]>
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> | |
33 | ||
a322f70c DW |
34 | /* a non-official T10 name for creation GUIDs */ |
35 | static char T10[] = "Linux-MD"; | |
36 | ||
37 | /* DDF timestamps are 1980 based, so we need to add | |
38 | * second-in-decade-of-seventies to convert to linux timestamps. | |
39 | * 10 years with 2 leap years. | |
40 | */ | |
41 | #define DECADE (3600*24*(365*10+2)) | |
42 | unsigned long crc32( | |
43 | unsigned long crc, | |
44 | const unsigned char *buf, | |
45 | unsigned len); | |
46 | ||
bedbf68a | 47 | #define DDF_NOTFOUND (~0U) |
48 | #define DDF_CONTAINER (DDF_NOTFOUND-1) | |
49 | ||
5684fff6 | 50 | /* Default for safe_mode_delay. Same value as for IMSM. |
51 | */ | |
52 | static const int DDF_SAFE_MODE_DELAY = 4000; | |
53 | ||
a322f70c DW |
54 | /* The DDF metadata handling. |
55 | * DDF metadata lives at the end of the device. | |
56 | * The last 512 byte block provides an 'anchor' which is used to locate | |
57 | * the rest of the metadata which usually lives immediately behind the anchor. | |
58 | * | |
59 | * Note: | |
60 | * - all multibyte numeric fields are bigendian. | |
61 | * - all strings are space padded. | |
62 | * | |
63 | */ | |
64 | ||
4d1bdc18 | 65 | typedef struct __be16 { |
66 | __u16 _v16; | |
67 | } be16; | |
68 | #define be16_eq(x, y) ((x)._v16 == (y)._v16) | |
a8173e43 | 69 | #define be16_and(x, y) ((x)._v16 & (y)._v16) |
70 | #define be16_or(x, y) ((x)._v16 | (y)._v16) | |
71 | #define be16_clear(x, y) ((x)._v16 &= ~(y)._v16) | |
72 | #define be16_set(x, y) ((x)._v16 |= (y)._v16) | |
4d1bdc18 | 73 | |
74 | typedef struct __be32 { | |
75 | __u32 _v32; | |
76 | } be32; | |
77 | #define be32_eq(x, y) ((x)._v32 == (y)._v32) | |
78 | ||
79 | typedef struct __be64 { | |
80 | __u64 _v64; | |
81 | } be64; | |
82 | #define be64_eq(x, y) ((x)._v64 == (y)._v64) | |
83 | ||
84 | #define be16_to_cpu(be) __be16_to_cpu((be)._v16) | |
85 | static inline be16 cpu_to_be16(__u16 x) | |
86 | { | |
87 | be16 be = { ._v16 = __cpu_to_be16(x) }; | |
88 | return be; | |
89 | } | |
90 | ||
91 | #define be32_to_cpu(be) __be32_to_cpu((be)._v32) | |
92 | static inline be32 cpu_to_be32(__u32 x) | |
93 | { | |
94 | be32 be = { ._v32 = __cpu_to_be32(x) }; | |
95 | return be; | |
96 | } | |
97 | ||
98 | #define be64_to_cpu(be) __be64_to_cpu((be)._v64) | |
99 | static inline be64 cpu_to_be64(__u64 x) | |
100 | { | |
101 | be64 be = { ._v64 = __cpu_to_be64(x) }; | |
102 | return be; | |
103 | } | |
104 | ||
a322f70c DW |
105 | /* Primary Raid Level (PRL) */ |
106 | #define DDF_RAID0 0x00 | |
107 | #define DDF_RAID1 0x01 | |
108 | #define DDF_RAID3 0x03 | |
109 | #define DDF_RAID4 0x04 | |
110 | #define DDF_RAID5 0x05 | |
111 | #define DDF_RAID1E 0x11 | |
112 | #define DDF_JBOD 0x0f | |
113 | #define DDF_CONCAT 0x1f | |
114 | #define DDF_RAID5E 0x15 | |
115 | #define DDF_RAID5EE 0x25 | |
59e36268 | 116 | #define DDF_RAID6 0x06 |
a322f70c DW |
117 | |
118 | /* Raid Level Qualifier (RLQ) */ | |
119 | #define DDF_RAID0_SIMPLE 0x00 | |
120 | #define DDF_RAID1_SIMPLE 0x00 /* just 2 devices in this plex */ | |
121 | #define DDF_RAID1_MULTI 0x01 /* exactly 3 devices in this plex */ | |
122 | #define DDF_RAID3_0 0x00 /* parity in first extent */ | |
123 | #define DDF_RAID3_N 0x01 /* parity in last extent */ | |
124 | #define DDF_RAID4_0 0x00 /* parity in first extent */ | |
125 | #define DDF_RAID4_N 0x01 /* parity in last extent */ | |
126 | /* these apply to raid5e and raid5ee as well */ | |
127 | #define DDF_RAID5_0_RESTART 0x00 /* same as 'right asymmetric' - layout 1 */ | |
59e36268 | 128 | #define DDF_RAID6_0_RESTART 0x01 /* raid6 different from raid5 here!!! */ |
a322f70c DW |
129 | #define DDF_RAID5_N_RESTART 0x02 /* same as 'left asymmetric' - layout 0 */ |
130 | #define DDF_RAID5_N_CONTINUE 0x03 /* same as 'left symmetric' - layout 2 */ | |
131 | ||
132 | #define DDF_RAID1E_ADJACENT 0x00 /* raid10 nearcopies==2 */ | |
133 | #define DDF_RAID1E_OFFSET 0x01 /* raid10 offsetcopies==2 */ | |
134 | ||
135 | /* Secondary RAID Level (SRL) */ | |
136 | #define DDF_2STRIPED 0x00 /* This is weirder than RAID0 !! */ | |
137 | #define DDF_2MIRRORED 0x01 | |
138 | #define DDF_2CONCAT 0x02 | |
139 | #define DDF_2SPANNED 0x03 /* This is also weird - be careful */ | |
140 | ||
141 | /* Magic numbers */ | |
60931cf9 | 142 | #define DDF_HEADER_MAGIC cpu_to_be32(0xDE11DE11) |
143 | #define DDF_CONTROLLER_MAGIC cpu_to_be32(0xAD111111) | |
144 | #define DDF_PHYS_RECORDS_MAGIC cpu_to_be32(0x22222222) | |
145 | #define DDF_PHYS_DATA_MAGIC cpu_to_be32(0x33333333) | |
146 | #define DDF_VIRT_RECORDS_MAGIC cpu_to_be32(0xDDDDDDDD) | |
147 | #define DDF_VD_CONF_MAGIC cpu_to_be32(0xEEEEEEEE) | |
148 | #define DDF_SPARE_ASSIGN_MAGIC cpu_to_be32(0x55555555) | |
149 | #define DDF_VU_CONF_MAGIC cpu_to_be32(0x88888888) | |
150 | #define DDF_VENDOR_LOG_MAGIC cpu_to_be32(0x01dBEEF0) | |
151 | #define DDF_BBM_LOG_MAGIC cpu_to_be32(0xABADB10C) | |
a322f70c DW |
152 | |
153 | #define DDF_GUID_LEN 24 | |
59e36268 NB |
154 | #define DDF_REVISION_0 "01.00.00" |
155 | #define DDF_REVISION_2 "01.02.00" | |
a322f70c DW |
156 | |
157 | struct ddf_header { | |
60931cf9 | 158 | be32 magic; /* DDF_HEADER_MAGIC */ |
159 | be32 crc; | |
a322f70c | 160 | char guid[DDF_GUID_LEN]; |
59e36268 | 161 | char revision[8]; /* 01.02.00 */ |
60931cf9 | 162 | be32 seq; /* starts at '1' */ |
163 | be32 timestamp; | |
a322f70c DW |
164 | __u8 openflag; |
165 | __u8 foreignflag; | |
166 | __u8 enforcegroups; | |
167 | __u8 pad0; /* 0xff */ | |
168 | __u8 pad1[12]; /* 12 * 0xff */ | |
169 | /* 64 bytes so far */ | |
170 | __u8 header_ext[32]; /* reserved: fill with 0xff */ | |
9d0c6b70 | 171 | be64 primary_lba; |
172 | be64 secondary_lba; | |
a322f70c DW |
173 | __u8 type; |
174 | __u8 pad2[3]; /* 0xff */ | |
60931cf9 | 175 | be32 workspace_len; /* sectors for vendor space - |
a322f70c | 176 | * at least 32768(sectors) */ |
9d0c6b70 | 177 | be64 workspace_lba; |
a8173e43 | 178 | be16 max_pd_entries; /* one of 15, 63, 255, 1023, 4095 */ |
179 | be16 max_vd_entries; /* 2^(4,6,8,10,12)-1 : i.e. as above */ | |
180 | be16 max_partitions; /* i.e. max num of configuration | |
a322f70c | 181 | record entries per disk */ |
a8173e43 | 182 | be16 config_record_len; /* 1 +ROUNDUP(max_primary_element_entries |
a322f70c | 183 | *12/512) */ |
a8173e43 | 184 | be16 max_primary_element_entries; /* 16, 64, 256, 1024, or 4096 */ |
a322f70c DW |
185 | __u8 pad3[54]; /* 0xff */ |
186 | /* 192 bytes so far */ | |
60931cf9 | 187 | be32 controller_section_offset; |
188 | be32 controller_section_length; | |
189 | be32 phys_section_offset; | |
190 | be32 phys_section_length; | |
191 | be32 virt_section_offset; | |
192 | be32 virt_section_length; | |
193 | be32 config_section_offset; | |
194 | be32 config_section_length; | |
195 | be32 data_section_offset; | |
196 | be32 data_section_length; | |
197 | be32 bbm_section_offset; | |
198 | be32 bbm_section_length; | |
199 | be32 diag_space_offset; | |
200 | be32 diag_space_length; | |
201 | be32 vendor_offset; | |
202 | be32 vendor_length; | |
a322f70c DW |
203 | /* 256 bytes so far */ |
204 | __u8 pad4[256]; /* 0xff */ | |
205 | }; | |
206 | ||
207 | /* type field */ | |
208 | #define DDF_HEADER_ANCHOR 0x00 | |
209 | #define DDF_HEADER_PRIMARY 0x01 | |
210 | #define DDF_HEADER_SECONDARY 0x02 | |
211 | ||
212 | /* The content of the 'controller section' - global scope */ | |
213 | struct ddf_controller_data { | |
60931cf9 | 214 | be32 magic; /* DDF_CONTROLLER_MAGIC */ |
215 | be32 crc; | |
a322f70c DW |
216 | char guid[DDF_GUID_LEN]; |
217 | struct controller_type { | |
a8173e43 | 218 | be16 vendor_id; |
219 | be16 device_id; | |
220 | be16 sub_vendor_id; | |
221 | be16 sub_device_id; | |
a322f70c DW |
222 | } type; |
223 | char product_id[16]; | |
224 | __u8 pad[8]; /* 0xff */ | |
225 | __u8 vendor_data[448]; | |
226 | }; | |
227 | ||
228 | /* The content of phys_section - global scope */ | |
229 | struct phys_disk { | |
60931cf9 | 230 | be32 magic; /* DDF_PHYS_RECORDS_MAGIC */ |
231 | be32 crc; | |
217dead4 N |
232 | be16 used_pdes; /* This is a counter, not a max - the list |
233 | * of used entries may not be dense */ | |
a8173e43 | 234 | be16 max_pdes; |
a322f70c DW |
235 | __u8 pad[52]; |
236 | struct phys_disk_entry { | |
237 | char guid[DDF_GUID_LEN]; | |
60931cf9 | 238 | be32 refnum; |
a8173e43 | 239 | be16 type; |
240 | be16 state; | |
56cb05c4 N |
241 | be64 config_size; /* DDF structures must be after here */ |
242 | char path[18]; /* Another horrible structure really | |
243 | * but is "used for information | |
244 | * purposes only" */ | |
a322f70c DW |
245 | __u8 pad[6]; |
246 | } entries[0]; | |
247 | }; | |
248 | ||
249 | /* phys_disk_entry.type is a bitmap - bigendian remember */ | |
250 | #define DDF_Forced_PD_GUID 1 | |
251 | #define DDF_Active_in_VD 2 | |
88c164f4 | 252 | #define DDF_Global_Spare 4 /* VD_CONF records are ignored */ |
a322f70c DW |
253 | #define DDF_Spare 8 /* overrides Global_spare */ |
254 | #define DDF_Foreign 16 | |
255 | #define DDF_Legacy 32 /* no DDF on this device */ | |
256 | ||
257 | #define DDF_Interface_mask 0xf00 | |
258 | #define DDF_Interface_SCSI 0x100 | |
259 | #define DDF_Interface_SAS 0x200 | |
260 | #define DDF_Interface_SATA 0x300 | |
261 | #define DDF_Interface_FC 0x400 | |
262 | ||
263 | /* phys_disk_entry.state is a bigendian bitmap */ | |
264 | #define DDF_Online 1 | |
265 | #define DDF_Failed 2 /* overrides 1,4,8 */ | |
266 | #define DDF_Rebuilding 4 | |
267 | #define DDF_Transition 8 | |
268 | #define DDF_SMART 16 | |
269 | #define DDF_ReadErrors 32 | |
270 | #define DDF_Missing 64 | |
271 | ||
272 | /* The content of the virt_section global scope */ | |
273 | struct virtual_disk { | |
60931cf9 | 274 | be32 magic; /* DDF_VIRT_RECORDS_MAGIC */ |
275 | be32 crc; | |
a8173e43 | 276 | be16 populated_vdes; |
277 | be16 max_vdes; | |
a322f70c DW |
278 | __u8 pad[52]; |
279 | struct virtual_entry { | |
280 | char guid[DDF_GUID_LEN]; | |
a8173e43 | 281 | be16 unit; |
a322f70c | 282 | __u16 pad0; /* 0xffff */ |
a8173e43 | 283 | be16 guid_crc; |
284 | be16 type; | |
a322f70c DW |
285 | __u8 state; |
286 | __u8 init_state; | |
287 | __u8 pad1[14]; | |
288 | char name[16]; | |
289 | } entries[0]; | |
290 | }; | |
291 | ||
292 | /* virtual_entry.type is a bitmap - bigendian */ | |
293 | #define DDF_Shared 1 | |
294 | #define DDF_Enforce_Groups 2 | |
295 | #define DDF_Unicode 4 | |
296 | #define DDF_Owner_Valid 8 | |
297 | ||
298 | /* virtual_entry.state is a bigendian bitmap */ | |
299 | #define DDF_state_mask 0x7 | |
300 | #define DDF_state_optimal 0x0 | |
301 | #define DDF_state_degraded 0x1 | |
302 | #define DDF_state_deleted 0x2 | |
303 | #define DDF_state_missing 0x3 | |
304 | #define DDF_state_failed 0x4 | |
7a7cc504 | 305 | #define DDF_state_part_optimal 0x5 |
a322f70c DW |
306 | |
307 | #define DDF_state_morphing 0x8 | |
308 | #define DDF_state_inconsistent 0x10 | |
309 | ||
310 | /* virtual_entry.init_state is a bigendian bitmap */ | |
311 | #define DDF_initstate_mask 0x03 | |
312 | #define DDF_init_not 0x00 | |
7a7cc504 NB |
313 | #define DDF_init_quick 0x01 /* initialisation is progress. |
314 | * i.e. 'state_inconsistent' */ | |
a322f70c DW |
315 | #define DDF_init_full 0x02 |
316 | ||
317 | #define DDF_access_mask 0xc0 | |
318 | #define DDF_access_rw 0x00 | |
319 | #define DDF_access_ro 0x80 | |
320 | #define DDF_access_blocked 0xc0 | |
321 | ||
322 | /* The content of the config_section - local scope | |
323 | * It has multiple records each config_record_len sectors | |
324 | * They can be vd_config or spare_assign | |
325 | */ | |
326 | ||
327 | struct vd_config { | |
60931cf9 | 328 | be32 magic; /* DDF_VD_CONF_MAGIC */ |
329 | be32 crc; | |
a322f70c | 330 | char guid[DDF_GUID_LEN]; |
60931cf9 | 331 | be32 timestamp; |
332 | be32 seqnum; | |
a322f70c | 333 | __u8 pad0[24]; |
a8173e43 | 334 | be16 prim_elmnt_count; |
a322f70c DW |
335 | __u8 chunk_shift; /* 0 == 512, 1==1024 etc */ |
336 | __u8 prl; | |
337 | __u8 rlq; | |
338 | __u8 sec_elmnt_count; | |
339 | __u8 sec_elmnt_seq; | |
340 | __u8 srl; | |
9d0c6b70 | 341 | be64 blocks; /* blocks per component could be different |
598f0d58 NB |
342 | * on different component devices...(only |
343 | * for concat I hope) */ | |
9d0c6b70 | 344 | be64 array_blocks; /* blocks in array */ |
a322f70c | 345 | __u8 pad1[8]; |
2770f45a N |
346 | be32 spare_refs[8]; /* This is used to detect missing spares. |
347 | * As we don't have an interface for that | |
348 | * the values are ignored. | |
349 | */ | |
a322f70c DW |
350 | __u8 cache_pol[8]; |
351 | __u8 bg_rate; | |
352 | __u8 pad2[3]; | |
353 | __u8 pad3[52]; | |
354 | __u8 pad4[192]; | |
355 | __u8 v0[32]; /* reserved- 0xff */ | |
356 | __u8 v1[32]; /* reserved- 0xff */ | |
357 | __u8 v2[16]; /* reserved- 0xff */ | |
358 | __u8 v3[16]; /* reserved- 0xff */ | |
359 | __u8 vendor[32]; | |
60931cf9 | 360 | be32 phys_refnum[0]; /* refnum of each disk in sequence */ |
a322f70c DW |
361 | /*__u64 lba_offset[0]; LBA offset in each phys. Note extents in a |
362 | bvd are always the same size */ | |
363 | }; | |
9d0c6b70 | 364 | #define LBA_OFFSET(ddf, vd) ((be64 *) &(vd)->phys_refnum[(ddf)->mppe]) |
a322f70c DW |
365 | |
366 | /* vd_config.cache_pol[7] is a bitmap */ | |
367 | #define DDF_cache_writeback 1 /* else writethrough */ | |
368 | #define DDF_cache_wadaptive 2 /* only applies if writeback */ | |
369 | #define DDF_cache_readahead 4 | |
370 | #define DDF_cache_radaptive 8 /* only if doing read-ahead */ | |
371 | #define DDF_cache_ifnobatt 16 /* even to write cache if battery is poor */ | |
372 | #define DDF_cache_wallowed 32 /* enable write caching */ | |
373 | #define DDF_cache_rallowed 64 /* enable read caching */ | |
374 | ||
375 | struct spare_assign { | |
60931cf9 | 376 | be32 magic; /* DDF_SPARE_ASSIGN_MAGIC */ |
377 | be32 crc; | |
378 | be32 timestamp; | |
a322f70c DW |
379 | __u8 reserved[7]; |
380 | __u8 type; | |
a8173e43 | 381 | be16 populated; /* SAEs used */ |
382 | be16 max; /* max SAEs */ | |
a322f70c DW |
383 | __u8 pad[8]; |
384 | struct spare_assign_entry { | |
385 | char guid[DDF_GUID_LEN]; | |
a8173e43 | 386 | be16 secondary_element; |
a322f70c DW |
387 | __u8 pad[6]; |
388 | } spare_ents[0]; | |
389 | }; | |
390 | /* spare_assign.type is a bitmap */ | |
391 | #define DDF_spare_dedicated 0x1 /* else global */ | |
392 | #define DDF_spare_revertible 0x2 /* else committable */ | |
393 | #define DDF_spare_active 0x4 /* else not active */ | |
394 | #define DDF_spare_affinity 0x8 /* enclosure affinity */ | |
395 | ||
396 | /* The data_section contents - local scope */ | |
397 | struct disk_data { | |
60931cf9 | 398 | be32 magic; /* DDF_PHYS_DATA_MAGIC */ |
399 | be32 crc; | |
a322f70c | 400 | char guid[DDF_GUID_LEN]; |
60931cf9 | 401 | be32 refnum; /* crc of some magic drive data ... */ |
a322f70c DW |
402 | __u8 forced_ref; /* set when above was not result of magic */ |
403 | __u8 forced_guid; /* set if guid was forced rather than magic */ | |
404 | __u8 vendor[32]; | |
405 | __u8 pad[442]; | |
406 | }; | |
407 | ||
408 | /* bbm_section content */ | |
409 | struct bad_block_log { | |
60931cf9 | 410 | be32 magic; |
411 | be32 crc; | |
a8173e43 | 412 | be16 entry_count; |
60931cf9 | 413 | be32 spare_count; |
a322f70c | 414 | __u8 pad[10]; |
9d0c6b70 | 415 | be64 first_spare; |
a322f70c | 416 | struct mapped_block { |
9d0c6b70 | 417 | be64 defective_start; |
60931cf9 | 418 | be32 replacement_start; |
a8173e43 | 419 | be16 remap_count; |
a322f70c DW |
420 | __u8 pad[2]; |
421 | } entries[0]; | |
422 | }; | |
423 | ||
424 | /* Struct for internally holding ddf structures */ | |
425 | /* The DDF structure stored on each device is potentially | |
426 | * quite different, as some data is global and some is local. | |
427 | * The global data is: | |
428 | * - ddf header | |
429 | * - controller_data | |
430 | * - Physical disk records | |
431 | * - Virtual disk records | |
432 | * The local data is: | |
433 | * - Configuration records | |
434 | * - Physical Disk data section | |
435 | * ( and Bad block and vendor which I don't care about yet). | |
436 | * | |
437 | * The local data is parsed into separate lists as it is read | |
438 | * and reconstructed for writing. This means that we only need | |
439 | * to make config changes once and they are automatically | |
440 | * propagated to all devices. | |
56cb05c4 N |
441 | * The global (config and disk data) records are each in a list |
442 | * of separate data structures. When writing we find the entry | |
443 | * or entries applicable to the particular device. | |
a322f70c DW |
444 | */ |
445 | struct ddf_super { | |
56cb05c4 | 446 | struct ddf_header anchor, primary, secondary; |
a322f70c | 447 | struct ddf_controller_data controller; |
56cb05c4 | 448 | struct ddf_header *active; |
a322f70c DW |
449 | struct phys_disk *phys; |
450 | struct virtual_disk *virt; | |
3921e41a | 451 | char *conf; |
56cb05c4 N |
452 | int pdsize, vdsize; |
453 | unsigned int max_part, mppe, conf_rec_len; | |
454 | int currentdev; | |
455 | int updates_pending; | |
a322f70c | 456 | struct vcl { |
6416d527 NB |
457 | union { |
458 | char space[512]; | |
459 | struct { | |
460 | struct vcl *next; | |
f21e18ca | 461 | unsigned int vcnum; /* index into ->virt */ |
56cb05c4 N |
462 | /* For an array with a secondary level there are |
463 | * multiple vd_config structures, all with the same | |
464 | * guid but with different sec_elmnt_seq. | |
465 | * One of these structures is in 'conf' below. | |
466 | * The others are in other_bvds, not in any | |
467 | * particular order. | |
468 | */ | |
8ec5d685 | 469 | struct vd_config **other_bvds; |
6416d527 NB |
470 | __u64 *block_sizes; /* NULL if all the same */ |
471 | }; | |
472 | }; | |
a322f70c | 473 | struct vd_config conf; |
d2ca6449 | 474 | } *conflist, *currentconf; |
a322f70c | 475 | struct dl { |
6416d527 NB |
476 | union { |
477 | char space[512]; | |
478 | struct { | |
479 | struct dl *next; | |
480 | int major, minor; | |
481 | char *devname; | |
482 | int fd; | |
483 | unsigned long long size; /* sectors */ | |
9d0c6b70 | 484 | be64 primary_lba; /* sectors */ |
485 | be64 secondary_lba; /* sectors */ | |
486 | be64 workspace_lba; /* sectors */ | |
6416d527 NB |
487 | int pdnum; /* index in ->phys */ |
488 | struct spare_assign *spare; | |
8592f29d N |
489 | void *mdupdate; /* hold metadata update */ |
490 | ||
491 | /* These fields used by auto-layout */ | |
492 | int raiddisk; /* slot to fill in autolayout */ | |
493 | __u64 esize; | |
d2ec75fb | 494 | int displayed; |
6416d527 NB |
495 | }; |
496 | }; | |
a322f70c | 497 | struct disk_data disk; |
b2280677 | 498 | struct vcl *vlist[0]; /* max_part in size */ |
2cc2983d | 499 | } *dlist, *add_list; |
a322f70c DW |
500 | }; |
501 | ||
56cb05c4 N |
502 | #ifndef MDASSEMBLE |
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); | |
520 | #endif | |
521 | ||
522 | static void free_super_ddf(struct supertype *st); | |
523 | static int all_ff(const char *guid); | |
524 | static unsigned int get_pd_index_from_refnum(const struct vcl *vc, | |
525 | be32 refnum, unsigned int nmax, | |
526 | const struct vd_config **bvd, | |
527 | unsigned int *idx); | |
528 | static void getinfo_super_ddf(struct supertype *st, struct mdinfo *info, char *map); | |
529 | static void uuid_from_ddf_guid(const char *guid, int uuid[4]); | |
530 | static void uuid_from_super_ddf(struct supertype *st, int uuid[4]); | |
531 | static void _ddf_array_name(char *name, const struct ddf_super *ddf, int i); | |
532 | static void getinfo_super_ddf_bvd(struct supertype *st, struct mdinfo *info, char *map); | |
533 | static int init_super_ddf_bvd(struct supertype *st, | |
534 | mdu_array_info_t *info, | |
535 | unsigned long long size, | |
536 | char *name, char *homehost, | |
537 | int *uuid, unsigned long long data_offset); | |
538 | ||
a322f70c DW |
539 | #ifndef offsetof |
540 | #define offsetof(t,f) ((size_t)&(((t*)0)->f)) | |
541 | #endif | |
542 | ||
7d5a7ff3 | 543 | #if DEBUG |
7d5a7ff3 | 544 | static void pr_state(struct ddf_super *ddf, const char *msg) |
545 | { | |
546 | unsigned int i; | |
547 | dprintf("%s/%s: ", __func__, msg); | |
a8173e43 | 548 | for (i = 0; i < be16_to_cpu(ddf->active->max_vd_entries); i++) { |
7d5a7ff3 | 549 | if (all_ff(ddf->virt->entries[i].guid)) |
550 | continue; | |
551 | dprintf("%u(s=%02x i=%02x) ", i, | |
552 | ddf->virt->entries[i].state, | |
553 | ddf->virt->entries[i].init_state); | |
554 | } | |
555 | dprintf("\n"); | |
556 | } | |
557 | #else | |
558 | static void pr_state(const struct ddf_super *ddf, const char *msg) {} | |
559 | #endif | |
560 | ||
609ce161 N |
561 | static void _ddf_set_updates_pending(struct ddf_super *ddf, struct vd_config *vc, |
562 | const char *func) | |
35c3606d | 563 | { |
609ce161 N |
564 | if (vc) { |
565 | vc->timestamp = cpu_to_be32(time(0)-DECADE); | |
566 | vc->seqnum = cpu_to_be32(be32_to_cpu(vc->seqnum) + 1); | |
567 | } | |
5a46fcd7 N |
568 | if (ddf->updates_pending) |
569 | return; | |
35c3606d | 570 | ddf->updates_pending = 1; |
60931cf9 | 571 | ddf->active->seq = cpu_to_be32((be32_to_cpu(ddf->active->seq)+1)); |
35c3606d | 572 | pr_state(ddf, func); |
573 | } | |
574 | ||
609ce161 | 575 | #define ddf_set_updates_pending(x,v) _ddf_set_updates_pending((x), (v), __func__) |
7d5a7ff3 | 576 | |
60931cf9 | 577 | static be32 calc_crc(void *buf, int len) |
a322f70c DW |
578 | { |
579 | /* crcs are always at the same place as in the ddf_header */ | |
580 | struct ddf_header *ddf = buf; | |
60931cf9 | 581 | be32 oldcrc = ddf->crc; |
a322f70c | 582 | __u32 newcrc; |
60931cf9 | 583 | ddf->crc = cpu_to_be32(0xffffffff); |
a322f70c DW |
584 | |
585 | newcrc = crc32(0, buf, len); | |
586 | ddf->crc = oldcrc; | |
56cb05c4 | 587 | /* The crc is stored (like everything) bigendian, so convert |
4abe6b70 N |
588 | * here for simplicity |
589 | */ | |
60931cf9 | 590 | return cpu_to_be32(newcrc); |
a322f70c DW |
591 | } |
592 | ||
a3163bf0 | 593 | #define DDF_INVALID_LEVEL 0xff |
594 | #define DDF_NO_SECONDARY 0xff | |
595 | static int err_bad_md_layout(const mdu_array_info_t *array) | |
596 | { | |
597 | pr_err("RAID%d layout %x with %d disks is unsupported for DDF\n", | |
598 | array->level, array->layout, array->raid_disks); | |
2aba583f | 599 | return -1; |
a3163bf0 | 600 | } |
601 | ||
602 | static int layout_md2ddf(const mdu_array_info_t *array, | |
603 | struct vd_config *conf) | |
604 | { | |
a8173e43 | 605 | be16 prim_elmnt_count = cpu_to_be16(array->raid_disks); |
a3163bf0 | 606 | __u8 prl = DDF_INVALID_LEVEL, rlq = 0; |
607 | __u8 sec_elmnt_count = 1; | |
608 | __u8 srl = DDF_NO_SECONDARY; | |
609 | ||
610 | switch (array->level) { | |
611 | case LEVEL_LINEAR: | |
612 | prl = DDF_CONCAT; | |
613 | break; | |
614 | case 0: | |
615 | rlq = DDF_RAID0_SIMPLE; | |
616 | prl = DDF_RAID0; | |
617 | break; | |
618 | case 1: | |
619 | switch (array->raid_disks) { | |
620 | case 2: | |
621 | rlq = DDF_RAID1_SIMPLE; | |
622 | break; | |
623 | case 3: | |
624 | rlq = DDF_RAID1_MULTI; | |
625 | break; | |
626 | default: | |
627 | return err_bad_md_layout(array); | |
628 | } | |
629 | prl = DDF_RAID1; | |
630 | break; | |
631 | case 4: | |
632 | if (array->layout != 0) | |
633 | return err_bad_md_layout(array); | |
634 | rlq = DDF_RAID4_N; | |
635 | prl = DDF_RAID4; | |
636 | break; | |
637 | case 5: | |
638 | switch (array->layout) { | |
639 | case ALGORITHM_LEFT_ASYMMETRIC: | |
640 | rlq = DDF_RAID5_N_RESTART; | |
641 | break; | |
642 | case ALGORITHM_RIGHT_ASYMMETRIC: | |
643 | rlq = DDF_RAID5_0_RESTART; | |
644 | break; | |
645 | case ALGORITHM_LEFT_SYMMETRIC: | |
646 | rlq = DDF_RAID5_N_CONTINUE; | |
647 | break; | |
648 | case ALGORITHM_RIGHT_SYMMETRIC: | |
649 | /* not mentioned in standard */ | |
650 | default: | |
651 | return err_bad_md_layout(array); | |
652 | } | |
653 | prl = DDF_RAID5; | |
654 | break; | |
655 | case 6: | |
656 | switch (array->layout) { | |
657 | case ALGORITHM_ROTATING_N_RESTART: | |
658 | rlq = DDF_RAID5_N_RESTART; | |
659 | break; | |
660 | case ALGORITHM_ROTATING_ZERO_RESTART: | |
661 | rlq = DDF_RAID6_0_RESTART; | |
662 | break; | |
663 | case ALGORITHM_ROTATING_N_CONTINUE: | |
664 | rlq = DDF_RAID5_N_CONTINUE; | |
665 | break; | |
666 | default: | |
667 | return err_bad_md_layout(array); | |
668 | } | |
669 | prl = DDF_RAID6; | |
670 | break; | |
671 | case 10: | |
672 | if (array->raid_disks % 2 == 0 && array->layout == 0x102) { | |
673 | rlq = DDF_RAID1_SIMPLE; | |
a8173e43 | 674 | prim_elmnt_count = cpu_to_be16(2); |
a3163bf0 | 675 | sec_elmnt_count = array->raid_disks / 2; |
f70d549f N |
676 | srl = DDF_2SPANNED; |
677 | prl = DDF_RAID1; | |
a3163bf0 | 678 | } else if (array->raid_disks % 3 == 0 |
679 | && array->layout == 0x103) { | |
680 | rlq = DDF_RAID1_MULTI; | |
a8173e43 | 681 | prim_elmnt_count = cpu_to_be16(3); |
a3163bf0 | 682 | sec_elmnt_count = array->raid_disks / 3; |
f70d549f N |
683 | srl = DDF_2SPANNED; |
684 | prl = DDF_RAID1; | |
685 | } else if (array->layout == 0x201) { | |
686 | prl = DDF_RAID1E; | |
687 | rlq = DDF_RAID1E_OFFSET; | |
688 | } else if (array->layout == 0x102) { | |
689 | prl = DDF_RAID1E; | |
690 | rlq = DDF_RAID1E_ADJACENT; | |
a3163bf0 | 691 | } else |
692 | return err_bad_md_layout(array); | |
a3163bf0 | 693 | break; |
694 | default: | |
695 | return err_bad_md_layout(array); | |
696 | } | |
697 | conf->prl = prl; | |
698 | conf->prim_elmnt_count = prim_elmnt_count; | |
699 | conf->rlq = rlq; | |
700 | conf->srl = srl; | |
701 | conf->sec_elmnt_count = sec_elmnt_count; | |
702 | return 0; | |
703 | } | |
704 | ||
8a2848a7 | 705 | static int err_bad_ddf_layout(const struct vd_config *conf) |
706 | { | |
707 | pr_err("DDF RAID %u qualifier %u with %u disks is unsupported\n", | |
a8173e43 | 708 | conf->prl, conf->rlq, be16_to_cpu(conf->prim_elmnt_count)); |
