]>
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 | \ | |
bbab0940 TM |
84 | MPB_ATTRIB_EXP_STRIPE_SIZE | \ |
85 | MPB_ATTRIB_BBM) | |
418f9b36 N |
86 | |
87 | /* Define attributes that are unused but not harmful */ | |
88 | #define MPB_ATTRIB_IGNORED (MPB_ATTRIB_NEVER_USE) | |
fe7ed8cb | 89 | |
8e59f3d8 | 90 | #define MPB_SECTOR_CNT 2210 |
c2c087e6 | 91 | #define IMSM_RESERVED_SECTORS 4096 |
b81221b7 | 92 | #define NUM_BLOCKS_DIRTY_STRIPE_REGION 2056 |
979d38be | 93 | #define SECT_PER_MB_SHIFT 11 |
f36a9ecd | 94 | #define MAX_SECTOR_SIZE 4096 |
cdddbdbc DW |
95 | |
96 | /* Disk configuration info. */ | |
97 | #define IMSM_MAX_DEVICES 255 | |
98 | struct imsm_disk { | |
99 | __u8 serial[MAX_RAID_SERIAL_LEN];/* 0xD8 - 0xE7 ascii serial number */ | |
5551b113 | 100 | __u32 total_blocks_lo; /* 0xE8 - 0xEB total blocks lo */ |
cdddbdbc | 101 | __u32 scsi_id; /* 0xEC - 0xEF scsi ID */ |
f2f27e63 DW |
102 | #define SPARE_DISK __cpu_to_le32(0x01) /* Spare */ |
103 | #define CONFIGURED_DISK __cpu_to_le32(0x02) /* Member of some RaidDev */ | |
104 | #define FAILED_DISK __cpu_to_le32(0x04) /* Permanent failure */ | |
cdddbdbc | 105 | __u32 status; /* 0xF0 - 0xF3 */ |
1011e834 | 106 | __u32 owner_cfg_num; /* which config 0,1,2... owns this disk */ |
5551b113 CA |
107 | __u32 total_blocks_hi; /* 0xF4 - 0xF5 total blocks hi */ |
108 | #define IMSM_DISK_FILLERS 3 | |
109 | __u32 filler[IMSM_DISK_FILLERS]; /* 0xF5 - 0x107 MPB_DISK_FILLERS for future expansion */ | |
cdddbdbc DW |
110 | }; |
111 | ||
3b451610 AK |
112 | /* map selector for map managment |
113 | */ | |
238c0a71 AK |
114 | #define MAP_0 0 |
115 | #define MAP_1 1 | |
116 | #define MAP_X -1 | |
3b451610 | 117 | |
cdddbdbc DW |
118 | /* RAID map configuration infos. */ |
119 | struct imsm_map { | |
5551b113 CA |
120 | __u32 pba_of_lba0_lo; /* start address of partition */ |
121 | __u32 blocks_per_member_lo;/* blocks per member */ | |
122 | __u32 num_data_stripes_lo; /* number of data stripes */ | |
cdddbdbc DW |
123 | __u16 blocks_per_strip; |
124 | __u8 map_state; /* Normal, Uninitialized, Degraded, Failed */ | |
125 | #define IMSM_T_STATE_NORMAL 0 | |
126 | #define IMSM_T_STATE_UNINITIALIZED 1 | |
e3bba0e0 DW |
127 | #define IMSM_T_STATE_DEGRADED 2 |
128 | #define IMSM_T_STATE_FAILED 3 | |
cdddbdbc DW |
129 | __u8 raid_level; |
130 | #define IMSM_T_RAID0 0 | |
131 | #define IMSM_T_RAID1 1 | |
132 | #define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */ | |
133 | __u8 num_members; /* number of member disks */ | |
fe7ed8cb DW |
134 | __u8 num_domains; /* number of parity domains */ |
135 | __u8 failed_disk_num; /* valid only when state is degraded */ | |
252d23c0 | 136 | __u8 ddf; |
5551b113 CA |
137 | __u32 pba_of_lba0_hi; |
138 | __u32 blocks_per_member_hi; | |
139 | __u32 num_data_stripes_hi; | |
140 | __u32 filler[4]; /* expansion area */ | |
7eef0453 | 141 | #define IMSM_ORD_REBUILD (1 << 24) |
cdddbdbc | 142 | __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members], |
7eef0453 DW |
143 | * top byte contains some flags |
144 | */ | |
cdddbdbc DW |
145 | } __attribute__ ((packed)); |
146 | ||
147 | struct imsm_vol { | |
f8f603f1 | 148 | __u32 curr_migr_unit; |
fe7ed8cb | 149 | __u32 checkpoint_id; /* id to access curr_migr_unit */ |
cdddbdbc | 150 | __u8 migr_state; /* Normal or Migrating */ |
e3bba0e0 DW |
151 | #define MIGR_INIT 0 |
152 | #define MIGR_REBUILD 1 | |
153 | #define MIGR_VERIFY 2 /* analagous to echo check > sync_action */ | |
154 | #define MIGR_GEN_MIGR 3 | |
155 | #define MIGR_STATE_CHANGE 4 | |
1484e727 | 156 | #define MIGR_REPAIR 5 |
cdddbdbc DW |
157 | __u8 migr_type; /* Initializing, Rebuilding, ... */ |
158 | __u8 dirty; | |
fe7ed8cb DW |
159 | __u8 fs_state; /* fast-sync state for CnG (0xff == disabled) */ |
160 | __u16 verify_errors; /* number of mismatches */ | |
161 | __u16 bad_blocks; /* number of bad blocks during verify */ | |
162 | __u32 filler[4]; | |
cdddbdbc DW |
163 | struct imsm_map map[1]; |
164 | /* here comes another one if migr_state */ | |
165 | } __attribute__ ((packed)); | |
166 | ||
167 | struct imsm_dev { | |
fe7ed8cb | 168 | __u8 volume[MAX_RAID_SERIAL_LEN]; |
cdddbdbc DW |
169 | __u32 size_low; |
170 | __u32 size_high; | |
fe7ed8cb DW |
171 | #define DEV_BOOTABLE __cpu_to_le32(0x01) |
172 | #define DEV_BOOT_DEVICE __cpu_to_le32(0x02) | |
173 | #define DEV_READ_COALESCING __cpu_to_le32(0x04) | |
174 | #define DEV_WRITE_COALESCING __cpu_to_le32(0x08) | |
175 | #define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10) | |
176 | #define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20) | |
177 | #define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40) | |
178 | #define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80) | |
179 | #define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100) | |
180 | #define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200) | |
181 | #define DEV_CLONE_N_GO __cpu_to_le32(0x400) | |
182 | #define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800) | |
183 | #define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000) | |
cdddbdbc DW |
184 | __u32 status; /* Persistent RaidDev status */ |
185 | __u32 reserved_blocks; /* Reserved blocks at beginning of volume */ | |
fe7ed8cb DW |
186 | __u8 migr_priority; |
187 | __u8 num_sub_vols; | |
188 | __u8 tid; | |
189 | __u8 cng_master_disk; | |
190 | __u16 cache_policy; | |
191 | __u8 cng_state; | |
192 | __u8 cng_sub_state; | |
193 | #define IMSM_DEV_FILLERS 10 | |
cdddbdbc DW |
194 | __u32 filler[IMSM_DEV_FILLERS]; |
195 | struct imsm_vol vol; | |
196 | } __attribute__ ((packed)); | |
197 | ||
198 | struct imsm_super { | |
199 | __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */ | |
200 | __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */ | |
201 | __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */ | |
202 | __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */ | |
203 | __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */ | |
604b746f JD |
204 | __u32 error_log_size; /* 0x30 - 0x33 in bytes */ |
205 | __u32 attributes; /* 0x34 - 0x37 */ | |
cdddbdbc DW |
206 | __u8 num_disks; /* 0x38 Number of configured disks */ |
207 | __u8 num_raid_devs; /* 0x39 Number of configured volumes */ | |
604b746f JD |
208 | __u8 error_log_pos; /* 0x3A */ |
209 | __u8 fill[1]; /* 0x3B */ | |
210 | __u32 cache_size; /* 0x3c - 0x40 in mb */ | |
211 | __u32 orig_family_num; /* 0x40 - 0x43 original family num */ | |
212 | __u32 pwr_cycle_count; /* 0x44 - 0x47 simulated power cycle count for array */ | |
213 | __u32 bbm_log_size; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */ | |
214 | #define IMSM_FILLERS 35 | |
215 | __u32 filler[IMSM_FILLERS]; /* 0x4C - 0xD7 RAID_MPB_FILLERS */ | |
cdddbdbc DW |
216 | struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */ |
217 | /* here comes imsm_dev[num_raid_devs] */ | |
604b746f | 218 | /* here comes BBM logs */ |
cdddbdbc DW |
219 | } __attribute__ ((packed)); |
220 | ||
604b746f | 221 | #define BBM_LOG_MAX_ENTRIES 254 |
8d67477f TM |
222 | #define BBM_LOG_MAX_LBA_ENTRY_VAL 256 /* Represents 256 LBAs */ |
223 | #define BBM_LOG_SIGNATURE 0xabadb10c | |
224 | ||
225 | struct bbm_log_block_addr { | |
226 | __u16 w1; | |
227 | __u32 dw1; | |
228 | } __attribute__ ((__packed__)); | |
604b746f JD |
229 | |
230 | struct bbm_log_entry { | |
8d67477f TM |
231 | __u8 marked_count; /* Number of blocks marked - 1 */ |
232 | __u8 disk_ordinal; /* Disk entry within the imsm_super */ | |
233 | struct bbm_log_block_addr defective_block_start; | |
604b746f JD |
234 | } __attribute__ ((__packed__)); |
235 | ||
236 | struct bbm_log { | |
237 | __u32 signature; /* 0xABADB10C */ | |
238 | __u32 entry_count; | |
8d67477f | 239 | struct bbm_log_entry marked_block_entries[BBM_LOG_MAX_ENTRIES]; |
604b746f JD |
240 | } __attribute__ ((__packed__)); |
241 | ||
cdddbdbc DW |
242 | #ifndef MDASSEMBLE |
243 | static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" }; | |
244 | #endif | |
245 | ||
8e59f3d8 AK |
246 | #define RAID_DISK_RESERVED_BLOCKS_IMSM_HI 2209 |
247 | ||
248 | #define GEN_MIGR_AREA_SIZE 2048 /* General Migration Copy Area size in blocks */ | |
249 | ||
de44e46f PB |
250 | #define MIGR_REC_BUF_SECTORS 1 /* size of migr_record i/o buffer in sectors */ |
251 | #define MIGR_REC_SECTOR_POSITION 1 /* migr_record position offset on disk, | |
252 | * MIGR_REC_BUF_SECTORS <= MIGR_REC_SECTOR_POS | |
17a4eaf9 AK |
253 | */ |
254 | ||
8e59f3d8 AK |
255 | #define UNIT_SRC_NORMAL 0 /* Source data for curr_migr_unit must |
256 | * be recovered using srcMap */ | |
257 | #define UNIT_SRC_IN_CP_AREA 1 /* Source data for curr_migr_unit has | |
258 | * already been migrated and must | |
259 | * be recovered from checkpoint area */ | |
260 | struct migr_record { | |
261 | __u32 rec_status; /* Status used to determine how to restart | |
262 | * migration in case it aborts | |
263 | * in some fashion */ | |
264 | __u32 curr_migr_unit; /* 0..numMigrUnits-1 */ | |
265 | __u32 family_num; /* Family number of MPB | |
266 | * containing the RaidDev | |
267 | * that is migrating */ | |
268 | __u32 ascending_migr; /* True if migrating in increasing | |
269 | * order of lbas */ | |
270 | __u32 blocks_per_unit; /* Num disk blocks per unit of operation */ | |
271 | __u32 dest_depth_per_unit; /* Num member blocks each destMap | |
272 | * member disk | |
273 | * advances per unit-of-operation */ | |
274 | __u32 ckpt_area_pba; /* Pba of first block of ckpt copy area */ | |
275 | __u32 dest_1st_member_lba; /* First member lba on first | |
276 | * stripe of destination */ | |
277 | __u32 num_migr_units; /* Total num migration units-of-op */ | |
278 | __u32 post_migr_vol_cap; /* Size of volume after | |
279 | * migration completes */ | |
280 | __u32 post_migr_vol_cap_hi; /* Expansion space for LBA64 */ | |
281 | __u32 ckpt_read_disk_num; /* Which member disk in destSubMap[0] the | |
282 | * migration ckpt record was read from | |
283 | * (for recovered migrations) */ | |
284 | } __attribute__ ((__packed__)); | |
285 | ||
ec50f7b6 LM |
286 | struct md_list { |
287 | /* usage marker: | |
288 | * 1: load metadata | |
289 | * 2: metadata does not match | |
290 | * 4: already checked | |
291 | */ | |
292 | int used; | |
293 | char *devname; | |
294 | int found; | |
295 | int container; | |
296 | dev_t st_rdev; | |
297 | struct md_list *next; | |
298 | }; | |
299 | ||
e7b84f9d | 300 | #define pr_vrb(fmt, arg...) (void) (verbose && pr_err(fmt, ##arg)) |
ec50f7b6 | 301 | |
1484e727 DW |
302 | static __u8 migr_type(struct imsm_dev *dev) |
303 | { | |
304 | if (dev->vol.migr_type == MIGR_VERIFY && | |
305 | dev->status & DEV_VERIFY_AND_FIX) | |
306 | return MIGR_REPAIR; | |
307 | else | |
308 | return dev->vol.migr_type; | |
309 | } | |
310 | ||
311 | static void set_migr_type(struct imsm_dev *dev, __u8 migr_type) | |
312 | { | |
313 | /* for compatibility with older oroms convert MIGR_REPAIR, into | |
314 | * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status | |
315 | */ | |
316 | if (migr_type == MIGR_REPAIR) { | |
317 | dev->vol.migr_type = MIGR_VERIFY; | |
318 | dev->status |= DEV_VERIFY_AND_FIX; | |
319 | } else { | |
320 | dev->vol.migr_type = migr_type; | |
321 | dev->status &= ~DEV_VERIFY_AND_FIX; | |
322 | } | |
323 | } | |
324 | ||
f36a9ecd | 325 | static unsigned int sector_count(__u32 bytes, unsigned int sector_size) |
cdddbdbc | 326 | { |
f36a9ecd | 327 | return ROUND_UP(bytes, sector_size) / sector_size; |
87eb16df | 328 | } |
cdddbdbc | 329 | |
f36a9ecd PB |
330 | static unsigned int mpb_sectors(struct imsm_super *mpb, |
331 | unsigned int sector_size) | |
87eb16df | 332 | { |
f36a9ecd | 333 | return sector_count(__le32_to_cpu(mpb->mpb_size), sector_size); |
cdddbdbc DW |
334 | } |
335 | ||
ba2de7ba DW |
336 | struct intel_dev { |
337 | struct imsm_dev *dev; | |
338 | struct intel_dev *next; | |
f21e18ca | 339 | unsigned index; |
ba2de7ba DW |
340 | }; |
341 | ||
88654014 LM |
342 | struct intel_hba { |
343 | enum sys_dev_type type; | |
344 | char *path; | |
345 | char *pci_id; | |
346 | struct intel_hba *next; | |
347 | }; | |
348 | ||
1a64be56 LM |
349 | enum action { |
350 | DISK_REMOVE = 1, | |
351 | DISK_ADD | |
352 | }; | |
cdddbdbc DW |
353 | /* internal representation of IMSM metadata */ |
354 | struct intel_super { | |
355 | union { | |
949c47a0 DW |
356 | void *buf; /* O_DIRECT buffer for reading/writing metadata */ |
357 | struct imsm_super *anchor; /* immovable parameters */ | |
cdddbdbc | 358 | }; |
8e59f3d8 AK |
359 | union { |
360 | void *migr_rec_buf; /* buffer for I/O operations */ | |
361 | struct migr_record *migr_rec; /* migration record */ | |
362 | }; | |
51d83f5d AK |
363 | int clean_migration_record_by_mdmon; /* when reshape is switched to next |
364 | array, it indicates that mdmon is allowed to clean migration | |
365 | record */ | |
949c47a0 | 366 | size_t len; /* size of the 'buf' allocation */ |
bbab0940 | 367 | size_t extra_space; /* extra space in 'buf' that is not used yet */ |
4d7b1503 DW |
368 | void *next_buf; /* for realloc'ing buf from the manager */ |
369 | size_t next_len; | |
c2c087e6 | 370 | int updates_pending; /* count of pending updates for mdmon */ |
bf5a934a | 371 | int current_vol; /* index of raid device undergoing creation */ |
5551b113 | 372 | unsigned long long create_offset; /* common start for 'current_vol' */ |
148acb7b | 373 | __u32 random; /* random data for seeding new family numbers */ |
ba2de7ba | 374 | struct intel_dev *devlist; |
fa7bb6f8 | 375 | unsigned int sector_size; /* sector size of used member drives */ |
cdddbdbc DW |
376 | struct dl { |
377 | struct dl *next; | |
378 | int index; | |
379 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
380 | int major, minor; | |
381 | char *devname; | |
b9f594fe | 382 | struct imsm_disk disk; |
cdddbdbc | 383 | int fd; |
0dcecb2e DW |
384 | int extent_cnt; |
385 | struct extent *e; /* for determining freespace @ create */ | |
efb30e7f | 386 | int raiddisk; /* slot to fill in autolayout */ |
1a64be56 | 387 | enum action action; |
ca0748fa | 388 | } *disks, *current_disk; |
1a64be56 LM |
389 | struct dl *disk_mgmt_list; /* list of disks to add/remove while mdmon |
390 | active */ | |
47ee5a45 | 391 | struct dl *missing; /* disks removed while we weren't looking */ |
43dad3d6 | 392 | struct bbm_log *bbm_log; |
88654014 | 393 | struct intel_hba *hba; /* device path of the raid controller for this metadata */ |
88c32bb1 | 394 | const struct imsm_orom *orom; /* platform firmware support */ |
a2b97981 | 395 | struct intel_super *next; /* (temp) list for disambiguating family_num */ |
928f1424 | 396 | struct md_bb bb; /* memory for get_bad_blocks call */ |
a2b97981 DW |
397 | }; |
398 | ||
399 | struct intel_disk { | |
400 | struct imsm_disk disk; | |
401 | #define IMSM_UNKNOWN_OWNER (-1) | |
402 | int owner; | |
403 | struct intel_disk *next; | |
cdddbdbc DW |
404 | }; |
405 | ||
c2c087e6 DW |
406 | struct extent { |
407 | unsigned long long start, size; | |
408 | }; | |
409 | ||
694575e7 KW |
410 | /* definitions of reshape process types */ |
411 | enum imsm_reshape_type { | |
412 | CH_TAKEOVER, | |
b5347799 | 413 | CH_MIGRATION, |
7abc9871 | 414 | CH_ARRAY_SIZE, |
694575e7 KW |
415 | }; |
416 | ||
88758e9d DW |
417 | /* definition of messages passed to imsm_process_update */ |
418 | enum imsm_update_type { | |
419 | update_activate_spare, | |
8273f55e | 420 | update_create_array, |
33414a01 | 421 | update_kill_array, |
aa534678 | 422 | update_rename_array, |
1a64be56 | 423 | update_add_remove_disk, |
78b10e66 | 424 | update_reshape_container_disks, |
48c5303a | 425 | update_reshape_migration, |
2d40f3a1 AK |
426 | update_takeover, |
427 | update_general_migration_checkpoint, | |
f3871fdc | 428 | update_size_change, |
bbab0940 | 429 | update_prealloc_badblocks_mem, |
88758e9d DW |
430 | }; |
431 | ||
432 | struct imsm_update_activate_spare { | |
433 | enum imsm_update_type type; | |
d23fe947 | 434 | struct dl *dl; |
88758e9d DW |
435 | int slot; |
436 | int array; | |
437 | struct imsm_update_activate_spare *next; | |
438 | }; | |
439 | ||
78b10e66 | 440 | struct geo_params { |
4dd2df09 | 441 | char devnm[32]; |
78b10e66 | 442 | char *dev_name; |
d04f65f4 | 443 | unsigned long long size; |
78b10e66 N |
444 | int level; |
445 | int layout; | |
446 | int chunksize; | |
447 | int raid_disks; | |
448 | }; | |
449 | ||
bb025c2f KW |
450 | enum takeover_direction { |
451 | R10_TO_R0, | |
452 | R0_TO_R10 | |
453 | }; | |
454 | struct imsm_update_takeover { | |
455 | enum imsm_update_type type; | |
456 | int subarray; | |
457 | enum takeover_direction direction; | |
458 | }; | |
78b10e66 N |
459 | |
460 | struct imsm_update_reshape { | |
461 | enum imsm_update_type type; | |
462 | int old_raid_disks; | |
463 | int new_raid_disks; | |
48c5303a PC |
464 | |
465 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ | |
466 | }; | |
467 | ||
468 | struct imsm_update_reshape_migration { | |
469 | enum imsm_update_type type; | |
470 | int old_raid_disks; | |
471 | int new_raid_disks; | |
472 | /* fields for array migration changes | |
473 | */ | |
474 | int subdev; | |
475 | int new_level; | |
476 | int new_layout; | |
4bba0439 | 477 | int new_chunksize; |
48c5303a | 478 | |
d195167d | 479 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ |
78b10e66 N |
480 | }; |
481 | ||
f3871fdc AK |
482 | struct imsm_update_size_change { |
483 | enum imsm_update_type type; | |
484 | int subdev; | |
485 | long long new_size; | |
486 | }; | |
487 | ||
2d40f3a1 AK |
488 | struct imsm_update_general_migration_checkpoint { |
489 | enum imsm_update_type type; | |
490 | __u32 curr_migr_unit; | |
491 | }; | |
492 | ||
54c2c1ea DW |
493 | struct disk_info { |
494 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
495 | }; | |
496 | ||
8273f55e DW |
497 | struct imsm_update_create_array { |
498 | enum imsm_update_type type; | |
8273f55e | 499 | int dev_idx; |
6a3e913e | 500 | struct imsm_dev dev; |
8273f55e DW |
501 | }; |
502 | ||
33414a01 DW |
503 | struct imsm_update_kill_array { |
504 | enum imsm_update_type type; | |
505 | int dev_idx; | |
506 | }; | |
507 | ||
aa534678 DW |
508 | struct imsm_update_rename_array { |
509 | enum imsm_update_type type; | |
510 | __u8 name[MAX_RAID_SERIAL_LEN]; | |
511 | int dev_idx; | |
512 | }; | |
513 | ||
1a64be56 | 514 | struct imsm_update_add_remove_disk { |
43dad3d6 DW |
515 | enum imsm_update_type type; |
516 | }; | |
517 | ||
bbab0940 TM |
518 | struct imsm_update_prealloc_bb_mem { |
519 | enum imsm_update_type type; | |
520 | }; | |
521 | ||
88654014 LM |
522 | static const char *_sys_dev_type[] = { |
523 | [SYS_DEV_UNKNOWN] = "Unknown", | |
524 | [SYS_DEV_SAS] = "SAS", | |
614902f6 | 525 | [SYS_DEV_SATA] = "SATA", |
60f0f54d PB |
526 | [SYS_DEV_NVME] = "NVMe", |
527 | [SYS_DEV_VMD] = "VMD" | |
88654014 LM |
528 | }; |
529 | ||
530 | const char *get_sys_dev_type(enum sys_dev_type type) | |
531 | { | |
532 | if (type >= SYS_DEV_MAX) | |
533 | type = SYS_DEV_UNKNOWN; | |
534 | ||
535 | return _sys_dev_type[type]; | |
536 | } | |
537 | ||
538 | static struct intel_hba * alloc_intel_hba(struct sys_dev *device) | |
539 | { | |
503975b9 N |
540 | struct intel_hba *result = xmalloc(sizeof(*result)); |
541 | ||
542 | result->type = device->type; | |
543 | result->path = xstrdup(device->path); | |
544 | result->next = NULL; | |
545 | if (result->path && (result->pci_id = strrchr(result->path, '/')) != NULL) | |
546 | result->pci_id++; | |
547 | ||
88654014 LM |
548 | return result; |
549 | } | |
550 | ||
551 | static struct intel_hba * find_intel_hba(struct intel_hba *hba, struct sys_dev *device) | |
552 | { | |
594dc1b8 JS |
553 | struct intel_hba *result; |
554 | ||
88654014 LM |
555 | for (result = hba; result; result = result->next) { |
556 | if (result->type == device->type && strcmp(result->path, device->path) == 0) | |
557 | break; | |
558 | } | |
559 | return result; | |
560 | } | |
561 | ||
b4cf4cba | 562 | static int attach_hba_to_super(struct intel_super *super, struct sys_dev *device) |
88654014 LM |
563 | { |
564 | struct intel_hba *hba; | |
565 | ||
566 | /* check if disk attached to Intel HBA */ | |
567 | hba = find_intel_hba(super->hba, device); | |
568 | if (hba != NULL) | |
569 | return 1; | |
570 | /* Check if HBA is already attached to super */ | |
571 | if (super->hba == NULL) { | |
572 | super->hba = alloc_intel_hba(device); | |
573 | return 1; | |
6b781d33 AP |
574 | } |
575 | ||
576 | hba = super->hba; | |
577 | /* Intel metadata allows for all disks attached to the same type HBA. | |
614902f6 | 578 | * Do not support HBA types mixing |
6b781d33 AP |
579 | */ |
580 | if (device->type != hba->type) | |
88654014 | 581 | return 2; |
6b781d33 AP |
582 | |
583 | /* Multiple same type HBAs can be used if they share the same OROM */ | |
584 | const struct imsm_orom *device_orom = get_orom_by_device_id(device->dev_id); | |
585 | ||
586 | if (device_orom != super->orom) | |
587 | return 2; | |
588 | ||
589 | while (hba->next) | |
590 | hba = hba->next; | |
591 | ||
592 | hba->next = alloc_intel_hba(device); | |
593 | return 1; | |
88654014 LM |
594 | } |
595 | ||
596 | static struct sys_dev* find_disk_attached_hba(int fd, const char *devname) | |
597 | { | |
9bc4ae77 | 598 | struct sys_dev *list, *elem; |
88654014 LM |
599 | char *disk_path; |
600 | ||
601 | if ((list = find_intel_devices()) == NULL) | |
602 | return 0; | |
603 | ||
604 | if (fd < 0) | |
605 | disk_path = (char *) devname; | |
606 | else | |
607 | disk_path = diskfd_to_devpath(fd); | |
608 | ||
9bc4ae77 | 609 | if (!disk_path) |
88654014 | 610 | return 0; |
88654014 | 611 | |
9bc4ae77 N |
612 | for (elem = list; elem; elem = elem->next) |
613 | if (path_attached_to_hba(disk_path, elem->path)) | |
88654014 | 614 | return elem; |
9bc4ae77 | 615 | |
88654014 LM |
616 | if (disk_path != devname) |
617 | free(disk_path); | |
88654014 LM |
618 | |
619 | return NULL; | |
620 | } | |
621 | ||
d424212e N |
622 | static int find_intel_hba_capability(int fd, struct intel_super *super, |
623 | char *devname); | |
f2f5c343 | 624 | |
cdddbdbc DW |
625 | static struct supertype *match_metadata_desc_imsm(char *arg) |
626 | { | |
627 | struct supertype *st; | |
628 | ||
629 | if (strcmp(arg, "imsm") != 0 && | |
630 | strcmp(arg, "default") != 0 | |
631 | ) | |
632 | return NULL; | |
633 | ||
503975b9 | 634 | st = xcalloc(1, sizeof(*st)); |
cdddbdbc DW |
635 | st->ss = &super_imsm; |
636 | st->max_devs = IMSM_MAX_DEVICES; | |
637 | st->minor_version = 0; | |
638 | st->sb = NULL; | |
639 | return st; | |
640 | } | |
641 | ||
0e600426 | 642 | #ifndef MDASSEMBLE |
cdddbdbc DW |
643 | static __u8 *get_imsm_version(struct imsm_super *mpb) |
644 | { | |
645 | return &mpb->sig[MPB_SIG_LEN]; | |
646 | } | |
9e2d750d | 647 | #endif |
cdddbdbc | 648 | |
949c47a0 DW |
649 | /* retrieve a disk directly from the anchor when the anchor is known to be |
650 | * up-to-date, currently only at load time | |
651 | */ | |
652 | static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index) | |
cdddbdbc | 653 | { |
949c47a0 | 654 | if (index >= mpb->num_disks) |
cdddbdbc DW |
655 | return NULL; |
656 | return &mpb->disk[index]; | |
657 | } | |
658 | ||
95d07a2c LM |
659 | /* retrieve the disk description based on a index of the disk |
660 | * in the sub-array | |
661 | */ | |
662 | static struct dl *get_imsm_dl_disk(struct intel_super *super, __u8 index) | |
949c47a0 | 663 | { |
b9f594fe DW |
664 | struct dl *d; |
665 | ||
666 | for (d = super->disks; d; d = d->next) | |
667 | if (d->index == index) | |
95d07a2c LM |
668 | return d; |
669 | ||
670 | return NULL; | |
671 | } | |
672 | /* retrieve a disk from the parsed metadata */ | |
673 | static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index) | |
674 | { | |
675 | struct dl *dl; | |
676 | ||
677 | dl = get_imsm_dl_disk(super, index); | |
678 | if (dl) | |
679 | return &dl->disk; | |
680 | ||
b9f594fe | 681 | return NULL; |
949c47a0 DW |
682 | } |
683 | ||
684 | /* generate a checksum directly from the anchor when the anchor is known to be | |
685 | * up-to-date, currently only at load or write_super after coalescing | |
686 | */ | |
687 | static __u32 __gen_imsm_checksum(struct imsm_super *mpb) | |
cdddbdbc DW |
688 | { |
689 | __u32 end = mpb->mpb_size / sizeof(end); | |
690 | __u32 *p = (__u32 *) mpb; | |
691 | __u32 sum = 0; | |
692 | ||
5d500228 N |
693 | while (end--) { |
694 | sum += __le32_to_cpu(*p); | |
97f734fd N |
695 | p++; |
696 | } | |
cdddbdbc | 697 | |
5d500228 | 698 | return sum - __le32_to_cpu(mpb->check_sum); |
cdddbdbc DW |
699 | } |
700 | ||
a965f303 DW |
701 | static size_t sizeof_imsm_map(struct imsm_map *map) |
702 | { | |
703 | return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1); | |
704 | } | |
705 | ||
706 | struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map) | |
cdddbdbc | 707 | { |
5e7b0330 AK |
708 | /* A device can have 2 maps if it is in the middle of a migration. |
709 | * If second_map is: | |
238c0a71 AK |
710 | * MAP_0 - we return the first map |
711 | * MAP_1 - we return the second map if it exists, else NULL | |
712 | * MAP_X - we return the second map if it exists, else the first | |
5e7b0330 | 713 | */ |
a965f303 | 714 | struct imsm_map *map = &dev->vol.map[0]; |
9535fc47 | 715 | struct imsm_map *map2 = NULL; |
a965f303 | 716 | |
9535fc47 AK |
717 | if (dev->vol.migr_state) |
718 | map2 = (void *)map + sizeof_imsm_map(map); | |
a965f303 | 719 | |
9535fc47 | 720 | switch (second_map) { |
3b451610 | 721 | case MAP_0: |
9535fc47 | 722 | break; |
3b451610 | 723 | case MAP_1: |
9535fc47 AK |
724 | map = map2; |
725 | break; | |
238c0a71 | 726 | case MAP_X: |
9535fc47 AK |
727 | if (map2) |
728 | map = map2; | |
729 | break; | |
9535fc47 AK |
730 | default: |
731 | map = NULL; | |
732 | } | |
733 | return map; | |
5e7b0330 | 734 | |
a965f303 | 735 | } |
cdddbdbc | 736 | |
3393c6af DW |
737 | /* return the size of the device. |
738 | * migr_state increases the returned size if map[0] were to be duplicated | |
739 | */ | |
740 | static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state) | |
a965f303 DW |
741 | { |
742 | size_t size = sizeof(*dev) - sizeof(struct imsm_map) + | |
238c0a71 | 743 | sizeof_imsm_map(get_imsm_map(dev, MAP_0)); |
cdddbdbc DW |
744 | |
745 | /* migrating means an additional map */ | |
a965f303 | 746 | if (dev->vol.migr_state) |
238c0a71 | 747 | size += sizeof_imsm_map(get_imsm_map(dev, MAP_1)); |
3393c6af | 748 | else if (migr_state) |
238c0a71 | 749 | size += sizeof_imsm_map(get_imsm_map(dev, MAP_0)); |
cdddbdbc DW |
750 | |
751 | return size; | |
752 | } | |
753 | ||
54c2c1ea DW |
754 | #ifndef MDASSEMBLE |
755 | /* retrieve disk serial number list from a metadata update */ | |
756 | static struct disk_info *get_disk_info(struct imsm_update_create_array *update) | |
757 | { | |
758 | void *u = update; | |
759 | struct disk_info *inf; | |
760 | ||
761 | inf = u + sizeof(*update) - sizeof(struct imsm_dev) + | |
762 | sizeof_imsm_dev(&update->dev, 0); | |
763 | ||
764 | return inf; | |
765 | } | |
766 | #endif | |
767 | ||
949c47a0 | 768 | static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index) |
cdddbdbc DW |
769 | { |
770 | int offset; | |
771 | int i; | |
772 | void *_mpb = mpb; | |
773 | ||
949c47a0 | 774 | if (index >= mpb->num_raid_devs) |
cdddbdbc DW |
775 | return NULL; |
776 | ||
777 | /* devices start after all disks */ | |
778 | offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb; | |
779 | ||
780 | for (i = 0; i <= index; i++) | |
781 | if (i == index) | |
