[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/zfs_context.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},
	{"ZPL_DXATTR", 0, SA_UINT8_ARRAY, 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()) {
		VERIFY0(dmu_set_bonus(db, len + ZFS_OLD_ZNODE_PHYS_SIZE, tx));
		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);
		VERIFY0(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));
	}
}

int
zfs_sa_get_xattr(znode_t *zp)
{
	zfs_sb_t *zsb = ZTOZSB(zp);
	char *obj;
	int size;
	int error;

	ASSERT(RW_LOCK_HELD(&zp->z_xattr_lock));
	ASSERT(!zp->z_xattr_cached);
	ASSERT(zp->z_is_sa);

	error = sa_size(zp->z_sa_hdl, SA_ZPL_DXATTR(zsb), &size);
	if (error) {
		if (error == ENOENT)
			return nvlist_alloc(&zp->z_xattr_cached,
			    NV_UNIQUE_NAME, KM_SLEEP);
		else
			return (error);
	}

	obj = zio_buf_alloc(size);

	error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DXATTR(zsb), obj, size);
	if (error == 0)
		error = nvlist_unpack(obj, size, &zp->z_xattr_cached, KM_SLEEP);

	zio_buf_free(obj, size);

	return (error);
}

int
zfs_sa_set_xattr(znode_t *zp)
{
	zfs_sb_t *zsb = ZTOZSB(zp);
	dmu_tx_t *tx;
	char *obj;
	size_t size;
	int error;

	ASSERT(RW_WRITE_HELD(&zp->z_xattr_lock));
	ASSERT(zp->z_xattr_cached);
	ASSERT(zp->z_is_sa);

	error = nvlist_size(zp->z_xattr_cached, &size, NV_ENCODE_XDR);
	if ((error == 0) && (size > SA_ATTR_MAX_LEN))
		error = EFBIG;
	if (error)
		goto out;

	obj = zio_buf_alloc(size);

	error = nvlist_pack(zp->z_xattr_cached, &obj, &size,
	    NV_ENCODE_XDR, KM_SLEEP);
	if (error)
		goto out_free;

	tx = dmu_tx_create(zsb->z_os);
	dmu_tx_hold_sa_create(tx, size);
	dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);

	error = dmu_tx_assign(tx, TXG_WAIT);
	if (error) {
		dmu_tx_abort(tx);
	} else {
		VERIFY0(sa_update(zp->z_sa_hdl, SA_ZPL_DXATTR(zsb),
		    obj, size, tx));
		dmu_tx_commit(tx);
	}
out_free:
	zio_buf_free(obj, size);
out:
	return (error);
}

