i_d.di_mode & (S_ISGID | S_IXGRP)) ==
		(S_ISGID | S_IXGRP))) &&
	     !capable(CAP_FSETID)) {
		error = xfs_write_clear_setuid(xip);
		if (likely(!error))
			error = -remove_suid(file->f_dentry);
		if (unlikely(error)) {
			xfs_iunlock(xip, iolock);
			goto out_unlock_isem;
		}
	}

retry:
	/* We can write back this queue in page reclaim */
	current->backing_dev_info = mapping->backing_dev_info;

	if ((ioflags & IO_ISDIRECT)) {
		if (need_flush) {
			xfs_inval_cached_trace(io, pos, -1,
					ctooff(offtoct(pos)), -1);
			VOP_FLUSHINVAL_PAGES(vp, ctooff(offtoct(pos)),
					-1, FI_REMAPF_LOCKED);
		}

		if (need_isem) {
			/* demote the lock now the cached pages are gone */
			XFS_ILOCK_DEMOTE(mp, io, XFS_IOLOCK_EXCL);
			up(&inode->i_sem);

			iolock = XFS_IOLOCK_SHARED;
			locktype = VRWLOCK_WRITE_DIRECT;
			need_isem = 0;
		}

 		xfs_rw_enter_trace(XFS_DIOWR_ENTER, io, (void *)iovp, segs,
				*offset, ioflags);
		ret = generic_file_direct_write(iocb, iovp,
				&segs, pos, offset, count, ocount);

		/*
		 * direct-io write to a hole: fall through to buffered I/O
		 * for completing the rest of the request.
		 */
		if (ret >= 0 && ret != count) {
			XFS_STATS_ADD(xs_write_bytes, ret);

			pos += ret;
			count -= ret;

			need_isem = 1;
			ioflags &= ~IO_ISDIRECT;
			xfs_iunlock(xip, iolock);
			goto relock;
		}
	} else {
		xfs_rw_enter_trace(XFS_WRITE_ENTER, io, (void *)iovp, segs,
				*offset, ioflags);
		ret = generic_file_buffered_write(iocb, iovp, segs,
				pos, offset, count, ret);
	}

	current->backing_dev_info = NULL;

	if (ret == -EIOCBQUEUED && !(ioflags & IO_ISAIO))
		ret = wait_on_sync_kiocb(iocb);

	if ((ret == -ENOSPC) &&
	    DM_EVENT_ENABLED(vp->v_vfsp, xip, DM_EVENT_NOSPACE) &&
	    !(ioflags & IO_INVIS)) {

		xfs_rwunlock(bdp, locktype);
		if (need_isem)
			up(&inode->i_sem);
		error = XFS_SEND_NAMESP(xip->i_mount, DM_EVENT_NOSPACE, vp,
				DM_RIGHT_NULL, vp, DM_RIGHT_NULL, NULL, NULL,
				0, 0, 0); /* Delay flag intentionally  unused */
		if (error)
			goto out_nounlocks;
		if (need_isem)
			down(&inode->i_sem);
		xfs_rwlock(bdp, locktype);
		pos = xip->i_d.di_size;
		ret = 0;
		goto retry;
	}

	if (*offset > xip->i_d.di_size) {
		xfs_ilock(xip, XFS_ILOCK_EXCL);
		if (*offset > xip->i_d.di_size) {
			xip->i_d.di_size = *offset;
			i_size_write(inode, *offset);
			xip->i_update_core = 1;
			xip->i_update_size = 1;
		}
		xfs_iunlock(xip, XFS_ILOCK_EXCL);
	}

	error = -ret;
	if (ret <= 0)
		goto out_unlock_internal;

	XFS_STATS_ADD(xs_write_bytes, ret);

	/* Handle various SYNC-type writes */
	if ((file->f_flags & O_SYNC) || IS_SYNC(inode)) {
		/*
		 * If we're treating this as O_DSYNC and we have not updated the
		 * size, force the log.
		 */
		if (!(mp->m_flags & XFS_MOUNT_OSYNCISOSYNC) &&
		    !(xip->i_update_size)) {
			xfs_inode_log_item_t	*iip = xip->i_itemp;

			/*
			 * If an allocation transaction occurred
			 * without extending the size, then we have to force
			 * the log up the proper point to ensure that the
			 * allocation is permanent.  We can't count on
			 * the fact that buffered writes lock out direct I/O
			 * writes - the direct I/O write could have extended
			 * the size nontransactionally, then finished before
			 * we started.  xfs_write_file will think that the file
			 * didn't grow but the update isn't safe unless the
			 * size change is logged.
			 *
			 * Force the log if we've committed a transaction
			 * against the inode or if someone else has and
			 * the commit record hasn't gone to disk (e.g.
			 * the inode is pinned).  This guarantees that
			 * all changes affecting the inode are permanent
			 * when we return.
			 */
			if (iip && iip->ili_last_lsn) {
				xfs_log_force(mp, iip->ili_last_lsn,
						XFS_LOG_FORCE | XFS_LOG_SYNC);
			} else if (xfs_ipincount(xip) > 0) {
				xfs_log_force(mp, (xfs_lsn_t)0,
						XFS_LOG_FORCE | XFS_LOG_SYNC);
			}

