]> git.ipfire.org Git - thirdparty/mdadm.git/blame - super-ddf.c
detail/wait: better handling of monitoring sync action.
[thirdparty/mdadm.git] / super-ddf.c
CommitLineData
a322f70c
DW
1/*
2 * mdadm - manage Linux "md" devices aka RAID arrays.
3 *
e736b623 4 * Copyright (C) 2006-2009 Neil Brown <neilb@suse.de>
a322f70c
DW
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"
549e9569 30#include "mdmon.h"
a322f70c
DW
31#include "sha1.h"
32#include <values.h>
33
a322f70c
DW
34/* a non-official T10 name for creation GUIDs */
35static char T10[] = "Linux-MD";
36
37/* DDF timestamps are 1980 based, so we need to add
38 * second-in-decade-of-seventies to convert to linux timestamps.
39 * 10 years with 2 leap years.
40 */
41#define DECADE (3600*24*(365*10+2))
42unsigned long crc32(
43 unsigned long crc,
44 const unsigned char *buf,
45 unsigned len);
46
47/* The DDF metadata handling.
48 * DDF metadata lives at the end of the device.
49 * The last 512 byte block provides an 'anchor' which is used to locate
50 * the rest of the metadata which usually lives immediately behind the anchor.
51 *
52 * Note:
53 * - all multibyte numeric fields are bigendian.
54 * - all strings are space padded.
55 *
56 */
57
58/* Primary Raid Level (PRL) */
59#define DDF_RAID0 0x00
60#define DDF_RAID1 0x01
61#define DDF_RAID3 0x03
62#define DDF_RAID4 0x04
63#define DDF_RAID5 0x05
64#define DDF_RAID1E 0x11
65#define DDF_JBOD 0x0f
66#define DDF_CONCAT 0x1f
67#define DDF_RAID5E 0x15
68#define DDF_RAID5EE 0x25
59e36268 69#define DDF_RAID6 0x06
a322f70c
DW
70
71/* Raid Level Qualifier (RLQ) */
72#define DDF_RAID0_SIMPLE 0x00
73#define DDF_RAID1_SIMPLE 0x00 /* just 2 devices in this plex */
74#define DDF_RAID1_MULTI 0x01 /* exactly 3 devices in this plex */
75#define DDF_RAID3_0 0x00 /* parity in first extent */
76#define DDF_RAID3_N 0x01 /* parity in last extent */
77#define DDF_RAID4_0 0x00 /* parity in first extent */
78#define DDF_RAID4_N 0x01 /* parity in last extent */
79/* these apply to raid5e and raid5ee as well */
80#define DDF_RAID5_0_RESTART 0x00 /* same as 'right asymmetric' - layout 1 */
59e36268 81#define DDF_RAID6_0_RESTART 0x01 /* raid6 different from raid5 here!!! */
a322f70c
DW
82#define DDF_RAID5_N_RESTART 0x02 /* same as 'left asymmetric' - layout 0 */
83#define DDF_RAID5_N_CONTINUE 0x03 /* same as 'left symmetric' - layout 2 */
84
85#define DDF_RAID1E_ADJACENT 0x00 /* raid10 nearcopies==2 */
86#define DDF_RAID1E_OFFSET 0x01 /* raid10 offsetcopies==2 */
87
88/* Secondary RAID Level (SRL) */
89#define DDF_2STRIPED 0x00 /* This is weirder than RAID0 !! */
90#define DDF_2MIRRORED 0x01
91#define DDF_2CONCAT 0x02
92#define DDF_2SPANNED 0x03 /* This is also weird - be careful */
93
94/* Magic numbers */
95#define DDF_HEADER_MAGIC __cpu_to_be32(0xDE11DE11)
96#define DDF_CONTROLLER_MAGIC __cpu_to_be32(0xAD111111)
97#define DDF_PHYS_RECORDS_MAGIC __cpu_to_be32(0x22222222)
98#define DDF_PHYS_DATA_MAGIC __cpu_to_be32(0x33333333)
99#define DDF_VIRT_RECORDS_MAGIC __cpu_to_be32(0xDDDDDDDD)
100#define DDF_VD_CONF_MAGIC __cpu_to_be32(0xEEEEEEEE)
101#define DDF_SPARE_ASSIGN_MAGIC __cpu_to_be32(0x55555555)
102#define DDF_VU_CONF_MAGIC __cpu_to_be32(0x88888888)
103#define DDF_VENDOR_LOG_MAGIC __cpu_to_be32(0x01dBEEF0)
104#define DDF_BBM_LOG_MAGIC __cpu_to_be32(0xABADB10C)
105
106#define DDF_GUID_LEN 24
59e36268
NB
107#define DDF_REVISION_0 "01.00.00"
108#define DDF_REVISION_2 "01.02.00"
a322f70c
DW
109
110struct ddf_header {
88c164f4 111 __u32 magic; /* DDF_HEADER_MAGIC */
a322f70c
DW
112 __u32 crc;
113 char guid[DDF_GUID_LEN];
59e36268 114 char revision[8]; /* 01.02.00 */
a322f70c
DW
115 __u32 seq; /* starts at '1' */
116 __u32 timestamp;
117 __u8 openflag;
118 __u8 foreignflag;
119 __u8 enforcegroups;
120 __u8 pad0; /* 0xff */
121 __u8 pad1[12]; /* 12 * 0xff */
122 /* 64 bytes so far */
123 __u8 header_ext[32]; /* reserved: fill with 0xff */
124 __u64 primary_lba;
125 __u64 secondary_lba;
126 __u8 type;
127 __u8 pad2[3]; /* 0xff */
128 __u32 workspace_len; /* sectors for vendor space -
129 * at least 32768(sectors) */
130 __u64 workspace_lba;
131 __u16 max_pd_entries; /* one of 15, 63, 255, 1023, 4095 */
132 __u16 max_vd_entries; /* 2^(4,6,8,10,12)-1 : i.e. as above */
133 __u16 max_partitions; /* i.e. max num of configuration
134 record entries per disk */
135 __u16 config_record_len; /* 1 +ROUNDUP(max_primary_element_entries
136 *12/512) */
137 __u16 max_primary_element_entries; /* 16, 64, 256, 1024, or 4096 */
138 __u8 pad3[54]; /* 0xff */
139 /* 192 bytes so far */
140 __u32 controller_section_offset;
141 __u32 controller_section_length;
142 __u32 phys_section_offset;
143 __u32 phys_section_length;
144 __u32 virt_section_offset;
145 __u32 virt_section_length;
146 __u32 config_section_offset;
147 __u32 config_section_length;
148 __u32 data_section_offset;
149 __u32 data_section_length;
150 __u32 bbm_section_offset;
151 __u32 bbm_section_length;
152 __u32 diag_space_offset;
153 __u32 diag_space_length;
154 __u32 vendor_offset;
155 __u32 vendor_length;
156 /* 256 bytes so far */
157 __u8 pad4[256]; /* 0xff */
158};
159
160/* type field */
161#define DDF_HEADER_ANCHOR 0x00
162#define DDF_HEADER_PRIMARY 0x01
163#define DDF_HEADER_SECONDARY 0x02
164
165/* The content of the 'controller section' - global scope */
166struct ddf_controller_data {
88c164f4 167 __u32 magic; /* DDF_CONTROLLER_MAGIC */
a322f70c
DW
168 __u32 crc;
169 char guid[DDF_GUID_LEN];
170 struct controller_type {
171 __u16 vendor_id;
172 __u16 device_id;
173 __u16 sub_vendor_id;
174 __u16 sub_device_id;
175 } type;
176 char product_id[16];
177 __u8 pad[8]; /* 0xff */
178 __u8 vendor_data[448];
179};
180
181/* The content of phys_section - global scope */
182struct phys_disk {
88c164f4 183 __u32 magic; /* DDF_PHYS_RECORDS_MAGIC */
a322f70c
DW
184 __u32 crc;
185 __u16 used_pdes;
186 __u16 max_pdes;
187 __u8 pad[52];
188 struct phys_disk_entry {
189 char guid[DDF_GUID_LEN];
190 __u32 refnum;
191 __u16 type;
192 __u16 state;
193 __u64 config_size; /* DDF structures must be after here */
194 char path[18]; /* another horrible structure really */
195 __u8 pad[6];
196 } entries[0];
197};
198
199/* phys_disk_entry.type is a bitmap - bigendian remember */
200#define DDF_Forced_PD_GUID 1
201#define DDF_Active_in_VD 2
88c164f4 202#define DDF_Global_Spare 4 /* VD_CONF records are ignored */
a322f70c
DW
203#define DDF_Spare 8 /* overrides Global_spare */
204#define DDF_Foreign 16
205#define DDF_Legacy 32 /* no DDF on this device */
206
207#define DDF_Interface_mask 0xf00
208#define DDF_Interface_SCSI 0x100
209#define DDF_Interface_SAS 0x200
210#define DDF_Interface_SATA 0x300
211#define DDF_Interface_FC 0x400
212
213/* phys_disk_entry.state is a bigendian bitmap */
214#define DDF_Online 1
215#define DDF_Failed 2 /* overrides 1,4,8 */
216#define DDF_Rebuilding 4
217#define DDF_Transition 8
218#define DDF_SMART 16
219#define DDF_ReadErrors 32
220#define DDF_Missing 64
221
222/* The content of the virt_section global scope */
223struct virtual_disk {
88c164f4 224 __u32 magic; /* DDF_VIRT_RECORDS_MAGIC */
a322f70c
DW
225 __u32 crc;
226 __u16 populated_vdes;
227 __u16 max_vdes;
228 __u8 pad[52];
229 struct virtual_entry {
230 char guid[DDF_GUID_LEN];
231 __u16 unit;
232 __u16 pad0; /* 0xffff */
233 __u16 guid_crc;
234 __u16 type;
235 __u8 state;
236 __u8 init_state;
237 __u8 pad1[14];
238 char name[16];
239 } entries[0];
240};
241
242/* virtual_entry.type is a bitmap - bigendian */
243#define DDF_Shared 1
244#define DDF_Enforce_Groups 2
245#define DDF_Unicode 4
246#define DDF_Owner_Valid 8
247
248/* virtual_entry.state is a bigendian bitmap */
249#define DDF_state_mask 0x7
250#define DDF_state_optimal 0x0
251#define DDF_state_degraded 0x1
252#define DDF_state_deleted 0x2
253#define DDF_state_missing 0x3
254#define DDF_state_failed 0x4
7a7cc504 255#define DDF_state_part_optimal 0x5
a322f70c
DW
256
257#define DDF_state_morphing 0x8
258#define DDF_state_inconsistent 0x10
259
260/* virtual_entry.init_state is a bigendian bitmap */
261#define DDF_initstate_mask 0x03
262#define DDF_init_not 0x00
7a7cc504
NB
263#define DDF_init_quick 0x01 /* initialisation is progress.
264 * i.e. 'state_inconsistent' */
a322f70c
DW
265#define DDF_init_full 0x02
266
267#define DDF_access_mask 0xc0
268#define DDF_access_rw 0x00
269#define DDF_access_ro 0x80
270#define DDF_access_blocked 0xc0
271
272/* The content of the config_section - local scope
273 * It has multiple records each config_record_len sectors
274 * They can be vd_config or spare_assign
275 */
276
277struct vd_config {
88c164f4 278 __u32 magic; /* DDF_VD_CONF_MAGIC */
a322f70c
DW
279 __u32 crc;
280 char guid[DDF_GUID_LEN];
281 __u32 timestamp;
282 __u32 seqnum;
283 __u8 pad0[24];
284 __u16 prim_elmnt_count;
285 __u8 chunk_shift; /* 0 == 512, 1==1024 etc */
286 __u8 prl;
287 __u8 rlq;
288 __u8 sec_elmnt_count;
289 __u8 sec_elmnt_seq;
290 __u8 srl;
598f0d58
NB
291 __u64 blocks; /* blocks per component could be different
292 * on different component devices...(only
293 * for concat I hope) */
294 __u64 array_blocks; /* blocks in array */
a322f70c
DW
295 __u8 pad1[8];
296 __u32 spare_refs[8];
297 __u8 cache_pol[8];
298 __u8 bg_rate;
299 __u8 pad2[3];
300 __u8 pad3[52];
301 __u8 pad4[192];
302 __u8 v0[32]; /* reserved- 0xff */
303 __u8 v1[32]; /* reserved- 0xff */
304 __u8 v2[16]; /* reserved- 0xff */
305 __u8 v3[16]; /* reserved- 0xff */
306 __u8 vendor[32];
307 __u32 phys_refnum[0]; /* refnum of each disk in sequence */
308 /*__u64 lba_offset[0]; LBA offset in each phys. Note extents in a
309 bvd are always the same size */
310};
311
312/* vd_config.cache_pol[7] is a bitmap */
313#define DDF_cache_writeback 1 /* else writethrough */
314#define DDF_cache_wadaptive 2 /* only applies if writeback */
315#define DDF_cache_readahead 4
316#define DDF_cache_radaptive 8 /* only if doing read-ahead */
317#define DDF_cache_ifnobatt 16 /* even to write cache if battery is poor */
318#define DDF_cache_wallowed 32 /* enable write caching */
319#define DDF_cache_rallowed 64 /* enable read caching */
320
321struct spare_assign {
88c164f4 322 __u32 magic; /* DDF_SPARE_ASSIGN_MAGIC */
a322f70c
DW
323 __u32 crc;
324 __u32 timestamp;
325 __u8 reserved[7];
326 __u8 type;
327 __u16 populated; /* SAEs used */
328 __u16 max; /* max SAEs */
329 __u8 pad[8];
330 struct spare_assign_entry {
331 char guid[DDF_GUID_LEN];
332 __u16 secondary_element;
333 __u8 pad[6];
334 } spare_ents[0];
335};
336/* spare_assign.type is a bitmap */
337#define DDF_spare_dedicated 0x1 /* else global */
338#define DDF_spare_revertible 0x2 /* else committable */
339#define DDF_spare_active 0x4 /* else not active */
340#define DDF_spare_affinity 0x8 /* enclosure affinity */
341
342/* The data_section contents - local scope */
343struct disk_data {
88c164f4 344 __u32 magic; /* DDF_PHYS_DATA_MAGIC */
a322f70c
DW
345 __u32 crc;
346 char guid[DDF_GUID_LEN];
347 __u32 refnum; /* crc of some magic drive data ... */
348 __u8 forced_ref; /* set when above was not result of magic */
349 __u8 forced_guid; /* set if guid was forced rather than magic */
350 __u8 vendor[32];
351 __u8 pad[442];
352};
353
354/* bbm_section content */
355struct bad_block_log {
356 __u32 magic;
357 __u32 crc;
358 __u16 entry_count;
359 __u32 spare_count;
360 __u8 pad[10];
361 __u64 first_spare;
362 struct mapped_block {
363 __u64 defective_start;
364 __u32 replacement_start;
365 __u16 remap_count;
366 __u8 pad[2];
367 } entries[0];
368};
369
370/* Struct for internally holding ddf structures */
371/* The DDF structure stored on each device is potentially
372 * quite different, as some data is global and some is local.
373 * The global data is:
374 * - ddf header
375 * - controller_data
376 * - Physical disk records
377 * - Virtual disk records
378 * The local data is:
379 * - Configuration records
380 * - Physical Disk data section
381 * ( and Bad block and vendor which I don't care about yet).
382 *
383 * The local data is parsed into separate lists as it is read
384 * and reconstructed for writing. This means that we only need
385 * to make config changes once and they are automatically
386 * propagated to all devices.
387 * Note that the ddf_super has space of the conf and disk data
388 * for this disk and also for a list of all such data.
389 * The list is only used for the superblock that is being
390 * built in Create or Assemble to describe the whole array.
