]> 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
mdmon: pass symbolic name to mdmon instead of device name.
[thirdparty/mdadm.git] / super-ddf.c
CommitLineData
a322f70c
DW		 1/*
2 * mdadm - manage Linux "md" devices aka RAID arrays.
3 *
4 * Copyright (C) 2006-2007 Neil Brown <neilb@suse.de>
5 *
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2 of the License, or
10 * (at your option) any later version.
11 *
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
16 *
17 * You should have received a copy of the GNU General Public License
18 * along with this program; if not, write to the Free Software
19 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20 *
21 * Author: Neil Brown
22 * Email: <neil@brown.name>
23 *
24 * Specifications for DDF takes from Common RAID DDF Specification Revision 1.2
25 * (July 28 2006). Reused by permission of SNIA.
26 */
27
28#define HAVE_STDINT_H 1
29#include "mdadm.h"
549e9569 30#include "mdmon.h"
a322f70c
DW		 31#include "sha1.h"
32#include <values.h>
33
a322f70c
DW		 34/* a non-official T10 name for creation GUIDs */
35static char T10[] = "Linux-MD";
36
37/* DDF timestamps are 1980 based, so we need to add
38 * second-in-decade-of-seventies to convert to linux timestamps.
39 * 10 years with 2 leap years.
40 */
41#define DECADE (3600*24*(365*10+2))
42unsigned long crc32(
43 unsigned long crc,
44 const unsigned char *buf,
45 unsigned len);
46
47/* The DDF metadata handling.
48 * DDF metadata lives at the end of the device.
49 * The last 512 byte block provides an 'anchor' which is used to locate
50 * the rest of the metadata which usually lives immediately behind the anchor.
51 *
52 * Note:
53 * - all multibyte numeric fields are bigendian.
54 * - all strings are space padded.
55 *
56 */
57
58/* Primary Raid Level (PRL) */
59#define DDF_RAID0 0x00
60#define DDF_RAID1 0x01
61#define DDF_RAID3 0x03
62#define DDF_RAID4 0x04
63#define DDF_RAID5 0x05
64#define DDF_RAID1E 0x11
65#define DDF_JBOD 0x0f
66#define DDF_CONCAT 0x1f
67#define DDF_RAID5E 0x15
68#define DDF_RAID5EE 0x25
59e36268 69#define DDF_RAID6 0x06
a322f70c
DW		 70
71/* Raid Level Qualifier (RLQ) */
72#define DDF_RAID0_SIMPLE 0x00
73#define DDF_RAID1_SIMPLE 0x00 /* just 2 devices in this plex */
74#define DDF_RAID1_MULTI 0x01 /* exactly 3 devices in this plex */
75#define DDF_RAID3_0 0x00 /* parity in first extent */
76#define DDF_RAID3_N 0x01 /* parity in last extent */
77#define DDF_RAID4_0 0x00 /* parity in first extent */
78#define DDF_RAID4_N 0x01 /* parity in last extent */
79/* these apply to raid5e and raid5ee as well */
80#define DDF_RAID5_0_RESTART 0x00 /* same as 'right asymmetric' - layout 1 */
59e36268 81#define DDF_RAID6_0_RESTART 0x01 /* raid6 different from raid5 here!!! */
a322f70c
DW		 82#define DDF_RAID5_N_RESTART 0x02 /* same as 'left asymmetric' - layout 0 */
83#define DDF_RAID5_N_CONTINUE 0x03 /* same as 'left symmetric' - layout 2 */
84
85#define DDF_RAID1E_ADJACENT 0x00 /* raid10 nearcopies==2 */
86#define DDF_RAID1E_OFFSET 0x01 /* raid10 offsetcopies==2 */
87
88/* Secondary RAID Level (SRL) */
89#define DDF_2STRIPED 0x00 /* This is weirder than RAID0 !! */
90#define DDF_2MIRRORED 0x01
91#define DDF_2CONCAT 0x02
92#define DDF_2SPANNED 0x03 /* This is also weird - be careful */
93
94/* Magic numbers */
95#define DDF_HEADER_MAGIC __cpu_to_be32(0xDE11DE11)
96#define DDF_CONTROLLER_MAGIC __cpu_to_be32(0xAD111111)
97#define DDF_PHYS_RECORDS_MAGIC __cpu_to_be32(0x22222222)
98#define DDF_PHYS_DATA_MAGIC __cpu_to_be32(0x33333333)
99#define DDF_VIRT_RECORDS_MAGIC __cpu_to_be32(0xDDDDDDDD)
100#define DDF_VD_CONF_MAGIC __cpu_to_be32(0xEEEEEEEE)
101#define DDF_SPARE_ASSIGN_MAGIC __cpu_to_be32(0x55555555)
102#define DDF_VU_CONF_MAGIC __cpu_to_be32(0x88888888)
103#define DDF_VENDOR_LOG_MAGIC __cpu_to_be32(0x01dBEEF0)
104#define DDF_BBM_LOG_MAGIC __cpu_to_be32(0xABADB10C)
105
106#define DDF_GUID_LEN 24
59e36268
NB		 107#define DDF_REVISION_0 "01.00.00"
108#define DDF_REVISION_2 "01.02.00"
a322f70c
DW		 109
110struct ddf_header {
88c164f4 111 __u32 magic; /* DDF_HEADER_MAGIC */
a322f70c
DW		 112 __u32 crc;
113 char guid[DDF_GUID_LEN];
59e36268 114 char revision[8]; /* 01.02.00 */
a322f70c
DW		 115 __u32 seq; /* starts at '1' */
116 __u32 timestamp;
117 __u8 openflag;
118 __u8 foreignflag;
119 __u8 enforcegroups;
120 __u8 pad0; /* 0xff */
121 __u8 pad1[12]; /* 12 * 0xff */
122 /* 64 bytes so far */
123 __u8 header_ext[32]; /* reserved: fill with 0xff */
124 __u64 primary_lba;
125 __u64 secondary_lba;
126 __u8 type;
127 __u8 pad2[3]; /* 0xff */
128 __u32 workspace_len; /* sectors for vendor space -
129 * at least 32768(sectors) */
130 __u64 workspace_lba;
131 __u16 max_pd_entries; /* one of 15, 63, 255, 1023, 4095 */
132 __u16 max_vd_entries; /* 2^(4,6,8,10,12)-1 : i.e. as above */
133 __u16 max_partitions; /* i.e. max num of configuration
134 record entries per disk */
135 __u16 config_record_len; /* 1 +ROUNDUP(max_primary_element_entries
136 *12/512) */
137 __u16 max_primary_element_entries; /* 16, 64, 256, 1024, or 4096 */
138 __u8 pad3[54]; /* 0xff */
139 /* 192 bytes so far */
140 __u32 controller_section_offset;
141 __u32 controller_section_length;
142 __u32 phys_section_offset;
143 __u32 phys_section_length;
144 __u32 virt_section_offset;
145 __u32 virt_section_length;
146 __u32 config_section_offset;
147 __u32 config_section_length;
148 __u32 data_section_offset;
149 __u32 data_section_length;
150 __u32 bbm_section_offset;
151 __u32 bbm_section_length;
152 __u32 diag_space_offset;
153 __u32 diag_space_length;
154 __u32 vendor_offset;
155 __u32 vendor_length;
156 /* 256 bytes so far */
157 __u8 pad4[256]; /* 0xff */
158};
159
160/* type field */
161#define DDF_HEADER_ANCHOR 0x00
162#define DDF_HEADER_PRIMARY 0x01
163#define DDF_HEADER_SECONDARY 0x02
164
165/* The content of the 'controller section' - global scope */
166struct ddf_controller_data {
88c164f4 167 __u32 magic; /* DDF_CONTROLLER_MAGIC */
a322f70c
DW		 168 __u32 crc;
169 char guid[DDF_GUID_LEN];
170 struct controller_type {
171 __u16 vendor_id;
172 __u16 device_id;
173 __u16 sub_vendor_id;
174 __u16 sub_device_id;
175 } type;
176 char product_id[16];
177 __u8 pad[8]; /* 0xff */
178 __u8 vendor_data[448];
179};
180
181/* The content of phys_section - global scope */
182struct phys_disk {
88c164f4 183 __u32 magic; /* DDF_PHYS_RECORDS_MAGIC */
a322f70c
DW		 184 __u32 crc;
185 __u16 used_pdes;
186 __u16 max_pdes;
187 __u8 pad[52];
188 struct phys_disk_entry {
189 char guid[DDF_GUID_LEN];
190 __u32 refnum;
191 __u16 type;
192 __u16 state;
193 __u64 config_size; /* DDF structures must be after here */
194 char path[18]; /* another horrible structure really */
195 __u8 pad[6];
196 } entries[0];
197};
198
199/* phys_disk_entry.type is a bitmap - bigendian remember */
200#define DDF_Forced_PD_GUID 1
201#define DDF_Active_in_VD 2
88c164f4 202#define DDF_Global_Spare 4 /* VD_CONF records are ignored */
a322f70c
DW		 203#define DDF_Spare 8 /* overrides Global_spare */
204#define DDF_Foreign 16
205#define DDF_Legacy 32 /* no DDF on this device */
206
207#define DDF_Interface_mask 0xf00
208#define DDF_Interface_SCSI 0x100
209#define DDF_Interface_SAS 0x200
210#define DDF_Interface_SATA 0x300
211#define DDF_Interface_FC 0x400
212
213/* phys_disk_entry.state is a bigendian bitmap */
214#define DDF_Online 1
215#define DDF_Failed 2 /* overrides 1,4,8 */
216#define DDF_Rebuilding 4
217#define DDF_Transition 8
218#define DDF_SMART 16
219#define DDF_ReadErrors 32
220#define DDF_Missing 64
221
222/* The content of the virt_section global scope */
223struct virtual_disk {
88c164f4 224 __u32 magic; /* DDF_VIRT_RECORDS_MAGIC */
a322f70c
DW		 225 __u32 crc;
226 __u16 populated_vdes;
227 __u16 max_vdes;
228 __u8 pad[52];
229 struct virtual_entry {
230 char guid[DDF_GUID_LEN];
231 __u16 unit;
232 __u16 pad0; /* 0xffff */
233 __u16 guid_crc;
234 __u16 type;
235 __u8 state;
236 __u8 init_state;
237 __u8 pad1[14];
238 char name[16];
239 } entries[0];
240};
241
242/* virtual_entry.type is a bitmap - bigendian */
243#define DDF_Shared 1
244#define DDF_Enforce_Groups 2
245#define DDF_Unicode 4
246#define DDF_Owner_Valid 8
247
248/* virtual_entry.state is a bigendian bitmap */
249#define DDF_state_mask 0x7
250#define DDF_state_optimal 0x0
251#define DDF_state_degraded 0x1
252#define DDF_state_deleted 0x2
253#define DDF_state_missing 0x3
254#define DDF_state_failed 0x4
7a7cc504 255#define DDF_state_part_optimal 0x5
a322f70c
DW		 256
257#define DDF_state_morphing 0x8
258#define DDF_state_inconsistent 0x10
259
260/* virtual_entry.init_state is a bigendian bitmap */
261#define DDF_initstate_mask 0x03
262#define DDF_init_not 0x00
7a7cc504
NB		 263#define DDF_init_quick 0x01 /* initialisation is progress.
264 * i.e. 'state_inconsistent' */
a322f70c
DW		 265#define DDF_init_full 0x02
266
267#define DDF_access_mask 0xc0
268#define DDF_access_rw 0x00
269#define DDF_access_ro 0x80
270#define DDF_access_blocked 0xc0
271
272/* The content of the config_section - local scope
273 * It has multiple records each config_record_len sectors
274 * They can be vd_config or spare_assign
275 */
276
277struct vd_config {
88c164f4 278 __u32 magic; /* DDF_VD_CONF_MAGIC */
a322f70c
DW		 279 __u32 crc;
280 char guid[DDF_GUID_LEN];
281 __u32 timestamp;
282 __u32 seqnum;
283 __u8 pad0[24];
284 __u16 prim_elmnt_count;
285 __u8 chunk_shift; /* 0 == 512, 1==1024 etc */
286 __u8 prl;
287 __u8 rlq;
288 __u8 sec_elmnt_count;
289 __u8 sec_elmnt_seq;
290 __u8 srl;
598f0d58
NB		 291 __u64 blocks; /* blocks per component could be different
292 * on different component devices...(only
293 * for concat I hope) */
294 __u64 array_blocks; /* blocks in array */
a322f70c
DW		 295 __u8 pad1[8];
296 __u32 spare_refs[8];
297 __u8 cache_pol[8];
298 __u8 bg_rate;
299 __u8 pad2[3];
300 __u8 pad3[52];
301 __u8 pad4[192];
302 __u8 v0[32]; /* reserved- 0xff */
303 __u8 v1[32]; /* reserved- 0xff */
304 __u8 v2[16]; /* reserved- 0xff */
305 __u8 v3[16]; /* reserved- 0xff */
306 __u8 vendor[32];
307 __u32 phys_refnum[0]; /* refnum of each disk in sequence */
308 /*__u64 lba_offset[0]; LBA offset in each phys. Note extents in a
309 bvd are always the same size */
310};
311
312/* vd_config.cache_pol[7] is a bitmap */
313#define DDF_cache_writeback 1 /* else writethrough */
314#define DDF_cache_wadaptive 2 /* only applies if writeback */
315#define DDF_cache_readahead 4
316#define DDF_cache_radaptive 8 /* only if doing read-ahead */
317#define DDF_cache_ifnobatt 16 /* even to write cache if battery is poor */
318#define DDF_cache_wallowed 32 /* enable write caching */
319#define DDF_cache_rallowed 64 /* enable read caching */
320
321struct spare_assign {
88c164f4 322 __u32 magic; /* DDF_SPARE_ASSIGN_MAGIC */
a322f70c
DW		 323 __u32 crc;
324 __u32 timestamp;
325 __u8 reserved[7];
326 __u8 type;
327 __u16 populated; /* SAEs used */
328 __u16 max; /* max SAEs */
329 __u8 pad[8];
330 struct spare_assign_entry {
331 char guid[DDF_GUID_LEN];
332 __u16 secondary_element;
333 __u8 pad[6];
334 } spare_ents[0];
335};
336/* spare_assign.type is a bitmap */
337#define DDF_spare_dedicated 0x1 /* else global */
338#define DDF_spare_revertible 0x2 /* else committable */
339#define DDF_spare_active 0x4 /* else not active */
340#define DDF_spare_affinity 0x8 /* enclosure affinity */
341
342/* The data_section contents - local scope */
343struct disk_data {
88c164f4 344 __u32 magic; /* DDF_PHYS_DATA_MAGIC */
a322f70c
DW		 345 __u32 crc;
346 char guid[DDF_GUID_LEN];
347 __u32 refnum; /* crc of some magic drive data ... */
348 __u8 forced_ref; /* set when above was not result of magic */
349 __u8 forced_guid; /* set if guid was forced rather than magic */
350 __u8 vendor[32];
351 __u8 pad[442];
352};
353
354/* bbm_section content */
355struct bad_block_log {
356 __u32 magic;
357 __u32 crc;
358 __u16 entry_count;
359 __u32 spare_count;
360 __u8 pad[10];
361 __u64 first_spare;
362 struct mapped_block {
363 __u64 defective_start;
364 __u32 replacement_start;
365 __u16 remap_count;
366 __u8 pad[2];
367 } entries[0];
368};
369
370/* Struct for internally holding ddf structures */
371/* The DDF structure stored on each device is potentially
372 * quite different, as some data is global and some is local.
373 * The global data is:
374 * - ddf header
375 * - controller_data
376 * - Physical disk records
377 * - Virtual disk records
378 * The local data is:
379 * - Configuration records
380 * - Physical Disk data section
381 * ( and Bad block and vendor which I don't care about yet).
382 *
383 * The local data is parsed into separate lists as it is read
384 * and reconstructed for writing. This means that we only need
385 * to make config changes once and they are automatically
386 * propagated to all devices.
387 * Note that the ddf_super has space of the conf and disk data
388 * for this disk and also for a list of all such data.
389 * The list is only used for the superblock that is being
390 * built in Create or Assemble to describe the whole array.
