2 * Copyright (c) 2004-2006 Pawel Jakub Dawidek <pjd@FreeBSD.org>
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 #include <sys/endian.h>
35 #define G_RAID3_CLASS_NAME "RAID3"
37 #define G_RAID3_MAGIC "GEOM::RAID3"
40 * 0 - Initial version number.
41 * 1 - Added 'round-robin reading' algorithm.
42 * 2 - Added 'verify reading' algorithm.
43 * 3 - Added md_genid field to metadata.
44 * 4 - Added md_provsize field to metadata.
45 * 5 - Added 'no failure synchronization' flag.
47 #define G_RAID3_VERSION 5
49 #define G_RAID3_DISK_FLAG_DIRTY 0x0000000000000001ULL
50 #define G_RAID3_DISK_FLAG_SYNCHRONIZING 0x0000000000000002ULL
51 #define G_RAID3_DISK_FLAG_FORCE_SYNC 0x0000000000000004ULL
52 #define G_RAID3_DISK_FLAG_HARDCODED 0x0000000000000008ULL
53 #define G_RAID3_DISK_FLAG_BROKEN 0x0000000000000010ULL
54 #define G_RAID3_DISK_FLAG_MASK (G_RAID3_DISK_FLAG_DIRTY | \
55 G_RAID3_DISK_FLAG_SYNCHRONIZING | \
56 G_RAID3_DISK_FLAG_FORCE_SYNC)
58 #define G_RAID3_DEVICE_FLAG_NOAUTOSYNC 0x0000000000000001ULL
59 #define G_RAID3_DEVICE_FLAG_ROUND_ROBIN 0x0000000000000002ULL
60 #define G_RAID3_DEVICE_FLAG_VERIFY 0x0000000000000004ULL
61 #define G_RAID3_DEVICE_FLAG_NOFAILSYNC 0x0000000000000008ULL
62 #define G_RAID3_DEVICE_FLAG_MASK (G_RAID3_DEVICE_FLAG_NOAUTOSYNC | \
63 G_RAID3_DEVICE_FLAG_ROUND_ROBIN | \
64 G_RAID3_DEVICE_FLAG_VERIFY | \
65 G_RAID3_DEVICE_FLAG_NOFAILSYNC)
68 extern u_int g_raid3_debug;
70 #define G_RAID3_DEBUG(lvl, ...) do { \
71 if (g_raid3_debug >= (lvl)) { \
72 printf("GEOM_RAID3"); \
73 if (g_raid3_debug > 0) \
74 printf("[%u]", lvl); \
76 printf(__VA_ARGS__); \
80 #define G_RAID3_LOGREQ(lvl, bp, ...) do { \
81 if (g_raid3_debug >= (lvl)) { \
82 printf("GEOM_RAID3"); \
83 if (g_raid3_debug > 0) \
84 printf("[%u]", lvl); \
86 printf(__VA_ARGS__); \
93 #define G_RAID3_BIO_CFLAG_REGULAR 0x01
94 #define G_RAID3_BIO_CFLAG_SYNC 0x02
95 #define G_RAID3_BIO_CFLAG_PARITY 0x04
96 #define G_RAID3_BIO_CFLAG_NODISK 0x08
97 #define G_RAID3_BIO_CFLAG_REGSYNC 0x10
98 #define G_RAID3_BIO_CFLAG_MASK (G_RAID3_BIO_CFLAG_REGULAR | \
99 G_RAID3_BIO_CFLAG_SYNC | \
100 G_RAID3_BIO_CFLAG_PARITY | \
101 G_RAID3_BIO_CFLAG_NODISK | \
102 G_RAID3_BIO_CFLAG_REGSYNC)
104 #define G_RAID3_BIO_PFLAG_DEGRADED 0x01
105 #define G_RAID3_BIO_PFLAG_NOPARITY 0x02
106 #define G_RAID3_BIO_PFLAG_VERIFY 0x04
107 #define G_RAID3_BIO_PFLAG_MASK (G_RAID3_BIO_PFLAG_DEGRADED | \
108 G_RAID3_BIO_PFLAG_NOPARITY | \
109 G_RAID3_BIO_PFLAG_VERIFY)
112 * Informations needed for synchronization.
