]> git.ipfire.org Git - thirdparty/mdadm.git/blame - super-ddf.c
imsm: fix spare promotion
[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;
8c3b8c2c 399 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 */
409 int vcnum; /* index into ->virt */
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
DW
442
443static int calc_crc(void *buf, int len)
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
525 if (lseek64(fd, offset<<9, 0) != (offset<<9)) {
526 if (dofree)
527 free(buf);
528 return NULL;
529 }
530 if (read(fd, buf, len<<9) != (len<<9)) {
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;
645 int i;
e223334f 646 int confsec;
b2280677 647 int vnum;
59e36268 648 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
NB
675 dl->spare = NULL;
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;
5575e7d9
NB
680 for (i=0; i < __be16_to_cpu(super->active->max_pd_entries); i++)
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
786 /* 32M is a lower bound */
787 if (dsize <= 32*1024*1024) {
97320d7c 788 if (devname)
a322f70c
DW
789 fprintf(stderr,
790 Name ": %s is too small for ddf: "
791 "size is %llu sectors.\n",
792 devname, dsize>>9);
97320d7c 793 return 1;
a322f70c
DW
794 }
795 if (dsize & 511) {
97320d7c 796 if (devname)
a322f70c
DW
797 fprintf(stderr,
798 Name ": %s is an odd size for ddf: "
799 "size is %llu bytes.\n",
800 devname, dsize);
97320d7c 801 return 1;
a322f70c
DW
802 }
803
37424f13
DW
804 free_super_ddf(st);
805
6416d527 806 if (posix_memalign((void**)&super, 512, sizeof(*super))!= 0) {
a322f70c
DW
807 fprintf(stderr, Name ": malloc of %zu failed.\n",
808 sizeof(*super));
809 return 1;
810 }
a2349791 811 memset(super, 0, sizeof(*super));
a322f70c
DW
812
813 rv = load_ddf_headers(fd, super, devname);
814 if (rv) {
815 free(super);
816 return rv;
817 }
818
819 /* Have valid headers and have chosen the best. Let's read in the rest*/
820
821 rv = load_ddf_global(fd, super, devname);
822
823 if (rv) {
824 if (devname)
825 fprintf(stderr,
826 Name ": Failed to load all information "
827 "sections on %s\n", devname);
828 free(super);
829 return rv;
830 }
831
3d2c4fc7
DW
832 rv = load_ddf_local(fd, super, devname, 0);
833
834 if (rv) {
835 if (devname)
836 fprintf(stderr,
837 Name ": Failed to load all information "
838 "sections on %s\n", devname);
839 free(super);
840 return rv;
841 }
a322f70c 842
af99d9ca
DW
843 if (st->subarray[0]) {
844 struct vcl *v;
845
846 for (v = super->conflist; v; v = v->next)
847 if (v->vcnum == atoi(st->subarray))
848 super->currentconf = v;
849 if (!super->currentconf) {
850 free(super);
851 return 1;
852 }
853 }
854
a322f70c
DW
855 /* Should possibly check the sections .... */
856
857 st->sb = super;
858 if (st->ss == NULL) {
859 st->ss = &super_ddf;
860 st->minor_version = 0;
861 st->max_devs = 512;
862 }
352452c3 863 st->loaded_container = 0;
a322f70c
DW
864 return 0;
865
866}
867
868static void free_super_ddf(struct supertype *st)
869{
870 struct ddf_super *ddf = st->sb;
871 if (ddf == NULL)
872 return;
873 free(ddf->phys);
874 free(ddf->virt);
875 while (ddf->conflist) {
876 struct vcl *v = ddf->conflist;
877 ddf->conflist = v->next;
59e36268
NB
878 if (v->block_sizes)
879 free(v->block_sizes);
a322f70c
DW
880 free(v);
881 }
882 while (ddf->dlist) {
883 struct dl *d = ddf->dlist;
884 ddf->dlist = d->next;
885 if (d->fd >= 0)
886 close(d->fd);
b2280677
NB
887 if (d->spare)
888 free(d->spare);
a322f70c
DW
889 free(d);
890 }
891 free(ddf);
892 st->sb = NULL;
893}
894
895static struct supertype *match_metadata_desc_ddf(char *arg)
896{
897 /* 'ddf' only support containers */
898 struct supertype *st;
899 if (strcmp(arg, "ddf") != 0 &&
900 strcmp(arg, "default") != 0
901 )
902 return NULL;
903
904 st = malloc(sizeof(*st));
ef609477 905 memset(st, 0, sizeof(*st));
a322f70c
DW
906 st->ss = &super_ddf;
907 st->max_devs = 512;
908 st->minor_version = 0;
909 st->sb = NULL;
910 return st;
911}
912
a322f70c
DW
913
914#ifndef MDASSEMBLE
915
916static mapping_t ddf_state[] = {
917 { "Optimal", 0},
918 { "Degraded", 1},
919 { "Deleted", 2},
920 { "Missing", 3},
921 { "Failed", 4},
922 { "Partially Optimal", 5},
923 { "-reserved-", 6},
924 { "-reserved-", 7},
925 { NULL, 0}
926};
927
928static mapping_t ddf_init_state[] = {
929 { "Not Initialised", 0},
930 { "QuickInit in Progress", 1},
931 { "Fully Initialised", 2},
932 { "*UNKNOWN*", 3},
933 { NULL, 0}
934};
935static mapping_t ddf_access[] = {
936 { "Read/Write", 0},
937 { "Reserved", 1},
938 { "Read Only", 2},
939 { "Blocked (no access)", 3},
940 { NULL ,0}
941};
942
943static mapping_t ddf_level[] = {
944 { "RAID0", DDF_RAID0},
945 { "RAID1", DDF_RAID1},
946 { "RAID3", DDF_RAID3},
947 { "RAID4", DDF_RAID4},
948 { "RAID5", DDF_RAID5},
949 { "RAID1E",DDF_RAID1E},
950 { "JBOD", DDF_JBOD},
951 { "CONCAT",DDF_CONCAT},
952 { "RAID5E",DDF_RAID5E},
953 { "RAID5EE",DDF_RAID5EE},
954 { "RAID6", DDF_RAID6},
955 { NULL, 0}
956};
957static mapping_t ddf_sec_level[] = {
958 { "Striped", DDF_2STRIPED},
959 { "Mirrored", DDF_2MIRRORED},
960 { "Concat", DDF_2CONCAT},
961 { "Spanned", DDF_2SPANNED},
962 { NULL, 0}
963};
964#endif
965
966struct num_mapping {
967 int num1, num2;
968};
969static struct num_mapping ddf_level_num[] = {
970 { DDF_RAID0, 0 },
971 { DDF_RAID1, 1 },
972 { DDF_RAID3, LEVEL_UNSUPPORTED },
60f18132
NB
973 { DDF_RAID4, 4 },
974 { DDF_RAID5, 5 },
a322f70c
DW
975 { DDF_RAID1E, LEVEL_UNSUPPORTED },
976 { DDF_JBOD, LEVEL_UNSUPPORTED },
977 { DDF_CONCAT, LEVEL_LINEAR },
978 { DDF_RAID5E, LEVEL_UNSUPPORTED },
979 { DDF_RAID5EE, LEVEL_UNSUPPORTED },
980 { DDF_RAID6, 6},
981 { MAXINT, MAXINT }
982};
983
984static int map_num1(struct num_mapping *map, int num)
985{
986 int i;
987 for (i=0 ; map[i].num1 != MAXINT; i++)
988 if (map[i].num1 == num)
989 break;
990 return map[i].num2;
991}
992
42dc2744
N
993static int all_ff(char *guid)
994{
995 int i;
996 for (i = 0; i < DDF_GUID_LEN; i++)
997 if (guid[i] != (char)0xff)
998 return 0;
999 return 1;
1000}
1001
a322f70c
DW
1002#ifndef MDASSEMBLE
1003static void print_guid(char *guid, int tstamp)
1004{
1005 /* A GUIDs are part (or all) ASCII and part binary.
1006 * They tend to be space padded.
59e36268
NB
1007 * We print the GUID in HEX, then in parentheses add
1008 * any initial ASCII sequence, and a possible
1009 * time stamp from bytes 16-19
a322f70c
DW
1010 */
1011 int l = DDF_GUID_LEN;
1012 int i;
59e36268
NB
1013
1014 for (i=0 ; i<DDF_GUID_LEN ; i++) {
1015 if ((i&3)==0 && i != 0) printf(":");
1016 printf("%02X", guid[i]&255);
1017 }
1018
cfccea8c 1019 printf("\n (");
a322f70c
DW
1020 while (l && guid[l-1] == ' ')
1021 l--;
1022 for (i=0 ; i<l ; i++) {
1023 if (guid[i] >= 0x20 && guid[i] < 0x7f)
1024 fputc(guid[i], stdout);
1025 else
59e36268 1026 break;
a322f70c
DW
1027 }
1028 if (tstamp) {
1029 time_t then = __be32_to_cpu(*(__u32*)(guid+16)) + DECADE;
1030 char tbuf[100];
1031 struct tm *tm;
1032 tm = localtime(&then);
59e36268 1033 strftime(tbuf, 100, " %D %T",tm);
a322f70c
DW
1034 fputs(tbuf, stdout);
1035 }
59e36268 1036 printf(")");
a322f70c
DW
1037}
1038
1039static void examine_vd(int n, struct ddf_super *sb, char *guid)
1040{
8c3b8c2c 1041 int crl = sb->conf_rec_len;
a322f70c
DW
1042 struct vcl *vcl;
1043
1044 for (vcl = sb->conflist ; vcl ; vcl = vcl->next) {
b06e3095 1045 int i;
a322f70c
DW
1046 struct vd_config *vc = &vcl->conf;
1047
1048 if (calc_crc(vc, crl*512) != vc->crc)
1049 continue;
1050 if (memcmp(vc->guid, guid, DDF_GUID_LEN) != 0)
1051 continue;
1052
1053 /* Ok, we know about this VD, let's give more details */
b06e3095 1054 printf(" Raid Devices[%d] : %d (", n,
a322f70c 1055 __be16_to_cpu(vc->prim_elmnt_count));
b06e3095
N
1056 for (i=0; i<__be16_to_cpu(vc->prim_elmnt_count); i++) {
1057 int j;
1058 int cnt = __be16_to_cpu(sb->phys->used_pdes);
1059 for (j=0; j<cnt; j++)
1060 if (vc->phys_refnum[i] == sb->phys->entries[j].refnum)
1061 break;
1062 if (i) printf(" ");
1063 if (j < cnt)
1064 printf("%d", j);
1065 else
1066 printf("--");
1067 }
1068 printf(")\n");
1069 if (vc->chunk_shift != 255)
a322f70c
DW
1070 printf(" Chunk Size[%d] : %d sectors\n", n,
1071 1 << vc->chunk_shift);
1072 printf(" Raid Level[%d] : %s\n", n,
1073 map_num(ddf_level, vc->prl)?:"-unknown-");
1074 if (vc->sec_elmnt_count != 1) {
1075 printf(" Secondary Position[%d] : %d of %d\n", n,
1076 vc->sec_elmnt_seq, vc->sec_elmnt_count);
1077 printf(" Secondary Level[%d] : %s\n", n,
1078 map_num(ddf_sec_level, vc->srl) ?: "-unknown-");
1079 }
1080 printf(" Device Size[%d] : %llu\n", n,
c9b6907b 1081 (unsigned long long)__be64_to_cpu(vc->blocks)/2);
a322f70c 1082 printf(" Array Size[%d] : %llu\n", n,
c9b6907b 1083 (unsigned long long)__be64_to_cpu(vc->array_blocks)/2);
a322f70c
DW
1084 }
1085}
1086
1087static void examine_vds(struct ddf_super *sb)
1088{
1089 int cnt = __be16_to_cpu(sb->virt->populated_vdes);
1090 int i;
1091 printf(" Virtual Disks : %d\n", cnt);
1092
1093 for (i=0; i<cnt; i++) {
1094 struct virtual_entry *ve = &sb->virt->entries[i];
b06e3095 1095 printf("\n");
a322f70c
DW
1096 printf(" VD GUID[%d] : ", i); print_guid(ve->guid, 1);
1097 printf("\n");
1098 printf(" unit[%d] : %d\n", i, __be16_to_cpu(ve->unit));
1099 printf(" state[%d] : %s, %s%s\n", i,
1100 map_num(ddf_state, ve->state & 7),
1101 (ve->state & 8) ? "Morphing, ": "",
1102 (ve->state & 16)? "Not Consistent" : "Consistent");
1103 printf(" init state[%d] : %s\n", i,
1104 map_num(ddf_init_state, ve->init_state&3));
1105 printf(" access[%d] : %s\n", i,
1106 map_num(ddf_access, (ve->init_state>>6) & 3));
1107 printf(" Name[%d] : %.16s\n", i, ve->name);
1108 examine_vd(i, sb, ve->guid);
1109 }
1110 if (cnt) printf("\n");
1111}
1112
1113static void examine_pds(struct ddf_super *sb)
1114{
1115 int cnt = __be16_to_cpu(sb->phys->used_pdes);
1116 int i;
1117 struct dl *dl;
1118 printf(" Physical Disks : %d\n", cnt);
962371a5 1119 printf(" Number RefNo Size Device Type/State\n");
a322f70c
DW
1120
1121 for (i=0 ; i<cnt ; i++) {
1122 struct phys_disk_entry *pd = &sb->phys->entries[i];
1123 int type = __be16_to_cpu(pd->type);
1124 int state = __be16_to_cpu(pd->state);
1125
b06e3095
N
1126 //printf(" PD GUID[%d] : ", i); print_guid(pd->guid, 0);
1127 //printf("\n");
1128 printf(" %3d %08x ", i,
a322f70c 1129 __be32_to_cpu(pd->refnum));
c9b6907b
DL
1130 printf("%8lluK ",
1131 (unsigned long long)__be64_to_cpu(pd->config_size)>>1);
b06e3095
N
1132 for (dl = sb->dlist; dl ; dl = dl->next) {
1133 if (dl->disk.refnum == pd->refnum) {
1134 char *dv = map_dev(dl->major, dl->minor, 0);
1135 if (dv) {
962371a5 1136 printf("%-15s", dv);
b06e3095
N
1137 break;
1138 }
1139 }
1140 }
1141 if (!dl)
962371a5 1142 printf("%15s","");
b06e3095 1143 printf(" %s%s%s%s%s",
a322f70c 1144 (type&2) ? "active":"",
b06e3095 1145 (type&4) ? "Global-Spare":"",
a322f70c
DW
1146 (type&8) ? "spare" : "",
1147 (type&16)? ", foreign" : "",
1148 (type&32)? "pass-through" : "");
b06e3095 1149 printf("/%s%s%s%s%s%s%s",
a322f70c
DW
1150 (state&1)? "Online": "Offline",
1151 (state&2)? ", Failed": "",
1152 (state&4)? ", Rebuilding": "",
1153 (state&8)? ", in-transition": "",
b06e3095
N
1154 (state&16)? ", SMART-errors": "",
1155 (state&32)? ", Unrecovered-Read-Errors": "",
a322f70c 1156 (state&64)? ", Missing" : "");
a322f70c
DW
1157 printf("\n");
1158 }
1159}
1160
1161static void examine_super_ddf(struct supertype *st, char *homehost)
1162{
1163 struct ddf_super *sb = st->sb;
1164
1165 printf(" Magic : %08x\n", __be32_to_cpu(sb->anchor.magic));
1166 printf(" Version : %.8s\n", sb->anchor.revision);
598f0d58
NB
1167 printf("Controller GUID : "); print_guid(sb->controller.guid, 0);
1168 printf("\n");
1169 printf(" Container GUID : "); print_guid(sb->anchor.guid, 1);
a322f70c
DW
1170 printf("\n");
1171 printf(" Seq : %08x\n", __be32_to_cpu(sb->active->seq));
1172 printf(" Redundant hdr : %s\n", sb->secondary.magic == DDF_HEADER_MAGIC
1173 ?"yes" : "no");
1174 examine_vds(sb);
1175 examine_pds(sb);
1176}
1177
ff54de6e
N
1178static void getinfo_super_ddf(struct supertype *st, struct mdinfo *info);
1179
42dc2744 1180static void uuid_from_super_ddf(struct supertype *st, int uuid[4]);
ff54de6e 1181
061f2c6a 1182static void brief_examine_super_ddf(struct supertype *st, int verbose)
4737ae25
N
1183{
1184 /* We just write a generic DDF ARRAY entry
1185 */
1186 struct mdinfo info;
1187 char nbuf[64];
1188 getinfo_super_ddf(st, &info);
1189 fname_from_uuid(st, &info, nbuf, ':');
1190
1191 printf("ARRAY metadata=ddf UUID=%s\n", nbuf + 5);
1192}
1193
1194static void brief_examine_subarrays_ddf(struct supertype *st, int verbose)
a322f70c
DW
1195{
1196 /* We just write a generic DDF ARRAY entry
a322f70c 1197 */
42dc2744 1198 struct ddf_super *ddf = st->sb;
ff54de6e 1199 struct mdinfo info;
42dc2744 1200 int i;
ff54de6e
N
1201 char nbuf[64];
1202 getinfo_super_ddf(st, &info);
1203 fname_from_uuid(st, &info, nbuf, ':');
42dc2744
N
1204
1205 for (i=0; i<__be16_to_cpu(ddf->virt->max_vdes); i++) {
1206 struct virtual_entry *ve = &ddf->virt->entries[i];
1207 struct vcl vcl;
1208 char nbuf1[64];
1209 if (all_ff(ve->guid))
1210 continue;
1211 memcpy(vcl.conf.guid, ve->guid, DDF_GUID_LEN);
1212 ddf->currentconf =&vcl;
1213 uuid_from_super_ddf(st, info.uuid);
1214 fname_from_uuid(st, &info, nbuf1, ':');
1215 printf("ARRAY container=%s member=%d UUID=%s\n",
1216 nbuf+5, i, nbuf1+5);
1217 }
a322f70c
DW
1218}
1219
bceedeec
N
1220static void export_examine_super_ddf(struct supertype *st)
1221{
1222 struct mdinfo info;
1223 char nbuf[64];
1224 getinfo_super_ddf(st, &info);
1225 fname_from_uuid(st, &info, nbuf, ':');
1226 printf("MD_METADATA=ddf\n");
1227 printf("MD_LEVEL=container\n");
1228 printf("MD_UUID=%s\n", nbuf+5);
1229}
1230
1231
a322f70c
DW
1232static void detail_super_ddf(struct supertype *st, char *homehost)
1233{
1234 /* FIXME later
1235 * Could print DDF GUID
1236 * Need to find which array
1237 * If whole, briefly list all arrays
1238 * If one, give name
1239 */
1240}
1241
1242static void brief_detail_super_ddf(struct supertype *st)
1243{
1244 /* FIXME I really need to know which array we are detailing.
