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.
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.
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]
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2013 by Delphix. All rights reserved.
24 * Copyright 2013 Saso Kiselkov. All rights reserved.
27 #include <sys/zfs_context.h>
29 #include <sys/spa_impl.h>
31 #include <sys/zio_checksum.h>
33 #include <zfs_fletcher.h>
38 * In the SPA, everything is checksummed. We support checksum vectors
39 * for three distinct reasons:
41 * 1. Different kinds of data need different levels of protection.
42 * For SPA metadata, we always want a very strong checksum.
43 * For user data, we let users make the trade-off between speed
44 * and checksum strength.
46 * 2. Cryptographic hash and MAC algorithms are an area of active research.
47 * It is likely that in future hash functions will be at least as strong
48 * as current best-of-breed, and may be substantially faster as well.
49 * We want the ability to take advantage of these new hashes as soon as
50 * they become available.
52 * 3. If someone develops hardware that can compute a strong hash quickly,
53 * we want the ability to take advantage of that hardware.
55 * Of course, we don't want a checksum upgrade to invalidate existing
56 * data, so we store the checksum *function* in eight bits of the bp.
57 * This gives us room for up to 256 different checksum functions.
59 * When writing a block, we always checksum it with the latest-and-greatest
60 * checksum function of the appropriate strength. When reading a block,
61 * we compare the expected checksum against the actual checksum, which we
62 * compute via the checksum function specified by BP_GET_CHECKSUM(bp).
66 * To enable the use of less secure hash algorithms with dedup, we
67 * introduce the notion of salted checksums (MACs, really). A salted
68 * checksum is fed both a random 256-bit value (the salt) and the data
69 * to be checksummed. This salt is kept secret (stored on the pool, but
70 * never shown to the user). Thus even if an attacker knew of collision
71 * weaknesses in the hash algorithm, they won't be able to mount a known
72 * plaintext attack on the DDT, since the actual hash value cannot be
73 * known ahead of time. How the salt is used is algorithm-specific
74 * (some might simply prefix it to the data block, others might need to
75 * utilize a full-blown HMAC). On disk the salt is stored in a ZAP
76 * object in the MOS (DMU_POOL_CHECKSUM_SALT).
80 * Some hashing algorithms need to perform a substantial amount of
81 * initialization work (e.g. salted checksums above may need to pre-hash
82 * the salt) before being able to process data. Performing this
83 * redundant work for each block would be wasteful, so we instead allow
84 * a checksum algorithm to do the work once (the first time it's used)
85 * and then keep this pre-initialized context as a template inside the
86 * spa_t (spa_cksum_tmpls). If the zio_checksum_info_t contains
87 * non-NULL ci_tmpl_init and ci_tmpl_free callbacks, they are used to
88 * construct and destruct the pre-initialized checksum context. The
89 * pre-initialized context is then reused during each checksum
90 * invocation and passed to the checksum function.
95 zio_checksum_off(const void *buf
, uint64_t size
,
96 const void *ctx_template
, zio_cksum_t
*zcp
)
98 ZIO_SET_CHECKSUM(zcp
, 0, 0, 0, 0);
101 zio_checksum_info_t zio_checksum_table
