[mirror_ubuntu-bionic-kernel.git] / fs / xfs / xfs_rw.c

/*
 * Copyright (c) 2000-2006 Silicon Graphics, Inc.
 * All Rights Reserved.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License as
 * published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it would be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write the Free Software Foundation,
 * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
 */
#include "xfs.h"
#include "xfs_fs.h"
#include "xfs_types.h"
#include "xfs_bit.h"
#include "xfs_log.h"
#include "xfs_inum.h"
#include "xfs_trans.h"
#include "xfs_sb.h"
#include "xfs_ag.h"
#include "xfs_dir2.h"
#include "xfs_dmapi.h"
#include "xfs_mount.h"
#include "xfs_bmap_btree.h"
#include "xfs_alloc_btree.h"
#include "xfs_ialloc_btree.h"
#include "xfs_dir2_sf.h"
#include "xfs_attr_sf.h"
#include "xfs_dinode.h"
#include "xfs_inode.h"
#include "xfs_inode_item.h"
#include "xfs_itable.h"
#include "xfs_btree.h"
#include "xfs_alloc.h"
#include "xfs_ialloc.h"
#include "xfs_attr.h"
#include "xfs_bmap.h"
#include "xfs_error.h"
#include "xfs_buf_item.h"
#include "xfs_rw.h"

/*
 * This is a subroutine for xfs_write() and other writers (xfs_ioctl)
 * which clears the setuid and setgid bits when a file is written.
 */
int
xfs_write_clear_setuid(
	xfs_inode_t	*ip)
{
	xfs_mount_t	*mp;
	xfs_trans_t	*tp;
	int		error;

	mp = ip->i_mount;
	tp = xfs_trans_alloc(mp, XFS_TRANS_WRITEID);
	if ((error = xfs_trans_reserve(tp, 0,
				      XFS_WRITEID_LOG_RES(mp),
				      0, 0, 0))) {
		xfs_trans_cancel(tp, 0);
		return error;
	}
	xfs_ilock(ip, XFS_ILOCK_EXCL);
	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
	xfs_trans_ihold(tp, ip);
	ip->i_d.di_mode &= ~S_ISUID;

	/*
	 * Note that we don't have to worry about mandatory
	 * file locking being disabled here because we only
	 * clear the S_ISGID bit if the Group execute bit is
	 * on, but if it was on then mandatory locking wouldn't
	 * have been enabled.
	 */
	if (ip->i_d.di_mode & S_IXGRP) {
		ip->i_d.di_mode &= ~S_ISGID;
	}
	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
	xfs_trans_set_sync(tp);
	error = xfs_trans_commit(tp, 0);
	xfs_iunlock(ip, XFS_ILOCK_EXCL);
	return 0;
}

/*
 * Handle logging requirements of various synchronous types of write.
 */
int
xfs_write_sync_logforce(
	xfs_mount_t	*mp,
	xfs_inode_t	*ip)
{
	int		error = 0;

	/*
	 * 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) &&
	    !(ip->i_update_size)) {
		xfs_inode_log_item_t	*iip = ip->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) {
			error = _xfs_log_force(mp, iip->ili_last_lsn,
					XFS_LOG_FORCE | XFS_LOG_SYNC, NULL);
		} else if (xfs_ipincount(ip) > 0) {
			error = _xfs_log_force(mp, (xfs_lsn_t)0,
					XFS_LOG_FORCE | XFS_LOG_SYNC, NULL);
		}

	} 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(ip, XFS_ILOCK_EXCL);
			xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
			xfs_trans_ihold(tp, ip);
			xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
			xfs_trans_set_sync(tp);
			error = xfs_trans_commit(tp, 0);
			xfs_iunlock(ip, XFS_ILOCK_EXCL);
		}
	}

	return error;
}

/*
 * Force a shutdown of the filesystem instantly while keeping
 * the filesystem consistent. We don't do an unmount here; just shutdown
 * the shop, make sure that absolutely nothing persistent happens to
 * this filesystem after this point.
 */
void
xfs_do_force_shutdown(
	xfs_mount_t	*mp,
	int		flags,
	char		*fname,
	int		lnnum)
{
	int		logerror;

	logerror = flags & SHUTDOWN_LOG_IO_ERROR;

	if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
		cmn_err(CE_NOTE, "xfs_force_shutdown(%s,0x%x) called from "
				 "line %d of file %s.  Return address = 0x%p",
			mp->m_fsname, flags, lnnum, fname, __return_address);
	}
	/*
	 * No need to duplicate efforts.
	 */
	if (XFS_FORCED_SHUTDOWN(mp) && !logerror)
		return;

