]> 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
Create: user container_dev rather than subarray for some tests.
[thirdparty/mdadm.git] / super-ddf.c
CommitLineData
a322f70c
DW		 1/*
2 * mdadm - manage Linux "md" devices aka RAID arrays.
3 *
e736b623 4 * Copyright (C) 2006-2009 Neil Brown <neilb@suse.de>
a322f70c
DW		 5 *
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2 of the License, or
10 * (at your option) any later version.
11 *
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
16 *
17 * You should have received a copy of the GNU General Public License
18 * along with this program; if not, write to the Free Software
19 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20 *
21 * Author: Neil Brown
22 * Email: <neil@brown.name>
23 *
24 * Specifications for DDF takes from Common RAID DDF Specification Revision 1.2
25 * (July 28 2006). Reused by permission of SNIA.
26 */
27
28#define HAVE_STDINT_H 1
29#include "mdadm.h"
549e9569 30#include "mdmon.h"
a322f70c
DW		 31#include "sha1.h"
32#include <values.h>
33
a322f70c
DW		 34/* a non-official T10 name for creation GUIDs */
35static char T10[] = "Linux-MD";
36
37/* DDF timestamps are 1980 based, so we need to add
38 * second-in-decade-of-seventies to convert to linux timestamps.
39 * 10 years with 2 leap years.
40 */
41#define DECADE (3600*24*(365*10+2))
42unsigned long crc32(
43 unsigned long crc,
44 const unsigned char *buf,
45 unsigned len);
46
47/* The DDF metadata handling.
48 * DDF metadata lives at the end of the device.
49 * The last 512 byte block provides an 'anchor' which is used to locate
50 * the rest of the metadata which usually lives immediately behind the anchor.
51 *
52 * Note:
53 * - all multibyte numeric fields are bigendian.
54 * - all strings are space padded.
55 *
56 */
57
58/* Primary Raid Level (PRL) */
59#define DDF_RAID0 0x00
60#define DDF_RAID1 0x01
61#define DDF_RAID3 0x03
62#define DDF_RAID4 0x04
63#define DDF_RAID5 0x05
64#define DDF_RAID1E 0x11
65#define DDF_JBOD 0x0f
66#define DDF_CONCAT 0x1f
67#define DDF_RAID5E 0x15
68#define DDF_RAID5EE 0x25
59e36268 69#define DDF_RAID6 0x06
a322f70c
DW		 70
71/* Raid Level Qualifier (RLQ) */
72#define DDF_RAID0_SIMPLE 0x00
73#define DDF_RAID1_SIMPLE 0x00 /* just 2 devices in this plex */
74#define DDF_RAID1_MULTI 0x01 /* exactly 3 devices in this plex */
75#define DDF_RAID3_0 0x00 /* parity in first extent */
76#define DDF_RAID3_N 0x01 /* parity in last extent */
77#define DDF_RAID4_0 0x00 /* parity in first extent */
78#define DDF_RAID4_N 0x01 /* parity in last extent */
79/* these apply to raid5e and raid5ee as well */
80#define DDF_RAID5_0_RESTART 0x00 /* same as 'right asymmetric' - layout 1 */
59e36268 81#define DDF_RAID6_0_RESTART 0x01 /* raid6 different from raid5 here!!! */
a322f70c
DW		 82#define DDF_RAID5_N_RESTART 0x02 /* same as 'left asymmetric' - layout 0 */
83#define DDF_RAID5_N_CONTINUE 0x03 /* same as 'left symmetric' - layout 2 */
84
85#define DDF_RAID1E_ADJACENT 0x00 /* raid10 nearcopies==2 */
86#define DDF_RAID1E_OFFSET 0x01 /* raid10 offsetcopies==2 */
87
88/* Secondary RAID Level (SRL) */
89#define DDF_2STRIPED 0x00 /* This is weirder than RAID0 !! */
90#define DDF_2MIRRORED 0x01
91#define DDF_2CONCAT 0x02
92#define DDF_2SPANNED 0x03 /* This is also weird - be careful */
93
94/* Magic numbers */
95#define DDF_HEADER_MAGIC __cpu_to_be32(0xDE11DE11)
96#define DDF_CONTROLLER_MAGIC __cpu_to_be32(0xAD111111)
97#define DDF_PHYS_RECORDS_MAGIC __cpu_to_be32(0x22222222)
98#define DDF_PHYS_DATA_MAGIC __cpu_to_be32(0x33333333)
99#define DDF_VIRT_RECORDS_MAGIC __cpu_to_be32(0xDDDDDDDD)
100#define DDF_VD_CONF_MAGIC __cpu_to_be32(0xEEEEEEEE)
101#define DDF_SPARE_ASSIGN_MAGIC __cpu_to_be32(0x55555555)
102#define DDF_VU_CONF_MAGIC __cpu_to_be32(0x88888888)
103#define DDF_VENDOR_LOG_MAGIC __cpu_to_be32(0x01dBEEF0)
104#define DDF_BBM_LOG_MAGIC __cpu_to_be32(0xABADB10C)
105
106#define DDF_GUID_LEN 24
59e36268
NB		 107#define DDF_REVISION_0 "01.00.00"
108#define DDF_REVISION_2 "01.02.00"
a322f70c
DW		 109
110struct ddf_header {
88c164f4 111 __u32 magic; /* DDF_HEADER_MAGIC */
a322f70c
DW		 112 __u32 crc;
113 char guid[DDF_GUID_LEN];
59e36268 114 char revision[8]; /* 01.02.00 */
a322f70c
DW		 115 __u32 seq; /* starts at '1' */
116 __u32 timestamp;
117 __u8 openflag;
118 __u8 foreignflag;
119 __u8 enforcegroups;
120 __u8 pad0; /* 0xff */
121 __u8 pad1[12]; /* 12 * 0xff */
122 /* 64 bytes so far */
123 __u8 header_ext[32]; /* reserved: fill with 0xff */
124 __u64 primary_lba;
125 __u64 secondary_lba;
126 __u8 type;
127 __u8 pad2[3]; /* 0xff */
128 __u32 workspace_len; /* sectors for vendor space -
129 * at least 32768(sectors) */
130 __u64 workspace_lba;
131 __u16 max_pd_entries; /* one of 15, 63, 255, 1023, 4095 */
132 __u16 max_vd_entries; /* 2^(4,6,8,10,12)-1 : i.e. as above */
133 __u16 max_partitions; /* i.e. max num of configuration
134 record entries per disk */
135 __u16 config_record_len; /* 1 +ROUNDUP(max_primary_element_entries
136 *12/512) */
137 __u16 max_primary_element_entries; /* 16, 64, 256, 1024, or 4096 */
138 __u8 pad3[54]; /* 0xff */
139 /* 192 bytes so far */
140 __u32 controller_section_offset;
141 __u32 controller_section_length;
142 __u32 phys_section_offset;
143 __u32 phys_section_length;
144 __u32 virt_section_offset;
145 __u32 virt_section_length;
146 __u32 config_section_offset;
147 __u32 config_section_length;
148 __u32 data_section_offset;
149 __u32 data_section_length;
150 __u32 bbm_section_offset;
151 __u32 bbm_section_length;
152 __u32 diag_space_offset;
153 __u32 diag_space_length;
154 __u32 vendor_offset;
155 __u32 vendor_length;
156 /* 256 bytes so far */
157 __u8 pad4[256]; /* 0xff */
158};
159
160/* type field */
161#define DDF_HEADER_ANCHOR 0x00
162#define DDF_HEADER_PRIMARY 0x01
163#define DDF_HEADER_SECONDARY 0x02
164
165/* The content of the 'controller section' - global scope */
166struct ddf_controller_data {
88c164f4 167 __u32 magic; /* DDF_CONTROLLER_MAGIC */
a322f70c
DW		 168 __u32 crc;
169 char guid[DDF_GUID_LEN];
170 struct controller_type {
171 __u16 vendor_id;
172 __u16 device_id;
173 __u16 sub_vendor_id;
174 __u16 sub_device_id;
175 } type;
176 char product_id[16];
177 __u8 pad[8]; /* 0xff */
178 __u8 vendor_data[448];
179};
180
181/* The content of phys_section - global scope */
182struct phys_disk {
88c164f4 183 __u32 magic; /* DDF_PHYS_RECORDS_MAGIC */
a322f70c
DW		 184 __u32 crc;
185 __u16 used_pdes;
186 __u16 max_pdes;
187 __u8 pad[52];
188 struct phys_disk_entry {
189 char guid[DDF_GUID_LEN];
190 __u32 refnum;
191 __u16 type;
192 __u16 state;
193 __u64 config_size; /* DDF structures must be after here */
194 char path[18]; /* another horrible structure really */
195 __u8 pad[6];
196 } entries[0];
197};
198
199/* phys_disk_entry.type is a bitmap - bigendian remember */
200#define DDF_Forced_PD_GUID 1
201#define DDF_Active_in_VD 2
88c164f4 202#define DDF_Global_Spare 4 /* VD_CONF records are ignored */
a322f70c
DW		 203#define DDF_Spare 8 /* overrides Global_spare */
204#define DDF_Foreign 16
205#define DDF_Legacy 32 /* no DDF on this device */
206
207#define DDF_Interface_mask 0xf00
208#define DDF_Interface_SCSI 0x100
209#define DDF_Interface_SAS 0x200
210#define DDF_Interface_SATA 0x300
211#define DDF_Interface_FC 0x400
212
213/* phys_disk_entry.state is a bigendian bitmap */
214#define DDF_Online 1
215#define DDF_Failed 2 /* overrides 1,4,8 */
216#define DDF_Rebuilding 4
217#define DDF_Transition 8
218#define DDF_SMART 16
219#define DDF_ReadErrors 32
220#define DDF_Missing 64
221
222/* The content of the virt_section global scope */
223struct virtual_disk {
88c164f4 224 __u32 magic; /* DDF_VIRT_RECORDS_MAGIC */
a322f70c
DW		 225 __u32 crc;
226 __u16 populated_vdes;
227 __u16 max_vdes;
228 __u8 pad[52];
229 struct virtual_entry {
230 char guid[DDF_GUID_LEN];
231 __u16 unit;
232 __u16 pad0; /* 0xffff */
233 __u16 guid_crc;
234 __u16 type;
235 __u8 state;
236 __u8 init_state;
237 __u8 pad1[14];
238 char name[16];
239 } entries[0];
240};
241
242/* virtual_entry.type is a bitmap - bigendian */
243#define DDF_Shared 1
244#define DDF_Enforce_Groups 2
245#define DDF_Unicode 4
246#define DDF_Owner_Valid 8
247
248/* virtual_entry.state is a bigendian bitmap */
249#define DDF_state_mask 0x7
250#define DDF_state_optimal 0x0
251#define DDF_state_degraded 0x1
252#define DDF_state_deleted 0x2
253#define DDF_state_missing 0x3
254#define DDF_state_failed 0x4
7a7cc504 255#define DDF_state_part_optimal 0x5
a322f70c
DW		 256
257#define DDF_state_morphing 0x8
258#define DDF_state_inconsistent 0x10
259
260/* virtual_entry.init_state is a bigendian bitmap */
261#define DDF_initstate_mask 0x03
262#define DDF_init_not 0x00
7a7cc504
NB		 263#define DDF_init_quick 0x01 /* initialisation is progress.
264 * i.e. 'state_inconsistent' */
a322f70c
DW		 265#define DDF_init_full 0x02
266
267#define DDF_access_mask 0xc0
268#define DDF_access_rw 0x00
269#define DDF_access_ro 0x80
270#define DDF_access_blocked 0xc0
271
272/* The content of the config_section - local scope
273 * It has multiple records each config_record_len sectors
274 * They can be vd_config or spare_assign
275 */
276
277struct vd_config {
88c164f4 278 __u32 magic; /* DDF_VD_CONF_MAGIC */
a322f70c
DW		 279 __u32 crc;
280 char guid[DDF_GUID_LEN];
281 __u32 timestamp;
282 __u32 seqnum;
283 __u8 pad0[24];
284 __u16 prim_elmnt_count;
285 __u8 chunk_shift; /* 0 == 512, 1==1024 etc */
286 __u8 prl;
287 __u8 rlq;
288 __u8 sec_elmnt_count;
289 __u8 sec_elmnt_seq;
290 __u8 srl;
598f0d58
NB		 291 __u64 blocks; /* blocks per component could be different
292 * on different component devices...(only
293 * for concat I hope) */
294 __u64 array_blocks; /* blocks in array */
a322f70c
DW		 295 __u8 pad1[8];
296 __u32 spare_refs[8];
297 __u8 cache_pol[8];
298 __u8 bg_rate;
299 __u8 pad2[3];
300 __u8 pad3[52];
301 __u8 pad4[192];
302 __u8 v0[32]; /* reserved- 0xff */
303 __u8 v1[32]; /* reserved- 0xff */
304 __u8 v2[16]; /* reserved- 0xff */
305 __u8 v3[16]; /* reserved- 0xff */
306 __u8 vendor[32];
307 __u32 phys_refnum[0]; /* refnum of each disk in sequence */
308 /*__u64 lba_offset[0]; LBA offset in each phys. Note extents in a
309 bvd are always the same size */
310};
311
312/* vd_config.cache_pol[7] is a bitmap */
313#define DDF_cache_writeback 1 /* else writethrough */
314#define DDF_cache_wadaptive 2 /* only applies if writeback */
315#define DDF_cache_readahead 4
316#define DDF_cache_radaptive 8 /* only if doing read-ahead */
317#define DDF_cache_ifnobatt 16 /* even to write cache if battery is poor */
318#define DDF_cache_wallowed 32 /* enable write caching */
319#define DDF_cache_rallowed 64 /* enable read caching */
320
321struct spare_assign {
88c164f4 322 __u32 magic; /* DDF_SPARE_ASSIGN_MAGIC */
a322f70c
DW		 323 __u32 crc;
324 __u32 timestamp;
325 __u8 reserved[7];
326 __u8 type;
327 __u16 populated; /* SAEs used */
328 __u16 max; /* max SAEs */
329 __u8 pad[8];
330 struct spare_assign_entry {
331 char guid[DDF_GUID_LEN];
332 __u16 secondary_element;
333 __u8 pad[6];
334 } spare_ents[0];
335};
336/* spare_assign.type is a bitmap */
337#define DDF_spare_dedicated 0x1 /* else global */
338#define DDF_spare_revertible 0x2 /* else committable */
339#define DDF_spare_active 0x4 /* else not active */
340#define DDF_spare_affinity 0x8 /* enclosure affinity */
341
342/* The data_section contents - local scope */
343struct disk_data {
88c164f4 344 __u32 magic; /* DDF_PHYS_DATA_MAGIC */
a322f70c
DW		 345 __u32 crc;
346 char guid[DDF_GUID_LEN];
347 __u32 refnum; /* crc of some magic drive data ... */
348 __u8 forced_ref; /* set when above was not result of magic */
349 __u8 forced_guid; /* set if guid was forced rather than magic */
350 __u8 vendor[32];
351 __u8 pad[442];
352};
353
354/* bbm_section content */
355struct bad_block_log {
356 __u32 magic;
357 __u32 crc;
358 __u16 entry_count;
359 __u32 spare_count;
360 __u8 pad[10];
361 __u64 first_spare;
362 struct mapped_block {
363 __u64 defective_start;
364 __u32 replacement_start;
365 __u16 remap_count;
366 __u8 pad[2];
367 } entries[0];
368};
369
370/* Struct for internally holding ddf structures */
371/* The DDF structure stored on each device is potentially
372 * quite different, as some data is global and some is local.
373 * The global data is:
374 * - ddf header
375 * - controller_data
376 * - Physical disk records
377 * - Virtual disk records
378 * The local data is:
379 * - Configuration records
380 * - Physical Disk data section
381 * ( and Bad block and vendor which I don't care about yet).
382 *
383 * The local data is parsed into separate lists as it is read
384 * and reconstructed for writing. This means that we only need
385 * to make config changes once and they are automatically
386 * propagated to all devices.
387 * Note that the ddf_super has space of the conf and disk data
388 * for this disk and also for a list of all such data.
389 * The list is only used for the superblock that is being
390 * built in Create or Assemble to describe the whole array.
