]> git.ipfire.org Git - thirdparty/mdadm.git/blame - super-ddf.c
projects / thirdparty / mdadm.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
IMSM/DDF: don't recognised these metadata on partitions.
[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
691c6ee1
N		 786 if (test_partition(fd))
787 /* DDF is not allowed on partitions */
788 return 1;
789
a322f70c
DW		 790 /* 32M is a lower bound */
791 if (dsize <= 32*1024*1024) {
97320d7c 792 if (devname)
a322f70c
DW		 793 fprintf(stderr,
794 Name ": %s is too small for ddf: "
795 "size is %llu sectors.\n",
796 devname, dsize>>9);
97320d7c 797 return 1;
a322f70c
DW		 798 }
799 if (dsize & 511) {
97320d7c 800 if (devname)
a322f70c
DW		 801 fprintf(stderr,
802 Name ": %s is an odd size for ddf: "
803 "size is %llu bytes.\n",
804 devname, dsize);
97320d7c 805 return 1;
a322f70c
DW		 806 }
807
37424f13
DW		 808 free_super_ddf(st);
809
6416d527 810 if (posix_memalign((void**)&super, 512, sizeof(*super))!= 0) {
a322f70c
DW		 811 fprintf(stderr, Name ": malloc of %zu failed.\n",
812 sizeof(*super));
813 return 1;
814 }
a2349791 815 memset(super, 0, sizeof(*super));
a322f70c
DW		 816
817 rv = load_ddf_headers(fd, super, devname);
818 if (rv) {
819 free(super);
820 return rv;
821 }
822
823 /* Have valid headers and have chosen the best. Let's read in the rest*/
824
825 rv = load_ddf_global(fd, super, devname);
826
827 if (rv) {
828 if (devname)
829 fprintf(stderr,
830 Name ": Failed to load all information "
831 "sections on %s\n", devname);
832 free(super);
833 return rv;
834 }
835
3d2c4fc7
DW		 836 rv = load_ddf_local(fd, super, devname, 0);
837
838 if (rv) {
839 if (devname)
840 fprintf(stderr,
841 Name ": Failed to load all information "
842 "sections on %s\n", devname);
843 free(super);
844 return rv;
845 }
a322f70c 846
af99d9ca
DW		 847 if (st->subarray[0]) {
848 struct vcl *v;
849
850 for (v = super->conflist; v; v = v->next)
851 if (v->vcnum == atoi(st->subarray))
852 super->currentconf = v;
853 if (!super->currentconf) {
854 free(super);
855 return 1;
856 }
857 }
858
a322f70c
DW		 859 /* Should possibly check the sections .... */
860
861 st->sb = super;
862 if (st->ss == NULL) {
863 st->ss = &super_ddf;
864 st->minor_version = 0;
865 st->max_devs = 512;
866 }
352452c3 867 st->loaded_container = 0;
a322f70c
DW		 868 return 0;
869
870}
871
872static void free_super_ddf(struct supertype *st)
873{
874 struct ddf_super *ddf = st->sb;
875 if (ddf == NULL)
876 return;
877 free(ddf->phys);
878 free(ddf->virt);
879 while (ddf->conflist) {
880 struct vcl *v = ddf->conflist;
881 ddf->conflist = v->next;
59e36268
NB		 882 if (v->block_sizes)
883 free(v->block_sizes);
a322f70c
DW		 884 free(v);
885 }
886 while (ddf->dlist) {
887 struct dl *d = ddf->dlist;
888 ddf->dlist = d->next;
889 if (d->fd >= 0)
890 close(d->fd);
b2280677
NB		 891 if (d->spare)
892 free(d->spare);
a322f70c
DW		 893 free(d);
894 }
895 free(ddf);
896 st->sb = NULL;
897}
898
899static struct supertype *match_metadata_desc_ddf(char *arg)
900{
901 /* 'ddf' only support containers */
902 struct supertype *st;
903 if (strcmp(arg, "ddf") != 0 &&
904 strcmp(arg, "default") != 0
905 )
906 return NULL;
907
908 st = malloc(sizeof(*st));
ef609477 909 memset(st, 0, sizeof(*st));
a322f70c
DW		 910 st->ss = &super_ddf;
911 st->max_devs = 512;
912 st->minor_version = 0;
913 st->sb = NULL;
914 return st;
915}
916
a322f70c
DW		 917
918#ifndef MDASSEMBLE
919
920static mapping_t ddf_state[] = {
921 { "Optimal", 0},
922 { "Degraded", 1},
923 { "Deleted", 2},
924 { "Missing", 3},
925 { "Failed", 4},
926 { "Partially Optimal", 5},
927 { "-reserved-", 6},
928 { "-reserved-", 7},
929 { NULL, 0}
930};
931
932static mapping_t ddf_init_state[] = {
933 { "Not Initialised", 0},
934 { "QuickInit in Progress", 1},
935 { "Fully Initialised", 2},
936 { "*UNKNOWN*", 3},
937 { NULL, 0}
938};
939static mapping_t ddf_access[] = {
940 { "Read/Write", 0},
941 { "Reserved", 1},
942 { "Read Only", 2},
943 { "Blocked (no access)", 3},
944 { NULL ,0}
945};
946
947static mapping_t ddf_level[] = {
948 { "RAID0", DDF_RAID0},
949 { "RAID1", DDF_RAID1},
950 { "RAID3", DDF_RAID3},
951 { "RAID4", DDF_RAID4},
952 { "RAID5", DDF_RAID5},
953 { "RAID1E",DDF_RAID1E},
954 { "JBOD", DDF_JBOD},
955 { "CONCAT",DDF_CONCAT},
956 { "RAID5E",DDF_RAID5E},
957 { "RAID5EE",DDF_RAID5EE},
958 { "RAID6", DDF_RAID6},
959 { NULL, 0}
960};
961static mapping_t ddf_sec_level[] = {
962 { "Striped", DDF_2STRIPED},
963 { "Mirrored", DDF_2MIRRORED},
964 { "Concat", DDF_2CONCAT},
965 { "Spanned", DDF_2SPANNED},
966 { NULL, 0}
967};
968#endif
969
970struct num_mapping {
971 int num1, num2;
972};
973static struct num_mapping ddf_level_num[] = {
974 { DDF_RAID0, 0 },
975 { DDF_RAID1, 1 },
976 { DDF_RAID3, LEVEL_UNSUPPORTED },
60f18132
NB		 977 { DDF_RAID4, 4 },
978 { DDF_RAID5, 5 },
a322f70c
DW		 979 { DDF_RAID1E, LEVEL_UNSUPPORTED },
980 { DDF_JBOD, LEVEL_UNSUPPORTED },
981 { DDF_CONCAT, LEVEL_LINEAR },
982 { DDF_RAID5E, LEVEL_UNSUPPORTED },
983 { DDF_RAID5EE, LEVEL_UNSUPPORTED },
984 { DDF_RAID6, 6},
985 { MAXINT, MAXINT }
986};
987
988static int map_num1(struct num_mapping *map, int num)
989{
990 int i;
991 for (i=0 ; map[i].num1 != MAXINT; i++)
992 if (map[i].num1 == num)
993 break;
994 return map[i].num2;
995}
996
42dc2744
N		 997static int all_ff(char *guid)
998{
999 int i;
1000 for (i = 0; i < DDF_GUID_LEN; i++)
1001 if (guid[i] != (char)0xff)
1002 return 0;
1003 return 1;
1004}
1005
a322f70c
DW		 1006#ifndef MDASSEMBLE
1007static void print_guid(char *guid, int tstamp)
1008{
1009 /* A GUIDs are part (or all) ASCII and part binary.
1010 * They tend to be space padded.
59e36268
NB		 1011 * We print the GUID in HEX, then in parentheses add
1012 * any initial ASCII sequence, and a possible
1013 * time stamp from bytes 16-19
a322f70c
DW		 1014 */
1015 int l = DDF_GUID_LEN;
1016 int i;
59e36268
NB		 1017
1018 for (i=0 ; i<DDF_GUID_LEN ; i++) {
1019 if ((i&3)==0 && i != 0) printf(":");
1020 printf("%02X", guid[i]&255);
1021 }
1022
cfccea8c 1023 printf("\n (");
a322f70c
DW		 1024 while (l && guid[l-1] == ' ')
1025 l--;
1026 for (i=0 ; i<l ; i++) {
1027 if (guid[i] >= 0x20 && guid[i] < 0x7f)
1028 fputc(guid[i], stdout);
1029 else
59e36268 1030 break;
a322f70c
DW		 1031 }
1032 if (tstamp) {
1033 time_t then = __be32_to_cpu(*(__u32*)(guid+16)) + DECADE;
1034 char tbuf[100];
1035 struct tm *tm;
1036 tm = localtime(&then);
59e36268 1037 strftime(tbuf, 100, " %D %T",tm);
a322f70c
DW		 1038 fputs(tbuf, stdout);
1039 }
59e36268 1040 printf(")");
a322f70c
DW		 1041}
1042
1043static void examine_vd(int n, struct ddf_super *sb, char *guid)
1044{
8c3b8c2c 1045 int crl = sb->conf_rec_len;
a322f70c
DW		 1046 struct vcl *vcl;
1047
1048 for (vcl = sb->conflist ; vcl ; vcl = vcl->next) {
b06e3095 1049 int i;
a322f70c
DW		 1050 struct vd_config *vc = &vcl->conf;
1051
1052 if (calc_crc(vc, crl*512) != vc->crc)
1053 continue;
1054 if (memcmp(vc->guid, guid, DDF_GUID_LEN) != 0)
1055 continue;
1056
1057 /* Ok, we know about this VD, let's give more details */
b06e3095 1058 printf(" Raid Devices[%d] : %d (", n,
a322f70c 1059 __be16_to_cpu(vc->prim_elmnt_count));
b06e3095
N		 1060 for (i=0; i<__be16_to_cpu(vc->prim_elmnt_count); i++) {
1061 int j;
1062 int cnt = __be16_to_cpu(sb->phys->used_pdes);
1063 for (j=0; j<cnt; j++)
1064 if (vc->phys_refnum[i] == sb->phys->entries[j].refnum)
1065 break;
1066 if (i) printf(" ");
1067 if (j < cnt)
1068 printf("%d", j);
1069 else
1070 printf("--");
1071 }
1072 printf(")\n");
1073 if (vc->chunk_shift != 255)
a322f70c
DW		 1074 printf(" Chunk Size[%d] : %d sectors\n", n,
1075 1 << vc->chunk_shift);
1076 printf(" Raid Level[%d] : %s\n", n,
1077 map_num(ddf_level, vc->prl)?:"-unknown-");
1078 if (vc->sec_elmnt_count != 1) {
1079 printf(" Secondary Position[%d] : %d of %d\n", n,
1080 vc->sec_elmnt_seq, vc->sec_elmnt_count);
1081 printf(" Secondary Level[%d] : %s\n", n,
1082 map_num(ddf_sec_level, vc->srl) ?: "-unknown-");
1083 }
1084 printf(" Device Size[%d] : %llu\n", n,
c9b6907b 1085 (unsigned long long)__be64_to_cpu(vc->blocks)/2);
a322f70c 1086 printf(" Array Size[%d] : %llu\n", n,
c9b6907b 1087 (unsigned long long)__be64_to_cpu(vc->array_blocks)/2);
a322f70c
DW		 1088 }
1089}
1090
1091static void examine_vds(struct ddf_super *sb)
1092{
1093 int cnt = __be16_to_cpu(sb->virt->populated_vdes);
1094 int i;
1095 printf(" Virtual Disks : %d\n", cnt);
1096
1097 for (i=0; i<cnt; i++) {
1098 struct virtual_entry *ve = &sb->virt->entries[i];
b06e3095 1099 printf("\n");
a322f70c
DW		 1100 printf(" VD GUID[%d] : ", i); print_guid(ve->guid, 1);
1101 printf("\n");
1102 printf(" unit[%d] : %d\n", i, __be16_to_cpu(ve->unit));
1103 printf(" state[%d] : %s, %s%s\n", i,
1104 map_num(ddf_state, ve->state & 7),
1105 (ve->state & 8) ? "Morphing, ": "",
1106 (ve->state & 16)? "Not Consistent" : "Consistent");
1107 printf(" init state[%d] : %s\n", i,
1108 map_num(ddf_init_state, ve->init_state&3));
1109 printf(" access[%d] : %s\n", i,
1110 map_num(ddf_access, (ve->init_state>>6) & 3));
1111 printf(" Name[%d] : %.16s\n", i, ve->name);
1112 examine_vd(i, sb, ve->guid);
1113 }
1114 if (cnt) printf("\n");
1115}
1116
1117static void examine_pds(struct ddf_super *sb)
1118{
1119 int cnt = __be16_to_cpu(sb->phys->used_pdes);
1120 int i;
1121 struct dl *dl;
1122 printf(" Physical Disks : %d\n", cnt);
962371a5 1123 printf(" Number RefNo Size Device Type/State\n");
a322f70c
DW		 1124
1125 for (i=0 ; i<cnt ; i++) {
1126 struct phys_disk_entry *pd = &sb->phys->entries[i];
1127 int type = __be16_to_cpu(pd->type);
1128 int state = __be16_to_cpu(pd->state);
1129
b06e3095
N		 1130 //printf(" PD GUID[%d] : ", i); print_guid(pd->guid, 0);
1131 //printf("\n");
1132 printf(" %3d %08x ", i,
a322f70c 1133 __be32_to_cpu(pd->refnum));
c9b6907b
DL		 1134 printf("%8lluK ",
1135 (unsigned long long)__be64_to_cpu(pd->config_size)>>1);
b06e3095
N		 1136 for (dl = sb->dlist; dl ; dl = dl->next) {
1137 if (dl->disk.refnum == pd->refnum) {
1138 char *dv = map_dev(dl->major, dl->minor, 0);
1139 if (dv) {
962371a5 1140 printf("%-15s", dv);
b06e3095
N		 1141 break;
1142 }
1143 }
1144 }
1145 if (!dl)
962371a5 1146 printf("%15s","");
b06e3095 1147 printf(" %s%s%s%s%s",
a322f70c 1148 (type&2) ? "active":"",
b06e3095 1149 (type&4) ? "Global-Spare":"",
a322f70c
DW		 1150 (type&8) ? "spare" : "",
1151 (type&16)? ", foreign" : "",
1152 (type&32)? "pass-through" : "");
b06e3095 1153 printf("/%s%s%s%s%s%s%s",
a322f70c
DW		 1154 (state&1)? "Online": "Offline",
1155 (state&2)? ", Failed": "",
1156 (state&4)? ", Rebuilding": "",
1157 (state&8)? ", in-transition": "",
b06e3095
N		 1158 (state&16)? ", SMART-errors": "",
1159 (state&32)? ", Unrecovered-Read-Errors": "",
a322f70c 1160 (state&64)? ", Missing" : "");
a322f70c
DW		 1161 printf("\n");
1162 }
1163}
1164
1165static void examine_super_ddf(struct supertype *st, char *homehost)
1166{
1167 struct ddf_super *sb = st->sb;
1168
1169 printf(" Magic : %08x\n", __be32_to_cpu(sb->anchor.magic));
1170 printf(" Version : %.8s\n", sb->anchor.revision);
598f0d58
NB		 1171 printf("Controller GUID : "); print_guid(sb->controller.guid, 0);
1172 printf("\n");
1173 printf(" Container GUID : "); print_guid(sb->anchor.guid, 1);
a322f70c
DW		 1174 printf("\n");
1175 printf(" Seq : %08x\n", __be32_to_cpu(sb->active->seq));
1176 printf(" Redundant hdr : %s\n", sb->secondary.magic == DDF_HEADER_MAGIC
1177 ?"yes" : "no");
1178 examine_vds(sb);
1179 examine_pds(sb);
1180}
1181
ff54de6e
N		 1182static void getinfo_super_ddf(struct supertype *st, struct mdinfo *info);
1183
42dc2744 1184static void uuid_from_super_ddf(struct supertype *st, int uuid[4]);
ff54de6e 1185
061f2c6a 1186static void brief_examine_super_ddf(struct supertype *st, int verbose)
4737ae25
N		 1187{
1188 /* We just write a generic DDF ARRAY entry
1189 */
1190 struct mdinfo info;
1191 char nbuf[64];
1192 getinfo_super_ddf(st, &info);
1193 fname_from_uuid(st, &info, nbuf, ':');
1194
1195 printf("ARRAY metadata=ddf UUID=%s\n", nbuf + 5);
1196}
1197
1198static void brief_examine_subarrays_ddf(struct supertype *st, int verbose)
a322f70c
DW		 1199{
1200 /* We just write a generic DDF ARRAY entry
a322f70c 1201 */
42dc2744 1202 struct ddf_super *ddf = st->sb;
ff54de6e 1203 struct mdinfo info;
42dc2744 1204 int i;
ff54de6e
N		 1205 char nbuf[64];
1206 getinfo_super_ddf(st, &info);
1207 fname_from_uuid(st, &info, nbuf, ':');
42dc2744
N		 1208
1209 for (i=0; i<__be16_to_cpu(ddf->virt->max_vdes); i++) {
1210 struct virtual_entry *ve = &ddf->virt->entries[i];
1211 struct vcl vcl;
1212 char nbuf1[64];
1213 if (all_ff(ve->guid))
1214 continue;
1215 memcpy(vcl.conf.guid, ve->guid, DDF_GUID_LEN);
1216 ddf->currentconf =&vcl;
1217 uuid_from_super_ddf(st, info.uuid);
1218 fname_from_uuid(st, &info, nbuf1, ':');
1219 printf("ARRAY container=%s member=%d UUID=%s\n",
1220 nbuf+5, i, nbuf1+5);
1221 }
a322f70c
DW		 1222}
1223
bceedeec
N		 1224static void export_examine_super_ddf(struct supertype *st)
1225{
1226 struct mdinfo info;
1227 char nbuf[64];
1228 getinfo_super_ddf(st, &info);
1229 fname_from_uuid(st, &info, nbuf, ':');
1230 printf("MD_METADATA=ddf\n");
1231 printf("MD_LEVEL=container\n");
1232 printf("MD_UUID=%s\n", nbuf+5);
1233}
1234
1235
a322f70c
DW		 1236static void detail_super_ddf(struct supertype *st, char *homehost)
1237{
1238 /* FIXME later
1239 * Could print DDF GUID
1240 * Need to find which array
1241 * If whole, briefly list all arrays
1242 * If one, give name
1243 */
1244}
1245
1246static void brief_detail_super_ddf(struct supertype *st)
1247{
1248 /* FIXME I really need to know which array we are detailing.