	/*
	 * This flags XFS_MOUNT_FS_SHUTDOWN, makes sure that we don't
	 * queue up anybody new on the log reservations, and wakes up
	 * everybody who's sleeping on log reservations to tell them
	 * the bad news.
	 */
	if (xfs_log_force_umount(mp, logerror))
		return;

	if (flags & SHUTDOWN_CORRUPT_INCORE) {
		xfs_cmn_err(XFS_PTAG_SHUTDOWN_CORRUPT, CE_ALERT, mp,
    "Corruption of in-memory data detected.  Shutting down filesystem: %s",
			mp->m_fsname);
		if (XFS_ERRLEVEL_HIGH <= xfs_error_level) {
			xfs_stack_trace();
		}
	} else if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
		if (logerror) {
			xfs_cmn_err(XFS_PTAG_SHUTDOWN_LOGERROR, CE_ALERT, mp,
		"Log I/O Error Detected.  Shutting down filesystem: %s",
				mp->m_fsname);
		} else if (flags & SHUTDOWN_DEVICE_REQ) {
			xfs_cmn_err(XFS_PTAG_SHUTDOWN_IOERROR, CE_ALERT, mp,
		"All device paths lost.  Shutting down filesystem: %s",
				mp->m_fsname);
		} else if (!(flags & SHUTDOWN_REMOTE_REQ)) {
			xfs_cmn_err(XFS_PTAG_SHUTDOWN_IOERROR, CE_ALERT, mp,
		"I/O Error Detected.  Shutting down filesystem: %s",
				mp->m_fsname);
		}
	}
	if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
		cmn_err(CE_ALERT, "Please umount the filesystem, "
				  "and rectify the problem(s)");
	}
}


/*
 * Called when we want to stop a buffer from getting written or read.
 * We attach the EIO error, muck with its flags, and call biodone
 * so that the proper iodone callbacks get called.
 */
int
xfs_bioerror(
	xfs_buf_t *bp)
{

#ifdef XFSERRORDEBUG
	ASSERT(XFS_BUF_ISREAD(bp) || bp->b_iodone);
#endif

	/*
	 * No need to wait until the buffer is unpinned.
	 * We aren't flushing it.
	 */
	xfs_buftrace("XFS IOERROR", bp);
	XFS_BUF_ERROR(bp, EIO);
	/*
	 * We're calling biodone, so delete B_DONE flag. Either way
	 * we have to call the iodone callback, and calling biodone
	 * probably is the best way since it takes care of
	 * GRIO as well.
	 */
	XFS_BUF_UNREAD(bp);
	XFS_BUF_UNDELAYWRITE(bp);
	XFS_BUF_UNDONE(bp);
	XFS_BUF_STALE(bp);

	XFS_BUF_CLR_BDSTRAT_FUNC(bp);
	xfs_biodone(bp);

	return (EIO);
}

/*
 * Same as xfs_bioerror, except that we are releasing the buffer
 * here ourselves, and avoiding the biodone call.
 * This is meant for userdata errors; metadata bufs come with
 * iodone functions attached, so that we can track down errors.
 */
int
xfs_bioerror_relse(
	xfs_buf_t *bp)
{
	int64_t fl;

	ASSERT(XFS_BUF_IODONE_FUNC(bp) != xfs_buf_iodone_callbacks);
	ASSERT(XFS_BUF_IODONE_FUNC(bp) != xlog_iodone);

	xfs_buftrace("XFS IOERRELSE", bp);
	fl = XFS_BUF_BFLAGS(bp);
	/*
	 * No need to wait until the buffer is unpinned.
	 * We aren't flushing it.
	 *
	 * chunkhold expects B_DONE to be set, whether
	 * we actually finish the I/O or not. We don't want to
	 * change that interface.
	 */
	XFS_BUF_UNREAD(bp);
	XFS_BUF_UNDELAYWRITE(bp);
	XFS_BUF_DONE(bp);
	XFS_BUF_STALE(bp);
	XFS_BUF_CLR_IODONE_FUNC(bp);
	XFS_BUF_CLR_BDSTRAT_FUNC(bp);
	if (!(fl & XFS_B_ASYNC)) {
		/*
		 * Mark b_error and B_ERROR _both_.
		 * Lot's of chunkcache code assumes that.
		 * There's no reason to mark error for
		 * ASYNC buffers.
		 */
		XFS_BUF_ERROR(bp, EIO);
		XFS_BUF_FINISH_IOWAIT(bp);
	} else {
		xfs_buf_relse(bp);
	}
	return (EIO);
}