391 */
392struct ddf_super {
6416d527 393 struct ddf_header anchor, primary, secondary;
a322f70c 394 struct ddf_controller_data controller;
6416d527 395 struct ddf_header *active;
a322f70c
DW		 396 struct phys_disk *phys;
397 struct virtual_disk *virt;
398 int pdsize, vdsize;
8c3b8c2c 399 int max_part, mppe, conf_rec_len;
d2ca6449 400 int currentdev;
18a2f463 401 int updates_pending;
a322f70c 402 struct vcl {
6416d527
NB		 403 union {
404 char space[512];
405 struct {
406 struct vcl *next;
407 __u64 *lba_offset; /* location in 'conf' of
408 * the lba table */
409 int vcnum; /* index into ->virt */
410 __u64 *block_sizes; /* NULL if all the same */
411 };
412 };
a322f70c 413 struct vd_config conf;
d2ca6449 414 } *conflist, *currentconf;
a322f70c 415 struct dl {
6416d527
NB		 416 union {
417 char space[512];
418 struct {
419 struct dl *next;
420 int major, minor;
421 char *devname;
422 int fd;
423 unsigned long long size; /* sectors */
424 int pdnum; /* index in ->phys */
425 struct spare_assign *spare;
426 };
427 };
a322f70c 428 struct disk_data disk;
2cc2983d 429 void *mdupdate; /* hold metadata update */
b2280677 430 struct vcl *vlist[0]; /* max_part in size */
2cc2983d 431 } *dlist, *add_list;
a322f70c
DW		 432};
433
434#ifndef offsetof
435#define offsetof(t,f) ((size_t)&(((t*)0)->f))
436#endif
437
a322f70c
DW		 438
439static int calc_crc(void *buf, int len)
440{
441 /* crcs are always at the same place as in the ddf_header */
442 struct ddf_header *ddf = buf;
443 __u32 oldcrc = ddf->crc;
444 __u32 newcrc;
445 ddf->crc = 0xffffffff;
446
447 newcrc = crc32(0, buf, len);
448 ddf->crc = oldcrc;
4abe6b70
N		 449 /* The crc is store (like everything) bigendian, so convert
450 * here for simplicity
451 */
452 return __cpu_to_be32(newcrc);
a322f70c
DW		 453}
454
455static int load_ddf_header(int fd, unsigned long long lba,
456 unsigned long long size,
457 int type,
458 struct ddf_header *hdr, struct ddf_header *anchor)
459{
460 /* read a ddf header (primary or secondary) from fd/lba
461 * and check that it is consistent with anchor
462 * Need to check:
463 * magic, crc, guid, rev, and LBA's header_type, and
464 * everything after header_type must be the same
465 */
466 if (lba >= size-1)
467 return 0;
468
469 if (lseek64(fd, lba<<9, 0) < 0)
470 return 0;
471
472 if (read(fd, hdr, 512) != 512)
473 return 0;
474
475 if (hdr->magic != DDF_HEADER_MAGIC)
476 return 0;
477 if (calc_crc(hdr, 512) != hdr->crc)
478 return 0;
479 if (memcmp(anchor->guid, hdr->guid, DDF_GUID_LEN) != 0 ||
480 memcmp(anchor->revision, hdr->revision, 8) != 0 ||
481 anchor->primary_lba != hdr->primary_lba ||
482 anchor->secondary_lba != hdr->secondary_lba ||
483 hdr->type != type ||
484 memcmp(anchor->pad2, hdr->pad2, 512 -
485 offsetof(struct ddf_header, pad2)) != 0)
486 return 0;
487
488 /* Looks good enough to me... */
489 return 1;
490}
491
492static void *load_section(int fd, struct ddf_super *super, void *buf,
493 __u32 offset_be, __u32 len_be, int check)
494{
495 unsigned long long offset = __be32_to_cpu(offset_be);
496 unsigned long long len = __be32_to_cpu(len_be);
497 int dofree = (buf == NULL);
498
499 if (check)
500 if (len != 2 && len != 8 && len != 32
501 && len != 128 && len != 512)
502 return NULL;
503
504 if (len > 1024)
505 return NULL;
506 if (buf) {
507 /* All pre-allocated sections are a single block */
508 if (len != 1)
509 return NULL;
3d2c4fc7
DW		 510 } else if (posix_memalign(&buf, 512, len<<9) != 0)
511 buf = NULL;
6416d527 512
a322f70c
DW		 513 if (!buf)
514 return NULL;
515
516 if (super->active->type == 1)
517 offset += __be64_to_cpu(super->active->primary_lba);
518 else
519 offset += __be64_to_cpu(super->active->secondary_lba);
520
521 if (lseek64(fd, offset<<9, 0) != (offset<<9)) {
522 if (dofree)
523 free(buf);
524 return NULL;
525 }
526 if (read(fd, buf, len<<9) != (len<<9)) {
527 if (dofree)
528 free(buf);
529 return NULL;
530 }
531 return buf;
532}
533
534static int load_ddf_headers(int fd, struct ddf_super *super, char *devname)
535{
536 unsigned long long dsize;
537
538 get_dev_size(fd, NULL, &dsize);
539
540 if (lseek64(fd, dsize-512, 0) < 0) {
541 if (devname)
542 fprintf(stderr,
543 Name": Cannot seek to anchor block on %s: %s\n",
544 devname, strerror(errno));
545 return 1;
546 }
547 if (read(fd, &super->anchor, 512) != 512) {
548 if (devname)
549 fprintf(stderr,
550 Name ": Cannot read anchor block on %s: %s\n",
551 devname, strerror(errno));
552 return 1;
553 }
554 if (super->anchor.magic != DDF_HEADER_MAGIC) {
555 if (devname)
556 fprintf(stderr, Name ": no DDF anchor found on %s\n",
557 devname);
558 return 2;
559 }
560 if (calc_crc(&super->anchor, 512) != super->anchor.crc) {
561 if (devname)
562 fprintf(stderr, Name ": bad CRC on anchor on %s\n",
563 devname);
564 return 2;
565 }
59e36268
NB		 566 if (memcmp(super->anchor.revision, DDF_REVISION_0, 8) != 0 &&
567 memcmp(super->anchor.revision, DDF_REVISION_2, 8) != 0) {
a322f70c
DW		 568 if (devname)
569 fprintf(stderr, Name ": can only support super revision"
59e36268
NB		 570 " %.8s and earlier, not %.8s on %s\n",
571 DDF_REVISION_2, super->anchor.revision,devname);
a322f70c
DW		 572 return 2;
573 }
574 if (load_ddf_header(fd, __be64_to_cpu(super->anchor.primary_lba),
575 dsize >> 9, 1,
576 &super->primary, &super->anchor) == 0) {
577 if (devname)
578 fprintf(stderr,
579 Name ": Failed to load primary DDF header "
580 "on %s\n", devname);
581 return 2;
582 }
583 super->active = &super->primary;
584 if (load_ddf_header(fd, __be64_to_cpu(super->anchor.secondary_lba),
585 dsize >> 9, 2,
586 &super->secondary, &super->anchor)) {
587 if ((__be32_to_cpu(super->primary.seq)
588 < __be32_to_cpu(super->secondary.seq) &&
589 !super->secondary.openflag)
590 || (__be32_to_cpu(super->primary.seq)
591 == __be32_to_cpu(super->secondary.seq) &&
592 super->primary.openflag && !super->secondary.openflag)
593 )
594 super->active = &super->secondary;
595 }
596 return 0;
597}
598
599static int load_ddf_global(int fd, struct ddf_super *super, char *devname)
600{
601 void *ok;
602 ok = load_section(fd, super, &super->controller,
603 super->active->controller_section_offset,
604 super->active->controller_section_length,
605 0);
606 super->phys = load_section(fd, super, NULL,
607 super->active->phys_section_offset,
608 super->active->phys_section_length,
609 1);
610 super->pdsize = __be32_to_cpu(super->active->phys_section_length) * 512;
611
612 super->virt = load_section(fd, super, NULL,
613 super->active->virt_section_offset,
614 super->active->virt_section_length,
615 1);
616 super->vdsize = __be32_to_cpu(super->active->virt_section_length) * 512;
617 if (!ok ||
618 !super->phys ||
619 !super->virt) {
620 free(super->phys);
621 free(super->virt);
a2349791
NB		 622 super->phys = NULL;
623 super->virt = NULL;
a322f70c
DW		 624 return 2;
625 }
626 super->conflist = NULL;
627 super->dlist = NULL;
8c3b8c2c
NB		 628
629 super->max_part = __be16_to_cpu(super->active->max_partitions);
630 super->mppe = __be16_to_cpu(super->active->max_primary_element_entries);
631 super->conf_rec_len = __be16_to_cpu(super->active->config_record_len);
a322f70c
DW		 632 return 0;
633}
634
635static int load_ddf_local(int fd, struct ddf_super *super,
636 char *devname, int keep)
637{
638 struct dl *dl;
639 struct stat stb;
640 char *conf;
641 int i;
b2280677 642 int vnum;
59e36268 643 int max_virt_disks = __be16_to_cpu(super->active->max_vd_entries);
d2ca6449 644 unsigned long long dsize;
a322f70c
DW		 645
646 /* First the local disk info */
3d2c4fc7 647 if (posix_memalign((void**)&dl, 512,
6416d527 648 sizeof(*dl) +
3d2c4fc7
DW		 649 (super->max_part) * sizeof(dl->vlist[0])) != 0) {
650 fprintf(stderr, Name ": %s could not allocate disk info buffer\n",
651 __func__);
652 return 1;
653 }
a322f70c
DW		 654
655 load_section(fd, super, &dl->disk,
656 super->active->data_section_offset,
657 super->active->data_section_length,
658 0);
659 dl->devname = devname ? strdup(devname) : NULL;
598f0d58 660
a322f70c
DW		 661 fstat(fd, &stb);
662 dl->major = major(stb.st_rdev);
663 dl->minor = minor(stb.st_rdev);
664 dl->next = super->dlist;
665 dl->fd = keep ? fd : -1;
d2ca6449
NB		 666
667 dl->size = 0;
668 if (get_dev_size(fd, devname, &dsize))
669 dl->size = dsize >> 9;
b2280677
NB		 670 dl->spare = NULL;
671 for (i=0 ; i < super->max_part ; i++)
a322f70c
DW		 672 dl->vlist[i] = NULL;
673 super->dlist = dl;
59e36268 674 dl->pdnum = -1;
5575e7d9
NB		 675 for (i=0; i < __be16_to_cpu(super->active->max_pd_entries); i++)
676 if (memcmp(super->phys->entries[i].guid,
677 dl->disk.guid, DDF_GUID_LEN) == 0)
678 dl->pdnum = i;
679
a322f70c
DW		 680 /* Now the config list. */
681 /* 'conf' is an array of config entries, some of which are
682 * probably invalid. Those which are good need to be copied into
683 * the conflist
684 */
a322f70c
DW		 685
686 conf = load_section(fd, super, NULL,
687 super->active->config_section_offset,
688 super->active->config_section_length,
689 0);
690
b2280677 691 vnum = 0;
a322f70c
DW		 692 for (i = 0;
693 i < __be32_to_cpu(super->active->config_section_length);
8c3b8c2c 694 i += super->conf_rec_len) {
a322f70c
DW		 695 struct vd_config *vd =
696 (struct vd_config *)((char*)conf + i*512);
697 struct vcl *vcl;
698
b2280677
NB		 699 if (vd->magic == DDF_SPARE_ASSIGN_MAGIC) {
700 if (dl->spare)
701 continue;
3d2c4fc7
DW		 702 if (posix_memalign((void**)&dl->spare, 512,
703 super->conf_rec_len*512) != 0) {
704 fprintf(stderr, Name
705 ": %s could not allocate spare info buf\n",
706 __func__);
707 return 1;
708 }
709
b2280677
NB		 710 memcpy(dl->spare, vd, super->conf_rec_len*512);
711 continue;
712 }
a322f70c
DW		 713 if (vd->magic != DDF_VD_CONF_MAGIC)
714 continue;
715 for (vcl = super->conflist; vcl; vcl = vcl->next) {
716 if (memcmp(vcl->conf.guid,
717 vd->guid, DDF_GUID_LEN) == 0)
718 break;
719 }
720
721 if (vcl) {
b2280677 722 dl->vlist[vnum++] = vcl;
a322f70c
DW		 723 if (__be32_to_cpu(vd->seqnum) <=
724 __be32_to_cpu(vcl->conf.seqnum))
725 continue;
59e36268 726 } else {
3d2c4fc7 727 if (posix_memalign((void**)&vcl, 512,
6416d527 728 (super->conf_rec_len*512 +
3d2c4fc7
DW		 729 offsetof(struct vcl, conf))) != 0) {
730 fprintf(stderr, Name
731 ": %s could not allocate vcl buf\n",
732 __func__);
733 return 1;
734 }
a322f70c 735 vcl->next = super->conflist;
59e36268 736 vcl->block_sizes = NULL; /* FIXME not for CONCAT */
a322f70c 737 super->conflist = vcl;
b2280677 738 dl->vlist[vnum++] = vcl;
a322f70c 739 }
8c3b8c2c 740 memcpy(&vcl->conf, vd, super->conf_rec_len*512);
a322f70c 741 vcl->lba_offset = (__u64*)
8c3b8c2c 742 &vcl->conf.phys_refnum[super->mppe];
59e36268
NB		 743
744 for (i=0; i < max_virt_disks ; i++)
745 if (memcmp(super->virt->entries[i].guid,
746 vcl->conf.guid, DDF_GUID_LEN)==0)
747 break;
748 if (i < max_virt_disks)
749 vcl->vcnum = i;
a322f70c
DW		 750 }
751 free(conf);
752
753 return 0;
754}
755
756#ifndef MDASSEMBLE
757static int load_super_ddf_all(struct supertype *st, int fd,
758 void **sbp, char *devname, int keep_fd);
759#endif
760static int load_super_ddf(struct supertype *st, int fd,
761 char *devname)
762{
763 unsigned long long dsize;
764 struct ddf_super *super;
765 int rv;
766
767#ifndef MDASSEMBLE
59e36268 768 /* if 'fd' is a container, load metadata from all the devices */
3dbccbcf 769 if (load_super_ddf_all(st, fd, &st->sb, devname, 1) == 0)
a322f70c
DW		 770 return 0;
771#endif
f7e7067b
NB		 772 if (st->subarray[0])
773 return 1; /* FIXME Is this correct */
a322f70c
DW		 774
775 if (get_dev_size(fd, devname, &dsize) == 0)
776 return 1;
777
778 /* 32M is a lower bound */
779 if (dsize <= 32*1024*1024) {
780 if (devname) {
781 fprintf(stderr,
782 Name ": %s is too small for ddf: "
783 "size is %llu sectors.\n",
784 devname, dsize>>9);
785 return 1;
786 }
787 }
788 if (dsize & 511) {
789 if (devname) {
790 fprintf(stderr,
791 Name ": %s is an odd size for ddf: "
792 "size is %llu bytes.\n",
793 devname, dsize);
794 return 1;
795 }
796 }
797
6416d527 798 if (posix_memalign((void**)&super, 512, sizeof(*super))!= 0) {
a322f70c
DW		 799 fprintf(stderr, Name ": malloc of %zu failed.\n",
800 sizeof(*super));
801 return 1;
802 }
a2349791 803 memset(super, 0, sizeof(*super));
a322f70c
DW		 804
805 rv = load_ddf_headers(fd, super, devname);
806 if (rv) {
807 free(super);
808 return rv;
809 }
810
811 /* Have valid headers and have chosen the best. Let's read in the rest*/
812
813 rv = load_ddf_global(fd, super, devname);
814
815 if (rv) {
816 if (devname)
817 fprintf(stderr,
818 Name ": Failed to load all information "
819 "sections on %s\n", devname);
820 free(super);
821 return rv;
822 }
823
3d2c4fc7
DW		 824 rv = load_ddf_local(fd, super, devname, 0);
825
826 if (rv) {
827 if (devname)
828 fprintf(stderr,
829 Name ": Failed to load all information "
830 "sections on %s\n", devname);
831 free(super);
832 return rv;
833 }
a322f70c
DW		 834
835 /* Should possibly check the sections .... */
836
837 st->sb = super;
838 if (st->ss == NULL) {
839 st->ss = &super_ddf;
840 st->minor_version = 0;
841 st->max_devs = 512;
842 }
352452c3 843 st->loaded_container = 0;
a322f70c
DW		 844 return 0;
845
846}
847
848static void free_super_ddf(struct supertype *st)
849{
850 struct ddf_super *ddf = st->sb;
851 if (ddf == NULL)
852 return;
853 free(ddf->phys);
854 free(ddf->virt);
855 while (ddf->conflist) {
856 struct vcl *v = ddf->conflist;
857 ddf->conflist = v->next;
59e36268
NB		 858 if (v->block_sizes)
859 free(v->block_sizes);
a322f70c
DW		 860 free(v);
861 }
862 while (ddf->dlist) {
863 struct dl *d = ddf->dlist;
864 ddf->dlist = d->next;
865 if (d->fd >= 0)
866 close(d->fd);
b2280677
NB		 867 if (d->spare)
868 free(d->spare);
a322f70c
DW		 869 free(d);
870 }
871 free(ddf);
872 st->sb = NULL;
873}
874
875static struct supertype *match_metadata_desc_ddf(char *arg)
876{
877 /* 'ddf' only support containers */
878 struct supertype *st;
879 if (strcmp(arg, "ddf") != 0 &&
880 strcmp(arg, "default") != 0
881 )
882 return NULL;
883
884 st = malloc(sizeof(*st));
ef609477 885 memset(st, 0, sizeof(*st));
a322f70c
DW		 886 st->ss = &super_ddf;
887 st->max_devs = 512;
888 st->minor_version = 0;
889 st->sb = NULL;
890 return st;
891}
892
a322f70c
DW		 893
894#ifndef MDASSEMBLE
895
896static mapping_t ddf_state[] = {
897 { "Optimal", 0},
898 { "Degraded", 1},
899 { "Deleted", 2},
900 { "Missing", 3},
901 { "Failed", 4},
902 { "Partially Optimal", 5},
903 { "-reserved-", 6},
904 { "-reserved-", 7},
905 { NULL, 0}
906};
907
908static mapping_t ddf_init_state[] = {
909 { "Not Initialised", 0},
910 { "QuickInit in Progress", 1},
911 { "Fully Initialised", 2},
912 { "*UNKNOWN*", 3},
913 { NULL, 0}
914};
915static mapping_t ddf_access[] = {
916 { "Read/Write", 0},
917 { "Reserved", 1},
918 { "Read Only", 2},
919 { "Blocked (no access)", 3},
920 { NULL ,0}
921};
922
923static mapping_t ddf_level[] = {
924 { "RAID0", DDF_RAID0},
925 { "RAID1", DDF_RAID1},
926 { "RAID3", DDF_RAID3},
927 { "RAID4", DDF_RAID4},
928 { "RAID5", DDF_RAID5},
929 { "RAID1E",DDF_RAID1E},
930 { "JBOD", DDF_JBOD},
931 { "CONCAT",DDF_CONCAT},
932 { "RAID5E",DDF_RAID5E},
933 { "RAID5EE",DDF_RAID5EE},
934 { "RAID6", DDF_RAID6},
935 { NULL, 0}
936};
937static mapping_t ddf_sec_level[] = {
938 { "Striped", DDF_2STRIPED},
939 { "Mirrored", DDF_2MIRRORED},
940 { "Concat", DDF_2CONCAT},
941 { "Spanned", DDF_2SPANNED},
942 { NULL, 0}
943};
944#endif
945
946struct num_mapping {
947 int num1, num2;
948};
949static struct num_mapping ddf_level_num[] = {
950 { DDF_RAID0, 0 },
951 { DDF_RAID1, 1 },
952 { DDF_RAID3, LEVEL_UNSUPPORTED },
60f18132
NB		 953 { DDF_RAID4, 4 },
954 { DDF_RAID5, 5 },
a322f70c
DW		 955 { DDF_RAID1E, LEVEL_UNSUPPORTED },
956 { DDF_JBOD, LEVEL_UNSUPPORTED },
957 { DDF_CONCAT, LEVEL_LINEAR },
958 { DDF_RAID5E, LEVEL_UNSUPPORTED },
959 { DDF_RAID5EE, LEVEL_UNSUPPORTED },
960 { DDF_RAID6, 6},
961 { MAXINT, MAXINT }
962};
963
964static int map_num1(struct num_mapping *map, int num)
965{
966 int i;
967 for (i=0 ; map[i].num1 != MAXINT; i++)
968 if (map[i].num1 == num)
969 break;
970 return map[i].num2;
971}
972
42dc2744
N		 973static int all_ff(char *guid)
974{
975 int i;
976 for (i = 0; i < DDF_GUID_LEN; i++)
977 if (guid[i] != (char)0xff)
978 return 0;
979 return 1;
980}
981
a322f70c
DW		 982#ifndef MDASSEMBLE
983static void print_guid(char *guid, int tstamp)
984{
985 /* A GUIDs are part (or all) ASCII and part binary.
