]>
Commit | Line | Data |
---|---|---|
a322f70c DW |
1 | /* |
2 | * mdadm - manage Linux "md" devices aka RAID arrays. | |
3 | * | |
e736b623 | 4 | * Copyright (C) 2006-2009 Neil Brown <neilb@suse.de> |
a322f70c DW |
5 | * |
6 | * | |
7 | * This program is free software; you can redistribute it and/or modify | |
8 | * it under the terms of the GNU General Public License as published by | |
9 | * the Free Software Foundation; either version 2 of the License, or | |
10 | * (at your option) any later version. | |
11 | * | |
12 | * This program is distributed in the hope that it will be useful, | |
13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
15 | * GNU General Public License for more details. | |
16 | * | |
17 | * You should have received a copy of the GNU General Public License | |
18 | * along with this program; if not, write to the Free Software | |
19 | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA | |
20 | * | |
21 | * Author: Neil Brown | |
22 | * Email: <neil@brown.name> | |
23 | * | |
24 | * Specifications for DDF takes from Common RAID DDF Specification Revision 1.2 | |
25 | * (July 28 2006). Reused by permission of SNIA. | |
26 | */ | |
27 | ||
28 | #define HAVE_STDINT_H 1 | |
29 | #include "mdadm.h" | |
549e9569 | 30 | #include "mdmon.h" |
a322f70c DW |
31 | #include "sha1.h" |
32 | #include <values.h> | |
33 | ||
a322f70c DW |
34 | /* a non-official T10 name for creation GUIDs */ |
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 | ||
bedbf68a | 47 | #define DDF_NOTFOUND (~0U) |
48 | #define DDF_CONTAINER (DDF_NOTFOUND-1) | |
49 | ||
a322f70c DW |
50 | /* The DDF metadata handling. |
51 | * DDF metadata lives at the end of the device. | |
52 | * The last 512 byte block provides an 'anchor' which is used to locate | |
53 | * the rest of the metadata which usually lives immediately behind the anchor. | |
54 | * | |
55 | * Note: | |
56 | * - all multibyte numeric fields are bigendian. | |
57 | * - all strings are space padded. | |
58 | * | |
59 | */ | |
60 | ||
61 | /* Primary Raid Level (PRL) */ | |
62 | #define DDF_RAID0 0x00 | |
63 | #define DDF_RAID1 0x01 | |
64 | #define DDF_RAID3 0x03 | |
65 | #define DDF_RAID4 0x04 | |
66 | #define DDF_RAID5 0x05 | |
67 | #define DDF_RAID1E 0x11 | |
68 | #define DDF_JBOD 0x0f | |
69 | #define DDF_CONCAT 0x1f | |
70 | #define DDF_RAID5E 0x15 | |
71 | #define DDF_RAID5EE 0x25 | |
59e36268 | 72 | #define DDF_RAID6 0x06 |
a322f70c DW |
73 | |
74 | /* Raid Level Qualifier (RLQ) */ | |
75 | #define DDF_RAID0_SIMPLE 0x00 | |
76 | #define DDF_RAID1_SIMPLE 0x00 /* just 2 devices in this plex */ | |
77 | #define DDF_RAID1_MULTI 0x01 /* exactly 3 devices in this plex */ | |
78 | #define DDF_RAID3_0 0x00 /* parity in first extent */ | |
79 | #define DDF_RAID3_N 0x01 /* parity in last extent */ | |
80 | #define DDF_RAID4_0 0x00 /* parity in first extent */ | |
81 | #define DDF_RAID4_N 0x01 /* parity in last extent */ | |
82 | /* these apply to raid5e and raid5ee as well */ | |
83 | #define DDF_RAID5_0_RESTART 0x00 /* same as 'right asymmetric' - layout 1 */ | |
59e36268 | 84 | #define DDF_RAID6_0_RESTART 0x01 /* raid6 different from raid5 here!!! */ |
a322f70c DW |
85 | #define DDF_RAID5_N_RESTART 0x02 /* same as 'left asymmetric' - layout 0 */ |
86 | #define DDF_RAID5_N_CONTINUE 0x03 /* same as 'left symmetric' - layout 2 */ | |
87 | ||
88 | #define DDF_RAID1E_ADJACENT 0x00 /* raid10 nearcopies==2 */ | |
89 | #define DDF_RAID1E_OFFSET 0x01 /* raid10 offsetcopies==2 */ | |
90 | ||
91 | /* Secondary RAID Level (SRL) */ | |
92 | #define DDF_2STRIPED 0x00 /* This is weirder than RAID0 !! */ | |
93 | #define DDF_2MIRRORED 0x01 | |
94 | #define DDF_2CONCAT 0x02 | |
95 | #define DDF_2SPANNED 0x03 /* This is also weird - be careful */ | |
96 | ||
97 | /* Magic numbers */ | |
98 | #define DDF_HEADER_MAGIC __cpu_to_be32(0xDE11DE11) | |
99 | #define DDF_CONTROLLER_MAGIC __cpu_to_be32(0xAD111111) | |
100 | #define DDF_PHYS_RECORDS_MAGIC __cpu_to_be32(0x22222222) | |
101 | #define DDF_PHYS_DATA_MAGIC __cpu_to_be32(0x33333333) | |
102 | #define DDF_VIRT_RECORDS_MAGIC __cpu_to_be32(0xDDDDDDDD) | |
103 | #define DDF_VD_CONF_MAGIC __cpu_to_be32(0xEEEEEEEE) | |
104 | #define DDF_SPARE_ASSIGN_MAGIC __cpu_to_be32(0x55555555) | |
105 | #define DDF_VU_CONF_MAGIC __cpu_to_be32(0x88888888) | |
106 | #define DDF_VENDOR_LOG_MAGIC __cpu_to_be32(0x01dBEEF0) | |
107 | #define DDF_BBM_LOG_MAGIC __cpu_to_be32(0xABADB10C) | |
108 | ||
109 | #define DDF_GUID_LEN 24 | |
59e36268 NB |
110 | #define DDF_REVISION_0 "01.00.00" |
111 | #define DDF_REVISION_2 "01.02.00" | |
a322f70c DW |
112 | |
113 | struct ddf_header { | |
88c164f4 | 114 | __u32 magic; /* DDF_HEADER_MAGIC */ |
a322f70c DW |
115 | __u32 crc; |
116 | char guid[DDF_GUID_LEN]; | |
59e36268 | 117 | char revision[8]; /* 01.02.00 */ |
a322f70c DW |
118 | __u32 seq; /* starts at '1' */ |
119 | __u32 timestamp; | |
120 | __u8 openflag; | |
121 | __u8 foreignflag; | |
122 | __u8 enforcegroups; | |
123 | __u8 pad0; /* 0xff */ | |
124 | __u8 pad1[12]; /* 12 * 0xff */ | |
125 | /* 64 bytes so far */ | |
126 | __u8 header_ext[32]; /* reserved: fill with 0xff */ | |
127 | __u64 primary_lba; | |
128 | __u64 secondary_lba; | |
129 | __u8 type; | |
130 | __u8 pad2[3]; /* 0xff */ | |
131 | __u32 workspace_len; /* sectors for vendor space - | |
132 | * at least 32768(sectors) */ | |
133 | __u64 workspace_lba; | |
134 | __u16 max_pd_entries; /* one of 15, 63, 255, 1023, 4095 */ | |
135 | __u16 max_vd_entries; /* 2^(4,6,8,10,12)-1 : i.e. as above */ | |
136 | __u16 max_partitions; /* i.e. max num of configuration | |
137 | record entries per disk */ | |
138 | __u16 config_record_len; /* 1 +ROUNDUP(max_primary_element_entries | |
139 | *12/512) */ | |
140 | __u16 max_primary_element_entries; /* 16, 64, 256, 1024, or 4096 */ | |
141 | __u8 pad3[54]; /* 0xff */ | |
142 | /* 192 bytes so far */ | |
143 | __u32 controller_section_offset; | |
144 | __u32 controller_section_length; | |
145 | __u32 phys_section_offset; | |
146 | __u32 phys_section_length; | |
147 | __u32 virt_section_offset; | |
148 | __u32 virt_section_length; | |
149 | __u32 config_section_offset; | |
150 | __u32 config_section_length; | |
151 | __u32 data_section_offset; | |
152 | __u32 data_section_length; | |
153 | __u32 bbm_section_offset; | |
154 | __u32 bbm_section_length; | |
155 | __u32 diag_space_offset; | |
156 | __u32 diag_space_length; | |
157 | __u32 vendor_offset; | |
158 | __u32 vendor_length; | |
159 | /* 256 bytes so far */ | |
160 | __u8 pad4[256]; /* 0xff */ | |
161 | }; | |
162 | ||
163 | /* type field */ | |
164 | #define DDF_HEADER_ANCHOR 0x00 | |
165 | #define DDF_HEADER_PRIMARY 0x01 | |
166 | #define DDF_HEADER_SECONDARY 0x02 | |
167 | ||
168 | /* The content of the 'controller section' - global scope */ | |
169 | struct ddf_controller_data { | |
88c164f4 | 170 | __u32 magic; /* DDF_CONTROLLER_MAGIC */ |
a322f70c DW |
171 | __u32 crc; |
172 | char guid[DDF_GUID_LEN]; | |
173 | struct controller_type { | |
174 | __u16 vendor_id; | |
175 | __u16 device_id; | |
176 | __u16 sub_vendor_id; | |
177 | __u16 sub_device_id; | |
178 | } type; | |
179 | char product_id[16]; | |
180 | __u8 pad[8]; /* 0xff */ | |
181 | __u8 vendor_data[448]; | |
182 | }; | |
183 | ||
184 | /* The content of phys_section - global scope */ | |
185 | struct phys_disk { | |
88c164f4 | 186 | __u32 magic; /* DDF_PHYS_RECORDS_MAGIC */ |
a322f70c DW |
187 | __u32 crc; |
188 | __u16 used_pdes; | |
189 | __u16 max_pdes; | |
190 | __u8 pad[52]; | |
191 | struct phys_disk_entry { | |
192 | char guid[DDF_GUID_LEN]; | |
193 | __u32 refnum; | |
194 | __u16 type; | |
195 | __u16 state; | |
196 | __u64 config_size; /* DDF structures must be after here */ | |
197 | char path[18]; /* another horrible structure really */ | |
198 | __u8 pad[6]; | |
199 | } entries[0]; | |
200 | }; | |
201 | ||
202 | /* phys_disk_entry.type is a bitmap - bigendian remember */ | |
203 | #define DDF_Forced_PD_GUID 1 | |
204 | #define DDF_Active_in_VD 2 | |
88c164f4 | 205 | #define DDF_Global_Spare 4 /* VD_CONF records are ignored */ |
a322f70c DW |
206 | #define DDF_Spare 8 /* overrides Global_spare */ |
207 | #define DDF_Foreign 16 | |
208 | #define DDF_Legacy 32 /* no DDF on this device */ | |
209 | ||
210 | #define DDF_Interface_mask 0xf00 | |
211 | #define DDF_Interface_SCSI 0x100 | |
212 | #define DDF_Interface_SAS 0x200 | |
213 | #define DDF_Interface_SATA 0x300 | |
214 | #define DDF_Interface_FC 0x400 | |
215 | ||
216 | /* phys_disk_entry.state is a bigendian bitmap */ | |
217 | #define DDF_Online 1 | |
218 | #define DDF_Failed 2 /* overrides 1,4,8 */ | |
219 | #define DDF_Rebuilding 4 | |
220 | #define DDF_Transition 8 | |
221 | #define DDF_SMART 16 | |
222 | #define DDF_ReadErrors 32 | |
223 | #define DDF_Missing 64 | |
224 | ||
225 | /* The content of the virt_section global scope */ | |
226 | struct virtual_disk { | |
88c164f4 | 227 | __u32 magic; /* DDF_VIRT_RECORDS_MAGIC */ |
a322f70c DW |
228 | __u32 crc; |
229 | __u16 populated_vdes; | |
230 | __u16 max_vdes; | |
231 | __u8 pad[52]; | |
232 | struct virtual_entry { | |
233 | char guid[DDF_GUID_LEN]; | |
234 | __u16 unit; | |
235 | __u16 pad0; /* 0xffff */ | |
236 | __u16 guid_crc; | |
237 | __u16 type; | |
238 | __u8 state; | |
239 | __u8 init_state; | |
240 | __u8 pad1[14]; | |
241 | char name[16]; | |
242 | } entries[0]; | |
243 | }; | |
244 | ||
245 | /* virtual_entry.type is a bitmap - bigendian */ | |
246 | #define DDF_Shared 1 | |
247 | #define DDF_Enforce_Groups 2 | |
248 | #define DDF_Unicode 4 | |
249 | #define DDF_Owner_Valid 8 | |
250 | ||
251 | /* virtual_entry.state is a bigendian bitmap */ | |
252 | #define DDF_state_mask 0x7 | |
253 | #define DDF_state_optimal 0x0 | |
254 | #define DDF_state_degraded 0x1 | |
255 | #define DDF_state_deleted 0x2 | |
256 | #define DDF_state_missing 0x3 | |
257 | #define DDF_state_failed 0x4 | |
7a7cc504 | 258 | #define DDF_state_part_optimal 0x5 |
a322f70c DW |
259 | |
260 | #define DDF_state_morphing 0x8 | |
261 | #define DDF_state_inconsistent 0x10 | |
262 | ||
263 | /* virtual_entry.init_state is a bigendian bitmap */ | |
264 | #define DDF_initstate_mask 0x03 | |
265 | #define DDF_init_not 0x00 | |
7a7cc504 NB |
266 | #define DDF_init_quick 0x01 /* initialisation is progress. |
267 | * i.e. 'state_inconsistent' */ | |
a322f70c DW |
268 | #define DDF_init_full 0x02 |
269 | ||
270 | #define DDF_access_mask 0xc0 | |
271 | #define DDF_access_rw 0x00 | |
272 | #define DDF_access_ro 0x80 | |
273 | #define DDF_access_blocked 0xc0 | |
274 | ||
275 | /* The content of the config_section - local scope | |
276 | * It has multiple records each config_record_len sectors | |
277 | * They can be vd_config or spare_assign | |
278 | */ | |
279 | ||
280 | struct vd_config { | |
88c164f4 | 281 | __u32 magic; /* DDF_VD_CONF_MAGIC */ |
a322f70c DW |
282 | __u32 crc; |
283 | char guid[DDF_GUID_LEN]; | |
284 | __u32 timestamp; | |
285 | __u32 seqnum; | |
286 | __u8 pad0[24]; | |
287 | __u16 prim_elmnt_count; | |
288 | __u8 chunk_shift; /* 0 == 512, 1==1024 etc */ | |
289 | __u8 prl; | |
290 | __u8 rlq; | |
291 | __u8 sec_elmnt_count; | |
292 | __u8 sec_elmnt_seq; | |
293 | __u8 srl; | |
598f0d58 NB |
294 | __u64 blocks; /* blocks per component could be different |
295 | * on different component devices...(only | |
296 | * for concat I hope) */ | |
297 | __u64 array_blocks; /* blocks in array */ | |
a322f70c DW |
298 | __u8 pad1[8]; |
299 | __u32 spare_refs[8]; | |
300 | __u8 cache_pol[8]; | |
301 | __u8 bg_rate; | |
302 | __u8 pad2[3]; | |
303 | __u8 pad3[52]; | |
304 | __u8 pad4[192]; | |
305 | __u8 v0[32]; /* reserved- 0xff */ | |
306 | __u8 v1[32]; /* reserved- 0xff */ | |
307 | __u8 v2[16]; /* reserved- 0xff */ | |
308 | __u8 v3[16]; /* reserved- 0xff */ | |
309 | __u8 vendor[32]; | |
310 | __u32 phys_refnum[0]; /* refnum of each disk in sequence */ | |
311 | /*__u64 lba_offset[0]; LBA offset in each phys. Note extents in a | |
312 | bvd are always the same size */ | |
313 | }; | |
314 | ||
315 | /* vd_config.cache_pol[7] is a bitmap */ | |
316 | #define DDF_cache_writeback 1 /* else writethrough */ | |
317 | #define DDF_cache_wadaptive 2 /* only applies if writeback */ | |
318 | #define DDF_cache_readahead 4 | |
319 | #define DDF_cache_radaptive 8 /* only if doing read-ahead */ | |
320 | #define DDF_cache_ifnobatt 16 /* even to write cache if battery is poor */ | |
321 | #define DDF_cache_wallowed 32 /* enable write caching */ | |
322 | #define DDF_cache_rallowed 64 /* enable read caching */ | |
323 | ||
324 | struct spare_assign { | |
88c164f4 | 325 | __u32 magic; /* DDF_SPARE_ASSIGN_MAGIC */ |
a322f70c DW |
326 | __u32 crc; |
327 | __u32 timestamp; | |
328 | __u8 reserved[7]; | |
329 | __u8 type; | |
330 | __u16 populated; /* SAEs used */ | |
331 | __u16 max; /* max SAEs */ | |
332 | __u8 pad[8]; | |
333 | struct spare_assign_entry { | |
334 | char guid[DDF_GUID_LEN]; | |
335 | __u16 secondary_element; | |
336 | __u8 pad[6]; | |
337 | } spare_ents[0]; | |
338 | }; | |
339 | /* spare_assign.type is a bitmap */ | |
340 | #define DDF_spare_dedicated 0x1 /* else global */ | |
341 | #define DDF_spare_revertible 0x2 /* else committable */ | |
342 | #define DDF_spare_active 0x4 /* else not active */ | |
343 | #define DDF_spare_affinity 0x8 /* enclosure affinity */ | |
344 | ||
345 | /* The data_section contents - local scope */ | |
346 | struct disk_data { | |
88c164f4 | 347 | __u32 magic; /* DDF_PHYS_DATA_MAGIC */ |
a322f70c DW |
348 | __u32 crc; |
349 | char guid[DDF_GUID_LEN]; | |
350 | __u32 refnum; /* crc of some magic drive data ... */ | |
351 | __u8 forced_ref; /* set when above was not result of magic */ | |
352 | __u8 forced_guid; /* set if guid was forced rather than magic */ | |
353 | __u8 vendor[32]; | |
354 | __u8 pad[442]; | |
355 | }; | |
356 | ||
357 | /* bbm_section content */ | |
358 | struct bad_block_log { | |
359 | __u32 magic; | |
360 | __u32 crc; | |
361 | __u16 entry_count; | |
362 | __u32 spare_count; | |
363 | __u8 pad[10]; | |
364 | __u64 first_spare; | |
365 | struct mapped_block { | |
366 | __u64 defective_start; | |
367 | __u32 replacement_start; | |
368 | __u16 remap_count; | |
369 | __u8 pad[2]; | |
370 | } entries[0]; | |
371 | }; | |
372 | ||
373 | /* Struct for internally holding ddf structures */ | |
374 | /* The DDF structure stored on each device is potentially | |
375 | * quite different, as some data is global and some is local. | |
376 | * The global data is: | |
377 | * - ddf header | |
378 | * - controller_data | |
379 | * - Physical disk records | |
380 | * - Virtual disk records | |
381 | * The local data is: | |
382 | * - Configuration records | |
383 | * - Physical Disk data section | |
384 | * ( and Bad block and vendor which I don't care about yet). | |
385 | * | |
386 | * The local data is parsed into separate lists as it is read | |
387 | * and reconstructed for writing. This means that we only need | |
388 | * to make config changes once and they are automatically | |
389 | * propagated to all devices. | |
390 | * Note that the ddf_super has space of the conf and disk data | |
391 | * for this disk and also for a list of all such data. | |
392 | * The list is only used for the superblock that is being | |
393 | * built in Create or Assemble to describe the whole array. | |
394 | */ | |
395 | struct ddf_super { | |
6416d527 | 396 | struct ddf_header anchor, primary, secondary; |
a322f70c | 397 | struct ddf_controller_data controller; |
6416d527 | 398 | struct ddf_header *active; |
a322f70c DW |
399 | struct phys_disk *phys; |
400 | struct virtual_disk *virt; | |
401 | int pdsize, vdsize; | |
f21e18ca | 402 | unsigned int max_part, mppe, conf_rec_len; |
d2ca6449 | 403 | int currentdev; |
18a2f463 | 404 | int updates_pending; |
a322f70c | 405 | struct vcl { |
6416d527 NB |
406 | union { |
407 | char space[512]; | |
408 | struct { | |
409 | struct vcl *next; | |
410 | __u64 *lba_offset; /* location in 'conf' of | |
411 | * the lba table */ | |
f21e18ca | 412 | unsigned int vcnum; /* index into ->virt */ |
8ec5d685 | 413 | struct vd_config **other_bvds; |
6416d527 NB |
414 | __u64 *block_sizes; /* NULL if all the same */ |
415 | }; | |
416 | }; | |
a322f70c | 417 | struct vd_config conf; |
d2ca6449 | 418 | } *conflist, *currentconf; |
a322f70c | 419 | struct dl { |
6416d527 NB |
420 | union { |
421 | char space[512]; | |
422 | struct { | |
423 | struct dl *next; | |
424 | int major, minor; | |
425 | char *devname; | |
426 | int fd; | |
427 | unsigned long long size; /* sectors */ | |
097bcf00 | 428 | unsigned long long primary_lba; /* sectors */ |
429 | unsigned long long secondary_lba; /* sectors */ | |
430 | unsigned long long workspace_lba; /* sectors */ | |
6416d527 NB |
431 | int pdnum; /* index in ->phys */ |
432 | struct spare_assign *spare; | |
8592f29d N |
433 | void *mdupdate; /* hold metadata update */ |
434 | ||
435 | /* These fields used by auto-layout */ | |
436 | int raiddisk; /* slot to fill in autolayout */ | |
437 | __u64 esize; | |
6416d527 NB |
438 | }; |
439 | }; | |
a322f70c | 440 | struct disk_data disk; |
b2280677 | 441 | struct vcl *vlist[0]; /* max_part in size */ |
2cc2983d | 442 | } *dlist, *add_list; |
a322f70c DW |
443 | }; |
444 | ||
445 | #ifndef offsetof | |
446 | #define offsetof(t,f) ((size_t)&(((t*)0)->f)) | |
447 | #endif | |
448 | ||
f21e18ca | 449 | static unsigned int calc_crc(void *buf, int len) |
a322f70c DW |
450 | { |
451 | /* crcs are always at the same place as in the ddf_header */ | |
452 | struct ddf_header *ddf = buf; | |
453 | __u32 oldcrc = ddf->crc; | |
454 | __u32 newcrc; | |
455 | ddf->crc = 0xffffffff; | |
456 | ||
457 | newcrc = crc32(0, buf, len); | |
458 | ddf->crc = oldcrc; | |
4abe6b70 N |
459 | /* The crc is store (like everything) bigendian, so convert |
460 | * here for simplicity | |
461 | */ | |
462 | return __cpu_to_be32(newcrc); | |
a322f70c DW |
463 | } |
464 | ||
465 | static int load_ddf_header(int fd, unsigned long long lba, | |
466 | unsigned long long size, | |
467 | int type, | |
468 | struct ddf_header *hdr, struct ddf_header *anchor) | |
469 | { | |
470 | /* read a ddf header (primary or secondary) from fd/lba | |
471 | * and check that it is consistent with anchor | |
472 | * Need to check: | |
473 | * magic, crc, guid, rev, and LBA's header_type, and | |
474 | * everything after header_type must be the same | |
475 | */ | |
476 | if (lba >= size-1) | |
477 | return 0; | |
478 | ||
479 | if (lseek64(fd, lba<<9, 0) < 0) | |
480 | return 0; | |
481 | ||
482 | if (read(fd, hdr, 512) != 512) | |
483 | return 0; | |
484 | ||
485 | if (hdr->magic != DDF_HEADER_MAGIC) | |
486 | return 0; | |
487 | if (calc_crc(hdr, 512) != hdr->crc) | |
488 | return 0; | |
489 | if (memcmp(anchor->guid, hdr->guid, DDF_GUID_LEN) != 0 || | |
490 | memcmp(anchor->revision, hdr->revision, 8) != 0 || | |
491 | anchor->primary_lba != hdr->primary_lba || | |
492 | anchor->secondary_lba != hdr->secondary_lba || | |
493 | hdr->type != type || | |
494 | memcmp(anchor->pad2, hdr->pad2, 512 - | |
495 | offsetof(struct ddf_header, pad2)) != 0) | |
496 | return 0; | |
497 | ||
498 | /* Looks good enough to me... */ | |
499 | return 1; | |
500 | } | |
501 | ||
502 | static void *load_section(int fd, struct ddf_super *super, void *buf, | |
503 | __u32 offset_be, __u32 len_be, int check) | |
504 | { | |
505 | unsigned long long offset = __be32_to_cpu(offset_be); | |
506 | unsigned long long len = __be32_to_cpu(len_be); | |
507 | int dofree = (buf == NULL); | |
508 | ||
509 | if (check) | |
510 | if (len != 2 && len != 8 && len != 32 | |
511 | && len != 128 && len != 512) | |
512 | return NULL; | |
513 | ||
514 | if (len > 1024) | |
515 | return NULL; | |
516 | if (buf) { | |
517 | /* All pre-allocated sections are a single block */ | |
518 | if (len != 1) | |
519 | return NULL; | |
3d2c4fc7 DW |
520 | } else if (posix_memalign(&buf, 512, len<<9) != 0) |
521 | buf = NULL; | |
6416d527 | 522 | |
a322f70c DW |
523 | if (!buf) |
524 | return NULL; | |
525 | ||
526 | if (super->active->type == 1) | |
527 | offset += __be64_to_cpu(super->active->primary_lba); | |
528 | else | |
529 | offset += __be64_to_cpu(super->active->secondary_lba); | |
530 | ||
f21e18ca | 531 | if ((unsigned long long)lseek64(fd, offset<<9, 0) != (offset<<9)) { |
a322f70c DW |
532 | if (dofree) |
533 | free(buf); | |
534 | return NULL; | |
535 | } | |
f21e18ca | 536 | if ((unsigned long long)read(fd, buf, len<<9) != (len<<9)) { |
a322f70c DW |
537 | if (dofree) |
538 | free(buf); | |
539 | return NULL; | |
540 | } | |
541 | return buf; | |
542 | } | |
543 | ||
544 | static int load_ddf_headers(int fd, struct ddf_super *super, char *devname) | |
545 | { | |
546 | unsigned long long dsize; | |
547 | ||
548 | get_dev_size(fd, NULL, &dsize); | |