1245 * Can that be stored in ddf_super??
1246 */
1247// struct ddf_super *ddf = st->sb;
ff54de6e
N
1248 struct mdinfo info;
1249 char nbuf[64];
1250 getinfo_super_ddf(st, &info);
1251 fname_from_uuid(st, &info, nbuf,':');
1252 printf(" UUID=%s", nbuf + 5);
a322f70c 1253}
a322f70c
DW
1254#endif
1255
1256static int match_home_ddf(struct supertype *st, char *homehost)
1257{
1258 /* It matches 'this' host if the controller is a
1259 * Linux-MD controller with vendor_data matching
1260 * the hostname
1261 */
1262 struct ddf_super *ddf = st->sb;
1263 int len = strlen(homehost);
1264
1265 return (memcmp(ddf->controller.guid, T10, 8) == 0 &&
1266 len < sizeof(ddf->controller.vendor_data) &&
1267 memcmp(ddf->controller.vendor_data, homehost,len) == 0 &&
1268 ddf->controller.vendor_data[len] == 0);
1269}
1270
0e600426 1271#ifndef MDASSEMBLE
7a7cc504 1272static struct vd_config *find_vdcr(struct ddf_super *ddf, int inst)
a322f70c 1273{
7a7cc504 1274 struct vcl *v;
59e36268 1275
7a7cc504 1276 for (v = ddf->conflist; v; v = v->next)
59e36268 1277 if (inst == v->vcnum)
7a7cc504
NB
1278 return &v->conf;
1279 return NULL;
1280}
0e600426 1281#endif
7a7cc504
NB
1282
1283static int find_phys(struct ddf_super *ddf, __u32 phys_refnum)
1284{
1285 /* Find the entry in phys_disk which has the given refnum
1286 * and return it's index
1287 */
1288 int i;
1289 for (i=0; i < __be16_to_cpu(ddf->phys->max_pdes); i++)
1290 if (ddf->phys->entries[i].refnum == phys_refnum)
1291 return i;
1292 return -1;
a322f70c
DW
1293}
1294
1295static void uuid_from_super_ddf(struct supertype *st, int uuid[4])
1296{
1297 /* The uuid returned here is used for:
1298 * uuid to put into bitmap file (Create, Grow)
1299 * uuid for backup header when saving critical section (Grow)
1300 * comparing uuids when re-adding a device into an array
51006d85
N
1301 * In these cases the uuid required is that of the data-array,
1302 * not the device-set.
1303 * uuid to recognise same set when adding a missing device back
1304 * to an array. This is a uuid for the device-set.
1305 *
a322f70c
DW
1306 * For each of these we can make do with a truncated
1307 * or hashed uuid rather than the original, as long as
1308 * everyone agrees.
a322f70c
DW
1309 * In the case of SVD we assume the BVD is of interest,
1310 * though that might be the case if a bitmap were made for
1311 * a mirrored SVD - worry about that later.
1312 * So we need to find the VD configuration record for the
1313 * relevant BVD and extract the GUID and Secondary_Element_Seq.
1314 * The first 16 bytes of the sha1 of these is used.
1315 */
1316 struct ddf_super *ddf = st->sb;
d2ca6449 1317 struct vcl *vcl = ddf->currentconf;
c5afc314
N
1318 char *guid;
1319 char buf[20];
1320 struct sha1_ctx ctx;
a322f70c 1321
c5afc314
N
1322 if (vcl)
1323 guid = vcl->conf.guid;
1324 else
1325 guid = ddf->anchor.guid;
1326
1327 sha1_init_ctx(&ctx);
1328 sha1_process_bytes(guid, DDF_GUID_LEN, &ctx);
c5afc314
N
1329 sha1_finish_ctx(&ctx, buf);
1330 memcpy(uuid, buf, 4*4);
a322f70c
DW
1331}
1332
78e44928
NB
1333static void getinfo_super_ddf_bvd(struct supertype *st, struct mdinfo *info);
1334
a322f70c
DW
1335static void getinfo_super_ddf(struct supertype *st, struct mdinfo *info)
1336{
1337 struct ddf_super *ddf = st->sb;
1338
78e44928
NB
1339 if (ddf->currentconf) {
1340 getinfo_super_ddf_bvd(st, info);
1341 return;
1342 }
1343
a322f70c
DW
1344 info->array.raid_disks = __be16_to_cpu(ddf->phys->used_pdes);
1345 info->array.level = LEVEL_CONTAINER;
1346 info->array.layout = 0;
1347 info->array.md_minor = -1;
1348 info->array.ctime = DECADE + __be32_to_cpu(*(__u32*)
1349 (ddf->anchor.guid+16));
1350 info->array.utime = 0;
1351 info->array.chunk_size = 0;
1352
a322f70c
DW
1353
1354 info->disk.major = 0;
1355 info->disk.minor = 0;
cba0191b
NB
1356 if (ddf->dlist) {
1357 info->disk.number = __be32_to_cpu(ddf->dlist->disk.refnum);
59e36268 1358 info->disk.raid_disk = find_phys(ddf, ddf->dlist->disk.refnum);
d2ca6449
NB
1359
1360 info->data_offset = __be64_to_cpu(ddf->phys->
1361 entries[info->disk.raid_disk].
1362 config_size);
1363 info->component_size = ddf->dlist->size - info->data_offset;
cba0191b
NB
1364 } else {
1365 info->disk.number = -1;
661dce36 1366 info->disk.raid_disk = -1;
cba0191b
NB
1367// info->disk.raid_disk = find refnum in the table and use index;
1368 }
f22385f9 1369 info->disk.state = (1 << MD_DISK_SYNC) | (1 << MD_DISK_ACTIVE);
a19c88b8 1370
d2ca6449 1371
a19c88b8 1372 info->reshape_active = 0;
c5afc314 1373 info->name[0] = 0;
a322f70c 1374
f35f2525
N
1375 info->array.major_version = -1;
1376 info->array.minor_version = -2;
159c3a1a 1377 strcpy(info->text_version, "ddf");
a67dd8cc 1378 info->safe_mode_delay = 0;
159c3a1a 1379
c5afc314 1380 uuid_from_super_ddf(st, info->uuid);
a322f70c 1381
a322f70c
DW
1382}
1383
598f0d58
NB
1384static int rlq_to_layout(int rlq, int prl, int raiddisks);
1385
a322f70c
DW
1386static void getinfo_super_ddf_bvd(struct supertype *st, struct mdinfo *info)
1387{
1388 struct ddf_super *ddf = st->sb;
d2ca6449
NB
1389 struct vcl *vc = ddf->currentconf;
1390 int cd = ddf->currentdev;
db42fa9b 1391 int j;
8592f29d 1392 struct dl *dl;
a322f70c
DW
1393
1394 /* FIXME this returns BVD info - what if we want SVD ?? */
1395
d2ca6449
NB
1396 info->array.raid_disks = __be16_to_cpu(vc->conf.prim_elmnt_count);
1397 info->array.level = map_num1(ddf_level_num, vc->conf.prl);
1398 info->array.layout = rlq_to_layout(vc->conf.rlq, vc->conf.prl,
598f0d58 1399 info->array.raid_disks);
a322f70c 1400 info->array.md_minor = -1;
d2ca6449
NB
1401 info->array.ctime = DECADE +
1402 __be32_to_cpu(*(__u32*)(vc->conf.guid+16));
1403 info->array.utime = DECADE + __be32_to_cpu(vc->conf.timestamp);
1404 info->array.chunk_size = 512 << vc->conf.chunk_shift;
da9b4a62 1405 info->custom_array_size = 0;
d2ca6449
NB
1406
1407 if (cd >= 0 && cd < ddf->mppe) {
1408 info->data_offset = __be64_to_cpu(vc->lba_offset[cd]);
1409 if (vc->block_sizes)
1410 info->component_size = vc->block_sizes[cd];
1411 else
1412 info->component_size = __be64_to_cpu(vc->conf.blocks);
1413 }
a322f70c 1414
8592f29d
N
1415 for (dl = ddf->dlist; dl ; dl = dl->next)
1416 if (dl->raiddisk == info->disk.raid_disk)
1417 break;
a322f70c
DW
1418 info->disk.major = 0;
1419 info->disk.minor = 0;
8592f29d
N
1420 if (dl) {
1421 info->disk.major = dl->major;
1422 info->disk.minor = dl->minor;
1423 }
a322f70c
DW
1424// info->disk.number = __be32_to_cpu(ddf->disk.refnum);
1425// info->disk.raid_disk = find refnum in the table and use index;
1426// info->disk.state = ???;
1427
103f2410
NB
1428 info->container_member = ddf->currentconf->vcnum;
1429
80d26cb2
NB
1430 info->resync_start = 0;
1431 if (!(ddf->virt->entries[info->container_member].state
1432 & DDF_state_inconsistent) &&
1433 (ddf->virt->entries[info->container_member].init_state
1434 & DDF_initstate_mask)
1435 == DDF_init_full)
1436 info->resync_start = ~0ULL;
1437
a322f70c
DW
1438 uuid_from_super_ddf(st, info->uuid);
1439
3576e302 1440 info->container_member = atoi(st->subarray);
f35f2525
N
1441 info->array.major_version = -1;
1442 info->array.minor_version = -2;
3576e302 1443 sprintf(info->text_version, "/%s/%s",
159c3a1a 1444 devnum2devname(st->container_dev),
3576e302 1445 st->subarray);
a67dd8cc 1446 info->safe_mode_delay = 200;
159c3a1a 1447
db42fa9b
N
1448 memcpy(info->name, ddf->virt->entries[info->container_member].name, 16);
1449 info->name[16]=0;
1450 for(j=0; j<16; j++)
1451 if (info->name[j] == ' ')
1452 info->name[j] = 0;
a322f70c
DW
1453}
1454
598f0d58 1455
a322f70c
DW
1456static int update_super_ddf(struct supertype *st, struct mdinfo *info,
1457 char *update,
1458 char *devname, int verbose,
1459 int uuid_set, char *homehost)
1460{
1461 /* For 'assemble' and 'force' we need to return non-zero if any
1462 * change was made. For others, the return value is ignored.
1463 * Update options are:
1464 * force-one : This device looks a bit old but needs to be included,
1465 * update age info appropriately.
1466 * assemble: clear any 'faulty' flag to allow this device to
1467 * be assembled.
1468 * force-array: Array is degraded but being forced, mark it clean
1469 * if that will be needed to assemble it.
1470 *
1471 * newdev: not used ????
1472 * grow: Array has gained a new device - this is currently for
1473 * linear only
1474 * resync: mark as dirty so a resync will happen.
59e36268 1475 * uuid: Change the uuid of the array to match what is given
a322f70c
DW
1476 * homehost: update the recorded homehost
1477 * name: update the name - preserving the homehost
1478 * _reshape_progress: record new reshape_progress position.
1479 *
1480 * Following are not relevant for this version:
1481 * sparc2.2 : update from old dodgey metadata
1482 * super-minor: change the preferred_minor number
1483 * summaries: update redundant counters.
1484 */
1485 int rv = 0;
1486// struct ddf_super *ddf = st->sb;
7a7cc504 1487// struct vd_config *vd = find_vdcr(ddf, info->container_member);
a322f70c
DW
1488// struct virtual_entry *ve = find_ve(ddf);
1489
a322f70c
DW
1490 /* we don't need to handle "force-*" or "assemble" as
1491 * there is no need to 'trick' the kernel. We the metadata is
1492 * first updated to activate the array, all the implied modifications
1493 * will just happen.
