Remove st->text_version in favour of info->text_version
[thirdparty/mdadm.git] / super-ddf.c
1 /*
2  * mdadm - manage Linux "md" devices aka RAID arrays.
3  *
4  * Copyright (C) 2006-2007 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: <neil@brown.name>
23  *
24  * Specifications for DDF takes from Common RAID DDF Specification Revision 1.2
25  * (July 28 2006).  Reused by permission of SNIA.
26  */
27
28 #define HAVE_STDINT_H 1
29 #include "mdadm.h"
30 #include "mdmon.h"
31 #include "sha1.h"
32 #include <values.h>
33
34 static inline int ROUND_UP(int a, int base)
35 {
36         return ((a+base-1)/base)*base;
37 }
38
39 /* a non-official T10 name for creation GUIDs */
40 static char T10[] = "Linux-MD";
41
42 /* DDF timestamps are 1980 based, so we need to add
43  * second-in-decade-of-seventies to convert to linux timestamps.
44  * 10 years with 2 leap years.
45  */
46 #define DECADE (3600*24*(365*10+2))
47 unsigned long crc32(
48         unsigned long crc,
49         const unsigned char *buf,
50         unsigned len);
51
52 /* The DDF metadata handling.
53  * DDF metadata lives at the end of the device.
54  * The last 512 byte block provides an 'anchor' which is used to locate
55  * the rest of the metadata which usually lives immediately behind the anchor.
56  *
57  * Note:
58  *  - all multibyte numeric fields are bigendian.
59  *  - all strings are space padded.
60  *
61  */
62
63 /* Primary Raid Level (PRL) */
64 #define DDF_RAID0       0x00
65 #define DDF_RAID1       0x01
66 #define DDF_RAID3       0x03
67 #define DDF_RAID4       0x04
68 #define DDF_RAID5       0x05
69 #define DDF_RAID1E      0x11
70 #define DDF_JBOD        0x0f
71 #define DDF_CONCAT      0x1f
72 #define DDF_RAID5E      0x15
73 #define DDF_RAID5EE     0x25
74 #define DDF_RAID6       0x16    /* Vendor unique layout */
75
76 /* Raid Level Qualifier (RLQ) */
77 #define DDF_RAID0_SIMPLE        0x00
78 #define DDF_RAID1_SIMPLE        0x00 /* just 2 devices in this plex */
79 #define DDF_RAID1_MULTI         0x01 /* exactly 3 devices in this plex */
80 #define DDF_RAID3_0             0x00 /* parity in first extent */
81 #define DDF_RAID3_N             0x01 /* parity in last extent */
82 #define DDF_RAID4_0             0x00 /* parity in first extent */
83 #define DDF_RAID4_N             0x01 /* parity in last extent */
84 /* these apply to raid5e and raid5ee as well */
85 #define DDF_RAID5_0_RESTART     0x00 /* same as 'right asymmetric' - layout 1 */
86 #define DDF_RAID5_N_RESTART     0x02 /* same as 'left asymmetric' - layout 0 */
87 #define DDF_RAID5_N_CONTINUE    0x03 /* same as 'left symmetric' - layout 2 */
88
89 #define DDF_RAID1E_ADJACENT     0x00 /* raid10 nearcopies==2 */
90 #define DDF_RAID1E_OFFSET       0x01 /* raid10 offsetcopies==2 */
91
92 /* Secondary RAID Level (SRL) */
93 #define DDF_2STRIPED    0x00    /* This is weirder than RAID0 !! */
94 #define DDF_2MIRRORED   0x01
95 #define DDF_2CONCAT     0x02
96 #define DDF_2SPANNED    0x03    /* This is also weird - be careful */
97
98 /* Magic numbers */
99 #define DDF_HEADER_MAGIC        __cpu_to_be32(0xDE11DE11)
100 #define DDF_CONTROLLER_MAGIC    __cpu_to_be32(0xAD111111)
101 #define DDF_PHYS_RECORDS_MAGIC  __cpu_to_be32(0x22222222)
102 #define DDF_PHYS_DATA_MAGIC     __cpu_to_be32(0x33333333)
103 #define DDF_VIRT_RECORDS_MAGIC  __cpu_to_be32(0xDDDDDDDD)
104 #define DDF_VD_CONF_MAGIC       __cpu_to_be32(0xEEEEEEEE)
105 #define DDF_SPARE_ASSIGN_MAGIC  __cpu_to_be32(0x55555555)
106 #define DDF_VU_CONF_MAGIC       __cpu_to_be32(0x88888888)
107 #define DDF_VENDOR_LOG_MAGIC    __cpu_to_be32(0x01dBEEF0)
108 #define DDF_BBM_LOG_MAGIC       __cpu_to_be32(0xABADB10C)
109
110 #define DDF_GUID_LEN    24
111 #define DDF_REVISION    "01.00.00"
112
113 struct ddf_header {
114         __u32   magic;
115         __u32   crc;
116         char    guid[DDF_GUID_LEN];
117         char    revision[8];    /* 01.00.00 */
118         __u32   seq;            /* starts at '1' */
119         __u32   timestamp;
120         __u8    openflag;
121         __u8    foreignflag;
122         __u8    enforcegroups;
123         __u8    pad0;           /* 0xff */
124         __u8    pad1[12];       /* 12 * 0xff */
125         /* 64 bytes so far */
126         __u8    header_ext[32]; /* reserved: fill with 0xff */
127         __u64   primary_lba;
128         __u64   secondary_lba;
129         __u8    type;
130         __u8    pad2[3];        /* 0xff */
131         __u32   workspace_len;  /* sectors for vendor space -
132                                  * at least 32768(sectors) */
133         __u64   workspace_lba;
134         __u16   max_pd_entries; /* one of 15, 63, 255, 1023, 4095 */
135         __u16   max_vd_entries; /* 2^(4,6,8,10,12)-1 : i.e. as above */
136         __u16   max_partitions; /* i.e. max num of configuration
137                                    record entries per disk */
138         __u16   config_record_len; /* 1 +ROUNDUP(max_primary_element_entries
139                                                  *12/512) */
140         __u16   max_primary_element_entries; /* 16, 64, 256, 1024, or 4096 */
141         __u8    pad3[54];       /* 0xff */
142         /* 192 bytes so far */
143         __u32   controller_section_offset;
144         __u32   controller_section_length;
145         __u32   phys_section_offset;
146         __u32   phys_section_length;
147         __u32   virt_section_offset;
148         __u32   virt_section_length;
149         __u32   config_section_offset;
150         __u32   config_section_length;
151         __u32   data_section_offset;
152         __u32   data_section_length;
153         __u32   bbm_section_offset;
154         __u32   bbm_section_length;
155         __u32   diag_space_offset;
156         __u32   diag_space_length;
157         __u32   vendor_offset;
158         __u32   vendor_length;
159         /* 256 bytes so far */
160         __u8    pad4[256];      /* 0xff */
161 };
162
163 /* type field */
164 #define DDF_HEADER_ANCHOR       0x00
165 #define DDF_HEADER_PRIMARY      0x01
166 #define DDF_HEADER_SECONDARY    0x02
167
168 /* The content of the 'controller section' - global scope */
169 struct ddf_controller_data {
170         __u32   magic;
171         __u32   crc;
172         char    guid[DDF_GUID_LEN];
173         struct controller_type {
174                 __u16 vendor_id;
175                 __u16 device_id;
176                 __u16 sub_vendor_id;
177                 __u16 sub_device_id;
178         } type;
179         char    product_id[16];
180         __u8    pad[8]; /* 0xff */
181         __u8    vendor_data[448];
182 };
183
184 /* The content of phys_section - global scope */
185 struct phys_disk {
186         __u32   magic;
187         __u32   crc;
188         __u16   used_pdes;
189         __u16   max_pdes;
190         __u8    pad[52];
191         struct phys_disk_entry {
192                 char    guid[DDF_GUID_LEN];
193                 __u32   refnum;
194                 __u16   type;
195                 __u16   state;
196                 __u64   config_size; /* DDF structures must be after here */
197                 char    path[18];       /* another horrible structure really */
198                 __u8    pad[6];
199         } entries[0];
200 };
201
202 /* phys_disk_entry.type is a bitmap - bigendian remember */
203 #define DDF_Forced_PD_GUID              1
204 #define DDF_Active_in_VD                2
205 #define DDF_Global_Spare                4
206 #define DDF_Spare                       8 /* overrides Global_spare */
207 #define DDF_Foreign                     16
208 #define DDF_Legacy                      32 /* no DDF on this device */
209
210 #define DDF_Interface_mask              0xf00
211 #define DDF_Interface_SCSI              0x100
212 #define DDF_Interface_SAS               0x200
213 #define DDF_Interface_SATA              0x300
214 #define DDF_Interface_FC                0x400
215
216 /* phys_disk_entry.state is a bigendian bitmap */
217 #define DDF_Online                      1
218 #define DDF_Failed                      2 /* overrides  1,4,8 */
219 #define DDF_Rebuilding                  4
220 #define DDF_Transition                  8
221 #define DDF_SMART                       16
222 #define DDF_ReadErrors                  32
223 #define DDF_Missing                     64
224
225 /* The content of the virt_section global scope */
226 struct virtual_disk {
227         __u32   magic;
228         __u32   crc;
229         __u16   populated_vdes;
230         __u16   max_vdes;
231         __u8    pad[52];
232         struct virtual_entry {
233                 char    guid[DDF_GUID_LEN];
234                 __u16   unit;
235                 __u16   pad0;   /* 0xffff */
236                 __u16   guid_crc;
237                 __u16   type;
238                 __u8    state;
239                 __u8    init_state;
240                 __u8    pad1[14];
241                 char    name[16];
242         } entries[0];
243 };
244
245 /* virtual_entry.type is a bitmap - bigendian */
246 #define DDF_Shared              1
247 #define DDF_Enforce_Groups      2
248 #define DDF_Unicode             4
249 #define DDF_Owner_Valid         8
250
251 /* virtual_entry.state is a bigendian bitmap */
252 #define DDF_state_mask          0x7
253 #define DDF_state_optimal       0x0
254 #define DDF_state_degraded      0x1
255 #define DDF_state_deleted       0x2
256 #define DDF_state_missing       0x3
257 #define DDF_state_failed        0x4
258 #define DDF_state_part_optimal  0x5
259
260 #define DDF_state_morphing      0x8
261 #define DDF_state_inconsistent  0x10
262
263 /* virtual_entry.init_state is a bigendian bitmap */
264 #define DDF_initstate_mask      0x03
265 #define DDF_init_not            0x00
266 #define DDF_init_quick          0x01 /* initialisation is progress.
267                                       * i.e. 'state_inconsistent' */
268 #define DDF_init_full           0x02
269
270 #define DDF_access_mask         0xc0
271 #define DDF_access_rw           0x00
272 #define DDF_access_ro           0x80
273 #define DDF_access_blocked      0xc0
274
275 /* The content of the config_section - local scope
276  * It has multiple records each config_record_len sectors
277  * They can be vd_config or spare_assign
278  */
279
280 struct vd_config {
281         __u32   magic;
282         __u32   crc;
283         char    guid[DDF_GUID_LEN];
284         __u32   timestamp;
285         __u32   seqnum;
286         __u8    pad0[24];
287         __u16   prim_elmnt_count;
288         __u8    chunk_shift;    /* 0 == 512, 1==1024 etc */
289         __u8    prl;
290         __u8    rlq;
291         __u8    sec_elmnt_count;
292         __u8    sec_elmnt_seq;
293         __u8    srl;
294         __u64   blocks;         /* blocks per component could be different
295                                  * on different component devices...(only
296                                  * for concat I hope) */
297         __u64   array_blocks;   /* blocks in array */
298         __u8    pad1[8];
299         __u32   spare_refs[8];
300         __u8    cache_pol[8];
301         __u8    bg_rate;
302         __u8    pad2[3];
303         __u8    pad3[52];
304         __u8    pad4[192];
305         __u8    v0[32]; /* reserved- 0xff */
306         __u8    v1[32]; /* reserved- 0xff */
307         __u8    v2[16]; /* reserved- 0xff */
308         __u8    v3[16]; /* reserved- 0xff */
309         __u8    vendor[32];
310         __u32   phys_refnum[0]; /* refnum of each disk in sequence */
311       /*__u64   lba_offset[0];  LBA offset in each phys.  Note extents in a
312                                 bvd are always the same size */
313 };
314
315 /* vd_config.cache_pol[7] is a bitmap */
316 #define DDF_cache_writeback     1       /* else writethrough */
317 #define DDF_cache_wadaptive     2       /* only applies if writeback */
318 #define DDF_cache_readahead     4
319 #define DDF_cache_radaptive     8       /* only if doing read-ahead */
320 #define DDF_cache_ifnobatt      16      /* even to write cache if battery is poor */
321 #define DDF_cache_wallowed      32      /* enable write caching */
322 #define DDF_cache_rallowed      64      /* enable read caching */
323
324 struct spare_assign {
325         __u32   magic;
326         __u32   crc;
327         __u32   timestamp;
328         __u8    reserved[7];
329         __u8    type;
330         __u16   populated;      /* SAEs used */
331         __u16   max;            /* max SAEs */
332         __u8    pad[8];
333         struct spare_assign_entry {
334                 char    guid[DDF_GUID_LEN];
335                 __u16   secondary_element;
336                 __u8    pad[6];
337         } spare_ents[0];
338 };
339 /* spare_assign.type is a bitmap */
340 #define DDF_spare_dedicated     0x1     /* else global */
341 #define DDF_spare_revertible    0x2     /* else committable */
342 #define DDF_spare_active        0x4     /* else not active */
343 #define DDF_spare_affinity      0x8     /* enclosure affinity */
344
345 /* The data_section contents - local scope */
346 struct disk_data {
347         __u32   magic;
348         __u32   crc;
349         char    guid[DDF_GUID_LEN];
350         __u32   refnum;         /* crc of some magic drive data ... */
351         __u8    forced_ref;     /* set when above was not result of magic */
352         __u8    forced_guid;    /* set if guid was forced rather than magic */
353         __u8    vendor[32];
354         __u8    pad[442];
355 };
356
357 /* bbm_section content */
358 struct bad_block_log {
359         __u32   magic;
360         __u32   crc;
361         __u16   entry_count;
362         __u32   spare_count;
363         __u8    pad[10];
364         __u64   first_spare;
365         struct mapped_block {
366                 __u64   defective_start;
367                 __u32   replacement_start;
368                 __u16   remap_count;
369                 __u8    pad[2];
370         } entries[0];
371 };
372
373 /* Struct for internally holding ddf structures */
374 /* The DDF structure stored on each device is potentially
375  * quite different, as some data is global and some is local.
