xref: /illumos-gate/usr/src/uts/common/fs/zfs/sys/sa_impl.h (revision e914ace2)
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  * Copyright (c) 2013 by Delphix. All rights reserved.
24  * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
25  */
26 
27 #ifndef	_SYS_SA_IMPL_H
28 #define	_SYS_SA_IMPL_H
29 
30 #include <sys/dmu.h>
31 #include <sys/refcount.h>
32 #include <sys/list.h>
33 
34 /*
35  * Array of known attributes and their
36  * various characteristics.
37  */
38 typedef struct sa_attr_table {
39 	sa_attr_type_t	sa_attr;
40 	uint8_t sa_registered;
41 	uint16_t sa_length;
42 	sa_bswap_type_t sa_byteswap;
43 	char *sa_name;
44 } sa_attr_table_t;
45 
46 /*
47  * Zap attribute format for attribute registration
48  *
49  * 64      56      48      40      32      24      16      8       0
50  * +-------+-------+-------+-------+-------+-------+-------+-------+
51  * |        unused         |      len      | bswap |   attr num    |
52  * +-------+-------+-------+-------+-------+-------+-------+-------+
53  *
54  * Zap attribute format for layout information.
55  *
56  * layout information is stored as an array of attribute numbers
57  * The name of the attribute is the layout number (0, 1, 2, ...)
58  *
59  * 16       0
60  * +---- ---+
61  * | attr # |
62  * +--------+
63  * | attr # |
64  * +--- ----+
65  *  ......
66  *
67  */
68 
69 #define	ATTR_BSWAP(x)	BF32_GET(x, 16, 8)
70 #define	ATTR_LENGTH(x)	BF32_GET(x, 24, 16)
71 #define	ATTR_NUM(x)	BF32_GET(x, 0, 16)
72 #define	ATTR_ENCODE(x, attr, length, bswap) \
73 { \
74 	BF64_SET(x, 24, 16, length); \
75 	BF64_SET(x, 16, 8, bswap); \
76 	BF64_SET(x, 0, 16, attr); \
77 }
78 
79 #define	TOC_OFF(x)		BF32_GET(x, 0, 23)
80 #define	TOC_ATTR_PRESENT(x)	BF32_GET(x, 31, 1)
81 #define	TOC_LEN_IDX(x)		BF32_GET(x, 24, 4)
82 #define	TOC_ATTR_ENCODE(x, len_idx, offset) \
83 { \
84 	BF32_SET(x, 31, 1, 1); \
85 	BF32_SET(x, 24, 7, len_idx); \
86 	BF32_SET(x, 0, 24, offset); \
87 }
88 
89 #define	SA_LAYOUTS	"LAYOUTS"
90 #define	SA_REGISTRY	"REGISTRY"
91 
92 /*
93  * Each unique layout will have their own table
94  * sa_lot (layout_table)
95  */
96 typedef struct sa_lot {
97 	avl_node_t lot_num_node;
98 	avl_node_t lot_hash_node;
99 	uint64_t lot_num;
100 	uint64_t lot_hash;
101 	sa_attr_type_t *lot_attrs;	/* array of attr #'s */
102 	uint32_t lot_var_sizes;	/* how many aren't fixed size */
103 	uint32_t lot_attr_count;	/* total attr count */
104 	list_t	lot_idx_tab;	/* should be only a couple of entries */
105 	int	lot_instance;	/* used with lot_hash to identify entry */
106 } sa_lot_t;
107 
108 /* index table of offsets */
109 typedef struct sa_idx_tab {
110 	list_node_t	sa_next;
111 	sa_lot_t	*sa_layout;
112 	uint16_t	*sa_variable_lengths;
113 	zfs_refcount_t	sa_refcount;
114 	uint32_t	*sa_idx_tab;	/* array of offsets */
115 } sa_idx_tab_t;
116 
117 /*
118  * Since the offset/index information into the actual data
119  * will usually be identical we can share that information with
120  * all handles that have the exact same offsets.
121  *
122  * You would typically only have a large number of different table of
123  * contents if you had a several variable sized attributes.