549 | ||
550 | if (lseek64(fd, dsize-512, 0) < 0) { | |
551 | if (devname) | |
e7b84f9d N |
552 | pr_err("Cannot seek to anchor block on %s: %s\n", |
553 | devname, strerror(errno)); | |
a322f70c DW |
554 | return 1; |
555 | } | |
556 | if (read(fd, &super->anchor, 512) != 512) { | |
557 | if (devname) | |
e7b84f9d N |
558 | pr_err("Cannot read anchor block on %s: %s\n", |
559 | devname, strerror(errno)); | |
a322f70c DW |
560 | return 1; |
561 | } | |
562 | if (super->anchor.magic != DDF_HEADER_MAGIC) { | |
563 | if (devname) | |
e7b84f9d | 564 | pr_err("no DDF anchor found on %s\n", |
a322f70c DW |
565 | devname); |
566 | return 2; | |
567 | } | |
568 | if (calc_crc(&super->anchor, 512) != super->anchor.crc) { | |
569 | if (devname) | |
e7b84f9d | 570 | pr_err("bad CRC on anchor on %s\n", |
a322f70c DW |
571 | devname); |
572 | return 2; | |
573 | } | |
59e36268 NB |
574 | if (memcmp(super->anchor.revision, DDF_REVISION_0, 8) != 0 && |
575 | memcmp(super->anchor.revision, DDF_REVISION_2, 8) != 0) { | |
a322f70c | 576 | if (devname) |
e7b84f9d | 577 | pr_err("can only support super revision" |
59e36268 NB |
578 | " %.8s and earlier, not %.8s on %s\n", |
579 | DDF_REVISION_2, super->anchor.revision,devname); | |
a322f70c DW |
580 | return 2; |
581 | } | |
582 | if (load_ddf_header(fd, __be64_to_cpu(super->anchor.primary_lba), | |
583 | dsize >> 9, 1, | |
584 | &super->primary, &super->anchor) == 0) { | |
585 | if (devname) | |
e7b84f9d N |
586 | pr_err("Failed to load primary DDF header " |
587 | "on %s\n", devname); | |
a322f70c DW |
588 | return 2; |
589 | } | |
590 | super->active = &super->primary; | |
591 | if (load_ddf_header(fd, __be64_to_cpu(super->anchor.secondary_lba), | |
592 | dsize >> 9, 2, | |
593 | &super->secondary, &super->anchor)) { | |
594 | if ((__be32_to_cpu(super->primary.seq) | |
595 | < __be32_to_cpu(super->secondary.seq) && | |
596 | !super->secondary.openflag) | |
597 | || (__be32_to_cpu(super->primary.seq) | |
598 | == __be32_to_cpu(super->secondary.seq) && | |
599 | super->primary.openflag && !super->secondary.openflag) | |
600 | ) | |
601 | super->active = &super->secondary; | |
602 | } | |
603 | return 0; | |
604 | } | |
605 | ||
606 | static int load_ddf_global(int fd, struct ddf_super *super, char *devname) | |
607 | { | |
608 | void *ok; | |
609 | ok = load_section(fd, super, &super->controller, | |
610 | super->active->controller_section_offset, | |
611 | super->active->controller_section_length, | |
612 | 0); | |
613 | super->phys = load_section(fd, super, NULL, | |
614 | super->active->phys_section_offset, | |
615 | super->active->phys_section_length, | |
616 | 1); | |
617 | super->pdsize = __be32_to_cpu(super->active->phys_section_length) * 512; | |
618 | ||
619 | super->virt = load_section(fd, super, NULL, | |
620 | super->active->virt_section_offset, | |
621 | super->active->virt_section_length, | |
622 | 1); | |
623 | super->vdsize = __be32_to_cpu(super->active->virt_section_length) * 512; | |
624 | if (!ok || | |
625 | !super->phys || | |
626 | !super->virt) { | |
627 | free(super->phys); | |
628 | free(super->virt); | |
a2349791 NB |
629 | super->phys = NULL; |
630 | super->virt = NULL; | |
a322f70c DW |
631 | return 2; |
632 | } | |
633 | super->conflist = NULL; | |
634 | super->dlist = NULL; | |
8c3b8c2c NB |
635 | |
636 | super->max_part = __be16_to_cpu(super->active->max_partitions); | |
637 | super->mppe = __be16_to_cpu(super->active->max_primary_element_entries); | |
638 | super->conf_rec_len = __be16_to_cpu(super->active->config_record_len); | |
a322f70c DW |
639 | return 0; |
640 | } | |
641 | ||
3dc821b0 | 642 | static void add_other_bvd(struct vcl *vcl, struct vd_config *vd, |
643 | unsigned int len) | |
644 | { | |
645 | int i; | |
646 | for (i = 0; i < vcl->conf.sec_elmnt_count-1; i++) | |
647 | if (vcl->other_bvds[i] != NULL && | |
648 | vcl->other_bvds[i]->sec_elmnt_seq == vd->sec_elmnt_seq) | |
649 | break; | |
650 | ||
651 | if (i < vcl->conf.sec_elmnt_count-1) { | |
652 | if (vd->seqnum <= vcl->other_bvds[i]->seqnum) | |
653 | return; | |
654 | } else { | |
655 | for (i = 0; i < vcl->conf.sec_elmnt_count-1; i++) | |
656 | if (vcl->other_bvds[i] == NULL) | |
657 | break; | |
658 | if (i == vcl->conf.sec_elmnt_count-1) { | |
659 | pr_err("no space for sec level config %u, count is %u\n", | |
660 | vd->sec_elmnt_seq, vcl->conf.sec_elmnt_count); | |
661 | return; | |
662 | } | |
663 | if (posix_memalign((void **)&vcl->other_bvds[i], 512, len) | |
664 | != 0) { | |
665 | pr_err("%s could not allocate vd buf\n", __func__); | |
666 | return; | |
667 | } | |
668 | } | |
669 | memcpy(vcl->other_bvds[i], vd, len); | |
670 | } | |
671 | ||
a322f70c DW |
672 | static int load_ddf_local(int fd, struct ddf_super *super, |
673 | char *devname, int keep) | |
674 | { | |
675 | struct dl *dl; | |
676 | struct stat stb; | |
677 | char *conf; | |
f21e18ca N |
678 | unsigned int i; |
679 | unsigned int confsec; | |
b2280677 | 680 | int vnum; |
f21e18ca | 681 | unsigned int max_virt_disks = __be16_to_cpu(super->active->max_vd_entries); |
d2ca6449 | 682 | unsigned long long dsize; |
a322f70c DW |
683 | |
684 | /* First the local disk info */ | |
3d2c4fc7 | 685 | if (posix_memalign((void**)&dl, 512, |
6416d527 | 686 | sizeof(*dl) + |
3d2c4fc7 | 687 | (super->max_part) * sizeof(dl->vlist[0])) != 0) { |
e7b84f9d | 688 | pr_err("%s could not allocate disk info buffer\n", |
3d2c4fc7 DW |
689 | __func__); |
690 | return 1; | |
691 | } | |
a322f70c DW |
692 | |
693 | load_section(fd, super, &dl->disk, | |
694 | super->active->data_section_offset, | |
695 | super->active->data_section_length, | |
696 | 0); | |
503975b9 | 697 | dl->devname = devname ? xstrdup(devname) : NULL; |
598f0d58 | 698 | |
a322f70c DW |
699 | fstat(fd, &stb); |
700 | dl->major = major(stb.st_rdev); | |
701 | dl->minor = minor(stb.st_rdev); | |
702 | dl->next = super->dlist; | |
703 | dl->fd = keep ? fd : -1; | |
d2ca6449 NB |
704 | |
705 | dl->size = 0; | |
706 | if (get_dev_size(fd, devname, &dsize)) | |
707 | dl->size = dsize >> 9; | |
097bcf00 | 708 | /* If the disks have different sizes, the LBAs will differ |
709 | * between phys disks. | |
710 | * At this point here, the values in super->active must be valid | |
711 | * for this phys disk. */ | |
712 | dl->primary_lba = super->active->primary_lba; | |
713 | dl->secondary_lba = super->active->secondary_lba; | |
714 | dl->workspace_lba = super->active->workspace_lba; | |
b2280677 | 715 | dl->spare = NULL; |
f21e18ca | 716 | for (i = 0 ; i < super->max_part ; i++) |
a322f70c DW |
717 | dl->vlist[i] = NULL; |
718 | super->dlist = dl; | |
59e36268 | 719 | dl->pdnum = -1; |
f21e18ca | 720 | for (i = 0; i < __be16_to_cpu(super->active->max_pd_entries); i++) |
5575e7d9 NB |
721 | if (memcmp(super->phys->entries[i].guid, |
722 | dl->disk.guid, DDF_GUID_LEN) == 0) | |
723 | dl->pdnum = i; | |
724 | ||
a322f70c DW |
725 | /* Now the config list. */ |
726 | /* 'conf' is an array of config entries, some of which are | |
727 | * probably invalid. Those which are good need to be copied into | |
728 | * the conflist | |
729 | */ | |
a322f70c DW |
730 | |
731 | conf = load_section(fd, super, NULL, | |
732 | super->active->config_section_offset, | |
733 | super->active->config_section_length, | |
734 | 0); | |
735 | ||
b2280677 | 736 | vnum = 0; |
e223334f N |
737 | for (confsec = 0; |
738 | confsec < __be32_to_cpu(super->active->config_section_length); | |
739 | confsec += super->conf_rec_len) { | |
a322f70c | 740 | struct vd_config *vd = |
e223334f | 741 | (struct vd_config *)((char*)conf + confsec*512); |
a322f70c DW |
742 | struct vcl *vcl; |
743 | ||
b2280677 NB |
744 | if (vd->magic == DDF_SPARE_ASSIGN_MAGIC) { |
745 | if (dl->spare) | |
746 | continue; | |
3d2c4fc7 DW |
747 | if (posix_memalign((void**)&dl->spare, 512, |
748 | super->conf_rec_len*512) != 0) { | |
e7b84f9d N |
749 | pr_err("%s could not allocate spare info buf\n", |
750 | __func__); | |
3d2c4fc7 DW |
751 | return 1; |
752 | } | |
613b0d17 | 753 | |
b2280677 NB |
754 | memcpy(dl->spare, vd, super->conf_rec_len*512); |
755 | continue; | |
756 | } | |
a322f70c DW |
757 | if (vd->magic != DDF_VD_CONF_MAGIC) |
758 | continue; | |
759 | for (vcl = super->conflist; vcl; vcl = vcl->next) { | |
760 | if (memcmp(vcl->conf.guid, | |
761 | vd->guid, DDF_GUID_LEN) == 0) | |
762 | break; | |
763 | } | |
764 | ||
765 | if (vcl) { | |
b2280677 | 766 | dl->vlist[vnum++] = vcl; |
3dc821b0 | 767 | if (vcl->other_bvds != NULL && |
768 | vcl->conf.sec_elmnt_seq != vd->sec_elmnt_seq) { | |
769 | add_other_bvd(vcl, vd, super->conf_rec_len*512); | |
770 | continue; | |
771 | } | |
a322f70c DW |
772 | if (__be32_to_cpu(vd->seqnum) <= |
773 | __be32_to_cpu(vcl->conf.seqnum)) | |
774 | continue; | |
59e36268 | 775 | } else { |
3d2c4fc7 | 776 | if (posix_memalign((void**)&vcl, 512, |
6416d527 | 777 | (super->conf_rec_len*512 + |
3d2c4fc7 | 778 | offsetof(struct vcl, conf))) != 0) { |
e7b84f9d N |
779 | pr_err("%s could not allocate vcl buf\n", |
780 | __func__); | |
3d2c4fc7 DW |
781 | return 1; |
782 | } | |
a322f70c | 783 | vcl->next = super->conflist; |
59e36268 | 784 | vcl->block_sizes = NULL; /* FIXME not for CONCAT */ |
8ec5d685 | 785 | if (vd->sec_elmnt_count > 1) |
786 | vcl->other_bvds = | |
787 | xcalloc(vd->sec_elmnt_count - 1, | |
788 | sizeof(struct vd_config *)); | |
789 | else | |
790 | vcl->other_bvds = NULL; | |
a322f70c | 791 | super->conflist = vcl; |
b2280677 | 792 | dl->vlist[vnum++] = vcl; |
a322f70c | 793 | } |
8c3b8c2c | 794 | memcpy(&vcl->conf, vd, super->conf_rec_len*512); |
a322f70c | 795 | vcl->lba_offset = (__u64*) |
8c3b8c2c | 796 | &vcl->conf.phys_refnum[super->mppe]; |
59e36268 NB |
797 | |
798 | for (i=0; i < max_virt_disks ; i++) | |
799 | if (memcmp(super->virt->entries[i].guid, | |
800 | vcl->conf.guid, DDF_GUID_LEN)==0) | |
801 | break; | |
802 | if (i < max_virt_disks) | |
803 | vcl->vcnum = i; | |
a322f70c DW |
804 | } |
805 | free(conf); | |
806 | ||
807 | return 0; | |
808 | } | |
809 | ||
810 | #ifndef MDASSEMBLE | |
811 | static int load_super_ddf_all(struct supertype *st, int fd, | |
e1902a7b | 812 | void **sbp, char *devname); |
a322f70c | 813 | #endif |
37424f13 DW |
814 | |
815 | static void free_super_ddf(struct supertype *st); | |
816 | ||
a322f70c DW |
817 | static int load_super_ddf(struct supertype *st, int fd, |
818 | char *devname) | |
819 | { | |
820 | unsigned long long dsize; | |
821 | struct ddf_super *super; | |
822 | int rv; | |
823 | ||
a322f70c DW |
824 | if (get_dev_size(fd, devname, &dsize) == 0) |
825 | return 1; | |
826 | ||
691c6ee1 N |
827 | if (test_partition(fd)) |
828 | /* DDF is not allowed on partitions */ | |
829 | return 1; | |
830 | ||
a322f70c DW |
831 | /* 32M is a lower bound */ |
832 | if (dsize <= 32*1024*1024) { | |
97320d7c | 833 | if (devname) |
e7b84f9d N |
834 | pr_err("%s is too small for ddf: " |
835 | "size is %llu sectors.\n", | |
836 | devname, dsize>>9); | |
97320d7c | 837 | return 1; |
a322f70c DW |
838 | } |
839 | if (dsize & 511) { | |
97320d7c | 840 | if (devname) |
e7b84f9d N |
841 | pr_err("%s is an odd size for ddf: " |
842 | "size is %llu bytes.\n", | |
843 | devname, dsize); | |
97320d7c | 844 | return 1; |
a322f70c DW |
845 | } |
846 | ||
37424f13 DW |
847 | free_super_ddf(st); |
848 | ||
6416d527 | 849 | if (posix_memalign((void**)&super, 512, sizeof(*super))!= 0) { |
e7b84f9d | 850 | pr_err("malloc of %zu failed.\n", |
a322f70c DW |
851 | sizeof(*super)); |
852 | return 1; | |
853 | } | |
a2349791 | 854 | memset(super, 0, sizeof(*super)); |
a322f70c DW |
855 | |
856 | rv = load_ddf_headers(fd, super, devname); | |
857 | if (rv) { | |
858 | free(super); | |
859 | return rv; | |
860 | } | |
861 | ||
862 | /* Have valid headers and have chosen the best. Let's read in the rest*/ | |
863 | ||
864 | rv = load_ddf_global(fd, super, devname); | |
865 | ||
866 | if (rv) { | |
867 | if (devname) | |
e7b84f9d N |
868 | pr_err("Failed to load all information " |
869 | "sections on %s\n", devname); | |
a322f70c DW |
870 | free(super); |
871 | return rv; | |
872 | } | |
873 | ||
3d2c4fc7 DW |
874 | rv = load_ddf_local(fd, super, devname, 0); |
875 | ||
876 | if (rv) { | |
877 | if (devname) | |
e7b84f9d N |
878 | pr_err("Failed to load all information " |
879 | "sections on %s\n", devname); | |
3d2c4fc7 DW |
880 | free(super); |
881 | return rv; | |
882 | } | |
a322f70c DW |
883 | |
884 | /* Should possibly check the sections .... */ | |
885 | ||
886 | st->sb = super; | |
887 | if (st->ss == NULL) { | |
888 | st->ss = &super_ddf; | |
889 | st->minor_version = 0; | |
890 | st->max_devs = 512; | |
891 | } | |
892 | return 0; | |
893 | ||
894 | } | |
895 | ||
896 | static void free_super_ddf(struct supertype *st) | |
897 | { | |
898 | struct ddf_super *ddf = st->sb; | |
899 | if (ddf == NULL) | |
900 | return; | |
901 | free(ddf->phys); | |
902 | free(ddf->virt); | |
903 | while (ddf->conflist) { | |
904 | struct vcl *v = ddf->conflist; | |
905 | ddf->conflist = v->next; | |
59e36268 NB |
906 | if (v->block_sizes) |
907 | free(v->block_sizes); | |
3dc821b0 | 908 | if (v->other_bvds) { |
909 | int i; | |
910 | for (i = 0; i < v->conf.sec_elmnt_count-1; i++) | |
911 | if (v->other_bvds[i] != NULL) | |
912 | free(v->other_bvds[i]); | |
8ec5d685 | 913 | free(v->other_bvds); |
3dc821b0 | 914 | } |
a322f70c DW |
915 | free(v); |
916 | } | |
917 | while (ddf->dlist) { | |
918 | struct dl *d = ddf->dlist; | |
919 | ddf->dlist = d->next; | |
920 | if (d->fd >= 0) | |
921 | close(d->fd); | |
b2280677 NB |
922 | if (d->spare) |
923 | free(d->spare); | |
a322f70c DW |
924 | free(d); |
925 | } | |
8a38cb04 N |
926 | while (ddf->add_list) { |
927 | struct dl *d = ddf->add_list; | |
928 | ddf->add_list = d->next; | |
929 | if (d->fd >= 0) | |
930 | close(d->fd); | |
931 | if (d->spare) | |
932 | free(d->spare); | |
933 | free(d); | |
934 | } | |
a322f70c DW |
935 | free(ddf); |
936 | st->sb = NULL; | |
937 | } | |
938 | ||
939 | static struct supertype *match_metadata_desc_ddf(char *arg) | |
940 | { | |
941 | /* 'ddf' only support containers */ | |
942 | struct supertype *st; | |
943 | if (strcmp(arg, "ddf") != 0 && | |
944 | strcmp(arg, "default") != 0 | |
945 | ) | |
946 | return NULL; | |
947 | ||
503975b9 | 948 | st = xcalloc(1, sizeof(*st)); |
a322f70c DW |
949 | st->ss = &super_ddf; |
950 | st->max_devs = 512; | |
951 | st->minor_version = 0; | |
952 | st->sb = NULL; | |
953 | return st; | |
954 | } | |
955 | ||
a322f70c DW |
956 | #ifndef MDASSEMBLE |
957 | ||
958 | static mapping_t ddf_state[] = { | |
959 | { "Optimal", 0}, | |
960 | { "Degraded", 1}, | |
961 | { "Deleted", 2}, | |
962 | { "Missing", 3}, | |
963 | { "Failed", 4}, | |
964 | { "Partially Optimal", 5}, | |
965 | { "-reserved-", 6}, | |
966 | { "-reserved-", 7}, | |
967 | { NULL, 0} | |
968 | }; | |
969 | ||
970 | static mapping_t ddf_init_state[] = { | |
971 | { "Not Initialised", 0}, | |
972 | { "QuickInit in Progress", 1}, | |
973 | { "Fully Initialised", 2}, | |
974 | { "*UNKNOWN*", 3}, | |
975 | { NULL, 0} | |
976 | }; | |
977 | static mapping_t ddf_access[] = { | |
978 | { "Read/Write", 0}, | |
979 | { "Reserved", 1}, | |
980 | { "Read Only", 2}, | |
981 | { "Blocked (no access)", 3}, | |
982 | { NULL ,0} | |
983 | }; | |
984 | ||
985 | static mapping_t ddf_level[] = { | |
986 | { "RAID0", DDF_RAID0}, | |
987 | { "RAID1", DDF_RAID1}, | |
988 | { "RAID3", DDF_RAID3}, | |
989 | { "RAID4", DDF_RAID4}, | |
990 | { "RAID5", DDF_RAID5}, | |
991 | { "RAID1E",DDF_RAID1E}, | |
992 | { "JBOD", DDF_JBOD}, | |
993 | { "CONCAT",DDF_CONCAT}, | |
994 | { "RAID5E",DDF_RAID5E}, | |
995 | { "RAID5EE",DDF_RAID5EE}, | |
996 | { "RAID6", DDF_RAID6}, | |
997 | { NULL, 0} | |
998 | }; | |
999 | static mapping_t ddf_sec_level[] = { | |
1000 | { "Striped", DDF_2STRIPED}, | |
1001 | { "Mirrored", DDF_2MIRRORED}, | |
1002 | { "Concat", DDF_2CONCAT}, | |
1003 | { "Spanned", DDF_2SPANNED}, | |
1004 | { NULL, 0} | |
1005 | }; | |
1006 | #endif | |
1007 | ||
1008 | struct num_mapping { | |
1009 | int num1, num2; | |
1010 | }; | |
1011 | static struct num_mapping ddf_level_num[] = { | |
1012 | { DDF_RAID0, 0 }, | |
1013 | { DDF_RAID1, 1 }, | |
1014 | { DDF_RAID3, LEVEL_UNSUPPORTED }, | |
60f18132 NB |
1015 | { DDF_RAID4, 4 }, |
1016 | { DDF_RAID5, 5 }, | |
a322f70c DW |
1017 | { DDF_RAID1E, LEVEL_UNSUPPORTED }, |
1018 | { DDF_JBOD, LEVEL_UNSUPPORTED }, | |
1019 | { DDF_CONCAT, LEVEL_LINEAR }, | |
1020 | { DDF_RAID5E, LEVEL_UNSUPPORTED }, | |
1021 | { DDF_RAID5EE, LEVEL_UNSUPPORTED }, | |
1022 | { DDF_RAID6, 6}, | |
1023 | { MAXINT, MAXINT } | |
1024 | }; | |
1025 | ||
1026 | static int map_num1(struct num_mapping *map, int num) | |
1027 | { | |
1028 | int i; | |
1029 | for (i=0 ; map[i].num1 != MAXINT; i++) | |
1030 | if (map[i].num1 == num) | |
1031 | break; | |
1032 | return map[i].num2; | |
1033 | } | |
1034 | ||
42dc2744 N |
1035 | static int all_ff(char *guid) |
1036 | { | |
1037 | int i; | |
1038 | for (i = 0; i < DDF_GUID_LEN; i++) | |
1039 | if (guid[i] != (char)0xff) | |
1040 | return 0; | |
1041 | return 1; | |
1042 | } | |
1043 | ||
a322f70c DW |
1044 | #ifndef MDASSEMBLE |
1045 | static void print_guid(char *guid, int tstamp) | |
1046 | { | |
1047 | /* A GUIDs are part (or all) ASCII and part binary. | |
1048 | * They tend to be space padded. | |
59e36268 NB |
1049 | * We print the GUID in HEX, then in parentheses add |
1050 | * any initial ASCII sequence, and a possible | |
1051 | * time stamp from bytes 16-19 | |
a322f70c DW |
1052 | */ |
1053 | int l = DDF_GUID_LEN; | |
1054 | int i; | |
59e36268 NB |
1055 | |
1056 | for (i=0 ; i<DDF_GUID_LEN ; i++) { | |
1057 | if ((i&3)==0 && i != 0) printf(":"); | |
1058 | printf("%02X", guid[i]&255); | |
1059 | } | |
1060 | ||
cfccea8c | 1061 | printf("\n ("); |
a322f70c DW |
1062 | while (l && guid[l-1] == ' ') |
1063 | l--; | |
1064 | for (i=0 ; i<l ; i++) { | |
1065 | if (guid[i] >= 0x20 && guid[i] < 0x7f) | |
1066 | fputc(guid[i], stdout); | |
1067 | else | |
59e36268 | 1068 | break; |
a322f70c DW |
1069 | } |
1070 | if (tstamp) { | |
1071 | time_t then = __be32_to_cpu(*(__u32*)(guid+16)) + DECADE; | |
1072 | char tbuf[100]; | |
1073 | struct tm *tm; | |
1074 | tm = localtime(&then); | |
59e36268 | 1075 | strftime(tbuf, 100, " %D %T",tm); |
a322f70c DW |
1076 | fputs(tbuf, stdout); |
1077 | } | |
59e36268 | 1078 | printf(")"); |
a322f70c DW |
1079 | } |
1080 | ||
1081 | static void examine_vd(int n, struct ddf_super *sb, char *guid) | |
1082 | { | |
8c3b8c2c | 1083 | int crl = sb->conf_rec_len; |
a322f70c DW |
1084 | struct vcl *vcl; |
1085 | ||
1086 | for (vcl = sb->conflist ; vcl ; vcl = vcl->next) { | |
f21e18ca | 1087 | unsigned int i; |
a322f70c DW |
1088 | struct vd_config *vc = &vcl->conf; |
1089 | ||
1090 | if (calc_crc(vc, crl*512) != vc->crc) | |
1091 | continue; | |
1092 | if (memcmp(vc->guid, guid, DDF_GUID_LEN) != 0) | |
1093 | continue; | |
1094 | ||
1095 | /* Ok, we know about this VD, let's give more details */ | |
b06e3095 | 1096 | printf(" Raid Devices[%d] : %d (", n, |
a322f70c | 1097 | __be16_to_cpu(vc->prim_elmnt_count)); |
f21e18ca | 1098 | for (i = 0; i < __be16_to_cpu(vc->prim_elmnt_count); i++) { |
b06e3095 N |
1099 | int j; |
1100 | int cnt = __be16_to_cpu(sb->phys->used_pdes); | |
1101 | for (j=0; j<cnt; j++) | |
1102 | if (vc->phys_refnum[i] == sb->phys->entries[j].refnum) | |
1103 | break; | |
1104 | if (i) printf(" "); | |
1105 | if (j < cnt) | |
1106 | printf("%d", j); | |
1107 | else | |
1108 | printf("--"); | |
1109 | } | |
1110 | printf(")\n"); | |
1111 | if (vc->chunk_shift != 255) | |
613b0d17 N |
1112 | printf(" Chunk Size[%d] : %d sectors\n", n, |
1113 | 1 << vc->chunk_shift); | |
a322f70c DW |
1114 | printf(" Raid Level[%d] : %s\n", n, |
1115 | map_num(ddf_level, vc->prl)?:"-unknown-"); | |
1116 | if (vc->sec_elmnt_count != 1) { | |
1117 | printf(" Secondary Position[%d] : %d of %d\n", n, | |
1118 | vc->sec_elmnt_seq, vc->sec_elmnt_count); | |
1119 | printf(" Secondary Level[%d] : %s\n", n, | |
1120 | map_num(ddf_sec_level, vc->srl) ?: "-unknown-"); | |
1121 | } | |
1122 | printf(" Device Size[%d] : %llu\n", n, | |
c9b6907b | 1123 | (unsigned long long)__be64_to_cpu(vc->blocks)/2); |
a322f70c | 1124 | printf(" Array Size[%d] : %llu\n", n, |
c9b6907b | 1125 | (unsigned long long)__be64_to_cpu(vc->array_blocks)/2); |
a322f70c DW |
1126 | } |
1127 | } | |
1128 | ||
1129 | static void examine_vds(struct ddf_super *sb) | |
1130 | { | |
1131 | int cnt = __be16_to_cpu(sb->virt->populated_vdes); | |
1132 | int i; | |
1133 | printf(" Virtual Disks : %d\n", cnt); | |
1134 | ||
1135 | for (i=0; i<cnt; i++) { | |
1136 | struct virtual_entry *ve = &sb->virt->entries[i]; | |
b06e3095 | 1137 | printf("\n"); |
a322f70c DW |
1138 | printf(" VD GUID[%d] : ", i); print_guid(ve->guid, 1); |