/*
 * Prints out an ALERT message about I/O error.
 */
void
xfs_ioerror_alert(
	char			*func,
	struct xfs_mount	*mp,
	xfs_buf_t		*bp,
	xfs_daddr_t		blkno)
{
	cmn_err(CE_ALERT,
 "I/O error in filesystem (\"%s\") meta-data dev %s block 0x%llx"
 "       (\"%s\") error %d buf count %zd",
		(!mp || !mp->m_fsname) ? "(fs name not set)" : mp->m_fsname,
		XFS_BUFTARG_NAME(XFS_BUF_TARGET(bp)),
		(__uint64_t)blkno, func,
		XFS_BUF_GETERROR(bp), XFS_BUF_COUNT(bp));
}

/*
 * This isn't an absolute requirement, but it is
 * just a good idea to call xfs_read_buf instead of
 * directly doing a read_buf call. For one, we shouldn't
 * be doing this disk read if we are in SHUTDOWN state anyway,
 * so this stops that from happening. Secondly, this does all
 * the error checking stuff and the brelse if appropriate for
 * the caller, so the code can be a little leaner.
 */

int
xfs_read_buf(
	struct xfs_mount *mp,
	xfs_buftarg_t	 *target,
	xfs_daddr_t	 blkno,
	int              len,
	uint             flags,
	xfs_buf_t	 **bpp)
{
	xfs_buf_t	 *bp;
	int		 error;

	if (flags)
		bp = xfs_buf_read_flags(target, blkno, len, flags);
	else
		bp = xfs_buf_read(target, blkno, len, flags);
	if (!bp)
		return XFS_ERROR(EIO);
	error = XFS_BUF_GETERROR(bp);
	if (bp && !error && !XFS_FORCED_SHUTDOWN(mp)) {
		*bpp = bp;
	} else {
		*bpp = NULL;
		if (error) {
			xfs_ioerror_alert("xfs_read_buf", mp, bp, XFS_BUF_ADDR(bp));
		} else {
			error = XFS_ERROR(EIO);
		}
		if (bp) {
			XFS_BUF_UNDONE(bp);
			XFS_BUF_UNDELAYWRITE(bp);
			XFS_BUF_STALE(bp);
			/*
			 * brelse clears B_ERROR and b_error
			 */
			xfs_buf_relse(bp);
		}
	}
	return (error);
}

/*
 * Wrapper around bwrite() so that we can trap
 * write errors, and act accordingly.
 */
int
xfs_bwrite(
	struct xfs_mount *mp,
	struct xfs_buf	 *bp)
{
	int	error;

	/*
	 * XXXsup how does this work for quotas.
	 */
	XFS_BUF_SET_BDSTRAT_FUNC(bp, xfs_bdstrat_cb);
	bp->b_mount = mp;
	XFS_BUF_WRITE(bp);

