]>
Commit | Line | Data |
---|---|---|
a322f70c DW |
1 | /* |
2 | * mdadm - manage Linux "md" devices aka RAID arrays. | |
3 | * | |
56cb05c4 | 4 | * Copyright (C) 2006-2014 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 | * | |
56cb05c4 | 24 | * Specifications for DDF taken from Common RAID DDF Specification Revision 1.2 |
a322f70c DW |
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> | |
04f903b2 | 33 | #include <stddef.h> |
a322f70c | 34 | |
a322f70c DW |
35 | /* a non-official T10 name for creation GUIDs */ |
36 | static char T10[] = "Linux-MD"; | |
37 | ||
38 | /* DDF timestamps are 1980 based, so we need to add | |
39 | * second-in-decade-of-seventies to convert to linux timestamps. | |
40 | * 10 years with 2 leap years. | |
41 | */ | |
42 | #define DECADE (3600*24*(365*10+2)) | |
43 | unsigned long crc32( | |
44 | unsigned long crc, | |
45 | const unsigned char *buf, | |
46 | unsigned len); | |
47 | ||
bedbf68a | 48 | #define DDF_NOTFOUND (~0U) |
49 | #define DDF_CONTAINER (DDF_NOTFOUND-1) | |
50 | ||
5684fff6 | 51 | /* Default for safe_mode_delay. Same value as for IMSM. |
52 | */ | |
53 | static const int DDF_SAFE_MODE_DELAY = 4000; | |
54 | ||
a322f70c DW |
55 | /* The DDF metadata handling. |
56 | * DDF metadata lives at the end of the device. | |
57 | * The last 512 byte block provides an 'anchor' which is used to locate | |
58 | * the rest of the metadata which usually lives immediately behind the anchor. | |
59 | * | |
60 | * Note: | |
61 | * - all multibyte numeric fields are bigendian. | |
62 | * - all strings are space padded. | |
63 | * | |
64 | */ | |
65 | ||
4d1bdc18 | 66 | typedef struct __be16 { |
67 | __u16 _v16; | |
68 | } be16; | |
69 | #define be16_eq(x, y) ((x)._v16 == (y)._v16) | |
a8173e43 | 70 | #define be16_and(x, y) ((x)._v16 & (y)._v16) |
71 | #define be16_or(x, y) ((x)._v16 | (y)._v16) | |
72 | #define be16_clear(x, y) ((x)._v16 &= ~(y)._v16) | |
73 | #define be16_set(x, y) ((x)._v16 |= (y)._v16) | |
4d1bdc18 | 74 | |
75 | typedef struct __be32 { | |
76 | __u32 _v32; | |
77 | } be32; | |
78 | #define be32_eq(x, y) ((x)._v32 == (y)._v32) | |
79 | ||
80 | typedef struct __be64 { | |
81 | __u64 _v64; | |
82 | } be64; | |
83 | #define be64_eq(x, y) ((x)._v64 == (y)._v64) | |
84 | ||
85 | #define be16_to_cpu(be) __be16_to_cpu((be)._v16) | |
86 | static inline be16 cpu_to_be16(__u16 x) | |
87 | { | |
88 | be16 be = { ._v16 = __cpu_to_be16(x) }; | |
89 | return be; | |
90 | } | |
91 | ||
92 | #define be32_to_cpu(be) __be32_to_cpu((be)._v32) | |
93 | static inline be32 cpu_to_be32(__u32 x) | |
94 | { | |
95 | be32 be = { ._v32 = __cpu_to_be32(x) }; | |
96 | return be; | |
97 | } | |
98 | ||
99 | #define be64_to_cpu(be) __be64_to_cpu((be)._v64) | |
100 | static inline be64 cpu_to_be64(__u64 x) | |
101 | { | |
102 | be64 be = { ._v64 = __cpu_to_be64(x) }; | |
103 | return be; | |
104 | } | |
105 | ||
a322f70c DW |
106 | /* Primary Raid Level (PRL) */ |
107 | #define DDF_RAID0 0x00 | |
108 | #define DDF_RAID1 0x01 | |
109 | #define DDF_RAID3 0x03 | |
110 | #define DDF_RAID4 0x04 | |
111 | #define DDF_RAID5 0x05 | |
112 | #define DDF_RAID1E 0x11 | |
113 | #define DDF_JBOD 0x0f | |
114 | #define DDF_CONCAT 0x1f | |
115 | #define DDF_RAID5E 0x15 | |
116 | #define DDF_RAID5EE 0x25 | |
59e36268 | 117 | #define DDF_RAID6 0x06 |
a322f70c DW |
118 | |
119 | /* Raid Level Qualifier (RLQ) */ | |
120 | #define DDF_RAID0_SIMPLE 0x00 | |
121 | #define DDF_RAID1_SIMPLE 0x00 /* just 2 devices in this plex */ | |
122 | #define DDF_RAID1_MULTI 0x01 /* exactly 3 devices in this plex */ | |
123 | #define DDF_RAID3_0 0x00 /* parity in first extent */ | |
124 | #define DDF_RAID3_N 0x01 /* parity in last extent */ | |
125 | #define DDF_RAID4_0 0x00 /* parity in first extent */ | |
126 | #define DDF_RAID4_N 0x01 /* parity in last extent */ | |
127 | /* these apply to raid5e and raid5ee as well */ | |
128 | #define DDF_RAID5_0_RESTART 0x00 /* same as 'right asymmetric' - layout 1 */ | |
59e36268 | 129 | #define DDF_RAID6_0_RESTART 0x01 /* raid6 different from raid5 here!!! */ |
a322f70c DW |
130 | #define DDF_RAID5_N_RESTART 0x02 /* same as 'left asymmetric' - layout 0 */ |
131 | #define DDF_RAID5_N_CONTINUE 0x03 /* same as 'left symmetric' - layout 2 */ | |
132 | ||
133 | #define DDF_RAID1E_ADJACENT 0x00 /* raid10 nearcopies==2 */ | |
134 | #define DDF_RAID1E_OFFSET 0x01 /* raid10 offsetcopies==2 */ | |
135 | ||
136 | /* Secondary RAID Level (SRL) */ | |
137 | #define DDF_2STRIPED 0x00 /* This is weirder than RAID0 !! */ | |
138 | #define DDF_2MIRRORED 0x01 | |
139 | #define DDF_2CONCAT 0x02 | |
140 | #define DDF_2SPANNED 0x03 /* This is also weird - be careful */ | |
141 | ||
142 | /* Magic numbers */ | |
60931cf9 | 143 | #define DDF_HEADER_MAGIC cpu_to_be32(0xDE11DE11) |
144 | #define DDF_CONTROLLER_MAGIC cpu_to_be32(0xAD111111) | |
145 | #define DDF_PHYS_RECORDS_MAGIC cpu_to_be32(0x22222222) | |
146 | #define DDF_PHYS_DATA_MAGIC cpu_to_be32(0x33333333) | |
147 | #define DDF_VIRT_RECORDS_MAGIC cpu_to_be32(0xDDDDDDDD) | |
148 | #define DDF_VD_CONF_MAGIC cpu_to_be32(0xEEEEEEEE) | |
149 | #define DDF_SPARE_ASSIGN_MAGIC cpu_to_be32(0x55555555) | |
150 | #define DDF_VU_CONF_MAGIC cpu_to_be32(0x88888888) | |
151 | #define DDF_VENDOR_LOG_MAGIC cpu_to_be32(0x01dBEEF0) | |
152 | #define DDF_BBM_LOG_MAGIC cpu_to_be32(0xABADB10C) | |
a322f70c DW |
153 | |
154 | #define DDF_GUID_LEN 24 | |
59e36268 NB |
155 | #define DDF_REVISION_0 "01.00.00" |
156 | #define DDF_REVISION_2 "01.02.00" | |
a322f70c DW |
157 | |
158 | struct ddf_header { | |
60931cf9 | 159 | be32 magic; /* DDF_HEADER_MAGIC */ |
160 | be32 crc; | |
a322f70c | 161 | char guid[DDF_GUID_LEN]; |
59e36268 | 162 | char revision[8]; /* 01.02.00 */ |
60931cf9 | 163 | be32 seq; /* starts at '1' */ |
164 | be32 timestamp; | |
a322f70c DW |
165 | __u8 openflag; |
166 | __u8 foreignflag; | |
167 | __u8 enforcegroups; | |
168 | __u8 pad0; /* 0xff */ | |
169 | __u8 pad1[12]; /* 12 * 0xff */ | |
170 | /* 64 bytes so far */ | |
171 | __u8 header_ext[32]; /* reserved: fill with 0xff */ | |
9d0c6b70 | 172 | be64 primary_lba; |
173 | be64 secondary_lba; | |
a322f70c DW |
174 | __u8 type; |
175 | __u8 pad2[3]; /* 0xff */ | |
60931cf9 | 176 | be32 workspace_len; /* sectors for vendor space - |
a322f70c | 177 | * at least 32768(sectors) */ |
9d0c6b70 | 178 | be64 workspace_lba; |
a8173e43 | 179 | be16 max_pd_entries; /* one of 15, 63, 255, 1023, 4095 */ |
180 | be16 max_vd_entries; /* 2^(4,6,8,10,12)-1 : i.e. as above */ | |
181 | be16 max_partitions; /* i.e. max num of configuration | |
a322f70c | 182 | record entries per disk */ |
a8173e43 | 183 | be16 config_record_len; /* 1 +ROUNDUP(max_primary_element_entries |
a322f70c | 184 | *12/512) */ |
a8173e43 | 185 | be16 max_primary_element_entries; /* 16, 64, 256, 1024, or 4096 */ |
a322f70c DW |
186 | __u8 pad3[54]; /* 0xff */ |
187 | /* 192 bytes so far */ | |
60931cf9 | 188 | be32 controller_section_offset; |
189 | be32 controller_section_length; | |
190 | be32 phys_section_offset; | |
191 | be32 phys_section_length; | |
192 | be32 virt_section_offset; | |
193 | be32 virt_section_length; | |
194 | be32 config_section_offset; | |
195 | be32 config_section_length; | |
196 | be32 data_section_offset; | |
197 | be32 data_section_length; | |
198 | be32 bbm_section_offset; | |
199 | be32 bbm_section_length; | |
200 | be32 diag_space_offset; | |
201 | be32 diag_space_length; | |
202 | be32 vendor_offset; | |
203 | be32 vendor_length; | |
a322f70c DW |
204 | /* 256 bytes so far */ |
205 | __u8 pad4[256]; /* 0xff */ | |
206 | }; | |
207 | ||
208 | /* type field */ | |
209 | #define DDF_HEADER_ANCHOR 0x00 | |
210 | #define DDF_HEADER_PRIMARY 0x01 | |
211 | #define DDF_HEADER_SECONDARY 0x02 | |
212 | ||
213 | /* The content of the 'controller section' - global scope */ | |
214 | struct ddf_controller_data { | |
60931cf9 | 215 | be32 magic; /* DDF_CONTROLLER_MAGIC */ |
216 | be32 crc; | |
a322f70c DW |
217 | char guid[DDF_GUID_LEN]; |
218 | struct controller_type { | |
a8173e43 | 219 | be16 vendor_id; |
220 | be16 device_id; | |
221 | be16 sub_vendor_id; | |
222 | be16 sub_device_id; | |
a322f70c DW |
223 | } type; |
224 | char product_id[16]; | |
225 | __u8 pad[8]; /* 0xff */ | |
226 | __u8 vendor_data[448]; | |
227 | }; | |
228 | ||
229 | /* The content of phys_section - global scope */ | |
230 | struct phys_disk { | |
60931cf9 | 231 | be32 magic; /* DDF_PHYS_RECORDS_MAGIC */ |
232 | be32 crc; | |
217dead4 N |
233 | be16 used_pdes; /* This is a counter, not a max - the list |
234 | * of used entries may not be dense */ | |
a8173e43 | 235 | be16 max_pdes; |
a322f70c DW |
236 | __u8 pad[52]; |
237 | struct phys_disk_entry { | |
238 | char guid[DDF_GUID_LEN]; | |
60931cf9 | 239 | be32 refnum; |
a8173e43 | 240 | be16 type; |
241 | be16 state; | |
56cb05c4 N |
242 | be64 config_size; /* DDF structures must be after here */ |
243 | char path[18]; /* Another horrible structure really | |
244 | * but is "used for information | |
245 | * purposes only" */ | |
a322f70c DW |
246 | __u8 pad[6]; |
247 | } entries[0]; | |
248 | }; | |
249 | ||
250 | /* phys_disk_entry.type is a bitmap - bigendian remember */ | |
251 | #define DDF_Forced_PD_GUID 1 | |
252 | #define DDF_Active_in_VD 2 | |
88c164f4 | 253 | #define DDF_Global_Spare 4 /* VD_CONF records are ignored */ |
a322f70c DW |
254 | #define DDF_Spare 8 /* overrides Global_spare */ |
255 | #define DDF_Foreign 16 | |
256 | #define DDF_Legacy 32 /* no DDF on this device */ | |
257 | ||
258 | #define DDF_Interface_mask 0xf00 | |
259 | #define DDF_Interface_SCSI 0x100 | |
260 | #define DDF_Interface_SAS 0x200 | |
261 | #define DDF_Interface_SATA 0x300 | |
262 | #define DDF_Interface_FC 0x400 | |
263 | ||
264 | /* phys_disk_entry.state is a bigendian bitmap */ | |
265 | #define DDF_Online 1 | |
266 | #define DDF_Failed 2 /* overrides 1,4,8 */ | |
267 | #define DDF_Rebuilding 4 | |
268 | #define DDF_Transition 8 | |
269 | #define DDF_SMART 16 | |
270 | #define DDF_ReadErrors 32 | |
271 | #define DDF_Missing 64 | |
272 | ||
273 | /* The content of the virt_section global scope */ | |
274 | struct virtual_disk { | |
60931cf9 | 275 | be32 magic; /* DDF_VIRT_RECORDS_MAGIC */ |
276 | be32 crc; | |
a8173e43 | 277 | be16 populated_vdes; |
278 | be16 max_vdes; | |
a322f70c DW |
279 | __u8 pad[52]; |
280 | struct virtual_entry { | |
281 | char guid[DDF_GUID_LEN]; | |
a8173e43 | 282 | be16 unit; |
a322f70c | 283 | __u16 pad0; /* 0xffff */ |
a8173e43 | 284 | be16 guid_crc; |
285 | be16 type; | |
a322f70c DW |
286 | __u8 state; |
287 | __u8 init_state; | |
288 | __u8 pad1[14]; | |
289 | char name[16]; | |
290 | } entries[0]; | |
291 | }; | |
292 | ||
293 | /* virtual_entry.type is a bitmap - bigendian */ | |
294 | #define DDF_Shared 1 | |
295 | #define DDF_Enforce_Groups 2 | |
296 | #define DDF_Unicode 4 | |
297 | #define DDF_Owner_Valid 8 | |
298 | ||
299 | /* virtual_entry.state is a bigendian bitmap */ | |
300 | #define DDF_state_mask 0x7 | |
301 | #define DDF_state_optimal 0x0 | |
302 | #define DDF_state_degraded 0x1 | |
303 | #define DDF_state_deleted 0x2 | |
304 | #define DDF_state_missing 0x3 | |
305 | #define DDF_state_failed 0x4 | |
7a7cc504 | 306 | #define DDF_state_part_optimal 0x5 |
a322f70c DW |
307 | |
308 | #define DDF_state_morphing 0x8 | |
309 | #define DDF_state_inconsistent 0x10 | |
310 | ||
311 | /* virtual_entry.init_state is a bigendian bitmap */ | |
312 | #define DDF_initstate_mask 0x03 | |
313 | #define DDF_init_not 0x00 | |
7a7cc504 NB |
314 | #define DDF_init_quick 0x01 /* initialisation is progress. |
315 | * i.e. 'state_inconsistent' */ | |
a322f70c DW |
316 | #define DDF_init_full 0x02 |
317 | ||
318 | #define DDF_access_mask 0xc0 | |
319 | #define DDF_access_rw 0x00 | |
320 | #define DDF_access_ro 0x80 | |
321 | #define DDF_access_blocked 0xc0 | |
322 | ||
323 | /* The content of the config_section - local scope | |
324 | * It has multiple records each config_record_len sectors | |
325 | * They can be vd_config or spare_assign | |
326 | */ | |
327 | ||
328 | struct vd_config { | |
60931cf9 | 329 | be32 magic; /* DDF_VD_CONF_MAGIC */ |
330 | be32 crc; | |
a322f70c | 331 | char guid[DDF_GUID_LEN]; |
60931cf9 | 332 | be32 timestamp; |
333 | be32 seqnum; | |
a322f70c | 334 | __u8 pad0[24]; |
a8173e43 | 335 | be16 prim_elmnt_count; |
a322f70c DW |
336 | __u8 chunk_shift; /* 0 == 512, 1==1024 etc */ |
337 | __u8 prl; | |
338 | __u8 rlq; | |
339 | __u8 sec_elmnt_count; | |
340 | __u8 sec_elmnt_seq; | |
341 | __u8 srl; | |
9d0c6b70 | 342 | be64 blocks; /* blocks per component could be different |
598f0d58 NB |
343 | * on different component devices...(only |
344 | * for concat I hope) */ | |
9d0c6b70 | 345 | be64 array_blocks; /* blocks in array */ |
a322f70c | 346 | __u8 pad1[8]; |
2770f45a N |
347 | be32 spare_refs[8]; /* This is used to detect missing spares. |
348 | * As we don't have an interface for that | |
349 | * the values are ignored. | |
350 | */ | |
a322f70c DW |
351 | __u8 cache_pol[8]; |
352 | __u8 bg_rate; | |
353 | __u8 pad2[3]; | |
354 | __u8 pad3[52]; | |
355 | __u8 pad4[192]; | |
356 | __u8 v0[32]; /* reserved- 0xff */ | |
357 | __u8 v1[32]; /* reserved- 0xff */ | |
358 | __u8 v2[16]; /* reserved- 0xff */ | |
359 | __u8 v3[16]; /* reserved- 0xff */ | |
360 | __u8 vendor[32]; | |
60931cf9 | 361 | be32 phys_refnum[0]; /* refnum of each disk in sequence */ |
a322f70c DW |
362 | /*__u64 lba_offset[0]; LBA offset in each phys. Note extents in a |
363 | bvd are always the same size */ | |
364 | }; | |
9d0c6b70 | 365 | #define LBA_OFFSET(ddf, vd) ((be64 *) &(vd)->phys_refnum[(ddf)->mppe]) |
a322f70c DW |
366 | |
367 | /* vd_config.cache_pol[7] is a bitmap */ | |
368 | #define DDF_cache_writeback 1 /* else writethrough */ | |
369 | #define DDF_cache_wadaptive 2 /* only applies if writeback */ | |
370 | #define DDF_cache_readahead 4 | |
371 | #define DDF_cache_radaptive 8 /* only if doing read-ahead */ | |
372 | #define DDF_cache_ifnobatt 16 /* even to write cache if battery is poor */ | |
373 | #define DDF_cache_wallowed 32 /* enable write caching */ | |
374 | #define DDF_cache_rallowed 64 /* enable read caching */ | |
375 | ||
376 | struct spare_assign { | |
60931cf9 | 377 | be32 magic; /* DDF_SPARE_ASSIGN_MAGIC */ |
378 | be32 crc; | |
379 | be32 timestamp; | |
a322f70c DW |
380 | __u8 reserved[7]; |
381 | __u8 type; | |
a8173e43 | 382 | be16 populated; /* SAEs used */ |
383 | be16 max; /* max SAEs */ | |
a322f70c DW |
384 | __u8 pad[8]; |
385 | struct spare_assign_entry { | |
386 | char guid[DDF_GUID_LEN]; | |
a8173e43 | 387 | be16 secondary_element; |
a322f70c DW |
388 | __u8 pad[6]; |
389 | } spare_ents[0]; | |
390 | }; | |
391 | /* spare_assign.type is a bitmap */ | |
392 | #define DDF_spare_dedicated 0x1 /* else global */ | |
393 | #define DDF_spare_revertible 0x2 /* else committable */ | |
394 | #define DDF_spare_active 0x4 /* else not active */ | |
395 | #define DDF_spare_affinity 0x8 /* enclosure affinity */ | |
396 | ||
397 | /* The data_section contents - local scope */ | |
398 | struct disk_data { | |
60931cf9 | 399 | be32 magic; /* DDF_PHYS_DATA_MAGIC */ |
400 | be32 crc; | |
a322f70c | 401 | char guid[DDF_GUID_LEN]; |
60931cf9 | 402 | be32 refnum; /* crc of some magic drive data ... */ |
a322f70c DW |
403 | __u8 forced_ref; /* set when above was not result of magic */ |
404 | __u8 forced_guid; /* set if guid was forced rather than magic */ | |
405 | __u8 vendor[32]; | |
406 | __u8 pad[442]; | |
407 | }; | |
408 | ||
409 | /* bbm_section content */ | |
410 | struct bad_block_log { | |
60931cf9 | 411 | be32 magic; |
412 | be32 crc; | |
a8173e43 | 413 | be16 entry_count; |
60931cf9 | 414 | be32 spare_count; |
a322f70c | 415 | __u8 pad[10]; |
9d0c6b70 | 416 | be64 first_spare; |
a322f70c | 417 | struct mapped_block { |
9d0c6b70 | 418 | be64 defective_start; |
60931cf9 | 419 | be32 replacement_start; |
a8173e43 | 420 | be16 remap_count; |
a322f70c DW |
421 | __u8 pad[2]; |
422 | } entries[0]; | |
423 | }; | |
424 | ||
425 | /* Struct for internally holding ddf structures */ | |
426 | /* The DDF structure stored on each device is potentially | |
427 | * quite different, as some data is global and some is local. | |
428 | * The global data is: | |
429 | * - ddf header | |
430 | * - controller_data | |
431 | * - Physical disk records | |
432 | * - Virtual disk records | |
433 | * The local data is: | |
434 | * - Configuration records | |
435 | * - Physical Disk data section | |
436 | * ( and Bad block and vendor which I don't care about yet). | |
437 | * | |
438 | * The local data is parsed into separate lists as it is read | |
439 | * and reconstructed for writing. This means that we only need | |
440 | * to make config changes once and they are automatically | |
441 | * propagated to all devices. | |
56cb05c4 N |
442 | * The global (config and disk data) records are each in a list |
443 | * of separate data structures. When writing we find the entry | |
444 | * or entries applicable to the particular device. | |
a322f70c DW |
445 | */ |
446 | struct ddf_super { | |
56cb05c4 | 447 | struct ddf_header anchor, primary, secondary; |
a322f70c | 448 | struct ddf_controller_data controller; |
56cb05c4 | 449 | struct ddf_header *active; |
a322f70c DW |
450 | struct phys_disk *phys; |
451 | struct virtual_disk *virt; | |
3921e41a | 452 | char *conf; |
56cb05c4 N |
453 | int pdsize, vdsize; |
454 | unsigned int max_part, mppe, conf_rec_len; | |
455 | int currentdev; | |
456 | int updates_pending; | |
a322f70c | 457 | struct vcl { |
6416d527 NB |
458 | union { |
459 | char space[512]; | |
460 | struct { | |
461 | struct vcl *next; | |
f21e18ca | 462 | unsigned int vcnum; /* index into ->virt */ |
56cb05c4 N |
463 | /* For an array with a secondary level there are |
464 | * multiple vd_config structures, all with the same | |
465 | * guid but with different sec_elmnt_seq. | |
466 | * One of these structures is in 'conf' below. | |
467 | * The others are in other_bvds, not in any | |
468 | * particular order. | |
469 | */ | |
8ec5d685 | 470 | struct vd_config **other_bvds; |
6416d527 NB |
471 | __u64 *block_sizes; /* NULL if all the same */ |
472 | }; | |
473 | }; | |
a322f70c | 474 | struct vd_config conf; |
d2ca6449 | 475 | } *conflist, *currentconf; |
a322f70c | 476 | struct dl { |
6416d527 NB |
477 | union { |
478 | char space[512]; | |
479 | struct { | |
480 | struct dl *next; | |
481 | int major, minor; | |
482 | char *devname; | |
483 | int fd; | |
484 | unsigned long long size; /* sectors */ | |
9d0c6b70 | 485 | be64 primary_lba; /* sectors */ |
486 | be64 secondary_lba; /* sectors */ | |
487 | be64 workspace_lba; /* sectors */ | |
6416d527 NB |
488 | int pdnum; /* index in ->phys */ |
489 | struct spare_assign *spare; | |
8592f29d N |
490 | void *mdupdate; /* hold metadata update */ |
491 | ||
492 | /* These fields used by auto-layout */ | |
493 | int raiddisk; /* slot to fill in autolayout */ | |
494 | __u64 esize; | |
d2ec75fb | 495 | int displayed; |
6416d527 NB |
496 | }; |
497 | }; | |
a322f70c | 498 | struct disk_data disk; |
b2280677 | 499 | struct vcl *vlist[0]; /* max_part in size */ |
2cc2983d | 500 | } *dlist, *add_list; |
a322f70c DW |
501 | }; |
502 | ||
56cb05c4 N |
503 | static int load_super_ddf_all(struct supertype *st, int fd, |
504 | void **sbp, char *devname); | |
505 | static int get_svd_state(const struct ddf_super *, const struct vcl *); | |
506 | static int | |
507 | validate_geometry_ddf_container(struct supertype *st, | |
508 | int level, int layout, int raiddisks, | |
509 | int chunk, unsigned long long size, | |
510 | unsigned long long data_offset, | |
511 | char *dev, unsigned long long *freesize, | |
512 | int verbose); | |
513 | ||
514 | static int validate_geometry_ddf_bvd(struct supertype *st, | |
515 | int level, int layout, int raiddisks, | |
516 | int *chunk, unsigned long long size, | |
517 | unsigned long long data_offset, | |
518 | char *dev, unsigned long long *freesize, | |
519 | int verbose); | |
56cb05c4 N |
520 | |
521 | static void free_super_ddf(struct supertype *st); | |
522 | static int all_ff(const char *guid); | |
523 | static unsigned int get_pd_index_from_refnum(const struct vcl *vc, | |
524 | be32 refnum, unsigned int nmax, | |
525 | const struct vd_config **bvd, | |
526 | unsigned int *idx); | |
527 | static void getinfo_super_ddf(struct supertype *st, struct mdinfo *info, char *map); | |
528 | static void uuid_from_ddf_guid(const char *guid, int uuid[4]); | |
529 | static void uuid_from_super_ddf(struct supertype *st, int uuid[4]); | |
530 | static void _ddf_array_name(char *name, const struct ddf_super *ddf, int i); | |
531 | static void getinfo_super_ddf_bvd(struct supertype *st, struct mdinfo *info, char *map); | |
532 | static int init_super_ddf_bvd(struct supertype *st, | |
533 | mdu_array_info_t *info, | |
534 | unsigned long long size, | |
535 | char *name, char *homehost, | |
536 | int *uuid, unsigned long long data_offset); | |
537 | ||
7d5a7ff3 | 538 | #if DEBUG |
7d5a7ff3 | 539 | static void pr_state(struct ddf_super *ddf, const char *msg) |
540 | { | |
541 | unsigned int i; | |
1ade5cc1 | 542 | dprintf("%s: ", msg); |
a8173e43 | 543 | for (i = 0; i < be16_to_cpu(ddf->active->max_vd_entries); i++) { |
7d5a7ff3 | 544 | if (all_ff(ddf->virt->entries[i].guid)) |
545 | continue; | |
1ade5cc1 | 546 | dprintf_cont("%u(s=%02x i=%02x) ", i, |
7d5a7ff3 | 547 | ddf->virt->entries[i].state, |
548 | ddf->virt->entries[i].init_state); | |
549 | } | |
1ade5cc1 | 550 | dprintf_cont("\n"); |
7d5a7ff3 | 551 | } |
552 | #else | |
553 | static void pr_state(const struct ddf_super *ddf, const char *msg) {} | |
554 | #endif | |
555 | ||
609ce161 N |
556 | static void _ddf_set_updates_pending(struct ddf_super *ddf, struct vd_config *vc, |
557 | const char *func) | |
35c3606d | 558 | { |
609ce161 N |
559 | if (vc) { |
560 | vc->timestamp = cpu_to_be32(time(0)-DECADE); | |
561 | vc->seqnum = cpu_to_be32(be32_to_cpu(vc->seqnum) + 1); | |
562 | } | |
5a46fcd7 N |
563 | if (ddf->updates_pending) |
564 | return; | |
35c3606d | 565 | ddf->updates_pending = 1; |
60931cf9 | 566 | ddf->active->seq = cpu_to_be32((be32_to_cpu(ddf->active->seq)+1)); |
35c3606d | 567 | pr_state(ddf, func); |
568 | } | |
569 | ||
609ce161 | 570 | #define ddf_set_updates_pending(x,v) _ddf_set_updates_pending((x), (v), __func__) |
7d5a7ff3 | 571 | |
60931cf9 | 572 | static be32 calc_crc(void *buf, int len) |
a322f70c DW |
573 | { |
574 | /* crcs are always at the same place as in the ddf_header */ | |
575 | struct ddf_header *ddf = buf; | |
60931cf9 | 576 | be32 oldcrc = ddf->crc; |
a322f70c | 577 | __u32 newcrc; |
60931cf9 | 578 | ddf->crc = cpu_to_be32(0xffffffff); |
a322f70c DW |
579 | |
580 | newcrc = crc32(0, buf, len); | |
581 | ddf->crc = oldcrc; | |
56cb05c4 | 582 | /* The crc is stored (like everything) bigendian, so convert |
4abe6b70 N |
583 | * here for simplicity |
584 | */ | |
60931cf9 | 585 | return cpu_to_be32(newcrc); |
a322f70c DW |
586 | } |
587 | ||
a3163bf0 | 588 | #define DDF_INVALID_LEVEL 0xff |
589 | #define DDF_NO_SECONDARY 0xff | |
590 | static int err_bad_md_layout(const mdu_array_info_t *array) | |
591 | { | |
592 | pr_err("RAID%d layout %x with %d disks is unsupported for DDF\n", | |
593 | array->level, array->layout, array->raid_disks); | |
2aba583f | 594 | return -1; |
a3163bf0 | 595 | } |
596 | ||
597 | static int layout_md2ddf(const mdu_array_info_t *array, | |
598 | struct vd_config *conf) | |
599 | { | |
a8173e43 | 600 | be16 prim_elmnt_count = cpu_to_be16(array->raid_disks); |
a3163bf0 | 601 | __u8 prl = DDF_INVALID_LEVEL, rlq = 0; |
602 | __u8 sec_elmnt_count = 1; | |
603 | __u8 srl = DDF_NO_SECONDARY; | |
604 | ||
605 | switch (array->level) { | |
606 | case LEVEL_LINEAR: | |
607 | prl = DDF_CONCAT; | |
608 | break; | |
609 | case 0: | |
610 | rlq = DDF_RAID0_SIMPLE; | |
611 | prl = DDF_RAID0; | |
612 | break; | |
613 | case 1: | |
614 | switch (array->raid_disks) { | |
615 | case 2: | |
616 | rlq = DDF_RAID1_SIMPLE; | |
617 | break; | |
618 | case 3: | |
619 | rlq = DDF_RAID1_MULTI; | |
620 | break; | |
621 | default: | |
622 | return err_bad_md_layout(array); | |
623 | } | |
624 | prl = DDF_RAID1; | |
625 | break; | |
626 | case 4: | |
627 | if (array->layout != 0) | |
628 | return err_bad_md_layout(array); | |
629 | rlq = DDF_RAID4_N; | |
630 | prl = DDF_RAID4; | |
631 | break; | |
632 | case 5: | |
633 | switch (array->layout) { | |
634 | case ALGORITHM_LEFT_ASYMMETRIC: | |
635 | rlq = DDF_RAID5_N_RESTART; | |
636 | break; | |
637 | case ALGORITHM_RIGHT_ASYMMETRIC: | |
638 | rlq = DDF_RAID5_0_RESTART; | |
639 | break; | |
640 | case ALGORITHM_LEFT_SYMMETRIC: | |
641 | rlq = DDF_RAID5_N_CONTINUE; | |
642 | break; | |
643 | case ALGORITHM_RIGHT_SYMMETRIC: | |
644 | /* not mentioned in standard */ | |
645 | default: | |
646 | return err_bad_md_layout(array); | |
647 | } | |
648 | prl = DDF_RAID5; | |
649 | break; | |
650 | case 6: | |
651 | switch (array->layout) { | |
652 | case ALGORITHM_ROTATING_N_RESTART: | |
653 | rlq = DDF_RAID5_N_RESTART; | |
654 | break; | |
655 | case ALGORITHM_ROTATING_ZERO_RESTART: | |
656 | rlq = DDF_RAID6_0_RESTART; | |
657 | break; | |
658 | case ALGORITHM_ROTATING_N_CONTINUE: | |
659 | rlq = DDF_RAID5_N_CONTINUE; | |
660 | break; | |
661 | default: | |
662 | return err_bad_md_layout(array); | |
663 | } | |
664 | prl = DDF_RAID6; | |
665 | break; | |
666 | case 10: | |
667 | if (array->raid_disks % 2 == 0 && array->layout == 0x102) { | |
668 | rlq = DDF_RAID1_SIMPLE; | |
a8173e43 | 669 | prim_elmnt_count = cpu_to_be16(2); |
a3163bf0 | 670 | sec_elmnt_count = array->raid_disks / 2; |
f70d549f N |
671 | srl = DDF_2SPANNED; |
672 | prl = DDF_RAID1; | |
a3163bf0 | 673 | } else if (array->raid_disks % 3 == 0 |
674 | && array->layout == 0x103) { | |
675 | rlq = DDF_RAID1_MULTI; | |
a8173e43 | 676 | prim_elmnt_count = cpu_to_be16(3); |
