]>
Commit | Line | Data |
---|---|---|
cdddbdbc DW |
1 | /* |
2 | * mdadm - Intel(R) Matrix Storage Manager Support | |
3 | * | |
a54d5262 | 4 | * Copyright (C) 2002-2008 Intel Corporation |
cdddbdbc DW |
5 | * |
6 | * This program is free software; you can redistribute it and/or modify it | |
7 | * under the terms and conditions of the GNU General Public License, | |
8 | * version 2, as published by the Free Software Foundation. | |
9 | * | |
10 | * This program is distributed in the hope it will be useful, but WITHOUT | |
11 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
12 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for | |
13 | * more details. | |
14 | * | |
15 | * You should have received a copy of the GNU General Public License along with | |
16 | * this program; if not, write to the Free Software Foundation, Inc., | |
17 | * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA. | |
18 | */ | |
19 | ||
51006d85 | 20 | #define HAVE_STDINT_H 1 |
cdddbdbc | 21 | #include "mdadm.h" |
c2a1e7da | 22 | #include "mdmon.h" |
51006d85 | 23 | #include "sha1.h" |
88c32bb1 | 24 | #include "platform-intel.h" |
cdddbdbc DW |
25 | #include <values.h> |
26 | #include <scsi/sg.h> | |
27 | #include <ctype.h> | |
d665cc31 | 28 | #include <dirent.h> |
cdddbdbc DW |
29 | |
30 | /* MPB == Metadata Parameter Block */ | |
31 | #define MPB_SIGNATURE "Intel Raid ISM Cfg Sig. " | |
32 | #define MPB_SIG_LEN (strlen(MPB_SIGNATURE)) | |
33 | #define MPB_VERSION_RAID0 "1.0.00" | |
34 | #define MPB_VERSION_RAID1 "1.1.00" | |
fe7ed8cb DW |
35 | #define MPB_VERSION_MANY_VOLUMES_PER_ARRAY "1.2.00" |
36 | #define MPB_VERSION_3OR4_DISK_ARRAY "1.2.01" | |
cdddbdbc | 37 | #define MPB_VERSION_RAID5 "1.2.02" |
fe7ed8cb DW |
38 | #define MPB_VERSION_5OR6_DISK_ARRAY "1.2.04" |
39 | #define MPB_VERSION_CNG "1.2.06" | |
40 | #define MPB_VERSION_ATTRIBS "1.3.00" | |
cdddbdbc DW |
41 | #define MAX_SIGNATURE_LENGTH 32 |
42 | #define MAX_RAID_SERIAL_LEN 16 | |
fe7ed8cb | 43 | |
19482bcc AK |
44 | /* supports RAID0 */ |
45 | #define MPB_ATTRIB_RAID0 __cpu_to_le32(0x00000001) | |
46 | /* supports RAID1 */ | |
47 | #define MPB_ATTRIB_RAID1 __cpu_to_le32(0x00000002) | |
48 | /* supports RAID10 */ | |
49 | #define MPB_ATTRIB_RAID10 __cpu_to_le32(0x00000004) | |
50 | /* supports RAID1E */ | |
51 | #define MPB_ATTRIB_RAID1E __cpu_to_le32(0x00000008) | |
52 | /* supports RAID5 */ | |
53 | #define MPB_ATTRIB_RAID5 __cpu_to_le32(0x00000010) | |
54 | /* supports RAID CNG */ | |
55 | #define MPB_ATTRIB_RAIDCNG __cpu_to_le32(0x00000020) | |
56 | /* supports expanded stripe sizes of 256K, 512K and 1MB */ | |
57 | #define MPB_ATTRIB_EXP_STRIPE_SIZE __cpu_to_le32(0x00000040) | |
58 | ||
59 | /* The OROM Support RST Caching of Volumes */ | |
60 | #define MPB_ATTRIB_NVM __cpu_to_le32(0x02000000) | |
61 | /* The OROM supports creating disks greater than 2TB */ | |
62 | #define MPB_ATTRIB_2TB_DISK __cpu_to_le32(0x04000000) | |
63 | /* The OROM supports Bad Block Management */ | |
64 | #define MPB_ATTRIB_BBM __cpu_to_le32(0x08000000) | |
65 | ||
66 | /* THe OROM Supports NVM Caching of Volumes */ | |
67 | #define MPB_ATTRIB_NEVER_USE2 __cpu_to_le32(0x10000000) | |
68 | /* The OROM supports creating volumes greater than 2TB */ | |
69 | #define MPB_ATTRIB_2TB __cpu_to_le32(0x20000000) | |
70 | /* originally for PMP, now it's wasted b/c. Never use this bit! */ | |
71 | #define MPB_ATTRIB_NEVER_USE __cpu_to_le32(0x40000000) | |
72 | /* Verify MPB contents against checksum after reading MPB */ | |
73 | #define MPB_ATTRIB_CHECKSUM_VERIFY __cpu_to_le32(0x80000000) | |
74 | ||
75 | /* Define all supported attributes that have to be accepted by mdadm | |
76 | */ | |
418f9b36 | 77 | #define MPB_ATTRIB_SUPPORTED (MPB_ATTRIB_CHECKSUM_VERIFY | \ |
19482bcc AK |
78 | MPB_ATTRIB_2TB | \ |
79 | MPB_ATTRIB_2TB_DISK | \ | |
80 | MPB_ATTRIB_RAID0 | \ | |
81 | MPB_ATTRIB_RAID1 | \ | |
82 | MPB_ATTRIB_RAID10 | \ | |
83 | MPB_ATTRIB_RAID5 | \ | |
418f9b36 N |
84 | MPB_ATTRIB_EXP_STRIPE_SIZE) |
85 | ||
86 | /* Define attributes that are unused but not harmful */ | |
87 | #define MPB_ATTRIB_IGNORED (MPB_ATTRIB_NEVER_USE) | |
fe7ed8cb | 88 | |
8e59f3d8 | 89 | #define MPB_SECTOR_CNT 2210 |
c2c087e6 | 90 | #define IMSM_RESERVED_SECTORS 4096 |
b81221b7 | 91 | #define NUM_BLOCKS_DIRTY_STRIPE_REGION 2056 |
979d38be | 92 | #define SECT_PER_MB_SHIFT 11 |
cdddbdbc DW |
93 | |
94 | /* Disk configuration info. */ | |
95 | #define IMSM_MAX_DEVICES 255 | |
96 | struct imsm_disk { | |
97 | __u8 serial[MAX_RAID_SERIAL_LEN];/* 0xD8 - 0xE7 ascii serial number */ | |
98 | __u32 total_blocks; /* 0xE8 - 0xEB total blocks */ | |
99 | __u32 scsi_id; /* 0xEC - 0xEF scsi ID */ | |
f2f27e63 DW |
100 | #define SPARE_DISK __cpu_to_le32(0x01) /* Spare */ |
101 | #define CONFIGURED_DISK __cpu_to_le32(0x02) /* Member of some RaidDev */ | |
102 | #define FAILED_DISK __cpu_to_le32(0x04) /* Permanent failure */ | |
cdddbdbc | 103 | __u32 status; /* 0xF0 - 0xF3 */ |
fe7ed8cb DW |
104 | __u32 owner_cfg_num; /* which config 0,1,2... owns this disk */ |
105 | #define IMSM_DISK_FILLERS 4 | |
cdddbdbc DW |
106 | __u32 filler[IMSM_DISK_FILLERS]; /* 0xF4 - 0x107 MPB_DISK_FILLERS for future expansion */ |
107 | }; | |
108 | ||
109 | /* RAID map configuration infos. */ | |
110 | struct imsm_map { | |
111 | __u32 pba_of_lba0; /* start address of partition */ | |
112 | __u32 blocks_per_member;/* blocks per member */ | |
113 | __u32 num_data_stripes; /* number of data stripes */ | |
114 | __u16 blocks_per_strip; | |
115 | __u8 map_state; /* Normal, Uninitialized, Degraded, Failed */ | |
116 | #define IMSM_T_STATE_NORMAL 0 | |
117 | #define IMSM_T_STATE_UNINITIALIZED 1 | |
e3bba0e0 DW |
118 | #define IMSM_T_STATE_DEGRADED 2 |
119 | #define IMSM_T_STATE_FAILED 3 | |
cdddbdbc DW |
120 | __u8 raid_level; |
121 | #define IMSM_T_RAID0 0 | |
122 | #define IMSM_T_RAID1 1 | |
123 | #define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */ | |
124 | __u8 num_members; /* number of member disks */ | |
fe7ed8cb DW |
125 | __u8 num_domains; /* number of parity domains */ |
126 | __u8 failed_disk_num; /* valid only when state is degraded */ | |
252d23c0 | 127 | __u8 ddf; |
cdddbdbc | 128 | __u32 filler[7]; /* expansion area */ |
7eef0453 | 129 | #define IMSM_ORD_REBUILD (1 << 24) |
cdddbdbc | 130 | __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members], |
7eef0453 DW |
131 | * top byte contains some flags |
132 | */ | |
cdddbdbc DW |
133 | } __attribute__ ((packed)); |
134 | ||
135 | struct imsm_vol { | |
f8f603f1 | 136 | __u32 curr_migr_unit; |
fe7ed8cb | 137 | __u32 checkpoint_id; /* id to access curr_migr_unit */ |
cdddbdbc | 138 | __u8 migr_state; /* Normal or Migrating */ |
e3bba0e0 DW |
139 | #define MIGR_INIT 0 |
140 | #define MIGR_REBUILD 1 | |
141 | #define MIGR_VERIFY 2 /* analagous to echo check > sync_action */ | |
142 | #define MIGR_GEN_MIGR 3 | |
143 | #define MIGR_STATE_CHANGE 4 | |
1484e727 | 144 | #define MIGR_REPAIR 5 |
cdddbdbc DW |
145 | __u8 migr_type; /* Initializing, Rebuilding, ... */ |
146 | __u8 dirty; | |
fe7ed8cb DW |
147 | __u8 fs_state; /* fast-sync state for CnG (0xff == disabled) */ |
148 | __u16 verify_errors; /* number of mismatches */ | |
149 | __u16 bad_blocks; /* number of bad blocks during verify */ | |
150 | __u32 filler[4]; | |
cdddbdbc DW |
151 | struct imsm_map map[1]; |
152 | /* here comes another one if migr_state */ | |
153 | } __attribute__ ((packed)); | |
154 | ||
155 | struct imsm_dev { | |
fe7ed8cb | 156 | __u8 volume[MAX_RAID_SERIAL_LEN]; |
cdddbdbc DW |
157 | __u32 size_low; |
158 | __u32 size_high; | |
fe7ed8cb DW |
159 | #define DEV_BOOTABLE __cpu_to_le32(0x01) |
160 | #define DEV_BOOT_DEVICE __cpu_to_le32(0x02) | |
161 | #define DEV_READ_COALESCING __cpu_to_le32(0x04) | |
162 | #define DEV_WRITE_COALESCING __cpu_to_le32(0x08) | |
163 | #define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10) | |
164 | #define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20) | |
165 | #define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40) | |
166 | #define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80) | |
167 | #define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100) | |
168 | #define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200) | |
169 | #define DEV_CLONE_N_GO __cpu_to_le32(0x400) | |
170 | #define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800) | |
171 | #define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000) | |
cdddbdbc DW |
172 | __u32 status; /* Persistent RaidDev status */ |
173 | __u32 reserved_blocks; /* Reserved blocks at beginning of volume */ | |
fe7ed8cb DW |
174 | __u8 migr_priority; |
175 | __u8 num_sub_vols; | |
176 | __u8 tid; | |
177 | __u8 cng_master_disk; | |
178 | __u16 cache_policy; | |
179 | __u8 cng_state; | |
180 | __u8 cng_sub_state; | |
181 | #define IMSM_DEV_FILLERS 10 | |
cdddbdbc DW |
182 | __u32 filler[IMSM_DEV_FILLERS]; |
183 | struct imsm_vol vol; | |
184 | } __attribute__ ((packed)); | |
185 | ||
186 | struct imsm_super { | |
187 | __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */ | |
188 | __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */ | |
189 | __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */ | |
190 | __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */ | |
191 | __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */ | |
604b746f JD |
192 | __u32 error_log_size; /* 0x30 - 0x33 in bytes */ |
193 | __u32 attributes; /* 0x34 - 0x37 */ | |
cdddbdbc DW |
194 | __u8 num_disks; /* 0x38 Number of configured disks */ |
195 | __u8 num_raid_devs; /* 0x39 Number of configured volumes */ | |
604b746f JD |
196 | __u8 error_log_pos; /* 0x3A */ |
197 | __u8 fill[1]; /* 0x3B */ | |
198 | __u32 cache_size; /* 0x3c - 0x40 in mb */ | |
199 | __u32 orig_family_num; /* 0x40 - 0x43 original family num */ | |
200 | __u32 pwr_cycle_count; /* 0x44 - 0x47 simulated power cycle count for array */ | |
201 | __u32 bbm_log_size; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */ | |
202 | #define IMSM_FILLERS 35 | |
203 | __u32 filler[IMSM_FILLERS]; /* 0x4C - 0xD7 RAID_MPB_FILLERS */ | |
cdddbdbc DW |
204 | struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */ |
205 | /* here comes imsm_dev[num_raid_devs] */ | |
604b746f | 206 | /* here comes BBM logs */ |
cdddbdbc DW |
207 | } __attribute__ ((packed)); |
208 | ||
604b746f JD |
209 | #define BBM_LOG_MAX_ENTRIES 254 |
210 | ||
211 | struct bbm_log_entry { | |
212 | __u64 defective_block_start; | |
213 | #define UNREADABLE 0xFFFFFFFF | |
214 | __u32 spare_block_offset; | |
215 | __u16 remapped_marked_count; | |
216 | __u16 disk_ordinal; | |
217 | } __attribute__ ((__packed__)); | |
218 | ||
219 | struct bbm_log { | |
220 | __u32 signature; /* 0xABADB10C */ | |
221 | __u32 entry_count; | |
222 | __u32 reserved_spare_block_count; /* 0 */ | |
223 | __u32 reserved; /* 0xFFFF */ | |
224 | __u64 first_spare_lba; | |
225 | struct bbm_log_entry mapped_block_entries[BBM_LOG_MAX_ENTRIES]; | |
226 | } __attribute__ ((__packed__)); | |
227 | ||
228 | ||
cdddbdbc DW |
229 | #ifndef MDASSEMBLE |
230 | static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" }; | |
231 | #endif | |
232 | ||
8e59f3d8 AK |
233 | #define RAID_DISK_RESERVED_BLOCKS_IMSM_HI 2209 |
234 | ||
235 | #define GEN_MIGR_AREA_SIZE 2048 /* General Migration Copy Area size in blocks */ | |
236 | ||
237 | #define UNIT_SRC_NORMAL 0 /* Source data for curr_migr_unit must | |
238 | * be recovered using srcMap */ | |
239 | #define UNIT_SRC_IN_CP_AREA 1 /* Source data for curr_migr_unit has | |
240 | * already been migrated and must | |
241 | * be recovered from checkpoint area */ | |
242 | struct migr_record { | |
243 | __u32 rec_status; /* Status used to determine how to restart | |
244 | * migration in case it aborts | |
245 | * in some fashion */ | |
246 | __u32 curr_migr_unit; /* 0..numMigrUnits-1 */ | |
247 | __u32 family_num; /* Family number of MPB | |
248 | * containing the RaidDev | |
249 | * that is migrating */ | |
250 | __u32 ascending_migr; /* True if migrating in increasing | |
251 | * order of lbas */ | |
252 | __u32 blocks_per_unit; /* Num disk blocks per unit of operation */ | |
253 | __u32 dest_depth_per_unit; /* Num member blocks each destMap | |
254 | * member disk | |
255 | * advances per unit-of-operation */ | |
256 | __u32 ckpt_area_pba; /* Pba of first block of ckpt copy area */ | |
257 | __u32 dest_1st_member_lba; /* First member lba on first | |
258 | * stripe of destination */ | |
259 | __u32 num_migr_units; /* Total num migration units-of-op */ | |
260 | __u32 post_migr_vol_cap; /* Size of volume after | |
261 | * migration completes */ | |
262 | __u32 post_migr_vol_cap_hi; /* Expansion space for LBA64 */ | |
263 | __u32 ckpt_read_disk_num; /* Which member disk in destSubMap[0] the | |
264 | * migration ckpt record was read from | |
265 | * (for recovered migrations) */ | |
266 | } __attribute__ ((__packed__)); | |
267 | ||
1484e727 DW |
268 | static __u8 migr_type(struct imsm_dev *dev) |
269 | { | |
270 | if (dev->vol.migr_type == MIGR_VERIFY && | |
271 | dev->status & DEV_VERIFY_AND_FIX) | |
272 | return MIGR_REPAIR; | |
273 | else | |
274 | return dev->vol.migr_type; | |
275 | } | |
276 | ||
277 | static void set_migr_type(struct imsm_dev *dev, __u8 migr_type) | |
278 | { | |
279 | /* for compatibility with older oroms convert MIGR_REPAIR, into | |
280 | * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status | |
281 | */ | |
282 | if (migr_type == MIGR_REPAIR) { | |
283 | dev->vol.migr_type = MIGR_VERIFY; | |
284 | dev->status |= DEV_VERIFY_AND_FIX; | |
285 | } else { | |
286 | dev->vol.migr_type = migr_type; | |
287 | dev->status &= ~DEV_VERIFY_AND_FIX; | |
288 | } | |
289 | } | |
290 | ||
87eb16df | 291 | static unsigned int sector_count(__u32 bytes) |
cdddbdbc | 292 | { |
87eb16df DW |
293 | return ((bytes + (512-1)) & (~(512-1))) / 512; |
294 | } | |
cdddbdbc | 295 | |
87eb16df DW |
296 | static unsigned int mpb_sectors(struct imsm_super *mpb) |
297 | { | |
298 | return sector_count(__le32_to_cpu(mpb->mpb_size)); | |
cdddbdbc DW |
299 | } |
300 | ||
ba2de7ba DW |
301 | struct intel_dev { |
302 | struct imsm_dev *dev; | |
303 | struct intel_dev *next; | |
f21e18ca | 304 | unsigned index; |
ba2de7ba DW |
305 | }; |
306 | ||
88654014 LM |
307 | struct intel_hba { |
308 | enum sys_dev_type type; | |
309 | char *path; | |
310 | char *pci_id; | |
311 | struct intel_hba *next; | |
312 | }; | |
313 | ||
1a64be56 LM |
314 | enum action { |
315 | DISK_REMOVE = 1, | |
316 | DISK_ADD | |
317 | }; | |
cdddbdbc DW |
318 | /* internal representation of IMSM metadata */ |
319 | struct intel_super { | |
320 | union { | |
949c47a0 DW |
321 | void *buf; /* O_DIRECT buffer for reading/writing metadata */ |
322 | struct imsm_super *anchor; /* immovable parameters */ | |
cdddbdbc | 323 | }; |
8e59f3d8 AK |
324 | union { |
325 | void *migr_rec_buf; /* buffer for I/O operations */ | |
326 | struct migr_record *migr_rec; /* migration record */ | |
327 | }; | |
949c47a0 | 328 | size_t len; /* size of the 'buf' allocation */ |
4d7b1503 DW |
329 | void *next_buf; /* for realloc'ing buf from the manager */ |
330 | size_t next_len; | |
c2c087e6 | 331 | int updates_pending; /* count of pending updates for mdmon */ |
bf5a934a | 332 | int current_vol; /* index of raid device undergoing creation */ |
0dcecb2e | 333 | __u32 create_offset; /* common start for 'current_vol' */ |
148acb7b | 334 | __u32 random; /* random data for seeding new family numbers */ |
ba2de7ba | 335 | struct intel_dev *devlist; |
cdddbdbc DW |
336 | struct dl { |
337 | struct dl *next; | |
338 | int index; | |
339 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
340 | int major, minor; | |
341 | char *devname; | |
b9f594fe | 342 | struct imsm_disk disk; |
cdddbdbc | 343 | int fd; |
0dcecb2e DW |
344 | int extent_cnt; |
345 | struct extent *e; /* for determining freespace @ create */ | |
efb30e7f | 346 | int raiddisk; /* slot to fill in autolayout */ |
1a64be56 | 347 | enum action action; |
ca0748fa | 348 | } *disks, *current_disk; |
1a64be56 LM |
349 | struct dl *disk_mgmt_list; /* list of disks to add/remove while mdmon |
350 | active */ | |
47ee5a45 | 351 | struct dl *missing; /* disks removed while we weren't looking */ |
43dad3d6 | 352 | struct bbm_log *bbm_log; |
88654014 | 353 | struct intel_hba *hba; /* device path of the raid controller for this metadata */ |
88c32bb1 | 354 | const struct imsm_orom *orom; /* platform firmware support */ |
a2b97981 DW |
355 | struct intel_super *next; /* (temp) list for disambiguating family_num */ |
356 | }; | |
357 | ||
358 | struct intel_disk { | |
359 | struct imsm_disk disk; | |
360 | #define IMSM_UNKNOWN_OWNER (-1) | |
361 | int owner; | |
362 | struct intel_disk *next; | |
cdddbdbc DW |
363 | }; |
364 | ||
c2c087e6 DW |
365 | struct extent { |
366 | unsigned long long start, size; | |
367 | }; | |
368 | ||
694575e7 KW |
369 | /* definitions of reshape process types */ |
370 | enum imsm_reshape_type { | |
371 | CH_TAKEOVER, | |
b5347799 | 372 | CH_MIGRATION, |
694575e7 KW |
373 | }; |
374 | ||
88758e9d DW |
375 | /* definition of messages passed to imsm_process_update */ |
376 | enum imsm_update_type { | |
377 | update_activate_spare, | |
8273f55e | 378 | update_create_array, |
33414a01 | 379 | update_kill_array, |
aa534678 | 380 | update_rename_array, |
1a64be56 | 381 | update_add_remove_disk, |
78b10e66 | 382 | update_reshape_container_disks, |
48c5303a | 383 | update_reshape_migration, |
2d40f3a1 AK |
384 | update_takeover, |
385 | update_general_migration_checkpoint, | |
88758e9d DW |
386 | }; |
387 | ||
388 | struct imsm_update_activate_spare { | |
389 | enum imsm_update_type type; | |
d23fe947 | 390 | struct dl *dl; |
88758e9d DW |
391 | int slot; |
392 | int array; | |
393 | struct imsm_update_activate_spare *next; | |
394 | }; | |
395 | ||
78b10e66 N |
396 | struct geo_params { |
397 | int dev_id; | |
398 | char *dev_name; | |
399 | long long size; | |
400 | int level; | |
401 | int layout; | |
402 | int chunksize; | |
403 | int raid_disks; | |
404 | }; | |
405 | ||
bb025c2f KW |
406 | enum takeover_direction { |
407 | R10_TO_R0, | |
408 | R0_TO_R10 | |
409 | }; | |
410 | struct imsm_update_takeover { | |
411 | enum imsm_update_type type; | |
412 | int subarray; | |
413 | enum takeover_direction direction; | |
414 | }; | |
78b10e66 N |
415 | |
416 | struct imsm_update_reshape { | |
417 | enum imsm_update_type type; | |
418 | int old_raid_disks; | |
419 | int new_raid_disks; | |
48c5303a PC |
420 | |
421 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ | |
422 | }; | |
423 | ||
424 | struct imsm_update_reshape_migration { | |
425 | enum imsm_update_type type; | |
426 | int old_raid_disks; | |
427 | int new_raid_disks; | |
428 | /* fields for array migration changes | |
429 | */ | |
430 | int subdev; | |
431 | int new_level; | |
432 | int new_layout; | |
4bba0439 | 433 | int new_chunksize; |
48c5303a | 434 | |
d195167d | 435 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ |
78b10e66 N |
436 | }; |
437 | ||
2d40f3a1 AK |
438 | struct imsm_update_general_migration_checkpoint { |
439 | enum imsm_update_type type; | |
440 | __u32 curr_migr_unit; | |
441 | }; | |
442 | ||
54c2c1ea DW |
443 | struct disk_info { |
444 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
445 | }; | |
446 | ||
8273f55e DW |
447 | struct imsm_update_create_array { |
448 | enum imsm_update_type type; | |
8273f55e | 449 | int dev_idx; |
6a3e913e | 450 | struct imsm_dev dev; |
8273f55e DW |
451 | }; |
452 | ||
33414a01 DW |
453 | struct imsm_update_kill_array { |
454 | enum imsm_update_type type; | |
455 | int dev_idx; | |
456 | }; | |
457 | ||
aa534678 DW |
458 | struct imsm_update_rename_array { |
459 | enum imsm_update_type type; | |
460 | __u8 name[MAX_RAID_SERIAL_LEN]; | |
461 | int dev_idx; | |
462 | }; | |
463 | ||
1a64be56 | 464 | struct imsm_update_add_remove_disk { |
43dad3d6 DW |
465 | enum imsm_update_type type; |
466 | }; | |
467 | ||
88654014 LM |
468 | |
469 | static const char *_sys_dev_type[] = { | |
470 | [SYS_DEV_UNKNOWN] = "Unknown", | |
471 | [SYS_DEV_SAS] = "SAS", | |
472 | [SYS_DEV_SATA] = "SATA" | |
473 | }; | |
474 | ||
475 | const char *get_sys_dev_type(enum sys_dev_type type) | |
476 | { | |
477 | if (type >= SYS_DEV_MAX) | |
478 | type = SYS_DEV_UNKNOWN; | |
479 | ||
480 | return _sys_dev_type[type]; | |
481 | } | |
482 | ||
483 | static struct intel_hba * alloc_intel_hba(struct sys_dev *device) | |
484 | { | |
485 | struct intel_hba *result = malloc(sizeof(*result)); | |
486 | if (result) { | |
487 | result->type = device->type; | |
488 | result->path = strdup(device->path); | |
489 | result->next = NULL; | |
490 | if (result->path && (result->pci_id = strrchr(result->path, '/')) != NULL) | |
491 | result->pci_id++; | |
492 | } | |
493 | return result; | |
494 | } | |
495 | ||
496 | static struct intel_hba * find_intel_hba(struct intel_hba *hba, struct sys_dev *device) | |
497 | { | |
498 | struct intel_hba *result=NULL; | |
499 | for (result = hba; result; result = result->next) { | |
500 | if (result->type == device->type && strcmp(result->path, device->path) == 0) | |
501 | break; | |
502 | } | |
503 | return result; | |
504 | } | |
505 | ||
b4cf4cba | 506 | static int attach_hba_to_super(struct intel_super *super, struct sys_dev *device) |
88654014 LM |
507 | { |
508 | struct intel_hba *hba; | |
509 | ||
510 | /* check if disk attached to Intel HBA */ | |
511 | hba = find_intel_hba(super->hba, device); | |
512 | if (hba != NULL) | |
513 | return 1; | |
514 | /* Check if HBA is already attached to super */ | |
515 | if (super->hba == NULL) { | |
516 | super->hba = alloc_intel_hba(device); | |
517 | return 1; | |
518 | } | |
519 | ||
520 | hba = super->hba; | |
521 | /* Intel metadata allows for all disks attached to the same type HBA. | |
522 | * Do not sypport odf HBA types mixing | |
523 | */ | |
524 | if (device->type != hba->type) | |
525 | return 2; | |
526 | ||
527 | while (hba->next) | |
528 | hba = hba->next; | |
529 | ||
530 | hba->next = alloc_intel_hba(device); | |
531 | return 1; | |
532 | } | |
533 | ||
534 | static struct sys_dev* find_disk_attached_hba(int fd, const char *devname) | |
535 | { | |
536 | struct sys_dev *list, *elem, *prev; | |
537 | char *disk_path; | |
538 | ||
539 | if ((list = find_intel_devices()) == NULL) | |
540 | return 0; | |
541 | ||
542 | if (fd < 0) | |
543 | disk_path = (char *) devname; | |
544 | else | |
545 | disk_path = diskfd_to_devpath(fd); | |
546 | ||
547 | if (!disk_path) { | |
548 | free_sys_dev(&list); | |
549 | return 0; | |
550 | } | |
551 | ||
552 | for (prev = NULL, elem = list; elem; prev = elem, elem = elem->next) { | |
553 | if (path_attached_to_hba(disk_path, elem->path)) { | |
554 | if (prev == NULL) | |
555 | list = list->next; | |
556 | else | |
557 | prev->next = elem->next; | |
558 | elem->next = NULL; | |
559 | if (disk_path != devname) | |
560 | free(disk_path); | |
561 | free_sys_dev(&list); | |
562 | return elem; | |
563 | } | |
564 | } | |
565 | if (disk_path != devname) | |
566 | free(disk_path); | |
567 | free_sys_dev(&list); | |
568 | ||
569 | return NULL; | |
570 | } | |
571 | ||
572 | ||
d424212e N |
573 | static int find_intel_hba_capability(int fd, struct intel_super *super, |
574 | char *devname); | |
f2f5c343 | 575 | |
cdddbdbc DW |
576 | static struct supertype *match_metadata_desc_imsm(char *arg) |
577 | { | |
578 | struct supertype *st; | |
579 | ||
580 | if (strcmp(arg, "imsm") != 0 && | |
581 | strcmp(arg, "default") != 0 | |
582 | ) | |
583 | return NULL; | |
584 | ||
585 | st = malloc(sizeof(*st)); | |
4e9d2186 AW |
586 | if (!st) |
587 | return NULL; | |
ef609477 | 588 | memset(st, 0, sizeof(*st)); |
d1d599ea | 589 | st->container_dev = NoMdDev; |
cdddbdbc DW |
590 | st->ss = &super_imsm; |
591 | st->max_devs = IMSM_MAX_DEVICES; | |
592 | st->minor_version = 0; | |
593 | st->sb = NULL; | |
594 | return st; | |
595 | } | |
596 | ||
0e600426 | 597 | #ifndef MDASSEMBLE |
cdddbdbc DW |
598 | static __u8 *get_imsm_version(struct imsm_super *mpb) |
599 | { | |
600 | return &mpb->sig[MPB_SIG_LEN]; | |
601 | } | |
9e2d750d | 602 | #endif |
cdddbdbc | 603 | |
949c47a0 DW |
604 | /* retrieve a disk directly from the anchor when the anchor is known to be |
605 | * up-to-date, currently only at load time | |
606 | */ | |
607 | static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index) | |
cdddbdbc | 608 | { |
949c47a0 | 609 | if (index >= mpb->num_disks) |
cdddbdbc DW |
610 | return NULL; |
611 | return &mpb->disk[index]; | |
612 | } | |
613 | ||
95d07a2c LM |
614 | /* retrieve the disk description based on a index of the disk |
615 | * in the sub-array | |
616 | */ | |
617 | static struct dl *get_imsm_dl_disk(struct intel_super *super, __u8 index) | |
949c47a0 | 618 | { |
b9f594fe DW |
619 | struct dl *d; |
620 | ||
621 | for (d = super->disks; d; d = d->next) | |
622 | if (d->index == index) | |
95d07a2c LM |
623 | return d; |
624 | ||
625 | return NULL; | |
626 | } | |
627 | /* retrieve a disk from the parsed metadata */ | |
628 | static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index) | |
629 | { | |
630 | struct dl *dl; | |
631 | ||
632 | dl = get_imsm_dl_disk(super, index); | |
633 | if (dl) | |
634 | return &dl->disk; | |
635 | ||
b9f594fe | 636 | return NULL; |
949c47a0 DW |
637 | } |
638 | ||
639 | /* generate a checksum directly from the anchor when the anchor is known to be | |
640 | * up-to-date, currently only at load or write_super after coalescing | |
641 | */ | |
642 | static __u32 __gen_imsm_checksum(struct imsm_super *mpb) | |
cdddbdbc DW |
643 | { |
644 | __u32 end = mpb->mpb_size / sizeof(end); | |
645 | __u32 *p = (__u32 *) mpb; | |
646 | __u32 sum = 0; | |
647 | ||
97f734fd N |
648 | while (end--) { |
649 | sum += __le32_to_cpu(*p); | |
650 | p++; | |
651 | } | |
cdddbdbc DW |
652 | |
653 | return sum - __le32_to_cpu(mpb->check_sum); | |
654 | } | |
655 | ||
a965f303 DW |
656 | static size_t sizeof_imsm_map(struct imsm_map *map) |
657 | { | |
658 | return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1); | |
659 | } | |
660 | ||
661 | struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map) | |
cdddbdbc | 662 | { |
5e7b0330 AK |
663 | /* A device can have 2 maps if it is in the middle of a migration. |
664 | * If second_map is: | |
665 | * 0 - we return the first map | |
666 | * 1 - we return the second map if it exists, else NULL | |
667 | * -1 - we return the second map if it exists, else the first | |
668 | */ | |
a965f303 DW |
669 | struct imsm_map *map = &dev->vol.map[0]; |
670 | ||
5e7b0330 | 671 | if (second_map == 1 && !dev->vol.migr_state) |
a965f303 | 672 | return NULL; |
5e7b0330 AK |
673 | else if (second_map == 1 || |
674 | (second_map < 0 && dev->vol.migr_state)) { | |
a965f303 DW |
675 | void *ptr = map; |
676 | ||
677 | return ptr + sizeof_imsm_map(map); | |
678 | } else | |
679 | return map; | |
5e7b0330 | 680 | |
a965f303 | 681 | } |
cdddbdbc | 682 | |
3393c6af DW |
683 | /* return the size of the device. |
684 | * migr_state increases the returned size if map[0] were to be duplicated | |
685 | */ | |
686 | static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state) | |
a965f303 DW |
687 | { |
688 | size_t size = sizeof(*dev) - sizeof(struct imsm_map) + | |
689 | sizeof_imsm_map(get_imsm_map(dev, 0)); | |
cdddbdbc DW |
690 | |
691 | /* migrating means an additional map */ | |
a965f303 DW |
692 | if (dev->vol.migr_state) |
693 | size += sizeof_imsm_map(get_imsm_map(dev, 1)); | |
3393c6af DW |
694 | else if (migr_state) |
695 | size += sizeof_imsm_map(get_imsm_map(dev, 0)); | |
cdddbdbc DW |
696 | |
697 | return size; | |
698 | } | |
699 | ||
54c2c1ea DW |
700 | #ifndef MDASSEMBLE |
701 | /* retrieve disk serial number list from a metadata update */ | |
702 | static struct disk_info *get_disk_info(struct imsm_update_create_array *update) | |
703 | { | |
704 | void *u = update; | |
705 | struct disk_info *inf; | |
706 | ||
707 | inf = u + sizeof(*update) - sizeof(struct imsm_dev) + | |
708 | sizeof_imsm_dev(&update->dev, 0); | |
709 | ||
710 | return inf; | |
711 | } | |
712 | #endif | |
713 | ||
949c47a0 | 714 | static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index) |
cdddbdbc DW |
715 | { |
716 | int offset; | |
717 | int i; | |
718 | void *_mpb = mpb; | |
719 | ||
949c47a0 | 720 | if (index >= mpb->num_raid_devs) |
cdddbdbc DW |
721 | return NULL; |
722 | ||
723 | /* devices start after all disks */ | |
724 | offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb; | |
725 | ||
726 | for (i = 0; i <= index; i++) | |
727 | if (i == index) | |
728 | return _mpb + offset; | |
729 | else | |
3393c6af | 730 | offset += sizeof_imsm_dev(_mpb + offset, 0); |
cdddbdbc DW |
731 | |
732 | return NULL; | |
733 | } | |
734 | ||
949c47a0 DW |
735 | static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index) |
736 | { | |
ba2de7ba DW |
737 | struct intel_dev *dv; |
738 | ||
949c47a0 DW |
739 | if (index >= super->anchor->num_raid_devs) |
740 | return NULL; | |
ba2de7ba DW |
741 | for (dv = super->devlist; dv; dv = dv->next) |
742 | if (dv->index == index) | |
743 | return dv->dev; | |
744 | return NULL; | |
949c47a0 DW |
745 | } |
746 | ||
98130f40 AK |
747 | /* |
748 | * for second_map: | |
749 | * == 0 get first map | |
750 | * == 1 get second map | |
751 | * == -1 than get map according to the current migr_state | |
752 | */ | |
753 | static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev, | |
754 | int slot, | |
755 | int second_map) | |
7eef0453 DW |
756 | { |
757 | struct imsm_map *map; | |
758 | ||
5e7b0330 | 759 | map = get_imsm_map(dev, second_map); |
7eef0453 | 760 | |
ff077194 DW |
761 | /* top byte identifies disk under rebuild */ |
762 | return __le32_to_cpu(map->disk_ord_tbl[slot]); | |
763 | } | |
764 | ||
765 | #define ord_to_idx(ord) (((ord) << 8) >> 8) | |
98130f40 | 766 | static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot, int second_map) |
ff077194 | 767 | { |
98130f40 | 768 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, second_map); |
ff077194 DW |
769 | |
770 | return ord_to_idx(ord); | |
7eef0453 DW |
771 | } |
772 | ||
be73972f DW |
773 | static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord) |
774 | { | |
775 | map->disk_ord_tbl[slot] = __cpu_to_le32(ord); | |
776 | } | |
777 | ||
f21e18ca | 778 | static int get_imsm_disk_slot(struct imsm_map *map, unsigned idx) |
620b1713 DW |
779 | { |
780 | int slot; | |
781 | __u32 ord; | |
782 | ||
783 | for (slot = 0; slot < map->num_members; slot++) { | |
784 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
785 | if (ord_to_idx(ord) == idx) | |
786 | return slot; | |
787 | } | |
788 | ||
789 | return -1; | |
790 | } | |
791 | ||
cdddbdbc DW |
792 | static int get_imsm_raid_level(struct imsm_map *map) |
793 | { | |
794 | if (map->raid_level == 1) { | |
795 | if (map->num_members == 2) | |
796 | return 1; | |
797 | else | |
798 | return 10; | |
799 | } | |
800 | ||
801 | return map->raid_level; | |
802 | } | |
803 | ||
c2c087e6 DW |
804 | static int cmp_extent(const void *av, const void *bv) |
805 | { | |
806 | const struct extent *a = av; | |
807 | const struct extent *b = bv; | |
808 | if (a->start < b->start) | |
809 | return -1; | |
810 | if (a->start > b->start) | |
811 | return 1; | |
812 | return 0; | |
813 | } | |
814 | ||
0dcecb2e | 815 | static int count_memberships(struct dl *dl, struct intel_super *super) |
c2c087e6 | 816 | { |
c2c087e6 | 817 | int memberships = 0; |
620b1713 | 818 | int i; |
c2c087e6 | 819 | |
949c47a0 DW |
820 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
821 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
a965f303 | 822 | struct imsm_map *map = get_imsm_map(dev, 0); |
c2c087e6 | 823 | |
620b1713 DW |
824 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
825 | memberships++; | |
c2c087e6 | 826 | } |
0dcecb2e DW |
827 | |
828 | return memberships; | |
829 | } | |
830 | ||
b81221b7 CA |
831 | static __u32 imsm_min_reserved_sectors(struct intel_super *super); |
832 | ||
0dcecb2e DW |
833 | static struct extent *get_extents(struct intel_super *super, struct dl *dl) |
834 | { | |
835 | /* find a list of used extents on the given physical device */ | |
836 | struct extent *rv, *e; | |
620b1713 | 837 | int i; |
0dcecb2e | 838 | int memberships = count_memberships(dl, super); |
b276dd33 DW |
839 | __u32 reservation; |
840 | ||
841 | /* trim the reserved area for spares, so they can join any array | |
842 | * regardless of whether the OROM has assigned sectors from the | |
843 | * IMSM_RESERVED_SECTORS region | |
844 | */ | |
845 | if (dl->index == -1) | |
b81221b7 | 846 | reservation = imsm_min_reserved_sectors(super); |
b276dd33 DW |
847 | else |
848 | reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
0dcecb2e | 849 | |
c2c087e6 DW |
850 | rv = malloc(sizeof(struct extent) * (memberships + 1)); |
851 | if (!rv) | |
852 | return NULL; | |
853 | e = rv; | |
854 | ||
949c47a0 DW |
855 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
856 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
a965f303 | 857 | struct imsm_map *map = get_imsm_map(dev, 0); |
c2c087e6 | 858 | |
620b1713 DW |
859 | if (get_imsm_disk_slot(map, dl->index) >= 0) { |
860 | e->start = __le32_to_cpu(map->pba_of_lba0); | |
861 | e->size = __le32_to_cpu(map->blocks_per_member); | |
862 | e++; | |
c2c087e6 DW |
863 | } |
864 | } | |
865 | qsort(rv, memberships, sizeof(*rv), cmp_extent); | |
866 | ||
14e8215b DW |
867 | /* determine the start of the metadata |
868 | * when no raid devices are defined use the default | |
869 | * ...otherwise allow the metadata to truncate the value | |
870 | * as is the case with older versions of imsm | |
871 | */ | |
872 | if (memberships) { | |
873 | struct extent *last = &rv[memberships - 1]; | |
874 | __u32 remainder; | |
875 | ||
876 | remainder = __le32_to_cpu(dl->disk.total_blocks) - | |
877 | (last->start + last->size); | |
dda5855f DW |
878 | /* round down to 1k block to satisfy precision of the kernel |
879 | * 'size' interface | |
880 | */ | |
881 | remainder &= ~1UL; | |
882 | /* make sure remainder is still sane */ | |
f21e18ca | 883 | if (remainder < (unsigned)ROUND_UP(super->len, 512) >> 9) |
dda5855f | 884 | remainder = ROUND_UP(super->len, 512) >> 9; |
14e8215b DW |
885 | if (reservation > remainder) |
886 | reservation = remainder; | |
887 | } | |
888 | e->start = __le32_to_cpu(dl->disk.total_blocks) - reservation; | |
c2c087e6 DW |
889 | e->size = 0; |
890 | return rv; | |
891 | } | |
892 | ||
14e8215b DW |
893 | /* try to determine how much space is reserved for metadata from |
894 | * the last get_extents() entry, otherwise fallback to the | |
895 | * default | |
896 | */ | |
897 | static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl) | |
898 | { | |
899 | struct extent *e; | |
900 | int i; | |
901 | __u32 rv; | |
902 | ||
903 | /* for spares just return a minimal reservation which will grow | |
904 | * once the spare is picked up by an array | |
905 | */ | |
906 | if (dl->index == -1) | |
907 | return MPB_SECTOR_CNT; | |
908 | ||
909 | e = get_extents(super, dl); | |
910 | if (!e) | |
911 | return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
912 | ||
913 | /* scroll to last entry */ | |
914 | for (i = 0; e[i].size; i++) | |
915 | continue; | |
916 | ||
917 | rv = __le32_to_cpu(dl->disk.total_blocks) - e[i].start; | |
918 | ||
919 | free(e); | |
920 | ||
921 | return rv; | |
922 | } | |
923 | ||
25ed7e59 DW |
924 | static int is_spare(struct imsm_disk *disk) |
925 | { | |
926 | return (disk->status & SPARE_DISK) == SPARE_DISK; | |
927 | } | |
928 | ||
929 | static int is_configured(struct imsm_disk *disk) | |
930 | { | |
931 | return (disk->status & CONFIGURED_DISK) == CONFIGURED_DISK; | |
932 | } | |
933 | ||
934 | static int is_failed(struct imsm_disk *disk) | |
935 | { | |
936 | return (disk->status & FAILED_DISK) == FAILED_DISK; | |
937 | } | |
938 | ||
b81221b7 CA |
939 | /* try to determine how much space is reserved for metadata from |
940 | * the last get_extents() entry on the smallest active disk, | |
941 | * otherwise fallback to the default | |
942 | */ | |
943 | static __u32 imsm_min_reserved_sectors(struct intel_super *super) | |
944 | { | |
945 | struct extent *e; | |
946 | int i; | |
947 | __u32 min_active, remainder; | |
948 | __u32 rv = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
949 | struct dl *dl, *dl_min = NULL; | |
950 | ||
951 | if (!super) | |
952 | return rv; | |
953 | ||
954 | min_active = 0; | |
955 | for (dl = super->disks; dl; dl = dl->next) { | |
956 | if (dl->index < 0) | |
957 | continue; | |
958 | if (dl->disk.total_blocks < min_active || min_active == 0) { | |
959 | dl_min = dl; | |
960 | min_active = dl->disk.total_blocks; | |
961 | } | |
962 | } | |
963 | if (!dl_min) | |
964 | return rv; | |
965 | ||
966 | /* find last lba used by subarrays on the smallest active disk */ | |
967 | e = get_extents(super, dl_min); | |
968 | if (!e) | |
969 | return rv; | |
970 | for (i = 0; e[i].size; i++) | |
971 | continue; | |
972 | ||
973 | remainder = min_active - e[i].start; | |
974 | free(e); | |
975 | ||
976 | /* to give priority to recovery we should not require full | |
977 | IMSM_RESERVED_SECTORS from the spare */ | |
978 | rv = MPB_SECTOR_CNT + NUM_BLOCKS_DIRTY_STRIPE_REGION; | |
979 | ||
980 | /* if real reservation is smaller use that value */ | |
981 | return (remainder < rv) ? remainder : rv; | |
982 | } | |
983 | ||
80e7f8c3 AC |
984 | /* Return minimum size of a spare that can be used in this array*/ |
985 | static unsigned long long min_acceptable_spare_size_imsm(struct supertype *st) | |
986 | { | |
987 | struct intel_super *super = st->sb; | |
988 | struct dl *dl; | |
989 | struct extent *e; | |
990 | int i; | |
991 | unsigned long long rv = 0; | |
b81221b7 | 992 | __u32 reservation; |
80e7f8c3 AC |
993 | |
994 | if (!super) | |
995 | return rv; | |
996 | /* find first active disk in array */ | |
997 | dl = super->disks; | |
998 | while (dl && (is_failed(&dl->disk) || dl->index == -1)) | |
999 | dl = dl->next; | |
1000 | if (!dl) | |
1001 | return rv; | |
1002 | /* find last lba used by subarrays */ | |
1003 | e = get_extents(super, dl); | |
1004 | if (!e) | |
1005 | return rv; | |
1006 | for (i = 0; e[i].size; i++) | |
1007 | continue; | |
1008 | if (i > 0) | |
1009 | rv = e[i-1].start + e[i-1].size; | |
b81221b7 | 1010 | reservation = __le32_to_cpu(dl->disk.total_blocks) - e[i].start; |
80e7f8c3 | 1011 | free(e); |
b81221b7 | 1012 | |
80e7f8c3 | 1013 | /* add the amount of space needed for metadata */ |
b81221b7 CA |
1014 | rv = rv + imsm_min_reserved_sectors(super); |
1015 | ||
80e7f8c3 AC |
1016 | return rv * 512; |
1017 | } | |
1018 | ||
1799c9e8 | 1019 | #ifndef MDASSEMBLE |
c47b0ff6 AK |
1020 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
1021 | struct imsm_dev *dev); | |
1e5c6983 | 1022 | |
c47b0ff6 AK |
1023 | static void print_imsm_dev(struct intel_super *super, |
1024 | struct imsm_dev *dev, | |
1025 | char *uuid, | |
1026 | int disk_idx) | |
cdddbdbc DW |
1027 | { |
1028 | __u64 sz; | |
0d80bb2f | 1029 | int slot, i; |
a965f303 | 1030 | struct imsm_map *map = get_imsm_map(dev, 0); |
dd8bcb3b | 1031 | struct imsm_map *map2 = get_imsm_map(dev, 1); |
b10b37b8 | 1032 | __u32 ord; |
cdddbdbc DW |
1033 | |
1034 | printf("\n"); | |
1e7bc0ed | 1035 | printf("[%.16s]:\n", dev->volume); |
44470971 | 1036 | printf(" UUID : %s\n", uuid); |
dd8bcb3b AK |
1037 | printf(" RAID Level : %d", get_imsm_raid_level(map)); |
1038 | if (map2) | |
1039 | printf(" <-- %d", get_imsm_raid_level(map2)); | |
1040 | printf("\n"); | |
1041 | printf(" Members : %d", map->num_members); | |
1042 | if (map2) | |
1043 | printf(" <-- %d", map2->num_members); | |
1044 | printf("\n"); | |
0d80bb2f DW |
1045 | printf(" Slots : ["); |
1046 | for (i = 0; i < map->num_members; i++) { | |
dd8bcb3b | 1047 | ord = get_imsm_ord_tbl_ent(dev, i, 0); |
0d80bb2f DW |
1048 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
1049 | } | |
dd8bcb3b AK |
1050 | printf("]"); |
1051 | if (map2) { | |
1052 | printf(" <-- ["); | |
1053 | for (i = 0; i < map2->num_members; i++) { | |
1054 | ord = get_imsm_ord_tbl_ent(dev, i, 1); | |
1055 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); | |
1056 | } | |
1057 | printf("]"); | |
1058 | } | |
1059 | printf("\n"); | |
7095bccb AK |
1060 | printf(" Failed disk : "); |
1061 | if (map->failed_disk_num == 0xff) | |
1062 | printf("none"); | |
1063 | else | |
1064 | printf("%i", map->failed_disk_num); | |
1065 | printf("\n"); | |
620b1713 DW |
1066 | slot = get_imsm_disk_slot(map, disk_idx); |
1067 | if (slot >= 0) { | |
98130f40 | 1068 | ord = get_imsm_ord_tbl_ent(dev, slot, -1); |
b10b37b8 DW |
1069 | printf(" This Slot : %d%s\n", slot, |
1070 | ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : ""); | |
1071 | } else | |
cdddbdbc DW |
1072 | printf(" This Slot : ?\n"); |
1073 | sz = __le32_to_cpu(dev->size_high); | |
1074 | sz <<= 32; | |
1075 | sz += __le32_to_cpu(dev->size_low); | |
1076 | printf(" Array Size : %llu%s\n", (unsigned long long)sz, | |
1077 | human_size(sz * 512)); | |
1078 | sz = __le32_to_cpu(map->blocks_per_member); | |
1079 | printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz, | |
1080 | human_size(sz * 512)); | |
1081 | printf(" Sector Offset : %u\n", | |
1082 | __le32_to_cpu(map->pba_of_lba0)); | |
1083 | printf(" Num Stripes : %u\n", | |
1084 | __le32_to_cpu(map->num_data_stripes)); | |
dd8bcb3b | 1085 | printf(" Chunk Size : %u KiB", |
cdddbdbc | 1086 | __le16_to_cpu(map->blocks_per_strip) / 2); |
dd8bcb3b AK |
1087 | if (map2) |
1088 | printf(" <-- %u KiB", | |
1089 | __le16_to_cpu(map2->blocks_per_strip) / 2); | |
1090 | printf("\n"); | |
cdddbdbc | 1091 | printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks)); |
8655a7b1 | 1092 | printf(" Migrate State : "); |
1484e727 DW |
1093 | if (dev->vol.migr_state) { |
1094 | if (migr_type(dev) == MIGR_INIT) | |
8655a7b1 | 1095 | printf("initialize\n"); |
1484e727 | 1096 | else if (migr_type(dev) == MIGR_REBUILD) |
8655a7b1 | 1097 | printf("rebuild\n"); |
1484e727 | 1098 | else if (migr_type(dev) == MIGR_VERIFY) |
8655a7b1 | 1099 | printf("check\n"); |
1484e727 | 1100 | else if (migr_type(dev) == MIGR_GEN_MIGR) |
8655a7b1 | 1101 | printf("general migration\n"); |
1484e727 | 1102 | else if (migr_type(dev) == MIGR_STATE_CHANGE) |
8655a7b1 | 1103 | printf("state change\n"); |
1484e727 | 1104 | else if (migr_type(dev) == MIGR_REPAIR) |
8655a7b1 | 1105 | printf("repair\n"); |
1484e727 | 1106 | else |
8655a7b1 DW |
1107 | printf("<unknown:%d>\n", migr_type(dev)); |
1108 | } else | |
1109 | printf("idle\n"); | |
3393c6af DW |
1110 | printf(" Map State : %s", map_state_str[map->map_state]); |
1111 | if (dev->vol.migr_state) { | |
1112 | struct imsm_map *map = get_imsm_map(dev, 1); | |
1e5c6983 | 1113 | |
b10b37b8 | 1114 | printf(" <-- %s", map_state_str[map->map_state]); |
1e5c6983 DW |
1115 | printf("\n Checkpoint : %u (%llu)", |
1116 | __le32_to_cpu(dev->vol.curr_migr_unit), | |
c47b0ff6 | 1117 | (unsigned long long)blocks_per_migr_unit(super, dev)); |
3393c6af DW |
1118 | } |
1119 | printf("\n"); | |
cdddbdbc | 1120 | printf(" Dirty State : %s\n", dev->vol.dirty ? "dirty" : "clean"); |
cdddbdbc DW |
1121 | } |
1122 | ||
0ec1f4e8 | 1123 | static void print_imsm_disk(struct imsm_disk *disk, int index, __u32 reserved) |
cdddbdbc | 1124 | { |
1f24f035 | 1125 | char str[MAX_RAID_SERIAL_LEN + 1]; |
cdddbdbc DW |
1126 | __u64 sz; |
1127 | ||
0ec1f4e8 | 1128 | if (index < -1 || !disk) |
e9d82038 DW |
1129 | return; |
1130 | ||
cdddbdbc | 1131 | printf("\n"); |
1f24f035 | 1132 | snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial); |
0ec1f4e8 DW |
1133 | if (index >= 0) |
1134 | printf(" Disk%02d Serial : %s\n", index, str); | |
1135 | else | |
1136 | printf(" Disk Serial : %s\n", str); | |
25ed7e59 DW |
1137 | printf(" State :%s%s%s\n", is_spare(disk) ? " spare" : "", |