8a2848a7 | 709 | return -1; |
710 | } | |
711 | ||
712 | static int layout_ddf2md(const struct vd_config *conf, | |
713 | mdu_array_info_t *array) | |
714 | { | |
715 | int level = LEVEL_UNSUPPORTED; | |
716 | int layout = 0; | |
a8173e43 | 717 | int raiddisks = be16_to_cpu(conf->prim_elmnt_count); |
8a2848a7 | 718 | |
719 | if (conf->sec_elmnt_count > 1) { | |
720 | /* see also check_secondary() */ | |
721 | if (conf->prl != DDF_RAID1 || | |
722 | (conf->srl != DDF_2STRIPED && conf->srl != DDF_2SPANNED)) { | |
723 | pr_err("Unsupported secondary RAID level %u/%u\n", | |
724 | conf->prl, conf->srl); | |
725 | return -1; | |
726 | } | |
727 | if (raiddisks == 2 && conf->rlq == DDF_RAID1_SIMPLE) | |
728 | layout = 0x102; | |
729 | else if (raiddisks == 3 && conf->rlq == DDF_RAID1_MULTI) | |
730 | layout = 0x103; | |
731 | else | |
732 | return err_bad_ddf_layout(conf); | |
733 | raiddisks *= conf->sec_elmnt_count; | |
734 | level = 10; | |
735 | goto good; | |
736 | } | |
737 | ||
738 | switch (conf->prl) { | |
739 | case DDF_CONCAT: | |
740 | level = LEVEL_LINEAR; | |
741 | break; | |
742 | case DDF_RAID0: | |
743 | if (conf->rlq != DDF_RAID0_SIMPLE) | |
744 | return err_bad_ddf_layout(conf); | |
745 | level = 0; | |
746 | break; | |
747 | case DDF_RAID1: | |
748 | if (!((conf->rlq == DDF_RAID1_SIMPLE && raiddisks == 2) || | |
749 | (conf->rlq == DDF_RAID1_MULTI && raiddisks == 3))) | |
750 | return err_bad_ddf_layout(conf); | |
751 | level = 1; | |
752 | break; | |
f70d549f N |
753 | case DDF_RAID1E: |
754 | if (conf->rlq == DDF_RAID1E_ADJACENT) | |
755 | layout = 0x102; | |
756 | else if (conf->rlq == DDF_RAID1E_OFFSET) | |
757 | layout = 0x201; | |
758 | else | |
759 | return err_bad_ddf_layout(conf); | |
760 | level = 10; | |
761 | break; | |
8a2848a7 | 762 | case DDF_RAID4: |
763 | if (conf->rlq != DDF_RAID4_N) | |
764 | return err_bad_ddf_layout(conf); | |
765 | level = 4; | |
766 | break; | |
767 | case DDF_RAID5: | |
768 | switch (conf->rlq) { | |
769 | case DDF_RAID5_N_RESTART: | |
770 | layout = ALGORITHM_LEFT_ASYMMETRIC; | |
771 | break; | |
772 | case DDF_RAID5_0_RESTART: | |
773 | layout = ALGORITHM_RIGHT_ASYMMETRIC; | |
774 | break; | |
775 | case DDF_RAID5_N_CONTINUE: | |
776 | layout = ALGORITHM_LEFT_SYMMETRIC; | |
777 | break; | |
778 | default: | |
779 | return err_bad_ddf_layout(conf); | |
780 | } | |
781 | level = 5; | |
782 | break; | |
783 | case DDF_RAID6: | |
784 | switch (conf->rlq) { | |
785 | case DDF_RAID5_N_RESTART: | |
786 | layout = ALGORITHM_ROTATING_N_RESTART; | |
787 | break; | |
788 | case DDF_RAID6_0_RESTART: | |
789 | layout = ALGORITHM_ROTATING_ZERO_RESTART; | |
790 | break; | |
791 | case DDF_RAID5_N_CONTINUE: | |
792 | layout = ALGORITHM_ROTATING_N_CONTINUE; | |
793 | break; | |
794 | default: | |
795 | return err_bad_ddf_layout(conf); | |
796 | } | |
797 | level = 6; | |
798 | break; | |
799 | default: | |
800 | return err_bad_ddf_layout(conf); | |
801 | }; | |
802 | ||
803 | good: | |
804 | array->level = level; | |
805 | array->layout = layout; | |
806 | array->raid_disks = raiddisks; | |
807 | return 0; | |
808 | } | |
809 | ||
a322f70c DW |
810 | static int load_ddf_header(int fd, unsigned long long lba, |
811 | unsigned long long size, | |
812 | int type, | |
813 | struct ddf_header *hdr, struct ddf_header *anchor) | |
814 | { | |
815 | /* read a ddf header (primary or secondary) from fd/lba | |
816 | * and check that it is consistent with anchor | |
817 | * Need to check: | |
818 | * magic, crc, guid, rev, and LBA's header_type, and | |
819 | * everything after header_type must be the same | |
820 | */ | |
821 | if (lba >= size-1) | |
822 | return 0; | |
823 | ||
824 | if (lseek64(fd, lba<<9, 0) < 0) | |
825 | return 0; | |
826 | ||
827 | if (read(fd, hdr, 512) != 512) | |
828 | return 0; | |
829 | ||
0e5fa862 MW |
830 | if (!be32_eq(hdr->magic, DDF_HEADER_MAGIC)) { |
831 | pr_err("%s: bad header magic\n", __func__); | |
a322f70c | 832 | return 0; |
0e5fa862 MW |
833 | } |
834 | if (!be32_eq(calc_crc(hdr, 512), hdr->crc)) { | |
835 | pr_err("%s: bad CRC\n", __func__); | |
a322f70c | 836 | return 0; |
0e5fa862 | 837 | } |
a322f70c DW |
838 | if (memcmp(anchor->guid, hdr->guid, DDF_GUID_LEN) != 0 || |
839 | memcmp(anchor->revision, hdr->revision, 8) != 0 || | |
9d0c6b70 | 840 | !be64_eq(anchor->primary_lba, hdr->primary_lba) || |
841 | !be64_eq(anchor->secondary_lba, hdr->secondary_lba) || | |
a322f70c DW |
842 | hdr->type != type || |
843 | memcmp(anchor->pad2, hdr->pad2, 512 - | |
0e5fa862 MW |
844 | offsetof(struct ddf_header, pad2)) != 0) { |
845 | pr_err("%s: header mismatch\n", __func__); | |
a322f70c | 846 | return 0; |
0e5fa862 | 847 | } |
a322f70c DW |
848 | |
849 | /* Looks good enough to me... */ | |
850 | return 1; | |
851 | } | |
852 | ||
853 | static void *load_section(int fd, struct ddf_super *super, void *buf, | |
60931cf9 | 854 | be32 offset_be, be32 len_be, int check) |
a322f70c | 855 | { |
60931cf9 | 856 | unsigned long long offset = be32_to_cpu(offset_be); |
857 | unsigned long long len = be32_to_cpu(len_be); | |
a322f70c DW |
858 | int dofree = (buf == NULL); |
859 | ||
860 | if (check) | |
861 | if (len != 2 && len != 8 && len != 32 | |
862 | && len != 128 && len != 512) | |
863 | return NULL; | |
864 | ||
865 | if (len > 1024) | |
866 | return NULL; | |
3921e41a | 867 | if (!buf && posix_memalign(&buf, 512, len<<9) != 0) |
3d2c4fc7 | 868 | buf = NULL; |
6416d527 | 869 | |
a322f70c DW |
870 | if (!buf) |
871 | return NULL; | |
872 | ||
873 | if (super->active->type == 1) | |
9d0c6b70 | 874 | offset += be64_to_cpu(super->active->primary_lba); |
a322f70c | 875 | else |
9d0c6b70 | 876 | offset += be64_to_cpu(super->active->secondary_lba); |
a322f70c | 877 | |
f21e18ca | 878 | if ((unsigned long long)lseek64(fd, offset<<9, 0) != (offset<<9)) { |
a322f70c DW |
879 | if (dofree) |
880 | free(buf); | |
881 | return NULL; | |
882 | } | |
f21e18ca | 883 | if ((unsigned long long)read(fd, buf, len<<9) != (len<<9)) { |
a322f70c DW |
884 | if (dofree) |
885 | free(buf); | |
886 | return NULL; | |
887 | } | |
888 | return buf; | |
889 | } | |
890 | ||
891 | static int load_ddf_headers(int fd, struct ddf_super *super, char *devname) | |
892 | { | |
893 | unsigned long long dsize; | |
894 | ||
895 | get_dev_size(fd, NULL, &dsize); | |
896 | ||
897 | if (lseek64(fd, dsize-512, 0) < 0) { | |
898 | if (devname) | |
e7b84f9d N |
899 | pr_err("Cannot seek to anchor block on %s: %s\n", |
900 | devname, strerror(errno)); | |
a322f70c DW |
901 | return 1; |
902 | } | |
903 | if (read(fd, &super->anchor, 512) != 512) { | |
904 | if (devname) | |
e7b84f9d N |
905 | pr_err("Cannot read anchor block on %s: %s\n", |
906 | devname, strerror(errno)); | |
a322f70c DW |
907 | return 1; |
908 | } | |
60931cf9 | 909 | if (!be32_eq(super->anchor.magic, DDF_HEADER_MAGIC)) { |
a322f70c | 910 | if (devname) |
e7b84f9d | 911 | pr_err("no DDF anchor found on %s\n", |
a322f70c DW |
912 | devname); |
913 | return 2; | |
914 | } | |
60931cf9 | 915 | if (!be32_eq(calc_crc(&super->anchor, 512), super->anchor.crc)) { |
a322f70c | 916 | if (devname) |
e7b84f9d | 917 | pr_err("bad CRC on anchor on %s\n", |
a322f70c DW |
918 | devname); |
919 | return 2; | |
920 | } | |
59e36268 NB |
921 | if (memcmp(super->anchor.revision, DDF_REVISION_0, 8) != 0 && |
922 | memcmp(super->anchor.revision, DDF_REVISION_2, 8) != 0) { | |
a322f70c | 923 | if (devname) |
e7b84f9d | 924 | pr_err("can only support super revision" |
59e36268 NB |
925 | " %.8s and earlier, not %.8s on %s\n", |
926 | DDF_REVISION_2, super->anchor.revision,devname); | |
a322f70c DW |
927 | return 2; |
928 | } | |
dbeb699a | 929 | super->active = NULL; |
9d0c6b70 | 930 | if (load_ddf_header(fd, be64_to_cpu(super->anchor.primary_lba), |
a322f70c DW |
931 | dsize >> 9, 1, |
932 | &super->primary, &super->anchor) == 0) { | |
933 | if (devname) | |
e7b84f9d N |
934 | pr_err("Failed to load primary DDF header " |
935 | "on %s\n", devname); | |
dbeb699a | 936 | } else |
937 | super->active = &super->primary; | |
60931cf9 | 938 | |
9d0c6b70 | 939 | if (load_ddf_header(fd, be64_to_cpu(super->anchor.secondary_lba), |
a322f70c DW |
940 | dsize >> 9, 2, |
941 | &super->secondary, &super->anchor)) { | |
3eff7c1d | 942 | if (super->active == NULL |
60931cf9 | 943 | || (be32_to_cpu(super->primary.seq) |
944 | < be32_to_cpu(super->secondary.seq) && | |
3eff7c1d | 945 | !super->secondary.openflag) |
60931cf9 | 946 | || (be32_to_cpu(super->primary.seq) |
947 | == be32_to_cpu(super->secondary.seq) && | |
a322f70c DW |
948 | super->primary.openflag && !super->secondary.openflag) |
949 | ) | |
950 | super->active = &super->secondary; | |
b95cb4b9 N |
951 | } else if (devname && |
952 | be64_to_cpu(super->anchor.secondary_lba) != ~(__u64)0) | |
dbeb699a | 953 | pr_err("Failed to load secondary DDF header on %s\n", |
954 | devname); | |
955 | if (super->active == NULL) | |
956 | return 2; | |
a322f70c DW |
957 | return 0; |
958 | } | |
959 | ||
960 | static int load_ddf_global(int fd, struct ddf_super *super, char *devname) | |
961 | { | |
962 | void *ok; | |
963 | ok = load_section(fd, super, &super->controller, | |
964 | super->active->controller_section_offset, | |
965 | super->active->controller_section_length, | |
966 | 0); | |
967 | super->phys = load_section(fd, super, NULL, | |
968 | super->active->phys_section_offset, | |
969 | super->active->phys_section_length, | |
970 | 1); | |
60931cf9 | 971 | super->pdsize = be32_to_cpu(super->active->phys_section_length) * 512; |
a322f70c DW |
972 | |
973 | super->virt = load_section(fd, super, NULL, | |
974 | super->active->virt_section_offset, | |
975 | super->active->virt_section_length, | |
976 | 1); | |
60931cf9 | 977 | super->vdsize = be32_to_cpu(super->active->virt_section_length) * 512; |
a322f70c DW |
978 | if (!ok || |
979 | !super->phys || | |
980 | !super->virt) { | |
981 | free(super->phys); | |
982 | free(super->virt); | |
a2349791 NB |
983 | super->phys = NULL; |
984 | super->virt = NULL; | |
a322f70c DW |
985 | return 2; |
986 | } | |
987 | super->conflist = NULL; | |
988 | super->dlist = NULL; | |
8c3b8c2c | 989 | |
a8173e43 | 990 | super->max_part = be16_to_cpu(super->active->max_partitions); |
991 | super->mppe = be16_to_cpu(super->active->max_primary_element_entries); | |
992 | super->conf_rec_len = be16_to_cpu(super->active->config_record_len); | |
a322f70c DW |
993 | return 0; |
994 | } | |
995 | ||
3c48f7be | 996 | #define DDF_UNUSED_BVD 0xff |
997 | static int alloc_other_bvds(const struct ddf_super *ddf, struct vcl *vcl) | |
998 | { | |
999 | unsigned int n_vds = vcl->conf.sec_elmnt_count - 1; | |
1000 | unsigned int i, vdsize; | |
1001 | void *p; | |
1002 | if (n_vds == 0) { | |
1003 | vcl->other_bvds = NULL; | |
1004 | return 0; | |
1005 | } | |
1006 | vdsize = ddf->conf_rec_len * 512; | |
1007 | if (posix_memalign(&p, 512, n_vds * | |
1008 | (vdsize + sizeof(struct vd_config *))) != 0) | |
1009 | return -1; | |
1010 | vcl->other_bvds = (struct vd_config **) (p + n_vds * vdsize); | |
1011 | for (i = 0; i < n_vds; i++) { | |
1012 | vcl->other_bvds[i] = p + i * vdsize; | |
1013 | memset(vcl->other_bvds[i], 0, vdsize); | |
1014 | vcl->other_bvds[i]->sec_elmnt_seq = DDF_UNUSED_BVD; | |
1015 | } | |
1016 | return 0; | |
1017 | } | |
1018 | ||
3dc821b0 | 1019 | static void add_other_bvd(struct vcl *vcl, struct vd_config *vd, |
1020 | unsigned int len) | |
1021 | { | |
1022 | int i; | |
1023 | for (i = 0; i < vcl->conf.sec_elmnt_count-1; i++) | |
3c48f7be | 1024 | if (vcl->other_bvds[i]->sec_elmnt_seq == vd->sec_elmnt_seq) |
3dc821b0 | 1025 | break; |
1026 | ||
1027 | if (i < vcl->conf.sec_elmnt_count-1) { | |
60931cf9 | 1028 | if (be32_to_cpu(vd->seqnum) <= |
1029 | be32_to_cpu(vcl->other_bvds[i]->seqnum)) | |
3dc821b0 | 1030 | return; |
1031 | } else { | |
1032 | for (i = 0; i < vcl->conf.sec_elmnt_count-1; i++) | |
3c48f7be | 1033 | if (vcl->other_bvds[i]->sec_elmnt_seq == DDF_UNUSED_BVD) |
3dc821b0 | 1034 | break; |
1035 | if (i == vcl->conf.sec_elmnt_count-1) { | |
1036 | pr_err("no space for sec level config %u, count is %u\n", | |
1037 | vd->sec_elmnt_seq, vcl->conf.sec_elmnt_count); | |
1038 | return; | |
1039 | } | |
3dc821b0 | 1040 | } |
1041 | memcpy(vcl->other_bvds[i], vd, len); | |
1042 | } | |
1043 | ||
a322f70c DW |
1044 | static int load_ddf_local(int fd, struct ddf_super *super, |
1045 | char *devname, int keep) | |
1046 | { | |
1047 | struct dl *dl; | |
1048 | struct stat stb; | |
1049 | char *conf; | |
f21e18ca N |
1050 | unsigned int i; |
1051 | unsigned int confsec; | |
b2280677 | 1052 | int vnum; |
56cb05c4 N |
1053 | unsigned int max_virt_disks = |
1054 | be16_to_cpu(super->active->max_vd_entries); | |
d2ca6449 | 1055 | unsigned long long dsize; |
a322f70c DW |
1056 | |
1057 | /* First the local disk info */ | |
3d2c4fc7 | 1058 | if (posix_memalign((void**)&dl, 512, |
56cb05c4 N |
1059 | sizeof(*dl) + |
1060 | (super->max_part) * sizeof(dl->vlist[0])) != 0) { | |
e7b84f9d | 1061 | pr_err("%s could not allocate disk info buffer\n", |
56cb05c4 | 1062 | __func__); |
3d2c4fc7 DW |
1063 | return 1; |
1064 | } | |
a322f70c DW |
1065 | |
1066 | load_section(fd, super, &dl->disk, | |
1067 | super->active->data_section_offset, | |
1068 | super->active->data_section_length, | |
1069 | 0); | |
503975b9 | 1070 | dl->devname = devname ? xstrdup(devname) : NULL; |
598f0d58 | 1071 | |
a322f70c DW |
1072 | fstat(fd, &stb); |
1073 | dl->major = major(stb.st_rdev); | |
1074 | dl->minor = minor(stb.st_rdev); | |
1075 | dl->next = super->dlist; | |
1076 | dl->fd = keep ? fd : -1; | |
d2ca6449 NB |
1077 | |
1078 | dl->size = 0; | |
1079 | if (get_dev_size(fd, devname, &dsize)) | |
1080 | dl->size = dsize >> 9; | |
097bcf00 | 1081 | /* If the disks have different sizes, the LBAs will differ |
1082 | * between phys disks. | |
1083 | * At this point here, the values in super->active must be valid | |
1084 | * for this phys disk. */ | |
1085 | dl->primary_lba = super->active->primary_lba; | |
1086 | dl->secondary_lba = super->active->secondary_lba; | |
1087 | dl->workspace_lba = super->active->workspace_lba; | |
b2280677 | 1088 | dl->spare = NULL; |
f21e18ca | 1089 | for (i = 0 ; i < super->max_part ; i++) |
a322f70c DW |
1090 | dl->vlist[i] = NULL; |
1091 | super->dlist = dl; | |
59e36268 | 1092 | dl->pdnum = -1; |
a8173e43 | 1093 | for (i = 0; i < be16_to_cpu(super->active->max_pd_entries); i++) |
5575e7d9 NB |
1094 | if (memcmp(super->phys->entries[i].guid, |
1095 | dl->disk.guid, DDF_GUID_LEN) == 0) | |
1096 | dl->pdnum = i; | |
1097 | ||
a322f70c DW |
1098 | /* Now the config list. */ |
1099 | /* 'conf' is an array of config entries, some of which are | |
1100 | * probably invalid. Those which are good need to be copied into | |
1101 | * the conflist | |
1102 | */ | |
a322f70c | 1103 | |
3921e41a | 1104 | conf = load_section(fd, super, super->conf, |
a322f70c DW |
1105 | super->active->config_section_offset, |
1106 | super->active->config_section_length, | |
1107 | 0); | |
3921e41a | 1108 | super->conf = conf; |
b2280677 | 1109 | vnum = 0; |
e223334f | 1110 | for (confsec = 0; |
60931cf9 | 1111 | confsec < be32_to_cpu(super->active->config_section_length); |
e223334f | 1112 | confsec += super->conf_rec_len) { |
a322f70c | 1113 | struct vd_config *vd = |
e223334f | 1114 | (struct vd_config *)((char*)conf + confsec*512); |
a322f70c DW |
1115 | struct vcl *vcl; |
1116 | ||
60931cf9 | 1117 | if (be32_eq(vd->magic, DDF_SPARE_ASSIGN_MAGIC)) { |
b2280677 NB |
1118 | if (dl->spare) |
1119 | continue; | |
3d2c4fc7 | 1120 | if (posix_memalign((void**)&dl->spare, 512, |
56cb05c4 | 1121 | super->conf_rec_len*512) != 0) { |
e7b84f9d N |
1122 | pr_err("%s could not allocate spare info buf\n", |
1123 | __func__); | |
3d2c4fc7 DW |
1124 | return 1; |
1125 | } | |
613b0d17 | 1126 | |
b2280677 NB |
1127 | memcpy(dl->spare, vd, super->conf_rec_len*512); |
1128 | continue; | |
1129 | } | |
60931cf9 | 1130 | if (!be32_eq(vd->magic, DDF_VD_CONF_MAGIC)) |
56cb05c4 | 1131 | /* Must be vendor-unique - I cannot handle those */ |
a322f70c | 1132 | continue; |
56cb05c4 | 1133 | |
a322f70c DW |
1134 | for (vcl = super->conflist; vcl; vcl = vcl->next) { |
1135 | if (memcmp(vcl->conf.guid, | |
1136 | vd->guid, DDF_GUID_LEN) == 0) | |
1137 | break; | |
1138 | } | |
1139 | ||
1140 | if (vcl) { | |
b2280677 | 1141 | dl->vlist[vnum++] = vcl; |
3dc821b0 | 1142 | if (vcl->other_bvds != NULL && |
1143 | vcl->conf.sec_elmnt_seq != vd->sec_elmnt_seq) { | |
1144 | add_other_bvd(vcl, vd, super->conf_rec_len*512); | |
1145 | continue; | |
1146 | } | |
60931cf9 | 1147 | if (be32_to_cpu(vd->seqnum) <= |
1148 | be32_to_cpu(vcl->conf.seqnum)) | |
a322f70c | 1149 | continue; |
59e36268 | 1150 | } else { |
3d2c4fc7 | 1151 | if (posix_memalign((void**)&vcl, 512, |
56cb05c4 N |
1152 | (super->conf_rec_len*512 + |
1153 | offsetof(struct vcl, conf))) != 0) { | |
e7b84f9d N |
1154 | pr_err("%s could not allocate vcl buf\n", |
1155 | __func__); | |
3d2c4fc7 DW |
1156 | return 1; |
1157 | } | |
a322f70c | 1158 | vcl->next = super->conflist; |
59e36268 | 1159 | vcl->block_sizes = NULL; /* FIXME not for CONCAT */ |
3c48f7be | 1160 | vcl->conf.sec_elmnt_count = vd->sec_elmnt_count; |
1161 | if (alloc_other_bvds(super, vcl) != 0) { | |
1162 | pr_err("%s could not allocate other bvds\n", | |
1163 | __func__); | |
1164 | free(vcl); | |
1165 | return 1; | |
1166 | }; | |
a322f70c | 1167 | super->conflist = vcl; |
b2280677 | 1168 | dl->vlist[vnum++] = vcl; |
a322f70c | 1169 | } |
8c3b8c2c | 1170 | memcpy(&vcl->conf, vd, super->conf_rec_len*512); |
59e36268 NB |
1171 | for (i=0; i < max_virt_disks ; i++) |
1172 | if (memcmp(super->virt->entries[i].guid, | |
1173 | vcl->conf.guid, DDF_GUID_LEN)==0) | |
1174 | break; | |
1175 | if (i < max_virt_disks) | |
1176 | vcl->vcnum = i; | |
a322f70c | 1177 | } |
a322f70c DW |
1178 | |
1179 | return 0; | |
1180 | } | |
1181 | ||
a322f70c DW |
1182 | static int load_super_ddf(struct supertype *st, int fd, |
1183 | char *devname) | |
1184 | { | |
1185 | unsigned long long dsize; | |
1186 | struct ddf_super *super; | |
1187 | int rv; | |
1188 | ||
a322f70c DW |
1189 | if (get_dev_size(fd, devname, &dsize) == 0) |
1190 | return 1; | |
1191 | ||
a34fea0e | 1192 | if (test_partition(fd)) |
691c6ee1 N |
1193 | /* DDF is not allowed on partitions */ |
1194 | return 1; | |
1195 | ||
a322f70c DW |
1196 | /* 32M is a lower bound */ |
1197 | if (dsize <= 32*1024*1024) { | |
97320d7c | 1198 | if (devname) |
e7b84f9d N |
1199 | pr_err("%s is too small for ddf: " |
1200 | "size is %llu sectors.\n", | |
1201 | devname, dsize>>9); | |
97320d7c | 1202 | return 1; |
a322f70c DW |
1203 | } |
1204 | if (dsize & 511) { | |
97320d7c | 1205 | if (devname) |
e7b84f9d N |
1206 | pr_err("%s is an odd size for ddf: " |
1207 | "size is %llu bytes.\n", | |
1208 | devname, dsize); | |
97320d7c | 1209 | return 1; |
a322f70c DW |
1210 | } |
1211 | ||
37424f13 DW |
1212 | free_super_ddf(st); |
1213 | ||
6416d527 | 1214 | if (posix_memalign((void**)&super, 512, sizeof(*super))!= 0) { |
e7b84f9d | 1215 | pr_err("malloc of %zu failed.\n", |
a322f70c DW |
1216 | sizeof(*super)); |
1217 | return 1; | |
1218 | } | |
a2349791 | 1219 | memset(super, 0, sizeof(*super)); |
a322f70c DW |
1220 | |
1221 | rv = load_ddf_headers(fd, super, devname); | |
1222 | if (rv) { | |
1223 | free(super); | |
1224 | return rv; | |
1225 | } | |
1226 | ||
1227 | /* Have valid headers and have chosen the best. Let's read in the rest*/ | |
1228 | ||
1229 | rv = load_ddf_global(fd, super, devname); | |
1230 | ||
1231 | if (rv) { | |
1232 | if (devname) | |
e7b84f9d N |
1233 | pr_err("Failed to load all information " |
1234 | "sections on %s\n", devname); | |
a322f70c DW |
1235 | free(super); |
1236 | return rv; | |
1237 | } | |
1238 | ||
3d2c4fc7 DW |
1239 | rv = load_ddf_local(fd, super, devname, 0); |
1240 | ||
1241 | if (rv) { | |
1242 | if (devname) | |
e7b84f9d N |
1243 | pr_err("Failed to load all information " |
1244 | "sections on %s\n", devname); | |
3d2c4fc7 DW |
1245 | free(super); |
1246 | return rv; | |
1247 | } | |
a322f70c DW |
1248 | |
1249 | /* Should possibly check the sections .... */ | |
1250 | ||
1251 | st->sb = super; | |
1252 | if (st->ss == NULL) { | |
1253 | st->ss = &super_ddf; | |
1254 | st->minor_version = 0; | |
1255 | st->max_devs = 512; | |
1256 | } | |
1257 | return 0; | |
1258 | ||
1259 | } | |
1260 | ||
1261 | static void free_super_ddf(struct supertype *st) | |
1262 | { | |
1263 | struct ddf_super *ddf = st->sb; | |
1264 | if (ddf == NULL) | |
1265 | return; | |
1266 | free(ddf->phys); | |
1267 | free(ddf->virt); | |
3921e41a | 1268 | free(ddf->conf); |
a322f70c DW |
1269 | while (ddf->conflist) { |
1270 | struct vcl *v = ddf->conflist; | |
1271 | ddf->conflist = v->next; | |
59e36268 NB |
1272 | if (v->block_sizes) |
1273 | free(v->block_sizes); | |
3c48f7be | 1274 | if (v->other_bvds) |
1275 | /* | |
1276 | v->other_bvds[0] points to beginning of buffer, | |
1277 | see alloc_other_bvds() | |
1278 | */ | |
1279 | free(v->other_bvds[0]); | |
a322f70c DW |
1280 | free(v); |
1281 | } | |
1282 | while (ddf->dlist) { | |
1283 | struct dl *d = ddf->dlist; | |
1284 | ddf->dlist = d->next; | |
1285 | if (d->fd >= 0) | |
1286 | close(d->fd); | |
b2280677 NB |
1287 | if (d->spare) |
1288 | free(d->spare); | |
a322f70c DW |
1289 | free(d); |
1290 | } | |
8a38cb04 N |
1291 | while (ddf->add_list) { |
1292 | struct dl *d = ddf->add_list; | |
1293 | ddf->add_list = d->next; | |
1294 | if (d->fd >= 0) | |
1295 | close(d->fd); | |
1296 | if (d->spare) | |
1297 | free(d->spare); | |
1298 | free(d); | |
1299 | } | |
a322f70c DW |
1300 | free(ddf); |
1301 | st->sb = NULL; | |
1302 | } | |
1303 | ||
1304 | static struct supertype *match_metadata_desc_ddf(char *arg) | |
1305 | { | |
56cb05c4 | 1306 | /* 'ddf' only supports containers */ |
a322f70c DW |
1307 | struct supertype *st; |
1308 | if (strcmp(arg, "ddf") != 0 && | |
1309 | strcmp(arg, "default") != 0 | |
1310 | ) | |
1311 | return NULL; | |
1312 | ||
503975b9 | 1313 | st = xcalloc(1, sizeof(*st)); |
a322f70c DW |
1314 | st->ss = &super_ddf; |
1315 | st->max_devs = 512; | |
1316 | st->minor_version = 0; | |
1317 | st->sb = NULL; | |
1318 | return st; | |
1319 | } | |
1320 | ||
a322f70c DW |
1321 | #ifndef MDASSEMBLE |
1322 | ||
1323 | static mapping_t ddf_state[] = { | |
1324 | { "Optimal", 0}, | |
1325 | { "Degraded", 1}, | |
1326 | { "Deleted", 2}, | |
1327 | { "Missing", 3}, | |
1328 | { "Failed", 4}, | |
1329 | { "Partially Optimal", 5}, | |
1330 | { "-reserved-", 6}, | |
1331 | { "-reserved-", 7}, | |
1332 | { NULL, 0} | |
1333 | }; | |
1334 | ||
1335 | static mapping_t ddf_init_state[] = { | |
1336 | { "Not Initialised", 0}, | |
1337 | { "QuickInit in Progress", 1}, | |
1338 | { "Fully Initialised", 2}, | |
1339 | { "*UNKNOWN*", 3}, | |
1340 | { NULL, 0} | |
1341 | }; | |
1342 | static mapping_t ddf_access[] = { | |
1343 | { "Read/Write", 0}, | |
1344 | { "Reserved", 1}, | |
1345 | { "Read Only", 2}, | |
1346 | { "Blocked (no access)", 3}, | |
1347 | { NULL ,0} | |
1348 | }; | |
1349 | ||
1350 | static mapping_t ddf_level[] = { | |
1351 | { "RAID0", DDF_RAID0}, | |
1352 | { "RAID1", DDF_RAID1}, | |
1353 | { "RAID3", DDF_RAID3}, | |
1354 | { "RAID4", DDF_RAID4}, | |
1355 | { "RAID5", DDF_RAID5}, | |
1356 | { "RAID1E",DDF_RAID1E}, | |
1357 | { "JBOD", DDF_JBOD}, | |
1358 | { "CONCAT",DDF_CONCAT}, | |
1359 | { "RAID5E",DDF_RAID5E}, | |
1360 | { "RAID5EE",DDF_RAID5EE}, | |
1361 | { "RAID6", DDF_RAID6}, | |
1362 | { NULL, 0} | |
1363 | }; | |
1364 | static mapping_t ddf_sec_level[] = { | |
1365 | { "Striped", DDF_2STRIPED}, | |
1366 | { "Mirrored", DDF_2MIRRORED}, | |
1367 | { "Concat", DDF_2CONCAT}, | |
1368 | { "Spanned", DDF_2SPANNED}, | |
1369 | { NULL, 0} | |
1370 | }; | |
1371 | #endif | |
1372 | ||
fb9d0acb | 1373 | static int all_ff(const char *guid) |
42dc2744 N |
1374 | { |
1375 | int i; | |
1376 | for (i = 0; i < DDF_GUID_LEN; i++) | |
1377 | if (guid[i] != (char)0xff) | |
1378 | return 0; | |
1379 | return 1; | |
1380 | } | |
1381 | ||
4441541f N |
1382 | static const char *guid_str(const char *guid) |
1383 | { | |
1384 | static char buf[DDF_GUID_LEN*2+1]; | |
1385 | int i; | |
1386 | char *p = buf; | |
1387 | for (i = 0; i < DDF_GUID_LEN; i++) { | |
1388 | unsigned char c = guid[i]; | |
1389 | if (c >= 32 && c < 127) | |
1390 | p += sprintf(p, "%c", c); | |
1391 | else | |
1392 | p += sprintf(p, "%02x", c); | |
1393 | } | |
1394 | *p = '\0'; | |
1395 | return (const char *) buf; | |
1396 | } | |
1397 | ||
a322f70c DW |
1398 | #ifndef MDASSEMBLE |
1399 | static void print_guid(char *guid, int tstamp) | |
1400 | { | |
1401 | /* A GUIDs are part (or all) ASCII and part binary. | |
1402 | * They tend to be space padded. | |
59e36268 NB |
1403 | * We print the GUID in HEX, then in parentheses add |
1404 | * any initial ASCII sequence, and a possible | |
1405 | * time stamp from bytes 16-19 | |
a322f70c DW |
1406 | */ |
1407 | int l = DDF_GUID_LEN; | |
1408 | int i; | |
59e36268 NB |
1409 | |
1410 | for (i=0 ; i<DDF_GUID_LEN ; i++) { | |
1411 | if ((i&3)==0 && i != 0) printf(":"); | |
1412 | printf("%02X", guid[i]&255); | |
1413 | } | |
1414 | ||
cfccea8c | 1415 | printf("\n ("); |
a322f70c DW |
1416 | while (l && guid[l-1] == ' ') |
1417 | l--; | |
1418 | for (i=0 ; i<l ; i++) { | |
1419 | if (guid[i] >= 0x20 && guid[i] < 0x7f) | |
1420 | fputc(guid[i], stdout); | |
1421 | else | |
59e36268 | 1422 | break; |
a322f70c DW |
1423 | } |
1424 | if (tstamp) { | |
1425 | time_t then = __be32_to_cpu(*(__u32*)(guid+16)) + DECADE; | |
1426 | char tbuf[100]; | |
1427 | struct tm *tm; | |
1428 | tm = localtime(&then); | |
59e36268 | 1429 | strftime(tbuf, 100, " %D %T",tm); |
a322f70c DW |
1430 | fputs(tbuf, stdout); |
1431 | } | |
59e36268 | 1432 | printf(")"); |
a322f70c DW |
1433 | } |
