]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * mdadm - manage Linux "md" devices aka RAID arrays. | |
3 | * | |
4 | * Copyright (C) 2001-2009 Neil Brown <neilb@suse.de> | |
5 | * | |
6 | * | |
7 | * This program is free software; you can redistribute it and/or modify | |
8 | * it under the terms of the GNU General Public License as published by | |
9 | * the Free Software Foundation; either version 2 of the License, or | |
10 | * (at your option) any later version. | |
11 | * | |
12 | * This program is distributed in the hope that it will be useful, | |
13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
15 | * GNU General Public License for more details. | |
16 | * | |
17 | * You should have received a copy of the GNU General Public License | |
18 | * along with this program; if not, write to the Free Software | |
19 | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA | |
20 | * | |
21 | * Author: Neil Brown | |
22 | * Email: <neilb@suse.de> | |
23 | */ | |
24 | ||
25 | #define _GNU_SOURCE | |
26 | #define _FILE_OFFSET_BITS 64 | |
27 | #include <unistd.h> | |
28 | #if !defined(__dietlibc__) && !defined(__KLIBC__) | |
29 | extern __off64_t lseek64 __P ((int __fd, __off64_t __offset, int __whence)); | |
30 | #else | |
31 | # if defined(__NO_STAT64) || __WORDSIZE != 32 | |
32 | # define lseek64 lseek | |
33 | # endif | |
34 | #endif | |
35 | ||
36 | #include <sys/types.h> | |
37 | #include <sys/stat.h> | |
38 | #include <stdlib.h> | |
39 | #include <time.h> | |
40 | #include <sys/time.h> | |
41 | #include <getopt.h> | |
42 | #include <fcntl.h> | |
43 | #include <stdio.h> | |
44 | #include <errno.h> | |
45 | #include <string.h> | |
46 | #include <syslog.h> | |
47 | #ifdef __dietlibc__ | |
48 | #include <strings.h> | |
49 | /* dietlibc has deprecated random and srandom!! */ | |
50 | #define random rand | |
51 | #define srandom srand | |
52 | #endif | |
53 | ||
54 | ||
55 | #include <linux/kdev_t.h> | |
56 | /*#include <linux/fs.h> */ | |
57 | #include <sys/mount.h> | |
58 | #include <asm/types.h> | |
59 | #include <sys/ioctl.h> | |
60 | #define MD_MAJOR 9 | |
61 | #define MdpMinorShift 6 | |
62 | ||
63 | #ifndef BLKGETSIZE64 | |
64 | #define BLKGETSIZE64 _IOR(0x12,114,size_t) /* return device size in bytes (u64 *arg) */ | |
65 | #endif | |
66 | ||
67 | #define DEFAULT_CHUNK 512 | |
68 | #define DEFAULT_BITMAP_CHUNK 4096 | |
69 | #define DEFAULT_BITMAP_DELAY 5 | |
70 | #define DEFAULT_MAX_WRITE_BEHIND 256 | |
71 | ||
72 | /* MAP_DIR should be somewhere that persists across the pivotroot | |
73 | * from early boot to late boot. | |
74 | * /run seems to have emerged as the best standard. | |
75 | */ | |
76 | #ifndef MAP_DIR | |
77 | #define MAP_DIR "/run/mdadm" | |
78 | #endif /* MAP_DIR */ | |
79 | /* MAP_FILE is what we name the map file we put in MAP_DIR, in case you | |
80 | * want something other than the default of "map" | |
81 | */ | |
82 | #ifndef MAP_FILE | |
83 | #define MAP_FILE "map" | |
84 | #endif /* MAP_FILE */ | |
85 | /* MDMON_DIR is where pid and socket files used for communicating | |
86 | * with mdmon normally live. Best is /var/run/mdadm as | |
87 | * mdmon is needed at early boot then it needs to write there prior | |
88 | * to /var/run being mounted read/write, and it also then needs to | |
89 | * persist beyond when /var/run is mounter read-only. So, to be | |
90 | * safe, the default is somewhere that is read/write early in the | |
91 | * boot process and stays up as long as possible during shutdown. | |
92 | */ | |
93 | #ifndef MDMON_DIR | |
94 | #define MDMON_DIR "/run/mdadm" | |
95 | #endif /* MDMON_DIR */ | |
96 | ||
97 | /* FAILED_SLOTS is where to save files storing recent removal of array | |
98 | * member in order to allow future reuse of disk inserted in the same | |
99 | * slot for array recovery | |
100 | */ | |
101 | #ifndef FAILED_SLOTS_DIR | |
102 | #define FAILED_SLOTS_DIR "/run/mdadm/failed-slots" | |
103 | #endif /* FAILED_SLOTS */ | |
104 | ||
105 | #include "md_u.h" | |
106 | #include "md_p.h" | |
107 | #include "bitmap.h" | |
108 | #include "msg.h" | |
109 | ||
110 | #include <endian.h> | |
111 | /* Redhat don't like to #include <asm/byteorder.h>, and | |
112 | * some time include <linux/byteorder/xxx_endian.h> isn't enough, | |
113 | * and there is no standard conversion function so... */ | |
114 | /* And dietlibc doesn't think byteswap is ok, so.. */ | |
115 | /* #include <byteswap.h> */ | |
116 | #define bswap_16(x) (((x) & 0x00ffU) << 8 | \ | |
117 | ((x) & 0xff00U) >> 8) | |
118 | #define bswap_32(x) (((x) & 0x000000ffU) << 24 | \ | |
119 | ((x) & 0xff000000U) >> 24 | \ | |
120 | ((x) & 0x0000ff00U) << 8 | \ | |
121 | ((x) & 0x00ff0000U) >> 8) | |
122 | #define bswap_64(x) (((x) & 0x00000000000000ffULL) << 56 | \ | |
123 | ((x) & 0xff00000000000000ULL) >> 56 | \ | |
124 | ((x) & 0x000000000000ff00ULL) << 40 | \ | |
125 | ((x) & 0x00ff000000000000ULL) >> 40 | \ | |
126 | ((x) & 0x0000000000ff0000ULL) << 24 | \ | |
127 | ((x) & 0x0000ff0000000000ULL) >> 24 | \ | |
128 | ((x) & 0x00000000ff000000ULL) << 8 | \ | |
129 | ((x) & 0x000000ff00000000ULL) >> 8) | |
130 | ||
131 | #if !defined(__KLIBC__) | |
132 | #if BYTE_ORDER == LITTLE_ENDIAN | |
133 | #define __cpu_to_le16(_x) (_x) | |
134 | #define __cpu_to_le32(_x) (_x) | |
135 | #define __cpu_to_le64(_x) (_x) | |
136 | #define __le16_to_cpu(_x) (_x) | |
137 | #define __le32_to_cpu(_x) (_x) | |
138 | #define __le64_to_cpu(_x) (_x) | |
139 | ||
140 | #define __cpu_to_be16(_x) bswap_16(_x) | |
141 | #define __cpu_to_be32(_x) bswap_32(_x) | |
142 | #define __cpu_to_be64(_x) bswap_64(_x) | |
143 | #define __be16_to_cpu(_x) bswap_16(_x) | |
144 | #define __be32_to_cpu(_x) bswap_32(_x) | |
145 | #define __be64_to_cpu(_x) bswap_64(_x) | |
146 | #elif BYTE_ORDER == BIG_ENDIAN | |
147 | #define __cpu_to_le16(_x) bswap_16(_x) | |
148 | #define __cpu_to_le32(_x) bswap_32(_x) | |
149 | #define __cpu_to_le64(_x) bswap_64(_x) | |
150 | #define __le16_to_cpu(_x) bswap_16(_x) | |
151 | #define __le32_to_cpu(_x) bswap_32(_x) | |
152 | #define __le64_to_cpu(_x) bswap_64(_x) | |
153 | ||
154 | #define __cpu_to_be16(_x) (_x) | |
155 | #define __cpu_to_be32(_x) (_x) | |
156 | #define __cpu_to_be64(_x) (_x) | |
157 | #define __be16_to_cpu(_x) (_x) | |
158 | #define __be32_to_cpu(_x) (_x) | |
159 | #define __be64_to_cpu(_x) (_x) | |
160 | #else | |
161 | # error "unknown endianness." | |
162 | #endif | |
163 | #endif /* __KLIBC__ */ | |
164 | ||
165 | ||
166 | /* | |
167 | * min()/max()/clamp() macros that also do | |
168 | * strict type-checking.. See the | |
169 | * "unnecessary" pointer comparison. | |
170 | */ | |
171 | #define min(x, y) ({ \ | |
172 | typeof(x) _min1 = (x); \ | |
173 | typeof(y) _min2 = (y); \ | |
174 | (void) (&_min1 == &_min2); \ | |
175 | _min1 < _min2 ? _min1 : _min2; }) | |
176 | ||
177 | #define max(x, y) ({ \ | |
178 | typeof(x) _max1 = (x); \ | |
179 | typeof(y) _max2 = (y); \ | |
180 | (void) (&_max1 == &_max2); \ | |
181 | _max1 > _max2 ? _max1 : _max2; }) | |
182 | ||
183 | /* general information that might be extracted from a superblock */ | |
184 | struct mdinfo { | |
185 | mdu_array_info_t array; | |
186 | mdu_disk_info_t disk; | |
187 | __u64 events; | |
188 | int uuid[4]; | |
189 | char name[33]; | |
190 | unsigned long long data_offset; | |
191 | unsigned long long component_size; /* same as array.size, except in | |
192 | * sectors and up to 64bits. | |
193 | */ | |
194 | unsigned long long custom_array_size; /* size for non-default sized | |
195 | * arrays (in sectors) | |
196 | */ | |
197 | #define NO_RESHAPE 0 | |
198 | #define VOLUME_RESHAPE 1 | |
199 | #define CONTAINER_RESHAPE 2 | |