1138 | is_configured(disk) ? " active" : "", | |
1139 | is_failed(disk) ? " failed" : ""); | |
cdddbdbc | 1140 | printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id)); |
14e8215b | 1141 | sz = __le32_to_cpu(disk->total_blocks) - reserved; |
cdddbdbc DW |
1142 | printf(" Usable Size : %llu%s\n", (unsigned long long)sz, |
1143 | human_size(sz * 512)); | |
1144 | } | |
1145 | ||
520e69e2 AK |
1146 | static int is_gen_migration(struct imsm_dev *dev); |
1147 | ||
1148 | void examine_migr_rec_imsm(struct intel_super *super) | |
1149 | { | |
1150 | struct migr_record *migr_rec = super->migr_rec; | |
1151 | struct imsm_super *mpb = super->anchor; | |
1152 | int i; | |
1153 | ||
1154 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
1155 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1156 | if (is_gen_migration(dev) == 0) | |
1157 | continue; | |
1158 | ||
1159 | printf("\nMigration Record Information:"); | |
1160 | if (super->disks->index > 1) { | |
1161 | printf(" Empty\n "); | |
1162 | printf("Examine one of first two disks in array\n"); | |
1163 | break; | |
1164 | } | |
1165 | printf("\n Status : "); | |
1166 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
1167 | printf("Normal\n"); | |
1168 | else | |
1169 | printf("Contains Data\n"); | |
1170 | printf(" Current Unit : %u\n", | |
1171 | __le32_to_cpu(migr_rec->curr_migr_unit)); | |
1172 | printf(" Family : %u\n", | |
1173 | __le32_to_cpu(migr_rec->family_num)); | |
1174 | printf(" Ascending : %u\n", | |
1175 | __le32_to_cpu(migr_rec->ascending_migr)); | |
1176 | printf(" Blocks Per Unit : %u\n", | |
1177 | __le32_to_cpu(migr_rec->blocks_per_unit)); | |
1178 | printf(" Dest. Depth Per Unit : %u\n", | |
1179 | __le32_to_cpu(migr_rec->dest_depth_per_unit)); | |
1180 | printf(" Checkpoint Area pba : %u\n", | |
1181 | __le32_to_cpu(migr_rec->ckpt_area_pba)); | |
1182 | printf(" First member lba : %u\n", | |
1183 | __le32_to_cpu(migr_rec->dest_1st_member_lba)); | |
1184 | printf(" Total Number of Units : %u\n", | |
1185 | __le32_to_cpu(migr_rec->num_migr_units)); | |
1186 | printf(" Size of volume : %u\n", | |
1187 | __le32_to_cpu(migr_rec->post_migr_vol_cap)); | |
1188 | printf(" Expansion space for LBA64 : %u\n", | |
1189 | __le32_to_cpu(migr_rec->post_migr_vol_cap_hi)); | |
1190 | printf(" Record was read from : %u\n", | |
1191 | __le32_to_cpu(migr_rec->ckpt_read_disk_num)); | |
1192 | ||
1193 | break; | |
1194 | } | |
1195 | } | |
9e2d750d | 1196 | #endif /* MDASSEMBLE */ |
19482bcc AK |
1197 | /******************************************************************************* |
1198 | * function: imsm_check_attributes | |
1199 | * Description: Function checks if features represented by attributes flags | |
1200 | * are supported by mdadm. | |
1201 | * Parameters: | |
1202 | * attributes - Attributes read from metadata | |
1203 | * Returns: | |
1204 | * 0 - passed attributes contains unsupported features flags | |
1205 | * 1 - all features are supported | |
1206 | ******************************************************************************/ | |
1207 | static int imsm_check_attributes(__u32 attributes) | |
1208 | { | |
1209 | int ret_val = 1; | |
418f9b36 N |
1210 | __u32 not_supported = MPB_ATTRIB_SUPPORTED^0xffffffff; |
1211 | ||
1212 | not_supported &= ~MPB_ATTRIB_IGNORED; | |
19482bcc AK |
1213 | |
1214 | not_supported &= attributes; | |
1215 | if (not_supported) { | |
418f9b36 N |
1216 | fprintf(stderr, Name "(IMSM): Unsupported attributes : %x\n", |
1217 | (unsigned)__le32_to_cpu(not_supported)); | |
19482bcc AK |
1218 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { |
1219 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY \n"); | |
1220 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
1221 | } | |
1222 | if (not_supported & MPB_ATTRIB_2TB) { | |
1223 | dprintf("\t\tMPB_ATTRIB_2TB\n"); | |
1224 | not_supported ^= MPB_ATTRIB_2TB; | |
1225 | } | |
1226 | if (not_supported & MPB_ATTRIB_RAID0) { | |
1227 | dprintf("\t\tMPB_ATTRIB_RAID0\n"); | |
1228 | not_supported ^= MPB_ATTRIB_RAID0; | |
1229 | } | |
1230 | if (not_supported & MPB_ATTRIB_RAID1) { | |
1231 | dprintf("\t\tMPB_ATTRIB_RAID1\n"); | |
1232 | not_supported ^= MPB_ATTRIB_RAID1; | |
1233 | } | |
1234 | if (not_supported & MPB_ATTRIB_RAID10) { | |
1235 | dprintf("\t\tMPB_ATTRIB_RAID10\n"); | |
1236 | not_supported ^= MPB_ATTRIB_RAID10; | |
1237 | } | |
1238 | if (not_supported & MPB_ATTRIB_RAID1E) { | |
1239 | dprintf("\t\tMPB_ATTRIB_RAID1E\n"); | |
1240 | not_supported ^= MPB_ATTRIB_RAID1E; | |
1241 | } | |
1242 | if (not_supported & MPB_ATTRIB_RAID5) { | |
1243 | dprintf("\t\tMPB_ATTRIB_RAID5\n"); | |
1244 | not_supported ^= MPB_ATTRIB_RAID5; | |
1245 | } | |
1246 | if (not_supported & MPB_ATTRIB_RAIDCNG) { | |
1247 | dprintf("\t\tMPB_ATTRIB_RAIDCNG\n"); | |
1248 | not_supported ^= MPB_ATTRIB_RAIDCNG; | |
1249 | } | |
1250 | if (not_supported & MPB_ATTRIB_BBM) { | |
1251 | dprintf("\t\tMPB_ATTRIB_BBM\n"); | |
1252 | not_supported ^= MPB_ATTRIB_BBM; | |
1253 | } | |
1254 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { | |
1255 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY (== MPB_ATTRIB_LEGACY)\n"); | |
1256 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
1257 | } | |
1258 | if (not_supported & MPB_ATTRIB_EXP_STRIPE_SIZE) { | |
1259 | dprintf("\t\tMPB_ATTRIB_EXP_STRIP_SIZE\n"); | |
1260 | not_supported ^= MPB_ATTRIB_EXP_STRIPE_SIZE; | |
1261 | } | |
1262 | if (not_supported & MPB_ATTRIB_2TB_DISK) { | |
1263 | dprintf("\t\tMPB_ATTRIB_2TB_DISK\n"); | |
1264 | not_supported ^= MPB_ATTRIB_2TB_DISK; | |
1265 | } | |
1266 | if (not_supported & MPB_ATTRIB_NEVER_USE2) { | |
1267 | dprintf("\t\tMPB_ATTRIB_NEVER_USE2\n"); | |
1268 | not_supported ^= MPB_ATTRIB_NEVER_USE2; | |
1269 | } | |
1270 | if (not_supported & MPB_ATTRIB_NEVER_USE) { | |
1271 | dprintf("\t\tMPB_ATTRIB_NEVER_USE\n"); | |
1272 | not_supported ^= MPB_ATTRIB_NEVER_USE; | |
1273 | } | |
1274 | ||
1275 | if (not_supported) | |
1276 | dprintf(Name "(IMSM): Unknown attributes : %x\n", not_supported); | |
1277 | ||
1278 | ret_val = 0; | |
1279 | } | |
1280 | ||
1281 | return ret_val; | |
1282 | } | |
1283 | ||
9e2d750d | 1284 | #ifndef MDASSEMBLE |
a5d85af7 | 1285 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map); |
44470971 | 1286 | |
cdddbdbc DW |
1287 | static void examine_super_imsm(struct supertype *st, char *homehost) |
1288 | { | |
1289 | struct intel_super *super = st->sb; | |
949c47a0 | 1290 | struct imsm_super *mpb = super->anchor; |
cdddbdbc DW |
1291 | char str[MAX_SIGNATURE_LENGTH]; |
1292 | int i; | |
27fd6274 DW |
1293 | struct mdinfo info; |
1294 | char nbuf[64]; | |
cdddbdbc | 1295 | __u32 sum; |
14e8215b | 1296 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
94827db3 | 1297 | struct dl *dl; |
27fd6274 | 1298 | |
cdddbdbc DW |
1299 | snprintf(str, MPB_SIG_LEN, "%s", mpb->sig); |
1300 | printf(" Magic : %s\n", str); | |
1301 | snprintf(str, strlen(MPB_VERSION_RAID0), "%s", get_imsm_version(mpb)); | |
1302 | printf(" Version : %s\n", get_imsm_version(mpb)); | |
148acb7b | 1303 | printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num)); |
cdddbdbc DW |
1304 | printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num)); |
1305 | printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num)); | |
19482bcc AK |
1306 | printf(" Attributes : "); |
1307 | if (imsm_check_attributes(mpb->attributes)) | |
1308 | printf("All supported\n"); | |
1309 | else | |
1310 | printf("not supported\n"); | |
a5d85af7 | 1311 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1312 | fname_from_uuid(st, &info, nbuf, ':'); |
27fd6274 | 1313 | printf(" UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
1314 | sum = __le32_to_cpu(mpb->check_sum); |
1315 | printf(" Checksum : %08x %s\n", sum, | |
949c47a0 | 1316 | __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect"); |
87eb16df | 1317 | printf(" MPB Sectors : %d\n", mpb_sectors(mpb)); |
cdddbdbc DW |
1318 | printf(" Disks : %d\n", mpb->num_disks); |
1319 | printf(" RAID Devices : %d\n", mpb->num_raid_devs); | |
0ec1f4e8 | 1320 | print_imsm_disk(__get_imsm_disk(mpb, super->disks->index), super->disks->index, reserved); |
604b746f JD |
1321 | if (super->bbm_log) { |
1322 | struct bbm_log *log = super->bbm_log; | |
1323 | ||
1324 | printf("\n"); | |
1325 | printf("Bad Block Management Log:\n"); | |
1326 | printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size)); | |
1327 | printf(" Signature : %x\n", __le32_to_cpu(log->signature)); | |
1328 | printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count)); | |
1329 | printf(" Spare Blocks : %d\n", __le32_to_cpu(log->reserved_spare_block_count)); | |
13a3b65d N |
1330 | printf(" First Spare : %llx\n", |
1331 | (unsigned long long) __le64_to_cpu(log->first_spare_lba)); | |
604b746f | 1332 | } |
44470971 DW |
1333 | for (i = 0; i < mpb->num_raid_devs; i++) { |
1334 | struct mdinfo info; | |
1335 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1336 | ||
1337 | super->current_vol = i; | |
a5d85af7 | 1338 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1339 | fname_from_uuid(st, &info, nbuf, ':'); |
c47b0ff6 | 1340 | print_imsm_dev(super, dev, nbuf + 5, super->disks->index); |
44470971 | 1341 | } |
cdddbdbc DW |
1342 | for (i = 0; i < mpb->num_disks; i++) { |
1343 | if (i == super->disks->index) | |
1344 | continue; | |
0ec1f4e8 | 1345 | print_imsm_disk(__get_imsm_disk(mpb, i), i, reserved); |
cdddbdbc | 1346 | } |
94827db3 | 1347 | |
0ec1f4e8 DW |
1348 | for (dl = super->disks; dl; dl = dl->next) |
1349 | if (dl->index == -1) | |
1350 | print_imsm_disk(&dl->disk, -1, reserved); | |
520e69e2 AK |
1351 | |
1352 | examine_migr_rec_imsm(super); | |
cdddbdbc DW |
1353 | } |
1354 | ||
061f2c6a | 1355 | static void brief_examine_super_imsm(struct supertype *st, int verbose) |
cdddbdbc | 1356 | { |
27fd6274 | 1357 | /* We just write a generic IMSM ARRAY entry */ |
ff54de6e N |
1358 | struct mdinfo info; |
1359 | char nbuf[64]; | |
1e7bc0ed | 1360 | struct intel_super *super = st->sb; |
1e7bc0ed | 1361 | |
0d5a423f DW |
1362 | if (!super->anchor->num_raid_devs) { |
1363 | printf("ARRAY metadata=imsm\n"); | |
1e7bc0ed | 1364 | return; |
0d5a423f | 1365 | } |
ff54de6e | 1366 | |
a5d85af7 | 1367 | getinfo_super_imsm(st, &info, NULL); |
4737ae25 N |
1368 | fname_from_uuid(st, &info, nbuf, ':'); |
1369 | printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5); | |
1370 | } | |
1371 | ||
1372 | static void brief_examine_subarrays_imsm(struct supertype *st, int verbose) | |
1373 | { | |
1374 | /* We just write a generic IMSM ARRAY entry */ | |
1375 | struct mdinfo info; | |
1376 | char nbuf[64]; | |
1377 | char nbuf1[64]; | |
1378 | struct intel_super *super = st->sb; | |
1379 | int i; | |
1380 | ||
1381 | if (!super->anchor->num_raid_devs) | |
1382 | return; | |
1383 | ||
a5d85af7 | 1384 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1385 | fname_from_uuid(st, &info, nbuf, ':'); |
1e7bc0ed DW |
1386 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
1387 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
1388 | ||
1389 | super->current_vol = i; | |
a5d85af7 | 1390 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1391 | fname_from_uuid(st, &info, nbuf1, ':'); |
1124b3cf | 1392 | printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n", |
cf8de691 | 1393 | dev->volume, nbuf + 5, i, nbuf1 + 5); |
1e7bc0ed | 1394 | } |
cdddbdbc DW |
1395 | } |
1396 | ||
9d84c8ea DW |
1397 | static void export_examine_super_imsm(struct supertype *st) |
1398 | { | |
1399 | struct intel_super *super = st->sb; | |
1400 | struct imsm_super *mpb = super->anchor; | |
1401 | struct mdinfo info; | |
1402 | char nbuf[64]; | |
1403 | ||
a5d85af7 | 1404 | getinfo_super_imsm(st, &info, NULL); |
9d84c8ea DW |
1405 | fname_from_uuid(st, &info, nbuf, ':'); |
1406 | printf("MD_METADATA=imsm\n"); | |
1407 | printf("MD_LEVEL=container\n"); | |
1408 | printf("MD_UUID=%s\n", nbuf+5); | |
1409 | printf("MD_DEVICES=%u\n", mpb->num_disks); | |
1410 | } | |
1411 | ||
cdddbdbc DW |
1412 | static void detail_super_imsm(struct supertype *st, char *homehost) |
1413 | { | |
3ebe00a1 DW |
1414 | struct mdinfo info; |
1415 | char nbuf[64]; | |
1416 | ||
a5d85af7 | 1417 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1418 | fname_from_uuid(st, &info, nbuf, ':'); |
3ebe00a1 | 1419 | printf("\n UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
1420 | } |
1421 | ||
1422 | static void brief_detail_super_imsm(struct supertype *st) | |
1423 | { | |
ff54de6e N |
1424 | struct mdinfo info; |
1425 | char nbuf[64]; | |
a5d85af7 | 1426 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1427 | fname_from_uuid(st, &info, nbuf, ':'); |
ff54de6e | 1428 | printf(" UUID=%s", nbuf + 5); |
cdddbdbc | 1429 | } |
d665cc31 DW |
1430 | |
1431 | static int imsm_read_serial(int fd, char *devname, __u8 *serial); | |
1432 | static void fd2devname(int fd, char *name); | |
1433 | ||
120dc887 | 1434 | static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose) |
d665cc31 | 1435 | { |
120dc887 LM |
1436 | /* dump an unsorted list of devices attached to AHCI Intel storage |
1437 | * controller, as well as non-connected ports | |
d665cc31 DW |
1438 | */ |
1439 | int hba_len = strlen(hba_path) + 1; | |
1440 | struct dirent *ent; | |
1441 | DIR *dir; | |
1442 | char *path = NULL; | |
1443 | int err = 0; | |
1444 | unsigned long port_mask = (1 << port_count) - 1; | |
1445 | ||
f21e18ca | 1446 | if (port_count > (int)sizeof(port_mask) * 8) { |
d665cc31 DW |
1447 | if (verbose) |
1448 | fprintf(stderr, Name ": port_count %d out of range\n", port_count); | |
1449 | return 2; | |
1450 | } | |
1451 | ||
1452 | /* scroll through /sys/dev/block looking for devices attached to | |
1453 | * this hba | |
1454 | */ | |
1455 | dir = opendir("/sys/dev/block"); | |
1456 | for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) { | |
1457 | int fd; | |
1458 | char model[64]; | |
1459 | char vendor[64]; | |
1460 | char buf[1024]; | |
1461 | int major, minor; | |
1462 | char *device; | |
1463 | char *c; | |
1464 | int port; | |
1465 | int type; | |
1466 | ||
1467 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
1468 | continue; | |
1469 | path = devt_to_devpath(makedev(major, minor)); | |
1470 | if (!path) | |
1471 | continue; | |
1472 | if (!path_attached_to_hba(path, hba_path)) { | |
1473 | free(path); | |
1474 | path = NULL; | |
1475 | continue; | |
1476 | } | |
1477 | ||
1478 | /* retrieve the scsi device type */ | |
1479 | if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) { | |
1480 | if (verbose) | |
1481 | fprintf(stderr, Name ": failed to allocate 'device'\n"); | |
1482 | err = 2; | |
1483 | break; | |
1484 | } | |
1485 | sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor); | |
1486 | if (load_sys(device, buf) != 0) { | |
1487 | if (verbose) | |
1488 | fprintf(stderr, Name ": failed to read device type for %s\n", | |
1489 | path); | |
1490 | err = 2; | |
1491 | free(device); | |
1492 | break; | |
1493 | } | |
1494 | type = strtoul(buf, NULL, 10); | |
1495 | ||
1496 | /* if it's not a disk print the vendor and model */ | |
1497 | if (!(type == 0 || type == 7 || type == 14)) { | |
1498 | vendor[0] = '\0'; | |
1499 | model[0] = '\0'; | |
1500 | sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor); | |
1501 | if (load_sys(device, buf) == 0) { | |
1502 | strncpy(vendor, buf, sizeof(vendor)); | |
1503 | vendor[sizeof(vendor) - 1] = '\0'; | |
1504 | c = (char *) &vendor[sizeof(vendor) - 1]; | |
1505 | while (isspace(*c) || *c == '\0') | |
1506 | *c-- = '\0'; | |
1507 | ||
1508 | } | |
1509 | sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor); | |
1510 | if (load_sys(device, buf) == 0) { | |
1511 | strncpy(model, buf, sizeof(model)); | |
1512 | model[sizeof(model) - 1] = '\0'; | |
1513 | c = (char *) &model[sizeof(model) - 1]; | |
1514 | while (isspace(*c) || *c == '\0') | |
1515 | *c-- = '\0'; | |
1516 | } | |
1517 | ||
1518 | if (vendor[0] && model[0]) | |
1519 | sprintf(buf, "%.64s %.64s", vendor, model); | |
1520 | else | |
1521 | switch (type) { /* numbers from hald/linux/device.c */ | |
1522 | case 1: sprintf(buf, "tape"); break; | |
1523 | case 2: sprintf(buf, "printer"); break; | |
1524 | case 3: sprintf(buf, "processor"); break; | |
1525 | case 4: | |
1526 | case 5: sprintf(buf, "cdrom"); break; | |
1527 | case 6: sprintf(buf, "scanner"); break; | |
1528 | case 8: sprintf(buf, "media_changer"); break; | |
1529 | case 9: sprintf(buf, "comm"); break; | |
1530 | case 12: sprintf(buf, "raid"); break; | |
1531 | default: sprintf(buf, "unknown"); | |
1532 | } | |
1533 | } else | |
1534 | buf[0] = '\0'; | |
1535 | free(device); | |
1536 | ||
1537 | /* chop device path to 'host%d' and calculate the port number */ | |
1538 | c = strchr(&path[hba_len], '/'); | |
4e5e717d AW |
1539 | if (!c) { |
1540 | if (verbose) | |
1541 | fprintf(stderr, Name ": %s - invalid path name\n", path + hba_len); | |
1542 | err = 2; | |
1543 | break; | |
1544 | } | |
d665cc31 DW |
1545 | *c = '\0'; |
1546 | if (sscanf(&path[hba_len], "host%d", &port) == 1) | |
1547 | port -= host_base; | |
1548 | else { | |
1549 | if (verbose) { | |
1550 | *c = '/'; /* repair the full string */ | |
1551 | fprintf(stderr, Name ": failed to determine port number for %s\n", | |
1552 | path); | |
1553 | } | |
1554 | err = 2; | |
1555 | break; | |
1556 | } | |
1557 | ||
1558 | /* mark this port as used */ | |
1559 | port_mask &= ~(1 << port); | |
1560 | ||
1561 | /* print out the device information */ | |
1562 | if (buf[0]) { | |
1563 | printf(" Port%d : - non-disk device (%s) -\n", port, buf); | |
1564 | continue; | |
1565 | } | |
1566 | ||
1567 | fd = dev_open(ent->d_name, O_RDONLY); | |
1568 | if (fd < 0) | |
1569 | printf(" Port%d : - disk info unavailable -\n", port); | |
1570 | else { | |
1571 | fd2devname(fd, buf); | |
1572 | printf(" Port%d : %s", port, buf); | |
1573 | if (imsm_read_serial(fd, NULL, (__u8 *) buf) == 0) | |
664d5325 | 1574 | printf(" (%.*s)\n", MAX_RAID_SERIAL_LEN, buf); |
d665cc31 | 1575 | else |
664d5325 | 1576 | printf(" ()\n"); |
d665cc31 DW |
1577 | } |
1578 | close(fd); | |
1579 | free(path); | |
1580 | path = NULL; | |
1581 | } | |
1582 | if (path) | |
1583 | free(path); | |
1584 | if (dir) | |
1585 | closedir(dir); | |
1586 | if (err == 0) { | |
1587 | int i; | |
1588 | ||
1589 | for (i = 0; i < port_count; i++) | |
1590 | if (port_mask & (1 << i)) | |
1591 | printf(" Port%d : - no device attached -\n", i); | |
1592 | } | |
1593 | ||
1594 | return err; | |
1595 | } | |
1596 | ||
120dc887 LM |
1597 | static void print_found_intel_controllers(struct sys_dev *elem) |
1598 | { | |
1599 | for (; elem; elem = elem->next) { | |
1600 | fprintf(stderr, Name ": found Intel(R) "); | |
1601 | if (elem->type == SYS_DEV_SATA) | |
1602 | fprintf(stderr, "SATA "); | |
155cbb4c LM |
1603 | else if (elem->type == SYS_DEV_SAS) |
1604 | fprintf(stderr, "SAS "); | |
120dc887 LM |
1605 | fprintf(stderr, "RAID controller"); |
1606 | if (elem->pci_id) | |
1607 | fprintf(stderr, " at %s", elem->pci_id); | |
1608 | fprintf(stderr, ".\n"); | |
1609 | } | |
1610 | fflush(stderr); | |
1611 | } | |
1612 | ||
120dc887 LM |
1613 | static int ahci_get_port_count(const char *hba_path, int *port_count) |
1614 | { | |
1615 | struct dirent *ent; | |
1616 | DIR *dir; | |
1617 | int host_base = -1; | |
1618 | ||
1619 | *port_count = 0; | |
1620 | if ((dir = opendir(hba_path)) == NULL) | |
1621 | return -1; | |
1622 | ||
1623 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
1624 | int host; | |
1625 | ||
1626 | if (sscanf(ent->d_name, "host%d", &host) != 1) | |
1627 | continue; | |
1628 | if (*port_count == 0) | |
1629 | host_base = host; | |
1630 | else if (host < host_base) | |
1631 | host_base = host; | |
1632 | ||
1633 | if (host + 1 > *port_count + host_base) | |
1634 | *port_count = host + 1 - host_base; | |
1635 | } | |
1636 | closedir(dir); | |
1637 | return host_base; | |
1638 | } | |
1639 | ||
a891a3c2 LM |
1640 | static void print_imsm_capability(const struct imsm_orom *orom) |
1641 | { | |
1642 | printf(" Platform : Intel(R) Matrix Storage Manager\n"); | |
1643 | printf(" Version : %d.%d.%d.%d\n", orom->major_ver, orom->minor_ver, | |
1644 | orom->hotfix_ver, orom->build); | |
1645 | printf(" RAID Levels :%s%s%s%s%s\n", | |
1646 | imsm_orom_has_raid0(orom) ? " raid0" : "", | |
1647 | imsm_orom_has_raid1(orom) ? " raid1" : "", | |
1648 | imsm_orom_has_raid1e(orom) ? " raid1e" : "", | |
1649 | imsm_orom_has_raid10(orom) ? " raid10" : "", | |
1650 | imsm_orom_has_raid5(orom) ? " raid5" : ""); | |
1651 | printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", | |
1652 | imsm_orom_has_chunk(orom, 2) ? " 2k" : "", | |
1653 | imsm_orom_has_chunk(orom, 4) ? " 4k" : "", | |
1654 | imsm_orom_has_chunk(orom, 8) ? " 8k" : "", | |
1655 | imsm_orom_has_chunk(orom, 16) ? " 16k" : "", | |
1656 | imsm_orom_has_chunk(orom, 32) ? " 32k" : "", | |
1657 | imsm_orom_has_chunk(orom, 64) ? " 64k" : "", | |
1658 | imsm_orom_has_chunk(orom, 128) ? " 128k" : "", | |
1659 | imsm_orom_has_chunk(orom, 256) ? " 256k" : "", | |
1660 | imsm_orom_has_chunk(orom, 512) ? " 512k" : "", | |
1661 | imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "", | |
1662 | imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "", | |
1663 | imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "", | |
1664 | imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "", | |
1665 | imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "", | |
1666 | imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "", | |
1667 | imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : ""); | |
1668 | printf(" Max Disks : %d\n", orom->tds); | |
1669 | printf(" Max Volumes : %d\n", orom->vpa); | |
1670 | return; | |
1671 | } | |
1672 | ||
5615172f | 1673 | static int detail_platform_imsm(int verbose, int enumerate_only) |
d665cc31 DW |
1674 | { |
1675 | /* There are two components to imsm platform support, the ahci SATA | |
1676 | * controller and the option-rom. To find the SATA controller we | |
1677 | * simply look in /sys/bus/pci/drivers/ahci to see if an ahci | |
1678 | * controller with the Intel vendor id is present. This approach | |
1679 | * allows mdadm to leverage the kernel's ahci detection logic, with the | |
1680 | * caveat that if ahci.ko is not loaded mdadm will not be able to | |
1681 | * detect platform raid capabilities. The option-rom resides in a | |
1682 | * platform "Adapter ROM". We scan for its signature to retrieve the | |
1683 | * platform capabilities. If raid support is disabled in the BIOS the | |
1684 | * option-rom capability structure will not be available. | |
1685 | */ | |
1686 | const struct imsm_orom *orom; | |
1687 | struct sys_dev *list, *hba; | |
d665cc31 DW |
1688 | int host_base = 0; |
1689 | int port_count = 0; | |
120dc887 | 1690 | int result=0; |
d665cc31 | 1691 | |
5615172f | 1692 | if (enumerate_only) { |
a891a3c2 | 1693 | if (check_env("IMSM_NO_PLATFORM")) |
5615172f | 1694 | return 0; |
a891a3c2 LM |
1695 | list = find_intel_devices(); |
1696 | if (!list) | |
1697 | return 2; | |
1698 | for (hba = list; hba; hba = hba->next) { | |
1699 | orom = find_imsm_capability(hba->type); | |
1700 | if (!orom) { | |
1701 | result = 2; | |
1702 | break; | |
1703 | } | |
1704 | } | |
1705 | free_sys_dev(&list); | |
1706 | return result; | |
5615172f DW |
1707 | } |
1708 | ||
155cbb4c LM |
1709 | list = find_intel_devices(); |
1710 | if (!list) { | |
d665cc31 | 1711 | if (verbose) |
155cbb4c LM |
1712 | fprintf(stderr, Name ": no active Intel(R) RAID " |
1713 | "controller found.\n"); | |
d665cc31 DW |
1714 | free_sys_dev(&list); |
1715 | return 2; | |
1716 | } else if (verbose) | |
155cbb4c | 1717 | print_found_intel_controllers(list); |
d665cc31 | 1718 | |
a891a3c2 LM |
1719 | for (hba = list; hba; hba = hba->next) { |
1720 | orom = find_imsm_capability(hba->type); | |
1721 | if (!orom) | |
1722 | fprintf(stderr, Name ": imsm capabilities not found for controller: %s (type %s)\n", | |
1723 | hba->path, get_sys_dev_type(hba->type)); | |
1724 | else | |
1725 | print_imsm_capability(orom); | |
d665cc31 DW |
1726 | } |
1727 | ||
120dc887 LM |
1728 | for (hba = list; hba; hba = hba->next) { |
1729 | printf(" I/O Controller : %s (%s)\n", | |
1730 | hba->path, get_sys_dev_type(hba->type)); | |
d665cc31 | 1731 | |
120dc887 LM |
1732 | if (hba->type == SYS_DEV_SATA) { |
1733 | host_base = ahci_get_port_count(hba->path, &port_count); | |
1734 | if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) { | |
1735 | if (verbose) | |
1736 | fprintf(stderr, Name ": failed to enumerate " | |
1737 | "ports on SATA controller at %s.", hba->pci_id); | |
1738 | result |= 2; | |
1739 | } | |
1740 | } | |
d665cc31 | 1741 | } |
155cbb4c | 1742 | |
120dc887 LM |
1743 | free_sys_dev(&list); |
1744 | return result; | |
d665cc31 | 1745 | } |
cdddbdbc DW |
1746 | #endif |
1747 | ||
1748 | static int match_home_imsm(struct supertype *st, char *homehost) | |
1749 | { | |
5115ca67 DW |
1750 | /* the imsm metadata format does not specify any host |
1751 | * identification information. We return -1 since we can never | |
1752 | * confirm nor deny whether a given array is "meant" for this | |
148acb7b | 1753 | * host. We rely on compare_super and the 'family_num' fields to |
5115ca67 DW |
1754 | * exclude member disks that do not belong, and we rely on |
1755 | * mdadm.conf to specify the arrays that should be assembled. | |
1756 | * Auto-assembly may still pick up "foreign" arrays. | |
1757 | */ | |
cdddbdbc | 1758 | |
9362c1c8 | 1759 | return -1; |
cdddbdbc DW |
1760 | } |
1761 | ||
1762 | static void uuid_from_super_imsm(struct supertype *st, int uuid[4]) | |
1763 | { | |
51006d85 N |
1764 | /* The uuid returned here is used for: |
1765 | * uuid to put into bitmap file (Create, Grow) | |
1766 | * uuid for backup header when saving critical section (Grow) | |
1767 | * comparing uuids when re-adding a device into an array | |
1768 | * In these cases the uuid required is that of the data-array, | |
1769 | * not the device-set. | |
1770 | * uuid to recognise same set when adding a missing device back | |
1771 | * to an array. This is a uuid for the device-set. | |
1772 | * | |
1773 | * For each of these we can make do with a truncated | |
1774 | * or hashed uuid rather than the original, as long as | |
1775 | * everyone agrees. | |
1776 | * In each case the uuid required is that of the data-array, | |
1777 | * not the device-set. | |
43dad3d6 | 1778 | */ |
51006d85 N |
1779 | /* imsm does not track uuid's so we synthesis one using sha1 on |
1780 | * - The signature (Which is constant for all imsm array, but no matter) | |
148acb7b | 1781 | * - the orig_family_num of the container |
51006d85 N |
1782 | * - the index number of the volume |
1783 | * - the 'serial' number of the volume. | |
1784 | * Hopefully these are all constant. | |
1785 | */ | |
1786 | struct intel_super *super = st->sb; | |
43dad3d6 | 1787 | |
51006d85 N |
1788 | char buf[20]; |
1789 | struct sha1_ctx ctx; | |
1790 | struct imsm_dev *dev = NULL; | |
148acb7b | 1791 | __u32 family_num; |
51006d85 | 1792 | |
148acb7b DW |
1793 | /* some mdadm versions failed to set ->orig_family_num, in which |
1794 | * case fall back to ->family_num. orig_family_num will be | |
1795 | * fixed up with the first metadata update. | |
1796 | */ | |
1797 | family_num = super->anchor->orig_family_num; | |
1798 | if (family_num == 0) | |
1799 | family_num = super->anchor->family_num; | |
51006d85 | 1800 | sha1_init_ctx(&ctx); |
92bd8f8d | 1801 | sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx); |
148acb7b | 1802 | sha1_process_bytes(&family_num, sizeof(__u32), &ctx); |
51006d85 N |
1803 | if (super->current_vol >= 0) |
1804 | dev = get_imsm_dev(super, super->current_vol); | |
1805 | if (dev) { | |
1806 | __u32 vol = super->current_vol; | |
1807 | sha1_process_bytes(&vol, sizeof(vol), &ctx); | |
1808 | sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx); | |
1809 | } | |
1810 | sha1_finish_ctx(&ctx, buf); | |
1811 | memcpy(uuid, buf, 4*4); | |
cdddbdbc DW |
1812 | } |
1813 | ||
0d481d37 | 1814 | #if 0 |
4f5bc454 DW |
1815 | static void |
1816 | get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p) | |
cdddbdbc | 1817 | { |
cdddbdbc DW |
1818 | __u8 *v = get_imsm_version(mpb); |
1819 | __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH; | |
1820 | char major[] = { 0, 0, 0 }; | |
1821 | char minor[] = { 0 ,0, 0 }; | |
1822 | char patch[] = { 0, 0, 0 }; | |
1823 | char *ver_parse[] = { major, minor, patch }; | |
1824 | int i, j; | |
1825 | ||
1826 | i = j = 0; | |
1827 | while (*v != '\0' && v < end) { | |
1828 | if (*v != '.' && j < 2) | |
1829 | ver_parse[i][j++] = *v; | |
1830 | else { | |
1831 | i++; | |
1832 | j = 0; | |
1833 | } | |
1834 | v++; | |
1835 | } | |
1836 | ||
4f5bc454 DW |
1837 | *m = strtol(minor, NULL, 0); |
1838 | *p = strtol(patch, NULL, 0); | |
1839 | } | |
0d481d37 | 1840 | #endif |
4f5bc454 | 1841 | |
1e5c6983 DW |
1842 | static __u32 migr_strip_blocks_resync(struct imsm_dev *dev) |
1843 | { | |
1844 | /* migr_strip_size when repairing or initializing parity */ | |
1845 | struct imsm_map *map = get_imsm_map(dev, 0); | |
1846 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); | |
1847 | ||
1848 | switch (get_imsm_raid_level(map)) { | |
1849 | case 5: | |
1850 | case 10: | |
1851 | return chunk; | |
1852 | default: | |
1853 | return 128*1024 >> 9; | |
1854 | } | |
1855 | } | |
1856 | ||
1857 | static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev) | |
1858 | { | |
1859 | /* migr_strip_size when rebuilding a degraded disk, no idea why | |
1860 | * this is different than migr_strip_size_resync(), but it's good | |
1861 | * to be compatible | |
1862 | */ | |
1863 | struct imsm_map *map = get_imsm_map(dev, 1); | |
1864 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); | |
1865 | ||
1866 | switch (get_imsm_raid_level(map)) { | |
1867 | case 1: | |
1868 | case 10: | |
1869 | if (map->num_members % map->num_domains == 0) | |
1870 | return 128*1024 >> 9; | |
1871 | else | |
1872 | return chunk; | |
1873 | case 5: | |
1874 | return max((__u32) 64*1024 >> 9, chunk); | |
1875 | default: | |
1876 | return 128*1024 >> 9; | |
1877 | } | |
1878 | } | |
1879 | ||
1880 | static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev) | |
1881 | { | |
1882 | struct imsm_map *lo = get_imsm_map(dev, 0); | |
1883 | struct imsm_map *hi = get_imsm_map(dev, 1); | |
1884 | __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip); | |
1885 | __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip); | |
1886 | ||
1887 | return max((__u32) 1, hi_chunk / lo_chunk); | |
1888 | } | |
1889 | ||
1890 | static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev) | |
1891 | { | |
1892 | struct imsm_map *lo = get_imsm_map(dev, 0); | |
1893 | int level = get_imsm_raid_level(lo); | |
1894 | ||
1895 | if (level == 1 || level == 10) { | |
1896 | struct imsm_map *hi = get_imsm_map(dev, 1); | |
1897 | ||
1898 | return hi->num_domains; | |
1899 | } else | |
1900 | return num_stripes_per_unit_resync(dev); | |
1901 | } | |
1902 | ||
98130f40 | 1903 | static __u8 imsm_num_data_members(struct imsm_dev *dev, int second_map) |
1e5c6983 DW |
1904 | { |
1905 | /* named 'imsm_' because raid0, raid1 and raid10 | |
1906 | * counter-intuitively have the same number of data disks | |
1907 | */ | |
98130f40 | 1908 | struct imsm_map *map = get_imsm_map(dev, second_map); |
1e5c6983 DW |
1909 | |
1910 | switch (get_imsm_raid_level(map)) { | |
1911 | case 0: | |
1912 | case 1: | |
1913 | case 10: | |
1914 | return map->num_members; | |
1915 | case 5: | |
1916 | return map->num_members - 1; | |
1917 | default: | |
1918 | dprintf("%s: unsupported raid level\n", __func__); | |
1919 | return 0; | |
1920 | } | |
1921 | } | |
1922 | ||
1923 | static __u32 parity_segment_depth(struct imsm_dev *dev) | |
1924 | { | |
1925 | struct imsm_map *map = get_imsm_map(dev, 0); | |
1926 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); | |
1927 | ||
1928 | switch(get_imsm_raid_level(map)) { | |
1929 | case 1: | |
1930 | case 10: | |
1931 | return chunk * map->num_domains; | |
1932 | case 5: | |
1933 | return chunk * map->num_members; | |
1934 | default: | |
1935 | return chunk; | |
1936 | } | |
1937 | } | |
1938 | ||
1939 | static __u32 map_migr_block(struct imsm_dev *dev, __u32 block) | |
1940 | { | |
1941 | struct imsm_map *map = get_imsm_map(dev, 1); | |
1942 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); | |
1943 | __u32 strip = block / chunk; | |
1944 | ||
1945 | switch (get_imsm_raid_level(map)) { | |
1946 | case 1: | |
1947 | case 10: { | |
1948 | __u32 vol_strip = (strip * map->num_domains) + 1; | |
1949 | __u32 vol_stripe = vol_strip / map->num_members; | |
1950 | ||
1951 | return vol_stripe * chunk + block % chunk; | |
1952 | } case 5: { | |
1953 | __u32 stripe = strip / (map->num_members - 1); | |
1954 | ||
1955 | return stripe * chunk + block % chunk; | |
1956 | } | |
1957 | default: | |
1958 | return 0; | |
1959 | } | |
1960 | } | |
1961 | ||
c47b0ff6 AK |
1962 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
1963 | struct imsm_dev *dev) | |
1e5c6983 DW |
1964 | { |
1965 | /* calculate the conversion factor between per member 'blocks' | |
1966 | * (md/{resync,rebuild}_start) and imsm migration units, return | |
1967 | * 0 for the 'not migrating' and 'unsupported migration' cases | |
1968 | */ | |
1969 | if (!dev->vol.migr_state) | |
1970 | return 0; | |
1971 | ||
1972 | switch (migr_type(dev)) { | |
c47b0ff6 AK |
1973 | case MIGR_GEN_MIGR: { |
1974 | struct migr_record *migr_rec = super->migr_rec; | |
1975 | return __le32_to_cpu(migr_rec->blocks_per_unit); | |
1976 | } | |
1e5c6983 DW |
1977 | case MIGR_VERIFY: |
1978 | case MIGR_REPAIR: | |
1979 | case MIGR_INIT: { | |
1980 | struct imsm_map *map = get_imsm_map(dev, 0); | |
1981 | __u32 stripes_per_unit; | |
1982 | __u32 blocks_per_unit; | |
1983 | __u32 parity_depth; | |
1984 | __u32 migr_chunk; | |
1985 | __u32 block_map; | |
1986 | __u32 block_rel; | |
1987 | __u32 segment; | |
1988 | __u32 stripe; | |
1989 | __u8 disks; | |
1990 | ||
1991 | /* yes, this is really the translation of migr_units to | |
1992 | * per-member blocks in the 'resync' case | |
1993 | */ | |
1994 | stripes_per_unit = num_stripes_per_unit_resync(dev); | |
1995 | migr_chunk = migr_strip_blocks_resync(dev); | |
98130f40 | 1996 | disks = imsm_num_data_members(dev, 0); |
1e5c6983 | 1997 | blocks_per_unit = stripes_per_unit * migr_chunk * disks; |
7b1ab482 | 1998 | stripe = __le16_to_cpu(map->blocks_per_strip) * disks; |
1e5c6983 DW |
1999 | segment = blocks_per_unit / stripe; |
2000 | block_rel = blocks_per_unit - segment * stripe; | |
2001 | parity_depth = parity_segment_depth(dev); | |
2002 | block_map = map_migr_block(dev, block_rel); | |
2003 | return block_map + parity_depth * segment; | |
2004 | } | |
2005 | case MIGR_REBUILD: { | |
2006 | __u32 stripes_per_unit; | |
2007 | __u32 migr_chunk; | |
2008 | ||
2009 | stripes_per_unit = num_stripes_per_unit_rebuild(dev); | |
2010 | migr_chunk = migr_strip_blocks_rebuild(dev); | |
2011 | return migr_chunk * stripes_per_unit; | |
2012 | } | |
1e5c6983 DW |
2013 | case MIGR_STATE_CHANGE: |
2014 | default: | |
2015 | return 0; | |
2016 | } | |
2017 | } | |
2018 | ||
c2c087e6 DW |
2019 | static int imsm_level_to_layout(int level) |
2020 | { | |
2021 | switch (level) { | |
2022 | case 0: | |
2023 | case 1: | |
2024 | return 0; | |
2025 | case 5: | |
2026 | case 6: | |
a380c027 | 2027 | return ALGORITHM_LEFT_ASYMMETRIC; |
c2c087e6 | 2028 | case 10: |
c92a2527 | 2029 | return 0x102; |
c2c087e6 | 2030 | } |
a18a888e | 2031 | return UnSet; |
c2c087e6 DW |
2032 | } |
2033 | ||
8e59f3d8 AK |
2034 | /******************************************************************************* |
2035 | * Function: read_imsm_migr_rec | |
2036 | * Description: Function reads imsm migration record from last sector of disk | |
2037 | * Parameters: | |
2038 | * fd : disk descriptor | |
2039 | * super : metadata info | |
2040 | * Returns: | |
2041 | * 0 : success, | |
2042 | * -1 : fail | |
2043 | ******************************************************************************/ | |
2044 | static int read_imsm_migr_rec(int fd, struct intel_super *super) | |
2045 | { | |
2046 | int ret_val = -1; | |
2047 | unsigned long long dsize; | |
2048 | ||
2049 | get_dev_size(fd, NULL, &dsize); | |
2050 | if (lseek64(fd, dsize - 512, SEEK_SET) < 0) { | |
2051 | fprintf(stderr, | |
2052 | Name ": Cannot seek to anchor block: %s\n", | |
2053 | strerror(errno)); | |
2054 | goto out; | |
2055 | } | |
2056 | if (read(fd, super->migr_rec_buf, 512) != 512) { | |
2057 | fprintf(stderr, | |
2058 | Name ": Cannot read migr record block: %s\n", | |
2059 | strerror(errno)); | |
2060 | goto out; | |
2061 | } | |
2062 | ret_val = 0; | |
2063 | ||
2064 | out: | |
2065 | return ret_val; | |
2066 | } | |
2067 | ||
2068 | /******************************************************************************* | |
2069 | * Function: load_imsm_migr_rec | |
2070 | * Description: Function reads imsm migration record (it is stored at the last | |
2071 | * sector of disk) | |
2072 | * Parameters: | |
2073 | * super : imsm internal array info | |
2074 | * info : general array info | |
2075 | * Returns: | |
2076 | * 0 : success | |
2077 | * -1 : fail | |
2078 | ******************************************************************************/ | |
2079 | static int load_imsm_migr_rec(struct intel_super *super, struct mdinfo *info) | |
2080 | { | |
2081 | struct mdinfo *sd; | |
2082 | struct dl *dl = NULL; | |
2083 | char nm[30]; | |
2084 | int retval = -1; | |
2085 | int fd = -1; | |
2086 | ||
2087 | if (info) { | |
2088 | for (sd = info->devs ; sd ; sd = sd->next) { | |
2089 | /* read only from one of the first two slots */ | |
2090 | if ((sd->disk.raid_disk > 1) || | |
2091 | (sd->disk.raid_disk < 0)) | |
2092 | continue; | |
2093 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
2094 | fd = dev_open(nm, O_RDONLY); | |
2095 | if (fd >= 0) | |
2096 | break; | |
2097 | } | |
2098 | } | |
2099 | if (fd < 0) { | |
2100 | for (dl = super->disks; dl; dl = dl->next) { | |
2101 | /* read only from one of the first two slots */ | |
2102 | if (dl->index > 1) | |
2103 | continue; | |
2104 | sprintf(nm, "%d:%d", dl->major, dl->minor); | |
2105 | fd = dev_open(nm, O_RDONLY); | |
2106 | if (fd >= 0) | |
2107 | break; | |
2108 | } | |
2109 | } | |
2110 | if (fd < 0) | |
2111 | goto out; | |
2112 | retval = read_imsm_migr_rec(fd, super); | |
2113 | ||
2114 | out: | |
2115 | if (fd >= 0) | |
2116 | close(fd); | |
2117 | return retval; | |
2118 | } | |
2119 | ||
9e2d750d | 2120 | #ifndef MDASSEMBLE |
c17608ea AK |
2121 | /******************************************************************************* |
2122 | * function: imsm_create_metadata_checkpoint_update | |
2123 | * Description: It creates update for checkpoint change. | |
2124 | * Parameters: | |
2125 | * super : imsm internal array info | |
2126 | * u : pointer to prepared update | |
2127 | * Returns: | |
2128 | * Uptate length. | |
2129 | * If length is equal to 0, input pointer u contains no update | |
2130 | ******************************************************************************/ | |
2131 | static int imsm_create_metadata_checkpoint_update( | |
2132 | struct intel_super *super, | |
2133 | struct imsm_update_general_migration_checkpoint **u) | |
2134 | { | |
2135 | ||
2136 | int update_memory_size = 0; | |
2137 | ||
2138 | dprintf("imsm_create_metadata_checkpoint_update(enter)\n"); | |
2139 | ||
2140 | if (u == NULL) | |
2141 | return 0; | |
2142 | *u = NULL; | |
2143 | ||
2144 | /* size of all update data without anchor */ | |
2145 | update_memory_size = | |
2146 | sizeof(struct imsm_update_general_migration_checkpoint); | |
2147 | ||
2148 | *u = calloc(1, update_memory_size); | |
2149 | if (*u == NULL) { | |
2150 | dprintf("error: cannot get memory for " | |
2151 | "imsm_create_metadata_checkpoint_update update\n"); | |
2152 | return 0; | |
2153 | } | |
2154 | (*u)->type = update_general_migration_checkpoint; | |
2155 | (*u)->curr_migr_unit = __le32_to_cpu(super->migr_rec->curr_migr_unit); | |
2156 | dprintf("imsm_create_metadata_checkpoint_update: prepared for %u\n", | |
2157 | (*u)->curr_migr_unit); | |
2158 | ||
2159 | return update_memory_size; | |
2160 | } | |
2161 | ||
2162 | ||
2163 | static void imsm_update_metadata_locally(struct supertype *st, | |
2164 | void *buf, int len); | |
2165 | ||
687629c2 AK |
2166 | /******************************************************************************* |
2167 | * Function: write_imsm_migr_rec | |
2168 | * Description: Function writes imsm migration record | |
2169 | * (at the last sector of disk) | |
2170 | * Parameters: | |
2171 | * super : imsm internal array info | |
2172 | * Returns: | |
2173 | * 0 : success | |
2174 | * -1 : if fail | |
2175 | ******************************************************************************/ | |
2176 | static int write_imsm_migr_rec(struct supertype *st) | |
2177 | { | |
2178 | struct intel_super *super = st->sb; | |
2179 | unsigned long long dsize; | |
2180 | char nm[30]; | |
2181 | int fd = -1; | |
2182 | int retval = -1; | |
2183 | struct dl *sd; | |
c17608ea AK |
2184 | int len; |
2185 | struct imsm_update_general_migration_checkpoint *u; | |
687629c2 AK |
2186 | |
2187 | for (sd = super->disks ; sd ; sd = sd->next) { | |
2188 | /* write to 2 first slots only */ | |
2189 | if ((sd->index < 0) || (sd->index > 1)) | |
2190 | continue; | |
2191 | sprintf(nm, "%d:%d", sd->major, sd->minor); | |
2192 | fd = dev_open(nm, O_RDWR); | |
2193 | if (fd < 0) | |
2194 | continue; | |
2195 | get_dev_size(fd, NULL, &dsize); | |
2196 | if (lseek64(fd, dsize - 512, SEEK_SET) < 0) { | |
2197 | fprintf(stderr, | |
2198 | Name ": Cannot seek to anchor block: %s\n", | |
2199 | strerror(errno)); | |
2200 | goto out; | |
2201 | } | |
2202 | if (write(fd, super->migr_rec_buf, 512) != 512) { | |
2203 | fprintf(stderr, | |
2204 | Name ": Cannot write migr record block: %s\n", | |
2205 | strerror(errno)); | |
2206 | goto out; | |
2207 | } | |
2208 | close(fd); | |
2209 | fd = -1; | |
2210 | } | |
c17608ea AK |
2211 | /* update checkpoint information in metadata */ |
2212 | len = imsm_create_metadata_checkpoint_update(super, &u); | |
2213 | ||
2214 | if (len <= 0) { | |
2215 | dprintf("imsm: Cannot prepare update\n"); | |
2216 | goto out; | |
2217 | } | |
2218 | /* update metadata locally */ | |
2219 | imsm_update_metadata_locally(st, u, len); | |
2220 | /* and possibly remotely */ | |
2221 | if (st->update_tail) { | |
2222 | append_metadata_update(st, u, len); | |
2223 | /* during reshape we do all work inside metadata handler | |
2224 | * manage_reshape(), so metadata update has to be triggered | |
2225 | * insida it | |
2226 | */ | |
2227 | flush_metadata_updates(st); | |
2228 | st->update_tail = &st->updates; | |
2229 | } else | |
2230 | free(u); | |
687629c2 AK |
2231 | |
2232 | retval = 0; | |
2233 | out: | |
2234 | if (fd >= 0) | |
2235 | close(fd); | |
2236 | return retval; | |
2237 | } | |
9e2d750d | 2238 | #endif /* MDASSEMBLE */ |
687629c2 | 2239 | |
e2962bfc AK |
2240 | /* spare/missing disks activations are not allowe when |
2241 | * array/container performs reshape operation, because | |
2242 | * all arrays in container works on the same disks set | |
2243 | */ | |
2244 | int imsm_reshape_blocks_arrays_changes(struct intel_super *super) | |
2245 | { | |
2246 | int rv = 0; | |
2247 | struct intel_dev *i_dev; | |
2248 | struct imsm_dev *dev; | |
2249 | ||
2250 | /* check whole container | |
2251 | */ | |
2252 | for (i_dev = super->devlist; i_dev; i_dev = i_dev->next) { | |
2253 | dev = i_dev->dev; | |
3ad25638 | 2254 | if (is_gen_migration(dev)) { |
e2962bfc AK |
2255 | /* No repair during any migration in container |
2256 | */ | |
2257 | rv = 1; | |
2258 | break; | |
2259 | } | |
2260 | } | |
2261 | return rv; | |
2262 | } | |
2263 | ||