986 * They tend to be space padded.
59e36268
NB		 987 * We print the GUID in HEX, then in parentheses add
988 * any initial ASCII sequence, and a possible
989 * time stamp from bytes 16-19
a322f70c
DW		 990 */
991 int l = DDF_GUID_LEN;
992 int i;
59e36268
NB		 993
994 for (i=0 ; i<DDF_GUID_LEN ; i++) {
995 if ((i&3)==0 && i != 0) printf(":");
996 printf("%02X", guid[i]&255);
997 }
998
999 printf(" (");
a322f70c
DW		 1000 while (l && guid[l-1] == ' ')
1001 l--;
1002 for (i=0 ; i<l ; i++) {
1003 if (guid[i] >= 0x20 && guid[i] < 0x7f)
1004 fputc(guid[i], stdout);
1005 else
59e36268 1006 break;
a322f70c
DW		 1007 }
1008 if (tstamp) {
1009 time_t then = __be32_to_cpu(*(__u32*)(guid+16)) + DECADE;
1010 char tbuf[100];
1011 struct tm *tm;
1012 tm = localtime(&then);
59e36268 1013 strftime(tbuf, 100, " %D %T",tm);
a322f70c
DW		 1014 fputs(tbuf, stdout);
1015 }
59e36268 1016 printf(")");
a322f70c
DW		 1017}
1018
1019static void examine_vd(int n, struct ddf_super *sb, char *guid)
1020{
8c3b8c2c 1021 int crl = sb->conf_rec_len;
a322f70c
DW		 1022 struct vcl *vcl;
1023
1024 for (vcl = sb->conflist ; vcl ; vcl = vcl->next) {
1025 struct vd_config *vc = &vcl->conf;
1026
1027 if (calc_crc(vc, crl*512) != vc->crc)
1028 continue;
1029 if (memcmp(vc->guid, guid, DDF_GUID_LEN) != 0)
1030 continue;
1031
1032 /* Ok, we know about this VD, let's give more details */
1033 printf(" Raid Devices[%d] : %d\n", n,
1034 __be16_to_cpu(vc->prim_elmnt_count));
1035 printf(" Chunk Size[%d] : %d sectors\n", n,
1036 1 << vc->chunk_shift);
1037 printf(" Raid Level[%d] : %s\n", n,
1038 map_num(ddf_level, vc->prl)?:"-unknown-");
1039 if (vc->sec_elmnt_count != 1) {
1040 printf(" Secondary Position[%d] : %d of %d\n", n,
1041 vc->sec_elmnt_seq, vc->sec_elmnt_count);
1042 printf(" Secondary Level[%d] : %s\n", n,
1043 map_num(ddf_sec_level, vc->srl) ?: "-unknown-");
1044 }
1045 printf(" Device Size[%d] : %llu\n", n,
1046 __be64_to_cpu(vc->blocks)/2);
1047 printf(" Array Size[%d] : %llu\n", n,
1048 __be64_to_cpu(vc->array_blocks)/2);
1049 }
1050}
1051
1052static void examine_vds(struct ddf_super *sb)
1053{
1054 int cnt = __be16_to_cpu(sb->virt->populated_vdes);
1055 int i;
1056 printf(" Virtual Disks : %d\n", cnt);
1057
1058 for (i=0; i<cnt; i++) {
1059 struct virtual_entry *ve = &sb->virt->entries[i];
1060 printf(" VD GUID[%d] : ", i); print_guid(ve->guid, 1);
1061 printf("\n");
1062 printf(" unit[%d] : %d\n", i, __be16_to_cpu(ve->unit));
1063 printf(" state[%d] : %s, %s%s\n", i,
1064 map_num(ddf_state, ve->state & 7),
1065 (ve->state & 8) ? "Morphing, ": "",
1066 (ve->state & 16)? "Not Consistent" : "Consistent");
1067 printf(" init state[%d] : %s\n", i,
1068 map_num(ddf_init_state, ve->init_state&3));
1069 printf(" access[%d] : %s\n", i,
1070 map_num(ddf_access, (ve->init_state>>6) & 3));
1071 printf(" Name[%d] : %.16s\n", i, ve->name);
1072 examine_vd(i, sb, ve->guid);
1073 }
1074 if (cnt) printf("\n");
1075}
1076
1077static void examine_pds(struct ddf_super *sb)
1078{
1079 int cnt = __be16_to_cpu(sb->phys->used_pdes);
1080 int i;
1081 struct dl *dl;
1082 printf(" Physical Disks : %d\n", cnt);
1083
1084 for (i=0 ; i<cnt ; i++) {
1085 struct phys_disk_entry *pd = &sb->phys->entries[i];
1086 int type = __be16_to_cpu(pd->type);
1087 int state = __be16_to_cpu(pd->state);
1088
1089 printf(" PD GUID[%d] : ", i); print_guid(pd->guid, 0);
1090 printf("\n");
1091 printf(" ref[%d] : %08x\n", i,
1092 __be32_to_cpu(pd->refnum));
1093 printf(" mode[%d] : %s%s%s%s%s\n", i,
1094 (type&2) ? "active":"",
1095 (type&4) ? "Global Spare":"",
1096 (type&8) ? "spare" : "",
1097 (type&16)? ", foreign" : "",
1098 (type&32)? "pass-through" : "");
1099 printf(" state[%d] : %s%s%s%s%s%s%s\n", i,
1100 (state&1)? "Online": "Offline",
1101 (state&2)? ", Failed": "",
1102 (state&4)? ", Rebuilding": "",
1103 (state&8)? ", in-transition": "",
1104 (state&16)? ", SMART errors": "",
1105 (state&32)? ", Unrecovered Read Errors": "",
1106 (state&64)? ", Missing" : "");
1107 printf(" Avail Size[%d] : %llu K\n", i,
1108 __be64_to_cpu(pd->config_size)>>1);
1109 for (dl = sb->dlist; dl ; dl = dl->next) {
1110 if (dl->disk.refnum == pd->refnum) {
1111 char *dv = map_dev(dl->major, dl->minor, 0);
1112 if (dv)
1113 printf(" Device[%d] : %s\n",
1114 i, dv);
1115 }
1116 }
1117 printf("\n");
1118 }
1119}
1120
1121static void examine_super_ddf(struct supertype *st, char *homehost)
1122{
1123 struct ddf_super *sb = st->sb;
1124
1125 printf(" Magic : %08x\n", __be32_to_cpu(sb->anchor.magic));
1126 printf(" Version : %.8s\n", sb->anchor.revision);
598f0d58
NB		 1127 printf("Controller GUID : "); print_guid(sb->controller.guid, 0);
1128 printf("\n");
1129 printf(" Container GUID : "); print_guid(sb->anchor.guid, 1);
a322f70c
DW		 1130 printf("\n");
1131 printf(" Seq : %08x\n", __be32_to_cpu(sb->active->seq));
1132 printf(" Redundant hdr : %s\n", sb->secondary.magic == DDF_HEADER_MAGIC
1133 ?"yes" : "no");
1134 examine_vds(sb);
1135 examine_pds(sb);
1136}
1137
ff54de6e
N		 1138static void getinfo_super_ddf(struct supertype *st, struct mdinfo *info);
1139
42dc2744 1140static void uuid_from_super_ddf(struct supertype *st, int uuid[4]);
ff54de6e 1141
a322f70c
DW		 1142static void brief_examine_super_ddf(struct supertype *st)
1143{
1144 /* We just write a generic DDF ARRAY entry
a322f70c 1145 */
42dc2744 1146 struct ddf_super *ddf = st->sb;
ff54de6e 1147 struct mdinfo info;
42dc2744 1148 int i;
ff54de6e
N		 1149 char nbuf[64];
1150 getinfo_super_ddf(st, &info);
1151 fname_from_uuid(st, &info, nbuf, ':');
cf8de691 1152 printf("ARRAY metadata=ddf UUID=%s\n", nbuf + 5);
42dc2744
N		 1153
1154 for (i=0; i<__be16_to_cpu(ddf->virt->max_vdes); i++) {
1155 struct virtual_entry *ve = &ddf->virt->entries[i];
1156 struct vcl vcl;
1157 char nbuf1[64];
1158 if (all_ff(ve->guid))
1159 continue;
1160 memcpy(vcl.conf.guid, ve->guid, DDF_GUID_LEN);
1161 ddf->currentconf =&vcl;
1162 uuid_from_super_ddf(st, info.uuid);
1163 fname_from_uuid(st, &info, nbuf1, ':');
1164 printf("ARRAY container=%s member=%d UUID=%s\n",
1165 nbuf+5, i, nbuf1+5);
1166 }
a322f70c
DW		 1167}
1168
1169static void detail_super_ddf(struct supertype *st, char *homehost)
1170{
1171 /* FIXME later
1172 * Could print DDF GUID
1173 * Need to find which array
1174 * If whole, briefly list all arrays
1175 * If one, give name
1176 */
1177}
1178
1179static void brief_detail_super_ddf(struct supertype *st)
1180{
1181 /* FIXME I really need to know which array we are detailing.
1182 * Can that be stored in ddf_super??
1183 */
1184// struct ddf_super *ddf = st->sb;
ff54de6e
N		 1185 struct mdinfo info;
1186 char nbuf[64];
1187 getinfo_super_ddf(st, &info);
1188 fname_from_uuid(st, &info, nbuf,':');
1189 printf(" UUID=%s", nbuf + 5);
a322f70c 1190}
a322f70c
DW		 1191#endif
1192
1193static int match_home_ddf(struct supertype *st, char *homehost)
1194{
1195 /* It matches 'this' host if the controller is a
1196 * Linux-MD controller with vendor_data matching
1197 * the hostname
1198 */
1199 struct ddf_super *ddf = st->sb;
1200 int len = strlen(homehost);
1201
1202 return (memcmp(ddf->controller.guid, T10, 8) == 0 &&
1203 len < sizeof(ddf->controller.vendor_data) &&
1204 memcmp(ddf->controller.vendor_data, homehost,len) == 0 &&
1205 ddf->controller.vendor_data[len] == 0);
1206}
1207
0e600426 1208#ifndef MDASSEMBLE
7a7cc504 1209static struct vd_config *find_vdcr(struct ddf_super *ddf, int inst)
a322f70c 1210{
7a7cc504 1211 struct vcl *v;
59e36268 1212
7a7cc504 1213 for (v = ddf->conflist; v; v = v->next)
59e36268 1214 if (inst == v->vcnum)
7a7cc504
NB		 1215 return &v->conf;
1216 return NULL;
1217}
0e600426 1218#endif
7a7cc504
NB		 1219
1220static int find_phys(struct ddf_super *ddf, __u32 phys_refnum)
1221{
1222 /* Find the entry in phys_disk which has the given refnum
1223 * and return it's index
1224 */
1225 int i;
1226 for (i=0; i < __be16_to_cpu(ddf->phys->max_pdes); i++)
1227 if (ddf->phys->entries[i].refnum == phys_refnum)
1228 return i;
1229 return -1;
a322f70c
DW		 1230}
1231
1232static void uuid_from_super_ddf(struct supertype *st, int uuid[4])
1233{
1234 /* The uuid returned here is used for:
1235 * uuid to put into bitmap file (Create, Grow)
1236 * uuid for backup header when saving critical section (Grow)
1237 * comparing uuids when re-adding a device into an array
51006d85
N		 1238 * In these cases the uuid required is that of the data-array,
1239 * not the device-set.
1240 * uuid to recognise same set when adding a missing device back
1241 * to an array. This is a uuid for the device-set.
1242 *
a322f70c
DW		 1243 * For each of these we can make do with a truncated
1244 * or hashed uuid rather than the original, as long as
1245 * everyone agrees.
a322f70c
DW		 1246 * In the case of SVD we assume the BVD is of interest,
1247 * though that might be the case if a bitmap were made for
1248 * a mirrored SVD - worry about that later.
1249 * So we need to find the VD configuration record for the
1250 * relevant BVD and extract the GUID and Secondary_Element_Seq.
1251 * The first 16 bytes of the sha1 of these is used.
1252 */
1253 struct ddf_super *ddf = st->sb;
d2ca6449 1254 struct vcl *vcl = ddf->currentconf;
c5afc314
N		 1255 char *guid;
1256 char buf[20];
1257 struct sha1_ctx ctx;
a322f70c 1258
c5afc314
N		 1259 if (vcl)
1260 guid = vcl->conf.guid;
1261 else
1262 guid = ddf->anchor.guid;
1263
1264 sha1_init_ctx(&ctx);
1265 sha1_process_bytes(guid, DDF_GUID_LEN, &ctx);
c5afc314
N		 1266 sha1_finish_ctx(&ctx, buf);
1267 memcpy(uuid, buf, 4*4);
a322f70c
DW		 1268}
1269
78e44928
NB		 1270static void getinfo_super_ddf_bvd(struct supertype *st, struct mdinfo *info);
1271
a322f70c
DW		 1272static void getinfo_super_ddf(struct supertype *st, struct mdinfo *info)
1273{
1274 struct ddf_super *ddf = st->sb;
1275
78e44928
NB		 1276 if (ddf->currentconf) {
1277 getinfo_super_ddf_bvd(st, info);
1278 return;
1279 }
1280
a322f70c
DW		 1281 info->array.raid_disks = __be16_to_cpu(ddf->phys->used_pdes);
1282 info->array.level = LEVEL_CONTAINER;
1283 info->array.layout = 0;
1284 info->array.md_minor = -1;
1285 info->array.ctime = DECADE + __be32_to_cpu(*(__u32*)
1286 (ddf->anchor.guid+16));
1287 info->array.utime = 0;
1288 info->array.chunk_size = 0;
1289
a322f70c
DW		 1290
1291 info->disk.major = 0;
1292 info->disk.minor = 0;
cba0191b
NB		 1293 if (ddf->dlist) {
1294 info->disk.number = __be32_to_cpu(ddf->dlist->disk.refnum);
59e36268 1295 info->disk.raid_disk = find_phys(ddf, ddf->dlist->disk.refnum);
d2ca6449
NB		 1296
1297 info->data_offset = __be64_to_cpu(ddf->phys->
1298 entries[info->disk.raid_disk].
1299 config_size);
1300 info->component_size = ddf->dlist->size - info->data_offset;
cba0191b
NB		 1301 } else {
1302 info->disk.number = -1;
1303// info->disk.raid_disk = find refnum in the table and use index;
1304 }
a19c88b8
NB		 1305 info->disk.state = (1 << MD_DISK_SYNC);
1306
d2ca6449 1307
a19c88b8 1308 info->reshape_active = 0;
c5afc314 1309 info->name[0] = 0;
a322f70c 1310
f35f2525
N		 1311 info->array.major_version = -1;
1312 info->array.minor_version = -2;
159c3a1a 1313 strcpy(info->text_version, "ddf");
a67dd8cc 1314 info->safe_mode_delay = 0;
159c3a1a 1315
c5afc314 1316 uuid_from_super_ddf(st, info->uuid);
a322f70c 1317
a322f70c
DW		 1318}
1319
598f0d58
NB		 1320static int rlq_to_layout(int rlq, int prl, int raiddisks);
1321
a322f70c
DW		 1322static void getinfo_super_ddf_bvd(struct supertype *st, struct mdinfo *info)
1323{
1324 struct ddf_super *ddf = st->sb;
d2ca6449
NB		 1325 struct vcl *vc = ddf->currentconf;
1326 int cd = ddf->currentdev;
db42fa9b 1327 int j;
a322f70c
DW		 1328
1329 /* FIXME this returns BVD info - what if we want SVD ?? */
1330
d2ca6449
NB		 1331 info->array.raid_disks = __be16_to_cpu(vc->conf.prim_elmnt_count);
1332 info->array.level = map_num1(ddf_level_num, vc->conf.prl);
1333 info->array.layout = rlq_to_layout(vc->conf.rlq, vc->conf.prl,
598f0d58 1334 info->array.raid_disks);
a322f70c 1335 info->array.md_minor = -1;
d2ca6449
NB		 1336 info->array.ctime = DECADE +
1337 __be32_to_cpu(*(__u32*)(vc->conf.guid+16));
1338 info->array.utime = DECADE + __be32_to_cpu(vc->conf.timestamp);
1339 info->array.chunk_size = 512 << vc->conf.chunk_shift;
1340
1341 if (cd >= 0 && cd < ddf->mppe) {
1342 info->data_offset = __be64_to_cpu(vc->lba_offset[cd]);
1343 if (vc->block_sizes)
1344 info->component_size = vc->block_sizes[cd];
1345 else
1346 info->component_size = __be64_to_cpu(vc->conf.blocks);
1347 }
a322f70c
DW		 1348
1349 info->disk.major = 0;
1350 info->disk.minor = 0;
1351// info->disk.number = __be32_to_cpu(ddf->disk.refnum);
1352// info->disk.raid_disk = find refnum in the table and use index;
1353// info->disk.state = ???;
1354
103f2410
NB		 1355 info->container_member = ddf->currentconf->vcnum;
1356
80d26cb2
NB		 1357 info->resync_start = 0;
1358 if (!(ddf->virt->entries[info->container_member].state
1359 & DDF_state_inconsistent) &&
1360 (ddf->virt->entries[info->container_member].init_state
1361 & DDF_initstate_mask)
1362 == DDF_init_full)
1363 info->resync_start = ~0ULL;
1364
a322f70c
DW		 1365 uuid_from_super_ddf(st, info->uuid);
1366
3576e302 1367 info->container_member = atoi(st->subarray);
f35f2525
N		 1368 info->array.major_version = -1;
1369 info->array.minor_version = -2;
3576e302 1370 sprintf(info->text_version, "/%s/%s",
159c3a1a 1371 devnum2devname(st->container_dev),
3576e302 1372 st->subarray);
a67dd8cc 1373 info->safe_mode_delay = 200;
159c3a1a 1374
db42fa9b
N		 1375 memcpy(info->name, ddf->virt->entries[info->container_member].name, 16);
1376 info->name[16]=0;
1377 for(j=0; j<16; j++)
1378 if (info->name[j] == ' ')
1379 info->name[j] = 0;
a322f70c
DW		 1380}
1381
598f0d58 1382
a322f70c
DW		 1383static int update_super_ddf(struct supertype *st, struct mdinfo *info,
1384 char *update,
1385 char *devname, int verbose,
1386 int uuid_set, char *homehost)
1387{
1388 /* For 'assemble' and 'force' we need to return non-zero if any
1389 * change was made. For others, the return value is ignored.
