]>
Commit | Line | Data |
---|---|---|
a322f70c DW |
1 | /* |
2 | * mdadm - manage Linux "md" devices aka RAID arrays. | |
3 | * | |
e736b623 | 4 | * Copyright (C) 2006-2009 Neil Brown <neilb@suse.de> |
a322f70c DW |
5 | * |
6 | * | |
7 | * This program is free software; you can redistribute it and/or modify | |
8 | * it under the terms of the GNU General Public License as published by | |
9 | * the Free Software Foundation; either version 2 of the License, or | |
10 | * (at your option) any later version. | |
11 | * | |
12 | * This program is distributed in the hope that it will be useful, | |
13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
15 | * GNU General Public License for more details. | |
16 | * | |
17 | * You should have received a copy of the GNU General Public License | |
18 | * along with this program; if not, write to the Free Software | |
19 | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA | |
20 | * | |
21 | * Author: Neil Brown | |
22 | * Email: <neil@brown.name> | |
23 | * | |
24 | * Specifications for DDF takes from Common RAID DDF Specification Revision 1.2 | |
25 | * (July 28 2006). Reused by permission of SNIA. | |
26 | */ | |
27 | ||
28 | #define HAVE_STDINT_H 1 | |
29 | #include "mdadm.h" | |
549e9569 | 30 | #include "mdmon.h" |
a322f70c DW |
31 | #include "sha1.h" |
32 | #include <values.h> | |
33 | ||
a322f70c DW |
34 | /* a non-official T10 name for creation GUIDs */ |
35 | static char T10[] = "Linux-MD"; | |
36 | ||
37 | /* DDF timestamps are 1980 based, so we need to add | |
38 | * second-in-decade-of-seventies to convert to linux timestamps. | |
39 | * 10 years with 2 leap years. | |
40 | */ | |
41 | #define DECADE (3600*24*(365*10+2)) | |
42 | unsigned long crc32( | |
43 | unsigned long crc, | |
44 | const unsigned char *buf, | |
45 | unsigned len); | |
46 | ||
bedbf68a | 47 | #define DDF_NOTFOUND (~0U) |
48 | #define DDF_CONTAINER (DDF_NOTFOUND-1) | |
49 | ||
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; | |
a8173e43 | 232 | be16 used_pdes; |
233 | be16 max_pdes; | |
a322f70c DW |
234 | __u8 pad[52]; |
235 | struct phys_disk_entry { | |
236 | char guid[DDF_GUID_LEN]; | |
60931cf9 | 237 | be32 refnum; |
a8173e43 | 238 | be16 type; |
239 | be16 state; | |
9d0c6b70 | 240 | be64 config_size; /* DDF structures must be after here */ |
a322f70c DW |
241 | char path[18]; /* another horrible structure really */ |
242 | __u8 pad[6]; | |
243 | } entries[0]; | |
244 | }; | |
245 | ||
246 | /* phys_disk_entry.type is a bitmap - bigendian remember */ | |
247 | #define DDF_Forced_PD_GUID 1 | |
248 | #define DDF_Active_in_VD 2 | |
88c164f4 | 249 | #define DDF_Global_Spare 4 /* VD_CONF records are ignored */ |
a322f70c DW |
250 | #define DDF_Spare 8 /* overrides Global_spare */ |
251 | #define DDF_Foreign 16 | |
252 | #define DDF_Legacy 32 /* no DDF on this device */ | |
253 | ||
254 | #define DDF_Interface_mask 0xf00 | |
255 | #define DDF_Interface_SCSI 0x100 | |
256 | #define DDF_Interface_SAS 0x200 | |
257 | #define DDF_Interface_SATA 0x300 | |
258 | #define DDF_Interface_FC 0x400 | |
259 | ||
260 | /* phys_disk_entry.state is a bigendian bitmap */ | |
261 | #define DDF_Online 1 | |
262 | #define DDF_Failed 2 /* overrides 1,4,8 */ | |
263 | #define DDF_Rebuilding 4 | |
264 | #define DDF_Transition 8 | |
265 | #define DDF_SMART 16 | |
266 | #define DDF_ReadErrors 32 | |
267 | #define DDF_Missing 64 | |
268 | ||
269 | /* The content of the virt_section global scope */ | |
270 | struct virtual_disk { | |
60931cf9 | 271 | be32 magic; /* DDF_VIRT_RECORDS_MAGIC */ |
272 | be32 crc; | |
a8173e43 | 273 | be16 populated_vdes; |
274 | be16 max_vdes; | |
a322f70c DW |
275 | __u8 pad[52]; |
276 | struct virtual_entry { | |
277 | char guid[DDF_GUID_LEN]; | |
a8173e43 | 278 | be16 unit; |
a322f70c | 279 | __u16 pad0; /* 0xffff */ |
a8173e43 | 280 | be16 guid_crc; |
281 | be16 type; | |
a322f70c DW |
282 | __u8 state; |
283 | __u8 init_state; | |
284 | __u8 pad1[14]; | |
285 | char name[16]; | |
286 | } entries[0]; | |
287 | }; | |
288 | ||
289 | /* virtual_entry.type is a bitmap - bigendian */ | |
290 | #define DDF_Shared 1 | |
291 | #define DDF_Enforce_Groups 2 | |
292 | #define DDF_Unicode 4 | |
293 | #define DDF_Owner_Valid 8 | |
294 | ||
295 | /* virtual_entry.state is a bigendian bitmap */ | |
296 | #define DDF_state_mask 0x7 | |
297 | #define DDF_state_optimal 0x0 | |
298 | #define DDF_state_degraded 0x1 | |
299 | #define DDF_state_deleted 0x2 | |
300 | #define DDF_state_missing 0x3 | |
301 | #define DDF_state_failed 0x4 | |
7a7cc504 | 302 | #define DDF_state_part_optimal 0x5 |
a322f70c DW |
303 | |
304 | #define DDF_state_morphing 0x8 | |
305 | #define DDF_state_inconsistent 0x10 | |
306 | ||
307 | /* virtual_entry.init_state is a bigendian bitmap */ | |
308 | #define DDF_initstate_mask 0x03 | |
309 | #define DDF_init_not 0x00 | |
7a7cc504 NB |
310 | #define DDF_init_quick 0x01 /* initialisation is progress. |
311 | * i.e. 'state_inconsistent' */ | |
a322f70c DW |
312 | #define DDF_init_full 0x02 |
313 | ||
314 | #define DDF_access_mask 0xc0 | |
315 | #define DDF_access_rw 0x00 | |
316 | #define DDF_access_ro 0x80 | |
317 | #define DDF_access_blocked 0xc0 | |
318 | ||
319 | /* The content of the config_section - local scope | |
320 | * It has multiple records each config_record_len sectors | |
321 | * They can be vd_config or spare_assign | |
322 | */ | |
323 | ||
324 | struct vd_config { | |
60931cf9 | 325 | be32 magic; /* DDF_VD_CONF_MAGIC */ |
326 | be32 crc; | |
a322f70c | 327 | char guid[DDF_GUID_LEN]; |
60931cf9 | 328 | be32 timestamp; |
329 | be32 seqnum; | |
a322f70c | 330 | __u8 pad0[24]; |
a8173e43 | 331 | be16 prim_elmnt_count; |
a322f70c DW |
332 | __u8 chunk_shift; /* 0 == 512, 1==1024 etc */ |
333 | __u8 prl; | |
334 | __u8 rlq; | |
335 | __u8 sec_elmnt_count; | |
336 | __u8 sec_elmnt_seq; | |
337 | __u8 srl; | |
9d0c6b70 | 338 | be64 blocks; /* blocks per component could be different |
598f0d58 NB |
339 | * on different component devices...(only |
340 | * for concat I hope) */ | |
9d0c6b70 | 341 | be64 array_blocks; /* blocks in array */ |
a322f70c | 342 | __u8 pad1[8]; |
60931cf9 | 343 | be32 spare_refs[8]; |
a322f70c DW |
344 | __u8 cache_pol[8]; |
345 | __u8 bg_rate; | |
346 | __u8 pad2[3]; | |
347 | __u8 pad3[52]; | |
348 | __u8 pad4[192]; | |
349 | __u8 v0[32]; /* reserved- 0xff */ | |
350 | __u8 v1[32]; /* reserved- 0xff */ | |
351 | __u8 v2[16]; /* reserved- 0xff */ | |
352 | __u8 v3[16]; /* reserved- 0xff */ | |
353 | __u8 vendor[32]; | |
60931cf9 | 354 | be32 phys_refnum[0]; /* refnum of each disk in sequence */ |
a322f70c DW |
355 | /*__u64 lba_offset[0]; LBA offset in each phys. Note extents in a |
356 | bvd are always the same size */ | |
357 | }; | |
9d0c6b70 | 358 | #define LBA_OFFSET(ddf, vd) ((be64 *) &(vd)->phys_refnum[(ddf)->mppe]) |
a322f70c DW |
359 | |
360 | /* vd_config.cache_pol[7] is a bitmap */ | |
361 | #define DDF_cache_writeback 1 /* else writethrough */ | |
362 | #define DDF_cache_wadaptive 2 /* only applies if writeback */ | |
363 | #define DDF_cache_readahead 4 | |
364 | #define DDF_cache_radaptive 8 /* only if doing read-ahead */ | |
365 | #define DDF_cache_ifnobatt 16 /* even to write cache if battery is poor */ | |
366 | #define DDF_cache_wallowed 32 /* enable write caching */ | |
367 | #define DDF_cache_rallowed 64 /* enable read caching */ | |
368 | ||
369 | struct spare_assign { | |
60931cf9 | 370 | be32 magic; /* DDF_SPARE_ASSIGN_MAGIC */ |
371 | be32 crc; | |
372 | be32 timestamp; | |
a322f70c DW |
373 | __u8 reserved[7]; |
374 | __u8 type; | |
a8173e43 | 375 | be16 populated; /* SAEs used */ |
376 | be16 max; /* max SAEs */ | |
a322f70c DW |
377 | __u8 pad[8]; |
378 | struct spare_assign_entry { | |
379 | char guid[DDF_GUID_LEN]; | |
a8173e43 | 380 | be16 secondary_element; |
a322f70c DW |
381 | __u8 pad[6]; |
382 | } spare_ents[0]; | |
383 | }; | |
384 | /* spare_assign.type is a bitmap */ | |
385 | #define DDF_spare_dedicated 0x1 /* else global */ | |
386 | #define DDF_spare_revertible 0x2 /* else committable */ | |
387 | #define DDF_spare_active 0x4 /* else not active */ | |
388 | #define DDF_spare_affinity 0x8 /* enclosure affinity */ | |
389 | ||
390 | /* The data_section contents - local scope */ | |
391 | struct disk_data { | |
60931cf9 | 392 | be32 magic; /* DDF_PHYS_DATA_MAGIC */ |
393 | be32 crc; | |
a322f70c | 394 | char guid[DDF_GUID_LEN]; |
60931cf9 | 395 | be32 refnum; /* crc of some magic drive data ... */ |
a322f70c DW |
396 | __u8 forced_ref; /* set when above was not result of magic */ |
397 | __u8 forced_guid; /* set if guid was forced rather than magic */ | |
398 | __u8 vendor[32]; | |
399 | __u8 pad[442]; | |
400 | }; | |
401 | ||
402 | /* bbm_section content */ | |
403 | struct bad_block_log { | |
60931cf9 | 404 | be32 magic; |
405 | be32 crc; | |
a8173e43 | 406 | be16 entry_count; |
60931cf9 | 407 | be32 spare_count; |
a322f70c | 408 | __u8 pad[10]; |
9d0c6b70 | 409 | be64 first_spare; |
a322f70c | 410 | struct mapped_block { |
9d0c6b70 | 411 | be64 defective_start; |
60931cf9 | 412 | be32 replacement_start; |
a8173e43 | 413 | be16 remap_count; |
a322f70c DW |
414 | __u8 pad[2]; |
415 | } entries[0]; | |
416 | }; | |
417 | ||
418 | /* Struct for internally holding ddf structures */ | |
419 | /* The DDF structure stored on each device is potentially | |
420 | * quite different, as some data is global and some is local. | |
421 | * The global data is: | |
422 | * - ddf header | |
423 | * - controller_data | |
424 | * - Physical disk records | |
425 | * - Virtual disk records | |
426 | * The local data is: | |
427 | * - Configuration records | |
428 | * - Physical Disk data section | |
429 | * ( and Bad block and vendor which I don't care about yet). | |
430 | * | |
431 | * The local data is parsed into separate lists as it is read | |
432 | * and reconstructed for writing. This means that we only need | |
433 | * to make config changes once and they are automatically | |
434 | * propagated to all devices. | |
435 | * Note that the ddf_super has space of the conf and disk data | |
436 | * for this disk and also for a list of all such data. | |
437 | * The list is only used for the superblock that is being | |
438 | * built in Create or Assemble to describe the whole array. | |
439 | */ | |
440 | struct ddf_super { | |
6416d527 | 441 | struct ddf_header anchor, primary, secondary; |
a322f70c | 442 | struct ddf_controller_data controller; |
6416d527 | 443 | struct ddf_header *active; |
a322f70c DW |
444 | struct phys_disk *phys; |
445 | struct virtual_disk *virt; | |
3921e41a | 446 | char *conf; |
a322f70c | 447 | int pdsize, vdsize; |
f21e18ca | 448 | unsigned int max_part, mppe, conf_rec_len; |
d2ca6449 | 449 | int currentdev; |
18a2f463 | 450 | int updates_pending; |
a322f70c | 451 | struct vcl { |
6416d527 NB |
452 | union { |
453 | char space[512]; | |
454 | struct { | |
455 | struct vcl *next; | |
f21e18ca | 456 | unsigned int vcnum; /* index into ->virt */ |
8ec5d685 | 457 | struct vd_config **other_bvds; |
6416d527 NB |
458 | __u64 *block_sizes; /* NULL if all the same */ |
459 | }; | |
460 | }; | |
a322f70c | 461 | struct vd_config conf; |
d2ca6449 | 462 | } *conflist, *currentconf; |
a322f70c | 463 | struct dl { |
6416d527 NB |
464 | union { |
465 | char space[512]; | |
466 | struct { | |
467 | struct dl *next; | |
468 | int major, minor; | |
469 | char *devname; | |
470 | int fd; | |
471 | unsigned long long size; /* sectors */ | |
9d0c6b70 | 472 | be64 primary_lba; /* sectors */ |
473 | be64 secondary_lba; /* sectors */ | |
474 | be64 workspace_lba; /* sectors */ | |
6416d527 NB |
475 | int pdnum; /* index in ->phys */ |
476 | struct spare_assign *spare; | |
8592f29d N |
477 | void *mdupdate; /* hold metadata update */ |
478 | ||
479 | /* These fields used by auto-layout */ | |
480 | int raiddisk; /* slot to fill in autolayout */ | |
481 | __u64 esize; | |
6416d527 NB |
482 | }; |
483 | }; | |
a322f70c | 484 | struct disk_data disk; |
b2280677 | 485 | struct vcl *vlist[0]; /* max_part in size */ |
2cc2983d | 486 | } *dlist, *add_list; |
a322f70c DW |
487 | }; |
488 | ||
489 | #ifndef offsetof | |
490 | #define offsetof(t,f) ((size_t)&(((t*)0)->f)) | |
491 | #endif | |
492 | ||
7d5a7ff3 | 493 | #if DEBUG |
fb9d0acb | 494 | static int all_ff(const char *guid); |
7d5a7ff3 | 495 | static void pr_state(struct ddf_super *ddf, const char *msg) |
496 | { | |
497 | unsigned int i; | |
498 | dprintf("%s/%s: ", __func__, msg); | |
a8173e43 | 499 | for (i = 0; i < be16_to_cpu(ddf->active->max_vd_entries); i++) { |
7d5a7ff3 | 500 | if (all_ff(ddf->virt->entries[i].guid)) |
501 | continue; | |
502 | dprintf("%u(s=%02x i=%02x) ", i, | |
503 | ddf->virt->entries[i].state, | |
504 | ddf->virt->entries[i].init_state); | |
505 | } | |
506 | dprintf("\n"); | |
507 | } | |
508 | #else | |
509 | static void pr_state(const struct ddf_super *ddf, const char *msg) {} | |
510 | #endif | |
511 | ||
35c3606d | 512 | static void _ddf_set_updates_pending(struct ddf_super *ddf, const char *func) |
513 | { | |
514 | ddf->updates_pending = 1; | |
60931cf9 | 515 | ddf->active->seq = cpu_to_be32((be32_to_cpu(ddf->active->seq)+1)); |
35c3606d | 516 | pr_state(ddf, func); |
517 | } | |
518 | ||
519 | #define ddf_set_updates_pending(x) _ddf_set_updates_pending((x), __func__) | |
7d5a7ff3 | 520 | |
fcc22180 | 521 | static unsigned int get_pd_index_from_refnum(const struct vcl *vc, |
60931cf9 | 522 | be32 refnum, unsigned int nmax, |
fcc22180 | 523 | const struct vd_config **bvd, |
524 | unsigned int *idx); | |
525 | ||
60931cf9 | 526 | static be32 calc_crc(void *buf, int len) |
a322f70c DW |
527 | { |
528 | /* crcs are always at the same place as in the ddf_header */ | |
529 | struct ddf_header *ddf = buf; | |
60931cf9 | 530 | be32 oldcrc = ddf->crc; |
a322f70c | 531 | __u32 newcrc; |
60931cf9 | 532 | ddf->crc = cpu_to_be32(0xffffffff); |
a322f70c DW |
533 | |
534 | newcrc = crc32(0, buf, len); | |
535 | ddf->crc = oldcrc; | |
4abe6b70 N |
536 | /* The crc is store (like everything) bigendian, so convert |
537 | * here for simplicity | |
538 | */ | |
60931cf9 | 539 | return cpu_to_be32(newcrc); |
a322f70c DW |
540 | } |
541 | ||
a3163bf0 | 542 | #define DDF_INVALID_LEVEL 0xff |
543 | #define DDF_NO_SECONDARY 0xff | |
544 | static int err_bad_md_layout(const mdu_array_info_t *array) | |
545 | { | |
546 | pr_err("RAID%d layout %x with %d disks is unsupported for DDF\n", | |
547 | array->level, array->layout, array->raid_disks); | |
2aba583f | 548 | return -1; |
a3163bf0 | 549 | } |
550 | ||
551 | static int layout_md2ddf(const mdu_array_info_t *array, | |
552 | struct vd_config *conf) | |
553 | { | |
a8173e43 | 554 | be16 prim_elmnt_count = cpu_to_be16(array->raid_disks); |
a3163bf0 | 555 | __u8 prl = DDF_INVALID_LEVEL, rlq = 0; |
556 | __u8 sec_elmnt_count = 1; | |
557 | __u8 srl = DDF_NO_SECONDARY; | |
558 | ||
559 | switch (array->level) { | |
560 | case LEVEL_LINEAR: | |
561 | prl = DDF_CONCAT; | |
562 | break; | |
563 | case 0: | |
564 | rlq = DDF_RAID0_SIMPLE; | |
565 | prl = DDF_RAID0; | |
566 | break; | |
567 | case 1: | |
568 | switch (array->raid_disks) { | |
569 | case 2: | |
570 | rlq = DDF_RAID1_SIMPLE; | |
571 | break; | |
572 | case 3: | |
573 | rlq = DDF_RAID1_MULTI; | |
574 | break; | |
575 | default: | |
576 | return err_bad_md_layout(array); | |
577 | } | |
578 | prl = DDF_RAID1; | |
579 | break; | |
580 | case 4: | |
581 | if (array->layout != 0) | |
582 | return err_bad_md_layout(array); | |
583 | rlq = DDF_RAID4_N; | |
584 | prl = DDF_RAID4; | |
585 | break; | |
586 | case 5: | |
587 | switch (array->layout) { | |
588 | case ALGORITHM_LEFT_ASYMMETRIC: | |
589 | rlq = DDF_RAID5_N_RESTART; | |
590 | break; | |
591 | case ALGORITHM_RIGHT_ASYMMETRIC: | |
592 | rlq = DDF_RAID5_0_RESTART; | |
593 | break; | |
594 | case ALGORITHM_LEFT_SYMMETRIC: | |
595 | rlq = DDF_RAID5_N_CONTINUE; | |
596 | break; | |
597 | case ALGORITHM_RIGHT_SYMMETRIC: | |
598 | /* not mentioned in standard */ | |
599 | default: | |
600 | return err_bad_md_layout(array); | |
601 | } | |
602 | prl = DDF_RAID5; | |
603 | break; | |
604 | case 6: | |
605 | switch (array->layout) { | |
606 | case ALGORITHM_ROTATING_N_RESTART: | |
607 | rlq = DDF_RAID5_N_RESTART; | |
608 | break; | |
609 | case ALGORITHM_ROTATING_ZERO_RESTART: | |
610 | rlq = DDF_RAID6_0_RESTART; | |
611 | break; | |
612 | case ALGORITHM_ROTATING_N_CONTINUE: | |
613 | rlq = DDF_RAID5_N_CONTINUE; | |
614 | break; | |
615 | default: | |
616 | return err_bad_md_layout(array); | |
617 | } | |
618 | prl = DDF_RAID6; | |
619 | break; | |
620 | case 10: | |
621 | if (array->raid_disks % 2 == 0 && array->layout == 0x102) { | |
622 | rlq = DDF_RAID1_SIMPLE; | |
a8173e43 | 623 | prim_elmnt_count = cpu_to_be16(2); |
a3163bf0 | 624 | sec_elmnt_count = array->raid_disks / 2; |
625 | } else if (array->raid_disks % 3 == 0 | |
626 | && array->layout == 0x103) { | |
627 | rlq = DDF_RAID1_MULTI; | |
a8173e43 | 628 | prim_elmnt_count = cpu_to_be16(3); |
a3163bf0 | 629 | sec_elmnt_count = array->raid_disks / 3; |
630 | } else | |
631 | return err_bad_md_layout(array); | |
632 | srl = DDF_2SPANNED; | |
633 | prl = DDF_RAID1; | |
634 | break; | |
635 | default: | |
636 | return err_bad_md_layout(array); | |
637 | } | |
638 | conf->prl = prl; | |
639 | conf->prim_elmnt_count = prim_elmnt_count; | |
640 | conf->rlq = rlq; | |
641 | conf->srl = srl; | |
642 | conf->sec_elmnt_count = sec_elmnt_count; | |
643 | return 0; | |
644 | } | |
645 | ||
8a2848a7 | 646 | static int err_bad_ddf_layout(const struct vd_config *conf) |
647 | { | |
648 | pr_err("DDF RAID %u qualifier %u with %u disks is unsupported\n", | |
a8173e43 | 649 | conf->prl, conf->rlq, be16_to_cpu(conf->prim_elmnt_count)); |
8a2848a7 | 650 | return -1; |
651 | } | |
652 | ||
653 | static int layout_ddf2md(const struct vd_config *conf, | |
654 | mdu_array_info_t *array) | |
655 | { | |
656 | int level = LEVEL_UNSUPPORTED; | |
657 | int layout = 0; | |
a8173e43 | 658 | int raiddisks = be16_to_cpu(conf->prim_elmnt_count); |
8a2848a7 | 659 | |
660 | if (conf->sec_elmnt_count > 1) { | |
661 | /* see also check_secondary() */ | |
662 | if (conf->prl != DDF_RAID1 || | |
663 | (conf->srl != DDF_2STRIPED && conf->srl != DDF_2SPANNED)) { | |
664 | pr_err("Unsupported secondary RAID level %u/%u\n", | |
665 | conf->prl, conf->srl); | |
666 | return -1; | |
667 | } | |
668 | if (raiddisks == 2 && conf->rlq == DDF_RAID1_SIMPLE) | |
669 | layout = 0x102; | |
670 | else if (raiddisks == 3 && conf->rlq == DDF_RAID1_MULTI) | |
671 | layout = 0x103; | |
672 | else | |
673 | return err_bad_ddf_layout(conf); | |
674 | raiddisks *= conf->sec_elmnt_count; | |
675 | level = 10; | |
676 | goto good; | |
677 | } | |
678 | ||
679 | switch (conf->prl) { | |
680 | case DDF_CONCAT: | |
681 | level = LEVEL_LINEAR; | |
682 | break; | |
683 | case DDF_RAID0: | |
684 | if (conf->rlq != DDF_RAID0_SIMPLE) | |
685 | return err_bad_ddf_layout(conf); | |
686 | level = 0; | |
687 | break; | |
688 | case DDF_RAID1: | |
689 | if (!((conf->rlq == DDF_RAID1_SIMPLE && raiddisks == 2) || | |
690 | (conf->rlq == DDF_RAID1_MULTI && raiddisks == 3))) | |
691 | return err_bad_ddf_layout(conf); | |
692 | level = 1; | |
693 | break; | |
694 | case DDF_RAID4: | |
695 | if (conf->rlq != DDF_RAID4_N) | |
696 | return err_bad_ddf_layout(conf); | |
697 | level = 4; | |
698 | break; | |
699 | case DDF_RAID5: | |
700 | switch (conf->rlq) { | |
701 | case DDF_RAID5_N_RESTART: | |
702 | layout = ALGORITHM_LEFT_ASYMMETRIC; | |
703 | break; | |
704 | case DDF_RAID5_0_RESTART: | |
705 | layout = ALGORITHM_RIGHT_ASYMMETRIC; | |
706 | break; | |
707 | case DDF_RAID5_N_CONTINUE: | |
708 | layout = ALGORITHM_LEFT_SYMMETRIC; | |
709 | break; | |
710 | default: | |
711 | return err_bad_ddf_layout(conf); | |
712 | } | |
713 | level = 5; | |
714 | break; | |
715 | case DDF_RAID6: | |
716 | switch (conf->rlq) { | |
717 | case DDF_RAID5_N_RESTART: | |
718 | layout = ALGORITHM_ROTATING_N_RESTART; | |
719 | break; | |
720 | case DDF_RAID6_0_RESTART: | |
721 | layout = ALGORITHM_ROTATING_ZERO_RESTART; | |
722 | break; | |
723 | case DDF_RAID5_N_CONTINUE: | |
724 | layout = ALGORITHM_ROTATING_N_CONTINUE; | |
725 | break; | |
726 | default: | |
727 | return err_bad_ddf_layout(conf); | |
728 | } | |
729 | level = 6; | |
730 | break; | |
731 | default: | |
732 | return err_bad_ddf_layout(conf); | |
733 | }; | |
734 | ||
735 | good: | |
736 | array->level = level; | |
737 | array->layout = layout; | |
738 | array->raid_disks = raiddisks; | |
739 | return 0; | |
740 | } | |
741 | ||
a322f70c DW |
742 | static int load_ddf_header(int fd, unsigned long long lba, |
743 | unsigned long long size, | |
744 | int type, | |
745 | struct ddf_header *hdr, struct ddf_header *anchor) | |
746 | { | |
747 | /* read a ddf header (primary or secondary) from fd/lba | |
748 | * and check that it is consistent with anchor | |
749 | * Need to check: | |
750 | * magic, crc, guid, rev, and LBA's header_type, and | |
751 | * everything after header_type must be the same | |
752 | */ | |
753 | if (lba >= size-1) | |
754 | return 0; | |
755 | ||
756 | if (lseek64(fd, lba<<9, 0) < 0) | |
757 | return 0; | |
758 | ||
759 | if (read(fd, hdr, 512) != 512) | |
760 | return 0; | |
761 | ||
0e5fa862 MW |
762 | if (!be32_eq(hdr->magic, DDF_HEADER_MAGIC)) { |
763 | pr_err("%s: bad header magic\n", __func__); | |
a322f70c | 764 | return 0; |
0e5fa862 MW |
765 | } |
766 | if (!be32_eq(calc_crc(hdr, 512), hdr->crc)) { | |
767 | pr_err("%s: bad CRC\n", __func__); | |
a322f70c | 768 | return 0; |
0e5fa862 | 769 | } |
a322f70c DW |
770 | if (memcmp(anchor->guid, hdr->guid, DDF_GUID_LEN) != 0 || |
771 | memcmp(anchor->revision, hdr->revision, 8) != 0 || | |
9d0c6b70 | 772 | !be64_eq(anchor->primary_lba, hdr->primary_lba) || |
773 | !be64_eq(anchor->secondary_lba, hdr->secondary_lba) || | |
a322f70c DW |
774 | hdr->type != type || |
775 | memcmp(anchor->pad2, hdr->pad2, 512 - | |
0e5fa862 MW |
776 | offsetof(struct ddf_header, pad2)) != 0) { |
777 | pr_err("%s: header mismatch\n", __func__); | |
a322f70c | 778 | return 0; |
0e5fa862 | 779 | } |
a322f70c DW |
780 | |
781 | /* Looks good enough to me... */ | |
782 | return 1; | |
783 | } | |
784 | ||
785 | static void *load_section(int fd, struct ddf_super *super, void *buf, | |
60931cf9 | 786 | be32 offset_be, be32 len_be, int check) |
a322f70c | 787 | { |
60931cf9 | 788 | unsigned long long offset = be32_to_cpu(offset_be); |
789 | unsigned long long len = be32_to_cpu(len_be); | |
a322f70c DW |
790 | int dofree = (buf == NULL); |
791 | ||
792 | if (check) | |
793 | if (len != 2 && len != 8 && len != 32 | |
794 | && len != 128 && len != 512) | |
795 | return NULL; | |
796 | ||
797 | if (len > 1024) | |
798 | return NULL; | |
3921e41a | 799 | if (!buf && posix_memalign(&buf, 512, len<<9) != 0) |
3d2c4fc7 | 800 | buf = NULL; |
6416d527 | 801 | |
a322f70c DW |
802 | if (!buf) |
803 | return NULL; | |
804 | ||
805 | if (super->active->type == 1) | |
9d0c6b70 | 806 | offset += be64_to_cpu(super->active->primary_lba); |
a322f70c | 807 | else |
9d0c6b70 | 808 | offset += be64_to_cpu(super->active->secondary_lba); |
a322f70c | 809 | |
f21e18ca | 810 | if ((unsigned long long)lseek64(fd, offset<<9, 0) != (offset<<9)) { |
a322f70c DW |
811 | if (dofree) |
812 | free(buf); | |
813 | return NULL; | |
814 | } | |
f21e18ca | 815 | if ((unsigned long long)read(fd, buf, len<<9) != (len<<9)) { |
a322f70c DW |
816 | if (dofree) |
817 | free(buf); | |
818 | return NULL; | |
819 | } | |
820 | return buf; | |
821 | } | |
822 | ||
823 | static int load_ddf_headers(int fd, struct ddf_super *super, char *devname) | |
824 | { | |
825 | unsigned long long dsize; | |
826 | ||
827 | get_dev_size(fd, NULL, &dsize); | |
828 | ||
829 | if (lseek64(fd, dsize-512, 0) < 0) { | |
830 | if (devname) | |
e7b84f9d N |
831 | pr_err("Cannot seek to anchor block on %s: %s\n", |
832 | devname, strerror(errno)); | |
a322f70c DW |
833 | return 1; |
834 | } | |
835 | if (read(fd, &super->anchor, 512) != 512) { | |
836 | if (devname) | |
e7b84f9d N |
837 | pr_err("Cannot read anchor block on %s: %s\n", |
838 | devname, strerror(errno)); | |
a322f70c DW |
839 | return 1; |
840 | } | |
60931cf9 | 841 | if (!be32_eq(super->anchor.magic, DDF_HEADER_MAGIC)) { |
a322f70c | 842 | if (devname) |
e7b84f9d | 843 | pr_err("no DDF anchor found on %s\n", |
a322f70c DW |
844 | devname); |