376  * The global data is:
377  *   - ddf header
378  *   - controller_data
379  *   - Physical disk records
380  *   - Virtual disk records
381  * The local data is:
382  *   - Configuration records
383  *   - Physical Disk data section
384  *  (  and Bad block and vendor which I don't care about yet).
385  *
386  * The local data is parsed into separate lists as it is read
387  * and reconstructed for writing.  This means that we only need
388  * to make config changes once and they are automatically
389  * propagated to all devices.
390  * Note that the ddf_super has space of the conf and disk data
391  * for this disk and also for a list of all such data.
392  * The list is only used for the superblock that is being
393  * built in Create or Assemble to describe the whole array.
394  */
395 struct ddf_super {
396         struct ddf_header anchor, primary, secondary, *active;
397         struct ddf_controller_data controller;
398         struct phys_disk        *phys;
399         struct virtual_disk     *virt;
400         int pdsize, vdsize;
401         int max_part;
402         struct vcl {
403                 struct vcl      *next;
404                 __u64           *lba_offset; /* location in 'conf' of
405                                               * the lba table */
406                 struct vd_config conf;
407         } *conflist, *newconf;
408         struct dl {
409                 struct dl       *next;
410                 struct disk_data disk;
411                 int major, minor;
412                 char *devname;
413                 int fd;
414                 struct vcl *vlist[0]; /* max_part+1 in size */
415         } *dlist;
416 };
417
418 #ifndef offsetof
419 #define offsetof(t,f) ((size_t)&(((t*)0)->f))
420 #endif
421
422 extern struct superswitch super_ddf_container, super_ddf_bvd, super_ddf;
423
424 static int calc_crc(void *buf, int len)
425 {
426         /* crcs are always at the same place as in the ddf_header */
427         struct ddf_header *ddf = buf;
428         __u32 oldcrc = ddf->crc;
429         __u32 newcrc;
430         ddf->crc = 0xffffffff;
431
432         newcrc = crc32(0, buf, len);
433         ddf->crc = oldcrc;
434         return newcrc;
435 }
436
437 static int load_ddf_header(int fd, unsigned long long lba,
438                            unsigned long long size,
439                            int type,
440                            struct ddf_header *hdr, struct ddf_header *anchor)
441 {
442         /* read a ddf header (primary or secondary) from fd/lba
443          * and check that it is consistent with anchor
444          * Need to check:
445          *   magic, crc, guid, rev, and LBA's header_type, and
446          *  everything after header_type must be the same
447          */
448         if (lba >= size-1)
449                 return 0;
450
451         if (lseek64(fd, lba<<9, 0) < 0)
452                 return 0;
453
454         if (read(fd, hdr, 512) != 512)
455                 return 0;
456
457         if (hdr->magic != DDF_HEADER_MAGIC)
458                 return 0;
459         if (calc_crc(hdr, 512) != hdr->crc)
460                 return 0;
461         if (memcmp(anchor->guid, hdr->guid, DDF_GUID_LEN) != 0 ||
462             memcmp(anchor->revision, hdr->revision, 8) != 0 ||
463             anchor->primary_lba != hdr->primary_lba ||
464             anchor->secondary_lba != hdr->secondary_lba ||
465             hdr->type != type ||
466             memcmp(anchor->pad2, hdr->pad2, 512 -
467                    offsetof(struct ddf_header, pad2)) != 0)
468                 return 0;
469
470         /* Looks good enough to me... */
471         return 1;
472 }
473
474 static void *load_section(int fd, struct ddf_super *super, void *buf,
475                           __u32 offset_be, __u32 len_be, int check)
476 {
477         unsigned long long offset = __be32_to_cpu(offset_be);
478         unsigned long long len = __be32_to_cpu(len_be);
479         int dofree = (buf == NULL);
480
481         if (check)
482                 if (len != 2 && len != 8 && len != 32
483                     && len != 128 && len != 512)
484                         return NULL;
485
486         if (len > 1024)
487                 return NULL;
488         if (buf) {
489                 /* All pre-allocated sections are a single block */
490                 if (len != 1)
491                         return NULL;
492         } else
493                 buf = malloc(len<<9);
494         if (!buf)
495                 return NULL;
496
497         if (super->active->type == 1)
498                 offset += __be64_to_cpu(super->active->primary_lba);
499         else
500                 offset += __be64_to_cpu(super->active->secondary_lba);
501
502         if (lseek64(fd, offset<<9, 0) != (offset<<9)) {
503                 if (dofree)
504                         free(buf);
505                 return NULL;
506         }
507         if (read(fd, buf, len<<9) != (len<<9)) {
508                 if (dofree)
509                         free(buf);
510                 return NULL;
511         }
512         return buf;
513 }
514
515 static int load_ddf_headers(int fd, struct ddf_super *super, char *devname)
516 {
517         unsigned long long dsize;
518
519         get_dev_size(fd, NULL, &dsize);
520
521         if (lseek64(fd, dsize-512, 0) < 0) {
522                 if (devname)
523                         fprintf(stderr,
524                                 Name": Cannot seek to anchor block on %s: %s\n",
525                                 devname, strerror(errno));
526                 return 1;
527         }
528         if (read(fd, &super->anchor, 512) != 512) {
529                 if (devname)
530                         fprintf(stderr,
531                                 Name ": Cannot read anchor block on %s: %s\n",
532                                 devname, strerror(errno));
533                 return 1;
534         }
535         if (super->anchor.magic != DDF_HEADER_MAGIC) {
536                 if (devname)
537                         fprintf(stderr, Name ": no DDF anchor found on %s\n",
538                                 devname);
539                 return 2;
540         }
541         if (calc_crc(&super->anchor, 512) != super->anchor.crc) {
542                 if (devname)
543                         fprintf(stderr, Name ": bad CRC on anchor on %s\n",
544                                 devname);
545                 return 2;
546         }
547         if (memcmp(super->anchor.revision, DDF_REVISION, 8) != 0) {
548                 if (devname)
549                         fprintf(stderr, Name ": can only support super revision"
550                                 " %.8s, not %.8s on %s\n",
551                                 DDF_REVISION, super->anchor.revision, devname);
552                 return 2;
553         }
554         if (load_ddf_header(fd, __be64_to_cpu(super->anchor.primary_lba),
555                             dsize >> 9,  1,
556                             &super->primary, &super->anchor) == 0) {
557                 if (devname)
558                         fprintf(stderr,
559                                 Name ": Failed to load primary DDF header "
560                                 "on %s\n", devname);
561                 return 2;
562         }
563         super->active = &super->primary;
564         if (load_ddf_header(fd, __be64_to_cpu(super->anchor.secondary_lba),
565                             dsize >> 9,  2,
566                             &super->secondary, &super->anchor)) {
567                 if ((__be32_to_cpu(super->primary.seq)
568                      < __be32_to_cpu(super->secondary.seq) &&
569                      !super->secondary.openflag)
570                     || (__be32_to_cpu(super->primary.seq)
571                         == __be32_to_cpu(super->secondary.seq) &&
572                         super->primary.openflag && !super->secondary.openflag)
573                         )
574                         super->active = &super->secondary;
575         }
576         return 0;
577 }
578
579 static int load_ddf_global(int fd, struct ddf_super *super, char *devname)
580 {
581         void *ok;
582         ok = load_section(fd, super, &super->controller,
583                           super->active->controller_section_offset,
584                           super->active->controller_section_length,
585                           0);
586         super->phys = load_section(fd, super, NULL,
587                                    super->active->phys_section_offset,
588                                    super->active->phys_section_length,
589                                    1);
590         super->pdsize = __be32_to_cpu(super->active->phys_section_length) * 512;
591
592         super->virt = load_section(fd, super, NULL,
593                                    super->active->virt_section_offset,
594                                    super->active->virt_section_length,
595                                    1);
596         super->vdsize = __be32_to_cpu(super->active->virt_section_length) * 512;
597         if (!ok ||
598             !super->phys ||
599             !super->virt) {
600                 free(super->phys);
601                 free(super->virt);
602                 super->phys = NULL;
603                 super->virt = NULL;
604                 return 2;
605         }
606         super->conflist = NULL;
607         super->dlist = NULL;
608         return 0;
609 }
610
611 static int load_ddf_local(int fd, struct ddf_super *super,
612                           char *devname, int keep)
613 {
614         struct dl *dl;
615         struct stat stb;
616         char *conf;
617         int i;
618         int conflen;
619         int mppe;
620
621         /* First the local disk info */
622         super->max_part = __be16_to_cpu(super->active->max_partitions);
623         dl = malloc(sizeof(*dl) +
624                     (super->max_part+1) * sizeof(dl->vlist[0]));
625
626         load_section(fd, super, &dl->disk,
627                      super->active->data_section_offset,
628                      super->active->data_section_length,
629                      0);
630         dl->devname = devname ? strdup(devname) : NULL;
631
632         fstat(fd, &stb);
633         dl->major = major(stb.st_rdev);
634         dl->minor = minor(stb.st_rdev);
635         dl->next = super->dlist;
636         dl->fd = keep ? fd : -1;
637         for (i=0 ; i < super->max_part + 1 ; i++)
638                 dl->vlist[i] = NULL;
639         super->dlist = dl;
640
641         /* Now the config list. */
642         /* 'conf' is an array of config entries, some of which are
643          * probably invalid.  Those which are good need to be copied into
644          * the conflist
645          */
646         conflen =  __be16_to_cpu(super->active->config_record_len);
647
648         conf = load_section(fd, super, NULL,
649                             super->active->config_section_offset,
650                             super->active->config_section_length,
651                             0);
652
653         for (i = 0;
654              i < __be32_to_cpu(super->active->config_section_length);
655              i += conflen) {
656                 struct vd_config *vd =
657                         (struct vd_config *)((char*)conf + i*512);
658                 struct vcl *vcl;
659
660                 if (vd->magic != DDF_VD_CONF_MAGIC)
661                         continue;
662                 for (vcl = super->conflist; vcl; vcl = vcl->next) {
663                         if (memcmp(vcl->conf.guid,
664                                    vd->guid, DDF_GUID_LEN) == 0)
665                                 break;
666                 }
667
668                 if (vcl) {
669                         dl->vlist[i/conflen] = vcl;
670                         if (__be32_to_cpu(vd->seqnum) <=
671                             __be32_to_cpu(vcl->conf.seqnum))
672                                 continue;
673                 } else {
674                         vcl = malloc(conflen*512 + offsetof(struct vcl, conf));
675                         vcl->next = super->conflist;
676                         super->conflist = vcl;
677                 }
678                 memcpy(&vcl->conf, vd, conflen*512);
679                 mppe = __be16_to_cpu(super->anchor.max_primary_element_entries);
680                 vcl->lba_offset = (__u64*)
681                         &vcl->conf.phys_refnum[mppe];
682                 dl->vlist[i/conflen] = vcl;
683         }
684         free(conf);
685
686         return 0;
687 }
688
689 #ifndef MDASSEMBLE
690 static int load_super_ddf_all(struct supertype *st, int fd,
691                               void **sbp, char *devname, int keep_fd);
692 #endif
693 static int load_super_ddf(struct supertype *st, int fd,
694                           char *devname)
695 {
696         unsigned long long dsize;
697         struct ddf_super *super;
698         int rv;
699
700 #ifndef MDASSEMBLE
701         if (load_super_ddf_all(st, fd, &st->sb, devname, 1) == 0)
702                 return 0;
703 #endif
704
705         if (get_dev_size(fd, devname, &dsize) == 0)
706                 return 1;
707
708         /* 32M is a lower bound */
709         if (dsize <= 32*1024*1024) {
710                 if (devname) {
711                         fprintf(stderr,
712                                 Name ": %s is too small for ddf: "
713                                 "size is %llu sectors.\n",
714                                 devname, dsize>>9);
715                         return 1;
716                 }
717         }
718         if (dsize & 511) {
719                 if (devname) {
720                         fprintf(stderr,
721                                 Name ": %s is an odd size for ddf: "