a3163bf0 | 677 | sec_elmnt_count = array->raid_disks / 3; |
f70d549f N |
678 | srl = DDF_2SPANNED; |
679 | prl = DDF_RAID1; | |
680 | } else if (array->layout == 0x201) { | |
681 | prl = DDF_RAID1E; | |
682 | rlq = DDF_RAID1E_OFFSET; | |
683 | } else if (array->layout == 0x102) { | |
684 | prl = DDF_RAID1E; | |
685 | rlq = DDF_RAID1E_ADJACENT; | |
a3163bf0 | 686 | } else |
687 | return err_bad_md_layout(array); | |
a3163bf0 | 688 | break; |
689 | default: | |
690 | return err_bad_md_layout(array); | |
691 | } | |
692 | conf->prl = prl; | |
693 | conf->prim_elmnt_count = prim_elmnt_count; | |
694 | conf->rlq = rlq; | |
695 | conf->srl = srl; | |
696 | conf->sec_elmnt_count = sec_elmnt_count; | |
697 | return 0; | |
698 | } | |
699 | ||
8a2848a7 | 700 | static int err_bad_ddf_layout(const struct vd_config *conf) |
701 | { | |
702 | pr_err("DDF RAID %u qualifier %u with %u disks is unsupported\n", | |
a8173e43 | 703 | conf->prl, conf->rlq, be16_to_cpu(conf->prim_elmnt_count)); |
8a2848a7 | 704 | return -1; |
705 | } | |
706 | ||
707 | static int layout_ddf2md(const struct vd_config *conf, | |
708 | mdu_array_info_t *array) | |
709 | { | |
710 | int level = LEVEL_UNSUPPORTED; | |
711 | int layout = 0; | |
a8173e43 | 712 | int raiddisks = be16_to_cpu(conf->prim_elmnt_count); |
8a2848a7 | 713 | |
714 | if (conf->sec_elmnt_count > 1) { | |
715 | /* see also check_secondary() */ | |
716 | if (conf->prl != DDF_RAID1 || | |
717 | (conf->srl != DDF_2STRIPED && conf->srl != DDF_2SPANNED)) { | |
718 | pr_err("Unsupported secondary RAID level %u/%u\n", | |
719 | conf->prl, conf->srl); | |
720 | return -1; | |
721 | } | |
722 | if (raiddisks == 2 && conf->rlq == DDF_RAID1_SIMPLE) | |
723 | layout = 0x102; | |
724 | else if (raiddisks == 3 && conf->rlq == DDF_RAID1_MULTI) | |
725 | layout = 0x103; | |
726 | else | |
727 | return err_bad_ddf_layout(conf); | |
728 | raiddisks *= conf->sec_elmnt_count; | |
729 | level = 10; | |
730 | goto good; | |
731 | } | |
732 | ||
733 | switch (conf->prl) { | |
734 | case DDF_CONCAT: | |
735 | level = LEVEL_LINEAR; | |
736 | break; | |
737 | case DDF_RAID0: | |
738 | if (conf->rlq != DDF_RAID0_SIMPLE) | |
739 | return err_bad_ddf_layout(conf); | |
740 | level = 0; | |
741 | break; | |
742 | case DDF_RAID1: | |
743 | if (!((conf->rlq == DDF_RAID1_SIMPLE && raiddisks == 2) || | |
744 | (conf->rlq == DDF_RAID1_MULTI && raiddisks == 3))) | |
745 | return err_bad_ddf_layout(conf); | |
746 | level = 1; | |
747 | break; | |
f70d549f N |
748 | case DDF_RAID1E: |
749 | if (conf->rlq == DDF_RAID1E_ADJACENT) | |
750 | layout = 0x102; | |
751 | else if (conf->rlq == DDF_RAID1E_OFFSET) | |
752 | layout = 0x201; | |
753 | else | |
754 | return err_bad_ddf_layout(conf); | |
755 | level = 10; | |
756 | break; | |
8a2848a7 | 757 | case DDF_RAID4: |
758 | if (conf->rlq != DDF_RAID4_N) | |
759 | return err_bad_ddf_layout(conf); | |
760 | level = 4; | |
761 | break; | |
762 | case DDF_RAID5: | |
763 | switch (conf->rlq) { | |
764 | case DDF_RAID5_N_RESTART: | |
765 | layout = ALGORITHM_LEFT_ASYMMETRIC; | |
766 | break; | |
767 | case DDF_RAID5_0_RESTART: | |
768 | layout = ALGORITHM_RIGHT_ASYMMETRIC; | |
769 | break; | |
770 | case DDF_RAID5_N_CONTINUE: | |
771 | layout = ALGORITHM_LEFT_SYMMETRIC; | |
772 | break; | |
773 | default: | |
774 | return err_bad_ddf_layout(conf); | |
775 | } | |
776 | level = 5; | |
777 | break; | |
778 | case DDF_RAID6: | |
779 | switch (conf->rlq) { | |
780 | case DDF_RAID5_N_RESTART: | |
781 | layout = ALGORITHM_ROTATING_N_RESTART; | |
782 | break; | |
783 | case DDF_RAID6_0_RESTART: | |
784 | layout = ALGORITHM_ROTATING_ZERO_RESTART; | |
785 | break; | |
786 | case DDF_RAID5_N_CONTINUE: | |
787 | layout = ALGORITHM_ROTATING_N_CONTINUE; | |
788 | break; | |
789 | default: | |
790 | return err_bad_ddf_layout(conf); | |
791 | } | |
792 | level = 6; | |
793 | break; | |
794 | default: | |
795 | return err_bad_ddf_layout(conf); | |
796 | }; | |
797 | ||
798 | good: | |
799 | array->level = level; | |
800 | array->layout = layout; | |
801 | array->raid_disks = raiddisks; | |
802 | return 0; | |
803 | } | |
804 | ||
a322f70c DW |
805 | static int load_ddf_header(int fd, unsigned long long lba, |
806 | unsigned long long size, | |
807 | int type, | |
808 | struct ddf_header *hdr, struct ddf_header *anchor) | |
809 | { | |
810 | /* read a ddf header (primary or secondary) from fd/lba | |
811 | * and check that it is consistent with anchor | |
812 | * Need to check: | |
813 | * magic, crc, guid, rev, and LBA's header_type, and | |
814 | * everything after header_type must be the same | |
815 | */ | |
816 | if (lba >= size-1) | |
817 | return 0; | |
818 | ||
819 | if (lseek64(fd, lba<<9, 0) < 0) | |
820 | return 0; | |
821 | ||
822 | if (read(fd, hdr, 512) != 512) | |
823 | return 0; | |
824 | ||
0e5fa862 | 825 | if (!be32_eq(hdr->magic, DDF_HEADER_MAGIC)) { |
1ade5cc1 | 826 | pr_err("bad header magic\n"); |
a322f70c | 827 | return 0; |
0e5fa862 MW |
828 | } |
829 | if (!be32_eq(calc_crc(hdr, 512), hdr->crc)) { | |
1ade5cc1 | 830 | pr_err("bad CRC\n"); |
a322f70c | 831 | return 0; |
0e5fa862 | 832 | } |
a322f70c DW |
833 | if (memcmp(anchor->guid, hdr->guid, DDF_GUID_LEN) != 0 || |
834 | memcmp(anchor->revision, hdr->revision, 8) != 0 || | |
9d0c6b70 | 835 | !be64_eq(anchor->primary_lba, hdr->primary_lba) || |
836 | !be64_eq(anchor->secondary_lba, hdr->secondary_lba) || | |
a322f70c DW |
837 | hdr->type != type || |
838 | memcmp(anchor->pad2, hdr->pad2, 512 - | |
0e5fa862 | 839 | offsetof(struct ddf_header, pad2)) != 0) { |
1ade5cc1 | 840 | pr_err("header mismatch\n"); |
a322f70c | 841 | return 0; |
0e5fa862 | 842 | } |
a322f70c DW |
843 | |
844 | /* Looks good enough to me... */ | |
845 | return 1; | |
846 | } | |
847 | ||
848 | static void *load_section(int fd, struct ddf_super *super, void *buf, | |
60931cf9 | 849 | be32 offset_be, be32 len_be, int check) |
a322f70c | 850 | { |
60931cf9 | 851 | unsigned long long offset = be32_to_cpu(offset_be); |
852 | unsigned long long len = be32_to_cpu(len_be); | |
a322f70c DW |
853 | int dofree = (buf == NULL); |
854 | ||
855 | if (check) | |
856 | if (len != 2 && len != 8 && len != 32 | |
857 | && len != 128 && len != 512) | |
858 | return NULL; | |
859 | ||
860 | if (len > 1024) | |
861 | return NULL; | |
3921e41a | 862 | if (!buf && posix_memalign(&buf, 512, len<<9) != 0) |
3d2c4fc7 | 863 | buf = NULL; |
6416d527 | 864 | |
a322f70c DW |
865 | if (!buf) |
866 | return NULL; | |
867 | ||
868 | if (super->active->type == 1) | |
9d0c6b70 | 869 | offset += be64_to_cpu(super->active->primary_lba); |
a322f70c | 870 | else |
9d0c6b70 | 871 | offset += be64_to_cpu(super->active->secondary_lba); |
a322f70c | 872 | |
f21e18ca | 873 | if ((unsigned long long)lseek64(fd, offset<<9, 0) != (offset<<9)) { |
a322f70c DW |
874 | if (dofree) |
875 | free(buf); | |
876 | return NULL; | |
877 | } | |
f21e18ca | 878 | if ((unsigned long long)read(fd, buf, len<<9) != (len<<9)) { |
a322f70c DW |
879 | if (dofree) |
880 | free(buf); | |
881 | return NULL; | |
882 | } | |
883 | return buf; | |
884 | } | |
885 | ||
886 | static int load_ddf_headers(int fd, struct ddf_super *super, char *devname) | |
887 | { | |
888 | unsigned long long dsize; | |
889 | ||
890 | get_dev_size(fd, NULL, &dsize); | |
891 | ||
892 | if (lseek64(fd, dsize-512, 0) < 0) { | |
893 | if (devname) | |
e7b84f9d N |
894 | pr_err("Cannot seek to anchor block on %s: %s\n", |
895 | devname, strerror(errno)); | |
a322f70c DW |
896 | return 1; |
897 | } | |
898 | if (read(fd, &super->anchor, 512) != 512) { | |
899 | if (devname) | |
e7b84f9d N |
900 | pr_err("Cannot read anchor block on %s: %s\n", |
901 | devname, strerror(errno)); | |
a322f70c DW |
902 | return 1; |
903 | } | |
60931cf9 | 904 | if (!be32_eq(super->anchor.magic, DDF_HEADER_MAGIC)) { |
a322f70c | 905 | if (devname) |
e7b84f9d | 906 | pr_err("no DDF anchor found on %s\n", |
a322f70c DW |
907 | devname); |
908 | return 2; | |
909 | } | |
60931cf9 | 910 | if (!be32_eq(calc_crc(&super->anchor, 512), super->anchor.crc)) { |
a322f70c | 911 | if (devname) |
e7b84f9d | 912 | pr_err("bad CRC on anchor on %s\n", |
a322f70c DW |
913 | devname); |
914 | return 2; | |
915 | } | |
59e36268 NB |
916 | if (memcmp(super->anchor.revision, DDF_REVISION_0, 8) != 0 && |
917 | memcmp(super->anchor.revision, DDF_REVISION_2, 8) != 0) { | |
a322f70c | 918 | if (devname) |
7a862a02 | 919 | pr_err("can only support super revision %.8s and earlier, not %.8s on %s\n", |
59e36268 | 920 | DDF_REVISION_2, super->anchor.revision,devname); |
a322f70c DW |
921 | return 2; |
922 | } | |
dbeb699a | 923 | super->active = NULL; |
9d0c6b70 | 924 | if (load_ddf_header(fd, be64_to_cpu(super->anchor.primary_lba), |
a322f70c DW |
925 | dsize >> 9, 1, |
926 | &super->primary, &super->anchor) == 0) { | |
927 | if (devname) | |
7a862a02 | 928 | pr_err("Failed to load primary DDF header on %s\n", devname); |
dbeb699a | 929 | } else |
930 | super->active = &super->primary; | |
60931cf9 | 931 | |
9d0c6b70 | 932 | if (load_ddf_header(fd, be64_to_cpu(super->anchor.secondary_lba), |
a322f70c DW |
933 | dsize >> 9, 2, |
934 | &super->secondary, &super->anchor)) { | |
3eff7c1d | 935 | if (super->active == NULL |
60931cf9 | 936 | || (be32_to_cpu(super->primary.seq) |
937 | < be32_to_cpu(super->secondary.seq) && | |
3eff7c1d | 938 | !super->secondary.openflag) |
60931cf9 | 939 | || (be32_to_cpu(super->primary.seq) |
940 | == be32_to_cpu(super->secondary.seq) && | |
a322f70c DW |
941 | super->primary.openflag && !super->secondary.openflag) |
942 | ) | |
943 | super->active = &super->secondary; | |
b95cb4b9 N |
944 | } else if (devname && |
945 | be64_to_cpu(super->anchor.secondary_lba) != ~(__u64)0) | |
dbeb699a | 946 | pr_err("Failed to load secondary DDF header on %s\n", |
947 | devname); | |
948 | if (super->active == NULL) | |
949 | return 2; | |
a322f70c DW |
950 | return 0; |
951 | } | |
952 | ||
953 | static int load_ddf_global(int fd, struct ddf_super *super, char *devname) | |
954 | { | |
955 | void *ok; | |
956 | ok = load_section(fd, super, &super->controller, | |
957 | super->active->controller_section_offset, | |
958 | super->active->controller_section_length, | |
959 | 0); | |
960 | super->phys = load_section(fd, super, NULL, | |
961 | super->active->phys_section_offset, | |
962 | super->active->phys_section_length, | |
963 | 1); | |
60931cf9 | 964 | super->pdsize = be32_to_cpu(super->active->phys_section_length) * 512; |
a322f70c DW |
965 | |
966 | super->virt = load_section(fd, super, NULL, | |
967 | super->active->virt_section_offset, | |
968 | super->active->virt_section_length, | |
969 | 1); | |
60931cf9 | 970 | super->vdsize = be32_to_cpu(super->active->virt_section_length) * 512; |
a322f70c DW |
971 | if (!ok || |
972 | !super->phys || | |
973 | !super->virt) { | |
974 | free(super->phys); | |
975 | free(super->virt); | |
a2349791 NB |
976 | super->phys = NULL; |
977 | super->virt = NULL; | |
a322f70c DW |
978 | return 2; |
979 | } | |
980 | super->conflist = NULL; | |
981 | super->dlist = NULL; | |
8c3b8c2c | 982 | |
a8173e43 | 983 | super->max_part = be16_to_cpu(super->active->max_partitions); |
984 | super->mppe = be16_to_cpu(super->active->max_primary_element_entries); | |
985 | super->conf_rec_len = be16_to_cpu(super->active->config_record_len); | |
a322f70c DW |
986 | return 0; |
987 | } | |
988 | ||
3c48f7be | 989 | #define DDF_UNUSED_BVD 0xff |
990 | static int alloc_other_bvds(const struct ddf_super *ddf, struct vcl *vcl) | |
991 | { | |
992 | unsigned int n_vds = vcl->conf.sec_elmnt_count - 1; | |
993 | unsigned int i, vdsize; | |
994 | void *p; | |
995 | if (n_vds == 0) { | |
996 | vcl->other_bvds = NULL; | |
997 | return 0; | |
998 | } | |
999 | vdsize = ddf->conf_rec_len * 512; | |
1000 | if (posix_memalign(&p, 512, n_vds * | |
1001 | (vdsize + sizeof(struct vd_config *))) != 0) | |
1002 | return -1; | |
1003 | vcl->other_bvds = (struct vd_config **) (p + n_vds * vdsize); | |
1004 | for (i = 0; i < n_vds; i++) { | |
1005 | vcl->other_bvds[i] = p + i * vdsize; | |
1006 | memset(vcl->other_bvds[i], 0, vdsize); | |
1007 | vcl->other_bvds[i]->sec_elmnt_seq = DDF_UNUSED_BVD; | |
1008 | } | |
1009 | return 0; | |
1010 | } | |
1011 | ||
3dc821b0 | 1012 | static void add_other_bvd(struct vcl *vcl, struct vd_config *vd, |
1013 | unsigned int len) | |
1014 | { | |
1015 | int i; | |
1016 | for (i = 0; i < vcl->conf.sec_elmnt_count-1; i++) | |
3c48f7be | 1017 | if (vcl->other_bvds[i]->sec_elmnt_seq == vd->sec_elmnt_seq) |
3dc821b0 | 1018 | break; |
1019 | ||
1020 | if (i < vcl->conf.sec_elmnt_count-1) { | |
60931cf9 | 1021 | if (be32_to_cpu(vd->seqnum) <= |
1022 | be32_to_cpu(vcl->other_bvds[i]->seqnum)) | |
3dc821b0 | 1023 | return; |
1024 | } else { | |
1025 | for (i = 0; i < vcl->conf.sec_elmnt_count-1; i++) | |
3c48f7be | 1026 | if (vcl->other_bvds[i]->sec_elmnt_seq == DDF_UNUSED_BVD) |
3dc821b0 | 1027 | break; |
1028 | if (i == vcl->conf.sec_elmnt_count-1) { | |
1029 | pr_err("no space for sec level config %u, count is %u\n", | |
1030 | vd->sec_elmnt_seq, vcl->conf.sec_elmnt_count); | |
1031 | return; | |
1032 | } | |
3dc821b0 | 1033 | } |
1034 | memcpy(vcl->other_bvds[i], vd, len); | |
1035 | } | |
1036 | ||
a322f70c DW |
1037 | static int load_ddf_local(int fd, struct ddf_super *super, |
1038 | char *devname, int keep) | |
1039 | { | |
1040 | struct dl *dl; | |
1041 | struct stat stb; | |
1042 | char *conf; | |
f21e18ca N |
1043 | unsigned int i; |
1044 | unsigned int confsec; | |
b2280677 | 1045 | int vnum; |
56cb05c4 N |
1046 | unsigned int max_virt_disks = |
1047 | be16_to_cpu(super->active->max_vd_entries); | |
d2ca6449 | 1048 | unsigned long long dsize; |
a322f70c DW |
1049 | |
1050 | /* First the local disk info */ | |
3d2c4fc7 | 1051 | if (posix_memalign((void**)&dl, 512, |
56cb05c4 N |
1052 | sizeof(*dl) + |
1053 | (super->max_part) * sizeof(dl->vlist[0])) != 0) { | |
1ade5cc1 | 1054 | pr_err("could not allocate disk info buffer\n"); |
3d2c4fc7 DW |
1055 | return 1; |
1056 | } | |
a322f70c DW |
1057 | |
1058 | load_section(fd, super, &dl->disk, | |
1059 | super->active->data_section_offset, | |
1060 | super->active->data_section_length, | |
1061 | 0); | |
503975b9 | 1062 | dl->devname = devname ? xstrdup(devname) : NULL; |
598f0d58 | 1063 | |
a322f70c DW |
1064 | fstat(fd, &stb); |
1065 | dl->major = major(stb.st_rdev); | |
1066 | dl->minor = minor(stb.st_rdev); | |
1067 | dl->next = super->dlist; | |
1068 | dl->fd = keep ? fd : -1; | |
d2ca6449 NB |
1069 | |
1070 | dl->size = 0; | |
1071 | if (get_dev_size(fd, devname, &dsize)) | |
1072 | dl->size = dsize >> 9; | |
097bcf00 | 1073 | /* If the disks have different sizes, the LBAs will differ |
1074 | * between phys disks. | |
1075 | * At this point here, the values in super->active must be valid | |
1076 | * for this phys disk. */ | |
1077 | dl->primary_lba = super->active->primary_lba; | |
1078 | dl->secondary_lba = super->active->secondary_lba; | |
1079 | dl->workspace_lba = super->active->workspace_lba; | |
b2280677 | 1080 | dl->spare = NULL; |
f21e18ca | 1081 | for (i = 0 ; i < super->max_part ; i++) |
a322f70c DW |
1082 | dl->vlist[i] = NULL; |
1083 | super->dlist = dl; | |
59e36268 | 1084 | dl->pdnum = -1; |
a8173e43 | 1085 | for (i = 0; i < be16_to_cpu(super->active->max_pd_entries); i++) |
5575e7d9 NB |
1086 | if (memcmp(super->phys->entries[i].guid, |
1087 | dl->disk.guid, DDF_GUID_LEN) == 0) | |
1088 | dl->pdnum = i; | |
1089 | ||
a322f70c DW |
1090 | /* Now the config list. */ |
1091 | /* 'conf' is an array of config entries, some of which are | |
1092 | * probably invalid. Those which are good need to be copied into | |
1093 | * the conflist | |
1094 | */ | |
a322f70c | 1095 | |
3921e41a | 1096 | conf = load_section(fd, super, super->conf, |
a322f70c DW |
1097 | super->active->config_section_offset, |
1098 | super->active->config_section_length, | |
1099 | 0); | |
3921e41a | 1100 | super->conf = conf; |
b2280677 | 1101 | vnum = 0; |
e223334f | 1102 | for (confsec = 0; |
60931cf9 | 1103 | confsec < be32_to_cpu(super->active->config_section_length); |
e223334f | 1104 | confsec += super->conf_rec_len) { |
a322f70c | 1105 | struct vd_config *vd = |
e223334f | 1106 | (struct vd_config *)((char*)conf + confsec*512); |
a322f70c DW |
1107 | struct vcl *vcl; |
1108 | ||
60931cf9 | 1109 | if (be32_eq(vd->magic, DDF_SPARE_ASSIGN_MAGIC)) { |
b2280677 NB |
1110 | if (dl->spare) |
1111 | continue; | |
3d2c4fc7 | 1112 | if (posix_memalign((void**)&dl->spare, 512, |
56cb05c4 | 1113 | super->conf_rec_len*512) != 0) { |
1ade5cc1 | 1114 | pr_err("could not allocate spare info buf\n"); |
3d2c4fc7 DW |
1115 | return 1; |
1116 | } | |
613b0d17 | 1117 | |
b2280677 NB |
1118 | memcpy(dl->spare, vd, super->conf_rec_len*512); |
1119 | continue; | |
1120 | } | |
60931cf9 | 1121 | if (!be32_eq(vd->magic, DDF_VD_CONF_MAGIC)) |
56cb05c4 | 1122 | /* Must be vendor-unique - I cannot handle those */ |
a322f70c | 1123 | continue; |
56cb05c4 | 1124 | |
a322f70c DW |
1125 | for (vcl = super->conflist; vcl; vcl = vcl->next) { |
1126 | if (memcmp(vcl->conf.guid, | |
1127 | vd->guid, DDF_GUID_LEN) == 0) | |
1128 | break; | |
1129 | } | |
1130 | ||
1131 | if (vcl) { | |
b2280677 | 1132 | dl->vlist[vnum++] = vcl; |
3dc821b0 | 1133 | if (vcl->other_bvds != NULL && |
1134 | vcl->conf.sec_elmnt_seq != vd->sec_elmnt_seq) { | |
1135 | add_other_bvd(vcl, vd, super->conf_rec_len*512); | |
1136 | continue; | |
1137 | } | |
60931cf9 | 1138 | if (be32_to_cpu(vd->seqnum) <= |
1139 | be32_to_cpu(vcl->conf.seqnum)) | |
a322f70c | 1140 | continue; |
59e36268 | 1141 | } else { |
3d2c4fc7 | 1142 | if (posix_memalign((void**)&vcl, 512, |
56cb05c4 N |
1143 | (super->conf_rec_len*512 + |
1144 | offsetof(struct vcl, conf))) != 0) { | |
1ade5cc1 | 1145 | pr_err("could not allocate vcl buf\n"); |
3d2c4fc7 DW |
1146 | return 1; |
1147 | } | |
a322f70c | 1148 | vcl->next = super->conflist; |
59e36268 | 1149 | vcl->block_sizes = NULL; /* FIXME not for CONCAT */ |
3c48f7be | 1150 | vcl->conf.sec_elmnt_count = vd->sec_elmnt_count; |
1151 | if (alloc_other_bvds(super, vcl) != 0) { | |
1ade5cc1 | 1152 | pr_err("could not allocate other bvds\n"); |
3c48f7be | 1153 | free(vcl); |
1154 | return 1; | |
1155 | }; | |
a322f70c | 1156 | super->conflist = vcl; |
b2280677 | 1157 | dl->vlist[vnum++] = vcl; |
a322f70c | 1158 | } |
8c3b8c2c | 1159 | memcpy(&vcl->conf, vd, super->conf_rec_len*512); |
59e36268 NB |
1160 | for (i=0; i < max_virt_disks ; i++) |
1161 | if (memcmp(super->virt->entries[i].guid, | |
1162 | vcl->conf.guid, DDF_GUID_LEN)==0) | |
1163 | break; | |
1164 | if (i < max_virt_disks) | |
1165 | vcl->vcnum = i; | |
a322f70c | 1166 | } |
a322f70c DW |
1167 | |
1168 | return 0; | |
1169 | } | |
1170 | ||
a322f70c DW |
1171 | static int load_super_ddf(struct supertype *st, int fd, |
1172 | char *devname) | |
1173 | { | |
1174 | unsigned long long dsize; | |
1175 | struct ddf_super *super; | |
1176 | int rv; | |
1177 | ||
a322f70c DW |
1178 | if (get_dev_size(fd, devname, &dsize) == 0) |
1179 | return 1; | |
1180 | ||
a34fea0e | 1181 | if (test_partition(fd)) |
691c6ee1 N |
1182 | /* DDF is not allowed on partitions */ |
1183 | return 1; | |
1184 | ||
a322f70c DW |
1185 | /* 32M is a lower bound */ |
1186 | if (dsize <= 32*1024*1024) { | |
97320d7c | 1187 | if (devname) |
7a862a02 | 1188 | pr_err("%s is too small for ddf: size is %llu sectors.\n", |
e7b84f9d | 1189 | devname, dsize>>9); |
97320d7c | 1190 | return 1; |
a322f70c DW |
1191 | } |
1192 | if (dsize & 511) { | |
97320d7c | 1193 | if (devname) |
7a862a02 | 1194 | pr_err("%s is an odd size for ddf: size is %llu bytes.\n", |
e7b84f9d | 1195 | devname, dsize); |
97320d7c | 1196 | return 1; |
a322f70c DW |
1197 | } |
1198 | ||
37424f13 DW |
1199 | free_super_ddf(st); |
1200 | ||
6416d527 | 1201 | if (posix_memalign((void**)&super, 512, sizeof(*super))!= 0) { |
e7b84f9d | 1202 | pr_err("malloc of %zu failed.\n", |
a322f70c DW |
1203 | sizeof(*super)); |
1204 | return 1; | |
1205 | } | |
a2349791 | 1206 | memset(super, 0, sizeof(*super)); |
a322f70c DW |
1207 | |
1208 | rv = load_ddf_headers(fd, super, devname); | |
1209 | if (rv) { | |
1210 | free(super); | |
1211 | return rv; | |
1212 | } | |
1213 | ||
1214 | /* Have valid headers and have chosen the best. Let's read in the rest*/ | |
1215 | ||
1216 | rv = load_ddf_global(fd, super, devname); | |
1217 | ||
1218 | if (rv) { | |
1219 | if (devname) | |
7a862a02 | 1220 | pr_err("Failed to load all information sections on %s\n", devname); |
a322f70c DW |
1221 | free(super); |
1222 | return rv; | |
1223 | } | |
1224 | ||
3d2c4fc7 DW |
1225 | rv = load_ddf_local(fd, super, devname, 0); |
1226 | ||
1227 | if (rv) { | |
1228 | if (devname) | |
7a862a02 | 1229 | pr_err("Failed to load all information sections on %s\n", devname); |
3d2c4fc7 DW |
1230 | free(super); |
1231 | return rv; | |
1232 | } | |
a322f70c DW |
1233 | |
1234 | /* Should possibly check the sections .... */ | |
1235 | ||
1236 | st->sb = super; | |
1237 | if (st->ss == NULL) { | |
1238 | st->ss = &super_ddf; | |
1239 | st->minor_version = 0; | |
1240 | st->max_devs = 512; | |
1241 | } | |
1242 | return 0; | |
1243 | ||
1244 | } | |
1245 | ||
1246 | static void free_super_ddf(struct supertype *st) | |
1247 | { | |
1248 | struct ddf_super *ddf = st->sb; | |
1249 | if (ddf == NULL) | |
1250 | return; | |
1251 | free(ddf->phys); | |
1252 | free(ddf->virt); | |
3921e41a | 1253 | free(ddf->conf); |
a322f70c DW |
1254 | while (ddf->conflist) { |
1255 | struct vcl *v = ddf->conflist; | |
1256 | ddf->conflist = v->next; | |
59e36268 NB |
1257 | if (v->block_sizes) |
1258 | free(v->block_sizes); | |
3c48f7be | 1259 | if (v->other_bvds) |
1260 | /* | |
1261 | v->other_bvds[0] points to beginning of buffer, | |
1262 | see alloc_other_bvds() | |
1263 | */ | |
1264 | free(v->other_bvds[0]); | |
a322f70c DW |
1265 | free(v); |
1266 | } | |
1267 | while (ddf->dlist) { | |
1268 | struct dl *d = ddf->dlist; | |
1269 | ddf->dlist = d->next; | |
1270 | if (d->fd >= 0) | |
1271 | close(d->fd); | |
b2280677 NB |
1272 | if (d->spare) |
1273 | free(d->spare); | |
a322f70c DW |
1274 | free(d); |
1275 | } | |
8a38cb04 N |
1276 | while (ddf->add_list) { |
1277 | struct dl *d = ddf->add_list; | |
1278 | ddf->add_list = d->next; | |
1279 | if (d->fd >= 0) | |
1280 | close(d->fd); | |
1281 | if (d->spare) | |
1282 | free(d->spare); | |
1283 | free(d); | |
1284 | } | |
a322f70c DW |
1285 | free(ddf); |
1286 | st->sb = NULL; | |
1287 | } | |
1288 | ||
1289 | static struct supertype *match_metadata_desc_ddf(char *arg) | |
1290 | { | |
56cb05c4 | 1291 | /* 'ddf' only supports containers */ |
a322f70c DW |
1292 | struct supertype *st; |
1293 | if (strcmp(arg, "ddf") != 0 && | |
1294 | strcmp(arg, "default") != 0 | |
1295 | ) | |
1296 | return NULL; | |
1297 | ||
503975b9 | 1298 | st = xcalloc(1, sizeof(*st)); |
a322f70c DW |
1299 | st->ss = &super_ddf; |
1300 | st->max_devs = 512; | |
1301 | st->minor_version = 0; | |
1302 | st->sb = NULL; | |
1303 | return st; | |
1304 | } | |
1305 | ||
a322f70c DW |
1306 | static mapping_t ddf_state[] = { |
1307 | { "Optimal", 0}, | |
1308 | { "Degraded", 1}, | |
1309 | { "Deleted", 2}, | |
1310 | { "Missing", 3}, | |
1311 | { "Failed", 4}, | |
1312 | { "Partially Optimal", 5}, | |
1313 | { "-reserved-", 6}, | |
1314 | { "-reserved-", 7}, | |
1315 | { NULL, 0} | |
1316 | }; | |
1317 | ||
1318 | static mapping_t ddf_init_state[] = { | |
1319 | { "Not Initialised", 0}, | |
1320 | { "QuickInit in Progress", 1}, | |
1321 | { "Fully Initialised", 2}, | |
1322 | { "*UNKNOWN*", 3}, | |
1323 | { NULL, 0} | |
1324 | }; | |
1325 | static mapping_t ddf_access[] = { | |
1326 | { "Read/Write", 0}, | |
1327 | { "Reserved", 1}, | |
1328 | { "Read Only", 2}, | |
1329 | { "Blocked (no access)", 3}, | |
1330 | { NULL ,0} | |
1331 | }; | |
1332 | ||
1333 | static mapping_t ddf_level[] = { | |
1334 | { "RAID0", DDF_RAID0}, | |
1335 | { "RAID1", DDF_RAID1}, | |
1336 | { "RAID3", DDF_RAID3}, | |
1337 | { "RAID4", DDF_RAID4}, | |
1338 | { "RAID5", DDF_RAID5}, | |
1339 | { "RAID1E",DDF_RAID1E}, | |
1340 | { "JBOD", DDF_JBOD}, | |
1341 | { "CONCAT",DDF_CONCAT}, | |
1342 | { "RAID5E",DDF_RAID5E}, | |
1343 | { "RAID5EE",DDF_RAID5EE}, | |
1344 | { "RAID6", DDF_RAID6}, | |
1345 | { NULL, 0} | |
1346 | }; | |
1347 | static mapping_t ddf_sec_level[] = { | |
1348 | { "Striped", DDF_2STRIPED}, | |
1349 | { "Mirrored", DDF_2MIRRORED}, | |
1350 | { "Concat", DDF_2CONCAT}, | |
1351 | { "Spanned", DDF_2SPANNED}, | |
1352 | { NULL, 0} | |
1353 | }; | |
a322f70c | 1354 | |
fb9d0acb | 1355 | static int all_ff(const char *guid) |
42dc2744 N |
1356 | { |
1357 | int i; | |
1358 | for (i = 0; i < DDF_GUID_LEN; i++) | |
1359 | if (guid[i] != (char)0xff) | |
1360 | return 0; | |
1361 | return 1; | |
1362 | } | |
1363 | ||
4441541f N |
1364 | static const char *guid_str(const char *guid) |
1365 | { | |
1366 | static char buf[DDF_GUID_LEN*2+1]; | |
1367 | int i; | |
1368 | char *p = buf; | |
1369 | for (i = 0; i < DDF_GUID_LEN; i++) { | |
1370 | unsigned char c = guid[i]; | |
1371 | if (c >= 32 && c < 127) | |
1372 | p += sprintf(p, "%c", c); | |
1373 | else | |
1374 | p += sprintf(p, "%02x", c); | |
1375 | } | |
1376 | *p = '\0'; | |
1377 | return (const char *) buf; | |
1378 | } | |
1379 | ||
a322f70c DW |
1380 | static void print_guid(char *guid, int tstamp) |
1381 | { | |
1382 | /* A GUIDs are part (or all) ASCII and part binary. | |
1383 | * They tend to be space padded. | |
59e36268 NB |
1384 | * We print the GUID in HEX, then in parentheses add |
1385 | * any initial ASCII sequence, and a possible | |
1386 | * time stamp from bytes 16-19 | |
a322f70c DW |
1387 | */ |
1388 | int l = DDF_GUID_LEN; | |
1389 | int i; | |
59e36268 NB |
1390 | |
1391 | for (i=0 ; i<DDF_GUID_LEN ; i++) { | |
1392 | if ((i&3)==0 && i != 0) printf(":"); | |
1393 | printf("%02X", guid[i]&255); | |
1394 | } | |
1395 | ||
