[mirror_zfs.git] / module / zfs / zfs_sa.c

/*
 * CDDL HEADER START
 *
 * The contents of this file are subject to the terms of the
 * Common Development and Distribution License (the "License").
 * You may not use this file except in compliance with the License.
 *
 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
 * or http://www.opensolaris.org/os/licensing.
 * See the License for the specific language governing permissions
 * and limitations under the License.
 *
 * When distributing Covered Code, include this CDDL HEADER in each
 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
 * If applicable, add the following below this CDDL HEADER, with the
 * fields enclosed by brackets "[]" replaced with your own identifying
 * information: Portions Copyright [yyyy] [name of copyright owner]
 *
 * CDDL HEADER END
 */
/*
 * Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved.
 */

#include <sys/types.h>
#include <sys/param.h>
#include <sys/vnode.h>
#include <sys/sa.h>
#include <sys/zfs_acl.h>
#include <sys/zfs_sa.h>

/*
 * ZPL attribute registration table.
 * Order of attributes doesn't matter
 * a unique value will be assigned for each
 * attribute that is file system specific
 *
 * This is just the set of ZPL attributes that this
 * version of ZFS deals with natively.  The file system
 * could have other attributes stored in files, but they will be
 * ignored.  The SA framework will preserve them, just that
 * this version of ZFS won't change or delete them.
 */

sa_attr_reg_t zfs_attr_table[ZPL_END+1] = {
	{"ZPL_ATIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 0},
	{"ZPL_MTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 1},
	{"ZPL_CTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 2},
	{"ZPL_CRTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 3},
	{"ZPL_GEN", sizeof (uint64_t), SA_UINT64_ARRAY, 4},
	{"ZPL_MODE", sizeof (uint64_t), SA_UINT64_ARRAY, 5},
	{"ZPL_SIZE", sizeof (uint64_t), SA_UINT64_ARRAY, 6},
	{"ZPL_PARENT", sizeof (uint64_t), SA_UINT64_ARRAY, 7},
	{"ZPL_LINKS", sizeof (uint64_t), SA_UINT64_ARRAY, 8},
	{"ZPL_XATTR", sizeof (uint64_t), SA_UINT64_ARRAY, 9},
	{"ZPL_RDEV", sizeof (uint64_t), SA_UINT64_ARRAY, 10},
	{"ZPL_FLAGS", sizeof (uint64_t), SA_UINT64_ARRAY, 11},
	{"ZPL_UID", sizeof (uint64_t), SA_UINT64_ARRAY, 12},
	{"ZPL_GID", sizeof (uint64_t), SA_UINT64_ARRAY, 13},
	{"ZPL_PAD", sizeof (uint64_t) * 4, SA_UINT64_ARRAY, 14},
	{"ZPL_ZNODE_ACL", 88, SA_UINT8_ARRAY, 15},
	{"ZPL_DACL_COUNT", sizeof (uint64_t), SA_UINT64_ARRAY, 0},
	{"ZPL_SYMLINK", 0, SA_UINT8_ARRAY, 0},
	{"ZPL_SCANSTAMP", 32, SA_UINT8_ARRAY, 0},
	{"ZPL_DACL_ACES", 0, SA_ACL, 0},
	{NULL, 0, 0, 0}
};

#ifdef _KERNEL
int
zfs_sa_readlink(znode_t *zp, uio_t *uio)
{
	dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
	size_t bufsz;
	int error;

	bufsz = zp->z_size;
	if (bufsz + ZFS_OLD_ZNODE_PHYS_SIZE <= db->db_size) {
		error = uiomove((caddr_t)db->db_data +
		    ZFS_OLD_ZNODE_PHYS_SIZE,
		    MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio);
	} else {
		dmu_buf_t *dbp;
		if ((error = dmu_buf_hold(ZTOZSB(zp)->z_os, zp->z_id,
		    0, FTAG, &dbp, DMU_READ_NO_PREFETCH)) == 0) {
			error = uiomove(dbp->db_data,
			    MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio);
			dmu_buf_rele(dbp, FTAG);
		}
	}
	return (error);
}

void
zfs_sa_symlink(znode_t *zp, char *link, int len, dmu_tx_t *tx)
{
	dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);

	if (ZFS_OLD_ZNODE_PHYS_SIZE + len <= dmu_bonus_max()) {
		VERIFY(dmu_set_bonus(db,
		    len + ZFS_OLD_ZNODE_PHYS_SIZE, tx) == 0);
		if (len) {
			bcopy(link, (caddr_t)db->db_data +
			    ZFS_OLD_ZNODE_PHYS_SIZE, len);
		}
	} else {
		dmu_buf_t *dbp;

		zfs_grow_blocksize(zp, len, tx);
		VERIFY(0 == dmu_buf_hold(ZTOZSB(zp)->z_os,
		    zp->z_id, 0, FTAG, &dbp, DMU_READ_NO_PREFETCH));

		dmu_buf_will_dirty(dbp, tx);