114 struct g_raid3_disk_sync {
115 struct g_consumer *ds_consumer; /* Consumer connected to our device. */
116 off_t ds_offset; /* Offset of next request to send. */
117 off_t ds_offset_done; /* Offset of already synchronized
119 off_t ds_resync; /* Resynchronize from this offset. */
120 u_int ds_syncid; /* Disk's synchronization ID. */
121 u_int ds_inflight; /* Number of in-flight sync requests. */
122 struct bio **ds_bios; /* BIOs for synchronization I/O. */
126 * Informations needed for synchronization.
128 struct g_raid3_device_sync {
129 struct g_geom *ds_geom; /* Synchronization geom. */
132 #define G_RAID3_DISK_STATE_NODISK 0
133 #define G_RAID3_DISK_STATE_NONE 1
134 #define G_RAID3_DISK_STATE_NEW 2
135 #define G_RAID3_DISK_STATE_ACTIVE 3
136 #define G_RAID3_DISK_STATE_STALE 4
137 #define G_RAID3_DISK_STATE_SYNCHRONIZING 5
138 #define G_RAID3_DISK_STATE_DISCONNECTED 6
139 #define G_RAID3_DISK_STATE_DESTROY 7
140 struct g_raid3_disk {
141 u_int d_no; /* Disk number. */
142 struct g_consumer *d_consumer; /* Consumer. */
143 struct g_raid3_softc *d_softc; /* Back-pointer to softc. */
144 int d_state; /* Disk state. */
145 uint64_t d_flags; /* Additional flags. */
146 u_int d_genid; /* Disk's generation ID. */
147 struct g_raid3_disk_sync d_sync; /* Sync information. */
148 LIST_ENTRY(g_raid3_disk) d_next;
150 #define d_name d_consumer->provider->name
152 #define G_RAID3_EVENT_DONTWAIT 0x1
153 #define G_RAID3_EVENT_WAIT 0x2
154 #define G_RAID3_EVENT_DEVICE 0x4
155 #define G_RAID3_EVENT_DONE 0x8
156 struct g_raid3_event {
157 struct g_raid3_disk *e_disk;
161 TAILQ_ENTRY(g_raid3_event) e_next;
164 #define G_RAID3_DEVICE_FLAG_DESTROY 0x0100000000000000ULL
165 #define G_RAID3_DEVICE_FLAG_WAIT 0x0200000000000000ULL
166 #define G_RAID3_DEVICE_FLAG_DESTROYING 0x0400000000000000ULL
168 #define G_RAID3_DEVICE_STATE_STARTING 0
169 #define G_RAID3_DEVICE_STATE_DEGRADED 1
170 #define G_RAID3_DEVICE_STATE_COMPLETE 2
172 /* Bump syncid on first write. */
173 #define G_RAID3_BUMP_SYNCID 0x1
174 /* Bump genid immediately. */
175 #define G_RAID3_BUMP_GENID 0x2
184 static __inline enum g_raid3_zones
185 g_raid3_zone(size_t nbytes) {
187 return (G_RAID3_ZONE_64K);
188 else if (nbytes > 4096)
189 return (G_RAID3_ZONE_16K);
191 return (G_RAID3_ZONE_4K);
194 struct g_raid3_softc {
195 u_int sc_state; /* Device state. */
196 uint64_t sc_mediasize; /* Device size. */
197 uint32_t sc_sectorsize; /* Sector size. */
198 uint64_t sc_flags; /* Additional flags. */
200 struct g_geom *sc_geom;
201 struct g_provider *sc_provider;
203 uint32_t sc_id; /* Device unique ID. */
206 struct bio_queue_head sc_queue;
207 struct mtx sc_queue_mtx;
208 struct proc *sc_worker;
209 struct bio_queue_head sc_regular_delayed; /* Delayed I/O requests due
212 struct bio_queue_head sc_inflight; /* In-flight regular write
214 struct bio_queue_head sc_sync_delayed; /* Delayed sync requests due
215 collision with regular
218 struct g_raid3_disk *sc_disks;
219 u_int sc_ndisks; /* Number of disks. */
220 u_int sc_round_robin;
221 struct g_raid3_disk *sc_syncdisk;
223 struct g_raid3_zone {
229 } sc_zones[G_RAID3_NUM_ZONES];
231 u_int sc_genid; /* Generation ID. */
232 u_int sc_syncid; /* Synchronization ID. */
234 struct g_raid3_device_sync sc_sync;
235 int sc_idle; /* DIRTY flags removed. */
236 time_t sc_last_write;