1434 | ||
1435 | static void examine_vd(int n, struct ddf_super *sb, char *guid) | |
1436 | { | |
8c3b8c2c | 1437 | int crl = sb->conf_rec_len; |
a322f70c DW |
1438 | struct vcl *vcl; |
1439 | ||
1440 | for (vcl = sb->conflist ; vcl ; vcl = vcl->next) { | |
f21e18ca | 1441 | unsigned int i; |
a322f70c DW |
1442 | struct vd_config *vc = &vcl->conf; |
1443 | ||
60931cf9 | 1444 | if (!be32_eq(calc_crc(vc, crl*512), vc->crc)) |
a322f70c DW |
1445 | continue; |
1446 | if (memcmp(vc->guid, guid, DDF_GUID_LEN) != 0) | |
1447 | continue; | |
1448 | ||
1449 | /* Ok, we know about this VD, let's give more details */ | |
b06e3095 | 1450 | printf(" Raid Devices[%d] : %d (", n, |
a8173e43 | 1451 | be16_to_cpu(vc->prim_elmnt_count)); |
1452 | for (i = 0; i < be16_to_cpu(vc->prim_elmnt_count); i++) { | |
b06e3095 | 1453 | int j; |
217dead4 | 1454 | int cnt = be16_to_cpu(sb->phys->max_pdes); |
b06e3095 | 1455 | for (j=0; j<cnt; j++) |
60931cf9 | 1456 | if (be32_eq(vc->phys_refnum[i], |
1457 | sb->phys->entries[j].refnum)) | |
b06e3095 N |
1458 | break; |
1459 | if (i) printf(" "); | |
1460 | if (j < cnt) | |
1461 | printf("%d", j); | |
1462 | else | |
1463 | printf("--"); | |
1464 | } | |
1465 | printf(")\n"); | |
1466 | if (vc->chunk_shift != 255) | |
613b0d17 N |
1467 | printf(" Chunk Size[%d] : %d sectors\n", n, |
1468 | 1 << vc->chunk_shift); | |
a322f70c DW |
1469 | printf(" Raid Level[%d] : %s\n", n, |
1470 | map_num(ddf_level, vc->prl)?:"-unknown-"); | |
1471 | if (vc->sec_elmnt_count != 1) { | |
1472 | printf(" Secondary Position[%d] : %d of %d\n", n, | |
1473 | vc->sec_elmnt_seq, vc->sec_elmnt_count); | |
1474 | printf(" Secondary Level[%d] : %s\n", n, | |
1475 | map_num(ddf_sec_level, vc->srl) ?: "-unknown-"); | |
1476 | } | |
1477 | printf(" Device Size[%d] : %llu\n", n, | |
9d0c6b70 | 1478 | be64_to_cpu(vc->blocks)/2); |
a322f70c | 1479 | printf(" Array Size[%d] : %llu\n", n, |
9d0c6b70 | 1480 | be64_to_cpu(vc->array_blocks)/2); |
a322f70c DW |
1481 | } |
1482 | } | |
1483 | ||
1484 | static void examine_vds(struct ddf_super *sb) | |
1485 | { | |
a8173e43 | 1486 | int cnt = be16_to_cpu(sb->virt->populated_vdes); |
fb9d0acb | 1487 | unsigned int i; |
a322f70c DW |
1488 | printf(" Virtual Disks : %d\n", cnt); |
1489 | ||
a8173e43 | 1490 | for (i = 0; i < be16_to_cpu(sb->virt->max_vdes); i++) { |
a322f70c | 1491 | struct virtual_entry *ve = &sb->virt->entries[i]; |
fb9d0acb | 1492 | if (all_ff(ve->guid)) |
1493 | continue; | |
b06e3095 | 1494 | printf("\n"); |
a322f70c DW |
1495 | printf(" VD GUID[%d] : ", i); print_guid(ve->guid, 1); |
1496 | printf("\n"); | |
a8173e43 | 1497 | printf(" unit[%d] : %d\n", i, be16_to_cpu(ve->unit)); |
a322f70c DW |
1498 | printf(" state[%d] : %s, %s%s\n", i, |
1499 | map_num(ddf_state, ve->state & 7), | |
cc83a819 N |
1500 | (ve->state & DDF_state_morphing) ? "Morphing, ": "", |
1501 | (ve->state & DDF_state_inconsistent)? "Not Consistent" : "Consistent"); | |
a322f70c | 1502 | printf(" init state[%d] : %s\n", i, |
cc83a819 | 1503 | map_num(ddf_init_state, ve->init_state&DDF_initstate_mask)); |
a322f70c | 1504 | printf(" access[%d] : %s\n", i, |
cc83a819 | 1505 | map_num(ddf_access, (ve->init_state & DDF_access_mask) >> 6)); |
a322f70c DW |
1506 | printf(" Name[%d] : %.16s\n", i, ve->name); |
1507 | examine_vd(i, sb, ve->guid); | |
1508 | } | |
1509 | if (cnt) printf("\n"); | |
1510 | } | |
1511 | ||
1512 | static void examine_pds(struct ddf_super *sb) | |
1513 | { | |
217dead4 | 1514 | int cnt = be16_to_cpu(sb->phys->max_pdes); |
a322f70c DW |
1515 | int i; |
1516 | struct dl *dl; | |
d2ec75fb | 1517 | int unlisted = 0; |
a322f70c | 1518 | printf(" Physical Disks : %d\n", cnt); |
962371a5 | 1519 | printf(" Number RefNo Size Device Type/State\n"); |
a322f70c | 1520 | |
d2ec75fb N |
1521 | for (dl = sb->dlist; dl; dl = dl->next) |
1522 | dl->displayed = 0; | |
1523 | ||
a322f70c DW |
1524 | for (i=0 ; i<cnt ; i++) { |
1525 | struct phys_disk_entry *pd = &sb->phys->entries[i]; | |
a8173e43 | 1526 | int type = be16_to_cpu(pd->type); |
1527 | int state = be16_to_cpu(pd->state); | |
a322f70c | 1528 | |
217dead4 N |
1529 | if (be32_to_cpu(pd->refnum) == 0xffffffff) |
1530 | /* Not in use */ | |
1531 | continue; | |
b06e3095 N |
1532 | //printf(" PD GUID[%d] : ", i); print_guid(pd->guid, 0); |
1533 | //printf("\n"); | |
1534 | printf(" %3d %08x ", i, | |
60931cf9 | 1535 | be32_to_cpu(pd->refnum)); |
613b0d17 | 1536 | printf("%8lluK ", |
9d0c6b70 | 1537 | be64_to_cpu(pd->config_size)>>1); |
b06e3095 | 1538 | for (dl = sb->dlist; dl ; dl = dl->next) { |
60931cf9 | 1539 | if (be32_eq(dl->disk.refnum, pd->refnum)) { |
b06e3095 N |
1540 | char *dv = map_dev(dl->major, dl->minor, 0); |
1541 | if (dv) { | |
962371a5 | 1542 | printf("%-15s", dv); |
b06e3095 N |
1543 | break; |
1544 | } | |
1545 | } | |
1546 | } | |
1547 | if (!dl) | |
962371a5 | 1548 | printf("%15s",""); |
d2ec75fb N |
1549 | else |
1550 | dl->displayed = 1; | |
b06e3095 | 1551 | printf(" %s%s%s%s%s", |
a322f70c | 1552 | (type&2) ? "active":"", |
b06e3095 | 1553 | (type&4) ? "Global-Spare":"", |
a322f70c DW |
1554 | (type&8) ? "spare" : "", |
1555 | (type&16)? ", foreign" : "", | |
1556 | (type&32)? "pass-through" : ""); | |
18cb4496 N |
1557 | if (state & DDF_Failed) |
1558 | /* This over-rides these three */ | |
1559 | state &= ~(DDF_Online|DDF_Rebuilding|DDF_Transition); | |
b06e3095 | 1560 | printf("/%s%s%s%s%s%s%s", |
a322f70c DW |
1561 | (state&1)? "Online": "Offline", |
1562 | (state&2)? ", Failed": "", | |
1563 | (state&4)? ", Rebuilding": "", | |
1564 | (state&8)? ", in-transition": "", | |
b06e3095 N |
1565 | (state&16)? ", SMART-errors": "", |
1566 | (state&32)? ", Unrecovered-Read-Errors": "", | |
a322f70c | 1567 | (state&64)? ", Missing" : ""); |
a322f70c DW |
1568 | printf("\n"); |
1569 | } | |
d2ec75fb N |
1570 | for (dl = sb->dlist; dl; dl = dl->next) { |
1571 | char *dv; | |
1572 | if (dl->displayed) | |
1573 | continue; | |
1574 | if (!unlisted) | |
1575 | printf(" Physical disks not in metadata!:\n"); | |
1576 | unlisted = 1; | |
1577 | dv = map_dev(dl->major, dl->minor, 0); | |
1578 | printf(" %08x %s\n", be32_to_cpu(dl->disk.refnum), | |
1579 | dv ? dv : "-unknown-"); | |
1580 | } | |
1581 | if (unlisted) | |
1582 | printf("\n"); | |
a322f70c DW |
1583 | } |
1584 | ||
1585 | static void examine_super_ddf(struct supertype *st, char *homehost) | |
1586 | { | |
1587 | struct ddf_super *sb = st->sb; | |
1588 | ||
60931cf9 | 1589 | printf(" Magic : %08x\n", be32_to_cpu(sb->anchor.magic)); |
a322f70c | 1590 | printf(" Version : %.8s\n", sb->anchor.revision); |
598f0d58 NB |
1591 | printf("Controller GUID : "); print_guid(sb->controller.guid, 0); |
1592 | printf("\n"); | |
1593 | printf(" Container GUID : "); print_guid(sb->anchor.guid, 1); | |
a322f70c | 1594 | printf("\n"); |
60931cf9 | 1595 | printf(" Seq : %08x\n", be32_to_cpu(sb->active->seq)); |
56cb05c4 | 1596 | printf(" Redundant hdr : %s\n", (be32_eq(sb->secondary.magic, |
60931cf9 | 1597 | DDF_HEADER_MAGIC) |
56cb05c4 | 1598 | ?"yes" : "no")); |
a322f70c DW |
1599 | examine_vds(sb); |
1600 | examine_pds(sb); | |
1601 | } | |
1602 | ||
bedbf68a | 1603 | static unsigned int get_vd_num_of_subarray(struct supertype *st) |
1604 | { | |
1605 | /* | |
1606 | * Figure out the VD number for this supertype. | |
1607 | * Returns DDF_CONTAINER for the container itself, | |
1608 | * and DDF_NOTFOUND on error. | |
1609 | */ | |
1610 | struct ddf_super *ddf = st->sb; | |
1611 | struct mdinfo *sra; | |
1612 | char *sub, *end; | |
1613 | unsigned int vcnum; | |
1614 | ||
1615 | if (*st->container_devnm == '\0') | |
1616 | return DDF_CONTAINER; | |
1617 | ||
1618 | sra = sysfs_read(-1, st->devnm, GET_VERSION); | |
1619 | if (!sra || sra->array.major_version != -1 || | |
1620 | sra->array.minor_version != -2 || | |
1621 | !is_subarray(sra->text_version)) | |
1622 | return DDF_NOTFOUND; | |
1623 | ||
1624 | sub = strchr(sra->text_version + 1, '/'); | |
1625 | if (sub != NULL) | |
1626 | vcnum = strtoul(sub + 1, &end, 10); | |
1627 | if (sub == NULL || *sub == '\0' || *end != '\0' || | |
a8173e43 | 1628 | vcnum >= be16_to_cpu(ddf->active->max_vd_entries)) |
bedbf68a | 1629 | return DDF_NOTFOUND; |
1630 | ||
1631 | return vcnum; | |
1632 | } | |
1633 | ||
061f2c6a | 1634 | static void brief_examine_super_ddf(struct supertype *st, int verbose) |
4737ae25 N |
1635 | { |
1636 | /* We just write a generic DDF ARRAY entry | |
1637 | */ | |
1638 | struct mdinfo info; | |
1639 | char nbuf[64]; | |
a5d85af7 | 1640 | getinfo_super_ddf(st, &info, NULL); |
4737ae25 N |
1641 | fname_from_uuid(st, &info, nbuf, ':'); |
1642 | ||
1643 | printf("ARRAY metadata=ddf UUID=%s\n", nbuf + 5); | |
1644 | } | |
1645 | ||
1646 | static void brief_examine_subarrays_ddf(struct supertype *st, int verbose) | |
a322f70c | 1647 | { |
56cb05c4 N |
1648 | /* We write a DDF ARRAY member entry for each vd, identifying container |
1649 | * by uuid and member by unit number and uuid. | |
a322f70c | 1650 | */ |
42dc2744 | 1651 | struct ddf_super *ddf = st->sb; |
ff54de6e | 1652 | struct mdinfo info; |
f21e18ca | 1653 | unsigned int i; |
ff54de6e | 1654 | char nbuf[64]; |
a5d85af7 | 1655 | getinfo_super_ddf(st, &info, NULL); |
ff54de6e | 1656 | fname_from_uuid(st, &info, nbuf, ':'); |
42dc2744 | 1657 | |
a8173e43 | 1658 | for (i = 0; i < be16_to_cpu(ddf->virt->max_vdes); i++) { |
42dc2744 N |
1659 | struct virtual_entry *ve = &ddf->virt->entries[i]; |
1660 | struct vcl vcl; | |
1661 | char nbuf1[64]; | |
a8b25633 | 1662 | char namebuf[17]; |
42dc2744 N |
1663 | if (all_ff(ve->guid)) |
1664 | continue; | |
1665 | memcpy(vcl.conf.guid, ve->guid, DDF_GUID_LEN); | |
1666 | ddf->currentconf =&vcl; | |
7087f02b | 1667 | vcl.vcnum = i; |
42dc2744 N |
1668 | uuid_from_super_ddf(st, info.uuid); |
1669 | fname_from_uuid(st, &info, nbuf1, ':'); | |
a8b25633 | 1670 | _ddf_array_name(namebuf, ddf, i); |
1671 | printf("ARRAY%s%s container=%s member=%d UUID=%s\n", | |
1672 | namebuf[0] == '\0' ? "" : " /dev/md/", namebuf, | |
42dc2744 N |
1673 | nbuf+5, i, nbuf1+5); |
1674 | } | |
a322f70c DW |
1675 | } |
1676 | ||
bceedeec N |
1677 | static void export_examine_super_ddf(struct supertype *st) |
1678 | { | |
1679 | struct mdinfo info; | |
1680 | char nbuf[64]; | |
a5d85af7 | 1681 | getinfo_super_ddf(st, &info, NULL); |
bceedeec N |
1682 | fname_from_uuid(st, &info, nbuf, ':'); |
1683 | printf("MD_METADATA=ddf\n"); | |
1684 | printf("MD_LEVEL=container\n"); | |
1685 | printf("MD_UUID=%s\n", nbuf+5); | |
cc9bfd9e | 1686 | printf("MD_DEVICES=%u\n", |
1687 | be16_to_cpu(((struct ddf_super *)st->sb)->phys->used_pdes)); | |
bceedeec | 1688 | } |
bceedeec | 1689 | |
74db60b0 N |
1690 | static int copy_metadata_ddf(struct supertype *st, int from, int to) |
1691 | { | |
1692 | void *buf; | |
1693 | unsigned long long dsize, offset; | |
1694 | int bytes; | |
1695 | struct ddf_header *ddf; | |
1696 | int written = 0; | |
1697 | ||
1698 | /* The meta consists of an anchor, a primary, and a secondary. | |
1699 | * This all lives at the end of the device. | |
1700 | * So it is easiest to find the earliest of primary and | |
1701 | * secondary, and copy everything from there. | |
1702 | * | |
56cb05c4 | 1703 | * Anchor is 512 from end. It contains primary_lba and secondary_lba |
74db60b0 N |
1704 | * we choose one of those |
1705 | */ | |
1706 | ||
1707 | if (posix_memalign(&buf, 4096, 4096) != 0) | |
1708 | return 1; | |
1709 | ||
1710 | if (!get_dev_size(from, NULL, &dsize)) | |
1711 | goto err; | |
1712 | ||
1713 | if (lseek64(from, dsize-512, 0) < 0) | |
1714 | goto err; | |
1715 | if (read(from, buf, 512) != 512) | |
1716 | goto err; | |
1717 | ddf = buf; | |
60931cf9 | 1718 | if (!be32_eq(ddf->magic, DDF_HEADER_MAGIC) || |
1719 | !be32_eq(calc_crc(ddf, 512), ddf->crc) || | |
74db60b0 N |
1720 | (memcmp(ddf->revision, DDF_REVISION_0, 8) != 0 && |
1721 | memcmp(ddf->revision, DDF_REVISION_2, 8) != 0)) | |
1722 | goto err; | |
1723 | ||
1724 | offset = dsize - 512; | |
9d0c6b70 | 1725 | if ((be64_to_cpu(ddf->primary_lba) << 9) < offset) |
1726 | offset = be64_to_cpu(ddf->primary_lba) << 9; | |
1727 | if ((be64_to_cpu(ddf->secondary_lba) << 9) < offset) | |
1728 | offset = be64_to_cpu(ddf->secondary_lba) << 9; | |
74db60b0 N |
1729 | |
1730 | bytes = dsize - offset; | |
1731 | ||
1732 | if (lseek64(from, offset, 0) < 0 || | |
1733 | lseek64(to, offset, 0) < 0) | |
1734 | goto err; | |
1735 | while (written < bytes) { | |
1736 | int n = bytes - written; | |
1737 | if (n > 4096) | |
1738 | n = 4096; | |
1739 | if (read(from, buf, n) != n) | |
1740 | goto err; | |
1741 | if (write(to, buf, n) != n) | |
1742 | goto err; | |
1743 | written += n; | |
1744 | } | |
1745 | free(buf); | |
1746 | return 0; | |
1747 | err: | |
1748 | free(buf); | |
1749 | return 1; | |
1750 | } | |
1751 | ||
a322f70c DW |
1752 | static void detail_super_ddf(struct supertype *st, char *homehost) |
1753 | { | |
ff84d052 N |
1754 | struct ddf_super *sb = st->sb; |
1755 | int cnt = be16_to_cpu(sb->virt->populated_vdes); | |
1756 | ||
1757 | printf(" Container GUID : "); print_guid(sb->anchor.guid, 1); | |
1758 | printf("\n"); | |
1759 | printf(" Seq : %08x\n", be32_to_cpu(sb->active->seq)); | |
1760 | printf(" Virtual Disks : %d\n", cnt); | |
1761 | printf("\n"); | |
a322f70c DW |
1762 | } |
1763 | ||
7087f02b | 1764 | static const char *vendors_with_variable_volume_UUID[] = { |
1765 | "LSI ", | |
1766 | }; | |
1767 | ||
1768 | static int volume_id_is_reliable(const struct ddf_super *ddf) | |
1769 | { | |
1c0aebc2 | 1770 | int n = ARRAY_SIZE(vendors_with_variable_volume_UUID); |
7087f02b | 1771 | int i; |
1772 | for (i = 0; i < n; i++) | |
1773 | if (!memcmp(ddf->controller.guid, | |
1774 | vendors_with_variable_volume_UUID[i], 8)) | |
1775 | return 0; | |
1776 | return 1; | |
1777 | } | |
1778 | ||
1779 | static void uuid_of_ddf_subarray(const struct ddf_super *ddf, | |
1780 | unsigned int vcnum, int uuid[4]) | |
1781 | { | |
1782 | char buf[DDF_GUID_LEN+18], sha[20], *p; | |
1783 | struct sha1_ctx ctx; | |
1784 | if (volume_id_is_reliable(ddf)) { | |
1785 | uuid_from_ddf_guid(ddf->virt->entries[vcnum].guid, uuid); | |
1786 | return; | |
1787 | } | |
1788 | /* | |
1789 | * Some fake RAID BIOSes (in particular, LSI ones) change the | |
1790 | * VD GUID at every boot. These GUIDs are not suitable for | |
1791 | * identifying an array. Luckily the header GUID appears to | |
1792 | * remain constant. | |
1793 | * We construct a pseudo-UUID from the header GUID and those | |
1794 | * properties of the subarray that we expect to remain constant. | |
1795 | */ | |
1796 | memset(buf, 0, sizeof(buf)); | |
1797 | p = buf; | |
1798 | memcpy(p, ddf->anchor.guid, DDF_GUID_LEN); | |
1799 | p += DDF_GUID_LEN; | |
1800 | memcpy(p, ddf->virt->entries[vcnum].name, 16); | |
1801 | p += 16; | |
1802 | *((__u16 *) p) = vcnum; | |
1803 | sha1_init_ctx(&ctx); | |
1804 | sha1_process_bytes(buf, sizeof(buf), &ctx); | |
1805 | sha1_finish_ctx(&ctx, sha); | |
1806 | memcpy(uuid, sha, 4*4); | |
1807 | } | |
1808 | ||
a322f70c DW |
1809 | static void brief_detail_super_ddf(struct supertype *st) |
1810 | { | |
ff54de6e N |
1811 | struct mdinfo info; |
1812 | char nbuf[64]; | |
bedbf68a | 1813 | struct ddf_super *ddf = st->sb; |
1814 | unsigned int vcnum = get_vd_num_of_subarray(st); | |
1815 | if (vcnum == DDF_CONTAINER) | |
1816 | uuid_from_super_ddf(st, info.uuid); | |
1817 | else if (vcnum == DDF_NOTFOUND) | |
1818 | return; | |
1819 | else | |
7087f02b | 1820 | uuid_of_ddf_subarray(ddf, vcnum, info.uuid); |
ff54de6e N |
1821 | fname_from_uuid(st, &info, nbuf,':'); |
1822 | printf(" UUID=%s", nbuf + 5); | |
a322f70c | 1823 | } |
a322f70c DW |
1824 | #endif |
1825 | ||
1826 | static int match_home_ddf(struct supertype *st, char *homehost) | |
1827 | { | |
1828 | /* It matches 'this' host if the controller is a | |
1829 | * Linux-MD controller with vendor_data matching | |
56cb05c4 N |
1830 | * the hostname. It would be nice if we could |
1831 | * test against controller found in /sys or somewhere... | |
a322f70c DW |
1832 | */ |
1833 | struct ddf_super *ddf = st->sb; | |
f21e18ca | 1834 | unsigned int len; |
d1d3482b N |
1835 | |
1836 | if (!homehost) | |
1837 | return 0; | |
1838 | len = strlen(homehost); | |
a322f70c DW |
1839 | |
1840 | return (memcmp(ddf->controller.guid, T10, 8) == 0 && | |
1841 | len < sizeof(ddf->controller.vendor_data) && | |
1842 | memcmp(ddf->controller.vendor_data, homehost,len) == 0 && | |
1843 | ddf->controller.vendor_data[len] == 0); | |
1844 | } | |
1845 | ||
0e600426 | 1846 | #ifndef MDASSEMBLE |
baba3f4e | 1847 | static int find_index_in_bvd(const struct ddf_super *ddf, |
1848 | const struct vd_config *conf, unsigned int n, | |
1849 | unsigned int *n_bvd) | |
1850 | { | |
1851 | /* | |
56cb05c4 N |
1852 | * Find the index of the n-th valid physical disk in this BVD. |
1853 | * Unused entries can be sprinkled in with the used entries, | |
1854 | * but don't count. | |
baba3f4e | 1855 | */ |
1856 | unsigned int i, j; | |
56cb05c4 N |
1857 | for (i = 0, j = 0; |
1858 | i < ddf->mppe && j < be16_to_cpu(conf->prim_elmnt_count); | |
1859 | i++) { | |
60931cf9 | 1860 | if (be32_to_cpu(conf->phys_refnum[i]) != 0xffffffff) { |
baba3f4e | 1861 | if (n == j) { |
1862 | *n_bvd = i; | |
1863 | return 1; | |
1864 | } | |
1865 | j++; | |
1866 | } | |
1867 | } | |
1868 | dprintf("%s: couldn't find BVD member %u (total %u)\n", | |
a8173e43 | 1869 | __func__, n, be16_to_cpu(conf->prim_elmnt_count)); |
baba3f4e | 1870 | return 0; |
1871 | } | |
1872 | ||
56cb05c4 N |
1873 | /* Given a member array instance number, and a raid disk within that instance, |
1874 | * find the vd_config structure. The offset of the given disk in the phys_refnum | |
1875 | * table is returned in n_bvd. | |
1876 | * For two-level members with a secondary raid level the vd_config for | |
1877 | * the appropriate BVD is returned. | |
1878 | * The return value is always &vlc->conf, where vlc is returned in last pointer. | |
1879 | */ | |
baba3f4e | 1880 | static struct vd_config *find_vdcr(struct ddf_super *ddf, unsigned int inst, |
1881 | unsigned int n, | |
1882 | unsigned int *n_bvd, struct vcl **vcl) | |
a322f70c | 1883 | { |
7a7cc504 | 1884 | struct vcl *v; |
59e36268 | 1885 | |
baba3f4e | 1886 | for (v = ddf->conflist; v; v = v->next) { |
84e32e19 | 1887 | unsigned int nsec, ibvd = 0; |
baba3f4e | 1888 | struct vd_config *conf; |
1889 | if (inst != v->vcnum) | |
1890 | continue; | |
1891 | conf = &v->conf; | |
1892 | if (conf->sec_elmnt_count == 1) { | |
1893 | if (find_index_in_bvd(ddf, conf, n, n_bvd)) { | |
1894 | *vcl = v; | |
1895 | return conf; | |
1896 | } else | |
1897 | goto bad; | |
1898 | } | |
1899 | if (v->other_bvds == NULL) { | |
1900 | pr_err("%s: BUG: other_bvds is NULL, nsec=%u\n", | |
1901 | __func__, conf->sec_elmnt_count); | |
1902 | goto bad; | |
1903 | } | |
a8173e43 | 1904 | nsec = n / be16_to_cpu(conf->prim_elmnt_count); |
baba3f4e | 1905 | if (conf->sec_elmnt_seq != nsec) { |
1906 | for (ibvd = 1; ibvd < conf->sec_elmnt_count; ibvd++) { | |
baba3f4e | 1907 | if (v->other_bvds[ibvd-1]->sec_elmnt_seq |
1908 | == nsec) | |
1909 | break; | |
1910 | } | |
1911 | if (ibvd == conf->sec_elmnt_count) | |
1912 | goto bad; | |
1913 | conf = v->other_bvds[ibvd-1]; | |
1914 | } | |
1915 | if (!find_index_in_bvd(ddf, conf, | |
1916 | n - nsec*conf->sec_elmnt_count, n_bvd)) | |
1917 | goto bad; | |
1918 | dprintf("%s: found disk %u as member %u in bvd %d of array %u\n" | |
84e32e19 | 1919 | , __func__, n, *n_bvd, ibvd, inst); |
baba3f4e | 1920 | *vcl = v; |
1921 | return conf; | |
1922 | } | |
1923 | bad: | |
1924 | pr_err("%s: Could't find disk %d in array %u\n", __func__, n, inst); | |
7a7cc504 NB |
1925 | return NULL; |
1926 | } | |
0e600426 | 1927 | #endif |
7a7cc504 | 1928 | |
60931cf9 | 1929 | static int find_phys(const struct ddf_super *ddf, be32 phys_refnum) |
7a7cc504 NB |
1930 | { |
1931 | /* Find the entry in phys_disk which has the given refnum | |
1932 | * and return it's index | |
1933 | */ | |
f21e18ca | 1934 | unsigned int i; |
a8173e43 | 1935 | for (i = 0; i < be16_to_cpu(ddf->phys->max_pdes); i++) |
60931cf9 | 1936 | if (be32_eq(ddf->phys->entries[i].refnum, phys_refnum)) |
7a7cc504 NB |
1937 | return i; |
1938 | return -1; | |
a322f70c DW |
1939 | } |
1940 | ||
bedbf68a | 1941 | static void uuid_from_ddf_guid(const char *guid, int uuid[4]) |
1942 | { | |
1943 | char buf[20]; | |
1944 | struct sha1_ctx ctx; | |
1945 | sha1_init_ctx(&ctx); | |
1946 | sha1_process_bytes(guid, DDF_GUID_LEN, &ctx); | |
1947 | sha1_finish_ctx(&ctx, buf); | |
1948 | memcpy(uuid, buf, 4*4); | |
1949 | } | |
1950 | ||
a322f70c DW |
1951 | static void uuid_from_super_ddf(struct supertype *st, int uuid[4]) |
1952 | { | |
1953 | /* The uuid returned here is used for: | |
1954 | * uuid to put into bitmap file (Create, Grow) | |
1955 | * uuid for backup header when saving critical section (Grow) | |
1956 | * comparing uuids when re-adding a device into an array | |
51006d85 N |
1957 | * In these cases the uuid required is that of the data-array, |
1958 | * not the device-set. | |
1959 | * uuid to recognise same set when adding a missing device back | |
1960 | * to an array. This is a uuid for the device-set. | |
613b0d17 | 1961 | * |
a322f70c DW |
1962 | * For each of these we can make do with a truncated |
1963 | * or hashed uuid rather than the original, as long as | |
1964 | * everyone agrees. | |
a322f70c DW |
1965 | * In the case of SVD we assume the BVD is of interest, |
1966 | * though that might be the case if a bitmap were made for | |
1967 | * a mirrored SVD - worry about that later. | |
1968 | * So we need to find the VD configuration record for the | |
1969 | * relevant BVD and extract the GUID and Secondary_Element_Seq. | |
1970 | * The first 16 bytes of the sha1 of these is used. | |
1971 | */ | |
1972 | struct ddf_super *ddf = st->sb; | |
d2ca6449 | 1973 | struct vcl *vcl = ddf->currentconf; |
a322f70c | 1974 | |
c5afc314 | 1975 | if (vcl) |
7087f02b | 1976 | uuid_of_ddf_subarray(ddf, vcl->vcnum, uuid); |
c5afc314 | 1977 | else |
7087f02b | 1978 | uuid_from_ddf_guid(ddf->anchor.guid, uuid); |
a322f70c DW |
1979 | } |
1980 | ||
a5d85af7 | 1981 | static void getinfo_super_ddf(struct supertype *st, struct mdinfo *info, char *map) |
a322f70c DW |
1982 | { |
1983 | struct ddf_super *ddf = st->sb; | |
a5d85af7 | 1984 | int map_disks = info->array.raid_disks; |
90fa1a29 | 1985 | __u32 *cptr; |
a322f70c | 1986 | |
78e44928 | 1987 | if (ddf->currentconf) { |
a5d85af7 | 1988 | getinfo_super_ddf_bvd(st, info, map); |
78e44928 NB |
1989 | return; |
1990 | } | |
95eeceeb | 1991 | memset(info, 0, sizeof(*info)); |
78e44928 | 1992 | |
a8173e43 | 1993 | info->array.raid_disks = be16_to_cpu(ddf->phys->used_pdes); |
a322f70c DW |
1994 | info->array.level = LEVEL_CONTAINER; |
1995 | info->array.layout = 0; | |
1996 | info->array.md_minor = -1; | |
90fa1a29 JS |
1997 | cptr = (__u32 *)(ddf->anchor.guid + 16); |
1998 | info->array.ctime = DECADE + __be32_to_cpu(*cptr); | |
1999 | ||
a322f70c | 2000 | info->array.chunk_size = 0; |
510242aa | 2001 | info->container_enough = 1; |
a322f70c | 2002 | |
56cb05c4 N |
2003 | info->disk.major = 0; |
2004 | info->disk.minor = 0; | |
cba0191b | 2005 | if (ddf->dlist) { |
f0e876ce | 2006 | struct phys_disk_entry *pde = NULL; |
60931cf9 | 2007 | info->disk.number = be32_to_cpu(ddf->dlist->disk.refnum); |
59e36268 | 2008 | info->disk.raid_disk = find_phys(ddf, ddf->dlist->disk.refnum); |
d2ca6449 | 2009 | |
9d0c6b70 | 2010 | info->data_offset = be64_to_cpu(ddf->phys-> |
613b0d17 N |
2011 | entries[info->disk.raid_disk]. |
2012 | config_size); | |
d2ca6449 | 2013 | info->component_size = ddf->dlist->size - info->data_offset; |
f0e876ce N |
2014 | if (info->disk.raid_disk >= 0) |
2015 | pde = ddf->phys->entries + info->disk.raid_disk; | |
2016 | if (pde && | |
4fe903aa N |
2017 | !(be16_to_cpu(pde->state) & DDF_Failed) && |
2018 | !(be16_to_cpu(pde->state) & DDF_Missing)) | |
f0e876ce N |
2019 | info->disk.state = (1 << MD_DISK_SYNC) | (1 << MD_DISK_ACTIVE); |
2020 | else | |
2021 | info->disk.state = 1 << MD_DISK_FAULTY; | |
eba2859f | 2022 | |
cba0191b | 2023 | } else { |
25532b88 | 2024 | /* There should always be a dlist, but just in case...*/ |
cba0191b | 2025 | info->disk.number = -1; |
661dce36 | 2026 | info->disk.raid_disk = -1; |
f0e876ce | 2027 | info->disk.state = (1 << MD_DISK_SYNC) | (1 << MD_DISK_ACTIVE); |
cba0191b | 2028 | } |
2a351cea N |
2029 | info->events = be32_to_cpu(ddf->active->seq); |
2030 | info->array.utime = DECADE + be32_to_cpu(ddf->active->timestamp); | |
a19c88b8 | 2031 | |
921d9e16 | 2032 | info->recovery_start = MaxSector; |
a19c88b8 | 2033 | info->reshape_active = 0; |
6e75048b | 2034 | info->recovery_blocked = 0; |
c5afc314 | 2035 | info->name[0] = 0; |
a322f70c | 2036 | |
f35f2525 N |
2037 | info->array.major_version = -1; |
2038 | info->array.minor_version = -2; | |
159c3a1a | 2039 | strcpy(info->text_version, "ddf"); |
a67dd8cc | 2040 | info->safe_mode_delay = 0; |
159c3a1a | 2041 | |
c5afc314 | 2042 | uuid_from_super_ddf(st, info->uuid); |
a322f70c | 2043 | |
a5d85af7 | 2044 | if (map) { |
708997ff N |
2045 | int i, e = 0; |
2046 | int max = be16_to_cpu(ddf->phys->max_pdes); | |
2047 | for (i = e = 0 ; i < map_disks ; i++, e++) { | |
2048 | while (e < max && | |
2049 | be32_to_cpu(ddf->phys->entries[e].refnum) == 0xffffffff) | |
2050 | e++; | |
2051 | if (i < info->array.raid_disks && e < max && | |
2052 | !(be16_to_cpu(ddf->phys->entries[e].state) | |
a8173e43 | 2053 | & DDF_Failed)) |
a5d85af7 N |
2054 | map[i] = 1; |
2055 | else | |
2056 | map[i] = 0; | |
2057 | } | |
2058 | } | |
a322f70c DW |
2059 | } |
2060 | ||
8bf989d8 | 2061 | /* size of name must be at least 17 bytes! */ |
2062 | static void _ddf_array_name(char *name, const struct ddf_super *ddf, int i) | |
