]> git.ipfire.org Git - thirdparty/mdadm.git/blame_incremental - mdadm.h
Add space_before/space_after fields to mdinfo
[thirdparty/mdadm.git] / mdadm.h
... / ...
CommitLineData
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__)
29extern __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 */
184struct 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 /* During reshape we can sometimes change the data_offset to avoid
210 * over-writing still-valid data. We need to know if there is space.
211 * So getinfo_super will fill in space_before and space_after in sectors.
212 * data_offset can be increased or decreased by this amount.
213 */
214 unsigned long long space_before, space_after;
215 union {
216 unsigned long long resync_start; /* per-array resync position */
217 unsigned long long recovery_start; /* per-device rebuild position */
218 #define MaxSector (~0ULL) /* resync/recovery complete position */
219 };
220 long bitmap_offset; /* 0 == none, 1 == a file */
221 unsigned long safe_mode_delay; /* ms delay to mark clean */
222 int new_level, delta_disks, new_layout, new_chunk;
223 int errors;
224 unsigned long cache_size; /* size of raid456 stripe cache*/
225 int mismatch_cnt;
226 char text_version[50];
227
228 int container_member; /* for assembling external-metatdata arrays
229 * This is to be used internally by metadata
230 * handler only */
231 int container_enough; /* flag external handlers can set to
232 * indicate that subarrays have not enough (-1),
233 * enough to start (0), or all expected disks (1) */
234 char sys_name[20];
235 struct mdinfo *devs;
236 struct mdinfo *next;
237
238 /* Device info for mdmon: */
239 int recovery_fd;
240 int state_fd;
241 #define DS_FAULTY 1
242 #define DS_INSYNC 2
243 #define DS_WRITE_MOSTLY 4
244 #define DS_SPARE 8
245 #define DS_BLOCKED 16
246 #define DS_REMOVE 1024
247 #define DS_UNBLOCK 2048
248 int prev_state, curr_state, next_state;
249
250};
251
252struct createinfo {
253 int uid;
254 int gid;
255 int autof;
256 int mode;
257 int symlinks;
258 struct supertype *supertype;
259};
260
261#define Name "mdadm"
262
263enum mode {
264 ASSEMBLE=1,
265 BUILD,
266 CREATE,
267 MANAGE,
268 MISC,
269 MONITOR,
270 GROW,
271 INCREMENTAL,
272 AUTODETECT,
273 mode_count
274};
275
276extern char short_options[];
277extern char short_bitmap_options[];
278extern char short_bitmap_auto_options[];
279extern struct option long_options[];
280extern char Version[], Usage[], Help[], OptionHelp[],
281 *mode_help[],
282 Help_create[], Help_build[], Help_assemble[], Help_grow[],
283 Help_incr[],
284 Help_manage[], Help_misc[], Help_monitor[], Help_config[];
285
286/* for option that don't have short equivilents, we assign arbitrary
287 * numbers later than any 'short' character option.
288 */
289enum special_options {
290 AssumeClean = 300,
291 BitmapChunk,
292 WriteBehind,
293 ReAdd,
294 NoDegraded,
295 Sparc22,
296 BackupFile,
297 HomeHost,
298 AutoHomeHost,
299 Symlinks,
300 AutoDetect,
301 Waitclean,
302 DetailPlatform,
303 KillSubarray,
304 UpdateSubarray,
305 IncrementalPath,
306 NoSharing,
307 HelpOptions,
308 Brief,
309 ManageOpt,
310 Add,
311 Remove,
312 Fail,
313 MiscOpt,
314 WaitOpt,
315 ConfigFile,
316 ChunkSize,
317 WriteMostly,
318 Layout,
319 Auto,
320 Force,
321 SuperMinor,
322 EMail,
323 ProgramOpt,
324 Increment,
325 Fork,
326 Bitmap,
327 RebuildMapOpt,
328 InvalidBackup,
329 UdevRules,
330 FreezeReshape,
331 Continue,
332 OffRootOpt,
333 Prefer,
334 KillOpt,
335 DataOffset,
336};
337
338enum prefix_standard {
339 JEDEC,
340 IEC
341};
342
343/* structures read from config file */
344/* List of mddevice names and identifiers
345 * Identifiers can be:
346 * uuid=128-hex-uuid
347 * super-minor=decimal-minor-number-from-superblock
348 * devices=comma,separated,list,of,device,names,with,wildcards
349 *
350 * If multiple fields are present, the intersection of all matching
351 * devices is considered
352 */
353#define UnSet (0xfffe)
354struct mddev_ident {
355 char *devname;
356
357 int uuid_set;
358 int uuid[4];
359 char name[33];
360
361 int super_minor;
362
363 char *devices; /* comma separated list of device
364 * names with wild cards
365 */
366 int level;
367 int raid_disks;
368 int spare_disks;
369 struct supertype *st;
370 int autof; /* 1 for normal, 2 for partitioned */
371 char *spare_group;
372 char *bitmap_file;
373 int bitmap_fd;
374
375 char *container; /* /dev/whatever name of container, or
376 * uuid of container. You would expect
377 * this to be the 'devname' or UUID
378 * of some other entry.
379 */
380 char *member; /* subarray within a container */
381
382 struct mddev_ident *next;
383 union {
384 /* fields needed by different users of this structure */
385 int assembled; /* set when assembly succeeds */
386 };
387};
388
389struct context {
390 int readonly;
391 int runstop;
392 int verbose;
393 int brief;
394 int force;
395 char *homehost;
396 int require_homehost;
397 char *prefer;
398 int export;
399 int test;
400 char *subarray;
401 char *update;
402 int scan;
403 int SparcAdjust;
404 int autof;
405 int delay;
406 int freeze_reshape;
407 char *backup_file;
408 int invalid_backup;
409};
410
411struct shape {
412 int raiddisks;
413 int sparedisks;
414 int level;
415 int layout;
416 char *layout_str;
417 int chunk;
418 int bitmap_chunk;
419 char *bitmap_file;
420 int assume_clean;
421 int write_behind;
422 unsigned long long size;
423};
424
425/* List of device names - wildcards expanded */
426struct mddev_dev {
427 char *devname;
428 int disposition; /* 'a' for add, 'r' for remove, 'f' for fail,
429 * 'A' for re_add.