845 | return 2; | |
846 | } | |
60931cf9 | 847 | if (!be32_eq(calc_crc(&super->anchor, 512), super->anchor.crc)) { |
a322f70c | 848 | if (devname) |
e7b84f9d | 849 | pr_err("bad CRC on anchor on %s\n", |
a322f70c DW |
850 | devname); |
851 | return 2; | |
852 | } | |
59e36268 NB |
853 | if (memcmp(super->anchor.revision, DDF_REVISION_0, 8) != 0 && |
854 | memcmp(super->anchor.revision, DDF_REVISION_2, 8) != 0) { | |
a322f70c | 855 | if (devname) |
e7b84f9d | 856 | pr_err("can only support super revision" |
59e36268 NB |
857 | " %.8s and earlier, not %.8s on %s\n", |
858 | DDF_REVISION_2, super->anchor.revision,devname); | |
a322f70c DW |
859 | return 2; |
860 | } | |
dbeb699a | 861 | super->active = NULL; |
9d0c6b70 | 862 | if (load_ddf_header(fd, be64_to_cpu(super->anchor.primary_lba), |
a322f70c DW |
863 | dsize >> 9, 1, |
864 | &super->primary, &super->anchor) == 0) { | |
865 | if (devname) | |
e7b84f9d N |
866 | pr_err("Failed to load primary DDF header " |
867 | "on %s\n", devname); | |
dbeb699a | 868 | } else |
869 | super->active = &super->primary; | |
60931cf9 | 870 | |
9d0c6b70 | 871 | if (load_ddf_header(fd, be64_to_cpu(super->anchor.secondary_lba), |
a322f70c DW |
872 | dsize >> 9, 2, |
873 | &super->secondary, &super->anchor)) { | |
3eff7c1d | 874 | if (super->active == NULL |
60931cf9 | 875 | || (be32_to_cpu(super->primary.seq) |
876 | < be32_to_cpu(super->secondary.seq) && | |
3eff7c1d | 877 | !super->secondary.openflag) |
60931cf9 | 878 | || (be32_to_cpu(super->primary.seq) |
879 | == be32_to_cpu(super->secondary.seq) && | |
a322f70c DW |
880 | super->primary.openflag && !super->secondary.openflag) |
881 | ) | |
882 | super->active = &super->secondary; | |
b95cb4b9 N |
883 | } else if (devname && |
884 | be64_to_cpu(super->anchor.secondary_lba) != ~(__u64)0) | |
dbeb699a | 885 | pr_err("Failed to load secondary DDF header on %s\n", |
886 | devname); | |
887 | if (super->active == NULL) | |
888 | return 2; | |
a322f70c DW |
889 | return 0; |
890 | } | |
891 | ||
892 | static int load_ddf_global(int fd, struct ddf_super *super, char *devname) | |
893 | { | |
894 | void *ok; | |
895 | ok = load_section(fd, super, &super->controller, | |
896 | super->active->controller_section_offset, | |
897 | super->active->controller_section_length, | |
898 | 0); | |
899 | super->phys = load_section(fd, super, NULL, | |
900 | super->active->phys_section_offset, | |
901 | super->active->phys_section_length, | |
902 | 1); | |
60931cf9 | 903 | super->pdsize = be32_to_cpu(super->active->phys_section_length) * 512; |
a322f70c DW |
904 | |
905 | super->virt = load_section(fd, super, NULL, | |
906 | super->active->virt_section_offset, | |
907 | super->active->virt_section_length, | |
908 | 1); | |
60931cf9 | 909 | super->vdsize = be32_to_cpu(super->active->virt_section_length) * 512; |
a322f70c DW |
910 | if (!ok || |
911 | !super->phys || | |
912 | !super->virt) { | |
913 | free(super->phys); | |
914 | free(super->virt); | |
a2349791 NB |
915 | super->phys = NULL; |
916 | super->virt = NULL; | |
a322f70c DW |
917 | return 2; |
918 | } | |
919 | super->conflist = NULL; | |
920 | super->dlist = NULL; | |
8c3b8c2c | 921 | |
a8173e43 | 922 | super->max_part = be16_to_cpu(super->active->max_partitions); |
923 | super->mppe = be16_to_cpu(super->active->max_primary_element_entries); | |
924 | super->conf_rec_len = be16_to_cpu(super->active->config_record_len); | |
a322f70c DW |
925 | return 0; |
926 | } | |
927 | ||
3c48f7be | 928 | #define DDF_UNUSED_BVD 0xff |
929 | static int alloc_other_bvds(const struct ddf_super *ddf, struct vcl *vcl) | |
930 | { | |
931 | unsigned int n_vds = vcl->conf.sec_elmnt_count - 1; | |
932 | unsigned int i, vdsize; | |
933 | void *p; | |
934 | if (n_vds == 0) { | |
935 | vcl->other_bvds = NULL; | |
936 | return 0; | |
937 | } | |
938 | vdsize = ddf->conf_rec_len * 512; | |
939 | if (posix_memalign(&p, 512, n_vds * | |
940 | (vdsize + sizeof(struct vd_config *))) != 0) | |
941 | return -1; | |
942 | vcl->other_bvds = (struct vd_config **) (p + n_vds * vdsize); | |
943 | for (i = 0; i < n_vds; i++) { | |
944 | vcl->other_bvds[i] = p + i * vdsize; | |
945 | memset(vcl->other_bvds[i], 0, vdsize); | |
946 | vcl->other_bvds[i]->sec_elmnt_seq = DDF_UNUSED_BVD; | |
947 | } | |
948 | return 0; | |
949 | } | |
950 | ||
3dc821b0 | 951 | static void add_other_bvd(struct vcl *vcl, struct vd_config *vd, |
952 | unsigned int len) | |
953 | { | |
954 | int i; | |
955 | for (i = 0; i < vcl->conf.sec_elmnt_count-1; i++) | |
3c48f7be | 956 | if (vcl->other_bvds[i]->sec_elmnt_seq == vd->sec_elmnt_seq) |
3dc821b0 | 957 | break; |
958 | ||
959 | if (i < vcl->conf.sec_elmnt_count-1) { | |
60931cf9 | 960 | if (be32_to_cpu(vd->seqnum) <= |
961 | be32_to_cpu(vcl->other_bvds[i]->seqnum)) | |
3dc821b0 | 962 | return; |
963 | } else { | |
964 | for (i = 0; i < vcl->conf.sec_elmnt_count-1; i++) | |
3c48f7be | 965 | if (vcl->other_bvds[i]->sec_elmnt_seq == DDF_UNUSED_BVD) |
3dc821b0 | 966 | break; |
967 | if (i == vcl->conf.sec_elmnt_count-1) { | |
968 | pr_err("no space for sec level config %u, count is %u\n", | |
969 | vd->sec_elmnt_seq, vcl->conf.sec_elmnt_count); | |
970 | return; | |
971 | } | |
3dc821b0 | 972 | } |
973 | memcpy(vcl->other_bvds[i], vd, len); | |
974 | } | |
975 | ||
a322f70c DW |
976 | static int load_ddf_local(int fd, struct ddf_super *super, |
977 | char *devname, int keep) | |
978 | { | |
979 | struct dl *dl; | |
980 | struct stat stb; | |
981 | char *conf; | |
f21e18ca N |
982 | unsigned int i; |
983 | unsigned int confsec; | |
b2280677 | 984 | int vnum; |
a8173e43 | 985 | unsigned int max_virt_disks = be16_to_cpu |
986 | (super->active->max_vd_entries); | |
d2ca6449 | 987 | unsigned long long dsize; |
a322f70c DW |
988 | |
989 | /* First the local disk info */ | |
3d2c4fc7 | 990 | if (posix_memalign((void**)&dl, 512, |
6416d527 | 991 | sizeof(*dl) + |
3d2c4fc7 | 992 | (super->max_part) * sizeof(dl->vlist[0])) != 0) { |
e7b84f9d | 993 | pr_err("%s could not allocate disk info buffer\n", |
3d2c4fc7 DW |
994 | __func__); |
995 | return 1; | |
996 | } | |
a322f70c DW |
997 | |
998 | load_section(fd, super, &dl->disk, | |
999 | super->active->data_section_offset, | |
1000 | super->active->data_section_length, | |
1001 | 0); | |
503975b9 | 1002 | dl->devname = devname ? xstrdup(devname) : NULL; |
598f0d58 | 1003 | |
a322f70c DW |
1004 | fstat(fd, &stb); |
1005 | dl->major = major(stb.st_rdev); | |
1006 | dl->minor = minor(stb.st_rdev); | |
1007 | dl->next = super->dlist; | |
1008 | dl->fd = keep ? fd : -1; | |
d2ca6449 NB |
1009 | |
1010 | dl->size = 0; | |
1011 | if (get_dev_size(fd, devname, &dsize)) | |
1012 | dl->size = dsize >> 9; | |
097bcf00 | 1013 | /* If the disks have different sizes, the LBAs will differ |
1014 | * between phys disks. | |
1015 | * At this point here, the values in super->active must be valid | |
1016 | * for this phys disk. */ | |
1017 | dl->primary_lba = super->active->primary_lba; | |
1018 | dl->secondary_lba = super->active->secondary_lba; | |
1019 | dl->workspace_lba = super->active->workspace_lba; | |
b2280677 | 1020 | dl->spare = NULL; |
f21e18ca | 1021 | for (i = 0 ; i < super->max_part ; i++) |
a322f70c DW |
1022 | dl->vlist[i] = NULL; |
1023 | super->dlist = dl; | |
59e36268 | 1024 | dl->pdnum = -1; |
a8173e43 | 1025 | for (i = 0; i < be16_to_cpu(super->active->max_pd_entries); i++) |
5575e7d9 NB |
1026 | if (memcmp(super->phys->entries[i].guid, |
1027 | dl->disk.guid, DDF_GUID_LEN) == 0) | |
1028 | dl->pdnum = i; | |
1029 | ||
a322f70c DW |
1030 | /* Now the config list. */ |
1031 | /* 'conf' is an array of config entries, some of which are | |
1032 | * probably invalid. Those which are good need to be copied into | |
1033 | * the conflist | |
1034 | */ | |
a322f70c | 1035 | |
3921e41a | 1036 | conf = load_section(fd, super, super->conf, |
a322f70c DW |
1037 | super->active->config_section_offset, |
1038 | super->active->config_section_length, | |
1039 | 0); | |
3921e41a | 1040 | super->conf = conf; |
b2280677 | 1041 | vnum = 0; |
e223334f | 1042 | for (confsec = 0; |
60931cf9 | 1043 | confsec < be32_to_cpu(super->active->config_section_length); |
e223334f | 1044 | confsec += super->conf_rec_len) { |
a322f70c | 1045 | struct vd_config *vd = |
e223334f | 1046 | (struct vd_config *)((char*)conf + confsec*512); |
a322f70c DW |
1047 | struct vcl *vcl; |
1048 | ||
60931cf9 | 1049 | if (be32_eq(vd->magic, DDF_SPARE_ASSIGN_MAGIC)) { |
b2280677 NB |
1050 | if (dl->spare) |
1051 | continue; | |
3d2c4fc7 DW |
1052 | if (posix_memalign((void**)&dl->spare, 512, |
1053 | super->conf_rec_len*512) != 0) { | |
e7b84f9d N |
1054 | pr_err("%s could not allocate spare info buf\n", |
1055 | __func__); | |
3d2c4fc7 DW |
1056 | return 1; |
1057 | } | |
613b0d17 | 1058 | |
b2280677 NB |
1059 | memcpy(dl->spare, vd, super->conf_rec_len*512); |
1060 | continue; | |
1061 | } | |
60931cf9 | 1062 | if (!be32_eq(vd->magic, DDF_VD_CONF_MAGIC)) |
a322f70c DW |
1063 | continue; |
1064 | for (vcl = super->conflist; vcl; vcl = vcl->next) { | |
1065 | if (memcmp(vcl->conf.guid, | |
1066 | vd->guid, DDF_GUID_LEN) == 0) | |
1067 | break; | |
1068 | } | |
1069 | ||
1070 | if (vcl) { | |
b2280677 | 1071 | dl->vlist[vnum++] = vcl; |
3dc821b0 | 1072 | if (vcl->other_bvds != NULL && |
1073 | vcl->conf.sec_elmnt_seq != vd->sec_elmnt_seq) { | |
1074 | add_other_bvd(vcl, vd, super->conf_rec_len*512); | |
1075 | continue; | |
1076 | } | |
60931cf9 | 1077 | if (be32_to_cpu(vd->seqnum) <= |
1078 | be32_to_cpu(vcl->conf.seqnum)) | |
a322f70c | 1079 | continue; |
59e36268 | 1080 | } else { |
3d2c4fc7 | 1081 | if (posix_memalign((void**)&vcl, 512, |
6416d527 | 1082 | (super->conf_rec_len*512 + |
3d2c4fc7 | 1083 | offsetof(struct vcl, conf))) != 0) { |
e7b84f9d N |
1084 | pr_err("%s could not allocate vcl buf\n", |
1085 | __func__); | |
3d2c4fc7 DW |
1086 | return 1; |
1087 | } | |
a322f70c | 1088 | vcl->next = super->conflist; |
59e36268 | 1089 | vcl->block_sizes = NULL; /* FIXME not for CONCAT */ |
3c48f7be | 1090 | vcl->conf.sec_elmnt_count = vd->sec_elmnt_count; |
1091 | if (alloc_other_bvds(super, vcl) != 0) { | |
1092 | pr_err("%s could not allocate other bvds\n", | |
1093 | __func__); | |
1094 | free(vcl); | |
1095 | return 1; | |
1096 | }; | |
a322f70c | 1097 | super->conflist = vcl; |
b2280677 | 1098 | dl->vlist[vnum++] = vcl; |
a322f70c | 1099 | } |
8c3b8c2c | 1100 | memcpy(&vcl->conf, vd, super->conf_rec_len*512); |
59e36268 NB |
1101 | for (i=0; i < max_virt_disks ; i++) |
1102 | if (memcmp(super->virt->entries[i].guid, | |
1103 | vcl->conf.guid, DDF_GUID_LEN)==0) | |
1104 | break; | |
1105 | if (i < max_virt_disks) | |
1106 | vcl->vcnum = i; | |
a322f70c | 1107 | } |
a322f70c DW |
1108 | |
1109 | return 0; | |
1110 | } | |
1111 | ||
1112 | #ifndef MDASSEMBLE | |
1113 | static int load_super_ddf_all(struct supertype *st, int fd, | |
e1902a7b | 1114 | void **sbp, char *devname); |
a322f70c | 1115 | #endif |
37424f13 DW |
1116 | |
1117 | static void free_super_ddf(struct supertype *st); | |
1118 | ||
a322f70c DW |
1119 | static int load_super_ddf(struct supertype *st, int fd, |
1120 | char *devname) | |
1121 | { | |
1122 | unsigned long long dsize; | |
1123 | struct ddf_super *super; | |
1124 | int rv; | |
1125 | ||
a322f70c DW |
1126 | if (get_dev_size(fd, devname, &dsize) == 0) |
1127 | return 1; | |
1128 | ||
a34fea0e | 1129 | if (test_partition(fd)) |
691c6ee1 N |
1130 | /* DDF is not allowed on partitions */ |
1131 | return 1; | |
1132 | ||
a322f70c DW |
1133 | /* 32M is a lower bound */ |
1134 | if (dsize <= 32*1024*1024) { | |
97320d7c | 1135 | if (devname) |
e7b84f9d N |
1136 | pr_err("%s is too small for ddf: " |
1137 | "size is %llu sectors.\n", | |
1138 | devname, dsize>>9); | |
97320d7c | 1139 | return 1; |
a322f70c DW |
1140 | } |
1141 | if (dsize & 511) { | |
97320d7c | 1142 | if (devname) |
e7b84f9d N |
1143 | pr_err("%s is an odd size for ddf: " |
1144 | "size is %llu bytes.\n", | |
1145 | devname, dsize); | |
97320d7c | 1146 | return 1; |
a322f70c DW |
1147 | } |
1148 | ||
37424f13 DW |
1149 | free_super_ddf(st); |
1150 | ||
6416d527 | 1151 | if (posix_memalign((void**)&super, 512, sizeof(*super))!= 0) { |
e7b84f9d | 1152 | pr_err("malloc of %zu failed.\n", |
a322f70c DW |
1153 | sizeof(*super)); |
1154 | return 1; | |
1155 | } | |
a2349791 | 1156 | memset(super, 0, sizeof(*super)); |
a322f70c DW |
1157 | |
1158 | rv = load_ddf_headers(fd, super, devname); | |
1159 | if (rv) { | |
1160 | free(super); | |
1161 | return rv; | |
1162 | } | |
1163 | ||
1164 | /* Have valid headers and have chosen the best. Let's read in the rest*/ | |
1165 | ||
1166 | rv = load_ddf_global(fd, super, devname); | |
1167 | ||
1168 | if (rv) { | |
1169 | if (devname) | |
e7b84f9d N |
1170 | pr_err("Failed to load all information " |
1171 | "sections on %s\n", devname); | |
a322f70c DW |
1172 | free(super); |
1173 | return rv; | |
1174 | } | |
1175 | ||
3d2c4fc7 DW |
1176 | rv = load_ddf_local(fd, super, devname, 0); |
1177 | ||
1178 | if (rv) { | |
1179 | if (devname) | |
e7b84f9d N |
1180 | pr_err("Failed to load all information " |
1181 | "sections on %s\n", devname); | |
3d2c4fc7 DW |
1182 | free(super); |
1183 | return rv; | |
1184 | } | |
a322f70c DW |
1185 | |
1186 | /* Should possibly check the sections .... */ | |
1187 | ||
1188 | st->sb = super; | |
1189 | if (st->ss == NULL) { | |
1190 | st->ss = &super_ddf; | |
1191 | st->minor_version = 0; | |
1192 | st->max_devs = 512; | |
1193 | } | |
1194 | return 0; | |
1195 | ||
1196 | } | |
1197 | ||
1198 | static void free_super_ddf(struct supertype *st) | |
1199 | { | |
1200 | struct ddf_super *ddf = st->sb; | |
1201 | if (ddf == NULL) | |
1202 | return; | |
1203 | free(ddf->phys); | |
1204 | free(ddf->virt); | |
3921e41a | 1205 | free(ddf->conf); |
a322f70c DW |
1206 | while (ddf->conflist) { |
1207 | struct vcl *v = ddf->conflist; | |
1208 | ddf->conflist = v->next; | |
59e36268 NB |
1209 | if (v->block_sizes) |
1210 | free(v->block_sizes); | |
3c48f7be | 1211 | if (v->other_bvds) |
1212 | /* | |
1213 | v->other_bvds[0] points to beginning of buffer, | |
1214 | see alloc_other_bvds() | |
1215 | */ | |
1216 | free(v->other_bvds[0]); | |
a322f70c DW |
1217 | free(v); |
1218 | } | |
1219 | while (ddf->dlist) { | |
1220 | struct dl *d = ddf->dlist; | |
1221 | ddf->dlist = d->next; | |
1222 | if (d->fd >= 0) | |
1223 | close(d->fd); | |
b2280677 NB |
1224 | if (d->spare) |
1225 | free(d->spare); | |
a322f70c DW |
1226 | free(d); |
1227 | } | |
8a38cb04 N |
1228 | while (ddf->add_list) { |
1229 | struct dl *d = ddf->add_list; | |
1230 | ddf->add_list = d->next; | |
1231 | if (d->fd >= 0) | |
1232 | close(d->fd); | |
1233 | if (d->spare) | |
1234 | free(d->spare); | |
1235 | free(d); | |
1236 | } | |
a322f70c DW |
1237 | free(ddf); |
1238 | st->sb = NULL; | |
1239 | } | |
1240 | ||
1241 | static struct supertype *match_metadata_desc_ddf(char *arg) | |
1242 | { | |
1243 | /* 'ddf' only support containers */ | |
1244 | struct supertype *st; | |
1245 | if (strcmp(arg, "ddf") != 0 && | |
1246 | strcmp(arg, "default") != 0 | |
1247 | ) | |
1248 | return NULL; | |
1249 | ||
503975b9 | 1250 | st = xcalloc(1, sizeof(*st)); |
a322f70c DW |
1251 | st->ss = &super_ddf; |
1252 | st->max_devs = 512; | |
1253 | st->minor_version = 0; | |
1254 | st->sb = NULL; | |
1255 | return st; | |
1256 | } | |
1257 | ||
a322f70c DW |
1258 | #ifndef MDASSEMBLE |
1259 | ||
1260 | static mapping_t ddf_state[] = { | |
1261 | { "Optimal", 0}, | |
1262 | { "Degraded", 1}, | |
1263 | { "Deleted", 2}, | |
1264 | { "Missing", 3}, | |
1265 | { "Failed", 4}, | |
1266 | { "Partially Optimal", 5}, | |
1267 | { "-reserved-", 6}, | |
1268 | { "-reserved-", 7}, | |
1269 | { NULL, 0} | |
1270 | }; | |
1271 | ||
1272 | static mapping_t ddf_init_state[] = { | |
1273 | { "Not Initialised", 0}, | |
1274 | { "QuickInit in Progress", 1}, | |
1275 | { "Fully Initialised", 2}, | |
1276 | { "*UNKNOWN*", 3}, | |
1277 | { NULL, 0} | |
1278 | }; | |
1279 | static mapping_t ddf_access[] = { | |
1280 | { "Read/Write", 0}, | |
1281 | { "Reserved", 1}, | |
1282 | { "Read Only", 2}, | |
1283 | { "Blocked (no access)", 3}, | |
1284 | { NULL ,0} | |
1285 | }; | |
1286 | ||
1287 | static mapping_t ddf_level[] = { | |
1288 | { "RAID0", DDF_RAID0}, | |
1289 | { "RAID1", DDF_RAID1}, | |
1290 | { "RAID3", DDF_RAID3}, | |
1291 | { "RAID4", DDF_RAID4}, | |
1292 | { "RAID5", DDF_RAID5}, | |
1293 | { "RAID1E",DDF_RAID1E}, | |
1294 | { "JBOD", DDF_JBOD}, | |
1295 | { "CONCAT",DDF_CONCAT}, | |
1296 | { "RAID5E",DDF_RAID5E}, | |
1297 | { "RAID5EE",DDF_RAID5EE}, | |
1298 | { "RAID6", DDF_RAID6}, | |
1299 | { NULL, 0} | |
1300 | }; | |
1301 | static mapping_t ddf_sec_level[] = { | |
1302 | { "Striped", DDF_2STRIPED}, | |
1303 | { "Mirrored", DDF_2MIRRORED}, | |
1304 | { "Concat", DDF_2CONCAT}, | |
1305 | { "Spanned", DDF_2SPANNED}, | |
1306 | { NULL, 0} | |
1307 | }; | |
1308 | #endif | |
1309 | ||
fb9d0acb | 1310 | static int all_ff(const char *guid) |
42dc2744 N |
1311 | { |
1312 | int i; | |
1313 | for (i = 0; i < DDF_GUID_LEN; i++) | |
1314 | if (guid[i] != (char)0xff) | |
1315 | return 0; | |
1316 | return 1; | |
1317 | } | |
1318 | ||
4441541f N |
1319 | static const char *guid_str(const char *guid) |
1320 | { | |
1321 | static char buf[DDF_GUID_LEN*2+1]; | |
1322 | int i; | |
1323 | char *p = buf; | |
1324 | for (i = 0; i < DDF_GUID_LEN; i++) { | |
1325 | unsigned char c = guid[i]; | |
1326 | if (c >= 32 && c < 127) | |
1327 | p += sprintf(p, "%c", c); | |
1328 | else | |
1329 | p += sprintf(p, "%02x", c); | |
1330 | } | |
1331 | *p = '\0'; | |
1332 | return (const char *) buf; | |
1333 | } | |
1334 | ||
a322f70c DW |
1335 | #ifndef MDASSEMBLE |
1336 | static void print_guid(char *guid, int tstamp) | |
1337 | { | |
1338 | /* A GUIDs are part (or all) ASCII and part binary. | |
1339 | * They tend to be space padded. | |
59e36268 NB |
1340 | * We print the GUID in HEX, then in parentheses add |
1341 | * any initial ASCII sequence, and a possible | |
1342 | * time stamp from bytes 16-19 | |
a322f70c DW |
1343 | */ |
1344 | int l = DDF_GUID_LEN; | |
1345 | int i; | |
59e36268 NB |
1346 | |
1347 | for (i=0 ; i<DDF_GUID_LEN ; i++) { | |
1348 | if ((i&3)==0 && i != 0) printf(":"); | |
1349 | printf("%02X", guid[i]&255); | |
1350 | } | |
1351 | ||
cfccea8c | 1352 | printf("\n ("); |
a322f70c DW |
1353 | while (l && guid[l-1] == ' ') |
1354 | l--; | |
1355 | for (i=0 ; i<l ; i++) { | |
1356 | if (guid[i] >= 0x20 && guid[i] < 0x7f) | |
1357 | fputc(guid[i], stdout); | |
1358 | else | |
59e36268 | 1359 | break; |
a322f70c DW |
1360 | } |
1361 | if (tstamp) { | |
1362 | time_t then = __be32_to_cpu(*(__u32*)(guid+16)) + DECADE; | |
1363 | char tbuf[100]; | |
1364 | struct tm *tm; | |
1365 | tm = localtime(&then); | |
59e36268 | 1366 | strftime(tbuf, 100, " %D %T",tm); |
a322f70c DW |
1367 | fputs(tbuf, stdout); |
1368 | } | |
59e36268 | 1369 | printf(")"); |
a322f70c DW |
1370 | } |
1371 | ||
1372 | static void examine_vd(int n, struct ddf_super *sb, char *guid) | |
1373 | { | |
8c3b8c2c | 1374 | int crl = sb->conf_rec_len; |
a322f70c DW |
1375 | struct vcl *vcl; |
1376 | ||
1377 | for (vcl = sb->conflist ; vcl ; vcl = vcl->next) { | |
f21e18ca | 1378 | unsigned int i; |
a322f70c DW |
1379 | struct vd_config *vc = &vcl->conf; |
1380 | ||
60931cf9 | 1381 | if (!be32_eq(calc_crc(vc, crl*512), vc->crc)) |
a322f70c DW |
1382 | continue; |
1383 | if (memcmp(vc->guid, guid, DDF_GUID_LEN) != 0) | |
1384 | continue; | |
1385 | ||
1386 | /* Ok, we know about this VD, let's give more details */ | |
b06e3095 | 1387 | printf(" Raid Devices[%d] : %d (", n, |
a8173e43 | 1388 | be16_to_cpu(vc->prim_elmnt_count)); |
1389 | for (i = 0; i < be16_to_cpu(vc->prim_elmnt_count); i++) { | |
b06e3095 | 1390 | int j; |
a8173e43 | 1391 | int cnt = be16_to_cpu(sb->phys->used_pdes); |
b06e3095 | 1392 | for (j=0; j<cnt; j++) |
60931cf9 | 1393 | if (be32_eq(vc->phys_refnum[i], |
1394 | sb->phys->entries[j].refnum)) | |
b06e3095 N |
1395 | break; |
1396 | if (i) printf(" "); | |
1397 | if (j < cnt) | |
1398 | printf("%d", j); | |
1399 | else | |
1400 | printf("--"); | |
1401 | } | |
1402 | printf(")\n"); | |
1403 | if (vc->chunk_shift != 255) | |
613b0d17 N |
1404 | printf(" Chunk Size[%d] : %d sectors\n", n, |
1405 | 1 << vc->chunk_shift); | |
a322f70c DW |
1406 | printf(" Raid Level[%d] : %s\n", n, |
1407 | map_num(ddf_level, vc->prl)?:"-unknown-"); | |
1408 | if (vc->sec_elmnt_count != 1) { | |
1409 | printf(" Secondary Position[%d] : %d of %d\n", n, | |
1410 | vc->sec_elmnt_seq, vc->sec_elmnt_count); | |
1411 | printf(" Secondary Level[%d] : %s\n", n, | |
1412 | map_num(ddf_sec_level, vc->srl) ?: "-unknown-"); | |
1413 | } | |
1414 | printf(" Device Size[%d] : %llu\n", n, | |
9d0c6b70 | 1415 | be64_to_cpu(vc->blocks)/2); |
a322f70c | 1416 | printf(" Array Size[%d] : %llu\n", n, |
9d0c6b70 | 1417 | be64_to_cpu(vc->array_blocks)/2); |
a322f70c DW |
1418 | } |
1419 | } | |
1420 | ||
1421 | static void examine_vds(struct ddf_super *sb) | |
1422 | { | |
a8173e43 | 1423 | int cnt = be16_to_cpu(sb->virt->populated_vdes); |
fb9d0acb | 1424 | unsigned int i; |
a322f70c DW |
1425 | printf(" Virtual Disks : %d\n", cnt); |
1426 | ||
a8173e43 | 1427 | for (i = 0; i < be16_to_cpu(sb->virt->max_vdes); i++) { |
a322f70c | 1428 | struct virtual_entry *ve = &sb->virt->entries[i]; |
fb9d0acb | 1429 | if (all_ff(ve->guid)) |
1430 | continue; | |
b06e3095 | 1431 | printf("\n"); |
a322f70c DW |
1432 | printf(" VD GUID[%d] : ", i); print_guid(ve->guid, 1); |
1433 | printf("\n"); | |
a8173e43 | 1434 | printf(" unit[%d] : %d\n", i, be16_to_cpu(ve->unit)); |
a322f70c DW |
1435 | printf(" state[%d] : %s, %s%s\n", i, |
1436 | map_num(ddf_state, ve->state & 7), | |
cc83a819 N |
1437 | (ve->state & DDF_state_morphing) ? "Morphing, ": "", |
1438 | (ve->state & DDF_state_inconsistent)? "Not Consistent" : "Consistent"); | |
a322f70c | 1439 | printf(" init state[%d] : %s\n", i, |
cc83a819 | 1440 | map_num(ddf_init_state, ve->init_state&DDF_initstate_mask)); |
a322f70c | 1441 | printf(" access[%d] : %s\n", i, |
cc83a819 | 1442 | map_num(ddf_access, (ve->init_state & DDF_access_mask) >> 6)); |
a322f70c DW |
1443 | printf(" Name[%d] : %.16s\n", i, ve->name); |
1444 | examine_vd(i, sb, ve->guid); | |
1445 | } | |
1446 | if (cnt) printf("\n"); | |
1447 | } | |
1448 | ||
1449 | static void examine_pds(struct ddf_super *sb) | |
1450 | { | |
a8173e43 | 1451 | int cnt = be16_to_cpu(sb->phys->used_pdes); |
a322f70c DW |
1452 | int i; |
1453 | struct dl *dl; | |
1454 | printf(" Physical Disks : %d\n", cnt); | |
962371a5 | 1455 | printf(" Number RefNo Size Device Type/State\n"); |
a322f70c DW |
1456 | |
1457 | for (i=0 ; i<cnt ; i++) { | |
1458 | struct phys_disk_entry *pd = &sb->phys->entries[i]; | |
a8173e43 | 1459 | int type = be16_to_cpu(pd->type); |
1460 | int state = be16_to_cpu(pd->state); | |
a322f70c | 1461 | |
b06e3095 N |
1462 | //printf(" PD GUID[%d] : ", i); print_guid(pd->guid, 0); |
1463 | //printf("\n"); | |
1464 | printf(" %3d %08x ", i, | |
60931cf9 | 1465 | be32_to_cpu(pd->refnum)); |
613b0d17 | 1466 | printf("%8lluK ", |
9d0c6b70 | 1467 | be64_to_cpu(pd->config_size)>>1); |
b06e3095 | 1468 | for (dl = sb->dlist; dl ; dl = dl->next) { |
60931cf9 | 1469 | if (be32_eq(dl->disk.refnum, pd->refnum)) { |
b06e3095 N |
1470 | char *dv = map_dev(dl->major, dl->minor, 0); |
1471 | if (dv) { | |
962371a5 | 1472 | printf("%-15s", dv); |
b06e3095 N |
1473 | break; |
1474 | } | |
1475 | } | |
1476 | } | |
1477 | if (!dl) | |
962371a5 | 1478 | printf("%15s",""); |
b06e3095 | 1479 | printf(" %s%s%s%s%s", |
a322f70c | 1480 | (type&2) ? "active":"", |
b06e3095 | 1481 | (type&4) ? "Global-Spare":"", |
a322f70c DW |
1482 | (type&8) ? "spare" : "", |
1483 | (type&16)? ", foreign" : "", | |
1484 | (type&32)? "pass-through" : ""); | |
18cb4496 N |
1485 | if (state & DDF_Failed) |
1486 | /* This over-rides these three */ | |
1487 | state &= ~(DDF_Online|DDF_Rebuilding|DDF_Transition); | |
b06e3095 | 1488 | printf("/%s%s%s%s%s%s%s", |
a322f70c DW |
1489 | (state&1)? "Online": "Offline", |
1490 | (state&2)? ", Failed": "", | |
1491 | (state&4)? ", Rebuilding": "", | |
1492 | (state&8)? ", in-transition": "", | |
b06e3095 N |
1493 | (state&16)? ", SMART-errors": "", |
1494 | (state&32)? ", Unrecovered-Read-Errors": "", | |
a322f70c | 1495 | (state&64)? ", Missing" : ""); |
a322f70c DW |
1496 | printf("\n"); |
1497 | } | |
1498 | } | |
