Home | History | Annotate | Download | only in zfs
      1 /* SPDX-License-Identifier: GPL-2.0+ */
      2 /*
      3  *  GRUB  --  GRand Unified Bootloader
      4  *  Copyright (C) 1999,2000,2001,2002,2003,2004  Free Software Foundation, Inc.
      5  */
      6 /*
      7  * Copyright (c) 2008, 2011, Oracle and/or its affiliates. All rights reserved.
      8  */
      9 
     10 #ifndef ZFS_SPA_HEADER
     11 #define	ZFS_SPA_HEADER 1
     12 
     13 
     14 /*
     15  * General-purpose 32-bit and 64-bit bitfield encodings.
     16  */
     17 #define	BF32_DECODE(x, low, len)	P2PHASE((x) >> (low), 1U << (len))
     18 #define	BF64_DECODE(x, low, len)	P2PHASE((x) >> (low), 1ULL << (len))
     19 #define	BF32_ENCODE(x, low, len)	(P2PHASE((x), 1U << (len)) << (low))
     20 #define	BF64_ENCODE(x, low, len)	(P2PHASE((x), 1ULL << (len)) << (low))
     21 
     22 #define	BF32_GET(x, low, len)		BF32_DECODE(x, low, len)
     23 #define	BF64_GET(x, low, len)		BF64_DECODE(x, low, len)
     24 
     25 #define	BF32_SET(x, low, len, val)						\
     26 	((x) ^= BF32_ENCODE((x >> low) ^ (val), low, len))
     27 #define	BF64_SET(x, low, len, val)						\
     28 	((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len))
     29 
     30 #define	BF32_GET_SB(x, low, len, shift, bias)		\
     31 	((BF32_GET(x, low, len) + (bias)) << (shift))
     32 #define	BF64_GET_SB(x, low, len, shift, bias)		\
     33 	((BF64_GET(x, low, len) + (bias)) << (shift))
     34 
     35 #define	BF32_SET_SB(x, low, len, shift, bias, val)		\
     36 	BF32_SET(x, low, len, ((val) >> (shift)) - (bias))
     37 #define	BF64_SET_SB(x, low, len, shift, bias, val)		\
     38 	BF64_SET(x, low, len, ((val) >> (shift)) - (bias))
     39 
     40 /*
     41  * We currently support nine block sizes, from 512 bytes to 128K.
     42  * We could go higher, but the benefits are near-zero and the cost
     43  * of COWing a giant block to modify one byte would become excessive.
     44  */
     45 #define	SPA_MINBLOCKSHIFT	9
     46 #define	SPA_MAXBLOCKSHIFT	17
     47 #define	SPA_MINBLOCKSIZE	(1ULL << SPA_MINBLOCKSHIFT)
     48 #define	SPA_MAXBLOCKSIZE	(1ULL << SPA_MAXBLOCKSHIFT)
     49 
     50 #define	SPA_BLOCKSIZES		(SPA_MAXBLOCKSHIFT - SPA_MINBLOCKSHIFT + 1)
     51 
     52 /*
     53  * Size of block to hold the configuration data (a packed nvlist)
     54  */
     55 #define	SPA_CONFIG_BLOCKSIZE	(1 << 14)
     56 
     57 /*
     58  * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB.
     59  * The ASIZE encoding should be at least 64 times larger (6 more bits)
     60  * to support up to 4-way RAID-Z mirror mode with worst-case gang block
     61  * overhead, three DVAs per bp, plus one more bit in case we do anything
     62  * else that expands the ASIZE.
     63  */
     64 #define	SPA_LSIZEBITS		16	/* LSIZE up to 32M (2^16 * 512)	*/
     65 #define	SPA_PSIZEBITS		16	/* PSIZE up to 32M (2^16 * 512)	*/
     66 #define	SPA_ASIZEBITS		24	/* ASIZE up to 64 times larger	*/
     67 
     68 /*
     69  * All SPA data is represented by 128-bit data virtual addresses (DVAs).
     70  * The members of the dva_t should be considered opaque outside the SPA.
     71  */
     72 typedef struct dva {
     73 	uint64_t	dva_word[2];
     74 } dva_t;
     75 
     76 /*
     77  * Each block has a 256-bit checksum -- strong enough for cryptographic hashes.