430 * Not set for names read from .config
431 */
432 char writemostly; /* 1 for 'set writemostly', 2 for 'clear writemostly' */
433 char used; /* set when used */
434 struct mddev_dev *next;
435};
436
437typedef struct mapping {
438 char *name;
439 int num;
440} mapping_t;
441
442
443struct mdstat_ent {
444 char *dev;
445 int devnum;
446 int active;
447 char *level;
448 char *pattern; /* U or up, _ for down */
449 int percent; /* -1 if no resync */
450 int resync; /* 3 if check, 2 if reshape, 1 if resync, 0 if recovery */
451 int devcnt;
452 int raid_disks;
453 char * metadata_version;
454 struct dev_member {
455 char *name;
456 struct dev_member *next;
457 } *members;
458 struct mdstat_ent *next;
459};
460
461extern struct mdstat_ent *mdstat_read(int hold, int start);
462extern void free_mdstat(struct mdstat_ent *ms);
463extern void mdstat_wait(int seconds);
464extern void mdstat_wait_fd(int fd, const sigset_t *sigmask);
465extern int mddev_busy(int devnum);
466extern struct mdstat_ent *mdstat_by_component(char *name);
467extern struct mdstat_ent *mdstat_by_subdev(char *subdev, int container);
468
469struct map_ent {
470 struct map_ent *next;
471 int devnum;
472 char metadata[20];
473 int uuid[4];
474 int bad;
475 char *path;
476};
477extern int map_update(struct map_ent **mpp, int devnum, char *metadata,
478 int uuid[4], char *path);
479extern void map_remove(struct map_ent **map, int devnum);
480extern struct map_ent *map_by_uuid(struct map_ent **map, int uuid[4]);
481extern struct map_ent *map_by_devnum(struct map_ent **map, int devnum);
482extern struct map_ent *map_by_name(struct map_ent **map, char *name);
483extern void map_read(struct map_ent **melp);
484extern int map_write(struct map_ent *mel);
485extern void map_delete(struct map_ent **mapp, int devnum);
486extern void map_free(struct map_ent *map);
487extern void map_add(struct map_ent **melp,
488 int devnum, char *metadata, int uuid[4], char *path);
489extern int map_lock(struct map_ent **melp);
490extern void map_unlock(struct map_ent **melp);
491extern void map_fork(void);
492
493/* various details can be requested */
494enum sysfs_read_flags {
495 GET_LEVEL = (1 << 0),
496 GET_LAYOUT = (1 << 1),
497 GET_COMPONENT = (1 << 2),
498 GET_CHUNK = (1 << 3),
499 GET_CACHE = (1 << 4),
500 GET_MISMATCH = (1 << 5),
501 GET_VERSION = (1 << 6),
502 GET_DISKS = (1 << 7),
503 GET_DEGRADED = (1 << 8),
504 GET_SAFEMODE = (1 << 9),
505 GET_BITMAP_LOCATION = (1 << 10),
506
507 GET_DEVS = (1 << 20), /* gets role, major, minor */
508 GET_OFFSET = (1 << 21),
509 GET_SIZE = (1 << 22),
510 GET_STATE = (1 << 23),
511 GET_ERROR = (1 << 24),
512};
513
514/* If fd >= 0, get the array it is open on,
515 * else use devnum. >=0 -> major9. <0.....
516 */
517extern int sysfs_open(int devnum, char *devname, char *attr);
518extern void sysfs_init(struct mdinfo *mdi, int fd, int devnum);
519extern void sysfs_free(struct mdinfo *sra);
520extern struct mdinfo *sysfs_read(int fd, int devnum, unsigned long options);
521extern int sysfs_attr_match(const char *attr, const char *str);
522extern int sysfs_match_word(const char *word, char **list);
523extern int sysfs_set_str(struct mdinfo *sra, struct mdinfo *dev,
524 char *name, char *val);
525extern int sysfs_set_num(struct mdinfo *sra, struct mdinfo *dev,
526 char *name, unsigned long long val);
527extern int sysfs_set_num_signed(struct mdinfo *sra, struct mdinfo *dev,
528 char *name, long long val);
529extern int sysfs_uevent(struct mdinfo *sra, char *event);
530extern int sysfs_get_fd(struct mdinfo *sra, struct mdinfo *dev,
531 char *name);
532extern int sysfs_fd_get_ll(int fd, unsigned long long *val);
533extern int sysfs_get_ll(struct mdinfo *sra, struct mdinfo *dev,
534 char *name, unsigned long long *val);
535extern int sysfs_fd_get_str(int fd, char *val, int size);
536extern int sysfs_attribute_available(struct mdinfo *sra, struct mdinfo *dev,
537 char *name);
538extern int sysfs_get_str(struct mdinfo *sra, struct mdinfo *dev,
539 char *name, char *val, int size);
540extern int sysfs_set_safemode(struct mdinfo *sra, unsigned long ms);
541extern int sysfs_set_array(struct mdinfo *info, int vers);
542extern int sysfs_add_disk(struct mdinfo *sra, struct mdinfo *sd, int resume);
543extern int sysfs_disk_to_scsi_id(int fd, __u32 *id);
544extern int sysfs_unique_holder(int devnum, long rdev);
545extern int sysfs_freeze_array(struct mdinfo *sra);
546extern int load_sys(char *path, char *buf);
547extern int reshape_prepare_fdlist(char *devname,
548 struct mdinfo *sra,
549 int raid_disks,
550 int nrdisks,
551 unsigned long blocks,
552 char *backup_file,
553 int *fdlist,
554 unsigned long long *offsets);
555extern void reshape_free_fdlist(int *fdlist,
556 unsigned long long *offsets,
557 int size);
558extern int reshape_open_backup_file(char *backup,
559 int fd,
560 char *devname,
561 long blocks,
562 int *fdlist,
563 unsigned long long *offsets,
564 int restart);
565extern unsigned long compute_backup_blocks(int nchunk, int ochunk,
566 unsigned int ndata, unsigned int odata);
567
568extern int save_stripes(int *source, unsigned long long *offsets,
569 int raid_disks, int chunk_size, int level, int layout,
570 int nwrites, int *dest,
571 unsigned long long start, unsigned long long length,
572 char *buf);
573extern int restore_stripes(int *dest, unsigned long long *offsets,
574 int raid_disks, int chunk_size, int level, int layout,
575 int source, unsigned long long read_offset,
576 unsigned long long start, unsigned long long length,
577 char *src_buf);
578
579#ifndef Sendmail
580#define Sendmail "/usr/lib/sendmail -t"
581#endif
582
583#define SYSLOG_FACILITY LOG_DAEMON
584
585extern char *map_num(mapping_t *map, int num);
586extern int map_name(mapping_t *map, char *name);
587extern mapping_t r5layout[], r6layout[], pers[], modes[], faultylayout[];
588
589extern char *map_dev_preferred(int major, int minor, int create,
590 char *prefer);
591static inline char *map_dev(int major, int minor, int create)
592{
593 return map_dev_preferred(major, minor, create, NULL);
594}
595
596struct active_array;
597struct metadata_update;
598
599
600/* 'struct reshape' records the intermediate states of
601 * a general reshape.
