]> git.ipfire.org Git - thirdparty/mdadm.git/blame_incremental - mdadm.h
build(deps): bump actions/checkout from 4 to 5
[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#define __USE_LARGEFILE64 1
28
29#include <unistd.h>
30#include <assert.h>
31#include <asm/byteorder.h>
32#include <sys/types.h>
33#include <sys/stat.h>
34#include <stdarg.h>
35#include <stdint.h>
36#include <stdlib.h>
37#include <time.h>
38#include <sys/time.h>
39#include <getopt.h>
40#include <fcntl.h>
41#include <ftw.h>
42#include <stdio.h>
43#include <errno.h>
44#include <string.h>
45#include <syslog.h>
46#include <stdbool.h>
47#include <signal.h>
48/* Newer glibc requires sys/sysmacros.h directly for makedev() */
49#include <sys/sysmacros.h>
50#ifdef __dietlibc__
51#include <strings.h>
52/* dietlibc has deprecated random and srandom!! */
53#define random rand
54#define srandom srand
55#endif
56
57#ifdef NO_COROSYNC
58#define CS_OK 1
59typedef uint64_t cmap_handle_t;
60#else
61#include <corosync/cmap.h>
62#endif
63
64#ifndef NO_DLM
65#include <libdlm.h>
66#include <errno.h>
67#else
68#define LKF_NOQUEUE 0x00000001
69#define LKM_PWMODE 4
70#define EUNLOCK 0x10002
71
72typedef void *dlm_lshandle_t;
73
74struct dlm_lksb {
75 int sb_status;
76 uint32_t sb_lkid;
77 char sb_flags;
78 char *sb_lvbptr;
79};
80#endif
81
82#include <linux/kdev_t.h>
83#include <sys/mount.h>
84#include <asm/types.h>
85#include <sys/ioctl.h>
86#define MD_MAJOR 9
87#define MdpMinorShift 6
88
89#ifndef BLKGETSIZE64
90#define BLKGETSIZE64 _IOR(0x12,114,size_t) /* return device size in bytes (u64 *arg) */
91#endif
92
93#define DEFAULT_CHUNK 512
94#define DEFAULT_BITMAP_CHUNK 4096
95#define DEFAULT_BITMAP_DELAY 5
96#define DEFAULT_MAX_WRITE_BEHIND 256
97
98#ifndef DEV_DIR
99#define DEV_DIR "/dev/"
100#define DEV_DIR_LEN (sizeof(DEV_DIR) - 1)
101#endif /* DEV_DIR */
102
103/* DEV_NUM_PREF is a subpath to numbered MD devices, e.g. /dev/md1 or directory name.
104 * DEV_NUM_PREF_LEN is a length with Null byte excluded.
105 */
106#ifndef DEV_NUM_PREF
107#define DEV_NUM_PREF "/dev/md"
108#define DEV_NUM_PREF_LEN (sizeof(DEV_NUM_PREF) - 1)
109#endif /* DEV_NUM_PREF */
110
111/* DEV_MD_DIR points to named MD devices directory.
112 * DEV_MD_DIR_LEN is a length with Null byte excluded.
113 */
114#ifndef DEV_MD_DIR
115#define DEV_MD_DIR "/dev/md/"
116#define DEV_MD_DIR_LEN (sizeof(DEV_MD_DIR) - 1)
117#endif /* DEV_MD_DIR */
118
119/* MAP_DIR should be somewhere that persists across the pivotroot
120 * from early boot to late boot.
121 * /run seems to have emerged as the best standard.
122 */
123#ifndef MAP_DIR
124#define MAP_DIR "/run/mdadm"
125#endif /* MAP_DIR */
126/* MAP_FILE is what we name the map file we put in MAP_DIR, in case you
127 * want something other than the default of "map"
128 */
129#ifndef MAP_FILE
130#define MAP_FILE "map"
131#endif /* MAP_FILE */
132/* MDMON_DIR is where pid and socket files used for communicating
133 * with mdmon normally live. Best is /var/run/mdadm as
134 * mdmon is needed at early boot then it needs to write there prior
135 * to /var/run being mounted read/write, and it also then needs to
136 * persist beyond when /var/run is mounter read-only. So, to be
137 * safe, the default is somewhere that is read/write early in the
138 * boot process and stays up as long as possible during shutdown.
139 */
140#ifndef MDMON_DIR
141#define MDMON_DIR "/run/mdadm"
142#endif /* MDMON_DIR */
143
144/* FAILED_SLOTS is where to save files storing recent removal of array
145 * member in order to allow future reuse of disk inserted in the same
146 * slot for array recovery
147 */
148#ifndef FAILED_SLOTS_DIR
149#define FAILED_SLOTS_DIR "/run/mdadm/failed-slots"
150#endif /* FAILED_SLOTS */
151
152#ifndef MDMON_SERVICE
153#define MDMON_SERVICE "mdmon"
154#endif /* MDMON_SERVICE */
155
156#ifndef GROW_SERVICE
157#define GROW_SERVICE "mdadm-grow-continue"
158#endif /* GROW_SERVICE */
159
160#include <linux/raid/md_u.h>
161#include <linux/raid/md_p.h>
162
163/* These defines might be missing in raid headers*/
164#ifndef MD_SB_BLOCK_CONTAINER_RESHAPE
165#define MD_SB_BLOCK_CONTAINER_RESHAPE 3
166#endif
167#ifndef MD_SB_BLOCK_VOLUME
168#define MD_SB_BLOCK_VOLUME 4
169#endif
170#ifndef MD_DISK_REPLACEMENT
171#define MD_DISK_REPLACEMENT 17
172#endif
173
174#include "bitmap.h"
175#include "msg.h"
176#include "mdadm_status.h"
177
178/*
179 * Partially stolen from include/linux/unaligned/packed_struct.h
180 */
181struct __una_u16 { __u16 x; } __attribute__ ((packed));
182struct __una_u32 { __u32 x; } __attribute__ ((packed));
183
184/*
185 * Ensure GNU basename behavior on GLIBC less systems.
186 */
187#ifndef __GLIBC__
188#define basename(path) \
189 (strrchr((path), '/') ? strrchr((path),'/') + 1 : (path))
190#endif
191
192static inline __u16 __get_unaligned16(const void *p)
193{
194 const struct __una_u16 *ptr = (const struct __una_u16 *)p;
195 return ptr->x;
196}
197
198static inline __u32 __get_unaligned32(const void *p)
199{
200 const struct __una_u32 *ptr = (const struct __una_u32 *)p;
201 return ptr->x;
202}
203
204static inline void __put_unaligned16(__u16 val, void *p)
205{
206 struct __una_u16 *ptr = (struct __una_u16 *)p;
207 ptr->x = val;
208}
209
210static inline void __put_unaligned32(__u32 val, void *p)
211{
212 struct __una_u32 *ptr = (struct __una_u32 *)p;
213 ptr->x = val;
214}
215
216/*
217 * Check at compile time that something is of a particular type.
218 * Always evaluates to 1 so you may use it easily in comparisons.
219*/
220
221#define typecheck(type,x) \
222({ type __dummy; \
223 typeof(x) __dummy2; \
224 (void)(&__dummy == &__dummy2); \
225 1; \
226})
227
228/*
229 * These inlines deal with timer wrapping correctly.
230 *
231 * time_after(a,b) returns true if the time a is after time b.
232*/
233
234#define time_after(a,b) \
235 (typecheck(unsigned int, a) && \
236 typecheck(unsigned int, b) && \
237 ((int)((b) - (a)) < 0))
238
239#define time_before(a,b) time_after(b,a)
240
241/*
242 * min()/max()/clamp() macros that also do
243 * strict type-checking.. See the
244 * "unnecessary" pointer comparison.
245 */
246#define min(x, y) ({ \
247 typeof(x) _min1 = (x); \
248 typeof(y) _min2 = (y); \
249 (void) (&_min1 == &_min2); \
250 _min1 < _min2 ? _min1 : _min2; })
251
252#define max(x, y) ({ \
253 typeof(x) _max1 = (x); \
254 typeof(y) _max2 = (y); \
255 (void) (&_max1 == &_max2); \
256 _max1 > _max2 ? _max1 : _max2; })
257
258#define ARRAY_SIZE(x) (sizeof(x)/sizeof(x[0]))
259
260#define KIB_TO_BYTES(x) ((x) << 10)
261#define SEC_TO_BYTES(x) ((x) << 9)
262
263/**
264 * This is true for native and DDF, IMSM allows 16.
265 */
266#define MD_NAME_MAX 32
267
268extern const char Name[];
269
270struct md_bb_entry {
271 unsigned long long sector;
272 int length;
273};
274
275struct md_bb {
276 int supported;
277 int count;
278 struct md_bb_entry *entries;
279};
280
281/* general information that might be extracted from a superblock */
282struct mdinfo {
283 mdu_array_info_t array;
284 mdu_disk_info_t disk;
285 __u64 events;
286 int uuid[4];
287 char name[33];
288 unsigned long long data_offset;
289 unsigned long long new_data_offset;
290 unsigned long long component_size; /* same as array.size, except in
291 * sectors and up to 64bits.
292 */
293 unsigned long long custom_array_size; /* size for non-default sized
294 * arrays (in sectors)
295 */
296#define NO_RESHAPE 0
297#define VOLUME_RESHAPE 1
298#define CONTAINER_RESHAPE 2
299#define RESHAPE_NO_BACKUP 16 /* Mask 'or'ed in */
300 int reshape_active;
301 unsigned long long reshape_progress;
302 int recovery_blocked; /* for external metadata it
303 * indicates that there is
304 * reshape in progress in
305 * container,
306 * for native metadata it is
307 * reshape_active field mirror
308 */
309 int journal_device_required;
310 int journal_clean;
311
312 enum {
313 CONSISTENCY_POLICY_UNKNOWN,
314 CONSISTENCY_POLICY_NONE,
315 CONSISTENCY_POLICY_RESYNC,
316 CONSISTENCY_POLICY_BITMAP,
317 CONSISTENCY_POLICY_JOURNAL,
318 CONSISTENCY_POLICY_PPL,
319 } consistency_policy;
320
321 /* During reshape we can sometimes change the data_offset to avoid
322 * over-writing still-valid data. We need to know if there is space.