cfccea8c | 1396 | printf("\n ("); |
a322f70c DW |
1397 | while (l && guid[l-1] == ' ') |
1398 | l--; | |
1399 | for (i=0 ; i<l ; i++) { | |
1400 | if (guid[i] >= 0x20 && guid[i] < 0x7f) | |
1401 | fputc(guid[i], stdout); | |
1402 | else | |
59e36268 | 1403 | break; |
a322f70c DW |
1404 | } |
1405 | if (tstamp) { | |
1406 | time_t then = __be32_to_cpu(*(__u32*)(guid+16)) + DECADE; | |
1407 | char tbuf[100]; | |
1408 | struct tm *tm; | |
1409 | tm = localtime(&then); | |
59e36268 | 1410 | strftime(tbuf, 100, " %D %T",tm); |
a322f70c DW |
1411 | fputs(tbuf, stdout); |
1412 | } | |
59e36268 | 1413 | printf(")"); |
a322f70c DW |
1414 | } |
1415 | ||
1416 | static void examine_vd(int n, struct ddf_super *sb, char *guid) | |
1417 | { | |
8c3b8c2c | 1418 | int crl = sb->conf_rec_len; |
a322f70c DW |
1419 | struct vcl *vcl; |
1420 | ||
1421 | for (vcl = sb->conflist ; vcl ; vcl = vcl->next) { | |
f21e18ca | 1422 | unsigned int i; |
a322f70c DW |
1423 | struct vd_config *vc = &vcl->conf; |
1424 | ||
60931cf9 | 1425 | if (!be32_eq(calc_crc(vc, crl*512), vc->crc)) |
a322f70c DW |
1426 | continue; |
1427 | if (memcmp(vc->guid, guid, DDF_GUID_LEN) != 0) | |
1428 | continue; | |
1429 | ||
1430 | /* Ok, we know about this VD, let's give more details */ | |
b06e3095 | 1431 | printf(" Raid Devices[%d] : %d (", n, |
a8173e43 | 1432 | be16_to_cpu(vc->prim_elmnt_count)); |
1433 | for (i = 0; i < be16_to_cpu(vc->prim_elmnt_count); i++) { | |
b06e3095 | 1434 | int j; |
217dead4 | 1435 | int cnt = be16_to_cpu(sb->phys->max_pdes); |
b06e3095 | 1436 | for (j=0; j<cnt; j++) |
60931cf9 | 1437 | if (be32_eq(vc->phys_refnum[i], |
1438 | sb->phys->entries[j].refnum)) | |
b06e3095 N |
1439 | break; |
1440 | if (i) printf(" "); | |
1441 | if (j < cnt) | |
1442 | printf("%d", j); | |
1443 | else | |
1444 | printf("--"); | |
31bc5466 | 1445 | printf("@%lluK", (unsigned long long) be64_to_cpu(LBA_OFFSET(sb, vc)[i])/2); |
b06e3095 N |
1446 | } |
1447 | printf(")\n"); | |
1448 | if (vc->chunk_shift != 255) | |
613b0d17 N |
1449 | printf(" Chunk Size[%d] : %d sectors\n", n, |
1450 | 1 << vc->chunk_shift); | |
a322f70c DW |
1451 | printf(" Raid Level[%d] : %s\n", n, |
1452 | map_num(ddf_level, vc->prl)?:"-unknown-"); | |
1453 | if (vc->sec_elmnt_count != 1) { | |
1454 | printf(" Secondary Position[%d] : %d of %d\n", n, | |
1455 | vc->sec_elmnt_seq, vc->sec_elmnt_count); | |
1456 | printf(" Secondary Level[%d] : %s\n", n, | |
1457 | map_num(ddf_sec_level, vc->srl) ?: "-unknown-"); | |
1458 | } | |
1459 | printf(" Device Size[%d] : %llu\n", n, | |
9d0c6b70 | 1460 | be64_to_cpu(vc->blocks)/2); |
a322f70c | 1461 | printf(" Array Size[%d] : %llu\n", n, |
9d0c6b70 | 1462 | be64_to_cpu(vc->array_blocks)/2); |
a322f70c DW |
1463 | } |
1464 | } | |
1465 | ||
1466 | static void examine_vds(struct ddf_super *sb) | |
1467 | { | |
a8173e43 | 1468 | int cnt = be16_to_cpu(sb->virt->populated_vdes); |
fb9d0acb | 1469 | unsigned int i; |
a322f70c DW |
1470 | printf(" Virtual Disks : %d\n", cnt); |
1471 | ||
a8173e43 | 1472 | for (i = 0; i < be16_to_cpu(sb->virt->max_vdes); i++) { |
a322f70c | 1473 | struct virtual_entry *ve = &sb->virt->entries[i]; |
fb9d0acb | 1474 | if (all_ff(ve->guid)) |
1475 | continue; | |
b06e3095 | 1476 | printf("\n"); |
a322f70c DW |
1477 | printf(" VD GUID[%d] : ", i); print_guid(ve->guid, 1); |
1478 | printf("\n"); | |
a8173e43 | 1479 | printf(" unit[%d] : %d\n", i, be16_to_cpu(ve->unit)); |
a322f70c DW |
1480 | printf(" state[%d] : %s, %s%s\n", i, |
1481 | map_num(ddf_state, ve->state & 7), | |
cc83a819 N |
1482 | (ve->state & DDF_state_morphing) ? "Morphing, ": "", |
1483 | (ve->state & DDF_state_inconsistent)? "Not Consistent" : "Consistent"); | |
a322f70c | 1484 | printf(" init state[%d] : %s\n", i, |
cc83a819 | 1485 | map_num(ddf_init_state, ve->init_state&DDF_initstate_mask)); |
a322f70c | 1486 | printf(" access[%d] : %s\n", i, |
cc83a819 | 1487 | map_num(ddf_access, (ve->init_state & DDF_access_mask) >> 6)); |
a322f70c DW |
1488 | printf(" Name[%d] : %.16s\n", i, ve->name); |
1489 | examine_vd(i, sb, ve->guid); | |
1490 | } | |
1491 | if (cnt) printf("\n"); | |
1492 | } | |
1493 | ||
1494 | static void examine_pds(struct ddf_super *sb) | |
1495 | { | |
217dead4 | 1496 | int cnt = be16_to_cpu(sb->phys->max_pdes); |
a322f70c DW |
1497 | int i; |
1498 | struct dl *dl; | |
d2ec75fb | 1499 | int unlisted = 0; |
a322f70c | 1500 | printf(" Physical Disks : %d\n", cnt); |
962371a5 | 1501 | printf(" Number RefNo Size Device Type/State\n"); |
a322f70c | 1502 | |
d2ec75fb N |
1503 | for (dl = sb->dlist; dl; dl = dl->next) |
1504 | dl->displayed = 0; | |
1505 | ||
a322f70c DW |
1506 | for (i=0 ; i<cnt ; i++) { |
1507 | struct phys_disk_entry *pd = &sb->phys->entries[i]; | |
a8173e43 | 1508 | int type = be16_to_cpu(pd->type); |
1509 | int state = be16_to_cpu(pd->state); | |
a322f70c | 1510 | |
217dead4 N |
1511 | if (be32_to_cpu(pd->refnum) == 0xffffffff) |
1512 | /* Not in use */ | |
1513 | continue; | |
b06e3095 N |
1514 | //printf(" PD GUID[%d] : ", i); print_guid(pd->guid, 0); |
1515 | //printf("\n"); | |
1516 | printf(" %3d %08x ", i, | |
60931cf9 | 1517 | be32_to_cpu(pd->refnum)); |
613b0d17 | 1518 | printf("%8lluK ", |
9d0c6b70 | 1519 | be64_to_cpu(pd->config_size)>>1); |
b06e3095 | 1520 | for (dl = sb->dlist; dl ; dl = dl->next) { |
60931cf9 | 1521 | if (be32_eq(dl->disk.refnum, pd->refnum)) { |
b06e3095 N |
1522 | char *dv = map_dev(dl->major, dl->minor, 0); |
1523 | if (dv) { | |
962371a5 | 1524 | printf("%-15s", dv); |
b06e3095 N |
1525 | break; |
1526 | } | |
1527 | } | |
1528 | } | |
1529 | if (!dl) | |
962371a5 | 1530 | printf("%15s",""); |
d2ec75fb N |
1531 | else |
1532 | dl->displayed = 1; | |
b06e3095 | 1533 | printf(" %s%s%s%s%s", |
a322f70c | 1534 | (type&2) ? "active":"", |
b06e3095 | 1535 | (type&4) ? "Global-Spare":"", |
a322f70c DW |
1536 | (type&8) ? "spare" : "", |
1537 | (type&16)? ", foreign" : "", | |
1538 | (type&32)? "pass-through" : ""); | |
18cb4496 N |
1539 | if (state & DDF_Failed) |
1540 | /* This over-rides these three */ | |
1541 | state &= ~(DDF_Online|DDF_Rebuilding|DDF_Transition); | |
b06e3095 | 1542 | printf("/%s%s%s%s%s%s%s", |
a322f70c DW |
1543 | (state&1)? "Online": "Offline", |
1544 | (state&2)? ", Failed": "", | |
1545 | (state&4)? ", Rebuilding": "", | |
1546 | (state&8)? ", in-transition": "", | |
b06e3095 N |
1547 | (state&16)? ", SMART-errors": "", |
1548 | (state&32)? ", Unrecovered-Read-Errors": "", | |
a322f70c | 1549 | (state&64)? ", Missing" : ""); |
a322f70c DW |
1550 | printf("\n"); |
1551 | } | |
d2ec75fb N |
1552 | for (dl = sb->dlist; dl; dl = dl->next) { |
1553 | char *dv; | |
1554 | if (dl->displayed) | |
1555 | continue; | |
1556 | if (!unlisted) | |
1557 | printf(" Physical disks not in metadata!:\n"); | |
1558 | unlisted = 1; | |
1559 | dv = map_dev(dl->major, dl->minor, 0); | |
1560 | printf(" %08x %s\n", be32_to_cpu(dl->disk.refnum), | |
1561 | dv ? dv : "-unknown-"); | |
1562 | } | |
1563 | if (unlisted) | |
1564 | printf("\n"); | |
a322f70c DW |
1565 | } |
1566 | ||
1567 | static void examine_super_ddf(struct supertype *st, char *homehost) | |
1568 | { | |
1569 | struct ddf_super *sb = st->sb; | |
1570 | ||
60931cf9 | 1571 | printf(" Magic : %08x\n", be32_to_cpu(sb->anchor.magic)); |
a322f70c | 1572 | printf(" Version : %.8s\n", sb->anchor.revision); |
598f0d58 NB |
1573 | printf("Controller GUID : "); print_guid(sb->controller.guid, 0); |
1574 | printf("\n"); | |
1575 | printf(" Container GUID : "); print_guid(sb->anchor.guid, 1); | |
a322f70c | 1576 | printf("\n"); |
60931cf9 | 1577 | printf(" Seq : %08x\n", be32_to_cpu(sb->active->seq)); |
56cb05c4 | 1578 | printf(" Redundant hdr : %s\n", (be32_eq(sb->secondary.magic, |
60931cf9 | 1579 | DDF_HEADER_MAGIC) |
56cb05c4 | 1580 | ?"yes" : "no")); |
a322f70c DW |
1581 | examine_vds(sb); |
1582 | examine_pds(sb); | |
1583 | } | |
1584 | ||
bedbf68a | 1585 | static unsigned int get_vd_num_of_subarray(struct supertype *st) |
1586 | { | |
1587 | /* | |
1588 | * Figure out the VD number for this supertype. | |
1589 | * Returns DDF_CONTAINER for the container itself, | |
1590 | * and DDF_NOTFOUND on error. | |
1591 | */ | |
1592 | struct ddf_super *ddf = st->sb; | |
1593 | struct mdinfo *sra; | |
1594 | char *sub, *end; | |
1595 | unsigned int vcnum; | |
1596 | ||
1597 | if (*st->container_devnm == '\0') | |
1598 | return DDF_CONTAINER; | |
1599 | ||
1600 | sra = sysfs_read(-1, st->devnm, GET_VERSION); | |
1601 | if (!sra || sra->array.major_version != -1 || | |
1602 | sra->array.minor_version != -2 || | |
1603 | !is_subarray(sra->text_version)) | |
1604 | return DDF_NOTFOUND; | |
1605 | ||
1606 | sub = strchr(sra->text_version + 1, '/'); | |
1607 | if (sub != NULL) | |
1608 | vcnum = strtoul(sub + 1, &end, 10); | |
1609 | if (sub == NULL || *sub == '\0' || *end != '\0' || | |
a8173e43 | 1610 | vcnum >= be16_to_cpu(ddf->active->max_vd_entries)) |
bedbf68a | 1611 | return DDF_NOTFOUND; |
1612 | ||
1613 | return vcnum; | |
1614 | } | |
1615 | ||
061f2c6a | 1616 | static void brief_examine_super_ddf(struct supertype *st, int verbose) |
4737ae25 N |
1617 | { |
1618 | /* We just write a generic DDF ARRAY entry | |
1619 | */ | |
1620 | struct mdinfo info; | |
1621 | char nbuf[64]; | |
a5d85af7 | 1622 | getinfo_super_ddf(st, &info, NULL); |
4737ae25 N |
1623 | fname_from_uuid(st, &info, nbuf, ':'); |
1624 | ||
1625 | printf("ARRAY metadata=ddf UUID=%s\n", nbuf + 5); | |
1626 | } | |
1627 | ||
1628 | static void brief_examine_subarrays_ddf(struct supertype *st, int verbose) | |
a322f70c | 1629 | { |
56cb05c4 N |
1630 | /* We write a DDF ARRAY member entry for each vd, identifying container |
1631 | * by uuid and member by unit number and uuid. | |
a322f70c | 1632 | */ |
42dc2744 | 1633 | struct ddf_super *ddf = st->sb; |
ff54de6e | 1634 | struct mdinfo info; |
f21e18ca | 1635 | unsigned int i; |
ff54de6e | 1636 | char nbuf[64]; |
a5d85af7 | 1637 | getinfo_super_ddf(st, &info, NULL); |
ff54de6e | 1638 | fname_from_uuid(st, &info, nbuf, ':'); |
42dc2744 | 1639 | |
a8173e43 | 1640 | for (i = 0; i < be16_to_cpu(ddf->virt->max_vdes); i++) { |
42dc2744 N |
1641 | struct virtual_entry *ve = &ddf->virt->entries[i]; |
1642 | struct vcl vcl; | |
1643 | char nbuf1[64]; | |
a8b25633 | 1644 | char namebuf[17]; |
42dc2744 N |
1645 | if (all_ff(ve->guid)) |
1646 | continue; | |
1647 | memcpy(vcl.conf.guid, ve->guid, DDF_GUID_LEN); | |
1648 | ddf->currentconf =&vcl; | |
7087f02b | 1649 | vcl.vcnum = i; |
42dc2744 N |
1650 | uuid_from_super_ddf(st, info.uuid); |
1651 | fname_from_uuid(st, &info, nbuf1, ':'); | |
a8b25633 | 1652 | _ddf_array_name(namebuf, ddf, i); |
1653 | printf("ARRAY%s%s container=%s member=%d UUID=%s\n", | |
1654 | namebuf[0] == '\0' ? "" : " /dev/md/", namebuf, | |
42dc2744 N |
1655 | nbuf+5, i, nbuf1+5); |
1656 | } | |
a322f70c DW |
1657 | } |
1658 | ||
bceedeec N |
1659 | static void export_examine_super_ddf(struct supertype *st) |
1660 | { | |
1661 | struct mdinfo info; | |
1662 | char nbuf[64]; | |
a5d85af7 | 1663 | getinfo_super_ddf(st, &info, NULL); |
bceedeec N |
1664 | fname_from_uuid(st, &info, nbuf, ':'); |
1665 | printf("MD_METADATA=ddf\n"); | |
1666 | printf("MD_LEVEL=container\n"); | |
1667 | printf("MD_UUID=%s\n", nbuf+5); | |
cc9bfd9e | 1668 | printf("MD_DEVICES=%u\n", |
1669 | be16_to_cpu(((struct ddf_super *)st->sb)->phys->used_pdes)); | |
bceedeec | 1670 | } |
bceedeec | 1671 | |
74db60b0 N |
1672 | static int copy_metadata_ddf(struct supertype *st, int from, int to) |
1673 | { | |
1674 | void *buf; | |
1675 | unsigned long long dsize, offset; | |
1676 | int bytes; | |
1677 | struct ddf_header *ddf; | |
1678 | int written = 0; | |
1679 | ||
1680 | /* The meta consists of an anchor, a primary, and a secondary. | |
1681 | * This all lives at the end of the device. | |
1682 | * So it is easiest to find the earliest of primary and | |
1683 | * secondary, and copy everything from there. | |
1684 | * | |
56cb05c4 | 1685 | * Anchor is 512 from end. It contains primary_lba and secondary_lba |
74db60b0 N |
1686 | * we choose one of those |
1687 | */ | |
1688 | ||
1689 | if (posix_memalign(&buf, 4096, 4096) != 0) | |
1690 | return 1; | |
1691 | ||
1692 | if (!get_dev_size(from, NULL, &dsize)) | |
1693 | goto err; | |
1694 | ||
1695 | if (lseek64(from, dsize-512, 0) < 0) | |
1696 | goto err; | |
1697 | if (read(from, buf, 512) != 512) | |
1698 | goto err; | |
1699 | ddf = buf; | |
60931cf9 | 1700 | if (!be32_eq(ddf->magic, DDF_HEADER_MAGIC) || |
1701 | !be32_eq(calc_crc(ddf, 512), ddf->crc) || | |
74db60b0 N |
1702 | (memcmp(ddf->revision, DDF_REVISION_0, 8) != 0 && |
1703 | memcmp(ddf->revision, DDF_REVISION_2, 8) != 0)) | |
1704 | goto err; | |
1705 | ||
1706 | offset = dsize - 512; | |
9d0c6b70 | 1707 | if ((be64_to_cpu(ddf->primary_lba) << 9) < offset) |
1708 | offset = be64_to_cpu(ddf->primary_lba) << 9; | |
1709 | if ((be64_to_cpu(ddf->secondary_lba) << 9) < offset) | |
1710 | offset = be64_to_cpu(ddf->secondary_lba) << 9; | |
74db60b0 N |
1711 | |
1712 | bytes = dsize - offset; | |
1713 | ||
1714 | if (lseek64(from, offset, 0) < 0 || | |
1715 | lseek64(to, offset, 0) < 0) | |
1716 | goto err; | |
1717 | while (written < bytes) { | |
1718 | int n = bytes - written; | |
1719 | if (n > 4096) | |
1720 | n = 4096; | |
1721 | if (read(from, buf, n) != n) | |
1722 | goto err; | |
1723 | if (write(to, buf, n) != n) | |
1724 | goto err; | |
1725 | written += n; | |
1726 | } | |
1727 | free(buf); | |
1728 | return 0; | |
1729 | err: | |
1730 | free(buf); | |
1731 | return 1; | |
1732 | } | |
1733 | ||
a322f70c DW |
1734 | static void detail_super_ddf(struct supertype *st, char *homehost) |
1735 | { | |
ff84d052 N |
1736 | struct ddf_super *sb = st->sb; |
1737 | int cnt = be16_to_cpu(sb->virt->populated_vdes); | |
1738 | ||
65884368 | 1739 | printf(" Container GUID : "); print_guid(sb->anchor.guid, 1); |
ff84d052 | 1740 | printf("\n"); |
65884368 AP |
1741 | printf(" Seq : %08x\n", be32_to_cpu(sb->active->seq)); |
1742 | printf(" Virtual Disks : %d\n", cnt); | |
ff84d052 | 1743 | printf("\n"); |
a322f70c DW |
1744 | } |
1745 | ||
7087f02b | 1746 | static const char *vendors_with_variable_volume_UUID[] = { |
1747 | "LSI ", | |
1748 | }; | |
1749 | ||
1750 | static int volume_id_is_reliable(const struct ddf_super *ddf) | |
1751 | { | |
1c0aebc2 | 1752 | int n = ARRAY_SIZE(vendors_with_variable_volume_UUID); |
7087f02b | 1753 | int i; |
1754 | for (i = 0; i < n; i++) | |
1755 | if (!memcmp(ddf->controller.guid, | |
1756 | vendors_with_variable_volume_UUID[i], 8)) | |
1757 | return 0; | |
1758 | return 1; | |
1759 | } | |
1760 | ||
1761 | static void uuid_of_ddf_subarray(const struct ddf_super *ddf, | |
1762 | unsigned int vcnum, int uuid[4]) | |
1763 | { | |
1764 | char buf[DDF_GUID_LEN+18], sha[20], *p; | |
1765 | struct sha1_ctx ctx; | |
1766 | if (volume_id_is_reliable(ddf)) { | |
1767 | uuid_from_ddf_guid(ddf->virt->entries[vcnum].guid, uuid); | |
1768 | return; | |
1769 | } | |
1770 | /* | |
1771 | * Some fake RAID BIOSes (in particular, LSI ones) change the | |
1772 | * VD GUID at every boot. These GUIDs are not suitable for | |
1773 | * identifying an array. Luckily the header GUID appears to | |
1774 | * remain constant. | |
1775 | * We construct a pseudo-UUID from the header GUID and those | |
1776 | * properties of the subarray that we expect to remain constant. | |
1777 | */ | |
1778 | memset(buf, 0, sizeof(buf)); | |
1779 | p = buf; | |
1780 | memcpy(p, ddf->anchor.guid, DDF_GUID_LEN); | |
1781 | p += DDF_GUID_LEN; | |
1782 | memcpy(p, ddf->virt->entries[vcnum].name, 16); | |
1783 | p += 16; | |
1784 | *((__u16 *) p) = vcnum; | |
1785 | sha1_init_ctx(&ctx); | |
1786 | sha1_process_bytes(buf, sizeof(buf), &ctx); | |
1787 | sha1_finish_ctx(&ctx, sha); | |
1788 | memcpy(uuid, sha, 4*4); | |
1789 | } | |
1790 | ||
a322f70c DW |
1791 | static void brief_detail_super_ddf(struct supertype *st) |
1792 | { | |
ff54de6e N |
1793 | struct mdinfo info; |
1794 | char nbuf[64]; | |
bedbf68a | 1795 | struct ddf_super *ddf = st->sb; |
1796 | unsigned int vcnum = get_vd_num_of_subarray(st); | |
1797 | if (vcnum == DDF_CONTAINER) | |
1798 | uuid_from_super_ddf(st, info.uuid); | |
1799 | else if (vcnum == DDF_NOTFOUND) | |
1800 | return; | |
1801 | else | |
7087f02b | 1802 | uuid_of_ddf_subarray(ddf, vcnum, info.uuid); |
ff54de6e N |
1803 | fname_from_uuid(st, &info, nbuf,':'); |
1804 | printf(" UUID=%s", nbuf + 5); | |
a322f70c | 1805 | } |
a322f70c DW |
1806 | |
1807 | static int match_home_ddf(struct supertype *st, char *homehost) | |
1808 | { | |
1809 | /* It matches 'this' host if the controller is a | |
1810 | * Linux-MD controller with vendor_data matching | |
56cb05c4 N |
1811 | * the hostname. It would be nice if we could |
1812 | * test against controller found in /sys or somewhere... | |
a322f70c DW |
1813 | */ |
1814 | struct ddf_super *ddf = st->sb; | |
f21e18ca | 1815 | unsigned int len; |
d1d3482b N |
1816 | |
1817 | if (!homehost) | |
1818 | return 0; | |
1819 | len = strlen(homehost); | |
a322f70c DW |
1820 | |
1821 | return (memcmp(ddf->controller.guid, T10, 8) == 0 && | |
1822 | len < sizeof(ddf->controller.vendor_data) && | |
1823 | memcmp(ddf->controller.vendor_data, homehost,len) == 0 && | |
1824 | ddf->controller.vendor_data[len] == 0); | |
1825 | } | |
1826 | ||
baba3f4e | 1827 | static int find_index_in_bvd(const struct ddf_super *ddf, |
1828 | const struct vd_config *conf, unsigned int n, | |
1829 | unsigned int *n_bvd) | |
1830 | { | |
1831 | /* | |
56cb05c4 N |
1832 | * Find the index of the n-th valid physical disk in this BVD. |
1833 | * Unused entries can be sprinkled in with the used entries, | |
1834 | * but don't count. | |
baba3f4e | 1835 | */ |
1836 | unsigned int i, j; | |
56cb05c4 N |
1837 | for (i = 0, j = 0; |
1838 | i < ddf->mppe && j < be16_to_cpu(conf->prim_elmnt_count); | |
1839 | i++) { | |
60931cf9 | 1840 | if (be32_to_cpu(conf->phys_refnum[i]) != 0xffffffff) { |
baba3f4e | 1841 | if (n == j) { |
1842 | *n_bvd = i; | |
1843 | return 1; | |
1844 | } | |
1845 | j++; | |
1846 | } | |
1847 | } | |
1ade5cc1 N |
1848 | dprintf("couldn't find BVD member %u (total %u)\n", |
1849 | n, be16_to_cpu(conf->prim_elmnt_count)); | |
baba3f4e | 1850 | return 0; |
1851 | } | |
1852 | ||
56cb05c4 N |
1853 | /* Given a member array instance number, and a raid disk within that instance, |
1854 | * find the vd_config structure. The offset of the given disk in the phys_refnum | |
1855 | * table is returned in n_bvd. | |
1856 | * For two-level members with a secondary raid level the vd_config for | |
1857 | * the appropriate BVD is returned. | |
1858 | * The return value is always &vlc->conf, where vlc is returned in last pointer. | |
1859 | */ | |
baba3f4e | 1860 | static struct vd_config *find_vdcr(struct ddf_super *ddf, unsigned int inst, |
1861 | unsigned int n, | |
1862 | unsigned int *n_bvd, struct vcl **vcl) | |
a322f70c | 1863 | { |
7a7cc504 | 1864 | struct vcl *v; |
59e36268 | 1865 | |
baba3f4e | 1866 | for (v = ddf->conflist; v; v = v->next) { |
84e32e19 | 1867 | unsigned int nsec, ibvd = 0; |
baba3f4e | 1868 | struct vd_config *conf; |
1869 | if (inst != v->vcnum) | |
1870 | continue; | |
1871 | conf = &v->conf; | |
1872 | if (conf->sec_elmnt_count == 1) { | |
1873 | if (find_index_in_bvd(ddf, conf, n, n_bvd)) { | |
1874 | *vcl = v; | |
1875 | return conf; | |
1876 | } else | |
1877 | goto bad; | |
1878 | } | |
1879 | if (v->other_bvds == NULL) { | |
1ade5cc1 N |
1880 | pr_err("BUG: other_bvds is NULL, nsec=%u\n", |
1881 | conf->sec_elmnt_count); | |
baba3f4e | 1882 | goto bad; |
1883 | } | |
a8173e43 | 1884 | nsec = n / be16_to_cpu(conf->prim_elmnt_count); |
baba3f4e | 1885 | if (conf->sec_elmnt_seq != nsec) { |
1886 | for (ibvd = 1; ibvd < conf->sec_elmnt_count; ibvd++) { | |
baba3f4e | 1887 | if (v->other_bvds[ibvd-1]->sec_elmnt_seq |
1888 | == nsec) | |
1889 | break; | |
1890 | } | |
1891 | if (ibvd == conf->sec_elmnt_count) | |
1892 | goto bad; | |
1893 | conf = v->other_bvds[ibvd-1]; | |
1894 | } | |
1895 | if (!find_index_in_bvd(ddf, conf, | |
1896 | n - nsec*conf->sec_elmnt_count, n_bvd)) | |
1897 | goto bad; | |
1ade5cc1 N |
1898 | dprintf("found disk %u as member %u in bvd %d of array %u\n", |
1899 | n, *n_bvd, ibvd, inst); | |
baba3f4e | 1900 | *vcl = v; |
1901 | return conf; | |
1902 | } | |
1903 | bad: | |
1ade5cc1 | 1904 | pr_err("Could't find disk %d in array %u\n", n, inst); |
7a7cc504 NB |
1905 | return NULL; |
1906 | } | |
1907 | ||
60931cf9 | 1908 | static int find_phys(const struct ddf_super *ddf, be32 phys_refnum) |
7a7cc504 NB |
1909 | { |
1910 | /* Find the entry in phys_disk which has the given refnum | |
1911 | * and return it's index | |
1912 | */ | |
f21e18ca | 1913 | unsigned int i; |
a8173e43 | 1914 | for (i = 0; i < be16_to_cpu(ddf->phys->max_pdes); i++) |
60931cf9 | 1915 | if (be32_eq(ddf->phys->entries[i].refnum, phys_refnum)) |
7a7cc504 NB |
1916 | return i; |
1917 | return -1; | |
a322f70c DW |
1918 | } |
1919 | ||
bedbf68a | 1920 | static void uuid_from_ddf_guid(const char *guid, int uuid[4]) |
1921 | { | |
1922 | char buf[20]; | |
1923 | struct sha1_ctx ctx; | |
1924 | sha1_init_ctx(&ctx); | |
1925 | sha1_process_bytes(guid, DDF_GUID_LEN, &ctx); | |
1926 | sha1_finish_ctx(&ctx, buf); | |
1927 | memcpy(uuid, buf, 4*4); | |
1928 | } | |
1929 | ||
a322f70c DW |
1930 | static void uuid_from_super_ddf(struct supertype *st, int uuid[4]) |
1931 | { | |
1932 | /* The uuid returned here is used for: | |
1933 | * uuid to put into bitmap file (Create, Grow) | |
1934 | * uuid for backup header when saving critical section (Grow) | |
1935 | * comparing uuids when re-adding a device into an array | |
51006d85 N |
1936 | * In these cases the uuid required is that of the data-array, |
1937 | * not the device-set. | |
1938 | * uuid to recognise same set when adding a missing device back | |
1939 | * to an array. This is a uuid for the device-set. | |
613b0d17 | 1940 | * |
a322f70c DW |
1941 | * For each of these we can make do with a truncated |
1942 | * or hashed uuid rather than the original, as long as | |
1943 | * everyone agrees. | |
a322f70c DW |
1944 | * In the case of SVD we assume the BVD is of interest, |
1945 | * though that might be the case if a bitmap were made for | |
1946 | * a mirrored SVD - worry about that later. | |
1947 | * So we need to find the VD configuration record for the | |
1948 | * relevant BVD and extract the GUID and Secondary_Element_Seq. | |
1949 | * The first 16 bytes of the sha1 of these is used. | |
1950 | */ | |
1951 | struct ddf_super *ddf = st->sb; | |
d2ca6449 | 1952 | struct vcl *vcl = ddf->currentconf; |
a322f70c | 1953 | |
c5afc314 | 1954 | if (vcl) |
7087f02b | 1955 | uuid_of_ddf_subarray(ddf, vcl->vcnum, uuid); |
c5afc314 | 1956 | else |
7087f02b | 1957 | uuid_from_ddf_guid(ddf->anchor.guid, uuid); |
a322f70c DW |
1958 | } |
1959 | ||
a5d85af7 | 1960 | static void getinfo_super_ddf(struct supertype *st, struct mdinfo *info, char *map) |
a322f70c DW |
1961 | { |
1962 | struct ddf_super *ddf = st->sb; | |
a5d85af7 | 1963 | int map_disks = info->array.raid_disks; |
90fa1a29 | 1964 | __u32 *cptr; |
a322f70c | 1965 | |
78e44928 | 1966 | if (ddf->currentconf) { |
a5d85af7 | 1967 | getinfo_super_ddf_bvd(st, info, map); |
78e44928 NB |
1968 | return; |
1969 | } | |
95eeceeb | 1970 | memset(info, 0, sizeof(*info)); |
78e44928 | 1971 | |
a8173e43 | 1972 | info->array.raid_disks = be16_to_cpu(ddf->phys->used_pdes); |
a322f70c DW |
1973 | info->array.level = LEVEL_CONTAINER; |
1974 | info->array.layout = 0; | |
1975 | info->array.md_minor = -1; | |
90fa1a29 JS |
1976 | cptr = (__u32 *)(ddf->anchor.guid + 16); |
1977 | info->array.ctime = DECADE + __be32_to_cpu(*cptr); | |
1978 | ||
a322f70c | 1979 | info->array.chunk_size = 0; |
510242aa | 1980 | info->container_enough = 1; |
a322f70c | 1981 | |
56cb05c4 N |
1982 | info->disk.major = 0; |
1983 | info->disk.minor = 0; | |
cba0191b | 1984 | if (ddf->dlist) { |
f0e876ce | 1985 | struct phys_disk_entry *pde = NULL; |
60931cf9 | 1986 | info->disk.number = be32_to_cpu(ddf->dlist->disk.refnum); |
59e36268 | 1987 | info->disk.raid_disk = find_phys(ddf, ddf->dlist->disk.refnum); |
d2ca6449 | 1988 | |
9d0c6b70 | 1989 | info->data_offset = be64_to_cpu(ddf->phys-> |
613b0d17 N |
1990 | entries[info->disk.raid_disk]. |
1991 | config_size); | |
d2ca6449 | 1992 | info->component_size = ddf->dlist->size - info->data_offset; |
f0e876ce N |
1993 | if (info->disk.raid_disk >= 0) |
1994 | pde = ddf->phys->entries + info->disk.raid_disk; | |
1995 | if (pde && | |
4fe903aa N |
1996 | !(be16_to_cpu(pde->state) & DDF_Failed) && |
1997 | !(be16_to_cpu(pde->state) & DDF_Missing)) | |
f0e876ce N |
1998 | info->disk.state = (1 << MD_DISK_SYNC) | (1 << MD_DISK_ACTIVE); |
1999 | else | |
2000 | info->disk.state = 1 << MD_DISK_FAULTY; | |
eba2859f | 2001 | |
cba0191b | 2002 | } else { |
25532b88 | 2003 | /* There should always be a dlist, but just in case...*/ |
cba0191b | 2004 | info->disk.number = -1; |
661dce36 | 2005 | info->disk.raid_disk = -1; |
f0e876ce | 2006 | info->disk.state = (1 << MD_DISK_SYNC) | (1 << MD_DISK_ACTIVE); |
cba0191b | 2007 | } |
2a351cea N |
2008 | info->events = be32_to_cpu(ddf->active->seq); |
2009 | info->array.utime = DECADE + be32_to_cpu(ddf->active->timestamp); | |
a19c88b8 | 2010 | |
921d9e16 | 2011 | info->recovery_start = MaxSector; |
a19c88b8 | 2012 | info->reshape_active = 0; |
6e75048b | 2013 | info->recovery_blocked = 0; |
c5afc314 | 2014 | info->name[0] = 0; |
a322f70c | 2015 | |
f35f2525 N |
2016 | info->array.major_version = -1; |
2017 | info->array.minor_version = -2; | |
159c3a1a | 2018 | strcpy(info->text_version, "ddf"); |
a67dd8cc | 2019 | info->safe_mode_delay = 0; |
159c3a1a | 2020 | |
c5afc314 | 2021 | uuid_from_super_ddf(st, info->uuid); |
a322f70c | 2022 | |
a5d85af7 | 2023 | if (map) { |
708997ff N |
2024 | int i, e = 0; |