1499 | ||
1500 | static void examine_super_ddf(struct supertype *st, char *homehost) | |
1501 | { | |
1502 | struct ddf_super *sb = st->sb; | |
1503 | ||
60931cf9 | 1504 | printf(" Magic : %08x\n", be32_to_cpu(sb->anchor.magic)); |
a322f70c | 1505 | printf(" Version : %.8s\n", sb->anchor.revision); |
598f0d58 NB |
1506 | printf("Controller GUID : "); print_guid(sb->controller.guid, 0); |
1507 | printf("\n"); | |
1508 | printf(" Container GUID : "); print_guid(sb->anchor.guid, 1); | |
a322f70c | 1509 | printf("\n"); |
60931cf9 | 1510 | printf(" Seq : %08x\n", be32_to_cpu(sb->active->seq)); |
1511 | printf(" Redundant hdr : %s\n", be32_eq(sb->secondary.magic, | |
1512 | DDF_HEADER_MAGIC) | |
a322f70c DW |
1513 | ?"yes" : "no"); |
1514 | examine_vds(sb); | |
1515 | examine_pds(sb); | |
1516 | } | |
1517 | ||
a5d85af7 | 1518 | static void getinfo_super_ddf(struct supertype *st, struct mdinfo *info, char *map); |
ff54de6e | 1519 | |
bedbf68a | 1520 | static void uuid_from_ddf_guid(const char *guid, int uuid[4]); |
42dc2744 | 1521 | static void uuid_from_super_ddf(struct supertype *st, int uuid[4]); |
a8b25633 | 1522 | static void _ddf_array_name(char *name, const struct ddf_super *ddf, int i); |
ff54de6e | 1523 | |
bedbf68a | 1524 | static unsigned int get_vd_num_of_subarray(struct supertype *st) |
1525 | { | |
1526 | /* | |
1527 | * Figure out the VD number for this supertype. | |
1528 | * Returns DDF_CONTAINER for the container itself, | |
1529 | * and DDF_NOTFOUND on error. | |
1530 | */ | |
1531 | struct ddf_super *ddf = st->sb; | |
1532 | struct mdinfo *sra; | |
1533 | char *sub, *end; | |
1534 | unsigned int vcnum; | |
1535 | ||
1536 | if (*st->container_devnm == '\0') | |
1537 | return DDF_CONTAINER; | |
1538 | ||
1539 | sra = sysfs_read(-1, st->devnm, GET_VERSION); | |
1540 | if (!sra || sra->array.major_version != -1 || | |
1541 | sra->array.minor_version != -2 || | |
1542 | !is_subarray(sra->text_version)) | |
1543 | return DDF_NOTFOUND; | |
1544 | ||
1545 | sub = strchr(sra->text_version + 1, '/'); | |
1546 | if (sub != NULL) | |
1547 | vcnum = strtoul(sub + 1, &end, 10); | |
1548 | if (sub == NULL || *sub == '\0' || *end != '\0' || | |
a8173e43 | 1549 | vcnum >= be16_to_cpu(ddf->active->max_vd_entries)) |
bedbf68a | 1550 | return DDF_NOTFOUND; |
1551 | ||
1552 | return vcnum; | |
1553 | } | |
1554 | ||
061f2c6a | 1555 | static void brief_examine_super_ddf(struct supertype *st, int verbose) |
4737ae25 N |
1556 | { |
1557 | /* We just write a generic DDF ARRAY entry | |
1558 | */ | |
1559 | struct mdinfo info; | |
1560 | char nbuf[64]; | |
a5d85af7 | 1561 | getinfo_super_ddf(st, &info, NULL); |
4737ae25 N |
1562 | fname_from_uuid(st, &info, nbuf, ':'); |
1563 | ||
1564 | printf("ARRAY metadata=ddf UUID=%s\n", nbuf + 5); | |
1565 | } | |
1566 | ||
1567 | static void brief_examine_subarrays_ddf(struct supertype *st, int verbose) | |
a322f70c DW |
1568 | { |
1569 | /* We just write a generic DDF ARRAY entry | |
a322f70c | 1570 | */ |
42dc2744 | 1571 | struct ddf_super *ddf = st->sb; |
ff54de6e | 1572 | struct mdinfo info; |
f21e18ca | 1573 | unsigned int i; |
ff54de6e | 1574 | char nbuf[64]; |
a5d85af7 | 1575 | getinfo_super_ddf(st, &info, NULL); |
ff54de6e | 1576 | fname_from_uuid(st, &info, nbuf, ':'); |
42dc2744 | 1577 | |
a8173e43 | 1578 | for (i = 0; i < be16_to_cpu(ddf->virt->max_vdes); i++) { |
42dc2744 N |
1579 | struct virtual_entry *ve = &ddf->virt->entries[i]; |
1580 | struct vcl vcl; | |
1581 | char nbuf1[64]; | |
a8b25633 | 1582 | char namebuf[17]; |
42dc2744 N |
1583 | if (all_ff(ve->guid)) |
1584 | continue; | |
1585 | memcpy(vcl.conf.guid, ve->guid, DDF_GUID_LEN); | |
1586 | ddf->currentconf =&vcl; | |
7087f02b | 1587 | vcl.vcnum = i; |
42dc2744 N |
1588 | uuid_from_super_ddf(st, info.uuid); |
1589 | fname_from_uuid(st, &info, nbuf1, ':'); | |
a8b25633 | 1590 | _ddf_array_name(namebuf, ddf, i); |
1591 | printf("ARRAY%s%s container=%s member=%d UUID=%s\n", | |
1592 | namebuf[0] == '\0' ? "" : " /dev/md/", namebuf, | |
42dc2744 N |
1593 | nbuf+5, i, nbuf1+5); |
1594 | } | |
a322f70c DW |
1595 | } |
1596 | ||
bceedeec N |
1597 | static void export_examine_super_ddf(struct supertype *st) |
1598 | { | |
1599 | struct mdinfo info; | |
1600 | char nbuf[64]; | |
a5d85af7 | 1601 | getinfo_super_ddf(st, &info, NULL); |
bceedeec N |
1602 | fname_from_uuid(st, &info, nbuf, ':'); |
1603 | printf("MD_METADATA=ddf\n"); | |
1604 | printf("MD_LEVEL=container\n"); | |
1605 | printf("MD_UUID=%s\n", nbuf+5); | |
cc9bfd9e | 1606 | printf("MD_DEVICES=%u\n", |
1607 | be16_to_cpu(((struct ddf_super *)st->sb)->phys->used_pdes)); | |
bceedeec | 1608 | } |
bceedeec | 1609 | |
74db60b0 N |
1610 | static int copy_metadata_ddf(struct supertype *st, int from, int to) |
1611 | { | |
1612 | void *buf; | |
1613 | unsigned long long dsize, offset; | |
1614 | int bytes; | |
1615 | struct ddf_header *ddf; | |
1616 | int written = 0; | |
1617 | ||
1618 | /* The meta consists of an anchor, a primary, and a secondary. | |
1619 | * This all lives at the end of the device. | |
1620 | * So it is easiest to find the earliest of primary and | |
1621 | * secondary, and copy everything from there. | |
1622 | * | |
1623 | * Anchor is 512 from end It contains primary_lba and secondary_lba | |
1624 | * we choose one of those | |
1625 | */ | |
1626 | ||
1627 | if (posix_memalign(&buf, 4096, 4096) != 0) | |
1628 | return 1; | |
1629 | ||
1630 | if (!get_dev_size(from, NULL, &dsize)) | |
1631 | goto err; | |
1632 | ||
1633 | if (lseek64(from, dsize-512, 0) < 0) | |
1634 | goto err; | |
1635 | if (read(from, buf, 512) != 512) | |
1636 | goto err; | |
1637 | ddf = buf; | |
60931cf9 | 1638 | if (!be32_eq(ddf->magic, DDF_HEADER_MAGIC) || |
1639 | !be32_eq(calc_crc(ddf, 512), ddf->crc) || | |
74db60b0 N |
1640 | (memcmp(ddf->revision, DDF_REVISION_0, 8) != 0 && |
1641 | memcmp(ddf->revision, DDF_REVISION_2, 8) != 0)) | |
1642 | goto err; | |
1643 | ||
1644 | offset = dsize - 512; | |
9d0c6b70 | 1645 | if ((be64_to_cpu(ddf->primary_lba) << 9) < offset) |
1646 | offset = be64_to_cpu(ddf->primary_lba) << 9; | |
1647 | if ((be64_to_cpu(ddf->secondary_lba) << 9) < offset) | |
1648 | offset = be64_to_cpu(ddf->secondary_lba) << 9; | |
74db60b0 N |
1649 | |
1650 | bytes = dsize - offset; | |
1651 | ||
1652 | if (lseek64(from, offset, 0) < 0 || | |
1653 | lseek64(to, offset, 0) < 0) | |
1654 | goto err; | |
1655 | while (written < bytes) { | |
1656 | int n = bytes - written; | |
1657 | if (n > 4096) | |
1658 | n = 4096; | |
1659 | if (read(from, buf, n) != n) | |
1660 | goto err; | |
1661 | if (write(to, buf, n) != n) | |
1662 | goto err; | |
1663 | written += n; | |
1664 | } | |
1665 | free(buf); | |
1666 | return 0; | |
1667 | err: | |
1668 | free(buf); | |
1669 | return 1; | |
1670 | } | |
1671 | ||
a322f70c DW |
1672 | static void detail_super_ddf(struct supertype *st, char *homehost) |
1673 | { | |
1674 | /* FIXME later | |
1675 | * Could print DDF GUID | |
1676 | * Need to find which array | |
1677 | * If whole, briefly list all arrays | |
1678 | * If one, give name | |
1679 | */ | |
1680 | } | |
1681 | ||
7087f02b | 1682 | static const char *vendors_with_variable_volume_UUID[] = { |
1683 | "LSI ", | |
1684 | }; | |
1685 | ||
1686 | static int volume_id_is_reliable(const struct ddf_super *ddf) | |
1687 | { | |
1c0aebc2 | 1688 | int n = ARRAY_SIZE(vendors_with_variable_volume_UUID); |
7087f02b | 1689 | int i; |
1690 | for (i = 0; i < n; i++) | |
1691 | if (!memcmp(ddf->controller.guid, | |
1692 | vendors_with_variable_volume_UUID[i], 8)) | |
1693 | return 0; | |
1694 | return 1; | |
1695 | } | |
1696 | ||
1697 | static void uuid_of_ddf_subarray(const struct ddf_super *ddf, | |
1698 | unsigned int vcnum, int uuid[4]) | |
1699 | { | |
1700 | char buf[DDF_GUID_LEN+18], sha[20], *p; | |
1701 | struct sha1_ctx ctx; | |
1702 | if (volume_id_is_reliable(ddf)) { | |
1703 | uuid_from_ddf_guid(ddf->virt->entries[vcnum].guid, uuid); | |
1704 | return; | |
1705 | } | |
1706 | /* | |
1707 | * Some fake RAID BIOSes (in particular, LSI ones) change the | |
1708 | * VD GUID at every boot. These GUIDs are not suitable for | |
1709 | * identifying an array. Luckily the header GUID appears to | |
1710 | * remain constant. | |
1711 | * We construct a pseudo-UUID from the header GUID and those | |
1712 | * properties of the subarray that we expect to remain constant. | |
1713 | */ | |
1714 | memset(buf, 0, sizeof(buf)); | |
1715 | p = buf; | |
1716 | memcpy(p, ddf->anchor.guid, DDF_GUID_LEN); | |
1717 | p += DDF_GUID_LEN; | |
1718 | memcpy(p, ddf->virt->entries[vcnum].name, 16); | |
1719 | p += 16; | |
1720 | *((__u16 *) p) = vcnum; | |
1721 | sha1_init_ctx(&ctx); | |
1722 | sha1_process_bytes(buf, sizeof(buf), &ctx); | |
1723 | sha1_finish_ctx(&ctx, sha); | |
1724 | memcpy(uuid, sha, 4*4); | |
1725 | } | |
1726 | ||
a322f70c DW |
1727 | static void brief_detail_super_ddf(struct supertype *st) |
1728 | { | |
ff54de6e N |
1729 | struct mdinfo info; |
1730 | char nbuf[64]; | |
bedbf68a | 1731 | struct ddf_super *ddf = st->sb; |
1732 | unsigned int vcnum = get_vd_num_of_subarray(st); | |
1733 | if (vcnum == DDF_CONTAINER) | |
1734 | uuid_from_super_ddf(st, info.uuid); | |
1735 | else if (vcnum == DDF_NOTFOUND) | |
1736 | return; | |
1737 | else | |
7087f02b | 1738 | uuid_of_ddf_subarray(ddf, vcnum, info.uuid); |
ff54de6e N |
1739 | fname_from_uuid(st, &info, nbuf,':'); |
1740 | printf(" UUID=%s", nbuf + 5); | |
a322f70c | 1741 | } |
a322f70c DW |
1742 | #endif |
1743 | ||
1744 | static int match_home_ddf(struct supertype *st, char *homehost) | |
1745 | { | |
1746 | /* It matches 'this' host if the controller is a | |
1747 | * Linux-MD controller with vendor_data matching | |
1748 | * the hostname | |
1749 | */ | |
1750 | struct ddf_super *ddf = st->sb; | |
f21e18ca | 1751 | unsigned int len; |
d1d3482b N |
1752 | |
1753 | if (!homehost) | |
1754 | return 0; | |
1755 | len = strlen(homehost); | |
a322f70c DW |
1756 | |
1757 | return (memcmp(ddf->controller.guid, T10, 8) == 0 && | |
1758 | len < sizeof(ddf->controller.vendor_data) && | |
1759 | memcmp(ddf->controller.vendor_data, homehost,len) == 0 && | |
1760 | ddf->controller.vendor_data[len] == 0); | |
1761 | } | |
1762 | ||
0e600426 | 1763 | #ifndef MDASSEMBLE |
baba3f4e | 1764 | static int find_index_in_bvd(const struct ddf_super *ddf, |
1765 | const struct vd_config *conf, unsigned int n, | |
1766 | unsigned int *n_bvd) | |
1767 | { | |
1768 | /* | |
1769 | * Find the index of the n-th valid physical disk in this BVD | |
1770 | */ | |
1771 | unsigned int i, j; | |
1772 | for (i = 0, j = 0; i < ddf->mppe && | |
a8173e43 | 1773 | j < be16_to_cpu(conf->prim_elmnt_count); i++) { |
60931cf9 | 1774 | if (be32_to_cpu(conf->phys_refnum[i]) != 0xffffffff) { |
baba3f4e | 1775 | if (n == j) { |
1776 | *n_bvd = i; | |
1777 | return 1; | |
1778 | } | |
1779 | j++; | |
1780 | } | |
1781 | } | |
1782 | dprintf("%s: couldn't find BVD member %u (total %u)\n", | |
a8173e43 | 1783 | __func__, n, be16_to_cpu(conf->prim_elmnt_count)); |
baba3f4e | 1784 | return 0; |
1785 | } | |
1786 | ||
1787 | static struct vd_config *find_vdcr(struct ddf_super *ddf, unsigned int inst, | |
1788 | unsigned int n, | |
1789 | unsigned int *n_bvd, struct vcl **vcl) | |
a322f70c | 1790 | { |
7a7cc504 | 1791 | struct vcl *v; |
59e36268 | 1792 | |
baba3f4e | 1793 | for (v = ddf->conflist; v; v = v->next) { |
84e32e19 | 1794 | unsigned int nsec, ibvd = 0; |
baba3f4e | 1795 | struct vd_config *conf; |
1796 | if (inst != v->vcnum) | |
1797 | continue; | |
1798 | conf = &v->conf; | |
1799 | if (conf->sec_elmnt_count == 1) { | |
1800 | if (find_index_in_bvd(ddf, conf, n, n_bvd)) { | |
1801 | *vcl = v; | |
1802 | return conf; | |
1803 | } else | |
1804 | goto bad; | |
1805 | } | |
1806 | if (v->other_bvds == NULL) { | |
1807 | pr_err("%s: BUG: other_bvds is NULL, nsec=%u\n", | |
1808 | __func__, conf->sec_elmnt_count); | |
1809 | goto bad; | |
1810 | } | |
a8173e43 | 1811 | nsec = n / be16_to_cpu(conf->prim_elmnt_count); |
baba3f4e | 1812 | if (conf->sec_elmnt_seq != nsec) { |
1813 | for (ibvd = 1; ibvd < conf->sec_elmnt_count; ibvd++) { | |
baba3f4e | 1814 | if (v->other_bvds[ibvd-1]->sec_elmnt_seq |
1815 | == nsec) | |
1816 | break; | |
1817 | } | |
1818 | if (ibvd == conf->sec_elmnt_count) | |
1819 | goto bad; | |
1820 | conf = v->other_bvds[ibvd-1]; | |
1821 | } | |
1822 | if (!find_index_in_bvd(ddf, conf, | |
1823 | n - nsec*conf->sec_elmnt_count, n_bvd)) | |
1824 | goto bad; | |
1825 | dprintf("%s: found disk %u as member %u in bvd %d of array %u\n" | |
84e32e19 | 1826 | , __func__, n, *n_bvd, ibvd, inst); |
baba3f4e | 1827 | *vcl = v; |
1828 | return conf; | |
1829 | } | |
1830 | bad: | |
1831 | pr_err("%s: Could't find disk %d in array %u\n", __func__, n, inst); | |
7a7cc504 NB |
1832 | return NULL; |
1833 | } | |
0e600426 | 1834 | #endif |
7a7cc504 | 1835 | |
60931cf9 | 1836 | static int find_phys(const struct ddf_super *ddf, be32 phys_refnum) |
7a7cc504 NB |
1837 | { |
1838 | /* Find the entry in phys_disk which has the given refnum | |
1839 | * and return it's index | |
1840 | */ | |
f21e18ca | 1841 | unsigned int i; |
a8173e43 | 1842 | for (i = 0; i < be16_to_cpu(ddf->phys->max_pdes); i++) |
60931cf9 | 1843 | if (be32_eq(ddf->phys->entries[i].refnum, phys_refnum)) |
7a7cc504 NB |
1844 | return i; |
1845 | return -1; | |
a322f70c DW |
1846 | } |
1847 | ||
bedbf68a | 1848 | static void uuid_from_ddf_guid(const char *guid, int uuid[4]) |
1849 | { | |
1850 | char buf[20]; | |
1851 | struct sha1_ctx ctx; | |
1852 | sha1_init_ctx(&ctx); | |
1853 | sha1_process_bytes(guid, DDF_GUID_LEN, &ctx); | |
1854 | sha1_finish_ctx(&ctx, buf); | |
1855 | memcpy(uuid, buf, 4*4); | |
1856 | } | |
1857 | ||
a322f70c DW |
1858 | static void uuid_from_super_ddf(struct supertype *st, int uuid[4]) |
1859 | { | |
1860 | /* The uuid returned here is used for: | |
1861 | * uuid to put into bitmap file (Create, Grow) | |
1862 | * uuid for backup header when saving critical section (Grow) | |
1863 | * comparing uuids when re-adding a device into an array | |
51006d85 N |
1864 | * In these cases the uuid required is that of the data-array, |
1865 | * not the device-set. | |
1866 | * uuid to recognise same set when adding a missing device back | |
1867 | * to an array. This is a uuid for the device-set. | |
613b0d17 | 1868 | * |
a322f70c DW |
1869 | * For each of these we can make do with a truncated |
1870 | * or hashed uuid rather than the original, as long as | |
1871 | * everyone agrees. | |
a322f70c DW |
1872 | * In the case of SVD we assume the BVD is of interest, |
1873 | * though that might be the case if a bitmap were made for | |
1874 | * a mirrored SVD - worry about that later. | |
1875 | * So we need to find the VD configuration record for the | |
1876 | * relevant BVD and extract the GUID and Secondary_Element_Seq. | |
1877 | * The first 16 bytes of the sha1 of these is used. | |
1878 | */ | |
1879 | struct ddf_super *ddf = st->sb; | |
d2ca6449 | 1880 | struct vcl *vcl = ddf->currentconf; |
a322f70c | 1881 | |
c5afc314 | 1882 | if (vcl) |
7087f02b | 1883 | uuid_of_ddf_subarray(ddf, vcl->vcnum, uuid); |
c5afc314 | 1884 | else |
7087f02b | 1885 | uuid_from_ddf_guid(ddf->anchor.guid, uuid); |
a322f70c DW |
1886 | } |
1887 | ||
a5d85af7 | 1888 | static void getinfo_super_ddf_bvd(struct supertype *st, struct mdinfo *info, char *map); |
78e44928 | 1889 | |
a5d85af7 | 1890 | static void getinfo_super_ddf(struct supertype *st, struct mdinfo *info, char *map) |
a322f70c DW |
1891 | { |
1892 | struct ddf_super *ddf = st->sb; | |
a5d85af7 | 1893 | int map_disks = info->array.raid_disks; |
90fa1a29 | 1894 | __u32 *cptr; |
a322f70c | 1895 | |
78e44928 | 1896 | if (ddf->currentconf) { |
a5d85af7 | 1897 | getinfo_super_ddf_bvd(st, info, map); |
78e44928 NB |
1898 | return; |
1899 | } | |
95eeceeb | 1900 | memset(info, 0, sizeof(*info)); |
78e44928 | 1901 | |
a8173e43 | 1902 | info->array.raid_disks = be16_to_cpu(ddf->phys->used_pdes); |
a322f70c DW |
1903 | info->array.level = LEVEL_CONTAINER; |
1904 | info->array.layout = 0; | |
1905 | info->array.md_minor = -1; | |
90fa1a29 JS |
1906 | cptr = (__u32 *)(ddf->anchor.guid + 16); |
1907 | info->array.ctime = DECADE + __be32_to_cpu(*cptr); | |
1908 | ||
a322f70c DW |
1909 | info->array.utime = 0; |
1910 | info->array.chunk_size = 0; | |
510242aa | 1911 | info->container_enough = 1; |
a322f70c | 1912 | |
a322f70c DW |
1913 | info->disk.major = 0; |
1914 | info->disk.minor = 0; | |
cba0191b | 1915 | if (ddf->dlist) { |
f0e876ce | 1916 | struct phys_disk_entry *pde = NULL; |
60931cf9 | 1917 | info->disk.number = be32_to_cpu(ddf->dlist->disk.refnum); |
59e36268 | 1918 | info->disk.raid_disk = find_phys(ddf, ddf->dlist->disk.refnum); |
d2ca6449 | 1919 | |
9d0c6b70 | 1920 | info->data_offset = be64_to_cpu(ddf->phys-> |
613b0d17 N |
1921 | entries[info->disk.raid_disk]. |
1922 | config_size); | |
d2ca6449 | 1923 | info->component_size = ddf->dlist->size - info->data_offset; |
f0e876ce N |
1924 | if (info->disk.raid_disk >= 0) |
1925 | pde = ddf->phys->entries + info->disk.raid_disk; | |
1926 | if (pde && | |
1927 | !(be16_to_cpu(pde->state) & DDF_Failed)) | |
1928 | info->disk.state = (1 << MD_DISK_SYNC) | (1 << MD_DISK_ACTIVE); | |
1929 | else | |
1930 | info->disk.state = 1 << MD_DISK_FAULTY; | |
eba2859f N |
1931 | |
1932 | info->events = be32_to_cpu(ddf->active->seq); | |
cba0191b NB |
1933 | } else { |
1934 | info->disk.number = -1; | |
661dce36 | 1935 | info->disk.raid_disk = -1; |
cba0191b | 1936 | // info->disk.raid_disk = find refnum in the table and use index; |
f0e876ce | 1937 | info->disk.state = (1 << MD_DISK_SYNC) | (1 << MD_DISK_ACTIVE); |
cba0191b | 1938 | } |
a19c88b8 | 1939 | |
921d9e16 | 1940 | info->recovery_start = MaxSector; |
a19c88b8 | 1941 | info->reshape_active = 0; |
6e75048b | 1942 | info->recovery_blocked = 0; |
c5afc314 | 1943 | info->name[0] = 0; |
a322f70c | 1944 | |
f35f2525 N |
1945 | info->array.major_version = -1; |
1946 | info->array.minor_version = -2; | |
159c3a1a | 1947 | strcpy(info->text_version, "ddf"); |
a67dd8cc | 1948 | info->safe_mode_delay = 0; |
159c3a1a | 1949 | |
c5afc314 | 1950 | uuid_from_super_ddf(st, info->uuid); |
a322f70c | 1951 | |
a5d85af7 N |
1952 | if (map) { |
1953 | int i; | |
1954 | for (i = 0 ; i < map_disks; i++) { | |
1955 | if (i < info->array.raid_disks && | |
a8173e43 | 1956 | !(be16_to_cpu(ddf->phys->entries[i].state) |
1957 | & DDF_Failed)) | |
a5d85af7 N |
1958 | map[i] = 1; |
1959 | else | |
1960 | map[i] = 0; | |
1961 | } | |
1962 | } | |
a322f70c DW |
1963 | } |
1964 | ||
8bf989d8 | 1965 | /* size of name must be at least 17 bytes! */ |
1966 | static void _ddf_array_name(char *name, const struct ddf_super *ddf, int i) | |
1967 | { | |
1968 | int j; | |
1969 | memcpy(name, ddf->virt->entries[i].name, 16); | |
1970 | name[16] = 0; | |
1971 | for(j = 0; j < 16; j++) | |
1972 | if (name[j] == ' ') | |
1973 | name[j] = 0; | |
1974 | } | |
1975 | ||
a5d85af7 | 1976 | static void getinfo_super_ddf_bvd(struct supertype *st, struct mdinfo *info, char *map) |
a322f70c DW |
1977 | { |
1978 | struct ddf_super *ddf = st->sb; | |
d2ca6449 NB |
1979 | struct vcl *vc = ddf->currentconf; |
1980 | int cd = ddf->currentdev; | |
ddf94a43 | 1981 | int n_prim; |
db42fa9b | 1982 | int j; |
8592f29d | 1983 | struct dl *dl; |
a5d85af7 | 1984 | int map_disks = info->array.raid_disks; |
90fa1a29 | 1985 | __u32 *cptr; |
ddf94a43 | 1986 | struct vd_config *conf; |
a322f70c | 1987 | |
95eeceeb | 1988 | memset(info, 0, sizeof(*info)); |
8a2848a7 | 1989 | if (layout_ddf2md(&vc->conf, &info->array) == -1) |
1990 | return; | |
a322f70c | 1991 | info->array.md_minor = -1; |
90fa1a29 JS |
1992 | cptr = (__u32 *)(vc->conf.guid + 16); |
1993 | info->array.ctime = DECADE + __be32_to_cpu(*cptr); | |
60931cf9 | 1994 | info->array.utime = DECADE + be32_to_cpu(vc->conf.timestamp); |
d2ca6449 | 1995 | info->array.chunk_size = 512 << vc->conf.chunk_shift; |
da9b4a62 | 1996 | info->custom_array_size = 0; |
d2ca6449 | 1997 | |
ddf94a43 | 1998 | conf = &vc->conf; |
a8173e43 | 1999 | n_prim = be16_to_cpu(conf->prim_elmnt_count); |
ddf94a43 | 2000 | if (conf->sec_elmnt_count > 1 && cd >= n_prim) { |
2001 | int ibvd = cd / n_prim - 1; | |
2002 | cd %= n_prim; | |
2003 | conf = vc->other_bvds[ibvd]; | |
2004 | } | |
2005 | ||
f21e18ca | 2006 | if (cd >= 0 && (unsigned)cd < ddf->mppe) { |
57a66662 | 2007 | info->data_offset = |
9d0c6b70 | 2008 | be64_to_cpu(LBA_OFFSET(ddf, conf)[cd]); |
d2ca6449 NB |
2009 | if (vc->block_sizes) |
2010 | info->component_size = vc->block_sizes[cd]; | |
2011 | else | |
9d0c6b70 | 2012 | info->component_size = be64_to_cpu(conf->blocks); |
d2ca6449 | 2013 | } |
a322f70c | 2014 | |
fb204fb2 | 2015 | for (dl = ddf->dlist; dl ; dl = dl->next) |
60931cf9 | 2016 | if (be32_eq(dl->disk.refnum, conf->phys_refnum[cd])) |
fb204fb2 N |
2017 | break; |
2018 | ||
a322f70c DW |
2019 | info->disk.major = 0; |
2020 | info->disk.minor = 0; | |
fb204fb2 | 2021 | info->disk.state = 0; |
8592f29d N |
2022 | if (dl) { |
2023 | info->disk.major = dl->major; | |
2024 | info->disk.minor = dl->minor; | |
7c3fb3ec | 2025 | info->disk.raid_disk = cd + conf->sec_elmnt_seq |
a8173e43 | 2026 | * be16_to_cpu(conf->prim_elmnt_count); |
fb204fb2 | 2027 | info->disk.number = dl->pdnum; |
f0e876ce N |
2028 | info->disk.state = 0; |
2029 | if (info->disk.number >= 0 && | |
2030 | (be16_to_cpu(ddf->phys->entries[info->disk.number].state) & DDF_Online) && | |
2031 | !(be16_to_cpu(ddf->phys->entries[info->disk.number].state) & DDF_Failed)) | |
2032 | info->disk.state = (1<<MD_DISK_SYNC)|(1<<MD_DISK_ACTIVE); | |
eba2859f | 2033 | info->events = be32_to_cpu(ddf->active->seq); |
8592f29d | 2034 | } |
a322f70c | 2035 | |
103f2410 NB |
2036 | info->container_member = ddf->currentconf->vcnum; |
2037 | ||
921d9e16 | 2038 | info->recovery_start = MaxSector; |
80d26cb2 | 2039 | info->resync_start = 0; |
624c5ad4 | 2040 | info->reshape_active = 0; |
6e75048b | 2041 | info->recovery_blocked = 0; |
80d26cb2 NB |
2042 | if (!(ddf->virt->entries[info->container_member].state |
2043 | & DDF_state_inconsistent) && | |
2044 | (ddf->virt->entries[info->container_member].init_state | |
2045 | & DDF_initstate_mask) | |
2046 | == DDF_init_full) | |
b7528a20 | 2047 | info->resync_start = MaxSector; |
80d26cb2 | 2048 | |
a322f70c DW |
2049 | uuid_from_super_ddf(st, info->uuid); |
2050 | ||
f35f2525 N |
2051 | info->array.major_version = -1; |
2052 | info->array.minor_version = -2; | |
9b63e648 | 2053 | sprintf(info->text_version, "/%s/%d", |
4dd2df09 | 2054 | st->container_devnm, |
9b63e648 | 2055 | info->container_member); |
5684fff6 | 2056 | info->safe_mode_delay = DDF_SAFE_MODE_DELAY; |
159c3a1a | 2057 | |
8bf989d8 | 2058 | _ddf_array_name(info->name, ddf, info->container_member); |
a5d85af7 N |
2059 | |
2060 | if (map) | |
2061 | for (j = 0; j < map_disks; j++) { | |
2062 | map[j] = 0; | |
2063 | if (j < info->array.raid_disks) { | |
2064 | int i = find_phys(ddf, vc->conf.phys_refnum[j]); | |
613b0d17 | 2065 | if (i >= 0 && |
a8173e43 | 2066 | (be16_to_cpu(ddf->phys->entries[i].state) |
2067 | & DDF_Online) && | |
2068 | !(be16_to_cpu(ddf->phys->entries[i].state) | |
2069 | & DDF_Failed)) | |
a5d85af7 N |
2070 | map[i] = 1; |
2071 | } | |
2072 | } | |
a322f70c DW |
2073 | } |
2074 | ||
2075 | static int update_super_ddf(struct supertype *st, struct mdinfo *info, | |
2076 | char *update, | |
2077 | char *devname, int verbose, | |
2078 | int uuid_set, char *homehost) | |
2079 | { | |
2080 | /* For 'assemble' and 'force' we need to return non-zero if any | |
2081 | * change was made. For others, the return value is ignored. | |
2082 | * Update options are: | |
2083 | * force-one : This device looks a bit old but needs to be included, | |
2084 | * update age info appropriately. | |
2085 | * assemble: clear any 'faulty' flag to allow this device to | |
2086 | * be assembled. | |
2087 | * force-array: Array is degraded but being forced, mark it clean | |
2088 | * if that will be needed to assemble it. | |
2089 | * | |
2090 | * newdev: not used ???? | |
2091 | * grow: Array has gained a new device - this is currently for | |
2092 | * linear only | |
2093 | * resync: mark as dirty so a resync will happen. | |
59e36268 | 2094 | * uuid: Change the uuid of the array to match what is given |
a322f70c DW |
2095 | * homehost: update the recorded homehost |