		ASSERT3U(len, <=, dbp->db_size);
		bcopy(link, dbp->db_data, len);
		dmu_buf_rele(dbp, FTAG);
	}
}

void
zfs_sa_get_scanstamp(znode_t *zp, xvattr_t *xvap)
{
	zfs_sb_t *zsb = ZTOZSB(zp);
	xoptattr_t *xoap;

	ASSERT(MUTEX_HELD(&zp->z_lock));
	VERIFY((xoap = xva_getxoptattr(xvap)) != NULL);
	if (zp->z_is_sa) {
		if (sa_lookup(zp->z_sa_hdl, SA_ZPL_SCANSTAMP(zsb),
		    &xoap->xoa_av_scanstamp,
		    sizeof (xoap->xoa_av_scanstamp)) != 0)
			return;
	} else {
		dmu_object_info_t doi;
		dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
		int len;

		if (!(zp->z_pflags & ZFS_BONUS_SCANSTAMP))
			return;

		sa_object_info(zp->z_sa_hdl, &doi);
		len = sizeof (xoap->xoa_av_scanstamp) +
		    ZFS_OLD_ZNODE_PHYS_SIZE;

		if (len <= doi.doi_bonus_size) {
			(void) memcpy(xoap->xoa_av_scanstamp,
			    (caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
			    sizeof (xoap->xoa_av_scanstamp));
		}
	}
	XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
}

void
zfs_sa_set_scanstamp(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx)
{
	zfs_sb_t *zsb = ZTOZSB(zp);
	xoptattr_t *xoap;

	ASSERT(MUTEX_HELD(&zp->z_lock));
	VERIFY((xoap = xva_getxoptattr(xvap)) != NULL);
	if (zp->z_is_sa)
		VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SCANSTAMP(zsb),
		    &xoap->xoa_av_scanstamp,
		    sizeof (xoap->xoa_av_scanstamp), tx));
	else {
		dmu_object_info_t doi;
		dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
		int len;

		sa_object_info(zp->z_sa_hdl, &doi);
		len = sizeof (xoap->xoa_av_scanstamp) +
		    ZFS_OLD_ZNODE_PHYS_SIZE;
		if (len > doi.doi_bonus_size)
			VERIFY(dmu_set_bonus(db, len, tx) == 0);
		(void) memcpy((caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
		    xoap->xoa_av_scanstamp, sizeof (xoap->xoa_av_scanstamp));

		zp->z_pflags |= ZFS_BONUS_SCANSTAMP;
		VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_FLAGS(zsb),
		    &zp->z_pflags, sizeof (uint64_t), tx));
	}
}

/*
 * I'm not convinced we should do any of this upgrade.
 * since the SA code can read both old/new znode formats
 * with probably little to know performance difference.
 *
 * All new files will be created with the new format.
 */

void
zfs_sa_upgrade(sa_handle_t *hdl, dmu_tx_t *tx)
{
	dmu_buf_t *db = sa_get_db(hdl);
	znode_t *zp = sa_get_userdata(hdl);
	zfs_sb_t *zsb = ZTOZSB(zp);
	int count = 0;
	sa_bulk_attr_t *bulk, *sa_attrs;
	zfs_acl_locator_cb_t locate = { 0 };
	uint64_t uid, gid, mode, rdev, xattr, parent;
	uint64_t crtime[2], mtime[2], ctime[2];
	zfs_acl_phys_t znode_acl;
	char scanstamp[AV_SCANSTAMP_SZ];
	boolean_t drop_lock = B_FALSE;

	/*
	 * No upgrade if ACL isn't cached
	 * since we won't know which locks are held
	 * and ready the ACL would require special "locked"
	 * interfaces that would be messy
	 */
	if (zp->z_acl_cached == NULL || S_ISLNK(ZTOI(zp)->i_mode))
		return;