2025 | int max = be16_to_cpu(ddf->phys->max_pdes); | |
2026 | for (i = e = 0 ; i < map_disks ; i++, e++) { | |
2027 | while (e < max && | |
2028 | be32_to_cpu(ddf->phys->entries[e].refnum) == 0xffffffff) | |
2029 | e++; | |
2030 | if (i < info->array.raid_disks && e < max && | |
2031 | !(be16_to_cpu(ddf->phys->entries[e].state) | |
a8173e43 | 2032 | & DDF_Failed)) |
a5d85af7 N |
2033 | map[i] = 1; |
2034 | else | |
2035 | map[i] = 0; | |
2036 | } | |
2037 | } | |
a322f70c DW |
2038 | } |
2039 | ||
8bf989d8 | 2040 | /* size of name must be at least 17 bytes! */ |
2041 | static void _ddf_array_name(char *name, const struct ddf_super *ddf, int i) | |
2042 | { | |
2043 | int j; | |
2044 | memcpy(name, ddf->virt->entries[i].name, 16); | |
2045 | name[16] = 0; | |
2046 | for(j = 0; j < 16; j++) | |
2047 | if (name[j] == ' ') | |
2048 | name[j] = 0; | |
2049 | } | |
2050 | ||
a5d85af7 | 2051 | static void getinfo_super_ddf_bvd(struct supertype *st, struct mdinfo *info, char *map) |
a322f70c DW |
2052 | { |
2053 | struct ddf_super *ddf = st->sb; | |
d2ca6449 NB |
2054 | struct vcl *vc = ddf->currentconf; |
2055 | int cd = ddf->currentdev; | |
ddf94a43 | 2056 | int n_prim; |
db42fa9b | 2057 | int j; |
1e60caeb | 2058 | struct dl *dl = NULL; |
a5d85af7 | 2059 | int map_disks = info->array.raid_disks; |
90fa1a29 | 2060 | __u32 *cptr; |
ddf94a43 | 2061 | struct vd_config *conf; |
a322f70c | 2062 | |
95eeceeb | 2063 | memset(info, 0, sizeof(*info)); |
8a2848a7 | 2064 | if (layout_ddf2md(&vc->conf, &info->array) == -1) |
2065 | return; | |
a322f70c | 2066 | info->array.md_minor = -1; |
90fa1a29 JS |
2067 | cptr = (__u32 *)(vc->conf.guid + 16); |
2068 | info->array.ctime = DECADE + __be32_to_cpu(*cptr); | |
60931cf9 | 2069 | info->array.utime = DECADE + be32_to_cpu(vc->conf.timestamp); |
d2ca6449 | 2070 | info->array.chunk_size = 512 << vc->conf.chunk_shift; |
55e00074 | 2071 | info->custom_array_size = be64_to_cpu(vc->conf.array_blocks); |
d2ca6449 | 2072 | |
ddf94a43 | 2073 | conf = &vc->conf; |
a8173e43 | 2074 | n_prim = be16_to_cpu(conf->prim_elmnt_count); |
ddf94a43 | 2075 | if (conf->sec_elmnt_count > 1 && cd >= n_prim) { |
2076 | int ibvd = cd / n_prim - 1; | |
2077 | cd %= n_prim; | |
2078 | conf = vc->other_bvds[ibvd]; | |
2079 | } | |
2080 | ||
f21e18ca | 2081 | if (cd >= 0 && (unsigned)cd < ddf->mppe) { |
57a66662 | 2082 | info->data_offset = |
9d0c6b70 | 2083 | be64_to_cpu(LBA_OFFSET(ddf, conf)[cd]); |
d2ca6449 NB |
2084 | if (vc->block_sizes) |
2085 | info->component_size = vc->block_sizes[cd]; | |
2086 | else | |
9d0c6b70 | 2087 | info->component_size = be64_to_cpu(conf->blocks); |
a322f70c | 2088 | |
0d255ff8 N |
2089 | for (dl = ddf->dlist; dl ; dl = dl->next) |
2090 | if (be32_eq(dl->disk.refnum, conf->phys_refnum[cd])) | |
2091 | break; | |
2092 | } | |
fb204fb2 | 2093 | |
a322f70c DW |
2094 | info->disk.major = 0; |
2095 | info->disk.minor = 0; | |
fb204fb2 | 2096 | info->disk.state = 0; |
41bcbc14 | 2097 | if (dl && dl->pdnum >= 0) { |
8592f29d N |
2098 | info->disk.major = dl->major; |
2099 | info->disk.minor = dl->minor; | |
7c3fb3ec | 2100 | info->disk.raid_disk = cd + conf->sec_elmnt_seq |
a8173e43 | 2101 | * be16_to_cpu(conf->prim_elmnt_count); |
fb204fb2 | 2102 | info->disk.number = dl->pdnum; |
f0e876ce N |
2103 | info->disk.state = 0; |
2104 | if (info->disk.number >= 0 && | |
2105 | (be16_to_cpu(ddf->phys->entries[info->disk.number].state) & DDF_Online) && | |
2106 | !(be16_to_cpu(ddf->phys->entries[info->disk.number].state) & DDF_Failed)) | |
2107 | info->disk.state = (1<<MD_DISK_SYNC)|(1<<MD_DISK_ACTIVE); | |
eba2859f | 2108 | info->events = be32_to_cpu(ddf->active->seq); |
8592f29d | 2109 | } |
a322f70c | 2110 | |
103f2410 NB |
2111 | info->container_member = ddf->currentconf->vcnum; |
2112 | ||
921d9e16 | 2113 | info->recovery_start = MaxSector; |
80d26cb2 | 2114 | info->resync_start = 0; |
624c5ad4 | 2115 | info->reshape_active = 0; |
6e75048b | 2116 | info->recovery_blocked = 0; |
80d26cb2 NB |
2117 | if (!(ddf->virt->entries[info->container_member].state |
2118 | & DDF_state_inconsistent) && | |
2119 | (ddf->virt->entries[info->container_member].init_state | |
2120 | & DDF_initstate_mask) | |
2121 | == DDF_init_full) | |
b7528a20 | 2122 | info->resync_start = MaxSector; |
80d26cb2 | 2123 | |
a322f70c DW |
2124 | uuid_from_super_ddf(st, info->uuid); |
2125 | ||
f35f2525 N |
2126 | info->array.major_version = -1; |
2127 | info->array.minor_version = -2; | |
9b63e648 | 2128 | sprintf(info->text_version, "/%s/%d", |
4dd2df09 | 2129 | st->container_devnm, |
9b63e648 | 2130 | info->container_member); |
5684fff6 | 2131 | info->safe_mode_delay = DDF_SAFE_MODE_DELAY; |
159c3a1a | 2132 | |
8bf989d8 | 2133 | _ddf_array_name(info->name, ddf, info->container_member); |
a5d85af7 N |
2134 | |
2135 | if (map) | |
2136 | for (j = 0; j < map_disks; j++) { | |
2137 | map[j] = 0; | |
2138 | if (j < info->array.raid_disks) { | |
2139 | int i = find_phys(ddf, vc->conf.phys_refnum[j]); | |
613b0d17 | 2140 | if (i >= 0 && |
a8173e43 | 2141 | (be16_to_cpu(ddf->phys->entries[i].state) |
2142 | & DDF_Online) && | |
2143 | !(be16_to_cpu(ddf->phys->entries[i].state) | |
2144 | & DDF_Failed)) | |
a5d85af7 N |
2145 | map[i] = 1; |
2146 | } | |
2147 | } | |
a322f70c DW |
2148 | } |
2149 | ||
2150 | static int update_super_ddf(struct supertype *st, struct mdinfo *info, | |
2151 | char *update, | |
2152 | char *devname, int verbose, | |
2153 | int uuid_set, char *homehost) | |
2154 | { | |
2155 | /* For 'assemble' and 'force' we need to return non-zero if any | |
2156 | * change was made. For others, the return value is ignored. | |
2157 | * Update options are: | |
2158 | * force-one : This device looks a bit old but needs to be included, | |
2159 | * update age info appropriately. | |
2160 | * assemble: clear any 'faulty' flag to allow this device to | |
2161 | * be assembled. | |
2162 | * force-array: Array is degraded but being forced, mark it clean | |
2163 | * if that will be needed to assemble it. | |
2164 | * | |
2165 | * newdev: not used ???? | |
2166 | * grow: Array has gained a new device - this is currently for | |
2167 | * linear only | |
2168 | * resync: mark as dirty so a resync will happen. | |
59e36268 | 2169 | * uuid: Change the uuid of the array to match what is given |
a322f70c DW |
2170 | * homehost: update the recorded homehost |
2171 | * name: update the name - preserving the homehost | |
2172 | * _reshape_progress: record new reshape_progress position. | |
2173 | * | |
2174 | * Following are not relevant for this version: | |
2175 | * sparc2.2 : update from old dodgey metadata | |
2176 | * super-minor: change the preferred_minor number | |
2177 | * summaries: update redundant counters. | |
2178 | */ | |
2179 | int rv = 0; | |
2180 | // struct ddf_super *ddf = st->sb; | |
7a7cc504 | 2181 | // struct vd_config *vd = find_vdcr(ddf, info->container_member); |
a322f70c DW |
2182 | // struct virtual_entry *ve = find_ve(ddf); |
2183 | ||
a322f70c | 2184 | /* we don't need to handle "force-*" or "assemble" as |
56cb05c4 | 2185 | * there is no need to 'trick' the kernel. When the metadata is |
a322f70c DW |
2186 | * first updated to activate the array, all the implied modifications |
2187 | * will just happen. | |
2188 | */ | |
2189 | ||
2190 | if (strcmp(update, "grow") == 0) { | |
2191 | /* FIXME */ | |
1e2b2765 | 2192 | } else if (strcmp(update, "resync") == 0) { |
a322f70c | 2193 | // info->resync_checkpoint = 0; |
1e2b2765 | 2194 | } else if (strcmp(update, "homehost") == 0) { |
a322f70c DW |
2195 | /* homehost is stored in controller->vendor_data, |
2196 | * or it is when we are the vendor | |
2197 | */ | |
2198 | // if (info->vendor_is_local) | |
2199 | // strcpy(ddf->controller.vendor_data, homehost); | |
1e2b2765 | 2200 | rv = -1; |
f49208ec | 2201 | } else if (strcmp(update, "name") == 0) { |
a322f70c DW |
2202 | /* name is stored in virtual_entry->name */ |
2203 | // memset(ve->name, ' ', 16); | |
2204 | // strncpy(ve->name, info->name, 16); | |
1e2b2765 | 2205 | rv = -1; |
f49208ec | 2206 | } else if (strcmp(update, "_reshape_progress") == 0) { |
a322f70c | 2207 | /* We don't support reshape yet */ |
f49208ec N |
2208 | } else if (strcmp(update, "assemble") == 0 ) { |
2209 | /* Do nothing, just succeed */ | |
2210 | rv = 0; | |
1e2b2765 N |
2211 | } else |
2212 | rv = -1; | |
a322f70c DW |
2213 | |
2214 | // update_all_csum(ddf); | |
2215 | ||
2216 | return rv; | |
2217 | } | |
2218 | ||
5f8097be NB |
2219 | static void make_header_guid(char *guid) |
2220 | { | |
60931cf9 | 2221 | be32 stamp; |
5f8097be NB |
2222 | /* Create a DDF Header of Virtual Disk GUID */ |
2223 | ||
2224 | /* 24 bytes of fiction required. | |
2225 | * first 8 are a 'vendor-id' - "Linux-MD" | |
2226 | * next 8 are controller type.. how about 0X DEAD BEEF 0000 0000 | |
2227 | * Remaining 8 random number plus timestamp | |
2228 | */ | |
2229 | memcpy(guid, T10, sizeof(T10)); | |
60931cf9 | 2230 | stamp = cpu_to_be32(0xdeadbeef); |
5f8097be | 2231 | memcpy(guid+8, &stamp, 4); |
60931cf9 | 2232 | stamp = cpu_to_be32(0); |
5f8097be | 2233 | memcpy(guid+12, &stamp, 4); |
60931cf9 | 2234 | stamp = cpu_to_be32(time(0) - DECADE); |
5f8097be | 2235 | memcpy(guid+16, &stamp, 4); |
60931cf9 | 2236 | stamp._v32 = random32(); |
5f8097be | 2237 | memcpy(guid+20, &stamp, 4); |
5f8097be | 2238 | } |
59e36268 | 2239 | |
fb9d0acb | 2240 | static unsigned int find_unused_vde(const struct ddf_super *ddf) |
2241 | { | |
2242 | unsigned int i; | |
a8173e43 | 2243 | for (i = 0; i < be16_to_cpu(ddf->virt->max_vdes); i++) { |
fb9d0acb | 2244 | if (all_ff(ddf->virt->entries[i].guid)) |
2245 | return i; | |
2246 | } | |
2247 | return DDF_NOTFOUND; | |
2248 | } | |
2249 | ||
2250 | static unsigned int find_vde_by_name(const struct ddf_super *ddf, | |
2251 | const char *name) | |
2252 | { | |
2253 | unsigned int i; | |
2254 | if (name == NULL) | |
2255 | return DDF_NOTFOUND; | |
a8173e43 | 2256 | for (i = 0; i < be16_to_cpu(ddf->virt->max_vdes); i++) { |
fb9d0acb | 2257 | if (all_ff(ddf->virt->entries[i].guid)) |
2258 | continue; | |
2259 | if (!strncmp(name, ddf->virt->entries[i].name, | |
2260 | sizeof(ddf->virt->entries[i].name))) | |
2261 | return i; | |
2262 | } | |
2263 | return DDF_NOTFOUND; | |
2264 | } | |
2265 | ||
2266 | static unsigned int find_vde_by_guid(const struct ddf_super *ddf, | |
2267 | const char *guid) | |
2268 | { | |
2269 | unsigned int i; | |
2270 | if (guid == NULL || all_ff(guid)) | |
2271 | return DDF_NOTFOUND; | |
a8173e43 | 2272 | for (i = 0; i < be16_to_cpu(ddf->virt->max_vdes); i++) |
fb9d0acb | 2273 | if (!memcmp(ddf->virt->entries[i].guid, guid, DDF_GUID_LEN)) |
2274 | return i; | |
2275 | return DDF_NOTFOUND; | |
2276 | } | |
2277 | ||
a322f70c DW |
2278 | static int init_super_ddf(struct supertype *st, |
2279 | mdu_array_info_t *info, | |
5308f117 | 2280 | struct shape *s, char *name, char *homehost, |
83cd1e97 | 2281 | int *uuid, unsigned long long data_offset) |
a322f70c DW |
2282 | { |
2283 | /* This is primarily called by Create when creating a new array. | |
2284 | * We will then get add_to_super called for each component, and then | |
2285 | * write_init_super called to write it out to each device. | |
2286 | * For DDF, Create can create on fresh devices or on a pre-existing | |
2287 | * array. | |
2288 | * To create on a pre-existing array a different method will be called. | |
2289 | * This one is just for fresh drives. | |
2290 | * | |
2291 | * We need to create the entire 'ddf' structure which includes: | |
2292 | * DDF headers - these are easy. | |
2293 | * Controller data - a Sector describing this controller .. not that | |
56cb05c4 | 2294 | * this is a controller exactly. |
a322f70c | 2295 | * Physical Disk Record - one entry per device, so |
56cb05c4 | 2296 | * leave plenty of space. |
a322f70c | 2297 | * Virtual Disk Records - again, just leave plenty of space. |
56cb05c4 N |
2298 | * This just lists VDs, doesn't give details. |
2299 | * Config records - describe the VDs that use this disk | |
a322f70c DW |
2300 | * DiskData - describes 'this' device. |
2301 | * BadBlockManagement - empty | |
2302 | * Diag Space - empty | |
2303 | * Vendor Logs - Could we put bitmaps here? | |
2304 | * | |
2305 | */ | |
2306 | struct ddf_super *ddf; | |
2307 | char hostname[17]; | |
2308 | int hostlen; | |
a322f70c DW |
2309 | int max_phys_disks, max_virt_disks; |
2310 | unsigned long long sector; | |
2311 | int clen; | |
2312 | int i; | |
2313 | int pdsize, vdsize; | |
2314 | struct phys_disk *pd; | |
2315 | struct virtual_disk *vd; | |
2316 | ||
78e44928 | 2317 | if (st->sb) |
5308f117 | 2318 | return init_super_ddf_bvd(st, info, s->size, name, homehost, uuid, |
83cd1e97 | 2319 | data_offset); |
ba7eb04f | 2320 | |
3d2c4fc7 | 2321 | if (posix_memalign((void**)&ddf, 512, sizeof(*ddf)) != 0) { |
1ade5cc1 | 2322 | pr_err("could not allocate superblock\n"); |
3d2c4fc7 DW |
2323 | return 0; |
2324 | } | |
6264b437 | 2325 | memset(ddf, 0, sizeof(*ddf)); |
955e9ea1 DW |
2326 | st->sb = ddf; |
2327 | ||
2328 | if (info == NULL) { | |
2329 | /* zeroing superblock */ | |
2330 | return 0; | |
2331 | } | |
a322f70c DW |
2332 | |
2333 | /* At least 32MB *must* be reserved for the ddf. So let's just | |
2334 | * start 32MB from the end, and put the primary header there. | |
2335 | * Don't do secondary for now. | |
2336 | * We don't know exactly where that will be yet as it could be | |
56cb05c4 | 2337 | * different on each device. So just set up the lengths. |
a322f70c DW |
2338 | */ |
2339 | ||
2340 | ddf->anchor.magic = DDF_HEADER_MAGIC; | |
5f8097be | 2341 | make_header_guid(ddf->anchor.guid); |
a322f70c | 2342 | |
59e36268 | 2343 | memcpy(ddf->anchor.revision, DDF_REVISION_2, 8); |
60931cf9 | 2344 | ddf->anchor.seq = cpu_to_be32(1); |
2345 | ddf->anchor.timestamp = cpu_to_be32(time(0) - DECADE); | |
a322f70c DW |
2346 | ddf->anchor.openflag = 0xFF; |
2347 | ddf->anchor.foreignflag = 0; | |
2348 | ddf->anchor.enforcegroups = 0; /* Is this best?? */ | |
2349 | ddf->anchor.pad0 = 0xff; | |
2350 | memset(ddf->anchor.pad1, 0xff, 12); | |
2351 | memset(ddf->anchor.header_ext, 0xff, 32); | |
9d0c6b70 | 2352 | ddf->anchor.primary_lba = cpu_to_be64(~(__u64)0); |
2353 | ddf->anchor.secondary_lba = cpu_to_be64(~(__u64)0); | |
a322f70c DW |
2354 | ddf->anchor.type = DDF_HEADER_ANCHOR; |
2355 | memset(ddf->anchor.pad2, 0xff, 3); | |
60931cf9 | 2356 | ddf->anchor.workspace_len = cpu_to_be32(32768); /* Must be reserved */ |
9d0c6b70 | 2357 | /* Put this at bottom of 32M reserved.. */ |
2358 | ddf->anchor.workspace_lba = cpu_to_be64(~(__u64)0); | |
56cb05c4 | 2359 | max_phys_disks = 1023; /* Should be enough, 4095 is also allowed */ |
a8173e43 | 2360 | ddf->anchor.max_pd_entries = cpu_to_be16(max_phys_disks); |
56cb05c4 N |
2361 | max_virt_disks = 255; /* 15, 63, 255, 1024, 4095 are all allowed */ |
2362 | ddf->anchor.max_vd_entries = cpu_to_be16(max_virt_disks); | |
a322f70c | 2363 | ddf->max_part = 64; |
56cb05c4 N |
2364 | ddf->anchor.max_partitions = cpu_to_be16(ddf->max_part); |
2365 | ddf->mppe = 256; /* 16, 64, 256, 1024, 4096 are all allowed */ | |
59e36268 | 2366 | ddf->conf_rec_len = 1 + ROUND_UP(ddf->mppe * (4+8), 512)/512; |
a8173e43 | 2367 | ddf->anchor.config_record_len = cpu_to_be16(ddf->conf_rec_len); |
2368 | ddf->anchor.max_primary_element_entries = cpu_to_be16(ddf->mppe); | |
a322f70c | 2369 | memset(ddf->anchor.pad3, 0xff, 54); |
56cb05c4 | 2370 | /* Controller section is one sector long immediately |
a322f70c DW |
2371 | * after the ddf header */ |
2372 | sector = 1; | |
60931cf9 | 2373 | ddf->anchor.controller_section_offset = cpu_to_be32(sector); |
2374 | ddf->anchor.controller_section_length = cpu_to_be32(1); | |
a322f70c DW |
2375 | sector += 1; |
2376 | ||
2377 | /* phys is 8 sectors after that */ | |
2378 | pdsize = ROUND_UP(sizeof(struct phys_disk) + | |
2379 | sizeof(struct phys_disk_entry)*max_phys_disks, | |
2380 | 512); | |
2381 | switch(pdsize/512) { | |
2382 | case 2: case 8: case 32: case 128: case 512: break; | |
2383 | default: abort(); | |
2384 | } | |
60931cf9 | 2385 | ddf->anchor.phys_section_offset = cpu_to_be32(sector); |
a322f70c | 2386 | ddf->anchor.phys_section_length = |
60931cf9 | 2387 | cpu_to_be32(pdsize/512); /* max_primary_element_entries/8 */ |
a322f70c DW |
2388 | sector += pdsize/512; |
2389 | ||
2390 | /* virt is another 32 sectors */ | |
2391 | vdsize = ROUND_UP(sizeof(struct virtual_disk) + | |
2392 | sizeof(struct virtual_entry) * max_virt_disks, | |
2393 | 512); | |
2394 | switch(vdsize/512) { | |
2395 | case 2: case 8: case 32: case 128: case 512: break; | |
2396 | default: abort(); | |
2397 | } | |
60931cf9 | 2398 | ddf->anchor.virt_section_offset = cpu_to_be32(sector); |
a322f70c | 2399 | ddf->anchor.virt_section_length = |
60931cf9 | 2400 | cpu_to_be32(vdsize/512); /* max_vd_entries/8 */ |
a322f70c DW |
2401 | sector += vdsize/512; |
2402 | ||
59e36268 | 2403 | clen = ddf->conf_rec_len * (ddf->max_part+1); |
60931cf9 | 2404 | ddf->anchor.config_section_offset = cpu_to_be32(sector); |
2405 | ddf->anchor.config_section_length = cpu_to_be32(clen); | |
a322f70c DW |
2406 | sector += clen; |
2407 | ||
60931cf9 | 2408 | ddf->anchor.data_section_offset = cpu_to_be32(sector); |
2409 | ddf->anchor.data_section_length = cpu_to_be32(1); | |
a322f70c DW |
2410 | sector += 1; |
2411 | ||
60931cf9 | 2412 | ddf->anchor.bbm_section_length = cpu_to_be32(0); |
2413 | ddf->anchor.bbm_section_offset = cpu_to_be32(0xFFFFFFFF); | |
2414 | ddf->anchor.diag_space_length = cpu_to_be32(0); | |
2415 | ddf->anchor.diag_space_offset = cpu_to_be32(0xFFFFFFFF); | |
2416 | ddf->anchor.vendor_length = cpu_to_be32(0); | |
2417 | ddf->anchor.vendor_offset = cpu_to_be32(0xFFFFFFFF); | |
a322f70c DW |
2418 | |
2419 | memset(ddf->anchor.pad4, 0xff, 256); | |
2420 | ||
2421 | memcpy(&ddf->primary, &ddf->anchor, 512); | |
2422 | memcpy(&ddf->secondary, &ddf->anchor, 512); | |
2423 | ||
2424 | ddf->primary.openflag = 1; /* I guess.. */ | |
2425 | ddf->primary.type = DDF_HEADER_PRIMARY; | |
2426 | ||
2427 | ddf->secondary.openflag = 1; /* I guess.. */ | |
2428 | ddf->secondary.type = DDF_HEADER_SECONDARY; | |
2429 | ||
2430 | ddf->active = &ddf->primary; | |
2431 | ||
2432 | ddf->controller.magic = DDF_CONTROLLER_MAGIC; | |
2433 | ||
2434 | /* 24 more bytes of fiction required. | |
2435 | * first 8 are a 'vendor-id' - "Linux-MD" | |
2436 | * Remaining 16 are serial number.... maybe a hostname would do? | |
2437 | */ | |
2438 | memcpy(ddf->controller.guid, T10, sizeof(T10)); | |
1ba6bff9 DW |
2439 | gethostname(hostname, sizeof(hostname)); |
2440 | hostname[sizeof(hostname) - 1] = 0; | |
a322f70c DW |
2441 | hostlen = strlen(hostname); |
2442 | memcpy(ddf->controller.guid + 24 - hostlen, hostname, hostlen); | |
2443 | for (i = strlen(T10) ; i+hostlen < 24; i++) | |
2444 | ddf->controller.guid[i] = ' '; | |
2445 | ||
a8173e43 | 2446 | ddf->controller.type.vendor_id = cpu_to_be16(0xDEAD); |
2447 | ddf->controller.type.device_id = cpu_to_be16(0xBEEF); | |
2448 | ddf->controller.type.sub_vendor_id = cpu_to_be16(0); | |
2449 | ddf->controller.type.sub_device_id = cpu_to_be16(0); | |
a322f70c DW |
2450 | memcpy(ddf->controller.product_id, "What Is My PID??", 16); |
2451 | memset(ddf->controller.pad, 0xff, 8); | |
2452 | memset(ddf->controller.vendor_data, 0xff, 448); | |
a9e1c11d N |
2453 | if (homehost && strlen(homehost) < 440) |
2454 | strcpy((char*)ddf->controller.vendor_data, homehost); | |
a322f70c | 2455 | |
3d2c4fc7 | 2456 | if (posix_memalign((void**)&pd, 512, pdsize) != 0) { |
1ade5cc1 | 2457 | pr_err("could not allocate pd\n"); |
3d2c4fc7 DW |
2458 | return 0; |
2459 | } | |
6416d527 | 2460 | ddf->phys = pd; |
a322f70c DW |
2461 | ddf->pdsize = pdsize; |
2462 | ||
2463 | memset(pd, 0xff, pdsize); | |
2464 | memset(pd, 0, sizeof(*pd)); | |
076515ba | 2465 | pd->magic = DDF_PHYS_RECORDS_MAGIC; |
a8173e43 | 2466 | pd->used_pdes = cpu_to_be16(0); |
2467 | pd->max_pdes = cpu_to_be16(max_phys_disks); | |
a322f70c | 2468 | memset(pd->pad, 0xff, 52); |
4a3ca8ac | 2469 | for (i = 0; i < max_phys_disks; i++) |
2470 | memset(pd->entries[i].guid, 0xff, DDF_GUID_LEN); | |
a322f70c | 2471 | |
3d2c4fc7 | 2472 | if (posix_memalign((void**)&vd, 512, vdsize) != 0) { |
1ade5cc1 | 2473 | pr_err("could not allocate vd\n"); |
3d2c4fc7 DW |
2474 | return 0; |
2475 | } | |
6416d527 | 2476 | ddf->virt = vd; |
a322f70c DW |
2477 | ddf->vdsize = vdsize; |
2478 | memset(vd, 0, vdsize); | |
2479 | vd->magic = DDF_VIRT_RECORDS_MAGIC; | |
a8173e43 | 2480 | vd->populated_vdes = cpu_to_be16(0); |
2481 | vd->max_vdes = cpu_to_be16(max_virt_disks); | |
a322f70c DW |
2482 | memset(vd->pad, 0xff, 52); |
2483 | ||
5f8097be NB |
2484 | for (i=0; i<max_virt_disks; i++) |
2485 | memset(&vd->entries[i], 0xff, sizeof(struct virtual_entry)); | |
2486 | ||
a322f70c | 2487 | st->sb = ddf; |
609ce161 | 2488 | ddf_set_updates_pending(ddf, NULL); |
a322f70c DW |
2489 | return 1; |
2490 | } | |
2491 | ||
5f8097be NB |
2492 | static int chunk_to_shift(int chunksize) |
2493 | { | |
2494 | return ffs(chunksize/512)-1; | |
2495 | } | |
2496 | ||
59e36268 NB |
2497 | struct extent { |
2498 | unsigned long long start, size; | |
2499 | }; | |
78e44928 | 2500 | static int cmp_extent(const void *av, const void *bv) |
59e36268 NB |
2501 | { |
2502 | const struct extent *a = av; | |
2503 | const struct extent *b = bv; | |
2504 | if (a->start < b->start) | |
2505 | return -1; | |
2506 | if (a->start > b->start) | |
2507 | return 1; | |
2508 | return 0; | |
2509 | } | |
2510 | ||
78e44928 | 2511 | static struct extent *get_extents(struct ddf_super *ddf, struct dl *dl) |
59e36268 | 2512 | { |
fca65520 | 2513 | /* Find a list of used extents on the given physical device |
59e36268 NB |
2514 | * (dnum) of the given ddf. |
2515 | * Return a malloced array of 'struct extent' | |
59e36268 NB |
2516 | */ |
2517 | struct extent *rv; | |
2518 | int n = 0; | |
fcc22180 | 2519 | unsigned int i; |
a44e993e N |
2520 | __u16 state; |
2521 | ||
2522 | if (dl->pdnum < 0) | |
2523 | return NULL; | |
2524 | state = be16_to_cpu(ddf->phys->entries[dl->pdnum].state); | |
60056e1c | 2525 | |
2526 | if ((state & (DDF_Online|DDF_Failed|DDF_Missing)) != DDF_Online) | |
2527 | return NULL; | |
59e36268 | 2528 | |
503975b9 | 2529 | rv = xmalloc(sizeof(struct extent) * (ddf->max_part + 2)); |
59e36268 NB |
2530 | |
2531 | for (i = 0; i < ddf->max_part; i++) { | |
fcc22180 | 2532 | const struct vd_config *bvd; |
2533 | unsigned int ibvd; | |
59e36268 | 2534 | struct vcl *v = dl->vlist[i]; |
fcc22180 | 2535 | if (v == NULL || |
2536 | get_pd_index_from_refnum(v, dl->disk.refnum, ddf->mppe, | |
2537 | &bvd, &ibvd) == DDF_NOTFOUND) | |
59e36268 | 2538 | continue; |
9d0c6b70 | 2539 | rv[n].start = be64_to_cpu(LBA_OFFSET(ddf, bvd)[ibvd]); |
2540 | rv[n].size = be64_to_cpu(bvd->blocks); | |
fcc22180 | 2541 | n++; |
59e36268 NB |
2542 | } |
2543 | qsort(rv, n, sizeof(*rv), cmp_extent); | |
2544 | ||
9d0c6b70 | 2545 | rv[n].start = be64_to_cpu(ddf->phys->entries[dl->pdnum].config_size); |
59e36268 NB |
2546 | rv[n].size = 0; |
2547 | return rv; | |
2548 | } | |
fca65520 N |
2549 | |
2550 | static unsigned long long find_space( | |
2551 | struct ddf_super *ddf, struct dl *dl, | |
2552 | unsigned long long data_offset, | |
2553 | unsigned long long *size) | |
2554 | { | |
2555 | /* Find if the requested amount of space is available. | |
2556 | * If it is, return start. | |
2557 | * If not, set *size to largest space. | |
2558 | * If data_offset != INVALID_SECTORS, then the space must start | |
2559 | * at this location. | |
2560 | */ | |
2561 | struct extent *e = get_extents(ddf, dl); | |
2562 | int i = 0; | |
2563 | unsigned long long pos = 0; | |
2564 | unsigned long long max_size = 0; | |
2565 | ||
2566 | if (!e) { | |
2567 | *size = 0; | |
2568 | return INVALID_SECTORS; | |
2569 | } | |
2570 | do { | |
2571 | unsigned long long esize = e[i].start - pos; | |
2572 | if (data_offset != INVALID_SECTORS && | |
2573 | pos <= data_offset && | |
2574 | e[i].start > data_offset) { | |
2575 | pos = data_offset; | |
2576 | esize = e[i].start - pos; | |
2577 | } | |
2578 | if (data_offset != INVALID_SECTORS && | |
2579 | pos != data_offset) { | |
2580 | i++; | |
2581 | continue; | |
2582 | } | |
2583 | if (esize >= *size) { | |
2584 | /* Found! */ | |
2585 | free(e); | |
2586 | return pos; | |
2587 | } | |
2588 | if (esize > max_size) | |
2589 | max_size = esize; | |
2590 | pos = e[i].start + e[i].size; | |
2591 | i++; | |
2592 | } while (e[i-1].size); | |
2593 | *size = max_size; | |
2594 | free(e); | |
2595 | return INVALID_SECTORS; | |
2596 | } | |
59e36268 | 2597 | |
5f8097be NB |
2598 | static int init_super_ddf_bvd(struct supertype *st, |
2599 | mdu_array_info_t *info, | |
2600 | unsigned long long size, | |
2601 | char *name, char *homehost, | |
83cd1e97 | 2602 | int *uuid, unsigned long long data_offset) |
5f8097be NB |
2603 | { |
2604 | /* We are creating a BVD inside a pre-existing container. | |
2605 | * so st->sb is already set. | |
2606 | * We need to create a new vd_config and a new virtual_entry | |
2607 | */ | |
2608 | struct ddf_super *ddf = st->sb; | |
5aaf6c7b | 2609 | unsigned int venum, i; |
5f8097be NB |
2610 | struct virtual_entry *ve; |
2611 | struct vcl *vcl; | |
2612 | struct vd_config *vc; | |
5f8097be | 2613 | |
fb9d0acb | 2614 | if (find_vde_by_name(ddf, name) != DDF_NOTFOUND) { |
2615 | pr_err("This ddf already has an array called %s\n", name); | |
5f8097be NB |
2616 | return 0; |
2617 | } | |
fb9d0acb | 2618 | venum = find_unused_vde(ddf); |
2619 | if (venum == DDF_NOTFOUND) { | |
2620 | pr_err("Cannot find spare slot for virtual disk\n"); | |
5f8097be NB |
2621 | return 0; |
2622 | } | |
2623 | ve = &ddf->virt->entries[venum]; | |
2624 | ||
2625 | /* A Virtual Disk GUID contains the T10 Vendor ID, controller type, | |
2626 | * timestamp, random number | |