2096 | * name: update the name - preserving the homehost | |
2097 | * _reshape_progress: record new reshape_progress position. | |
2098 | * | |
2099 | * Following are not relevant for this version: | |
2100 | * sparc2.2 : update from old dodgey metadata | |
2101 | * super-minor: change the preferred_minor number | |
2102 | * summaries: update redundant counters. | |
2103 | */ | |
2104 | int rv = 0; | |
2105 | // struct ddf_super *ddf = st->sb; | |
7a7cc504 | 2106 | // struct vd_config *vd = find_vdcr(ddf, info->container_member); |
a322f70c DW |
2107 | // struct virtual_entry *ve = find_ve(ddf); |
2108 | ||
a322f70c DW |
2109 | /* we don't need to handle "force-*" or "assemble" as |
2110 | * there is no need to 'trick' the kernel. We the metadata is | |
2111 | * first updated to activate the array, all the implied modifications | |
2112 | * will just happen. | |
2113 | */ | |
2114 | ||
2115 | if (strcmp(update, "grow") == 0) { | |
2116 | /* FIXME */ | |
1e2b2765 | 2117 | } else if (strcmp(update, "resync") == 0) { |
a322f70c | 2118 | // info->resync_checkpoint = 0; |
1e2b2765 | 2119 | } else if (strcmp(update, "homehost") == 0) { |
a322f70c DW |
2120 | /* homehost is stored in controller->vendor_data, |
2121 | * or it is when we are the vendor | |
2122 | */ | |
2123 | // if (info->vendor_is_local) | |
2124 | // strcpy(ddf->controller.vendor_data, homehost); | |
1e2b2765 | 2125 | rv = -1; |
f49208ec | 2126 | } else if (strcmp(update, "name") == 0) { |
a322f70c DW |
2127 | /* name is stored in virtual_entry->name */ |
2128 | // memset(ve->name, ' ', 16); | |
2129 | // strncpy(ve->name, info->name, 16); | |
1e2b2765 | 2130 | rv = -1; |
f49208ec | 2131 | } else if (strcmp(update, "_reshape_progress") == 0) { |
a322f70c | 2132 | /* We don't support reshape yet */ |
f49208ec N |
2133 | } else if (strcmp(update, "assemble") == 0 ) { |
2134 | /* Do nothing, just succeed */ | |
2135 | rv = 0; | |
1e2b2765 N |
2136 | } else |
2137 | rv = -1; | |
a322f70c DW |
2138 | |
2139 | // update_all_csum(ddf); | |
2140 | ||
2141 | return rv; | |
2142 | } | |
2143 | ||
5f8097be NB |
2144 | static void make_header_guid(char *guid) |
2145 | { | |
60931cf9 | 2146 | be32 stamp; |
5f8097be NB |
2147 | /* Create a DDF Header of Virtual Disk GUID */ |
2148 | ||
2149 | /* 24 bytes of fiction required. | |
2150 | * first 8 are a 'vendor-id' - "Linux-MD" | |
2151 | * next 8 are controller type.. how about 0X DEAD BEEF 0000 0000 | |
2152 | * Remaining 8 random number plus timestamp | |
2153 | */ | |
2154 | memcpy(guid, T10, sizeof(T10)); | |
60931cf9 | 2155 | stamp = cpu_to_be32(0xdeadbeef); |
5f8097be | 2156 | memcpy(guid+8, &stamp, 4); |
60931cf9 | 2157 | stamp = cpu_to_be32(0); |
5f8097be | 2158 | memcpy(guid+12, &stamp, 4); |
60931cf9 | 2159 | stamp = cpu_to_be32(time(0) - DECADE); |
5f8097be | 2160 | memcpy(guid+16, &stamp, 4); |
60931cf9 | 2161 | stamp._v32 = random32(); |
5f8097be | 2162 | memcpy(guid+20, &stamp, 4); |
5f8097be | 2163 | } |
59e36268 | 2164 | |
fb9d0acb | 2165 | static unsigned int find_unused_vde(const struct ddf_super *ddf) |
2166 | { | |
2167 | unsigned int i; | |
a8173e43 | 2168 | for (i = 0; i < be16_to_cpu(ddf->virt->max_vdes); i++) { |
fb9d0acb | 2169 | if (all_ff(ddf->virt->entries[i].guid)) |
2170 | return i; | |
2171 | } | |
2172 | return DDF_NOTFOUND; | |
2173 | } | |
2174 | ||
2175 | static unsigned int find_vde_by_name(const struct ddf_super *ddf, | |
2176 | const char *name) | |
2177 | { | |
2178 | unsigned int i; | |
2179 | if (name == NULL) | |
2180 | return DDF_NOTFOUND; | |
a8173e43 | 2181 | for (i = 0; i < be16_to_cpu(ddf->virt->max_vdes); i++) { |
fb9d0acb | 2182 | if (all_ff(ddf->virt->entries[i].guid)) |
2183 | continue; | |
2184 | if (!strncmp(name, ddf->virt->entries[i].name, | |
2185 | sizeof(ddf->virt->entries[i].name))) | |
2186 | return i; | |
2187 | } | |
2188 | return DDF_NOTFOUND; | |
2189 | } | |
2190 | ||
4441541f | 2191 | #ifndef MDASSEMBLE |
fb9d0acb | 2192 | static unsigned int find_vde_by_guid(const struct ddf_super *ddf, |
2193 | const char *guid) | |
2194 | { | |
2195 | unsigned int i; | |
2196 | if (guid == NULL || all_ff(guid)) | |
2197 | return DDF_NOTFOUND; | |
a8173e43 | 2198 | for (i = 0; i < be16_to_cpu(ddf->virt->max_vdes); i++) |
fb9d0acb | 2199 | if (!memcmp(ddf->virt->entries[i].guid, guid, DDF_GUID_LEN)) |
2200 | return i; | |
2201 | return DDF_NOTFOUND; | |
2202 | } | |
4441541f | 2203 | #endif |
fb9d0acb | 2204 | |
78e44928 NB |
2205 | static int init_super_ddf_bvd(struct supertype *st, |
2206 | mdu_array_info_t *info, | |
2207 | unsigned long long size, | |
2208 | char *name, char *homehost, | |
83cd1e97 | 2209 | int *uuid, unsigned long long data_offset); |
78e44928 | 2210 | |
a322f70c DW |
2211 | static int init_super_ddf(struct supertype *st, |
2212 | mdu_array_info_t *info, | |
2213 | unsigned long long size, char *name, char *homehost, | |
83cd1e97 | 2214 | int *uuid, unsigned long long data_offset) |
a322f70c DW |
2215 | { |
2216 | /* This is primarily called by Create when creating a new array. | |
2217 | * We will then get add_to_super called for each component, and then | |
2218 | * write_init_super called to write it out to each device. | |
2219 | * For DDF, Create can create on fresh devices or on a pre-existing | |
2220 | * array. | |
2221 | * To create on a pre-existing array a different method will be called. | |
2222 | * This one is just for fresh drives. | |
2223 | * | |
2224 | * We need to create the entire 'ddf' structure which includes: | |
2225 | * DDF headers - these are easy. | |
2226 | * Controller data - a Sector describing this controller .. not that | |
2227 | * this is a controller exactly. | |
2228 | * Physical Disk Record - one entry per device, so | |
2229 | * leave plenty of space. | |
2230 | * Virtual Disk Records - again, just leave plenty of space. | |
2231 | * This just lists VDs, doesn't give details | |
2232 | * Config records - describes the VDs that use this disk | |
2233 | * DiskData - describes 'this' device. | |
2234 | * BadBlockManagement - empty | |
2235 | * Diag Space - empty | |
2236 | * Vendor Logs - Could we put bitmaps here? | |
2237 | * | |
2238 | */ | |
2239 | struct ddf_super *ddf; | |
2240 | char hostname[17]; | |
2241 | int hostlen; | |
a322f70c DW |
2242 | int max_phys_disks, max_virt_disks; |
2243 | unsigned long long sector; | |
2244 | int clen; | |
2245 | int i; | |
2246 | int pdsize, vdsize; | |
2247 | struct phys_disk *pd; | |
2248 | struct virtual_disk *vd; | |
2249 | ||
83cd1e97 | 2250 | if (data_offset != INVALID_SECTORS) { |
ed503f89 | 2251 | pr_err("data-offset not supported by DDF\n"); |
83cd1e97 N |
2252 | return 0; |
2253 | } | |
2254 | ||
78e44928 | 2255 | if (st->sb) |
83cd1e97 N |
2256 | return init_super_ddf_bvd(st, info, size, name, homehost, uuid, |
2257 | data_offset); | |
ba7eb04f | 2258 | |
3d2c4fc7 | 2259 | if (posix_memalign((void**)&ddf, 512, sizeof(*ddf)) != 0) { |
e7b84f9d | 2260 | pr_err("%s could not allocate superblock\n", __func__); |
3d2c4fc7 DW |
2261 | return 0; |
2262 | } | |
6264b437 | 2263 | memset(ddf, 0, sizeof(*ddf)); |
a322f70c DW |
2264 | ddf->dlist = NULL; /* no physical disks yet */ |
2265 | ddf->conflist = NULL; /* No virtual disks yet */ | |
955e9ea1 DW |
2266 | st->sb = ddf; |
2267 | ||
2268 | if (info == NULL) { | |
2269 | /* zeroing superblock */ | |
2270 | return 0; | |
2271 | } | |
a322f70c DW |
2272 | |
2273 | /* At least 32MB *must* be reserved for the ddf. So let's just | |
2274 | * start 32MB from the end, and put the primary header there. | |
2275 | * Don't do secondary for now. | |
2276 | * We don't know exactly where that will be yet as it could be | |
2277 | * different on each device. To just set up the lengths. | |
2278 | * | |
2279 | */ | |
2280 | ||
2281 | ddf->anchor.magic = DDF_HEADER_MAGIC; | |
5f8097be | 2282 | make_header_guid(ddf->anchor.guid); |
a322f70c | 2283 | |
59e36268 | 2284 | memcpy(ddf->anchor.revision, DDF_REVISION_2, 8); |
60931cf9 | 2285 | ddf->anchor.seq = cpu_to_be32(1); |
2286 | ddf->anchor.timestamp = cpu_to_be32(time(0) - DECADE); | |
a322f70c DW |
2287 | ddf->anchor.openflag = 0xFF; |
2288 | ddf->anchor.foreignflag = 0; | |
2289 | ddf->anchor.enforcegroups = 0; /* Is this best?? */ | |
2290 | ddf->anchor.pad0 = 0xff; | |
2291 | memset(ddf->anchor.pad1, 0xff, 12); | |
2292 | memset(ddf->anchor.header_ext, 0xff, 32); | |
9d0c6b70 | 2293 | ddf->anchor.primary_lba = cpu_to_be64(~(__u64)0); |
2294 | ddf->anchor.secondary_lba = cpu_to_be64(~(__u64)0); | |
a322f70c DW |
2295 | ddf->anchor.type = DDF_HEADER_ANCHOR; |
2296 | memset(ddf->anchor.pad2, 0xff, 3); | |
60931cf9 | 2297 | ddf->anchor.workspace_len = cpu_to_be32(32768); /* Must be reserved */ |
9d0c6b70 | 2298 | /* Put this at bottom of 32M reserved.. */ |
2299 | ddf->anchor.workspace_lba = cpu_to_be64(~(__u64)0); | |
a322f70c | 2300 | max_phys_disks = 1023; /* Should be enough */ |
a8173e43 | 2301 | ddf->anchor.max_pd_entries = cpu_to_be16(max_phys_disks); |
a322f70c | 2302 | max_virt_disks = 255; |
a8173e43 | 2303 | ddf->anchor.max_vd_entries = cpu_to_be16(max_virt_disks); /* ?? */ |
2304 | ddf->anchor.max_partitions = cpu_to_be16(64); /* ?? */ | |
a322f70c | 2305 | ddf->max_part = 64; |
8c3b8c2c | 2306 | ddf->mppe = 256; |
59e36268 | 2307 | ddf->conf_rec_len = 1 + ROUND_UP(ddf->mppe * (4+8), 512)/512; |
a8173e43 | 2308 | ddf->anchor.config_record_len = cpu_to_be16(ddf->conf_rec_len); |
2309 | ddf->anchor.max_primary_element_entries = cpu_to_be16(ddf->mppe); | |
a322f70c | 2310 | memset(ddf->anchor.pad3, 0xff, 54); |
a322f70c DW |
2311 | /* controller sections is one sector long immediately |
2312 | * after the ddf header */ | |
2313 | sector = 1; | |
60931cf9 | 2314 | ddf->anchor.controller_section_offset = cpu_to_be32(sector); |
2315 | ddf->anchor.controller_section_length = cpu_to_be32(1); | |
a322f70c DW |
2316 | sector += 1; |
2317 | ||
2318 | /* phys is 8 sectors after that */ | |
2319 | pdsize = ROUND_UP(sizeof(struct phys_disk) + | |
2320 | sizeof(struct phys_disk_entry)*max_phys_disks, | |
2321 | 512); | |
2322 | switch(pdsize/512) { | |
2323 | case 2: case 8: case 32: case 128: case 512: break; | |
2324 | default: abort(); | |
2325 | } | |
60931cf9 | 2326 | ddf->anchor.phys_section_offset = cpu_to_be32(sector); |
a322f70c | 2327 | ddf->anchor.phys_section_length = |
60931cf9 | 2328 | cpu_to_be32(pdsize/512); /* max_primary_element_entries/8 */ |
a322f70c DW |
2329 | sector += pdsize/512; |
2330 | ||
2331 | /* virt is another 32 sectors */ | |
2332 | vdsize = ROUND_UP(sizeof(struct virtual_disk) + | |
2333 | sizeof(struct virtual_entry) * max_virt_disks, | |
2334 | 512); | |
2335 | switch(vdsize/512) { | |
2336 | case 2: case 8: case 32: case 128: case 512: break; | |
2337 | default: abort(); | |
2338 | } | |
60931cf9 | 2339 | ddf->anchor.virt_section_offset = cpu_to_be32(sector); |
a322f70c | 2340 | ddf->anchor.virt_section_length = |
60931cf9 | 2341 | cpu_to_be32(vdsize/512); /* max_vd_entries/8 */ |
a322f70c DW |
2342 | sector += vdsize/512; |
2343 | ||
59e36268 | 2344 | clen = ddf->conf_rec_len * (ddf->max_part+1); |
60931cf9 | 2345 | ddf->anchor.config_section_offset = cpu_to_be32(sector); |
2346 | ddf->anchor.config_section_length = cpu_to_be32(clen); | |
a322f70c DW |
2347 | sector += clen; |
2348 | ||
60931cf9 | 2349 | ddf->anchor.data_section_offset = cpu_to_be32(sector); |
2350 | ddf->anchor.data_section_length = cpu_to_be32(1); | |
a322f70c DW |
2351 | sector += 1; |
2352 | ||
60931cf9 | 2353 | ddf->anchor.bbm_section_length = cpu_to_be32(0); |
2354 | ddf->anchor.bbm_section_offset = cpu_to_be32(0xFFFFFFFF); | |
2355 | ddf->anchor.diag_space_length = cpu_to_be32(0); | |
2356 | ddf->anchor.diag_space_offset = cpu_to_be32(0xFFFFFFFF); | |
2357 | ddf->anchor.vendor_length = cpu_to_be32(0); | |
2358 | ddf->anchor.vendor_offset = cpu_to_be32(0xFFFFFFFF); | |
a322f70c DW |
2359 | |
2360 | memset(ddf->anchor.pad4, 0xff, 256); | |
2361 | ||
2362 | memcpy(&ddf->primary, &ddf->anchor, 512); | |
2363 | memcpy(&ddf->secondary, &ddf->anchor, 512); | |
2364 | ||
2365 | ddf->primary.openflag = 1; /* I guess.. */ | |
2366 | ddf->primary.type = DDF_HEADER_PRIMARY; | |
2367 | ||
2368 | ddf->secondary.openflag = 1; /* I guess.. */ | |
2369 | ddf->secondary.type = DDF_HEADER_SECONDARY; | |
2370 | ||
2371 | ddf->active = &ddf->primary; | |
2372 | ||
2373 | ddf->controller.magic = DDF_CONTROLLER_MAGIC; | |
2374 | ||
2375 | /* 24 more bytes of fiction required. | |
2376 | * first 8 are a 'vendor-id' - "Linux-MD" | |
2377 | * Remaining 16 are serial number.... maybe a hostname would do? | |
2378 | */ | |
2379 | memcpy(ddf->controller.guid, T10, sizeof(T10)); | |
1ba6bff9 DW |
2380 | gethostname(hostname, sizeof(hostname)); |
2381 | hostname[sizeof(hostname) - 1] = 0; | |
a322f70c DW |
2382 | hostlen = strlen(hostname); |
2383 | memcpy(ddf->controller.guid + 24 - hostlen, hostname, hostlen); | |
2384 | for (i = strlen(T10) ; i+hostlen < 24; i++) | |
2385 | ddf->controller.guid[i] = ' '; | |
2386 | ||
a8173e43 | 2387 | ddf->controller.type.vendor_id = cpu_to_be16(0xDEAD); |
2388 | ddf->controller.type.device_id = cpu_to_be16(0xBEEF); | |
2389 | ddf->controller.type.sub_vendor_id = cpu_to_be16(0); | |
2390 | ddf->controller.type.sub_device_id = cpu_to_be16(0); | |
a322f70c DW |
2391 | memcpy(ddf->controller.product_id, "What Is My PID??", 16); |
2392 | memset(ddf->controller.pad, 0xff, 8); | |
2393 | memset(ddf->controller.vendor_data, 0xff, 448); | |
a9e1c11d N |
2394 | if (homehost && strlen(homehost) < 440) |
2395 | strcpy((char*)ddf->controller.vendor_data, homehost); | |
a322f70c | 2396 | |
3d2c4fc7 | 2397 | if (posix_memalign((void**)&pd, 512, pdsize) != 0) { |
e7b84f9d | 2398 | pr_err("%s could not allocate pd\n", __func__); |
3d2c4fc7 DW |
2399 | return 0; |
2400 | } | |
6416d527 | 2401 | ddf->phys = pd; |
a322f70c DW |
2402 | ddf->pdsize = pdsize; |
2403 | ||
2404 | memset(pd, 0xff, pdsize); | |
2405 | memset(pd, 0, sizeof(*pd)); | |
076515ba | 2406 | pd->magic = DDF_PHYS_RECORDS_MAGIC; |
a8173e43 | 2407 | pd->used_pdes = cpu_to_be16(0); |
2408 | pd->max_pdes = cpu_to_be16(max_phys_disks); | |
a322f70c | 2409 | memset(pd->pad, 0xff, 52); |
4a3ca8ac | 2410 | for (i = 0; i < max_phys_disks; i++) |
2411 | memset(pd->entries[i].guid, 0xff, DDF_GUID_LEN); | |
a322f70c | 2412 | |
3d2c4fc7 | 2413 | if (posix_memalign((void**)&vd, 512, vdsize) != 0) { |
e7b84f9d | 2414 | pr_err("%s could not allocate vd\n", __func__); |
3d2c4fc7 DW |
2415 | return 0; |
2416 | } | |
6416d527 | 2417 | ddf->virt = vd; |
a322f70c DW |
2418 | ddf->vdsize = vdsize; |
2419 | memset(vd, 0, vdsize); | |
2420 | vd->magic = DDF_VIRT_RECORDS_MAGIC; | |
a8173e43 | 2421 | vd->populated_vdes = cpu_to_be16(0); |
2422 | vd->max_vdes = cpu_to_be16(max_virt_disks); | |
a322f70c DW |
2423 | memset(vd->pad, 0xff, 52); |
2424 | ||
5f8097be NB |
2425 | for (i=0; i<max_virt_disks; i++) |
2426 | memset(&vd->entries[i], 0xff, sizeof(struct virtual_entry)); | |
2427 | ||
a322f70c | 2428 | st->sb = ddf; |
7d5a7ff3 | 2429 | ddf_set_updates_pending(ddf); |
a322f70c DW |
2430 | return 1; |
2431 | } | |
2432 | ||
5f8097be NB |
2433 | static int chunk_to_shift(int chunksize) |
2434 | { | |
2435 | return ffs(chunksize/512)-1; | |
2436 | } | |
2437 | ||
0e600426 | 2438 | #ifndef MDASSEMBLE |
59e36268 NB |
2439 | struct extent { |
2440 | unsigned long long start, size; | |
2441 | }; | |
78e44928 | 2442 | static int cmp_extent(const void *av, const void *bv) |
59e36268 NB |
2443 | { |
2444 | const struct extent *a = av; | |
2445 | const struct extent *b = bv; | |
2446 | if (a->start < b->start) | |
2447 | return -1; | |
2448 | if (a->start > b->start) | |
2449 | return 1; | |
2450 | return 0; | |
2451 | } | |
2452 | ||
78e44928 | 2453 | static struct extent *get_extents(struct ddf_super *ddf, struct dl *dl) |
59e36268 NB |
2454 | { |
2455 | /* find a list of used extents on the give physical device | |
2456 | * (dnum) of the given ddf. | |
2457 | * Return a malloced array of 'struct extent' | |
2458 | ||
613b0d17 | 2459 | * FIXME ignore DDF_Legacy devices? |
59e36268 NB |
2460 | |
2461 | */ | |
2462 | struct extent *rv; | |
2463 | int n = 0; | |
fcc22180 | 2464 | unsigned int i; |
60056e1c | 2465 | __u16 state = be16_to_cpu(ddf->phys->entries[dl->pdnum].state); |
2466 | ||
2467 | if ((state & (DDF_Online|DDF_Failed|DDF_Missing)) != DDF_Online) | |
2468 | return NULL; | |
59e36268 | 2469 | |
503975b9 | 2470 | rv = xmalloc(sizeof(struct extent) * (ddf->max_part + 2)); |
59e36268 NB |
2471 | |
2472 | for (i = 0; i < ddf->max_part; i++) { | |
fcc22180 | 2473 | const struct vd_config *bvd; |
2474 | unsigned int ibvd; | |
59e36268 | 2475 | struct vcl *v = dl->vlist[i]; |
fcc22180 | 2476 | if (v == NULL || |
2477 | get_pd_index_from_refnum(v, dl->disk.refnum, ddf->mppe, | |
2478 | &bvd, &ibvd) == DDF_NOTFOUND) | |
59e36268 | 2479 | continue; |
9d0c6b70 | 2480 | rv[n].start = be64_to_cpu(LBA_OFFSET(ddf, bvd)[ibvd]); |
2481 | rv[n].size = be64_to_cpu(bvd->blocks); | |
fcc22180 | 2482 | n++; |
59e36268 NB |
2483 | } |
2484 | qsort(rv, n, sizeof(*rv), cmp_extent); | |
2485 | ||
9d0c6b70 | 2486 | rv[n].start = be64_to_cpu(ddf->phys->entries[dl->pdnum].config_size); |
59e36268 NB |
2487 | rv[n].size = 0; |
2488 | return rv; | |
2489 | } | |
0e600426 | 2490 | #endif |
59e36268 | 2491 | |
5f8097be NB |
2492 | static int init_super_ddf_bvd(struct supertype *st, |
2493 | mdu_array_info_t *info, | |
2494 | unsigned long long size, | |
2495 | char *name, char *homehost, | |
83cd1e97 | 2496 | int *uuid, unsigned long long data_offset) |
5f8097be NB |
2497 | { |
2498 | /* We are creating a BVD inside a pre-existing container. | |
2499 | * so st->sb is already set. | |
2500 | * We need to create a new vd_config and a new virtual_entry | |
2501 | */ | |
2502 | struct ddf_super *ddf = st->sb; | |
5aaf6c7b | 2503 | unsigned int venum, i; |
5f8097be NB |
2504 | struct virtual_entry *ve; |
2505 | struct vcl *vcl; | |
2506 | struct vd_config *vc; | |
5f8097be | 2507 | |
fb9d0acb | 2508 | if (find_vde_by_name(ddf, name) != DDF_NOTFOUND) { |
2509 | pr_err("This ddf already has an array called %s\n", name); | |
5f8097be NB |
2510 | return 0; |
2511 | } | |
fb9d0acb | 2512 | venum = find_unused_vde(ddf); |
2513 | if (venum == DDF_NOTFOUND) { | |
2514 | pr_err("Cannot find spare slot for virtual disk\n"); | |
5f8097be NB |
2515 | return 0; |
2516 | } | |
2517 | ve = &ddf->virt->entries[venum]; | |
2518 | ||
2519 | /* A Virtual Disk GUID contains the T10 Vendor ID, controller type, | |
2520 | * timestamp, random number | |
2521 | */ | |
2522 | make_header_guid(ve->guid); | |
a8173e43 | 2523 | ve->unit = cpu_to_be16(info->md_minor); |
5f8097be | 2524 | ve->pad0 = 0xFFFF; |
a8173e43 | 2525 | ve->guid_crc._v16 = crc32(0, (unsigned char *)ddf->anchor.guid, |
2526 | DDF_GUID_LEN); | |
2527 | ve->type = cpu_to_be16(0); | |
7a7cc504 NB |
2528 | ve->state = DDF_state_degraded; /* Will be modified as devices are added */ |
2529 | if (info->state & 1) /* clean */ | |
2530 | ve->init_state = DDF_init_full; | |
2531 | else | |
2532 | ve->init_state = DDF_init_not; | |
2533 | ||
5f8097be NB |
2534 | memset(ve->pad1, 0xff, 14); |
2535 | memset(ve->name, ' ', 16); | |
2536 | if (name) | |
2537 | strncpy(ve->name, name, 16); | |
2538 | ddf->virt->populated_vdes = | |
a8173e43 | 2539 | cpu_to_be16(be16_to_cpu(ddf->virt->populated_vdes)+1); |
5f8097be NB |
2540 | |
2541 | /* Now create a new vd_config */ | |
3d2c4fc7 DW |
2542 | if (posix_memalign((void**)&vcl, 512, |
2543 | (offsetof(struct vcl, conf) + ddf->conf_rec_len * 512)) != 0) { | |
e7b84f9d | 2544 | pr_err("%s could not allocate vd_config\n", __func__); |
3d2c4fc7 DW |
2545 | return 0; |
2546 | } | |
59e36268 NB |
2547 | vcl->vcnum = venum; |
2548 | vcl->block_sizes = NULL; /* FIXME not for CONCAT */ | |
5f8097be NB |
2549 | vc = &vcl->conf; |
2550 | ||
2551 | vc->magic = DDF_VD_CONF_MAGIC; | |
2552 | memcpy(vc->guid, ve->guid, DDF_GUID_LEN); | |
60931cf9 | 2553 | vc->timestamp = cpu_to_be32(time(0)-DECADE); |
2554 | vc->seqnum = cpu_to_be32(1); | |
5f8097be | 2555 | memset(vc->pad0, 0xff, 24); |
5f8097be | 2556 | vc->chunk_shift = chunk_to_shift(info->chunk_size); |
a3163bf0 | 2557 | if (layout_md2ddf(info, vc) == -1 || |
a8173e43 | 2558 | be16_to_cpu(vc->prim_elmnt_count) > ddf->mppe) { |
a3163bf0 | 2559 | pr_err("%s: unsupported RAID level/layout %d/%d with %d disks\n", |
2560 | __func__, info->level, info->layout, info->raid_disks); | |
2561 | free(vcl); | |
2562 | return 0; | |
2563 | } | |
5f8097be | 2564 | vc->sec_elmnt_seq = 0; |
3c48f7be | 2565 | if (alloc_other_bvds(ddf, vcl) != 0) { |
2566 | pr_err("%s could not allocate other bvds\n", | |
2567 | __func__); | |
2568 | free(vcl); | |
2569 | return 0; | |
2570 | } | |
9d0c6b70 | 2571 | vc->blocks = cpu_to_be64(info->size * 2); |
2572 | vc->array_blocks = cpu_to_be64( | |
5f8097be NB |
2573 | calc_array_size(info->level, info->raid_disks, info->layout, |
2574 | info->chunk_size, info->size*2)); | |
2575 | memset(vc->pad1, 0xff, 8); | |
60931cf9 | 2576 | vc->spare_refs[0] = cpu_to_be32(0xffffffff); |
2577 | vc->spare_refs[1] = cpu_to_be32(0xffffffff); | |
2578 | vc->spare_refs[2] = cpu_to_be32(0xffffffff); | |
2579 | vc->spare_refs[3] = cpu_to_be32(0xffffffff); | |
2580 | vc->spare_refs[4] = cpu_to_be32(0xffffffff); | |
2581 | vc->spare_refs[5] = cpu_to_be32(0xffffffff); | |
2582 | vc->spare_refs[6] = cpu_to_be32(0xffffffff); | |
2583 | vc->spare_refs[7] = cpu_to_be32(0xffffffff); | |
5f8097be NB |
2584 | memset(vc->cache_pol, 0, 8); |
2585 | vc->bg_rate = 0x80; | |
2586 | memset(vc->pad2, 0xff, 3); | |
2587 | memset(vc->pad3, 0xff, 52); | |
2588 | memset(vc->pad4, 0xff, 192); | |
2589 | memset(vc->v0, 0xff, 32); | |
2590 | memset(vc->v1, 0xff, 32); | |
2591 | memset(vc->v2, 0xff, 16); | |
2592 | memset(vc->v3, 0xff, 16); | |
2593 | memset(vc->vendor, 0xff, 32); | |
598f0d58 | 2594 | |
8c3b8c2c | 2595 | memset(vc->phys_refnum, 0xff, 4*ddf->mppe); |
e5a2a3cf | 2596 | memset(vc->phys_refnum+ddf->mppe, 0x00, 8*ddf->mppe); |
5f8097be | 2597 | |
5aaf6c7b | 2598 | for (i = 1; i < vc->sec_elmnt_count; i++) { |
2599 | memcpy(vcl->other_bvds[i-1], vc, ddf->conf_rec_len * 512); | |
2600 | vcl->other_bvds[i-1]->sec_elmnt_seq = i; | |
2601 | } | |
2602 | ||
5f8097be NB |
2603 | vcl->next = ddf->conflist; |
2604 | ddf->conflist = vcl; | |
d2ca6449 | 2605 | ddf->currentconf = vcl; |
7d5a7ff3 | 2606 | ddf_set_updates_pending(ddf); |
5f8097be NB |
2607 | return 1; |
2608 | } | |
2609 | ||
63eb2454 | 2610 | |
0e600426 | 2611 | #ifndef MDASSEMBLE |
4441541f N |
2612 | static int get_svd_state(const struct ddf_super *, const struct vcl *); |
2613 | ||
5f8097be NB |
2614 | static void add_to_super_ddf_bvd(struct supertype *st, |
2615 | mdu_disk_info_t *dk, int fd, char *devname) | |
2616 | { | |
2617 | /* fd and devname identify a device with-in the ddf container (st). | |
2618 | * dk identifies a location in the new BVD. | |
2619 | * We need to find suitable free space in that device and update | |
2620 | * the phys_refnum and lba_offset for the newly created vd_config. | |
2621 | * We might also want to update the type in the phys_disk | |
5575e7d9 | 2622 | * section. |
8592f29d N |
2623 | * |
2624 | * Alternately: fd == -1 and we have already chosen which device to | |
2625 | * use and recorded in dlist->raid_disk; | |
5f8097be NB |
2626 | */ |
2627 | struct dl *dl; | |
2628 | struct ddf_super *ddf = st->sb; | |
2629 | struct vd_config *vc; | |
f21e18ca | 2630 | unsigned int i; |
59e36268 NB |
2631 | unsigned long long blocks, pos, esize; |
2632 | struct extent *ex; | |
475ccbdb | 2633 | unsigned int raid_disk = dk->raid_disk; |
5f8097be | 2634 | |
8592f29d N |
2635 | if (fd == -1) { |
2636 | for (dl = ddf->dlist; dl ; dl = dl->next) | |
2637 | if (dl->raiddisk == dk->raid_disk) | |
2638 | break; | |
2639 | } else { | |
2640 | for (dl = ddf->dlist; dl ; dl = dl->next) | |
2641 | if (dl->major == dk->major && | |
2642 | dl->minor == dk->minor) | |
2643 | break; | |
2644 | } | |
5f8097be NB |
2645 | if (!dl || ! (dk->state & (1<<MD_DISK_SYNC))) |
2646 | return; | |
2647 | ||
d2ca6449 | 2648 | vc = &ddf->currentconf->conf; |
475ccbdb | 2649 | if (vc->sec_elmnt_count > 1) { |
a8173e43 | 2650 | unsigned int n = be16_to_cpu(vc->prim_elmnt_count); |
475ccbdb | 2651 | if (raid_disk >= n) |
2652 | vc = ddf->currentconf->other_bvds[raid_disk / n - 1]; | |
2653 | raid_disk %= n; | |
2654 | } | |
59e36268 NB |
2655 | |