[ZIO_CHECKSUM_FUNCTIONS
] = {
102 {{NULL
, NULL
}, NULL
, NULL
, 0, "inherit"},
103 {{NULL
, NULL
}, NULL
, NULL
, 0, "on"},
104 {{zio_checksum_off
, zio_checksum_off
},
105 NULL
, NULL
, 0, "off"},
106 {{zio_checksum_SHA256
, zio_checksum_SHA256
},
107 NULL
, NULL
, ZCHECKSUM_FLAG_METADATA
| ZCHECKSUM_FLAG_EMBEDDED
,
109 {{zio_checksum_SHA256
, zio_checksum_SHA256
},
110 NULL
, NULL
, ZCHECKSUM_FLAG_METADATA
| ZCHECKSUM_FLAG_EMBEDDED
,
112 {{fletcher_2_native
, fletcher_2_byteswap
},
113 NULL
, NULL
, ZCHECKSUM_FLAG_EMBEDDED
, "zilog"},
114 {{fletcher_2_native
, fletcher_2_byteswap
},
115 NULL
, NULL
, 0, "fletcher2"},
116 {{fletcher_4_native
, fletcher_4_byteswap
},
117 NULL
, NULL
, ZCHECKSUM_FLAG_METADATA
, "fletcher4"},
118 {{zio_checksum_SHA256
, zio_checksum_SHA256
},
119 NULL
, NULL
, ZCHECKSUM_FLAG_METADATA
| ZCHECKSUM_FLAG_DEDUP
|
120 ZCHECKSUM_FLAG_NOPWRITE
, "sha256"},
121 {{fletcher_4_native
, fletcher_4_byteswap
},
122 NULL
, NULL
, ZCHECKSUM_FLAG_EMBEDDED
, "zilog2"},
123 {{zio_checksum_off
, zio_checksum_off
},
124 NULL
, NULL
, 0, "noparity"},
125 {{zio_checksum_SHA512_native
, zio_checksum_SHA512_byteswap
},
126 NULL
, NULL
, ZCHECKSUM_FLAG_METADATA
| ZCHECKSUM_FLAG_DEDUP
|
127 ZCHECKSUM_FLAG_NOPWRITE
, "sha512"},
128 {{zio_checksum_skein_native
, zio_checksum_skein_byteswap
},
129 zio_checksum_skein_tmpl_init
, zio_checksum_skein_tmpl_free
,
130 ZCHECKSUM_FLAG_METADATA
| ZCHECKSUM_FLAG_DEDUP
|
131 ZCHECKSUM_FLAG_SALTED
| ZCHECKSUM_FLAG_NOPWRITE
, "skein"},
132 {{zio_checksum_edonr_native
, zio_checksum_edonr_byteswap
},
133 zio_checksum_edonr_tmpl_init
, zio_checksum_edonr_tmpl_free
,
134 ZCHECKSUM_FLAG_METADATA
| ZCHECKSUM_FLAG_SALTED
|
135 ZCHECKSUM_FLAG_NOPWRITE
, "edonr"},
139 * The flag corresponding to the "verify" in dedup=[checksum,]verify
140 * must be cleared first, so callers should use ZIO_CHECKSUM_MASK.
143 zio_checksum_to_feature(enum zio_checksum cksum
)
145 VERIFY((cksum
& ~ZIO_CHECKSUM_MASK
) == 0);
148 case ZIO_CHECKSUM_SHA512
:
149 return (SPA_FEATURE_SHA512
);
150 case ZIO_CHECKSUM_SKEIN
:
151 return (SPA_FEATURE_SKEIN
);
152 case ZIO_CHECKSUM_EDONR
:
153 return (SPA_FEATURE_EDONR
);
155 return (SPA_FEATURE_NONE
);
160 zio_checksum_select(enum zio_checksum child
, enum zio_checksum parent
)
162 ASSERT(child
< ZIO_CHECKSUM_FUNCTIONS
);
163 ASSERT(parent
< ZIO_CHECKSUM_FUNCTIONS
);
164 ASSERT(parent
!= ZIO_CHECKSUM_INHERIT
&& parent
!= ZIO_CHECKSUM_ON
);
166 if (child
== ZIO_CHECKSUM_INHERIT
)
169 if (child
== ZIO_CHECKSUM_ON
)
170 return (ZIO_CHECKSUM_ON_VALUE
);
176 zio_checksum_dedup_select(spa_t
*spa
, enum zio_checksum child
,
177 enum zio_checksum parent
)
179 ASSERT((child
& ZIO_CHECKSUM_MASK
) < ZIO_CHECKSUM_FUNCTIONS
);
180 ASSERT((parent
& ZIO_CHECKSUM_MASK
) < ZIO_CHECKSUM_FUNCTIONS
);
181 ASSERT(parent
!= ZIO_CHECKSUM_INHERIT
&& parent
!= ZIO_CHECKSUM_ON
);
183 if (child
== ZIO_CHECKSUM_INHERIT
)
186 if (child
== ZIO_CHECKSUM_ON
)
187 return (spa_dedup_checksum(spa
));
189 if (child
== (ZIO_CHECKSUM_ON
| ZIO_CHECKSUM_VERIFY
))
190 return (spa_dedup_checksum(spa
) | ZIO_CHECKSUM_VERIFY
);
192 ASSERT((zio_checksum_table
[child
& ZIO_CHECKSUM_MASK
].ci_flags
&
193 ZCHECKSUM_FLAG_DEDUP
) ||
194 (child
& ZIO_CHECKSUM_VERIFY
) || child
== ZIO_CHECKSUM_OFF
);
200 * Set the external verifier for a gang block based on <vdev, offset, txg>,
201 * a tuple which is guaranteed to be unique for the life of the pool.