2063 | { | |
2064 | int j; | |
2065 | memcpy(name, ddf->virt->entries[i].name, 16); | |
2066 | name[16] = 0; | |
2067 | for(j = 0; j < 16; j++) | |
2068 | if (name[j] == ' ') | |
2069 | name[j] = 0; | |
2070 | } | |
2071 | ||
a5d85af7 | 2072 | static void getinfo_super_ddf_bvd(struct supertype *st, struct mdinfo *info, char *map) |
a322f70c DW |
2073 | { |
2074 | struct ddf_super *ddf = st->sb; | |
d2ca6449 NB |
2075 | struct vcl *vc = ddf->currentconf; |
2076 | int cd = ddf->currentdev; | |
ddf94a43 | 2077 | int n_prim; |
db42fa9b | 2078 | int j; |
8592f29d | 2079 | struct dl *dl; |
a5d85af7 | 2080 | int map_disks = info->array.raid_disks; |
90fa1a29 | 2081 | __u32 *cptr; |
ddf94a43 | 2082 | struct vd_config *conf; |
a322f70c | 2083 | |
95eeceeb | 2084 | memset(info, 0, sizeof(*info)); |
8a2848a7 | 2085 | if (layout_ddf2md(&vc->conf, &info->array) == -1) |
2086 | return; | |
a322f70c | 2087 | info->array.md_minor = -1; |
90fa1a29 JS |
2088 | cptr = (__u32 *)(vc->conf.guid + 16); |
2089 | info->array.ctime = DECADE + __be32_to_cpu(*cptr); | |
60931cf9 | 2090 | info->array.utime = DECADE + be32_to_cpu(vc->conf.timestamp); |
d2ca6449 | 2091 | info->array.chunk_size = 512 << vc->conf.chunk_shift; |
55e00074 | 2092 | info->custom_array_size = be64_to_cpu(vc->conf.array_blocks); |
d2ca6449 | 2093 | |
ddf94a43 | 2094 | conf = &vc->conf; |
a8173e43 | 2095 | n_prim = be16_to_cpu(conf->prim_elmnt_count); |
ddf94a43 | 2096 | if (conf->sec_elmnt_count > 1 && cd >= n_prim) { |
2097 | int ibvd = cd / n_prim - 1; | |
2098 | cd %= n_prim; | |
2099 | conf = vc->other_bvds[ibvd]; | |
2100 | } | |
2101 | ||
f21e18ca | 2102 | if (cd >= 0 && (unsigned)cd < ddf->mppe) { |
57a66662 | 2103 | info->data_offset = |
9d0c6b70 | 2104 | be64_to_cpu(LBA_OFFSET(ddf, conf)[cd]); |
d2ca6449 NB |
2105 | if (vc->block_sizes) |
2106 | info->component_size = vc->block_sizes[cd]; | |
2107 | else | |
9d0c6b70 | 2108 | info->component_size = be64_to_cpu(conf->blocks); |
a322f70c | 2109 | |
0d255ff8 N |
2110 | for (dl = ddf->dlist; dl ; dl = dl->next) |
2111 | if (be32_eq(dl->disk.refnum, conf->phys_refnum[cd])) | |
2112 | break; | |
2113 | } | |
fb204fb2 | 2114 | |
a322f70c DW |
2115 | info->disk.major = 0; |
2116 | info->disk.minor = 0; | |
fb204fb2 | 2117 | info->disk.state = 0; |
41bcbc14 | 2118 | if (dl && dl->pdnum >= 0) { |
8592f29d N |
2119 | info->disk.major = dl->major; |
2120 | info->disk.minor = dl->minor; | |
7c3fb3ec | 2121 | info->disk.raid_disk = cd + conf->sec_elmnt_seq |
a8173e43 | 2122 | * be16_to_cpu(conf->prim_elmnt_count); |
fb204fb2 | 2123 | info->disk.number = dl->pdnum; |
f0e876ce N |
2124 | info->disk.state = 0; |
2125 | if (info->disk.number >= 0 && | |
2126 | (be16_to_cpu(ddf->phys->entries[info->disk.number].state) & DDF_Online) && | |
2127 | !(be16_to_cpu(ddf->phys->entries[info->disk.number].state) & DDF_Failed)) | |
2128 | info->disk.state = (1<<MD_DISK_SYNC)|(1<<MD_DISK_ACTIVE); | |
eba2859f | 2129 | info->events = be32_to_cpu(ddf->active->seq); |
8592f29d | 2130 | } |
a322f70c | 2131 | |
103f2410 NB |
2132 | info->container_member = ddf->currentconf->vcnum; |
2133 | ||
921d9e16 | 2134 | info->recovery_start = MaxSector; |
80d26cb2 | 2135 | info->resync_start = 0; |
624c5ad4 | 2136 | info->reshape_active = 0; |
6e75048b | 2137 | info->recovery_blocked = 0; |
80d26cb2 NB |
2138 | if (!(ddf->virt->entries[info->container_member].state |
2139 | & DDF_state_inconsistent) && | |
2140 | (ddf->virt->entries[info->container_member].init_state | |
2141 | & DDF_initstate_mask) | |
2142 | == DDF_init_full) | |
b7528a20 | 2143 | info->resync_start = MaxSector; |
80d26cb2 | 2144 | |
a322f70c DW |
2145 | uuid_from_super_ddf(st, info->uuid); |
2146 | ||
f35f2525 N |
2147 | info->array.major_version = -1; |
2148 | info->array.minor_version = -2; | |
9b63e648 | 2149 | sprintf(info->text_version, "/%s/%d", |
4dd2df09 | 2150 | st->container_devnm, |
9b63e648 | 2151 | info->container_member); |
5684fff6 | 2152 | info->safe_mode_delay = DDF_SAFE_MODE_DELAY; |
159c3a1a | 2153 | |
8bf989d8 | 2154 | _ddf_array_name(info->name, ddf, info->container_member); |
a5d85af7 N |
2155 | |
2156 | if (map) | |
2157 | for (j = 0; j < map_disks; j++) { | |
2158 | map[j] = 0; | |
2159 | if (j < info->array.raid_disks) { | |
2160 | int i = find_phys(ddf, vc->conf.phys_refnum[j]); | |
613b0d17 | 2161 | if (i >= 0 && |
a8173e43 | 2162 | (be16_to_cpu(ddf->phys->entries[i].state) |
2163 | & DDF_Online) && | |
2164 | !(be16_to_cpu(ddf->phys->entries[i].state) | |
2165 | & DDF_Failed)) | |
a5d85af7 N |
2166 | map[i] = 1; |
2167 | } | |
2168 | } | |
a322f70c DW |
2169 | } |
2170 | ||
2171 | static int update_super_ddf(struct supertype *st, struct mdinfo *info, | |
2172 | char *update, | |
2173 | char *devname, int verbose, | |
2174 | int uuid_set, char *homehost) | |
2175 | { | |
2176 | /* For 'assemble' and 'force' we need to return non-zero if any | |
2177 | * change was made. For others, the return value is ignored. | |
2178 | * Update options are: | |
2179 | * force-one : This device looks a bit old but needs to be included, | |
2180 | * update age info appropriately. | |
2181 | * assemble: clear any 'faulty' flag to allow this device to | |
2182 | * be assembled. | |
2183 | * force-array: Array is degraded but being forced, mark it clean | |
2184 | * if that will be needed to assemble it. | |
2185 | * | |
2186 | * newdev: not used ???? | |
2187 | * grow: Array has gained a new device - this is currently for | |
2188 | * linear only | |
2189 | * resync: mark as dirty so a resync will happen. | |
59e36268 | 2190 | * uuid: Change the uuid of the array to match what is given |
a322f70c DW |
2191 | * homehost: update the recorded homehost |
2192 | * name: update the name - preserving the homehost | |
2193 | * _reshape_progress: record new reshape_progress position. | |
2194 | * | |
2195 | * Following are not relevant for this version: | |
2196 | * sparc2.2 : update from old dodgey metadata | |
2197 | * super-minor: change the preferred_minor number | |
2198 | * summaries: update redundant counters. | |
2199 | */ | |
2200 | int rv = 0; | |
2201 | // struct ddf_super *ddf = st->sb; | |
7a7cc504 | 2202 | // struct vd_config *vd = find_vdcr(ddf, info->container_member); |
a322f70c DW |
2203 | // struct virtual_entry *ve = find_ve(ddf); |
2204 | ||
a322f70c | 2205 | /* we don't need to handle "force-*" or "assemble" as |
56cb05c4 | 2206 | * there is no need to 'trick' the kernel. When the metadata is |
a322f70c DW |
2207 | * first updated to activate the array, all the implied modifications |
2208 | * will just happen. | |
2209 | */ | |
2210 | ||
2211 | if (strcmp(update, "grow") == 0) { | |
2212 | /* FIXME */ | |
1e2b2765 | 2213 | } else if (strcmp(update, "resync") == 0) { |
a322f70c | 2214 | // info->resync_checkpoint = 0; |
1e2b2765 | 2215 | } else if (strcmp(update, "homehost") == 0) { |
a322f70c DW |
2216 | /* homehost is stored in controller->vendor_data, |
2217 | * or it is when we are the vendor | |
2218 | */ | |
2219 | // if (info->vendor_is_local) | |
2220 | // strcpy(ddf->controller.vendor_data, homehost); | |
1e2b2765 | 2221 | rv = -1; |
f49208ec | 2222 | } else if (strcmp(update, "name") == 0) { |
a322f70c DW |
2223 | /* name is stored in virtual_entry->name */ |
2224 | // memset(ve->name, ' ', 16); | |
2225 | // strncpy(ve->name, info->name, 16); | |
1e2b2765 | 2226 | rv = -1; |
f49208ec | 2227 | } else if (strcmp(update, "_reshape_progress") == 0) { |
a322f70c | 2228 | /* We don't support reshape yet */ |
f49208ec N |
2229 | } else if (strcmp(update, "assemble") == 0 ) { |
2230 | /* Do nothing, just succeed */ | |
2231 | rv = 0; | |
1e2b2765 N |
2232 | } else |
2233 | rv = -1; | |
a322f70c DW |
2234 | |
2235 | // update_all_csum(ddf); | |
2236 | ||
2237 | return rv; | |
2238 | } | |
2239 | ||
5f8097be NB |
2240 | static void make_header_guid(char *guid) |
2241 | { | |
60931cf9 | 2242 | be32 stamp; |
5f8097be NB |
2243 | /* Create a DDF Header of Virtual Disk GUID */ |
2244 | ||
2245 | /* 24 bytes of fiction required. | |
2246 | * first 8 are a 'vendor-id' - "Linux-MD" | |
2247 | * next 8 are controller type.. how about 0X DEAD BEEF 0000 0000 | |
2248 | * Remaining 8 random number plus timestamp | |
2249 | */ | |
2250 | memcpy(guid, T10, sizeof(T10)); | |
60931cf9 | 2251 | stamp = cpu_to_be32(0xdeadbeef); |
5f8097be | 2252 | memcpy(guid+8, &stamp, 4); |
60931cf9 | 2253 | stamp = cpu_to_be32(0); |
5f8097be | 2254 | memcpy(guid+12, &stamp, 4); |
60931cf9 | 2255 | stamp = cpu_to_be32(time(0) - DECADE); |
5f8097be | 2256 | memcpy(guid+16, &stamp, 4); |
60931cf9 | 2257 | stamp._v32 = random32(); |
5f8097be | 2258 | memcpy(guid+20, &stamp, 4); |
5f8097be | 2259 | } |
59e36268 | 2260 | |
fb9d0acb | 2261 | static unsigned int find_unused_vde(const struct ddf_super *ddf) |
2262 | { | |
2263 | unsigned int i; | |
a8173e43 | 2264 | for (i = 0; i < be16_to_cpu(ddf->virt->max_vdes); i++) { |
fb9d0acb | 2265 | if (all_ff(ddf->virt->entries[i].guid)) |
2266 | return i; | |
2267 | } | |
2268 | return DDF_NOTFOUND; | |
2269 | } | |
2270 | ||
2271 | static unsigned int find_vde_by_name(const struct ddf_super *ddf, | |
2272 | const char *name) | |
2273 | { | |
2274 | unsigned int i; | |
2275 | if (name == NULL) | |
2276 | return DDF_NOTFOUND; | |
a8173e43 | 2277 | for (i = 0; i < be16_to_cpu(ddf->virt->max_vdes); i++) { |
fb9d0acb | 2278 | if (all_ff(ddf->virt->entries[i].guid)) |
2279 | continue; | |
2280 | if (!strncmp(name, ddf->virt->entries[i].name, | |
2281 | sizeof(ddf->virt->entries[i].name))) | |
2282 | return i; | |
2283 | } | |
2284 | return DDF_NOTFOUND; | |
2285 | } | |
2286 | ||
4441541f | 2287 | #ifndef MDASSEMBLE |
fb9d0acb | 2288 | static unsigned int find_vde_by_guid(const struct ddf_super *ddf, |
2289 | const char *guid) | |
2290 | { | |
2291 | unsigned int i; | |
2292 | if (guid == NULL || all_ff(guid)) | |
2293 | return DDF_NOTFOUND; | |
a8173e43 | 2294 | for (i = 0; i < be16_to_cpu(ddf->virt->max_vdes); i++) |
fb9d0acb | 2295 | if (!memcmp(ddf->virt->entries[i].guid, guid, DDF_GUID_LEN)) |
2296 | return i; | |
2297 | return DDF_NOTFOUND; | |
2298 | } | |
4441541f | 2299 | #endif |
fb9d0acb | 2300 | |
a322f70c DW |
2301 | static int init_super_ddf(struct supertype *st, |
2302 | mdu_array_info_t *info, | |
2303 | unsigned long long size, char *name, char *homehost, | |
83cd1e97 | 2304 | int *uuid, unsigned long long data_offset) |
a322f70c DW |
2305 | { |
2306 | /* This is primarily called by Create when creating a new array. | |
2307 | * We will then get add_to_super called for each component, and then | |
2308 | * write_init_super called to write it out to each device. | |
2309 | * For DDF, Create can create on fresh devices or on a pre-existing | |
2310 | * array. | |
2311 | * To create on a pre-existing array a different method will be called. | |
2312 | * This one is just for fresh drives. | |
2313 | * | |
2314 | * We need to create the entire 'ddf' structure which includes: | |
2315 | * DDF headers - these are easy. | |
2316 | * Controller data - a Sector describing this controller .. not that | |
56cb05c4 | 2317 | * this is a controller exactly. |
a322f70c | 2318 | * Physical Disk Record - one entry per device, so |
56cb05c4 | 2319 | * leave plenty of space. |
a322f70c | 2320 | * Virtual Disk Records - again, just leave plenty of space. |
56cb05c4 N |
2321 | * This just lists VDs, doesn't give details. |
2322 | * Config records - describe the VDs that use this disk | |
a322f70c DW |
2323 | * DiskData - describes 'this' device. |
2324 | * BadBlockManagement - empty | |
2325 | * Diag Space - empty | |
2326 | * Vendor Logs - Could we put bitmaps here? | |
2327 | * | |
2328 | */ | |
2329 | struct ddf_super *ddf; | |
2330 | char hostname[17]; | |
2331 | int hostlen; | |
a322f70c DW |
2332 | int max_phys_disks, max_virt_disks; |
2333 | unsigned long long sector; | |
2334 | int clen; | |
2335 | int i; | |
2336 | int pdsize, vdsize; | |
2337 | struct phys_disk *pd; | |
2338 | struct virtual_disk *vd; | |
2339 | ||
83cd1e97 | 2340 | if (data_offset != INVALID_SECTORS) { |
ed503f89 | 2341 | pr_err("data-offset not supported by DDF\n"); |
83cd1e97 N |
2342 | return 0; |
2343 | } | |
2344 | ||
78e44928 | 2345 | if (st->sb) |
83cd1e97 N |
2346 | return init_super_ddf_bvd(st, info, size, name, homehost, uuid, |
2347 | data_offset); | |
ba7eb04f | 2348 | |
3d2c4fc7 | 2349 | if (posix_memalign((void**)&ddf, 512, sizeof(*ddf)) != 0) { |
e7b84f9d | 2350 | pr_err("%s could not allocate superblock\n", __func__); |
3d2c4fc7 DW |
2351 | return 0; |
2352 | } | |
6264b437 | 2353 | memset(ddf, 0, sizeof(*ddf)); |
955e9ea1 DW |
2354 | st->sb = ddf; |
2355 | ||
2356 | if (info == NULL) { | |
2357 | /* zeroing superblock */ | |
2358 | return 0; | |
2359 | } | |
a322f70c DW |
2360 | |
2361 | /* At least 32MB *must* be reserved for the ddf. So let's just | |
2362 | * start 32MB from the end, and put the primary header there. | |
2363 | * Don't do secondary for now. | |
2364 | * We don't know exactly where that will be yet as it could be | |
56cb05c4 | 2365 | * different on each device. So just set up the lengths. |
a322f70c DW |
2366 | */ |
2367 | ||
2368 | ddf->anchor.magic = DDF_HEADER_MAGIC; | |
5f8097be | 2369 | make_header_guid(ddf->anchor.guid); |
a322f70c | 2370 | |
59e36268 | 2371 | memcpy(ddf->anchor.revision, DDF_REVISION_2, 8); |
60931cf9 | 2372 | ddf->anchor.seq = cpu_to_be32(1); |
2373 | ddf->anchor.timestamp = cpu_to_be32(time(0) - DECADE); | |
a322f70c DW |
2374 | ddf->anchor.openflag = 0xFF; |
2375 | ddf->anchor.foreignflag = 0; | |
2376 | ddf->anchor.enforcegroups = 0; /* Is this best?? */ | |
2377 | ddf->anchor.pad0 = 0xff; | |
2378 | memset(ddf->anchor.pad1, 0xff, 12); | |
2379 | memset(ddf->anchor.header_ext, 0xff, 32); | |
9d0c6b70 | 2380 | ddf->anchor.primary_lba = cpu_to_be64(~(__u64)0); |
2381 | ddf->anchor.secondary_lba = cpu_to_be64(~(__u64)0); | |
a322f70c DW |
2382 | ddf->anchor.type = DDF_HEADER_ANCHOR; |
2383 | memset(ddf->anchor.pad2, 0xff, 3); | |
60931cf9 | 2384 | ddf->anchor.workspace_len = cpu_to_be32(32768); /* Must be reserved */ |
9d0c6b70 | 2385 | /* Put this at bottom of 32M reserved.. */ |
2386 | ddf->anchor.workspace_lba = cpu_to_be64(~(__u64)0); | |
56cb05c4 | 2387 | max_phys_disks = 1023; /* Should be enough, 4095 is also allowed */ |
a8173e43 | 2388 | ddf->anchor.max_pd_entries = cpu_to_be16(max_phys_disks); |
56cb05c4 N |
2389 | max_virt_disks = 255; /* 15, 63, 255, 1024, 4095 are all allowed */ |
2390 | ddf->anchor.max_vd_entries = cpu_to_be16(max_virt_disks); | |
a322f70c | 2391 | ddf->max_part = 64; |
56cb05c4 N |
2392 | ddf->anchor.max_partitions = cpu_to_be16(ddf->max_part); |
2393 | ddf->mppe = 256; /* 16, 64, 256, 1024, 4096 are all allowed */ | |
59e36268 | 2394 | ddf->conf_rec_len = 1 + ROUND_UP(ddf->mppe * (4+8), 512)/512; |
a8173e43 | 2395 | ddf->anchor.config_record_len = cpu_to_be16(ddf->conf_rec_len); |
2396 | ddf->anchor.max_primary_element_entries = cpu_to_be16(ddf->mppe); | |
a322f70c | 2397 | memset(ddf->anchor.pad3, 0xff, 54); |
56cb05c4 | 2398 | /* Controller section is one sector long immediately |
a322f70c DW |
2399 | * after the ddf header */ |
2400 | sector = 1; | |
60931cf9 | 2401 | ddf->anchor.controller_section_offset = cpu_to_be32(sector); |
2402 | ddf->anchor.controller_section_length = cpu_to_be32(1); | |
a322f70c DW |
2403 | sector += 1; |
2404 | ||
2405 | /* phys is 8 sectors after that */ | |
2406 | pdsize = ROUND_UP(sizeof(struct phys_disk) + | |
2407 | sizeof(struct phys_disk_entry)*max_phys_disks, | |
2408 | 512); | |
2409 | switch(pdsize/512) { | |
2410 | case 2: case 8: case 32: case 128: case 512: break; | |
2411 | default: abort(); | |
2412 | } | |
60931cf9 | 2413 | ddf->anchor.phys_section_offset = cpu_to_be32(sector); |
a322f70c | 2414 | ddf->anchor.phys_section_length = |
60931cf9 | 2415 | cpu_to_be32(pdsize/512); /* max_primary_element_entries/8 */ |
a322f70c DW |
2416 | sector += pdsize/512; |
2417 | ||
2418 | /* virt is another 32 sectors */ | |
2419 | vdsize = ROUND_UP(sizeof(struct virtual_disk) + | |
2420 | sizeof(struct virtual_entry) * max_virt_disks, | |
2421 | 512); | |
2422 | switch(vdsize/512) { | |
2423 | case 2: case 8: case 32: case 128: case 512: break; | |
2424 | default: abort(); | |
2425 | } | |
60931cf9 | 2426 | ddf->anchor.virt_section_offset = cpu_to_be32(sector); |
a322f70c | 2427 | ddf->anchor.virt_section_length = |
60931cf9 | 2428 | cpu_to_be32(vdsize/512); /* max_vd_entries/8 */ |
a322f70c DW |
2429 | sector += vdsize/512; |
2430 | ||
59e36268 | 2431 | clen = ddf->conf_rec_len * (ddf->max_part+1); |
60931cf9 | 2432 | ddf->anchor.config_section_offset = cpu_to_be32(sector); |
2433 | ddf->anchor.config_section_length = cpu_to_be32(clen); | |
a322f70c DW |
2434 | sector += clen; |
2435 | ||
60931cf9 | 2436 | ddf->anchor.data_section_offset = cpu_to_be32(sector); |
2437 | ddf->anchor.data_section_length = cpu_to_be32(1); | |
a322f70c DW |
2438 | sector += 1; |
2439 | ||
60931cf9 | 2440 | ddf->anchor.bbm_section_length = cpu_to_be32(0); |
2441 | ddf->anchor.bbm_section_offset = cpu_to_be32(0xFFFFFFFF); | |
2442 | ddf->anchor.diag_space_length = cpu_to_be32(0); | |
2443 | ddf->anchor.diag_space_offset = cpu_to_be32(0xFFFFFFFF); | |
2444 | ddf->anchor.vendor_length = cpu_to_be32(0); | |
2445 | ddf->anchor.vendor_offset = cpu_to_be32(0xFFFFFFFF); | |
a322f70c DW |
2446 | |
2447 | memset(ddf->anchor.pad4, 0xff, 256); | |
2448 | ||
2449 | memcpy(&ddf->primary, &ddf->anchor, 512); | |
2450 | memcpy(&ddf->secondary, &ddf->anchor, 512); | |
2451 | ||
2452 | ddf->primary.openflag = 1; /* I guess.. */ | |
2453 | ddf->primary.type = DDF_HEADER_PRIMARY; | |
2454 | ||
2455 | ddf->secondary.openflag = 1; /* I guess.. */ | |
2456 | ddf->secondary.type = DDF_HEADER_SECONDARY; | |
2457 | ||
2458 | ddf->active = &ddf->primary; | |
2459 | ||
2460 | ddf->controller.magic = DDF_CONTROLLER_MAGIC; | |
2461 | ||
2462 | /* 24 more bytes of fiction required. | |
2463 | * first 8 are a 'vendor-id' - "Linux-MD" | |
2464 | * Remaining 16 are serial number.... maybe a hostname would do? | |
2465 | */ | |
2466 | memcpy(ddf->controller.guid, T10, sizeof(T10)); | |
1ba6bff9 DW |
2467 | gethostname(hostname, sizeof(hostname)); |
2468 | hostname[sizeof(hostname) - 1] = 0; | |
a322f70c DW |
2469 | hostlen = strlen(hostname); |
2470 | memcpy(ddf->controller.guid + 24 - hostlen, hostname, hostlen); | |
2471 | for (i = strlen(T10) ; i+hostlen < 24; i++) | |
2472 | ddf->controller.guid[i] = ' '; | |
2473 | ||
a8173e43 | 2474 | ddf->controller.type.vendor_id = cpu_to_be16(0xDEAD); |
2475 | ddf->controller.type.device_id = cpu_to_be16(0xBEEF); | |
2476 | ddf->controller.type.sub_vendor_id = cpu_to_be16(0); | |
2477 | ddf->controller.type.sub_device_id = cpu_to_be16(0); | |
a322f70c DW |
2478 | memcpy(ddf->controller.product_id, "What Is My PID??", 16); |
2479 | memset(ddf->controller.pad, 0xff, 8); | |
2480 | memset(ddf->controller.vendor_data, 0xff, 448); | |
a9e1c11d N |
2481 | if (homehost && strlen(homehost) < 440) |
2482 | strcpy((char*)ddf->controller.vendor_data, homehost); | |
a322f70c | 2483 | |
3d2c4fc7 | 2484 | if (posix_memalign((void**)&pd, 512, pdsize) != 0) { |
e7b84f9d | 2485 | pr_err("%s could not allocate pd\n", __func__); |
3d2c4fc7 DW |
2486 | return 0; |
2487 | } | |
6416d527 | 2488 | ddf->phys = pd; |
a322f70c DW |
2489 | ddf->pdsize = pdsize; |
2490 | ||
2491 | memset(pd, 0xff, pdsize); | |
2492 | memset(pd, 0, sizeof(*pd)); | |
076515ba | 2493 | pd->magic = DDF_PHYS_RECORDS_MAGIC; |
a8173e43 | 2494 | pd->used_pdes = cpu_to_be16(0); |
2495 | pd->max_pdes = cpu_to_be16(max_phys_disks); | |
a322f70c | 2496 | memset(pd->pad, 0xff, 52); |
4a3ca8ac | 2497 | for (i = 0; i < max_phys_disks; i++) |
2498 | memset(pd->entries[i].guid, 0xff, DDF_GUID_LEN); | |
a322f70c | 2499 | |
3d2c4fc7 | 2500 | if (posix_memalign((void**)&vd, 512, vdsize) != 0) { |
e7b84f9d | 2501 | pr_err("%s could not allocate vd\n", __func__); |
3d2c4fc7 DW |
2502 | return 0; |
2503 | } | |
6416d527 | 2504 | ddf->virt = vd; |
a322f70c DW |
2505 | ddf->vdsize = vdsize; |
2506 | memset(vd, 0, vdsize); | |
2507 | vd->magic = DDF_VIRT_RECORDS_MAGIC; | |
a8173e43 | 2508 | vd->populated_vdes = cpu_to_be16(0); |
2509 | vd->max_vdes = cpu_to_be16(max_virt_disks); | |
a322f70c DW |
2510 | memset(vd->pad, 0xff, 52); |
2511 | ||
5f8097be NB |
2512 | for (i=0; i<max_virt_disks; i++) |
2513 | memset(&vd->entries[i], 0xff, sizeof(struct virtual_entry)); | |
2514 | ||
a322f70c | 2515 | st->sb = ddf; |
609ce161 | 2516 | ddf_set_updates_pending(ddf, NULL); |
a322f70c DW |
2517 | return 1; |
2518 | } | |
2519 | ||
5f8097be NB |
2520 | static int chunk_to_shift(int chunksize) |
2521 | { | |
2522 | return ffs(chunksize/512)-1; | |
2523 | } | |
2524 | ||
0e600426 | 2525 | #ifndef MDASSEMBLE |
59e36268 NB |
2526 | struct extent { |
2527 | unsigned long long start, size; | |
2528 | }; | |
78e44928 | 2529 | static int cmp_extent(const void *av, const void *bv) |
59e36268 NB |
2530 | { |
2531 | const struct extent *a = av; | |
2532 | const struct extent *b = bv; | |
2533 | if (a->start < b->start) | |
2534 | return -1; | |
2535 | if (a->start > b->start) | |
2536 | return 1; | |
2537 | return 0; | |
2538 | } | |
2539 | ||
78e44928 | 2540 | static struct extent *get_extents(struct ddf_super *ddf, struct dl *dl) |
59e36268 | 2541 | { |
fca65520 | 2542 | /* Find a list of used extents on the given physical device |
59e36268 NB |
2543 | * (dnum) of the given ddf. |
2544 | * Return a malloced array of 'struct extent' | |
59e36268 NB |
2545 | */ |
2546 | struct extent *rv; | |
2547 | int n = 0; | |
fcc22180 | 2548 | unsigned int i; |
a44e993e N |
2549 | __u16 state; |
2550 | ||
2551 | if (dl->pdnum < 0) | |
2552 | return NULL; | |
2553 | state = be16_to_cpu(ddf->phys->entries[dl->pdnum].state); | |
60056e1c | 2554 | |
2555 | if ((state & (DDF_Online|DDF_Failed|DDF_Missing)) != DDF_Online) | |
2556 | return NULL; | |
59e36268 | 2557 | |
503975b9 | 2558 | rv = xmalloc(sizeof(struct extent) * (ddf->max_part + 2)); |
59e36268 NB |
2559 | |
2560 | for (i = 0; i < ddf->max_part; i++) { | |
fcc22180 | 2561 | const struct vd_config *bvd; |
2562 | unsigned int ibvd; | |
59e36268 | 2563 | struct vcl *v = dl->vlist[i]; |
fcc22180 | 2564 | if (v == NULL || |
2565 | get_pd_index_from_refnum(v, dl->disk.refnum, ddf->mppe, | |
2566 | &bvd, &ibvd) == DDF_NOTFOUND) | |
59e36268 | 2567 | continue; |
9d0c6b70 | 2568 | rv[n].start = be64_to_cpu(LBA_OFFSET(ddf, bvd)[ibvd]); |
2569 | rv[n].size = be64_to_cpu(bvd->blocks); | |
fcc22180 | 2570 | n++; |
59e36268 NB |
2571 | } |
2572 | qsort(rv, n, sizeof(*rv), cmp_extent); | |
2573 | ||
9d0c6b70 | 2574 | rv[n].start = be64_to_cpu(ddf->phys->entries[dl->pdnum].config_size); |
59e36268 NB |
2575 | rv[n].size = 0; |
2576 | return rv; | |
2577 | } | |
fca65520 N |
2578 | |
2579 | static unsigned long long find_space( | |
2580 | struct ddf_super *ddf, struct dl *dl, | |
2581 | unsigned long long data_offset, | |
2582 | unsigned long long *size) | |
2583 | { | |
2584 | /* Find if the requested amount of space is available. | |
2585 | * If it is, return start. | |
2586 | * If not, set *size to largest space. | |
2587 | * If data_offset != INVALID_SECTORS, then the space must start | |
2588 | * at this location. | |
2589 | */ | |
2590 | struct extent *e = get_extents(ddf, dl); | |
2591 | int i = 0; | |
2592 | unsigned long long pos = 0; | |
2593 | unsigned long long max_size = 0; | |
2594 | ||
2595 | if (!e) { | |
2596 | *size = 0; | |
2597 | return INVALID_SECTORS; | |
2598 | } | |
2599 | do { | |
2600 | unsigned long long esize = e[i].start - pos; | |
2601 | if (data_offset != INVALID_SECTORS && | |
2602 | pos <= data_offset && | |
2603 | e[i].start > data_offset) { | |
2604 | pos = data_offset; | |
2605 | esize = e[i].start - pos; | |
2606 | } | |
2607 | if (data_offset != INVALID_SECTORS && | |
2608 | pos != data_offset) { | |
2609 | i++; | |
2610 | continue; | |
2611 | } | |
2612 | if (esize >= *size) { | |
2613 | /* Found! */ | |
2614 | free(e); | |
2615 | return pos; | |
2616 | } | |
2617 | if (esize > max_size) | |
2618 | max_size = esize; | |
2619 | pos = e[i].start + e[i].size; | |
2620 | i++; | |
2621 | } while (e[i-1].size); | |
2622 | *size = max_size; | |
2623 | free(e); | |
2624 | return INVALID_SECTORS; | |
2625 | } | |
0e600426 | 2626 | #endif |
59e36268 | 2627 | |
5f8097be NB |
2628 | static int init_super_ddf_bvd(struct supertype *st, |
2629 | mdu_array_info_t *info, | |
2630 | unsigned long long size, | |
2631 | char *name, char *homehost, | |
83cd1e97 | 2632 | int *uuid, unsigned long long data_offset) |
5f8097be NB |
2633 | { |
2634 | /* We are creating a BVD inside a pre-existing container. | |
2635 | * so st->sb is already set. | |
2636 | * We need to create a new vd_config and a new virtual_entry | |
2637 | */ | |
2638 | struct ddf_super *ddf = st->sb; | |
5aaf6c7b | 2639 | unsigned int venum, i; |
5f8097be NB |
2640 | struct virtual_entry *ve; |
2641 | struct vcl *vcl; | |
2642 | struct vd_config *vc; | |
5f8097be | 2643 | |
fb9d0acb | 2644 | if (find_vde_by_name(ddf, name) != DDF_NOTFOUND) { |
2645 | pr_err("This ddf already has an array called %s\n", name); | |
5f8097be NB |
2646 | return 0; |
2647 | } | |
fb9d0acb | 2648 | venum = find_unused_vde(ddf); |
2649 | if (venum == DDF_NOTFOUND) { | |
2650 | pr_err("Cannot find spare slot for virtual disk\n"); | |
5f8097be NB |
2651 | return 0; |
2652 | } | |
2653 | ve = &ddf->virt->entries[venum]; | |
2654 | ||
2655 | /* A Virtual Disk GUID contains the T10 Vendor ID, controller type, | |
2656 | * timestamp, random number | |
2657 | */ | |
2658 | make_header_guid(ve->guid); | |
a8173e43 | 2659 | ve->unit = cpu_to_be16(info->md_minor); |
5f8097be | 2660 | ve->pad0 = 0xFFFF; |
a8173e43 | 2661 | ve->guid_crc._v16 = crc32(0, (unsigned char *)ddf->anchor.guid, |