1494 */
1495
1496 if (strcmp(update, "grow") == 0) {
1497 /* FIXME */
1498 }
1499 if (strcmp(update, "resync") == 0) {
1500// info->resync_checkpoint = 0;
1501 }
1502 /* We ignore UUID updates as they make even less sense
1503 * with DDF
1504 */
1505 if (strcmp(update, "homehost") == 0) {
1506 /* homehost is stored in controller->vendor_data,
1507 * or it is when we are the vendor
1508 */
1509// if (info->vendor_is_local)
1510// strcpy(ddf->controller.vendor_data, homehost);
1511 }
1512 if (strcmp(update, "name") == 0) {
1513 /* name is stored in virtual_entry->name */
1514// memset(ve->name, ' ', 16);
1515// strncpy(ve->name, info->name, 16);
1516 }
1517 if (strcmp(update, "_reshape_progress") == 0) {
1518 /* We don't support reshape yet */
1519 }
1520
1521// update_all_csum(ddf);
1522
1523 return rv;
1524}
1525
5f8097be
NB
1526static void make_header_guid(char *guid)
1527{
1528 __u32 stamp;
5f8097be
NB
1529 /* Create a DDF Header of Virtual Disk GUID */
1530
1531 /* 24 bytes of fiction required.
1532 * first 8 are a 'vendor-id' - "Linux-MD"
1533 * next 8 are controller type.. how about 0X DEAD BEEF 0000 0000
1534 * Remaining 8 random number plus timestamp
1535 */
1536 memcpy(guid, T10, sizeof(T10));
1537 stamp = __cpu_to_be32(0xdeadbeef);
1538 memcpy(guid+8, &stamp, 4);
1539 stamp = __cpu_to_be32(0);
1540 memcpy(guid+12, &stamp, 4);
1541 stamp = __cpu_to_be32(time(0) - DECADE);
1542 memcpy(guid+16, &stamp, 4);
bfb7ea78 1543 stamp = random32();
5f8097be 1544 memcpy(guid+20, &stamp, 4);
5f8097be 1545}
59e36268 1546
78e44928
NB
1547static int init_super_ddf_bvd(struct supertype *st,
1548 mdu_array_info_t *info,
1549 unsigned long long size,
1550 char *name, char *homehost,
1551 int *uuid);
1552
a322f70c
DW
1553static int init_super_ddf(struct supertype *st,
1554 mdu_array_info_t *info,
1555 unsigned long long size, char *name, char *homehost,
1556 int *uuid)
1557{
1558 /* This is primarily called by Create when creating a new array.
1559 * We will then get add_to_super called for each component, and then
1560 * write_init_super called to write it out to each device.
1561 * For DDF, Create can create on fresh devices or on a pre-existing
1562 * array.
1563 * To create on a pre-existing array a different method will be called.
1564 * This one is just for fresh drives.
1565 *
1566 * We need to create the entire 'ddf' structure which includes:
1567 * DDF headers - these are easy.
1568 * Controller data - a Sector describing this controller .. not that
1569 * this is a controller exactly.
1570 * Physical Disk Record - one entry per device, so
1571 * leave plenty of space.
1572 * Virtual Disk Records - again, just leave plenty of space.
1573 * This just lists VDs, doesn't give details
1574 * Config records - describes the VDs that use this disk
1575 * DiskData - describes 'this' device.
1576 * BadBlockManagement - empty
1577 * Diag Space - empty
1578 * Vendor Logs - Could we put bitmaps here?
1579 *
1580 */
1581 struct ddf_super *ddf;
1582 char hostname[17];
1583 int hostlen;
a322f70c
DW
1584 int max_phys_disks, max_virt_disks;
1585 unsigned long long sector;
1586 int clen;
1587 int i;
1588 int pdsize, vdsize;
1589 struct phys_disk *pd;
1590 struct virtual_disk *vd;
1591
78e44928 1592 if (st->sb)
955e9ea1 1593 return init_super_ddf_bvd(st, info, size, name, homehost, uuid);
ba7eb04f 1594
3d2c4fc7
DW
1595 if (posix_memalign((void**)&ddf, 512, sizeof(*ddf)) != 0) {
1596 fprintf(stderr, Name ": %s could not allocate superblock\n", __func__);
1597 return 0;
1598 }
6264b437 1599 memset(ddf, 0, sizeof(*ddf));
a322f70c
DW
1600 ddf->dlist = NULL; /* no physical disks yet */
1601 ddf->conflist = NULL; /* No virtual disks yet */
955e9ea1
DW
1602 st->sb = ddf;
1603
1604 if (info == NULL) {
1605 /* zeroing superblock */
1606 return 0;
1607 }
a322f70c
DW
1608
1609 /* At least 32MB *must* be reserved for the ddf. So let's just
1610 * start 32MB from the end, and put the primary header there.
1611 * Don't do secondary for now.
1612 * We don't know exactly where that will be yet as it could be
1613 * different on each device. To just set up the lengths.
1614 *
1615 */
1616
1617 ddf->anchor.magic = DDF_HEADER_MAGIC;
5f8097be 1618 make_header_guid(ddf->anchor.guid);
a322f70c 1619
59e36268 1620 memcpy(ddf->anchor.revision, DDF_REVISION_2, 8);
a322f70c
DW
1621 ddf->anchor.seq = __cpu_to_be32(1);
1622 ddf->anchor.timestamp = __cpu_to_be32(time(0) - DECADE);
1623 ddf->anchor.openflag = 0xFF;
1624 ddf->anchor.foreignflag = 0;
1625 ddf->anchor.enforcegroups = 0; /* Is this best?? */
1626 ddf->anchor.pad0 = 0xff;
1627 memset(ddf->anchor.pad1, 0xff, 12);
1628 memset(ddf->anchor.header_ext, 0xff, 32);
1629 ddf->anchor.primary_lba = ~(__u64)0;
1630 ddf->anchor.secondary_lba = ~(__u64)0;
1631 ddf->anchor.type = DDF_HEADER_ANCHOR;
1632 memset(ddf->anchor.pad2, 0xff, 3);
1633 ddf->anchor.workspace_len = __cpu_to_be32(32768); /* Must be reserved */
1634 ddf->anchor.workspace_lba = ~(__u64)0; /* Put this at bottom
1635 of 32M reserved.. */
1636 max_phys_disks = 1023; /* Should be enough */
1637 ddf->anchor.max_pd_entries = __cpu_to_be16(max_phys_disks);
1638 max_virt_disks = 255;
1639 ddf->anchor.max_vd_entries = __cpu_to_be16(max_virt_disks); /* ?? */
1640 ddf->anchor.max_partitions = __cpu_to_be16(64); /* ?? */
1641 ddf->max_part = 64;
8c3b8c2c 1642 ddf->mppe = 256;
59e36268
NB
1643 ddf->conf_rec_len = 1 + ROUND_UP(ddf->mppe * (4+8), 512)/512;
1644 ddf->anchor.config_record_len = __cpu_to_be16(ddf->conf_rec_len);
1645 ddf->anchor.max_primary_element_entries = __cpu_to_be16(ddf->mppe);
a322f70c 1646 memset(ddf->anchor.pad3, 0xff, 54);
a322f70c
DW
1647 /* controller sections is one sector long immediately
1648 * after the ddf header */
1649 sector = 1;
1650 ddf->anchor.controller_section_offset = __cpu_to_be32(sector);
1651 ddf->anchor.controller_section_length = __cpu_to_be32(1);
1652 sector += 1;
1653
1654 /* phys is 8 sectors after that */
1655 pdsize = ROUND_UP(sizeof(struct phys_disk) +
1656 sizeof(struct phys_disk_entry)*max_phys_disks,
1657 512);
1658 switch(pdsize/512) {
1659 case 2: case 8: case 32: case 128: case 512: break;
1660 default: abort();
1661 }
1662 ddf->anchor.phys_section_offset = __cpu_to_be32(sector);
1663 ddf->anchor.phys_section_length =
1664 __cpu_to_be32(pdsize/512); /* max_primary_element_entries/8 */
1665 sector += pdsize/512;
1666
1667 /* virt is another 32 sectors */
1668 vdsize = ROUND_UP(sizeof(struct virtual_disk) +
1669 sizeof(struct virtual_entry) * max_virt_disks,
1670 512);
1671 switch(vdsize/512) {
1672 case 2: case 8: case 32: case 128: case 512: break;
1673 default: abort();
1674 }
1675 ddf->anchor.virt_section_offset = __cpu_to_be32(sector);
1676 ddf->anchor.virt_section_length =
1677 __cpu_to_be32(vdsize/512); /* max_vd_entries/8 */
1678 sector += vdsize/512;
1679
59e36268 1680 clen = ddf->conf_rec_len * (ddf->max_part+1);
a322f70c
DW
1681 ddf->anchor.config_section_offset = __cpu_to_be32(sector);
1682 ddf->anchor.config_section_length = __cpu_to_be32(clen);
1683 sector += clen;
1684
1685 ddf->anchor.data_section_offset = __cpu_to_be32(sector);
1686 ddf->anchor.data_section_length = __cpu_to_be32(1);
1687 sector += 1;
1688
1689 ddf->anchor.bbm_section_length = __cpu_to_be32(0);
1690 ddf->anchor.bbm_section_offset = __cpu_to_be32(0xFFFFFFFF);
1691 ddf->anchor.diag_space_length = __cpu_to_be32(0);
1692 ddf->anchor.diag_space_offset = __cpu_to_be32(0xFFFFFFFF);
1693 ddf->anchor.vendor_length = __cpu_to_be32(0);
1694 ddf->anchor.vendor_offset = __cpu_to_be32(0xFFFFFFFF);
1695
1696 memset(ddf->anchor.pad4, 0xff, 256);
1697
1698 memcpy(&ddf->primary, &ddf->anchor, 512);
1699 memcpy(&ddf->secondary, &ddf->anchor, 512);
1700
1701 ddf->primary.openflag = 1; /* I guess.. */
1702 ddf->primary.type = DDF_HEADER_PRIMARY;
1703
1704 ddf->secondary.openflag = 1; /* I guess.. */
1705 ddf->secondary.type = DDF_HEADER_SECONDARY;
1706
1707 ddf->active = &ddf->primary;
1708
1709 ddf->controller.magic = DDF_CONTROLLER_MAGIC;
1710
1711 /* 24 more bytes of fiction required.
1712 * first 8 are a 'vendor-id' - "Linux-MD"
1713 * Remaining 16 are serial number.... maybe a hostname would do?
1714 */
1715 memcpy(ddf->controller.guid, T10, sizeof(T10));
1ba6bff9
DW
1716 gethostname(hostname, sizeof(hostname));
1717 hostname[sizeof(hostname) - 1] = 0;
a322f70c
DW
1718 hostlen = strlen(hostname);
1719 memcpy(ddf->controller.guid + 24 - hostlen, hostname, hostlen);
1720 for (i = strlen(T10) ; i+hostlen < 24; i++)
1721 ddf->controller.guid[i] = ' ';
1722
1723 ddf->controller.type.vendor_id = __cpu_to_be16(0xDEAD);
1724 ddf->controller.type.device_id = __cpu_to_be16(0xBEEF);
1725 ddf->controller.type.sub_vendor_id = 0;
1726 ddf->controller.type.sub_device_id = 0;
1727 memcpy(ddf->controller.product_id, "What Is My PID??", 16);
1728 memset(ddf->controller.pad, 0xff, 8);
1729 memset(ddf->controller.vendor_data, 0xff, 448);
a9e1c11d
N
1730 if (homehost && strlen(homehost) < 440)
1731 strcpy((char*)ddf->controller.vendor_data, homehost);
a322f70c 1732
3d2c4fc7
DW
1733 if (posix_memalign((void**)&pd, 512, pdsize) != 0) {
1734 fprintf(stderr, Name ": %s could not allocate pd\n", __func__);
1735 return 0;
1736 }
6416d527 1737 ddf->phys = pd;
a322f70c
DW
1738 ddf->pdsize = pdsize;
1739
1740 memset(pd, 0xff, pdsize);
1741 memset(pd, 0, sizeof(*pd));
1742 pd->magic = DDF_PHYS_DATA_MAGIC;
1743 pd->used_pdes = __cpu_to_be16(0);
1744 pd->max_pdes = __cpu_to_be16(max_phys_disks);
1745 memset(pd->pad, 0xff, 52);
1746
3d2c4fc7
DW
1747 if (posix_memalign((void**)&vd, 512, vdsize) != 0) {
1748 fprintf(stderr, Name ": %s could not allocate vd\n", __func__);
1749 return 0;
1750 }
6416d527 1751 ddf->virt = vd;
a322f70c
DW
1752 ddf->vdsize = vdsize;
1753 memset(vd, 0, vdsize);
1754 vd->magic = DDF_VIRT_RECORDS_MAGIC;
1755 vd->populated_vdes = __cpu_to_be16(0);
1756 vd->max_vdes = __cpu_to_be16(max_virt_disks);
1757 memset(vd->pad, 0xff, 52);
1758
5f8097be
NB
1759 for (i=0; i<max_virt_disks; i++)
1760 memset(&vd->entries[i], 0xff, sizeof(struct virtual_entry));
1761
a322f70c 1762 st->sb = ddf;
18a2f463 1763 ddf->updates_pending = 1;
a322f70c
DW
1764 return 1;
1765}
1766
5f8097be
NB
1767static int chunk_to_shift(int chunksize)
1768{
1769 return ffs(chunksize/512)-1;
1770}
1771
1772static int level_to_prl(int level)
1773{
1774 switch (level) {
1775 case LEVEL_LINEAR: return DDF_CONCAT;
1776 case 0: return DDF_RAID0;
1777 case 1: return DDF_RAID1;
1778 case 4: return DDF_RAID4;
1779 case 5: return DDF_RAID5;
1780 case 6: return DDF_RAID6;
1781 default: return -1;
1782 }
1783}
1784static int layout_to_rlq(int level, int layout, int raiddisks)
1785{
1786 switch(level) {
1787 case 0:
1788 return DDF_RAID0_SIMPLE;
1789 case 1:
1790 switch(raiddisks) {
1791 case 2: return DDF_RAID1_SIMPLE;
1792 case 3: return DDF_RAID1_MULTI;
1793 default: return -1;
1794 }
1795 case 4:
1796 switch(layout) {
1797 case 0: return DDF_RAID4_N;
1798 }
1799 break;
1800 case 5:
5f8097be
NB
1801 switch(layout) {
1802 case ALGORITHM_LEFT_ASYMMETRIC:
1803 return DDF_RAID5_N_RESTART;
1804 case ALGORITHM_RIGHT_ASYMMETRIC:
b640a252 1805 return DDF_RAID5_0_RESTART;
5f8097be
NB
1806 case ALGORITHM_LEFT_SYMMETRIC:
1807 return DDF_RAID5_N_CONTINUE;
1808 case ALGORITHM_RIGHT_SYMMETRIC:
1809 return -1; /* not mentioned in standard */
1810 }
b640a252
N
1811 case 6:
1812 switch(layout) {
1813 case ALGORITHM_ROTATING_N_RESTART:
1814 return DDF_RAID5_N_RESTART;
1815 case ALGORITHM_ROTATING_ZERO_RESTART:
1816 return DDF_RAID6_0_RESTART;
1817 case ALGORITHM_ROTATING_N_CONTINUE:
1818 return DDF_RAID5_N_CONTINUE;
1819 }
5f8097be
NB
1820 }
1821 return -1;
1822}
1823
598f0d58
NB
1824static int rlq_to_layout(int rlq, int prl, int raiddisks)
1825{
1826 switch(prl) {
1827 case DDF_RAID0:
1828 return 0; /* hopefully rlq == DDF_RAID0_SIMPLE */
1829 case DDF_RAID1:
1830 return 0; /* hopefully rlq == SIMPLE or MULTI depending
1831 on raiddisks*/
1832 case DDF_RAID4:
1833 switch(rlq) {
1834 case DDF_RAID4_N:
1835 return 0;
1836 default:
1837 /* not supported */
1838 return -1; /* FIXME this isn't checked */
1839 }
1840 case DDF_RAID5:
598f0d58
NB
1841 switch(rlq) {
1842 case DDF_RAID5_N_RESTART:
1843 return ALGORITHM_LEFT_ASYMMETRIC;
1844 case DDF_RAID5_0_RESTART:
1845 return ALGORITHM_RIGHT_ASYMMETRIC;
1846 case DDF_RAID5_N_CONTINUE:
1847 return ALGORITHM_LEFT_SYMMETRIC;
1848 default:
1849 return -1;
1850 }
59e36268
NB
1851 case DDF_RAID6:
1852 switch(rlq) {
1853 case DDF_RAID5_N_RESTART:
b640a252 1854 return ALGORITHM_ROTATING_N_RESTART;
59e36268 1855 case DDF_RAID6_0_RESTART:
b640a252 1856 return ALGORITHM_ROTATING_ZERO_RESTART;
59e36268 1857 case DDF_RAID5_N_CONTINUE:
b640a252 1858 return ALGORITHM_ROTATING_N_CONTINUE;
59e36268
NB
1859 default:
1860 return -1;
1861 }
598f0d58
NB
1862 }
1863 return -1;
1864}
1865
0e600426 1866#ifndef MDASSEMBLE
59e36268
NB
1867struct extent {
1868 unsigned long long start, size;
1869};
78e44928 1870static int cmp_extent(const void *av, const void *bv)
59e36268
NB
1871{
1872 const struct extent *a = av;
1873 const struct extent *b = bv;
1874 if (a->start < b->start)
1875 return -1;
1876 if (a->start > b->start)
1877 return 1;
1878 return 0;
1879}
1880
78e44928 1881static struct extent *get_extents(struct ddf_super *ddf, struct dl *dl)
59e36268
NB
1882{
1883 /* find a list of used extents on the give physical device
1884 * (dnum) of the given ddf.