		} else {
			xfs_trans_t	*tp;

			/*
			 * O_SYNC or O_DSYNC _with_ a size update are handled
			 * the same way.
			 *
			 * If the write was synchronous then we need to make
			 * sure that the inode modification time is permanent.
			 * We'll have updated the timestamp above, so here
			 * we use a synchronous transaction to log the inode.
			 * It's not fast, but it's necessary.
			 *
			 * If this a dsync write and the size got changed
			 * non-transactionally, then we need to ensure that
			 * the size change gets logged in a synchronous
			 * transaction.
			 */

			tp = xfs_trans_alloc(mp, XFS_TRANS_WRITE_SYNC);
			if ((error = xfs_trans_reserve(tp, 0,
						      XFS_SWRITE_LOG_RES(mp),
						      0, 0, 0))) {
				/* Transaction reserve failed */
				xfs_trans_cancel(tp, 0);
			} else {
				/* Transaction reserve successful */
				xfs_ilock(xip, XFS_ILOCK_EXCL);
				xfs_trans_ijoin(tp, xip, XFS_ILOCK_EXCL);
				xfs_trans_ihold(tp, xip);
				xfs_trans_log_inode(tp, xip, XFS_ILOG_CORE);
				xfs_trans_set_sync(tp);
				error = xfs_trans_commit(tp, 0, NULL);
				xfs_iunlock(xip, XFS_ILOCK_EXCL);
			}
			if (error)
				goto out_unlock_internal;
		}
	
		xfs_rwunlock(bdp, locktype);
		if (need_isem)
			up(&inode->i_sem);

		error = sync_page_range(inode, mapping, pos, ret);
		if (!error)
			error = ret;
		return error;
	}

 out_unlock_internal:
	xfs_rwunlock(bdp, locktype);
 out_unlock_isem:
	if (need_isem)
		up(&inode->i_sem);
 out_nounlocks:
	return -error;
}

/*
 * All xfs metadata buffers except log state machine buffers
 * get this attached as their b_bdstrat callback function.
 * This is so that we can catch a buffer
 * after prematurely unpinning it to forcibly shutdown the filesystem.
 */
int
xfs_bdstrat_cb(struct xfs_buf *bp)
{
	xfs_mount_t	*mp;

	mp = XFS_BUF_FSPRIVATE3(bp, xfs_mount_t *);
	if (!XFS_FORCED_SHUTDOWN(mp)) {
		pagebuf_iorequest(bp);
		return 0;
	} else {
		xfs_buftrace("XFS__BDSTRAT IOERROR", bp);
		/*
		 * Metadata write that didn't get logged but
		 * written delayed anyway. These aren't associated
		 * with a transaction, and can be ignored.
		 */
		if (XFS_BUF_IODONE_FUNC(bp) == NULL &&
		    (XFS_BUF_ISREAD(bp)) == 0)
			return (xfs_bioerror_relse(bp));
		else
			return (xfs_bioerror(bp));
	}
}


int
xfs_bmap(bhv_desc_t	*bdp,
	xfs_off_t	offset,
	ssize_t		count,
	int		flags,
	xfs_iomap_t	*iomapp,
	int		*niomaps)
{
	xfs_inode_t	*ip = XFS_BHVTOI(bdp);
	xfs_iocore_t	*io = &ip->i_iocore;

	ASSERT((ip->i_d.di_mode & S_IFMT) == S_IFREG);
	ASSERT(((ip->i_d.di_flags & XFS_DIFLAG_REALTIME) != 0) ==
	       ((ip->i_iocore.io_flags & XFS_IOCORE_RT) != 0));

	return xfs_iomap(io, offset, count, flags, iomapp, niomaps);
}

/*
 * Wrapper around bdstrat so that we can stop data
 * from going to disk in case we are shutting down the filesystem.
 * Typically user data goes thru this path; one of the exceptions
 * is the superblock.
 */
int
xfsbdstrat(
	struct xfs_mount	*mp,
	struct xfs_buf		*bp)
{
	ASSERT(mp);
	if (!XFS_FORCED_SHUTDOWN(mp)) {
		/* Grio redirection would go here
		 * if (XFS_BUF_IS_GRIO(bp)) {
		 */

		pagebuf_iorequest(bp);
		return 0;
	}

	xfs_buftrace("XFSBDSTRAT IOERROR", bp);
	return (xfs_bioerror_relse(bp));
}

/*
 * If the underlying (data/log/rt) device is readonly, there are some
 * operations that cannot proceed.
 */
int
xfs_dev_is_read_only(
	xfs_mount_t		*mp,
	char			*message)
{
	if (xfs_readonly_buftarg(mp->m_ddev_targp) ||
	    xfs_readonly_buftarg(mp->m_logdev_targp) ||
	    (mp->m_rtdev_targp && xfs_readonly_buftarg(mp->m_rtdev_targp))) {
		cmn_err(CE_NOTE,
			"XFS: %s required on read-only device.", message);
		cmn_err(CE_NOTE,
			"XFS: write access unavailable, cannot proceed.");
		return EROFS;
	}
	return 0;
}