a5d85af7 | 2264 | static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap) |
bf5a934a DW |
2265 | { |
2266 | struct intel_super *super = st->sb; | |
c47b0ff6 | 2267 | struct migr_record *migr_rec = super->migr_rec; |
949c47a0 | 2268 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); |
a965f303 | 2269 | struct imsm_map *map = get_imsm_map(dev, 0); |
81ac8b4d | 2270 | struct imsm_map *prev_map = get_imsm_map(dev, 1); |
b335e593 | 2271 | struct imsm_map *map_to_analyse = map; |
efb30e7f | 2272 | struct dl *dl; |
e207da2f | 2273 | char *devname; |
139dae11 | 2274 | unsigned int component_size_alligment; |
a5d85af7 | 2275 | int map_disks = info->array.raid_disks; |
bf5a934a | 2276 | |
95eeceeb | 2277 | memset(info, 0, sizeof(*info)); |
b335e593 AK |
2278 | if (prev_map) |
2279 | map_to_analyse = prev_map; | |
2280 | ||
ca0748fa | 2281 | dl = super->current_disk; |
9894ec0d | 2282 | |
bf5a934a | 2283 | info->container_member = super->current_vol; |
cd0430a1 | 2284 | info->array.raid_disks = map->num_members; |
b335e593 | 2285 | info->array.level = get_imsm_raid_level(map_to_analyse); |
bf5a934a DW |
2286 | info->array.layout = imsm_level_to_layout(info->array.level); |
2287 | info->array.md_minor = -1; | |
2288 | info->array.ctime = 0; | |
2289 | info->array.utime = 0; | |
b335e593 AK |
2290 | info->array.chunk_size = |
2291 | __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9; | |
301406c9 | 2292 | info->array.state = !dev->vol.dirty; |
da9b4a62 DW |
2293 | info->custom_array_size = __le32_to_cpu(dev->size_high); |
2294 | info->custom_array_size <<= 32; | |
2295 | info->custom_array_size |= __le32_to_cpu(dev->size_low); | |
3ad25638 AK |
2296 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
2297 | ||
3f83228a N |
2298 | if (prev_map && map->map_state == prev_map->map_state) { |
2299 | info->reshape_active = 1; | |
b335e593 AK |
2300 | info->new_level = get_imsm_raid_level(map); |
2301 | info->new_layout = imsm_level_to_layout(info->new_level); | |
2302 | info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9; | |
3f83228a | 2303 | info->delta_disks = map->num_members - prev_map->num_members; |
493f5dd6 N |
2304 | if (info->delta_disks) { |
2305 | /* this needs to be applied to every array | |
2306 | * in the container. | |
2307 | */ | |
2308 | info->reshape_active = 2; | |
2309 | } | |
3f83228a N |
2310 | /* We shape information that we give to md might have to be |
2311 | * modify to cope with md's requirement for reshaping arrays. | |
2312 | * For example, when reshaping a RAID0, md requires it to be | |
2313 | * presented as a degraded RAID4. | |
2314 | * Also if a RAID0 is migrating to a RAID5 we need to specify | |
2315 | * the array as already being RAID5, but the 'before' layout | |
2316 | * is a RAID4-like layout. | |
2317 | */ | |
2318 | switch (info->array.level) { | |
2319 | case 0: | |
2320 | switch(info->new_level) { | |
2321 | case 0: | |
2322 | /* conversion is happening as RAID4 */ | |
2323 | info->array.level = 4; | |
2324 | info->array.raid_disks += 1; | |
2325 | break; | |
2326 | case 5: | |
2327 | /* conversion is happening as RAID5 */ | |
2328 | info->array.level = 5; | |
2329 | info->array.layout = ALGORITHM_PARITY_N; | |
3f83228a N |
2330 | info->delta_disks -= 1; |
2331 | break; | |
2332 | default: | |
2333 | /* FIXME error message */ | |
2334 | info->array.level = UnSet; | |
2335 | break; | |
2336 | } | |
2337 | break; | |
2338 | } | |
b335e593 AK |
2339 | } else { |
2340 | info->new_level = UnSet; | |
2341 | info->new_layout = UnSet; | |
2342 | info->new_chunk = info->array.chunk_size; | |
3f83228a | 2343 | info->delta_disks = 0; |
b335e593 | 2344 | } |
ca0748fa | 2345 | |
efb30e7f DW |
2346 | if (dl) { |
2347 | info->disk.major = dl->major; | |
2348 | info->disk.minor = dl->minor; | |
ca0748fa | 2349 | info->disk.number = dl->index; |
656b6b5a N |
2350 | info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse, |
2351 | dl->index); | |
efb30e7f | 2352 | } |
bf5a934a | 2353 | |
b335e593 AK |
2354 | info->data_offset = __le32_to_cpu(map_to_analyse->pba_of_lba0); |
2355 | info->component_size = | |
2356 | __le32_to_cpu(map_to_analyse->blocks_per_member); | |
139dae11 AK |
2357 | |
2358 | /* check component size aligment | |
2359 | */ | |
2360 | component_size_alligment = | |
2361 | info->component_size % (info->array.chunk_size/512); | |
2362 | ||
2363 | if (component_size_alligment && | |
2364 | (info->array.level != 1) && (info->array.level != UnSet)) { | |
2365 | dprintf("imsm: reported component size alligned from %llu ", | |
2366 | info->component_size); | |
2367 | info->component_size -= component_size_alligment; | |
2368 | dprintf("to %llu (%i).\n", | |
2369 | info->component_size, component_size_alligment); | |
2370 | } | |
2371 | ||
301406c9 | 2372 | memset(info->uuid, 0, sizeof(info->uuid)); |
921d9e16 | 2373 | info->recovery_start = MaxSector; |
bf5a934a | 2374 | |
d2e6d5d6 | 2375 | info->reshape_progress = 0; |
b6796ce1 | 2376 | info->resync_start = MaxSector; |
b9172665 AK |
2377 | if ((map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED || |
2378 | dev->vol.dirty) && | |
2379 | imsm_reshape_blocks_arrays_changes(super) == 0) { | |
301406c9 | 2380 | info->resync_start = 0; |
b6796ce1 AK |
2381 | } |
2382 | if (dev->vol.migr_state) { | |
1e5c6983 DW |
2383 | switch (migr_type(dev)) { |
2384 | case MIGR_REPAIR: | |
2385 | case MIGR_INIT: { | |
c47b0ff6 AK |
2386 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
2387 | dev); | |
1e5c6983 DW |
2388 | __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit); |
2389 | ||
2390 | info->resync_start = blocks_per_unit * units; | |
2391 | break; | |
2392 | } | |
d2e6d5d6 | 2393 | case MIGR_GEN_MIGR: { |
c47b0ff6 AK |
2394 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
2395 | dev); | |
2396 | __u64 units = __le32_to_cpu(migr_rec->curr_migr_unit); | |
04fa9523 AK |
2397 | unsigned long long array_blocks; |
2398 | int used_disks; | |
d2e6d5d6 | 2399 | |
befb629b AK |
2400 | if (__le32_to_cpu(migr_rec->ascending_migr) && |
2401 | (units < | |
2402 | (__le32_to_cpu(migr_rec->num_migr_units)-1)) && | |
2403 | (super->migr_rec->rec_status == | |
2404 | __cpu_to_le32(UNIT_SRC_IN_CP_AREA))) | |
2405 | units++; | |
2406 | ||
d2e6d5d6 | 2407 | info->reshape_progress = blocks_per_unit * units; |
6289d1e0 | 2408 | |
d2e6d5d6 AK |
2409 | dprintf("IMSM: General Migration checkpoint : %llu " |
2410 | "(%llu) -> read reshape progress : %llu\n", | |
19986c72 MB |
2411 | (unsigned long long)units, |
2412 | (unsigned long long)blocks_per_unit, | |
2413 | info->reshape_progress); | |
75156c46 AK |
2414 | |
2415 | used_disks = imsm_num_data_members(dev, 1); | |
2416 | if (used_disks > 0) { | |
2417 | array_blocks = map->blocks_per_member * | |
2418 | used_disks; | |
2419 | /* round array size down to closest MB | |
2420 | */ | |
2421 | info->custom_array_size = (array_blocks | |
2422 | >> SECT_PER_MB_SHIFT) | |
2423 | << SECT_PER_MB_SHIFT; | |
2424 | } | |
d2e6d5d6 | 2425 | } |
1e5c6983 DW |
2426 | case MIGR_VERIFY: |
2427 | /* we could emulate the checkpointing of | |
2428 | * 'sync_action=check' migrations, but for now | |
2429 | * we just immediately complete them | |
2430 | */ | |
2431 | case MIGR_REBUILD: | |
2432 | /* this is handled by container_content_imsm() */ | |
1e5c6983 DW |
2433 | case MIGR_STATE_CHANGE: |
2434 | /* FIXME handle other migrations */ | |
2435 | default: | |
2436 | /* we are not dirty, so... */ | |
2437 | info->resync_start = MaxSector; | |
2438 | } | |
b6796ce1 | 2439 | } |
301406c9 DW |
2440 | |
2441 | strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN); | |
2442 | info->name[MAX_RAID_SERIAL_LEN] = 0; | |
bf5a934a | 2443 | |
f35f2525 N |
2444 | info->array.major_version = -1; |
2445 | info->array.minor_version = -2; | |
e207da2f AW |
2446 | devname = devnum2devname(st->container_dev); |
2447 | *info->text_version = '\0'; | |
2448 | if (devname) | |
2449 | sprintf(info->text_version, "/%s/%d", devname, info->container_member); | |
2450 | free(devname); | |
a67dd8cc | 2451 | info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */ |
51006d85 | 2452 | uuid_from_super_imsm(st, info->uuid); |
a5d85af7 N |
2453 | |
2454 | if (dmap) { | |
2455 | int i, j; | |
2456 | for (i=0; i<map_disks; i++) { | |
2457 | dmap[i] = 0; | |
2458 | if (i < info->array.raid_disks) { | |
2459 | struct imsm_disk *dsk; | |
98130f40 | 2460 | j = get_imsm_disk_idx(dev, i, -1); |
a5d85af7 N |
2461 | dsk = get_imsm_disk(super, j); |
2462 | if (dsk && (dsk->status & CONFIGURED_DISK)) | |
2463 | dmap[i] = 1; | |
2464 | } | |
2465 | } | |
2466 | } | |
81ac8b4d | 2467 | } |
bf5a934a | 2468 | |
97b4d0e9 DW |
2469 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, int failed); |
2470 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev); | |
2471 | ||
2472 | static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index) | |
2473 | { | |
2474 | struct dl *d; | |
2475 | ||
2476 | for (d = super->missing; d; d = d->next) | |
2477 | if (d->index == index) | |
2478 | return &d->disk; | |
2479 | return NULL; | |
2480 | } | |
2481 | ||
a5d85af7 | 2482 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map) |
4f5bc454 DW |
2483 | { |
2484 | struct intel_super *super = st->sb; | |
4f5bc454 | 2485 | struct imsm_disk *disk; |
a5d85af7 | 2486 | int map_disks = info->array.raid_disks; |
ab3cb6b3 N |
2487 | int max_enough = -1; |
2488 | int i; | |
2489 | struct imsm_super *mpb; | |
4f5bc454 | 2490 | |
bf5a934a | 2491 | if (super->current_vol >= 0) { |
a5d85af7 | 2492 | getinfo_super_imsm_volume(st, info, map); |
bf5a934a DW |
2493 | return; |
2494 | } | |
95eeceeb | 2495 | memset(info, 0, sizeof(*info)); |
d23fe947 DW |
2496 | |
2497 | /* Set raid_disks to zero so that Assemble will always pull in valid | |
2498 | * spares | |
2499 | */ | |
2500 | info->array.raid_disks = 0; | |
cdddbdbc DW |
2501 | info->array.level = LEVEL_CONTAINER; |
2502 | info->array.layout = 0; | |
2503 | info->array.md_minor = -1; | |
c2c087e6 | 2504 | info->array.ctime = 0; /* N/A for imsm */ |
cdddbdbc DW |
2505 | info->array.utime = 0; |
2506 | info->array.chunk_size = 0; | |
2507 | ||
2508 | info->disk.major = 0; | |
2509 | info->disk.minor = 0; | |
cdddbdbc | 2510 | info->disk.raid_disk = -1; |
c2c087e6 | 2511 | info->reshape_active = 0; |
f35f2525 N |
2512 | info->array.major_version = -1; |
2513 | info->array.minor_version = -2; | |
c2c087e6 | 2514 | strcpy(info->text_version, "imsm"); |
a67dd8cc | 2515 | info->safe_mode_delay = 0; |
c2c087e6 DW |
2516 | info->disk.number = -1; |
2517 | info->disk.state = 0; | |
c5afc314 | 2518 | info->name[0] = 0; |
921d9e16 | 2519 | info->recovery_start = MaxSector; |
3ad25638 | 2520 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
c2c087e6 | 2521 | |
97b4d0e9 | 2522 | /* do we have the all the insync disks that we expect? */ |
ab3cb6b3 | 2523 | mpb = super->anchor; |
97b4d0e9 | 2524 | |
ab3cb6b3 N |
2525 | for (i = 0; i < mpb->num_raid_devs; i++) { |
2526 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
2527 | int failed, enough, j, missing = 0; | |
2528 | struct imsm_map *map; | |
2529 | __u8 state; | |
97b4d0e9 | 2530 | |
ab3cb6b3 N |
2531 | failed = imsm_count_failed(super, dev); |
2532 | state = imsm_check_degraded(super, dev, failed); | |
2533 | map = get_imsm_map(dev, dev->vol.migr_state); | |
2534 | ||
2535 | /* any newly missing disks? | |
2536 | * (catches single-degraded vs double-degraded) | |
2537 | */ | |
2538 | for (j = 0; j < map->num_members; j++) { | |
98130f40 | 2539 | __u32 ord = get_imsm_ord_tbl_ent(dev, i, -1); |
ab3cb6b3 N |
2540 | __u32 idx = ord_to_idx(ord); |
2541 | ||
2542 | if (!(ord & IMSM_ORD_REBUILD) && | |
2543 | get_imsm_missing(super, idx)) { | |
2544 | missing = 1; | |
2545 | break; | |
2546 | } | |
97b4d0e9 | 2547 | } |
ab3cb6b3 N |
2548 | |
2549 | if (state == IMSM_T_STATE_FAILED) | |
2550 | enough = -1; | |
2551 | else if (state == IMSM_T_STATE_DEGRADED && | |
2552 | (state != map->map_state || missing)) | |
2553 | enough = 0; | |
2554 | else /* we're normal, or already degraded */ | |
2555 | enough = 1; | |
2556 | ||
2557 | /* in the missing/failed disk case check to see | |
2558 | * if at least one array is runnable | |
2559 | */ | |
2560 | max_enough = max(max_enough, enough); | |
2561 | } | |
2562 | dprintf("%s: enough: %d\n", __func__, max_enough); | |
2563 | info->container_enough = max_enough; | |
97b4d0e9 | 2564 | |
4a04ec6c | 2565 | if (super->disks) { |
14e8215b DW |
2566 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
2567 | ||
b9f594fe | 2568 | disk = &super->disks->disk; |
14e8215b DW |
2569 | info->data_offset = __le32_to_cpu(disk->total_blocks) - reserved; |
2570 | info->component_size = reserved; | |
25ed7e59 | 2571 | info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0; |
df474657 DW |
2572 | /* we don't change info->disk.raid_disk here because |
2573 | * this state will be finalized in mdmon after we have | |
2574 | * found the 'most fresh' version of the metadata | |
2575 | */ | |
25ed7e59 DW |
2576 | info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; |
2577 | info->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
cdddbdbc | 2578 | } |
a575e2a7 DW |
2579 | |
2580 | /* only call uuid_from_super_imsm when this disk is part of a populated container, | |
2581 | * ->compare_super may have updated the 'num_raid_devs' field for spares | |
2582 | */ | |
2583 | if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs) | |
36ba7d48 | 2584 | uuid_from_super_imsm(st, info->uuid); |
22e263f6 AC |
2585 | else |
2586 | memcpy(info->uuid, uuid_zero, sizeof(uuid_zero)); | |
a5d85af7 N |
2587 | |
2588 | /* I don't know how to compute 'map' on imsm, so use safe default */ | |
2589 | if (map) { | |
2590 | int i; | |
2591 | for (i = 0; i < map_disks; i++) | |
2592 | map[i] = 1; | |
2593 | } | |
2594 | ||
cdddbdbc DW |
2595 | } |
2596 | ||
5c4cd5da AC |
2597 | /* allocates memory and fills disk in mdinfo structure |
2598 | * for each disk in array */ | |
2599 | struct mdinfo *getinfo_super_disks_imsm(struct supertype *st) | |
2600 | { | |
2601 | struct mdinfo *mddev = NULL; | |
2602 | struct intel_super *super = st->sb; | |
2603 | struct imsm_disk *disk; | |
2604 | int count = 0; | |
2605 | struct dl *dl; | |
2606 | if (!super || !super->disks) | |
2607 | return NULL; | |
2608 | dl = super->disks; | |
2609 | mddev = malloc(sizeof(*mddev)); | |
2610 | if (!mddev) { | |
2611 | fprintf(stderr, Name ": Failed to allocate memory.\n"); | |
2612 | return NULL; | |
2613 | } | |
2614 | memset(mddev, 0, sizeof(*mddev)); | |
2615 | while (dl) { | |
2616 | struct mdinfo *tmp; | |
2617 | disk = &dl->disk; | |
2618 | tmp = malloc(sizeof(*tmp)); | |
2619 | if (!tmp) { | |
2620 | fprintf(stderr, Name ": Failed to allocate memory.\n"); | |
2621 | if (mddev) | |
2622 | sysfs_free(mddev); | |
2623 | return NULL; | |
2624 | } | |
2625 | memset(tmp, 0, sizeof(*tmp)); | |
2626 | if (mddev->devs) | |
2627 | tmp->next = mddev->devs; | |
2628 | mddev->devs = tmp; | |
2629 | tmp->disk.number = count++; | |
2630 | tmp->disk.major = dl->major; | |
2631 | tmp->disk.minor = dl->minor; | |
2632 | tmp->disk.state = is_configured(disk) ? | |
2633 | (1 << MD_DISK_ACTIVE) : 0; | |
2634 | tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; | |
2635 | tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
2636 | tmp->disk.raid_disk = -1; | |
2637 | dl = dl->next; | |
2638 | } | |
2639 | return mddev; | |
2640 | } | |
2641 | ||
cdddbdbc DW |
2642 | static int update_super_imsm(struct supertype *st, struct mdinfo *info, |
2643 | char *update, char *devname, int verbose, | |
2644 | int uuid_set, char *homehost) | |
2645 | { | |
f352c545 DW |
2646 | /* For 'assemble' and 'force' we need to return non-zero if any |
2647 | * change was made. For others, the return value is ignored. | |
2648 | * Update options are: | |
2649 | * force-one : This device looks a bit old but needs to be included, | |
2650 | * update age info appropriately. | |
2651 | * assemble: clear any 'faulty' flag to allow this device to | |
2652 | * be assembled. | |
2653 | * force-array: Array is degraded but being forced, mark it clean | |
2654 | * if that will be needed to assemble it. | |
2655 | * | |
2656 | * newdev: not used ???? | |
2657 | * grow: Array has gained a new device - this is currently for | |
2658 | * linear only | |
2659 | * resync: mark as dirty so a resync will happen. | |
2660 | * name: update the name - preserving the homehost | |
6e46bf34 | 2661 | * uuid: Change the uuid of the array to match watch is given |
f352c545 DW |
2662 | * |
2663 | * Following are not relevant for this imsm: | |
2664 | * sparc2.2 : update from old dodgey metadata | |
2665 | * super-minor: change the preferred_minor number | |
2666 | * summaries: update redundant counters. | |
f352c545 DW |
2667 | * homehost: update the recorded homehost |
2668 | * _reshape_progress: record new reshape_progress position. | |
2669 | */ | |
6e46bf34 DW |
2670 | int rv = 1; |
2671 | struct intel_super *super = st->sb; | |
2672 | struct imsm_super *mpb; | |
f352c545 | 2673 | |
6e46bf34 DW |
2674 | /* we can only update container info */ |
2675 | if (!super || super->current_vol >= 0 || !super->anchor) | |
2676 | return 1; | |
2677 | ||
2678 | mpb = super->anchor; | |
2679 | ||
2680 | if (strcmp(update, "uuid") == 0 && uuid_set && !info->update_private) | |
1e2b2765 | 2681 | rv = -1; |
6e46bf34 DW |
2682 | else if (strcmp(update, "uuid") == 0 && uuid_set && info->update_private) { |
2683 | mpb->orig_family_num = *((__u32 *) info->update_private); | |
2684 | rv = 0; | |
2685 | } else if (strcmp(update, "uuid") == 0) { | |
2686 | __u32 *new_family = malloc(sizeof(*new_family)); | |
2687 | ||
2688 | /* update orig_family_number with the incoming random | |
2689 | * data, report the new effective uuid, and store the | |
2690 | * new orig_family_num for future updates. | |
2691 | */ | |
2692 | if (new_family) { | |
2693 | memcpy(&mpb->orig_family_num, info->uuid, sizeof(__u32)); | |
2694 | uuid_from_super_imsm(st, info->uuid); | |
2695 | *new_family = mpb->orig_family_num; | |
2696 | info->update_private = new_family; | |
2697 | rv = 0; | |
2698 | } | |
2699 | } else if (strcmp(update, "assemble") == 0) | |
2700 | rv = 0; | |
2701 | else | |
1e2b2765 | 2702 | rv = -1; |
f352c545 | 2703 | |
6e46bf34 DW |
2704 | /* successful update? recompute checksum */ |
2705 | if (rv == 0) | |
2706 | mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb)); | |
f352c545 DW |
2707 | |
2708 | return rv; | |
cdddbdbc DW |
2709 | } |
2710 | ||
c2c087e6 | 2711 | static size_t disks_to_mpb_size(int disks) |
cdddbdbc | 2712 | { |
c2c087e6 | 2713 | size_t size; |
cdddbdbc | 2714 | |
c2c087e6 DW |
2715 | size = sizeof(struct imsm_super); |
2716 | size += (disks - 1) * sizeof(struct imsm_disk); | |
2717 | size += 2 * sizeof(struct imsm_dev); | |
2718 | /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */ | |
2719 | size += (4 - 2) * sizeof(struct imsm_map); | |
2720 | /* 4 possible disk_ord_tbl's */ | |
2721 | size += 4 * (disks - 1) * sizeof(__u32); | |
2722 | ||
2723 | return size; | |
2724 | } | |
2725 | ||
2726 | static __u64 avail_size_imsm(struct supertype *st, __u64 devsize) | |
2727 | { | |
2728 | if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS)) | |
2729 | return 0; | |
2730 | ||
2731 | return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); | |
cdddbdbc DW |
2732 | } |
2733 | ||
ba2de7ba DW |
2734 | static void free_devlist(struct intel_super *super) |
2735 | { | |
2736 | struct intel_dev *dv; | |
2737 | ||
2738 | while (super->devlist) { | |
2739 | dv = super->devlist->next; | |
2740 | free(super->devlist->dev); | |
2741 | free(super->devlist); | |
2742 | super->devlist = dv; | |
2743 | } | |
2744 | } | |
2745 | ||
2746 | static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src) | |
2747 | { | |
2748 | memcpy(dest, src, sizeof_imsm_dev(src, 0)); | |
2749 | } | |
2750 | ||
cdddbdbc DW |
2751 | static int compare_super_imsm(struct supertype *st, struct supertype *tst) |
2752 | { | |
2753 | /* | |
2754 | * return: | |
2755 | * 0 same, or first was empty, and second was copied | |
2756 | * 1 second had wrong number | |
2757 | * 2 wrong uuid | |
2758 | * 3 wrong other info | |
2759 | */ | |
2760 | struct intel_super *first = st->sb; | |
2761 | struct intel_super *sec = tst->sb; | |
2762 | ||
2763 | if (!first) { | |
2764 | st->sb = tst->sb; | |
2765 | tst->sb = NULL; | |
2766 | return 0; | |
2767 | } | |
8603ea6f LM |
2768 | /* in platform dependent environment test if the disks |
2769 | * use the same Intel hba | |
2770 | */ | |
2771 | if (!check_env("IMSM_NO_PLATFORM")) { | |
ea2bc72b LM |
2772 | if (!first->hba || !sec->hba || |
2773 | (first->hba->type != sec->hba->type)) { | |
8603ea6f LM |
2774 | fprintf(stderr, |
2775 | "HBAs of devices does not match %s != %s\n", | |
ea2bc72b LM |
2776 | first->hba ? get_sys_dev_type(first->hba->type) : NULL, |
2777 | sec->hba ? get_sys_dev_type(sec->hba->type) : NULL); | |
8603ea6f LM |
2778 | return 3; |
2779 | } | |
2780 | } | |
cdddbdbc | 2781 | |
d23fe947 DW |
2782 | /* if an anchor does not have num_raid_devs set then it is a free |
2783 | * floating spare | |
2784 | */ | |
2785 | if (first->anchor->num_raid_devs > 0 && | |
2786 | sec->anchor->num_raid_devs > 0) { | |
a2b97981 DW |
2787 | /* Determine if these disks might ever have been |
2788 | * related. Further disambiguation can only take place | |
2789 | * in load_super_imsm_all | |
2790 | */ | |
2791 | __u32 first_family = first->anchor->orig_family_num; | |
2792 | __u32 sec_family = sec->anchor->orig_family_num; | |
2793 | ||
f796af5d DW |
2794 | if (memcmp(first->anchor->sig, sec->anchor->sig, |
2795 | MAX_SIGNATURE_LENGTH) != 0) | |
2796 | return 3; | |
2797 | ||
a2b97981 DW |
2798 | if (first_family == 0) |
2799 | first_family = first->anchor->family_num; | |
2800 | if (sec_family == 0) | |
2801 | sec_family = sec->anchor->family_num; | |
2802 | ||
2803 | if (first_family != sec_family) | |
d23fe947 | 2804 | return 3; |
f796af5d | 2805 | |
d23fe947 | 2806 | } |
cdddbdbc | 2807 | |
f796af5d | 2808 | |
3e372e5a DW |
2809 | /* if 'first' is a spare promote it to a populated mpb with sec's |
2810 | * family number | |
2811 | */ | |
2812 | if (first->anchor->num_raid_devs == 0 && | |
2813 | sec->anchor->num_raid_devs > 0) { | |
78d30f94 | 2814 | int i; |
ba2de7ba DW |
2815 | struct intel_dev *dv; |
2816 | struct imsm_dev *dev; | |
78d30f94 DW |
2817 | |
2818 | /* we need to copy raid device info from sec if an allocation | |
2819 | * fails here we don't associate the spare | |
2820 | */ | |
2821 | for (i = 0; i < sec->anchor->num_raid_devs; i++) { | |
ba2de7ba DW |
2822 | dv = malloc(sizeof(*dv)); |
2823 | if (!dv) | |
2824 | break; | |
2825 | dev = malloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1)); | |
2826 | if (!dev) { | |
2827 | free(dv); | |
2828 | break; | |
78d30f94 | 2829 | } |
ba2de7ba DW |
2830 | dv->dev = dev; |
2831 | dv->index = i; | |
2832 | dv->next = first->devlist; | |
2833 | first->devlist = dv; | |
78d30f94 | 2834 | } |
709743c5 | 2835 | if (i < sec->anchor->num_raid_devs) { |
ba2de7ba DW |
2836 | /* allocation failure */ |
2837 | free_devlist(first); | |
2838 | fprintf(stderr, "imsm: failed to associate spare\n"); | |
2839 | return 3; | |
78d30f94 | 2840 | } |
3e372e5a | 2841 | first->anchor->num_raid_devs = sec->anchor->num_raid_devs; |
148acb7b | 2842 | first->anchor->orig_family_num = sec->anchor->orig_family_num; |
3e372e5a | 2843 | first->anchor->family_num = sec->anchor->family_num; |
ac6449be | 2844 | memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH); |
709743c5 DW |
2845 | for (i = 0; i < sec->anchor->num_raid_devs; i++) |
2846 | imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i)); | |
3e372e5a DW |
2847 | } |
2848 | ||
cdddbdbc DW |
2849 | return 0; |
2850 | } | |
2851 | ||
0030e8d6 DW |
2852 | static void fd2devname(int fd, char *name) |
2853 | { | |
2854 | struct stat st; | |
2855 | char path[256]; | |
33a6535d | 2856 | char dname[PATH_MAX]; |
0030e8d6 DW |
2857 | char *nm; |
2858 | int rv; | |
2859 | ||
2860 | name[0] = '\0'; | |
2861 | if (fstat(fd, &st) != 0) | |
2862 | return; | |
2863 | sprintf(path, "/sys/dev/block/%d:%d", | |
2864 | major(st.st_rdev), minor(st.st_rdev)); | |
2865 | ||
2866 | rv = readlink(path, dname, sizeof(dname)); | |
2867 | if (rv <= 0) | |
2868 | return; | |
2869 | ||
2870 | dname[rv] = '\0'; | |
2871 | nm = strrchr(dname, '/'); | |
2872 | nm++; | |
2873 | snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm); | |
2874 | } | |
2875 | ||
cdddbdbc DW |
2876 | extern int scsi_get_serial(int fd, void *buf, size_t buf_len); |
2877 | ||
2878 | static int imsm_read_serial(int fd, char *devname, | |
2879 | __u8 serial[MAX_RAID_SERIAL_LEN]) | |
2880 | { | |
2881 | unsigned char scsi_serial[255]; | |
cdddbdbc DW |
2882 | int rv; |
2883 | int rsp_len; | |
1f24f035 | 2884 | int len; |
316e2bf4 DW |
2885 | char *dest; |
2886 | char *src; | |
2887 | char *rsp_buf; | |
2888 | int i; | |
cdddbdbc DW |
2889 | |
2890 | memset(scsi_serial, 0, sizeof(scsi_serial)); | |
cdddbdbc | 2891 | |
f9ba0ff1 DW |
2892 | rv = scsi_get_serial(fd, scsi_serial, sizeof(scsi_serial)); |
2893 | ||
40ebbb9c | 2894 | if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) { |
f9ba0ff1 DW |
2895 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
2896 | fd2devname(fd, (char *) serial); | |
0030e8d6 DW |
2897 | return 0; |
2898 | } | |
2899 | ||
cdddbdbc DW |
2900 | if (rv != 0) { |
2901 | if (devname) | |
2902 | fprintf(stderr, | |
2903 | Name ": Failed to retrieve serial for %s\n", | |
2904 | devname); | |
2905 | return rv; | |
2906 | } | |
2907 | ||
2908 | rsp_len = scsi_serial[3]; | |
03cd4cc8 DW |
2909 | if (!rsp_len) { |
2910 | if (devname) | |
2911 | fprintf(stderr, | |
2912 | Name ": Failed to retrieve serial for %s\n", | |
2913 | devname); | |
2914 | return 2; | |
2915 | } | |
1f24f035 | 2916 | rsp_buf = (char *) &scsi_serial[4]; |
5c3db629 | 2917 | |
316e2bf4 DW |
2918 | /* trim all whitespace and non-printable characters and convert |
2919 | * ':' to ';' | |
2920 | */ | |
2921 | for (i = 0, dest = rsp_buf; i < rsp_len; i++) { | |
2922 | src = &rsp_buf[i]; | |
2923 | if (*src > 0x20) { | |
2924 | /* ':' is reserved for use in placeholder serial | |
2925 | * numbers for missing disks | |
2926 | */ | |
2927 | if (*src == ':') | |
2928 | *dest++ = ';'; | |
2929 | else | |
2930 | *dest++ = *src; | |
2931 | } | |
2932 | } | |
2933 | len = dest - rsp_buf; | |
2934 | dest = rsp_buf; | |
2935 | ||
2936 | /* truncate leading characters */ | |
2937 | if (len > MAX_RAID_SERIAL_LEN) { | |
2938 | dest += len - MAX_RAID_SERIAL_LEN; | |
1f24f035 | 2939 | len = MAX_RAID_SERIAL_LEN; |
316e2bf4 | 2940 | } |
5c3db629 | 2941 | |
5c3db629 | 2942 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
316e2bf4 | 2943 | memcpy(serial, dest, len); |
cdddbdbc DW |
2944 | |
2945 | return 0; | |
2946 | } | |
2947 | ||
1f24f035 DW |
2948 | static int serialcmp(__u8 *s1, __u8 *s2) |
2949 | { | |
2950 | return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN); | |
2951 | } | |
2952 | ||
2953 | static void serialcpy(__u8 *dest, __u8 *src) | |
2954 | { | |
2955 | strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN); | |
2956 | } | |
2957 | ||
54c2c1ea DW |
2958 | static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super) |
2959 | { | |
2960 | struct dl *dl; | |
2961 | ||
2962 | for (dl = super->disks; dl; dl = dl->next) | |
2963 | if (serialcmp(dl->serial, serial) == 0) | |
2964 | break; | |
2965 | ||
2966 | return dl; | |
2967 | } | |
2968 | ||
a2b97981 DW |
2969 | static struct imsm_disk * |
2970 | __serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx) | |
2971 | { | |
2972 | int i; | |
2973 | ||
2974 | for (i = 0; i < mpb->num_disks; i++) { | |
2975 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
2976 | ||
2977 | if (serialcmp(disk->serial, serial) == 0) { | |
2978 | if (idx) | |
2979 | *idx = i; | |
2980 | return disk; | |
2981 | } | |
2982 | } | |
2983 | ||
2984 | return NULL; | |
2985 | } | |
2986 | ||
cdddbdbc DW |
2987 | static int |
2988 | load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd) | |
2989 | { | |
a2b97981 | 2990 | struct imsm_disk *disk; |
cdddbdbc DW |
2991 | struct dl *dl; |
2992 | struct stat stb; | |
cdddbdbc | 2993 | int rv; |
a2b97981 | 2994 | char name[40]; |
d23fe947 DW |
2995 | __u8 serial[MAX_RAID_SERIAL_LEN]; |
2996 | ||
2997 | rv = imsm_read_serial(fd, devname, serial); | |
2998 | ||
2999 | if (rv != 0) | |
3000 | return 2; | |
3001 | ||
a2b97981 | 3002 | dl = calloc(1, sizeof(*dl)); |
b9f594fe | 3003 | if (!dl) { |
cdddbdbc DW |
3004 | if (devname) |
3005 | fprintf(stderr, | |
3006 | Name ": failed to allocate disk buffer for %s\n", | |
3007 | devname); | |
3008 | return 2; | |
3009 | } | |
cdddbdbc | 3010 | |
a2b97981 DW |
3011 | fstat(fd, &stb); |
3012 | dl->major = major(stb.st_rdev); | |
3013 | dl->minor = minor(stb.st_rdev); | |
3014 | dl->next = super->disks; | |
3015 | dl->fd = keep_fd ? fd : -1; | |
3016 | assert(super->disks == NULL); | |
3017 | super->disks = dl; | |
3018 | serialcpy(dl->serial, serial); | |
3019 | dl->index = -2; | |
3020 | dl->e = NULL; | |
3021 | fd2devname(fd, name); | |
3022 | if (devname) | |
3023 | dl->devname = strdup(devname); | |
3024 | else | |
3025 | dl->devname = strdup(name); | |
cdddbdbc | 3026 | |
d23fe947 | 3027 | /* look up this disk's index in the current anchor */ |
a2b97981 DW |
3028 | disk = __serial_to_disk(dl->serial, super->anchor, &dl->index); |
3029 | if (disk) { | |
3030 | dl->disk = *disk; | |
3031 | /* only set index on disks that are a member of a | |
3032 | * populated contianer, i.e. one with raid_devs | |
3033 | */ | |
3034 | if (is_failed(&dl->disk)) | |
3f6efecc | 3035 | dl->index = -2; |
a2b97981 DW |
3036 | else if (is_spare(&dl->disk)) |
3037 | dl->index = -1; | |
3f6efecc DW |
3038 | } |
3039 | ||
949c47a0 DW |
3040 | return 0; |
3041 | } | |
3042 | ||
0e600426 | 3043 | #ifndef MDASSEMBLE |
0c046afd DW |
3044 | /* When migrating map0 contains the 'destination' state while map1 |
3045 | * contains the current state. When not migrating map0 contains the | |
3046 | * current state. This routine assumes that map[0].map_state is set to | |
3047 | * the current array state before being called. | |
3048 | * | |
3049 | * Migration is indicated by one of the following states | |
3050 | * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed) | |
e3bba0e0 | 3051 | * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal |
0c046afd | 3052 | * map1state=unitialized) |
1484e727 | 3053 | * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal |
0c046afd | 3054 | * map1state=normal) |
e3bba0e0 | 3055 | * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal |
0c046afd | 3056 | * map1state=degraded) |
8e59f3d8 AK |
3057 | * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal |
3058 | * map1state=normal) | |
0c046afd | 3059 | */ |
8e59f3d8 AK |
3060 | static void migrate(struct imsm_dev *dev, struct intel_super *super, |
3061 | __u8 to_state, int migr_type) | |
3393c6af | 3062 | { |
0c046afd | 3063 | struct imsm_map *dest; |
3393c6af DW |
3064 | struct imsm_map *src = get_imsm_map(dev, 0); |
3065 | ||
0c046afd | 3066 | dev->vol.migr_state = 1; |
1484e727 | 3067 | set_migr_type(dev, migr_type); |
f8f603f1 | 3068 | dev->vol.curr_migr_unit = 0; |
0c046afd DW |
3069 | dest = get_imsm_map(dev, 1); |
3070 | ||
0556e1a2 | 3071 | /* duplicate and then set the target end state in map[0] */ |
3393c6af | 3072 | memcpy(dest, src, sizeof_imsm_map(src)); |
28bce06f AK |
3073 | if ((migr_type == MIGR_REBUILD) || |
3074 | (migr_type == MIGR_GEN_MIGR)) { | |
0556e1a2 DW |
3075 | __u32 ord; |
3076 | int i; | |
3077 | ||
3078 | for (i = 0; i < src->num_members; i++) { | |
3079 | ord = __le32_to_cpu(src->disk_ord_tbl[i]); | |
3080 | set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord)); | |
3081 | } | |
3082 | } | |
3083 | ||
8e59f3d8 AK |
3084 | if (migr_type == MIGR_GEN_MIGR) |
3085 | /* Clear migration record */ | |
3086 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
3087 | ||
0c046afd | 3088 | src->map_state = to_state; |
949c47a0 | 3089 | } |
f8f603f1 DW |
3090 | |
3091 | static void end_migration(struct imsm_dev *dev, __u8 map_state) | |
3092 | { | |
3093 | struct imsm_map *map = get_imsm_map(dev, 0); | |
0556e1a2 | 3094 | struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state); |
28bce06f | 3095 | int i, j; |
0556e1a2 DW |
3096 | |
3097 | /* merge any IMSM_ORD_REBUILD bits that were not successfully | |
3098 | * completed in the last migration. | |
3099 | * | |
28bce06f | 3100 | * FIXME add support for raid-level-migration |
0556e1a2 DW |
3101 | */ |
3102 | for (i = 0; i < prev->num_members; i++) | |
28bce06f AK |
3103 | for (j = 0; j < map->num_members; j++) |
3104 | /* during online capacity expansion | |
3105 | * disks position can be changed if takeover is used | |
3106 | */ | |
3107 | if (ord_to_idx(map->disk_ord_tbl[j]) == | |
3108 | ord_to_idx(prev->disk_ord_tbl[i])) { | |
3109 | map->disk_ord_tbl[j] |= prev->disk_ord_tbl[i]; | |
3110 | break; | |
3111 | } | |
f8f603f1 DW |
3112 | |
3113 | dev->vol.migr_state = 0; | |
ea672ee1 | 3114 | set_migr_type(dev, 0); |
f8f603f1 DW |
3115 | dev->vol.curr_migr_unit = 0; |
3116 | map->map_state = map_state; | |
3117 | } | |
0e600426 | 3118 | #endif |
949c47a0 DW |
3119 | |
3120 | static int parse_raid_devices(struct intel_super *super) | |
3121 | { | |
3122 | int i; | |
3123 | struct imsm_dev *dev_new; | |
4d7b1503 | 3124 | size_t len, len_migr; |
401d313b | 3125 | size_t max_len = 0; |
4d7b1503 DW |
3126 | size_t space_needed = 0; |
3127 | struct imsm_super *mpb = super->anchor; | |
949c47a0 DW |
3128 | |
3129 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
3130 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
ba2de7ba | 3131 | struct intel_dev *dv; |
949c47a0 | 3132 | |
4d7b1503 DW |
3133 | len = sizeof_imsm_dev(dev_iter, 0); |
3134 | len_migr = sizeof_imsm_dev(dev_iter, 1); | |
3135 | if (len_migr > len) | |
3136 | space_needed += len_migr - len; | |
3137 | ||
ba2de7ba DW |
3138 | dv = malloc(sizeof(*dv)); |
3139 | if (!dv) | |
3140 | return 1; | |
401d313b AK |
3141 | if (max_len < len_migr) |
3142 | max_len = len_migr; | |
3143 | if (max_len > len_migr) | |
3144 | space_needed += max_len - len_migr; | |
3145 | dev_new = malloc(max_len); | |
ba2de7ba DW |
3146 | if (!dev_new) { |
3147 | free(dv); | |
949c47a0 | 3148 | return 1; |
ba2de7ba | 3149 | } |
949c47a0 | 3150 | imsm_copy_dev(dev_new, dev_iter); |
ba2de7ba DW |
3151 | dv->dev = dev_new; |
3152 | dv->index = i; | |
3153 | dv->next = super->devlist; | |
3154 | super->devlist = dv; | |
949c47a0 | 3155 | } |
cdddbdbc | 3156 | |
4d7b1503 DW |
3157 | /* ensure that super->buf is large enough when all raid devices |
3158 | * are migrating | |
3159 | */ | |
3160 | if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) { | |
3161 | void *buf; | |
3162 | ||
3163 | len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed, 512); | |
3164 | if (posix_memalign(&buf, 512, len) != 0) | |
3165 | return 1; | |
3166 | ||
1f45a8ad DW |
3167 | memcpy(buf, super->buf, super->len); |
3168 | memset(buf + super->len, 0, len - super->len); | |
4d7b1503 DW |
3169 | free(super->buf); |
3170 | super->buf = buf; | |
3171 | super->len = len; | |
3172 | } | |
3173 | ||
cdddbdbc DW |
3174 | return 0; |
3175 | } | |
3176 | ||
604b746f JD |
3177 | /* retrieve a pointer to the bbm log which starts after all raid devices */ |
3178 | struct bbm_log *__get_imsm_bbm_log(struct imsm_super *mpb) | |
3179 | { | |
3180 | void *ptr = NULL; | |
3181 | ||
3182 | if (__le32_to_cpu(mpb->bbm_log_size)) { | |
3183 | ptr = mpb; | |
3184 | ptr += mpb->mpb_size - __le32_to_cpu(mpb->bbm_log_size); | |
3185 | } | |
3186 | ||
3187 | return ptr; | |
3188 | } | |
3189 | ||
e2f41b2c AK |
3190 | /******************************************************************************* |
3191 | * Function: check_mpb_migr_compatibility | |
3192 | * Description: Function checks for unsupported migration features: | |
3193 | * - migration optimization area (pba_of_lba0) | |
3194 | * - descending reshape (ascending_migr) | |
3195 | * Parameters: | |
3196 | * super : imsm metadata information | |
3197 | * Returns: | |
3198 | * 0 : migration is compatible | |
3199 | * -1 : migration is not compatible | |
3200 | ******************************************************************************/ | |
3201 | int check_mpb_migr_compatibility(struct intel_super *super) | |
3202 | { | |
3203 | struct imsm_map *map0, *map1; | |
3204 | struct migr_record *migr_rec = super->migr_rec; | |
3205 | int i; | |
3206 | ||
3207 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
3208 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
3209 | ||
3210 | if (dev_iter && | |
3211 | dev_iter->vol.migr_state == 1 && | |
3212 | dev_iter->vol.migr_type == MIGR_GEN_MIGR) { | |
3213 | /* This device is migrating */ | |
3214 | map0 = get_imsm_map(dev_iter, 0); | |
3215 | map1 = get_imsm_map(dev_iter, 1); | |
3216 | if (map0->pba_of_lba0 != map1->pba_of_lba0) | |
3217 | /* migration optimization area was used */ | |
3218 | return -1; | |
3219 | if (migr_rec->ascending_migr == 0 | |
3220 | && migr_rec->dest_depth_per_unit > 0) | |
3221 | /* descending reshape not supported yet */ | |
3222 | return -1; | |
3223 | } | |
3224 | } | |
3225 | return 0; | |
3226 | } | |
3227 | ||
d23fe947 | 3228 | static void __free_imsm(struct intel_super *super, int free_disks); |
9ca2c81c | 3229 | |
cdddbdbc | 3230 | /* load_imsm_mpb - read matrix metadata |
f2f5c343 | 3231 | * allocates super->mpb to be freed by free_imsm |
cdddbdbc DW |
3232 | */ |
3233 | static int load_imsm_mpb(int fd, struct intel_super *super, char *devname) | |
3234 | { | |
3235 | unsigned long long dsize; | |
cdddbdbc DW |
3236 | unsigned long long sectors; |
3237 | struct stat; | |
6416d527 | 3238 | struct imsm_super *anchor; |
cdddbdbc DW |
3239 | __u32 check_sum; |
3240 | ||
cdddbdbc | 3241 | get_dev_size(fd, NULL, &dsize); |
64436f06 N |
3242 | if (dsize < 1024) { |
3243 | if (devname) | |
3244 | fprintf(stderr, | |
3245 | Name ": %s: device to small for imsm\n", | |
3246 | devname); | |
3247 | return 1; | |
3248 | } | |
cdddbdbc DW |
3249 | |
3250 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) { | |
3251 | if (devname) | |
2e062e82 AK |
3252 | fprintf(stderr, Name |
3253 | ": Cannot seek to anchor block on %s: %s\n", | |
cdddbdbc DW |
3254 | devname, strerror(errno)); |
3255 | return 1; | |
3256 | } | |
3257 | ||
949c47a0 | 3258 | if (posix_memalign((void**)&anchor, 512, 512) != 0) { |
ad97895e DW |
3259 | if (devname) |
3260 | fprintf(stderr, | |
3261 | Name ": Failed to allocate imsm anchor buffer" | |
3262 | " on %s\n", devname); | |
3263 | return 1; | |
3264 | } | |
949c47a0 | 3265 | if (read(fd, anchor, 512) != 512) { |
cdddbdbc DW |
3266 | if (devname) |
3267 | fprintf(stderr, | |
3268 | Name ": Cannot read anchor block on %s: %s\n", | |
3269 | devname, strerror(errno)); | |
6416d527 | 3270 | free(anchor); |
cdddbdbc DW |
3271 | return 1; |
3272 | } | |
3273 | ||
6416d527 | 3274 | if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) { |
cdddbdbc DW |
3275 | if (devname) |
3276 | fprintf(stderr, | |
3277 | Name ": no IMSM anchor on %s\n", devname); | |
6416d527 | 3278 | free(anchor); |
cdddbdbc DW |
3279 | return 2; |
3280 | } | |
3281 | ||
d23fe947 | 3282 | __free_imsm(super, 0); |
f2f5c343 LM |
3283 | /* reload capability and hba */ |
3284 | ||
3285 | /* capability and hba must be updated with new super allocation */ | |
d424212e | 3286 | find_intel_hba_capability(fd, super, devname); |
949c47a0 DW |
3287 | super->len = ROUND_UP(anchor->mpb_size, 512); |
3288 | if (posix_memalign(&super->buf, 512, super->len) != 0) { | |
cdddbdbc DW |
3289 | if (devname) |
3290 | fprintf(stderr, | |
3291 | Name ": unable to allocate %zu byte mpb buffer\n", | |
949c47a0 | 3292 | super->len); |
6416d527 | 3293 | free(anchor); |
cdddbdbc DW |
3294 | return 2; |
3295 | } | |
949c47a0 | 3296 | memcpy(super->buf, anchor, 512); |
cdddbdbc | 3297 | |
6416d527 NB |
3298 | sectors = mpb_sectors(anchor) - 1; |
3299 | free(anchor); | |
8e59f3d8 AK |
3300 | |
3301 | if (posix_memalign(&super->migr_rec_buf, 512, 512) != 0) { | |
3302 | fprintf(stderr, Name | |
3303 | ": %s could not allocate migr_rec buffer\n", __func__); | |
3304 | free(super->buf); | |
3305 | return 2; | |
3306 | } | |
3307 | ||
949c47a0 | 3308 | if (!sectors) { |
ecf45690 DW |
3309 | check_sum = __gen_imsm_checksum(super->anchor); |
3310 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
3311 | if (devname) | |
3312 | fprintf(stderr, | |
3313 | Name ": IMSM checksum %x != %x on %s\n", | |
3314 | check_sum, | |
3315 | __le32_to_cpu(super->anchor->check_sum), | |
3316 | devname); | |
3317 | return 2; | |
3318 | } | |
3319 | ||
a2b97981 | 3320 | return 0; |
949c47a0 | 3321 | } |
cdddbdbc DW |
3322 | |
3323 | /* read the extended mpb */ | |
3324 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) { | |
3325 | if (devname) | |
3326 | fprintf(stderr, | |
3327 | Name ": Cannot seek to extended mpb on %s: %s\n", | |
3328 | devname, strerror(errno)); | |
3329 | return 1; | |
3330 | } | |
3331 | ||
f21e18ca | 3332 | if ((unsigned)read(fd, super->buf + 512, super->len - 512) != super->len - 512) { |
cdddbdbc DW |
3333 | if (devname) |
3334 | fprintf(stderr, | |
3335 | Name ": Cannot read extended mpb on %s: %s\n", | |
3336 | devname, strerror(errno)); | |
3337 | return 2; | |
3338 | } | |
3339 | ||
949c47a0 DW |
3340 | check_sum = __gen_imsm_checksum(super->anchor); |
3341 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
cdddbdbc DW |
3342 | if (devname) |
3343 | fprintf(stderr, | |
3344 | Name ": IMSM checksum %x != %x on %s\n", | |
949c47a0 | 3345 | check_sum, __le32_to_cpu(super->anchor->check_sum), |
cdddbdbc | 3346 | devname); |
db575f3b | 3347 | return 3; |
cdddbdbc DW |
3348 | } |
3349 | ||
604b746f JD |
3350 | /* FIXME the BBM log is disk specific so we cannot use this global |
3351 | * buffer for all disks. Ok for now since we only look at the global | |
3352 | * bbm_log_size parameter to gate assembly | |
3353 | */ | |
3354 | super->bbm_log = __get_imsm_bbm_log(super->anchor); | |
3355 | ||
a2b97981 DW |
3356 | return 0; |
3357 | } | |
3358 | ||
8e59f3d8 AK |
3359 | static int read_imsm_migr_rec(int fd, struct intel_super *super); |
3360 | ||
a2b97981 DW |
3361 | static int |
3362 | load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd) | |
3363 | { | |
3364 | int err; | |
3365 | ||
3366 | err = load_imsm_mpb(fd, super, devname); | |
3367 | if (err) | |
3368 | return err; | |
3369 | err = load_imsm_disk(fd, super, devname, keep_fd); | |
3370 | if (err) | |
3371 | return err; | |
3372 | err = parse_raid_devices(super); | |
4d7b1503 | 3373 | |
a2b97981 | 3374 | return err; |
cdddbdbc DW |
3375 | } |
3376 | ||
ae6aad82 DW |
3377 | static void __free_imsm_disk(struct dl *d) |
3378 | { | |
3379 | if (d->fd >= 0) | |
3380 | close(d->fd); | |
3381 | if (d->devname) | |
3382 | free(d->devname); | |
0dcecb2e DW |
3383 | if (d->e) |
3384 | free(d->e); | |
ae6aad82 DW |
3385 | free(d); |
3386 | ||
3387 | } | |
1a64be56 | 3388 | |
cdddbdbc DW |
3389 | static void free_imsm_disks(struct intel_super *super) |
3390 | { | |
47ee5a45 | 3391 | struct dl *d; |
cdddbdbc | 3392 | |
47ee5a45 DW |
3393 | while (super->disks) { |
3394 | d = super->disks; | |
cdddbdbc | 3395 | super->disks = d->next; |
ae6aad82 | 3396 | __free_imsm_disk(d); |
cdddbdbc | 3397 | } |
cb82edca AK |
3398 | while (super->disk_mgmt_list) { |
3399 | d = super->disk_mgmt_list; | |
3400 | super->disk_mgmt_list = d->next; | |
3401 | __free_imsm_disk(d); | |
3402 | } | |
47ee5a45 DW |
3403 | while (super->missing) { |
3404 | d = super->missing; | |
3405 | super->missing = d->next; | |
3406 | __free_imsm_disk(d); | |
3407 | } | |
3408 | ||
cdddbdbc DW |
3409 | } |
3410 | ||