1885 * Return a malloced array of 'struct extent'
1886
1887FIXME ignore DDF_Legacy devices?
1888
1889 */
1890 struct extent *rv;
1891 int n = 0;
1892 int i, j;
1893
1894 rv = malloc(sizeof(struct extent) * (ddf->max_part + 2));
1895 if (!rv)
1896 return NULL;
1897
1898 for (i = 0; i < ddf->max_part; i++) {
1899 struct vcl *v = dl->vlist[i];
1900 if (v == NULL)
1901 continue;
1902 for (j=0; j < v->conf.prim_elmnt_count; j++)
1903 if (v->conf.phys_refnum[j] == dl->disk.refnum) {
1904 /* This device plays role 'j' in 'v'. */
1905 rv[n].start = __be64_to_cpu(v->lba_offset[j]);
1906 rv[n].size = __be64_to_cpu(v->conf.blocks);
1907 n++;
1908 break;
1909 }
1910 }
1911 qsort(rv, n, sizeof(*rv), cmp_extent);
1912
1913 rv[n].start = __be64_to_cpu(ddf->phys->entries[dl->pdnum].config_size);
1914 rv[n].size = 0;
1915 return rv;
1916}
0e600426 1917#endif
59e36268 1918
5f8097be
NB
1919static int init_super_ddf_bvd(struct supertype *st,
1920 mdu_array_info_t *info,
1921 unsigned long long size,
1922 char *name, char *homehost,
1923 int *uuid)
1924{
1925 /* We are creating a BVD inside a pre-existing container.
1926 * so st->sb is already set.
1927 * We need to create a new vd_config and a new virtual_entry
1928 */
1929 struct ddf_super *ddf = st->sb;
1930 int venum;
1931 struct virtual_entry *ve;
1932 struct vcl *vcl;
1933 struct vd_config *vc;
5f8097be
NB
1934
1935 if (__be16_to_cpu(ddf->virt->populated_vdes)
1936 >= __be16_to_cpu(ddf->virt->max_vdes)) {
1937 fprintf(stderr, Name": This ddf already has the "
1938 "maximum of %d virtual devices\n",
1939 __be16_to_cpu(ddf->virt->max_vdes));
1940 return 0;
1941 }
1942
1943 for (venum = 0; venum < __be16_to_cpu(ddf->virt->max_vdes); venum++)
1944 if (all_ff(ddf->virt->entries[venum].guid))
1945 break;
1946 if (venum == __be16_to_cpu(ddf->virt->max_vdes)) {
1947 fprintf(stderr, Name ": Cannot find spare slot for "
1948 "virtual disk - DDF is corrupt\n");
1949 return 0;
1950 }
1951 ve = &ddf->virt->entries[venum];
1952
1953 /* A Virtual Disk GUID contains the T10 Vendor ID, controller type,
1954 * timestamp, random number
1955 */
1956 make_header_guid(ve->guid);
1957 ve->unit = __cpu_to_be16(info->md_minor);
1958 ve->pad0 = 0xFFFF;
1959 ve->guid_crc = crc32(0, (unsigned char*)ddf->anchor.guid, DDF_GUID_LEN);
1960 ve->type = 0;
7a7cc504
NB
1961 ve->state = DDF_state_degraded; /* Will be modified as devices are added */
1962 if (info->state & 1) /* clean */
1963 ve->init_state = DDF_init_full;
1964 else
1965 ve->init_state = DDF_init_not;
1966
5f8097be
NB
1967 memset(ve->pad1, 0xff, 14);
1968 memset(ve->name, ' ', 16);
1969 if (name)
1970 strncpy(ve->name, name, 16);
1971 ddf->virt->populated_vdes =
1972 __cpu_to_be16(__be16_to_cpu(ddf->virt->populated_vdes)+1);
1973
1974 /* Now create a new vd_config */
3d2c4fc7
DW
1975 if (posix_memalign((void**)&vcl, 512,
1976 (offsetof(struct vcl, conf) + ddf->conf_rec_len * 512)) != 0) {
1977 fprintf(stderr, Name ": %s could not allocate vd_config\n", __func__);
1978 return 0;
1979 }
8c3b8c2c 1980 vcl->lba_offset = (__u64*) &vcl->conf.phys_refnum[ddf->mppe];
59e36268 1981 vcl->vcnum = venum;
f7e7067b 1982 sprintf(st->subarray, "%d", venum);
59e36268 1983 vcl->block_sizes = NULL; /* FIXME not for CONCAT */
5f8097be
NB
1984
1985 vc = &vcl->conf;
1986
1987 vc->magic = DDF_VD_CONF_MAGIC;
1988 memcpy(vc->guid, ve->guid, DDF_GUID_LEN);
1989 vc->timestamp = __cpu_to_be32(time(0)-DECADE);
1990 vc->seqnum = __cpu_to_be32(1);
1991 memset(vc->pad0, 0xff, 24);
1992 vc->prim_elmnt_count = __cpu_to_be16(info->raid_disks);
1993 vc->chunk_shift = chunk_to_shift(info->chunk_size);
1994 vc->prl = level_to_prl(info->level);
1995 vc->rlq = layout_to_rlq(info->level, info->layout, info->raid_disks);
1996 vc->sec_elmnt_count = 1;
1997 vc->sec_elmnt_seq = 0;
1998 vc->srl = 0;
1999 vc->blocks = __cpu_to_be64(info->size * 2);
2000 vc->array_blocks = __cpu_to_be64(
2001 calc_array_size(info->level, info->raid_disks, info->layout,
2002 info->chunk_size, info->size*2));
2003 memset(vc->pad1, 0xff, 8);
2004 vc->spare_refs[0] = 0xffffffff;
2005 vc->spare_refs[1] = 0xffffffff;
2006 vc->spare_refs[2] = 0xffffffff;
2007 vc->spare_refs[3] = 0xffffffff;
2008 vc->spare_refs[4] = 0xffffffff;
2009 vc->spare_refs[5] = 0xffffffff;
2010 vc->spare_refs[6] = 0xffffffff;
2011 vc->spare_refs[7] = 0xffffffff;
2012 memset(vc->cache_pol, 0, 8);
2013 vc->bg_rate = 0x80;
2014 memset(vc->pad2, 0xff, 3);
2015 memset(vc->pad3, 0xff, 52);
2016 memset(vc->pad4, 0xff, 192);
2017 memset(vc->v0, 0xff, 32);
2018 memset(vc->v1, 0xff, 32);
2019 memset(vc->v2, 0xff, 16);
2020 memset(vc->v3, 0xff, 16);
2021 memset(vc->vendor, 0xff, 32);
598f0d58 2022
8c3b8c2c 2023 memset(vc->phys_refnum, 0xff, 4*ddf->mppe);
e5a2a3cf 2024 memset(vc->phys_refnum+ddf->mppe, 0x00, 8*ddf->mppe);
5f8097be
NB
2025
2026 vcl->next = ddf->conflist;
2027 ddf->conflist = vcl;
d2ca6449 2028 ddf->currentconf = vcl;
18a2f463 2029 ddf->updates_pending = 1;
5f8097be
NB
2030 return 1;
2031}
2032
0e600426 2033#ifndef MDASSEMBLE
5f8097be
NB
2034static void add_to_super_ddf_bvd(struct supertype *st,
2035 mdu_disk_info_t *dk, int fd, char *devname)
2036{
2037 /* fd and devname identify a device with-in the ddf container (st).
2038 * dk identifies a location in the new BVD.
2039 * We need to find suitable free space in that device and update
2040 * the phys_refnum and lba_offset for the newly created vd_config.
2041 * We might also want to update the type in the phys_disk
5575e7d9 2042 * section.
8592f29d
N
2043 *
2044 * Alternately: fd == -1 and we have already chosen which device to
2045 * use and recorded in dlist->raid_disk;
5f8097be
NB
2046 */
2047 struct dl *dl;
2048 struct ddf_super *ddf = st->sb;
2049 struct vd_config *vc;
2050 __u64 *lba_offset;
7a7cc504 2051 int working;
5575e7d9 2052 int i;
59e36268
NB
2053 unsigned long long blocks, pos, esize;
2054 struct extent *ex;
5f8097be 2055
8592f29d
N
2056 if (fd == -1) {
2057 for (dl = ddf->dlist; dl ; dl = dl->next)
2058 if (dl->raiddisk == dk->raid_disk)
2059 break;
2060 } else {
2061 for (dl = ddf->dlist; dl ; dl = dl->next)
2062 if (dl->major == dk->major &&
2063 dl->minor == dk->minor)
2064 break;
2065 }
5f8097be
NB
2066 if (!dl || ! (dk->state & (1<<MD_DISK_SYNC)))
2067 return;
2068
d2ca6449
NB
2069 vc = &ddf->currentconf->conf;
2070 lba_offset = ddf->currentconf->lba_offset;
59e36268
NB
2071
2072 ex = get_extents(ddf, dl);
2073 if (!ex)
2074 return;
2075
2076 i = 0; pos = 0;
2077 blocks = __be64_to_cpu(vc->blocks);
d2ca6449
NB
2078 if (ddf->currentconf->block_sizes)
2079 blocks = ddf->currentconf->block_sizes[dk->raid_disk];
59e36268
NB
2080
2081 do {
2082 esize = ex[i].start - pos;
2083 if (esize >= blocks)
2084 break;
2085 pos = ex[i].start + ex[i].size;
2086 i++;
2087 } while (ex[i-1].size);
2088
2089 free(ex);
2090 if (esize < blocks)
2091 return;
2092
d2ca6449 2093 ddf->currentdev = dk->raid_disk;
5f8097be 2094 vc->phys_refnum[dk->raid_disk] = dl->disk.refnum;
59e36268 2095 lba_offset[dk->raid_disk] = __cpu_to_be64(pos);
5f8097be 2096
5575e7d9
NB
2097 for (i=0; i < ddf->max_part ; i++)
2098 if (dl->vlist[i] == NULL)
2099 break;
2100 if (i == ddf->max_part)
2101 return;
d2ca6449 2102 dl->vlist[i] = ddf->currentconf;
5f8097be 2103
8592f29d
N
2104 if (fd >= 0)
2105 dl->fd = fd;
2106 if (devname)
2107 dl->devname = devname;
7a7cc504
NB
2108
2109 /* Check how many working raid_disks, and if we can mark
2110 * array as optimal yet
2111 */
2112 working = 0;
5575e7d9 2113
7a7cc504
NB
2114 for (i=0; i < __be16_to_cpu(vc->prim_elmnt_count); i++)
2115 if (vc->phys_refnum[i] != 0xffffffff)
2116 working++;
59e36268 2117
5575e7d9 2118 /* Find which virtual_entry */
d2ca6449 2119 i = ddf->currentconf->vcnum;
7a7cc504 2120 if (working == __be16_to_cpu(vc->prim_elmnt_count))
5575e7d9
NB
2121 ddf->virt->entries[i].state =
2122 (ddf->virt->entries[i].state & ~DDF_state_mask)
7a7cc504
NB
2123 | DDF_state_optimal;
2124
2125 if (vc->prl == DDF_RAID6 &&
2126 working+1 == __be16_to_cpu(vc->prim_elmnt_count))
5575e7d9
NB
2127 ddf->virt->entries[i].state =
2128 (ddf->virt->entries[i].state & ~DDF_state_mask)
7a7cc504 2129 | DDF_state_part_optimal;
5575e7d9
NB
2130
2131 ddf->phys->entries[dl->pdnum].type &= ~__cpu_to_be16(DDF_Global_Spare);
2132 ddf->phys->entries[dl->pdnum].type |= __cpu_to_be16(DDF_Active_in_VD);
18a2f463 2133 ddf->updates_pending = 1;
5f8097be
NB
2134}
2135
a322f70c
DW
2136/* add a device to a container, either while creating it or while
2137 * expanding a pre-existing container
2138 */
f20c3968 2139static int add_to_super_ddf(struct supertype *st,
a322f70c
DW
2140 mdu_disk_info_t *dk, int fd, char *devname)
2141{
2142 struct ddf_super *ddf = st->sb;
2143 struct dl *dd;
2144 time_t now;
2145 struct tm *tm;
2146 unsigned long long size;
2147 struct phys_disk_entry *pde;
2148 int n, i;
2149 struct stat stb;
2150
78e44928
NB
2151 if (ddf->currentconf) {
2152 add_to_super_ddf_bvd(st, dk, fd, devname);
f20c3968 2153 return 0;
78e44928
NB
2154 }
2155
a322f70c
DW
2156 /* This is device numbered dk->number. We need to create
2157 * a phys_disk entry and a more detailed disk_data entry.