     78  */
     79 typedef struct zio_cksum {
     80 	uint64_t	zc_word[4];
     81 } zio_cksum_t;
     82 
     83 /*
     84  * Each block is described by its DVAs, time of birth, checksum, etc.
     85  * The word-by-word, bit-by-bit layout of the blkptr is as follows:
     86  *
     87  *	64	56	48	40	32	24	16	8	0
     88  *	+-------+-------+-------+-------+-------+-------+-------+-------+
     89  * 0	|		vdev1		| GRID	|	  ASIZE		|
     90  *	+-------+-------+-------+-------+-------+-------+-------+-------+
     91  * 1	|G|			 offset1				|
     92  *	+-------+-------+-------+-------+-------+-------+-------+-------+
     93  * 2	|		vdev2		| GRID	|	  ASIZE		|
     94  *	+-------+-------+-------+-------+-------+-------+-------+-------+
     95  * 3	|G|			 offset2				|
     96  *	+-------+-------+-------+-------+-------+-------+-------+-------+
     97  * 4	|		vdev3		| GRID	|	  ASIZE		|
     98  *	+-------+-------+-------+-------+-------+-------+-------+-------+
     99  * 5	|G|			 offset3				|
    100  *	+-------+-------+-------+-------+-------+-------+-------+-------+
    101  * 6	|BDX|lvl| type	| cksum | comp	|	  PSIZE	|	  LSIZE	|
    102  *	+-------+-------+-------+-------+-------+-------+-------+-------+
    103  * 7	|			padding					|
    104  *	+-------+-------+-------+-------+-------+-------+-------+-------+
    105  * 8	|			padding					|
    106  *	+-------+-------+-------+-------+-------+-------+-------+-------+
    107  * 9	|			physical birth txg			|
    108  *	+-------+-------+-------+-------+-------+-------+-------+-------+
    109  * a	|			logical birth txg			|
    110  *	+-------+-------+-------+-------+-------+-------+-------+-------+
    111  * b	|			fill count				|
    112  *	+-------+-------+-------+-------+-------+-------+-------+-------+
    113  * c	|			checksum[0]				|
    114  *	+-------+-------+-------+-------+-------+-------+-------+-------+
    115  * d	|			checksum[1]				|
    116  *	+-------+-------+-------+-------+-------+-------+-------+-------+
    117  * e	|			checksum[2]				|
    118  *	+-------+-------+-------+-------+-------+-------+-------+-------+
    119  * f	|			checksum[3]				|
    120  *	+-------+-------+-------+-------+-------+-------+-------+-------+
    121  *
    122  * Legend:
    123  *
    124  * vdev		virtual device ID
    125  * offset	offset into virtual device
    126  * LSIZE	logical size
    127  * PSIZE	physical size (after compression)
    128  * ASIZE	allocated size (including RAID-Z parity and gang block headers)
    129  * GRID		RAID-Z layout information (reserved for future use)
    130  * cksum	checksum function
    131  * comp		compression function
    132  * G		gang block indicator
    133  * B		byteorder (endianness)
    134  * D		dedup
    135  * X		unused
    136  * lvl		level of indirection
    137  * type		DMU object type
    138  * phys birth	txg of block allocation; zero if same as logical birth txg
    139  * log. birth	transaction group in which the block was logically born
    140  * fill count	number of non-zero blocks under this bp
    141  * checksum[4]	256-bit checksum of the data this bp describes
    142  */
    143 #define	SPA_BLKPTRSHIFT	7		/* blkptr_t is 128 bytes	*/
    144 #define	SPA_DVAS_PER_BP	3		/* Number of DVAs in a bp	*/
    145 
    146 typedef struct blkptr {
    147 	dva_t		blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */
    148 	uint64_t	blk_prop;	/* size, compression, type, etc		*/
    149 	uint64_t	blk_pad[2];	/* Extra space for the future		*/
    150 	uint64_t	blk_phys_birth;	/* txg when block was allocated		*/
    151 	uint64_t	blk_birth;	/* transaction group at birth		*/
    152 	uint64_t	blk_fill;	/* fill count				*/
    153 	zio_cksum_t	blk_cksum;	/* 256-bit checksum			*/
    154 } blkptr_t;
    155 
    156 /*
    157  * Macros to get and set fields in a bp or DVA.