391 */
392struct ddf_super {
6416d527 393 struct ddf_header anchor, primary, secondary;
a322f70c 394 struct ddf_controller_data controller;
6416d527 395 struct ddf_header *active;
a322f70c
DW
396 struct phys_disk *phys;
397 struct virtual_disk *virt;
398 int pdsize, vdsize;
f21e18ca 399 unsigned int max_part, mppe, conf_rec_len;
d2ca6449 400 int currentdev;
18a2f463 401 int updates_pending;
a322f70c 402 struct vcl {
6416d527
NB
403 union {
404 char space[512];
405 struct {
406 struct vcl *next;
407 __u64 *lba_offset; /* location in 'conf' of
408 * the lba table */
f21e18ca 409 unsigned int vcnum; /* index into ->virt */
6416d527
NB
410 __u64 *block_sizes; /* NULL if all the same */
411 };
412 };
a322f70c 413 struct vd_config conf;
d2ca6449 414 } *conflist, *currentconf;
a322f70c 415 struct dl {
6416d527
NB
416 union {
417 char space[512];
418 struct {
419 struct dl *next;
420 int major, minor;
421 char *devname;
422 int fd;
423 unsigned long long size; /* sectors */
424 int pdnum; /* index in ->phys */
425 struct spare_assign *spare;
8592f29d
N
426 void *mdupdate; /* hold metadata update */
427
428 /* These fields used by auto-layout */
429 int raiddisk; /* slot to fill in autolayout */
430 __u64 esize;
6416d527
NB
431 };
432 };
a322f70c 433 struct disk_data disk;
b2280677 434 struct vcl *vlist[0]; /* max_part in size */
2cc2983d 435 } *dlist, *add_list;
a322f70c
DW
436};
437
438#ifndef offsetof
439#define offsetof(t,f) ((size_t)&(((t*)0)->f))
440#endif
441
a322f70c 442
f21e18ca 443static unsigned int calc_crc(void *buf, int len)
a322f70c
DW
444{
445 /* crcs are always at the same place as in the ddf_header */
446 struct ddf_header *ddf = buf;
447 __u32 oldcrc = ddf->crc;
448 __u32 newcrc;
449 ddf->crc = 0xffffffff;
450
451 newcrc = crc32(0, buf, len);
452 ddf->crc = oldcrc;
4abe6b70
N
453 /* The crc is store (like everything) bigendian, so convert
454 * here for simplicity
455 */
456 return __cpu_to_be32(newcrc);
a322f70c
DW
457}
458
459static int load_ddf_header(int fd, unsigned long long lba,
460 unsigned long long size,
461 int type,
462 struct ddf_header *hdr, struct ddf_header *anchor)
463{
464 /* read a ddf header (primary or secondary) from fd/lba
465 * and check that it is consistent with anchor
466 * Need to check:
467 * magic, crc, guid, rev, and LBA's header_type, and
468 * everything after header_type must be the same
469 */
470 if (lba >= size-1)
471 return 0;
472
473 if (lseek64(fd, lba<<9, 0) < 0)
474 return 0;
475
476 if (read(fd, hdr, 512) != 512)
477 return 0;
478
479 if (hdr->magic != DDF_HEADER_MAGIC)
480 return 0;
481 if (calc_crc(hdr, 512) != hdr->crc)
482 return 0;
483 if (memcmp(anchor->guid, hdr->guid, DDF_GUID_LEN) != 0 ||
484 memcmp(anchor->revision, hdr->revision, 8) != 0 ||
485 anchor->primary_lba != hdr->primary_lba ||
486 anchor->secondary_lba != hdr->secondary_lba ||
487 hdr->type != type ||
488 memcmp(anchor->pad2, hdr->pad2, 512 -
489 offsetof(struct ddf_header, pad2)) != 0)
490 return 0;
491
492 /* Looks good enough to me... */
493 return 1;
494}
495
496static void *load_section(int fd, struct ddf_super *super, void *buf,
497 __u32 offset_be, __u32 len_be, int check)
498{
499 unsigned long long offset = __be32_to_cpu(offset_be);
500 unsigned long long len = __be32_to_cpu(len_be);
501 int dofree = (buf == NULL);
502
503 if (check)
504 if (len != 2 && len != 8 && len != 32
505 && len != 128 && len != 512)
506 return NULL;
507
508 if (len > 1024)
509 return NULL;
510 if (buf) {
511 /* All pre-allocated sections are a single block */
512 if (len != 1)
513 return NULL;
3d2c4fc7
DW
514 } else if (posix_memalign(&buf, 512, len<<9) != 0)
515 buf = NULL;
6416d527 516
a322f70c
DW
517 if (!buf)
518 return NULL;
519
520 if (super->active->type == 1)
521 offset += __be64_to_cpu(super->active->primary_lba);
522 else
523 offset += __be64_to_cpu(super->active->secondary_lba);
524
f21e18ca 525 if ((unsigned long long)lseek64(fd, offset<<9, 0) != (offset<<9)) {
a322f70c
DW
526 if (dofree)
527 free(buf);
528 return NULL;
529 }
f21e18ca 530 if ((unsigned long long)read(fd, buf, len<<9) != (len<<9)) {
a322f70c
DW
531 if (dofree)
532 free(buf);
533 return NULL;
534 }
535 return buf;
536}
537
538static int load_ddf_headers(int fd, struct ddf_super *super, char *devname)
539{
540 unsigned long long dsize;
541
542 get_dev_size(fd, NULL, &dsize);
543
544 if (lseek64(fd, dsize-512, 0) < 0) {
545 if (devname)
546 fprintf(stderr,
547 Name": Cannot seek to anchor block on %s: %s\n",
548 devname, strerror(errno));
549 return 1;
550 }
551 if (read(fd, &super->anchor, 512) != 512) {
552 if (devname)
553 fprintf(stderr,
554 Name ": Cannot read anchor block on %s: %s\n",
555 devname, strerror(errno));
556 return 1;
557 }
558 if (super->anchor.magic != DDF_HEADER_MAGIC) {
559 if (devname)
560 fprintf(stderr, Name ": no DDF anchor found on %s\n",
561 devname);
562 return 2;
563 }
564 if (calc_crc(&super->anchor, 512) != super->anchor.crc) {
565 if (devname)
566 fprintf(stderr, Name ": bad CRC on anchor on %s\n",
567 devname);
568 return 2;
569 }
59e36268
NB
570 if (memcmp(super->anchor.revision, DDF_REVISION_0, 8) != 0 &&
571 memcmp(super->anchor.revision, DDF_REVISION_2, 8) != 0) {
a322f70c
DW
572 if (devname)
573 fprintf(stderr, Name ": can only support super revision"
59e36268
NB
574 " %.8s and earlier, not %.8s on %s\n",
575 DDF_REVISION_2, super->anchor.revision,devname);
a322f70c
DW
576 return 2;
577 }
578 if (load_ddf_header(fd, __be64_to_cpu(super->anchor.primary_lba),
579 dsize >> 9, 1,
580 &super->primary, &super->anchor) == 0) {
581 if (devname)
582 fprintf(stderr,
583 Name ": Failed to load primary DDF header "
584 "on %s\n", devname);
585 return 2;
586 }
587 super->active = &super->primary;
588 if (load_ddf_header(fd, __be64_to_cpu(super->anchor.secondary_lba),
589 dsize >> 9, 2,
590 &super->secondary, &super->anchor)) {
591 if ((__be32_to_cpu(super->primary.seq)
592 < __be32_to_cpu(super->secondary.seq) &&
593 !super->secondary.openflag)
594 || (__be32_to_cpu(super->primary.seq)
595 == __be32_to_cpu(super->secondary.seq) &&
596 super->primary.openflag && !super->secondary.openflag)
597 )
598 super->active = &super->secondary;
599 }
600 return 0;
601}
602
603static int load_ddf_global(int fd, struct ddf_super *super, char *devname)
604{
605 void *ok;
606 ok = load_section(fd, super, &super->controller,
607 super->active->controller_section_offset,
608 super->active->controller_section_length,
609 0);
610 super->phys = load_section(fd, super, NULL,
611 super->active->phys_section_offset,
612 super->active->phys_section_length,
613 1);
614 super->pdsize = __be32_to_cpu(super->active->phys_section_length) * 512;
615
616 super->virt = load_section(fd, super, NULL,
617 super->active->virt_section_offset,
618 super->active->virt_section_length,
619 1);
620 super->vdsize = __be32_to_cpu(super->active->virt_section_length) * 512;
621 if (!ok ||
622 !super->phys ||
623 !super->virt) {
624 free(super->phys);
625 free(super->virt);
a2349791
NB
626 super->phys = NULL;
627 super->virt = NULL;
a322f70c
DW
628 return 2;
629 }
630 super->conflist = NULL;
631 super->dlist = NULL;
8c3b8c2c
NB
632
633 super->max_part = __be16_to_cpu(super->active->max_partitions);
634 super->mppe = __be16_to_cpu(super->active->max_primary_element_entries);
635 super->conf_rec_len = __be16_to_cpu(super->active->config_record_len);
a322f70c
DW
636 return 0;
637}
638
639static int load_ddf_local(int fd, struct ddf_super *super,
640 char *devname, int keep)
641{
642 struct dl *dl;
643 struct stat stb;
644 char *conf;
f21e18ca
N
645 unsigned int i;
646 unsigned int confsec;
b2280677 647 int vnum;
f21e18ca 648 unsigned int max_virt_disks = __be16_to_cpu(super->active->max_vd_entries);
d2ca6449 649 unsigned long long dsize;
a322f70c
DW
650
651 /* First the local disk info */
3d2c4fc7 652 if (posix_memalign((void**)&dl, 512,
6416d527 653 sizeof(*dl) +
3d2c4fc7
DW
654 (super->max_part) * sizeof(dl->vlist[0])) != 0) {
655 fprintf(stderr, Name ": %s could not allocate disk info buffer\n",
656 __func__);
657 return 1;
658 }
a322f70c
DW
659
660 load_section(fd, super, &dl->disk,
661 super->active->data_section_offset,
662 super->active->data_section_length,
663 0);
664 dl->devname = devname ? strdup(devname) : NULL;
598f0d58 665
a322f70c
DW
666 fstat(fd, &stb);
667 dl->major = major(stb.st_rdev);
668 dl->minor = minor(stb.st_rdev);
669 dl->next = super->dlist;
670 dl->fd = keep ? fd : -1;
d2ca6449
NB
671
672 dl->size = 0;
673 if (get_dev_size(fd, devname, &dsize))
674 dl->size = dsize >> 9;
b2280677 675 dl->spare = NULL;
f21e18ca 676 for (i = 0 ; i < super->max_part ; i++)
a322f70c
DW
677 dl->vlist[i] = NULL;
678 super->dlist = dl;
59e36268 679 dl->pdnum = -1;
f21e18ca 680 for (i = 0; i < __be16_to_cpu(super->active->max_pd_entries); i++)
5575e7d9
NB
681 if (memcmp(super->phys->entries[i].guid,
682 dl->disk.guid, DDF_GUID_LEN) == 0)
683 dl->pdnum = i;
684
a322f70c
DW
685 /* Now the config list. */
686 /* 'conf' is an array of config entries, some of which are
687 * probably invalid. Those which are good need to be copied into
688 * the conflist
689 */
a322f70c
DW
690
691 conf = load_section(fd, super, NULL,
692 super->active->config_section_offset,
693 super->active->config_section_length,
694 0);
695
b2280677 696 vnum = 0;
e223334f
N
697 for (confsec = 0;
698 confsec < __be32_to_cpu(super->active->config_section_length);
699 confsec += super->conf_rec_len) {
a322f70c 700 struct vd_config *vd =
e223334f 701 (struct vd_config *)((char*)conf + confsec*512);
a322f70c
DW
702 struct vcl *vcl;
703
b2280677
NB
704 if (vd->magic == DDF_SPARE_ASSIGN_MAGIC) {
705 if (dl->spare)
706 continue;
3d2c4fc7
DW
707 if (posix_memalign((void**)&dl->spare, 512,
708 super->conf_rec_len*512) != 0) {
709 fprintf(stderr, Name
710 ": %s could not allocate spare info buf\n",
711 __func__);
712 return 1;
713 }
714
b2280677
NB
715 memcpy(dl->spare, vd, super->conf_rec_len*512);
716 continue;
717 }
a322f70c
DW
718 if (vd->magic != DDF_VD_CONF_MAGIC)
719 continue;
720 for (vcl = super->conflist; vcl; vcl = vcl->next) {
721 if (memcmp(vcl->conf.guid,
722 vd->guid, DDF_GUID_LEN) == 0)
723 break;
724 }
725
726 if (vcl) {
b2280677 727 dl->vlist[vnum++] = vcl;
a322f70c
DW
728 if (__be32_to_cpu(vd->seqnum) <=
729 __be32_to_cpu(vcl->conf.seqnum))
730 continue;
59e36268 731 } else {
3d2c4fc7 732 if (posix_memalign((void**)&vcl, 512,
6416d527 733 (super->conf_rec_len*512 +
3d2c4fc7
DW
734 offsetof(struct vcl, conf))) != 0) {
735 fprintf(stderr, Name
736 ": %s could not allocate vcl buf\n",
737 __func__);
738 return 1;
739 }
a322f70c 740 vcl->next = super->conflist;
59e36268 741 vcl->block_sizes = NULL; /* FIXME not for CONCAT */
a322f70c 742 super->conflist = vcl;
b2280677 743 dl->vlist[vnum++] = vcl;
a322f70c 744 }
8c3b8c2c 745 memcpy(&vcl->conf, vd, super->conf_rec_len*512);
a322f70c 746 vcl->lba_offset = (__u64*)
8c3b8c2c 747 &vcl->conf.phys_refnum[super->mppe];
59e36268
NB
748
749 for (i=0; i < max_virt_disks ; i++)
750 if (memcmp(super->virt->entries[i].guid,
751 vcl->conf.guid, DDF_GUID_LEN)==0)
752 break;
753 if (i < max_virt_disks)
754 vcl->vcnum = i;
a322f70c
DW
755 }
756 free(conf);
757
758 return 0;
759}
760
761#ifndef MDASSEMBLE
762static int load_super_ddf_all(struct supertype *st, int fd,
763 void **sbp, char *devname, int keep_fd);
764#endif
37424f13
DW
765
766static void free_super_ddf(struct supertype *st);
767
a322f70c
DW
768static int load_super_ddf(struct supertype *st, int fd,
769 char *devname)
770{
771 unsigned long long dsize;
772 struct ddf_super *super;
773 int rv;
774
775#ifndef MDASSEMBLE
59e36268 776 /* if 'fd' is a container, load metadata from all the devices */
3dbccbcf 777 if (load_super_ddf_all(st, fd, &st->sb, devname, 1) == 0)
a322f70c
DW
778 return 0;
779#endif
f7e7067b
NB
780 if (st->subarray[0])
781 return 1; /* FIXME Is this correct */
a322f70c
DW
782
783 if (get_dev_size(fd, devname, &dsize) == 0)
784 return 1;
785
691c6ee1
N
786 if (test_partition(fd))
787 /* DDF is not allowed on partitions */
788 return 1;
789
a322f70c
DW
790 /* 32M is a lower bound */
791 if (dsize <= 32*1024*1024) {
97320d7c 792 if (devname)
a322f70c
DW
793 fprintf(stderr,
794 Name ": %s is too small for ddf: "
795 "size is %llu sectors.\n",
796 devname, dsize>>9);
97320d7c 797 return 1;
a322f70c
DW
798 }
799 if (dsize & 511) {
97320d7c 800 if (devname)
a322f70c
DW
801 fprintf(stderr,
802 Name ": %s is an odd size for ddf: "
803 "size is %llu bytes.\n",
804 devname, dsize);
97320d7c 805 return 1;
a322f70c
DW
806 }
807
37424f13
DW
808 free_super_ddf(st);
809
6416d527 810 if (posix_memalign((void**)&super, 512, sizeof(*super))!= 0) {
a322f70c
DW
811 fprintf(stderr, Name ": malloc of %zu failed.\n",
812 sizeof(*super));
813 return 1;
814 }
a2349791 815 memset(super, 0, sizeof(*super));
a322f70c
DW
816
817 rv = load_ddf_headers(fd, super, devname);
818 if (rv) {
819 free(super);
820 return rv;
821 }
822
823 /* Have valid headers and have chosen the best. Let's read in the rest*/
824
825 rv = load_ddf_global(fd, super, devname);
826
827 if (rv) {
828 if (devname)
829 fprintf(stderr,
830 Name ": Failed to load all information "
831 "sections on %s\n", devname);
832 free(super);
833 return rv;
834 }
835
3d2c4fc7
DW
836 rv = load_ddf_local(fd, super, devname, 0);
837
838 if (rv) {
839 if (devname)
840 fprintf(stderr,
841 Name ": Failed to load all information "
842 "sections on %s\n", devname);
843 free(super);
844 return rv;
845 }
a322f70c 846
af99d9ca 847 if (st->subarray[0]) {
33414a01 848 unsigned long val;
af99d9ca 849 struct vcl *v;
33414a01
DW
850 char *ep;
851
852 val = strtoul(st->subarray, &ep, 10);
853 if (*ep != '\0') {
854 free(super);
855 return 1;
856 }
af99d9ca
DW
857
858 for (v = super->conflist; v; v = v->next)
33414a01 859 if (v->vcnum == val)
af99d9ca
DW
860 super->currentconf = v;
861 if (!super->currentconf) {
862 free(super);
863 return 1;
864 }
865 }
866
a322f70c
DW
867 /* Should possibly check the sections .... */
868
869 st->sb = super;
870 if (st->ss == NULL) {
871 st->ss = &super_ddf;
872 st->minor_version = 0;
873 st->max_devs = 512;
874 }
352452c3 875 st->loaded_container = 0;
a322f70c
DW
876 return 0;
877
878}
879
880static void free_super_ddf(struct supertype *st)
881{
882 struct ddf_super *ddf = st->sb;
883 if (ddf == NULL)
884 return;
885 free(ddf->phys);
886 free(ddf->virt);
887 while (ddf->conflist) {
888 struct vcl *v = ddf->conflist;
889 ddf->conflist = v->next;
59e36268
NB
890 if (v->block_sizes)
891 free(v->block_sizes);
a322f70c
DW
892 free(v);
893 }
894 while (ddf->dlist) {
895 struct dl *d = ddf->dlist;
896 ddf->dlist = d->next;
897 if (d->fd >= 0)
898 close(d->fd);
b2280677
NB
899 if (d->spare)
900 free(d->spare);
a322f70c
DW
901 free(d);
902 }
903 free(ddf);
904 st->sb = NULL;
905}
906
907static struct supertype *match_metadata_desc_ddf(char *arg)
908{
909 /* 'ddf' only support containers */
910 struct supertype *st;
911 if (strcmp(arg, "ddf") != 0 &&
912 strcmp(arg, "default") != 0
913 )
914 return NULL;
915
916 st = malloc(sizeof(*st));
ef609477 917 memset(st, 0, sizeof(*st));
a322f70c
DW
918 st->ss = &super_ddf;
919 st->max_devs = 512;
920 st->minor_version = 0;
921 st->sb = NULL;
922 return st;
923}
924
a322f70c
DW
925
926#ifndef MDASSEMBLE
927
928static mapping_t ddf_state[] = {
929 { "Optimal", 0},
930 { "Degraded", 1},
931 { "Deleted", 2},
932 { "Missing", 3},
933 { "Failed", 4},
934 { "Partially Optimal", 5},
935 { "-reserved-", 6},
936 { "-reserved-", 7},
937 { NULL, 0}
938};
939
940static mapping_t ddf_init_state[] = {
941 { "Not Initialised", 0},
942 { "QuickInit in Progress", 1},
943 { "Fully Initialised", 2},
944 { "*UNKNOWN*", 3},
945 { NULL, 0}
946};
947static mapping_t ddf_access[] = {
948 { "Read/Write", 0},
949 { "Reserved", 1},
950 { "Read Only", 2},
951 { "Blocked (no access)", 3},
952 { NULL ,0}
953};
954
955static mapping_t ddf_level[] = {
956 { "RAID0", DDF_RAID0},
957 { "RAID1", DDF_RAID1},
958 { "RAID3", DDF_RAID3},
959 { "RAID4", DDF_RAID4},
960 { "RAID5", DDF_RAID5},
961 { "RAID1E",DDF_RAID1E},
962 { "JBOD", DDF_JBOD},
963 { "CONCAT",DDF_CONCAT},
964 { "RAID5E",DDF_RAID5E},
965 { "RAID5EE",DDF_RAID5EE},
966 { "RAID6", DDF_RAID6},
967 { NULL, 0}
968};
969static mapping_t ddf_sec_level[] = {
970 { "Striped", DDF_2STRIPED},
971 { "Mirrored", DDF_2MIRRORED},
972 { "Concat", DDF_2CONCAT},
973 { "Spanned", DDF_2SPANNED},
974 { NULL, 0}
975};
976#endif
977
978struct num_mapping {
979 int num1, num2;
980};
981static struct num_mapping ddf_level_num[] = {
982 { DDF_RAID0, 0 },
983 { DDF_RAID1, 1 },
984 { DDF_RAID3, LEVEL_UNSUPPORTED },
60f18132
NB
985 { DDF_RAID4, 4 },
986 { DDF_RAID5, 5 },
a322f70c
DW
987 { DDF_RAID1E, LEVEL_UNSUPPORTED },
988 { DDF_JBOD, LEVEL_UNSUPPORTED },
989 { DDF_CONCAT, LEVEL_LINEAR },
990 { DDF_RAID5E, LEVEL_UNSUPPORTED },
991 { DDF_RAID5EE, LEVEL_UNSUPPORTED },
992 { DDF_RAID6, 6},
993 { MAXINT, MAXINT }
994};
995
996static int map_num1(struct num_mapping *map, int num)
997{
998 int i;
999 for (i=0 ; map[i].num1 != MAXINT; i++)
1000 if (map[i].num1 == num)
1001 break;
1002 return map[i].num2;
1003}
1004
42dc2744
N
1005static int all_ff(char *guid)
1006{
1007 int i;
1008 for (i = 0; i < DDF_GUID_LEN; i++)
1009 if (guid[i] != (char)0xff)
1010 return 0;
1011 return 1;
1012}
1013
a322f70c
DW
1014#ifndef MDASSEMBLE
1015static void print_guid(char *guid, int tstamp)
1016{
1017 /* A GUIDs are part (or all) ASCII and part binary.
1018 * They tend to be space padded.