200 | int reshape_active; | |
201 | unsigned long long reshape_progress; | |
202 | int recovery_blocked; /* for external metadata it | |
203 | * indicates that there is | |
204 | * reshape in progress in | |
205 | * container, | |
206 | * for native metadata it is | |
207 | * reshape_active field mirror | |
208 | */ | |
209 | union { | |
210 | unsigned long long resync_start; /* per-array resync position */ | |
211 | unsigned long long recovery_start; /* per-device rebuild position */ | |
212 | #define MaxSector (~0ULL) /* resync/recovery complete position */ | |
213 | }; | |
214 | long bitmap_offset; /* 0 == none, 1 == a file */ | |
215 | unsigned long safe_mode_delay; /* ms delay to mark clean */ | |
216 | int new_level, delta_disks, new_layout, new_chunk; | |
217 | int errors; | |
218 | unsigned long cache_size; /* size of raid456 stripe cache*/ | |
219 | int mismatch_cnt; | |
220 | char text_version[50]; | |
221 | ||
222 | int container_member; /* for assembling external-metatdata arrays | |
223 | * This is to be used internally by metadata | |
224 | * handler only */ | |
225 | int container_enough; /* flag external handlers can set to | |
226 | * indicate that subarrays have not enough (-1), | |
227 | * enough to start (0), or all expected disks (1) */ | |
228 | char sys_name[20]; | |
229 | struct mdinfo *devs; | |
230 | struct mdinfo *next; | |
231 | ||
232 | /* Device info for mdmon: */ | |
233 | int recovery_fd; | |
234 | int state_fd; | |
235 | #define DS_FAULTY 1 | |
236 | #define DS_INSYNC 2 | |
237 | #define DS_WRITE_MOSTLY 4 | |
238 | #define DS_SPARE 8 | |
239 | #define DS_BLOCKED 16 | |
240 | #define DS_REMOVE 1024 | |
241 | #define DS_UNBLOCK 2048 | |
242 | int prev_state, curr_state, next_state; | |
243 | ||
244 | }; | |
245 | ||
246 | struct createinfo { | |
247 | int uid; | |
248 | int gid; | |
249 | int autof; | |
250 | int mode; | |
251 | int symlinks; | |
252 | struct supertype *supertype; | |
253 | }; | |
254 | ||
255 | #define Name "mdadm" | |
256 | ||
257 | enum mode { | |
258 | ASSEMBLE=1, | |
259 | BUILD, | |
260 | CREATE, | |
261 | MANAGE, | |
262 | MISC, | |
263 | MONITOR, | |
264 | GROW, | |
265 | INCREMENTAL, | |
266 | AUTODETECT, | |
267 | mode_count | |
268 | }; | |
269 | ||
270 | extern char short_options[]; | |
271 | extern char short_bitmap_options[]; | |
272 | extern char short_bitmap_auto_options[]; | |
273 | extern struct option long_options[]; | |
274 | extern char Version[], Usage[], Help[], OptionHelp[], | |
275 | *mode_help[], | |
276 | Help_create[], Help_build[], Help_assemble[], Help_grow[], | |
277 | Help_incr[], | |
278 | Help_manage[], Help_misc[], Help_monitor[], Help_config[]; | |
279 | ||
280 | /* for option that don't have short equivilents, we assign arbitrary | |
281 | * numbers later than any 'short' character option. | |
282 | */ | |
283 | enum special_options { | |
284 | AssumeClean = 300, | |
285 | BitmapChunk, | |
286 | WriteBehind, | |
287 | ReAdd, | |
288 | NoDegraded, | |
289 | Sparc22, | |
290 | BackupFile, | |
291 | HomeHost, | |
292 | AutoHomeHost, | |
293 | Symlinks, | |
294 | AutoDetect, | |
295 | Waitclean, | |
296 | DetailPlatform, | |
297 | KillSubarray, | |
298 | UpdateSubarray, | |
299 | IncrementalPath, | |
300 | NoSharing, | |
301 | HelpOptions, | |
302 | Brief, | |
303 | ManageOpt, | |
304 | Add, | |
305 | Remove, | |
306 | Fail, | |
307 | MiscOpt, | |
308 | WaitOpt, | |
309 | ConfigFile, | |
310 | ChunkSize, | |
311 | WriteMostly, | |
312 | Layout, | |
313 | Auto, | |
314 | Force, | |
315 | SuperMinor, | |
316 | EMail, | |
317 | ProgramOpt, | |
318 | Increment, | |
319 | Fork, | |
320 | Bitmap, | |
321 | RebuildMapOpt, | |
322 | InvalidBackup, | |
323 | UdevRules, | |
324 | FreezeReshape, | |
325 | Continue, | |
326 | OffRootOpt, | |
327 | Prefer, | |
328 | KillOpt, | |
329 | }; | |
330 | ||
331 | enum prefix_standard { | |
332 | JEDEC, | |
333 | IEC | |
334 | }; | |
335 | ||
336 | /* structures read from config file */ | |
337 | /* List of mddevice names and identifiers | |
338 | * Identifiers can be: | |
339 | * uuid=128-hex-uuid | |
340 | * super-minor=decimal-minor-number-from-superblock | |
341 | * devices=comma,separated,list,of,device,names,with,wildcards | |
342 | * | |
343 | * If multiple fields are present, the intersection of all matching | |
344 | * devices is considered | |
345 | */ | |
346 | #define UnSet (0xfffe) | |
347 | struct mddev_ident { | |
348 | char *devname; | |
349 | ||
350 | int uuid_set; | |
351 | int uuid[4]; | |
352 | char name[33]; | |
353 | ||
354 | int super_minor; | |
355 | ||
356 | char *devices; /* comma separated list of device | |
357 | * names with wild cards | |
358 | */ | |
359 | int level; | |
360 | int raid_disks; | |
361 | int spare_disks; | |
362 | struct supertype *st; | |
363 | int autof; /* 1 for normal, 2 for partitioned */ | |
364 | char *spare_group; | |
365 | char *bitmap_file; | |
366 | int bitmap_fd; | |
367 | ||
368 | char *container; /* /dev/whatever name of container, or | |
369 | * uuid of container. You would expect | |
370 | * this to be the 'devname' or UUID | |
371 | * of some other entry. | |
372 | */ | |
373 | char *member; /* subarray within a container */ | |
374 | ||
375 | struct mddev_ident *next; | |
376 | union { | |
377 | /* fields needed by different users of this structure */ | |
378 | int assembled; /* set when assembly succeeds */ | |
379 | }; | |
380 | }; | |
381 | ||
382 | struct context { | |
383 | int readonly; | |
384 | int runstop; | |
385 | int verbose; | |
386 | int brief; | |
387 | int force; | |
388 | char *homehost; | |
389 | int require_homehost; | |
390 | char *prefer; | |
391 | int export; | |
392 | int test; | |
393 | char *subarray; | |
394 | char *update; | |
395 | int scan; | |
396 | int SparcAdjust; | |
397 | int autof; | |
398 | int delay; | |
399 | int freeze_reshape; | |
400 | char *backup_file; | |
401 | int invalid_backup; | |
402 | }; | |
403 | ||
404 | struct shape { | |
405 | int raiddisks; | |
406 | int sparedisks; | |
407 | int level; | |
408 | int layout; | |
409 | char *layout_str; | |
410 | int chunk; | |
411 | int bitmap_chunk; | |
412 | char *bitmap_file; | |
413 | int assume_clean; | |
414 | int write_behind; | |
415 | unsigned long long size; | |
416 | }; | |
417 | ||
418 | /* List of device names - wildcards expanded */ | |
419 | struct mddev_dev { | |
420 | char *devname; | |
421 | int disposition; /* 'a' for add, 'r' for remove, 'f' for fail, | |
422 | * 'A' for re_add. | |
423 | * Not set for names read from .config | |
424 | */ | |
425 | char writemostly; /* 1 for 'set writemostly', 2 for 'clear writemostly' */ | |
426 | char used; /* set when used */ | |
427 | struct mddev_dev *next; | |
428 | }; | |
429 | ||
430 | typedef struct mapping { | |
431 | char *name; | |
432 | int num; | |
433 | } mapping_t; | |
434 | ||
435 | ||
436 | struct mdstat_ent { | |
437 | char *dev; | |
438 | int devnum; | |
439 | int active; | |
440 | char *level; | |
441 | char *pattern; /* U or up, _ for down */ | |
442 | int percent; /* -1 if no resync */ | |
443 | int resync; /* 3 if check, 2 if reshape, 1 if resync, 0 if recovery */ | |
444 | int devcnt; | |
445 | int raid_disks; | |
446 | char * metadata_version; | |
447 | struct dev_member { | |
448 | char *name; | |
449 | struct dev_member *next; | |
450 | } *members; | |
451 | struct mdstat_ent *next; | |
452 | }; | |
453 | ||
454 | extern struct mdstat_ent *mdstat_read(int hold, int start); | |
455 | extern void free_mdstat(struct mdstat_ent *ms); | |
456 | extern void mdstat_wait(int seconds); | |
457 | extern void mdstat_wait_fd(int fd, const sigset_t *sigmask); | |
458 | extern int mddev_busy(int devnum); | |