2158 */
2159 fstat(fd, &stb);
3d2c4fc7
DW
2160 if (posix_memalign((void**)&dd, 512,
2161 sizeof(*dd) + sizeof(dd->vlist[0]) * ddf->max_part) != 0) {
2162 fprintf(stderr, Name
2163 ": %s could allocate buffer for new disk, aborting\n",
2164 __func__);
f20c3968 2165 return 1;
3d2c4fc7 2166 }
a322f70c
DW
2167 dd->major = major(stb.st_rdev);
2168 dd->minor = minor(stb.st_rdev);
2169 dd->devname = devname;
a322f70c 2170 dd->fd = fd;
b2280677 2171 dd->spare = NULL;
a322f70c
DW
2172
2173 dd->disk.magic = DDF_PHYS_DATA_MAGIC;
2174 now = time(0);
2175 tm = localtime(&now);
2176 sprintf(dd->disk.guid, "%8s%04d%02d%02d",
2177 T10, tm->tm_year+1900, tm->tm_mon+1, tm->tm_mday);
bfb7ea78
N
2178 *(__u32*)(dd->disk.guid + 16) = random32();
2179 *(__u32*)(dd->disk.guid + 20) = random32();
a322f70c 2180
59e36268
NB
2181 do {
2182 /* Cannot be bothered finding a CRC of some irrelevant details*/
bfb7ea78 2183 dd->disk.refnum = random32();
59e36268
NB
2184 for (i = __be16_to_cpu(ddf->active->max_pd_entries) - 1;
2185 i >= 0; i--)
2186 if (ddf->phys->entries[i].refnum == dd->disk.refnum)
2187 break;
2188 } while (i >= 0);
2189
a322f70c
DW
2190 dd->disk.forced_ref = 1;
2191 dd->disk.forced_guid = 1;
2192 memset(dd->disk.vendor, ' ', 32);
2193 memcpy(dd->disk.vendor, "Linux", 5);
2194 memset(dd->disk.pad, 0xff, 442);
b2280677 2195 for (i = 0; i < ddf->max_part ; i++)
a322f70c
DW
2196 dd->vlist[i] = NULL;
2197
2198 n = __be16_to_cpu(ddf->phys->used_pdes);
2199 pde = &ddf->phys->entries[n];
5575e7d9
NB
2200 dd->pdnum = n;
2201
2cc2983d
N
2202 if (st->update_tail) {
2203 int len = (sizeof(struct phys_disk) +
2204 sizeof(struct phys_disk_entry));
2205 struct phys_disk *pd;
2206
2207 pd = malloc(len);
2208 pd->magic = DDF_PHYS_RECORDS_MAGIC;
2209 pd->used_pdes = __cpu_to_be16(n);
2210 pde = &pd->entries[0];
2211 dd->mdupdate = pd;
2212 } else {
2213 n++;
2214 ddf->phys->used_pdes = __cpu_to_be16(n);
2215 }
a322f70c
DW
2216
2217 memcpy(pde->guid, dd->disk.guid, DDF_GUID_LEN);
2218 pde->refnum = dd->disk.refnum;
5575e7d9 2219 pde->type = __cpu_to_be16(DDF_Forced_PD_GUID | DDF_Global_Spare);
a322f70c
DW
2220 pde->state = __cpu_to_be16(DDF_Online);
2221 get_dev_size(fd, NULL, &size);
2222 /* We are required to reserve 32Meg, and record the size in sectors */
2223 pde->config_size = __cpu_to_be64( (size - 32*1024*1024) / 512);
2224 sprintf(pde->path, "%17.17s","Information: nil") ;
2225 memset(pde->pad, 0xff, 6);
2226
d2ca6449 2227 dd->size = size >> 9;
2cc2983d
N
2228 if (st->update_tail) {
2229 dd->next = ddf->add_list;
2230 ddf->add_list = dd;
2231 } else {
2232 dd->next = ddf->dlist;
2233 ddf->dlist = dd;
2234 ddf->updates_pending = 1;
2235 }
f20c3968
DW
2236
2237 return 0;
a322f70c
DW
2238}
2239
2240/*
2241 * This is the write_init_super method for a ddf container. It is
2242 * called when creating a container or adding another device to a
2243 * container.
2244 */
2245
6416d527 2246static unsigned char null_conf[4096+512];
18a2f463 2247
7a7cc504 2248static int __write_init_super_ddf(struct supertype *st, int do_close)
a322f70c
DW
2249{
2250
2251 struct ddf_super *ddf = st->sb;
2252 int i;
2253 struct dl *d;
2254 int n_config;
2255 int conf_size;
175593bf
DW
2256 int attempts = 0;
2257 int successes = 0;
a322f70c
DW
2258 unsigned long long size, sector;
2259
175593bf
DW
2260 /* try to write updated metadata,
2261 * if we catch a failure move on to the next disk
2262 */
a322f70c
DW
2263 for (d = ddf->dlist; d; d=d->next) {
2264 int fd = d->fd;
2265
2266 if (fd < 0)
2267 continue;
2268
175593bf 2269 attempts++;
a322f70c
DW
2270 /* We need to fill in the primary, (secondary) and workspace
2271 * lba's in the headers, set their checksums,
2272 * Also checksum phys, virt....
2273 *
2274 * Then write everything out, finally the anchor is written.
2275 */
2276 get_dev_size(fd, NULL, &size);
2277 size /= 512;
2278 ddf->anchor.workspace_lba = __cpu_to_be64(size - 32*1024*2);
2279 ddf->anchor.primary_lba = __cpu_to_be64(size - 16*1024*2);
2280 ddf->anchor.seq = __cpu_to_be32(1);
2281 memcpy(&ddf->primary, &ddf->anchor, 512);
2282 memcpy(&ddf->secondary, &ddf->anchor, 512);
2283
2284 ddf->anchor.openflag = 0xFF; /* 'open' means nothing */
2285 ddf->anchor.seq = 0xFFFFFFFF; /* no sequencing in anchor */
2286 ddf->anchor.crc = calc_crc(&ddf->anchor, 512);
2287
2288 ddf->primary.openflag = 0;
2289 ddf->primary.type = DDF_HEADER_PRIMARY;
2290
2291 ddf->secondary.openflag = 0;
2292 ddf->secondary.type = DDF_HEADER_SECONDARY;
2293
2294 ddf->primary.crc = calc_crc(&ddf->primary, 512);
2295 ddf->secondary.crc = calc_crc(&ddf->secondary, 512);
2296
2297 sector = size - 16*1024*2;
2298 lseek64(fd, sector<<9, 0);
175593bf
DW
2299 if (write(fd, &ddf->primary, 512) < 0)
2300 continue;
a322f70c
DW
2301
2302 ddf->controller.crc = calc_crc(&ddf->controller, 512);
175593bf
DW
2303 if (write(fd, &ddf->controller, 512) < 0)
2304 continue;
a322f70c
DW
2305
2306 ddf->phys->crc = calc_crc(ddf->phys, ddf->pdsize);
2307
175593bf
DW
2308 if (write(fd, ddf->phys, ddf->pdsize) < 0)
2309 continue;
a322f70c
DW
2310
2311 ddf->virt->crc = calc_crc(ddf->virt, ddf->vdsize);
175593bf
DW
2312 if (write(fd, ddf->virt, ddf->vdsize) < 0)
2313 continue;
a322f70c
DW
2314
2315 /* Now write lots of config records. */
8c3b8c2c
NB
2316 n_config = ddf->max_part;
2317 conf_size = ddf->conf_rec_len * 512;
a322f70c
DW
2318 for (i = 0 ; i <= n_config ; i++) {
2319 struct vcl *c = d->vlist[i];
b2280677
NB
2320 if (i == n_config)
2321 c = (struct vcl*)d->spare;
a322f70c
DW
2322
2323 if (c) {
2324 c->conf.crc = calc_crc(&c->conf, conf_size);
175593bf
DW
2325 if (write(fd, &c->conf, conf_size) < 0)
2326 break;
a322f70c 2327 } else {
6416d527 2328 char *null_aligned = (char*)((((unsigned long)null_conf)+511)&~511UL);
18a2f463
NB
2329 if (null_conf[0] != 0xff)
2330 memset(null_conf, 0xff, sizeof(null_conf));
2331 int togo = conf_size;
6416d527 2332 while (togo > sizeof(null_conf)-512) {
175593bf
DW
2333 if (write(fd, null_aligned, sizeof(null_conf)-512) < 0)
2334 break;
6416d527 2335 togo -= sizeof(null_conf)-512;
18a2f463 2336 }
175593bf
DW
2337 if (write(fd, null_aligned, togo) < 0)
2338 break;
a322f70c
DW
2339 }
2340 }
175593bf
DW
2341 if (i <= n_config)
2342 continue;
a322f70c 2343 d->disk.crc = calc_crc(&d->disk, 512);
175593bf
DW
2344 if (write(fd, &d->disk, 512) < 0)
2345 continue;
a322f70c
DW
2346
2347 /* Maybe do the same for secondary */
2348
2349 lseek64(fd, (size-1)*512, SEEK_SET);
175593bf
DW
2350 if (write(fd, &ddf->anchor, 512) < 0)
2351 continue;
2352 successes++;
2353 }
2354
2355 if (do_close)
2356 for (d = ddf->dlist; d; d=d->next) {
2357 close(d->fd);
7a7cc504
NB
2358 d->fd = -1;
2359 }
175593bf
DW
2360
2361 return attempts != successes;
a322f70c 2362}
7a7cc504
NB
2363
2364static int write_init_super_ddf(struct supertype *st)
2365{
9b1fb677
DW
2366 struct ddf_super *ddf = st->sb;
2367 struct vcl *currentconf = ddf->currentconf;
2368
2369 /* we are done with currentconf reset it to point st at the container */
2370 ddf->currentconf = NULL;
edd8d13c
NB
2371
2372 if (st->update_tail) {
2373 /* queue the virtual_disk and vd_config as metadata updates */
2374 struct virtual_disk *vd;
2375 struct vd_config *vc;
edd8d13c
NB
2376 int len;
2377
9b1fb677 2378 if (!currentconf) {
2cc2983d
N
2379 int len = (sizeof(struct phys_disk) +
2380 sizeof(struct phys_disk_entry));
2381
2382 /* adding a disk to the container. */
2383 if (!ddf->add_list)
2384 return 0;
2385
2386 append_metadata_update(st, ddf->add_list->mdupdate, len);
2387 ddf->add_list->mdupdate = NULL;
2388 return 0;
2389 }
2390
2391 /* Newly created VD */
2392
edd8d13c
NB
2393 /* First the virtual disk. We have a slightly fake header */
2394 len = sizeof(struct virtual_disk) + sizeof(struct virtual_entry);
2395 vd = malloc(len);
2396 *vd = *ddf->virt;
9b1fb677
DW
2397 vd->entries[0] = ddf->virt->entries[currentconf->vcnum];
2398 vd->populated_vdes = __cpu_to_be16(currentconf->vcnum);
edd8d13c
NB
2399 append_metadata_update(st, vd, len);
2400
2401 /* Then the vd_config */
2402 len = ddf->conf_rec_len * 512;
2403 vc = malloc(len);
9b1fb677 2404 memcpy(vc, &currentconf->conf, len);
edd8d13c
NB
2405 append_metadata_update(st, vc, len);
2406
2407 /* FIXME I need to close the fds! */
2408 return 0;
2409 } else
2410 return __write_init_super_ddf(st, 1);
7a7cc504
NB
2411}
2412
a322f70c
DW
2413#endif
2414
2415static __u64 avail_size_ddf(struct supertype *st, __u64 devsize)
2416{
2417 /* We must reserve the last 32Meg */
2418 if (devsize <= 32*1024*2)
2419 return 0;
2420 return devsize - 32*1024*2;
2421}
2422
2423#ifndef MDASSEMBLE
8592f29d
N
2424
2425static int reserve_space(struct supertype *st, int raiddisks,
2426 unsigned long long size, int chunk,
2427 unsigned long long *freesize)
2428{
2429 /* Find 'raiddisks' spare extents at least 'size' big (but
2430 * only caring about multiples of 'chunk') and remember
2431 * them.
2432 * If the cannot be found, fail.
2433 */
2434 struct dl *dl;
2435 struct ddf_super *ddf = st->sb;
2436 int cnt = 0;
2437
2438 for (dl = ddf->dlist; dl ; dl=dl->next) {
2439 dl->raiddisk = -1;
2440 dl->esize = 0;
2441 }
2442 /* Now find largest extent on each device */
2443 for (dl = ddf->dlist ; dl ; dl=dl->next) {
2444 struct extent *e = get_extents(ddf, dl);
2445 unsigned long long pos = 0;
2446 int i = 0;
2447 int found = 0;
2448 unsigned long long minsize = size;
2449
2450 if (size == 0)
2451 minsize = chunk;
2452
2453 if (!e)
2454 continue;
2455 do {
2456 unsigned long long esize;
2457 esize = e[i].start - pos;
2458 if (esize >= minsize) {
2459 found = 1;
2460 minsize = esize;
2461 }
2462 pos = e[i].start + e[i].size;
2463 i++;
2464 } while (e[i-1].size);
2465 if (found) {
2466 cnt++;
2467 dl->esize = minsize;
2468 }
2469 free(e);
2470 }
2471 if (cnt < raiddisks) {
2472 fprintf(stderr, Name ": not enough devices with space to create array.\n");
2473 return 0; /* No enough free spaces large enough */
2474 }
2475 if (size == 0) {
2476 /* choose the largest size of which there are at least 'raiddisk' */
2477 for (dl = ddf->dlist ; dl ; dl=dl->next) {
2478 struct dl *dl2;
2479 if (dl->esize <= size)
2480 continue;
2481 /* This is bigger than 'size', see if there are enough */
2482 cnt = 0;
2483 for (dl2 = dl; dl2 ; dl2=dl2->next)
2484 if (dl2->esize >= dl->esize)
2485 cnt++;
2486 if (cnt >= raiddisks)
2487 size = dl->esize;
2488 }
2489 if (chunk) {
2490 size = size / chunk;
2491 size *= chunk;
2492 }
2493 *freesize = size;
2494 if (size < 32) {
2495 fprintf(stderr, Name ": not enough spare devices to create array.\n");
2496 return 0;
2497 }
2498 }
2499 /* We have a 'size' of which there are enough spaces.
2500 * We simply do a first-fit */
2501 cnt = 0;
2502 for (dl = ddf->dlist ; dl && cnt < raiddisks ; dl=dl->next) {
2503 if (dl->esize < size)
2504 continue;
2505
2506 dl->raiddisk = cnt;
2507 cnt++;
2508 }
2509 return 1;
2510}
2511
2512
2513
2c514b71
NB
2514static int
2515validate_geometry_ddf_container(struct supertype *st,
2516 int level, int layout, int raiddisks,
2517 int chunk, unsigned long long size,
2518 char *dev, unsigned long long *freesize,
2519 int verbose);
78e44928
NB
2520
2521static int validate_geometry_ddf_bvd(struct supertype *st,
2522 int level, int layout, int raiddisks,
2523 int chunk, unsigned long long size,
2c514b71
NB
2524 char *dev, unsigned long long *freesize,
2525 int verbose);
78e44928
NB
2526
2527static int validate_geometry_ddf(struct supertype *st,
2c514b71
NB
2528 int level, int layout, int raiddisks,
2529 int chunk, unsigned long long size,
2530 char *dev, unsigned long long *freesize,
2531 int verbose)
a322f70c
DW
2532{
2533 int fd;
2534 struct mdinfo *sra;
2535 int cfd;
2536
2537 /* ddf potentially supports lots of things, but it depends on
2538 * what devices are offered (and maybe kernel version?)
2539 * If given unused devices, we will make a container.
2540 * If given devices in a container, we will make a BVD.
2541 * If given BVDs, we make an SVD, changing all the GUIDs in the process.