1139 | printf("\n"); | |
1140 | printf(" unit[%d] : %d\n", i, __be16_to_cpu(ve->unit)); | |
1141 | printf(" state[%d] : %s, %s%s\n", i, | |
1142 | map_num(ddf_state, ve->state & 7), | |
1143 | (ve->state & 8) ? "Morphing, ": "", | |
1144 | (ve->state & 16)? "Not Consistent" : "Consistent"); | |
1145 | printf(" init state[%d] : %s\n", i, | |
1146 | map_num(ddf_init_state, ve->init_state&3)); | |
1147 | printf(" access[%d] : %s\n", i, | |
1148 | map_num(ddf_access, (ve->init_state>>6) & 3)); | |
1149 | printf(" Name[%d] : %.16s\n", i, ve->name); | |
1150 | examine_vd(i, sb, ve->guid); | |
1151 | } | |
1152 | if (cnt) printf("\n"); | |
1153 | } | |
1154 | ||
1155 | static void examine_pds(struct ddf_super *sb) | |
1156 | { | |
1157 | int cnt = __be16_to_cpu(sb->phys->used_pdes); | |
1158 | int i; | |
1159 | struct dl *dl; | |
1160 | printf(" Physical Disks : %d\n", cnt); | |
962371a5 | 1161 | printf(" Number RefNo Size Device Type/State\n"); |
a322f70c DW |
1162 | |
1163 | for (i=0 ; i<cnt ; i++) { | |
1164 | struct phys_disk_entry *pd = &sb->phys->entries[i]; | |
1165 | int type = __be16_to_cpu(pd->type); | |
1166 | int state = __be16_to_cpu(pd->state); | |
1167 | ||
b06e3095 N |
1168 | //printf(" PD GUID[%d] : ", i); print_guid(pd->guid, 0); |
1169 | //printf("\n"); | |
1170 | printf(" %3d %08x ", i, | |
a322f70c | 1171 | __be32_to_cpu(pd->refnum)); |
613b0d17 | 1172 | printf("%8lluK ", |
c9b6907b | 1173 | (unsigned long long)__be64_to_cpu(pd->config_size)>>1); |
b06e3095 N |
1174 | for (dl = sb->dlist; dl ; dl = dl->next) { |
1175 | if (dl->disk.refnum == pd->refnum) { | |
1176 | char *dv = map_dev(dl->major, dl->minor, 0); | |
1177 | if (dv) { | |
962371a5 | 1178 | printf("%-15s", dv); |
b06e3095 N |
1179 | break; |
1180 | } | |
1181 | } | |
1182 | } | |
1183 | if (!dl) | |
962371a5 | 1184 | printf("%15s",""); |
b06e3095 | 1185 | printf(" %s%s%s%s%s", |
a322f70c | 1186 | (type&2) ? "active":"", |
b06e3095 | 1187 | (type&4) ? "Global-Spare":"", |
a322f70c DW |
1188 | (type&8) ? "spare" : "", |
1189 | (type&16)? ", foreign" : "", | |
1190 | (type&32)? "pass-through" : ""); | |
18cb4496 N |
1191 | if (state & DDF_Failed) |
1192 | /* This over-rides these three */ | |
1193 | state &= ~(DDF_Online|DDF_Rebuilding|DDF_Transition); | |
b06e3095 | 1194 | printf("/%s%s%s%s%s%s%s", |
a322f70c DW |
1195 | (state&1)? "Online": "Offline", |
1196 | (state&2)? ", Failed": "", | |
1197 | (state&4)? ", Rebuilding": "", | |
1198 | (state&8)? ", in-transition": "", | |
b06e3095 N |
1199 | (state&16)? ", SMART-errors": "", |
1200 | (state&32)? ", Unrecovered-Read-Errors": "", | |
a322f70c | 1201 | (state&64)? ", Missing" : ""); |
a322f70c DW |
1202 | printf("\n"); |
1203 | } | |
1204 | } | |
1205 | ||
1206 | static void examine_super_ddf(struct supertype *st, char *homehost) | |
1207 | { | |
1208 | struct ddf_super *sb = st->sb; | |
1209 | ||
1210 | printf(" Magic : %08x\n", __be32_to_cpu(sb->anchor.magic)); | |
1211 | printf(" Version : %.8s\n", sb->anchor.revision); | |
598f0d58 NB |
1212 | printf("Controller GUID : "); print_guid(sb->controller.guid, 0); |
1213 | printf("\n"); | |
1214 | printf(" Container GUID : "); print_guid(sb->anchor.guid, 1); | |
a322f70c DW |
1215 | printf("\n"); |
1216 | printf(" Seq : %08x\n", __be32_to_cpu(sb->active->seq)); | |
1217 | printf(" Redundant hdr : %s\n", sb->secondary.magic == DDF_HEADER_MAGIC | |
1218 | ?"yes" : "no"); | |
1219 | examine_vds(sb); | |
1220 | examine_pds(sb); | |
1221 | } | |
1222 | ||
a5d85af7 | 1223 | static void getinfo_super_ddf(struct supertype *st, struct mdinfo *info, char *map); |
ff54de6e | 1224 | |
bedbf68a | 1225 | static void uuid_from_ddf_guid(const char *guid, int uuid[4]); |
42dc2744 | 1226 | static void uuid_from_super_ddf(struct supertype *st, int uuid[4]); |
ff54de6e | 1227 | |
bedbf68a | 1228 | static unsigned int get_vd_num_of_subarray(struct supertype *st) |
1229 | { | |
1230 | /* | |
1231 | * Figure out the VD number for this supertype. | |
1232 | * Returns DDF_CONTAINER for the container itself, | |
1233 | * and DDF_NOTFOUND on error. | |
1234 | */ | |
1235 | struct ddf_super *ddf = st->sb; | |
1236 | struct mdinfo *sra; | |
1237 | char *sub, *end; | |
1238 | unsigned int vcnum; | |
1239 | ||
1240 | if (*st->container_devnm == '\0') | |
1241 | return DDF_CONTAINER; | |
1242 | ||
1243 | sra = sysfs_read(-1, st->devnm, GET_VERSION); | |
1244 | if (!sra || sra->array.major_version != -1 || | |
1245 | sra->array.minor_version != -2 || | |
1246 | !is_subarray(sra->text_version)) | |
1247 | return DDF_NOTFOUND; | |
1248 | ||
1249 | sub = strchr(sra->text_version + 1, '/'); | |
1250 | if (sub != NULL) | |
1251 | vcnum = strtoul(sub + 1, &end, 10); | |
1252 | if (sub == NULL || *sub == '\0' || *end != '\0' || | |
1253 | vcnum >= __be16_to_cpu(ddf->active->max_vd_entries)) | |
1254 | return DDF_NOTFOUND; | |
1255 | ||
1256 | return vcnum; | |
1257 | } | |
1258 | ||
061f2c6a | 1259 | static void brief_examine_super_ddf(struct supertype *st, int verbose) |
4737ae25 N |
1260 | { |
1261 | /* We just write a generic DDF ARRAY entry | |
1262 | */ | |
1263 | struct mdinfo info; | |
1264 | char nbuf[64]; | |
a5d85af7 | 1265 | getinfo_super_ddf(st, &info, NULL); |
4737ae25 N |
1266 | fname_from_uuid(st, &info, nbuf, ':'); |
1267 | ||
1268 | printf("ARRAY metadata=ddf UUID=%s\n", nbuf + 5); | |
1269 | } | |
1270 | ||
1271 | static void brief_examine_subarrays_ddf(struct supertype *st, int verbose) | |
a322f70c DW |
1272 | { |
1273 | /* We just write a generic DDF ARRAY entry | |
a322f70c | 1274 | */ |
42dc2744 | 1275 | struct ddf_super *ddf = st->sb; |
ff54de6e | 1276 | struct mdinfo info; |
f21e18ca | 1277 | unsigned int i; |
ff54de6e | 1278 | char nbuf[64]; |
a5d85af7 | 1279 | getinfo_super_ddf(st, &info, NULL); |
ff54de6e | 1280 | fname_from_uuid(st, &info, nbuf, ':'); |
42dc2744 | 1281 | |
f21e18ca | 1282 | for (i = 0; i < __be16_to_cpu(ddf->virt->max_vdes); i++) { |
42dc2744 N |
1283 | struct virtual_entry *ve = &ddf->virt->entries[i]; |
1284 | struct vcl vcl; | |
1285 | char nbuf1[64]; | |
1286 | if (all_ff(ve->guid)) | |
1287 | continue; | |
1288 | memcpy(vcl.conf.guid, ve->guid, DDF_GUID_LEN); | |
1289 | ddf->currentconf =&vcl; | |
1290 | uuid_from_super_ddf(st, info.uuid); | |
1291 | fname_from_uuid(st, &info, nbuf1, ':'); | |
1292 | printf("ARRAY container=%s member=%d UUID=%s\n", | |
1293 | nbuf+5, i, nbuf1+5); | |
1294 | } | |
a322f70c DW |
1295 | } |
1296 | ||
bceedeec N |
1297 | static void export_examine_super_ddf(struct supertype *st) |
1298 | { | |
1299 | struct mdinfo info; | |
1300 | char nbuf[64]; | |
a5d85af7 | 1301 | getinfo_super_ddf(st, &info, NULL); |
bceedeec N |
1302 | fname_from_uuid(st, &info, nbuf, ':'); |
1303 | printf("MD_METADATA=ddf\n"); | |
1304 | printf("MD_LEVEL=container\n"); | |
1305 | printf("MD_UUID=%s\n", nbuf+5); | |
1306 | } | |
bceedeec | 1307 | |
a322f70c DW |
1308 | static void detail_super_ddf(struct supertype *st, char *homehost) |
1309 | { | |
1310 | /* FIXME later | |
1311 | * Could print DDF GUID | |
1312 | * Need to find which array | |
1313 | * If whole, briefly list all arrays | |
1314 | * If one, give name | |
1315 | */ | |
1316 | } | |
1317 | ||
1318 | static void brief_detail_super_ddf(struct supertype *st) | |
1319 | { | |
ff54de6e N |
1320 | struct mdinfo info; |
1321 | char nbuf[64]; | |
bedbf68a | 1322 | struct ddf_super *ddf = st->sb; |
1323 | unsigned int vcnum = get_vd_num_of_subarray(st); | |
1324 | if (vcnum == DDF_CONTAINER) | |
1325 | uuid_from_super_ddf(st, info.uuid); | |
1326 | else if (vcnum == DDF_NOTFOUND) | |
1327 | return; | |
1328 | else | |
1329 | uuid_from_ddf_guid(ddf->virt->entries[vcnum].guid, info.uuid); | |
ff54de6e N |
1330 | fname_from_uuid(st, &info, nbuf,':'); |
1331 | printf(" UUID=%s", nbuf + 5); | |
a322f70c | 1332 | } |
a322f70c DW |
1333 | #endif |
1334 | ||
1335 | static int match_home_ddf(struct supertype *st, char *homehost) | |
1336 | { | |
1337 | /* It matches 'this' host if the controller is a | |
1338 | * Linux-MD controller with vendor_data matching | |
1339 | * the hostname | |
1340 | */ | |
1341 | struct ddf_super *ddf = st->sb; | |
f21e18ca | 1342 | unsigned int len; |
d1d3482b N |
1343 | |
1344 | if (!homehost) | |
1345 | return 0; | |
1346 | len = strlen(homehost); | |
a322f70c DW |
1347 | |
1348 | return (memcmp(ddf->controller.guid, T10, 8) == 0 && | |
1349 | len < sizeof(ddf->controller.vendor_data) && | |
1350 | memcmp(ddf->controller.vendor_data, homehost,len) == 0 && | |
1351 | ddf->controller.vendor_data[len] == 0); | |
1352 | } | |
1353 | ||
0e600426 | 1354 | #ifndef MDASSEMBLE |
f21e18ca | 1355 | static struct vd_config *find_vdcr(struct ddf_super *ddf, unsigned int inst) |
a322f70c | 1356 | { |
7a7cc504 | 1357 | struct vcl *v; |
59e36268 | 1358 | |
7a7cc504 | 1359 | for (v = ddf->conflist; v; v = v->next) |
59e36268 | 1360 | if (inst == v->vcnum) |
7a7cc504 NB |
1361 | return &v->conf; |
1362 | return NULL; | |
1363 | } | |
0e600426 | 1364 | #endif |
7a7cc504 NB |
1365 | |
1366 | static int find_phys(struct ddf_super *ddf, __u32 phys_refnum) | |
1367 | { | |
1368 | /* Find the entry in phys_disk which has the given refnum | |
1369 | * and return it's index | |
1370 | */ | |
f21e18ca N |
1371 | unsigned int i; |
1372 | for (i = 0; i < __be16_to_cpu(ddf->phys->max_pdes); i++) | |
7a7cc504 NB |
1373 | if (ddf->phys->entries[i].refnum == phys_refnum) |
1374 | return i; | |
1375 | return -1; | |
a322f70c DW |
1376 | } |
1377 | ||
bedbf68a | 1378 | static void uuid_from_ddf_guid(const char *guid, int uuid[4]) |
1379 | { | |
1380 | char buf[20]; | |
1381 | struct sha1_ctx ctx; | |
1382 | sha1_init_ctx(&ctx); | |
1383 | sha1_process_bytes(guid, DDF_GUID_LEN, &ctx); | |
1384 | sha1_finish_ctx(&ctx, buf); | |
1385 | memcpy(uuid, buf, 4*4); | |
1386 | } | |
1387 | ||
a322f70c DW |
1388 | static void uuid_from_super_ddf(struct supertype *st, int uuid[4]) |
1389 | { | |
1390 | /* The uuid returned here is used for: | |
1391 | * uuid to put into bitmap file (Create, Grow) | |
1392 | * uuid for backup header when saving critical section (Grow) | |
1393 | * comparing uuids when re-adding a device into an array | |
51006d85 N |
1394 | * In these cases the uuid required is that of the data-array, |
1395 | * not the device-set. | |
1396 | * uuid to recognise same set when adding a missing device back | |
1397 | * to an array. This is a uuid for the device-set. | |
613b0d17 | 1398 | * |
a322f70c DW |
1399 | * For each of these we can make do with a truncated |
1400 | * or hashed uuid rather than the original, as long as | |
1401 | * everyone agrees. | |
a322f70c DW |
1402 | * In the case of SVD we assume the BVD is of interest, |
1403 | * though that might be the case if a bitmap were made for | |
1404 | * a mirrored SVD - worry about that later. | |
1405 | * So we need to find the VD configuration record for the | |
1406 | * relevant BVD and extract the GUID and Secondary_Element_Seq. | |
1407 | * The first 16 bytes of the sha1 of these is used. | |
1408 | */ | |
1409 | struct ddf_super *ddf = st->sb; | |
d2ca6449 | 1410 | struct vcl *vcl = ddf->currentconf; |
c5afc314 | 1411 | char *guid; |
a322f70c | 1412 | |
c5afc314 N |
1413 | if (vcl) |
1414 | guid = vcl->conf.guid; | |
1415 | else | |
1416 | guid = ddf->anchor.guid; | |
bedbf68a | 1417 | uuid_from_ddf_guid(guid, uuid); |
a322f70c DW |
1418 | } |
1419 | ||
a5d85af7 | 1420 | static void getinfo_super_ddf_bvd(struct supertype *st, struct mdinfo *info, char *map); |
78e44928 | 1421 | |
a5d85af7 | 1422 | static void getinfo_super_ddf(struct supertype *st, struct mdinfo *info, char *map) |
a322f70c DW |
1423 | { |
1424 | struct ddf_super *ddf = st->sb; | |
a5d85af7 | 1425 | int map_disks = info->array.raid_disks; |
90fa1a29 | 1426 | __u32 *cptr; |
a322f70c | 1427 | |
78e44928 | 1428 | if (ddf->currentconf) { |
a5d85af7 | 1429 | getinfo_super_ddf_bvd(st, info, map); |
78e44928 NB |
1430 | return; |
1431 | } | |
95eeceeb | 1432 | memset(info, 0, sizeof(*info)); |
78e44928 | 1433 | |
a322f70c DW |
1434 | info->array.raid_disks = __be16_to_cpu(ddf->phys->used_pdes); |
1435 | info->array.level = LEVEL_CONTAINER; | |
1436 | info->array.layout = 0; | |
1437 | info->array.md_minor = -1; | |
90fa1a29 JS |
1438 | cptr = (__u32 *)(ddf->anchor.guid + 16); |
1439 | info->array.ctime = DECADE + __be32_to_cpu(*cptr); | |
1440 | ||
a322f70c DW |
1441 | info->array.utime = 0; |
1442 | info->array.chunk_size = 0; | |
510242aa | 1443 | info->container_enough = 1; |
a322f70c | 1444 | |
a322f70c DW |
1445 | info->disk.major = 0; |
1446 | info->disk.minor = 0; | |
cba0191b NB |
1447 | if (ddf->dlist) { |
1448 | info->disk.number = __be32_to_cpu(ddf->dlist->disk.refnum); | |
59e36268 | 1449 | info->disk.raid_disk = find_phys(ddf, ddf->dlist->disk.refnum); |
d2ca6449 NB |
1450 | |
1451 | info->data_offset = __be64_to_cpu(ddf->phys-> | |
613b0d17 N |
1452 | entries[info->disk.raid_disk]. |
1453 | config_size); | |
d2ca6449 | 1454 | info->component_size = ddf->dlist->size - info->data_offset; |
cba0191b NB |
1455 | } else { |
1456 | info->disk.number = -1; | |
661dce36 | 1457 | info->disk.raid_disk = -1; |
cba0191b NB |
1458 | // info->disk.raid_disk = find refnum in the table and use index; |
1459 | } | |
f22385f9 | 1460 | info->disk.state = (1 << MD_DISK_SYNC) | (1 << MD_DISK_ACTIVE); |
a19c88b8 | 1461 | |
921d9e16 | 1462 | info->recovery_start = MaxSector; |
a19c88b8 | 1463 | info->reshape_active = 0; |
6e75048b | 1464 | info->recovery_blocked = 0; |
c5afc314 | 1465 | info->name[0] = 0; |
a322f70c | 1466 | |
f35f2525 N |
1467 | info->array.major_version = -1; |
1468 | info->array.minor_version = -2; | |
159c3a1a | 1469 | strcpy(info->text_version, "ddf"); |
a67dd8cc | 1470 | info->safe_mode_delay = 0; |
159c3a1a | 1471 | |
c5afc314 | 1472 | uuid_from_super_ddf(st, info->uuid); |
a322f70c | 1473 | |
a5d85af7 N |
1474 | if (map) { |
1475 | int i; | |
1476 | for (i = 0 ; i < map_disks; i++) { | |
1477 | if (i < info->array.raid_disks && | |
1478 | (__be16_to_cpu(ddf->phys->entries[i].state) & DDF_Online) && | |
1479 | !(__be16_to_cpu(ddf->phys->entries[i].state) & DDF_Failed)) | |
1480 | map[i] = 1; | |
1481 | else | |
1482 | map[i] = 0; | |
1483 | } | |
1484 | } | |
a322f70c DW |
1485 | } |
1486 | ||
598f0d58 NB |
1487 | static int rlq_to_layout(int rlq, int prl, int raiddisks); |
1488 | ||
a5d85af7 | 1489 | static void getinfo_super_ddf_bvd(struct supertype *st, struct mdinfo *info, char *map) |
a322f70c DW |
1490 | { |
1491 | struct ddf_super *ddf = st->sb; | |
d2ca6449 NB |
1492 | struct vcl *vc = ddf->currentconf; |
1493 | int cd = ddf->currentdev; | |
db42fa9b | 1494 | int j; |
8592f29d | 1495 | struct dl *dl; |
a5d85af7 | 1496 | int map_disks = info->array.raid_disks; |
90fa1a29 | 1497 | __u32 *cptr; |
a322f70c | 1498 | |
95eeceeb | 1499 | memset(info, 0, sizeof(*info)); |
a322f70c DW |
1500 | /* FIXME this returns BVD info - what if we want SVD ?? */ |
1501 | ||
d2ca6449 NB |
1502 | info->array.raid_disks = __be16_to_cpu(vc->conf.prim_elmnt_count); |
1503 | info->array.level = map_num1(ddf_level_num, vc->conf.prl); | |
1504 | info->array.layout = rlq_to_layout(vc->conf.rlq, vc->conf.prl, | |
598f0d58 | 1505 | info->array.raid_disks); |
a322f70c | 1506 | info->array.md_minor = -1; |
90fa1a29 JS |
1507 | cptr = (__u32 *)(vc->conf.guid + 16); |
1508 | info->array.ctime = DECADE + __be32_to_cpu(*cptr); | |
d2ca6449 NB |
1509 | info->array.utime = DECADE + __be32_to_cpu(vc->conf.timestamp); |
1510 | info->array.chunk_size = 512 << vc->conf.chunk_shift; | |
da9b4a62 | 1511 | info->custom_array_size = 0; |
d2ca6449 | 1512 | |
f21e18ca | 1513 | if (cd >= 0 && (unsigned)cd < ddf->mppe) { |
d2ca6449 NB |
1514 | info->data_offset = __be64_to_cpu(vc->lba_offset[cd]); |
1515 | if (vc->block_sizes) | |
1516 | info->component_size = vc->block_sizes[cd]; | |
1517 | else | |
1518 | info->component_size = __be64_to_cpu(vc->conf.blocks); | |
1519 | } | |
a322f70c | 1520 | |
fb204fb2 N |
1521 | for (dl = ddf->dlist; dl ; dl = dl->next) |
1522 | if (dl->raiddisk == ddf->currentdev) | |
1523 | break; | |
1524 | ||
a322f70c DW |
1525 | info->disk.major = 0; |
1526 | info->disk.minor = 0; | |
fb204fb2 | 1527 | info->disk.state = 0; |
8592f29d N |
1528 | if (dl) { |
1529 | info->disk.major = dl->major; | |
1530 | info->disk.minor = dl->minor; | |
fb204fb2 N |
1531 | info->disk.raid_disk = dl->raiddisk; |
1532 | info->disk.number = dl->pdnum; | |
1533 | info->disk.state = (1<<MD_DISK_SYNC)|(1<<MD_DISK_ACTIVE); | |
8592f29d | 1534 | } |
a322f70c | 1535 | |
103f2410 NB |
1536 | info->container_member = ddf->currentconf->vcnum; |
1537 | ||
921d9e16 | 1538 | info->recovery_start = MaxSector; |
80d26cb2 | 1539 | info->resync_start = 0; |
624c5ad4 | 1540 | info->reshape_active = 0; |
6e75048b | 1541 | info->recovery_blocked = 0; |
80d26cb2 NB |
1542 | if (!(ddf->virt->entries[info->container_member].state |
1543 | & DDF_state_inconsistent) && | |
1544 | (ddf->virt->entries[info->container_member].init_state | |
1545 | & DDF_initstate_mask) | |
1546 | == DDF_init_full) | |
b7528a20 | 1547 | info->resync_start = MaxSector; |
80d26cb2 | 1548 | |
a322f70c DW |
1549 | uuid_from_super_ddf(st, info->uuid); |
1550 | ||
f35f2525 N |
1551 | info->array.major_version = -1; |
1552 | info->array.minor_version = -2; | |
9b63e648 | 1553 | sprintf(info->text_version, "/%s/%d", |
4dd2df09 | 1554 | st->container_devnm, |
9b63e648 | 1555 | info->container_member); |
a67dd8cc | 1556 | info->safe_mode_delay = 200; |
159c3a1a | 1557 | |
db42fa9b N |
1558 | memcpy(info->name, ddf->virt->entries[info->container_member].name, 16); |
1559 | info->name[16]=0; | |
1560 | for(j=0; j<16; j++) | |
1561 | if (info->name[j] == ' ') | |
1562 | info->name[j] = 0; | |
a5d85af7 N |
1563 | |
1564 | if (map) | |
1565 | for (j = 0; j < map_disks; j++) { | |
1566 | map[j] = 0; | |
1567 | if (j < info->array.raid_disks) { | |
1568 | int i = find_phys(ddf, vc->conf.phys_refnum[j]); | |
613b0d17 | 1569 | if (i >= 0 && |
a5d85af7 N |
1570 | (__be16_to_cpu(ddf->phys->entries[i].state) & DDF_Online) && |
1571 | !(__be16_to_cpu(ddf->phys->entries[i].state) & DDF_Failed)) | |
1572 | map[i] = 1; | |
1573 | } | |
1574 | } | |
a322f70c DW |
1575 | } |
1576 | ||
1577 | static int update_super_ddf(struct supertype *st, struct mdinfo *info, | |
1578 | char *update, | |
1579 | char *devname, int verbose, | |
1580 | int uuid_set, char *homehost) | |
1581 | { | |
1582 | /* For 'assemble' and 'force' we need to return non-zero if any | |
1583 | * change was made. For others, the return value is ignored. | |
1584 | * Update options are: | |
1585 | * force-one : This device looks a bit old but needs to be included, | |
1586 | * update age info appropriately. | |
1587 | * assemble: clear any 'faulty' flag to allow this device to | |
1588 | * be assembled. | |
1589 | * force-array: Array is degraded but being forced, mark it clean | |
1590 | * if that will be needed to assemble it. | |
1591 | * | |
1592 | * newdev: not used ???? | |
1593 | * grow: Array has gained a new device - this is currently for | |
1594 | * linear only | |
1595 | * resync: mark as dirty so a resync will happen. | |
59e36268 | 1596 | * uuid: Change the uuid of the array to match what is given |
a322f70c DW |
1597 | * homehost: update the recorded homehost |
1598 | * name: update the name - preserving the homehost | |
1599 | * _reshape_progress: record new reshape_progress position. | |
1600 | * | |
1601 | * Following are not relevant for this version: | |
1602 | * sparc2.2 : update from old dodgey metadata | |
1603 | * super-minor: change the preferred_minor number | |
1604 | * summaries: update redundant counters. | |
1605 | */ | |
1606 | int rv = 0; | |
1607 | // struct ddf_super *ddf = st->sb; | |
7a7cc504 | 1608 | // struct vd_config *vd = find_vdcr(ddf, info->container_member); |
a322f70c DW |
1609 | // struct virtual_entry *ve = find_ve(ddf); |
1610 | ||
a322f70c DW |
1611 | /* we don't need to handle "force-*" or "assemble" as |
1612 | * there is no need to 'trick' the kernel. We the metadata is | |
1613 | * first updated to activate the array, all the implied modifications | |
1614 | * will just happen. | |
1615 | */ | |
1616 | ||
1617 | if (strcmp(update, "grow") == 0) { | |
1618 | /* FIXME */ | |
1e2b2765 | 1619 | } else if (strcmp(update, "resync") == 0) { |
a322f70c | 1620 | // info->resync_checkpoint = 0; |
1e2b2765 | 1621 | } else if (strcmp(update, "homehost") == 0) { |
a322f70c DW |
1622 | /* homehost is stored in controller->vendor_data, |
1623 | * or it is when we are the vendor | |
1624 | */ | |
1625 | // if (info->vendor_is_local) | |
1626 | // strcpy(ddf->controller.vendor_data, homehost); | |
1e2b2765 | 1627 | rv = -1; |
f49208ec | 1628 | } else if (strcmp(update, "name") == 0) { |