391 */
392struct ddf_super {
6416d527 393 struct ddf_header anchor, primary, secondary;
a322f70c 394 struct ddf_controller_data controller;
6416d527 395 struct ddf_header *active;
a322f70c
DW		 396 struct phys_disk *phys;
397 struct virtual_disk *virt;
398 int pdsize, vdsize;
f21e18ca 399 unsigned int max_part, mppe, conf_rec_len;
d2ca6449 400 int currentdev;
18a2f463 401 int updates_pending;
a322f70c 402 struct vcl {
6416d527
NB		 403 union {
404 char space[512];
405 struct {
406 struct vcl *next;
407 __u64 *lba_offset; /* location in 'conf' of
408 * the lba table */
f21e18ca 409 unsigned int vcnum; /* index into ->virt */
6416d527
NB		 410 __u64 *block_sizes; /* NULL if all the same */
411 };
412 };
a322f70c 413 struct vd_config conf;
d2ca6449 414 } *conflist, *currentconf;
a322f70c 415 struct dl {
6416d527
NB		 416 union {
417 char space[512];
418 struct {
419 struct dl *next;
420 int major, minor;
421 char *devname;
422 int fd;
423 unsigned long long size; /* sectors */
424 int pdnum; /* index in ->phys */
425 struct spare_assign *spare;
8592f29d
N		 426 void *mdupdate; /* hold metadata update */
427
428 /* These fields used by auto-layout */
429 int raiddisk; /* slot to fill in autolayout */
430 __u64 esize;
6416d527
NB		 431 };
432 };
a322f70c 433 struct disk_data disk;
b2280677 434 struct vcl *vlist[0]; /* max_part in size */
2cc2983d 435 } *dlist, *add_list;
a322f70c
DW		 436};
437
438#ifndef offsetof
439#define offsetof(t,f) ((size_t)&(((t*)0)->f))
440#endif
441
a322f70c 442
f21e18ca 443static unsigned int calc_crc(void *buf, int len)
a322f70c
DW		 444{
445 /* crcs are always at the same place as in the ddf_header */
446 struct ddf_header *ddf = buf;
447 __u32 oldcrc = ddf->crc;
448 __u32 newcrc;
449 ddf->crc = 0xffffffff;
450
451 newcrc = crc32(0, buf, len);
452 ddf->crc = oldcrc;
4abe6b70
N		 453 /* The crc is store (like everything) bigendian, so convert
454 * here for simplicity
455 */
456 return __cpu_to_be32(newcrc);
a322f70c
DW		 457}
458
459static int load_ddf_header(int fd, unsigned long long lba,
460 unsigned long long size,
461 int type,
462 struct ddf_header *hdr, struct ddf_header *anchor)
463{
464 /* read a ddf header (primary or secondary) from fd/lba
465 * and check that it is consistent with anchor
466 * Need to check:
467 * magic, crc, guid, rev, and LBA's header_type, and
468 * everything after header_type must be the same
469 */
470 if (lba >= size-1)
471 return 0;
472
473 if (lseek64(fd, lba<<9, 0) < 0)
474 return 0;
475
476 if (read(fd, hdr, 512) != 512)
477 return 0;
478
479 if (hdr->magic != DDF_HEADER_MAGIC)
480 return 0;
481 if (calc_crc(hdr, 512) != hdr->crc)
482 return 0;
483 if (memcmp(anchor->guid, hdr->guid, DDF_GUID_LEN) != 0 ||
484 memcmp(anchor->revision, hdr->revision, 8) != 0 ||
485 anchor->primary_lba != hdr->primary_lba ||
486 anchor->secondary_lba != hdr->secondary_lba ||
487 hdr->type != type ||
488 memcmp(anchor->pad2, hdr->pad2, 512 -
489 offsetof(struct ddf_header, pad2)) != 0)
490 return 0;
491
492 /* Looks good enough to me... */
493 return 1;
494}
495
496static void *load_section(int fd, struct ddf_super *super, void *buf,
497 __u32 offset_be, __u32 len_be, int check)
498{
499 unsigned long long offset = __be32_to_cpu(offset_be);
500 unsigned long long len = __be32_to_cpu(len_be);
501 int dofree = (buf == NULL);
502
503 if (check)
504 if (len != 2 && len != 8 && len != 32
505 && len != 128 && len != 512)
506 return NULL;
507
508 if (len > 1024)
509 return NULL;
510 if (buf) {
511 /* All pre-allocated sections are a single block */
512 if (len != 1)
513 return NULL;
3d2c4fc7
DW		 514 } else if (posix_memalign(&buf, 512, len<<9) != 0)
515 buf = NULL;
6416d527 516
a322f70c
DW		 517 if (!buf)
518 return NULL;
519
520 if (super->active->type == 1)
521 offset += __be64_to_cpu(super->active->primary_lba);
522 else
523 offset += __be64_to_cpu(super->active->secondary_lba);
524
f21e18ca 525 if ((unsigned long long)lseek64(fd, offset<<9, 0) != (offset<<9)) {
a322f70c
DW		 526 if (dofree)
527 free(buf);
528 return NULL;
529 }
f21e18ca 530 if ((unsigned long long)read(fd, buf, len<<9) != (len<<9)) {
a322f70c
DW		 531 if (dofree)
532 free(buf);
533 return NULL;
534 }
535 return buf;
536}
537
538static int load_ddf_headers(int fd, struct ddf_super *super, char *devname)
539{
540 unsigned long long dsize;
541
542 get_dev_size(fd, NULL, &dsize);
543
544 if (lseek64(fd, dsize-512, 0) < 0) {
545 if (devname)
546 fprintf(stderr,
547 Name": Cannot seek to anchor block on %s: %s\n",
548 devname, strerror(errno));
549 return 1;
550 }
551 if (read(fd, &super->anchor, 512) != 512) {
552 if (devname)
553 fprintf(stderr,
554 Name ": Cannot read anchor block on %s: %s\n",
555 devname, strerror(errno));
556 return 1;
557 }
558 if (super->anchor.magic != DDF_HEADER_MAGIC) {
559 if (devname)
560 fprintf(stderr, Name ": no DDF anchor found on %s\n",
561 devname);
562 return 2;
563 }
564 if (calc_crc(&super->anchor, 512) != super->anchor.crc) {
565 if (devname)
566 fprintf(stderr, Name ": bad CRC on anchor on %s\n",
567 devname);
568 return 2;
569 }
59e36268
NB		 570 if (memcmp(super->anchor.revision, DDF_REVISION_0, 8) != 0 &&
571 memcmp(super->anchor.revision, DDF_REVISION_2, 8) != 0) {
a322f70c
DW		 572 if (devname)
573 fprintf(stderr, Name ": can only support super revision"
59e36268
NB		 574 " %.8s and earlier, not %.8s on %s\n",
575 DDF_REVISION_2, super->anchor.revision,devname);
a322f70c
DW		 576 return 2;
577 }
578 if (load_ddf_header(fd, __be64_to_cpu(super->anchor.primary_lba),
579 dsize >> 9, 1,
580 &super->primary, &super->anchor) == 0) {
581 if (devname)
582 fprintf(stderr,
583 Name ": Failed to load primary DDF header "
584 "on %s\n", devname);
585 return 2;
586 }
587 super->active = &super->primary;
588 if (load_ddf_header(fd, __be64_to_cpu(super->anchor.secondary_lba),
589 dsize >> 9, 2,
590 &super->secondary, &super->anchor)) {
591 if ((__be32_to_cpu(super->primary.seq)
592 < __be32_to_cpu(super->secondary.seq) &&
593 !super->secondary.openflag)
594 || (__be32_to_cpu(super->primary.seq)
595 == __be32_to_cpu(super->secondary.seq) &&
596 super->primary.openflag && !super->secondary.openflag)
597 )
598 super->active = &super->secondary;
599 }
600 return 0;
601}
602
603static int load_ddf_global(int fd, struct ddf_super *super, char *devname)
604{
605 void *ok;
606 ok = load_section(fd, super, &super->controller,
607 super->active->controller_section_offset,
608 super->active->controller_section_length,
609 0);
610 super->phys = load_section(fd, super, NULL,
611 super->active->phys_section_offset,
612 super->active->phys_section_length,
613 1);
614 super->pdsize = __be32_to_cpu(super->active->phys_section_length) * 512;
615
616 super->virt = load_section(fd, super, NULL,
617 super->active->virt_section_offset,
618 super->active->virt_section_length,
619 1);
620 super->vdsize = __be32_to_cpu(super->active->virt_section_length) * 512;
621 if (!ok ||
622 !super->phys ||
623 !super->virt) {
624 free(super->phys);
625 free(super->virt);
a2349791
NB		 626 super->phys = NULL;
627 super->virt = NULL;
a322f70c
DW		 628 return 2;
629 }
630 super->conflist = NULL;
631 super->dlist = NULL;
8c3b8c2c
NB		 632
633 super->max_part = __be16_to_cpu(super->active->max_partitions);
634 super->mppe = __be16_to_cpu(super->active->max_primary_element_entries);
635 super->conf_rec_len = __be16_to_cpu(super->active->config_record_len);
a322f70c
DW		 636 return 0;
637}
638
639static int load_ddf_local(int fd, struct ddf_super *super,
640 char *devname, int keep)
641{
642 struct dl *dl;
643 struct stat stb;
644 char *conf;
f21e18ca
N		 645 unsigned int i;
646 unsigned int confsec;
b2280677 647 int vnum;
f21e18ca 648 unsigned int max_virt_disks = __be16_to_cpu(super->active->max_vd_entries);
d2ca6449 649 unsigned long long dsize;
a322f70c
DW		 650
651 /* First the local disk info */
3d2c4fc7 652 if (posix_memalign((void**)&dl, 512,
6416d527 653 sizeof(*dl) +
3d2c4fc7
DW		 654 (super->max_part) * sizeof(dl->vlist[0])) != 0) {
655 fprintf(stderr, Name ": %s could not allocate disk info buffer\n",
656 __func__);
657 return 1;
658 }
a322f70c
DW		 659
660 load_section(fd, super, &dl->disk,
661 super->active->data_section_offset,
662 super->active->data_section_length,
663 0);
664 dl->devname = devname ? strdup(devname) : NULL;
598f0d58 665
a322f70c
DW		 666 fstat(fd, &stb);
667 dl->major = major(stb.st_rdev);
668 dl->minor = minor(stb.st_rdev);
669 dl->next = super->dlist;
670 dl->fd = keep ? fd : -1;
d2ca6449
NB		 671
672 dl->size = 0;
673 if (get_dev_size(fd, devname, &dsize))
674 dl->size = dsize >> 9;
b2280677 675 dl->spare = NULL;
f21e18ca 676 for (i = 0 ; i < super->max_part ; i++)
a322f70c
DW		 677 dl->vlist[i] = NULL;
678 super->dlist = dl;
59e36268 679 dl->pdnum = -1;
f21e18ca 680 for (i = 0; i < __be16_to_cpu(super->active->max_pd_entries); i++)
5575e7d9
NB		 681 if (memcmp(super->phys->entries[i].guid,
682 dl->disk.guid, DDF_GUID_LEN) == 0)
683 dl->pdnum = i;
684
a322f70c
DW		 685 /* Now the config list. */
686 /* 'conf' is an array of config entries, some of which are
687 * probably invalid. Those which are good need to be copied into
688 * the conflist
689 */
a322f70c
DW		 690
691 conf = load_section(fd, super, NULL,
692 super->active->config_section_offset,
693 super->active->config_section_length,
694 0);
695
b2280677 696 vnum = 0;
e223334f
N		 697 for (confsec = 0;
698 confsec < __be32_to_cpu(super->active->config_section_length);
699 confsec += super->conf_rec_len) {
a322f70c 700 struct vd_config *vd =
e223334f 701 (struct vd_config *)((char*)conf + confsec*512);
a322f70c
DW		 702 struct vcl *vcl;
703
b2280677
NB		 704 if (vd->magic == DDF_SPARE_ASSIGN_MAGIC) {
705 if (dl->spare)
706 continue;
3d2c4fc7
DW		 707 if (posix_memalign((void**)&dl->spare, 512,
708 super->conf_rec_len*512) != 0) {
709 fprintf(stderr, Name
710 ": %s could not allocate spare info buf\n",
711 __func__);
712 return 1;
713 }
714
b2280677
NB		 715 memcpy(dl->spare, vd, super->conf_rec_len*512);
716 continue;
717 }
a322f70c
DW		 718 if (vd->magic != DDF_VD_CONF_MAGIC)
719 continue;
720 for (vcl = super->conflist; vcl; vcl = vcl->next) {
721 if (memcmp(vcl->conf.guid,
722 vd->guid, DDF_GUID_LEN) == 0)
723 break;
724 }
725
726 if (vcl) {
b2280677 727 dl->vlist[vnum++] = vcl;
a322f70c
DW		 728 if (__be32_to_cpu(vd->seqnum) <=
729 __be32_to_cpu(vcl->conf.seqnum))
730 continue;
59e36268 731 } else {
3d2c4fc7 732 if (posix_memalign((void**)&vcl, 512,
6416d527 733 (super->conf_rec_len*512 +
3d2c4fc7
DW		 734 offsetof(struct vcl, conf))) != 0) {
735 fprintf(stderr, Name
736 ": %s could not allocate vcl buf\n",
737 __func__);
738 return 1;
739 }
a322f70c 740 vcl->next = super->conflist;
59e36268 741 vcl->block_sizes = NULL; /* FIXME not for CONCAT */
a322f70c 742 super->conflist = vcl;
b2280677 743 dl->vlist[vnum++] = vcl;
a322f70c 744 }
8c3b8c2c 745 memcpy(&vcl->conf, vd, super->conf_rec_len*512);
a322f70c 746 vcl->lba_offset = (__u64*)
8c3b8c2c 747 &vcl->conf.phys_refnum[super->mppe];
59e36268
NB		 748
749 for (i=0; i < max_virt_disks ; i++)
750 if (memcmp(super->virt->entries[i].guid,
751 vcl->conf.guid, DDF_GUID_LEN)==0)
752 break;
753 if (i < max_virt_disks)
754 vcl->vcnum = i;
a322f70c
DW		 755 }
756 free(conf);
757
758 return 0;
759}
760
761#ifndef MDASSEMBLE
762static int load_super_ddf_all(struct supertype *st, int fd,
763 void **sbp, char *devname, int keep_fd);
764#endif
37424f13
DW		 765
766static void free_super_ddf(struct supertype *st);
767
a322f70c
DW		 768static int load_super_ddf(struct supertype *st, int fd,
769 char *devname)
770{
771 unsigned long long dsize;
772 struct ddf_super *super;
773 int rv;
774
775#ifndef MDASSEMBLE
59e36268 776 /* if 'fd' is a container, load metadata from all the devices */
3dbccbcf 777 if (load_super_ddf_all(st, fd, &st->sb, devname, 1) == 0)
a322f70c
DW		 778 return 0;
779#endif
780
781 if (get_dev_size(fd, devname, &dsize) == 0)
782 return 1;
783
691c6ee1
N		 784 if (test_partition(fd))
785 /* DDF is not allowed on partitions */
786 return 1;
787
a322f70c
DW		 788 /* 32M is a lower bound */
789 if (dsize <= 32*1024*1024) {
97320d7c 790 if (devname)
a322f70c
DW		 791 fprintf(stderr,
792 Name ": %s is too small for ddf: "
793 "size is %llu sectors.\n",
794 devname, dsize>>9);
97320d7c 795 return 1;
a322f70c
DW		 796 }
797 if (dsize & 511) {
97320d7c 798 if (devname)
a322f70c
DW		 799 fprintf(stderr,
800 Name ": %s is an odd size for ddf: "
801 "size is %llu bytes.\n",
802 devname, dsize);
97320d7c 803 return 1;
a322f70c
DW		 804 }
805
37424f13
DW		 806 free_super_ddf(st);
807
6416d527 808 if (posix_memalign((void**)&super, 512, sizeof(*super))!= 0) {
a322f70c
DW		 809 fprintf(stderr, Name ": malloc of %zu failed.\n",
810 sizeof(*super));
811 return 1;
812 }
a2349791 813 memset(super, 0, sizeof(*super));
a322f70c
DW		 814
815 rv = load_ddf_headers(fd, super, devname);
816 if (rv) {
817 free(super);
818 return rv;
819 }
820
821 /* Have valid headers and have chosen the best. Let's read in the rest*/
822
823 rv = load_ddf_global(fd, super, devname);
824
825 if (rv) {
826 if (devname)
827 fprintf(stderr,
828 Name ": Failed to load all information "
829 "sections on %s\n", devname);
830 free(super);
831 return rv;
832 }
833
3d2c4fc7
DW		 834 rv = load_ddf_local(fd, super, devname, 0);
835
836 if (rv) {
837 if (devname)
838 fprintf(stderr,
839 Name ": Failed to load all information "
840 "sections on %s\n", devname);
841 free(super);
842 return rv;
843 }
a322f70c
DW		 844
845 /* Should possibly check the sections .... */
846
847 st->sb = super;
848 if (st->ss == NULL) {
849 st->ss = &super_ddf;
850 st->minor_version = 0;
851 st->max_devs = 512;
852 }
352452c3 853 st->loaded_container = 0;
a322f70c
DW		 854 return 0;
855
856}
857
858static void free_super_ddf(struct supertype *st)
859{
860 struct ddf_super *ddf = st->sb;
861 if (ddf == NULL)
862 return;
863 free(ddf->phys);
864 free(ddf->virt);
865 while (ddf->conflist) {
866 struct vcl *v = ddf->conflist;
867 ddf->conflist = v->next;
59e36268
NB		 868 if (v->block_sizes)
869 free(v->block_sizes);
a322f70c
DW		 870 free(v);
871 }
872 while (ddf->dlist) {
873 struct dl *d = ddf->dlist;
874 ddf->dlist = d->next;
875 if (d->fd >= 0)
876 close(d->fd);
b2280677
NB		 877 if (d->spare)
878 free(d->spare);
a322f70c
DW		 879 free(d);
880 }
881 free(ddf);
882 st->sb = NULL;
883}
884
885static struct supertype *match_metadata_desc_ddf(char *arg)
886{
887 /* 'ddf' only support containers */
888 struct supertype *st;
889 if (strcmp(arg, "ddf") != 0 &&
890 strcmp(arg, "default") != 0
891 )
892 return NULL;
893
894 st = malloc(sizeof(*st));
ef609477 895 memset(st, 0, sizeof(*st));
d1d599ea 896 st->container_dev = NoMdDev;
a322f70c
DW		 897 st->ss = &super_ddf;
898 st->max_devs = 512;
899 st->minor_version = 0;
900 st->sb = NULL;
901 return st;
902}
903
a322f70c
DW		 904
905#ifndef MDASSEMBLE
906
907static mapping_t ddf_state[] = {
908 { "Optimal", 0},
909 { "Degraded", 1},
910 { "Deleted", 2},
911 { "Missing", 3},
912 { "Failed", 4},
913 { "Partially Optimal", 5},
914 { "-reserved-", 6},
915 { "-reserved-", 7},
916 { NULL, 0}
917};
918
919static mapping_t ddf_init_state[] = {
920 { "Not Initialised", 0},
921 { "QuickInit in Progress", 1},
922 { "Fully Initialised", 2},
923 { "*UNKNOWN*", 3},
924 { NULL, 0}
925};
926static mapping_t ddf_access[] = {
927 { "Read/Write", 0},
928 { "Reserved", 1},
929 { "Read Only", 2},
930 { "Blocked (no access)", 3},
931 { NULL ,0}
932};
933
934static mapping_t ddf_level[] = {
935 { "RAID0", DDF_RAID0},
936 { "RAID1", DDF_RAID1},
937 { "RAID3", DDF_RAID3},
938 { "RAID4", DDF_RAID4},
939 { "RAID5", DDF_RAID5},
940 { "RAID1E",DDF_RAID1E},
941 { "JBOD", DDF_JBOD},
942 { "CONCAT",DDF_CONCAT},
943 { "RAID5E",DDF_RAID5E},
944 { "RAID5EE",DDF_RAID5EE},
945 { "RAID6", DDF_RAID6},
946 { NULL, 0}
947};
948static mapping_t ddf_sec_level[] = {
949 { "Striped", DDF_2STRIPED},
950 { "Mirrored", DDF_2MIRRORED},
951 { "Concat", DDF_2CONCAT},
952 { "Spanned", DDF_2SPANNED},
953 { NULL, 0}
954};
955#endif
956
957struct num_mapping {
958 int num1, num2;
959};
960static struct num_mapping ddf_level_num[] = {
961 { DDF_RAID0, 0 },
962 { DDF_RAID1, 1 },
963 { DDF_RAID3, LEVEL_UNSUPPORTED },
60f18132
NB		 964 { DDF_RAID4, 4 },
965 { DDF_RAID5, 5 },
a322f70c
DW		 966 { DDF_RAID1E, LEVEL_UNSUPPORTED },
967 { DDF_JBOD, LEVEL_UNSUPPORTED },
968 { DDF_CONCAT, LEVEL_LINEAR },
969 { DDF_RAID5E, LEVEL_UNSUPPORTED },
970 { DDF_RAID5EE, LEVEL_UNSUPPORTED },
971 { DDF_RAID6, 6},
972 { MAXINT, MAXINT }
973};
974
975static int map_num1(struct num_mapping *map, int num)
976{
977 int i;
978 for (i=0 ; map[i].num1 != MAXINT; i++)
979 if (map[i].num1 == num)
980 break;
981 return map[i].num2;
982}
983
42dc2744
N		 984static int all_ff(char *guid)
985{
986 int i;
987 for (i = 0; i < DDF_GUID_LEN; i++)
988 if (guid[i] != (char)0xff)
989 return 0;
990 return 1;
991}
992
a322f70c
DW		 993#ifndef MDASSEMBLE
994static void print_guid(char *guid, int tstamp)
995{
996 /* A GUIDs are part (or all) ASCII and part binary.
997 * They tend to be space padded.