	/*
	 * If the z_lock is held and we aren't the owner
	 * the just return since we don't want to deadlock
	 * trying to update the status of z_is_sa.  This
	 * file can then be upgraded at a later time.
	 *
	 * Otherwise, we know we are doing the
	 * sa_update() that caused us to enter this function.
	 */
	if (mutex_owner(&zp->z_lock) != curthread) {
		if (mutex_tryenter(&zp->z_lock) == 0)
			return;
		else
			drop_lock = B_TRUE;
	}

	/* First do a bulk query of the attributes that aren't cached */
	bulk = kmem_alloc(sizeof(sa_bulk_attr_t) * 20, KM_SLEEP);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, &mtime, 16);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, &ctime, 16);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CRTIME(zsb), NULL, &crtime, 16);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, &mode, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zsb), NULL, &parent, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_XATTR(zsb), NULL, &xattr, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_RDEV(zsb), NULL, &rdev, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, &uid, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL, &gid, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zsb), NULL,
	    &znode_acl, 88);

	if (sa_bulk_lookup_locked(hdl, bulk, count) != 0) {
		kmem_free(bulk, sizeof(sa_bulk_attr_t) * 20);
		goto done;
	}

	/*
	 * While the order here doesn't matter its best to try and organize
	 * it is such a way to pick up an already existing layout number
	 */
	count = 0;
	sa_attrs = kmem_zalloc(sizeof(sa_bulk_attr_t) * 20, KM_SLEEP);
	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_MODE(zsb), NULL, &mode, 8);
	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_SIZE(zsb), NULL,
	    &zp->z_size, 8);
	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_GEN(zsb),
	    NULL, &zp->z_gen, 8);
	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_UID(zsb), NULL, &uid, 8);
	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_GID(zsb), NULL, &gid, 8);
	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_PARENT(zsb),
	    NULL, &parent, 8);
	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_FLAGS(zsb), NULL,
	    &zp->z_pflags, 8);
	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_ATIME(zsb), NULL,
	    zp->z_atime, 16);
	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_MTIME(zsb), NULL,
	    &mtime, 16);
	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_CTIME(zsb), NULL,
	    &ctime, 16);
	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_CRTIME(zsb), NULL,
	    &crtime, 16);
	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_LINKS(zsb), NULL,
	    &zp->z_links, 8);
	if (S_ISBLK(ZTOI(zp)->i_mode) || S_ISCHR(ZTOI(zp)->i_mode))
		SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_RDEV(zsb), NULL,
		    &rdev, 8);
	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_DACL_COUNT(zsb), NULL,
	    &zp->z_acl_cached->z_acl_count, 8);

	if (zp->z_acl_cached->z_version < ZFS_ACL_VERSION_FUID)
		zfs_acl_xform(zp, zp->z_acl_cached, CRED());

	locate.cb_aclp = zp->z_acl_cached;
	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_DACL_ACES(zsb),
	    zfs_acl_data_locator, &locate, zp->z_acl_cached->z_acl_bytes);

	if (xattr)
		SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_XATTR(zsb),
		    NULL, &xattr, 8);

	/* if scanstamp then add scanstamp */

	if (zp->z_pflags & ZFS_BONUS_SCANSTAMP) {
		bcopy((caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
		    scanstamp, AV_SCANSTAMP_SZ);
		SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_SCANSTAMP(zsb),
		    NULL, scanstamp, AV_SCANSTAMP_SZ);
		zp->z_pflags &= ~ZFS_BONUS_SCANSTAMP;
	}

	VERIFY(dmu_set_bonustype(db, DMU_OT_SA, tx) == 0);
	VERIFY(sa_replace_all_by_template_locked(hdl, sa_attrs,
	    count, tx) == 0);
	if (znode_acl.z_acl_extern_obj)
		VERIFY(0 == dmu_object_free(zsb->z_os,
		    znode_acl.z_acl_extern_obj, tx));

	zp->z_is_sa = B_TRUE;
	kmem_free(sa_attrs, sizeof(sa_bulk_attr_t) * 20);
	kmem_free(bulk, sizeof(sa_bulk_attr_t) * 20);
done:
	if (drop_lock)
		mutex_exit(&zp->z_lock);
}

void
zfs_sa_upgrade_txholds(dmu_tx_t *tx, znode_t *zp)
{
	if (!ZTOZSB(zp)->z_use_sa || zp->z_is_sa)
		return;


	dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);

	if (zfs_external_acl(zp)) {
		dmu_tx_hold_free(tx, zfs_external_acl(zp), 0,
		    DMU_OBJECT_END);
	}
}