a322f70c DW |
1629 | /* name is stored in virtual_entry->name */ |
1630 | // memset(ve->name, ' ', 16); | |
1631 | // strncpy(ve->name, info->name, 16); | |
1e2b2765 | 1632 | rv = -1; |
f49208ec | 1633 | } else if (strcmp(update, "_reshape_progress") == 0) { |
a322f70c | 1634 | /* We don't support reshape yet */ |
f49208ec N |
1635 | } else if (strcmp(update, "assemble") == 0 ) { |
1636 | /* Do nothing, just succeed */ | |
1637 | rv = 0; | |
1e2b2765 N |
1638 | } else |
1639 | rv = -1; | |
a322f70c DW |
1640 | |
1641 | // update_all_csum(ddf); | |
1642 | ||
1643 | return rv; | |
1644 | } | |
1645 | ||
5f8097be NB |
1646 | static void make_header_guid(char *guid) |
1647 | { | |
1648 | __u32 stamp; | |
5f8097be NB |
1649 | /* Create a DDF Header of Virtual Disk GUID */ |
1650 | ||
1651 | /* 24 bytes of fiction required. | |
1652 | * first 8 are a 'vendor-id' - "Linux-MD" | |
1653 | * next 8 are controller type.. how about 0X DEAD BEEF 0000 0000 | |
1654 | * Remaining 8 random number plus timestamp | |
1655 | */ | |
1656 | memcpy(guid, T10, sizeof(T10)); | |
1657 | stamp = __cpu_to_be32(0xdeadbeef); | |
1658 | memcpy(guid+8, &stamp, 4); | |
1659 | stamp = __cpu_to_be32(0); | |
1660 | memcpy(guid+12, &stamp, 4); | |
1661 | stamp = __cpu_to_be32(time(0) - DECADE); | |
1662 | memcpy(guid+16, &stamp, 4); | |
bfb7ea78 | 1663 | stamp = random32(); |
5f8097be | 1664 | memcpy(guid+20, &stamp, 4); |
5f8097be | 1665 | } |
59e36268 | 1666 | |
78e44928 NB |
1667 | static int init_super_ddf_bvd(struct supertype *st, |
1668 | mdu_array_info_t *info, | |
1669 | unsigned long long size, | |
1670 | char *name, char *homehost, | |
83cd1e97 | 1671 | int *uuid, unsigned long long data_offset); |
78e44928 | 1672 | |
a322f70c DW |
1673 | static int init_super_ddf(struct supertype *st, |
1674 | mdu_array_info_t *info, | |
1675 | unsigned long long size, char *name, char *homehost, | |
83cd1e97 | 1676 | int *uuid, unsigned long long data_offset) |
a322f70c DW |
1677 | { |
1678 | /* This is primarily called by Create when creating a new array. | |
1679 | * We will then get add_to_super called for each component, and then | |
1680 | * write_init_super called to write it out to each device. | |
1681 | * For DDF, Create can create on fresh devices or on a pre-existing | |
1682 | * array. | |
1683 | * To create on a pre-existing array a different method will be called. | |
1684 | * This one is just for fresh drives. | |
1685 | * | |
1686 | * We need to create the entire 'ddf' structure which includes: | |
1687 | * DDF headers - these are easy. | |
1688 | * Controller data - a Sector describing this controller .. not that | |
1689 | * this is a controller exactly. | |
1690 | * Physical Disk Record - one entry per device, so | |
1691 | * leave plenty of space. | |
1692 | * Virtual Disk Records - again, just leave plenty of space. | |
1693 | * This just lists VDs, doesn't give details | |
1694 | * Config records - describes the VDs that use this disk | |
1695 | * DiskData - describes 'this' device. | |
1696 | * BadBlockManagement - empty | |
1697 | * Diag Space - empty | |
1698 | * Vendor Logs - Could we put bitmaps here? | |
1699 | * | |
1700 | */ | |
1701 | struct ddf_super *ddf; | |
1702 | char hostname[17]; | |
1703 | int hostlen; | |
a322f70c DW |
1704 | int max_phys_disks, max_virt_disks; |
1705 | unsigned long long sector; | |
1706 | int clen; | |
1707 | int i; | |
1708 | int pdsize, vdsize; | |
1709 | struct phys_disk *pd; | |
1710 | struct virtual_disk *vd; | |
1711 | ||
83cd1e97 N |
1712 | if (data_offset != INVALID_SECTORS) { |
1713 | fprintf(stderr, Name ": data-offset not supported by DDF\n"); | |
1714 | return 0; | |
1715 | } | |
1716 | ||
78e44928 | 1717 | if (st->sb) |
83cd1e97 N |
1718 | return init_super_ddf_bvd(st, info, size, name, homehost, uuid, |
1719 | data_offset); | |
ba7eb04f | 1720 | |
3d2c4fc7 | 1721 | if (posix_memalign((void**)&ddf, 512, sizeof(*ddf)) != 0) { |
e7b84f9d | 1722 | pr_err("%s could not allocate superblock\n", __func__); |
3d2c4fc7 DW |
1723 | return 0; |
1724 | } | |
6264b437 | 1725 | memset(ddf, 0, sizeof(*ddf)); |
a322f70c DW |
1726 | ddf->dlist = NULL; /* no physical disks yet */ |
1727 | ddf->conflist = NULL; /* No virtual disks yet */ | |
955e9ea1 DW |
1728 | st->sb = ddf; |
1729 | ||
1730 | if (info == NULL) { | |
1731 | /* zeroing superblock */ | |
1732 | return 0; | |
1733 | } | |
a322f70c DW |
1734 | |
1735 | /* At least 32MB *must* be reserved for the ddf. So let's just | |
1736 | * start 32MB from the end, and put the primary header there. | |
1737 | * Don't do secondary for now. | |
1738 | * We don't know exactly where that will be yet as it could be | |
1739 | * different on each device. To just set up the lengths. | |
1740 | * | |
1741 | */ | |
1742 | ||
1743 | ddf->anchor.magic = DDF_HEADER_MAGIC; | |
5f8097be | 1744 | make_header_guid(ddf->anchor.guid); |
a322f70c | 1745 | |
59e36268 | 1746 | memcpy(ddf->anchor.revision, DDF_REVISION_2, 8); |
a322f70c DW |
1747 | ddf->anchor.seq = __cpu_to_be32(1); |
1748 | ddf->anchor.timestamp = __cpu_to_be32(time(0) - DECADE); | |
1749 | ddf->anchor.openflag = 0xFF; | |
1750 | ddf->anchor.foreignflag = 0; | |
1751 | ddf->anchor.enforcegroups = 0; /* Is this best?? */ | |
1752 | ddf->anchor.pad0 = 0xff; | |
1753 | memset(ddf->anchor.pad1, 0xff, 12); | |
1754 | memset(ddf->anchor.header_ext, 0xff, 32); | |
1755 | ddf->anchor.primary_lba = ~(__u64)0; | |
1756 | ddf->anchor.secondary_lba = ~(__u64)0; | |
1757 | ddf->anchor.type = DDF_HEADER_ANCHOR; | |
1758 | memset(ddf->anchor.pad2, 0xff, 3); | |
1759 | ddf->anchor.workspace_len = __cpu_to_be32(32768); /* Must be reserved */ | |
1760 | ddf->anchor.workspace_lba = ~(__u64)0; /* Put this at bottom | |
1761 | of 32M reserved.. */ | |
1762 | max_phys_disks = 1023; /* Should be enough */ | |
1763 | ddf->anchor.max_pd_entries = __cpu_to_be16(max_phys_disks); | |
1764 | max_virt_disks = 255; | |
1765 | ddf->anchor.max_vd_entries = __cpu_to_be16(max_virt_disks); /* ?? */ | |
1766 | ddf->anchor.max_partitions = __cpu_to_be16(64); /* ?? */ | |
1767 | ddf->max_part = 64; | |
8c3b8c2c | 1768 | ddf->mppe = 256; |
59e36268 NB |
1769 | ddf->conf_rec_len = 1 + ROUND_UP(ddf->mppe * (4+8), 512)/512; |
1770 | ddf->anchor.config_record_len = __cpu_to_be16(ddf->conf_rec_len); | |
1771 | ddf->anchor.max_primary_element_entries = __cpu_to_be16(ddf->mppe); | |
a322f70c | 1772 | memset(ddf->anchor.pad3, 0xff, 54); |
a322f70c DW |
1773 | /* controller sections is one sector long immediately |
1774 | * after the ddf header */ | |
1775 | sector = 1; | |
1776 | ddf->anchor.controller_section_offset = __cpu_to_be32(sector); | |
1777 | ddf->anchor.controller_section_length = __cpu_to_be32(1); | |
1778 | sector += 1; | |
1779 | ||
1780 | /* phys is 8 sectors after that */ | |
1781 | pdsize = ROUND_UP(sizeof(struct phys_disk) + | |
1782 | sizeof(struct phys_disk_entry)*max_phys_disks, | |
1783 | 512); | |
1784 | switch(pdsize/512) { | |
1785 | case 2: case 8: case 32: case 128: case 512: break; | |
1786 | default: abort(); | |
1787 | } | |
1788 | ddf->anchor.phys_section_offset = __cpu_to_be32(sector); | |
1789 | ddf->anchor.phys_section_length = | |
1790 | __cpu_to_be32(pdsize/512); /* max_primary_element_entries/8 */ | |
1791 | sector += pdsize/512; | |
1792 | ||
1793 | /* virt is another 32 sectors */ | |
1794 | vdsize = ROUND_UP(sizeof(struct virtual_disk) + | |
1795 | sizeof(struct virtual_entry) * max_virt_disks, | |
1796 | 512); | |
1797 | switch(vdsize/512) { | |
1798 | case 2: case 8: case 32: case 128: case 512: break; | |
1799 | default: abort(); | |
1800 | } | |
1801 | ddf->anchor.virt_section_offset = __cpu_to_be32(sector); | |
1802 | ddf->anchor.virt_section_length = | |
1803 | __cpu_to_be32(vdsize/512); /* max_vd_entries/8 */ | |
1804 | sector += vdsize/512; | |
1805 | ||
59e36268 | 1806 | clen = ddf->conf_rec_len * (ddf->max_part+1); |
a322f70c DW |
1807 | ddf->anchor.config_section_offset = __cpu_to_be32(sector); |
1808 | ddf->anchor.config_section_length = __cpu_to_be32(clen); | |
1809 | sector += clen; | |
1810 | ||
1811 | ddf->anchor.data_section_offset = __cpu_to_be32(sector); | |
1812 | ddf->anchor.data_section_length = __cpu_to_be32(1); | |
1813 | sector += 1; | |
1814 | ||
1815 | ddf->anchor.bbm_section_length = __cpu_to_be32(0); | |
1816 | ddf->anchor.bbm_section_offset = __cpu_to_be32(0xFFFFFFFF); | |
1817 | ddf->anchor.diag_space_length = __cpu_to_be32(0); | |
1818 | ddf->anchor.diag_space_offset = __cpu_to_be32(0xFFFFFFFF); | |
1819 | ddf->anchor.vendor_length = __cpu_to_be32(0); | |
1820 | ddf->anchor.vendor_offset = __cpu_to_be32(0xFFFFFFFF); | |
1821 | ||
1822 | memset(ddf->anchor.pad4, 0xff, 256); | |
1823 | ||
1824 | memcpy(&ddf->primary, &ddf->anchor, 512); | |
1825 | memcpy(&ddf->secondary, &ddf->anchor, 512); | |
1826 | ||
1827 | ddf->primary.openflag = 1; /* I guess.. */ | |
1828 | ddf->primary.type = DDF_HEADER_PRIMARY; | |
1829 | ||
1830 | ddf->secondary.openflag = 1; /* I guess.. */ | |
1831 | ddf->secondary.type = DDF_HEADER_SECONDARY; | |
1832 | ||
1833 | ddf->active = &ddf->primary; | |
1834 | ||
1835 | ddf->controller.magic = DDF_CONTROLLER_MAGIC; | |
1836 | ||
1837 | /* 24 more bytes of fiction required. | |
1838 | * first 8 are a 'vendor-id' - "Linux-MD" | |
1839 | * Remaining 16 are serial number.... maybe a hostname would do? | |
1840 | */ | |
1841 | memcpy(ddf->controller.guid, T10, sizeof(T10)); | |
1ba6bff9 DW |
1842 | gethostname(hostname, sizeof(hostname)); |
1843 | hostname[sizeof(hostname) - 1] = 0; | |
a322f70c DW |
1844 | hostlen = strlen(hostname); |
1845 | memcpy(ddf->controller.guid + 24 - hostlen, hostname, hostlen); | |
1846 | for (i = strlen(T10) ; i+hostlen < 24; i++) | |
1847 | ddf->controller.guid[i] = ' '; | |
1848 | ||
1849 | ddf->controller.type.vendor_id = __cpu_to_be16(0xDEAD); | |
1850 | ddf->controller.type.device_id = __cpu_to_be16(0xBEEF); | |
1851 | ddf->controller.type.sub_vendor_id = 0; | |
1852 | ddf->controller.type.sub_device_id = 0; | |
1853 | memcpy(ddf->controller.product_id, "What Is My PID??", 16); | |
1854 | memset(ddf->controller.pad, 0xff, 8); | |
1855 | memset(ddf->controller.vendor_data, 0xff, 448); | |
a9e1c11d N |
1856 | if (homehost && strlen(homehost) < 440) |
1857 | strcpy((char*)ddf->controller.vendor_data, homehost); | |
a322f70c | 1858 | |
3d2c4fc7 | 1859 | if (posix_memalign((void**)&pd, 512, pdsize) != 0) { |
e7b84f9d | 1860 | pr_err("%s could not allocate pd\n", __func__); |
3d2c4fc7 DW |
1861 | return 0; |
1862 | } | |
6416d527 | 1863 | ddf->phys = pd; |
a322f70c DW |
1864 | ddf->pdsize = pdsize; |
1865 | ||
1866 | memset(pd, 0xff, pdsize); | |
1867 | memset(pd, 0, sizeof(*pd)); | |
076515ba | 1868 | pd->magic = DDF_PHYS_RECORDS_MAGIC; |
a322f70c DW |
1869 | pd->used_pdes = __cpu_to_be16(0); |
1870 | pd->max_pdes = __cpu_to_be16(max_phys_disks); | |
1871 | memset(pd->pad, 0xff, 52); | |
1872 | ||
3d2c4fc7 | 1873 | if (posix_memalign((void**)&vd, 512, vdsize) != 0) { |
e7b84f9d | 1874 | pr_err("%s could not allocate vd\n", __func__); |
3d2c4fc7 DW |
1875 | return 0; |
1876 | } | |
6416d527 | 1877 | ddf->virt = vd; |
a322f70c DW |
1878 | ddf->vdsize = vdsize; |
1879 | memset(vd, 0, vdsize); | |
1880 | vd->magic = DDF_VIRT_RECORDS_MAGIC; | |
1881 | vd->populated_vdes = __cpu_to_be16(0); | |
1882 | vd->max_vdes = __cpu_to_be16(max_virt_disks); | |
1883 | memset(vd->pad, 0xff, 52); | |
1884 | ||
5f8097be NB |
1885 | for (i=0; i<max_virt_disks; i++) |
1886 | memset(&vd->entries[i], 0xff, sizeof(struct virtual_entry)); | |
1887 | ||
a322f70c | 1888 | st->sb = ddf; |
18a2f463 | 1889 | ddf->updates_pending = 1; |
a322f70c DW |
1890 | return 1; |
1891 | } | |
1892 | ||
5f8097be NB |
1893 | static int chunk_to_shift(int chunksize) |
1894 | { | |
1895 | return ffs(chunksize/512)-1; | |
1896 | } | |
1897 | ||
1898 | static int level_to_prl(int level) | |
1899 | { | |
1900 | switch (level) { | |
1901 | case LEVEL_LINEAR: return DDF_CONCAT; | |
1902 | case 0: return DDF_RAID0; | |
1903 | case 1: return DDF_RAID1; | |
1904 | case 4: return DDF_RAID4; | |
1905 | case 5: return DDF_RAID5; | |
1906 | case 6: return DDF_RAID6; | |
1907 | default: return -1; | |
1908 | } | |
1909 | } | |
613b0d17 | 1910 | |
5f8097be NB |
1911 | static int layout_to_rlq(int level, int layout, int raiddisks) |
1912 | { | |
1913 | switch(level) { | |
1914 | case 0: | |
1915 | return DDF_RAID0_SIMPLE; | |
1916 | case 1: | |
1917 | switch(raiddisks) { | |
1918 | case 2: return DDF_RAID1_SIMPLE; | |
1919 | case 3: return DDF_RAID1_MULTI; | |
1920 | default: return -1; | |
1921 | } | |
1922 | case 4: | |
1923 | switch(layout) { | |
1924 | case 0: return DDF_RAID4_N; | |
1925 | } | |
1926 | break; | |
1927 | case 5: | |
5f8097be NB |
1928 | switch(layout) { |
1929 | case ALGORITHM_LEFT_ASYMMETRIC: | |
1930 | return DDF_RAID5_N_RESTART; | |
1931 | case ALGORITHM_RIGHT_ASYMMETRIC: | |
b640a252 | 1932 | return DDF_RAID5_0_RESTART; |
5f8097be NB |
1933 | case ALGORITHM_LEFT_SYMMETRIC: |
1934 | return DDF_RAID5_N_CONTINUE; | |
1935 | case ALGORITHM_RIGHT_SYMMETRIC: | |
1936 | return -1; /* not mentioned in standard */ | |
1937 | } | |
b640a252 N |
1938 | case 6: |
1939 | switch(layout) { | |
1940 | case ALGORITHM_ROTATING_N_RESTART: | |
1941 | return DDF_RAID5_N_RESTART; | |
1942 | case ALGORITHM_ROTATING_ZERO_RESTART: | |
1943 | return DDF_RAID6_0_RESTART; | |
1944 | case ALGORITHM_ROTATING_N_CONTINUE: | |
1945 | return DDF_RAID5_N_CONTINUE; | |
1946 | } | |
5f8097be NB |
1947 | } |
1948 | return -1; | |
1949 | } | |
1950 | ||
598f0d58 NB |
1951 | static int rlq_to_layout(int rlq, int prl, int raiddisks) |
1952 | { | |
1953 | switch(prl) { | |
1954 | case DDF_RAID0: | |
1955 | return 0; /* hopefully rlq == DDF_RAID0_SIMPLE */ | |
1956 | case DDF_RAID1: | |
1957 | return 0; /* hopefully rlq == SIMPLE or MULTI depending | |
1958 | on raiddisks*/ | |
1959 | case DDF_RAID4: | |
1960 | switch(rlq) { | |
1961 | case DDF_RAID4_N: | |
1962 | return 0; | |
1963 | default: | |
1964 | /* not supported */ | |
1965 | return -1; /* FIXME this isn't checked */ | |
1966 | } | |
1967 | case DDF_RAID5: | |
598f0d58 NB |
1968 | switch(rlq) { |
1969 | case DDF_RAID5_N_RESTART: | |
1970 | return ALGORITHM_LEFT_ASYMMETRIC; | |
1971 | case DDF_RAID5_0_RESTART: | |
1972 | return ALGORITHM_RIGHT_ASYMMETRIC; | |
1973 | case DDF_RAID5_N_CONTINUE: | |
1974 | return ALGORITHM_LEFT_SYMMETRIC; | |
1975 | default: | |
1976 | return -1; | |
1977 | } | |
59e36268 NB |
1978 | case DDF_RAID6: |
1979 | switch(rlq) { | |
1980 | case DDF_RAID5_N_RESTART: | |
b640a252 | 1981 | return ALGORITHM_ROTATING_N_RESTART; |
59e36268 | 1982 | case DDF_RAID6_0_RESTART: |
b640a252 | 1983 | return ALGORITHM_ROTATING_ZERO_RESTART; |
59e36268 | 1984 | case DDF_RAID5_N_CONTINUE: |
b640a252 | 1985 | return ALGORITHM_ROTATING_N_CONTINUE; |
59e36268 NB |
1986 | default: |
1987 | return -1; | |
1988 | } | |
598f0d58 NB |
1989 | } |
1990 | return -1; | |
1991 | } | |
1992 | ||
0e600426 | 1993 | #ifndef MDASSEMBLE |
59e36268 NB |
1994 | struct extent { |
1995 | unsigned long long start, size; | |
1996 | }; | |
78e44928 | 1997 | static int cmp_extent(const void *av, const void *bv) |
59e36268 NB |
1998 | { |
1999 | const struct extent *a = av; | |
2000 | const struct extent *b = bv; | |
2001 | if (a->start < b->start) | |
2002 | return -1; | |
2003 | if (a->start > b->start) | |
2004 | return 1; | |
2005 | return 0; | |
2006 | } | |
2007 | ||
78e44928 | 2008 | static struct extent *get_extents(struct ddf_super *ddf, struct dl *dl) |
59e36268 NB |
2009 | { |
2010 | /* find a list of used extents on the give physical device | |
2011 | * (dnum) of the given ddf. | |
2012 | * Return a malloced array of 'struct extent' | |
2013 | ||
613b0d17 | 2014 | * FIXME ignore DDF_Legacy devices? |
59e36268 NB |
2015 | |
2016 | */ | |
2017 | struct extent *rv; | |
2018 | int n = 0; | |
f21e18ca | 2019 | unsigned int i, j; |
59e36268 | 2020 | |
503975b9 | 2021 | rv = xmalloc(sizeof(struct extent) * (ddf->max_part + 2)); |
59e36268 NB |
2022 | |
2023 | for (i = 0; i < ddf->max_part; i++) { | |
2024 | struct vcl *v = dl->vlist[i]; | |
2025 | if (v == NULL) | |
2026 | continue; | |
f21e18ca | 2027 | for (j = 0; j < v->conf.prim_elmnt_count; j++) |
59e36268 NB |
2028 | if (v->conf.phys_refnum[j] == dl->disk.refnum) { |
2029 | /* This device plays role 'j' in 'v'. */ | |
2030 | rv[n].start = __be64_to_cpu(v->lba_offset[j]); | |
2031 | rv[n].size = __be64_to_cpu(v->conf.blocks); | |
2032 | n++; | |
2033 | break; | |
2034 | } | |
2035 | } | |
2036 | qsort(rv, n, sizeof(*rv), cmp_extent); | |
2037 | ||
2038 | rv[n].start = __be64_to_cpu(ddf->phys->entries[dl->pdnum].config_size); | |
2039 | rv[n].size = 0; | |
2040 | return rv; | |
2041 | } | |
0e600426 | 2042 | #endif |
59e36268 | 2043 | |
5f8097be NB |
2044 | static int init_super_ddf_bvd(struct supertype *st, |
2045 | mdu_array_info_t *info, | |
2046 | unsigned long long size, | |
2047 | char *name, char *homehost, | |
83cd1e97 | 2048 | int *uuid, unsigned long long data_offset) |
5f8097be NB |
2049 | { |
2050 | /* We are creating a BVD inside a pre-existing container. | |
2051 | * so st->sb is already set. | |
2052 | * We need to create a new vd_config and a new virtual_entry | |
2053 | */ | |
2054 | struct ddf_super *ddf = st->sb; | |
f21e18ca | 2055 | unsigned int venum; |
5f8097be NB |
2056 | struct virtual_entry *ve; |
2057 | struct vcl *vcl; | |
2058 | struct vd_config *vc; | |
5f8097be NB |
2059 | |
2060 | if (__be16_to_cpu(ddf->virt->populated_vdes) | |
2061 | >= __be16_to_cpu(ddf->virt->max_vdes)) { | |
e7b84f9d N |
2062 | pr_err("This ddf already has the " |
2063 | "maximum of %d virtual devices\n", | |
2064 | __be16_to_cpu(ddf->virt->max_vdes)); | |
5f8097be NB |
2065 | return 0; |
2066 | } | |
2067 | ||
97c9c100 N |
2068 | if (name) |
2069 | for (venum = 0; venum < __be16_to_cpu(ddf->virt->max_vdes); venum++) | |
2070 | if (!all_ff(ddf->virt->entries[venum].guid)) { | |
2071 | char *n = ddf->virt->entries[venum].name; | |
2072 | ||
2073 | if (strncmp(name, n, 16) == 0) { | |
e7b84f9d N |
2074 | pr_err("This ddf already" |
2075 | " has an array called %s\n", | |
2076 | name); | |
97c9c100 N |
2077 | return 0; |
2078 | } | |
2079 | } | |
2080 | ||
5f8097be NB |
2081 | for (venum = 0; venum < __be16_to_cpu(ddf->virt->max_vdes); venum++) |
2082 | if (all_ff(ddf->virt->entries[venum].guid)) | |
2083 | break; | |
2084 | if (venum == __be16_to_cpu(ddf->virt->max_vdes)) { | |
e7b84f9d | 2085 | pr_err("Cannot find spare slot for " |
613b0d17 | 2086 | "virtual disk - DDF is corrupt\n"); |
5f8097be NB |
2087 | return 0; |
2088 | } | |
2089 | ve = &ddf->virt->entries[venum]; | |
2090 | ||
2091 | /* A Virtual Disk GUID contains the T10 Vendor ID, controller type, | |
2092 | * timestamp, random number | |
2093 | */ | |
2094 | make_header_guid(ve->guid); | |
2095 | ve->unit = __cpu_to_be16(info->md_minor); | |
2096 | ve->pad0 = 0xFFFF; | |
2097 | ve->guid_crc = crc32(0, (unsigned char*)ddf->anchor.guid, DDF_GUID_LEN); | |
2098 | ve->type = 0; | |
7a7cc504 NB |
2099 | ve->state = DDF_state_degraded; /* Will be modified as devices are added */ |
2100 | if (info->state & 1) /* clean */ | |
2101 | ve->init_state = DDF_init_full; | |
2102 | else | |
2103 | ve->init_state = DDF_init_not; | |
2104 | ||
5f8097be NB |
2105 | memset(ve->pad1, 0xff, 14); |
2106 | memset(ve->name, ' ', 16); | |
2107 | if (name) | |
2108 | strncpy(ve->name, name, 16); | |
2109 | ddf->virt->populated_vdes = | |
2110 | __cpu_to_be16(__be16_to_cpu(ddf->virt->populated_vdes)+1); | |
2111 | ||
2112 | /* Now create a new vd_config */ | |
3d2c4fc7 DW |
2113 | if (posix_memalign((void**)&vcl, 512, |
2114 | (offsetof(struct vcl, conf) + ddf->conf_rec_len * 512)) != 0) { | |
e7b84f9d | 2115 | pr_err("%s could not allocate vd_config\n", __func__); |
3d2c4fc7 DW |
2116 | return 0; |
2117 | } | |
8c3b8c2c | 2118 | vcl->lba_offset = (__u64*) &vcl->conf.phys_refnum[ddf->mppe]; |
59e36268 NB |
2119 | vcl->vcnum = venum; |
2120 | vcl->block_sizes = NULL; /* FIXME not for CONCAT */ | |
8ec5d685 | 2121 | vcl->other_bvds = NULL; |
5f8097be NB |
2122 | |
2123 | vc = &vcl->conf; | |
2124 | ||
2125 | vc->magic = DDF_VD_CONF_MAGIC; | |
2126 | memcpy(vc->guid, ve->guid, DDF_GUID_LEN); | |
2127 | vc->timestamp = __cpu_to_be32(time(0)-DECADE); | |
2128 | vc->seqnum = __cpu_to_be32(1); | |
2129 | memset(vc->pad0, 0xff, 24); | |
2130 | vc->prim_elmnt_count = __cpu_to_be16(info->raid_disks); | |
2131 | vc->chunk_shift = chunk_to_shift(info->chunk_size); | |
2132 | vc->prl = level_to_prl(info->level); | |
2133 | vc->rlq = layout_to_rlq(info->level, info->layout, info->raid_disks); | |
2134 | vc->sec_elmnt_count = 1; | |
2135 | vc->sec_elmnt_seq = 0; | |
2136 | vc->srl = 0; | |
2137 | vc->blocks = __cpu_to_be64(info->size * 2); | |
