[mirror_zfs-debian.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)
{
	zfsvfs_t *zfsvfs = 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(zfsvfs),
		    &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)
{
	zfsvfs_t *zfsvfs = 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(zfsvfs),
		    &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(zfsvfs),
		    &zp->z_pflags, sizeof (uint64_t), tx));
	}
}

int
zfs_sa_get_xattr(znode_t *zp)
{
	zfsvfs_t *zfsvfs = 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(zfsvfs), &size);
	if (error) {
		if (error == ENOENT)
			return nvlist_alloc(&zp->z_xattr_cached,
			    NV_UNIQUE_NAME, KM_SLEEP);
		else
			return (error);
	}

	obj = vmem_alloc(size, KM_SLEEP);

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

	vmem_free(obj, size);

	return (error);
}

int
zfs_sa_set_xattr(znode_t *zp)
{
	zfsvfs_t *zfsvfs = 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 = vmem_alloc(size, KM_SLEEP);

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

	tx = dmu_tx_create(zfsvfs->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 {
		int count = 0;
		sa_bulk_attr_t bulk[2];
		uint64_t ctime[2];

		zfs_tstamp_update_setup(zp, STATE_CHANGED, NULL, ctime);
		SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DXATTR(zfsvfs),
		    NULL, obj, size);
		SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs),
		    NULL, &ctime, 16);
		VERIFY0(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx));

		dmu_tx_commit(tx);
	}
out_free:
	vmem_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);
	zfsvfs_t *zfsvfs = 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];
	uint64_t links;
	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(zfsvfs), NULL, &atime, 16);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CRTIME(zfsvfs), NULL, &crtime, 16);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_XATTR(zfsvfs), NULL, &xattr, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_RDEV(zfsvfs), NULL, &rdev, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, &uid, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, &gid, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, &tmp_gen, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zfsvfs), 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(zfsvfs), NULL, &mode, 8);
	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_SIZE(zfsvfs), NULL,
	    &zp->z_size, 8);
	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_GEN(zfsvfs),
	    NULL, &tmp_gen, 8);
	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_UID(zfsvfs), NULL, &uid, 8);
	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_GID(zfsvfs), NULL, &gid, 8);
	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_PARENT(zfsvfs),
	    NULL, &parent, 8);
	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_FLAGS(zfsvfs), NULL,
	    &zp->z_pflags, 8);
	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_ATIME(zfsvfs), NULL,
	    &atime, 16);
	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_MTIME(zfsvfs), NULL,
	    &mtime, 16);
	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_CTIME(zfsvfs), NULL,
	    &ctime, 16);
	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_CRTIME(zfsvfs), NULL,
	    &crtime, 16);
	links = ZTOI(zp)->i_nlink;
	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_LINKS(zfsvfs), NULL,
	    &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(zfsvfs), NULL,
		    &rdev, 8);
	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_DACL_COUNT(zfsvfs), 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(zfsvfs),
	    zfs_acl_data_locator, &locate, zp->z_acl_cached->z_acl_bytes);