EXPORT_SYMBOL(zfs_sa_readlink);
EXPORT_SYMBOL(zfs_sa_symlink);
EXPORT_SYMBOL(zfs_sa_get_scanstamp);
EXPORT_SYMBOL(zfs_sa_set_scanstamp);
EXPORT_SYMBOL(zfs_sa_upgrade);
EXPORT_SYMBOL(zfs_sa_upgrade_txholds);

#endif
Commit	Line	Data
428870ff BB	1	/*
	2	* CDDL HEADER START
	3	*
	4	* The contents of this file are subject to the terms of the
	5	* Common Development and Distribution License (the "License").
	6	* You may not use this file except in compliance with the License.
	7	*
	8	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	9	* or http://www.opensolaris.org/os/licensing.
	10	* See the License for the specific language governing permissions
	11	* and limitations under the License.
	12	*
	13	* When distributing Covered Code, include this CDDL HEADER in each
	14	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	15	* If applicable, add the following below this CDDL HEADER, with the
	16	* fields enclosed by brackets "[]" replaced with your own identifying
	17	* information: Portions Copyright [yyyy] [name of copyright owner]
	18	*
	19	* CDDL HEADER END
	20	*/
	21	/*
	22	* Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved.
	23	*/
	24
	25	#include <sys/types.h>
	26	#include <sys/param.h>
	27	#include <sys/vnode.h>
	28	#include <sys/sa.h>
	29	#include <sys/zfs_acl.h>
	30	#include <sys/zfs_sa.h>
	31
	32	/*
	33	* ZPL attribute registration table.
	34	* Order of attributes doesn't matter
	35	* a unique value will be assigned for each
	36	* attribute that is file system specific
	37	*
	38	* This is just the set of ZPL attributes that this
	39	* version of ZFS deals with natively. The file system
	40	* could have other attributes stored in files, but they will be
	41	* ignored. The SA framework will preserve them, just that
	42	* this version of ZFS won't change or delete them.
	43	*/
	44
	45	sa_attr_reg_t zfs_attr_table[ZPL_END+1] = {
	46	{"ZPL_ATIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 0},
	47	{"ZPL_MTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 1},
	48	{"ZPL_CTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 2},
	49	{"ZPL_CRTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 3},
	50	{"ZPL_GEN", sizeof (uint64_t), SA_UINT64_ARRAY, 4},
	51	{"ZPL_MODE", sizeof (uint64_t), SA_UINT64_ARRAY, 5},
	52	{"ZPL_SIZE", sizeof (uint64_t), SA_UINT64_ARRAY, 6},
	53	{"ZPL_PARENT", sizeof (uint64_t), SA_UINT64_ARRAY, 7},
	54	{"ZPL_LINKS", sizeof (uint64_t), SA_UINT64_ARRAY, 8},
	55	{"ZPL_XATTR", sizeof (uint64_t), SA_UINT64_ARRAY, 9},
	56	{"ZPL_RDEV", sizeof (uint64_t), SA_UINT64_ARRAY, 10},
	57	{"ZPL_FLAGS", sizeof (uint64_t), SA_UINT64_ARRAY, 11},
	58	{"ZPL_UID", sizeof (uint64_t), SA_UINT64_ARRAY, 12},
	59	{"ZPL_GID", sizeof (uint64_t), SA_UINT64_ARRAY, 13},
	60	{"ZPL_PAD", sizeof (uint64_t) * 4, SA_UINT64_ARRAY, 14},
	61	{"ZPL_ZNODE_ACL", 88, SA_UINT8_ARRAY, 15},
	62	{"ZPL_DACL_COUNT", sizeof (uint64_t), SA_UINT64_ARRAY, 0},
	63	{"ZPL_SYMLINK", 0, SA_UINT8_ARRAY, 0},
	64	{"ZPL_SCANSTAMP", 32, SA_UINT8_ARRAY, 0},
65	{"ZPL_DACL_ACES", 0, SA_ACL, 0},
66	{NULL, 0, 0, 0}
67	};
68
69	#ifdef _KERNEL
428870ff BB	70	int
	71	zfs_sa_readlink(znode_t zp, uio_t uio)
	72	{
	73	dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
	74	size_t bufsz;
	75	int error;
	76
	77	bufsz = zp->z_size;
	78	if (bufsz + ZFS_OLD_ZNODE_PHYS_SIZE <= db->db_size) {
	79	error = uiomove((caddr_t)db->db_data +