1249 * Can that be stored in ddf_super??
1250 */
1251// struct ddf_super *ddf = st->sb;
ff54de6e
N		 1252 struct mdinfo info;
1253 char nbuf[64];
1254 getinfo_super_ddf(st, &info);
1255 fname_from_uuid(st, &info, nbuf,':');
1256 printf(" UUID=%s", nbuf + 5);
a322f70c 1257}
a322f70c
DW		 1258#endif
1259
1260static int match_home_ddf(struct supertype *st, char *homehost)
1261{
1262 /* It matches 'this' host if the controller is a
1263 * Linux-MD controller with vendor_data matching
1264 * the hostname
1265 */
1266 struct ddf_super *ddf = st->sb;
d1d3482b
N		 1267 int len;
1268
1269 if (!homehost)
1270 return 0;
1271 len = strlen(homehost);
a322f70c
DW		 1272
1273 return (memcmp(ddf->controller.guid, T10, 8) == 0 &&
1274 len < sizeof(ddf->controller.vendor_data) &&
1275 memcmp(ddf->controller.vendor_data, homehost,len) == 0 &&
1276 ddf->controller.vendor_data[len] == 0);
1277}
1278
0e600426 1279#ifndef MDASSEMBLE
7a7cc504 1280static struct vd_config *find_vdcr(struct ddf_super *ddf, int inst)
a322f70c 1281{
7a7cc504 1282 struct vcl *v;
59e36268 1283
7a7cc504 1284 for (v = ddf->conflist; v; v = v->next)
59e36268 1285 if (inst == v->vcnum)
7a7cc504
NB		 1286 return &v->conf;
1287 return NULL;
1288}
0e600426 1289#endif
7a7cc504
NB		 1290
1291static int find_phys(struct ddf_super *ddf, __u32 phys_refnum)
1292{
1293 /* Find the entry in phys_disk which has the given refnum
1294 * and return it's index
1295 */
1296 int i;
1297 for (i=0; i < __be16_to_cpu(ddf->phys->max_pdes); i++)
1298 if (ddf->phys->entries[i].refnum == phys_refnum)
1299 return i;
1300 return -1;
a322f70c
DW		 1301}
1302
1303static void uuid_from_super_ddf(struct supertype *st, int uuid[4])
1304{
1305 /* The uuid returned here is used for:
1306 * uuid to put into bitmap file (Create, Grow)
1307 * uuid for backup header when saving critical section (Grow)
1308 * comparing uuids when re-adding a device into an array
51006d85
N		 1309 * In these cases the uuid required is that of the data-array,
1310 * not the device-set.
1311 * uuid to recognise same set when adding a missing device back
1312 * to an array. This is a uuid for the device-set.
1313 *
a322f70c
DW		 1314 * For each of these we can make do with a truncated
1315 * or hashed uuid rather than the original, as long as
1316 * everyone agrees.
a322f70c
DW		 1317 * In the case of SVD we assume the BVD is of interest,
1318 * though that might be the case if a bitmap were made for
1319 * a mirrored SVD - worry about that later.
1320 * So we need to find the VD configuration record for the
1321 * relevant BVD and extract the GUID and Secondary_Element_Seq.
1322 * The first 16 bytes of the sha1 of these is used.
1323 */
1324 struct ddf_super *ddf = st->sb;
d2ca6449 1325 struct vcl *vcl = ddf->currentconf;
c5afc314
N		 1326 char *guid;
1327 char buf[20];
1328 struct sha1_ctx ctx;
a322f70c 1329
c5afc314
N		 1330 if (vcl)
1331 guid = vcl->conf.guid;
1332 else
1333 guid = ddf->anchor.guid;
1334
1335 sha1_init_ctx(&ctx);
1336 sha1_process_bytes(guid, DDF_GUID_LEN, &ctx);
c5afc314
N		 1337 sha1_finish_ctx(&ctx, buf);
1338 memcpy(uuid, buf, 4*4);
a322f70c
DW		 1339}
1340
78e44928
NB		 1341static void getinfo_super_ddf_bvd(struct supertype *st, struct mdinfo *info);
1342
a322f70c
DW		 1343static void getinfo_super_ddf(struct supertype *st, struct mdinfo *info)
1344{
1345 struct ddf_super *ddf = st->sb;
1346
78e44928
NB		 1347 if (ddf->currentconf) {
1348 getinfo_super_ddf_bvd(st, info);
1349 return;
1350 }
1351
a322f70c
DW		 1352 info->array.raid_disks = __be16_to_cpu(ddf->phys->used_pdes);
1353 info->array.level = LEVEL_CONTAINER;
1354 info->array.layout = 0;
1355 info->array.md_minor = -1;
1356 info->array.ctime = DECADE + __be32_to_cpu(*(__u32*)
1357 (ddf->anchor.guid+16));
1358 info->array.utime = 0;
1359 info->array.chunk_size = 0;
1360
a322f70c
DW		 1361
1362 info->disk.major = 0;
1363 info->disk.minor = 0;
cba0191b
NB		 1364 if (ddf->dlist) {
1365 info->disk.number = __be32_to_cpu(ddf->dlist->disk.refnum);
59e36268 1366 info->disk.raid_disk = find_phys(ddf, ddf->dlist->disk.refnum);
d2ca6449
NB		 1367
1368 info->data_offset = __be64_to_cpu(ddf->phys->
1369 entries[info->disk.raid_disk].
1370 config_size);
1371 info->component_size = ddf->dlist->size - info->data_offset;
cba0191b
NB		 1372 } else {
1373 info->disk.number = -1;
661dce36 1374 info->disk.raid_disk = -1;
cba0191b
NB		 1375// info->disk.raid_disk = find refnum in the table and use index;
1376 }
f22385f9 1377 info->disk.state = (1 << MD_DISK_SYNC) | (1 << MD_DISK_ACTIVE);
a19c88b8 1378
d2ca6449 1379
921d9e16 1380 info->recovery_start = MaxSector;
a19c88b8 1381 info->reshape_active = 0;
c5afc314 1382 info->name[0] = 0;
a322f70c 1383
f35f2525
N		 1384 info->array.major_version = -1;
1385 info->array.minor_version = -2;
159c3a1a 1386 strcpy(info->text_version, "ddf");
a67dd8cc 1387 info->safe_mode_delay = 0;
159c3a1a 1388
c5afc314 1389 uuid_from_super_ddf(st, info->uuid);
a322f70c 1390
a322f70c
DW		 1391}
1392
598f0d58
NB		 1393static int rlq_to_layout(int rlq, int prl, int raiddisks);
1394
a322f70c
DW		 1395static void getinfo_super_ddf_bvd(struct supertype *st, struct mdinfo *info)
1396{
1397 struct ddf_super *ddf = st->sb;
d2ca6449
NB		 1398 struct vcl *vc = ddf->currentconf;
1399 int cd = ddf->currentdev;
db42fa9b 1400 int j;
8592f29d 1401 struct dl *dl;
a322f70c
DW		 1402
1403 /* FIXME this returns BVD info - what if we want SVD ?? */
1404
d2ca6449
NB		 1405 info->array.raid_disks = __be16_to_cpu(vc->conf.prim_elmnt_count);
1406 info->array.level = map_num1(ddf_level_num, vc->conf.prl);
1407 info->array.layout = rlq_to_layout(vc->conf.rlq, vc->conf.prl,
598f0d58 1408 info->array.raid_disks);
a322f70c 1409 info->array.md_minor = -1;
d2ca6449
NB		 1410 info->array.ctime = DECADE +
1411 __be32_to_cpu(*(__u32*)(vc->conf.guid+16));
1412 info->array.utime = DECADE + __be32_to_cpu(vc->conf.timestamp);
1413 info->array.chunk_size = 512 << vc->conf.chunk_shift;
da9b4a62 1414 info->custom_array_size = 0;
d2ca6449
NB		 1415
1416 if (cd >= 0 && cd < ddf->mppe) {
1417 info->data_offset = __be64_to_cpu(vc->lba_offset[cd]);
1418 if (vc->block_sizes)
1419 info->component_size = vc->block_sizes[cd];
1420 else
1421 info->component_size = __be64_to_cpu(vc->conf.blocks);
1422 }
a322f70c 1423
8592f29d
N		 1424 for (dl = ddf->dlist; dl ; dl = dl->next)
1425 if (dl->raiddisk == info->disk.raid_disk)
1426 break;
a322f70c
DW		 1427 info->disk.major = 0;
1428 info->disk.minor = 0;
8592f29d
N		 1429 if (dl) {
1430 info->disk.major = dl->major;
1431 info->disk.minor = dl->minor;
1432 }
a322f70c
DW		 1433// info->disk.number = __be32_to_cpu(ddf->disk.refnum);
1434// info->disk.raid_disk = find refnum in the table and use index;
1435// info->disk.state = ???;
1436
103f2410
NB		 1437 info->container_member = ddf->currentconf->vcnum;
1438
921d9e16 1439 info->recovery_start = MaxSector;
80d26cb2 1440 info->resync_start = 0;
624c5ad4 1441 info->reshape_active = 0;
80d26cb2
NB		 1442 if (!(ddf->virt->entries[info->container_member].state
1443 & DDF_state_inconsistent) &&
1444 (ddf->virt->entries[info->container_member].init_state
1445 & DDF_initstate_mask)
1446 == DDF_init_full)
b7528a20 1447 info->resync_start = MaxSector;
80d26cb2 1448
a322f70c
DW		 1449 uuid_from_super_ddf(st, info->uuid);
1450
3576e302 1451 info->container_member = atoi(st->subarray);
f35f2525
N		 1452 info->array.major_version = -1;
1453 info->array.minor_version = -2;
3576e302 1454 sprintf(info->text_version, "/%s/%s",
159c3a1a 1455 devnum2devname(st->container_dev),
3576e302 1456 st->subarray);
a67dd8cc 1457 info->safe_mode_delay = 200;
159c3a1a 1458
db42fa9b
N		 1459 memcpy(info->name, ddf->virt->entries[info->container_member].name, 16);
1460 info->name[16]=0;
1461 for(j=0; j<16; j++)
1462 if (info->name[j] == ' ')
1463 info->name[j] = 0;
a322f70c
DW		 1464}
1465
598f0d58 1466
a322f70c
DW		 1467static int update_super_ddf(struct supertype *st, struct mdinfo *info,
1468 char *update,
1469 char *devname, int verbose,
1470 int uuid_set, char *homehost)
1471{
1472 /* For 'assemble' and 'force' we need to return non-zero if any
1473 * change was made. For others, the return value is ignored.
1474 * Update options are:
1475 * force-one : This device looks a bit old but needs to be included,
1476 * update age info appropriately.
1477 * assemble: clear any 'faulty' flag to allow this device to
1478 * be assembled.
1479 * force-array: Array is degraded but being forced, mark it clean
1480 * if that will be needed to assemble it.
1481 *
1482 * newdev: not used ????
1483 * grow: Array has gained a new device - this is currently for
1484 * linear only
1485 * resync: mark as dirty so a resync will happen.
59e36268 1486 * uuid: Change the uuid of the array to match what is given
a322f70c
DW		 1487 * homehost: update the recorded homehost
1488 * name: update the name - preserving the homehost
1489 * _reshape_progress: record new reshape_progress position.
1490 *
1491 * Following are not relevant for this version:
1492 * sparc2.2 : update from old dodgey metadata
1493 * super-minor: change the preferred_minor number
1494 * summaries: update redundant counters.
1495 */
1496 int rv = 0;
1497// struct ddf_super *ddf = st->sb;
7a7cc504 1498// struct vd_config *vd = find_vdcr(ddf, info->container_member);
a322f70c
DW		 1499// struct virtual_entry *ve = find_ve(ddf);
1500
a322f70c
DW		 1501 /* we don't need to handle "force-*" or "assemble" as
1502 * there is no need to 'trick' the kernel. We the metadata is
1503 * first updated to activate the array, all the implied modifications
1504 * will just happen.