59e36268
NB
1019 * We print the GUID in HEX, then in parentheses add
1020 * any initial ASCII sequence, and a possible
1021 * time stamp from bytes 16-19
a322f70c
DW
1022 */
1023 int l = DDF_GUID_LEN;
1024 int i;
59e36268
NB
1025
1026 for (i=0 ; i<DDF_GUID_LEN ; i++) {
1027 if ((i&3)==0 && i != 0) printf(":");
1028 printf("%02X", guid[i]&255);
1029 }
1030
cfccea8c 1031 printf("\n (");
a322f70c
DW
1032 while (l && guid[l-1] == ' ')
1033 l--;
1034 for (i=0 ; i<l ; i++) {
1035 if (guid[i] >= 0x20 && guid[i] < 0x7f)
1036 fputc(guid[i], stdout);
1037 else
59e36268 1038 break;
a322f70c
DW
1039 }
1040 if (tstamp) {
1041 time_t then = __be32_to_cpu(*(__u32*)(guid+16)) + DECADE;
1042 char tbuf[100];
1043 struct tm *tm;
1044 tm = localtime(&then);
59e36268 1045 strftime(tbuf, 100, " %D %T",tm);
a322f70c
DW
1046 fputs(tbuf, stdout);
1047 }
59e36268 1048 printf(")");
a322f70c
DW
1049}
1050
1051static void examine_vd(int n, struct ddf_super *sb, char *guid)
1052{
8c3b8c2c 1053 int crl = sb->conf_rec_len;
a322f70c
DW
1054 struct vcl *vcl;
1055
1056 for (vcl = sb->conflist ; vcl ; vcl = vcl->next) {
f21e18ca 1057 unsigned int i;
a322f70c
DW
1058 struct vd_config *vc = &vcl->conf;
1059
1060 if (calc_crc(vc, crl*512) != vc->crc)
1061 continue;
1062 if (memcmp(vc->guid, guid, DDF_GUID_LEN) != 0)
1063 continue;
1064
1065 /* Ok, we know about this VD, let's give more details */
b06e3095 1066 printf(" Raid Devices[%d] : %d (", n,
a322f70c 1067 __be16_to_cpu(vc->prim_elmnt_count));
f21e18ca 1068 for (i = 0; i < __be16_to_cpu(vc->prim_elmnt_count); i++) {
b06e3095
N
1069 int j;
1070 int cnt = __be16_to_cpu(sb->phys->used_pdes);
1071 for (j=0; j<cnt; j++)
1072 if (vc->phys_refnum[i] == sb->phys->entries[j].refnum)
1073 break;
1074 if (i) printf(" ");
1075 if (j < cnt)
1076 printf("%d", j);
1077 else
1078 printf("--");
1079 }
1080 printf(")\n");
1081 if (vc->chunk_shift != 255)
a322f70c
DW
1082 printf(" Chunk Size[%d] : %d sectors\n", n,
1083 1 << vc->chunk_shift);
1084 printf(" Raid Level[%d] : %s\n", n,
1085 map_num(ddf_level, vc->prl)?:"-unknown-");
1086 if (vc->sec_elmnt_count != 1) {
1087 printf(" Secondary Position[%d] : %d of %d\n", n,
1088 vc->sec_elmnt_seq, vc->sec_elmnt_count);
1089 printf(" Secondary Level[%d] : %s\n", n,
1090 map_num(ddf_sec_level, vc->srl) ?: "-unknown-");
1091 }
1092 printf(" Device Size[%d] : %llu\n", n,
c9b6907b 1093 (unsigned long long)__be64_to_cpu(vc->blocks)/2);
a322f70c 1094 printf(" Array Size[%d] : %llu\n", n,
c9b6907b 1095 (unsigned long long)__be64_to_cpu(vc->array_blocks)/2);
a322f70c
DW
1096 }
1097}
1098
1099static void examine_vds(struct ddf_super *sb)
1100{
1101 int cnt = __be16_to_cpu(sb->virt->populated_vdes);
1102 int i;
1103 printf(" Virtual Disks : %d\n", cnt);
1104
1105 for (i=0; i<cnt; i++) {
1106 struct virtual_entry *ve = &sb->virt->entries[i];
b06e3095 1107 printf("\n");
a322f70c
DW
1108 printf(" VD GUID[%d] : ", i); print_guid(ve->guid, 1);
1109 printf("\n");
1110 printf(" unit[%d] : %d\n", i, __be16_to_cpu(ve->unit));
1111 printf(" state[%d] : %s, %s%s\n", i,
1112 map_num(ddf_state, ve->state & 7),
1113 (ve->state & 8) ? "Morphing, ": "",
1114 (ve->state & 16)? "Not Consistent" : "Consistent");
1115 printf(" init state[%d] : %s\n", i,
1116 map_num(ddf_init_state, ve->init_state&3));
1117 printf(" access[%d] : %s\n", i,
1118 map_num(ddf_access, (ve->init_state>>6) & 3));
1119 printf(" Name[%d] : %.16s\n", i, ve->name);
1120 examine_vd(i, sb, ve->guid);
1121 }
1122 if (cnt) printf("\n");
1123}
1124
1125static void examine_pds(struct ddf_super *sb)
1126{
1127 int cnt = __be16_to_cpu(sb->phys->used_pdes);
1128 int i;
1129 struct dl *dl;
1130 printf(" Physical Disks : %d\n", cnt);
962371a5 1131 printf(" Number RefNo Size Device Type/State\n");
a322f70c
DW
1132
1133 for (i=0 ; i<cnt ; i++) {
1134 struct phys_disk_entry *pd = &sb->phys->entries[i];
1135 int type = __be16_to_cpu(pd->type);
1136 int state = __be16_to_cpu(pd->state);
1137
b06e3095
N
1138 //printf(" PD GUID[%d] : ", i); print_guid(pd->guid, 0);
1139 //printf("\n");
1140 printf(" %3d %08x ", i,
a322f70c 1141 __be32_to_cpu(pd->refnum));
c9b6907b
DL
1142 printf("%8lluK ",
1143 (unsigned long long)__be64_to_cpu(pd->config_size)>>1);
b06e3095
N
1144 for (dl = sb->dlist; dl ; dl = dl->next) {
1145 if (dl->disk.refnum == pd->refnum) {
1146 char *dv = map_dev(dl->major, dl->minor, 0);
1147 if (dv) {
962371a5 1148 printf("%-15s", dv);
b06e3095
N
1149 break;
1150 }
1151 }
1152 }
1153 if (!dl)
962371a5 1154 printf("%15s","");
b06e3095 1155 printf(" %s%s%s%s%s",
a322f70c 1156 (type&2) ? "active":"",
b06e3095 1157 (type&4) ? "Global-Spare":"",
a322f70c
DW
1158 (type&8) ? "spare" : "",
1159 (type&16)? ", foreign" : "",
1160 (type&32)? "pass-through" : "");
b06e3095 1161 printf("/%s%s%s%s%s%s%s",
a322f70c
DW
1162 (state&1)? "Online": "Offline",
1163 (state&2)? ", Failed": "",
1164 (state&4)? ", Rebuilding": "",
1165 (state&8)? ", in-transition": "",
b06e3095
N
1166 (state&16)? ", SMART-errors": "",
1167 (state&32)? ", Unrecovered-Read-Errors": "",
a322f70c 1168 (state&64)? ", Missing" : "");
a322f70c
DW
1169 printf("\n");
1170 }
1171}
1172
1173static void examine_super_ddf(struct supertype *st, char *homehost)
1174{
1175 struct ddf_super *sb = st->sb;
1176
1177 printf(" Magic : %08x\n", __be32_to_cpu(sb->anchor.magic));
1178 printf(" Version : %.8s\n", sb->anchor.revision);
598f0d58
NB
1179 printf("Controller GUID : "); print_guid(sb->controller.guid, 0);
1180 printf("\n");
1181 printf(" Container GUID : "); print_guid(sb->anchor.guid, 1);
a322f70c
DW
1182 printf("\n");
1183 printf(" Seq : %08x\n", __be32_to_cpu(sb->active->seq));
1184 printf(" Redundant hdr : %s\n", sb->secondary.magic == DDF_HEADER_MAGIC
1185 ?"yes" : "no");
1186 examine_vds(sb);
1187 examine_pds(sb);
1188}
1189
ff54de6e
N
1190static void getinfo_super_ddf(struct supertype *st, struct mdinfo *info);
1191
42dc2744 1192static void uuid_from_super_ddf(struct supertype *st, int uuid[4]);
ff54de6e 1193
061f2c6a 1194static void brief_examine_super_ddf(struct supertype *st, int verbose)
4737ae25
N
1195{
1196 /* We just write a generic DDF ARRAY entry
1197 */
1198 struct mdinfo info;
1199 char nbuf[64];
1200 getinfo_super_ddf(st, &info);
1201 fname_from_uuid(st, &info, nbuf, ':');
1202
1203 printf("ARRAY metadata=ddf UUID=%s\n", nbuf + 5);
1204}
1205
1206static void brief_examine_subarrays_ddf(struct supertype *st, int verbose)
a322f70c
DW
1207{
1208 /* We just write a generic DDF ARRAY entry
a322f70c 1209 */
42dc2744 1210 struct ddf_super *ddf = st->sb;
ff54de6e 1211 struct mdinfo info;
f21e18ca 1212 unsigned int i;
ff54de6e
N
1213 char nbuf[64];
1214 getinfo_super_ddf(st, &info);
1215 fname_from_uuid(st, &info, nbuf, ':');
42dc2744 1216
f21e18ca 1217 for (i = 0; i < __be16_to_cpu(ddf->virt->max_vdes); i++) {
42dc2744
N
1218 struct virtual_entry *ve = &ddf->virt->entries[i];
1219 struct vcl vcl;
1220 char nbuf1[64];
1221 if (all_ff(ve->guid))
1222 continue;
1223 memcpy(vcl.conf.guid, ve->guid, DDF_GUID_LEN);
1224 ddf->currentconf =&vcl;
1225 uuid_from_super_ddf(st, info.uuid);
1226 fname_from_uuid(st, &info, nbuf1, ':');
1227 printf("ARRAY container=%s member=%d UUID=%s\n",
1228 nbuf+5, i, nbuf1+5);
1229 }
a322f70c
DW
1230}
1231
bceedeec
N
1232static void export_examine_super_ddf(struct supertype *st)
1233{
1234 struct mdinfo info;
1235 char nbuf[64];
1236 getinfo_super_ddf(st, &info);
1237 fname_from_uuid(st, &info, nbuf, ':');
1238 printf("MD_METADATA=ddf\n");
1239 printf("MD_LEVEL=container\n");
1240 printf("MD_UUID=%s\n", nbuf+5);
1241}
1242
1243
a322f70c
DW
1244static void detail_super_ddf(struct supertype *st, char *homehost)
1245{
1246 /* FIXME later
1247 * Could print DDF GUID
1248 * Need to find which array
1249 * If whole, briefly list all arrays
1250 * If one, give name
1251 */
1252}
1253
1254static void brief_detail_super_ddf(struct supertype *st)
1255{
1256 /* FIXME I really need to know which array we are detailing.
1257 * Can that be stored in ddf_super??
1258 */
1259// struct ddf_super *ddf = st->sb;
ff54de6e
N
1260 struct mdinfo info;
1261 char nbuf[64];
1262 getinfo_super_ddf(st, &info);
1263 fname_from_uuid(st, &info, nbuf,':');
1264 printf(" UUID=%s", nbuf + 5);
a322f70c 1265}
a322f70c
DW
1266#endif
1267
1268static int match_home_ddf(struct supertype *st, char *homehost)
1269{
1270 /* It matches 'this' host if the controller is a
1271 * Linux-MD controller with vendor_data matching
1272 * the hostname
1273 */
1274 struct ddf_super *ddf = st->sb;
f21e18ca 1275 unsigned int len;
d1d3482b
N
1276
1277 if (!homehost)
1278 return 0;
1279 len = strlen(homehost);
a322f70c
DW
1280
1281 return (memcmp(ddf->controller.guid, T10, 8) == 0 &&
1282 len < sizeof(ddf->controller.vendor_data) &&
1283 memcmp(ddf->controller.vendor_data, homehost,len) == 0 &&
1284 ddf->controller.vendor_data[len] == 0);
1285}
1286
0e600426 1287#ifndef MDASSEMBLE
f21e18ca 1288static struct vd_config *find_vdcr(struct ddf_super *ddf, unsigned int inst)
a322f70c 1289{
7a7cc504 1290 struct vcl *v;
59e36268 1291
7a7cc504 1292 for (v = ddf->conflist; v; v = v->next)
59e36268 1293 if (inst == v->vcnum)
7a7cc504
NB
1294 return &v->conf;
1295 return NULL;
1296}
0e600426 1297#endif
7a7cc504
NB
1298
1299static int find_phys(struct ddf_super *ddf, __u32 phys_refnum)
1300{
1301 /* Find the entry in phys_disk which has the given refnum
1302 * and return it's index
1303 */
f21e18ca
N
1304 unsigned int i;
1305 for (i = 0; i < __be16_to_cpu(ddf->phys->max_pdes); i++)
7a7cc504
NB
1306 if (ddf->phys->entries[i].refnum == phys_refnum)
1307 return i;
1308 return -1;
a322f70c
DW
1309}
1310
1311static void uuid_from_super_ddf(struct supertype *st, int uuid[4])
1312{
1313 /* The uuid returned here is used for:
1314 * uuid to put into bitmap file (Create, Grow)
1315 * uuid for backup header when saving critical section (Grow)
1316 * comparing uuids when re-adding a device into an array
51006d85
N
1317 * In these cases the uuid required is that of the data-array,
1318 * not the device-set.
1319 * uuid to recognise same set when adding a missing device back
1320 * to an array. This is a uuid for the device-set.
1321 *
a322f70c
DW
1322 * For each of these we can make do with a truncated
1323 * or hashed uuid rather than the original, as long as
1324 * everyone agrees.
a322f70c
DW
1325 * In the case of SVD we assume the BVD is of interest,
1326 * though that might be the case if a bitmap were made for
1327 * a mirrored SVD - worry about that later.
1328 * So we need to find the VD configuration record for the
1329 * relevant BVD and extract the GUID and Secondary_Element_Seq.
1330 * The first 16 bytes of the sha1 of these is used.
1331 */
1332 struct ddf_super *ddf = st->sb;
d2ca6449 1333 struct vcl *vcl = ddf->currentconf;
c5afc314
N
1334 char *guid;
1335 char buf[20];
1336 struct sha1_ctx ctx;
a322f70c 1337
c5afc314
N
1338 if (vcl)
1339 guid = vcl->conf.guid;
1340 else
1341 guid = ddf->anchor.guid;
1342
1343 sha1_init_ctx(&ctx);
1344 sha1_process_bytes(guid, DDF_GUID_LEN, &ctx);
c5afc314
N
1345 sha1_finish_ctx(&ctx, buf);
1346 memcpy(uuid, buf, 4*4);
a322f70c
DW
1347}
1348
78e44928
NB
1349static void getinfo_super_ddf_bvd(struct supertype *st, struct mdinfo *info);
1350
a322f70c
DW
1351static void getinfo_super_ddf(struct supertype *st, struct mdinfo *info)
1352{
1353 struct ddf_super *ddf = st->sb;
1354
78e44928
NB
1355 if (ddf->currentconf) {
1356 getinfo_super_ddf_bvd(st, info);
1357 return;
1358 }
1359
a322f70c
DW
1360 info->array.raid_disks = __be16_to_cpu(ddf->phys->used_pdes);
1361 info->array.level = LEVEL_CONTAINER;
1362 info->array.layout = 0;
1363 info->array.md_minor = -1;
1364 info->array.ctime = DECADE + __be32_to_cpu(*(__u32*)
1365 (ddf->anchor.guid+16));
1366 info->array.utime = 0;
1367 info->array.chunk_size = 0;
510242aa 1368 info->container_enough = 1;
a322f70c 1369
a322f70c
DW
1370
1371 info->disk.major = 0;
1372 info->disk.minor = 0;
cba0191b
NB
1373 if (ddf->dlist) {
1374 info->disk.number = __be32_to_cpu(ddf->dlist->disk.refnum);
59e36268 1375 info->disk.raid_disk = find_phys(ddf, ddf->dlist->disk.refnum);
d2ca6449
NB
1376
1377 info->data_offset = __be64_to_cpu(ddf->phys->
1378 entries[info->disk.raid_disk].
1379 config_size);
1380 info->component_size = ddf->dlist->size - info->data_offset;
cba0191b
NB
1381 } else {
1382 info->disk.number = -1;
661dce36 1383 info->disk.raid_disk = -1;
cba0191b
NB
1384// info->disk.raid_disk = find refnum in the table and use index;
1385 }
f22385f9 1386 info->disk.state = (1 << MD_DISK_SYNC) | (1 << MD_DISK_ACTIVE);
a19c88b8 1387
d2ca6449 1388
921d9e16 1389 info->recovery_start = MaxSector;
a19c88b8 1390 info->reshape_active = 0;
c5afc314 1391 info->name[0] = 0;
a322f70c 1392
f35f2525
N
1393 info->array.major_version = -1;
1394 info->array.minor_version = -2;
159c3a1a 1395 strcpy(info->text_version, "ddf");
a67dd8cc 1396 info->safe_mode_delay = 0;
159c3a1a 1397
c5afc314 1398 uuid_from_super_ddf(st, info->uuid);
a322f70c 1399
a322f70c
DW
1400}
1401
598f0d58
NB
1402static int rlq_to_layout(int rlq, int prl, int raiddisks);
1403
a322f70c
DW
1404static void getinfo_super_ddf_bvd(struct supertype *st, struct mdinfo *info)
1405{
1406 struct ddf_super *ddf = st->sb;
d2ca6449
NB
1407 struct vcl *vc = ddf->currentconf;
1408 int cd = ddf->currentdev;
db42fa9b 1409 int j;
8592f29d 1410 struct dl *dl;
a322f70c
DW
1411
1412 /* FIXME this returns BVD info - what if we want SVD ?? */
1413
d2ca6449
NB
1414 info->array.raid_disks = __be16_to_cpu(vc->conf.prim_elmnt_count);
1415 info->array.level = map_num1(ddf_level_num, vc->conf.prl);
1416 info->array.layout = rlq_to_layout(vc->conf.rlq, vc->conf.prl,
598f0d58 1417 info->array.raid_disks);
a322f70c 1418 info->array.md_minor = -1;
d2ca6449
NB
1419 info->array.ctime = DECADE +
1420 __be32_to_cpu(*(__u32*)(vc->conf.guid+16));
1421 info->array.utime = DECADE + __be32_to_cpu(vc->conf.timestamp);
1422 info->array.chunk_size = 512 << vc->conf.chunk_shift;
da9b4a62 1423 info->custom_array_size = 0;
d2ca6449 1424
f21e18ca 1425 if (cd >= 0 && (unsigned)cd < ddf->mppe) {
d2ca6449
NB
1426 info->data_offset = __be64_to_cpu(vc->lba_offset[cd]);
1427 if (vc->block_sizes)
1428 info->component_size = vc->block_sizes[cd];
1429 else
1430 info->component_size = __be64_to_cpu(vc->conf.blocks);
1431 }
a322f70c 1432
8592f29d
N
1433 for (dl = ddf->dlist; dl ; dl = dl->next)
1434 if (dl->raiddisk == info->disk.raid_disk)
1435 break;
a322f70c
DW
1436 info->disk.major = 0;
1437 info->disk.minor = 0;
8592f29d
N
1438 if (dl) {
1439 info->disk.major = dl->major;
1440 info->disk.minor = dl->minor;
1441 }
a322f70c
DW
1442// info->disk.number = __be32_to_cpu(ddf->disk.refnum);
1443// info->disk.raid_disk = find refnum in the table and use index;
1444// info->disk.state = ???;
1445
103f2410
NB
1446 info->container_member = ddf->currentconf->vcnum;
1447
921d9e16 1448 info->recovery_start = MaxSector;
80d26cb2 1449 info->resync_start = 0;
624c5ad4 1450 info->reshape_active = 0;
80d26cb2
NB
1451 if (!(ddf->virt->entries[info->container_member].state
1452 & DDF_state_inconsistent) &&
1453 (ddf->virt->entries[info->container_member].init_state
1454 & DDF_initstate_mask)
1455 == DDF_init_full)
b7528a20 1456 info->resync_start = MaxSector;
80d26cb2 1457
a322f70c
DW
1458 uuid_from_super_ddf(st, info->uuid);
1459
3576e302 1460 info->container_member = atoi(st->subarray);
f35f2525
N
1461 info->array.major_version = -1;
1462 info->array.minor_version = -2;
3576e302 1463 sprintf(info->text_version, "/%s/%s",
159c3a1a 1464 devnum2devname(st->container_dev),
3576e302 1465 st->subarray);
a67dd8cc 1466 info->safe_mode_delay = 200;
159c3a1a 1467
db42fa9b
N
1468 memcpy(info->name, ddf->virt->entries[info->container_member].name, 16);
1469 info->name[16]=0;
1470 for(j=0; j<16; j++)
1471 if (info->name[j] == ' ')
1472 info->name[j] = 0;
a322f70c
DW
1473}
1474
598f0d58 1475
a322f70c
DW
1476static int update_super_ddf(struct supertype *st, struct mdinfo *info,
1477 char *update,
1478 char *devname, int verbose,
1479 int uuid_set, char *homehost)
1480{
1481 /* For 'assemble' and 'force' we need to return non-zero if any
1482 * change was made. For others, the return value is ignored.