1390 * Update options are:
1391 * force-one : This device looks a bit old but needs to be included,
1392 * update age info appropriately.
1393 * assemble: clear any 'faulty' flag to allow this device to
1394 * be assembled.
1395 * force-array: Array is degraded but being forced, mark it clean
1396 * if that will be needed to assemble it.
1397 *
1398 * newdev: not used ????
1399 * grow: Array has gained a new device - this is currently for
1400 * linear only
1401 * resync: mark as dirty so a resync will happen.
59e36268 1402 * uuid: Change the uuid of the array to match what is given
a322f70c
DW		 1403 * homehost: update the recorded homehost
1404 * name: update the name - preserving the homehost
1405 * _reshape_progress: record new reshape_progress position.
1406 *
1407 * Following are not relevant for this version:
1408 * sparc2.2 : update from old dodgey metadata
1409 * super-minor: change the preferred_minor number
1410 * summaries: update redundant counters.
1411 */
1412 int rv = 0;
1413// struct ddf_super *ddf = st->sb;
7a7cc504 1414// struct vd_config *vd = find_vdcr(ddf, info->container_member);
a322f70c
DW		 1415// struct virtual_entry *ve = find_ve(ddf);
1416
a322f70c
DW		 1417 /* we don't need to handle "force-*" or "assemble" as
1418 * there is no need to 'trick' the kernel. We the metadata is
1419 * first updated to activate the array, all the implied modifications
1420 * will just happen.
1421 */
1422
1423 if (strcmp(update, "grow") == 0) {
1424 /* FIXME */
1425 }
1426 if (strcmp(update, "resync") == 0) {
1427// info->resync_checkpoint = 0;
1428 }
1429 /* We ignore UUID updates as they make even less sense
1430 * with DDF
1431 */
1432 if (strcmp(update, "homehost") == 0) {
1433 /* homehost is stored in controller->vendor_data,
1434 * or it is when we are the vendor
1435 */
1436// if (info->vendor_is_local)
1437// strcpy(ddf->controller.vendor_data, homehost);
1438 }
1439 if (strcmp(update, "name") == 0) {
1440 /* name is stored in virtual_entry->name */
1441// memset(ve->name, ' ', 16);
1442// strncpy(ve->name, info->name, 16);
1443 }
1444 if (strcmp(update, "_reshape_progress") == 0) {
1445 /* We don't support reshape yet */
1446 }
1447
1448// update_all_csum(ddf);
1449
1450 return rv;
1451}
1452
5f8097be
NB		 1453static void make_header_guid(char *guid)
1454{
1455 __u32 stamp;
1456 int rfd;
1457 /* Create a DDF Header of Virtual Disk GUID */
1458
1459 /* 24 bytes of fiction required.
1460 * first 8 are a 'vendor-id' - "Linux-MD"
1461 * next 8 are controller type.. how about 0X DEAD BEEF 0000 0000
1462 * Remaining 8 random number plus timestamp
1463 */
1464 memcpy(guid, T10, sizeof(T10));
1465 stamp = __cpu_to_be32(0xdeadbeef);
1466 memcpy(guid+8, &stamp, 4);
1467 stamp = __cpu_to_be32(0);
1468 memcpy(guid+12, &stamp, 4);
1469 stamp = __cpu_to_be32(time(0) - DECADE);
1470 memcpy(guid+16, &stamp, 4);
1471 rfd = open("/dev/urandom", O_RDONLY);
1472 if (rfd < 0 || read(rfd, &stamp, 4) != 4)
1473 stamp = random();
1474 memcpy(guid+20, &stamp, 4);
1475 if (rfd >= 0) close(rfd);
1476}
59e36268 1477
78e44928
NB		 1478static int init_super_ddf_bvd(struct supertype *st,
1479 mdu_array_info_t *info,
1480 unsigned long long size,
1481 char *name, char *homehost,
1482 int *uuid);
1483
a322f70c
DW		 1484static int init_super_ddf(struct supertype *st,
1485 mdu_array_info_t *info,
1486 unsigned long long size, char *name, char *homehost,
1487 int *uuid)
1488{
1489 /* This is primarily called by Create when creating a new array.
1490 * We will then get add_to_super called for each component, and then
1491 * write_init_super called to write it out to each device.
1492 * For DDF, Create can create on fresh devices or on a pre-existing
1493 * array.
1494 * To create on a pre-existing array a different method will be called.
1495 * This one is just for fresh drives.
1496 *
1497 * We need to create the entire 'ddf' structure which includes:
1498 * DDF headers - these are easy.
1499 * Controller data - a Sector describing this controller .. not that
1500 * this is a controller exactly.
1501 * Physical Disk Record - one entry per device, so
1502 * leave plenty of space.
1503 * Virtual Disk Records - again, just leave plenty of space.
1504 * This just lists VDs, doesn't give details
1505 * Config records - describes the VDs that use this disk
1506 * DiskData - describes 'this' device.
1507 * BadBlockManagement - empty
1508 * Diag Space - empty
1509 * Vendor Logs - Could we put bitmaps here?
1510 *
1511 */
1512 struct ddf_super *ddf;
1513 char hostname[17];
1514 int hostlen;
a322f70c
DW		 1515 int max_phys_disks, max_virt_disks;
1516 unsigned long long sector;
1517 int clen;
1518 int i;
1519 int pdsize, vdsize;
1520 struct phys_disk *pd;
1521 struct virtual_disk *vd;
1522
ba7eb04f
NB		 1523 if (!info) {
1524 st->sb = NULL;
1525 return 0;
1526 }
78e44928
NB		 1527 if (st->sb)
1528 return init_super_ddf_bvd(st, info, size, name, homehost,
1529 uuid);
ba7eb04f 1530
3d2c4fc7
DW		 1531 if (posix_memalign((void**)&ddf, 512, sizeof(*ddf)) != 0) {
1532 fprintf(stderr, Name ": %s could not allocate superblock\n", __func__);
1533 return 0;
1534 }
6264b437 1535 memset(ddf, 0, sizeof(*ddf));
a322f70c
DW		 1536 ddf->dlist = NULL; /* no physical disks yet */
1537 ddf->conflist = NULL; /* No virtual disks yet */
1538
1539 /* At least 32MB *must* be reserved for the ddf. So let's just
1540 * start 32MB from the end, and put the primary header there.
1541 * Don't do secondary for now.
1542 * We don't know exactly where that will be yet as it could be
1543 * different on each device. To just set up the lengths.
1544 *
1545 */
1546
1547 ddf->anchor.magic = DDF_HEADER_MAGIC;
5f8097be 1548 make_header_guid(ddf->anchor.guid);
a322f70c 1549
59e36268 1550 memcpy(ddf->anchor.revision, DDF_REVISION_2, 8);
a322f70c
DW		 1551 ddf->anchor.seq = __cpu_to_be32(1);
1552 ddf->anchor.timestamp = __cpu_to_be32(time(0) - DECADE);
1553 ddf->anchor.openflag = 0xFF;
1554 ddf->anchor.foreignflag = 0;
1555 ddf->anchor.enforcegroups = 0; /* Is this best?? */
1556 ddf->anchor.pad0 = 0xff;
1557 memset(ddf->anchor.pad1, 0xff, 12);
1558 memset(ddf->anchor.header_ext, 0xff, 32);
1559 ddf->anchor.primary_lba = ~(__u64)0;
1560 ddf->anchor.secondary_lba = ~(__u64)0;
1561 ddf->anchor.type = DDF_HEADER_ANCHOR;
1562 memset(ddf->anchor.pad2, 0xff, 3);
1563 ddf->anchor.workspace_len = __cpu_to_be32(32768); /* Must be reserved */
1564 ddf->anchor.workspace_lba = ~(__u64)0; /* Put this at bottom
1565 of 32M reserved.. */
1566 max_phys_disks = 1023; /* Should be enough */
1567 ddf->anchor.max_pd_entries = __cpu_to_be16(max_phys_disks);
1568 max_virt_disks = 255;
1569 ddf->anchor.max_vd_entries = __cpu_to_be16(max_virt_disks); /* ?? */
1570 ddf->anchor.max_partitions = __cpu_to_be16(64); /* ?? */
1571 ddf->max_part = 64;
8c3b8c2c 1572 ddf->mppe = 256;
59e36268
NB		 1573 ddf->conf_rec_len = 1 + ROUND_UP(ddf->mppe * (4+8), 512)/512;
1574 ddf->anchor.config_record_len = __cpu_to_be16(ddf->conf_rec_len);
1575 ddf->anchor.max_primary_element_entries = __cpu_to_be16(ddf->mppe);
a322f70c 1576 memset(ddf->anchor.pad3, 0xff, 54);
a322f70c
DW		 1577 /* controller sections is one sector long immediately
1578 * after the ddf header */
1579 sector = 1;
1580 ddf->anchor.controller_section_offset = __cpu_to_be32(sector);
1581 ddf->anchor.controller_section_length = __cpu_to_be32(1);
1582 sector += 1;
1583
1584 /* phys is 8 sectors after that */
1585 pdsize = ROUND_UP(sizeof(struct phys_disk) +
1586 sizeof(struct phys_disk_entry)*max_phys_disks,
1587 512);
1588 switch(pdsize/512) {
1589 case 2: case 8: case 32: case 128: case 512: break;
1590 default: abort();
1591 }
1592 ddf->anchor.phys_section_offset = __cpu_to_be32(sector);
1593 ddf->anchor.phys_section_length =
1594 __cpu_to_be32(pdsize/512); /* max_primary_element_entries/8 */
1595 sector += pdsize/512;
1596
1597 /* virt is another 32 sectors */
1598 vdsize = ROUND_UP(sizeof(struct virtual_disk) +
1599 sizeof(struct virtual_entry) * max_virt_disks,
1600 512);
1601 switch(vdsize/512) {
1602 case 2: case 8: case 32: case 128: case 512: break;
1603 default: abort();
1604 }
1605 ddf->anchor.virt_section_offset = __cpu_to_be32(sector);
1606 ddf->anchor.virt_section_length =
1607 __cpu_to_be32(vdsize/512); /* max_vd_entries/8 */
1608 sector += vdsize/512;
1609
59e36268 1610 clen = ddf->conf_rec_len * (ddf->max_part+1);
a322f70c
DW		 1611 ddf->anchor.config_section_offset = __cpu_to_be32(sector);
1612 ddf->anchor.config_section_length = __cpu_to_be32(clen);
1613 sector += clen;
1614
1615 ddf->anchor.data_section_offset = __cpu_to_be32(sector);
1616 ddf->anchor.data_section_length = __cpu_to_be32(1);
1617 sector += 1;
1618
1619 ddf->anchor.bbm_section_length = __cpu_to_be32(0);
1620 ddf->anchor.bbm_section_offset = __cpu_to_be32(0xFFFFFFFF);
1621 ddf->anchor.diag_space_length = __cpu_to_be32(0);
1622 ddf->anchor.diag_space_offset = __cpu_to_be32(0xFFFFFFFF);
1623 ddf->anchor.vendor_length = __cpu_to_be32(0);
1624 ddf->anchor.vendor_offset = __cpu_to_be32(0xFFFFFFFF);
1625
1626 memset(ddf->anchor.pad4, 0xff, 256);
1627
1628 memcpy(&ddf->primary, &ddf->anchor, 512);
1629 memcpy(&ddf->secondary, &ddf->anchor, 512);
1630
1631 ddf->primary.openflag = 1; /* I guess.. */
1632 ddf->primary.type = DDF_HEADER_PRIMARY;
1633
1634 ddf->secondary.openflag = 1; /* I guess.. */
1635 ddf->secondary.type = DDF_HEADER_SECONDARY;
1636
1637 ddf->active = &ddf->primary;
1638
1639 ddf->controller.magic = DDF_CONTROLLER_MAGIC;
1640
1641 /* 24 more bytes of fiction required.
1642 * first 8 are a 'vendor-id' - "Linux-MD"
1643 * Remaining 16 are serial number.... maybe a hostname would do?
1644 */
1645 memcpy(ddf->controller.guid, T10, sizeof(T10));
1ba6bff9
DW		 1646 gethostname(hostname, sizeof(hostname));
1647 hostname[sizeof(hostname) - 1] = 0;
a322f70c
DW		 1648 hostlen = strlen(hostname);
1649 memcpy(ddf->controller.guid + 24 - hostlen, hostname, hostlen);
1650 for (i = strlen(T10) ; i+hostlen < 24; i++)
1651 ddf->controller.guid[i] = ' ';
1652
1653 ddf->controller.type.vendor_id = __cpu_to_be16(0xDEAD);
1654 ddf->controller.type.device_id = __cpu_to_be16(0xBEEF);
1655 ddf->controller.type.sub_vendor_id = 0;
1656 ddf->controller.type.sub_device_id = 0;
1657 memcpy(ddf->controller.product_id, "What Is My PID??", 16);
1658 memset(ddf->controller.pad, 0xff, 8);
1659 memset(ddf->controller.vendor_data, 0xff, 448);
a9e1c11d
N		 1660 if (homehost && strlen(homehost) < 440)
1661 strcpy((char*)ddf->controller.vendor_data, homehost);
a322f70c 1662
3d2c4fc7
DW		 1663 if (posix_memalign((void**)&pd, 512, pdsize) != 0) {
1664 fprintf(stderr, Name ": %s could not allocate pd\n", __func__);
1665 return 0;
1666 }
6416d527 1667 ddf->phys = pd;
a322f70c
DW		 1668 ddf->pdsize = pdsize;
1669
1670 memset(pd, 0xff, pdsize);
1671 memset(pd, 0, sizeof(*pd));
1672 pd->magic = DDF_PHYS_DATA_MAGIC;
1673 pd->used_pdes = __cpu_to_be16(0);
1674 pd->max_pdes = __cpu_to_be16(max_phys_disks);
1675 memset(pd->pad, 0xff, 52);
1676
3d2c4fc7
DW		 1677 if (posix_memalign((void**)&vd, 512, vdsize) != 0) {
1678 fprintf(stderr, Name ": %s could not allocate vd\n", __func__);
1679 return 0;
1680 }
6416d527 1681 ddf->virt = vd;
a322f70c
DW		 1682 ddf->vdsize = vdsize;
1683 memset(vd, 0, vdsize);
1684 vd->magic = DDF_VIRT_RECORDS_MAGIC;
1685 vd->populated_vdes = __cpu_to_be16(0);
1686 vd->max_vdes = __cpu_to_be16(max_virt_disks);
1687 memset(vd->pad, 0xff, 52);
1688
5f8097be
NB		 1689 for (i=0; i<max_virt_disks; i++)
1690 memset(&vd->entries[i], 0xff, sizeof(struct virtual_entry));
1691
a322f70c 1692 st->sb = ddf;
18a2f463 1693 ddf->updates_pending = 1;
a322f70c
DW		 1694 return 1;
1695}
1696
5f8097be
NB		 1697static int chunk_to_shift(int chunksize)
1698{
1699 return ffs(chunksize/512)-1;
1700}
1701
1702static int level_to_prl(int level)
1703{
1704 switch (level) {
1705 case LEVEL_LINEAR: return DDF_CONCAT;
1706 case 0: return DDF_RAID0;
1707 case 1: return DDF_RAID1;
1708 case 4: return DDF_RAID4;
1709 case 5: return DDF_RAID5;
1710 case 6: return DDF_RAID6;
1711 default: return -1;
1712 }
1713}
1714static int layout_to_rlq(int level, int layout, int raiddisks)
1715{
1716 switch(level) {
1717 case 0:
1718 return DDF_RAID0_SIMPLE;
1719 case 1:
1720 switch(raiddisks) {
1721 case 2: return DDF_RAID1_SIMPLE;
1722 case 3: return DDF_RAID1_MULTI;
1723 default: return -1;
1724 }
1725 case 4:
1726 switch(layout) {
1727 case 0: return DDF_RAID4_N;
1728 }
1729 break;
1730 case 5:
1731 case 6:
1732 switch(layout) {
1733 case ALGORITHM_LEFT_ASYMMETRIC:
1734 return DDF_RAID5_N_RESTART;
1735 case ALGORITHM_RIGHT_ASYMMETRIC:
59e36268
NB		 1736 if (level == 5)
1737 return DDF_RAID5_0_RESTART;
1738 else
1739 return DDF_RAID6_0_RESTART;
5f8097be
NB		 1740 case ALGORITHM_LEFT_SYMMETRIC:
1741 return DDF_RAID5_N_CONTINUE;
1742 case ALGORITHM_RIGHT_SYMMETRIC:
1743 return -1; /* not mentioned in standard */
1744 }
1745 }
1746 return -1;
1747}
1748
598f0d58
NB		 1749static int rlq_to_layout(int rlq, int prl, int raiddisks)
1750{
1751 switch(prl) {
1752 case DDF_RAID0:
1753 return 0; /* hopefully rlq == DDF_RAID0_SIMPLE */
1754 case DDF_RAID1:
1755 return 0; /* hopefully rlq == SIMPLE or MULTI depending
1756 on raiddisks*/
1757 case DDF_RAID4:
1758 switch(rlq) {
1759 case DDF_RAID4_N:
1760 return 0;
1761 default:
1762 /* not supported */
1763 return -1; /* FIXME this isn't checked */
1764 }
1765 case DDF_RAID5:
598f0d58
NB		 1766 switch(rlq) {
1767 case DDF_RAID5_N_RESTART:
1768 return ALGORITHM_LEFT_ASYMMETRIC;
1769 case DDF_RAID5_0_RESTART:
1770 return ALGORITHM_RIGHT_ASYMMETRIC;
1771 case DDF_RAID5_N_CONTINUE:
1772 return ALGORITHM_LEFT_SYMMETRIC;
1773 default:
1774 return -1;
1775 }
59e36268
NB		 1776 case DDF_RAID6:
1777 switch(rlq) {
1778 case DDF_RAID5_N_RESTART:
1779 return ALGORITHM_LEFT_ASYMMETRIC;
1780 case DDF_RAID6_0_RESTART:
1781 return ALGORITHM_RIGHT_ASYMMETRIC;
1782 case DDF_RAID5_N_CONTINUE:
1783 return ALGORITHM_LEFT_SYMMETRIC;
1784 default:
1785 return -1;
1786 }
598f0d58
NB		 1787 }
1788 return -1;
1789}
1790
0e600426 1791#ifndef MDASSEMBLE
59e36268
NB		 1792struct extent {
1793 unsigned long long start, size;
1794};
78e44928 1795static int cmp_extent(const void *av, const void *bv)
59e36268
NB		 1796{
1797 const struct extent *a = av;
1798 const struct extent *b = bv;
1799 if (a->start < b->start)
1800 return -1;
1801 if (a->start > b->start)
1802 return 1;
1803 return 0;
1804}
1805
78e44928 1806static struct extent *get_extents(struct ddf_super *ddf, struct dl *dl)
59e36268
NB		 1807{
1808 /* find a list of used extents on the give physical device
1809 * (dnum) of the given ddf.