	if ((error = XFS_bwrite(bp))) {
		ASSERT(mp);
		/*
		 * Cannot put a buftrace here since if the buffer is not
		 * B_HOLD then we will brelse() the buffer before returning
		 * from bwrite and we could be tracing a buffer that has
		 * been reused.
		 */
		xfs_force_shutdown(mp, SHUTDOWN_META_IO_ERROR);
	}
	return (error);
}
Commit	Line	Data
1da177e4	1	/*
1e69dd0e	2	* Copyright (c) 2000-2006 Silicon Graphics, Inc.
7b718769	3	* All Rights Reserved.
1da177e4	4	*
7b718769 NS	5	* This program is free software; you can redistribute it and/or
7b718769 NS	6	* modify it under the terms of the GNU General Public License as
1da177e4 LT	7	* published by the Free Software Foundation.
1da177e4 LT	8	*
7b718769 NS	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.
1da177e4	13	*
7b718769 NS	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
1da177e4	17	*/
1da177e4	18	#include "xfs.h"
a844f451	19	#include "xfs_fs.h"
1da177e4	20	#include "xfs_types.h"
a844f451	21	#include "xfs_bit.h"
1da177e4	22	#include "xfs_log.h"
a844f451	23	#include "xfs_inum.h"
1da177e4 LT	24	#include "xfs_trans.h"
	25	#include "xfs_sb.h"
	26	#include "xfs_ag.h"
1da177e4 LT	27	#include "xfs_dir2.h"
	28	#include "xfs_dmapi.h"
	29	#include "xfs_mount.h"
1da177e4	30	#include "xfs_bmap_btree.h"
a844f451	31	#include "xfs_alloc_btree.h"
1da177e4	32	#include "xfs_ialloc_btree.h"
a844f451 NS	33	#include "xfs_dir2_sf.h"
	34	#include "xfs_attr_sf.h"
	35	#include "xfs_dinode.h"
	36	#include "xfs_inode.h"
	37	#include "xfs_inode_item.h"
1da177e4 LT	38	#include "xfs_itable.h"
	39	#include "xfs_btree.h"
	40	#include "xfs_alloc.h"
	41	#include "xfs_ialloc.h"
	42	#include "xfs_attr.h"
1da177e4	43	#include "xfs_bmap.h"
1da177e4 LT	44	#include "xfs_error.h"
	45	#include "xfs_buf_item.h"
	46	#include "xfs_rw.h"
	47
	48	/*
	49	* This is a subroutine for xfs_write() and other writers (xfs_ioctl)
	50	* which clears the setuid and setgid bits when a file is written.
	51	*/
	52	int
	53	xfs_write_clear_setuid(
	54	xfs_inode_t *ip)
	55	{
	56	xfs_mount_t *mp;
	57	xfs_trans_t *tp;
	58	int error;
	59
	60	mp = ip->i_mount;
	61	tp = xfs_trans_alloc(mp, XFS_TRANS_WRITEID);
	62	if ((error = xfs_trans_reserve(tp, 0,
	63	XFS_WRITEID_LOG_RES(mp),
	64	0, 0, 0))) {
	65	xfs_trans_cancel(tp, 0);
	66	return error;
	67	}
	68	xfs_ilock(ip, XFS_ILOCK_EXCL);
	69	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
	70	xfs_trans_ihold(tp, ip);
	71	ip->i_d.di_mode &= ~S_ISUID;
	72
	73	/*
	74	* Note that we don't have to worry about mandatory
	75	* file locking being disabled here because we only
	76	* clear the S_ISGID bit if the Group execute bit is
	77	* on, but if it was on then mandatory locking wouldn't
	78	* have been enabled.
	79	*/
	80	if (ip->i_d.di_mode & S_IXGRP) {
	81	ip->i_d.di_mode &= ~S_ISGID;
	82	}
	83	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
	84	xfs_trans_set_sync(tp);
1c72bf90	85	error = xfs_trans_commit(tp, 0);
1da177e4 LT	86	xfs_iunlock(ip, XFS_ILOCK_EXCL);
	87	return 0;
	88	}
	89
1e69dd0e NS	90	/*
	91	* Handle logging requirements of various synchronous types of write.
	92	*/
	93	int
	94	xfs_write_sync_logforce(
	95	xfs_mount_t *mp,
	96	xfs_inode_t *ip)
	97	{
	98	int error = 0;
	99
	100	/*
	101	* If we're treating this as O_DSYNC and we have not updated the
	102	* size, force the log.
	103	*/