2662 | DDF_GUID_LEN); | |
2663 | ve->type = cpu_to_be16(0); | |
7a7cc504 NB |
2664 | ve->state = DDF_state_degraded; /* Will be modified as devices are added */ |
2665 | if (info->state & 1) /* clean */ | |
2666 | ve->init_state = DDF_init_full; | |
2667 | else | |
2668 | ve->init_state = DDF_init_not; | |
2669 | ||
5f8097be NB |
2670 | memset(ve->pad1, 0xff, 14); |
2671 | memset(ve->name, ' ', 16); | |
2672 | if (name) | |
2673 | strncpy(ve->name, name, 16); | |
2674 | ddf->virt->populated_vdes = | |
a8173e43 | 2675 | cpu_to_be16(be16_to_cpu(ddf->virt->populated_vdes)+1); |
5f8097be NB |
2676 | |
2677 | /* Now create a new vd_config */ | |
3d2c4fc7 DW |
2678 | if (posix_memalign((void**)&vcl, 512, |
2679 | (offsetof(struct vcl, conf) + ddf->conf_rec_len * 512)) != 0) { | |
e7b84f9d | 2680 | pr_err("%s could not allocate vd_config\n", __func__); |
3d2c4fc7 DW |
2681 | return 0; |
2682 | } | |
59e36268 NB |
2683 | vcl->vcnum = venum; |
2684 | vcl->block_sizes = NULL; /* FIXME not for CONCAT */ | |
5f8097be NB |
2685 | vc = &vcl->conf; |
2686 | ||
2687 | vc->magic = DDF_VD_CONF_MAGIC; | |
2688 | memcpy(vc->guid, ve->guid, DDF_GUID_LEN); | |
60931cf9 | 2689 | vc->timestamp = cpu_to_be32(time(0)-DECADE); |
2690 | vc->seqnum = cpu_to_be32(1); | |
5f8097be | 2691 | memset(vc->pad0, 0xff, 24); |
5f8097be | 2692 | vc->chunk_shift = chunk_to_shift(info->chunk_size); |
a3163bf0 | 2693 | if (layout_md2ddf(info, vc) == -1 || |
a8173e43 | 2694 | be16_to_cpu(vc->prim_elmnt_count) > ddf->mppe) { |
a3163bf0 | 2695 | pr_err("%s: unsupported RAID level/layout %d/%d with %d disks\n", |
2696 | __func__, info->level, info->layout, info->raid_disks); | |
2697 | free(vcl); | |
2698 | return 0; | |
2699 | } | |
5f8097be | 2700 | vc->sec_elmnt_seq = 0; |
3c48f7be | 2701 | if (alloc_other_bvds(ddf, vcl) != 0) { |
2702 | pr_err("%s could not allocate other bvds\n", | |
2703 | __func__); | |
2704 | free(vcl); | |
2705 | return 0; | |
2706 | } | |
9d0c6b70 | 2707 | vc->blocks = cpu_to_be64(info->size * 2); |
2708 | vc->array_blocks = cpu_to_be64( | |
5f8097be NB |
2709 | calc_array_size(info->level, info->raid_disks, info->layout, |
2710 | info->chunk_size, info->size*2)); | |
2711 | memset(vc->pad1, 0xff, 8); | |
60931cf9 | 2712 | vc->spare_refs[0] = cpu_to_be32(0xffffffff); |
2713 | vc->spare_refs[1] = cpu_to_be32(0xffffffff); | |
2714 | vc->spare_refs[2] = cpu_to_be32(0xffffffff); | |
2715 | vc->spare_refs[3] = cpu_to_be32(0xffffffff); | |
2716 | vc->spare_refs[4] = cpu_to_be32(0xffffffff); | |
2717 | vc->spare_refs[5] = cpu_to_be32(0xffffffff); | |
2718 | vc->spare_refs[6] = cpu_to_be32(0xffffffff); | |
2719 | vc->spare_refs[7] = cpu_to_be32(0xffffffff); | |
5f8097be NB |
2720 | memset(vc->cache_pol, 0, 8); |
2721 | vc->bg_rate = 0x80; | |
2722 | memset(vc->pad2, 0xff, 3); | |
2723 | memset(vc->pad3, 0xff, 52); | |
2724 | memset(vc->pad4, 0xff, 192); | |
2725 | memset(vc->v0, 0xff, 32); | |
2726 | memset(vc->v1, 0xff, 32); | |
2727 | memset(vc->v2, 0xff, 16); | |
2728 | memset(vc->v3, 0xff, 16); | |
2729 | memset(vc->vendor, 0xff, 32); | |
598f0d58 | 2730 | |
8c3b8c2c | 2731 | memset(vc->phys_refnum, 0xff, 4*ddf->mppe); |
e5a2a3cf | 2732 | memset(vc->phys_refnum+ddf->mppe, 0x00, 8*ddf->mppe); |
5f8097be | 2733 | |
5aaf6c7b | 2734 | for (i = 1; i < vc->sec_elmnt_count; i++) { |
2735 | memcpy(vcl->other_bvds[i-1], vc, ddf->conf_rec_len * 512); | |
2736 | vcl->other_bvds[i-1]->sec_elmnt_seq = i; | |
2737 | } | |
2738 | ||
5f8097be NB |
2739 | vcl->next = ddf->conflist; |
2740 | ddf->conflist = vcl; | |
d2ca6449 | 2741 | ddf->currentconf = vcl; |
609ce161 | 2742 | ddf_set_updates_pending(ddf, NULL); |
5f8097be NB |
2743 | return 1; |
2744 | } | |
2745 | ||
0e600426 | 2746 | #ifndef MDASSEMBLE |
5f8097be NB |
2747 | static void add_to_super_ddf_bvd(struct supertype *st, |
2748 | mdu_disk_info_t *dk, int fd, char *devname) | |
2749 | { | |
56cb05c4 | 2750 | /* fd and devname identify a device within the ddf container (st). |
5f8097be NB |
2751 | * dk identifies a location in the new BVD. |
2752 | * We need to find suitable free space in that device and update | |
2753 | * the phys_refnum and lba_offset for the newly created vd_config. | |
2754 | * We might also want to update the type in the phys_disk | |
5575e7d9 | 2755 | * section. |
8592f29d N |
2756 | * |
2757 | * Alternately: fd == -1 and we have already chosen which device to | |
2758 | * use and recorded in dlist->raid_disk; | |
5f8097be NB |
2759 | */ |
2760 | struct dl *dl; | |
2761 | struct ddf_super *ddf = st->sb; | |
2762 | struct vd_config *vc; | |
f21e18ca | 2763 | unsigned int i; |
fca65520 | 2764 | unsigned long long blocks, pos; |
475ccbdb | 2765 | unsigned int raid_disk = dk->raid_disk; |
5f8097be | 2766 | |
8592f29d N |
2767 | if (fd == -1) { |
2768 | for (dl = ddf->dlist; dl ; dl = dl->next) | |
2769 | if (dl->raiddisk == dk->raid_disk) | |
2770 | break; | |
2771 | } else { | |
2772 | for (dl = ddf->dlist; dl ; dl = dl->next) | |
2773 | if (dl->major == dk->major && | |
2774 | dl->minor == dk->minor) | |
2775 | break; | |
2776 | } | |
41bcbc14 | 2777 | if (!dl || dl->pdnum < 0 || ! (dk->state & (1<<MD_DISK_SYNC))) |
5f8097be NB |
2778 | return; |
2779 | ||
d2ca6449 | 2780 | vc = &ddf->currentconf->conf; |
475ccbdb | 2781 | if (vc->sec_elmnt_count > 1) { |
a8173e43 | 2782 | unsigned int n = be16_to_cpu(vc->prim_elmnt_count); |
475ccbdb | 2783 | if (raid_disk >= n) |
2784 | vc = ddf->currentconf->other_bvds[raid_disk / n - 1]; | |
2785 | raid_disk %= n; | |
2786 | } | |
59e36268 | 2787 | |
9d0c6b70 | 2788 | blocks = be64_to_cpu(vc->blocks); |
d2ca6449 NB |
2789 | if (ddf->currentconf->block_sizes) |
2790 | blocks = ddf->currentconf->block_sizes[dk->raid_disk]; | |
59e36268 | 2791 | |
fca65520 N |
2792 | pos = find_space(ddf, dl, INVALID_SECTORS, &blocks); |
2793 | if (pos == INVALID_SECTORS) | |
59e36268 NB |
2794 | return; |
2795 | ||
d2ca6449 | 2796 | ddf->currentdev = dk->raid_disk; |
475ccbdb | 2797 | vc->phys_refnum[raid_disk] = dl->disk.refnum; |
9d0c6b70 | 2798 | LBA_OFFSET(ddf, vc)[raid_disk] = cpu_to_be64(pos); |
5f8097be | 2799 | |
f21e18ca | 2800 | for (i = 0; i < ddf->max_part ; i++) |
5575e7d9 NB |
2801 | if (dl->vlist[i] == NULL) |
2802 | break; | |
2803 | if (i == ddf->max_part) | |
2804 | return; | |
d2ca6449 | 2805 | dl->vlist[i] = ddf->currentconf; |
5f8097be | 2806 | |
8592f29d N |
2807 | if (fd >= 0) |
2808 | dl->fd = fd; | |
2809 | if (devname) | |
2810 | dl->devname = devname; | |
7a7cc504 | 2811 | |
63eb2454 | 2812 | /* Check if we can mark array as optimal yet */ |
d2ca6449 | 2813 | i = ddf->currentconf->vcnum; |
63eb2454 | 2814 | ddf->virt->entries[i].state = |
2815 | (ddf->virt->entries[i].state & ~DDF_state_mask) | |
2816 | | get_svd_state(ddf, ddf->currentconf); | |
a8173e43 | 2817 | be16_clear(ddf->phys->entries[dl->pdnum].type, |
2818 | cpu_to_be16(DDF_Global_Spare)); | |
2819 | be16_set(ddf->phys->entries[dl->pdnum].type, | |
2820 | cpu_to_be16(DDF_Active_in_VD)); | |
4f9bbe63 | 2821 | dprintf("%s: added disk %d/%08x to VD %d/%s as disk %d\n", |
60931cf9 | 2822 | __func__, dl->pdnum, be32_to_cpu(dl->disk.refnum), |
4f9bbe63 | 2823 | ddf->currentconf->vcnum, guid_str(vc->guid), |
2824 | dk->raid_disk); | |
609ce161 | 2825 | ddf_set_updates_pending(ddf, vc); |
5f8097be NB |
2826 | } |
2827 | ||
4a3ca8ac | 2828 | static unsigned int find_unused_pde(const struct ddf_super *ddf) |
2829 | { | |
2830 | unsigned int i; | |
a8173e43 | 2831 | for (i = 0; i < be16_to_cpu(ddf->phys->max_pdes); i++) { |
4a3ca8ac | 2832 | if (all_ff(ddf->phys->entries[i].guid)) |
2833 | return i; | |
2834 | } | |
2835 | return DDF_NOTFOUND; | |
2836 | } | |
2837 | ||
105e6e93 | 2838 | static void _set_config_size(struct phys_disk_entry *pde, const struct dl *dl) |
2839 | { | |
2840 | __u64 cfs, t; | |
2841 | cfs = min(dl->size - 32*1024*2ULL, be64_to_cpu(dl->primary_lba)); | |
2842 | t = be64_to_cpu(dl->secondary_lba); | |
2843 | if (t != ~(__u64)0) | |
2844 | cfs = min(cfs, t); | |
2845 | /* | |
2846 | * Some vendor DDF structures interpret workspace_lba | |
56cb05c4 | 2847 | * very differently than we do: Make a sanity check on the value. |
105e6e93 | 2848 | */ |
2849 | t = be64_to_cpu(dl->workspace_lba); | |
2850 | if (t < cfs) { | |
2851 | __u64 wsp = cfs - t; | |
2852 | if (wsp > 1024*1024*2ULL && wsp > dl->size / 16) { | |
2853 | pr_err("%s: %x:%x: workspace size 0x%llx too big, ignoring\n", | |
2854 | __func__, dl->major, dl->minor, wsp); | |
2855 | } else | |
2856 | cfs = t; | |
2857 | } | |
2858 | pde->config_size = cpu_to_be64(cfs); | |
2859 | dprintf("%s: %x:%x config_size %llx, DDF structure is %llx blocks\n", | |
2860 | __func__, dl->major, dl->minor, cfs, dl->size-cfs); | |
2861 | } | |
2862 | ||
56cb05c4 | 2863 | /* Add a device to a container, either while creating it or while |
a322f70c DW |
2864 | * expanding a pre-existing container |
2865 | */ | |
f20c3968 | 2866 | static int add_to_super_ddf(struct supertype *st, |
72ca9bcf N |
2867 | mdu_disk_info_t *dk, int fd, char *devname, |
2868 | unsigned long long data_offset) | |
a322f70c DW |
2869 | { |
2870 | struct ddf_super *ddf = st->sb; | |
2871 | struct dl *dd; | |
2872 | time_t now; | |
2873 | struct tm *tm; | |
2874 | unsigned long long size; | |
2875 | struct phys_disk_entry *pde; | |
f21e18ca | 2876 | unsigned int n, i; |
a322f70c | 2877 | struct stat stb; |
90fa1a29 | 2878 | __u32 *tptr; |
a322f70c | 2879 | |
78e44928 NB |
2880 | if (ddf->currentconf) { |
2881 | add_to_super_ddf_bvd(st, dk, fd, devname); | |
f20c3968 | 2882 | return 0; |
78e44928 NB |
2883 | } |
2884 | ||
a322f70c DW |
2885 | /* This is device numbered dk->number. We need to create |
2886 | * a phys_disk entry and a more detailed disk_data entry. | |
2887 | */ | |
2888 | fstat(fd, &stb); | |
4a3ca8ac | 2889 | n = find_unused_pde(ddf); |
2890 | if (n == DDF_NOTFOUND) { | |
2891 | pr_err("%s: No free slot in array, cannot add disk\n", | |
2892 | __func__); | |
2893 | return 1; | |
2894 | } | |
2895 | pde = &ddf->phys->entries[n]; | |
4ee8cca9 | 2896 | get_dev_size(fd, NULL, &size); |
2897 | if (size <= 32*1024*1024) { | |
2898 | pr_err("%s: device size must be at least 32MB\n", | |
2899 | __func__); | |
2900 | return 1; | |
2901 | } | |
2902 | size >>= 9; | |
4a3ca8ac | 2903 | |
3d2c4fc7 DW |
2904 | if (posix_memalign((void**)&dd, 512, |
2905 | sizeof(*dd) + sizeof(dd->vlist[0]) * ddf->max_part) != 0) { | |
e7b84f9d N |
2906 | pr_err("%s could allocate buffer for new disk, aborting\n", |
2907 | __func__); | |
f20c3968 | 2908 | return 1; |
3d2c4fc7 | 2909 | } |
a322f70c DW |
2910 | dd->major = major(stb.st_rdev); |
2911 | dd->minor = minor(stb.st_rdev); | |
2912 | dd->devname = devname; | |
a322f70c | 2913 | dd->fd = fd; |
b2280677 | 2914 | dd->spare = NULL; |
a322f70c DW |
2915 | |
2916 | dd->disk.magic = DDF_PHYS_DATA_MAGIC; | |
2917 | now = time(0); | |
2918 | tm = localtime(&now); | |
2919 | sprintf(dd->disk.guid, "%8s%04d%02d%02d", | |
2920 | T10, tm->tm_year+1900, tm->tm_mon+1, tm->tm_mday); | |
90fa1a29 JS |
2921 | tptr = (__u32 *)(dd->disk.guid + 16); |
2922 | *tptr++ = random32(); | |
2923 | *tptr = random32(); | |
a322f70c | 2924 | |
59e36268 NB |
2925 | do { |
2926 | /* Cannot be bothered finding a CRC of some irrelevant details*/ | |
60931cf9 | 2927 | dd->disk.refnum._v32 = random32(); |
a8173e43 | 2928 | for (i = be16_to_cpu(ddf->active->max_pd_entries); |
f21e18ca | 2929 | i > 0; i--) |
60931cf9 | 2930 | if (be32_eq(ddf->phys->entries[i-1].refnum, |
2931 | dd->disk.refnum)) | |
59e36268 | 2932 | break; |
f21e18ca | 2933 | } while (i > 0); |
59e36268 | 2934 | |
a322f70c DW |
2935 | dd->disk.forced_ref = 1; |
2936 | dd->disk.forced_guid = 1; | |
2937 | memset(dd->disk.vendor, ' ', 32); | |
2938 | memcpy(dd->disk.vendor, "Linux", 5); | |
2939 | memset(dd->disk.pad, 0xff, 442); | |
b2280677 | 2940 | for (i = 0; i < ddf->max_part ; i++) |
a322f70c DW |
2941 | dd->vlist[i] = NULL; |
2942 | ||
5575e7d9 NB |
2943 | dd->pdnum = n; |
2944 | ||
2cc2983d N |
2945 | if (st->update_tail) { |
2946 | int len = (sizeof(struct phys_disk) + | |
2947 | sizeof(struct phys_disk_entry)); | |
2948 | struct phys_disk *pd; | |
2949 | ||
503975b9 | 2950 | pd = xmalloc(len); |
2cc2983d | 2951 | pd->magic = DDF_PHYS_RECORDS_MAGIC; |
a8173e43 | 2952 | pd->used_pdes = cpu_to_be16(n); |
2cc2983d N |
2953 | pde = &pd->entries[0]; |
2954 | dd->mdupdate = pd; | |
4a3ca8ac | 2955 | } else |
a8173e43 | 2956 | ddf->phys->used_pdes = cpu_to_be16( |
2957 | 1 + be16_to_cpu(ddf->phys->used_pdes)); | |
a322f70c DW |
2958 | |
2959 | memcpy(pde->guid, dd->disk.guid, DDF_GUID_LEN); | |
2960 | pde->refnum = dd->disk.refnum; | |
a8173e43 | 2961 | pde->type = cpu_to_be16(DDF_Forced_PD_GUID | DDF_Global_Spare); |
2962 | pde->state = cpu_to_be16(DDF_Online); | |
4ee8cca9 | 2963 | dd->size = size; |
2964 | /* | |
2965 | * If there is already a device in dlist, try to reserve the same | |
2966 | * amount of workspace. Otherwise, use 32MB. | |
2967 | * We checked disk size above already. | |
2968 | */ | |
2969 | #define __calc_lba(new, old, lba, mb) do { \ | |
2970 | unsigned long long dif; \ | |
2971 | if ((old) != NULL) \ | |
9d0c6b70 | 2972 | dif = (old)->size - be64_to_cpu((old)->lba); \ |
4ee8cca9 | 2973 | else \ |
2974 | dif = (new)->size; \ | |
2975 | if ((new)->size > dif) \ | |
9d0c6b70 | 2976 | (new)->lba = cpu_to_be64((new)->size - dif); \ |
4ee8cca9 | 2977 | else \ |
9d0c6b70 | 2978 | (new)->lba = cpu_to_be64((new)->size - (mb*1024*2)); \ |
4ee8cca9 | 2979 | } while (0) |
2980 | __calc_lba(dd, ddf->dlist, workspace_lba, 32); | |
2981 | __calc_lba(dd, ddf->dlist, primary_lba, 16); | |
b95cb4b9 N |
2982 | if (ddf->dlist == NULL || |
2983 | be64_to_cpu(ddf->dlist->secondary_lba) != ~(__u64)0) | |
2984 | __calc_lba(dd, ddf->dlist, secondary_lba, 32); | |
105e6e93 | 2985 | _set_config_size(pde, dd); |
4ee8cca9 | 2986 | |
a322f70c DW |
2987 | sprintf(pde->path, "%17.17s","Information: nil") ; |
2988 | memset(pde->pad, 0xff, 6); | |
2989 | ||
2cc2983d N |
2990 | if (st->update_tail) { |
2991 | dd->next = ddf->add_list; | |
2992 | ddf->add_list = dd; | |
2993 | } else { | |
2994 | dd->next = ddf->dlist; | |
2995 | ddf->dlist = dd; | |
609ce161 | 2996 | ddf_set_updates_pending(ddf, NULL); |
2cc2983d | 2997 | } |
f20c3968 DW |
2998 | |
2999 | return 0; | |
a322f70c DW |
3000 | } |
3001 | ||
4dd968cc N |
3002 | static int remove_from_super_ddf(struct supertype *st, mdu_disk_info_t *dk) |
3003 | { | |
3004 | struct ddf_super *ddf = st->sb; | |
3005 | struct dl *dl; | |
3006 | ||
3007 | /* mdmon has noticed that this disk (dk->major/dk->minor) has | |
3008 | * disappeared from the container. | |
3009 | * We need to arrange that it disappears from the metadata and | |
3010 | * internal data structures too. | |
3011 | * Most of the work is done by ddf_process_update which edits | |
3012 | * the metadata and closes the file handle and attaches the memory | |
3013 | * where free_updates will free it. | |
3014 | */ | |
3015 | for (dl = ddf->dlist; dl ; dl = dl->next) | |
3016 | if (dl->major == dk->major && | |
3017 | dl->minor == dk->minor) | |
3018 | break; | |
a44e993e | 3019 | if (!dl || dl->pdnum < 0) |
4dd968cc N |
3020 | return -1; |
3021 | ||
3022 | if (st->update_tail) { | |
3023 | int len = (sizeof(struct phys_disk) + | |
3024 | sizeof(struct phys_disk_entry)); | |
3025 | struct phys_disk *pd; | |
3026 | ||
503975b9 | 3027 | pd = xmalloc(len); |
4dd968cc | 3028 | pd->magic = DDF_PHYS_RECORDS_MAGIC; |
a8173e43 | 3029 | pd->used_pdes = cpu_to_be16(dl->pdnum); |
3030 | pd->entries[0].state = cpu_to_be16(DDF_Missing); | |
4dd968cc N |
3031 | append_metadata_update(st, pd, len); |
3032 | } | |
3033 | return 0; | |
3034 | } | |
4441541f | 3035 | #endif |
4dd968cc | 3036 | |
a322f70c DW |
3037 | /* |
3038 | * This is the write_init_super method for a ddf container. It is | |
3039 | * called when creating a container or adding another device to a | |
3040 | * container. | |
3041 | */ | |
18a2f463 | 3042 | |
3921e41a | 3043 | static int __write_ddf_structure(struct dl *d, struct ddf_super *ddf, __u8 type) |
a322f70c | 3044 | { |
7f798aca | 3045 | unsigned long long sector; |
3046 | struct ddf_header *header; | |
3921e41a | 3047 | int fd, i, n_config, conf_size, buf_size; |
a4057a88 | 3048 | int ret = 0; |
3921e41a | 3049 | char *conf; |
8e9387ac | 3050 | |
7f798aca | 3051 | fd = d->fd; |
3052 | ||
3053 | switch (type) { | |
3054 | case DDF_HEADER_PRIMARY: | |
3055 | header = &ddf->primary; | |
9d0c6b70 | 3056 | sector = be64_to_cpu(header->primary_lba); |
7f798aca | 3057 | break; |
3058 | case DDF_HEADER_SECONDARY: | |
3059 | header = &ddf->secondary; | |
9d0c6b70 | 3060 | sector = be64_to_cpu(header->secondary_lba); |
7f798aca | 3061 | break; |
3062 | default: | |
3063 | return 0; | |
3064 | } | |
b95cb4b9 N |
3065 | if (sector == ~(__u64)0) |
3066 | return 0; | |
7f798aca | 3067 | |
3068 | header->type = type; | |
a4057a88 | 3069 | header->openflag = 1; |
7f798aca | 3070 | header->crc = calc_crc(header, 512); |
3071 | ||
3072 | lseek64(fd, sector<<9, 0); | |
3073 | if (write(fd, header, 512) < 0) | |
a4057a88 | 3074 | goto out; |
7f798aca | 3075 | |
3076 | ddf->controller.crc = calc_crc(&ddf->controller, 512); | |
3077 | if (write(fd, &ddf->controller, 512) < 0) | |
a4057a88 | 3078 | goto out; |
a322f70c | 3079 | |
7f798aca | 3080 | ddf->phys->crc = calc_crc(ddf->phys, ddf->pdsize); |
3081 | if (write(fd, ddf->phys, ddf->pdsize) < 0) | |
a4057a88 | 3082 | goto out; |
7f798aca | 3083 | ddf->virt->crc = calc_crc(ddf->virt, ddf->vdsize); |
3084 | if (write(fd, ddf->virt, ddf->vdsize) < 0) | |
a4057a88 | 3085 | goto out; |
7f798aca | 3086 | |
3087 | /* Now write lots of config records. */ | |
3088 | n_config = ddf->max_part; | |
3089 | conf_size = ddf->conf_rec_len * 512; | |
3921e41a N |
3090 | conf = ddf->conf; |
3091 | buf_size = conf_size * (n_config + 1); | |
3092 | if (!conf) { | |
3093 | if (posix_memalign((void**)&conf, 512, buf_size) != 0) | |
3094 | goto out; | |
3095 | ddf->conf = conf; | |
3096 | } | |
7f798aca | 3097 | for (i = 0 ; i <= n_config ; i++) { |
e3c2a365 | 3098 | struct vcl *c; |
3099 | struct vd_config *vdc = NULL; | |
3100 | if (i == n_config) { | |
7f798aca | 3101 | c = (struct vcl *)d->spare; |
e3c2a365 | 3102 | if (c) |
3103 | vdc = &c->conf; | |
3104 | } else { | |
3105 | unsigned int dummy; | |
3106 | c = d->vlist[i]; | |
3107 | if (c) | |
3108 | get_pd_index_from_refnum( | |
3109 | c, d->disk.refnum, | |
3110 | ddf->mppe, | |
3111 | (const struct vd_config **)&vdc, | |
3112 | &dummy); | |
3113 | } | |
188d31ed | 3114 | if (vdc) { |
be9b9ef4 | 3115 | dprintf("writing conf record %i on disk %08x for %s/%u\n", |
60931cf9 | 3116 | i, be32_to_cpu(d->disk.refnum), |
ad60eea1 | 3117 | guid_str(vdc->guid), |
be9b9ef4 | 3118 | vdc->sec_elmnt_seq); |
e3c2a365 | 3119 | vdc->crc = calc_crc(vdc, conf_size); |
3921e41a | 3120 | memcpy(conf + i*conf_size, vdc, conf_size); |
ce45c819 | 3121 | } else |
3921e41a | 3122 | memset(conf + i*conf_size, 0xff, conf_size); |
7f798aca | 3123 | } |
3921e41a | 3124 | if (write(fd, conf, buf_size) != buf_size) |
a4057a88 | 3125 | goto out; |
7f798aca | 3126 | |
3127 | d->disk.crc = calc_crc(&d->disk, 512); | |
3128 | if (write(fd, &d->disk, 512) < 0) | |
a4057a88 | 3129 | goto out; |
7f798aca | 3130 | |
a4057a88 | 3131 | ret = 1; |
3132 | out: | |
3133 | header->openflag = 0; | |
3134 | header->crc = calc_crc(header, 512); | |
3135 | ||
3136 | lseek64(fd, sector<<9, 0); | |
3137 | if (write(fd, header, 512) < 0) | |
3138 | ret = 0; | |
3139 | ||
3140 | return ret; | |
7f798aca | 3141 | } |
3142 | ||
3921e41a | 3143 | static int _write_super_to_disk(struct ddf_super *ddf, struct dl *d) |
9bf38704 | 3144 | { |
3145 | unsigned long long size; | |
3146 | int fd = d->fd; | |
3147 | if (fd < 0) | |
3148 | return 0; | |
3149 | ||
3150 | /* We need to fill in the primary, (secondary) and workspace | |
3151 | * lba's in the headers, set their checksums, | |
3152 | * Also checksum phys, virt.... | |
3153 | * | |
3154 | * Then write everything out, finally the anchor is written. | |
3155 | */ | |
3156 | get_dev_size(fd, NULL, &size); | |
3157 | size /= 512; | |
30bee020 | 3158 | memcpy(&ddf->anchor, ddf->active, 512); |
9d0c6b70 | 3159 | if (be64_to_cpu(d->workspace_lba) != 0ULL) |
9bf38704 | 3160 | ddf->anchor.workspace_lba = d->workspace_lba; |
3161 | else | |
3162 | ddf->anchor.workspace_lba = | |
9d0c6b70 | 3163 | cpu_to_be64(size - 32*1024*2); |
3164 | if (be64_to_cpu(d->primary_lba) != 0ULL) | |
9bf38704 | 3165 | ddf->anchor.primary_lba = d->primary_lba; |
3166 | else | |
3167 | ddf->anchor.primary_lba = | |
9d0c6b70 | 3168 | cpu_to_be64(size - 16*1024*2); |
3169 | if (be64_to_cpu(d->secondary_lba) != 0ULL) | |
9bf38704 | 3170 | ddf->anchor.secondary_lba = d->secondary_lba; |
3171 | else | |
3172 | ddf->anchor.secondary_lba = | |
9d0c6b70 | 3173 | cpu_to_be64(size - 32*1024*2); |
07de2684 | 3174 | ddf->anchor.timestamp = cpu_to_be32(time(0) - DECADE); |
9bf38704 | 3175 | memcpy(&ddf->primary, &ddf->anchor, 512); |
3176 | memcpy(&ddf->secondary, &ddf->anchor, 512); | |
3177 | ||
3178 | ddf->anchor.openflag = 0xFF; /* 'open' means nothing */ | |
60931cf9 | 3179 | ddf->anchor.seq = cpu_to_be32(0xFFFFFFFF); /* no sequencing in anchor */ |
9bf38704 | 3180 | ddf->anchor.crc = calc_crc(&ddf->anchor, 512); |
3181 | ||
3921e41a | 3182 | if (!__write_ddf_structure(d, ddf, DDF_HEADER_PRIMARY)) |
9bf38704 | 3183 | return 0; |
3184 | ||
3921e41a | 3185 | if (!__write_ddf_structure(d, ddf, DDF_HEADER_SECONDARY)) |
9bf38704 | 3186 | return 0; |
3187 | ||
3188 | lseek64(fd, (size-1)*512, SEEK_SET); | |
3189 | if (write(fd, &ddf->anchor, 512) < 0) | |
3190 | return 0; | |
3191 | ||
3192 | return 1; | |
3193 | } | |
3194 | ||
4441541f | 3195 | #ifndef MDASSEMBLE |
3921e41a | 3196 | static int __write_init_super_ddf(struct supertype *st) |
7f798aca | 3197 | { |
a322f70c | 3198 | struct ddf_super *ddf = st->sb; |
a322f70c | 3199 | struct dl *d; |
175593bf DW |
3200 | int attempts = 0; |
3201 | int successes = 0; | |
42d5dfd9 | 3202 | |
7d5a7ff3 | 3203 | pr_state(ddf, __func__); |
a322f70c | 3204 | |
175593bf DW |
3205 | /* try to write updated metadata, |
3206 | * if we catch a failure move on to the next disk | |
3207 | */ | |
a322f70c | 3208 | for (d = ddf->dlist; d; d=d->next) { |
175593bf | 3209 | attempts++; |
3921e41a | 3210 | successes += _write_super_to_disk(ddf, d); |
175593bf DW |
3211 | } |
3212 | ||
175593bf | 3213 | return attempts != successes; |
a322f70c | 3214 | } |
7a7cc504 NB |
3215 | |
3216 | static int write_init_super_ddf(struct supertype *st) | |
3217 | { | |
9b1fb677 DW |
3218 | struct ddf_super *ddf = st->sb; |
3219 | struct vcl *currentconf = ddf->currentconf; | |
3220 | ||
56cb05c4 | 3221 | /* We are done with currentconf - reset it so st refers to the container */ |
9b1fb677 | 3222 | ddf->currentconf = NULL; |
edd8d13c NB |
3223 | |
3224 | if (st->update_tail) { | |
3225 | /* queue the virtual_disk and vd_config as metadata updates */ | |
3226 | struct virtual_disk *vd; | |
3227 | struct vd_config *vc; | |
c5943560 | 3228 | int len, tlen; |
3229 | unsigned int i; | |
edd8d13c | 3230 | |
9b1fb677 | 3231 | if (!currentconf) { |
56cb05c4 | 3232 | /* Must be adding a physical disk to the container */ |
2cc2983d N |
3233 | int len = (sizeof(struct phys_disk) + |
3234 | sizeof(struct phys_disk_entry)); | |
3235 | ||
3236 | /* adding a disk to the container. */ | |
3237 | if (!ddf->add_list) | |
3238 | return 0; | |
3239 | ||
3240 | append_metadata_update(st, ddf->add_list->mdupdate, len); | |
3241 | ddf->add_list->mdupdate = NULL; | |
3242 | return 0; | |
3243 | } | |
3244 | ||
3245 | /* Newly created VD */ | |
3246 | ||
edd8d13c NB |
3247 | /* First the virtual disk. We have a slightly fake header */ |
3248 | len = sizeof(struct virtual_disk) + sizeof(struct virtual_entry); | |
503975b9 | 3249 | vd = xmalloc(len); |
edd8d13c | 3250 | *vd = *ddf->virt; |
9b1fb677 | 3251 | vd->entries[0] = ddf->virt->entries[currentconf->vcnum]; |
a8173e43 | 3252 | vd->populated_vdes = cpu_to_be16(currentconf->vcnum); |
edd8d13c NB |
3253 | append_metadata_update(st, vd, len); |
3254 | ||
3255 | /* Then the vd_config */ | |
3256 | len = ddf->conf_rec_len * 512; | |
c5943560 | 3257 | tlen = len * currentconf->conf.sec_elmnt_count; |
3258 | vc = xmalloc(tlen); | |
9b1fb677 | 3259 | memcpy(vc, ¤tconf->conf, len); |
c5943560 | 3260 | for (i = 1; i < currentconf->conf.sec_elmnt_count; i++) |
3261 | memcpy((char *)vc + i*len, currentconf->other_bvds[i-1], | |
3262 | len); | |
3263 | append_metadata_update(st, vc, tlen); | |
edd8d13c NB |
3264 | |
3265 | /* FIXME I need to close the fds! */ | |
3266 | return 0; | |
613b0d17 | 3267 | } else { |
d682f344 | 3268 | struct dl *d; |
19041058 | 3269 | if (!currentconf) |
3270 | for (d = ddf->dlist; d; d=d->next) | |
3271 | while (Kill(d->devname, NULL, 0, -1, 1) == 0); | |
3921e41a | 3272 | return __write_init_super_ddf(st); |
d682f344 | 3273 | } |
7a7cc504 NB |
3274 | } |
3275 | ||
a322f70c DW |
3276 | #endif |
3277 | ||
387fcd59 N |
3278 | static __u64 avail_size_ddf(struct supertype *st, __u64 devsize, |
3279 | unsigned long long data_offset) | |
a322f70c DW |
3280 | { |
3281 | /* We must reserve the last 32Meg */ | |
3282 | if (devsize <= 32*1024*2) | |
3283 | return 0; | |
3284 | return devsize - 32*1024*2; | |
3285 | } | |
3286 | ||
3287 | #ifndef MDASSEMBLE | |
8592f29d N |
3288 | |
3289 | static int reserve_space(struct supertype *st, int raiddisks, | |
3290 | unsigned long long size, int chunk, | |
3291 | unsigned long long *freesize) | |
3292 | { | |
3293 | /* Find 'raiddisks' spare extents at least 'size' big (but | |
3294 | * only caring about multiples of 'chunk') and remember | |
56cb05c4 N |
3295 | * them. If size==0, find the largest size possible. |
3296 | * Report available size in *freesize | |