2656 | ex = get_extents(ddf, dl); | |
2657 | if (!ex) | |
2658 | return; | |
2659 | ||
2660 | i = 0; pos = 0; | |
9d0c6b70 | 2661 | blocks = be64_to_cpu(vc->blocks); |
d2ca6449 NB |
2662 | if (ddf->currentconf->block_sizes) |
2663 | blocks = ddf->currentconf->block_sizes[dk->raid_disk]; | |
59e36268 NB |
2664 | |
2665 | do { | |
2666 | esize = ex[i].start - pos; | |
2667 | if (esize >= blocks) | |
2668 | break; | |
2669 | pos = ex[i].start + ex[i].size; | |
2670 | i++; | |
2671 | } while (ex[i-1].size); | |
2672 | ||
2673 | free(ex); | |
2674 | if (esize < blocks) | |
2675 | return; | |
2676 | ||
d2ca6449 | 2677 | ddf->currentdev = dk->raid_disk; |
475ccbdb | 2678 | vc->phys_refnum[raid_disk] = dl->disk.refnum; |
9d0c6b70 | 2679 | LBA_OFFSET(ddf, vc)[raid_disk] = cpu_to_be64(pos); |
5f8097be | 2680 | |
f21e18ca | 2681 | for (i = 0; i < ddf->max_part ; i++) |
5575e7d9 NB |
2682 | if (dl->vlist[i] == NULL) |
2683 | break; | |
2684 | if (i == ddf->max_part) | |
2685 | return; | |
d2ca6449 | 2686 | dl->vlist[i] = ddf->currentconf; |
5f8097be | 2687 | |
8592f29d N |
2688 | if (fd >= 0) |
2689 | dl->fd = fd; | |
2690 | if (devname) | |
2691 | dl->devname = devname; | |
7a7cc504 | 2692 | |
63eb2454 | 2693 | /* Check if we can mark array as optimal yet */ |
d2ca6449 | 2694 | i = ddf->currentconf->vcnum; |
63eb2454 | 2695 | ddf->virt->entries[i].state = |
2696 | (ddf->virt->entries[i].state & ~DDF_state_mask) | |
2697 | | get_svd_state(ddf, ddf->currentconf); | |
a8173e43 | 2698 | be16_clear(ddf->phys->entries[dl->pdnum].type, |
2699 | cpu_to_be16(DDF_Global_Spare)); | |
2700 | be16_set(ddf->phys->entries[dl->pdnum].type, | |
2701 | cpu_to_be16(DDF_Active_in_VD)); | |
4f9bbe63 | 2702 | dprintf("%s: added disk %d/%08x to VD %d/%s as disk %d\n", |
60931cf9 | 2703 | __func__, dl->pdnum, be32_to_cpu(dl->disk.refnum), |
4f9bbe63 | 2704 | ddf->currentconf->vcnum, guid_str(vc->guid), |
2705 | dk->raid_disk); | |
7d5a7ff3 | 2706 | ddf_set_updates_pending(ddf); |
5f8097be NB |
2707 | } |
2708 | ||
4a3ca8ac | 2709 | static unsigned int find_unused_pde(const struct ddf_super *ddf) |
2710 | { | |
2711 | unsigned int i; | |
a8173e43 | 2712 | for (i = 0; i < be16_to_cpu(ddf->phys->max_pdes); i++) { |
4a3ca8ac | 2713 | if (all_ff(ddf->phys->entries[i].guid)) |
2714 | return i; | |
2715 | } | |
2716 | return DDF_NOTFOUND; | |
2717 | } | |
2718 | ||
105e6e93 | 2719 | static void _set_config_size(struct phys_disk_entry *pde, const struct dl *dl) |
2720 | { | |
2721 | __u64 cfs, t; | |
2722 | cfs = min(dl->size - 32*1024*2ULL, be64_to_cpu(dl->primary_lba)); | |
2723 | t = be64_to_cpu(dl->secondary_lba); | |
2724 | if (t != ~(__u64)0) | |
2725 | cfs = min(cfs, t); | |
2726 | /* | |
2727 | * Some vendor DDF structures interpret workspace_lba | |
2728 | * very differently then us. Make a sanity check on the value. | |
2729 | */ | |
2730 | t = be64_to_cpu(dl->workspace_lba); | |
2731 | if (t < cfs) { | |
2732 | __u64 wsp = cfs - t; | |
2733 | if (wsp > 1024*1024*2ULL && wsp > dl->size / 16) { | |
2734 | pr_err("%s: %x:%x: workspace size 0x%llx too big, ignoring\n", | |
2735 | __func__, dl->major, dl->minor, wsp); | |
2736 | } else | |
2737 | cfs = t; | |
2738 | } | |
2739 | pde->config_size = cpu_to_be64(cfs); | |
2740 | dprintf("%s: %x:%x config_size %llx, DDF structure is %llx blocks\n", | |
2741 | __func__, dl->major, dl->minor, cfs, dl->size-cfs); | |
2742 | } | |
2743 | ||
a322f70c DW |
2744 | /* add a device to a container, either while creating it or while |
2745 | * expanding a pre-existing container | |
2746 | */ | |
f20c3968 | 2747 | static int add_to_super_ddf(struct supertype *st, |
72ca9bcf N |
2748 | mdu_disk_info_t *dk, int fd, char *devname, |
2749 | unsigned long long data_offset) | |
a322f70c DW |
2750 | { |
2751 | struct ddf_super *ddf = st->sb; | |
2752 | struct dl *dd; | |
2753 | time_t now; | |
2754 | struct tm *tm; | |
2755 | unsigned long long size; | |
2756 | struct phys_disk_entry *pde; | |
f21e18ca | 2757 | unsigned int n, i; |
a322f70c | 2758 | struct stat stb; |
90fa1a29 | 2759 | __u32 *tptr; |
a322f70c | 2760 | |
78e44928 NB |
2761 | if (ddf->currentconf) { |
2762 | add_to_super_ddf_bvd(st, dk, fd, devname); | |
f20c3968 | 2763 | return 0; |
78e44928 NB |
2764 | } |
2765 | ||
a322f70c DW |
2766 | /* This is device numbered dk->number. We need to create |
2767 | * a phys_disk entry and a more detailed disk_data entry. | |
2768 | */ | |
2769 | fstat(fd, &stb); | |
4a3ca8ac | 2770 | n = find_unused_pde(ddf); |
2771 | if (n == DDF_NOTFOUND) { | |
2772 | pr_err("%s: No free slot in array, cannot add disk\n", | |
2773 | __func__); | |
2774 | return 1; | |
2775 | } | |
2776 | pde = &ddf->phys->entries[n]; | |
4ee8cca9 | 2777 | get_dev_size(fd, NULL, &size); |
2778 | if (size <= 32*1024*1024) { | |
2779 | pr_err("%s: device size must be at least 32MB\n", | |
2780 | __func__); | |
2781 | return 1; | |
2782 | } | |
2783 | size >>= 9; | |
4a3ca8ac | 2784 | |
3d2c4fc7 DW |
2785 | if (posix_memalign((void**)&dd, 512, |
2786 | sizeof(*dd) + sizeof(dd->vlist[0]) * ddf->max_part) != 0) { | |
e7b84f9d N |
2787 | pr_err("%s could allocate buffer for new disk, aborting\n", |
2788 | __func__); | |
f20c3968 | 2789 | return 1; |
3d2c4fc7 | 2790 | } |
a322f70c DW |
2791 | dd->major = major(stb.st_rdev); |
2792 | dd->minor = minor(stb.st_rdev); | |
2793 | dd->devname = devname; | |
a322f70c | 2794 | dd->fd = fd; |
b2280677 | 2795 | dd->spare = NULL; |
a322f70c DW |
2796 | |
2797 | dd->disk.magic = DDF_PHYS_DATA_MAGIC; | |
2798 | now = time(0); | |
2799 | tm = localtime(&now); | |
2800 | sprintf(dd->disk.guid, "%8s%04d%02d%02d", | |
2801 | T10, tm->tm_year+1900, tm->tm_mon+1, tm->tm_mday); | |
90fa1a29 JS |
2802 | tptr = (__u32 *)(dd->disk.guid + 16); |
2803 | *tptr++ = random32(); | |
2804 | *tptr = random32(); | |
a322f70c | 2805 | |
59e36268 NB |
2806 | do { |
2807 | /* Cannot be bothered finding a CRC of some irrelevant details*/ | |
60931cf9 | 2808 | dd->disk.refnum._v32 = random32(); |
a8173e43 | 2809 | for (i = be16_to_cpu(ddf->active->max_pd_entries); |
f21e18ca | 2810 | i > 0; i--) |
60931cf9 | 2811 | if (be32_eq(ddf->phys->entries[i-1].refnum, |
2812 | dd->disk.refnum)) | |
59e36268 | 2813 | break; |
f21e18ca | 2814 | } while (i > 0); |
59e36268 | 2815 | |
a322f70c DW |
2816 | dd->disk.forced_ref = 1; |
2817 | dd->disk.forced_guid = 1; | |
2818 | memset(dd->disk.vendor, ' ', 32); | |
2819 | memcpy(dd->disk.vendor, "Linux", 5); | |
2820 | memset(dd->disk.pad, 0xff, 442); | |
b2280677 | 2821 | for (i = 0; i < ddf->max_part ; i++) |
a322f70c DW |
2822 | dd->vlist[i] = NULL; |
2823 | ||
5575e7d9 NB |
2824 | dd->pdnum = n; |
2825 | ||
2cc2983d N |
2826 | if (st->update_tail) { |
2827 | int len = (sizeof(struct phys_disk) + | |
2828 | sizeof(struct phys_disk_entry)); | |
2829 | struct phys_disk *pd; | |
2830 | ||
503975b9 | 2831 | pd = xmalloc(len); |
2cc2983d | 2832 | pd->magic = DDF_PHYS_RECORDS_MAGIC; |
a8173e43 | 2833 | pd->used_pdes = cpu_to_be16(n); |
2cc2983d N |
2834 | pde = &pd->entries[0]; |
2835 | dd->mdupdate = pd; | |
4a3ca8ac | 2836 | } else |
a8173e43 | 2837 | ddf->phys->used_pdes = cpu_to_be16( |
2838 | 1 + be16_to_cpu(ddf->phys->used_pdes)); | |
a322f70c DW |
2839 | |
2840 | memcpy(pde->guid, dd->disk.guid, DDF_GUID_LEN); | |
2841 | pde->refnum = dd->disk.refnum; | |
a8173e43 | 2842 | pde->type = cpu_to_be16(DDF_Forced_PD_GUID | DDF_Global_Spare); |
2843 | pde->state = cpu_to_be16(DDF_Online); | |
4ee8cca9 | 2844 | dd->size = size; |
2845 | /* | |
2846 | * If there is already a device in dlist, try to reserve the same | |
2847 | * amount of workspace. Otherwise, use 32MB. | |
2848 | * We checked disk size above already. | |
2849 | */ | |
2850 | #define __calc_lba(new, old, lba, mb) do { \ | |
2851 | unsigned long long dif; \ | |
2852 | if ((old) != NULL) \ | |
9d0c6b70 | 2853 | dif = (old)->size - be64_to_cpu((old)->lba); \ |
4ee8cca9 | 2854 | else \ |
2855 | dif = (new)->size; \ | |
2856 | if ((new)->size > dif) \ | |
9d0c6b70 | 2857 | (new)->lba = cpu_to_be64((new)->size - dif); \ |
4ee8cca9 | 2858 | else \ |
9d0c6b70 | 2859 | (new)->lba = cpu_to_be64((new)->size - (mb*1024*2)); \ |
4ee8cca9 | 2860 | } while (0) |
2861 | __calc_lba(dd, ddf->dlist, workspace_lba, 32); | |
2862 | __calc_lba(dd, ddf->dlist, primary_lba, 16); | |
b95cb4b9 N |
2863 | if (ddf->dlist == NULL || |
2864 | be64_to_cpu(ddf->dlist->secondary_lba) != ~(__u64)0) | |
2865 | __calc_lba(dd, ddf->dlist, secondary_lba, 32); | |
105e6e93 | 2866 | _set_config_size(pde, dd); |
4ee8cca9 | 2867 | |
a322f70c DW |
2868 | sprintf(pde->path, "%17.17s","Information: nil") ; |
2869 | memset(pde->pad, 0xff, 6); | |
2870 | ||
2cc2983d N |
2871 | if (st->update_tail) { |
2872 | dd->next = ddf->add_list; | |
2873 | ddf->add_list = dd; | |
2874 | } else { | |
2875 | dd->next = ddf->dlist; | |
2876 | ddf->dlist = dd; | |
7d5a7ff3 | 2877 | ddf_set_updates_pending(ddf); |
2cc2983d | 2878 | } |
f20c3968 DW |
2879 | |
2880 | return 0; | |
a322f70c DW |
2881 | } |
2882 | ||
4dd968cc N |
2883 | static int remove_from_super_ddf(struct supertype *st, mdu_disk_info_t *dk) |
2884 | { | |
2885 | struct ddf_super *ddf = st->sb; | |
2886 | struct dl *dl; | |
2887 | ||
2888 | /* mdmon has noticed that this disk (dk->major/dk->minor) has | |
2889 | * disappeared from the container. | |
2890 | * We need to arrange that it disappears from the metadata and | |
2891 | * internal data structures too. | |
2892 | * Most of the work is done by ddf_process_update which edits | |
2893 | * the metadata and closes the file handle and attaches the memory | |
2894 | * where free_updates will free it. | |
2895 | */ | |
2896 | for (dl = ddf->dlist; dl ; dl = dl->next) | |
2897 | if (dl->major == dk->major && | |
2898 | dl->minor == dk->minor) | |
2899 | break; | |
2900 | if (!dl) | |
2901 | return -1; | |
2902 | ||
2903 | if (st->update_tail) { | |
2904 | int len = (sizeof(struct phys_disk) + | |
2905 | sizeof(struct phys_disk_entry)); | |
2906 | struct phys_disk *pd; | |
2907 | ||
503975b9 | 2908 | pd = xmalloc(len); |
4dd968cc | 2909 | pd->magic = DDF_PHYS_RECORDS_MAGIC; |
a8173e43 | 2910 | pd->used_pdes = cpu_to_be16(dl->pdnum); |
2911 | pd->entries[0].state = cpu_to_be16(DDF_Missing); | |
4dd968cc N |
2912 | append_metadata_update(st, pd, len); |
2913 | } | |
2914 | return 0; | |
2915 | } | |
4441541f | 2916 | #endif |
4dd968cc | 2917 | |
a322f70c DW |
2918 | /* |
2919 | * This is the write_init_super method for a ddf container. It is | |
2920 | * called when creating a container or adding another device to a | |
2921 | * container. | |
2922 | */ | |
42d5dfd9 | 2923 | #define NULL_CONF_SZ 4096 |
18a2f463 | 2924 | |
3921e41a | 2925 | static int __write_ddf_structure(struct dl *d, struct ddf_super *ddf, __u8 type) |
a322f70c | 2926 | { |
7f798aca | 2927 | unsigned long long sector; |
2928 | struct ddf_header *header; | |
3921e41a | 2929 | int fd, i, n_config, conf_size, buf_size; |
a4057a88 | 2930 | int ret = 0; |
3921e41a | 2931 | char *conf; |
8e9387ac | 2932 | |
7f798aca | 2933 | fd = d->fd; |
2934 | ||
2935 | switch (type) { | |
2936 | case DDF_HEADER_PRIMARY: | |
2937 | header = &ddf->primary; | |
9d0c6b70 | 2938 | sector = be64_to_cpu(header->primary_lba); |
7f798aca | 2939 | break; |
2940 | case DDF_HEADER_SECONDARY: | |
2941 | header = &ddf->secondary; | |
9d0c6b70 | 2942 | sector = be64_to_cpu(header->secondary_lba); |
7f798aca | 2943 | break; |
2944 | default: | |
2945 | return 0; | |
2946 | } | |
b95cb4b9 N |
2947 | if (sector == ~(__u64)0) |
2948 | return 0; | |
7f798aca | 2949 | |
2950 | header->type = type; | |
a4057a88 | 2951 | header->openflag = 1; |
7f798aca | 2952 | header->crc = calc_crc(header, 512); |
2953 | ||
2954 | lseek64(fd, sector<<9, 0); | |
2955 | if (write(fd, header, 512) < 0) | |
a4057a88 | 2956 | goto out; |
7f798aca | 2957 | |
2958 | ddf->controller.crc = calc_crc(&ddf->controller, 512); | |
2959 | if (write(fd, &ddf->controller, 512) < 0) | |
a4057a88 | 2960 | goto out; |
a322f70c | 2961 | |
7f798aca | 2962 | ddf->phys->crc = calc_crc(ddf->phys, ddf->pdsize); |
2963 | if (write(fd, ddf->phys, ddf->pdsize) < 0) | |
a4057a88 | 2964 | goto out; |
7f798aca | 2965 | ddf->virt->crc = calc_crc(ddf->virt, ddf->vdsize); |
2966 | if (write(fd, ddf->virt, ddf->vdsize) < 0) | |
a4057a88 | 2967 | goto out; |
7f798aca | 2968 | |
2969 | /* Now write lots of config records. */ | |
2970 | n_config = ddf->max_part; | |
2971 | conf_size = ddf->conf_rec_len * 512; | |
3921e41a N |
2972 | conf = ddf->conf; |
2973 | buf_size = conf_size * (n_config + 1); | |
2974 | if (!conf) { | |
2975 | if (posix_memalign((void**)&conf, 512, buf_size) != 0) | |
2976 | goto out; | |
2977 | ddf->conf = conf; | |
2978 | } | |
7f798aca | 2979 | for (i = 0 ; i <= n_config ; i++) { |
e3c2a365 | 2980 | struct vcl *c; |
2981 | struct vd_config *vdc = NULL; | |
2982 | if (i == n_config) { | |
7f798aca | 2983 | c = (struct vcl *)d->spare; |
e3c2a365 | 2984 | if (c) |
2985 | vdc = &c->conf; | |
2986 | } else { | |
2987 | unsigned int dummy; | |
2988 | c = d->vlist[i]; | |
2989 | if (c) | |
2990 | get_pd_index_from_refnum( | |
2991 | c, d->disk.refnum, | |
2992 | ddf->mppe, | |
2993 | (const struct vd_config **)&vdc, | |
2994 | &dummy); | |
2995 | } | |
7f798aca | 2996 | if (c) { |
be9b9ef4 | 2997 | dprintf("writing conf record %i on disk %08x for %s/%u\n", |
60931cf9 | 2998 | i, be32_to_cpu(d->disk.refnum), |
ad60eea1 | 2999 | guid_str(vdc->guid), |
be9b9ef4 | 3000 | vdc->sec_elmnt_seq); |
dacf3dc5 | 3001 | vdc->seqnum = header->seq; |
e3c2a365 | 3002 | vdc->crc = calc_crc(vdc, conf_size); |
3921e41a | 3003 | memcpy(conf + i*conf_size, vdc, conf_size); |
ce45c819 | 3004 | } else |
3921e41a | 3005 | memset(conf + i*conf_size, 0xff, conf_size); |
7f798aca | 3006 | } |
3921e41a | 3007 | if (write(fd, conf, buf_size) != buf_size) |
a4057a88 | 3008 | goto out; |
7f798aca | 3009 | |
3010 | d->disk.crc = calc_crc(&d->disk, 512); | |
3011 | if (write(fd, &d->disk, 512) < 0) | |
a4057a88 | 3012 | goto out; |
7f798aca | 3013 | |
a4057a88 | 3014 | ret = 1; |
3015 | out: | |
3016 | header->openflag = 0; | |
3017 | header->crc = calc_crc(header, 512); | |
3018 | ||
3019 | lseek64(fd, sector<<9, 0); | |
3020 | if (write(fd, header, 512) < 0) | |
3021 | ret = 0; | |
3022 | ||
3023 | return ret; | |
7f798aca | 3024 | } |
3025 | ||
3921e41a | 3026 | static int _write_super_to_disk(struct ddf_super *ddf, struct dl *d) |
9bf38704 | 3027 | { |
3028 | unsigned long long size; | |
3029 | int fd = d->fd; | |
3030 | if (fd < 0) | |
3031 | return 0; | |
3032 | ||
3033 | /* We need to fill in the primary, (secondary) and workspace | |
3034 | * lba's in the headers, set their checksums, | |
3035 | * Also checksum phys, virt.... | |
3036 | * | |
3037 | * Then write everything out, finally the anchor is written. | |
3038 | */ | |
3039 | get_dev_size(fd, NULL, &size); | |
3040 | size /= 512; | |
9d0c6b70 | 3041 | if (be64_to_cpu(d->workspace_lba) != 0ULL) |
9bf38704 | 3042 | ddf->anchor.workspace_lba = d->workspace_lba; |
3043 | else | |
3044 | ddf->anchor.workspace_lba = | |
9d0c6b70 | 3045 | cpu_to_be64(size - 32*1024*2); |
3046 | if (be64_to_cpu(d->primary_lba) != 0ULL) | |
9bf38704 | 3047 | ddf->anchor.primary_lba = d->primary_lba; |
3048 | else | |
3049 | ddf->anchor.primary_lba = | |
9d0c6b70 | 3050 | cpu_to_be64(size - 16*1024*2); |
3051 | if (be64_to_cpu(d->secondary_lba) != 0ULL) | |
9bf38704 | 3052 | ddf->anchor.secondary_lba = d->secondary_lba; |
3053 | else | |
3054 | ddf->anchor.secondary_lba = | |
9d0c6b70 | 3055 | cpu_to_be64(size - 32*1024*2); |
9bf38704 | 3056 | ddf->anchor.seq = ddf->active->seq; |
3057 | memcpy(&ddf->primary, &ddf->anchor, 512); | |
3058 | memcpy(&ddf->secondary, &ddf->anchor, 512); | |
3059 | ||
3060 | ddf->anchor.openflag = 0xFF; /* 'open' means nothing */ | |
60931cf9 | 3061 | ddf->anchor.seq = cpu_to_be32(0xFFFFFFFF); /* no sequencing in anchor */ |
9bf38704 | 3062 | ddf->anchor.crc = calc_crc(&ddf->anchor, 512); |
3063 | ||
3921e41a | 3064 | if (!__write_ddf_structure(d, ddf, DDF_HEADER_PRIMARY)) |
9bf38704 | 3065 | return 0; |
3066 | ||
3921e41a | 3067 | if (!__write_ddf_structure(d, ddf, DDF_HEADER_SECONDARY)) |
9bf38704 | 3068 | return 0; |
3069 | ||
3070 | lseek64(fd, (size-1)*512, SEEK_SET); | |
3071 | if (write(fd, &ddf->anchor, 512) < 0) | |
3072 | return 0; | |
3073 | ||
3074 | return 1; | |
3075 | } | |
3076 | ||
4441541f | 3077 | #ifndef MDASSEMBLE |
3921e41a | 3078 | static int __write_init_super_ddf(struct supertype *st) |
7f798aca | 3079 | { |
a322f70c | 3080 | struct ddf_super *ddf = st->sb; |
a322f70c | 3081 | struct dl *d; |
175593bf DW |
3082 | int attempts = 0; |
3083 | int successes = 0; | |
42d5dfd9 | 3084 | |
7d5a7ff3 | 3085 | pr_state(ddf, __func__); |
a322f70c | 3086 | |
175593bf DW |
3087 | /* try to write updated metadata, |
3088 | * if we catch a failure move on to the next disk | |
3089 | */ | |
a322f70c | 3090 | for (d = ddf->dlist; d; d=d->next) { |
175593bf | 3091 | attempts++; |
3921e41a | 3092 | successes += _write_super_to_disk(ddf, d); |
175593bf DW |
3093 | } |
3094 | ||
175593bf | 3095 | return attempts != successes; |
a322f70c | 3096 | } |
7a7cc504 NB |
3097 | |
3098 | static int write_init_super_ddf(struct supertype *st) | |
3099 | { | |
9b1fb677 DW |
3100 | struct ddf_super *ddf = st->sb; |
3101 | struct vcl *currentconf = ddf->currentconf; | |
3102 | ||
3103 | /* we are done with currentconf reset it to point st at the container */ | |
3104 | ddf->currentconf = NULL; | |
edd8d13c NB |
3105 | |
3106 | if (st->update_tail) { | |
3107 | /* queue the virtual_disk and vd_config as metadata updates */ | |
3108 | struct virtual_disk *vd; | |
3109 | struct vd_config *vc; | |
c5943560 | 3110 | int len, tlen; |
3111 | unsigned int i; | |
edd8d13c | 3112 | |
9b1fb677 | 3113 | if (!currentconf) { |
2cc2983d N |
3114 | int len = (sizeof(struct phys_disk) + |
3115 | sizeof(struct phys_disk_entry)); | |
3116 | ||
3117 | /* adding a disk to the container. */ | |
3118 | if (!ddf->add_list) | |
3119 | return 0; | |
3120 | ||
3121 | append_metadata_update(st, ddf->add_list->mdupdate, len); | |
3122 | ddf->add_list->mdupdate = NULL; | |
3123 | return 0; | |
3124 | } | |
3125 | ||
3126 | /* Newly created VD */ | |
3127 | ||
edd8d13c NB |
3128 | /* First the virtual disk. We have a slightly fake header */ |
3129 | len = sizeof(struct virtual_disk) + sizeof(struct virtual_entry); | |
503975b9 | 3130 | vd = xmalloc(len); |
edd8d13c | 3131 | *vd = *ddf->virt; |
9b1fb677 | 3132 | vd->entries[0] = ddf->virt->entries[currentconf->vcnum]; |
a8173e43 | 3133 | vd->populated_vdes = cpu_to_be16(currentconf->vcnum); |
edd8d13c NB |
3134 | append_metadata_update(st, vd, len); |
3135 | ||
3136 | /* Then the vd_config */ | |
3137 | len = ddf->conf_rec_len * 512; | |
c5943560 | 3138 | tlen = len * currentconf->conf.sec_elmnt_count; |
3139 | vc = xmalloc(tlen); | |
9b1fb677 | 3140 | memcpy(vc, ¤tconf->conf, len); |
c5943560 | 3141 | for (i = 1; i < currentconf->conf.sec_elmnt_count; i++) |
3142 | memcpy((char *)vc + i*len, currentconf->other_bvds[i-1], | |
3143 | len); | |
3144 | append_metadata_update(st, vc, tlen); | |
edd8d13c NB |
3145 | |
3146 | /* FIXME I need to close the fds! */ | |
3147 | return 0; | |
613b0d17 | 3148 | } else { |
d682f344 | 3149 | struct dl *d; |
19041058 | 3150 | if (!currentconf) |
3151 | for (d = ddf->dlist; d; d=d->next) | |
3152 | while (Kill(d->devname, NULL, 0, -1, 1) == 0); | |
3921e41a | 3153 | return __write_init_super_ddf(st); |
d682f344 | 3154 | } |
7a7cc504 NB |
3155 | } |
3156 | ||
a322f70c DW |
3157 | #endif |
3158 | ||
387fcd59 N |
3159 | static __u64 avail_size_ddf(struct supertype *st, __u64 devsize, |
3160 | unsigned long long data_offset) | |
a322f70c DW |
3161 | { |
3162 | /* We must reserve the last 32Meg */ | |
3163 | if (devsize <= 32*1024*2) | |
3164 | return 0; | |
3165 | return devsize - 32*1024*2; | |
3166 | } | |
3167 | ||
3168 | #ifndef MDASSEMBLE | |
8592f29d N |
3169 | |
3170 | static int reserve_space(struct supertype *st, int raiddisks, | |
3171 | unsigned long long size, int chunk, | |
3172 | unsigned long long *freesize) | |
3173 | { | |
3174 | /* Find 'raiddisks' spare extents at least 'size' big (but | |
3175 | * only caring about multiples of 'chunk') and remember | |
3176 | * them. | |
3177 | * If the cannot be found, fail. | |
3178 | */ | |
3179 | struct dl *dl; | |
3180 | struct ddf_super *ddf = st->sb; | |
3181 | int cnt = 0; | |
3182 | ||
3183 | for (dl = ddf->dlist; dl ; dl=dl->next) { | |
613b0d17 | 3184 | dl->raiddisk = -1; |
8592f29d N |
3185 | dl->esize = 0; |
3186 | } | |
3187 | /* Now find largest extent on each device */ | |
3188 | for (dl = ddf->dlist ; dl ; dl=dl->next) { | |
3189 | struct extent *e = get_extents(ddf, dl); | |
3190 | unsigned long long pos = 0; | |
3191 | int i = 0; | |
3192 | int found = 0; | |
3193 | unsigned long long minsize = size; | |
3194 | ||
3195 | if (size == 0) | |
3196 | minsize = chunk; | |
3197 | ||
3198 | if (!e) | |
3199 | continue; | |
3200 | do { | |
3201 | unsigned long long esize; | |
3202 | esize = e[i].start - pos; | |
3203 | if (esize >= minsize) { | |
3204 | found = 1; | |
3205 | minsize = esize; | |
3206 | } | |
3207 | pos = e[i].start + e[i].size; | |
3208 | i++; | |
3209 | } while (e[i-1].size); | |
3210 | if (found) { | |
3211 | cnt++; | |
3212 | dl->esize = minsize; | |
3213 | } | |
3214 | free(e); | |
3215 | } | |
3216 | if (cnt < raiddisks) { | |
e7b84f9d | 3217 | pr_err("not enough devices with space to create array.\n"); |
8592f29d N |
3218 | return 0; /* No enough free spaces large enough */ |
3219 | } | |
3220 | if (size == 0) { | |
3221 | /* choose the largest size of which there are at least 'raiddisk' */ | |
3222 | for (dl = ddf->dlist ; dl ; dl=dl->next) { | |
3223 | struct dl *dl2; | |
3224 | if (dl->esize <= size) | |
3225 | continue; | |
3226 | /* This is bigger than 'size', see if there are enough */ | |
3227 | cnt = 0; | |
7b80ad6a | 3228 | for (dl2 = ddf->dlist; dl2 ; dl2=dl2->next) |
8592f29d N |
3229 | if (dl2->esize >= dl->esize) |
3230 | cnt++; | |
3231 | if (cnt >= raiddisks) | |
3232 | size = dl->esize; | |
3233 | } | |
3234 | if (chunk) { | |
3235 | size = size / chunk; | |
3236 | size *= chunk; | |
3237 | } | |
3238 | *freesize = size; | |
3239 | if (size < 32) { | |
e7b84f9d | 3240 | pr_err("not enough spare devices to create array.\n"); |
8592f29d N |
3241 | return 0; |
3242 | } | |
3243 | } | |
3244 | /* We have a 'size' of which there are enough spaces. | |
3245 | * We simply do a first-fit */ | |
3246 | cnt = 0; | |
3247 | for (dl = ddf->dlist ; dl && cnt < raiddisks ; dl=dl->next) { | |
3248 | if (dl->esize < size) | |
3249 | continue; | |
613b0d17 | 3250 | |
8592f29d N |
3251 | dl->raiddisk = cnt; |
3252 | cnt++; | |
3253 | } | |
3254 | return 1; | |
3255 | } | |
3256 | ||
2c514b71 NB |
3257 | static int |
3258 | validate_geometry_ddf_container(struct supertype *st, | |
3259 | int level, int layout, int raiddisks, | |
3260 | int chunk, unsigned long long size, | |
af4348dd | 3261 | unsigned long long data_offset, |
2c514b71 NB |
3262 | char *dev, unsigned long long *freesize, |
3263 | int verbose); | |
78e44928 NB |
3264 | |
3265 | static int validate_geometry_ddf_bvd(struct supertype *st, | |
3266 | int level, int layout, int raiddisks, | |
c21e737b | 3267 | int *chunk, unsigned long long size, |
af4348dd | 3268 | unsigned long long data_offset, |
2c514b71 NB |
3269 | char *dev, unsigned long long *freesize, |
3270 | int verbose); | |
78e44928 NB |
3271 | |
3272 | static int validate_geometry_ddf(struct supertype *st, | |
2c514b71 | 3273 | int level, int layout, int raiddisks, |
c21e737b | 3274 | int *chunk, unsigned long long size, |
af4348dd | 3275 | unsigned long long data_offset, |
2c514b71 NB |
3276 | char *dev, unsigned long long *freesize, |
3277 | int verbose) | |
a322f70c DW |
3278 | { |
3279 | int fd; | |
3280 | struct mdinfo *sra; | |
3281 | int cfd; | |
3282 | ||
3283 | /* ddf potentially supports lots of things, but it depends on | |
3284 | * what devices are offered (and maybe kernel version?) | |
3285 | * If given unused devices, we will make a container. | |
3286 | * If given devices in a container, we will make a BVD. | |
3287 | * If given BVDs, we make an SVD, changing all the GUIDs in the process. | |
3288 | */ | |
3289 | ||
7ccc4cc4 | 3290 | if (*chunk == UnSet) |
bb7295f1 N |
3291 | *chunk = DEFAULT_CHUNK; |
3292 | ||