2542 */
2543
2544 if (level == LEVEL_CONTAINER) {
78e44928
NB
2545 /* Must be a fresh device to add to a container */
2546 return validate_geometry_ddf_container(st, level, layout,
2c514b71
NB
2547 raiddisks, chunk,
2548 size, dev, freesize,
2549 verbose);
5f8097be
NB
2550 }
2551
78e44928
NB
2552 if (!dev) {
2553 /* Initial sanity check. Exclude illegal levels. */
2554 int i;
2555 for (i=0; ddf_level_num[i].num1 != MAXINT; i++)
2556 if (ddf_level_num[i].num2 == level)
2557 break;
b42f577a
N
2558 if (ddf_level_num[i].num1 == MAXINT) {
2559 if (verbose)
2560 fprintf(stderr, Name ": DDF does not support level %d arrays\n",
2561 level);
78e44928 2562 return 0;
b42f577a 2563 }
78e44928 2564 /* Should check layout? etc */
8592f29d
N
2565
2566 if (st->sb && freesize) {
2567 /* --create was given a container to create in.
2568 * So we need to check that there are enough
2569 * free spaces and return the amount of space.
2570 * We may as well remember which drives were
2571 * chosen so that add_to_super/getinfo_super
2572 * can return them.
2573 */
2574 return reserve_space(st, raiddisks, size, chunk, freesize);
2575 }
a322f70c 2576 return 1;
78e44928 2577 }
a322f70c 2578
8592f29d
N
2579 if (st->sb) {
2580 /* A container has already been opened, so we are
2581 * creating in there. Maybe a BVD, maybe an SVD.
2582 * Should make a distinction one day.
2583 */
2584 return validate_geometry_ddf_bvd(st, level, layout, raiddisks,
2585 chunk, size, dev, freesize,
2586 verbose);
2587 }
78e44928
NB
2588 /* This is the first device for the array.
2589 * If it is a container, we read it in and do automagic allocations,
2590 * no other devices should be given.
2591 * Otherwise it must be a member device of a container, and we
2592 * do manual allocation.
2593 * Later we should check for a BVD and make an SVD.
a322f70c 2594 */
a322f70c
DW
2595 fd = open(dev, O_RDONLY|O_EXCL, 0);
2596 if (fd >= 0) {
2597 sra = sysfs_read(fd, 0, GET_VERSION);
2598 close(fd);
2599 if (sra && sra->array.major_version == -1 &&
78e44928
NB
2600 strcmp(sra->text_version, "ddf") == 0) {
2601
2602 /* load super */
2603 /* find space for 'n' devices. */
2604 /* remember the devices */
2605 /* Somehow return the fact that we have enough */
a322f70c
DW
2606 }
2607
2c514b71
NB
2608 if (verbose)
2609 fprintf(stderr,
2610 Name ": ddf: Cannot create this array "
b42f577a 2611 "on device %s - a container is required.\n",
2c514b71 2612 dev);
a322f70c
DW
2613 return 0;
2614 }
2615 if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) {
2c514b71
NB
2616 if (verbose)
2617 fprintf(stderr, Name ": ddf: Cannot open %s: %s\n",
2618 dev, strerror(errno));
a322f70c
DW
2619 return 0;
2620 }
2621 /* Well, it is in use by someone, maybe a 'ddf' container. */
2622 cfd = open_container(fd);
2623 if (cfd < 0) {
2624 close(fd);
2c514b71
NB
2625 if (verbose)
2626 fprintf(stderr, Name ": ddf: Cannot use %s: %s\n",
2627 dev, strerror(EBUSY));
a322f70c
DW
2628 return 0;
2629 }
2630 sra = sysfs_read(cfd, 0, GET_VERSION);
2631 close(fd);
2632 if (sra && sra->array.major_version == -1 &&
2633 strcmp(sra->text_version, "ddf") == 0) {
2634 /* This is a member of a ddf container. Load the container
2635 * and try to create a bvd
2636 */
2637 struct ddf_super *ddf;
a322f70c 2638 if (load_super_ddf_all(st, cfd, (void **)&ddf, NULL, 1) == 0) {
5f8097be 2639 st->sb = ddf;
2f6079dc 2640 st->container_dev = fd2devnum(cfd);
a322f70c 2641 close(cfd);
78e44928 2642 return validate_geometry_ddf_bvd(st, level, layout,
a322f70c 2643 raiddisks, chunk, size,
2c514b71
NB
2644 dev, freesize,
2645 verbose);
a322f70c
DW
2646 }
2647 close(cfd);
c42ec1ed
DW
2648 } else /* device may belong to a different container */
2649 return 0;
2650
a322f70c
DW
2651 return 1;
2652}
2653
2c514b71
NB
2654static int
2655validate_geometry_ddf_container(struct supertype *st,
2656 int level, int layout, int raiddisks,
2657 int chunk, unsigned long long size,
2658 char *dev, unsigned long long *freesize,
2659 int verbose)
a322f70c
DW
2660{
2661 int fd;
2662 unsigned long long ldsize;
2663
2664 if (level != LEVEL_CONTAINER)
2665 return 0;
2666 if (!dev)
2667 return 1;
2668
2669 fd = open(dev, O_RDONLY|O_EXCL, 0);
2670 if (fd < 0) {
2c514b71
NB
2671 if (verbose)
2672 fprintf(stderr, Name ": ddf: Cannot open %s: %s\n",
2673 dev, strerror(errno));
a322f70c
DW
2674 return 0;
2675 }
2676 if (!get_dev_size(fd, dev, &ldsize)) {
2677 close(fd);
2678 return 0;
2679 }
2680 close(fd);
2681
56aca704 2682 *freesize = avail_size_ddf(st, ldsize >> 9);
ea17e7aa
N
2683 if (*freesize == 0)
2684 return 0;
a322f70c
DW
2685
2686 return 1;
2687}
2688
78e44928
NB
2689static int validate_geometry_ddf_bvd(struct supertype *st,
2690 int level, int layout, int raiddisks,
2691 int chunk, unsigned long long size,
2c514b71
NB
2692 char *dev, unsigned long long *freesize,
2693 int verbose)
a322f70c
DW
2694{
2695 struct stat stb;
2696 struct ddf_super *ddf = st->sb;
2697 struct dl *dl;
5f8097be
NB
2698 unsigned long long pos = 0;
2699 unsigned long long maxsize;
2700 struct extent *e;
2701 int i;
a322f70c 2702 /* ddf/bvd supports lots of things, but not containers */
b42f577a
N
2703 if (level == LEVEL_CONTAINER) {
2704 if (verbose)
2705 fprintf(stderr, Name ": DDF cannot create a container within an container\n");
a322f70c 2706 return 0;
b42f577a 2707 }
a322f70c
DW
2708 /* We must have the container info already read in. */
2709 if (!ddf)
2710 return 0;
2711
5f8097be
NB
2712 if (!dev) {
2713 /* General test: make sure there is space for
2714 * 'raiddisks' device extents of size 'size'.
2715 */
2716 unsigned long long minsize = size;
2717 int dcnt = 0;
2718 if (minsize == 0)
2719 minsize = 8;
2720 for (dl = ddf->dlist; dl ; dl = dl->next)
2721 {
2722 int found = 0;
7e1432fb 2723 pos = 0;
5f8097be
NB
2724
2725 i = 0;
2726 e = get_extents(ddf, dl);
2727 if (!e) continue;
2728 do {
2729 unsigned long long esize;
2730 esize = e[i].start - pos;
2731 if (esize >= minsize)
2732 found = 1;
2733 pos = e[i].start + e[i].size;
2734 i++;
2735 } while (e[i-1].size);
2736 if (found)
2737 dcnt++;
2738 free(e);
2739 }
2740 if (dcnt < raiddisks) {
2c514b71
NB
2741 if (verbose)
2742 fprintf(stderr,
2743 Name ": ddf: Not enough devices with "
2744 "space for this array (%d < %d)\n",
2745 dcnt, raiddisks);
5f8097be
NB
2746 return 0;
2747 }
2748 return 1;
2749 }
a322f70c
DW
2750 /* This device must be a member of the set */
2751 if (stat(dev, &stb) < 0)
2752 return 0;
2753 if ((S_IFMT & stb.st_mode) != S_IFBLK)
2754 return 0;
2755 for (dl = ddf->dlist ; dl ; dl = dl->next) {
2756 if (dl->major == major(stb.st_rdev) &&
2757 dl->minor == minor(stb.st_rdev))
2758 break;
2759 }
5f8097be 2760 if (!dl) {
2c514b71
NB
2761 if (verbose)
2762 fprintf(stderr, Name ": ddf: %s is not in the "
2763 "same DDF set\n",
2764 dev);
5f8097be
NB
2765 return 0;
2766 }
2767 e = get_extents(ddf, dl);
2768 maxsize = 0;
2769 i = 0;
2770 if (e) do {
2771 unsigned long long esize;
2772 esize = e[i].start - pos;
2773 if (esize >= maxsize)
2774 maxsize = esize;
2775 pos = e[i].start + e[i].size;
2776 i++;
2777 } while (e[i-1].size);
2778 *freesize = maxsize;
a322f70c
DW
2779 // FIXME here I am
2780
2781 return 1;
2782}
59e36268 2783
a322f70c
DW
2784static int load_super_ddf_all(struct supertype *st, int fd,
2785 void **sbp, char *devname, int keep_fd)
2786{
2787 struct mdinfo *sra;
2788 struct ddf_super *super;
2789 struct mdinfo *sd, *best = NULL;
2790 int bestseq = 0;
2791 int seq;
2792 char nm[20];
2793 int dfd;
dab4a513
DW
2794 int devnum = fd2devnum(fd);
2795 enum sysfs_read_flags flags;
a322f70c 2796
dab4a513
DW
2797 flags = GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE;
2798 if (mdmon_running(devnum))
2799 flags |= SKIP_GONE_DEVS;
2800
2801 sra = sysfs_read(fd, 0, flags);
a322f70c
DW
2802 if (!sra)
2803 return 1;
2804 if (sra->array.major_version != -1 ||
2805 sra->array.minor_version != -2 ||
2806 strcmp(sra->text_version, "ddf") != 0)
2807 return 1;
2808
6416d527 2809 if (posix_memalign((void**)&super, 512, sizeof(*super)) != 0)
a322f70c 2810 return 1;
a2349791 2811 memset(super, 0, sizeof(*super));
a322f70c
DW
2812
2813 /* first, try each device, and choose the best ddf */
2814 for (sd = sra->devs ; sd ; sd = sd->next) {
2815 int rv;
2816 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
7a7cc504
NB
2817 dfd = dev_open(nm, O_RDONLY);
2818 if (dfd < 0)
a322f70c
DW
2819 return 2;
2820 rv = load_ddf_headers(dfd, super, NULL);
7a7cc504 2821 close(dfd);
a322f70c
DW
2822 if (rv == 0) {
2823 seq = __be32_to_cpu(super->active->seq);
2824 if (super->active->openflag)
2825 seq--;
2826 if (!best || seq > bestseq) {
2827 bestseq = seq;
2828 best = sd;
2829 }
2830 }
2831 }
2832 if (!best)
2833 return 1;
2834 /* OK, load this ddf */
2835 sprintf(nm, "%d:%d", best->disk.major, best->disk.minor);
2836 dfd = dev_open(nm, O_RDONLY);
7a7cc504 2837 if (dfd < 0)
a322f70c
DW
2838 return 1;
2839 load_ddf_headers(dfd, super, NULL);
2840 load_ddf_global(dfd, super, NULL);
2841 close(dfd);
2842 /* Now we need the device-local bits */
2843 for (sd = sra->devs ; sd ; sd = sd->next) {
3d2c4fc7
DW
2844 int rv;
2845
a322f70c 2846 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
5f8097be 2847 dfd = dev_open(nm, keep_fd? O_RDWR : O_RDONLY);
7a7cc504 2848 if (dfd < 0)
a322f70c 2849 return 2;
3d2c4fc7
DW
2850 rv = load_ddf_headers(dfd, super, NULL);
2851 if (rv == 0)
2852 rv = load_ddf_local(dfd, super, NULL, keep_fd);
5f8097be 2853 if (!keep_fd) close(dfd);
3d2c4fc7
DW
2854 if (rv)
2855 return 1;
a322f70c 2856 }
f7e7067b
NB
2857 if (st->subarray[0]) {
2858 struct vcl *v;
2859
2860 for (v = super->conflist; v; v = v->next)
2861 if (v->vcnum == atoi(st->subarray))
d2ca6449
NB
2862 super->currentconf = v;
2863 if (!super->currentconf)
f7e7067b
NB
2864 return 1;
2865 }
a322f70c
DW
2866 *sbp = super;
2867 if (st->ss == NULL) {
78e44928 2868 st->ss = &super_ddf;
a322f70c
DW
2869 st->minor_version = 0;
2870 st->max_devs = 512;
75aa18b5 2871 st->container_dev = fd2devnum(fd);
a322f70c 2872 }
352452c3 2873 st->loaded_container = 1;
a322f70c
DW
2874 return 0;
2875}
0e600426 2876#endif /* MDASSEMBLE */
a322f70c 2877
598f0d58
NB
2878static struct mdinfo *container_content_ddf(struct supertype *st)
2879{
2880 /* Given a container loaded by load_super_ddf_all,
2881 * extract information about all the arrays into
2882 * an mdinfo tree.
2883 *
2884 * For each vcl in conflist: create an mdinfo, fill it in,
2885 * then look for matching devices (phys_refnum) in dlist
2886 * and create appropriate device mdinfo.