9ca2c81c | 3411 | /* free all the pieces hanging off of a super pointer */ |
d23fe947 | 3412 | static void __free_imsm(struct intel_super *super, int free_disks) |
cdddbdbc | 3413 | { |
88654014 LM |
3414 | struct intel_hba *elem, *next; |
3415 | ||
9ca2c81c | 3416 | if (super->buf) { |
949c47a0 | 3417 | free(super->buf); |
9ca2c81c DW |
3418 | super->buf = NULL; |
3419 | } | |
f2f5c343 LM |
3420 | /* unlink capability description */ |
3421 | super->orom = NULL; | |
8e59f3d8 AK |
3422 | if (super->migr_rec_buf) { |
3423 | free(super->migr_rec_buf); | |
3424 | super->migr_rec_buf = NULL; | |
3425 | } | |
d23fe947 DW |
3426 | if (free_disks) |
3427 | free_imsm_disks(super); | |
ba2de7ba | 3428 | free_devlist(super); |
88654014 LM |
3429 | elem = super->hba; |
3430 | while (elem) { | |
3431 | if (elem->path) | |
3432 | free((void *)elem->path); | |
3433 | next = elem->next; | |
3434 | free(elem); | |
3435 | elem = next; | |
88c32bb1 | 3436 | } |
88654014 | 3437 | super->hba = NULL; |
cdddbdbc DW |
3438 | } |
3439 | ||
9ca2c81c DW |
3440 | static void free_imsm(struct intel_super *super) |
3441 | { | |
d23fe947 | 3442 | __free_imsm(super, 1); |
9ca2c81c DW |
3443 | free(super); |
3444 | } | |
cdddbdbc DW |
3445 | |
3446 | static void free_super_imsm(struct supertype *st) | |
3447 | { | |
3448 | struct intel_super *super = st->sb; | |
3449 | ||
3450 | if (!super) | |
3451 | return; | |
3452 | ||
3453 | free_imsm(super); | |
3454 | st->sb = NULL; | |
3455 | } | |
3456 | ||
49133e57 | 3457 | static struct intel_super *alloc_super(void) |
c2c087e6 DW |
3458 | { |
3459 | struct intel_super *super = malloc(sizeof(*super)); | |
3460 | ||
3461 | if (super) { | |
3462 | memset(super, 0, sizeof(*super)); | |
bf5a934a | 3463 | super->current_vol = -1; |
0dcecb2e | 3464 | super->create_offset = ~((__u32 ) 0); |
c2c087e6 | 3465 | } |
c2c087e6 DW |
3466 | return super; |
3467 | } | |
3468 | ||
f0f5a016 LM |
3469 | /* |
3470 | * find and allocate hba and OROM/EFI based on valid fd of RAID component device | |
3471 | */ | |
d424212e | 3472 | static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname) |
f0f5a016 LM |
3473 | { |
3474 | struct sys_dev *hba_name; | |
3475 | int rv = 0; | |
3476 | ||
3477 | if ((fd < 0) || check_env("IMSM_NO_PLATFORM")) { | |
f2f5c343 | 3478 | super->orom = NULL; |
f0f5a016 LM |
3479 | super->hba = NULL; |
3480 | return 0; | |
3481 | } | |
3482 | hba_name = find_disk_attached_hba(fd, NULL); | |
3483 | if (!hba_name) { | |
d424212e | 3484 | if (devname) |
f0f5a016 LM |
3485 | fprintf(stderr, |
3486 | Name ": %s is not attached to Intel(R) RAID controller.\n", | |
d424212e | 3487 | devname); |
f0f5a016 LM |
3488 | return 1; |
3489 | } | |
3490 | rv = attach_hba_to_super(super, hba_name); | |
3491 | if (rv == 2) { | |
d424212e N |
3492 | if (devname) { |
3493 | struct intel_hba *hba = super->hba; | |
f0f5a016 | 3494 | |
f0f5a016 LM |
3495 | fprintf(stderr, Name ": %s is attached to Intel(R) %s RAID " |
3496 | "controller (%s),\n" | |
3497 | " but the container is assigned to Intel(R) " | |
3498 | "%s RAID controller (", | |
d424212e | 3499 | devname, |
f0f5a016 LM |
3500 | hba_name->path, |
3501 | hba_name->pci_id ? : "Err!", | |
3502 | get_sys_dev_type(hba_name->type)); | |
3503 | ||
f0f5a016 LM |
3504 | while (hba) { |
3505 | fprintf(stderr, "%s", hba->pci_id ? : "Err!"); | |
3506 | if (hba->next) | |
3507 | fprintf(stderr, ", "); | |
3508 | hba = hba->next; | |
3509 | } | |
3510 | ||
3511 | fprintf(stderr, ").\n" | |
3512 | " Mixing devices attached to different controllers " | |
3513 | "is not allowed.\n"); | |
3514 | } | |
3515 | free_sys_dev(&hba_name); | |
3516 | return 2; | |
3517 | } | |
f2f5c343 | 3518 | super->orom = find_imsm_capability(hba_name->type); |
f0f5a016 | 3519 | free_sys_dev(&hba_name); |
f2f5c343 LM |
3520 | if (!super->orom) |
3521 | return 3; | |
f0f5a016 LM |
3522 | return 0; |
3523 | } | |
3524 | ||
47ee5a45 DW |
3525 | /* find_missing - helper routine for load_super_imsm_all that identifies |
3526 | * disks that have disappeared from the system. This routine relies on | |
3527 | * the mpb being uptodate, which it is at load time. | |
3528 | */ | |
3529 | static int find_missing(struct intel_super *super) | |
3530 | { | |
3531 | int i; | |
3532 | struct imsm_super *mpb = super->anchor; | |
3533 | struct dl *dl; | |
3534 | struct imsm_disk *disk; | |
47ee5a45 DW |
3535 | |
3536 | for (i = 0; i < mpb->num_disks; i++) { | |
3537 | disk = __get_imsm_disk(mpb, i); | |
54c2c1ea | 3538 | dl = serial_to_dl(disk->serial, super); |
47ee5a45 DW |
3539 | if (dl) |
3540 | continue; | |
47ee5a45 DW |
3541 | |
3542 | dl = malloc(sizeof(*dl)); | |
3543 | if (!dl) | |
3544 | return 1; | |
3545 | dl->major = 0; | |
3546 | dl->minor = 0; | |
3547 | dl->fd = -1; | |
3548 | dl->devname = strdup("missing"); | |
3549 | dl->index = i; | |
3550 | serialcpy(dl->serial, disk->serial); | |
3551 | dl->disk = *disk; | |
689c9bf3 | 3552 | dl->e = NULL; |
47ee5a45 DW |
3553 | dl->next = super->missing; |
3554 | super->missing = dl; | |
3555 | } | |
3556 | ||
3557 | return 0; | |
3558 | } | |
3559 | ||
3960e579 | 3560 | #ifndef MDASSEMBLE |
a2b97981 DW |
3561 | static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list) |
3562 | { | |
3563 | struct intel_disk *idisk = disk_list; | |
3564 | ||
3565 | while (idisk) { | |
3566 | if (serialcmp(idisk->disk.serial, serial) == 0) | |
3567 | break; | |
3568 | idisk = idisk->next; | |
3569 | } | |
3570 | ||
3571 | return idisk; | |
3572 | } | |
3573 | ||
3574 | static int __prep_thunderdome(struct intel_super **table, int tbl_size, | |
3575 | struct intel_super *super, | |
3576 | struct intel_disk **disk_list) | |
3577 | { | |
3578 | struct imsm_disk *d = &super->disks->disk; | |
3579 | struct imsm_super *mpb = super->anchor; | |
3580 | int i, j; | |
3581 | ||
3582 | for (i = 0; i < tbl_size; i++) { | |
3583 | struct imsm_super *tbl_mpb = table[i]->anchor; | |
3584 | struct imsm_disk *tbl_d = &table[i]->disks->disk; | |
3585 | ||
3586 | if (tbl_mpb->family_num == mpb->family_num) { | |
3587 | if (tbl_mpb->check_sum == mpb->check_sum) { | |
3588 | dprintf("%s: mpb from %d:%d matches %d:%d\n", | |
3589 | __func__, super->disks->major, | |
3590 | super->disks->minor, | |
3591 | table[i]->disks->major, | |
3592 | table[i]->disks->minor); | |
3593 | break; | |
3594 | } | |
3595 | ||
3596 | if (((is_configured(d) && !is_configured(tbl_d)) || | |
3597 | is_configured(d) == is_configured(tbl_d)) && | |
3598 | tbl_mpb->generation_num < mpb->generation_num) { | |
3599 | /* current version of the mpb is a | |
3600 | * better candidate than the one in | |
3601 | * super_table, but copy over "cross | |
3602 | * generational" status | |
3603 | */ | |
3604 | struct intel_disk *idisk; | |
3605 | ||
3606 | dprintf("%s: mpb from %d:%d replaces %d:%d\n", | |
3607 | __func__, super->disks->major, | |
3608 | super->disks->minor, | |
3609 | table[i]->disks->major, | |
3610 | table[i]->disks->minor); | |
3611 | ||
3612 | idisk = disk_list_get(tbl_d->serial, *disk_list); | |
3613 | if (idisk && is_failed(&idisk->disk)) | |
3614 | tbl_d->status |= FAILED_DISK; | |
3615 | break; | |
3616 | } else { | |
3617 | struct intel_disk *idisk; | |
3618 | struct imsm_disk *disk; | |
3619 | ||
3620 | /* tbl_mpb is more up to date, but copy | |
3621 | * over cross generational status before | |
3622 | * returning | |
3623 | */ | |
3624 | disk = __serial_to_disk(d->serial, mpb, NULL); | |
3625 | if (disk && is_failed(disk)) | |
3626 | d->status |= FAILED_DISK; | |
3627 | ||
3628 | idisk = disk_list_get(d->serial, *disk_list); | |
3629 | if (idisk) { | |
3630 | idisk->owner = i; | |
3631 | if (disk && is_configured(disk)) | |
3632 | idisk->disk.status |= CONFIGURED_DISK; | |
3633 | } | |
3634 | ||
3635 | dprintf("%s: mpb from %d:%d prefer %d:%d\n", | |
3636 | __func__, super->disks->major, | |
3637 | super->disks->minor, | |
3638 | table[i]->disks->major, | |
3639 | table[i]->disks->minor); | |
3640 | ||
3641 | return tbl_size; | |
3642 | } | |
3643 | } | |
3644 | } | |
3645 | ||
3646 | if (i >= tbl_size) | |
3647 | table[tbl_size++] = super; | |
3648 | else | |
3649 | table[i] = super; | |
3650 | ||
3651 | /* update/extend the merged list of imsm_disk records */ | |
3652 | for (j = 0; j < mpb->num_disks; j++) { | |
3653 | struct imsm_disk *disk = __get_imsm_disk(mpb, j); | |
3654 | struct intel_disk *idisk; | |
3655 | ||
3656 | idisk = disk_list_get(disk->serial, *disk_list); | |
3657 | if (idisk) { | |
3658 | idisk->disk.status |= disk->status; | |
3659 | if (is_configured(&idisk->disk) || | |
3660 | is_failed(&idisk->disk)) | |
3661 | idisk->disk.status &= ~(SPARE_DISK); | |
3662 | } else { | |
3663 | idisk = calloc(1, sizeof(*idisk)); | |
3664 | if (!idisk) | |
3665 | return -1; | |
3666 | idisk->owner = IMSM_UNKNOWN_OWNER; | |
3667 | idisk->disk = *disk; | |
3668 | idisk->next = *disk_list; | |
3669 | *disk_list = idisk; | |
3670 | } | |
3671 | ||
3672 | if (serialcmp(idisk->disk.serial, d->serial) == 0) | |
3673 | idisk->owner = i; | |
3674 | } | |
3675 | ||
3676 | return tbl_size; | |
3677 | } | |
3678 | ||
3679 | static struct intel_super * | |
3680 | validate_members(struct intel_super *super, struct intel_disk *disk_list, | |
3681 | const int owner) | |
3682 | { | |
3683 | struct imsm_super *mpb = super->anchor; | |
3684 | int ok_count = 0; | |
3685 | int i; | |
3686 | ||
3687 | for (i = 0; i < mpb->num_disks; i++) { | |
3688 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
3689 | struct intel_disk *idisk; | |
3690 | ||
3691 | idisk = disk_list_get(disk->serial, disk_list); | |
3692 | if (idisk) { | |
3693 | if (idisk->owner == owner || | |
3694 | idisk->owner == IMSM_UNKNOWN_OWNER) | |
3695 | ok_count++; | |
3696 | else | |
3697 | dprintf("%s: '%.16s' owner %d != %d\n", | |
3698 | __func__, disk->serial, idisk->owner, | |
3699 | owner); | |
3700 | } else { | |
3701 | dprintf("%s: unknown disk %x [%d]: %.16s\n", | |
3702 | __func__, __le32_to_cpu(mpb->family_num), i, | |
3703 | disk->serial); | |
3704 | break; | |
3705 | } | |
3706 | } | |
3707 | ||
3708 | if (ok_count == mpb->num_disks) | |
3709 | return super; | |
3710 | return NULL; | |
3711 | } | |
3712 | ||
3713 | static void show_conflicts(__u32 family_num, struct intel_super *super_list) | |
3714 | { | |
3715 | struct intel_super *s; | |
3716 | ||
3717 | for (s = super_list; s; s = s->next) { | |
3718 | if (family_num != s->anchor->family_num) | |
3719 | continue; | |
3720 | fprintf(stderr, "Conflict, offlining family %#x on '%s'\n", | |
3721 | __le32_to_cpu(family_num), s->disks->devname); | |
3722 | } | |
3723 | } | |
3724 | ||
3725 | static struct intel_super * | |
3726 | imsm_thunderdome(struct intel_super **super_list, int len) | |
3727 | { | |
3728 | struct intel_super *super_table[len]; | |
3729 | struct intel_disk *disk_list = NULL; | |
3730 | struct intel_super *champion, *spare; | |
3731 | struct intel_super *s, **del; | |
3732 | int tbl_size = 0; | |
3733 | int conflict; | |
3734 | int i; | |
3735 | ||
3736 | memset(super_table, 0, sizeof(super_table)); | |
3737 | for (s = *super_list; s; s = s->next) | |
3738 | tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list); | |
3739 | ||
3740 | for (i = 0; i < tbl_size; i++) { | |
3741 | struct imsm_disk *d; | |
3742 | struct intel_disk *idisk; | |
3743 | struct imsm_super *mpb = super_table[i]->anchor; | |
3744 | ||
3745 | s = super_table[i]; | |
3746 | d = &s->disks->disk; | |
3747 | ||
3748 | /* 'd' must appear in merged disk list for its | |
3749 | * configuration to be valid | |
3750 | */ | |
3751 | idisk = disk_list_get(d->serial, disk_list); | |
3752 | if (idisk && idisk->owner == i) | |
3753 | s = validate_members(s, disk_list, i); | |
3754 | else | |
3755 | s = NULL; | |
3756 | ||
3757 | if (!s) | |
3758 | dprintf("%s: marking family: %#x from %d:%d offline\n", | |
3759 | __func__, mpb->family_num, | |
3760 | super_table[i]->disks->major, | |
3761 | super_table[i]->disks->minor); | |
3762 | super_table[i] = s; | |
3763 | } | |
3764 | ||
3765 | /* This is where the mdadm implementation differs from the Windows | |
3766 | * driver which has no strict concept of a container. We can only | |
3767 | * assemble one family from a container, so when returning a prodigal | |
3768 | * array member to this system the code will not be able to disambiguate | |
3769 | * the container contents that should be assembled ("foreign" versus | |
3770 | * "local"). It requires user intervention to set the orig_family_num | |
3771 | * to a new value to establish a new container. The Windows driver in | |
3772 | * this situation fixes up the volume name in place and manages the | |
3773 | * foreign array as an independent entity. | |
3774 | */ | |
3775 | s = NULL; | |
3776 | spare = NULL; | |
3777 | conflict = 0; | |
3778 | for (i = 0; i < tbl_size; i++) { | |
3779 | struct intel_super *tbl_ent = super_table[i]; | |
3780 | int is_spare = 0; | |
3781 | ||
3782 | if (!tbl_ent) | |
3783 | continue; | |
3784 | ||
3785 | if (tbl_ent->anchor->num_raid_devs == 0) { | |
3786 | spare = tbl_ent; | |
3787 | is_spare = 1; | |
3788 | } | |
3789 | ||
3790 | if (s && !is_spare) { | |
3791 | show_conflicts(tbl_ent->anchor->family_num, *super_list); | |
3792 | conflict++; | |
3793 | } else if (!s && !is_spare) | |
3794 | s = tbl_ent; | |
3795 | } | |
3796 | ||
3797 | if (!s) | |
3798 | s = spare; | |
3799 | if (!s) { | |
3800 | champion = NULL; | |
3801 | goto out; | |
3802 | } | |
3803 | champion = s; | |
3804 | ||
3805 | if (conflict) | |
3806 | fprintf(stderr, "Chose family %#x on '%s', " | |
3807 | "assemble conflicts to new container with '--update=uuid'\n", | |
3808 | __le32_to_cpu(s->anchor->family_num), s->disks->devname); | |
3809 | ||
3810 | /* collect all dl's onto 'champion', and update them to | |
3811 | * champion's version of the status | |
3812 | */ | |
3813 | for (s = *super_list; s; s = s->next) { | |
3814 | struct imsm_super *mpb = champion->anchor; | |
3815 | struct dl *dl = s->disks; | |
3816 | ||
3817 | if (s == champion) | |
3818 | continue; | |
3819 | ||
3820 | for (i = 0; i < mpb->num_disks; i++) { | |
3821 | struct imsm_disk *disk; | |
3822 | ||
3823 | disk = __serial_to_disk(dl->serial, mpb, &dl->index); | |
3824 | if (disk) { | |
3825 | dl->disk = *disk; | |
3826 | /* only set index on disks that are a member of | |
3827 | * a populated contianer, i.e. one with | |
3828 | * raid_devs | |
3829 | */ | |
3830 | if (is_failed(&dl->disk)) | |
3831 | dl->index = -2; | |
3832 | else if (is_spare(&dl->disk)) | |
3833 | dl->index = -1; | |
3834 | break; | |
3835 | } | |
3836 | } | |
3837 | ||
3838 | if (i >= mpb->num_disks) { | |
3839 | struct intel_disk *idisk; | |
3840 | ||
3841 | idisk = disk_list_get(dl->serial, disk_list); | |
ecf408e9 | 3842 | if (idisk && is_spare(&idisk->disk) && |
a2b97981 DW |
3843 | !is_failed(&idisk->disk) && !is_configured(&idisk->disk)) |
3844 | dl->index = -1; | |
3845 | else { | |
3846 | dl->index = -2; | |
3847 | continue; | |
3848 | } | |
3849 | } | |
3850 | ||
3851 | dl->next = champion->disks; | |
3852 | champion->disks = dl; | |
3853 | s->disks = NULL; | |
3854 | } | |
3855 | ||
3856 | /* delete 'champion' from super_list */ | |
3857 | for (del = super_list; *del; ) { | |
3858 | if (*del == champion) { | |
3859 | *del = (*del)->next; | |
3860 | break; | |
3861 | } else | |
3862 | del = &(*del)->next; | |
3863 | } | |
3864 | champion->next = NULL; | |
3865 | ||
3866 | out: | |
3867 | while (disk_list) { | |
3868 | struct intel_disk *idisk = disk_list; | |
3869 | ||
3870 | disk_list = disk_list->next; | |
3871 | free(idisk); | |
3872 | } | |
3873 | ||
3874 | return champion; | |
3875 | } | |
3876 | ||
cdddbdbc | 3877 | static int load_super_imsm_all(struct supertype *st, int fd, void **sbp, |
e1902a7b | 3878 | char *devname) |
cdddbdbc DW |
3879 | { |
3880 | struct mdinfo *sra; | |
a2b97981 DW |
3881 | struct intel_super *super_list = NULL; |
3882 | struct intel_super *super = NULL; | |
db575f3b | 3883 | int devnum = fd2devnum(fd); |
a2b97981 | 3884 | struct mdinfo *sd; |
db575f3b | 3885 | int retry; |
a2b97981 DW |
3886 | int err = 0; |
3887 | int i; | |
dab4a513 DW |
3888 | |
3889 | /* check if 'fd' an opened container */ | |
b526e52d | 3890 | sra = sysfs_read(fd, 0, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE); |
cdddbdbc DW |
3891 | if (!sra) |
3892 | return 1; | |
3893 | ||
3894 | if (sra->array.major_version != -1 || | |
3895 | sra->array.minor_version != -2 || | |
1602d52c AW |
3896 | strcmp(sra->text_version, "imsm") != 0) { |
3897 | err = 1; | |
3898 | goto error; | |
3899 | } | |
a2b97981 DW |
3900 | /* load all mpbs */ |
3901 | for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) { | |
49133e57 | 3902 | struct intel_super *s = alloc_super(); |
7a6ecd55 | 3903 | char nm[32]; |
a2b97981 | 3904 | int dfd; |
f2f5c343 | 3905 | int rv; |
a2b97981 DW |
3906 | |
3907 | err = 1; | |
3908 | if (!s) | |
3909 | goto error; | |
3910 | s->next = super_list; | |
3911 | super_list = s; | |
cdddbdbc | 3912 | |
a2b97981 | 3913 | err = 2; |
cdddbdbc | 3914 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); |
e1902a7b | 3915 | dfd = dev_open(nm, O_RDWR); |
a2b97981 DW |
3916 | if (dfd < 0) |
3917 | goto error; | |
3918 | ||
d424212e | 3919 | rv = find_intel_hba_capability(dfd, s, devname); |
f2f5c343 LM |
3920 | /* no orom/efi or non-intel hba of the disk */ |
3921 | if (rv != 0) | |
3922 | goto error; | |
3923 | ||
e1902a7b | 3924 | err = load_and_parse_mpb(dfd, s, NULL, 1); |
db575f3b DW |
3925 | |
3926 | /* retry the load if we might have raced against mdmon */ | |
a2b97981 | 3927 | if (err == 3 && mdmon_running(devnum)) |
db575f3b DW |
3928 | for (retry = 0; retry < 3; retry++) { |
3929 | usleep(3000); | |
e1902a7b | 3930 | err = load_and_parse_mpb(dfd, s, NULL, 1); |
a2b97981 | 3931 | if (err != 3) |
db575f3b DW |
3932 | break; |
3933 | } | |
a2b97981 DW |
3934 | if (err) |
3935 | goto error; | |
cdddbdbc DW |
3936 | } |
3937 | ||
a2b97981 DW |
3938 | /* all mpbs enter, maybe one leaves */ |
3939 | super = imsm_thunderdome(&super_list, i); | |
3940 | if (!super) { | |
3941 | err = 1; | |
3942 | goto error; | |
cdddbdbc DW |
3943 | } |
3944 | ||
47ee5a45 DW |
3945 | if (find_missing(super) != 0) { |
3946 | free_imsm(super); | |
a2b97981 DW |
3947 | err = 2; |
3948 | goto error; | |
47ee5a45 | 3949 | } |
8e59f3d8 AK |
3950 | |
3951 | /* load migration record */ | |
3952 | err = load_imsm_migr_rec(super, NULL); | |
3953 | if (err) { | |
3954 | err = 4; | |
3955 | goto error; | |
3956 | } | |
e2f41b2c AK |
3957 | |
3958 | /* Check migration compatibility */ | |
3959 | if (check_mpb_migr_compatibility(super) != 0) { | |
3960 | fprintf(stderr, Name ": Unsupported migration detected"); | |
3961 | if (devname) | |
3962 | fprintf(stderr, " on %s\n", devname); | |
3963 | else | |
3964 | fprintf(stderr, " (IMSM).\n"); | |
3965 | ||
3966 | err = 5; | |
3967 | goto error; | |
3968 | } | |
3969 | ||
a2b97981 DW |
3970 | err = 0; |
3971 | ||
3972 | error: | |
3973 | while (super_list) { | |
3974 | struct intel_super *s = super_list; | |
3975 | ||
3976 | super_list = super_list->next; | |
3977 | free_imsm(s); | |
3978 | } | |
1602d52c | 3979 | sysfs_free(sra); |
a2b97981 DW |
3980 | |
3981 | if (err) | |
3982 | return err; | |
f7e7067b | 3983 | |
cdddbdbc | 3984 | *sbp = super; |
db575f3b | 3985 | st->container_dev = devnum; |
a2b97981 | 3986 | if (err == 0 && st->ss == NULL) { |
bf5a934a | 3987 | st->ss = &super_imsm; |
cdddbdbc DW |
3988 | st->minor_version = 0; |
3989 | st->max_devs = IMSM_MAX_DEVICES; | |
3990 | } | |
cdddbdbc DW |
3991 | return 0; |
3992 | } | |
2b959fbf N |
3993 | |
3994 | static int load_container_imsm(struct supertype *st, int fd, char *devname) | |
3995 | { | |
3996 | return load_super_imsm_all(st, fd, &st->sb, devname); | |
3997 | } | |
cdddbdbc DW |
3998 | #endif |
3999 | ||
4000 | static int load_super_imsm(struct supertype *st, int fd, char *devname) | |
4001 | { | |
4002 | struct intel_super *super; | |
4003 | int rv; | |
4004 | ||
691c6ee1 N |
4005 | if (test_partition(fd)) |
4006 | /* IMSM not allowed on partitions */ | |
4007 | return 1; | |
4008 | ||
37424f13 DW |
4009 | free_super_imsm(st); |
4010 | ||
49133e57 | 4011 | super = alloc_super(); |
cdddbdbc DW |
4012 | if (!super) { |
4013 | fprintf(stderr, | |
4014 | Name ": malloc of %zu failed.\n", | |
4015 | sizeof(*super)); | |
4016 | return 1; | |
4017 | } | |
ea2bc72b LM |
4018 | /* Load hba and capabilities if they exist. |
4019 | * But do not preclude loading metadata in case capabilities or hba are | |
4020 | * non-compliant and ignore_hw_compat is set. | |
4021 | */ | |
d424212e | 4022 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 4023 | /* no orom/efi or non-intel hba of the disk */ |
ea2bc72b | 4024 | if ((rv != 0) && (st->ignore_hw_compat == 0)) { |
f2f5c343 LM |
4025 | if (devname) |
4026 | fprintf(stderr, | |
4027 | Name ": No OROM/EFI properties for %s\n", devname); | |
4028 | free_imsm(super); | |
4029 | return 2; | |
4030 | } | |
a2b97981 | 4031 | rv = load_and_parse_mpb(fd, super, devname, 0); |
cdddbdbc DW |
4032 | |
4033 | if (rv) { | |
4034 | if (devname) | |
4035 | fprintf(stderr, | |
4036 | Name ": Failed to load all information " | |
4037 | "sections on %s\n", devname); | |
4038 | free_imsm(super); | |
4039 | return rv; | |
4040 | } | |
4041 | ||
4042 | st->sb = super; | |
4043 | if (st->ss == NULL) { | |
4044 | st->ss = &super_imsm; | |
4045 | st->minor_version = 0; | |
4046 | st->max_devs = IMSM_MAX_DEVICES; | |
4047 | } | |
8e59f3d8 AK |
4048 | |
4049 | /* load migration record */ | |
2e062e82 AK |
4050 | if (load_imsm_migr_rec(super, NULL) == 0) { |
4051 | /* Check for unsupported migration features */ | |
4052 | if (check_mpb_migr_compatibility(super) != 0) { | |
4053 | fprintf(stderr, | |
4054 | Name ": Unsupported migration detected"); | |
4055 | if (devname) | |
4056 | fprintf(stderr, " on %s\n", devname); | |
4057 | else | |
4058 | fprintf(stderr, " (IMSM).\n"); | |
4059 | return 3; | |
4060 | } | |
e2f41b2c AK |
4061 | } |
4062 | ||
cdddbdbc DW |
4063 | return 0; |
4064 | } | |
4065 | ||
ef6ffade DW |
4066 | static __u16 info_to_blocks_per_strip(mdu_array_info_t *info) |
4067 | { | |
4068 | if (info->level == 1) | |
4069 | return 128; | |
4070 | return info->chunk_size >> 9; | |
4071 | } | |
4072 | ||
ff596308 | 4073 | static __u32 info_to_num_data_stripes(mdu_array_info_t *info, int num_domains) |
ef6ffade DW |
4074 | { |
4075 | __u32 num_stripes; | |
4076 | ||
4077 | num_stripes = (info->size * 2) / info_to_blocks_per_strip(info); | |
ff596308 | 4078 | num_stripes /= num_domains; |
ef6ffade DW |
4079 | |
4080 | return num_stripes; | |
4081 | } | |
4082 | ||
fcfd9599 DW |
4083 | static __u32 info_to_blocks_per_member(mdu_array_info_t *info) |
4084 | { | |
4025c288 DW |
4085 | if (info->level == 1) |
4086 | return info->size * 2; | |
4087 | else | |
4088 | return (info->size * 2) & ~(info_to_blocks_per_strip(info) - 1); | |
fcfd9599 DW |
4089 | } |
4090 | ||
4d1313e9 DW |
4091 | static void imsm_update_version_info(struct intel_super *super) |
4092 | { | |
4093 | /* update the version and attributes */ | |
4094 | struct imsm_super *mpb = super->anchor; | |
4095 | char *version; | |
4096 | struct imsm_dev *dev; | |
4097 | struct imsm_map *map; | |
4098 | int i; | |
4099 | ||
4100 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
4101 | dev = get_imsm_dev(super, i); | |
4102 | map = get_imsm_map(dev, 0); | |
4103 | if (__le32_to_cpu(dev->size_high) > 0) | |
4104 | mpb->attributes |= MPB_ATTRIB_2TB; | |
4105 | ||
4106 | /* FIXME detect when an array spans a port multiplier */ | |
4107 | #if 0 | |
4108 | mpb->attributes |= MPB_ATTRIB_PM; | |
4109 | #endif | |
4110 | ||
4111 | if (mpb->num_raid_devs > 1 || | |
4112 | mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) { | |
4113 | version = MPB_VERSION_ATTRIBS; | |
4114 | switch (get_imsm_raid_level(map)) { | |
4115 | case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break; | |
4116 | case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break; | |
4117 | case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break; | |
4118 | case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break; | |
4119 | } | |
4120 | } else { | |
4121 | if (map->num_members >= 5) | |
4122 | version = MPB_VERSION_5OR6_DISK_ARRAY; | |
4123 | else if (dev->status == DEV_CLONE_N_GO) | |
4124 | version = MPB_VERSION_CNG; | |
4125 | else if (get_imsm_raid_level(map) == 5) | |
4126 | version = MPB_VERSION_RAID5; | |
4127 | else if (map->num_members >= 3) | |
4128 | version = MPB_VERSION_3OR4_DISK_ARRAY; | |
4129 | else if (get_imsm_raid_level(map) == 1) | |
4130 | version = MPB_VERSION_RAID1; | |
4131 | else | |
4132 | version = MPB_VERSION_RAID0; | |
4133 | } | |
4134 | strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version); | |
4135 | } | |
4136 | } | |
4137 | ||
aa534678 DW |
4138 | static int check_name(struct intel_super *super, char *name, int quiet) |
4139 | { | |
4140 | struct imsm_super *mpb = super->anchor; | |
4141 | char *reason = NULL; | |
4142 | int i; | |
4143 | ||
4144 | if (strlen(name) > MAX_RAID_SERIAL_LEN) | |
4145 | reason = "must be 16 characters or less"; | |
4146 | ||
4147 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
4148 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
4149 | ||
4150 | if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) { | |
4151 | reason = "already exists"; | |
4152 | break; | |
4153 | } | |
4154 | } | |
4155 | ||
4156 | if (reason && !quiet) | |
4157 | fprintf(stderr, Name ": imsm volume name %s\n", reason); | |
4158 | ||
4159 | return !reason; | |
4160 | } | |
4161 | ||
8b353278 DW |
4162 | static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info, |
4163 | unsigned long long size, char *name, | |
4164 | char *homehost, int *uuid) | |
cdddbdbc | 4165 | { |
c2c087e6 DW |
4166 | /* We are creating a volume inside a pre-existing container. |
4167 | * so st->sb is already set. | |
4168 | */ | |
4169 | struct intel_super *super = st->sb; | |
949c47a0 | 4170 | struct imsm_super *mpb = super->anchor; |
ba2de7ba | 4171 | struct intel_dev *dv; |
c2c087e6 DW |
4172 | struct imsm_dev *dev; |
4173 | struct imsm_vol *vol; | |
4174 | struct imsm_map *map; | |
4175 | int idx = mpb->num_raid_devs; | |
4176 | int i; | |
4177 | unsigned long long array_blocks; | |
2c092cad | 4178 | size_t size_old, size_new; |
ff596308 | 4179 | __u32 num_data_stripes; |
cdddbdbc | 4180 | |
88c32bb1 | 4181 | if (super->orom && mpb->num_raid_devs >= super->orom->vpa) { |
c2c087e6 | 4182 | fprintf(stderr, Name": This imsm-container already has the " |
88c32bb1 | 4183 | "maximum of %d volumes\n", super->orom->vpa); |
c2c087e6 DW |
4184 | return 0; |
4185 | } | |
4186 | ||
2c092cad DW |
4187 | /* ensure the mpb is large enough for the new data */ |
4188 | size_old = __le32_to_cpu(mpb->mpb_size); | |
4189 | size_new = disks_to_mpb_size(info->nr_disks); | |
4190 | if (size_new > size_old) { | |
4191 | void *mpb_new; | |
4192 | size_t size_round = ROUND_UP(size_new, 512); | |
4193 | ||
4194 | if (posix_memalign(&mpb_new, 512, size_round) != 0) { | |
4195 | fprintf(stderr, Name": could not allocate new mpb\n"); | |
4196 | return 0; | |
4197 | } | |
8e59f3d8 AK |
4198 | if (posix_memalign(&super->migr_rec_buf, 512, 512) != 0) { |
4199 | fprintf(stderr, Name | |
4200 | ": %s could not allocate migr_rec buffer\n", | |
4201 | __func__); | |
4202 | free(super->buf); | |
4203 | free(super); | |
4204 | return 0; | |
4205 | } | |
2c092cad DW |
4206 | memcpy(mpb_new, mpb, size_old); |
4207 | free(mpb); | |
4208 | mpb = mpb_new; | |
949c47a0 | 4209 | super->anchor = mpb_new; |
2c092cad DW |
4210 | mpb->mpb_size = __cpu_to_le32(size_new); |
4211 | memset(mpb_new + size_old, 0, size_round - size_old); | |
4212 | } | |
bf5a934a | 4213 | super->current_vol = idx; |
3960e579 DW |
4214 | |
4215 | /* handle 'failed_disks' by either: | |
4216 | * a) create dummy disk entries in the table if this the first | |
4217 | * volume in the array. We add them here as this is the only | |
4218 | * opportunity to add them. add_to_super_imsm_volume() | |
4219 | * handles the non-failed disks and continues incrementing | |
4220 | * mpb->num_disks. | |
4221 | * b) validate that 'failed_disks' matches the current number | |
4222 | * of missing disks if the container is populated | |
d23fe947 | 4223 | */ |
3960e579 | 4224 | if (super->current_vol == 0) { |
d23fe947 | 4225 | mpb->num_disks = 0; |
3960e579 DW |
4226 | for (i = 0; i < info->failed_disks; i++) { |
4227 | struct imsm_disk *disk; | |
4228 | ||
4229 | mpb->num_disks++; | |
4230 | disk = __get_imsm_disk(mpb, i); | |
4231 | disk->status = CONFIGURED_DISK | FAILED_DISK; | |
4232 | disk->scsi_id = __cpu_to_le32(~(__u32)0); | |
4233 | snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN, | |
4234 | "missing:%d", i); | |
4235 | } | |
4236 | find_missing(super); | |
4237 | } else { | |
4238 | int missing = 0; | |
4239 | struct dl *d; | |
4240 | ||
4241 | for (d = super->missing; d; d = d->next) | |
4242 | missing++; | |
4243 | if (info->failed_disks > missing) { | |
4244 | fprintf(stderr, Name": unable to add 'missing' disk to container\n"); | |
4245 | return 0; | |
4246 | } | |
4247 | } | |
5a038140 | 4248 | |
aa534678 DW |
4249 | if (!check_name(super, name, 0)) |
4250 | return 0; | |
ba2de7ba DW |
4251 | dv = malloc(sizeof(*dv)); |
4252 | if (!dv) { | |
4253 | fprintf(stderr, Name ": failed to allocate device list entry\n"); | |
4254 | return 0; | |
4255 | } | |
1a2487c2 | 4256 | dev = calloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1)); |
949c47a0 | 4257 | if (!dev) { |
ba2de7ba | 4258 | free(dv); |
949c47a0 DW |
4259 | fprintf(stderr, Name": could not allocate raid device\n"); |
4260 | return 0; | |
4261 | } | |
1a2487c2 | 4262 | |
c2c087e6 | 4263 | strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN); |
03bcbc65 DW |
4264 | if (info->level == 1) |
4265 | array_blocks = info_to_blocks_per_member(info); | |
4266 | else | |
4267 | array_blocks = calc_array_size(info->level, info->raid_disks, | |
4268 | info->layout, info->chunk_size, | |
4269 | info->size*2); | |
979d38be DW |
4270 | /* round array size down to closest MB */ |
4271 | array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; | |
4272 | ||
c2c087e6 DW |
4273 | dev->size_low = __cpu_to_le32((__u32) array_blocks); |
4274 | dev->size_high = __cpu_to_le32((__u32) (array_blocks >> 32)); | |
1a2487c2 | 4275 | dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING); |
c2c087e6 DW |
4276 | vol = &dev->vol; |
4277 | vol->migr_state = 0; | |
1484e727 | 4278 | set_migr_type(dev, MIGR_INIT); |
3960e579 | 4279 | vol->dirty = !info->state; |
f8f603f1 | 4280 | vol->curr_migr_unit = 0; |
a965f303 | 4281 | map = get_imsm_map(dev, 0); |
0dcecb2e | 4282 | map->pba_of_lba0 = __cpu_to_le32(super->create_offset); |
fcfd9599 | 4283 | map->blocks_per_member = __cpu_to_le32(info_to_blocks_per_member(info)); |
ef6ffade | 4284 | map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info)); |
0556e1a2 | 4285 | map->failed_disk_num = ~0; |
3960e579 | 4286 | map->map_state = info->failed_disks ? IMSM_T_STATE_DEGRADED : IMSM_T_STATE_NORMAL; |
252d23c0 | 4287 | map->ddf = 1; |
ef6ffade DW |
4288 | |
4289 | if (info->level == 1 && info->raid_disks > 2) { | |
38950822 AW |
4290 | free(dev); |
4291 | free(dv); | |
ef6ffade DW |
4292 | fprintf(stderr, Name": imsm does not support more than 2 disks" |
4293 | "in a raid1 volume\n"); | |
4294 | return 0; | |
4295 | } | |
81062a36 DW |
4296 | |
4297 | map->raid_level = info->level; | |
4d1313e9 | 4298 | if (info->level == 10) { |
c2c087e6 | 4299 | map->raid_level = 1; |
4d1313e9 | 4300 | map->num_domains = info->raid_disks / 2; |
81062a36 DW |
4301 | } else if (info->level == 1) |
4302 | map->num_domains = info->raid_disks; | |
4303 | else | |
ff596308 | 4304 | map->num_domains = 1; |
81062a36 | 4305 | |
ff596308 DW |
4306 | num_data_stripes = info_to_num_data_stripes(info, map->num_domains); |
4307 | map->num_data_stripes = __cpu_to_le32(num_data_stripes); | |
ef6ffade | 4308 | |
c2c087e6 DW |
4309 | map->num_members = info->raid_disks; |
4310 | for (i = 0; i < map->num_members; i++) { | |
4311 | /* initialized in add_to_super */ | |
4eb26970 | 4312 | set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD); |
c2c087e6 | 4313 | } |
949c47a0 | 4314 | mpb->num_raid_devs++; |
ba2de7ba DW |
4315 | |
4316 | dv->dev = dev; | |
4317 | dv->index = super->current_vol; | |
4318 | dv->next = super->devlist; | |
4319 | super->devlist = dv; | |
c2c087e6 | 4320 | |
4d1313e9 DW |
4321 | imsm_update_version_info(super); |
4322 | ||
c2c087e6 | 4323 | return 1; |
cdddbdbc DW |
4324 | } |
4325 | ||
bf5a934a DW |
4326 | static int init_super_imsm(struct supertype *st, mdu_array_info_t *info, |
4327 | unsigned long long size, char *name, | |
4328 | char *homehost, int *uuid) | |
4329 | { | |
4330 | /* This is primarily called by Create when creating a new array. | |
4331 | * We will then get add_to_super called for each component, and then | |
4332 | * write_init_super called to write it out to each device. | |
4333 | * For IMSM, Create can create on fresh devices or on a pre-existing | |
4334 | * array. | |
4335 | * To create on a pre-existing array a different method will be called. | |
4336 | * This one is just for fresh drives. | |
4337 | */ | |
4338 | struct intel_super *super; | |
4339 | struct imsm_super *mpb; | |
4340 | size_t mpb_size; | |
4d1313e9 | 4341 | char *version; |
bf5a934a | 4342 | |
bf5a934a | 4343 | if (st->sb) |
e683ca88 DW |
4344 | return init_super_imsm_volume(st, info, size, name, homehost, uuid); |
4345 | ||
4346 | if (info) | |
4347 | mpb_size = disks_to_mpb_size(info->nr_disks); | |
4348 | else | |
4349 | mpb_size = 512; | |
bf5a934a | 4350 | |
49133e57 | 4351 | super = alloc_super(); |
e683ca88 | 4352 | if (super && posix_memalign(&super->buf, 512, mpb_size) != 0) { |
bf5a934a | 4353 | free(super); |
e683ca88 DW |
4354 | super = NULL; |
4355 | } | |
4356 | if (!super) { | |
4357 | fprintf(stderr, Name | |
4358 | ": %s could not allocate superblock\n", __func__); | |
bf5a934a DW |
4359 | return 0; |
4360 | } | |
8e59f3d8 AK |
4361 | if (posix_memalign(&super->migr_rec_buf, 512, 512) != 0) { |
4362 | fprintf(stderr, Name | |
4363 | ": %s could not allocate migr_rec buffer\n", __func__); | |
4364 | free(super->buf); | |
4365 | free(super); | |
4366 | return 0; | |
4367 | } | |
e683ca88 | 4368 | memset(super->buf, 0, mpb_size); |
ef649044 | 4369 | mpb = super->buf; |
e683ca88 DW |
4370 | mpb->mpb_size = __cpu_to_le32(mpb_size); |
4371 | st->sb = super; | |
4372 | ||
4373 | if (info == NULL) { | |
4374 | /* zeroing superblock */ | |
4375 | return 0; | |
4376 | } | |
bf5a934a | 4377 | |
4d1313e9 DW |
4378 | mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; |
4379 | ||
4380 | version = (char *) mpb->sig; | |
4381 | strcpy(version, MPB_SIGNATURE); | |
4382 | version += strlen(MPB_SIGNATURE); | |
4383 | strcpy(version, MPB_VERSION_RAID0); | |
bf5a934a | 4384 | |
bf5a934a DW |
4385 | return 1; |
4386 | } | |
4387 | ||
0e600426 | 4388 | #ifndef MDASSEMBLE |
f20c3968 | 4389 | static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk, |
bf5a934a DW |
4390 | int fd, char *devname) |
4391 | { | |
4392 | struct intel_super *super = st->sb; | |
d23fe947 | 4393 | struct imsm_super *mpb = super->anchor; |
3960e579 | 4394 | struct imsm_disk *_disk; |
bf5a934a DW |
4395 | struct imsm_dev *dev; |
4396 | struct imsm_map *map; | |
3960e579 | 4397 | struct dl *dl, *df; |
4eb26970 | 4398 | int slot; |
bf5a934a | 4399 | |
949c47a0 | 4400 | dev = get_imsm_dev(super, super->current_vol); |
a965f303 | 4401 | map = get_imsm_map(dev, 0); |
bf5a934a | 4402 | |
208933a7 N |
4403 | if (! (dk->state & (1<<MD_DISK_SYNC))) { |
4404 | fprintf(stderr, Name ": %s: Cannot add spare devices to IMSM volume\n", | |
4405 | devname); | |
4406 | return 1; | |
4407 | } | |
4408 | ||
efb30e7f DW |
4409 | if (fd == -1) { |
4410 | /* we're doing autolayout so grab the pre-marked (in | |
4411 | * validate_geometry) raid_disk | |
4412 | */ | |
4413 | for (dl = super->disks; dl; dl = dl->next) | |
4414 | if (dl->raiddisk == dk->raid_disk) | |
4415 | break; | |
4416 | } else { | |
4417 | for (dl = super->disks; dl ; dl = dl->next) | |
4418 | if (dl->major == dk->major && | |
4419 | dl->minor == dk->minor) | |
4420 | break; | |
4421 | } | |
d23fe947 | 4422 | |
208933a7 N |
4423 | if (!dl) { |
4424 | fprintf(stderr, Name ": %s is not a member of the same container\n", devname); | |
f20c3968 | 4425 | return 1; |
208933a7 | 4426 | } |
bf5a934a | 4427 | |
d23fe947 DW |
4428 | /* add a pristine spare to the metadata */ |
4429 | if (dl->index < 0) { | |
4430 | dl->index = super->anchor->num_disks; | |
4431 | super->anchor->num_disks++; | |
4432 | } | |
4eb26970 DW |
4433 | /* Check the device has not already been added */ |
4434 | slot = get_imsm_disk_slot(map, dl->index); | |
4435 | if (slot >= 0 && | |
98130f40 | 4436 | (get_imsm_ord_tbl_ent(dev, slot, -1) & IMSM_ORD_REBUILD) == 0) { |
4eb26970 DW |
4437 | fprintf(stderr, Name ": %s has been included in this array twice\n", |
4438 | devname); | |
4439 | return 1; | |
4440 | } | |
656b6b5a | 4441 | set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index); |
ee5aad5a | 4442 | dl->disk.status = CONFIGURED_DISK; |
d23fe947 | 4443 | |
3960e579 DW |
4444 | /* update size of 'missing' disks to be at least as large as the |
4445 | * largest acitve member (we only have dummy missing disks when | |
4446 | * creating the first volume) | |
4447 | */ | |
4448 | if (super->current_vol == 0) { | |
4449 | for (df = super->missing; df; df = df->next) { | |
4450 | if (dl->disk.total_blocks > df->disk.total_blocks) | |
4451 | df->disk.total_blocks = dl->disk.total_blocks; | |
4452 | _disk = __get_imsm_disk(mpb, df->index); | |
4453 | *_disk = df->disk; | |
4454 | } | |
4455 | } | |
4456 | ||
4457 | /* refresh unset/failed slots to point to valid 'missing' entries */ | |
4458 | for (df = super->missing; df; df = df->next) | |
4459 | for (slot = 0; slot < mpb->num_disks; slot++) { | |
4460 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, -1); | |
4461 | ||
4462 | if ((ord & IMSM_ORD_REBUILD) == 0) | |
4463 | continue; | |
4464 | set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD); | |
4465 | dprintf("set slot:%d to missing disk:%d\n", slot, df->index); | |
4466 | break; | |
4467 | } | |
4468 | ||
d23fe947 DW |
4469 | /* if we are creating the first raid device update the family number */ |
4470 | if (super->current_vol == 0) { | |
4471 | __u32 sum; | |
4472 | struct imsm_dev *_dev = __get_imsm_dev(mpb, 0); | |
d23fe947 | 4473 | |
3960e579 | 4474 | _disk = __get_imsm_disk(mpb, dl->index); |
791b666a AW |
4475 | if (!_dev || !_disk) { |
4476 | fprintf(stderr, Name ": BUG mpb setup error\n"); | |
4477 | return 1; | |
4478 | } | |
d23fe947 DW |
4479 | *_dev = *dev; |
4480 | *_disk = dl->disk; | |
148acb7b DW |
4481 | sum = random32(); |
4482 | sum += __gen_imsm_checksum(mpb); | |
d23fe947 | 4483 | mpb->family_num = __cpu_to_le32(sum); |
148acb7b | 4484 | mpb->orig_family_num = mpb->family_num; |
d23fe947 | 4485 | } |
ca0748fa | 4486 | super->current_disk = dl; |
f20c3968 | 4487 | return 0; |
bf5a934a DW |
4488 | } |
4489 | ||
a8619d23 AK |
4490 | /* mark_spare() |
4491 | * Function marks disk as spare and restores disk serial | |
4492 | * in case it was previously marked as failed by takeover operation | |
4493 | * reruns: | |
4494 | * -1 : critical error | |
4495 | * 0 : disk is marked as spare but serial is not set | |
4496 | * 1 : success | |
4497 | */ | |
4498 | int mark_spare(struct dl *disk) | |
4499 | { | |
4500 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
4501 | int ret_val = -1; | |
4502 | ||
4503 | if (!disk) | |
4504 | return ret_val; | |
4505 | ||
4506 | ret_val = 0; | |
4507 | if (!imsm_read_serial(disk->fd, NULL, serial)) { | |
4508 | /* Restore disk serial number, because takeover marks disk | |
4509 | * as failed and adds to serial ':0' before it becomes | |
4510 | * a spare disk. | |
4511 | */ | |
4512 | serialcpy(disk->serial, serial); | |
4513 | serialcpy(disk->disk.serial, serial); | |
4514 | ret_val = 1; | |
4515 | } | |
4516 | disk->disk.status = SPARE_DISK; | |
4517 | disk->index = -1; | |
4518 | ||
4519 | return ret_val; | |
4520 | } | |
88654014 | 4521 | |
f20c3968 | 4522 | static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk, |
88654014 | 4523 | int fd, char *devname) |
cdddbdbc | 4524 | { |
c2c087e6 | 4525 | struct intel_super *super = st->sb; |
c2c087e6 DW |
4526 | struct dl *dd; |
4527 | unsigned long long size; | |
f2f27e63 | 4528 | __u32 id; |
c2c087e6 DW |
4529 | int rv; |
4530 | struct stat stb; | |
4531 | ||
88654014 LM |
4532 | /* If we are on an RAID enabled platform check that the disk is |
4533 | * attached to the raid controller. | |
4534 | * We do not need to test disks attachment for container based additions, | |
4535 | * they shall be already tested when container was created/assembled. | |
88c32bb1 | 4536 | */ |
d424212e | 4537 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 4538 | /* no orom/efi or non-intel hba of the disk */ |
f0f5a016 LM |
4539 | if (rv != 0) { |
4540 | dprintf("capability: %p fd: %d ret: %d\n", | |
4541 | super->orom, fd, rv); | |
4542 | return 1; | |
88c32bb1 DW |
4543 | } |
4544 | ||
f20c3968 DW |
4545 | if (super->current_vol >= 0) |
4546 | return add_to_super_imsm_volume(st, dk, fd, devname); | |
bf5a934a | 4547 | |
c2c087e6 DW |
4548 | fstat(fd, &stb); |
4549 | dd = malloc(sizeof(*dd)); | |
b9f594fe | 4550 | if (!dd) { |
c2c087e6 DW |
4551 | fprintf(stderr, |
4552 | Name ": malloc failed %s:%d.\n", __func__, __LINE__); | |
f20c3968 | 4553 | return 1; |
c2c087e6 DW |
4554 | } |
4555 | memset(dd, 0, sizeof(*dd)); | |
4556 | dd->major = major(stb.st_rdev); | |
4557 | dd->minor = minor(stb.st_rdev); | |
c2c087e6 | 4558 | dd->devname = devname ? strdup(devname) : NULL; |
c2c087e6 | 4559 | dd->fd = fd; |
689c9bf3 | 4560 | dd->e = NULL; |
1a64be56 | 4561 | dd->action = DISK_ADD; |
c2c087e6 | 4562 | rv = imsm_read_serial(fd, devname, dd->serial); |
32ba9157 | 4563 | if (rv) { |
c2c087e6 | 4564 | fprintf(stderr, |
0030e8d6 | 4565 | Name ": failed to retrieve scsi serial, aborting\n"); |
949c47a0 | 4566 | free(dd); |
0030e8d6 | 4567 | abort(); |
c2c087e6 DW |
4568 | } |
4569 | ||
c2c087e6 DW |
4570 | get_dev_size(fd, NULL, &size); |
4571 | size /= 512; | |
1f24f035 | 4572 | serialcpy(dd->disk.serial, dd->serial); |
b9f594fe | 4573 | dd->disk.total_blocks = __cpu_to_le32(size); |
a8619d23 | 4574 | mark_spare(dd); |
c2c087e6 | 4575 | if (sysfs_disk_to_scsi_id(fd, &id) == 0) |
b9f594fe | 4576 | dd->disk.scsi_id = __cpu_to_le32(id); |
c2c087e6 | 4577 | else |
b9f594fe | 4578 | dd->disk.scsi_id = __cpu_to_le32(0); |
43dad3d6 DW |
4579 | |
4580 | if (st->update_tail) { | |
1a64be56 LM |
4581 | dd->next = super->disk_mgmt_list; |
4582 | super->disk_mgmt_list = dd; | |
43dad3d6 DW |
4583 | } else { |
4584 | dd->next = super->disks; | |
4585 | super->disks = dd; | |
ceaf0ee1 | 4586 | super->updates_pending++; |
43dad3d6 | 4587 | } |
f20c3968 DW |
4588 | |
4589 | return 0; | |
cdddbdbc DW |
4590 | } |
4591 | ||
1a64be56 LM |
4592 | |
4593 | static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk) | |
4594 | { | |
4595 | struct intel_super *super = st->sb; | |
4596 | struct dl *dd; | |
4597 | ||
4598 | /* remove from super works only in mdmon - for communication | |
4599 | * manager - monitor. Check if communication memory buffer | |
4600 | * is prepared. | |
4601 | */ | |
4602 | if (!st->update_tail) { | |
4603 | fprintf(stderr, | |
4604 | Name ": %s shall be used in mdmon context only" | |
4605 | "(line %d).\n", __func__, __LINE__); | |
4606 | return 1; | |
4607 | } | |
4608 | dd = malloc(sizeof(*dd)); | |
4609 | if (!dd) { | |
4610 | fprintf(stderr, | |
4611 | Name ": malloc failed %s:%d.\n", __func__, __LINE__); | |
4612 | return 1; | |
4613 | } | |
4614 | memset(dd, 0, sizeof(*dd)); | |