1505 */
1506
1507 if (strcmp(update, "grow") == 0) {
1508 /* FIXME */
1509 }
1510 if (strcmp(update, "resync") == 0) {
1511// info->resync_checkpoint = 0;
1512 }
1513 /* We ignore UUID updates as they make even less sense
1514 * with DDF
1515 */
1516 if (strcmp(update, "homehost") == 0) {
1517 /* homehost is stored in controller->vendor_data,
1518 * or it is when we are the vendor
1519 */
1520// if (info->vendor_is_local)
1521// strcpy(ddf->controller.vendor_data, homehost);
1522 }
1523 if (strcmp(update, "name") == 0) {
1524 /* name is stored in virtual_entry->name */
1525// memset(ve->name, ' ', 16);
1526// strncpy(ve->name, info->name, 16);
1527 }
1528 if (strcmp(update, "_reshape_progress") == 0) {
1529 /* We don't support reshape yet */
1530 }
1531
1532// update_all_csum(ddf);
1533
1534 return rv;
1535}
1536
5f8097be
NB		 1537static void make_header_guid(char *guid)
1538{
1539 __u32 stamp;
5f8097be
NB		 1540 /* Create a DDF Header of Virtual Disk GUID */
1541
1542 /* 24 bytes of fiction required.
1543 * first 8 are a 'vendor-id' - "Linux-MD"
1544 * next 8 are controller type.. how about 0X DEAD BEEF 0000 0000
1545 * Remaining 8 random number plus timestamp
1546 */
1547 memcpy(guid, T10, sizeof(T10));
1548 stamp = __cpu_to_be32(0xdeadbeef);
1549 memcpy(guid+8, &stamp, 4);
1550 stamp = __cpu_to_be32(0);
1551 memcpy(guid+12, &stamp, 4);
1552 stamp = __cpu_to_be32(time(0) - DECADE);
1553 memcpy(guid+16, &stamp, 4);
bfb7ea78 1554 stamp = random32();
5f8097be 1555 memcpy(guid+20, &stamp, 4);
5f8097be 1556}
59e36268 1557
78e44928
NB		 1558static int init_super_ddf_bvd(struct supertype *st,
1559 mdu_array_info_t *info,
1560 unsigned long long size,
1561 char *name, char *homehost,
1562 int *uuid);
1563
a322f70c
DW		 1564static int init_super_ddf(struct supertype *st,
1565 mdu_array_info_t *info,
1566 unsigned long long size, char *name, char *homehost,
1567 int *uuid)
1568{
1569 /* This is primarily called by Create when creating a new array.
1570 * We will then get add_to_super called for each component, and then
1571 * write_init_super called to write it out to each device.
1572 * For DDF, Create can create on fresh devices or on a pre-existing
1573 * array.
1574 * To create on a pre-existing array a different method will be called.
1575 * This one is just for fresh drives.
1576 *
1577 * We need to create the entire 'ddf' structure which includes:
1578 * DDF headers - these are easy.
1579 * Controller data - a Sector describing this controller .. not that
1580 * this is a controller exactly.
1581 * Physical Disk Record - one entry per device, so
1582 * leave plenty of space.
1583 * Virtual Disk Records - again, just leave plenty of space.
1584 * This just lists VDs, doesn't give details
1585 * Config records - describes the VDs that use this disk
1586 * DiskData - describes 'this' device.
1587 * BadBlockManagement - empty
1588 * Diag Space - empty
1589 * Vendor Logs - Could we put bitmaps here?
1590 *
1591 */
1592 struct ddf_super *ddf;
1593 char hostname[17];
1594 int hostlen;
a322f70c
DW		 1595 int max_phys_disks, max_virt_disks;
1596 unsigned long long sector;
1597 int clen;
1598 int i;
1599 int pdsize, vdsize;
1600 struct phys_disk *pd;
1601 struct virtual_disk *vd;
1602
78e44928 1603 if (st->sb)
955e9ea1 1604 return init_super_ddf_bvd(st, info, size, name, homehost, uuid);
ba7eb04f 1605
3d2c4fc7
DW		 1606 if (posix_memalign((void**)&ddf, 512, sizeof(*ddf)) != 0) {
1607 fprintf(stderr, Name ": %s could not allocate superblock\n", __func__);
1608 return 0;
1609 }
6264b437 1610 memset(ddf, 0, sizeof(*ddf));
a322f70c
DW		 1611 ddf->dlist = NULL; /* no physical disks yet */
1612 ddf->conflist = NULL; /* No virtual disks yet */
955e9ea1
DW		 1613 st->sb = ddf;
1614
1615 if (info == NULL) {
1616 /* zeroing superblock */
1617 return 0;
1618 }
a322f70c
DW		 1619
1620 /* At least 32MB *must* be reserved for the ddf. So let's just
1621 * start 32MB from the end, and put the primary header there.
1622 * Don't do secondary for now.
1623 * We don't know exactly where that will be yet as it could be
1624 * different on each device. To just set up the lengths.
1625 *
1626 */
1627
1628 ddf->anchor.magic = DDF_HEADER_MAGIC;
5f8097be 1629 make_header_guid(ddf->anchor.guid);
a322f70c 1630
59e36268 1631 memcpy(ddf->anchor.revision, DDF_REVISION_2, 8);
a322f70c
DW		 1632 ddf->anchor.seq = __cpu_to_be32(1);
1633 ddf->anchor.timestamp = __cpu_to_be32(time(0) - DECADE);
1634 ddf->anchor.openflag = 0xFF;
1635 ddf->anchor.foreignflag = 0;
1636 ddf->anchor.enforcegroups = 0; /* Is this best?? */
1637 ddf->anchor.pad0 = 0xff;
1638 memset(ddf->anchor.pad1, 0xff, 12);
1639 memset(ddf->anchor.header_ext, 0xff, 32);
1640 ddf->anchor.primary_lba = ~(__u64)0;
1641 ddf->anchor.secondary_lba = ~(__u64)0;
1642 ddf->anchor.type = DDF_HEADER_ANCHOR;
1643 memset(ddf->anchor.pad2, 0xff, 3);
1644 ddf->anchor.workspace_len = __cpu_to_be32(32768); /* Must be reserved */
1645 ddf->anchor.workspace_lba = ~(__u64)0; /* Put this at bottom
1646 of 32M reserved.. */
1647 max_phys_disks = 1023; /* Should be enough */
1648 ddf->anchor.max_pd_entries = __cpu_to_be16(max_phys_disks);
1649 max_virt_disks = 255;
1650 ddf->anchor.max_vd_entries = __cpu_to_be16(max_virt_disks); /* ?? */
1651 ddf->anchor.max_partitions = __cpu_to_be16(64); /* ?? */
1652 ddf->max_part = 64;
8c3b8c2c 1653 ddf->mppe = 256;
59e36268
NB		 1654 ddf->conf_rec_len = 1 + ROUND_UP(ddf->mppe * (4+8), 512)/512;
1655 ddf->anchor.config_record_len = __cpu_to_be16(ddf->conf_rec_len);
1656 ddf->anchor.max_primary_element_entries = __cpu_to_be16(ddf->mppe);
a322f70c 1657 memset(ddf->anchor.pad3, 0xff, 54);
a322f70c
DW		 1658 /* controller sections is one sector long immediately
1659 * after the ddf header */
1660 sector = 1;
1661 ddf->anchor.controller_section_offset = __cpu_to_be32(sector);
1662 ddf->anchor.controller_section_length = __cpu_to_be32(1);
1663 sector += 1;
1664
1665 /* phys is 8 sectors after that */
1666 pdsize = ROUND_UP(sizeof(struct phys_disk) +
1667 sizeof(struct phys_disk_entry)*max_phys_disks,
1668 512);
1669 switch(pdsize/512) {
1670 case 2: case 8: case 32: case 128: case 512: break;
1671 default: abort();
1672 }
1673 ddf->anchor.phys_section_offset = __cpu_to_be32(sector);
1674 ddf->anchor.phys_section_length =
1675 __cpu_to_be32(pdsize/512); /* max_primary_element_entries/8 */
1676 sector += pdsize/512;
1677
1678 /* virt is another 32 sectors */
1679 vdsize = ROUND_UP(sizeof(struct virtual_disk) +
1680 sizeof(struct virtual_entry) * max_virt_disks,
1681 512);
1682 switch(vdsize/512) {
1683 case 2: case 8: case 32: case 128: case 512: break;
1684 default: abort();
1685 }
1686 ddf->anchor.virt_section_offset = __cpu_to_be32(sector);
1687 ddf->anchor.virt_section_length =
1688 __cpu_to_be32(vdsize/512); /* max_vd_entries/8 */
1689 sector += vdsize/512;
1690
59e36268 1691 clen = ddf->conf_rec_len * (ddf->max_part+1);
a322f70c
DW		 1692 ddf->anchor.config_section_offset = __cpu_to_be32(sector);
1693 ddf->anchor.config_section_length = __cpu_to_be32(clen);
1694 sector += clen;
1695
1696 ddf->anchor.data_section_offset = __cpu_to_be32(sector);
1697 ddf->anchor.data_section_length = __cpu_to_be32(1);
1698 sector += 1;
1699
1700 ddf->anchor.bbm_section_length = __cpu_to_be32(0);
1701 ddf->anchor.bbm_section_offset = __cpu_to_be32(0xFFFFFFFF);
1702 ddf->anchor.diag_space_length = __cpu_to_be32(0);
1703 ddf->anchor.diag_space_offset = __cpu_to_be32(0xFFFFFFFF);
1704 ddf->anchor.vendor_length = __cpu_to_be32(0);
1705 ddf->anchor.vendor_offset = __cpu_to_be32(0xFFFFFFFF);
1706
1707 memset(ddf->anchor.pad4, 0xff, 256);
1708
1709 memcpy(&ddf->primary, &ddf->anchor, 512);
1710 memcpy(&ddf->secondary, &ddf->anchor, 512);
1711
1712 ddf->primary.openflag = 1; /* I guess.. */
1713 ddf->primary.type = DDF_HEADER_PRIMARY;
1714
1715 ddf->secondary.openflag = 1; /* I guess.. */
1716 ddf->secondary.type = DDF_HEADER_SECONDARY;
1717
1718 ddf->active = &ddf->primary;
1719
1720 ddf->controller.magic = DDF_CONTROLLER_MAGIC;
1721
1722 /* 24 more bytes of fiction required.
1723 * first 8 are a 'vendor-id' - "Linux-MD"
1724 * Remaining 16 are serial number.... maybe a hostname would do?
1725 */
1726 memcpy(ddf->controller.guid, T10, sizeof(T10));
1ba6bff9
DW		 1727 gethostname(hostname, sizeof(hostname));
1728 hostname[sizeof(hostname) - 1] = 0;
a322f70c
DW		 1729 hostlen = strlen(hostname);
1730 memcpy(ddf->controller.guid + 24 - hostlen, hostname, hostlen);
1731 for (i = strlen(T10) ; i+hostlen < 24; i++)
1732 ddf->controller.guid[i] = ' ';
1733
1734 ddf->controller.type.vendor_id = __cpu_to_be16(0xDEAD);
1735 ddf->controller.type.device_id = __cpu_to_be16(0xBEEF);
1736 ddf->controller.type.sub_vendor_id = 0;
1737 ddf->controller.type.sub_device_id = 0;
1738 memcpy(ddf->controller.product_id, "What Is My PID??", 16);
1739 memset(ddf->controller.pad, 0xff, 8);
1740 memset(ddf->controller.vendor_data, 0xff, 448);
a9e1c11d
N		 1741 if (homehost && strlen(homehost) < 440)
1742 strcpy((char*)ddf->controller.vendor_data, homehost);
a322f70c 1743
3d2c4fc7
DW		 1744 if (posix_memalign((void**)&pd, 512, pdsize) != 0) {
1745 fprintf(stderr, Name ": %s could not allocate pd\n", __func__);
1746 return 0;
1747 }
6416d527 1748 ddf->phys = pd;
a322f70c
DW		 1749 ddf->pdsize = pdsize;
1750
1751 memset(pd, 0xff, pdsize);
1752 memset(pd, 0, sizeof(*pd));
076515ba 1753 pd->magic = DDF_PHYS_RECORDS_MAGIC;
a322f70c
DW		 1754 pd->used_pdes = __cpu_to_be16(0);
1755 pd->max_pdes = __cpu_to_be16(max_phys_disks);
1756 memset(pd->pad, 0xff, 52);
1757
3d2c4fc7
DW		 1758 if (posix_memalign((void**)&vd, 512, vdsize) != 0) {
1759 fprintf(stderr, Name ": %s could not allocate vd\n", __func__);
1760 return 0;
1761 }
6416d527 1762 ddf->virt = vd;
a322f70c
DW		 1763 ddf->vdsize = vdsize;
1764 memset(vd, 0, vdsize);
1765 vd->magic = DDF_VIRT_RECORDS_MAGIC;
1766 vd->populated_vdes = __cpu_to_be16(0);
1767 vd->max_vdes = __cpu_to_be16(max_virt_disks);
1768 memset(vd->pad, 0xff, 52);
1769
5f8097be
NB		 1770 for (i=0; i<max_virt_disks; i++)
1771 memset(&vd->entries[i], 0xff, sizeof(struct virtual_entry));
1772
a322f70c 1773 st->sb = ddf;
18a2f463 1774 ddf->updates_pending = 1;
a322f70c
DW		 1775 return 1;
1776}
1777
5f8097be
NB		 1778static int chunk_to_shift(int chunksize)
1779{
1780 return ffs(chunksize/512)-1;
1781}
1782
1783static int level_to_prl(int level)
1784{
1785 switch (level) {
1786 case LEVEL_LINEAR: return DDF_CONCAT;
1787 case 0: return DDF_RAID0;
1788 case 1: return DDF_RAID1;
1789 case 4: return DDF_RAID4;
1790 case 5: return DDF_RAID5;
1791 case 6: return DDF_RAID6;
1792 default: return -1;
1793 }
1794}
1795static int layout_to_rlq(int level, int layout, int raiddisks)
1796{
1797 switch(level) {
1798 case 0:
1799 return DDF_RAID0_SIMPLE;
1800 case 1:
1801 switch(raiddisks) {
1802 case 2: return DDF_RAID1_SIMPLE;
1803 case 3: return DDF_RAID1_MULTI;
1804 default: return -1;
1805 }
1806 case 4:
1807 switch(layout) {
1808 case 0: return DDF_RAID4_N;
1809 }
1810 break;
1811 case 5:
5f8097be
NB		 1812 switch(layout) {
1813 case ALGORITHM_LEFT_ASYMMETRIC:
1814 return DDF_RAID5_N_RESTART;
1815 case ALGORITHM_RIGHT_ASYMMETRIC:
b640a252 1816 return DDF_RAID5_0_RESTART;
5f8097be
NB		 1817 case ALGORITHM_LEFT_SYMMETRIC:
1818 return DDF_RAID5_N_CONTINUE;
1819 case ALGORITHM_RIGHT_SYMMETRIC:
1820 return -1; /* not mentioned in standard */
1821 }
b640a252
N		 1822 case 6:
1823 switch(layout) {
1824 case ALGORITHM_ROTATING_N_RESTART:
1825 return DDF_RAID5_N_RESTART;
1826 case ALGORITHM_ROTATING_ZERO_RESTART:
1827 return DDF_RAID6_0_RESTART;
1828 case ALGORITHM_ROTATING_N_CONTINUE:
1829 return DDF_RAID5_N_CONTINUE;
1830 }
5f8097be
NB		 1831 }
1832 return -1;
1833}
1834
598f0d58
NB		 1835static int rlq_to_layout(int rlq, int prl, int raiddisks)
1836{
1837 switch(prl) {
1838 case DDF_RAID0:
1839 return 0; /* hopefully rlq == DDF_RAID0_SIMPLE */
1840 case DDF_RAID1:
1841 return 0; /* hopefully rlq == SIMPLE or MULTI depending
1842 on raiddisks*/
1843 case DDF_RAID4:
1844 switch(rlq) {
1845 case DDF_RAID4_N:
1846 return 0;
1847 default:
1848 /* not supported */
1849 return -1; /* FIXME this isn't checked */
1850 }
1851 case DDF_RAID5:
598f0d58
NB		 1852 switch(rlq) {
1853 case DDF_RAID5_N_RESTART:
1854 return ALGORITHM_LEFT_ASYMMETRIC;
1855 case DDF_RAID5_0_RESTART:
1856 return ALGORITHM_RIGHT_ASYMMETRIC;
1857 case DDF_RAID5_N_CONTINUE:
1858 return ALGORITHM_LEFT_SYMMETRIC;
1859 default:
1860 return -1;
1861 }
59e36268
NB		 1862 case DDF_RAID6:
1863 switch(rlq) {
1864 case DDF_RAID5_N_RESTART:
b640a252 1865 return ALGORITHM_ROTATING_N_RESTART;
59e36268 1866 case DDF_RAID6_0_RESTART:
b640a252 1867 return ALGORITHM_ROTATING_ZERO_RESTART;
59e36268 1868 case DDF_RAID5_N_CONTINUE:
b640a252 1869 return ALGORITHM_ROTATING_N_CONTINUE;
59e36268
NB		 1870 default:
1871 return -1;
1872 }
598f0d58
NB		 1873 }
1874 return -1;
1875}
1876
0e600426 1877#ifndef MDASSEMBLE
59e36268
NB		 1878struct extent {
1879 unsigned long long start, size;
1880};
78e44928 1881static int cmp_extent(const void *av, const void *bv)
59e36268
NB		 1882{
1883 const struct extent *a = av;
1884 const struct extent *b = bv;
1885 if (a->start < b->start)
1886 return -1;
1887 if (a->start > b->start)
1888 return 1;
1889 return 0;
1890}
1891
78e44928 1892static struct extent *get_extents(struct ddf_super *ddf, struct dl *dl)
59e36268
NB		 1893{
1894 /* find a list of used extents on the give physical device
1895 * (dnum) of the given ddf.