2627 | */ | |
2628 | make_header_guid(ve->guid); | |
a8173e43 | 2629 | ve->unit = cpu_to_be16(info->md_minor); |
5f8097be | 2630 | ve->pad0 = 0xFFFF; |
a8173e43 | 2631 | ve->guid_crc._v16 = crc32(0, (unsigned char *)ddf->anchor.guid, |
2632 | DDF_GUID_LEN); | |
2633 | ve->type = cpu_to_be16(0); | |
7a7cc504 NB |
2634 | ve->state = DDF_state_degraded; /* Will be modified as devices are added */ |
2635 | if (info->state & 1) /* clean */ | |
2636 | ve->init_state = DDF_init_full; | |
2637 | else | |
2638 | ve->init_state = DDF_init_not; | |
2639 | ||
5f8097be NB |
2640 | memset(ve->pad1, 0xff, 14); |
2641 | memset(ve->name, ' ', 16); | |
2642 | if (name) | |
2643 | strncpy(ve->name, name, 16); | |
2644 | ddf->virt->populated_vdes = | |
a8173e43 | 2645 | cpu_to_be16(be16_to_cpu(ddf->virt->populated_vdes)+1); |
5f8097be NB |
2646 | |
2647 | /* Now create a new vd_config */ | |
3d2c4fc7 DW |
2648 | if (posix_memalign((void**)&vcl, 512, |
2649 | (offsetof(struct vcl, conf) + ddf->conf_rec_len * 512)) != 0) { | |
1ade5cc1 | 2650 | pr_err("could not allocate vd_config\n"); |
3d2c4fc7 DW |
2651 | return 0; |
2652 | } | |
59e36268 NB |
2653 | vcl->vcnum = venum; |
2654 | vcl->block_sizes = NULL; /* FIXME not for CONCAT */ | |
5f8097be NB |
2655 | vc = &vcl->conf; |
2656 | ||
2657 | vc->magic = DDF_VD_CONF_MAGIC; | |
2658 | memcpy(vc->guid, ve->guid, DDF_GUID_LEN); | |
60931cf9 | 2659 | vc->timestamp = cpu_to_be32(time(0)-DECADE); |
2660 | vc->seqnum = cpu_to_be32(1); | |
5f8097be | 2661 | memset(vc->pad0, 0xff, 24); |
5f8097be | 2662 | vc->chunk_shift = chunk_to_shift(info->chunk_size); |
a3163bf0 | 2663 | if (layout_md2ddf(info, vc) == -1 || |
a8173e43 | 2664 | be16_to_cpu(vc->prim_elmnt_count) > ddf->mppe) { |
1ade5cc1 N |
2665 | pr_err("unsupported RAID level/layout %d/%d with %d disks\n", |
2666 | info->level, info->layout, info->raid_disks); | |
a3163bf0 | 2667 | free(vcl); |
2668 | return 0; | |
2669 | } | |
5f8097be | 2670 | vc->sec_elmnt_seq = 0; |
3c48f7be | 2671 | if (alloc_other_bvds(ddf, vcl) != 0) { |
1ade5cc1 | 2672 | pr_err("could not allocate other bvds\n"); |
3c48f7be | 2673 | free(vcl); |
2674 | return 0; | |
2675 | } | |
eddaacc3 | 2676 | vc->blocks = cpu_to_be64(size * 2); |
9d0c6b70 | 2677 | vc->array_blocks = cpu_to_be64( |
5f8097be | 2678 | calc_array_size(info->level, info->raid_disks, info->layout, |
eddaacc3 | 2679 | info->chunk_size, size * 2)); |
5f8097be | 2680 | memset(vc->pad1, 0xff, 8); |
60931cf9 | 2681 | vc->spare_refs[0] = cpu_to_be32(0xffffffff); |
2682 | vc->spare_refs[1] = cpu_to_be32(0xffffffff); | |
2683 | vc->spare_refs[2] = cpu_to_be32(0xffffffff); | |
2684 | vc->spare_refs[3] = cpu_to_be32(0xffffffff); | |
2685 | vc->spare_refs[4] = cpu_to_be32(0xffffffff); | |
2686 | vc->spare_refs[5] = cpu_to_be32(0xffffffff); | |
2687 | vc->spare_refs[6] = cpu_to_be32(0xffffffff); | |
2688 | vc->spare_refs[7] = cpu_to_be32(0xffffffff); | |
5f8097be NB |
2689 | memset(vc->cache_pol, 0, 8); |
2690 | vc->bg_rate = 0x80; | |
2691 | memset(vc->pad2, 0xff, 3); | |
2692 | memset(vc->pad3, 0xff, 52); | |
2693 | memset(vc->pad4, 0xff, 192); | |
2694 | memset(vc->v0, 0xff, 32); | |
2695 | memset(vc->v1, 0xff, 32); | |
2696 | memset(vc->v2, 0xff, 16); | |
2697 | memset(vc->v3, 0xff, 16); | |
2698 | memset(vc->vendor, 0xff, 32); | |
598f0d58 | 2699 | |
8c3b8c2c | 2700 | memset(vc->phys_refnum, 0xff, 4*ddf->mppe); |
e5a2a3cf | 2701 | memset(vc->phys_refnum+ddf->mppe, 0x00, 8*ddf->mppe); |
5f8097be | 2702 | |
5aaf6c7b | 2703 | for (i = 1; i < vc->sec_elmnt_count; i++) { |
2704 | memcpy(vcl->other_bvds[i-1], vc, ddf->conf_rec_len * 512); | |
2705 | vcl->other_bvds[i-1]->sec_elmnt_seq = i; | |
2706 | } | |
2707 | ||
5f8097be NB |
2708 | vcl->next = ddf->conflist; |
2709 | ddf->conflist = vcl; | |
d2ca6449 | 2710 | ddf->currentconf = vcl; |
609ce161 | 2711 | ddf_set_updates_pending(ddf, NULL); |
5f8097be NB |
2712 | return 1; |
2713 | } | |
2714 | ||
2715 | static void add_to_super_ddf_bvd(struct supertype *st, | |
476066a3 N |
2716 | mdu_disk_info_t *dk, int fd, char *devname, |
2717 | unsigned long long data_offset) | |
5f8097be | 2718 | { |
56cb05c4 | 2719 | /* fd and devname identify a device within the ddf container (st). |
5f8097be NB |
2720 | * dk identifies a location in the new BVD. |
2721 | * We need to find suitable free space in that device and update | |
2722 | * the phys_refnum and lba_offset for the newly created vd_config. | |
2723 | * We might also want to update the type in the phys_disk | |
5575e7d9 | 2724 | * section. |
8592f29d N |
2725 | * |
2726 | * Alternately: fd == -1 and we have already chosen which device to | |
2727 | * use and recorded in dlist->raid_disk; | |
5f8097be NB |
2728 | */ |
2729 | struct dl *dl; | |
2730 | struct ddf_super *ddf = st->sb; | |
2731 | struct vd_config *vc; | |
f21e18ca | 2732 | unsigned int i; |
fca65520 | 2733 | unsigned long long blocks, pos; |
475ccbdb | 2734 | unsigned int raid_disk = dk->raid_disk; |
5f8097be | 2735 | |
8592f29d N |
2736 | if (fd == -1) { |
2737 | for (dl = ddf->dlist; dl ; dl = dl->next) | |
2738 | if (dl->raiddisk == dk->raid_disk) | |
2739 | break; | |
2740 | } else { | |
2741 | for (dl = ddf->dlist; dl ; dl = dl->next) | |
2742 | if (dl->major == dk->major && | |
2743 | dl->minor == dk->minor) | |
2744 | break; | |
2745 | } | |
41bcbc14 | 2746 | if (!dl || dl->pdnum < 0 || ! (dk->state & (1<<MD_DISK_SYNC))) |
5f8097be NB |
2747 | return; |
2748 | ||
d2ca6449 | 2749 | vc = &ddf->currentconf->conf; |
475ccbdb | 2750 | if (vc->sec_elmnt_count > 1) { |
a8173e43 | 2751 | unsigned int n = be16_to_cpu(vc->prim_elmnt_count); |
475ccbdb | 2752 | if (raid_disk >= n) |
2753 | vc = ddf->currentconf->other_bvds[raid_disk / n - 1]; | |
2754 | raid_disk %= n; | |
2755 | } | |
59e36268 | 2756 | |
9d0c6b70 | 2757 | blocks = be64_to_cpu(vc->blocks); |
d2ca6449 NB |
2758 | if (ddf->currentconf->block_sizes) |
2759 | blocks = ddf->currentconf->block_sizes[dk->raid_disk]; | |
59e36268 | 2760 | |
476066a3 | 2761 | pos = find_space(ddf, dl, data_offset, &blocks); |
fca65520 | 2762 | if (pos == INVALID_SECTORS) |
59e36268 NB |
2763 | return; |
2764 | ||
d2ca6449 | 2765 | ddf->currentdev = dk->raid_disk; |
475ccbdb | 2766 | vc->phys_refnum[raid_disk] = dl->disk.refnum; |
9d0c6b70 | 2767 | LBA_OFFSET(ddf, vc)[raid_disk] = cpu_to_be64(pos); |
5f8097be | 2768 | |
f21e18ca | 2769 | for (i = 0; i < ddf->max_part ; i++) |
5575e7d9 NB |
2770 | if (dl->vlist[i] == NULL) |
2771 | break; | |
2772 | if (i == ddf->max_part) | |
2773 | return; | |
d2ca6449 | 2774 | dl->vlist[i] = ddf->currentconf; |
5f8097be | 2775 | |
8592f29d N |
2776 | if (fd >= 0) |
2777 | dl->fd = fd; | |
2778 | if (devname) | |
2779 | dl->devname = devname; | |
7a7cc504 | 2780 | |
63eb2454 | 2781 | /* Check if we can mark array as optimal yet */ |
d2ca6449 | 2782 | i = ddf->currentconf->vcnum; |
63eb2454 | 2783 | ddf->virt->entries[i].state = |
2784 | (ddf->virt->entries[i].state & ~DDF_state_mask) | |
2785 | | get_svd_state(ddf, ddf->currentconf); | |
a8173e43 | 2786 | be16_clear(ddf->phys->entries[dl->pdnum].type, |
2787 | cpu_to_be16(DDF_Global_Spare)); | |
2788 | be16_set(ddf->phys->entries[dl->pdnum].type, | |
2789 | cpu_to_be16(DDF_Active_in_VD)); | |
1ade5cc1 N |
2790 | dprintf("added disk %d/%08x to VD %d/%s as disk %d\n", |
2791 | dl->pdnum, be32_to_cpu(dl->disk.refnum), | |
4f9bbe63 | 2792 | ddf->currentconf->vcnum, guid_str(vc->guid), |
2793 | dk->raid_disk); | |
609ce161 | 2794 | ddf_set_updates_pending(ddf, vc); |
5f8097be NB |
2795 | } |
2796 | ||
4a3ca8ac | 2797 | static unsigned int find_unused_pde(const struct ddf_super *ddf) |
2798 | { | |
2799 | unsigned int i; | |
a8173e43 | 2800 | for (i = 0; i < be16_to_cpu(ddf->phys->max_pdes); i++) { |
4a3ca8ac | 2801 | if (all_ff(ddf->phys->entries[i].guid)) |
2802 | return i; | |
2803 | } | |
2804 | return DDF_NOTFOUND; | |
2805 | } | |
2806 | ||
105e6e93 | 2807 | static void _set_config_size(struct phys_disk_entry *pde, const struct dl *dl) |
2808 | { | |
2809 | __u64 cfs, t; | |
2810 | cfs = min(dl->size - 32*1024*2ULL, be64_to_cpu(dl->primary_lba)); | |
2811 | t = be64_to_cpu(dl->secondary_lba); | |
2812 | if (t != ~(__u64)0) | |
2813 | cfs = min(cfs, t); | |
2814 | /* | |
2815 | * Some vendor DDF structures interpret workspace_lba | |
56cb05c4 | 2816 | * very differently than we do: Make a sanity check on the value. |
105e6e93 | 2817 | */ |
2818 | t = be64_to_cpu(dl->workspace_lba); | |
2819 | if (t < cfs) { | |
2820 | __u64 wsp = cfs - t; | |
2821 | if (wsp > 1024*1024*2ULL && wsp > dl->size / 16) { | |
1ade5cc1 N |
2822 | pr_err("%x:%x: workspace size 0x%llx too big, ignoring\n", |
2823 | dl->major, dl->minor, (unsigned long long)wsp); | |
105e6e93 | 2824 | } else |
2825 | cfs = t; | |
2826 | } | |
2827 | pde->config_size = cpu_to_be64(cfs); | |
1ade5cc1 N |
2828 | dprintf("%x:%x config_size %llx, DDF structure is %llx blocks\n", |
2829 | dl->major, dl->minor, | |
d13566f9 | 2830 | (unsigned long long)cfs, (unsigned long long)(dl->size-cfs)); |
105e6e93 | 2831 | } |
2832 | ||
56cb05c4 | 2833 | /* Add a device to a container, either while creating it or while |
a322f70c DW |
2834 | * expanding a pre-existing container |
2835 | */ | |
f20c3968 | 2836 | static int add_to_super_ddf(struct supertype *st, |
72ca9bcf N |
2837 | mdu_disk_info_t *dk, int fd, char *devname, |
2838 | unsigned long long data_offset) | |
a322f70c DW |
2839 | { |
2840 | struct ddf_super *ddf = st->sb; | |
2841 | struct dl *dd; | |
2842 | time_t now; | |
2843 | struct tm *tm; | |
2844 | unsigned long long size; | |
2845 | struct phys_disk_entry *pde; | |
f21e18ca | 2846 | unsigned int n, i; |
a322f70c | 2847 | struct stat stb; |
90fa1a29 | 2848 | __u32 *tptr; |
a322f70c | 2849 | |
78e44928 | 2850 | if (ddf->currentconf) { |
476066a3 | 2851 | add_to_super_ddf_bvd(st, dk, fd, devname, data_offset); |
f20c3968 | 2852 | return 0; |
78e44928 NB |
2853 | } |
2854 | ||
a322f70c DW |
2855 | /* This is device numbered dk->number. We need to create |
2856 | * a phys_disk entry and a more detailed disk_data entry. | |
2857 | */ | |
2858 | fstat(fd, &stb); | |
4a3ca8ac | 2859 | n = find_unused_pde(ddf); |
2860 | if (n == DDF_NOTFOUND) { | |
1ade5cc1 | 2861 | pr_err("No free slot in array, cannot add disk\n"); |
4a3ca8ac | 2862 | return 1; |
2863 | } | |
2864 | pde = &ddf->phys->entries[n]; | |
4ee8cca9 | 2865 | get_dev_size(fd, NULL, &size); |
2866 | if (size <= 32*1024*1024) { | |
1ade5cc1 | 2867 | pr_err("device size must be at least 32MB\n"); |
4ee8cca9 | 2868 | return 1; |
2869 | } | |
2870 | size >>= 9; | |
4a3ca8ac | 2871 | |
3d2c4fc7 DW |
2872 | if (posix_memalign((void**)&dd, 512, |
2873 | sizeof(*dd) + sizeof(dd->vlist[0]) * ddf->max_part) != 0) { | |
1ade5cc1 | 2874 | pr_err("could allocate buffer for new disk, aborting\n"); |
f20c3968 | 2875 | return 1; |
3d2c4fc7 | 2876 | } |
a322f70c DW |
2877 | dd->major = major(stb.st_rdev); |
2878 | dd->minor = minor(stb.st_rdev); | |
2879 | dd->devname = devname; | |
a322f70c | 2880 | dd->fd = fd; |
b2280677 | 2881 | dd->spare = NULL; |
a322f70c DW |
2882 | |
2883 | dd->disk.magic = DDF_PHYS_DATA_MAGIC; | |
2884 | now = time(0); | |
2885 | tm = localtime(&now); | |
2886 | sprintf(dd->disk.guid, "%8s%04d%02d%02d", | |
2887 | T10, tm->tm_year+1900, tm->tm_mon+1, tm->tm_mday); | |
90fa1a29 JS |
2888 | tptr = (__u32 *)(dd->disk.guid + 16); |
2889 | *tptr++ = random32(); | |
2890 | *tptr = random32(); | |
a322f70c | 2891 | |
59e36268 NB |
2892 | do { |
2893 | /* Cannot be bothered finding a CRC of some irrelevant details*/ | |
60931cf9 | 2894 | dd->disk.refnum._v32 = random32(); |
a8173e43 | 2895 | for (i = be16_to_cpu(ddf->active->max_pd_entries); |
f21e18ca | 2896 | i > 0; i--) |
60931cf9 | 2897 | if (be32_eq(ddf->phys->entries[i-1].refnum, |
2898 | dd->disk.refnum)) | |
59e36268 | 2899 | break; |
f21e18ca | 2900 | } while (i > 0); |
59e36268 | 2901 | |
a322f70c DW |
2902 | dd->disk.forced_ref = 1; |
2903 | dd->disk.forced_guid = 1; | |
2904 | memset(dd->disk.vendor, ' ', 32); | |
2905 | memcpy(dd->disk.vendor, "Linux", 5); | |
2906 | memset(dd->disk.pad, 0xff, 442); | |
b2280677 | 2907 | for (i = 0; i < ddf->max_part ; i++) |
a322f70c DW |
2908 | dd->vlist[i] = NULL; |
2909 | ||
5575e7d9 NB |
2910 | dd->pdnum = n; |
2911 | ||
2cc2983d N |
2912 | if (st->update_tail) { |
2913 | int len = (sizeof(struct phys_disk) + | |
2914 | sizeof(struct phys_disk_entry)); | |
2915 | struct phys_disk *pd; | |
2916 | ||
503975b9 | 2917 | pd = xmalloc(len); |
2cc2983d | 2918 | pd->magic = DDF_PHYS_RECORDS_MAGIC; |
a8173e43 | 2919 | pd->used_pdes = cpu_to_be16(n); |
2cc2983d N |
2920 | pde = &pd->entries[0]; |
2921 | dd->mdupdate = pd; | |
4a3ca8ac | 2922 | } else |
a8173e43 | 2923 | ddf->phys->used_pdes = cpu_to_be16( |
2924 | 1 + be16_to_cpu(ddf->phys->used_pdes)); | |
a322f70c DW |
2925 | |
2926 | memcpy(pde->guid, dd->disk.guid, DDF_GUID_LEN); | |
2927 | pde->refnum = dd->disk.refnum; | |
a8173e43 | 2928 | pde->type = cpu_to_be16(DDF_Forced_PD_GUID | DDF_Global_Spare); |
2929 | pde->state = cpu_to_be16(DDF_Online); | |
4ee8cca9 | 2930 | dd->size = size; |
2931 | /* | |
2932 | * If there is already a device in dlist, try to reserve the same | |
2933 | * amount of workspace. Otherwise, use 32MB. | |
2934 | * We checked disk size above already. | |
2935 | */ | |
2936 | #define __calc_lba(new, old, lba, mb) do { \ | |
2937 | unsigned long long dif; \ | |
2938 | if ((old) != NULL) \ | |
9d0c6b70 | 2939 | dif = (old)->size - be64_to_cpu((old)->lba); \ |
4ee8cca9 | 2940 | else \ |
2941 | dif = (new)->size; \ | |
2942 | if ((new)->size > dif) \ | |
9d0c6b70 | 2943 | (new)->lba = cpu_to_be64((new)->size - dif); \ |
4ee8cca9 | 2944 | else \ |
9d0c6b70 | 2945 | (new)->lba = cpu_to_be64((new)->size - (mb*1024*2)); \ |
4ee8cca9 | 2946 | } while (0) |
2947 | __calc_lba(dd, ddf->dlist, workspace_lba, 32); | |
2948 | __calc_lba(dd, ddf->dlist, primary_lba, 16); | |
b95cb4b9 N |
2949 | if (ddf->dlist == NULL || |
2950 | be64_to_cpu(ddf->dlist->secondary_lba) != ~(__u64)0) | |
2951 | __calc_lba(dd, ddf->dlist, secondary_lba, 32); | |
105e6e93 | 2952 | _set_config_size(pde, dd); |
4ee8cca9 | 2953 | |
a322f70c DW |
2954 | sprintf(pde->path, "%17.17s","Information: nil") ; |
2955 | memset(pde->pad, 0xff, 6); | |
2956 | ||
2cc2983d N |
2957 | if (st->update_tail) { |
2958 | dd->next = ddf->add_list; | |
2959 | ddf->add_list = dd; | |
2960 | } else { | |
2961 | dd->next = ddf->dlist; | |
2962 | ddf->dlist = dd; | |
609ce161 | 2963 | ddf_set_updates_pending(ddf, NULL); |
2cc2983d | 2964 | } |
f20c3968 DW |
2965 | |
2966 | return 0; | |
a322f70c DW |
2967 | } |
2968 | ||
4dd968cc N |
2969 | static int remove_from_super_ddf(struct supertype *st, mdu_disk_info_t *dk) |
2970 | { | |
2971 | struct ddf_super *ddf = st->sb; | |
2972 | struct dl *dl; | |
2973 | ||
2974 | /* mdmon has noticed that this disk (dk->major/dk->minor) has | |
2975 | * disappeared from the container. | |
2976 | * We need to arrange that it disappears from the metadata and | |
2977 | * internal data structures too. | |
2978 | * Most of the work is done by ddf_process_update which edits | |
2979 | * the metadata and closes the file handle and attaches the memory | |
2980 | * where free_updates will free it. | |
2981 | */ | |
2982 | for (dl = ddf->dlist; dl ; dl = dl->next) | |
2983 | if (dl->major == dk->major && | |
2984 | dl->minor == dk->minor) | |
2985 | break; | |
a44e993e | 2986 | if (!dl || dl->pdnum < 0) |
4dd968cc N |
2987 | return -1; |
2988 | ||
2989 | if (st->update_tail) { | |
2990 | int len = (sizeof(struct phys_disk) + | |
2991 | sizeof(struct phys_disk_entry)); | |
2992 | struct phys_disk *pd; | |
2993 | ||
503975b9 | 2994 | pd = xmalloc(len); |
4dd968cc | 2995 | pd->magic = DDF_PHYS_RECORDS_MAGIC; |
a8173e43 | 2996 | pd->used_pdes = cpu_to_be16(dl->pdnum); |
2997 | pd->entries[0].state = cpu_to_be16(DDF_Missing); | |
4dd968cc N |
2998 | append_metadata_update(st, pd, len); |
2999 | } | |
3000 | return 0; | |
3001 | } | |
3002 | ||
a322f70c DW |
3003 | /* |
3004 | * This is the write_init_super method for a ddf container. It is | |
3005 | * called when creating a container or adding another device to a | |
3006 | * container. | |
3007 | */ | |
18a2f463 | 3008 | |
3921e41a | 3009 | static int __write_ddf_structure(struct dl *d, struct ddf_super *ddf, __u8 type) |
a322f70c | 3010 | { |
7f798aca | 3011 | unsigned long long sector; |
3012 | struct ddf_header *header; | |
3921e41a | 3013 | int fd, i, n_config, conf_size, buf_size; |
a4057a88 | 3014 | int ret = 0; |
3921e41a | 3015 | char *conf; |
8e9387ac | 3016 | |
7f798aca | 3017 | fd = d->fd; |
3018 | ||
3019 | switch (type) { | |
3020 | case DDF_HEADER_PRIMARY: | |
3021 | header = &ddf->primary; | |
9d0c6b70 | 3022 | sector = be64_to_cpu(header->primary_lba); |
7f798aca | 3023 | break; |
3024 | case DDF_HEADER_SECONDARY: | |
3025 | header = &ddf->secondary; | |
9d0c6b70 | 3026 | sector = be64_to_cpu(header->secondary_lba); |
7f798aca | 3027 | break; |
3028 | default: | |
3029 | return 0; | |
3030 | } | |
b95cb4b9 N |
3031 | if (sector == ~(__u64)0) |
3032 | return 0; | |
7f798aca | 3033 | |
3034 | header->type = type; | |
a4057a88 | 3035 | header->openflag = 1; |
7f798aca | 3036 | header->crc = calc_crc(header, 512); |
3037 | ||
3038 | lseek64(fd, sector<<9, 0); | |
3039 | if (write(fd, header, 512) < 0) | |
a4057a88 | 3040 | goto out; |
7f798aca | 3041 | |
3042 | ddf->controller.crc = calc_crc(&ddf->controller, 512); | |
3043 | if (write(fd, &ddf->controller, 512) < 0) | |
a4057a88 | 3044 | goto out; |
a322f70c | 3045 | |
7f798aca | 3046 | ddf->phys->crc = calc_crc(ddf->phys, ddf->pdsize); |
3047 | if (write(fd, ddf->phys, ddf->pdsize) < 0) | |
a4057a88 | 3048 | goto out; |
7f798aca | 3049 | ddf->virt->crc = calc_crc(ddf->virt, ddf->vdsize); |
3050 | if (write(fd, ddf->virt, ddf->vdsize) < 0) | |
a4057a88 | 3051 | goto out; |
7f798aca | 3052 | |
3053 | /* Now write lots of config records. */ | |
3054 | n_config = ddf->max_part; | |
3055 | conf_size = ddf->conf_rec_len * 512; | |
3921e41a N |
3056 | conf = ddf->conf; |
3057 | buf_size = conf_size * (n_config + 1); | |
3058 | if (!conf) { | |
3059 | if (posix_memalign((void**)&conf, 512, buf_size) != 0) | |
3060 | goto out; | |
3061 | ddf->conf = conf; | |
3062 | } | |
7f798aca | 3063 | for (i = 0 ; i <= n_config ; i++) { |
e3c2a365 | 3064 | struct vcl *c; |
3065 | struct vd_config *vdc = NULL; | |
3066 | if (i == n_config) { | |
7f798aca | 3067 | c = (struct vcl *)d->spare; |
e3c2a365 | 3068 | if (c) |
3069 | vdc = &c->conf; | |
3070 | } else { | |
3071 | unsigned int dummy; | |
3072 | c = d->vlist[i]; | |
3073 | if (c) | |
3074 | get_pd_index_from_refnum( | |
3075 | c, d->disk.refnum, | |
3076 | ddf->mppe, | |
3077 | (const struct vd_config **)&vdc, | |
3078 | &dummy); | |
3079 | } | |
188d31ed | 3080 | if (vdc) { |
be9b9ef4 | 3081 | dprintf("writing conf record %i on disk %08x for %s/%u\n", |
60931cf9 | 3082 | i, be32_to_cpu(d->disk.refnum), |
ad60eea1 | 3083 | guid_str(vdc->guid), |
be9b9ef4 | 3084 | vdc->sec_elmnt_seq); |
e3c2a365 | 3085 | vdc->crc = calc_crc(vdc, conf_size); |
3921e41a | 3086 | memcpy(conf + i*conf_size, vdc, conf_size); |
ce45c819 | 3087 | } else |
3921e41a | 3088 | memset(conf + i*conf_size, 0xff, conf_size); |
7f798aca | 3089 | } |
3921e41a | 3090 | if (write(fd, conf, buf_size) != buf_size) |
a4057a88 | 3091 | goto out; |
7f798aca | 3092 | |
3093 | d->disk.crc = calc_crc(&d->disk, 512); | |
3094 | if (write(fd, &d->disk, 512) < 0) | |
a4057a88 | 3095 | goto out; |
7f798aca | 3096 | |
a4057a88 | 3097 | ret = 1; |
3098 | out: | |
3099 | header->openflag = 0; | |
3100 | header->crc = calc_crc(header, 512); | |
3101 | ||
3102 | lseek64(fd, sector<<9, 0); | |
3103 | if (write(fd, header, 512) < 0) | |
3104 | ret = 0; | |
3105 | ||
3106 | return ret; | |
7f798aca | 3107 | } |
3108 | ||
3921e41a | 3109 | static int _write_super_to_disk(struct ddf_super *ddf, struct dl *d) |
9bf38704 | 3110 | { |
3111 | unsigned long long size; | |
3112 | int fd = d->fd; | |
3113 | if (fd < 0) | |
3114 | return 0; | |
3115 | ||
3116 | /* We need to fill in the primary, (secondary) and workspace | |
3117 | * lba's in the headers, set their checksums, | |
3118 | * Also checksum phys, virt.... | |
3119 | * | |
3120 | * Then write everything out, finally the anchor is written. | |
3121 | */ | |
3122 | get_dev_size(fd, NULL, &size); | |
3123 | size /= 512; | |
30bee020 | 3124 | memcpy(&ddf->anchor, ddf->active, 512); |
9d0c6b70 | 3125 | if (be64_to_cpu(d->workspace_lba) != 0ULL) |
9bf38704 | 3126 | ddf->anchor.workspace_lba = d->workspace_lba; |
3127 | else | |
3128 | ddf->anchor.workspace_lba = | |
9d0c6b70 | 3129 | cpu_to_be64(size - 32*1024*2); |
3130 | if (be64_to_cpu(d->primary_lba) != 0ULL) | |
9bf38704 | 3131 | ddf->anchor.primary_lba = d->primary_lba; |
3132 | else | |
3133 | ddf->anchor.primary_lba = | |
9d0c6b70 | 3134 | cpu_to_be64(size - 16*1024*2); |
3135 | if (be64_to_cpu(d->secondary_lba) != 0ULL) | |
9bf38704 | 3136 | ddf->anchor.secondary_lba = d->secondary_lba; |
3137 | else | |
3138 | ddf->anchor.secondary_lba = | |
9d0c6b70 | 3139 | cpu_to_be64(size - 32*1024*2); |
07de2684 | 3140 | ddf->anchor.timestamp = cpu_to_be32(time(0) - DECADE); |
9bf38704 | 3141 | memcpy(&ddf->primary, &ddf->anchor, 512); |
3142 | memcpy(&ddf->secondary, &ddf->anchor, 512); | |
3143 | ||
b87fdf4e | 3144 | ddf->anchor.type = DDF_HEADER_ANCHOR; |
9bf38704 | 3145 | ddf->anchor.openflag = 0xFF; /* 'open' means nothing */ |
60931cf9 | 3146 | ddf->anchor.seq = cpu_to_be32(0xFFFFFFFF); /* no sequencing in anchor */ |
9bf38704 | 3147 | ddf->anchor.crc = calc_crc(&ddf->anchor, 512); |
3148 | ||
3921e41a | 3149 | if (!__write_ddf_structure(d, ddf, DDF_HEADER_PRIMARY)) |
9bf38704 | 3150 | return 0; |
3151 | ||
3921e41a | 3152 | if (!__write_ddf_structure(d, ddf, DDF_HEADER_SECONDARY)) |
9bf38704 | 3153 | return 0; |
3154 | ||
3155 | lseek64(fd, (size-1)*512, SEEK_SET); | |
3156 | if (write(fd, &ddf->anchor, 512) < 0) | |
3157 | return 0; | |
3158 | ||
3159 | return 1; | |
3160 | } | |
3161 | ||
3921e41a | 3162 | static int __write_init_super_ddf(struct supertype *st) |
7f798aca | 3163 | { |
a322f70c | 3164 | struct ddf_super *ddf = st->sb; |
a322f70c | 3165 | struct dl *d; |
175593bf DW |
3166 | int attempts = 0; |
3167 | int successes = 0; | |
42d5dfd9 | 3168 | |
7d5a7ff3 | 3169 | pr_state(ddf, __func__); |
a322f70c | 3170 | |
175593bf DW |
3171 | /* try to write updated metadata, |
3172 | * if we catch a failure move on to the next disk | |
3173 | */ | |
a322f70c | 3174 | for (d = ddf->dlist; d; d=d->next) { |
175593bf | 3175 | attempts++; |
3921e41a | 3176 | successes += _write_super_to_disk(ddf, d); |
175593bf DW |
3177 | } |
3178 | ||
175593bf | 3179 | return attempts != successes; |
a322f70c | 3180 | } |
7a7cc504 NB |
3181 | |
3182 | static int write_init_super_ddf(struct supertype *st) | |
3183 | { | |
9b1fb677 DW |
3184 | struct ddf_super *ddf = st->sb; |
3185 | struct vcl *currentconf = ddf->currentconf; | |
3186 | ||
56cb05c4 | 3187 | /* We are done with currentconf - reset it so st refers to the container */ |
9b1fb677 | 3188 | ddf->currentconf = NULL; |
edd8d13c NB |
3189 | |
3190 | if (st->update_tail) { | |
3191 | /* queue the virtual_disk and vd_config as metadata updates */ | |
3192 | struct virtual_disk *vd; | |
3193 | struct vd_config *vc; | |
c5943560 | 3194 | int len, tlen; |
3195 | unsigned int i; | |
edd8d13c | 3196 | |
9b1fb677 | 3197 | if (!currentconf) { |
56cb05c4 | 3198 | /* Must be adding a physical disk to the container */ |
2cc2983d N |
3199 | int len = (sizeof(struct phys_disk) + |
3200 | sizeof(struct phys_disk_entry)); | |
3201 | ||
3202 | /* adding a disk to the container. */ | |
3203 | if (!ddf->add_list) | |
3204 | return 0; | |
3205 | ||
3206 | append_metadata_update(st, ddf->add_list->mdupdate, len); | |
3207 | ddf->add_list->mdupdate = NULL; | |
3208 | return 0; | |
3209 | } | |
3210 | ||
3211 | /* Newly created VD */ | |
3212 | ||
edd8d13c NB |
3213 | /* First the virtual disk. We have a slightly fake header */ |
3214 | len = sizeof(struct virtual_disk) + sizeof(struct virtual_entry); | |
503975b9 | 3215 | vd = xmalloc(len); |
edd8d13c | 3216 | *vd = *ddf->virt; |
9b1fb677 | 3217 | vd->entries[0] = ddf->virt->entries[currentconf->vcnum]; |
a8173e43 | 3218 | vd->populated_vdes = cpu_to_be16(currentconf->vcnum); |
edd8d13c NB |
3219 | append_metadata_update(st, vd, len); |
3220 | ||
3221 | /* Then the vd_config */ | |
3222 | len = ddf->conf_rec_len * 512; | |
c5943560 | 3223 | tlen = len * currentconf->conf.sec_elmnt_count; |
3224 | vc = xmalloc(tlen); | |
9b1fb677 | 3225 | memcpy(vc, ¤tconf->conf, len); |
c5943560 | 3226 | for (i = 1; i < currentconf->conf.sec_elmnt_count; i++) |
3227 | memcpy((char *)vc + i*len, currentconf->other_bvds[i-1], | |
3228 | len); | |
3229 | append_metadata_update(st, vc, tlen); | |
edd8d13c | 3230 | |
edd8d13c | 3231 | return 0; |
613b0d17 | 3232 | } else { |
d682f344 | 3233 | struct dl *d; |
19041058 | 3234 | if (!currentconf) |
3235 | for (d = ddf->dlist; d; d=d->next) | |
3236 | while (Kill(d->devname, NULL, 0, -1, 1) == 0); | |
733eedc8 N |
3237 | /* Note: we don't close the fd's now, but a subsequent |
3238 | * ->free_super() will | |
3239 | */ | |
3921e41a | 3240 | return __write_init_super_ddf(st); |
d682f344 | 3241 | } |
7a7cc504 NB |
3242 | } |
3243 | ||
387fcd59 N |
3244 | static __u64 avail_size_ddf(struct supertype *st, __u64 devsize, |
3245 | unsigned long long data_offset) | |
a322f70c DW |
3246 | { |
3247 | /* We must reserve the last 32Meg */ | |
3248 | if (devsize <= 32*1024*2) | |
3249 | return 0; | |
3250 | return devsize - 32*1024*2; | |
3251 | } | |
3252 | ||
8592f29d N |
3253 | static int reserve_space(struct supertype *st, int raiddisks, |
3254 | unsigned long long size, int chunk, | |
476066a3 | 3255 | unsigned long long data_offset, |
8592f29d N |