2138 | vc->array_blocks = __cpu_to_be64( | |
2139 | calc_array_size(info->level, info->raid_disks, info->layout, | |
2140 | info->chunk_size, info->size*2)); | |
2141 | memset(vc->pad1, 0xff, 8); | |
2142 | vc->spare_refs[0] = 0xffffffff; | |
2143 | vc->spare_refs[1] = 0xffffffff; | |
2144 | vc->spare_refs[2] = 0xffffffff; | |
2145 | vc->spare_refs[3] = 0xffffffff; | |
2146 | vc->spare_refs[4] = 0xffffffff; | |
2147 | vc->spare_refs[5] = 0xffffffff; | |
2148 | vc->spare_refs[6] = 0xffffffff; | |
2149 | vc->spare_refs[7] = 0xffffffff; | |
2150 | memset(vc->cache_pol, 0, 8); | |
2151 | vc->bg_rate = 0x80; | |
2152 | memset(vc->pad2, 0xff, 3); | |
2153 | memset(vc->pad3, 0xff, 52); | |
2154 | memset(vc->pad4, 0xff, 192); | |
2155 | memset(vc->v0, 0xff, 32); | |
2156 | memset(vc->v1, 0xff, 32); | |
2157 | memset(vc->v2, 0xff, 16); | |
2158 | memset(vc->v3, 0xff, 16); | |
2159 | memset(vc->vendor, 0xff, 32); | |
598f0d58 | 2160 | |
8c3b8c2c | 2161 | memset(vc->phys_refnum, 0xff, 4*ddf->mppe); |
e5a2a3cf | 2162 | memset(vc->phys_refnum+ddf->mppe, 0x00, 8*ddf->mppe); |
5f8097be NB |
2163 | |
2164 | vcl->next = ddf->conflist; | |
2165 | ddf->conflist = vcl; | |
d2ca6449 | 2166 | ddf->currentconf = vcl; |
18a2f463 | 2167 | ddf->updates_pending = 1; |
5f8097be NB |
2168 | return 1; |
2169 | } | |
2170 | ||
0e600426 | 2171 | #ifndef MDASSEMBLE |
5f8097be NB |
2172 | static void add_to_super_ddf_bvd(struct supertype *st, |
2173 | mdu_disk_info_t *dk, int fd, char *devname) | |
2174 | { | |
2175 | /* fd and devname identify a device with-in the ddf container (st). | |
2176 | * dk identifies a location in the new BVD. | |
2177 | * We need to find suitable free space in that device and update | |
2178 | * the phys_refnum and lba_offset for the newly created vd_config. | |
2179 | * We might also want to update the type in the phys_disk | |
5575e7d9 | 2180 | * section. |
8592f29d N |
2181 | * |
2182 | * Alternately: fd == -1 and we have already chosen which device to | |
2183 | * use and recorded in dlist->raid_disk; | |
5f8097be NB |
2184 | */ |
2185 | struct dl *dl; | |
2186 | struct ddf_super *ddf = st->sb; | |
2187 | struct vd_config *vc; | |
2188 | __u64 *lba_offset; | |
f21e18ca N |
2189 | unsigned int working; |
2190 | unsigned int i; | |
59e36268 NB |
2191 | unsigned long long blocks, pos, esize; |
2192 | struct extent *ex; | |
5f8097be | 2193 | |
8592f29d N |
2194 | if (fd == -1) { |
2195 | for (dl = ddf->dlist; dl ; dl = dl->next) | |
2196 | if (dl->raiddisk == dk->raid_disk) | |
2197 | break; | |
2198 | } else { | |
2199 | for (dl = ddf->dlist; dl ; dl = dl->next) | |
2200 | if (dl->major == dk->major && | |
2201 | dl->minor == dk->minor) | |
2202 | break; | |
2203 | } | |
5f8097be NB |
2204 | if (!dl || ! (dk->state & (1<<MD_DISK_SYNC))) |
2205 | return; | |
2206 | ||
d2ca6449 NB |
2207 | vc = &ddf->currentconf->conf; |
2208 | lba_offset = ddf->currentconf->lba_offset; | |
59e36268 NB |
2209 | |
2210 | ex = get_extents(ddf, dl); | |
2211 | if (!ex) | |
2212 | return; | |
2213 | ||
2214 | i = 0; pos = 0; | |
2215 | blocks = __be64_to_cpu(vc->blocks); | |
d2ca6449 NB |
2216 | if (ddf->currentconf->block_sizes) |
2217 | blocks = ddf->currentconf->block_sizes[dk->raid_disk]; | |
59e36268 NB |
2218 | |
2219 | do { | |
2220 | esize = ex[i].start - pos; | |
2221 | if (esize >= blocks) | |
2222 | break; | |
2223 | pos = ex[i].start + ex[i].size; | |
2224 | i++; | |
2225 | } while (ex[i-1].size); | |
2226 | ||
2227 | free(ex); | |
2228 | if (esize < blocks) | |
2229 | return; | |
2230 | ||
d2ca6449 | 2231 | ddf->currentdev = dk->raid_disk; |
5f8097be | 2232 | vc->phys_refnum[dk->raid_disk] = dl->disk.refnum; |
59e36268 | 2233 | lba_offset[dk->raid_disk] = __cpu_to_be64(pos); |
5f8097be | 2234 | |
f21e18ca | 2235 | for (i = 0; i < ddf->max_part ; i++) |
5575e7d9 NB |
2236 | if (dl->vlist[i] == NULL) |
2237 | break; | |
2238 | if (i == ddf->max_part) | |
2239 | return; | |
d2ca6449 | 2240 | dl->vlist[i] = ddf->currentconf; |
5f8097be | 2241 | |
8592f29d N |
2242 | if (fd >= 0) |
2243 | dl->fd = fd; | |
2244 | if (devname) | |
2245 | dl->devname = devname; | |
7a7cc504 NB |
2246 | |
2247 | /* Check how many working raid_disks, and if we can mark | |
2248 | * array as optimal yet | |
2249 | */ | |
2250 | working = 0; | |
5575e7d9 | 2251 | |
f21e18ca | 2252 | for (i = 0; i < __be16_to_cpu(vc->prim_elmnt_count); i++) |
7a7cc504 NB |
2253 | if (vc->phys_refnum[i] != 0xffffffff) |
2254 | working++; | |
59e36268 | 2255 | |
5575e7d9 | 2256 | /* Find which virtual_entry */ |
d2ca6449 | 2257 | i = ddf->currentconf->vcnum; |
7a7cc504 | 2258 | if (working == __be16_to_cpu(vc->prim_elmnt_count)) |
5575e7d9 NB |
2259 | ddf->virt->entries[i].state = |
2260 | (ddf->virt->entries[i].state & ~DDF_state_mask) | |
7a7cc504 NB |
2261 | | DDF_state_optimal; |
2262 | ||
2263 | if (vc->prl == DDF_RAID6 && | |
2264 | working+1 == __be16_to_cpu(vc->prim_elmnt_count)) | |
5575e7d9 NB |
2265 | ddf->virt->entries[i].state = |
2266 | (ddf->virt->entries[i].state & ~DDF_state_mask) | |
7a7cc504 | 2267 | | DDF_state_part_optimal; |
5575e7d9 NB |
2268 | |
2269 | ddf->phys->entries[dl->pdnum].type &= ~__cpu_to_be16(DDF_Global_Spare); | |
2270 | ddf->phys->entries[dl->pdnum].type |= __cpu_to_be16(DDF_Active_in_VD); | |
18a2f463 | 2271 | ddf->updates_pending = 1; |
5f8097be NB |
2272 | } |
2273 | ||
a322f70c DW |
2274 | /* add a device to a container, either while creating it or while |
2275 | * expanding a pre-existing container | |
2276 | */ | |
f20c3968 | 2277 | static int add_to_super_ddf(struct supertype *st, |
72ca9bcf N |
2278 | mdu_disk_info_t *dk, int fd, char *devname, |
2279 | unsigned long long data_offset) | |
a322f70c DW |
2280 | { |
2281 | struct ddf_super *ddf = st->sb; | |
2282 | struct dl *dd; | |
2283 | time_t now; | |
2284 | struct tm *tm; | |
2285 | unsigned long long size; | |
2286 | struct phys_disk_entry *pde; | |
f21e18ca | 2287 | unsigned int n, i; |
a322f70c | 2288 | struct stat stb; |
90fa1a29 | 2289 | __u32 *tptr; |
a322f70c | 2290 | |
78e44928 NB |
2291 | if (ddf->currentconf) { |
2292 | add_to_super_ddf_bvd(st, dk, fd, devname); | |
f20c3968 | 2293 | return 0; |
78e44928 NB |
2294 | } |
2295 | ||
a322f70c DW |
2296 | /* This is device numbered dk->number. We need to create |
2297 | * a phys_disk entry and a more detailed disk_data entry. | |
2298 | */ | |
2299 | fstat(fd, &stb); | |
3d2c4fc7 DW |
2300 | if (posix_memalign((void**)&dd, 512, |
2301 | sizeof(*dd) + sizeof(dd->vlist[0]) * ddf->max_part) != 0) { | |
e7b84f9d N |
2302 | pr_err("%s could allocate buffer for new disk, aborting\n", |
2303 | __func__); | |
f20c3968 | 2304 | return 1; |
3d2c4fc7 | 2305 | } |
a322f70c DW |
2306 | dd->major = major(stb.st_rdev); |
2307 | dd->minor = minor(stb.st_rdev); | |
2308 | dd->devname = devname; | |
a322f70c | 2309 | dd->fd = fd; |
b2280677 | 2310 | dd->spare = NULL; |
a322f70c DW |
2311 | |
2312 | dd->disk.magic = DDF_PHYS_DATA_MAGIC; | |
2313 | now = time(0); | |
2314 | tm = localtime(&now); | |
2315 | sprintf(dd->disk.guid, "%8s%04d%02d%02d", | |
2316 | T10, tm->tm_year+1900, tm->tm_mon+1, tm->tm_mday); | |
90fa1a29 JS |
2317 | tptr = (__u32 *)(dd->disk.guid + 16); |
2318 | *tptr++ = random32(); | |
2319 | *tptr = random32(); | |
a322f70c | 2320 | |
59e36268 NB |
2321 | do { |
2322 | /* Cannot be bothered finding a CRC of some irrelevant details*/ | |
bfb7ea78 | 2323 | dd->disk.refnum = random32(); |
f21e18ca N |
2324 | for (i = __be16_to_cpu(ddf->active->max_pd_entries); |
2325 | i > 0; i--) | |
2326 | if (ddf->phys->entries[i-1].refnum == dd->disk.refnum) | |
59e36268 | 2327 | break; |
f21e18ca | 2328 | } while (i > 0); |
59e36268 | 2329 | |
a322f70c DW |
2330 | dd->disk.forced_ref = 1; |
2331 | dd->disk.forced_guid = 1; | |
2332 | memset(dd->disk.vendor, ' ', 32); | |
2333 | memcpy(dd->disk.vendor, "Linux", 5); | |
2334 | memset(dd->disk.pad, 0xff, 442); | |
b2280677 | 2335 | for (i = 0; i < ddf->max_part ; i++) |
a322f70c DW |
2336 | dd->vlist[i] = NULL; |
2337 | ||
2338 | n = __be16_to_cpu(ddf->phys->used_pdes); | |
2339 | pde = &ddf->phys->entries[n]; | |
5575e7d9 NB |
2340 | dd->pdnum = n; |
2341 | ||
2cc2983d N |
2342 | if (st->update_tail) { |
2343 | int len = (sizeof(struct phys_disk) + | |
2344 | sizeof(struct phys_disk_entry)); | |
2345 | struct phys_disk *pd; | |
2346 | ||
503975b9 | 2347 | pd = xmalloc(len); |
2cc2983d N |
2348 | pd->magic = DDF_PHYS_RECORDS_MAGIC; |
2349 | pd->used_pdes = __cpu_to_be16(n); | |
2350 | pde = &pd->entries[0]; | |
2351 | dd->mdupdate = pd; | |
2352 | } else { | |
2353 | n++; | |
2354 | ddf->phys->used_pdes = __cpu_to_be16(n); | |
2355 | } | |
a322f70c DW |
2356 | |
2357 | memcpy(pde->guid, dd->disk.guid, DDF_GUID_LEN); | |
2358 | pde->refnum = dd->disk.refnum; | |
5575e7d9 | 2359 | pde->type = __cpu_to_be16(DDF_Forced_PD_GUID | DDF_Global_Spare); |
a322f70c DW |
2360 | pde->state = __cpu_to_be16(DDF_Online); |
2361 | get_dev_size(fd, NULL, &size); | |
2362 | /* We are required to reserve 32Meg, and record the size in sectors */ | |
2363 | pde->config_size = __cpu_to_be64( (size - 32*1024*1024) / 512); | |
2364 | sprintf(pde->path, "%17.17s","Information: nil") ; | |
2365 | memset(pde->pad, 0xff, 6); | |
2366 | ||
d2ca6449 | 2367 | dd->size = size >> 9; |
2cc2983d N |
2368 | if (st->update_tail) { |
2369 | dd->next = ddf->add_list; | |
2370 | ddf->add_list = dd; | |
2371 | } else { | |
2372 | dd->next = ddf->dlist; | |
2373 | ddf->dlist = dd; | |
2374 | ddf->updates_pending = 1; | |
2375 | } | |
f20c3968 DW |
2376 | |
2377 | return 0; | |
a322f70c DW |
2378 | } |
2379 | ||
4dd968cc N |
2380 | static int remove_from_super_ddf(struct supertype *st, mdu_disk_info_t *dk) |
2381 | { | |
2382 | struct ddf_super *ddf = st->sb; | |
2383 | struct dl *dl; | |
2384 | ||
2385 | /* mdmon has noticed that this disk (dk->major/dk->minor) has | |
2386 | * disappeared from the container. | |
2387 | * We need to arrange that it disappears from the metadata and | |
2388 | * internal data structures too. | |
2389 | * Most of the work is done by ddf_process_update which edits | |
2390 | * the metadata and closes the file handle and attaches the memory | |
2391 | * where free_updates will free it. | |
2392 | */ | |
2393 | for (dl = ddf->dlist; dl ; dl = dl->next) | |
2394 | if (dl->major == dk->major && | |
2395 | dl->minor == dk->minor) | |
2396 | break; | |
2397 | if (!dl) | |
2398 | return -1; | |
2399 | ||
2400 | if (st->update_tail) { | |
2401 | int len = (sizeof(struct phys_disk) + | |
2402 | sizeof(struct phys_disk_entry)); | |
2403 | struct phys_disk *pd; | |
2404 | ||
503975b9 | 2405 | pd = xmalloc(len); |
4dd968cc N |
2406 | pd->magic = DDF_PHYS_RECORDS_MAGIC; |
2407 | pd->used_pdes = __cpu_to_be16(dl->pdnum); | |
2408 | pd->entries[0].state = __cpu_to_be16(DDF_Missing); | |
2409 | append_metadata_update(st, pd, len); | |
2410 | } | |
2411 | return 0; | |
2412 | } | |
2413 | ||
a322f70c DW |
2414 | /* |
2415 | * This is the write_init_super method for a ddf container. It is | |
2416 | * called when creating a container or adding another device to a | |
2417 | * container. | |
2418 | */ | |
42d5dfd9 | 2419 | #define NULL_CONF_SZ 4096 |
18a2f463 | 2420 | |
e3c2a365 | 2421 | static unsigned int get_pd_index_from_refnum(const struct vcl *vc, |
2422 | __u32 refnum, unsigned int nmax, | |
2423 | const struct vd_config **bvd, | |
2424 | unsigned int *idx); | |
2425 | ||
7f798aca | 2426 | static int __write_ddf_structure(struct dl *d, struct ddf_super *ddf, __u8 type, |
2427 | char *null_aligned) | |
a322f70c | 2428 | { |
7f798aca | 2429 | unsigned long long sector; |
2430 | struct ddf_header *header; | |
2431 | int fd, i, n_config, conf_size; | |
2432 | ||
2433 | fd = d->fd; | |
2434 | ||
2435 | switch (type) { | |
2436 | case DDF_HEADER_PRIMARY: | |
2437 | header = &ddf->primary; | |
2438 | sector = __be64_to_cpu(header->primary_lba); | |
2439 | break; | |
2440 | case DDF_HEADER_SECONDARY: | |
2441 | header = &ddf->secondary; | |
2442 | sector = __be64_to_cpu(header->secondary_lba); | |
2443 | break; | |
2444 | default: | |
2445 | return 0; | |
2446 | } | |
2447 | ||
2448 | header->type = type; | |
2449 | header->openflag = 0; | |
2450 | header->crc = calc_crc(header, 512); | |
2451 | ||
2452 | lseek64(fd, sector<<9, 0); | |
2453 | if (write(fd, header, 512) < 0) | |
2454 | return 0; | |
2455 | ||
2456 | ddf->controller.crc = calc_crc(&ddf->controller, 512); | |
2457 | if (write(fd, &ddf->controller, 512) < 0) | |
2458 | return 0; | |
a322f70c | 2459 | |
7f798aca | 2460 | ddf->phys->crc = calc_crc(ddf->phys, ddf->pdsize); |
2461 | if (write(fd, ddf->phys, ddf->pdsize) < 0) | |
2462 | return 0; | |
2463 | ddf->virt->crc = calc_crc(ddf->virt, ddf->vdsize); | |
2464 | if (write(fd, ddf->virt, ddf->vdsize) < 0) | |
2465 | return 0; | |
2466 | ||
2467 | /* Now write lots of config records. */ | |
2468 | n_config = ddf->max_part; | |
2469 | conf_size = ddf->conf_rec_len * 512; | |
2470 | for (i = 0 ; i <= n_config ; i++) { | |
e3c2a365 | 2471 | struct vcl *c; |
2472 | struct vd_config *vdc = NULL; | |
2473 | if (i == n_config) { | |
7f798aca | 2474 | c = (struct vcl *)d->spare; |
e3c2a365 | 2475 | if (c) |
2476 | vdc = &c->conf; | |
2477 | } else { | |
2478 | unsigned int dummy; | |
2479 | c = d->vlist[i]; | |
2480 | if (c) | |
2481 | get_pd_index_from_refnum( | |
2482 | c, d->disk.refnum, | |
2483 | ddf->mppe, | |
2484 | (const struct vd_config **)&vdc, | |
2485 | &dummy); | |
2486 | } | |
7f798aca | 2487 | if (c) { |
dacf3dc5 | 2488 | vdc->seqnum = header->seq; |
e3c2a365 | 2489 | vdc->crc = calc_crc(vdc, conf_size); |
2490 | if (write(fd, vdc, conf_size) < 0) | |
7f798aca | 2491 | break; |
2492 | } else { | |
2493 | unsigned int togo = conf_size; | |
2494 | while (togo > NULL_CONF_SZ) { | |
2495 | if (write(fd, null_aligned, NULL_CONF_SZ) < 0) | |
2496 | break; | |
2497 | togo -= NULL_CONF_SZ; | |
2498 | } | |
2499 | if (write(fd, null_aligned, togo) < 0) | |
2500 | break; | |
2501 | } | |
2502 | } | |
2503 | if (i <= n_config) | |
2504 | return 0; | |
2505 | ||
2506 | d->disk.crc = calc_crc(&d->disk, 512); | |
2507 | if (write(fd, &d->disk, 512) < 0) | |
2508 | return 0; | |
2509 | ||
2510 | return 1; | |
2511 | } | |
2512 | ||
2513 | static int __write_init_super_ddf(struct supertype *st) | |
2514 | { | |
a322f70c | 2515 | struct ddf_super *ddf = st->sb; |
a322f70c | 2516 | struct dl *d; |
175593bf DW |
2517 | int attempts = 0; |
2518 | int successes = 0; | |
7f798aca | 2519 | unsigned long long size; |
42d5dfd9 | 2520 | char *null_aligned; |
0175cbf6 | 2521 | __u32 seq; |
42d5dfd9 JS |
2522 | |
2523 | if (posix_memalign((void**)&null_aligned, 4096, NULL_CONF_SZ) != 0) { | |
2524 | return -ENOMEM; | |
2525 | } | |
2526 | memset(null_aligned, 0xff, NULL_CONF_SZ); | |
a322f70c | 2527 | |
dc9e279c | 2528 | seq = ddf->active->seq + 1; |
0175cbf6 | 2529 | |
175593bf DW |
2530 | /* try to write updated metadata, |
2531 | * if we catch a failure move on to the next disk | |
2532 | */ | |
a322f70c DW |
2533 | for (d = ddf->dlist; d; d=d->next) { |
2534 | int fd = d->fd; | |
2535 | ||
2536 | if (fd < 0) | |
2537 | continue; | |
2538 | ||
175593bf | 2539 | attempts++; |
a322f70c DW |
2540 | /* We need to fill in the primary, (secondary) and workspace |
2541 | * lba's in the headers, set their checksums, | |
2542 | * Also checksum phys, virt.... | |
2543 | * | |
2544 | * Then write everything out, finally the anchor is written. | |
2545 | */ | |
2546 | get_dev_size(fd, NULL, &size); | |
2547 | size /= 512; | |
097bcf00 | 2548 | if (d->workspace_lba != 0) |
2549 | ddf->anchor.workspace_lba = d->workspace_lba; | |
2550 | else | |
2551 | ddf->anchor.workspace_lba = | |
2552 | __cpu_to_be64(size - 32*1024*2); | |
2553 | if (d->primary_lba != 0) | |
2554 | ddf->anchor.primary_lba = d->primary_lba; | |
2555 | else | |
2556 | ddf->anchor.primary_lba = | |
2557 | __cpu_to_be64(size - 16*1024*2); | |
2558 | if (d->secondary_lba != 0) | |
2559 | ddf->anchor.secondary_lba = d->secondary_lba; | |
2560 | else | |
2561 | ddf->anchor.secondary_lba = | |
2562 | __cpu_to_be64(size - 32*1024*2); | |
0175cbf6 | 2563 | ddf->anchor.seq = seq; |
a322f70c DW |
2564 | memcpy(&ddf->primary, &ddf->anchor, 512); |
2565 | memcpy(&ddf->secondary, &ddf->anchor, 512); | |
2566 | ||
2567 | ddf->anchor.openflag = 0xFF; /* 'open' means nothing */ | |
2568 | ddf->anchor.seq = 0xFFFFFFFF; /* no sequencing in anchor */ | |
2569 | ddf->anchor.crc = calc_crc(&ddf->anchor, 512); | |
2570 | ||
7f798aca | 2571 | if (!__write_ddf_structure(d, ddf, DDF_HEADER_PRIMARY, |
2572 | null_aligned)) | |
175593bf | 2573 | continue; |
a322f70c | 2574 | |
7f798aca | 2575 | if (!__write_ddf_structure(d, ddf, DDF_HEADER_SECONDARY, |
2576 | null_aligned)) | |
175593bf | 2577 | continue; |
a322f70c | 2578 | |
a322f70c | 2579 | lseek64(fd, (size-1)*512, SEEK_SET); |
175593bf DW |
2580 | if (write(fd, &ddf->anchor, 512) < 0) |
2581 | continue; | |
2582 | successes++; | |
2583 | } | |
42d5dfd9 | 2584 | free(null_aligned); |
175593bf | 2585 | |
175593bf | 2586 | return attempts != successes; |
a322f70c | 2587 | } |
7a7cc504 NB |
2588 | |
2589 | static int write_init_super_ddf(struct supertype *st) | |
2590 | { | |
9b1fb677 DW |
2591 | struct ddf_super *ddf = st->sb; |
2592 | struct vcl *currentconf = ddf->currentconf; | |
2593 | ||
2594 | /* we are done with currentconf reset it to point st at the container */ | |
2595 | ddf->currentconf = NULL; | |
edd8d13c NB |
2596 | |
2597 | if (st->update_tail) { | |
2598 | /* queue the virtual_disk and vd_config as metadata updates */ | |
2599 | struct virtual_disk *vd; | |
2600 | struct vd_config *vc; | |
edd8d13c NB |
2601 | int len; |
2602 | ||
9b1fb677 | 2603 | if (!currentconf) { |
2cc2983d N |
2604 | int len = (sizeof(struct phys_disk) + |
2605 | sizeof(struct phys_disk_entry)); | |
2606 | ||
2607 | /* adding a disk to the container. */ | |
2608 | if (!ddf->add_list) | |
2609 | return 0; | |
2610 | ||
2611 | append_metadata_update(st, ddf->add_list->mdupdate, len); | |
2612 | ddf->add_list->mdupdate = NULL; | |
2613 | return 0; | |
2614 | } | |
2615 | ||
2616 | /* Newly created VD */ | |
2617 | ||
edd8d13c NB |
2618 | /* First the virtual disk. We have a slightly fake header */ |
2619 | len = sizeof(struct virtual_disk) + sizeof(struct virtual_entry); | |
503975b9 | 2620 | vd = xmalloc(len); |
edd8d13c | 2621 | *vd = *ddf->virt; |
9b1fb677 DW |
2622 | vd->entries[0] = ddf->virt->entries[currentconf->vcnum]; |
2623 | vd->populated_vdes = __cpu_to_be16(currentconf->vcnum); | |
edd8d13c NB |
2624 | append_metadata_update(st, vd, len); |
2625 | ||
2626 | /* Then the vd_config */ | |
2627 | len = ddf->conf_rec_len * 512; | |
503975b9 | 2628 | vc = xmalloc(len); |
9b1fb677 | 2629 | memcpy(vc, ¤tconf->conf, len); |
edd8d13c NB |
2630 | append_metadata_update(st, vc, len); |
2631 | ||
2632 | /* FIXME I need to close the fds! */ | |
2633 | return 0; | |
613b0d17 | 2634 | } else { |
d682f344 N |
2635 | struct dl *d; |
2636 | for (d = ddf->dlist; d; d=d->next) | |
ba728be7 | 2637 | while (Kill(d->devname, NULL, 0, -1, 1) == 0); |
1cc7f4fe | 2638 | return __write_init_super_ddf(st); |
d682f344 | 2639 | } |
7a7cc504 NB |
2640 | } |
2641 | ||
a322f70c DW |
2642 | #endif |
2643 | ||
387fcd59 N |
2644 | static __u64 avail_size_ddf(struct supertype *st, __u64 devsize, |
2645 | unsigned long long data_offset) | |
a322f70c DW |
2646 | { |
2647 | /* We must reserve the last 32Meg */ | |
2648 | if (devsize <= 32*1024*2) | |
2649 | return 0; | |
2650 | return devsize - 32*1024*2; | |
2651 | } | |
2652 | ||
2653 | #ifndef MDASSEMBLE | |
8592f29d N |
2654 | |
2655 | static int reserve_space(struct supertype *st, int raiddisks, | |
2656 | unsigned long long size, int chunk, | |
2657 | unsigned long long *freesize) | |
2658 | { | |
2659 | /* Find 'raiddisks' spare extents at least 'size' big (but | |
2660 | * only caring about multiples of 'chunk') and remember | |
2661 | * them. | |
2662 | * If the cannot be found, fail. | |
2663 | */ | |
2664 | struct dl *dl; | |
2665 | struct ddf_super *ddf = st->sb; | |
2666 | int cnt = 0; | |
2667 | ||
2668 | for (dl = ddf->dlist; dl ; dl=dl->next) { | |
613b0d17 | 2669 | dl->raiddisk = -1; |
8592f29d N |
2670 | dl->esize = 0; |
2671 | } | |
2672 | /* Now find largest extent on each device */ | |
2673 | for (dl = ddf->dlist ; dl ; dl=dl->next) { | |
2674 | struct extent *e = get_extents(ddf, dl); | |
2675 | unsigned long long pos = 0; | |
2676 | int i = 0; | |
2677 | int found = 0; | |
2678 | unsigned long long minsize = size; | |
2679 | ||
2680 | if (size == 0) | |
2681 | minsize = chunk; | |
2682 | ||
2683 | if (!e) | |
2684 | continue; | |
2685 | do { | |
2686 | unsigned long long esize; | |
2687 | esize = e[i].start - pos; | |
2688 | if (esize >= minsize) { | |