782 | return _mpb + offset; | |
783 | else | |
3393c6af | 784 | offset += sizeof_imsm_dev(_mpb + offset, 0); |
cdddbdbc DW |
785 | |
786 | return NULL; | |
787 | } | |
788 | ||
949c47a0 DW |
789 | static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index) |
790 | { | |
ba2de7ba DW |
791 | struct intel_dev *dv; |
792 | ||
949c47a0 DW |
793 | if (index >= super->anchor->num_raid_devs) |
794 | return NULL; | |
ba2de7ba DW |
795 | for (dv = super->devlist; dv; dv = dv->next) |
796 | if (dv->index == index) | |
797 | return dv->dev; | |
798 | return NULL; | |
949c47a0 DW |
799 | } |
800 | ||
8d67477f TM |
801 | static inline unsigned long long __le48_to_cpu(const struct bbm_log_block_addr |
802 | *addr) | |
803 | { | |
804 | return ((((__u64)__le32_to_cpu(addr->dw1)) << 16) | | |
805 | __le16_to_cpu(addr->w1)); | |
806 | } | |
807 | ||
808 | static inline struct bbm_log_block_addr __cpu_to_le48(unsigned long long sec) | |
809 | { | |
810 | struct bbm_log_block_addr addr; | |
811 | ||
812 | addr.w1 = __cpu_to_le16((__u16)(sec & 0xffff)); | |
813 | addr.dw1 = __cpu_to_le32((__u32)(sec >> 16) & 0xffffffff); | |
814 | return addr; | |
815 | } | |
816 | ||
817 | #ifndef MDASSEMBLE | |
818 | /* get size of the bbm log */ | |
819 | static __u32 get_imsm_bbm_log_size(struct bbm_log *log) | |
820 | { | |
821 | if (!log || log->entry_count == 0) | |
822 | return 0; | |
823 | ||
824 | return sizeof(log->signature) + | |
825 | sizeof(log->entry_count) + | |
826 | log->entry_count * sizeof(struct bbm_log_entry); | |
827 | } | |
6f50473f TM |
828 | |
829 | /* check if bad block is not partially stored in bbm log */ | |
830 | static int is_stored_in_bbm(struct bbm_log *log, const __u8 idx, const unsigned | |
831 | long long sector, const int length, __u32 *pos) | |
832 | { | |
833 | __u32 i; | |
834 | ||
835 | for (i = *pos; i < log->entry_count; i++) { | |
836 | struct bbm_log_entry *entry = &log->marked_block_entries[i]; | |
837 | unsigned long long bb_start; | |
838 | unsigned long long bb_end; | |
839 | ||
840 | bb_start = __le48_to_cpu(&entry->defective_block_start); | |
841 | bb_end = bb_start + (entry->marked_count + 1); | |
842 | ||
843 | if ((entry->disk_ordinal == idx) && (bb_start >= sector) && | |
844 | (bb_end <= sector + length)) { | |
845 | *pos = i; | |
846 | return 1; | |
847 | } | |
848 | } | |
849 | return 0; | |
850 | } | |
851 | ||
852 | /* record new bad block in bbm log */ | |
853 | static int record_new_badblock(struct bbm_log *log, const __u8 idx, unsigned | |
854 | long long sector, int length) | |
855 | { | |
856 | int new_bb = 0; | |
857 | __u32 pos = 0; | |
858 | struct bbm_log_entry *entry = NULL; | |
859 | ||
860 | while (is_stored_in_bbm(log, idx, sector, length, &pos)) { | |
861 | struct bbm_log_entry *e = &log->marked_block_entries[pos]; | |
862 | ||
863 | if ((e->marked_count + 1 == BBM_LOG_MAX_LBA_ENTRY_VAL) && | |
864 | (__le48_to_cpu(&e->defective_block_start) == sector)) { | |
865 | sector += BBM_LOG_MAX_LBA_ENTRY_VAL; | |
866 | length -= BBM_LOG_MAX_LBA_ENTRY_VAL; | |
867 | pos = pos + 1; | |
868 | continue; | |
869 | } | |
870 | entry = e; | |
871 | break; | |
872 | } | |
873 | ||
874 | if (entry) { | |
875 | int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length : | |
876 | BBM_LOG_MAX_LBA_ENTRY_VAL; | |
877 | entry->defective_block_start = __cpu_to_le48(sector); | |
878 | entry->marked_count = cnt - 1; | |
879 | if (cnt == length) | |
880 | return 1; | |
881 | sector += cnt; | |
882 | length -= cnt; | |
883 | } | |
884 | ||
885 | new_bb = ROUND_UP(length, BBM_LOG_MAX_LBA_ENTRY_VAL) / | |
886 | BBM_LOG_MAX_LBA_ENTRY_VAL; | |
887 | if (log->entry_count + new_bb > BBM_LOG_MAX_ENTRIES) | |
888 | return 0; | |
889 | ||
890 | while (length > 0) { | |
891 | int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length : | |
892 | BBM_LOG_MAX_LBA_ENTRY_VAL; | |
893 | struct bbm_log_entry *entry = | |
894 | &log->marked_block_entries[log->entry_count]; | |
895 | ||
896 | entry->defective_block_start = __cpu_to_le48(sector); | |
897 | entry->marked_count = cnt - 1; | |
898 | entry->disk_ordinal = idx; | |
899 | ||
900 | sector += cnt; | |
901 | length -= cnt; | |
902 | ||
903 | log->entry_count++; | |
904 | } | |
905 | ||
906 | return new_bb; | |
907 | } | |
c07a5a4f | 908 | |
4c9e8c1e TM |
909 | /* clear all bad blocks for given disk */ |
910 | static void clear_disk_badblocks(struct bbm_log *log, const __u8 idx) | |
911 | { | |
912 | __u32 i = 0; | |
913 | ||
914 | while (i < log->entry_count) { | |
915 | struct bbm_log_entry *entries = log->marked_block_entries; | |
916 | ||
917 | if (entries[i].disk_ordinal == idx) { | |
918 | if (i < log->entry_count - 1) | |
919 | entries[i] = entries[log->entry_count - 1]; | |
920 | log->entry_count--; | |
921 | } else { | |
922 | i++; | |
923 | } | |
924 | } | |
925 | } | |
926 | ||
c07a5a4f TM |
927 | /* clear given bad block */ |
928 | static int clear_badblock(struct bbm_log *log, const __u8 idx, const unsigned | |
929 | long long sector, const int length) { | |
930 | __u32 i = 0; | |
931 | ||
932 | while (i < log->entry_count) { | |
933 | struct bbm_log_entry *entries = log->marked_block_entries; | |
934 | ||
935 | if ((entries[i].disk_ordinal == idx) && | |
936 | (__le48_to_cpu(&entries[i].defective_block_start) == | |
937 | sector) && (entries[i].marked_count + 1 == length)) { | |
938 | if (i < log->entry_count - 1) | |
939 | entries[i] = entries[log->entry_count - 1]; | |
940 | log->entry_count--; | |
941 | break; | |
942 | } | |
943 | i++; | |
944 | } | |
945 | ||
946 | return 1; | |
947 | } | |
8d67477f TM |
948 | #endif /* MDASSEMBLE */ |
949 | ||
950 | /* allocate and load BBM log from metadata */ | |
951 | static int load_bbm_log(struct intel_super *super) | |
952 | { | |
953 | struct imsm_super *mpb = super->anchor; | |
954 | __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size); | |
955 | ||
956 | super->bbm_log = xcalloc(1, sizeof(struct bbm_log)); | |
957 | if (!super->bbm_log) | |
958 | return 1; | |
959 | ||
960 | if (bbm_log_size) { | |
961 | struct bbm_log *log = (void *)mpb + | |
962 | __le32_to_cpu(mpb->mpb_size) - bbm_log_size; | |
963 | ||
964 | __u32 entry_count; | |
965 | ||
966 | if (bbm_log_size < sizeof(log->signature) + | |
967 | sizeof(log->entry_count)) | |
968 | return 2; | |
969 | ||
970 | entry_count = __le32_to_cpu(log->entry_count); | |
971 | if ((__le32_to_cpu(log->signature) != BBM_LOG_SIGNATURE) || | |
972 | (entry_count > BBM_LOG_MAX_ENTRIES)) | |
973 | return 3; | |
974 | ||
975 | if (bbm_log_size != | |
976 | sizeof(log->signature) + sizeof(log->entry_count) + | |
977 | entry_count * sizeof(struct bbm_log_entry)) | |
978 | return 4; | |
979 | ||
980 | memcpy(super->bbm_log, log, bbm_log_size); | |
981 | } else { | |
982 | super->bbm_log->signature = __cpu_to_le32(BBM_LOG_SIGNATURE); | |
983 | super->bbm_log->entry_count = 0; | |
984 | } | |
985 | ||
986 | return 0; | |
987 | } | |
988 | ||
b12796be TM |
989 | /* checks if bad block is within volume boundaries */ |
990 | static int is_bad_block_in_volume(const struct bbm_log_entry *entry, | |
991 | const unsigned long long start_sector, | |
992 | const unsigned long long size) | |
993 | { | |
994 | unsigned long long bb_start; | |
995 | unsigned long long bb_end; | |
996 | ||
997 | bb_start = __le48_to_cpu(&entry->defective_block_start); | |
998 | bb_end = bb_start + (entry->marked_count + 1); | |
999 | ||
1000 | if (((bb_start >= start_sector) && (bb_start < start_sector + size)) || | |
1001 | ((bb_end >= start_sector) && (bb_end <= start_sector + size))) | |
1002 | return 1; | |
1003 | ||
1004 | return 0; | |
1005 | } | |
1006 | ||
1007 | /* get list of bad blocks on a drive for a volume */ | |
1008 | static void get_volume_badblocks(const struct bbm_log *log, const __u8 idx, | |
1009 | const unsigned long long start_sector, | |
1010 | const unsigned long long size, | |
1011 | struct md_bb *bbs) | |
1012 | { | |
1013 | __u32 count = 0; | |
1014 | __u32 i; | |
1015 | ||
1016 | for (i = 0; i < log->entry_count; i++) { | |
1017 | const struct bbm_log_entry *ent = | |
1018 | &log->marked_block_entries[i]; | |
1019 | struct md_bb_entry *bb; | |
1020 | ||
1021 | if ((ent->disk_ordinal == idx) && | |
1022 | is_bad_block_in_volume(ent, start_sector, size)) { | |
1023 | ||
1024 | if (!bbs->entries) { | |
1025 | bbs->entries = xmalloc(BBM_LOG_MAX_ENTRIES * | |
1026 | sizeof(*bb)); | |
1027 | if (!bbs->entries) | |
1028 | break; | |
1029 | } | |
1030 | ||
1031 | bb = &bbs->entries[count++]; | |
1032 | bb->sector = __le48_to_cpu(&ent->defective_block_start); | |
1033 | bb->length = ent->marked_count + 1; | |
1034 | } | |
1035 | } | |
1036 | bbs->count = count; | |
1037 | } | |
1038 | ||
98130f40 AK |
1039 | /* |
1040 | * for second_map: | |
238c0a71 AK |
1041 | * == MAP_0 get first map |
1042 | * == MAP_1 get second map | |
1043 | * == MAP_X than get map according to the current migr_state | |
98130f40 AK |
1044 | */ |
1045 | static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev, | |
1046 | int slot, | |
1047 | int second_map) | |
7eef0453 DW |
1048 | { |
1049 | struct imsm_map *map; | |
1050 | ||
5e7b0330 | 1051 | map = get_imsm_map(dev, second_map); |
7eef0453 | 1052 | |
ff077194 DW |
1053 | /* top byte identifies disk under rebuild */ |
1054 | return __le32_to_cpu(map->disk_ord_tbl[slot]); | |
1055 | } | |
1056 | ||
1057 | #define ord_to_idx(ord) (((ord) << 8) >> 8) | |
98130f40 | 1058 | static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot, int second_map) |
ff077194 | 1059 | { |
98130f40 | 1060 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, second_map); |
ff077194 DW |
1061 | |
1062 | return ord_to_idx(ord); | |
7eef0453 DW |
1063 | } |
1064 | ||
be73972f DW |
1065 | static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord) |
1066 | { | |
1067 | map->disk_ord_tbl[slot] = __cpu_to_le32(ord); | |
1068 | } | |
1069 | ||
f21e18ca | 1070 | static int get_imsm_disk_slot(struct imsm_map *map, unsigned idx) |
620b1713 DW |
1071 | { |
1072 | int slot; | |
1073 | __u32 ord; | |
1074 | ||
1075 | for (slot = 0; slot < map->num_members; slot++) { | |
1076 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
1077 | if (ord_to_idx(ord) == idx) | |
1078 | return slot; | |
1079 | } | |
1080 | ||
1081 | return -1; | |
1082 | } | |
1083 | ||
cdddbdbc DW |
1084 | static int get_imsm_raid_level(struct imsm_map *map) |
1085 | { | |
1086 | if (map->raid_level == 1) { | |
1087 | if (map->num_members == 2) | |
1088 | return 1; | |
1089 | else | |
1090 | return 10; | |
1091 | } | |
1092 | ||
1093 | return map->raid_level; | |
1094 | } | |
1095 | ||
c2c087e6 DW |
1096 | static int cmp_extent(const void *av, const void *bv) |
1097 | { | |
1098 | const struct extent *a = av; | |
1099 | const struct extent *b = bv; | |
1100 | if (a->start < b->start) | |
1101 | return -1; | |
1102 | if (a->start > b->start) | |
1103 | return 1; | |
1104 | return 0; | |
1105 | } | |
1106 | ||
0dcecb2e | 1107 | static int count_memberships(struct dl *dl, struct intel_super *super) |
c2c087e6 | 1108 | { |
c2c087e6 | 1109 | int memberships = 0; |
620b1713 | 1110 | int i; |
c2c087e6 | 1111 | |
949c47a0 DW |
1112 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
1113 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 1114 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
c2c087e6 | 1115 | |
620b1713 DW |
1116 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
1117 | memberships++; | |
c2c087e6 | 1118 | } |
0dcecb2e DW |
1119 | |
1120 | return memberships; | |
1121 | } | |
1122 | ||
b81221b7 CA |
1123 | static __u32 imsm_min_reserved_sectors(struct intel_super *super); |
1124 | ||
5551b113 CA |
1125 | static int split_ull(unsigned long long n, __u32 *lo, __u32 *hi) |
1126 | { | |
1127 | if (lo == 0 || hi == 0) | |
1128 | return 1; | |
1129 | *lo = __le32_to_cpu((unsigned)n); | |
1130 | *hi = __le32_to_cpu((unsigned)(n >> 32)); | |
1131 | return 0; | |
1132 | } | |
1133 | ||
1134 | static unsigned long long join_u32(__u32 lo, __u32 hi) | |
1135 | { | |
1136 | return (unsigned long long)__le32_to_cpu(lo) | | |
1137 | (((unsigned long long)__le32_to_cpu(hi)) << 32); | |
1138 | } | |
1139 | ||
1140 | static unsigned long long total_blocks(struct imsm_disk *disk) | |
1141 | { | |
1142 | if (disk == NULL) | |
1143 | return 0; | |
1144 | return join_u32(disk->total_blocks_lo, disk->total_blocks_hi); | |
1145 | } | |
1146 | ||
1147 | static unsigned long long pba_of_lba0(struct imsm_map *map) | |
1148 | { | |
1149 | if (map == NULL) | |
1150 | return 0; | |
1151 | return join_u32(map->pba_of_lba0_lo, map->pba_of_lba0_hi); | |
1152 | } | |
1153 | ||
1154 | static unsigned long long blocks_per_member(struct imsm_map *map) | |
1155 | { | |
1156 | if (map == NULL) | |
1157 | return 0; | |
1158 | return join_u32(map->blocks_per_member_lo, map->blocks_per_member_hi); | |
1159 | } | |
1160 | ||
1161 | static unsigned long long num_data_stripes(struct imsm_map *map) | |
1162 | { | |
1163 | if (map == NULL) | |
1164 | return 0; | |
1165 | return join_u32(map->num_data_stripes_lo, map->num_data_stripes_hi); | |
1166 | } | |
1167 | ||
1168 | static void set_total_blocks(struct imsm_disk *disk, unsigned long long n) | |
1169 | { | |
1170 | split_ull(n, &disk->total_blocks_lo, &disk->total_blocks_hi); | |
1171 | } | |
1172 | ||
1173 | static void set_pba_of_lba0(struct imsm_map *map, unsigned long long n) | |
1174 | { | |
1175 | split_ull(n, &map->pba_of_lba0_lo, &map->pba_of_lba0_hi); | |
1176 | } | |
1177 | ||
1178 | static void set_blocks_per_member(struct imsm_map *map, unsigned long long n) | |
1179 | { | |
1180 | split_ull(n, &map->blocks_per_member_lo, &map->blocks_per_member_hi); | |
1181 | } | |
1182 | ||
1183 | static void set_num_data_stripes(struct imsm_map *map, unsigned long long n) | |
1184 | { | |
1185 | split_ull(n, &map->num_data_stripes_lo, &map->num_data_stripes_hi); | |
1186 | } | |
1187 | ||
0dcecb2e DW |
1188 | static struct extent *get_extents(struct intel_super *super, struct dl *dl) |
1189 | { | |
1190 | /* find a list of used extents on the given physical device */ | |
1191 | struct extent *rv, *e; | |
620b1713 | 1192 | int i; |
0dcecb2e | 1193 | int memberships = count_memberships(dl, super); |
b276dd33 DW |
1194 | __u32 reservation; |
1195 | ||
1196 | /* trim the reserved area for spares, so they can join any array | |
1197 | * regardless of whether the OROM has assigned sectors from the | |
1198 | * IMSM_RESERVED_SECTORS region | |
1199 | */ | |
1200 | if (dl->index == -1) | |
b81221b7 | 1201 | reservation = imsm_min_reserved_sectors(super); |
b276dd33 DW |
1202 | else |
1203 | reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
0dcecb2e | 1204 | |
503975b9 | 1205 | rv = xcalloc(sizeof(struct extent), (memberships + 1)); |
c2c087e6 DW |
1206 | e = rv; |
1207 | ||
949c47a0 DW |
1208 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
1209 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 1210 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
c2c087e6 | 1211 | |
620b1713 | 1212 | if (get_imsm_disk_slot(map, dl->index) >= 0) { |
5551b113 CA |
1213 | e->start = pba_of_lba0(map); |
1214 | e->size = blocks_per_member(map); | |
620b1713 | 1215 | e++; |
c2c087e6 DW |
1216 | } |
1217 | } | |
1218 | qsort(rv, memberships, sizeof(*rv), cmp_extent); | |
1219 | ||
1011e834 | 1220 | /* determine the start of the metadata |
14e8215b DW |
1221 | * when no raid devices are defined use the default |
1222 | * ...otherwise allow the metadata to truncate the value | |
1223 | * as is the case with older versions of imsm | |
1224 | */ | |
1225 | if (memberships) { | |
1226 | struct extent *last = &rv[memberships - 1]; | |
5551b113 | 1227 | unsigned long long remainder; |
14e8215b | 1228 | |
5551b113 | 1229 | remainder = total_blocks(&dl->disk) - (last->start + last->size); |
dda5855f DW |
1230 | /* round down to 1k block to satisfy precision of the kernel |
1231 | * 'size' interface | |
1232 | */ | |
1233 | remainder &= ~1UL; | |
1234 | /* make sure remainder is still sane */ | |
f21e18ca | 1235 | if (remainder < (unsigned)ROUND_UP(super->len, 512) >> 9) |
dda5855f | 1236 | remainder = ROUND_UP(super->len, 512) >> 9; |
14e8215b DW |
1237 | if (reservation > remainder) |
1238 | reservation = remainder; | |
1239 | } | |
5551b113 | 1240 | e->start = total_blocks(&dl->disk) - reservation; |
c2c087e6 DW |
1241 | e->size = 0; |
1242 | return rv; | |
1243 | } | |
1244 | ||
14e8215b DW |
1245 | /* try to determine how much space is reserved for metadata from |
1246 | * the last get_extents() entry, otherwise fallback to the | |
1247 | * default | |
1248 | */ | |
1249 | static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl) | |
1250 | { | |
1251 | struct extent *e; | |
1252 | int i; | |
1253 | __u32 rv; | |
1254 | ||
1255 | /* for spares just return a minimal reservation which will grow | |
1256 | * once the spare is picked up by an array | |
1257 | */ | |
1258 | if (dl->index == -1) | |
1259 | return MPB_SECTOR_CNT; | |
1260 | ||
1261 | e = get_extents(super, dl); | |
1262 | if (!e) | |
1263 | return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
1264 | ||
1265 | /* scroll to last entry */ | |
1266 | for (i = 0; e[i].size; i++) | |
1267 | continue; | |
1268 | ||
5551b113 | 1269 | rv = total_blocks(&dl->disk) - e[i].start; |
14e8215b DW |
1270 | |
1271 | free(e); | |
1272 | ||
1273 | return rv; | |
1274 | } | |
1275 | ||
25ed7e59 DW |
1276 | static int is_spare(struct imsm_disk *disk) |
1277 | { | |
1278 | return (disk->status & SPARE_DISK) == SPARE_DISK; | |
1279 | } | |
1280 | ||
1281 | static int is_configured(struct imsm_disk *disk) | |
1282 | { | |
1283 | return (disk->status & CONFIGURED_DISK) == CONFIGURED_DISK; | |
1284 | } | |
1285 | ||
1286 | static int is_failed(struct imsm_disk *disk) | |
1287 | { | |
1288 | return (disk->status & FAILED_DISK) == FAILED_DISK; | |
1289 | } | |
1290 | ||
b81221b7 CA |
1291 | /* try to determine how much space is reserved for metadata from |
1292 | * the last get_extents() entry on the smallest active disk, | |
1293 | * otherwise fallback to the default | |
1294 | */ | |
1295 | static __u32 imsm_min_reserved_sectors(struct intel_super *super) | |
1296 | { | |
1297 | struct extent *e; | |
1298 | int i; | |
5551b113 CA |
1299 | unsigned long long min_active; |
1300 | __u32 remainder; | |
b81221b7 CA |
1301 | __u32 rv = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; |
1302 | struct dl *dl, *dl_min = NULL; | |
1303 | ||
1304 | if (!super) | |
1305 | return rv; | |
1306 | ||
1307 | min_active = 0; | |
1308 | for (dl = super->disks; dl; dl = dl->next) { | |
1309 | if (dl->index < 0) | |
1310 | continue; | |
5551b113 CA |
1311 | unsigned long long blocks = total_blocks(&dl->disk); |
1312 | if (blocks < min_active || min_active == 0) { | |
b81221b7 | 1313 | dl_min = dl; |
5551b113 | 1314 | min_active = blocks; |
b81221b7 CA |
1315 | } |
1316 | } | |
1317 | if (!dl_min) | |
1318 | return rv; | |
1319 | ||
1320 | /* find last lba used by subarrays on the smallest active disk */ | |
1321 | e = get_extents(super, dl_min); | |
1322 | if (!e) | |
1323 | return rv; | |
1324 | for (i = 0; e[i].size; i++) | |
1325 | continue; | |
1326 | ||
1327 | remainder = min_active - e[i].start; | |
1328 | free(e); | |
1329 | ||
1330 | /* to give priority to recovery we should not require full | |
1331 | IMSM_RESERVED_SECTORS from the spare */ | |
1332 | rv = MPB_SECTOR_CNT + NUM_BLOCKS_DIRTY_STRIPE_REGION; | |
1333 | ||
1334 | /* if real reservation is smaller use that value */ | |
1335 | return (remainder < rv) ? remainder : rv; | |
1336 | } | |
1337 | ||
80e7f8c3 AC |
1338 | /* Return minimum size of a spare that can be used in this array*/ |
1339 | static unsigned long long min_acceptable_spare_size_imsm(struct supertype *st) | |
1340 | { | |
1341 | struct intel_super *super = st->sb; | |
1342 | struct dl *dl; | |
1343 | struct extent *e; | |
1344 | int i; | |
1345 | unsigned long long rv = 0; | |
1346 | ||
1347 | if (!super) | |
1348 | return rv; | |
1349 | /* find first active disk in array */ | |
1350 | dl = super->disks; | |
1351 | while (dl && (is_failed(&dl->disk) || dl->index == -1)) | |
1352 | dl = dl->next; | |
1353 | if (!dl) | |
1354 | return rv; | |
1355 | /* find last lba used by subarrays */ | |
1356 | e = get_extents(super, dl); | |
1357 | if (!e) | |
1358 | return rv; | |
1359 | for (i = 0; e[i].size; i++) | |
1360 | continue; | |
1361 | if (i > 0) | |
1362 | rv = e[i-1].start + e[i-1].size; | |
1363 | free(e); | |
b81221b7 | 1364 | |
80e7f8c3 | 1365 | /* add the amount of space needed for metadata */ |
b81221b7 CA |
1366 | rv = rv + imsm_min_reserved_sectors(super); |
1367 | ||
80e7f8c3 AC |
1368 | return rv * 512; |
1369 | } | |
1370 | ||
d1e02575 AK |
1371 | static int is_gen_migration(struct imsm_dev *dev); |
1372 | ||
f36a9ecd PB |
1373 | #define IMSM_4K_DIV 8 |
1374 | ||
1799c9e8 | 1375 | #ifndef MDASSEMBLE |
c47b0ff6 AK |
1376 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
1377 | struct imsm_dev *dev); | |
1e5c6983 | 1378 | |
c47b0ff6 AK |
1379 | static void print_imsm_dev(struct intel_super *super, |
1380 | struct imsm_dev *dev, | |
1381 | char *uuid, | |
1382 | int disk_idx) | |
cdddbdbc DW |
1383 | { |
1384 | __u64 sz; | |
0d80bb2f | 1385 | int slot, i; |
238c0a71 AK |
1386 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1387 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); | |
b10b37b8 | 1388 | __u32 ord; |
cdddbdbc DW |
1389 | |
1390 | printf("\n"); | |
1e7bc0ed | 1391 | printf("[%.16s]:\n", dev->volume); |
44470971 | 1392 | printf(" UUID : %s\n", uuid); |
dd8bcb3b AK |
1393 | printf(" RAID Level : %d", get_imsm_raid_level(map)); |
1394 | if (map2) | |
1395 | printf(" <-- %d", get_imsm_raid_level(map2)); | |
1396 | printf("\n"); | |
1397 | printf(" Members : %d", map->num_members); | |
1398 | if (map2) | |
1399 | printf(" <-- %d", map2->num_members); | |
1400 | printf("\n"); | |
0d80bb2f DW |
1401 | printf(" Slots : ["); |
1402 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 1403 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_0); |
0d80bb2f DW |
1404 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
1405 | } | |
dd8bcb3b AK |
1406 | printf("]"); |
1407 | if (map2) { | |
1408 | printf(" <-- ["); | |
1409 | for (i = 0; i < map2->num_members; i++) { | |
238c0a71 | 1410 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_1); |
dd8bcb3b AK |
1411 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
1412 | } | |
1413 | printf("]"); | |
1414 | } | |
1415 | printf("\n"); | |
7095bccb AK |
1416 | printf(" Failed disk : "); |
1417 | if (map->failed_disk_num == 0xff) | |
1418 | printf("none"); | |
1419 | else | |
1420 | printf("%i", map->failed_disk_num); | |
1421 | printf("\n"); | |
620b1713 DW |
1422 | slot = get_imsm_disk_slot(map, disk_idx); |
1423 | if (slot >= 0) { | |
238c0a71 | 1424 | ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); |
b10b37b8 DW |
1425 | printf(" This Slot : %d%s\n", slot, |
1426 | ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : ""); | |
1427 | } else | |
cdddbdbc DW |
1428 | printf(" This Slot : ?\n"); |
1429 | sz = __le32_to_cpu(dev->size_high); | |
1430 | sz <<= 32; | |
1431 | sz += __le32_to_cpu(dev->size_low); | |
1432 | printf(" Array Size : %llu%s\n", (unsigned long long)sz, | |
1433 | human_size(sz * 512)); | |
5551b113 | 1434 | sz = blocks_per_member(map); |
cdddbdbc DW |
1435 | printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz, |
1436 | human_size(sz * 512)); | |
5551b113 CA |
1437 | printf(" Sector Offset : %llu\n", |
1438 | pba_of_lba0(map)); | |
1439 | printf(" Num Stripes : %llu\n", | |
1440 | num_data_stripes(map)); | |
dd8bcb3b | 1441 | printf(" Chunk Size : %u KiB", |
cdddbdbc | 1442 | __le16_to_cpu(map->blocks_per_strip) / 2); |
dd8bcb3b AK |
1443 | if (map2) |
1444 | printf(" <-- %u KiB", | |
1445 | __le16_to_cpu(map2->blocks_per_strip) / 2); | |
1446 | printf("\n"); | |
cdddbdbc | 1447 | printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks)); |
8655a7b1 | 1448 | printf(" Migrate State : "); |
1484e727 DW |
1449 | if (dev->vol.migr_state) { |
1450 | if (migr_type(dev) == MIGR_INIT) | |
8655a7b1 | 1451 | printf("initialize\n"); |
1484e727 | 1452 | else if (migr_type(dev) == MIGR_REBUILD) |
8655a7b1 | 1453 | printf("rebuild\n"); |
1484e727 | 1454 | else if (migr_type(dev) == MIGR_VERIFY) |
8655a7b1 | 1455 | printf("check\n"); |
1484e727 | 1456 | else if (migr_type(dev) == MIGR_GEN_MIGR) |
8655a7b1 | 1457 | printf("general migration\n"); |
1484e727 | 1458 | else if (migr_type(dev) == MIGR_STATE_CHANGE) |
8655a7b1 | 1459 | printf("state change\n"); |
1484e727 | 1460 | else if (migr_type(dev) == MIGR_REPAIR) |
8655a7b1 | 1461 | printf("repair\n"); |
1484e727 | 1462 | else |
8655a7b1 DW |
1463 | printf("<unknown:%d>\n", migr_type(dev)); |
1464 | } else | |
1465 | printf("idle\n"); | |
3393c6af DW |
1466 | printf(" Map State : %s", map_state_str[map->map_state]); |
1467 | if (dev->vol.migr_state) { | |
238c0a71 | 1468 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 | 1469 | |
b10b37b8 | 1470 | printf(" <-- %s", map_state_str[map->map_state]); |
464d40e8 LD |
1471 | printf("\n Checkpoint : %u ", |
1472 | __le32_to_cpu(dev->vol.curr_migr_unit)); | |
089f9d79 | 1473 | if (is_gen_migration(dev) && (slot > 1 || slot < 0)) |
464d40e8 LD |
1474 | printf("(N/A)"); |
1475 | else | |
1476 | printf("(%llu)", (unsigned long long) | |
1477 | blocks_per_migr_unit(super, dev)); | |
3393c6af DW |
1478 | } |
1479 | printf("\n"); | |
cdddbdbc | 1480 | printf(" Dirty State : %s\n", dev->vol.dirty ? "dirty" : "clean"); |
cdddbdbc DW |
1481 | } |
1482 | ||
ef5c214e MK |
1483 | static void print_imsm_disk(struct imsm_disk *disk, |
1484 | int index, | |
1485 | __u32 reserved, | |
1486 | unsigned int sector_size) { | |
1f24f035 | 1487 | char str[MAX_RAID_SERIAL_LEN + 1]; |
cdddbdbc DW |
1488 | __u64 sz; |
1489 | ||
0ec1f4e8 | 1490 | if (index < -1 || !disk) |
e9d82038 DW |
1491 | return; |
1492 | ||
cdddbdbc | 1493 | printf("\n"); |
1f24f035 | 1494 | snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial); |
0ec1f4e8 DW |
1495 | if (index >= 0) |
1496 | printf(" Disk%02d Serial : %s\n", index, str); | |
1497 | else | |
1498 | printf(" Disk Serial : %s\n", str); | |
25ed7e59 DW |
1499 | printf(" State :%s%s%s\n", is_spare(disk) ? " spare" : "", |
1500 | is_configured(disk) ? " active" : "", | |
1501 | is_failed(disk) ? " failed" : ""); | |
cdddbdbc | 1502 | printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id)); |
5551b113 | 1503 | sz = total_blocks(disk) - reserved; |
ef5c214e MK |
1504 | printf(" Usable Size : %llu%s\n", |
1505 | (unsigned long long)sz * 512 / sector_size, | |
cdddbdbc DW |
1506 | human_size(sz * 512)); |
1507 | } | |
1508 | ||
de44e46f PB |
1509 | void convert_to_4k_imsm_migr_rec(struct intel_super *super) |
1510 | { | |
1511 | struct migr_record *migr_rec = super->migr_rec; | |
1512 | ||
1513 | migr_rec->blocks_per_unit /= IMSM_4K_DIV; | |
1514 | migr_rec->ckpt_area_pba /= IMSM_4K_DIV; | |
1515 | migr_rec->dest_1st_member_lba /= IMSM_4K_DIV; | |
1516 | migr_rec->dest_depth_per_unit /= IMSM_4K_DIV; | |
1517 | split_ull((join_u32(migr_rec->post_migr_vol_cap, | |
1518 | migr_rec->post_migr_vol_cap_hi) / IMSM_4K_DIV), | |
1519 | &migr_rec->post_migr_vol_cap, &migr_rec->post_migr_vol_cap_hi); | |
1520 | } | |
1521 | ||
f36a9ecd PB |
1522 | void convert_to_4k_imsm_disk(struct imsm_disk *disk) |
1523 | { | |
1524 | set_total_blocks(disk, (total_blocks(disk)/IMSM_4K_DIV)); | |
1525 | } | |
1526 | ||
1527 | void convert_to_4k(struct intel_super *super) | |