1810 * Return a malloced array of 'struct extent'
1811
1812FIXME ignore DDF_Legacy devices?
1813
1814 */
1815 struct extent *rv;
1816 int n = 0;
1817 int i, j;
1818
1819 rv = malloc(sizeof(struct extent) * (ddf->max_part + 2));
1820 if (!rv)
1821 return NULL;
1822
1823 for (i = 0; i < ddf->max_part; i++) {
1824 struct vcl *v = dl->vlist[i];
1825 if (v == NULL)
1826 continue;
1827 for (j=0; j < v->conf.prim_elmnt_count; j++)
1828 if (v->conf.phys_refnum[j] == dl->disk.refnum) {
1829 /* This device plays role 'j' in 'v'. */
1830 rv[n].start = __be64_to_cpu(v->lba_offset[j]);
1831 rv[n].size = __be64_to_cpu(v->conf.blocks);
1832 n++;
1833 break;
1834 }
1835 }
1836 qsort(rv, n, sizeof(*rv), cmp_extent);
1837
1838 rv[n].start = __be64_to_cpu(ddf->phys->entries[dl->pdnum].config_size);
1839 rv[n].size = 0;
1840 return rv;
1841}
0e600426 1842#endif
59e36268 1843
5f8097be
NB		 1844static int init_super_ddf_bvd(struct supertype *st,
1845 mdu_array_info_t *info,
1846 unsigned long long size,
1847 char *name, char *homehost,
1848 int *uuid)
1849{
1850 /* We are creating a BVD inside a pre-existing container.
1851 * so st->sb is already set.
1852 * We need to create a new vd_config and a new virtual_entry
1853 */
1854 struct ddf_super *ddf = st->sb;
1855 int venum;
1856 struct virtual_entry *ve;
1857 struct vcl *vcl;
1858 struct vd_config *vc;
5f8097be
NB		 1859
1860 if (__be16_to_cpu(ddf->virt->populated_vdes)
1861 >= __be16_to_cpu(ddf->virt->max_vdes)) {
1862 fprintf(stderr, Name": This ddf already has the "
1863 "maximum of %d virtual devices\n",
1864 __be16_to_cpu(ddf->virt->max_vdes));
1865 return 0;
1866 }
1867
1868 for (venum = 0; venum < __be16_to_cpu(ddf->virt->max_vdes); venum++)
1869 if (all_ff(ddf->virt->entries[venum].guid))
1870 break;
1871 if (venum == __be16_to_cpu(ddf->virt->max_vdes)) {
1872 fprintf(stderr, Name ": Cannot find spare slot for "
1873 "virtual disk - DDF is corrupt\n");
1874 return 0;
1875 }
1876 ve = &ddf->virt->entries[venum];
1877
1878 /* A Virtual Disk GUID contains the T10 Vendor ID, controller type,
1879 * timestamp, random number
1880 */
1881 make_header_guid(ve->guid);
1882 ve->unit = __cpu_to_be16(info->md_minor);
1883 ve->pad0 = 0xFFFF;
1884 ve->guid_crc = crc32(0, (unsigned char*)ddf->anchor.guid, DDF_GUID_LEN);
1885 ve->type = 0;
7a7cc504
NB		 1886 ve->state = DDF_state_degraded; /* Will be modified as devices are added */
1887 if (info->state & 1) /* clean */
1888 ve->init_state = DDF_init_full;
1889 else
1890 ve->init_state = DDF_init_not;
1891
5f8097be
NB		 1892 memset(ve->pad1, 0xff, 14);
1893 memset(ve->name, ' ', 16);
1894 if (name)
1895 strncpy(ve->name, name, 16);
1896 ddf->virt->populated_vdes =
1897 __cpu_to_be16(__be16_to_cpu(ddf->virt->populated_vdes)+1);
1898
1899 /* Now create a new vd_config */
3d2c4fc7
DW		 1900 if (posix_memalign((void**)&vcl, 512,
1901 (offsetof(struct vcl, conf) + ddf->conf_rec_len * 512)) != 0) {
1902 fprintf(stderr, Name ": %s could not allocate vd_config\n", __func__);
1903 return 0;
1904 }
8c3b8c2c 1905 vcl->lba_offset = (__u64*) &vcl->conf.phys_refnum[ddf->mppe];
59e36268 1906 vcl->vcnum = venum;
f7e7067b 1907 sprintf(st->subarray, "%d", venum);
59e36268 1908 vcl->block_sizes = NULL; /* FIXME not for CONCAT */
5f8097be
NB		 1909
1910 vc = &vcl->conf;
1911
1912 vc->magic = DDF_VD_CONF_MAGIC;
1913 memcpy(vc->guid, ve->guid, DDF_GUID_LEN);
1914 vc->timestamp = __cpu_to_be32(time(0)-DECADE);
1915 vc->seqnum = __cpu_to_be32(1);
1916 memset(vc->pad0, 0xff, 24);
1917 vc->prim_elmnt_count = __cpu_to_be16(info->raid_disks);
1918 vc->chunk_shift = chunk_to_shift(info->chunk_size);
1919 vc->prl = level_to_prl(info->level);
1920 vc->rlq = layout_to_rlq(info->level, info->layout, info->raid_disks);
1921 vc->sec_elmnt_count = 1;
1922 vc->sec_elmnt_seq = 0;
1923 vc->srl = 0;
1924 vc->blocks = __cpu_to_be64(info->size * 2);
1925 vc->array_blocks = __cpu_to_be64(
1926 calc_array_size(info->level, info->raid_disks, info->layout,
1927 info->chunk_size, info->size*2));
1928 memset(vc->pad1, 0xff, 8);
1929 vc->spare_refs[0] = 0xffffffff;
1930 vc->spare_refs[1] = 0xffffffff;
1931 vc->spare_refs[2] = 0xffffffff;
1932 vc->spare_refs[3] = 0xffffffff;
1933 vc->spare_refs[4] = 0xffffffff;
1934 vc->spare_refs[5] = 0xffffffff;
1935 vc->spare_refs[6] = 0xffffffff;
1936 vc->spare_refs[7] = 0xffffffff;
1937 memset(vc->cache_pol, 0, 8);
1938 vc->bg_rate = 0x80;
1939 memset(vc->pad2, 0xff, 3);
1940 memset(vc->pad3, 0xff, 52);
1941 memset(vc->pad4, 0xff, 192);
1942 memset(vc->v0, 0xff, 32);
1943 memset(vc->v1, 0xff, 32);
1944 memset(vc->v2, 0xff, 16);
1945 memset(vc->v3, 0xff, 16);
1946 memset(vc->vendor, 0xff, 32);
598f0d58 1947
8c3b8c2c
NB		 1948 memset(vc->phys_refnum, 0xff, 4*ddf->mppe);
1949 memset(vc->phys_refnum+(ddf->mppe * 4), 0x00, 8*ddf->mppe);
5f8097be
NB		 1950
1951 vcl->next = ddf->conflist;
1952 ddf->conflist = vcl;
d2ca6449 1953 ddf->currentconf = vcl;
18a2f463 1954 ddf->updates_pending = 1;
5f8097be
NB		 1955 return 1;
1956}
1957
0e600426 1958#ifndef MDASSEMBLE
5f8097be
NB		 1959static void add_to_super_ddf_bvd(struct supertype *st,
1960 mdu_disk_info_t *dk, int fd, char *devname)
1961{
1962 /* fd and devname identify a device with-in the ddf container (st).
1963 * dk identifies a location in the new BVD.
1964 * We need to find suitable free space in that device and update
1965 * the phys_refnum and lba_offset for the newly created vd_config.
1966 * We might also want to update the type in the phys_disk
5575e7d9 1967 * section.
5f8097be
NB		 1968 */
1969 struct dl *dl;
1970 struct ddf_super *ddf = st->sb;
1971 struct vd_config *vc;
1972 __u64 *lba_offset;
7a7cc504 1973 int working;
5575e7d9 1974 int i;
59e36268
NB		 1975 unsigned long long blocks, pos, esize;
1976 struct extent *ex;
5f8097be
NB		 1977
1978 for (dl = ddf->dlist; dl ; dl = dl->next)
1979 if (dl->major == dk->major &&
1980 dl->minor == dk->minor)
1981 break;
1982 if (!dl || ! (dk->state & (1<<MD_DISK_SYNC)))
1983 return;
1984
d2ca6449
NB		 1985 vc = &ddf->currentconf->conf;
1986 lba_offset = ddf->currentconf->lba_offset;
59e36268
NB		 1987
1988 ex = get_extents(ddf, dl);
1989 if (!ex)
1990 return;
1991
1992 i = 0; pos = 0;
1993 blocks = __be64_to_cpu(vc->blocks);
d2ca6449
NB		 1994 if (ddf->currentconf->block_sizes)
1995 blocks = ddf->currentconf->block_sizes[dk->raid_disk];
59e36268
NB		 1996
1997 do {
1998 esize = ex[i].start - pos;
1999 if (esize >= blocks)
2000 break;
2001 pos = ex[i].start + ex[i].size;
2002 i++;
2003 } while (ex[i-1].size);
2004
2005 free(ex);
2006 if (esize < blocks)
2007 return;
2008
d2ca6449 2009 ddf->currentdev = dk->raid_disk;
5f8097be 2010 vc->phys_refnum[dk->raid_disk] = dl->disk.refnum;
59e36268 2011 lba_offset[dk->raid_disk] = __cpu_to_be64(pos);
5f8097be 2012
5575e7d9
NB		 2013 for (i=0; i < ddf->max_part ; i++)
2014 if (dl->vlist[i] == NULL)
2015 break;
2016 if (i == ddf->max_part)
2017 return;
d2ca6449 2018 dl->vlist[i] = ddf->currentconf;
5f8097be
NB		 2019
2020 dl->fd = fd;
2021 dl->devname = devname;
7a7cc504
NB		 2022
2023 /* Check how many working raid_disks, and if we can mark
2024 * array as optimal yet
2025 */
2026 working = 0;
5575e7d9 2027
7a7cc504
NB		 2028 for (i=0; i < __be16_to_cpu(vc->prim_elmnt_count); i++)
2029 if (vc->phys_refnum[i] != 0xffffffff)
2030 working++;
59e36268 2031
5575e7d9 2032 /* Find which virtual_entry */
d2ca6449 2033 i = ddf->currentconf->vcnum;
7a7cc504 2034 if (working == __be16_to_cpu(vc->prim_elmnt_count))
5575e7d9
NB		 2035 ddf->virt->entries[i].state =
2036 (ddf->virt->entries[i].state & ~DDF_state_mask)
7a7cc504
NB		 2037 | DDF_state_optimal;
2038
2039 if (vc->prl == DDF_RAID6 &&
2040 working+1 == __be16_to_cpu(vc->prim_elmnt_count))
5575e7d9
NB		 2041 ddf->virt->entries[i].state =
2042 (ddf->virt->entries[i].state & ~DDF_state_mask)
7a7cc504 2043 | DDF_state_part_optimal;
5575e7d9
NB		 2044
2045 ddf->phys->entries[dl->pdnum].type &= ~__cpu_to_be16(DDF_Global_Spare);
2046 ddf->phys->entries[dl->pdnum].type |= __cpu_to_be16(DDF_Active_in_VD);
18a2f463 2047 ddf->updates_pending = 1;
5f8097be
NB		 2048}
2049
a322f70c
DW		 2050/* add a device to a container, either while creating it or while
2051 * expanding a pre-existing container
2052 */
2053static void add_to_super_ddf(struct supertype *st,
2054 mdu_disk_info_t *dk, int fd, char *devname)
2055{
2056 struct ddf_super *ddf = st->sb;
2057 struct dl *dd;
2058 time_t now;
2059 struct tm *tm;
2060 unsigned long long size;
2061 struct phys_disk_entry *pde;
2062 int n, i;
2063 struct stat stb;
2064
78e44928
NB		 2065 if (ddf->currentconf) {
2066 add_to_super_ddf_bvd(st, dk, fd, devname);
2067 return;
2068 }
2069
a322f70c
DW		 2070 /* This is device numbered dk->number. We need to create
2071 * a phys_disk entry and a more detailed disk_data entry.
2072 */
2073 fstat(fd, &stb);
3d2c4fc7
DW		 2074 if (posix_memalign((void**)&dd, 512,
2075 sizeof(*dd) + sizeof(dd->vlist[0]) * ddf->max_part) != 0) {
2076 fprintf(stderr, Name
2077 ": %s could allocate buffer for new disk, aborting\n",
2078 __func__);
2079 abort();
2080 }
a322f70c
DW		 2081 dd->major = major(stb.st_rdev);
2082 dd->minor = minor(stb.st_rdev);
2083 dd->devname = devname;
a322f70c 2084 dd->fd = fd;
b2280677 2085 dd->spare = NULL;
a322f70c
DW		 2086
2087 dd->disk.magic = DDF_PHYS_DATA_MAGIC;
2088 now = time(0);
2089 tm = localtime(&now);
2090 sprintf(dd->disk.guid, "%8s%04d%02d%02d",
2091 T10, tm->tm_year+1900, tm->tm_mon+1, tm->tm_mday);
2092 *(__u32*)(dd->disk.guid + 16) = random();
2093 *(__u32*)(dd->disk.guid + 20) = random();
2094
59e36268
NB		 2095 do {
2096 /* Cannot be bothered finding a CRC of some irrelevant details*/
2097 dd->disk.refnum = random();
2098 for (i = __be16_to_cpu(ddf->active->max_pd_entries) - 1;
2099 i >= 0; i--)
2100 if (ddf->phys->entries[i].refnum == dd->disk.refnum)
2101 break;
2102 } while (i >= 0);
2103
a322f70c
DW		 2104 dd->disk.forced_ref = 1;
2105 dd->disk.forced_guid = 1;
2106 memset(dd->disk.vendor, ' ', 32);
2107 memcpy(dd->disk.vendor, "Linux", 5);
2108 memset(dd->disk.pad, 0xff, 442);
b2280677 2109 for (i = 0; i < ddf->max_part ; i++)
a322f70c
DW		 2110 dd->vlist[i] = NULL;
2111
2112 n = __be16_to_cpu(ddf->phys->used_pdes);
2113 pde = &ddf->phys->entries[n];
5575e7d9
NB		 2114 dd->pdnum = n;
2115
2cc2983d
N		 2116 if (st->update_tail) {
2117 int len = (sizeof(struct phys_disk) +
2118 sizeof(struct phys_disk_entry));
2119 struct phys_disk *pd;
2120
2121 pd = malloc(len);
2122 pd->magic = DDF_PHYS_RECORDS_MAGIC;
2123 pd->used_pdes = __cpu_to_be16(n);
2124 pde = &pd->entries[0];
2125 dd->mdupdate = pd;
2126 } else {
2127 n++;
2128 ddf->phys->used_pdes = __cpu_to_be16(n);
2129 }
a322f70c
DW		 2130
2131 memcpy(pde->guid, dd->disk.guid, DDF_GUID_LEN);
2132 pde->refnum = dd->disk.refnum;
5575e7d9 2133 pde->type = __cpu_to_be16(DDF_Forced_PD_GUID | DDF_Global_Spare);
a322f70c
DW		 2134 pde->state = __cpu_to_be16(DDF_Online);
2135 get_dev_size(fd, NULL, &size);
2136 /* We are required to reserve 32Meg, and record the size in sectors */
2137 pde->config_size = __cpu_to_be64( (size - 32*1024*1024) / 512);
2138 sprintf(pde->path, "%17.17s","Information: nil") ;
2139 memset(pde->pad, 0xff, 6);
2140
d2ca6449 2141 dd->size = size >> 9;
2cc2983d
N		 2142 if (st->update_tail) {
2143 dd->next = ddf->add_list;
2144 ddf->add_list = dd;
2145 } else {
2146 dd->next = ddf->dlist;
2147 ddf->dlist = dd;
2148 ddf->updates_pending = 1;
2149 }
a322f70c
DW		 2150}
2151
2152/*
2153 * This is the write_init_super method for a ddf container. It is
2154 * called when creating a container or adding another device to a
2155 * container.