3256 | unsigned long long *freesize) |
3257 | { | |
3258 | /* Find 'raiddisks' spare extents at least 'size' big (but | |
3259 | * only caring about multiples of 'chunk') and remember | |
56cb05c4 N |
3260 | * them. If size==0, find the largest size possible. |
3261 | * Report available size in *freesize | |
3262 | * If space cannot be found, fail. | |
8592f29d N |
3263 | */ |
3264 | struct dl *dl; | |
3265 | struct ddf_super *ddf = st->sb; | |
3266 | int cnt = 0; | |
3267 | ||
3268 | for (dl = ddf->dlist; dl ; dl=dl->next) { | |
613b0d17 | 3269 | dl->raiddisk = -1; |
8592f29d N |
3270 | dl->esize = 0; |
3271 | } | |
3272 | /* Now find largest extent on each device */ | |
3273 | for (dl = ddf->dlist ; dl ; dl=dl->next) { | |
fca65520 | 3274 | unsigned long long minsize = ULLONG_MAX; |
8592f29d | 3275 | |
476066a3 | 3276 | find_space(ddf, dl, data_offset, &minsize); |
fca65520 | 3277 | if (minsize >= size && minsize >= (unsigned)chunk) { |
8592f29d N |
3278 | cnt++; |
3279 | dl->esize = minsize; | |
3280 | } | |
8592f29d N |
3281 | } |
3282 | if (cnt < raiddisks) { | |
e7b84f9d | 3283 | pr_err("not enough devices with space to create array.\n"); |
8592f29d N |
3284 | return 0; /* No enough free spaces large enough */ |
3285 | } | |
3286 | if (size == 0) { | |
3287 | /* choose the largest size of which there are at least 'raiddisk' */ | |
3288 | for (dl = ddf->dlist ; dl ; dl=dl->next) { | |
3289 | struct dl *dl2; | |
3290 | if (dl->esize <= size) | |
3291 | continue; | |
3292 | /* This is bigger than 'size', see if there are enough */ | |
3293 | cnt = 0; | |
7b80ad6a | 3294 | for (dl2 = ddf->dlist; dl2 ; dl2=dl2->next) |
8592f29d N |
3295 | if (dl2->esize >= dl->esize) |
3296 | cnt++; | |
3297 | if (cnt >= raiddisks) | |
3298 | size = dl->esize; | |
3299 | } | |
3300 | if (chunk) { | |
3301 | size = size / chunk; | |
3302 | size *= chunk; | |
3303 | } | |
3304 | *freesize = size; | |
3305 | if (size < 32) { | |
e7b84f9d | 3306 | pr_err("not enough spare devices to create array.\n"); |
8592f29d N |
3307 | return 0; |
3308 | } | |
3309 | } | |
3310 | /* We have a 'size' of which there are enough spaces. | |
3311 | * We simply do a first-fit */ | |
3312 | cnt = 0; | |
3313 | for (dl = ddf->dlist ; dl && cnt < raiddisks ; dl=dl->next) { | |
3314 | if (dl->esize < size) | |
3315 | continue; | |
613b0d17 | 3316 | |
8592f29d N |
3317 | dl->raiddisk = cnt; |
3318 | cnt++; | |
3319 | } | |
3320 | return 1; | |
3321 | } | |
3322 | ||
78e44928 | 3323 | static int validate_geometry_ddf(struct supertype *st, |
2c514b71 | 3324 | int level, int layout, int raiddisks, |
c21e737b | 3325 | int *chunk, unsigned long long size, |
af4348dd | 3326 | unsigned long long data_offset, |
2c514b71 | 3327 | char *dev, unsigned long long *freesize, |
5308f117 | 3328 | int consistency_policy, int verbose) |
a322f70c DW |
3329 | { |
3330 | int fd; | |
3331 | struct mdinfo *sra; | |
3332 | int cfd; | |
3333 | ||
3334 | /* ddf potentially supports lots of things, but it depends on | |
3335 | * what devices are offered (and maybe kernel version?) | |
3336 | * If given unused devices, we will make a container. | |
3337 | * If given devices in a container, we will make a BVD. | |
3338 | * If given BVDs, we make an SVD, changing all the GUIDs in the process. | |
3339 | */ | |
3340 | ||
7ccc4cc4 | 3341 | if (*chunk == UnSet) |
bb7295f1 N |
3342 | *chunk = DEFAULT_CHUNK; |
3343 | ||
56cb05c4 N |
3344 | if (level == LEVEL_NONE) |
3345 | level = LEVEL_CONTAINER; | |
a322f70c | 3346 | if (level == LEVEL_CONTAINER) { |
78e44928 NB |
3347 | /* Must be a fresh device to add to a container */ |
3348 | return validate_geometry_ddf_container(st, level, layout, | |
7ccc4cc4 | 3349 | raiddisks, *chunk, |
af4348dd N |
3350 | size, data_offset, dev, |
3351 | freesize, | |
2c514b71 | 3352 | verbose); |
5f8097be NB |
3353 | } |
3354 | ||
78e44928 | 3355 | if (!dev) { |
a3163bf0 | 3356 | mdu_array_info_t array = { |
56cb05c4 N |
3357 | .level = level, |
3358 | .layout = layout, | |
a3163bf0 | 3359 | .raid_disks = raiddisks |
3360 | }; | |
3361 | struct vd_config conf; | |
3362 | if (layout_md2ddf(&array, &conf) == -1) { | |
b42f577a | 3363 | if (verbose) |
94b08b7c | 3364 | pr_err("DDF does not support level %d /layout %d arrays with %d disks\n", |
3365 | level, layout, raiddisks); | |
78e44928 | 3366 | return 0; |
b42f577a | 3367 | } |
78e44928 | 3368 | /* Should check layout? etc */ |
8592f29d N |
3369 | |
3370 | if (st->sb && freesize) { | |
3371 | /* --create was given a container to create in. | |
3372 | * So we need to check that there are enough | |
3373 | * free spaces and return the amount of space. | |
3374 | * We may as well remember which drives were | |
3375 | * chosen so that add_to_super/getinfo_super | |
3376 | * can return them. | |
3377 | */ | |
476066a3 N |
3378 | return reserve_space(st, raiddisks, size, *chunk, |
3379 | data_offset, freesize); | |
8592f29d | 3380 | } |
a322f70c | 3381 | return 1; |
78e44928 | 3382 | } |
a322f70c | 3383 | |
8592f29d N |
3384 | if (st->sb) { |
3385 | /* A container has already been opened, so we are | |
3386 | * creating in there. Maybe a BVD, maybe an SVD. | |
3387 | * Should make a distinction one day. | |
3388 | */ | |
3389 | return validate_geometry_ddf_bvd(st, level, layout, raiddisks, | |
af4348dd N |
3390 | chunk, size, data_offset, dev, |
3391 | freesize, | |
8592f29d N |
3392 | verbose); |
3393 | } | |
78e44928 NB |
3394 | /* This is the first device for the array. |
3395 | * If it is a container, we read it in and do automagic allocations, | |
3396 | * no other devices should be given. | |
3397 | * Otherwise it must be a member device of a container, and we | |
3398 | * do manual allocation. | |
3399 | * Later we should check for a BVD and make an SVD. | |
a322f70c | 3400 | */ |
a322f70c DW |
3401 | fd = open(dev, O_RDONLY|O_EXCL, 0); |
3402 | if (fd >= 0) { | |
a322f70c | 3403 | close(fd); |
2ecda5a3 | 3404 | /* Just a bare device, no good to us */ |
2c514b71 | 3405 | if (verbose) |
7a862a02 | 3406 | pr_err("ddf: Cannot create this array on device %s - a container is required.\n", |
e7b84f9d | 3407 | dev); |
a322f70c DW |
3408 | return 0; |
3409 | } | |
3410 | if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) { | |
2c514b71 | 3411 | if (verbose) |
e7b84f9d | 3412 | pr_err("ddf: Cannot open %s: %s\n", |
613b0d17 | 3413 | dev, strerror(errno)); |
a322f70c DW |
3414 | return 0; |
3415 | } | |
3416 | /* Well, it is in use by someone, maybe a 'ddf' container. */ | |
3417 | cfd = open_container(fd); | |
3418 | if (cfd < 0) { | |
3419 | close(fd); | |
2c514b71 | 3420 | if (verbose) |
e7b84f9d | 3421 | pr_err("ddf: Cannot use %s: %s\n", |
613b0d17 | 3422 | dev, strerror(EBUSY)); |
a322f70c DW |
3423 | return 0; |
3424 | } | |
4dd2df09 | 3425 | sra = sysfs_read(cfd, NULL, GET_VERSION); |
a322f70c DW |
3426 | close(fd); |
3427 | if (sra && sra->array.major_version == -1 && | |
3428 | strcmp(sra->text_version, "ddf") == 0) { | |
3429 | /* This is a member of a ddf container. Load the container | |
3430 | * and try to create a bvd | |
3431 | */ | |
3432 | struct ddf_super *ddf; | |
e1902a7b | 3433 | if (load_super_ddf_all(st, cfd, (void **)&ddf, NULL) == 0) { |
5f8097be | 3434 | st->sb = ddf; |
4dd2df09 | 3435 | strcpy(st->container_devnm, fd2devnm(cfd)); |
a322f70c | 3436 | close(cfd); |
78e44928 | 3437 | return validate_geometry_ddf_bvd(st, level, layout, |
a322f70c | 3438 | raiddisks, chunk, size, |
af4348dd | 3439 | data_offset, |
2c514b71 NB |
3440 | dev, freesize, |
3441 | verbose); | |
a322f70c DW |
3442 | } |
3443 | close(cfd); | |
c42ec1ed DW |
3444 | } else /* device may belong to a different container */ |
3445 | return 0; | |
3446 | ||
a322f70c DW |
3447 | return 1; |
3448 | } | |
3449 | ||
2c514b71 NB |
3450 | static int |
3451 | validate_geometry_ddf_container(struct supertype *st, | |
3452 | int level, int layout, int raiddisks, | |
3453 | int chunk, unsigned long long size, | |
af4348dd | 3454 | unsigned long long data_offset, |
2c514b71 NB |
3455 | char *dev, unsigned long long *freesize, |
3456 | int verbose) | |
a322f70c DW |
3457 | { |
3458 | int fd; | |
3459 | unsigned long long ldsize; | |
3460 | ||
3461 | if (level != LEVEL_CONTAINER) | |
3462 | return 0; | |
3463 | if (!dev) | |
3464 | return 1; | |
3465 | ||
3466 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
3467 | if (fd < 0) { | |
2c514b71 | 3468 | if (verbose) |
e7b84f9d | 3469 | pr_err("ddf: Cannot open %s: %s\n", |
613b0d17 | 3470 | dev, strerror(errno)); |
a322f70c DW |
3471 | return 0; |
3472 | } | |
3473 | if (!get_dev_size(fd, dev, &ldsize)) { | |
3474 | close(fd); | |
3475 | return 0; | |
3476 | } | |
3477 | close(fd); | |
3478 | ||
387fcd59 | 3479 | *freesize = avail_size_ddf(st, ldsize >> 9, INVALID_SECTORS); |
ea17e7aa N |
3480 | if (*freesize == 0) |
3481 | return 0; | |
a322f70c DW |
3482 | |
3483 | return 1; | |
3484 | } | |
3485 | ||
78e44928 NB |
3486 | static int validate_geometry_ddf_bvd(struct supertype *st, |
3487 | int level, int layout, int raiddisks, | |
c21e737b | 3488 | int *chunk, unsigned long long size, |
af4348dd | 3489 | unsigned long long data_offset, |
2c514b71 NB |
3490 | char *dev, unsigned long long *freesize, |
3491 | int verbose) | |
a322f70c DW |
3492 | { |
3493 | struct stat stb; | |
3494 | struct ddf_super *ddf = st->sb; | |
3495 | struct dl *dl; | |
5f8097be | 3496 | unsigned long long maxsize; |
a322f70c | 3497 | /* ddf/bvd supports lots of things, but not containers */ |
b42f577a N |
3498 | if (level == LEVEL_CONTAINER) { |
3499 | if (verbose) | |
e7b84f9d | 3500 | pr_err("DDF cannot create a container within an container\n"); |
a322f70c | 3501 | return 0; |
b42f577a | 3502 | } |
a322f70c DW |
3503 | /* We must have the container info already read in. */ |
3504 | if (!ddf) | |
3505 | return 0; | |
3506 | ||
5f8097be NB |
3507 | if (!dev) { |
3508 | /* General test: make sure there is space for | |
3509 | * 'raiddisks' device extents of size 'size'. | |
3510 | */ | |
3511 | unsigned long long minsize = size; | |
3512 | int dcnt = 0; | |
3513 | if (minsize == 0) | |
3514 | minsize = 8; | |
56cb05c4 | 3515 | for (dl = ddf->dlist; dl ; dl = dl->next) { |
476066a3 | 3516 | if (find_space(ddf, dl, data_offset, &minsize) |
fca65520 | 3517 | != INVALID_SECTORS) |
5f8097be | 3518 | dcnt++; |
5f8097be NB |
3519 | } |
3520 | if (dcnt < raiddisks) { | |
2c514b71 | 3521 | if (verbose) |
7a862a02 | 3522 | pr_err("ddf: Not enough devices with space for this array (%d < %d)\n", |
e7b84f9d | 3523 | dcnt, raiddisks); |
5f8097be NB |
3524 | return 0; |
3525 | } | |
3526 | return 1; | |
3527 | } | |
a322f70c DW |
3528 | /* This device must be a member of the set */ |
3529 | if (stat(dev, &stb) < 0) | |
3530 | return 0; | |
3531 | if ((S_IFMT & stb.st_mode) != S_IFBLK) | |
3532 | return 0; | |
3533 | for (dl = ddf->dlist ; dl ; dl = dl->next) { | |
f21e18ca N |
3534 | if (dl->major == (int)major(stb.st_rdev) && |
3535 | dl->minor == (int)minor(stb.st_rdev)) | |
a322f70c DW |
3536 | break; |
3537 | } | |
5f8097be | 3538 | if (!dl) { |
2c514b71 | 3539 | if (verbose) |
7a862a02 | 3540 | pr_err("ddf: %s is not in the same DDF set\n", |
613b0d17 | 3541 | dev); |
5f8097be NB |
3542 | return 0; |
3543 | } | |
fca65520 | 3544 | maxsize = ULLONG_MAX; |
476066a3 | 3545 | find_space(ddf, dl, data_offset, &maxsize); |
5f8097be | 3546 | *freesize = maxsize; |
a322f70c DW |
3547 | |
3548 | return 1; | |
3549 | } | |
59e36268 | 3550 | |
a322f70c | 3551 | static int load_super_ddf_all(struct supertype *st, int fd, |
e1902a7b | 3552 | void **sbp, char *devname) |
a322f70c DW |
3553 | { |
3554 | struct mdinfo *sra; | |
3555 | struct ddf_super *super; | |
3556 | struct mdinfo *sd, *best = NULL; | |
3557 | int bestseq = 0; | |
3558 | int seq; | |
3559 | char nm[20]; | |
3560 | int dfd; | |
3561 | ||
b526e52d | 3562 | sra = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE); |
a322f70c DW |
3563 | if (!sra) |
3564 | return 1; | |
3565 | if (sra->array.major_version != -1 || | |
3566 | sra->array.minor_version != -2 || | |
3567 | strcmp(sra->text_version, "ddf") != 0) | |
3568 | return 1; | |
3569 | ||
6416d527 | 3570 | if (posix_memalign((void**)&super, 512, sizeof(*super)) != 0) |
a322f70c | 3571 | return 1; |
a2349791 | 3572 | memset(super, 0, sizeof(*super)); |
a322f70c DW |
3573 | |
3574 | /* first, try each device, and choose the best ddf */ | |
3575 | for (sd = sra->devs ; sd ; sd = sd->next) { | |
3576 | int rv; | |
3577 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
7a7cc504 NB |
3578 | dfd = dev_open(nm, O_RDONLY); |
3579 | if (dfd < 0) | |
a322f70c DW |
3580 | return 2; |
3581 | rv = load_ddf_headers(dfd, super, NULL); | |
7a7cc504 | 3582 | close(dfd); |
a322f70c | 3583 | if (rv == 0) { |
60931cf9 | 3584 | seq = be32_to_cpu(super->active->seq); |
a322f70c DW |
3585 | if (super->active->openflag) |
3586 | seq--; | |
3587 | if (!best || seq > bestseq) { | |
3588 | bestseq = seq; | |
3589 | best = sd; | |
3590 | } | |
3591 | } | |
3592 | } | |
3593 | if (!best) | |
3594 | return 1; | |
3595 | /* OK, load this ddf */ | |
3596 | sprintf(nm, "%d:%d", best->disk.major, best->disk.minor); | |
3597 | dfd = dev_open(nm, O_RDONLY); | |
7a7cc504 | 3598 | if (dfd < 0) |
a322f70c DW |
3599 | return 1; |
3600 | load_ddf_headers(dfd, super, NULL); | |
3601 | load_ddf_global(dfd, super, NULL); | |
3602 | close(dfd); | |
3603 | /* Now we need the device-local bits */ | |
3604 | for (sd = sra->devs ; sd ; sd = sd->next) { | |
3d2c4fc7 DW |
3605 | int rv; |
3606 | ||
a322f70c | 3607 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); |
e1902a7b | 3608 | dfd = dev_open(nm, O_RDWR); |
7a7cc504 | 3609 | if (dfd < 0) |
a322f70c | 3610 | return 2; |
3d2c4fc7 DW |
3611 | rv = load_ddf_headers(dfd, super, NULL); |
3612 | if (rv == 0) | |
e1902a7b | 3613 | rv = load_ddf_local(dfd, super, NULL, 1); |
3d2c4fc7 DW |
3614 | if (rv) |
3615 | return 1; | |
a322f70c | 3616 | } |
33414a01 | 3617 | |
a322f70c DW |
3618 | *sbp = super; |
3619 | if (st->ss == NULL) { | |
78e44928 | 3620 | st->ss = &super_ddf; |
a322f70c DW |
3621 | st->minor_version = 0; |
3622 | st->max_devs = 512; | |
3623 | } | |
4dd2df09 | 3624 | strcpy(st->container_devnm, fd2devnm(fd)); |
a322f70c DW |
3625 | return 0; |
3626 | } | |
2b959fbf N |
3627 | |
3628 | static int load_container_ddf(struct supertype *st, int fd, | |
3629 | char *devname) | |
3630 | { | |
3631 | return load_super_ddf_all(st, fd, &st->sb, devname); | |
3632 | } | |
3633 | ||
a5c7adb3 | 3634 | static int check_secondary(const struct vcl *vc) |
3635 | { | |
3636 | const struct vd_config *conf = &vc->conf; | |
3637 | int i; | |
3638 | ||
3639 | /* The only DDF secondary RAID level md can support is | |
3640 | * RAID 10, if the stripe sizes and Basic volume sizes | |
3641 | * are all equal. | |
3642 | * Other configurations could in theory be supported by exposing | |
3643 | * the BVDs to user space and using device mapper for the secondary | |
3644 | * mapping. So far we don't support that. | |
3645 | */ | |
3646 | ||
3647 | __u64 sec_elements[4] = {0, 0, 0, 0}; | |
3648 | #define __set_sec_seen(n) (sec_elements[(n)>>6] |= (1<<((n)&63))) | |
3649 | #define __was_sec_seen(n) ((sec_elements[(n)>>6] & (1<<((n)&63))) != 0) | |
3650 | ||
3651 | if (vc->other_bvds == NULL) { | |
3652 | pr_err("No BVDs for secondary RAID found\n"); | |
3653 | return -1; | |
3654 | } | |
3655 | if (conf->prl != DDF_RAID1) { | |
3656 | pr_err("Secondary RAID level only supported for mirrored BVD\n"); | |
3657 | return -1; | |
3658 | } | |
3659 | if (conf->srl != DDF_2STRIPED && conf->srl != DDF_2SPANNED) { | |
3660 | pr_err("Secondary RAID level %d is unsupported\n", | |
3661 | conf->srl); | |
3662 | return -1; | |
3663 | } | |
3664 | __set_sec_seen(conf->sec_elmnt_seq); | |
3665 | for (i = 0; i < conf->sec_elmnt_count-1; i++) { | |
3666 | const struct vd_config *bvd = vc->other_bvds[i]; | |
3c48f7be | 3667 | if (bvd->sec_elmnt_seq == DDF_UNUSED_BVD) |
c98567ba | 3668 | continue; |
a5c7adb3 | 3669 | if (bvd->srl != conf->srl) { |
3670 | pr_err("Inconsistent secondary RAID level across BVDs\n"); | |
3671 | return -1; | |
3672 | } | |
3673 | if (bvd->prl != conf->prl) { | |
3674 | pr_err("Different RAID levels for BVDs are unsupported\n"); | |
3675 | return -1; | |
3676 | } | |
a8173e43 | 3677 | if (!be16_eq(bvd->prim_elmnt_count, conf->prim_elmnt_count)) { |
a5c7adb3 | 3678 | pr_err("All BVDs must have the same number of primary elements\n"); |
3679 | return -1; | |
3680 | } | |
3681 | if (bvd->chunk_shift != conf->chunk_shift) { | |
3682 | pr_err("Different strip sizes for BVDs are unsupported\n"); | |
3683 | return -1; | |
3684 | } | |
9d0c6b70 | 3685 | if (!be64_eq(bvd->array_blocks, conf->array_blocks)) { |
a5c7adb3 | 3686 | pr_err("Different BVD sizes are unsupported\n"); |
3687 | return -1; | |
3688 | } | |
3689 | __set_sec_seen(bvd->sec_elmnt_seq); | |
3690 | } | |
3691 | for (i = 0; i < conf->sec_elmnt_count; i++) { | |
3692 | if (!__was_sec_seen(i)) { | |
20d430ca | 3693 | /* pr_err("BVD %d is missing\n", i); */ |
a5c7adb3 | 3694 | return -1; |
3695 | } | |
3696 | } | |
3697 | return 0; | |
3698 | } | |
3699 | ||
8a38db86 | 3700 | static unsigned int get_pd_index_from_refnum(const struct vcl *vc, |
60931cf9 | 3701 | be32 refnum, unsigned int nmax, |
4e587018 | 3702 | const struct vd_config **bvd, |
3703 | unsigned int *idx) | |
8a38db86 | 3704 | { |
4e587018 | 3705 | unsigned int i, j, n, sec, cnt; |
3706 | ||
a8173e43 | 3707 | cnt = be16_to_cpu(vc->conf.prim_elmnt_count); |
4e587018 | 3708 | sec = (vc->conf.sec_elmnt_count == 1 ? 0 : vc->conf.sec_elmnt_seq); |
3709 | ||
3710 | for (i = 0, j = 0 ; i < nmax ; i++) { | |
3711 | /* j counts valid entries for this BVD */ | |
60931cf9 | 3712 | if (be32_eq(vc->conf.phys_refnum[i], refnum)) { |
4e587018 | 3713 | *bvd = &vc->conf; |
3714 | *idx = i; | |
56cb05c4 | 3715 | return sec * cnt + j; |
4e587018 | 3716 | } |
56cb05c4 N |
3717 | if (be32_to_cpu(vc->conf.phys_refnum[i]) != 0xffffffff) |
3718 | j++; | |
4e587018 | 3719 | } |
3720 | if (vc->other_bvds == NULL) | |
3721 | goto bad; | |
3722 | ||
3723 | for (n = 1; n < vc->conf.sec_elmnt_count; n++) { | |
3724 | struct vd_config *vd = vc->other_bvds[n-1]; | |
4e587018 | 3725 | sec = vd->sec_elmnt_seq; |
3c48f7be | 3726 | if (sec == DDF_UNUSED_BVD) |
3727 | continue; | |
4e587018 | 3728 | for (i = 0, j = 0 ; i < nmax ; i++) { |
60931cf9 | 3729 | if (be32_eq(vd->phys_refnum[i], refnum)) { |
4e587018 | 3730 | *bvd = vd; |
3731 | *idx = i; | |
56cb05c4 | 3732 | return sec * cnt + j; |
4e587018 | 3733 | } |
56cb05c4 N |
3734 | if (be32_to_cpu(vd->phys_refnum[i]) != 0xffffffff) |
3735 | j++; | |
4e587018 | 3736 | } |
3737 | } | |
3738 | bad: | |
3739 | *bvd = NULL; | |
d6e7b083 | 3740 | return DDF_NOTFOUND; |
8a38db86 | 3741 | } |
3742 | ||
00bbdbda | 3743 | static struct mdinfo *container_content_ddf(struct supertype *st, char *subarray) |
598f0d58 NB |
3744 | { |
3745 | /* Given a container loaded by load_super_ddf_all, | |
3746 | * extract information about all the arrays into | |
3747 | * an mdinfo tree. | |
3748 | * | |
3749 | * For each vcl in conflist: create an mdinfo, fill it in, | |
3750 | * then look for matching devices (phys_refnum) in dlist | |
3751 | * and create appropriate device mdinfo. | |
3752 | */ | |
3753 | struct ddf_super *ddf = st->sb; | |
3754 | struct mdinfo *rest = NULL; | |
3755 | struct vcl *vc; | |
3756 | ||
56cb05c4 | 3757 | for (vc = ddf->conflist ; vc ; vc=vc->next) { |
f21e18ca | 3758 | unsigned int i; |
598f0d58 | 3759 | struct mdinfo *this; |
00bbdbda | 3760 | char *ep; |
90fa1a29 | 3761 | __u32 *cptr; |
8a38db86 | 3762 | unsigned int pd; |
00bbdbda N |
3763 | |
3764 | if (subarray && | |
3765 | (strtoul(subarray, &ep, 10) != vc->vcnum || | |
3766 | *ep != '\0')) | |
3767 | continue; | |
3768 | ||
a5c7adb3 | 3769 | if (vc->conf.sec_elmnt_count > 1) { |
3770 | if (check_secondary(vc) != 0) | |
3771 | continue; | |
3772 | } | |
3773 | ||
503975b9 | 3774 | this = xcalloc(1, sizeof(*this)); |
598f0d58 NB |
3775 | this->next = rest; |
3776 | rest = this; | |
3777 | ||
8a2848a7 | 3778 | if (layout_ddf2md(&vc->conf, &this->array)) |
3779 | continue; | |
598f0d58 | 3780 | this->array.md_minor = -1; |
f35f2525 N |
3781 | this->array.major_version = -1; |
3782 | this->array.minor_version = -2; | |
56cb05c4 | 3783 | this->safe_mode_delay = DDF_SAFE_MODE_DELAY; |
90fa1a29 JS |
3784 | cptr = (__u32 *)(vc->conf.guid + 16); |
3785 | this->array.ctime = DECADE + __be32_to_cpu(*cptr); | |
598f0d58 | 3786 | this->array.utime = DECADE + |
60931cf9 | 3787 | be32_to_cpu(vc->conf.timestamp); |
598f0d58 NB |
3788 | this->array.chunk_size = 512 << vc->conf.chunk_shift; |
3789 | ||
59e36268 | 3790 | i = vc->vcnum; |
7a7cc504 NB |
3791 | if ((ddf->virt->entries[i].state & DDF_state_inconsistent) || |
3792 | (ddf->virt->entries[i].init_state & DDF_initstate_mask) != | |
ed9d66aa | 3793 | DDF_init_full) { |
598f0d58 | 3794 | this->array.state = 0; |
ed9d66aa NB |
3795 | this->resync_start = 0; |
3796 | } else { | |
598f0d58 | 3797 | this->array.state = 1; |
b7528a20 | 3798 | this->resync_start = MaxSector; |
ed9d66aa | 3799 | } |
8bf989d8 | 3800 | _ddf_array_name(this->name, ddf, i); |
598f0d58 | 3801 | memset(this->uuid, 0, sizeof(this->uuid)); |
56cb05c4 N |
3802 | this->component_size = be64_to_cpu(vc->conf.blocks); |
3803 | this->array.size = this->component_size / 2; | |
3804 | this->container_member = i; | |
598f0d58 | 3805 | |
c5afc314 N |
3806 | ddf->currentconf = vc; |
3807 | uuid_from_super_ddf(st, this->uuid); | |
f646805e | 3808 | if (!subarray) |
3809 | ddf->currentconf = NULL; | |
c5afc314 | 3810 | |
60f18132 | 3811 | sprintf(this->text_version, "/%s/%d", |
4dd2df09 | 3812 | st->container_devnm, this->container_member); |
60f18132 | 3813 | |
217dead4 | 3814 | for (pd = 0; pd < be16_to_cpu(ddf->phys->max_pdes); pd++) { |
598f0d58 NB |
3815 | struct mdinfo *dev; |
3816 | struct dl *d; | |
4e587018 | 3817 | const struct vd_config *bvd; |
3818 | unsigned int iphys; | |
fa033bec | 3819 | int stt; |
598f0d58 | 3820 | |
60931cf9 | 3821 | if (be32_to_cpu(ddf->phys->entries[pd].refnum) |
3822 | == 0xFFFFFFFF) | |
bc17324f | 3823 | continue; |
0cf5ef67 | 3824 | |
a8173e43 | 3825 | stt = be16_to_cpu(ddf->phys->entries[pd].state); |
fa033bec N |
3826 | if ((stt & (DDF_Online|DDF_Failed|DDF_Rebuilding)) |
3827 | != DDF_Online) | |
3828 | continue; | |
3829 | ||
8a38db86 | 3830 | i = get_pd_index_from_refnum( |
4e587018 | 3831 | vc, ddf->phys->entries[pd].refnum, |
3832 | ddf->mppe, &bvd, &iphys); | |
d6e7b083 | 3833 | if (i == DDF_NOTFOUND) |
8a38db86 | 3834 | continue; |
3835 | ||
fa033bec | 3836 | this->array.working_disks++; |
bc17324f | 3837 | |
0cf5ef67 | 3838 | for (d = ddf->dlist; d ; d=d->next) |
60931cf9 | 3839 | if (be32_eq(d->disk.refnum, |
3840 | ddf->phys->entries[pd].refnum)) | |
0cf5ef67 N |
3841 | break; |
3842 | if (d == NULL) | |
3843 | /* Haven't found that one yet, maybe there are others */ | |
3844 | continue; | |
3845 | ||
503975b9 | 3846 | dev = xcalloc(1, sizeof(*dev)); |
56cb05c4 N |
3847 | dev->next = this->devs; |
3848 | this->devs = dev; | |
598f0d58 | 3849 | |
60931cf9 | 3850 | dev->disk.number = be32_to_cpu(d->disk.refnum); |
56cb05c4 N |
3851 | dev->disk.major = d->major; |
3852 | dev->disk.minor = d->minor; | |
598f0d58 | 3853 | dev->disk.raid_disk = i; |
56cb05c4 | 3854 | dev->disk.state = (1<<MD_DISK_SYNC)|(1<<MD_DISK_ACTIVE); |
d23534e4 | 3855 | dev->recovery_start = MaxSector; |
598f0d58 | 3856 | |
56cb05c4 | 3857 | dev->events = be32_to_cpu(ddf->active->seq); |
57a66662 | 3858 | dev->data_offset = |
9d0c6b70 | 3859 | be64_to_cpu(LBA_OFFSET(ddf, bvd)[iphys]); |
3860 | dev->component_size = be64_to_cpu(bvd->blocks); | |
598f0d58 NB |
3861 | if (d->devname) |
3862 | strcpy(dev->name, d->devname); | |
3863 | } | |
3864 | } | |
3865 | return rest; | |
3866 | } | |
3867 | ||
955e9ea1 | 3868 | static int store_super_ddf(struct supertype *st, int fd) |
a322f70c | 3869 | { |
955e9ea1 | 3870 | struct ddf_super *ddf = st->sb; |
a322f70c | 3871 | unsigned long long dsize; |
6416d527 | 3872 | void *buf; |
3d2c4fc7 | 3873 | int rc; |
a322f70c | 3874 | |
955e9ea1 DW |
3875 | if (!ddf) |
3876 | return 1; | |
3877 | ||
a322f70c DW |
3878 | if (!get_dev_size(fd, NULL, &dsize)) |
3879 | return 1; | |
3880 | ||
dbf98368 | 3881 | if (ddf->dlist || ddf->conflist) { |
3882 | struct stat sta; | |
3883 | struct dl *dl; | |
3884 | int ofd, ret; | |
3885 | ||
3886 | if (fstat(fd, &sta) == -1 || !S_ISBLK(sta.st_mode)) { | |
1ade5cc1 | 3887 | pr_err("file descriptor for invalid device\n"); |
dbf98368 | 3888 | return 1; |
3889 | } | |
3890 | for (dl = ddf->dlist; dl; dl = dl->next) | |
3891 | if (dl->major == (int)major(sta.st_rdev) && | |
3892 | dl->minor == (int)minor(sta.st_rdev)) | |
3893 | break; | |
3894 | if (!dl) { | |
1ade5cc1 | 3895 | pr_err("couldn't find disk %d/%d\n", |
dbf98368 | 3896 | (int)major(sta.st_rdev), |
3897 | (int)minor(sta.st_rdev)); | |
3898 | return 1; | |
3899 | } | |
dbf98368 | 3900 | ofd = dl->fd; |
3901 | dl->fd = fd; | |
3921e41a | 3902 | ret = (_write_super_to_disk(ddf, dl) != 1); |
dbf98368 | 3903 | dl->fd = ofd; |
3904 | return ret; | |
3905 | } | |
3906 | ||
3d2c4fc7 DW |
3907 | if (posix_memalign(&buf, 512, 512) != 0) |
3908 | return 1; | |
6416d527 NB |
3909 | memset(buf, 0, 512); |
3910 | ||
a322f70c | 3911 | lseek64(fd, dsize-512, 0); |
3d2c4fc7 | 3912 | rc = write(fd, buf, 512); |
6416d527 | 3913 | free(buf); |
3d2c4fc7 DW |
3914 | if (rc < 0) |
3915 | return 1; | |
a322f70c DW |
3916 | return 0; |
3917 | } | |
3918 | ||
a19c88b8 NB |
3919 | static int compare_super_ddf(struct supertype *st, struct supertype *tst) |
3920 | { | |
3921 | /* | |
3922 | * return: | |
3923 | * 0 same, or first was empty, and second was copied | |
56cb05c4 | 3924 | * 1 second had wrong magic number - but that isn't possible |