2689 | found = 1; | |
2690 | minsize = esize; | |
2691 | } | |
2692 | pos = e[i].start + e[i].size; | |
2693 | i++; | |
2694 | } while (e[i-1].size); | |
2695 | if (found) { | |
2696 | cnt++; | |
2697 | dl->esize = minsize; | |
2698 | } | |
2699 | free(e); | |
2700 | } | |
2701 | if (cnt < raiddisks) { | |
e7b84f9d | 2702 | pr_err("not enough devices with space to create array.\n"); |
8592f29d N |
2703 | return 0; /* No enough free spaces large enough */ |
2704 | } | |
2705 | if (size == 0) { | |
2706 | /* choose the largest size of which there are at least 'raiddisk' */ | |
2707 | for (dl = ddf->dlist ; dl ; dl=dl->next) { | |
2708 | struct dl *dl2; | |
2709 | if (dl->esize <= size) | |
2710 | continue; | |
2711 | /* This is bigger than 'size', see if there are enough */ | |
2712 | cnt = 0; | |
7b80ad6a | 2713 | for (dl2 = ddf->dlist; dl2 ; dl2=dl2->next) |
8592f29d N |
2714 | if (dl2->esize >= dl->esize) |
2715 | cnt++; | |
2716 | if (cnt >= raiddisks) | |
2717 | size = dl->esize; | |
2718 | } | |
2719 | if (chunk) { | |
2720 | size = size / chunk; | |
2721 | size *= chunk; | |
2722 | } | |
2723 | *freesize = size; | |
2724 | if (size < 32) { | |
e7b84f9d | 2725 | pr_err("not enough spare devices to create array.\n"); |
8592f29d N |
2726 | return 0; |
2727 | } | |
2728 | } | |
2729 | /* We have a 'size' of which there are enough spaces. | |
2730 | * We simply do a first-fit */ | |
2731 | cnt = 0; | |
2732 | for (dl = ddf->dlist ; dl && cnt < raiddisks ; dl=dl->next) { | |
2733 | if (dl->esize < size) | |
2734 | continue; | |
613b0d17 | 2735 | |
8592f29d N |
2736 | dl->raiddisk = cnt; |
2737 | cnt++; | |
2738 | } | |
2739 | return 1; | |
2740 | } | |
2741 | ||
2c514b71 NB |
2742 | static int |
2743 | validate_geometry_ddf_container(struct supertype *st, | |
2744 | int level, int layout, int raiddisks, | |
2745 | int chunk, unsigned long long size, | |
af4348dd | 2746 | unsigned long long data_offset, |
2c514b71 NB |
2747 | char *dev, unsigned long long *freesize, |
2748 | int verbose); | |
78e44928 NB |
2749 | |
2750 | static int validate_geometry_ddf_bvd(struct supertype *st, | |
2751 | int level, int layout, int raiddisks, | |
c21e737b | 2752 | int *chunk, unsigned long long size, |
af4348dd | 2753 | unsigned long long data_offset, |
2c514b71 NB |
2754 | char *dev, unsigned long long *freesize, |
2755 | int verbose); | |
78e44928 NB |
2756 | |
2757 | static int validate_geometry_ddf(struct supertype *st, | |
2c514b71 | 2758 | int level, int layout, int raiddisks, |
c21e737b | 2759 | int *chunk, unsigned long long size, |
af4348dd | 2760 | unsigned long long data_offset, |
2c514b71 NB |
2761 | char *dev, unsigned long long *freesize, |
2762 | int verbose) | |
a322f70c DW |
2763 | { |
2764 | int fd; | |
2765 | struct mdinfo *sra; | |
2766 | int cfd; | |
2767 | ||
2768 | /* ddf potentially supports lots of things, but it depends on | |
2769 | * what devices are offered (and maybe kernel version?) | |
2770 | * If given unused devices, we will make a container. | |
2771 | * If given devices in a container, we will make a BVD. | |
2772 | * If given BVDs, we make an SVD, changing all the GUIDs in the process. | |
2773 | */ | |
2774 | ||
bb7295f1 N |
2775 | if (chunk && *chunk == UnSet) |
2776 | *chunk = DEFAULT_CHUNK; | |
2777 | ||
542ef4ec | 2778 | if (level == -1000000) level = LEVEL_CONTAINER; |
a322f70c | 2779 | if (level == LEVEL_CONTAINER) { |
78e44928 NB |
2780 | /* Must be a fresh device to add to a container */ |
2781 | return validate_geometry_ddf_container(st, level, layout, | |
c21e737b | 2782 | raiddisks, chunk?*chunk:0, |
af4348dd N |
2783 | size, data_offset, dev, |
2784 | freesize, | |
2c514b71 | 2785 | verbose); |
5f8097be NB |
2786 | } |
2787 | ||
78e44928 NB |
2788 | if (!dev) { |
2789 | /* Initial sanity check. Exclude illegal levels. */ | |
2790 | int i; | |
2791 | for (i=0; ddf_level_num[i].num1 != MAXINT; i++) | |
2792 | if (ddf_level_num[i].num2 == level) | |
2793 | break; | |
b42f577a N |
2794 | if (ddf_level_num[i].num1 == MAXINT) { |
2795 | if (verbose) | |
e7b84f9d | 2796 | pr_err("DDF does not support level %d arrays\n", |
613b0d17 | 2797 | level); |
78e44928 | 2798 | return 0; |
b42f577a | 2799 | } |
78e44928 | 2800 | /* Should check layout? etc */ |
8592f29d N |
2801 | |
2802 | if (st->sb && freesize) { | |
2803 | /* --create was given a container to create in. | |
2804 | * So we need to check that there are enough | |
2805 | * free spaces and return the amount of space. | |
2806 | * We may as well remember which drives were | |
2807 | * chosen so that add_to_super/getinfo_super | |
2808 | * can return them. | |
2809 | */ | |
c21e737b | 2810 | return reserve_space(st, raiddisks, size, chunk?*chunk:0, freesize); |
8592f29d | 2811 | } |
a322f70c | 2812 | return 1; |
78e44928 | 2813 | } |
a322f70c | 2814 | |
8592f29d N |
2815 | if (st->sb) { |
2816 | /* A container has already been opened, so we are | |
2817 | * creating in there. Maybe a BVD, maybe an SVD. | |
2818 | * Should make a distinction one day. | |
2819 | */ | |
2820 | return validate_geometry_ddf_bvd(st, level, layout, raiddisks, | |
af4348dd N |
2821 | chunk, size, data_offset, dev, |
2822 | freesize, | |
8592f29d N |
2823 | verbose); |
2824 | } | |
78e44928 NB |
2825 | /* This is the first device for the array. |
2826 | * If it is a container, we read it in and do automagic allocations, | |
2827 | * no other devices should be given. | |
2828 | * Otherwise it must be a member device of a container, and we | |
2829 | * do manual allocation. | |
2830 | * Later we should check for a BVD and make an SVD. | |
a322f70c | 2831 | */ |
a322f70c DW |
2832 | fd = open(dev, O_RDONLY|O_EXCL, 0); |
2833 | if (fd >= 0) { | |
4dd2df09 | 2834 | sra = sysfs_read(fd, NULL, GET_VERSION); |
a322f70c DW |
2835 | close(fd); |
2836 | if (sra && sra->array.major_version == -1 && | |
78e44928 NB |
2837 | strcmp(sra->text_version, "ddf") == 0) { |
2838 | ||
2839 | /* load super */ | |
2840 | /* find space for 'n' devices. */ | |
2841 | /* remember the devices */ | |
2842 | /* Somehow return the fact that we have enough */ | |
a322f70c DW |
2843 | } |
2844 | ||
2c514b71 | 2845 | if (verbose) |
e7b84f9d N |
2846 | pr_err("ddf: Cannot create this array " |
2847 | "on device %s - a container is required.\n", | |
2848 | dev); | |
a322f70c DW |
2849 | return 0; |
2850 | } | |
2851 | if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) { | |
2c514b71 | 2852 | if (verbose) |
e7b84f9d | 2853 | pr_err("ddf: Cannot open %s: %s\n", |
613b0d17 | 2854 | dev, strerror(errno)); |
a322f70c DW |
2855 | return 0; |
2856 | } | |
2857 | /* Well, it is in use by someone, maybe a 'ddf' container. */ | |
2858 | cfd = open_container(fd); | |
2859 | if (cfd < 0) { | |
2860 | close(fd); | |
2c514b71 | 2861 | if (verbose) |
e7b84f9d | 2862 | pr_err("ddf: Cannot use %s: %s\n", |
613b0d17 | 2863 | dev, strerror(EBUSY)); |
a322f70c DW |
2864 | return 0; |
2865 | } | |
4dd2df09 | 2866 | sra = sysfs_read(cfd, NULL, GET_VERSION); |
a322f70c DW |
2867 | close(fd); |
2868 | if (sra && sra->array.major_version == -1 && | |
2869 | strcmp(sra->text_version, "ddf") == 0) { | |
2870 | /* This is a member of a ddf container. Load the container | |
2871 | * and try to create a bvd | |
2872 | */ | |
2873 | struct ddf_super *ddf; | |
e1902a7b | 2874 | if (load_super_ddf_all(st, cfd, (void **)&ddf, NULL) == 0) { |
5f8097be | 2875 | st->sb = ddf; |
4dd2df09 | 2876 | strcpy(st->container_devnm, fd2devnm(cfd)); |
a322f70c | 2877 | close(cfd); |
78e44928 | 2878 | return validate_geometry_ddf_bvd(st, level, layout, |
a322f70c | 2879 | raiddisks, chunk, size, |
af4348dd | 2880 | data_offset, |
2c514b71 NB |
2881 | dev, freesize, |
2882 | verbose); | |
a322f70c DW |
2883 | } |
2884 | close(cfd); | |
c42ec1ed DW |
2885 | } else /* device may belong to a different container */ |
2886 | return 0; | |
2887 | ||
a322f70c DW |
2888 | return 1; |
2889 | } | |
2890 | ||
2c514b71 NB |
2891 | static int |
2892 | validate_geometry_ddf_container(struct supertype *st, | |
2893 | int level, int layout, int raiddisks, | |
2894 | int chunk, unsigned long long size, | |
af4348dd | 2895 | unsigned long long data_offset, |
2c514b71 NB |
2896 | char *dev, unsigned long long *freesize, |
2897 | int verbose) | |
a322f70c DW |
2898 | { |
2899 | int fd; | |
2900 | unsigned long long ldsize; | |
2901 | ||
2902 | if (level != LEVEL_CONTAINER) | |
2903 | return 0; | |
2904 | if (!dev) | |
2905 | return 1; | |
2906 | ||
2907 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
2908 | if (fd < 0) { | |
2c514b71 | 2909 | if (verbose) |
e7b84f9d | 2910 | pr_err("ddf: Cannot open %s: %s\n", |
613b0d17 | 2911 | dev, strerror(errno)); |
a322f70c DW |
2912 | return 0; |
2913 | } | |
2914 | if (!get_dev_size(fd, dev, &ldsize)) { | |
2915 | close(fd); | |
2916 | return 0; | |
2917 | } | |
2918 | close(fd); | |
2919 | ||
387fcd59 | 2920 | *freesize = avail_size_ddf(st, ldsize >> 9, INVALID_SECTORS); |
ea17e7aa N |
2921 | if (*freesize == 0) |
2922 | return 0; | |
a322f70c DW |
2923 | |
2924 | return 1; | |
2925 | } | |
2926 | ||
78e44928 NB |
2927 | static int validate_geometry_ddf_bvd(struct supertype *st, |
2928 | int level, int layout, int raiddisks, | |
c21e737b | 2929 | int *chunk, unsigned long long size, |
af4348dd | 2930 | unsigned long long data_offset, |
2c514b71 NB |
2931 | char *dev, unsigned long long *freesize, |
2932 | int verbose) | |
a322f70c DW |
2933 | { |
2934 | struct stat stb; | |
2935 | struct ddf_super *ddf = st->sb; | |
2936 | struct dl *dl; | |
5f8097be NB |
2937 | unsigned long long pos = 0; |
2938 | unsigned long long maxsize; | |
2939 | struct extent *e; | |
2940 | int i; | |
a322f70c | 2941 | /* ddf/bvd supports lots of things, but not containers */ |
b42f577a N |
2942 | if (level == LEVEL_CONTAINER) { |
2943 | if (verbose) | |
e7b84f9d | 2944 | pr_err("DDF cannot create a container within an container\n"); |
a322f70c | 2945 | return 0; |
b42f577a | 2946 | } |
a322f70c DW |
2947 | /* We must have the container info already read in. */ |
2948 | if (!ddf) | |
2949 | return 0; | |
2950 | ||
5f8097be NB |
2951 | if (!dev) { |
2952 | /* General test: make sure there is space for | |
2953 | * 'raiddisks' device extents of size 'size'. | |
2954 | */ | |
2955 | unsigned long long minsize = size; | |
2956 | int dcnt = 0; | |
2957 | if (minsize == 0) | |
2958 | minsize = 8; | |
2959 | for (dl = ddf->dlist; dl ; dl = dl->next) | |
2960 | { | |
2961 | int found = 0; | |
7e1432fb | 2962 | pos = 0; |
5f8097be NB |
2963 | |
2964 | i = 0; | |
2965 | e = get_extents(ddf, dl); | |
2966 | if (!e) continue; | |
2967 | do { | |
2968 | unsigned long long esize; | |
2969 | esize = e[i].start - pos; | |
2970 | if (esize >= minsize) | |
2971 | found = 1; | |
2972 | pos = e[i].start + e[i].size; | |
2973 | i++; | |
2974 | } while (e[i-1].size); | |
2975 | if (found) | |
2976 | dcnt++; | |
2977 | free(e); | |
2978 | } | |
2979 | if (dcnt < raiddisks) { | |
2c514b71 | 2980 | if (verbose) |
e7b84f9d N |
2981 | pr_err("ddf: Not enough devices with " |
2982 | "space for this array (%d < %d)\n", | |
2983 | dcnt, raiddisks); | |
5f8097be NB |
2984 | return 0; |
2985 | } | |
2986 | return 1; | |
2987 | } | |
a322f70c DW |
2988 | /* This device must be a member of the set */ |
2989 | if (stat(dev, &stb) < 0) | |
2990 | return 0; | |
2991 | if ((S_IFMT & stb.st_mode) != S_IFBLK) | |
2992 | return 0; | |
2993 | for (dl = ddf->dlist ; dl ; dl = dl->next) { | |
f21e18ca N |
2994 | if (dl->major == (int)major(stb.st_rdev) && |
2995 | dl->minor == (int)minor(stb.st_rdev)) | |
a322f70c DW |
2996 | break; |
2997 | } | |
5f8097be | 2998 | if (!dl) { |
2c514b71 | 2999 | if (verbose) |
e7b84f9d | 3000 | pr_err("ddf: %s is not in the " |
613b0d17 N |
3001 | "same DDF set\n", |
3002 | dev); | |
5f8097be NB |
3003 | return 0; |
3004 | } | |
3005 | e = get_extents(ddf, dl); | |
3006 | maxsize = 0; | |
3007 | i = 0; | |
3008 | if (e) do { | |
613b0d17 N |
3009 | unsigned long long esize; |
3010 | esize = e[i].start - pos; | |
3011 | if (esize >= maxsize) | |
3012 | maxsize = esize; | |
3013 | pos = e[i].start + e[i].size; | |
3014 | i++; | |
3015 | } while (e[i-1].size); | |
5f8097be | 3016 | *freesize = maxsize; |
a322f70c DW |
3017 | // FIXME here I am |
3018 | ||
3019 | return 1; | |
3020 | } | |
59e36268 | 3021 | |
a322f70c | 3022 | static int load_super_ddf_all(struct supertype *st, int fd, |
e1902a7b | 3023 | void **sbp, char *devname) |
a322f70c DW |
3024 | { |
3025 | struct mdinfo *sra; | |
3026 | struct ddf_super *super; | |
3027 | struct mdinfo *sd, *best = NULL; | |
3028 | int bestseq = 0; | |
3029 | int seq; | |
3030 | char nm[20]; | |
3031 | int dfd; | |
3032 | ||
b526e52d | 3033 | sra = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE); |
a322f70c DW |
3034 | if (!sra) |
3035 | return 1; | |
3036 | if (sra->array.major_version != -1 || | |
3037 | sra->array.minor_version != -2 || | |
3038 | strcmp(sra->text_version, "ddf") != 0) | |
3039 | return 1; | |
3040 | ||
6416d527 | 3041 | if (posix_memalign((void**)&super, 512, sizeof(*super)) != 0) |
a322f70c | 3042 | return 1; |
a2349791 | 3043 | memset(super, 0, sizeof(*super)); |
a322f70c DW |
3044 | |
3045 | /* first, try each device, and choose the best ddf */ | |
3046 | for (sd = sra->devs ; sd ; sd = sd->next) { | |
3047 | int rv; | |
3048 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
7a7cc504 NB |
3049 | dfd = dev_open(nm, O_RDONLY); |
3050 | if (dfd < 0) | |
a322f70c DW |
3051 | return 2; |
3052 | rv = load_ddf_headers(dfd, super, NULL); | |
7a7cc504 | 3053 | close(dfd); |
a322f70c DW |
3054 | if (rv == 0) { |
3055 | seq = __be32_to_cpu(super->active->seq); | |
3056 | if (super->active->openflag) | |
3057 | seq--; | |
3058 | if (!best || seq > bestseq) { | |
3059 | bestseq = seq; | |
3060 | best = sd; | |
3061 | } | |
3062 | } | |
3063 | } | |
3064 | if (!best) | |
3065 | return 1; | |
3066 | /* OK, load this ddf */ | |
3067 | sprintf(nm, "%d:%d", best->disk.major, best->disk.minor); | |
3068 | dfd = dev_open(nm, O_RDONLY); | |
7a7cc504 | 3069 | if (dfd < 0) |
a322f70c DW |
3070 | return 1; |
3071 | load_ddf_headers(dfd, super, NULL); | |
3072 | load_ddf_global(dfd, super, NULL); | |
3073 | close(dfd); | |
3074 | /* Now we need the device-local bits */ | |
3075 | for (sd = sra->devs ; sd ; sd = sd->next) { | |
3d2c4fc7 DW |
3076 | int rv; |
3077 | ||
a322f70c | 3078 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); |
e1902a7b | 3079 | dfd = dev_open(nm, O_RDWR); |
7a7cc504 | 3080 | if (dfd < 0) |
a322f70c | 3081 | return 2; |
3d2c4fc7 DW |
3082 | rv = load_ddf_headers(dfd, super, NULL); |
3083 | if (rv == 0) | |
e1902a7b | 3084 | rv = load_ddf_local(dfd, super, NULL, 1); |
3d2c4fc7 DW |
3085 | if (rv) |
3086 | return 1; | |
a322f70c | 3087 | } |
33414a01 | 3088 | |
a322f70c DW |
3089 | *sbp = super; |
3090 | if (st->ss == NULL) { | |
78e44928 | 3091 | st->ss = &super_ddf; |
a322f70c DW |
3092 | st->minor_version = 0; |
3093 | st->max_devs = 512; | |
3094 | } | |
4dd2df09 | 3095 | strcpy(st->container_devnm, fd2devnm(fd)); |
a322f70c DW |
3096 | return 0; |
3097 | } | |
2b959fbf N |
3098 | |
3099 | static int load_container_ddf(struct supertype *st, int fd, | |
3100 | char *devname) | |
3101 | { | |
3102 | return load_super_ddf_all(st, fd, &st->sb, devname); | |
3103 | } | |
3104 | ||
0e600426 | 3105 | #endif /* MDASSEMBLE */ |
a322f70c | 3106 | |
a5c7adb3 | 3107 | static int check_secondary(const struct vcl *vc) |
3108 | { | |
3109 | const struct vd_config *conf = &vc->conf; | |
3110 | int i; | |
3111 | ||
3112 | /* The only DDF secondary RAID level md can support is | |
3113 | * RAID 10, if the stripe sizes and Basic volume sizes | |
3114 | * are all equal. | |
3115 | * Other configurations could in theory be supported by exposing | |
3116 | * the BVDs to user space and using device mapper for the secondary | |
3117 | * mapping. So far we don't support that. | |
3118 | */ | |
3119 | ||
3120 | __u64 sec_elements[4] = {0, 0, 0, 0}; | |
3121 | #define __set_sec_seen(n) (sec_elements[(n)>>6] |= (1<<((n)&63))) | |
3122 | #define __was_sec_seen(n) ((sec_elements[(n)>>6] & (1<<((n)&63))) != 0) | |
3123 | ||
3124 | if (vc->other_bvds == NULL) { | |
3125 | pr_err("No BVDs for secondary RAID found\n"); | |
3126 | return -1; | |
3127 | } | |
3128 | if (conf->prl != DDF_RAID1) { | |
3129 | pr_err("Secondary RAID level only supported for mirrored BVD\n"); | |
3130 | return -1; | |
3131 | } | |
3132 | if (conf->srl != DDF_2STRIPED && conf->srl != DDF_2SPANNED) { | |
3133 | pr_err("Secondary RAID level %d is unsupported\n", | |
3134 | conf->srl); | |
3135 | return -1; | |
3136 | } | |
3137 | __set_sec_seen(conf->sec_elmnt_seq); | |
3138 | for (i = 0; i < conf->sec_elmnt_count-1; i++) { | |
3139 | const struct vd_config *bvd = vc->other_bvds[i]; | |
3140 | if (bvd == NULL) { | |
e3c2a365 | 3141 | pr_err("BVD %d is missing\n", i+1); |
a5c7adb3 | 3142 | return -1; |
3143 | } | |
3144 | if (bvd->srl != conf->srl) { | |
3145 | pr_err("Inconsistent secondary RAID level across BVDs\n"); | |
3146 | return -1; | |
3147 | } | |
3148 | if (bvd->prl != conf->prl) { | |
3149 | pr_err("Different RAID levels for BVDs are unsupported\n"); | |
3150 | return -1; | |
3151 | } | |
3152 | if (bvd->prim_elmnt_count != conf->prim_elmnt_count) { | |
3153 | pr_err("All BVDs must have the same number of primary elements\n"); | |
3154 | return -1; | |
3155 | } | |
3156 | if (bvd->chunk_shift != conf->chunk_shift) { | |
3157 | pr_err("Different strip sizes for BVDs are unsupported\n"); | |
3158 | return -1; | |
3159 | } | |
3160 | if (bvd->array_blocks != conf->array_blocks) { | |
3161 | pr_err("Different BVD sizes are unsupported\n"); | |
3162 | return -1; | |
3163 | } | |
3164 | __set_sec_seen(bvd->sec_elmnt_seq); | |
3165 | } | |
3166 | for (i = 0; i < conf->sec_elmnt_count; i++) { | |
3167 | if (!__was_sec_seen(i)) { | |
3168 | pr_err("BVD %d is missing\n", i); | |
3169 | return -1; | |
3170 | } | |
3171 | } | |
3172 | return 0; | |
3173 | } | |
3174 | ||
8a38db86 | 3175 | #define NO_SUCH_REFNUM (0xFFFFFFFF) |
3176 | static unsigned int get_pd_index_from_refnum(const struct vcl *vc, | |
4e587018 | 3177 | __u32 refnum, unsigned int nmax, |
3178 | const struct vd_config **bvd, | |
3179 | unsigned int *idx) | |
8a38db86 | 3180 | { |
4e587018 | 3181 | unsigned int i, j, n, sec, cnt; |
3182 | ||
3183 | cnt = __be16_to_cpu(vc->conf.prim_elmnt_count); | |
3184 | sec = (vc->conf.sec_elmnt_count == 1 ? 0 : vc->conf.sec_elmnt_seq); | |
3185 | ||
3186 | for (i = 0, j = 0 ; i < nmax ; i++) { | |
3187 | /* j counts valid entries for this BVD */ | |
3188 | if (vc->conf.phys_refnum[i] != 0xffffffff) | |
3189 | j++; | |
3190 | if (vc->conf.phys_refnum[i] == refnum) { | |
3191 | *bvd = &vc->conf; | |
3192 | *idx = i; | |
3193 | return sec * cnt + j - 1; | |
3194 | } | |
3195 | } | |
3196 | if (vc->other_bvds == NULL) | |
3197 | goto bad; | |
3198 | ||
3199 | for (n = 1; n < vc->conf.sec_elmnt_count; n++) { | |
3200 | struct vd_config *vd = vc->other_bvds[n-1]; | |
3201 | if (vd == NULL) | |
3202 | continue; | |
3203 | sec = vd->sec_elmnt_seq; | |
3204 | for (i = 0, j = 0 ; i < nmax ; i++) { | |
3205 | if (vd->phys_refnum[i] != 0xffffffff) | |
3206 | j++; | |
3207 | if (vd->phys_refnum[i] == refnum) { | |
3208 | *bvd = vd; | |
3209 | *idx = i; | |
3210 | return sec * cnt + j - 1; | |
3211 | } | |
3212 | } | |
3213 | } | |
3214 | bad: | |
3215 | *bvd = NULL; | |
8a38db86 | 3216 | return NO_SUCH_REFNUM; |
3217 | } | |
3218 | ||
00bbdbda | 3219 | static struct mdinfo *container_content_ddf(struct supertype *st, char *subarray) |
598f0d58 NB |
3220 | { |
3221 | /* Given a container loaded by load_super_ddf_all, | |
3222 | * extract information about all the arrays into | |
3223 | * an mdinfo tree. | |
3224 | * | |
3225 | * For each vcl in conflist: create an mdinfo, fill it in, | |
3226 | * then look for matching devices (phys_refnum) in dlist | |
3227 | * and create appropriate device mdinfo. | |
3228 | */ | |
3229 | struct ddf_super *ddf = st->sb; | |
3230 | struct mdinfo *rest = NULL; | |
3231 | struct vcl *vc; | |
3232 | ||
3233 | for (vc = ddf->conflist ; vc ; vc=vc->next) | |
3234 | { | |
f21e18ca N |
3235 | unsigned int i; |
3236 | unsigned int j; | |
598f0d58 | 3237 | struct mdinfo *this; |
00bbdbda | 3238 | char *ep; |
90fa1a29 | 3239 | __u32 *cptr; |
8a38db86 | 3240 | unsigned int pd; |
00bbdbda N |
3241 | |
3242 | if (subarray && | |
3243 | (strtoul(subarray, &ep, 10) != vc->vcnum || | |
3244 | *ep != '\0')) | |
3245 | continue; | |
3246 | ||
a5c7adb3 | 3247 | if (vc->conf.sec_elmnt_count > 1) { |
3248 | if (check_secondary(vc) != 0) | |
3249 | continue; | |
3250 | } | |
3251 | ||
503975b9 | 3252 | this = xcalloc(1, sizeof(*this)); |
598f0d58 NB |
3253 | this->next = rest; |