722                                 "size is %llu bytes.\n",
723                                 devname, dsize);
724                         return 1;
725                 }
726         }
727
728         super = malloc(sizeof(*super));
729         if (!super) {
730                 fprintf(stderr, Name ": malloc of %zu failed.\n",
731                         sizeof(*super));
732                 return 1;
733         }
734         memset(super, 0, sizeof(*super));
735
736         rv = load_ddf_headers(fd, super, devname);
737         if (rv) {
738                 free(super);
739                 return rv;
740         }
741
742         /* Have valid headers and have chosen the best. Let's read in the rest*/
743
744         rv = load_ddf_global(fd, super, devname);
745
746         if (rv) {
747                 if (devname)
748                         fprintf(stderr,
749                                 Name ": Failed to load all information "
750                                 "sections on %s\n", devname);
751                 free(super);
752                 return rv;
753         }
754
755         load_ddf_local(fd, super, devname, 0);
756
757         /* Should possibly check the sections .... */
758
759         st->sb = super;
760         if (st->ss == NULL) {
761                 st->ss = &super_ddf;
762                 st->minor_version = 0;
763                 st->max_devs = 512;
764         }
765         return 0;
766
767 }
768
769 static void free_super_ddf(struct supertype *st)
770 {
771         struct ddf_super *ddf = st->sb;
772         if (ddf == NULL)
773                 return;
774         free(ddf->phys);
775         free(ddf->virt);
776         while (ddf->conflist) {
777                 struct vcl *v = ddf->conflist;
778                 ddf->conflist = v->next;
779                 free(v);
780         }
781         while (ddf->dlist) {
782                 struct dl *d = ddf->dlist;
783                 ddf->dlist = d->next;
784                 if (d->fd >= 0)
785                         close(d->fd);
786                 free(d);
787         }
788         free(ddf);
789         st->sb = NULL;
790 }
791
792 static struct supertype *match_metadata_desc_ddf(char *arg)
793 {
794         /* 'ddf' only support containers */
795         struct supertype *st;
796         if (strcmp(arg, "ddf") != 0 &&
797             strcmp(arg, "default") != 0
798                 )
799                 return NULL;
800
801         st = malloc(sizeof(*st));
802         st->ss = &super_ddf;
803         st->max_devs = 512;
804         st->minor_version = 0;
805         st->sb = NULL;
806         return st;
807 }
808
809 static struct supertype *match_metadata_desc_ddf_bvd(char *arg)
810 {
811         struct supertype *st;
812         if (strcmp(arg, "ddf/bvd") != 0 &&
813             strcmp(arg, "bvd") != 0 &&
814             strcmp(arg, "default") != 0
815                 )
816                 return NULL;
817
818         st = malloc(sizeof(*st));
819         st->ss = &super_ddf_bvd;
820         st->max_devs = 512;
821         st->minor_version = 0;
822         st->sb = NULL;
823         return st;
824 }
825 static struct supertype *match_metadata_desc_ddf_svd(char *arg)
826 {
827         struct supertype *st;
828         if (strcmp(arg, "ddf/svd") != 0 &&
829             strcmp(arg, "svd") != 0 &&
830             strcmp(arg, "default") != 0
831                 )
832                 return NULL;
833
834         st = malloc(sizeof(*st));
835         st->ss = &super_ddf_svd;
836         st->max_devs = 512;
837         st->minor_version = 0;
838         st->sb = NULL;
839         return st;
840 }
841
842 #ifndef MDASSEMBLE
843
844 static mapping_t ddf_state[] = {
845         { "Optimal", 0},
846         { "Degraded", 1},
847         { "Deleted", 2},
848         { "Missing", 3},
849         { "Failed", 4},
850         { "Partially Optimal", 5},
851         { "-reserved-", 6},
852         { "-reserved-", 7},
853         { NULL, 0}
854 };
855
856 static mapping_t ddf_init_state[] = {
857         { "Not Initialised", 0},
858         { "QuickInit in Progress", 1},
859         { "Fully Initialised", 2},
860         { "*UNKNOWN*", 3},
861         { NULL, 0}
862 };
863 static mapping_t ddf_access[] = {
864         { "Read/Write", 0},
865         { "Reserved", 1},
866         { "Read Only", 2},
867         { "Blocked (no access)", 3},
868         { NULL ,0}
869 };
870
871 static mapping_t ddf_level[] = {
872         { "RAID0", DDF_RAID0},
873         { "RAID1", DDF_RAID1},
874         { "RAID3", DDF_RAID3},
875         { "RAID4", DDF_RAID4},
876         { "RAID5", DDF_RAID5},
877         { "RAID1E",DDF_RAID1E},
878         { "JBOD",  DDF_JBOD},
879         { "CONCAT",DDF_CONCAT},
880         { "RAID5E",DDF_RAID5E},
881         { "RAID5EE",DDF_RAID5EE},
882         { "RAID6", DDF_RAID6},
883         { NULL, 0}
884 };
885 static mapping_t ddf_sec_level[] = {
886         { "Striped", DDF_2STRIPED},
887         { "Mirrored", DDF_2MIRRORED},
888         { "Concat", DDF_2CONCAT},
889         { "Spanned", DDF_2SPANNED},
890         { NULL, 0}
891 };
892 #endif
893
894 struct num_mapping {
895         int num1, num2;
896 };
897 static struct num_mapping ddf_level_num[] = {
898         { DDF_RAID0, 0 },
899         { DDF_RAID1, 1 },
900         { DDF_RAID3, LEVEL_UNSUPPORTED },
901         { DDF_RAID5, 4 },
902         { DDF_RAID1E, LEVEL_UNSUPPORTED },
903         { DDF_JBOD, LEVEL_UNSUPPORTED },
904         { DDF_CONCAT, LEVEL_LINEAR },
905         { DDF_RAID5E, LEVEL_UNSUPPORTED },
906         { DDF_RAID5EE, LEVEL_UNSUPPORTED },
907         { DDF_RAID6, 6},
908         { MAXINT, MAXINT }
909 };
910
911 static int map_num1(struct num_mapping *map, int num)
912 {
913         int i;
914         for (i=0 ; map[i].num1 != MAXINT; i++)
915                 if (map[i].num1 == num)
916                         break;
917         return map[i].num2;
918 }
919
920 #ifndef MDASSEMBLE
921 static void print_guid(char *guid, int tstamp)
922 {
923         /* A GUIDs are part (or all) ASCII and part binary.
924          * They tend to be space padded.
925          * We ignore trailing spaces and print numbers
926          * <0x20 and >=0x7f as \xXX
927          * Some GUIDs have a time stamp in bytes 16-19.
928          * We print that if appropriate
929          */
930         int l = DDF_GUID_LEN;
931         int i;
932         while (l && guid[l-1] == ' ')
933                 l--;
934         for (i=0 ; i<l ; i++) {
935                 if (guid[i] >= 0x20 && guid[i] < 0x7f)
936                         fputc(guid[i], stdout);
937                 else
938                         fprintf(stdout, "\\x%02x", guid[i]&255);
939         }
940         if (tstamp) {
941                 time_t then = __be32_to_cpu(*(__u32*)(guid+16)) + DECADE;
942                 char tbuf[100];
943                 struct tm *tm;
944                 tm = localtime(&then);
945                 strftime(tbuf, 100, " (%D %T)",tm);
946                 fputs(tbuf, stdout);
947         }
948 }
949
950 static void examine_vd(int n, struct ddf_super *sb, char *guid)
951 {
952         int crl = __be16_to_cpu(sb->anchor.config_record_len);
953         struct vcl *vcl;
954
955         for (vcl = sb->conflist ; vcl ; vcl = vcl->next) {
956                 struct vd_config *vc = &vcl->conf;
957
958                 if (calc_crc(vc, crl*512) != vc->crc)
959                         continue;
960                 if (memcmp(vc->guid, guid, DDF_GUID_LEN) != 0)
961                         continue;
962
963                 /* Ok, we know about this VD, let's give more details */
964                 printf(" Raid Devices[%d] : %d\n", n,
965                        __be16_to_cpu(vc->prim_elmnt_count));
966                 printf("   Chunk Size[%d] : %d sectors\n", n,
967                        1 << vc->chunk_shift);
968                 printf("   Raid Level[%d] : %s\n", n,
969                        map_num(ddf_level, vc->prl)?:"-unknown-");
970                 if (vc->sec_elmnt_count != 1) {
971                         printf("  Secondary Position[%d] : %d of %d\n", n,
972                                vc->sec_elmnt_seq, vc->sec_elmnt_count);
973                         printf("  Secondary Level[%d] : %s\n", n,
974                                map_num(ddf_sec_level, vc->srl) ?: "-unknown-");
975                 }
976                 printf("  Device Size[%d] : %llu\n", n,
977                        __be64_to_cpu(vc->blocks)/2);
978                 printf("   Array Size[%d] : %llu\n", n,
979                        __be64_to_cpu(vc->array_blocks)/2);
980         }
981 }
982
983 static void examine_vds(struct ddf_super *sb)
984 {
985         int cnt = __be16_to_cpu(sb->virt->populated_vdes);
986         int i;
987         printf("  Virtual Disks : %d\n", cnt);
988
989         for (i=0; i<cnt; i++) {
990                 struct virtual_entry *ve = &sb->virt->entries[i];
991                 printf("      VD GUID[%d] : ", i); print_guid(ve->guid, 1);
992                 printf("\n");
993                 printf("         unit[%d] : %d\n", i, __be16_to_cpu(ve->unit));
994                 printf("        state[%d] : %s, %s%s\n", i,
995                        map_num(ddf_state, ve->state & 7),
996                        (ve->state & 8) ? "Morphing, ": "",
997                        (ve->state & 16)? "Not Consistent" : "Consistent");
998                 printf("   init state[%d] : %s\n", i,
999                        map_num(ddf_init_state, ve->init_state&3));
1000                 printf("       access[%d] : %s\n", i,
1001                        map_num(ddf_access, (ve->init_state>>6) & 3));
1002                 printf("         Name[%d] : %.16s\n", i, ve->name);
1003                 examine_vd(i, sb, ve->guid);
1004         }
1005         if (cnt) printf("\n");
1006 }
1007
1008 static void examine_pds(struct ddf_super *sb)
1009 {
1010         int cnt = __be16_to_cpu(sb->phys->used_pdes);
1011         int i;
1012         struct dl *dl;
1013         printf(" Physical Disks : %d\n", cnt);
1014
1015         for (i=0 ; i<cnt ; i++) {
1016                 struct phys_disk_entry *pd = &sb->phys->entries[i];
1017                 int type = __be16_to_cpu(pd->type);
1018                 int state = __be16_to_cpu(pd->state);
1019
1020                 printf("      PD GUID[%d] : ", i); print_guid(pd->guid, 0);
1021                 printf("\n");
1022                 printf("          ref[%d] : %08x\n", i,
1023                        __be32_to_cpu(pd->refnum));
1024                 printf("         mode[%d] : %s%s%s%s%s\n", i,
1025                        (type&2) ? "active":"",
1026                        (type&4) ? "Global Spare":"",
1027                        (type&8) ? "spare" : "",
1028                        (type&16)? ", foreign" : "",
1029                        (type&32)? "pass-through" : "");
1030                 printf("        state[%d] : %s%s%s%s%s%s%s\n", i,
1031                        (state&1)? "Online": "Offline",
1032                        (state&2)? ", Failed": "",
1033                        (state&4)? ", Rebuilding": "",
1034                        (state&8)? ", in-transition": "",
1035                        (state&16)? ", SMART errors": "",
1036                        (state&32)? ", Unrecovered Read Errors": "",
1037                        (state&64)? ", Missing" : "");
1038                 printf("   Avail Size[%d] : %llu K\n", i,
1039                        __be64_to_cpu(pd->config_size)>>1);
1040                 for (dl = sb->dlist; dl ; dl = dl->next) {
1041                         if (dl->disk.refnum == pd->refnum) {
1042                                 char *dv = map_dev(dl->major, dl->minor, 0);
1043                                 if (dv)
1044                                         printf("       Device[%d] : %s\n",
1045                                                i, dv);
1046                         }
1047                 }
1048                 printf("\n");
1049         }
1050 }
1051
1052 static void examine_super_ddf(struct supertype *st, char *homehost)
1053 {
1054         struct ddf_super *sb = st->sb;
1055
1056         printf("          Magic : %08x\n", __be32_to_cpu(sb->anchor.magic));
1057         printf("        Version : %.8s\n", sb->anchor.revision);
1058         printf("Controller GUID : "); print_guid(sb->controller.guid, 0);
1059         printf("\n");
1060         printf(" Container GUID : "); print_guid(sb->anchor.guid, 1);
1061         printf("\n");
1062         printf("            Seq : %08x\n", __be32_to_cpu(sb->active->seq));
1063         printf("  Redundant hdr : %s\n", sb->secondary.magic == DDF_HEADER_MAGIC
1064                ?"yes" : "no");
1065         examine_vds(sb);
1066         examine_pds(sb);
1067 }
1068
1069 static void brief_examine_super_ddf(struct supertype *st)
1070 {
1071         /* We just write a generic DDF ARRAY entry
1072          * The uuid is all hex, 6 groups of 4 bytes
1073          */
1074         struct ddf_super *ddf = st->sb;
1075         int i;
1076         printf("ARRAY /dev/ddf UUID=");
1077         for (i = 0; i < DDF_GUID_LEN; i++) {
1078                 printf("%02x", ddf->anchor.guid[i]);
1079                 if ((i&3) == 0 && i != 0)
1080                         printf(":");
1081         }
1082         printf("\n");
1083 }
1084
1085 static void detail_super_ddf(struct supertype *st, char *homehost)
1086 {
1087         /* FIXME later
1088          * Could print DDF GUID
1089          * Need to find which array
1090          *  If whole, briefly list all arrays
1091          *  If one, give name
1092          */
1093 }
1094
1095 static void brief_detail_super_ddf(struct supertype *st)
1096 {
1097         /* FIXME I really need to know which array we are detailing.
1098          * Can that be stored in ddf_super??