	80	ZFS_OLD_ZNODE_PHYS_SIZE,
	81	MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio);
	82	} else {
	83	dmu_buf_t *dbp;
3558fd73	84	if ((error = dmu_buf_hold(ZTOZSB(zp)->z_os, zp->z_id,
428870ff BB	85	0, FTAG, &dbp, DMU_READ_NO_PREFETCH)) == 0) {
	86	error = uiomove(dbp->db_data,
	87	MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio);
	88	dmu_buf_rele(dbp, FTAG);
	89	}
	90	}
	91	return (error);
	92	}
	93
	94	void
	95	zfs_sa_symlink(znode_t zp, char link, int len, dmu_tx_t *tx)
	96	{
	97	dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
	98
	99	if (ZFS_OLD_ZNODE_PHYS_SIZE + len <= dmu_bonus_max()) {
	100	VERIFY(dmu_set_bonus(db,
	101	len + ZFS_OLD_ZNODE_PHYS_SIZE, tx) == 0);
	102	if (len) {
	103	bcopy(link, (caddr_t)db->db_data +
	104	ZFS_OLD_ZNODE_PHYS_SIZE, len);
	105	}
	106	} else {
	107	dmu_buf_t *dbp;
	108
	109	zfs_grow_blocksize(zp, len, tx);
3558fd73	110	VERIFY(0 == dmu_buf_hold(ZTOZSB(zp)->z_os,
428870ff BB	111	zp->z_id, 0, FTAG, &dbp, DMU_READ_NO_PREFETCH));
	112
	113	dmu_buf_will_dirty(dbp, tx);
	114
	115	ASSERT3U(len, <=, dbp->db_size);
	116	bcopy(link, dbp->db_data, len);
	117	dmu_buf_rele(dbp, FTAG);
	118	}
	119	}
	120
	121	void
	122	zfs_sa_get_scanstamp(znode_t zp, xvattr_t xvap)
	123	{
3558fd73	124	zfs_sb_t *zsb = ZTOZSB(zp);
428870ff BB	125	xoptattr_t *xoap;
428870ff BB	126
572e2857	127	ASSERT(MUTEX_HELD(&zp->z_lock));
428870ff BB	128	VERIFY((xoap = xva_getxoptattr(xvap)) != NULL);
428870ff BB	129	if (zp->z_is_sa) {
3558fd73	130	if (sa_lookup(zp->z_sa_hdl, SA_ZPL_SCANSTAMP(zsb),
428870ff BB	131	&xoap->xoa_av_scanstamp,
	132	sizeof (xoap->xoa_av_scanstamp)) != 0)
	133	return;
	134	} else {
	135	dmu_object_info_t doi;
	136	dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
	137	int len;
	138
	139	if (!(zp->z_pflags & ZFS_BONUS_SCANSTAMP))
	140	return;
	141
	142	sa_object_info(zp->z_sa_hdl, &doi);
	143	len = sizeof (xoap->xoa_av_scanstamp) +
	144	ZFS_OLD_ZNODE_PHYS_SIZE;
	145
	146	if (len <= doi.doi_bonus_size) {
	147	(void) memcpy(xoap->xoa_av_scanstamp,
	148	(caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
	149	sizeof (xoap->xoa_av_scanstamp));
	150	}
	151	}
	152	XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
	153	}
	154
	155	void
	156	zfs_sa_set_scanstamp(znode_t zp, xvattr_t xvap, dmu_tx_t *tx)
	157	{
3558fd73	158	zfs_sb_t *zsb = ZTOZSB(zp);
428870ff BB	159	xoptattr_t *xoap;
428870ff BB	160
572e2857	161	ASSERT(MUTEX_HELD(&zp->z_lock));
428870ff BB	162	VERIFY((xoap = xva_getxoptattr(xvap)) != NULL);
428870ff BB	163	if (zp->z_is_sa)
3558fd73	164	VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SCANSTAMP(zsb),
428870ff BB	165	&xoap->xoa_av_scanstamp,
	166	sizeof (xoap->xoa_av_scanstamp), tx));
	167	else {
	168	dmu_object_info_t doi;
	169	dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
	170	int len;
	171
	172	sa_object_info(zp->z_sa_hdl, &doi);
	173	len = sizeof (xoap->xoa_av_scanstamp) +
	174	ZFS_OLD_ZNODE_PHYS_SIZE;
	175	if (len > doi.doi_bonus_size)
	176	VERIFY(dmu_set_bonus(db, len, tx) == 0);
	177	(void) memcpy((caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
	178	xoap->xoa_av_scanstamp, sizeof (xoap->xoa_av_scanstamp));
	179
	180	zp->z_pflags \|= ZFS_BONUS_SCANSTAMP;
3558fd73	181	VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_FLAGS(zsb),
428870ff BB	182	&zp->z_pflags, sizeof (uint64_t), tx));
	183	}
	184	}
	185
	186	/*