59e36268
NB		 998 * We print the GUID in HEX, then in parentheses add
999 * any initial ASCII sequence, and a possible
1000 * time stamp from bytes 16-19
a322f70c
DW		 1001 */
1002 int l = DDF_GUID_LEN;
1003 int i;
59e36268
NB		 1004
1005 for (i=0 ; i<DDF_GUID_LEN ; i++) {
1006 if ((i&3)==0 && i != 0) printf(":");
1007 printf("%02X", guid[i]&255);
1008 }
1009
cfccea8c 1010 printf("\n (");
a322f70c
DW		 1011 while (l && guid[l-1] == ' ')
1012 l--;
1013 for (i=0 ; i<l ; i++) {
1014 if (guid[i] >= 0x20 && guid[i] < 0x7f)
1015 fputc(guid[i], stdout);
1016 else
59e36268 1017 break;
a322f70c
DW		 1018 }
1019 if (tstamp) {
1020 time_t then = __be32_to_cpu(*(__u32*)(guid+16)) + DECADE;
1021 char tbuf[100];
1022 struct tm *tm;
1023 tm = localtime(&then);
59e36268 1024 strftime(tbuf, 100, " %D %T",tm);
a322f70c
DW		 1025 fputs(tbuf, stdout);
1026 }
59e36268 1027 printf(")");
a322f70c
DW		 1028}
1029
1030static void examine_vd(int n, struct ddf_super *sb, char *guid)
1031{
8c3b8c2c 1032 int crl = sb->conf_rec_len;
a322f70c
DW		 1033 struct vcl *vcl;
1034
1035 for (vcl = sb->conflist ; vcl ; vcl = vcl->next) {
f21e18ca 1036 unsigned int i;
a322f70c
DW		 1037 struct vd_config *vc = &vcl->conf;
1038
1039 if (calc_crc(vc, crl*512) != vc->crc)
1040 continue;
1041 if (memcmp(vc->guid, guid, DDF_GUID_LEN) != 0)
1042 continue;
1043
1044 /* Ok, we know about this VD, let's give more details */
b06e3095 1045 printf(" Raid Devices[%d] : %d (", n,
a322f70c 1046 __be16_to_cpu(vc->prim_elmnt_count));
f21e18ca 1047 for (i = 0; i < __be16_to_cpu(vc->prim_elmnt_count); i++) {
b06e3095
N		 1048 int j;
1049 int cnt = __be16_to_cpu(sb->phys->used_pdes);
1050 for (j=0; j<cnt; j++)
1051 if (vc->phys_refnum[i] == sb->phys->entries[j].refnum)
1052 break;
1053 if (i) printf(" ");
1054 if (j < cnt)
1055 printf("%d", j);
1056 else
1057 printf("--");
1058 }
1059 printf(")\n");
1060 if (vc->chunk_shift != 255)
a322f70c
DW		 1061 printf(" Chunk Size[%d] : %d sectors\n", n,
1062 1 << vc->chunk_shift);
1063 printf(" Raid Level[%d] : %s\n", n,
1064 map_num(ddf_level, vc->prl)?:"-unknown-");
1065 if (vc->sec_elmnt_count != 1) {
1066 printf(" Secondary Position[%d] : %d of %d\n", n,
1067 vc->sec_elmnt_seq, vc->sec_elmnt_count);
1068 printf(" Secondary Level[%d] : %s\n", n,
1069 map_num(ddf_sec_level, vc->srl) ?: "-unknown-");
1070 }
1071 printf(" Device Size[%d] : %llu\n", n,
c9b6907b 1072 (unsigned long long)__be64_to_cpu(vc->blocks)/2);
a322f70c 1073 printf(" Array Size[%d] : %llu\n", n,
c9b6907b 1074 (unsigned long long)__be64_to_cpu(vc->array_blocks)/2);
a322f70c
DW		 1075 }
1076}
1077
1078static void examine_vds(struct ddf_super *sb)
1079{
1080 int cnt = __be16_to_cpu(sb->virt->populated_vdes);
1081 int i;
1082 printf(" Virtual Disks : %d\n", cnt);
1083
1084 for (i=0; i<cnt; i++) {
1085 struct virtual_entry *ve = &sb->virt->entries[i];
b06e3095 1086 printf("\n");
a322f70c
DW		 1087 printf(" VD GUID[%d] : ", i); print_guid(ve->guid, 1);
1088 printf("\n");
1089 printf(" unit[%d] : %d\n", i, __be16_to_cpu(ve->unit));
1090 printf(" state[%d] : %s, %s%s\n", i,
1091 map_num(ddf_state, ve->state & 7),
1092 (ve->state & 8) ? "Morphing, ": "",
1093 (ve->state & 16)? "Not Consistent" : "Consistent");
1094 printf(" init state[%d] : %s\n", i,
1095 map_num(ddf_init_state, ve->init_state&3));
1096 printf(" access[%d] : %s\n", i,
1097 map_num(ddf_access, (ve->init_state>>6) & 3));
1098 printf(" Name[%d] : %.16s\n", i, ve->name);
1099 examine_vd(i, sb, ve->guid);
1100 }
1101 if (cnt) printf("\n");
1102}
1103
1104static void examine_pds(struct ddf_super *sb)
1105{
1106 int cnt = __be16_to_cpu(sb->phys->used_pdes);
1107 int i;
1108 struct dl *dl;
1109 printf(" Physical Disks : %d\n", cnt);
962371a5 1110 printf(" Number RefNo Size Device Type/State\n");
a322f70c
DW		 1111
1112 for (i=0 ; i<cnt ; i++) {
1113 struct phys_disk_entry *pd = &sb->phys->entries[i];
1114 int type = __be16_to_cpu(pd->type);
1115 int state = __be16_to_cpu(pd->state);
1116
b06e3095
N		 1117 //printf(" PD GUID[%d] : ", i); print_guid(pd->guid, 0);
1118 //printf("\n");
1119 printf(" %3d %08x ", i,
a322f70c 1120 __be32_to_cpu(pd->refnum));
c9b6907b
DL		 1121 printf("%8lluK ",
1122 (unsigned long long)__be64_to_cpu(pd->config_size)>>1);
b06e3095
N		 1123 for (dl = sb->dlist; dl ; dl = dl->next) {
1124 if (dl->disk.refnum == pd->refnum) {
1125 char *dv = map_dev(dl->major, dl->minor, 0);
1126 if (dv) {
962371a5 1127 printf("%-15s", dv);
b06e3095
N		 1128 break;
1129 }
1130 }
1131 }
1132 if (!dl)
962371a5 1133 printf("%15s","");
b06e3095 1134 printf(" %s%s%s%s%s",
a322f70c 1135 (type&2) ? "active":"",
b06e3095 1136 (type&4) ? "Global-Spare":"",
a322f70c
DW		 1137 (type&8) ? "spare" : "",
1138 (type&16)? ", foreign" : "",
1139 (type&32)? "pass-through" : "");
b06e3095 1140 printf("/%s%s%s%s%s%s%s",
a322f70c
DW		 1141 (state&1)? "Online": "Offline",
1142 (state&2)? ", Failed": "",
1143 (state&4)? ", Rebuilding": "",
1144 (state&8)? ", in-transition": "",
b06e3095
N		 1145 (state&16)? ", SMART-errors": "",
1146 (state&32)? ", Unrecovered-Read-Errors": "",
a322f70c 1147 (state&64)? ", Missing" : "");
a322f70c
DW		 1148 printf("\n");
1149 }
1150}
1151
1152static void examine_super_ddf(struct supertype *st, char *homehost)
1153{
1154 struct ddf_super *sb = st->sb;
1155
1156 printf(" Magic : %08x\n", __be32_to_cpu(sb->anchor.magic));
1157 printf(" Version : %.8s\n", sb->anchor.revision);
598f0d58
NB		 1158 printf("Controller GUID : "); print_guid(sb->controller.guid, 0);
1159 printf("\n");
1160 printf(" Container GUID : "); print_guid(sb->anchor.guid, 1);
a322f70c
DW		 1161 printf("\n");
1162 printf(" Seq : %08x\n", __be32_to_cpu(sb->active->seq));
1163 printf(" Redundant hdr : %s\n", sb->secondary.magic == DDF_HEADER_MAGIC
1164 ?"yes" : "no");
1165 examine_vds(sb);
1166 examine_pds(sb);
1167}
1168
a5d85af7 1169static void getinfo_super_ddf(struct supertype *st, struct mdinfo *info, char *map);
ff54de6e 1170
42dc2744 1171static void uuid_from_super_ddf(struct supertype *st, int uuid[4]);
ff54de6e 1172
061f2c6a 1173static void brief_examine_super_ddf(struct supertype *st, int verbose)
4737ae25
N		 1174{
1175 /* We just write a generic DDF ARRAY entry
1176 */
1177 struct mdinfo info;
1178 char nbuf[64];
a5d85af7 1179 getinfo_super_ddf(st, &info, NULL);
4737ae25
N		 1180 fname_from_uuid(st, &info, nbuf, ':');
1181
1182 printf("ARRAY metadata=ddf UUID=%s\n", nbuf + 5);
1183}
1184
1185static void brief_examine_subarrays_ddf(struct supertype *st, int verbose)
a322f70c
DW		 1186{
1187 /* We just write a generic DDF ARRAY entry
a322f70c 1188 */
42dc2744 1189 struct ddf_super *ddf = st->sb;
ff54de6e 1190 struct mdinfo info;
f21e18ca 1191 unsigned int i;
ff54de6e 1192 char nbuf[64];
a5d85af7 1193 getinfo_super_ddf(st, &info, NULL);
ff54de6e 1194 fname_from_uuid(st, &info, nbuf, ':');
42dc2744 1195
f21e18ca 1196 for (i = 0; i < __be16_to_cpu(ddf->virt->max_vdes); i++) {
42dc2744
N		 1197 struct virtual_entry *ve = &ddf->virt->entries[i];
1198 struct vcl vcl;
1199 char nbuf1[64];
1200 if (all_ff(ve->guid))
1201 continue;
1202 memcpy(vcl.conf.guid, ve->guid, DDF_GUID_LEN);
1203 ddf->currentconf =&vcl;
1204 uuid_from_super_ddf(st, info.uuid);
1205 fname_from_uuid(st, &info, nbuf1, ':');
1206 printf("ARRAY container=%s member=%d UUID=%s\n",
1207 nbuf+5, i, nbuf1+5);
1208 }
a322f70c
DW		 1209}
1210
bceedeec
N		 1211static void export_examine_super_ddf(struct supertype *st)
1212{
1213 struct mdinfo info;
1214 char nbuf[64];
a5d85af7 1215 getinfo_super_ddf(st, &info, NULL);
bceedeec
N		 1216 fname_from_uuid(st, &info, nbuf, ':');
1217 printf("MD_METADATA=ddf\n");
1218 printf("MD_LEVEL=container\n");
1219 printf("MD_UUID=%s\n", nbuf+5);
1220}
1221
1222
a322f70c
DW		 1223static void detail_super_ddf(struct supertype *st, char *homehost)
1224{
1225 /* FIXME later
1226 * Could print DDF GUID
1227 * Need to find which array
1228 * If whole, briefly list all arrays
1229 * If one, give name
1230 */
1231}
1232
1233static void brief_detail_super_ddf(struct supertype *st)
1234{
1235 /* FIXME I really need to know which array we are detailing.
1236 * Can that be stored in ddf_super??
1237 */
1238// struct ddf_super *ddf = st->sb;
ff54de6e
N		 1239 struct mdinfo info;
1240 char nbuf[64];
a5d85af7 1241 getinfo_super_ddf(st, &info, NULL);
ff54de6e
N		 1242 fname_from_uuid(st, &info, nbuf,':');
1243 printf(" UUID=%s", nbuf + 5);
a322f70c 1244}
a322f70c
DW		 1245#endif
1246
1247static int match_home_ddf(struct supertype *st, char *homehost)
1248{
1249 /* It matches 'this' host if the controller is a
1250 * Linux-MD controller with vendor_data matching
1251 * the hostname
1252 */
1253 struct ddf_super *ddf = st->sb;
f21e18ca 1254 unsigned int len;
d1d3482b
N		 1255
1256 if (!homehost)
1257 return 0;
1258 len = strlen(homehost);
a322f70c
DW		 1259
1260 return (memcmp(ddf->controller.guid, T10, 8) == 0 &&
1261 len < sizeof(ddf->controller.vendor_data) &&
1262 memcmp(ddf->controller.vendor_data, homehost,len) == 0 &&
1263 ddf->controller.vendor_data[len] == 0);
1264}
1265
0e600426 1266#ifndef MDASSEMBLE
f21e18ca 1267static struct vd_config *find_vdcr(struct ddf_super *ddf, unsigned int inst)
a322f70c 1268{
7a7cc504 1269 struct vcl *v;
59e36268 1270
7a7cc504 1271 for (v = ddf->conflist; v; v = v->next)
59e36268 1272 if (inst == v->vcnum)
7a7cc504
NB		 1273 return &v->conf;
1274 return NULL;
1275}
0e600426 1276#endif
7a7cc504
NB		 1277
1278static int find_phys(struct ddf_super *ddf, __u32 phys_refnum)
1279{
1280 /* Find the entry in phys_disk which has the given refnum
1281 * and return it's index
1282 */
f21e18ca
N		 1283 unsigned int i;
1284 for (i = 0; i < __be16_to_cpu(ddf->phys->max_pdes); i++)
7a7cc504
NB		 1285 if (ddf->phys->entries[i].refnum == phys_refnum)
1286 return i;
1287 return -1;
a322f70c
DW		 1288}
1289
1290static void uuid_from_super_ddf(struct supertype *st, int uuid[4])
1291{
1292 /* The uuid returned here is used for:
1293 * uuid to put into bitmap file (Create, Grow)
1294 * uuid for backup header when saving critical section (Grow)
1295 * comparing uuids when re-adding a device into an array
51006d85
N		 1296 * In these cases the uuid required is that of the data-array,
1297 * not the device-set.
1298 * uuid to recognise same set when adding a missing device back
1299 * to an array. This is a uuid for the device-set.
1300 *
a322f70c
DW		 1301 * For each of these we can make do with a truncated
1302 * or hashed uuid rather than the original, as long as
1303 * everyone agrees.
a322f70c
DW		 1304 * In the case of SVD we assume the BVD is of interest,
1305 * though that might be the case if a bitmap were made for
1306 * a mirrored SVD - worry about that later.
1307 * So we need to find the VD configuration record for the
1308 * relevant BVD and extract the GUID and Secondary_Element_Seq.
1309 * The first 16 bytes of the sha1 of these is used.
1310 */
1311 struct ddf_super *ddf = st->sb;
d2ca6449 1312 struct vcl *vcl = ddf->currentconf;
c5afc314
N		 1313 char *guid;
1314 char buf[20];
1315 struct sha1_ctx ctx;
a322f70c 1316
c5afc314
N		 1317 if (vcl)
1318 guid = vcl->conf.guid;
1319 else
1320 guid = ddf->anchor.guid;
1321
1322 sha1_init_ctx(&ctx);
1323 sha1_process_bytes(guid, DDF_GUID_LEN, &ctx);
c5afc314
N		 1324 sha1_finish_ctx(&ctx, buf);
1325 memcpy(uuid, buf, 4*4);
a322f70c
DW		 1326}
1327
a5d85af7 1328static void getinfo_super_ddf_bvd(struct supertype *st, struct mdinfo *info, char *map);
78e44928 1329
a5d85af7 1330static void getinfo_super_ddf(struct supertype *st, struct mdinfo *info, char *map)
a322f70c
DW		 1331{
1332 struct ddf_super *ddf = st->sb;
a5d85af7 1333 int map_disks = info->array.raid_disks;
a322f70c 1334
78e44928 1335 if (ddf->currentconf) {
a5d85af7 1336 getinfo_super_ddf_bvd(st, info, map);
78e44928
NB		 1337 return;
1338 }
1339
a322f70c
DW		 1340 info->array.raid_disks = __be16_to_cpu(ddf->phys->used_pdes);
1341 info->array.level = LEVEL_CONTAINER;
1342 info->array.layout = 0;
1343 info->array.md_minor = -1;
1344 info->array.ctime = DECADE + __be32_to_cpu(*(__u32*)
1345 (ddf->anchor.guid+16));
1346 info->array.utime = 0;
1347 info->array.chunk_size = 0;
510242aa 1348 info->container_enough = 1;
a322f70c 1349
a322f70c
DW		 1350
1351 info->disk.major = 0;
1352 info->disk.minor = 0;
cba0191b
NB		 1353 if (ddf->dlist) {
1354 info->disk.number = __be32_to_cpu(ddf->dlist->disk.refnum);
59e36268 1355 info->disk.raid_disk = find_phys(ddf, ddf->dlist->disk.refnum);
d2ca6449
NB		 1356
1357 info->data_offset = __be64_to_cpu(ddf->phys->
1358 entries[info->disk.raid_disk].
1359 config_size);
1360 info->component_size = ddf->dlist->size - info->data_offset;
cba0191b
NB		 1361 } else {
1362 info->disk.number = -1;
661dce36 1363 info->disk.raid_disk = -1;
cba0191b
NB		 1364// info->disk.raid_disk = find refnum in the table and use index;
1365 }
f22385f9 1366 info->disk.state = (1 << MD_DISK_SYNC) | (1 << MD_DISK_ACTIVE);
a19c88b8 1367
d2ca6449 1368
921d9e16 1369 info->recovery_start = MaxSector;
a19c88b8 1370 info->reshape_active = 0;
c5afc314 1371 info->name[0] = 0;
a322f70c 1372
f35f2525
N		 1373 info->array.major_version = -1;
1374 info->array.minor_version = -2;
159c3a1a 1375 strcpy(info->text_version, "ddf");
a67dd8cc 1376 info->safe_mode_delay = 0;
159c3a1a 1377
c5afc314 1378 uuid_from_super_ddf(st, info->uuid);
a322f70c 1379
a5d85af7
N		 1380 if (map) {
1381 int i;
1382 for (i = 0 ; i < map_disks; i++) {
1383 if (i < info->array.raid_disks &&
1384 (__be16_to_cpu(ddf->phys->entries[i].state) & DDF_Online) &&
1385 !(__be16_to_cpu(ddf->phys->entries[i].state) & DDF_Failed))
1386 map[i] = 1;
1387 else
1388 map[i] = 0;
1389 }
1390 }
a322f70c
DW		 1391}
1392
598f0d58
NB		 1393static int rlq_to_layout(int rlq, int prl, int raiddisks);
1394
a5d85af7 1395static void getinfo_super_ddf_bvd(struct supertype *st, struct mdinfo *info, char *map)
a322f70c
DW		 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;
a5d85af7 1402 int map_disks = info->array.raid_disks;
a322f70c
DW		 1403
1404 /* FIXME this returns BVD info - what if we want SVD ?? */
1405
d2ca6449
NB		 1406 info->array.raid_disks = __be16_to_cpu(vc->conf.prim_elmnt_count);
1407 info->array.level = map_num1(ddf_level_num, vc->conf.prl);
1408 info->array.layout = rlq_to_layout(vc->conf.rlq, vc->conf.prl,
598f0d58 1409 info->array.raid_disks);
a322f70c 1410 info->array.md_minor = -1;
d2ca6449
NB		 1411 info->array.ctime = DECADE +
1412 __be32_to_cpu(*(__u32*)(vc->conf.guid+16));
1413 info->array.utime = DECADE + __be32_to_cpu(vc->conf.timestamp);
1414 info->array.chunk_size = 512 << vc->conf.chunk_shift;
da9b4a62 1415 info->custom_array_size = 0;
d2ca6449 1416
f21e18ca 1417 if (cd >= 0 && (unsigned)cd < ddf->mppe) {
d2ca6449
NB		 1418 info->data_offset = __be64_to_cpu(vc->lba_offset[cd]);
1419 if (vc->block_sizes)
1420 info->component_size = vc->block_sizes[cd];
1421 else
1422 info->component_size = __be64_to_cpu(vc->conf.blocks);
1423 }
a322f70c 1424
8592f29d
N		 1425 for (dl = ddf->dlist; dl ; dl = dl->next)
1426 if (dl->raiddisk == info->disk.raid_disk)
1427 break;
a322f70c
DW		 1428 info->disk.major = 0;
1429 info->disk.minor = 0;
8592f29d
N		 1430 if (dl) {
1431 info->disk.major = dl->major;
1432 info->disk.minor = dl->minor;
1433 }
a322f70c
DW		 1434// info->disk.number = __be32_to_cpu(ddf->disk.refnum);
1435// info->disk.raid_disk = find refnum in the table and use index;
1436// info->disk.state = ???;
1437
103f2410
NB		 1438 info->container_member = ddf->currentconf->vcnum;
1439
921d9e16 1440 info->recovery_start = MaxSector;
80d26cb2 1441 info->resync_start = 0;
624c5ad4 1442 info->reshape_active = 0;
80d26cb2
NB		 1443 if (!(ddf->virt->entries[info->container_member].state
1444 & DDF_state_inconsistent) &&
1445 (ddf->virt->entries[info->container_member].init_state
1446 & DDF_initstate_mask)
1447 == DDF_init_full)
b7528a20 1448 info->resync_start = MaxSector;
80d26cb2 1449
a322f70c
DW		 1450 uuid_from_super_ddf(st, info->uuid);
1451
f35f2525
N		 1452 info->array.major_version = -1;
1453 info->array.minor_version = -2;
9b63e648 1454 sprintf(info->text_version, "/%s/%d",
159c3a1a 1455 devnum2devname(st->container_dev),
9b63e648 1456 info->container_member);
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;
a5d85af7
N		 1464
1465 if (map)
1466 for (j = 0; j < map_disks; j++) {
1467 map[j] = 0;
1468 if (j < info->array.raid_disks) {
1469 int i = find_phys(ddf, vc->conf.phys_refnum[j]);
1470 if (i >= 0 &&
1471 (__be16_to_cpu(ddf->phys->entries[i].state) & DDF_Online) &&
1472 !(__be16_to_cpu(ddf->phys->entries[i].state) & DDF_Failed))
1473 map[i] = 1;
1474 }
1475 }
a322f70c
DW		 1476}
1477
598f0d58 1478
a322f70c
DW		 1479static int update_super_ddf(struct supertype *st, struct mdinfo *info,
1480 char *update,
1481 char *devname, int verbose,
1482 int uuid_set, char *homehost)
1483{
1484 /* For 'assemble' and 'force' we need to return non-zero if any
1485 * change was made. For others, the return value is ignored.