1896 * Return a malloced array of 'struct extent'
1897
1898FIXME ignore DDF_Legacy devices?
1899
1900 */
1901 struct extent *rv;
1902 int n = 0;
1903 int i, j;
1904
1905 rv = malloc(sizeof(struct extent) * (ddf->max_part + 2));
1906 if (!rv)
1907 return NULL;
1908
1909 for (i = 0; i < ddf->max_part; i++) {
1910 struct vcl *v = dl->vlist[i];
1911 if (v == NULL)
1912 continue;
1913 for (j=0; j < v->conf.prim_elmnt_count; j++)
1914 if (v->conf.phys_refnum[j] == dl->disk.refnum) {
1915 /* This device plays role 'j' in 'v'. */
1916 rv[n].start = __be64_to_cpu(v->lba_offset[j]);
1917 rv[n].size = __be64_to_cpu(v->conf.blocks);
1918 n++;
1919 break;
1920 }
1921 }
1922 qsort(rv, n, sizeof(*rv), cmp_extent);
1923
1924 rv[n].start = __be64_to_cpu(ddf->phys->entries[dl->pdnum].config_size);
1925 rv[n].size = 0;
1926 return rv;
1927}
0e600426 1928#endif
59e36268 1929
5f8097be
NB		 1930static int init_super_ddf_bvd(struct supertype *st,
1931 mdu_array_info_t *info,
1932 unsigned long long size,
1933 char *name, char *homehost,
1934 int *uuid)
1935{
1936 /* We are creating a BVD inside a pre-existing container.
1937 * so st->sb is already set.
1938 * We need to create a new vd_config and a new virtual_entry
1939 */
1940 struct ddf_super *ddf = st->sb;
1941 int venum;
1942 struct virtual_entry *ve;
1943 struct vcl *vcl;
1944 struct vd_config *vc;
5f8097be
NB		 1945
1946 if (__be16_to_cpu(ddf->virt->populated_vdes)
1947 >= __be16_to_cpu(ddf->virt->max_vdes)) {
1948 fprintf(stderr, Name": This ddf already has the "
1949 "maximum of %d virtual devices\n",
1950 __be16_to_cpu(ddf->virt->max_vdes));
1951 return 0;
1952 }
1953
1954 for (venum = 0; venum < __be16_to_cpu(ddf->virt->max_vdes); venum++)
1955 if (all_ff(ddf->virt->entries[venum].guid))
1956 break;
1957 if (venum == __be16_to_cpu(ddf->virt->max_vdes)) {
1958 fprintf(stderr, Name ": Cannot find spare slot for "
1959 "virtual disk - DDF is corrupt\n");
1960 return 0;
1961 }
1962 ve = &ddf->virt->entries[venum];
1963
1964 /* A Virtual Disk GUID contains the T10 Vendor ID, controller type,
1965 * timestamp, random number
1966 */
1967 make_header_guid(ve->guid);
1968 ve->unit = __cpu_to_be16(info->md_minor);
1969 ve->pad0 = 0xFFFF;
1970 ve->guid_crc = crc32(0, (unsigned char*)ddf->anchor.guid, DDF_GUID_LEN);
1971 ve->type = 0;
7a7cc504
NB		 1972 ve->state = DDF_state_degraded; /* Will be modified as devices are added */
1973 if (info->state & 1) /* clean */
1974 ve->init_state = DDF_init_full;
1975 else
1976 ve->init_state = DDF_init_not;
1977
5f8097be
NB		 1978 memset(ve->pad1, 0xff, 14);
1979 memset(ve->name, ' ', 16);
1980 if (name)
1981 strncpy(ve->name, name, 16);
1982 ddf->virt->populated_vdes =
1983 __cpu_to_be16(__be16_to_cpu(ddf->virt->populated_vdes)+1);
1984
1985 /* Now create a new vd_config */
3d2c4fc7
DW		 1986 if (posix_memalign((void**)&vcl, 512,
1987 (offsetof(struct vcl, conf) + ddf->conf_rec_len * 512)) != 0) {
1988 fprintf(stderr, Name ": %s could not allocate vd_config\n", __func__);
1989 return 0;
1990 }
8c3b8c2c 1991 vcl->lba_offset = (__u64*) &vcl->conf.phys_refnum[ddf->mppe];
59e36268 1992 vcl->vcnum = venum;
f7e7067b 1993 sprintf(st->subarray, "%d", venum);
59e36268 1994 vcl->block_sizes = NULL; /* FIXME not for CONCAT */
5f8097be
NB		 1995
1996 vc = &vcl->conf;
1997
1998 vc->magic = DDF_VD_CONF_MAGIC;
1999 memcpy(vc->guid, ve->guid, DDF_GUID_LEN);
2000 vc->timestamp = __cpu_to_be32(time(0)-DECADE);
2001 vc->seqnum = __cpu_to_be32(1);
2002 memset(vc->pad0, 0xff, 24);
2003 vc->prim_elmnt_count = __cpu_to_be16(info->raid_disks);
2004 vc->chunk_shift = chunk_to_shift(info->chunk_size);
2005 vc->prl = level_to_prl(info->level);
2006 vc->rlq = layout_to_rlq(info->level, info->layout, info->raid_disks);
2007 vc->sec_elmnt_count = 1;
2008 vc->sec_elmnt_seq = 0;
2009 vc->srl = 0;
2010 vc->blocks = __cpu_to_be64(info->size * 2);
2011 vc->array_blocks = __cpu_to_be64(
2012 calc_array_size(info->level, info->raid_disks, info->layout,
2013 info->chunk_size, info->size*2));
2014 memset(vc->pad1, 0xff, 8);
2015 vc->spare_refs[0] = 0xffffffff;
2016 vc->spare_refs[1] = 0xffffffff;
2017 vc->spare_refs[2] = 0xffffffff;
2018 vc->spare_refs[3] = 0xffffffff;
2019 vc->spare_refs[4] = 0xffffffff;
2020 vc->spare_refs[5] = 0xffffffff;
2021 vc->spare_refs[6] = 0xffffffff;
2022 vc->spare_refs[7] = 0xffffffff;
2023 memset(vc->cache_pol, 0, 8);
2024 vc->bg_rate = 0x80;
2025 memset(vc->pad2, 0xff, 3);
2026 memset(vc->pad3, 0xff, 52);
2027 memset(vc->pad4, 0xff, 192);
2028 memset(vc->v0, 0xff, 32);
2029 memset(vc->v1, 0xff, 32);
2030 memset(vc->v2, 0xff, 16);
2031 memset(vc->v3, 0xff, 16);
2032 memset(vc->vendor, 0xff, 32);
598f0d58 2033
8c3b8c2c 2034 memset(vc->phys_refnum, 0xff, 4*ddf->mppe);
e5a2a3cf 2035 memset(vc->phys_refnum+ddf->mppe, 0x00, 8*ddf->mppe);
5f8097be
NB		 2036
2037 vcl->next = ddf->conflist;
2038 ddf->conflist = vcl;
d2ca6449 2039 ddf->currentconf = vcl;
18a2f463 2040 ddf->updates_pending = 1;
5f8097be
NB		 2041 return 1;
2042}
2043
0e600426 2044#ifndef MDASSEMBLE
5f8097be
NB		 2045static void add_to_super_ddf_bvd(struct supertype *st,
2046 mdu_disk_info_t *dk, int fd, char *devname)
2047{
2048 /* fd and devname identify a device with-in the ddf container (st).
2049 * dk identifies a location in the new BVD.
2050 * We need to find suitable free space in that device and update
2051 * the phys_refnum and lba_offset for the newly created vd_config.
2052 * We might also want to update the type in the phys_disk
5575e7d9 2053 * section.
8592f29d
N		 2054 *
2055 * Alternately: fd == -1 and we have already chosen which device to
2056 * use and recorded in dlist->raid_disk;
5f8097be
NB		 2057 */
2058 struct dl *dl;
2059 struct ddf_super *ddf = st->sb;
2060 struct vd_config *vc;
2061 __u64 *lba_offset;
7a7cc504 2062 int working;
5575e7d9 2063 int i;
59e36268
NB		 2064 unsigned long long blocks, pos, esize;
2065 struct extent *ex;
5f8097be 2066
8592f29d
N		 2067 if (fd == -1) {
2068 for (dl = ddf->dlist; dl ; dl = dl->next)
2069 if (dl->raiddisk == dk->raid_disk)
2070 break;
2071 } else {
2072 for (dl = ddf->dlist; dl ; dl = dl->next)
2073 if (dl->major == dk->major &&
2074 dl->minor == dk->minor)
2075 break;
2076 }
5f8097be
NB		 2077 if (!dl || ! (dk->state & (1<<MD_DISK_SYNC)))
2078 return;
2079
d2ca6449
NB		 2080 vc = &ddf->currentconf->conf;
2081 lba_offset = ddf->currentconf->lba_offset;
59e36268
NB		 2082
2083 ex = get_extents(ddf, dl);
2084 if (!ex)
2085 return;
2086
2087 i = 0; pos = 0;
2088 blocks = __be64_to_cpu(vc->blocks);
d2ca6449
NB		 2089 if (ddf->currentconf->block_sizes)
2090 blocks = ddf->currentconf->block_sizes[dk->raid_disk];
59e36268
NB		 2091
2092 do {
2093 esize = ex[i].start - pos;
2094 if (esize >= blocks)
2095 break;
2096 pos = ex[i].start + ex[i].size;
2097 i++;
2098 } while (ex[i-1].size);
2099
2100 free(ex);
2101 if (esize < blocks)
2102 return;
2103
d2ca6449 2104 ddf->currentdev = dk->raid_disk;
5f8097be 2105 vc->phys_refnum[dk->raid_disk] = dl->disk.refnum;
59e36268 2106 lba_offset[dk->raid_disk] = __cpu_to_be64(pos);
5f8097be 2107
5575e7d9
NB		 2108 for (i=0; i < ddf->max_part ; i++)
2109 if (dl->vlist[i] == NULL)
2110 break;
2111 if (i == ddf->max_part)
2112 return;
d2ca6449 2113 dl->vlist[i] = ddf->currentconf;
5f8097be 2114
8592f29d
N		 2115 if (fd >= 0)
2116 dl->fd = fd;
2117 if (devname)
2118 dl->devname = devname;
7a7cc504
NB		 2119
2120 /* Check how many working raid_disks, and if we can mark
2121 * array as optimal yet
2122 */
2123 working = 0;
5575e7d9 2124
7a7cc504
NB		 2125 for (i=0; i < __be16_to_cpu(vc->prim_elmnt_count); i++)
2126 if (vc->phys_refnum[i] != 0xffffffff)
2127 working++;
59e36268 2128
5575e7d9 2129 /* Find which virtual_entry */
d2ca6449 2130 i = ddf->currentconf->vcnum;
7a7cc504 2131 if (working == __be16_to_cpu(vc->prim_elmnt_count))
5575e7d9
NB		 2132 ddf->virt->entries[i].state =
2133 (ddf->virt->entries[i].state & ~DDF_state_mask)
7a7cc504
NB		 2134 | DDF_state_optimal;
2135
2136 if (vc->prl == DDF_RAID6 &&
2137 working+1 == __be16_to_cpu(vc->prim_elmnt_count))
5575e7d9
NB		 2138 ddf->virt->entries[i].state =
2139 (ddf->virt->entries[i].state & ~DDF_state_mask)
7a7cc504 2140 | DDF_state_part_optimal;
5575e7d9
NB		 2141
2142 ddf->phys->entries[dl->pdnum].type &= ~__cpu_to_be16(DDF_Global_Spare);
2143 ddf->phys->entries[dl->pdnum].type |= __cpu_to_be16(DDF_Active_in_VD);
18a2f463 2144 ddf->updates_pending = 1;
5f8097be
NB		 2145}
2146
a322f70c
DW		 2147/* add a device to a container, either while creating it or while
2148 * expanding a pre-existing container
2149 */
f20c3968 2150static int add_to_super_ddf(struct supertype *st,
a322f70c
DW		 2151 mdu_disk_info_t *dk, int fd, char *devname)
2152{
2153 struct ddf_super *ddf = st->sb;
2154 struct dl *dd;
2155 time_t now;
2156 struct tm *tm;
2157 unsigned long long size;
2158 struct phys_disk_entry *pde;
2159 int n, i;
2160 struct stat stb;
2161
78e44928
NB		 2162 if (ddf->currentconf) {
2163 add_to_super_ddf_bvd(st, dk, fd, devname);
f20c3968 2164 return 0;
78e44928
NB		 2165 }
2166
a322f70c
DW		 2167 /* This is device numbered dk->number. We need to create
2168 * a phys_disk entry and a more detailed disk_data entry.