1528 | { | |
1529 | struct imsm_super *mpb = super->anchor; | |
1530 | struct imsm_disk *disk; | |
1531 | int i; | |
e4467bc7 | 1532 | __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size); |
f36a9ecd PB |
1533 | |
1534 | for (i = 0; i < mpb->num_disks ; i++) { | |
1535 | disk = __get_imsm_disk(mpb, i); | |
1536 | /* disk */ | |
1537 | convert_to_4k_imsm_disk(disk); | |
1538 | } | |
1539 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
1540 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1541 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
1542 | /* dev */ | |
1543 | split_ull((join_u32(dev->size_low, dev->size_high)/IMSM_4K_DIV), | |
1544 | &dev->size_low, &dev->size_high); | |
1545 | dev->vol.curr_migr_unit /= IMSM_4K_DIV; | |
1546 | ||
1547 | /* map0 */ | |
1548 | set_blocks_per_member(map, blocks_per_member(map)/IMSM_4K_DIV); | |
1549 | map->blocks_per_strip /= IMSM_4K_DIV; | |
1550 | set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV); | |
1551 | ||
1552 | if (dev->vol.migr_state) { | |
1553 | /* map1 */ | |
1554 | map = get_imsm_map(dev, MAP_1); | |
1555 | set_blocks_per_member(map, | |
1556 | blocks_per_member(map)/IMSM_4K_DIV); | |
1557 | map->blocks_per_strip /= IMSM_4K_DIV; | |
1558 | set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV); | |
1559 | } | |
1560 | } | |
e4467bc7 TM |
1561 | if (bbm_log_size) { |
1562 | struct bbm_log *log = (void *)mpb + | |
1563 | __le32_to_cpu(mpb->mpb_size) - bbm_log_size; | |
1564 | __u32 i; | |
1565 | ||
1566 | for (i = 0; i < log->entry_count; i++) { | |
1567 | struct bbm_log_entry *entry = | |
1568 | &log->marked_block_entries[i]; | |
1569 | ||
1570 | __u8 count = entry->marked_count + 1; | |
1571 | unsigned long long sector = | |
1572 | __le48_to_cpu(&entry->defective_block_start); | |
1573 | ||
1574 | entry->defective_block_start = | |
1575 | __cpu_to_le48(sector/IMSM_4K_DIV); | |
1576 | entry->marked_count = max(count/IMSM_4K_DIV, 1) - 1; | |
1577 | } | |
1578 | } | |
f36a9ecd PB |
1579 | |
1580 | mpb->check_sum = __gen_imsm_checksum(mpb); | |
1581 | } | |
1582 | ||
520e69e2 AK |
1583 | void examine_migr_rec_imsm(struct intel_super *super) |
1584 | { | |
1585 | struct migr_record *migr_rec = super->migr_rec; | |
1586 | struct imsm_super *mpb = super->anchor; | |
1587 | int i; | |
1588 | ||
1589 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
1590 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
3136abe5 | 1591 | struct imsm_map *map; |
b4ab44d8 | 1592 | int slot = -1; |
3136abe5 | 1593 | |
520e69e2 AK |
1594 | if (is_gen_migration(dev) == 0) |
1595 | continue; | |
1596 | ||
1597 | printf("\nMigration Record Information:"); | |
3136abe5 | 1598 | |
44bfe6df AK |
1599 | /* first map under migration */ |
1600 | map = get_imsm_map(dev, MAP_0); | |
3136abe5 AK |
1601 | if (map) |
1602 | slot = get_imsm_disk_slot(map, super->disks->index); | |
089f9d79 | 1603 | if (map == NULL || slot > 1 || slot < 0) { |
520e69e2 AK |
1604 | printf(" Empty\n "); |
1605 | printf("Examine one of first two disks in array\n"); | |
1606 | break; | |
1607 | } | |
1608 | printf("\n Status : "); | |
1609 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
1610 | printf("Normal\n"); | |
1611 | else | |
1612 | printf("Contains Data\n"); | |
1613 | printf(" Current Unit : %u\n", | |
1614 | __le32_to_cpu(migr_rec->curr_migr_unit)); | |
1615 | printf(" Family : %u\n", | |
1616 | __le32_to_cpu(migr_rec->family_num)); | |
1617 | printf(" Ascending : %u\n", | |
1618 | __le32_to_cpu(migr_rec->ascending_migr)); | |
1619 | printf(" Blocks Per Unit : %u\n", | |
1620 | __le32_to_cpu(migr_rec->blocks_per_unit)); | |
1621 | printf(" Dest. Depth Per Unit : %u\n", | |
1622 | __le32_to_cpu(migr_rec->dest_depth_per_unit)); | |
1623 | printf(" Checkpoint Area pba : %u\n", | |
1624 | __le32_to_cpu(migr_rec->ckpt_area_pba)); | |
1625 | printf(" First member lba : %u\n", | |
1626 | __le32_to_cpu(migr_rec->dest_1st_member_lba)); | |
1627 | printf(" Total Number of Units : %u\n", | |
1628 | __le32_to_cpu(migr_rec->num_migr_units)); | |
1629 | printf(" Size of volume : %u\n", | |
1630 | __le32_to_cpu(migr_rec->post_migr_vol_cap)); | |
1631 | printf(" Expansion space for LBA64 : %u\n", | |
1632 | __le32_to_cpu(migr_rec->post_migr_vol_cap_hi)); | |
1633 | printf(" Record was read from : %u\n", | |
1634 | __le32_to_cpu(migr_rec->ckpt_read_disk_num)); | |
1635 | ||
1636 | break; | |
1637 | } | |
1638 | } | |
9e2d750d | 1639 | #endif /* MDASSEMBLE */ |
f36a9ecd | 1640 | |
de44e46f PB |
1641 | void convert_from_4k_imsm_migr_rec(struct intel_super *super) |
1642 | { | |
1643 | struct migr_record *migr_rec = super->migr_rec; | |
1644 | ||
1645 | migr_rec->blocks_per_unit *= IMSM_4K_DIV; | |
1646 | migr_rec->ckpt_area_pba *= IMSM_4K_DIV; | |
1647 | migr_rec->dest_1st_member_lba *= IMSM_4K_DIV; | |
1648 | migr_rec->dest_depth_per_unit *= IMSM_4K_DIV; | |
1649 | split_ull((join_u32(migr_rec->post_migr_vol_cap, | |
1650 | migr_rec->post_migr_vol_cap_hi) * IMSM_4K_DIV), | |
1651 | &migr_rec->post_migr_vol_cap, | |
1652 | &migr_rec->post_migr_vol_cap_hi); | |
1653 | } | |
1654 | ||
f36a9ecd PB |
1655 | void convert_from_4k(struct intel_super *super) |
1656 | { | |
1657 | struct imsm_super *mpb = super->anchor; | |
1658 | struct imsm_disk *disk; | |
1659 | int i; | |
e4467bc7 | 1660 | __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size); |
f36a9ecd PB |
1661 | |
1662 | for (i = 0; i < mpb->num_disks ; i++) { | |
1663 | disk = __get_imsm_disk(mpb, i); | |
1664 | /* disk */ | |
1665 | set_total_blocks(disk, (total_blocks(disk)*IMSM_4K_DIV)); | |
1666 | } | |
1667 | ||
1668 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
1669 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1670 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
1671 | /* dev */ | |
1672 | split_ull((join_u32(dev->size_low, dev->size_high)*IMSM_4K_DIV), | |
1673 | &dev->size_low, &dev->size_high); | |
1674 | dev->vol.curr_migr_unit *= IMSM_4K_DIV; | |
1675 | ||
1676 | /* map0 */ | |
1677 | set_blocks_per_member(map, blocks_per_member(map)*IMSM_4K_DIV); | |
1678 | map->blocks_per_strip *= IMSM_4K_DIV; | |
1679 | set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV); | |
1680 | ||
1681 | if (dev->vol.migr_state) { | |
1682 | /* map1 */ | |
1683 | map = get_imsm_map(dev, MAP_1); | |
1684 | set_blocks_per_member(map, | |
1685 | blocks_per_member(map)*IMSM_4K_DIV); | |
1686 | map->blocks_per_strip *= IMSM_4K_DIV; | |
1687 | set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV); | |
1688 | } | |
1689 | } | |
e4467bc7 TM |
1690 | if (bbm_log_size) { |
1691 | struct bbm_log *log = (void *)mpb + | |
1692 | __le32_to_cpu(mpb->mpb_size) - bbm_log_size; | |
1693 | __u32 i; | |
1694 | ||
1695 | for (i = 0; i < log->entry_count; i++) { | |
1696 | struct bbm_log_entry *entry = | |
1697 | &log->marked_block_entries[i]; | |
1698 | ||
1699 | __u8 count = entry->marked_count + 1; | |
1700 | unsigned long long sector = | |
1701 | __le48_to_cpu(&entry->defective_block_start); | |
1702 | ||
1703 | entry->defective_block_start = | |
1704 | __cpu_to_le48(sector*IMSM_4K_DIV); | |
1705 | entry->marked_count = count*IMSM_4K_DIV - 1; | |
1706 | } | |
1707 | } | |
f36a9ecd PB |
1708 | |
1709 | mpb->check_sum = __gen_imsm_checksum(mpb); | |
1710 | } | |
1711 | ||
19482bcc AK |
1712 | /******************************************************************************* |
1713 | * function: imsm_check_attributes | |
1714 | * Description: Function checks if features represented by attributes flags | |
1011e834 | 1715 | * are supported by mdadm. |
19482bcc AK |
1716 | * Parameters: |
1717 | * attributes - Attributes read from metadata | |
1718 | * Returns: | |
1011e834 N |
1719 | * 0 - passed attributes contains unsupported features flags |
1720 | * 1 - all features are supported | |
19482bcc AK |
1721 | ******************************************************************************/ |
1722 | static int imsm_check_attributes(__u32 attributes) | |
1723 | { | |
1724 | int ret_val = 1; | |
418f9b36 N |
1725 | __u32 not_supported = MPB_ATTRIB_SUPPORTED^0xffffffff; |
1726 | ||
1727 | not_supported &= ~MPB_ATTRIB_IGNORED; | |
19482bcc AK |
1728 | |
1729 | not_supported &= attributes; | |
1730 | if (not_supported) { | |
e7b84f9d | 1731 | pr_err("(IMSM): Unsupported attributes : %x\n", |
418f9b36 | 1732 | (unsigned)__le32_to_cpu(not_supported)); |
19482bcc AK |
1733 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { |
1734 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY \n"); | |
1735 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
1736 | } | |
1737 | if (not_supported & MPB_ATTRIB_2TB) { | |
1738 | dprintf("\t\tMPB_ATTRIB_2TB\n"); | |
1739 | not_supported ^= MPB_ATTRIB_2TB; | |
1740 | } | |
1741 | if (not_supported & MPB_ATTRIB_RAID0) { | |
1742 | dprintf("\t\tMPB_ATTRIB_RAID0\n"); | |
1743 | not_supported ^= MPB_ATTRIB_RAID0; | |
1744 | } | |
1745 | if (not_supported & MPB_ATTRIB_RAID1) { | |
1746 | dprintf("\t\tMPB_ATTRIB_RAID1\n"); | |
1747 | not_supported ^= MPB_ATTRIB_RAID1; | |
1748 | } | |
1749 | if (not_supported & MPB_ATTRIB_RAID10) { | |
1750 | dprintf("\t\tMPB_ATTRIB_RAID10\n"); | |
1751 | not_supported ^= MPB_ATTRIB_RAID10; | |
1752 | } | |
1753 | if (not_supported & MPB_ATTRIB_RAID1E) { | |
1754 | dprintf("\t\tMPB_ATTRIB_RAID1E\n"); | |
1755 | not_supported ^= MPB_ATTRIB_RAID1E; | |
1756 | } | |
1757 | if (not_supported & MPB_ATTRIB_RAID5) { | |
1758 | dprintf("\t\tMPB_ATTRIB_RAID5\n"); | |
1759 | not_supported ^= MPB_ATTRIB_RAID5; | |
1760 | } | |
1761 | if (not_supported & MPB_ATTRIB_RAIDCNG) { | |
1762 | dprintf("\t\tMPB_ATTRIB_RAIDCNG\n"); | |
1763 | not_supported ^= MPB_ATTRIB_RAIDCNG; | |
1764 | } | |
1765 | if (not_supported & MPB_ATTRIB_BBM) { | |
1766 | dprintf("\t\tMPB_ATTRIB_BBM\n"); | |
1767 | not_supported ^= MPB_ATTRIB_BBM; | |
1768 | } | |
1769 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { | |
1770 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY (== MPB_ATTRIB_LEGACY)\n"); | |
1771 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
1772 | } | |
1773 | if (not_supported & MPB_ATTRIB_EXP_STRIPE_SIZE) { | |
1774 | dprintf("\t\tMPB_ATTRIB_EXP_STRIP_SIZE\n"); | |
1775 | not_supported ^= MPB_ATTRIB_EXP_STRIPE_SIZE; | |
1776 | } | |
1777 | if (not_supported & MPB_ATTRIB_2TB_DISK) { | |
1778 | dprintf("\t\tMPB_ATTRIB_2TB_DISK\n"); | |
1779 | not_supported ^= MPB_ATTRIB_2TB_DISK; | |
1780 | } | |
1781 | if (not_supported & MPB_ATTRIB_NEVER_USE2) { | |
1782 | dprintf("\t\tMPB_ATTRIB_NEVER_USE2\n"); | |
1783 | not_supported ^= MPB_ATTRIB_NEVER_USE2; | |
1784 | } | |
1785 | if (not_supported & MPB_ATTRIB_NEVER_USE) { | |
1786 | dprintf("\t\tMPB_ATTRIB_NEVER_USE\n"); | |
1787 | not_supported ^= MPB_ATTRIB_NEVER_USE; | |
1788 | } | |
1789 | ||
1790 | if (not_supported) | |
1ade5cc1 | 1791 | dprintf("(IMSM): Unknown attributes : %x\n", not_supported); |
19482bcc AK |
1792 | |
1793 | ret_val = 0; | |
1794 | } | |
1795 | ||
1796 | return ret_val; | |
1797 | } | |
1798 | ||
9e2d750d | 1799 | #ifndef MDASSEMBLE |
a5d85af7 | 1800 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map); |
44470971 | 1801 | |
cdddbdbc DW |
1802 | static void examine_super_imsm(struct supertype *st, char *homehost) |
1803 | { | |
1804 | struct intel_super *super = st->sb; | |
949c47a0 | 1805 | struct imsm_super *mpb = super->anchor; |
cdddbdbc DW |
1806 | char str[MAX_SIGNATURE_LENGTH]; |
1807 | int i; | |
27fd6274 DW |
1808 | struct mdinfo info; |
1809 | char nbuf[64]; | |
cdddbdbc | 1810 | __u32 sum; |
14e8215b | 1811 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
94827db3 | 1812 | struct dl *dl; |
27fd6274 | 1813 | |
cdddbdbc DW |
1814 | snprintf(str, MPB_SIG_LEN, "%s", mpb->sig); |
1815 | printf(" Magic : %s\n", str); | |
1816 | snprintf(str, strlen(MPB_VERSION_RAID0), "%s", get_imsm_version(mpb)); | |
1817 | printf(" Version : %s\n", get_imsm_version(mpb)); | |
148acb7b | 1818 | printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num)); |
cdddbdbc DW |
1819 | printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num)); |
1820 | printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num)); | |
19482bcc AK |
1821 | printf(" Attributes : "); |
1822 | if (imsm_check_attributes(mpb->attributes)) | |
1823 | printf("All supported\n"); | |
1824 | else | |
1825 | printf("not supported\n"); | |
a5d85af7 | 1826 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1827 | fname_from_uuid(st, &info, nbuf, ':'); |
27fd6274 | 1828 | printf(" UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
1829 | sum = __le32_to_cpu(mpb->check_sum); |
1830 | printf(" Checksum : %08x %s\n", sum, | |
949c47a0 | 1831 | __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect"); |
f36a9ecd | 1832 | printf(" MPB Sectors : %d\n", mpb_sectors(mpb, super->sector_size)); |
cdddbdbc DW |
1833 | printf(" Disks : %d\n", mpb->num_disks); |
1834 | printf(" RAID Devices : %d\n", mpb->num_raid_devs); | |
ef5c214e MK |
1835 | print_imsm_disk(__get_imsm_disk(mpb, super->disks->index), |
1836 | super->disks->index, reserved, super->sector_size); | |
8d67477f | 1837 | if (get_imsm_bbm_log_size(super->bbm_log)) { |
604b746f JD |
1838 | struct bbm_log *log = super->bbm_log; |
1839 | ||
1840 | printf("\n"); | |
1841 | printf("Bad Block Management Log:\n"); | |
1842 | printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size)); | |
1843 | printf(" Signature : %x\n", __le32_to_cpu(log->signature)); | |
1844 | printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count)); | |
604b746f | 1845 | } |
44470971 DW |
1846 | for (i = 0; i < mpb->num_raid_devs; i++) { |
1847 | struct mdinfo info; | |
1848 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1849 | ||
1850 | super->current_vol = i; | |
a5d85af7 | 1851 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1852 | fname_from_uuid(st, &info, nbuf, ':'); |
c47b0ff6 | 1853 | print_imsm_dev(super, dev, nbuf + 5, super->disks->index); |
44470971 | 1854 | } |
cdddbdbc DW |
1855 | for (i = 0; i < mpb->num_disks; i++) { |
1856 | if (i == super->disks->index) | |
1857 | continue; | |
ef5c214e MK |
1858 | print_imsm_disk(__get_imsm_disk(mpb, i), i, reserved, |
1859 | super->sector_size); | |
cdddbdbc | 1860 | } |
94827db3 | 1861 | |
0ec1f4e8 DW |
1862 | for (dl = super->disks; dl; dl = dl->next) |
1863 | if (dl->index == -1) | |
ef5c214e MK |
1864 | print_imsm_disk(&dl->disk, -1, reserved, |
1865 | super->sector_size); | |
520e69e2 AK |
1866 | |
1867 | examine_migr_rec_imsm(super); | |
cdddbdbc DW |
1868 | } |
1869 | ||
061f2c6a | 1870 | static void brief_examine_super_imsm(struct supertype *st, int verbose) |
cdddbdbc | 1871 | { |
27fd6274 | 1872 | /* We just write a generic IMSM ARRAY entry */ |
ff54de6e N |
1873 | struct mdinfo info; |
1874 | char nbuf[64]; | |
1e7bc0ed | 1875 | struct intel_super *super = st->sb; |
1e7bc0ed | 1876 | |
0d5a423f DW |
1877 | if (!super->anchor->num_raid_devs) { |
1878 | printf("ARRAY metadata=imsm\n"); | |
1e7bc0ed | 1879 | return; |
0d5a423f | 1880 | } |
ff54de6e | 1881 | |
a5d85af7 | 1882 | getinfo_super_imsm(st, &info, NULL); |
4737ae25 N |
1883 | fname_from_uuid(st, &info, nbuf, ':'); |
1884 | printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5); | |
1885 | } | |
1886 | ||
1887 | static void brief_examine_subarrays_imsm(struct supertype *st, int verbose) | |
1888 | { | |
1889 | /* We just write a generic IMSM ARRAY entry */ | |
1890 | struct mdinfo info; | |
1891 | char nbuf[64]; | |
1892 | char nbuf1[64]; | |
1893 | struct intel_super *super = st->sb; | |
1894 | int i; | |
1895 | ||
1896 | if (!super->anchor->num_raid_devs) | |
1897 | return; | |
1898 | ||
a5d85af7 | 1899 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1900 | fname_from_uuid(st, &info, nbuf, ':'); |
1e7bc0ed DW |
1901 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
1902 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
1903 | ||
1904 | super->current_vol = i; | |
a5d85af7 | 1905 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1906 | fname_from_uuid(st, &info, nbuf1, ':'); |
1124b3cf | 1907 | printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n", |
cf8de691 | 1908 | dev->volume, nbuf + 5, i, nbuf1 + 5); |
1e7bc0ed | 1909 | } |
cdddbdbc DW |
1910 | } |
1911 | ||
9d84c8ea DW |
1912 | static void export_examine_super_imsm(struct supertype *st) |
1913 | { | |
1914 | struct intel_super *super = st->sb; | |
1915 | struct imsm_super *mpb = super->anchor; | |
1916 | struct mdinfo info; | |
1917 | char nbuf[64]; | |
1918 | ||
a5d85af7 | 1919 | getinfo_super_imsm(st, &info, NULL); |
9d84c8ea DW |
1920 | fname_from_uuid(st, &info, nbuf, ':'); |
1921 | printf("MD_METADATA=imsm\n"); | |
1922 | printf("MD_LEVEL=container\n"); | |
1923 | printf("MD_UUID=%s\n", nbuf+5); | |
1924 | printf("MD_DEVICES=%u\n", mpb->num_disks); | |
1925 | } | |
1926 | ||
74db60b0 N |
1927 | static int copy_metadata_imsm(struct supertype *st, int from, int to) |
1928 | { | |
f36a9ecd | 1929 | /* The second last sector of the device contains |
74db60b0 N |
1930 | * the "struct imsm_super" metadata. |
1931 | * This contains mpb_size which is the size in bytes of the | |
1932 | * extended metadata. This is located immediately before | |
1933 | * the imsm_super. | |
1934 | * We want to read all that, plus the last sector which | |
1935 | * may contain a migration record, and write it all | |
1936 | * to the target. | |
1937 | */ | |
1938 | void *buf; | |
1939 | unsigned long long dsize, offset; | |
1940 | int sectors; | |
1941 | struct imsm_super *sb; | |
f36a9ecd PB |
1942 | struct intel_super *super = st->sb; |
1943 | unsigned int sector_size = super->sector_size; | |
1944 | unsigned int written = 0; | |
74db60b0 | 1945 | |
de44e46f | 1946 | if (posix_memalign(&buf, MAX_SECTOR_SIZE, MAX_SECTOR_SIZE) != 0) |
74db60b0 N |
1947 | return 1; |
1948 | ||
1949 | if (!get_dev_size(from, NULL, &dsize)) | |
1950 | goto err; | |
1951 | ||
f36a9ecd | 1952 | if (lseek64(from, dsize-(2*sector_size), 0) < 0) |
74db60b0 | 1953 | goto err; |
466070ad | 1954 | if ((unsigned int)read(from, buf, sector_size) != sector_size) |
74db60b0 N |
1955 | goto err; |
1956 | sb = buf; | |
1957 | if (strncmp((char*)sb->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) | |
1958 | goto err; | |
1959 | ||
f36a9ecd PB |
1960 | sectors = mpb_sectors(sb, sector_size) + 2; |
1961 | offset = dsize - sectors * sector_size; | |
74db60b0 N |
1962 | if (lseek64(from, offset, 0) < 0 || |
1963 | lseek64(to, offset, 0) < 0) | |
1964 | goto err; | |
f36a9ecd PB |
1965 | while (written < sectors * sector_size) { |
1966 | int n = sectors*sector_size - written; | |
74db60b0 N |
1967 | if (n > 4096) |
1968 | n = 4096; | |
1969 | if (read(from, buf, n) != n) | |
1970 | goto err; | |
1971 | if (write(to, buf, n) != n) | |
1972 | goto err; | |
1973 | written += n; | |
1974 | } | |
1975 | free(buf); | |
1976 | return 0; | |
1977 | err: | |
1978 | free(buf); | |
1979 | return 1; | |
1980 | } | |
1981 | ||
cdddbdbc DW |
1982 | static void detail_super_imsm(struct supertype *st, char *homehost) |
1983 | { | |
3ebe00a1 DW |
1984 | struct mdinfo info; |
1985 | char nbuf[64]; | |
1986 | ||
a5d85af7 | 1987 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1988 | fname_from_uuid(st, &info, nbuf, ':'); |
3ebe00a1 | 1989 | printf("\n UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
1990 | } |
1991 | ||
1992 | static void brief_detail_super_imsm(struct supertype *st) | |
1993 | { | |
ff54de6e N |
1994 | struct mdinfo info; |
1995 | char nbuf[64]; | |
a5d85af7 | 1996 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1997 | fname_from_uuid(st, &info, nbuf, ':'); |
ff54de6e | 1998 | printf(" UUID=%s", nbuf + 5); |
cdddbdbc | 1999 | } |
d665cc31 DW |
2000 | |
2001 | static int imsm_read_serial(int fd, char *devname, __u8 *serial); | |
2002 | static void fd2devname(int fd, char *name); | |
2003 | ||
120dc887 | 2004 | static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose) |
d665cc31 | 2005 | { |
120dc887 LM |
2006 | /* dump an unsorted list of devices attached to AHCI Intel storage |
2007 | * controller, as well as non-connected ports | |
d665cc31 DW |
2008 | */ |
2009 | int hba_len = strlen(hba_path) + 1; | |
2010 | struct dirent *ent; | |
2011 | DIR *dir; | |
2012 | char *path = NULL; | |
2013 | int err = 0; | |
2014 | unsigned long port_mask = (1 << port_count) - 1; | |
2015 | ||
f21e18ca | 2016 | if (port_count > (int)sizeof(port_mask) * 8) { |
ba728be7 | 2017 | if (verbose > 0) |
e7b84f9d | 2018 | pr_err("port_count %d out of range\n", port_count); |
d665cc31 DW |
2019 | return 2; |
2020 | } | |
2021 | ||
2022 | /* scroll through /sys/dev/block looking for devices attached to | |
2023 | * this hba | |
2024 | */ | |
2025 | dir = opendir("/sys/dev/block"); | |
1a6dd6b9 PB |
2026 | if (!dir) |
2027 | return 1; | |
2028 | ||
2029 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
d665cc31 DW |
2030 | int fd; |
2031 | char model[64]; | |
2032 | char vendor[64]; | |
2033 | char buf[1024]; | |
2034 | int major, minor; | |
2035 | char *device; | |
2036 | char *c; | |
2037 | int port; | |
2038 | int type; | |
2039 | ||
2040 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
2041 | continue; | |
2042 | path = devt_to_devpath(makedev(major, minor)); | |
2043 | if (!path) | |
2044 | continue; | |
2045 | if (!path_attached_to_hba(path, hba_path)) { | |
2046 | free(path); | |
2047 | path = NULL; | |
2048 | continue; | |
2049 | } | |
2050 | ||
2051 | /* retrieve the scsi device type */ | |
2052 | if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) { | |
ba728be7 | 2053 | if (verbose > 0) |
e7b84f9d | 2054 | pr_err("failed to allocate 'device'\n"); |
d665cc31 DW |
2055 | err = 2; |
2056 | break; | |
2057 | } | |
2058 | sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor); | |
193b6c0b | 2059 | if (load_sys(device, buf, sizeof(buf)) != 0) { |
ba728be7 | 2060 | if (verbose > 0) |
e7b84f9d | 2061 | pr_err("failed to read device type for %s\n", |
d665cc31 DW |
2062 | path); |
2063 | err = 2; | |
2064 | free(device); | |
2065 | break; | |
2066 | } | |
2067 | type = strtoul(buf, NULL, 10); | |
2068 | ||
2069 | /* if it's not a disk print the vendor and model */ | |
2070 | if (!(type == 0 || type == 7 || type == 14)) { | |
2071 | vendor[0] = '\0'; | |
2072 | model[0] = '\0'; | |
2073 | sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor); | |
193b6c0b | 2074 | if (load_sys(device, buf, sizeof(buf)) == 0) { |
d665cc31 DW |
2075 | strncpy(vendor, buf, sizeof(vendor)); |
2076 | vendor[sizeof(vendor) - 1] = '\0'; | |
2077 | c = (char *) &vendor[sizeof(vendor) - 1]; | |
2078 | while (isspace(*c) || *c == '\0') | |
2079 | *c-- = '\0'; | |
2080 | ||
2081 | } | |
2082 | sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor); | |
193b6c0b | 2083 | if (load_sys(device, buf, sizeof(buf)) == 0) { |
d665cc31 DW |
2084 | strncpy(model, buf, sizeof(model)); |
2085 | model[sizeof(model) - 1] = '\0'; | |
2086 | c = (char *) &model[sizeof(model) - 1]; | |
2087 | while (isspace(*c) || *c == '\0') | |
2088 | *c-- = '\0'; | |
2089 | } | |
2090 | ||
2091 | if (vendor[0] && model[0]) | |
2092 | sprintf(buf, "%.64s %.64s", vendor, model); | |
2093 | else | |
2094 | switch (type) { /* numbers from hald/linux/device.c */ | |
2095 | case 1: sprintf(buf, "tape"); break; | |
2096 | case 2: sprintf(buf, "printer"); break; | |
2097 | case 3: sprintf(buf, "processor"); break; | |
2098 | case 4: | |
2099 | case 5: sprintf(buf, "cdrom"); break; | |
2100 | case 6: sprintf(buf, "scanner"); break; | |
2101 | case 8: sprintf(buf, "media_changer"); break; | |
2102 | case 9: sprintf(buf, "comm"); break; | |
2103 | case 12: sprintf(buf, "raid"); break; | |
2104 | default: sprintf(buf, "unknown"); | |
2105 | } | |
2106 | } else | |
2107 | buf[0] = '\0'; | |
2108 | free(device); | |
2109 | ||
2110 | /* chop device path to 'host%d' and calculate the port number */ | |
2111 | c = strchr(&path[hba_len], '/'); | |
4e5e717d | 2112 | if (!c) { |
ba728be7 | 2113 | if (verbose > 0) |
e7b84f9d | 2114 | pr_err("%s - invalid path name\n", path + hba_len); |
4e5e717d AW |
2115 | err = 2; |
2116 | break; | |
2117 | } | |
d665cc31 | 2118 | *c = '\0'; |
0858eccf AP |
2119 | if ((sscanf(&path[hba_len], "ata%d", &port) == 1) || |
2120 | ((sscanf(&path[hba_len], "host%d", &port) == 1))) | |
d665cc31 DW |
2121 | port -= host_base; |
2122 | else { | |
ba728be7 | 2123 | if (verbose > 0) { |
d665cc31 | 2124 | *c = '/'; /* repair the full string */ |
e7b84f9d | 2125 | pr_err("failed to determine port number for %s\n", |
d665cc31 DW |
2126 | path); |
2127 | } | |
2128 | err = 2; | |
2129 | break; | |
2130 | } | |
2131 | ||
2132 | /* mark this port as used */ | |
2133 | port_mask &= ~(1 << port); | |
2134 | ||
2135 | /* print out the device information */ | |
2136 | if (buf[0]) { | |
2137 | printf(" Port%d : - non-disk device (%s) -\n", port, buf); | |
2138 | continue; | |
2139 | } | |
2140 | ||
2141 | fd = dev_open(ent->d_name, O_RDONLY); | |
2142 | if (fd < 0) | |
2143 | printf(" Port%d : - disk info unavailable -\n", port); | |
2144 | else { | |
2145 | fd2devname(fd, buf); | |
2146 | printf(" Port%d : %s", port, buf); | |
2147 | if (imsm_read_serial(fd, NULL, (__u8 *) buf) == 0) | |
664d5325 | 2148 | printf(" (%.*s)\n", MAX_RAID_SERIAL_LEN, buf); |
d665cc31 | 2149 | else |
664d5325 | 2150 | printf(" ()\n"); |
4dab422a | 2151 | close(fd); |
d665cc31 | 2152 | } |
d665cc31 DW |
2153 | free(path); |
2154 | path = NULL; | |
2155 | } | |
2156 | if (path) | |
2157 | free(path); | |
2158 | if (dir) | |
2159 | closedir(dir); | |
2160 | if (err == 0) { | |
2161 | int i; | |
2162 | ||
2163 | for (i = 0; i < port_count; i++) | |
2164 | if (port_mask & (1 << i)) | |
2165 | printf(" Port%d : - no device attached -\n", i); | |
2166 | } | |
2167 | ||
2168 | return err; | |
2169 | } | |
2170 | ||
b5eece69 | 2171 | static int print_vmd_attached_devs(struct sys_dev *hba) |
60f0f54d PB |
2172 | { |
2173 | struct dirent *ent; | |
2174 | DIR *dir; | |
2175 | char path[292]; | |
2176 | char link[256]; | |
2177 | char *c, *rp; | |
2178 | ||
2179 | if (hba->type != SYS_DEV_VMD) | |
b5eece69 | 2180 | return 1; |
60f0f54d PB |
2181 | |
2182 | /* scroll through /sys/dev/block looking for devices attached to | |
2183 | * this hba | |
2184 | */ | |
2185 | dir = opendir("/sys/bus/pci/drivers/nvme"); | |
b9135011 | 2186 | if (!dir) |
b5eece69 | 2187 | return 1; |
b9135011 JS |
2188 | |
2189 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
60f0f54d PB |
2190 | int n; |
2191 | ||
2192 | /* is 'ent' a device? check that the 'subsystem' link exists and | |
2193 | * that its target matches 'bus' | |
2194 | */ | |
2195 | sprintf(path, "/sys/bus/pci/drivers/nvme/%s/subsystem", | |
2196 | ent->d_name); | |
2197 | n = readlink(path, link, sizeof(link)); | |
2198 | if (n < 0 || n >= (int)sizeof(link)) | |
2199 | continue; | |
2200 | link[n] = '\0'; | |
2201 | c = strrchr(link, '/'); | |
2202 | if (!c) | |
2203 | continue; | |
2204 | if (strncmp("pci", c+1, strlen("pci")) != 0) | |
2205 | continue; | |
2206 | ||
2207 | sprintf(path, "/sys/bus/pci/drivers/nvme/%s", ent->d_name); | |
60f0f54d PB |
2208 | |
2209 | rp = realpath(path, NULL); | |
2210 | if (!rp) | |
2211 | continue; | |
2212 | ||
2213 | if (path_attached_to_hba(rp, hba->path)) { | |
2214 | printf(" NVMe under VMD : %s\n", rp); | |
2215 | } | |
2216 | free(rp); | |
2217 | } | |
2218 | ||
b9135011 | 2219 | closedir(dir); |
b5eece69 | 2220 | return 0; |
60f0f54d PB |
2221 | } |
2222 | ||
120dc887 LM |
2223 | static void print_found_intel_controllers(struct sys_dev *elem) |
2224 | { | |