1486 * Update options are:
1487 * force-one : This device looks a bit old but needs to be included,
1488 * update age info appropriately.
1489 * assemble: clear any 'faulty' flag to allow this device to
1490 * be assembled.
1491 * force-array: Array is degraded but being forced, mark it clean
1492 * if that will be needed to assemble it.
1493 *
1494 * newdev: not used ????
1495 * grow: Array has gained a new device - this is currently for
1496 * linear only
1497 * resync: mark as dirty so a resync will happen.
59e36268 1498 * uuid: Change the uuid of the array to match what is given
a322f70c
DW		 1499 * homehost: update the recorded homehost
1500 * name: update the name - preserving the homehost
1501 * _reshape_progress: record new reshape_progress position.
1502 *
1503 * Following are not relevant for this version:
1504 * sparc2.2 : update from old dodgey metadata
1505 * super-minor: change the preferred_minor number
1506 * summaries: update redundant counters.
1507 */
1508 int rv = 0;
1509// struct ddf_super *ddf = st->sb;
7a7cc504 1510// struct vd_config *vd = find_vdcr(ddf, info->container_member);
a322f70c
DW		 1511// struct virtual_entry *ve = find_ve(ddf);
1512
a322f70c
DW		 1513 /* we don't need to handle "force-*" or "assemble" as
1514 * there is no need to 'trick' the kernel. We the metadata is
1515 * first updated to activate the array, all the implied modifications
1516 * will just happen.
1517 */
1518
1519 if (strcmp(update, "grow") == 0) {
1520 /* FIXME */
1e2b2765 1521 } else if (strcmp(update, "resync") == 0) {
a322f70c 1522// info->resync_checkpoint = 0;
1e2b2765 1523 } else if (strcmp(update, "homehost") == 0) {
a322f70c
DW		 1524 /* homehost is stored in controller->vendor_data,
1525 * or it is when we are the vendor
1526 */
1527// if (info->vendor_is_local)
1528// strcpy(ddf->controller.vendor_data, homehost);
1e2b2765
N		 1529 rv = -1;
1530 } if (strcmp(update, "name") == 0) {
a322f70c
DW		 1531 /* name is stored in virtual_entry->name */
1532// memset(ve->name, ' ', 16);
1533// strncpy(ve->name, info->name, 16);
1e2b2765
N		 1534 rv = -1;
1535 } if (strcmp(update, "_reshape_progress") == 0) {
a322f70c 1536 /* We don't support reshape yet */
1e2b2765
N		 1537 } else
1538 rv = -1;
a322f70c
DW		 1539
1540// update_all_csum(ddf);
1541
1542 return rv;
1543}
1544
5f8097be
NB		 1545static void make_header_guid(char *guid)
1546{
1547 __u32 stamp;
5f8097be
NB		 1548 /* Create a DDF Header of Virtual Disk GUID */
1549
1550 /* 24 bytes of fiction required.
1551 * first 8 are a 'vendor-id' - "Linux-MD"
1552 * next 8 are controller type.. how about 0X DEAD BEEF 0000 0000
1553 * Remaining 8 random number plus timestamp
1554 */
1555 memcpy(guid, T10, sizeof(T10));
1556 stamp = __cpu_to_be32(0xdeadbeef);
1557 memcpy(guid+8, &stamp, 4);
1558 stamp = __cpu_to_be32(0);
1559 memcpy(guid+12, &stamp, 4);
1560 stamp = __cpu_to_be32(time(0) - DECADE);
1561 memcpy(guid+16, &stamp, 4);
bfb7ea78 1562 stamp = random32();
5f8097be 1563 memcpy(guid+20, &stamp, 4);
5f8097be 1564}
59e36268 1565
78e44928
NB		 1566static int init_super_ddf_bvd(struct supertype *st,
1567 mdu_array_info_t *info,
1568 unsigned long long size,
1569 char *name, char *homehost,
1570 int *uuid);
1571
a322f70c
DW		 1572static int init_super_ddf(struct supertype *st,
1573 mdu_array_info_t *info,
1574 unsigned long long size, char *name, char *homehost,
1575 int *uuid)
1576{
1577 /* This is primarily called by Create when creating a new array.
1578 * We will then get add_to_super called for each component, and then
1579 * write_init_super called to write it out to each device.
1580 * For DDF, Create can create on fresh devices or on a pre-existing
1581 * array.
1582 * To create on a pre-existing array a different method will be called.
1583 * This one is just for fresh drives.
1584 *
1585 * We need to create the entire 'ddf' structure which includes:
1586 * DDF headers - these are easy.
1587 * Controller data - a Sector describing this controller .. not that
1588 * this is a controller exactly.
1589 * Physical Disk Record - one entry per device, so
1590 * leave plenty of space.
1591 * Virtual Disk Records - again, just leave plenty of space.
1592 * This just lists VDs, doesn't give details
1593 * Config records - describes the VDs that use this disk
1594 * DiskData - describes 'this' device.
1595 * BadBlockManagement - empty
1596 * Diag Space - empty
1597 * Vendor Logs - Could we put bitmaps here?
1598 *
1599 */
1600 struct ddf_super *ddf;
1601 char hostname[17];
1602 int hostlen;
a322f70c
DW		 1603 int max_phys_disks, max_virt_disks;
1604 unsigned long long sector;
1605 int clen;
1606 int i;
1607 int pdsize, vdsize;
1608 struct phys_disk *pd;
1609 struct virtual_disk *vd;
1610
78e44928 1611 if (st->sb)
955e9ea1 1612 return init_super_ddf_bvd(st, info, size, name, homehost, uuid);
ba7eb04f 1613
3d2c4fc7
DW		 1614 if (posix_memalign((void**)&ddf, 512, sizeof(*ddf)) != 0) {
1615 fprintf(stderr, Name ": %s could not allocate superblock\n", __func__);
1616 return 0;
1617 }
6264b437 1618 memset(ddf, 0, sizeof(*ddf));
a322f70c
DW		 1619 ddf->dlist = NULL; /* no physical disks yet */
1620 ddf->conflist = NULL; /* No virtual disks yet */
955e9ea1
DW		 1621 st->sb = ddf;
1622
1623 if (info == NULL) {
1624 /* zeroing superblock */
1625 return 0;
1626 }
a322f70c
DW		 1627
1628 /* At least 32MB *must* be reserved for the ddf. So let's just
1629 * start 32MB from the end, and put the primary header there.
1630 * Don't do secondary for now.
1631 * We don't know exactly where that will be yet as it could be
1632 * different on each device. To just set up the lengths.
1633 *
1634 */
1635
1636 ddf->anchor.magic = DDF_HEADER_MAGIC;
5f8097be 1637 make_header_guid(ddf->anchor.guid);
a322f70c 1638
59e36268 1639 memcpy(ddf->anchor.revision, DDF_REVISION_2, 8);
a322f70c
DW		 1640 ddf->anchor.seq = __cpu_to_be32(1);
1641 ddf->anchor.timestamp = __cpu_to_be32(time(0) - DECADE);
1642 ddf->anchor.openflag = 0xFF;
1643 ddf->anchor.foreignflag = 0;
1644 ddf->anchor.enforcegroups = 0; /* Is this best?? */
1645 ddf->anchor.pad0 = 0xff;
1646 memset(ddf->anchor.pad1, 0xff, 12);
1647 memset(ddf->anchor.header_ext, 0xff, 32);
1648 ddf->anchor.primary_lba = ~(__u64)0;
1649 ddf->anchor.secondary_lba = ~(__u64)0;
1650 ddf->anchor.type = DDF_HEADER_ANCHOR;
1651 memset(ddf->anchor.pad2, 0xff, 3);
1652 ddf->anchor.workspace_len = __cpu_to_be32(32768); /* Must be reserved */
1653 ddf->anchor.workspace_lba = ~(__u64)0; /* Put this at bottom
1654 of 32M reserved.. */
1655 max_phys_disks = 1023; /* Should be enough */
1656 ddf->anchor.max_pd_entries = __cpu_to_be16(max_phys_disks);
1657 max_virt_disks = 255;
1658 ddf->anchor.max_vd_entries = __cpu_to_be16(max_virt_disks); /* ?? */
1659 ddf->anchor.max_partitions = __cpu_to_be16(64); /* ?? */
1660 ddf->max_part = 64;
8c3b8c2c 1661 ddf->mppe = 256;
59e36268
NB		 1662 ddf->conf_rec_len = 1 + ROUND_UP(ddf->mppe * (4+8), 512)/512;
1663 ddf->anchor.config_record_len = __cpu_to_be16(ddf->conf_rec_len);
1664 ddf->anchor.max_primary_element_entries = __cpu_to_be16(ddf->mppe);
a322f70c 1665 memset(ddf->anchor.pad3, 0xff, 54);
a322f70c
DW		 1666 /* controller sections is one sector long immediately
1667 * after the ddf header */
1668 sector = 1;
1669 ddf->anchor.controller_section_offset = __cpu_to_be32(sector);
1670 ddf->anchor.controller_section_length = __cpu_to_be32(1);
1671 sector += 1;
1672
1673 /* phys is 8 sectors after that */
1674 pdsize = ROUND_UP(sizeof(struct phys_disk) +
1675 sizeof(struct phys_disk_entry)*max_phys_disks,
1676 512);
1677 switch(pdsize/512) {
1678 case 2: case 8: case 32: case 128: case 512: break;
1679 default: abort();
1680 }
1681 ddf->anchor.phys_section_offset = __cpu_to_be32(sector);
1682 ddf->anchor.phys_section_length =
1683 __cpu_to_be32(pdsize/512); /* max_primary_element_entries/8 */
1684 sector += pdsize/512;
1685
1686 /* virt is another 32 sectors */
1687 vdsize = ROUND_UP(sizeof(struct virtual_disk) +
1688 sizeof(struct virtual_entry) * max_virt_disks,
1689 512);
1690 switch(vdsize/512) {
1691 case 2: case 8: case 32: case 128: case 512: break;
1692 default: abort();
1693 }
1694 ddf->anchor.virt_section_offset = __cpu_to_be32(sector);
1695 ddf->anchor.virt_section_length =
1696 __cpu_to_be32(vdsize/512); /* max_vd_entries/8 */
1697 sector += vdsize/512;
1698
59e36268 1699 clen = ddf->conf_rec_len * (ddf->max_part+1);
a322f70c
DW		 1700 ddf->anchor.config_section_offset = __cpu_to_be32(sector);
1701 ddf->anchor.config_section_length = __cpu_to_be32(clen);
1702 sector += clen;
1703
1704 ddf->anchor.data_section_offset = __cpu_to_be32(sector);
1705 ddf->anchor.data_section_length = __cpu_to_be32(1);
1706 sector += 1;
1707
1708 ddf->anchor.bbm_section_length = __cpu_to_be32(0);
1709 ddf->anchor.bbm_section_offset = __cpu_to_be32(0xFFFFFFFF);
1710 ddf->anchor.diag_space_length = __cpu_to_be32(0);
1711 ddf->anchor.diag_space_offset = __cpu_to_be32(0xFFFFFFFF);
1712 ddf->anchor.vendor_length = __cpu_to_be32(0);
1713 ddf->anchor.vendor_offset = __cpu_to_be32(0xFFFFFFFF);
1714
1715 memset(ddf->anchor.pad4, 0xff, 256);
1716
1717 memcpy(&ddf->primary, &ddf->anchor, 512);
1718 memcpy(&ddf->secondary, &ddf->anchor, 512);
1719
1720 ddf->primary.openflag = 1; /* I guess.. */
1721 ddf->primary.type = DDF_HEADER_PRIMARY;
1722
1723 ddf->secondary.openflag = 1; /* I guess.. */
1724 ddf->secondary.type = DDF_HEADER_SECONDARY;
1725
1726 ddf->active = &ddf->primary;
1727
1728 ddf->controller.magic = DDF_CONTROLLER_MAGIC;
1729
1730 /* 24 more bytes of fiction required.
1731 * first 8 are a 'vendor-id' - "Linux-MD"
1732 * Remaining 16 are serial number.... maybe a hostname would do?
1733 */
1734 memcpy(ddf->controller.guid, T10, sizeof(T10));
1ba6bff9
DW		 1735 gethostname(hostname, sizeof(hostname));
1736 hostname[sizeof(hostname) - 1] = 0;
a322f70c
DW		 1737 hostlen = strlen(hostname);
1738 memcpy(ddf->controller.guid + 24 - hostlen, hostname, hostlen);
1739 for (i = strlen(T10) ; i+hostlen < 24; i++)
1740 ddf->controller.guid[i] = ' ';
1741
1742 ddf->controller.type.vendor_id = __cpu_to_be16(0xDEAD);
1743 ddf->controller.type.device_id = __cpu_to_be16(0xBEEF);
1744 ddf->controller.type.sub_vendor_id = 0;
1745 ddf->controller.type.sub_device_id = 0;
1746 memcpy(ddf->controller.product_id, "What Is My PID??", 16);
1747 memset(ddf->controller.pad, 0xff, 8);
1748 memset(ddf->controller.vendor_data, 0xff, 448);
a9e1c11d
N		 1749 if (homehost && strlen(homehost) < 440)
1750 strcpy((char*)ddf->controller.vendor_data, homehost);
a322f70c 1751
3d2c4fc7
DW		 1752 if (posix_memalign((void**)&pd, 512, pdsize) != 0) {
1753 fprintf(stderr, Name ": %s could not allocate pd\n", __func__);
1754 return 0;
1755 }
6416d527 1756 ddf->phys = pd;
a322f70c
DW		 1757 ddf->pdsize = pdsize;
1758
1759 memset(pd, 0xff, pdsize);
1760 memset(pd, 0, sizeof(*pd));
076515ba 1761 pd->magic = DDF_PHYS_RECORDS_MAGIC;
a322f70c
DW		 1762 pd->used_pdes = __cpu_to_be16(0);
1763 pd->max_pdes = __cpu_to_be16(max_phys_disks);
1764 memset(pd->pad, 0xff, 52);
1765
3d2c4fc7
DW		 1766 if (posix_memalign((void**)&vd, 512, vdsize) != 0) {
1767 fprintf(stderr, Name ": %s could not allocate vd\n", __func__);
1768 return 0;
1769 }
6416d527 1770 ddf->virt = vd;
a322f70c
DW		 1771 ddf->vdsize = vdsize;
1772 memset(vd, 0, vdsize);
1773 vd->magic = DDF_VIRT_RECORDS_MAGIC;
1774 vd->populated_vdes = __cpu_to_be16(0);
1775 vd->max_vdes = __cpu_to_be16(max_virt_disks);
1776 memset(vd->pad, 0xff, 52);
1777
5f8097be
NB		 1778 for (i=0; i<max_virt_disks; i++)
1779 memset(&vd->entries[i], 0xff, sizeof(struct virtual_entry));
1780
a322f70c 1781 st->sb = ddf;
18a2f463 1782 ddf->updates_pending = 1;
a322f70c
DW		 1783 return 1;
1784}
1785
5f8097be
NB		 1786static int chunk_to_shift(int chunksize)
1787{
1788 return ffs(chunksize/512)-1;
1789}
1790
1791static int level_to_prl(int level)
1792{
1793 switch (level) {
1794 case LEVEL_LINEAR: return DDF_CONCAT;
1795 case 0: return DDF_RAID0;
1796 case 1: return DDF_RAID1;
1797 case 4: return DDF_RAID4;
1798 case 5: return DDF_RAID5;
1799 case 6: return DDF_RAID6;
1800 default: return -1;
1801 }
1802}
1803static int layout_to_rlq(int level, int layout, int raiddisks)
1804{
1805 switch(level) {
1806 case 0:
1807 return DDF_RAID0_SIMPLE;
1808 case 1:
1809 switch(raiddisks) {
1810 case 2: return DDF_RAID1_SIMPLE;
1811 case 3: return DDF_RAID1_MULTI;
1812 default: return -1;
1813 }
1814 case 4:
1815 switch(layout) {
1816 case 0: return DDF_RAID4_N;
1817 }
1818 break;
1819 case 5:
5f8097be
NB		 1820 switch(layout) {
1821 case ALGORITHM_LEFT_ASYMMETRIC:
1822 return DDF_RAID5_N_RESTART;
1823 case ALGORITHM_RIGHT_ASYMMETRIC:
b640a252 1824 return DDF_RAID5_0_RESTART;
5f8097be
NB		 1825 case ALGORITHM_LEFT_SYMMETRIC:
1826 return DDF_RAID5_N_CONTINUE;
1827 case ALGORITHM_RIGHT_SYMMETRIC:
1828 return -1; /* not mentioned in standard */
1829 }
b640a252
N		 1830 case 6:
1831 switch(layout) {
1832 case ALGORITHM_ROTATING_N_RESTART:
1833 return DDF_RAID5_N_RESTART;
1834 case ALGORITHM_ROTATING_ZERO_RESTART:
1835 return DDF_RAID6_0_RESTART;
1836 case ALGORITHM_ROTATING_N_CONTINUE:
1837 return DDF_RAID5_N_CONTINUE;
1838 }
5f8097be
NB		 1839 }
1840 return -1;
1841}
1842
598f0d58
NB		 1843static int rlq_to_layout(int rlq, int prl, int raiddisks)
1844{
1845 switch(prl) {
1846 case DDF_RAID0:
1847 return 0; /* hopefully rlq == DDF_RAID0_SIMPLE */
1848 case DDF_RAID1:
1849 return 0; /* hopefully rlq == SIMPLE or MULTI depending
1850 on raiddisks*/
1851 case DDF_RAID4:
1852 switch(rlq) {
1853 case DDF_RAID4_N:
1854 return 0;
1855 default:
1856 /* not supported */
1857 return -1; /* FIXME this isn't checked */
1858 }
1859 case DDF_RAID5:
598f0d58
NB		 1860 switch(rlq) {
1861 case DDF_RAID5_N_RESTART:
1862 return ALGORITHM_LEFT_ASYMMETRIC;
1863 case DDF_RAID5_0_RESTART:
1864 return ALGORITHM_RIGHT_ASYMMETRIC;
1865 case DDF_RAID5_N_CONTINUE:
1866 return ALGORITHM_LEFT_SYMMETRIC;
1867 default:
1868 return -1;
1869 }
59e36268
NB		 1870 case DDF_RAID6:
1871 switch(rlq) {
1872 case DDF_RAID5_N_RESTART:
b640a252 1873 return ALGORITHM_ROTATING_N_RESTART;
59e36268 1874 case DDF_RAID6_0_RESTART:
b640a252 1875 return ALGORITHM_ROTATING_ZERO_RESTART;
59e36268 1876 case DDF_RAID5_N_CONTINUE:
b640a252 1877 return ALGORITHM_ROTATING_N_CONTINUE;
59e36268
NB		 1878 default:
1879 return -1;
1880 }
598f0d58
NB		 1881 }
1882 return -1;
1883}
1884
0e600426 1885#ifndef MDASSEMBLE
59e36268
NB		 1886struct extent {
1887 unsigned long long start, size;
1888};
78e44928 1889static int cmp_extent(const void *av, const void *bv)
59e36268
NB		 1890{
1891 const struct extent *a = av;
1892 const struct extent *b = bv;
1893 if (a->start < b->start)
1894 return -1;
1895 if (a->start > b->start)
1896 return 1;
1897 return 0;
1898}
1899
78e44928 1900static struct extent *get_extents(struct ddf_super *ddf, struct dl *dl)
59e36268
NB		 1901{
1902 /* find a list of used extents on the give physical device
1903 * (dnum) of the given ddf.