602 * The starting geometry is converted to the 'before' geometry
603 * by at most an atomic level change. They could be the same.
604 * Similarly the 'after' geometry is converted to the final
605 * geometry by at most a level change.
606 * Note that 'before' and 'after' must have the same level.
607 * 'blocks' is the minimum number of sectors for a reshape unit.
608 * This will be a multiple of the stripe size in each of the
609 * 'before' and 'after' geometries.
610 * If 'blocks' is 0, no restriping is necessary.
611 */
612struct reshape {
613 int level;
614 int parity; /* number of parity blocks/devices */
615 struct {
616 int layout;
617 int data_disks;
618 } before, after;
619 unsigned long long backup_blocks;
620 unsigned long long stripes; /* number of old stripes that comprise 'blocks'*/
621 unsigned long long new_size; /* New size of array in sectors */
622};
623
624/* A superswitch provides entry point the a metadata handler.
625 *
626 * The super_switch primarily operates on some "metadata" that
627 * is accessed via the 'supertype'.
628 * This metadata has one of three possible sources.
629 * 1/ It is read from a single device. In this case it may not completely
630 * describe the array or arrays as some information might be on other
631 * devices.
632 * 2/ It is read from all devices in a container. In this case all
633 * information is present.
634 * 3/ It is created by ->init_super / ->add_to_super. In this case it will
635 * be complete once enough ->add_to_super calls have completed.
636 *
637 * When creating an array inside a container, the metadata will be
638 * formed by a combination of 2 and 3. The metadata or the array is read,
639 * then new information is added.
640 *
641 * The metadata must sometimes have a concept of a 'current' array
642 * and a 'current' device.
643 * The 'current' array is set by init_super to be the newly created array,
644 * or is set by super_by_fd when it finds it is looking at an array inside
645 * a container.
646 *
647 * The 'current' device is either the device that the metadata was read from
648 * in case 1, or the last device added by add_to_super in case 3.
649 * Case 2 does not identify a 'current' device.
650 */
651extern struct superswitch {
652
653 /* Used to report details of metadata read from a component
654 * device. ->load_super has been called.
655 */
656 void (*examine_super)(struct supertype *st, char *homehost);
657 void (*brief_examine_super)(struct supertype *st, int verbose);
658 void (*brief_examine_subarrays)(struct supertype *st, int verbose);
659 void (*export_examine_super)(struct supertype *st);
660
661 /* Used to report details of an active array.
662 * ->load_super was possibly given a 'component' string.
663 */
664 void (*detail_super)(struct supertype *st, char *homehost);
665 void (*brief_detail_super)(struct supertype *st);
666 void (*export_detail_super)(struct supertype *st);
667
668 /* Optional: platform hardware / firmware details */
669 int (*detail_platform)(int verbose, int enumerate_only, char *controller_path);
670 int (*export_detail_platform)(int verbose, char *controller_path);
671
672 /* Used:
673 * to get uuid to storing in bitmap metadata
674 * and 'reshape' backup-data metadata
675 * To see if a device is being re-added to an array it was part of.
676 */
677 void (*uuid_from_super)(struct supertype *st, int uuid[4]);
678
679 /* Extract generic details from metadata. This could be details about
680 * the container, or about an individual array within the container.
681 * The determination is made either by:
682 * load_super being given a 'component' string.
683 * validate_geometry determining what to create.
684 * The info includes both array information and device information.
685 * The particular device should be:
686 * The last device added by add_to_super
687 * The device the metadata was loaded from by load_super
688 * If 'map' is present, then it is an array raid_disks long
689 * (raid_disk must already be set and correct) and it is filled
690 * with 1 for slots that are thought to be active and 0 for slots which
691 * appear to be failed/missing.
692 * *info is zeroed out before data is added.
693 */
694 void (*getinfo_super)(struct supertype *st, struct mdinfo *info, char *map);
695 struct mdinfo *(*getinfo_super_disks)(struct supertype *st);
696 /* Check if the given metadata is flagged as belonging to "this"
697 * host. 0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
698 */
699 int (*match_home)(struct supertype *st, char *homehost);
700
701 /* Make one of several generic modifications to metadata
702 * prior to assembly (or other times).
703 * sparc2.2 - first bug in early 0.90 metadata
704 * super-minor - change name of 0.90 metadata
705 * summaries - 'correct' any redundant data
706 * resync - mark array as dirty to trigger a resync.
707 * uuid - set new uuid - only 0.90 or 1.x
708 * name - change the name of the array (where supported)
709 * homehost - change which host this array is tied to.
710 * devicesize - If metadata is at start of device, change recorded
711 * device size to match actual device size
712 * byteorder - swap bytes for 0.90 metadata
713 *
714 * force-one - mark that device as uptodate, not old or failed.
715 * force-array - mark array as clean if it would not otherwise
716 * assemble
717 * assemble - not sure how this is different from force-one...
718 * linear-grow-new - add a new device to a linear array, but don't
719 * change the size: so superblock still matches
720 * linear-grow-update - now change the size of the array.