459 | extern struct mdstat_ent *mdstat_by_component(char *name); | |
460 | extern struct mdstat_ent *mdstat_by_subdev(char *subdev, int container); | |
461 | ||
462 | struct map_ent { | |
463 | struct map_ent *next; | |
464 | int devnum; | |
465 | char metadata[20]; | |
466 | int uuid[4]; | |
467 | int bad; | |
468 | char *path; | |
469 | }; | |
470 | extern int map_update(struct map_ent **mpp, int devnum, char *metadata, | |
471 | int uuid[4], char *path); | |
472 | extern void map_remove(struct map_ent **map, int devnum); | |
473 | extern struct map_ent *map_by_uuid(struct map_ent **map, int uuid[4]); | |
474 | extern struct map_ent *map_by_devnum(struct map_ent **map, int devnum); | |
475 | extern struct map_ent *map_by_name(struct map_ent **map, char *name); | |
476 | extern void map_read(struct map_ent **melp); | |
477 | extern int map_write(struct map_ent *mel); | |
478 | extern void map_delete(struct map_ent **mapp, int devnum); | |
479 | extern void map_free(struct map_ent *map); | |
480 | extern void map_add(struct map_ent **melp, | |
481 | int devnum, char *metadata, int uuid[4], char *path); | |
482 | extern int map_lock(struct map_ent **melp); | |
483 | extern void map_unlock(struct map_ent **melp); | |
484 | extern void map_fork(void); | |
485 | ||
486 | /* various details can be requested */ | |
487 | enum sysfs_read_flags { | |
488 | GET_LEVEL = (1 << 0), | |
489 | GET_LAYOUT = (1 << 1), | |
490 | GET_COMPONENT = (1 << 2), | |
491 | GET_CHUNK = (1 << 3), | |
492 | GET_CACHE = (1 << 4), | |
493 | GET_MISMATCH = (1 << 5), | |
494 | GET_VERSION = (1 << 6), | |
495 | GET_DISKS = (1 << 7), | |
496 | GET_DEGRADED = (1 << 8), | |
497 | GET_SAFEMODE = (1 << 9), | |
498 | GET_BITMAP_LOCATION = (1 << 10), | |
499 | ||
500 | GET_DEVS = (1 << 20), /* gets role, major, minor */ | |
501 | GET_OFFSET = (1 << 21), | |
502 | GET_SIZE = (1 << 22), | |
503 | GET_STATE = (1 << 23), | |
504 | GET_ERROR = (1 << 24), | |
505 | }; | |
506 | ||
507 | /* If fd >= 0, get the array it is open on, | |
508 | * else use devnum. >=0 -> major9. <0..... | |
509 | */ | |
510 | extern int sysfs_open(int devnum, char *devname, char *attr); | |
511 | extern void sysfs_init(struct mdinfo *mdi, int fd, int devnum); | |
512 | extern void sysfs_free(struct mdinfo *sra); | |
513 | extern struct mdinfo *sysfs_read(int fd, int devnum, unsigned long options); | |
514 | extern int sysfs_attr_match(const char *attr, const char *str); | |
515 | extern int sysfs_match_word(const char *word, char **list); | |
516 | extern int sysfs_set_str(struct mdinfo *sra, struct mdinfo *dev, | |
517 | char *name, char *val); | |
518 | extern int sysfs_set_num(struct mdinfo *sra, struct mdinfo *dev, | |
519 | char *name, unsigned long long val); | |
520 | extern int sysfs_set_num_signed(struct mdinfo *sra, struct mdinfo *dev, | |
521 | char *name, long long val); | |
522 | extern int sysfs_uevent(struct mdinfo *sra, char *event); | |
523 | extern int sysfs_get_fd(struct mdinfo *sra, struct mdinfo *dev, | |
524 | char *name); | |
525 | extern int sysfs_fd_get_ll(int fd, unsigned long long *val); | |
526 | extern int sysfs_get_ll(struct mdinfo *sra, struct mdinfo *dev, | |
527 | char *name, unsigned long long *val); | |
528 | extern int sysfs_fd_get_str(int fd, char *val, int size); | |
529 | extern int sysfs_attribute_available(struct mdinfo *sra, struct mdinfo *dev, | |
530 | char *name); | |
531 | extern int sysfs_get_str(struct mdinfo *sra, struct mdinfo *dev, | |
532 | char *name, char *val, int size); | |
533 | extern int sysfs_set_safemode(struct mdinfo *sra, unsigned long ms); | |
534 | extern int sysfs_set_array(struct mdinfo *info, int vers); | |
535 | extern int sysfs_add_disk(struct mdinfo *sra, struct mdinfo *sd, int resume); | |
536 | extern int sysfs_disk_to_scsi_id(int fd, __u32 *id); | |
537 | extern int sysfs_unique_holder(int devnum, long rdev); | |
538 | extern int sysfs_freeze_array(struct mdinfo *sra); | |
539 | extern int load_sys(char *path, char *buf); | |
540 | extern int reshape_prepare_fdlist(char *devname, | |
541 | struct mdinfo *sra, | |
542 | int raid_disks, | |
543 | int nrdisks, | |
544 | unsigned long blocks, | |
545 | char *backup_file, | |
546 | int *fdlist, | |
547 | unsigned long long *offsets); | |
548 | extern void reshape_free_fdlist(int *fdlist, | |
549 | unsigned long long *offsets, | |
550 | int size); | |
551 | extern int reshape_open_backup_file(char *backup, | |
552 | int fd, | |
553 | char *devname, | |
554 | long blocks, | |
555 | int *fdlist, | |
556 | unsigned long long *offsets, | |
557 | int restart); | |
558 | extern unsigned long compute_backup_blocks(int nchunk, int ochunk, | |
559 | unsigned int ndata, unsigned int odata); | |
560 | ||
561 | extern int save_stripes(int *source, unsigned long long *offsets, | |
562 | int raid_disks, int chunk_size, int level, int layout, | |
563 | int nwrites, int *dest, | |
564 | unsigned long long start, unsigned long long length, | |
565 | char *buf); | |
566 | extern int restore_stripes(int *dest, unsigned long long *offsets, | |
567 | int raid_disks, int chunk_size, int level, int layout, | |
568 | int source, unsigned long long read_offset, | |
569 | unsigned long long start, unsigned long long length, | |
570 | char *src_buf); | |
571 | ||
572 | #ifndef Sendmail | |
573 | #define Sendmail "/usr/lib/sendmail -t" | |
574 | #endif | |
575 | ||
576 | #define SYSLOG_FACILITY LOG_DAEMON | |
577 | ||
578 | extern char *map_num(mapping_t *map, int num); | |
579 | extern int map_name(mapping_t *map, char *name); | |
580 | extern mapping_t r5layout[], r6layout[], pers[], modes[], faultylayout[]; | |
581 | ||
582 | extern char *map_dev_preferred(int major, int minor, int create, | |
583 | char *prefer); | |
584 | static inline char *map_dev(int major, int minor, int create) | |
585 | { | |
586 | return map_dev_preferred(major, minor, create, NULL); | |
587 | } | |
588 | ||
589 | struct active_array; | |
590 | struct metadata_update; | |
591 | ||
592 | ||
593 | /* 'struct reshape' records the intermediate states | |
594 | * a general reshape. | |
595 | * The starting geometry is converted to the 'before' geometry | |
596 | * by at most an atomic level change. They could be the same. | |
597 | * Similarly the 'after' geometry is converted to the final | |
598 | * geometry by at most a level change. | |
599 | * Note that 'before' and 'after' must have the same level. | |
600 | * 'blocks' is the minimum number of sectors for a reshape unit. | |
601 | * This will be a multiple of the stripe size in each of the | |
602 | * 'before' and 'after' geometries. | |
603 | * If 'blocks' is 0, no restriping is necessary. | |
604 | */ | |
605 | struct reshape { | |
606 | int level; | |
607 | int parity; /* number of parity blocks/devices */ | |
608 | struct { | |
609 | int layout; | |
610 | int data_disks; | |
611 | } before, after; | |
612 | unsigned long long backup_blocks; | |
613 | unsigned long long stripes; /* number of old stripes that comprise 'blocks'*/ | |
614 | unsigned long long new_size; /* New size of array in sectors */ | |
615 | }; | |
616 | ||
617 | /* A superswitch provides entry point the a metadata handler. | |
618 | * | |
619 | * The super_switch primarily operates on some "metadata" that | |
620 | * is accessed via the 'supertype'. | |
621 | * This metadata has one of three possible sources. | |
622 | * 1/ It is read from a single device. In this case it may not completely | |
623 | * describe the array or arrays as some information might be on other | |
624 | * devices. | |
625 | * 2/ It is read from all devices in a container. In this case all | |
626 | * information is present. | |
