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 * Copyright (c) 2013 by Delphix. All rights reserved.
24 * Copyright (c) 2013, Joyent, Inc. All rights reserved.
25 */
26
27 #include <sys/zfs_context.h>
28 #include <sys/spa.h>
29 #include <sys/zio.h>
30 #include <sys/zio_checksum.h>
31 #include <sys/zil.h>
32 #include <zfs_fletcher.h>
33
34 /*
35 * Checksum vectors.
36 *
37 * In the SPA, everything is checksummed. We support checksum vectors
38 * for three distinct reasons:
39 *
40 * 1. Different kinds of data need different levels of protection.
41 * For SPA metadata, we always want a very strong checksum.
42 * For user data, we let users make the trade-off between speed
43 * and checksum strength.
44 *
45 * 2. Cryptographic hash and MAC algorithms are an area of active research.
46 * It is likely that in future hash functions will be at least as strong
47 * as current best-of-breed, and may be substantially faster as well.
48 * We want the ability to take advantage of these new hashes as soon as
49 * they become available.
50 *
51 * 3. If someone develops hardware that can compute a strong hash quickly,
52 * we want the ability to take advantage of that hardware.
53 *
54 * Of course, we don't want a checksum upgrade to invalidate existing
55 * data, so we store the checksum *function* in eight bits of the bp.
56 * This gives us room for up to 256 different checksum functions.
57 *
58 * When writing a block, we always checksum it with the latest-and-greatest
59 * checksum function of the appropriate strength. When reading a block,
60 * we compare the expected checksum against the actual checksum, which we
61 * compute via the checksum function specified by BP_GET_CHECKSUM(bp).
62 */
63
64 /*ARGSUSED*/
65 static void
66 zio_checksum_off(const void *buf, uint64_t size, zio_cksum_t *zcp)
67 {
68 ZIO_SET_CHECKSUM(zcp, 0, 0, 0, 0);
69 }
70
71 zio_checksum_info_t zio_checksum_table[ZIO_CHECKSUM_FUNCTIONS] = {
72 {{NULL, NULL}, 0, 0, 0, "inherit"},
73 {{NULL, NULL}, 0, 0, 0, "on"},
74 {{zio_checksum_off, zio_checksum_off}, 0, 0, 0, "off"},
75 {{zio_checksum_SHA256, zio_checksum_SHA256}, 1, 1, 0, "label"},
76 {{zio_checksum_SHA256, zio_checksum_SHA256}, 1, 1, 0, "gang_header"},
77 {{fletcher_2_native, fletcher_2_byteswap}, 0, 1, 0, "zilog"},
78 {{fletcher_2_native, fletcher_2_byteswap}, 0, 0, 0, "fletcher2"},
79 {{fletcher_4_native, fletcher_4_byteswap}, 1, 0, 0, "fletcher4"},
80 {{zio_checksum_SHA256, zio_checksum_SHA256}, 1, 0, 1, "sha256"},
81 {{fletcher_4_native, fletcher_4_byteswap}, 0, 1, 0, "zilog2"},
82 {{zio_checksum_off, zio_checksum_off}, 0, 0, 0, "noparity"},
83 };
84
85 enum zio_checksum
86 zio_checksum_select(enum zio_checksum child, enum zio_checksum parent)
87 {
88 ASSERT(child < ZIO_CHECKSUM_FUNCTIONS);