721 * writemostly - set the WriteMostly1 bit in the superblock devflags
722 * readwrite - clear the WriteMostly1 bit in the superblock devflags
723 */
724 int (*update_super)(struct supertype *st, struct mdinfo *info,
725 char *update,
726 char *devname, int verbose,
727 int uuid_set, char *homehost);
728
729 /* Create new metadata for new array as described. This could
730 * be a new container, or an array in a pre-existing container.
731 * Also used to zero metadata prior to writing it to invalidate old
732 * metadata.
733 */
734 int (*init_super)(struct supertype *st, mdu_array_info_t *info,
735 unsigned long long size, char *name,
736 char *homehost, int *uuid,
737 unsigned long long data_offset);
738
739 /* update the metadata to include new device, either at create or
740 * when hot-adding a spare.
741 */
742 int (*add_to_super)(struct supertype *st, mdu_disk_info_t *dinfo,
743 int fd, char *devname);
744 /* update the metadata to delete a device,
745 * when hot-removing.
746 */
747 int (*remove_from_super)(struct supertype *st, mdu_disk_info_t *dinfo);
748
749 /* Write metadata to one device when fixing problems or adding
750 * a new device.
751 */
752 int (*store_super)(struct supertype *st, int fd);
753
754 /* Write all metadata for this array.
755 */
756 int (*write_init_super)(struct supertype *st);
757 int (*compare_super)(struct supertype *st, struct supertype *tst);
758 int (*load_super)(struct supertype *st, int fd, char *devname);
759 int (*load_container)(struct supertype *st, int fd, char *devname);
760 struct supertype * (*match_metadata_desc)(char *arg);
761 __u64 (*avail_size)(struct supertype *st, __u64 size,
762 unsigned long long data_offset);
763 unsigned long long (*min_acceptable_spare_size)(struct supertype *st);
764 int (*add_internal_bitmap)(struct supertype *st, int *chunkp,
765 int delay, int write_behind,
766 unsigned long long size, int may_change, int major);
767 void (*locate_bitmap)(struct supertype *st, int fd);
768 int (*write_bitmap)(struct supertype *st, int fd);
769 void (*free_super)(struct supertype *st);
770
771 /* validate_geometry is called with an st returned by
772 * match_metadata_desc.
773 * It should check that the geometry described in compatible with
774 * the metadata type. It will be called repeatedly as devices
775 * added to validate changing size and new devices. If there are
776 * inter-device dependencies, it should record sufficient details
777 * so these can be validated.
778 * Both 'size' and '*freesize' are in sectors. chunk is KiB.
779 * Return value is:
780 * 1: everything is OK
781 * 0: not OK for some reason - if 'verbose', then error was reported.
782 * -1: st->sb was NULL, 'subdev' is a member of a container of this
783 * types, but array is not acceptable for some reason
784 * message was reported even if verbose is 0.
785 */
786 int (*validate_geometry)(struct supertype *st, int level, int layout,
787 int raiddisks,
788 int *chunk, unsigned long long size,
789 unsigned long long data_offset,
790 char *subdev, unsigned long long *freesize,
791 int verbose);
792
793 struct mdinfo *(*container_content)(struct supertype *st, char *subarray);
794 /* query the supertype for default geometry */
795 void (*default_geometry)(struct supertype *st, int *level, int *layout, int *chunk); /* optional */
796 /* Permit subarray's to be deleted from inactive containers */
797 int (*kill_subarray)(struct supertype *st); /* optional */
798 /* Permit subarray's to be modified */
799 int (*update_subarray)(struct supertype *st, char *subarray,
800 char *update, struct mddev_ident *ident); /* optional */
801 /* Check if reshape is supported for this external format.
802 * st is obtained from super_by_fd() where st->subarray[0] is
803 * initialized to indicate if reshape is being performed at the
804 * container or subarray level
805 */
806#define APPLY_METADATA_CHANGES 1
807#define ROLLBACK_METADATA_CHANGES 0
808
809 int (*reshape_super)(struct supertype *st,
810 unsigned long long size, int level,
811 int layout, int chunksize, int raid_disks,
812 int delta_disks, char *backup, char *dev,
813 int direction,
814 int verbose); /* optional */
815 int (*manage_reshape)( /* optional */
816 int afd, struct mdinfo *sra, struct reshape *reshape,
817 struct supertype *st, unsigned long blocks,
818 int *fds, unsigned long long *offsets,
819 int dests, int *destfd, unsigned long long *destoffsets);
820
821/* for mdmon */
822 int (*open_new)(struct supertype *c, struct active_array *a,
823 char *inst);
824
825 /* Tell the metadata handler the current state of the array.
826 * This covers whether it is known to be consistent (no pending writes)
827 * and how far along a resync is known to have progressed
828 * (in a->resync_start).
829 * resync status is really irrelevant if the array is not consistent,
830 * but some metadata (DDF!) have a place to record the distinction.
831 * If 'consistent' is '2', then the array can mark it dirty if a
832 * resync/recovery/whatever is required, or leave it clean if not.
833 * Return value is 0 dirty (not consistent) and 1 if clean.
834 * it is only really important if consistent is passed in as '2'.
835 */
836 int (*set_array_state)(struct active_array *a, int consistent);
837
838 /* When the state of a device might have changed, we call set_disk to
839 * tell the metadata what the current state is.
840 * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
841 * transitions.
842 * set_disk might be called when the state of the particular disk has
843 * not in fact changed.
844 */
845 void (*set_disk)(struct active_array *a, int n, int state);
846 void (*sync_metadata)(struct supertype *st);
847 void (*process_update)(struct supertype *st,
848 struct metadata_update *update);
849 void (*prepare_update)(struct supertype *st,
850 struct metadata_update *update);
851
852 /* activate_spare will check if the array is degraded and, if it
853 * is, try to find some spare space in the container.
854 * On success, it add appropriate updates (For process_update) to
855 * to the 'updates' list and returns a list of 'mdinfo' identifying
856 * the device, or devices as there might be multiple missing
857 * devices and multiple spares available.