	104	if (!(mp->m_flags & XFS_MOUNT_OSYNCISOSYNC) &&
	105	!(ip->i_update_size)) {
	106	xfs_inode_log_item_t *iip = ip->i_itemp;
	107
	108	/*
	109	* If an allocation transaction occurred
	110	* without extending the size, then we have to force
	111	* the log up the proper point to ensure that the
	112	* allocation is permanent. We can't count on
	113	* the fact that buffered writes lock out direct I/O
	114	* writes - the direct I/O write could have extended
	115	* the size nontransactionally, then finished before
	116	* we started. xfs_write_file will think that the file
	117	* didn't grow but the update isn't safe unless the
	118	* size change is logged.
	119	*
	120	* Force the log if we've committed a transaction
	121	* against the inode or if someone else has and
	122	* the commit record hasn't gone to disk (e.g.
	123	* the inode is pinned). This guarantees that
	124	* all changes affecting the inode are permanent
	125	* when we return.
	126	*/
	127	if (iip && iip->ili_last_lsn) {
b911ca04 DC	128	error = _xfs_log_force(mp, iip->ili_last_lsn,
b911ca04 DC	129	XFS_LOG_FORCE \| XFS_LOG_SYNC, NULL);
1e69dd0e	130	} else if (xfs_ipincount(ip) > 0) {
b911ca04 DC	131	error = _xfs_log_force(mp, (xfs_lsn_t)0,
b911ca04 DC	132	XFS_LOG_FORCE \| XFS_LOG_SYNC, NULL);
1e69dd0e NS	133	}
	134
	135	} else {
	136	xfs_trans_t *tp;
	137
	138	/*
	139	* O_SYNC or O_DSYNC _with_ a size update are handled
	140	* the same way.
	141	*
	142	* If the write was synchronous then we need to make
	143	* sure that the inode modification time is permanent.
	144	* We'll have updated the timestamp above, so here
	145	* we use a synchronous transaction to log the inode.
	146	* It's not fast, but it's necessary.
	147	*
	148	* If this a dsync write and the size got changed
	149	* non-transactionally, then we need to ensure that
	150	* the size change gets logged in a synchronous
	151	* transaction.
	152	*/
	153	tp = xfs_trans_alloc(mp, XFS_TRANS_WRITE_SYNC);
	154	if ((error = xfs_trans_reserve(tp, 0,
	155	XFS_SWRITE_LOG_RES(mp),
	156	0, 0, 0))) {
	157	/* Transaction reserve failed */
	158	xfs_trans_cancel(tp, 0);
	159	} else {
	160	/* Transaction reserve successful */
	161	xfs_ilock(ip, XFS_ILOCK_EXCL);
	162	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
	163	xfs_trans_ihold(tp, ip);
	164	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
	165	xfs_trans_set_sync(tp);
1c72bf90	166	error = xfs_trans_commit(tp, 0);
1e69dd0e NS	167	xfs_iunlock(ip, XFS_ILOCK_EXCL);
	168	}
	169	}
	170
	171	return error;
	172	}
	173
1da177e4 LT	174	/*
	175	* Force a shutdown of the filesystem instantly while keeping
	176	* the filesystem consistent. We don't do an unmount here; just shutdown
	177	* the shop, make sure that absolutely nothing persistent happens to
	178	* this filesystem after this point.
	179	*/
1da177e4 LT	180	void
1da177e4 LT	181	xfs_do_force_shutdown(
48c872a9	182	xfs_mount_t *mp,
1da177e4 LT	183	int flags,
	184	char *fname,
	185	int lnnum)
	186	{
	187	int logerror;
1da177e4	188
7d04a335	189	logerror = flags & SHUTDOWN_LOG_IO_ERROR;
1da177e4	190
7d04a335 NS	191	if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
	192	cmn_err(CE_NOTE, "xfs_force_shutdown(%s,0x%x) called from "
	193	"line %d of file %s. Return address = 0x%p",
	194	mp->m_fsname, flags, lnnum, fname, __return_address);