323 * So getinfo_super will fill in space_before and space_after in sectors.
324 * data_offset can be increased or decreased by this amount.
325 */
326 unsigned long long space_before, space_after;
327 union {
328 unsigned long long resync_start; /* per-array resync position */
329 unsigned long long recovery_start; /* per-device rebuild position */
330 #define MaxSector (~0ULL) /* resync/recovery complete position */
331 };
332 long bitmap_offset; /* 0 == none, 1 == a file */
333 unsigned int ppl_size;
334 int ppl_offset;
335 unsigned long long ppl_sector;
336 unsigned long safe_mode_delay; /* ms delay to mark clean */
337 int new_level, delta_disks, new_layout, new_chunk;
338 int errors;
339 unsigned long cache_size; /* size of raid456 stripe cache*/
340 int mismatch_cnt;
341 char text_version[50];
342
343 int container_member; /* for assembling external-metatdata arrays
344 * This is to be used internally by metadata
345 * handler only */
346 /**
347 * flag external handlers can set to indicate that subarrays have:
348 * - not enough disks to start (-1),
349 * - enough disks to start (0),
350 * - all expected disks (1).
351 */
352 int container_enough;
353 char sys_name[32];
354 struct mdinfo *devs;
355 struct mdinfo *next;
356
357 /* Device info for mdmon: */
358 int recovery_fd;
359 int state_fd;
360 int bb_fd;
361 int ubb_fd;
362 #define DS_FAULTY 1
363 #define DS_INSYNC 2
364 #define DS_WRITE_MOSTLY 4
365 #define DS_SPARE 8
366 #define DS_BLOCKED 16
367 #define DS_REMOVE 1024
368 #define DS_UNBLOCK 2048
369 #define DS_EXTERNAL_BB 4096
370 int prev_state, curr_state, next_state;
371
372 /* If set by monitor, managemon needs to remove faulty device */
373 bool man_disk_to_remove : 1;
374
375 /* Managemon cannot close descriptors if monitor is using them for select() */
376 bool mon_descriptors_not_used : 1;
377
378 /* info read from sysfs */
379 enum {
380 ARRAY_CLEAR,
381 ARRAY_INACTIVE,
382 ARRAY_SUSPENDED,
383 ARRAY_READONLY,
384 ARRAY_READ_AUTO,
385 ARRAY_CLEAN,
386 ARRAY_ACTIVE,
387 ARRAY_WRITE_PENDING,
388 ARRAY_ACTIVE_IDLE,
389 ARRAY_BROKEN,
390 ARRAY_UNKNOWN_STATE,
391 } array_state;
392 struct md_bb bb;
393};
394
395struct createinfo {
396 int uid;
397 int gid;
398 int autof;
399 int mode;
400 int names;
401 int bblist;
402 struct supertype *supertype;
403};
404
405struct spare_criteria {
406 bool criteria_set;
407 unsigned long long min_size;
408 unsigned int sector_size;
409 struct dev_policy *pols;
410};
411
412enum mode {
413 ASSEMBLE=1,
414 BUILD,
415 CREATE,
416 MANAGE,
417 MISC,
418 MONITOR,
419 GROW,
420 INCREMENTAL,
421 AUTODETECT,
422 mode_count
423};
424
425extern char short_opts[], short_monitor_opts[], short_bitmap_opts[], short_bitmap_auto_opts[];
426
427extern struct option long_options[];
428extern char Version[], Usage[], Help[], OptionHelp[],
429 *mode_help[],
430 Help_create[], Help_build[], Help_assemble[], Help_grow[],
431 Help_incr[],
432 Help_manage[], Help_misc[], Help_monitor[], Help_config[];
433
434/* for option that don't have short equivilents, we assign arbitrary
435 * numbers later than any 'short' character option.
436 */
437enum special_options {
438 AssumeClean = 300,
439 WriteZeroes,
440 BitmapChunk,
441 WriteBehind,
442 ReAdd,
443 NoDegraded,
444 Sparc22,
445 BackupFile,
446 HomeHost,
447 AutoHomeHost,
448 AutoDetect,
449 Waitclean,
450 DetailPlatform,
451 KillSubarray,
452 UpdateSubarray,
453 IncrementalPath,
454 NoSharing,
455 HelpOptions,
456 Brief,
457 NoDevices,
458 ManageOpt,
459 Add,
460 AddSpare,
461 AddJournal,
462 Remove,
463 Fail,
464 Replace,
465 With,
466 MiscOpt,
467 WaitOpt,
468 ConfigFile,
469 ChunkSize,
470 WriteMostly,
471 FailFast,
472 NoFailFast,
473 Layout,
474 Auto,
475 Force,
476 SuperMinor,
477 EMail,
478 ProgramOpt,
479 Increment,
480 Fork,
481 Bitmap,
482 RebuildMapOpt,
483 InvalidBackup,
484 UdevRules,
485 Continue,
486 OffRootOpt,
487 Prefer,
488 KillOpt,
489 DataOffset,
490 ExamineBB,
491 Dump,
492 Restore,
493 Action,
494 Nodes,
495 ClusterName,
496 ClusterConfirm,
497 WriteJournal,
498 ConsistencyPolicy,
499};
500
501enum update_opt {
502 UOPT_UNDEFINED = 0,
503 UOPT_NAME,
504 UOPT_PPL,
505 UOPT_NO_PPL,
506 UOPT_BITMAP,
507 UOPT_NO_BITMAP,
508 UOPT_SUBARRAY_ONLY,
509 UOPT_SPARC22,
510 UOPT_SUPER_MINOR,
511 UOPT_SUMMARIES,
512 UOPT_RESYNC,
513 UOPT_UUID,
514 UOPT_HOMEHOST,
515 UOPT_HOME_CLUSTER,
516 UOPT_NODES,
517 UOPT_DEVICESIZE,
518 UOPT_BBL,
519 UOPT_NO_BBL,
520 UOPT_FORCE_NO_BBL,
521 UOPT_METADATA,
522 UOPT_REVERT_RESHAPE,
523 UOPT_LAYOUT_ORIGINAL,
524 UOPT_LAYOUT_ALTERNATE,
525 UOPT_LAYOUT_UNSPECIFIED,
526 UOPT_BYTEORDER,
527 UOPT_HELP,
528 UOPT_USER_ONLY,
529 /*
530 * Code specific options, cannot be set by the user
531 */
532 UOPT_SPEC_FORCE_ONE,
533 UOPT_SPEC_FORCE_ARRAY,
534 UOPT_SPEC_ASSEMBLE,
535 UOPT_SPEC_LINEAR_GROW_NEW,
536 UOPT_SPEC_LINEAR_GROW_UPDATE,
537 UOPT_SPEC__RESHAPE_PROGRESS,
538 UOPT_SPEC_WRITEMOSTLY,
539 UOPT_SPEC_READWRITE,
540 UOPT_SPEC_FAILFAST,
541 UOPT_SPEC_NOFAILFAST,
542 UOPT_SPEC_REVERT_RESHAPE_NOBACKUP,
543};
544extern void fprint_update_options(FILE *outf, enum update_opt update_mode);
545
546enum prefix_standard {
547 JEDEC,
548 IEC
549};
550
551enum bitmap_update {
552 NoUpdate,
553 NameUpdate,
554 NodeNumUpdate,
555};
556
557enum bitmap_type {
558 BitmapNone,
559 BitmapInternal,
560 BitmapCluster,
561 BitmapUnknown,
562};
563
564enum flag_mode {
565 FlagDefault, FlagSet, FlagClear,
566};
567
568typedef enum {
569 ROLLBACK_METADATA_CHANGES,
570 APPLY_METADATA_CHANGES
571} change_dir_t;
572
573/* structures read from config file */
574/* List of mddevice names and identifiers
575 * Identifiers can be:
576 * uuid=128-hex-uuid
577 * super-minor=decimal-minor-number-from-superblock
578 * devices=comma,separated,list,of,device,names,with,wildcards
579 *
580 * If multiple fields are present, the intersection of all matching
581 * devices is considered
582 */
583#define UnSet (0xfffe)
584struct mddev_ident {
585 char *devname;
586
587 int uuid_set;
588 int uuid[4];
589 char name[MD_NAME_MAX + 1];
590
591 int super_minor;
592
593 char *devices; /* comma separated list of device
594 * names with wild cards
595 */
596 int level;
597 int raid_disks;
598 int spare_disks;
599 struct supertype *st;
600 char *spare_group;
601 enum bitmap_type btype;
602
603 char *container; /* /dev/whatever name of container, or
604 * uuid of container. You would expect
605 * this to be the 'devname' or UUID
606 * of some other entry.
607 */
608 char *member; /* subarray within a container */
609
610 struct mddev_ident *next;
611 union {
612 /* fields needed by different users of this structure */
613 int assembled; /* set when assembly succeeds */
614 };
615};
616
617struct context {
618 int readonly;
619 int runstop;
620 int verbose;
621 int brief;
622 int no_devices;
623 int force;
624 char *homehost;
625 int require_homehost;
626 char *prefer;
627 int export;
628 int test;
629 char *subarray;
630 enum update_opt update;
631 int scan;
632 int SparcAdjust;
633 int delay;
634 char *backup_file;
635 int invalid_backup;
636 char *action;
637 int nodes;
638 char *homecluster;
639};
640
641struct shape {
642 char *dev;
643 int raiddisks;
644 int delta_disks;
645 int sparedisks;
646 int journaldisks;
647 int level;
648 int layout;
649 char *layout_str;
650 int chunk;
651 int bitmap_chunk;
652 enum bitmap_type btype;
653 int assume_clean;
654 bool write_zeroes;
655 int write_behind;
656 unsigned long long size;
657 unsigned long long data_offset;
658 int consistency_policy;
659 change_dir_t direction;
660};
661
662/* List of device names - wildcards expanded */
663struct mddev_dev {
664 char *devname;
665 int disposition; /* 'a' for add, 'r' for remove, 'f' for fail,
666 * 'A' for re_add.