2169 */
2170 fstat(fd, &stb);
3d2c4fc7
DW		 2171 if (posix_memalign((void**)&dd, 512,
2172 sizeof(*dd) + sizeof(dd->vlist[0]) * ddf->max_part) != 0) {
2173 fprintf(stderr, Name
2174 ": %s could allocate buffer for new disk, aborting\n",
2175 __func__);
f20c3968 2176 return 1;
3d2c4fc7 2177 }
a322f70c
DW		 2178 dd->major = major(stb.st_rdev);
2179 dd->minor = minor(stb.st_rdev);
2180 dd->devname = devname;
a322f70c 2181 dd->fd = fd;
b2280677 2182 dd->spare = NULL;
a322f70c
DW		 2183
2184 dd->disk.magic = DDF_PHYS_DATA_MAGIC;
2185 now = time(0);
2186 tm = localtime(&now);
2187 sprintf(dd->disk.guid, "%8s%04d%02d%02d",
2188 T10, tm->tm_year+1900, tm->tm_mon+1, tm->tm_mday);
bfb7ea78
N		 2189 *(__u32*)(dd->disk.guid + 16) = random32();
2190 *(__u32*)(dd->disk.guid + 20) = random32();
a322f70c 2191
59e36268
NB		 2192 do {
2193 /* Cannot be bothered finding a CRC of some irrelevant details*/
bfb7ea78 2194 dd->disk.refnum = random32();
59e36268
NB		 2195 for (i = __be16_to_cpu(ddf->active->max_pd_entries) - 1;
2196 i >= 0; i--)
2197 if (ddf->phys->entries[i].refnum == dd->disk.refnum)
2198 break;
2199 } while (i >= 0);
2200
a322f70c
DW		 2201 dd->disk.forced_ref = 1;
2202 dd->disk.forced_guid = 1;
2203 memset(dd->disk.vendor, ' ', 32);
2204 memcpy(dd->disk.vendor, "Linux", 5);
2205 memset(dd->disk.pad, 0xff, 442);
b2280677 2206 for (i = 0; i < ddf->max_part ; i++)
a322f70c
DW		 2207 dd->vlist[i] = NULL;
2208
2209 n = __be16_to_cpu(ddf->phys->used_pdes);
2210 pde = &ddf->phys->entries[n];
5575e7d9
NB		 2211 dd->pdnum = n;
2212
2cc2983d
N		 2213 if (st->update_tail) {
2214 int len = (sizeof(struct phys_disk) +
2215 sizeof(struct phys_disk_entry));
2216 struct phys_disk *pd;
2217
2218 pd = malloc(len);
2219 pd->magic = DDF_PHYS_RECORDS_MAGIC;
2220 pd->used_pdes = __cpu_to_be16(n);
2221 pde = &pd->entries[0];
2222 dd->mdupdate = pd;
2223 } else {
2224 n++;
2225 ddf->phys->used_pdes = __cpu_to_be16(n);
2226 }
a322f70c
DW		 2227
2228 memcpy(pde->guid, dd->disk.guid, DDF_GUID_LEN);
2229 pde->refnum = dd->disk.refnum;
5575e7d9 2230 pde->type = __cpu_to_be16(DDF_Forced_PD_GUID | DDF_Global_Spare);
a322f70c
DW		 2231 pde->state = __cpu_to_be16(DDF_Online);
2232 get_dev_size(fd, NULL, &size);
2233 /* We are required to reserve 32Meg, and record the size in sectors */
2234 pde->config_size = __cpu_to_be64( (size - 32*1024*1024) / 512);
2235 sprintf(pde->path, "%17.17s","Information: nil") ;
2236 memset(pde->pad, 0xff, 6);
2237
d2ca6449 2238 dd->size = size >> 9;
2cc2983d
N		 2239 if (st->update_tail) {
2240 dd->next = ddf->add_list;
2241 ddf->add_list = dd;
2242 } else {
2243 dd->next = ddf->dlist;
2244 ddf->dlist = dd;
2245 ddf->updates_pending = 1;
2246 }
f20c3968
DW		 2247
2248 return 0;
a322f70c
DW		 2249}
2250
2251/*
2252 * This is the write_init_super method for a ddf container. It is
2253 * called when creating a container or adding another device to a
2254 * container.
2255 */
2256
6416d527 2257static unsigned char null_conf[4096+512];
18a2f463 2258
7a7cc504 2259static int __write_init_super_ddf(struct supertype *st, int do_close)
a322f70c
DW		 2260{
2261
2262 struct ddf_super *ddf = st->sb;
2263 int i;
2264 struct dl *d;
2265 int n_config;
2266 int conf_size;
175593bf
DW		 2267 int attempts = 0;
2268 int successes = 0;
a322f70c
DW		 2269 unsigned long long size, sector;
2270
175593bf
DW		 2271 /* try to write updated metadata,
2272 * if we catch a failure move on to the next disk
2273 */
a322f70c
DW		 2274 for (d = ddf->dlist; d; d=d->next) {
2275 int fd = d->fd;
2276
2277 if (fd < 0)
2278 continue;
2279
175593bf 2280 attempts++;
a322f70c
DW		 2281 /* We need to fill in the primary, (secondary) and workspace
2282 * lba's in the headers, set their checksums,
2283 * Also checksum phys, virt....
2284 *
2285 * Then write everything out, finally the anchor is written.
2286 */
2287 get_dev_size(fd, NULL, &size);
2288 size /= 512;
2289 ddf->anchor.workspace_lba = __cpu_to_be64(size - 32*1024*2);
2290 ddf->anchor.primary_lba = __cpu_to_be64(size - 16*1024*2);
2291 ddf->anchor.seq = __cpu_to_be32(1);
2292 memcpy(&ddf->primary, &ddf->anchor, 512);
2293 memcpy(&ddf->secondary, &ddf->anchor, 512);
2294
2295 ddf->anchor.openflag = 0xFF; /* 'open' means nothing */
2296 ddf->anchor.seq = 0xFFFFFFFF; /* no sequencing in anchor */
2297 ddf->anchor.crc = calc_crc(&ddf->anchor, 512);
2298
2299 ddf->primary.openflag = 0;
2300 ddf->primary.type = DDF_HEADER_PRIMARY;
2301
2302 ddf->secondary.openflag = 0;
2303 ddf->secondary.type = DDF_HEADER_SECONDARY;
2304
2305 ddf->primary.crc = calc_crc(&ddf->primary, 512);
2306 ddf->secondary.crc = calc_crc(&ddf->secondary, 512);
2307
2308 sector = size - 16*1024*2;
2309 lseek64(fd, sector<<9, 0);
175593bf
DW		 2310 if (write(fd, &ddf->primary, 512) < 0)
2311 continue;
a322f70c
DW		 2312
2313 ddf->controller.crc = calc_crc(&ddf->controller, 512);
175593bf
DW		 2314 if (write(fd, &ddf->controller, 512) < 0)
2315 continue;
a322f70c
DW		 2316
2317 ddf->phys->crc = calc_crc(ddf->phys, ddf->pdsize);
2318
175593bf
DW		 2319 if (write(fd, ddf->phys, ddf->pdsize) < 0)
2320 continue;
a322f70c
DW		 2321
2322 ddf->virt->crc = calc_crc(ddf->virt, ddf->vdsize);
175593bf
DW		 2323 if (write(fd, ddf->virt, ddf->vdsize) < 0)
2324 continue;
a322f70c
DW		 2325
2326 /* Now write lots of config records. */
8c3b8c2c
NB		 2327 n_config = ddf->max_part;
2328 conf_size = ddf->conf_rec_len * 512;
a322f70c
DW		 2329 for (i = 0 ; i <= n_config ; i++) {
2330 struct vcl *c = d->vlist[i];
b2280677
NB		 2331 if (i == n_config)
2332 c = (struct vcl*)d->spare;
a322f70c
DW		 2333
2334 if (c) {
2335 c->conf.crc = calc_crc(&c->conf, conf_size);
175593bf
DW		 2336 if (write(fd, &c->conf, conf_size) < 0)
2337 break;
a322f70c 2338 } else {
6416d527 2339 char *null_aligned = (char*)((((unsigned long)null_conf)+511)&~511UL);
18a2f463
NB		 2340 if (null_conf[0] != 0xff)
2341 memset(null_conf, 0xff, sizeof(null_conf));
2342 int togo = conf_size;
6416d527 2343 while (togo > sizeof(null_conf)-512) {
175593bf
DW		 2344 if (write(fd, null_aligned, sizeof(null_conf)-512) < 0)
2345 break;
6416d527 2346 togo -= sizeof(null_conf)-512;
18a2f463 2347 }
175593bf
DW		 2348 if (write(fd, null_aligned, togo) < 0)
2349 break;
a322f70c
DW		 2350 }
2351 }
175593bf
DW		 2352 if (i <= n_config)
2353 continue;
a322f70c 2354 d->disk.crc = calc_crc(&d->disk, 512);
175593bf
DW		 2355 if (write(fd, &d->disk, 512) < 0)
2356 continue;
a322f70c
DW		 2357
2358 /* Maybe do the same for secondary */
2359
2360 lseek64(fd, (size-1)*512, SEEK_SET);
175593bf
DW		 2361 if (write(fd, &ddf->anchor, 512) < 0)
2362 continue;
2363 successes++;
2364 }
2365
2366 if (do_close)
2367 for (d = ddf->dlist; d; d=d->next) {
2368 close(d->fd);
7a7cc504
NB		 2369 d->fd = -1;
2370 }
175593bf
DW		 2371
2372 return attempts != successes;
a322f70c 2373}
7a7cc504
NB		 2374
2375static int write_init_super_ddf(struct supertype *st)
2376{
9b1fb677
DW		 2377 struct ddf_super *ddf = st->sb;
2378 struct vcl *currentconf = ddf->currentconf;
2379
2380 /* we are done with currentconf reset it to point st at the container */
2381 ddf->currentconf = NULL;
edd8d13c
NB		 2382
2383 if (st->update_tail) {
2384 /* queue the virtual_disk and vd_config as metadata updates */
2385 struct virtual_disk *vd;
2386 struct vd_config *vc;
edd8d13c
NB		 2387 int len;
2388
9b1fb677 2389 if (!currentconf) {
2cc2983d
N		 2390 int len = (sizeof(struct phys_disk) +
2391 sizeof(struct phys_disk_entry));
2392
2393 /* adding a disk to the container. */
2394 if (!ddf->add_list)
2395 return 0;
2396
2397 append_metadata_update(st, ddf->add_list->mdupdate, len);
2398 ddf->add_list->mdupdate = NULL;
2399 return 0;
2400 }
2401
2402 /* Newly created VD */
2403
edd8d13c
NB		 2404 /* First the virtual disk. We have a slightly fake header */
2405 len = sizeof(struct virtual_disk) + sizeof(struct virtual_entry);
2406 vd = malloc(len);
2407 *vd = *ddf->virt;
9b1fb677
DW		 2408 vd->entries[0] = ddf->virt->entries[currentconf->vcnum];
2409 vd->populated_vdes = __cpu_to_be16(currentconf->vcnum);
edd8d13c
NB		 2410 append_metadata_update(st, vd, len);
2411
2412 /* Then the vd_config */
2413 len = ddf->conf_rec_len * 512;
2414 vc = malloc(len);
9b1fb677 2415 memcpy(vc, &currentconf->conf, len);
edd8d13c
NB		 2416 append_metadata_update(st, vc, len);
2417
2418 /* FIXME I need to close the fds! */
2419 return 0;
d682f344
N		 2420 } else {
2421 struct dl *d;
2422 for (d = ddf->dlist; d; d=d->next)
2423 while (Kill(d->devname, NULL, 0, 1, 1) == 0);
edd8d13c 2424 return __write_init_super_ddf(st, 1);
d682f344 2425 }
7a7cc504
NB		 2426}
2427
a322f70c
DW		 2428#endif
2429
2430static __u64 avail_size_ddf(struct supertype *st, __u64 devsize)
2431{
2432 /* We must reserve the last 32Meg */
2433 if (devsize <= 32*1024*2)
2434 return 0;
2435 return devsize - 32*1024*2;
2436}
2437
2438#ifndef MDASSEMBLE
8592f29d
N		 2439
2440static int reserve_space(struct supertype *st, int raiddisks,
2441 unsigned long long size, int chunk,
2442 unsigned long long *freesize)
2443{
2444 /* Find 'raiddisks' spare extents at least 'size' big (but
2445 * only caring about multiples of 'chunk') and remember
2446 * them.
2447 * If the cannot be found, fail.
2448 */
2449 struct dl *dl;
2450 struct ddf_super *ddf = st->sb;
2451 int cnt = 0;
2452
2453 for (dl = ddf->dlist; dl ; dl=dl->next) {
2454 dl->raiddisk = -1;
2455 dl->esize = 0;
2456 }
2457 /* Now find largest extent on each device */
2458 for (dl = ddf->dlist ; dl ; dl=dl->next) {
2459 struct extent *e = get_extents(ddf, dl);
2460 unsigned long long pos = 0;
2461 int i = 0;
2462 int found = 0;
2463 unsigned long long minsize = size;
2464
2465 if (size == 0)
2466 minsize = chunk;
2467
2468 if (!e)
2469 continue;
2470 do {
2471 unsigned long long esize;
2472 esize = e[i].start - pos;
2473 if (esize >= minsize) {
2474 found = 1;
2475 minsize = esize;
2476 }
2477 pos = e[i].start + e[i].size;
2478 i++;
2479 } while (e[i-1].size);
2480 if (found) {
2481 cnt++;
2482 dl->esize = minsize;
2483 }
2484 free(e);
2485 }
2486 if (cnt < raiddisks) {
2487 fprintf(stderr, Name ": not enough devices with space to create array.\n");
2488 return 0; /* No enough free spaces large enough */
2489 }
2490 if (size == 0) {
2491 /* choose the largest size of which there are at least 'raiddisk' */
2492 for (dl = ddf->dlist ; dl ; dl=dl->next) {
2493 struct dl *dl2;
2494 if (dl->esize <= size)
2495 continue;
2496 /* This is bigger than 'size', see if there are enough */
2497 cnt = 0;
2498 for (dl2 = dl; dl2 ; dl2=dl2->next)
2499 if (dl2->esize >= dl->esize)
2500 cnt++;
2501 if (cnt >= raiddisks)
2502 size = dl->esize;
2503 }
2504 if (chunk) {
2505 size = size / chunk;
2506 size *= chunk;
2507 }
2508 *freesize = size;
2509 if (size < 32) {
2510 fprintf(stderr, Name ": not enough spare devices to create array.\n");
2511 return 0;
2512 }
2513 }
2514 /* We have a 'size' of which there are enough spaces.
2515 * We simply do a first-fit */
2516 cnt = 0;
2517 for (dl = ddf->dlist ; dl && cnt < raiddisks ; dl=dl->next) {
2518 if (dl->esize < size)
2519 continue;
2520
2521 dl->raiddisk = cnt;
2522 cnt++;
2523 }
2524 return 1;
2525}
2526
2527
2528
2c514b71
NB		 2529static int
2530validate_geometry_ddf_container(struct supertype *st,
2531 int level, int layout, int raiddisks,
2532 int chunk, unsigned long long size,
2533 char *dev, unsigned long long *freesize,
2534 int verbose);
78e44928
NB		 2535
2536static int validate_geometry_ddf_bvd(struct supertype *st,
2537 int level, int layout, int raiddisks,
2538 int chunk, unsigned long long size,
2c514b71
NB		 2539 char *dev, unsigned long long *freesize,
2540 int verbose);
78e44928
NB		 2541
2542static int validate_geometry_ddf(struct supertype *st,
2c514b71
NB		 2543 int level, int layout, int raiddisks,
2544 int chunk, unsigned long long size,
2545 char *dev, unsigned long long *freesize,
2546 int verbose)
a322f70c
DW		 2547{
2548 int fd;
2549 struct mdinfo *sra;
2550 int cfd;
2551
2552 /* ddf potentially supports lots of things, but it depends on
2553 * what devices are offered (and maybe kernel version?)
2554 * If given unused devices, we will make a container.
2555 * If given devices in a container, we will make a BVD.
2556 * If given BVDs, we make an SVD, changing all the GUIDs in the process.