2156 */
2157
6416d527 2158static unsigned char null_conf[4096+512];
18a2f463 2159
7a7cc504 2160static int __write_init_super_ddf(struct supertype *st, int do_close)
a322f70c
DW		 2161{
2162
2163 struct ddf_super *ddf = st->sb;
2164 int i;
2165 struct dl *d;
2166 int n_config;
2167 int conf_size;
175593bf
DW		 2168 int attempts = 0;
2169 int successes = 0;
a322f70c
DW		 2170 unsigned long long size, sector;
2171
175593bf
DW		 2172 /* try to write updated metadata,
2173 * if we catch a failure move on to the next disk
2174 */
a322f70c
DW		 2175 for (d = ddf->dlist; d; d=d->next) {
2176 int fd = d->fd;
2177
2178 if (fd < 0)
2179 continue;
2180
175593bf 2181 attempts++;
a322f70c
DW		 2182 /* We need to fill in the primary, (secondary) and workspace
2183 * lba's in the headers, set their checksums,
2184 * Also checksum phys, virt....
2185 *
2186 * Then write everything out, finally the anchor is written.
2187 */
2188 get_dev_size(fd, NULL, &size);
2189 size /= 512;
2190 ddf->anchor.workspace_lba = __cpu_to_be64(size - 32*1024*2);
2191 ddf->anchor.primary_lba = __cpu_to_be64(size - 16*1024*2);
2192 ddf->anchor.seq = __cpu_to_be32(1);
2193 memcpy(&ddf->primary, &ddf->anchor, 512);
2194 memcpy(&ddf->secondary, &ddf->anchor, 512);
2195
2196 ddf->anchor.openflag = 0xFF; /* 'open' means nothing */
2197 ddf->anchor.seq = 0xFFFFFFFF; /* no sequencing in anchor */
2198 ddf->anchor.crc = calc_crc(&ddf->anchor, 512);
2199
2200 ddf->primary.openflag = 0;
2201 ddf->primary.type = DDF_HEADER_PRIMARY;
2202
2203 ddf->secondary.openflag = 0;
2204 ddf->secondary.type = DDF_HEADER_SECONDARY;
2205
2206 ddf->primary.crc = calc_crc(&ddf->primary, 512);
2207 ddf->secondary.crc = calc_crc(&ddf->secondary, 512);
2208
2209 sector = size - 16*1024*2;
2210 lseek64(fd, sector<<9, 0);
175593bf
DW		 2211 if (write(fd, &ddf->primary, 512) < 0)
2212 continue;
a322f70c
DW		 2213
2214 ddf->controller.crc = calc_crc(&ddf->controller, 512);
175593bf
DW		 2215 if (write(fd, &ddf->controller, 512) < 0)
2216 continue;
a322f70c
DW		 2217
2218 ddf->phys->crc = calc_crc(ddf->phys, ddf->pdsize);
2219
175593bf
DW		 2220 if (write(fd, ddf->phys, ddf->pdsize) < 0)
2221 continue;
a322f70c
DW		 2222
2223 ddf->virt->crc = calc_crc(ddf->virt, ddf->vdsize);
175593bf
DW		 2224 if (write(fd, ddf->virt, ddf->vdsize) < 0)
2225 continue;
a322f70c
DW		 2226
2227 /* Now write lots of config records. */
8c3b8c2c
NB		 2228 n_config = ddf->max_part;
2229 conf_size = ddf->conf_rec_len * 512;
a322f70c
DW		 2230 for (i = 0 ; i <= n_config ; i++) {
2231 struct vcl *c = d->vlist[i];
b2280677
NB		 2232 if (i == n_config)
2233 c = (struct vcl*)d->spare;
a322f70c
DW		 2234
2235 if (c) {
2236 c->conf.crc = calc_crc(&c->conf, conf_size);
175593bf
DW		 2237 if (write(fd, &c->conf, conf_size) < 0)
2238 break;
a322f70c 2239 } else {
6416d527 2240 char *null_aligned = (char*)((((unsigned long)null_conf)+511)&~511UL);
18a2f463
NB		 2241 if (null_conf[0] != 0xff)
2242 memset(null_conf, 0xff, sizeof(null_conf));
2243 int togo = conf_size;
6416d527 2244 while (togo > sizeof(null_conf)-512) {
175593bf
DW		 2245 if (write(fd, null_aligned, sizeof(null_conf)-512) < 0)
2246 break;
6416d527 2247 togo -= sizeof(null_conf)-512;
18a2f463 2248 }
175593bf
DW		 2249 if (write(fd, null_aligned, togo) < 0)
2250 break;
a322f70c
DW		 2251 }
2252 }
175593bf
DW		 2253 if (i <= n_config)
2254 continue;
a322f70c 2255 d->disk.crc = calc_crc(&d->disk, 512);
175593bf
DW		 2256 if (write(fd, &d->disk, 512) < 0)
2257 continue;
a322f70c
DW		 2258
2259 /* Maybe do the same for secondary */
2260
2261 lseek64(fd, (size-1)*512, SEEK_SET);
175593bf
DW		 2262 if (write(fd, &ddf->anchor, 512) < 0)
2263 continue;
2264 successes++;
2265 }
2266
2267 if (do_close)
2268 for (d = ddf->dlist; d; d=d->next) {
2269 close(d->fd);
7a7cc504
NB		 2270 d->fd = -1;
2271 }
175593bf
DW		 2272
2273 return attempts != successes;
a322f70c 2274}
7a7cc504
NB		 2275
2276static int write_init_super_ddf(struct supertype *st)
2277{
edd8d13c
NB		 2278
2279 if (st->update_tail) {
2280 /* queue the virtual_disk and vd_config as metadata updates */
2281 struct virtual_disk *vd;
2282 struct vd_config *vc;
2283 struct ddf_super *ddf = st->sb;
2284 int len;
2285
2cc2983d
N		 2286 if (!ddf->currentconf) {
2287 int len = (sizeof(struct phys_disk) +
2288 sizeof(struct phys_disk_entry));
2289
2290 /* adding a disk to the container. */
2291 if (!ddf->add_list)
2292 return 0;
2293
2294 append_metadata_update(st, ddf->add_list->mdupdate, len);
2295 ddf->add_list->mdupdate = NULL;
2296 return 0;
2297 }
2298
2299 /* Newly created VD */
2300
edd8d13c
NB		 2301 /* First the virtual disk. We have a slightly fake header */
2302 len = sizeof(struct virtual_disk) + sizeof(struct virtual_entry);
2303 vd = malloc(len);
2304 *vd = *ddf->virt;
2305 vd->entries[0] = ddf->virt->entries[ddf->currentconf->vcnum];
2306 vd->populated_vdes = __cpu_to_be16(ddf->currentconf->vcnum);
2307 append_metadata_update(st, vd, len);
2308
2309 /* Then the vd_config */
2310 len = ddf->conf_rec_len * 512;
2311 vc = malloc(len);
2312 memcpy(vc, &ddf->currentconf->conf, len);
2313 append_metadata_update(st, vc, len);
2314
2315 /* FIXME I need to close the fds! */
2316 return 0;
2317 } else
2318 return __write_init_super_ddf(st, 1);
7a7cc504
NB		 2319}
2320
a322f70c
DW		 2321#endif
2322
2323static __u64 avail_size_ddf(struct supertype *st, __u64 devsize)
2324{
2325 /* We must reserve the last 32Meg */
2326 if (devsize <= 32*1024*2)
2327 return 0;
2328 return devsize - 32*1024*2;
2329}
2330
2331#ifndef MDASSEMBLE
2c514b71
NB		 2332static int
2333validate_geometry_ddf_container(struct supertype *st,
2334 int level, int layout, int raiddisks,
2335 int chunk, unsigned long long size,
2336 char *dev, unsigned long long *freesize,
2337 int verbose);
78e44928
NB		 2338
2339static int validate_geometry_ddf_bvd(struct supertype *st,
2340 int level, int layout, int raiddisks,
2341 int chunk, unsigned long long size,
2c514b71
NB		 2342 char *dev, unsigned long long *freesize,
2343 int verbose);
78e44928
NB		 2344
2345static int validate_geometry_ddf(struct supertype *st,
2c514b71
NB		 2346 int level, int layout, int raiddisks,
2347 int chunk, unsigned long long size,
2348 char *dev, unsigned long long *freesize,
2349 int verbose)
a322f70c
DW		 2350{
2351 int fd;
2352 struct mdinfo *sra;
2353 int cfd;
2354
2355 /* ddf potentially supports lots of things, but it depends on
2356 * what devices are offered (and maybe kernel version?)
2357 * If given unused devices, we will make a container.
2358 * If given devices in a container, we will make a BVD.
2359 * If given BVDs, we make an SVD, changing all the GUIDs in the process.
2360 */
2361
2362 if (level == LEVEL_CONTAINER) {
78e44928
NB		 2363 /* Must be a fresh device to add to a container */
2364 return validate_geometry_ddf_container(st, level, layout,
2c514b71
NB		 2365 raiddisks, chunk,
2366 size, dev, freesize,
2367 verbose);
5f8097be
NB		 2368 }
2369
2370 if (st->sb) {
78e44928
NB		 2371 /* A container has already been opened, so we are
2372 * creating in there. Maybe a BVD, maybe an SVD.
2373 * Should make a distinction one day.
2374 */
2375 return validate_geometry_ddf_bvd(st, level, layout, raiddisks,
2c514b71
NB		 2376 chunk, size, dev, freesize,
2377 verbose);
a322f70c 2378 }
78e44928
NB		 2379 if (!dev) {
2380 /* Initial sanity check. Exclude illegal levels. */
2381 int i;
2382 for (i=0; ddf_level_num[i].num1 != MAXINT; i++)
2383 if (ddf_level_num[i].num2 == level)
2384 break;
2385 if (ddf_level_num[i].num1 == MAXINT)
2386 return 0;
2387 /* Should check layout? etc */
a322f70c 2388 return 1;
78e44928 2389 }
a322f70c 2390
78e44928
NB		 2391 /* This is the first device for the array.
2392 * If it is a container, we read it in and do automagic allocations,
2393 * no other devices should be given.
2394 * Otherwise it must be a member device of a container, and we
2395 * do manual allocation.
2396 * Later we should check for a BVD and make an SVD.
a322f70c 2397 */
a322f70c
DW		 2398 fd = open(dev, O_RDONLY|O_EXCL, 0);
2399 if (fd >= 0) {
2400 sra = sysfs_read(fd, 0, GET_VERSION);
2401 close(fd);
2402 if (sra && sra->array.major_version == -1 &&
78e44928
NB		 2403 strcmp(sra->text_version, "ddf") == 0) {
2404
2405 /* load super */
2406 /* find space for 'n' devices. */
2407 /* remember the devices */
2408 /* Somehow return the fact that we have enough */
a322f70c
DW		 2409 }
2410
2c514b71
NB		 2411 if (verbose)
2412 fprintf(stderr,
2413 Name ": ddf: Cannot create this array "
2414 "on device %s\n",
2415 dev);
a322f70c
DW		 2416 return 0;
2417 }
2418 if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) {
2c514b71
NB		 2419 if (verbose)
2420 fprintf(stderr, Name ": ddf: Cannot open %s: %s\n",
2421 dev, strerror(errno));
a322f70c
DW		 2422 return 0;
2423 }
2424 /* Well, it is in use by someone, maybe a 'ddf' container. */
2425 cfd = open_container(fd);
2426 if (cfd < 0) {
2427 close(fd);
2c514b71
NB		 2428 if (verbose)
2429 fprintf(stderr, Name ": ddf: Cannot use %s: %s\n",
2430 dev, strerror(EBUSY));
a322f70c
DW		 2431 return 0;
2432 }
2433 sra = sysfs_read(cfd, 0, GET_VERSION);
2434 close(fd);
2435 if (sra && sra->array.major_version == -1 &&
2436 strcmp(sra->text_version, "ddf") == 0) {
2437 /* This is a member of a ddf container. Load the container
2438 * and try to create a bvd
2439 */
2440 struct ddf_super *ddf;
a322f70c 2441 if (load_super_ddf_all(st, cfd, (void **)&ddf, NULL, 1) == 0) {
5f8097be 2442 st->sb = ddf;
2f6079dc 2443 st->container_dev = fd2devnum(cfd);
a322f70c 2444 close(cfd);
78e44928 2445 return validate_geometry_ddf_bvd(st, level, layout,
a322f70c 2446 raiddisks, chunk, size,
2c514b71
NB		 2447 dev, freesize,
2448 verbose);
a322f70c
DW		 2449 }
2450 close(cfd);
c42ec1ed
DW		 2451 } else /* device may belong to a different container */
2452 return 0;
2453
a322f70c
DW		 2454 return 1;
2455}
2456
2c514b71
NB		 2457static int
2458validate_geometry_ddf_container(struct supertype *st,
2459 int level, int layout, int raiddisks,
2460 int chunk, unsigned long long size,
2461 char *dev, unsigned long long *freesize,
2462 int verbose)
a322f70c
DW		 2463{
2464 int fd;
2465 unsigned long long ldsize;
2466
2467 if (level != LEVEL_CONTAINER)
2468 return 0;
2469 if (!dev)
2470 return 1;
2471
2472 fd = open(dev, O_RDONLY|O_EXCL, 0);
2473 if (fd < 0) {
2c514b71
NB		 2474 if (verbose)
2475 fprintf(stderr, Name ": ddf: Cannot open %s: %s\n",
2476 dev, strerror(errno));
a322f70c
DW		 2477 return 0;
2478 }
2479 if (!get_dev_size(fd, dev, &ldsize)) {
2480 close(fd);
2481 return 0;
2482 }
2483 close(fd);
2484
56aca704 2485 *freesize = avail_size_ddf(st, ldsize >> 9);
a322f70c
DW		 2486
2487 return 1;
2488}
2489
78e44928
NB		 2490static int validate_geometry_ddf_bvd(struct supertype *st,
2491 int level, int layout, int raiddisks,
2492 int chunk, unsigned long long size,
2c514b71
NB		 2493 char *dev, unsigned long long *freesize,
2494 int verbose)
a322f70c
DW		 2495{
2496 struct stat stb;
2497 struct ddf_super *ddf = st->sb;
2498 struct dl *dl;
5f8097be
NB		 2499 unsigned long long pos = 0;
2500 unsigned long long maxsize;
2501 struct extent *e;
2502 int i;
a322f70c
DW		 2503 /* ddf/bvd supports lots of things, but not containers */
2504 if (level == LEVEL_CONTAINER)
2505 return 0;
2506 /* We must have the container info already read in. */
2507 if (!ddf)
2508 return 0;
2509
5f8097be
NB		 2510 if (!dev) {
2511 /* General test: make sure there is space for
2512 * 'raiddisks' device extents of size 'size'.