a19c88b8 NB |
3925 | * 2 wrong uuid |
3926 | * 3 wrong other info | |
3927 | */ | |
3928 | struct ddf_super *first = st->sb; | |
3929 | struct ddf_super *second = tst->sb; | |
4eefd651 | 3930 | struct dl *dl1, *dl2; |
3931 | struct vcl *vl1, *vl2; | |
2d210697 | 3932 | unsigned int max_vds, max_pds, pd, vd; |
a19c88b8 NB |
3933 | |
3934 | if (!first) { | |
3935 | st->sb = tst->sb; | |
3936 | tst->sb = NULL; | |
3937 | return 0; | |
3938 | } | |
3939 | ||
3940 | if (memcmp(first->anchor.guid, second->anchor.guid, DDF_GUID_LEN) != 0) | |
3941 | return 2; | |
3942 | ||
f43f5b32 N |
3943 | /* It is only OK to compare info in the anchor. Anything else |
3944 | * could be changing due to a reconfig so must be ignored. | |
3945 | * guid really should be enough anyway. | |
3946 | */ | |
2d210697 | 3947 | |
f43f5b32 | 3948 | if (!be32_eq(first->active->seq, second->active->seq)) { |
1ade5cc1 | 3949 | dprintf("sequence number mismatch %u<->%u\n", |
f43f5b32 N |
3950 | be32_to_cpu(first->active->seq), |
3951 | be32_to_cpu(second->active->seq)); | |
3952 | return 0; | |
2d210697 | 3953 | } |
2d210697 | 3954 | |
4eefd651 | 3955 | /* |
56cb05c4 N |
3956 | * At this point we are fairly sure that the meta data matches. |
3957 | * But the new disk may contain additional local data. | |
3958 | * Add it to the super block. | |
4eefd651 | 3959 | */ |
f43f5b32 | 3960 | max_vds = be16_to_cpu(first->active->max_vd_entries); |
217dead4 | 3961 | max_pds = be16_to_cpu(first->phys->max_pdes); |
4eefd651 | 3962 | for (vl2 = second->conflist; vl2; vl2 = vl2->next) { |
3963 | for (vl1 = first->conflist; vl1; vl1 = vl1->next) | |
3964 | if (!memcmp(vl1->conf.guid, vl2->conf.guid, | |
3965 | DDF_GUID_LEN)) | |
3966 | break; | |
3967 | if (vl1) { | |
3968 | if (vl1->other_bvds != NULL && | |
3969 | vl1->conf.sec_elmnt_seq != | |
3970 | vl2->conf.sec_elmnt_seq) { | |
1ade5cc1 | 3971 | dprintf("adding BVD %u\n", |
4eefd651 | 3972 | vl2->conf.sec_elmnt_seq); |
3973 | add_other_bvd(vl1, &vl2->conf, | |
3974 | first->conf_rec_len*512); | |
3975 | } | |
3976 | continue; | |
3977 | } | |
3978 | ||
3979 | if (posix_memalign((void **)&vl1, 512, | |
3980 | (first->conf_rec_len*512 + | |
3981 | offsetof(struct vcl, conf))) != 0) { | |
1ade5cc1 | 3982 | pr_err("could not allocate vcl buf\n"); |
4eefd651 | 3983 | return 3; |
3984 | } | |
3985 | ||
3986 | vl1->next = first->conflist; | |
3987 | vl1->block_sizes = NULL; | |
4eefd651 | 3988 | memcpy(&vl1->conf, &vl2->conf, first->conf_rec_len*512); |
3c48f7be | 3989 | if (alloc_other_bvds(first, vl1) != 0) { |
1ade5cc1 | 3990 | pr_err("could not allocate other bvds\n"); |
3c48f7be | 3991 | free(vl1); |
3992 | return 3; | |
3993 | } | |
4eefd651 | 3994 | for (vd = 0; vd < max_vds; vd++) |
3995 | if (!memcmp(first->virt->entries[vd].guid, | |
3996 | vl1->conf.guid, DDF_GUID_LEN)) | |
3997 | break; | |
3998 | vl1->vcnum = vd; | |
1ade5cc1 | 3999 | dprintf("added config for VD %u\n", vl1->vcnum); |
4eefd651 | 4000 | first->conflist = vl1; |
4001 | } | |
4002 | ||
4003 | for (dl2 = second->dlist; dl2; dl2 = dl2->next) { | |
4004 | for (dl1 = first->dlist; dl1; dl1 = dl1->next) | |
60931cf9 | 4005 | if (be32_eq(dl1->disk.refnum, dl2->disk.refnum)) |
4eefd651 | 4006 | break; |
4007 | if (dl1) | |
4008 | continue; | |
4009 | ||
4010 | if (posix_memalign((void **)&dl1, 512, | |
4011 | sizeof(*dl1) + (first->max_part) * sizeof(dl1->vlist[0])) | |
4012 | != 0) { | |
1ade5cc1 | 4013 | pr_err("could not allocate disk info buffer\n"); |
4eefd651 | 4014 | return 3; |
4015 | } | |
4016 | memcpy(dl1, dl2, sizeof(*dl1)); | |
4017 | dl1->mdupdate = NULL; | |
4018 | dl1->next = first->dlist; | |
4019 | dl1->fd = -1; | |
4020 | for (pd = 0; pd < max_pds; pd++) | |
60931cf9 | 4021 | if (be32_eq(first->phys->entries[pd].refnum, |
4022 | dl1->disk.refnum)) | |
4eefd651 | 4023 | break; |
41bcbc14 | 4024 | dl1->pdnum = pd < max_pds ? (int)pd : -1; |
4eefd651 | 4025 | if (dl2->spare) { |
4026 | if (posix_memalign((void **)&dl1->spare, 512, | |
4027 | first->conf_rec_len*512) != 0) { | |
1ade5cc1 | 4028 | pr_err("could not allocate spare info buf\n"); |
4eefd651 | 4029 | return 3; |
4030 | } | |
4031 | memcpy(dl1->spare, dl2->spare, first->conf_rec_len*512); | |
4032 | } | |
4033 | for (vd = 0 ; vd < first->max_part ; vd++) { | |
4034 | if (!dl2->vlist[vd]) { | |
4035 | dl1->vlist[vd] = NULL; | |
4036 | continue; | |
4037 | } | |
4038 | for (vl1 = first->conflist; vl1; vl1 = vl1->next) { | |
4039 | if (!memcmp(vl1->conf.guid, | |
4040 | dl2->vlist[vd]->conf.guid, | |
4041 | DDF_GUID_LEN)) | |
4042 | break; | |
4043 | dl1->vlist[vd] = vl1; | |
4044 | } | |
4045 | } | |
4046 | first->dlist = dl1; | |
1ade5cc1 | 4047 | dprintf("added disk %d: %08x\n", dl1->pdnum, |
60931cf9 | 4048 | be32_to_cpu(dl1->disk.refnum)); |
4eefd651 | 4049 | } |
4050 | ||
a19c88b8 NB |
4051 | return 0; |
4052 | } | |
4053 | ||
4e5528c6 NB |
4054 | /* |
4055 | * A new array 'a' has been started which claims to be instance 'inst' | |
4056 | * within container 'c'. | |
4057 | * We need to confirm that the array matches the metadata in 'c' so | |
4058 | * that we don't corrupt any metadata. | |
4059 | */ | |
cba0191b | 4060 | static int ddf_open_new(struct supertype *c, struct active_array *a, char *inst) |
549e9569 | 4061 | { |
a2aa439e | 4062 | struct ddf_super *ddf = c->sb; |
4063 | int n = atoi(inst); | |
5daa35ac | 4064 | struct mdinfo *dev; |
4065 | struct dl *dl; | |
4066 | static const char faulty[] = "faulty"; | |
4067 | ||
fb9d0acb | 4068 | if (all_ff(ddf->virt->entries[n].guid)) { |
1ade5cc1 | 4069 | pr_err("subarray %d doesn't exist\n", n); |
a2aa439e | 4070 | return -ENODEV; |
4071 | } | |
1ade5cc1 | 4072 | dprintf("new subarray %d, GUID: %s\n", n, |
5daa35ac | 4073 | guid_str(ddf->virt->entries[n].guid)); |
4074 | for (dev = a->info.devs; dev; dev = dev->next) { | |
4075 | for (dl = ddf->dlist; dl; dl = dl->next) | |
4076 | if (dl->major == dev->disk.major && | |
4077 | dl->minor == dev->disk.minor) | |
4078 | break; | |
a44e993e | 4079 | if (!dl || dl->pdnum < 0) { |
1ade5cc1 N |
4080 | pr_err("device %d/%d of subarray %d not found in meta data\n", |
4081 | dev->disk.major, dev->disk.minor, n); | |
5daa35ac | 4082 | return -1; |
4083 | } | |
4084 | if ((be16_to_cpu(ddf->phys->entries[dl->pdnum].state) & | |
4085 | (DDF_Online|DDF_Missing|DDF_Failed)) != DDF_Online) { | |
1ade5cc1 N |
4086 | pr_err("new subarray %d contains broken device %d/%d (%02x)\n", |
4087 | n, dl->major, dl->minor, | |
4088 | be16_to_cpu(ddf->phys->entries[dl->pdnum].state)); | |
5daa35ac | 4089 | if (write(dev->state_fd, faulty, sizeof(faulty)-1) != |
4090 | sizeof(faulty) - 1) | |
4091 | pr_err("Write to state_fd failed\n"); | |
4092 | dev->curr_state = DS_FAULTY; | |
4093 | } | |
4094 | } | |
a2aa439e | 4095 | a->info.container_member = n; |
549e9569 NB |
4096 | return 0; |
4097 | } | |
4098 | ||
e5a03804 | 4099 | static void handle_missing(struct ddf_super *ddf, struct active_array *a, int inst) |
5a46fcd7 N |
4100 | { |
4101 | /* This member array is being activated. If any devices | |
4102 | * are missing they must now be marked as failed. | |
4103 | */ | |
4104 | struct vd_config *vc; | |
4105 | unsigned int n_bvd; | |
4106 | struct vcl *vcl; | |
4107 | struct dl *dl; | |
e5a03804 | 4108 | int pd; |
5a46fcd7 | 4109 | int n; |
e5a03804 | 4110 | int state; |
5a46fcd7 N |
4111 | |
4112 | for (n = 0; ; n++) { | |
4113 | vc = find_vdcr(ddf, inst, n, &n_bvd, &vcl); | |
4114 | if (!vc) | |
4115 | break; | |
4116 | for (dl = ddf->dlist; dl; dl = dl->next) | |
4117 | if (be32_eq(dl->disk.refnum, vc->phys_refnum[n_bvd])) | |
4118 | break; | |
4119 | if (dl) | |
4120 | /* Found this disk, so not missing */ | |
4121 | continue; | |
e5a03804 N |
4122 | |
4123 | /* Mark the device as failed/missing. */ | |
4124 | pd = find_phys(ddf, vc->phys_refnum[n_bvd]); | |
4125 | if (pd >= 0 && be16_and(ddf->phys->entries[pd].state, | |
4126 | cpu_to_be16(DDF_Online))) { | |
4127 | be16_clear(ddf->phys->entries[pd].state, | |
4128 | cpu_to_be16(DDF_Online)); | |
4129 | be16_set(ddf->phys->entries[pd].state, | |
4130 | cpu_to_be16(DDF_Failed|DDF_Missing)); | |
4131 | vc->phys_refnum[n_bvd] = cpu_to_be32(0); | |
609ce161 | 4132 | ddf_set_updates_pending(ddf, vc); |
e5a03804 N |
4133 | } |
4134 | ||
4135 | /* Mark the array as Degraded */ | |
4136 | state = get_svd_state(ddf, vcl); | |
4137 | if (ddf->virt->entries[inst].state != | |
4138 | ((ddf->virt->entries[inst].state & ~DDF_state_mask) | |
4139 | | state)) { | |
4140 | ddf->virt->entries[inst].state = | |
4141 | (ddf->virt->entries[inst].state & ~DDF_state_mask) | |
4142 | | state; | |
4143 | a->check_degraded = 1; | |
609ce161 | 4144 | ddf_set_updates_pending(ddf, vc); |
e5a03804 | 4145 | } |
5a46fcd7 N |
4146 | } |
4147 | } | |
4148 | ||
4e5528c6 NB |
4149 | /* |
4150 | * The array 'a' is to be marked clean in the metadata. | |
ed9d66aa | 4151 | * If '->resync_start' is not ~(unsigned long long)0, then the array is only |
4e5528c6 NB |
4152 | * clean up to the point (in sectors). If that cannot be recorded in the |
4153 | * metadata, then leave it as dirty. | |
4154 | * | |
4155 | * For DDF, we need to clear the DDF_state_inconsistent bit in the | |
4156 | * !global! virtual_disk.virtual_entry structure. | |
4157 | */ | |
01f157d7 | 4158 | static int ddf_set_array_state(struct active_array *a, int consistent) |
549e9569 | 4159 | { |
4e5528c6 NB |
4160 | struct ddf_super *ddf = a->container->sb; |
4161 | int inst = a->info.container_member; | |
18a2f463 | 4162 | int old = ddf->virt->entries[inst].state; |
01f157d7 | 4163 | if (consistent == 2) { |
e5a03804 | 4164 | handle_missing(ddf, a, inst); |
01f157d7 | 4165 | consistent = 1; |
b7941fd6 | 4166 | if (!is_resync_complete(&a->info)) |
01f157d7 N |
4167 | consistent = 0; |
4168 | } | |
ed9d66aa NB |
4169 | if (consistent) |
4170 | ddf->virt->entries[inst].state &= ~DDF_state_inconsistent; | |
4171 | else | |
4e5528c6 | 4172 | ddf->virt->entries[inst].state |= DDF_state_inconsistent; |
18a2f463 | 4173 | if (old != ddf->virt->entries[inst].state) |
609ce161 | 4174 | ddf_set_updates_pending(ddf, NULL); |
18a2f463 NB |
4175 | |
4176 | old = ddf->virt->entries[inst].init_state; | |
ed9d66aa | 4177 | ddf->virt->entries[inst].init_state &= ~DDF_initstate_mask; |
b7941fd6 | 4178 | if (is_resync_complete(&a->info)) |
ed9d66aa | 4179 | ddf->virt->entries[inst].init_state |= DDF_init_full; |
b7941fd6 | 4180 | else if (a->info.resync_start == 0) |
ed9d66aa | 4181 | ddf->virt->entries[inst].init_state |= DDF_init_not; |
4e5528c6 | 4182 | else |
ed9d66aa | 4183 | ddf->virt->entries[inst].init_state |= DDF_init_quick; |
18a2f463 | 4184 | if (old != ddf->virt->entries[inst].init_state) |
609ce161 | 4185 | ddf_set_updates_pending(ddf, NULL); |
ed9d66aa | 4186 | |
b27336a2 | 4187 | dprintf("ddf mark %d/%s (%d) %s %llu\n", inst, |
4188 | guid_str(ddf->virt->entries[inst].guid), a->curr_state, | |
4189 | consistent?"clean":"dirty", | |
b7941fd6 | 4190 | a->info.resync_start); |
01f157d7 | 4191 | return consistent; |
fd7cde1b DW |
4192 | } |
4193 | ||
5ec636b7 | 4194 | static int get_bvd_state(const struct ddf_super *ddf, |
4195 | const struct vd_config *vc) | |
4196 | { | |
4197 | unsigned int i, n_bvd, working = 0; | |
a8173e43 | 4198 | unsigned int n_prim = be16_to_cpu(vc->prim_elmnt_count); |
5ec636b7 | 4199 | int pd, st, state; |
f70d549f N |
4200 | char *avail = xcalloc(1, n_prim); |
4201 | mdu_array_info_t array; | |
4202 | ||
4203 | layout_ddf2md(vc, &array); | |
4204 | ||
5ec636b7 | 4205 | for (i = 0; i < n_prim; i++) { |
4206 | if (!find_index_in_bvd(ddf, vc, i, &n_bvd)) | |
4207 | continue; | |
4208 | pd = find_phys(ddf, vc->phys_refnum[n_bvd]); | |
4209 | if (pd < 0) | |
4210 | continue; | |
a8173e43 | 4211 | st = be16_to_cpu(ddf->phys->entries[pd].state); |
5ec636b7 | 4212 | if ((st & (DDF_Online|DDF_Failed|DDF_Rebuilding)) |
f70d549f | 4213 | == DDF_Online) { |
5ec636b7 | 4214 | working++; |
f70d549f N |
4215 | avail[i] = 1; |
4216 | } | |
5ec636b7 | 4217 | } |
4218 | ||
4219 | state = DDF_state_degraded; | |
4220 | if (working == n_prim) | |
4221 | state = DDF_state_optimal; | |
4222 | else | |
4223 | switch (vc->prl) { | |
4224 | case DDF_RAID0: | |
4225 | case DDF_CONCAT: | |
4226 | case DDF_JBOD: | |
4227 | state = DDF_state_failed; | |
4228 | break; | |
4229 | case DDF_RAID1: | |
4230 | if (working == 0) | |
4231 | state = DDF_state_failed; | |
4232 | else if (working >= 2) | |
4233 | state = DDF_state_part_optimal; | |
4234 | break; | |
f70d549f N |
4235 | case DDF_RAID1E: |
4236 | if (!enough(10, n_prim, array.layout, 1, avail)) | |
4237 | state = DDF_state_failed; | |
4238 | break; | |
5ec636b7 | 4239 | case DDF_RAID4: |
4240 | case DDF_RAID5: | |
4241 | if (working < n_prim - 1) | |
4242 | state = DDF_state_failed; | |
4243 | break; | |
4244 | case DDF_RAID6: | |
4245 | if (working < n_prim - 2) | |
4246 | state = DDF_state_failed; | |
4247 | else if (working == n_prim - 1) | |
4248 | state = DDF_state_part_optimal; | |
4249 | break; | |
4250 | } | |
4251 | return state; | |
4252 | } | |
4253 | ||
0777d17d | 4254 | static int secondary_state(int state, int other, int seclevel) |
4255 | { | |
4256 | if (state == DDF_state_optimal && other == DDF_state_optimal) | |
4257 | return DDF_state_optimal; | |
4258 | if (seclevel == DDF_2MIRRORED) { | |
4259 | if (state == DDF_state_optimal || other == DDF_state_optimal) | |
4260 | return DDF_state_part_optimal; | |
4261 | if (state == DDF_state_failed && other == DDF_state_failed) | |
4262 | return DDF_state_failed; | |
4263 | return DDF_state_degraded; | |
4264 | } else { | |
4265 | if (state == DDF_state_failed || other == DDF_state_failed) | |
4266 | return DDF_state_failed; | |
4267 | if (state == DDF_state_degraded || other == DDF_state_degraded) | |
4268 | return DDF_state_degraded; | |
4269 | return DDF_state_part_optimal; | |
4270 | } | |
4271 | } | |
4272 | ||
4273 | static int get_svd_state(const struct ddf_super *ddf, const struct vcl *vcl) | |
4274 | { | |
4275 | int state = get_bvd_state(ddf, &vcl->conf); | |
4276 | unsigned int i; | |
4277 | for (i = 1; i < vcl->conf.sec_elmnt_count; i++) { | |
4278 | state = secondary_state( | |
4279 | state, | |
4280 | get_bvd_state(ddf, vcl->other_bvds[i-1]), | |
4281 | vcl->conf.srl); | |
4282 | } | |
4283 | return state; | |
4284 | } | |
4285 | ||
7a7cc504 NB |
4286 | /* |
4287 | * The state of each disk is stored in the global phys_disk structure | |
4288 | * in phys_disk.entries[n].state. | |
4289 | * This makes various combinations awkward. | |
4290 | * - When a device fails in any array, it must be failed in all arrays | |
4291 | * that include a part of this device. | |
4292 | * - When a component is rebuilding, we cannot include it officially in the | |
4293 | * array unless this is the only array that uses the device. | |
4294 | * | |
4295 | * So: when transitioning: | |
4296 | * Online -> failed, just set failed flag. monitor will propagate | |
4297 | * spare -> online, the device might need to be added to the array. | |
4298 | * spare -> failed, just set failed. Don't worry if in array or not. | |
4299 | */ | |
8d45d196 | 4300 | static void ddf_set_disk(struct active_array *a, int n, int state) |
549e9569 | 4301 | { |
7a7cc504 | 4302 | struct ddf_super *ddf = a->container->sb; |
baba3f4e | 4303 | unsigned int inst = a->info.container_member, n_bvd; |
4304 | struct vcl *vcl; | |
4305 | struct vd_config *vc = find_vdcr(ddf, inst, (unsigned int)n, | |
4306 | &n_bvd, &vcl); | |
4307 | int pd; | |
e1316fab N |
4308 | struct mdinfo *mdi; |
4309 | struct dl *dl; | |
609ce161 | 4310 | int update = 0; |
7a7cc504 | 4311 | |
1ade5cc1 | 4312 | dprintf("%d to %x\n", n, state); |
7a7cc504 | 4313 | if (vc == NULL) { |
2c514b71 | 4314 | dprintf("ddf: cannot find instance %d!!\n", inst); |
7a7cc504 NB |
4315 | return; |
4316 | } | |
e1316fab N |
4317 | /* Find the matching slot in 'info'. */ |
4318 | for (mdi = a->info.devs; mdi; mdi = mdi->next) | |
4319 | if (mdi->disk.raid_disk == n) | |
4320 | break; | |
ce6844b9 | 4321 | if (!mdi) { |
1ade5cc1 | 4322 | pr_err("cannot find raid disk %d\n", n); |
e1316fab | 4323 | return; |
ce6844b9 | 4324 | } |
e1316fab N |
4325 | |
4326 | /* and find the 'dl' entry corresponding to that. */ | |
4327 | for (dl = ddf->dlist; dl; dl = dl->next) | |
77632af9 N |
4328 | if (mdi->state_fd >= 0 && |
4329 | mdi->disk.major == dl->major && | |
e1316fab N |
4330 | mdi->disk.minor == dl->minor) |
4331 | break; | |
ce6844b9 | 4332 | if (!dl) { |
1ade5cc1 N |
4333 | pr_err("cannot find raid disk %d (%d/%d)\n", |
4334 | n, mdi->disk.major, mdi->disk.minor); | |
e1316fab | 4335 | return; |
ce6844b9 | 4336 | } |
e1316fab | 4337 | |
baba3f4e | 4338 | pd = find_phys(ddf, vc->phys_refnum[n_bvd]); |
e1316fab N |
4339 | if (pd < 0 || pd != dl->pdnum) { |
4340 | /* disk doesn't currently exist or has changed. | |
4341 | * If it is now in_sync, insert it. */ | |
1ade5cc1 N |
4342 | dprintf("phys disk not found for %d: %d/%d ref %08x\n", |
4343 | dl->pdnum, dl->major, dl->minor, | |
60931cf9 | 4344 | be32_to_cpu(dl->disk.refnum)); |
1ade5cc1 N |
4345 | dprintf("array %u disk %u ref %08x pd %d\n", |
4346 | inst, n_bvd, | |
60931cf9 | 4347 | be32_to_cpu(vc->phys_refnum[n_bvd]), pd); |
98fbc0ff N |
4348 | if ((state & DS_INSYNC) && ! (state & DS_FAULTY) && |
4349 | dl->pdnum >= 0) { | |
4350 | pd = dl->pdnum; | |
baba3f4e | 4351 | vc->phys_refnum[n_bvd] = dl->disk.refnum; |
57a66662 | 4352 | LBA_OFFSET(ddf, vc)[n_bvd] = |
9d0c6b70 | 4353 | cpu_to_be64(mdi->data_offset); |
a8173e43 | 4354 | be16_clear(ddf->phys->entries[pd].type, |
4355 | cpu_to_be16(DDF_Global_Spare)); | |
4356 | be16_set(ddf->phys->entries[pd].type, | |
4357 | cpu_to_be16(DDF_Active_in_VD)); | |
609ce161 | 4358 | update = 1; |
7a7cc504 NB |
4359 | } |
4360 | } else { | |
a8173e43 | 4361 | be16 old = ddf->phys->entries[pd].state; |
7a7cc504 | 4362 | if (state & DS_FAULTY) |
a8173e43 | 4363 | be16_set(ddf->phys->entries[pd].state, |
4364 | cpu_to_be16(DDF_Failed)); | |
7a7cc504 | 4365 | if (state & DS_INSYNC) { |
a8173e43 | 4366 | be16_set(ddf->phys->entries[pd].state, |
4367 | cpu_to_be16(DDF_Online)); | |
4368 | be16_clear(ddf->phys->entries[pd].state, | |
4369 | cpu_to_be16(DDF_Rebuilding)); | |
7a7cc504 | 4370 | } |
a8173e43 | 4371 | if (!be16_eq(old, ddf->phys->entries[pd].state)) |
609ce161 | 4372 | update = 1; |
7a7cc504 NB |
4373 | } |
4374 | ||
ce6844b9 MW |
4375 | dprintf("ddf: set_disk %d (%08x) to %x->%02x\n", n, |
4376 | be32_to_cpu(dl->disk.refnum), state, | |
4377 | be16_to_cpu(ddf->phys->entries[pd].state)); | |
7e1432fb | 4378 | |
7a7cc504 NB |
4379 | /* Now we need to check the state of the array and update |
4380 | * virtual_disk.entries[n].state. | |
4381 | * It needs to be one of "optimal", "degraded", "failed". | |
4382 | * I don't understand 'deleted' or 'missing'. | |
4383 | */ | |
0777d17d | 4384 | state = get_svd_state(ddf, vcl); |
7a7cc504 | 4385 | |
18a2f463 NB |
4386 | if (ddf->virt->entries[inst].state != |
4387 | ((ddf->virt->entries[inst].state & ~DDF_state_mask) | |
4388 | | state)) { | |
18a2f463 NB |
4389 | ddf->virt->entries[inst].state = |
4390 | (ddf->virt->entries[inst].state & ~DDF_state_mask) | |
4391 | | state; | |
609ce161 | 4392 | update = 1; |
18a2f463 | 4393 | } |
609ce161 N |
4394 | if (update) |
4395 | ddf_set_updates_pending(ddf, vc); | |
549e9569 NB |
4396 | } |
4397 | ||
2e735d19 | 4398 | static void ddf_sync_metadata(struct supertype *st) |
549e9569 | 4399 | { |
7a7cc504 NB |
4400 | /* |
4401 | * Write all data to all devices. | |
4402 | * Later, we might be able to track whether only local changes | |
4403 | * have been made, or whether any global data has been changed, | |
4404 | * but ddf is sufficiently weird that it probably always | |
4405 | * changes global data .... | |
4406 | */ | |
18a2f463 NB |
4407 | struct ddf_super *ddf = st->sb; |
4408 | if (!ddf->updates_pending) | |
4409 | return; | |
4410 | ddf->updates_pending = 0; | |
3921e41a | 4411 | __write_init_super_ddf(st); |
2c514b71 | 4412 | dprintf("ddf: sync_metadata\n"); |
549e9569 NB |
4413 | } |
4414 | ||
f646805e | 4415 | static int del_from_conflist(struct vcl **list, const char *guid) |
4416 | { | |
4417 | struct vcl **p; | |
4418 | int found = 0; | |
4419 | for (p = list; p && *p; p = &((*p)->next)) | |
4420 | if (!memcmp((*p)->conf.guid, guid, DDF_GUID_LEN)) { | |
4421 | found = 1; | |
4422 | *p = (*p)->next; | |
4423 | } | |
4424 | return found; | |
4425 | } | |
4426 | ||
4427 | static int _kill_subarray_ddf(struct ddf_super *ddf, const char *guid) | |
4428 | { | |
4429 | struct dl *dl; | |
4430 | unsigned int vdnum, i; | |
4431 | vdnum = find_vde_by_guid(ddf, guid); | |
4432 | if (vdnum == DDF_NOTFOUND) { | |
1ade5cc1 | 4433 | pr_err("could not find VD %s\n", guid_str(guid)); |
f646805e | 4434 | return -1; |
4435 | } | |
4436 | if (del_from_conflist(&ddf->conflist, guid) == 0) { | |
1ade5cc1 | 4437 | pr_err("could not find conf %s\n", guid_str(guid)); |
f646805e | 4438 | return -1; |
4439 | } | |
4440 | for (dl = ddf->dlist; dl; dl = dl->next) | |
4441 | for (i = 0; i < ddf->max_part; i++) | |
4442 | if (dl->vlist[i] != NULL && | |
4443 | !memcmp(dl->vlist[i]->conf.guid, guid, | |
4444 | DDF_GUID_LEN)) | |
4445 | dl->vlist[i] = NULL; | |
4446 | memset(ddf->virt->entries[vdnum].guid, 0xff, DDF_GUID_LEN); | |
1ade5cc1 | 4447 | dprintf("deleted %s\n", guid_str(guid)); |
f646805e | 4448 | return 0; |
4449 | } | |
4450 | ||
4451 | static int kill_subarray_ddf(struct supertype *st) | |
4452 | { | |
4453 | struct ddf_super *ddf = st->sb; | |
4454 | /* | |
4455 | * currentconf is set in container_content_ddf, | |
4456 | * called with subarray arg | |
4457 | */ | |
4458 | struct vcl *victim = ddf->currentconf; | |
4459 | struct vd_config *conf; | |
f646805e | 4460 | unsigned int vdnum; |
56cb05c4 N |
4461 | |
4462 | ddf->currentconf = NULL; | |
f646805e | 4463 | if (!victim) { |
1ade5cc1 | 4464 | pr_err("nothing to kill\n"); |
f646805e | 4465 | return -1; |
4466 | } | |
4467 | conf = &victim->conf; | |
4468 | vdnum = find_vde_by_guid(ddf, conf->guid); | |
4469 | if (vdnum == DDF_NOTFOUND) { | |
1ade5cc1 | 4470 | pr_err("could not find VD %s\n", guid_str(conf->guid)); |
f646805e | 4471 | return -1; |
4472 | } | |
4473 | if (st->update_tail) { | |
4474 | struct virtual_disk *vd; | |
4475 | int len = sizeof(struct virtual_disk) | |
4476 | + sizeof(struct virtual_entry); | |
4477 | vd = xmalloc(len); | |
4478 | if (vd == NULL) { | |
1ade5cc1 | 4479 | pr_err("failed to allocate %d bytes\n", len); |
f646805e | 4480 | return -1; |
4481 | } | |
4482 | memset(vd, 0 , len); | |
4483 | vd->magic = DDF_VIRT_RECORDS_MAGIC; | |
a8173e43 | 4484 | vd->populated_vdes = cpu_to_be16(0); |
f646805e | 4485 | memcpy(vd->entries[0].guid, conf->guid, DDF_GUID_LEN); |
4486 | /* we use DDF_state_deleted as marker */ | |
4487 | vd->entries[0].state = DDF_state_deleted; | |
4488 | append_metadata_update(st, vd, len); | |
6a350d82 | 4489 | } else { |
f646805e | 4490 | _kill_subarray_ddf(ddf, conf->guid); |
609ce161 | 4491 | ddf_set_updates_pending(ddf, NULL); |
6a350d82 | 4492 | ddf_sync_metadata(st); |
4493 | } | |
f646805e | 4494 | return 0; |
4495 | } | |
4496 | ||
c5943560 | 4497 | static void copy_matching_bvd(struct ddf_super *ddf, |
4498 | struct vd_config *conf, | |
4499 | const struct metadata_update *update) | |
4500 | { | |
4501 | unsigned int mppe = | |
a8173e43 | 4502 | be16_to_cpu(ddf->anchor.max_primary_element_entries); |
c5943560 | 4503 | unsigned int len = ddf->conf_rec_len * 512; |
4504 | char *p; | |
4505 | struct vd_config *vc; | |
4506 | for (p = update->buf; p < update->buf + update->len; p += len) { | |
4507 | vc = (struct vd_config *) p; | |
4508 | if (vc->sec_elmnt_seq == conf->sec_elmnt_seq) { | |
4509 | memcpy(conf->phys_refnum, vc->phys_refnum, | |
4510 | mppe * (sizeof(__u32) + sizeof(__u64))); | |
4511 | return; | |
4512 | } | |
4513 | } | |
1ade5cc1 | 4514 | pr_err("no match for BVD %d of %s in update\n", |
c5943560 | 4515 | conf->sec_elmnt_seq, guid_str(conf->guid)); |
4516 | } | |
4517 | ||
fea6a6c0 N |
4518 | static void ddf_process_phys_update(struct supertype *st, |
4519 | struct metadata_update *update) | |
4520 | { | |
4521 | struct ddf_super *ddf = st->sb; | |
4522 | struct phys_disk *pd; | |
4523 | unsigned int ent; | |
4524 | ||
4525 | pd = (struct phys_disk*)update->buf; | |
4526 | ent = be16_to_cpu(pd->used_pdes); | |
4527 | if (ent >= be16_to_cpu(ddf->phys->max_pdes)) | |
4528 | return; | |
4529 | if (be16_and(pd->entries[0].state, cpu_to_be16(DDF_Missing))) { | |
4530 | struct dl **dlp; | |
4531 | /* removing this disk. */ | |
4532 | be16_set(ddf->phys->entries[ent].state, | |
4533 | cpu_to_be16(DDF_Missing)); | |
4534 | for (dlp = &ddf->dlist; *dlp; dlp = &(*dlp)->next) { | |
4535 | struct dl *dl = *dlp; | |
4536 | if (dl->pdnum == (signed)ent) { | |
4537 | close(dl->fd); | |
4538 | dl->fd = -1; | |
fea6a6c0 | 4539 | *dlp = dl->next; |
de910774 N |
4540 | update->space = dl->devname; |
4541 | *(void**)dl = update->space_list; | |
4542 | update->space_list = (void**)dl; | |
fea6a6c0 N |
4543 | break; |
4544 | } | |
4545 | } | |
4546 | ddf_set_updates_pending(ddf, NULL); | |
4547 | return; | |
4548 | } | |
4549 | if (!all_ff(ddf->phys->entries[ent].guid)) | |
4550 | return; | |
4551 | ddf->phys->entries[ent] = pd->entries[0]; | |
4552 | ddf->phys->used_pdes = cpu_to_be16 | |
4553 | (1 + be16_to_cpu(ddf->phys->used_pdes)); | |
4554 | ddf_set_updates_pending(ddf, NULL); | |
4555 | if (ddf->add_list) { | |
4556 | struct active_array *a; | |
4557 | struct dl *al = ddf->add_list; | |
4558 | ddf->add_list = al->next; | |
4559 | ||
4560 | al->next = ddf->dlist; | |
4561 | ddf->dlist = al; | |
4562 | ||
4563 | /* As a device has been added, we should check | |
4564 | * for any degraded devices that might make | |
4565 | * use of this spare */ | |