4615 | dd->major = dk->major; | |
4616 | dd->minor = dk->minor; | |
1a64be56 | 4617 | dd->fd = -1; |
a8619d23 | 4618 | mark_spare(dd); |
1a64be56 LM |
4619 | dd->action = DISK_REMOVE; |
4620 | ||
4621 | dd->next = super->disk_mgmt_list; | |
4622 | super->disk_mgmt_list = dd; | |
4623 | ||
4624 | ||
4625 | return 0; | |
4626 | } | |
4627 | ||
f796af5d DW |
4628 | static int store_imsm_mpb(int fd, struct imsm_super *mpb); |
4629 | ||
4630 | static union { | |
4631 | char buf[512]; | |
4632 | struct imsm_super anchor; | |
4633 | } spare_record __attribute__ ((aligned(512))); | |
c2c087e6 | 4634 | |
d23fe947 DW |
4635 | /* spare records have their own family number and do not have any defined raid |
4636 | * devices | |
4637 | */ | |
4638 | static int write_super_imsm_spares(struct intel_super *super, int doclose) | |
4639 | { | |
d23fe947 | 4640 | struct imsm_super *mpb = super->anchor; |
f796af5d | 4641 | struct imsm_super *spare = &spare_record.anchor; |
d23fe947 DW |
4642 | __u32 sum; |
4643 | struct dl *d; | |
4644 | ||
f796af5d DW |
4645 | spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super)), |
4646 | spare->generation_num = __cpu_to_le32(1UL), | |
4647 | spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; | |
4648 | spare->num_disks = 1, | |
4649 | spare->num_raid_devs = 0, | |
4650 | spare->cache_size = mpb->cache_size, | |
4651 | spare->pwr_cycle_count = __cpu_to_le32(1), | |
4652 | ||
4653 | snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH, | |
4654 | MPB_SIGNATURE MPB_VERSION_RAID0); | |
d23fe947 DW |
4655 | |
4656 | for (d = super->disks; d; d = d->next) { | |
8796fdc4 | 4657 | if (d->index != -1) |
d23fe947 DW |
4658 | continue; |
4659 | ||
f796af5d DW |
4660 | spare->disk[0] = d->disk; |
4661 | sum = __gen_imsm_checksum(spare); | |
4662 | spare->family_num = __cpu_to_le32(sum); | |
4663 | spare->orig_family_num = 0; | |
4664 | sum = __gen_imsm_checksum(spare); | |
4665 | spare->check_sum = __cpu_to_le32(sum); | |
d23fe947 | 4666 | |
f796af5d | 4667 | if (store_imsm_mpb(d->fd, spare)) { |
d23fe947 DW |
4668 | fprintf(stderr, "%s: failed for device %d:%d %s\n", |
4669 | __func__, d->major, d->minor, strerror(errno)); | |
e74255d9 | 4670 | return 1; |
d23fe947 DW |
4671 | } |
4672 | if (doclose) { | |
4673 | close(d->fd); | |
4674 | d->fd = -1; | |
4675 | } | |
4676 | } | |
4677 | ||
e74255d9 | 4678 | return 0; |
d23fe947 DW |
4679 | } |
4680 | ||
36988a3d | 4681 | static int write_super_imsm(struct supertype *st, int doclose) |
cdddbdbc | 4682 | { |
36988a3d | 4683 | struct intel_super *super = st->sb; |
949c47a0 | 4684 | struct imsm_super *mpb = super->anchor; |
c2c087e6 DW |
4685 | struct dl *d; |
4686 | __u32 generation; | |
4687 | __u32 sum; | |
d23fe947 | 4688 | int spares = 0; |
949c47a0 | 4689 | int i; |
a48ac0a8 | 4690 | __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk); |
36988a3d | 4691 | int num_disks = 0; |
146c6260 | 4692 | int clear_migration_record = 1; |
cdddbdbc | 4693 | |
c2c087e6 DW |
4694 | /* 'generation' is incremented everytime the metadata is written */ |
4695 | generation = __le32_to_cpu(mpb->generation_num); | |
4696 | generation++; | |
4697 | mpb->generation_num = __cpu_to_le32(generation); | |
4698 | ||
148acb7b DW |
4699 | /* fix up cases where previous mdadm releases failed to set |
4700 | * orig_family_num | |
4701 | */ | |
4702 | if (mpb->orig_family_num == 0) | |
4703 | mpb->orig_family_num = mpb->family_num; | |
4704 | ||
d23fe947 | 4705 | for (d = super->disks; d; d = d->next) { |
8796fdc4 | 4706 | if (d->index == -1) |
d23fe947 | 4707 | spares++; |
36988a3d | 4708 | else { |
d23fe947 | 4709 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
4710 | num_disks++; |
4711 | } | |
d23fe947 | 4712 | } |
36988a3d | 4713 | for (d = super->missing; d; d = d->next) { |
47ee5a45 | 4714 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
4715 | num_disks++; |
4716 | } | |
4717 | mpb->num_disks = num_disks; | |
4718 | mpb_size += sizeof(struct imsm_disk) * mpb->num_disks; | |
b9f594fe | 4719 | |
949c47a0 DW |
4720 | for (i = 0; i < mpb->num_raid_devs; i++) { |
4721 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
36988a3d AK |
4722 | struct imsm_dev *dev2 = get_imsm_dev(super, i); |
4723 | if (dev && dev2) { | |
4724 | imsm_copy_dev(dev, dev2); | |
4725 | mpb_size += sizeof_imsm_dev(dev, 0); | |
4726 | } | |
146c6260 AK |
4727 | if (is_gen_migration(dev2)) |
4728 | clear_migration_record = 0; | |
949c47a0 | 4729 | } |
a48ac0a8 DW |
4730 | mpb_size += __le32_to_cpu(mpb->bbm_log_size); |
4731 | mpb->mpb_size = __cpu_to_le32(mpb_size); | |
949c47a0 | 4732 | |
c2c087e6 | 4733 | /* recalculate checksum */ |
949c47a0 | 4734 | sum = __gen_imsm_checksum(mpb); |
c2c087e6 DW |
4735 | mpb->check_sum = __cpu_to_le32(sum); |
4736 | ||
146c6260 AK |
4737 | if (clear_migration_record) |
4738 | memset(super->migr_rec_buf, 0, 512); | |
4739 | ||
d23fe947 | 4740 | /* write the mpb for disks that compose raid devices */ |
c2c087e6 | 4741 | for (d = super->disks; d ; d = d->next) { |
86c54047 | 4742 | if (d->index < 0 || is_failed(&d->disk)) |
d23fe947 | 4743 | continue; |
f796af5d | 4744 | if (store_imsm_mpb(d->fd, mpb)) |
c2c087e6 DW |
4745 | fprintf(stderr, "%s: failed for device %d:%d %s\n", |
4746 | __func__, d->major, d->minor, strerror(errno)); | |
146c6260 AK |
4747 | if (clear_migration_record) { |
4748 | unsigned long long dsize; | |
4749 | ||
4750 | get_dev_size(d->fd, NULL, &dsize); | |
4751 | if (lseek64(d->fd, dsize - 512, SEEK_SET) >= 0) { | |
9e2d750d N |
4752 | if (write(d->fd, super->migr_rec_buf, 512) != 512) |
4753 | perror("Write migr_rec failed"); | |
146c6260 AK |
4754 | } |
4755 | } | |
c2c087e6 DW |
4756 | if (doclose) { |
4757 | close(d->fd); | |
4758 | d->fd = -1; | |
4759 | } | |
4760 | } | |
4761 | ||
d23fe947 DW |
4762 | if (spares) |
4763 | return write_super_imsm_spares(super, doclose); | |
4764 | ||
e74255d9 | 4765 | return 0; |
c2c087e6 DW |
4766 | } |
4767 | ||
0e600426 | 4768 | |
9b1fb677 | 4769 | static int create_array(struct supertype *st, int dev_idx) |
43dad3d6 DW |
4770 | { |
4771 | size_t len; | |
4772 | struct imsm_update_create_array *u; | |
4773 | struct intel_super *super = st->sb; | |
9b1fb677 | 4774 | struct imsm_dev *dev = get_imsm_dev(super, dev_idx); |
54c2c1ea DW |
4775 | struct imsm_map *map = get_imsm_map(dev, 0); |
4776 | struct disk_info *inf; | |
4777 | struct imsm_disk *disk; | |
4778 | int i; | |
43dad3d6 | 4779 | |
54c2c1ea DW |
4780 | len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) + |
4781 | sizeof(*inf) * map->num_members; | |
43dad3d6 DW |
4782 | u = malloc(len); |
4783 | if (!u) { | |
4784 | fprintf(stderr, "%s: failed to allocate update buffer\n", | |
4785 | __func__); | |
4786 | return 1; | |
4787 | } | |
4788 | ||
4789 | u->type = update_create_array; | |
9b1fb677 | 4790 | u->dev_idx = dev_idx; |
43dad3d6 | 4791 | imsm_copy_dev(&u->dev, dev); |
54c2c1ea DW |
4792 | inf = get_disk_info(u); |
4793 | for (i = 0; i < map->num_members; i++) { | |
98130f40 | 4794 | int idx = get_imsm_disk_idx(dev, i, -1); |
9b1fb677 | 4795 | |
54c2c1ea DW |
4796 | disk = get_imsm_disk(super, idx); |
4797 | serialcpy(inf[i].serial, disk->serial); | |
4798 | } | |
43dad3d6 DW |
4799 | append_metadata_update(st, u, len); |
4800 | ||
4801 | return 0; | |
4802 | } | |
4803 | ||
1a64be56 | 4804 | static int mgmt_disk(struct supertype *st) |
43dad3d6 DW |
4805 | { |
4806 | struct intel_super *super = st->sb; | |
4807 | size_t len; | |
1a64be56 | 4808 | struct imsm_update_add_remove_disk *u; |
43dad3d6 | 4809 | |
1a64be56 | 4810 | if (!super->disk_mgmt_list) |
43dad3d6 DW |
4811 | return 0; |
4812 | ||
4813 | len = sizeof(*u); | |
4814 | u = malloc(len); | |
4815 | if (!u) { | |
4816 | fprintf(stderr, "%s: failed to allocate update buffer\n", | |
4817 | __func__); | |
4818 | return 1; | |
4819 | } | |
4820 | ||
1a64be56 | 4821 | u->type = update_add_remove_disk; |
43dad3d6 DW |
4822 | append_metadata_update(st, u, len); |
4823 | ||
4824 | return 0; | |
4825 | } | |
4826 | ||
c2c087e6 DW |
4827 | static int write_init_super_imsm(struct supertype *st) |
4828 | { | |
9b1fb677 DW |
4829 | struct intel_super *super = st->sb; |
4830 | int current_vol = super->current_vol; | |
4831 | ||
4832 | /* we are done with current_vol reset it to point st at the container */ | |
4833 | super->current_vol = -1; | |
4834 | ||
8273f55e | 4835 | if (st->update_tail) { |
43dad3d6 DW |
4836 | /* queue the recently created array / added disk |
4837 | * as a metadata update */ | |
43dad3d6 | 4838 | int rv; |
8273f55e | 4839 | |
43dad3d6 | 4840 | /* determine if we are creating a volume or adding a disk */ |
9b1fb677 | 4841 | if (current_vol < 0) { |
1a64be56 LM |
4842 | /* in the mgmt (add/remove) disk case we are running |
4843 | * in mdmon context, so don't close fd's | |
43dad3d6 | 4844 | */ |
1a64be56 | 4845 | return mgmt_disk(st); |
43dad3d6 | 4846 | } else |
9b1fb677 | 4847 | rv = create_array(st, current_vol); |
8273f55e | 4848 | |
43dad3d6 | 4849 | return rv; |
d682f344 N |
4850 | } else { |
4851 | struct dl *d; | |
4852 | for (d = super->disks; d; d = d->next) | |
4853 | Kill(d->devname, NULL, 0, 1, 1); | |
36988a3d | 4854 | return write_super_imsm(st, 1); |
d682f344 | 4855 | } |
cdddbdbc | 4856 | } |
0e600426 | 4857 | #endif |
cdddbdbc | 4858 | |
e683ca88 | 4859 | static int store_super_imsm(struct supertype *st, int fd) |
cdddbdbc | 4860 | { |
e683ca88 DW |
4861 | struct intel_super *super = st->sb; |
4862 | struct imsm_super *mpb = super ? super->anchor : NULL; | |
551c80c1 | 4863 | |
e683ca88 | 4864 | if (!mpb) |
ad97895e DW |
4865 | return 1; |
4866 | ||
1799c9e8 | 4867 | #ifndef MDASSEMBLE |
e683ca88 | 4868 | return store_imsm_mpb(fd, mpb); |
1799c9e8 N |
4869 | #else |
4870 | return 1; | |
4871 | #endif | |
cdddbdbc DW |
4872 | } |
4873 | ||
0e600426 N |
4874 | static int imsm_bbm_log_size(struct imsm_super *mpb) |
4875 | { | |
4876 | return __le32_to_cpu(mpb->bbm_log_size); | |
4877 | } | |
4878 | ||
4879 | #ifndef MDASSEMBLE | |
cdddbdbc DW |
4880 | static int validate_geometry_imsm_container(struct supertype *st, int level, |
4881 | int layout, int raiddisks, int chunk, | |
c2c087e6 | 4882 | unsigned long long size, char *dev, |
2c514b71 NB |
4883 | unsigned long long *freesize, |
4884 | int verbose) | |
cdddbdbc | 4885 | { |
c2c087e6 DW |
4886 | int fd; |
4887 | unsigned long long ldsize; | |
f2f5c343 LM |
4888 | struct intel_super *super=NULL; |
4889 | int rv = 0; | |
cdddbdbc | 4890 | |
c2c087e6 DW |
4891 | if (level != LEVEL_CONTAINER) |
4892 | return 0; | |
4893 | if (!dev) | |
4894 | return 1; | |
4895 | ||
4896 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
4897 | if (fd < 0) { | |
2c514b71 NB |
4898 | if (verbose) |
4899 | fprintf(stderr, Name ": imsm: Cannot open %s: %s\n", | |
4900 | dev, strerror(errno)); | |
c2c087e6 DW |
4901 | return 0; |
4902 | } | |
4903 | if (!get_dev_size(fd, dev, &ldsize)) { | |
4904 | close(fd); | |
4905 | return 0; | |
4906 | } | |
f2f5c343 LM |
4907 | |
4908 | /* capabilities retrieve could be possible | |
4909 | * note that there is no fd for the disks in array. | |
4910 | */ | |
4911 | super = alloc_super(); | |
4912 | if (!super) { | |
4913 | fprintf(stderr, | |
4914 | Name ": malloc of %zu failed.\n", | |
4915 | sizeof(*super)); | |
4916 | close(fd); | |
4917 | return 0; | |
4918 | } | |
4919 | ||
d424212e | 4920 | rv = find_intel_hba_capability(fd, super, verbose ? dev : NULL); |
f2f5c343 LM |
4921 | if (rv != 0) { |
4922 | #if DEBUG | |
4923 | char str[256]; | |
4924 | fd2devname(fd, str); | |
4925 | dprintf("validate_geometry_imsm_container: fd: %d %s orom: %p rv: %d raiddisk: %d\n", | |
4926 | fd, str, super->orom, rv, raiddisks); | |
4927 | #endif | |
4928 | /* no orom/efi or non-intel hba of the disk */ | |
4929 | close(fd); | |
4930 | free_imsm(super); | |
4931 | return 0; | |
4932 | } | |
c2c087e6 | 4933 | close(fd); |
f2f5c343 LM |
4934 | if (super->orom && raiddisks > super->orom->tds) { |
4935 | if (verbose) | |
4936 | fprintf(stderr, Name ": %d exceeds maximum number of" | |
4937 | " platform supported disks: %d\n", | |
4938 | raiddisks, super->orom->tds); | |
4939 | ||
4940 | free_imsm(super); | |
4941 | return 0; | |
4942 | } | |
c2c087e6 DW |
4943 | |
4944 | *freesize = avail_size_imsm(st, ldsize >> 9); | |
f2f5c343 | 4945 | free_imsm(super); |
c2c087e6 DW |
4946 | |
4947 | return 1; | |
cdddbdbc DW |
4948 | } |
4949 | ||
0dcecb2e DW |
4950 | static unsigned long long find_size(struct extent *e, int *idx, int num_extents) |
4951 | { | |
4952 | const unsigned long long base_start = e[*idx].start; | |
4953 | unsigned long long end = base_start + e[*idx].size; | |
4954 | int i; | |
4955 | ||
4956 | if (base_start == end) | |
4957 | return 0; | |
4958 | ||
4959 | *idx = *idx + 1; | |
4960 | for (i = *idx; i < num_extents; i++) { | |
4961 | /* extend overlapping extents */ | |
4962 | if (e[i].start >= base_start && | |
4963 | e[i].start <= end) { | |
4964 | if (e[i].size == 0) | |
4965 | return 0; | |
4966 | if (e[i].start + e[i].size > end) | |
4967 | end = e[i].start + e[i].size; | |
4968 | } else if (e[i].start > end) { | |
4969 | *idx = i; | |
4970 | break; | |
4971 | } | |
4972 | } | |
4973 | ||
4974 | return end - base_start; | |
4975 | } | |
4976 | ||
4977 | static unsigned long long merge_extents(struct intel_super *super, int sum_extents) | |
4978 | { | |
4979 | /* build a composite disk with all known extents and generate a new | |
4980 | * 'maxsize' given the "all disks in an array must share a common start | |
4981 | * offset" constraint | |
4982 | */ | |
4983 | struct extent *e = calloc(sum_extents, sizeof(*e)); | |
4984 | struct dl *dl; | |
4985 | int i, j; | |
4986 | int start_extent; | |
4987 | unsigned long long pos; | |
b9d77223 | 4988 | unsigned long long start = 0; |
0dcecb2e DW |
4989 | unsigned long long maxsize; |
4990 | unsigned long reserve; | |
4991 | ||
4992 | if (!e) | |
a7dd165b | 4993 | return 0; |
0dcecb2e DW |
4994 | |
4995 | /* coalesce and sort all extents. also, check to see if we need to | |
4996 | * reserve space between member arrays | |
4997 | */ | |
4998 | j = 0; | |
4999 | for (dl = super->disks; dl; dl = dl->next) { | |
5000 | if (!dl->e) | |
5001 | continue; | |
5002 | for (i = 0; i < dl->extent_cnt; i++) | |
5003 | e[j++] = dl->e[i]; | |
5004 | } | |
5005 | qsort(e, sum_extents, sizeof(*e), cmp_extent); | |
5006 | ||
5007 | /* merge extents */ | |
5008 | i = 0; | |
5009 | j = 0; | |
5010 | while (i < sum_extents) { | |
5011 | e[j].start = e[i].start; | |
5012 | e[j].size = find_size(e, &i, sum_extents); | |
5013 | j++; | |
5014 | if (e[j-1].size == 0) | |
5015 | break; | |
5016 | } | |
5017 | ||
5018 | pos = 0; | |
5019 | maxsize = 0; | |
5020 | start_extent = 0; | |
5021 | i = 0; | |
5022 | do { | |
5023 | unsigned long long esize; | |
5024 | ||
5025 | esize = e[i].start - pos; | |
5026 | if (esize >= maxsize) { | |
5027 | maxsize = esize; | |
5028 | start = pos; | |
5029 | start_extent = i; | |
5030 | } | |
5031 | pos = e[i].start + e[i].size; | |
5032 | i++; | |
5033 | } while (e[i-1].size); | |
5034 | free(e); | |
5035 | ||
a7dd165b DW |
5036 | if (maxsize == 0) |
5037 | return 0; | |
5038 | ||
5039 | /* FIXME assumes volume at offset 0 is the first volume in a | |
5040 | * container | |
5041 | */ | |
0dcecb2e DW |
5042 | if (start_extent > 0) |
5043 | reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */ | |
5044 | else | |
5045 | reserve = 0; | |
5046 | ||
5047 | if (maxsize < reserve) | |
a7dd165b | 5048 | return 0; |
0dcecb2e DW |
5049 | |
5050 | super->create_offset = ~((__u32) 0); | |
5051 | if (start + reserve > super->create_offset) | |
a7dd165b | 5052 | return 0; /* start overflows create_offset */ |
0dcecb2e DW |
5053 | super->create_offset = start + reserve; |
5054 | ||
5055 | return maxsize - reserve; | |
5056 | } | |
5057 | ||
88c32bb1 DW |
5058 | static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks) |
5059 | { | |
5060 | if (level < 0 || level == 6 || level == 4) | |
5061 | return 0; | |
5062 | ||
5063 | /* if we have an orom prevent invalid raid levels */ | |
5064 | if (orom) | |
5065 | switch (level) { | |
5066 | case 0: return imsm_orom_has_raid0(orom); | |
5067 | case 1: | |
5068 | if (raiddisks > 2) | |
5069 | return imsm_orom_has_raid1e(orom); | |
1c556e92 DW |
5070 | return imsm_orom_has_raid1(orom) && raiddisks == 2; |
5071 | case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4; | |
5072 | case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2; | |
88c32bb1 DW |
5073 | } |
5074 | else | |
5075 | return 1; /* not on an Intel RAID platform so anything goes */ | |
5076 | ||
5077 | return 0; | |
5078 | } | |
5079 | ||
cd9d1ac7 DW |
5080 | static int imsm_default_chunk(const struct imsm_orom *orom) |
5081 | { | |
5082 | /* up to 512 if the plaform supports it, otherwise the platform max. | |
5083 | * 128 if no platform detected | |
5084 | */ | |
5085 | int fs = max(7, orom ? fls(orom->sss) : 0); | |
5086 | ||
5087 | return min(512, (1 << fs)); | |
5088 | } | |
73408129 | 5089 | |
35f81cbb | 5090 | #define pr_vrb(fmt, arg...) (void) (verbose && fprintf(stderr, Name fmt, ##arg)) |
6592ce37 DW |
5091 | static int |
5092 | validate_geometry_imsm_orom(struct intel_super *super, int level, int layout, | |
c21e737b | 5093 | int raiddisks, int *chunk, int verbose) |
6592ce37 | 5094 | { |
660260d0 DW |
5095 | /* check/set platform and metadata limits/defaults */ |
5096 | if (super->orom && raiddisks > super->orom->dpa) { | |
5097 | pr_vrb(": platform supports a maximum of %d disks per array\n", | |
5098 | super->orom->dpa); | |
73408129 LM |
5099 | return 0; |
5100 | } | |
5101 | ||
5102 | /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */ | |
660260d0 | 5103 | if (!is_raid_level_supported(super->orom, level, raiddisks)) { |
6592ce37 DW |
5104 | pr_vrb(": platform does not support raid%d with %d disk%s\n", |
5105 | level, raiddisks, raiddisks > 1 ? "s" : ""); | |
5106 | return 0; | |
5107 | } | |
cd9d1ac7 DW |
5108 | |
5109 | if (chunk && (*chunk == 0 || *chunk == UnSet)) | |
5110 | *chunk = imsm_default_chunk(super->orom); | |
5111 | ||
5112 | if (super->orom && chunk && !imsm_orom_has_chunk(super->orom, *chunk)) { | |
5113 | pr_vrb(": platform does not support a chunk size of: " | |
5114 | "%d\n", *chunk); | |
5115 | return 0; | |
6592ce37 | 5116 | } |
cd9d1ac7 | 5117 | |
6592ce37 DW |
5118 | if (layout != imsm_level_to_layout(level)) { |
5119 | if (level == 5) | |
5120 | pr_vrb(": imsm raid 5 only supports the left-asymmetric layout\n"); | |
5121 | else if (level == 10) | |
5122 | pr_vrb(": imsm raid 10 only supports the n2 layout\n"); | |
5123 | else | |
5124 | pr_vrb(": imsm unknown layout %#x for this raid level %d\n", | |
5125 | layout, level); | |
5126 | return 0; | |
5127 | } | |
6592ce37 DW |
5128 | return 1; |
5129 | } | |
5130 | ||
c2c087e6 DW |
5131 | /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd |
5132 | * FIX ME add ahci details | |
5133 | */ | |
8b353278 | 5134 | static int validate_geometry_imsm_volume(struct supertype *st, int level, |
c21e737b | 5135 | int layout, int raiddisks, int *chunk, |
c2c087e6 | 5136 | unsigned long long size, char *dev, |
2c514b71 NB |
5137 | unsigned long long *freesize, |
5138 | int verbose) | |
cdddbdbc | 5139 | { |
c2c087e6 DW |
5140 | struct stat stb; |
5141 | struct intel_super *super = st->sb; | |
a20d2ba5 | 5142 | struct imsm_super *mpb = super->anchor; |
c2c087e6 DW |
5143 | struct dl *dl; |
5144 | unsigned long long pos = 0; | |
5145 | unsigned long long maxsize; | |
5146 | struct extent *e; | |
5147 | int i; | |
cdddbdbc | 5148 | |
88c32bb1 DW |
5149 | /* We must have the container info already read in. */ |
5150 | if (!super) | |
c2c087e6 DW |
5151 | return 0; |
5152 | ||
e7cb06c8 LO |
5153 | if (mpb->num_raid_devs > 0 && mpb->num_disks != raiddisks) { |
5154 | fprintf(stderr, Name ": the option-rom requires all " | |
5155 | "member disks to be a member of all volumes.\n"); | |
5156 | return 0; | |
5157 | } | |
5158 | ||
d54559f0 LM |
5159 | if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, verbose)) { |
5160 | fprintf(stderr, Name ": RAID gemetry validation failed. " | |
5161 | "Cannot proceed with the action(s).\n"); | |
c2c087e6 | 5162 | return 0; |
d54559f0 | 5163 | } |
c2c087e6 DW |
5164 | if (!dev) { |
5165 | /* General test: make sure there is space for | |
2da8544a DW |
5166 | * 'raiddisks' device extents of size 'size' at a given |
5167 | * offset | |
c2c087e6 | 5168 | */ |
e46273eb | 5169 | unsigned long long minsize = size; |
b7528a20 | 5170 | unsigned long long start_offset = MaxSector; |
c2c087e6 DW |
5171 | int dcnt = 0; |
5172 | if (minsize == 0) | |
5173 | minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
5174 | for (dl = super->disks; dl ; dl = dl->next) { | |
5175 | int found = 0; | |
5176 | ||
bf5a934a | 5177 | pos = 0; |
c2c087e6 DW |
5178 | i = 0; |
5179 | e = get_extents(super, dl); | |
5180 | if (!e) continue; | |
5181 | do { | |
5182 | unsigned long long esize; | |
5183 | esize = e[i].start - pos; | |
5184 | if (esize >= minsize) | |
5185 | found = 1; | |
b7528a20 | 5186 | if (found && start_offset == MaxSector) { |
2da8544a DW |
5187 | start_offset = pos; |
5188 | break; | |
5189 | } else if (found && pos != start_offset) { | |
5190 | found = 0; | |
5191 | break; | |
5192 | } | |
c2c087e6 DW |
5193 | pos = e[i].start + e[i].size; |
5194 | i++; | |
5195 | } while (e[i-1].size); | |
5196 | if (found) | |
5197 | dcnt++; | |
5198 | free(e); | |
5199 | } | |
5200 | if (dcnt < raiddisks) { | |
2c514b71 NB |
5201 | if (verbose) |
5202 | fprintf(stderr, Name ": imsm: Not enough " | |
5203 | "devices with space for this array " | |
5204 | "(%d < %d)\n", | |
5205 | dcnt, raiddisks); | |
c2c087e6 DW |
5206 | return 0; |
5207 | } | |
5208 | return 1; | |
5209 | } | |
0dcecb2e | 5210 | |
c2c087e6 DW |
5211 | /* This device must be a member of the set */ |
5212 | if (stat(dev, &stb) < 0) | |
5213 | return 0; | |
5214 | if ((S_IFMT & stb.st_mode) != S_IFBLK) | |
5215 | return 0; | |
5216 | for (dl = super->disks ; dl ; dl = dl->next) { | |
f21e18ca N |
5217 | if (dl->major == (int)major(stb.st_rdev) && |
5218 | dl->minor == (int)minor(stb.st_rdev)) | |
c2c087e6 DW |
5219 | break; |
5220 | } | |
5221 | if (!dl) { | |
2c514b71 NB |
5222 | if (verbose) |
5223 | fprintf(stderr, Name ": %s is not in the " | |
5224 | "same imsm set\n", dev); | |
c2c087e6 | 5225 | return 0; |
a20d2ba5 DW |
5226 | } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) { |
5227 | /* If a volume is present then the current creation attempt | |
5228 | * cannot incorporate new spares because the orom may not | |
5229 | * understand this configuration (all member disks must be | |
5230 | * members of each array in the container). | |
5231 | */ | |
5232 | fprintf(stderr, Name ": %s is a spare and a volume" | |
5233 | " is already defined for this container\n", dev); | |
5234 | fprintf(stderr, Name ": The option-rom requires all member" | |
5235 | " disks to be a member of all volumes\n"); | |
5236 | return 0; | |
c2c087e6 | 5237 | } |
0dcecb2e DW |
5238 | |
5239 | /* retrieve the largest free space block */ | |
c2c087e6 DW |
5240 | e = get_extents(super, dl); |
5241 | maxsize = 0; | |
5242 | i = 0; | |
0dcecb2e DW |
5243 | if (e) { |
5244 | do { | |
5245 | unsigned long long esize; | |
5246 | ||
5247 | esize = e[i].start - pos; | |
5248 | if (esize >= maxsize) | |
5249 | maxsize = esize; | |
5250 | pos = e[i].start + e[i].size; | |
5251 | i++; | |
5252 | } while (e[i-1].size); | |
5253 | dl->e = e; | |
5254 | dl->extent_cnt = i; | |
5255 | } else { | |
5256 | if (verbose) | |
5257 | fprintf(stderr, Name ": unable to determine free space for: %s\n", | |
5258 | dev); | |
5259 | return 0; | |
5260 | } | |
5261 | if (maxsize < size) { | |
5262 | if (verbose) | |
5263 | fprintf(stderr, Name ": %s not enough space (%llu < %llu)\n", | |
5264 | dev, maxsize, size); | |
5265 | return 0; | |
5266 | } | |
5267 | ||
5268 | /* count total number of extents for merge */ | |
5269 | i = 0; | |
5270 | for (dl = super->disks; dl; dl = dl->next) | |
5271 | if (dl->e) | |
5272 | i += dl->extent_cnt; | |
5273 | ||
5274 | maxsize = merge_extents(super, i); | |
a7dd165b | 5275 | if (maxsize < size || maxsize == 0) { |
0dcecb2e DW |
5276 | if (verbose) |
5277 | fprintf(stderr, Name ": not enough space after merge (%llu < %llu)\n", | |
5278 | maxsize, size); | |
5279 | return 0; | |
0dcecb2e DW |
5280 | } |
5281 | ||
c2c087e6 DW |
5282 | *freesize = maxsize; |
5283 | ||
5284 | return 1; | |
cdddbdbc DW |
5285 | } |
5286 | ||
efb30e7f DW |
5287 | static int reserve_space(struct supertype *st, int raiddisks, |
5288 | unsigned long long size, int chunk, | |
5289 | unsigned long long *freesize) | |
5290 | { | |
5291 | struct intel_super *super = st->sb; | |
5292 | struct imsm_super *mpb = super->anchor; | |
5293 | struct dl *dl; | |
5294 | int i; | |
5295 | int extent_cnt; | |
5296 | struct extent *e; | |
5297 | unsigned long long maxsize; | |
5298 | unsigned long long minsize; | |
5299 | int cnt; | |
5300 | int used; | |
5301 | ||
5302 | /* find the largest common start free region of the possible disks */ | |
5303 | used = 0; | |
5304 | extent_cnt = 0; | |
5305 | cnt = 0; | |
5306 | for (dl = super->disks; dl; dl = dl->next) { | |
5307 | dl->raiddisk = -1; | |
5308 | ||
5309 | if (dl->index >= 0) | |
5310 | used++; | |
5311 | ||
5312 | /* don't activate new spares if we are orom constrained | |
5313 | * and there is already a volume active in the container | |
5314 | */ | |
5315 | if (super->orom && dl->index < 0 && mpb->num_raid_devs) | |
5316 | continue; | |
5317 | ||
5318 | e = get_extents(super, dl); | |
5319 | if (!e) | |
5320 | continue; | |
5321 | for (i = 1; e[i-1].size; i++) | |
5322 | ; | |
5323 | dl->e = e; | |
5324 | dl->extent_cnt = i; | |
5325 | extent_cnt += i; | |
5326 | cnt++; | |
5327 | } | |
5328 | ||
5329 | maxsize = merge_extents(super, extent_cnt); | |
5330 | minsize = size; | |
5331 | if (size == 0) | |
612e59d8 CA |
5332 | /* chunk is in K */ |
5333 | minsize = chunk * 2; | |
efb30e7f DW |
5334 | |
5335 | if (cnt < raiddisks || | |
5336 | (super->orom && used && used != raiddisks) || | |
a7dd165b DW |
5337 | maxsize < minsize || |
5338 | maxsize == 0) { | |
efb30e7f DW |
5339 | fprintf(stderr, Name ": not enough devices with space to create array.\n"); |
5340 | return 0; /* No enough free spaces large enough */ | |
5341 | } | |
5342 | ||
5343 | if (size == 0) { | |
5344 | size = maxsize; | |
5345 | if (chunk) { | |
612e59d8 CA |
5346 | size /= 2 * chunk; |
5347 | size *= 2 * chunk; | |
efb30e7f DW |
5348 | } |
5349 | } | |
5350 | ||
5351 | cnt = 0; | |
5352 | for (dl = super->disks; dl; dl = dl->next) | |
5353 | if (dl->e) | |
5354 | dl->raiddisk = cnt++; | |
5355 | ||
5356 | *freesize = size; | |
5357 | ||
5358 | return 1; | |
5359 | } | |
5360 | ||
bf5a934a | 5361 | static int validate_geometry_imsm(struct supertype *st, int level, int layout, |
c21e737b | 5362 | int raiddisks, int *chunk, unsigned long long size, |
bf5a934a DW |
5363 | char *dev, unsigned long long *freesize, |
5364 | int verbose) | |
5365 | { | |
5366 | int fd, cfd; | |
5367 | struct mdinfo *sra; | |
20cbe8d2 | 5368 | int is_member = 0; |
bf5a934a | 5369 | |
d54559f0 LM |
5370 | /* load capability |
5371 | * if given unused devices create a container | |
bf5a934a DW |
5372 | * if given given devices in a container create a member volume |
5373 | */ | |
5374 | if (level == LEVEL_CONTAINER) { | |
5375 | /* Must be a fresh device to add to a container */ | |
5376 | return validate_geometry_imsm_container(st, level, layout, | |
c21e737b CA |
5377 | raiddisks, |
5378 | chunk?*chunk:0, size, | |
bf5a934a DW |
5379 | dev, freesize, |
5380 | verbose); | |
5381 | } | |
5382 | ||
8592f29d N |
5383 | if (!dev) { |
5384 | if (st->sb && freesize) { | |
efb30e7f DW |
5385 | /* we are being asked to automatically layout a |
5386 | * new volume based on the current contents of | |
5387 | * the container. If the the parameters can be | |
5388 | * satisfied reserve_space will record the disks, | |
5389 | * start offset, and size of the volume to be | |
5390 | * created. add_to_super and getinfo_super | |
5391 | * detect when autolayout is in progress. | |
5392 | */ | |
6592ce37 DW |
5393 | if (!validate_geometry_imsm_orom(st->sb, level, layout, |
5394 | raiddisks, chunk, | |
5395 | verbose)) | |
5396 | return 0; | |
c21e737b CA |
5397 | return reserve_space(st, raiddisks, size, |
5398 | chunk?*chunk:0, freesize); | |
8592f29d N |
5399 | } |
5400 | return 1; | |
5401 | } | |
bf5a934a DW |
5402 | if (st->sb) { |
5403 | /* creating in a given container */ | |
5404 | return validate_geometry_imsm_volume(st, level, layout, | |
5405 | raiddisks, chunk, size, | |
5406 | dev, freesize, verbose); | |
5407 | } | |
5408 | ||
bf5a934a DW |
5409 | /* This device needs to be a device in an 'imsm' container */ |
5410 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
5411 | if (fd >= 0) { | |
5412 | if (verbose) | |
5413 | fprintf(stderr, | |
5414 | Name ": Cannot create this array on device %s\n", | |
5415 | dev); | |
5416 | close(fd); | |
5417 | return 0; | |
5418 | } | |
5419 | if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) { | |
5420 | if (verbose) | |
5421 | fprintf(stderr, Name ": Cannot open %s: %s\n", | |
5422 | dev, strerror(errno)); | |
5423 | return 0; | |
5424 | } | |
5425 | /* Well, it is in use by someone, maybe an 'imsm' container. */ | |
5426 | cfd = open_container(fd); | |
20cbe8d2 | 5427 | close(fd); |
bf5a934a | 5428 | if (cfd < 0) { |
bf5a934a DW |
5429 | if (verbose) |
5430 | fprintf(stderr, Name ": Cannot use %s: It is busy\n", | |
5431 | dev); | |
5432 | return 0; | |
5433 | } | |
5434 | sra = sysfs_read(cfd, 0, GET_VERSION); | |
bf5a934a | 5435 | if (sra && sra->array.major_version == -1 && |
20cbe8d2 AW |
5436 | strcmp(sra->text_version, "imsm") == 0) |
5437 | is_member = 1; | |
5438 | sysfs_free(sra); | |
5439 | if (is_member) { | |
bf5a934a DW |
5440 | /* This is a member of a imsm container. Load the container |
5441 | * and try to create a volume | |
5442 | */ | |
5443 | struct intel_super *super; | |
5444 | ||
e1902a7b | 5445 | if (load_super_imsm_all(st, cfd, (void **) &super, NULL) == 0) { |
bf5a934a DW |
5446 | st->sb = super; |
5447 | st->container_dev = fd2devnum(cfd); | |
5448 | close(cfd); | |
5449 | return validate_geometry_imsm_volume(st, level, layout, | |
5450 | raiddisks, chunk, | |
5451 | size, dev, | |
ecbd9e81 N |
5452 | freesize, 1) |
5453 | ? 1 : -1; | |
bf5a934a | 5454 | } |
20cbe8d2 | 5455 | } |
bf5a934a | 5456 | |
20cbe8d2 AW |
5457 | if (verbose) |
5458 | fprintf(stderr, Name ": failed container membership check\n"); | |
5459 | ||
5460 | close(cfd); | |
5461 | return 0; | |
bf5a934a | 5462 | } |
0bd16cf2 | 5463 | |
30f58b22 | 5464 | static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk) |
0bd16cf2 DJ |
5465 | { |
5466 | struct intel_super *super = st->sb; | |
5467 | ||
30f58b22 DW |
5468 | if (level && *level == UnSet) |
5469 | *level = LEVEL_CONTAINER; | |
5470 | ||
5471 | if (level && layout && *layout == UnSet) | |
5472 | *layout = imsm_level_to_layout(*level); | |
0bd16cf2 | 5473 | |
cd9d1ac7 DW |
5474 | if (chunk && (*chunk == UnSet || *chunk == 0)) |
5475 | *chunk = imsm_default_chunk(super->orom); | |
0bd16cf2 DJ |
5476 | } |
5477 | ||
33414a01 DW |
5478 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev); |
5479 | ||
5480 | static int kill_subarray_imsm(struct supertype *st) | |
5481 | { | |
5482 | /* remove the subarray currently referenced by ->current_vol */ | |
5483 | __u8 i; | |
5484 | struct intel_dev **dp; | |
5485 | struct intel_super *super = st->sb; | |
5486 | __u8 current_vol = super->current_vol; | |
5487 | struct imsm_super *mpb = super->anchor; | |
5488 | ||
5489 | if (super->current_vol < 0) | |
5490 | return 2; | |
5491 | super->current_vol = -1; /* invalidate subarray cursor */ | |
5492 | ||
5493 | /* block deletions that would change the uuid of active subarrays | |
5494 | * | |
5495 | * FIXME when immutable ids are available, but note that we'll | |
5496 | * also need to fixup the invalidated/active subarray indexes in | |
5497 | * mdstat | |
5498 | */ | |
5499 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
5500 | char subarray[4]; | |
5501 | ||
5502 | if (i < current_vol) | |
5503 | continue; | |
5504 | sprintf(subarray, "%u", i); | |
5505 | if (is_subarray_active(subarray, st->devname)) { | |
5506 | fprintf(stderr, | |
5507 | Name ": deleting subarray-%d would change the UUID of active subarray-%d, aborting\n", | |
5508 | current_vol, i); | |
5509 | ||
5510 | return 2; | |
5511 | } | |
5512 | } | |
5513 | ||
5514 | if (st->update_tail) { | |
5515 | struct imsm_update_kill_array *u = malloc(sizeof(*u)); | |
5516 | ||
5517 | if (!u) | |
5518 | return 2; | |
5519 | u->type = update_kill_array; | |
5520 | u->dev_idx = current_vol; | |
5521 | append_metadata_update(st, u, sizeof(*u)); | |
5522 | ||
5523 | return 0; | |
5524 | } | |
5525 | ||
5526 | for (dp = &super->devlist; *dp;) | |
5527 | if ((*dp)->index == current_vol) { | |
5528 | *dp = (*dp)->next; | |
5529 | } else { | |
5530 | handle_missing(super, (*dp)->dev); | |
5531 | if ((*dp)->index > current_vol) | |
5532 | (*dp)->index--; | |
5533 | dp = &(*dp)->next; | |
5534 | } | |
5535 | ||
5536 | /* no more raid devices, all active components are now spares, | |
5537 | * but of course failed are still failed | |
5538 | */ | |
5539 | if (--mpb->num_raid_devs == 0) { | |
5540 | struct dl *d; | |
5541 | ||
5542 | for (d = super->disks; d; d = d->next) | |
a8619d23 AK |
5543 | if (d->index > -2) |
5544 | mark_spare(d); | |
33414a01 DW |
5545 | } |
5546 | ||
5547 | super->updates_pending++; | |
5548 | ||
5549 | return 0; | |
5550 | } | |
aa534678 | 5551 | |
a951a4f7 | 5552 | static int update_subarray_imsm(struct supertype *st, char *subarray, |
fa56eddb | 5553 | char *update, struct mddev_ident *ident) |
aa534678 DW |
5554 | { |
5555 | /* update the subarray currently referenced by ->current_vol */ | |
5556 | struct intel_super *super = st->sb; | |
5557 | struct imsm_super *mpb = super->anchor; | |
5558 | ||
aa534678 DW |
5559 | if (strcmp(update, "name") == 0) { |
5560 | char *name = ident->name; | |
a951a4f7 N |
5561 | char *ep; |
5562 | int vol; | |
aa534678 | 5563 | |
a951a4f7 | 5564 | if (is_subarray_active(subarray, st->devname)) { |
aa534678 DW |
5565 | fprintf(stderr, |
5566 | Name ": Unable to update name of active subarray\n"); | |
5567 | return 2; | |
5568 | } | |
5569 | ||
5570 | if (!check_name(super, name, 0)) | |
5571 | return 2; | |
5572 | ||
a951a4f7 N |
5573 | vol = strtoul(subarray, &ep, 10); |
5574 | if (*ep != '\0' || vol >= super->anchor->num_raid_devs) | |
5575 | return 2; | |
5576 | ||
aa534678 DW |
5577 | if (st->update_tail) { |
5578 | struct imsm_update_rename_array *u = malloc(sizeof(*u)); | |
5579 | ||
5580 | if (!u) | |
5581 | return 2; | |
5582 | u->type = update_rename_array; | |
a951a4f7 | 5583 | u->dev_idx = vol; |
aa534678 DW |
5584 | snprintf((char *) u->name, MAX_RAID_SERIAL_LEN, "%s", name); |
5585 | append_metadata_update(st, u, sizeof(*u)); | |
5586 | } else { | |
5587 | struct imsm_dev *dev; | |
5588 | int i; | |
5589 | ||
a951a4f7 | 5590 | dev = get_imsm_dev(super, vol); |
aa534678 DW |
5591 | snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name); |
5592 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
5593 | dev = get_imsm_dev(super, i); | |
5594 | handle_missing(super, dev); | |
5595 | } | |
5596 | super->updates_pending++; | |
5597 | } | |
5598 | } else | |
5599 | return 2; | |
5600 | ||
5601 | return 0; | |
5602 | } | |
bf5a934a | 5603 | |
28bce06f AK |
5604 | static int is_gen_migration(struct imsm_dev *dev) |
5605 | { | |
7534230b AK |
5606 | if (dev == NULL) |
5607 | return 0; | |
5608 | ||
28bce06f AK |
5609 | if (!dev->vol.migr_state) |
5610 | return 0; | |
5611 | ||
5612 | if (migr_type(dev) == MIGR_GEN_MIGR) | |
5613 | return 1; | |
5614 | ||
5615 | return 0; | |
5616 | } | |
71204a50 | 5617 | #endif /* MDASSEMBLE */ |
28bce06f | 5618 | |
1e5c6983 DW |
5619 | static int is_rebuilding(struct imsm_dev *dev) |
5620 | { | |
5621 | struct imsm_map *migr_map; | |
5622 | ||
5623 | if (!dev->vol.migr_state) | |
5624 | return 0; | |
5625 | ||
5626 | if (migr_type(dev) != MIGR_REBUILD) | |
5627 | return 0; | |
5628 | ||
5629 | migr_map = get_imsm_map(dev, 1); | |
5630 | ||
5631 | if (migr_map->map_state == IMSM_T_STATE_DEGRADED) | |
5632 | return 1; | |
5633 | else | |
5634 | return 0; | |
5635 | } | |
5636 | ||
c47b0ff6 AK |
5637 | static void update_recovery_start(struct intel_super *super, |
5638 | struct imsm_dev *dev, | |
5639 | struct mdinfo *array) | |
1e5c6983 DW |
5640 | { |
5641 | struct mdinfo *rebuild = NULL; | |
5642 | struct mdinfo *d; | |
5643 | __u32 units; | |
5644 | ||
5645 | if (!is_rebuilding(dev)) | |
5646 | return; | |
5647 | ||
5648 | /* Find the rebuild target, but punt on the dual rebuild case */ | |
5649 | for (d = array->devs; d; d = d->next) | |
5650 | if (d->recovery_start == 0) { | |
5651 | if (rebuild) | |
5652 | return; | |
5653 | rebuild = d; | |
5654 | } | |
5655 | ||
4363fd80 DW |
5656 | if (!rebuild) { |
5657 | /* (?) none of the disks are marked with | |
5658 | * IMSM_ORD_REBUILD, so assume they are missing and the | |
5659 | * disk_ord_tbl was not correctly updated | |
5660 | */ | |
5661 | dprintf("%s: failed to locate out-of-sync disk\n", __func__); | |
5662 | return; | |
5663 | } | |
5664 | ||
1e5c6983 | 5665 | units = __le32_to_cpu(dev->vol.curr_migr_unit); |
c47b0ff6 | 5666 | rebuild->recovery_start = units * blocks_per_migr_unit(super, dev); |
1e5c6983 DW |
5667 | } |
5668 | ||
9e2d750d | 5669 | #ifndef MDASSEMBLE |
276d77db | 5670 | static int recover_backup_imsm(struct supertype *st, struct mdinfo *info); |
9e2d750d | 5671 | #endif |
1e5c6983 | 5672 | |
00bbdbda | 5673 | static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray) |
cdddbdbc | 5674 | { |
4f5bc454 DW |
5675 | /* Given a container loaded by load_super_imsm_all, |
5676 | * extract information about all the arrays into | |
5677 | * an mdinfo tree. | |
00bbdbda | 5678 | * If 'subarray' is given, just extract info about that array. |
4f5bc454 DW |
5679 | * |
5680 | * For each imsm_dev create an mdinfo, fill it in, | |
5681 | * then look for matching devices in super->disks | |
5682 | * and create appropriate device mdinfo. | |
5683 | */ | |
5684 | struct intel_super *super = st->sb; | |
949c47a0 | 5685 | struct imsm_super *mpb = super->anchor; |
4f5bc454 | 5686 | struct mdinfo *rest = NULL; |
00bbdbda | 5687 | unsigned int i; |
a06d022d | 5688 | int bbm_errors = 0; |
abef11a3 AK |
5689 | struct dl *d; |
5690 | int spare_disks = 0; | |
cdddbdbc | 5691 | |
19482bcc AK |
5692 | /* do not assemble arrays when not all attributes are supported */ |
5693 | if (imsm_check_attributes(mpb->attributes) == 0) { | |
5694 | fprintf(stderr, Name ": IMSM metadata loading not allowed " | |
5695 | "due to attributes incompatibility.\n"); | |
5696 | return NULL; | |
5697 | } | |
5698 | ||
a06d022d KW |
5699 | /* check for bad blocks */ |
5700 | if (imsm_bbm_log_size(super->anchor)) | |
5701 | bbm_errors = 1; | |
604b746f | 5702 | |
abef11a3 AK |
5703 | /* count spare devices, not used in maps |
5704 | */ | |
5705 | for (d = super->disks; d; d = d->next) | |
5706 | if (d->index == -1) | |
5707 | spare_disks++; | |
5708 | ||
4f5bc454 | 5709 | for (i = 0; i < mpb->num_raid_devs; i++) { |
00bbdbda N |
5710 | struct imsm_dev *dev; |
5711 | struct imsm_map *map; | |
86e3692b | 5712 | struct imsm_map *map2; |
4f5bc454 | 5713 | struct mdinfo *this; |
2db86302 | 5714 | int slot, chunk; |
00bbdbda N |
5715 | char *ep; |
5716 | ||
5717 | if (subarray && | |
5718 | (i != strtoul(subarray, &ep, 10) || *ep != '\0')) | |
5719 | continue; | |
5720 | ||
5721 | dev = get_imsm_dev(super, i); | |
5722 | map = get_imsm_map(dev, 0); | |
86e3692b | 5723 | map2 = get_imsm_map(dev, 1); |
4f5bc454 | 5724 | |
1ce0101c DW |
5725 | /* do not publish arrays that are in the middle of an |
5726 | * unsupported migration | |
5727 | */ | |
5728 | if (dev->vol.migr_state && | |
28bce06f | 5729 | (migr_type(dev) == MIGR_STATE_CHANGE)) { |
1ce0101c DW |
5730 | fprintf(stderr, Name ": cannot assemble volume '%.16s':" |
5731 | " unsupported migration in progress\n", | |
5732 | dev->volume); | |
5733 | continue; | |
5734 | } | |
2db86302 LM |
5735 | /* do not publish arrays that are not support by controller's |
5736 | * OROM/EFI | |
5737 | */ | |
1ce0101c | 5738 | |
2db86302 | 5739 | chunk = __le16_to_cpu(map->blocks_per_strip) >> 1; |
7b0bbd0f | 5740 | #ifndef MDASSEMBLE |
2db86302 LM |
5741 | if (!validate_geometry_imsm_orom(super, |
5742 | get_imsm_raid_level(map), /* RAID level */ | |
5743 | imsm_level_to_layout(get_imsm_raid_level(map)), | |
5744 | map->num_members, /* raid disks */ | |
5745 | &chunk, | |
5746 | 1 /* verbose */)) { | |
5747 | fprintf(stderr, Name ": RAID gemetry validation failed. " | |
5748 | "Cannot proceed with the action(s).\n"); | |
5749 | continue; | |
5750 | } | |
7b0bbd0f | 5751 | #endif /* MDASSEMBLE */ |
4f5bc454 | 5752 | this = malloc(sizeof(*this)); |
0fbd635c | 5753 | if (!this) { |
cf1be220 | 5754 | fprintf(stderr, Name ": failed to allocate %zu bytes\n", |
0fbd635c AW |
5755 | sizeof(*this)); |
5756 | break; | |
5757 | } | |
4f5bc454 | 5758 | |
301406c9 | 5759 | super->current_vol = i; |
a5d85af7 | 5760 | getinfo_super_imsm_volume(st, this, NULL); |
9894ec0d | 5761 | this->next = rest; |
4f5bc454 | 5762 | for (slot = 0 ; slot < map->num_members; slot++) { |
1e5c6983 | 5763 | unsigned long long recovery_start; |
4f5bc454 DW |
5764 | struct mdinfo *info_d; |
5765 | struct dl *d; | |
5766 | int idx; | |
9a1608e5 | 5767 | int skip; |
7eef0453 | 5768 | __u32 ord; |
4f5bc454 | 5769 | |
9a1608e5 | 5770 | skip = 0; |
98130f40 | 5771 | idx = get_imsm_disk_idx(dev, slot, 0); |
196b0d44 | 5772 | ord = get_imsm_ord_tbl_ent(dev, slot, -1); |
4f5bc454 DW |
5773 | for (d = super->disks; d ; d = d->next) |
5774 | if (d->index == idx) | |
0fbd635c | 5775 | break; |
4f5bc454 | 5776 | |
1e5c6983 | 5777 | recovery_start = MaxSector; |
4f5bc454 | 5778 | if (d == NULL) |
9a1608e5 | 5779 | skip = 1; |
25ed7e59 | 5780 | if (d && is_failed(&d->disk)) |
9a1608e5 | 5781 | skip = 1; |
7eef0453 | 5782 | if (ord & IMSM_ORD_REBUILD) |
1e5c6983 | 5783 | recovery_start = 0; |
9a1608e5 DW |
5784 | |
5785 | /* | |
5786 | * if we skip some disks the array will be assmebled degraded; | |
1e5c6983 DW |
5787 | * reset resync start to avoid a dirty-degraded |
5788 | * situation when performing the intial sync | |
9a1608e5 DW |
5789 | * |
5790 | * FIXME handle dirty degraded | |
5791 | */ | |
1e5c6983 | 5792 | if ((skip || recovery_start == 0) && !dev->vol.dirty) |
b7528a20 | 5793 | this->resync_start = MaxSector; |
9a1608e5 DW |
5794 | if (skip) |
5795 | continue; | |
4f5bc454 | 5796 | |
1e5c6983 | 5797 | info_d = calloc(1, sizeof(*info_d)); |
9a1608e5 DW |
5798 | if (!info_d) { |
5799 | fprintf(stderr, Name ": failed to allocate disk" | |
1ce0101c | 5800 | " for volume %.16s\n", dev->volume); |
1e5c6983 DW |
5801 | info_d = this->devs; |
5802 | while (info_d) { | |
5803 | struct mdinfo *d = info_d->next; | |
5804 | ||
5805 | free(info_d); | |
5806 | info_d = d; | |
5807 | } | |
9a1608e5 DW |
5808 | free(this); |
5809 | this = rest; | |
5810 | break; | |
5811 | } | |
4f5bc454 DW |
5812 | info_d->next = this->devs; |
5813 | this->devs = info_d; | |
5814 | ||
4f5bc454 DW |
5815 | info_d->disk.number = d->index; |
5816 | info_d->disk.major = d->major; | |
5817 | info_d->disk.minor = d->minor; | |
5818 | info_d->disk.raid_disk = slot; | |
1e5c6983 | 5819 | info_d->recovery_start = recovery_start; |
86e3692b AK |
5820 | if (map2) { |
5821 | if (slot < map2->num_members) | |
5822 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
5823 | else |
5824 | this->array.spare_disks++; | |
86e3692b AK |
5825 | } else { |
5826 | if (slot < map->num_members) | |