2225 | for (; elem; elem = elem->next) { | |
e7b84f9d | 2226 | pr_err("found Intel(R) "); |
120dc887 LM |
2227 | if (elem->type == SYS_DEV_SATA) |
2228 | fprintf(stderr, "SATA "); | |
155cbb4c LM |
2229 | else if (elem->type == SYS_DEV_SAS) |
2230 | fprintf(stderr, "SAS "); | |
0858eccf AP |
2231 | else if (elem->type == SYS_DEV_NVME) |
2232 | fprintf(stderr, "NVMe "); | |
60f0f54d PB |
2233 | |
2234 | if (elem->type == SYS_DEV_VMD) | |
2235 | fprintf(stderr, "VMD domain"); | |
2236 | else | |
2237 | fprintf(stderr, "RAID controller"); | |
2238 | ||
120dc887 LM |
2239 | if (elem->pci_id) |
2240 | fprintf(stderr, " at %s", elem->pci_id); | |
2241 | fprintf(stderr, ".\n"); | |
2242 | } | |
2243 | fflush(stderr); | |
2244 | } | |
2245 | ||
120dc887 LM |
2246 | static int ahci_get_port_count(const char *hba_path, int *port_count) |
2247 | { | |
2248 | struct dirent *ent; | |
2249 | DIR *dir; | |
2250 | int host_base = -1; | |
2251 | ||
2252 | *port_count = 0; | |
2253 | if ((dir = opendir(hba_path)) == NULL) | |
2254 | return -1; | |
2255 | ||
2256 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
2257 | int host; | |
2258 | ||
0858eccf AP |
2259 | if ((sscanf(ent->d_name, "ata%d", &host) != 1) && |
2260 | ((sscanf(ent->d_name, "host%d", &host) != 1))) | |
120dc887 LM |
2261 | continue; |
2262 | if (*port_count == 0) | |
2263 | host_base = host; | |
2264 | else if (host < host_base) | |
2265 | host_base = host; | |
2266 | ||
2267 | if (host + 1 > *port_count + host_base) | |
2268 | *port_count = host + 1 - host_base; | |
2269 | } | |
2270 | closedir(dir); | |
2271 | return host_base; | |
2272 | } | |
2273 | ||
a891a3c2 LM |
2274 | static void print_imsm_capability(const struct imsm_orom *orom) |
2275 | { | |
0858eccf AP |
2276 | printf(" Platform : Intel(R) "); |
2277 | if (orom->capabilities == 0 && orom->driver_features == 0) | |
2278 | printf("Matrix Storage Manager\n"); | |
2279 | else | |
2280 | printf("Rapid Storage Technology%s\n", | |
2281 | imsm_orom_is_enterprise(orom) ? " enterprise" : ""); | |
2282 | if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build) | |
2283 | printf(" Version : %d.%d.%d.%d\n", orom->major_ver, | |
2284 | orom->minor_ver, orom->hotfix_ver, orom->build); | |
a891a3c2 LM |
2285 | printf(" RAID Levels :%s%s%s%s%s\n", |
2286 | imsm_orom_has_raid0(orom) ? " raid0" : "", | |
2287 | imsm_orom_has_raid1(orom) ? " raid1" : "", | |
2288 | imsm_orom_has_raid1e(orom) ? " raid1e" : "", | |
2289 | imsm_orom_has_raid10(orom) ? " raid10" : "", | |
2290 | imsm_orom_has_raid5(orom) ? " raid5" : ""); | |
2291 | printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", | |
2292 | imsm_orom_has_chunk(orom, 2) ? " 2k" : "", | |
2293 | imsm_orom_has_chunk(orom, 4) ? " 4k" : "", | |
2294 | imsm_orom_has_chunk(orom, 8) ? " 8k" : "", | |
2295 | imsm_orom_has_chunk(orom, 16) ? " 16k" : "", | |
2296 | imsm_orom_has_chunk(orom, 32) ? " 32k" : "", | |
2297 | imsm_orom_has_chunk(orom, 64) ? " 64k" : "", | |
2298 | imsm_orom_has_chunk(orom, 128) ? " 128k" : "", | |
2299 | imsm_orom_has_chunk(orom, 256) ? " 256k" : "", | |
2300 | imsm_orom_has_chunk(orom, 512) ? " 512k" : "", | |
2301 | imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "", | |
2302 | imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "", | |
2303 | imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "", | |
2304 | imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "", | |
2305 | imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "", | |
2306 | imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "", | |
2307 | imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : ""); | |
29cd0821 CA |
2308 | printf(" 2TB volumes :%s supported\n", |
2309 | (orom->attr & IMSM_OROM_ATTR_2TB)?"":" not"); | |
2310 | printf(" 2TB disks :%s supported\n", | |
2311 | (orom->attr & IMSM_OROM_ATTR_2TB_DISK)?"":" not"); | |
0e7f69a8 | 2312 | printf(" Max Disks : %d\n", orom->tds); |
0858eccf AP |
2313 | printf(" Max Volumes : %d per array, %d per %s\n", |
2314 | orom->vpa, orom->vphba, | |
2315 | imsm_orom_is_nvme(orom) ? "platform" : "controller"); | |
a891a3c2 LM |
2316 | return; |
2317 | } | |
2318 | ||
e50cf220 MN |
2319 | static void print_imsm_capability_export(const struct imsm_orom *orom) |
2320 | { | |
2321 | printf("MD_FIRMWARE_TYPE=imsm\n"); | |
0858eccf AP |
2322 | if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build) |
2323 | printf("IMSM_VERSION=%d.%d.%d.%d\n", orom->major_ver, orom->minor_ver, | |
2324 | orom->hotfix_ver, orom->build); | |
e50cf220 MN |
2325 | printf("IMSM_SUPPORTED_RAID_LEVELS=%s%s%s%s%s\n", |
2326 | imsm_orom_has_raid0(orom) ? "raid0 " : "", | |
2327 | imsm_orom_has_raid1(orom) ? "raid1 " : "", | |
2328 | imsm_orom_has_raid1e(orom) ? "raid1e " : "", | |
2329 | imsm_orom_has_raid5(orom) ? "raid10 " : "", | |
2330 | imsm_orom_has_raid10(orom) ? "raid5 " : ""); | |
2331 | printf("IMSM_SUPPORTED_CHUNK_SIZES=%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", | |
2332 | imsm_orom_has_chunk(orom, 2) ? "2k " : "", | |
2333 | imsm_orom_has_chunk(orom, 4) ? "4k " : "", | |
2334 | imsm_orom_has_chunk(orom, 8) ? "8k " : "", | |
2335 | imsm_orom_has_chunk(orom, 16) ? "16k " : "", | |
2336 | imsm_orom_has_chunk(orom, 32) ? "32k " : "", | |
2337 | imsm_orom_has_chunk(orom, 64) ? "64k " : "", | |
2338 | imsm_orom_has_chunk(orom, 128) ? "128k " : "", | |
2339 | imsm_orom_has_chunk(orom, 256) ? "256k " : "", | |
2340 | imsm_orom_has_chunk(orom, 512) ? "512k " : "", | |
2341 | imsm_orom_has_chunk(orom, 1024*1) ? "1M " : "", | |
2342 | imsm_orom_has_chunk(orom, 1024*2) ? "2M " : "", | |
2343 | imsm_orom_has_chunk(orom, 1024*4) ? "4M " : "", | |
2344 | imsm_orom_has_chunk(orom, 1024*8) ? "8M " : "", | |
2345 | imsm_orom_has_chunk(orom, 1024*16) ? "16M " : "", | |
2346 | imsm_orom_has_chunk(orom, 1024*32) ? "32M " : "", | |
2347 | imsm_orom_has_chunk(orom, 1024*64) ? "64M " : ""); | |
2348 | printf("IMSM_2TB_VOLUMES=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB) ? "yes" : "no"); | |
2349 | printf("IMSM_2TB_DISKS=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB_DISK) ? "yes" : "no"); | |
2350 | printf("IMSM_MAX_DISKS=%d\n",orom->tds); | |
2351 | printf("IMSM_MAX_VOLUMES_PER_ARRAY=%d\n",orom->vpa); | |
2352 | printf("IMSM_MAX_VOLUMES_PER_CONTROLLER=%d\n",orom->vphba); | |
2353 | } | |
2354 | ||
9eafa1de | 2355 | static int detail_platform_imsm(int verbose, int enumerate_only, char *controller_path) |
d665cc31 DW |
2356 | { |
2357 | /* There are two components to imsm platform support, the ahci SATA | |
2358 | * controller and the option-rom. To find the SATA controller we | |
2359 | * simply look in /sys/bus/pci/drivers/ahci to see if an ahci | |
2360 | * controller with the Intel vendor id is present. This approach | |
2361 | * allows mdadm to leverage the kernel's ahci detection logic, with the | |
2362 | * caveat that if ahci.ko is not loaded mdadm will not be able to | |
2363 | * detect platform raid capabilities. The option-rom resides in a | |
2364 | * platform "Adapter ROM". We scan for its signature to retrieve the | |
2365 | * platform capabilities. If raid support is disabled in the BIOS the | |
2366 | * option-rom capability structure will not be available. | |
2367 | */ | |
d665cc31 | 2368 | struct sys_dev *list, *hba; |
d665cc31 DW |
2369 | int host_base = 0; |
2370 | int port_count = 0; | |
9eafa1de | 2371 | int result=1; |
d665cc31 | 2372 | |
5615172f | 2373 | if (enumerate_only) { |
a891a3c2 | 2374 | if (check_env("IMSM_NO_PLATFORM")) |
5615172f | 2375 | return 0; |
a891a3c2 LM |
2376 | list = find_intel_devices(); |
2377 | if (!list) | |
2378 | return 2; | |
2379 | for (hba = list; hba; hba = hba->next) { | |
6b781d33 AP |
2380 | if (find_imsm_capability(hba)) { |
2381 | result = 0; | |
a891a3c2 LM |
2382 | break; |
2383 | } | |
9eafa1de | 2384 | else |
6b781d33 | 2385 | result = 2; |
a891a3c2 | 2386 | } |
a891a3c2 | 2387 | return result; |
5615172f DW |
2388 | } |
2389 | ||
155cbb4c LM |
2390 | list = find_intel_devices(); |
2391 | if (!list) { | |
ba728be7 | 2392 | if (verbose > 0) |
7a862a02 | 2393 | pr_err("no active Intel(R) RAID controller found.\n"); |
d665cc31 | 2394 | return 2; |
ba728be7 | 2395 | } else if (verbose > 0) |
155cbb4c | 2396 | print_found_intel_controllers(list); |
d665cc31 | 2397 | |
a891a3c2 | 2398 | for (hba = list; hba; hba = hba->next) { |
0858eccf | 2399 | if (controller_path && (compare_paths(hba->path, controller_path) != 0)) |
9eafa1de | 2400 | continue; |
0858eccf | 2401 | if (!find_imsm_capability(hba)) { |
60f0f54d | 2402 | char buf[PATH_MAX]; |
e7b84f9d | 2403 | pr_err("imsm capabilities not found for controller: %s (type %s)\n", |
60f0f54d PB |
2404 | hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path, |
2405 | get_sys_dev_type(hba->type)); | |
0858eccf AP |
2406 | continue; |
2407 | } | |
2408 | result = 0; | |
2409 | } | |
2410 | ||
2411 | if (controller_path && result == 1) { | |
2412 | pr_err("no active Intel(R) RAID controller found under %s\n", | |
2413 | controller_path); | |
2414 | return result; | |
2415 | } | |
2416 | ||
5e1d6128 | 2417 | const struct orom_entry *entry; |
0858eccf | 2418 | |
5e1d6128 | 2419 | for (entry = orom_entries; entry; entry = entry->next) { |
60f0f54d | 2420 | if (entry->type == SYS_DEV_VMD) { |
07cb1e57 | 2421 | print_imsm_capability(&entry->orom); |
32716c51 PB |
2422 | printf(" 3rd party NVMe :%s supported\n", |
2423 | imsm_orom_has_tpv_support(&entry->orom)?"":" not"); | |
60f0f54d PB |
2424 | for (hba = list; hba; hba = hba->next) { |
2425 | if (hba->type == SYS_DEV_VMD) { | |
2426 | char buf[PATH_MAX]; | |
60f0f54d PB |
2427 | printf(" I/O Controller : %s (%s)\n", |
2428 | vmd_domain_to_controller(hba, buf), get_sys_dev_type(hba->type)); | |
b5eece69 PB |
2429 | if (print_vmd_attached_devs(hba)) { |
2430 | if (verbose > 0) | |
2431 | pr_err("failed to get devices attached to VMD domain.\n"); | |
2432 | result |= 2; | |
2433 | } | |
60f0f54d PB |
2434 | } |
2435 | } | |
07cb1e57 | 2436 | printf("\n"); |
60f0f54d PB |
2437 | continue; |
2438 | } | |
0858eccf | 2439 | |
60f0f54d PB |
2440 | print_imsm_capability(&entry->orom); |
2441 | if (entry->type == SYS_DEV_NVME) { | |
0858eccf AP |
2442 | for (hba = list; hba; hba = hba->next) { |
2443 | if (hba->type == SYS_DEV_NVME) | |
2444 | printf(" NVMe Device : %s\n", hba->path); | |
2445 | } | |
60f0f54d | 2446 | printf("\n"); |
0858eccf AP |
2447 | continue; |
2448 | } | |
2449 | ||
2450 | struct devid_list *devid; | |
5e1d6128 | 2451 | for (devid = entry->devid_list; devid; devid = devid->next) { |
0858eccf AP |
2452 | hba = device_by_id(devid->devid); |
2453 | if (!hba) | |
2454 | continue; | |
2455 | ||
9eafa1de MN |
2456 | printf(" I/O Controller : %s (%s)\n", |
2457 | hba->path, get_sys_dev_type(hba->type)); | |
2458 | if (hba->type == SYS_DEV_SATA) { | |
2459 | host_base = ahci_get_port_count(hba->path, &port_count); | |
2460 | if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) { | |
2461 | if (verbose > 0) | |
7a862a02 | 2462 | pr_err("failed to enumerate ports on SATA controller at %s.\n", hba->pci_id); |
9eafa1de MN |
2463 | result |= 2; |
2464 | } | |
120dc887 LM |
2465 | } |
2466 | } | |
0858eccf | 2467 | printf("\n"); |
d665cc31 | 2468 | } |
155cbb4c | 2469 | |
120dc887 | 2470 | return result; |
d665cc31 | 2471 | } |
e50cf220 | 2472 | |
9eafa1de | 2473 | static int export_detail_platform_imsm(int verbose, char *controller_path) |
e50cf220 | 2474 | { |
e50cf220 MN |
2475 | struct sys_dev *list, *hba; |
2476 | int result=1; | |
2477 | ||
2478 | list = find_intel_devices(); | |
2479 | if (!list) { | |
2480 | if (verbose > 0) | |
2481 | pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_INTEL_DEVICES\n"); | |
2482 | result = 2; | |
e50cf220 MN |
2483 | return result; |
2484 | } | |
2485 | ||
2486 | for (hba = list; hba; hba = hba->next) { | |
9eafa1de MN |
2487 | if (controller_path && (compare_paths(hba->path,controller_path) != 0)) |
2488 | continue; | |
60f0f54d PB |
2489 | if (!find_imsm_capability(hba) && verbose > 0) { |
2490 | char buf[PATH_MAX]; | |
2491 | pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_IMSM_CAPABLE_DEVICE_UNDER_%s\n", | |
2492 | hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path); | |
2493 | } | |
0858eccf | 2494 | else |
e50cf220 | 2495 | result = 0; |
e50cf220 MN |
2496 | } |
2497 | ||
5e1d6128 | 2498 | const struct orom_entry *entry; |
0858eccf | 2499 | |
60f0f54d PB |
2500 | for (entry = orom_entries; entry; entry = entry->next) { |
2501 | if (entry->type == SYS_DEV_VMD) { | |
2502 | for (hba = list; hba; hba = hba->next) | |
2503 | print_imsm_capability_export(&entry->orom); | |
2504 | continue; | |
2505 | } | |
5e1d6128 | 2506 | print_imsm_capability_export(&entry->orom); |
60f0f54d | 2507 | } |
0858eccf | 2508 | |
e50cf220 MN |
2509 | return result; |
2510 | } | |
2511 | ||
cdddbdbc DW |
2512 | #endif |
2513 | ||
2514 | static int match_home_imsm(struct supertype *st, char *homehost) | |
2515 | { | |
5115ca67 DW |
2516 | /* the imsm metadata format does not specify any host |
2517 | * identification information. We return -1 since we can never | |
2518 | * confirm nor deny whether a given array is "meant" for this | |
148acb7b | 2519 | * host. We rely on compare_super and the 'family_num' fields to |
5115ca67 DW |
2520 | * exclude member disks that do not belong, and we rely on |
2521 | * mdadm.conf to specify the arrays that should be assembled. | |
2522 | * Auto-assembly may still pick up "foreign" arrays. | |
2523 | */ | |
cdddbdbc | 2524 | |
9362c1c8 | 2525 | return -1; |
cdddbdbc DW |
2526 | } |
2527 | ||
2528 | static void uuid_from_super_imsm(struct supertype *st, int uuid[4]) | |
2529 | { | |
51006d85 N |
2530 | /* The uuid returned here is used for: |
2531 | * uuid to put into bitmap file (Create, Grow) | |
2532 | * uuid for backup header when saving critical section (Grow) | |
2533 | * comparing uuids when re-adding a device into an array | |
2534 | * In these cases the uuid required is that of the data-array, | |
2535 | * not the device-set. | |
2536 | * uuid to recognise same set when adding a missing device back | |
2537 | * to an array. This is a uuid for the device-set. | |
1011e834 | 2538 | * |
51006d85 N |
2539 | * For each of these we can make do with a truncated |
2540 | * or hashed uuid rather than the original, as long as | |
2541 | * everyone agrees. | |
2542 | * In each case the uuid required is that of the data-array, | |
2543 | * not the device-set. | |
43dad3d6 | 2544 | */ |
51006d85 N |
2545 | /* imsm does not track uuid's so we synthesis one using sha1 on |
2546 | * - The signature (Which is constant for all imsm array, but no matter) | |
148acb7b | 2547 | * - the orig_family_num of the container |
51006d85 N |
2548 | * - the index number of the volume |
2549 | * - the 'serial' number of the volume. | |
2550 | * Hopefully these are all constant. | |
2551 | */ | |
2552 | struct intel_super *super = st->sb; | |
43dad3d6 | 2553 | |
51006d85 N |
2554 | char buf[20]; |
2555 | struct sha1_ctx ctx; | |
2556 | struct imsm_dev *dev = NULL; | |
148acb7b | 2557 | __u32 family_num; |
51006d85 | 2558 | |
148acb7b DW |
2559 | /* some mdadm versions failed to set ->orig_family_num, in which |
2560 | * case fall back to ->family_num. orig_family_num will be | |
2561 | * fixed up with the first metadata update. | |
2562 | */ | |
2563 | family_num = super->anchor->orig_family_num; | |
2564 | if (family_num == 0) | |
2565 | family_num = super->anchor->family_num; | |
51006d85 | 2566 | sha1_init_ctx(&ctx); |
92bd8f8d | 2567 | sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx); |
148acb7b | 2568 | sha1_process_bytes(&family_num, sizeof(__u32), &ctx); |
51006d85 N |
2569 | if (super->current_vol >= 0) |
2570 | dev = get_imsm_dev(super, super->current_vol); | |
2571 | if (dev) { | |
2572 | __u32 vol = super->current_vol; | |
2573 | sha1_process_bytes(&vol, sizeof(vol), &ctx); | |
2574 | sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx); | |
2575 | } | |
2576 | sha1_finish_ctx(&ctx, buf); | |
2577 | memcpy(uuid, buf, 4*4); | |
cdddbdbc DW |
2578 | } |
2579 | ||
0d481d37 | 2580 | #if 0 |
4f5bc454 DW |
2581 | static void |
2582 | get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p) | |
cdddbdbc | 2583 | { |
cdddbdbc DW |
2584 | __u8 *v = get_imsm_version(mpb); |
2585 | __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH; | |
2586 | char major[] = { 0, 0, 0 }; | |
2587 | char minor[] = { 0 ,0, 0 }; | |
2588 | char patch[] = { 0, 0, 0 }; | |
2589 | char *ver_parse[] = { major, minor, patch }; | |
2590 | int i, j; | |
2591 | ||
2592 | i = j = 0; | |
2593 | while (*v != '\0' && v < end) { | |
2594 | if (*v != '.' && j < 2) | |
2595 | ver_parse[i][j++] = *v; | |
2596 | else { | |
2597 | i++; | |
2598 | j = 0; | |
2599 | } | |
2600 | v++; | |
2601 | } | |
2602 | ||
4f5bc454 DW |
2603 | *m = strtol(minor, NULL, 0); |
2604 | *p = strtol(patch, NULL, 0); | |
2605 | } | |
0d481d37 | 2606 | #endif |
4f5bc454 | 2607 | |
1e5c6983 DW |
2608 | static __u32 migr_strip_blocks_resync(struct imsm_dev *dev) |
2609 | { | |
2610 | /* migr_strip_size when repairing or initializing parity */ | |
238c0a71 | 2611 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2612 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2613 | ||
2614 | switch (get_imsm_raid_level(map)) { | |
2615 | case 5: | |
2616 | case 10: | |
2617 | return chunk; | |
2618 | default: | |
2619 | return 128*1024 >> 9; | |
2620 | } | |
2621 | } | |
2622 | ||
2623 | static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev) | |
2624 | { | |
2625 | /* migr_strip_size when rebuilding a degraded disk, no idea why | |
2626 | * this is different than migr_strip_size_resync(), but it's good | |
2627 | * to be compatible | |
2628 | */ | |
238c0a71 | 2629 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2630 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2631 | ||
2632 | switch (get_imsm_raid_level(map)) { | |
2633 | case 1: | |
2634 | case 10: | |
2635 | if (map->num_members % map->num_domains == 0) | |
2636 | return 128*1024 >> 9; | |
2637 | else | |
2638 | return chunk; | |
2639 | case 5: | |
2640 | return max((__u32) 64*1024 >> 9, chunk); | |
2641 | default: | |
2642 | return 128*1024 >> 9; | |
2643 | } | |
2644 | } | |
2645 | ||
2646 | static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev) | |
2647 | { | |
238c0a71 AK |
2648 | struct imsm_map *lo = get_imsm_map(dev, MAP_0); |
2649 | struct imsm_map *hi = get_imsm_map(dev, MAP_1); | |
1e5c6983 DW |
2650 | __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip); |
2651 | __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip); | |
2652 | ||
2653 | return max((__u32) 1, hi_chunk / lo_chunk); | |
2654 | } | |
2655 | ||
2656 | static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev) | |
2657 | { | |
238c0a71 | 2658 | struct imsm_map *lo = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2659 | int level = get_imsm_raid_level(lo); |
2660 | ||
2661 | if (level == 1 || level == 10) { | |
238c0a71 | 2662 | struct imsm_map *hi = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2663 | |
2664 | return hi->num_domains; | |
2665 | } else | |
2666 | return num_stripes_per_unit_resync(dev); | |
2667 | } | |
2668 | ||
98130f40 | 2669 | static __u8 imsm_num_data_members(struct imsm_dev *dev, int second_map) |
1e5c6983 DW |
2670 | { |
2671 | /* named 'imsm_' because raid0, raid1 and raid10 | |
2672 | * counter-intuitively have the same number of data disks | |
2673 | */ | |
98130f40 | 2674 | struct imsm_map *map = get_imsm_map(dev, second_map); |
1e5c6983 DW |
2675 | |
2676 | switch (get_imsm_raid_level(map)) { | |
2677 | case 0: | |
36fd8ccc AK |
2678 | return map->num_members; |
2679 | break; | |
1e5c6983 DW |
2680 | case 1: |
2681 | case 10: | |
36fd8ccc | 2682 | return map->num_members/2; |
1e5c6983 DW |
2683 | case 5: |
2684 | return map->num_members - 1; | |
2685 | default: | |
1ade5cc1 | 2686 | dprintf("unsupported raid level\n"); |
1e5c6983 DW |
2687 | return 0; |
2688 | } | |
2689 | } | |
2690 | ||
2691 | static __u32 parity_segment_depth(struct imsm_dev *dev) | |
2692 | { | |
238c0a71 | 2693 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2694 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2695 | ||
2696 | switch(get_imsm_raid_level(map)) { | |
2697 | case 1: | |
2698 | case 10: | |
2699 | return chunk * map->num_domains; | |
2700 | case 5: | |
2701 | return chunk * map->num_members; | |
2702 | default: | |
2703 | return chunk; | |
2704 | } | |
2705 | } | |
2706 | ||
2707 | static __u32 map_migr_block(struct imsm_dev *dev, __u32 block) | |
2708 | { | |
238c0a71 | 2709 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2710 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2711 | __u32 strip = block / chunk; | |
2712 | ||
2713 | switch (get_imsm_raid_level(map)) { | |
2714 | case 1: | |
2715 | case 10: { | |
2716 | __u32 vol_strip = (strip * map->num_domains) + 1; | |
2717 | __u32 vol_stripe = vol_strip / map->num_members; | |
2718 | ||
2719 | return vol_stripe * chunk + block % chunk; | |
2720 | } case 5: { | |
2721 | __u32 stripe = strip / (map->num_members - 1); | |
2722 | ||
2723 | return stripe * chunk + block % chunk; | |
2724 | } | |
2725 | default: | |
2726 | return 0; | |
2727 | } | |
2728 | } | |
2729 | ||
c47b0ff6 AK |
2730 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
2731 | struct imsm_dev *dev) | |
1e5c6983 DW |
2732 | { |
2733 | /* calculate the conversion factor between per member 'blocks' | |
2734 | * (md/{resync,rebuild}_start) and imsm migration units, return | |
2735 | * 0 for the 'not migrating' and 'unsupported migration' cases | |
2736 | */ | |
2737 | if (!dev->vol.migr_state) | |
2738 | return 0; | |
2739 | ||
2740 | switch (migr_type(dev)) { | |
c47b0ff6 AK |
2741 | case MIGR_GEN_MIGR: { |
2742 | struct migr_record *migr_rec = super->migr_rec; | |
2743 | return __le32_to_cpu(migr_rec->blocks_per_unit); | |
2744 | } | |
1e5c6983 DW |
2745 | case MIGR_VERIFY: |
2746 | case MIGR_REPAIR: | |
2747 | case MIGR_INIT: { | |
238c0a71 | 2748 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2749 | __u32 stripes_per_unit; |
2750 | __u32 blocks_per_unit; | |
2751 | __u32 parity_depth; | |
2752 | __u32 migr_chunk; | |
2753 | __u32 block_map; | |
2754 | __u32 block_rel; | |
2755 | __u32 segment; | |
2756 | __u32 stripe; | |
2757 | __u8 disks; | |
2758 | ||
2759 | /* yes, this is really the translation of migr_units to | |
2760 | * per-member blocks in the 'resync' case | |
2761 | */ | |
2762 | stripes_per_unit = num_stripes_per_unit_resync(dev); | |
2763 | migr_chunk = migr_strip_blocks_resync(dev); | |
238c0a71 | 2764 | disks = imsm_num_data_members(dev, MAP_0); |
1e5c6983 | 2765 | blocks_per_unit = stripes_per_unit * migr_chunk * disks; |
7b1ab482 | 2766 | stripe = __le16_to_cpu(map->blocks_per_strip) * disks; |
1e5c6983 DW |
2767 | segment = blocks_per_unit / stripe; |
2768 | block_rel = blocks_per_unit - segment * stripe; | |
2769 | parity_depth = parity_segment_depth(dev); | |
2770 | block_map = map_migr_block(dev, block_rel); | |
2771 | return block_map + parity_depth * segment; | |
2772 | } | |
2773 | case MIGR_REBUILD: { | |
2774 | __u32 stripes_per_unit; | |
2775 | __u32 migr_chunk; | |
2776 | ||
2777 | stripes_per_unit = num_stripes_per_unit_rebuild(dev); | |
2778 | migr_chunk = migr_strip_blocks_rebuild(dev); | |
2779 | return migr_chunk * stripes_per_unit; | |
2780 | } | |
1e5c6983 DW |
2781 | case MIGR_STATE_CHANGE: |
2782 | default: | |
2783 | return 0; | |
2784 | } | |
2785 | } | |
2786 | ||
c2c087e6 DW |
2787 | static int imsm_level_to_layout(int level) |
2788 | { | |
2789 | switch (level) { | |
2790 | case 0: | |
2791 | case 1: | |
2792 | return 0; | |
2793 | case 5: | |
2794 | case 6: | |
a380c027 | 2795 | return ALGORITHM_LEFT_ASYMMETRIC; |
c2c087e6 | 2796 | case 10: |
c92a2527 | 2797 | return 0x102; |
c2c087e6 | 2798 | } |
a18a888e | 2799 | return UnSet; |
c2c087e6 DW |
2800 | } |
2801 | ||
8e59f3d8 AK |
2802 | /******************************************************************************* |
2803 | * Function: read_imsm_migr_rec | |
2804 | * Description: Function reads imsm migration record from last sector of disk | |
2805 | * Parameters: | |
2806 | * fd : disk descriptor | |
2807 | * super : metadata info | |
2808 | * Returns: | |
2809 | * 0 : success, | |
2810 | * -1 : fail | |
2811 | ******************************************************************************/ | |
2812 | static int read_imsm_migr_rec(int fd, struct intel_super *super) | |
2813 | { | |
2814 | int ret_val = -1; | |
de44e46f | 2815 | unsigned int sector_size = super->sector_size; |
8e59f3d8 AK |
2816 | unsigned long long dsize; |
2817 | ||
2818 | get_dev_size(fd, NULL, &dsize); | |
de44e46f PB |
2819 | if (lseek64(fd, dsize - (sector_size*MIGR_REC_SECTOR_POSITION), |
2820 | SEEK_SET) < 0) { | |
e7b84f9d N |
2821 | pr_err("Cannot seek to anchor block: %s\n", |
2822 | strerror(errno)); | |
8e59f3d8 AK |
2823 | goto out; |
2824 | } | |
466070ad | 2825 | if ((unsigned int)read(fd, super->migr_rec_buf, |
de44e46f PB |
2826 | MIGR_REC_BUF_SECTORS*sector_size) != |
2827 | MIGR_REC_BUF_SECTORS*sector_size) { | |
e7b84f9d N |
2828 | pr_err("Cannot read migr record block: %s\n", |
2829 | strerror(errno)); | |
8e59f3d8 AK |
2830 | goto out; |
2831 | } | |
2832 | ret_val = 0; | |
de44e46f PB |
2833 | if (sector_size == 4096) |
2834 | convert_from_4k_imsm_migr_rec(super); | |
8e59f3d8 AK |
2835 | |
2836 | out: | |
2837 | return ret_val; | |
2838 | } | |
2839 | ||
3136abe5 AK |
2840 | static struct imsm_dev *imsm_get_device_during_migration( |
2841 | struct intel_super *super) | |
2842 | { | |
2843 | ||
2844 | struct intel_dev *dv; | |
2845 | ||
2846 | for (dv = super->devlist; dv; dv = dv->next) { | |
2847 | if (is_gen_migration(dv->dev)) | |
2848 | return dv->dev; | |
2849 | } | |
2850 | return NULL; | |
2851 | } | |
2852 | ||
8e59f3d8 AK |
2853 | /******************************************************************************* |
2854 | * Function: load_imsm_migr_rec | |
2855 | * Description: Function reads imsm migration record (it is stored at the last | |
2856 | * sector of disk) | |
2857 | * Parameters: | |
2858 | * super : imsm internal array info | |
2859 | * info : general array info | |
2860 | * Returns: | |
2861 | * 0 : success | |
2862 | * -1 : fail | |
4c965cc9 | 2863 | * -2 : no migration in progress |
8e59f3d8 AK |
2864 | ******************************************************************************/ |
2865 | static int load_imsm_migr_rec(struct intel_super *super, struct mdinfo *info) | |
2866 | { | |
2867 | struct mdinfo *sd; | |
594dc1b8 | 2868 | struct dl *dl; |
8e59f3d8 AK |
2869 | char nm[30]; |
2870 | int retval = -1; | |
2871 | int fd = -1; | |
3136abe5 | 2872 | struct imsm_dev *dev; |
594dc1b8 | 2873 | struct imsm_map *map; |
b4ab44d8 | 2874 | int slot = -1; |
3136abe5 AK |
2875 | |
2876 | /* find map under migration */ | |
2877 | dev = imsm_get_device_during_migration(super); | |
2878 | /* nothing to load,no migration in progress? | |
2879 | */ | |
2880 | if (dev == NULL) | |
4c965cc9 | 2881 | return -2; |
8e59f3d8 AK |
2882 | |
2883 | if (info) { | |
2884 | for (sd = info->devs ; sd ; sd = sd->next) { | |
2885 | /* read only from one of the first two slots */ | |
12fe93e9 TM |
2886 | if ((sd->disk.raid_disk < 0) || |
2887 | (sd->disk.raid_disk > 1)) | |
8e59f3d8 | 2888 | continue; |
3136abe5 | 2889 | |
8e59f3d8 AK |
2890 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); |
2891 | fd = dev_open(nm, O_RDONLY); | |
2892 | if (fd >= 0) | |
2893 | break; | |
2894 | } | |