4566 | for (a = st->arrays ; a; a=a->next) | |
4567 | a->check_degraded = 1; | |
4568 | } | |
4569 | } | |
4570 | ||
4571 | static void ddf_process_virt_update(struct supertype *st, | |
4572 | struct metadata_update *update) | |
4573 | { | |
4574 | struct ddf_super *ddf = st->sb; | |
4575 | struct virtual_disk *vd; | |
4576 | unsigned int ent; | |
4577 | ||
4578 | vd = (struct virtual_disk*)update->buf; | |
4579 | ||
4580 | if (vd->entries[0].state == DDF_state_deleted) { | |
4581 | if (_kill_subarray_ddf(ddf, vd->entries[0].guid)) | |
4582 | return; | |
4583 | } else { | |
4584 | ent = find_vde_by_guid(ddf, vd->entries[0].guid); | |
4585 | if (ent != DDF_NOTFOUND) { | |
1ade5cc1 N |
4586 | dprintf("VD %s exists already in slot %d\n", |
4587 | guid_str(vd->entries[0].guid), | |
fea6a6c0 N |
4588 | ent); |
4589 | return; | |
4590 | } | |
4591 | ent = find_unused_vde(ddf); | |
4592 | if (ent == DDF_NOTFOUND) | |
4593 | return; | |
4594 | ddf->virt->entries[ent] = vd->entries[0]; | |
4595 | ddf->virt->populated_vdes = | |
4596 | cpu_to_be16( | |
4597 | 1 + be16_to_cpu( | |
4598 | ddf->virt->populated_vdes)); | |
1ade5cc1 N |
4599 | dprintf("added VD %s in slot %d(s=%02x i=%02x)\n", |
4600 | guid_str(vd->entries[0].guid), ent, | |
fea6a6c0 N |
4601 | ddf->virt->entries[ent].state, |
4602 | ddf->virt->entries[ent].init_state); | |
4603 | } | |
4604 | ddf_set_updates_pending(ddf, NULL); | |
4605 | } | |
4606 | ||
4607 | static void ddf_remove_failed(struct ddf_super *ddf) | |
4608 | { | |
4609 | /* Now remove any 'Failed' devices that are not part | |
4610 | * of any VD. They will have the Transition flag set. | |
4611 | * Once done, we need to update all dl->pdnum numbers. | |
4612 | */ | |
4613 | unsigned int pdnum; | |
4614 | unsigned int pd2 = 0; | |
4615 | struct dl *dl; | |
4616 | ||
4617 | for (pdnum = 0; pdnum < be16_to_cpu(ddf->phys->max_pdes); | |
4618 | pdnum++) { | |
4619 | if (be32_to_cpu(ddf->phys->entries[pdnum].refnum) == | |
4620 | 0xFFFFFFFF) | |
4621 | continue; | |
4622 | if (be16_and(ddf->phys->entries[pdnum].state, | |
4623 | cpu_to_be16(DDF_Failed)) | |
4624 | && be16_and(ddf->phys->entries[pdnum].state, | |
4625 | cpu_to_be16(DDF_Transition))) { | |
4626 | /* skip this one unless in dlist*/ | |
4627 | for (dl = ddf->dlist; dl; dl = dl->next) | |
4628 | if (dl->pdnum == (int)pdnum) | |
4629 | break; | |
4630 | if (!dl) | |
4631 | continue; | |
4632 | } | |
4633 | if (pdnum == pd2) | |
4634 | pd2++; | |
4635 | else { | |
4636 | ddf->phys->entries[pd2] = | |
4637 | ddf->phys->entries[pdnum]; | |
4638 | for (dl = ddf->dlist; dl; dl = dl->next) | |
4639 | if (dl->pdnum == (int)pdnum) | |
4640 | dl->pdnum = pd2; | |
4641 | pd2++; | |
4642 | } | |
4643 | } | |
4644 | ddf->phys->used_pdes = cpu_to_be16(pd2); | |
4645 | while (pd2 < pdnum) { | |
4646 | memset(ddf->phys->entries[pd2].guid, 0xff, | |
4647 | DDF_GUID_LEN); | |
4648 | pd2++; | |
4649 | } | |
4650 | } | |
4651 | ||
4652 | static void ddf_update_vlist(struct ddf_super *ddf, struct dl *dl) | |
4653 | { | |
4654 | struct vcl *vcl; | |
4655 | unsigned int vn = 0; | |
4656 | int in_degraded = 0; | |
4657 | ||
4658 | if (dl->pdnum < 0) | |
4659 | return; | |
4660 | for (vcl = ddf->conflist; vcl ; vcl = vcl->next) { | |
4661 | unsigned int dn, ibvd; | |
4662 | const struct vd_config *conf; | |
4663 | int vstate; | |
4664 | dn = get_pd_index_from_refnum(vcl, | |
4665 | dl->disk.refnum, | |
4666 | ddf->mppe, | |
4667 | &conf, &ibvd); | |
4668 | if (dn == DDF_NOTFOUND) | |
4669 | continue; | |
4670 | dprintf("dev %d/%08x has %s (sec=%u) at %d\n", | |
4671 | dl->pdnum, | |
4672 | be32_to_cpu(dl->disk.refnum), | |
4673 | guid_str(conf->guid), | |
4674 | conf->sec_elmnt_seq, vn); | |
4675 | /* Clear the Transition flag */ | |
4676 | if (be16_and | |
4677 | (ddf->phys->entries[dl->pdnum].state, | |
4678 | cpu_to_be16(DDF_Failed))) | |
4679 | be16_clear(ddf->phys | |
4680 | ->entries[dl->pdnum].state, | |
4681 | cpu_to_be16(DDF_Transition)); | |
4682 | dl->vlist[vn++] = vcl; | |
4683 | vstate = ddf->virt->entries[vcl->vcnum].state | |
4684 | & DDF_state_mask; | |
4685 | if (vstate == DDF_state_degraded || | |
4686 | vstate == DDF_state_part_optimal) | |
4687 | in_degraded = 1; | |
4688 | } | |
4689 | while (vn < ddf->max_part) | |
4690 | dl->vlist[vn++] = NULL; | |
4691 | if (dl->vlist[0]) { | |
4692 | be16_clear(ddf->phys->entries[dl->pdnum].type, | |
4693 | cpu_to_be16(DDF_Global_Spare)); | |
4694 | if (!be16_and(ddf->phys | |
4695 | ->entries[dl->pdnum].type, | |
4696 | cpu_to_be16(DDF_Active_in_VD))) { | |
4697 | be16_set(ddf->phys | |
4698 | ->entries[dl->pdnum].type, | |
4699 | cpu_to_be16(DDF_Active_in_VD)); | |
4700 | if (in_degraded) | |
4701 | be16_set(ddf->phys | |
4702 | ->entries[dl->pdnum] | |
4703 | .state, | |
4704 | cpu_to_be16 | |
4705 | (DDF_Rebuilding)); | |
4706 | } | |
4707 | } | |
4708 | if (dl->spare) { | |
4709 | be16_clear(ddf->phys->entries[dl->pdnum].type, | |
4710 | cpu_to_be16(DDF_Global_Spare)); | |
4711 | be16_set(ddf->phys->entries[dl->pdnum].type, | |
4712 | cpu_to_be16(DDF_Spare)); | |
4713 | } | |
4714 | if (!dl->vlist[0] && !dl->spare) { | |
4715 | be16_set(ddf->phys->entries[dl->pdnum].type, | |
4716 | cpu_to_be16(DDF_Global_Spare)); | |
4717 | be16_clear(ddf->phys->entries[dl->pdnum].type, | |
4718 | cpu_to_be16(DDF_Spare)); | |
4719 | be16_clear(ddf->phys->entries[dl->pdnum].type, | |
4720 | cpu_to_be16(DDF_Active_in_VD)); | |
4721 | } | |
4722 | } | |
4723 | ||
4724 | static void ddf_process_conf_update(struct supertype *st, | |
4725 | struct metadata_update *update) | |
4726 | { | |
4727 | struct ddf_super *ddf = st->sb; | |
4728 | struct vd_config *vc; | |
4729 | struct vcl *vcl; | |
4730 | struct dl *dl; | |
4731 | unsigned int ent; | |
4732 | unsigned int pdnum, len; | |
4733 | ||
4734 | vc = (struct vd_config*)update->buf; | |
4735 | len = ddf->conf_rec_len * 512; | |
4736 | if ((unsigned int)update->len != len * vc->sec_elmnt_count) { | |
1ade5cc1 N |
4737 | pr_err("%s: insufficient data (%d) for %u BVDs\n", |
4738 | guid_str(vc->guid), update->len, | |
fea6a6c0 N |
4739 | vc->sec_elmnt_count); |
4740 | return; | |
4741 | } | |
4742 | for (vcl = ddf->conflist; vcl ; vcl = vcl->next) | |
4743 | if (memcmp(vcl->conf.guid, vc->guid, DDF_GUID_LEN) == 0) | |
4744 | break; | |
1ade5cc1 | 4745 | dprintf("conf update for %s (%s)\n", |
fea6a6c0 N |
4746 | guid_str(vc->guid), (vcl ? "old" : "new")); |
4747 | if (vcl) { | |
4748 | /* An update, just copy the phys_refnum and lba_offset | |
4749 | * fields | |
4750 | */ | |
4751 | unsigned int i; | |
4752 | unsigned int k; | |
4753 | copy_matching_bvd(ddf, &vcl->conf, update); | |
4754 | for (k = 0; k < be16_to_cpu(vc->prim_elmnt_count); k++) | |
4755 | dprintf("BVD %u has %08x at %llu\n", 0, | |
4756 | be32_to_cpu(vcl->conf.phys_refnum[k]), | |
4757 | be64_to_cpu(LBA_OFFSET(ddf, | |
4758 | &vcl->conf)[k])); | |
4759 | for (i = 1; i < vc->sec_elmnt_count; i++) { | |
4760 | copy_matching_bvd(ddf, vcl->other_bvds[i-1], | |
4761 | update); | |
4762 | for (k = 0; k < be16_to_cpu( | |
4763 | vc->prim_elmnt_count); k++) | |
4764 | dprintf("BVD %u has %08x at %llu\n", i, | |
4765 | be32_to_cpu | |
4766 | (vcl->other_bvds[i-1]-> | |
4767 | phys_refnum[k]), | |
4768 | be64_to_cpu | |
4769 | (LBA_OFFSET | |
4770 | (ddf, | |
4771 | vcl->other_bvds[i-1])[k])); | |
4772 | } | |
4773 | } else { | |
4774 | /* A new VD_CONF */ | |
4775 | unsigned int i; | |
4776 | if (!update->space) | |
4777 | return; | |
4778 | vcl = update->space; | |
4779 | update->space = NULL; | |
4780 | vcl->next = ddf->conflist; | |
4781 | memcpy(&vcl->conf, vc, len); | |
4782 | ent = find_vde_by_guid(ddf, vc->guid); | |
4783 | if (ent == DDF_NOTFOUND) | |
4784 | return; | |
4785 | vcl->vcnum = ent; | |
4786 | ddf->conflist = vcl; | |
4787 | for (i = 1; i < vc->sec_elmnt_count; i++) | |
4788 | memcpy(vcl->other_bvds[i-1], | |
4789 | update->buf + len * i, len); | |
4790 | } | |
4791 | /* Set DDF_Transition on all Failed devices - to help | |
4792 | * us detect those that are no longer in use | |
4793 | */ | |
4794 | for (pdnum = 0; pdnum < be16_to_cpu(ddf->phys->max_pdes); | |
4795 | pdnum++) | |
4796 | if (be16_and(ddf->phys->entries[pdnum].state, | |
4797 | cpu_to_be16(DDF_Failed))) | |
4798 | be16_set(ddf->phys->entries[pdnum].state, | |
4799 | cpu_to_be16(DDF_Transition)); | |
4800 | ||
4801 | /* Now make sure vlist is correct for each dl. */ | |
4802 | for (dl = ddf->dlist; dl; dl = dl->next) | |
4803 | ddf_update_vlist(ddf, dl); | |
4804 | ddf_remove_failed(ddf); | |
4805 | ||
4806 | ddf_set_updates_pending(ddf, vc); | |
4807 | } | |
4808 | ||
88c164f4 NB |
4809 | static void ddf_process_update(struct supertype *st, |
4810 | struct metadata_update *update) | |
4811 | { | |
4812 | /* Apply this update to the metadata. | |
4813 | * The first 4 bytes are a DDF_*_MAGIC which guides | |
4814 | * our actions. | |
4815 | * Possible update are: | |
4816 | * DDF_PHYS_RECORDS_MAGIC | |
4dd968cc N |
4817 | * Add a new physical device or remove an old one. |
4818 | * Changes to this record only happen implicitly. | |
88c164f4 NB |
4819 | * used_pdes is the device number. |
4820 | * DDF_VIRT_RECORDS_MAGIC | |
4821 | * Add a new VD. Possibly also change the 'access' bits. | |
4822 | * populated_vdes is the entry number. | |
4823 | * DDF_VD_CONF_MAGIC | |
4824 | * New or updated VD. the VIRT_RECORD must already | |
4825 | * exist. For an update, phys_refnum and lba_offset | |
4826 | * (at least) are updated, and the VD_CONF must | |
4827 | * be written to precisely those devices listed with | |
4828 | * a phys_refnum. | |
4829 | * DDF_SPARE_ASSIGN_MAGIC | |
4830 | * replacement Spare Assignment Record... but for which device? | |
4831 | * | |
4832 | * So, e.g.: | |
4833 | * - to create a new array, we send a VIRT_RECORD and | |
4834 | * a VD_CONF. Then assemble and start the array. | |
4835 | * - to activate a spare we send a VD_CONF to add the phys_refnum | |
4836 | * and offset. This will also mark the spare as active with | |
4837 | * a spare-assignment record. | |
4838 | */ | |
60931cf9 | 4839 | be32 *magic = (be32 *)update->buf; |
88c164f4 | 4840 | |
60931cf9 | 4841 | dprintf("Process update %x\n", be32_to_cpu(*magic)); |
7e1432fb | 4842 | |
60931cf9 | 4843 | if (be32_eq(*magic, DDF_PHYS_RECORDS_MAGIC)) { |
fea6a6c0 | 4844 | if (update->len == (sizeof(struct phys_disk) + |
88c164f4 | 4845 | sizeof(struct phys_disk_entry))) |
fea6a6c0 | 4846 | ddf_process_phys_update(st, update); |
60931cf9 | 4847 | } else if (be32_eq(*magic, DDF_VIRT_RECORDS_MAGIC)) { |
fea6a6c0 | 4848 | if (update->len == (sizeof(struct virtual_disk) + |
88c164f4 | 4849 | sizeof(struct virtual_entry))) |
fea6a6c0 N |
4850 | ddf_process_virt_update(st, update); |
4851 | } else if (be32_eq(*magic, DDF_VD_CONF_MAGIC)) { | |
4852 | ddf_process_conf_update(st, update); | |
88c164f4 | 4853 | } |
60931cf9 | 4854 | /* case DDF_SPARE_ASSIGN_MAGIC */ |
88c164f4 NB |
4855 | } |
4856 | ||
5fe6f031 N |
4857 | static int ddf_prepare_update(struct supertype *st, |
4858 | struct metadata_update *update) | |
edd8d13c NB |
4859 | { |
4860 | /* This update arrived at managemon. | |
4861 | * We are about to pass it to monitor. | |
4862 | * If a malloc is needed, do it here. | |
4863 | */ | |
4864 | struct ddf_super *ddf = st->sb; | |
1f17f96b N |
4865 | be32 *magic; |
4866 | if (update->len < 4) | |
4867 | return 0; | |
4868 | magic = (be32 *)update->buf; | |
60931cf9 | 4869 | if (be32_eq(*magic, DDF_VD_CONF_MAGIC)) { |
c5943560 | 4870 | struct vcl *vcl; |
1f17f96b N |
4871 | struct vd_config *conf; |
4872 | if (update->len < (int)sizeof(*conf)) | |
4873 | return 0; | |
4874 | conf = (struct vd_config *) update->buf; | |
e6b9548d | 4875 | if (posix_memalign(&update->space, 512, |
613b0d17 | 4876 | offsetof(struct vcl, conf) |
c5943560 | 4877 | + ddf->conf_rec_len * 512) != 0) { |
4878 | update->space = NULL; | |
5fe6f031 | 4879 | return 0; |
c5943560 | 4880 | } |
4881 | vcl = update->space; | |
4882 | vcl->conf.sec_elmnt_count = conf->sec_elmnt_count; | |
4883 | if (alloc_other_bvds(ddf, vcl) != 0) { | |
4884 | free(update->space); | |
e6b9548d | 4885 | update->space = NULL; |
5fe6f031 | 4886 | return 0; |
c5943560 | 4887 | } |
4888 | } | |
5fe6f031 | 4889 | return 1; |
edd8d13c NB |
4890 | } |
4891 | ||
7733b91d | 4892 | /* |
4893 | * Check degraded state of a RAID10. | |
4894 | * returns 2 for good, 1 for degraded, 0 for failed, and -1 for error | |
4895 | */ | |
4896 | static int raid10_degraded(struct mdinfo *info) | |
4897 | { | |
4898 | int n_prim, n_bvds; | |
4899 | int i; | |
9591a2de | 4900 | struct mdinfo *d; |
7733b91d | 4901 | char *found; |
4902 | int ret = -1; | |
4903 | ||
7733b91d | 4904 | n_prim = info->array.layout & ~0x100; |
4905 | n_bvds = info->array.raid_disks / n_prim; | |
4906 | found = xmalloc(n_bvds); | |
4907 | if (found == NULL) | |
4908 | return ret; | |
4909 | memset(found, 0, n_bvds); | |
4910 | for (d = info->devs; d; d = d->next) { | |
4911 | i = d->disk.raid_disk / n_prim; | |
4912 | if (i >= n_bvds) { | |
1ade5cc1 | 4913 | pr_err("BUG: invalid raid disk\n"); |
7733b91d | 4914 | goto out; |
4915 | } | |
4916 | if (d->state_fd > 0) | |
4917 | found[i]++; | |
4918 | } | |
4919 | ret = 2; | |
4920 | for (i = 0; i < n_bvds; i++) | |
4921 | if (!found[i]) { | |
1ade5cc1 | 4922 | dprintf("BVD %d/%d failed\n", i, n_bvds); |
7733b91d | 4923 | ret = 0; |
4924 | goto out; | |
4925 | } else if (found[i] < n_prim) { | |
1ade5cc1 | 4926 | dprintf("BVD %d/%d degraded\n", i, n_bvds); |
7733b91d | 4927 | ret = 1; |
4928 | } | |
4929 | out: | |
4930 | free(found); | |
4931 | return ret; | |
4932 | } | |
4933 | ||
7e1432fb NB |
4934 | /* |
4935 | * Check if the array 'a' is degraded but not failed. | |
4936 | * If it is, find as many spares as are available and needed and | |
4937 | * arrange for their inclusion. | |
4938 | * We only choose devices which are not already in the array, | |
4939 | * and prefer those with a spare-assignment to this array. | |
56cb05c4 | 4940 | * Otherwise we choose global spares - assuming always that |
7e1432fb NB |
4941 | * there is enough room. |
4942 | * For each spare that we assign, we return an 'mdinfo' which | |
4943 | * describes the position for the device in the array. | |
4944 | * We also add to 'updates' a DDF_VD_CONF_MAGIC update with | |
4945 | * the new phys_refnum and lba_offset values. | |
4946 | * | |
4947 | * Only worry about BVDs at the moment. | |
4948 | */ | |
4949 | static struct mdinfo *ddf_activate_spare(struct active_array *a, | |
4950 | struct metadata_update **updates) | |
4951 | { | |
4952 | int working = 0; | |
4953 | struct mdinfo *d; | |
4954 | struct ddf_super *ddf = a->container->sb; | |
4955 | int global_ok = 0; | |
4956 | struct mdinfo *rv = NULL; | |
4957 | struct mdinfo *di; | |
4958 | struct metadata_update *mu; | |
4959 | struct dl *dl; | |
4960 | int i; | |
0c78849f | 4961 | unsigned int j; |
baba3f4e | 4962 | struct vcl *vcl; |
7e1432fb | 4963 | struct vd_config *vc; |
baba3f4e | 4964 | unsigned int n_bvd; |
7e1432fb | 4965 | |
7e1432fb NB |
4966 | for (d = a->info.devs ; d ; d = d->next) { |
4967 | if ((d->curr_state & DS_FAULTY) && | |
613b0d17 | 4968 | d->state_fd >= 0) |
7e1432fb NB |
4969 | /* wait for Removal to happen */ |
4970 | return NULL; | |
4971 | if (d->state_fd >= 0) | |
4972 | working ++; | |
4973 | } | |
4974 | ||
1ade5cc1 | 4975 | dprintf("working=%d (%d) level=%d\n", working, |
a8173e43 | 4976 | a->info.array.raid_disks, |
2c514b71 | 4977 | a->info.array.level); |
7e1432fb NB |
4978 | if (working == a->info.array.raid_disks) |
4979 | return NULL; /* array not degraded */ | |
4980 | switch (a->info.array.level) { | |
4981 | case 1: | |
4982 | if (working == 0) | |
4983 | return NULL; /* failed */ | |
4984 | break; | |
4985 | case 4: | |
4986 | case 5: | |
4987 | if (working < a->info.array.raid_disks - 1) | |
4988 | return NULL; /* failed */ | |
4989 | break; | |
4990 | case 6: | |
4991 | if (working < a->info.array.raid_disks - 2) | |
4992 | return NULL; /* failed */ | |
4993 | break; | |
7733b91d | 4994 | case 10: |
4995 | if (raid10_degraded(&a->info) < 1) | |
4996 | return NULL; | |
4997 | break; | |
7e1432fb NB |
4998 | default: /* concat or stripe */ |
4999 | return NULL; /* failed */ | |
5000 | } | |
5001 | ||
5002 | /* For each slot, if it is not working, find a spare */ | |
5003 | dl = ddf->dlist; | |
5004 | for (i = 0; i < a->info.array.raid_disks; i++) { | |
5005 | for (d = a->info.devs ; d ; d = d->next) | |
5006 | if (d->disk.raid_disk == i) | |
5007 | break; | |
2c514b71 | 5008 | dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0); |
7e1432fb NB |
5009 | if (d && (d->state_fd >= 0)) |
5010 | continue; | |
5011 | ||
5012 | /* OK, this device needs recovery. Find a spare */ | |
5013 | again: | |
5014 | for ( ; dl ; dl = dl->next) { | |
5015 | unsigned long long esize; | |
5016 | unsigned long long pos; | |
5017 | struct mdinfo *d2; | |
5018 | int is_global = 0; | |
5019 | int is_dedicated = 0; | |
a44e993e N |
5020 | be16 state; |
5021 | ||
5022 | if (dl->pdnum < 0) | |
5023 | continue; | |
5024 | state = ddf->phys->entries[dl->pdnum].state; | |
6f56dbb9 MW |
5025 | if (be16_and(state, |
5026 | cpu_to_be16(DDF_Failed|DDF_Missing)) || | |
5027 | !be16_and(state, | |
5028 | cpu_to_be16(DDF_Online))) | |
5029 | continue; | |
5030 | ||
7e1432fb NB |
5031 | /* If in this array, skip */ |
5032 | for (d2 = a->info.devs ; d2 ; d2 = d2->next) | |
7590d562 N |
5033 | if (d2->state_fd >= 0 && |
5034 | d2->disk.major == dl->major && | |
7e1432fb | 5035 | d2->disk.minor == dl->minor) { |
2a645ee2 MW |
5036 | dprintf("%x:%x (%08x) already in array\n", |
5037 | dl->major, dl->minor, | |
5038 | be32_to_cpu(dl->disk.refnum)); | |
7e1432fb NB |
5039 | break; |
5040 | } | |
5041 | if (d2) | |
5042 | continue; | |
a8173e43 | 5043 | if (be16_and(ddf->phys->entries[dl->pdnum].type, |
5044 | cpu_to_be16(DDF_Spare))) { | |
7e1432fb NB |
5045 | /* Check spare assign record */ |
5046 | if (dl->spare) { | |
5047 | if (dl->spare->type & DDF_spare_dedicated) { | |
5048 | /* check spare_ents for guid */ | |
fca65520 | 5049 | unsigned int j; |
7e1432fb | 5050 | for (j = 0 ; |
a8173e43 | 5051 | j < be16_to_cpu |
5052 | (dl->spare | |
5053 | ->populated); | |
7e1432fb NB |
5054 | j++) { |
5055 | if (memcmp(dl->spare->spare_ents[j].guid, | |
5056 | ddf->virt->entries[a->info.container_member].guid, | |
5057 | DDF_GUID_LEN) == 0) | |
5058 | is_dedicated = 1; | |
5059 | } | |
5060 | } else | |
5061 | is_global = 1; | |
5062 | } | |
a8173e43 | 5063 | } else if (be16_and(ddf->phys->entries[dl->pdnum].type, |
5064 | cpu_to_be16(DDF_Global_Spare))) { | |
7e1432fb | 5065 | is_global = 1; |
a8173e43 | 5066 | } else if (!be16_and(ddf->phys |
5067 | ->entries[dl->pdnum].state, | |
5068 | cpu_to_be16(DDF_Failed))) { | |
e0e7aeaa N |
5069 | /* we can possibly use some of this */ |
5070 | is_global = 1; | |
7e1432fb NB |
5071 | } |
5072 | if ( ! (is_dedicated || | |
5073 | (is_global && global_ok))) { | |
2c514b71 | 5074 | dprintf("%x:%x not suitable: %d %d\n", dl->major, dl->minor, |
613b0d17 | 5075 | is_dedicated, is_global); |
7e1432fb NB |
5076 | continue; |
5077 | } | |
5078 | ||
5079 | /* We are allowed to use this device - is there space? | |
5080 | * We need a->info.component_size sectors */ | |
fca65520 N |
5081 | esize = a->info.component_size; |
5082 | pos = find_space(ddf, dl, INVALID_SECTORS, &esize); | |
7e1432fb | 5083 | |
7e1432fb | 5084 | if (esize < a->info.component_size) { |
e5cc7d46 N |
5085 | dprintf("%x:%x has no room: %llu %llu\n", |
5086 | dl->major, dl->minor, | |
2c514b71 | 5087 | esize, a->info.component_size); |
7e1432fb NB |
5088 | /* No room */ |
5089 | continue; | |
5090 | } | |
5091 | ||
5092 | /* Cool, we have a device with some space at pos */ | |
503975b9 | 5093 | di = xcalloc(1, sizeof(*di)); |
7e1432fb NB |
5094 | di->disk.number = i; |
5095 | di->disk.raid_disk = i; | |
5096 | di->disk.major = dl->major; | |
5097 | di->disk.minor = dl->minor; | |
5098 | di->disk.state = 0; | |
d23534e4 | 5099 | di->recovery_start = 0; |
7e1432fb NB |
5100 | di->data_offset = pos; |
5101 | di->component_size = a->info.component_size; | |
7e1432fb NB |
5102 | di->next = rv; |
5103 | rv = di; | |
2a645ee2 MW |
5104 | dprintf("%x:%x (%08x) to be %d at %llu\n", |
5105 | dl->major, dl->minor, | |
5106 | be32_to_cpu(dl->disk.refnum), i, pos); | |
7e1432fb NB |
5107 | |
5108 | break; | |
5109 | } | |
5110 | if (!dl && ! global_ok) { | |
5111 | /* not enough dedicated spares, try global */ | |
5112 | global_ok = 1; | |
5113 | dl = ddf->dlist; | |
5114 | goto again; | |
5115 | } | |
5116 | } | |
5117 | ||
5118 | if (!rv) | |
5119 | /* No spares found */ | |
5120 | return rv; | |
5121 | /* Now 'rv' has a list of devices to return. | |
5122 | * Create a metadata_update record to update the | |
5123 | * phys_refnum and lba_offset values | |
5124 | */ | |
bb925ff0 | 5125 | vc = find_vdcr(ddf, a->info.container_member, rv->disk.raid_disk, |
0c78849f | 5126 | &n_bvd, &vcl); |
5127 | if (vc == NULL) | |
5128 | return NULL; | |
5129 | ||
503975b9 N |
5130 | mu = xmalloc(sizeof(*mu)); |
5131 | if (posix_memalign(&mu->space, 512, sizeof(struct vcl)) != 0) { | |
79244939 DW |
5132 | free(mu); |
5133 | mu = NULL; | |
5134 | } | |
0c78849f | 5135 | |
5136 | mu->len = ddf->conf_rec_len * 512 * vcl->conf.sec_elmnt_count; | |
5137 | mu->buf = xmalloc(mu->len); | |
7590d562 | 5138 | mu->space = NULL; |
f50ae22e | 5139 | mu->space_list = NULL; |
7e1432fb | 5140 | mu->next = *updates; |
0c78849f | 5141 | memcpy(mu->buf, &vcl->conf, ddf->conf_rec_len * 512); |
5142 | for (j = 1; j < vcl->conf.sec_elmnt_count; j++) | |
5143 | memcpy(mu->buf + j * ddf->conf_rec_len * 512, | |
5144 | vcl->other_bvds[j-1], ddf->conf_rec_len * 512); | |
7e1432fb NB |
5145 | |
5146 | vc = (struct vd_config*)mu->buf; | |
7e1432fb | 5147 | for (di = rv ; di ; di = di->next) { |
0c78849f | 5148 | unsigned int i_sec, i_prim; |
5149 | i_sec = di->disk.raid_disk | |
5150 | / be16_to_cpu(vcl->conf.prim_elmnt_count); | |
5151 | i_prim = di->disk.raid_disk | |
5152 | % be16_to_cpu(vcl->conf.prim_elmnt_count); | |
5153 | vc = (struct vd_config *)(mu->buf | |
5154 | + i_sec * ddf->conf_rec_len * 512); | |
5155 | for (dl = ddf->dlist; dl; dl = dl->next) | |
5156 | if (dl->major == di->disk.major | |
5157 | && dl->minor == di->disk.minor) | |
5158 | break; | |
a44e993e | 5159 | if (!dl || dl->pdnum < 0) { |
1ade5cc1 N |
5160 | pr_err("BUG: can't find disk %d (%d/%d)\n", |
5161 | di->disk.raid_disk, | |
0c78849f | 5162 | di->disk.major, di->disk.minor); |
5163 | return NULL; | |
5164 | } | |
5165 | vc->phys_refnum[i_prim] = ddf->phys->entries[dl->pdnum].refnum; | |
5166 | LBA_OFFSET(ddf, vc)[i_prim] = cpu_to_be64(di->data_offset); | |
2a645ee2 MW |
5167 | dprintf("BVD %u gets %u: %08x at %llu\n", i_sec, i_prim, |
5168 | be32_to_cpu(vc->phys_refnum[i_prim]), | |
5169 | be64_to_cpu(LBA_OFFSET(ddf, vc)[i_prim])); | |
7e1432fb NB |
5170 | } |
5171 | *updates = mu; | |
5172 | return rv; | |
5173 | } | |
5174 | ||
b640a252 N |
5175 | static int ddf_level_to_layout(int level) |
5176 | { | |
5177 | switch(level) { | |
5178 | case 0: | |
5179 | case 1: | |
5180 | return 0; | |
5181 | case 5: | |
5182 | return ALGORITHM_LEFT_SYMMETRIC; | |
5183 | case 6: | |
5184 | return ALGORITHM_ROTATING_N_CONTINUE; | |
5185 | case 10: | |
5186 | return 0x102; | |
5187 | default: | |
5188 | return UnSet; | |
5189 | } | |
5190 | } | |
5191 | ||
30f58b22 DW |
5192 | static void default_geometry_ddf(struct supertype *st, int *level, int *layout, int *chunk) |
5193 | { | |
5194 | if (level && *level == UnSet) | |
5195 | *level = LEVEL_CONTAINER; | |
5196 | ||
5197 | if (level && layout && *layout == UnSet) | |
5198 | *layout = ddf_level_to_layout(*level); | |
5199 | } | |
5200 | ||
a322f70c | 5201 | struct superswitch super_ddf = { |
a322f70c DW |
5202 | .examine_super = examine_super_ddf, |
5203 | .brief_examine_super = brief_examine_super_ddf, | |
4737ae25 | 5204 | .brief_examine_subarrays = brief_examine_subarrays_ddf, |
bceedeec | 5205 | .export_examine_super = export_examine_super_ddf, |
a322f70c DW |
5206 | .detail_super = detail_super_ddf, |
5207 | .brief_detail_super = brief_detail_super_ddf, | |
5208 | .validate_geometry = validate_geometry_ddf, | |
78e44928 | 5209 | .write_init_super = write_init_super_ddf, |
0e600426 | 5210 | .add_to_super = add_to_super_ddf, |
4dd968cc | 5211 | .remove_from_super = remove_from_super_ddf, |
2b959fbf | 5212 | .load_container = load_container_ddf, |
74db60b0 | 5213 | .copy_metadata = copy_metadata_ddf, |
4441541f | 5214 | .kill_subarray = kill_subarray_ddf, |
a322f70c DW |
5215 | .match_home = match_home_ddf, |
5216 | .uuid_from_super= uuid_from_super_ddf, | |
5217 | .getinfo_super = getinfo_super_ddf, | |
5218 | .update_super = update_super_ddf, | |
5219 | ||
5220 | .avail_size = avail_size_ddf, | |
5221 | ||
a19c88b8 NB |
5222 | .compare_super = compare_super_ddf, |
5223 | ||
a322f70c | 5224 | .load_super = load_super_ddf, |
ba7eb04f | 5225 | .init_super = init_super_ddf, |
955e9ea1 | 5226 | .store_super = store_super_ddf, |
a322f70c DW |
5227 | .free_super = free_super_ddf, |
5228 | .match_metadata_desc = match_metadata_desc_ddf, | |
78e44928 | 5229 | .container_content = container_content_ddf, |
30f58b22 | 5230 | .default_geometry = default_geometry_ddf, |
a322f70c | 5231 | |
a322f70c | 5232 | .external = 1, |
549e9569 NB |
5233 | |
5234 | /* for mdmon */ | |
5235 | .open_new = ddf_open_new, | |
ed9d66aa | 5236 | .set_array_state= ddf_set_array_state, |
549e9569 NB |
5237 | .set_disk = ddf_set_disk, |
5238 | .sync_metadata = ddf_sync_metadata, | |
88c164f4 | 5239 | .process_update = ddf_process_update, |
edd8d13c | 5240 | .prepare_update = ddf_prepare_update, |
7e1432fb | 5241 | .activate_spare = ddf_activate_spare, |
4cce4069 | 5242 | .name = "ddf", |
a322f70c | 5243 | }; |