1904 * Return a malloced array of 'struct extent'
1905
1906FIXME ignore DDF_Legacy devices?
1907
1908 */
1909 struct extent *rv;
1910 int n = 0;
f21e18ca 1911 unsigned int i, j;
59e36268
NB		 1912
1913 rv = malloc(sizeof(struct extent) * (ddf->max_part + 2));
1914 if (!rv)
1915 return NULL;
1916
1917 for (i = 0; i < ddf->max_part; i++) {
1918 struct vcl *v = dl->vlist[i];
1919 if (v == NULL)
1920 continue;
f21e18ca 1921 for (j = 0; j < v->conf.prim_elmnt_count; j++)
59e36268
NB		 1922 if (v->conf.phys_refnum[j] == dl->disk.refnum) {
1923 /* This device plays role 'j' in 'v'. */
1924 rv[n].start = __be64_to_cpu(v->lba_offset[j]);
1925 rv[n].size = __be64_to_cpu(v->conf.blocks);
1926 n++;
1927 break;
1928 }
1929 }
1930 qsort(rv, n, sizeof(*rv), cmp_extent);
1931
1932 rv[n].start = __be64_to_cpu(ddf->phys->entries[dl->pdnum].config_size);
1933 rv[n].size = 0;
1934 return rv;
1935}
0e600426 1936#endif
59e36268 1937
5f8097be
NB		 1938static int init_super_ddf_bvd(struct supertype *st,
1939 mdu_array_info_t *info,
1940 unsigned long long size,
1941 char *name, char *homehost,
1942 int *uuid)
1943{
1944 /* We are creating a BVD inside a pre-existing container.
1945 * so st->sb is already set.
1946 * We need to create a new vd_config and a new virtual_entry
1947 */
1948 struct ddf_super *ddf = st->sb;
f21e18ca 1949 unsigned int venum;
5f8097be
NB		 1950 struct virtual_entry *ve;
1951 struct vcl *vcl;
1952 struct vd_config *vc;
5f8097be
NB		 1953
1954 if (__be16_to_cpu(ddf->virt->populated_vdes)
1955 >= __be16_to_cpu(ddf->virt->max_vdes)) {
1956 fprintf(stderr, Name": This ddf already has the "
1957 "maximum of %d virtual devices\n",
1958 __be16_to_cpu(ddf->virt->max_vdes));
1959 return 0;
1960 }
1961
1962 for (venum = 0; venum < __be16_to_cpu(ddf->virt->max_vdes); venum++)
1963 if (all_ff(ddf->virt->entries[venum].guid))
1964 break;
1965 if (venum == __be16_to_cpu(ddf->virt->max_vdes)) {
1966 fprintf(stderr, Name ": Cannot find spare slot for "
1967 "virtual disk - DDF is corrupt\n");
1968 return 0;
1969 }
1970 ve = &ddf->virt->entries[venum];
1971
1972 /* A Virtual Disk GUID contains the T10 Vendor ID, controller type,
1973 * timestamp, random number
1974 */
1975 make_header_guid(ve->guid);
1976 ve->unit = __cpu_to_be16(info->md_minor);
1977 ve->pad0 = 0xFFFF;
1978 ve->guid_crc = crc32(0, (unsigned char*)ddf->anchor.guid, DDF_GUID_LEN);
1979 ve->type = 0;
7a7cc504
NB		 1980 ve->state = DDF_state_degraded; /* Will be modified as devices are added */
1981 if (info->state & 1) /* clean */
1982 ve->init_state = DDF_init_full;
1983 else
1984 ve->init_state = DDF_init_not;
1985
5f8097be
NB		 1986 memset(ve->pad1, 0xff, 14);
1987 memset(ve->name, ' ', 16);
1988 if (name)
1989 strncpy(ve->name, name, 16);
1990 ddf->virt->populated_vdes =
1991 __cpu_to_be16(__be16_to_cpu(ddf->virt->populated_vdes)+1);
1992
1993 /* Now create a new vd_config */
3d2c4fc7
DW		 1994 if (posix_memalign((void**)&vcl, 512,
1995 (offsetof(struct vcl, conf) + ddf->conf_rec_len * 512)) != 0) {
1996 fprintf(stderr, Name ": %s could not allocate vd_config\n", __func__);
1997 return 0;
1998 }
8c3b8c2c 1999 vcl->lba_offset = (__u64*) &vcl->conf.phys_refnum[ddf->mppe];
59e36268 2000 vcl->vcnum = venum;
f7e7067b 2001 sprintf(st->subarray, "%d", venum);
59e36268 2002 vcl->block_sizes = NULL; /* FIXME not for CONCAT */
5f8097be
NB		 2003
2004 vc = &vcl->conf;
2005
2006 vc->magic = DDF_VD_CONF_MAGIC;
2007 memcpy(vc->guid, ve->guid, DDF_GUID_LEN);
2008 vc->timestamp = __cpu_to_be32(time(0)-DECADE);
2009 vc->seqnum = __cpu_to_be32(1);
2010 memset(vc->pad0, 0xff, 24);
2011 vc->prim_elmnt_count = __cpu_to_be16(info->raid_disks);
2012 vc->chunk_shift = chunk_to_shift(info->chunk_size);
2013 vc->prl = level_to_prl(info->level);
2014 vc->rlq = layout_to_rlq(info->level, info->layout, info->raid_disks);
2015 vc->sec_elmnt_count = 1;
2016 vc->sec_elmnt_seq = 0;
2017 vc->srl = 0;
2018 vc->blocks = __cpu_to_be64(info->size * 2);
2019 vc->array_blocks = __cpu_to_be64(
2020 calc_array_size(info->level, info->raid_disks, info->layout,
2021 info->chunk_size, info->size*2));
2022 memset(vc->pad1, 0xff, 8);
2023 vc->spare_refs[0] = 0xffffffff;
2024 vc->spare_refs[1] = 0xffffffff;
2025 vc->spare_refs[2] = 0xffffffff;
2026 vc->spare_refs[3] = 0xffffffff;
2027 vc->spare_refs[4] = 0xffffffff;
2028 vc->spare_refs[5] = 0xffffffff;
2029 vc->spare_refs[6] = 0xffffffff;
2030 vc->spare_refs[7] = 0xffffffff;
2031 memset(vc->cache_pol, 0, 8);
2032 vc->bg_rate = 0x80;
2033 memset(vc->pad2, 0xff, 3);
2034 memset(vc->pad3, 0xff, 52);
2035 memset(vc->pad4, 0xff, 192);
2036 memset(vc->v0, 0xff, 32);
2037 memset(vc->v1, 0xff, 32);
2038 memset(vc->v2, 0xff, 16);
2039 memset(vc->v3, 0xff, 16);
2040 memset(vc->vendor, 0xff, 32);
598f0d58 2041
8c3b8c2c 2042 memset(vc->phys_refnum, 0xff, 4*ddf->mppe);
e5a2a3cf 2043 memset(vc->phys_refnum+ddf->mppe, 0x00, 8*ddf->mppe);
5f8097be
NB		 2044
2045 vcl->next = ddf->conflist;
2046 ddf->conflist = vcl;
d2ca6449 2047 ddf->currentconf = vcl;
18a2f463 2048 ddf->updates_pending = 1;
5f8097be
NB		 2049 return 1;
2050}
2051
0e600426 2052#ifndef MDASSEMBLE
5f8097be
NB		 2053static void add_to_super_ddf_bvd(struct supertype *st,
2054 mdu_disk_info_t *dk, int fd, char *devname)
2055{
2056 /* fd and devname identify a device with-in the ddf container (st).
2057 * dk identifies a location in the new BVD.
2058 * We need to find suitable free space in that device and update
2059 * the phys_refnum and lba_offset for the newly created vd_config.
2060 * We might also want to update the type in the phys_disk
5575e7d9 2061 * section.
8592f29d
N		 2062 *
2063 * Alternately: fd == -1 and we have already chosen which device to
2064 * use and recorded in dlist->raid_disk;
5f8097be
NB		 2065 */
2066 struct dl *dl;
2067 struct ddf_super *ddf = st->sb;
2068 struct vd_config *vc;
2069 __u64 *lba_offset;
f21e18ca
N		 2070 unsigned int working;
2071 unsigned int i;
59e36268
NB		 2072 unsigned long long blocks, pos, esize;
2073 struct extent *ex;
5f8097be 2074
8592f29d
N		 2075 if (fd == -1) {
2076 for (dl = ddf->dlist; dl ; dl = dl->next)
2077 if (dl->raiddisk == dk->raid_disk)
2078 break;
2079 } else {
2080 for (dl = ddf->dlist; dl ; dl = dl->next)
2081 if (dl->major == dk->major &&
2082 dl->minor == dk->minor)
2083 break;
2084 }
5f8097be
NB		 2085 if (!dl || ! (dk->state & (1<<MD_DISK_SYNC)))
2086 return;
2087
d2ca6449
NB		 2088 vc = &ddf->currentconf->conf;
2089 lba_offset = ddf->currentconf->lba_offset;
59e36268
NB		 2090
2091 ex = get_extents(ddf, dl);
2092 if (!ex)
2093 return;
2094
2095 i = 0; pos = 0;
2096 blocks = __be64_to_cpu(vc->blocks);
d2ca6449
NB		 2097 if (ddf->currentconf->block_sizes)
2098 blocks = ddf->currentconf->block_sizes[dk->raid_disk];
59e36268
NB		 2099
2100 do {
2101 esize = ex[i].start - pos;
2102 if (esize >= blocks)
2103 break;
2104 pos = ex[i].start + ex[i].size;
2105 i++;
2106 } while (ex[i-1].size);
2107
2108 free(ex);
2109 if (esize < blocks)
2110 return;
2111
d2ca6449 2112 ddf->currentdev = dk->raid_disk;
5f8097be 2113 vc->phys_refnum[dk->raid_disk] = dl->disk.refnum;
59e36268 2114 lba_offset[dk->raid_disk] = __cpu_to_be64(pos);
5f8097be 2115
f21e18ca 2116 for (i = 0; i < ddf->max_part ; i++)
5575e7d9
NB		 2117 if (dl->vlist[i] == NULL)
2118 break;
2119 if (i == ddf->max_part)
2120 return;
d2ca6449 2121 dl->vlist[i] = ddf->currentconf;
5f8097be 2122
8592f29d
N		 2123 if (fd >= 0)
2124 dl->fd = fd;
2125 if (devname)
2126 dl->devname = devname;
7a7cc504
NB		 2127
2128 /* Check how many working raid_disks, and if we can mark
2129 * array as optimal yet
2130 */
2131 working = 0;
5575e7d9 2132
f21e18ca 2133 for (i = 0; i < __be16_to_cpu(vc->prim_elmnt_count); i++)
7a7cc504
NB		 2134 if (vc->phys_refnum[i] != 0xffffffff)
2135 working++;
59e36268 2136
5575e7d9 2137 /* Find which virtual_entry */
d2ca6449 2138 i = ddf->currentconf->vcnum;
7a7cc504 2139 if (working == __be16_to_cpu(vc->prim_elmnt_count))
5575e7d9
NB		 2140 ddf->virt->entries[i].state =
2141 (ddf->virt->entries[i].state & ~DDF_state_mask)
7a7cc504
NB		 2142 | DDF_state_optimal;
2143
2144 if (vc->prl == DDF_RAID6 &&
2145 working+1 == __be16_to_cpu(vc->prim_elmnt_count))
5575e7d9
NB		 2146 ddf->virt->entries[i].state =
2147 (ddf->virt->entries[i].state & ~DDF_state_mask)
7a7cc504 2148 | DDF_state_part_optimal;
5575e7d9
NB		 2149
2150 ddf->phys->entries[dl->pdnum].type &= ~__cpu_to_be16(DDF_Global_Spare);
2151 ddf->phys->entries[dl->pdnum].type |= __cpu_to_be16(DDF_Active_in_VD);
18a2f463 2152 ddf->updates_pending = 1;
5f8097be
NB		 2153}
2154
a322f70c
DW		 2155/* add a device to a container, either while creating it or while
2156 * expanding a pre-existing container
2157 */
f20c3968 2158static int add_to_super_ddf(struct supertype *st,
a322f70c
DW		 2159 mdu_disk_info_t *dk, int fd, char *devname)
2160{
2161 struct ddf_super *ddf = st->sb;
2162 struct dl *dd;
2163 time_t now;
2164 struct tm *tm;
2165 unsigned long long size;
2166 struct phys_disk_entry *pde;
f21e18ca 2167 unsigned int n, i;
a322f70c
DW		 2168 struct stat stb;
2169
78e44928
NB		 2170 if (ddf->currentconf) {
2171 add_to_super_ddf_bvd(st, dk, fd, devname);
f20c3968 2172 return 0;
78e44928
NB		 2173 }
2174
a322f70c
DW		 2175 /* This is device numbered dk->number. We need to create
2176 * a phys_disk entry and a more detailed disk_data entry.
2177 */
2178 fstat(fd, &stb);
3d2c4fc7
DW		 2179 if (posix_memalign((void**)&dd, 512,
2180 sizeof(*dd) + sizeof(dd->vlist[0]) * ddf->max_part) != 0) {
2181 fprintf(stderr, Name
2182 ": %s could allocate buffer for new disk, aborting\n",
2183 __func__);
f20c3968 2184 return 1;
3d2c4fc7 2185 }
a322f70c
DW		 2186 dd->major = major(stb.st_rdev);
2187 dd->minor = minor(stb.st_rdev);
2188 dd->devname = devname;
a322f70c 2189 dd->fd = fd;
b2280677 2190 dd->spare = NULL;
a322f70c
DW		 2191
2192 dd->disk.magic = DDF_PHYS_DATA_MAGIC;
2193 now = time(0);
2194 tm = localtime(&now);
2195 sprintf(dd->disk.guid, "%8s%04d%02d%02d",
2196 T10, tm->tm_year+1900, tm->tm_mon+1, tm->tm_mday);
bfb7ea78
N		 2197 *(__u32*)(dd->disk.guid + 16) = random32();
2198 *(__u32*)(dd->disk.guid + 20) = random32();
a322f70c 2199
59e36268
NB		 2200 do {
2201 /* Cannot be bothered finding a CRC of some irrelevant details*/
bfb7ea78 2202 dd->disk.refnum = random32();
f21e18ca
N		 2203 for (i = __be16_to_cpu(ddf->active->max_pd_entries);
2204 i > 0; i--)
2205 if (ddf->phys->entries[i-1].refnum == dd->disk.refnum)
59e36268 2206 break;
f21e18ca 2207 } while (i > 0);
59e36268 2208
a322f70c
DW		 2209 dd->disk.forced_ref = 1;
2210 dd->disk.forced_guid = 1;
2211 memset(dd->disk.vendor, ' ', 32);
2212 memcpy(dd->disk.vendor, "Linux", 5);
2213 memset(dd->disk.pad, 0xff, 442);
b2280677 2214 for (i = 0; i < ddf->max_part ; i++)
a322f70c
DW		 2215 dd->vlist[i] = NULL;
2216
2217 n = __be16_to_cpu(ddf->phys->used_pdes);
2218 pde = &ddf->phys->entries[n];
5575e7d9
NB		 2219 dd->pdnum = n;
2220
2cc2983d
N		 2221 if (st->update_tail) {
2222 int len = (sizeof(struct phys_disk) +
2223 sizeof(struct phys_disk_entry));
2224 struct phys_disk *pd;
2225
2226 pd = malloc(len);
2227 pd->magic = DDF_PHYS_RECORDS_MAGIC;
2228 pd->used_pdes = __cpu_to_be16(n);
2229 pde = &pd->entries[0];
2230 dd->mdupdate = pd;
2231 } else {
2232 n++;
2233 ddf->phys->used_pdes = __cpu_to_be16(n);
2234 }
a322f70c
DW		 2235
2236 memcpy(pde->guid, dd->disk.guid, DDF_GUID_LEN);
2237 pde->refnum = dd->disk.refnum;
5575e7d9 2238 pde->type = __cpu_to_be16(DDF_Forced_PD_GUID | DDF_Global_Spare);
a322f70c
DW		 2239 pde->state = __cpu_to_be16(DDF_Online);
2240 get_dev_size(fd, NULL, &size);
2241 /* We are required to reserve 32Meg, and record the size in sectors */
2242 pde->config_size = __cpu_to_be64( (size - 32*1024*1024) / 512);
2243 sprintf(pde->path, "%17.17s","Information: nil") ;
2244 memset(pde->pad, 0xff, 6);
2245
d2ca6449 2246 dd->size = size >> 9;
2cc2983d
N		 2247 if (st->update_tail) {
2248 dd->next = ddf->add_list;
2249 ddf->add_list = dd;
2250 } else {
2251 dd->next = ddf->dlist;
2252 ddf->dlist = dd;
2253 ddf->updates_pending = 1;
2254 }
f20c3968
DW		 2255
2256 return 0;
a322f70c
DW		 2257}
2258
2259/*
2260 * This is the write_init_super method for a ddf container. It is
2261 * called when creating a container or adding another device to a
2262 * container.