5827 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
5828 | else |
5829 | this->array.spare_disks++; | |
86e3692b | 5830 | } |
1e5c6983 DW |
5831 | if (info_d->recovery_start == MaxSector) |
5832 | this->array.working_disks++; | |
4f5bc454 DW |
5833 | |
5834 | info_d->events = __le32_to_cpu(mpb->generation_num); | |
5835 | info_d->data_offset = __le32_to_cpu(map->pba_of_lba0); | |
5836 | info_d->component_size = __le32_to_cpu(map->blocks_per_member); | |
4f5bc454 | 5837 | } |
1e5c6983 | 5838 | /* now that the disk list is up-to-date fixup recovery_start */ |
c47b0ff6 | 5839 | update_recovery_start(super, dev, this); |
abef11a3 | 5840 | this->array.spare_disks += spare_disks; |
276d77db | 5841 | |
9e2d750d | 5842 | #ifndef MDASSEMBLE |
276d77db AK |
5843 | /* check for reshape */ |
5844 | if (this->reshape_active == 1) | |
5845 | recover_backup_imsm(st, this); | |
9e2d750d | 5846 | #endif |
9a1608e5 | 5847 | rest = this; |
4f5bc454 DW |
5848 | } |
5849 | ||
a06d022d KW |
5850 | /* if array has bad blocks, set suitable bit in array status */ |
5851 | if (bbm_errors) | |
5852 | rest->array.state |= (1<<MD_SB_BBM_ERRORS); | |
5853 | ||
4f5bc454 | 5854 | return rest; |
cdddbdbc DW |
5855 | } |
5856 | ||
845dea95 | 5857 | |
fb49eef2 | 5858 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, int failed) |
c2a1e7da | 5859 | { |
a965f303 | 5860 | struct imsm_map *map = get_imsm_map(dev, 0); |
c2a1e7da DW |
5861 | |
5862 | if (!failed) | |
3393c6af DW |
5863 | return map->map_state == IMSM_T_STATE_UNINITIALIZED ? |
5864 | IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL; | |
c2a1e7da DW |
5865 | |
5866 | switch (get_imsm_raid_level(map)) { | |
5867 | case 0: | |
5868 | return IMSM_T_STATE_FAILED; | |
5869 | break; | |
5870 | case 1: | |
5871 | if (failed < map->num_members) | |
5872 | return IMSM_T_STATE_DEGRADED; | |
5873 | else | |
5874 | return IMSM_T_STATE_FAILED; | |
5875 | break; | |
5876 | case 10: | |
5877 | { | |
5878 | /** | |
c92a2527 DW |
5879 | * check to see if any mirrors have failed, otherwise we |
5880 | * are degraded. Even numbered slots are mirrored on | |
5881 | * slot+1 | |
c2a1e7da | 5882 | */ |
c2a1e7da | 5883 | int i; |
d9b420a5 N |
5884 | /* gcc -Os complains that this is unused */ |
5885 | int insync = insync; | |
c2a1e7da DW |
5886 | |
5887 | for (i = 0; i < map->num_members; i++) { | |
98130f40 | 5888 | __u32 ord = get_imsm_ord_tbl_ent(dev, i, -1); |
c92a2527 DW |
5889 | int idx = ord_to_idx(ord); |
5890 | struct imsm_disk *disk; | |
c2a1e7da | 5891 | |
c92a2527 DW |
5892 | /* reset the potential in-sync count on even-numbered |
5893 | * slots. num_copies is always 2 for imsm raid10 | |
5894 | */ | |
5895 | if ((i & 1) == 0) | |
5896 | insync = 2; | |
c2a1e7da | 5897 | |
c92a2527 | 5898 | disk = get_imsm_disk(super, idx); |
25ed7e59 | 5899 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) |
c92a2527 | 5900 | insync--; |
c2a1e7da | 5901 | |
c92a2527 DW |
5902 | /* no in-sync disks left in this mirror the |
5903 | * array has failed | |
5904 | */ | |
5905 | if (insync == 0) | |
5906 | return IMSM_T_STATE_FAILED; | |
c2a1e7da DW |
5907 | } |
5908 | ||
5909 | return IMSM_T_STATE_DEGRADED; | |
5910 | } | |
5911 | case 5: | |
5912 | if (failed < 2) | |
5913 | return IMSM_T_STATE_DEGRADED; | |
5914 | else | |
5915 | return IMSM_T_STATE_FAILED; | |
5916 | break; | |
5917 | default: | |
5918 | break; | |
5919 | } | |
5920 | ||
5921 | return map->map_state; | |
5922 | } | |
5923 | ||
ff077194 | 5924 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev) |
c2a1e7da DW |
5925 | { |
5926 | int i; | |
5927 | int failed = 0; | |
5928 | struct imsm_disk *disk; | |
ff077194 | 5929 | struct imsm_map *map = get_imsm_map(dev, 0); |
0556e1a2 DW |
5930 | struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state); |
5931 | __u32 ord; | |
5932 | int idx; | |
c2a1e7da | 5933 | |
0556e1a2 DW |
5934 | /* at the beginning of migration we set IMSM_ORD_REBUILD on |
5935 | * disks that are being rebuilt. New failures are recorded to | |
5936 | * map[0]. So we look through all the disks we started with and | |
5937 | * see if any failures are still present, or if any new ones | |
5938 | * have arrived | |
5939 | * | |
5940 | * FIXME add support for online capacity expansion and | |
5941 | * raid-level-migration | |
5942 | */ | |
5943 | for (i = 0; i < prev->num_members; i++) { | |
5944 | ord = __le32_to_cpu(prev->disk_ord_tbl[i]); | |
5945 | ord |= __le32_to_cpu(map->disk_ord_tbl[i]); | |
5946 | idx = ord_to_idx(ord); | |
c2a1e7da | 5947 | |
949c47a0 | 5948 | disk = get_imsm_disk(super, idx); |
25ed7e59 | 5949 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) |
fcb84475 | 5950 | failed++; |
c2a1e7da DW |
5951 | } |
5952 | ||
5953 | return failed; | |
845dea95 NB |
5954 | } |
5955 | ||
97b4d0e9 DW |
5956 | #ifndef MDASSEMBLE |
5957 | static int imsm_open_new(struct supertype *c, struct active_array *a, | |
5958 | char *inst) | |
5959 | { | |
5960 | struct intel_super *super = c->sb; | |
5961 | struct imsm_super *mpb = super->anchor; | |
5962 | ||
5963 | if (atoi(inst) >= mpb->num_raid_devs) { | |
5964 | fprintf(stderr, "%s: subarry index %d, out of range\n", | |
5965 | __func__, atoi(inst)); | |
5966 | return -ENODEV; | |
5967 | } | |
5968 | ||
5969 | dprintf("imsm: open_new %s\n", inst); | |
5970 | a->info.container_member = atoi(inst); | |
5971 | return 0; | |
5972 | } | |
5973 | ||
0c046afd DW |
5974 | static int is_resyncing(struct imsm_dev *dev) |
5975 | { | |
5976 | struct imsm_map *migr_map; | |
5977 | ||
5978 | if (!dev->vol.migr_state) | |
5979 | return 0; | |
5980 | ||
1484e727 DW |
5981 | if (migr_type(dev) == MIGR_INIT || |
5982 | migr_type(dev) == MIGR_REPAIR) | |
0c046afd DW |
5983 | return 1; |
5984 | ||
4c9bc37b AK |
5985 | if (migr_type(dev) == MIGR_GEN_MIGR) |
5986 | return 0; | |
5987 | ||
0c046afd DW |
5988 | migr_map = get_imsm_map(dev, 1); |
5989 | ||
4c9bc37b AK |
5990 | if ((migr_map->map_state == IMSM_T_STATE_NORMAL) && |
5991 | (dev->vol.migr_type != MIGR_GEN_MIGR)) | |
0c046afd DW |
5992 | return 1; |
5993 | else | |
5994 | return 0; | |
5995 | } | |
5996 | ||
0556e1a2 DW |
5997 | /* return true if we recorded new information */ |
5998 | static int mark_failure(struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
47ee5a45 | 5999 | { |
0556e1a2 DW |
6000 | __u32 ord; |
6001 | int slot; | |
6002 | struct imsm_map *map; | |
86c54047 DW |
6003 | char buf[MAX_RAID_SERIAL_LEN+3]; |
6004 | unsigned int len, shift = 0; | |
0556e1a2 DW |
6005 | |
6006 | /* new failures are always set in map[0] */ | |
6007 | map = get_imsm_map(dev, 0); | |
6008 | ||
6009 | slot = get_imsm_disk_slot(map, idx); | |
6010 | if (slot < 0) | |
6011 | return 0; | |
6012 | ||
6013 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
25ed7e59 | 6014 | if (is_failed(disk) && (ord & IMSM_ORD_REBUILD)) |
0556e1a2 DW |
6015 | return 0; |
6016 | ||
86c54047 DW |
6017 | sprintf(buf, "%s:0", disk->serial); |
6018 | if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN) | |
6019 | shift = len - MAX_RAID_SERIAL_LEN + 1; | |
6020 | strncpy((char *)disk->serial, &buf[shift], MAX_RAID_SERIAL_LEN); | |
6021 | ||
f2f27e63 | 6022 | disk->status |= FAILED_DISK; |
0556e1a2 | 6023 | set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD); |
f21e18ca | 6024 | if (map->failed_disk_num == 0xff) |
0556e1a2 DW |
6025 | map->failed_disk_num = slot; |
6026 | return 1; | |
6027 | } | |
6028 | ||
6029 | static void mark_missing(struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
6030 | { | |
6031 | mark_failure(dev, disk, idx); | |
6032 | ||
6033 | if (disk->scsi_id == __cpu_to_le32(~(__u32)0)) | |
6034 | return; | |
6035 | ||
47ee5a45 DW |
6036 | disk->scsi_id = __cpu_to_le32(~(__u32)0); |
6037 | memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1); | |
6038 | } | |
6039 | ||
33414a01 DW |
6040 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev) |
6041 | { | |
6042 | __u8 map_state; | |
6043 | struct dl *dl; | |
6044 | int failed; | |
6045 | ||
6046 | if (!super->missing) | |
6047 | return; | |
6048 | failed = imsm_count_failed(super, dev); | |
6049 | map_state = imsm_check_degraded(super, dev, failed); | |
6050 | ||
6051 | dprintf("imsm: mark missing\n"); | |
6052 | end_migration(dev, map_state); | |
6053 | for (dl = super->missing; dl; dl = dl->next) | |
6054 | mark_missing(dev, &dl->disk, dl->index); | |
6055 | super->updates_pending++; | |
6056 | } | |
6057 | ||
70bdf0dc AK |
6058 | static unsigned long long imsm_set_array_size(struct imsm_dev *dev) |
6059 | { | |
6060 | int used_disks = imsm_num_data_members(dev, 0); | |
6061 | unsigned long long array_blocks; | |
6062 | struct imsm_map *map; | |
6063 | ||
6064 | if (used_disks == 0) { | |
6065 | /* when problems occures | |
6066 | * return current array_blocks value | |
6067 | */ | |
6068 | array_blocks = __le32_to_cpu(dev->size_high); | |
6069 | array_blocks = array_blocks << 32; | |
6070 | array_blocks += __le32_to_cpu(dev->size_low); | |
6071 | ||
6072 | return array_blocks; | |
6073 | } | |
6074 | ||
6075 | /* set array size in metadata | |
6076 | */ | |
6077 | map = get_imsm_map(dev, 0); | |
6078 | array_blocks = map->blocks_per_member * used_disks; | |
6079 | ||
6080 | /* round array size down to closest MB | |
6081 | */ | |
6082 | array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; | |
6083 | dev->size_low = __cpu_to_le32((__u32)array_blocks); | |
6084 | dev->size_high = __cpu_to_le32((__u32)(array_blocks >> 32)); | |
6085 | ||
6086 | return array_blocks; | |
6087 | } | |
6088 | ||
28bce06f AK |
6089 | static void imsm_set_disk(struct active_array *a, int n, int state); |
6090 | ||
0e2d1a4e AK |
6091 | static void imsm_progress_container_reshape(struct intel_super *super) |
6092 | { | |
6093 | /* if no device has a migr_state, but some device has a | |
6094 | * different number of members than the previous device, start | |
6095 | * changing the number of devices in this device to match | |
6096 | * previous. | |
6097 | */ | |
6098 | struct imsm_super *mpb = super->anchor; | |
6099 | int prev_disks = -1; | |
6100 | int i; | |
1dfaa380 | 6101 | int copy_map_size; |
0e2d1a4e AK |
6102 | |
6103 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
6104 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
6105 | struct imsm_map *map = get_imsm_map(dev, 0); | |
6106 | struct imsm_map *map2; | |
6107 | int prev_num_members; | |
0e2d1a4e AK |
6108 | |
6109 | if (dev->vol.migr_state) | |
6110 | return; | |
6111 | ||
6112 | if (prev_disks == -1) | |
6113 | prev_disks = map->num_members; | |
6114 | if (prev_disks == map->num_members) | |
6115 | continue; | |
6116 | ||
6117 | /* OK, this array needs to enter reshape mode. | |
6118 | * i.e it needs a migr_state | |
6119 | */ | |
6120 | ||
1dfaa380 | 6121 | copy_map_size = sizeof_imsm_map(map); |
0e2d1a4e AK |
6122 | prev_num_members = map->num_members; |
6123 | map->num_members = prev_disks; | |
6124 | dev->vol.migr_state = 1; | |
6125 | dev->vol.curr_migr_unit = 0; | |
ea672ee1 | 6126 | set_migr_type(dev, MIGR_GEN_MIGR); |
0e2d1a4e AK |
6127 | for (i = prev_num_members; |
6128 | i < map->num_members; i++) | |
6129 | set_imsm_ord_tbl_ent(map, i, i); | |
6130 | map2 = get_imsm_map(dev, 1); | |
6131 | /* Copy the current map */ | |
1dfaa380 | 6132 | memcpy(map2, map, copy_map_size); |
0e2d1a4e AK |
6133 | map2->num_members = prev_num_members; |
6134 | ||
70bdf0dc | 6135 | imsm_set_array_size(dev); |
0e2d1a4e AK |
6136 | super->updates_pending++; |
6137 | } | |
6138 | } | |
6139 | ||
aad6f216 | 6140 | /* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild |
0c046afd DW |
6141 | * states are handled in imsm_set_disk() with one exception, when a |
6142 | * resync is stopped due to a new failure this routine will set the | |
6143 | * 'degraded' state for the array. | |
6144 | */ | |
01f157d7 | 6145 | static int imsm_set_array_state(struct active_array *a, int consistent) |
a862209d DW |
6146 | { |
6147 | int inst = a->info.container_member; | |
6148 | struct intel_super *super = a->container->sb; | |
949c47a0 | 6149 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
a965f303 | 6150 | struct imsm_map *map = get_imsm_map(dev, 0); |
0c046afd DW |
6151 | int failed = imsm_count_failed(super, dev); |
6152 | __u8 map_state = imsm_check_degraded(super, dev, failed); | |
1e5c6983 | 6153 | __u32 blocks_per_unit; |
a862209d | 6154 | |
1af97990 AK |
6155 | if (dev->vol.migr_state && |
6156 | dev->vol.migr_type == MIGR_GEN_MIGR) { | |
6157 | /* array state change is blocked due to reshape action | |
aad6f216 N |
6158 | * We might need to |
6159 | * - abort the reshape (if last_checkpoint is 0 and action!= reshape) | |
6160 | * - finish the reshape (if last_checkpoint is big and action != reshape) | |
6161 | * - update curr_migr_unit | |
1af97990 | 6162 | */ |
aad6f216 N |
6163 | if (a->curr_action == reshape) { |
6164 | /* still reshaping, maybe update curr_migr_unit */ | |
633b5610 | 6165 | goto mark_checkpoint; |
aad6f216 N |
6166 | } else { |
6167 | if (a->last_checkpoint == 0 && a->prev_action == reshape) { | |
6168 | /* for some reason we aborted the reshape. | |
b66e591b AK |
6169 | * |
6170 | * disable automatic metadata rollback | |
6171 | * user action is required to recover process | |
aad6f216 | 6172 | */ |
b66e591b | 6173 | if (0) { |
aad6f216 N |
6174 | struct imsm_map *map2 = get_imsm_map(dev, 1); |
6175 | dev->vol.migr_state = 0; | |
ea672ee1 | 6176 | set_migr_type(dev, 0); |
aad6f216 N |
6177 | dev->vol.curr_migr_unit = 0; |
6178 | memcpy(map, map2, sizeof_imsm_map(map2)); | |
6179 | super->updates_pending++; | |
b66e591b | 6180 | } |
aad6f216 N |
6181 | } |
6182 | if (a->last_checkpoint >= a->info.component_size) { | |
6183 | unsigned long long array_blocks; | |
6184 | int used_disks; | |
e154ced3 | 6185 | struct mdinfo *mdi; |
aad6f216 | 6186 | |
9653001d | 6187 | used_disks = imsm_num_data_members(dev, 0); |
d55adef9 AK |
6188 | if (used_disks > 0) { |
6189 | array_blocks = | |
6190 | map->blocks_per_member * | |
6191 | used_disks; | |
6192 | /* round array size down to closest MB | |
6193 | */ | |
6194 | array_blocks = (array_blocks | |
6195 | >> SECT_PER_MB_SHIFT) | |
6196 | << SECT_PER_MB_SHIFT; | |
d55adef9 AK |
6197 | a->info.custom_array_size = array_blocks; |
6198 | /* encourage manager to update array | |
6199 | * size | |
6200 | */ | |
e154ced3 | 6201 | |
d55adef9 | 6202 | a->check_reshape = 1; |
633b5610 | 6203 | } |
e154ced3 AK |
6204 | /* finalize online capacity expansion/reshape */ |
6205 | for (mdi = a->info.devs; mdi; mdi = mdi->next) | |
6206 | imsm_set_disk(a, | |
6207 | mdi->disk.raid_disk, | |
6208 | mdi->curr_state); | |
6209 | ||
0e2d1a4e | 6210 | imsm_progress_container_reshape(super); |
e154ced3 | 6211 | } |
aad6f216 | 6212 | } |
1af97990 AK |
6213 | } |
6214 | ||
47ee5a45 | 6215 | /* before we activate this array handle any missing disks */ |
33414a01 DW |
6216 | if (consistent == 2) |
6217 | handle_missing(super, dev); | |
1e5c6983 | 6218 | |
0c046afd | 6219 | if (consistent == 2 && |
b7941fd6 | 6220 | (!is_resync_complete(&a->info) || |
0c046afd DW |
6221 | map_state != IMSM_T_STATE_NORMAL || |
6222 | dev->vol.migr_state)) | |
01f157d7 | 6223 | consistent = 0; |
272906ef | 6224 | |
b7941fd6 | 6225 | if (is_resync_complete(&a->info)) { |
0c046afd | 6226 | /* complete intialization / resync, |
0556e1a2 DW |
6227 | * recovery and interrupted recovery is completed in |
6228 | * ->set_disk | |
0c046afd DW |
6229 | */ |
6230 | if (is_resyncing(dev)) { | |
6231 | dprintf("imsm: mark resync done\n"); | |
f8f603f1 | 6232 | end_migration(dev, map_state); |
115c3803 | 6233 | super->updates_pending++; |
484240d8 | 6234 | a->last_checkpoint = 0; |
115c3803 | 6235 | } |
b9172665 AK |
6236 | } else if ((!is_resyncing(dev) && !failed) && |
6237 | (imsm_reshape_blocks_arrays_changes(super) == 0)) { | |
0c046afd | 6238 | /* mark the start of the init process if nothing is failed */ |
b7941fd6 | 6239 | dprintf("imsm: mark resync start\n"); |
1484e727 | 6240 | if (map->map_state == IMSM_T_STATE_UNINITIALIZED) |
8e59f3d8 | 6241 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT); |
1484e727 | 6242 | else |
8e59f3d8 | 6243 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR); |
3393c6af | 6244 | super->updates_pending++; |
115c3803 | 6245 | } |
a862209d | 6246 | |
633b5610 | 6247 | mark_checkpoint: |
5b83bacf AK |
6248 | /* skip checkpointing for general migration, |
6249 | * it is controlled in mdadm | |
6250 | */ | |
6251 | if (is_gen_migration(dev)) | |
6252 | goto skip_mark_checkpoint; | |
6253 | ||
1e5c6983 | 6254 | /* check if we can update curr_migr_unit from resync_start, recovery_start */ |
c47b0ff6 | 6255 | blocks_per_unit = blocks_per_migr_unit(super, dev); |
4f0a7acc | 6256 | if (blocks_per_unit) { |
1e5c6983 DW |
6257 | __u32 units32; |
6258 | __u64 units; | |
6259 | ||
4f0a7acc | 6260 | units = a->last_checkpoint / blocks_per_unit; |
1e5c6983 DW |
6261 | units32 = units; |
6262 | ||
6263 | /* check that we did not overflow 32-bits, and that | |
6264 | * curr_migr_unit needs updating | |
6265 | */ | |
6266 | if (units32 == units && | |
bfd80a56 | 6267 | units32 != 0 && |
1e5c6983 DW |
6268 | __le32_to_cpu(dev->vol.curr_migr_unit) != units32) { |
6269 | dprintf("imsm: mark checkpoint (%u)\n", units32); | |
6270 | dev->vol.curr_migr_unit = __cpu_to_le32(units32); | |
6271 | super->updates_pending++; | |
6272 | } | |
6273 | } | |
f8f603f1 | 6274 | |
5b83bacf | 6275 | skip_mark_checkpoint: |
3393c6af | 6276 | /* mark dirty / clean */ |
0c046afd | 6277 | if (dev->vol.dirty != !consistent) { |
b7941fd6 | 6278 | dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty"); |
0c046afd DW |
6279 | if (consistent) |
6280 | dev->vol.dirty = 0; | |
6281 | else | |
6282 | dev->vol.dirty = 1; | |
a862209d DW |
6283 | super->updates_pending++; |
6284 | } | |
28bce06f | 6285 | |
01f157d7 | 6286 | return consistent; |
a862209d DW |
6287 | } |
6288 | ||
8d45d196 | 6289 | static void imsm_set_disk(struct active_array *a, int n, int state) |
845dea95 | 6290 | { |
8d45d196 DW |
6291 | int inst = a->info.container_member; |
6292 | struct intel_super *super = a->container->sb; | |
949c47a0 | 6293 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
a965f303 | 6294 | struct imsm_map *map = get_imsm_map(dev, 0); |
8d45d196 | 6295 | struct imsm_disk *disk; |
0c046afd | 6296 | int failed; |
b10b37b8 | 6297 | __u32 ord; |
0c046afd | 6298 | __u8 map_state; |
8d45d196 DW |
6299 | |
6300 | if (n > map->num_members) | |
6301 | fprintf(stderr, "imsm: set_disk %d out of range 0..%d\n", | |
6302 | n, map->num_members - 1); | |
6303 | ||
6304 | if (n < 0) | |
6305 | return; | |
6306 | ||
4e6e574a | 6307 | dprintf("imsm: set_disk %d:%x\n", n, state); |
8d45d196 | 6308 | |
98130f40 | 6309 | ord = get_imsm_ord_tbl_ent(dev, n, -1); |
b10b37b8 | 6310 | disk = get_imsm_disk(super, ord_to_idx(ord)); |
8d45d196 | 6311 | |
5802a811 | 6312 | /* check for new failures */ |
0556e1a2 DW |
6313 | if (state & DS_FAULTY) { |
6314 | if (mark_failure(dev, disk, ord_to_idx(ord))) | |
6315 | super->updates_pending++; | |
8d45d196 | 6316 | } |
47ee5a45 | 6317 | |
19859edc | 6318 | /* check if in_sync */ |
0556e1a2 | 6319 | if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) { |
b10b37b8 DW |
6320 | struct imsm_map *migr_map = get_imsm_map(dev, 1); |
6321 | ||
6322 | set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord)); | |
19859edc DW |
6323 | super->updates_pending++; |
6324 | } | |
8d45d196 | 6325 | |
0c046afd DW |
6326 | failed = imsm_count_failed(super, dev); |
6327 | map_state = imsm_check_degraded(super, dev, failed); | |
5802a811 | 6328 | |
0c046afd DW |
6329 | /* check if recovery complete, newly degraded, or failed */ |
6330 | if (map_state == IMSM_T_STATE_NORMAL && is_rebuilding(dev)) { | |
f8f603f1 | 6331 | end_migration(dev, map_state); |
0556e1a2 DW |
6332 | map = get_imsm_map(dev, 0); |
6333 | map->failed_disk_num = ~0; | |
0c046afd | 6334 | super->updates_pending++; |
484240d8 | 6335 | a->last_checkpoint = 0; |
0c046afd DW |
6336 | } else if (map_state == IMSM_T_STATE_DEGRADED && |
6337 | map->map_state != map_state && | |
6338 | !dev->vol.migr_state) { | |
6339 | dprintf("imsm: mark degraded\n"); | |
6340 | map->map_state = map_state; | |
6341 | super->updates_pending++; | |
484240d8 | 6342 | a->last_checkpoint = 0; |
0c046afd DW |
6343 | } else if (map_state == IMSM_T_STATE_FAILED && |
6344 | map->map_state != map_state) { | |
6345 | dprintf("imsm: mark failed\n"); | |
f8f603f1 | 6346 | end_migration(dev, map_state); |
0c046afd | 6347 | super->updates_pending++; |
484240d8 | 6348 | a->last_checkpoint = 0; |
28bce06f AK |
6349 | } else if (is_gen_migration(dev)) { |
6350 | dprintf("imsm: Detected General Migration in state: "); | |
6351 | if (map_state == IMSM_T_STATE_NORMAL) { | |
6352 | end_migration(dev, map_state); | |
6353 | map = get_imsm_map(dev, 0); | |
6354 | map->failed_disk_num = ~0; | |
6355 | dprintf("normal\n"); | |
6356 | } else { | |
6357 | if (map_state == IMSM_T_STATE_DEGRADED) { | |
6358 | printf("degraded\n"); | |
6359 | end_migration(dev, map_state); | |
6360 | } else { | |
6361 | dprintf("failed\n"); | |
6362 | } | |
6363 | map->map_state = map_state; | |
6364 | } | |
6365 | super->updates_pending++; | |
5802a811 | 6366 | } |
845dea95 NB |
6367 | } |
6368 | ||
f796af5d | 6369 | static int store_imsm_mpb(int fd, struct imsm_super *mpb) |
c2a1e7da | 6370 | { |
f796af5d | 6371 | void *buf = mpb; |
c2a1e7da DW |
6372 | __u32 mpb_size = __le32_to_cpu(mpb->mpb_size); |
6373 | unsigned long long dsize; | |
6374 | unsigned long long sectors; | |
6375 | ||
6376 | get_dev_size(fd, NULL, &dsize); | |
6377 | ||
272f648f DW |
6378 | if (mpb_size > 512) { |
6379 | /* -1 to account for anchor */ | |
6380 | sectors = mpb_sectors(mpb) - 1; | |
c2a1e7da | 6381 | |
272f648f DW |
6382 | /* write the extended mpb to the sectors preceeding the anchor */ |
6383 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) | |
6384 | return 1; | |
c2a1e7da | 6385 | |
f21e18ca N |
6386 | if ((unsigned long long)write(fd, buf + 512, 512 * sectors) |
6387 | != 512 * sectors) | |
272f648f DW |
6388 | return 1; |
6389 | } | |
c2a1e7da | 6390 | |
272f648f DW |
6391 | /* first block is stored on second to last sector of the disk */ |
6392 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) | |
c2a1e7da DW |
6393 | return 1; |
6394 | ||
f796af5d | 6395 | if (write(fd, buf, 512) != 512) |
c2a1e7da DW |
6396 | return 1; |
6397 | ||
c2a1e7da DW |
6398 | return 0; |
6399 | } | |
6400 | ||
2e735d19 | 6401 | static void imsm_sync_metadata(struct supertype *container) |
845dea95 | 6402 | { |
2e735d19 | 6403 | struct intel_super *super = container->sb; |
c2a1e7da | 6404 | |
1a64be56 | 6405 | dprintf("sync metadata: %d\n", super->updates_pending); |
c2a1e7da DW |
6406 | if (!super->updates_pending) |
6407 | return; | |
6408 | ||
36988a3d | 6409 | write_super_imsm(container, 0); |
c2a1e7da DW |
6410 | |
6411 | super->updates_pending = 0; | |
845dea95 NB |
6412 | } |
6413 | ||
272906ef DW |
6414 | static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a) |
6415 | { | |
6416 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
98130f40 | 6417 | int i = get_imsm_disk_idx(dev, idx, -1); |
272906ef DW |
6418 | struct dl *dl; |
6419 | ||
6420 | for (dl = super->disks; dl; dl = dl->next) | |
6421 | if (dl->index == i) | |
6422 | break; | |
6423 | ||
25ed7e59 | 6424 | if (dl && is_failed(&dl->disk)) |
272906ef DW |
6425 | dl = NULL; |
6426 | ||
6427 | if (dl) | |
6428 | dprintf("%s: found %x:%x\n", __func__, dl->major, dl->minor); | |
6429 | ||
6430 | return dl; | |
6431 | } | |
6432 | ||
a20d2ba5 | 6433 | static struct dl *imsm_add_spare(struct intel_super *super, int slot, |
8ba77d32 AK |
6434 | struct active_array *a, int activate_new, |
6435 | struct mdinfo *additional_test_list) | |
272906ef DW |
6436 | { |
6437 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
98130f40 | 6438 | int idx = get_imsm_disk_idx(dev, slot, -1); |
a20d2ba5 DW |
6439 | struct imsm_super *mpb = super->anchor; |
6440 | struct imsm_map *map; | |
272906ef DW |
6441 | unsigned long long pos; |
6442 | struct mdinfo *d; | |
6443 | struct extent *ex; | |
a20d2ba5 | 6444 | int i, j; |
272906ef | 6445 | int found; |
569cc43f DW |
6446 | __u32 array_start = 0; |
6447 | __u32 array_end = 0; | |
272906ef | 6448 | struct dl *dl; |
6c932028 | 6449 | struct mdinfo *test_list; |
272906ef DW |
6450 | |
6451 | for (dl = super->disks; dl; dl = dl->next) { | |
6452 | /* If in this array, skip */ | |
6453 | for (d = a->info.devs ; d ; d = d->next) | |
e553d2a4 DW |
6454 | if (d->state_fd >= 0 && |
6455 | d->disk.major == dl->major && | |
272906ef | 6456 | d->disk.minor == dl->minor) { |
8ba77d32 AK |
6457 | dprintf("%x:%x already in array\n", |
6458 | dl->major, dl->minor); | |
272906ef DW |
6459 | break; |
6460 | } | |
6461 | if (d) | |
6462 | continue; | |
6c932028 AK |
6463 | test_list = additional_test_list; |
6464 | while (test_list) { | |
6465 | if (test_list->disk.major == dl->major && | |
6466 | test_list->disk.minor == dl->minor) { | |
8ba77d32 AK |
6467 | dprintf("%x:%x already in additional test list\n", |
6468 | dl->major, dl->minor); | |
6469 | break; | |
6470 | } | |
6c932028 | 6471 | test_list = test_list->next; |
8ba77d32 | 6472 | } |
6c932028 | 6473 | if (test_list) |
8ba77d32 | 6474 | continue; |
272906ef | 6475 | |
e553d2a4 | 6476 | /* skip in use or failed drives */ |
25ed7e59 | 6477 | if (is_failed(&dl->disk) || idx == dl->index || |
df474657 DW |
6478 | dl->index == -2) { |
6479 | dprintf("%x:%x status (failed: %d index: %d)\n", | |
25ed7e59 | 6480 | dl->major, dl->minor, is_failed(&dl->disk), idx); |
9a1608e5 DW |
6481 | continue; |
6482 | } | |
6483 | ||
a20d2ba5 DW |
6484 | /* skip pure spares when we are looking for partially |
6485 | * assimilated drives | |
6486 | */ | |
6487 | if (dl->index == -1 && !activate_new) | |
6488 | continue; | |
6489 | ||
272906ef | 6490 | /* Does this unused device have the requisite free space? |
a20d2ba5 | 6491 | * It needs to be able to cover all member volumes |
272906ef DW |
6492 | */ |
6493 | ex = get_extents(super, dl); | |
6494 | if (!ex) { | |
6495 | dprintf("cannot get extents\n"); | |
6496 | continue; | |
6497 | } | |
a20d2ba5 DW |
6498 | for (i = 0; i < mpb->num_raid_devs; i++) { |
6499 | dev = get_imsm_dev(super, i); | |
6500 | map = get_imsm_map(dev, 0); | |
272906ef | 6501 | |
a20d2ba5 DW |
6502 | /* check if this disk is already a member of |
6503 | * this array | |
272906ef | 6504 | */ |
620b1713 | 6505 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
a20d2ba5 DW |
6506 | continue; |
6507 | ||
6508 | found = 0; | |
6509 | j = 0; | |
6510 | pos = 0; | |
6511 | array_start = __le32_to_cpu(map->pba_of_lba0); | |
329c8278 DW |
6512 | array_end = array_start + |
6513 | __le32_to_cpu(map->blocks_per_member) - 1; | |
a20d2ba5 DW |
6514 | |
6515 | do { | |
6516 | /* check that we can start at pba_of_lba0 with | |
6517 | * blocks_per_member of space | |
6518 | */ | |
329c8278 | 6519 | if (array_start >= pos && array_end < ex[j].start) { |
a20d2ba5 DW |
6520 | found = 1; |
6521 | break; | |
6522 | } | |
6523 | pos = ex[j].start + ex[j].size; | |
6524 | j++; | |
6525 | } while (ex[j-1].size); | |
6526 | ||
6527 | if (!found) | |
272906ef | 6528 | break; |
a20d2ba5 | 6529 | } |
272906ef DW |
6530 | |
6531 | free(ex); | |
a20d2ba5 | 6532 | if (i < mpb->num_raid_devs) { |
329c8278 DW |
6533 | dprintf("%x:%x does not have %u to %u available\n", |
6534 | dl->major, dl->minor, array_start, array_end); | |
272906ef DW |
6535 | /* No room */ |
6536 | continue; | |
a20d2ba5 DW |
6537 | } |
6538 | return dl; | |
272906ef DW |
6539 | } |
6540 | ||
6541 | return dl; | |
6542 | } | |
6543 | ||
95d07a2c LM |
6544 | |
6545 | static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed) | |
6546 | { | |
6547 | struct imsm_dev *dev2; | |
6548 | struct imsm_map *map; | |
6549 | struct dl *idisk; | |
6550 | int slot; | |
6551 | int idx; | |
6552 | __u8 state; | |
6553 | ||
6554 | dev2 = get_imsm_dev(cont->sb, dev_idx); | |
6555 | if (dev2) { | |
6556 | state = imsm_check_degraded(cont->sb, dev2, failed); | |
6557 | if (state == IMSM_T_STATE_FAILED) { | |
6558 | map = get_imsm_map(dev2, 0); | |
6559 | if (!map) | |
6560 | return 1; | |
6561 | for (slot = 0; slot < map->num_members; slot++) { | |
6562 | /* | |
6563 | * Check if failed disks are deleted from intel | |
6564 | * disk list or are marked to be deleted | |
6565 | */ | |
98130f40 | 6566 | idx = get_imsm_disk_idx(dev2, slot, -1); |
95d07a2c LM |
6567 | idisk = get_imsm_dl_disk(cont->sb, idx); |
6568 | /* | |
6569 | * Do not rebuild the array if failed disks | |
6570 | * from failed sub-array are not removed from | |
6571 | * container. | |
6572 | */ | |
6573 | if (idisk && | |
6574 | is_failed(&idisk->disk) && | |
6575 | (idisk->action != DISK_REMOVE)) | |
6576 | return 0; | |
6577 | } | |
6578 | } | |
6579 | } | |
6580 | return 1; | |
6581 | } | |
6582 | ||
88758e9d DW |
6583 | static struct mdinfo *imsm_activate_spare(struct active_array *a, |
6584 | struct metadata_update **updates) | |
6585 | { | |
6586 | /** | |
d23fe947 DW |
6587 | * Find a device with unused free space and use it to replace a |
6588 | * failed/vacant region in an array. We replace failed regions one a | |
6589 | * array at a time. The result is that a new spare disk will be added | |
6590 | * to the first failed array and after the monitor has finished | |
6591 | * propagating failures the remainder will be consumed. | |
88758e9d | 6592 | * |
d23fe947 DW |
6593 | * FIXME add a capability for mdmon to request spares from another |
6594 | * container. | |
88758e9d DW |
6595 | */ |
6596 | ||
6597 | struct intel_super *super = a->container->sb; | |
88758e9d | 6598 | int inst = a->info.container_member; |
949c47a0 | 6599 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
a965f303 | 6600 | struct imsm_map *map = get_imsm_map(dev, 0); |
88758e9d DW |
6601 | int failed = a->info.array.raid_disks; |
6602 | struct mdinfo *rv = NULL; | |
6603 | struct mdinfo *d; | |
6604 | struct mdinfo *di; | |
6605 | struct metadata_update *mu; | |
6606 | struct dl *dl; | |
6607 | struct imsm_update_activate_spare *u; | |
6608 | int num_spares = 0; | |
6609 | int i; | |
95d07a2c | 6610 | int allowed; |
88758e9d DW |
6611 | |
6612 | for (d = a->info.devs ; d ; d = d->next) { | |
6613 | if ((d->curr_state & DS_FAULTY) && | |
6614 | d->state_fd >= 0) | |
6615 | /* wait for Removal to happen */ | |
6616 | return NULL; | |
6617 | if (d->state_fd >= 0) | |
6618 | failed--; | |
6619 | } | |
6620 | ||
6621 | dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n", | |
6622 | inst, failed, a->info.array.raid_disks, a->info.array.level); | |
1af97990 | 6623 | |
e2962bfc AK |
6624 | if (imsm_reshape_blocks_arrays_changes(super)) |
6625 | return NULL; | |
1af97990 | 6626 | |
89c67882 AK |
6627 | if (a->info.array.level == 4) |
6628 | /* No repair for takeovered array | |
6629 | * imsm doesn't support raid4 | |
6630 | */ | |
6631 | return NULL; | |
6632 | ||
fb49eef2 | 6633 | if (imsm_check_degraded(super, dev, failed) != IMSM_T_STATE_DEGRADED) |
88758e9d DW |
6634 | return NULL; |
6635 | ||
95d07a2c LM |
6636 | /* |
6637 | * If there are any failed disks check state of the other volume. | |
6638 | * Block rebuild if the another one is failed until failed disks | |
6639 | * are removed from container. | |
6640 | */ | |
6641 | if (failed) { | |
6642 | dprintf("found failed disks in %s, check if there another" | |
6643 | "failed sub-array.\n", | |
6644 | dev->volume); | |
6645 | /* check if states of the other volumes allow for rebuild */ | |
6646 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
6647 | if (i != inst) { | |
6648 | allowed = imsm_rebuild_allowed(a->container, | |
6649 | i, failed); | |
6650 | if (!allowed) | |
6651 | return NULL; | |
6652 | } | |
6653 | } | |
6654 | } | |
6655 | ||
88758e9d | 6656 | /* For each slot, if it is not working, find a spare */ |
88758e9d DW |
6657 | for (i = 0; i < a->info.array.raid_disks; i++) { |
6658 | for (d = a->info.devs ; d ; d = d->next) | |
6659 | if (d->disk.raid_disk == i) | |
6660 | break; | |
6661 | dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0); | |
6662 | if (d && (d->state_fd >= 0)) | |
6663 | continue; | |
6664 | ||
272906ef | 6665 | /* |
a20d2ba5 DW |
6666 | * OK, this device needs recovery. Try to re-add the |
6667 | * previous occupant of this slot, if this fails see if | |
6668 | * we can continue the assimilation of a spare that was | |
6669 | * partially assimilated, finally try to activate a new | |
6670 | * spare. | |
272906ef DW |
6671 | */ |
6672 | dl = imsm_readd(super, i, a); | |
6673 | if (!dl) | |
b303fe21 | 6674 | dl = imsm_add_spare(super, i, a, 0, rv); |
a20d2ba5 | 6675 | if (!dl) |
b303fe21 | 6676 | dl = imsm_add_spare(super, i, a, 1, rv); |
272906ef DW |
6677 | if (!dl) |
6678 | continue; | |
6679 | ||
6680 | /* found a usable disk with enough space */ | |
6681 | di = malloc(sizeof(*di)); | |
79244939 DW |
6682 | if (!di) |
6683 | continue; | |
272906ef DW |
6684 | memset(di, 0, sizeof(*di)); |
6685 | ||
6686 | /* dl->index will be -1 in the case we are activating a | |
6687 | * pristine spare. imsm_process_update() will create a | |
6688 | * new index in this case. Once a disk is found to be | |
6689 | * failed in all member arrays it is kicked from the | |
6690 | * metadata | |
6691 | */ | |
6692 | di->disk.number = dl->index; | |
d23fe947 | 6693 | |
272906ef DW |
6694 | /* (ab)use di->devs to store a pointer to the device |
6695 | * we chose | |
6696 | */ | |
6697 | di->devs = (struct mdinfo *) dl; | |
6698 | ||
6699 | di->disk.raid_disk = i; | |
6700 | di->disk.major = dl->major; | |
6701 | di->disk.minor = dl->minor; | |
6702 | di->disk.state = 0; | |
d23534e4 | 6703 | di->recovery_start = 0; |
272906ef DW |
6704 | di->data_offset = __le32_to_cpu(map->pba_of_lba0); |
6705 | di->component_size = a->info.component_size; | |
6706 | di->container_member = inst; | |
148acb7b | 6707 | super->random = random32(); |
272906ef DW |
6708 | di->next = rv; |
6709 | rv = di; | |
6710 | num_spares++; | |
6711 | dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor, | |
6712 | i, di->data_offset); | |
88758e9d DW |
6713 | } |
6714 | ||
6715 | if (!rv) | |
6716 | /* No spares found */ | |
6717 | return rv; | |
6718 | /* Now 'rv' has a list of devices to return. | |
6719 | * Create a metadata_update record to update the | |
6720 | * disk_ord_tbl for the array | |
6721 | */ | |
6722 | mu = malloc(sizeof(*mu)); | |
79244939 DW |
6723 | if (mu) { |
6724 | mu->buf = malloc(sizeof(struct imsm_update_activate_spare) * num_spares); | |
6725 | if (mu->buf == NULL) { | |
6726 | free(mu); | |
6727 | mu = NULL; | |
6728 | } | |
6729 | } | |
6730 | if (!mu) { | |
6731 | while (rv) { | |
6732 | struct mdinfo *n = rv->next; | |
6733 | ||
6734 | free(rv); | |
6735 | rv = n; | |
6736 | } | |
6737 | return NULL; | |
6738 | } | |
6739 | ||
88758e9d | 6740 | mu->space = NULL; |
cb23f1f4 | 6741 | mu->space_list = NULL; |
88758e9d DW |
6742 | mu->len = sizeof(struct imsm_update_activate_spare) * num_spares; |
6743 | mu->next = *updates; | |
6744 | u = (struct imsm_update_activate_spare *) mu->buf; | |
6745 | ||
6746 | for (di = rv ; di ; di = di->next) { | |
6747 | u->type = update_activate_spare; | |
d23fe947 DW |
6748 | u->dl = (struct dl *) di->devs; |
6749 | di->devs = NULL; | |
88758e9d DW |
6750 | u->slot = di->disk.raid_disk; |
6751 | u->array = inst; | |
6752 | u->next = u + 1; | |
6753 | u++; | |
6754 | } | |
6755 | (u-1)->next = NULL; | |
6756 | *updates = mu; | |
6757 | ||
6758 | return rv; | |
6759 | } | |
6760 | ||
54c2c1ea | 6761 | static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u) |
8273f55e | 6762 | { |
54c2c1ea DW |
6763 | struct imsm_dev *dev = get_imsm_dev(super, idx); |
6764 | struct imsm_map *map = get_imsm_map(dev, 0); | |
6765 | struct imsm_map *new_map = get_imsm_map(&u->dev, 0); | |
6766 | struct disk_info *inf = get_disk_info(u); | |
6767 | struct imsm_disk *disk; | |
8273f55e DW |
6768 | int i; |
6769 | int j; | |
8273f55e | 6770 | |
54c2c1ea | 6771 | for (i = 0; i < map->num_members; i++) { |
98130f40 | 6772 | disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, -1)); |
54c2c1ea DW |
6773 | for (j = 0; j < new_map->num_members; j++) |
6774 | if (serialcmp(disk->serial, inf[j].serial) == 0) | |
8273f55e DW |
6775 | return 1; |
6776 | } | |
6777 | ||
6778 | return 0; | |
6779 | } | |
6780 | ||
1a64be56 LM |
6781 | |
6782 | static struct dl *get_disk_super(struct intel_super *super, int major, int minor) | |
6783 | { | |
6784 | struct dl *dl = NULL; | |
6785 | for (dl = super->disks; dl; dl = dl->next) | |
6786 | if ((dl->major == major) && (dl->minor == minor)) | |
6787 | return dl; | |
6788 | return NULL; | |
6789 | } | |
6790 | ||
6791 | static int remove_disk_super(struct intel_super *super, int major, int minor) | |
6792 | { | |
6793 | struct dl *prev = NULL; | |
6794 | struct dl *dl; | |
6795 | ||
6796 | prev = NULL; | |
6797 | for (dl = super->disks; dl; dl = dl->next) { | |
6798 | if ((dl->major == major) && (dl->minor == minor)) { | |
6799 | /* remove */ | |
6800 | if (prev) | |
6801 | prev->next = dl->next; | |
6802 | else | |
6803 | super->disks = dl->next; | |
6804 | dl->next = NULL; | |
6805 | __free_imsm_disk(dl); | |
6806 | dprintf("%s: removed %x:%x\n", | |
6807 | __func__, major, minor); | |
6808 | break; | |
6809 | } | |
6810 | prev = dl; | |
6811 | } | |
6812 | return 0; | |
6813 | } | |
6814 | ||
f21e18ca | 6815 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index); |
ae6aad82 | 6816 | |
1a64be56 LM |
6817 | static int add_remove_disk_update(struct intel_super *super) |
6818 | { | |
6819 | int check_degraded = 0; | |
6820 | struct dl *disk = NULL; | |
6821 | /* add/remove some spares to/from the metadata/contrainer */ | |
6822 | while (super->disk_mgmt_list) { | |
6823 | struct dl *disk_cfg; | |
6824 | ||
6825 | disk_cfg = super->disk_mgmt_list; | |
6826 | super->disk_mgmt_list = disk_cfg->next; | |
6827 | disk_cfg->next = NULL; | |
6828 | ||
6829 | if (disk_cfg->action == DISK_ADD) { | |
6830 | disk_cfg->next = super->disks; | |
6831 | super->disks = disk_cfg; | |
6832 | check_degraded = 1; | |
6833 | dprintf("%s: added %x:%x\n", | |
6834 | __func__, disk_cfg->major, | |
6835 | disk_cfg->minor); | |
6836 | } else if (disk_cfg->action == DISK_REMOVE) { | |
6837 | dprintf("Disk remove action processed: %x.%x\n", | |
6838 | disk_cfg->major, disk_cfg->minor); | |
6839 | disk = get_disk_super(super, | |
6840 | disk_cfg->major, | |
6841 | disk_cfg->minor); | |
6842 | if (disk) { | |
6843 | /* store action status */ | |
6844 | disk->action = DISK_REMOVE; | |
6845 | /* remove spare disks only */ | |
6846 | if (disk->index == -1) { | |
6847 | remove_disk_super(super, | |
6848 | disk_cfg->major, | |
6849 | disk_cfg->minor); | |
6850 | } | |
6851 | } | |
6852 | /* release allocate disk structure */ | |
6853 | __free_imsm_disk(disk_cfg); | |
6854 | } | |
6855 | } | |
6856 | return check_degraded; | |
6857 | } | |
6858 | ||
a29911da PC |
6859 | |
6860 | static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u, | |
6861 | struct intel_super *super, | |
6862 | void ***space_list) | |
6863 | { | |
6864 | struct intel_dev *id; | |
6865 | void **tofree = NULL; | |
6866 | int ret_val = 0; | |
6867 | ||
6868 | dprintf("apply_reshape_migration_update()\n"); | |
6869 | if ((u->subdev < 0) || | |
6870 | (u->subdev > 1)) { | |
6871 | dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev); | |
6872 | return ret_val; | |
6873 | } | |
6874 | if ((space_list == NULL) || (*space_list == NULL)) { | |
6875 | dprintf("imsm: Error: Memory is not allocated\n"); | |
6876 | return ret_val; | |
6877 | } | |
6878 | ||
6879 | for (id = super->devlist ; id; id = id->next) { | |
6880 | if (id->index == (unsigned)u->subdev) { | |
6881 | struct imsm_dev *dev = get_imsm_dev(super, u->subdev); | |
6882 | struct imsm_map *map; | |
6883 | struct imsm_dev *new_dev = | |
6884 | (struct imsm_dev *)*space_list; | |
6885 | struct imsm_map *migr_map = get_imsm_map(dev, 1); | |
6886 | int to_state; | |
6887 | struct dl *new_disk; | |
6888 | ||
6889 | if (new_dev == NULL) | |
6890 | return ret_val; | |
6891 | *space_list = **space_list; | |
6892 | memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0)); | |
6893 | map = get_imsm_map(new_dev, 0); | |
6894 | if (migr_map) { | |
6895 | dprintf("imsm: Error: migration in progress"); | |
6896 | return ret_val; | |
6897 | } | |
6898 | ||
6899 | to_state = map->map_state; | |
6900 | if ((u->new_level == 5) && (map->raid_level == 0)) { | |
6901 | map->num_members++; | |
6902 | /* this should not happen */ | |
6903 | if (u->new_disks[0] < 0) { | |
6904 | map->failed_disk_num = | |
6905 | map->num_members - 1; | |
6906 | to_state = IMSM_T_STATE_DEGRADED; | |
6907 | } else | |
6908 | to_state = IMSM_T_STATE_NORMAL; | |
6909 | } | |
8e59f3d8 | 6910 | migrate(new_dev, super, to_state, MIGR_GEN_MIGR); |
a29911da PC |
6911 | if (u->new_level > -1) |
6912 | map->raid_level = u->new_level; | |
6913 | migr_map = get_imsm_map(new_dev, 1); | |
6914 | if ((u->new_level == 5) && | |
6915 | (migr_map->raid_level == 0)) { | |
6916 | int ord = map->num_members - 1; | |
6917 | migr_map->num_members--; | |
6918 | if (u->new_disks[0] < 0) | |
6919 | ord |= IMSM_ORD_REBUILD; | |
6920 | set_imsm_ord_tbl_ent(map, | |
6921 | map->num_members - 1, | |
6922 | ord); | |
6923 | } | |
6924 | id->dev = new_dev; | |
6925 | tofree = (void **)dev; | |
6926 | ||
4bba0439 PC |
6927 | /* update chunk size |
6928 | */ | |
6929 | if (u->new_chunksize > 0) | |
6930 | map->blocks_per_strip = | |
6931 | __cpu_to_le16(u->new_chunksize * 2); | |
6932 | ||
a29911da PC |
6933 | /* add disk |
6934 | */ | |
6935 | if ((u->new_level != 5) || | |
6936 | (migr_map->raid_level != 0) || | |
6937 | (migr_map->raid_level == map->raid_level)) | |
6938 | goto skip_disk_add; | |
6939 | ||
6940 | if (u->new_disks[0] >= 0) { | |
6941 | /* use passes spare | |
6942 | */ | |
6943 | new_disk = get_disk_super(super, | |
6944 | major(u->new_disks[0]), | |
6945 | minor(u->new_disks[0])); | |
6946 | dprintf("imsm: new disk for reshape is: %i:%i " | |
6947 | "(%p, index = %i)\n", | |
6948 | major(u->new_disks[0]), | |
6949 | minor(u->new_disks[0]), | |
6950 | new_disk, new_disk->index); | |
6951 | if (new_disk == NULL) | |
6952 | goto error_disk_add; | |
6953 | ||
6954 | new_disk->index = map->num_members - 1; | |
6955 | /* slot to fill in autolayout | |
6956 | */ | |
6957 | new_disk->raiddisk = new_disk->index; | |
6958 | new_disk->disk.status |= CONFIGURED_DISK; | |
6959 | new_disk->disk.status &= ~SPARE_DISK; | |
6960 | } else | |
6961 | goto error_disk_add; | |
6962 | ||
6963 | skip_disk_add: | |
6964 | *tofree = *space_list; | |
6965 | /* calculate new size | |
6966 | */ | |
6967 | imsm_set_array_size(new_dev); | |
6968 | ||
6969 | ret_val = 1; | |
6970 | } | |
6971 | } | |
6972 | ||
6973 | if (tofree) | |
6974 | *space_list = tofree; | |
6975 | return ret_val; | |
6976 | ||
6977 | error_disk_add: | |
6978 | dprintf("Error: imsm: Cannot find disk.\n"); | |
6979 | return ret_val; | |
6980 | } | |
6981 | ||
6982 | ||
2e5dc010 N |
6983 | static int apply_reshape_container_disks_update(struct imsm_update_reshape *u, |
6984 | struct intel_super *super, | |
6985 | void ***space_list) | |
6986 | { | |
6987 | struct dl *new_disk; | |
6988 | struct intel_dev *id; | |
6989 | int i; | |
6990 | int delta_disks = u->new_raid_disks - u->old_raid_disks; | |
ee4beede | 6991 | int disk_count = u->old_raid_disks; |
2e5dc010 N |
6992 | void **tofree = NULL; |
6993 | int devices_to_reshape = 1; | |
6994 | struct imsm_super *mpb = super->anchor; | |
6995 | int ret_val = 0; | |
d098291a | 6996 | unsigned int dev_id; |
2e5dc010 | 6997 | |
ed7333bd | 6998 | dprintf("imsm: apply_reshape_container_disks_update()\n"); |
2e5dc010 N |
6999 | |
7000 | /* enable spares to use in array */ | |