	if (xattr)
		SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_XATTR(zfsvfs),
		    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(zfsvfs),
		    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(zfsvfs->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
e10b0808	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()) {
cae5b340	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);
cae5b340 AX	109	VERIFY0(dmu_buf_hold(ZTOZSB(zp)->z_os, zp->z_id, 0, FTAG, &dbp,
cae5b340 AX	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	{
cae5b340	123	zfsvfs_t *zfsvfs = 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) {
cae5b340	129	if (sa_lookup(zp->z_sa_hdl, SA_ZPL_SCANSTAMP(zfsvfs),
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	{
cae5b340	157	zfsvfs_t *zfsvfs = 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)
cae5b340	163	VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SCANSTAMP(zfsvfs),
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;
cae5b340	180	VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs),
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	{
cae5b340	188	zfsvfs_t *zfsvfs = ZTOZSB(zp);
82a37189 BB	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
cae5b340	197	error = sa_size(zp->z_sa_hdl, SA_ZPL_DXATTR(zfsvfs), &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
cae5b340	206	obj = vmem_alloc(size, KM_SLEEP);
82a37189	207
cae5b340	208	error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DXATTR(zfsvfs), obj, size);
82a37189 BB	209	if (error == 0)
	210	error = nvlist_unpack(obj, size, &zp->z_xattr_cached, KM_SLEEP);
	211
cae5b340	212	vmem_free(obj, size);
82a37189 BB	213
	214	return (error);
	215	}
	216
	217	int
	218	zfs_sa_set_xattr(znode_t *zp)
	219	{
cae5b340	220	zfsvfs_t *zfsvfs = ZTOZSB(zp);
82a37189 BB	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);
4d815aed AX	231	if ((error == 0) && (size > SA_ATTR_MAX_LEN))
4d815aed AX	232	error = EFBIG;
82a37189 BB	233	if (error)
	234	goto out;
	235
cae5b340	236	obj = vmem_alloc(size, KM_SLEEP);
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
cae5b340	243	tx = dmu_tx_create(zfsvfs->z_os);
82a37189	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 {
cae5b340 AX	251	int count = 0;
	252	sa_bulk_attr_t bulk[2];
	253	uint64_t ctime[2];
	254
	255	zfs_tstamp_update_setup(zp, STATE_CHANGED, NULL, ctime);
	256	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_DXATTR(zfsvfs),
	257	NULL, obj, size);
	258	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs),
	259	NULL, &ctime, 16);
	260	VERIFY0(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx));
	261
4d815aed	262	dmu_tx_commit(tx);
82a37189 BB	263	}
82a37189 BB	264	out_free:
cae5b340	265	vmem_free(obj, size);
82a37189 BB	266	out:
	267	return (error);
	268	}
	269
428870ff BB	270	/*
	271	* I'm not convinced we should do any of this upgrade.
	272	* since the SA code can read both old/new znode formats
a08ee875	273	* with probably little to no performance difference.
428870ff BB	274	*
	275	* All new files will be created with the new format.
	276	*/
	277
	278	void
	279	zfs_sa_upgrade(sa_handle_t hdl, dmu_tx_t tx)
	280	{
	281	dmu_buf_t *db = sa_get_db(hdl);
	282	znode_t *zp = sa_get_userdata(hdl);
cae5b340	283	zfsvfs_t *zfsvfs = ZTOZSB(zp);
428870ff	284	int count = 0;
1ee1b767	285	sa_bulk_attr_t bulk, sa_attrs;
428870ff	286	zfs_acl_locator_cb_t locate = { 0 };
cae5b340	287	uint64_t uid, gid, mode, rdev, xattr, parent, tmp_gen;
68d83c55	288	uint64_t crtime[2], mtime[2], ctime[2], atime[2];
cae5b340	289	uint64_t links;
428870ff BB	290	zfs_acl_phys_t znode_acl;
428870ff BB	291	char scanstamp[AV_SCANSTAMP_SZ];
572e2857	292	boolean_t drop_lock = B_FALSE;
428870ff BB	293
	294	/*
	295	* No upgrade if ACL isn't cached
	296	* since we won't know which locks are held
	297	* and ready the ACL would require special "locked"
	298	* interfaces that would be messy
	299	*/
3558fd73	300	if (zp->z_acl_cached == NULL \|\| S_ISLNK(ZTOI(zp)->i_mode))
428870ff BB	301	return;
428870ff BB	302
572e2857 BB	303	/*
	304	* If the z_lock is held and we aren't the owner
	305	* the just return since we don't want to deadlock
	306	* trying to update the status of z_is_sa. This
	307	* file can then be upgraded at a later time.
	308	*
	309	* Otherwise, we know we are doing the
	310	* sa_update() that caused us to enter this function.
	311	*/
	312	if (mutex_owner(&zp->z_lock) != curthread) {
	313	if (mutex_tryenter(&zp->z_lock) == 0)
	314	return;
	315	else
	316	drop_lock = B_TRUE;
	317	}
	318
428870ff	319	/* First do a bulk query of the attributes that aren't cached */
a08ee875	320	bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * 20, KM_SLEEP);
cae5b340 AX	321	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16);
	322	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16);
	323	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16);
	324	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CRTIME(zfsvfs), NULL, &crtime, 16);