3297 | * If space cannot be found, fail. | |
8592f29d N |
3298 | */ |
3299 | struct dl *dl; | |
3300 | struct ddf_super *ddf = st->sb; | |
3301 | int cnt = 0; | |
3302 | ||
3303 | for (dl = ddf->dlist; dl ; dl=dl->next) { | |
613b0d17 | 3304 | dl->raiddisk = -1; |
8592f29d N |
3305 | dl->esize = 0; |
3306 | } | |
3307 | /* Now find largest extent on each device */ | |
3308 | for (dl = ddf->dlist ; dl ; dl=dl->next) { | |
fca65520 | 3309 | unsigned long long minsize = ULLONG_MAX; |
8592f29d | 3310 | |
fca65520 N |
3311 | find_space(ddf, dl, INVALID_SECTORS, &minsize); |
3312 | if (minsize >= size && minsize >= (unsigned)chunk) { | |
8592f29d N |
3313 | cnt++; |
3314 | dl->esize = minsize; | |
3315 | } | |
8592f29d N |
3316 | } |
3317 | if (cnt < raiddisks) { | |
e7b84f9d | 3318 | pr_err("not enough devices with space to create array.\n"); |
8592f29d N |
3319 | return 0; /* No enough free spaces large enough */ |
3320 | } | |
3321 | if (size == 0) { | |
3322 | /* choose the largest size of which there are at least 'raiddisk' */ | |
3323 | for (dl = ddf->dlist ; dl ; dl=dl->next) { | |
3324 | struct dl *dl2; | |
3325 | if (dl->esize <= size) | |
3326 | continue; | |
3327 | /* This is bigger than 'size', see if there are enough */ | |
3328 | cnt = 0; | |
7b80ad6a | 3329 | for (dl2 = ddf->dlist; dl2 ; dl2=dl2->next) |
8592f29d N |
3330 | if (dl2->esize >= dl->esize) |
3331 | cnt++; | |
3332 | if (cnt >= raiddisks) | |
3333 | size = dl->esize; | |
3334 | } | |
3335 | if (chunk) { | |
3336 | size = size / chunk; | |
3337 | size *= chunk; | |
3338 | } | |
3339 | *freesize = size; | |
3340 | if (size < 32) { | |
e7b84f9d | 3341 | pr_err("not enough spare devices to create array.\n"); |
8592f29d N |
3342 | return 0; |
3343 | } | |
3344 | } | |
3345 | /* We have a 'size' of which there are enough spaces. | |
3346 | * We simply do a first-fit */ | |
3347 | cnt = 0; | |
3348 | for (dl = ddf->dlist ; dl && cnt < raiddisks ; dl=dl->next) { | |
3349 | if (dl->esize < size) | |
3350 | continue; | |
613b0d17 | 3351 | |
8592f29d N |
3352 | dl->raiddisk = cnt; |
3353 | cnt++; | |
3354 | } | |
3355 | return 1; | |
3356 | } | |
3357 | ||
78e44928 | 3358 | static int validate_geometry_ddf(struct supertype *st, |
2c514b71 | 3359 | int level, int layout, int raiddisks, |
c21e737b | 3360 | int *chunk, unsigned long long size, |
af4348dd | 3361 | unsigned long long data_offset, |
2c514b71 NB |
3362 | char *dev, unsigned long long *freesize, |
3363 | int verbose) | |
a322f70c DW |
3364 | { |
3365 | int fd; | |
3366 | struct mdinfo *sra; | |
3367 | int cfd; | |
3368 | ||
3369 | /* ddf potentially supports lots of things, but it depends on | |
3370 | * what devices are offered (and maybe kernel version?) | |
3371 | * If given unused devices, we will make a container. | |
3372 | * If given devices in a container, we will make a BVD. | |
3373 | * If given BVDs, we make an SVD, changing all the GUIDs in the process. | |
3374 | */ | |
3375 | ||
7ccc4cc4 | 3376 | if (*chunk == UnSet) |
bb7295f1 N |
3377 | *chunk = DEFAULT_CHUNK; |
3378 | ||
56cb05c4 N |
3379 | if (level == LEVEL_NONE) |
3380 | level = LEVEL_CONTAINER; | |
a322f70c | 3381 | if (level == LEVEL_CONTAINER) { |
78e44928 NB |
3382 | /* Must be a fresh device to add to a container */ |
3383 | return validate_geometry_ddf_container(st, level, layout, | |
7ccc4cc4 | 3384 | raiddisks, *chunk, |
af4348dd N |
3385 | size, data_offset, dev, |
3386 | freesize, | |
2c514b71 | 3387 | verbose); |
5f8097be NB |
3388 | } |
3389 | ||
78e44928 | 3390 | if (!dev) { |
a3163bf0 | 3391 | mdu_array_info_t array = { |
56cb05c4 N |
3392 | .level = level, |
3393 | .layout = layout, | |
a3163bf0 | 3394 | .raid_disks = raiddisks |
3395 | }; | |
3396 | struct vd_config conf; | |
3397 | if (layout_md2ddf(&array, &conf) == -1) { | |
b42f577a | 3398 | if (verbose) |
94b08b7c | 3399 | pr_err("DDF does not support level %d /layout %d arrays with %d disks\n", |
3400 | level, layout, raiddisks); | |
78e44928 | 3401 | return 0; |
b42f577a | 3402 | } |
78e44928 | 3403 | /* Should check layout? etc */ |
8592f29d N |
3404 | |
3405 | if (st->sb && freesize) { | |
3406 | /* --create was given a container to create in. | |
3407 | * So we need to check that there are enough | |
3408 | * free spaces and return the amount of space. | |
3409 | * We may as well remember which drives were | |
3410 | * chosen so that add_to_super/getinfo_super | |
3411 | * can return them. | |
3412 | */ | |
7ccc4cc4 | 3413 | return reserve_space(st, raiddisks, size, *chunk, freesize); |
8592f29d | 3414 | } |
a322f70c | 3415 | return 1; |
78e44928 | 3416 | } |
a322f70c | 3417 | |
8592f29d N |
3418 | if (st->sb) { |
3419 | /* A container has already been opened, so we are | |
3420 | * creating in there. Maybe a BVD, maybe an SVD. | |
3421 | * Should make a distinction one day. | |
3422 | */ | |
3423 | return validate_geometry_ddf_bvd(st, level, layout, raiddisks, | |
af4348dd N |
3424 | chunk, size, data_offset, dev, |
3425 | freesize, | |
8592f29d N |
3426 | verbose); |
3427 | } | |
78e44928 NB |
3428 | /* This is the first device for the array. |
3429 | * If it is a container, we read it in and do automagic allocations, | |
3430 | * no other devices should be given. | |
3431 | * Otherwise it must be a member device of a container, and we | |
3432 | * do manual allocation. | |
3433 | * Later we should check for a BVD and make an SVD. | |
a322f70c | 3434 | */ |
a322f70c DW |
3435 | fd = open(dev, O_RDONLY|O_EXCL, 0); |
3436 | if (fd >= 0) { | |
4dd2df09 | 3437 | sra = sysfs_read(fd, NULL, GET_VERSION); |
a322f70c DW |
3438 | close(fd); |
3439 | if (sra && sra->array.major_version == -1 && | |
78e44928 | 3440 | strcmp(sra->text_version, "ddf") == 0) { |
78e44928 NB |
3441 | /* load super */ |
3442 | /* find space for 'n' devices. */ | |
3443 | /* remember the devices */ | |
3444 | /* Somehow return the fact that we have enough */ | |
a322f70c DW |
3445 | } |
3446 | ||
2c514b71 | 3447 | if (verbose) |
e7b84f9d N |
3448 | pr_err("ddf: Cannot create this array " |
3449 | "on device %s - a container is required.\n", | |
3450 | dev); | |
a322f70c DW |
3451 | return 0; |
3452 | } | |
3453 | if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) { | |
2c514b71 | 3454 | if (verbose) |
e7b84f9d | 3455 | pr_err("ddf: Cannot open %s: %s\n", |
613b0d17 | 3456 | dev, strerror(errno)); |
a322f70c DW |
3457 | return 0; |
3458 | } | |
3459 | /* Well, it is in use by someone, maybe a 'ddf' container. */ | |
3460 | cfd = open_container(fd); | |
3461 | if (cfd < 0) { | |
3462 | close(fd); | |
2c514b71 | 3463 | if (verbose) |
e7b84f9d | 3464 | pr_err("ddf: Cannot use %s: %s\n", |
613b0d17 | 3465 | dev, strerror(EBUSY)); |
a322f70c DW |
3466 | return 0; |
3467 | } | |
4dd2df09 | 3468 | sra = sysfs_read(cfd, NULL, GET_VERSION); |
a322f70c DW |
3469 | close(fd); |
3470 | if (sra && sra->array.major_version == -1 && | |
3471 | strcmp(sra->text_version, "ddf") == 0) { | |
3472 | /* This is a member of a ddf container. Load the container | |
3473 | * and try to create a bvd | |
3474 | */ | |
3475 | struct ddf_super *ddf; | |
e1902a7b | 3476 | if (load_super_ddf_all(st, cfd, (void **)&ddf, NULL) == 0) { |
5f8097be | 3477 | st->sb = ddf; |
4dd2df09 | 3478 | strcpy(st->container_devnm, fd2devnm(cfd)); |
a322f70c | 3479 | close(cfd); |
78e44928 | 3480 | return validate_geometry_ddf_bvd(st, level, layout, |
a322f70c | 3481 | raiddisks, chunk, size, |
af4348dd | 3482 | data_offset, |
2c514b71 NB |
3483 | dev, freesize, |
3484 | verbose); | |
a322f70c DW |
3485 | } |
3486 | close(cfd); | |
c42ec1ed DW |
3487 | } else /* device may belong to a different container */ |
3488 | return 0; | |
3489 | ||
a322f70c DW |
3490 | return 1; |
3491 | } | |
3492 | ||
2c514b71 NB |
3493 | static int |
3494 | validate_geometry_ddf_container(struct supertype *st, | |
3495 | int level, int layout, int raiddisks, | |
3496 | int chunk, unsigned long long size, | |
af4348dd | 3497 | unsigned long long data_offset, |
2c514b71 NB |
3498 | char *dev, unsigned long long *freesize, |
3499 | int verbose) | |
a322f70c DW |
3500 | { |
3501 | int fd; | |
3502 | unsigned long long ldsize; | |
3503 | ||
3504 | if (level != LEVEL_CONTAINER) | |
3505 | return 0; | |
3506 | if (!dev) | |
3507 | return 1; | |
3508 | ||
3509 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
3510 | if (fd < 0) { | |
2c514b71 | 3511 | if (verbose) |
e7b84f9d | 3512 | pr_err("ddf: Cannot open %s: %s\n", |
613b0d17 | 3513 | dev, strerror(errno)); |
a322f70c DW |
3514 | return 0; |
3515 | } | |
3516 | if (!get_dev_size(fd, dev, &ldsize)) { | |
3517 | close(fd); | |
3518 | return 0; | |
3519 | } | |
3520 | close(fd); | |
3521 | ||
387fcd59 | 3522 | *freesize = avail_size_ddf(st, ldsize >> 9, INVALID_SECTORS); |
ea17e7aa N |
3523 | if (*freesize == 0) |
3524 | return 0; | |
a322f70c DW |
3525 | |
3526 | return 1; | |
3527 | } | |
3528 | ||
78e44928 NB |
3529 | static int validate_geometry_ddf_bvd(struct supertype *st, |
3530 | int level, int layout, int raiddisks, | |
c21e737b | 3531 | int *chunk, unsigned long long size, |
af4348dd | 3532 | unsigned long long data_offset, |
2c514b71 NB |
3533 | char *dev, unsigned long long *freesize, |
3534 | int verbose) | |
a322f70c DW |
3535 | { |
3536 | struct stat stb; | |
3537 | struct ddf_super *ddf = st->sb; | |
3538 | struct dl *dl; | |
5f8097be | 3539 | unsigned long long maxsize; |
a322f70c | 3540 | /* ddf/bvd supports lots of things, but not containers */ |
b42f577a N |
3541 | if (level == LEVEL_CONTAINER) { |
3542 | if (verbose) | |
e7b84f9d | 3543 | pr_err("DDF cannot create a container within an container\n"); |
a322f70c | 3544 | return 0; |
b42f577a | 3545 | } |
a322f70c DW |
3546 | /* We must have the container info already read in. */ |
3547 | if (!ddf) | |
3548 | return 0; | |
3549 | ||
5f8097be NB |
3550 | if (!dev) { |
3551 | /* General test: make sure there is space for | |
3552 | * 'raiddisks' device extents of size 'size'. | |
3553 | */ | |
3554 | unsigned long long minsize = size; | |
3555 | int dcnt = 0; | |
3556 | if (minsize == 0) | |
3557 | minsize = 8; | |
56cb05c4 | 3558 | for (dl = ddf->dlist; dl ; dl = dl->next) { |
fca65520 N |
3559 | if (find_space(ddf, dl, INVALID_SECTORS, &minsize) |
3560 | != INVALID_SECTORS) | |
5f8097be | 3561 | dcnt++; |
5f8097be NB |
3562 | } |
3563 | if (dcnt < raiddisks) { | |
2c514b71 | 3564 | if (verbose) |
e7b84f9d N |
3565 | pr_err("ddf: Not enough devices with " |
3566 | "space for this array (%d < %d)\n", | |
3567 | dcnt, raiddisks); | |
5f8097be NB |
3568 | return 0; |
3569 | } | |
3570 | return 1; | |
3571 | } | |
a322f70c DW |
3572 | /* This device must be a member of the set */ |
3573 | if (stat(dev, &stb) < 0) | |
3574 | return 0; | |
3575 | if ((S_IFMT & stb.st_mode) != S_IFBLK) | |
3576 | return 0; | |
3577 | for (dl = ddf->dlist ; dl ; dl = dl->next) { | |
f21e18ca N |
3578 | if (dl->major == (int)major(stb.st_rdev) && |
3579 | dl->minor == (int)minor(stb.st_rdev)) | |
a322f70c DW |
3580 | break; |
3581 | } | |
5f8097be | 3582 | if (!dl) { |
2c514b71 | 3583 | if (verbose) |
e7b84f9d | 3584 | pr_err("ddf: %s is not in the " |
613b0d17 N |
3585 | "same DDF set\n", |
3586 | dev); | |
5f8097be NB |
3587 | return 0; |
3588 | } | |
fca65520 N |
3589 | maxsize = ULLONG_MAX; |
3590 | find_space(ddf, dl, INVALID_SECTORS, &maxsize); | |
5f8097be | 3591 | *freesize = maxsize; |
a322f70c DW |
3592 | // FIXME here I am |
3593 | ||
3594 | return 1; | |
3595 | } | |
59e36268 | 3596 | |
a322f70c | 3597 | static int load_super_ddf_all(struct supertype *st, int fd, |
e1902a7b | 3598 | void **sbp, char *devname) |
a322f70c DW |
3599 | { |
3600 | struct mdinfo *sra; | |
3601 | struct ddf_super *super; | |
3602 | struct mdinfo *sd, *best = NULL; | |
3603 | int bestseq = 0; | |
3604 | int seq; | |
3605 | char nm[20]; | |
3606 | int dfd; | |
3607 | ||
b526e52d | 3608 | sra = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE); |
a322f70c DW |
3609 | if (!sra) |
3610 | return 1; | |
3611 | if (sra->array.major_version != -1 || | |
3612 | sra->array.minor_version != -2 || | |
3613 | strcmp(sra->text_version, "ddf") != 0) | |
3614 | return 1; | |
3615 | ||
6416d527 | 3616 | if (posix_memalign((void**)&super, 512, sizeof(*super)) != 0) |
a322f70c | 3617 | return 1; |
a2349791 | 3618 | memset(super, 0, sizeof(*super)); |
a322f70c DW |
3619 | |
3620 | /* first, try each device, and choose the best ddf */ | |
3621 | for (sd = sra->devs ; sd ; sd = sd->next) { | |
3622 | int rv; | |
3623 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
7a7cc504 NB |
3624 | dfd = dev_open(nm, O_RDONLY); |
3625 | if (dfd < 0) | |
a322f70c DW |
3626 | return 2; |
3627 | rv = load_ddf_headers(dfd, super, NULL); | |
7a7cc504 | 3628 | close(dfd); |
a322f70c | 3629 | if (rv == 0) { |
60931cf9 | 3630 | seq = be32_to_cpu(super->active->seq); |
a322f70c DW |
3631 | if (super->active->openflag) |
3632 | seq--; | |
3633 | if (!best || seq > bestseq) { | |
3634 | bestseq = seq; | |
3635 | best = sd; | |
3636 | } | |
3637 | } | |
3638 | } | |
3639 | if (!best) | |
3640 | return 1; | |
3641 | /* OK, load this ddf */ | |
3642 | sprintf(nm, "%d:%d", best->disk.major, best->disk.minor); | |
3643 | dfd = dev_open(nm, O_RDONLY); | |
7a7cc504 | 3644 | if (dfd < 0) |
a322f70c DW |
3645 | return 1; |
3646 | load_ddf_headers(dfd, super, NULL); | |
3647 | load_ddf_global(dfd, super, NULL); | |
3648 | close(dfd); | |
3649 | /* Now we need the device-local bits */ | |
3650 | for (sd = sra->devs ; sd ; sd = sd->next) { | |
3d2c4fc7 DW |
3651 | int rv; |
3652 | ||
a322f70c | 3653 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); |
e1902a7b | 3654 | dfd = dev_open(nm, O_RDWR); |
7a7cc504 | 3655 | if (dfd < 0) |
a322f70c | 3656 | return 2; |
3d2c4fc7 DW |
3657 | rv = load_ddf_headers(dfd, super, NULL); |
3658 | if (rv == 0) | |
e1902a7b | 3659 | rv = load_ddf_local(dfd, super, NULL, 1); |
3d2c4fc7 DW |
3660 | if (rv) |
3661 | return 1; | |
a322f70c | 3662 | } |
33414a01 | 3663 | |
a322f70c DW |
3664 | *sbp = super; |
3665 | if (st->ss == NULL) { | |
78e44928 | 3666 | st->ss = &super_ddf; |
a322f70c DW |
3667 | st->minor_version = 0; |
3668 | st->max_devs = 512; | |
3669 | } | |
4dd2df09 | 3670 | strcpy(st->container_devnm, fd2devnm(fd)); |
a322f70c DW |
3671 | return 0; |
3672 | } | |
2b959fbf N |
3673 | |
3674 | static int load_container_ddf(struct supertype *st, int fd, | |
3675 | char *devname) | |
3676 | { | |
3677 | return load_super_ddf_all(st, fd, &st->sb, devname); | |
3678 | } | |
3679 | ||
0e600426 | 3680 | #endif /* MDASSEMBLE */ |
a322f70c | 3681 | |
a5c7adb3 | 3682 | static int check_secondary(const struct vcl *vc) |
3683 | { | |
3684 | const struct vd_config *conf = &vc->conf; | |
3685 | int i; | |
3686 | ||
3687 | /* The only DDF secondary RAID level md can support is | |
3688 | * RAID 10, if the stripe sizes and Basic volume sizes | |
3689 | * are all equal. | |
3690 | * Other configurations could in theory be supported by exposing | |
3691 | * the BVDs to user space and using device mapper for the secondary | |
3692 | * mapping. So far we don't support that. | |
3693 | */ | |
3694 | ||
3695 | __u64 sec_elements[4] = {0, 0, 0, 0}; | |
3696 | #define __set_sec_seen(n) (sec_elements[(n)>>6] |= (1<<((n)&63))) | |
3697 | #define __was_sec_seen(n) ((sec_elements[(n)>>6] & (1<<((n)&63))) != 0) | |
3698 | ||
3699 | if (vc->other_bvds == NULL) { | |
3700 | pr_err("No BVDs for secondary RAID found\n"); | |
3701 | return -1; | |
3702 | } | |
3703 | if (conf->prl != DDF_RAID1) { | |
3704 | pr_err("Secondary RAID level only supported for mirrored BVD\n"); | |
3705 | return -1; | |
3706 | } | |
3707 | if (conf->srl != DDF_2STRIPED && conf->srl != DDF_2SPANNED) { | |
3708 | pr_err("Secondary RAID level %d is unsupported\n", | |
3709 | conf->srl); | |
3710 | return -1; | |
3711 | } | |
3712 | __set_sec_seen(conf->sec_elmnt_seq); | |
3713 | for (i = 0; i < conf->sec_elmnt_count-1; i++) { | |
3714 | const struct vd_config *bvd = vc->other_bvds[i]; | |
3c48f7be | 3715 | if (bvd->sec_elmnt_seq == DDF_UNUSED_BVD) |
c98567ba | 3716 | continue; |
a5c7adb3 | 3717 | if (bvd->srl != conf->srl) { |
3718 | pr_err("Inconsistent secondary RAID level across BVDs\n"); | |
3719 | return -1; | |
3720 | } | |
3721 | if (bvd->prl != conf->prl) { | |
3722 | pr_err("Different RAID levels for BVDs are unsupported\n"); | |
3723 | return -1; | |
3724 | } | |
a8173e43 | 3725 | if (!be16_eq(bvd->prim_elmnt_count, conf->prim_elmnt_count)) { |
a5c7adb3 | 3726 | pr_err("All BVDs must have the same number of primary elements\n"); |
3727 | return -1; | |
3728 | } | |
3729 | if (bvd->chunk_shift != conf->chunk_shift) { | |
3730 | pr_err("Different strip sizes for BVDs are unsupported\n"); | |
3731 | return -1; | |
3732 | } | |
9d0c6b70 | 3733 | if (!be64_eq(bvd->array_blocks, conf->array_blocks)) { |
a5c7adb3 | 3734 | pr_err("Different BVD sizes are unsupported\n"); |
3735 | return -1; | |
3736 | } | |
3737 | __set_sec_seen(bvd->sec_elmnt_seq); | |
3738 | } | |
3739 | for (i = 0; i < conf->sec_elmnt_count; i++) { | |
3740 | if (!__was_sec_seen(i)) { | |
3741 | pr_err("BVD %d is missing\n", i); | |
3742 | return -1; | |
3743 | } | |
3744 | } | |
3745 | return 0; | |
3746 | } | |
3747 | ||
8a38db86 | 3748 | static unsigned int get_pd_index_from_refnum(const struct vcl *vc, |
60931cf9 | 3749 | be32 refnum, unsigned int nmax, |
4e587018 | 3750 | const struct vd_config **bvd, |
3751 | unsigned int *idx) | |
8a38db86 | 3752 | { |
4e587018 | 3753 | unsigned int i, j, n, sec, cnt; |
3754 | ||
a8173e43 | 3755 | cnt = be16_to_cpu(vc->conf.prim_elmnt_count); |
4e587018 | 3756 | sec = (vc->conf.sec_elmnt_count == 1 ? 0 : vc->conf.sec_elmnt_seq); |
3757 | ||
3758 | for (i = 0, j = 0 ; i < nmax ; i++) { | |
3759 | /* j counts valid entries for this BVD */ | |
60931cf9 | 3760 | if (be32_eq(vc->conf.phys_refnum[i], refnum)) { |
4e587018 | 3761 | *bvd = &vc->conf; |
3762 | *idx = i; | |
56cb05c4 | 3763 | return sec * cnt + j; |
4e587018 | 3764 | } |
56cb05c4 N |
3765 | if (be32_to_cpu(vc->conf.phys_refnum[i]) != 0xffffffff) |
3766 | j++; | |
4e587018 | 3767 | } |
3768 | if (vc->other_bvds == NULL) | |
3769 | goto bad; | |
3770 | ||
3771 | for (n = 1; n < vc->conf.sec_elmnt_count; n++) { | |
3772 | struct vd_config *vd = vc->other_bvds[n-1]; | |
4e587018 | 3773 | sec = vd->sec_elmnt_seq; |
3c48f7be | 3774 | if (sec == DDF_UNUSED_BVD) |
3775 | continue; | |
4e587018 | 3776 | for (i = 0, j = 0 ; i < nmax ; i++) { |
60931cf9 | 3777 | if (be32_eq(vd->phys_refnum[i], refnum)) { |
4e587018 | 3778 | *bvd = vd; |
3779 | *idx = i; | |
56cb05c4 | 3780 | return sec * cnt + j; |
4e587018 | 3781 | } |
56cb05c4 N |
3782 | if (be32_to_cpu(vd->phys_refnum[i]) != 0xffffffff) |
3783 | j++; | |
4e587018 | 3784 | } |
3785 | } | |
3786 | bad: | |
3787 | *bvd = NULL; | |
d6e7b083 | 3788 | return DDF_NOTFOUND; |
8a38db86 | 3789 | } |
3790 | ||
00bbdbda | 3791 | static struct mdinfo *container_content_ddf(struct supertype *st, char *subarray) |
598f0d58 NB |
3792 | { |
3793 | /* Given a container loaded by load_super_ddf_all, | |
3794 | * extract information about all the arrays into | |
3795 | * an mdinfo tree. | |
3796 | * | |
3797 | * For each vcl in conflist: create an mdinfo, fill it in, | |
3798 | * then look for matching devices (phys_refnum) in dlist | |
3799 | * and create appropriate device mdinfo. | |
3800 | */ | |
3801 | struct ddf_super *ddf = st->sb; | |
3802 | struct mdinfo *rest = NULL; | |
3803 | struct vcl *vc; | |
3804 | ||
56cb05c4 | 3805 | for (vc = ddf->conflist ; vc ; vc=vc->next) { |
f21e18ca | 3806 | unsigned int i; |
598f0d58 | 3807 | struct mdinfo *this; |
00bbdbda | 3808 | char *ep; |
90fa1a29 | 3809 | __u32 *cptr; |
8a38db86 | 3810 | unsigned int pd; |
00bbdbda N |
3811 | |
3812 | if (subarray && | |
3813 | (strtoul(subarray, &ep, 10) != vc->vcnum || | |
3814 | *ep != '\0')) | |
3815 | continue; | |
3816 | ||
a5c7adb3 | 3817 | if (vc->conf.sec_elmnt_count > 1) { |
3818 | if (check_secondary(vc) != 0) | |
3819 | continue; | |
3820 | } | |
3821 | ||
503975b9 | 3822 | this = xcalloc(1, sizeof(*this)); |
598f0d58 NB |
3823 | this->next = rest; |
3824 | rest = this; | |
3825 | ||
8a2848a7 | 3826 | if (layout_ddf2md(&vc->conf, &this->array)) |
3827 | continue; | |
598f0d58 | 3828 | this->array.md_minor = -1; |
f35f2525 N |
3829 | this->array.major_version = -1; |
3830 | this->array.minor_version = -2; | |
56cb05c4 | 3831 | this->safe_mode_delay = DDF_SAFE_MODE_DELAY; |
90fa1a29 JS |
3832 | cptr = (__u32 *)(vc->conf.guid + 16); |
3833 | this->array.ctime = DECADE + __be32_to_cpu(*cptr); | |
598f0d58 | 3834 | this->array.utime = DECADE + |
60931cf9 | 3835 | be32_to_cpu(vc->conf.timestamp); |
598f0d58 NB |
3836 | this->array.chunk_size = 512 << vc->conf.chunk_shift; |
3837 | ||
59e36268 | 3838 | i = vc->vcnum; |
7a7cc504 NB |
3839 | if ((ddf->virt->entries[i].state & DDF_state_inconsistent) || |
3840 | (ddf->virt->entries[i].init_state & DDF_initstate_mask) != | |
ed9d66aa | 3841 | DDF_init_full) { |
598f0d58 | 3842 | this->array.state = 0; |
ed9d66aa NB |
3843 | this->resync_start = 0; |
3844 | } else { | |
598f0d58 | 3845 | this->array.state = 1; |
b7528a20 | 3846 | this->resync_start = MaxSector; |
ed9d66aa | 3847 | } |
8bf989d8 | 3848 | _ddf_array_name(this->name, ddf, i); |
598f0d58 | 3849 | memset(this->uuid, 0, sizeof(this->uuid)); |
56cb05c4 N |
3850 | this->component_size = be64_to_cpu(vc->conf.blocks); |
3851 | this->array.size = this->component_size / 2; | |
3852 | this->container_member = i; | |
598f0d58 | 3853 | |
c5afc314 N |
3854 | ddf->currentconf = vc; |
3855 | uuid_from_super_ddf(st, this->uuid); | |
f646805e | 3856 | if (!subarray) |
3857 | ddf->currentconf = NULL; | |
c5afc314 | 3858 | |
60f18132 | 3859 | sprintf(this->text_version, "/%s/%d", |
4dd2df09 | 3860 | st->container_devnm, this->container_member); |
60f18132 | 3861 | |
217dead4 | 3862 | for (pd = 0; pd < be16_to_cpu(ddf->phys->max_pdes); pd++) { |
598f0d58 NB |
3863 | struct mdinfo *dev; |
3864 | struct dl *d; | |
4e587018 | 3865 | const struct vd_config *bvd; |
3866 | unsigned int iphys; | |
fa033bec | 3867 | int stt; |
598f0d58 | 3868 | |
60931cf9 | 3869 | if (be32_to_cpu(ddf->phys->entries[pd].refnum) |
3870 | == 0xFFFFFFFF) | |
bc17324f | 3871 | continue; |
0cf5ef67 | 3872 | |
a8173e43 | 3873 | stt = be16_to_cpu(ddf->phys->entries[pd].state); |
fa033bec N |
3874 | if ((stt & (DDF_Online|DDF_Failed|DDF_Rebuilding)) |
3875 | != DDF_Online) | |
3876 | continue; | |
3877 | ||
8a38db86 | 3878 | i = get_pd_index_from_refnum( |
4e587018 | 3879 | vc, ddf->phys->entries[pd].refnum, |
3880 | ddf->mppe, &bvd, &iphys); | |
d6e7b083 | 3881 | if (i == DDF_NOTFOUND) |
8a38db86 | 3882 | continue; |
3883 | ||
fa033bec | 3884 | this->array.working_disks++; |
bc17324f | 3885 | |
0cf5ef67 | 3886 | for (d = ddf->dlist; d ; d=d->next) |
60931cf9 | 3887 | if (be32_eq(d->disk.refnum, |
3888 | ddf->phys->entries[pd].refnum)) | |
0cf5ef67 N |
3889 | break; |
3890 | if (d == NULL) | |
3891 | /* Haven't found that one yet, maybe there are others */ | |
3892 | continue; | |
3893 | ||
503975b9 | 3894 | dev = xcalloc(1, sizeof(*dev)); |
56cb05c4 N |
3895 | dev->next = this->devs; |
3896 | this->devs = dev; | |
598f0d58 | 3897 | |
60931cf9 | 3898 | dev->disk.number = be32_to_cpu(d->disk.refnum); |
56cb05c4 N |
3899 | dev->disk.major = d->major; |
3900 | dev->disk.minor = d->minor; | |
598f0d58 | 3901 | dev->disk.raid_disk = i; |
56cb05c4 | 3902 | dev->disk.state = (1<<MD_DISK_SYNC)|(1<<MD_DISK_ACTIVE); |
d23534e4 | 3903 | dev->recovery_start = MaxSector; |
598f0d58 | 3904 | |
56cb05c4 | 3905 | dev->events = be32_to_cpu(ddf->active->seq); |
57a66662 | 3906 | dev->data_offset = |
9d0c6b70 | 3907 | be64_to_cpu(LBA_OFFSET(ddf, bvd)[iphys]); |
3908 | dev->component_size = be64_to_cpu(bvd->blocks); | |
598f0d58 NB |
3909 | if (d->devname) |
3910 | strcpy(dev->name, d->devname); | |
3911 | } | |
3912 | } | |
3913 | return rest; | |
3914 | } | |
3915 | ||
955e9ea1 | 3916 | static int store_super_ddf(struct supertype *st, int fd) |
a322f70c | 3917 | { |
955e9ea1 | 3918 | struct ddf_super *ddf = st->sb; |
a322f70c | 3919 | unsigned long long dsize; |
6416d527 | 3920 | void *buf; |
3d2c4fc7 | 3921 | int rc; |
a322f70c | 3922 | |
955e9ea1 DW |
3923 | if (!ddf) |
3924 | return 1; | |
3925 | ||
a322f70c DW |
3926 | if (!get_dev_size(fd, NULL, &dsize)) |
3927 | return 1; | |
3928 | ||
dbf98368 | 3929 | if (ddf->dlist || ddf->conflist) { |
3930 | struct stat sta; | |
3931 | struct dl *dl; | |
3932 | int ofd, ret; | |
3933 | ||
3934 | if (fstat(fd, &sta) == -1 || !S_ISBLK(sta.st_mode)) { | |
3935 | pr_err("%s: file descriptor for invalid device\n", | |
3936 | __func__); | |
3937 | return 1; | |
3938 | } | |
3939 | for (dl = ddf->dlist; dl; dl = dl->next) | |
3940 | if (dl->major == (int)major(sta.st_rdev) && | |
3941 | dl->minor == (int)minor(sta.st_rdev)) | |
3942 | break; | |
3943 | if (!dl) { | |
3944 | pr_err("%s: couldn't find disk %d/%d\n", __func__, | |
3945 | (int)major(sta.st_rdev), | |
3946 | (int)minor(sta.st_rdev)); | |
3947 | return 1; | |
3948 | } | |
dbf98368 | 3949 | ofd = dl->fd; |
3950 | dl->fd = fd; | |
3921e41a | 3951 | ret = (_write_super_to_disk(ddf, dl) != 1); |
dbf98368 | 3952 | dl->fd = ofd; |
3953 | return ret; | |
3954 | } | |
3955 | ||
3d2c4fc7 DW |
3956 | if (posix_memalign(&buf, 512, 512) != 0) |
3957 | return 1; | |
6416d527 NB |
3958 | memset(buf, 0, 512); |
3959 | ||
a322f70c | 3960 | lseek64(fd, dsize-512, 0); |
3d2c4fc7 | 3961 | rc = write(fd, buf, 512); |
6416d527 | 3962 | free(buf); |
3d2c4fc7 DW |
3963 | if (rc < 0) |
3964 | return 1; | |
a322f70c DW |
3965 | return 0; |
3966 | } | |
3967 | ||
a19c88b8 NB |
3968 | static int compare_super_ddf(struct supertype *st, struct supertype *tst) |
3969 | { | |
3970 | /* | |
3971 | * return: | |
3972 | * 0 same, or first was empty, and second was copied | |