7001 | for (i = 0; i < delta_disks; i++) { | |
7002 | new_disk = get_disk_super(super, | |
7003 | major(u->new_disks[i]), | |
7004 | minor(u->new_disks[i])); | |
ed7333bd AK |
7005 | dprintf("imsm: new disk for reshape is: %i:%i " |
7006 | "(%p, index = %i)\n", | |
2e5dc010 N |
7007 | major(u->new_disks[i]), minor(u->new_disks[i]), |
7008 | new_disk, new_disk->index); | |
7009 | if ((new_disk == NULL) || | |
7010 | ((new_disk->index >= 0) && | |
7011 | (new_disk->index < u->old_raid_disks))) | |
7012 | goto update_reshape_exit; | |
ee4beede | 7013 | new_disk->index = disk_count++; |
2e5dc010 N |
7014 | /* slot to fill in autolayout |
7015 | */ | |
7016 | new_disk->raiddisk = new_disk->index; | |
7017 | new_disk->disk.status |= | |
7018 | CONFIGURED_DISK; | |
7019 | new_disk->disk.status &= ~SPARE_DISK; | |
7020 | } | |
7021 | ||
ed7333bd AK |
7022 | dprintf("imsm: volume set mpb->num_raid_devs = %i\n", |
7023 | mpb->num_raid_devs); | |
2e5dc010 N |
7024 | /* manage changes in volume |
7025 | */ | |
d098291a | 7026 | for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) { |
2e5dc010 N |
7027 | void **sp = *space_list; |
7028 | struct imsm_dev *newdev; | |
7029 | struct imsm_map *newmap, *oldmap; | |
7030 | ||
d098291a AK |
7031 | for (id = super->devlist ; id; id = id->next) { |
7032 | if (id->index == dev_id) | |
7033 | break; | |
7034 | } | |
7035 | if (id == NULL) | |
7036 | break; | |
2e5dc010 N |
7037 | if (!sp) |
7038 | continue; | |
7039 | *space_list = *sp; | |
7040 | newdev = (void*)sp; | |
7041 | /* Copy the dev, but not (all of) the map */ | |
7042 | memcpy(newdev, id->dev, sizeof(*newdev)); | |
7043 | oldmap = get_imsm_map(id->dev, 0); | |
7044 | newmap = get_imsm_map(newdev, 0); | |
7045 | /* Copy the current map */ | |
7046 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); | |
7047 | /* update one device only | |
7048 | */ | |
7049 | if (devices_to_reshape) { | |
ed7333bd AK |
7050 | dprintf("imsm: modifying subdev: %i\n", |
7051 | id->index); | |
2e5dc010 N |
7052 | devices_to_reshape--; |
7053 | newdev->vol.migr_state = 1; | |
7054 | newdev->vol.curr_migr_unit = 0; | |
ea672ee1 | 7055 | set_migr_type(newdev, MIGR_GEN_MIGR); |
2e5dc010 N |
7056 | newmap->num_members = u->new_raid_disks; |
7057 | for (i = 0; i < delta_disks; i++) { | |
7058 | set_imsm_ord_tbl_ent(newmap, | |
7059 | u->old_raid_disks + i, | |
7060 | u->old_raid_disks + i); | |
7061 | } | |
7062 | /* New map is correct, now need to save old map | |
7063 | */ | |
7064 | newmap = get_imsm_map(newdev, 1); | |
7065 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); | |
7066 | ||
70bdf0dc | 7067 | imsm_set_array_size(newdev); |
2e5dc010 N |
7068 | } |
7069 | ||
7070 | sp = (void **)id->dev; | |
7071 | id->dev = newdev; | |
7072 | *sp = tofree; | |
7073 | tofree = sp; | |
8e59f3d8 AK |
7074 | |
7075 | /* Clear migration record */ | |
7076 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
2e5dc010 | 7077 | } |
819bc634 AK |
7078 | if (tofree) |
7079 | *space_list = tofree; | |
2e5dc010 N |
7080 | ret_val = 1; |
7081 | ||
7082 | update_reshape_exit: | |
7083 | ||
7084 | return ret_val; | |
7085 | } | |
7086 | ||
bb025c2f | 7087 | static int apply_takeover_update(struct imsm_update_takeover *u, |
8ca6df95 KW |
7088 | struct intel_super *super, |
7089 | void ***space_list) | |
bb025c2f KW |
7090 | { |
7091 | struct imsm_dev *dev = NULL; | |
8ca6df95 KW |
7092 | struct intel_dev *dv; |
7093 | struct imsm_dev *dev_new; | |
bb025c2f KW |
7094 | struct imsm_map *map; |
7095 | struct dl *dm, *du; | |
8ca6df95 | 7096 | int i; |
bb025c2f KW |
7097 | |
7098 | for (dv = super->devlist; dv; dv = dv->next) | |
7099 | if (dv->index == (unsigned int)u->subarray) { | |
7100 | dev = dv->dev; | |
7101 | break; | |
7102 | } | |
7103 | ||
7104 | if (dev == NULL) | |
7105 | return 0; | |
7106 | ||
7107 | map = get_imsm_map(dev, 0); | |
7108 | ||
7109 | if (u->direction == R10_TO_R0) { | |
43d5ec18 KW |
7110 | /* Number of failed disks must be half of initial disk number */ |
7111 | if (imsm_count_failed(super, dev) != (map->num_members / 2)) | |
7112 | return 0; | |
7113 | ||
bb025c2f KW |
7114 | /* iterate through devices to mark removed disks as spare */ |
7115 | for (dm = super->disks; dm; dm = dm->next) { | |
7116 | if (dm->disk.status & FAILED_DISK) { | |
7117 | int idx = dm->index; | |
7118 | /* update indexes on the disk list */ | |
7119 | /* FIXME this loop-with-the-loop looks wrong, I'm not convinced | |
7120 | the index values will end up being correct.... NB */ | |
7121 | for (du = super->disks; du; du = du->next) | |
7122 | if (du->index > idx) | |
7123 | du->index--; | |
7124 | /* mark as spare disk */ | |
a8619d23 | 7125 | mark_spare(dm); |
bb025c2f KW |
7126 | } |
7127 | } | |
bb025c2f KW |
7128 | /* update map */ |
7129 | map->num_members = map->num_members / 2; | |
7130 | map->map_state = IMSM_T_STATE_NORMAL; | |
7131 | map->num_domains = 1; | |
7132 | map->raid_level = 0; | |
7133 | map->failed_disk_num = -1; | |
7134 | } | |
7135 | ||
8ca6df95 KW |
7136 | if (u->direction == R0_TO_R10) { |
7137 | void **space; | |
7138 | /* update slots in current disk list */ | |
7139 | for (dm = super->disks; dm; dm = dm->next) { | |
7140 | if (dm->index >= 0) | |
7141 | dm->index *= 2; | |
7142 | } | |
7143 | /* create new *missing* disks */ | |
7144 | for (i = 0; i < map->num_members; i++) { | |
7145 | space = *space_list; | |
7146 | if (!space) | |
7147 | continue; | |
7148 | *space_list = *space; | |
7149 | du = (void *)space; | |
7150 | memcpy(du, super->disks, sizeof(*du)); | |
8ca6df95 KW |
7151 | du->fd = -1; |
7152 | du->minor = 0; | |
7153 | du->major = 0; | |
7154 | du->index = (i * 2) + 1; | |
7155 | sprintf((char *)du->disk.serial, | |
7156 | " MISSING_%d", du->index); | |
7157 | sprintf((char *)du->serial, | |
7158 | "MISSING_%d", du->index); | |
7159 | du->next = super->missing; | |
7160 | super->missing = du; | |
7161 | } | |
7162 | /* create new dev and map */ | |
7163 | space = *space_list; | |
7164 | if (!space) | |
7165 | return 0; | |
7166 | *space_list = *space; | |
7167 | dev_new = (void *)space; | |
7168 | memcpy(dev_new, dev, sizeof(*dev)); | |
7169 | /* update new map */ | |
7170 | map = get_imsm_map(dev_new, 0); | |
8ca6df95 | 7171 | map->num_members = map->num_members * 2; |
1a2487c2 | 7172 | map->map_state = IMSM_T_STATE_DEGRADED; |
8ca6df95 KW |
7173 | map->num_domains = 2; |
7174 | map->raid_level = 1; | |
7175 | /* replace dev<->dev_new */ | |
7176 | dv->dev = dev_new; | |
7177 | } | |
bb025c2f KW |
7178 | /* update disk order table */ |
7179 | for (du = super->disks; du; du = du->next) | |
7180 | if (du->index >= 0) | |
7181 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
8ca6df95 | 7182 | for (du = super->missing; du; du = du->next) |
1a2487c2 KW |
7183 | if (du->index >= 0) { |
7184 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
e4c72d1d | 7185 | mark_missing(dv->dev, &du->disk, du->index); |
1a2487c2 | 7186 | } |
bb025c2f KW |
7187 | |
7188 | return 1; | |
7189 | } | |
7190 | ||
e8319a19 DW |
7191 | static void imsm_process_update(struct supertype *st, |
7192 | struct metadata_update *update) | |
7193 | { | |
7194 | /** | |
7195 | * crack open the metadata_update envelope to find the update record | |
7196 | * update can be one of: | |
d195167d AK |
7197 | * update_reshape_container_disks - all the arrays in the container |
7198 | * are being reshaped to have more devices. We need to mark | |
7199 | * the arrays for general migration and convert selected spares | |
7200 | * into active devices. | |
7201 | * update_activate_spare - a spare device has replaced a failed | |
e8319a19 DW |
7202 | * device in an array, update the disk_ord_tbl. If this disk is |
7203 | * present in all member arrays then also clear the SPARE_DISK | |
7204 | * flag | |
d195167d AK |
7205 | * update_create_array |
7206 | * update_kill_array | |
7207 | * update_rename_array | |
7208 | * update_add_remove_disk | |
e8319a19 DW |
7209 | */ |
7210 | struct intel_super *super = st->sb; | |
4d7b1503 | 7211 | struct imsm_super *mpb; |
e8319a19 DW |
7212 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
7213 | ||
4d7b1503 DW |
7214 | /* update requires a larger buf but the allocation failed */ |
7215 | if (super->next_len && !super->next_buf) { | |
7216 | super->next_len = 0; | |
7217 | return; | |
7218 | } | |
7219 | ||
7220 | if (super->next_buf) { | |
7221 | memcpy(super->next_buf, super->buf, super->len); | |
7222 | free(super->buf); | |
7223 | super->len = super->next_len; | |
7224 | super->buf = super->next_buf; | |
7225 | ||
7226 | super->next_len = 0; | |
7227 | super->next_buf = NULL; | |
7228 | } | |
7229 | ||
7230 | mpb = super->anchor; | |
7231 | ||
e8319a19 | 7232 | switch (type) { |
0ec5d470 AK |
7233 | case update_general_migration_checkpoint: { |
7234 | struct intel_dev *id; | |
7235 | struct imsm_update_general_migration_checkpoint *u = | |
7236 | (void *)update->buf; | |
7237 | ||
7238 | dprintf("imsm: process_update() " | |
7239 | "for update_general_migration_checkpoint called\n"); | |
7240 | ||
7241 | /* find device under general migration */ | |
7242 | for (id = super->devlist ; id; id = id->next) { | |
7243 | if (is_gen_migration(id->dev)) { | |
7244 | id->dev->vol.curr_migr_unit = | |
7245 | __cpu_to_le32(u->curr_migr_unit); | |
7246 | super->updates_pending++; | |
7247 | } | |
7248 | } | |
7249 | break; | |
7250 | } | |
bb025c2f KW |
7251 | case update_takeover: { |
7252 | struct imsm_update_takeover *u = (void *)update->buf; | |
1a2487c2 KW |
7253 | if (apply_takeover_update(u, super, &update->space_list)) { |
7254 | imsm_update_version_info(super); | |
bb025c2f | 7255 | super->updates_pending++; |
1a2487c2 | 7256 | } |
bb025c2f KW |
7257 | break; |
7258 | } | |
7259 | ||
78b10e66 | 7260 | case update_reshape_container_disks: { |
d195167d | 7261 | struct imsm_update_reshape *u = (void *)update->buf; |
2e5dc010 N |
7262 | if (apply_reshape_container_disks_update( |
7263 | u, super, &update->space_list)) | |
7264 | super->updates_pending++; | |
78b10e66 N |
7265 | break; |
7266 | } | |
48c5303a | 7267 | case update_reshape_migration: { |
a29911da PC |
7268 | struct imsm_update_reshape_migration *u = (void *)update->buf; |
7269 | if (apply_reshape_migration_update( | |
7270 | u, super, &update->space_list)) | |
7271 | super->updates_pending++; | |
48c5303a PC |
7272 | break; |
7273 | } | |
e8319a19 DW |
7274 | case update_activate_spare: { |
7275 | struct imsm_update_activate_spare *u = (void *) update->buf; | |
949c47a0 | 7276 | struct imsm_dev *dev = get_imsm_dev(super, u->array); |
a965f303 | 7277 | struct imsm_map *map = get_imsm_map(dev, 0); |
0c046afd | 7278 | struct imsm_map *migr_map; |
e8319a19 DW |
7279 | struct active_array *a; |
7280 | struct imsm_disk *disk; | |
0c046afd | 7281 | __u8 to_state; |
e8319a19 | 7282 | struct dl *dl; |
e8319a19 | 7283 | unsigned int found; |
0c046afd | 7284 | int failed; |
98130f40 | 7285 | int victim = get_imsm_disk_idx(dev, u->slot, -1); |
e8319a19 DW |
7286 | int i; |
7287 | ||
7288 | for (dl = super->disks; dl; dl = dl->next) | |
d23fe947 | 7289 | if (dl == u->dl) |
e8319a19 DW |
7290 | break; |
7291 | ||
7292 | if (!dl) { | |
7293 | fprintf(stderr, "error: imsm_activate_spare passed " | |
1f24f035 DW |
7294 | "an unknown disk (index: %d)\n", |
7295 | u->dl->index); | |
e8319a19 DW |
7296 | return; |
7297 | } | |
7298 | ||
7299 | super->updates_pending++; | |
0c046afd DW |
7300 | /* count failures (excluding rebuilds and the victim) |
7301 | * to determine map[0] state | |
7302 | */ | |
7303 | failed = 0; | |
7304 | for (i = 0; i < map->num_members; i++) { | |
7305 | if (i == u->slot) | |
7306 | continue; | |
98130f40 AK |
7307 | disk = get_imsm_disk(super, |
7308 | get_imsm_disk_idx(dev, i, -1)); | |
25ed7e59 | 7309 | if (!disk || is_failed(disk)) |
0c046afd DW |
7310 | failed++; |
7311 | } | |
7312 | ||
d23fe947 DW |
7313 | /* adding a pristine spare, assign a new index */ |
7314 | if (dl->index < 0) { | |
7315 | dl->index = super->anchor->num_disks; | |
7316 | super->anchor->num_disks++; | |
7317 | } | |
d23fe947 | 7318 | disk = &dl->disk; |
f2f27e63 DW |
7319 | disk->status |= CONFIGURED_DISK; |
7320 | disk->status &= ~SPARE_DISK; | |
e8319a19 | 7321 | |
0c046afd DW |
7322 | /* mark rebuild */ |
7323 | to_state = imsm_check_degraded(super, dev, failed); | |
7324 | map->map_state = IMSM_T_STATE_DEGRADED; | |
8e59f3d8 | 7325 | migrate(dev, super, to_state, MIGR_REBUILD); |
0c046afd DW |
7326 | migr_map = get_imsm_map(dev, 1); |
7327 | set_imsm_ord_tbl_ent(map, u->slot, dl->index); | |
7328 | set_imsm_ord_tbl_ent(migr_map, u->slot, dl->index | IMSM_ORD_REBUILD); | |
7329 | ||
148acb7b DW |
7330 | /* update the family_num to mark a new container |
7331 | * generation, being careful to record the existing | |
7332 | * family_num in orig_family_num to clean up after | |
7333 | * earlier mdadm versions that neglected to set it. | |
7334 | */ | |
7335 | if (mpb->orig_family_num == 0) | |
7336 | mpb->orig_family_num = mpb->family_num; | |
7337 | mpb->family_num += super->random; | |
7338 | ||
e8319a19 DW |
7339 | /* count arrays using the victim in the metadata */ |
7340 | found = 0; | |
7341 | for (a = st->arrays; a ; a = a->next) { | |
949c47a0 | 7342 | dev = get_imsm_dev(super, a->info.container_member); |
620b1713 DW |
7343 | map = get_imsm_map(dev, 0); |
7344 | ||
7345 | if (get_imsm_disk_slot(map, victim) >= 0) | |
7346 | found++; | |
e8319a19 DW |
7347 | } |
7348 | ||
24565c9a | 7349 | /* delete the victim if it is no longer being |
e8319a19 DW |
7350 | * utilized anywhere |
7351 | */ | |
e8319a19 | 7352 | if (!found) { |
ae6aad82 | 7353 | struct dl **dlp; |
24565c9a | 7354 | |
47ee5a45 DW |
7355 | /* We know that 'manager' isn't touching anything, |
7356 | * so it is safe to delete | |
7357 | */ | |
24565c9a | 7358 | for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next) |
ae6aad82 DW |
7359 | if ((*dlp)->index == victim) |
7360 | break; | |
47ee5a45 DW |
7361 | |
7362 | /* victim may be on the missing list */ | |
7363 | if (!*dlp) | |
7364 | for (dlp = &super->missing; *dlp; dlp = &(*dlp)->next) | |
7365 | if ((*dlp)->index == victim) | |
7366 | break; | |
24565c9a | 7367 | imsm_delete(super, dlp, victim); |
e8319a19 | 7368 | } |
8273f55e DW |
7369 | break; |
7370 | } | |
7371 | case update_create_array: { | |
7372 | /* someone wants to create a new array, we need to be aware of | |
7373 | * a few races/collisions: | |
7374 | * 1/ 'Create' called by two separate instances of mdadm | |
7375 | * 2/ 'Create' versus 'activate_spare': mdadm has chosen | |
7376 | * devices that have since been assimilated via | |
7377 | * activate_spare. | |
7378 | * In the event this update can not be carried out mdadm will | |
7379 | * (FIX ME) notice that its update did not take hold. | |
7380 | */ | |
7381 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 7382 | struct intel_dev *dv; |
8273f55e DW |
7383 | struct imsm_dev *dev; |
7384 | struct imsm_map *map, *new_map; | |
7385 | unsigned long long start, end; | |
7386 | unsigned long long new_start, new_end; | |
7387 | int i; | |
54c2c1ea DW |
7388 | struct disk_info *inf; |
7389 | struct dl *dl; | |
8273f55e DW |
7390 | |
7391 | /* handle racing creates: first come first serve */ | |
7392 | if (u->dev_idx < mpb->num_raid_devs) { | |
7393 | dprintf("%s: subarray %d already defined\n", | |
7394 | __func__, u->dev_idx); | |
ba2de7ba | 7395 | goto create_error; |
8273f55e DW |
7396 | } |
7397 | ||
7398 | /* check update is next in sequence */ | |
7399 | if (u->dev_idx != mpb->num_raid_devs) { | |
6a3e913e DW |
7400 | dprintf("%s: can not create array %d expected index %d\n", |
7401 | __func__, u->dev_idx, mpb->num_raid_devs); | |
ba2de7ba | 7402 | goto create_error; |
8273f55e DW |
7403 | } |
7404 | ||
a965f303 | 7405 | new_map = get_imsm_map(&u->dev, 0); |
8273f55e DW |
7406 | new_start = __le32_to_cpu(new_map->pba_of_lba0); |
7407 | new_end = new_start + __le32_to_cpu(new_map->blocks_per_member); | |
54c2c1ea | 7408 | inf = get_disk_info(u); |
8273f55e DW |
7409 | |
7410 | /* handle activate_spare versus create race: | |
7411 | * check to make sure that overlapping arrays do not include | |
7412 | * overalpping disks | |
7413 | */ | |
7414 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
949c47a0 | 7415 | dev = get_imsm_dev(super, i); |
a965f303 | 7416 | map = get_imsm_map(dev, 0); |
8273f55e DW |
7417 | start = __le32_to_cpu(map->pba_of_lba0); |
7418 | end = start + __le32_to_cpu(map->blocks_per_member); | |
7419 | if ((new_start >= start && new_start <= end) || | |
7420 | (start >= new_start && start <= new_end)) | |
54c2c1ea DW |
7421 | /* overlap */; |
7422 | else | |
7423 | continue; | |
7424 | ||
7425 | if (disks_overlap(super, i, u)) { | |
8273f55e | 7426 | dprintf("%s: arrays overlap\n", __func__); |
ba2de7ba | 7427 | goto create_error; |
8273f55e DW |
7428 | } |
7429 | } | |
8273f55e | 7430 | |
949c47a0 DW |
7431 | /* check that prepare update was successful */ |
7432 | if (!update->space) { | |
7433 | dprintf("%s: prepare update failed\n", __func__); | |
ba2de7ba | 7434 | goto create_error; |
949c47a0 DW |
7435 | } |
7436 | ||
54c2c1ea DW |
7437 | /* check that all disks are still active before committing |
7438 | * changes. FIXME: could we instead handle this by creating a | |
7439 | * degraded array? That's probably not what the user expects, | |
7440 | * so better to drop this update on the floor. | |
7441 | */ | |
7442 | for (i = 0; i < new_map->num_members; i++) { | |
7443 | dl = serial_to_dl(inf[i].serial, super); | |
7444 | if (!dl) { | |
7445 | dprintf("%s: disk disappeared\n", __func__); | |
ba2de7ba | 7446 | goto create_error; |
54c2c1ea | 7447 | } |
949c47a0 DW |
7448 | } |
7449 | ||
8273f55e | 7450 | super->updates_pending++; |
54c2c1ea DW |
7451 | |
7452 | /* convert spares to members and fixup ord_tbl */ | |
7453 | for (i = 0; i < new_map->num_members; i++) { | |
7454 | dl = serial_to_dl(inf[i].serial, super); | |
7455 | if (dl->index == -1) { | |
7456 | dl->index = mpb->num_disks; | |
7457 | mpb->num_disks++; | |
7458 | dl->disk.status |= CONFIGURED_DISK; | |
7459 | dl->disk.status &= ~SPARE_DISK; | |
7460 | } | |
7461 | set_imsm_ord_tbl_ent(new_map, i, dl->index); | |
7462 | } | |
7463 | ||
ba2de7ba DW |
7464 | dv = update->space; |
7465 | dev = dv->dev; | |
949c47a0 DW |
7466 | update->space = NULL; |
7467 | imsm_copy_dev(dev, &u->dev); | |
ba2de7ba DW |
7468 | dv->index = u->dev_idx; |
7469 | dv->next = super->devlist; | |
7470 | super->devlist = dv; | |
8273f55e | 7471 | mpb->num_raid_devs++; |
8273f55e | 7472 | |
4d1313e9 | 7473 | imsm_update_version_info(super); |
8273f55e | 7474 | break; |
ba2de7ba DW |
7475 | create_error: |
7476 | /* mdmon knows how to release update->space, but not | |
7477 | * ((struct intel_dev *) update->space)->dev | |
7478 | */ | |
7479 | if (update->space) { | |
7480 | dv = update->space; | |
7481 | free(dv->dev); | |
7482 | } | |
8273f55e | 7483 | break; |
e8319a19 | 7484 | } |
33414a01 DW |
7485 | case update_kill_array: { |
7486 | struct imsm_update_kill_array *u = (void *) update->buf; | |
7487 | int victim = u->dev_idx; | |
7488 | struct active_array *a; | |
7489 | struct intel_dev **dp; | |
7490 | struct imsm_dev *dev; | |
7491 | ||
7492 | /* sanity check that we are not affecting the uuid of | |
7493 | * active arrays, or deleting an active array | |
7494 | * | |
7495 | * FIXME when immutable ids are available, but note that | |
7496 | * we'll also need to fixup the invalidated/active | |
7497 | * subarray indexes in mdstat | |
7498 | */ | |
7499 | for (a = st->arrays; a; a = a->next) | |
7500 | if (a->info.container_member >= victim) | |
7501 | break; | |
7502 | /* by definition if mdmon is running at least one array | |
7503 | * is active in the container, so checking | |
7504 | * mpb->num_raid_devs is just extra paranoia | |
7505 | */ | |
7506 | dev = get_imsm_dev(super, victim); | |
7507 | if (a || !dev || mpb->num_raid_devs == 1) { | |
7508 | dprintf("failed to delete subarray-%d\n", victim); | |
7509 | break; | |
7510 | } | |
7511 | ||
7512 | for (dp = &super->devlist; *dp;) | |
f21e18ca | 7513 | if ((*dp)->index == (unsigned)super->current_vol) { |
33414a01 DW |
7514 | *dp = (*dp)->next; |
7515 | } else { | |
f21e18ca | 7516 | if ((*dp)->index > (unsigned)victim) |
33414a01 DW |
7517 | (*dp)->index--; |
7518 | dp = &(*dp)->next; | |
7519 | } | |
7520 | mpb->num_raid_devs--; | |
7521 | super->updates_pending++; | |
7522 | break; | |
7523 | } | |
aa534678 DW |
7524 | case update_rename_array: { |
7525 | struct imsm_update_rename_array *u = (void *) update->buf; | |
7526 | char name[MAX_RAID_SERIAL_LEN+1]; | |
7527 | int target = u->dev_idx; | |
7528 | struct active_array *a; | |
7529 | struct imsm_dev *dev; | |
7530 | ||
7531 | /* sanity check that we are not affecting the uuid of | |
7532 | * an active array | |
7533 | */ | |
7534 | snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name); | |
7535 | name[MAX_RAID_SERIAL_LEN] = '\0'; | |
7536 | for (a = st->arrays; a; a = a->next) | |
7537 | if (a->info.container_member == target) | |
7538 | break; | |
7539 | dev = get_imsm_dev(super, u->dev_idx); | |
7540 | if (a || !dev || !check_name(super, name, 1)) { | |
7541 | dprintf("failed to rename subarray-%d\n", target); | |
7542 | break; | |
7543 | } | |
7544 | ||
cdbe98cd | 7545 | snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name); |
aa534678 DW |
7546 | super->updates_pending++; |
7547 | break; | |
7548 | } | |
1a64be56 | 7549 | case update_add_remove_disk: { |
43dad3d6 | 7550 | /* we may be able to repair some arrays if disks are |
1a64be56 LM |
7551 | * being added, check teh status of add_remove_disk |
7552 | * if discs has been added. | |
7553 | */ | |
7554 | if (add_remove_disk_update(super)) { | |
43dad3d6 | 7555 | struct active_array *a; |
072b727f DW |
7556 | |
7557 | super->updates_pending++; | |
1a64be56 | 7558 | for (a = st->arrays; a; a = a->next) |
43dad3d6 DW |
7559 | a->check_degraded = 1; |
7560 | } | |
43dad3d6 | 7561 | break; |
e8319a19 | 7562 | } |
1a64be56 LM |
7563 | default: |
7564 | fprintf(stderr, "error: unsuported process update type:" | |
7565 | "(type: %d)\n", type); | |
7566 | } | |
e8319a19 | 7567 | } |
88758e9d | 7568 | |
bc0b9d34 PC |
7569 | static struct mdinfo *get_spares_for_grow(struct supertype *st); |
7570 | ||
8273f55e DW |
7571 | static void imsm_prepare_update(struct supertype *st, |
7572 | struct metadata_update *update) | |
7573 | { | |
949c47a0 | 7574 | /** |
4d7b1503 DW |
7575 | * Allocate space to hold new disk entries, raid-device entries or a new |
7576 | * mpb if necessary. The manager synchronously waits for updates to | |
7577 | * complete in the monitor, so new mpb buffers allocated here can be | |
7578 | * integrated by the monitor thread without worrying about live pointers | |
7579 | * in the manager thread. | |
8273f55e | 7580 | */ |
949c47a0 | 7581 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
4d7b1503 DW |
7582 | struct intel_super *super = st->sb; |
7583 | struct imsm_super *mpb = super->anchor; | |
7584 | size_t buf_len; | |
7585 | size_t len = 0; | |
949c47a0 DW |
7586 | |
7587 | switch (type) { | |
0ec5d470 AK |
7588 | case update_general_migration_checkpoint: |
7589 | dprintf("imsm: prepare_update() " | |
7590 | "for update_general_migration_checkpoint called\n"); | |
7591 | break; | |
abedf5fc KW |
7592 | case update_takeover: { |
7593 | struct imsm_update_takeover *u = (void *)update->buf; | |
7594 | if (u->direction == R0_TO_R10) { | |
7595 | void **tail = (void **)&update->space_list; | |
7596 | struct imsm_dev *dev = get_imsm_dev(super, u->subarray); | |
7597 | struct imsm_map *map = get_imsm_map(dev, 0); | |
7598 | int num_members = map->num_members; | |
7599 | void *space; | |
7600 | int size, i; | |
7601 | int err = 0; | |
7602 | /* allocate memory for added disks */ | |
7603 | for (i = 0; i < num_members; i++) { | |
7604 | size = sizeof(struct dl); | |
7605 | space = malloc(size); | |
7606 | if (!space) { | |
7607 | err++; | |
7608 | break; | |
7609 | } | |
7610 | *tail = space; | |
7611 | tail = space; | |
7612 | *tail = NULL; | |
7613 | } | |
7614 | /* allocate memory for new device */ | |
7615 | size = sizeof_imsm_dev(super->devlist->dev, 0) + | |
7616 | (num_members * sizeof(__u32)); | |
7617 | space = malloc(size); | |
7618 | if (!space) | |
7619 | err++; | |
7620 | else { | |
7621 | *tail = space; | |
7622 | tail = space; | |
7623 | *tail = NULL; | |
7624 | } | |
7625 | if (!err) { | |
7626 | len = disks_to_mpb_size(num_members * 2); | |
7627 | } else { | |
7628 | /* if allocation didn't success, free buffer */ | |
7629 | while (update->space_list) { | |
7630 | void **sp = update->space_list; | |
7631 | update->space_list = *sp; | |
7632 | free(sp); | |
7633 | } | |
7634 | } | |
7635 | } | |
7636 | ||
7637 | break; | |
7638 | } | |
78b10e66 | 7639 | case update_reshape_container_disks: { |
d195167d AK |
7640 | /* Every raid device in the container is about to |
7641 | * gain some more devices, and we will enter a | |
7642 | * reconfiguration. | |
7643 | * So each 'imsm_map' will be bigger, and the imsm_vol | |
7644 | * will now hold 2 of them. | |
7645 | * Thus we need new 'struct imsm_dev' allocations sized | |
7646 | * as sizeof_imsm_dev but with more devices in both maps. | |
7647 | */ | |
7648 | struct imsm_update_reshape *u = (void *)update->buf; | |
7649 | struct intel_dev *dl; | |
7650 | void **space_tail = (void**)&update->space_list; | |
7651 | ||
7652 | dprintf("imsm: imsm_prepare_update() for update_reshape\n"); | |
7653 | ||
7654 | for (dl = super->devlist; dl; dl = dl->next) { | |
7655 | int size = sizeof_imsm_dev(dl->dev, 1); | |
7656 | void *s; | |
d677e0b8 AK |
7657 | if (u->new_raid_disks > u->old_raid_disks) |
7658 | size += sizeof(__u32)*2* | |
7659 | (u->new_raid_disks - u->old_raid_disks); | |
d195167d AK |
7660 | s = malloc(size); |
7661 | if (!s) | |
7662 | break; | |
7663 | *space_tail = s; | |
7664 | space_tail = s; | |
7665 | *space_tail = NULL; | |
7666 | } | |
7667 | ||
7668 | len = disks_to_mpb_size(u->new_raid_disks); | |
7669 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
78b10e66 N |
7670 | break; |
7671 | } | |
48c5303a | 7672 | case update_reshape_migration: { |
bc0b9d34 PC |
7673 | /* for migration level 0->5 we need to add disks |
7674 | * so the same as for container operation we will copy | |
7675 | * device to the bigger location. | |
7676 | * in memory prepared device and new disk area are prepared | |
7677 | * for usage in process update | |
7678 | */ | |
7679 | struct imsm_update_reshape_migration *u = (void *)update->buf; | |
7680 | struct intel_dev *id; | |
7681 | void **space_tail = (void **)&update->space_list; | |
7682 | int size; | |
7683 | void *s; | |
7684 | int current_level = -1; | |
7685 | ||
7686 | dprintf("imsm: imsm_prepare_update() for update_reshape\n"); | |
7687 | ||
7688 | /* add space for bigger array in update | |
7689 | */ | |
7690 | for (id = super->devlist; id; id = id->next) { | |
7691 | if (id->index == (unsigned)u->subdev) { | |
7692 | size = sizeof_imsm_dev(id->dev, 1); | |
7693 | if (u->new_raid_disks > u->old_raid_disks) | |
7694 | size += sizeof(__u32)*2* | |
7695 | (u->new_raid_disks - u->old_raid_disks); | |
7696 | s = malloc(size); | |
7697 | if (!s) | |
7698 | break; | |
7699 | *space_tail = s; | |
7700 | space_tail = s; | |
7701 | *space_tail = NULL; | |
7702 | break; | |
7703 | } | |
7704 | } | |
7705 | if (update->space_list == NULL) | |
7706 | break; | |
7707 | ||
7708 | /* add space for disk in update | |
7709 | */ | |
7710 | size = sizeof(struct dl); | |
7711 | s = malloc(size); | |
7712 | if (!s) { | |
7713 | free(update->space_list); | |
7714 | update->space_list = NULL; | |
7715 | break; | |
7716 | } | |
7717 | *space_tail = s; | |
7718 | space_tail = s; | |
7719 | *space_tail = NULL; | |
7720 | ||
7721 | /* add spare device to update | |
7722 | */ | |
7723 | for (id = super->devlist ; id; id = id->next) | |
7724 | if (id->index == (unsigned)u->subdev) { | |
7725 | struct imsm_dev *dev; | |
7726 | struct imsm_map *map; | |
7727 | ||
7728 | dev = get_imsm_dev(super, u->subdev); | |
7729 | map = get_imsm_map(dev, 0); | |
7730 | current_level = map->raid_level; | |
7731 | break; | |
7732 | } | |
7733 | if ((u->new_level == 5) && (u->new_level != current_level)) { | |
7734 | struct mdinfo *spares; | |
7735 | ||
7736 | spares = get_spares_for_grow(st); | |
7737 | if (spares) { | |
7738 | struct dl *dl; | |
7739 | struct mdinfo *dev; | |
7740 | ||
7741 | dev = spares->devs; | |
7742 | if (dev) { | |
7743 | u->new_disks[0] = | |
7744 | makedev(dev->disk.major, | |
7745 | dev->disk.minor); | |
7746 | dl = get_disk_super(super, | |
7747 | dev->disk.major, | |
7748 | dev->disk.minor); | |
7749 | dl->index = u->old_raid_disks; | |
7750 | dev = dev->next; | |
7751 | } | |
7752 | sysfs_free(spares); | |
7753 | } | |
7754 | } | |
7755 | len = disks_to_mpb_size(u->new_raid_disks); | |
7756 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
48c5303a PC |
7757 | break; |
7758 | } | |
949c47a0 DW |
7759 | case update_create_array: { |
7760 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 7761 | struct intel_dev *dv; |
54c2c1ea DW |
7762 | struct imsm_dev *dev = &u->dev; |
7763 | struct imsm_map *map = get_imsm_map(dev, 0); | |
7764 | struct dl *dl; | |
7765 | struct disk_info *inf; | |
7766 | int i; | |
7767 | int activate = 0; | |
949c47a0 | 7768 | |
54c2c1ea DW |
7769 | inf = get_disk_info(u); |
7770 | len = sizeof_imsm_dev(dev, 1); | |
ba2de7ba DW |
7771 | /* allocate a new super->devlist entry */ |
7772 | dv = malloc(sizeof(*dv)); | |
7773 | if (dv) { | |
7774 | dv->dev = malloc(len); | |
7775 | if (dv->dev) | |
7776 | update->space = dv; | |
7777 | else { | |
7778 | free(dv); | |
7779 | update->space = NULL; | |
7780 | } | |
7781 | } | |
949c47a0 | 7782 | |
54c2c1ea DW |
7783 | /* count how many spares will be converted to members */ |
7784 | for (i = 0; i < map->num_members; i++) { | |
7785 | dl = serial_to_dl(inf[i].serial, super); | |
7786 | if (!dl) { | |
7787 | /* hmm maybe it failed?, nothing we can do about | |
7788 | * it here | |
7789 | */ | |
7790 | continue; | |
7791 | } | |
7792 | if (count_memberships(dl, super) == 0) | |
7793 | activate++; | |
7794 | } | |
7795 | len += activate * sizeof(struct imsm_disk); | |
949c47a0 DW |
7796 | break; |
7797 | default: | |
7798 | break; | |
7799 | } | |
7800 | } | |
8273f55e | 7801 | |
4d7b1503 DW |
7802 | /* check if we need a larger metadata buffer */ |
7803 | if (super->next_buf) | |
7804 | buf_len = super->next_len; | |
7805 | else | |
7806 | buf_len = super->len; | |
7807 | ||
7808 | if (__le32_to_cpu(mpb->mpb_size) + len > buf_len) { | |
7809 | /* ok we need a larger buf than what is currently allocated | |
7810 | * if this allocation fails process_update will notice that | |
7811 | * ->next_len is set and ->next_buf is NULL | |
7812 | */ | |
7813 | buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + len, 512); | |
7814 | if (super->next_buf) | |
7815 | free(super->next_buf); | |
7816 | ||
7817 | super->next_len = buf_len; | |
1f45a8ad DW |
7818 | if (posix_memalign(&super->next_buf, 512, buf_len) == 0) |
7819 | memset(super->next_buf, 0, buf_len); | |
7820 | else | |
4d7b1503 DW |
7821 | super->next_buf = NULL; |
7822 | } | |
8273f55e DW |
7823 | } |
7824 | ||
ae6aad82 | 7825 | /* must be called while manager is quiesced */ |
f21e18ca | 7826 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index) |
ae6aad82 DW |
7827 | { |
7828 | struct imsm_super *mpb = super->anchor; | |
ae6aad82 DW |
7829 | struct dl *iter; |
7830 | struct imsm_dev *dev; | |
7831 | struct imsm_map *map; | |
24565c9a DW |
7832 | int i, j, num_members; |
7833 | __u32 ord; | |
ae6aad82 | 7834 | |
24565c9a DW |
7835 | dprintf("%s: deleting device[%d] from imsm_super\n", |
7836 | __func__, index); | |
ae6aad82 DW |
7837 | |
7838 | /* shift all indexes down one */ | |
7839 | for (iter = super->disks; iter; iter = iter->next) | |
f21e18ca | 7840 | if (iter->index > (int)index) |
ae6aad82 | 7841 | iter->index--; |
47ee5a45 | 7842 | for (iter = super->missing; iter; iter = iter->next) |
f21e18ca | 7843 | if (iter->index > (int)index) |
47ee5a45 | 7844 | iter->index--; |
ae6aad82 DW |
7845 | |
7846 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
7847 | dev = get_imsm_dev(super, i); | |
7848 | map = get_imsm_map(dev, 0); | |
24565c9a DW |
7849 | num_members = map->num_members; |
7850 | for (j = 0; j < num_members; j++) { | |
7851 | /* update ord entries being careful not to propagate | |
7852 | * ord-flags to the first map | |
7853 | */ | |
98130f40 | 7854 | ord = get_imsm_ord_tbl_ent(dev, j, -1); |
ae6aad82 | 7855 | |
24565c9a DW |
7856 | if (ord_to_idx(ord) <= index) |
7857 | continue; | |
ae6aad82 | 7858 | |
24565c9a DW |
7859 | map = get_imsm_map(dev, 0); |
7860 | set_imsm_ord_tbl_ent(map, j, ord_to_idx(ord - 1)); | |
7861 | map = get_imsm_map(dev, 1); | |
7862 | if (map) | |
7863 | set_imsm_ord_tbl_ent(map, j, ord - 1); | |
ae6aad82 DW |
7864 | } |
7865 | } | |
7866 | ||
7867 | mpb->num_disks--; | |
7868 | super->updates_pending++; | |
24565c9a DW |
7869 | if (*dlp) { |
7870 | struct dl *dl = *dlp; | |
7871 | ||
7872 | *dlp = (*dlp)->next; | |
7873 | __free_imsm_disk(dl); | |
7874 | } | |
ae6aad82 | 7875 | } |
9e2d750d | 7876 | #endif /* MDASSEMBLE */ |
687629c2 AK |
7877 | /******************************************************************************* |
7878 | * Function: open_backup_targets | |
7879 | * Description: Function opens file descriptors for all devices given in | |
7880 | * info->devs | |
7881 | * Parameters: | |
7882 | * info : general array info | |
7883 | * raid_disks : number of disks | |
7884 | * raid_fds : table of device's file descriptors | |
7885 | * Returns: | |
7886 | * 0 : success | |
7887 | * -1 : fail | |
7888 | ******************************************************************************/ | |
7889 | int open_backup_targets(struct mdinfo *info, int raid_disks, int *raid_fds) | |
7890 | { | |
7891 | struct mdinfo *sd; | |
7892 | ||
7893 | for (sd = info->devs ; sd ; sd = sd->next) { | |
7894 | char *dn; | |
7895 | ||
7896 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) { | |
7897 | dprintf("disk is faulty!!\n"); | |
7898 | continue; | |
7899 | } | |
7900 | ||
7901 | if ((sd->disk.raid_disk >= raid_disks) || | |
7902 | (sd->disk.raid_disk < 0)) | |
7903 | continue; | |
7904 | ||
7905 | dn = map_dev(sd->disk.major, | |
7906 | sd->disk.minor, 1); | |
7907 | raid_fds[sd->disk.raid_disk] = dev_open(dn, O_RDWR); | |
7908 | if (raid_fds[sd->disk.raid_disk] < 0) { | |
7909 | fprintf(stderr, "cannot open component\n"); | |
7910 | return -1; | |
7911 | } | |
7912 | } | |
7913 | return 0; | |
7914 | } | |
7915 | ||
9e2d750d | 7916 | #ifndef MDASSEMBLE |
687629c2 AK |
7917 | /******************************************************************************* |
7918 | * Function: init_migr_record_imsm | |
7919 | * Description: Function inits imsm migration record | |
7920 | * Parameters: | |
7921 | * super : imsm internal array info | |
7922 | * dev : device under migration | |
7923 | * info : general array info to find the smallest device | |
7924 | * Returns: | |
7925 | * none | |
7926 | ******************************************************************************/ | |
7927 | void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev, | |
7928 | struct mdinfo *info) | |
7929 | { | |
7930 | struct intel_super *super = st->sb; | |
7931 | struct migr_record *migr_rec = super->migr_rec; | |
7932 | int new_data_disks; | |
7933 | unsigned long long dsize, dev_sectors; | |
7934 | long long unsigned min_dev_sectors = -1LLU; | |
7935 | struct mdinfo *sd; | |
7936 | char nm[30]; | |
7937 | int fd; | |
7938 | struct imsm_map *map_dest = get_imsm_map(dev, 0); | |
7939 | struct imsm_map *map_src = get_imsm_map(dev, 1); | |
7940 | unsigned long long num_migr_units; | |
3ef4403c | 7941 | unsigned long long array_blocks; |
687629c2 AK |
7942 | |
7943 | memset(migr_rec, 0, sizeof(struct migr_record)); | |
7944 | migr_rec->family_num = __cpu_to_le32(super->anchor->family_num); | |
7945 | ||
7946 | /* only ascending reshape supported now */ | |
7947 | migr_rec->ascending_migr = __cpu_to_le32(1); | |
7948 | ||
7949 | migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE / | |
7950 | max(map_dest->blocks_per_strip, map_src->blocks_per_strip); | |
7951 | migr_rec->dest_depth_per_unit *= map_dest->blocks_per_strip; | |
7952 | new_data_disks = imsm_num_data_members(dev, 0); | |
7953 | migr_rec->blocks_per_unit = | |
7954 | __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks); | |
7955 | migr_rec->dest_depth_per_unit = | |
7956 | __cpu_to_le32(migr_rec->dest_depth_per_unit); | |
3ef4403c | 7957 | array_blocks = info->component_size * new_data_disks; |
687629c2 AK |
7958 | num_migr_units = |
7959 | array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit); | |
7960 | ||
7961 | if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit)) | |
7962 | num_migr_units++; | |
7963 | migr_rec->num_migr_units = __cpu_to_le32(num_migr_units); | |
7964 | ||
7965 | migr_rec->post_migr_vol_cap = dev->size_low; | |
7966 | migr_rec->post_migr_vol_cap_hi = dev->size_high; | |
7967 | ||
7968 | ||
7969 | /* Find the smallest dev */ | |
7970 | for (sd = info->devs ; sd ; sd = sd->next) { | |
7971 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
7972 | fd = dev_open(nm, O_RDONLY); | |
7973 | if (fd < 0) | |
7974 | continue; | |
7975 | get_dev_size(fd, NULL, &dsize); | |
7976 | dev_sectors = dsize / 512; | |
7977 | if (dev_sectors < min_dev_sectors) | |
7978 | min_dev_sectors = dev_sectors; | |
7979 | close(fd); | |
7980 | } | |
7981 | migr_rec->ckpt_area_pba = __cpu_to_le32(min_dev_sectors - | |
7982 | RAID_DISK_RESERVED_BLOCKS_IMSM_HI); | |
7983 | ||
7984 | write_imsm_migr_rec(st); | |
7985 | ||
7986 | return; | |
7987 | } | |
7988 | ||
7989 | /******************************************************************************* | |
7990 | * Function: save_backup_imsm | |
7991 | * Description: Function saves critical data stripes to Migration Copy Area | |
7992 | * and updates the current migration unit status. | |
7993 | * Use restore_stripes() to form a destination stripe, | |
7994 | * and to write it to the Copy Area. | |
7995 | * Parameters: | |
7996 | * st : supertype information | |
aea93171 | 7997 | * dev : imsm device that backup is saved for |
687629c2 AK |
7998 | * info : general array info |
7999 | * buf : input buffer | |
687629c2 AK |
8000 | * length : length of data to backup (blocks_per_unit) |
8001 | * Returns: | |
8002 | * 0 : success | |
8003 | *, -1 : fail | |
8004 | ******************************************************************************/ | |
8005 | int save_backup_imsm(struct supertype *st, | |
8006 | struct imsm_dev *dev, | |
8007 | struct mdinfo *info, | |
8008 | void *buf, | |
687629c2 AK |
8009 | int length) |
8010 | { | |
8011 | int rv = -1; | |
8012 | struct intel_super *super = st->sb; | |
8013 | unsigned long long *target_offsets = NULL; | |
8014 | int *targets = NULL; | |
8015 | int i; | |
8016 | struct imsm_map *map_dest = get_imsm_map(dev, 0); | |
8017 | int new_disks = map_dest->num_members; | |
ab724b98 AK |
8018 | int dest_layout = 0; |
8019 | int dest_chunk; | |
d1877f69 AK |
8020 | unsigned long long start; |
8021 | int data_disks = imsm_num_data_members(dev, 0); | |
687629c2 AK |
8022 | |
8023 | targets = malloc(new_disks * sizeof(int)); | |
8024 | if (!targets) | |
8025 | goto abort; | |
8026 | ||
7e45b550 AK |
8027 | for (i = 0; i < new_disks; i++) |
8028 | targets[i] = -1; | |
8029 | ||
687629c2 AK |
8030 | target_offsets = malloc(new_disks * sizeof(unsigned long long)); |
8031 | if (!target_offsets) | |
8032 | goto abort; | |
8033 | ||
d1877f69 | 8034 | start = info->reshape_progress * 512; |
687629c2 | 8035 | for (i = 0; i < new_disks; i++) { |
687629c2 AK |
8036 | target_offsets[i] = (unsigned long long) |
8037 | __le32_to_cpu(super->migr_rec->ckpt_area_pba) * 512; | |
d1877f69 AK |
8038 | /* move back copy area adderss, it will be moved forward |
8039 | * in restore_stripes() using start input variable | |
8040 | */ | |
8041 | target_offsets[i] -= start/data_disks; | |
687629c2 AK |
8042 | } |
8043 | ||
8044 | if (open_backup_targets(info, new_disks, targets)) | |
8045 | goto abort; | |
8046 | ||
68eb8bc6 | 8047 | dest_layout = imsm_level_to_layout(map_dest->raid_level); |
ab724b98 AK |
8048 | dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512; |
8049 | ||
687629c2 AK |
8050 | if (restore_stripes(targets, /* list of dest devices */ |
8051 | target_offsets, /* migration record offsets */ | |
8052 | new_disks, | |
ab724b98 AK |
8053 | dest_chunk, |
8054 | map_dest->raid_level, | |
8055 | dest_layout, | |
8056 | -1, /* source backup file descriptor */ | |
8057 | 0, /* input buf offset | |
8058 | * always 0 buf is already offseted */ | |
d1877f69 | 8059 | start, |
687629c2 AK |
8060 | length, |
8061 | buf) != 0) { | |
8062 | fprintf(stderr, Name ": Error restoring stripes\n"); | |
8063 | goto abort; | |
8064 | } | |
8065 | ||
8066 | rv = 0; | |
8067 | ||
8068 | abort: | |
8069 | if (targets) { | |
8070 | for (i = 0; i < new_disks; i++) | |
8071 | if (targets[i] >= 0) | |
8072 | close(targets[i]); | |
8073 | free(targets); | |
8074 | } | |
8075 | free(target_offsets); | |
8076 | ||
8077 | return rv; | |
8078 | } | |
8079 | ||
8080 | /******************************************************************************* | |
8081 | * Function: save_checkpoint_imsm | |
8082 | * Description: Function called for current unit status update | |
8083 | * in the migration record. It writes it to disk. | |
8084 | * Parameters: | |
8085 | * super : imsm internal array info | |
8086 | * info : general array info | |
8087 | * Returns: | |
8088 | * 0: success | |
8089 | * 1: failure | |
0228d92c AK |
8090 | * 2: failure, means no valid migration record |
8091 | * / no general migration in progress / | |
687629c2 AK |
8092 | ******************************************************************************/ |
8093 | int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state) | |
8094 | { | |
8095 | struct intel_super *super = st->sb; | |
f8b72ef5 AK |
8096 | unsigned long long blocks_per_unit; |
8097 | unsigned long long curr_migr_unit; | |
8098 | ||
2e062e82 AK |
8099 | if (load_imsm_migr_rec(super, info) != 0) { |
8100 | dprintf("imsm: ERROR: Cannot read migration record " | |
8101 | "for checkpoint save.\n"); | |
8102 | return 1; | |
8103 | } | |
8104 | ||
f8b72ef5 AK |
8105 | blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit); |
8106 | if (blocks_per_unit == 0) { | |
0228d92c AK |
8107 | dprintf("imsm: no migration in progress.\n"); |
8108 | return 2; | |
687629c2 | 8109 | } |
f8b72ef5 AK |
8110 | curr_migr_unit = info->reshape_progress / blocks_per_unit; |
8111 | /* check if array is alligned to copy area | |
8112 | * if it is not alligned, add one to current migration unit value | |
8113 | * this can happend on array reshape finish only | |
8114 | */ | |
8115 | if (info->reshape_progress % blocks_per_unit) | |
8116 | curr_migr_unit++; | |
687629c2 AK |
8117 | |
8118 | super->migr_rec->curr_migr_unit = | |
f8b72ef5 | 8119 | __cpu_to_le32(curr_migr_unit); |
687629c2 AK |
8120 | super->migr_rec->rec_status = __cpu_to_le32(state); |
8121 | super->migr_rec->dest_1st_member_lba = | |
f8b72ef5 AK |
8122 | __cpu_to_le32(curr_migr_unit * |
8123 | __le32_to_cpu(super->migr_rec->dest_depth_per_unit)); | |
687629c2 AK |
8124 | if (write_imsm_migr_rec(st) < 0) { |
8125 | dprintf("imsm: Cannot write migration record " | |
8126 | "outside backup area\n"); | |
8127 | return 1; | |
8128 | } | |
8129 | ||
8130 | return 0; | |
8131 | } | |
8132 | ||
276d77db AK |
8133 | /******************************************************************************* |
8134 | * Function: recover_backup_imsm | |