2513 */
2514 unsigned long long minsize = size;
2515 int dcnt = 0;
2516 if (minsize == 0)
2517 minsize = 8;
2518 for (dl = ddf->dlist; dl ; dl = dl->next)
2519 {
2520 int found = 0;
7e1432fb 2521 pos = 0;
5f8097be
NB		 2522
2523 i = 0;
2524 e = get_extents(ddf, dl);
2525 if (!e) continue;
2526 do {
2527 unsigned long long esize;
2528 esize = e[i].start - pos;
2529 if (esize >= minsize)
2530 found = 1;
2531 pos = e[i].start + e[i].size;
2532 i++;
2533 } while (e[i-1].size);
2534 if (found)
2535 dcnt++;
2536 free(e);
2537 }
2538 if (dcnt < raiddisks) {
2c514b71
NB		 2539 if (verbose)
2540 fprintf(stderr,
2541 Name ": ddf: Not enough devices with "
2542 "space for this array (%d < %d)\n",
2543 dcnt, raiddisks);
5f8097be
NB		 2544 return 0;
2545 }
2546 return 1;
2547 }
a322f70c
DW		 2548 /* This device must be a member of the set */
2549 if (stat(dev, &stb) < 0)
2550 return 0;
2551 if ((S_IFMT & stb.st_mode) != S_IFBLK)
2552 return 0;
2553 for (dl = ddf->dlist ; dl ; dl = dl->next) {
2554 if (dl->major == major(stb.st_rdev) &&
2555 dl->minor == minor(stb.st_rdev))
2556 break;
2557 }
5f8097be 2558 if (!dl) {
2c514b71
NB		 2559 if (verbose)
2560 fprintf(stderr, Name ": ddf: %s is not in the "
2561 "same DDF set\n",
2562 dev);
5f8097be
NB		 2563 return 0;
2564 }
2565 e = get_extents(ddf, dl);
2566 maxsize = 0;
2567 i = 0;
2568 if (e) do {
2569 unsigned long long esize;
2570 esize = e[i].start - pos;
2571 if (esize >= maxsize)
2572 maxsize = esize;
2573 pos = e[i].start + e[i].size;
2574 i++;
2575 } while (e[i-1].size);
2576 *freesize = maxsize;
a322f70c
DW		 2577 // FIXME here I am
2578
2579 return 1;
2580}
59e36268 2581
a322f70c
DW		 2582static int load_super_ddf_all(struct supertype *st, int fd,
2583 void **sbp, char *devname, int keep_fd)
2584{
2585 struct mdinfo *sra;
2586 struct ddf_super *super;
2587 struct mdinfo *sd, *best = NULL;
2588 int bestseq = 0;
2589 int seq;
2590 char nm[20];
2591 int dfd;
2592
2593 sra = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE);
2594 if (!sra)
2595 return 1;
2596 if (sra->array.major_version != -1 ||
2597 sra->array.minor_version != -2 ||
2598 strcmp(sra->text_version, "ddf") != 0)
2599 return 1;
2600
6416d527 2601 if (posix_memalign((void**)&super, 512, sizeof(*super)) != 0)
a322f70c 2602 return 1;
a2349791 2603 memset(super, 0, sizeof(*super));
a322f70c
DW		 2604
2605 /* first, try each device, and choose the best ddf */
2606 for (sd = sra->devs ; sd ; sd = sd->next) {
2607 int rv;
2608 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
7a7cc504
NB		 2609 dfd = dev_open(nm, O_RDONLY);
2610 if (dfd < 0)
a322f70c
DW		 2611 return 2;
2612 rv = load_ddf_headers(dfd, super, NULL);
7a7cc504 2613 close(dfd);
a322f70c
DW		 2614 if (rv == 0) {
2615 seq = __be32_to_cpu(super->active->seq);
2616 if (super->active->openflag)
2617 seq--;
2618 if (!best || seq > bestseq) {
2619 bestseq = seq;
2620 best = sd;
2621 }
2622 }
2623 }
2624 if (!best)
2625 return 1;
2626 /* OK, load this ddf */
2627 sprintf(nm, "%d:%d", best->disk.major, best->disk.minor);
2628 dfd = dev_open(nm, O_RDONLY);
7a7cc504 2629 if (dfd < 0)
a322f70c
DW		 2630 return 1;
2631 load_ddf_headers(dfd, super, NULL);
2632 load_ddf_global(dfd, super, NULL);
2633 close(dfd);
2634 /* Now we need the device-local bits */
2635 for (sd = sra->devs ; sd ; sd = sd->next) {
3d2c4fc7
DW		 2636 int rv;
2637
a322f70c 2638 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
5f8097be 2639 dfd = dev_open(nm, keep_fd? O_RDWR : O_RDONLY);
7a7cc504 2640 if (dfd < 0)
a322f70c 2641 return 2;
3d2c4fc7
DW		 2642 rv = load_ddf_headers(dfd, super, NULL);
2643 if (rv == 0)
2644 rv = load_ddf_local(dfd, super, NULL, keep_fd);
5f8097be 2645 if (!keep_fd) close(dfd);
3d2c4fc7
DW		 2646 if (rv)
2647 return 1;
a322f70c 2648 }
f7e7067b
NB		 2649 if (st->subarray[0]) {
2650 struct vcl *v;
2651
2652 for (v = super->conflist; v; v = v->next)
2653 if (v->vcnum == atoi(st->subarray))
d2ca6449
NB		 2654 super->currentconf = v;
2655 if (!super->currentconf)
f7e7067b
NB		 2656 return 1;
2657 }
a322f70c
DW		 2658 *sbp = super;
2659 if (st->ss == NULL) {
78e44928 2660 st->ss = &super_ddf;
a322f70c
DW		 2661 st->minor_version = 0;
2662 st->max_devs = 512;
75aa18b5 2663 st->container_dev = fd2devnum(fd);
a322f70c 2664 }
352452c3 2665 st->loaded_container = 1;
a322f70c
DW		 2666 return 0;
2667}
0e600426 2668#endif /* MDASSEMBLE */
a322f70c 2669
598f0d58
NB		 2670static struct mdinfo *container_content_ddf(struct supertype *st)
2671{
2672 /* Given a container loaded by load_super_ddf_all,
2673 * extract information about all the arrays into
2674 * an mdinfo tree.
2675 *
2676 * For each vcl in conflist: create an mdinfo, fill it in,
2677 * then look for matching devices (phys_refnum) in dlist
2678 * and create appropriate device mdinfo.
2679 */
2680 struct ddf_super *ddf = st->sb;
2681 struct mdinfo *rest = NULL;
2682 struct vcl *vc;
2683
2684 for (vc = ddf->conflist ; vc ; vc=vc->next)
2685 {
598f0d58 2686 int i;
db42fa9b 2687 int j;
598f0d58
NB		 2688 struct mdinfo *this;
2689 this = malloc(sizeof(*this));
2690 memset(this, 0, sizeof(*this));
2691 this->next = rest;
2692 rest = this;
2693
598f0d58
NB		 2694 this->array.level = map_num1(ddf_level_num, vc->conf.prl);
2695 this->array.raid_disks =
2696 __be16_to_cpu(vc->conf.prim_elmnt_count);
60f18132
NB		 2697 this->array.layout = rlq_to_layout(vc->conf.rlq, vc->conf.prl,
2698 this->array.raid_disks);
598f0d58 2699 this->array.md_minor = -1;
f35f2525
N		 2700 this->array.major_version = -1;
2701 this->array.minor_version = -2;
598f0d58
NB		 2702 this->array.ctime = DECADE +
2703 __be32_to_cpu(*(__u32*)(vc->conf.guid+16));
2704 this->array.utime = DECADE +
2705 __be32_to_cpu(vc->conf.timestamp);
2706 this->array.chunk_size = 512 << vc->conf.chunk_shift;
2707
59e36268 2708 i = vc->vcnum;
7a7cc504
NB		 2709 if ((ddf->virt->entries[i].state & DDF_state_inconsistent) ||
2710 (ddf->virt->entries[i].init_state & DDF_initstate_mask) !=
ed9d66aa 2711 DDF_init_full) {
598f0d58 2712 this->array.state = 0;
ed9d66aa
NB		 2713 this->resync_start = 0;
2714 } else {
598f0d58 2715 this->array.state = 1;
ed9d66aa
NB		 2716 this->resync_start = ~0ULL;
2717 }
db42fa9b
N		 2718 memcpy(this->name, ddf->virt->entries[i].name, 16);
2719 this->name[16]=0;
2720 for(j=0; j<16; j++)
2721 if (this->name[j] == ' ')
2722 this->name[j] = 0;
598f0d58
NB		 2723
2724 memset(this->uuid, 0, sizeof(this->uuid));
2725 this->component_size = __be64_to_cpu(vc->conf.blocks);
2726 this->array.size = this->component_size / 2;
5f2aace8 2727 this->container_member = i;
598f0d58 2728
c5afc314
N		 2729 ddf->currentconf = vc;
2730 uuid_from_super_ddf(st, this->uuid);
2731 ddf->currentconf = NULL;
2732
60f18132
NB		 2733 sprintf(this->text_version, "/%s/%d",
2734 devnum2devname(st->container_dev),
2735 this->container_member);
2736
8c3b8c2c 2737 for (i=0 ; i < ddf->mppe ; i++) {
598f0d58
NB		 2738 struct mdinfo *dev;
2739 struct dl *d;
2740
2741 if (vc->conf.phys_refnum[i] == 0xFFFFFFFF)
2742 continue;
2743
2744 this->array.working_disks++;
2745
2746 for (d = ddf->dlist; d ; d=d->next)
2747 if (d->disk.refnum == vc->conf.phys_refnum[i])
2748 break;
2749 if (d == NULL)
2750 break;
2751
2752 dev = malloc(sizeof(*dev));
2753 memset(dev, 0, sizeof(*dev));
2754 dev->next = this->devs;
2755 this->devs = dev;
2756
2757 dev->disk.number = __be32_to_cpu(d->disk.refnum);
2758 dev->disk.major = d->major;
2759 dev->disk.minor = d->minor;
2760 dev->disk.raid_disk = i;
2761 dev->disk.state = (1<<MD_DISK_SYNC)|(1<<MD_DISK_ACTIVE);
2762
120f7677
NB		 2763 dev->events = __be32_to_cpu(ddf->primary.seq);
2764 dev->data_offset = __be64_to_cpu(vc->lba_offset[i]);
598f0d58
NB		 2765 dev->component_size = __be64_to_cpu(vc->conf.blocks);
2766 if (d->devname)
2767 strcpy(dev->name, d->devname);
2768 }
2769 }
2770 return rest;
2771}
2772
a322f70c
DW		 2773static int store_zero_ddf(struct supertype *st, int fd)
2774{
2775 unsigned long long dsize;
6416d527 2776 void *buf;
3d2c4fc7 2777 int rc;
a322f70c 2778
a322f70c
DW		 2779 if (!get_dev_size(fd, NULL, &dsize))
2780 return 1;
2781
3d2c4fc7
DW		 2782 if (posix_memalign(&buf, 512, 512) != 0)
2783 return 1;
6416d527
NB		 2784 memset(buf, 0, 512);
2785
a322f70c 2786 lseek64(fd, dsize-512, 0);
3d2c4fc7 2787 rc = write(fd, buf, 512);
6416d527 2788 free(buf);
3d2c4fc7
DW		 2789 if (rc < 0)
2790 return 1;
a322f70c
DW		 2791 return 0;
2792}
2793
a19c88b8
NB		 2794static int compare_super_ddf(struct supertype *st, struct supertype *tst)
2795{
2796 /*
2797 * return:
2798 * 0 same, or first was empty, and second was copied
2799 * 1 second had wrong number
2800 * 2 wrong uuid
2801 * 3 wrong other info
2802 */
2803 struct ddf_super *first = st->sb;
2804 struct ddf_super *second = tst->sb;
2805
2806 if (!first) {
2807 st->sb = tst->sb;
2808 tst->sb = NULL;
2809 return 0;
2810 }
2811
2812 if (memcmp(first->anchor.guid, second->anchor.guid, DDF_GUID_LEN) != 0)
2813 return 2;
2814
2815 /* FIXME should I look at anything else? */
2816 return 0;
2817}
2818
0e600426 2819#ifndef MDASSEMBLE
4e5528c6
NB		 2820/*
2821 * A new array 'a' has been started which claims to be instance 'inst'
2822 * within container 'c'.
2823 * We need to confirm that the array matches the metadata in 'c' so
2824 * that we don't corrupt any metadata.
2825 */
cba0191b 2826static int ddf_open_new(struct supertype *c, struct active_array *a, char *inst)
549e9569 2827{
2c514b71 2828 dprintf("ddf: open_new %s\n", inst);
cba0191b 2829 a->info.container_member = atoi(inst);
549e9569
NB		 2830 return 0;
2831}
2832
4e5528c6
NB		 2833/*
2834 * The array 'a' is to be marked clean in the metadata.
ed9d66aa 2835 * If '->resync_start' is not ~(unsigned long long)0, then the array is only
4e5528c6
NB		 2836 * clean up to the point (in sectors). If that cannot be recorded in the
2837 * metadata, then leave it as dirty.
2838 *
2839 * For DDF, we need to clear the DDF_state_inconsistent bit in the
2840 * !global! virtual_disk.virtual_entry structure.
2841 */
01f157d7 2842static int ddf_set_array_state(struct active_array *a, int consistent)
549e9569 2843{
4e5528c6
NB		 2844 struct ddf_super *ddf = a->container->sb;
2845 int inst = a->info.container_member;
18a2f463 2846 int old = ddf->virt->entries[inst].state;
01f157d7
N		 2847 if (consistent == 2) {
2848 /* Should check if a recovery should be started FIXME */
2849 consistent = 1;
593add1b 2850 if (!is_resync_complete(a))
01f157d7
N		 2851 consistent = 0;
2852 }
ed9d66aa
NB		 2853 if (consistent)
2854 ddf->virt->entries[inst].state &= ~DDF_state_inconsistent;
2855 else
4e5528c6 2856 ddf->virt->entries[inst].state |= DDF_state_inconsistent;
18a2f463
NB		 2857 if (old != ddf->virt->entries[inst].state)
2858 ddf->updates_pending = 1;
2859
2860 old = ddf->virt->entries[inst].init_state;
ed9d66aa 2861 ddf->virt->entries[inst].init_state &= ~DDF_initstate_mask;
593add1b 2862 if (is_resync_complete(a))
ed9d66aa
NB		 2863 ddf->virt->entries[inst].init_state |= DDF_init_full;
2864 else if (a->resync_start == 0)
2865 ddf->virt->entries[inst].init_state |= DDF_init_not;
4e5528c6 2866 else
ed9d66aa 2867 ddf->virt->entries[inst].init_state |= DDF_init_quick;
18a2f463
NB		 2868 if (old != ddf->virt->entries[inst].init_state)
2869 ddf->updates_pending = 1;
ed9d66aa 2870
2c514b71
NB		 2871 dprintf("ddf mark %d %s %llu\n", inst, consistent?"clean":"dirty",
2872 a->resync_start);
01f157d7 2873 return consistent;
fd7cde1b
DW		 2874}
2875
7a7cc504
NB		 2876/*
2877 * The state of each disk is stored in the global phys_disk structure
2878 * in phys_disk.entries[n].state.
2879 * This makes various combinations awkward.
2880 * - When a device fails in any array, it must be failed in all arrays
2881 * that include a part of this device.
2882 * - When a component is rebuilding, we cannot include it officially in the
2883 * array unless this is the only array that uses the device.
2884 *
2885 * So: when transitioning:
2886 * Online -> failed, just set failed flag. monitor will propagate
2887 * spare -> online, the device might need to be added to the array.
2888 * spare -> failed, just set failed. Don't worry if in array or not.
2889 */
8d45d196 2890static void ddf_set_disk(struct active_array *a, int n, int state)
549e9569 2891{
7a7cc504
NB		 2892 struct ddf_super *ddf = a->container->sb;
2893 int inst = a->info.container_member;
2894 struct vd_config *vc = find_vdcr(ddf, inst);
2895 int pd = find_phys(ddf, vc->phys_refnum[n]);
2896 int i, st, working;
2897
2898 if (vc == NULL) {
2c514b71 2899 dprintf("ddf: cannot find instance %d!!\n", inst);
7a7cc504
NB		 2900 return;
2901 }
2902 if (pd < 0) {
2903 /* disk doesn't currently exist. If it is now in_sync,
2904 * insert it. */
2905 if ((state & DS_INSYNC) && ! (state & DS_FAULTY)) {
2906 /* Find dev 'n' in a->info->devs, determine the
2907 * ddf refnum, and set vc->phys_refnum and update
2908 * phys->entries[]
2909 */
2910 /* FIXME */
2911 }
2912 } else {
18a2f463 2913 int old = ddf->phys->entries[pd].state;
7a7cc504
NB		 2914 if (state & DS_FAULTY)
2915 ddf->phys->entries[pd].state |= __cpu_to_be16(DDF_Failed);
2916 if (state & DS_INSYNC) {
2917 ddf->phys->entries[pd].state |= __cpu_to_be16(DDF_Online);
2918 ddf->phys->entries[pd].state &= __cpu_to_be16(~DDF_Rebuilding);
2919 }
18a2f463
NB		 2920 if (old != ddf->phys->entries[pd].state)
2921 ddf->updates_pending = 1;
7a7cc504
NB		 2922 }
2923
2c514b71 2924 dprintf("ddf: set_disk %d to %x\n", n, state);
7e1432fb 2925
7a7cc504
NB		 2926 /* Now we need to check the state of the array and update
2927 * virtual_disk.entries[n].state.
2928 * It needs to be one of "optimal", "degraded", "failed".
2929 * I don't understand 'deleted' or 'missing'.
2930 */
2931 working = 0;
2932 for (i=0; i < a->info.array.raid_disks; i++) {
2933 pd = find_phys(ddf, vc->phys_refnum[i]);
2934 if (pd < 0)
2935 continue;
57632f4a
NB		 2936 st = __be16_to_cpu(ddf->phys->entries[pd].state);
2937 if ((st & (DDF_Online|DDF_Failed|DDF_Rebuilding))
7a7cc504
NB		 2938 == DDF_Online)
2939 working++;
2940 }
2941 state = DDF_state_degraded;
2942 if (working == a->info.array.raid_disks)
2943 state = DDF_state_optimal;
2944 else switch(vc->prl) {
2945 case DDF_RAID0:
2946 case DDF_CONCAT:
2947 case DDF_JBOD:
2948 state = DDF_state_failed;
2949 break;
2950 case DDF_RAID1:
2951 if (working == 0)
2952 state = DDF_state_failed;
2953 break;
2954 case DDF_RAID4:
2955 case DDF_RAID5:
2956 if (working < a->info.array.raid_disks-1)
2957 state = DDF_state_failed;
2958 break;
2959 case DDF_RAID6:
2960 if (working < a->info.array.raid_disks-2)
2961 state = DDF_state_failed;
2962 else if (working == a->info.array.raid_disks-1)
2963 state = DDF_state_part_optimal;
2964 break;
2965 }
2966
18a2f463
NB		 2967 if (ddf->virt->entries[inst].state !=
2968 ((ddf->virt->entries[inst].state & ~DDF_state_mask)
2969 | state)) {
2970
2971 ddf->virt->entries[inst].state =
2972 (ddf->virt->entries[inst].state & ~DDF_state_mask)
2973 | state;
2974 ddf->updates_pending = 1;
2975 }
7a7cc504 2976
549e9569
NB		 2977}
2978
2e735d19 2979static void ddf_sync_metadata(struct supertype *st)
549e9569 2980{
7a7cc504
NB		 2981
2982 /*
2983 * Write all data to all devices.
2984 * Later, we might be able to track whether only local changes
2985 * have been made, or whether any global data has been changed,
2986 * but ddf is sufficiently weird that it probably always
2987 * changes global data ....
2988 */
18a2f463
NB		 2989 struct ddf_super *ddf = st->sb;
2990 if (!ddf->updates_pending)
2991 return;
2992 ddf->updates_pending = 0;
2e735d19 2993 __write_init_super_ddf(st, 0);
2c514b71 2994 dprintf("ddf: sync_metadata\n");
549e9569
NB		 2995}
2996
88c164f4
NB		 2997static void ddf_process_update(struct supertype *st,
2998 struct metadata_update *update)
2999{
3000 /* Apply this update to the metadata.