3254 | rest = this; | |
3255 | ||
4e587018 | 3256 | if (vc->conf.sec_elmnt_count == 1) { |
3257 | this->array.level = map_num1(ddf_level_num, | |
3258 | vc->conf.prl); | |
3259 | this->array.raid_disks = | |
3260 | __be16_to_cpu(vc->conf.prim_elmnt_count); | |
3261 | this->array.layout = | |
3262 | rlq_to_layout(vc->conf.rlq, vc->conf.prl, | |
3263 | this->array.raid_disks); | |
3264 | } else { | |
3265 | /* The only supported layout is RAID 10. | |
3266 | * Compatibility has been checked in check_secondary() | |
3267 | * above. | |
3268 | */ | |
3269 | this->array.level = 10; | |
3270 | this->array.raid_disks = | |
3271 | __be16_to_cpu(vc->conf.prim_elmnt_count) | |
3272 | * vc->conf.sec_elmnt_count; | |
3273 | this->array.layout = 0x100 | | |
3274 | __be16_to_cpu(vc->conf.prim_elmnt_count); | |
3275 | } | |
598f0d58 | 3276 | this->array.md_minor = -1; |
f35f2525 N |
3277 | this->array.major_version = -1; |
3278 | this->array.minor_version = -2; | |
90fa1a29 JS |
3279 | cptr = (__u32 *)(vc->conf.guid + 16); |
3280 | this->array.ctime = DECADE + __be32_to_cpu(*cptr); | |
598f0d58 NB |
3281 | this->array.utime = DECADE + |
3282 | __be32_to_cpu(vc->conf.timestamp); | |
3283 | this->array.chunk_size = 512 << vc->conf.chunk_shift; | |
3284 | ||
59e36268 | 3285 | i = vc->vcnum; |
7a7cc504 NB |
3286 | if ((ddf->virt->entries[i].state & DDF_state_inconsistent) || |
3287 | (ddf->virt->entries[i].init_state & DDF_initstate_mask) != | |
ed9d66aa | 3288 | DDF_init_full) { |
598f0d58 | 3289 | this->array.state = 0; |
ed9d66aa NB |
3290 | this->resync_start = 0; |
3291 | } else { | |
598f0d58 | 3292 | this->array.state = 1; |
b7528a20 | 3293 | this->resync_start = MaxSector; |
ed9d66aa | 3294 | } |
db42fa9b N |
3295 | memcpy(this->name, ddf->virt->entries[i].name, 16); |
3296 | this->name[16]=0; | |
3297 | for(j=0; j<16; j++) | |
3298 | if (this->name[j] == ' ') | |
3299 | this->name[j] = 0; | |
598f0d58 NB |
3300 | |
3301 | memset(this->uuid, 0, sizeof(this->uuid)); | |
3302 | this->component_size = __be64_to_cpu(vc->conf.blocks); | |
3303 | this->array.size = this->component_size / 2; | |
5f2aace8 | 3304 | this->container_member = i; |
598f0d58 | 3305 | |
c5afc314 N |
3306 | ddf->currentconf = vc; |
3307 | uuid_from_super_ddf(st, this->uuid); | |
3308 | ddf->currentconf = NULL; | |
3309 | ||
60f18132 | 3310 | sprintf(this->text_version, "/%s/%d", |
4dd2df09 | 3311 | st->container_devnm, this->container_member); |
60f18132 | 3312 | |
8a38db86 | 3313 | for (pd = 0; pd < __be16_to_cpu(ddf->phys->used_pdes); pd++) { |
598f0d58 NB |
3314 | struct mdinfo *dev; |
3315 | struct dl *d; | |
4e587018 | 3316 | const struct vd_config *bvd; |
3317 | unsigned int iphys; | |
3318 | __u64 *lba_offset; | |
fa033bec | 3319 | int stt; |
598f0d58 | 3320 | |
8a38db86 | 3321 | if (ddf->phys->entries[pd].refnum == 0xFFFFFFFF) |
bc17324f | 3322 | continue; |
0cf5ef67 N |
3323 | |
3324 | stt = __be16_to_cpu(ddf->phys->entries[pd].state); | |
fa033bec N |
3325 | if ((stt & (DDF_Online|DDF_Failed|DDF_Rebuilding)) |
3326 | != DDF_Online) | |
3327 | continue; | |
3328 | ||
8a38db86 | 3329 | i = get_pd_index_from_refnum( |
4e587018 | 3330 | vc, ddf->phys->entries[pd].refnum, |
3331 | ddf->mppe, &bvd, &iphys); | |
8a38db86 | 3332 | if (i == NO_SUCH_REFNUM) |
3333 | continue; | |
3334 | ||
fa033bec | 3335 | this->array.working_disks++; |
bc17324f | 3336 | |
0cf5ef67 | 3337 | for (d = ddf->dlist; d ; d=d->next) |
8a38db86 | 3338 | if (d->disk.refnum == |
3339 | ddf->phys->entries[pd].refnum) | |
0cf5ef67 N |
3340 | break; |
3341 | if (d == NULL) | |
3342 | /* Haven't found that one yet, maybe there are others */ | |
3343 | continue; | |
3344 | ||
503975b9 | 3345 | dev = xcalloc(1, sizeof(*dev)); |
598f0d58 NB |
3346 | dev->next = this->devs; |
3347 | this->devs = dev; | |
3348 | ||
3349 | dev->disk.number = __be32_to_cpu(d->disk.refnum); | |
3350 | dev->disk.major = d->major; | |
3351 | dev->disk.minor = d->minor; | |
3352 | dev->disk.raid_disk = i; | |
3353 | dev->disk.state = (1<<MD_DISK_SYNC)|(1<<MD_DISK_ACTIVE); | |
d23534e4 | 3354 | dev->recovery_start = MaxSector; |
598f0d58 | 3355 | |
120f7677 | 3356 | dev->events = __be32_to_cpu(ddf->primary.seq); |
4e587018 | 3357 | lba_offset = (__u64 *)&bvd->phys_refnum[ddf->mppe]; |
3358 | dev->data_offset = __be64_to_cpu(lba_offset[iphys]); | |
3359 | dev->component_size = __be64_to_cpu(bvd->blocks); | |
598f0d58 NB |
3360 | if (d->devname) |
3361 | strcpy(dev->name, d->devname); | |
3362 | } | |
3363 | } | |
3364 | return rest; | |
3365 | } | |
3366 | ||
955e9ea1 | 3367 | static int store_super_ddf(struct supertype *st, int fd) |
a322f70c | 3368 | { |
955e9ea1 | 3369 | struct ddf_super *ddf = st->sb; |
a322f70c | 3370 | unsigned long long dsize; |
6416d527 | 3371 | void *buf; |
3d2c4fc7 | 3372 | int rc; |
a322f70c | 3373 | |
955e9ea1 DW |
3374 | if (!ddf) |
3375 | return 1; | |
3376 | ||
3377 | /* ->dlist and ->conflist will be set for updates, currently not | |
3378 | * supported | |
3379 | */ | |
3380 | if (ddf->dlist || ddf->conflist) | |
3381 | return 1; | |
3382 | ||
a322f70c DW |
3383 | if (!get_dev_size(fd, NULL, &dsize)) |
3384 | return 1; | |
3385 | ||
3d2c4fc7 DW |
3386 | if (posix_memalign(&buf, 512, 512) != 0) |
3387 | return 1; | |
6416d527 NB |
3388 | memset(buf, 0, 512); |
3389 | ||
a322f70c | 3390 | lseek64(fd, dsize-512, 0); |
3d2c4fc7 | 3391 | rc = write(fd, buf, 512); |
6416d527 | 3392 | free(buf); |
3d2c4fc7 DW |
3393 | if (rc < 0) |
3394 | return 1; | |
a322f70c DW |
3395 | return 0; |
3396 | } | |
3397 | ||
a19c88b8 NB |
3398 | static int compare_super_ddf(struct supertype *st, struct supertype *tst) |
3399 | { | |
3400 | /* | |
3401 | * return: | |
3402 | * 0 same, or first was empty, and second was copied | |
3403 | * 1 second had wrong number | |
3404 | * 2 wrong uuid | |
3405 | * 3 wrong other info | |
3406 | */ | |
3407 | struct ddf_super *first = st->sb; | |
3408 | struct ddf_super *second = tst->sb; | |
4eefd651 | 3409 | struct dl *dl1, *dl2; |
3410 | struct vcl *vl1, *vl2; | |
2d210697 | 3411 | unsigned int max_vds, max_pds, pd, vd; |
a19c88b8 NB |
3412 | |
3413 | if (!first) { | |
3414 | st->sb = tst->sb; | |
3415 | tst->sb = NULL; | |
3416 | return 0; | |
3417 | } | |
3418 | ||
3419 | if (memcmp(first->anchor.guid, second->anchor.guid, DDF_GUID_LEN) != 0) | |
3420 | return 2; | |
3421 | ||
2d210697 | 3422 | if (first->anchor.seq != second->anchor.seq) { |
3423 | dprintf("%s: sequence number mismatch %u/%u\n", __func__, | |
3424 | __be32_to_cpu(first->anchor.seq), | |
3425 | __be32_to_cpu(second->anchor.seq)); | |
3426 | return 3; | |
3427 | } | |
3428 | if (first->max_part != second->max_part || | |
3429 | first->phys->used_pdes != second->phys->used_pdes || | |
3430 | first->virt->populated_vdes != second->virt->populated_vdes) { | |
3431 | dprintf("%s: PD/VD number mismatch\n", __func__); | |
3432 | return 3; | |
3433 | } | |
3434 | ||
3435 | max_pds = __be16_to_cpu(first->phys->used_pdes); | |
3436 | for (dl2 = second->dlist; dl2; dl2 = dl2->next) { | |
3437 | for (pd = 0; pd < max_pds; pd++) | |
3438 | if (first->phys->entries[pd].refnum == dl2->disk.refnum) | |
3439 | break; | |
3440 | if (pd == max_pds) { | |
3441 | dprintf("%s: no match for disk %08x\n", __func__, | |
3442 | __be32_to_cpu(dl2->disk.refnum)); | |
3443 | return 3; | |
3444 | } | |
3445 | } | |
3446 | ||
3447 | max_vds = __be16_to_cpu(first->active->max_vd_entries); | |
3448 | for (vl2 = second->conflist; vl2; vl2 = vl2->next) { | |
3449 | if (vl2->conf.magic != DDF_VD_CONF_MAGIC) | |
3450 | continue; | |
3451 | for (vd = 0; vd < max_vds; vd++) | |
3452 | if (!memcmp(first->virt->entries[vd].guid, | |
3453 | vl2->conf.guid, DDF_GUID_LEN)) | |
3454 | break; | |
3455 | if (vd == max_vds) { | |
3456 | dprintf("%s: no match for VD config\n", __func__); | |
3457 | return 3; | |
3458 | } | |
3459 | } | |
a19c88b8 | 3460 | /* FIXME should I look at anything else? */ |
2d210697 | 3461 | |
4eefd651 | 3462 | /* |
3463 | At this point we are fairly sure that the meta data matches. | |
3464 | But the new disk may contain additional local data. | |
3465 | Add it to the super block. | |
3466 | */ | |
3467 | for (vl2 = second->conflist; vl2; vl2 = vl2->next) { | |
3468 | for (vl1 = first->conflist; vl1; vl1 = vl1->next) | |
3469 | if (!memcmp(vl1->conf.guid, vl2->conf.guid, | |
3470 | DDF_GUID_LEN)) | |
3471 | break; | |
3472 | if (vl1) { | |
3473 | if (vl1->other_bvds != NULL && | |
3474 | vl1->conf.sec_elmnt_seq != | |
3475 | vl2->conf.sec_elmnt_seq) { | |
3476 | dprintf("%s: adding BVD %u\n", __func__, | |
3477 | vl2->conf.sec_elmnt_seq); | |
3478 | add_other_bvd(vl1, &vl2->conf, | |
3479 | first->conf_rec_len*512); | |
3480 | } | |
3481 | continue; | |
3482 | } | |
3483 | ||
3484 | if (posix_memalign((void **)&vl1, 512, | |
3485 | (first->conf_rec_len*512 + | |
3486 | offsetof(struct vcl, conf))) != 0) { | |
3487 | pr_err("%s could not allocate vcl buf\n", | |
3488 | __func__); | |
3489 | return 3; | |
3490 | } | |
3491 | ||
3492 | vl1->next = first->conflist; | |
3493 | vl1->block_sizes = NULL; | |
3494 | if (vl1->conf.sec_elmnt_count > 1) { | |
3495 | vl1->other_bvds = xcalloc(vl2->conf.sec_elmnt_count - 1, | |
3496 | sizeof(struct vd_config *)); | |
3497 | } else | |
3498 | vl1->other_bvds = NULL; | |
3499 | memcpy(&vl1->conf, &vl2->conf, first->conf_rec_len*512); | |
3500 | vl1->lba_offset = (__u64 *) | |
3501 | &vl1->conf.phys_refnum[first->mppe]; | |
3502 | for (vd = 0; vd < max_vds; vd++) | |
3503 | if (!memcmp(first->virt->entries[vd].guid, | |
3504 | vl1->conf.guid, DDF_GUID_LEN)) | |
3505 | break; | |
3506 | vl1->vcnum = vd; | |
3507 | dprintf("%s: added config for VD %u\n", __func__, vl1->vcnum); | |
3508 | first->conflist = vl1; | |
3509 | } | |
3510 | ||
3511 | for (dl2 = second->dlist; dl2; dl2 = dl2->next) { | |
3512 | for (dl1 = first->dlist; dl1; dl1 = dl1->next) | |
3513 | if (dl1->disk.refnum == dl2->disk.refnum) | |
3514 | break; | |
3515 | if (dl1) | |
3516 | continue; | |
3517 | ||
3518 | if (posix_memalign((void **)&dl1, 512, | |
3519 | sizeof(*dl1) + (first->max_part) * sizeof(dl1->vlist[0])) | |
3520 | != 0) { | |
3521 | pr_err("%s could not allocate disk info buffer\n", | |
3522 | __func__); | |
3523 | return 3; | |
3524 | } | |
3525 | memcpy(dl1, dl2, sizeof(*dl1)); | |
3526 | dl1->mdupdate = NULL; | |
3527 | dl1->next = first->dlist; | |
3528 | dl1->fd = -1; | |
3529 | for (pd = 0; pd < max_pds; pd++) | |
3530 | if (first->phys->entries[pd].refnum == dl1->disk.refnum) | |
3531 | break; | |
3532 | dl1->pdnum = pd; | |
3533 | if (dl2->spare) { | |
3534 | if (posix_memalign((void **)&dl1->spare, 512, | |
3535 | first->conf_rec_len*512) != 0) { | |
3536 | pr_err("%s could not allocate spare info buf\n", | |
3537 | __func__); | |
3538 | return 3; | |
3539 | } | |
3540 | memcpy(dl1->spare, dl2->spare, first->conf_rec_len*512); | |
3541 | } | |
3542 | for (vd = 0 ; vd < first->max_part ; vd++) { | |
3543 | if (!dl2->vlist[vd]) { | |
3544 | dl1->vlist[vd] = NULL; | |
3545 | continue; | |
3546 | } | |
3547 | for (vl1 = first->conflist; vl1; vl1 = vl1->next) { | |
3548 | if (!memcmp(vl1->conf.guid, | |
3549 | dl2->vlist[vd]->conf.guid, | |
3550 | DDF_GUID_LEN)) | |
3551 | break; | |
3552 | dl1->vlist[vd] = vl1; | |
3553 | } | |
3554 | } | |
3555 | first->dlist = dl1; | |
3556 | dprintf("%s: added disk %d: %08x\n", __func__, dl1->pdnum, | |
3557 | dl1->disk.refnum); | |
3558 | } | |
3559 | ||
a19c88b8 NB |
3560 | return 0; |
3561 | } | |
3562 | ||
0e600426 | 3563 | #ifndef MDASSEMBLE |
4e5528c6 NB |
3564 | /* |
3565 | * A new array 'a' has been started which claims to be instance 'inst' | |
3566 | * within container 'c'. | |
3567 | * We need to confirm that the array matches the metadata in 'c' so | |
3568 | * that we don't corrupt any metadata. | |
3569 | */ | |
cba0191b | 3570 | static int ddf_open_new(struct supertype *c, struct active_array *a, char *inst) |
549e9569 | 3571 | { |
2c514b71 | 3572 | dprintf("ddf: open_new %s\n", inst); |
cba0191b | 3573 | a->info.container_member = atoi(inst); |
549e9569 NB |
3574 | return 0; |
3575 | } | |
3576 | ||
4e5528c6 NB |
3577 | /* |
3578 | * The array 'a' is to be marked clean in the metadata. | |
ed9d66aa | 3579 | * If '->resync_start' is not ~(unsigned long long)0, then the array is only |
4e5528c6 NB |
3580 | * clean up to the point (in sectors). If that cannot be recorded in the |
3581 | * metadata, then leave it as dirty. | |
3582 | * | |
3583 | * For DDF, we need to clear the DDF_state_inconsistent bit in the | |
3584 | * !global! virtual_disk.virtual_entry structure. | |
3585 | */ | |
01f157d7 | 3586 | static int ddf_set_array_state(struct active_array *a, int consistent) |
549e9569 | 3587 | { |
4e5528c6 NB |
3588 | struct ddf_super *ddf = a->container->sb; |
3589 | int inst = a->info.container_member; | |
18a2f463 | 3590 | int old = ddf->virt->entries[inst].state; |
01f157d7 N |
3591 | if (consistent == 2) { |
3592 | /* Should check if a recovery should be started FIXME */ | |
3593 | consistent = 1; | |
b7941fd6 | 3594 | if (!is_resync_complete(&a->info)) |
01f157d7 N |
3595 | consistent = 0; |
3596 | } | |
ed9d66aa NB |
3597 | if (consistent) |
3598 | ddf->virt->entries[inst].state &= ~DDF_state_inconsistent; | |
3599 | else | |
4e5528c6 | 3600 | ddf->virt->entries[inst].state |= DDF_state_inconsistent; |
18a2f463 NB |
3601 | if (old != ddf->virt->entries[inst].state) |
3602 | ddf->updates_pending = 1; | |
3603 | ||
3604 | old = ddf->virt->entries[inst].init_state; | |
ed9d66aa | 3605 | ddf->virt->entries[inst].init_state &= ~DDF_initstate_mask; |
b7941fd6 | 3606 | if (is_resync_complete(&a->info)) |
ed9d66aa | 3607 | ddf->virt->entries[inst].init_state |= DDF_init_full; |
b7941fd6 | 3608 | else if (a->info.resync_start == 0) |
ed9d66aa | 3609 | ddf->virt->entries[inst].init_state |= DDF_init_not; |
4e5528c6 | 3610 | else |
ed9d66aa | 3611 | ddf->virt->entries[inst].init_state |= DDF_init_quick; |
18a2f463 NB |
3612 | if (old != ddf->virt->entries[inst].init_state) |
3613 | ddf->updates_pending = 1; | |
ed9d66aa | 3614 | |
2c514b71 | 3615 | dprintf("ddf mark %d %s %llu\n", inst, consistent?"clean":"dirty", |
b7941fd6 | 3616 | a->info.resync_start); |
01f157d7 | 3617 | return consistent; |
fd7cde1b DW |
3618 | } |
3619 | ||
e1316fab | 3620 | #define container_of(ptr, type, member) ({ \ |
5d500228 N |
3621 | const typeof( ((type *)0)->member ) *__mptr = (ptr); \ |
3622 | (type *)( (char *)__mptr - offsetof(type,member) );}) | |
7a7cc504 NB |
3623 | /* |
3624 | * The state of each disk is stored in the global phys_disk structure | |
3625 | * in phys_disk.entries[n].state. | |
3626 | * This makes various combinations awkward. | |
3627 | * - When a device fails in any array, it must be failed in all arrays | |
3628 | * that include a part of this device. | |
3629 | * - When a component is rebuilding, we cannot include it officially in the | |
3630 | * array unless this is the only array that uses the device. | |
3631 | * | |
3632 | * So: when transitioning: | |
3633 | * Online -> failed, just set failed flag. monitor will propagate | |
3634 | * spare -> online, the device might need to be added to the array. | |
3635 | * spare -> failed, just set failed. Don't worry if in array or not. | |
3636 | */ | |
8d45d196 | 3637 | static void ddf_set_disk(struct active_array *a, int n, int state) |
549e9569 | 3638 | { |
7a7cc504 | 3639 | struct ddf_super *ddf = a->container->sb; |
f21e18ca | 3640 | unsigned int inst = a->info.container_member; |
7a7cc504 NB |
3641 | struct vd_config *vc = find_vdcr(ddf, inst); |
3642 | int pd = find_phys(ddf, vc->phys_refnum[n]); | |
3643 | int i, st, working; | |
e1316fab N |
3644 | struct mdinfo *mdi; |
3645 | struct dl *dl; | |
7a7cc504 NB |
3646 | |
3647 | if (vc == NULL) { | |
2c514b71 | 3648 | dprintf("ddf: cannot find instance %d!!\n", inst); |
7a7cc504 NB |
3649 | return; |
3650 | } | |
e1316fab N |
3651 | /* Find the matching slot in 'info'. */ |
3652 | for (mdi = a->info.devs; mdi; mdi = mdi->next) | |
3653 | if (mdi->disk.raid_disk == n) | |
3654 | break; | |
3655 | if (!mdi) | |
3656 | return; | |
3657 | ||
3658 | /* and find the 'dl' entry corresponding to that. */ | |
3659 | for (dl = ddf->dlist; dl; dl = dl->next) | |
77632af9 N |
3660 | if (mdi->state_fd >= 0 && |
3661 | mdi->disk.major == dl->major && | |
e1316fab N |
3662 | mdi->disk.minor == dl->minor) |
3663 | break; | |
3664 | if (!dl) | |
3665 | return; | |
3666 | ||
3667 | if (pd < 0 || pd != dl->pdnum) { | |
3668 | /* disk doesn't currently exist or has changed. | |
3669 | * If it is now in_sync, insert it. */ | |
7a7cc504 | 3670 | if ((state & DS_INSYNC) && ! (state & DS_FAULTY)) { |
e1316fab N |
3671 | struct vcl *vcl; |
3672 | pd = dl->pdnum; | |
3673 | vc->phys_refnum[n] = dl->disk.refnum; | |
3674 | vcl = container_of(vc, struct vcl, conf); | |
3675 | vcl->lba_offset[n] = mdi->data_offset; | |
3676 | ddf->phys->entries[pd].type &= | |
3677 | ~__cpu_to_be16(DDF_Global_Spare); | |
3678 | ddf->phys->entries[pd].type |= | |
3679 | __cpu_to_be16(DDF_Active_in_VD); | |
3680 | ddf->updates_pending = 1; | |
7a7cc504 NB |
3681 | } |
3682 | } else { | |
18a2f463 | 3683 | int old = ddf->phys->entries[pd].state; |
7a7cc504 NB |
3684 | if (state & DS_FAULTY) |
3685 | ddf->phys->entries[pd].state |= __cpu_to_be16(DDF_Failed); | |
3686 | if (state & DS_INSYNC) { | |
3687 | ddf->phys->entries[pd].state |= __cpu_to_be16(DDF_Online); | |
3688 | ddf->phys->entries[pd].state &= __cpu_to_be16(~DDF_Rebuilding); | |
3689 | } | |
18a2f463 NB |
3690 | if (old != ddf->phys->entries[pd].state) |
3691 | ddf->updates_pending = 1; | |
7a7cc504 NB |
3692 | } |
3693 | ||
2c514b71 | 3694 | dprintf("ddf: set_disk %d to %x\n", n, state); |
7e1432fb | 3695 | |
7a7cc504 NB |
3696 | /* Now we need to check the state of the array and update |
3697 | * virtual_disk.entries[n].state. | |
3698 | * It needs to be one of "optimal", "degraded", "failed". | |
3699 | * I don't understand 'deleted' or 'missing'. | |
3700 | */ | |
3701 | working = 0; | |
3702 | for (i=0; i < a->info.array.raid_disks; i++) { | |
3703 | pd = find_phys(ddf, vc->phys_refnum[i]); | |
3704 | if (pd < 0) | |
3705 | continue; | |
57632f4a NB |
3706 | st = __be16_to_cpu(ddf->phys->entries[pd].state); |
3707 | if ((st & (DDF_Online|DDF_Failed|DDF_Rebuilding)) | |
7a7cc504 NB |
3708 | == DDF_Online) |
3709 | working++; | |
3710 | } | |
3711 | state = DDF_state_degraded; | |
3712 | if (working == a->info.array.raid_disks) | |
3713 | state = DDF_state_optimal; | |
3714 | else switch(vc->prl) { | |
613b0d17 N |
3715 | case DDF_RAID0: |
3716 | case DDF_CONCAT: | |
3717 | case DDF_JBOD: | |
7a7cc504 | 3718 | state = DDF_state_failed; |
613b0d17 N |
3719 | break; |
3720 | case DDF_RAID1: | |
3721 | if (working == 0) | |
3722 | state = DDF_state_failed; | |
3723 | else if (working == 2 && state == DDF_state_degraded) | |
3724 | state = DDF_state_part_optimal; | |
3725 | break; | |
3726 | case DDF_RAID4: | |
3727 | case DDF_RAID5: | |
3728 | if (working < a->info.array.raid_disks-1) | |
3729 | state = DDF_state_failed; | |
3730 | break; | |
3731 | case DDF_RAID6: | |
3732 | if (working < a->info.array.raid_disks-2) | |
3733 | state = DDF_state_failed; | |
3734 | else if (working == a->info.array.raid_disks-1) | |
3735 | state = DDF_state_part_optimal; | |
3736 | break; | |
3737 | } | |
7a7cc504 | 3738 | |
18a2f463 NB |
3739 | if (ddf->virt->entries[inst].state != |
3740 | ((ddf->virt->entries[inst].state & ~DDF_state_mask) | |
3741 | | state)) { | |
3742 | ||
3743 | ddf->virt->entries[inst].state = | |
3744 | (ddf->virt->entries[inst].state & ~DDF_state_mask) | |
3745 | | state; | |
3746 | ddf->updates_pending = 1; | |
3747 | } | |
7a7cc504 | 3748 | |
549e9569 NB |
3749 | } |
3750 | ||
2e735d19 | 3751 | static void ddf_sync_metadata(struct supertype *st) |
549e9569 | 3752 | { |
7a7cc504 NB |
3753 | |
3754 | /* | |
3755 | * Write all data to all devices. | |
3756 | * Later, we might be able to track whether only local changes | |
3757 | * have been made, or whether any global data has been changed, | |
3758 | * but ddf is sufficiently weird that it probably always | |
3759 | * changes global data .... | |
3760 | */ | |
18a2f463 NB |
3761 | struct ddf_super *ddf = st->sb; |
3762 | if (!ddf->updates_pending) | |
3763 | return; | |
3764 | ddf->updates_pending = 0; | |
1cc7f4fe | 3765 | __write_init_super_ddf(st); |
2c514b71 | 3766 | dprintf("ddf: sync_metadata\n"); |
549e9569 NB |
3767 | } |
3768 | ||
88c164f4 NB |
3769 | static void ddf_process_update(struct supertype *st, |
3770 | struct metadata_update *update) | |
3771 | { | |
3772 | /* Apply this update to the metadata. | |
3773 | * The first 4 bytes are a DDF_*_MAGIC which guides | |
3774 | * our actions. | |
3775 | * Possible update are: | |