1483 * Update options are:
1484 * force-one : This device looks a bit old but needs to be included,
1485 * update age info appropriately.
1486 * assemble: clear any 'faulty' flag to allow this device to
1487 * be assembled.
1488 * force-array: Array is degraded but being forced, mark it clean
1489 * if that will be needed to assemble it.
1490 *
1491 * newdev: not used ????
1492 * grow: Array has gained a new device - this is currently for
1493 * linear only
1494 * resync: mark as dirty so a resync will happen.
59e36268 1495 * uuid: Change the uuid of the array to match what is given
a322f70c
DW
1496 * homehost: update the recorded homehost
1497 * name: update the name - preserving the homehost
1498 * _reshape_progress: record new reshape_progress position.
1499 *
1500 * Following are not relevant for this version:
1501 * sparc2.2 : update from old dodgey metadata
1502 * super-minor: change the preferred_minor number
1503 * summaries: update redundant counters.
1504 */
1505 int rv = 0;
1506// struct ddf_super *ddf = st->sb;
7a7cc504 1507// struct vd_config *vd = find_vdcr(ddf, info->container_member);
a322f70c
DW
1508// struct virtual_entry *ve = find_ve(ddf);
1509
a322f70c
DW
1510 /* we don't need to handle "force-*" or "assemble" as
1511 * there is no need to 'trick' the kernel. We the metadata is
1512 * first updated to activate the array, all the implied modifications
1513 * will just happen.
1514 */
1515
1516 if (strcmp(update, "grow") == 0) {
1517 /* FIXME */
1e2b2765 1518 } else if (strcmp(update, "resync") == 0) {
a322f70c 1519// info->resync_checkpoint = 0;
1e2b2765 1520 } else if (strcmp(update, "homehost") == 0) {
a322f70c
DW
1521 /* homehost is stored in controller->vendor_data,
1522 * or it is when we are the vendor
1523 */
1524// if (info->vendor_is_local)
1525// strcpy(ddf->controller.vendor_data, homehost);
1e2b2765
N
1526 rv = -1;
1527 } if (strcmp(update, "name") == 0) {
a322f70c
DW
1528 /* name is stored in virtual_entry->name */
1529// memset(ve->name, ' ', 16);
1530// strncpy(ve->name, info->name, 16);
1e2b2765
N
1531 rv = -1;
1532 } if (strcmp(update, "_reshape_progress") == 0) {
a322f70c 1533 /* We don't support reshape yet */
1e2b2765
N
1534 } else
1535 rv = -1;
a322f70c
DW
1536
1537// update_all_csum(ddf);
1538
1539 return rv;
1540}
1541
5f8097be
NB
1542static void make_header_guid(char *guid)
1543{
1544 __u32 stamp;
5f8097be
NB
1545 /* Create a DDF Header of Virtual Disk GUID */
1546
1547 /* 24 bytes of fiction required.
1548 * first 8 are a 'vendor-id' - "Linux-MD"
1549 * next 8 are controller type.. how about 0X DEAD BEEF 0000 0000
1550 * Remaining 8 random number plus timestamp
1551 */
1552 memcpy(guid, T10, sizeof(T10));
1553 stamp = __cpu_to_be32(0xdeadbeef);
1554 memcpy(guid+8, &stamp, 4);
1555 stamp = __cpu_to_be32(0);
1556 memcpy(guid+12, &stamp, 4);
1557 stamp = __cpu_to_be32(time(0) - DECADE);
1558 memcpy(guid+16, &stamp, 4);
bfb7ea78 1559 stamp = random32();
5f8097be 1560 memcpy(guid+20, &stamp, 4);
5f8097be 1561}
59e36268 1562
78e44928
NB
1563static int init_super_ddf_bvd(struct supertype *st,
1564 mdu_array_info_t *info,
1565 unsigned long long size,
1566 char *name, char *homehost,
1567 int *uuid);
1568
a322f70c
DW
1569static int init_super_ddf(struct supertype *st,
1570 mdu_array_info_t *info,
1571 unsigned long long size, char *name, char *homehost,
1572 int *uuid)
1573{
1574 /* This is primarily called by Create when creating a new array.
1575 * We will then get add_to_super called for each component, and then
1576 * write_init_super called to write it out to each device.
1577 * For DDF, Create can create on fresh devices or on a pre-existing
1578 * array.
1579 * To create on a pre-existing array a different method will be called.
1580 * This one is just for fresh drives.
1581 *
1582 * We need to create the entire 'ddf' structure which includes:
1583 * DDF headers - these are easy.
1584 * Controller data - a Sector describing this controller .. not that
1585 * this is a controller exactly.
1586 * Physical Disk Record - one entry per device, so
1587 * leave plenty of space.
1588 * Virtual Disk Records - again, just leave plenty of space.
1589 * This just lists VDs, doesn't give details
1590 * Config records - describes the VDs that use this disk
1591 * DiskData - describes 'this' device.
1592 * BadBlockManagement - empty
1593 * Diag Space - empty
1594 * Vendor Logs - Could we put bitmaps here?
1595 *
1596 */
1597 struct ddf_super *ddf;
1598 char hostname[17];
1599 int hostlen;
a322f70c
DW
1600 int max_phys_disks, max_virt_disks;
1601 unsigned long long sector;
1602 int clen;
1603 int i;
1604 int pdsize, vdsize;
1605 struct phys_disk *pd;
1606 struct virtual_disk *vd;
1607
78e44928 1608 if (st->sb)
955e9ea1 1609 return init_super_ddf_bvd(st, info, size, name, homehost, uuid);
ba7eb04f 1610
3d2c4fc7
DW
1611 if (posix_memalign((void**)&ddf, 512, sizeof(*ddf)) != 0) {
1612 fprintf(stderr, Name ": %s could not allocate superblock\n", __func__);
1613 return 0;
1614 }
6264b437 1615 memset(ddf, 0, sizeof(*ddf));
a322f70c
DW
1616 ddf->dlist = NULL; /* no physical disks yet */
1617 ddf->conflist = NULL; /* No virtual disks yet */
955e9ea1
DW
1618 st->sb = ddf;
1619
1620 if (info == NULL) {
1621 /* zeroing superblock */
1622 return 0;
1623 }
a322f70c
DW
1624
1625 /* At least 32MB *must* be reserved for the ddf. So let's just
1626 * start 32MB from the end, and put the primary header there.
1627 * Don't do secondary for now.
1628 * We don't know exactly where that will be yet as it could be
1629 * different on each device. To just set up the lengths.
1630 *
1631 */
1632
1633 ddf->anchor.magic = DDF_HEADER_MAGIC;
5f8097be 1634 make_header_guid(ddf->anchor.guid);
a322f70c 1635
59e36268 1636 memcpy(ddf->anchor.revision, DDF_REVISION_2, 8);
a322f70c
DW
1637 ddf->anchor.seq = __cpu_to_be32(1);
1638 ddf->anchor.timestamp = __cpu_to_be32(time(0) - DECADE);
1639 ddf->anchor.openflag = 0xFF;
1640 ddf->anchor.foreignflag = 0;
1641 ddf->anchor.enforcegroups = 0; /* Is this best?? */
1642 ddf->anchor.pad0 = 0xff;
1643 memset(ddf->anchor.pad1, 0xff, 12);
1644 memset(ddf->anchor.header_ext, 0xff, 32);
1645 ddf->anchor.primary_lba = ~(__u64)0;
1646 ddf->anchor.secondary_lba = ~(__u64)0;
1647 ddf->anchor.type = DDF_HEADER_ANCHOR;
1648 memset(ddf->anchor.pad2, 0xff, 3);
1649 ddf->anchor.workspace_len = __cpu_to_be32(32768); /* Must be reserved */
1650 ddf->anchor.workspace_lba = ~(__u64)0; /* Put this at bottom
1651 of 32M reserved.. */
1652 max_phys_disks = 1023; /* Should be enough */
1653 ddf->anchor.max_pd_entries = __cpu_to_be16(max_phys_disks);
1654 max_virt_disks = 255;
1655 ddf->anchor.max_vd_entries = __cpu_to_be16(max_virt_disks); /* ?? */
1656 ddf->anchor.max_partitions = __cpu_to_be16(64); /* ?? */
1657 ddf->max_part = 64;
8c3b8c2c 1658 ddf->mppe = 256;
59e36268
NB
1659 ddf->conf_rec_len = 1 + ROUND_UP(ddf->mppe * (4+8), 512)/512;
1660 ddf->anchor.config_record_len = __cpu_to_be16(ddf->conf_rec_len);
1661 ddf->anchor.max_primary_element_entries = __cpu_to_be16(ddf->mppe);
a322f70c 1662 memset(ddf->anchor.pad3, 0xff, 54);
a322f70c
DW
1663 /* controller sections is one sector long immediately
1664 * after the ddf header */
1665 sector = 1;
1666 ddf->anchor.controller_section_offset = __cpu_to_be32(sector);
1667 ddf->anchor.controller_section_length = __cpu_to_be32(1);
1668 sector += 1;
1669
1670 /* phys is 8 sectors after that */
1671 pdsize = ROUND_UP(sizeof(struct phys_disk) +
1672 sizeof(struct phys_disk_entry)*max_phys_disks,
1673 512);
1674 switch(pdsize/512) {
1675 case 2: case 8: case 32: case 128: case 512: break;
1676 default: abort();
1677 }
1678 ddf->anchor.phys_section_offset = __cpu_to_be32(sector);
1679 ddf->anchor.phys_section_length =
1680 __cpu_to_be32(pdsize/512); /* max_primary_element_entries/8 */
1681 sector += pdsize/512;
1682
1683 /* virt is another 32 sectors */
1684 vdsize = ROUND_UP(sizeof(struct virtual_disk) +
1685 sizeof(struct virtual_entry) * max_virt_disks,
1686 512);
1687 switch(vdsize/512) {
1688 case 2: case 8: case 32: case 128: case 512: break;
1689 default: abort();
1690 }
1691 ddf->anchor.virt_section_offset = __cpu_to_be32(sector);
1692 ddf->anchor.virt_section_length =
1693 __cpu_to_be32(vdsize/512); /* max_vd_entries/8 */
1694 sector += vdsize/512;
1695
59e36268 1696 clen = ddf->conf_rec_len * (ddf->max_part+1);
a322f70c
DW
1697 ddf->anchor.config_section_offset = __cpu_to_be32(sector);
1698 ddf->anchor.config_section_length = __cpu_to_be32(clen);
1699 sector += clen;
1700
1701 ddf->anchor.data_section_offset = __cpu_to_be32(sector);
1702 ddf->anchor.data_section_length = __cpu_to_be32(1);
1703 sector += 1;
1704
1705 ddf->anchor.bbm_section_length = __cpu_to_be32(0);
1706 ddf->anchor.bbm_section_offset = __cpu_to_be32(0xFFFFFFFF);
1707 ddf->anchor.diag_space_length = __cpu_to_be32(0);
1708 ddf->anchor.diag_space_offset = __cpu_to_be32(0xFFFFFFFF);
1709 ddf->anchor.vendor_length = __cpu_to_be32(0);
1710 ddf->anchor.vendor_offset = __cpu_to_be32(0xFFFFFFFF);
1711
1712 memset(ddf->anchor.pad4, 0xff, 256);
1713
1714 memcpy(&ddf->primary, &ddf->anchor, 512);
1715 memcpy(&ddf->secondary, &ddf->anchor, 512);
1716
1717 ddf->primary.openflag = 1; /* I guess.. */
1718 ddf->primary.type = DDF_HEADER_PRIMARY;
1719
1720 ddf->secondary.openflag = 1; /* I guess.. */
1721 ddf->secondary.type = DDF_HEADER_SECONDARY;
1722
1723 ddf->active = &ddf->primary;
1724
1725 ddf->controller.magic = DDF_CONTROLLER_MAGIC;
1726
1727 /* 24 more bytes of fiction required.
1728 * first 8 are a 'vendor-id' - "Linux-MD"
1729 * Remaining 16 are serial number.... maybe a hostname would do?
1730 */
1731 memcpy(ddf->controller.guid, T10, sizeof(T10));
1ba6bff9
DW
1732 gethostname(hostname, sizeof(hostname));
1733 hostname[sizeof(hostname) - 1] = 0;
a322f70c
DW
1734 hostlen = strlen(hostname);
1735 memcpy(ddf->controller.guid + 24 - hostlen, hostname, hostlen);
1736 for (i = strlen(T10) ; i+hostlen < 24; i++)
1737 ddf->controller.guid[i] = ' ';
1738
1739 ddf->controller.type.vendor_id = __cpu_to_be16(0xDEAD);
1740 ddf->controller.type.device_id = __cpu_to_be16(0xBEEF);
1741 ddf->controller.type.sub_vendor_id = 0;
1742 ddf->controller.type.sub_device_id = 0;
1743 memcpy(ddf->controller.product_id, "What Is My PID??", 16);
1744 memset(ddf->controller.pad, 0xff, 8);
1745 memset(ddf->controller.vendor_data, 0xff, 448);
a9e1c11d
N
1746 if (homehost && strlen(homehost) < 440)
1747 strcpy((char*)ddf->controller.vendor_data, homehost);
a322f70c 1748
3d2c4fc7
DW
1749 if (posix_memalign((void**)&pd, 512, pdsize) != 0) {
1750 fprintf(stderr, Name ": %s could not allocate pd\n", __func__);
1751 return 0;
1752 }
6416d527 1753 ddf->phys = pd;
a322f70c
DW
1754 ddf->pdsize = pdsize;
1755
1756 memset(pd, 0xff, pdsize);
1757 memset(pd, 0, sizeof(*pd));
076515ba 1758 pd->magic = DDF_PHYS_RECORDS_MAGIC;
a322f70c
DW
1759 pd->used_pdes = __cpu_to_be16(0);
1760 pd->max_pdes = __cpu_to_be16(max_phys_disks);
1761 memset(pd->pad, 0xff, 52);
1762
3d2c4fc7
DW
1763 if (posix_memalign((void**)&vd, 512, vdsize) != 0) {
1764 fprintf(stderr, Name ": %s could not allocate vd\n", __func__);
1765 return 0;
1766 }
6416d527 1767 ddf->virt = vd;
a322f70c
DW
1768 ddf->vdsize = vdsize;
1769 memset(vd, 0, vdsize);
1770 vd->magic = DDF_VIRT_RECORDS_MAGIC;
1771 vd->populated_vdes = __cpu_to_be16(0);
1772 vd->max_vdes = __cpu_to_be16(max_virt_disks);
1773 memset(vd->pad, 0xff, 52);
1774
5f8097be
NB
1775 for (i=0; i<max_virt_disks; i++)
1776 memset(&vd->entries[i], 0xff, sizeof(struct virtual_entry));
1777
a322f70c 1778 st->sb = ddf;
18a2f463 1779 ddf->updates_pending = 1;
a322f70c
DW
1780 return 1;
1781}
1782
5f8097be
NB
1783static int chunk_to_shift(int chunksize)
1784{
1785 return ffs(chunksize/512)-1;
1786}
1787
1788static int level_to_prl(int level)
1789{
1790 switch (level) {
1791 case LEVEL_LINEAR: return DDF_CONCAT;
1792 case 0: return DDF_RAID0;
1793 case 1: return DDF_RAID1;
1794 case 4: return DDF_RAID4;
1795 case 5: return DDF_RAID5;
1796 case 6: return DDF_RAID6;
1797 default: return -1;
1798 }
1799}
1800static int layout_to_rlq(int level, int layout, int raiddisks)
1801{
1802 switch(level) {
1803 case 0:
1804 return DDF_RAID0_SIMPLE;
1805 case 1:
1806 switch(raiddisks) {
1807 case 2: return DDF_RAID1_SIMPLE;
1808 case 3: return DDF_RAID1_MULTI;
1809 default: return -1;
1810 }
1811 case 4:
1812 switch(layout) {
1813 case 0: return DDF_RAID4_N;
1814 }
1815 break;
1816 case 5:
5f8097be
NB
1817 switch(layout) {
1818 case ALGORITHM_LEFT_ASYMMETRIC:
1819 return DDF_RAID5_N_RESTART;
1820 case ALGORITHM_RIGHT_ASYMMETRIC:
b640a252 1821 return DDF_RAID5_0_RESTART;
5f8097be
NB
1822 case ALGORITHM_LEFT_SYMMETRIC:
1823 return DDF_RAID5_N_CONTINUE;
1824 case ALGORITHM_RIGHT_SYMMETRIC:
1825 return -1; /* not mentioned in standard */
1826 }
b640a252
N
1827 case 6:
1828 switch(layout) {
1829 case ALGORITHM_ROTATING_N_RESTART:
1830 return DDF_RAID5_N_RESTART;
1831 case ALGORITHM_ROTATING_ZERO_RESTART:
1832 return DDF_RAID6_0_RESTART;
1833 case ALGORITHM_ROTATING_N_CONTINUE:
1834 return DDF_RAID5_N_CONTINUE;
1835 }
5f8097be
NB
1836 }
1837 return -1;
1838}
1839
598f0d58
NB
1840static int rlq_to_layout(int rlq, int prl, int raiddisks)
1841{
1842 switch(prl) {
1843 case DDF_RAID0:
1844 return 0; /* hopefully rlq == DDF_RAID0_SIMPLE */
1845 case DDF_RAID1:
1846 return 0; /* hopefully rlq == SIMPLE or MULTI depending
1847 on raiddisks*/
1848 case DDF_RAID4:
1849 switch(rlq) {
1850 case DDF_RAID4_N:
1851 return 0;
1852 default:
1853 /* not supported */
1854 return -1; /* FIXME this isn't checked */
1855 }
1856 case DDF_RAID5:
598f0d58
NB
1857 switch(rlq) {
1858 case DDF_RAID5_N_RESTART:
1859 return ALGORITHM_LEFT_ASYMMETRIC;
1860 case DDF_RAID5_0_RESTART:
1861 return ALGORITHM_RIGHT_ASYMMETRIC;
1862 case DDF_RAID5_N_CONTINUE:
1863 return ALGORITHM_LEFT_SYMMETRIC;
1864 default:
1865 return -1;
1866 }
59e36268
NB
1867 case DDF_RAID6:
1868 switch(rlq) {
1869 case DDF_RAID5_N_RESTART:
b640a252 1870 return ALGORITHM_ROTATING_N_RESTART;
59e36268 1871 case DDF_RAID6_0_RESTART:
b640a252 1872 return ALGORITHM_ROTATING_ZERO_RESTART;
59e36268 1873 case DDF_RAID5_N_CONTINUE:
b640a252 1874 return ALGORITHM_ROTATING_N_CONTINUE;
59e36268
NB
1875 default:
1876 return -1;
1877 }
598f0d58
NB
1878 }
1879 return -1;
1880}
1881
0e600426 1882#ifndef MDASSEMBLE
59e36268
NB
1883struct extent {
1884 unsigned long long start, size;
1885};
78e44928 1886static int cmp_extent(const void *av, const void *bv)
59e36268
NB
1887{
1888 const struct extent *a = av;
1889 const struct extent *b = bv;
1890 if (a->start < b->start)
1891 return -1;
1892 if (a->start > b->start)
1893 return 1;
1894 return 0;
1895}
1896
78e44928 1897static struct extent *get_extents(struct ddf_super *ddf, struct dl *dl)
59e36268
NB
1898{
1899 /* find a list of used extents on the give physical device
1900 * (dnum) of the given ddf.