124  *
125  * Two AVL trees are used to track the attribute layout numbers.
126  * one is keyed by number and will be consulted when a DMU_OT_SA
127  * object is first read.  The second tree is keyed by the hash signature
128  * of the attributes and will be consulted when an attribute is added
129  * to determine if we already have an instance of that layout.  Both
130  * of these tree's are interconnected.  The only difference is that
131  * when an entry is found in the "hash" tree the list of attributes will
132  * need to be compared against the list of attributes you have in hand.
133  * The assumption is that typically attributes will just be updated and
134  * adding a completely new attribute is a very rare operation.
135  */
136 struct sa_os {
137 	kmutex_t	sa_lock;
138 	boolean_t	sa_need_attr_registration;
139 	boolean_t	sa_force_spill;
140 	uint64_t	sa_master_obj;
141 	uint64_t	sa_reg_attr_obj;
142 	uint64_t	sa_layout_attr_obj;
143 	int		sa_num_attrs;
144 	sa_attr_table_t *sa_attr_table;	 /* private attr table */
145 	sa_update_cb_t	*sa_update_cb;
146 	avl_tree_t	sa_layout_num_tree;  /* keyed by layout number */
147 	avl_tree_t	sa_layout_hash_tree; /* keyed by layout hash value */
148 	int		sa_user_table_sz;
149 	sa_attr_type_t	*sa_user_table; /* user name->attr mapping table */
150 };
151 
152 /*
153  * header for all bonus and spill buffers.
154  *
155  * The header has a fixed portion with a variable number
156  * of "lengths" depending on the number of variable sized
157  * attributes which are determined by the "layout number"
158  */
159 
160 #define	SA_MAGIC	0x2F505A  /* ZFS SA */
161 typedef struct sa_hdr_phys {
162 	uint32_t sa_magic;
163 	/* BEGIN CSTYLED */
164 	/*
165 	 * Encoded with hdrsize and layout number as follows:
166 	 * 16      10       0
167 	 * +--------+-------+
168 	 * | hdrsz  |layout |
169 	 * +--------+-------+
170 	 *
171 	 * Bits 0-10 are the layout number
172 	 * Bits 11-16 are the size of the header.
173 	 * The hdrsize is the number * 8
174 	 *
175 	 * For example.
176 	 * hdrsz of 1 ==> 8 byte header
177 	 *          2 ==> 16 byte header
178 	 *
179 	 */
180 	/* END CSTYLED */
181 	uint16_t sa_layout_info;
182 	uint16_t sa_lengths[1];	/* optional sizes for variable length attrs */
183 	/* ... Data follows the lengths.  */
184 } sa_hdr_phys_t;
185 
186 #define	SA_HDR_LAYOUT_NUM(hdr) BF32_GET(hdr->sa_layout_info, 0, 10)
187 #define	SA_HDR_SIZE(hdr) BF32_GET_SB(hdr->sa_layout_info, 10, 6, 3, 0)
188 #define	SA_HDR_LAYOUT_INFO_ENCODE(x, num, size) \
189 { \
190 	BF32_SET_SB(x, 10, 6, 3, 0, size); \
191 	BF32_SET(x, 0, 10, num); \
192 }
193 
194 typedef enum sa_buf_type {
195 	SA_BONUS = 1,
196 	SA_SPILL = 2
197 } sa_buf_type_t;
198 
199 typedef enum sa_data_op {
200 	SA_LOOKUP,
201 	SA_UPDATE,
202 	SA_ADD,
203 	SA_REPLACE,
204 	SA_REMOVE
205 } sa_data_op_t;
206 
207 /*
208  * Opaque handle used for most sa functions
209  *
210  * This needs to be kept as small as possible.