1da177e4 LT	195	}
	196	/*
	197	* No need to duplicate efforts.
	198	*/
	199	if (XFS_FORCED_SHUTDOWN(mp) && !logerror)
	200	return;
	201
	202	/*
	203	* This flags XFS_MOUNT_FS_SHUTDOWN, makes sure that we don't
	204	* queue up anybody new on the log reservations, and wakes up
7d04a335 NS	205	* everybody who's sleeping on log reservations to tell them
7d04a335 NS	206	* the bad news.
1da177e4 LT	207	*/
	208	if (xfs_log_force_umount(mp, logerror))
	209	return;
	210
7d04a335	211	if (flags & SHUTDOWN_CORRUPT_INCORE) {
1da177e4 LT	212	xfs_cmn_err(XFS_PTAG_SHUTDOWN_CORRUPT, CE_ALERT, mp,
	213	"Corruption of in-memory data detected. Shutting down filesystem: %s",
	214	mp->m_fsname);
	215	if (XFS_ERRLEVEL_HIGH <= xfs_error_level) {
	216	xfs_stack_trace();
	217	}
7d04a335	218	} else if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
1da177e4 LT	219	if (logerror) {
1da177e4 LT	220	xfs_cmn_err(XFS_PTAG_SHUTDOWN_LOGERROR, CE_ALERT, mp,
7d04a335	221	"Log I/O Error Detected. Shutting down filesystem: %s",
1da177e4	222	mp->m_fsname);
7d04a335	223	} else if (flags & SHUTDOWN_DEVICE_REQ) {
1da177e4	224	xfs_cmn_err(XFS_PTAG_SHUTDOWN_IOERROR, CE_ALERT, mp,
7d04a335 NS	225	"All device paths lost. Shutting down filesystem: %s",
	226	mp->m_fsname);
	227	} else if (!(flags & SHUTDOWN_REMOTE_REQ)) {
	228	xfs_cmn_err(XFS_PTAG_SHUTDOWN_IOERROR, CE_ALERT, mp,
	229	"I/O Error Detected. Shutting down filesystem: %s",
1da177e4 LT	230	mp->m_fsname);
	231	}
	232	}
7d04a335 NS	233	if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
	234	cmn_err(CE_ALERT, "Please umount the filesystem, "
	235	"and rectify the problem(s)");
1da177e4 LT	236	}
	237	}
	238
	239
	240	/*
	241	* Called when we want to stop a buffer from getting written or read.
	242	* We attach the EIO error, muck with its flags, and call biodone
	243	* so that the proper iodone callbacks get called.
	244	*/
	245	int
	246	xfs_bioerror(
	247	xfs_buf_t *bp)
	248	{
	249
	250	#ifdef XFSERRORDEBUG
	251	ASSERT(XFS_BUF_ISREAD(bp) \|\| bp->b_iodone);
	252	#endif
	253
	254	/*
	255	* No need to wait until the buffer is unpinned.
	256	* We aren't flushing it.
	257	*/
	258	xfs_buftrace("XFS IOERROR", bp);
	259	XFS_BUF_ERROR(bp, EIO);
	260	/*
	261	* We're calling biodone, so delete B_DONE flag. Either way
	262	* we have to call the iodone callback, and calling biodone
	263	* probably is the best way since it takes care of
	264	* GRIO as well.
	265	*/
	266	XFS_BUF_UNREAD(bp);
	267	XFS_BUF_UNDELAYWRITE(bp);
	268	XFS_BUF_UNDONE(bp);
	269	XFS_BUF_STALE(bp);
	270
	271	XFS_BUF_CLR_BDSTRAT_FUNC(bp);
	272	xfs_biodone(bp);
	273
	274	return (EIO);
	275	}
	276
	277	/*
	278	* Same as xfs_bioerror, except that we are releasing the buffer
	279	* here ourselves, and avoiding the biodone call.
	280	* This is meant for userdata errors; metadata bufs come with
	281	* iodone functions attached, so that we can track down errors.
	282	*/
	283	int
	284	xfs_bioerror_relse(
	285	xfs_buf_t *bp)
	286	{
	287	int64_t fl;
	288
	289	ASSERT(XFS_BUF_IODONE_FUNC(bp) != xfs_buf_iodone_callbacks);
	290	ASSERT(XFS_BUF_IODONE_FUNC(bp) != xlog_iodone);
	291
	292	xfs_buftrace("XFS IOERRELSE", bp);
	293	fl = XFS_BUF_BFLAGS(bp);
	294	/*
	295	* No need to wait until the buffer is unpinned.
	296	* We aren't flushing it.
	297	*
	298	* chunkhold expects B_DONE to be set, whether