858 */
859 struct mdinfo *(*activate_spare)(struct active_array *a,
860 struct metadata_update **updates);
861 /*
862 * Return statically allocated string that represents metadata specific
863 * controller domain of the disk. The domain is used in disk domain
864 * matching functions. Disks belong to the same domain if the they have
865 * the same domain from mdadm.conf and belong the same metadata domain.
866 * Returning NULL or not providing this handler means that metadata
867 * does not distinguish the differences between disks that belong to
868 * different controllers. They are in the domain specified by
869 * configuration file (mdadm.conf).
870 * In case when the metadata has the notion of domains based on disk
871 * it shall return NULL for disks that do not belong to the controller
872 * the supported domains. Such disks will form another domain and won't
873 * be mixed with supported ones.
874 */
875 const char *(*get_disk_controller_domain)(const char *path);
876
877 /* for external backup area */
878 int (*recover_backup)(struct supertype *st, struct mdinfo *info);
879
880 int swapuuid; /* true if uuid is bigending rather than hostendian */
881 int external;
882 const char *name; /* canonical metadata name */
883} *superlist[];
884
885extern struct superswitch super0, super1;
886extern struct superswitch super_imsm, super_ddf;
887extern struct superswitch mbr, gpt;
888
889struct metadata_update {
890 int len;
891 char *buf;
892 void *space; /* allocated space that monitor will use */
893 void **space_list; /* list of allocated spaces that monitor can
894 * use or that it returned.
895 */
896 struct metadata_update *next;
897};
898
899/* A supertype holds a particular collection of metadata.
900 * It identifies the metadata type by the superswitch, and the particular
901 * sub-version of that metadata type.
902 * metadata read in or created is stored in 'sb' and 'info'.
903 * There are also fields used by mdmon to track containers.
904 *
905 * A supertype may refer to:
906 * Just an array, possibly in a container
907 * A container, not identifying any particular array
908 * Info read from just one device, not yet fully describing the array/container.
909 *
910 *
911 * A supertype is created by:
912 * super_by_fd
913 * guess_super
914 * dup_super
915 */
916struct supertype {
917 struct superswitch *ss;
918 int minor_version;
919 int max_devs;
920 int container_dev; /* devnum of container */
921 void *sb;
922 void *info;
923 int ignore_hw_compat; /* used to inform metadata handlers that it should ignore
924 HW/firmware related incompatability to load metadata.
925 Used when examining metadata to display content of disk
926 when user has no hw/firmare compatible system.
927 */
928 struct metadata_update *updates;
929 struct metadata_update **update_tail;
930
931 /* extra stuff used by mdmon */
932 struct active_array *arrays;
933 int sock; /* listen to external programs */
934 int devnum;
935 char *devname; /* e.g. md0. This appears in metadata_verison:
936 * external:/md0/12
937 */
938 int devcnt;
939 int retry_soon;
940
941 struct mdinfo *devs;
942
943};
944
945extern struct supertype *super_by_fd(int fd, char **subarray);
946enum guess_types { guess_any, guess_array, guess_partitions };
947extern struct supertype *guess_super_type(int fd, enum guess_types guess_type);
948static inline struct supertype *guess_super(int fd) {
949 return guess_super_type(fd, guess_any);
950}
951extern struct supertype *dup_super(struct supertype *st);
952extern int get_dev_size(int fd, char *dname, unsigned long long *sizep);
953extern int must_be_container(int fd);
954extern int dev_size_from_id(dev_t id, unsigned long long *size);
955extern void get_one_disk(int mdfd, mdu_array_info_t *ainf,
956 mdu_disk_info_t *disk);
957void wait_for(char *dev, int fd);
958
959/*
960 * Data structures for policy management.
961 * Each device can have a policy structure that lists
962 * various name/value pairs each possibly with a metadata associated.
963 * The policy list is sorted by name/value/metadata
964 */
965struct dev_policy {
966 struct dev_policy *next;
967 char *name; /* None of these strings are allocated. They are
968 * all just references to strings which are known
969 * to exist elsewhere.
970 * name and metadata can be compared by address equality.
971 */
972 const char *metadata;
973 const char *value;
974};
975
976extern char pol_act[], pol_domain[], pol_metadata[], pol_auto[];
977
978/* iterate over the sublist starting at list, having the same
979 * 'name' as 'list', and matching the given metadata (Where
980 * NULL matches anything
981 */
982#define pol_for_each(item, list, _metadata) \
983 for (item = list; \
984 item && item->name == list->name; \
985 item = item->next) \
986 if (!(!_metadata || !item->metadata || _metadata == item->metadata)) \
987 ; else
988
989/*
990 * policy records read from mdadm are largely just name-value pairs.