2263 */
2264
6416d527 2265static unsigned char null_conf[4096+512];
18a2f463 2266
7a7cc504 2267static int __write_init_super_ddf(struct supertype *st, int do_close)
a322f70c
DW		 2268{
2269
2270 struct ddf_super *ddf = st->sb;
2271 int i;
2272 struct dl *d;
2273 int n_config;
2274 int conf_size;
175593bf
DW		 2275 int attempts = 0;
2276 int successes = 0;
a322f70c
DW		 2277 unsigned long long size, sector;
2278
175593bf
DW		 2279 /* try to write updated metadata,
2280 * if we catch a failure move on to the next disk
2281 */
a322f70c
DW		 2282 for (d = ddf->dlist; d; d=d->next) {
2283 int fd = d->fd;
2284
2285 if (fd < 0)
2286 continue;
2287
175593bf 2288 attempts++;
a322f70c
DW		 2289 /* We need to fill in the primary, (secondary) and workspace
2290 * lba's in the headers, set their checksums,
2291 * Also checksum phys, virt....
2292 *
2293 * Then write everything out, finally the anchor is written.
2294 */
2295 get_dev_size(fd, NULL, &size);
2296 size /= 512;
2297 ddf->anchor.workspace_lba = __cpu_to_be64(size - 32*1024*2);
2298 ddf->anchor.primary_lba = __cpu_to_be64(size - 16*1024*2);
2299 ddf->anchor.seq = __cpu_to_be32(1);
2300 memcpy(&ddf->primary, &ddf->anchor, 512);
2301 memcpy(&ddf->secondary, &ddf->anchor, 512);
2302
2303 ddf->anchor.openflag = 0xFF; /* 'open' means nothing */
2304 ddf->anchor.seq = 0xFFFFFFFF; /* no sequencing in anchor */
2305 ddf->anchor.crc = calc_crc(&ddf->anchor, 512);
2306
2307 ddf->primary.openflag = 0;
2308 ddf->primary.type = DDF_HEADER_PRIMARY;
2309
2310 ddf->secondary.openflag = 0;
2311 ddf->secondary.type = DDF_HEADER_SECONDARY;
2312
2313 ddf->primary.crc = calc_crc(&ddf->primary, 512);
2314 ddf->secondary.crc = calc_crc(&ddf->secondary, 512);
2315
2316 sector = size - 16*1024*2;
2317 lseek64(fd, sector<<9, 0);
175593bf
DW		 2318 if (write(fd, &ddf->primary, 512) < 0)
2319 continue;
a322f70c
DW		 2320
2321 ddf->controller.crc = calc_crc(&ddf->controller, 512);
175593bf
DW		 2322 if (write(fd, &ddf->controller, 512) < 0)
2323 continue;
a322f70c
DW		 2324
2325 ddf->phys->crc = calc_crc(ddf->phys, ddf->pdsize);
2326
175593bf
DW		 2327 if (write(fd, ddf->phys, ddf->pdsize) < 0)
2328 continue;
a322f70c
DW		 2329
2330 ddf->virt->crc = calc_crc(ddf->virt, ddf->vdsize);
175593bf
DW		 2331 if (write(fd, ddf->virt, ddf->vdsize) < 0)
2332 continue;
a322f70c
DW		 2333
2334 /* Now write lots of config records. */
8c3b8c2c
NB		 2335 n_config = ddf->max_part;
2336 conf_size = ddf->conf_rec_len * 512;
a322f70c
DW		 2337 for (i = 0 ; i <= n_config ; i++) {
2338 struct vcl *c = d->vlist[i];
b2280677
NB		 2339 if (i == n_config)
2340 c = (struct vcl*)d->spare;
a322f70c
DW		 2341
2342 if (c) {
2343 c->conf.crc = calc_crc(&c->conf, conf_size);
175593bf
DW		 2344 if (write(fd, &c->conf, conf_size) < 0)
2345 break;
a322f70c 2346 } else {
6416d527 2347 char *null_aligned = (char*)((((unsigned long)null_conf)+511)&~511UL);
18a2f463
NB		 2348 if (null_conf[0] != 0xff)
2349 memset(null_conf, 0xff, sizeof(null_conf));
f21e18ca 2350 unsigned int togo = conf_size;
6416d527 2351 while (togo > sizeof(null_conf)-512) {
175593bf
DW		 2352 if (write(fd, null_aligned, sizeof(null_conf)-512) < 0)
2353 break;
6416d527 2354 togo -= sizeof(null_conf)-512;
18a2f463 2355 }
175593bf
DW		 2356 if (write(fd, null_aligned, togo) < 0)
2357 break;
a322f70c
DW		 2358 }
2359 }
175593bf
DW		 2360 if (i <= n_config)
2361 continue;
a322f70c 2362 d->disk.crc = calc_crc(&d->disk, 512);
175593bf
DW		 2363 if (write(fd, &d->disk, 512) < 0)
2364 continue;
a322f70c
DW		 2365
2366 /* Maybe do the same for secondary */
2367
2368 lseek64(fd, (size-1)*512, SEEK_SET);
175593bf
DW		 2369 if (write(fd, &ddf->anchor, 512) < 0)
2370 continue;
2371 successes++;
2372 }
2373
2374 if (do_close)
2375 for (d = ddf->dlist; d; d=d->next) {
2376 close(d->fd);
7a7cc504
NB		 2377 d->fd = -1;
2378 }
175593bf
DW		 2379
2380 return attempts != successes;
a322f70c 2381}
7a7cc504
NB		 2382
2383static int write_init_super_ddf(struct supertype *st)
2384{
9b1fb677
DW		 2385 struct ddf_super *ddf = st->sb;
2386 struct vcl *currentconf = ddf->currentconf;
2387
2388 /* we are done with currentconf reset it to point st at the container */
2389 ddf->currentconf = NULL;
edd8d13c
NB		 2390
2391 if (st->update_tail) {
2392 /* queue the virtual_disk and vd_config as metadata updates */
2393 struct virtual_disk *vd;
2394 struct vd_config *vc;
edd8d13c
NB		 2395 int len;
2396
9b1fb677 2397 if (!currentconf) {
2cc2983d
N		 2398 int len = (sizeof(struct phys_disk) +
2399 sizeof(struct phys_disk_entry));
2400
2401 /* adding a disk to the container. */
2402 if (!ddf->add_list)
2403 return 0;
2404
2405 append_metadata_update(st, ddf->add_list->mdupdate, len);
2406 ddf->add_list->mdupdate = NULL;
2407 return 0;
2408 }
2409
2410 /* Newly created VD */
2411
edd8d13c
NB		 2412 /* First the virtual disk. We have a slightly fake header */
2413 len = sizeof(struct virtual_disk) + sizeof(struct virtual_entry);
2414 vd = malloc(len);
2415 *vd = *ddf->virt;
9b1fb677
DW		 2416 vd->entries[0] = ddf->virt->entries[currentconf->vcnum];
2417 vd->populated_vdes = __cpu_to_be16(currentconf->vcnum);
edd8d13c
NB		 2418 append_metadata_update(st, vd, len);
2419
2420 /* Then the vd_config */
2421 len = ddf->conf_rec_len * 512;
2422 vc = malloc(len);
9b1fb677 2423 memcpy(vc, &currentconf->conf, len);
edd8d13c
NB		 2424 append_metadata_update(st, vc, len);
2425
2426 /* FIXME I need to close the fds! */
2427 return 0;
d682f344
N		 2428 } else {
2429 struct dl *d;
2430 for (d = ddf->dlist; d; d=d->next)
2431 while (Kill(d->devname, NULL, 0, 1, 1) == 0);
edd8d13c 2432 return __write_init_super_ddf(st, 1);
d682f344 2433 }
7a7cc504
NB		 2434}
2435
a322f70c
DW		 2436#endif
2437
2438static __u64 avail_size_ddf(struct supertype *st, __u64 devsize)
2439{
2440 /* We must reserve the last 32Meg */
2441 if (devsize <= 32*1024*2)
2442 return 0;
2443 return devsize - 32*1024*2;
2444}
2445
2446#ifndef MDASSEMBLE
8592f29d
N		 2447
2448static int reserve_space(struct supertype *st, int raiddisks,
2449 unsigned long long size, int chunk,
2450 unsigned long long *freesize)
2451{
2452 /* Find 'raiddisks' spare extents at least 'size' big (but
2453 * only caring about multiples of 'chunk') and remember
2454 * them.
2455 * If the cannot be found, fail.
2456 */
2457 struct dl *dl;
2458 struct ddf_super *ddf = st->sb;
2459 int cnt = 0;
2460
2461 for (dl = ddf->dlist; dl ; dl=dl->next) {
2462 dl->raiddisk = -1;
2463 dl->esize = 0;
2464 }
2465 /* Now find largest extent on each device */
2466 for (dl = ddf->dlist ; dl ; dl=dl->next) {
2467 struct extent *e = get_extents(ddf, dl);
2468 unsigned long long pos = 0;
2469 int i = 0;
2470 int found = 0;
2471 unsigned long long minsize = size;
2472
2473 if (size == 0)
2474 minsize = chunk;
2475
2476 if (!e)
2477 continue;
2478 do {
2479 unsigned long long esize;
2480 esize = e[i].start - pos;
2481 if (esize >= minsize) {
2482 found = 1;
2483 minsize = esize;
2484 }
2485 pos = e[i].start + e[i].size;
2486 i++;
2487 } while (e[i-1].size);
2488 if (found) {
2489 cnt++;
2490 dl->esize = minsize;
2491 }
2492 free(e);
2493 }
2494 if (cnt < raiddisks) {
2495 fprintf(stderr, Name ": not enough devices with space to create array.\n");
2496 return 0; /* No enough free spaces large enough */
2497 }
2498 if (size == 0) {
2499 /* choose the largest size of which there are at least 'raiddisk' */
2500 for (dl = ddf->dlist ; dl ; dl=dl->next) {
2501 struct dl *dl2;
2502 if (dl->esize <= size)
2503 continue;
2504 /* This is bigger than 'size', see if there are enough */
2505 cnt = 0;
2506 for (dl2 = dl; dl2 ; dl2=dl2->next)
2507 if (dl2->esize >= dl->esize)
2508 cnt++;
2509 if (cnt >= raiddisks)
2510 size = dl->esize;
2511 }
2512 if (chunk) {
2513 size = size / chunk;
2514 size *= chunk;
2515 }
2516 *freesize = size;
2517 if (size < 32) {
2518 fprintf(stderr, Name ": not enough spare devices to create array.\n");
2519 return 0;
2520 }
2521 }
2522 /* We have a 'size' of which there are enough spaces.
2523 * We simply do a first-fit */
2524 cnt = 0;
2525 for (dl = ddf->dlist ; dl && cnt < raiddisks ; dl=dl->next) {
2526 if (dl->esize < size)
2527 continue;
2528
2529 dl->raiddisk = cnt;
2530 cnt++;
2531 }
2532 return 1;
2533}
2534
2535
2536
2c514b71
NB		 2537static int
2538validate_geometry_ddf_container(struct supertype *st,
2539 int level, int layout, int raiddisks,
2540 int chunk, unsigned long long size,
2541 char *dev, unsigned long long *freesize,
2542 int verbose);
78e44928
NB		 2543
2544static int validate_geometry_ddf_bvd(struct supertype *st,
2545 int level, int layout, int raiddisks,
2546 int chunk, unsigned long long size,
2c514b71
NB		 2547 char *dev, unsigned long long *freesize,
2548 int verbose);
78e44928
NB		 2549
2550static int validate_geometry_ddf(struct supertype *st,
2c514b71
NB		 2551 int level, int layout, int raiddisks,
2552 int chunk, unsigned long long size,
2553 char *dev, unsigned long long *freesize,
2554 int verbose)
a322f70c
DW		 2555{
2556 int fd;
2557 struct mdinfo *sra;
2558 int cfd;
2559
2560 /* ddf potentially supports lots of things, but it depends on
2561 * what devices are offered (and maybe kernel version?)
2562 * If given unused devices, we will make a container.
2563 * If given devices in a container, we will make a BVD.
2564 * If given BVDs, we make an SVD, changing all the GUIDs in the process.
2565 */
2566
2567 if (level == LEVEL_CONTAINER) {
78e44928
NB		 2568 /* Must be a fresh device to add to a container */
2569 return validate_geometry_ddf_container(st, level, layout,
2c514b71
NB		 2570 raiddisks, chunk,
2571 size, dev, freesize,
2572 verbose);
5f8097be
NB		 2573 }
2574
78e44928
NB		 2575 if (!dev) {
2576 /* Initial sanity check. Exclude illegal levels. */
2577 int i;
2578 for (i=0; ddf_level_num[i].num1 != MAXINT; i++)
2579 if (ddf_level_num[i].num2 == level)
2580 break;
b42f577a
N		 2581 if (ddf_level_num[i].num1 == MAXINT) {
2582 if (verbose)
2583 fprintf(stderr, Name ": DDF does not support level %d arrays\n",
2584 level);
78e44928 2585 return 0;
b42f577a 2586 }
78e44928 2587 /* Should check layout? etc */
8592f29d
N		 2588
2589 if (st->sb && freesize) {
2590 /* --create was given a container to create in.
2591 * So we need to check that there are enough
2592 * free spaces and return the amount of space.
2593 * We may as well remember which drives were
2594 * chosen so that add_to_super/getinfo_super
2595 * can return them.
2596 */
2597 return reserve_space(st, raiddisks, size, chunk, freesize);
2598 }
a322f70c 2599 return 1;
78e44928 2600 }
a322f70c 2601
8592f29d
N		 2602 if (st->sb) {
2603 /* A container has already been opened, so we are
2604 * creating in there. Maybe a BVD, maybe an SVD.
2605 * Should make a distinction one day.
2606 */
2607 return validate_geometry_ddf_bvd(st, level, layout, raiddisks,
2608 chunk, size, dev, freesize,
2609 verbose);
2610 }
78e44928
NB		 2611 /* This is the first device for the array.
2612 * If it is a container, we read it in and do automagic allocations,
2613 * no other devices should be given.
2614 * Otherwise it must be a member device of a container, and we
2615 * do manual allocation.
2616 * Later we should check for a BVD and make an SVD.
a322f70c 2617 */
a322f70c
DW		 2618 fd = open(dev, O_RDONLY|O_EXCL, 0);
2619 if (fd >= 0) {
2620 sra = sysfs_read(fd, 0, GET_VERSION);
2621 close(fd);
2622 if (sra && sra->array.major_version == -1 &&
78e44928
NB		 2623 strcmp(sra->text_version, "ddf") == 0) {
2624
2625 /* load super */
2626 /* find space for 'n' devices. */
2627 /* remember the devices */
2628 /* Somehow return the fact that we have enough */
a322f70c
DW		 2629 }
2630
2c514b71
NB		 2631 if (verbose)
2632 fprintf(stderr,
2633 Name ": ddf: Cannot create this array "
b42f577a 2634 "on device %s - a container is required.\n",
2c514b71 2635 dev);
a322f70c
DW		 2636 return 0;
2637 }
2638 if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) {
2c514b71
NB		 2639 if (verbose)
2640 fprintf(stderr, Name ": ddf: Cannot open %s: %s\n",
2641 dev, strerror(errno));
a322f70c
DW		 2642 return 0;
2643 }
2644 /* Well, it is in use by someone, maybe a 'ddf' container. */
2645 cfd = open_container(fd);
2646 if (cfd < 0) {
2647 close(fd);
2c514b71
NB		 2648 if (verbose)
2649 fprintf(stderr, Name ": ddf: Cannot use %s: %s\n",
2650 dev, strerror(EBUSY));
a322f70c
DW		 2651 return 0;
2652 }
2653 sra = sysfs_read(cfd, 0, GET_VERSION);
2654 close(fd);
2655 if (sra && sra->array.major_version == -1 &&
2656 strcmp(sra->text_version, "ddf") == 0) {
2657 /* This is a member of a ddf container. Load the container
2658 * and try to create a bvd
2659 */
2660 struct ddf_super *ddf;
a322f70c 2661 if (load_super_ddf_all(st, cfd, (void **)&ddf, NULL, 1) == 0) {
5f8097be 2662 st->sb = ddf;
2f6079dc 2663 st->container_dev = fd2devnum(cfd);
a322f70c 2664 close(cfd);
78e44928 2665 return validate_geometry_ddf_bvd(st, level, layout,
a322f70c 2666 raiddisks, chunk, size,
2c514b71
NB		 2667 dev, freesize,
2668 verbose);
a322f70c
DW		 2669 }
2670 close(cfd);
c42ec1ed
DW		 2671 } else /* device may belong to a different container */
2672 return 0;
2673
a322f70c
DW		 2674 return 1;
2675}
2676
2c514b71
NB		 2677static int
2678validate_geometry_ddf_container(struct supertype *st,
2679 int level, int layout, int raiddisks,
2680 int chunk, unsigned long long size,
2681 char *dev, unsigned long long *freesize,
2682 int verbose)
a322f70c
DW		 2683{
2684 int fd;
2685 unsigned long long ldsize;
2686
2687 if (level != LEVEL_CONTAINER)
2688 return 0;
2689 if (!dev)
2690 return 1;
2691
2692 fd = open(dev, O_RDONLY|O_EXCL, 0);
2693 if (fd < 0) {
2c514b71
NB		 2694 if (verbose)
2695 fprintf(stderr, Name ": ddf: Cannot open %s: %s\n",
2696 dev, strerror(errno));
a322f70c
DW		 2697 return 0;
2698 }
2699 if (!get_dev_size(fd, dev, &ldsize)) {
2700 close(fd);
2701 return 0;
2702 }
2703 close(fd);
2704
56aca704 2705 *freesize = avail_size_ddf(st, ldsize >> 9);
ea17e7aa
N		 2706 if (*freesize == 0)
2707 return 0;
a322f70c
DW		 2708
2709 return 1;
2710}
2711
78e44928
NB		 2712static int validate_geometry_ddf_bvd(struct supertype *st,
2713 int level, int layout, int raiddisks,
2714 int chunk, unsigned long long size,
2c514b71
NB		 2715 char *dev, unsigned long long *freesize,
2716 int verbose)
a322f70c
DW		 2717{
2718 struct stat stb;
2719 struct ddf_super *ddf = st->sb;
2720 struct dl *dl;
5f8097be
NB		 2721 unsigned long long pos = 0;
2722 unsigned long long maxsize;
2723 struct extent *e;
2724 int i;
a322f70c 2725 /* ddf/bvd supports lots of things, but not containers */
b42f577a
N		 2726 if (level == LEVEL_CONTAINER) {
2727 if (verbose)
2728 fprintf(stderr, Name ": DDF cannot create a container within an container\n");
a322f70c 2729 return 0;
b42f577a 2730 }
a322f70c
DW		 2731 /* We must have the container info already read in. */
2732 if (!ddf)
2733 return 0;
2734
5f8097be
NB		 2735 if (!dev) {
2736 /* General test: make sure there is space for
2737 * 'raiddisks' device extents of size 'size'.