542ef4ec | 3293 | if (level == -1000000) level = LEVEL_CONTAINER; |
a322f70c | 3294 | if (level == LEVEL_CONTAINER) { |
78e44928 NB |
3295 | /* Must be a fresh device to add to a container */ |
3296 | return validate_geometry_ddf_container(st, level, layout, | |
7ccc4cc4 | 3297 | raiddisks, *chunk, |
af4348dd N |
3298 | size, data_offset, dev, |
3299 | freesize, | |
2c514b71 | 3300 | verbose); |
5f8097be NB |
3301 | } |
3302 | ||
78e44928 | 3303 | if (!dev) { |
a3163bf0 | 3304 | mdu_array_info_t array = { |
3305 | .level = level, .layout = layout, | |
3306 | .raid_disks = raiddisks | |
3307 | }; | |
3308 | struct vd_config conf; | |
3309 | if (layout_md2ddf(&array, &conf) == -1) { | |
b42f577a | 3310 | if (verbose) |
94b08b7c | 3311 | pr_err("DDF does not support level %d /layout %d arrays with %d disks\n", |
3312 | level, layout, raiddisks); | |
78e44928 | 3313 | return 0; |
b42f577a | 3314 | } |
78e44928 | 3315 | /* Should check layout? etc */ |
8592f29d N |
3316 | |
3317 | if (st->sb && freesize) { | |
3318 | /* --create was given a container to create in. | |
3319 | * So we need to check that there are enough | |
3320 | * free spaces and return the amount of space. | |
3321 | * We may as well remember which drives were | |
3322 | * chosen so that add_to_super/getinfo_super | |
3323 | * can return them. | |
3324 | */ | |
7ccc4cc4 | 3325 | return reserve_space(st, raiddisks, size, *chunk, freesize); |
8592f29d | 3326 | } |
a322f70c | 3327 | return 1; |
78e44928 | 3328 | } |
a322f70c | 3329 | |
8592f29d N |
3330 | if (st->sb) { |
3331 | /* A container has already been opened, so we are | |
3332 | * creating in there. Maybe a BVD, maybe an SVD. | |
3333 | * Should make a distinction one day. | |
3334 | */ | |
3335 | return validate_geometry_ddf_bvd(st, level, layout, raiddisks, | |
af4348dd N |
3336 | chunk, size, data_offset, dev, |
3337 | freesize, | |
8592f29d N |
3338 | verbose); |
3339 | } | |
78e44928 NB |
3340 | /* This is the first device for the array. |
3341 | * If it is a container, we read it in and do automagic allocations, | |
3342 | * no other devices should be given. | |
3343 | * Otherwise it must be a member device of a container, and we | |
3344 | * do manual allocation. | |
3345 | * Later we should check for a BVD and make an SVD. | |
a322f70c | 3346 | */ |
a322f70c DW |
3347 | fd = open(dev, O_RDONLY|O_EXCL, 0); |
3348 | if (fd >= 0) { | |
4dd2df09 | 3349 | sra = sysfs_read(fd, NULL, GET_VERSION); |
a322f70c DW |
3350 | close(fd); |
3351 | if (sra && sra->array.major_version == -1 && | |
78e44928 NB |
3352 | strcmp(sra->text_version, "ddf") == 0) { |
3353 | ||
3354 | /* load super */ | |
3355 | /* find space for 'n' devices. */ | |
3356 | /* remember the devices */ | |
3357 | /* Somehow return the fact that we have enough */ | |
a322f70c DW |
3358 | } |
3359 | ||
2c514b71 | 3360 | if (verbose) |
e7b84f9d N |
3361 | pr_err("ddf: Cannot create this array " |
3362 | "on device %s - a container is required.\n", | |
3363 | dev); | |
a322f70c DW |
3364 | return 0; |
3365 | } | |
3366 | if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) { | |
2c514b71 | 3367 | if (verbose) |
e7b84f9d | 3368 | pr_err("ddf: Cannot open %s: %s\n", |
613b0d17 | 3369 | dev, strerror(errno)); |
a322f70c DW |
3370 | return 0; |
3371 | } | |
3372 | /* Well, it is in use by someone, maybe a 'ddf' container. */ | |
3373 | cfd = open_container(fd); | |
3374 | if (cfd < 0) { | |
3375 | close(fd); | |
2c514b71 | 3376 | if (verbose) |
e7b84f9d | 3377 | pr_err("ddf: Cannot use %s: %s\n", |
613b0d17 | 3378 | dev, strerror(EBUSY)); |
a322f70c DW |
3379 | return 0; |
3380 | } | |
4dd2df09 | 3381 | sra = sysfs_read(cfd, NULL, GET_VERSION); |
a322f70c DW |
3382 | close(fd); |
3383 | if (sra && sra->array.major_version == -1 && | |
3384 | strcmp(sra->text_version, "ddf") == 0) { | |
3385 | /* This is a member of a ddf container. Load the container | |
3386 | * and try to create a bvd | |
3387 | */ | |
3388 | struct ddf_super *ddf; | |
e1902a7b | 3389 | if (load_super_ddf_all(st, cfd, (void **)&ddf, NULL) == 0) { |
5f8097be | 3390 | st->sb = ddf; |
4dd2df09 | 3391 | strcpy(st->container_devnm, fd2devnm(cfd)); |
a322f70c | 3392 | close(cfd); |
78e44928 | 3393 | return validate_geometry_ddf_bvd(st, level, layout, |
a322f70c | 3394 | raiddisks, chunk, size, |
af4348dd | 3395 | data_offset, |
2c514b71 NB |
3396 | dev, freesize, |
3397 | verbose); | |
a322f70c DW |
3398 | } |
3399 | close(cfd); | |
c42ec1ed DW |
3400 | } else /* device may belong to a different container */ |
3401 | return 0; | |
3402 | ||
a322f70c DW |
3403 | return 1; |
3404 | } | |
3405 | ||
2c514b71 NB |
3406 | static int |
3407 | validate_geometry_ddf_container(struct supertype *st, | |
3408 | int level, int layout, int raiddisks, | |
3409 | int chunk, unsigned long long size, | |
af4348dd | 3410 | unsigned long long data_offset, |
2c514b71 NB |
3411 | char *dev, unsigned long long *freesize, |
3412 | int verbose) | |
a322f70c DW |
3413 | { |
3414 | int fd; | |
3415 | unsigned long long ldsize; | |
3416 | ||
3417 | if (level != LEVEL_CONTAINER) | |
3418 | return 0; | |
3419 | if (!dev) | |
3420 | return 1; | |
3421 | ||
3422 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
3423 | if (fd < 0) { | |
2c514b71 | 3424 | if (verbose) |
e7b84f9d | 3425 | pr_err("ddf: Cannot open %s: %s\n", |
613b0d17 | 3426 | dev, strerror(errno)); |
a322f70c DW |
3427 | return 0; |
3428 | } | |
3429 | if (!get_dev_size(fd, dev, &ldsize)) { | |
3430 | close(fd); | |
3431 | return 0; | |
3432 | } | |
3433 | close(fd); | |
3434 | ||
387fcd59 | 3435 | *freesize = avail_size_ddf(st, ldsize >> 9, INVALID_SECTORS); |
ea17e7aa N |
3436 | if (*freesize == 0) |
3437 | return 0; | |
a322f70c DW |
3438 | |
3439 | return 1; | |
3440 | } | |
3441 | ||
78e44928 NB |
3442 | static int validate_geometry_ddf_bvd(struct supertype *st, |
3443 | int level, int layout, int raiddisks, | |
c21e737b | 3444 | int *chunk, unsigned long long size, |
af4348dd | 3445 | unsigned long long data_offset, |
2c514b71 NB |
3446 | char *dev, unsigned long long *freesize, |
3447 | int verbose) | |
a322f70c DW |
3448 | { |
3449 | struct stat stb; | |
3450 | struct ddf_super *ddf = st->sb; | |
3451 | struct dl *dl; | |
5f8097be NB |
3452 | unsigned long long pos = 0; |
3453 | unsigned long long maxsize; | |
3454 | struct extent *e; | |
3455 | int i; | |
a322f70c | 3456 | /* ddf/bvd supports lots of things, but not containers */ |
b42f577a N |
3457 | if (level == LEVEL_CONTAINER) { |
3458 | if (verbose) | |
e7b84f9d | 3459 | pr_err("DDF cannot create a container within an container\n"); |
a322f70c | 3460 | return 0; |
b42f577a | 3461 | } |
a322f70c DW |
3462 | /* We must have the container info already read in. */ |
3463 | if (!ddf) | |
3464 | return 0; | |
3465 | ||
5f8097be NB |
3466 | if (!dev) { |
3467 | /* General test: make sure there is space for | |
3468 | * 'raiddisks' device extents of size 'size'. | |
3469 | */ | |
3470 | unsigned long long minsize = size; | |
3471 | int dcnt = 0; | |
3472 | if (minsize == 0) | |
3473 | minsize = 8; | |
3474 | for (dl = ddf->dlist; dl ; dl = dl->next) | |
3475 | { | |
3476 | int found = 0; | |
7e1432fb | 3477 | pos = 0; |
5f8097be NB |
3478 | |
3479 | i = 0; | |
3480 | e = get_extents(ddf, dl); | |
3481 | if (!e) continue; | |
3482 | do { | |
3483 | unsigned long long esize; | |
3484 | esize = e[i].start - pos; | |
3485 | if (esize >= minsize) | |
3486 | found = 1; | |
3487 | pos = e[i].start + e[i].size; | |
3488 | i++; | |
3489 | } while (e[i-1].size); | |
3490 | if (found) | |
3491 | dcnt++; | |
3492 | free(e); | |
3493 | } | |
3494 | if (dcnt < raiddisks) { | |
2c514b71 | 3495 | if (verbose) |
e7b84f9d N |
3496 | pr_err("ddf: Not enough devices with " |
3497 | "space for this array (%d < %d)\n", | |
3498 | dcnt, raiddisks); | |
5f8097be NB |
3499 | return 0; |
3500 | } | |
3501 | return 1; | |
3502 | } | |
a322f70c DW |
3503 | /* This device must be a member of the set */ |
3504 | if (stat(dev, &stb) < 0) | |
3505 | return 0; | |
3506 | if ((S_IFMT & stb.st_mode) != S_IFBLK) | |
3507 | return 0; | |
3508 | for (dl = ddf->dlist ; dl ; dl = dl->next) { | |
f21e18ca N |
3509 | if (dl->major == (int)major(stb.st_rdev) && |
3510 | dl->minor == (int)minor(stb.st_rdev)) | |
a322f70c DW |
3511 | break; |
3512 | } | |
5f8097be | 3513 | if (!dl) { |
2c514b71 | 3514 | if (verbose) |
e7b84f9d | 3515 | pr_err("ddf: %s is not in the " |
613b0d17 N |
3516 | "same DDF set\n", |
3517 | dev); | |
5f8097be NB |
3518 | return 0; |
3519 | } | |
3520 | e = get_extents(ddf, dl); | |
3521 | maxsize = 0; | |
3522 | i = 0; | |
3523 | if (e) do { | |
613b0d17 N |
3524 | unsigned long long esize; |
3525 | esize = e[i].start - pos; | |
3526 | if (esize >= maxsize) | |
3527 | maxsize = esize; | |
3528 | pos = e[i].start + e[i].size; | |
3529 | i++; | |
3530 | } while (e[i-1].size); | |
5f8097be | 3531 | *freesize = maxsize; |
a322f70c DW |
3532 | // FIXME here I am |
3533 | ||
3534 | return 1; | |
3535 | } | |
59e36268 | 3536 | |
a322f70c | 3537 | static int load_super_ddf_all(struct supertype *st, int fd, |
e1902a7b | 3538 | void **sbp, char *devname) |
a322f70c DW |
3539 | { |
3540 | struct mdinfo *sra; | |
3541 | struct ddf_super *super; | |
3542 | struct mdinfo *sd, *best = NULL; | |
3543 | int bestseq = 0; | |
3544 | int seq; | |
3545 | char nm[20]; | |
3546 | int dfd; | |
3547 | ||
b526e52d | 3548 | sra = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE); |
a322f70c DW |
3549 | if (!sra) |
3550 | return 1; | |
3551 | if (sra->array.major_version != -1 || | |
3552 | sra->array.minor_version != -2 || | |
3553 | strcmp(sra->text_version, "ddf") != 0) | |
3554 | return 1; | |
3555 | ||
6416d527 | 3556 | if (posix_memalign((void**)&super, 512, sizeof(*super)) != 0) |
a322f70c | 3557 | return 1; |
a2349791 | 3558 | memset(super, 0, sizeof(*super)); |
a322f70c DW |
3559 | |
3560 | /* first, try each device, and choose the best ddf */ | |
3561 | for (sd = sra->devs ; sd ; sd = sd->next) { | |
3562 | int rv; | |
3563 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
7a7cc504 NB |
3564 | dfd = dev_open(nm, O_RDONLY); |
3565 | if (dfd < 0) | |
a322f70c DW |
3566 | return 2; |
3567 | rv = load_ddf_headers(dfd, super, NULL); | |
7a7cc504 | 3568 | close(dfd); |
a322f70c | 3569 | if (rv == 0) { |
60931cf9 | 3570 | seq = be32_to_cpu(super->active->seq); |
a322f70c DW |
3571 | if (super->active->openflag) |
3572 | seq--; | |
3573 | if (!best || seq > bestseq) { | |
3574 | bestseq = seq; | |
3575 | best = sd; | |
3576 | } | |
3577 | } | |
3578 | } | |
3579 | if (!best) | |
3580 | return 1; | |
3581 | /* OK, load this ddf */ | |
3582 | sprintf(nm, "%d:%d", best->disk.major, best->disk.minor); | |
3583 | dfd = dev_open(nm, O_RDONLY); | |
7a7cc504 | 3584 | if (dfd < 0) |
a322f70c DW |
3585 | return 1; |
3586 | load_ddf_headers(dfd, super, NULL); | |
3587 | load_ddf_global(dfd, super, NULL); | |
3588 | close(dfd); | |
3589 | /* Now we need the device-local bits */ | |
3590 | for (sd = sra->devs ; sd ; sd = sd->next) { | |
3d2c4fc7 DW |
3591 | int rv; |
3592 | ||
a322f70c | 3593 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); |
e1902a7b | 3594 | dfd = dev_open(nm, O_RDWR); |
7a7cc504 | 3595 | if (dfd < 0) |
a322f70c | 3596 | return 2; |
3d2c4fc7 DW |
3597 | rv = load_ddf_headers(dfd, super, NULL); |
3598 | if (rv == 0) | |
e1902a7b | 3599 | rv = load_ddf_local(dfd, super, NULL, 1); |
3d2c4fc7 DW |
3600 | if (rv) |
3601 | return 1; | |
a322f70c | 3602 | } |
33414a01 | 3603 | |
a322f70c DW |
3604 | *sbp = super; |
3605 | if (st->ss == NULL) { | |
78e44928 | 3606 | st->ss = &super_ddf; |
a322f70c DW |
3607 | st->minor_version = 0; |
3608 | st->max_devs = 512; | |
3609 | } | |
4dd2df09 | 3610 | strcpy(st->container_devnm, fd2devnm(fd)); |
a322f70c DW |
3611 | return 0; |
3612 | } | |
2b959fbf N |
3613 | |
3614 | static int load_container_ddf(struct supertype *st, int fd, | |
3615 | char *devname) | |
3616 | { | |
3617 | return load_super_ddf_all(st, fd, &st->sb, devname); | |
3618 | } | |
3619 | ||
0e600426 | 3620 | #endif /* MDASSEMBLE */ |
a322f70c | 3621 | |
a5c7adb3 | 3622 | static int check_secondary(const struct vcl *vc) |
3623 | { | |
3624 | const struct vd_config *conf = &vc->conf; | |
3625 | int i; | |
3626 | ||
3627 | /* The only DDF secondary RAID level md can support is | |
3628 | * RAID 10, if the stripe sizes and Basic volume sizes | |
3629 | * are all equal. | |
3630 | * Other configurations could in theory be supported by exposing | |
3631 | * the BVDs to user space and using device mapper for the secondary | |
3632 | * mapping. So far we don't support that. | |
3633 | */ | |
3634 | ||
3635 | __u64 sec_elements[4] = {0, 0, 0, 0}; | |
3636 | #define __set_sec_seen(n) (sec_elements[(n)>>6] |= (1<<((n)&63))) | |
3637 | #define __was_sec_seen(n) ((sec_elements[(n)>>6] & (1<<((n)&63))) != 0) | |
3638 | ||
3639 | if (vc->other_bvds == NULL) { | |
3640 | pr_err("No BVDs for secondary RAID found\n"); | |
3641 | return -1; | |
3642 | } | |
3643 | if (conf->prl != DDF_RAID1) { | |
3644 | pr_err("Secondary RAID level only supported for mirrored BVD\n"); | |
3645 | return -1; | |
3646 | } | |
3647 | if (conf->srl != DDF_2STRIPED && conf->srl != DDF_2SPANNED) { | |
3648 | pr_err("Secondary RAID level %d is unsupported\n", | |
3649 | conf->srl); | |
3650 | return -1; | |
3651 | } | |
3652 | __set_sec_seen(conf->sec_elmnt_seq); | |
3653 | for (i = 0; i < conf->sec_elmnt_count-1; i++) { | |
3654 | const struct vd_config *bvd = vc->other_bvds[i]; | |
3c48f7be | 3655 | if (bvd->sec_elmnt_seq == DDF_UNUSED_BVD) |
c98567ba | 3656 | continue; |
a5c7adb3 | 3657 | if (bvd->srl != conf->srl) { |
3658 | pr_err("Inconsistent secondary RAID level across BVDs\n"); | |
3659 | return -1; | |
3660 | } | |
3661 | if (bvd->prl != conf->prl) { | |
3662 | pr_err("Different RAID levels for BVDs are unsupported\n"); | |
3663 | return -1; | |
3664 | } | |
a8173e43 | 3665 | if (!be16_eq(bvd->prim_elmnt_count, conf->prim_elmnt_count)) { |
a5c7adb3 | 3666 | pr_err("All BVDs must have the same number of primary elements\n"); |
3667 | return -1; | |
3668 | } | |
3669 | if (bvd->chunk_shift != conf->chunk_shift) { | |
3670 | pr_err("Different strip sizes for BVDs are unsupported\n"); | |
3671 | return -1; | |
3672 | } | |
9d0c6b70 | 3673 | if (!be64_eq(bvd->array_blocks, conf->array_blocks)) { |
a5c7adb3 | 3674 | pr_err("Different BVD sizes are unsupported\n"); |
3675 | return -1; | |
3676 | } | |
3677 | __set_sec_seen(bvd->sec_elmnt_seq); | |
3678 | } | |
3679 | for (i = 0; i < conf->sec_elmnt_count; i++) { | |
3680 | if (!__was_sec_seen(i)) { | |
3681 | pr_err("BVD %d is missing\n", i); | |
3682 | return -1; | |
3683 | } | |
3684 | } | |
3685 | return 0; | |
3686 | } | |
3687 | ||
8a38db86 | 3688 | static unsigned int get_pd_index_from_refnum(const struct vcl *vc, |
60931cf9 | 3689 | be32 refnum, unsigned int nmax, |
4e587018 | 3690 | const struct vd_config **bvd, |
3691 | unsigned int *idx) | |
8a38db86 | 3692 | { |
4e587018 | 3693 | unsigned int i, j, n, sec, cnt; |
3694 | ||
a8173e43 | 3695 | cnt = be16_to_cpu(vc->conf.prim_elmnt_count); |
4e587018 | 3696 | sec = (vc->conf.sec_elmnt_count == 1 ? 0 : vc->conf.sec_elmnt_seq); |
3697 | ||
3698 | for (i = 0, j = 0 ; i < nmax ; i++) { | |
3699 | /* j counts valid entries for this BVD */ | |
60931cf9 | 3700 | if (be32_to_cpu(vc->conf.phys_refnum[i]) != 0xffffffff) |
4e587018 | 3701 | j++; |
60931cf9 | 3702 | if (be32_eq(vc->conf.phys_refnum[i], refnum)) { |
4e587018 | 3703 | *bvd = &vc->conf; |
3704 | *idx = i; | |
3705 | return sec * cnt + j - 1; | |
3706 | } | |
3707 | } | |
3708 | if (vc->other_bvds == NULL) | |
3709 | goto bad; | |
3710 | ||
3711 | for (n = 1; n < vc->conf.sec_elmnt_count; n++) { | |
3712 | struct vd_config *vd = vc->other_bvds[n-1]; | |
4e587018 | 3713 | sec = vd->sec_elmnt_seq; |
3c48f7be | 3714 | if (sec == DDF_UNUSED_BVD) |
3715 | continue; | |
4e587018 | 3716 | for (i = 0, j = 0 ; i < nmax ; i++) { |
60931cf9 | 3717 | if (be32_to_cpu(vd->phys_refnum[i]) != 0xffffffff) |
4e587018 | 3718 | j++; |
60931cf9 | 3719 | if (be32_eq(vd->phys_refnum[i], refnum)) { |
4e587018 | 3720 | *bvd = vd; |
3721 | *idx = i; | |
3722 | return sec * cnt + j - 1; | |
3723 | } | |
3724 | } | |
3725 | } | |
3726 | bad: | |
3727 | *bvd = NULL; | |
d6e7b083 | 3728 | return DDF_NOTFOUND; |
8a38db86 | 3729 | } |
3730 | ||
00bbdbda | 3731 | static struct mdinfo *container_content_ddf(struct supertype *st, char *subarray) |
598f0d58 NB |
3732 | { |
3733 | /* Given a container loaded by load_super_ddf_all, | |
3734 | * extract information about all the arrays into | |
3735 | * an mdinfo tree. | |
3736 | * | |
3737 | * For each vcl in conflist: create an mdinfo, fill it in, | |
3738 | * then look for matching devices (phys_refnum) in dlist | |
3739 | * and create appropriate device mdinfo. | |
3740 | */ | |
3741 | struct ddf_super *ddf = st->sb; | |
3742 | struct mdinfo *rest = NULL; | |
3743 | struct vcl *vc; | |
3744 | ||
3745 | for (vc = ddf->conflist ; vc ; vc=vc->next) | |
3746 | { | |
f21e18ca | 3747 | unsigned int i; |
598f0d58 | 3748 | struct mdinfo *this; |
00bbdbda | 3749 | char *ep; |
90fa1a29 | 3750 | __u32 *cptr; |
8a38db86 | 3751 | unsigned int pd; |
00bbdbda N |
3752 | |
3753 | if (subarray && | |
3754 | (strtoul(subarray, &ep, 10) != vc->vcnum || | |
3755 | *ep != '\0')) | |
3756 | continue; | |
3757 | ||
a5c7adb3 | 3758 | if (vc->conf.sec_elmnt_count > 1) { |
3759 | if (check_secondary(vc) != 0) | |
3760 | continue; | |
3761 | } | |
3762 | ||
503975b9 | 3763 | this = xcalloc(1, sizeof(*this)); |
598f0d58 NB |
3764 | this->next = rest; |
3765 | rest = this; | |
3766 | ||
8a2848a7 | 3767 | if (layout_ddf2md(&vc->conf, &this->array)) |
3768 | continue; | |
598f0d58 | 3769 | this->array.md_minor = -1; |
f35f2525 N |
3770 | this->array.major_version = -1; |
3771 | this->array.minor_version = -2; | |
5684fff6 | 3772 | this->safe_mode_delay = DDF_SAFE_MODE_DELAY; |
90fa1a29 JS |
3773 | cptr = (__u32 *)(vc->conf.guid + 16); |
3774 | this->array.ctime = DECADE + __be32_to_cpu(*cptr); | |
598f0d58 | 3775 | this->array.utime = DECADE + |
60931cf9 | 3776 | be32_to_cpu(vc->conf.timestamp); |
598f0d58 NB |
3777 | this->array.chunk_size = 512 << vc->conf.chunk_shift; |
3778 | ||
59e36268 | 3779 | i = vc->vcnum; |
7a7cc504 NB |
3780 | if ((ddf->virt->entries[i].state & DDF_state_inconsistent) || |
3781 | (ddf->virt->entries[i].init_state & DDF_initstate_mask) != | |
ed9d66aa | 3782 | DDF_init_full) { |
598f0d58 | 3783 | this->array.state = 0; |
ed9d66aa NB |
3784 | this->resync_start = 0; |
3785 | } else { | |
598f0d58 | 3786 | this->array.state = 1; |
b7528a20 | 3787 | this->resync_start = MaxSector; |
ed9d66aa | 3788 | } |
8bf989d8 | 3789 | _ddf_array_name(this->name, ddf, i); |
598f0d58 | 3790 | memset(this->uuid, 0, sizeof(this->uuid)); |
9d0c6b70 | 3791 | this->component_size = be64_to_cpu(vc->conf.blocks); |
598f0d58 | 3792 | this->array.size = this->component_size / 2; |
5f2aace8 | 3793 | this->container_member = i; |
598f0d58 | 3794 | |
c5afc314 N |
3795 | ddf->currentconf = vc; |
3796 | uuid_from_super_ddf(st, this->uuid); | |
f646805e | 3797 | if (!subarray) |
3798 | ddf->currentconf = NULL; | |
c5afc314 | 3799 | |
60f18132 | 3800 | sprintf(this->text_version, "/%s/%d", |
4dd2df09 | 3801 | st->container_devnm, this->container_member); |
60f18132 | 3802 | |
a8173e43 | 3803 | for (pd = 0; pd < be16_to_cpu(ddf->phys->used_pdes); pd++) { |
598f0d58 NB |
3804 | struct mdinfo *dev; |
3805 | struct dl *d; | |
4e587018 | 3806 | const struct vd_config *bvd; |
3807 | unsigned int iphys; | |
fa033bec | 3808 | int stt; |
598f0d58 | 3809 | |
60931cf9 | 3810 | if (be32_to_cpu(ddf->phys->entries[pd].refnum) |
3811 | == 0xFFFFFFFF) | |
bc17324f | 3812 | continue; |
0cf5ef67 | 3813 | |
a8173e43 | 3814 | stt = be16_to_cpu(ddf->phys->entries[pd].state); |
fa033bec N |
3815 | if ((stt & (DDF_Online|DDF_Failed|DDF_Rebuilding)) |
3816 | != DDF_Online) | |
3817 | continue; | |
3818 | ||
8a38db86 | 3819 | i = get_pd_index_from_refnum( |
4e587018 | 3820 | vc, ddf->phys->entries[pd].refnum, |
3821 | ddf->mppe, &bvd, &iphys); | |
d6e7b083 | 3822 | if (i == DDF_NOTFOUND) |
8a38db86 | 3823 | continue; |
3824 | ||
fa033bec | 3825 | this->array.working_disks++; |
bc17324f | 3826 | |
0cf5ef67 | 3827 | for (d = ddf->dlist; d ; d=d->next) |
60931cf9 | 3828 | if (be32_eq(d->disk.refnum, |
3829 | ddf->phys->entries[pd].refnum)) | |
0cf5ef67 N |
3830 | break; |
3831 | if (d == NULL) | |
3832 | /* Haven't found that one yet, maybe there are others */ | |
3833 | continue; | |
3834 | ||
503975b9 | 3835 | dev = xcalloc(1, sizeof(*dev)); |
598f0d58 NB |
3836 | dev->next = this->devs; |
3837 | this->devs = dev; | |
3838 | ||
60931cf9 | 3839 | dev->disk.number = be32_to_cpu(d->disk.refnum); |
598f0d58 NB |
3840 | dev->disk.major = d->major; |
3841 | dev->disk.minor = d->minor; | |
3842 | dev->disk.raid_disk = i; | |
3843 | dev->disk.state = (1<<MD_DISK_SYNC)|(1<<MD_DISK_ACTIVE); | |
d23534e4 | 3844 | dev->recovery_start = MaxSector; |
598f0d58 | 3845 | |
eba2859f | 3846 | dev->events = be32_to_cpu(ddf->active->seq); |
57a66662 | 3847 | dev->data_offset = |
9d0c6b70 | 3848 | be64_to_cpu(LBA_OFFSET(ddf, bvd)[iphys]); |
3849 | dev->component_size = be64_to_cpu(bvd->blocks); | |
598f0d58 NB |
3850 | if (d->devname) |
3851 | strcpy(dev->name, d->devname); | |
3852 | } | |
3853 | } | |
3854 | return rest; | |
3855 | } | |
3856 | ||
955e9ea1 | 3857 | static int store_super_ddf(struct supertype *st, int fd) |
a322f70c | 3858 | { |
955e9ea1 | 3859 | struct ddf_super *ddf = st->sb; |
a322f70c | 3860 | unsigned long long dsize; |
6416d527 | 3861 | void *buf; |
3d2c4fc7 | 3862 | int rc; |
a322f70c | 3863 | |
955e9ea1 DW |
3864 | if (!ddf) |
3865 | return 1; | |
3866 | ||
a322f70c DW |
3867 | if (!get_dev_size(fd, NULL, &dsize)) |
3868 | return 1; | |
3869 | ||
dbf98368 | 3870 | if (ddf->dlist || ddf->conflist) { |
3871 | struct stat sta; | |
3872 | struct dl *dl; | |
3873 | int ofd, ret; | |
3874 | ||
3875 | if (fstat(fd, &sta) == -1 || !S_ISBLK(sta.st_mode)) { | |
3876 | pr_err("%s: file descriptor for invalid device\n", | |
3877 | __func__); | |
3878 | return 1; | |
3879 | } | |
3880 | for (dl = ddf->dlist; dl; dl = dl->next) | |
3881 | if (dl->major == (int)major(sta.st_rdev) && | |
3882 | dl->minor == (int)minor(sta.st_rdev)) | |
3883 | break; | |
3884 | if (!dl) { | |
3885 | pr_err("%s: couldn't find disk %d/%d\n", __func__, | |
3886 | (int)major(sta.st_rdev), | |
3887 | (int)minor(sta.st_rdev)); | |
3888 | return 1; | |
3889 | } | |
dbf98368 | 3890 | ofd = dl->fd; |
3891 | dl->fd = fd; | |
3921e41a | 3892 | ret = (_write_super_to_disk(ddf, dl) != 1); |
dbf98368 | 3893 | dl->fd = ofd; |
3894 | return ret; | |
3895 | } | |
3896 | ||
3d2c4fc7 DW |
3897 | if (posix_memalign(&buf, 512, 512) != 0) |
3898 | return 1; | |
6416d527 NB |
3899 | memset(buf, 0, 512); |
3900 | ||
a322f70c | 3901 | lseek64(fd, dsize-512, 0); |
3d2c4fc7 | 3902 | rc = write(fd, buf, 512); |
6416d527 | 3903 | free(buf); |
3d2c4fc7 DW |
3904 | if (rc < 0) |
3905 | return 1; | |
a322f70c DW |
3906 | return 0; |
3907 | } | |
3908 | ||
a19c88b8 NB |
3909 | static int compare_super_ddf(struct supertype *st, struct supertype *tst) |
3910 | { | |
3911 | /* | |
3912 | * return: | |
3913 | * 0 same, or first was empty, and second was copied | |
3914 | * 1 second had wrong number | |
3915 | * 2 wrong uuid | |
3916 | * 3 wrong other info | |
3917 | */ | |
3918 | struct ddf_super *first = st->sb; | |
3919 | struct ddf_super *second = tst->sb; | |
4eefd651 | 3920 | struct dl *dl1, *dl2; |
3921 | struct vcl *vl1, *vl2; | |
2d210697 | 3922 | unsigned int max_vds, max_pds, pd, vd; |
a19c88b8 NB |
3923 | |
3924 | if (!first) { | |
3925 | st->sb = tst->sb; | |
3926 | tst->sb = NULL; | |
3927 | return 0; | |
3928 | } | |
3929 | ||
3930 | if (memcmp(first->anchor.guid, second->anchor.guid, DDF_GUID_LEN) != 0) | |
3931 | return 2; | |
3932 | ||
2d210697 | 3933 | if (first->max_part != second->max_part || |
a8173e43 | 3934 | !be16_eq(first->phys->used_pdes, second->phys->used_pdes) || |
3935 | !be16_eq(first->virt->populated_vdes, | |
3936 | second->virt->populated_vdes)) { | |
2d210697 | 3937 | dprintf("%s: PD/VD number mismatch\n", __func__); |
3938 | return 3; | |
3939 | } | |
3940 | ||
a8173e43 | 3941 | max_pds = be16_to_cpu(first->phys->used_pdes); |
2d210697 | 3942 | for (dl2 = second->dlist; dl2; dl2 = dl2->next) { |