8135 | * Description: Function recovers critical data from the Migration Copy Area | |
8136 | * while assembling an array. | |
8137 | * Parameters: | |
8138 | * super : imsm internal array info | |
8139 | * info : general array info | |
8140 | * Returns: | |
8141 | * 0 : success (or there is no data to recover) | |
8142 | * 1 : fail | |
8143 | ******************************************************************************/ | |
8144 | int recover_backup_imsm(struct supertype *st, struct mdinfo *info) | |
8145 | { | |
8146 | struct intel_super *super = st->sb; | |
8147 | struct migr_record *migr_rec = super->migr_rec; | |
8148 | struct imsm_map *map_dest = NULL; | |
8149 | struct intel_dev *id = NULL; | |
8150 | unsigned long long read_offset; | |
8151 | unsigned long long write_offset; | |
8152 | unsigned unit_len; | |
8153 | int *targets = NULL; | |
8154 | int new_disks, i, err; | |
8155 | char *buf = NULL; | |
8156 | int retval = 1; | |
8157 | unsigned long curr_migr_unit = __le32_to_cpu(migr_rec->curr_migr_unit); | |
8158 | unsigned long num_migr_units = __le32_to_cpu(migr_rec->num_migr_units); | |
276d77db | 8159 | char buffer[20]; |
6c3560c0 AK |
8160 | int skipped_disks = 0; |
8161 | int max_degradation; | |
276d77db AK |
8162 | |
8163 | err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20); | |
8164 | if (err < 1) | |
8165 | return 1; | |
8166 | ||
8167 | /* recover data only during assemblation */ | |
8168 | if (strncmp(buffer, "inactive", 8) != 0) | |
8169 | return 0; | |
8170 | /* no data to recover */ | |
8171 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
8172 | return 0; | |
8173 | if (curr_migr_unit >= num_migr_units) | |
8174 | return 1; | |
8175 | ||
8176 | /* find device during reshape */ | |
8177 | for (id = super->devlist; id; id = id->next) | |
8178 | if (is_gen_migration(id->dev)) | |
8179 | break; | |
8180 | if (id == NULL) | |
8181 | return 1; | |
8182 | ||
8183 | map_dest = get_imsm_map(id->dev, 0); | |
8184 | new_disks = map_dest->num_members; | |
6c3560c0 | 8185 | max_degradation = new_disks - imsm_num_data_members(id->dev, 0); |
276d77db AK |
8186 | |
8187 | read_offset = (unsigned long long) | |
8188 | __le32_to_cpu(migr_rec->ckpt_area_pba) * 512; | |
8189 | ||
8190 | write_offset = ((unsigned long long) | |
8191 | __le32_to_cpu(migr_rec->dest_1st_member_lba) + | |
75b69ea4 | 8192 | __le32_to_cpu(map_dest->pba_of_lba0)) * 512; |
276d77db AK |
8193 | |
8194 | unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
8195 | if (posix_memalign((void **)&buf, 512, unit_len) != 0) | |
8196 | goto abort; | |
8197 | targets = malloc(new_disks * sizeof(int)); | |
8198 | if (!targets) | |
8199 | goto abort; | |
8200 | ||
8201 | open_backup_targets(info, new_disks, targets); | |
8202 | ||
8203 | for (i = 0; i < new_disks; i++) { | |
6c3560c0 AK |
8204 | if (targets[i] < 0) { |
8205 | skipped_disks++; | |
8206 | continue; | |
8207 | } | |
276d77db AK |
8208 | if (lseek64(targets[i], read_offset, SEEK_SET) < 0) { |
8209 | fprintf(stderr, | |
8210 | Name ": Cannot seek to block: %s\n", | |
8211 | strerror(errno)); | |
8212 | goto abort; | |
8213 | } | |
9ec11d1a | 8214 | if ((unsigned)read(targets[i], buf, unit_len) != unit_len) { |
276d77db AK |
8215 | fprintf(stderr, |
8216 | Name ": Cannot read copy area block: %s\n", | |
8217 | strerror(errno)); | |
8218 | goto abort; | |
8219 | } | |
8220 | if (lseek64(targets[i], write_offset, SEEK_SET) < 0) { | |
8221 | fprintf(stderr, | |
8222 | Name ": Cannot seek to block: %s\n", | |
8223 | strerror(errno)); | |
8224 | goto abort; | |
8225 | } | |
9ec11d1a | 8226 | if ((unsigned)write(targets[i], buf, unit_len) != unit_len) { |
276d77db AK |
8227 | fprintf(stderr, |
8228 | Name ": Cannot restore block: %s\n", | |
8229 | strerror(errno)); | |
8230 | goto abort; | |
8231 | } | |
8232 | } | |
8233 | ||
6c3560c0 AK |
8234 | if (skipped_disks > max_degradation) { |
8235 | fprintf(stderr, | |
8236 | Name ": Cannot restore data from backup." | |
8237 | " Too many failed disks\n"); | |
8238 | goto abort; | |
8239 | } | |
8240 | ||
befb629b AK |
8241 | if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) { |
8242 | /* ignore error == 2, this can mean end of reshape here | |
8243 | */ | |
8244 | dprintf("imsm: Cannot write checkpoint to " | |
8245 | "migration record (UNIT_SRC_NORMAL) during restart\n"); | |
8246 | } else | |
276d77db | 8247 | retval = 0; |
276d77db AK |
8248 | |
8249 | abort: | |
8250 | if (targets) { | |
8251 | for (i = 0; i < new_disks; i++) | |
8252 | if (targets[i]) | |
8253 | close(targets[i]); | |
8254 | free(targets); | |
8255 | } | |
8256 | free(buf); | |
8257 | return retval; | |
8258 | } | |
8259 | ||
2cda7640 ML |
8260 | static char disk_by_path[] = "/dev/disk/by-path/"; |
8261 | ||
8262 | static const char *imsm_get_disk_controller_domain(const char *path) | |
8263 | { | |
2cda7640 | 8264 | char disk_path[PATH_MAX]; |
96234762 LM |
8265 | char *drv=NULL; |
8266 | struct stat st; | |
2cda7640 | 8267 | |
96234762 LM |
8268 | strncpy(disk_path, disk_by_path, PATH_MAX - 1); |
8269 | strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1); | |
8270 | if (stat(disk_path, &st) == 0) { | |
8271 | struct sys_dev* hba; | |
8272 | char *path=NULL; | |
8273 | ||
8274 | path = devt_to_devpath(st.st_rdev); | |
8275 | if (path == NULL) | |
8276 | return "unknown"; | |
8277 | hba = find_disk_attached_hba(-1, path); | |
8278 | if (hba && hba->type == SYS_DEV_SAS) | |
8279 | drv = "isci"; | |
8280 | else if (hba && hba->type == SYS_DEV_SATA) | |
8281 | drv = "ahci"; | |
8282 | else | |
8283 | drv = "unknown"; | |
8284 | dprintf("path: %s hba: %s attached: %s\n", | |
8285 | path, (hba) ? hba->path : "NULL", drv); | |
8286 | free(path); | |
8287 | if (hba) | |
8288 | free_sys_dev(&hba); | |
2cda7640 | 8289 | } |
96234762 | 8290 | return drv; |
2cda7640 ML |
8291 | } |
8292 | ||
78b10e66 N |
8293 | static int imsm_find_array_minor_by_subdev(int subdev, int container, int *minor) |
8294 | { | |
8295 | char subdev_name[20]; | |
8296 | struct mdstat_ent *mdstat; | |
8297 | ||
8298 | sprintf(subdev_name, "%d", subdev); | |
8299 | mdstat = mdstat_by_subdev(subdev_name, container); | |
8300 | if (!mdstat) | |
8301 | return -1; | |
8302 | ||
8303 | *minor = mdstat->devnum; | |
8304 | free_mdstat(mdstat); | |
8305 | return 0; | |
8306 | } | |
8307 | ||
8308 | static int imsm_reshape_is_allowed_on_container(struct supertype *st, | |
8309 | struct geo_params *geo, | |
8310 | int *old_raid_disks) | |
8311 | { | |
694575e7 KW |
8312 | /* currently we only support increasing the number of devices |
8313 | * for a container. This increases the number of device for each | |
8314 | * member array. They must all be RAID0 or RAID5. | |
8315 | */ | |
78b10e66 N |
8316 | int ret_val = 0; |
8317 | struct mdinfo *info, *member; | |
8318 | int devices_that_can_grow = 0; | |
8319 | ||
8320 | dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): " | |
8321 | "st->devnum = (%i)\n", | |
8322 | st->devnum); | |
8323 | ||
8324 | if (geo->size != -1 || | |
8325 | geo->level != UnSet || | |
8326 | geo->layout != UnSet || | |
8327 | geo->chunksize != 0 || | |
8328 | geo->raid_disks == UnSet) { | |
8329 | dprintf("imsm: Container operation is allowed for " | |
8330 | "raid disks number change only.\n"); | |
8331 | return ret_val; | |
8332 | } | |
8333 | ||
8334 | info = container_content_imsm(st, NULL); | |
8335 | for (member = info; member; member = member->next) { | |
8336 | int result; | |
8337 | int minor; | |
8338 | ||
8339 | dprintf("imsm: checking device_num: %i\n", | |
8340 | member->container_member); | |
8341 | ||
d7d205bd | 8342 | if (geo->raid_disks <= member->array.raid_disks) { |
78b10e66 N |
8343 | /* we work on container for Online Capacity Expansion |
8344 | * only so raid_disks has to grow | |
8345 | */ | |
8346 | dprintf("imsm: for container operation raid disks " | |
8347 | "increase is required\n"); | |
8348 | break; | |
8349 | } | |
8350 | ||
8351 | if ((info->array.level != 0) && | |
8352 | (info->array.level != 5)) { | |
8353 | /* we cannot use this container with other raid level | |
8354 | */ | |
690aae1a | 8355 | dprintf("imsm: for container operation wrong" |
78b10e66 N |
8356 | " raid level (%i) detected\n", |
8357 | info->array.level); | |
8358 | break; | |
8359 | } else { | |
8360 | /* check for platform support | |
8361 | * for this raid level configuration | |
8362 | */ | |
8363 | struct intel_super *super = st->sb; | |
8364 | if (!is_raid_level_supported(super->orom, | |
8365 | member->array.level, | |
8366 | geo->raid_disks)) { | |
690aae1a | 8367 | dprintf("platform does not support raid%d with" |
78b10e66 N |
8368 | " %d disk%s\n", |
8369 | info->array.level, | |
8370 | geo->raid_disks, | |
8371 | geo->raid_disks > 1 ? "s" : ""); | |
8372 | break; | |
8373 | } | |
2a4a08e7 AK |
8374 | /* check if component size is aligned to chunk size |
8375 | */ | |
8376 | if (info->component_size % | |
8377 | (info->array.chunk_size/512)) { | |
8378 | dprintf("Component size is not aligned to " | |
8379 | "chunk size\n"); | |
8380 | break; | |
8381 | } | |
78b10e66 N |
8382 | } |
8383 | ||
8384 | if (*old_raid_disks && | |
8385 | info->array.raid_disks != *old_raid_disks) | |
8386 | break; | |
8387 | *old_raid_disks = info->array.raid_disks; | |
8388 | ||
8389 | /* All raid5 and raid0 volumes in container | |
8390 | * have to be ready for Online Capacity Expansion | |
8391 | * so they need to be assembled. We have already | |
8392 | * checked that no recovery etc is happening. | |
8393 | */ | |
8394 | result = imsm_find_array_minor_by_subdev(member->container_member, | |
8395 | st->container_dev, | |
8396 | &minor); | |
8397 | if (result < 0) { | |
8398 | dprintf("imsm: cannot find array\n"); | |
8399 | break; | |
8400 | } | |
8401 | devices_that_can_grow++; | |
8402 | } | |
8403 | sysfs_free(info); | |
8404 | if (!member && devices_that_can_grow) | |
8405 | ret_val = 1; | |
8406 | ||
8407 | if (ret_val) | |
8408 | dprintf("\tContainer operation allowed\n"); | |
8409 | else | |
8410 | dprintf("\tError: %i\n", ret_val); | |
8411 | ||
8412 | return ret_val; | |
8413 | } | |
8414 | ||
8415 | /* Function: get_spares_for_grow | |
8416 | * Description: Allocates memory and creates list of spare devices | |
8417 | * avaliable in container. Checks if spare drive size is acceptable. | |
8418 | * Parameters: Pointer to the supertype structure | |
8419 | * Returns: Pointer to the list of spare devices (mdinfo structure) on success, | |
8420 | * NULL if fail | |
8421 | */ | |
8422 | static struct mdinfo *get_spares_for_grow(struct supertype *st) | |
8423 | { | |
78b10e66 | 8424 | unsigned long long min_size = min_acceptable_spare_size_imsm(st); |
326727d9 | 8425 | return container_choose_spares(st, min_size, NULL, NULL, NULL, 0); |
78b10e66 N |
8426 | } |
8427 | ||
8428 | /****************************************************************************** | |
8429 | * function: imsm_create_metadata_update_for_reshape | |
8430 | * Function creates update for whole IMSM container. | |
8431 | * | |
8432 | ******************************************************************************/ | |
8433 | static int imsm_create_metadata_update_for_reshape( | |
8434 | struct supertype *st, | |
8435 | struct geo_params *geo, | |
8436 | int old_raid_disks, | |
8437 | struct imsm_update_reshape **updatep) | |
8438 | { | |
8439 | struct intel_super *super = st->sb; | |
8440 | struct imsm_super *mpb = super->anchor; | |
8441 | int update_memory_size = 0; | |
8442 | struct imsm_update_reshape *u = NULL; | |
8443 | struct mdinfo *spares = NULL; | |
8444 | int i; | |
8445 | int delta_disks = 0; | |
bbd24d86 | 8446 | struct mdinfo *dev; |
78b10e66 N |
8447 | |
8448 | dprintf("imsm_update_metadata_for_reshape(enter) raid_disks = %i\n", | |
8449 | geo->raid_disks); | |
8450 | ||
8451 | delta_disks = geo->raid_disks - old_raid_disks; | |
8452 | ||
8453 | /* size of all update data without anchor */ | |
8454 | update_memory_size = sizeof(struct imsm_update_reshape); | |
8455 | ||
8456 | /* now add space for spare disks that we need to add. */ | |
8457 | update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1); | |
8458 | ||
8459 | u = calloc(1, update_memory_size); | |
8460 | if (u == NULL) { | |
8461 | dprintf("error: " | |
8462 | "cannot get memory for imsm_update_reshape update\n"); | |
8463 | return 0; | |
8464 | } | |
8465 | u->type = update_reshape_container_disks; | |
8466 | u->old_raid_disks = old_raid_disks; | |
8467 | u->new_raid_disks = geo->raid_disks; | |
8468 | ||
8469 | /* now get spare disks list | |
8470 | */ | |
8471 | spares = get_spares_for_grow(st); | |
8472 | ||
8473 | if (spares == NULL | |
8474 | || delta_disks > spares->array.spare_disks) { | |
e14e5960 KW |
8475 | fprintf(stderr, Name ": imsm: ERROR: Cannot get spare devices " |
8476 | "for %s.\n", geo->dev_name); | |
e4c72d1d | 8477 | i = -1; |
78b10e66 N |
8478 | goto abort; |
8479 | } | |
8480 | ||
8481 | /* we have got spares | |
8482 | * update disk list in imsm_disk list table in anchor | |
8483 | */ | |
8484 | dprintf("imsm: %i spares are available.\n\n", | |
8485 | spares->array.spare_disks); | |
8486 | ||
bbd24d86 | 8487 | dev = spares->devs; |
78b10e66 | 8488 | for (i = 0; i < delta_disks; i++) { |
78b10e66 N |
8489 | struct dl *dl; |
8490 | ||
bbd24d86 AK |
8491 | if (dev == NULL) |
8492 | break; | |
78b10e66 N |
8493 | u->new_disks[i] = makedev(dev->disk.major, |
8494 | dev->disk.minor); | |
8495 | dl = get_disk_super(super, dev->disk.major, dev->disk.minor); | |
ee4beede AK |
8496 | dl->index = mpb->num_disks; |
8497 | mpb->num_disks++; | |
bbd24d86 | 8498 | dev = dev->next; |
78b10e66 | 8499 | } |
78b10e66 N |
8500 | |
8501 | abort: | |
8502 | /* free spares | |
8503 | */ | |
8504 | sysfs_free(spares); | |
8505 | ||
d677e0b8 | 8506 | dprintf("imsm: reshape update preparation :"); |
78b10e66 | 8507 | if (i == delta_disks) { |
d677e0b8 | 8508 | dprintf(" OK\n"); |
78b10e66 N |
8509 | *updatep = u; |
8510 | return update_memory_size; | |
8511 | } | |
8512 | free(u); | |
d677e0b8 | 8513 | dprintf(" Error\n"); |
78b10e66 N |
8514 | |
8515 | return 0; | |
8516 | } | |
8517 | ||
48c5303a PC |
8518 | /****************************************************************************** |
8519 | * function: imsm_create_metadata_update_for_migration() | |
8520 | * Creates update for IMSM array. | |
8521 | * | |
8522 | ******************************************************************************/ | |
8523 | static int imsm_create_metadata_update_for_migration( | |
8524 | struct supertype *st, | |
8525 | struct geo_params *geo, | |
8526 | struct imsm_update_reshape_migration **updatep) | |
8527 | { | |
8528 | struct intel_super *super = st->sb; | |
8529 | int update_memory_size = 0; | |
8530 | struct imsm_update_reshape_migration *u = NULL; | |
8531 | struct imsm_dev *dev; | |
8532 | int previous_level = -1; | |
8533 | ||
8534 | dprintf("imsm_create_metadata_update_for_migration(enter)" | |
8535 | " New Level = %i\n", geo->level); | |
8536 | ||
8537 | /* size of all update data without anchor */ | |
8538 | update_memory_size = sizeof(struct imsm_update_reshape_migration); | |
8539 | ||
8540 | u = calloc(1, update_memory_size); | |
8541 | if (u == NULL) { | |
8542 | dprintf("error: cannot get memory for " | |
8543 | "imsm_create_metadata_update_for_migration\n"); | |
8544 | return 0; | |
8545 | } | |
8546 | u->type = update_reshape_migration; | |
8547 | u->subdev = super->current_vol; | |
8548 | u->new_level = geo->level; | |
8549 | u->new_layout = geo->layout; | |
8550 | u->new_raid_disks = u->old_raid_disks = geo->raid_disks; | |
8551 | u->new_disks[0] = -1; | |
4bba0439 | 8552 | u->new_chunksize = -1; |
48c5303a PC |
8553 | |
8554 | dev = get_imsm_dev(super, u->subdev); | |
8555 | if (dev) { | |
8556 | struct imsm_map *map; | |
8557 | ||
8558 | map = get_imsm_map(dev, 0); | |
4bba0439 PC |
8559 | if (map) { |
8560 | int current_chunk_size = | |
8561 | __le16_to_cpu(map->blocks_per_strip) / 2; | |
8562 | ||
8563 | if (geo->chunksize != current_chunk_size) { | |
8564 | u->new_chunksize = geo->chunksize / 1024; | |
8565 | dprintf("imsm: " | |
8566 | "chunk size change from %i to %i\n", | |
8567 | current_chunk_size, u->new_chunksize); | |
8568 | } | |
48c5303a | 8569 | previous_level = map->raid_level; |
4bba0439 | 8570 | } |
48c5303a PC |
8571 | } |
8572 | if ((geo->level == 5) && (previous_level == 0)) { | |
8573 | struct mdinfo *spares = NULL; | |
8574 | ||
8575 | u->new_raid_disks++; | |
8576 | spares = get_spares_for_grow(st); | |
8577 | if ((spares == NULL) || (spares->array.spare_disks < 1)) { | |
8578 | free(u); | |
8579 | sysfs_free(spares); | |
8580 | update_memory_size = 0; | |
8581 | dprintf("error: cannot get spare device " | |
8582 | "for requested migration"); | |
8583 | return 0; | |
8584 | } | |
8585 | sysfs_free(spares); | |
8586 | } | |
8587 | dprintf("imsm: reshape update preparation : OK\n"); | |
8588 | *updatep = u; | |
8589 | ||
8590 | return update_memory_size; | |
8591 | } | |
8592 | ||
8dd70bce AK |
8593 | static void imsm_update_metadata_locally(struct supertype *st, |
8594 | void *buf, int len) | |
8595 | { | |
8596 | struct metadata_update mu; | |
8597 | ||
8598 | mu.buf = buf; | |
8599 | mu.len = len; | |
8600 | mu.space = NULL; | |
8601 | mu.space_list = NULL; | |
8602 | mu.next = NULL; | |
8603 | imsm_prepare_update(st, &mu); | |
8604 | imsm_process_update(st, &mu); | |
8605 | ||
8606 | while (mu.space_list) { | |
8607 | void **space = mu.space_list; | |
8608 | mu.space_list = *space; | |
8609 | free(space); | |
8610 | } | |
8611 | } | |
78b10e66 | 8612 | |
471bceb6 | 8613 | /*************************************************************************** |
694575e7 | 8614 | * Function: imsm_analyze_change |
471bceb6 KW |
8615 | * Description: Function analyze change for single volume |
8616 | * and validate if transition is supported | |
694575e7 KW |
8617 | * Parameters: Geometry parameters, supertype structure |
8618 | * Returns: Operation type code on success, -1 if fail | |
471bceb6 KW |
8619 | ****************************************************************************/ |
8620 | enum imsm_reshape_type imsm_analyze_change(struct supertype *st, | |
8621 | struct geo_params *geo) | |
694575e7 | 8622 | { |
471bceb6 KW |
8623 | struct mdinfo info; |
8624 | int change = -1; | |
8625 | int check_devs = 0; | |
c21e737b | 8626 | int chunk; |
471bceb6 KW |
8627 | |
8628 | getinfo_super_imsm_volume(st, &info, NULL); | |
471bceb6 KW |
8629 | if ((geo->level != info.array.level) && |
8630 | (geo->level >= 0) && | |
8631 | (geo->level != UnSet)) { | |
8632 | switch (info.array.level) { | |
8633 | case 0: | |
8634 | if (geo->level == 5) { | |
b5347799 | 8635 | change = CH_MIGRATION; |
e13ce846 AK |
8636 | if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) { |
8637 | fprintf(stderr, | |
8638 | Name " Error. Requested Layout " | |
8639 | "not supported (left-asymmetric layout " | |
8640 | "is supported only)!\n"); | |
8641 | change = -1; | |
8642 | goto analyse_change_exit; | |
8643 | } | |
471bceb6 KW |
8644 | check_devs = 1; |
8645 | } | |
8646 | if (geo->level == 10) { | |
8647 | change = CH_TAKEOVER; | |
8648 | check_devs = 1; | |
8649 | } | |
dfe77a9e KW |
8650 | break; |
8651 | case 1: | |
8652 | if (geo->level == 0) { | |
8653 | change = CH_TAKEOVER; | |
8654 | check_devs = 1; | |
8655 | } | |
471bceb6 | 8656 | break; |
471bceb6 KW |
8657 | case 10: |
8658 | if (geo->level == 0) { | |
8659 | change = CH_TAKEOVER; | |
8660 | check_devs = 1; | |
8661 | } | |
8662 | break; | |
8663 | } | |
8664 | if (change == -1) { | |
8665 | fprintf(stderr, | |
8666 | Name " Error. Level Migration from %d to %d " | |
8667 | "not supported!\n", | |
8668 | info.array.level, geo->level); | |
8669 | goto analyse_change_exit; | |
8670 | } | |
8671 | } else | |
8672 | geo->level = info.array.level; | |
8673 | ||
8674 | if ((geo->layout != info.array.layout) | |
8675 | && ((geo->layout != UnSet) && (geo->layout != -1))) { | |
b5347799 | 8676 | change = CH_MIGRATION; |
471bceb6 KW |
8677 | if ((info.array.layout == 0) |
8678 | && (info.array.level == 5) | |
8679 | && (geo->layout == 5)) { | |
8680 | /* reshape 5 -> 4 */ | |
8681 | } else if ((info.array.layout == 5) | |
8682 | && (info.array.level == 5) | |
8683 | && (geo->layout == 0)) { | |
8684 | /* reshape 4 -> 5 */ | |
8685 | geo->layout = 0; | |
8686 | geo->level = 5; | |
8687 | } else { | |
8688 | fprintf(stderr, | |
8689 | Name " Error. Layout Migration from %d to %d " | |
8690 | "not supported!\n", | |
8691 | info.array.layout, geo->layout); | |
8692 | change = -1; | |
8693 | goto analyse_change_exit; | |
8694 | } | |
8695 | } else | |
8696 | geo->layout = info.array.layout; | |
8697 | ||
8698 | if ((geo->chunksize > 0) && (geo->chunksize != UnSet) | |
8699 | && (geo->chunksize != info.array.chunk_size)) | |
b5347799 | 8700 | change = CH_MIGRATION; |
471bceb6 KW |
8701 | else |
8702 | geo->chunksize = info.array.chunk_size; | |
8703 | ||
c21e737b | 8704 | chunk = geo->chunksize / 1024; |
471bceb6 KW |
8705 | if (!validate_geometry_imsm(st, |
8706 | geo->level, | |
8707 | geo->layout, | |
8708 | geo->raid_disks, | |
c21e737b | 8709 | &chunk, |
471bceb6 KW |
8710 | geo->size, |
8711 | 0, 0, 1)) | |
8712 | change = -1; | |
8713 | ||
8714 | if (check_devs) { | |
8715 | struct intel_super *super = st->sb; | |
8716 | struct imsm_super *mpb = super->anchor; | |
8717 | ||
8718 | if (mpb->num_raid_devs > 1) { | |
8719 | fprintf(stderr, | |
8720 | Name " Error. Cannot perform operation on %s" | |
8721 | "- for this operation it MUST be single " | |
8722 | "array in container\n", | |
8723 | geo->dev_name); | |
8724 | change = -1; | |
8725 | } | |
8726 | } | |
8727 | ||
8728 | analyse_change_exit: | |
8729 | ||
8730 | return change; | |
694575e7 KW |
8731 | } |
8732 | ||
bb025c2f KW |
8733 | int imsm_takeover(struct supertype *st, struct geo_params *geo) |
8734 | { | |
8735 | struct intel_super *super = st->sb; | |
8736 | struct imsm_update_takeover *u; | |
8737 | ||
8738 | u = malloc(sizeof(struct imsm_update_takeover)); | |
8739 | if (u == NULL) | |
8740 | return 1; | |
8741 | ||
8742 | u->type = update_takeover; | |
8743 | u->subarray = super->current_vol; | |
8744 | ||
8745 | /* 10->0 transition */ | |
8746 | if (geo->level == 0) | |
8747 | u->direction = R10_TO_R0; | |
8748 | ||
0529c688 KW |
8749 | /* 0->10 transition */ |
8750 | if (geo->level == 10) | |
8751 | u->direction = R0_TO_R10; | |
8752 | ||
bb025c2f KW |
8753 | /* update metadata locally */ |
8754 | imsm_update_metadata_locally(st, u, | |
8755 | sizeof(struct imsm_update_takeover)); | |
8756 | /* and possibly remotely */ | |
8757 | if (st->update_tail) | |
8758 | append_metadata_update(st, u, | |
8759 | sizeof(struct imsm_update_takeover)); | |
8760 | else | |
8761 | free(u); | |
8762 | ||
8763 | return 0; | |
8764 | } | |
8765 | ||
78b10e66 N |
8766 | static int imsm_reshape_super(struct supertype *st, long long size, int level, |
8767 | int layout, int chunksize, int raid_disks, | |
41784c88 AK |
8768 | int delta_disks, char *backup, char *dev, |
8769 | int verbose) | |
78b10e66 | 8770 | { |
78b10e66 N |
8771 | int ret_val = 1; |
8772 | struct geo_params geo; | |
8773 | ||
8774 | dprintf("imsm: reshape_super called.\n"); | |
8775 | ||
71204a50 | 8776 | memset(&geo, 0, sizeof(struct geo_params)); |
78b10e66 N |
8777 | |
8778 | geo.dev_name = dev; | |
694575e7 | 8779 | geo.dev_id = st->devnum; |
78b10e66 N |
8780 | geo.size = size; |
8781 | geo.level = level; | |
8782 | geo.layout = layout; | |
8783 | geo.chunksize = chunksize; | |
8784 | geo.raid_disks = raid_disks; | |
41784c88 AK |
8785 | if (delta_disks != UnSet) |
8786 | geo.raid_disks += delta_disks; | |
78b10e66 N |
8787 | |
8788 | dprintf("\tfor level : %i\n", geo.level); | |
8789 | dprintf("\tfor raid_disks : %i\n", geo.raid_disks); | |
8790 | ||
8791 | if (experimental() == 0) | |
8792 | return ret_val; | |
8793 | ||
78b10e66 | 8794 | if (st->container_dev == st->devnum) { |
694575e7 KW |
8795 | /* On container level we can only increase number of devices. */ |
8796 | dprintf("imsm: info: Container operation\n"); | |
78b10e66 | 8797 | int old_raid_disks = 0; |
6dc0be30 | 8798 | |
78b10e66 N |
8799 | if (imsm_reshape_is_allowed_on_container( |
8800 | st, &geo, &old_raid_disks)) { | |
8801 | struct imsm_update_reshape *u = NULL; | |
8802 | int len; | |
8803 | ||
8804 | len = imsm_create_metadata_update_for_reshape( | |
8805 | st, &geo, old_raid_disks, &u); | |
8806 | ||
ed08d51c AK |
8807 | if (len <= 0) { |
8808 | dprintf("imsm: Cannot prepare update\n"); | |
8809 | goto exit_imsm_reshape_super; | |
8810 | } | |
8811 | ||
8dd70bce AK |
8812 | ret_val = 0; |
8813 | /* update metadata locally */ | |
8814 | imsm_update_metadata_locally(st, u, len); | |
8815 | /* and possibly remotely */ | |
8816 | if (st->update_tail) | |
8817 | append_metadata_update(st, u, len); | |
8818 | else | |
ed08d51c | 8819 | free(u); |
8dd70bce | 8820 | |
694575e7 | 8821 | } else { |
e7ff7e40 AK |
8822 | fprintf(stderr, Name ": (imsm) Operation " |
8823 | "is not allowed on this container\n"); | |
694575e7 KW |
8824 | } |
8825 | } else { | |
8826 | /* On volume level we support following operations | |
471bceb6 KW |
8827 | * - takeover: raid10 -> raid0; raid0 -> raid10 |
8828 | * - chunk size migration | |
8829 | * - migration: raid5 -> raid0; raid0 -> raid5 | |
8830 | */ | |
8831 | struct intel_super *super = st->sb; | |
8832 | struct intel_dev *dev = super->devlist; | |
8833 | int change, devnum; | |
694575e7 | 8834 | dprintf("imsm: info: Volume operation\n"); |
471bceb6 KW |
8835 | /* find requested device */ |
8836 | while (dev) { | |
19986c72 MB |
8837 | if (imsm_find_array_minor_by_subdev( |
8838 | dev->index, st->container_dev, &devnum) == 0 | |
8839 | && devnum == geo.dev_id) | |
471bceb6 KW |
8840 | break; |
8841 | dev = dev->next; | |
8842 | } | |
8843 | if (dev == NULL) { | |
8844 | fprintf(stderr, Name " Cannot find %s (%i) subarray\n", | |
8845 | geo.dev_name, geo.dev_id); | |
8846 | goto exit_imsm_reshape_super; | |
8847 | } | |
8848 | super->current_vol = dev->index; | |
694575e7 KW |
8849 | change = imsm_analyze_change(st, &geo); |
8850 | switch (change) { | |
471bceb6 | 8851 | case CH_TAKEOVER: |
bb025c2f | 8852 | ret_val = imsm_takeover(st, &geo); |
694575e7 | 8853 | break; |
48c5303a PC |
8854 | case CH_MIGRATION: { |
8855 | struct imsm_update_reshape_migration *u = NULL; | |
8856 | int len = | |
8857 | imsm_create_metadata_update_for_migration( | |
8858 | st, &geo, &u); | |
8859 | if (len < 1) { | |
8860 | dprintf("imsm: " | |
8861 | "Cannot prepare update\n"); | |
8862 | break; | |
8863 | } | |
471bceb6 | 8864 | ret_val = 0; |
48c5303a PC |
8865 | /* update metadata locally */ |
8866 | imsm_update_metadata_locally(st, u, len); | |
8867 | /* and possibly remotely */ | |
8868 | if (st->update_tail) | |
8869 | append_metadata_update(st, u, len); | |
8870 | else | |
8871 | free(u); | |
8872 | } | |
8873 | break; | |
471bceb6 KW |
8874 | default: |
8875 | ret_val = 1; | |
694575e7 | 8876 | } |
694575e7 | 8877 | } |
78b10e66 | 8878 | |
ed08d51c | 8879 | exit_imsm_reshape_super: |
78b10e66 N |
8880 | dprintf("imsm: reshape_super Exit code = %i\n", ret_val); |
8881 | return ret_val; | |
8882 | } | |
2cda7640 | 8883 | |
eee67a47 AK |
8884 | /******************************************************************************* |
8885 | * Function: wait_for_reshape_imsm | |
8886 | * Description: Function writes new sync_max value and waits until | |
8887 | * reshape process reach new position | |
8888 | * Parameters: | |
8889 | * sra : general array info | |
eee67a47 AK |
8890 | * ndata : number of disks in new array's layout |
8891 | * Returns: | |
8892 | * 0 : success, | |
8893 | * 1 : there is no reshape in progress, | |
8894 | * -1 : fail | |
8895 | ******************************************************************************/ | |
ae9f01f8 | 8896 | int wait_for_reshape_imsm(struct mdinfo *sra, int ndata) |
eee67a47 AK |
8897 | { |
8898 | int fd = sysfs_get_fd(sra, NULL, "reshape_position"); | |
8899 | unsigned long long completed; | |
ae9f01f8 AK |
8900 | /* to_complete : new sync_max position */ |
8901 | unsigned long long to_complete = sra->reshape_progress; | |
8902 | unsigned long long position_to_set = to_complete / ndata; | |
eee67a47 | 8903 | |
ae9f01f8 AK |
8904 | if (fd < 0) { |
8905 | dprintf("imsm: wait_for_reshape_imsm() " | |
8906 | "cannot open reshape_position\n"); | |
eee67a47 | 8907 | return 1; |
ae9f01f8 | 8908 | } |
eee67a47 | 8909 | |
ae9f01f8 AK |
8910 | if (sysfs_fd_get_ll(fd, &completed) < 0) { |
8911 | dprintf("imsm: wait_for_reshape_imsm() " | |
8912 | "cannot read reshape_position (no reshape in progres)\n"); | |
8913 | close(fd); | |
8914 | return 0; | |
8915 | } | |
eee67a47 | 8916 | |
ae9f01f8 AK |
8917 | if (completed > to_complete) { |
8918 | dprintf("imsm: wait_for_reshape_imsm() " | |
8919 | "wrong next position to set %llu (%llu)\n", | |
8920 | to_complete, completed); | |
8921 | close(fd); | |
8922 | return -1; | |
8923 | } | |
8924 | dprintf("Position set: %llu\n", position_to_set); | |
8925 | if (sysfs_set_num(sra, NULL, "sync_max", | |
8926 | position_to_set) != 0) { | |
8927 | dprintf("imsm: wait_for_reshape_imsm() " | |
8928 | "cannot set reshape position to %llu\n", | |
8929 | position_to_set); | |
8930 | close(fd); | |
8931 | return -1; | |
eee67a47 AK |
8932 | } |
8933 | ||
eee67a47 AK |
8934 | do { |
8935 | char action[20]; | |
8936 | fd_set rfds; | |
8937 | FD_ZERO(&rfds); | |
8938 | FD_SET(fd, &rfds); | |
a47e44fb AK |
8939 | select(fd+1, &rfds, NULL, NULL, NULL); |
8940 | if (sysfs_get_str(sra, NULL, "sync_action", | |
8941 | action, 20) > 0 && | |
8942 | strncmp(action, "reshape", 7) != 0) | |
8943 | break; | |
eee67a47 | 8944 | if (sysfs_fd_get_ll(fd, &completed) < 0) { |
ae9f01f8 AK |
8945 | dprintf("imsm: wait_for_reshape_imsm() " |
8946 | "cannot read reshape_position (in loop)\n"); | |
eee67a47 AK |
8947 | close(fd); |
8948 | return 1; | |
8949 | } | |
eee67a47 AK |
8950 | } while (completed < to_complete); |
8951 | close(fd); | |
8952 | return 0; | |
8953 | ||
8954 | } | |
8955 | ||
b915c95f AK |
8956 | /******************************************************************************* |
8957 | * Function: check_degradation_change | |
8958 | * Description: Check that array hasn't become failed. | |
8959 | * Parameters: | |
8960 | * info : for sysfs access | |
8961 | * sources : source disks descriptors | |
8962 | * degraded: previous degradation level | |
8963 | * Returns: | |
8964 | * degradation level | |
8965 | ******************************************************************************/ | |
8966 | int check_degradation_change(struct mdinfo *info, | |
8967 | int *sources, | |
8968 | int degraded) | |
8969 | { | |
8970 | unsigned long long new_degraded; | |
8971 | sysfs_get_ll(info, NULL, "degraded", &new_degraded); | |
8972 | if (new_degraded != (unsigned long long)degraded) { | |
8973 | /* check each device to ensure it is still working */ | |
8974 | struct mdinfo *sd; | |
8975 | new_degraded = 0; | |
8976 | for (sd = info->devs ; sd ; sd = sd->next) { | |
8977 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) | |
8978 | continue; | |
8979 | if (sd->disk.state & (1<<MD_DISK_SYNC)) { | |
8980 | char sbuf[20]; | |
8981 | if (sysfs_get_str(info, | |
8982 | sd, "state", sbuf, 20) < 0 || | |
8983 | strstr(sbuf, "faulty") || | |
8984 | strstr(sbuf, "in_sync") == NULL) { | |
8985 | /* this device is dead */ | |
8986 | sd->disk.state = (1<<MD_DISK_FAULTY); | |
8987 | if (sd->disk.raid_disk >= 0 && | |
8988 | sources[sd->disk.raid_disk] >= 0) { | |
8989 | close(sources[ | |
8990 | sd->disk.raid_disk]); | |
8991 | sources[sd->disk.raid_disk] = | |
8992 | -1; | |
8993 | } | |
8994 | new_degraded++; | |
8995 | } | |
8996 | } | |
8997 | } | |
8998 | } | |
8999 | ||
9000 | return new_degraded; | |
9001 | } | |
9002 | ||
10f22854 AK |
9003 | /******************************************************************************* |
9004 | * Function: imsm_manage_reshape | |
9005 | * Description: Function finds array under reshape and it manages reshape | |
9006 | * process. It creates stripes backups (if required) and sets | |
9007 | * checheckpoits. | |
9008 | * Parameters: | |
9009 | * afd : Backup handle (nattive) - not used | |
9010 | * sra : general array info | |
9011 | * reshape : reshape parameters - not used | |
9012 | * st : supertype structure | |
9013 | * blocks : size of critical section [blocks] | |
9014 | * fds : table of source device descriptor | |
9015 | * offsets : start of array (offest per devices) | |
9016 | * dests : not used | |
9017 | * destfd : table of destination device descriptor | |
9018 | * destoffsets : table of destination offsets (per device) | |
9019 | * Returns: | |
9020 | * 1 : success, reshape is done | |
9021 | * 0 : fail | |
9022 | ******************************************************************************/ | |
999b4972 N |
9023 | static int imsm_manage_reshape( |
9024 | int afd, struct mdinfo *sra, struct reshape *reshape, | |
10f22854 | 9025 | struct supertype *st, unsigned long backup_blocks, |
999b4972 N |
9026 | int *fds, unsigned long long *offsets, |
9027 | int dests, int *destfd, unsigned long long *destoffsets) | |
9028 | { | |
10f22854 AK |
9029 | int ret_val = 0; |
9030 | struct intel_super *super = st->sb; | |
9031 | struct intel_dev *dv = NULL; | |
9032 | struct imsm_dev *dev = NULL; | |
a6b6d984 | 9033 | struct imsm_map *map_src; |
10f22854 AK |
9034 | int migr_vol_qan = 0; |
9035 | int ndata, odata; /* [bytes] */ | |
9036 | int chunk; /* [bytes] */ | |
9037 | struct migr_record *migr_rec; | |
9038 | char *buf = NULL; | |
9039 | unsigned int buf_size; /* [bytes] */ | |
9040 | unsigned long long max_position; /* array size [bytes] */ | |
9041 | unsigned long long next_step; /* [blocks]/[bytes] */ | |
9042 | unsigned long long old_data_stripe_length; | |
10f22854 AK |
9043 | unsigned long long start_src; /* [bytes] */ |
9044 | unsigned long long start; /* [bytes] */ | |
9045 | unsigned long long start_buf_shift; /* [bytes] */ | |
b915c95f | 9046 | int degraded = 0; |
ab724b98 | 9047 | int source_layout = 0; |
10f22854 | 9048 | |
1ab242d8 | 9049 | if (!fds || !offsets || !sra) |
10f22854 AK |
9050 | goto abort; |
9051 | ||
9052 | /* Find volume during the reshape */ | |
9053 | for (dv = super->devlist; dv; dv = dv->next) { | |
9054 | if (dv->dev->vol.migr_type == MIGR_GEN_MIGR | |
9055 | && dv->dev->vol.migr_state == 1) { | |
9056 | dev = dv->dev; | |
9057 | migr_vol_qan++; | |
9058 | } | |
9059 | } | |
9060 | /* Only one volume can migrate at the same time */ | |
9061 | if (migr_vol_qan != 1) { | |
9062 | fprintf(stderr, Name " : %s", migr_vol_qan ? | |
9063 | "Number of migrating volumes greater than 1\n" : | |
9064 | "There is no volume during migrationg\n"); | |
9065 | goto abort; | |
9066 | } | |
9067 | ||
9068 | map_src = get_imsm_map(dev, 1); | |
9069 | if (map_src == NULL) | |
9070 | goto abort; | |
10f22854 AK |
9071 | |
9072 | ndata = imsm_num_data_members(dev, 0); | |
9073 | odata = imsm_num_data_members(dev, 1); | |
9074 | ||
7b1ab482 | 9075 | chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512; |
10f22854 AK |
9076 | old_data_stripe_length = odata * chunk; |
9077 | ||
9078 | migr_rec = super->migr_rec; | |
9079 | ||
10f22854 AK |
9080 | /* initialize migration record for start condition */ |
9081 | if (sra->reshape_progress == 0) | |
9082 | init_migr_record_imsm(st, dev, sra); | |
b2c59438 AK |
9083 | else { |
9084 | if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) { | |
9085 | dprintf("imsm: cannot restart migration when data " | |
9086 | "are present in copy area.\n"); | |
9087 | goto abort; | |
9088 | } | |
9089 | } | |
10f22854 AK |
9090 | |
9091 | /* size for data */ | |
9092 | buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512; | |
9093 | /* extend buffer size for parity disk */ | |
9094 | buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
9095 | /* add space for stripe aligment */ | |
9096 | buf_size += old_data_stripe_length; | |
9097 | if (posix_memalign((void **)&buf, 4096, buf_size)) { | |
9098 | dprintf("imsm: Cannot allocate checpoint buffer\n"); | |
9099 | goto abort; | |
9100 | } | |
9101 | ||
3ef4403c | 9102 | max_position = sra->component_size * ndata; |
68eb8bc6 | 9103 | source_layout = imsm_level_to_layout(map_src->raid_level); |
10f22854 AK |
9104 | |
9105 | while (__le32_to_cpu(migr_rec->curr_migr_unit) < | |
9106 | __le32_to_cpu(migr_rec->num_migr_units)) { | |
9107 | /* current reshape position [blocks] */ | |
9108 | unsigned long long current_position = | |
9109 | __le32_to_cpu(migr_rec->blocks_per_unit) | |
9110 | * __le32_to_cpu(migr_rec->curr_migr_unit); | |
9111 | unsigned long long border; | |
9112 | ||
b915c95f AK |
9113 | /* Check that array hasn't become failed. |
9114 | */ | |
9115 | degraded = check_degradation_change(sra, fds, degraded); | |
9116 | if (degraded > 1) { | |
9117 | dprintf("imsm: Abort reshape due to degradation" | |
9118 | " level (%i)\n", degraded); | |
9119 | goto abort; | |
9120 | } | |
9121 | ||
10f22854 AK |
9122 | next_step = __le32_to_cpu(migr_rec->blocks_per_unit); |
9123 | ||
9124 | if ((current_position + next_step) > max_position) | |
9125 | next_step = max_position - current_position; | |
9126 | ||
92144abf | 9127 | start = current_position * 512; |
10f22854 AK |
9128 | |
9129 | /* allign reading start to old geometry */ | |
9130 | start_buf_shift = start % old_data_stripe_length; | |
9131 | start_src = start - start_buf_shift; | |
9132 | ||
9133 | border = (start_src / odata) - (start / ndata); | |
9134 | border /= 512; | |
9135 | if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) { | |
9136 | /* save critical stripes to buf | |
9137 | * start - start address of current unit | |
9138 | * to backup [bytes] | |
9139 | * start_src - start address of current unit | |
9140 | * to backup alligned to source array | |
9141 | * [bytes] | |
9142 | */ | |
9143 | unsigned long long next_step_filler = 0; | |
9144 | unsigned long long copy_length = next_step * 512; | |
9145 | ||
9146 | /* allign copy area length to stripe in old geometry */ | |
9147 | next_step_filler = ((copy_length + start_buf_shift) | |
9148 | % old_data_stripe_length); | |
9149 | if (next_step_filler) | |
9150 | next_step_filler = (old_data_stripe_length | |
9151 | - next_step_filler); | |
9152 | dprintf("save_stripes() parameters: start = %llu," | |
9153 | "\tstart_src = %llu,\tnext_step*512 = %llu," | |
9154 | "\tstart_in_buf_shift = %llu," | |
9155 | "\tnext_step_filler = %llu\n", | |
9156 | start, start_src, copy_length, | |
9157 | start_buf_shift, next_step_filler); | |
9158 | ||
9159 | if (save_stripes(fds, offsets, map_src->num_members, | |
ab724b98 AK |
9160 | chunk, map_src->raid_level, |
9161 | source_layout, 0, NULL, start_src, | |
10f22854 AK |
9162 | copy_length + |
9163 | next_step_filler + start_buf_shift, | |
9164 | buf)) { | |
9165 | dprintf("imsm: Cannot save stripes" | |
9166 | " to buffer\n"); | |
9167 | goto abort; | |
9168 | } | |
9169 | /* Convert data to destination format and store it | |
9170 | * in backup general migration area | |
9171 | */ | |
9172 | if (save_backup_imsm(st, dev, sra, | |
aea93171 | 9173 | buf + start_buf_shift, copy_length)) { |
10f22854 AK |
9174 | dprintf("imsm: Cannot save stripes to " |
9175 | "target devices\n"); | |
9176 | goto abort; | |
9177 | } | |
9178 | if (save_checkpoint_imsm(st, sra, | |
9179 | UNIT_SRC_IN_CP_AREA)) { | |
9180 | dprintf("imsm: Cannot write checkpoint to " | |
9181 | "migration record (UNIT_SRC_IN_CP_AREA)\n"); | |
9182 | goto abort; | |
9183 | } | |
8016a6d4 AK |
9184 | } else { |
9185 | /* set next step to use whole border area */ | |
9186 | border /= next_step; | |
9187 | if (border > 1) | |
9188 | next_step *= border; | |
10f22854 AK |
9189 | } |
9190 | /* When data backed up, checkpoint stored, | |
9191 | * kick the kernel to reshape unit of data | |
9192 | */ | |
9193 | next_step = next_step + sra->reshape_progress; | |
8016a6d4 AK |
9194 | /* limit next step to array max position */ |
9195 | if (next_step > max_position) | |
9196 | next_step = max_position; | |
10f22854 AK |
9197 | sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress); |
9198 | sysfs_set_num(sra, NULL, "suspend_hi", next_step); | |
ae9f01f8 | 9199 | sra->reshape_progress = next_step; |
10f22854 AK |
9200 | |
9201 | /* wait until reshape finish */ | |
ae9f01f8 | 9202 | if (wait_for_reshape_imsm(sra, ndata) < 0) { |
c47b0ff6 AK |
9203 | dprintf("wait_for_reshape_imsm returned error!\n"); |
9204 | goto abort; | |
9205 | } | |
10f22854 | 9206 | |
0228d92c AK |
9207 | if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) { |
9208 | /* ignore error == 2, this can mean end of reshape here | |
9209 | */ | |
10f22854 AK |
9210 | dprintf("imsm: Cannot write checkpoint to " |
9211 | "migration record (UNIT_SRC_NORMAL)\n"); | |
9212 | goto abort; | |
9213 | } | |
9214 | ||
9215 | } | |
9216 | ||
9217 | /* return '1' if done */ | |
9218 | ret_val = 1; | |
9219 | abort: | |
9220 | free(buf); | |
9221 | abort_reshape(sra); | |
9222 | ||
9223 | return ret_val; | |
999b4972 | 9224 | } |
71204a50 | 9225 | #endif /* MDASSEMBLE */ |
999b4972 | 9226 | |
cdddbdbc DW |
9227 | struct superswitch super_imsm = { |
9228 | #ifndef MDASSEMBLE | |
9229 | .examine_super = examine_super_imsm, | |
9230 | .brief_examine_super = brief_examine_super_imsm, | |
4737ae25 | 9231 | .brief_examine_subarrays = brief_examine_subarrays_imsm, |
9d84c8ea | 9232 | .export_examine_super = export_examine_super_imsm, |
cdddbdbc DW |
9233 | .detail_super = detail_super_imsm, |
9234 | .brief_detail_super = brief_detail_super_imsm, | |
bf5a934a | 9235 | .write_init_super = write_init_super_imsm, |
0e600426 N |
9236 | .validate_geometry = validate_geometry_imsm, |
9237 | .add_to_super = add_to_super_imsm, | |
1a64be56 | 9238 | .remove_from_super = remove_from_super_imsm, |
d665cc31 | 9239 | .detail_platform = detail_platform_imsm, |
33414a01 | 9240 | .kill_subarray = kill_subarray_imsm, |
aa534678 | 9241 | .update_subarray = update_subarray_imsm, |
2b959fbf | 9242 | .load_container = load_container_imsm, |
71204a50 N |
9243 | .default_geometry = default_geometry_imsm, |
9244 | .get_disk_controller_domain = imsm_get_disk_controller_domain, | |
9245 | .reshape_super = imsm_reshape_super, | |
9246 | .manage_reshape = imsm_manage_reshape, | |
9e2d750d | 9247 | .recover_backup = recover_backup_imsm, |
cdddbdbc DW |
9248 | #endif |
9249 | .match_home = match_home_imsm, | |
9250 | .uuid_from_super= uuid_from_super_imsm, | |
9251 | .getinfo_super = getinfo_super_imsm, | |
5c4cd5da | 9252 | .getinfo_super_disks = getinfo_super_disks_imsm, |
cdddbdbc DW |
9253 | .update_super = update_super_imsm, |
9254 | ||
9255 | .avail_size = avail_size_imsm, | |
80e7f8c3 | 9256 | .min_acceptable_spare_size = min_acceptable_spare_size_imsm, |
cdddbdbc DW |
9257 | |
9258 | .compare_super = compare_super_imsm, | |
9259 | ||
9260 | .load_super = load_super_imsm, | |
bf5a934a | 9261 | .init_super = init_super_imsm, |
e683ca88 | 9262 | .store_super = store_super_imsm, |
cdddbdbc DW |
9263 | .free_super = free_super_imsm, |
9264 | .match_metadata_desc = match_metadata_desc_imsm, | |
bf5a934a | 9265 | .container_content = container_content_imsm, |
cdddbdbc | 9266 | |
276d77db | 9267 | |
cdddbdbc | 9268 | .external = 1, |
4cce4069 | 9269 | .name = "imsm", |
845dea95 | 9270 | |
0e600426 | 9271 | #ifndef MDASSEMBLE |
845dea95 NB |
9272 | /* for mdmon */ |
9273 | .open_new = imsm_open_new, | |
ed9d66aa | 9274 | .set_array_state= imsm_set_array_state, |
845dea95 NB |
9275 | .set_disk = imsm_set_disk, |
9276 | .sync_metadata = imsm_sync_metadata, | |
88758e9d | 9277 | .activate_spare = imsm_activate_spare, |
e8319a19 | 9278 | .process_update = imsm_process_update, |
8273f55e | 9279 | .prepare_update = imsm_prepare_update, |
0e600426 | 9280 | #endif /* MDASSEMBLE */ |
cdddbdbc | 9281 | }; |