667 * Not set for names read from .config
668 */
669 enum flag_mode writemostly;
670 enum flag_mode failfast;
671 int used; /* set when used */
672 long long data_offset;
673 struct mddev_dev *next;
674};
675
676typedef struct mapping {
677 char *name;
678 int num;
679} mapping_t;
680
681struct mdstat_ent {
682 char devnm[32];
683 int active;
684 char *level;
685 char *pattern; /* U for up, _ for down */
686 int percent; /* -1 if no resync */
687 int resync; /* 3 if check, 2 if reshape, 1 if resync, 0 if recovery */
688 int devcnt;
689 int raid_disks;
690 char * metadata_version;
691 struct dev_member {
692 char *name;
693 struct dev_member *next;
694 } *members;
695 struct mdstat_ent *next;
696};
697
698extern struct mdstat_ent *mdstat_read(int hold, int start);
699extern void mdstat_close(void);
700extern void free_mdstat(struct mdstat_ent *ms);
701extern int mdstat_wait(int seconds);
702extern void mdstat_wait_fd(int fd, const sigset_t *sigmask);
703extern int mddev_busy(char *devnm);
704extern struct mdstat_ent *mdstat_by_component(char *name);
705extern struct mdstat_ent *mdstat_find_by_member_name(struct mdstat_ent *mdstat, char *member_devnm);
706extern struct mdstat_ent *mdstat_by_subdev(char *subdev, char *container);
707
708extern bool is_mdstat_ent_external(struct mdstat_ent *ent);
709extern bool is_mdstat_ent_subarray(struct mdstat_ent *ent);
710
711struct map_ent {
712 struct map_ent *next;
713 char devnm[32];
714 char metadata[20];
715 int uuid[4];
716 int bad;
717 char *path;
718};
719extern int map_update(struct map_ent **mpp, char *devnm, char *metadata,
720 int uuid[4], char *path);
721extern void map_remove(struct map_ent **map, char *devnm);
722extern struct map_ent *map_by_uuid(struct map_ent **map, int uuid[4]);
723extern struct map_ent *map_by_devnm(struct map_ent **map, char *devnm);
724extern void map_free(struct map_ent *map);
725extern struct map_ent *map_by_name(struct map_ent **map, char *name);
726extern void map_read(struct map_ent **melp);
727extern int map_write(struct map_ent *mel);
728extern void map_delete(struct map_ent **mapp, char *devnm);
729extern void map_add(struct map_ent **melp,
730 char *devnm, char *metadata, int uuid[4], char *path);
731extern int map_lock(struct map_ent **melp);
732extern void map_unlock(struct map_ent **melp);
733extern void map_fork(void);
734
735/* various details can be requested */
736enum sysfs_read_flags {
737 GET_LEVEL = (1 << 0),
738 GET_LAYOUT = (1 << 1),
739 GET_COMPONENT = (1 << 2),
740 GET_CHUNK = (1 << 3),
741 GET_CACHE = (1 << 4),
742 GET_MISMATCH = (1 << 5),
743 GET_VERSION = (1 << 6),
744 GET_DISKS = (1 << 7),
745 GET_SAFEMODE = (1 << 9),
746 GET_BITMAP_LOCATION = (1 << 10),
747
748 GET_DEVS = (1 << 20), /* gets role, major, minor */
749 GET_OFFSET = (1 << 21),
750 GET_SIZE = (1 << 22),
751 GET_STATE = (1 << 23),
752 GET_ERROR = (1 << 24),
753 GET_ARRAY_STATE = (1 << 25),
754 GET_CONSISTENCY_POLICY = (1 << 26),
755 GET_DEVS_ALL = (1 << 27),
756};
757
758#define SYSFS_MAX_BUF_SIZE 64
759
760/**
761 * Defines md/<disk>/state possible values.
762 * Note that remove can't be read-back from the file.
763 *
764 * This is not complete list.
765 */
766typedef enum memb_state {
767 MEMB_STATE_EXTERNAL_BBL,
768 MEMB_STATE_BLOCKED,
769 MEMB_STATE_SPARE,
770 MEMB_STATE_WRITE_MOSTLY,
771 MEMB_STATE_IN_SYNC,
772 MEMB_STATE_FAULTY,
773 MEMB_STATE_REMOVE,
774 MEMB_STATE_UNKNOWN
775} memb_state_t;
776char *map_memb_state(memb_state_t state);
777
778extern mdadm_status_t sysfs_write_descriptor(const int fd, const char *value,
779 const ssize_t len, int *errno_p);
780extern mdadm_status_t write_attr(const char *value, const int fd);
781extern mdadm_status_t sysfs_set_memb_state_fd(int fd, memb_state_t state, int *err);
782extern mdadm_status_t sysfs_set_memb_state(char *array_devnm, char *memb_devnm, memb_state_t state);
783extern void sysfs_get_container_devnm(struct mdinfo *mdi, char *buf);
784
785extern int sysfs_open(char *devnm, char *devname, char *attr);
786extern int sysfs_open_memb_attr(char *array_devnm, char *memb_devnm, char *attr, int oflag);
787extern int sysfs_init(struct mdinfo *mdi, int fd, char *devnm);
788extern void sysfs_init_dev(struct mdinfo *mdi, dev_t devid);
789extern void sysfs_free(struct mdinfo *sra);
790
791extern struct mdinfo *sysfs_read(int fd, char *devnm, unsigned long options);
792extern int sysfs_attr_match(const char *attr, const char *str);
793extern int sysfs_match_word(const char *word, char **list);
794extern int sysfs_set_str(struct mdinfo *sra, struct mdinfo *dev,
795 char *name, char *val);
796extern int sysfs_set_num(struct mdinfo *sra, struct mdinfo *dev,
797 char *name, unsigned long long val);
798extern int sysfs_set_num_signed(struct mdinfo *sra, struct mdinfo *dev,
799 char *name, long long val);
800extern int sysfs_uevent(struct mdinfo *sra, char *event);
801extern int sysfs_get_fd(struct mdinfo *sra, struct mdinfo *dev,
802 char *name);
803extern int sysfs_fd_get_ll(int fd, unsigned long long *val);
804extern int sysfs_get_ll(struct mdinfo *sra, struct mdinfo *dev,
805 char *name, unsigned long long *val);
806extern int sysfs_fd_get_two(int fd, unsigned long long *v1, unsigned long long *v2);
807extern int sysfs_get_two(struct mdinfo *sra, struct mdinfo *dev,
808 char *name, unsigned long long *v1, unsigned long long *v2);
809extern int sysfs_fd_get_str(int fd, char *val, int size);
810extern int sysfs_attribute_available(struct mdinfo *sra, struct mdinfo *dev,
811 char *name);
812extern int sysfs_get_str(struct mdinfo *sra, struct mdinfo *dev,
813 char *name, char *val, int size);
814extern int sysfs_set_safemode(struct mdinfo *sra, unsigned long ms);
815extern int sysfs_set_array(struct mdinfo *info);
816extern int sysfs_add_disk(struct mdinfo *sra, struct mdinfo *sd, int resume);
817extern int sysfs_disk_to_scsi_id(int fd, __u32 *id);
818extern int sysfs_unique_holder(char *devnm, long rdev);
819extern int sysfs_freeze_array(struct mdinfo *sra);
820extern int sysfs_wait(int fd, int *msec);
821extern int load_sys(char *path, char *buf, int len);
822extern int zero_disk_range(int fd, unsigned long long sector, size_t count);
823extern int reshape_prepare_fdlist(char *devname,
824 struct mdinfo *sra,
825 int raid_disks,
826 int nrdisks,
827 unsigned long blocks,
828 char *backup_file,
829 int *fdlist,
830 unsigned long long *offsets);
831extern void reshape_free_fdlist(int *fdlist,
832 unsigned long long *offsets,
833 int size);
834extern int reshape_open_backup_file(char *backup,
835 int fd,
836 char *devname,
837 long blocks,
838 int *fdlist,
839 unsigned long long *offsets,
840 char *sysfs_name,
841 int restart);
842extern unsigned long compute_backup_blocks(int nchunk, int ochunk,
843 unsigned int ndata, unsigned int odata);
844extern char *locate_backup(char *name);
845extern char *make_backup(char *name);
846
847extern int save_stripes(int *source, unsigned long long *offsets,
848 int raid_disks, int chunk_size, int level, int layout,
849 int nwrites, int *dest,
850 unsigned long long start, unsigned long long length,
851 char *buf);
852extern int restore_stripes(int *dest, unsigned long long *offsets,
853 int raid_disks, int chunk_size, int level, int layout,
854 int source, unsigned long long read_offset,
855 unsigned long long start, unsigned long long length,
856 char *src_buf);
857extern bool sysfs_is_libata_allow_tpm_enabled(const int verbose);
858
859#ifndef Sendmail
860#define Sendmail "/usr/lib/sendmail -t"
861#endif
862
863#define SYSLOG_FACILITY LOG_DAEMON
864extern char *map_num_s(mapping_t *map, int num);
865extern char *map_num(mapping_t *map, int num);
866extern int map_name(mapping_t *map, char *name);
867extern mapping_t r0layout[], r5layout[], r6layout[],
868 pers[], modes[], faultylayout[];
869extern mapping_t consistency_policies[], sysfs_array_states[], update_options[];
870
871extern char *map_dev_preferred(int major, int minor, int create,
872 char *prefer);
873static inline char *map_dev(int major, int minor, int create)
874{
875 return map_dev_preferred(major, minor, create, NULL);
876}
877
878/**
879 * is_fd_valid() - check file descriptor.