	299	* we actually finish the I/O or not. We don't want to
300	* change that interface.
301	*/
302	XFS_BUF_UNREAD(bp);
303	XFS_BUF_UNDELAYWRITE(bp);
304	XFS_BUF_DONE(bp);
305	XFS_BUF_STALE(bp);
306	XFS_BUF_CLR_IODONE_FUNC(bp);
307	XFS_BUF_CLR_BDSTRAT_FUNC(bp);
308	if (!(fl & XFS_B_ASYNC)) {
309	/*
310	* Mark b_error and B_ERROR _both_.
311	* Lot's of chunkcache code assumes that.
312	* There's no reason to mark error for
313	* ASYNC buffers.
314	*/
315	XFS_BUF_ERROR(bp, EIO);
b4dd330b	316	XFS_BUF_FINISH_IOWAIT(bp);
1da177e4 LT	317	} else {
	318	xfs_buf_relse(bp);
	319	}
	320	return (EIO);
	321	}
ce8e922c	322
1da177e4 LT	323	/*
	324	* Prints out an ALERT message about I/O error.
	325	*/
	326	void
	327	xfs_ioerror_alert(
	328	char *func,
	329	struct xfs_mount *mp,
	330	xfs_buf_t *bp,
	331	xfs_daddr_t blkno)
	332	{
	333	cmn_err(CE_ALERT,
	334	"I/O error in filesystem (\"%s\") meta-data dev %s block 0x%llx"
da1650a5	335	" (\"%s\") error %d buf count %zd",
1da177e4	336	(!mp \|\| !mp->m_fsname) ? "(fs name not set)" : mp->m_fsname,
ce8e922c NS	337	XFS_BUFTARG_NAME(XFS_BUF_TARGET(bp)),
	338	(__uint64_t)blkno, func,
	339	XFS_BUF_GETERROR(bp), XFS_BUF_COUNT(bp));
1da177e4 LT	340	}
	341
	342	/*
	343	* This isn't an absolute requirement, but it is
	344	* just a good idea to call xfs_read_buf instead of
	345	* directly doing a read_buf call. For one, we shouldn't
	346	* be doing this disk read if we are in SHUTDOWN state anyway,
	347	* so this stops that from happening. Secondly, this does all
	348	* the error checking stuff and the brelse if appropriate for
	349	* the caller, so the code can be a little leaner.
	350	*/
	351
	352	int
	353	xfs_read_buf(
	354	struct xfs_mount *mp,
	355	xfs_buftarg_t *target,
	356	xfs_daddr_t blkno,
	357	int len,
	358	uint flags,
	359	xfs_buf_t **bpp)
	360	{
	361	xfs_buf_t *bp;
	362	int error;
	363
	364	if (flags)
	365	bp = xfs_buf_read_flags(target, blkno, len, flags);
	366	else
	367	bp = xfs_buf_read(target, blkno, len, flags);
	368	if (!bp)
	369	return XFS_ERROR(EIO);
	370	error = XFS_BUF_GETERROR(bp);
	371	if (bp && !error && !XFS_FORCED_SHUTDOWN(mp)) {
	372	*bpp = bp;
	373	} else {
	374	*bpp = NULL;
	375	if (error) {
	376	xfs_ioerror_alert("xfs_read_buf", mp, bp, XFS_BUF_ADDR(bp));
	377	} else {
	378	error = XFS_ERROR(EIO);
	379	}
	380	if (bp) {
	381	XFS_BUF_UNDONE(bp);
	382	XFS_BUF_UNDELAYWRITE(bp);
	383	XFS_BUF_STALE(bp);
	384	/*
	385	* brelse clears B_ERROR and b_error
	386	*/
	387	xfs_buf_relse(bp);
	388	}
	389	}
	390	return (error);
	391	}
	392
	393	/*
	394	* Wrapper around bwrite() so that we can trap
	395	* write errors, and act accordingly.
	396	*/
	397	int
	398	xfs_bwrite(
	399	struct xfs_mount *mp,
	400	struct xfs_buf *bp)
	401	{
	402	int error;
	403
404	/*
405	* XXXsup how does this work for quotas.
406	*/
407	XFS_BUF_SET_BDSTRAT_FUNC(bp, xfs_bdstrat_cb);
15ac08a8	408	bp->b_mount = mp;
1da177e4 LT	409	XFS_BUF_WRITE(bp);
	410
	411	if ((error = XFS_bwrite(bp))) {
	412	ASSERT(mp);
	413	/*
	414	* Cannot put a buftrace here since if the buffer is not
	415	* B_HOLD then we will brelse() the buffer before returning
	416	* from bwrite and we could be tracing a buffer that has
	417	* been reused.
	418	*/
7d04a335	419	xfs_force_shutdown(mp, SHUTDOWN_META_IO_ERROR);
1da177e4 LT	420	}
	421	return (error);
	422	}