	187	* I'm not convinced we should do any of this upgrade.
	188	* since the SA code can read both old/new znode formats
	189	* with probably little to know performance difference.
	190	*
	191	* All new files will be created with the new format.
	192	*/
	193
	194	void
	195	zfs_sa_upgrade(sa_handle_t hdl, dmu_tx_t tx)
	196	{
	197	dmu_buf_t *db = sa_get_db(hdl);
	198	znode_t *zp = sa_get_userdata(hdl);
3558fd73	199	zfs_sb_t *zsb = ZTOZSB(zp);
428870ff	200	int count = 0;
1ee1b767	201	sa_bulk_attr_t bulk, sa_attrs;
428870ff BB	202	zfs_acl_locator_cb_t locate = { 0 };
	203	uint64_t uid, gid, mode, rdev, xattr, parent;
	204	uint64_t crtime[2], mtime[2], ctime[2];
	205	zfs_acl_phys_t znode_acl;
	206	char scanstamp[AV_SCANSTAMP_SZ];
572e2857	207	boolean_t drop_lock = B_FALSE;
428870ff BB	208
	209	/*
	210	* No upgrade if ACL isn't cached
	211	* since we won't know which locks are held
	212	* and ready the ACL would require special "locked"
	213	* interfaces that would be messy
	214	*/
3558fd73	215	if (zp->z_acl_cached == NULL \|\| S_ISLNK(ZTOI(zp)->i_mode))
428870ff BB	216	return;
428870ff BB	217
572e2857 BB	218	/*
	219	* If the z_lock is held and we aren't the owner
	220	* the just return since we don't want to deadlock
	221	* trying to update the status of z_is_sa. This
	222	* file can then be upgraded at a later time.
	223	*
	224	* Otherwise, we know we are doing the
	225	* sa_update() that caused us to enter this function.
	226	*/
	227	if (mutex_owner(&zp->z_lock) != curthread) {
	228	if (mutex_tryenter(&zp->z_lock) == 0)
	229	return;
	230	else
	231	drop_lock = B_TRUE;
	232	}
	233
428870ff	234	/* First do a bulk query of the attributes that aren't cached */
1ee1b767	235	bulk = kmem_alloc(sizeof(sa_bulk_attr_t) * 20, KM_SLEEP);
3558fd73 BB	236	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, &mtime, 16);
	237	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, &ctime, 16);
	238	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CRTIME(zsb), NULL, &crtime, 16);
	239	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, &mode, 8);
	240	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zsb), NULL, &parent, 8);
	241	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_XATTR(zsb), NULL, &xattr, 8);
	242	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_RDEV(zsb), NULL, &rdev, 8);
	243	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, &uid, 8);
	244	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL, &gid, 8);
	245	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zsb), NULL,
428870ff BB	246	&znode_acl, 88);
428870ff BB	247
1ee1b767 BB	248	if (sa_bulk_lookup_locked(hdl, bulk, count) != 0) {
1ee1b767 BB	249	kmem_free(bulk, sizeof(sa_bulk_attr_t) * 20);
572e2857	250	goto done;
1ee1b767	251	}
428870ff BB	252
	253	/*
	254	* While the order here doesn't matter its best to try and organize
	255	* it is such a way to pick up an already existing layout number
	256	*/
	257	count = 0;
1ee1b767	258	sa_attrs = kmem_zalloc(sizeof(sa_bulk_attr_t) * 20, KM_SLEEP);
3558fd73 BB	259	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_MODE(zsb), NULL, &mode, 8);
3558fd73 BB	260	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_SIZE(zsb), NULL,
428870ff	261	&zp->z_size, 8);
3558fd73	262	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_GEN(zsb),
428870ff	263	NULL, &zp->z_gen, 8);