1099          */
1100 //      struct ddf_super *ddf = st->sb;
1101 }
1102
1103
1104 #endif
1105
1106 static int match_home_ddf(struct supertype *st, char *homehost)
1107 {
1108         /* It matches 'this' host if the controller is a
1109          * Linux-MD controller with vendor_data matching
1110          * the hostname
1111          */
1112         struct ddf_super *ddf = st->sb;
1113         int len = strlen(homehost);
1114
1115         return (memcmp(ddf->controller.guid, T10, 8) == 0 &&
1116                 len < sizeof(ddf->controller.vendor_data) &&
1117                 memcmp(ddf->controller.vendor_data, homehost,len) == 0 &&
1118                 ddf->controller.vendor_data[len] == 0);
1119 }
1120
1121 static struct vd_config *find_vdcr(struct ddf_super *ddf, int inst)
1122 {
1123         struct vcl *v;
1124         if (inst < 0 || inst > __be16_to_cpu(ddf->virt->populated_vdes))
1125                 return NULL;
1126         for (v = ddf->conflist; v; v = v->next)
1127                 if (memcmp(v->conf.guid,
1128                            ddf->virt->entries[inst].guid,
1129                            DDF_GUID_LEN) == 0)
1130                         return &v->conf;
1131         return NULL;
1132 }
1133
1134 static int find_phys(struct ddf_super *ddf, __u32 phys_refnum)
1135 {
1136         /* Find the entry in phys_disk which has the given refnum
1137          * and return it's index
1138          */
1139         int i;
1140         for (i=0; i < __be16_to_cpu(ddf->phys->max_pdes); i++)
1141                 if (ddf->phys->entries[i].refnum == phys_refnum)
1142                         return i;
1143         return -1;
1144 }
1145
1146 static void uuid_from_super_ddf(struct supertype *st, int uuid[4])
1147 {
1148         /* The uuid returned here is used for:
1149          *  uuid to put into bitmap file (Create, Grow)
1150          *  uuid for backup header when saving critical section (Grow)
1151          *  comparing uuids when re-adding a device into an array
1152          * For each of these we can make do with a truncated
1153          * or hashed uuid rather than the original, as long as
1154          * everyone agrees.
1155          * In each case the uuid required is that of the data-array,
1156          * not the device-set.
1157          * In the case of SVD we assume the BVD is of interest,
1158          * though that might be the case if a bitmap were made for
1159          * a mirrored SVD - worry about that later.
1160          * So we need to find the VD configuration record for the
1161          * relevant BVD and extract the GUID and Secondary_Element_Seq.
1162          * The first 16 bytes of the sha1 of these is used.
1163          */
1164         struct ddf_super *ddf = st->sb;
1165         struct vd_config *vd = find_vdcr(ddf, st->container_member);
1166
1167         if (!vd)
1168                 memset(uuid, 0, sizeof (uuid));
1169         else {
1170                 char buf[20];
1171                 struct sha1_ctx ctx;
1172                 sha1_init_ctx(&ctx);
1173                 sha1_process_bytes(&vd->guid, DDF_GUID_LEN, &ctx);
1174                 if (vd->sec_elmnt_count > 1)
1175                         sha1_process_bytes(&vd->sec_elmnt_seq, 1, &ctx);
1176                 sha1_finish_ctx(&ctx, buf);
1177                 memcpy(uuid, buf, sizeof(uuid));
1178         }
1179 }
1180
1181 static void getinfo_super_ddf(struct supertype *st, struct mdinfo *info)
1182 {
1183         struct ddf_super *ddf = st->sb;
1184         int i;
1185
1186         info->array.major_version = 1000;
1187         info->array.minor_version = 0; /* FIXME use ddf->revision somehow */
1188         info->array.patch_version = 0;
1189         info->array.raid_disks    = __be16_to_cpu(ddf->phys->used_pdes);
1190         info->array.level         = LEVEL_CONTAINER;
1191         info->array.layout        = 0;
1192         info->array.md_minor      = -1;
1193         info->array.ctime         = DECADE + __be32_to_cpu(*(__u32*)
1194                                                          (ddf->anchor.guid+16));
1195         info->array.utime         = 0;
1196         info->array.chunk_size    = 0;
1197
1198 //      info->data_offset         = ???;
1199 //      info->component_size      = ???;
1200
1201         info->disk.major = 0;
1202         info->disk.minor = 0;
1203         info->disk.number = __be32_to_cpu(ddf->dlist->disk.refnum);
1204 //      info->disk.raid_disk = find refnum in the table and use index;
1205         info->disk.raid_disk = -1;
1206         for (i = 0; i < __be16_to_cpu(ddf->phys->max_pdes) ; i++)
1207                 if (ddf->phys->entries[i].refnum == ddf->dlist->disk.refnum) {
1208                         info->disk.raid_disk = i;
1209                         break;
1210                 }
1211         info->disk.state = (1 << MD_DISK_SYNC);
1212
1213         info->reshape_active = 0;
1214
1215         strcpy(info->text_version, "ddf");
1216
1217 //      uuid_from_super_ddf(info->uuid, sbv);
1218
1219 //      info->name[] ?? ;
1220 }
1221
1222 static void getinfo_super_n_container(struct supertype *st, struct mdinfo *info)
1223 {
1224         /* just need offset and size */
1225         struct ddf_super *ddf = st->sb;
1226         int n = info->disk.number;
1227
1228         info->data_offset = __be64_to_cpu(ddf->phys->entries[n].config_size);
1229         info->component_size = 32*1024*1024 / 512;
1230 }
1231
1232 static int rlq_to_layout(int rlq, int prl, int raiddisks);
1233
1234 static void getinfo_super_ddf_bvd(struct supertype *st, struct mdinfo *info)
1235 {
1236         struct ddf_super *ddf = st->sb;
1237         struct vd_config *vd = find_vdcr(ddf, info->container_member);
1238
1239         /* FIXME this returns BVD info - what if we want SVD ?? */
1240
1241         info->array.major_version = 1000;
1242         info->array.minor_version = 0; /* FIXME use ddf->revision somehow */
1243         info->array.patch_version = 0;
1244         info->array.raid_disks    = __be16_to_cpu(vd->prim_elmnt_count);
1245         info->array.level         = map_num1(ddf_level_num, vd->prl);
1246         info->array.layout        = rlq_to_layout(vd->rlq, vd->prl,
1247                                                   info->array.raid_disks);
1248         info->array.md_minor      = -1;
1249         info->array.ctime         = DECADE + __be32_to_cpu(*(__u32*)(vd->guid+16));
1250         info->array.utime         = DECADE + __be32_to_cpu(vd->timestamp);
1251         info->array.chunk_size    = 512 << vd->chunk_shift;
1252
1253 //      info->data_offset         = ???;
1254 //      info->component_size      = ???;
1255
1256         info->disk.major = 0;
1257         info->disk.minor = 0;
1258 //      info->disk.number = __be32_to_cpu(ddf->disk.refnum);
1259 //      info->disk.raid_disk = find refnum in the table and use index;
1260 //      info->disk.state = ???;
1261
1262         uuid_from_super_ddf(st, info->uuid);
1263
1264         sprintf(info->text_version, "/%s/%d",
1265                 devnum2devname(st->container_dev),
1266                 info->container_member);
1267
1268 //      info->name[] ?? ;
1269 }
1270
1271 static void getinfo_super_n_bvd(struct supertype *st, struct mdinfo *info)
1272 {
1273         /* Find the particular details for info->disk.raid_disk.
1274          * This includes data_offset, component_size,
1275          */
1276         struct ddf_super *ddf = st->sb;
1277         __u64 *lba_offset = ddf->newconf->lba_offset;
1278         struct vd_config *conf = &ddf->newconf->conf;
1279         info->data_offset = __be64_to_cpu(lba_offset[info->disk.raid_disk]);
1280         info->component_size = __be64_to_cpu(conf->blocks);
1281 }
1282
1283 static int update_super_ddf(struct supertype *st, struct mdinfo *info,
1284                             char *update,
1285                             char *devname, int verbose,
1286                             int uuid_set, char *homehost)
1287 {
1288         /* For 'assemble' and 'force' we need to return non-zero if any
1289          * change was made.  For others, the return value is ignored.
1290          * Update options are:
1291          *  force-one : This device looks a bit old but needs to be included,
1292          *        update age info appropriately.
1293          *  assemble: clear any 'faulty' flag to allow this device to
1294          *              be assembled.
1295          *  force-array: Array is degraded but being forced, mark it clean
1296          *         if that will be needed to assemble it.
1297          *
1298          *  newdev:  not used ????
1299          *  grow:  Array has gained a new device - this is currently for
1300          *              linear only
1301          *  resync: mark as dirty so a resync will happen.
1302          *  uuid:  Change the uuid of the array to match watch is given
1303          *  homehost:  update the recorded homehost
1304          *  name:  update the name - preserving the homehost
1305          *  _reshape_progress: record new reshape_progress position.
1306          *
1307          * Following are not relevant for this version:
1308          *  sparc2.2 : update from old dodgey metadata
1309          *  super-minor: change the preferred_minor number
1310          *  summaries:  update redundant counters.
1311          */
1312         int rv = 0;
1313 //      struct ddf_super *ddf = st->sb;
1314 //      struct vd_config *vd = find_vdcr(ddf, info->container_member);
1315 //      struct virtual_entry *ve = find_ve(ddf);
1316
1317
1318         /* we don't need to handle "force-*" or "assemble" as
1319          * there is no need to 'trick' the kernel.  We the metadata is
1320          * first updated to activate the array, all the implied modifications
1321          * will just happen.
1322          */
1323
1324         if (strcmp(update, "grow") == 0) {
1325                 /* FIXME */
1326         }
1327         if (strcmp(update, "resync") == 0) {
1328 //              info->resync_checkpoint = 0;
1329         }
1330         /* We ignore UUID updates as they make even less sense
1331          * with DDF
1332          */
1333         if (strcmp(update, "homehost") == 0) {
1334                 /* homehost is stored in controller->vendor_data,
1335                  * or it is when we are the vendor
1336                  */
1337 //              if (info->vendor_is_local)
1338 //                      strcpy(ddf->controller.vendor_data, homehost);
1339         }
1340         if (strcmp(update, "name") == 0) {
1341                 /* name is stored in virtual_entry->name */
1342 //              memset(ve->name, ' ', 16);
1343 //              strncpy(ve->name, info->name, 16);
1344         }
1345         if (strcmp(update, "_reshape_progress") == 0) {
1346                 /* We don't support reshape yet */
1347         }
1348
1349 //      update_all_csum(ddf);
1350
1351         return rv;
1352 }
1353
1354 static void make_header_guid(char *guid)
1355 {
1356         __u32 stamp;
1357         int rfd;
1358         /* Create a DDF Header of Virtual Disk GUID */
1359
1360         /* 24 bytes of fiction required.
1361          * first 8 are a 'vendor-id'  - "Linux-MD"
1362          * next 8 are controller type.. how about 0X DEAD BEEF 0000 0000
1363          * Remaining 8 random number plus timestamp
1364          */
1365         memcpy(guid, T10, sizeof(T10));
1366         stamp = __cpu_to_be32(0xdeadbeef);
1367         memcpy(guid+8, &stamp, 4);
1368         stamp = __cpu_to_be32(0);
1369         memcpy(guid+12, &stamp, 4);
1370         stamp = __cpu_to_be32(time(0) - DECADE);
1371         memcpy(guid+16, &stamp, 4);
1372         rfd = open("/dev/urandom", O_RDONLY);
1373         if (rfd < 0 || read(rfd, &stamp, 4) != 4)
1374                 stamp = random();
1375         memcpy(guid+20, &stamp, 4);
1376         if (rfd >= 0) close(rfd);
1377 }
1378 static int init_super_ddf(struct supertype *st,
1379                           mdu_array_info_t *info,
1380                           unsigned long long size, char *name, char *homehost,
1381                           int *uuid)
1382 {
1383         /* This is primarily called by Create when creating a new array.
1384          * We will then get add_to_super called for each component, and then
1385          * write_init_super called to write it out to each device.
1386          * For DDF, Create can create on fresh devices or on a pre-existing
1387          * array.
1388          * To create on a pre-existing array a different method will be called.
1389          * This one is just for fresh drives.
1390          *
1391          * We need to create the entire 'ddf' structure which includes:
1392          *  DDF headers - these are easy.
1393          *  Controller data - a Sector describing this controller .. not that
1394          *                  this is a controller exactly.
1395          *  Physical Disk Record - one entry per device, so
1396          *                      leave plenty of space.
1397          *  Virtual Disk Records - again, just leave plenty of space.
1398          *                   This just lists VDs, doesn't give details
1399          *  Config records - describes the VDs that use this disk
1400          *  DiskData  - describes 'this' device.
1401          *  BadBlockManagement - empty
1402          *  Diag Space - empty
1403          *  Vendor Logs - Could we put bitmaps here?
1404          *
1405          */
1406         struct ddf_super *ddf;
1407         char hostname[17];
1408         int hostlen;
1409         int max_phys_disks, max_virt_disks;
1410         unsigned long long sector;
1411         int clen;
1412         int i;
1413         int pdsize, vdsize;
1414         struct phys_disk *pd;
1415         struct virtual_disk *vd;
1416
1417         ddf = malloc(sizeof(*ddf));
1418         ddf->dlist = NULL; /* no physical disks yet */
1419         ddf->conflist = NULL; /* No virtual disks yet */
1420
1421         /* At least 32MB *must* be reserved for the ddf.  So let's just
1422          * start 32MB from the end, and put the primary header there.