2895 | } | |
2896 | if (fd < 0) { | |
12fe93e9 | 2897 | map = get_imsm_map(dev, MAP_0); |
8e59f3d8 | 2898 | for (dl = super->disks; dl; dl = dl->next) { |
3136abe5 AK |
2899 | /* skip spare and failed disks |
2900 | */ | |
2901 | if (dl->index < 0) | |
2902 | continue; | |
8e59f3d8 | 2903 | /* read only from one of the first two slots */ |
3136abe5 AK |
2904 | if (map) |
2905 | slot = get_imsm_disk_slot(map, dl->index); | |
089f9d79 | 2906 | if (map == NULL || slot > 1 || slot < 0) |
8e59f3d8 AK |
2907 | continue; |
2908 | sprintf(nm, "%d:%d", dl->major, dl->minor); | |
2909 | fd = dev_open(nm, O_RDONLY); | |
2910 | if (fd >= 0) | |
2911 | break; | |
2912 | } | |
2913 | } | |
2914 | if (fd < 0) | |
2915 | goto out; | |
2916 | retval = read_imsm_migr_rec(fd, super); | |
2917 | ||
2918 | out: | |
2919 | if (fd >= 0) | |
2920 | close(fd); | |
2921 | return retval; | |
2922 | } | |
2923 | ||
9e2d750d | 2924 | #ifndef MDASSEMBLE |
c17608ea AK |
2925 | /******************************************************************************* |
2926 | * function: imsm_create_metadata_checkpoint_update | |
2927 | * Description: It creates update for checkpoint change. | |
2928 | * Parameters: | |
2929 | * super : imsm internal array info | |
2930 | * u : pointer to prepared update | |
2931 | * Returns: | |
2932 | * Uptate length. | |
2933 | * If length is equal to 0, input pointer u contains no update | |
2934 | ******************************************************************************/ | |
2935 | static int imsm_create_metadata_checkpoint_update( | |
2936 | struct intel_super *super, | |
2937 | struct imsm_update_general_migration_checkpoint **u) | |
2938 | { | |
2939 | ||
2940 | int update_memory_size = 0; | |
2941 | ||
1ade5cc1 | 2942 | dprintf("(enter)\n"); |
c17608ea AK |
2943 | |
2944 | if (u == NULL) | |
2945 | return 0; | |
2946 | *u = NULL; | |
2947 | ||
2948 | /* size of all update data without anchor */ | |
2949 | update_memory_size = | |
2950 | sizeof(struct imsm_update_general_migration_checkpoint); | |
2951 | ||
503975b9 | 2952 | *u = xcalloc(1, update_memory_size); |
c17608ea | 2953 | if (*u == NULL) { |
1ade5cc1 | 2954 | dprintf("error: cannot get memory\n"); |
c17608ea AK |
2955 | return 0; |
2956 | } | |
2957 | (*u)->type = update_general_migration_checkpoint; | |
2958 | (*u)->curr_migr_unit = __le32_to_cpu(super->migr_rec->curr_migr_unit); | |
1ade5cc1 | 2959 | dprintf("prepared for %u\n", (*u)->curr_migr_unit); |
c17608ea AK |
2960 | |
2961 | return update_memory_size; | |
2962 | } | |
2963 | ||
c17608ea AK |
2964 | static void imsm_update_metadata_locally(struct supertype *st, |
2965 | void *buf, int len); | |
2966 | ||
687629c2 AK |
2967 | /******************************************************************************* |
2968 | * Function: write_imsm_migr_rec | |
2969 | * Description: Function writes imsm migration record | |
2970 | * (at the last sector of disk) | |
2971 | * Parameters: | |
2972 | * super : imsm internal array info | |
2973 | * Returns: | |
2974 | * 0 : success | |
2975 | * -1 : if fail | |
2976 | ******************************************************************************/ | |
2977 | static int write_imsm_migr_rec(struct supertype *st) | |
2978 | { | |
2979 | struct intel_super *super = st->sb; | |
de44e46f | 2980 | unsigned int sector_size = super->sector_size; |
687629c2 AK |
2981 | unsigned long long dsize; |
2982 | char nm[30]; | |
2983 | int fd = -1; | |
2984 | int retval = -1; | |
2985 | struct dl *sd; | |
c17608ea AK |
2986 | int len; |
2987 | struct imsm_update_general_migration_checkpoint *u; | |
3136abe5 | 2988 | struct imsm_dev *dev; |
594dc1b8 | 2989 | struct imsm_map *map; |
3136abe5 AK |
2990 | |
2991 | /* find map under migration */ | |
2992 | dev = imsm_get_device_during_migration(super); | |
2993 | /* if no migration, write buffer anyway to clear migr_record | |
2994 | * on disk based on first available device | |
2995 | */ | |
2996 | if (dev == NULL) | |
2997 | dev = get_imsm_dev(super, super->current_vol < 0 ? 0 : | |
2998 | super->current_vol); | |
2999 | ||
44bfe6df | 3000 | map = get_imsm_map(dev, MAP_0); |
687629c2 | 3001 | |
de44e46f PB |
3002 | if (sector_size == 4096) |
3003 | convert_to_4k_imsm_migr_rec(super); | |
687629c2 | 3004 | for (sd = super->disks ; sd ; sd = sd->next) { |
b4ab44d8 | 3005 | int slot = -1; |
3136abe5 AK |
3006 | |
3007 | /* skip failed and spare devices */ | |
3008 | if (sd->index < 0) | |
3009 | continue; | |
687629c2 | 3010 | /* write to 2 first slots only */ |
3136abe5 AK |
3011 | if (map) |
3012 | slot = get_imsm_disk_slot(map, sd->index); | |
089f9d79 | 3013 | if (map == NULL || slot > 1 || slot < 0) |
687629c2 | 3014 | continue; |
3136abe5 | 3015 | |
687629c2 AK |
3016 | sprintf(nm, "%d:%d", sd->major, sd->minor); |
3017 | fd = dev_open(nm, O_RDWR); | |
3018 | if (fd < 0) | |
3019 | continue; | |
3020 | get_dev_size(fd, NULL, &dsize); | |
de44e46f PB |
3021 | if (lseek64(fd, dsize - (MIGR_REC_SECTOR_POSITION*sector_size), |
3022 | SEEK_SET) < 0) { | |
e7b84f9d N |
3023 | pr_err("Cannot seek to anchor block: %s\n", |
3024 | strerror(errno)); | |
687629c2 AK |
3025 | goto out; |
3026 | } | |
466070ad | 3027 | if ((unsigned int)write(fd, super->migr_rec_buf, |
de44e46f PB |
3028 | MIGR_REC_BUF_SECTORS*sector_size) != |
3029 | MIGR_REC_BUF_SECTORS*sector_size) { | |
e7b84f9d N |
3030 | pr_err("Cannot write migr record block: %s\n", |
3031 | strerror(errno)); | |
687629c2 AK |
3032 | goto out; |
3033 | } | |
3034 | close(fd); | |
3035 | fd = -1; | |
3036 | } | |
de44e46f PB |
3037 | if (sector_size == 4096) |
3038 | convert_from_4k_imsm_migr_rec(super); | |
c17608ea AK |
3039 | /* update checkpoint information in metadata */ |
3040 | len = imsm_create_metadata_checkpoint_update(super, &u); | |
c17608ea AK |
3041 | if (len <= 0) { |
3042 | dprintf("imsm: Cannot prepare update\n"); | |
3043 | goto out; | |
3044 | } | |
3045 | /* update metadata locally */ | |
3046 | imsm_update_metadata_locally(st, u, len); | |
3047 | /* and possibly remotely */ | |
3048 | if (st->update_tail) { | |
3049 | append_metadata_update(st, u, len); | |
3050 | /* during reshape we do all work inside metadata handler | |
3051 | * manage_reshape(), so metadata update has to be triggered | |
3052 | * insida it | |
3053 | */ | |
3054 | flush_metadata_updates(st); | |
3055 | st->update_tail = &st->updates; | |
3056 | } else | |
3057 | free(u); | |
687629c2 AK |
3058 | |
3059 | retval = 0; | |
3060 | out: | |
3061 | if (fd >= 0) | |
3062 | close(fd); | |
3063 | return retval; | |
3064 | } | |
9e2d750d | 3065 | #endif /* MDASSEMBLE */ |
687629c2 | 3066 | |
e2962bfc AK |
3067 | /* spare/missing disks activations are not allowe when |
3068 | * array/container performs reshape operation, because | |
3069 | * all arrays in container works on the same disks set | |
3070 | */ | |
3071 | int imsm_reshape_blocks_arrays_changes(struct intel_super *super) | |
3072 | { | |
3073 | int rv = 0; | |
3074 | struct intel_dev *i_dev; | |
3075 | struct imsm_dev *dev; | |
3076 | ||
3077 | /* check whole container | |
3078 | */ | |
3079 | for (i_dev = super->devlist; i_dev; i_dev = i_dev->next) { | |
3080 | dev = i_dev->dev; | |
3ad25638 | 3081 | if (is_gen_migration(dev)) { |
e2962bfc AK |
3082 | /* No repair during any migration in container |
3083 | */ | |
3084 | rv = 1; | |
3085 | break; | |
3086 | } | |
3087 | } | |
3088 | return rv; | |
3089 | } | |
c41e00b2 AK |
3090 | static unsigned long long imsm_component_size_aligment_check(int level, |
3091 | int chunk_size, | |
f36a9ecd | 3092 | unsigned int sector_size, |
c41e00b2 AK |
3093 | unsigned long long component_size) |
3094 | { | |
3095 | unsigned int component_size_alligment; | |
3096 | ||
3097 | /* check component size aligment | |
3098 | */ | |
f36a9ecd | 3099 | component_size_alligment = component_size % (chunk_size/sector_size); |
c41e00b2 | 3100 | |
1ade5cc1 | 3101 | dprintf("(Level: %i, chunk_size = %i, component_size = %llu), component_size_alligment = %u\n", |
c41e00b2 AK |
3102 | level, chunk_size, component_size, |
3103 | component_size_alligment); | |
3104 | ||
3105 | if (component_size_alligment && (level != 1) && (level != UnSet)) { | |
3106 | dprintf("imsm: reported component size alligned from %llu ", | |
3107 | component_size); | |
3108 | component_size -= component_size_alligment; | |
1ade5cc1 | 3109 | dprintf_cont("to %llu (%i).\n", |
c41e00b2 AK |
3110 | component_size, component_size_alligment); |
3111 | } | |
3112 | ||
3113 | return component_size; | |
3114 | } | |
e2962bfc | 3115 | |
a5d85af7 | 3116 | static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap) |
bf5a934a DW |
3117 | { |
3118 | struct intel_super *super = st->sb; | |
c47b0ff6 | 3119 | struct migr_record *migr_rec = super->migr_rec; |
949c47a0 | 3120 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); |
238c0a71 AK |
3121 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
3122 | struct imsm_map *prev_map = get_imsm_map(dev, MAP_1); | |
b335e593 | 3123 | struct imsm_map *map_to_analyse = map; |
efb30e7f | 3124 | struct dl *dl; |
a5d85af7 | 3125 | int map_disks = info->array.raid_disks; |
bf5a934a | 3126 | |
95eeceeb | 3127 | memset(info, 0, sizeof(*info)); |
b335e593 AK |
3128 | if (prev_map) |
3129 | map_to_analyse = prev_map; | |
3130 | ||
ca0748fa | 3131 | dl = super->current_disk; |
9894ec0d | 3132 | |
bf5a934a | 3133 | info->container_member = super->current_vol; |
cd0430a1 | 3134 | info->array.raid_disks = map->num_members; |
b335e593 | 3135 | info->array.level = get_imsm_raid_level(map_to_analyse); |
bf5a934a DW |
3136 | info->array.layout = imsm_level_to_layout(info->array.level); |
3137 | info->array.md_minor = -1; | |
3138 | info->array.ctime = 0; | |
3139 | info->array.utime = 0; | |
b335e593 AK |
3140 | info->array.chunk_size = |
3141 | __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9; | |
301406c9 | 3142 | info->array.state = !dev->vol.dirty; |
da9b4a62 DW |
3143 | info->custom_array_size = __le32_to_cpu(dev->size_high); |
3144 | info->custom_array_size <<= 32; | |
3145 | info->custom_array_size |= __le32_to_cpu(dev->size_low); | |
3ad25638 AK |
3146 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
3147 | ||
3f510843 | 3148 | if (is_gen_migration(dev)) { |
3f83228a | 3149 | info->reshape_active = 1; |
b335e593 AK |
3150 | info->new_level = get_imsm_raid_level(map); |
3151 | info->new_layout = imsm_level_to_layout(info->new_level); | |
3152 | info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9; | |
3f83228a | 3153 | info->delta_disks = map->num_members - prev_map->num_members; |
493f5dd6 N |
3154 | if (info->delta_disks) { |
3155 | /* this needs to be applied to every array | |
3156 | * in the container. | |
3157 | */ | |
81219e70 | 3158 | info->reshape_active = CONTAINER_RESHAPE; |
493f5dd6 | 3159 | } |
3f83228a N |
3160 | /* We shape information that we give to md might have to be |
3161 | * modify to cope with md's requirement for reshaping arrays. | |
3162 | * For example, when reshaping a RAID0, md requires it to be | |
3163 | * presented as a degraded RAID4. | |
3164 | * Also if a RAID0 is migrating to a RAID5 we need to specify | |
3165 | * the array as already being RAID5, but the 'before' layout | |
3166 | * is a RAID4-like layout. | |
3167 | */ | |
3168 | switch (info->array.level) { | |
3169 | case 0: | |
3170 | switch(info->new_level) { | |
3171 | case 0: | |
3172 | /* conversion is happening as RAID4 */ | |
3173 | info->array.level = 4; | |
3174 | info->array.raid_disks += 1; | |
3175 | break; | |
3176 | case 5: | |
3177 | /* conversion is happening as RAID5 */ | |
3178 | info->array.level = 5; | |
3179 | info->array.layout = ALGORITHM_PARITY_N; | |
3f83228a N |
3180 | info->delta_disks -= 1; |
3181 | break; | |
3182 | default: | |
3183 | /* FIXME error message */ | |
3184 | info->array.level = UnSet; | |
3185 | break; | |
3186 | } | |
3187 | break; | |
3188 | } | |
b335e593 AK |
3189 | } else { |
3190 | info->new_level = UnSet; | |
3191 | info->new_layout = UnSet; | |
3192 | info->new_chunk = info->array.chunk_size; | |
3f83228a | 3193 | info->delta_disks = 0; |
b335e593 | 3194 | } |
ca0748fa | 3195 | |
efb30e7f DW |
3196 | if (dl) { |
3197 | info->disk.major = dl->major; | |
3198 | info->disk.minor = dl->minor; | |
ca0748fa | 3199 | info->disk.number = dl->index; |
656b6b5a N |
3200 | info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse, |
3201 | dl->index); | |
efb30e7f | 3202 | } |
bf5a934a | 3203 | |
5551b113 | 3204 | info->data_offset = pba_of_lba0(map_to_analyse); |
06fb291a PB |
3205 | |
3206 | if (info->array.level == 5) { | |
3207 | info->component_size = num_data_stripes(map_to_analyse) * | |
3208 | map_to_analyse->blocks_per_strip; | |
3209 | } else { | |
3210 | info->component_size = blocks_per_member(map_to_analyse); | |
3211 | } | |
139dae11 | 3212 | |
c41e00b2 AK |
3213 | info->component_size = imsm_component_size_aligment_check( |
3214 | info->array.level, | |
3215 | info->array.chunk_size, | |
f36a9ecd | 3216 | super->sector_size, |
c41e00b2 | 3217 | info->component_size); |
5e46202e | 3218 | info->bb.supported = 1; |
139dae11 | 3219 | |
301406c9 | 3220 | memset(info->uuid, 0, sizeof(info->uuid)); |
921d9e16 | 3221 | info->recovery_start = MaxSector; |
bf5a934a | 3222 | |
d2e6d5d6 | 3223 | info->reshape_progress = 0; |
b6796ce1 | 3224 | info->resync_start = MaxSector; |
b9172665 AK |
3225 | if ((map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED || |
3226 | dev->vol.dirty) && | |
3227 | imsm_reshape_blocks_arrays_changes(super) == 0) { | |
301406c9 | 3228 | info->resync_start = 0; |
b6796ce1 AK |
3229 | } |
3230 | if (dev->vol.migr_state) { | |
1e5c6983 DW |
3231 | switch (migr_type(dev)) { |
3232 | case MIGR_REPAIR: | |
3233 | case MIGR_INIT: { | |
c47b0ff6 AK |
3234 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
3235 | dev); | |
1e5c6983 DW |
3236 | __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit); |
3237 | ||
3238 | info->resync_start = blocks_per_unit * units; | |
3239 | break; | |
3240 | } | |
d2e6d5d6 | 3241 | case MIGR_GEN_MIGR: { |
c47b0ff6 AK |
3242 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
3243 | dev); | |
3244 | __u64 units = __le32_to_cpu(migr_rec->curr_migr_unit); | |
04fa9523 AK |
3245 | unsigned long long array_blocks; |
3246 | int used_disks; | |
d2e6d5d6 | 3247 | |
befb629b AK |
3248 | if (__le32_to_cpu(migr_rec->ascending_migr) && |
3249 | (units < | |
3250 | (__le32_to_cpu(migr_rec->num_migr_units)-1)) && | |
3251 | (super->migr_rec->rec_status == | |
3252 | __cpu_to_le32(UNIT_SRC_IN_CP_AREA))) | |
3253 | units++; | |
3254 | ||
d2e6d5d6 | 3255 | info->reshape_progress = blocks_per_unit * units; |
6289d1e0 | 3256 | |
7a862a02 | 3257 | dprintf("IMSM: General Migration checkpoint : %llu (%llu) -> read reshape progress : %llu\n", |
19986c72 MB |
3258 | (unsigned long long)units, |
3259 | (unsigned long long)blocks_per_unit, | |
3260 | info->reshape_progress); | |
75156c46 | 3261 | |
238c0a71 | 3262 | used_disks = imsm_num_data_members(dev, MAP_1); |
75156c46 | 3263 | if (used_disks > 0) { |
5551b113 | 3264 | array_blocks = blocks_per_member(map) * |
75156c46 AK |
3265 | used_disks; |
3266 | /* round array size down to closest MB | |
3267 | */ | |
3268 | info->custom_array_size = (array_blocks | |
3269 | >> SECT_PER_MB_SHIFT) | |
3270 | << SECT_PER_MB_SHIFT; | |
3271 | } | |
d2e6d5d6 | 3272 | } |
1e5c6983 DW |
3273 | case MIGR_VERIFY: |
3274 | /* we could emulate the checkpointing of | |
3275 | * 'sync_action=check' migrations, but for now | |
3276 | * we just immediately complete them | |
3277 | */ | |
3278 | case MIGR_REBUILD: | |
3279 | /* this is handled by container_content_imsm() */ | |
1e5c6983 DW |
3280 | case MIGR_STATE_CHANGE: |
3281 | /* FIXME handle other migrations */ | |
3282 | default: | |
3283 | /* we are not dirty, so... */ | |
3284 | info->resync_start = MaxSector; | |
3285 | } | |
b6796ce1 | 3286 | } |
301406c9 DW |
3287 | |
3288 | strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN); | |
3289 | info->name[MAX_RAID_SERIAL_LEN] = 0; | |
bf5a934a | 3290 | |
f35f2525 N |
3291 | info->array.major_version = -1; |
3292 | info->array.minor_version = -2; | |
4dd2df09 | 3293 | sprintf(info->text_version, "/%s/%d", st->container_devnm, info->container_member); |
a67dd8cc | 3294 | info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */ |
51006d85 | 3295 | uuid_from_super_imsm(st, info->uuid); |
a5d85af7 N |
3296 | |
3297 | if (dmap) { | |
3298 | int i, j; | |
3299 | for (i=0; i<map_disks; i++) { | |
3300 | dmap[i] = 0; | |
3301 | if (i < info->array.raid_disks) { | |
3302 | struct imsm_disk *dsk; | |
238c0a71 | 3303 | j = get_imsm_disk_idx(dev, i, MAP_X); |
a5d85af7 N |
3304 | dsk = get_imsm_disk(super, j); |
3305 | if (dsk && (dsk->status & CONFIGURED_DISK)) | |
3306 | dmap[i] = 1; | |
3307 | } | |
3308 | } | |
3309 | } | |
81ac8b4d | 3310 | } |
bf5a934a | 3311 | |
3b451610 AK |
3312 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, |
3313 | int failed, int look_in_map); | |
3314 | ||
3315 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev, | |
3316 | int look_in_map); | |
3317 | ||
b4ab44d8 | 3318 | #ifndef MDASSEMBLE |
3b451610 AK |
3319 | static void manage_second_map(struct intel_super *super, struct imsm_dev *dev) |
3320 | { | |
3321 | if (is_gen_migration(dev)) { | |
3322 | int failed; | |
3323 | __u8 map_state; | |
3324 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); | |
3325 | ||
3326 | failed = imsm_count_failed(super, dev, MAP_1); | |
238c0a71 | 3327 | map_state = imsm_check_degraded(super, dev, failed, MAP_1); |
3b451610 AK |
3328 | if (map2->map_state != map_state) { |
3329 | map2->map_state = map_state; | |
3330 | super->updates_pending++; | |
3331 | } | |
3332 | } | |
3333 | } | |
b4ab44d8 | 3334 | #endif |
97b4d0e9 DW |
3335 | |
3336 | static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index) | |
3337 | { | |
3338 | struct dl *d; | |
3339 | ||
3340 | for (d = super->missing; d; d = d->next) | |
3341 | if (d->index == index) | |
3342 | return &d->disk; | |
3343 | return NULL; | |
3344 | } | |
3345 | ||
a5d85af7 | 3346 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map) |
4f5bc454 DW |
3347 | { |
3348 | struct intel_super *super = st->sb; | |
4f5bc454 | 3349 | struct imsm_disk *disk; |
a5d85af7 | 3350 | int map_disks = info->array.raid_disks; |
ab3cb6b3 N |
3351 | int max_enough = -1; |
3352 | int i; | |
3353 | struct imsm_super *mpb; | |
4f5bc454 | 3354 | |
bf5a934a | 3355 | if (super->current_vol >= 0) { |
a5d85af7 | 3356 | getinfo_super_imsm_volume(st, info, map); |
bf5a934a DW |
3357 | return; |
3358 | } | |
95eeceeb | 3359 | memset(info, 0, sizeof(*info)); |
d23fe947 DW |
3360 | |
3361 | /* Set raid_disks to zero so that Assemble will always pull in valid | |
3362 | * spares | |
3363 | */ | |
3364 | info->array.raid_disks = 0; | |
cdddbdbc DW |
3365 | info->array.level = LEVEL_CONTAINER; |
3366 | info->array.layout = 0; | |
3367 | info->array.md_minor = -1; | |
1011e834 | 3368 | info->array.ctime = 0; /* N/A for imsm */ |
cdddbdbc DW |
3369 | info->array.utime = 0; |
3370 | info->array.chunk_size = 0; | |
3371 | ||
3372 | info->disk.major = 0; | |
3373 | info->disk.minor = 0; | |
cdddbdbc | 3374 | info->disk.raid_disk = -1; |
c2c087e6 | 3375 | info->reshape_active = 0; |
f35f2525 N |
3376 | info->array.major_version = -1; |
3377 | info->array.minor_version = -2; | |
c2c087e6 | 3378 | strcpy(info->text_version, "imsm"); |
a67dd8cc | 3379 | info->safe_mode_delay = 0; |
c2c087e6 DW |
3380 | info->disk.number = -1; |
3381 | info->disk.state = 0; | |
c5afc314 | 3382 | info->name[0] = 0; |
921d9e16 | 3383 | info->recovery_start = MaxSector; |
3ad25638 | 3384 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
5e46202e | 3385 | info->bb.supported = 1; |
c2c087e6 | 3386 | |
97b4d0e9 | 3387 | /* do we have the all the insync disks that we expect? */ |
ab3cb6b3 | 3388 | mpb = super->anchor; |
b7d81a38 | 3389 | info->events = __le32_to_cpu(mpb->generation_num); |
97b4d0e9 | 3390 | |
ab3cb6b3 N |
3391 | for (i = 0; i < mpb->num_raid_devs; i++) { |
3392 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
3393 | int failed, enough, j, missing = 0; | |
3394 | struct imsm_map *map; | |
3395 | __u8 state; | |
97b4d0e9 | 3396 | |
3b451610 AK |
3397 | failed = imsm_count_failed(super, dev, MAP_0); |
3398 | state = imsm_check_degraded(super, dev, failed, MAP_0); | |
238c0a71 | 3399 | map = get_imsm_map(dev, MAP_0); |
ab3cb6b3 N |
3400 | |
3401 | /* any newly missing disks? | |
3402 | * (catches single-degraded vs double-degraded) | |
3403 | */ | |
3404 | for (j = 0; j < map->num_members; j++) { | |
238c0a71 | 3405 | __u32 ord = get_imsm_ord_tbl_ent(dev, j, MAP_0); |
ab3cb6b3 N |
3406 | __u32 idx = ord_to_idx(ord); |
3407 | ||
3408 | if (!(ord & IMSM_ORD_REBUILD) && | |
3409 | get_imsm_missing(super, idx)) { | |
3410 | missing = 1; | |
3411 | break; | |
3412 | } | |
97b4d0e9 | 3413 | } |
ab3cb6b3 N |
3414 | |
3415 | if (state == IMSM_T_STATE_FAILED) | |
3416 | enough = -1; | |
3417 | else if (state == IMSM_T_STATE_DEGRADED && | |
3418 | (state != map->map_state || missing)) | |
3419 | enough = 0; | |
3420 | else /* we're normal, or already degraded */ | |
3421 | enough = 1; | |
d2bde6d3 AK |
3422 | if (is_gen_migration(dev) && missing) { |
3423 | /* during general migration we need all disks | |
3424 | * that process is running on. | |
3425 | * No new missing disk is allowed. | |
3426 | */ | |
3427 | max_enough = -1; | |
3428 | enough = -1; | |
3429 | /* no more checks necessary | |
3430 | */ | |
3431 | break; | |
3432 | } | |
ab3cb6b3 N |
3433 | /* in the missing/failed disk case check to see |
3434 | * if at least one array is runnable | |
3435 | */ | |
3436 | max_enough = max(max_enough, enough); | |
3437 | } | |
1ade5cc1 | 3438 | dprintf("enough: %d\n", max_enough); |
ab3cb6b3 | 3439 | info->container_enough = max_enough; |
97b4d0e9 | 3440 | |
4a04ec6c | 3441 | if (super->disks) { |
14e8215b DW |
3442 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
3443 | ||
b9f594fe | 3444 | disk = &super->disks->disk; |
5551b113 | 3445 | info->data_offset = total_blocks(&super->disks->disk) - reserved; |
14e8215b | 3446 | info->component_size = reserved; |
25ed7e59 | 3447 | info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0; |
df474657 DW |
3448 | /* we don't change info->disk.raid_disk here because |
3449 | * this state will be finalized in mdmon after we have | |
3450 | * found the 'most fresh' version of the metadata | |
3451 | */ | |
25ed7e59 DW |
3452 | info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; |
3453 | info->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
cdddbdbc | 3454 | } |
a575e2a7 DW |
3455 | |
3456 | /* only call uuid_from_super_imsm when this disk is part of a populated container, | |
3457 | * ->compare_super may have updated the 'num_raid_devs' field for spares | |
3458 | */ | |
3459 | if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs) | |
36ba7d48 | 3460 | uuid_from_super_imsm(st, info->uuid); |
22e263f6 AC |
3461 | else |
3462 | memcpy(info->uuid, uuid_zero, sizeof(uuid_zero)); | |
a5d85af7 N |
3463 | |
3464 | /* I don't know how to compute 'map' on imsm, so use safe default */ | |
3465 | if (map) { | |
3466 | int i; | |
3467 | for (i = 0; i < map_disks; i++) | |
3468 | map[i] = 1; | |
3469 | } | |
3470 | ||
cdddbdbc DW |
3471 | } |
3472 | ||
5c4cd5da AC |
3473 | /* allocates memory and fills disk in mdinfo structure |
3474 | * for each disk in array */ | |
3475 | struct mdinfo *getinfo_super_disks_imsm(struct supertype *st) | |
3476 | { | |
594dc1b8 | 3477 | struct mdinfo *mddev; |
5c4cd5da AC |
3478 | struct intel_super *super = st->sb; |
3479 | struct imsm_disk *disk; | |
3480 | int count = 0; | |
3481 | struct dl *dl; | |
3482 | if (!super || !super->disks) | |
3483 | return NULL; | |
3484 | dl = super->disks; | |
503975b9 | 3485 | mddev = xcalloc(1, sizeof(*mddev)); |
5c4cd5da AC |
3486 | while (dl) { |
3487 | struct mdinfo *tmp; | |
3488 | disk = &dl->disk; | |
503975b9 | 3489 | tmp = xcalloc(1, sizeof(*tmp)); |
5c4cd5da AC |
3490 | if (mddev->devs) |
3491 | tmp->next = mddev->devs; | |
3492 | mddev->devs = tmp; | |
3493 | tmp->disk.number = count++; | |
3494 | tmp->disk.major = dl->major; | |
3495 | tmp->disk.minor = dl->minor; | |
3496 | tmp->disk.state = is_configured(disk) ? | |
3497 | (1 << MD_DISK_ACTIVE) : 0; | |
3498 | tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; | |
3499 | tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
3500 | tmp->disk.raid_disk = -1; | |
3501 | dl = dl->next; | |
3502 | } | |
3503 | return mddev; | |
3504 | } | |
3505 | ||
cdddbdbc DW |
3506 | static int update_super_imsm(struct supertype *st, struct mdinfo *info, |
3507 | char *update, char *devname, int verbose, | |
3508 | int uuid_set, char *homehost) | |
3509 | { | |
f352c545 DW |
3510 | /* For 'assemble' and 'force' we need to return non-zero if any |
3511 | * change was made. For others, the return value is ignored. | |
3512 | * Update options are: | |
3513 | * force-one : This device looks a bit old but needs to be included, | |
3514 | * update age info appropriately. | |
3515 | * assemble: clear any 'faulty' flag to allow this device to | |
3516 | * be assembled. | |
3517 | * force-array: Array is degraded but being forced, mark it clean | |
3518 | * if that will be needed to assemble it. | |
3519 | * | |
3520 | * newdev: not used ???? | |
3521 | * grow: Array has gained a new device - this is currently for | |
3522 | * linear only | |
3523 | * resync: mark as dirty so a resync will happen. | |
3524 | * name: update the name - preserving the homehost | |
6e46bf34 | 3525 | * uuid: Change the uuid of the array to match watch is given |
f352c545 DW |
3526 | * |
3527 | * Following are not relevant for this imsm: | |
3528 | * sparc2.2 : update from old dodgey metadata | |
3529 | * super-minor: change the preferred_minor number | |
3530 | * summaries: update redundant counters. | |
f352c545 DW |
3531 | * homehost: update the recorded homehost |
3532 | * _reshape_progress: record new reshape_progress position. | |
3533 | */ | |
6e46bf34 DW |
3534 | int rv = 1; |
3535 | struct intel_super *super = st->sb; | |
3536 | struct imsm_super *mpb; | |
f352c545 | 3537 | |
6e46bf34 DW |
3538 | /* we can only update container info */ |
3539 | if (!super || super->current_vol >= 0 || !super->anchor) | |
3540 | return 1; | |
3541 | ||
3542 | mpb = super->anchor; | |
3543 | ||
81a5b4f5 N |
3544 | if (strcmp(update, "uuid") == 0) { |
3545 | /* We take this to mean that the family_num should be updated. | |
3546 | * However that is much smaller than the uuid so we cannot really | |
3547 | * allow an explicit uuid to be given. And it is hard to reliably | |
3548 | * know if one was. | |
3549 | * So if !uuid_set we know the current uuid is random and just used | |
3550 | * the first 'int' and copy it to the other 3 positions. | |
3551 | * Otherwise we require the 4 'int's to be the same as would be the | |
3552 | * case if we are using a random uuid. So an explicit uuid will be | |
3553 | * accepted as long as all for ints are the same... which shouldn't hurt | |
6e46bf34 | 3554 | */ |
81a5b4f5 N |
3555 | if (!uuid_set) { |
3556 | info->uuid[1] = info->uuid[2] = info->uuid[3] = info->uuid[0]; | |
6e46bf34 | 3557 | rv = 0; |
81a5b4f5 N |
3558 | } else { |
3559 | if (info->uuid[0] != info->uuid[1] || | |
3560 | info->uuid[1] != info->uuid[2] || | |
3561 | info->uuid[2] != info->uuid[3]) | |
3562 | rv = -1; | |
3563 | else | |
3564 | rv = 0; | |
6e46bf34 | 3565 | } |