211  */
212 
213 struct sa_handle {
214 	dmu_buf_user_t	sa_dbu;
215 	kmutex_t	sa_lock;
216 	dmu_buf_t	*sa_bonus;
217 	dmu_buf_t	*sa_spill;
218 	objset_t	*sa_os;
219 	void		*sa_userp;
220 	sa_idx_tab_t	*sa_bonus_tab;	 /* idx of bonus */
221 	sa_idx_tab_t	*sa_spill_tab; /* only present if spill activated */
222 };
223 
224 #define	SA_GET_DB(hdl, type)	\
225 	(dmu_buf_impl_t *)((type == SA_BONUS) ? hdl->sa_bonus : hdl->sa_spill)
226 
227 #define	SA_GET_HDR(hdl, type) \
228 	((sa_hdr_phys_t *)((dmu_buf_impl_t *)(SA_GET_DB(hdl, \
229 	type))->db.db_data))
230 
231 #define	SA_IDX_TAB_GET(hdl, type) \
232 	(type == SA_BONUS ? hdl->sa_bonus_tab : hdl->sa_spill_tab)
233 
234 #define	IS_SA_BONUSTYPE(a)	\
235 	((a == DMU_OT_SA) ? B_TRUE : B_FALSE)
236 
237 #define	SA_BONUSTYPE_FROM_DB(db) \
238 	(dmu_get_bonustype((dmu_buf_t *)db))
239 
240 #define	SA_BLKPTR_SPACE	(DN_OLD_MAX_BONUSLEN - sizeof (blkptr_t))
241 
242 #define	SA_LAYOUT_NUM(x, type) \
243 	((!IS_SA_BONUSTYPE(type) ? 0 : (((IS_SA_BONUSTYPE(type)) && \
244 	((SA_HDR_LAYOUT_NUM(x)) == 0)) ? 1 : SA_HDR_LAYOUT_NUM(x))))
245 
246 
247 #define	SA_REGISTERED_LEN(sa, attr) sa->sa_attr_table[attr].sa_length
248 
249 #define	SA_ATTR_LEN(sa, idx, attr, hdr) ((SA_REGISTERED_LEN(sa, attr) == 0) ?\
250 	hdr->sa_lengths[TOC_LEN_IDX(idx->sa_idx_tab[attr])] : \
251 	SA_REGISTERED_LEN(sa, attr))
252 
253 #define	SA_SET_HDR(hdr, num, size) \
254 	{ \
255 		hdr->sa_magic = SA_MAGIC; \
256 		SA_HDR_LAYOUT_INFO_ENCODE(hdr->sa_layout_info, num, size); \
257 	}
258 
259 #define	SA_ATTR_INFO(sa, idx, hdr, attr, bulk, type, hdl) \
260 	{ \
261 		bulk.sa_size = SA_ATTR_LEN(sa, idx, attr, hdr); \
262 		bulk.sa_buftype = type; \
263 		bulk.sa_addr = \
264 		    (void *)((uintptr_t)TOC_OFF(idx->sa_idx_tab[attr]) + \
265 		    (uintptr_t)hdr); \
266 }
267 
268 #define	SA_HDR_SIZE_MATCH_LAYOUT(hdr, tb) \
269 	(SA_HDR_SIZE(hdr) == (sizeof (sa_hdr_phys_t) + \
270 	(tb->lot_var_sizes > 1 ? P2ROUNDUP((tb->lot_var_sizes - 1) * \
271 	sizeof (uint16_t), 8) : 0)))
272 
273 int sa_add_impl(sa_handle_t *, sa_attr_type_t,
274     uint32_t, sa_data_locator_t, void *, dmu_tx_t *);
275 
276 void sa_register_update_callback_locked(objset_t *, sa_update_cb_t *);
277 int sa_size_locked(sa_handle_t *, sa_attr_type_t, int *);
278 
279 void sa_default_locator(void **, uint32_t *, uint32_t, boolean_t, void *);
280 int sa_attr_size(sa_os_t *, sa_idx_tab_t *, sa_attr_type_t,
281     uint16_t *, sa_hdr_phys_t *);
282 
283 #ifdef	__cplusplus
284 extern "C" {
285 #endif
286 
287 #ifdef	__cplusplus
288 }
289 #endif
290 
291 #endif	/* _SYS_SA_IMPL_H */
292