1423          * Don't do secondary for now.
1424          * We don't know exactly where that will be yet as it could be
1425          * different on each device.  To just set up the lengths.
1426          *
1427          */
1428
1429         ddf->anchor.magic = DDF_HEADER_MAGIC;
1430         make_header_guid(ddf->anchor.guid);
1431
1432         memcpy(ddf->anchor.revision, DDF_REVISION, 8);
1433         ddf->anchor.seq = __cpu_to_be32(1);
1434         ddf->anchor.timestamp = __cpu_to_be32(time(0) - DECADE);
1435         ddf->anchor.openflag = 0xFF;
1436         ddf->anchor.foreignflag = 0;
1437         ddf->anchor.enforcegroups = 0; /* Is this best?? */
1438         ddf->anchor.pad0 = 0xff;
1439         memset(ddf->anchor.pad1, 0xff, 12);
1440         memset(ddf->anchor.header_ext, 0xff, 32);
1441         ddf->anchor.primary_lba = ~(__u64)0;
1442         ddf->anchor.secondary_lba = ~(__u64)0;
1443         ddf->anchor.type = DDF_HEADER_ANCHOR;
1444         memset(ddf->anchor.pad2, 0xff, 3);
1445         ddf->anchor.workspace_len = __cpu_to_be32(32768); /* Must be reserved */
1446         ddf->anchor.workspace_lba = ~(__u64)0; /* Put this at bottom
1447                                                   of 32M reserved.. */
1448         max_phys_disks = 1023;   /* Should be enough */
1449         ddf->anchor.max_pd_entries = __cpu_to_be16(max_phys_disks);
1450         max_virt_disks = 255;
1451         ddf->anchor.max_vd_entries = __cpu_to_be16(max_virt_disks); /* ?? */
1452         ddf->anchor.max_partitions = __cpu_to_be16(64); /* ?? */
1453         ddf->max_part = 64;
1454         ddf->anchor.config_record_len = __cpu_to_be16(1 + 256*12/512);
1455         ddf->anchor.max_primary_element_entries = __cpu_to_be16(256);
1456         memset(ddf->anchor.pad3, 0xff, 54);
1457
1458         /* controller sections is one sector long immediately
1459          * after the ddf header */
1460         sector = 1;
1461         ddf->anchor.controller_section_offset = __cpu_to_be32(sector);
1462         ddf->anchor.controller_section_length = __cpu_to_be32(1);
1463         sector += 1;
1464
1465         /* phys is 8 sectors after that */
1466         pdsize = ROUND_UP(sizeof(struct phys_disk) +
1467                           sizeof(struct phys_disk_entry)*max_phys_disks,
1468                           512);
1469         switch(pdsize/512) {
1470         case 2: case 8: case 32: case 128: case 512: break;
1471         default: abort();
1472         }
1473         ddf->anchor.phys_section_offset = __cpu_to_be32(sector);
1474         ddf->anchor.phys_section_length =
1475                 __cpu_to_be32(pdsize/512); /* max_primary_element_entries/8 */
1476         sector += pdsize/512;
1477
1478         /* virt is another 32 sectors */
1479         vdsize = ROUND_UP(sizeof(struct virtual_disk) +
1480                           sizeof(struct virtual_entry) * max_virt_disks,
1481                           512);
1482         switch(vdsize/512) {
1483         case 2: case 8: case 32: case 128: case 512: break;
1484         default: abort();
1485         }
1486         ddf->anchor.virt_section_offset = __cpu_to_be32(sector);
1487         ddf->anchor.virt_section_length =
1488                 __cpu_to_be32(vdsize/512); /* max_vd_entries/8 */
1489         sector += vdsize/512;
1490
1491         clen = (1 + 256*12/512) * (64+1);
1492         ddf->anchor.config_section_offset = __cpu_to_be32(sector);
1493         ddf->anchor.config_section_length = __cpu_to_be32(clen);
1494         sector += clen;
1495
1496         ddf->anchor.data_section_offset = __cpu_to_be32(sector);
1497         ddf->anchor.data_section_length = __cpu_to_be32(1);
1498         sector += 1;
1499
1500         ddf->anchor.bbm_section_length = __cpu_to_be32(0);
1501         ddf->anchor.bbm_section_offset = __cpu_to_be32(0xFFFFFFFF);
1502         ddf->anchor.diag_space_length = __cpu_to_be32(0);
1503         ddf->anchor.diag_space_offset = __cpu_to_be32(0xFFFFFFFF);
1504         ddf->anchor.vendor_length = __cpu_to_be32(0);
1505         ddf->anchor.vendor_offset = __cpu_to_be32(0xFFFFFFFF);
1506
1507         memset(ddf->anchor.pad4, 0xff, 256);
1508
1509         memcpy(&ddf->primary, &ddf->anchor, 512);
1510         memcpy(&ddf->secondary, &ddf->anchor, 512);
1511
1512         ddf->primary.openflag = 1; /* I guess.. */
1513         ddf->primary.type = DDF_HEADER_PRIMARY;
1514
1515         ddf->secondary.openflag = 1; /* I guess.. */
1516         ddf->secondary.type = DDF_HEADER_SECONDARY;
1517
1518         ddf->active = &ddf->primary;
1519
1520         ddf->controller.magic = DDF_CONTROLLER_MAGIC;
1521
1522         /* 24 more bytes of fiction required.
1523          * first 8 are a 'vendor-id'  - "Linux-MD"
1524          * Remaining 16 are serial number.... maybe a hostname would do?
1525          */
1526         memcpy(ddf->controller.guid, T10, sizeof(T10));
1527         gethostname(hostname, 17);
1528         hostname[17] = 0;
1529         hostlen = strlen(hostname);
1530         memcpy(ddf->controller.guid + 24 - hostlen, hostname, hostlen);
1531         for (i = strlen(T10) ; i+hostlen < 24; i++)
1532                 ddf->controller.guid[i] = ' ';
1533
1534         ddf->controller.type.vendor_id = __cpu_to_be16(0xDEAD);
1535         ddf->controller.type.device_id = __cpu_to_be16(0xBEEF);
1536         ddf->controller.type.sub_vendor_id = 0;
1537         ddf->controller.type.sub_device_id = 0;
1538         memcpy(ddf->controller.product_id, "What Is My PID??", 16);
1539         memset(ddf->controller.pad, 0xff, 8);
1540         memset(ddf->controller.vendor_data, 0xff, 448);
1541
1542         pd = ddf->phys = malloc(pdsize);
1543         ddf->pdsize = pdsize;
1544
1545         memset(pd, 0xff, pdsize);
1546         memset(pd, 0, sizeof(*pd));
1547         pd->magic = DDF_PHYS_DATA_MAGIC;
1548         pd->used_pdes = __cpu_to_be16(0);
1549         pd->max_pdes = __cpu_to_be16(max_phys_disks);
1550         memset(pd->pad, 0xff, 52);
1551
1552         vd = ddf->virt = malloc(vdsize);
1553         ddf->vdsize = vdsize;
1554         memset(vd, 0, vdsize);
1555         vd->magic = DDF_VIRT_RECORDS_MAGIC;
1556         vd->populated_vdes = __cpu_to_be16(0);
1557         vd->max_vdes = __cpu_to_be16(max_virt_disks);
1558         memset(vd->pad, 0xff, 52);
1559
1560         for (i=0; i<max_virt_disks; i++)
1561                 memset(&vd->entries[i], 0xff, sizeof(struct virtual_entry));
1562
1563         st->sb = ddf;
1564         return 1;
1565 }
1566
1567 static int all_ff(char *guid)
1568 {
1569         int i;
1570         for (i = 0; i < DDF_GUID_LEN; i++)
1571                 if (guid[i] != (char)0xff)
1572                         return 0;
1573         return 1;
1574 }
1575 static int chunk_to_shift(int chunksize)
1576 {
1577         return ffs(chunksize/512)-1;
1578 }
1579
1580 static int level_to_prl(int level)
1581 {
1582         switch (level) {
1583         case LEVEL_LINEAR: return DDF_CONCAT;
1584         case 0: return DDF_RAID0;
1585         case 1: return DDF_RAID1;
1586         case 4: return DDF_RAID4;
1587         case 5: return DDF_RAID5;
1588         case 6: return DDF_RAID6;
1589         default: return -1;
1590         }
1591 }
1592 static int layout_to_rlq(int level, int layout, int raiddisks)
1593 {
1594         switch(level) {
1595         case 0:
1596                 return DDF_RAID0_SIMPLE;
1597         case 1:
1598                 switch(raiddisks) {
1599                 case 2: return DDF_RAID1_SIMPLE;
1600                 case 3: return DDF_RAID1_MULTI;
1601                 default: return -1;
1602                 }
1603         case 4:
1604                 switch(layout) {
1605                 case 0: return DDF_RAID4_N;
1606                 }
1607                 break;
1608         case 5:
1609         case 6:
1610                 switch(layout) {
1611                 case ALGORITHM_LEFT_ASYMMETRIC:
1612                         return DDF_RAID5_N_RESTART;
1613                 case ALGORITHM_RIGHT_ASYMMETRIC:
1614                         return DDF_RAID5_0_RESTART;
1615                 case ALGORITHM_LEFT_SYMMETRIC:
1616                         return DDF_RAID5_N_CONTINUE;
1617                 case ALGORITHM_RIGHT_SYMMETRIC:
1618                         return -1; /* not mentioned in standard */
1619                 }
1620         }
1621         return -1;
1622 }
1623
1624 static int rlq_to_layout(int rlq, int prl, int raiddisks)
1625 {
1626         switch(prl) {
1627         case DDF_RAID0:
1628                 return 0; /* hopefully rlq == DDF_RAID0_SIMPLE */
1629         case DDF_RAID1:
1630                 return 0; /* hopefully rlq == SIMPLE or MULTI depending
1631                              on raiddisks*/
1632         case DDF_RAID4:
1633                 switch(rlq) {
1634                 case DDF_RAID4_N:
1635                         return 0;
1636                 default:
1637                         /* not supported */
1638                         return -1; /* FIXME this isn't checked */
1639                 }
1640         case DDF_RAID5:
1641         case DDF_RAID6:
1642                 switch(rlq) {
1643                 case DDF_RAID5_N_RESTART:
1644                         return ALGORITHM_LEFT_ASYMMETRIC;
1645                 case DDF_RAID5_0_RESTART:
1646                         return ALGORITHM_RIGHT_ASYMMETRIC;
1647                 case DDF_RAID5_N_CONTINUE:
1648                         return ALGORITHM_LEFT_SYMMETRIC;
1649                 default:
1650                         return -1;
1651                 }
1652         }
1653         return -1;
1654 }
1655
1656 static int init_super_ddf_bvd(struct supertype *st,
1657                               mdu_array_info_t *info,
1658                               unsigned long long size,
1659                               char *name, char *homehost,
1660                               int *uuid)
1661 {
1662         /* We are creating a BVD inside a pre-existing container.
1663          * so st->sb is already set.