81a5b4f5 N |
3566 | if (rv == 0) |
3567 | mpb->orig_family_num = info->uuid[0]; | |
6e46bf34 DW |
3568 | } else if (strcmp(update, "assemble") == 0) |
3569 | rv = 0; | |
3570 | else | |
1e2b2765 | 3571 | rv = -1; |
f352c545 | 3572 | |
6e46bf34 DW |
3573 | /* successful update? recompute checksum */ |
3574 | if (rv == 0) | |
3575 | mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb)); | |
f352c545 DW |
3576 | |
3577 | return rv; | |
cdddbdbc DW |
3578 | } |
3579 | ||
c2c087e6 | 3580 | static size_t disks_to_mpb_size(int disks) |
cdddbdbc | 3581 | { |
c2c087e6 | 3582 | size_t size; |
cdddbdbc | 3583 | |
c2c087e6 DW |
3584 | size = sizeof(struct imsm_super); |
3585 | size += (disks - 1) * sizeof(struct imsm_disk); | |
3586 | size += 2 * sizeof(struct imsm_dev); | |
3587 | /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */ | |
3588 | size += (4 - 2) * sizeof(struct imsm_map); | |
3589 | /* 4 possible disk_ord_tbl's */ | |
3590 | size += 4 * (disks - 1) * sizeof(__u32); | |
bbab0940 TM |
3591 | /* maximum bbm log */ |
3592 | size += sizeof(struct bbm_log); | |
c2c087e6 DW |
3593 | |
3594 | return size; | |
3595 | } | |
3596 | ||
387fcd59 N |
3597 | static __u64 avail_size_imsm(struct supertype *st, __u64 devsize, |
3598 | unsigned long long data_offset) | |
c2c087e6 DW |
3599 | { |
3600 | if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS)) | |
3601 | return 0; | |
3602 | ||
3603 | return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); | |
cdddbdbc DW |
3604 | } |
3605 | ||
ba2de7ba DW |
3606 | static void free_devlist(struct intel_super *super) |
3607 | { | |
3608 | struct intel_dev *dv; | |
3609 | ||
3610 | while (super->devlist) { | |
3611 | dv = super->devlist->next; | |
3612 | free(super->devlist->dev); | |
3613 | free(super->devlist); | |
3614 | super->devlist = dv; | |
3615 | } | |
3616 | } | |
3617 | ||
3618 | static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src) | |
3619 | { | |
3620 | memcpy(dest, src, sizeof_imsm_dev(src, 0)); | |
3621 | } | |
3622 | ||
cdddbdbc DW |
3623 | static int compare_super_imsm(struct supertype *st, struct supertype *tst) |
3624 | { | |
3625 | /* | |
3626 | * return: | |
3627 | * 0 same, or first was empty, and second was copied | |
3628 | * 1 second had wrong number | |
3629 | * 2 wrong uuid | |
3630 | * 3 wrong other info | |
3631 | */ | |
3632 | struct intel_super *first = st->sb; | |
3633 | struct intel_super *sec = tst->sb; | |
3634 | ||
5d500228 N |
3635 | if (!first) { |
3636 | st->sb = tst->sb; | |
3637 | tst->sb = NULL; | |
3638 | return 0; | |
3639 | } | |
8603ea6f LM |
3640 | /* in platform dependent environment test if the disks |
3641 | * use the same Intel hba | |
cb8f6859 | 3642 | * If not on Intel hba at all, allow anything. |
8603ea6f | 3643 | */ |
6b781d33 AP |
3644 | if (!check_env("IMSM_NO_PLATFORM") && first->hba && sec->hba) { |
3645 | if (first->hba->type != sec->hba->type) { | |
8603ea6f | 3646 | fprintf(stderr, |
6b781d33 AP |
3647 | "HBAs of devices do not match %s != %s\n", |
3648 | get_sys_dev_type(first->hba->type), | |
3649 | get_sys_dev_type(sec->hba->type)); | |
3650 | return 3; | |
3651 | } | |
3652 | if (first->orom != sec->orom) { | |
3653 | fprintf(stderr, | |
3654 | "HBAs of devices do not match %s != %s\n", | |
3655 | first->hba->pci_id, sec->hba->pci_id); | |
8603ea6f LM |
3656 | return 3; |
3657 | } | |
3658 | } | |
cdddbdbc | 3659 | |
d23fe947 DW |
3660 | /* if an anchor does not have num_raid_devs set then it is a free |
3661 | * floating spare | |
3662 | */ | |
3663 | if (first->anchor->num_raid_devs > 0 && | |
3664 | sec->anchor->num_raid_devs > 0) { | |
a2b97981 DW |
3665 | /* Determine if these disks might ever have been |
3666 | * related. Further disambiguation can only take place | |
3667 | * in load_super_imsm_all | |
3668 | */ | |
3669 | __u32 first_family = first->anchor->orig_family_num; | |
3670 | __u32 sec_family = sec->anchor->orig_family_num; | |
3671 | ||
f796af5d DW |
3672 | if (memcmp(first->anchor->sig, sec->anchor->sig, |
3673 | MAX_SIGNATURE_LENGTH) != 0) | |
3674 | return 3; | |
3675 | ||
a2b97981 DW |
3676 | if (first_family == 0) |
3677 | first_family = first->anchor->family_num; | |
3678 | if (sec_family == 0) | |
3679 | sec_family = sec->anchor->family_num; | |
3680 | ||
3681 | if (first_family != sec_family) | |
d23fe947 | 3682 | return 3; |
f796af5d | 3683 | |
d23fe947 | 3684 | } |
cdddbdbc | 3685 | |
3e372e5a DW |
3686 | /* if 'first' is a spare promote it to a populated mpb with sec's |
3687 | * family number | |
3688 | */ | |
3689 | if (first->anchor->num_raid_devs == 0 && | |
3690 | sec->anchor->num_raid_devs > 0) { | |
78d30f94 | 3691 | int i; |
ba2de7ba DW |
3692 | struct intel_dev *dv; |
3693 | struct imsm_dev *dev; | |
78d30f94 DW |
3694 | |
3695 | /* we need to copy raid device info from sec if an allocation | |
3696 | * fails here we don't associate the spare | |
3697 | */ | |
3698 | for (i = 0; i < sec->anchor->num_raid_devs; i++) { | |
503975b9 N |
3699 | dv = xmalloc(sizeof(*dv)); |
3700 | dev = xmalloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1)); | |
ba2de7ba DW |
3701 | dv->dev = dev; |
3702 | dv->index = i; | |
3703 | dv->next = first->devlist; | |
3704 | first->devlist = dv; | |
78d30f94 | 3705 | } |
709743c5 | 3706 | if (i < sec->anchor->num_raid_devs) { |
ba2de7ba DW |
3707 | /* allocation failure */ |
3708 | free_devlist(first); | |
e12b3daa | 3709 | pr_err("imsm: failed to associate spare\n"); |
ba2de7ba | 3710 | return 3; |
78d30f94 | 3711 | } |
3e372e5a | 3712 | first->anchor->num_raid_devs = sec->anchor->num_raid_devs; |
148acb7b | 3713 | first->anchor->orig_family_num = sec->anchor->orig_family_num; |
3e372e5a | 3714 | first->anchor->family_num = sec->anchor->family_num; |
ac6449be | 3715 | memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH); |
709743c5 DW |
3716 | for (i = 0; i < sec->anchor->num_raid_devs; i++) |
3717 | imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i)); | |
3e372e5a DW |
3718 | } |
3719 | ||
cdddbdbc DW |
3720 | return 0; |
3721 | } | |
3722 | ||
0030e8d6 DW |
3723 | static void fd2devname(int fd, char *name) |
3724 | { | |
3725 | struct stat st; | |
3726 | char path[256]; | |
33a6535d | 3727 | char dname[PATH_MAX]; |
0030e8d6 DW |
3728 | char *nm; |
3729 | int rv; | |
3730 | ||
3731 | name[0] = '\0'; | |
3732 | if (fstat(fd, &st) != 0) | |
3733 | return; | |
3734 | sprintf(path, "/sys/dev/block/%d:%d", | |
3735 | major(st.st_rdev), minor(st.st_rdev)); | |
3736 | ||
9cf014ec | 3737 | rv = readlink(path, dname, sizeof(dname)-1); |
0030e8d6 DW |
3738 | if (rv <= 0) |
3739 | return; | |
9587c373 | 3740 | |
0030e8d6 DW |
3741 | dname[rv] = '\0'; |
3742 | nm = strrchr(dname, '/'); | |
7897de29 JS |
3743 | if (nm) { |
3744 | nm++; | |
3745 | snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm); | |
3746 | } | |
0030e8d6 DW |
3747 | } |
3748 | ||
21e9380b AP |
3749 | static int nvme_get_serial(int fd, void *buf, size_t buf_len) |
3750 | { | |
3751 | char path[60]; | |
3752 | char *name = fd2kname(fd); | |
3753 | ||
3754 | if (!name) | |
3755 | return 1; | |
3756 | ||
3757 | if (strncmp(name, "nvme", 4) != 0) | |
3758 | return 1; | |
3759 | ||
3760 | snprintf(path, sizeof(path) - 1, "/sys/block/%s/device/serial", name); | |
3761 | ||
3762 | return load_sys(path, buf, buf_len); | |
3763 | } | |
3764 | ||
cdddbdbc DW |
3765 | extern int scsi_get_serial(int fd, void *buf, size_t buf_len); |
3766 | ||
3767 | static int imsm_read_serial(int fd, char *devname, | |
3768 | __u8 serial[MAX_RAID_SERIAL_LEN]) | |
3769 | { | |
21e9380b | 3770 | char buf[50]; |
cdddbdbc | 3771 | int rv; |
1f24f035 | 3772 | int len; |
316e2bf4 DW |
3773 | char *dest; |
3774 | char *src; | |
21e9380b AP |
3775 | unsigned int i; |
3776 | ||
3777 | memset(buf, 0, sizeof(buf)); | |
cdddbdbc | 3778 | |
21e9380b | 3779 | rv = nvme_get_serial(fd, buf, sizeof(buf)); |
cdddbdbc | 3780 | |
21e9380b AP |
3781 | if (rv) |
3782 | rv = scsi_get_serial(fd, buf, sizeof(buf)); | |
f9ba0ff1 | 3783 | |
40ebbb9c | 3784 | if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) { |
f9ba0ff1 DW |
3785 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
3786 | fd2devname(fd, (char *) serial); | |
0030e8d6 DW |
3787 | return 0; |
3788 | } | |
3789 | ||
cdddbdbc DW |
3790 | if (rv != 0) { |
3791 | if (devname) | |
e7b84f9d N |
3792 | pr_err("Failed to retrieve serial for %s\n", |
3793 | devname); | |
cdddbdbc DW |
3794 | return rv; |
3795 | } | |
3796 | ||
316e2bf4 DW |
3797 | /* trim all whitespace and non-printable characters and convert |
3798 | * ':' to ';' | |
3799 | */ | |
21e9380b AP |
3800 | for (i = 0, dest = buf; i < sizeof(buf) && buf[i]; i++) { |
3801 | src = &buf[i]; | |
316e2bf4 DW |
3802 | if (*src > 0x20) { |
3803 | /* ':' is reserved for use in placeholder serial | |
3804 | * numbers for missing disks | |
3805 | */ | |
3806 | if (*src == ':') | |
3807 | *dest++ = ';'; | |
3808 | else | |
3809 | *dest++ = *src; | |
3810 | } | |
3811 | } | |
21e9380b AP |
3812 | len = dest - buf; |
3813 | dest = buf; | |
316e2bf4 DW |
3814 | |
3815 | /* truncate leading characters */ | |
3816 | if (len > MAX_RAID_SERIAL_LEN) { | |
3817 | dest += len - MAX_RAID_SERIAL_LEN; | |
1f24f035 | 3818 | len = MAX_RAID_SERIAL_LEN; |
316e2bf4 | 3819 | } |
5c3db629 | 3820 | |
5c3db629 | 3821 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
316e2bf4 | 3822 | memcpy(serial, dest, len); |
cdddbdbc DW |
3823 | |
3824 | return 0; | |
3825 | } | |
3826 | ||
1f24f035 DW |
3827 | static int serialcmp(__u8 *s1, __u8 *s2) |
3828 | { | |
3829 | return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN); | |
3830 | } | |
3831 | ||
3832 | static void serialcpy(__u8 *dest, __u8 *src) | |
3833 | { | |
3834 | strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN); | |
3835 | } | |
3836 | ||
54c2c1ea DW |
3837 | static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super) |
3838 | { | |
3839 | struct dl *dl; | |
3840 | ||
3841 | for (dl = super->disks; dl; dl = dl->next) | |
3842 | if (serialcmp(dl->serial, serial) == 0) | |
3843 | break; | |
3844 | ||
3845 | return dl; | |
3846 | } | |
3847 | ||
a2b97981 DW |
3848 | static struct imsm_disk * |
3849 | __serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx) | |
3850 | { | |
3851 | int i; | |
3852 | ||
3853 | for (i = 0; i < mpb->num_disks; i++) { | |
3854 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
3855 | ||
3856 | if (serialcmp(disk->serial, serial) == 0) { | |
3857 | if (idx) | |
3858 | *idx = i; | |
3859 | return disk; | |
3860 | } | |
3861 | } | |
3862 | ||
3863 | return NULL; | |
3864 | } | |
3865 | ||
cdddbdbc DW |
3866 | static int |
3867 | load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd) | |
3868 | { | |
a2b97981 | 3869 | struct imsm_disk *disk; |
cdddbdbc DW |
3870 | struct dl *dl; |
3871 | struct stat stb; | |
cdddbdbc | 3872 | int rv; |
a2b97981 | 3873 | char name[40]; |
d23fe947 DW |
3874 | __u8 serial[MAX_RAID_SERIAL_LEN]; |
3875 | ||
3876 | rv = imsm_read_serial(fd, devname, serial); | |
3877 | ||
3878 | if (rv != 0) | |
3879 | return 2; | |
3880 | ||
503975b9 | 3881 | dl = xcalloc(1, sizeof(*dl)); |
cdddbdbc | 3882 | |
a2b97981 DW |
3883 | fstat(fd, &stb); |
3884 | dl->major = major(stb.st_rdev); | |
3885 | dl->minor = minor(stb.st_rdev); | |
3886 | dl->next = super->disks; | |
3887 | dl->fd = keep_fd ? fd : -1; | |
3888 | assert(super->disks == NULL); | |
3889 | super->disks = dl; | |
3890 | serialcpy(dl->serial, serial); | |
3891 | dl->index = -2; | |
3892 | dl->e = NULL; | |
3893 | fd2devname(fd, name); | |
3894 | if (devname) | |
503975b9 | 3895 | dl->devname = xstrdup(devname); |
a2b97981 | 3896 | else |
503975b9 | 3897 | dl->devname = xstrdup(name); |
cdddbdbc | 3898 | |
d23fe947 | 3899 | /* look up this disk's index in the current anchor */ |
a2b97981 DW |
3900 | disk = __serial_to_disk(dl->serial, super->anchor, &dl->index); |
3901 | if (disk) { | |
3902 | dl->disk = *disk; | |
3903 | /* only set index on disks that are a member of a | |
3904 | * populated contianer, i.e. one with raid_devs | |
3905 | */ | |
3906 | if (is_failed(&dl->disk)) | |
3f6efecc | 3907 | dl->index = -2; |
a2b97981 DW |
3908 | else if (is_spare(&dl->disk)) |
3909 | dl->index = -1; | |
3f6efecc DW |
3910 | } |
3911 | ||
949c47a0 DW |
3912 | return 0; |
3913 | } | |
3914 | ||
0e600426 | 3915 | #ifndef MDASSEMBLE |
0c046afd DW |
3916 | /* When migrating map0 contains the 'destination' state while map1 |
3917 | * contains the current state. When not migrating map0 contains the | |
3918 | * current state. This routine assumes that map[0].map_state is set to | |
3919 | * the current array state before being called. | |
3920 | * | |
3921 | * Migration is indicated by one of the following states | |
3922 | * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed) | |
e3bba0e0 | 3923 | * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal |
0c046afd | 3924 | * map1state=unitialized) |
1484e727 | 3925 | * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal |
0c046afd | 3926 | * map1state=normal) |
e3bba0e0 | 3927 | * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal |
0c046afd | 3928 | * map1state=degraded) |
8e59f3d8 AK |
3929 | * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal |
3930 | * map1state=normal) | |
0c046afd | 3931 | */ |
8e59f3d8 AK |
3932 | static void migrate(struct imsm_dev *dev, struct intel_super *super, |
3933 | __u8 to_state, int migr_type) | |
3393c6af | 3934 | { |
0c046afd | 3935 | struct imsm_map *dest; |
238c0a71 | 3936 | struct imsm_map *src = get_imsm_map(dev, MAP_0); |
3393c6af | 3937 | |
0c046afd | 3938 | dev->vol.migr_state = 1; |
1484e727 | 3939 | set_migr_type(dev, migr_type); |
f8f603f1 | 3940 | dev->vol.curr_migr_unit = 0; |
238c0a71 | 3941 | dest = get_imsm_map(dev, MAP_1); |
0c046afd | 3942 | |
0556e1a2 | 3943 | /* duplicate and then set the target end state in map[0] */ |
3393c6af | 3944 | memcpy(dest, src, sizeof_imsm_map(src)); |
089f9d79 | 3945 | if (migr_type == MIGR_REBUILD || migr_type == MIGR_GEN_MIGR) { |
0556e1a2 DW |
3946 | __u32 ord; |
3947 | int i; | |
3948 | ||
3949 | for (i = 0; i < src->num_members; i++) { | |
3950 | ord = __le32_to_cpu(src->disk_ord_tbl[i]); | |
3951 | set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord)); | |
3952 | } | |
3953 | } | |
3954 | ||
8e59f3d8 AK |
3955 | if (migr_type == MIGR_GEN_MIGR) |
3956 | /* Clear migration record */ | |
3957 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
3958 | ||
0c046afd | 3959 | src->map_state = to_state; |
949c47a0 | 3960 | } |
f8f603f1 | 3961 | |
809da78e AK |
3962 | static void end_migration(struct imsm_dev *dev, struct intel_super *super, |
3963 | __u8 map_state) | |
f8f603f1 | 3964 | { |
238c0a71 AK |
3965 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
3966 | struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state == 0 ? | |
3967 | MAP_0 : MAP_1); | |
28bce06f | 3968 | int i, j; |
0556e1a2 DW |
3969 | |
3970 | /* merge any IMSM_ORD_REBUILD bits that were not successfully | |
3971 | * completed in the last migration. | |
3972 | * | |
28bce06f | 3973 | * FIXME add support for raid-level-migration |
0556e1a2 | 3974 | */ |
089f9d79 JS |
3975 | if (map_state != map->map_state && (is_gen_migration(dev) == 0) && |
3976 | prev->map_state != IMSM_T_STATE_UNINITIALIZED) { | |
809da78e AK |
3977 | /* when final map state is other than expected |
3978 | * merge maps (not for migration) | |
3979 | */ | |
3980 | int failed; | |
3981 | ||
3982 | for (i = 0; i < prev->num_members; i++) | |
3983 | for (j = 0; j < map->num_members; j++) | |
3984 | /* during online capacity expansion | |
3985 | * disks position can be changed | |
3986 | * if takeover is used | |
3987 | */ | |
3988 | if (ord_to_idx(map->disk_ord_tbl[j]) == | |
3989 | ord_to_idx(prev->disk_ord_tbl[i])) { | |
3990 | map->disk_ord_tbl[j] |= | |
3991 | prev->disk_ord_tbl[i]; | |
3992 | break; | |
3993 | } | |
3994 | failed = imsm_count_failed(super, dev, MAP_0); | |
3995 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
3996 | } | |
f8f603f1 DW |
3997 | |
3998 | dev->vol.migr_state = 0; | |
ea672ee1 | 3999 | set_migr_type(dev, 0); |
f8f603f1 DW |
4000 | dev->vol.curr_migr_unit = 0; |
4001 | map->map_state = map_state; | |
4002 | } | |
0e600426 | 4003 | #endif |
949c47a0 DW |
4004 | |
4005 | static int parse_raid_devices(struct intel_super *super) | |
4006 | { | |
4007 | int i; | |
4008 | struct imsm_dev *dev_new; | |
4d7b1503 | 4009 | size_t len, len_migr; |
401d313b | 4010 | size_t max_len = 0; |
4d7b1503 DW |
4011 | size_t space_needed = 0; |
4012 | struct imsm_super *mpb = super->anchor; | |
949c47a0 DW |
4013 | |
4014 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
4015 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
ba2de7ba | 4016 | struct intel_dev *dv; |
949c47a0 | 4017 | |
4d7b1503 DW |
4018 | len = sizeof_imsm_dev(dev_iter, 0); |
4019 | len_migr = sizeof_imsm_dev(dev_iter, 1); | |
4020 | if (len_migr > len) | |
4021 | space_needed += len_migr - len; | |
ca9de185 | 4022 | |
503975b9 | 4023 | dv = xmalloc(sizeof(*dv)); |
401d313b AK |
4024 | if (max_len < len_migr) |
4025 | max_len = len_migr; | |
4026 | if (max_len > len_migr) | |
4027 | space_needed += max_len - len_migr; | |
503975b9 | 4028 | dev_new = xmalloc(max_len); |
949c47a0 | 4029 | imsm_copy_dev(dev_new, dev_iter); |
ba2de7ba DW |
4030 | dv->dev = dev_new; |
4031 | dv->index = i; | |
4032 | dv->next = super->devlist; | |
4033 | super->devlist = dv; | |
949c47a0 | 4034 | } |
cdddbdbc | 4035 | |
4d7b1503 DW |
4036 | /* ensure that super->buf is large enough when all raid devices |
4037 | * are migrating | |
4038 | */ | |
4039 | if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) { | |
4040 | void *buf; | |
4041 | ||
f36a9ecd PB |
4042 | len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed, |
4043 | super->sector_size); | |
4044 | if (posix_memalign(&buf, MAX_SECTOR_SIZE, len) != 0) | |
4d7b1503 DW |
4045 | return 1; |
4046 | ||
1f45a8ad DW |
4047 | memcpy(buf, super->buf, super->len); |
4048 | memset(buf + super->len, 0, len - super->len); | |
4d7b1503 DW |
4049 | free(super->buf); |
4050 | super->buf = buf; | |
4051 | super->len = len; | |
4052 | } | |
ca9de185 | 4053 | |
bbab0940 TM |
4054 | super->extra_space += space_needed; |
4055 | ||
cdddbdbc DW |
4056 | return 0; |
4057 | } | |
4058 | ||
e2f41b2c AK |
4059 | /******************************************************************************* |
4060 | * Function: check_mpb_migr_compatibility | |
4061 | * Description: Function checks for unsupported migration features: | |
4062 | * - migration optimization area (pba_of_lba0) | |
4063 | * - descending reshape (ascending_migr) | |
4064 | * Parameters: | |
4065 | * super : imsm metadata information | |
4066 | * Returns: | |
4067 | * 0 : migration is compatible | |
4068 | * -1 : migration is not compatible | |
4069 | ******************************************************************************/ | |
4070 | int check_mpb_migr_compatibility(struct intel_super *super) | |
4071 | { | |
4072 | struct imsm_map *map0, *map1; | |
4073 | struct migr_record *migr_rec = super->migr_rec; | |
4074 | int i; | |
4075 | ||
4076 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
4077 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
4078 | ||
4079 | if (dev_iter && | |
4080 | dev_iter->vol.migr_state == 1 && | |
4081 | dev_iter->vol.migr_type == MIGR_GEN_MIGR) { | |
4082 | /* This device is migrating */ | |
238c0a71 AK |
4083 | map0 = get_imsm_map(dev_iter, MAP_0); |
4084 | map1 = get_imsm_map(dev_iter, MAP_1); | |
5551b113 | 4085 | if (pba_of_lba0(map0) != pba_of_lba0(map1)) |
e2f41b2c AK |
4086 | /* migration optimization area was used */ |
4087 | return -1; | |
4088 | if (migr_rec->ascending_migr == 0 | |
4089 | && migr_rec->dest_depth_per_unit > 0) | |
4090 | /* descending reshape not supported yet */ | |
4091 | return -1; | |
4092 | } | |
4093 | } | |
4094 | return 0; | |
4095 | } | |
4096 | ||
d23fe947 | 4097 | static void __free_imsm(struct intel_super *super, int free_disks); |
9ca2c81c | 4098 | |
cdddbdbc | 4099 | /* load_imsm_mpb - read matrix metadata |
f2f5c343 | 4100 | * allocates super->mpb to be freed by free_imsm |
cdddbdbc DW |
4101 | */ |
4102 | static int load_imsm_mpb(int fd, struct intel_super *super, char *devname) | |
4103 | { | |
4104 | unsigned long long dsize; | |
cdddbdbc | 4105 | unsigned long long sectors; |
f36a9ecd | 4106 | unsigned int sector_size = super->sector_size; |
cdddbdbc | 4107 | struct stat; |
6416d527 | 4108 | struct imsm_super *anchor; |
cdddbdbc DW |
4109 | __u32 check_sum; |
4110 | ||
cdddbdbc | 4111 | get_dev_size(fd, NULL, &dsize); |
f36a9ecd | 4112 | if (dsize < 2*sector_size) { |
64436f06 | 4113 | if (devname) |
e7b84f9d N |
4114 | pr_err("%s: device to small for imsm\n", |
4115 | devname); | |
64436f06 N |
4116 | return 1; |
4117 | } | |
cdddbdbc | 4118 | |
f36a9ecd | 4119 | if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0) { |
cdddbdbc | 4120 | if (devname) |
e7b84f9d N |
4121 | pr_err("Cannot seek to anchor block on %s: %s\n", |
4122 | devname, strerror(errno)); | |
cdddbdbc DW |
4123 | return 1; |
4124 | } | |
4125 | ||
f36a9ecd | 4126 | if (posix_memalign((void **)&anchor, sector_size, sector_size) != 0) { |
ad97895e | 4127 | if (devname) |
7a862a02 | 4128 | pr_err("Failed to allocate imsm anchor buffer on %s\n", devname); |
ad97895e DW |
4129 | return 1; |
4130 | } | |
466070ad | 4131 | if ((unsigned int)read(fd, anchor, sector_size) != sector_size) { |
cdddbdbc | 4132 | if (devname) |
e7b84f9d N |
4133 | pr_err("Cannot read anchor block on %s: %s\n", |
4134 | devname, strerror(errno)); | |
6416d527 | 4135 | free(anchor); |
cdddbdbc DW |
4136 | return 1; |
4137 | } | |
4138 | ||
6416d527 | 4139 | if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) { |
cdddbdbc | 4140 | if (devname) |
e7b84f9d | 4141 | pr_err("no IMSM anchor on %s\n", devname); |
6416d527 | 4142 | free(anchor); |
cdddbdbc DW |
4143 | return 2; |
4144 | } | |
4145 | ||
d23fe947 | 4146 | __free_imsm(super, 0); |
f2f5c343 LM |
4147 | /* reload capability and hba */ |
4148 | ||
4149 | /* capability and hba must be updated with new super allocation */ | |
d424212e | 4150 | find_intel_hba_capability(fd, super, devname); |
f36a9ecd PB |
4151 | super->len = ROUND_UP(anchor->mpb_size, sector_size); |
4152 | if (posix_memalign(&super->buf, MAX_SECTOR_SIZE, super->len) != 0) { | |
cdddbdbc | 4153 | if (devname) |
e7b84f9d N |
4154 | pr_err("unable to allocate %zu byte mpb buffer\n", |
4155 | super->len); | |
6416d527 | 4156 | free(anchor); |
cdddbdbc DW |
4157 | return 2; |
4158 | } | |
f36a9ecd | 4159 | memcpy(super->buf, anchor, sector_size); |
cdddbdbc | 4160 | |
f36a9ecd | 4161 | sectors = mpb_sectors(anchor, sector_size) - 1; |
6416d527 | 4162 | free(anchor); |
8e59f3d8 | 4163 | |
de44e46f PB |
4164 | if (posix_memalign(&super->migr_rec_buf, sector_size, |
4165 | MIGR_REC_BUF_SECTORS*sector_size) != 0) { | |
1ade5cc1 | 4166 | pr_err("could not allocate migr_rec buffer\n"); |
8e59f3d8 AK |
4167 | free(super->buf); |
4168 | return 2; | |
4169 | } | |
51d83f5d | 4170 | super->clean_migration_record_by_mdmon = 0; |
8e59f3d8 | 4171 | |
949c47a0 | 4172 | if (!sectors) { |
ecf45690 DW |
4173 | check_sum = __gen_imsm_checksum(super->anchor); |
4174 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
4175 | if (devname) | |
e7b84f9d N |
4176 | pr_err("IMSM checksum %x != %x on %s\n", |
4177 | check_sum, | |
4178 | __le32_to_cpu(super->anchor->check_sum), | |
4179 | devname); | |
ecf45690 DW |
4180 | return 2; |
4181 | } | |
4182 | ||
a2b97981 | 4183 | return 0; |
949c47a0 | 4184 | } |
cdddbdbc DW |
4185 | |
4186 | /* read the extended mpb */ | |
f36a9ecd | 4187 | if (lseek64(fd, dsize - (sector_size * (2 + sectors)), SEEK_SET) < 0) { |
cdddbdbc | 4188 | if (devname) |
e7b84f9d N |
4189 | pr_err("Cannot seek to extended mpb on %s: %s\n", |
4190 | devname, strerror(errno)); | |
cdddbdbc DW |
4191 | return 1; |
4192 | } | |
4193 | ||
f36a9ecd PB |
4194 | if ((unsigned int)read(fd, super->buf + sector_size, |
4195 | super->len - sector_size) != super->len - sector_size) { | |
cdddbdbc | 4196 | if (devname) |
e7b84f9d N |
4197 | pr_err("Cannot read extended mpb on %s: %s\n", |
4198 | devname, strerror(errno)); | |
cdddbdbc DW |
4199 | return 2; |
4200 | } | |
4201 | ||
949c47a0 DW |
4202 | check_sum = __gen_imsm_checksum(super->anchor); |
4203 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
cdddbdbc | 4204 | if (devname) |
e7b84f9d N |
4205 | pr_err("IMSM checksum %x != %x on %s\n", |
4206 | check_sum, __le32_to_cpu(super->anchor->check_sum), | |
4207 | devname); | |
db575f3b | 4208 | return 3; |
cdddbdbc DW |
4209 | } |
4210 | ||
a2b97981 DW |
4211 | return 0; |
4212 | } | |
4213 | ||
8e59f3d8 AK |
4214 | static int read_imsm_migr_rec(int fd, struct intel_super *super); |
4215 | ||
97f81ee2 CA |
4216 | /* clears hi bits in metadata if MPB_ATTRIB_2TB_DISK not set */ |
4217 | static void clear_hi(struct intel_super *super) | |
4218 | { | |
4219 | struct imsm_super *mpb = super->anchor; | |
4220 | int i, n; | |
4221 | if (mpb->attributes & MPB_ATTRIB_2TB_DISK) | |
4222 | return; | |
4223 | for (i = 0; i < mpb->num_disks; ++i) { | |
4224 | struct imsm_disk *disk = &mpb->disk[i]; | |
4225 | disk->total_blocks_hi = 0; | |
4226 | } | |
4227 | for (i = 0; i < mpb->num_raid_devs; ++i) { | |
4228 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
4229 | if (!dev) | |
4230 | return; | |
4231 | for (n = 0; n < 2; ++n) { | |
4232 | struct imsm_map *map = get_imsm_map(dev, n); | |
4233 | if (!map) | |
4234 | continue; | |
4235 | map->pba_of_lba0_hi = 0; | |
4236 | map->blocks_per_member_hi = 0; | |
4237 | map->num_data_stripes_hi = 0; | |
4238 | } | |
4239 | } | |
4240 | } | |
4241 | ||
a2b97981 DW |
4242 | static int |
4243 | load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd) | |
4244 | { | |
4245 | int err; | |
4246 | ||
4247 | err = load_imsm_mpb(fd, super, devname); | |
4248 | if (err) | |
4249 | return err; | |
f36a9ecd PB |
4250 | if (super->sector_size == 4096) |
4251 | convert_from_4k(super); | |
a2b97981 DW |
4252 | err = load_imsm_disk(fd, super, devname, keep_fd); |
4253 | if (err) | |
4254 | return err; | |
4255 | err = parse_raid_devices(super); | |
8d67477f TM |
4256 | if (err) |
4257 | return err; | |
4258 | err = load_bbm_log(super); | |
97f81ee2 | 4259 | clear_hi(super); |
a2b97981 | 4260 | return err; |
cdddbdbc DW |
4261 | } |
4262 | ||
ae6aad82 DW |
4263 | static void __free_imsm_disk(struct dl *d) |
4264 | { | |
4265 | if (d->fd >= 0) | |
4266 | close(d->fd); | |
4267 | if (d->devname) | |
4268 | free(d->devname); | |
0dcecb2e DW |
4269 | if (d->e) |
4270 | free(d->e); | |
ae6aad82 DW |
4271 | free(d); |
4272 | ||
4273 | } | |
1a64be56 | 4274 | |
cdddbdbc DW |
4275 | static void free_imsm_disks(struct intel_super *super) |
4276 | { | |
47ee5a45 | 4277 | struct dl *d; |
cdddbdbc | 4278 | |
47ee5a45 DW |
4279 | while (super->disks) { |
4280 | d = super->disks; | |
cdddbdbc | 4281 | super->disks = d->next; |
ae6aad82 | 4282 | __free_imsm_disk(d); |
cdddbdbc | 4283 | } |
cb82edca AK |
4284 | while (super->disk_mgmt_list) { |
4285 | d = super->disk_mgmt_list; | |
4286 | super->disk_mgmt_list = d->next; | |
4287 | __free_imsm_disk(d); | |
4288 | } | |
47ee5a45 DW |
4289 | while (super->missing) { |
4290 | d = super->missing; | |
4291 | super->missing = d->next; | |
4292 | __free_imsm_disk(d); | |
4293 | } | |
4294 | ||
cdddbdbc DW |
4295 | } |
4296 | ||
9ca2c81c | 4297 | /* free all the pieces hanging off of a super pointer */ |