880 * @fd: file descriptor.
881 *
882 * The function checks if @fd is nonnegative integer and shall be used only
883 * to verify open() result.
884 */
885static inline int is_fd_valid(int fd)
886{
887 return (fd > -1);
888}
889
890/**
891 * is_level456() - check whether given level is between inclusive 4 and 6.
892 * @level: level to check.
893 *
894 * Return: true if condition is met, false otherwise
895 */
896static inline bool is_level456(int level)
897{
898 return (level >= 4 && level <= 6);
899}
900
901/**
902 * close_fd() - verify, close and unset file descriptor.
903 * @fd: pointer to file descriptor.
904 *
905 * The function closes and invalidates file descriptor if appropriative. It
906 * ignores incorrect file descriptor quitely to simplify error handling.
907 */
908static inline void close_fd(int *fd)
909{
910 if (is_fd_valid(*fd) && close(*fd) == 0)
911 *fd = -1;
912}
913
914struct active_array;
915struct metadata_update;
916
917/* 'struct reshape' records the intermediate states of
918 * a general reshape.
919 * The starting geometry is converted to the 'before' geometry
920 * by at most an atomic level change. They could be the same.
921 * Similarly the 'after' geometry is converted to the final
922 * geometry by at most a level change.
923 * Note that 'before' and 'after' must have the same level.
924 * 'blocks' is the minimum number of sectors for a reshape unit.
925 * This will be a multiple of the stripe size in each of the
926 * 'before' and 'after' geometries.
927 * If 'blocks' is 0, no restriping is necessary.
928 * 'min_offset_change' is the minimum change to data_offset to
929 * allow the reshape to happen. It is at least the larger of
930 * the old and new chunk sizes, and typically the same as 'blocks'
931 * divided by number of data disks.
932 */
933struct reshape {
934 int level;
935 int parity; /* number of parity blocks/devices */
936 struct {
937 int layout;
938 int data_disks;
939 } before, after;
940 unsigned long long backup_blocks;
941 unsigned long long min_offset_change;
942 unsigned long long stripes; /* number of old stripes that comprise 'blocks'*/
943 unsigned long long new_size; /* New size of array in sectors */
944};
945
946/**
947 * struct dev_policy - Data structure for policy management.
948 * @next: pointer to next dev_policy.
949 * @name: policy name, category.
950 * @metadata: the metadata type it affects.
951 * @value: value of the policy.
952 *
953 * The functions to manipulate dev_policy lists do not free elements, so they must be statically
954 * allocated. @name and @metadata can be compared by address.
955 */
956typedef struct dev_policy {
957 struct dev_policy *next;
958 char *name;
959 const char *metadata;
960 const char *value;
961} dev_policy_t;
962
963/* A superswitch provides entry point to a metadata handler.
964 *
965 * The superswitch primarily operates on some "metadata" that
966 * is accessed via the 'supertype'.
967 * This metadata has one of three possible sources.
968 * 1/ It is read from a single device. In this case it may not completely
969 * describe the array or arrays as some information might be on other
970 * devices.
971 * 2/ It is read from all devices in a container. In this case all
972 * information is present.
973 * 3/ It is created by ->init_super / ->add_to_super. In this case it will
974 * be complete once enough ->add_to_super calls have completed.
975 *
976 * When creating an array inside a container, the metadata will be
977 * formed by a combination of 2 and 3. The metadata or the array is read,
978 * then new information is added.
979 *
980 * The metadata must sometimes have a concept of a 'current' array
981 * and a 'current' device.
982 * The 'current' array is set by init_super to be the newly created array,
983 * or is set by super_by_fd when it finds it is looking at an array inside
984 * a container.
985 *
986 * The 'current' device is either the device that the metadata was read from
987 * in case 1, or the last device added by add_to_super in case 3.
988 * Case 2 does not identify a 'current' device.
989 */
990extern struct superswitch {
991
992 /* Used to report details of metadata read from a component
993 * device. ->load_super has been called.
994 */
995 void (*examine_super)(struct supertype *st, char *homehost);
996 void (*brief_examine_super)(struct supertype *st, int verbose);
997 void (*brief_examine_subarrays)(struct supertype *st, int verbose);
998 void (*export_examine_super)(struct supertype *st);
999 int (*examine_badblocks)(struct supertype *st, int fd, char *devname);
1000 int (*copy_metadata)(struct supertype *st, int from, int to);
1001
1002 /* Used to report details of an active array.
1003 * ->load_super was possibly given a 'component' string.
1004 */
1005 void (*detail_super)(struct supertype *st, char *homehost,
1006 char *subarray);
1007 void (*brief_detail_super)(struct supertype *st, char *subarray);
1008 void (*export_detail_super)(struct supertype *st);
1009
1010 /* Optional: platform hardware / firmware details */
1011 int (*detail_platform)(int verbose, int enumerate_only, char *controller_path);
1012 int (*export_detail_platform)(int verbose, char *controller_path);
1013
1014 /* Used:
1015 * to get uuid to storing in bitmap metadata
1016 * and 'reshape' backup-data metadata
1017 * To see if a device is being re-added to an array it was part of.
1018 */
1019 void (*uuid_from_super)(struct supertype *st, int uuid[4]);
1020
1021 /* Extract generic details from metadata. This could be details about
1022 * the container, or about an individual array within the container.
1023 * The determination is made either by:
1024 * load_super being given a 'component' string.
1025 * validate_geometry determining what to create.
1026 * The info includes both array information and device information.
1027 * The particular device should be:
1028 * The last device added by add_to_super
1029 * The device the metadata was loaded from by load_super
1030 * If 'map' is present, then it is an array raid_disks long
1031 * (raid_disk must already be set and correct) and it is filled
1032 * with 1 for slots that are thought to be active and 0 for slots which
1033 * appear to be failed/missing.
1034 * *info is zeroed out before data is added.
1035 */
1036 void (*getinfo_super)(struct supertype *st, struct mdinfo *info, char *map);
1037 struct mdinfo *(*getinfo_super_disks)(struct supertype *st);
1038 /* Check if the given metadata is flagged as belonging to "this"
1039 * host. 0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
1040 */
1041 int (*match_home)(struct supertype *st, char *homehost);
1042
1043 /* Make one of several generic modifications to metadata
1044 * prior to assembly (or other times).
1045 * sparc2.2 - first bug in early 0.90 metadata
1046 * super-minor - change name of 0.90 metadata
1047 * summaries - 'correct' any redundant data
1048 * resync - mark array as dirty to trigger a resync.
1049 * uuid - set new uuid - only 0.90 or 1.x
1050 * name - change the name of the array (where supported)
1051 * homehost - change which host this array is tied to.
1052 * devicesize - If metadata is at start of device, change recorded
1053 * device size to match actual device size
1054 * byteorder - swap bytes for 0.90 metadata
1055 *
1056 * force-one - mark that device as uptodate, not old or failed.
1057 * force-array - mark array as clean if it would not otherwise
1058 * assemble
1059 * assemble - not sure how this is different from force-one...
1060 * linear-grow-new - add a new device to a linear array, but don't
1061 * change the size: so superblock still matches
1062 * linear-grow-update - now change the size of the array.
1063 * writemostly - set the WriteMostly1 bit in the superblock devflags
1064 * readwrite - clear the WriteMostly1 bit in the superblock devflags
1065 * failfast - set the FailFast1 bit in the superblock
1066 * nofailfast - clear the FailFast1 bit
1067 * no-bitmap - clear any record that a bitmap is present.
1068 * bbl - add a bad-block-log if possible
1069 * no-bbl - remove any bad-block-log is it is empty.
1070 * force-no-bbl - remove any bad-block-log even if empty.
1071 * revert-reshape - If a reshape is in progress, modify metadata so
1072 * it will resume going in the opposite direction.
1073 */
1074 int (*update_super)(struct supertype *st, struct mdinfo *info,
1075 enum update_opt update,
1076 char *devname, int verbose,
1077 int uuid_set, char *homehost);
1078
1079 /* Create new metadata for new array as described. This could
1080 * be a new container, or an array in a pre-existing container.
1081 * Also used to zero metadata prior to writing it to invalidate old
1082 * metadata.
1083 */
1084 int (*init_super)(struct supertype *st, mdu_array_info_t *info,
1085 struct shape *s, char *name,
1086 char *homehost, int *uuid,
1087 unsigned long long data_offset);
1088
1089 /* update the metadata to include new device, either at create or
1090 * when hot-adding a spare.
1091 */
1092 int (*add_to_super)(struct supertype *st, mdu_disk_info_t *dinfo,
1093 int fd, char *devname,
1094 unsigned long long data_offset);
1095 /* update the metadata to delete a device,
1096 * when hot-removing.
1097 */
1098 int (*remove_from_super)(struct supertype *st, mdu_disk_info_t *dinfo);
1099
1100 /* Write metadata to one device when fixing problems or adding
1101 * a new device.
1102 */
1103 int (*store_super)(struct supertype *st, int fd);
1104
1105 /* Write all metadata for this array.
1106 */
1107 int (*write_init_super)(struct supertype *st);
1108 /* Check if metadata read from one device is compatible with an array,
1109 * used when assembling an array, or pseudo-assembling was with
1110 * "--examine --brief"
1111 * If "st" has not yet been loaded the superblock from, "tst" is
1112 * moved in, otherwise the superblock in 'st' is compared with
1113 * 'tst'.
1114 */
1115 int (*compare_super)(struct supertype *st, struct supertype *tst,
1116 int verbose);
1117 /* Load metadata from a single device. If 'devname' is not NULL
1118 * print error messages as appropriate */
1119 int (*load_super)(struct supertype *st, int fd, char *devname);
1120 /* 'fd' is a 'container' md array - load array metadata from the
1121 * whole container.
1122 */
1123 int (*load_container)(struct supertype *st, int fd, char *devname);
1124 /* If 'arg' is a valid name of this metadata type, allocate and
1125 * return a 'supertype' for the particular minor version */
1126 struct supertype * (*match_metadata_desc)(char *arg);
1127 /* If a device has the given size, and the data_offset has been
1128 * requested - work out how much space is available for data.