3776 | * DDF_PHYS_RECORDS_MAGIC | |
4dd968cc N |
3777 | * Add a new physical device or remove an old one. |
3778 | * Changes to this record only happen implicitly. | |
88c164f4 NB |
3779 | * used_pdes is the device number. |
3780 | * DDF_VIRT_RECORDS_MAGIC | |
3781 | * Add a new VD. Possibly also change the 'access' bits. | |
3782 | * populated_vdes is the entry number. | |
3783 | * DDF_VD_CONF_MAGIC | |
3784 | * New or updated VD. the VIRT_RECORD must already | |
3785 | * exist. For an update, phys_refnum and lba_offset | |
3786 | * (at least) are updated, and the VD_CONF must | |
3787 | * be written to precisely those devices listed with | |
3788 | * a phys_refnum. | |
3789 | * DDF_SPARE_ASSIGN_MAGIC | |
3790 | * replacement Spare Assignment Record... but for which device? | |
3791 | * | |
3792 | * So, e.g.: | |
3793 | * - to create a new array, we send a VIRT_RECORD and | |
3794 | * a VD_CONF. Then assemble and start the array. | |
3795 | * - to activate a spare we send a VD_CONF to add the phys_refnum | |
3796 | * and offset. This will also mark the spare as active with | |
3797 | * a spare-assignment record. | |
3798 | */ | |
3799 | struct ddf_super *ddf = st->sb; | |
3800 | __u32 *magic = (__u32*)update->buf; | |
3801 | struct phys_disk *pd; | |
3802 | struct virtual_disk *vd; | |
3803 | struct vd_config *vc; | |
3804 | struct vcl *vcl; | |
3805 | struct dl *dl; | |
f21e18ca N |
3806 | unsigned int mppe; |
3807 | unsigned int ent; | |
c7079c84 | 3808 | unsigned int pdnum, pd2; |
88c164f4 | 3809 | |
2c514b71 | 3810 | dprintf("Process update %x\n", *magic); |
7e1432fb | 3811 | |
88c164f4 NB |
3812 | switch (*magic) { |
3813 | case DDF_PHYS_RECORDS_MAGIC: | |
3814 | ||
3815 | if (update->len != (sizeof(struct phys_disk) + | |
3816 | sizeof(struct phys_disk_entry))) | |
3817 | return; | |
3818 | pd = (struct phys_disk*)update->buf; | |
3819 | ||
3820 | ent = __be16_to_cpu(pd->used_pdes); | |
3821 | if (ent >= __be16_to_cpu(ddf->phys->max_pdes)) | |
3822 | return; | |
4dd968cc N |
3823 | if (pd->entries[0].state & __cpu_to_be16(DDF_Missing)) { |
3824 | struct dl **dlp; | |
3825 | /* removing this disk. */ | |
3826 | ddf->phys->entries[ent].state |= __cpu_to_be16(DDF_Missing); | |
3827 | for (dlp = &ddf->dlist; *dlp; dlp = &(*dlp)->next) { | |
3828 | struct dl *dl = *dlp; | |
3829 | if (dl->pdnum == (signed)ent) { | |
3830 | close(dl->fd); | |
3831 | dl->fd = -1; | |
3832 | /* FIXME this doesn't free | |
3833 | * dl->devname */ | |
3834 | update->space = dl; | |
3835 | *dlp = dl->next; | |
3836 | break; | |
3837 | } | |
3838 | } | |
3839 | ddf->updates_pending = 1; | |
3840 | return; | |
3841 | } | |
88c164f4 NB |
3842 | if (!all_ff(ddf->phys->entries[ent].guid)) |
3843 | return; | |
3844 | ddf->phys->entries[ent] = pd->entries[0]; | |
3845 | ddf->phys->used_pdes = __cpu_to_be16(1 + | |
613b0d17 | 3846 | __be16_to_cpu(ddf->phys->used_pdes)); |
18a2f463 | 3847 | ddf->updates_pending = 1; |
2cc2983d N |
3848 | if (ddf->add_list) { |
3849 | struct active_array *a; | |
3850 | struct dl *al = ddf->add_list; | |
3851 | ddf->add_list = al->next; | |
3852 | ||
3853 | al->next = ddf->dlist; | |
3854 | ddf->dlist = al; | |
3855 | ||
3856 | /* As a device has been added, we should check | |
3857 | * for any degraded devices that might make | |
3858 | * use of this spare */ | |
3859 | for (a = st->arrays ; a; a=a->next) | |
3860 | a->check_degraded = 1; | |
3861 | } | |
88c164f4 NB |
3862 | break; |
3863 | ||
3864 | case DDF_VIRT_RECORDS_MAGIC: | |
3865 | ||
3866 | if (update->len != (sizeof(struct virtual_disk) + | |
3867 | sizeof(struct virtual_entry))) | |
3868 | return; | |
3869 | vd = (struct virtual_disk*)update->buf; | |
3870 | ||
3871 | ent = __be16_to_cpu(vd->populated_vdes); | |
3872 | if (ent >= __be16_to_cpu(ddf->virt->max_vdes)) | |
3873 | return; | |
3874 | if (!all_ff(ddf->virt->entries[ent].guid)) | |
3875 | return; | |
3876 | ddf->virt->entries[ent] = vd->entries[0]; | |
3877 | ddf->virt->populated_vdes = __cpu_to_be16(1 + | |
613b0d17 | 3878 | __be16_to_cpu(ddf->virt->populated_vdes)); |
18a2f463 | 3879 | ddf->updates_pending = 1; |
88c164f4 NB |
3880 | break; |
3881 | ||
3882 | case DDF_VD_CONF_MAGIC: | |
2c514b71 | 3883 | dprintf("len %d %d\n", update->len, ddf->conf_rec_len); |
88c164f4 NB |
3884 | |
3885 | mppe = __be16_to_cpu(ddf->anchor.max_primary_element_entries); | |
f21e18ca | 3886 | if ((unsigned)update->len != ddf->conf_rec_len * 512) |
88c164f4 NB |
3887 | return; |
3888 | vc = (struct vd_config*)update->buf; | |
3889 | for (vcl = ddf->conflist; vcl ; vcl = vcl->next) | |
3890 | if (memcmp(vcl->conf.guid, vc->guid, DDF_GUID_LEN) == 0) | |
3891 | break; | |
2c514b71 | 3892 | dprintf("vcl = %p\n", vcl); |
88c164f4 NB |
3893 | if (vcl) { |
3894 | /* An update, just copy the phys_refnum and lba_offset | |
3895 | * fields | |
3896 | */ | |
3897 | memcpy(vcl->conf.phys_refnum, vc->phys_refnum, | |
3898 | mppe * (sizeof(__u32) + sizeof(__u64))); | |
3899 | } else { | |
3900 | /* A new VD_CONF */ | |
e6b9548d DW |
3901 | if (!update->space) |
3902 | return; | |
88c164f4 NB |
3903 | vcl = update->space; |
3904 | update->space = NULL; | |
3905 | vcl->next = ddf->conflist; | |
edd8d13c | 3906 | memcpy(&vcl->conf, vc, update->len); |
88c164f4 NB |
3907 | vcl->lba_offset = (__u64*) |
3908 | &vcl->conf.phys_refnum[mppe]; | |
1502a43a N |
3909 | for (ent = 0; |
3910 | ent < __be16_to_cpu(ddf->virt->populated_vdes); | |
3911 | ent++) | |
3912 | if (memcmp(vc->guid, ddf->virt->entries[ent].guid, | |
3913 | DDF_GUID_LEN) == 0) { | |
3914 | vcl->vcnum = ent; | |
3915 | break; | |
3916 | } | |
88c164f4 NB |
3917 | ddf->conflist = vcl; |
3918 | } | |
c7079c84 N |
3919 | /* Set DDF_Transition on all Failed devices - to help |
3920 | * us detect those that are no longer in use | |
3921 | */ | |
3922 | for (pdnum = 0; pdnum < __be16_to_cpu(ddf->phys->used_pdes); pdnum++) | |
3923 | if (ddf->phys->entries[pdnum].state | |
3924 | & __be16_to_cpu(DDF_Failed)) | |
3925 | ddf->phys->entries[pdnum].state | |
3926 | |= __be16_to_cpu(DDF_Transition); | |
88c164f4 NB |
3927 | /* Now make sure vlist is correct for each dl. */ |
3928 | for (dl = ddf->dlist; dl; dl = dl->next) { | |
f21e18ca N |
3929 | unsigned int dn; |
3930 | unsigned int vn = 0; | |
8401644c | 3931 | int in_degraded = 0; |
88c164f4 NB |
3932 | for (vcl = ddf->conflist; vcl ; vcl = vcl->next) |
3933 | for (dn=0; dn < ddf->mppe ; dn++) | |
3934 | if (vcl->conf.phys_refnum[dn] == | |
3935 | dl->disk.refnum) { | |
8401644c | 3936 | int vstate; |
2c514b71 NB |
3937 | dprintf("dev %d has %p at %d\n", |
3938 | dl->pdnum, vcl, vn); | |
c7079c84 N |
3939 | /* Clear the Transition flag */ |
3940 | if (ddf->phys->entries[dl->pdnum].state | |
3941 | & __be16_to_cpu(DDF_Failed)) | |
3942 | ddf->phys->entries[dl->pdnum].state &= | |
3943 | ~__be16_to_cpu(DDF_Transition); | |
3944 | ||
88c164f4 | 3945 | dl->vlist[vn++] = vcl; |
8401644c N |
3946 | vstate = ddf->virt->entries[vcl->vcnum].state |
3947 | & DDF_state_mask; | |
3948 | if (vstate == DDF_state_degraded || | |
3949 | vstate == DDF_state_part_optimal) | |
3950 | in_degraded = 1; | |
88c164f4 NB |
3951 | break; |
3952 | } | |
3953 | while (vn < ddf->max_part) | |
3954 | dl->vlist[vn++] = NULL; | |
7e1432fb NB |
3955 | if (dl->vlist[0]) { |
3956 | ddf->phys->entries[dl->pdnum].type &= | |
3957 | ~__cpu_to_be16(DDF_Global_Spare); | |
8401644c N |
3958 | if (!(ddf->phys->entries[dl->pdnum].type & |
3959 | __cpu_to_be16(DDF_Active_in_VD))) { | |
613b0d17 N |
3960 | ddf->phys->entries[dl->pdnum].type |= |
3961 | __cpu_to_be16(DDF_Active_in_VD); | |
3962 | if (in_degraded) | |
3963 | ddf->phys->entries[dl->pdnum].state |= | |
3964 | __cpu_to_be16(DDF_Rebuilding); | |
3965 | } | |
7e1432fb NB |
3966 | } |
3967 | if (dl->spare) { | |
3968 | ddf->phys->entries[dl->pdnum].type &= | |
3969 | ~__cpu_to_be16(DDF_Global_Spare); | |
3970 | ddf->phys->entries[dl->pdnum].type |= | |
3971 | __cpu_to_be16(DDF_Spare); | |
3972 | } | |
3973 | if (!dl->vlist[0] && !dl->spare) { | |
3974 | ddf->phys->entries[dl->pdnum].type |= | |
3975 | __cpu_to_be16(DDF_Global_Spare); | |
3976 | ddf->phys->entries[dl->pdnum].type &= | |
3977 | ~__cpu_to_be16(DDF_Spare | | |
3978 | DDF_Active_in_VD); | |
3979 | } | |
88c164f4 | 3980 | } |
c7079c84 N |
3981 | |
3982 | /* Now remove any 'Failed' devices that are not part | |
3983 | * of any VD. They will have the Transition flag set. | |
3984 | * Once done, we need to update all dl->pdnum numbers. | |
3985 | */ | |
3986 | pd2 = 0; | |
3987 | for (pdnum = 0; pdnum < __be16_to_cpu(ddf->phys->used_pdes); pdnum++) | |
3988 | if ((ddf->phys->entries[pdnum].state | |
3989 | & __be16_to_cpu(DDF_Failed)) | |
3990 | && (ddf->phys->entries[pdnum].state | |
3991 | & __be16_to_cpu(DDF_Transition))) | |
3992 | /* skip this one */; | |
3993 | else if (pdnum == pd2) | |
3994 | pd2++; | |
3995 | else { | |
3996 | ddf->phys->entries[pd2] = ddf->phys->entries[pdnum]; | |
3997 | for (dl = ddf->dlist; dl; dl = dl->next) | |
3998 | if (dl->pdnum == (int)pdnum) | |
3999 | dl->pdnum = pd2; | |
4000 | pd2++; | |
4001 | } | |
4002 | ddf->phys->used_pdes = __cpu_to_be16(pd2); | |
4003 | while (pd2 < pdnum) { | |
4004 | memset(ddf->phys->entries[pd2].guid, 0xff, DDF_GUID_LEN); | |
4005 | pd2++; | |
4006 | } | |
4007 | ||
18a2f463 | 4008 | ddf->updates_pending = 1; |
88c164f4 NB |
4009 | break; |
4010 | case DDF_SPARE_ASSIGN_MAGIC: | |
4011 | default: break; | |
4012 | } | |
4013 | } | |
4014 | ||
edd8d13c NB |
4015 | static void ddf_prepare_update(struct supertype *st, |
4016 | struct metadata_update *update) | |
4017 | { | |
4018 | /* This update arrived at managemon. | |
4019 | * We are about to pass it to monitor. | |
4020 | * If a malloc is needed, do it here. | |
4021 | */ | |
4022 | struct ddf_super *ddf = st->sb; | |
4023 | __u32 *magic = (__u32*)update->buf; | |
4024 | if (*magic == DDF_VD_CONF_MAGIC) | |
e6b9548d | 4025 | if (posix_memalign(&update->space, 512, |
613b0d17 N |
4026 | offsetof(struct vcl, conf) |
4027 | + ddf->conf_rec_len * 512) != 0) | |
e6b9548d | 4028 | update->space = NULL; |
edd8d13c NB |
4029 | } |
4030 | ||
7e1432fb NB |
4031 | /* |
4032 | * Check if the array 'a' is degraded but not failed. | |
4033 | * If it is, find as many spares as are available and needed and | |
4034 | * arrange for their inclusion. | |
4035 | * We only choose devices which are not already in the array, | |
4036 | * and prefer those with a spare-assignment to this array. | |
4037 | * otherwise we choose global spares - assuming always that | |
4038 | * there is enough room. | |
4039 | * For each spare that we assign, we return an 'mdinfo' which | |
4040 | * describes the position for the device in the array. | |
4041 | * We also add to 'updates' a DDF_VD_CONF_MAGIC update with | |
4042 | * the new phys_refnum and lba_offset values. | |
4043 | * | |
4044 | * Only worry about BVDs at the moment. | |
4045 | */ | |
4046 | static struct mdinfo *ddf_activate_spare(struct active_array *a, | |
4047 | struct metadata_update **updates) | |
4048 | { | |
4049 | int working = 0; | |
4050 | struct mdinfo *d; | |
4051 | struct ddf_super *ddf = a->container->sb; | |
4052 | int global_ok = 0; | |
4053 | struct mdinfo *rv = NULL; | |
4054 | struct mdinfo *di; | |
4055 | struct metadata_update *mu; | |
4056 | struct dl *dl; | |
4057 | int i; | |
4058 | struct vd_config *vc; | |
4059 | __u64 *lba; | |
4060 | ||
7e1432fb NB |
4061 | for (d = a->info.devs ; d ; d = d->next) { |
4062 | if ((d->curr_state & DS_FAULTY) && | |
613b0d17 | 4063 | d->state_fd >= 0) |
7e1432fb NB |
4064 | /* wait for Removal to happen */ |
4065 | return NULL; | |
4066 | if (d->state_fd >= 0) | |
4067 | working ++; | |
4068 | } | |
4069 | ||
2c514b71 NB |
4070 | dprintf("ddf_activate: working=%d (%d) level=%d\n", working, a->info.array.raid_disks, |
4071 | a->info.array.level); | |
7e1432fb NB |
4072 | if (working == a->info.array.raid_disks) |
4073 | return NULL; /* array not degraded */ | |
4074 | switch (a->info.array.level) { | |
4075 | case 1: | |
4076 | if (working == 0) | |
4077 | return NULL; /* failed */ | |
4078 | break; | |
4079 | case 4: | |
4080 | case 5: | |
4081 | if (working < a->info.array.raid_disks - 1) | |
4082 | return NULL; /* failed */ | |
4083 | break; | |
4084 | case 6: | |
4085 | if (working < a->info.array.raid_disks - 2) | |
4086 | return NULL; /* failed */ | |
4087 | break; | |
4088 | default: /* concat or stripe */ | |
4089 | return NULL; /* failed */ | |
4090 | } | |
4091 | ||
4092 | /* For each slot, if it is not working, find a spare */ | |
4093 | dl = ddf->dlist; | |
4094 | for (i = 0; i < a->info.array.raid_disks; i++) { | |
4095 | for (d = a->info.devs ; d ; d = d->next) | |
4096 | if (d->disk.raid_disk == i) | |
4097 | break; | |
2c514b71 | 4098 | dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0); |
7e1432fb NB |
4099 | if (d && (d->state_fd >= 0)) |
4100 | continue; | |
4101 | ||
4102 | /* OK, this device needs recovery. Find a spare */ | |
4103 | again: | |
4104 | for ( ; dl ; dl = dl->next) { | |
4105 | unsigned long long esize; | |
4106 | unsigned long long pos; | |
4107 | struct mdinfo *d2; | |
4108 | int is_global = 0; | |
4109 | int is_dedicated = 0; | |
4110 | struct extent *ex; | |
f21e18ca | 4111 | unsigned int j; |
7e1432fb NB |
4112 | /* If in this array, skip */ |
4113 | for (d2 = a->info.devs ; d2 ; d2 = d2->next) | |
7590d562 N |
4114 | if (d2->state_fd >= 0 && |
4115 | d2->disk.major == dl->major && | |
7e1432fb | 4116 | d2->disk.minor == dl->minor) { |
2c514b71 | 4117 | dprintf("%x:%x already in array\n", dl->major, dl->minor); |
7e1432fb NB |
4118 | break; |
4119 | } | |
4120 | if (d2) | |
4121 | continue; | |
4122 | if (ddf->phys->entries[dl->pdnum].type & | |
4123 | __cpu_to_be16(DDF_Spare)) { | |
4124 | /* Check spare assign record */ | |
4125 | if (dl->spare) { | |
4126 | if (dl->spare->type & DDF_spare_dedicated) { | |
4127 | /* check spare_ents for guid */ | |
4128 | for (j = 0 ; | |
4129 | j < __be16_to_cpu(dl->spare->populated); | |
4130 | j++) { | |
4131 | if (memcmp(dl->spare->spare_ents[j].guid, | |
4132 | ddf->virt->entries[a->info.container_member].guid, | |
4133 | DDF_GUID_LEN) == 0) | |
4134 | is_dedicated = 1; | |
4135 | } | |
4136 | } else | |
4137 | is_global = 1; | |
4138 | } | |
4139 | } else if (ddf->phys->entries[dl->pdnum].type & | |
4140 | __cpu_to_be16(DDF_Global_Spare)) { | |
4141 | is_global = 1; | |
e0e7aeaa N |
4142 | } else if (!(ddf->phys->entries[dl->pdnum].state & |
4143 | __cpu_to_be16(DDF_Failed))) { | |
4144 | /* we can possibly use some of this */ | |
4145 | is_global = 1; | |
7e1432fb NB |
4146 | } |
4147 | if ( ! (is_dedicated || | |
4148 | (is_global && global_ok))) { | |
2c514b71 | 4149 | dprintf("%x:%x not suitable: %d %d\n", dl->major, dl->minor, |
613b0d17 | 4150 | is_dedicated, is_global); |
7e1432fb NB |
4151 | continue; |
4152 | } | |
4153 | ||
4154 | /* We are allowed to use this device - is there space? | |
4155 | * We need a->info.component_size sectors */ | |
4156 | ex = get_extents(ddf, dl); | |
4157 | if (!ex) { | |
2c514b71 | 4158 | dprintf("cannot get extents\n"); |
7e1432fb NB |
4159 | continue; |
4160 | } | |
4161 | j = 0; pos = 0; | |
4162 | esize = 0; | |
4163 | ||
4164 | do { | |
4165 | esize = ex[j].start - pos; | |
4166 | if (esize >= a->info.component_size) | |
4167 | break; | |
e5cc7d46 N |
4168 | pos = ex[j].start + ex[j].size; |
4169 | j++; | |
4170 | } while (ex[j-1].size); | |
7e1432fb NB |
4171 | |
4172 | free(ex); | |
4173 | if (esize < a->info.component_size) { | |
e5cc7d46 N |
4174 | dprintf("%x:%x has no room: %llu %llu\n", |
4175 | dl->major, dl->minor, | |
2c514b71 | 4176 | esize, a->info.component_size); |
7e1432fb NB |
4177 | /* No room */ |
4178 | continue; | |
4179 | } | |
4180 | ||
4181 | /* Cool, we have a device with some space at pos */ | |
503975b9 | 4182 | di = xcalloc(1, sizeof(*di)); |
7e1432fb NB |
4183 | di->disk.number = i; |
4184 | di->disk.raid_disk = i; | |
4185 | di->disk.major = dl->major; | |
4186 | di->disk.minor = dl->minor; | |
4187 | di->disk.state = 0; | |
d23534e4 | 4188 | di->recovery_start = 0; |
7e1432fb NB |
4189 | di->data_offset = pos; |
4190 | di->component_size = a->info.component_size; | |
4191 | di->container_member = dl->pdnum; | |
4192 | di->next = rv; | |
4193 | rv = di; | |
2c514b71 NB |
4194 | dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor, |
4195 | i, pos); | |
7e1432fb NB |
4196 | |
4197 | break; | |
4198 | } | |
4199 | if (!dl && ! global_ok) { | |
4200 | /* not enough dedicated spares, try global */ | |
4201 | global_ok = 1; | |
4202 | dl = ddf->dlist; | |
4203 | goto again; | |
4204 | } | |
4205 | } | |
4206 | ||
4207 | if (!rv) | |
4208 | /* No spares found */ | |
4209 | return rv; | |
4210 | /* Now 'rv' has a list of devices to return. | |
4211 | * Create a metadata_update record to update the | |
4212 | * phys_refnum and lba_offset values | |
4213 | */ | |
503975b9 N |
4214 | mu = xmalloc(sizeof(*mu)); |
4215 | if (posix_memalign(&mu->space, 512, sizeof(struct vcl)) != 0) { | |
79244939 DW |
4216 | free(mu); |
4217 | mu = NULL; | |
4218 | } | |
503975b9 | 4219 | mu->buf = xmalloc(ddf->conf_rec_len * 512); |
7590d562 N |
4220 | mu->len = ddf->conf_rec_len * 512; |
4221 | mu->space = NULL; | |
f50ae22e | 4222 | mu->space_list = NULL; |
7e1432fb NB |
4223 | mu->next = *updates; |
4224 | vc = find_vdcr(ddf, a->info.container_member); | |
4225 | memcpy(mu->buf, vc, ddf->conf_rec_len * 512); | |
4226 | ||
4227 | vc = (struct vd_config*)mu->buf; | |
4228 | lba = (__u64*)&vc->phys_refnum[ddf->mppe]; | |
4229 | for (di = rv ; di ; di = di->next) { | |
4230 | vc->phys_refnum[di->disk.raid_disk] = | |
4231 | ddf->phys->entries[dl->pdnum].refnum; | |
4232 | lba[di->disk.raid_disk] = di->data_offset; | |
4233 | } | |
4234 | *updates = mu; | |
4235 | return rv; | |
4236 | } | |
0e600426 | 4237 | #endif /* MDASSEMBLE */ |
7e1432fb | 4238 | |
b640a252 N |
4239 | static int ddf_level_to_layout(int level) |
4240 | { | |
4241 | switch(level) { | |
4242 | case 0: | |
4243 | case 1: | |
4244 | return 0; | |
4245 | case 5: | |
4246 | return ALGORITHM_LEFT_SYMMETRIC; | |
4247 | case 6: | |
4248 | return ALGORITHM_ROTATING_N_CONTINUE; | |
4249 | case 10: | |
4250 | return 0x102; | |
4251 | default: | |
4252 | return UnSet; | |
4253 | } | |
4254 | } | |
4255 | ||
30f58b22 DW |
4256 | static void default_geometry_ddf(struct supertype *st, int *level, int *layout, int *chunk) |
4257 | { | |
4258 | if (level && *level == UnSet) | |
4259 | *level = LEVEL_CONTAINER; | |
4260 | ||
4261 | if (level && layout && *layout == UnSet) | |
4262 | *layout = ddf_level_to_layout(*level); | |
4263 | } | |
4264 | ||
a322f70c DW |
4265 | struct superswitch super_ddf = { |
4266 | #ifndef MDASSEMBLE | |
4267 | .examine_super = examine_super_ddf, | |
4268 | .brief_examine_super = brief_examine_super_ddf, | |
4737ae25 | 4269 | .brief_examine_subarrays = brief_examine_subarrays_ddf, |
bceedeec | 4270 | .export_examine_super = export_examine_super_ddf, |
a322f70c DW |
4271 | .detail_super = detail_super_ddf, |
4272 | .brief_detail_super = brief_detail_super_ddf, | |
4273 | .validate_geometry = validate_geometry_ddf, | |
78e44928 | 4274 | .write_init_super = write_init_super_ddf, |
0e600426 | 4275 | .add_to_super = add_to_super_ddf, |
4dd968cc | 4276 | .remove_from_super = remove_from_super_ddf, |
2b959fbf | 4277 | .load_container = load_container_ddf, |
a322f70c DW |
4278 | #endif |
4279 | .match_home = match_home_ddf, | |
4280 | .uuid_from_super= uuid_from_super_ddf, | |
4281 | .getinfo_super = getinfo_super_ddf, | |
4282 | .update_super = update_super_ddf, | |
4283 | ||
4284 | .avail_size = avail_size_ddf, | |
4285 | ||
a19c88b8 NB |
4286 | .compare_super = compare_super_ddf, |
4287 | ||
a322f70c | 4288 | .load_super = load_super_ddf, |
ba7eb04f | 4289 | .init_super = init_super_ddf, |
955e9ea1 | 4290 | .store_super = store_super_ddf, |
a322f70c DW |
4291 | .free_super = free_super_ddf, |
4292 | .match_metadata_desc = match_metadata_desc_ddf, | |
78e44928 | 4293 | .container_content = container_content_ddf, |
30f58b22 | 4294 | .default_geometry = default_geometry_ddf, |
a322f70c | 4295 | |
a322f70c | 4296 | .external = 1, |
549e9569 | 4297 | |
0e600426 | 4298 | #ifndef MDASSEMBLE |
549e9569 NB |
4299 | /* for mdmon */ |
4300 | .open_new = ddf_open_new, | |
ed9d66aa | 4301 | .set_array_state= ddf_set_array_state, |
549e9569 NB |
4302 | .set_disk = ddf_set_disk, |
4303 | .sync_metadata = ddf_sync_metadata, | |
88c164f4 | 4304 | .process_update = ddf_process_update, |
edd8d13c | 4305 | .prepare_update = ddf_prepare_update, |
7e1432fb | 4306 | .activate_spare = ddf_activate_spare, |
0e600426 | 4307 | #endif |
4cce4069 | 4308 | .name = "ddf", |
a322f70c | 4309 | }; |