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) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 */
24
25 #include <sys/zfs_context.h>
26 #include <sys/spa.h>
27 #include <sys/zio.h>
28 #include <sys/zio_checksum.h>
29 #include <sys/zil.h>
30 #include <zfs_fletcher.h>
31
32 /*
33 * Checksum vectors.
34 *
35 * In the SPA, everything is checksummed. We support checksum vectors
36 * for three distinct reasons:
37 *
38 * 1. Different kinds of data need different levels of protection.
39 * For SPA metadata, we always want a very strong checksum.
40 * For user data, we let users make the trade-off between speed
41 * and checksum strength.
42 *
43 * 2. Cryptographic hash and MAC algorithms are an area of active research.
44 * It is likely that in future hash functions will be at least as strong
45 * as current best-of-breed, and may be substantially faster as well.
46 * We want the ability to take advantage of these new hashes as soon as
47 * they become available.
48 *
49 * 3. If someone develops hardware that can compute a strong hash quickly,
50 * we want the ability to take advantage of that hardware.
51 *
52 * Of course, we don't want a checksum upgrade to invalidate existing
53 * data, so we store the checksum *function* in eight bits of the bp.
54 * This gives us room for up to 256 different checksum functions.
55 *
56 * When writing a block, we always checksum it with the latest-and-greatest
57 * checksum function of the appropriate strength. When reading a block,
58 * we compare the expected checksum against the actual checksum, which we
59 * compute via the checksum function specified by BP_GET_CHECKSUM(bp).
60 */
61
62 /*ARGSUSED*/
63 static void
64 zio_checksum_off(const void *buf, uint64_t size, zio_cksum_t *zcp)
65 {
66 ZIO_SET_CHECKSUM(zcp, 0, 0, 0, 0);
67 }
68
69 zio_checksum_info_t zio_checksum_table[ZIO_CHECKSUM_FUNCTIONS] = {
70 {{NULL, NULL}, 0, 0, 0, "inherit"},
71 {{NULL, NULL}, 0, 0, 0, "on"},
72 {{zio_checksum_off, zio_checksum_off}, 0, 0, 0, "off"},
73 {{zio_checksum_SHA256, zio_checksum_SHA256}, 1, 1, 0, "label"},
74 {{zio_checksum_SHA256, zio_checksum_SHA256}, 1, 1, 0, "gang_header"},
75 {{fletcher_2_native, fletcher_2_byteswap}, 0, 1, 0, "zilog"},
76 {{fletcher_2_native, fletcher_2_byteswap}, 0, 0, 0, "fletcher2"},
77 {{fletcher_4_native, fletcher_4_byteswap}, 1, 0, 0, "fletcher4"},
78 {{zio_checksum_SHA256, zio_checksum_SHA256}, 1, 0, 1, "sha256"},
79 {{fletcher_4_native, fletcher_4_byteswap}, 0, 1, 0, "zilog2"},
80 {{zio_checksum_EdonR512_256, zio_checksum_EdonR512_256_byteswap},
81 1, 0, 1, "edonr512/256"},
82 };
83
84 enum zio_checksum
85 zio_checksum_select(enum zio_checksum child, enum zio_checksum parent)
86 {
87 ASSERT(child < ZIO_CHECKSUM_FUNCTIONS);
88 ASSERT(parent < ZIO_CHECKSUM_FUNCTIONS);
89 ASSERT(parent != ZIO_CHECKSUM_INHERIT && parent != ZIO_CHECKSUM_ON);
90
91 if (child == ZIO_CHECKSUM_INHERIT)
92 return (parent);
93
94 if (child == ZIO_CHECKSUM_ON)
95 return (ZIO_CHECKSUM_ON_VALUE);
96
97 return (child);
98 }
99
100 enum zio_checksum
101 zio_checksum_dedup_select(spa_t *spa, enum zio_checksum child,
102 enum zio_checksum parent)
103 {
104 ASSERT((child & ZIO_CHECKSUM_MASK) < ZIO_CHECKSUM_FUNCTIONS);
105 ASSERT((parent & ZIO_CHECKSUM_MASK) < ZIO_CHECKSUM_FUNCTIONS);
106 ASSERT(parent != ZIO_CHECKSUM_INHERIT && parent != ZIO_CHECKSUM_ON);
107
108 if (child == ZIO_CHECKSUM_INHERIT)
109 return (parent);
110
111 if (child == ZIO_CHECKSUM_ON)
112 return (spa_dedup_checksum(spa));
113
114 if (child == (ZIO_CHECKSUM_ON | ZIO_CHECKSUM_VERIFY))
115 return (spa_dedup_checksum(spa) | ZIO_CHECKSUM_VERIFY);
116
117 ASSERT(zio_checksum_table[child & ZIO_CHECKSUM_MASK].ci_dedup ||
118 (child & ZIO_CHECKSUM_VERIFY) || child == ZIO_CHECKSUM_OFF);
119
120 return (child);
121 }
122
123 /*
124 * Set the external verifier for a gang block based on <vdev, offset, txg>,
125 * a tuple which is guaranteed to be unique for the life of the pool.
126 */
127 static void
128 zio_checksum_gang_verifier(zio_cksum_t *zcp, blkptr_t *bp)
129 {
130 dva_t *dva = BP_IDENTITY(bp);
131 uint64_t txg = BP_PHYSICAL_BIRTH(bp);
132
133 ASSERT(BP_IS_GANG(bp));
134
135 ZIO_SET_CHECKSUM(zcp, DVA_GET_VDEV(dva), DVA_GET_OFFSET(dva), txg, 0);
136 }
137
138 /*
139 * Set the external verifier for a label block based on its offset.