1664          * We need to create a new vd_config and a new virtual_entry
1665          */
1666         struct ddf_super *ddf = st->sb;
1667         int venum;
1668         struct virtual_entry *ve;
1669         struct vcl *vcl;
1670         struct vd_config *vc;
1671         int mppe;
1672         int conflen;
1673
1674         if (__be16_to_cpu(ddf->virt->populated_vdes)
1675             >= __be16_to_cpu(ddf->virt->max_vdes)) {
1676                 fprintf(stderr, Name": This ddf already has the "
1677                         "maximum of %d virtual devices\n",
1678                         __be16_to_cpu(ddf->virt->max_vdes));
1679                 return 0;
1680         }
1681
1682         for (venum = 0; venum < __be16_to_cpu(ddf->virt->max_vdes); venum++)
1683                 if (all_ff(ddf->virt->entries[venum].guid))
1684                         break;
1685         if (venum == __be16_to_cpu(ddf->virt->max_vdes)) {
1686                 fprintf(stderr, Name ": Cannot find spare slot for "
1687                         "virtual disk - DDF is corrupt\n");
1688                 return 0;
1689         }
1690         ve = &ddf->virt->entries[venum];
1691         st->container_member = venum;
1692
1693         /* A Virtual Disk GUID contains the T10 Vendor ID, controller type,
1694          * timestamp, random number
1695          */
1696         make_header_guid(ve->guid);
1697         ve->unit = __cpu_to_be16(info->md_minor);
1698         ve->pad0 = 0xFFFF;
1699         ve->guid_crc = crc32(0, (unsigned char*)ddf->anchor.guid, DDF_GUID_LEN);
1700         ve->type = 0;
1701         ve->state = DDF_state_degraded; /* Will be modified as devices are added */
1702         if (info->state & 1) /* clean */
1703                 ve->init_state = DDF_init_full;
1704         else
1705                 ve->init_state = DDF_init_not;
1706
1707         memset(ve->pad1, 0xff, 14);
1708         memset(ve->name, ' ', 16);
1709         if (name)
1710                 strncpy(ve->name, name, 16);
1711         ddf->virt->populated_vdes =
1712                 __cpu_to_be16(__be16_to_cpu(ddf->virt->populated_vdes)+1);
1713
1714         /* Now create a new vd_config */
1715         conflen =  __be16_to_cpu(ddf->active->config_record_len);
1716         vcl = malloc(offsetof(struct vcl, conf) + conflen * 512);
1717         mppe = __be16_to_cpu(ddf->anchor.max_primary_element_entries);
1718         vcl->lba_offset = (__u64*) &vcl->conf.phys_refnum[mppe];
1719
1720         vc = &vcl->conf;
1721
1722         vc->magic = DDF_VD_CONF_MAGIC;
1723         memcpy(vc->guid, ve->guid, DDF_GUID_LEN);
1724         vc->timestamp = __cpu_to_be32(time(0)-DECADE);
1725         vc->seqnum = __cpu_to_be32(1);
1726         memset(vc->pad0, 0xff, 24);
1727         vc->prim_elmnt_count = __cpu_to_be16(info->raid_disks);
1728         vc->chunk_shift = chunk_to_shift(info->chunk_size);
1729         vc->prl = level_to_prl(info->level);
1730         vc->rlq = layout_to_rlq(info->level, info->layout, info->raid_disks);
1731         vc->sec_elmnt_count = 1;
1732         vc->sec_elmnt_seq = 0;
1733         vc->srl = 0;
1734         vc->blocks = __cpu_to_be64(info->size * 2);
1735         vc->array_blocks = __cpu_to_be64(
1736                 calc_array_size(info->level, info->raid_disks, info->layout,
1737                                 info->chunk_size, info->size*2));
1738         memset(vc->pad1, 0xff, 8);
1739         vc->spare_refs[0] = 0xffffffff;
1740         vc->spare_refs[1] = 0xffffffff;
1741         vc->spare_refs[2] = 0xffffffff;
1742         vc->spare_refs[3] = 0xffffffff;
1743         vc->spare_refs[4] = 0xffffffff;
1744         vc->spare_refs[5] = 0xffffffff;
1745         vc->spare_refs[6] = 0xffffffff;
1746         vc->spare_refs[7] = 0xffffffff;
1747         memset(vc->cache_pol, 0, 8);
1748         vc->bg_rate = 0x80;
1749         memset(vc->pad2, 0xff, 3);
1750         memset(vc->pad3, 0xff, 52);
1751         memset(vc->pad4, 0xff, 192);
1752         memset(vc->v0, 0xff, 32);
1753         memset(vc->v1, 0xff, 32);
1754         memset(vc->v2, 0xff, 16);
1755         memset(vc->v3, 0xff, 16);
1756         memset(vc->vendor, 0xff, 32);
1757
1758         memset(vc->phys_refnum, 0xff, 4*mppe);
1759         memset(vc->phys_refnum+mppe, 0x00, 8*mppe);
1760
1761         vcl->next = ddf->conflist;
1762         ddf->conflist = vcl;
1763         ddf->newconf = vcl;
1764         return 1;
1765 }
1766
1767 static void add_to_super_ddf_bvd(struct supertype *st,
1768                                  mdu_disk_info_t *dk, int fd, char *devname)
1769 {
1770         /* fd and devname identify a device with-in the ddf container (st).
1771          * dk identifies a location in the new BVD.
1772          * We need to find suitable free space in that device and update
1773          * the phys_refnum and lba_offset for the newly created vd_config.
1774          * We might also want to update the type in the phys_disk
1775          * section. FIXME
1776          */
1777         struct dl *dl;
1778         struct ddf_super *ddf = st->sb;
1779         struct vd_config *vc;
1780         __u64 *lba_offset;
1781         int mppe;
1782         int working;
1783
1784         for (dl = ddf->dlist; dl ; dl = dl->next)
1785                 if (dl->major == dk->major &&
1786                     dl->minor == dk->minor)
1787                         break;
1788         if (!dl || ! (dk->state & (1<<MD_DISK_SYNC)))
1789                 return;
1790
1791         vc = &ddf->newconf->conf;
1792         vc->phys_refnum[dk->raid_disk] = dl->disk.refnum;
1793         mppe = __be16_to_cpu(ddf->anchor.max_primary_element_entries);
1794         lba_offset = (__u64*)(vc->phys_refnum + mppe);
1795         lba_offset[dk->raid_disk] = 0; /* FIXME */
1796
1797         dl->vlist[0] = ddf->newconf; /* FIXME */
1798
1799         dl->fd = fd;
1800         dl->devname = devname;
1801
1802         /* Check how many working raid_disks, and if we can mark
1803          * array as optimal yet
1804          */
1805         working = 0;
1806 #if 0
1807         for (i=0; i < __be16_to_cpu(vc->prim_elmnt_count); i++)
1808                 if (vc->phys_refnum[i] != 0xffffffff)
1809                         working++;
1810         if (working == __be16_to_cpu(vc->prim_elmnt_count))
1811                 ->entries[xx].state = (->entries[xx].state & ~DDF_state_mask)
1812                         | DDF_state_optimal;
1813
1814         if (vc->prl == DDF_RAID6 &&
1815             working+1 == __be16_to_cpu(vc->prim_elmnt_count))
1816                 ->entries[xx].state = (->entries[xx].state & ~DDF_state_mask)
1817                         | DDF_state_part_optimal;
1818 #endif
1819 }
1820
1821 /* add a device to a container, either while creating it or while
1822  * expanding a pre-existing container
1823  */
1824 static void add_to_super_ddf(struct supertype *st,
1825                              mdu_disk_info_t *dk, int fd, char *devname)
1826 {
1827         struct ddf_super *ddf = st->sb;
1828         struct dl *dd;
1829         time_t now;
1830         struct tm *tm;
1831         unsigned long long size;
1832         struct phys_disk_entry *pde;
1833         int n, i;
1834         struct stat stb;
1835
1836         /* This is device numbered dk->number.  We need to create
1837          * a phys_disk entry and a more detailed disk_data entry.
1838          */
1839         fstat(fd, &stb);
1840         dd = malloc(sizeof(*dd) + sizeof(dd->vlist[0]) * (ddf->max_part+1));
1841         dd->major = major(stb.st_rdev);
1842         dd->minor = minor(stb.st_rdev);
1843         dd->devname = devname;
1844         dd->next = ddf->dlist;
1845         dd->fd = fd;
1846
1847         dd->disk.magic = DDF_PHYS_DATA_MAGIC;
1848         now = time(0);
1849         tm = localtime(&now);
1850         sprintf(dd->disk.guid, "%8s%04d%02d%02d",
1851                 T10, tm->tm_year+1900, tm->tm_mon+1, tm->tm_mday);
1852         *(__u32*)(dd->disk.guid + 16) = random();
1853         *(__u32*)(dd->disk.guid + 20) = random();
1854
1855         dd->disk.refnum = random(); /* and hope for the best FIXME check this is unique!!*/
1856         dd->disk.forced_ref = 1;
1857         dd->disk.forced_guid = 1;
1858         memset(dd->disk.vendor, ' ', 32);
1859         memcpy(dd->disk.vendor, "Linux", 5);
1860         memset(dd->disk.pad, 0xff, 442);
1861         for (i = 0; i < ddf->max_part+1 ; i++)
1862                 dd->vlist[i] = NULL;
1863
1864         n = __be16_to_cpu(ddf->phys->used_pdes);
1865         pde = &ddf->phys->entries[n];
1866         n++;
1867         ddf->phys->used_pdes = __cpu_to_be16(n);
1868
1869         memcpy(pde->guid, dd->disk.guid, DDF_GUID_LEN);
1870         pde->refnum = dd->disk.refnum;
1871         pde->type = __cpu_to_be16(DDF_Forced_PD_GUID |DDF_Global_Spare);
1872         pde->state = __cpu_to_be16(DDF_Online);
1873         get_dev_size(fd, NULL, &size);
1874         /* We are required to reserve 32Meg, and record the size in sectors */
1875         pde->config_size = __cpu_to_be64( (size - 32*1024*1024) / 512);
1876         sprintf(pde->path, "%17.17s","Information: nil") ;
1877         memset(pde->pad, 0xff, 6);
1878
1879         ddf->dlist = dd;
1880 }
1881
1882 /*
1883  * This is the write_init_super method for a ddf container.  It is
1884  * called when creating a container or adding another device to a
1885  * container.
1886  */
1887
1888 #ifndef MDASSEMBLE
1889 static int __write_init_super_ddf(struct supertype *st, int do_close)
1890 {
1891
1892         struct ddf_super *ddf = st->sb;
1893         int i;
1894         struct dl *d;
1895         int n_config;
1896         int conf_size;
1897
1898         unsigned long long size, sector;
1899
1900         for (d = ddf->dlist; d; d=d->next) {
1901                 int fd = d->fd;
1902
1903                 if (fd < 0)
1904                         continue;
1905
1906                 /* We need to fill in the primary, (secondary) and workspace
1907                  * lba's in the headers, set their checksums,
1908                  * Also checksum phys, virt....
1909                  *
1910                  * Then write everything out, finally the anchor is written.
1911                  */
1912                 get_dev_size(fd, NULL, &size);
1913                 size /= 512;
1914                 ddf->anchor.workspace_lba = __cpu_to_be64(size - 32*1024*2);
1915                 ddf->anchor.primary_lba = __cpu_to_be64(size - 16*1024*2);
1916                 ddf->anchor.seq = __cpu_to_be32(1);
1917                 memcpy(&ddf->primary, &ddf->anchor, 512);
1918                 memcpy(&ddf->secondary, &ddf->anchor, 512);
1919
1920                 ddf->anchor.openflag = 0xFF; /* 'open' means nothing */
1921                 ddf->anchor.seq = 0xFFFFFFFF; /* no sequencing in anchor */
1922                 ddf->anchor.crc = calc_crc(&ddf->anchor, 512);
1923
1924                 ddf->primary.openflag = 0;
1925                 ddf->primary.type = DDF_HEADER_PRIMARY;
1926
1927                 ddf->secondary.openflag = 0;
1928                 ddf->secondary.type = DDF_HEADER_SECONDARY;
1929
1930                 ddf->primary.crc = calc_crc(&ddf->primary, 512);
1931                 ddf->secondary.crc = calc_crc(&ddf->secondary, 512);
1932
1933                 sector = size - 16*1024*2;
1934                 lseek64(fd, sector<<9, 0);
1935                 write(fd, &ddf->primary, 512);
1936
1937                 ddf->controller.crc = calc_crc(&ddf->controller, 512);
1938                 write(fd, &ddf->controller, 512);
1939
1940                 ddf->phys->crc = calc_crc(ddf->phys, ddf->pdsize);
1941
1942                 write(fd, ddf->phys, ddf->pdsize);
1943
1944                 ddf->virt->crc = calc_crc(ddf->virt, ddf->vdsize);
1945                 write(fd, ddf->virt, ddf->vdsize);
1946
1947                 /* Now write lots of config records. */
1948                 n_config = __be16_to_cpu(ddf->active->max_partitions);
1949                 conf_size = __be16_to_cpu(ddf->active->config_record_len) * 512;
1950                 for (i = 0 ; i <= n_config ; i++) {
1951                         struct vcl *c = d->vlist[i];
1952
1953                         if (c) {
1954                                 c->conf.crc = calc_crc(&c->conf, conf_size);
1955                                 write(fd, &c->conf, conf_size);
1956                         } else {
1957                                 __u32 sig = 0xffffffff;
1958                                 write(fd, &sig, 4);
1959                                 lseek64(fd, conf_size-4, SEEK_CUR);
1960                         }
1961                 }
1962                 d->disk.crc = calc_crc(&d->disk, 512);
1963                 write(fd, &d->disk, 512);
1964
1965                 /* Maybe do the same for secondary */
1966
1967                 lseek64(fd, (size-1)*512, SEEK_SET);
1968                 write(fd, &ddf->anchor, 512);
1969                 if (do_close) {
1970                         close(fd);
1971                         d->fd = -1;
1972                 }
1973         }
1974         return 1;
1975 }
1976
1977 static int write_init_super_ddf(struct supertype *st)
1978 {
1979         return __write_init_super_ddf(st, 1);
1980 }
1981
1982 #endif
1983
1984 static __u64 avail_size_ddf(struct supertype *st, __u64 devsize)
1985 {
1986         /* We must reserve the last 32Meg */
1987         if (devsize <= 32*1024*2)
1988                 return 0;
1989         return devsize - 32*1024*2;
1990 }
1991
1992 #ifndef MDASSEMBLE
1993 int validate_geometry_ddf(struct supertype *st,
1994                           int level, int layout, int raiddisks,
1995                           int chunk, unsigned long long size,
1996                           char *dev, unsigned long long *freesize)
1997 {
1998         int fd;
1999         struct mdinfo *sra;
2000         int cfd;
2001
2002         /* ddf potentially supports lots of things, but it depends on
2003          * what devices are offered (and maybe kernel version?)
2004          * If given unused devices, we will make a container.
2005          * If given devices in a container, we will make a BVD.
2006          * If given BVDs, we make an SVD, changing all the GUIDs in the process.
2007          */
2008
2009         if (level == LEVEL_CONTAINER) {
2010                 st->ss = &super_ddf_container;
2011                 if (dev) {
2012                         int rv =st->ss->validate_geometry(st, level, layout,
2013                                                           raiddisks, chunk,
2014                                                           size,
2015                                                           NULL, freesize);
2016                         if (rv)
2017                                 return rv;
2018                 }
2019                 return st->ss->validate_geometry(st, level, layout, raiddisks,
2020                                                  chunk, size, dev, freesize);
2021         }
2022
2023         if (st->sb) {
2024                 /* creating in a given container */
2025                 st->ss = &super_ddf_bvd;
2026                 if (dev) {
2027                         int rv =st->ss->validate_geometry(st, level, layout,
2028                                                           raiddisks, chunk,
2029                                                           size,
2030                                                           NULL, freesize);
2031                         if (rv)
2032                                 return rv;
2033                 }
2034                 return st->ss->validate_geometry(st, level, layout, raiddisks,
2035                                                  chunk, size, dev, freesize);
2036         }
2037         /* FIXME should exclude MULTIPATH, or more appropriately, allow
2038          * only known levels.