1129 * This involves adjusting for reserved space (e.g. bitmaps)
1130 * and for any rounding.
1131 * 'mdadm' only calls this for existing arrays where a possible
1132 * spare is being added. However some super-handlers call it
1133 * internally from validate_geometry when creating an array.
1134 */
1135 __u64 (*avail_size)(struct supertype *st, __u64 size,
1136 unsigned long long data_offset);
1137 /*
1138 * Return spare criteria for array:
1139 * - minimum disk size can be used in array;
1140 * - sector size can be used in array.
1141 */
1142 mdadm_status_t (*get_spare_criteria)(struct supertype *st, char *mddev_path,
1143 struct spare_criteria *sc);
1144 /* Find somewhere to put a bitmap - possibly auto-size it - and
1145 * update the metadata to record this. The array may be newly
1146 * created, in which case data_size may be updated, or it might
1147 * already exist. Metadata handler can know if init_super
1148 * has been called, but not write_init_super.
1149 * 0: Success
1150 * -Exxxx: On error
1151 */
1152 int (*add_internal_bitmap)(struct supertype *st, int *chunkp,
1153 int delay, int write_behind,
1154 unsigned long long size, int may_change, int major);
1155 /* Perform additional setup required to activate a bitmap.
1156 */
1157 int (*set_bitmap)(struct supertype *st, struct mdinfo *info);
1158 /* Seek 'fd' to start of write-intent-bitmap. Must be an
1159 * md-native format bitmap
1160 */
1161 int (*locate_bitmap)(struct supertype *st, int fd, int node_num);
1162 /* if add_internal_bitmap succeeded for existing array, this
1163 * writes it out.
1164 */
1165 int (*write_bitmap)(struct supertype *st, int fd, enum bitmap_update update);
1166 /* Free the superblock and any other allocated data */
1167 void (*free_super)(struct supertype *st);
1168
1169 /* validate_geometry is called with an st returned by
1170 * match_metadata_desc.
1171 * It should check that the geometry described is compatible with
1172 * the metadata type. It will be called repeatedly as devices
1173 * added to validate changing size and new devices. If there are
1174 * inter-device dependencies, it should record sufficient details
1175 * so these can be validated.
1176 * Both 'size' and '*freesize' are in sectors. chunk is KiB.
1177 * Return value is:
1178 * 1: everything is OK
1179 * 0: not OK for some reason - if 'verbose', then error was reported.
1180 * -1: st->sb was NULL, 'subdev' is a member of a container of this
1181 * type, but array is not acceptable for some reason
1182 * message was reported even if verbose is 0.
1183 */
1184 int (*validate_geometry)(struct supertype *st, int level, int layout,
1185 int raiddisks,
1186 int *chunk, unsigned long long size,
1187 unsigned long long data_offset,
1188 char *subdev, unsigned long long *freesize,
1189 int consistency_policy, int verbose);
1190
1191 /**
1192 * test_and_add_drive_policies() - test new and add custom policies from metadata handler.
1193 * @pols: list of currently recorded policies.
1194 * @disk_fd: file descriptor of the device to check.
1195 * @verbose: verbose flag.
1196 *
1197 * Used by IMSM to verify all drives in container/array, against requirements not recored
1198 * in superblock, like controller type for IMSM. It should check all drives even if
1199 * they are not actually used, because mdmon or kernel are free to use any drive assigned to
1200 * container automatically.
1201 *
1202 * Generating and comparison methods belong to metadata handler. It is not mandatory to be
1203 * implemented.
1204 *
1205 * Return: MDADM_STATUS_SUCCESS is expected on success.
1206 */
1207 mdadm_status_t (*test_and_add_drive_policies)(dev_policy_t **pols, int disk_fd,
1208 const int verbose);
1209
1210 /* Return a linked list of 'mdinfo' structures for all arrays
1211 * in the container. For non-containers, it is like
1212 * getinfo_super with an allocated mdinfo.*/
1213 struct mdinfo *(*container_content)(struct supertype *st, char *subarray);
1214 /* query the supertype for default geometry */
1215 void (*default_geometry)(struct supertype *st, int *level, int *layout, int *chunk); /* optional */
1216 /* Permit subarray's to be deleted from inactive containers */
1217 int (*kill_subarray)(struct supertype *st,
1218 char *subarray_id); /* optional */
1219 /**
1220 * update_subarray() - Permit subarray to be modified.
1221 * @st: Supertype.
1222 * @subarray: Subarray name.
1223 * @update: Update option.
1224 * @ident: Optional identifiers.
1225 */
1226 int (*update_subarray)(struct supertype *st, char *subarray,
1227 enum update_opt update, struct mddev_ident *ident);
1228 /* Check if reshape is supported for this external format.
1229 * st is obtained from super_by_fd() where st->subarray[0] is
1230 * initialized to indicate if reshape is being performed at the
1231 * container or subarray level
1232 */
1233
1234 int (*reshape_super)(struct supertype *st, struct shape *shape, struct context *c);
1235 int (*manage_reshape)( /* optional */
1236 int afd, struct mdinfo *sra, struct reshape *reshape,
1237 struct supertype *st, unsigned long blocks,
1238 int *fds, unsigned long long *offsets,
1239 int dests, int *destfd, unsigned long long *destoffsets);
1240
1241/* for mdmon */
1242 int (*open_new)(struct supertype *c, struct active_array *a,
1243 int inst);
1244
1245 /* Tell the metadata handler the current state of the array.
1246 * This covers whether it is known to be consistent (no pending writes)
1247 * and how far along a resync is known to have progressed
1248 * (in a->resync_start).
1249 * resync status is really irrelevant if the array is not consistent,
1250 * but some metadata (DDF!) have a place to record the distinction.
1251 * If 'consistent' is '2', then the array can mark it dirty if a
1252 * resync/recovery/whatever is required, or leave it clean if not.
1253 * Return value is 0 dirty (not consistent) and 1 if clean.
1254 * it is only really important if consistent is passed in as '2'.
1255 */
1256 int (*set_array_state)(struct active_array *a, int consistent);
1257
1258 /* When the state of a device might have changed, we call set_disk to
1259 * tell the metadata what the current state is.
1260 * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
1261 * transitions.
1262 * set_disk might be called when the state of the particular disk has
1263 * not in fact changed.
1264 */
1265 void (*set_disk)(struct active_array *a, int n, int state);
1266 void (*sync_metadata)(struct supertype *st);
1267 void (*process_update)(struct supertype *st,
1268 struct metadata_update *update);
1269 /* Prepare updates allocates extra memory that might be
1270 * needed. If the update cannot be understood, return 0.
1271 */
1272 int (*prepare_update)(struct supertype *st,
1273 struct metadata_update *update);
1274
1275 /* activate_spare will check if the array is degraded and, if it
1276 * is, try to find some spare space in the container.
1277 * On success, it add appropriate updates (For process_update) to
1278 * to the 'updates' list and returns a list of 'mdinfo' identifying
1279 * the device, or devices as there might be multiple missing
1280 * devices and multiple spares available.
1281 */
1282 struct mdinfo *(*activate_spare)(struct active_array *a,
1283 struct metadata_update **updates);
1284
1285 /* for external backup area */
1286 int (*recover_backup)(struct supertype *st, struct mdinfo *info);
1287
1288 /* validate container after assemble */
1289 int (*validate_container)(struct mdinfo *info);
1290
1291 /* write initial empty PPL on device */
1292 int (*write_init_ppl)(struct supertype *st, struct mdinfo *info, int fd);
1293
1294 /* validate ppl before assemble */
1295 int (*validate_ppl)(struct supertype *st, struct mdinfo *info,
1296 struct mdinfo *disk);
1297
1298 /* records new bad block in metadata */
1299 int (*record_bad_block)(struct active_array *a, int n,
1300 unsigned long long sector, int length);
1301
1302 /* clears bad block from metadata */
1303 int (*clear_bad_block)(struct active_array *a, int n,
1304 unsigned long long sector, int length);
1305
1306 /* get list of bad blocks from metadata */
1307 struct md_bb *(*get_bad_blocks)(struct active_array *a, int n);
1308
1309 int swapuuid; /* true if uuid is bigending rather than hostendian */
1310 int external;
1311 const char *name; /* canonical metadata name */
1312} *superlist[];
1313
1314extern struct superswitch super0, super1;
1315extern struct superswitch super_imsm, super_ddf;
1316extern struct superswitch mbr, gpt;
1317
1318void imsm_set_no_platform(int v);
1319
1320struct metadata_update {
1321 int len;
1322 char *buf;
1323 void *space; /* allocated space that monitor will use */
1324 void **space_list; /* list of allocated spaces that monitor can
1325 * use or that it returned.
1326 */
1327 struct metadata_update *next;
1328};
1329
1330/* A supertype holds a particular collection of metadata.
1331 * It identifies the metadata type by the superswitch, and the particular
1332 * sub-version of that metadata type.
1333 * metadata read in or created is stored in 'sb' and 'info'.
1334 * There are also fields used by mdmon to track containers.
1335 *
1336 * A supertype may refer to:
1337 * Just an array, possibly in a container
1338 * A container, not identifying any particular array
1339 * Info read from just one device, not yet fully describing the array/container.
1340 *
1341 *
1342 * A supertype is created by:
1343 * super_by_fd
1344 * guess_super
1345 * dup_super
1346 */
1347struct supertype {
1348 struct superswitch *ss;
1349 int minor_version;
1350 int max_devs;
1351 char container_devnm[32]; /* devnm of container */
1352 void *sb;
1353 void *info;
1354 void *other; /* Hack used to convert v0.90 to v1.0 */
1355 unsigned long long devsize;
1356 unsigned long long data_offset; /* used by v1.x only */
1357 int ignore_hw_compat; /* used to inform metadata handlers that it should ignore
1358 HW/firmware related incompatability to load metadata.
1359 Used when examining metadata to display content of disk
1360 when user has no hw/firmare compatible system.