204 zio_checksum_gang_verifier(zio_cksum_t
*zcp
, blkptr_t
*bp
)
206 const dva_t
*dva
= BP_IDENTITY(bp
);
207 uint64_t txg
= BP_PHYSICAL_BIRTH(bp
);
209 ASSERT(BP_IS_GANG(bp
));
211 ZIO_SET_CHECKSUM(zcp
, DVA_GET_VDEV(dva
), DVA_GET_OFFSET(dva
), txg
, 0);
215 * Set the external verifier for a label block based on its offset.
216 * The vdev is implicit, and the txg is unknowable at pool open time --
217 * hence the logic in vdev_uberblock_load() to find the most recent copy.
220 zio_checksum_label_verifier(zio_cksum_t
*zcp
, uint64_t offset
)
222 ZIO_SET_CHECKSUM(zcp
, offset
, 0, 0, 0);
226 * Calls the template init function of a checksum which supports context
227 * templates and installs the template into the spa_t.
230 zio_checksum_template_init(enum zio_checksum checksum
, spa_t
*spa
)
232 zio_checksum_info_t
*ci
= &zio_checksum_table
[checksum
];
234 if (ci
->ci_tmpl_init
== NULL
)
236 if (spa
->spa_cksum_tmpls
[checksum
] != NULL
)
239 VERIFY(ci
->ci_tmpl_free
!= NULL
);
240 mutex_enter(&spa
->spa_cksum_tmpls_lock
);
241 if (spa
->spa_cksum_tmpls
[checksum
] == NULL
) {
242 spa
->spa_cksum_tmpls
[checksum
] =
243 ci
->ci_tmpl_init(&spa
->spa_cksum_salt
);
244 VERIFY(spa
->spa_cksum_tmpls
[checksum
] != NULL
);
246 mutex_exit(&spa
->spa_cksum_tmpls_lock
);
250 * Generate the checksum.
253 zio_checksum_compute(zio_t
*zio
, enum zio_checksum checksum
,
254 void *data
, uint64_t size
)
256 blkptr_t
*bp
= zio
->io_bp
;
257 uint64_t offset
= zio
->io_offset
;
258 zio_checksum_info_t
*ci
= &zio_checksum_table
[checksum
];
260 spa_t
*spa
= zio
->io_spa
;
262 ASSERT((uint_t
)checksum
< ZIO_CHECKSUM_FUNCTIONS
);
263 ASSERT(ci
->ci_func
[0] != NULL
);
265 zio_checksum_template_init(checksum
, spa
);
267 if (ci
->ci_flags
& ZCHECKSUM_FLAG_EMBEDDED
) {
270 if (checksum
== ZIO_CHECKSUM_ZILOG2
) {
271 zil_chain_t
*zilc
= data
;
273 size
= P2ROUNDUP_TYPED(zilc
->zc_nused
, ZIL_MIN_BLKSZ
,
277 eck
= (zio_eck_t
*)((char *)data
+ size
) - 1;
279 if (checksum
== ZIO_CHECKSUM_GANG_HEADER
)
280 zio_checksum_gang_verifier(&eck
->zec_cksum
, bp
);
281 else if (checksum
== ZIO_CHECKSUM_LABEL
)
282 zio_checksum_label_verifier(&eck
->zec_cksum
, offset
);
284 bp
->blk_cksum
= eck
->zec_cksum
;
285 eck
->zec_magic
= ZEC_MAGIC
;
286 ci
->ci_func
[0](data
, size
, spa
->spa_cksum_tmpls
[checksum
],
288 eck
->zec_cksum
= cksum
;
290 ci
->ci_func
[0](data
, size
, spa
->spa_cksum_tmpls
[checksum
],
296 zio_checksum_error_impl(spa_t
*spa
, blkptr_t
*bp
, enum zio_checksum checksum
,
297 void *data
, uint64_t size
, uint64_t offset
, zio_bad_cksum_t
*info
)
299 zio_checksum_info_t
*ci
= &zio_checksum_table
[checksum
];
301 zio_cksum_t actual_cksum
, expected_cksum
;
303 if (checksum
>= ZIO_CHECKSUM_FUNCTIONS
|| ci
->ci_func
[0] == NULL
)
304 return (SET_ERROR(EINVAL
));
306 zio_checksum_template_init(checksum
, spa
);
308 if (ci
->ci_flags
& ZCHECKSUM_FLAG_EMBEDDED