/*
 * 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 no 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, tmp_gen;
	uint64_t crtime[2], mtime[2], ctime[2], atime[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_ATIME(zsb), NULL, &atime, 16);
	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_GEN(zsb), NULL, &tmp_gen, 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, &tmp_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,
	    &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_attr_table);
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_get_xattr);
EXPORT_SYMBOL(zfs_sa_set_xattr);
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
0c66c32d	25	#include <sys/zfs_context.h>
428870ff BB	26	#include <sys/vnode.h>
	27	#include <sys/sa.h>
	28	#include <sys/zfs_acl.h>
	29	#include <sys/zfs_sa.h>
	30
	31	/*
	32	* ZPL attribute registration table.
	33	* Order of attributes doesn't matter
	34	* a unique value will be assigned for each
	35	* attribute that is file system specific
	36	*
	37	* This is just the set of ZPL attributes that this
	38	* version of ZFS deals with natively. The file system
	39	* could have other attributes stored in files, but they will be
	40	* ignored. The SA framework will preserve them, just that
	41	* this version of ZFS won't change or delete them.
	42	*/
	43
	44	sa_attr_reg_t zfs_attr_table[ZPL_END+1] = {
	45	{"ZPL_ATIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 0},
	46	{"ZPL_MTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 1},
	47	{"ZPL_CTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 2},
	48	{"ZPL_CRTIME", sizeof (uint64_t) * 2, SA_UINT64_ARRAY, 3},
	49	{"ZPL_GEN", sizeof (uint64_t), SA_UINT64_ARRAY, 4},
	50	{"ZPL_MODE", sizeof (uint64_t), SA_UINT64_ARRAY, 5},
	51	{"ZPL_SIZE", sizeof (uint64_t), SA_UINT64_ARRAY, 6},
	52	{"ZPL_PARENT", sizeof (uint64_t), SA_UINT64_ARRAY, 7},
	53	{"ZPL_LINKS", sizeof (uint64_t), SA_UINT64_ARRAY, 8},
	54	{"ZPL_XATTR", sizeof (uint64_t), SA_UINT64_ARRAY, 9},
	55	{"ZPL_RDEV", sizeof (uint64_t), SA_UINT64_ARRAY, 10},
	56	{"ZPL_FLAGS", sizeof (uint64_t), SA_UINT64_ARRAY, 11},
	57	{"ZPL_UID", sizeof (uint64_t), SA_UINT64_ARRAY, 12},
	58	{"ZPL_GID", sizeof (uint64_t), SA_UINT64_ARRAY, 13},
	59	{"ZPL_PAD", sizeof (uint64_t) * 4, SA_UINT64_ARRAY, 14},
	60	{"ZPL_ZNODE_ACL", 88, SA_UINT8_ARRAY, 15},
	61	{"ZPL_DACL_COUNT", sizeof (uint64_t), SA_UINT64_ARRAY, 0},
	62	{"ZPL_SYMLINK", 0, SA_UINT8_ARRAY, 0},
	63	{"ZPL_SCANSTAMP", 32, SA_UINT8_ARRAY, 0},
	64	{"ZPL_DACL_ACES", 0, SA_ACL, 0},
82a37189	65	{"ZPL_DXATTR", 0, SA_UINT8_ARRAY, 0},
428870ff BB	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()) {
50c957f7	100	VERIFY0(dmu_set_bonus(db, len + ZFS_OLD_ZNODE_PHYS_SIZE, tx));
428870ff BB	101	if (len) {
	102	bcopy(link, (caddr_t)db->db_data +
	103	ZFS_OLD_ZNODE_PHYS_SIZE, len);
	104	}
	105	} else {
	106	dmu_buf_t *dbp;
	107
	108	zfs_grow_blocksize(zp, len, tx);
50c957f7 NB	109	VERIFY0(dmu_buf_hold(ZTOZSB(zp)->z_os, zp->z_id, 0, FTAG, &dbp,
50c957f7 NB	110	DMU_READ_NO_PREFETCH));
428870ff BB	111
	112	dmu_buf_will_dirty(dbp, tx);
	113
	114	ASSERT3U(len, <=, dbp->db_size);
	115	bcopy(link, dbp->db_data, len);
	116	dmu_buf_rele(dbp, FTAG);
	117	}
	118	}
	119
	120	void
	121	zfs_sa_get_scanstamp(znode_t zp, xvattr_t xvap)
	122	{
3558fd73	123	zfs_sb_t *zsb = ZTOZSB(zp);
428870ff BB	124	xoptattr_t *xoap;
428870ff BB	125
572e2857	126	ASSERT(MUTEX_HELD(&zp->z_lock));
428870ff BB	127	VERIFY((xoap = xva_getxoptattr(xvap)) != NULL);
428870ff BB	128	if (zp->z_is_sa) {
3558fd73	129	if (sa_lookup(zp->z_sa_hdl, SA_ZPL_SCANSTAMP(zsb),
428870ff BB	130	&xoap->xoa_av_scanstamp,