2557 */
2558
2559 if (level == LEVEL_CONTAINER) {
78e44928
NB		 2560 /* Must be a fresh device to add to a container */
2561 return validate_geometry_ddf_container(st, level, layout,
2c514b71
NB		 2562 raiddisks, chunk,
2563 size, dev, freesize,
2564 verbose);
5f8097be
NB		 2565 }
2566
78e44928
NB		 2567 if (!dev) {
2568 /* Initial sanity check. Exclude illegal levels. */
2569 int i;
2570 for (i=0; ddf_level_num[i].num1 != MAXINT; i++)
2571 if (ddf_level_num[i].num2 == level)
2572 break;
b42f577a
N		 2573 if (ddf_level_num[i].num1 == MAXINT) {
2574 if (verbose)
2575 fprintf(stderr, Name ": DDF does not support level %d arrays\n",
2576 level);
78e44928 2577 return 0;
b42f577a 2578 }
78e44928 2579 /* Should check layout? etc */
8592f29d
N		 2580
2581 if (st->sb && freesize) {
2582 /* --create was given a container to create in.
2583 * So we need to check that there are enough
2584 * free spaces and return the amount of space.
2585 * We may as well remember which drives were
2586 * chosen so that add_to_super/getinfo_super
2587 * can return them.
2588 */
2589 return reserve_space(st, raiddisks, size, chunk, freesize);
2590 }
a322f70c 2591 return 1;
78e44928 2592 }
a322f70c 2593
8592f29d
N		 2594 if (st->sb) {
2595 /* A container has already been opened, so we are
2596 * creating in there. Maybe a BVD, maybe an SVD.
2597 * Should make a distinction one day.
2598 */
2599 return validate_geometry_ddf_bvd(st, level, layout, raiddisks,
2600 chunk, size, dev, freesize,
2601 verbose);
2602 }
78e44928
NB		 2603 /* This is the first device for the array.
2604 * If it is a container, we read it in and do automagic allocations,
2605 * no other devices should be given.
2606 * Otherwise it must be a member device of a container, and we
2607 * do manual allocation.
2608 * Later we should check for a BVD and make an SVD.
a322f70c 2609 */
a322f70c
DW		 2610 fd = open(dev, O_RDONLY|O_EXCL, 0);
2611 if (fd >= 0) {
2612 sra = sysfs_read(fd, 0, GET_VERSION);
2613 close(fd);
2614 if (sra && sra->array.major_version == -1 &&
78e44928
NB		 2615 strcmp(sra->text_version, "ddf") == 0) {
2616
2617 /* load super */
2618 /* find space for 'n' devices. */
2619 /* remember the devices */
2620 /* Somehow return the fact that we have enough */
a322f70c
DW		 2621 }
2622
2c514b71
NB		 2623 if (verbose)
2624 fprintf(stderr,
2625 Name ": ddf: Cannot create this array "
b42f577a 2626 "on device %s - a container is required.\n",
2c514b71 2627 dev);
a322f70c
DW		 2628 return 0;
2629 }
2630 if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) {
2c514b71
NB		 2631 if (verbose)
2632 fprintf(stderr, Name ": ddf: Cannot open %s: %s\n",
2633 dev, strerror(errno));
a322f70c
DW		 2634 return 0;
2635 }
2636 /* Well, it is in use by someone, maybe a 'ddf' container. */
2637 cfd = open_container(fd);
2638 if (cfd < 0) {
2639 close(fd);
2c514b71
NB		 2640 if (verbose)
2641 fprintf(stderr, Name ": ddf: Cannot use %s: %s\n",
2642 dev, strerror(EBUSY));
a322f70c
DW		 2643 return 0;
2644 }
2645 sra = sysfs_read(cfd, 0, GET_VERSION);
2646 close(fd);
2647 if (sra && sra->array.major_version == -1 &&
2648 strcmp(sra->text_version, "ddf") == 0) {
2649 /* This is a member of a ddf container. Load the container
2650 * and try to create a bvd
2651 */
2652 struct ddf_super *ddf;
a322f70c 2653 if (load_super_ddf_all(st, cfd, (void **)&ddf, NULL, 1) == 0) {
5f8097be 2654 st->sb = ddf;
2f6079dc 2655 st->container_dev = fd2devnum(cfd);
a322f70c 2656 close(cfd);
78e44928 2657 return validate_geometry_ddf_bvd(st, level, layout,
a322f70c 2658 raiddisks, chunk, size,
2c514b71
NB		 2659 dev, freesize,
2660 verbose);
a322f70c
DW		 2661 }
2662 close(cfd);
c42ec1ed
DW		 2663 } else /* device may belong to a different container */
2664 return 0;
2665
a322f70c
DW		 2666 return 1;
2667}
2668
2c514b71
NB		 2669static int
2670validate_geometry_ddf_container(struct supertype *st,
2671 int level, int layout, int raiddisks,
2672 int chunk, unsigned long long size,
2673 char *dev, unsigned long long *freesize,
2674 int verbose)
a322f70c
DW		 2675{
2676 int fd;
2677 unsigned long long ldsize;
2678
2679 if (level != LEVEL_CONTAINER)
2680 return 0;
2681 if (!dev)
2682 return 1;
2683
2684 fd = open(dev, O_RDONLY|O_EXCL, 0);
2685 if (fd < 0) {
2c514b71
NB		 2686 if (verbose)
2687 fprintf(stderr, Name ": ddf: Cannot open %s: %s\n",
2688 dev, strerror(errno));
a322f70c
DW		 2689 return 0;
2690 }
2691 if (!get_dev_size(fd, dev, &ldsize)) {
2692 close(fd);
2693 return 0;
2694 }
2695 close(fd);
2696
56aca704 2697 *freesize = avail_size_ddf(st, ldsize >> 9);
ea17e7aa
N		 2698 if (*freesize == 0)
2699 return 0;
a322f70c
DW		 2700
2701 return 1;
2702}
2703
78e44928
NB		 2704static int validate_geometry_ddf_bvd(struct supertype *st,
2705 int level, int layout, int raiddisks,
2706 int chunk, unsigned long long size,
2c514b71
NB		 2707 char *dev, unsigned long long *freesize,
2708 int verbose)
a322f70c
DW		 2709{
2710 struct stat stb;
2711 struct ddf_super *ddf = st->sb;
2712 struct dl *dl;
5f8097be
NB		 2713 unsigned long long pos = 0;
2714 unsigned long long maxsize;
2715 struct extent *e;
2716 int i;
a322f70c 2717 /* ddf/bvd supports lots of things, but not containers */
b42f577a
N		 2718 if (level == LEVEL_CONTAINER) {
2719 if (verbose)
2720 fprintf(stderr, Name ": DDF cannot create a container within an container\n");
a322f70c 2721 return 0;
b42f577a 2722 }
a322f70c
DW		 2723 /* We must have the container info already read in. */
2724 if (!ddf)
2725 return 0;
2726
5f8097be
NB		 2727 if (!dev) {
2728 /* General test: make sure there is space for
2729 * 'raiddisks' device extents of size 'size'.
2730 */
2731 unsigned long long minsize = size;
2732 int dcnt = 0;
2733 if (minsize == 0)
2734 minsize = 8;
2735 for (dl = ddf->dlist; dl ; dl = dl->next)
2736 {
2737 int found = 0;
7e1432fb 2738 pos = 0;
5f8097be
NB		 2739
2740 i = 0;
2741 e = get_extents(ddf, dl);
2742 if (!e) continue;
2743 do {
2744 unsigned long long esize;
2745 esize = e[i].start - pos;
2746 if (esize >= minsize)
2747 found = 1;
2748 pos = e[i].start + e[i].size;
2749 i++;
2750 } while (e[i-1].size);
2751 if (found)
2752 dcnt++;
2753 free(e);
2754 }
2755 if (dcnt < raiddisks) {
2c514b71
NB		 2756 if (verbose)
2757 fprintf(stderr,
2758 Name ": ddf: Not enough devices with "
2759 "space for this array (%d < %d)\n",
2760 dcnt, raiddisks);
5f8097be
NB		 2761 return 0;
2762 }
2763 return 1;
2764 }
a322f70c
DW		 2765 /* This device must be a member of the set */
2766 if (stat(dev, &stb) < 0)
2767 return 0;
2768 if ((S_IFMT & stb.st_mode) != S_IFBLK)
2769 return 0;
2770 for (dl = ddf->dlist ; dl ; dl = dl->next) {
2771 if (dl->major == major(stb.st_rdev) &&
2772 dl->minor == minor(stb.st_rdev))
2773 break;
2774 }
5f8097be 2775 if (!dl) {
2c514b71
NB		 2776 if (verbose)
2777 fprintf(stderr, Name ": ddf: %s is not in the "
2778 "same DDF set\n",
2779 dev);
5f8097be
NB		 2780 return 0;
2781 }
2782 e = get_extents(ddf, dl);
2783 maxsize = 0;
2784 i = 0;
2785 if (e) do {
2786 unsigned long long esize;
2787 esize = e[i].start - pos;
2788 if (esize >= maxsize)
2789 maxsize = esize;
2790 pos = e[i].start + e[i].size;
2791 i++;
2792 } while (e[i-1].size);
2793 *freesize = maxsize;
a322f70c
DW		 2794 // FIXME here I am
2795
2796 return 1;
2797}
59e36268 2798
a322f70c
DW		 2799static int load_super_ddf_all(struct supertype *st, int fd,
2800 void **sbp, char *devname, int keep_fd)
2801{
2802 struct mdinfo *sra;
2803 struct ddf_super *super;
2804 struct mdinfo *sd, *best = NULL;
2805 int bestseq = 0;
2806 int seq;
2807 char nm[20];
2808 int dfd;
dab4a513
DW		 2809 int devnum = fd2devnum(fd);
2810 enum sysfs_read_flags flags;
a322f70c 2811
dab4a513
DW		 2812 flags = GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE;
2813 if (mdmon_running(devnum))
2814 flags |= SKIP_GONE_DEVS;
2815
2816 sra = sysfs_read(fd, 0, flags);
a322f70c
DW		 2817 if (!sra)
2818 return 1;
2819 if (sra->array.major_version != -1 ||
2820 sra->array.minor_version != -2 ||
2821 strcmp(sra->text_version, "ddf") != 0)
2822 return 1;
2823
6416d527 2824 if (posix_memalign((void**)&super, 512, sizeof(*super)) != 0)
a322f70c 2825 return 1;
a2349791 2826 memset(super, 0, sizeof(*super));
a322f70c
DW		 2827
2828 /* first, try each device, and choose the best ddf */
2829 for (sd = sra->devs ; sd ; sd = sd->next) {
2830 int rv;
2831 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
7a7cc504
NB		 2832 dfd = dev_open(nm, O_RDONLY);
2833 if (dfd < 0)
a322f70c
DW		 2834 return 2;
2835 rv = load_ddf_headers(dfd, super, NULL);
7a7cc504 2836 close(dfd);
a322f70c
DW		 2837 if (rv == 0) {
2838 seq = __be32_to_cpu(super->active->seq);
2839 if (super->active->openflag)
2840 seq--;
2841 if (!best || seq > bestseq) {
2842 bestseq = seq;
2843 best = sd;
2844 }
2845 }
2846 }
2847 if (!best)
2848 return 1;
2849 /* OK, load this ddf */
2850 sprintf(nm, "%d:%d", best->disk.major, best->disk.minor);
2851 dfd = dev_open(nm, O_RDONLY);
7a7cc504 2852 if (dfd < 0)
a322f70c
DW		 2853 return 1;
2854 load_ddf_headers(dfd, super, NULL);
2855 load_ddf_global(dfd, super, NULL);
2856 close(dfd);
2857 /* Now we need the device-local bits */
2858 for (sd = sra->devs ; sd ; sd = sd->next) {
3d2c4fc7
DW		 2859 int rv;
2860
a322f70c 2861 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
5f8097be 2862 dfd = dev_open(nm, keep_fd? O_RDWR : O_RDONLY);
7a7cc504 2863 if (dfd < 0)
a322f70c 2864 return 2;
3d2c4fc7
DW		 2865 rv = load_ddf_headers(dfd, super, NULL);
2866 if (rv == 0)
2867 rv = load_ddf_local(dfd, super, NULL, keep_fd);
5f8097be 2868 if (!keep_fd) close(dfd);
3d2c4fc7
DW		 2869 if (rv)
2870 return 1;
a322f70c 2871 }
f7e7067b
NB		 2872 if (st->subarray[0]) {
2873 struct vcl *v;
2874
2875 for (v = super->conflist; v; v = v->next)
2876 if (v->vcnum == atoi(st->subarray))
d2ca6449
NB		 2877 super->currentconf = v;
2878 if (!super->currentconf)
f7e7067b
NB		 2879 return 1;
2880 }
a322f70c
DW		 2881 *sbp = super;
2882 if (st->ss == NULL) {
78e44928 2883 st->ss = &super_ddf;
a322f70c
DW		 2884 st->minor_version = 0;
2885 st->max_devs = 512;
75aa18b5 2886 st->container_dev = fd2devnum(fd);
a322f70c 2887 }
352452c3 2888 st->loaded_container = 1;
a322f70c
DW		 2889 return 0;
2890}
0e600426 2891#endif /* MDASSEMBLE */
a322f70c 2892
598f0d58
NB		 2893static struct mdinfo *container_content_ddf(struct supertype *st)
2894{
2895 /* Given a container loaded by load_super_ddf_all,
2896 * extract information about all the arrays into
2897 * an mdinfo tree.
2898 *
2899 * For each vcl in conflist: create an mdinfo, fill it in,
2900 * then look for matching devices (phys_refnum) in dlist
2901 * and create appropriate device mdinfo.