140 * The vdev is implicit, and the txg is unknowable at pool open time --
141 * hence the logic in vdev_uberblock_load() to find the most recent copy.
142 */
143 static void
144 zio_checksum_label_verifier(zio_cksum_t *zcp, uint64_t offset)
145 {
146 ZIO_SET_CHECKSUM(zcp, offset, 0, 0, 0);
147 }
148
149 /*
150 * Generate the checksum.
151 */
152 void
153 zio_checksum_compute(zio_t *zio, enum zio_checksum checksum,
154 void *data, uint64_t size)
155 {
156 blkptr_t *bp = zio->io_bp;
157 uint64_t offset = zio->io_offset;
158 zio_checksum_info_t *ci = &zio_checksum_table[checksum];
159 zio_cksum_t cksum;
160
161 ASSERT((uint_t)checksum < ZIO_CHECKSUM_FUNCTIONS);
162 ASSERT(ci->ci_func[0] != NULL);
163
164 if (ci->ci_eck) {
165 zio_eck_t *eck;
166
167 if (checksum == ZIO_CHECKSUM_ZILOG2) {
168 zil_chain_t *zilc = data;
169
170 size = P2ROUNDUP_TYPED(zilc->zc_nused, ZIL_MIN_BLKSZ,
171 uint64_t);
172 eck = &zilc->zc_eck;
173 } else {
174 eck = (zio_eck_t *)((char *)data + size) - 1;
175 }
176 if (checksum == ZIO_CHECKSUM_GANG_HEADER)
177 zio_checksum_gang_verifier(&eck->zec_cksum, bp);
178 else if (checksum == ZIO_CHECKSUM_LABEL)
179 zio_checksum_label_verifier(&eck->zec_cksum, offset);
180 else
181 bp->blk_cksum = eck->zec_cksum;
182 eck->zec_magic = ZEC_MAGIC;
183 ci->ci_func[0](data, size, &cksum);
184 eck->zec_cksum = cksum;
185 } else {
186 ci->ci_func[0](data, size, &bp->blk_cksum);
187 }
188 }
189
190 int
191 zio_checksum_error(zio_t *zio, zio_bad_cksum_t *info)
192 {
193 blkptr_t *bp = zio->io_bp;
194 uint_t checksum = (bp == NULL ? zio->io_prop.zp_checksum :
195 (BP_IS_GANG(bp) ? ZIO_CHECKSUM_GANG_HEADER : BP_GET_CHECKSUM(bp)));
196 int byteswap;
197 int error;
198 uint64_t size = (bp == NULL ? zio->io_size :
199 (BP_IS_GANG(bp) ? SPA_GANGBLOCKSIZE : BP_GET_PSIZE(bp)));
200 uint64_t offset = zio->io_offset;
201 void *data = zio->io_data;
202 zio_checksum_info_t *ci = &zio_checksum_table[checksum];
203 zio_cksum_t actual_cksum, expected_cksum, verifier;
204
205 if (checksum >= ZIO_CHECKSUM_FUNCTIONS || ci->ci_func[0] == NULL)
206 return (EINVAL);
207
208 if (ci->ci_eck) {
209 zio_eck_t *eck;
210
211 if (checksum == ZIO_CHECKSUM_ZILOG2) {
212 zil_chain_t *zilc = data;
213 uint64_t nused;
214
215 eck = &zilc->zc_eck;
216 if (eck->zec_magic == ZEC_MAGIC)
217 nused = zilc->zc_nused;
218 else if (eck->zec_magic == BSWAP_64(ZEC_MAGIC))
219 nused = BSWAP_64(zilc->zc_nused);
220 else
221 return (ECKSUM);
222
223 if (nused > size)
224 return (ECKSUM);
225
226 size = P2ROUNDUP_TYPED(nused, ZIL_MIN_BLKSZ, uint64_t);
227 } else {
228 eck = (zio_eck_t *)((char *)data + size) - 1;
229 }
230
231 if (checksum == ZIO_CHECKSUM_GANG_HEADER)
232 zio_checksum_gang_verifier(&verifier, bp);
233 else if (checksum == ZIO_CHECKSUM_LABEL)
234 zio_checksum_label_verifier(&verifier, offset);
235 else
236 verifier = bp->blk_cksum;
237
238 byteswap = (eck->zec_magic == BSWAP_64(ZEC_MAGIC));
239
240 if (byteswap)
241 byteswap_uint64_array(&verifier, sizeof (zio_cksum_t));
242
243 expected_cksum = eck->zec_cksum;
244 eck->zec_cksum = verifier;
245 ci->ci_func[byteswap](data, size, &actual_cksum);
246 eck->zec_cksum = expected_cksum;
247
248 if (byteswap)
249 byteswap_uint64_array(&expected_cksum,
250 sizeof (zio_cksum_t));
251 } else {
252 ASSERT(!BP_IS_GANG(bp));
253 byteswap = BP_SHOULD_BYTESWAP(bp);
254 expected_cksum = bp->blk_cksum;
255 ci->ci_func[byteswap](data, size, &actual_cksum);
256 }
257
258 info->zbc_expected = expected_cksum;
259 info->zbc_actual = actual_cksum;
260 info->zbc_checksum_name = ci->ci_name;
261 info->zbc_byteswapped = byteswap;
262 info->zbc_injected = 0;
263 info->zbc_has_cksum = 1;
264
265 if (!ZIO_CHECKSUM_EQUAL(actual_cksum, expected_cksum))
266 return (ECKSUM);
267
268 if (zio_injection_enabled && !zio->io_error &&
269 (error = zio_handle_fault_injection(zio, ECKSUM)) != 0) {
270
271 info->zbc_injected = 1;
272 return (error);
273 }
274
275 return (0);
276 }