	131	sizeof (xoap->xoa_av_scanstamp)) != 0)
	132	return;
	133	} else {
	134	dmu_object_info_t doi;
	135	dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
	136	int len;
	137
	138	if (!(zp->z_pflags & ZFS_BONUS_SCANSTAMP))
	139	return;
	140
	141	sa_object_info(zp->z_sa_hdl, &doi);
	142	len = sizeof (xoap->xoa_av_scanstamp) +
	143	ZFS_OLD_ZNODE_PHYS_SIZE;
	144
	145	if (len <= doi.doi_bonus_size) {
	146	(void) memcpy(xoap->xoa_av_scanstamp,
	147	(caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
	148	sizeof (xoap->xoa_av_scanstamp));
	149	}
	150	}
	151	XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
	152	}
	153
	154	void
	155	zfs_sa_set_scanstamp(znode_t zp, xvattr_t xvap, dmu_tx_t *tx)
	156	{
3558fd73	157	zfs_sb_t *zsb = ZTOZSB(zp);
428870ff BB	158	xoptattr_t *xoap;
428870ff BB	159
572e2857	160	ASSERT(MUTEX_HELD(&zp->z_lock));
428870ff BB	161	VERIFY((xoap = xva_getxoptattr(xvap)) != NULL);
428870ff BB	162	if (zp->z_is_sa)
3558fd73	163	VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SCANSTAMP(zsb),
428870ff BB	164	&xoap->xoa_av_scanstamp,
	165	sizeof (xoap->xoa_av_scanstamp), tx));
	166	else {
	167	dmu_object_info_t doi;
	168	dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
	169	int len;
	170
	171	sa_object_info(zp->z_sa_hdl, &doi);
	172	len = sizeof (xoap->xoa_av_scanstamp) +
	173	ZFS_OLD_ZNODE_PHYS_SIZE;
	174	if (len > doi.doi_bonus_size)
	175	VERIFY(dmu_set_bonus(db, len, tx) == 0);
	176	(void) memcpy((caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
	177	xoap->xoa_av_scanstamp, sizeof (xoap->xoa_av_scanstamp));
	178
	179	zp->z_pflags \|= ZFS_BONUS_SCANSTAMP;
3558fd73	180	VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_FLAGS(zsb),
428870ff BB	181	&zp->z_pflags, sizeof (uint64_t), tx));
	182	}
	183	}
	184
82a37189 BB	185	int
	186	zfs_sa_get_xattr(znode_t *zp)
	187	{
	188	zfs_sb_t *zsb = ZTOZSB(zp);
	189	char *obj;
	190	int size;
	191	int error;
	192
	193	ASSERT(RW_LOCK_HELD(&zp->z_xattr_lock));
	194	ASSERT(!zp->z_xattr_cached);
	195	ASSERT(zp->z_is_sa);
	196
f828e63a	197	error = sa_size(zp->z_sa_hdl, SA_ZPL_DXATTR(zsb), &size);
82a37189 BB	198	if (error) {
	199	if (error == ENOENT)
	200	return nvlist_alloc(&zp->z_xattr_cached,
	201	NV_UNIQUE_NAME, KM_SLEEP);
	202	else
	203	return (error);
	204	}
	205
81971b13	206	obj = zio_buf_alloc(size);
82a37189	207
f828e63a	208	error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DXATTR(zsb), obj, size);
82a37189 BB	209	if (error == 0)
	210	error = nvlist_unpack(obj, size, &zp->z_xattr_cached, KM_SLEEP);
	211
81971b13	212	zio_buf_free(obj, size);
82a37189 BB	213
	214	return (error);
	215	}
	216
	217	int
	218	zfs_sa_set_xattr(znode_t *zp)
	219	{
	220	zfs_sb_t *zsb = ZTOZSB(zp);
	221	dmu_tx_t *tx;
	222	char *obj;
	223	size_t size;
	224	int error;
	225
	226	ASSERT(RW_WRITE_HELD(&zp->z_xattr_lock));
	227	ASSERT(zp->z_xattr_cached);
	228	ASSERT(zp->z_is_sa);
	229
	230	error = nvlist_size(zp->z_xattr_cached, &size, NV_ENCODE_XDR);
43b4935e NB	231	if ((error == 0) && (size > SA_ATTR_MAX_LEN))
43b4935e NB	232	error = EFBIG;
82a37189 BB	233	if (error)
	234	goto out;
	235
81971b13	236	obj = zio_buf_alloc(size);
82a37189 BB	237
	238	error = nvlist_pack(zp->z_xattr_cached, &obj, &size,
	239	NV_ENCODE_XDR, KM_SLEEP);
	240	if (error)
	241	goto out_free;
	242
	243	tx = dmu_tx_create(zsb->z_os);
	244	dmu_tx_hold_sa_create(tx, size);
f828e63a	245	dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
82a37189 BB	246
	247	error = dmu_tx_assign(tx, TXG_WAIT);
	248	if (error) {
	249	dmu_tx_abort(tx);
	250	} else {
43b4935e NB	251	VERIFY0(sa_update(zp->z_sa_hdl, SA_ZPL_DXATTR(zsb),
	252	obj, size, tx));
	253	dmu_tx_commit(tx);