d23fe947 | 4298 | static void __free_imsm(struct intel_super *super, int free_disks) |
cdddbdbc | 4299 | { |
88654014 LM |
4300 | struct intel_hba *elem, *next; |
4301 | ||
9ca2c81c | 4302 | if (super->buf) { |
949c47a0 | 4303 | free(super->buf); |
9ca2c81c DW |
4304 | super->buf = NULL; |
4305 | } | |
f2f5c343 LM |
4306 | /* unlink capability description */ |
4307 | super->orom = NULL; | |
8e59f3d8 AK |
4308 | if (super->migr_rec_buf) { |
4309 | free(super->migr_rec_buf); | |
4310 | super->migr_rec_buf = NULL; | |
4311 | } | |
d23fe947 DW |
4312 | if (free_disks) |
4313 | free_imsm_disks(super); | |
ba2de7ba | 4314 | free_devlist(super); |
88654014 LM |
4315 | elem = super->hba; |
4316 | while (elem) { | |
4317 | if (elem->path) | |
4318 | free((void *)elem->path); | |
4319 | next = elem->next; | |
4320 | free(elem); | |
4321 | elem = next; | |
88c32bb1 | 4322 | } |
8d67477f TM |
4323 | if (super->bbm_log) |
4324 | free(super->bbm_log); | |
88654014 | 4325 | super->hba = NULL; |
cdddbdbc DW |
4326 | } |
4327 | ||
9ca2c81c DW |
4328 | static void free_imsm(struct intel_super *super) |
4329 | { | |
d23fe947 | 4330 | __free_imsm(super, 1); |
928f1424 | 4331 | free(super->bb.entries); |
9ca2c81c DW |
4332 | free(super); |
4333 | } | |
cdddbdbc DW |
4334 | |
4335 | static void free_super_imsm(struct supertype *st) | |
4336 | { | |
4337 | struct intel_super *super = st->sb; | |
4338 | ||
4339 | if (!super) | |
4340 | return; | |
4341 | ||
4342 | free_imsm(super); | |
4343 | st->sb = NULL; | |
4344 | } | |
4345 | ||
49133e57 | 4346 | static struct intel_super *alloc_super(void) |
c2c087e6 | 4347 | { |
503975b9 | 4348 | struct intel_super *super = xcalloc(1, sizeof(*super)); |
c2c087e6 | 4349 | |
503975b9 N |
4350 | super->current_vol = -1; |
4351 | super->create_offset = ~((unsigned long long) 0); | |
928f1424 TM |
4352 | |
4353 | super->bb.entries = xmalloc(BBM_LOG_MAX_ENTRIES * | |
4354 | sizeof(struct md_bb_entry)); | |
4355 | if (!super->bb.entries) { | |
4356 | free(super); | |
4357 | return NULL; | |
4358 | } | |
4359 | ||
c2c087e6 DW |
4360 | return super; |
4361 | } | |
4362 | ||
f0f5a016 LM |
4363 | /* |
4364 | * find and allocate hba and OROM/EFI based on valid fd of RAID component device | |
4365 | */ | |
d424212e | 4366 | static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname) |
f0f5a016 LM |
4367 | { |
4368 | struct sys_dev *hba_name; | |
4369 | int rv = 0; | |
4370 | ||
089f9d79 | 4371 | if (fd < 0 || check_env("IMSM_NO_PLATFORM")) { |
f2f5c343 | 4372 | super->orom = NULL; |
f0f5a016 LM |
4373 | super->hba = NULL; |
4374 | return 0; | |
4375 | } | |
4376 | hba_name = find_disk_attached_hba(fd, NULL); | |
4377 | if (!hba_name) { | |
d424212e | 4378 | if (devname) |
e7b84f9d N |
4379 | pr_err("%s is not attached to Intel(R) RAID controller.\n", |
4380 | devname); | |
f0f5a016 LM |
4381 | return 1; |
4382 | } | |
4383 | rv = attach_hba_to_super(super, hba_name); | |
4384 | if (rv == 2) { | |
d424212e N |
4385 | if (devname) { |
4386 | struct intel_hba *hba = super->hba; | |
f0f5a016 | 4387 | |
60f0f54d PB |
4388 | pr_err("%s is attached to Intel(R) %s %s (%s),\n" |
4389 | " but the container is assigned to Intel(R) %s %s (", | |
d424212e | 4390 | devname, |
614902f6 | 4391 | get_sys_dev_type(hba_name->type), |
60f0f54d | 4392 | hba_name->type == SYS_DEV_VMD ? "domain" : "RAID controller", |
f0f5a016 | 4393 | hba_name->pci_id ? : "Err!", |
60f0f54d PB |
4394 | get_sys_dev_type(super->hba->type), |
4395 | hba->type == SYS_DEV_VMD ? "domain" : "RAID controller"); | |
f0f5a016 | 4396 | |
f0f5a016 LM |
4397 | while (hba) { |
4398 | fprintf(stderr, "%s", hba->pci_id ? : "Err!"); | |
4399 | if (hba->next) | |
4400 | fprintf(stderr, ", "); | |
4401 | hba = hba->next; | |
4402 | } | |
6b781d33 | 4403 | fprintf(stderr, ").\n" |
60f0f54d PB |
4404 | " Mixing devices attached to different %s is not allowed.\n", |
4405 | hba_name->type == SYS_DEV_VMD ? "VMD domains" : "controllers"); | |
f0f5a016 | 4406 | } |
f0f5a016 LM |
4407 | return 2; |
4408 | } | |
6b781d33 | 4409 | super->orom = find_imsm_capability(hba_name); |
f2f5c343 LM |
4410 | if (!super->orom) |
4411 | return 3; | |
614902f6 | 4412 | |
f0f5a016 LM |
4413 | return 0; |
4414 | } | |
4415 | ||
47ee5a45 DW |
4416 | /* find_missing - helper routine for load_super_imsm_all that identifies |
4417 | * disks that have disappeared from the system. This routine relies on | |
4418 | * the mpb being uptodate, which it is at load time. | |
4419 | */ | |
4420 | static int find_missing(struct intel_super *super) | |
4421 | { | |
4422 | int i; | |
4423 | struct imsm_super *mpb = super->anchor; | |
4424 | struct dl *dl; | |
4425 | struct imsm_disk *disk; | |
47ee5a45 DW |
4426 | |
4427 | for (i = 0; i < mpb->num_disks; i++) { | |
4428 | disk = __get_imsm_disk(mpb, i); | |
54c2c1ea | 4429 | dl = serial_to_dl(disk->serial, super); |
47ee5a45 DW |
4430 | if (dl) |
4431 | continue; | |
47ee5a45 | 4432 | |
503975b9 | 4433 | dl = xmalloc(sizeof(*dl)); |
47ee5a45 DW |
4434 | dl->major = 0; |
4435 | dl->minor = 0; | |
4436 | dl->fd = -1; | |
503975b9 | 4437 | dl->devname = xstrdup("missing"); |
47ee5a45 DW |
4438 | dl->index = i; |
4439 | serialcpy(dl->serial, disk->serial); | |
4440 | dl->disk = *disk; | |
689c9bf3 | 4441 | dl->e = NULL; |
47ee5a45 DW |
4442 | dl->next = super->missing; |
4443 | super->missing = dl; | |
4444 | } | |
4445 | ||
4446 | return 0; | |
4447 | } | |
4448 | ||
3960e579 | 4449 | #ifndef MDASSEMBLE |
a2b97981 DW |
4450 | static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list) |
4451 | { | |
4452 | struct intel_disk *idisk = disk_list; | |
4453 | ||
4454 | while (idisk) { | |
4455 | if (serialcmp(idisk->disk.serial, serial) == 0) | |
4456 | break; | |
4457 | idisk = idisk->next; | |
4458 | } | |
4459 | ||
4460 | return idisk; | |
4461 | } | |
4462 | ||
4463 | static int __prep_thunderdome(struct intel_super **table, int tbl_size, | |
4464 | struct intel_super *super, | |
4465 | struct intel_disk **disk_list) | |
4466 | { | |
4467 | struct imsm_disk *d = &super->disks->disk; | |
4468 | struct imsm_super *mpb = super->anchor; | |
4469 | int i, j; | |
4470 | ||
4471 | for (i = 0; i < tbl_size; i++) { | |
4472 | struct imsm_super *tbl_mpb = table[i]->anchor; | |
4473 | struct imsm_disk *tbl_d = &table[i]->disks->disk; | |
4474 | ||
4475 | if (tbl_mpb->family_num == mpb->family_num) { | |
4476 | if (tbl_mpb->check_sum == mpb->check_sum) { | |
1ade5cc1 N |
4477 | dprintf("mpb from %d:%d matches %d:%d\n", |
4478 | super->disks->major, | |
a2b97981 DW |
4479 | super->disks->minor, |
4480 | table[i]->disks->major, | |
4481 | table[i]->disks->minor); | |
4482 | break; | |
4483 | } | |
4484 | ||
4485 | if (((is_configured(d) && !is_configured(tbl_d)) || | |
4486 | is_configured(d) == is_configured(tbl_d)) && | |
4487 | tbl_mpb->generation_num < mpb->generation_num) { | |
4488 | /* current version of the mpb is a | |
4489 | * better candidate than the one in | |
4490 | * super_table, but copy over "cross | |
4491 | * generational" status | |
4492 | */ | |
4493 | struct intel_disk *idisk; | |
4494 | ||
1ade5cc1 N |
4495 | dprintf("mpb from %d:%d replaces %d:%d\n", |
4496 | super->disks->major, | |
a2b97981 DW |
4497 | super->disks->minor, |
4498 | table[i]->disks->major, | |
4499 | table[i]->disks->minor); | |
4500 | ||
4501 | idisk = disk_list_get(tbl_d->serial, *disk_list); | |
4502 | if (idisk && is_failed(&idisk->disk)) | |
4503 | tbl_d->status |= FAILED_DISK; | |
4504 | break; | |
4505 | } else { | |
4506 | struct intel_disk *idisk; | |
4507 | struct imsm_disk *disk; | |
4508 | ||
4509 | /* tbl_mpb is more up to date, but copy | |
4510 | * over cross generational status before | |
4511 | * returning | |
4512 | */ | |
4513 | disk = __serial_to_disk(d->serial, mpb, NULL); | |
4514 | if (disk && is_failed(disk)) | |
4515 | d->status |= FAILED_DISK; | |
4516 | ||
4517 | idisk = disk_list_get(d->serial, *disk_list); | |
4518 | if (idisk) { | |
4519 | idisk->owner = i; | |
4520 | if (disk && is_configured(disk)) | |
4521 | idisk->disk.status |= CONFIGURED_DISK; | |
4522 | } | |
4523 | ||
1ade5cc1 N |
4524 | dprintf("mpb from %d:%d prefer %d:%d\n", |
4525 | super->disks->major, | |
a2b97981 DW |
4526 | super->disks->minor, |
4527 | table[i]->disks->major, | |
4528 | table[i]->disks->minor); | |
4529 | ||
4530 | return tbl_size; | |
4531 | } | |
4532 | } | |
4533 | } | |
4534 | ||
4535 | if (i >= tbl_size) | |
4536 | table[tbl_size++] = super; | |
4537 | else | |
4538 | table[i] = super; | |
4539 | ||
4540 | /* update/extend the merged list of imsm_disk records */ | |
4541 | for (j = 0; j < mpb->num_disks; j++) { | |
4542 | struct imsm_disk *disk = __get_imsm_disk(mpb, j); | |
4543 | struct intel_disk *idisk; | |
4544 | ||
4545 | idisk = disk_list_get(disk->serial, *disk_list); | |
4546 | if (idisk) { | |
4547 | idisk->disk.status |= disk->status; | |
4548 | if (is_configured(&idisk->disk) || | |
4549 | is_failed(&idisk->disk)) | |
4550 | idisk->disk.status &= ~(SPARE_DISK); | |
4551 | } else { | |
503975b9 | 4552 | idisk = xcalloc(1, sizeof(*idisk)); |
a2b97981 DW |
4553 | idisk->owner = IMSM_UNKNOWN_OWNER; |
4554 | idisk->disk = *disk; | |
4555 | idisk->next = *disk_list; | |
4556 | *disk_list = idisk; | |
4557 | } | |
4558 | ||
4559 | if (serialcmp(idisk->disk.serial, d->serial) == 0) | |
4560 | idisk->owner = i; | |
4561 | } | |
4562 | ||
4563 | return tbl_size; | |
4564 | } | |
4565 | ||
4566 | static struct intel_super * | |
4567 | validate_members(struct intel_super *super, struct intel_disk *disk_list, | |
4568 | const int owner) | |
4569 | { | |
4570 | struct imsm_super *mpb = super->anchor; | |
4571 | int ok_count = 0; | |
4572 | int i; | |
4573 | ||
4574 | for (i = 0; i < mpb->num_disks; i++) { | |
4575 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
4576 | struct intel_disk *idisk; | |
4577 | ||
4578 | idisk = disk_list_get(disk->serial, disk_list); | |
4579 | if (idisk) { | |
4580 | if (idisk->owner == owner || | |
4581 | idisk->owner == IMSM_UNKNOWN_OWNER) | |
4582 | ok_count++; | |
4583 | else | |
1ade5cc1 N |
4584 | dprintf("'%.16s' owner %d != %d\n", |
4585 | disk->serial, idisk->owner, | |
a2b97981 DW |
4586 | owner); |
4587 | } else { | |
1ade5cc1 N |
4588 | dprintf("unknown disk %x [%d]: %.16s\n", |
4589 | __le32_to_cpu(mpb->family_num), i, | |
a2b97981 DW |
4590 | disk->serial); |
4591 | break; | |
4592 | } | |
4593 | } | |
4594 | ||
4595 | if (ok_count == mpb->num_disks) | |
4596 | return super; | |
4597 | return NULL; | |
4598 | } | |
4599 | ||
4600 | static void show_conflicts(__u32 family_num, struct intel_super *super_list) | |
4601 | { | |
4602 | struct intel_super *s; | |
4603 | ||
4604 | for (s = super_list; s; s = s->next) { | |
4605 | if (family_num != s->anchor->family_num) | |
4606 | continue; | |
e12b3daa | 4607 | pr_err("Conflict, offlining family %#x on '%s'\n", |
a2b97981 DW |
4608 | __le32_to_cpu(family_num), s->disks->devname); |
4609 | } | |
4610 | } | |
4611 | ||
4612 | static struct intel_super * | |
4613 | imsm_thunderdome(struct intel_super **super_list, int len) | |
4614 | { | |
4615 | struct intel_super *super_table[len]; | |
4616 | struct intel_disk *disk_list = NULL; | |
4617 | struct intel_super *champion, *spare; | |
4618 | struct intel_super *s, **del; | |
4619 | int tbl_size = 0; | |
4620 | int conflict; | |
4621 | int i; | |
4622 | ||
4623 | memset(super_table, 0, sizeof(super_table)); | |
4624 | for (s = *super_list; s; s = s->next) | |
4625 | tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list); | |
4626 | ||
4627 | for (i = 0; i < tbl_size; i++) { | |
4628 | struct imsm_disk *d; | |
4629 | struct intel_disk *idisk; | |
4630 | struct imsm_super *mpb = super_table[i]->anchor; | |
4631 | ||
4632 | s = super_table[i]; | |
4633 | d = &s->disks->disk; | |
4634 | ||
4635 | /* 'd' must appear in merged disk list for its | |
4636 | * configuration to be valid | |
4637 | */ | |
4638 | idisk = disk_list_get(d->serial, disk_list); | |
4639 | if (idisk && idisk->owner == i) | |
4640 | s = validate_members(s, disk_list, i); | |
4641 | else | |
4642 | s = NULL; | |
4643 | ||
4644 | if (!s) | |
1ade5cc1 N |
4645 | dprintf("marking family: %#x from %d:%d offline\n", |
4646 | mpb->family_num, | |
a2b97981 DW |
4647 | super_table[i]->disks->major, |
4648 | super_table[i]->disks->minor); | |
4649 | super_table[i] = s; | |
4650 | } | |
4651 | ||
4652 | /* This is where the mdadm implementation differs from the Windows | |
4653 | * driver which has no strict concept of a container. We can only | |
4654 | * assemble one family from a container, so when returning a prodigal | |
4655 | * array member to this system the code will not be able to disambiguate | |
4656 | * the container contents that should be assembled ("foreign" versus | |
4657 | * "local"). It requires user intervention to set the orig_family_num | |
4658 | * to a new value to establish a new container. The Windows driver in | |
4659 | * this situation fixes up the volume name in place and manages the | |
4660 | * foreign array as an independent entity. | |
4661 | */ | |
4662 | s = NULL; | |
4663 | spare = NULL; | |
4664 | conflict = 0; | |
4665 | for (i = 0; i < tbl_size; i++) { | |
4666 | struct intel_super *tbl_ent = super_table[i]; | |
4667 | int is_spare = 0; | |
4668 | ||
4669 | if (!tbl_ent) | |
4670 | continue; | |
4671 | ||
4672 | if (tbl_ent->anchor->num_raid_devs == 0) { | |
4673 | spare = tbl_ent; | |
4674 | is_spare = 1; | |
4675 | } | |
4676 | ||
4677 | if (s && !is_spare) { | |
4678 | show_conflicts(tbl_ent->anchor->family_num, *super_list); | |
4679 | conflict++; | |
4680 | } else if (!s && !is_spare) | |
4681 | s = tbl_ent; | |
4682 | } | |
4683 | ||
4684 | if (!s) | |
4685 | s = spare; | |
4686 | if (!s) { | |
4687 | champion = NULL; | |
4688 | goto out; | |
4689 | } | |
4690 | champion = s; | |
4691 | ||
4692 | if (conflict) | |
7a862a02 | 4693 | pr_err("Chose family %#x on '%s', assemble conflicts to new container with '--update=uuid'\n", |
a2b97981 DW |
4694 | __le32_to_cpu(s->anchor->family_num), s->disks->devname); |
4695 | ||
4696 | /* collect all dl's onto 'champion', and update them to | |
4697 | * champion's version of the status | |
4698 | */ | |
4699 | for (s = *super_list; s; s = s->next) { | |
4700 | struct imsm_super *mpb = champion->anchor; | |
4701 | struct dl *dl = s->disks; | |
4702 | ||
4703 | if (s == champion) | |
4704 | continue; | |
4705 | ||
5d7b407a CA |
4706 | mpb->attributes |= s->anchor->attributes & MPB_ATTRIB_2TB_DISK; |
4707 | ||
a2b97981 DW |
4708 | for (i = 0; i < mpb->num_disks; i++) { |
4709 | struct imsm_disk *disk; | |
4710 | ||
4711 | disk = __serial_to_disk(dl->serial, mpb, &dl->index); | |
4712 | if (disk) { | |
4713 | dl->disk = *disk; | |
4714 | /* only set index on disks that are a member of | |
4715 | * a populated contianer, i.e. one with | |
4716 | * raid_devs | |
4717 | */ | |
4718 | if (is_failed(&dl->disk)) | |
4719 | dl->index = -2; | |
4720 | else if (is_spare(&dl->disk)) | |
4721 | dl->index = -1; | |
4722 | break; | |
4723 | } | |
4724 | } | |
4725 | ||
4726 | if (i >= mpb->num_disks) { | |
4727 | struct intel_disk *idisk; | |
4728 | ||
4729 | idisk = disk_list_get(dl->serial, disk_list); | |
ecf408e9 | 4730 | if (idisk && is_spare(&idisk->disk) && |
a2b97981 DW |
4731 | !is_failed(&idisk->disk) && !is_configured(&idisk->disk)) |
4732 | dl->index = -1; | |
4733 | else { | |
4734 | dl->index = -2; | |
4735 | continue; | |
4736 | } | |
4737 | } | |
4738 | ||
4739 | dl->next = champion->disks; | |
4740 | champion->disks = dl; | |
4741 | s->disks = NULL; | |
4742 | } | |
4743 | ||
4744 | /* delete 'champion' from super_list */ | |
4745 | for (del = super_list; *del; ) { | |
4746 | if (*del == champion) { | |
4747 | *del = (*del)->next; | |
4748 | break; | |
4749 | } else | |
4750 | del = &(*del)->next; | |
4751 | } | |
4752 | champion->next = NULL; | |
4753 | ||
4754 | out: | |
4755 | while (disk_list) { | |
4756 | struct intel_disk *idisk = disk_list; | |
4757 | ||
4758 | disk_list = disk_list->next; | |
4759 | free(idisk); | |
4760 | } | |
4761 | ||
4762 | return champion; | |
4763 | } | |
4764 | ||
9587c373 LM |
4765 | static int |
4766 | get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd); | |
4dd2df09 | 4767 | static int get_super_block(struct intel_super **super_list, char *devnm, char *devname, |
9587c373 | 4768 | int major, int minor, int keep_fd); |
ec50f7b6 LM |
4769 | static int |
4770 | get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list, | |
4771 | int *max, int keep_fd); | |
4772 | ||
cdddbdbc | 4773 | static int load_super_imsm_all(struct supertype *st, int fd, void **sbp, |
ec50f7b6 LM |
4774 | char *devname, struct md_list *devlist, |
4775 | int keep_fd) | |
cdddbdbc | 4776 | { |
a2b97981 DW |
4777 | struct intel_super *super_list = NULL; |
4778 | struct intel_super *super = NULL; | |
a2b97981 | 4779 | int err = 0; |
9587c373 | 4780 | int i = 0; |
dab4a513 | 4781 | |
9587c373 LM |
4782 | if (fd >= 0) |
4783 | /* 'fd' is an opened container */ | |
4784 | err = get_sra_super_block(fd, &super_list, devname, &i, keep_fd); | |
4785 | else | |
ec50f7b6 LM |
4786 | /* get super block from devlist devices */ |
4787 | err = get_devlist_super_block(devlist, &super_list, &i, keep_fd); | |
9587c373 | 4788 | if (err) |
1602d52c | 4789 | goto error; |
a2b97981 DW |
4790 | /* all mpbs enter, maybe one leaves */ |
4791 | super = imsm_thunderdome(&super_list, i); | |
4792 | if (!super) { | |
4793 | err = 1; | |
4794 | goto error; | |
cdddbdbc DW |
4795 | } |
4796 | ||
47ee5a45 DW |
4797 | if (find_missing(super) != 0) { |
4798 | free_imsm(super); | |
a2b97981 DW |
4799 | err = 2; |
4800 | goto error; | |
47ee5a45 | 4801 | } |
8e59f3d8 AK |
4802 | |
4803 | /* load migration record */ | |
4804 | err = load_imsm_migr_rec(super, NULL); | |
4c965cc9 AK |
4805 | if (err == -1) { |
4806 | /* migration is in progress, | |
4807 | * but migr_rec cannot be loaded, | |
4808 | */ | |
8e59f3d8 AK |
4809 | err = 4; |
4810 | goto error; | |
4811 | } | |
e2f41b2c AK |
4812 | |
4813 | /* Check migration compatibility */ | |
089f9d79 | 4814 | if (err == 0 && check_mpb_migr_compatibility(super) != 0) { |
e7b84f9d | 4815 | pr_err("Unsupported migration detected"); |
e2f41b2c AK |
4816 | if (devname) |
4817 | fprintf(stderr, " on %s\n", devname); | |
4818 | else | |
4819 | fprintf(stderr, " (IMSM).\n"); | |
4820 | ||
4821 | err = 5; | |
4822 | goto error; | |
4823 | } | |
4824 | ||
a2b97981 DW |
4825 | err = 0; |
4826 | ||
4827 | error: | |
4828 | while (super_list) { | |
4829 | struct intel_super *s = super_list; | |
4830 | ||
4831 | super_list = super_list->next; | |
4832 | free_imsm(s); | |
4833 | } | |
9587c373 | 4834 | |
a2b97981 DW |
4835 | if (err) |
4836 | return err; | |
f7e7067b | 4837 | |
cdddbdbc | 4838 | *sbp = super; |
9587c373 | 4839 | if (fd >= 0) |
4dd2df09 | 4840 | strcpy(st->container_devnm, fd2devnm(fd)); |
9587c373 | 4841 | else |
4dd2df09 | 4842 | st->container_devnm[0] = 0; |
a2b97981 | 4843 | if (err == 0 && st->ss == NULL) { |
bf5a934a | 4844 | st->ss = &super_imsm; |
cdddbdbc DW |
4845 | st->minor_version = 0; |
4846 | st->max_devs = IMSM_MAX_DEVICES; | |
4847 | } | |
cdddbdbc DW |
4848 | return 0; |
4849 | } | |
2b959fbf | 4850 | |
ec50f7b6 LM |
4851 | static int |
4852 | get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list, | |
4853 | int *max, int keep_fd) | |
4854 | { | |
4855 | struct md_list *tmpdev; | |
4856 | int err = 0; | |
4857 | int i = 0; | |
9587c373 | 4858 | |
ec50f7b6 LM |
4859 | for (i = 0, tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { |
4860 | if (tmpdev->used != 1) | |
4861 | continue; | |
4862 | if (tmpdev->container == 1) { | |
ca9de185 | 4863 | int lmax = 0; |
ec50f7b6 LM |
4864 | int fd = dev_open(tmpdev->devname, O_RDONLY|O_EXCL); |
4865 | if (fd < 0) { | |
e7b84f9d | 4866 | pr_err("cannot open device %s: %s\n", |
ec50f7b6 LM |
4867 | tmpdev->devname, strerror(errno)); |
4868 | err = 8; | |
4869 | goto error; | |
4870 | } | |
4871 | err = get_sra_super_block(fd, super_list, | |
4872 | tmpdev->devname, &lmax, | |
4873 | keep_fd); | |
4874 | i += lmax; | |
4875 | close(fd); | |
4876 | if (err) { | |
4877 | err = 7; | |
4878 | goto error; | |
4879 | } | |
4880 | } else { | |
4881 | int major = major(tmpdev->st_rdev); | |
4882 | int minor = minor(tmpdev->st_rdev); | |
4883 | err = get_super_block(super_list, | |
4dd2df09 | 4884 | NULL, |
ec50f7b6 LM |
4885 | tmpdev->devname, |
4886 | major, minor, | |
4887 | keep_fd); | |
4888 | i++; | |
4889 | if (err) { | |
4890 | err = 6; | |
4891 | goto error; | |
4892 | } | |
4893 | } | |
4894 | } | |
4895 | error: | |
4896 | *max = i; | |
4897 | return err; | |
4898 | } | |
9587c373 | 4899 | |
4dd2df09 | 4900 | static int get_super_block(struct intel_super **super_list, char *devnm, char *devname, |
9587c373 LM |
4901 | int major, int minor, int keep_fd) |
4902 | { | |
594dc1b8 | 4903 | struct intel_super *s; |
9587c373 LM |
4904 | char nm[32]; |
4905 | int dfd = -1; | |
9587c373 LM |
4906 | int err = 0; |
4907 | int retry; | |
4908 | ||
4909 | s = alloc_super(); | |
4910 | if (!s) { | |
4911 | err = 1; | |
4912 | goto error; | |
4913 | } | |
4914 | ||
4915 | sprintf(nm, "%d:%d", major, minor); | |
4916 | dfd = dev_open(nm, O_RDWR); | |
4917 | if (dfd < 0) { | |
4918 | err = 2; | |
4919 | goto error; | |
4920 | } | |
4921 | ||
fa7bb6f8 | 4922 | get_dev_sector_size(dfd, NULL, &s->sector_size); |
cb8f6859 | 4923 | find_intel_hba_capability(dfd, s, devname); |
9587c373 LM |
4924 | err = load_and_parse_mpb(dfd, s, NULL, keep_fd); |
4925 | ||
4926 | /* retry the load if we might have raced against mdmon */ | |
4dd2df09 | 4927 | if (err == 3 && devnm && mdmon_running(devnm)) |
9587c373 LM |
4928 | for (retry = 0; retry < 3; retry++) { |
4929 | usleep(3000); | |
4930 | err = load_and_parse_mpb(dfd, s, NULL, keep_fd); | |
4931 | if (err != 3) | |
4932 | break; | |
4933 | } | |
4934 | error: | |
4935 | if (!err) { | |
4936 | s->next = *super_list; | |
4937 | *super_list = s; | |
4938 | } else { | |
4939 | if (s) | |
8d67477f | 4940 | free_imsm(s); |
36614e95 | 4941 | if (dfd >= 0) |
9587c373 LM |
4942 | close(dfd); |
4943 | } | |
089f9d79 | 4944 | if (dfd >= 0 && !keep_fd) |
9587c373 LM |
4945 | close(dfd); |
4946 | return err; | |
4947 | ||
4948 | } | |
4949 | ||
4950 | static int | |
4951 | get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd) | |
4952 | { | |
4953 | struct mdinfo *sra; | |
4dd2df09 | 4954 | char *devnm; |
9587c373 LM |
4955 | struct mdinfo *sd; |
4956 | int err = 0; | |
4957 | int i = 0; | |
4dd2df09 | 4958 | sra = sysfs_read(fd, NULL, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE); |
9587c373 LM |
4959 | if (!sra) |
4960 | return 1; | |
4961 | ||
4962 | if (sra->array.major_version != -1 || | |
4963 | sra->array.minor_version != -2 || | |
4964 | strcmp(sra->text_version, "imsm") != 0) { | |
4965 | err = 1; | |
4966 | goto error; | |
4967 | } | |
4968 | /* load all mpbs */ | |
4dd2df09 | 4969 | devnm = fd2devnm(fd); |
9587c373 | 4970 | for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) { |
4dd2df09 | 4971 | if (get_super_block(super_list, devnm, devname, |
9587c373 LM |
4972 | sd->disk.major, sd->disk.minor, keep_fd) != 0) { |
4973 | err = 7; | |
4974 | goto error; | |
4975 | } | |
4976 | } | |
4977 | error: | |
4978 | sysfs_free(sra); | |
4979 | *max = i; | |
4980 | return err; | |
4981 | } | |
4982 | ||
2b959fbf N |
4983 | static int load_container_imsm(struct supertype *st, int fd, char *devname) |
4984 | { | |
ec50f7b6 | 4985 | return load_super_imsm_all(st, fd, &st->sb, devname, NULL, 1); |
2b959fbf | 4986 | } |
cdddbdbc DW |
4987 | #endif |
4988 | ||
4989 | static int load_super_imsm(struct supertype *st, int fd, char *devname) | |
4990 | { | |
4991 | struct intel_super *super; | |
4992 | int rv; | |
8a3544f8 | 4993 | int retry; |
cdddbdbc | 4994 | |
357ac106 | 4995 | if (test_partition(fd)) |
691c6ee1 N |
4996 | /* IMSM not allowed on partitions */ |
4997 | return 1; | |
4998 | ||
37424f13 DW |
4999 | free_super_imsm(st); |
5000 | ||
49133e57 | 5001 | super = alloc_super(); |
fa7bb6f8 | 5002 | get_dev_sector_size(fd, NULL, &super->sector_size); |
8d67477f TM |
5003 | if (!super) |
5004 | return 1; | |
ea2bc72b LM |
5005 | /* Load hba and capabilities if they exist. |
5006 | * But do not preclude loading metadata in case capabilities or hba are | |
5007 | * non-compliant and ignore_hw_compat is set. | |
5008 | */ | |
d424212e | 5009 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 5010 | /* no orom/efi or non-intel hba of the disk */ |
089f9d79 | 5011 | if (rv != 0 && st->ignore_hw_compat == 0) { |
f2f5c343 | 5012 | if (devname) |
e7b84f9d | 5013 | pr_err("No OROM/EFI properties for %s\n", devname); |
f2f5c343 LM |
5014 | free_imsm(super); |
5015 | return 2; | |
5016 | } | |
a2b97981 | 5017 | rv = load_and_parse_mpb(fd, super, devname, 0); |
cdddbdbc | 5018 | |
8a3544f8 AP |
5019 | /* retry the load if we might have raced against mdmon */ |
5020 | if (rv == 3) { | |
f96b1302 AP |
5021 | struct mdstat_ent *mdstat = NULL; |
5022 | char *name = fd2kname(fd); | |
5023 | ||
5024 | if (name) | |
5025 | mdstat = mdstat_by_component(name); | |
8a3544f8 AP |
5026 | |
5027 | if (mdstat && mdmon_running(mdstat->devnm) && getpid() != mdmon_pid(mdstat->devnm)) { | |
5028 | for (retry = 0; retry < 3; retry++) { | |
5029 | usleep(3000); | |
5030 | rv = load_and_parse_mpb(fd, super, devname, 0); | |
5031 | if (rv != 3) | |
5032 | break; | |
5033 | } | |
5034 | } | |
5035 | ||
5036 | free_mdstat(mdstat); | |
5037 | } | |
5038 | ||
cdddbdbc DW |
5039 | if (rv) { |
5040 | if (devname) | |
7a862a02 | 5041 | pr_err("Failed to load all information sections on %s\n", devname); |
cdddbdbc DW |
5042 | free_imsm(super); |
5043 | return rv; | |
5044 | } | |
5045 | ||
5046 | st->sb = super; | |
5047 | if (st->ss == NULL) { | |
5048 | st->ss = &super_imsm; | |
5049 | st->minor_version = 0; | |
5050 | st->max_devs = IMSM_MAX_DEVICES; | |
5051 | } | |
8e59f3d8 AK |
5052 | |
5053 | /* load migration record */ | |
2e062e82 AK |
5054 | if (load_imsm_migr_rec(super, NULL) == 0) { |
5055 | /* Check for unsupported migration features */ | |
5056 | if (check_mpb_migr_compatibility(super) != 0) { | |
e7b84f9d | 5057 | pr_err("Unsupported migration detected"); |
2e062e82 AK |
5058 | if (devname) |
5059 | fprintf(stderr, " on %s\n", devname); | |
5060 | else | |
5061 | fprintf(stderr, " (IMSM).\n"); | |
5062 | return 3; | |
5063 | } | |
e2f41b2c AK |
5064 | } |
5065 | ||
cdddbdbc DW |
5066 | return 0; |
5067 | } | |
5068 | ||
ef6ffade DW |
5069 | static __u16 info_to_blocks_per_strip(mdu_array_info_t *info) |
5070 | { | |
5071 | if (info->level == 1) | |
5072 | return 128; | |
5073 | return info->chunk_size >> 9; | |
5074 | } | |
5075 | ||
5551b113 CA |
5076 | static unsigned long long info_to_blocks_per_member(mdu_array_info_t *info, |
5077 | unsigned long long size) | |
fcfd9599 | 5078 | { |
4025c288 | 5079 | if (info->level == 1) |
5551b113 | 5080 | return size * 2; |
4025c288 | 5081 | else |
5551b113 | 5082 | return (size * 2) & ~(info_to_blocks_per_strip(info) - 1); |
fcfd9599 DW |
5083 | } |
5084 | ||
4d1313e9 DW |
5085 | static void imsm_update_version_info(struct intel_super *super) |
5086 | { | |
5087 | /* update the version and attributes */ | |
5088 | struct imsm_super *mpb = super->anchor; | |
5089 | char *version; | |
5090 | struct imsm_dev *dev; | |
5091 | struct imsm_map *map; | |
5092 | int i; | |
5093 | ||
5094 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
5095 | dev = get_imsm_dev(super, i); | |
