[mirror_ubuntu-artful-kernel.git] / zfs / 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()) {
		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));
	}
}

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;
	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_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,
	    &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
87d546d8 TG	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/zfs_context.h>
	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},
65	{"ZPL_DXATTR", 0, SA_UINT8_ARRAY, 0},
66	{NULL, 0, 0, 0}
67	};
68
69	#ifdef _KERNEL
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;
84	if ((error = dmu_buf_hold(ZTOZSB(zp)->z_os, zp->z_id,
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);
110	VERIFY(0 == dmu_buf_hold(ZTOZSB(zp)->z_os,
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	{
124	zfs_sb_t *zsb = ZTOZSB(zp);
125	xoptattr_t *xoap;
126
127	ASSERT(MUTEX_HELD(&zp->z_lock));
128	VERIFY((xoap = xva_getxoptattr(xvap)) != NULL);
129	if (zp->z_is_sa) {
130	if (sa_lookup(zp->z_sa_hdl, SA_ZPL_SCANSTAMP(zsb),
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	{
158	zfs_sb_t *zsb = ZTOZSB(zp);
159	xoptattr_t *xoap;
160
161	ASSERT(MUTEX_HELD(&zp->z_lock));
162	VERIFY((xoap = xva_getxoptattr(xvap)) != NULL);
163	if (zp->z_is_sa)
164	VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SCANSTAMP(zsb),
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;
181	VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_FLAGS(zsb),
182	&zp->z_pflags, sizeof (uint64_t), tx));
183	}
184	}
185
186	int
187	zfs_sa_get_xattr(znode_t *zp)
188	{
189	zfs_sb_t *zsb = ZTOZSB(zp);
190	char *obj;
191	int size;
192	int error;
193
194	ASSERT(RW_LOCK_HELD(&zp->z_xattr_lock));
195	ASSERT(!zp->z_xattr_cached);
196	ASSERT(zp->z_is_sa);
197
198	error = sa_size(zp->z_sa_hdl, SA_ZPL_DXATTR(zsb), &size);
199	if (error) {
200	if (error == ENOENT)
201	return nvlist_alloc(&zp->z_xattr_cached,
202	NV_UNIQUE_NAME, KM_SLEEP);
203	else
204	return (error);
205	}
206
207	obj = zio_buf_alloc(size);
208
209	error = sa_lookup(zp->z_sa_hdl, SA_ZPL_DXATTR(zsb), obj, size);
210	if (error == 0)
211	error = nvlist_unpack(obj, size, &zp->z_xattr_cached, KM_SLEEP);
212
213	zio_buf_free(obj, size);
214
215	return (error);
216	}
217
218	int
219	zfs_sa_set_xattr(znode_t *zp)
220	{
221	zfs_sb_t *zsb = ZTOZSB(zp);
222	dmu_tx_t *tx;
223	char *obj;
224	size_t size;
225	int error;
226
227	ASSERT(RW_WRITE_HELD(&zp->z_xattr_lock));
228	ASSERT(zp->z_xattr_cached);
229	ASSERT(zp->z_is_sa);
230
231	error = nvlist_size(zp->z_xattr_cached, &size, NV_ENCODE_XDR);
232	if ((error == 0) && (size > SA_ATTR_MAX_LEN))
233	error = EFBIG;
234	if (error)
235	goto out;
236
237	obj = zio_buf_alloc(size);
238
239	error = nvlist_pack(zp->z_xattr_cached, &obj, &size,
240	NV_ENCODE_XDR, KM_SLEEP);
241	if (error)
242	goto out_free;
243
244	tx = dmu_tx_create(zsb->z_os);
245	dmu_tx_hold_sa_create(tx, size);
246	dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
247
248	error = dmu_tx_assign(tx, TXG_WAIT);
249	if (error) {
250	dmu_tx_abort(tx);
251	} else {
252	VERIFY0(sa_update(zp->z_sa_hdl, SA_ZPL_DXATTR(zsb),
253	obj, size, tx));
254	dmu_tx_commit(tx);
255	}
256	out_free:
257	zio_buf_free(obj, size);
258	out:
259	return (error);
260	}
261
262	/*
263	* I'm not convinced we should do any of this upgrade.
264	* since the SA code can read both old/new znode formats
265	* with probably little to no performance difference.
266	*
267	* All new files will be created with the new format.
268	*/
269
270	void
271	zfs_sa_upgrade(sa_handle_t hdl, dmu_tx_t tx)
272	{
273	dmu_buf_t *db = sa_get_db(hdl);
274	znode_t *zp = sa_get_userdata(hdl);
275	zfs_sb_t *zsb = ZTOZSB(zp);
276	int count = 0;
277	sa_bulk_attr_t bulk, sa_attrs;
278	zfs_acl_locator_cb_t locate = { 0 };
279	uint64_t uid, gid, mode, rdev, xattr, parent;
280	uint64_t crtime[2], mtime[2], ctime[2], atime[2];
281	zfs_acl_phys_t znode_acl;
282	char scanstamp[AV_SCANSTAMP_SZ];
283	boolean_t drop_lock = B_FALSE;
284
285	/*
286	* No upgrade if ACL isn't cached
287	* since we won't know which locks are held
288	* and ready the ACL would require special "locked"
289	* interfaces that would be messy
290	*/
291	if (zp->z_acl_cached == NULL \|\| S_ISLNK(ZTOI(zp)->i_mode))
292	return;
293
294	/*
295	* If the z_lock is held and we aren't the owner
296	* the just return since we don't want to deadlock
297	* trying to update the status of z_is_sa. This
298	* file can then be upgraded at a later time.