991 * The names are constants, not strdupped
992 */
993struct pol_rule {
994 struct pol_rule *next;
995 char *type; /* rule_policy or rule_part */
996 struct rule {
997 struct rule *next;
998 char *name;
999 char *value;
1000 char *dups; /* duplicates of 'value' with a partNN appended */
1001 } *rule;
1002};
1003
1004extern char rule_policy[], rule_part[];
1005extern char rule_path[], rule_type[];
1006extern char type_part[], type_disk[];
1007
1008extern void policyline(char *line, char *type);
1009extern void policy_add(char *type, ...);
1010extern void policy_free(void);
1011
1012extern struct dev_policy *path_policy(char *path, char *type);
1013extern struct dev_policy *disk_policy(struct mdinfo *disk);
1014extern struct dev_policy *devnum_policy(int dev);
1015extern void dev_policy_free(struct dev_policy *p);
1016
1017//extern void pol_new(struct dev_policy **pol, char *name, char *val, char *metadata);
1018extern void pol_add(struct dev_policy **pol, char *name, char *val, char *metadata);
1019extern struct dev_policy *pol_find(struct dev_policy *pol, char *name);
1020
1021enum policy_action {
1022 act_default,
1023 act_include,
1024 act_re_add,
1025 act_spare, /* This only applies to bare devices */
1026 act_spare_same_slot, /* this allows non-bare devices,
1027 * but only if recent removal */
1028 act_force_spare, /* this allow non-bare devices in any case */
1029 act_err
1030};
1031
1032extern int policy_action_allows(struct dev_policy *plist, const char *metadata,
1033 enum policy_action want);
1034extern int disk_action_allows(struct mdinfo *disk, const char *metadata,
1035 enum policy_action want);
1036
1037struct domainlist {
1038 struct domainlist *next;
1039 const char *dom;
1040};
1041
1042extern int domain_test(struct domainlist *dom, struct dev_policy *pol,
1043 const char *metadata);
1044extern struct domainlist *domain_from_array(struct mdinfo *mdi,
1045 const char *metadata);
1046extern void domainlist_add_dev(struct domainlist **dom, int devnum,
1047 const char *metadata);
1048extern void domain_free(struct domainlist *dl);
1049extern void domain_merge(struct domainlist **domp, struct dev_policy *pol,
1050 const char *metadata);
1051void domain_add(struct domainlist **domp, char *domain);
1052
1053extern void policy_save_path(char *id_path, struct map_ent *array);
1054extern int policy_check_path(struct mdinfo *disk, struct map_ent *array);
1055
1056#if __GNUC__ < 3
1057struct stat64;
1058#endif
1059
1060#define HAVE_NFTW we assume
1061#define HAVE_FTW
1062
1063#ifdef __UCLIBC__
1064# include <features.h>
1065# ifndef __UCLIBC_HAS_LFS__
1066# define lseek64 lseek
1067# endif
1068# ifndef __UCLIBC_HAS_FTW__
1069# undef HAVE_FTW
1070# undef HAVE_NFTW
1071# endif
1072#endif
1073
1074#ifdef __dietlibc__
1075# undef HAVE_NFTW
1076#endif
1077
1078#if defined(__KLIBC__)
1079# undef HAVE_NFTW
1080# undef HAVE_FTW
1081#endif
1082
1083#ifndef HAVE_NFTW
1084# define FTW_PHYS 1
1085# ifndef HAVE_FTW
1086 struct FTW {};
1087# endif
1088#endif
1089
1090#ifdef HAVE_FTW
1091# include <ftw.h>
1092#endif
1093
1094extern int add_dev(const char *name, const struct stat *stb, int flag, struct FTW *s);
1095
1096
1097extern int Manage_ro(char *devname, int fd, int readonly);
1098extern int Manage_runstop(char *devname, int fd, int runstop, int quiet,
1099 int will_retry);
1100extern int Manage_subdevs(char *devname, int fd,
1101 struct mddev_dev *devlist, int verbose, int test,
1102 char *update, int force);
1103extern int autodetect(void);
1104extern int Grow_Add_device(char *devname, int fd, char *newdev);
1105extern int Grow_addbitmap(char *devname, int fd,
1106 struct context *c, struct shape *s);
1107extern int Grow_reshape(char *devname, int fd,
1108 struct mddev_dev *devlist,
1109 unsigned long long data_offset,
1110 struct context *c, struct shape *s);
1111extern int Grow_restart(struct supertype *st, struct mdinfo *info,
1112 int *fdlist, int cnt, char *backup_file, int verbose);
1113extern int Grow_continue(int mdfd, struct supertype *st,
1114 struct mdinfo *info, char *backup_file,
1115 int freeze_reshape);
1116
1117extern int restore_backup(struct supertype *st,
1118 struct mdinfo *content,
1119 int working_disks,
1120 int spares,
1121 char *backup_file,
1122 int verbose);
1123extern int Grow_continue_command(char *devname, int fd,
1124 char *backup_file, int verbose);
1125
1126extern int Assemble(struct supertype *st, char *mddev,
1127 struct mddev_ident *ident,
1128 struct mddev_dev *devlist,
1129 struct context *c);
1130
1131extern int Build(char *mddev, struct mddev_dev *devlist,
1132 struct shape *s, struct context *c);
1133
1134extern int Create(struct supertype *st, char *mddev,
1135 char *name, int *uuid,
1136 int subdevs, struct mddev_dev *devlist,
1137 struct shape *s,
1138 struct context *c,
1139 unsigned long long data_offset);
1140
1141extern int Detail(char *dev, struct context *c);
1142extern int Detail_Platform(struct superswitch *ss, int scan, int verbose, int export, char *controller_path);
1143extern int Query(char *dev);
1144extern int Examine(struct mddev_dev *devlist, struct context *c,
1145 struct supertype *forcest);
1146extern int Monitor(struct mddev_dev *devlist,
1147 char *mailaddr, char *alert_cmd,
1148 struct context *c,
1149 int daemonise, int oneshot,
1150 int dosyslog, char *pidfile, int increments,
1151 int share);
1152
1153extern int Kill(char *dev, struct supertype *st, int force, int verbose, int noexcl);
1154extern int Kill_subarray(char *dev, char *subarray, int verbose);
1155extern int Update_subarray(char *dev, char *subarray, char *update, struct mddev_ident *ident, int quiet);
1156extern int Wait(char *dev);
1157extern int WaitClean(char *dev, int sock, int verbose);
1158
1159extern int Incremental(char *devname, struct context *c,
1160 struct supertype *st);
1161extern void RebuildMap(void);
1162extern int IncrementalScan(int verbose);
1163extern int IncrementalRemove(char *devname, char *path, int verbose);
1164extern int CreateBitmap(char *filename, int force, char uuid[16],
1165 unsigned long chunksize, unsigned long daemon_sleep,
1166 unsigned long write_behind,
1167 unsigned long long array_size,
1168 int major);
1169extern int ExamineBitmap(char *filename, int brief, struct supertype *st);
1170extern int Write_rules(char *rule_name);
1171extern int bitmap_update_uuid(int fd, int *uuid, int swap);
1172extern unsigned long bitmap_sectors(struct bitmap_super_s *bsb);
1173
1174extern int md_get_version(int fd);
1175extern int get_linux_version(void);
1176extern int mdadm_version(char *version);