1361 */
1362 struct metadata_update *updates;
1363 struct metadata_update **update_tail;
1364
1365 /* extra stuff used by mdmon */
1366 struct active_array *arrays;
1367 int sock; /* listen to external programs */
1368 char devnm[32]; /* e.g. md0. This appears in metadata_version:
1369 * external:/md0/12
1370 */
1371 int devcnt;
1372 int retry_soon;
1373 int nodes;
1374 char *cluster_name;
1375
1376 struct mdinfo *devs;
1377
1378};
1379
1380extern struct supertype *super_by_fd(int fd, char **subarray);
1381enum guess_types { guess_any, guess_array, guess_partitions };
1382extern struct supertype *guess_super_type(int fd, enum guess_types guess_type);
1383static inline struct supertype *guess_super(int fd) {
1384 return guess_super_type(fd, guess_any);
1385}
1386extern struct supertype *dup_super(struct supertype *st);
1387extern int get_dev_size(int fd, char *dname, unsigned long long *sizep);
1388extern int get_dev_sector_size(int fd, char *dname, unsigned int *sectsizep);
1389extern int must_be_container(int fd);
1390void wait_for(char *dev, int fd);
1391
1392extern char pol_act[], pol_domain[], pol_metadata[], pol_auto[];
1393
1394/* iterate over the sublist starting at list, having the same
1395 * 'name' as 'list', and matching the given metadata (Where
1396 * NULL matches anything
1397 */
1398#define pol_for_each(item, list, _metadata) \
1399 for (item = list; \
1400 item && item->name == list->name; \
1401 item = item->next) \
1402 if (!(!_metadata || !item->metadata || _metadata == item->metadata)) \
1403 ; else
1404
1405/*
1406 * policy records read from mdadm are largely just name-value pairs.
1407 * The names are constants, not strdupped
1408 */
1409struct pol_rule {
1410 struct pol_rule *next;
1411 char *type; /* rule_policy or rule_part */
1412 struct rule {
1413 struct rule *next;
1414 char *name;
1415 char *value;
1416 char *dups; /* duplicates of 'value' with a partNN appended */
1417 } *rule;
1418};
1419
1420extern char rule_policy[], rule_part[];
1421extern char rule_path[], rule_type[];
1422extern char type_part[], type_disk[];
1423
1424extern void policyline(char *line, char *type);
1425extern void policy_add(char *type, ...);
1426extern void policy_free(void);
1427
1428extern struct dev_policy *path_policy(char **paths, char *type);
1429extern struct dev_policy *disk_policy(struct mdinfo *disk);
1430extern struct dev_policy *devid_policy(int devid);
1431extern void dev_policy_free(struct dev_policy *p);
1432
1433extern void pol_add(struct dev_policy **pol, char *name, char *val, char *metadata);
1434extern struct dev_policy *pol_find(struct dev_policy *pol, char *name);
1435
1436extern mdadm_status_t drive_test_and_add_policies(struct supertype *st, dev_policy_t **pols,
1437 int fd, const int verbose);
1438extern mdadm_status_t sysfs_test_and_add_drive_policies(struct supertype *st, dev_policy_t **pols,
1439 struct mdinfo *mdi, const int verbose);
1440extern mdadm_status_t mddev_test_and_add_drive_policies(struct supertype *st, dev_policy_t **pols,
1441 int array_fd, const int verbose);
1442
1443enum policy_action {
1444 act_default,
1445 act_include,
1446 act_re_add,
1447 act_spare, /* This only applies to bare devices */
1448 act_spare_same_slot, /* this allows non-bare devices,
1449 * but only if recent removal */
1450 act_force_spare, /* this allow non-bare devices in any case */
1451 act_err
1452};
1453
1454extern int policy_action_allows(struct dev_policy *plist, const char *metadata,
1455 enum policy_action want);
1456extern int disk_action_allows(struct mdinfo *disk, const char *metadata,
1457 enum policy_action want);
1458
1459struct domainlist {
1460 struct domainlist *next;
1461 const char *dom;
1462};
1463
1464extern int domain_test(struct domainlist *dom, struct dev_policy *pol,
1465 const char *metadata);
1466extern struct domainlist *domain_from_array(struct mdinfo *mdi,
1467 const char *metadata);
1468extern void domainlist_add_dev(struct domainlist **dom, int devid,
1469 const char *metadata);
1470extern void domain_free(struct domainlist *dl);
1471extern void domain_merge(struct domainlist **domp, struct dev_policy *pol,
1472 const char *metadata);
1473void domain_add(struct domainlist **domp, char *domain);
1474
1475extern void policy_save_path(char *id_path, struct map_ent *array);
1476extern int policy_check_path(struct mdinfo *disk, struct map_ent *array);
1477
1478extern void sysfs_rules_apply(char *devnm, struct mdinfo *dev);
1479extern void sysfsline(char *line);
1480
1481#if __GNUC__ < 3
1482struct stat64;
1483#endif
1484
1485extern int add_dev(const char *name, const struct stat *stb, int flag, struct FTW *s);
1486
1487extern int Manage_ro(char *devname, int fd, int readonly);
1488extern int Manage_run(char *devname, int fd, struct context *c);
1489extern int Manage_stop(char *devname, int fd, int quiet,
1490 int will_retry);
1491extern int Manage_subdevs(char *devname, int fd,
1492 struct mddev_dev *devlist, int verbose, int test,
1493 enum update_opt update, int force);
1494extern int autodetect(void);
1495extern int Grow_Add_device(char *devname, int fd, char *newdev);
1496extern int Grow_addbitmap(char *devname, int fd,
1497 struct context *c, struct shape *s);
1498extern int Grow_reshape(char *devname, int fd,
1499 struct mddev_dev *devlist,
1500 struct context *c, struct shape *s);
1501extern int Grow_restart(struct supertype *st, struct mdinfo *info,
1502 int *fdlist, int cnt, char *backup_file, int verbose);
1503extern int Grow_continue(int mdfd, struct supertype *st,
1504 struct mdinfo *info, int forked, struct context *c);
1505extern int Grow_consistency_policy(char *devname, int fd,
1506 struct context *c, struct shape *s);
1507
1508extern int restore_backup(struct supertype *st,
1509 struct mdinfo *content,
1510 int working_disks,
1511 int spares,
1512 char **backup_filep,
1513 int verbose);
1514extern int Grow_continue_command(char *devname, int fd, struct context *c);
1515
1516extern int Assemble(struct supertype *st, char *mddev,
1517 struct mddev_ident *ident,
1518 struct mddev_dev *devlist,
1519 struct context *c);
1520
1521extern int Build(struct mddev_ident *ident, struct mddev_dev *devlist, struct shape *s,
1522 struct context *c);
1523
1524extern int Create(struct supertype *st, struct mddev_ident *ident, int subdevs,
1525 struct mddev_dev *devlist, struct shape *s, struct context *c);
1526
1527extern int Detail(char *dev, struct context *c);
1528extern int Detail_Platform(struct superswitch *ss, int scan, int verbose, int export, char *controller_path);
1529extern int Query(char *dev);
1530extern int ExamineBadblocks(char *devname, int brief, struct supertype *forcest);
1531extern int Examine(struct mddev_dev *devlist, struct context *c,
1532 struct supertype *forcest);
1533extern int Monitor(struct mddev_dev *devlist,
1534 char *mailaddr, char *alert_cmd,
1535 struct context *c,
1536 int daemonise, int oneshot,
1537 int dosyslog, char *pidfile, int increments,
1538 int share);
1539
1540extern int Kill(char *dev, struct supertype *st, int force, int verbose, int noexcl);
1541extern int Kill_subarray(char *dev, char *subarray, int verbose);
1542extern int Update_subarray(char *dev, char *subarray, enum update_opt update, struct mddev_ident *ident, int quiet);
1543extern int Wait(char *dev);
1544extern int WaitClean(char *dev, int verbose);
1545extern int SetAction(char *dev, char *action);
1546
1547extern int Incremental(struct mddev_dev *devlist, struct context *c,
1548 struct supertype *st);
1549extern void RebuildMap(void);
1550extern int IncrementalScan(struct context *c, char *devnm);
1551extern int Incremental_remove(char *devname, char *path, int verbose);
1552extern int CreateBitmap(char *filename, int force, char uuid[16],
1553 unsigned long chunksize, unsigned long daemon_sleep,
1554 unsigned long write_behind,
1555 unsigned long long array_size,
1556 int major);
1557extern int ExamineBitmap(char *filename, int brief, struct supertype *st);
1558extern int IsBitmapDirty(char *filename);
1559extern int Write_rules(char *rule_name);
1560extern int bitmap_update_uuid(int fd, int *uuid, int swap);
1561
1562/* calculate the size of the bitmap given the array size and bitmap chunksize */
1563static inline unsigned long long
1564bitmap_bits(unsigned long long array_size, unsigned long chunksize)
1565{
1566 return (array_size * 512 + chunksize - 1) / chunksize;
1567}
1568
1569extern int Dump_metadata(char *dev, char *dir, struct context *c,
1570 struct supertype *st);
1571extern int Restore_metadata(char *dev, char *dir, struct context *c,
1572 struct supertype *st, int only);
1573
1574int md_array_valid(int fd);
1575int md_array_active(int fd);
1576int md_array_is_active(struct mdinfo *info);
1577int md_get_array_info(int fd, struct mdu_array_info_s *array);
1578int md_set_array_info(int fd, struct mdu_array_info_s *array);
1579int md_get_disk_info(int fd, struct mdu_disk_info_s *disk);
1580extern int get_linux_version(void);
1581extern int mdadm_version(char *version);
1582extern unsigned long long parse_size(char *size);
1583extern int parse_uuid(char *str, int uuid[4]);
1584int default_layout(struct supertype *st, int level, int verbose);
1585extern int is_near_layout_10(int layout);
1586extern int parse_layout_10(char *layout);
1587extern int parse_layout_faulty(char *layout);
1588extern int parse_num(int *dest, const char *num);
1589extern int parse_cluster_confirm_arg(char *inp, char **devname, int *slot);
1590extern int check_ext2(int fd, char *name);
1591extern int check_reiser(int fd, char *name);
1592extern int check_raid(int fd, char *name);
1593extern int check_partitions(int fd, char *dname,
1594 unsigned long long freesize,
1595 unsigned long long size);
