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