2887 */
2888 struct ddf_super *ddf = st->sb;
2889 struct mdinfo *rest = NULL;
2890 struct vcl *vc;
2891
2892 for (vc = ddf->conflist ; vc ; vc=vc->next)
2893 {
598f0d58 2894 int i;
db42fa9b 2895 int j;
598f0d58
NB
2896 struct mdinfo *this;
2897 this = malloc(sizeof(*this));
2898 memset(this, 0, sizeof(*this));
2899 this->next = rest;
2900 rest = this;
2901
598f0d58
NB
2902 this->array.level = map_num1(ddf_level_num, vc->conf.prl);
2903 this->array.raid_disks =
2904 __be16_to_cpu(vc->conf.prim_elmnt_count);
60f18132
NB
2905 this->array.layout = rlq_to_layout(vc->conf.rlq, vc->conf.prl,
2906 this->array.raid_disks);
598f0d58 2907 this->array.md_minor = -1;
f35f2525
N
2908 this->array.major_version = -1;
2909 this->array.minor_version = -2;
598f0d58
NB
2910 this->array.ctime = DECADE +
2911 __be32_to_cpu(*(__u32*)(vc->conf.guid+16));
2912 this->array.utime = DECADE +
2913 __be32_to_cpu(vc->conf.timestamp);
2914 this->array.chunk_size = 512 << vc->conf.chunk_shift;
2915
59e36268 2916 i = vc->vcnum;
7a7cc504
NB
2917 if ((ddf->virt->entries[i].state & DDF_state_inconsistent) ||
2918 (ddf->virt->entries[i].init_state & DDF_initstate_mask) !=
ed9d66aa 2919 DDF_init_full) {
598f0d58 2920 this->array.state = 0;
ed9d66aa
NB
2921 this->resync_start = 0;
2922 } else {
598f0d58 2923 this->array.state = 1;
ed9d66aa
NB
2924 this->resync_start = ~0ULL;
2925 }
db42fa9b
N
2926 memcpy(this->name, ddf->virt->entries[i].name, 16);
2927 this->name[16]=0;
2928 for(j=0; j<16; j++)
2929 if (this->name[j] == ' ')
2930 this->name[j] = 0;
598f0d58
NB
2931
2932 memset(this->uuid, 0, sizeof(this->uuid));
2933 this->component_size = __be64_to_cpu(vc->conf.blocks);
2934 this->array.size = this->component_size / 2;
5f2aace8 2935 this->container_member = i;
598f0d58 2936
c5afc314
N
2937 ddf->currentconf = vc;
2938 uuid_from_super_ddf(st, this->uuid);
2939 ddf->currentconf = NULL;
2940
60f18132
NB
2941 sprintf(this->text_version, "/%s/%d",
2942 devnum2devname(st->container_dev),
2943 this->container_member);
2944
8c3b8c2c 2945 for (i=0 ; i < ddf->mppe ; i++) {
598f0d58
NB
2946 struct mdinfo *dev;
2947 struct dl *d;
2948
2949 if (vc->conf.phys_refnum[i] == 0xFFFFFFFF)
2950 continue;
2951
c0d50ca5
N
2952 this->array.working_disks++;
2953
598f0d58
NB
2954 for (d = ddf->dlist; d ; d=d->next)
2955 if (d->disk.refnum == vc->conf.phys_refnum[i])
2956 break;
2957 if (d == NULL)
bc17324f
N
2958 /* Haven't found that one yet, maybe there are others */
2959 continue;
2960
598f0d58
NB
2961 dev = malloc(sizeof(*dev));
2962 memset(dev, 0, sizeof(*dev));
2963 dev->next = this->devs;
2964 this->devs = dev;
2965
2966 dev->disk.number = __be32_to_cpu(d->disk.refnum);
2967 dev->disk.major = d->major;
2968 dev->disk.minor = d->minor;
2969 dev->disk.raid_disk = i;
2970 dev->disk.state = (1<<MD_DISK_SYNC)|(1<<MD_DISK_ACTIVE);
2971
120f7677
NB
2972 dev->events = __be32_to_cpu(ddf->primary.seq);
2973 dev->data_offset = __be64_to_cpu(vc->lba_offset[i]);
598f0d58
NB
2974 dev->component_size = __be64_to_cpu(vc->conf.blocks);
2975 if (d->devname)
2976 strcpy(dev->name, d->devname);
2977 }
2978 }
2979 return rest;
2980}
2981
955e9ea1 2982static int store_super_ddf(struct supertype *st, int fd)
a322f70c 2983{
955e9ea1 2984 struct ddf_super *ddf = st->sb;
a322f70c 2985 unsigned long long dsize;
6416d527 2986 void *buf;
3d2c4fc7 2987 int rc;
a322f70c 2988
955e9ea1
DW
2989 if (!ddf)
2990 return 1;
2991
2992 /* ->dlist and ->conflist will be set for updates, currently not
2993 * supported
2994 */
2995 if (ddf->dlist || ddf->conflist)
2996 return 1;
2997
a322f70c
DW
2998 if (!get_dev_size(fd, NULL, &dsize))
2999 return 1;
3000
3d2c4fc7
DW
3001 if (posix_memalign(&buf, 512, 512) != 0)
3002 return 1;
6416d527
NB
3003 memset(buf, 0, 512);
3004
a322f70c 3005 lseek64(fd, dsize-512, 0);
3d2c4fc7 3006 rc = write(fd, buf, 512);
6416d527 3007 free(buf);
3d2c4fc7
DW
3008 if (rc < 0)
3009 return 1;
a322f70c
DW
3010 return 0;
3011}
3012
a19c88b8
NB
3013static int compare_super_ddf(struct supertype *st, struct supertype *tst)
3014{
3015 /*
3016 * return:
3017 * 0 same, or first was empty, and second was copied
3018 * 1 second had wrong number
3019 * 2 wrong uuid
3020 * 3 wrong other info
3021 */
3022 struct ddf_super *first = st->sb;
3023 struct ddf_super *second = tst->sb;
3024
3025 if (!first) {
3026 st->sb = tst->sb;
3027 tst->sb = NULL;
3028 return 0;
3029 }
3030
3031 if (memcmp(first->anchor.guid, second->anchor.guid, DDF_GUID_LEN) != 0)
3032 return 2;
3033
3034 /* FIXME should I look at anything else? */
3035 return 0;
3036}
3037
0e600426 3038#ifndef MDASSEMBLE
4e5528c6
NB
3039/*
3040 * A new array 'a' has been started which claims to be instance 'inst'
3041 * within container 'c'.
3042 * We need to confirm that the array matches the metadata in 'c' so
3043 * that we don't corrupt any metadata.
3044 */
cba0191b 3045static int ddf_open_new(struct supertype *c, struct active_array *a, char *inst)
549e9569 3046{
2c514b71 3047 dprintf("ddf: open_new %s\n", inst);
cba0191b 3048 a->info.container_member = atoi(inst);
549e9569
NB
3049 return 0;
3050}
3051
4e5528c6
NB
3052/*
3053 * The array 'a' is to be marked clean in the metadata.
ed9d66aa 3054 * If '->resync_start' is not ~(unsigned long long)0, then the array is only
4e5528c6
NB
3055 * clean up to the point (in sectors). If that cannot be recorded in the
3056 * metadata, then leave it as dirty.
3057 *
3058 * For DDF, we need to clear the DDF_state_inconsistent bit in the
3059 * !global! virtual_disk.virtual_entry structure.
3060 */
01f157d7 3061static int ddf_set_array_state(struct active_array *a, int consistent)
549e9569 3062{
4e5528c6
NB
3063 struct ddf_super *ddf = a->container->sb;
3064 int inst = a->info.container_member;
18a2f463 3065 int old = ddf->virt->entries[inst].state;
01f157d7
N
3066 if (consistent == 2) {
3067 /* Should check if a recovery should be started FIXME */
3068 consistent = 1;
593add1b 3069 if (!is_resync_complete(a))
01f157d7
N
3070 consistent = 0;
3071 }
ed9d66aa
NB
3072 if (consistent)
3073 ddf->virt->entries[inst].state &= ~DDF_state_inconsistent;
3074 else
4e5528c6 3075 ddf->virt->entries[inst].state |= DDF_state_inconsistent;
18a2f463
NB
3076 if (old != ddf->virt->entries[inst].state)
3077 ddf->updates_pending = 1;
3078
3079 old = ddf->virt->entries[inst].init_state;
ed9d66aa 3080 ddf->virt->entries[inst].init_state &= ~DDF_initstate_mask;
593add1b 3081 if (is_resync_complete(a))
ed9d66aa
NB
3082 ddf->virt->entries[inst].init_state |= DDF_init_full;
3083 else if (a->resync_start == 0)
3084 ddf->virt->entries[inst].init_state |= DDF_init_not;
4e5528c6 3085 else
ed9d66aa 3086 ddf->virt->entries[inst].init_state |= DDF_init_quick;
18a2f463
NB
3087 if (old != ddf->virt->entries[inst].init_state)
3088 ddf->updates_pending = 1;
ed9d66aa 3089
2c514b71
NB
3090 dprintf("ddf mark %d %s %llu\n", inst, consistent?"clean":"dirty",
3091 a->resync_start);
01f157d7 3092 return consistent;
fd7cde1b
DW
3093}
3094
7a7cc504
NB
3095/*
3096 * The state of each disk is stored in the global phys_disk structure
3097 * in phys_disk.entries[n].state.
3098 * This makes various combinations awkward.
3099 * - When a device fails in any array, it must be failed in all arrays
3100 * that include a part of this device.
3101 * - When a component is rebuilding, we cannot include it officially in the
3102 * array unless this is the only array that uses the device.
3103 *
3104 * So: when transitioning:
3105 * Online -> failed, just set failed flag. monitor will propagate
3106 * spare -> online, the device might need to be added to the array.
3107 * spare -> failed, just set failed. Don't worry if in array or not.
3108 */
8d45d196 3109static void ddf_set_disk(struct active_array *a, int n, int state)
549e9569 3110{
7a7cc504
NB
3111 struct ddf_super *ddf = a->container->sb;
3112 int inst = a->info.container_member;
3113 struct vd_config *vc = find_vdcr(ddf, inst);
3114 int pd = find_phys(ddf, vc->phys_refnum[n]);
3115 int i, st, working;
3116
3117 if (vc == NULL) {
2c514b71 3118 dprintf("ddf: cannot find instance %d!!\n", inst);
7a7cc504
NB
3119 return;
3120 }
3121 if (pd < 0) {
3122 /* disk doesn't currently exist. If it is now in_sync,
3123 * insert it. */
3124 if ((state & DS_INSYNC) && ! (state & DS_FAULTY)) {
3125 /* Find dev 'n' in a->info->devs, determine the
3126 * ddf refnum, and set vc->phys_refnum and update
3127 * phys->entries[]
3128 */
3129 /* FIXME */
3130 }
3131 } else {
18a2f463 3132 int old = ddf->phys->entries[pd].state;
7a7cc504
NB
3133 if (state & DS_FAULTY)
3134 ddf->phys->entries[pd].state |= __cpu_to_be16(DDF_Failed);
3135 if (state & DS_INSYNC) {
3136 ddf->phys->entries[pd].state |= __cpu_to_be16(DDF_Online);
3137 ddf->phys->entries[pd].state &= __cpu_to_be16(~DDF_Rebuilding);
3138 }
18a2f463
NB
3139 if (old != ddf->phys->entries[pd].state)
3140 ddf->updates_pending = 1;
7a7cc504
NB
3141 }
3142
2c514b71 3143 dprintf("ddf: set_disk %d to %x\n", n, state);
7e1432fb 3144
7a7cc504
NB
3145 /* Now we need to check the state of the array and update
3146 * virtual_disk.entries[n].state.
3147 * It needs to be one of "optimal", "degraded", "failed".
3148 * I don't understand 'deleted' or 'missing'.
3149 */
3150 working = 0;
3151 for (i=0; i < a->info.array.raid_disks; i++) {
3152 pd = find_phys(ddf, vc->phys_refnum[i]);
3153 if (pd < 0)
3154 continue;
57632f4a
NB
3155 st = __be16_to_cpu(ddf->phys->entries[pd].state);
3156 if ((st & (DDF_Online|DDF_Failed|DDF_Rebuilding))
7a7cc504
NB
3157 == DDF_Online)
3158 working++;
3159 }
3160 state = DDF_state_degraded;
3161 if (working == a->info.array.raid_disks)
3162 state = DDF_state_optimal;
3163 else switch(vc->prl) {
3164 case DDF_RAID0:
3165 case DDF_CONCAT:
3166 case DDF_JBOD:
3167 state = DDF_state_failed;
3168 break;
3169 case DDF_RAID1:
3170 if (working == 0)
3171 state = DDF_state_failed;
3172 break;
3173 case DDF_RAID4:
3174 case DDF_RAID5:
3175 if (working < a->info.array.raid_disks-1)
3176 state = DDF_state_failed;
3177 break;
3178 case DDF_RAID6:
3179 if (working < a->info.array.raid_disks-2)
3180 state = DDF_state_failed;
3181 else if (working == a->info.array.raid_disks-1)
3182 state = DDF_state_part_optimal;
3183 break;
3184 }
3185
18a2f463
NB
3186 if (ddf->virt->entries[inst].state !=
3187 ((ddf->virt->entries[inst].state & ~DDF_state_mask)
3188 | state)) {
3189
3190 ddf->virt->entries[inst].state =
3191 (ddf->virt->entries[inst].state & ~DDF_state_mask)
3192 | state;
3193 ddf->updates_pending = 1;
3194 }
7a7cc504 3195
549e9569
NB
3196}
3197
2e735d19 3198static void ddf_sync_metadata(struct supertype *st)
549e9569 3199{
7a7cc504
NB
3200
3201 /*
3202 * Write all data to all devices.
3203 * Later, we might be able to track whether only local changes
3204 * have been made, or whether any global data has been changed,
3205 * but ddf is sufficiently weird that it probably always
3206 * changes global data ....
3207 */
18a2f463
NB
3208 struct ddf_super *ddf = st->sb;
3209 if (!ddf->updates_pending)
3210 return;
3211 ddf->updates_pending = 0;
2e735d19 3212 __write_init_super_ddf(st, 0);
2c514b71 3213 dprintf("ddf: sync_metadata\n");
549e9569
NB
3214}
3215
88c164f4
NB
3216static void ddf_process_update(struct supertype *st,
3217 struct metadata_update *update)
3218{
3219 /* Apply this update to the metadata.
3220 * The first 4 bytes are a DDF_*_MAGIC which guides
3221 * our actions.
3222 * Possible update are:
3223 * DDF_PHYS_RECORDS_MAGIC
3224 * Add a new physical device. Changes to this record
3225 * only happen implicitly.
3226 * used_pdes is the device number.
3227 * DDF_VIRT_RECORDS_MAGIC
3228 * Add a new VD. Possibly also change the 'access' bits.
3229 * populated_vdes is the entry number.
3230 * DDF_VD_CONF_MAGIC
3231 * New or updated VD. the VIRT_RECORD must already
3232 * exist. For an update, phys_refnum and lba_offset
3233 * (at least) are updated, and the VD_CONF must
3234 * be written to precisely those devices listed with
3235 * a phys_refnum.
3236 * DDF_SPARE_ASSIGN_MAGIC
3237 * replacement Spare Assignment Record... but for which device?
3238 *
3239 * So, e.g.:
3240 * - to create a new array, we send a VIRT_RECORD and
3241 * a VD_CONF. Then assemble and start the array.
3242 * - to activate a spare we send a VD_CONF to add the phys_refnum
3243 * and offset. This will also mark the spare as active with
3244 * a spare-assignment record.