627 | * 3/ It is created by ->init_super / ->add_to_super. In this case it will | |
628 | * be complete once enough ->add_to_super calls have completed. | |
629 | * | |
630 | * When creating an array inside a container, the metadata will be | |
631 | * formed by a combination of 2 and 3. The metadata or the array is read, | |
632 | * then new information is added. | |
633 | * | |
634 | * The metadata must sometimes have a concept of a 'current' array | |
635 | * and a 'current' device. | |
636 | * The 'current' array is set by init_super to be the newly created array, | |
637 | * or is set by super_by_fd when it finds it is looking at an array inside | |
638 | * a container. | |
639 | * | |
640 | * The 'current' device is either the device that the metadata was read from | |
641 | * in case 1, or the last device added by add_to_super in case 3. | |
642 | * Case 2 does not identify a 'current' device. | |
643 | */ | |
644 | extern struct superswitch { | |
645 | ||
646 | /* Used to report details of metadata read from a component | |
647 | * device. ->load_super has been called. | |
648 | */ | |
649 | void (*examine_super)(struct supertype *st, char *homehost); | |
650 | void (*brief_examine_super)(struct supertype *st, int verbose); | |
651 | void (*brief_examine_subarrays)(struct supertype *st, int verbose); | |
652 | void (*export_examine_super)(struct supertype *st); | |
653 | ||
654 | /* Used to report details of an active array. | |
655 | * ->load_super was possibly given a 'component' string. | |
656 | */ | |
657 | void (*detail_super)(struct supertype *st, char *homehost); | |
658 | void (*brief_detail_super)(struct supertype *st); | |
659 | void (*export_detail_super)(struct supertype *st); | |
660 | ||
661 | /* Optional: platform hardware / firmware details */ | |
662 | int (*detail_platform)(int verbose, int enumerate_only, char *controller_path); | |
663 | int (*export_detail_platform)(int verbose, char *controller_path); | |
664 | ||
665 | /* Used: | |
666 | * to get uuid to storing in bitmap metadata | |
667 | * and 'reshape' backup-data metadata | |
668 | * To see if a device is being re-added to an array it was part of. | |
669 | */ | |
670 | void (*uuid_from_super)(struct supertype *st, int uuid[4]); | |
671 | ||
672 | /* Extract generic details from metadata. This could be details about | |
673 | * the container, or about an individual array within the container. | |
674 | * The determination is made either by: | |
675 | * load_super being given a 'component' string. | |
676 | * validate_geometry determining what to create. | |
677 | * The info includes both array information and device information. | |
678 | * The particular device should be: | |
679 | * The last device added by add_to_super | |
680 | * The device the metadata was loaded from by load_super | |
681 | * If 'map' is present, then it is an array raid_disks long | |
682 | * (raid_disk must already be set and correct) and it is filled | |
683 | * with 1 for slots that are thought to be active and 0 for slots which | |
684 | * appear to be failed/missing. | |
685 | * *info is zeroed out before data is added. | |
686 | */ | |
687 | void (*getinfo_super)(struct supertype *st, struct mdinfo *info, char *map); | |
688 | struct mdinfo *(*getinfo_super_disks)(struct supertype *st); | |
689 | /* Check if the given metadata is flagged as belonging to "this" | |
690 | * host. 0 for 'no', 1 for 'yes', -1 for "Don't record homehost" | |
691 | */ | |
692 | int (*match_home)(struct supertype *st, char *homehost); | |
693 | ||
694 | /* Make one of several generic modifications to metadata | |
695 | * prior to assembly (or other times). | |
696 | * sparc2.2 - first bug in early 0.90 metadata | |
697 | * super-minor - change name of 0.90 metadata | |
698 | * summaries - 'correct' any redundant data | |
699 | * resync - mark array as dirty to trigger a resync. | |
700 | * uuid - set new uuid - only 0.90 or 1.x | |
701 | * name - change the name of the array (where supported) | |
702 | * homehost - change which host this array is tied to. | |
703 | * devicesize - If metadata is at start of device, change recorded | |
704 | * device size to match actual device size | |
705 | * byteorder - swap bytes for 0.90 metadata | |
706 | * | |
707 | * force-one - mark that device as uptodate, not old or failed. | |
708 | * force-array - mark array as clean if it would not otherwise | |
709 | * assemble | |
710 | * assemble - not sure how this is different from force-one... | |
711 | * linear-grow-new - add a new device to a linear array, but don't | |
712 | * change the size: so superblock still matches | |
713 | * linear-grow-update - now change the size of the array. | |
714 | * writemostly - set the WriteMostly1 bit in the superblock devflags | |
715 | * readwrite - clear the WriteMostly1 bit in the superblock devflags | |
716 | */ | |
717 | int (*update_super)(struct supertype *st, struct mdinfo *info, | |
718 | char *update, | |
719 | char *devname, int verbose, | |
720 | int uuid_set, char *homehost); | |
721 | ||
722 | /* Create new metadata for new array as described. This could | |
723 | * be a new container, or an array in a pre-existing container. | |
724 | * Also used to zero metadata prior to writing it to invalidate old | |
725 | * metadata. | |
726 | */ | |
727 | int (*init_super)(struct supertype *st, mdu_array_info_t *info, | |
728 | unsigned long long size, char *name, | |
729 | char *homehost, int *uuid); | |
730 | ||
731 | /* update the metadata to include new device, either at create or | |
732 | * when hot-adding a spare. | |
733 | */ | |
734 | int (*add_to_super)(struct supertype *st, mdu_disk_info_t *dinfo, | |
735 | int fd, char *devname); | |
736 | /* update the metadata to delete a device, | |
737 | * when hot-removing. | |
738 | */ | |
739 | int (*remove_from_super)(struct supertype *st, mdu_disk_info_t *dinfo); | |
740 | ||
741 | /* Write metadata to one device when fixing problems or adding | |
742 | * a new device. | |
743 | */ | |
744 | int (*store_super)(struct supertype *st, int fd); | |
745 | ||
746 | /* Write all metadata for this array. | |
747 | */ | |
748 | int (*write_init_super)(struct supertype *st); | |
749 | int (*compare_super)(struct supertype *st, struct supertype *tst); | |
750 | int (*load_super)(struct supertype *st, int fd, char *devname); | |
751 | int (*load_container)(struct supertype *st, int fd, char *devname); | |
752 | struct supertype * (*match_metadata_desc)(char *arg); | |
753 | __u64 (*avail_size)(struct supertype *st, __u64 size, | |
754 | unsigned long long data_offset); | |
755 | unsigned long long (*min_acceptable_spare_size)(struct supertype *st); | |
756 | int (*add_internal_bitmap)(struct supertype *st, int *chunkp, | |
757 | int delay, int write_behind, | |
758 | unsigned long long size, int may_change, int major); | |
759 | void (*locate_bitmap)(struct supertype *st, int fd); | |
760 | int (*write_bitmap)(struct supertype *st, int fd); | |
761 | void (*free_super)(struct supertype *st); | |
762 | ||
763 | /* validate_geometry is called with an st returned by | |
764 | * match_metadata_desc. | |
765 | * It should check that the geometry described in compatible with | |
766 | * the metadata type. It will be called repeatedly as devices | |
767 | * added to validate changing size and new devices. If there are | |
768 | * inter-device dependencies, it should record sufficient details | |
769 | * so these can be validated. | |
770 | * Both 'size' and '*freesize' are in sectors. chunk is KiB. | |
771 | * Return value is: | |
772 | * 1: everything is OK | |