3943 | for (pd = 0; pd < max_pds; pd++) | |
60931cf9 | 3944 | if (be32_eq(first->phys->entries[pd].refnum, |
3945 | dl2->disk.refnum)) | |
2d210697 | 3946 | break; |
3947 | if (pd == max_pds) { | |
3948 | dprintf("%s: no match for disk %08x\n", __func__, | |
60931cf9 | 3949 | be32_to_cpu(dl2->disk.refnum)); |
2d210697 | 3950 | return 3; |
3951 | } | |
3952 | } | |
3953 | ||
a8173e43 | 3954 | max_vds = be16_to_cpu(first->active->max_vd_entries); |
2d210697 | 3955 | for (vl2 = second->conflist; vl2; vl2 = vl2->next) { |
60931cf9 | 3956 | if (!be32_eq(vl2->conf.magic, DDF_VD_CONF_MAGIC)) |
2d210697 | 3957 | continue; |
3958 | for (vd = 0; vd < max_vds; vd++) | |
3959 | if (!memcmp(first->virt->entries[vd].guid, | |
3960 | vl2->conf.guid, DDF_GUID_LEN)) | |
3961 | break; | |
3962 | if (vd == max_vds) { | |
3963 | dprintf("%s: no match for VD config\n", __func__); | |
3964 | return 3; | |
3965 | } | |
3966 | } | |
a19c88b8 | 3967 | /* FIXME should I look at anything else? */ |
2d210697 | 3968 | |
4eefd651 | 3969 | /* |
3970 | At this point we are fairly sure that the meta data matches. | |
3971 | But the new disk may contain additional local data. | |
3972 | Add it to the super block. | |
3973 | */ | |
3974 | for (vl2 = second->conflist; vl2; vl2 = vl2->next) { | |
3975 | for (vl1 = first->conflist; vl1; vl1 = vl1->next) | |
3976 | if (!memcmp(vl1->conf.guid, vl2->conf.guid, | |
3977 | DDF_GUID_LEN)) | |
3978 | break; | |
3979 | if (vl1) { | |
3980 | if (vl1->other_bvds != NULL && | |
3981 | vl1->conf.sec_elmnt_seq != | |
3982 | vl2->conf.sec_elmnt_seq) { | |
3983 | dprintf("%s: adding BVD %u\n", __func__, | |
3984 | vl2->conf.sec_elmnt_seq); | |
3985 | add_other_bvd(vl1, &vl2->conf, | |
3986 | first->conf_rec_len*512); | |
3987 | } | |
3988 | continue; | |
3989 | } | |
3990 | ||
3991 | if (posix_memalign((void **)&vl1, 512, | |
3992 | (first->conf_rec_len*512 + | |
3993 | offsetof(struct vcl, conf))) != 0) { | |
3994 | pr_err("%s could not allocate vcl buf\n", | |
3995 | __func__); | |
3996 | return 3; | |
3997 | } | |
3998 | ||
3999 | vl1->next = first->conflist; | |
4000 | vl1->block_sizes = NULL; | |
4eefd651 | 4001 | memcpy(&vl1->conf, &vl2->conf, first->conf_rec_len*512); |
3c48f7be | 4002 | if (alloc_other_bvds(first, vl1) != 0) { |
4003 | pr_err("%s could not allocate other bvds\n", | |
4004 | __func__); | |
4005 | free(vl1); | |
4006 | return 3; | |
4007 | } | |
4eefd651 | 4008 | for (vd = 0; vd < max_vds; vd++) |
4009 | if (!memcmp(first->virt->entries[vd].guid, | |
4010 | vl1->conf.guid, DDF_GUID_LEN)) | |
4011 | break; | |
4012 | vl1->vcnum = vd; | |
4013 | dprintf("%s: added config for VD %u\n", __func__, vl1->vcnum); | |
4014 | first->conflist = vl1; | |
4015 | } | |
4016 | ||
4017 | for (dl2 = second->dlist; dl2; dl2 = dl2->next) { | |
4018 | for (dl1 = first->dlist; dl1; dl1 = dl1->next) | |
60931cf9 | 4019 | if (be32_eq(dl1->disk.refnum, dl2->disk.refnum)) |
4eefd651 | 4020 | break; |
4021 | if (dl1) | |
4022 | continue; | |
4023 | ||
4024 | if (posix_memalign((void **)&dl1, 512, | |
4025 | sizeof(*dl1) + (first->max_part) * sizeof(dl1->vlist[0])) | |
4026 | != 0) { | |
4027 | pr_err("%s could not allocate disk info buffer\n", | |
4028 | __func__); | |
4029 | return 3; | |
4030 | } | |
4031 | memcpy(dl1, dl2, sizeof(*dl1)); | |
4032 | dl1->mdupdate = NULL; | |
4033 | dl1->next = first->dlist; | |
4034 | dl1->fd = -1; | |
4035 | for (pd = 0; pd < max_pds; pd++) | |
60931cf9 | 4036 | if (be32_eq(first->phys->entries[pd].refnum, |
4037 | dl1->disk.refnum)) | |
4eefd651 | 4038 | break; |
4039 | dl1->pdnum = pd; | |
4040 | if (dl2->spare) { | |
4041 | if (posix_memalign((void **)&dl1->spare, 512, | |
4042 | first->conf_rec_len*512) != 0) { | |
4043 | pr_err("%s could not allocate spare info buf\n", | |
4044 | __func__); | |
4045 | return 3; | |
4046 | } | |
4047 | memcpy(dl1->spare, dl2->spare, first->conf_rec_len*512); | |
4048 | } | |
4049 | for (vd = 0 ; vd < first->max_part ; vd++) { | |
4050 | if (!dl2->vlist[vd]) { | |
4051 | dl1->vlist[vd] = NULL; | |
4052 | continue; | |
4053 | } | |
4054 | for (vl1 = first->conflist; vl1; vl1 = vl1->next) { | |
4055 | if (!memcmp(vl1->conf.guid, | |
4056 | dl2->vlist[vd]->conf.guid, | |
4057 | DDF_GUID_LEN)) | |
4058 | break; | |
4059 | dl1->vlist[vd] = vl1; | |
4060 | } | |
4061 | } | |
4062 | first->dlist = dl1; | |
4063 | dprintf("%s: added disk %d: %08x\n", __func__, dl1->pdnum, | |
60931cf9 | 4064 | be32_to_cpu(dl1->disk.refnum)); |
4eefd651 | 4065 | } |
4066 | ||
a19c88b8 NB |
4067 | return 0; |
4068 | } | |
4069 | ||
0e600426 | 4070 | #ifndef MDASSEMBLE |
4e5528c6 NB |
4071 | /* |
4072 | * A new array 'a' has been started which claims to be instance 'inst' | |
4073 | * within container 'c'. | |
4074 | * We need to confirm that the array matches the metadata in 'c' so | |
4075 | * that we don't corrupt any metadata. | |
4076 | */ | |
cba0191b | 4077 | static int ddf_open_new(struct supertype *c, struct active_array *a, char *inst) |
549e9569 | 4078 | { |
a2aa439e | 4079 | struct ddf_super *ddf = c->sb; |
4080 | int n = atoi(inst); | |
5daa35ac | 4081 | struct mdinfo *dev; |
4082 | struct dl *dl; | |
4083 | static const char faulty[] = "faulty"; | |
4084 | ||
fb9d0acb | 4085 | if (all_ff(ddf->virt->entries[n].guid)) { |
4086 | pr_err("%s: subarray %d doesn't exist\n", __func__, n); | |
a2aa439e | 4087 | return -ENODEV; |
4088 | } | |
5daa35ac | 4089 | dprintf("%s: new subarray %d, GUID: %s\n", __func__, n, |
4090 | guid_str(ddf->virt->entries[n].guid)); | |
4091 | for (dev = a->info.devs; dev; dev = dev->next) { | |
4092 | for (dl = ddf->dlist; dl; dl = dl->next) | |
4093 | if (dl->major == dev->disk.major && | |
4094 | dl->minor == dev->disk.minor) | |
4095 | break; | |
4096 | if (!dl) { | |
4097 | pr_err("%s: device %d/%d of subarray %d not found in meta data\n", | |
4098 | __func__, dev->disk.major, dev->disk.minor, n); | |
4099 | return -1; | |
4100 | } | |
4101 | if ((be16_to_cpu(ddf->phys->entries[dl->pdnum].state) & | |
4102 | (DDF_Online|DDF_Missing|DDF_Failed)) != DDF_Online) { | |
4103 | pr_err("%s: new subarray %d contains broken device %d/%d (%02x)\n", | |
4104 | __func__, n, dl->major, dl->minor, | |
4105 | be16_to_cpu( | |
4106 | ddf->phys->entries[dl->pdnum].state)); | |
4107 | if (write(dev->state_fd, faulty, sizeof(faulty)-1) != | |
4108 | sizeof(faulty) - 1) | |
4109 | pr_err("Write to state_fd failed\n"); | |
4110 | dev->curr_state = DS_FAULTY; | |
4111 | } | |
4112 | } | |
a2aa439e | 4113 | a->info.container_member = n; |
549e9569 NB |
4114 | return 0; |
4115 | } | |
4116 | ||
4e5528c6 NB |
4117 | /* |
4118 | * The array 'a' is to be marked clean in the metadata. | |
ed9d66aa | 4119 | * If '->resync_start' is not ~(unsigned long long)0, then the array is only |
4e5528c6 NB |
4120 | * clean up to the point (in sectors). If that cannot be recorded in the |
4121 | * metadata, then leave it as dirty. | |
4122 | * | |
4123 | * For DDF, we need to clear the DDF_state_inconsistent bit in the | |
4124 | * !global! virtual_disk.virtual_entry structure. | |
4125 | */ | |
01f157d7 | 4126 | static int ddf_set_array_state(struct active_array *a, int consistent) |
549e9569 | 4127 | { |
4e5528c6 NB |
4128 | struct ddf_super *ddf = a->container->sb; |
4129 | int inst = a->info.container_member; | |
18a2f463 | 4130 | int old = ddf->virt->entries[inst].state; |
01f157d7 N |
4131 | if (consistent == 2) { |
4132 | /* Should check if a recovery should be started FIXME */ | |
4133 | consistent = 1; | |
b7941fd6 | 4134 | if (!is_resync_complete(&a->info)) |
01f157d7 N |
4135 | consistent = 0; |
4136 | } | |
ed9d66aa NB |
4137 | if (consistent) |
4138 | ddf->virt->entries[inst].state &= ~DDF_state_inconsistent; | |
4139 | else | |
4e5528c6 | 4140 | ddf->virt->entries[inst].state |= DDF_state_inconsistent; |
18a2f463 | 4141 | if (old != ddf->virt->entries[inst].state) |
7d5a7ff3 | 4142 | ddf_set_updates_pending(ddf); |
18a2f463 NB |
4143 | |
4144 | old = ddf->virt->entries[inst].init_state; | |
ed9d66aa | 4145 | ddf->virt->entries[inst].init_state &= ~DDF_initstate_mask; |
b7941fd6 | 4146 | if (is_resync_complete(&a->info)) |
ed9d66aa | 4147 | ddf->virt->entries[inst].init_state |= DDF_init_full; |
b7941fd6 | 4148 | else if (a->info.resync_start == 0) |
ed9d66aa | 4149 | ddf->virt->entries[inst].init_state |= DDF_init_not; |
4e5528c6 | 4150 | else |
ed9d66aa | 4151 | ddf->virt->entries[inst].init_state |= DDF_init_quick; |
18a2f463 | 4152 | if (old != ddf->virt->entries[inst].init_state) |
7d5a7ff3 | 4153 | ddf_set_updates_pending(ddf); |
ed9d66aa | 4154 | |
b27336a2 | 4155 | dprintf("ddf mark %d/%s (%d) %s %llu\n", inst, |
4156 | guid_str(ddf->virt->entries[inst].guid), a->curr_state, | |
4157 | consistent?"clean":"dirty", | |
b7941fd6 | 4158 | a->info.resync_start); |
01f157d7 | 4159 | return consistent; |
fd7cde1b DW |
4160 | } |
4161 | ||
5ec636b7 | 4162 | static int get_bvd_state(const struct ddf_super *ddf, |
4163 | const struct vd_config *vc) | |
4164 | { | |
4165 | unsigned int i, n_bvd, working = 0; | |
a8173e43 | 4166 | unsigned int n_prim = be16_to_cpu(vc->prim_elmnt_count); |
5ec636b7 | 4167 | int pd, st, state; |
4168 | for (i = 0; i < n_prim; i++) { | |
4169 | if (!find_index_in_bvd(ddf, vc, i, &n_bvd)) | |
4170 | continue; | |
4171 | pd = find_phys(ddf, vc->phys_refnum[n_bvd]); | |
4172 | if (pd < 0) | |
4173 | continue; | |
a8173e43 | 4174 | st = be16_to_cpu(ddf->phys->entries[pd].state); |
5ec636b7 | 4175 | if ((st & (DDF_Online|DDF_Failed|DDF_Rebuilding)) |
4176 | == DDF_Online) | |
4177 | working++; | |
4178 | } | |
4179 | ||
4180 | state = DDF_state_degraded; | |
4181 | if (working == n_prim) | |
4182 | state = DDF_state_optimal; | |
4183 | else | |
4184 | switch (vc->prl) { | |
4185 | case DDF_RAID0: | |
4186 | case DDF_CONCAT: | |
4187 | case DDF_JBOD: | |
4188 | state = DDF_state_failed; | |
4189 | break; | |
4190 | case DDF_RAID1: | |
4191 | if (working == 0) | |
4192 | state = DDF_state_failed; | |
4193 | else if (working >= 2) | |
4194 | state = DDF_state_part_optimal; | |
4195 | break; | |
4196 | case DDF_RAID4: | |
4197 | case DDF_RAID5: | |
4198 | if (working < n_prim - 1) | |
4199 | state = DDF_state_failed; | |
4200 | break; | |
4201 | case DDF_RAID6: | |
4202 | if (working < n_prim - 2) | |
4203 | state = DDF_state_failed; | |
4204 | else if (working == n_prim - 1) | |
4205 | state = DDF_state_part_optimal; | |
4206 | break; | |
4207 | } | |
4208 | return state; | |
4209 | } | |
4210 | ||
0777d17d | 4211 | static int secondary_state(int state, int other, int seclevel) |
4212 | { | |
4213 | if (state == DDF_state_optimal && other == DDF_state_optimal) | |
4214 | return DDF_state_optimal; | |
4215 | if (seclevel == DDF_2MIRRORED) { | |
4216 | if (state == DDF_state_optimal || other == DDF_state_optimal) | |
4217 | return DDF_state_part_optimal; | |
4218 | if (state == DDF_state_failed && other == DDF_state_failed) | |
4219 | return DDF_state_failed; | |
4220 | return DDF_state_degraded; | |
4221 | } else { | |
4222 | if (state == DDF_state_failed || other == DDF_state_failed) | |
4223 | return DDF_state_failed; | |
4224 | if (state == DDF_state_degraded || other == DDF_state_degraded) | |
4225 | return DDF_state_degraded; | |
4226 | return DDF_state_part_optimal; | |
4227 | } | |
4228 | } | |
4229 | ||
4230 | static int get_svd_state(const struct ddf_super *ddf, const struct vcl *vcl) | |
4231 | { | |
4232 | int state = get_bvd_state(ddf, &vcl->conf); | |
4233 | unsigned int i; | |
4234 | for (i = 1; i < vcl->conf.sec_elmnt_count; i++) { | |
4235 | state = secondary_state( | |
4236 | state, | |
4237 | get_bvd_state(ddf, vcl->other_bvds[i-1]), | |
4238 | vcl->conf.srl); | |
4239 | } | |
4240 | return state; | |
4241 | } | |
4242 | ||
7a7cc504 NB |
4243 | /* |
4244 | * The state of each disk is stored in the global phys_disk structure | |
4245 | * in phys_disk.entries[n].state. | |
4246 | * This makes various combinations awkward. | |
4247 | * - When a device fails in any array, it must be failed in all arrays | |
4248 | * that include a part of this device. | |
4249 | * - When a component is rebuilding, we cannot include it officially in the | |
4250 | * array unless this is the only array that uses the device. | |
4251 | * | |
4252 | * So: when transitioning: | |
4253 | * Online -> failed, just set failed flag. monitor will propagate | |
4254 | * spare -> online, the device might need to be added to the array. | |
4255 | * spare -> failed, just set failed. Don't worry if in array or not. | |
4256 | */ | |
8d45d196 | 4257 | static void ddf_set_disk(struct active_array *a, int n, int state) |
549e9569 | 4258 | { |
7a7cc504 | 4259 | struct ddf_super *ddf = a->container->sb; |
baba3f4e | 4260 | unsigned int inst = a->info.container_member, n_bvd; |
4261 | struct vcl *vcl; | |
4262 | struct vd_config *vc = find_vdcr(ddf, inst, (unsigned int)n, | |
4263 | &n_bvd, &vcl); | |
4264 | int pd; | |
e1316fab N |
4265 | struct mdinfo *mdi; |
4266 | struct dl *dl; | |
7a7cc504 | 4267 | |
ce6844b9 | 4268 | dprintf("%s: %d to %x\n", __func__, n, state); |
7a7cc504 | 4269 | if (vc == NULL) { |
2c514b71 | 4270 | dprintf("ddf: cannot find instance %d!!\n", inst); |
7a7cc504 NB |
4271 | return; |
4272 | } | |
e1316fab N |
4273 | /* Find the matching slot in 'info'. */ |
4274 | for (mdi = a->info.devs; mdi; mdi = mdi->next) | |
4275 | if (mdi->disk.raid_disk == n) | |
4276 | break; | |
ce6844b9 MW |
4277 | if (!mdi) { |
4278 | pr_err("%s: cannot find raid disk %d\n", | |
4279 | __func__, n); | |
e1316fab | 4280 | return; |
ce6844b9 | 4281 | } |
e1316fab N |
4282 | |
4283 | /* and find the 'dl' entry corresponding to that. */ | |
4284 | for (dl = ddf->dlist; dl; dl = dl->next) | |
77632af9 N |
4285 | if (mdi->state_fd >= 0 && |
4286 | mdi->disk.major == dl->major && | |
e1316fab N |
4287 | mdi->disk.minor == dl->minor) |
4288 | break; | |
ce6844b9 MW |
4289 | if (!dl) { |
4290 | pr_err("%s: cannot find raid disk %d (%d/%d)\n", | |
4291 | __func__, n, | |
4292 | mdi->disk.major, mdi->disk.minor); | |
e1316fab | 4293 | return; |
ce6844b9 | 4294 | } |
e1316fab | 4295 | |
baba3f4e | 4296 | pd = find_phys(ddf, vc->phys_refnum[n_bvd]); |
e1316fab N |
4297 | if (pd < 0 || pd != dl->pdnum) { |
4298 | /* disk doesn't currently exist or has changed. | |
4299 | * If it is now in_sync, insert it. */ | |
baba3f4e | 4300 | dprintf("%s: phys disk not found for %d: %d/%d ref %08x\n", |
4301 | __func__, dl->pdnum, dl->major, dl->minor, | |
60931cf9 | 4302 | be32_to_cpu(dl->disk.refnum)); |
baba3f4e | 4303 | dprintf("%s: array %u disk %u ref %08x pd %d\n", |
60931cf9 | 4304 | __func__, inst, n_bvd, |
4305 | be32_to_cpu(vc->phys_refnum[n_bvd]), pd); | |
7a7cc504 | 4306 | if ((state & DS_INSYNC) && ! (state & DS_FAULTY)) { |
baba3f4e | 4307 | pd = dl->pdnum; /* FIXME: is this really correct ? */ |
4308 | vc->phys_refnum[n_bvd] = dl->disk.refnum; | |
57a66662 | 4309 | LBA_OFFSET(ddf, vc)[n_bvd] = |
9d0c6b70 | 4310 | cpu_to_be64(mdi->data_offset); |
a8173e43 | 4311 | be16_clear(ddf->phys->entries[pd].type, |
4312 | cpu_to_be16(DDF_Global_Spare)); | |
4313 | be16_set(ddf->phys->entries[pd].type, | |
4314 | cpu_to_be16(DDF_Active_in_VD)); | |
7d5a7ff3 | 4315 | ddf_set_updates_pending(ddf); |
7a7cc504 NB |
4316 | } |
4317 | } else { | |
a8173e43 | 4318 | be16 old = ddf->phys->entries[pd].state; |
7a7cc504 | 4319 | if (state & DS_FAULTY) |
a8173e43 | 4320 | be16_set(ddf->phys->entries[pd].state, |
4321 | cpu_to_be16(DDF_Failed)); | |
7a7cc504 | 4322 | if (state & DS_INSYNC) { |
a8173e43 | 4323 | be16_set(ddf->phys->entries[pd].state, |
4324 | cpu_to_be16(DDF_Online)); | |
4325 | be16_clear(ddf->phys->entries[pd].state, | |
4326 | cpu_to_be16(DDF_Rebuilding)); | |
7a7cc504 | 4327 | } |
a8173e43 | 4328 | if (!be16_eq(old, ddf->phys->entries[pd].state)) |
7d5a7ff3 | 4329 | ddf_set_updates_pending(ddf); |
7a7cc504 NB |
4330 | } |
4331 | ||
ce6844b9 MW |
4332 | dprintf("ddf: set_disk %d (%08x) to %x->%02x\n", n, |
4333 | be32_to_cpu(dl->disk.refnum), state, | |
4334 | be16_to_cpu(ddf->phys->entries[pd].state)); | |
7e1432fb | 4335 | |
7a7cc504 NB |
4336 | /* Now we need to check the state of the array and update |
4337 | * virtual_disk.entries[n].state. | |
4338 | * It needs to be one of "optimal", "degraded", "failed". | |
4339 | * I don't understand 'deleted' or 'missing'. | |
4340 | */ | |
0777d17d | 4341 | state = get_svd_state(ddf, vcl); |
7a7cc504 | 4342 | |
18a2f463 NB |
4343 | if (ddf->virt->entries[inst].state != |
4344 | ((ddf->virt->entries[inst].state & ~DDF_state_mask) | |
4345 | | state)) { | |
4346 | ||
4347 | ddf->virt->entries[inst].state = | |
4348 | (ddf->virt->entries[inst].state & ~DDF_state_mask) | |
4349 | | state; | |
7d5a7ff3 | 4350 | ddf_set_updates_pending(ddf); |
18a2f463 | 4351 | } |
7a7cc504 | 4352 | |
549e9569 NB |
4353 | } |
4354 | ||
2e735d19 | 4355 | static void ddf_sync_metadata(struct supertype *st) |
549e9569 | 4356 | { |
7a7cc504 NB |
4357 | |
4358 | /* | |
4359 | * Write all data to all devices. | |
4360 | * Later, we might be able to track whether only local changes | |
4361 | * have been made, or whether any global data has been changed, | |
4362 | * but ddf is sufficiently weird that it probably always | |
4363 | * changes global data .... | |
4364 | */ | |
18a2f463 NB |
4365 | struct ddf_super *ddf = st->sb; |
4366 | if (!ddf->updates_pending) | |
4367 | return; | |
4368 | ddf->updates_pending = 0; | |
3921e41a | 4369 | __write_init_super_ddf(st); |
2c514b71 | 4370 | dprintf("ddf: sync_metadata\n"); |
549e9569 NB |
4371 | } |
4372 | ||
f646805e | 4373 | static int del_from_conflist(struct vcl **list, const char *guid) |
4374 | { | |
4375 | struct vcl **p; | |
4376 | int found = 0; | |
4377 | for (p = list; p && *p; p = &((*p)->next)) | |
4378 | if (!memcmp((*p)->conf.guid, guid, DDF_GUID_LEN)) { | |
4379 | found = 1; | |
4380 | *p = (*p)->next; | |
4381 | } | |
4382 | return found; | |
4383 | } | |
4384 | ||
4385 | static int _kill_subarray_ddf(struct ddf_super *ddf, const char *guid) | |
4386 | { | |
4387 | struct dl *dl; | |
4388 | unsigned int vdnum, i; | |
4389 | vdnum = find_vde_by_guid(ddf, guid); | |
4390 | if (vdnum == DDF_NOTFOUND) { | |
4391 | pr_err("%s: could not find VD %s\n", __func__, | |
4392 | guid_str(guid)); | |
4393 | return -1; | |
4394 | } | |
4395 | if (del_from_conflist(&ddf->conflist, guid) == 0) { | |
4396 | pr_err("%s: could not find conf %s\n", __func__, | |
4397 | guid_str(guid)); | |
4398 | return -1; | |
4399 | } | |
4400 | for (dl = ddf->dlist; dl; dl = dl->next) | |
4401 | for (i = 0; i < ddf->max_part; i++) | |
4402 | if (dl->vlist[i] != NULL && | |
4403 | !memcmp(dl->vlist[i]->conf.guid, guid, | |
4404 | DDF_GUID_LEN)) | |
4405 | dl->vlist[i] = NULL; | |
4406 | memset(ddf->virt->entries[vdnum].guid, 0xff, DDF_GUID_LEN); | |
4407 | dprintf("%s: deleted %s\n", __func__, guid_str(guid)); | |
4408 | return 0; | |
4409 | } | |
4410 | ||
4411 | static int kill_subarray_ddf(struct supertype *st) | |
4412 | { | |
4413 | struct ddf_super *ddf = st->sb; | |
4414 | /* | |
4415 | * currentconf is set in container_content_ddf, | |
4416 | * called with subarray arg | |
4417 | */ | |
4418 | struct vcl *victim = ddf->currentconf; | |
4419 | struct vd_config *conf; | |
4420 | ddf->currentconf = NULL; | |
4421 | unsigned int vdnum; | |
4422 | if (!victim) { | |
4423 | pr_err("%s: nothing to kill\n", __func__); | |
4424 | return -1; | |
4425 | } | |
4426 | conf = &victim->conf; | |
4427 | vdnum = find_vde_by_guid(ddf, conf->guid); | |
4428 | if (vdnum == DDF_NOTFOUND) { | |
4429 | pr_err("%s: could not find VD %s\n", __func__, | |
4430 | guid_str(conf->guid)); | |
4431 | return -1; | |
4432 | } | |
4433 | if (st->update_tail) { | |
4434 | struct virtual_disk *vd; | |
4435 | int len = sizeof(struct virtual_disk) | |
4436 | + sizeof(struct virtual_entry); | |
4437 | vd = xmalloc(len); | |
4438 | if (vd == NULL) { | |
4439 | pr_err("%s: failed to allocate %d bytes\n", __func__, | |
4440 | len); | |
4441 | return -1; | |
4442 | } | |
4443 | memset(vd, 0 , len); | |
4444 | vd->magic = DDF_VIRT_RECORDS_MAGIC; | |
a8173e43 | 4445 | vd->populated_vdes = cpu_to_be16(0); |
f646805e | 4446 | memcpy(vd->entries[0].guid, conf->guid, DDF_GUID_LEN); |
4447 | /* we use DDF_state_deleted as marker */ | |
4448 | vd->entries[0].state = DDF_state_deleted; | |
4449 | append_metadata_update(st, vd, len); | |
6a350d82 | 4450 | } else { |
f646805e | 4451 | _kill_subarray_ddf(ddf, conf->guid); |
6a350d82 | 4452 | ddf_set_updates_pending(ddf); |
4453 | ddf_sync_metadata(st); | |
4454 | } | |
f646805e | 4455 | return 0; |
4456 | } | |
4457 | ||
c5943560 | 4458 | static void copy_matching_bvd(struct ddf_super *ddf, |
4459 | struct vd_config *conf, | |
4460 | const struct metadata_update *update) | |
4461 | { | |
4462 | unsigned int mppe = | |
a8173e43 | 4463 | be16_to_cpu(ddf->anchor.max_primary_element_entries); |
c5943560 | 4464 | unsigned int len = ddf->conf_rec_len * 512; |
4465 | char *p; | |
4466 | struct vd_config *vc; | |
4467 | for (p = update->buf; p < update->buf + update->len; p += len) { | |
4468 | vc = (struct vd_config *) p; | |
4469 | if (vc->sec_elmnt_seq == conf->sec_elmnt_seq) { | |
4470 | memcpy(conf->phys_refnum, vc->phys_refnum, | |
4471 | mppe * (sizeof(__u32) + sizeof(__u64))); | |
4472 | return; | |
4473 | } | |
4474 | } | |
4475 | pr_err("%s: no match for BVD %d of %s in update\n", __func__, | |
4476 | conf->sec_elmnt_seq, guid_str(conf->guid)); | |
4477 | } | |
4478 | ||
88c164f4 NB |
4479 | static void ddf_process_update(struct supertype *st, |
4480 | struct metadata_update *update) | |
4481 | { | |
4482 | /* Apply this update to the metadata. | |
4483 | * The first 4 bytes are a DDF_*_MAGIC which guides | |
4484 | * our actions. | |
4485 | * Possible update are: | |
4486 | * DDF_PHYS_RECORDS_MAGIC | |
4dd968cc N |
4487 | * Add a new physical device or remove an old one. |
4488 | * Changes to this record only happen implicitly. | |
88c164f4 NB |
4489 | * used_pdes is the device number. |
4490 | * DDF_VIRT_RECORDS_MAGIC | |
4491 | * Add a new VD. Possibly also change the 'access' bits. | |
4492 | * populated_vdes is the entry number. | |
4493 | * DDF_VD_CONF_MAGIC | |
4494 | * New or updated VD. the VIRT_RECORD must already | |
4495 | * exist. For an update, phys_refnum and lba_offset | |
4496 | * (at least) are updated, and the VD_CONF must | |
4497 | * be written to precisely those devices listed with | |
4498 | * a phys_refnum. | |
4499 | * DDF_SPARE_ASSIGN_MAGIC | |
4500 | * replacement Spare Assignment Record... but for which device? | |
4501 | * | |
4502 | * So, e.g.: | |
4503 | * - to create a new array, we send a VIRT_RECORD and | |
4504 | * a VD_CONF. Then assemble and start the array. | |
4505 | * - to activate a spare we send a VD_CONF to add the phys_refnum | |
4506 | * and offset. This will also mark the spare as active with | |
4507 | * a spare-assignment record. | |
4508 | */ | |
4509 | struct ddf_super *ddf = st->sb; | |
60931cf9 | 4510 | be32 *magic = (be32 *)update->buf; |
88c164f4 NB |
4511 | struct phys_disk *pd; |
4512 | struct virtual_disk *vd; | |
4513 | struct vd_config *vc; | |
4514 | struct vcl *vcl; | |
4515 | struct dl *dl; | |
f21e18ca | 4516 | unsigned int ent; |
c5943560 | 4517 | unsigned int pdnum, pd2, len; |
88c164f4 | 4518 | |
60931cf9 | 4519 | dprintf("Process update %x\n", be32_to_cpu(*magic)); |
7e1432fb | 4520 | |
60931cf9 | 4521 | if (be32_eq(*magic, DDF_PHYS_RECORDS_MAGIC)) { |
88c164f4 NB |
4522 | |
4523 | if (update->len != (sizeof(struct phys_disk) + | |
4524 | sizeof(struct phys_disk_entry))) | |
4525 | return; | |
4526 | pd = (struct phys_disk*)update->buf; | |
4527 | ||
a8173e43 | 4528 | ent = be16_to_cpu(pd->used_pdes); |
4529 | if (ent >= be16_to_cpu(ddf->phys->max_pdes)) | |
88c164f4 | 4530 | return; |
a8173e43 | 4531 | if (be16_and(pd->entries[0].state, cpu_to_be16(DDF_Missing))) { |
4dd968cc N |
4532 | struct dl **dlp; |
4533 | /* removing this disk. */ | |
a8173e43 | 4534 | be16_set(ddf->phys->entries[ent].state, |
4535 | cpu_to_be16(DDF_Missing)); | |
4dd968cc N |
4536 | for (dlp = &ddf->dlist; *dlp; dlp = &(*dlp)->next) { |
4537 | struct dl *dl = *dlp; | |
4538 | if (dl->pdnum == (signed)ent) { | |
4539 | close(dl->fd); | |
4540 | dl->fd = -1; | |
4541 | /* FIXME this doesn't free | |
4542 | * dl->devname */ | |
4543 | update->space = dl; | |