3245 */
3246 struct ddf_super *ddf = st->sb;
3247 __u32 *magic = (__u32*)update->buf;
3248 struct phys_disk *pd;
3249 struct virtual_disk *vd;
3250 struct vd_config *vc;
3251 struct vcl *vcl;
3252 struct dl *dl;
3253 int mppe;
3254 int ent;
3255
2c514b71 3256 dprintf("Process update %x\n", *magic);
7e1432fb 3257
88c164f4
NB
3258 switch (*magic) {
3259 case DDF_PHYS_RECORDS_MAGIC:
3260
3261 if (update->len != (sizeof(struct phys_disk) +
3262 sizeof(struct phys_disk_entry)))
3263 return;
3264 pd = (struct phys_disk*)update->buf;
3265
3266 ent = __be16_to_cpu(pd->used_pdes);
3267 if (ent >= __be16_to_cpu(ddf->phys->max_pdes))
3268 return;
3269 if (!all_ff(ddf->phys->entries[ent].guid))
3270 return;
3271 ddf->phys->entries[ent] = pd->entries[0];
3272 ddf->phys->used_pdes = __cpu_to_be16(1 +
3273 __be16_to_cpu(ddf->phys->used_pdes));
18a2f463 3274 ddf->updates_pending = 1;
2cc2983d
N
3275 if (ddf->add_list) {
3276 struct active_array *a;
3277 struct dl *al = ddf->add_list;
3278 ddf->add_list = al->next;
3279
3280 al->next = ddf->dlist;
3281 ddf->dlist = al;
3282
3283 /* As a device has been added, we should check
3284 * for any degraded devices that might make
3285 * use of this spare */
3286 for (a = st->arrays ; a; a=a->next)
3287 a->check_degraded = 1;
3288 }
88c164f4
NB
3289 break;
3290
3291 case DDF_VIRT_RECORDS_MAGIC:
3292
3293 if (update->len != (sizeof(struct virtual_disk) +
3294 sizeof(struct virtual_entry)))
3295 return;
3296 vd = (struct virtual_disk*)update->buf;
3297
3298 ent = __be16_to_cpu(vd->populated_vdes);
3299 if (ent >= __be16_to_cpu(ddf->virt->max_vdes))
3300 return;
3301 if (!all_ff(ddf->virt->entries[ent].guid))
3302 return;
3303 ddf->virt->entries[ent] = vd->entries[0];
3304 ddf->virt->populated_vdes = __cpu_to_be16(1 +
3305 __be16_to_cpu(ddf->virt->populated_vdes));
18a2f463 3306 ddf->updates_pending = 1;
88c164f4
NB
3307 break;
3308
3309 case DDF_VD_CONF_MAGIC:
2c514b71 3310 dprintf("len %d %d\n", update->len, ddf->conf_rec_len);
88c164f4
NB
3311
3312 mppe = __be16_to_cpu(ddf->anchor.max_primary_element_entries);
edd8d13c 3313 if (update->len != ddf->conf_rec_len * 512)
88c164f4
NB
3314 return;
3315 vc = (struct vd_config*)update->buf;
3316 for (vcl = ddf->conflist; vcl ; vcl = vcl->next)
3317 if (memcmp(vcl->conf.guid, vc->guid, DDF_GUID_LEN) == 0)
3318 break;
2c514b71 3319 dprintf("vcl = %p\n", vcl);
88c164f4
NB
3320 if (vcl) {
3321 /* An update, just copy the phys_refnum and lba_offset
3322 * fields
3323 */
3324 memcpy(vcl->conf.phys_refnum, vc->phys_refnum,
3325 mppe * (sizeof(__u32) + sizeof(__u64)));
3326 } else {
3327 /* A new VD_CONF */
e6b9548d
DW
3328 if (!update->space)
3329 return;
88c164f4
NB
3330 vcl = update->space;
3331 update->space = NULL;
3332 vcl->next = ddf->conflist;
edd8d13c 3333 memcpy(&vcl->conf, vc, update->len);
88c164f4
NB
3334 vcl->lba_offset = (__u64*)
3335 &vcl->conf.phys_refnum[mppe];
3336 ddf->conflist = vcl;
3337 }
3338 /* Now make sure vlist is correct for each dl. */
3339 for (dl = ddf->dlist; dl; dl = dl->next) {
3340 int dn;
3341 int vn = 0;
3342 for (vcl = ddf->conflist; vcl ; vcl = vcl->next)
3343 for (dn=0; dn < ddf->mppe ; dn++)
3344 if (vcl->conf.phys_refnum[dn] ==
3345 dl->disk.refnum) {
2c514b71
NB
3346 dprintf("dev %d has %p at %d\n",
3347 dl->pdnum, vcl, vn);
88c164f4
NB
3348 dl->vlist[vn++] = vcl;
3349 break;
3350 }
3351 while (vn < ddf->max_part)
3352 dl->vlist[vn++] = NULL;
7e1432fb
NB
3353 if (dl->vlist[0]) {
3354 ddf->phys->entries[dl->pdnum].type &=
3355 ~__cpu_to_be16(DDF_Global_Spare);
3356 ddf->phys->entries[dl->pdnum].type |=
3357 __cpu_to_be16(DDF_Active_in_VD);
3358 }
3359 if (dl->spare) {
3360 ddf->phys->entries[dl->pdnum].type &=
3361 ~__cpu_to_be16(DDF_Global_Spare);
3362 ddf->phys->entries[dl->pdnum].type |=
3363 __cpu_to_be16(DDF_Spare);
3364 }
3365 if (!dl->vlist[0] && !dl->spare) {
3366 ddf->phys->entries[dl->pdnum].type |=
3367 __cpu_to_be16(DDF_Global_Spare);
3368 ddf->phys->entries[dl->pdnum].type &=
3369 ~__cpu_to_be16(DDF_Spare |
3370 DDF_Active_in_VD);
3371 }
88c164f4 3372 }
18a2f463 3373 ddf->updates_pending = 1;
88c164f4
NB
3374 break;
3375 case DDF_SPARE_ASSIGN_MAGIC:
3376 default: break;
3377 }
3378}
3379
edd8d13c
NB
3380static void ddf_prepare_update(struct supertype *st,
3381 struct metadata_update *update)
3382{
3383 /* This update arrived at managemon.
3384 * We are about to pass it to monitor.
3385 * If a malloc is needed, do it here.
3386 */
3387 struct ddf_super *ddf = st->sb;
3388 __u32 *magic = (__u32*)update->buf;
3389 if (*magic == DDF_VD_CONF_MAGIC)
e6b9548d 3390 if (posix_memalign(&update->space, 512,
6416d527 3391 offsetof(struct vcl, conf)
e6b9548d
DW
3392 + ddf->conf_rec_len * 512) != 0)
3393 update->space = NULL;
edd8d13c
NB
3394}
3395
7e1432fb
NB
3396/*
3397 * Check if the array 'a' is degraded but not failed.
3398 * If it is, find as many spares as are available and needed and
3399 * arrange for their inclusion.
3400 * We only choose devices which are not already in the array,
3401 * and prefer those with a spare-assignment to this array.
3402 * otherwise we choose global spares - assuming always that
3403 * there is enough room.
3404 * For each spare that we assign, we return an 'mdinfo' which
3405 * describes the position for the device in the array.
3406 * We also add to 'updates' a DDF_VD_CONF_MAGIC update with
3407 * the new phys_refnum and lba_offset values.
3408 *
3409 * Only worry about BVDs at the moment.
3410 */
3411static struct mdinfo *ddf_activate_spare(struct active_array *a,
3412 struct metadata_update **updates)
3413{
3414 int working = 0;
3415 struct mdinfo *d;
3416 struct ddf_super *ddf = a->container->sb;
3417 int global_ok = 0;
3418 struct mdinfo *rv = NULL;
3419 struct mdinfo *di;
3420 struct metadata_update *mu;
3421 struct dl *dl;
3422 int i;
3423 struct vd_config *vc;
3424 __u64 *lba;
3425
7e1432fb
NB
3426 for (d = a->info.devs ; d ; d = d->next) {
3427 if ((d->curr_state & DS_FAULTY) &&
3428 d->state_fd >= 0)
3429 /* wait for Removal to happen */
3430 return NULL;
3431 if (d->state_fd >= 0)
3432 working ++;
3433 }
3434
2c514b71
NB
3435 dprintf("ddf_activate: working=%d (%d) level=%d\n", working, a->info.array.raid_disks,
3436 a->info.array.level);
7e1432fb
NB
3437 if (working == a->info.array.raid_disks)
3438 return NULL; /* array not degraded */
3439 switch (a->info.array.level) {
3440 case 1:
3441 if (working == 0)
3442 return NULL; /* failed */
3443 break;
3444 case 4:
3445 case 5:
3446 if (working < a->info.array.raid_disks - 1)
3447 return NULL; /* failed */
3448 break;
3449 case 6:
3450 if (working < a->info.array.raid_disks - 2)
3451 return NULL; /* failed */
3452 break;
3453 default: /* concat or stripe */
3454 return NULL; /* failed */
3455 }
3456
3457 /* For each slot, if it is not working, find a spare */
3458 dl = ddf->dlist;
3459 for (i = 0; i < a->info.array.raid_disks; i++) {
3460 for (d = a->info.devs ; d ; d = d->next)
3461 if (d->disk.raid_disk == i)
3462 break;
2c514b71 3463 dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0);
7e1432fb
NB
3464 if (d && (d->state_fd >= 0))
3465 continue;
3466
3467 /* OK, this device needs recovery. Find a spare */
3468 again:
3469 for ( ; dl ; dl = dl->next) {
3470 unsigned long long esize;
3471 unsigned long long pos;
3472 struct mdinfo *d2;
3473 int is_global = 0;
3474 int is_dedicated = 0;
3475 struct extent *ex;
3476 int j;
3477 /* If in this array, skip */
3478 for (d2 = a->info.devs ; d2 ; d2 = d2->next)
3479 if (d2->disk.major == dl->major &&
3480 d2->disk.minor == dl->minor) {
2c514b71 3481 dprintf("%x:%x already in array\n", dl->major, dl->minor);
7e1432fb
NB
3482 break;
3483 }
3484 if (d2)
3485 continue;
3486 if (ddf->phys->entries[dl->pdnum].type &
3487 __cpu_to_be16(DDF_Spare)) {
3488 /* Check spare assign record */
3489 if (dl->spare) {
3490 if (dl->spare->type & DDF_spare_dedicated) {
3491 /* check spare_ents for guid */
3492 for (j = 0 ;
3493 j < __be16_to_cpu(dl->spare->populated);
3494 j++) {
3495 if (memcmp(dl->spare->spare_ents[j].guid,
3496 ddf->virt->entries[a->info.container_member].guid,
3497 DDF_GUID_LEN) == 0)
3498 is_dedicated = 1;
3499 }
3500 } else
3501 is_global = 1;
3502 }
3503 } else if (ddf->phys->entries[dl->pdnum].type &
3504 __cpu_to_be16(DDF_Global_Spare)) {
3505 is_global = 1;
3506 }
3507 if ( ! (is_dedicated ||
3508 (is_global && global_ok))) {
2c514b71 3509 dprintf("%x:%x not suitable: %d %d\n", dl->major, dl->minor,
7e1432fb
NB
3510 is_dedicated, is_global);
3511 continue;
3512 }
3513
3514 /* We are allowed to use this device - is there space?
3515 * We need a->info.component_size sectors */
3516 ex = get_extents(ddf, dl);
3517 if (!ex) {
2c514b71 3518 dprintf("cannot get extents\n");
7e1432fb
NB
3519 continue;
3520 }
3521 j = 0; pos = 0;
3522 esize = 0;
3523
3524 do {
3525 esize = ex[j].start - pos;
3526 if (esize >= a->info.component_size)
3527 break;
3528 pos = ex[i].start + ex[i].size;
3529 i++;
3530 } while (ex[i-1].size);
3531
3532 free(ex);
3533 if (esize < a->info.component_size) {
2c514b71
NB
3534 dprintf("%x:%x has no room: %llu %llu\n", dl->major, dl->minor,
3535 esize, a->info.component_size);
7e1432fb
NB
3536 /* No room */
3537 continue;
3538 }
3539
3540 /* Cool, we have a device with some space at pos */
3541 di = malloc(sizeof(*di));
79244939
DW
3542 if (!di)
3543 continue;
7e1432fb
NB
3544 memset(di, 0, sizeof(*di));
3545 di->disk.number = i;
3546 di->disk.raid_disk = i;
3547 di->disk.major = dl->major;
3548 di->disk.minor = dl->minor;
3549 di->disk.state = 0;
3550 di->data_offset = pos;
3551 di->component_size = a->info.component_size;
3552 di->container_member = dl->pdnum;
3553 di->next = rv;
3554 rv = di;
2c514b71
NB
3555 dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor,
3556 i, pos);
7e1432fb
NB
3557
3558 break;
3559 }
3560 if (!dl && ! global_ok) {
3561 /* not enough dedicated spares, try global */
3562 global_ok = 1;
3563 dl = ddf->dlist;
3564 goto again;
3565 }
3566 }
3567
3568 if (!rv)
3569 /* No spares found */
3570 return rv;
3571 /* Now 'rv' has a list of devices to return.
3572 * Create a metadata_update record to update the
3573 * phys_refnum and lba_offset values
3574 */
904c1ef7 3575 mu = malloc(sizeof(*mu));
79244939
DW
3576 if (mu && posix_memalign(&mu->space, 512, sizeof(struct vcl)) != 0) {
3577 free(mu);
3578 mu = NULL;
3579 }
3580 if (!mu) {
3581 while (rv) {
3582 struct mdinfo *n = rv->next;
3583
3584 free(rv);
3585 rv = n;
3586 }
3587 return NULL;
3588 }
3589
904c1ef7 3590 mu->buf = malloc(ddf->conf_rec_len * 512);
7e1432fb
NB
3591 mu->len = ddf->conf_rec_len;
3592 mu->next = *updates;
3593 vc = find_vdcr(ddf, a->info.container_member);
3594 memcpy(mu->buf, vc, ddf->conf_rec_len * 512);
3595
3596 vc = (struct vd_config*)mu->buf;
3597 lba = (__u64*)&vc->phys_refnum[ddf->mppe];
3598 for (di = rv ; di ; di = di->next) {
3599 vc->phys_refnum[di->disk.raid_disk] =
3600 ddf->phys->entries[dl->pdnum].refnum;
3601 lba[di->disk.raid_disk] = di->data_offset;
3602 }
3603 *updates = mu;
3604 return rv;
3605}
0e600426 3606#endif /* MDASSEMBLE */
7e1432fb 3607
b640a252
N
3608static int ddf_level_to_layout(int level)
3609{
3610 switch(level) {
3611 case 0:
3612 case 1:
3613 return 0;
3614 case 5:
3615 return ALGORITHM_LEFT_SYMMETRIC;
3616 case 6:
3617 return ALGORITHM_ROTATING_N_CONTINUE;
3618 case 10:
3619 return 0x102;
3620 default:
3621 return UnSet;
3622 }
3623}
3624
a322f70c
DW
3625struct superswitch super_ddf = {
3626#ifndef MDASSEMBLE
3627 .examine_super = examine_super_ddf,
3628 .brief_examine_super = brief_examine_super_ddf,
4737ae25 3629 .brief_examine_subarrays = brief_examine_subarrays_ddf,
bceedeec 3630 .export_examine_super = export_examine_super_ddf,
a322f70c
DW
3631 .detail_super = detail_super_ddf,
3632 .brief_detail_super = brief_detail_super_ddf,
3633 .validate_geometry = validate_geometry_ddf,
78e44928 3634 .write_init_super = write_init_super_ddf,
0e600426 3635 .add_to_super = add_to_super_ddf,
a322f70c
DW
3636#endif
3637 .match_home = match_home_ddf,
3638 .uuid_from_super= uuid_from_super_ddf,
3639 .getinfo_super = getinfo_super_ddf,
3640 .update_super = update_super_ddf,
3641
3642 .avail_size = avail_size_ddf,
3643
a19c88b8
NB
3644 .compare_super = compare_super_ddf,
3645
a322f70c 3646 .load_super = load_super_ddf,
ba7eb04f 3647 .init_super = init_super_ddf,
955e9ea1 3648 .store_super = store_super_ddf,
a322f70c
DW
3649 .free_super = free_super_ddf,
3650 .match_metadata_desc = match_metadata_desc_ddf,
78e44928 3651 .container_content = container_content_ddf,
b640a252 3652 .default_layout = ddf_level_to_layout,
a322f70c 3653
a322f70c 3654 .external = 1,
549e9569 3655
0e600426 3656#ifndef MDASSEMBLE
549e9569
NB
3657/* for mdmon */
3658 .open_new = ddf_open_new,
ed9d66aa 3659 .set_array_state= ddf_set_array_state,
549e9569
NB
3660 .set_disk = ddf_set_disk,
3661 .sync_metadata = ddf_sync_metadata,
88c164f4 3662 .process_update = ddf_process_update,
edd8d13c 3663 .prepare_update = ddf_prepare_update,
7e1432fb 3664 .activate_spare = ddf_activate_spare,
0e600426 3665#endif
4cce4069 3666 .name = "ddf",
a322f70c 3667};