    158  */
    159 #define	DVA_GET_ASIZE(dva)											\
    160 	BF64_GET_SB((dva)->dva_word[0], 0, 24, SPA_MINBLOCKSHIFT, 0)
    161 #define	DVA_SET_ASIZE(dva, x)										\
    162 	BF64_SET_SB((dva)->dva_word[0], 0, 24, SPA_MINBLOCKSHIFT, 0, x)
    163 
    164 #define	DVA_GET_GRID(dva)	BF64_GET((dva)->dva_word[0], 24, 8)
    165 #define	DVA_SET_GRID(dva, x)	BF64_SET((dva)->dva_word[0], 24, 8, x)
    166 
    167 #define	DVA_GET_VDEV(dva)	BF64_GET((dva)->dva_word[0], 32, 32)
    168 #define	DVA_SET_VDEV(dva, x)	BF64_SET((dva)->dva_word[0], 32, 32, x)
    169 
    170 #define	DVA_GET_GANG(dva)	BF64_GET((dva)->dva_word[1], 63, 1)
    171 #define	DVA_SET_GANG(dva, x)	BF64_SET((dva)->dva_word[1], 63, 1, x)
    172 
    173 #define	BP_GET_LSIZE(bp)										\
    174 	BF64_GET_SB((bp)->blk_prop, 0, 16, SPA_MINBLOCKSHIFT, 1)
    175 #define	BP_SET_LSIZE(bp, x)										\
    176 	BF64_SET_SB((bp)->blk_prop, 0, 16, SPA_MINBLOCKSHIFT, 1, x)
    177 
    178 #define	BP_GET_COMPRESS(bp)		BF64_GET((bp)->blk_prop, 32, 8)
    179 #define	BP_SET_COMPRESS(bp, x)		BF64_SET((bp)->blk_prop, 32, 8, x)
    180 
    181 #define	BP_GET_CHECKSUM(bp)		BF64_GET((bp)->blk_prop, 40, 8)
    182 #define	BP_SET_CHECKSUM(bp, x)		BF64_SET((bp)->blk_prop, 40, 8, x)
    183 
    184 #define	BP_GET_TYPE(bp)			BF64_GET((bp)->blk_prop, 48, 8)
    185 #define	BP_SET_TYPE(bp, x)		BF64_SET((bp)->blk_prop, 48, 8, x)
    186 
    187 #define	BP_GET_LEVEL(bp)		BF64_GET((bp)->blk_prop, 56, 5)
    188 #define	BP_SET_LEVEL(bp, x)		BF64_SET((bp)->blk_prop, 56, 5, x)
    189 
    190 #define	BP_GET_PROP_BIT_61(bp)		BF64_GET((bp)->blk_prop, 61, 1)
    191 #define	BP_SET_PROP_BIT_61(bp, x)	BF64_SET((bp)->blk_prop, 61, 1, x)
    192 
    193 #define	BP_GET_DEDUP(bp)		BF64_GET((bp)->blk_prop, 62, 1)
    194 #define	BP_SET_DEDUP(bp, x)		BF64_SET((bp)->blk_prop, 62, 1, x)
    195 
    196 #define	BP_GET_BYTEORDER(bp)		(0 - BF64_GET((bp)->blk_prop, 63, 1))
    197 #define	BP_SET_BYTEORDER(bp, x)		BF64_SET((bp)->blk_prop, 63, 1, x)
    198 
    199 #define	BP_PHYSICAL_BIRTH(bp)										\
    200 	((bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth)
    201 
    202 #define	BP_SET_BIRTH(bp, logical, physical)								\
    203 	{																	\
    204 		(bp)->blk_birth = (logical);									\
    205 		(bp)->blk_phys_birth = ((logical) == (physical) ? 0 : (physical)); \
    206 	}
    207 
    208 #define	BP_GET_ASIZE(bp)												\
    209 	(DVA_GET_ASIZE(&(bp)->blk_dva[0]) + DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \
    210 	 DVA_GET_ASIZE(&(bp)->blk_dva[2]))
    211 
    212 #define	BP_GET_UCSIZE(bp)												\
    213 	((BP_GET_LEVEL(bp) > 0 || dmu_ot[BP_GET_TYPE(bp)].ot_metadata) ?	\