773 | * 0: not OK for some reason - if 'verbose', then error was reported. | |
774 | * -1: st->sb was NULL, 'subdev' is a member of a container of this | |
775 | * types, but array is not acceptable for some reason | |
776 | * message was reported even if verbose is 0. | |
777 | */ | |
778 | int (*validate_geometry)(struct supertype *st, int level, int layout, | |
779 | int raiddisks, | |
780 | int *chunk, unsigned long long size, | |
781 | unsigned long long data_offset, | |
782 | char *subdev, unsigned long long *freesize, | |
783 | int verbose); | |
784 | ||
785 | struct mdinfo *(*container_content)(struct supertype *st, char *subarray); | |
786 | /* query the supertype for default geometry */ | |
787 | void (*default_geometry)(struct supertype *st, int *level, int *layout, int *chunk); /* optional */ | |
788 | /* Permit subarray's to be deleted from inactive containers */ | |
789 | int (*kill_subarray)(struct supertype *st); /* optional */ | |
790 | /* Permit subarray's to be modified */ | |
791 | int (*update_subarray)(struct supertype *st, char *subarray, | |
792 | char *update, struct mddev_ident *ident); /* optional */ | |
793 | /* Check if reshape is supported for this external format. | |
794 | * st is obtained from super_by_fd() where st->subarray[0] is | |
795 | * initialized to indicate if reshape is being performed at the | |
796 | * container or subarray level | |
797 | */ | |
798 | #define APPLY_METADATA_CHANGES 1 | |
799 | #define ROLLBACK_METADATA_CHANGES 0 | |
800 | ||
801 | int (*reshape_super)(struct supertype *st, | |
802 | unsigned long long size, int level, | |
803 | int layout, int chunksize, int raid_disks, | |
804 | int delta_disks, char *backup, char *dev, | |
805 | int direction, | |
806 | int verbose); /* optional */ | |
807 | int (*manage_reshape)( /* optional */ | |
808 | int afd, struct mdinfo *sra, struct reshape *reshape, | |
809 | struct supertype *st, unsigned long blocks, | |
810 | int *fds, unsigned long long *offsets, | |
811 | int dests, int *destfd, unsigned long long *destoffsets); | |
812 | ||
813 | /* for mdmon */ | |
814 | int (*open_new)(struct supertype *c, struct active_array *a, | |
815 | char *inst); | |
816 | ||
817 | /* Tell the metadata handler the current state of the array. | |
818 | * This covers whether it is known to be consistent (no pending writes) | |
819 | * and how far along a resync is known to have progressed | |
820 | * (in a->resync_start). | |
821 | * resync status is really irrelevant if the array is not consistent, | |
822 | * but some metadata (DDF!) have a place to record the distinction. | |
823 | * If 'consistent' is '2', then the array can mark it dirty if a | |
824 | * resync/recovery/whatever is required, or leave it clean if not. | |
825 | * Return value is 0 dirty (not consistent) and 1 if clean. | |
826 | * it is only really important if consistent is passed in as '2'. | |
827 | */ | |
828 | int (*set_array_state)(struct active_array *a, int consistent); | |
829 | ||
830 | /* When the state of a device might have changed, we call set_disk to | |
831 | * tell the metadata what the current state is. | |
832 | * Typically this happens on spare->in_sync and (spare|in_sync)->faulty | |
833 | * transitions. | |
834 | * set_disk might be called when the state of the particular disk has | |
835 | * not in fact changed. | |
836 | */ | |
837 | void (*set_disk)(struct active_array *a, int n, int state); | |
838 | void (*sync_metadata)(struct supertype *st); | |
839 | void (*process_update)(struct supertype *st, | |
840 | struct metadata_update *update); | |
841 | void (*prepare_update)(struct supertype *st, | |
842 | struct metadata_update *update); | |
843 | ||
844 | /* activate_spare will check if the array is degraded and, if it | |
845 | * is, try to find some spare space in the container. | |
846 | * On success, it add appropriate updates (For process_update) to | |
847 | * to the 'updates' list and returns a list of 'mdinfo' identifying | |
848 | * the device, or devices as there might be multiple missing | |
849 | * devices and multiple spares available. | |
850 | */ | |
851 | struct mdinfo *(*activate_spare)(struct active_array *a, | |
852 | struct metadata_update **updates); | |
853 | /* | |
854 | * Return statically allocated string that represents metadata specific | |
855 | * controller domain of the disk. The domain is used in disk domain | |
856 | * matching functions. Disks belong to the same domain if the they have | |
857 | * the same domain from mdadm.conf and belong the same metadata domain. | |
858 | * Returning NULL or not providing this handler means that metadata | |
859 | * does not distinguish the differences between disks that belong to | |
860 | * different controllers. They are in the domain specified by | |
861 | * configuration file (mdadm.conf). | |
862 | * In case when the metadata has the notion of domains based on disk | |
863 | * it shall return NULL for disks that do not belong to the controller | |
864 | * the supported domains. Such disks will form another domain and won't | |
865 | * be mixed with supported ones. | |
866 | */ | |
867 | const char *(*get_disk_controller_domain)(const char *path); | |
868 | ||
869 | /* for external backup area */ | |
870 | int (*recover_backup)(struct supertype *st, struct mdinfo *info); | |
871 | ||
872 | int swapuuid; /* true if uuid is bigending rather than hostendian */ | |
873 | int external; | |
874 | const char *name; /* canonical metadata name */ | |
875 | } *superlist[]; | |
876 | ||
877 | extern struct superswitch super0, super1; | |
878 | extern struct superswitch super_imsm, super_ddf; | |
879 | extern struct superswitch mbr, gpt; | |
880 | ||
881 | struct metadata_update { | |
882 | int len; | |
883 | char *buf; | |
884 | void *space; /* allocated space that monitor will use */ | |
885 | void **space_list; /* list of allocated spaces that monitor can | |
886 | * use or that it returned. | |
887 | */ | |
888 | struct metadata_update *next; | |
889 | }; | |
890 | ||
891 | /* A supertype holds a particular collection of metadata. | |
892 | * It identifies the metadata type by the superswitch, and the particular | |
893 | * sub-version of that metadata type. | |
894 | * metadata read in or created is stored in 'sb' and 'info'. | |
895 | * There are also fields used by mdmon to track containers. | |
896 | * | |
897 | * A supertype may refer to: | |
898 | * Just an array, possibly in a container | |
899 | * A container, not identifying any particular array | |
900 | * Info read from just one device, not yet fully describing the array/container. | |
901 | * | |
902 | * | |
903 | * A supertype is created by: | |
904 | * super_by_fd | |
905 | * guess_super | |
906 | * dup_super | |
907 | */ | |
908 | struct supertype { | |
909 | struct superswitch *ss; | |
910 | int minor_version; | |
911 | int max_devs; | |
912 | int container_dev; /* devnum of container */ | |
913 | void *sb; | |
914 | void *info; | |
915 | int ignore_hw_compat; /* used to inform metadata handlers that it should ignore | |
916 | HW/firmware related incompatability to load metadata. | |
917 | Used when examining metadata to display content of disk | |
918 | when user has no hw/firmare compatible system. | |
919 | */ | |
920 | struct metadata_update *updates; | |
921 | struct metadata_update **update_tail; | |
922 | ||
923 | /* extra stuff used by mdmon */ | |
924 | struct active_array *arrays; | |
925 | int sock; /* listen to external programs */ | |
926 | int devnum; | |
927 | char *devname; /* e.g. md0. This appears in metadata_verison: | |
928 | * external:/md0/12 | |
929 | */ | |
930 | int devcnt; | |
931 | int retry_soon; | |
932 | ||