89 ASSERT(parent < ZIO_CHECKSUM_FUNCTIONS);
90 ASSERT(parent != ZIO_CHECKSUM_INHERIT && parent != ZIO_CHECKSUM_ON);
91
92 if (child == ZIO_CHECKSUM_INHERIT)
93 return (parent);
94
95 if (child == ZIO_CHECKSUM_ON)
96 return (ZIO_CHECKSUM_ON_VALUE);
97
98 return (child);
99 }
100
101 enum zio_checksum
102 zio_checksum_dedup_select(spa_t *spa, enum zio_checksum child,
103 enum zio_checksum parent)
104 {
105 ASSERT((child & ZIO_CHECKSUM_MASK) < ZIO_CHECKSUM_FUNCTIONS);
106 ASSERT((parent & ZIO_CHECKSUM_MASK) < ZIO_CHECKSUM_FUNCTIONS);
107 ASSERT(parent != ZIO_CHECKSUM_INHERIT && parent != ZIO_CHECKSUM_ON);
108
109 if (child == ZIO_CHECKSUM_INHERIT)
110 return (parent);
111
112 if (child == ZIO_CHECKSUM_ON)
113 return (spa_dedup_checksum(spa));
114
115 if (child == (ZIO_CHECKSUM_ON | ZIO_CHECKSUM_VERIFY))
116 return (spa_dedup_checksum(spa) | ZIO_CHECKSUM_VERIFY);
117
118 ASSERT(zio_checksum_table[child & ZIO_CHECKSUM_MASK].ci_dedup ||
119 (child & ZIO_CHECKSUM_VERIFY) || child == ZIO_CHECKSUM_OFF);
120
121 return (child);
122 }
123
124 /*
125 * Set the external verifier for a gang block based on <vdev, offset, txg>,
126 * a tuple which is guaranteed to be unique for the life of the pool.
127 */
128 static void
129 zio_checksum_gang_verifier(zio_cksum_t *zcp, blkptr_t *bp)
130 {
131 dva_t *dva = BP_IDENTITY(bp);
132 uint64_t txg = BP_PHYSICAL_BIRTH(bp);
133
134 ASSERT(BP_IS_GANG(bp));
135
136 ZIO_SET_CHECKSUM(zcp, DVA_GET_VDEV(dva), DVA_GET_OFFSET(dva), txg, 0);
137 }
138
139 /*
140 * Set the external verifier for a label block based on its offset.
141 * The vdev is implicit, and the txg is unknowable at pool open time --
142 * hence the logic in vdev_uberblock_load() to find the most recent copy.
143 */
144 static void
145 zio_checksum_label_verifier(zio_cksum_t *zcp, uint64_t offset)
146 {
147 ZIO_SET_CHECKSUM(zcp, offset, 0, 0, 0);
148 }
149
150 /*
151 * Generate the checksum.
152 */
153 void
154 zio_checksum_compute(zio_t *zio, enum zio_checksum checksum,
155 void *data, uint64_t size)
156 {
157 blkptr_t *bp = zio->io_bp;
158 uint64_t offset = zio->io_offset;
159 zio_checksum_info_t *ci = &zio_checksum_table[checksum];
160 zio_cksum_t cksum;
161
162 ASSERT((uint_t)checksum < ZIO_CHECKSUM_FUNCTIONS);
163 ASSERT(ci->ci_func[0] != NULL);
164
165 if (ci->ci_eck) {
166 zio_eck_t *eck;
167
168 if (checksum == ZIO_CHECKSUM_ZILOG2) {
169 zil_chain_t *zilc = data;
170
171 size = P2ROUNDUP_TYPED(zilc->zc_nused, ZIL_MIN_BLKSZ,
172 uint64_t);
173 eck = &zilc->zc_eck;
174 } else {
175 eck = (zio_eck_t *)((char *)data + size) - 1;
176 }
177 if (checksum == ZIO_CHECKSUM_GANG_HEADER)
178 zio_checksum_gang_verifier(&eck->zec_cksum, bp);
179 else if (checksum == ZIO_CHECKSUM_LABEL)
180 zio_checksum_label_verifier(&eck->zec_cksum, offset);
181 else