56cb05c4 | 3973 | * 1 second had wrong magic number - but that isn't possible |
a19c88b8 NB |
3974 | * 2 wrong uuid |
3975 | * 3 wrong other info | |
3976 | */ | |
3977 | struct ddf_super *first = st->sb; | |
3978 | struct ddf_super *second = tst->sb; | |
4eefd651 | 3979 | struct dl *dl1, *dl2; |
3980 | struct vcl *vl1, *vl2; | |
2d210697 | 3981 | unsigned int max_vds, max_pds, pd, vd; |
a19c88b8 NB |
3982 | |
3983 | if (!first) { | |
3984 | st->sb = tst->sb; | |
3985 | tst->sb = NULL; | |
3986 | return 0; | |
3987 | } | |
3988 | ||
3989 | if (memcmp(first->anchor.guid, second->anchor.guid, DDF_GUID_LEN) != 0) | |
3990 | return 2; | |
3991 | ||
f43f5b32 N |
3992 | /* It is only OK to compare info in the anchor. Anything else |
3993 | * could be changing due to a reconfig so must be ignored. | |
3994 | * guid really should be enough anyway. | |
3995 | */ | |
2d210697 | 3996 | |
f43f5b32 N |
3997 | if (!be32_eq(first->active->seq, second->active->seq)) { |
3998 | dprintf("%s: sequence number mismatch %u<->%u\n", __func__, | |
3999 | be32_to_cpu(first->active->seq), | |
4000 | be32_to_cpu(second->active->seq)); | |
4001 | return 0; | |
2d210697 | 4002 | } |
2d210697 | 4003 | |
4eefd651 | 4004 | /* |
56cb05c4 N |
4005 | * At this point we are fairly sure that the meta data matches. |
4006 | * But the new disk may contain additional local data. | |
4007 | * Add it to the super block. | |
4eefd651 | 4008 | */ |
f43f5b32 | 4009 | max_vds = be16_to_cpu(first->active->max_vd_entries); |
217dead4 | 4010 | max_pds = be16_to_cpu(first->phys->max_pdes); |
4eefd651 | 4011 | for (vl2 = second->conflist; vl2; vl2 = vl2->next) { |
4012 | for (vl1 = first->conflist; vl1; vl1 = vl1->next) | |
4013 | if (!memcmp(vl1->conf.guid, vl2->conf.guid, | |
4014 | DDF_GUID_LEN)) | |
4015 | break; | |
4016 | if (vl1) { | |
4017 | if (vl1->other_bvds != NULL && | |
4018 | vl1->conf.sec_elmnt_seq != | |
4019 | vl2->conf.sec_elmnt_seq) { | |
4020 | dprintf("%s: adding BVD %u\n", __func__, | |
4021 | vl2->conf.sec_elmnt_seq); | |
4022 | add_other_bvd(vl1, &vl2->conf, | |
4023 | first->conf_rec_len*512); | |
4024 | } | |
4025 | continue; | |
4026 | } | |
4027 | ||
4028 | if (posix_memalign((void **)&vl1, 512, | |
4029 | (first->conf_rec_len*512 + | |
4030 | offsetof(struct vcl, conf))) != 0) { | |
4031 | pr_err("%s could not allocate vcl buf\n", | |
4032 | __func__); | |
4033 | return 3; | |
4034 | } | |
4035 | ||
4036 | vl1->next = first->conflist; | |
4037 | vl1->block_sizes = NULL; | |
4eefd651 | 4038 | memcpy(&vl1->conf, &vl2->conf, first->conf_rec_len*512); |
3c48f7be | 4039 | if (alloc_other_bvds(first, vl1) != 0) { |
4040 | pr_err("%s could not allocate other bvds\n", | |
4041 | __func__); | |
4042 | free(vl1); | |
4043 | return 3; | |
4044 | } | |
4eefd651 | 4045 | for (vd = 0; vd < max_vds; vd++) |
4046 | if (!memcmp(first->virt->entries[vd].guid, | |
4047 | vl1->conf.guid, DDF_GUID_LEN)) | |
4048 | break; | |
4049 | vl1->vcnum = vd; | |
4050 | dprintf("%s: added config for VD %u\n", __func__, vl1->vcnum); | |
4051 | first->conflist = vl1; | |
4052 | } | |
4053 | ||
4054 | for (dl2 = second->dlist; dl2; dl2 = dl2->next) { | |
4055 | for (dl1 = first->dlist; dl1; dl1 = dl1->next) | |
60931cf9 | 4056 | if (be32_eq(dl1->disk.refnum, dl2->disk.refnum)) |
4eefd651 | 4057 | break; |
4058 | if (dl1) | |
4059 | continue; | |
4060 | ||
4061 | if (posix_memalign((void **)&dl1, 512, | |
4062 | sizeof(*dl1) + (first->max_part) * sizeof(dl1->vlist[0])) | |
4063 | != 0) { | |
4064 | pr_err("%s could not allocate disk info buffer\n", | |
4065 | __func__); | |
4066 | return 3; | |
4067 | } | |
4068 | memcpy(dl1, dl2, sizeof(*dl1)); | |
4069 | dl1->mdupdate = NULL; | |
4070 | dl1->next = first->dlist; | |
4071 | dl1->fd = -1; | |
4072 | for (pd = 0; pd < max_pds; pd++) | |
60931cf9 | 4073 | if (be32_eq(first->phys->entries[pd].refnum, |
4074 | dl1->disk.refnum)) | |
4eefd651 | 4075 | break; |
41bcbc14 | 4076 | dl1->pdnum = pd < max_pds ? (int)pd : -1; |
4eefd651 | 4077 | if (dl2->spare) { |
4078 | if (posix_memalign((void **)&dl1->spare, 512, | |
4079 | first->conf_rec_len*512) != 0) { | |
4080 | pr_err("%s could not allocate spare info buf\n", | |
4081 | __func__); | |
4082 | return 3; | |
4083 | } | |
4084 | memcpy(dl1->spare, dl2->spare, first->conf_rec_len*512); | |
4085 | } | |
4086 | for (vd = 0 ; vd < first->max_part ; vd++) { | |
4087 | if (!dl2->vlist[vd]) { | |
4088 | dl1->vlist[vd] = NULL; | |
4089 | continue; | |
4090 | } | |
4091 | for (vl1 = first->conflist; vl1; vl1 = vl1->next) { | |
4092 | if (!memcmp(vl1->conf.guid, | |
4093 | dl2->vlist[vd]->conf.guid, | |
4094 | DDF_GUID_LEN)) | |
4095 | break; | |
4096 | dl1->vlist[vd] = vl1; | |
4097 | } | |
4098 | } | |
4099 | first->dlist = dl1; | |
4100 | dprintf("%s: added disk %d: %08x\n", __func__, dl1->pdnum, | |
60931cf9 | 4101 | be32_to_cpu(dl1->disk.refnum)); |
4eefd651 | 4102 | } |
4103 | ||
a19c88b8 NB |
4104 | return 0; |
4105 | } | |
4106 | ||
0e600426 | 4107 | #ifndef MDASSEMBLE |
4e5528c6 NB |
4108 | /* |
4109 | * A new array 'a' has been started which claims to be instance 'inst' | |
4110 | * within container 'c'. | |
4111 | * We need to confirm that the array matches the metadata in 'c' so | |
4112 | * that we don't corrupt any metadata. | |
4113 | */ | |
cba0191b | 4114 | static int ddf_open_new(struct supertype *c, struct active_array *a, char *inst) |
549e9569 | 4115 | { |
a2aa439e | 4116 | struct ddf_super *ddf = c->sb; |
4117 | int n = atoi(inst); | |
5daa35ac | 4118 | struct mdinfo *dev; |
4119 | struct dl *dl; | |
4120 | static const char faulty[] = "faulty"; | |
4121 | ||
fb9d0acb | 4122 | if (all_ff(ddf->virt->entries[n].guid)) { |
4123 | pr_err("%s: subarray %d doesn't exist\n", __func__, n); | |
a2aa439e | 4124 | return -ENODEV; |
4125 | } | |
5daa35ac | 4126 | dprintf("%s: new subarray %d, GUID: %s\n", __func__, n, |
4127 | guid_str(ddf->virt->entries[n].guid)); | |
4128 | for (dev = a->info.devs; dev; dev = dev->next) { | |
4129 | for (dl = ddf->dlist; dl; dl = dl->next) | |
4130 | if (dl->major == dev->disk.major && | |
4131 | dl->minor == dev->disk.minor) | |
4132 | break; | |
a44e993e | 4133 | if (!dl || dl->pdnum < 0) { |
5daa35ac | 4134 | pr_err("%s: device %d/%d of subarray %d not found in meta data\n", |
4135 | __func__, dev->disk.major, dev->disk.minor, n); | |
4136 | return -1; | |
4137 | } | |
4138 | if ((be16_to_cpu(ddf->phys->entries[dl->pdnum].state) & | |
4139 | (DDF_Online|DDF_Missing|DDF_Failed)) != DDF_Online) { | |
4140 | pr_err("%s: new subarray %d contains broken device %d/%d (%02x)\n", | |
4141 | __func__, n, dl->major, dl->minor, | |
4142 | be16_to_cpu( | |
4143 | ddf->phys->entries[dl->pdnum].state)); | |
4144 | if (write(dev->state_fd, faulty, sizeof(faulty)-1) != | |
4145 | sizeof(faulty) - 1) | |
4146 | pr_err("Write to state_fd failed\n"); | |
4147 | dev->curr_state = DS_FAULTY; | |
4148 | } | |
4149 | } | |
a2aa439e | 4150 | a->info.container_member = n; |
549e9569 NB |
4151 | return 0; |
4152 | } | |
4153 | ||
e5a03804 | 4154 | static void handle_missing(struct ddf_super *ddf, struct active_array *a, int inst) |
5a46fcd7 N |
4155 | { |
4156 | /* This member array is being activated. If any devices | |
4157 | * are missing they must now be marked as failed. | |
4158 | */ | |
4159 | struct vd_config *vc; | |
4160 | unsigned int n_bvd; | |
4161 | struct vcl *vcl; | |
4162 | struct dl *dl; | |
e5a03804 | 4163 | int pd; |
5a46fcd7 | 4164 | int n; |
e5a03804 | 4165 | int state; |
5a46fcd7 N |
4166 | |
4167 | for (n = 0; ; n++) { | |
4168 | vc = find_vdcr(ddf, inst, n, &n_bvd, &vcl); | |
4169 | if (!vc) | |
4170 | break; | |
4171 | for (dl = ddf->dlist; dl; dl = dl->next) | |
4172 | if (be32_eq(dl->disk.refnum, vc->phys_refnum[n_bvd])) | |
4173 | break; | |
4174 | if (dl) | |
4175 | /* Found this disk, so not missing */ | |
4176 | continue; | |
e5a03804 N |
4177 | |
4178 | /* Mark the device as failed/missing. */ | |
4179 | pd = find_phys(ddf, vc->phys_refnum[n_bvd]); | |
4180 | if (pd >= 0 && be16_and(ddf->phys->entries[pd].state, | |
4181 | cpu_to_be16(DDF_Online))) { | |
4182 | be16_clear(ddf->phys->entries[pd].state, | |
4183 | cpu_to_be16(DDF_Online)); | |
4184 | be16_set(ddf->phys->entries[pd].state, | |
4185 | cpu_to_be16(DDF_Failed|DDF_Missing)); | |
4186 | vc->phys_refnum[n_bvd] = cpu_to_be32(0); | |
609ce161 | 4187 | ddf_set_updates_pending(ddf, vc); |
e5a03804 N |
4188 | } |
4189 | ||
4190 | /* Mark the array as Degraded */ | |
4191 | state = get_svd_state(ddf, vcl); | |
4192 | if (ddf->virt->entries[inst].state != | |
4193 | ((ddf->virt->entries[inst].state & ~DDF_state_mask) | |
4194 | | state)) { | |
4195 | ddf->virt->entries[inst].state = | |
4196 | (ddf->virt->entries[inst].state & ~DDF_state_mask) | |
4197 | | state; | |
4198 | a->check_degraded = 1; | |
609ce161 | 4199 | ddf_set_updates_pending(ddf, vc); |
e5a03804 | 4200 | } |
5a46fcd7 N |
4201 | } |
4202 | } | |
4203 | ||
4e5528c6 NB |
4204 | /* |
4205 | * The array 'a' is to be marked clean in the metadata. | |
ed9d66aa | 4206 | * If '->resync_start' is not ~(unsigned long long)0, then the array is only |
4e5528c6 NB |
4207 | * clean up to the point (in sectors). If that cannot be recorded in the |
4208 | * metadata, then leave it as dirty. | |
4209 | * | |
4210 | * For DDF, we need to clear the DDF_state_inconsistent bit in the | |
4211 | * !global! virtual_disk.virtual_entry structure. | |
4212 | */ | |
01f157d7 | 4213 | static int ddf_set_array_state(struct active_array *a, int consistent) |
549e9569 | 4214 | { |
4e5528c6 NB |
4215 | struct ddf_super *ddf = a->container->sb; |
4216 | int inst = a->info.container_member; | |
18a2f463 | 4217 | int old = ddf->virt->entries[inst].state; |
01f157d7 | 4218 | if (consistent == 2) { |
e5a03804 | 4219 | handle_missing(ddf, a, inst); |
01f157d7 N |
4220 | /* Should check if a recovery should be started FIXME */ |
4221 | consistent = 1; | |
b7941fd6 | 4222 | if (!is_resync_complete(&a->info)) |
01f157d7 N |
4223 | consistent = 0; |
4224 | } | |
ed9d66aa NB |
4225 | if (consistent) |
4226 | ddf->virt->entries[inst].state &= ~DDF_state_inconsistent; | |
4227 | else | |
4e5528c6 | 4228 | ddf->virt->entries[inst].state |= DDF_state_inconsistent; |
18a2f463 | 4229 | if (old != ddf->virt->entries[inst].state) |
609ce161 | 4230 | ddf_set_updates_pending(ddf, NULL); |
18a2f463 NB |
4231 | |
4232 | old = ddf->virt->entries[inst].init_state; | |
ed9d66aa | 4233 | ddf->virt->entries[inst].init_state &= ~DDF_initstate_mask; |
b7941fd6 | 4234 | if (is_resync_complete(&a->info)) |
ed9d66aa | 4235 | ddf->virt->entries[inst].init_state |= DDF_init_full; |
b7941fd6 | 4236 | else if (a->info.resync_start == 0) |
ed9d66aa | 4237 | ddf->virt->entries[inst].init_state |= DDF_init_not; |
4e5528c6 | 4238 | else |
ed9d66aa | 4239 | ddf->virt->entries[inst].init_state |= DDF_init_quick; |
18a2f463 | 4240 | if (old != ddf->virt->entries[inst].init_state) |
609ce161 | 4241 | ddf_set_updates_pending(ddf, NULL); |
ed9d66aa | 4242 | |
b27336a2 | 4243 | dprintf("ddf mark %d/%s (%d) %s %llu\n", inst, |
4244 | guid_str(ddf->virt->entries[inst].guid), a->curr_state, | |
4245 | consistent?"clean":"dirty", | |
b7941fd6 | 4246 | a->info.resync_start); |
01f157d7 | 4247 | return consistent; |
fd7cde1b DW |
4248 | } |
4249 | ||
5ec636b7 | 4250 | static int get_bvd_state(const struct ddf_super *ddf, |
4251 | const struct vd_config *vc) | |
4252 | { | |
4253 | unsigned int i, n_bvd, working = 0; | |
a8173e43 | 4254 | unsigned int n_prim = be16_to_cpu(vc->prim_elmnt_count); |
5ec636b7 | 4255 | int pd, st, state; |
f70d549f N |
4256 | char *avail = xcalloc(1, n_prim); |
4257 | mdu_array_info_t array; | |
4258 | ||
4259 | layout_ddf2md(vc, &array); | |
4260 | ||
5ec636b7 | 4261 | for (i = 0; i < n_prim; i++) { |
4262 | if (!find_index_in_bvd(ddf, vc, i, &n_bvd)) | |
4263 | continue; | |
4264 | pd = find_phys(ddf, vc->phys_refnum[n_bvd]); | |
4265 | if (pd < 0) | |
4266 | continue; | |
a8173e43 | 4267 | st = be16_to_cpu(ddf->phys->entries[pd].state); |
5ec636b7 | 4268 | if ((st & (DDF_Online|DDF_Failed|DDF_Rebuilding)) |
f70d549f | 4269 | == DDF_Online) { |
5ec636b7 | 4270 | working++; |
f70d549f N |
4271 | avail[i] = 1; |
4272 | } | |
5ec636b7 | 4273 | } |
4274 | ||
4275 | state = DDF_state_degraded; | |
4276 | if (working == n_prim) | |
4277 | state = DDF_state_optimal; | |
4278 | else | |
4279 | switch (vc->prl) { | |
4280 | case DDF_RAID0: | |
4281 | case DDF_CONCAT: | |
4282 | case DDF_JBOD: | |
4283 | state = DDF_state_failed; | |
4284 | break; | |
4285 | case DDF_RAID1: | |
4286 | if (working == 0) | |
4287 | state = DDF_state_failed; | |
4288 | else if (working >= 2) | |
4289 | state = DDF_state_part_optimal; | |
4290 | break; | |
f70d549f N |
4291 | case DDF_RAID1E: |
4292 | if (!enough(10, n_prim, array.layout, 1, avail)) | |
4293 | state = DDF_state_failed; | |
4294 | break; | |
5ec636b7 | 4295 | case DDF_RAID4: |
4296 | case DDF_RAID5: | |
4297 | if (working < n_prim - 1) | |
4298 | state = DDF_state_failed; | |
4299 | break; | |
4300 | case DDF_RAID6: | |
4301 | if (working < n_prim - 2) | |
4302 | state = DDF_state_failed; | |
4303 | else if (working == n_prim - 1) | |
4304 | state = DDF_state_part_optimal; | |
4305 | break; | |
4306 | } | |
4307 | return state; | |
4308 | } | |
4309 | ||
0777d17d | 4310 | static int secondary_state(int state, int other, int seclevel) |
4311 | { | |
4312 | if (state == DDF_state_optimal && other == DDF_state_optimal) | |
4313 | return DDF_state_optimal; | |
4314 | if (seclevel == DDF_2MIRRORED) { | |
4315 | if (state == DDF_state_optimal || other == DDF_state_optimal) | |
4316 | return DDF_state_part_optimal; | |
4317 | if (state == DDF_state_failed && other == DDF_state_failed) | |
4318 | return DDF_state_failed; | |
4319 | return DDF_state_degraded; | |
4320 | } else { | |
4321 | if (state == DDF_state_failed || other == DDF_state_failed) | |
4322 | return DDF_state_failed; | |
4323 | if (state == DDF_state_degraded || other == DDF_state_degraded) | |
4324 | return DDF_state_degraded; | |
4325 | return DDF_state_part_optimal; | |
4326 | } | |
4327 | } | |
4328 | ||
4329 | static int get_svd_state(const struct ddf_super *ddf, const struct vcl *vcl) | |
4330 | { | |
4331 | int state = get_bvd_state(ddf, &vcl->conf); | |
4332 | unsigned int i; | |
4333 | for (i = 1; i < vcl->conf.sec_elmnt_count; i++) { | |
4334 | state = secondary_state( | |
4335 | state, | |
4336 | get_bvd_state(ddf, vcl->other_bvds[i-1]), | |
4337 | vcl->conf.srl); | |
4338 | } | |
4339 | return state; | |
4340 | } | |
4341 | ||
7a7cc504 NB |
4342 | /* |
4343 | * The state of each disk is stored in the global phys_disk structure | |
4344 | * in phys_disk.entries[n].state. | |
4345 | * This makes various combinations awkward. | |
4346 | * - When a device fails in any array, it must be failed in all arrays | |
4347 | * that include a part of this device. | |
4348 | * - When a component is rebuilding, we cannot include it officially in the | |
4349 | * array unless this is the only array that uses the device. | |
4350 | * | |
4351 | * So: when transitioning: | |
4352 | * Online -> failed, just set failed flag. monitor will propagate | |
4353 | * spare -> online, the device might need to be added to the array. | |
4354 | * spare -> failed, just set failed. Don't worry if in array or not. | |
4355 | */ | |
8d45d196 | 4356 | static void ddf_set_disk(struct active_array *a, int n, int state) |
549e9569 | 4357 | { |
7a7cc504 | 4358 | struct ddf_super *ddf = a->container->sb; |
baba3f4e | 4359 | unsigned int inst = a->info.container_member, n_bvd; |
4360 | struct vcl *vcl; | |
4361 | struct vd_config *vc = find_vdcr(ddf, inst, (unsigned int)n, | |
4362 | &n_bvd, &vcl); | |
4363 | int pd; | |
e1316fab N |
4364 | struct mdinfo *mdi; |
4365 | struct dl *dl; | |
609ce161 | 4366 | int update = 0; |
7a7cc504 | 4367 | |
ce6844b9 | 4368 | dprintf("%s: %d to %x\n", __func__, n, state); |
7a7cc504 | 4369 | if (vc == NULL) { |
2c514b71 | 4370 | dprintf("ddf: cannot find instance %d!!\n", inst); |
7a7cc504 NB |
4371 | return; |
4372 | } | |
e1316fab N |
4373 | /* Find the matching slot in 'info'. */ |
4374 | for (mdi = a->info.devs; mdi; mdi = mdi->next) | |
4375 | if (mdi->disk.raid_disk == n) | |
4376 | break; | |
ce6844b9 MW |
4377 | if (!mdi) { |
4378 | pr_err("%s: cannot find raid disk %d\n", | |
4379 | __func__, n); | |
e1316fab | 4380 | return; |
ce6844b9 | 4381 | } |
e1316fab N |
4382 | |
4383 | /* and find the 'dl' entry corresponding to that. */ | |
4384 | for (dl = ddf->dlist; dl; dl = dl->next) | |
77632af9 N |
4385 | if (mdi->state_fd >= 0 && |
4386 | mdi->disk.major == dl->major && | |
e1316fab N |
4387 | mdi->disk.minor == dl->minor) |
4388 | break; | |
ce6844b9 MW |
4389 | if (!dl) { |
4390 | pr_err("%s: cannot find raid disk %d (%d/%d)\n", | |
4391 | __func__, n, | |
4392 | mdi->disk.major, mdi->disk.minor); | |
e1316fab | 4393 | return; |
ce6844b9 | 4394 | } |
e1316fab | 4395 | |
baba3f4e | 4396 | pd = find_phys(ddf, vc->phys_refnum[n_bvd]); |
e1316fab N |
4397 | if (pd < 0 || pd != dl->pdnum) { |
4398 | /* disk doesn't currently exist or has changed. | |
4399 | * If it is now in_sync, insert it. */ | |
baba3f4e | 4400 | dprintf("%s: phys disk not found for %d: %d/%d ref %08x\n", |
4401 | __func__, dl->pdnum, dl->major, dl->minor, | |
60931cf9 | 4402 | be32_to_cpu(dl->disk.refnum)); |
baba3f4e | 4403 | dprintf("%s: array %u disk %u ref %08x pd %d\n", |
60931cf9 | 4404 | __func__, inst, n_bvd, |
4405 | be32_to_cpu(vc->phys_refnum[n_bvd]), pd); | |
7a7cc504 | 4406 | if ((state & DS_INSYNC) && ! (state & DS_FAULTY)) { |
baba3f4e | 4407 | pd = dl->pdnum; /* FIXME: is this really correct ? */ |
4408 | vc->phys_refnum[n_bvd] = dl->disk.refnum; | |
57a66662 | 4409 | LBA_OFFSET(ddf, vc)[n_bvd] = |
9d0c6b70 | 4410 | cpu_to_be64(mdi->data_offset); |
a8173e43 | 4411 | be16_clear(ddf->phys->entries[pd].type, |
4412 | cpu_to_be16(DDF_Global_Spare)); | |
4413 | be16_set(ddf->phys->entries[pd].type, | |
4414 | cpu_to_be16(DDF_Active_in_VD)); | |
609ce161 | 4415 | update = 1; |
7a7cc504 NB |
4416 | } |
4417 | } else { | |
a8173e43 | 4418 | be16 old = ddf->phys->entries[pd].state; |
7a7cc504 | 4419 | if (state & DS_FAULTY) |
a8173e43 | 4420 | be16_set(ddf->phys->entries[pd].state, |
4421 | cpu_to_be16(DDF_Failed)); | |
7a7cc504 | 4422 | if (state & DS_INSYNC) { |
a8173e43 | 4423 | be16_set(ddf->phys->entries[pd].state, |
4424 | cpu_to_be16(DDF_Online)); | |
4425 | be16_clear(ddf->phys->entries[pd].state, | |
4426 | cpu_to_be16(DDF_Rebuilding)); | |
7a7cc504 | 4427 | } |
a8173e43 | 4428 | if (!be16_eq(old, ddf->phys->entries[pd].state)) |
609ce161 | 4429 | update = 1; |
7a7cc504 NB |
4430 | } |
4431 | ||
ce6844b9 MW |
4432 | dprintf("ddf: set_disk %d (%08x) to %x->%02x\n", n, |
4433 | be32_to_cpu(dl->disk.refnum), state, | |
4434 | be16_to_cpu(ddf->phys->entries[pd].state)); | |
7e1432fb | 4435 | |
7a7cc504 NB |
4436 | /* Now we need to check the state of the array and update |
4437 | * virtual_disk.entries[n].state. | |
4438 | * It needs to be one of "optimal", "degraded", "failed". | |
4439 | * I don't understand 'deleted' or 'missing'. | |
4440 | */ | |
0777d17d | 4441 | state = get_svd_state(ddf, vcl); |
7a7cc504 | 4442 | |
18a2f463 NB |
4443 | if (ddf->virt->entries[inst].state != |
4444 | ((ddf->virt->entries[inst].state & ~DDF_state_mask) | |
4445 | | state)) { | |
18a2f463 NB |
4446 | ddf->virt->entries[inst].state = |
4447 | (ddf->virt->entries[inst].state & ~DDF_state_mask) | |
4448 | | state; | |
609ce161 | 4449 | update = 1; |
18a2f463 | 4450 | } |
609ce161 N |
4451 | if (update) |
4452 | ddf_set_updates_pending(ddf, vc); | |
549e9569 NB |
4453 | } |
4454 | ||
2e735d19 | 4455 | static void ddf_sync_metadata(struct supertype *st) |
549e9569 | 4456 | { |
7a7cc504 NB |
4457 | /* |
4458 | * Write all data to all devices. | |
4459 | * Later, we might be able to track whether only local changes | |
4460 | * have been made, or whether any global data has been changed, | |
4461 | * but ddf is sufficiently weird that it probably always | |
4462 | * changes global data .... | |
4463 | */ | |
18a2f463 NB |
4464 | struct ddf_super *ddf = st->sb; |
4465 | if (!ddf->updates_pending) | |
4466 | return; | |
4467 | ddf->updates_pending = 0; | |
3921e41a | 4468 | __write_init_super_ddf(st); |
2c514b71 | 4469 | dprintf("ddf: sync_metadata\n"); |
549e9569 NB |
4470 | } |
4471 | ||
f646805e | 4472 | static int del_from_conflist(struct vcl **list, const char *guid) |
4473 | { | |
4474 | struct vcl **p; | |
4475 | int found = 0; | |
4476 | for (p = list; p && *p; p = &((*p)->next)) | |
4477 | if (!memcmp((*p)->conf.guid, guid, DDF_GUID_LEN)) { | |
4478 | found = 1; | |
4479 | *p = (*p)->next; | |
4480 | } | |
4481 | return found; | |
4482 | } | |
4483 | ||
4484 | static int _kill_subarray_ddf(struct ddf_super *ddf, const char *guid) | |
4485 | { | |
4486 | struct dl *dl; | |
4487 | unsigned int vdnum, i; | |
4488 | vdnum = find_vde_by_guid(ddf, guid); | |
4489 | if (vdnum == DDF_NOTFOUND) { | |
4490 | pr_err("%s: could not find VD %s\n", __func__, | |
4491 | guid_str(guid)); | |
4492 | return -1; | |
4493 | } | |
4494 | if (del_from_conflist(&ddf->conflist, guid) == 0) { | |
4495 | pr_err("%s: could not find conf %s\n", __func__, | |
4496 | guid_str(guid)); | |
4497 | return -1; | |
4498 | } | |
4499 | for (dl = ddf->dlist; dl; dl = dl->next) | |
4500 | for (i = 0; i < ddf->max_part; i++) | |
4501 | if (dl->vlist[i] != NULL && | |
4502 | !memcmp(dl->vlist[i]->conf.guid, guid, | |
4503 | DDF_GUID_LEN)) | |
4504 | dl->vlist[i] = NULL; | |
4505 | memset(ddf->virt->entries[vdnum].guid, 0xff, DDF_GUID_LEN); | |
4506 | dprintf("%s: deleted %s\n", __func__, guid_str(guid)); | |
4507 | return 0; | |
4508 | } | |
4509 | ||
4510 | static int kill_subarray_ddf(struct supertype *st) | |
4511 | { | |
4512 | struct ddf_super *ddf = st->sb; | |
4513 | /* | |
4514 | * currentconf is set in container_content_ddf, | |
4515 | * called with subarray arg | |
4516 | */ | |
4517 | struct vcl *victim = ddf->currentconf; | |
4518 | struct vd_config *conf; | |
f646805e | 4519 | unsigned int vdnum; |
56cb05c4 N |
4520 | |
4521 | ddf->currentconf = NULL; | |
f646805e | 4522 | if (!victim) { |
4523 | pr_err("%s: nothing to kill\n", __func__); | |
4524 | return -1; | |
4525 | } | |
4526 | conf = &victim->conf; | |
4527 | vdnum = find_vde_by_guid(ddf, conf->guid); | |
4528 | if (vdnum == DDF_NOTFOUND) { | |
4529 | pr_err("%s: could not find VD %s\n", __func__, | |
4530 | guid_str(conf->guid)); | |
4531 | return -1; | |
4532 | } | |
4533 | if (st->update_tail) { | |
4534 | struct virtual_disk *vd; | |
4535 | int len = sizeof(struct virtual_disk) | |
4536 | + sizeof(struct virtual_entry); | |
4537 | vd = xmalloc(len); | |
4538 | if (vd == NULL) { | |
4539 | pr_err("%s: failed to allocate %d bytes\n", __func__, | |
4540 | len); | |
4541 | return -1; | |
4542 | } | |
4543 | memset(vd, 0 , len); | |
4544 | vd->magic = DDF_VIRT_RECORDS_MAGIC; | |
a8173e43 | 4545 | vd->populated_vdes = cpu_to_be16(0); |
f646805e | 4546 | memcpy(vd->entries[0].guid, conf->guid, DDF_GUID_LEN); |
4547 | /* we use DDF_state_deleted as marker */ | |
4548 | vd->entries[0].state = DDF_state_deleted; | |
4549 | append_metadata_update(st, vd, len); | |
6a350d82 | 4550 | } else { |
f646805e | 4551 | _kill_subarray_ddf(ddf, conf->guid); |
609ce161 | 4552 | ddf_set_updates_pending(ddf, NULL); |
6a350d82 | 4553 | ddf_sync_metadata(st); |
4554 | } | |
f646805e | 4555 | return 0; |
4556 | } | |
4557 | ||
c5943560 | 4558 | static void copy_matching_bvd(struct ddf_super *ddf, |
4559 | struct vd_config *conf, | |
4560 | const struct metadata_update *update) | |
4561 | { | |
4562 | unsigned int mppe = | |
a8173e43 | 4563 | be16_to_cpu(ddf->anchor.max_primary_element_entries); |
c5943560 | 4564 | unsigned int len = ddf->conf_rec_len * 512; |
4565 | char *p; | |
4566 | struct vd_config *vc; | |
4567 | for (p = update->buf; p < update->buf + update->len; p += len) { | |
4568 | vc = (struct vd_config *) p; | |
4569 | if (vc->sec_elmnt_seq == conf->sec_elmnt_seq) { | |
4570 | memcpy(conf->phys_refnum, vc->phys_refnum, | |
4571 | mppe * (sizeof(__u32) + sizeof(__u64))); | |
4572 | return; | |
4573 | } | |
4574 | } | |
4575 | pr_err("%s: no match for BVD %d of %s in update\n", __func__, | |
4576 | conf->sec_elmnt_seq, guid_str(conf->guid)); | |
4577 | } | |
4578 | ||
88c164f4 NB |
4579 | static void ddf_process_update(struct supertype *st, |
4580 | struct metadata_update *update) | |
4581 | { | |
4582 | /* Apply this update to the metadata. | |
4583 | * The first 4 bytes are a DDF_*_MAGIC which guides | |
4584 | * our actions. | |
4585 | * Possible update are: | |
4586 | * DDF_PHYS_RECORDS_MAGIC | |
4dd968cc N |
4587 | * Add a new physical device or remove an old one. |
4588 | * Changes to this record only happen implicitly. | |
88c164f4 NB |
4589 | * used_pdes is the device number. |
4590 | * DDF_VIRT_RECORDS_MAGIC | |
4591 | * Add a new VD. Possibly also change the 'access' bits. | |
4592 | * populated_vdes is the entry number. | |
4593 | * DDF_VD_CONF_MAGIC | |
4594 | * New or updated VD. the VIRT_RECORD must already | |
4595 | * exist. For an update, phys_refnum and lba_offset | |
4596 | * (at least) are updated, and the VD_CONF must | |
4597 | * be written to precisely those devices listed with | |
4598 | * a phys_refnum. | |
4599 | * DDF_SPARE_ASSIGN_MAGIC | |
4600 | * replacement Spare Assignment Record... but for which device? | |
4601 | * | |
4602 | * So, e.g.: | |
4603 | * - to create a new array, we send a VIRT_RECORD and | |
4604 | * a VD_CONF. Then assemble and start the array. | |
4605 | * - to activate a spare we send a VD_CONF to add the phys_refnum | |
4606 | * and offset. This will also mark the spare as active with | |