3001 * The first 4 bytes are a DDF_*_MAGIC which guides
3002 * our actions.
3003 * Possible update are:
3004 * DDF_PHYS_RECORDS_MAGIC
3005 * Add a new physical device. Changes to this record
3006 * only happen implicitly.
3007 * used_pdes is the device number.
3008 * DDF_VIRT_RECORDS_MAGIC
3009 * Add a new VD. Possibly also change the 'access' bits.
3010 * populated_vdes is the entry number.
3011 * DDF_VD_CONF_MAGIC
3012 * New or updated VD. the VIRT_RECORD must already
3013 * exist. For an update, phys_refnum and lba_offset
3014 * (at least) are updated, and the VD_CONF must
3015 * be written to precisely those devices listed with
3016 * a phys_refnum.
3017 * DDF_SPARE_ASSIGN_MAGIC
3018 * replacement Spare Assignment Record... but for which device?
3019 *
3020 * So, e.g.:
3021 * - to create a new array, we send a VIRT_RECORD and
3022 * a VD_CONF. Then assemble and start the array.
3023 * - to activate a spare we send a VD_CONF to add the phys_refnum
3024 * and offset. This will also mark the spare as active with
3025 * a spare-assignment record.
3026 */
3027 struct ddf_super *ddf = st->sb;
3028 __u32 *magic = (__u32*)update->buf;
3029 struct phys_disk *pd;
3030 struct virtual_disk *vd;
3031 struct vd_config *vc;
3032 struct vcl *vcl;
3033 struct dl *dl;
3034 int mppe;
3035 int ent;
3036
2c514b71 3037 dprintf("Process update %x\n", *magic);
7e1432fb 3038
88c164f4
NB		 3039 switch (*magic) {
3040 case DDF_PHYS_RECORDS_MAGIC:
3041
3042 if (update->len != (sizeof(struct phys_disk) +
3043 sizeof(struct phys_disk_entry)))
3044 return;
3045 pd = (struct phys_disk*)update->buf;
3046
3047 ent = __be16_to_cpu(pd->used_pdes);
3048 if (ent >= __be16_to_cpu(ddf->phys->max_pdes))
3049 return;
3050 if (!all_ff(ddf->phys->entries[ent].guid))
3051 return;
3052 ddf->phys->entries[ent] = pd->entries[0];
3053 ddf->phys->used_pdes = __cpu_to_be16(1 +
3054 __be16_to_cpu(ddf->phys->used_pdes));
18a2f463 3055 ddf->updates_pending = 1;
2cc2983d
N		 3056 if (ddf->add_list) {
3057 struct active_array *a;
3058 struct dl *al = ddf->add_list;
3059 ddf->add_list = al->next;
3060
3061 al->next = ddf->dlist;
3062 ddf->dlist = al;
3063
3064 /* As a device has been added, we should check
3065 * for any degraded devices that might make
3066 * use of this spare */
3067 for (a = st->arrays ; a; a=a->next)
3068 a->check_degraded = 1;
3069 }
88c164f4
NB		 3070 break;
3071
3072 case DDF_VIRT_RECORDS_MAGIC:
3073
3074 if (update->len != (sizeof(struct virtual_disk) +
3075 sizeof(struct virtual_entry)))
3076 return;
3077 vd = (struct virtual_disk*)update->buf;
3078
3079 ent = __be16_to_cpu(vd->populated_vdes);
3080 if (ent >= __be16_to_cpu(ddf->virt->max_vdes))
3081 return;
3082 if (!all_ff(ddf->virt->entries[ent].guid))
3083 return;
3084 ddf->virt->entries[ent] = vd->entries[0];
3085 ddf->virt->populated_vdes = __cpu_to_be16(1 +
3086 __be16_to_cpu(ddf->virt->populated_vdes));
18a2f463 3087 ddf->updates_pending = 1;
88c164f4
NB		 3088 break;
3089
3090 case DDF_VD_CONF_MAGIC:
2c514b71 3091 dprintf("len %d %d\n", update->len, ddf->conf_rec_len);
88c164f4
NB		 3092
3093 mppe = __be16_to_cpu(ddf->anchor.max_primary_element_entries);
edd8d13c 3094 if (update->len != ddf->conf_rec_len * 512)
88c164f4
NB		 3095 return;
3096 vc = (struct vd_config*)update->buf;
3097 for (vcl = ddf->conflist; vcl ; vcl = vcl->next)
3098 if (memcmp(vcl->conf.guid, vc->guid, DDF_GUID_LEN) == 0)
3099 break;
2c514b71 3100 dprintf("vcl = %p\n", vcl);
88c164f4
NB		 3101 if (vcl) {
3102 /* An update, just copy the phys_refnum and lba_offset
3103 * fields
3104 */
3105 memcpy(vcl->conf.phys_refnum, vc->phys_refnum,
3106 mppe * (sizeof(__u32) + sizeof(__u64)));
3107 } else {
3108 /* A new VD_CONF */
e6b9548d
DW		 3109 if (!update->space)
3110 return;
88c164f4
NB		 3111 vcl = update->space;
3112 update->space = NULL;
3113 vcl->next = ddf->conflist;
edd8d13c 3114 memcpy(&vcl->conf, vc, update->len);
88c164f4
NB		 3115 vcl->lba_offset = (__u64*)
3116 &vcl->conf.phys_refnum[mppe];
3117 ddf->conflist = vcl;
3118 }
3119 /* Now make sure vlist is correct for each dl. */
3120 for (dl = ddf->dlist; dl; dl = dl->next) {
3121 int dn;
3122 int vn = 0;
3123 for (vcl = ddf->conflist; vcl ; vcl = vcl->next)
3124 for (dn=0; dn < ddf->mppe ; dn++)
3125 if (vcl->conf.phys_refnum[dn] ==
3126 dl->disk.refnum) {
2c514b71
NB		 3127 dprintf("dev %d has %p at %d\n",
3128 dl->pdnum, vcl, vn);
88c164f4
NB		 3129 dl->vlist[vn++] = vcl;
3130 break;
3131 }
3132 while (vn < ddf->max_part)
3133 dl->vlist[vn++] = NULL;
7e1432fb
NB		 3134 if (dl->vlist[0]) {
3135 ddf->phys->entries[dl->pdnum].type &=
3136 ~__cpu_to_be16(DDF_Global_Spare);
3137 ddf->phys->entries[dl->pdnum].type |=
3138 __cpu_to_be16(DDF_Active_in_VD);
3139 }
3140 if (dl->spare) {
3141 ddf->phys->entries[dl->pdnum].type &=
3142 ~__cpu_to_be16(DDF_Global_Spare);
3143 ddf->phys->entries[dl->pdnum].type |=
3144 __cpu_to_be16(DDF_Spare);
3145 }
3146 if (!dl->vlist[0] && !dl->spare) {
3147 ddf->phys->entries[dl->pdnum].type |=
3148 __cpu_to_be16(DDF_Global_Spare);
3149 ddf->phys->entries[dl->pdnum].type &=
3150 ~__cpu_to_be16(DDF_Spare |
3151 DDF_Active_in_VD);
3152 }
88c164f4 3153 }
18a2f463 3154 ddf->updates_pending = 1;
88c164f4
NB		 3155 break;
3156 case DDF_SPARE_ASSIGN_MAGIC:
3157 default: break;
3158 }
3159}
3160
edd8d13c
NB		 3161static void ddf_prepare_update(struct supertype *st,
3162 struct metadata_update *update)
3163{
3164 /* This update arrived at managemon.
3165 * We are about to pass it to monitor.
3166 * If a malloc is needed, do it here.
3167 */
3168 struct ddf_super *ddf = st->sb;
3169 __u32 *magic = (__u32*)update->buf;
3170 if (*magic == DDF_VD_CONF_MAGIC)
e6b9548d 3171 if (posix_memalign(&update->space, 512,
6416d527 3172 offsetof(struct vcl, conf)
e6b9548d
DW		 3173 + ddf->conf_rec_len * 512) != 0)
3174 update->space = NULL;
edd8d13c
NB		 3175}
3176
7e1432fb
NB		 3177/*
3178 * Check if the array 'a' is degraded but not failed.
3179 * If it is, find as many spares as are available and needed and
3180 * arrange for their inclusion.
3181 * We only choose devices which are not already in the array,
3182 * and prefer those with a spare-assignment to this array.
3183 * otherwise we choose global spares - assuming always that
3184 * there is enough room.
3185 * For each spare that we assign, we return an 'mdinfo' which
3186 * describes the position for the device in the array.
3187 * We also add to 'updates' a DDF_VD_CONF_MAGIC update with
3188 * the new phys_refnum and lba_offset values.
3189 *
3190 * Only worry about BVDs at the moment.
3191 */
3192static struct mdinfo *ddf_activate_spare(struct active_array *a,
3193 struct metadata_update **updates)
3194{
3195 int working = 0;
3196 struct mdinfo *d;
3197 struct ddf_super *ddf = a->container->sb;
3198 int global_ok = 0;
3199 struct mdinfo *rv = NULL;
3200 struct mdinfo *di;
3201 struct metadata_update *mu;
3202 struct dl *dl;
3203 int i;
3204 struct vd_config *vc;
3205 __u64 *lba;
3206
7e1432fb
NB		 3207 for (d = a->info.devs ; d ; d = d->next) {
3208 if ((d->curr_state & DS_FAULTY) &&
3209 d->state_fd >= 0)
3210 /* wait for Removal to happen */
3211 return NULL;
3212 if (d->state_fd >= 0)
3213 working ++;
3214 }
3215
2c514b71
NB		 3216 dprintf("ddf_activate: working=%d (%d) level=%d\n", working, a->info.array.raid_disks,
3217 a->info.array.level);
7e1432fb
NB		 3218 if (working == a->info.array.raid_disks)
3219 return NULL; /* array not degraded */
3220 switch (a->info.array.level) {
3221 case 1:
3222 if (working == 0)
3223 return NULL; /* failed */
3224 break;
3225 case 4:
3226 case 5:
3227 if (working < a->info.array.raid_disks - 1)
3228 return NULL; /* failed */
3229 break;
3230 case 6:
3231 if (working < a->info.array.raid_disks - 2)
3232 return NULL; /* failed */
3233 break;
3234 default: /* concat or stripe */
3235 return NULL; /* failed */
3236 }
3237
3238 /* For each slot, if it is not working, find a spare */
3239 dl = ddf->dlist;
3240 for (i = 0; i < a->info.array.raid_disks; i++) {
3241 for (d = a->info.devs ; d ; d = d->next)
3242 if (d->disk.raid_disk == i)
3243 break;
2c514b71 3244 dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0);
7e1432fb
NB		 3245 if (d && (d->state_fd >= 0))
3246 continue;
3247
3248 /* OK, this device needs recovery. Find a spare */
3249 again:
3250 for ( ; dl ; dl = dl->next) {
3251 unsigned long long esize;
3252 unsigned long long pos;
3253 struct mdinfo *d2;
3254 int is_global = 0;
3255 int is_dedicated = 0;
3256 struct extent *ex;
3257 int j;
3258 /* If in this array, skip */
3259 for (d2 = a->info.devs ; d2 ; d2 = d2->next)
3260 if (d2->disk.major == dl->major &&
3261 d2->disk.minor == dl->minor) {
2c514b71 3262 dprintf("%x:%x already in array\n", dl->major, dl->minor);
7e1432fb
NB		 3263 break;
3264 }
3265 if (d2)
3266 continue;
3267 if (ddf->phys->entries[dl->pdnum].type &
3268 __cpu_to_be16(DDF_Spare)) {
3269 /* Check spare assign record */
3270 if (dl->spare) {
3271 if (dl->spare->type & DDF_spare_dedicated) {
3272 /* check spare_ents for guid */
3273 for (j = 0 ;
3274 j < __be16_to_cpu(dl->spare->populated);
3275 j++) {
3276 if (memcmp(dl->spare->spare_ents[j].guid,
3277 ddf->virt->entries[a->info.container_member].guid,
3278 DDF_GUID_LEN) == 0)
3279 is_dedicated = 1;
3280 }
3281 } else
3282 is_global = 1;
3283 }
3284 } else if (ddf->phys->entries[dl->pdnum].type &
3285 __cpu_to_be16(DDF_Global_Spare)) {
3286 is_global = 1;
3287 }
3288 if ( ! (is_dedicated ||
3289 (is_global && global_ok))) {
2c514b71 3290 dprintf("%x:%x not suitable: %d %d\n", dl->major, dl->minor,
7e1432fb
NB		 3291 is_dedicated, is_global);
3292 continue;
3293 }
3294
3295 /* We are allowed to use this device - is there space?
3296 * We need a->info.component_size sectors */
3297 ex = get_extents(ddf, dl);
3298 if (!ex) {
2c514b71 3299 dprintf("cannot get extents\n");
7e1432fb
NB		 3300 continue;
3301 }
3302 j = 0; pos = 0;
3303 esize = 0;
3304
3305 do {
3306 esize = ex[j].start - pos;
3307 if (esize >= a->info.component_size)
3308 break;
3309 pos = ex[i].start + ex[i].size;
3310 i++;
3311 } while (ex[i-1].size);
3312
3313 free(ex);
3314 if (esize < a->info.component_size) {
2c514b71
NB		 3315 dprintf("%x:%x has no room: %llu %llu\n", dl->major, dl->minor,
3316 esize, a->info.component_size);
7e1432fb
NB		 3317 /* No room */
3318 continue;
3319 }
3320
3321 /* Cool, we have a device with some space at pos */
3322 di = malloc(sizeof(*di));
79244939
DW		 3323 if (!di)
3324 continue;
7e1432fb
NB		 3325 memset(di, 0, sizeof(*di));
3326 di->disk.number = i;
3327 di->disk.raid_disk = i;
3328 di->disk.major = dl->major;
3329 di->disk.minor = dl->minor;
3330 di->disk.state = 0;
3331 di->data_offset = pos;
3332 di->component_size = a->info.component_size;
3333 di->container_member = dl->pdnum;
3334 di->next = rv;
3335 rv = di;
2c514b71
NB		 3336 dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor,
3337 i, pos);
7e1432fb
NB		 3338
3339 break;
3340 }
3341 if (!dl && ! global_ok) {
3342 /* not enough dedicated spares, try global */
3343 global_ok = 1;
3344 dl = ddf->dlist;
3345 goto again;
3346 }
3347 }
3348
3349 if (!rv)
3350 /* No spares found */
3351 return rv;
3352 /* Now 'rv' has a list of devices to return.
3353 * Create a metadata_update record to update the
3354 * phys_refnum and lba_offset values
3355 */
904c1ef7 3356 mu = malloc(sizeof(*mu));
79244939
DW		 3357 if (mu && posix_memalign(&mu->space, 512, sizeof(struct vcl)) != 0) {
3358 free(mu);
3359 mu = NULL;
3360 }
3361 if (!mu) {
3362 while (rv) {
3363 struct mdinfo *n = rv->next;
3364
3365 free(rv);
3366 rv = n;
3367 }
3368 return NULL;
3369 }
3370
904c1ef7 3371 mu->buf = malloc(ddf->conf_rec_len * 512);
7e1432fb
NB		 3372 mu->len = ddf->conf_rec_len;
3373 mu->next = *updates;
3374 vc = find_vdcr(ddf, a->info.container_member);
3375 memcpy(mu->buf, vc, ddf->conf_rec_len * 512);
3376
3377 vc = (struct vd_config*)mu->buf;
3378 lba = (__u64*)&vc->phys_refnum[ddf->mppe];
3379 for (di = rv ; di ; di = di->next) {
3380 vc->phys_refnum[di->disk.raid_disk] =
3381 ddf->phys->entries[dl->pdnum].refnum;
3382 lba[di->disk.raid_disk] = di->data_offset;
3383 }
3384 *updates = mu;
3385 return rv;
3386}
0e600426 3387#endif /* MDASSEMBLE */
7e1432fb 3388
a322f70c
DW		 3389struct superswitch super_ddf = {
3390#ifndef MDASSEMBLE
3391 .examine_super = examine_super_ddf,
3392 .brief_examine_super = brief_examine_super_ddf,
3393 .detail_super = detail_super_ddf,
3394 .brief_detail_super = brief_detail_super_ddf,
3395 .validate_geometry = validate_geometry_ddf,
78e44928 3396 .write_init_super = write_init_super_ddf,
0e600426 3397 .add_to_super = add_to_super_ddf,
a322f70c
DW		 3398#endif
3399 .match_home = match_home_ddf,
3400 .uuid_from_super= uuid_from_super_ddf,
3401 .getinfo_super = getinfo_super_ddf,
3402 .update_super = update_super_ddf,
3403
3404 .avail_size = avail_size_ddf,
3405
a19c88b8
NB		 3406 .compare_super = compare_super_ddf,
3407
a322f70c 3408 .load_super = load_super_ddf,
ba7eb04f 3409 .init_super = init_super_ddf,
a322f70c
DW		 3410 .store_super = store_zero_ddf,
3411 .free_super = free_super_ddf,
3412 .match_metadata_desc = match_metadata_desc_ddf,
78e44928 3413 .container_content = container_content_ddf,
a322f70c 3414
a322f70c 3415 .external = 1,
549e9569 3416
0e600426 3417#ifndef MDASSEMBLE
549e9569
NB		 3418/* for mdmon */
3419 .open_new = ddf_open_new,
ed9d66aa 3420 .set_array_state= ddf_set_array_state,
549e9569
NB		 3421 .set_disk = ddf_set_disk,
3422 .sync_metadata = ddf_sync_metadata,
88c164f4 3423 .process_update = ddf_process_update,
edd8d13c 3424 .prepare_update = ddf_prepare_update,
7e1432fb 3425 .activate_spare = ddf_activate_spare,
0e600426 3426#endif
a322f70c 3427};
Unnamed repository; edit this file 'description' to name the repository.