933 | struct mdinfo *devs; | |
934 | ||
935 | }; | |
936 | ||
937 | extern struct supertype *super_by_fd(int fd, char **subarray); | |
938 | enum guess_types { guess_any, guess_array, guess_partitions }; | |
939 | extern struct supertype *guess_super_type(int fd, enum guess_types guess_type); | |
940 | static inline struct supertype *guess_super(int fd) { | |
941 | return guess_super_type(fd, guess_any); | |
942 | } | |
943 | extern struct supertype *dup_super(struct supertype *st); | |
944 | extern int get_dev_size(int fd, char *dname, unsigned long long *sizep); | |
945 | extern int must_be_container(int fd); | |
946 | extern int dev_size_from_id(dev_t id, unsigned long long *size); | |
947 | extern void get_one_disk(int mdfd, mdu_array_info_t *ainf, | |
948 | mdu_disk_info_t *disk); | |
949 | void wait_for(char *dev, int fd); | |
950 | ||
951 | /* | |
952 | * Data structures for policy management. | |
953 | * Each device can have a policy structure that lists | |
954 | * various name/value pairs each possibly with a metadata associated. | |
955 | * The policy list is sorted by name/value/metadata | |
956 | */ | |
957 | struct dev_policy { | |
958 | struct dev_policy *next; | |
959 | char *name; /* None of these strings are allocated. They are | |
960 | * all just references to strings which are known | |
961 | * to exist elsewhere. | |
962 | * name and metadata can be compared by address equality. | |
963 | */ | |
964 | const char *metadata; | |
965 | const char *value; | |
966 | }; | |
967 | ||
968 | extern char pol_act[], pol_domain[], pol_metadata[], pol_auto[]; | |
969 | ||
970 | /* iterate over the sublist starting at list, having the same | |
971 | * 'name' as 'list', and matching the given metadata (Where | |
972 | * NULL matches anything | |
973 | */ | |
974 | #define pol_for_each(item, list, _metadata) \ | |
975 | for (item = list; \ | |
976 | item && item->name == list->name; \ | |
977 | item = item->next) \ | |
978 | if (!(!_metadata || !item->metadata || _metadata == item->metadata)) \ | |
979 | ; else | |
980 | ||
981 | /* | |
982 | * policy records read from mdadm are largely just name-value pairs. | |
983 | * The names are constants, not strdupped | |
984 | */ | |
985 | struct pol_rule { | |
986 | struct pol_rule *next; | |
987 | char *type; /* rule_policy or rule_part */ | |
988 | struct rule { | |
989 | struct rule *next; | |
990 | char *name; | |
991 | char *value; | |
992 | char *dups; /* duplicates of 'value' with a partNN appended */ | |
993 | } *rule; | |
994 | }; | |
995 | ||
996 | extern char rule_policy[], rule_part[]; | |
997 | extern char rule_path[], rule_type[]; | |
998 | extern char type_part[], type_disk[]; | |
999 | ||
1000 | extern void policyline(char *line, char *type); | |
1001 | extern void policy_add(char *type, ...); | |
1002 | extern void policy_free(void); | |
1003 | ||
1004 | extern struct dev_policy *path_policy(char *path, char *type); | |
1005 | extern struct dev_policy *disk_policy(struct mdinfo *disk); | |
1006 | extern struct dev_policy *devnum_policy(int dev); | |
1007 | extern void dev_policy_free(struct dev_policy *p); | |
1008 | ||
1009 | //extern void pol_new(struct dev_policy **pol, char *name, char *val, char *metadata); | |
1010 | extern void pol_add(struct dev_policy **pol, char *name, char *val, char *metadata); | |
1011 | extern struct dev_policy *pol_find(struct dev_policy *pol, char *name); | |
1012 | ||
1013 | enum policy_action { | |
1014 | act_default, | |
1015 | act_include, | |
1016 | act_re_add, | |
1017 | act_spare, /* This only applies to bare devices */ | |
1018 | act_spare_same_slot, /* this allows non-bare devices, | |
1019 | * but only if recent removal */ | |
1020 | act_force_spare, /* this allow non-bare devices in any case */ | |
1021 | act_err | |
1022 | }; | |
1023 | ||
1024 | extern int policy_action_allows(struct dev_policy *plist, const char *metadata, | |
1025 | enum policy_action want); | |
1026 | extern int disk_action_allows(struct mdinfo *disk, const char *metadata, | |
1027 | enum policy_action want); | |
1028 | ||
1029 | struct domainlist { | |
1030 | struct domainlist *next; | |
1031 | const char *dom; | |
1032 | }; | |
1033 | ||
1034 | extern int domain_test(struct domainlist *dom, struct dev_policy *pol, | |
1035 | const char *metadata); | |
1036 | extern struct domainlist *domain_from_array(struct mdinfo *mdi, | |
1037 | const char *metadata); | |
1038 | extern void domainlist_add_dev(struct domainlist **dom, int devnum, | |
1039 | const char *metadata); | |
1040 | extern void domain_free(struct domainlist *dl); | |
1041 | extern void domain_merge(struct domainlist **domp, struct dev_policy *pol, | |
1042 | const char *metadata); | |
1043 | void domain_add(struct domainlist **domp, char *domain); | |
1044 | ||
1045 | extern void policy_save_path(char *id_path, struct map_ent *array); | |
1046 | extern int policy_check_path(struct mdinfo *disk, struct map_ent *array); | |
1047 | ||
1048 | #if __GNUC__ < 3 | |
1049 | struct stat64; | |
1050 | #endif | |
1051 | ||
1052 | #define HAVE_NFTW we assume | |
1053 | #define HAVE_FTW | |
1054 | ||
1055 | #ifdef __UCLIBC__ | |
1056 | # include <features.h> | |
1057 | # ifndef __UCLIBC_HAS_LFS__ | |
1058 | # define lseek64 lseek | |
1059 | # endif | |
1060 | # ifndef __UCLIBC_HAS_FTW__ | |
1061 | # undef HAVE_FTW | |
1062 | # undef HAVE_NFTW | |
1063 | # endif | |
1064 | #endif | |
1065 | ||
1066 | #ifdef __dietlibc__ | |
1067 | # undef HAVE_NFTW | |
1068 | #endif | |
1069 | ||
1070 | #if defined(__KLIBC__) | |
1071 | # undef HAVE_NFTW | |
1072 | # undef HAVE_FTW | |
1073 | #endif | |
1074 | ||
1075 | #ifndef HAVE_NFTW | |
1076 | # define FTW_PHYS 1 | |
1077 | # ifndef HAVE_FTW | |
1078 | struct FTW {}; | |
1079 | # endif | |
1080 | #endif | |
1081 | ||
1082 | #ifdef HAVE_FTW | |
1083 | # include <ftw.h> | |
1084 | #endif | |
1085 | ||
1086 | extern int add_dev(const char *name, const struct stat *stb, int flag, struct FTW *s); | |
1087 | ||
1088 | ||
1089 | extern int Manage_ro(char *devname, int fd, int readonly); | |
1090 | extern int Manage_runstop(char *devname, int fd, int runstop, int quiet, | |
1091 | int will_retry); | |
1092 | extern int Manage_subdevs(char *devname, int fd, | |
1093 | struct mddev_dev *devlist, int verbose, int test, | |
1094 | char *update, int force); | |
1095 | extern int autodetect(void); | |
1096 | extern int Grow_Add_device(char *devname, int fd, char *newdev); | |
1097 | extern int Grow_addbitmap(char *devname, int fd, | |
1098 | struct context *c, struct shape *s); | |
1099 | extern int Grow_reshape(char *devname, int fd, | |
1100 | struct mddev_dev *devlist, | |
1101 | struct context *c, struct shape *s); | |
1102 | extern int Grow_restart(struct supertype *st, struct mdinfo *info, | |
1103 | int *fdlist, int cnt, char *backup_file, int verbose); | |
1104 | extern int Grow_continue(int mdfd, struct supertype *st, | |
1105 | struct mdinfo *info, char *backup_file, | |
1106 | int freeze_reshape); | |
1107 | ||
1108 | extern int restore_backup(struct supertype *st, | |
1109 | struct mdinfo *content, | |
1110 | int working_disks, | |
1111 | int spares, | |
1112 | char *backup_file, | |
1113 | int verbose); | |
1114 | extern int Grow_continue_command(char *devname, int fd, | |
1115 | char *backup_file, int verbose); | |
1116 | ||
1117 | extern int Assemble(struct supertype *st, char *mddev, | |
1118 | struct mddev_ident *ident, | |
1119 | struct mddev_dev *devlist, | |
1120 | struct context *c); | |
1121 | ||
1122 | extern int Build(char *mddev, struct mddev_dev *devlist, | |
1123 | struct shape *s, struct context *c); | |
1124 | ||