1177extern unsigned long long parse_size(char *size);
1178extern int parse_uuid(char *str, int uuid[4]);
1179extern int parse_layout_10(char *layout);
1180extern int parse_layout_faulty(char *layout);
1181extern long parse_num(char *num);
1182extern int check_ext2(int fd, char *name);
1183extern int check_reiser(int fd, char *name);
1184extern int check_raid(int fd, char *name);
1185extern int check_partitions(int fd, char *dname,
1186 unsigned long long freesize,
1187 unsigned long long size);
1188
1189extern int get_mdp_major(void);
1190extern int dev_open(char *dev, int flags);
1191extern int open_dev(int devnum);
1192extern int open_dev_flags(int devnum, int flags);
1193extern int open_dev_excl(int devnum);
1194extern int is_standard(char *dev, int *nump);
1195extern int same_dev(char *one, char *two);
1196extern int compare_paths (char* path1,char* path2);
1197
1198extern int parse_auto(char *str, char *msg, int config);
1199extern struct mddev_ident *conf_get_ident(char *dev);
1200extern struct mddev_dev *conf_get_devs(void);
1201extern int conf_test_dev(char *devname);
1202extern int conf_test_metadata(const char *version, struct dev_policy *pol, int is_homehost);
1203extern struct createinfo *conf_get_create_info(void);
1204extern void set_conffile(char *file);
1205extern char *conf_get_mailaddr(void);
1206extern char *conf_get_mailfrom(void);
1207extern char *conf_get_program(void);
1208extern char *conf_get_homehost(int *require_homehostp);
1209extern char *conf_line(FILE *file);
1210extern char *conf_word(FILE *file, int allow_key);
1211extern void print_quoted(char *str);
1212extern void print_escape(char *str);
1213extern int conf_name_is_free(char *name);
1214extern int conf_verify_devnames(struct mddev_ident *array_list);
1215extern int devname_matches(char *name, char *match);
1216extern struct mddev_ident *conf_match(struct supertype *st,
1217 struct mdinfo *info,
1218 char *devname,
1219 int verbose, int *rvp);
1220extern int experimental(void);
1221
1222extern void free_line(char *line);
1223extern int match_oneof(char *devices, char *devname);
1224extern void uuid_from_super(int uuid[4], mdp_super_t *super);
1225extern const int uuid_zero[4];
1226extern int same_uuid(int a[4], int b[4], int swapuuid);
1227extern void copy_uuid(void *a, int b[4], int swapuuid);
1228extern char *__fname_from_uuid(int id[4], int swap, char *buf, char sep);
1229extern char *fname_from_uuid(struct supertype *st,
1230 struct mdinfo *info, char *buf, char sep);
1231extern unsigned long calc_csum(void *super, int bytes);
1232extern int enough(int level, int raid_disks, int layout, int clean,
1233 char *avail);
1234extern int enough_fd(int fd);
1235extern int ask(char *mesg);
1236extern unsigned long long get_component_size(int fd);
1237extern void remove_partitions(int fd);
1238extern int test_partition(int fd);
1239extern int test_partition_from_id(dev_t id);
1240extern int get_data_disks(int level, int layout, int raid_disks);
1241extern unsigned long long calc_array_size(int level, int raid_disks, int layout,
1242 int chunksize, unsigned long long devsize);
1243extern int flush_metadata_updates(struct supertype *st);
1244extern void append_metadata_update(struct supertype *st, void *buf, int len);
1245extern int assemble_container_content(struct supertype *st, int mdfd,
1246 struct mdinfo *content,
1247 struct context *c,
1248 char *chosen_name);
1249extern struct mdinfo *container_choose_spares(struct supertype *st,
1250 unsigned long long min_size,
1251 struct domainlist *domlist,
1252 char *spare_group,
1253 const char *metadata, int get_one);
1254extern int move_spare(char *from_devname, char *to_devname, dev_t devid);
1255extern int add_disk(int mdfd, struct supertype *st,
1256 struct mdinfo *sra, struct mdinfo *info);
1257extern int remove_disk(int mdfd, struct supertype *st,
1258 struct mdinfo *sra, struct mdinfo *info);
1259extern int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info);
1260unsigned long long min_recovery_start(struct mdinfo *array);
1261
1262extern char *human_size(long long bytes);
1263extern char *human_size_brief(long long bytes, int prefix);
1264extern void print_r10_layout(int layout);
1265
1266#define NoMdDev (1<<23)
1267extern int find_free_devnum(int use_partitions);
1268
1269extern void put_md_name(char *name);
1270extern char *get_md_name(int dev);
1271
1272extern char DefaultConfFile[];
1273
1274extern int create_mddev(char *dev, char *name, int autof, int trustworthy,
1275 char *chosen);
1276/* values for 'trustworthy' */
1277#define LOCAL 1
1278#define LOCAL_ANY 10
1279#define FOREIGN 2
1280#define METADATA 3
1281extern int open_mddev(char *dev, int report_errors);
1282extern int open_container(int fd);
1283extern int is_container_member(struct mdstat_ent *ent, char *devname);
1284extern int is_subarray_active(char *subarray, char *devname);
1285extern int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet);
1286extern struct superswitch *version_to_superswitch(char *vers);
1287
1288extern int mdmon_running(int devnum);
1289extern int mdmon_pid(int devnum);
1290extern int check_env(char *name);
1291extern __u32 random32(void);
1292extern int start_mdmon(int devnum);
1293
1294extern int child_monitor(int afd, struct mdinfo *sra, struct reshape *reshape,
1295 struct supertype *st, unsigned long stripes,
1296 int *fds, unsigned long long *offsets,
1297 int dests, int *destfd, unsigned long long *destoffsets);
1298void abort_reshape(struct mdinfo *sra);
1299
1300extern char *devnum2devname(int num);
1301extern void fmt_devname(char *name, int num);
1302extern int devname2devnum(char *name);
1303extern int stat2devnum(struct stat *st);
1304extern int fd2devnum(int fd);
1305
1306static inline int dev2major(int d)
1307{
1308 if (d >= 0)
1309 return MD_MAJOR;
1310 else
1311 return get_mdp_major();
1312}
1313
1314static inline int dev2minor(int d)
1315{
1316 if (d >= 0)
1317 return d;
1318 return (-1-d) << MdpMinorShift;
1319}
1320
1321#define _ROUND_UP(val, base) (((val) + (base) - 1) & ~(base - 1))
1322#define ROUND_UP(val, base) _ROUND_UP(val, (typeof(val))(base))
1323#define ROUND_UP_PTR(ptr, base) ((typeof(ptr)) \
1324 (ROUND_UP((unsigned long)(ptr), base)))
1325
1326static inline int is_subarray(char *vers)
1327{
1328 /* The version string for a 'subarray' (an array in a container)
1329 * is
1330 * /containername/componentname for normal read-write arrays
1331 * -containername/componentname for arrays which mdmon must not
1332 * reconfigure. They might be read-only
1333 * or might be undergoing reshape etc.