1901 * Return a malloced array of 'struct extent'
1902
1903FIXME ignore DDF_Legacy devices?
1904
1905 */
1906 struct extent *rv;
1907 int n = 0;
f21e18ca 1908 unsigned int i, j;
59e36268
NB
1909
1910 rv = malloc(sizeof(struct extent) * (ddf->max_part + 2));
1911 if (!rv)
1912 return NULL;
1913
1914 for (i = 0; i < ddf->max_part; i++) {
1915 struct vcl *v = dl->vlist[i];
1916 if (v == NULL)
1917 continue;
f21e18ca 1918 for (j = 0; j < v->conf.prim_elmnt_count; j++)
59e36268
NB
1919 if (v->conf.phys_refnum[j] == dl->disk.refnum) {
1920 /* This device plays role 'j' in 'v'. */
1921 rv[n].start = __be64_to_cpu(v->lba_offset[j]);
1922 rv[n].size = __be64_to_cpu(v->conf.blocks);
1923 n++;
1924 break;
1925 }
1926 }
1927 qsort(rv, n, sizeof(*rv), cmp_extent);
1928
1929 rv[n].start = __be64_to_cpu(ddf->phys->entries[dl->pdnum].config_size);
1930 rv[n].size = 0;
1931 return rv;
1932}
0e600426 1933#endif
59e36268 1934
5f8097be
NB
1935static int init_super_ddf_bvd(struct supertype *st,
1936 mdu_array_info_t *info,
1937 unsigned long long size,
1938 char *name, char *homehost,
1939 int *uuid)
1940{
1941 /* We are creating a BVD inside a pre-existing container.
1942 * so st->sb is already set.
1943 * We need to create a new vd_config and a new virtual_entry
1944 */
1945 struct ddf_super *ddf = st->sb;
f21e18ca 1946 unsigned int venum;
5f8097be
NB
1947 struct virtual_entry *ve;
1948 struct vcl *vcl;
1949 struct vd_config *vc;
5f8097be
NB
1950
1951 if (__be16_to_cpu(ddf->virt->populated_vdes)
1952 >= __be16_to_cpu(ddf->virt->max_vdes)) {
1953 fprintf(stderr, Name": This ddf already has the "
1954 "maximum of %d virtual devices\n",
1955 __be16_to_cpu(ddf->virt->max_vdes));
1956 return 0;
1957 }
1958
1959 for (venum = 0; venum < __be16_to_cpu(ddf->virt->max_vdes); venum++)
1960 if (all_ff(ddf->virt->entries[venum].guid))
1961 break;
1962 if (venum == __be16_to_cpu(ddf->virt->max_vdes)) {
1963 fprintf(stderr, Name ": Cannot find spare slot for "
1964 "virtual disk - DDF is corrupt\n");
1965 return 0;
1966 }
1967 ve = &ddf->virt->entries[venum];
1968
1969 /* A Virtual Disk GUID contains the T10 Vendor ID, controller type,
1970 * timestamp, random number
1971 */
1972 make_header_guid(ve->guid);
1973 ve->unit = __cpu_to_be16(info->md_minor);
1974 ve->pad0 = 0xFFFF;
1975 ve->guid_crc = crc32(0, (unsigned char*)ddf->anchor.guid, DDF_GUID_LEN);
1976 ve->type = 0;
7a7cc504
NB
1977 ve->state = DDF_state_degraded; /* Will be modified as devices are added */
1978 if (info->state & 1) /* clean */
1979 ve->init_state = DDF_init_full;
1980 else
1981 ve->init_state = DDF_init_not;
1982
5f8097be
NB
1983 memset(ve->pad1, 0xff, 14);
1984 memset(ve->name, ' ', 16);
1985 if (name)
1986 strncpy(ve->name, name, 16);
1987 ddf->virt->populated_vdes =
1988 __cpu_to_be16(__be16_to_cpu(ddf->virt->populated_vdes)+1);
1989
1990 /* Now create a new vd_config */
3d2c4fc7
DW
1991 if (posix_memalign((void**)&vcl, 512,
1992 (offsetof(struct vcl, conf) + ddf->conf_rec_len * 512)) != 0) {
1993 fprintf(stderr, Name ": %s could not allocate vd_config\n", __func__);
1994 return 0;
1995 }
8c3b8c2c 1996 vcl->lba_offset = (__u64*) &vcl->conf.phys_refnum[ddf->mppe];
59e36268 1997 vcl->vcnum = venum;
f7e7067b 1998 sprintf(st->subarray, "%d", venum);
59e36268 1999 vcl->block_sizes = NULL; /* FIXME not for CONCAT */
5f8097be
NB
2000
2001 vc = &vcl->conf;
2002
2003 vc->magic = DDF_VD_CONF_MAGIC;
2004 memcpy(vc->guid, ve->guid, DDF_GUID_LEN);
2005 vc->timestamp = __cpu_to_be32(time(0)-DECADE);
2006 vc->seqnum = __cpu_to_be32(1);
2007 memset(vc->pad0, 0xff, 24);
2008 vc->prim_elmnt_count = __cpu_to_be16(info->raid_disks);
2009 vc->chunk_shift = chunk_to_shift(info->chunk_size);
2010 vc->prl = level_to_prl(info->level);
2011 vc->rlq = layout_to_rlq(info->level, info->layout, info->raid_disks);
2012 vc->sec_elmnt_count = 1;
2013 vc->sec_elmnt_seq = 0;
2014 vc->srl = 0;
2015 vc->blocks = __cpu_to_be64(info->size * 2);
2016 vc->array_blocks = __cpu_to_be64(
2017 calc_array_size(info->level, info->raid_disks, info->layout,
2018 info->chunk_size, info->size*2));
2019 memset(vc->pad1, 0xff, 8);
2020 vc->spare_refs[0] = 0xffffffff;
2021 vc->spare_refs[1] = 0xffffffff;
2022 vc->spare_refs[2] = 0xffffffff;
2023 vc->spare_refs[3] = 0xffffffff;
2024 vc->spare_refs[4] = 0xffffffff;
2025 vc->spare_refs[5] = 0xffffffff;
2026 vc->spare_refs[6] = 0xffffffff;
2027 vc->spare_refs[7] = 0xffffffff;
2028 memset(vc->cache_pol, 0, 8);
2029 vc->bg_rate = 0x80;
2030 memset(vc->pad2, 0xff, 3);
2031 memset(vc->pad3, 0xff, 52);
2032 memset(vc->pad4, 0xff, 192);
2033 memset(vc->v0, 0xff, 32);
2034 memset(vc->v1, 0xff, 32);
2035 memset(vc->v2, 0xff, 16);
2036 memset(vc->v3, 0xff, 16);
2037 memset(vc->vendor, 0xff, 32);
598f0d58 2038
8c3b8c2c 2039 memset(vc->phys_refnum, 0xff, 4*ddf->mppe);
e5a2a3cf 2040 memset(vc->phys_refnum+ddf->mppe, 0x00, 8*ddf->mppe);
5f8097be
NB
2041
2042 vcl->next = ddf->conflist;
2043 ddf->conflist = vcl;
d2ca6449 2044 ddf->currentconf = vcl;
18a2f463 2045 ddf->updates_pending = 1;
5f8097be
NB
2046 return 1;
2047}
2048
0e600426 2049#ifndef MDASSEMBLE
5f8097be
NB
2050static void add_to_super_ddf_bvd(struct supertype *st,
2051 mdu_disk_info_t *dk, int fd, char *devname)
2052{
2053 /* fd and devname identify a device with-in the ddf container (st).
2054 * dk identifies a location in the new BVD.
2055 * We need to find suitable free space in that device and update
2056 * the phys_refnum and lba_offset for the newly created vd_config.
2057 * We might also want to update the type in the phys_disk
5575e7d9 2058 * section.
8592f29d
N
2059 *
2060 * Alternately: fd == -1 and we have already chosen which device to
2061 * use and recorded in dlist->raid_disk;
5f8097be
NB
2062 */
2063 struct dl *dl;
2064 struct ddf_super *ddf = st->sb;
2065 struct vd_config *vc;
2066 __u64 *lba_offset;
f21e18ca
N
2067 unsigned int working;
2068 unsigned int i;
59e36268
NB
2069 unsigned long long blocks, pos, esize;
2070 struct extent *ex;
5f8097be 2071
8592f29d
N
2072 if (fd == -1) {
2073 for (dl = ddf->dlist; dl ; dl = dl->next)
2074 if (dl->raiddisk == dk->raid_disk)
2075 break;
2076 } else {
2077 for (dl = ddf->dlist; dl ; dl = dl->next)
2078 if (dl->major == dk->major &&
2079 dl->minor == dk->minor)
2080 break;
2081 }
5f8097be
NB
2082 if (!dl || ! (dk->state & (1<<MD_DISK_SYNC)))
2083 return;
2084
d2ca6449
NB
2085 vc = &ddf->currentconf->conf;
2086 lba_offset = ddf->currentconf->lba_offset;
59e36268
NB
2087
2088 ex = get_extents(ddf, dl);
2089 if (!ex)
2090 return;
2091
2092 i = 0; pos = 0;
2093 blocks = __be64_to_cpu(vc->blocks);
d2ca6449
NB
2094 if (ddf->currentconf->block_sizes)
2095 blocks = ddf->currentconf->block_sizes[dk->raid_disk];
59e36268
NB
2096
2097 do {
2098 esize = ex[i].start - pos;
2099 if (esize >= blocks)
2100 break;
2101 pos = ex[i].start + ex[i].size;
2102 i++;
2103 } while (ex[i-1].size);
2104
2105 free(ex);
2106 if (esize < blocks)
2107 return;
2108
d2ca6449 2109 ddf->currentdev = dk->raid_disk;
5f8097be 2110 vc->phys_refnum[dk->raid_disk] = dl->disk.refnum;
59e36268 2111 lba_offset[dk->raid_disk] = __cpu_to_be64(pos);
5f8097be 2112
f21e18ca 2113 for (i = 0; i < ddf->max_part ; i++)
5575e7d9
NB
2114 if (dl->vlist[i] == NULL)
2115 break;
2116 if (i == ddf->max_part)
2117 return;
d2ca6449 2118 dl->vlist[i] = ddf->currentconf;
5f8097be 2119
8592f29d
N
2120 if (fd >= 0)
2121 dl->fd = fd;
2122 if (devname)
2123 dl->devname = devname;
7a7cc504
NB
2124
2125 /* Check how many working raid_disks, and if we can mark
2126 * array as optimal yet
2127 */
2128 working = 0;
5575e7d9 2129
f21e18ca 2130 for (i = 0; i < __be16_to_cpu(vc->prim_elmnt_count); i++)
7a7cc504
NB
2131 if (vc->phys_refnum[i] != 0xffffffff)
2132 working++;
59e36268 2133
5575e7d9 2134 /* Find which virtual_entry */
d2ca6449 2135 i = ddf->currentconf->vcnum;
7a7cc504 2136 if (working == __be16_to_cpu(vc->prim_elmnt_count))
5575e7d9
NB
2137 ddf->virt->entries[i].state =
2138 (ddf->virt->entries[i].state & ~DDF_state_mask)
7a7cc504
NB
2139 | DDF_state_optimal;
2140
2141 if (vc->prl == DDF_RAID6 &&
2142 working+1 == __be16_to_cpu(vc->prim_elmnt_count))
5575e7d9
NB
2143 ddf->virt->entries[i].state =
2144 (ddf->virt->entries[i].state & ~DDF_state_mask)
7a7cc504 2145 | DDF_state_part_optimal;
5575e7d9
NB
2146
2147 ddf->phys->entries[dl->pdnum].type &= ~__cpu_to_be16(DDF_Global_Spare);
2148 ddf->phys->entries[dl->pdnum].type |= __cpu_to_be16(DDF_Active_in_VD);
18a2f463 2149 ddf->updates_pending = 1;
5f8097be
NB
2150}
2151
a322f70c
DW
2152/* add a device to a container, either while creating it or while
2153 * expanding a pre-existing container
2154 */
f20c3968 2155static int add_to_super_ddf(struct supertype *st,
a322f70c
DW
2156 mdu_disk_info_t *dk, int fd, char *devname)
2157{
2158 struct ddf_super *ddf = st->sb;
2159 struct dl *dd;
2160 time_t now;
2161 struct tm *tm;
2162 unsigned long long size;
2163 struct phys_disk_entry *pde;
f21e18ca 2164 unsigned int n, i;
a322f70c
DW
2165 struct stat stb;
2166
78e44928
NB
2167 if (ddf->currentconf) {
2168 add_to_super_ddf_bvd(st, dk, fd, devname);
f20c3968 2169 return 0;
78e44928
NB
2170 }
2171
a322f70c
DW
2172 /* This is device numbered dk->number. We need to create
2173 * a phys_disk entry and a more detailed disk_data entry.
2174 */
2175 fstat(fd, &stb);
3d2c4fc7
DW
2176 if (posix_memalign((void**)&dd, 512,
2177 sizeof(*dd) + sizeof(dd->vlist[0]) * ddf->max_part) != 0) {
2178 fprintf(stderr, Name
2179 ": %s could allocate buffer for new disk, aborting\n",
2180 __func__);
f20c3968 2181 return 1;
3d2c4fc7 2182 }
a322f70c
DW
2183 dd->major = major(stb.st_rdev);
2184 dd->minor = minor(stb.st_rdev);
2185 dd->devname = devname;
a322f70c 2186 dd->fd = fd;
b2280677 2187 dd->spare = NULL;
a322f70c
DW
2188
2189 dd->disk.magic = DDF_PHYS_DATA_MAGIC;
2190 now = time(0);
2191 tm = localtime(&now);
2192 sprintf(dd->disk.guid, "%8s%04d%02d%02d",
2193 T10, tm->tm_year+1900, tm->tm_mon+1, tm->tm_mday);
bfb7ea78
N
2194 *(__u32*)(dd->disk.guid + 16) = random32();
2195 *(__u32*)(dd->disk.guid + 20) = random32();
a322f70c 2196
59e36268
NB
2197 do {
2198 /* Cannot be bothered finding a CRC of some irrelevant details*/
bfb7ea78 2199 dd->disk.refnum = random32();
f21e18ca
N
2200 for (i = __be16_to_cpu(ddf->active->max_pd_entries);
2201 i > 0; i--)
2202 if (ddf->phys->entries[i-1].refnum == dd->disk.refnum)
59e36268 2203 break;
f21e18ca 2204 } while (i > 0);
59e36268 2205
a322f70c
DW
2206 dd->disk.forced_ref = 1;
2207 dd->disk.forced_guid = 1;
2208 memset(dd->disk.vendor, ' ', 32);
2209 memcpy(dd->disk.vendor, "Linux", 5);
2210 memset(dd->disk.pad, 0xff, 442);
b2280677 2211 for (i = 0; i < ddf->max_part ; i++)
a322f70c
DW
2212 dd->vlist[i] = NULL;
2213
2214 n = __be16_to_cpu(ddf->phys->used_pdes);
2215 pde = &ddf->phys->entries[n];
5575e7d9
NB
2216 dd->pdnum = n;
2217
2cc2983d
N
2218 if (st->update_tail) {
2219 int len = (sizeof(struct phys_disk) +
2220 sizeof(struct phys_disk_entry));
2221 struct phys_disk *pd;
2222
2223 pd = malloc(len);
2224 pd->magic = DDF_PHYS_RECORDS_MAGIC;
2225 pd->used_pdes = __cpu_to_be16(n);
2226 pde = &pd->entries[0];
2227 dd->mdupdate = pd;
2228 } else {
2229 n++;
2230 ddf->phys->used_pdes = __cpu_to_be16(n);
2231 }
a322f70c
DW
2232
2233 memcpy(pde->guid, dd->disk.guid, DDF_GUID_LEN);
2234 pde->refnum = dd->disk.refnum;
5575e7d9 2235 pde->type = __cpu_to_be16(DDF_Forced_PD_GUID | DDF_Global_Spare);
a322f70c
DW
2236 pde->state = __cpu_to_be16(DDF_Online);
2237 get_dev_size(fd, NULL, &size);
2238 /* We are required to reserve 32Meg, and record the size in sectors */
2239 pde->config_size = __cpu_to_be64( (size - 32*1024*1024) / 512);
2240 sprintf(pde->path, "%17.17s","Information: nil") ;
2241 memset(pde->pad, 0xff, 6);
2242
d2ca6449 2243 dd->size = size >> 9;
2cc2983d
N
2244 if (st->update_tail) {
2245 dd->next = ddf->add_list;
2246 ddf->add_list = dd;
2247 } else {
2248 dd->next = ddf->dlist;
2249 ddf->dlist = dd;
2250 ddf->updates_pending = 1;
2251 }
f20c3968
DW
2252
2253 return 0;
a322f70c
DW
2254}
2255
2256/*
2257 * This is the write_init_super method for a ddf container. It is
2258 * called when creating a container or adding another device to a
2259 * container.
2260 */
2261
6416d527 2262static unsigned char null_conf[4096+512];
18a2f463 2263
7a7cc504 2264static int __write_init_super_ddf(struct supertype *st, int do_close)
a322f70c
DW
2265{
2266
2267 struct ddf_super *ddf = st->sb;
2268 int i;
2269 struct dl *d;
2270 int n_config;
2271 int conf_size;
175593bf
DW
2272 int attempts = 0;
2273 int successes = 0;
a322f70c
DW
2274 unsigned long long size, sector;
2275
175593bf
DW
2276 /* try to write updated metadata,
2277 * if we catch a failure move on to the next disk
2278 */
a322f70c
DW
2279 for (d = ddf->dlist; d; d=d->next) {
2280 int fd = d->fd;
2281
2282 if (fd < 0)
2283 continue;
2284
175593bf 2285 attempts++;
a322f70c
DW
2286 /* We need to fill in the primary, (secondary) and workspace
2287 * lba's in the headers, set their checksums,
2288 * Also checksum phys, virt....