2902 */
2903 struct ddf_super *ddf = st->sb;
2904 struct mdinfo *rest = NULL;
2905 struct vcl *vc;
2906
2907 for (vc = ddf->conflist ; vc ; vc=vc->next)
2908 {
598f0d58 2909 int i;
db42fa9b 2910 int j;
598f0d58
NB		 2911 struct mdinfo *this;
2912 this = malloc(sizeof(*this));
2913 memset(this, 0, sizeof(*this));
2914 this->next = rest;
2915 rest = this;
2916
598f0d58
NB		 2917 this->array.level = map_num1(ddf_level_num, vc->conf.prl);
2918 this->array.raid_disks =
2919 __be16_to_cpu(vc->conf.prim_elmnt_count);
60f18132
NB		 2920 this->array.layout = rlq_to_layout(vc->conf.rlq, vc->conf.prl,
2921 this->array.raid_disks);
598f0d58 2922 this->array.md_minor = -1;
f35f2525
N		 2923 this->array.major_version = -1;
2924 this->array.minor_version = -2;
598f0d58
NB		 2925 this->array.ctime = DECADE +
2926 __be32_to_cpu(*(__u32*)(vc->conf.guid+16));
2927 this->array.utime = DECADE +
2928 __be32_to_cpu(vc->conf.timestamp);
2929 this->array.chunk_size = 512 << vc->conf.chunk_shift;
2930
59e36268 2931 i = vc->vcnum;
7a7cc504
NB		 2932 if ((ddf->virt->entries[i].state & DDF_state_inconsistent) ||
2933 (ddf->virt->entries[i].init_state & DDF_initstate_mask) !=
ed9d66aa 2934 DDF_init_full) {
598f0d58 2935 this->array.state = 0;
ed9d66aa
NB		 2936 this->resync_start = 0;
2937 } else {
598f0d58 2938 this->array.state = 1;
b7528a20 2939 this->resync_start = MaxSector;
ed9d66aa 2940 }
db42fa9b
N		 2941 memcpy(this->name, ddf->virt->entries[i].name, 16);
2942 this->name[16]=0;
2943 for(j=0; j<16; j++)
2944 if (this->name[j] == ' ')
2945 this->name[j] = 0;
598f0d58
NB		 2946
2947 memset(this->uuid, 0, sizeof(this->uuid));
2948 this->component_size = __be64_to_cpu(vc->conf.blocks);
2949 this->array.size = this->component_size / 2;
5f2aace8 2950 this->container_member = i;
598f0d58 2951
c5afc314
N		 2952 ddf->currentconf = vc;
2953 uuid_from_super_ddf(st, this->uuid);
2954 ddf->currentconf = NULL;
2955
60f18132
NB		 2956 sprintf(this->text_version, "/%s/%d",
2957 devnum2devname(st->container_dev),
2958 this->container_member);
2959
8c3b8c2c 2960 for (i=0 ; i < ddf->mppe ; i++) {
598f0d58
NB		 2961 struct mdinfo *dev;
2962 struct dl *d;
2963
2964 if (vc->conf.phys_refnum[i] == 0xFFFFFFFF)
2965 continue;
2966
c0d50ca5
N		 2967 this->array.working_disks++;
2968
598f0d58
NB		 2969 for (d = ddf->dlist; d ; d=d->next)
2970 if (d->disk.refnum == vc->conf.phys_refnum[i])
2971 break;
2972 if (d == NULL)
bc17324f
N		 2973 /* Haven't found that one yet, maybe there are others */
2974 continue;
2975
598f0d58
NB		 2976 dev = malloc(sizeof(*dev));
2977 memset(dev, 0, sizeof(*dev));
2978 dev->next = this->devs;
2979 this->devs = dev;
2980
2981 dev->disk.number = __be32_to_cpu(d->disk.refnum);
2982 dev->disk.major = d->major;
2983 dev->disk.minor = d->minor;
2984 dev->disk.raid_disk = i;
2985 dev->disk.state = (1<<MD_DISK_SYNC)|(1<<MD_DISK_ACTIVE);
d23534e4 2986 dev->recovery_start = MaxSector;
598f0d58 2987
120f7677
NB		 2988 dev->events = __be32_to_cpu(ddf->primary.seq);
2989 dev->data_offset = __be64_to_cpu(vc->lba_offset[i]);
598f0d58
NB		 2990 dev->component_size = __be64_to_cpu(vc->conf.blocks);
2991 if (d->devname)
2992 strcpy(dev->name, d->devname);
2993 }
2994 }
2995 return rest;
2996}
2997
955e9ea1 2998static int store_super_ddf(struct supertype *st, int fd)
a322f70c 2999{
955e9ea1 3000 struct ddf_super *ddf = st->sb;
a322f70c 3001 unsigned long long dsize;
6416d527 3002 void *buf;
3d2c4fc7 3003 int rc;
a322f70c 3004
955e9ea1
DW		 3005 if (!ddf)
3006 return 1;
3007
3008 /* ->dlist and ->conflist will be set for updates, currently not
3009 * supported
3010 */
3011 if (ddf->dlist || ddf->conflist)
3012 return 1;
3013
a322f70c
DW		 3014 if (!get_dev_size(fd, NULL, &dsize))
3015 return 1;
3016
3d2c4fc7
DW		 3017 if (posix_memalign(&buf, 512, 512) != 0)
3018 return 1;
6416d527
NB		 3019 memset(buf, 0, 512);
3020
a322f70c 3021 lseek64(fd, dsize-512, 0);
3d2c4fc7 3022 rc = write(fd, buf, 512);
6416d527 3023 free(buf);
3d2c4fc7
DW		 3024 if (rc < 0)
3025 return 1;
a322f70c
DW		 3026 return 0;
3027}
3028
a19c88b8
NB		 3029static int compare_super_ddf(struct supertype *st, struct supertype *tst)
3030{
3031 /*
3032 * return:
3033 * 0 same, or first was empty, and second was copied
3034 * 1 second had wrong number
3035 * 2 wrong uuid
3036 * 3 wrong other info
3037 */
3038 struct ddf_super *first = st->sb;
3039 struct ddf_super *second = tst->sb;
3040
3041 if (!first) {
3042 st->sb = tst->sb;
3043 tst->sb = NULL;
3044 return 0;
3045 }
3046
3047 if (memcmp(first->anchor.guid, second->anchor.guid, DDF_GUID_LEN) != 0)
3048 return 2;
3049
3050 /* FIXME should I look at anything else? */
3051 return 0;
3052}
3053
0e600426 3054#ifndef MDASSEMBLE
4e5528c6
NB		 3055/*
3056 * A new array 'a' has been started which claims to be instance 'inst'
3057 * within container 'c'.
3058 * We need to confirm that the array matches the metadata in 'c' so
3059 * that we don't corrupt any metadata.
3060 */
cba0191b 3061static int ddf_open_new(struct supertype *c, struct active_array *a, char *inst)
549e9569 3062{
2c514b71 3063 dprintf("ddf: open_new %s\n", inst);
cba0191b 3064 a->info.container_member = atoi(inst);
549e9569
NB		 3065 return 0;
3066}
3067
4e5528c6
NB		 3068/*
3069 * The array 'a' is to be marked clean in the metadata.
ed9d66aa 3070 * If '->resync_start' is not ~(unsigned long long)0, then the array is only
4e5528c6
NB		 3071 * clean up to the point (in sectors). If that cannot be recorded in the
3072 * metadata, then leave it as dirty.
3073 *
3074 * For DDF, we need to clear the DDF_state_inconsistent bit in the
3075 * !global! virtual_disk.virtual_entry structure.
3076 */
01f157d7 3077static int ddf_set_array_state(struct active_array *a, int consistent)
549e9569 3078{
4e5528c6
NB		 3079 struct ddf_super *ddf = a->container->sb;
3080 int inst = a->info.container_member;
18a2f463 3081 int old = ddf->virt->entries[inst].state;
01f157d7
N		 3082 if (consistent == 2) {
3083 /* Should check if a recovery should be started FIXME */
3084 consistent = 1;
b7941fd6 3085 if (!is_resync_complete(&a->info))
01f157d7
N		 3086 consistent = 0;
3087 }
ed9d66aa
NB		 3088 if (consistent)
3089 ddf->virt->entries[inst].state &= ~DDF_state_inconsistent;
3090 else
4e5528c6 3091 ddf->virt->entries[inst].state |= DDF_state_inconsistent;
18a2f463
NB		 3092 if (old != ddf->virt->entries[inst].state)
3093 ddf->updates_pending = 1;
3094
3095 old = ddf->virt->entries[inst].init_state;
ed9d66aa 3096 ddf->virt->entries[inst].init_state &= ~DDF_initstate_mask;
b7941fd6 3097 if (is_resync_complete(&a->info))
ed9d66aa 3098 ddf->virt->entries[inst].init_state |= DDF_init_full;
b7941fd6 3099 else if (a->info.resync_start == 0)
ed9d66aa 3100 ddf->virt->entries[inst].init_state |= DDF_init_not;
4e5528c6 3101 else
ed9d66aa 3102 ddf->virt->entries[inst].init_state |= DDF_init_quick;
18a2f463
NB		 3103 if (old != ddf->virt->entries[inst].init_state)
3104 ddf->updates_pending = 1;
ed9d66aa 3105
2c514b71 3106 dprintf("ddf mark %d %s %llu\n", inst, consistent?"clean":"dirty",
b7941fd6 3107 a->info.resync_start);
01f157d7 3108 return consistent;
fd7cde1b
DW		 3109}
3110
7a7cc504
NB		 3111/*
3112 * The state of each disk is stored in the global phys_disk structure
3113 * in phys_disk.entries[n].state.
3114 * This makes various combinations awkward.
3115 * - When a device fails in any array, it must be failed in all arrays
3116 * that include a part of this device.
3117 * - When a component is rebuilding, we cannot include it officially in the
3118 * array unless this is the only array that uses the device.
3119 *
3120 * So: when transitioning:
3121 * Online -> failed, just set failed flag. monitor will propagate
3122 * spare -> online, the device might need to be added to the array.
3123 * spare -> failed, just set failed. Don't worry if in array or not.
3124 */
8d45d196 3125static void ddf_set_disk(struct active_array *a, int n, int state)
549e9569 3126{
7a7cc504
NB		 3127 struct ddf_super *ddf = a->container->sb;
3128 int inst = a->info.container_member;
3129 struct vd_config *vc = find_vdcr(ddf, inst);
3130 int pd = find_phys(ddf, vc->phys_refnum[n]);
3131 int i, st, working;
3132
3133 if (vc == NULL) {
2c514b71 3134 dprintf("ddf: cannot find instance %d!!\n", inst);
7a7cc504
NB		 3135 return;
3136 }
3137 if (pd < 0) {
3138 /* disk doesn't currently exist. If it is now in_sync,
3139 * insert it. */
3140 if ((state & DS_INSYNC) && ! (state & DS_FAULTY)) {
3141 /* Find dev 'n' in a->info->devs, determine the
3142 * ddf refnum, and set vc->phys_refnum and update
3143 * phys->entries[]
3144 */
3145 /* FIXME */
3146 }
3147 } else {
18a2f463 3148 int old = ddf->phys->entries[pd].state;
7a7cc504
NB		 3149 if (state & DS_FAULTY)
3150 ddf->phys->entries[pd].state |= __cpu_to_be16(DDF_Failed);
3151 if (state & DS_INSYNC) {
3152 ddf->phys->entries[pd].state |= __cpu_to_be16(DDF_Online);
3153 ddf->phys->entries[pd].state &= __cpu_to_be16(~DDF_Rebuilding);
3154 }
18a2f463
NB		 3155 if (old != ddf->phys->entries[pd].state)
3156 ddf->updates_pending = 1;
7a7cc504
NB		 3157 }
3158
2c514b71 3159 dprintf("ddf: set_disk %d to %x\n", n, state);
7e1432fb 3160
7a7cc504
NB		 3161 /* Now we need to check the state of the array and update
3162 * virtual_disk.entries[n].state.
3163 * It needs to be one of "optimal", "degraded", "failed".
3164 * I don't understand 'deleted' or 'missing'.
3165 */
3166 working = 0;
3167 for (i=0; i < a->info.array.raid_disks; i++) {
3168 pd = find_phys(ddf, vc->phys_refnum[i]);
3169 if (pd < 0)
3170 continue;
57632f4a
NB		 3171 st = __be16_to_cpu(ddf->phys->entries[pd].state);
3172 if ((st & (DDF_Online|DDF_Failed|DDF_Rebuilding))
7a7cc504
NB		 3173 == DDF_Online)
3174 working++;
3175 }
3176 state = DDF_state_degraded;
3177 if (working == a->info.array.raid_disks)
3178 state = DDF_state_optimal;
3179 else switch(vc->prl) {
3180 case DDF_RAID0:
3181 case DDF_CONCAT:
3182 case DDF_JBOD:
3183 state = DDF_state_failed;
3184 break;
3185 case DDF_RAID1:
3186 if (working == 0)
3187 state = DDF_state_failed;
3188 break;
3189 case DDF_RAID4:
3190 case DDF_RAID5:
3191 if (working < a->info.array.raid_disks-1)
3192 state = DDF_state_failed;
3193 break;
3194 case DDF_RAID6:
3195 if (working < a->info.array.raid_disks-2)
3196 state = DDF_state_failed;
3197 else if (working == a->info.array.raid_disks-1)
3198 state = DDF_state_part_optimal;
3199 break;
3200 }
3201
18a2f463
NB		 3202 if (ddf->virt->entries[inst].state !=
3203 ((ddf->virt->entries[inst].state & ~DDF_state_mask)
3204 | state)) {
3205
3206 ddf->virt->entries[inst].state =
3207 (ddf->virt->entries[inst].state & ~DDF_state_mask)
3208 | state;
3209 ddf->updates_pending = 1;
3210 }
7a7cc504 3211
549e9569
NB		 3212}
3213
2e735d19 3214static void ddf_sync_metadata(struct supertype *st)
549e9569 3215{
7a7cc504
NB		 3216
3217 /*
3218 * Write all data to all devices.
3219 * Later, we might be able to track whether only local changes
3220 * have been made, or whether any global data has been changed,
3221 * but ddf is sufficiently weird that it probably always
3222 * changes global data ....
3223 */
18a2f463
NB		 3224 struct ddf_super *ddf = st->sb;
3225 if (!ddf->updates_pending)
3226 return;
3227 ddf->updates_pending = 0;
2e735d19 3228 __write_init_super_ddf(st, 0);
2c514b71 3229 dprintf("ddf: sync_metadata\n");
549e9569
NB		 3230}
3231
88c164f4
NB		 3232static void ddf_process_update(struct supertype *st,
3233 struct metadata_update *update)
3234{
3235 /* Apply this update to the metadata.
3236 * The first 4 bytes are a DDF_*_MAGIC which guides
3237 * our actions.
3238 * Possible update are:
3239 * DDF_PHYS_RECORDS_MAGIC
3240 * Add a new physical device. Changes to this record
3241 * only happen implicitly.
3242 * used_pdes is the device number.
3243 * DDF_VIRT_RECORDS_MAGIC
3244 * Add a new VD. Possibly also change the 'access' bits.
3245 * populated_vdes is the entry number.
3246 * DDF_VD_CONF_MAGIC
3247 * New or updated VD. the VIRT_RECORD must already
3248 * exist. For an update, phys_refnum and lba_offset
3249 * (at least) are updated, and the VD_CONF must
3250 * be written to precisely those devices listed with
3251 * a phys_refnum.
3252 * DDF_SPARE_ASSIGN_MAGIC
3253 * replacement Spare Assignment Record... but for which device?
3254 *
3255 * So, e.g.:
3256 * - to create a new array, we send a VIRT_RECORD and
3257 * a VD_CONF. Then assemble and start the array.
3258 * - to activate a spare we send a VD_CONF to add the phys_refnum
3259 * and offset. This will also mark the spare as active with
3260 * a spare-assignment record.