239 TAILQ_HEAD(, g_raid3_event) sc_events;
240 struct mtx sc_events_mtx;
242 struct callout sc_callout;
244 struct root_hold_token *sc_rootmount;
246 #define sc_name sc_geom->name
248 const char *g_raid3_get_diskname(struct g_raid3_disk *disk);
249 u_int g_raid3_ndisks(struct g_raid3_softc *sc, int state);
250 #define G_RAID3_DESTROY_SOFT 0
251 #define G_RAID3_DESTROY_DELAYED 1
252 #define G_RAID3_DESTROY_HARD 2
253 int g_raid3_destroy(struct g_raid3_softc *sc, int how);
254 int g_raid3_event_send(void *arg, int state, int flags);
255 struct g_raid3_metadata;
256 int g_raid3_add_disk(struct g_raid3_softc *sc, struct g_provider *pp,
257 struct g_raid3_metadata *md);
258 int g_raid3_read_metadata(struct g_consumer *cp, struct g_raid3_metadata *md);
259 void g_raid3_fill_metadata(struct g_raid3_disk *disk,
260 struct g_raid3_metadata *md);
261 int g_raid3_clear_metadata(struct g_raid3_disk *disk);
262 void g_raid3_update_metadata(struct g_raid3_disk *disk);
264 g_ctl_req_t g_raid3_config;
267 struct g_raid3_metadata {
268 char md_magic[16]; /* Magic value. */
269 uint32_t md_version; /* Version number. */
270 char md_name[16]; /* Device name. */
271 uint32_t md_id; /* Device unique ID. */
272 uint16_t md_no; /* Component number. */
273 uint16_t md_all; /* Number of disks in device. */
274 uint32_t md_genid; /* Generation ID. */
275 uint32_t md_syncid; /* Synchronization ID. */
276 uint64_t md_mediasize; /* Size of whole device. */
277 uint32_t md_sectorsize; /* Sector size. */
278 uint64_t md_sync_offset; /* Synchronized offset. */
279 uint64_t md_mflags; /* Additional device flags. */
280 uint64_t md_dflags; /* Additional disk flags. */
281 char md_provider[16]; /* Hardcoded provider. */
282 uint64_t md_provsize; /* Provider's size. */
283 u_char md_hash[16]; /* MD5 hash. */
286 raid3_metadata_encode(struct g_raid3_metadata *md, u_char *data)
290 bcopy(md->md_magic, data, 16);
291 le32enc(data + 16, md->md_version);
292 bcopy(md->md_name, data + 20, 16);
293 le32enc(data + 36, md->md_id);
294 le16enc(data + 40, md->md_no);
295 le16enc(data + 42, md->md_all);
296 le32enc(data + 44, md->md_genid);
297 le32enc(data + 48, md->md_syncid);
298 le64enc(data + 52, md->md_mediasize);
299 le32enc(data + 60, md->md_sectorsize);
300 le64enc(data + 64, md->md_sync_offset);
301 le64enc(data + 72, md->md_mflags);
302 le64enc(data + 80, md->md_dflags);
303 bcopy(md->md_provider, data + 88, 16);
304 le64enc(data + 104, md->md_provsize);
306 MD5Update(&ctx, data, 112);
307 MD5Final(md->md_hash, &ctx);
308 bcopy(md->md_hash, data + 112, 16);
311 raid3_metadata_decode_v0v1v2(const u_char *data, struct g_raid3_metadata *md)
315 bcopy(data + 20, md->md_name, 16);
316 md->md_id = le32dec(data + 36);
317 md->md_no = le16dec(data + 40);
318 md->md_all = le16dec(data + 42);
319 md->md_syncid = le32dec(data + 44);
320 md->md_mediasize = le64dec(data + 48);
321 md->md_sectorsize = le32dec(data + 56);
322 md->md_sync_offset = le64dec(data + 60);
323 md->md_mflags = le64dec(data + 68);
324 md->md_dflags = le64dec(data + 76);
325 bcopy(data + 84, md->md_provider, 16);
326 bcopy(data + 100, md->md_hash, 16);
328 MD5Update(&ctx, data, 100);
329 MD5Final(md->md_hash, &ctx);
330 if (bcmp(md->md_hash, data + 100, 16) != 0)
340 raid3_metadata_decode_v3(const u_char *data, struct g_raid3_metadata *md)
344 bcopy(data + 20, md->md_name, 16);