2289 *
2290 * Then write everything out, finally the anchor is written.
2291 */
2292 get_dev_size(fd, NULL, &size);
2293 size /= 512;
2294 ddf->anchor.workspace_lba = __cpu_to_be64(size - 32*1024*2);
2295 ddf->anchor.primary_lba = __cpu_to_be64(size - 16*1024*2);
2296 ddf->anchor.seq = __cpu_to_be32(1);
2297 memcpy(&ddf->primary, &ddf->anchor, 512);
2298 memcpy(&ddf->secondary, &ddf->anchor, 512);
2299
2300 ddf->anchor.openflag = 0xFF; /* 'open' means nothing */
2301 ddf->anchor.seq = 0xFFFFFFFF; /* no sequencing in anchor */
2302 ddf->anchor.crc = calc_crc(&ddf->anchor, 512);
2303
2304 ddf->primary.openflag = 0;
2305 ddf->primary.type = DDF_HEADER_PRIMARY;
2306
2307 ddf->secondary.openflag = 0;
2308 ddf->secondary.type = DDF_HEADER_SECONDARY;
2309
2310 ddf->primary.crc = calc_crc(&ddf->primary, 512);
2311 ddf->secondary.crc = calc_crc(&ddf->secondary, 512);
2312
2313 sector = size - 16*1024*2;
2314 lseek64(fd, sector<<9, 0);
175593bf
DW
2315 if (write(fd, &ddf->primary, 512) < 0)
2316 continue;
a322f70c
DW
2317
2318 ddf->controller.crc = calc_crc(&ddf->controller, 512);
175593bf
DW
2319 if (write(fd, &ddf->controller, 512) < 0)
2320 continue;
a322f70c
DW
2321
2322 ddf->phys->crc = calc_crc(ddf->phys, ddf->pdsize);
2323
175593bf
DW
2324 if (write(fd, ddf->phys, ddf->pdsize) < 0)
2325 continue;
a322f70c
DW
2326
2327 ddf->virt->crc = calc_crc(ddf->virt, ddf->vdsize);
175593bf
DW
2328 if (write(fd, ddf->virt, ddf->vdsize) < 0)
2329 continue;
a322f70c
DW
2330
2331 /* Now write lots of config records. */
8c3b8c2c
NB
2332 n_config = ddf->max_part;
2333 conf_size = ddf->conf_rec_len * 512;
a322f70c
DW
2334 for (i = 0 ; i <= n_config ; i++) {
2335 struct vcl *c = d->vlist[i];
b2280677
NB
2336 if (i == n_config)
2337 c = (struct vcl*)d->spare;
a322f70c
DW
2338
2339 if (c) {
2340 c->conf.crc = calc_crc(&c->conf, conf_size);
175593bf
DW
2341 if (write(fd, &c->conf, conf_size) < 0)
2342 break;
a322f70c 2343 } else {
6416d527 2344 char *null_aligned = (char*)((((unsigned long)null_conf)+511)&~511UL);
18a2f463
NB
2345 if (null_conf[0] != 0xff)
2346 memset(null_conf, 0xff, sizeof(null_conf));
f21e18ca 2347 unsigned int togo = conf_size;
6416d527 2348 while (togo > sizeof(null_conf)-512) {
175593bf
DW
2349 if (write(fd, null_aligned, sizeof(null_conf)-512) < 0)
2350 break;
6416d527 2351 togo -= sizeof(null_conf)-512;
18a2f463 2352 }
175593bf
DW
2353 if (write(fd, null_aligned, togo) < 0)
2354 break;
a322f70c
DW
2355 }
2356 }
175593bf
DW
2357 if (i <= n_config)
2358 continue;
a322f70c 2359 d->disk.crc = calc_crc(&d->disk, 512);
175593bf
DW
2360 if (write(fd, &d->disk, 512) < 0)
2361 continue;
a322f70c
DW
2362
2363 /* Maybe do the same for secondary */
2364
2365 lseek64(fd, (size-1)*512, SEEK_SET);
175593bf
DW
2366 if (write(fd, &ddf->anchor, 512) < 0)
2367 continue;
2368 successes++;
2369 }
2370
2371 if (do_close)
2372 for (d = ddf->dlist; d; d=d->next) {
2373 close(d->fd);
7a7cc504
NB
2374 d->fd = -1;
2375 }
175593bf
DW
2376
2377 return attempts != successes;
a322f70c 2378}
7a7cc504
NB
2379
2380static int write_init_super_ddf(struct supertype *st)
2381{
9b1fb677
DW
2382 struct ddf_super *ddf = st->sb;
2383 struct vcl *currentconf = ddf->currentconf;
2384
2385 /* we are done with currentconf reset it to point st at the container */
2386 ddf->currentconf = NULL;
edd8d13c
NB
2387
2388 if (st->update_tail) {
2389 /* queue the virtual_disk and vd_config as metadata updates */
2390 struct virtual_disk *vd;
2391 struct vd_config *vc;
edd8d13c
NB
2392 int len;
2393
9b1fb677 2394 if (!currentconf) {
2cc2983d
N
2395 int len = (sizeof(struct phys_disk) +
2396 sizeof(struct phys_disk_entry));
2397
2398 /* adding a disk to the container. */
2399 if (!ddf->add_list)
2400 return 0;
2401
2402 append_metadata_update(st, ddf->add_list->mdupdate, len);
2403 ddf->add_list->mdupdate = NULL;
2404 return 0;
2405 }
2406
2407 /* Newly created VD */
2408
edd8d13c
NB
2409 /* First the virtual disk. We have a slightly fake header */
2410 len = sizeof(struct virtual_disk) + sizeof(struct virtual_entry);
2411 vd = malloc(len);
2412 *vd = *ddf->virt;
9b1fb677
DW
2413 vd->entries[0] = ddf->virt->entries[currentconf->vcnum];
2414 vd->populated_vdes = __cpu_to_be16(currentconf->vcnum);
edd8d13c
NB
2415 append_metadata_update(st, vd, len);
2416
2417 /* Then the vd_config */
2418 len = ddf->conf_rec_len * 512;
2419 vc = malloc(len);
9b1fb677 2420 memcpy(vc, &currentconf->conf, len);
edd8d13c
NB
2421 append_metadata_update(st, vc, len);
2422
2423 /* FIXME I need to close the fds! */
2424 return 0;
d682f344
N
2425 } else {
2426 struct dl *d;
2427 for (d = ddf->dlist; d; d=d->next)
2428 while (Kill(d->devname, NULL, 0, 1, 1) == 0);
edd8d13c 2429 return __write_init_super_ddf(st, 1);
d682f344 2430 }
7a7cc504
NB
2431}
2432
a322f70c
DW
2433#endif
2434
2435static __u64 avail_size_ddf(struct supertype *st, __u64 devsize)
2436{
2437 /* We must reserve the last 32Meg */
2438 if (devsize <= 32*1024*2)
2439 return 0;
2440 return devsize - 32*1024*2;
2441}
2442
2443#ifndef MDASSEMBLE
8592f29d
N
2444
2445static int reserve_space(struct supertype *st, int raiddisks,
2446 unsigned long long size, int chunk,
2447 unsigned long long *freesize)
2448{
2449 /* Find 'raiddisks' spare extents at least 'size' big (but
2450 * only caring about multiples of 'chunk') and remember
2451 * them.
2452 * If the cannot be found, fail.
2453 */
2454 struct dl *dl;
2455 struct ddf_super *ddf = st->sb;
2456 int cnt = 0;
2457
2458 for (dl = ddf->dlist; dl ; dl=dl->next) {
2459 dl->raiddisk = -1;
2460 dl->esize = 0;
2461 }
2462 /* Now find largest extent on each device */
2463 for (dl = ddf->dlist ; dl ; dl=dl->next) {
2464 struct extent *e = get_extents(ddf, dl);
2465 unsigned long long pos = 0;
2466 int i = 0;
2467 int found = 0;
2468 unsigned long long minsize = size;
2469
2470 if (size == 0)
2471 minsize = chunk;
2472
2473 if (!e)
2474 continue;
2475 do {
2476 unsigned long long esize;
2477 esize = e[i].start - pos;
2478 if (esize >= minsize) {
2479 found = 1;
2480 minsize = esize;
2481 }
2482 pos = e[i].start + e[i].size;
2483 i++;
2484 } while (e[i-1].size);
2485 if (found) {
2486 cnt++;
2487 dl->esize = minsize;
2488 }
2489 free(e);
2490 }
2491 if (cnt < raiddisks) {
2492 fprintf(stderr, Name ": not enough devices with space to create array.\n");
2493 return 0; /* No enough free spaces large enough */
2494 }
2495 if (size == 0) {
2496 /* choose the largest size of which there are at least 'raiddisk' */
2497 for (dl = ddf->dlist ; dl ; dl=dl->next) {
2498 struct dl *dl2;
2499 if (dl->esize <= size)
2500 continue;
2501 /* This is bigger than 'size', see if there are enough */
2502 cnt = 0;
2503 for (dl2 = dl; dl2 ; dl2=dl2->next)
2504 if (dl2->esize >= dl->esize)
2505 cnt++;
2506 if (cnt >= raiddisks)
2507 size = dl->esize;
2508 }
2509 if (chunk) {
2510 size = size / chunk;
2511 size *= chunk;
2512 }
2513 *freesize = size;
2514 if (size < 32) {
2515 fprintf(stderr, Name ": not enough spare devices to create array.\n");
2516 return 0;
2517 }
2518 }
2519 /* We have a 'size' of which there are enough spaces.
2520 * We simply do a first-fit */
2521 cnt = 0;
2522 for (dl = ddf->dlist ; dl && cnt < raiddisks ; dl=dl->next) {
2523 if (dl->esize < size)
2524 continue;
2525
2526 dl->raiddisk = cnt;
2527 cnt++;
2528 }
2529 return 1;
2530}
2531
2532
2533
2c514b71
NB
2534static int
2535validate_geometry_ddf_container(struct supertype *st,
2536 int level, int layout, int raiddisks,
2537 int chunk, unsigned long long size,
2538 char *dev, unsigned long long *freesize,
2539 int verbose);
78e44928
NB
2540
2541static int validate_geometry_ddf_bvd(struct supertype *st,
2542 int level, int layout, int raiddisks,
2543 int chunk, unsigned long long size,
2c514b71
NB
2544 char *dev, unsigned long long *freesize,
2545 int verbose);
78e44928
NB
2546
2547static int validate_geometry_ddf(struct supertype *st,
2c514b71
NB
2548 int level, int layout, int raiddisks,
2549 int chunk, unsigned long long size,
2550 char *dev, unsigned long long *freesize,
2551 int verbose)
a322f70c
DW
2552{
2553 int fd;
2554 struct mdinfo *sra;
2555 int cfd;
2556
2557 /* ddf potentially supports lots of things, but it depends on
2558 * what devices are offered (and maybe kernel version?)
2559 * If given unused devices, we will make a container.
2560 * If given devices in a container, we will make a BVD.
2561 * If given BVDs, we make an SVD, changing all the GUIDs in the process.
2562 */
2563
2564 if (level == LEVEL_CONTAINER) {
78e44928
NB
2565 /* Must be a fresh device to add to a container */
2566 return validate_geometry_ddf_container(st, level, layout,
2c514b71
NB
2567 raiddisks, chunk,
2568 size, dev, freesize,
2569 verbose);
5f8097be
NB
2570 }
2571
78e44928
NB
2572 if (!dev) {
2573 /* Initial sanity check. Exclude illegal levels. */
2574 int i;
2575 for (i=0; ddf_level_num[i].num1 != MAXINT; i++)
2576 if (ddf_level_num[i].num2 == level)
2577 break;
b42f577a
N
2578 if (ddf_level_num[i].num1 == MAXINT) {
2579 if (verbose)
2580 fprintf(stderr, Name ": DDF does not support level %d arrays\n",
2581 level);
78e44928 2582 return 0;
b42f577a 2583 }
78e44928 2584 /* Should check layout? etc */
8592f29d
N
2585
2586 if (st->sb && freesize) {
2587 /* --create was given a container to create in.
2588 * So we need to check that there are enough
2589 * free spaces and return the amount of space.
2590 * We may as well remember which drives were
2591 * chosen so that add_to_super/getinfo_super
2592 * can return them.
2593 */
2594 return reserve_space(st, raiddisks, size, chunk, freesize);
2595 }
a322f70c 2596 return 1;
78e44928 2597 }
a322f70c 2598
8592f29d
N
2599 if (st->sb) {
2600 /* A container has already been opened, so we are
2601 * creating in there. Maybe a BVD, maybe an SVD.
2602 * Should make a distinction one day.
2603 */
2604 return validate_geometry_ddf_bvd(st, level, layout, raiddisks,
2605 chunk, size, dev, freesize,
2606 verbose);
2607 }
78e44928
NB
2608 /* This is the first device for the array.
2609 * If it is a container, we read it in and do automagic allocations,
2610 * no other devices should be given.
2611 * Otherwise it must be a member device of a container, and we
2612 * do manual allocation.
2613 * Later we should check for a BVD and make an SVD.
a322f70c 2614 */
a322f70c
DW
2615 fd = open(dev, O_RDONLY|O_EXCL, 0);
2616 if (fd >= 0) {
2617 sra = sysfs_read(fd, 0, GET_VERSION);
2618 close(fd);
2619 if (sra && sra->array.major_version == -1 &&
78e44928
NB
2620 strcmp(sra->text_version, "ddf") == 0) {
2621
2622 /* load super */
2623 /* find space for 'n' devices. */
2624 /* remember the devices */
2625 /* Somehow return the fact that we have enough */
a322f70c
DW
2626 }
2627
2c514b71
NB
2628 if (verbose)
2629 fprintf(stderr,
2630 Name ": ddf: Cannot create this array "
b42f577a 2631 "on device %s - a container is required.\n",
2c514b71 2632 dev);
a322f70c
DW
2633 return 0;
2634 }
2635 if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) {
2c514b71
NB
2636 if (verbose)
2637 fprintf(stderr, Name ": ddf: Cannot open %s: %s\n",
2638 dev, strerror(errno));
a322f70c
DW
2639 return 0;
2640 }
2641 /* Well, it is in use by someone, maybe a 'ddf' container. */
2642 cfd = open_container(fd);
2643 if (cfd < 0) {
2644 close(fd);
2c514b71
NB
2645 if (verbose)
2646 fprintf(stderr, Name ": ddf: Cannot use %s: %s\n",
2647 dev, strerror(EBUSY));
a322f70c
DW
2648 return 0;
2649 }
2650 sra = sysfs_read(cfd, 0, GET_VERSION);
2651 close(fd);
2652 if (sra && sra->array.major_version == -1 &&
2653 strcmp(sra->text_version, "ddf") == 0) {
2654 /* This is a member of a ddf container. Load the container
2655 * and try to create a bvd
2656 */
2657 struct ddf_super *ddf;
a322f70c 2658 if (load_super_ddf_all(st, cfd, (void **)&ddf, NULL, 1) == 0) {
5f8097be 2659 st->sb = ddf;
2f6079dc 2660 st->container_dev = fd2devnum(cfd);
a322f70c 2661 close(cfd);
78e44928 2662 return validate_geometry_ddf_bvd(st, level, layout,
a322f70c 2663 raiddisks, chunk, size,
2c514b71
NB
2664 dev, freesize,
2665 verbose);
a322f70c
DW
2666 }
2667 close(cfd);
c42ec1ed
DW
2668 } else /* device may belong to a different container */
2669 return 0;
2670
a322f70c
DW
2671 return 1;
2672}
2673
2c514b71
NB
2674static int
2675validate_geometry_ddf_container(struct supertype *st,
2676 int level, int layout, int raiddisks,
2677 int chunk, unsigned long long size,
2678 char *dev, unsigned long long *freesize,
2679 int verbose)
a322f70c
DW
2680{
2681 int fd;
2682 unsigned long long ldsize;
2683
2684 if (level != LEVEL_CONTAINER)
2685 return 0;
2686 if (!dev)
2687 return 1;
2688
2689 fd = open(dev, O_RDONLY|O_EXCL, 0);
2690 if (fd < 0) {
2c514b71
NB
2691 if (verbose)
2692 fprintf(stderr, Name ": ddf: Cannot open %s: %s\n",
2693 dev, strerror(errno));
a322f70c
DW
2694 return 0;
2695 }
2696 if (!get_dev_size(fd, dev, &ldsize)) {
2697 close(fd);
2698 return 0;
2699 }
2700 close(fd);
2701
56aca704 2702 *freesize = avail_size_ddf(st, ldsize >> 9);
ea17e7aa
N
2703 if (*freesize == 0)
2704 return 0;
a322f70c
DW
2705
2706 return 1;
2707}
2708
78e44928
NB
2709static int validate_geometry_ddf_bvd(struct supertype *st,
2710 int level, int layout, int raiddisks,
2711 int chunk, unsigned long long size,
2c514b71
NB
2712 char *dev, unsigned long long *freesize,
2713 int verbose)
a322f70c
DW
2714{
2715 struct stat stb;
2716 struct ddf_super *ddf = st->sb;
2717 struct dl *dl;
5f8097be
NB
2718 unsigned long long pos = 0;
2719 unsigned long long maxsize;
2720 struct extent *e;
2721 int i;
a322f70c 2722 /* ddf/bvd supports lots of things, but not containers */
b42f577a
N
2723 if (level == LEVEL_CONTAINER) {
2724 if (verbose)
2725 fprintf(stderr, Name ": DDF cannot create a container within an container\n");
a322f70c 2726 return 0;
b42f577a 2727 }
a322f70c
DW
2728 /* We must have the container info already read in. */
2729 if (!ddf)
2730 return 0;
2731
5f8097be
NB
2732 if (!dev) {
2733 /* General test: make sure there is space for
2734 * 'raiddisks' device extents of size 'size'.