82a37189 BB	254	}
82a37189 BB	255	out_free:
81971b13	256	zio_buf_free(obj, size);
82a37189 BB	257	out:
	258	return (error);
	259	}
	260
428870ff BB	261	/*
	262	* I'm not convinced we should do any of this upgrade.
	263	* since the SA code can read both old/new znode formats
d3cc8b15	264	* with probably little to no performance difference.
428870ff BB	265	*
	266	* All new files will be created with the new format.
	267	*/
	268
	269	void
	270	zfs_sa_upgrade(sa_handle_t hdl, dmu_tx_t tx)
	271	{
	272	dmu_buf_t *db = sa_get_db(hdl);
	273	znode_t *zp = sa_get_userdata(hdl);
3558fd73	274	zfs_sb_t *zsb = ZTOZSB(zp);
428870ff	275	int count = 0;
1ee1b767	276	sa_bulk_attr_t bulk, sa_attrs;
428870ff	277	zfs_acl_locator_cb_t locate = { 0 };
278f2236	278	uint64_t uid, gid, mode, rdev, xattr, parent, tmp_gen;
0df9673f	279	uint64_t crtime[2], mtime[2], ctime[2], atime[2];
428870ff BB	280	zfs_acl_phys_t znode_acl;
428870ff BB	281	char scanstamp[AV_SCANSTAMP_SZ];
572e2857	282	boolean_t drop_lock = B_FALSE;
428870ff BB	283
	284	/*
	285	* No upgrade if ACL isn't cached
	286	* since we won't know which locks are held
	287	* and ready the ACL would require special "locked"
	288	* interfaces that would be messy
	289	*/
3558fd73	290	if (zp->z_acl_cached == NULL \|\| S_ISLNK(ZTOI(zp)->i_mode))
428870ff BB	291	return;
428870ff BB	292
572e2857 BB	293	/*
	294	* If the z_lock is held and we aren't the owner
	295	* the just return since we don't want to deadlock
	296	* trying to update the status of z_is_sa. This
	297	* file can then be upgraded at a later time.
	298	*
	299	* Otherwise, we know we are doing the
	300	* sa_update() that caused us to enter this function.
	301	*/
	302	if (mutex_owner(&zp->z_lock) != curthread) {
	303	if (mutex_tryenter(&zp->z_lock) == 0)
	304	return;
	305	else
	306	drop_lock = B_TRUE;
	307	}
	308
428870ff	309	/* First do a bulk query of the attributes that aren't cached */
d1d7e268	310	bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * 20, KM_SLEEP);
0df9673f	311	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL, &atime, 16);
3558fd73 BB	312	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, &mtime, 16);
	313	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, &ctime, 16);
	314	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CRTIME(zsb), NULL, &crtime, 16);
	315	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, &mode, 8);
	316	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zsb), NULL, &parent, 8);
	317	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_XATTR(zsb), NULL, &xattr, 8);
	318	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_RDEV(zsb), NULL, &rdev, 8);
	319	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, &uid, 8);
	320	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL, &gid, 8);
278f2236	321	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zsb), NULL, &tmp_gen, 8);
3558fd73	322	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zsb), NULL,
428870ff BB	323	&znode_acl, 88);
428870ff BB	324
1ee1b767	325	if (sa_bulk_lookup_locked(hdl, bulk, count) != 0) {
d1d7e268	326	kmem_free(bulk, sizeof (sa_bulk_attr_t) * 20);
572e2857	327	goto done;
1ee1b767	328	}
428870ff BB	329
	330	/*
	331	* While the order here doesn't matter its best to try and organize
	332	* it is such a way to pick up an already existing layout number
	333	*/
	334	count = 0;
d1d7e268	335	sa_attrs = kmem_zalloc(sizeof (sa_bulk_attr_t) * 20, KM_SLEEP);
3558fd73 BB	336	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_MODE(zsb), NULL, &mode, 8);
3558fd73 BB	337	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_SIZE(zsb), NULL,
428870ff	338	&zp->z_size, 8);