2039          */
2040         if (!dev)
2041                 return 1;
2042
2043         /* This device needs to be either a device in a 'ddf' container,
2044          * or it needs to be a 'ddf-bvd' array.
2045          */
2046
2047         fd = open(dev, O_RDONLY|O_EXCL, 0);
2048         if (fd >= 0) {
2049                 sra = sysfs_read(fd, 0, GET_VERSION);
2050                 close(fd);
2051                 if (sra && sra->array.major_version == -1 &&
2052                     strcmp(sra->text_version, "ddf-bvd") == 0) {
2053                         st->ss = &super_ddf_svd;
2054                         return st->ss->validate_geometry(st, level, layout,
2055                                                          raiddisks, chunk, size,
2056                                                          dev, freesize);
2057                 }
2058
2059                 fprintf(stderr,
2060                         Name ": Cannot create this array on device %s\n",
2061                         dev);
2062                 return 0;
2063         }
2064         if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) {
2065                 fprintf(stderr, Name ": Cannot open %s: %s\n",
2066                         dev, strerror(errno));
2067                 return 0;
2068         }
2069         /* Well, it is in use by someone, maybe a 'ddf' container. */
2070         cfd = open_container(fd);
2071         if (cfd < 0) {
2072                 close(fd);
2073                 fprintf(stderr, Name ": Cannot use %s: It is busy\n",
2074                         dev);
2075                 return 0;
2076         }
2077         sra = sysfs_read(cfd, 0, GET_VERSION);
2078         close(fd);
2079         if (sra && sra->array.major_version == -1 &&
2080             strcmp(sra->text_version, "ddf") == 0) {
2081                 /* This is a member of a ddf container.  Load the container
2082                  * and try to create a bvd
2083                  */
2084                 struct ddf_super *ddf;
2085                 st->ss = &super_ddf_bvd;
2086                 if (load_super_ddf_all(st, cfd, (void **)&ddf, NULL, 1) == 0) {
2087                         st->sb = ddf;
2088                         st->container_dev = fd2devnum(cfd);
2089                         st->container_member = 27; // FIXME
2090                         close(cfd);
2091                         return st->ss->validate_geometry(st, level, layout,
2092                                                          raiddisks, chunk, size,
2093                                                          dev, freesize);
2094                 }
2095                 close(cfd);
2096         }
2097         fprintf(stderr, Name ": Cannot use %s: Already in use\n",
2098                 dev);
2099         return 1;
2100 }
2101
2102 int validate_geometry_ddf_container(struct supertype *st,
2103                                     int level, int layout, int raiddisks,
2104                                     int chunk, unsigned long long size,
2105                                     char *dev, unsigned long long *freesize)
2106 {
2107         int fd;
2108         unsigned long long ldsize;
2109
2110         if (level != LEVEL_CONTAINER)
2111                 return 0;
2112         if (!dev)
2113                 return 1;
2114
2115         fd = open(dev, O_RDONLY|O_EXCL, 0);
2116         if (fd < 0) {
2117                 fprintf(stderr, Name ": Cannot open %s: %s\n",
2118                         dev, strerror(errno));
2119                 return 0;
2120         }
2121         if (!get_dev_size(fd, dev, &ldsize)) {
2122                 close(fd);
2123                 return 0;
2124         }
2125         close(fd);
2126
2127         *freesize = avail_size_ddf(st, ldsize);
2128
2129         return 1;
2130 }
2131
2132 struct extent {
2133         unsigned long long start, size;
2134 };
2135 int cmp_extent(const void *av, const void *bv)
2136 {
2137         const struct extent *a = av;
2138         const struct extent *b = bv;
2139         if (a->start < b->start)
2140                 return -1;
2141         if (a->start > b->start)
2142                 return 1;
2143         return 0;
2144 }
2145
2146 struct extent *get_extents(struct ddf_super *ddf, struct dl *dl)
2147 {
2148         /* find a list of used extents on the give physical device
2149          * (dnum) or the given ddf.
2150          * Return a malloced array of 'struct extent'
2151
2152 FIXME ignore DDF_Legacy devices?
2153
2154          */
2155         struct extent *rv;
2156         int n = 0;
2157         int dnum;
2158         int i, j;
2159
2160         for (dnum = 0; dnum < ddf->phys->used_pdes; dnum++)
2161                 if (memcmp(dl->disk.guid,
2162                            ddf->phys->entries[dnum].guid,
2163                            DDF_GUID_LEN) == 0)
2164                         break;
2165
2166         if (dnum == ddf->phys->used_pdes)
2167                 return NULL;
2168
2169         rv = malloc(sizeof(struct extent) * (ddf->max_part + 2));
2170         if (!rv)
2171                 return NULL;
2172
2173         for (i = 0; i < ddf->max_part+1; i++) {
2174                 struct vcl *v = dl->vlist[i];
2175                 if (v == NULL)
2176                         continue;
2177                 for (j=0; j < v->conf.prim_elmnt_count; j++)
2178                         if (v->conf.phys_refnum[j] == dl->disk.refnum) {
2179                                 /* This device plays role 'j' in  'v'. */
2180                                 rv[n].start = __be64_to_cpu(v->lba_offset[j]);
2181                                 rv[n].size = __be64_to_cpu(v->conf.blocks);
2182                                 n++;
2183                                 break;
2184                         }
2185         }
2186         qsort(rv, n, sizeof(*rv), cmp_extent);
2187
2188         rv[n].start = __be64_to_cpu(ddf->phys->entries[dnum].config_size);
2189         rv[n].size = 0;
2190         return rv;
2191 }
2192
2193 int validate_geometry_ddf_bvd(struct supertype *st,
2194                               int level, int layout, int raiddisks,
2195                               int chunk, unsigned long long size,
2196                               char *dev, unsigned long long *freesize)
2197 {
2198         struct stat stb;
2199         struct ddf_super *ddf = st->sb;
2200         struct dl *dl;
2201         unsigned long long pos = 0;
2202         unsigned long long maxsize;
2203         struct extent *e;
2204         int i;
2205         /* ddf/bvd supports lots of things, but not containers */
2206         if (level == LEVEL_CONTAINER)
2207                 return 0;
2208         /* We must have the container info already read in. */
2209         if (!ddf)
2210                 return 0;
2211
2212         if (!dev) {
2213                 /* General test:  make sure there is space for
2214                  * 'raiddisks' device extents of size 'size'.
2215                  */
2216                 unsigned long long minsize = size;
2217                 int dcnt = 0;
2218                 if (minsize == 0)
2219                         minsize = 8;
2220                 for (dl = ddf->dlist; dl ; dl = dl->next)
2221                 {
2222                         int found = 0;
2223
2224                         i = 0;
2225                         e = get_extents(ddf, dl);
2226                         if (!e) continue;
2227                         do {
2228                                 unsigned long long esize;
2229                                 esize = e[i].start - pos;
2230                                 if (esize >= minsize)
2231                                         found = 1;
2232                                 pos = e[i].start + e[i].size;
2233                                 i++;
2234                         } while (e[i-1].size);
2235                         if (found)
2236                                 dcnt++;
2237                         free(e);
2238                 }
2239                 if (dcnt < raiddisks) {
2240                         fprintf(stderr, Name ": Not enough devices with space "
2241                                 "for this array (%d < %d)\n",
2242                                 dcnt, raiddisks);
2243                         return 0;
2244                 }
2245                 return 1;
2246         }
2247         /* This device must be a member of the set */
2248         if (stat(dev, &stb) < 0)
2249                 return 0;
2250         if ((S_IFMT & stb.st_mode) != S_IFBLK)
2251                 return 0;
2252         for (dl = ddf->dlist ; dl ; dl = dl->next) {
2253                 if (dl->major == major(stb.st_rdev) &&
2254                     dl->minor == minor(stb.st_rdev))
2255                         break;
2256         }
2257         if (!dl) {
2258                 fprintf(stderr, Name ": %s is not in the same DDF set\n",
2259                         dev);
2260                 return 0;
2261         }
2262         e = get_extents(ddf, dl);
2263         maxsize = 0;
2264         i = 0;
2265         if (e) do {
2266                 unsigned long long esize;
2267                 esize = e[i].start - pos;
2268                 if (esize >= maxsize)
2269                         maxsize = esize;
2270                 pos = e[i].start + e[i].size;
2271                 i++;
2272         } while (e[i-1].size);
2273         *freesize = maxsize;
2274         // FIXME here I am
2275
2276         return 1;
2277 }
2278 int validate_geometry_ddf_svd(struct supertype *st,
2279                               int level, int layout, int raiddisks,
2280                               int chunk, unsigned long long size,
2281                               char *dev, unsigned long long *freesize)
2282 {
2283         /* dd/svd only supports striped, mirrored, concat, spanned... */
2284         if (level != LEVEL_LINEAR &&
2285             level != 0 &&
2286             level != 1)
2287                 return 0;
2288         return 1;
2289 }
2290
2291
2292 static int load_super_ddf_all(struct supertype *st, int fd,
2293                               void **sbp, char *devname, int keep_fd)
2294 {
2295         struct mdinfo *sra;
2296         struct ddf_super *super;
2297         struct mdinfo *sd, *best = NULL;
2298         int bestseq = 0;
2299         int seq;
2300         char nm[20];
2301         int dfd;
2302
2303         sra = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE);
2304         if (!sra)
2305                 return 1;
2306         if (sra->array.major_version != -1 ||
2307             sra->array.minor_version != -2 ||
2308             strcmp(sra->text_version, "ddf") != 0)
2309                 return 1;
2310
2311         super = malloc(sizeof(*super));
2312         if (!super)
2313                 return 1;
2314         memset(super, 0, sizeof(*super));
2315
2316         /* first, try each device, and choose the best ddf */
2317         for (sd = sra->devs ; sd ; sd = sd->next) {
2318                 int rv;
2319                 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
2320                 dfd = dev_open(nm, O_RDONLY);
2321                 if (dfd < 0)
2322                         return 2;
2323                 rv = load_ddf_headers(dfd, super, NULL);
2324                 close(dfd);
2325                 if (rv == 0) {
2326                         seq = __be32_to_cpu(super->active->seq);
2327                         if (super->active->openflag)
2328                                 seq--;
2329                         if (!best || seq > bestseq) {
2330                                 bestseq = seq;
2331                                 best = sd;
2332                         }
2333                 }
2334         }
2335         if (!best)
2336                 return 1;
2337         /* OK, load this ddf */
2338         sprintf(nm, "%d:%d", best->disk.major, best->disk.minor);
2339         dfd = dev_open(nm, O_RDONLY);
2340         if (dfd < 0)
2341                 return 1;
2342         load_ddf_headers(dfd, super, NULL);
2343         load_ddf_global(dfd, super, NULL);
2344         close(dfd);
2345         /* Now we need the device-local bits */
2346         for (sd = sra->devs ; sd ; sd = sd->next) {
2347                 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
2348                 dfd = dev_open(nm, keep_fd? O_RDWR : O_RDONLY);
2349                 if (dfd < 0)
2350                         return 2;
2351                 seq = load_ddf_local(dfd, super, NULL, keep_fd);
2352                 if (!keep_fd) close(dfd);
2353         }
2354         *sbp = super;
2355         if (st->ss == NULL) {
2356                 st->ss = &super_ddf_container;
2357                 st->minor_version = 0;
2358                 st->max_devs = 512;
2359                 st->container_dev = fd2devnum(fd);
2360         }
2361         return 0;
2362 }
2363 #endif
2364
2365
2366
2367 static struct mdinfo *container_content_ddf(struct supertype *st)
2368 {
2369         /* Given a container loaded by load_super_ddf_all,
2370          * extract information about all the arrays into
2371          * an mdinfo tree.
2372          *
2373          * For each vcl in conflist: create an mdinfo, fill it in,
2374          *  then look for matching devices (phys_refnum) in dlist
2375          *  and create appropriate device mdinfo.