3558fd73 BB	264	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_UID(zsb), NULL, &uid, 8);
	265	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_GID(zsb), NULL, &gid, 8);
	266	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_PARENT(zsb),
428870ff	267	NULL, &parent, 8);
3558fd73	268	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_FLAGS(zsb), NULL,
428870ff	269	&zp->z_pflags, 8);
3558fd73	270	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_ATIME(zsb), NULL,
428870ff	271	zp->z_atime, 16);
3558fd73	272	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_MTIME(zsb), NULL,
428870ff	273	&mtime, 16);
3558fd73	274	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_CTIME(zsb), NULL,
428870ff	275	&ctime, 16);
3558fd73	276	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_CRTIME(zsb), NULL,
428870ff	277	&crtime, 16);
3558fd73	278	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_LINKS(zsb), NULL,
428870ff	279	&zp->z_links, 8);
3558fd73 BB	280	if (S_ISBLK(ZTOI(zp)->i_mode) \|\| S_ISCHR(ZTOI(zp)->i_mode))
3558fd73 BB	281	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_RDEV(zsb), NULL,
428870ff	282	&rdev, 8);
3558fd73	283	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_DACL_COUNT(zsb), NULL,
428870ff BB	284	&zp->z_acl_cached->z_acl_count, 8);
	285
	286	if (zp->z_acl_cached->z_version < ZFS_ACL_VERSION_FUID)
	287	zfs_acl_xform(zp, zp->z_acl_cached, CRED());
	288
	289	locate.cb_aclp = zp->z_acl_cached;
3558fd73	290	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_DACL_ACES(zsb),
428870ff	291	zfs_acl_data_locator, &locate, zp->z_acl_cached->z_acl_bytes);
572e2857	292
428870ff	293	if (xattr)
3558fd73	294	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_XATTR(zsb),
572e2857	295	NULL, &xattr, 8);
428870ff BB	296
	297	/* if scanstamp then add scanstamp */
	298
	299	if (zp->z_pflags & ZFS_BONUS_SCANSTAMP) {
	300	bcopy((caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
	301	scanstamp, AV_SCANSTAMP_SZ);
3558fd73	302	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_SCANSTAMP(zsb),
428870ff BB	303	NULL, scanstamp, AV_SCANSTAMP_SZ);
	304	zp->z_pflags &= ~ZFS_BONUS_SCANSTAMP;
	305	}
	306
	307	VERIFY(dmu_set_bonustype(db, DMU_OT_SA, tx) == 0);
	308	VERIFY(sa_replace_all_by_template_locked(hdl, sa_attrs,
	309	count, tx) == 0);
	310	if (znode_acl.z_acl_extern_obj)
3558fd73	311	VERIFY(0 == dmu_object_free(zsb->z_os,
428870ff BB	312	znode_acl.z_acl_extern_obj, tx));
	313
	314	zp->z_is_sa = B_TRUE;
1ee1b767 BB	315	kmem_free(sa_attrs, sizeof(sa_bulk_attr_t) * 20);
1ee1b767 BB	316	kmem_free(bulk, sizeof(sa_bulk_attr_t) * 20);
572e2857 BB	317	done:
	318	if (drop_lock)
	319	mutex_exit(&zp->z_lock);
428870ff BB	320	}
	321
	322	void
	323	zfs_sa_upgrade_txholds(dmu_tx_t tx, znode_t zp)
	324	{
3558fd73	325	if (!ZTOZSB(zp)->z_use_sa \|\| zp->z_is_sa)
428870ff BB	326	return;
428870ff BB	327
428870ff BB	328
	329	dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
	330
572e2857 BB	331	if (zfs_external_acl(zp)) {
572e2857 BB	332	dmu_tx_hold_free(tx, zfs_external_acl(zp), 0,
428870ff BB	333	DMU_OBJECT_END);
	334	}
	335	}
	336
e45aa452 BB	337	EXPORT_SYMBOL(zfs_sa_readlink);
	338	EXPORT_SYMBOL(zfs_sa_symlink);
	339	EXPORT_SYMBOL(zfs_sa_get_scanstamp);
	340	EXPORT_SYMBOL(zfs_sa_set_scanstamp);
	341	EXPORT_SYMBOL(zfs_sa_upgrade);
	342	EXPORT_SYMBOL(zfs_sa_upgrade_txholds);
	343
428870ff	344	#endif