4607 | * a spare-assignment record. | |
4608 | */ | |
4609 | struct ddf_super *ddf = st->sb; | |
60931cf9 | 4610 | be32 *magic = (be32 *)update->buf; |
88c164f4 NB |
4611 | struct phys_disk *pd; |
4612 | struct virtual_disk *vd; | |
4613 | struct vd_config *vc; | |
4614 | struct vcl *vcl; | |
4615 | struct dl *dl; | |
f21e18ca | 4616 | unsigned int ent; |
c5943560 | 4617 | unsigned int pdnum, pd2, len; |
88c164f4 | 4618 | |
60931cf9 | 4619 | dprintf("Process update %x\n", be32_to_cpu(*magic)); |
7e1432fb | 4620 | |
60931cf9 | 4621 | if (be32_eq(*magic, DDF_PHYS_RECORDS_MAGIC)) { |
88c164f4 NB |
4622 | if (update->len != (sizeof(struct phys_disk) + |
4623 | sizeof(struct phys_disk_entry))) | |
4624 | return; | |
4625 | pd = (struct phys_disk*)update->buf; | |
4626 | ||
a8173e43 | 4627 | ent = be16_to_cpu(pd->used_pdes); |
4628 | if (ent >= be16_to_cpu(ddf->phys->max_pdes)) | |
88c164f4 | 4629 | return; |
a8173e43 | 4630 | if (be16_and(pd->entries[0].state, cpu_to_be16(DDF_Missing))) { |
4dd968cc N |
4631 | struct dl **dlp; |
4632 | /* removing this disk. */ | |
a8173e43 | 4633 | be16_set(ddf->phys->entries[ent].state, |
4634 | cpu_to_be16(DDF_Missing)); | |
4dd968cc N |
4635 | for (dlp = &ddf->dlist; *dlp; dlp = &(*dlp)->next) { |
4636 | struct dl *dl = *dlp; | |
4637 | if (dl->pdnum == (signed)ent) { | |
4638 | close(dl->fd); | |
4639 | dl->fd = -1; | |
4640 | /* FIXME this doesn't free | |
4641 | * dl->devname */ | |
4642 | update->space = dl; | |
4643 | *dlp = dl->next; | |
4644 | break; | |
4645 | } | |
4646 | } | |
609ce161 | 4647 | ddf_set_updates_pending(ddf, NULL); |
4dd968cc N |
4648 | return; |
4649 | } | |
88c164f4 NB |
4650 | if (!all_ff(ddf->phys->entries[ent].guid)) |
4651 | return; | |
4652 | ddf->phys->entries[ent] = pd->entries[0]; | |
a8173e43 | 4653 | ddf->phys->used_pdes = cpu_to_be16 |
4654 | (1 + be16_to_cpu(ddf->phys->used_pdes)); | |
609ce161 | 4655 | ddf_set_updates_pending(ddf, NULL); |
2cc2983d N |
4656 | if (ddf->add_list) { |
4657 | struct active_array *a; | |
4658 | struct dl *al = ddf->add_list; | |
4659 | ddf->add_list = al->next; | |
4660 | ||
4661 | al->next = ddf->dlist; | |
4662 | ddf->dlist = al; | |
4663 | ||
4664 | /* As a device has been added, we should check | |
4665 | * for any degraded devices that might make | |
4666 | * use of this spare */ | |
4667 | for (a = st->arrays ; a; a=a->next) | |
4668 | a->check_degraded = 1; | |
4669 | } | |
60931cf9 | 4670 | } else if (be32_eq(*magic, DDF_VIRT_RECORDS_MAGIC)) { |
88c164f4 NB |
4671 | if (update->len != (sizeof(struct virtual_disk) + |
4672 | sizeof(struct virtual_entry))) | |
4673 | return; | |
4674 | vd = (struct virtual_disk*)update->buf; | |
4675 | ||
f646805e | 4676 | if (vd->entries[0].state == DDF_state_deleted) { |
4677 | if (_kill_subarray_ddf(ddf, vd->entries[0].guid)) | |
4678 | return; | |
4679 | } else { | |
6a7e7ecc | 4680 | ent = find_vde_by_guid(ddf, vd->entries[0].guid); |
4681 | if (ent != DDF_NOTFOUND) { | |
4682 | dprintf("%s: VD %s exists already in slot %d\n", | |
4683 | __func__, guid_str(vd->entries[0].guid), | |
4684 | ent); | |
4685 | return; | |
4686 | } | |
f646805e | 4687 | ent = find_unused_vde(ddf); |
4688 | if (ent == DDF_NOTFOUND) | |
4689 | return; | |
4690 | ddf->virt->entries[ent] = vd->entries[0]; | |
4691 | ddf->virt->populated_vdes = | |
a8173e43 | 4692 | cpu_to_be16( |
4693 | 1 + be16_to_cpu( | |
f646805e | 4694 | ddf->virt->populated_vdes)); |
ed5ff7a2 | 4695 | dprintf("%s: added VD %s in slot %d(s=%02x i=%02x)\n", |
4696 | __func__, guid_str(vd->entries[0].guid), ent, | |
4697 | ddf->virt->entries[ent].state, | |
4698 | ddf->virt->entries[ent].init_state); | |
f646805e | 4699 | } |
609ce161 | 4700 | ddf_set_updates_pending(ddf, NULL); |
60931cf9 | 4701 | } |
88c164f4 | 4702 | |
60931cf9 | 4703 | else if (be32_eq(*magic, DDF_VD_CONF_MAGIC)) { |
88c164f4 | 4704 | vc = (struct vd_config*)update->buf; |
c5943560 | 4705 | len = ddf->conf_rec_len * 512; |
4706 | if ((unsigned int)update->len != len * vc->sec_elmnt_count) { | |
4707 | pr_err("%s: %s: insufficient data (%d) for %u BVDs\n", | |
4708 | __func__, guid_str(vc->guid), update->len, | |
4709 | vc->sec_elmnt_count); | |
4710 | return; | |
4711 | } | |
88c164f4 NB |
4712 | for (vcl = ddf->conflist; vcl ; vcl = vcl->next) |
4713 | if (memcmp(vcl->conf.guid, vc->guid, DDF_GUID_LEN) == 0) | |
4714 | break; | |
ed5ff7a2 | 4715 | dprintf("%s: conf update for %s (%s)\n", __func__, |
4716 | guid_str(vc->guid), (vcl ? "old" : "new")); | |
88c164f4 NB |
4717 | if (vcl) { |
4718 | /* An update, just copy the phys_refnum and lba_offset | |
4719 | * fields | |
4720 | */ | |
c5943560 | 4721 | unsigned int i; |
0847945b | 4722 | unsigned int k; |
c5943560 | 4723 | copy_matching_bvd(ddf, &vcl->conf, update); |
0847945b MW |
4724 | for (k = 0; k < be16_to_cpu(vc->prim_elmnt_count); k++) |
4725 | dprintf("BVD %u has %08x at %llu\n", 0, | |
4726 | be32_to_cpu(vcl->conf.phys_refnum[k]), | |
4727 | be64_to_cpu(LBA_OFFSET(ddf, | |
4728 | &vcl->conf)[k])); | |
4729 | for (i = 1; i < vc->sec_elmnt_count; i++) { | |
c5943560 | 4730 | copy_matching_bvd(ddf, vcl->other_bvds[i-1], |
4731 | update); | |
0847945b MW |
4732 | for (k = 0; k < be16_to_cpu( |
4733 | vc->prim_elmnt_count); k++) | |
4734 | dprintf("BVD %u has %08x at %llu\n", i, | |
4735 | be32_to_cpu | |
4736 | (vcl->other_bvds[i-1]-> | |
4737 | phys_refnum[k]), | |
4738 | be64_to_cpu | |
4739 | (LBA_OFFSET | |
4740 | (ddf, | |
4741 | vcl->other_bvds[i-1])[k])); | |
4742 | } | |
88c164f4 NB |
4743 | } else { |
4744 | /* A new VD_CONF */ | |
c5943560 | 4745 | unsigned int i; |
e6b9548d DW |
4746 | if (!update->space) |
4747 | return; | |
88c164f4 NB |
4748 | vcl = update->space; |
4749 | update->space = NULL; | |
4750 | vcl->next = ddf->conflist; | |
c5943560 | 4751 | memcpy(&vcl->conf, vc, len); |
fb9d0acb | 4752 | ent = find_vde_by_guid(ddf, vc->guid); |
4753 | if (ent == DDF_NOTFOUND) | |
4754 | return; | |
4755 | vcl->vcnum = ent; | |
88c164f4 | 4756 | ddf->conflist = vcl; |
c5943560 | 4757 | for (i = 1; i < vc->sec_elmnt_count; i++) |
4758 | memcpy(vcl->other_bvds[i-1], | |
4759 | update->buf + len * i, len); | |
88c164f4 | 4760 | } |
c7079c84 N |
4761 | /* Set DDF_Transition on all Failed devices - to help |
4762 | * us detect those that are no longer in use | |
4763 | */ | |
217dead4 | 4764 | for (pdnum = 0; pdnum < be16_to_cpu(ddf->phys->max_pdes); |
a8173e43 | 4765 | pdnum++) |
4766 | if (be16_and(ddf->phys->entries[pdnum].state, | |
4767 | cpu_to_be16(DDF_Failed))) | |
4768 | be16_set(ddf->phys->entries[pdnum].state, | |
4769 | cpu_to_be16(DDF_Transition)); | |
88c164f4 NB |
4770 | /* Now make sure vlist is correct for each dl. */ |
4771 | for (dl = ddf->dlist; dl; dl = dl->next) { | |
f21e18ca | 4772 | unsigned int vn = 0; |
8401644c | 4773 | int in_degraded = 0; |
a44e993e N |
4774 | |
4775 | if (dl->pdnum < 0) | |
4776 | continue; | |
5838fccd | 4777 | for (vcl = ddf->conflist; vcl ; vcl = vcl->next) { |
4778 | unsigned int dn, ibvd; | |
4779 | const struct vd_config *conf; | |
4780 | int vstate; | |
4781 | dn = get_pd_index_from_refnum(vcl, | |
4782 | dl->disk.refnum, | |
4783 | ddf->mppe, | |
4784 | &conf, &ibvd); | |
4785 | if (dn == DDF_NOTFOUND) | |
4786 | continue; | |
4787 | dprintf("dev %d/%08x has %s (sec=%u) at %d\n", | |
ad60eea1 | 4788 | dl->pdnum, |
60931cf9 | 4789 | be32_to_cpu(dl->disk.refnum), |
5838fccd | 4790 | guid_str(conf->guid), |
4791 | conf->sec_elmnt_seq, vn); | |
4792 | /* Clear the Transition flag */ | |
a8173e43 | 4793 | if (be16_and |
4794 | (ddf->phys->entries[dl->pdnum].state, | |
4795 | cpu_to_be16(DDF_Failed))) | |
4796 | be16_clear(ddf->phys | |
4797 | ->entries[dl->pdnum].state, | |
4798 | cpu_to_be16(DDF_Transition)); | |
5838fccd | 4799 | dl->vlist[vn++] = vcl; |
4800 | vstate = ddf->virt->entries[vcl->vcnum].state | |
4801 | & DDF_state_mask; | |
4802 | if (vstate == DDF_state_degraded || | |
56cb05c4 | 4803 | vstate == DDF_state_part_optimal) |
5838fccd | 4804 | in_degraded = 1; |
4805 | } | |
88c164f4 NB |
4806 | while (vn < ddf->max_part) |
4807 | dl->vlist[vn++] = NULL; | |
7e1432fb | 4808 | if (dl->vlist[0]) { |
a8173e43 | 4809 | be16_clear(ddf->phys->entries[dl->pdnum].type, |
4810 | cpu_to_be16(DDF_Global_Spare)); | |
4811 | if (!be16_and(ddf->phys | |
4812 | ->entries[dl->pdnum].type, | |
4813 | cpu_to_be16(DDF_Active_in_VD))) { | |
4814 | be16_set(ddf->phys | |
4815 | ->entries[dl->pdnum].type, | |
4816 | cpu_to_be16(DDF_Active_in_VD)); | |
613b0d17 | 4817 | if (in_degraded) |
a8173e43 | 4818 | be16_set(ddf->phys |
4819 | ->entries[dl->pdnum] | |
4820 | .state, | |
4821 | cpu_to_be16 | |
4822 | (DDF_Rebuilding)); | |
613b0d17 | 4823 | } |
7e1432fb NB |
4824 | } |
4825 | if (dl->spare) { | |
a8173e43 | 4826 | be16_clear(ddf->phys->entries[dl->pdnum].type, |
4827 | cpu_to_be16(DDF_Global_Spare)); | |
4828 | be16_set(ddf->phys->entries[dl->pdnum].type, | |
4829 | cpu_to_be16(DDF_Spare)); | |
7e1432fb NB |
4830 | } |
4831 | if (!dl->vlist[0] && !dl->spare) { | |
a8173e43 | 4832 | be16_set(ddf->phys->entries[dl->pdnum].type, |
4833 | cpu_to_be16(DDF_Global_Spare)); | |
4834 | be16_clear(ddf->phys->entries[dl->pdnum].type, | |
4835 | cpu_to_be16(DDF_Spare)); | |
4836 | be16_clear(ddf->phys->entries[dl->pdnum].type, | |
4837 | cpu_to_be16(DDF_Active_in_VD)); | |
7e1432fb | 4838 | } |
88c164f4 | 4839 | } |
c7079c84 N |
4840 | |
4841 | /* Now remove any 'Failed' devices that are not part | |
4842 | * of any VD. They will have the Transition flag set. | |
4843 | * Once done, we need to update all dl->pdnum numbers. | |
4844 | */ | |
4845 | pd2 = 0; | |
217dead4 | 4846 | for (pdnum = 0; pdnum < be16_to_cpu(ddf->phys->max_pdes); |
92939eb2 | 4847 | pdnum++) { |
217dead4 N |
4848 | if (be32_to_cpu(ddf->phys->entries[pdnum].refnum) == |
4849 | 0xFFFFFFFF) | |
4850 | continue; | |
a8173e43 | 4851 | if (be16_and(ddf->phys->entries[pdnum].state, |
4852 | cpu_to_be16(DDF_Failed)) | |
4853 | && be16_and(ddf->phys->entries[pdnum].state, | |
92939eb2 N |
4854 | cpu_to_be16(DDF_Transition))) { |
4855 | /* skip this one unless in dlist*/ | |
4856 | for (dl = ddf->dlist; dl; dl = dl->next) | |
4857 | if (dl->pdnum == (int)pdnum) | |
4858 | break; | |
4859 | if (!dl) | |
4860 | continue; | |
4861 | } | |
4862 | if (pdnum == pd2) | |
c7079c84 N |
4863 | pd2++; |
4864 | else { | |
a8173e43 | 4865 | ddf->phys->entries[pd2] = |
4866 | ddf->phys->entries[pdnum]; | |
c7079c84 N |
4867 | for (dl = ddf->dlist; dl; dl = dl->next) |
4868 | if (dl->pdnum == (int)pdnum) | |
4869 | dl->pdnum = pd2; | |
4870 | pd2++; | |
4871 | } | |
92939eb2 | 4872 | } |
a8173e43 | 4873 | ddf->phys->used_pdes = cpu_to_be16(pd2); |
c7079c84 | 4874 | while (pd2 < pdnum) { |
a8173e43 | 4875 | memset(ddf->phys->entries[pd2].guid, 0xff, |
4876 | DDF_GUID_LEN); | |
c7079c84 N |
4877 | pd2++; |
4878 | } | |
4879 | ||
609ce161 | 4880 | ddf_set_updates_pending(ddf, vc); |
88c164f4 | 4881 | } |
60931cf9 | 4882 | /* case DDF_SPARE_ASSIGN_MAGIC */ |
88c164f4 NB |
4883 | } |
4884 | ||
edd8d13c NB |
4885 | static void ddf_prepare_update(struct supertype *st, |
4886 | struct metadata_update *update) | |
4887 | { | |
4888 | /* This update arrived at managemon. | |
4889 | * We are about to pass it to monitor. | |
4890 | * If a malloc is needed, do it here. | |
4891 | */ | |
4892 | struct ddf_super *ddf = st->sb; | |
60931cf9 | 4893 | be32 *magic = (be32 *)update->buf; |
4894 | if (be32_eq(*magic, DDF_VD_CONF_MAGIC)) { | |
c5943560 | 4895 | struct vcl *vcl; |
4896 | struct vd_config *conf = (struct vd_config *) update->buf; | |
e6b9548d | 4897 | if (posix_memalign(&update->space, 512, |
613b0d17 | 4898 | offsetof(struct vcl, conf) |
c5943560 | 4899 | + ddf->conf_rec_len * 512) != 0) { |
4900 | update->space = NULL; | |
4901 | return; | |
4902 | } | |
4903 | vcl = update->space; | |
4904 | vcl->conf.sec_elmnt_count = conf->sec_elmnt_count; | |
4905 | if (alloc_other_bvds(ddf, vcl) != 0) { | |
4906 | free(update->space); | |
e6b9548d | 4907 | update->space = NULL; |
c5943560 | 4908 | } |
4909 | } | |
edd8d13c NB |
4910 | } |
4911 | ||
7733b91d | 4912 | /* |
4913 | * Check degraded state of a RAID10. | |
4914 | * returns 2 for good, 1 for degraded, 0 for failed, and -1 for error | |
4915 | */ | |
4916 | static int raid10_degraded(struct mdinfo *info) | |
4917 | { | |
4918 | int n_prim, n_bvds; | |
4919 | int i; | |
9591a2de | 4920 | struct mdinfo *d; |
7733b91d | 4921 | char *found; |
4922 | int ret = -1; | |
4923 | ||
7733b91d | 4924 | n_prim = info->array.layout & ~0x100; |
4925 | n_bvds = info->array.raid_disks / n_prim; | |
4926 | found = xmalloc(n_bvds); | |
4927 | if (found == NULL) | |
4928 | return ret; | |
4929 | memset(found, 0, n_bvds); | |
4930 | for (d = info->devs; d; d = d->next) { | |
4931 | i = d->disk.raid_disk / n_prim; | |
4932 | if (i >= n_bvds) { | |
4933 | pr_err("%s: BUG: invalid raid disk\n", __func__); | |
4934 | goto out; | |
4935 | } | |
4936 | if (d->state_fd > 0) | |
4937 | found[i]++; | |
4938 | } | |
4939 | ret = 2; | |
4940 | for (i = 0; i < n_bvds; i++) | |
4941 | if (!found[i]) { | |
4942 | dprintf("%s: BVD %d/%d failed\n", __func__, i, n_bvds); | |
4943 | ret = 0; | |
4944 | goto out; | |
4945 | } else if (found[i] < n_prim) { | |
4946 | dprintf("%s: BVD %d/%d degraded\n", __func__, i, | |
4947 | n_bvds); | |
4948 | ret = 1; | |
4949 | } | |
4950 | out: | |
4951 | free(found); | |
4952 | return ret; | |
4953 | } | |
4954 | ||
7e1432fb NB |
4955 | /* |
4956 | * Check if the array 'a' is degraded but not failed. | |
4957 | * If it is, find as many spares as are available and needed and | |
4958 | * arrange for their inclusion. | |
4959 | * We only choose devices which are not already in the array, | |
4960 | * and prefer those with a spare-assignment to this array. | |
56cb05c4 | 4961 | * Otherwise we choose global spares - assuming always that |
7e1432fb NB |
4962 | * there is enough room. |
4963 | * For each spare that we assign, we return an 'mdinfo' which | |
4964 | * describes the position for the device in the array. | |
4965 | * We also add to 'updates' a DDF_VD_CONF_MAGIC update with | |
4966 | * the new phys_refnum and lba_offset values. | |
4967 | * | |
4968 | * Only worry about BVDs at the moment. | |
4969 | */ | |
4970 | static struct mdinfo *ddf_activate_spare(struct active_array *a, | |
4971 | struct metadata_update **updates) | |
4972 | { | |
4973 | int working = 0; | |
4974 | struct mdinfo *d; | |
4975 | struct ddf_super *ddf = a->container->sb; | |
4976 | int global_ok = 0; | |
4977 | struct mdinfo *rv = NULL; | |
4978 | struct mdinfo *di; | |
4979 | struct metadata_update *mu; | |
4980 | struct dl *dl; | |
4981 | int i; | |
0c78849f | 4982 | unsigned int j; |
baba3f4e | 4983 | struct vcl *vcl; |
7e1432fb | 4984 | struct vd_config *vc; |
baba3f4e | 4985 | unsigned int n_bvd; |
7e1432fb | 4986 | |
7e1432fb NB |
4987 | for (d = a->info.devs ; d ; d = d->next) { |
4988 | if ((d->curr_state & DS_FAULTY) && | |
613b0d17 | 4989 | d->state_fd >= 0) |
7e1432fb NB |
4990 | /* wait for Removal to happen */ |
4991 | return NULL; | |
4992 | if (d->state_fd >= 0) | |
4993 | working ++; | |
4994 | } | |
4995 | ||
7733b91d | 4996 | dprintf("%s: working=%d (%d) level=%d\n", __func__, working, |
a8173e43 | 4997 | a->info.array.raid_disks, |
2c514b71 | 4998 | a->info.array.level); |
7e1432fb NB |
4999 | if (working == a->info.array.raid_disks) |
5000 | return NULL; /* array not degraded */ | |
5001 | switch (a->info.array.level) { | |
5002 | case 1: | |
5003 | if (working == 0) | |
5004 | return NULL; /* failed */ | |
5005 | break; | |
5006 | case 4: | |
5007 | case 5: | |
5008 | if (working < a->info.array.raid_disks - 1) | |
5009 | return NULL; /* failed */ | |
5010 | break; | |
5011 | case 6: | |
5012 | if (working < a->info.array.raid_disks - 2) | |
5013 | return NULL; /* failed */ | |
5014 | break; | |
7733b91d | 5015 | case 10: |
5016 | if (raid10_degraded(&a->info) < 1) | |
5017 | return NULL; | |
5018 | break; | |
7e1432fb NB |
5019 | default: /* concat or stripe */ |
5020 | return NULL; /* failed */ | |
5021 | } | |
5022 | ||
5023 | /* For each slot, if it is not working, find a spare */ | |
5024 | dl = ddf->dlist; | |
5025 | for (i = 0; i < a->info.array.raid_disks; i++) { | |
5026 | for (d = a->info.devs ; d ; d = d->next) | |
5027 | if (d->disk.raid_disk == i) | |
5028 | break; | |
2c514b71 | 5029 | dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0); |
7e1432fb NB |
5030 | if (d && (d->state_fd >= 0)) |
5031 | continue; | |
5032 | ||
5033 | /* OK, this device needs recovery. Find a spare */ | |
5034 | again: | |
5035 | for ( ; dl ; dl = dl->next) { | |
5036 | unsigned long long esize; | |
5037 | unsigned long long pos; | |
5038 | struct mdinfo *d2; | |
5039 | int is_global = 0; | |
5040 | int is_dedicated = 0; | |
a44e993e N |
5041 | be16 state; |
5042 | ||
5043 | if (dl->pdnum < 0) | |
5044 | continue; | |
5045 | state = ddf->phys->entries[dl->pdnum].state; | |
6f56dbb9 MW |
5046 | if (be16_and(state, |
5047 | cpu_to_be16(DDF_Failed|DDF_Missing)) || | |
5048 | !be16_and(state, | |
5049 | cpu_to_be16(DDF_Online))) | |
5050 | continue; | |
5051 | ||
7e1432fb NB |
5052 | /* If in this array, skip */ |
5053 | for (d2 = a->info.devs ; d2 ; d2 = d2->next) | |
7590d562 N |
5054 | if (d2->state_fd >= 0 && |
5055 | d2->disk.major == dl->major && | |
7e1432fb | 5056 | d2->disk.minor == dl->minor) { |
2a645ee2 MW |
5057 | dprintf("%x:%x (%08x) already in array\n", |
5058 | dl->major, dl->minor, | |
5059 | be32_to_cpu(dl->disk.refnum)); | |
7e1432fb NB |
5060 | break; |
5061 | } | |
5062 | if (d2) | |
5063 | continue; | |
a8173e43 | 5064 | if (be16_and(ddf->phys->entries[dl->pdnum].type, |
5065 | cpu_to_be16(DDF_Spare))) { | |
7e1432fb NB |
5066 | /* Check spare assign record */ |
5067 | if (dl->spare) { | |
5068 | if (dl->spare->type & DDF_spare_dedicated) { | |
5069 | /* check spare_ents for guid */ | |
fca65520 | 5070 | unsigned int j; |
7e1432fb | 5071 | for (j = 0 ; |
a8173e43 | 5072 | j < be16_to_cpu |
5073 | (dl->spare | |
5074 | ->populated); | |
7e1432fb NB |
5075 | j++) { |
5076 | if (memcmp(dl->spare->spare_ents[j].guid, | |
5077 | ddf->virt->entries[a->info.container_member].guid, | |
5078 | DDF_GUID_LEN) == 0) | |
5079 | is_dedicated = 1; | |
5080 | } | |
5081 | } else | |
5082 | is_global = 1; | |
5083 | } | |
a8173e43 | 5084 | } else if (be16_and(ddf->phys->entries[dl->pdnum].type, |
5085 | cpu_to_be16(DDF_Global_Spare))) { | |
7e1432fb | 5086 | is_global = 1; |
a8173e43 | 5087 | } else if (!be16_and(ddf->phys |
5088 | ->entries[dl->pdnum].state, | |
5089 | cpu_to_be16(DDF_Failed))) { | |
e0e7aeaa N |
5090 | /* we can possibly use some of this */ |
5091 | is_global = 1; | |
7e1432fb NB |
5092 | } |
5093 | if ( ! (is_dedicated || | |
5094 | (is_global && global_ok))) { | |
2c514b71 | 5095 | dprintf("%x:%x not suitable: %d %d\n", dl->major, dl->minor, |
613b0d17 | 5096 | is_dedicated, is_global); |
7e1432fb NB |
5097 | continue; |
5098 | } | |
5099 | ||
5100 | /* We are allowed to use this device - is there space? | |
5101 | * We need a->info.component_size sectors */ | |
fca65520 N |
5102 | esize = a->info.component_size; |
5103 | pos = find_space(ddf, dl, INVALID_SECTORS, &esize); | |
7e1432fb | 5104 | |
7e1432fb | 5105 | if (esize < a->info.component_size) { |
e5cc7d46 N |
5106 | dprintf("%x:%x has no room: %llu %llu\n", |
5107 | dl->major, dl->minor, | |
2c514b71 | 5108 | esize, a->info.component_size); |
7e1432fb NB |
5109 | /* No room */ |
5110 | continue; | |
5111 | } | |
5112 | ||
5113 | /* Cool, we have a device with some space at pos */ | |
503975b9 | 5114 | di = xcalloc(1, sizeof(*di)); |
7e1432fb NB |
5115 | di->disk.number = i; |
5116 | di->disk.raid_disk = i; | |
5117 | di->disk.major = dl->major; | |
5118 | di->disk.minor = dl->minor; | |
5119 | di->disk.state = 0; | |
d23534e4 | 5120 | di->recovery_start = 0; |
7e1432fb NB |
5121 | di->data_offset = pos; |
5122 | di->component_size = a->info.component_size; | |
7e1432fb NB |
5123 | di->next = rv; |
5124 | rv = di; | |
2a645ee2 MW |
5125 | dprintf("%x:%x (%08x) to be %d at %llu\n", |
5126 | dl->major, dl->minor, | |
5127 | be32_to_cpu(dl->disk.refnum), i, pos); | |
7e1432fb NB |
5128 | |
5129 | break; | |
5130 | } | |
5131 | if (!dl && ! global_ok) { | |
5132 | /* not enough dedicated spares, try global */ | |
5133 | global_ok = 1; | |
5134 | dl = ddf->dlist; | |
5135 | goto again; | |
5136 | } | |
5137 | } | |
5138 | ||
5139 | if (!rv) | |
5140 | /* No spares found */ | |
5141 | return rv; | |
5142 | /* Now 'rv' has a list of devices to return. | |
5143 | * Create a metadata_update record to update the | |
5144 | * phys_refnum and lba_offset values | |
5145 | */ | |
bb925ff0 | 5146 | vc = find_vdcr(ddf, a->info.container_member, rv->disk.raid_disk, |
0c78849f | 5147 | &n_bvd, &vcl); |
5148 | if (vc == NULL) | |
5149 | return NULL; | |
5150 | ||
503975b9 N |
5151 | mu = xmalloc(sizeof(*mu)); |
5152 | if (posix_memalign(&mu->space, 512, sizeof(struct vcl)) != 0) { | |
79244939 DW |
5153 | free(mu); |
5154 | mu = NULL; | |
5155 | } | |
0c78849f | 5156 | |
5157 | mu->len = ddf->conf_rec_len * 512 * vcl->conf.sec_elmnt_count; | |
5158 | mu->buf = xmalloc(mu->len); | |
7590d562 | 5159 | mu->space = NULL; |
f50ae22e | 5160 | mu->space_list = NULL; |
7e1432fb | 5161 | mu->next = *updates; |
0c78849f | 5162 | memcpy(mu->buf, &vcl->conf, ddf->conf_rec_len * 512); |
5163 | for (j = 1; j < vcl->conf.sec_elmnt_count; j++) | |
5164 | memcpy(mu->buf + j * ddf->conf_rec_len * 512, | |
5165 | vcl->other_bvds[j-1], ddf->conf_rec_len * 512); | |
7e1432fb NB |
5166 | |
5167 | vc = (struct vd_config*)mu->buf; | |
7e1432fb | 5168 | for (di = rv ; di ; di = di->next) { |
0c78849f | 5169 | unsigned int i_sec, i_prim; |
5170 | i_sec = di->disk.raid_disk | |
5171 | / be16_to_cpu(vcl->conf.prim_elmnt_count); | |
5172 | i_prim = di->disk.raid_disk | |
5173 | % be16_to_cpu(vcl->conf.prim_elmnt_count); | |
5174 | vc = (struct vd_config *)(mu->buf | |
5175 | + i_sec * ddf->conf_rec_len * 512); | |
5176 | for (dl = ddf->dlist; dl; dl = dl->next) | |
5177 | if (dl->major == di->disk.major | |
5178 | && dl->minor == di->disk.minor) | |
5179 | break; | |
a44e993e | 5180 | if (!dl || dl->pdnum < 0) { |
0c78849f | 5181 | pr_err("%s: BUG: can't find disk %d (%d/%d)\n", |
5182 | __func__, di->disk.raid_disk, | |
5183 | di->disk.major, di->disk.minor); | |
5184 | return NULL; | |
5185 | } | |
5186 | vc->phys_refnum[i_prim] = ddf->phys->entries[dl->pdnum].refnum; | |
5187 | LBA_OFFSET(ddf, vc)[i_prim] = cpu_to_be64(di->data_offset); | |
2a645ee2 MW |
5188 | dprintf("BVD %u gets %u: %08x at %llu\n", i_sec, i_prim, |
5189 | be32_to_cpu(vc->phys_refnum[i_prim]), | |
5190 | be64_to_cpu(LBA_OFFSET(ddf, vc)[i_prim])); | |
7e1432fb NB |
5191 | } |
5192 | *updates = mu; | |
5193 | return rv; | |
5194 | } | |
0e600426 | 5195 | #endif /* MDASSEMBLE */ |
7e1432fb | 5196 | |
b640a252 N |
5197 | static int ddf_level_to_layout(int level) |
5198 | { | |
5199 | switch(level) { | |
5200 | case 0: | |
5201 | case 1: | |
5202 | return 0; | |
5203 | case 5: | |
5204 | return ALGORITHM_LEFT_SYMMETRIC; | |
5205 | case 6: | |
5206 | return ALGORITHM_ROTATING_N_CONTINUE; | |
5207 | case 10: | |
5208 | return 0x102; | |
5209 | default: | |
5210 | return UnSet; | |
5211 | } | |
5212 | } | |
5213 | ||
30f58b22 DW |
5214 | static void default_geometry_ddf(struct supertype *st, int *level, int *layout, int *chunk) |
5215 | { | |
5216 | if (level && *level == UnSet) | |
5217 | *level = LEVEL_CONTAINER; | |
5218 | ||
5219 | if (level && layout && *layout == UnSet) | |
5220 | *layout = ddf_level_to_layout(*level); | |
5221 | } | |
5222 | ||
a322f70c DW |
5223 | struct superswitch super_ddf = { |
5224 | #ifndef MDASSEMBLE | |
5225 | .examine_super = examine_super_ddf, | |
5226 | .brief_examine_super = brief_examine_super_ddf, | |
4737ae25 | 5227 | .brief_examine_subarrays = brief_examine_subarrays_ddf, |
bceedeec | 5228 | .export_examine_super = export_examine_super_ddf, |
a322f70c DW |
5229 | .detail_super = detail_super_ddf, |
5230 | .brief_detail_super = brief_detail_super_ddf, | |
5231 | .validate_geometry = validate_geometry_ddf, | |
78e44928 | 5232 | .write_init_super = write_init_super_ddf, |
0e600426 | 5233 | .add_to_super = add_to_super_ddf, |
4dd968cc | 5234 | .remove_from_super = remove_from_super_ddf, |
2b959fbf | 5235 | .load_container = load_container_ddf, |
74db60b0 | 5236 | .copy_metadata = copy_metadata_ddf, |
4441541f | 5237 | .kill_subarray = kill_subarray_ddf, |
a322f70c DW |
5238 | #endif |
5239 | .match_home = match_home_ddf, | |
5240 | .uuid_from_super= uuid_from_super_ddf, | |
5241 | .getinfo_super = getinfo_super_ddf, | |
5242 | .update_super = update_super_ddf, | |
5243 | ||
5244 | .avail_size = avail_size_ddf, | |
5245 | ||
a19c88b8 NB |
5246 | .compare_super = compare_super_ddf, |
5247 | ||
a322f70c | 5248 | .load_super = load_super_ddf, |
ba7eb04f | 5249 | .init_super = init_super_ddf, |
955e9ea1 | 5250 | .store_super = store_super_ddf, |
a322f70c DW |
5251 | .free_super = free_super_ddf, |
5252 | .match_metadata_desc = match_metadata_desc_ddf, | |
78e44928 | 5253 | .container_content = container_content_ddf, |
30f58b22 | 5254 | .default_geometry = default_geometry_ddf, |
a322f70c | 5255 | |
a322f70c | 5256 | .external = 1, |
549e9569 | 5257 | |
0e600426 | 5258 | #ifndef MDASSEMBLE |
549e9569 NB |
5259 | /* for mdmon */ |
5260 | .open_new = ddf_open_new, | |
ed9d66aa | 5261 | .set_array_state= ddf_set_array_state, |
549e9569 NB |
5262 | .set_disk = ddf_set_disk, |
5263 | .sync_metadata = ddf_sync_metadata, | |
88c164f4 | 5264 | .process_update = ddf_process_update, |
edd8d13c | 5265 | .prepare_update = ddf_prepare_update, |
7e1432fb | 5266 | .activate_spare = ddf_activate_spare, |
0e600426 | 5267 | #endif |
4cce4069 | 5268 | .name = "ddf", |
a322f70c | 5269 | }; |