1596extern bool is_name_posix_compatible(const char *path);
1597extern int fstat_is_blkdev(int fd, char *devname, dev_t *rdev);
1598extern int stat_is_blkdev(char *devname, dev_t *rdev);
1599
1600extern bool is_string_lq(const char * const str, size_t max_len);
1601extern bool is_dev_alive(char *path);
1602extern int get_mdp_major(void);
1603extern int get_maj_min(char *dev, int *major, int *minor);
1604extern bool is_bit_set(int *val, unsigned char index);
1605extern int dev_open(char *dev, int flags);
1606extern int open_dev(char *devnm);
1607extern void reopen_mddev(int mdfd);
1608extern int open_dev_flags(char *devnm, int flags);
1609extern int open_dev_excl(char *devnm);
1610extern int is_standard(char *dev, int *nump);
1611extern int same_dev(char *one, char *two);
1612extern int compare_paths (char* path1,char* path2);
1613extern void enable_fds(int devices);
1614extern void manage_fork_fds(int close_all);
1615extern mdadm_status_t continue_via_systemd(char *devnm, char *service_name, char *prefix);
1616
1617extern void ident_init(struct mddev_ident *ident);
1618extern mdadm_status_t ident_set_devname(struct mddev_ident *ident, const char *devname);
1619extern mdadm_status_t ident_set_name(struct mddev_ident *ident, const char *name);
1620
1621extern int parse_auto(char *str, char *msg, int config);
1622extern struct mddev_ident *conf_get_ident(char *dev);
1623extern struct mddev_dev *conf_get_devs(void);
1624extern int conf_test_dev(char *devname);
1625extern int conf_test_metadata(const char *version, struct dev_policy *pol, int is_homehost);
1626extern struct createinfo *conf_get_create_info(void);
1627extern void set_conffile(char *file);
1628extern char *conf_get_mailaddr(void);
1629extern char *conf_get_mailfrom(void);
1630extern char *conf_get_program(void);
1631extern char *conf_get_homehost(int *require_homehostp);
1632extern char *conf_get_homecluster(void);
1633extern int conf_get_monitor_delay(void);
1634extern bool conf_get_sata_opal_encryption_no_verify(void);
1635extern char *conf_line(FILE *file);
1636extern char *conf_word(FILE *file, int allow_key);
1637extern void print_quoted(char *str);
1638extern int use_udev(void);
1639extern void print_escape(char *str);
1640extern unsigned long GCD(unsigned long a, unsigned long b);
1641extern int conf_name_is_free(char *name);
1642extern bool is_devname_ignore(const char *devname);
1643extern bool is_devname_md_numbered(const char *devname);
1644extern bool is_devname_md_d_numbered(const char *devname);
1645extern int conf_verify_devnames(struct mddev_ident *array_list);
1646extern int devname_matches(char *name, char *match);
1647extern struct mddev_ident *conf_match(struct supertype *st,
1648 struct mdinfo *info,
1649 char *devname,
1650 int verbose, int *rvp);
1651
1652extern void free_line(char *line);
1653extern int match_oneof(char *devices, char *devname);
1654extern void uuid_from_super(int uuid[4], mdp_super_t *super);
1655extern const int uuid_zero[4];
1656extern int same_uuid(int a[4], int b[4], int swapuuid);
1657extern void copy_uuid(void *a, int b[4], int swapuuid);
1658extern char *__fname_from_uuid(int id[4], int swap, char *buf, char sep);
1659extern char *fname_from_uuid(struct mdinfo *info, char *buf);
1660extern unsigned long calc_csum(void *super, int bytes);
1661extern int enough(int level, int raid_disks, int layout, int clean,
1662 char *avail);
1663extern int ask(char *mesg);
1664extern unsigned long long get_component_size(int fd);
1665extern void remove_partitions(int fd);
1666extern int test_partition(int fd);
1667extern int test_partition_from_id(dev_t id);
1668extern int get_data_disks(int level, int layout, int raid_disks);
1669extern unsigned long long calc_array_size(int level, int raid_disks, int layout,
1670 int chunksize, unsigned long long devsize);
1671extern int flush_metadata_updates(struct supertype *st);
1672extern void append_metadata_update(struct supertype *st, void *buf, int len);
1673extern int assemble_container_content(struct supertype *st, int mdfd,
1674 struct mdinfo *content,
1675 struct context *c,
1676 char *chosen_name, int *result);
1677#define INCR_NO 1
1678#define INCR_UNSAFE 2
1679#define INCR_ALREADY 4
1680#define INCR_YES 8
1681
1682extern bool devid_matches_criteria(struct supertype *st, dev_t devid, struct spare_criteria *sc);
1683extern bool disk_fd_matches_criteria(struct supertype *st, int disk_fd, struct spare_criteria *sc);
1684extern struct mdinfo *container_choose_spares(struct supertype *st,
1685 struct spare_criteria *criteria,
1686 struct domainlist *domlist,
1687 char *spare_group,
1688 const char *metadata, int get_one);
1689extern int move_spare(char *from_devname, char *to_devname, dev_t devid);
1690extern int add_disk(int mdfd, struct supertype *st,
1691 struct mdinfo *sra, struct mdinfo *info);
1692extern int remove_disk(int mdfd, struct supertype *st,
1693 struct mdinfo *sra, struct mdinfo *info);
1694extern int hot_remove_disk(int mdfd, unsigned long dev, int force);
1695extern int sys_hot_remove_disk(int statefd, int force);
1696extern int set_array_info(int mdfd, struct supertype *st, struct mdinfo *info);
1697unsigned long long min_recovery_start(struct mdinfo *array);
1698
1699extern char *human_size(long long bytes);
1700extern char *human_size_brief(long long bytes, int prefix);
1701extern void print_r10_layout(int layout);
1702
1703extern void put_md_name(char *name);
1704extern char *devid2kname(dev_t devid);
1705extern char *devid2devnm(dev_t devid);
1706extern dev_t devnm2devid(char *devnm);
1707extern char *get_md_name(char *devnm);
1708
1709extern char DefaultConfFile[];
1710
1711extern int create_mddev(char *dev, char *name, int trustworthy, char *chosen, int block_udev);
1712/* values for 'trustworthy' */
1713#define LOCAL 1
1714#define LOCAL_ANY 10
1715#define FOREIGN 2
1716#define METADATA 3
1717extern int open_mddev(char *dev, int report_errors);
1718extern int is_mddev(char *dev);
1719extern int open_container(int fd);
1720extern int metadata_container_matches(char *metadata, char *devnm);
1721extern int metadata_subdev_matches(char *metadata, char *devnm);
1722extern bool is_container_member(struct mdstat_ent *ent, char *devname);
1723extern int is_subarray_active(char *subarray, char *devname);
1724extern int open_subarray(char *dev, char *subarray, struct supertype *st, int quiet);
1725extern struct superswitch *version_to_superswitch(char *vers);
1726
1727extern mdadm_status_t wait_for_mdmon_control_socket(const char *container_devnm);
1728extern int mdmon_running(const char *devnm);
1729extern int mdmon_pid(const char *devnm);
1730extern mdadm_status_t wait_for_mdmon(const char *devnm);
1731
1732extern int check_env(char *name);
1733extern __u32 random32(void);
1734extern void random_uuid(__u8 *buf);
1735extern int start_mdmon(char *devnm);
1736
1737extern int child_monitor(int afd, struct mdinfo *sra, struct reshape *reshape,
1738 struct supertype *st, unsigned long stripes,
1739 int *fds, unsigned long long *offsets,
1740 int dests, int *destfd, unsigned long long *destoffsets);
1741void abort_reshape(struct mdinfo *sra);
1742
1743void *super1_make_v0(struct supertype *st, struct mdinfo *info, mdp_super_t *sb0);
1744
1745extern char *stat2kname(struct stat *st);
1746extern char *fd2kname(int fd);
1747extern char *stat2devnm(struct stat *st);
1748bool stat_is_md_dev(struct stat *st);
1749extern char *fd2devnm(int fd);
1750
1751extern int in_initrd(void);
1752
1753struct cmap_hooks {
1754 void *cmap_handle; /* corosync lib related */
1755
1756 int (*initialize)(cmap_handle_t *handle);
1757 int (*get_string)(cmap_handle_t handle,
1758 const char *string,
1759 char **name);
1760 int (*finalize)(cmap_handle_t handle);
1761};
1762
1763extern void set_cmap_hooks(void);
1764extern void set_hooks(void);
1765
1766struct dlm_hooks {
1767 void *dlm_handle; /* dlm lib related */
1768
1769 dlm_lshandle_t (*create_lockspace)(const char *name,
1770 unsigned int mode);
1771 dlm_lshandle_t (*open_lockspace)(const char *name);
1772 int (*release_lockspace)(const char *name, dlm_lshandle_t ls,
1773 int force);
1774 int (*ls_lock)(dlm_lshandle_t lockspace, uint32_t mode,
1775 struct dlm_lksb *lksb, uint32_t flags,
1776 const void *name, unsigned int namelen,
1777 uint32_t parent, void (*astaddr) (void *astarg),
1778 void *astarg, void (*bastaddr) (void *astarg),
1779 void *range);
1780 int (*ls_unlock_wait)(dlm_lshandle_t lockspace, uint32_t lkid,
1781 uint32_t flags, struct dlm_lksb *lksb);
1782 int (*ls_get_fd)(dlm_lshandle_t ls);
1783 int (*dispatch)(int fd);
1784};
1785
1786extern int get_cluster_name(char **name);
1787extern int dlm_funs_ready(void);
1788extern int cluster_get_dlmlock(void);
1789extern int cluster_release_dlmlock(void);
1790extern void set_dlm_hooks(void);
1791
1792#define MSEC_TO_NSEC(msec) ((msec) * 1000000)
1793#define USEC_TO_NSEC(usec) ((usec) * 1000)
1794extern void sleep_for(unsigned int sec, long nsec, bool wake_after_interrupt);
1795extern bool is_directory(const char *path);
1796extern bool is_file(const char *path);
1797extern int s_gethostname(char *buf, int buf_len);
1798
1799#define _ROUND_UP(val, base) (((val) + (base) - 1) & ~(base - 1))
1800#define ROUND_UP(val, base) _ROUND_UP(val, (typeof(val))(base))
1801#define ROUND_UP_PTR(ptr, base) ((typeof(ptr)) \
1802 (ROUND_UP((unsigned long)(ptr), base)))
1803
1804static inline int is_subarray(char *vers)
1805{
1806 /* The version string for a 'subarray' (an array in a container)
1807 * is
1808 * /containername/componentname for normal read-write arrays
1809 * -containername/componentname for arrays which mdmon must not
1810 * reconfigure. They might be read-only
1811 * or might be undergoing reshape etc.