299	*
300	* Otherwise, we know we are doing the
301	* sa_update() that caused us to enter this function.
302	*/
303	if (mutex_owner(&zp->z_lock) != curthread) {
304	if (mutex_tryenter(&zp->z_lock) == 0)
305	return;
306	else
307	drop_lock = B_TRUE;
308	}
309
310	/* First do a bulk query of the attributes that aren't cached */
311	bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * 20, KM_SLEEP);
312	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL, &atime, 16);
313	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, &mtime, 16);
314	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, &ctime, 16);
315	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CRTIME(zsb), NULL, &crtime, 16);
316	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, &mode, 8);
317	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zsb), NULL, &parent, 8);
318	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_XATTR(zsb), NULL, &xattr, 8);
319	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_RDEV(zsb), NULL, &rdev, 8);
320	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, &uid, 8);
321	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL, &gid, 8);
322	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ZNODE_ACL(zsb), NULL,
323	&znode_acl, 88);
324
325	if (sa_bulk_lookup_locked(hdl, bulk, count) != 0) {
326	kmem_free(bulk, sizeof (sa_bulk_attr_t) * 20);
327	goto done;
328	}
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;
335	sa_attrs = kmem_zalloc(sizeof (sa_bulk_attr_t) * 20, KM_SLEEP);
336	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_MODE(zsb), NULL, &mode, 8);
337	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_SIZE(zsb), NULL,
338	&zp->z_size, 8);
339	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_GEN(zsb),
340	NULL, &zp->z_gen, 8);
341	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_UID(zsb), NULL, &uid, 8);
342	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_GID(zsb), NULL, &gid, 8);
343	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_PARENT(zsb),
344	NULL, &parent, 8);
345	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_FLAGS(zsb), NULL,
346	&zp->z_pflags, 8);
347	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_ATIME(zsb), NULL,
348	&atime, 16);
349	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_MTIME(zsb), NULL,
350	&mtime, 16);
351	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_CTIME(zsb), NULL,
352	&ctime, 16);
353	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_CRTIME(zsb), NULL,
354	&crtime, 16);
355	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_LINKS(zsb), NULL,
356	&zp->z_links, 8);
357	if (S_ISBLK(ZTOI(zp)->i_mode) \|\| S_ISCHR(ZTOI(zp)->i_mode))
358	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_RDEV(zsb), NULL,
359	&rdev, 8);
360	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_DACL_COUNT(zsb), NULL,
361	&zp->z_acl_cached->z_acl_count, 8);
362
363	if (zp->z_acl_cached->z_version < ZFS_ACL_VERSION_FUID)
364	zfs_acl_xform(zp, zp->z_acl_cached, CRED());
365
366	locate.cb_aclp = zp->z_acl_cached;
367	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_DACL_ACES(zsb),
368	zfs_acl_data_locator, &locate, zp->z_acl_cached->z_acl_bytes);
369
370	if (xattr)
371	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_XATTR(zsb),
372	NULL, &xattr, 8);
373
374	/* if scanstamp then add scanstamp */
375
376	if (zp->z_pflags & ZFS_BONUS_SCANSTAMP) {
377	bcopy((caddr_t)db->db_data + ZFS_OLD_ZNODE_PHYS_SIZE,
378	scanstamp, AV_SCANSTAMP_SZ);
379	SA_ADD_BULK_ATTR(sa_attrs, count, SA_ZPL_SCANSTAMP(zsb),
380	NULL, scanstamp, AV_SCANSTAMP_SZ);
381	zp->z_pflags &= ~ZFS_BONUS_SCANSTAMP;
382	}
383
384	VERIFY(dmu_set_bonustype(db, DMU_OT_SA, tx) == 0);
385	VERIFY(sa_replace_all_by_template_locked(hdl, sa_attrs,
386	count, tx) == 0);
387	if (znode_acl.z_acl_extern_obj)
388	VERIFY(0 == dmu_object_free(zsb->z_os,
389	znode_acl.z_acl_extern_obj, tx));
390
391	zp->z_is_sa = B_TRUE;
392	kmem_free(sa_attrs, sizeof (sa_bulk_attr_t) * 20);
393	kmem_free(bulk, sizeof (sa_bulk_attr_t) * 20);
394	done:
395	if (drop_lock)
396	mutex_exit(&zp->z_lock);
397	}
398
399	void
400	zfs_sa_upgrade_txholds(dmu_tx_t tx, znode_t zp)
401	{
402	if (!ZTOZSB(zp)->z_use_sa \|\| zp->z_is_sa)
403	return;
404
405
406	dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
407
408	if (zfs_external_acl(zp)) {
409	dmu_tx_hold_free(tx, zfs_external_acl(zp), 0,
410	DMU_OBJECT_END);
411	}
412	}
413
414	EXPORT_SYMBOL(zfs_attr_table);
415	EXPORT_SYMBOL(zfs_sa_readlink);
416	EXPORT_SYMBOL(zfs_sa_symlink);
417	EXPORT_SYMBOL(zfs_sa_get_scanstamp);
418	EXPORT_SYMBOL(zfs_sa_set_scanstamp);
419	EXPORT_SYMBOL(zfs_sa_get_xattr);
420	EXPORT_SYMBOL(zfs_sa_set_xattr);
421	EXPORT_SYMBOL(zfs_sa_upgrade);
422	EXPORT_SYMBOL(zfs_sa_upgrade_txholds);
423
424	#endif