2738 */
2739 unsigned long long minsize = size;
2740 int dcnt = 0;
2741 if (minsize == 0)
2742 minsize = 8;
2743 for (dl = ddf->dlist; dl ; dl = dl->next)
2744 {
2745 int found = 0;
7e1432fb 2746 pos = 0;
5f8097be
NB		 2747
2748 i = 0;
2749 e = get_extents(ddf, dl);
2750 if (!e) continue;
2751 do {
2752 unsigned long long esize;
2753 esize = e[i].start - pos;
2754 if (esize >= minsize)
2755 found = 1;
2756 pos = e[i].start + e[i].size;
2757 i++;
2758 } while (e[i-1].size);
2759 if (found)
2760 dcnt++;
2761 free(e);
2762 }
2763 if (dcnt < raiddisks) {
2c514b71
NB		 2764 if (verbose)
2765 fprintf(stderr,
2766 Name ": ddf: Not enough devices with "
2767 "space for this array (%d < %d)\n",
2768 dcnt, raiddisks);
5f8097be
NB		 2769 return 0;
2770 }
2771 return 1;
2772 }
a322f70c
DW		 2773 /* This device must be a member of the set */
2774 if (stat(dev, &stb) < 0)
2775 return 0;
2776 if ((S_IFMT & stb.st_mode) != S_IFBLK)
2777 return 0;
2778 for (dl = ddf->dlist ; dl ; dl = dl->next) {
f21e18ca
N		 2779 if (dl->major == (int)major(stb.st_rdev) &&
2780 dl->minor == (int)minor(stb.st_rdev))
a322f70c
DW		 2781 break;
2782 }
5f8097be 2783 if (!dl) {
2c514b71
NB		 2784 if (verbose)
2785 fprintf(stderr, Name ": ddf: %s is not in the "
2786 "same DDF set\n",
2787 dev);
5f8097be
NB		 2788 return 0;
2789 }
2790 e = get_extents(ddf, dl);
2791 maxsize = 0;
2792 i = 0;
2793 if (e) do {
2794 unsigned long long esize;
2795 esize = e[i].start - pos;
2796 if (esize >= maxsize)
2797 maxsize = esize;
2798 pos = e[i].start + e[i].size;
2799 i++;
2800 } while (e[i-1].size);
2801 *freesize = maxsize;
a322f70c
DW		 2802 // FIXME here I am
2803
2804 return 1;
2805}
59e36268 2806
a322f70c
DW		 2807static int load_super_ddf_all(struct supertype *st, int fd,
2808 void **sbp, char *devname, int keep_fd)
2809{
2810 struct mdinfo *sra;
2811 struct ddf_super *super;
2812 struct mdinfo *sd, *best = NULL;
2813 int bestseq = 0;
2814 int seq;
2815 char nm[20];
2816 int dfd;
2817
b526e52d 2818 sra = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE);
a322f70c
DW		 2819 if (!sra)
2820 return 1;
2821 if (sra->array.major_version != -1 ||
2822 sra->array.minor_version != -2 ||
2823 strcmp(sra->text_version, "ddf") != 0)
2824 return 1;
2825
6416d527 2826 if (posix_memalign((void**)&super, 512, sizeof(*super)) != 0)
a322f70c 2827 return 1;
a2349791 2828 memset(super, 0, sizeof(*super));
a322f70c
DW		 2829
2830 /* first, try each device, and choose the best ddf */
2831 for (sd = sra->devs ; sd ; sd = sd->next) {
2832 int rv;
2833 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
7a7cc504
NB		 2834 dfd = dev_open(nm, O_RDONLY);
2835 if (dfd < 0)
a322f70c
DW		 2836 return 2;
2837 rv = load_ddf_headers(dfd, super, NULL);
7a7cc504 2838 close(dfd);
a322f70c
DW		 2839 if (rv == 0) {
2840 seq = __be32_to_cpu(super->active->seq);
2841 if (super->active->openflag)
2842 seq--;
2843 if (!best || seq > bestseq) {
2844 bestseq = seq;
2845 best = sd;
2846 }
2847 }
2848 }
2849 if (!best)
2850 return 1;
2851 /* OK, load this ddf */
2852 sprintf(nm, "%d:%d", best->disk.major, best->disk.minor);
2853 dfd = dev_open(nm, O_RDONLY);
7a7cc504 2854 if (dfd < 0)
a322f70c
DW		 2855 return 1;
2856 load_ddf_headers(dfd, super, NULL);
2857 load_ddf_global(dfd, super, NULL);
2858 close(dfd);
2859 /* Now we need the device-local bits */
2860 for (sd = sra->devs ; sd ; sd = sd->next) {
3d2c4fc7
DW		 2861 int rv;
2862
a322f70c 2863 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
5f8097be 2864 dfd = dev_open(nm, keep_fd? O_RDWR : O_RDONLY);
7a7cc504 2865 if (dfd < 0)
a322f70c 2866 return 2;
3d2c4fc7
DW		 2867 rv = load_ddf_headers(dfd, super, NULL);
2868 if (rv == 0)
2869 rv = load_ddf_local(dfd, super, NULL, keep_fd);
5f8097be 2870 if (!keep_fd) close(dfd);
3d2c4fc7
DW		 2871 if (rv)
2872 return 1;
a322f70c 2873 }
33414a01 2874
a322f70c
DW		 2875 *sbp = super;
2876 if (st->ss == NULL) {
78e44928 2877 st->ss = &super_ddf;
a322f70c
DW		 2878 st->minor_version = 0;
2879 st->max_devs = 512;
75aa18b5 2880 st->container_dev = fd2devnum(fd);
a322f70c 2881 }
352452c3 2882 st->loaded_container = 1;
a322f70c
DW		 2883 return 0;
2884}
0e600426 2885#endif /* MDASSEMBLE */
a322f70c 2886
00bbdbda 2887static struct mdinfo *container_content_ddf(struct supertype *st, char *subarray)
598f0d58
NB		 2888{
2889 /* Given a container loaded by load_super_ddf_all,
2890 * extract information about all the arrays into
2891 * an mdinfo tree.
2892 *
2893 * For each vcl in conflist: create an mdinfo, fill it in,
2894 * then look for matching devices (phys_refnum) in dlist
2895 * and create appropriate device mdinfo.
2896 */
2897 struct ddf_super *ddf = st->sb;
2898 struct mdinfo *rest = NULL;
2899 struct vcl *vc;
2900
2901 for (vc = ddf->conflist ; vc ; vc=vc->next)
2902 {
f21e18ca
N		 2903 unsigned int i;
2904 unsigned int j;
598f0d58 2905 struct mdinfo *this;
00bbdbda
N		 2906 char *ep;
2907
2908 if (subarray &&
2909 (strtoul(subarray, &ep, 10) != vc->vcnum ||
2910 *ep != '\0'))
2911 continue;
2912
598f0d58
NB		 2913 this = malloc(sizeof(*this));
2914 memset(this, 0, sizeof(*this));
2915 this->next = rest;
2916 rest = this;
2917
598f0d58
NB		 2918 this->array.level = map_num1(ddf_level_num, vc->conf.prl);
2919 this->array.raid_disks =
2920 __be16_to_cpu(vc->conf.prim_elmnt_count);
60f18132
NB		 2921 this->array.layout = rlq_to_layout(vc->conf.rlq, vc->conf.prl,
2922 this->array.raid_disks);
598f0d58 2923 this->array.md_minor = -1;
f35f2525
N		 2924 this->array.major_version = -1;
2925 this->array.minor_version = -2;
598f0d58
NB		 2926 this->array.ctime = DECADE +
2927 __be32_to_cpu(*(__u32*)(vc->conf.guid+16));
2928 this->array.utime = DECADE +
2929 __be32_to_cpu(vc->conf.timestamp);
2930 this->array.chunk_size = 512 << vc->conf.chunk_shift;
2931
59e36268 2932 i = vc->vcnum;
7a7cc504
NB		 2933 if ((ddf->virt->entries[i].state & DDF_state_inconsistent) ||
2934 (ddf->virt->entries[i].init_state & DDF_initstate_mask) !=
ed9d66aa 2935 DDF_init_full) {
598f0d58 2936 this->array.state = 0;
ed9d66aa
NB		 2937 this->resync_start = 0;
2938 } else {
598f0d58 2939 this->array.state = 1;
b7528a20 2940 this->resync_start = MaxSector;
ed9d66aa 2941 }
db42fa9b
N		 2942 memcpy(this->name, ddf->virt->entries[i].name, 16);
2943 this->name[16]=0;
2944 for(j=0; j<16; j++)
2945 if (this->name[j] == ' ')
2946 this->name[j] = 0;
598f0d58
NB		 2947
2948 memset(this->uuid, 0, sizeof(this->uuid));
2949 this->component_size = __be64_to_cpu(vc->conf.blocks);
2950 this->array.size = this->component_size / 2;
5f2aace8 2951 this->container_member = i;
598f0d58 2952
c5afc314
N		 2953 ddf->currentconf = vc;
2954 uuid_from_super_ddf(st, this->uuid);
2955 ddf->currentconf = NULL;
2956
60f18132
NB		 2957 sprintf(this->text_version, "/%s/%d",
2958 devnum2devname(st->container_dev),
2959 this->container_member);
2960
f21e18ca 2961 for (i = 0 ; i < ddf->mppe ; i++) {
598f0d58
NB		 2962 struct mdinfo *dev;
2963 struct dl *d;
2964
2965 if (vc->conf.phys_refnum[i] == 0xFFFFFFFF)
2966 continue;
2967
c0d50ca5
N		 2968 this->array.working_disks++;
2969
598f0d58
NB		 2970 for (d = ddf->dlist; d ; d=d->next)
2971 if (d->disk.refnum == vc->conf.phys_refnum[i])
2972 break;
2973 if (d == NULL)
bc17324f
N		 2974 /* Haven't found that one yet, maybe there are others */
2975 continue;
2976
598f0d58
NB		 2977 dev = malloc(sizeof(*dev));
2978 memset(dev, 0, sizeof(*dev));
2979 dev->next = this->devs;
2980 this->devs = dev;
2981
2982 dev->disk.number = __be32_to_cpu(d->disk.refnum);
2983 dev->disk.major = d->major;
2984 dev->disk.minor = d->minor;
2985 dev->disk.raid_disk = i;
2986 dev->disk.state = (1<<MD_DISK_SYNC)|(1<<MD_DISK_ACTIVE);
d23534e4 2987 dev->recovery_start = MaxSector;
598f0d58 2988
120f7677
NB		 2989 dev->events = __be32_to_cpu(ddf->primary.seq);
2990 dev->data_offset = __be64_to_cpu(vc->lba_offset[i]);
598f0d58
NB		 2991 dev->component_size = __be64_to_cpu(vc->conf.blocks);
2992 if (d->devname)
2993 strcpy(dev->name, d->devname);
2994 }
2995 }
2996 return rest;
2997}
2998
955e9ea1 2999static int store_super_ddf(struct supertype *st, int fd)
a322f70c 3000{
955e9ea1 3001 struct ddf_super *ddf = st->sb;
a322f70c 3002 unsigned long long dsize;
6416d527 3003 void *buf;
3d2c4fc7 3004 int rc;
a322f70c 3005
955e9ea1
DW		 3006 if (!ddf)
3007 return 1;
3008
3009 /* ->dlist and ->conflist will be set for updates, currently not
3010 * supported
3011 */
3012 if (ddf->dlist || ddf->conflist)
3013 return 1;
3014
a322f70c
DW		 3015 if (!get_dev_size(fd, NULL, &dsize))
3016 return 1;
3017
3d2c4fc7
DW		 3018 if (posix_memalign(&buf, 512, 512) != 0)
3019 return 1;
6416d527
NB		 3020 memset(buf, 0, 512);
3021
a322f70c 3022 lseek64(fd, dsize-512, 0);
3d2c4fc7 3023 rc = write(fd, buf, 512);
6416d527 3024 free(buf);
3d2c4fc7
DW		 3025 if (rc < 0)
3026 return 1;
a322f70c
DW		 3027 return 0;
3028}
3029
a19c88b8
NB		 3030static int compare_super_ddf(struct supertype *st, struct supertype *tst)
3031{
3032 /*
3033 * return:
3034 * 0 same, or first was empty, and second was copied
3035 * 1 second had wrong number
3036 * 2 wrong uuid
3037 * 3 wrong other info
3038 */
3039 struct ddf_super *first = st->sb;
3040 struct ddf_super *second = tst->sb;
3041
3042 if (!first) {
3043 st->sb = tst->sb;
3044 tst->sb = NULL;
3045 return 0;
3046 }
3047
3048 if (memcmp(first->anchor.guid, second->anchor.guid, DDF_GUID_LEN) != 0)
3049 return 2;
3050
3051 /* FIXME should I look at anything else? */
3052 return 0;
3053}
3054
0e600426 3055#ifndef MDASSEMBLE
4e5528c6
NB		 3056/*
3057 * A new array 'a' has been started which claims to be instance 'inst'
3058 * within container 'c'.
3059 * We need to confirm that the array matches the metadata in 'c' so
3060 * that we don't corrupt any metadata.
3061 */
cba0191b 3062static int ddf_open_new(struct supertype *c, struct active_array *a, char *inst)
549e9569 3063{
2c514b71 3064 dprintf("ddf: open_new %s\n", inst);
cba0191b 3065 a->info.container_member = atoi(inst);
549e9569
NB		 3066 return 0;
3067}
3068
4e5528c6
NB		 3069/*
3070 * The array 'a' is to be marked clean in the metadata.
ed9d66aa 3071 * If '->resync_start' is not ~(unsigned long long)0, then the array is only
4e5528c6
NB		 3072 * clean up to the point (in sectors). If that cannot be recorded in the
3073 * metadata, then leave it as dirty.
3074 *
3075 * For DDF, we need to clear the DDF_state_inconsistent bit in the
3076 * !global! virtual_disk.virtual_entry structure.
3077 */
01f157d7 3078static int ddf_set_array_state(struct active_array *a, int consistent)
549e9569 3079{
4e5528c6
NB		 3080 struct ddf_super *ddf = a->container->sb;
3081 int inst = a->info.container_member;
18a2f463 3082 int old = ddf->virt->entries[inst].state;
01f157d7
N		 3083 if (consistent == 2) {
3084 /* Should check if a recovery should be started FIXME */
3085 consistent = 1;
b7941fd6 3086 if (!is_resync_complete(&a->info))
01f157d7
N		 3087 consistent = 0;
3088 }
ed9d66aa
NB		 3089 if (consistent)
3090 ddf->virt->entries[inst].state &= ~DDF_state_inconsistent;
3091 else
4e5528c6 3092 ddf->virt->entries[inst].state |= DDF_state_inconsistent;
18a2f463
NB		 3093 if (old != ddf->virt->entries[inst].state)
3094 ddf->updates_pending = 1;
3095
3096 old = ddf->virt->entries[inst].init_state;
ed9d66aa 3097 ddf->virt->entries[inst].init_state &= ~DDF_initstate_mask;
b7941fd6 3098 if (is_resync_complete(&a->info))
ed9d66aa 3099 ddf->virt->entries[inst].init_state |= DDF_init_full;
b7941fd6 3100 else if (a->info.resync_start == 0)
ed9d66aa 3101 ddf->virt->entries[inst].init_state |= DDF_init_not;
4e5528c6 3102 else
ed9d66aa 3103 ddf->virt->entries[inst].init_state |= DDF_init_quick;
18a2f463
NB		 3104 if (old != ddf->virt->entries[inst].init_state)
3105 ddf->updates_pending = 1;
ed9d66aa 3106
2c514b71 3107 dprintf("ddf mark %d %s %llu\n", inst, consistent?"clean":"dirty",
b7941fd6 3108 a->info.resync_start);
01f157d7 3109 return consistent;
fd7cde1b
DW		 3110}
3111
7a7cc504
NB		 3112/*
3113 * The state of each disk is stored in the global phys_disk structure
3114 * in phys_disk.entries[n].state.
3115 * This makes various combinations awkward.
3116 * - When a device fails in any array, it must be failed in all arrays
3117 * that include a part of this device.
3118 * - When a component is rebuilding, we cannot include it officially in the
3119 * array unless this is the only array that uses the device.
3120 *
3121 * So: when transitioning:
3122 * Online -> failed, just set failed flag. monitor will propagate
3123 * spare -> online, the device might need to be added to the array.
3124 * spare -> failed, just set failed. Don't worry if in array or not.
3125 */
8d45d196 3126static void ddf_set_disk(struct active_array *a, int n, int state)
549e9569 3127{
7a7cc504 3128 struct ddf_super *ddf = a->container->sb;
f21e18ca 3129 unsigned int inst = a->info.container_member;
7a7cc504
NB		 3130 struct vd_config *vc = find_vdcr(ddf, inst);
3131 int pd = find_phys(ddf, vc->phys_refnum[n]);
3132 int i, st, working;
3133
3134 if (vc == NULL) {
2c514b71 3135 dprintf("ddf: cannot find instance %d!!\n", inst);
7a7cc504
NB		 3136 return;
3137 }
3138 if (pd < 0) {
3139 /* disk doesn't currently exist. If it is now in_sync,
3140 * insert it. */
3141 if ((state & DS_INSYNC) && ! (state & DS_FAULTY)) {
3142 /* Find dev 'n' in a->info->devs, determine the
3143 * ddf refnum, and set vc->phys_refnum and update
3144 * phys->entries[]
3145 */
3146 /* FIXME */
3147 }
3148 } else {
18a2f463 3149 int old = ddf->phys->entries[pd].state;
7a7cc504
NB		 3150 if (state & DS_FAULTY)
3151 ddf->phys->entries[pd].state |= __cpu_to_be16(DDF_Failed);
3152 if (state & DS_INSYNC) {
3153 ddf->phys->entries[pd].state |= __cpu_to_be16(DDF_Online);
3154 ddf->phys->entries[pd].state &= __cpu_to_be16(~DDF_Rebuilding);
3155 }
18a2f463
NB		 3156 if (old != ddf->phys->entries[pd].state)
3157 ddf->updates_pending = 1;
7a7cc504
NB		 3158 }
3159
2c514b71 3160 dprintf("ddf: set_disk %d to %x\n", n, state);
7e1432fb 3161
7a7cc504
NB		 3162 /* Now we need to check the state of the array and update
3163 * virtual_disk.entries[n].state.
3164 * It needs to be one of "optimal", "degraded", "failed".
3165 * I don't understand 'deleted' or 'missing'.
3166 */
3167 working = 0;
3168 for (i=0; i < a->info.array.raid_disks; i++) {
3169 pd = find_phys(ddf, vc->phys_refnum[i]);
3170 if (pd < 0)
3171 continue;
57632f4a
NB		 3172 st = __be16_to_cpu(ddf->phys->entries[pd].state);
3173 if ((st & (DDF_Online|DDF_Failed|DDF_Rebuilding))
7a7cc504
NB		 3174 == DDF_Online)
3175 working++;
3176 }
3177 state = DDF_state_degraded;
3178 if (working == a->info.array.raid_disks)
3179 state = DDF_state_optimal;
3180 else switch(vc->prl) {
3181 case DDF_RAID0:
3182 case DDF_CONCAT:
3183 case DDF_JBOD:
3184 state = DDF_state_failed;
3185 break;
3186 case DDF_RAID1:
3187 if (working == 0)
3188 state = DDF_state_failed;
3189 break;
3190 case DDF_RAID4:
3191 case DDF_RAID5:
3192 if (working < a->info.array.raid_disks-1)
3193 state = DDF_state_failed;
3194 break;
3195 case DDF_RAID6:
3196 if (working < a->info.array.raid_disks-2)
3197 state = DDF_state_failed;
3198 else if (working == a->info.array.raid_disks-1)
3199 state = DDF_state_part_optimal;
3200 break;
3201 }
3202
18a2f463
NB		 3203 if (ddf->virt->entries[inst].state !=
3204 ((ddf->virt->entries[inst].state & ~DDF_state_mask)
3205 | state)) {
3206
3207 ddf->virt->entries[inst].state =
3208 (ddf->virt->entries[inst].state & ~DDF_state_mask)
3209 | state;
3210 ddf->updates_pending = 1;
3211 }
7a7cc504 3212
549e9569
NB		 3213}
3214
2e735d19 3215static void ddf_sync_metadata(struct supertype *st)
549e9569 3216{
7a7cc504
NB		 3217
3218 /*
3219 * Write all data to all devices.