345 md->md_id = le32dec(data + 36);
346 md->md_no = le16dec(data + 40);
347 md->md_all = le16dec(data + 42);
348 md->md_genid = le32dec(data + 44);
349 md->md_syncid = le32dec(data + 48);
350 md->md_mediasize = le64dec(data + 52);
351 md->md_sectorsize = le32dec(data + 60);
352 md->md_sync_offset = le64dec(data + 64);
353 md->md_mflags = le64dec(data + 72);
354 md->md_dflags = le64dec(data + 80);
355 bcopy(data + 88, md->md_provider, 16);
356 bcopy(data + 104, md->md_hash, 16);
358 MD5Update(&ctx, data, 104);
359 MD5Final(md->md_hash, &ctx);
360 if (bcmp(md->md_hash, data + 104, 16) != 0)
369 raid3_metadata_decode_v4v5(const u_char *data, struct g_raid3_metadata *md)
373 bcopy(data + 20, md->md_name, 16);
374 md->md_id = le32dec(data + 36);
375 md->md_no = le16dec(data + 40);
376 md->md_all = le16dec(data + 42);
377 md->md_genid = le32dec(data + 44);
378 md->md_syncid = le32dec(data + 48);
379 md->md_mediasize = le64dec(data + 52);
380 md->md_sectorsize = le32dec(data + 60);
381 md->md_sync_offset = le64dec(data + 64);
382 md->md_mflags = le64dec(data + 72);
383 md->md_dflags = le64dec(data + 80);
384 bcopy(data + 88, md->md_provider, 16);
385 md->md_provsize = le64dec(data + 104);
386 bcopy(data + 112, md->md_hash, 16);
388 MD5Update(&ctx, data, 112);
389 MD5Final(md->md_hash, &ctx);
390 if (bcmp(md->md_hash, data + 112, 16) != 0)
395 raid3_metadata_decode(const u_char *data, struct g_raid3_metadata *md)
399 bcopy(data, md->md_magic, 16);
400 md->md_version = le32dec(data + 16);
401 switch (md->md_version) {
405 error = raid3_metadata_decode_v0v1v2(data, md);
408 error = raid3_metadata_decode_v3(data, md);
412 error = raid3_metadata_decode_v4v5(data, md);
422 raid3_metadata_dump(const struct g_raid3_metadata *md)
424 static const char hex[] = "0123456789abcdef";
425 char hash[16 * 2 + 1];
428 printf(" magic: %s\n", md->md_magic);
429 printf(" version: %u\n", (u_int)md->md_version);
430 printf(" name: %s\n", md->md_name);
431 printf(" id: %u\n", (u_int)md->md_id);
432 printf(" no: %u\n", (u_int)md->md_no);
433 printf(" all: %u\n", (u_int)md->md_all);
434 printf(" genid: %u\n", (u_int)md->md_genid);
435 printf(" syncid: %u\n", (u_int)md->md_syncid);
436 printf(" mediasize: %jd\n", (intmax_t)md->md_mediasize);
437 printf("sectorsize: %u\n", (u_int)md->md_sectorsize);
438 printf("syncoffset: %jd\n", (intmax_t)md->md_sync_offset);
440 if (md->md_mflags == 0)
443 if ((md->md_mflags & G_RAID3_DEVICE_FLAG_NOAUTOSYNC) != 0)
444 printf(" NOAUTOSYNC");
445 if ((md->md_mflags & G_RAID3_DEVICE_FLAG_ROUND_ROBIN) != 0)
446 printf(" ROUND-ROBIN");
447 if ((md->md_mflags & G_RAID3_DEVICE_FLAG_VERIFY) != 0)
449 if ((md->md_mflags & G_RAID3_DEVICE_FLAG_NOFAILSYNC) != 0)
450 printf(" NOFAILSYNC");
454 if (md->md_dflags == 0)
457 if ((md->md_dflags & G_RAID3_DISK_FLAG_DIRTY) != 0)
459 if ((md->md_dflags & G_RAID3_DISK_FLAG_SYNCHRONIZING) != 0)
460 printf(" SYNCHRONIZING");
461 if ((md->md_dflags & G_RAID3_DISK_FLAG_FORCE_SYNC) != 0)
462 printf(" FORCE_SYNC");
465 printf("hcprovider: %s\n", md->md_provider);
466 printf(" provsize: %ju\n", (uintmax_t)md->md_provsize);
467 bzero(hash, sizeof(hash));
468 for (i = 0; i < 16; i++) {
469 hash[i * 2] = hex[md->md_hash[i] >> 4];
470 hash[i * 2 + 1] = hex[md->md_hash[i] & 0x0f];
472 printf(" MD5 hash: %s\n", hash);
474 #endif /* !_G_RAID3_H_ */