    214 	 BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp));
    215 
    216 #define	BP_GET_NDVAS(bp)						\
    217 	(!!DVA_GET_ASIZE(&(bp)->blk_dva[0]) +		\
    218 	 !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) +		\
    219 	 !!DVA_GET_ASIZE(&(bp)->blk_dva[2]))
    220 
    221 #define	BP_COUNT_GANG(bp)						\
    222 	(DVA_GET_GANG(&(bp)->blk_dva[0]) +			\
    223 	 DVA_GET_GANG(&(bp)->blk_dva[1]) +			\
    224 	 DVA_GET_GANG(&(bp)->blk_dva[2]))
    225 
    226 #define	DVA_EQUAL(dva1, dva2)						\
    227 	((dva1)->dva_word[1] == (dva2)->dva_word[1] &&	\
    228 	 (dva1)->dva_word[0] == (dva2)->dva_word[0])
    229 
    230 #define	BP_EQUAL(bp1, bp2)									\
    231 	(BP_PHYSICAL_BIRTH(bp1) == BP_PHYSICAL_BIRTH(bp2) &&	\
    232 	 DVA_EQUAL(&(bp1)->blk_dva[0], &(bp2)->blk_dva[0]) &&	\
    233 	 DVA_EQUAL(&(bp1)->blk_dva[1], &(bp2)->blk_dva[1]) &&	\
    234 	 DVA_EQUAL(&(bp1)->blk_dva[2], &(bp2)->blk_dva[2]))
    235 
    236 #define	ZIO_CHECKSUM_EQUAL(zc1, zc2)				\
    237 	(0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) |	\
    238 		   ((zc1).zc_word[1] - (zc2).zc_word[1]) |	\
    239 		   ((zc1).zc_word[2] - (zc2).zc_word[2]) |	\
    240 		   ((zc1).zc_word[3] - (zc2).zc_word[3])))
    241 
    242 #define	DVA_IS_VALID(dva)	(DVA_GET_ASIZE(dva) != 0)
    243 
    244 #define	ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3)	\
    245 	{											\
    246 		(zcp)->zc_word[0] = w0;					\
    247 		(zcp)->zc_word[1] = w1;					\
    248 		(zcp)->zc_word[2] = w2;					\
    249 		(zcp)->zc_word[3] = w3;					\
    250 	}
    251 
    252 #define	BP_IDENTITY(bp)		(&(bp)->blk_dva[0])
    253 #define	BP_IS_GANG(bp)		DVA_GET_GANG(BP_IDENTITY(bp))
    254 #define	BP_IS_HOLE(bp)		((bp)->blk_birth == 0)
    255 
    256 /* BP_IS_RAIDZ(bp) assumes no block compression */
    257 #define	BP_IS_RAIDZ(bp)		(DVA_GET_ASIZE(&(bp)->blk_dva[0]) > \
    258 							 BP_GET_PSIZE(bp))
    259 
    260 #define	BP_ZERO(bp)										\
    261 	{													\
    262 		(bp)->blk_dva[0].dva_word[0] = 0;				\
    263 		(bp)->blk_dva[0].dva_word[1] = 0;				\
    264 		(bp)->blk_dva[1].dva_word[0] = 0;				\
    265 		(bp)->blk_dva[1].dva_word[1] = 0;				\
    266 		(bp)->blk_dva[2].dva_word[0] = 0;				\
    267 		(bp)->blk_dva[2].dva_word[1] = 0;				\
    268 		(bp)->blk_prop = 0;								\
    269 		(bp)->blk_pad[0] = 0;							\
    270 		(bp)->blk_pad[1] = 0;							\
    271 		(bp)->blk_phys_birth = 0;						\
    272 		(bp)->blk_birth = 0;							\
    273 		(bp)->blk_fill = 0;								\
    274 		ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0);	\
    275 	}
    276 
    277 #define	BP_SPRINTF_LEN	320
    278 
    279 #endif	/* ! ZFS_SPA_HEADER */
    280