3220 * Later, we might be able to track whether only local changes
3221 * have been made, or whether any global data has been changed,
3222 * but ddf is sufficiently weird that it probably always
3223 * changes global data ....
3224 */
18a2f463
NB		 3225 struct ddf_super *ddf = st->sb;
3226 if (!ddf->updates_pending)
3227 return;
3228 ddf->updates_pending = 0;
2e735d19 3229 __write_init_super_ddf(st, 0);
2c514b71 3230 dprintf("ddf: sync_metadata\n");
549e9569
NB		 3231}
3232
88c164f4
NB		 3233static void ddf_process_update(struct supertype *st,
3234 struct metadata_update *update)
3235{
3236 /* Apply this update to the metadata.
3237 * The first 4 bytes are a DDF_*_MAGIC which guides
3238 * our actions.
3239 * Possible update are:
3240 * DDF_PHYS_RECORDS_MAGIC
3241 * Add a new physical device. Changes to this record
3242 * only happen implicitly.
3243 * used_pdes is the device number.
3244 * DDF_VIRT_RECORDS_MAGIC
3245 * Add a new VD. Possibly also change the 'access' bits.
3246 * populated_vdes is the entry number.
3247 * DDF_VD_CONF_MAGIC
3248 * New or updated VD. the VIRT_RECORD must already
3249 * exist. For an update, phys_refnum and lba_offset
3250 * (at least) are updated, and the VD_CONF must
3251 * be written to precisely those devices listed with
3252 * a phys_refnum.
3253 * DDF_SPARE_ASSIGN_MAGIC
3254 * replacement Spare Assignment Record... but for which device?
3255 *
3256 * So, e.g.:
3257 * - to create a new array, we send a VIRT_RECORD and
3258 * a VD_CONF. Then assemble and start the array.
3259 * - to activate a spare we send a VD_CONF to add the phys_refnum
3260 * and offset. This will also mark the spare as active with
3261 * a spare-assignment record.
3262 */
3263 struct ddf_super *ddf = st->sb;
3264 __u32 *magic = (__u32*)update->buf;
3265 struct phys_disk *pd;
3266 struct virtual_disk *vd;
3267 struct vd_config *vc;
3268 struct vcl *vcl;
3269 struct dl *dl;
f21e18ca
N		 3270 unsigned int mppe;
3271 unsigned int ent;
88c164f4 3272
2c514b71 3273 dprintf("Process update %x\n", *magic);
7e1432fb 3274
88c164f4
NB		 3275 switch (*magic) {
3276 case DDF_PHYS_RECORDS_MAGIC:
3277
3278 if (update->len != (sizeof(struct phys_disk) +
3279 sizeof(struct phys_disk_entry)))
3280 return;
3281 pd = (struct phys_disk*)update->buf;
3282
3283 ent = __be16_to_cpu(pd->used_pdes);
3284 if (ent >= __be16_to_cpu(ddf->phys->max_pdes))
3285 return;
3286 if (!all_ff(ddf->phys->entries[ent].guid))
3287 return;
3288 ddf->phys->entries[ent] = pd->entries[0];
3289 ddf->phys->used_pdes = __cpu_to_be16(1 +
3290 __be16_to_cpu(ddf->phys->used_pdes));
18a2f463 3291 ddf->updates_pending = 1;
2cc2983d
N		 3292 if (ddf->add_list) {
3293 struct active_array *a;
3294 struct dl *al = ddf->add_list;
3295 ddf->add_list = al->next;
3296
3297 al->next = ddf->dlist;
3298 ddf->dlist = al;
3299
3300 /* As a device has been added, we should check
3301 * for any degraded devices that might make
3302 * use of this spare */
3303 for (a = st->arrays ; a; a=a->next)
3304 a->check_degraded = 1;
3305 }
88c164f4
NB		 3306 break;
3307
3308 case DDF_VIRT_RECORDS_MAGIC:
3309
3310 if (update->len != (sizeof(struct virtual_disk) +
3311 sizeof(struct virtual_entry)))
3312 return;
3313 vd = (struct virtual_disk*)update->buf;
3314
3315 ent = __be16_to_cpu(vd->populated_vdes);
3316 if (ent >= __be16_to_cpu(ddf->virt->max_vdes))
3317 return;
3318 if (!all_ff(ddf->virt->entries[ent].guid))
3319 return;
3320 ddf->virt->entries[ent] = vd->entries[0];
3321 ddf->virt->populated_vdes = __cpu_to_be16(1 +
3322 __be16_to_cpu(ddf->virt->populated_vdes));
18a2f463 3323 ddf->updates_pending = 1;
88c164f4
NB		 3324 break;
3325
3326 case DDF_VD_CONF_MAGIC:
2c514b71 3327 dprintf("len %d %d\n", update->len, ddf->conf_rec_len);
88c164f4
NB		 3328
3329 mppe = __be16_to_cpu(ddf->anchor.max_primary_element_entries);
f21e18ca 3330 if ((unsigned)update->len != ddf->conf_rec_len * 512)
88c164f4
NB		 3331 return;
3332 vc = (struct vd_config*)update->buf;
3333 for (vcl = ddf->conflist; vcl ; vcl = vcl->next)
3334 if (memcmp(vcl->conf.guid, vc->guid, DDF_GUID_LEN) == 0)
3335 break;
2c514b71 3336 dprintf("vcl = %p\n", vcl);
88c164f4
NB		 3337 if (vcl) {
3338 /* An update, just copy the phys_refnum and lba_offset
3339 * fields
3340 */
3341 memcpy(vcl->conf.phys_refnum, vc->phys_refnum,
3342 mppe * (sizeof(__u32) + sizeof(__u64)));
3343 } else {
3344 /* A new VD_CONF */
e6b9548d
DW		 3345 if (!update->space)
3346 return;
88c164f4
NB		 3347 vcl = update->space;
3348 update->space = NULL;
3349 vcl->next = ddf->conflist;
edd8d13c 3350 memcpy(&vcl->conf, vc, update->len);
88c164f4
NB		 3351 vcl->lba_offset = (__u64*)
3352 &vcl->conf.phys_refnum[mppe];
3353 ddf->conflist = vcl;
3354 }
3355 /* Now make sure vlist is correct for each dl. */
3356 for (dl = ddf->dlist; dl; dl = dl->next) {
f21e18ca
N		 3357 unsigned int dn;
3358 unsigned int vn = 0;
88c164f4
NB		 3359 for (vcl = ddf->conflist; vcl ; vcl = vcl->next)
3360 for (dn=0; dn < ddf->mppe ; dn++)
3361 if (vcl->conf.phys_refnum[dn] ==
3362 dl->disk.refnum) {
2c514b71
NB		 3363 dprintf("dev %d has %p at %d\n",
3364 dl->pdnum, vcl, vn);
88c164f4
NB		 3365 dl->vlist[vn++] = vcl;
3366 break;
3367 }
3368 while (vn < ddf->max_part)
3369 dl->vlist[vn++] = NULL;
7e1432fb
NB		 3370 if (dl->vlist[0]) {
3371 ddf->phys->entries[dl->pdnum].type &=
3372 ~__cpu_to_be16(DDF_Global_Spare);
3373 ddf->phys->entries[dl->pdnum].type |=
3374 __cpu_to_be16(DDF_Active_in_VD);
3375 }
3376 if (dl->spare) {
3377 ddf->phys->entries[dl->pdnum].type &=
3378 ~__cpu_to_be16(DDF_Global_Spare);
3379 ddf->phys->entries[dl->pdnum].type |=
3380 __cpu_to_be16(DDF_Spare);
3381 }
3382 if (!dl->vlist[0] && !dl->spare) {
3383 ddf->phys->entries[dl->pdnum].type |=
3384 __cpu_to_be16(DDF_Global_Spare);
3385 ddf->phys->entries[dl->pdnum].type &=
3386 ~__cpu_to_be16(DDF_Spare |
3387 DDF_Active_in_VD);
3388 }
88c164f4 3389 }
18a2f463 3390 ddf->updates_pending = 1;
88c164f4
NB		 3391 break;
3392 case DDF_SPARE_ASSIGN_MAGIC:
3393 default: break;
3394 }
3395}
3396
edd8d13c
NB		 3397static void ddf_prepare_update(struct supertype *st,
3398 struct metadata_update *update)
3399{
3400 /* This update arrived at managemon.
3401 * We are about to pass it to monitor.
3402 * If a malloc is needed, do it here.
3403 */
3404 struct ddf_super *ddf = st->sb;
3405 __u32 *magic = (__u32*)update->buf;
3406 if (*magic == DDF_VD_CONF_MAGIC)
e6b9548d 3407 if (posix_memalign(&update->space, 512,
6416d527 3408 offsetof(struct vcl, conf)
e6b9548d
DW		 3409 + ddf->conf_rec_len * 512) != 0)
3410 update->space = NULL;
edd8d13c
NB		 3411}
3412
7e1432fb
NB		 3413/*
3414 * Check if the array 'a' is degraded but not failed.
3415 * If it is, find as many spares as are available and needed and
3416 * arrange for their inclusion.
3417 * We only choose devices which are not already in the array,
3418 * and prefer those with a spare-assignment to this array.
3419 * otherwise we choose global spares - assuming always that
3420 * there is enough room.
3421 * For each spare that we assign, we return an 'mdinfo' which
3422 * describes the position for the device in the array.
3423 * We also add to 'updates' a DDF_VD_CONF_MAGIC update with
3424 * the new phys_refnum and lba_offset values.
3425 *
3426 * Only worry about BVDs at the moment.
3427 */
3428static struct mdinfo *ddf_activate_spare(struct active_array *a,
3429 struct metadata_update **updates)
3430{
3431 int working = 0;
3432 struct mdinfo *d;
3433 struct ddf_super *ddf = a->container->sb;
3434 int global_ok = 0;
3435 struct mdinfo *rv = NULL;
3436 struct mdinfo *di;
3437 struct metadata_update *mu;
3438 struct dl *dl;
3439 int i;
3440 struct vd_config *vc;
3441 __u64 *lba;
3442
7e1432fb
NB		 3443 for (d = a->info.devs ; d ; d = d->next) {
3444 if ((d->curr_state & DS_FAULTY) &&
3445 d->state_fd >= 0)
3446 /* wait for Removal to happen */
3447 return NULL;
3448 if (d->state_fd >= 0)
3449 working ++;
3450 }
3451
2c514b71
NB		 3452 dprintf("ddf_activate: working=%d (%d) level=%d\n", working, a->info.array.raid_disks,
3453 a->info.array.level);
7e1432fb
NB		 3454 if (working == a->info.array.raid_disks)
3455 return NULL; /* array not degraded */
3456 switch (a->info.array.level) {
3457 case 1:
3458 if (working == 0)
3459 return NULL; /* failed */
3460 break;
3461 case 4:
3462 case 5:
3463 if (working < a->info.array.raid_disks - 1)
3464 return NULL; /* failed */
3465 break;
3466 case 6:
3467 if (working < a->info.array.raid_disks - 2)
3468 return NULL; /* failed */
3469 break;
3470 default: /* concat or stripe */
3471 return NULL; /* failed */
3472 }
3473
3474 /* For each slot, if it is not working, find a spare */
3475 dl = ddf->dlist;
3476 for (i = 0; i < a->info.array.raid_disks; i++) {
3477 for (d = a->info.devs ; d ; d = d->next)
3478 if (d->disk.raid_disk == i)
3479 break;
2c514b71 3480 dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0);
7e1432fb
NB		 3481 if (d && (d->state_fd >= 0))
3482 continue;
3483
3484 /* OK, this device needs recovery. Find a spare */
3485 again:
3486 for ( ; dl ; dl = dl->next) {
3487 unsigned long long esize;
3488 unsigned long long pos;
3489 struct mdinfo *d2;
3490 int is_global = 0;
3491 int is_dedicated = 0;
3492 struct extent *ex;
f21e18ca 3493 unsigned int j;
7e1432fb
NB		 3494 /* If in this array, skip */
3495 for (d2 = a->info.devs ; d2 ; d2 = d2->next)
3496 if (d2->disk.major == dl->major &&
3497 d2->disk.minor == dl->minor) {
2c514b71 3498 dprintf("%x:%x already in array\n", dl->major, dl->minor);
7e1432fb
NB		 3499 break;
3500 }
3501 if (d2)
3502 continue;
3503 if (ddf->phys->entries[dl->pdnum].type &
3504 __cpu_to_be16(DDF_Spare)) {
3505 /* Check spare assign record */
3506 if (dl->spare) {
3507 if (dl->spare->type & DDF_spare_dedicated) {
3508 /* check spare_ents for guid */
3509 for (j = 0 ;
3510 j < __be16_to_cpu(dl->spare->populated);
3511 j++) {
3512 if (memcmp(dl->spare->spare_ents[j].guid,
3513 ddf->virt->entries[a->info.container_member].guid,
3514 DDF_GUID_LEN) == 0)
3515 is_dedicated = 1;
3516 }
3517 } else
3518 is_global = 1;
3519 }
3520 } else if (ddf->phys->entries[dl->pdnum].type &
3521 __cpu_to_be16(DDF_Global_Spare)) {
3522 is_global = 1;
3523 }
3524 if ( ! (is_dedicated ||
3525 (is_global && global_ok))) {
2c514b71 3526 dprintf("%x:%x not suitable: %d %d\n", dl->major, dl->minor,
7e1432fb
NB		 3527 is_dedicated, is_global);
3528 continue;
3529 }
3530
3531 /* We are allowed to use this device - is there space?
3532 * We need a->info.component_size sectors */
3533 ex = get_extents(ddf, dl);
3534 if (!ex) {
2c514b71 3535 dprintf("cannot get extents\n");
7e1432fb
NB		 3536 continue;
3537 }
3538 j = 0; pos = 0;
3539 esize = 0;
3540
3541 do {
3542 esize = ex[j].start - pos;
3543 if (esize >= a->info.component_size)
3544 break;
3545 pos = ex[i].start + ex[i].size;
3546 i++;
3547 } while (ex[i-1].size);
3548
3549 free(ex);
3550 if (esize < a->info.component_size) {
2c514b71
NB		 3551 dprintf("%x:%x has no room: %llu %llu\n", dl->major, dl->minor,
3552 esize, a->info.component_size);
7e1432fb
NB		 3553 /* No room */
3554 continue;
3555 }
3556
3557 /* Cool, we have a device with some space at pos */
3558 di = malloc(sizeof(*di));
79244939
DW		 3559 if (!di)
3560 continue;
7e1432fb
NB		 3561 memset(di, 0, sizeof(*di));
3562 di->disk.number = i;
3563 di->disk.raid_disk = i;
3564 di->disk.major = dl->major;
3565 di->disk.minor = dl->minor;
3566 di->disk.state = 0;
d23534e4 3567 di->recovery_start = 0;
7e1432fb
NB		 3568 di->data_offset = pos;
3569 di->component_size = a->info.component_size;
3570 di->container_member = dl->pdnum;
3571 di->next = rv;
3572 rv = di;
2c514b71
NB		 3573 dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor,
3574 i, pos);
7e1432fb
NB		 3575
3576 break;
3577 }
3578 if (!dl && ! global_ok) {
3579 /* not enough dedicated spares, try global */
3580 global_ok = 1;
3581 dl = ddf->dlist;
3582 goto again;
3583 }
3584 }
3585
3586 if (!rv)
3587 /* No spares found */
3588 return rv;
3589 /* Now 'rv' has a list of devices to return.
3590 * Create a metadata_update record to update the
3591 * phys_refnum and lba_offset values
3592 */
904c1ef7 3593 mu = malloc(sizeof(*mu));
79244939
DW		 3594 if (mu && posix_memalign(&mu->space, 512, sizeof(struct vcl)) != 0) {
3595 free(mu);
3596 mu = NULL;
3597 }
3598 if (!mu) {
3599 while (rv) {
3600 struct mdinfo *n = rv->next;
3601
3602 free(rv);
3603 rv = n;
3604 }
3605 return NULL;
3606 }
3607
904c1ef7 3608 mu->buf = malloc(ddf->conf_rec_len * 512);
7e1432fb
NB		 3609 mu->len = ddf->conf_rec_len;
3610 mu->next = *updates;
3611 vc = find_vdcr(ddf, a->info.container_member);
3612 memcpy(mu->buf, vc, ddf->conf_rec_len * 512);
3613
3614 vc = (struct vd_config*)mu->buf;
3615 lba = (__u64*)&vc->phys_refnum[ddf->mppe];
3616 for (di = rv ; di ; di = di->next) {
3617 vc->phys_refnum[di->disk.raid_disk] =
3618 ddf->phys->entries[dl->pdnum].refnum;
3619 lba[di->disk.raid_disk] = di->data_offset;
3620 }
3621 *updates = mu;
3622 return rv;
3623}
0e600426 3624#endif /* MDASSEMBLE */
7e1432fb 3625
b640a252
N		 3626static int ddf_level_to_layout(int level)
3627{
3628 switch(level) {
3629 case 0:
3630 case 1:
3631 return 0;
3632 case 5:
3633 return ALGORITHM_LEFT_SYMMETRIC;
3634 case 6:
3635 return ALGORITHM_ROTATING_N_CONTINUE;
3636 case 10:
3637 return 0x102;
3638 default:
3639 return UnSet;
3640 }
3641}
3642
a322f70c
DW		 3643struct superswitch super_ddf = {
3644#ifndef MDASSEMBLE
3645 .examine_super = examine_super_ddf,
3646 .brief_examine_super = brief_examine_super_ddf,
4737ae25 3647 .brief_examine_subarrays = brief_examine_subarrays_ddf,
bceedeec 3648 .export_examine_super = export_examine_super_ddf,
a322f70c
DW		 3649 .detail_super = detail_super_ddf,
3650 .brief_detail_super = brief_detail_super_ddf,
3651 .validate_geometry = validate_geometry_ddf,
78e44928 3652 .write_init_super = write_init_super_ddf,
0e600426 3653 .add_to_super = add_to_super_ddf,
a322f70c
DW		 3654#endif
3655 .match_home = match_home_ddf,
3656 .uuid_from_super= uuid_from_super_ddf,
3657 .getinfo_super = getinfo_super_ddf,
3658 .update_super = update_super_ddf,
3659
3660 .avail_size = avail_size_ddf,
3661
a19c88b8
NB		 3662 .compare_super = compare_super_ddf,
3663
a322f70c 3664 .load_super = load_super_ddf,
ba7eb04f 3665 .init_super = init_super_ddf,
955e9ea1 3666 .store_super = store_super_ddf,
a322f70c
DW		 3667 .free_super = free_super_ddf,
3668 .match_metadata_desc = match_metadata_desc_ddf,
78e44928 3669 .container_content = container_content_ddf,
b640a252 3670 .default_layout = ddf_level_to_layout,
a322f70c 3671
a322f70c 3672 .external = 1,
549e9569 3673
0e600426 3674#ifndef MDASSEMBLE
549e9569
NB		 3675/* for mdmon */
3676 .open_new = ddf_open_new,
ed9d66aa 3677 .set_array_state= ddf_set_array_state,
549e9569
NB		 3678 .set_disk = ddf_set_disk,
3679 .sync_metadata = ddf_sync_metadata,
88c164f4 3680 .process_update = ddf_process_update,
edd8d13c 3681 .prepare_update = ddf_prepare_update,
7e1432fb 3682 .activate_spare = ddf_activate_spare,
0e600426 3683#endif
4cce4069 3684 .name = "ddf",
a322f70c 3685};
Unnamed repository; edit this file 'description' to name the repository.