182 bp->blk_cksum = eck->zec_cksum;
183 eck->zec_magic = ZEC_MAGIC;
184 ci->ci_func[0](data, size, &cksum);
185 eck->zec_cksum = cksum;
186 } else {
187 ci->ci_func[0](data, size, &bp->blk_cksum);
188 }
189 }
190
191 int
192 zio_checksum_error(zio_t *zio, zio_bad_cksum_t *info)
193 {
194 blkptr_t *bp = zio->io_bp;
195 uint_t checksum = (bp == NULL ? zio->io_prop.zp_checksum :
196 (BP_IS_GANG(bp) ? ZIO_CHECKSUM_GANG_HEADER : BP_GET_CHECKSUM(bp)));
197 int byteswap;
198 int error;
199 uint64_t size = (bp == NULL ? zio->io_size :
200 (BP_IS_GANG(bp) ? SPA_GANGBLOCKSIZE : BP_GET_PSIZE(bp)));
201 uint64_t offset = zio->io_offset;
202 void *data = zio->io_data;
203 zio_checksum_info_t *ci = &zio_checksum_table[checksum];
204 zio_cksum_t actual_cksum, expected_cksum, verifier;
205
206 if (checksum >= ZIO_CHECKSUM_FUNCTIONS || ci->ci_func[0] == NULL)
207 return (SET_ERROR(EINVAL));
208
209 if (ci->ci_eck) {
210 zio_eck_t *eck;
211
212 if (checksum == ZIO_CHECKSUM_ZILOG2) {
213 zil_chain_t *zilc = data;
214 uint64_t nused;
215
216 eck = &zilc->zc_eck;
217 if (eck->zec_magic == ZEC_MAGIC)
218 nused = zilc->zc_nused;
219 else if (eck->zec_magic == BSWAP_64(ZEC_MAGIC))
220 nused = BSWAP_64(zilc->zc_nused);
221 else
222 return (SET_ERROR(ECKSUM));
223
224 if (nused > size)
225 return (SET_ERROR(ECKSUM));
226
227 size = P2ROUNDUP_TYPED(nused, ZIL_MIN_BLKSZ, uint64_t);
228 } else {
229 eck = (zio_eck_t *)((char *)data + size) - 1;
230 }
231
232 if (checksum == ZIO_CHECKSUM_GANG_HEADER)
233 zio_checksum_gang_verifier(&verifier, bp);
234 else if (checksum == ZIO_CHECKSUM_LABEL)
235 zio_checksum_label_verifier(&verifier, offset);
236 else
237 verifier = bp->blk_cksum;
238
239 byteswap = (eck->zec_magic == BSWAP_64(ZEC_MAGIC));
240
241 if (byteswap)
242 byteswap_uint64_array(&verifier, sizeof (zio_cksum_t));
243
244 expected_cksum = eck->zec_cksum;
245 eck->zec_cksum = verifier;
246 ci->ci_func[byteswap](data, size, &actual_cksum);
247 eck->zec_cksum = expected_cksum;
248
249 if (byteswap)
250 byteswap_uint64_array(&expected_cksum,
251 sizeof (zio_cksum_t));
252 } else {
253 ASSERT(!BP_IS_GANG(bp));
254 byteswap = BP_SHOULD_BYTESWAP(bp);
255 expected_cksum = bp->blk_cksum;
256 ci->ci_func[byteswap](data, size, &actual_cksum);
257 }
258
259 info->zbc_expected = expected_cksum;
260 info->zbc_actual = actual_cksum;
261 info->zbc_checksum_name = ci->ci_name;
262 info->zbc_byteswapped = byteswap;
263 info->zbc_injected = 0;
264 info->zbc_has_cksum = 1;
265
266 if (!ZIO_CHECKSUM_EQUAL(actual_cksum, expected_cksum))
267 return (SET_ERROR(ECKSUM));
268
269 if (zio_injection_enabled && !zio->io_error &&
270 (error = zio_handle_fault_injection(zio, ECKSUM)) != 0) {
271
272 info->zbc_injected = 1;
273 return (error);
274 }
275
276 return (0);
277 }