1334 * containername is e.g. md0, md_d1
1335 * componentname is dependant on the metadata. e.g. '1' 'S1' ...
1336 */
1337 return (*vers == '/' || *vers == '-');
1338}
1339
1340static inline char *to_subarray(struct mdstat_ent *ent, char *container)
1341{
1342 return &ent->metadata_version[10+strlen(container)+1];
1343}
1344
1345#ifdef DEBUG
1346#define dprintf(fmt, arg...) \
1347 fprintf(stderr, fmt, ##arg)
1348#else
1349#define dprintf(fmt, arg...) \
1350 ({ if (0) fprintf(stderr, fmt, ##arg); 0; })
1351#endif
1352#include <assert.h>
1353#include <stdarg.h>
1354static inline int xasprintf(char **strp, const char *fmt, ...) {
1355 va_list ap;
1356 int ret;
1357 va_start(ap, fmt);
1358 ret = vasprintf(strp, fmt, ap);
1359 va_end(ap);
1360 assert(ret >= 0);
1361 return ret;
1362}
1363
1364#define pr_err(fmt ...) fprintf(stderr, Name ": " fmt)
1365#define cont_err(fmt ...) fprintf(stderr, " " fmt)
1366
1367void *xmalloc(size_t len);
1368void *xrealloc(void *ptr, size_t len);
1369void *xcalloc(size_t num, size_t size);
1370char *xstrdup(const char *str);
1371
1372#define LEVEL_MULTIPATH (-4)
1373#define LEVEL_LINEAR (-1)
1374#define LEVEL_FAULTY (-5)
1375
1376/* kernel module doesn't know about these */
1377#define LEVEL_CONTAINER (-100)
1378#define LEVEL_UNSUPPORTED (-200)
1379
1380
1381/* faulty stuff */
1382
1383#define WriteTransient 0
1384#define ReadTransient 1
1385#define WritePersistent 2
1386#define ReadPersistent 3
1387#define WriteAll 4 /* doesn't go to device */
1388#define ReadFixable 5
1389#define Modes 6
1390
1391#define ClearErrors 31
1392#define ClearFaults 30
1393
1394#define AllPersist 100 /* internal use only */
1395#define NoPersist 101
1396
1397#define ModeMask 0x1f
1398#define ModeShift 5
1399
1400
1401#ifdef __TINYC__
1402#undef minor
1403#undef major
1404#undef makedev
1405#define minor(x) ((x)&0xff)
1406#define major(x) (((x)>>8)&0xff)
1407#define makedev(M,m) (((M)<<8) | (m))
1408#endif
1409
1410/* for raid4/5/6 */
1411#define ALGORITHM_LEFT_ASYMMETRIC 0
1412#define ALGORITHM_RIGHT_ASYMMETRIC 1
1413#define ALGORITHM_LEFT_SYMMETRIC 2
1414#define ALGORITHM_RIGHT_SYMMETRIC 3
1415
1416/* Define non-rotating (raid4) algorithms. These allow
1417 * conversion of raid4 to raid5.
1418 */
1419#define ALGORITHM_PARITY_0 4 /* P or P,Q are initial devices */
1420#define ALGORITHM_PARITY_N 5 /* P or P,Q are final devices. */
1421
1422/* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
1423 * Firstly, the exact positioning of the parity block is slightly
1424 * different between the 'LEFT_*' modes of md and the "_N_*" modes
1425 * of DDF.
1426 * Secondly, or order of datablocks over which the Q syndrome is computed
1427 * is different.
1428 * Consequently we have different layouts for DDF/raid6 than md/raid6.
1429 * These layouts are from the DDFv1.2 spec.
1430 * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
1431 * leaves RLQ=3 as 'Vendor Specific'
1432 */
1433
1434#define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
1435#define ALGORITHM_ROTATING_N_RESTART 9 /* DDF PRL=6 RLQ=2 */
1436#define ALGORITHM_ROTATING_N_CONTINUE 10 /*DDF PRL=6 RLQ=3 */
1437
1438
1439/* For every RAID5 algorithm we define a RAID6 algorithm
1440 * with exactly the same layout for data and parity, and
1441 * with the Q block always on the last device (N-1).
1442 * This allows trivial conversion from RAID5 to RAID6
1443 */
1444#define ALGORITHM_LEFT_ASYMMETRIC_6 16
1445#define ALGORITHM_RIGHT_ASYMMETRIC_6 17
1446#define ALGORITHM_LEFT_SYMMETRIC_6 18
1447#define ALGORITHM_RIGHT_SYMMETRIC_6 19
1448#define ALGORITHM_PARITY_0_6 20
1449#define ALGORITHM_PARITY_N_6 ALGORITHM_PARITY_N
1450
1451/* Define PATH_MAX in case we don't use glibc or standard library does
1452 * not have PATH_MAX defined. Assume max path length is 4K characters.
1453 */
1454#ifndef PATH_MAX
1455#define PATH_MAX 4096
1456#endif
1457
1458#define RESYNC_NONE -1
1459#define RESYNC_DELAYED -2
1460#define RESYNC_PENDING -3
1461#define RESYNC_UNKNOWN -4
1462
1463/* When using "GET_DISK_INFO" it isn't certain how high
1464 * we need to check. So we impose an absolute limit of
1465 * MAX_DISKS. This needs to be much more than the largest
1466 * number of devices any metadata can support. Currently
1467 * v1.x can support 1920
1468 */
1469#define MAX_DISKS 4096
1470
1471/* Sometimes the 'size' value passed needs to mean "Maximum".
1472 * In those cases with use MAX_SIZE
1473 */
1474#define MAX_SIZE 1
1475
1476/* We want to use unsigned numbers for sector counts, but need
1477 * a value for 'invalid'. Use '1'.
1478 */
1479#define INVALID_SECTORS 1
1480
1481extern int __offroot;