3261 */
3262 struct ddf_super *ddf = st->sb;
3263 __u32 *magic = (__u32*)update->buf;
3264 struct phys_disk *pd;
3265 struct virtual_disk *vd;
3266 struct vd_config *vc;
3267 struct vcl *vcl;
3268 struct dl *dl;
3269 int mppe;
3270 int ent;
3271
2c514b71 3272 dprintf("Process update %x\n", *magic);
7e1432fb 3273
88c164f4
NB		 3274 switch (*magic) {
3275 case DDF_PHYS_RECORDS_MAGIC:
3276
3277 if (update->len != (sizeof(struct phys_disk) +
3278 sizeof(struct phys_disk_entry)))
3279 return;
3280 pd = (struct phys_disk*)update->buf;
3281
3282 ent = __be16_to_cpu(pd->used_pdes);
3283 if (ent >= __be16_to_cpu(ddf->phys->max_pdes))
3284 return;
3285 if (!all_ff(ddf->phys->entries[ent].guid))
3286 return;
3287 ddf->phys->entries[ent] = pd->entries[0];
3288 ddf->phys->used_pdes = __cpu_to_be16(1 +
3289 __be16_to_cpu(ddf->phys->used_pdes));
18a2f463 3290 ddf->updates_pending = 1;
2cc2983d
N		 3291 if (ddf->add_list) {
3292 struct active_array *a;
3293 struct dl *al = ddf->add_list;
3294 ddf->add_list = al->next;
3295
3296 al->next = ddf->dlist;
3297 ddf->dlist = al;
3298
3299 /* As a device has been added, we should check
3300 * for any degraded devices that might make
3301 * use of this spare */
3302 for (a = st->arrays ; a; a=a->next)
3303 a->check_degraded = 1;
3304 }
88c164f4
NB		 3305 break;
3306
3307 case DDF_VIRT_RECORDS_MAGIC:
3308
3309 if (update->len != (sizeof(struct virtual_disk) +
3310 sizeof(struct virtual_entry)))
3311 return;
3312 vd = (struct virtual_disk*)update->buf;
3313
3314 ent = __be16_to_cpu(vd->populated_vdes);
3315 if (ent >= __be16_to_cpu(ddf->virt->max_vdes))
3316 return;
3317 if (!all_ff(ddf->virt->entries[ent].guid))
3318 return;
3319 ddf->virt->entries[ent] = vd->entries[0];
3320 ddf->virt->populated_vdes = __cpu_to_be16(1 +
3321 __be16_to_cpu(ddf->virt->populated_vdes));
18a2f463 3322 ddf->updates_pending = 1;
88c164f4
NB		 3323 break;
3324
3325 case DDF_VD_CONF_MAGIC:
2c514b71 3326 dprintf("len %d %d\n", update->len, ddf->conf_rec_len);
88c164f4
NB		 3327
3328 mppe = __be16_to_cpu(ddf->anchor.max_primary_element_entries);
edd8d13c 3329 if (update->len != ddf->conf_rec_len * 512)
88c164f4
NB		 3330 return;
3331 vc = (struct vd_config*)update->buf;
3332 for (vcl = ddf->conflist; vcl ; vcl = vcl->next)
3333 if (memcmp(vcl->conf.guid, vc->guid, DDF_GUID_LEN) == 0)
3334 break;
2c514b71 3335 dprintf("vcl = %p\n", vcl);
88c164f4
NB		 3336 if (vcl) {
3337 /* An update, just copy the phys_refnum and lba_offset
3338 * fields
3339 */
3340 memcpy(vcl->conf.phys_refnum, vc->phys_refnum,
3341 mppe * (sizeof(__u32) + sizeof(__u64)));
3342 } else {
3343 /* A new VD_CONF */
e6b9548d
DW		 3344 if (!update->space)
3345 return;
88c164f4
NB		 3346 vcl = update->space;
3347 update->space = NULL;
3348 vcl->next = ddf->conflist;
edd8d13c 3349 memcpy(&vcl->conf, vc, update->len);
88c164f4
NB		 3350 vcl->lba_offset = (__u64*)
3351 &vcl->conf.phys_refnum[mppe];
3352 ddf->conflist = vcl;
3353 }
3354 /* Now make sure vlist is correct for each dl. */
3355 for (dl = ddf->dlist; dl; dl = dl->next) {
3356 int dn;
3357 int vn = 0;
3358 for (vcl = ddf->conflist; vcl ; vcl = vcl->next)
3359 for (dn=0; dn < ddf->mppe ; dn++)
3360 if (vcl->conf.phys_refnum[dn] ==
3361 dl->disk.refnum) {
2c514b71
NB		 3362 dprintf("dev %d has %p at %d\n",
3363 dl->pdnum, vcl, vn);
88c164f4
NB		 3364 dl->vlist[vn++] = vcl;
3365 break;
3366 }
3367 while (vn < ddf->max_part)
3368 dl->vlist[vn++] = NULL;
7e1432fb
NB		 3369 if (dl->vlist[0]) {
3370 ddf->phys->entries[dl->pdnum].type &=
3371 ~__cpu_to_be16(DDF_Global_Spare);
3372 ddf->phys->entries[dl->pdnum].type |=
3373 __cpu_to_be16(DDF_Active_in_VD);
3374 }
3375 if (dl->spare) {
3376 ddf->phys->entries[dl->pdnum].type &=
3377 ~__cpu_to_be16(DDF_Global_Spare);
3378 ddf->phys->entries[dl->pdnum].type |=
3379 __cpu_to_be16(DDF_Spare);
3380 }
3381 if (!dl->vlist[0] && !dl->spare) {
3382 ddf->phys->entries[dl->pdnum].type |=
3383 __cpu_to_be16(DDF_Global_Spare);
3384 ddf->phys->entries[dl->pdnum].type &=
3385 ~__cpu_to_be16(DDF_Spare |
3386 DDF_Active_in_VD);
3387 }
88c164f4 3388 }
18a2f463 3389 ddf->updates_pending = 1;
88c164f4
NB		 3390 break;
3391 case DDF_SPARE_ASSIGN_MAGIC:
3392 default: break;
3393 }
3394}
3395
edd8d13c
NB		 3396static void ddf_prepare_update(struct supertype *st,
3397 struct metadata_update *update)
3398{
3399 /* This update arrived at managemon.
3400 * We are about to pass it to monitor.
3401 * If a malloc is needed, do it here.
3402 */
3403 struct ddf_super *ddf = st->sb;
3404 __u32 *magic = (__u32*)update->buf;
3405 if (*magic == DDF_VD_CONF_MAGIC)
e6b9548d 3406 if (posix_memalign(&update->space, 512,
6416d527 3407 offsetof(struct vcl, conf)
e6b9548d
DW		 3408 + ddf->conf_rec_len * 512) != 0)
3409 update->space = NULL;
edd8d13c
NB		 3410}
3411
7e1432fb
NB		 3412/*
3413 * Check if the array 'a' is degraded but not failed.
3414 * If it is, find as many spares as are available and needed and
3415 * arrange for their inclusion.
3416 * We only choose devices which are not already in the array,
3417 * and prefer those with a spare-assignment to this array.
3418 * otherwise we choose global spares - assuming always that
3419 * there is enough room.
3420 * For each spare that we assign, we return an 'mdinfo' which
3421 * describes the position for the device in the array.
3422 * We also add to 'updates' a DDF_VD_CONF_MAGIC update with
3423 * the new phys_refnum and lba_offset values.
3424 *
3425 * Only worry about BVDs at the moment.
3426 */
3427static struct mdinfo *ddf_activate_spare(struct active_array *a,
3428 struct metadata_update **updates)
3429{
3430 int working = 0;
3431 struct mdinfo *d;
3432 struct ddf_super *ddf = a->container->sb;
3433 int global_ok = 0;
3434 struct mdinfo *rv = NULL;
3435 struct mdinfo *di;
3436 struct metadata_update *mu;
3437 struct dl *dl;
3438 int i;
3439 struct vd_config *vc;
3440 __u64 *lba;
3441
7e1432fb
NB		 3442 for (d = a->info.devs ; d ; d = d->next) {
3443 if ((d->curr_state & DS_FAULTY) &&
3444 d->state_fd >= 0)
3445 /* wait for Removal to happen */
3446 return NULL;
3447 if (d->state_fd >= 0)
3448 working ++;
3449 }
3450
2c514b71
NB		 3451 dprintf("ddf_activate: working=%d (%d) level=%d\n", working, a->info.array.raid_disks,
3452 a->info.array.level);
7e1432fb
NB		 3453 if (working == a->info.array.raid_disks)
3454 return NULL; /* array not degraded */
3455 switch (a->info.array.level) {
3456 case 1:
3457 if (working == 0)
3458 return NULL; /* failed */
3459 break;
3460 case 4:
3461 case 5:
3462 if (working < a->info.array.raid_disks - 1)
3463 return NULL; /* failed */
3464 break;
3465 case 6:
3466 if (working < a->info.array.raid_disks - 2)
3467 return NULL; /* failed */
3468 break;
3469 default: /* concat or stripe */
3470 return NULL; /* failed */
3471 }
3472
3473 /* For each slot, if it is not working, find a spare */
3474 dl = ddf->dlist;
3475 for (i = 0; i < a->info.array.raid_disks; i++) {
3476 for (d = a->info.devs ; d ; d = d->next)
3477 if (d->disk.raid_disk == i)
3478 break;
2c514b71 3479 dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0);
7e1432fb
NB		 3480 if (d && (d->state_fd >= 0))
3481 continue;
3482
3483 /* OK, this device needs recovery. Find a spare */
3484 again:
3485 for ( ; dl ; dl = dl->next) {
3486 unsigned long long esize;
3487 unsigned long long pos;
3488 struct mdinfo *d2;
3489 int is_global = 0;
3490 int is_dedicated = 0;
3491 struct extent *ex;
3492 int j;
3493 /* If in this array, skip */
3494 for (d2 = a->info.devs ; d2 ; d2 = d2->next)
3495 if (d2->disk.major == dl->major &&
3496 d2->disk.minor == dl->minor) {
2c514b71 3497 dprintf("%x:%x already in array\n", dl->major, dl->minor);
7e1432fb
NB		 3498 break;
3499 }
3500 if (d2)
3501 continue;
3502 if (ddf->phys->entries[dl->pdnum].type &
3503 __cpu_to_be16(DDF_Spare)) {
3504 /* Check spare assign record */
3505 if (dl->spare) {
3506 if (dl->spare->type & DDF_spare_dedicated) {
3507 /* check spare_ents for guid */
3508 for (j = 0 ;
3509 j < __be16_to_cpu(dl->spare->populated);
3510 j++) {
3511 if (memcmp(dl->spare->spare_ents[j].guid,
3512 ddf->virt->entries[a->info.container_member].guid,
3513 DDF_GUID_LEN) == 0)
3514 is_dedicated = 1;
3515 }
3516 } else
3517 is_global = 1;
3518 }
3519 } else if (ddf->phys->entries[dl->pdnum].type &
3520 __cpu_to_be16(DDF_Global_Spare)) {
3521 is_global = 1;
3522 }
3523 if ( ! (is_dedicated ||
3524 (is_global && global_ok))) {
2c514b71 3525 dprintf("%x:%x not suitable: %d %d\n", dl->major, dl->minor,
7e1432fb
NB		 3526 is_dedicated, is_global);
3527 continue;
3528 }
3529
3530 /* We are allowed to use this device - is there space?
3531 * We need a->info.component_size sectors */
3532 ex = get_extents(ddf, dl);
3533 if (!ex) {
2c514b71 3534 dprintf("cannot get extents\n");
7e1432fb
NB		 3535 continue;
3536 }
3537 j = 0; pos = 0;
3538 esize = 0;
3539
3540 do {
3541 esize = ex[j].start - pos;
3542 if (esize >= a->info.component_size)
3543 break;
3544 pos = ex[i].start + ex[i].size;
3545 i++;
3546 } while (ex[i-1].size);
3547
3548 free(ex);
3549 if (esize < a->info.component_size) {
2c514b71
NB		 3550 dprintf("%x:%x has no room: %llu %llu\n", dl->major, dl->minor,
3551 esize, a->info.component_size);
7e1432fb
NB		 3552 /* No room */
3553 continue;
3554 }
3555
3556 /* Cool, we have a device with some space at pos */
3557 di = malloc(sizeof(*di));
79244939
DW		 3558 if (!di)
3559 continue;
7e1432fb
NB		 3560 memset(di, 0, sizeof(*di));
3561 di->disk.number = i;
3562 di->disk.raid_disk = i;
3563 di->disk.major = dl->major;
3564 di->disk.minor = dl->minor;
3565 di->disk.state = 0;
d23534e4 3566 di->recovery_start = 0;
7e1432fb
NB		 3567 di->data_offset = pos;
3568 di->component_size = a->info.component_size;
3569 di->container_member = dl->pdnum;
3570 di->next = rv;
3571 rv = di;
2c514b71
NB		 3572 dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor,
3573 i, pos);
7e1432fb
NB		 3574
3575 break;
3576 }
3577 if (!dl && ! global_ok) {
3578 /* not enough dedicated spares, try global */
3579 global_ok = 1;
3580 dl = ddf->dlist;
3581 goto again;
3582 }
3583 }
3584
3585 if (!rv)
3586 /* No spares found */
3587 return rv;
3588 /* Now 'rv' has a list of devices to return.
3589 * Create a metadata_update record to update the
3590 * phys_refnum and lba_offset values
3591 */
904c1ef7 3592 mu = malloc(sizeof(*mu));
79244939
DW		 3593 if (mu && posix_memalign(&mu->space, 512, sizeof(struct vcl)) != 0) {
3594 free(mu);
3595 mu = NULL;
3596 }
3597 if (!mu) {
3598 while (rv) {
3599 struct mdinfo *n = rv->next;
3600
3601 free(rv);
3602 rv = n;
3603 }
3604 return NULL;
3605 }
3606
904c1ef7 3607 mu->buf = malloc(ddf->conf_rec_len * 512);
7e1432fb
NB		 3608 mu->len = ddf->conf_rec_len;
3609 mu->next = *updates;
3610 vc = find_vdcr(ddf, a->info.container_member);
3611 memcpy(mu->buf, vc, ddf->conf_rec_len * 512);
3612
3613 vc = (struct vd_config*)mu->buf;
3614 lba = (__u64*)&vc->phys_refnum[ddf->mppe];
3615 for (di = rv ; di ; di = di->next) {
3616 vc->phys_refnum[di->disk.raid_disk] =
3617 ddf->phys->entries[dl->pdnum].refnum;
3618 lba[di->disk.raid_disk] = di->data_offset;
3619 }
3620 *updates = mu;
3621 return rv;
3622}
0e600426 3623#endif /* MDASSEMBLE */
7e1432fb 3624
b640a252
N		 3625static int ddf_level_to_layout(int level)
3626{
3627 switch(level) {
3628 case 0:
3629 case 1:
3630 return 0;
3631 case 5:
3632 return ALGORITHM_LEFT_SYMMETRIC;
3633 case 6:
3634 return ALGORITHM_ROTATING_N_CONTINUE;
3635 case 10:
3636 return 0x102;
3637 default:
3638 return UnSet;
3639 }
3640}
3641
a322f70c
DW		 3642struct superswitch super_ddf = {
3643#ifndef MDASSEMBLE
3644 .examine_super = examine_super_ddf,
3645 .brief_examine_super = brief_examine_super_ddf,
4737ae25 3646 .brief_examine_subarrays = brief_examine_subarrays_ddf,
bceedeec 3647 .export_examine_super = export_examine_super_ddf,
a322f70c
DW		 3648 .detail_super = detail_super_ddf,
3649 .brief_detail_super = brief_detail_super_ddf,
3650 .validate_geometry = validate_geometry_ddf,
78e44928 3651 .write_init_super = write_init_super_ddf,
0e600426 3652 .add_to_super = add_to_super_ddf,
a322f70c
DW		 3653#endif
3654 .match_home = match_home_ddf,
3655 .uuid_from_super= uuid_from_super_ddf,
3656 .getinfo_super = getinfo_super_ddf,
3657 .update_super = update_super_ddf,
3658
3659 .avail_size = avail_size_ddf,
3660
a19c88b8
NB		 3661 .compare_super = compare_super_ddf,
3662
a322f70c 3663 .load_super = load_super_ddf,
ba7eb04f 3664 .init_super = init_super_ddf,
955e9ea1 3665 .store_super = store_super_ddf,
a322f70c
DW		 3666 .free_super = free_super_ddf,
3667 .match_metadata_desc = match_metadata_desc_ddf,
78e44928 3668 .container_content = container_content_ddf,
b640a252 3669 .default_layout = ddf_level_to_layout,
a322f70c 3670
a322f70c 3671 .external = 1,
549e9569 3672
0e600426 3673#ifndef MDASSEMBLE
549e9569
NB		 3674/* for mdmon */
3675 .open_new = ddf_open_new,
ed9d66aa 3676 .set_array_state= ddf_set_array_state,
549e9569
NB		 3677 .set_disk = ddf_set_disk,
3678 .sync_metadata = ddf_sync_metadata,
88c164f4 3679 .process_update = ddf_process_update,
edd8d13c 3680 .prepare_update = ddf_prepare_update,
7e1432fb 3681 .activate_spare = ddf_activate_spare,
0e600426 3682#endif
4cce4069 3683 .name = "ddf",
a322f70c 3684};
Unnamed repository; edit this file 'description' to name the repository.