2735 */
2736 unsigned long long minsize = size;
2737 int dcnt = 0;
2738 if (minsize == 0)
2739 minsize = 8;
2740 for (dl = ddf->dlist; dl ; dl = dl->next)
2741 {
2742 int found = 0;
7e1432fb 2743 pos = 0;
5f8097be
NB
2744
2745 i = 0;
2746 e = get_extents(ddf, dl);
2747 if (!e) continue;
2748 do {
2749 unsigned long long esize;
2750 esize = e[i].start - pos;
2751 if (esize >= minsize)
2752 found = 1;
2753 pos = e[i].start + e[i].size;
2754 i++;
2755 } while (e[i-1].size);
2756 if (found)
2757 dcnt++;
2758 free(e);
2759 }
2760 if (dcnt < raiddisks) {
2c514b71
NB
2761 if (verbose)
2762 fprintf(stderr,
2763 Name ": ddf: Not enough devices with "
2764 "space for this array (%d < %d)\n",
2765 dcnt, raiddisks);
5f8097be
NB
2766 return 0;
2767 }
2768 return 1;
2769 }
a322f70c
DW
2770 /* This device must be a member of the set */
2771 if (stat(dev, &stb) < 0)
2772 return 0;
2773 if ((S_IFMT & stb.st_mode) != S_IFBLK)
2774 return 0;
2775 for (dl = ddf->dlist ; dl ; dl = dl->next) {
f21e18ca
N
2776 if (dl->major == (int)major(stb.st_rdev) &&
2777 dl->minor == (int)minor(stb.st_rdev))
a322f70c
DW
2778 break;
2779 }
5f8097be 2780 if (!dl) {
2c514b71
NB
2781 if (verbose)
2782 fprintf(stderr, Name ": ddf: %s is not in the "
2783 "same DDF set\n",
2784 dev);
5f8097be
NB
2785 return 0;
2786 }
2787 e = get_extents(ddf, dl);
2788 maxsize = 0;
2789 i = 0;
2790 if (e) do {
2791 unsigned long long esize;
2792 esize = e[i].start - pos;
2793 if (esize >= maxsize)
2794 maxsize = esize;
2795 pos = e[i].start + e[i].size;
2796 i++;
2797 } while (e[i-1].size);
2798 *freesize = maxsize;
a322f70c
DW
2799 // FIXME here I am
2800
2801 return 1;
2802}
59e36268 2803
a322f70c
DW
2804static int load_super_ddf_all(struct supertype *st, int fd,
2805 void **sbp, char *devname, int keep_fd)
2806{
2807 struct mdinfo *sra;
2808 struct ddf_super *super;
2809 struct mdinfo *sd, *best = NULL;
2810 int bestseq = 0;
2811 int seq;
2812 char nm[20];
2813 int dfd;
2814
b526e52d 2815 sra = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE);
a322f70c
DW
2816 if (!sra)
2817 return 1;
2818 if (sra->array.major_version != -1 ||
2819 sra->array.minor_version != -2 ||
2820 strcmp(sra->text_version, "ddf") != 0)
2821 return 1;
2822
6416d527 2823 if (posix_memalign((void**)&super, 512, sizeof(*super)) != 0)
a322f70c 2824 return 1;
a2349791 2825 memset(super, 0, sizeof(*super));
a322f70c
DW
2826
2827 /* first, try each device, and choose the best ddf */
2828 for (sd = sra->devs ; sd ; sd = sd->next) {
2829 int rv;
2830 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
7a7cc504
NB
2831 dfd = dev_open(nm, O_RDONLY);
2832 if (dfd < 0)
a322f70c
DW
2833 return 2;
2834 rv = load_ddf_headers(dfd, super, NULL);
7a7cc504 2835 close(dfd);
a322f70c
DW
2836 if (rv == 0) {
2837 seq = __be32_to_cpu(super->active->seq);
2838 if (super->active->openflag)
2839 seq--;
2840 if (!best || seq > bestseq) {
2841 bestseq = seq;
2842 best = sd;
2843 }
2844 }
2845 }
2846 if (!best)
2847 return 1;
2848 /* OK, load this ddf */
2849 sprintf(nm, "%d:%d", best->disk.major, best->disk.minor);
2850 dfd = dev_open(nm, O_RDONLY);
7a7cc504 2851 if (dfd < 0)
a322f70c
DW
2852 return 1;
2853 load_ddf_headers(dfd, super, NULL);
2854 load_ddf_global(dfd, super, NULL);
2855 close(dfd);
2856 /* Now we need the device-local bits */
2857 for (sd = sra->devs ; sd ; sd = sd->next) {
3d2c4fc7
DW
2858 int rv;
2859
a322f70c 2860 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
5f8097be 2861 dfd = dev_open(nm, keep_fd? O_RDWR : O_RDONLY);
7a7cc504 2862 if (dfd < 0)
a322f70c 2863 return 2;
3d2c4fc7
DW
2864 rv = load_ddf_headers(dfd, super, NULL);
2865 if (rv == 0)
2866 rv = load_ddf_local(dfd, super, NULL, keep_fd);
5f8097be 2867 if (!keep_fd) close(dfd);
3d2c4fc7
DW
2868 if (rv)
2869 return 1;
a322f70c 2870 }
f7e7067b 2871 if (st->subarray[0]) {
33414a01 2872 unsigned long val;
f7e7067b 2873 struct vcl *v;
33414a01
DW
2874 char *ep;
2875
2876 val = strtoul(st->subarray, &ep, 10);
2877 if (*ep != '\0') {
2878 free(super);
2879 return 1;
2880 }
f7e7067b
NB
2881
2882 for (v = super->conflist; v; v = v->next)
33414a01 2883 if (v->vcnum == val)
d2ca6449 2884 super->currentconf = v;
33414a01
DW
2885 if (!super->currentconf) {
2886 free(super);
f7e7067b 2887 return 1;
33414a01 2888 }
f7e7067b 2889 }
33414a01 2890
a322f70c
DW
2891 *sbp = super;
2892 if (st->ss == NULL) {
78e44928 2893 st->ss = &super_ddf;
a322f70c
DW
2894 st->minor_version = 0;
2895 st->max_devs = 512;
75aa18b5 2896 st->container_dev = fd2devnum(fd);
a322f70c 2897 }
352452c3 2898 st->loaded_container = 1;
a322f70c
DW
2899 return 0;
2900}
0e600426 2901#endif /* MDASSEMBLE */
a322f70c 2902
598f0d58
NB
2903static struct mdinfo *container_content_ddf(struct supertype *st)
2904{
2905 /* Given a container loaded by load_super_ddf_all,
2906 * extract information about all the arrays into
2907 * an mdinfo tree.
2908 *
2909 * For each vcl in conflist: create an mdinfo, fill it in,
2910 * then look for matching devices (phys_refnum) in dlist
2911 * and create appropriate device mdinfo.
2912 */
2913 struct ddf_super *ddf = st->sb;
2914 struct mdinfo *rest = NULL;
2915 struct vcl *vc;
2916
2917 for (vc = ddf->conflist ; vc ; vc=vc->next)
2918 {
f21e18ca
N
2919 unsigned int i;
2920 unsigned int j;
598f0d58
NB
2921 struct mdinfo *this;
2922 this = malloc(sizeof(*this));
2923 memset(this, 0, sizeof(*this));
2924 this->next = rest;
2925 rest = this;
2926
598f0d58
NB
2927 this->array.level = map_num1(ddf_level_num, vc->conf.prl);
2928 this->array.raid_disks =
2929 __be16_to_cpu(vc->conf.prim_elmnt_count);
60f18132
NB
2930 this->array.layout = rlq_to_layout(vc->conf.rlq, vc->conf.prl,
2931 this->array.raid_disks);
598f0d58 2932 this->array.md_minor = -1;
f35f2525
N
2933 this->array.major_version = -1;
2934 this->array.minor_version = -2;
598f0d58
NB
2935 this->array.ctime = DECADE +
2936 __be32_to_cpu(*(__u32*)(vc->conf.guid+16));
2937 this->array.utime = DECADE +
2938 __be32_to_cpu(vc->conf.timestamp);
2939 this->array.chunk_size = 512 << vc->conf.chunk_shift;
2940
59e36268 2941 i = vc->vcnum;
7a7cc504
NB
2942 if ((ddf->virt->entries[i].state & DDF_state_inconsistent) ||
2943 (ddf->virt->entries[i].init_state & DDF_initstate_mask) !=
ed9d66aa 2944 DDF_init_full) {
598f0d58 2945 this->array.state = 0;
ed9d66aa
NB
2946 this->resync_start = 0;
2947 } else {
598f0d58 2948 this->array.state = 1;
b7528a20 2949 this->resync_start = MaxSector;
ed9d66aa 2950 }
db42fa9b
N
2951 memcpy(this->name, ddf->virt->entries[i].name, 16);
2952 this->name[16]=0;
2953 for(j=0; j<16; j++)
2954 if (this->name[j] == ' ')
2955 this->name[j] = 0;
598f0d58
NB
2956
2957 memset(this->uuid, 0, sizeof(this->uuid));
2958 this->component_size = __be64_to_cpu(vc->conf.blocks);
2959 this->array.size = this->component_size / 2;
5f2aace8 2960 this->container_member = i;
598f0d58 2961
c5afc314
N
2962 ddf->currentconf = vc;
2963 uuid_from_super_ddf(st, this->uuid);
2964 ddf->currentconf = NULL;
2965
60f18132
NB
2966 sprintf(this->text_version, "/%s/%d",
2967 devnum2devname(st->container_dev),
2968 this->container_member);
2969
f21e18ca 2970 for (i = 0 ; i < ddf->mppe ; i++) {
598f0d58
NB
2971 struct mdinfo *dev;
2972 struct dl *d;
2973
2974 if (vc->conf.phys_refnum[i] == 0xFFFFFFFF)
2975 continue;
2976
c0d50ca5
N
2977 this->array.working_disks++;
2978
598f0d58
NB
2979 for (d = ddf->dlist; d ; d=d->next)
2980 if (d->disk.refnum == vc->conf.phys_refnum[i])
2981 break;
2982 if (d == NULL)
bc17324f
N
2983 /* Haven't found that one yet, maybe there are others */
2984 continue;
2985
598f0d58
NB
2986 dev = malloc(sizeof(*dev));
2987 memset(dev, 0, sizeof(*dev));
2988 dev->next = this->devs;
2989 this->devs = dev;
2990
2991 dev->disk.number = __be32_to_cpu(d->disk.refnum);
2992 dev->disk.major = d->major;
2993 dev->disk.minor = d->minor;
2994 dev->disk.raid_disk = i;
2995 dev->disk.state = (1<<MD_DISK_SYNC)|(1<<MD_DISK_ACTIVE);
d23534e4 2996 dev->recovery_start = MaxSector;
598f0d58 2997
120f7677
NB
2998 dev->events = __be32_to_cpu(ddf->primary.seq);
2999 dev->data_offset = __be64_to_cpu(vc->lba_offset[i]);
598f0d58
NB
3000 dev->component_size = __be64_to_cpu(vc->conf.blocks);
3001 if (d->devname)
3002 strcpy(dev->name, d->devname);
3003 }
3004 }
3005 return rest;
3006}
3007
955e9ea1 3008static int store_super_ddf(struct supertype *st, int fd)
a322f70c 3009{
955e9ea1 3010 struct ddf_super *ddf = st->sb;
a322f70c 3011 unsigned long long dsize;
6416d527 3012 void *buf;
3d2c4fc7 3013 int rc;
a322f70c 3014
955e9ea1
DW
3015 if (!ddf)
3016 return 1;
3017
3018 /* ->dlist and ->conflist will be set for updates, currently not
3019 * supported
3020 */
3021 if (ddf->dlist || ddf->conflist)
3022 return 1;
3023
a322f70c
DW
3024 if (!get_dev_size(fd, NULL, &dsize))
3025 return 1;
3026
3d2c4fc7
DW
3027 if (posix_memalign(&buf, 512, 512) != 0)
3028 return 1;
6416d527
NB
3029 memset(buf, 0, 512);
3030
a322f70c 3031 lseek64(fd, dsize-512, 0);
3d2c4fc7 3032 rc = write(fd, buf, 512);
6416d527 3033 free(buf);
3d2c4fc7
DW
3034 if (rc < 0)
3035 return 1;
a322f70c
DW
3036 return 0;
3037}
3038
a19c88b8
NB
3039static int compare_super_ddf(struct supertype *st, struct supertype *tst)
3040{
3041 /*
3042 * return:
3043 * 0 same, or first was empty, and second was copied
3044 * 1 second had wrong number
3045 * 2 wrong uuid
3046 * 3 wrong other info
3047 */
3048 struct ddf_super *first = st->sb;
3049 struct ddf_super *second = tst->sb;
3050
3051 if (!first) {
3052 st->sb = tst->sb;
3053 tst->sb = NULL;
3054 return 0;
3055 }
3056
3057 if (memcmp(first->anchor.guid, second->anchor.guid, DDF_GUID_LEN) != 0)
3058 return 2;
3059
3060 /* FIXME should I look at anything else? */
3061 return 0;
3062}
3063
0e600426 3064#ifndef MDASSEMBLE
4e5528c6
NB
3065/*
3066 * A new array 'a' has been started which claims to be instance 'inst'
3067 * within container 'c'.
3068 * We need to confirm that the array matches the metadata in 'c' so
3069 * that we don't corrupt any metadata.
3070 */
cba0191b 3071static int ddf_open_new(struct supertype *c, struct active_array *a, char *inst)
549e9569 3072{
2c514b71 3073 dprintf("ddf: open_new %s\n", inst);
cba0191b 3074 a->info.container_member = atoi(inst);
549e9569
NB
3075 return 0;
3076}
3077
4e5528c6
NB
3078/*
3079 * The array 'a' is to be marked clean in the metadata.
ed9d66aa 3080 * If '->resync_start' is not ~(unsigned long long)0, then the array is only
4e5528c6
NB
3081 * clean up to the point (in sectors). If that cannot be recorded in the
3082 * metadata, then leave it as dirty.
3083 *
3084 * For DDF, we need to clear the DDF_state_inconsistent bit in the
3085 * !global! virtual_disk.virtual_entry structure.
3086 */
01f157d7 3087static int ddf_set_array_state(struct active_array *a, int consistent)
549e9569 3088{
4e5528c6
NB
3089 struct ddf_super *ddf = a->container->sb;
3090 int inst = a->info.container_member;
18a2f463 3091 int old = ddf->virt->entries[inst].state;
01f157d7
N
3092 if (consistent == 2) {
3093 /* Should check if a recovery should be started FIXME */
3094 consistent = 1;
b7941fd6 3095 if (!is_resync_complete(&a->info))
01f157d7
N
3096 consistent = 0;
3097 }
ed9d66aa
NB
3098 if (consistent)
3099 ddf->virt->entries[inst].state &= ~DDF_state_inconsistent;
3100 else
4e5528c6 3101 ddf->virt->entries[inst].state |= DDF_state_inconsistent;
18a2f463
NB
3102 if (old != ddf->virt->entries[inst].state)
3103 ddf->updates_pending = 1;
3104
3105 old = ddf->virt->entries[inst].init_state;
ed9d66aa 3106 ddf->virt->entries[inst].init_state &= ~DDF_initstate_mask;
b7941fd6 3107 if (is_resync_complete(&a->info))
ed9d66aa 3108 ddf->virt->entries[inst].init_state |= DDF_init_full;
b7941fd6 3109 else if (a->info.resync_start == 0)
ed9d66aa 3110 ddf->virt->entries[inst].init_state |= DDF_init_not;
4e5528c6 3111 else
ed9d66aa 3112 ddf->virt->entries[inst].init_state |= DDF_init_quick;
18a2f463
NB
3113 if (old != ddf->virt->entries[inst].init_state)
3114 ddf->updates_pending = 1;
ed9d66aa 3115
2c514b71 3116 dprintf("ddf mark %d %s %llu\n", inst, consistent?"clean":"dirty",
b7941fd6 3117 a->info.resync_start);
01f157d7 3118 return consistent;
fd7cde1b
DW
3119}
3120
7a7cc504
NB
3121/*
3122 * The state of each disk is stored in the global phys_disk structure
3123 * in phys_disk.entries[n].state.
3124 * This makes various combinations awkward.
3125 * - When a device fails in any array, it must be failed in all arrays
3126 * that include a part of this device.
3127 * - When a component is rebuilding, we cannot include it officially in the
3128 * array unless this is the only array that uses the device.
3129 *
3130 * So: when transitioning:
3131 * Online -> failed, just set failed flag. monitor will propagate
3132 * spare -> online, the device might need to be added to the array.
3133 * spare -> failed, just set failed. Don't worry if in array or not.
3134 */
8d45d196 3135static void ddf_set_disk(struct active_array *a, int n, int state)
549e9569 3136{
7a7cc504 3137 struct ddf_super *ddf = a->container->sb;
f21e18ca 3138 unsigned int inst = a->info.container_member;
7a7cc504
NB
3139 struct vd_config *vc = find_vdcr(ddf, inst);
3140 int pd = find_phys(ddf, vc->phys_refnum[n]);
3141 int i, st, working;
3142
3143 if (vc == NULL) {
2c514b71 3144 dprintf("ddf: cannot find instance %d!!\n", inst);
7a7cc504
NB
3145 return;
3146 }
3147 if (pd < 0) {
3148 /* disk doesn't currently exist. If it is now in_sync,
3149 * insert it. */
3150 if ((state & DS_INSYNC) && ! (state & DS_FAULTY)) {
3151 /* Find dev 'n' in a->info->devs, determine the
3152 * ddf refnum, and set vc->phys_refnum and update
3153 * phys->entries[]
3154 */
3155 /* FIXME */
3156 }
3157 } else {
18a2f463 3158 int old = ddf->phys->entries[pd].state;
7a7cc504
NB
3159 if (state & DS_FAULTY)
3160 ddf->phys->entries[pd].state |= __cpu_to_be16(DDF_Failed);
3161 if (state & DS_INSYNC) {
3162 ddf->phys->entries[pd].state |= __cpu_to_be16(DDF_Online);
3163 ddf->phys->entries[pd].state &= __cpu_to_be16(~DDF_Rebuilding);
3164 }
18a2f463
NB
3165 if (old != ddf->phys->entries[pd].state)
3166 ddf->updates_pending = 1;
7a7cc504
NB
3167 }
3168
2c514b71 3169 dprintf("ddf: set_disk %d to %x\n", n, state);
7e1432fb 3170
7a7cc504
NB
3171 /* Now we need to check the state of the array and update
3172 * virtual_disk.entries[n].state.
3173 * It needs to be one of "optimal", "degraded", "failed".
3174 * I don't understand 'deleted' or 'missing'.
3175 */
3176 working = 0;
3177 for (i=0; i < a->info.array.raid_disks; i++) {
3178 pd = find_phys(ddf, vc->phys_refnum[i]);
3179 if (pd < 0)
3180 continue;
57632f4a
NB
3181 st = __be16_to_cpu(ddf->phys->entries[pd].state);
3182 if ((st & (DDF_Online|DDF_Failed|DDF_Rebuilding))
7a7cc504
NB
3183 == DDF_Online)
3184 working++;
3185 }
3186 state = DDF_state_degraded;
3187 if (working == a->info.array.raid_disks)
3188 state = DDF_state_optimal;
3189 else switch(vc->prl) {
3190 case DDF_RAID0:
3191 case DDF_CONCAT:
3192 case DDF_JBOD:
3193 state = DDF_state_failed;
3194 break;
3195 case DDF_RAID1:
3196 if (working == 0)
3197 state = DDF_state_failed;
3198 break;
3199 case DDF_RAID4:
3200 case DDF_RAID5:
3201 if (working < a->info.array.raid_disks-1)
3202 state = DDF_state_failed;
3203 break;
3204 case DDF_RAID6:
3205 if (working < a->info.array.raid_disks-2)
3206 state = DDF_state_failed;
3207 else if (working == a->info.array.raid_disks-1)
3208 state = DDF_state_part_optimal;
3209 break;
3210 }
3211
18a2f463
NB
3212 if (ddf->virt->entries[inst].state !=
3213 ((ddf->virt->entries[inst].state & ~DDF_state_mask)
3214 | state)) {
3215
3216 ddf->virt->entries[inst].state =
3217 (ddf->virt->entries[inst].state & ~DDF_state_mask)
3218 | state;
3219 ddf->updates_pending = 1;
3220 }
7a7cc504 3221
549e9569
NB
3222}
3223
2e735d19 3224static void ddf_sync_metadata(struct supertype *st)
549e9569 3225{
7a7cc504
NB
3226
3227 /*
3228 * Write all data to all devices.