	325	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
	326	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8);
	327	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_XATTR(zfsvfs), NULL, &xattr, 8);
	328	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_RDEV(zfsvfs), NULL, &rdev, 8);
	329	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, &uid, 8);
	330	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, &gid, 8);
	331	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, &tmp_gen, 8);
	332	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
428870ff BB	333	&znode_acl, 88);
428870ff BB	334
1ee1b767	335	if (sa_bulk_lookup_locked(hdl, bulk, count) != 0) {
a08ee875	336	kmem_free(bulk, sizeof (sa_bulk_attr_t) * 20);
572e2857	337	goto done;
1ee1b767	338	}
428870ff BB	339
	340	/*
	341	* While the order here doesn't matter its best to try and organize
	342	* it is such a way to pick up an already existing layout number
	343	*/
	344	count = 0;
a08ee875	345	sa_attrs = kmem_zalloc(sizeof (sa_bulk_attr_t) * 20, KM_SLEEP);
cae5b340 AX	346	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
cae5b340 AX	347	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_SIZE(zfsvfs), NULL,
428870ff	348	&zp->z_size, 8);
cae5b340 AX	349	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_GEN(zfsvfs),
	350	NULL, &tmp_gen, 8);
	351	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_UID(zfsvfs), NULL, &uid, 8);
	352	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_GID(zfsvfs), NULL, &gid, 8);
	353	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_PARENT(zfsvfs),
428870ff	354	NULL, &parent, 8);
cae5b340	355	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_FLAGS(zfsvfs), NULL,
428870ff	356	&zp->z_pflags, 8);
cae5b340	357	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_ATIME(zfsvfs), NULL,
68d83c55	358	&atime, 16);
cae5b340	359	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_MTIME(zfsvfs), NULL,
428870ff	360	&mtime, 16);
cae5b340	361	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_CTIME(zfsvfs), NULL,
428870ff	362	&ctime, 16);
cae5b340	363	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_CRTIME(zfsvfs), NULL,
428870ff	364	&crtime, 16);
cae5b340 AX	365	links = ZTOI(zp)->i_nlink;
	366	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_LINKS(zfsvfs), NULL,
	367	&links, 8);
3558fd73	368	if (S_ISBLK(ZTOI(zp)->i_mode) \|\| S_ISCHR(ZTOI(zp)->i_mode))
cae5b340	369	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_RDEV(zfsvfs), NULL,
428870ff	370	&rdev, 8);
cae5b340	371	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_DACL_COUNT(zfsvfs), NULL,
428870ff BB	372	&zp->z_acl_cached->z_acl_count, 8);
	373
	374	if (zp->z_acl_cached->z_version < ZFS_ACL_VERSION_FUID)
	375	zfs_acl_xform(zp, zp->z_acl_cached, CRED());
	376
	377	locate.cb_aclp = zp->z_acl_cached;
cae5b340	378	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_DACL_ACES(zfsvfs),
428870ff	379	zfs_acl_data_locator, &locate, zp->z_acl_cached->z_acl_bytes);
572e2857	380
428870ff	381	if (xattr)
cae5b340	382	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_XATTR(zfsvfs),
572e2857	383	NULL, &xattr, 8);
428870ff BB	384
	385	/* if scanstamp then add scanstamp */
	386
	387	if (zp->z_pflags & ZFS_BONUS_SCANSTAMP) {
	388	bcopy((caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
	389	scanstamp, AV_SCANSTAMP_SZ);
cae5b340	390	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_SCANSTAMP(zfsvfs),
428870ff BB	391	NULL, scanstamp, AV_SCANSTAMP_SZ);
	392	zp->z_pflags &= ~ZFS_BONUS_SCANSTAMP;
	393	}
	394
	395	VERIFY(dmu_set_bonustype(db, DMU_OT_SA, tx) == 0);
	396	VERIFY(sa_replace_all_by_template_locked(hdl, sa_attrs,
	397	count, tx) == 0);
	398	if (znode_acl.z_acl_extern_obj)
cae5b340	399	VERIFY(0 == dmu_object_free(zfsvfs->z_os,
428870ff BB	400	znode_acl.z_acl_extern_obj, tx));
	401
	402	zp->z_is_sa = B_TRUE;
a08ee875 LG	403	kmem_free(sa_attrs, sizeof (sa_bulk_attr_t) * 20);
a08ee875 LG	404	kmem_free(bulk, sizeof (sa_bulk_attr_t) * 20);
572e2857 BB	405	done:
	406	if (drop_lock)
	407	mutex_exit(&zp->z_lock);
428870ff BB	408	}
	409
	410	void
	411	zfs_sa_upgrade_txholds(dmu_tx_t tx, znode_t zp)
	412	{
3558fd73	413	if (!ZTOZSB(zp)->z_use_sa \|\| zp->z_is_sa)
428870ff BB	414	return;
428870ff BB	415
428870ff BB	416
	417	dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
	418
572e2857 BB	419	if (zfs_external_acl(zp)) {
572e2857 BB	420	dmu_tx_hold_free(tx, zfs_external_acl(zp), 0,
428870ff BB	421	DMU_OBJECT_END);
	422	}
	423	}
	424
e29be02e	425	EXPORT_SYMBOL(zfs_attr_table);
e45aa452 BB	426	EXPORT_SYMBOL(zfs_sa_readlink);
	427	EXPORT_SYMBOL(zfs_sa_symlink);
	428	EXPORT_SYMBOL(zfs_sa_get_scanstamp);
	429	EXPORT_SYMBOL(zfs_sa_set_scanstamp);
82a37189 BB	430	EXPORT_SYMBOL(zfs_sa_get_xattr);
82a37189 BB	431	EXPORT_SYMBOL(zfs_sa_set_xattr);
e45aa452 BB	432	EXPORT_SYMBOL(zfs_sa_upgrade);
	433	EXPORT_SYMBOL(zfs_sa_upgrade_txholds);
	434
428870ff	435	#endif