2376          */
2377         struct ddf_super *ddf = st->sb;
2378         struct mdinfo *rest = NULL;
2379         struct vcl *vc;
2380
2381         for (vc = ddf->conflist ; vc ; vc=vc->next)
2382         {
2383                 int mppe;
2384                 int i;
2385                 struct mdinfo *this;
2386                 this = malloc(sizeof(*this));
2387                 memset(this, 0, sizeof(*this));
2388                 this->next = rest;
2389                 rest = this;
2390
2391                 this->array.major_version = 1000;
2392                 this->array.minor_version = 0;
2393                 this->array.patch_version = 0;
2394                 this->array.level = map_num1(ddf_level_num, vc->conf.prl);
2395                 this->array.raid_disks =
2396                         __be16_to_cpu(vc->conf.prim_elmnt_count);
2397                 /* FIXME this should be mapped */
2398                 this->array.layout = vc->conf.rlq;
2399                 this->array.md_minor      = -1;
2400                 this->array.ctime         = DECADE +
2401                         __be32_to_cpu(*(__u32*)(vc->conf.guid+16));
2402                 this->array.utime         = DECADE +
2403                         __be32_to_cpu(vc->conf.timestamp);
2404                 this->array.chunk_size    = 512 << vc->conf.chunk_shift;
2405
2406                 for (i=0; i < __be16_to_cpu(ddf->virt->populated_vdes); i++)
2407                         if (memcmp(ddf->virt->entries[i].guid,
2408                                    vc->conf.guid, DDF_GUID_LEN) == 0)
2409                                 break;
2410                 if ((ddf->virt->entries[i].state & DDF_state_inconsistent) ||
2411                     (ddf->virt->entries[i].init_state & DDF_initstate_mask) !=
2412                     DDF_init_full) {
2413                         this->array.state = 0;
2414                         this->resync_start = 0;
2415                 } else {
2416                         this->array.state = 1;
2417                         this->resync_start = ~0ULL;
2418                 }
2419                 memcpy(this->name, ddf->virt->entries[i].name, 32);
2420                 this->name[33]=0;
2421
2422                 memset(this->uuid, 0, sizeof(this->uuid));
2423                 this->component_size = __be64_to_cpu(vc->conf.blocks);
2424                 this->array.size = this->component_size / 2;
2425                 this->container_member = i;
2426
2427                 mppe = __be16_to_cpu(ddf->anchor.max_primary_element_entries);
2428                 for (i=0 ; i < mppe ; i++) {
2429                         struct mdinfo *dev;
2430                         struct dl *d;
2431
2432                         if (vc->conf.phys_refnum[i] == 0xFFFFFFFF)
2433                                 continue;
2434
2435                         this->array.working_disks++;
2436
2437                         for (d = ddf->dlist; d ; d=d->next)
2438                                 if (d->disk.refnum == vc->conf.phys_refnum[i])
2439                                         break;
2440                         if (d == NULL)
2441                                 break;
2442
2443                         dev = malloc(sizeof(*dev));
2444                         memset(dev, 0, sizeof(*dev));
2445                         dev->next = this->devs;
2446                         this->devs = dev;
2447
2448                         dev->disk.number = __be32_to_cpu(d->disk.refnum);
2449                         dev->disk.major = d->major;
2450                         dev->disk.minor = d->minor;
2451                         dev->disk.raid_disk = i;
2452                         dev->disk.state = (1<<MD_DISK_SYNC)|(1<<MD_DISK_ACTIVE);
2453
2454                         dev->events = __le32_to_cpu(ddf->primary.seq);
2455                         dev->data_offset = vc->lba_offset[i];
2456                         dev->component_size = __be64_to_cpu(vc->conf.blocks);
2457                         if (d->devname)
2458                                 strcpy(dev->name, d->devname);
2459                 }
2460         }
2461         return rest;
2462 }
2463
2464 static int init_zero_ddf(struct supertype *st,
2465                          mdu_array_info_t *info,
2466                          unsigned long long size, char *name,
2467                          char *homehost, int *uuid)
2468 {
2469         st->sb = NULL;
2470         return 0;
2471 }
2472
2473 static int store_zero_ddf(struct supertype *st, int fd)
2474 {
2475         unsigned long long dsize;
2476         char buf[512];
2477         memset(buf, 0, 512);
2478
2479
2480         if (!get_dev_size(fd, NULL, &dsize))
2481                 return 1;
2482
2483         lseek64(fd, dsize-512, 0);
2484         write(fd, buf, 512);
2485         return 0;
2486 }
2487
2488 static int compare_super_ddf(struct supertype *st, struct supertype *tst)
2489 {
2490         /*
2491          * return:
2492          *  0 same, or first was empty, and second was copied
2493          *  1 second had wrong number
2494          *  2 wrong uuid
2495          *  3 wrong other info
2496          */
2497         struct ddf_super *first = st->sb;
2498         struct ddf_super *second = tst->sb;
2499
2500         if (!first) {
2501                 st->sb = tst->sb;
2502                 tst->sb = NULL;
2503                 return 0;
2504         }
2505
2506         if (memcmp(first->anchor.guid, second->anchor.guid, DDF_GUID_LEN) != 0)
2507                 return 2;
2508
2509         /* FIXME should I look at anything else? */
2510         return 0;
2511 }
2512
2513 /*
2514  * A new array 'a' has been started which claims to be instance 'inst'
2515  * within container 'c'.
2516  * We need to confirm that the array matches the metadata in 'c' so
2517  * that we don't corrupt any metadata.
2518  */
2519 static int ddf_open_new(struct supertype *c, struct active_array *a, int inst)
2520 {
2521         fprintf(stderr, "ddf: open_new %d\n", inst);
2522         return 0;
2523 }
2524
2525 /*
2526  * The array 'a' is to be marked clean in the metadata.
2527  * If '->resync_start' is not ~(unsigned long long)0, then the array is only
2528  * clean up to the point (in sectors).  If that cannot be recorded in the
2529  * metadata, then leave it as dirty.
2530  *
2531  * For DDF, we need to clear the DDF_state_inconsistent bit in the
2532  * !global! virtual_disk.virtual_entry structure.
2533  */
2534 static void ddf_set_array_state(struct active_array *a, int consistent)
2535 {
2536         struct ddf_super *ddf = a->container->sb;
2537         int inst = a->info.container_member;
2538         if (consistent)
2539                 ddf->virt->entries[inst].state &= ~DDF_state_inconsistent;
2540         else
2541                 ddf->virt->entries[inst].state |= DDF_state_inconsistent;
2542         ddf->virt->entries[inst].init_state &= ~DDF_initstate_mask;
2543         if (a->resync_start == ~0ULL)
2544                 ddf->virt->entries[inst].init_state |= DDF_init_full;
2545         else if (a->resync_start == 0)
2546                 ddf->virt->entries[inst].init_state |= DDF_init_not;
2547         else
2548                 ddf->virt->entries[inst].init_state |= DDF_init_quick;
2549
2550         printf("ddf mark %s %llu\n", consistent?"clean":"dirty",
2551                a->resync_start);
2552 }
2553
2554 /*
2555  * The state of each disk is stored in the global phys_disk structure
2556  * in phys_disk.entries[n].state.
2557  * This makes various combinations awkward.
2558  * - When a device fails in any array, it must be failed in all arrays
2559  *   that include a part of this device.
2560  * - When a component is rebuilding, we cannot include it officially in the
2561  *   array unless this is the only array that uses the device.
2562  *
2563  * So: when transitioning:
2564  *   Online -> failed,  just set failed flag.  monitor will propagate
2565  *   spare -> online,   the device might need to be added to the array.
2566  *   spare -> failed,   just set failed.  Don't worry if in array or not.
2567  */
2568 static void ddf_set_disk(struct active_array *a, int n, int state)
2569 {
2570         struct ddf_super *ddf = a->container->sb;
2571         int inst = a->info.container_member;
2572         struct vd_config *vc = find_vdcr(ddf, inst);
2573         int pd = find_phys(ddf, vc->phys_refnum[n]);
2574         int i, st, working;
2575
2576         if (vc == NULL) {
2577                 fprintf(stderr, "ddf: cannot find instance %d!!\n", inst);
2578                 return;
2579         }
2580         if (pd < 0) {
2581                 /* disk doesn't currently exist. If it is now in_sync,
2582                  * insert it. */
2583                 if ((state & DS_INSYNC) && ! (state & DS_FAULTY)) {
2584                         /* Find dev 'n' in a->info->devs, determine the
2585                          * ddf refnum, and set vc->phys_refnum and update
2586                          * phys->entries[]
2587                          */
2588                         /* FIXME */
2589                 }
2590         } else {
2591                 if (state & DS_FAULTY)
2592                         ddf->phys->entries[pd].state  |= __cpu_to_be16(DDF_Failed);
2593                 if (state & DS_INSYNC) {
2594                         ddf->phys->entries[pd].state  |= __cpu_to_be16(DDF_Online);
2595                         ddf->phys->entries[pd].state  &= __cpu_to_be16(~DDF_Rebuilding);
2596                 }
2597         }
2598
2599         /* Now we need to check the state of the array and update
2600          * virtual_disk.entries[n].state.
2601          * It needs to be one of "optimal", "degraded", "failed".
2602          * I don't understand 'deleted' or 'missing'.
2603          */
2604         working = 0;
2605         for (i=0; i < a->info.array.raid_disks; i++) {
2606                 pd = find_phys(ddf, vc->phys_refnum[i]);
2607                 if (pd < 0)
2608                         continue;
2609                 st = ddf->phys->entries[pd].state;
2610                 if ((state & (DDF_Online|DDF_Failed|DDF_Rebuilding))
2611                     == DDF_Online)
2612                         working++;
2613         }
2614         state = DDF_state_degraded;
2615         if (working == a->info.array.raid_disks)
2616                 state = DDF_state_optimal;
2617         else switch(vc->prl) {
2618         case DDF_RAID0:
2619         case DDF_CONCAT:
2620         case DDF_JBOD:
2621                 state = DDF_state_failed;
2622                 break;
2623         case DDF_RAID1:
2624                 if (working == 0)
2625                         state = DDF_state_failed;
2626                 break;
2627         case DDF_RAID4:
2628         case DDF_RAID5:
2629                 if (working < a->info.array.raid_disks-1)
2630                         state = DDF_state_failed;
2631                 break;
2632         case DDF_RAID6:
2633                 if (working < a->info.array.raid_disks-2)
2634                         state = DDF_state_failed;
2635                 else if (working == a->info.array.raid_disks-1)
2636                         state = DDF_state_part_optimal;
2637                 break;
2638         }
2639
2640         ddf->virt->entries[inst].state =
2641                 (ddf->virt->entries[inst].state & ~DDF_state_mask)
2642                 | state;
2643
2644         fprintf(stderr, "ddf: set_disk %d\n", n);
2645 }
2646
2647 static void ddf_sync_metadata(struct active_array *a)
2648 {
2649
2650         /*
2651          * Write all data to all devices.
2652          * Later, we might be able to track whether only local changes
2653          * have been made, or whether any global data has been changed,
2654          * but ddf is sufficiently weird that it probably always
2655          * changes global data ....
2656          */
2657         __write_init_super_ddf(a->container, 0);
2658         fprintf(stderr, "ddf: sync_metadata\n");
2659 }
2660
2661 struct superswitch super_ddf = {
2662 #ifndef MDASSEMBLE
2663         .examine_super  = examine_super_ddf,
2664         .brief_examine_super = brief_examine_super_ddf,
2665         .detail_super   = detail_super_ddf,
2666         .brief_detail_super = brief_detail_super_ddf,
2667         .validate_geometry = validate_geometry_ddf,
2668 #endif
2669         .match_home     = match_home_ddf,
2670         .uuid_from_super= uuid_from_super_ddf,
2671         .getinfo_super  = getinfo_super_ddf,
2672         .update_super   = update_super_ddf,
2673
2674         .avail_size     = avail_size_ddf,
2675
2676         .compare_super  = compare_super_ddf,
2677
2678         .load_super     = load_super_ddf,
2679         .init_super     = init_zero_ddf,
2680         .store_super    = store_zero_ddf,
2681         .free_super     = free_super_ddf,
2682         .match_metadata_desc = match_metadata_desc_ddf,
2683         .getinfo_super_n  = getinfo_super_n_container,
2684
2685
2686         .major          = 1000,
2687         .swapuuid       = 0,
2688         .external       = 1,
2689
2690 /* for mdmon */
2691         .open_new       = ddf_open_new,
2692         .set_array_state= ddf_set_array_state,
2693         .set_disk       = ddf_set_disk,
2694         .sync_metadata  = ddf_sync_metadata,
2695
2696
2697 };
2698
2699 /* Super_ddf_container is set by validate_geometry_ddf when given a
2700  * device that is not part of any array
2701  */
2702 struct superswitch super_ddf_container = {
2703 #ifndef MDASSEMBLE
2704         .validate_geometry = validate_geometry_ddf_container,
2705         .write_init_super = write_init_super_ddf,
2706 #endif
2707
2708         .load_super     = load_super_ddf,
2709         .init_super     = init_super_ddf,
2710         .add_to_super   = add_to_super_ddf,
2711         .getinfo_super  = getinfo_super_ddf,
2712
2713         .free_super     = free_super_ddf,
2714
2715         .container_content = container_content_ddf,
2716         .getinfo_super_n  = getinfo_super_n_container,
2717
2718         .major          = 1000,
2719         .swapuuid       = 0,
2720         .external       = 1,
2721 };
2722
2723 struct superswitch super_ddf_bvd = {
2724 #ifndef MDASSEMBLE
2725 //      .detail_super   = detail_super_ddf_bvd,
2726 //      .brief_detail_super = brief_detail_super_ddf_bvd,
2727         .validate_geometry = validate_geometry_ddf_bvd,
2728         .write_init_super = write_init_super_ddf,
2729 #endif
2730         .update_super   = update_super_ddf,
2731         .init_super     = init_super_ddf_bvd,
2732         .add_to_super   = add_to_super_ddf_bvd,
2733         .getinfo_super  = getinfo_super_ddf_bvd,
2734         .getinfo_super_n  = getinfo_super_n_bvd,
2735
2736         .load_super     = load_super_ddf,
2737         .free_super     = free_super_ddf,
2738         .match_metadata_desc = match_metadata_desc_ddf_bvd,
2739
2740
2741         .major          = 1001,
2742         .swapuuid       = 0,
2743         .external       = 2,
2744 };
2745
2746 struct superswitch super_ddf_svd = {
2747 #ifndef MDASSEMBLE
2748 //      .detail_super   = detail_super_ddf_svd,
2749 //      .brief_detail_super = brief_detail_super_ddf_svd,
2750         .validate_geometry = validate_geometry_ddf_svd,
2751 #endif
2752         .update_super   = update_super_ddf,
2753         .init_super     = init_super_ddf,
2754
2755         .load_super     = load_super_ddf,
2756         .free_super     = free_super_ddf,
2757         .match_metadata_desc = match_metadata_desc_ddf_svd,
2758
2759         .major          = 1002,
2760         .swapuuid       = 0,
2761         .external       = 2,
2762 };