1812 * containername is e.g. md0, md_d1
1813 * componentname is dependant on the metadata. e.g. '1' 'S1' ...
1814 */
1815 return (*vers == '/' || *vers == '-');
1816}
1817
1818static inline char *to_subarray(struct mdstat_ent *ent, char *container)
1819{
1820 return &ent->metadata_version[10+strlen(container)+1];
1821}
1822
1823/**
1824 * signal_s() - Wrapper for sigaction() with signal()-like interface.
1825 * @sig: The signal to set the signal handler to.
1826 * @handler: The signal handler.
1827 *
1828 * Return: previous handler or SIG_ERR on failure.
1829 */
1830static inline sighandler_t signal_s(int sig, sighandler_t handler)
1831{
1832 struct sigaction new_act = {0};
1833 struct sigaction old_act = {0};
1834
1835 new_act.sa_handler = handler;
1836
1837 if (sigaction(sig, &new_act, &old_act) == 0)
1838 return old_act.sa_handler;
1839
1840 return SIG_ERR;
1841}
1842
1843#ifdef DEBUG
1844#include <time.h>
1845
1846#define dprintf(fmt, arg...) \
1847 do { \
1848 struct timespec ts; \
1849 clock_gettime(CLOCK_MONOTONIC, &ts); \
1850 double timestamp = ts.tv_sec + ts.tv_nsec / 1e9; \
1851 fprintf(stderr, "[%10.5f] %s: %s: " fmt, timestamp, Name, __func__, ##arg); \
1852 } while (0)
1853
1854#define dprintf_cont(fmt, arg...) \
1855 fprintf(stderr, fmt, ##arg)
1856#else
1857#define dprintf(fmt, arg...) \
1858 do { } while (0)
1859#define dprintf_cont(fmt, arg...) \
1860 do { } while (0)
1861#endif
1862
1863static inline int xasprintf(char **strp, const char *fmt, ...) {
1864 va_list ap;
1865 int ret;
1866 va_start(ap, fmt);
1867 ret = vasprintf(strp, fmt, ap);
1868 va_end(ap);
1869 assert(ret >= 0);
1870 return ret;
1871}
1872
1873#ifdef DEBUG
1874#define pr_err(fmt, args...) fprintf(stderr, "%s: %s: "fmt, Name, __func__, ##args)
1875#else
1876#define pr_err(fmt, args...) fprintf(stderr, "%s: "fmt, Name, ##args)
1877#endif
1878#define cont_err(fmt ...) fprintf(stderr, " " fmt)
1879
1880#define pr_info(fmt, args...) printf("%s: "fmt, Name, ##args)
1881
1882#define pr_vrb(fmt, arg...) ((void)(verbose && pr_err(fmt, ##arg)))
1883
1884#define LEVEL_MULTIPATH (-4)
1885#define LEVEL_LINEAR (-1)
1886#define LEVEL_FAULTY (-5)
1887
1888/* kernel module doesn't know about these */
1889#define LEVEL_CONTAINER (-100)
1890#define LEVEL_UNSUPPORTED (-200)
1891
1892/* the kernel does know about this one ... */
1893#define LEVEL_NONE (-1000000)
1894
1895/* faulty stuff */
1896
1897#define WriteTransient 0
1898#define ReadTransient 1
1899#define WritePersistent 2
1900#define ReadPersistent 3
1901#define WriteAll 4 /* doesn't go to device */
1902#define ReadFixable 5
1903#define Modes 6
1904
1905#define ClearErrors 31
1906#define ClearFaults 30
1907
1908#define AllPersist 100 /* internal use only */
1909#define NoPersist 101
1910
1911#define ModeMask 0x1f
1912#define ModeShift 5
1913
1914#ifdef __TINYC__
1915#undef minor
1916#undef major
1917#undef makedev
1918#define minor(x) ((x)&0xff)
1919#define major(x) (((x)>>8)&0xff)
1920#define makedev(M,m) (((M)<<8) | (m))
1921#endif
1922
1923enum r0layout {
1924 RAID0_DANGEROUS_LAYOUT = 0, /* layout depends on kernel version */
1925 RAID0_ORIG_LAYOUT = 1,
1926 RAID0_ALT_MULTIZONE_LAYOUT = 2,
1927};
1928
1929/* for raid4/5/6 */
1930#define ALGORITHM_LEFT_ASYMMETRIC 0
1931#define ALGORITHM_RIGHT_ASYMMETRIC 1
1932#define ALGORITHM_LEFT_SYMMETRIC 2
1933#define ALGORITHM_RIGHT_SYMMETRIC 3
1934
1935/* Define non-rotating (raid4) algorithms. These allow
1936 * conversion of raid4 to raid5.
1937 */
1938#define ALGORITHM_PARITY_0 4 /* P or P,Q are initial devices */
1939#define ALGORITHM_PARITY_N 5 /* P or P,Q are final devices. */
1940
1941/* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
1942 * Firstly, the exact positioning of the parity block is slightly
1943 * different between the 'LEFT_*' modes of md and the "_N_*" modes
1944 * of DDF.
1945 * Secondly, or order of datablocks over which the Q syndrome is computed
1946 * is different.
1947 * Consequently we have different layouts for DDF/raid6 than md/raid6.
1948 * These layouts are from the DDFv1.2 spec.
1949 * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
1950 * leaves RLQ=3 as 'Vendor Specific'
1951 */
1952
1953#define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
1954#define ALGORITHM_ROTATING_N_RESTART 9 /* DDF PRL=6 RLQ=2 */
1955#define ALGORITHM_ROTATING_N_CONTINUE 10 /*DDF PRL=6 RLQ=3 */
1956
1957/* For every RAID5 algorithm we define a RAID6 algorithm
1958 * with exactly the same layout for data and parity, and
1959 * with the Q block always on the last device (N-1).
1960 * This allows trivial conversion from RAID5 to RAID6
1961 */
1962#define ALGORITHM_LEFT_ASYMMETRIC_6 16
1963#define ALGORITHM_RIGHT_ASYMMETRIC_6 17
1964#define ALGORITHM_LEFT_SYMMETRIC_6 18
1965#define ALGORITHM_RIGHT_SYMMETRIC_6 19
1966#define ALGORITHM_PARITY_0_6 20
1967#define ALGORITHM_PARITY_N_6 ALGORITHM_PARITY_N
1968
1969/* Define PATH_MAX in case we don't use glibc or standard library does
1970 * not have PATH_MAX defined. Assume max path length is 4K characters.
1971 */
1972#ifndef PATH_MAX
1973#define PATH_MAX 4096
1974#endif
1975
1976/* The max string length necessary for decimal conversion, cannot be longer than count of bits */
1977#define INT_2_DEC_STR_MAX (sizeof(int) * 8)
1978
1979#define RESYNC_NONE -1
1980#define RESYNC_DELAYED -2
1981#define RESYNC_PENDING -3
1982#define RESYNC_REMOTE -4
1983#define RESYNC_UNKNOWN -5
1984
1985/* When using "GET_DISK_INFO" it isn't certain how high
1986 * we need to check. So we impose an absolute limit of
1987 * MAX_DISKS. This needs to be much more than the largest
1988 * number of devices any metadata can support. Currently
1989 * v1.x can support 1920
1990 */
1991#define MAX_DISKS 4096
1992
1993/* Sometimes the 'size' value passed needs to mean "Maximum".
1994 * In those cases with use MAX_SIZE
1995 */
1996#define MAX_SIZE 1
1997
1998/* We want to use unsigned numbers for sector counts, but need
1999 * a value for 'invalid'. Use '1'.
2000 */
2001#define INVALID_SECTORS 1
2002/* And another special number needed for --data_offset=variable */
2003#define VARIABLE_OFFSET 3
2004
2005/**
2006 * is_container() - check if @level is &LEVEL_CONTAINER
2007 * @level: level value
2008 *
2009 * return:
2010 * 1 if level is equal to &LEVEL_CONTAINER, 0 otherwise.
2011 */
2012static inline int is_container(const int level)
2013{
2014 if (level == LEVEL_CONTAINER)
2015 return 1;
2016 return 0;
2017}
2018
2019#define STR_COMMON_NONE "none"
2020
2021/**
2022 * str_is_none() - check if @str starts with "none".
2023 * @str: string
2024 *
2025 * return:
2026 * true if string starts with "none", false otherwise.
2027 */
2028static inline bool str_is_none(char *str)
2029{
2030 if (strncmp(str, STR_COMMON_NONE, sizeof(STR_COMMON_NONE) - 1) == 0)
2031 return true;
2032 return false;
2033}