1125 | extern int Create(struct supertype *st, char *mddev, | |
1126 | char *name, int *uuid, | |
1127 | int subdevs, struct mddev_dev *devlist, | |
1128 | struct shape *s, | |
1129 | struct context *c); | |
1130 | ||
1131 | extern int Detail(char *dev, struct context *c); | |
1132 | extern int Detail_Platform(struct superswitch *ss, int scan, int verbose, int export, char *controller_path); | |
1133 | extern int Query(char *dev); | |
1134 | extern int Examine(struct mddev_dev *devlist, struct context *c, | |
1135 | struct supertype *forcest); | |
1136 | extern int Monitor(struct mddev_dev *devlist, | |
1137 | char *mailaddr, char *alert_cmd, | |
1138 | struct context *c, | |
1139 | int daemonise, int oneshot, | |
1140 | int dosyslog, char *pidfile, int increments, | |
1141 | int share); | |
1142 | ||
1143 | extern int Kill(char *dev, struct supertype *st, int force, int verbose, int noexcl); | |
1144 | extern int Kill_subarray(char *dev, char *subarray, int verbose); | |
1145 | extern int Update_subarray(char *dev, char *subarray, char *update, struct mddev_ident *ident, int quiet); | |
1146 | extern int Wait(char *dev); | |
1147 | extern int WaitClean(char *dev, int sock, int verbose); | |
1148 | ||
1149 | extern int Incremental(char *devname, struct context *c, | |
1150 | struct supertype *st); | |
1151 | extern void RebuildMap(void); | |
1152 | extern int IncrementalScan(int verbose); | |
1153 | extern int IncrementalRemove(char *devname, char *path, int verbose); | |
1154 | extern int CreateBitmap(char *filename, int force, char uuid[16], | |
1155 | unsigned long chunksize, unsigned long daemon_sleep, | |
1156 | unsigned long write_behind, | |
1157 | unsigned long long array_size, | |
1158 | int major); | |
1159 | extern int ExamineBitmap(char *filename, int brief, struct supertype *st); | |
1160 | extern int Write_rules(char *rule_name); | |
1161 | extern int bitmap_update_uuid(int fd, int *uuid, int swap); | |
1162 | extern unsigned long bitmap_sectors(struct bitmap_super_s *bsb); | |
1163 | ||
1164 | extern int md_get_version(int fd); | |
1165 | extern int get_linux_version(void); | |
1166 | extern int mdadm_version(char *version); | |
1167 | extern unsigned long long parse_size(char *size); | |
1168 | extern int parse_uuid(char *str, int uuid[4]); | |
1169 | extern int parse_layout_10(char *layout); | |
1170 | extern int parse_layout_faulty(char *layout); | |
1171 | extern long parse_num(char *num); | |
1172 | extern int check_ext2(int fd, char *name); | |
1173 | extern int check_reiser(int fd, char *name); | |
1174 | extern int check_raid(int fd, char *name); | |
1175 | extern int check_partitions(int fd, char *dname, | |
1176 | unsigned long long freesize, | |
1177 | unsigned long long size); | |
1178 | ||
1179 | extern int get_mdp_major(void); | |
1180 | extern int dev_open(char *dev, int flags); | |
1181 | extern int open_dev(int devnum); | |
1182 | extern int open_dev_flags(int devnum, int flags); | |
1183 | extern int open_dev_excl(int devnum); | |
1184 | extern int is_standard(char *dev, int *nump); | |
1185 | extern int same_dev(char *one, char *two); | |
1186 | extern int compare_paths (char* path1,char* path2); | |
1187 | ||
1188 | extern int parse_auto(char *str, char *msg, int config); | |
1189 | extern struct mddev_ident *conf_get_ident(char *dev); | |
1190 | extern struct mddev_dev *conf_get_devs(void); | |
1191 | extern int conf_test_dev(char *devname); | |
1192 | extern int conf_test_metadata(const char *version, struct dev_policy *pol, int is_homehost); | |
1193 | extern struct createinfo *conf_get_create_info(void); | |
1194 | extern void set_conffile(char *file); | |
1195 | extern char *conf_get_mailaddr(void); | |
1196 | extern char *conf_get_mailfrom(void); | |
1197 | extern char *conf_get_program(void); | |
1198 | extern char *conf_get_homehost(int *require_homehostp); | |
1199 | extern char *conf_line(FILE *file); | |
1200 | extern char *conf_word(FILE *file, int allow_key); | |
1201 | extern void print_quoted(char *str); | |
1202 | extern void print_escape(char *str); | |
1203 | extern int conf_name_is_free(char *name); | |
1204 | extern int conf_verify_devnames(struct mddev_ident *array_list); | |
1205 | extern int devname_matches(char *name, char *match); | |
1206 | extern struct mddev_ident *conf_match(struct supertype *st, | |
1207 | struct mdinfo *info, | |
1208 | char *devname, | |
1209 | int verbose, int *rvp); | |
1210 | extern int experimental(void); | |
1211 | ||
1212 | extern void free_line(char *line); | |
1213 | extern int match_oneof(char *devices, char *devname); | |
1214 | extern void uuid_from_super(int uuid[4], mdp_super_t *super); | |
1215 | extern const int uuid_zero[4]; | |
1216 | extern int same_uuid(int a[4], int b[4], int swapuuid); | |
1217 | extern void copy_uuid(void *a, int b[4], int swapuuid); | |
1218 | extern char *__fname_from_uuid(int id[4], int swap, char *buf, char sep); | |
1219 | extern char *fname_from_uuid(struct supertype *st, | |
1220 | struct mdinfo *info, char *buf, char sep); | |
1221 | extern unsigned long calc_csum(void *super, int bytes); | |
1222 | extern int enough(int level, int raid_disks, int layout, int clean, | |
1223 | char *avail); | |
1224 | extern int enough_fd(int fd); | |
1225 | extern int ask(char *mesg); | |
1226 | extern unsigned long long get_component_size(int fd); | |
1227 | extern void remove_partitions(int fd); | |
1228 | extern int test_partition(int fd); | |
1229 | extern int test_partition_from_id(dev_t id); | |
1230 | extern int get_data_disks(int level, int layout, int raid_disks); | |
1231 | extern unsigned long long calc_array_size(int level, int raid_disks, int layout, | |
1232 | int chunksize, unsigned long long devsize); | |
1233 | extern int flush_metadata_updates(struct supertype *st); | |
1234 | extern void append_metadata_update(struct supertype *st, void *buf, int len); | |
1235 | extern int assemble_container_content(struct supertype *st, int mdfd, | |
1236 | struct mdinfo *content, | |
1237 | struct context *c, | |
1238 | char *chosen_name); | |
1239 | extern struct mdinfo *container_choose_spares(struct supertype *st, | |
1240 | unsigned long long min_size, | |
1241 | struct domainlist *domlist, | |
1242 | char *spare_group, | |
1243 | const char *metadata, int get_one); | |
1244 | extern int move_spare(char *from_devname, char *to_devname, dev_t devid); | |
1245 | extern int add_disk(int mdfd, struct supertype *st, | |
1246 | struct mdinfo *sra, struct mdinfo *info); | |
1247 | extern int remove_disk(int mdfd, struct supertype *st, | |
1248 | struct mdinfo *sra, struct mdinfo *info); | |
1249 | extern int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info); | |
1250 | unsigned long long min_recovery_start(struct mdinfo *array); | |
1251 | ||
1252 | extern char *human_size(long long bytes); | |
1253 | extern char *human_size_brief(long long bytes, int prefix); | |
1254 | extern void print_r10_layout(int layout); | |
1255 | ||
1256 | #define NoMdDev (1<<23) | |
1257 | extern int find_free_devnum(int use_partitions); | |
1258 | ||
1259 | extern void put_md_name(char *name); | |
1260 | extern char *get_md_name(int dev); | |
1261 | ||
1262 | extern char DefaultConfFile[]; | |
1263 | ||
1264 | extern int create_mddev(char *dev, char *name, int autof, int trustworthy, | |
1265 | char *chosen); | |
1266 | /* values for 'trustworthy' */ | |
1267 | #define LOCAL 1 | |
1268 | #define LOCAL_ANY 10 | |
1269 | #define FOREIGN 2 | |
1270 | #define METADATA 3 | |
1271 | extern int open_mddev(char *dev, int report_errors); | |
1272 | extern int open_container(int fd); | |
1273 | extern int is_container_member(struct mdstat_ent *ent, char *devname); | |