278f2236	339	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_GEN(zsb), NULL, &tmp_gen, 8);
3558fd73 BB	340	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_UID(zsb), NULL, &uid, 8);
	341	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_GID(zsb), NULL, &gid, 8);
	342	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_PARENT(zsb),
428870ff	343	NULL, &parent, 8);
3558fd73	344	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_FLAGS(zsb), NULL,
428870ff	345	&zp->z_pflags, 8);
3558fd73	346	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_ATIME(zsb), NULL,
0df9673f	347	&atime, 16);
3558fd73	348	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_MTIME(zsb), NULL,
428870ff	349	&mtime, 16);
3558fd73	350	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_CTIME(zsb), NULL,
428870ff	351	&ctime, 16);
3558fd73	352	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_CRTIME(zsb), NULL,
428870ff	353	&crtime, 16);
3558fd73	354	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_LINKS(zsb), NULL,
428870ff	355	&zp->z_links, 8);
3558fd73 BB	356	if (S_ISBLK(ZTOI(zp)->i_mode) \|\| S_ISCHR(ZTOI(zp)->i_mode))
3558fd73 BB	357	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_RDEV(zsb), NULL,
428870ff	358	&rdev, 8);
3558fd73	359	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_DACL_COUNT(zsb), NULL,
428870ff BB	360	&zp->z_acl_cached->z_acl_count, 8);
	361
	362	if (zp->z_acl_cached->z_version < ZFS_ACL_VERSION_FUID)
	363	zfs_acl_xform(zp, zp->z_acl_cached, CRED());
	364
	365	locate.cb_aclp = zp->z_acl_cached;
3558fd73	366	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_DACL_ACES(zsb),
428870ff	367	zfs_acl_data_locator, &locate, zp->z_acl_cached->z_acl_bytes);
572e2857	368
428870ff	369	if (xattr)
3558fd73	370	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_XATTR(zsb),
572e2857	371	NULL, &xattr, 8);
428870ff BB	372
	373	/* if scanstamp then add scanstamp */
	374
	375	if (zp->z_pflags & ZFS_BONUS_SCANSTAMP) {
	376	bcopy((caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
	377	scanstamp, AV_SCANSTAMP_SZ);
3558fd73	378	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_SCANSTAMP(zsb),
428870ff BB	379	NULL, scanstamp, AV_SCANSTAMP_SZ);
	380	zp->z_pflags &= ~ZFS_BONUS_SCANSTAMP;
	381	}
	382
	383	VERIFY(dmu_set_bonustype(db, DMU_OT_SA, tx) == 0);
	384	VERIFY(sa_replace_all_by_template_locked(hdl, sa_attrs,
	385	count, tx) == 0);
	386	if (znode_acl.z_acl_extern_obj)
3558fd73	387	VERIFY(0 == dmu_object_free(zsb->z_os,
428870ff BB	388	znode_acl.z_acl_extern_obj, tx));
	389
	390	zp->z_is_sa = B_TRUE;
d1d7e268 MK	391	kmem_free(sa_attrs, sizeof (sa_bulk_attr_t) * 20);
d1d7e268 MK	392	kmem_free(bulk, sizeof (sa_bulk_attr_t) * 20);
572e2857 BB	393	done:
	394	if (drop_lock)
	395	mutex_exit(&zp->z_lock);
428870ff BB	396	}
	397
	398	void
	399	zfs_sa_upgrade_txholds(dmu_tx_t tx, znode_t zp)
	400	{
3558fd73	401	if (!ZTOZSB(zp)->z_use_sa \|\| zp->z_is_sa)
428870ff BB	402	return;
428870ff BB	403
428870ff BB	404
	405	dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
	406
572e2857 BB	407	if (zfs_external_acl(zp)) {
572e2857 BB	408	dmu_tx_hold_free(tx, zfs_external_acl(zp), 0,
428870ff BB	409	DMU_OBJECT_END);
	410	}
	411	}
	412
e29be02e	413	EXPORT_SYMBOL(zfs_attr_table);
e45aa452 BB	414	EXPORT_SYMBOL(zfs_sa_readlink);
	415	EXPORT_SYMBOL(zfs_sa_symlink);
	416	EXPORT_SYMBOL(zfs_sa_get_scanstamp);
	417	EXPORT_SYMBOL(zfs_sa_set_scanstamp);
82a37189 BB	418	EXPORT_SYMBOL(zfs_sa_get_xattr);
82a37189 BB	419	EXPORT_SYMBOL(zfs_sa_set_xattr);
e45aa452 BB	420	EXPORT_SYMBOL(zfs_sa_upgrade);
	421	EXPORT_SYMBOL(zfs_sa_upgrade_txholds);
	422
428870ff	423	#endif