4544 | *dlp = dl->next; | |
4545 | break; | |
4546 | } | |
4547 | } | |
7d5a7ff3 | 4548 | ddf_set_updates_pending(ddf); |
4dd968cc N |
4549 | return; |
4550 | } | |
88c164f4 NB |
4551 | if (!all_ff(ddf->phys->entries[ent].guid)) |
4552 | return; | |
4553 | ddf->phys->entries[ent] = pd->entries[0]; | |
a8173e43 | 4554 | ddf->phys->used_pdes = cpu_to_be16 |
4555 | (1 + be16_to_cpu(ddf->phys->used_pdes)); | |
7d5a7ff3 | 4556 | ddf_set_updates_pending(ddf); |
2cc2983d N |
4557 | if (ddf->add_list) { |
4558 | struct active_array *a; | |
4559 | struct dl *al = ddf->add_list; | |
4560 | ddf->add_list = al->next; | |
4561 | ||
4562 | al->next = ddf->dlist; | |
4563 | ddf->dlist = al; | |
4564 | ||
4565 | /* As a device has been added, we should check | |
4566 | * for any degraded devices that might make | |
4567 | * use of this spare */ | |
4568 | for (a = st->arrays ; a; a=a->next) | |
4569 | a->check_degraded = 1; | |
4570 | } | |
60931cf9 | 4571 | } else if (be32_eq(*magic, DDF_VIRT_RECORDS_MAGIC)) { |
88c164f4 NB |
4572 | |
4573 | if (update->len != (sizeof(struct virtual_disk) + | |
4574 | sizeof(struct virtual_entry))) | |
4575 | return; | |
4576 | vd = (struct virtual_disk*)update->buf; | |
4577 | ||
f646805e | 4578 | if (vd->entries[0].state == DDF_state_deleted) { |
4579 | if (_kill_subarray_ddf(ddf, vd->entries[0].guid)) | |
4580 | return; | |
4581 | } else { | |
4582 | ||
6a7e7ecc | 4583 | ent = find_vde_by_guid(ddf, vd->entries[0].guid); |
4584 | if (ent != DDF_NOTFOUND) { | |
4585 | dprintf("%s: VD %s exists already in slot %d\n", | |
4586 | __func__, guid_str(vd->entries[0].guid), | |
4587 | ent); | |
4588 | return; | |
4589 | } | |
f646805e | 4590 | ent = find_unused_vde(ddf); |
4591 | if (ent == DDF_NOTFOUND) | |
4592 | return; | |
4593 | ddf->virt->entries[ent] = vd->entries[0]; | |
4594 | ddf->virt->populated_vdes = | |
a8173e43 | 4595 | cpu_to_be16( |
4596 | 1 + be16_to_cpu( | |
f646805e | 4597 | ddf->virt->populated_vdes)); |
ed5ff7a2 | 4598 | dprintf("%s: added VD %s in slot %d(s=%02x i=%02x)\n", |
4599 | __func__, guid_str(vd->entries[0].guid), ent, | |
4600 | ddf->virt->entries[ent].state, | |
4601 | ddf->virt->entries[ent].init_state); | |
f646805e | 4602 | } |
7d5a7ff3 | 4603 | ddf_set_updates_pending(ddf); |
60931cf9 | 4604 | } |
88c164f4 | 4605 | |
60931cf9 | 4606 | else if (be32_eq(*magic, DDF_VD_CONF_MAGIC)) { |
88c164f4 | 4607 | vc = (struct vd_config*)update->buf; |
c5943560 | 4608 | len = ddf->conf_rec_len * 512; |
4609 | if ((unsigned int)update->len != len * vc->sec_elmnt_count) { | |
4610 | pr_err("%s: %s: insufficient data (%d) for %u BVDs\n", | |
4611 | __func__, guid_str(vc->guid), update->len, | |
4612 | vc->sec_elmnt_count); | |
4613 | return; | |
4614 | } | |
88c164f4 NB |
4615 | for (vcl = ddf->conflist; vcl ; vcl = vcl->next) |
4616 | if (memcmp(vcl->conf.guid, vc->guid, DDF_GUID_LEN) == 0) | |
4617 | break; | |
ed5ff7a2 | 4618 | dprintf("%s: conf update for %s (%s)\n", __func__, |
4619 | guid_str(vc->guid), (vcl ? "old" : "new")); | |
88c164f4 NB |
4620 | if (vcl) { |
4621 | /* An update, just copy the phys_refnum and lba_offset | |
4622 | * fields | |
4623 | */ | |
c5943560 | 4624 | unsigned int i; |
0847945b | 4625 | unsigned int k; |
c5943560 | 4626 | copy_matching_bvd(ddf, &vcl->conf, update); |
0847945b MW |
4627 | for (k = 0; k < be16_to_cpu(vc->prim_elmnt_count); k++) |
4628 | dprintf("BVD %u has %08x at %llu\n", 0, | |
4629 | be32_to_cpu(vcl->conf.phys_refnum[k]), | |
4630 | be64_to_cpu(LBA_OFFSET(ddf, | |
4631 | &vcl->conf)[k])); | |
4632 | for (i = 1; i < vc->sec_elmnt_count; i++) { | |
c5943560 | 4633 | copy_matching_bvd(ddf, vcl->other_bvds[i-1], |
4634 | update); | |
0847945b MW |
4635 | for (k = 0; k < be16_to_cpu( |
4636 | vc->prim_elmnt_count); k++) | |
4637 | dprintf("BVD %u has %08x at %llu\n", i, | |
4638 | be32_to_cpu | |
4639 | (vcl->other_bvds[i-1]-> | |
4640 | phys_refnum[k]), | |
4641 | be64_to_cpu | |
4642 | (LBA_OFFSET | |
4643 | (ddf, | |
4644 | vcl->other_bvds[i-1])[k])); | |
4645 | } | |
88c164f4 NB |
4646 | } else { |
4647 | /* A new VD_CONF */ | |
c5943560 | 4648 | unsigned int i; |
e6b9548d DW |
4649 | if (!update->space) |
4650 | return; | |
88c164f4 NB |
4651 | vcl = update->space; |
4652 | update->space = NULL; | |
4653 | vcl->next = ddf->conflist; | |
c5943560 | 4654 | memcpy(&vcl->conf, vc, len); |
fb9d0acb | 4655 | ent = find_vde_by_guid(ddf, vc->guid); |
4656 | if (ent == DDF_NOTFOUND) | |
4657 | return; | |
4658 | vcl->vcnum = ent; | |
88c164f4 | 4659 | ddf->conflist = vcl; |
c5943560 | 4660 | for (i = 1; i < vc->sec_elmnt_count; i++) |
4661 | memcpy(vcl->other_bvds[i-1], | |
4662 | update->buf + len * i, len); | |
88c164f4 | 4663 | } |
c7079c84 N |
4664 | /* Set DDF_Transition on all Failed devices - to help |
4665 | * us detect those that are no longer in use | |
4666 | */ | |
a8173e43 | 4667 | for (pdnum = 0; pdnum < be16_to_cpu(ddf->phys->used_pdes); |
4668 | pdnum++) | |
4669 | if (be16_and(ddf->phys->entries[pdnum].state, | |
4670 | cpu_to_be16(DDF_Failed))) | |
4671 | be16_set(ddf->phys->entries[pdnum].state, | |
4672 | cpu_to_be16(DDF_Transition)); | |
88c164f4 NB |
4673 | /* Now make sure vlist is correct for each dl. */ |
4674 | for (dl = ddf->dlist; dl; dl = dl->next) { | |
f21e18ca | 4675 | unsigned int vn = 0; |
8401644c | 4676 | int in_degraded = 0; |
5838fccd | 4677 | for (vcl = ddf->conflist; vcl ; vcl = vcl->next) { |
4678 | unsigned int dn, ibvd; | |
4679 | const struct vd_config *conf; | |
4680 | int vstate; | |
4681 | dn = get_pd_index_from_refnum(vcl, | |
4682 | dl->disk.refnum, | |
4683 | ddf->mppe, | |
4684 | &conf, &ibvd); | |
4685 | if (dn == DDF_NOTFOUND) | |
4686 | continue; | |
4687 | dprintf("dev %d/%08x has %s (sec=%u) at %d\n", | |
ad60eea1 | 4688 | dl->pdnum, |
60931cf9 | 4689 | be32_to_cpu(dl->disk.refnum), |
5838fccd | 4690 | guid_str(conf->guid), |
4691 | conf->sec_elmnt_seq, vn); | |
4692 | /* Clear the Transition flag */ | |
a8173e43 | 4693 | if (be16_and |
4694 | (ddf->phys->entries[dl->pdnum].state, | |
4695 | cpu_to_be16(DDF_Failed))) | |
4696 | be16_clear(ddf->phys | |
4697 | ->entries[dl->pdnum].state, | |
4698 | cpu_to_be16(DDF_Transition)); | |
5838fccd | 4699 | dl->vlist[vn++] = vcl; |
4700 | vstate = ddf->virt->entries[vcl->vcnum].state | |
4701 | & DDF_state_mask; | |
4702 | if (vstate == DDF_state_degraded || | |
4703 | vstate == DDF_state_part_optimal) | |
4704 | in_degraded = 1; | |
4705 | } | |
88c164f4 NB |
4706 | while (vn < ddf->max_part) |
4707 | dl->vlist[vn++] = NULL; | |
7e1432fb | 4708 | if (dl->vlist[0]) { |
a8173e43 | 4709 | be16_clear(ddf->phys->entries[dl->pdnum].type, |
4710 | cpu_to_be16(DDF_Global_Spare)); | |
4711 | if (!be16_and(ddf->phys | |
4712 | ->entries[dl->pdnum].type, | |
4713 | cpu_to_be16(DDF_Active_in_VD))) { | |
4714 | be16_set(ddf->phys | |
4715 | ->entries[dl->pdnum].type, | |
4716 | cpu_to_be16(DDF_Active_in_VD)); | |
613b0d17 | 4717 | if (in_degraded) |
a8173e43 | 4718 | be16_set(ddf->phys |
4719 | ->entries[dl->pdnum] | |
4720 | .state, | |
4721 | cpu_to_be16 | |
4722 | (DDF_Rebuilding)); | |
613b0d17 | 4723 | } |
7e1432fb NB |
4724 | } |
4725 | if (dl->spare) { | |
a8173e43 | 4726 | be16_clear(ddf->phys->entries[dl->pdnum].type, |
4727 | cpu_to_be16(DDF_Global_Spare)); | |
4728 | be16_set(ddf->phys->entries[dl->pdnum].type, | |
4729 | cpu_to_be16(DDF_Spare)); | |
7e1432fb NB |
4730 | } |
4731 | if (!dl->vlist[0] && !dl->spare) { | |
a8173e43 | 4732 | be16_set(ddf->phys->entries[dl->pdnum].type, |
4733 | cpu_to_be16(DDF_Global_Spare)); | |
4734 | be16_clear(ddf->phys->entries[dl->pdnum].type, | |
4735 | cpu_to_be16(DDF_Spare)); | |
4736 | be16_clear(ddf->phys->entries[dl->pdnum].type, | |
4737 | cpu_to_be16(DDF_Active_in_VD)); | |
7e1432fb | 4738 | } |
88c164f4 | 4739 | } |
c7079c84 N |
4740 | |
4741 | /* Now remove any 'Failed' devices that are not part | |
4742 | * of any VD. They will have the Transition flag set. | |
4743 | * Once done, we need to update all dl->pdnum numbers. | |
4744 | */ | |
4745 | pd2 = 0; | |
a8173e43 | 4746 | for (pdnum = 0; pdnum < be16_to_cpu(ddf->phys->used_pdes); |
92939eb2 | 4747 | pdnum++) { |
a8173e43 | 4748 | if (be16_and(ddf->phys->entries[pdnum].state, |
4749 | cpu_to_be16(DDF_Failed)) | |
4750 | && be16_and(ddf->phys->entries[pdnum].state, | |
92939eb2 N |
4751 | cpu_to_be16(DDF_Transition))) { |
4752 | /* skip this one unless in dlist*/ | |
4753 | for (dl = ddf->dlist; dl; dl = dl->next) | |
4754 | if (dl->pdnum == (int)pdnum) | |
4755 | break; | |
4756 | if (!dl) | |
4757 | continue; | |
4758 | } | |
4759 | if (pdnum == pd2) | |
c7079c84 N |
4760 | pd2++; |
4761 | else { | |
a8173e43 | 4762 | ddf->phys->entries[pd2] = |
4763 | ddf->phys->entries[pdnum]; | |
c7079c84 N |
4764 | for (dl = ddf->dlist; dl; dl = dl->next) |
4765 | if (dl->pdnum == (int)pdnum) | |
4766 | dl->pdnum = pd2; | |
4767 | pd2++; | |
4768 | } | |
92939eb2 | 4769 | } |
a8173e43 | 4770 | ddf->phys->used_pdes = cpu_to_be16(pd2); |
c7079c84 | 4771 | while (pd2 < pdnum) { |
a8173e43 | 4772 | memset(ddf->phys->entries[pd2].guid, 0xff, |
4773 | DDF_GUID_LEN); | |
c7079c84 N |
4774 | pd2++; |
4775 | } | |
4776 | ||
7d5a7ff3 | 4777 | ddf_set_updates_pending(ddf); |
88c164f4 | 4778 | } |
60931cf9 | 4779 | /* case DDF_SPARE_ASSIGN_MAGIC */ |
88c164f4 NB |
4780 | } |
4781 | ||
edd8d13c NB |
4782 | static void ddf_prepare_update(struct supertype *st, |
4783 | struct metadata_update *update) | |
4784 | { | |
4785 | /* This update arrived at managemon. | |
4786 | * We are about to pass it to monitor. | |
4787 | * If a malloc is needed, do it here. | |
4788 | */ | |
4789 | struct ddf_super *ddf = st->sb; | |
60931cf9 | 4790 | be32 *magic = (be32 *)update->buf; |
4791 | if (be32_eq(*magic, DDF_VD_CONF_MAGIC)) { | |
c5943560 | 4792 | struct vcl *vcl; |
4793 | struct vd_config *conf = (struct vd_config *) update->buf; | |
e6b9548d | 4794 | if (posix_memalign(&update->space, 512, |
613b0d17 | 4795 | offsetof(struct vcl, conf) |
c5943560 | 4796 | + ddf->conf_rec_len * 512) != 0) { |
4797 | update->space = NULL; | |
4798 | return; | |
4799 | } | |
4800 | vcl = update->space; | |
4801 | vcl->conf.sec_elmnt_count = conf->sec_elmnt_count; | |
4802 | if (alloc_other_bvds(ddf, vcl) != 0) { | |
4803 | free(update->space); | |
e6b9548d | 4804 | update->space = NULL; |
c5943560 | 4805 | } |
4806 | } | |
edd8d13c NB |
4807 | } |
4808 | ||
7733b91d | 4809 | /* |
4810 | * Check degraded state of a RAID10. | |
4811 | * returns 2 for good, 1 for degraded, 0 for failed, and -1 for error | |
4812 | */ | |
4813 | static int raid10_degraded(struct mdinfo *info) | |
4814 | { | |
4815 | int n_prim, n_bvds; | |
4816 | int i; | |
9591a2de | 4817 | struct mdinfo *d; |
7733b91d | 4818 | char *found; |
4819 | int ret = -1; | |
4820 | ||
7733b91d | 4821 | n_prim = info->array.layout & ~0x100; |
4822 | n_bvds = info->array.raid_disks / n_prim; | |
4823 | found = xmalloc(n_bvds); | |
4824 | if (found == NULL) | |
4825 | return ret; | |
4826 | memset(found, 0, n_bvds); | |
4827 | for (d = info->devs; d; d = d->next) { | |
4828 | i = d->disk.raid_disk / n_prim; | |
4829 | if (i >= n_bvds) { | |
4830 | pr_err("%s: BUG: invalid raid disk\n", __func__); | |
4831 | goto out; | |
4832 | } | |
4833 | if (d->state_fd > 0) | |
4834 | found[i]++; | |
4835 | } | |
4836 | ret = 2; | |
4837 | for (i = 0; i < n_bvds; i++) | |
4838 | if (!found[i]) { | |
4839 | dprintf("%s: BVD %d/%d failed\n", __func__, i, n_bvds); | |
4840 | ret = 0; | |
4841 | goto out; | |
4842 | } else if (found[i] < n_prim) { | |
4843 | dprintf("%s: BVD %d/%d degraded\n", __func__, i, | |
4844 | n_bvds); | |
4845 | ret = 1; | |
4846 | } | |
4847 | out: | |
4848 | free(found); | |
4849 | return ret; | |
4850 | } | |
4851 | ||
7e1432fb NB |
4852 | /* |
4853 | * Check if the array 'a' is degraded but not failed. | |
4854 | * If it is, find as many spares as are available and needed and | |
4855 | * arrange for their inclusion. | |
4856 | * We only choose devices which are not already in the array, | |
4857 | * and prefer those with a spare-assignment to this array. | |
4858 | * otherwise we choose global spares - assuming always that | |
4859 | * there is enough room. | |
4860 | * For each spare that we assign, we return an 'mdinfo' which | |
4861 | * describes the position for the device in the array. | |
4862 | * We also add to 'updates' a DDF_VD_CONF_MAGIC update with | |
4863 | * the new phys_refnum and lba_offset values. | |
4864 | * | |
4865 | * Only worry about BVDs at the moment. | |
4866 | */ | |
4867 | static struct mdinfo *ddf_activate_spare(struct active_array *a, | |
4868 | struct metadata_update **updates) | |
4869 | { | |
4870 | int working = 0; | |
4871 | struct mdinfo *d; | |
4872 | struct ddf_super *ddf = a->container->sb; | |
4873 | int global_ok = 0; | |
4874 | struct mdinfo *rv = NULL; | |
4875 | struct mdinfo *di; | |
4876 | struct metadata_update *mu; | |
4877 | struct dl *dl; | |
4878 | int i; | |
0c78849f | 4879 | unsigned int j; |
baba3f4e | 4880 | struct vcl *vcl; |
7e1432fb | 4881 | struct vd_config *vc; |
baba3f4e | 4882 | unsigned int n_bvd; |
7e1432fb | 4883 | |
7e1432fb NB |
4884 | for (d = a->info.devs ; d ; d = d->next) { |
4885 | if ((d->curr_state & DS_FAULTY) && | |
613b0d17 | 4886 | d->state_fd >= 0) |
7e1432fb NB |
4887 | /* wait for Removal to happen */ |
4888 | return NULL; | |
4889 | if (d->state_fd >= 0) | |
4890 | working ++; | |
4891 | } | |
4892 | ||
7733b91d | 4893 | dprintf("%s: working=%d (%d) level=%d\n", __func__, working, |
a8173e43 | 4894 | a->info.array.raid_disks, |
2c514b71 | 4895 | a->info.array.level); |
7e1432fb NB |
4896 | if (working == a->info.array.raid_disks) |
4897 | return NULL; /* array not degraded */ | |
4898 | switch (a->info.array.level) { | |
4899 | case 1: | |
4900 | if (working == 0) | |
4901 | return NULL; /* failed */ | |
4902 | break; | |
4903 | case 4: | |
4904 | case 5: | |
4905 | if (working < a->info.array.raid_disks - 1) | |
4906 | return NULL; /* failed */ | |
4907 | break; | |
4908 | case 6: | |
4909 | if (working < a->info.array.raid_disks - 2) | |
4910 | return NULL; /* failed */ | |
4911 | break; | |
7733b91d | 4912 | case 10: |
4913 | if (raid10_degraded(&a->info) < 1) | |
4914 | return NULL; | |
4915 | break; | |
7e1432fb NB |
4916 | default: /* concat or stripe */ |
4917 | return NULL; /* failed */ | |
4918 | } | |
4919 | ||
4920 | /* For each slot, if it is not working, find a spare */ | |
4921 | dl = ddf->dlist; | |
4922 | for (i = 0; i < a->info.array.raid_disks; i++) { | |
4923 | for (d = a->info.devs ; d ; d = d->next) | |
4924 | if (d->disk.raid_disk == i) | |
4925 | break; | |
2c514b71 | 4926 | dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0); |
7e1432fb NB |
4927 | if (d && (d->state_fd >= 0)) |
4928 | continue; | |
4929 | ||
4930 | /* OK, this device needs recovery. Find a spare */ | |
4931 | again: | |
4932 | for ( ; dl ; dl = dl->next) { | |
4933 | unsigned long long esize; | |
4934 | unsigned long long pos; | |
4935 | struct mdinfo *d2; | |
4936 | int is_global = 0; | |
4937 | int is_dedicated = 0; | |
4938 | struct extent *ex; | |
f21e18ca | 4939 | unsigned int j; |
6f56dbb9 MW |
4940 | be16 state = ddf->phys->entries[dl->pdnum].state; |
4941 | if (be16_and(state, | |
4942 | cpu_to_be16(DDF_Failed|DDF_Missing)) || | |
4943 | !be16_and(state, | |
4944 | cpu_to_be16(DDF_Online))) | |
4945 | continue; | |
4946 | ||
7e1432fb NB |
4947 | /* If in this array, skip */ |
4948 | for (d2 = a->info.devs ; d2 ; d2 = d2->next) | |
7590d562 N |
4949 | if (d2->state_fd >= 0 && |
4950 | d2->disk.major == dl->major && | |
7e1432fb | 4951 | d2->disk.minor == dl->minor) { |
2a645ee2 MW |
4952 | dprintf("%x:%x (%08x) already in array\n", |
4953 | dl->major, dl->minor, | |
4954 | be32_to_cpu(dl->disk.refnum)); | |
7e1432fb NB |
4955 | break; |
4956 | } | |
4957 | if (d2) | |
4958 | continue; | |
a8173e43 | 4959 | if (be16_and(ddf->phys->entries[dl->pdnum].type, |
4960 | cpu_to_be16(DDF_Spare))) { | |
7e1432fb NB |
4961 | /* Check spare assign record */ |
4962 | if (dl->spare) { | |
4963 | if (dl->spare->type & DDF_spare_dedicated) { | |
4964 | /* check spare_ents for guid */ | |
4965 | for (j = 0 ; | |
a8173e43 | 4966 | j < be16_to_cpu |
4967 | (dl->spare | |
4968 | ->populated); | |
7e1432fb NB |
4969 | j++) { |
4970 | if (memcmp(dl->spare->spare_ents[j].guid, | |
4971 | ddf->virt->entries[a->info.container_member].guid, | |
4972 | DDF_GUID_LEN) == 0) | |
4973 | is_dedicated = 1; | |
4974 | } | |
4975 | } else | |
4976 | is_global = 1; | |
4977 | } | |
a8173e43 | 4978 | } else if (be16_and(ddf->phys->entries[dl->pdnum].type, |
4979 | cpu_to_be16(DDF_Global_Spare))) { | |
7e1432fb | 4980 | is_global = 1; |
a8173e43 | 4981 | } else if (!be16_and(ddf->phys |
4982 | ->entries[dl->pdnum].state, | |
4983 | cpu_to_be16(DDF_Failed))) { | |
e0e7aeaa N |
4984 | /* we can possibly use some of this */ |
4985 | is_global = 1; | |
7e1432fb NB |
4986 | } |
4987 | if ( ! (is_dedicated || | |
4988 | (is_global && global_ok))) { | |
2c514b71 | 4989 | dprintf("%x:%x not suitable: %d %d\n", dl->major, dl->minor, |
613b0d17 | 4990 | is_dedicated, is_global); |
7e1432fb NB |
4991 | continue; |
4992 | } | |
4993 | ||
4994 | /* We are allowed to use this device - is there space? | |
4995 | * We need a->info.component_size sectors */ | |
4996 | ex = get_extents(ddf, dl); | |
4997 | if (!ex) { | |
2c514b71 | 4998 | dprintf("cannot get extents\n"); |
7e1432fb NB |
4999 | continue; |
5000 | } | |
5001 | j = 0; pos = 0; | |
5002 | esize = 0; | |
5003 | ||
5004 | do { | |
5005 | esize = ex[j].start - pos; | |
5006 | if (esize >= a->info.component_size) | |
5007 | break; | |
e5cc7d46 N |
5008 | pos = ex[j].start + ex[j].size; |
5009 | j++; | |
5010 | } while (ex[j-1].size); | |
7e1432fb NB |
5011 | |
5012 | free(ex); | |
5013 | if (esize < a->info.component_size) { | |
e5cc7d46 N |
5014 | dprintf("%x:%x has no room: %llu %llu\n", |
5015 | dl->major, dl->minor, | |
2c514b71 | 5016 | esize, a->info.component_size); |
7e1432fb NB |
5017 | /* No room */ |
5018 | continue; | |
5019 | } | |
5020 | ||
5021 | /* Cool, we have a device with some space at pos */ | |
503975b9 | 5022 | di = xcalloc(1, sizeof(*di)); |
7e1432fb NB |
5023 | di->disk.number = i; |
5024 | di->disk.raid_disk = i; | |
5025 | di->disk.major = dl->major; | |
5026 | di->disk.minor = dl->minor; | |
5027 | di->disk.state = 0; | |
d23534e4 | 5028 | di->recovery_start = 0; |
7e1432fb NB |
5029 | di->data_offset = pos; |
5030 | di->component_size = a->info.component_size; | |
5031 | di->container_member = dl->pdnum; | |
5032 | di->next = rv; | |
5033 | rv = di; | |
2a645ee2 MW |
5034 | dprintf("%x:%x (%08x) to be %d at %llu\n", |
5035 | dl->major, dl->minor, | |
5036 | be32_to_cpu(dl->disk.refnum), i, pos); | |
7e1432fb NB |
5037 | |
5038 | break; | |
5039 | } | |
5040 | if (!dl && ! global_ok) { | |
5041 | /* not enough dedicated spares, try global */ | |
5042 | global_ok = 1; | |
5043 | dl = ddf->dlist; | |
5044 | goto again; | |
5045 | } | |
5046 | } | |
5047 | ||
5048 | if (!rv) | |
5049 | /* No spares found */ | |
5050 | return rv; | |
5051 | /* Now 'rv' has a list of devices to return. | |
5052 | * Create a metadata_update record to update the | |
5053 | * phys_refnum and lba_offset values | |
5054 | */ | |
bb925ff0 | 5055 | vc = find_vdcr(ddf, a->info.container_member, rv->disk.raid_disk, |
0c78849f | 5056 | &n_bvd, &vcl); |
5057 | if (vc == NULL) | |
5058 | return NULL; | |
5059 | ||
503975b9 N |
5060 | mu = xmalloc(sizeof(*mu)); |
5061 | if (posix_memalign(&mu->space, 512, sizeof(struct vcl)) != 0) { | |
79244939 DW |
5062 | free(mu); |
5063 | mu = NULL; | |
5064 | } | |
0c78849f | 5065 | |
5066 | mu->len = ddf->conf_rec_len * 512 * vcl->conf.sec_elmnt_count; | |
5067 | mu->buf = xmalloc(mu->len); | |
7590d562 | 5068 | mu->space = NULL; |
f50ae22e | 5069 | mu->space_list = NULL; |
7e1432fb | 5070 | mu->next = *updates; |
0c78849f | 5071 | memcpy(mu->buf, &vcl->conf, ddf->conf_rec_len * 512); |
5072 | for (j = 1; j < vcl->conf.sec_elmnt_count; j++) | |
5073 | memcpy(mu->buf + j * ddf->conf_rec_len * 512, | |
5074 | vcl->other_bvds[j-1], ddf->conf_rec_len * 512); | |
7e1432fb NB |
5075 | |
5076 | vc = (struct vd_config*)mu->buf; | |
7e1432fb | 5077 | for (di = rv ; di ; di = di->next) { |
0c78849f | 5078 | unsigned int i_sec, i_prim; |
5079 | i_sec = di->disk.raid_disk | |
5080 | / be16_to_cpu(vcl->conf.prim_elmnt_count); | |
5081 | i_prim = di->disk.raid_disk | |
5082 | % be16_to_cpu(vcl->conf.prim_elmnt_count); | |
5083 | vc = (struct vd_config *)(mu->buf | |
5084 | + i_sec * ddf->conf_rec_len * 512); | |
5085 | for (dl = ddf->dlist; dl; dl = dl->next) | |
5086 | if (dl->major == di->disk.major | |
5087 | && dl->minor == di->disk.minor) | |
5088 | break; | |
5089 | if (!dl) { | |
5090 | pr_err("%s: BUG: can't find disk %d (%d/%d)\n", | |
5091 | __func__, di->disk.raid_disk, | |
5092 | di->disk.major, di->disk.minor); | |
5093 | return NULL; | |
5094 | } | |
5095 | vc->phys_refnum[i_prim] = ddf->phys->entries[dl->pdnum].refnum; | |
5096 | LBA_OFFSET(ddf, vc)[i_prim] = cpu_to_be64(di->data_offset); | |
2a645ee2 MW |
5097 | dprintf("BVD %u gets %u: %08x at %llu\n", i_sec, i_prim, |
5098 | be32_to_cpu(vc->phys_refnum[i_prim]), | |
5099 | be64_to_cpu(LBA_OFFSET(ddf, vc)[i_prim])); | |
7e1432fb NB |
5100 | } |
5101 | *updates = mu; | |
5102 | return rv; | |
5103 | } | |
0e600426 | 5104 | #endif /* MDASSEMBLE */ |
7e1432fb | 5105 | |
b640a252 N |
5106 | static int ddf_level_to_layout(int level) |
5107 | { | |
5108 | switch(level) { | |
5109 | case 0: | |
5110 | case 1: | |
5111 | return 0; | |
5112 | case 5: | |
5113 | return ALGORITHM_LEFT_SYMMETRIC; | |
5114 | case 6: | |
5115 | return ALGORITHM_ROTATING_N_CONTINUE; | |
5116 | case 10: | |
5117 | return 0x102; | |
5118 | default: | |
5119 | return UnSet; | |
5120 | } | |
5121 | } | |
5122 | ||
30f58b22 DW |
5123 | static void default_geometry_ddf(struct supertype *st, int *level, int *layout, int *chunk) |
5124 | { | |
5125 | if (level && *level == UnSet) | |
5126 | *level = LEVEL_CONTAINER; | |
5127 | ||
5128 | if (level && layout && *layout == UnSet) | |
5129 | *layout = ddf_level_to_layout(*level); | |
5130 | } | |
5131 | ||
a322f70c DW |
5132 | struct superswitch super_ddf = { |
5133 | #ifndef MDASSEMBLE | |
5134 | .examine_super = examine_super_ddf, | |
5135 | .brief_examine_super = brief_examine_super_ddf, | |
4737ae25 | 5136 | .brief_examine_subarrays = brief_examine_subarrays_ddf, |
bceedeec | 5137 | .export_examine_super = export_examine_super_ddf, |
a322f70c DW |
5138 | .detail_super = detail_super_ddf, |
5139 | .brief_detail_super = brief_detail_super_ddf, | |
5140 | .validate_geometry = validate_geometry_ddf, | |
78e44928 | 5141 | .write_init_super = write_init_super_ddf, |
0e600426 | 5142 | .add_to_super = add_to_super_ddf, |
4dd968cc | 5143 | .remove_from_super = remove_from_super_ddf, |
2b959fbf | 5144 | .load_container = load_container_ddf, |
74db60b0 | 5145 | .copy_metadata = copy_metadata_ddf, |
4441541f | 5146 | .kill_subarray = kill_subarray_ddf, |
a322f70c DW |
5147 | #endif |
5148 | .match_home = match_home_ddf, | |
5149 | .uuid_from_super= uuid_from_super_ddf, | |
5150 | .getinfo_super = getinfo_super_ddf, | |
5151 | .update_super = update_super_ddf, | |
5152 | ||
5153 | .avail_size = avail_size_ddf, | |
5154 | ||
a19c88b8 NB |
5155 | .compare_super = compare_super_ddf, |
5156 | ||
a322f70c | 5157 | .load_super = load_super_ddf, |
ba7eb04f | 5158 | .init_super = init_super_ddf, |
955e9ea1 | 5159 | .store_super = store_super_ddf, |
a322f70c DW |
5160 | .free_super = free_super_ddf, |
5161 | .match_metadata_desc = match_metadata_desc_ddf, | |
78e44928 | 5162 | .container_content = container_content_ddf, |
30f58b22 | 5163 | .default_geometry = default_geometry_ddf, |
a322f70c | 5164 | |
a322f70c | 5165 | .external = 1, |
549e9569 | 5166 | |
0e600426 | 5167 | #ifndef MDASSEMBLE |
549e9569 NB |
5168 | /* for mdmon */ |
5169 | .open_new = ddf_open_new, | |
ed9d66aa | 5170 | .set_array_state= ddf_set_array_state, |
549e9569 NB |
5171 | .set_disk = ddf_set_disk, |
5172 | .sync_metadata = ddf_sync_metadata, | |
88c164f4 | 5173 | .process_update = ddf_process_update, |
edd8d13c | 5174 | .prepare_update = ddf_prepare_update, |
7e1432fb | 5175 | .activate_spare = ddf_activate_spare, |
0e600426 | 5176 | #endif |
4cce4069 | 5177 | .name = "ddf", |
a322f70c | 5178 | }; |