3229 * Later, we might be able to track whether only local changes
3230 * have been made, or whether any global data has been changed,
3231 * but ddf is sufficiently weird that it probably always
3232 * changes global data ....
3233 */
18a2f463
NB
3234 struct ddf_super *ddf = st->sb;
3235 if (!ddf->updates_pending)
3236 return;
3237 ddf->updates_pending = 0;
2e735d19 3238 __write_init_super_ddf(st, 0);
2c514b71 3239 dprintf("ddf: sync_metadata\n");
549e9569
NB
3240}
3241
88c164f4
NB
3242static void ddf_process_update(struct supertype *st,
3243 struct metadata_update *update)
3244{
3245 /* Apply this update to the metadata.
3246 * The first 4 bytes are a DDF_*_MAGIC which guides
3247 * our actions.
3248 * Possible update are:
3249 * DDF_PHYS_RECORDS_MAGIC
3250 * Add a new physical device. Changes to this record
3251 * only happen implicitly.
3252 * used_pdes is the device number.
3253 * DDF_VIRT_RECORDS_MAGIC
3254 * Add a new VD. Possibly also change the 'access' bits.
3255 * populated_vdes is the entry number.
3256 * DDF_VD_CONF_MAGIC
3257 * New or updated VD. the VIRT_RECORD must already
3258 * exist. For an update, phys_refnum and lba_offset
3259 * (at least) are updated, and the VD_CONF must
3260 * be written to precisely those devices listed with
3261 * a phys_refnum.
3262 * DDF_SPARE_ASSIGN_MAGIC
3263 * replacement Spare Assignment Record... but for which device?
3264 *
3265 * So, e.g.:
3266 * - to create a new array, we send a VIRT_RECORD and
3267 * a VD_CONF. Then assemble and start the array.
3268 * - to activate a spare we send a VD_CONF to add the phys_refnum
3269 * and offset. This will also mark the spare as active with
3270 * a spare-assignment record.
3271 */
3272 struct ddf_super *ddf = st->sb;
3273 __u32 *magic = (__u32*)update->buf;
3274 struct phys_disk *pd;
3275 struct virtual_disk *vd;
3276 struct vd_config *vc;
3277 struct vcl *vcl;
3278 struct dl *dl;
f21e18ca
N
3279 unsigned int mppe;
3280 unsigned int ent;
88c164f4 3281
2c514b71 3282 dprintf("Process update %x\n", *magic);
7e1432fb 3283
88c164f4
NB
3284 switch (*magic) {
3285 case DDF_PHYS_RECORDS_MAGIC:
3286
3287 if (update->len != (sizeof(struct phys_disk) +
3288 sizeof(struct phys_disk_entry)))
3289 return;
3290 pd = (struct phys_disk*)update->buf;
3291
3292 ent = __be16_to_cpu(pd->used_pdes);
3293 if (ent >= __be16_to_cpu(ddf->phys->max_pdes))
3294 return;
3295 if (!all_ff(ddf->phys->entries[ent].guid))
3296 return;
3297 ddf->phys->entries[ent] = pd->entries[0];
3298 ddf->phys->used_pdes = __cpu_to_be16(1 +
3299 __be16_to_cpu(ddf->phys->used_pdes));
18a2f463 3300 ddf->updates_pending = 1;
2cc2983d
N
3301 if (ddf->add_list) {
3302 struct active_array *a;
3303 struct dl *al = ddf->add_list;
3304 ddf->add_list = al->next;
3305
3306 al->next = ddf->dlist;
3307 ddf->dlist = al;
3308
3309 /* As a device has been added, we should check
3310 * for any degraded devices that might make
3311 * use of this spare */
3312 for (a = st->arrays ; a; a=a->next)
3313 a->check_degraded = 1;
3314 }
88c164f4
NB
3315 break;
3316
3317 case DDF_VIRT_RECORDS_MAGIC:
3318
3319 if (update->len != (sizeof(struct virtual_disk) +
3320 sizeof(struct virtual_entry)))
3321 return;
3322 vd = (struct virtual_disk*)update->buf;
3323
3324 ent = __be16_to_cpu(vd->populated_vdes);
3325 if (ent >= __be16_to_cpu(ddf->virt->max_vdes))
3326 return;
3327 if (!all_ff(ddf->virt->entries[ent].guid))
3328 return;
3329 ddf->virt->entries[ent] = vd->entries[0];
3330 ddf->virt->populated_vdes = __cpu_to_be16(1 +
3331 __be16_to_cpu(ddf->virt->populated_vdes));
18a2f463 3332 ddf->updates_pending = 1;
88c164f4
NB
3333 break;
3334
3335 case DDF_VD_CONF_MAGIC:
2c514b71 3336 dprintf("len %d %d\n", update->len, ddf->conf_rec_len);
88c164f4
NB
3337
3338 mppe = __be16_to_cpu(ddf->anchor.max_primary_element_entries);
f21e18ca 3339 if ((unsigned)update->len != ddf->conf_rec_len * 512)
88c164f4
NB
3340 return;
3341 vc = (struct vd_config*)update->buf;
3342 for (vcl = ddf->conflist; vcl ; vcl = vcl->next)
3343 if (memcmp(vcl->conf.guid, vc->guid, DDF_GUID_LEN) == 0)
3344 break;
2c514b71 3345 dprintf("vcl = %p\n", vcl);
88c164f4
NB
3346 if (vcl) {
3347 /* An update, just copy the phys_refnum and lba_offset
3348 * fields
3349 */
3350 memcpy(vcl->conf.phys_refnum, vc->phys_refnum,
3351 mppe * (sizeof(__u32) + sizeof(__u64)));
3352 } else {
3353 /* A new VD_CONF */
e6b9548d
DW
3354 if (!update->space)
3355 return;
88c164f4
NB
3356 vcl = update->space;
3357 update->space = NULL;
3358 vcl->next = ddf->conflist;
edd8d13c 3359 memcpy(&vcl->conf, vc, update->len);
88c164f4
NB
3360 vcl->lba_offset = (__u64*)
3361 &vcl->conf.phys_refnum[mppe];
3362 ddf->conflist = vcl;
3363 }
3364 /* Now make sure vlist is correct for each dl. */
3365 for (dl = ddf->dlist; dl; dl = dl->next) {
f21e18ca
N
3366 unsigned int dn;
3367 unsigned int vn = 0;
88c164f4
NB
3368 for (vcl = ddf->conflist; vcl ; vcl = vcl->next)
3369 for (dn=0; dn < ddf->mppe ; dn++)
3370 if (vcl->conf.phys_refnum[dn] ==
3371 dl->disk.refnum) {
2c514b71
NB
3372 dprintf("dev %d has %p at %d\n",
3373 dl->pdnum, vcl, vn);
88c164f4
NB
3374 dl->vlist[vn++] = vcl;
3375 break;
3376 }
3377 while (vn < ddf->max_part)
3378 dl->vlist[vn++] = NULL;
7e1432fb
NB
3379 if (dl->vlist[0]) {
3380 ddf->phys->entries[dl->pdnum].type &=
3381 ~__cpu_to_be16(DDF_Global_Spare);
3382 ddf->phys->entries[dl->pdnum].type |=
3383 __cpu_to_be16(DDF_Active_in_VD);
3384 }
3385 if (dl->spare) {
3386 ddf->phys->entries[dl->pdnum].type &=
3387 ~__cpu_to_be16(DDF_Global_Spare);
3388 ddf->phys->entries[dl->pdnum].type |=
3389 __cpu_to_be16(DDF_Spare);
3390 }
3391 if (!dl->vlist[0] && !dl->spare) {
3392 ddf->phys->entries[dl->pdnum].type |=
3393 __cpu_to_be16(DDF_Global_Spare);
3394 ddf->phys->entries[dl->pdnum].type &=
3395 ~__cpu_to_be16(DDF_Spare |
3396 DDF_Active_in_VD);
3397 }
88c164f4 3398 }
18a2f463 3399 ddf->updates_pending = 1;
88c164f4
NB
3400 break;
3401 case DDF_SPARE_ASSIGN_MAGIC:
3402 default: break;
3403 }
3404}
3405
edd8d13c
NB
3406static void ddf_prepare_update(struct supertype *st,
3407 struct metadata_update *update)
3408{
3409 /* This update arrived at managemon.
3410 * We are about to pass it to monitor.
3411 * If a malloc is needed, do it here.
3412 */
3413 struct ddf_super *ddf = st->sb;
3414 __u32 *magic = (__u32*)update->buf;
3415 if (*magic == DDF_VD_CONF_MAGIC)
e6b9548d 3416 if (posix_memalign(&update->space, 512,
6416d527 3417 offsetof(struct vcl, conf)
e6b9548d
DW
3418 + ddf->conf_rec_len * 512) != 0)
3419 update->space = NULL;
edd8d13c
NB
3420}
3421
7e1432fb
NB
3422/*
3423 * Check if the array 'a' is degraded but not failed.
3424 * If it is, find as many spares as are available and needed and
3425 * arrange for their inclusion.
3426 * We only choose devices which are not already in the array,
3427 * and prefer those with a spare-assignment to this array.
3428 * otherwise we choose global spares - assuming always that
3429 * there is enough room.
3430 * For each spare that we assign, we return an 'mdinfo' which
3431 * describes the position for the device in the array.
3432 * We also add to 'updates' a DDF_VD_CONF_MAGIC update with
3433 * the new phys_refnum and lba_offset values.
3434 *
3435 * Only worry about BVDs at the moment.
3436 */
3437static struct mdinfo *ddf_activate_spare(struct active_array *a,
3438 struct metadata_update **updates)
3439{
3440 int working = 0;
3441 struct mdinfo *d;
3442 struct ddf_super *ddf = a->container->sb;
3443 int global_ok = 0;
3444 struct mdinfo *rv = NULL;
3445 struct mdinfo *di;
3446 struct metadata_update *mu;
3447 struct dl *dl;
3448 int i;
3449 struct vd_config *vc;
3450 __u64 *lba;
3451
7e1432fb
NB
3452 for (d = a->info.devs ; d ; d = d->next) {
3453 if ((d->curr_state & DS_FAULTY) &&
3454 d->state_fd >= 0)
3455 /* wait for Removal to happen */
3456 return NULL;
3457 if (d->state_fd >= 0)
3458 working ++;
3459 }
3460
2c514b71
NB
3461 dprintf("ddf_activate: working=%d (%d) level=%d\n", working, a->info.array.raid_disks,
3462 a->info.array.level);
7e1432fb
NB
3463 if (working == a->info.array.raid_disks)
3464 return NULL; /* array not degraded */
3465 switch (a->info.array.level) {
3466 case 1:
3467 if (working == 0)
3468 return NULL; /* failed */
3469 break;
3470 case 4:
3471 case 5:
3472 if (working < a->info.array.raid_disks - 1)
3473 return NULL; /* failed */
3474 break;
3475 case 6:
3476 if (working < a->info.array.raid_disks - 2)
3477 return NULL; /* failed */
3478 break;
3479 default: /* concat or stripe */
3480 return NULL; /* failed */
3481 }
3482
3483 /* For each slot, if it is not working, find a spare */
3484 dl = ddf->dlist;
3485 for (i = 0; i < a->info.array.raid_disks; i++) {
3486 for (d = a->info.devs ; d ; d = d->next)
3487 if (d->disk.raid_disk == i)
3488 break;
2c514b71 3489 dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0);
7e1432fb
NB
3490 if (d && (d->state_fd >= 0))
3491 continue;
3492
3493 /* OK, this device needs recovery. Find a spare */
3494 again:
3495 for ( ; dl ; dl = dl->next) {
3496 unsigned long long esize;
3497 unsigned long long pos;
3498 struct mdinfo *d2;
3499 int is_global = 0;
3500 int is_dedicated = 0;
3501 struct extent *ex;
f21e18ca 3502 unsigned int j;
7e1432fb
NB
3503 /* If in this array, skip */
3504 for (d2 = a->info.devs ; d2 ; d2 = d2->next)
3505 if (d2->disk.major == dl->major &&
3506 d2->disk.minor == dl->minor) {
2c514b71 3507 dprintf("%x:%x already in array\n", dl->major, dl->minor);
7e1432fb
NB
3508 break;
3509 }
3510 if (d2)
3511 continue;
3512 if (ddf->phys->entries[dl->pdnum].type &
3513 __cpu_to_be16(DDF_Spare)) {
3514 /* Check spare assign record */
3515 if (dl->spare) {
3516 if (dl->spare->type & DDF_spare_dedicated) {
3517 /* check spare_ents for guid */
3518 for (j = 0 ;
3519 j < __be16_to_cpu(dl->spare->populated);
3520 j++) {
3521 if (memcmp(dl->spare->spare_ents[j].guid,
3522 ddf->virt->entries[a->info.container_member].guid,
3523 DDF_GUID_LEN) == 0)
3524 is_dedicated = 1;
3525 }
3526 } else
3527 is_global = 1;
3528 }
3529 } else if (ddf->phys->entries[dl->pdnum].type &
3530 __cpu_to_be16(DDF_Global_Spare)) {
3531 is_global = 1;
3532 }
3533 if ( ! (is_dedicated ||
3534 (is_global && global_ok))) {
2c514b71 3535 dprintf("%x:%x not suitable: %d %d\n", dl->major, dl->minor,
7e1432fb
NB
3536 is_dedicated, is_global);
3537 continue;
3538 }
3539
3540 /* We are allowed to use this device - is there space?
3541 * We need a->info.component_size sectors */
3542 ex = get_extents(ddf, dl);
3543 if (!ex) {
2c514b71 3544 dprintf("cannot get extents\n");
7e1432fb
NB
3545 continue;
3546 }
3547 j = 0; pos = 0;
3548 esize = 0;
3549
3550 do {
3551 esize = ex[j].start - pos;
3552 if (esize >= a->info.component_size)
3553 break;
3554 pos = ex[i].start + ex[i].size;
3555 i++;
3556 } while (ex[i-1].size);
3557
3558 free(ex);
3559 if (esize < a->info.component_size) {
2c514b71
NB
3560 dprintf("%x:%x has no room: %llu %llu\n", dl->major, dl->minor,
3561 esize, a->info.component_size);
7e1432fb
NB
3562 /* No room */
3563 continue;
3564 }
3565
3566 /* Cool, we have a device with some space at pos */
3567 di = malloc(sizeof(*di));
79244939
DW
3568 if (!di)
3569 continue;
7e1432fb
NB
3570 memset(di, 0, sizeof(*di));
3571 di->disk.number = i;
3572 di->disk.raid_disk = i;
3573 di->disk.major = dl->major;
3574 di->disk.minor = dl->minor;
3575 di->disk.state = 0;
d23534e4 3576 di->recovery_start = 0;
7e1432fb
NB
3577 di->data_offset = pos;
3578 di->component_size = a->info.component_size;
3579 di->container_member = dl->pdnum;
3580 di->next = rv;
3581 rv = di;
2c514b71
NB
3582 dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor,
3583 i, pos);
7e1432fb
NB
3584
3585 break;
3586 }
3587 if (!dl && ! global_ok) {
3588 /* not enough dedicated spares, try global */
3589 global_ok = 1;
3590 dl = ddf->dlist;
3591 goto again;
3592 }
3593 }
3594
3595 if (!rv)
3596 /* No spares found */
3597 return rv;
3598 /* Now 'rv' has a list of devices to return.
3599 * Create a metadata_update record to update the
3600 * phys_refnum and lba_offset values
3601 */
904c1ef7 3602 mu = malloc(sizeof(*mu));
79244939
DW
3603 if (mu && posix_memalign(&mu->space, 512, sizeof(struct vcl)) != 0) {
3604 free(mu);
3605 mu = NULL;
3606 }
3607 if (!mu) {
3608 while (rv) {
3609 struct mdinfo *n = rv->next;
3610
3611 free(rv);
3612 rv = n;
3613 }
3614 return NULL;
3615 }
3616
904c1ef7 3617 mu->buf = malloc(ddf->conf_rec_len * 512);
7e1432fb
NB
3618 mu->len = ddf->conf_rec_len;
3619 mu->next = *updates;
3620 vc = find_vdcr(ddf, a->info.container_member);
3621 memcpy(mu->buf, vc, ddf->conf_rec_len * 512);
3622
3623 vc = (struct vd_config*)mu->buf;
3624 lba = (__u64*)&vc->phys_refnum[ddf->mppe];
3625 for (di = rv ; di ; di = di->next) {
3626 vc->phys_refnum[di->disk.raid_disk] =
3627 ddf->phys->entries[dl->pdnum].refnum;
3628 lba[di->disk.raid_disk] = di->data_offset;
3629 }
3630 *updates = mu;
3631 return rv;
3632}
0e600426 3633#endif /* MDASSEMBLE */
7e1432fb 3634
b640a252
N
3635static int ddf_level_to_layout(int level)
3636{
3637 switch(level) {
3638 case 0:
3639 case 1:
3640 return 0;
3641 case 5:
3642 return ALGORITHM_LEFT_SYMMETRIC;
3643 case 6:
3644 return ALGORITHM_ROTATING_N_CONTINUE;
3645 case 10:
3646 return 0x102;
3647 default:
3648 return UnSet;
3649 }
3650}
3651
a322f70c
DW
3652struct superswitch super_ddf = {
3653#ifndef MDASSEMBLE
3654 .examine_super = examine_super_ddf,
3655 .brief_examine_super = brief_examine_super_ddf,
4737ae25 3656 .brief_examine_subarrays = brief_examine_subarrays_ddf,
bceedeec 3657 .export_examine_super = export_examine_super_ddf,
a322f70c
DW
3658 .detail_super = detail_super_ddf,
3659 .brief_detail_super = brief_detail_super_ddf,
3660 .validate_geometry = validate_geometry_ddf,
78e44928 3661 .write_init_super = write_init_super_ddf,
0e600426 3662 .add_to_super = add_to_super_ddf,
a322f70c
DW
3663#endif
3664 .match_home = match_home_ddf,
3665 .uuid_from_super= uuid_from_super_ddf,
3666 .getinfo_super = getinfo_super_ddf,
3667 .update_super = update_super_ddf,
3668
3669 .avail_size = avail_size_ddf,
3670
a19c88b8
NB
3671 .compare_super = compare_super_ddf,
3672
a322f70c 3673 .load_super = load_super_ddf,
ba7eb04f 3674 .init_super = init_super_ddf,
955e9ea1 3675 .store_super = store_super_ddf,
a322f70c
DW
3676 .free_super = free_super_ddf,
3677 .match_metadata_desc = match_metadata_desc_ddf,
78e44928 3678 .container_content = container_content_ddf,
b640a252 3679 .default_layout = ddf_level_to_layout,
a322f70c 3680
a322f70c 3681 .external = 1,
549e9569 3682
0e600426 3683#ifndef MDASSEMBLE
549e9569
NB
3684/* for mdmon */
3685 .open_new = ddf_open_new,
ed9d66aa 3686 .set_array_state= ddf_set_array_state,
549e9569
NB
3687 .set_disk = ddf_set_disk,
3688 .sync_metadata = ddf_sync_metadata,
88c164f4 3689 .process_update = ddf_process_update,
edd8d13c 3690 .prepare_update = ddf_prepare_update,
7e1432fb 3691 .activate_spare = ddf_activate_spare,
0e600426 3692#endif
4cce4069 3693 .name = "ddf",
a322f70c 3694};