1274 | extern int is_subarray_active(char *subarray, char *devname); | |
1275 | extern int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet); | |
1276 | extern struct superswitch *version_to_superswitch(char *vers); | |
1277 | ||
1278 | extern int mdmon_running(int devnum); | |
1279 | extern int mdmon_pid(int devnum); | |
1280 | extern int check_env(char *name); | |
1281 | extern __u32 random32(void); | |
1282 | extern int start_mdmon(int devnum); | |
1283 | ||
1284 | extern int child_monitor(int afd, struct mdinfo *sra, struct reshape *reshape, | |
1285 | struct supertype *st, unsigned long stripes, | |
1286 | int *fds, unsigned long long *offsets, | |
1287 | int dests, int *destfd, unsigned long long *destoffsets); | |
1288 | void abort_reshape(struct mdinfo *sra); | |
1289 | ||
1290 | extern char *devnum2devname(int num); | |
1291 | extern void fmt_devname(char *name, int num); | |
1292 | extern int devname2devnum(char *name); | |
1293 | extern int stat2devnum(struct stat *st); | |
1294 | extern int fd2devnum(int fd); | |
1295 | ||
1296 | static inline int dev2major(int d) | |
1297 | { | |
1298 | if (d >= 0) | |
1299 | return MD_MAJOR; | |
1300 | else | |
1301 | return get_mdp_major(); | |
1302 | } | |
1303 | ||
1304 | static inline int dev2minor(int d) | |
1305 | { | |
1306 | if (d >= 0) | |
1307 | return d; | |
1308 | return (-1-d) << MdpMinorShift; | |
1309 | } | |
1310 | ||
1311 | #define _ROUND_UP(val, base) (((val) + (base) - 1) & ~(base - 1)) | |
1312 | #define ROUND_UP(val, base) _ROUND_UP(val, (typeof(val))(base)) | |
1313 | #define ROUND_UP_PTR(ptr, base) ((typeof(ptr)) \ | |
1314 | (ROUND_UP((unsigned long)(ptr), base))) | |
1315 | ||
1316 | static inline int is_subarray(char *vers) | |
1317 | { | |
1318 | /* The version string for a 'subarray' (an array in a container) | |
1319 | * is | |
1320 | * /containername/componentname for normal read-write arrays | |
1321 | * -containername/componentname for arrays which mdmon must not | |
1322 | * reconfigure. They might be read-only | |
1323 | * or might be undergoing reshape etc. | |
1324 | * containername is e.g. md0, md_d1 | |
1325 | * componentname is dependant on the metadata. e.g. '1' 'S1' ... | |
1326 | */ | |
1327 | return (*vers == '/' || *vers == '-'); | |
1328 | } | |
1329 | ||
1330 | static inline char *to_subarray(struct mdstat_ent *ent, char *container) | |
1331 | { | |
1332 | return &ent->metadata_version[10+strlen(container)+1]; | |
1333 | } | |
1334 | ||
1335 | #ifdef DEBUG | |
1336 | #define dprintf(fmt, arg...) \ | |
1337 | fprintf(stderr, fmt, ##arg) | |
1338 | #else | |
1339 | #define dprintf(fmt, arg...) \ | |
1340 | ({ if (0) fprintf(stderr, fmt, ##arg); 0; }) | |
1341 | #endif | |
1342 | #include <assert.h> | |
1343 | #include <stdarg.h> | |
1344 | static inline int xasprintf(char **strp, const char *fmt, ...) { | |
1345 | va_list ap; | |
1346 | int ret; | |
1347 | va_start(ap, fmt); | |
1348 | ret = vasprintf(strp, fmt, ap); | |
1349 | va_end(ap); | |
1350 | assert(ret >= 0); | |
1351 | return ret; | |
1352 | } | |
1353 | ||
1354 | #define pr_err(fmt ...) fprintf(stderr, Name ": " fmt) | |
1355 | #define cont_err(fmt ...) fprintf(stderr, " " fmt) | |
1356 | ||
1357 | void *xmalloc(size_t len); | |
1358 | void *xrealloc(void *ptr, size_t len); | |
1359 | void *xcalloc(size_t num, size_t size); | |
1360 | char *xstrdup(const char *str); | |
1361 | ||
1362 | #define LEVEL_MULTIPATH (-4) | |
1363 | #define LEVEL_LINEAR (-1) | |
1364 | #define LEVEL_FAULTY (-5) | |
1365 | ||
1366 | /* kernel module doesn't know about these */ | |
1367 | #define LEVEL_CONTAINER (-100) | |
1368 | #define LEVEL_UNSUPPORTED (-200) | |
1369 | ||
1370 | ||
1371 | /* faulty stuff */ | |
1372 | ||
1373 | #define WriteTransient 0 | |
1374 | #define ReadTransient 1 | |
1375 | #define WritePersistent 2 | |
1376 | #define ReadPersistent 3 | |
1377 | #define WriteAll 4 /* doesn't go to device */ | |
1378 | #define ReadFixable 5 | |
1379 | #define Modes 6 | |
1380 | ||
1381 | #define ClearErrors 31 | |
1382 | #define ClearFaults 30 | |
1383 | ||
1384 | #define AllPersist 100 /* internal use only */ | |
1385 | #define NoPersist 101 | |
1386 | ||
1387 | #define ModeMask 0x1f | |
1388 | #define ModeShift 5 | |
1389 | ||
1390 | ||
1391 | #ifdef __TINYC__ | |
1392 | #undef minor | |
1393 | #undef major | |
1394 | #undef makedev | |
1395 | #define minor(x) ((x)&0xff) | |
1396 | #define major(x) (((x)>>8)&0xff) | |
1397 | #define makedev(M,m) (((M)<<8) | (m)) | |
1398 | #endif | |
1399 | ||
1400 | /* for raid4/5/6 */ | |
1401 | #define ALGORITHM_LEFT_ASYMMETRIC 0 | |
1402 | #define ALGORITHM_RIGHT_ASYMMETRIC 1 | |
1403 | #define ALGORITHM_LEFT_SYMMETRIC 2 | |
1404 | #define ALGORITHM_RIGHT_SYMMETRIC 3 | |
1405 | ||
1406 | /* Define non-rotating (raid4) algorithms. These allow | |
1407 | * conversion of raid4 to raid5. | |
1408 | */ | |
1409 | #define ALGORITHM_PARITY_0 4 /* P or P,Q are initial devices */ | |
1410 | #define ALGORITHM_PARITY_N 5 /* P or P,Q are final devices. */ | |
1411 | ||
1412 | /* DDF RAID6 layouts differ from md/raid6 layouts in two ways. | |
1413 | * Firstly, the exact positioning of the parity block is slightly | |
1414 | * different between the 'LEFT_*' modes of md and the "_N_*" modes | |
1415 | * of DDF. | |
1416 | * Secondly, or order of datablocks over which the Q syndrome is computed | |
1417 | * is different. | |
1418 | * Consequently we have different layouts for DDF/raid6 than md/raid6. | |
1419 | * These layouts are from the DDFv1.2 spec. | |
1420 | * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but | |
1421 | * leaves RLQ=3 as 'Vendor Specific' | |
1422 | */ | |
1423 | ||
1424 | #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */ | |
1425 | #define ALGORITHM_ROTATING_N_RESTART 9 /* DDF PRL=6 RLQ=2 */ | |
1426 | #define ALGORITHM_ROTATING_N_CONTINUE 10 /*DDF PRL=6 RLQ=3 */ | |
1427 | ||
1428 | ||
1429 | /* For every RAID5 algorithm we define a RAID6 algorithm | |
1430 | * with exactly the same layout for data and parity, and | |
1431 | * with the Q block always on the last device (N-1). | |
1432 | * This allows trivial conversion from RAID5 to RAID6 | |
1433 | */ | |
1434 | #define ALGORITHM_LEFT_ASYMMETRIC_6 16 | |
1435 | #define ALGORITHM_RIGHT_ASYMMETRIC_6 17 | |
1436 | #define ALGORITHM_LEFT_SYMMETRIC_6 18 | |
1437 | #define ALGORITHM_RIGHT_SYMMETRIC_6 19 | |
1438 | #define ALGORITHM_PARITY_0_6 20 | |
1439 | #define ALGORITHM_PARITY_N_6 ALGORITHM_PARITY_N | |
1440 | ||
1441 | /* Define PATH_MAX in case we don't use glibc or standard library does | |
1442 | * not have PATH_MAX defined. Assume max path length is 4K characters. | |
1443 | */ | |
1444 | #ifndef PATH_MAX | |
1445 | #define PATH_MAX 4096 | |
1446 | #endif | |
1447 | ||
1448 | #define RESYNC_NONE -1 | |
1449 | #define RESYNC_DELAYED -2 | |
1450 | #define RESYNC_PENDING -3 | |
1451 | #define RESYNC_UNKNOWN -4 | |
1452 | ||
1453 | /* When using "GET_DISK_INFO" it isn't certain how high | |
1454 | * we need to check. So we impose an absolute limit of | |
1455 | * MAX_DISKS. This needs to be much more than the largest | |
1456 | * number of devices any metadata can support. Currently | |
1457 | * v1.x can support 1920 | |
1458 | */ | |
1459 | #define MAX_DISKS 4096 | |
1460 | ||
1461 | /* Sometimes the 'size' value passed needs to mean "Maximum". | |
1462 | * In those cases with use MAX_SIZE | |
1463 | */ | |
1464 | #define MAX_SIZE 1 | |
1465 | ||
1466 | /* We want to use unsigned numbers for sector counts, but need | |
1467 | * a value for 'invalid'. Use '1'. | |
1468 | */ | |
1469 | #define INVALID_SECTORS 1 | |
1470 | ||
1471 | extern int __offroot; |