) {
310 zio_cksum_t verifier
;
312 if (checksum
== ZIO_CHECKSUM_ZILOG2
) {
313 zil_chain_t
*zilc
= data
;
317 if (eck
->zec_magic
== ZEC_MAGIC
)
318 nused
= zilc
->zc_nused
;
319 else if (eck
->zec_magic
== BSWAP_64(ZEC_MAGIC
))
320 nused
= BSWAP_64(zilc
->zc_nused
);
322 return (SET_ERROR(ECKSUM
));
325 return (SET_ERROR(ECKSUM
));
327 size
= P2ROUNDUP_TYPED(nused
, ZIL_MIN_BLKSZ
, uint64_t);
329 eck
= (zio_eck_t
*)((char *)data
+ size
) - 1;
332 if (checksum
== ZIO_CHECKSUM_GANG_HEADER
)
333 zio_checksum_gang_verifier(&verifier
, bp
);
334 else if (checksum
== ZIO_CHECKSUM_LABEL
)
335 zio_checksum_label_verifier(&verifier
, offset
);
337 verifier
= bp
->blk_cksum
;
339 byteswap
= (eck
->zec_magic
== BSWAP_64(ZEC_MAGIC
));
342 byteswap_uint64_array(&verifier
, sizeof (zio_cksum_t
));
344 expected_cksum
= eck
->zec_cksum
;
345 eck
->zec_cksum
= verifier
;
346 ci
->ci_func
[byteswap
](data
, size
,
347 spa
->spa_cksum_tmpls
[checksum
], &actual_cksum
);
348 eck
->zec_cksum
= expected_cksum
;
351 byteswap_uint64_array(&expected_cksum
,
352 sizeof (zio_cksum_t
));
355 byteswap
= BP_SHOULD_BYTESWAP(bp
);
356 expected_cksum
= bp
->blk_cksum
;
357 ci
->ci_func
[byteswap
](data
, size
,
358 spa
->spa_cksum_tmpls
[checksum
], &actual_cksum
);
362 info
->zbc_expected
= expected_cksum
;
363 info
->zbc_actual
= actual_cksum
;
364 info
->zbc_checksum_name
= ci
->ci_name
;
365 info
->zbc_byteswapped
= byteswap
;
366 info
->zbc_injected
= 0;
367 info
->zbc_has_cksum
= 1;
370 if (!ZIO_CHECKSUM_EQUAL(actual_cksum
, expected_cksum
))
371 return (SET_ERROR(ECKSUM
));
377 zio_checksum_error(zio_t
*zio
, zio_bad_cksum_t
*info
)
379 blkptr_t
*bp
= zio
->io_bp
;
380 uint_t checksum
= (bp
== NULL
? zio
->io_prop
.zp_checksum
:
381 (BP_IS_GANG(bp
) ? ZIO_CHECKSUM_GANG_HEADER
: BP_GET_CHECKSUM(bp
)));
383 uint64_t size
= (bp
== NULL
? zio
->io_size
:
384 (BP_IS_GANG(bp
) ? SPA_GANGBLOCKSIZE
: BP_GET_PSIZE(bp
)));
385 uint64_t offset
= zio
->io_offset
;
386 void *data
= zio
->io_data
;
387 spa_t
*spa
= zio
->io_spa
;
389 error
= zio_checksum_error_impl(spa
, bp
, checksum
, data
, size
,
391 if (error
!= 0 && zio_injection_enabled
&& !zio
->io_error
&&
392 (error
= zio_handle_fault_injection(zio
, ECKSUM
)) != 0) {
394 info
->zbc_injected
= 1;
401 * Called by a spa_t that's about to be deallocated. This steps through
402 * all of the checksum context templates and deallocates any that were
403 * initialized using the algorithm-specific template init function.
406 zio_checksum_templates_free(spa_t
*spa
)
408 enum zio_checksum checksum
;
409 for (checksum
= 0; checksum
< ZIO_CHECKSUM_FUNCTIONS
;
411 if (spa
->spa_cksum_tmpls
[checksum
] != NULL
) {
412 zio_checksum_info_t
*ci
= &zio_checksum_table
[checksum
];
414 VERIFY(ci
->ci_tmpl_free
!= NULL
);
415 ci
->ci_tmpl_free(spa
->spa_cksum_tmpls
[checksum
]);
416 spa
->spa_cksum_tmpls
[checksum
] = NULL
;