238c0a71 | 5096 | map = get_imsm_map(dev, MAP_0); |
4d1313e9 DW |
5097 | if (__le32_to_cpu(dev->size_high) > 0) |
5098 | mpb->attributes |= MPB_ATTRIB_2TB; | |
5099 | ||
5100 | /* FIXME detect when an array spans a port multiplier */ | |
5101 | #if 0 | |
5102 | mpb->attributes |= MPB_ATTRIB_PM; | |
5103 | #endif | |
5104 | ||
5105 | if (mpb->num_raid_devs > 1 || | |
5106 | mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) { | |
5107 | version = MPB_VERSION_ATTRIBS; | |
5108 | switch (get_imsm_raid_level(map)) { | |
5109 | case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break; | |
5110 | case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break; | |
5111 | case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break; | |
5112 | case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break; | |
5113 | } | |
5114 | } else { | |
5115 | if (map->num_members >= 5) | |
5116 | version = MPB_VERSION_5OR6_DISK_ARRAY; | |
5117 | else if (dev->status == DEV_CLONE_N_GO) | |
5118 | version = MPB_VERSION_CNG; | |
5119 | else if (get_imsm_raid_level(map) == 5) | |
5120 | version = MPB_VERSION_RAID5; | |
5121 | else if (map->num_members >= 3) | |
5122 | version = MPB_VERSION_3OR4_DISK_ARRAY; | |
5123 | else if (get_imsm_raid_level(map) == 1) | |
5124 | version = MPB_VERSION_RAID1; | |
5125 | else | |
5126 | version = MPB_VERSION_RAID0; | |
5127 | } | |
5128 | strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version); | |
5129 | } | |
5130 | } | |
5131 | ||
aa534678 DW |
5132 | static int check_name(struct intel_super *super, char *name, int quiet) |
5133 | { | |
5134 | struct imsm_super *mpb = super->anchor; | |
5135 | char *reason = NULL; | |
5136 | int i; | |
5137 | ||
5138 | if (strlen(name) > MAX_RAID_SERIAL_LEN) | |
5139 | reason = "must be 16 characters or less"; | |
5140 | ||
5141 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
5142 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
5143 | ||
5144 | if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) { | |
5145 | reason = "already exists"; | |
5146 | break; | |
5147 | } | |
5148 | } | |
5149 | ||
5150 | if (reason && !quiet) | |
e7b84f9d | 5151 | pr_err("imsm volume name %s\n", reason); |
aa534678 DW |
5152 | |
5153 | return !reason; | |
5154 | } | |
5155 | ||
8b353278 DW |
5156 | static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info, |
5157 | unsigned long long size, char *name, | |
83cd1e97 N |
5158 | char *homehost, int *uuid, |
5159 | long long data_offset) | |
cdddbdbc | 5160 | { |
c2c087e6 DW |
5161 | /* We are creating a volume inside a pre-existing container. |
5162 | * so st->sb is already set. | |
5163 | */ | |
5164 | struct intel_super *super = st->sb; | |
f36a9ecd | 5165 | unsigned int sector_size = super->sector_size; |
949c47a0 | 5166 | struct imsm_super *mpb = super->anchor; |
ba2de7ba | 5167 | struct intel_dev *dv; |
c2c087e6 DW |
5168 | struct imsm_dev *dev; |
5169 | struct imsm_vol *vol; | |
5170 | struct imsm_map *map; | |
5171 | int idx = mpb->num_raid_devs; | |
5172 | int i; | |
5173 | unsigned long long array_blocks; | |
2c092cad | 5174 | size_t size_old, size_new; |
5551b113 | 5175 | unsigned long long num_data_stripes; |
cdddbdbc | 5176 | |
88c32bb1 | 5177 | if (super->orom && mpb->num_raid_devs >= super->orom->vpa) { |
7a862a02 | 5178 | pr_err("This imsm-container already has the maximum of %d volumes\n", super->orom->vpa); |
c2c087e6 DW |
5179 | return 0; |
5180 | } | |
5181 | ||
2c092cad DW |
5182 | /* ensure the mpb is large enough for the new data */ |
5183 | size_old = __le32_to_cpu(mpb->mpb_size); | |
5184 | size_new = disks_to_mpb_size(info->nr_disks); | |
5185 | if (size_new > size_old) { | |
5186 | void *mpb_new; | |
f36a9ecd | 5187 | size_t size_round = ROUND_UP(size_new, sector_size); |
2c092cad | 5188 | |
f36a9ecd | 5189 | if (posix_memalign(&mpb_new, sector_size, size_round) != 0) { |
e7b84f9d | 5190 | pr_err("could not allocate new mpb\n"); |
2c092cad DW |
5191 | return 0; |
5192 | } | |
de44e46f PB |
5193 | if (posix_memalign(&super->migr_rec_buf, sector_size, |
5194 | MIGR_REC_BUF_SECTORS*sector_size) != 0) { | |
1ade5cc1 | 5195 | pr_err("could not allocate migr_rec buffer\n"); |
8e59f3d8 AK |
5196 | free(super->buf); |
5197 | free(super); | |
ea944c8f | 5198 | free(mpb_new); |
8e59f3d8 AK |
5199 | return 0; |
5200 | } | |
2c092cad DW |
5201 | memcpy(mpb_new, mpb, size_old); |
5202 | free(mpb); | |
5203 | mpb = mpb_new; | |
949c47a0 | 5204 | super->anchor = mpb_new; |
2c092cad DW |
5205 | mpb->mpb_size = __cpu_to_le32(size_new); |
5206 | memset(mpb_new + size_old, 0, size_round - size_old); | |
bbab0940 | 5207 | super->len = size_round; |
2c092cad | 5208 | } |
bf5a934a | 5209 | super->current_vol = idx; |
3960e579 DW |
5210 | |
5211 | /* handle 'failed_disks' by either: | |
5212 | * a) create dummy disk entries in the table if this the first | |
5213 | * volume in the array. We add them here as this is the only | |
5214 | * opportunity to add them. add_to_super_imsm_volume() | |
5215 | * handles the non-failed disks and continues incrementing | |
5216 | * mpb->num_disks. | |
5217 | * b) validate that 'failed_disks' matches the current number | |
5218 | * of missing disks if the container is populated | |
d23fe947 | 5219 | */ |
3960e579 | 5220 | if (super->current_vol == 0) { |
d23fe947 | 5221 | mpb->num_disks = 0; |
3960e579 DW |
5222 | for (i = 0; i < info->failed_disks; i++) { |
5223 | struct imsm_disk *disk; | |
5224 | ||
5225 | mpb->num_disks++; | |
5226 | disk = __get_imsm_disk(mpb, i); | |
5227 | disk->status = CONFIGURED_DISK | FAILED_DISK; | |
5228 | disk->scsi_id = __cpu_to_le32(~(__u32)0); | |
5229 | snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN, | |
5230 | "missing:%d", i); | |
5231 | } | |
5232 | find_missing(super); | |
5233 | } else { | |
5234 | int missing = 0; | |
5235 | struct dl *d; | |
5236 | ||
5237 | for (d = super->missing; d; d = d->next) | |
5238 | missing++; | |
5239 | if (info->failed_disks > missing) { | |
e7b84f9d | 5240 | pr_err("unable to add 'missing' disk to container\n"); |
3960e579 DW |
5241 | return 0; |
5242 | } | |
5243 | } | |
5a038140 | 5244 | |
aa534678 DW |
5245 | if (!check_name(super, name, 0)) |
5246 | return 0; | |
503975b9 N |
5247 | dv = xmalloc(sizeof(*dv)); |
5248 | dev = xcalloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1)); | |
c2c087e6 | 5249 | strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN); |
e03640bd | 5250 | array_blocks = calc_array_size(info->level, info->raid_disks, |
03bcbc65 | 5251 | info->layout, info->chunk_size, |
5551b113 | 5252 | size * 2); |
979d38be DW |
5253 | /* round array size down to closest MB */ |
5254 | array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; | |
5255 | ||
c2c087e6 DW |
5256 | dev->size_low = __cpu_to_le32((__u32) array_blocks); |
5257 | dev->size_high = __cpu_to_le32((__u32) (array_blocks >> 32)); | |
1a2487c2 | 5258 | dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING); |
c2c087e6 DW |
5259 | vol = &dev->vol; |
5260 | vol->migr_state = 0; | |
1484e727 | 5261 | set_migr_type(dev, MIGR_INIT); |
3960e579 | 5262 | vol->dirty = !info->state; |
f8f603f1 | 5263 | vol->curr_migr_unit = 0; |
238c0a71 | 5264 | map = get_imsm_map(dev, MAP_0); |
5551b113 CA |
5265 | set_pba_of_lba0(map, super->create_offset); |
5266 | set_blocks_per_member(map, info_to_blocks_per_member(info, size)); | |
ef6ffade | 5267 | map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info)); |
0556e1a2 | 5268 | map->failed_disk_num = ~0; |
bf4442ab | 5269 | if (info->level > 0) |
fffaf1ff N |
5270 | map->map_state = (info->state ? IMSM_T_STATE_NORMAL |
5271 | : IMSM_T_STATE_UNINITIALIZED); | |
bf4442ab AK |
5272 | else |
5273 | map->map_state = info->failed_disks ? IMSM_T_STATE_FAILED : | |
5274 | IMSM_T_STATE_NORMAL; | |
252d23c0 | 5275 | map->ddf = 1; |
ef6ffade DW |
5276 | |
5277 | if (info->level == 1 && info->raid_disks > 2) { | |
38950822 AW |
5278 | free(dev); |
5279 | free(dv); | |
7a862a02 | 5280 | pr_err("imsm does not support more than 2 disksin a raid1 volume\n"); |
ef6ffade DW |
5281 | return 0; |
5282 | } | |
81062a36 DW |
5283 | |
5284 | map->raid_level = info->level; | |
4d1313e9 | 5285 | if (info->level == 10) { |
c2c087e6 | 5286 | map->raid_level = 1; |
4d1313e9 | 5287 | map->num_domains = info->raid_disks / 2; |
81062a36 DW |
5288 | } else if (info->level == 1) |
5289 | map->num_domains = info->raid_disks; | |
5290 | else | |
ff596308 | 5291 | map->num_domains = 1; |
81062a36 | 5292 | |
5551b113 CA |
5293 | /* info->size is only int so use the 'size' parameter instead */ |
5294 | num_data_stripes = (size * 2) / info_to_blocks_per_strip(info); | |
5295 | num_data_stripes /= map->num_domains; | |
5296 | set_num_data_stripes(map, num_data_stripes); | |
ef6ffade | 5297 | |
c2c087e6 DW |
5298 | map->num_members = info->raid_disks; |
5299 | for (i = 0; i < map->num_members; i++) { | |
5300 | /* initialized in add_to_super */ | |
4eb26970 | 5301 | set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD); |
c2c087e6 | 5302 | } |
949c47a0 | 5303 | mpb->num_raid_devs++; |
ba2de7ba DW |
5304 | |
5305 | dv->dev = dev; | |
5306 | dv->index = super->current_vol; | |
5307 | dv->next = super->devlist; | |
5308 | super->devlist = dv; | |
c2c087e6 | 5309 | |
4d1313e9 DW |
5310 | imsm_update_version_info(super); |
5311 | ||
c2c087e6 | 5312 | return 1; |
cdddbdbc DW |
5313 | } |
5314 | ||
bf5a934a DW |
5315 | static int init_super_imsm(struct supertype *st, mdu_array_info_t *info, |
5316 | unsigned long long size, char *name, | |
83cd1e97 N |
5317 | char *homehost, int *uuid, |
5318 | unsigned long long data_offset) | |
bf5a934a DW |
5319 | { |
5320 | /* This is primarily called by Create when creating a new array. | |
5321 | * We will then get add_to_super called for each component, and then | |
5322 | * write_init_super called to write it out to each device. | |
5323 | * For IMSM, Create can create on fresh devices or on a pre-existing | |
5324 | * array. | |
5325 | * To create on a pre-existing array a different method will be called. | |
5326 | * This one is just for fresh drives. | |
5327 | */ | |
5328 | struct intel_super *super; | |
5329 | struct imsm_super *mpb; | |
5330 | size_t mpb_size; | |
4d1313e9 | 5331 | char *version; |
bf5a934a | 5332 | |
83cd1e97 | 5333 | if (data_offset != INVALID_SECTORS) { |
ed503f89 | 5334 | pr_err("data-offset not supported by imsm\n"); |
83cd1e97 N |
5335 | return 0; |
5336 | } | |
5337 | ||
bf5a934a | 5338 | if (st->sb) |
83cd1e97 N |
5339 | return init_super_imsm_volume(st, info, size, name, homehost, uuid, |
5340 | data_offset); | |
e683ca88 DW |
5341 | |
5342 | if (info) | |
5343 | mpb_size = disks_to_mpb_size(info->nr_disks); | |
5344 | else | |
f36a9ecd | 5345 | mpb_size = MAX_SECTOR_SIZE; |
bf5a934a | 5346 | |
49133e57 | 5347 | super = alloc_super(); |
f36a9ecd PB |
5348 | if (super && |
5349 | posix_memalign(&super->buf, MAX_SECTOR_SIZE, mpb_size) != 0) { | |
8d67477f | 5350 | free_imsm(super); |
e683ca88 DW |
5351 | super = NULL; |
5352 | } | |
5353 | if (!super) { | |
1ade5cc1 | 5354 | pr_err("could not allocate superblock\n"); |
bf5a934a DW |
5355 | return 0; |
5356 | } | |
de44e46f PB |
5357 | if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE, |
5358 | MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) { | |
1ade5cc1 | 5359 | pr_err("could not allocate migr_rec buffer\n"); |
8e59f3d8 | 5360 | free(super->buf); |
8d67477f | 5361 | free_imsm(super); |
8e59f3d8 AK |
5362 | return 0; |
5363 | } | |
e683ca88 | 5364 | memset(super->buf, 0, mpb_size); |
ef649044 | 5365 | mpb = super->buf; |
e683ca88 DW |
5366 | mpb->mpb_size = __cpu_to_le32(mpb_size); |
5367 | st->sb = super; | |
5368 | ||
5369 | if (info == NULL) { | |
5370 | /* zeroing superblock */ | |
5371 | return 0; | |
5372 | } | |
bf5a934a | 5373 | |
4d1313e9 DW |
5374 | mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; |
5375 | ||
5376 | version = (char *) mpb->sig; | |
5377 | strcpy(version, MPB_SIGNATURE); | |
5378 | version += strlen(MPB_SIGNATURE); | |
5379 | strcpy(version, MPB_VERSION_RAID0); | |
bf5a934a | 5380 | |
bf5a934a DW |
5381 | return 1; |
5382 | } | |
5383 | ||
0e600426 | 5384 | #ifndef MDASSEMBLE |
f20c3968 | 5385 | static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk, |
bf5a934a DW |
5386 | int fd, char *devname) |
5387 | { | |
5388 | struct intel_super *super = st->sb; | |
d23fe947 | 5389 | struct imsm_super *mpb = super->anchor; |
3960e579 | 5390 | struct imsm_disk *_disk; |
bf5a934a DW |
5391 | struct imsm_dev *dev; |
5392 | struct imsm_map *map; | |
3960e579 | 5393 | struct dl *dl, *df; |
4eb26970 | 5394 | int slot; |
bf5a934a | 5395 | |
949c47a0 | 5396 | dev = get_imsm_dev(super, super->current_vol); |
238c0a71 | 5397 | map = get_imsm_map(dev, MAP_0); |
bf5a934a | 5398 | |
208933a7 | 5399 | if (! (dk->state & (1<<MD_DISK_SYNC))) { |
e7b84f9d | 5400 | pr_err("%s: Cannot add spare devices to IMSM volume\n", |
208933a7 N |
5401 | devname); |
5402 | return 1; | |
5403 | } | |
5404 | ||
efb30e7f DW |
5405 | if (fd == -1) { |
5406 | /* we're doing autolayout so grab the pre-marked (in | |
5407 | * validate_geometry) raid_disk | |
5408 | */ | |
5409 | for (dl = super->disks; dl; dl = dl->next) | |
5410 | if (dl->raiddisk == dk->raid_disk) | |
5411 | break; | |
5412 | } else { | |
5413 | for (dl = super->disks; dl ; dl = dl->next) | |
5414 | if (dl->major == dk->major && | |
5415 | dl->minor == dk->minor) | |
5416 | break; | |
5417 | } | |
d23fe947 | 5418 | |
208933a7 | 5419 | if (!dl) { |
e7b84f9d | 5420 | pr_err("%s is not a member of the same container\n", devname); |
f20c3968 | 5421 | return 1; |
208933a7 | 5422 | } |
bf5a934a | 5423 | |
d23fe947 DW |
5424 | /* add a pristine spare to the metadata */ |
5425 | if (dl->index < 0) { | |
5426 | dl->index = super->anchor->num_disks; | |
5427 | super->anchor->num_disks++; | |
5428 | } | |
4eb26970 DW |
5429 | /* Check the device has not already been added */ |
5430 | slot = get_imsm_disk_slot(map, dl->index); | |
5431 | if (slot >= 0 && | |
238c0a71 | 5432 | (get_imsm_ord_tbl_ent(dev, slot, MAP_X) & IMSM_ORD_REBUILD) == 0) { |
e7b84f9d | 5433 | pr_err("%s has been included in this array twice\n", |
4eb26970 DW |
5434 | devname); |
5435 | return 1; | |
5436 | } | |
656b6b5a | 5437 | set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index); |
ee5aad5a | 5438 | dl->disk.status = CONFIGURED_DISK; |
d23fe947 | 5439 | |
3960e579 DW |
5440 | /* update size of 'missing' disks to be at least as large as the |
5441 | * largest acitve member (we only have dummy missing disks when | |
5442 | * creating the first volume) | |
5443 | */ | |
5444 | if (super->current_vol == 0) { | |
5445 | for (df = super->missing; df; df = df->next) { | |
5551b113 CA |
5446 | if (total_blocks(&dl->disk) > total_blocks(&df->disk)) |
5447 | set_total_blocks(&df->disk, total_blocks(&dl->disk)); | |
3960e579 DW |
5448 | _disk = __get_imsm_disk(mpb, df->index); |
5449 | *_disk = df->disk; | |
5450 | } | |
5451 | } | |
5452 | ||
5453 | /* refresh unset/failed slots to point to valid 'missing' entries */ | |
5454 | for (df = super->missing; df; df = df->next) | |
5455 | for (slot = 0; slot < mpb->num_disks; slot++) { | |
238c0a71 | 5456 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); |
3960e579 DW |
5457 | |
5458 | if ((ord & IMSM_ORD_REBUILD) == 0) | |
5459 | continue; | |
5460 | set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD); | |
1ace8403 | 5461 | if (is_gen_migration(dev)) { |
238c0a71 AK |
5462 | struct imsm_map *map2 = get_imsm_map(dev, |
5463 | MAP_1); | |
0a108d63 | 5464 | int slot2 = get_imsm_disk_slot(map2, df->index); |
089f9d79 | 5465 | if (slot2 < map2->num_members && slot2 >= 0) { |
1ace8403 | 5466 | __u32 ord2 = get_imsm_ord_tbl_ent(dev, |
238c0a71 AK |
5467 | slot2, |
5468 | MAP_1); | |
1ace8403 AK |
5469 | if ((unsigned)df->index == |
5470 | ord_to_idx(ord2)) | |
5471 | set_imsm_ord_tbl_ent(map2, | |
0a108d63 | 5472 | slot2, |
1ace8403 AK |
5473 | df->index | |
5474 | IMSM_ORD_REBUILD); | |
5475 | } | |
5476 | } | |
3960e579 DW |
5477 | dprintf("set slot:%d to missing disk:%d\n", slot, df->index); |
5478 | break; | |
5479 | } | |
5480 | ||
d23fe947 DW |
5481 | /* if we are creating the first raid device update the family number */ |
5482 | if (super->current_vol == 0) { | |
5483 | __u32 sum; | |
5484 | struct imsm_dev *_dev = __get_imsm_dev(mpb, 0); | |
d23fe947 | 5485 | |
3960e579 | 5486 | _disk = __get_imsm_disk(mpb, dl->index); |
791b666a | 5487 | if (!_dev || !_disk) { |
e7b84f9d | 5488 | pr_err("BUG mpb setup error\n"); |
791b666a AW |
5489 | return 1; |
5490 | } | |
d23fe947 DW |
5491 | *_dev = *dev; |
5492 | *_disk = dl->disk; | |
148acb7b DW |
5493 | sum = random32(); |
5494 | sum += __gen_imsm_checksum(mpb); | |
d23fe947 | 5495 | mpb->family_num = __cpu_to_le32(sum); |
148acb7b | 5496 | mpb->orig_family_num = mpb->family_num; |
d23fe947 | 5497 | } |
ca0748fa | 5498 | super->current_disk = dl; |
f20c3968 | 5499 | return 0; |
bf5a934a DW |
5500 | } |
5501 | ||
a8619d23 AK |
5502 | /* mark_spare() |
5503 | * Function marks disk as spare and restores disk serial | |
5504 | * in case it was previously marked as failed by takeover operation | |
5505 | * reruns: | |
5506 | * -1 : critical error | |
5507 | * 0 : disk is marked as spare but serial is not set | |
5508 | * 1 : success | |
5509 | */ | |
5510 | int mark_spare(struct dl *disk) | |
5511 | { | |
5512 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
5513 | int ret_val = -1; | |
5514 | ||
5515 | if (!disk) | |
5516 | return ret_val; | |
5517 | ||
5518 | ret_val = 0; | |
5519 | if (!imsm_read_serial(disk->fd, NULL, serial)) { | |
5520 | /* Restore disk serial number, because takeover marks disk | |
5521 | * as failed and adds to serial ':0' before it becomes | |
5522 | * a spare disk. | |
5523 | */ | |
5524 | serialcpy(disk->serial, serial); | |
5525 | serialcpy(disk->disk.serial, serial); | |
5526 | ret_val = 1; | |
5527 | } | |
5528 | disk->disk.status = SPARE_DISK; | |
5529 | disk->index = -1; | |
5530 | ||
5531 | return ret_val; | |
5532 | } | |
88654014 | 5533 | |
f20c3968 | 5534 | static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk, |
72ca9bcf N |
5535 | int fd, char *devname, |
5536 | unsigned long long data_offset) | |
cdddbdbc | 5537 | { |
c2c087e6 | 5538 | struct intel_super *super = st->sb; |
c2c087e6 DW |
5539 | struct dl *dd; |
5540 | unsigned long long size; | |
fa7bb6f8 | 5541 | unsigned int member_sector_size; |
f2f27e63 | 5542 | __u32 id; |
c2c087e6 DW |
5543 | int rv; |
5544 | struct stat stb; | |
5545 | ||
88654014 LM |
5546 | /* If we are on an RAID enabled platform check that the disk is |
5547 | * attached to the raid controller. | |
5548 | * We do not need to test disks attachment for container based additions, | |
5549 | * they shall be already tested when container was created/assembled. | |
88c32bb1 | 5550 | */ |
d424212e | 5551 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 5552 | /* no orom/efi or non-intel hba of the disk */ |
f0f5a016 LM |
5553 | if (rv != 0) { |
5554 | dprintf("capability: %p fd: %d ret: %d\n", | |
5555 | super->orom, fd, rv); | |
5556 | return 1; | |
88c32bb1 DW |
5557 | } |
5558 | ||
f20c3968 DW |
5559 | if (super->current_vol >= 0) |
5560 | return add_to_super_imsm_volume(st, dk, fd, devname); | |
bf5a934a | 5561 | |
c2c087e6 | 5562 | fstat(fd, &stb); |
503975b9 | 5563 | dd = xcalloc(sizeof(*dd), 1); |
c2c087e6 DW |
5564 | dd->major = major(stb.st_rdev); |
5565 | dd->minor = minor(stb.st_rdev); | |
503975b9 | 5566 | dd->devname = devname ? xstrdup(devname) : NULL; |
c2c087e6 | 5567 | dd->fd = fd; |
689c9bf3 | 5568 | dd->e = NULL; |
1a64be56 | 5569 | dd->action = DISK_ADD; |
c2c087e6 | 5570 | rv = imsm_read_serial(fd, devname, dd->serial); |
32ba9157 | 5571 | if (rv) { |
e7b84f9d | 5572 | pr_err("failed to retrieve scsi serial, aborting\n"); |
20bee0f8 PB |
5573 | if (dd->devname) |
5574 | free(dd->devname); | |
949c47a0 | 5575 | free(dd); |
0030e8d6 | 5576 | abort(); |
c2c087e6 | 5577 | } |
20bee0f8 PB |
5578 | if (super->hba && ((super->hba->type == SYS_DEV_NVME) || |
5579 | (super->hba->type == SYS_DEV_VMD))) { | |
5580 | int i; | |
5581 | char *devpath = diskfd_to_devpath(fd); | |
5582 | char controller_path[PATH_MAX]; | |
5583 | ||
5584 | if (!devpath) { | |
5585 | pr_err("failed to get devpath, aborting\n"); | |
5586 | if (dd->devname) | |
5587 | free(dd->devname); | |
5588 | free(dd); | |
5589 | return 1; | |
5590 | } | |
5591 | ||
5592 | snprintf(controller_path, PATH_MAX-1, "%s/device", devpath); | |
5593 | free(devpath); | |
5594 | ||
5595 | if (devpath_to_vendor(controller_path) == 0x8086) { | |
5596 | /* | |
5597 | * If Intel's NVMe drive has serial ended with | |
5598 | * "-A","-B","-1" or "-2" it means that this is "x8" | |
5599 | * device (double drive on single PCIe card). | |
5600 | * User should be warned about potential data loss. | |
5601 | */ | |
5602 | for (i = MAX_RAID_SERIAL_LEN-1; i > 0; i--) { | |
5603 | /* Skip empty character at the end */ | |
5604 | if (dd->serial[i] == 0) | |
5605 | continue; | |
5606 | ||
5607 | if (((dd->serial[i] == 'A') || | |
5608 | (dd->serial[i] == 'B') || | |
5609 | (dd->serial[i] == '1') || | |
5610 | (dd->serial[i] == '2')) && | |
5611 | (dd->serial[i-1] == '-')) | |
5612 | pr_err("\tThe action you are about to take may put your data at risk.\n" | |
5613 | "\tPlease note that x8 devices may consist of two separate x4 devices " | |
5614 | "located on a single PCIe port.\n" | |
5615 | "\tRAID 0 is the only supported configuration for this type of x8 device.\n"); | |
5616 | break; | |
5617 | } | |
32716c51 PB |
5618 | } else if (super->hba->type == SYS_DEV_VMD && super->orom && |
5619 | !imsm_orom_has_tpv_support(super->orom)) { | |
5620 | pr_err("\tPlatform configuration does not support non-Intel NVMe drives.\n" | |
5621 | "\tPlease refer to Intel(R) RSTe user guide.\n"); | |
5622 | free(dd->devname); | |
5623 | free(dd); | |
5624 | return 1; | |
20bee0f8 PB |
5625 | } |
5626 | } | |
c2c087e6 | 5627 | |
c2c087e6 | 5628 | get_dev_size(fd, NULL, &size); |
fa7bb6f8 PB |
5629 | get_dev_sector_size(fd, NULL, &member_sector_size); |
5630 | ||
5631 | if (super->sector_size == 0) { | |
5632 | /* this a first device, so sector_size is not set yet */ | |
5633 | super->sector_size = member_sector_size; | |
5634 | } else if (member_sector_size != super->sector_size) { | |
5635 | pr_err("Mixing between different sector size is forbidden, aborting...\n"); | |
5636 | if (dd->devname) | |
5637 | free(dd->devname); | |
5638 | free(dd); | |
5639 | return 1; | |
5640 | } | |
5641 | ||
71e5411e | 5642 | /* clear migr_rec when adding disk to container */ |
de44e46f PB |
5643 | memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*super->sector_size); |
5644 | if (lseek64(fd, size - MIGR_REC_SECTOR_POSITION*super->sector_size, | |
5645 | SEEK_SET) >= 0) { | |
466070ad | 5646 | if ((unsigned int)write(fd, super->migr_rec_buf, |
de44e46f PB |
5647 | MIGR_REC_BUF_SECTORS*super->sector_size) != |
5648 | MIGR_REC_BUF_SECTORS*super->sector_size) | |
71e5411e PB |
5649 | perror("Write migr_rec failed"); |
5650 | } | |
5651 | ||
c2c087e6 | 5652 | size /= 512; |
1f24f035 | 5653 | serialcpy(dd->disk.serial, dd->serial); |
5551b113 CA |
5654 | set_total_blocks(&dd->disk, size); |
5655 | if (__le32_to_cpu(dd->disk.total_blocks_hi) > 0) { | |
5656 | struct imsm_super *mpb = super->anchor; | |
5657 | mpb->attributes |= MPB_ATTRIB_2TB_DISK; | |
5658 | } | |
a8619d23 | 5659 | mark_spare(dd); |
c2c087e6 | 5660 | if (sysfs_disk_to_scsi_id(fd, &id) == 0) |
b9f594fe | 5661 | dd->disk.scsi_id = __cpu_to_le32(id); |
c2c087e6 | 5662 | else |
b9f594fe | 5663 | dd->disk.scsi_id = __cpu_to_le32(0); |
43dad3d6 DW |
5664 | |
5665 | if (st->update_tail) { | |
1a64be56 LM |
5666 | dd->next = super->disk_mgmt_list; |
5667 | super->disk_mgmt_list = dd; | |
43dad3d6 DW |
5668 | } else { |
5669 | dd->next = super->disks; | |
5670 | super->disks = dd; | |
ceaf0ee1 | 5671 | super->updates_pending++; |
43dad3d6 | 5672 | } |
f20c3968 DW |
5673 | |
5674 | return 0; | |
cdddbdbc DW |
5675 | } |
5676 | ||
1a64be56 LM |
5677 | static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk) |
5678 | { | |
5679 | struct intel_super *super = st->sb; | |
5680 | struct dl *dd; | |
5681 | ||
5682 | /* remove from super works only in mdmon - for communication | |
5683 | * manager - monitor. Check if communication memory buffer | |
5684 | * is prepared. | |
5685 | */ | |
5686 | if (!st->update_tail) { | |
1ade5cc1 | 5687 | pr_err("shall be used in mdmon context only\n"); |
1a64be56 LM |
5688 | return 1; |
5689 | } | |
503975b9 | 5690 | dd = xcalloc(1, sizeof(*dd)); |
1a64be56 LM |
5691 | dd->major = dk->major; |
5692 | dd->minor = dk->minor; | |
1a64be56 | 5693 | dd->fd = -1; |
a8619d23 | 5694 | mark_spare(dd); |
1a64be56 LM |
5695 | dd->action = DISK_REMOVE; |
5696 | ||
5697 | dd->next = super->disk_mgmt_list; | |
5698 | super->disk_mgmt_list = dd; | |
5699 | ||
1a64be56 LM |
5700 | return 0; |
5701 | } | |
5702 | ||
f796af5d DW |
5703 | static int store_imsm_mpb(int fd, struct imsm_super *mpb); |
5704 | ||
5705 | static union { | |
f36a9ecd | 5706 | char buf[MAX_SECTOR_SIZE]; |
f796af5d | 5707 | struct imsm_super anchor; |
f36a9ecd | 5708 | } spare_record __attribute__ ((aligned(MAX_SECTOR_SIZE))); |
c2c087e6 | 5709 | |
d23fe947 DW |
5710 | /* spare records have their own family number and do not have any defined raid |
5711 | * devices | |
5712 | */ | |
5713 | static int write_super_imsm_spares(struct intel_super *super, int doclose) | |
5714 | { | |
d23fe947 | 5715 | struct imsm_super *mpb = super->anchor; |
f796af5d | 5716 | struct imsm_super *spare = &spare_record.anchor; |
d23fe947 DW |
5717 | __u32 sum; |
5718 | struct dl *d; | |
5719 | ||
68641cdb JS |
5720 | spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super)); |
5721 | spare->generation_num = __cpu_to_le32(1UL); | |
f796af5d | 5722 | spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; |
68641cdb JS |
5723 | spare->num_disks = 1; |
5724 | spare->num_raid_devs = 0; | |
5725 | spare->cache_size = mpb->cache_size; | |
5726 | spare->pwr_cycle_count = __cpu_to_le32(1); | |
f796af5d DW |
5727 | |
5728 | snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH, | |
5729 | MPB_SIGNATURE MPB_VERSION_RAID0); | |
d23fe947 DW |
5730 | |
5731 | for (d = super->disks; d; d = d->next) { | |
8796fdc4 | 5732 | if (d->index != -1) |
d23fe947 DW |
5733 | continue; |
5734 | ||
f796af5d | 5735 | spare->disk[0] = d->disk; |
027c374f CA |
5736 | if (__le32_to_cpu(d->disk.total_blocks_hi) > 0) |
5737 | spare->attributes |= MPB_ATTRIB_2TB_DISK; | |
5738 | ||
f36a9ecd PB |
5739 | if (super->sector_size == 4096) |
5740 | convert_to_4k_imsm_disk(&spare->disk[0]); | |
5741 | ||
f796af5d DW |
5742 | sum = __gen_imsm_checksum(spare); |
5743 | spare->family_num = __cpu_to_le32(sum); | |
5744 | spare->orig_family_num = 0; | |
5745 | sum = __gen_imsm_checksum(spare); | |
5746 | spare->check_sum = __cpu_to_le32(sum); | |
d23fe947 | 5747 | |
f796af5d | 5748 | if (store_imsm_mpb(d->fd, spare)) { |
1ade5cc1 N |
5749 | pr_err("failed for device %d:%d %s\n", |
5750 | d->major, d->minor, strerror(errno)); | |
e74255d9 | 5751 | return 1; |
d23fe947 DW |
5752 | } |
5753 | if (doclose) { | |
5754 | close(d->fd); | |
5755 | d->fd = -1; | |
5756 | } | |
5757 | } | |
5758 | ||
e74255d9 | 5759 | return 0; |
d23fe947 DW |
5760 | } |
5761 | ||
36988a3d | 5762 | static int write_super_imsm(struct supertype *st, int doclose) |
cdddbdbc | 5763 | { |
36988a3d | 5764 | struct intel_super *super = st->sb; |
f36a9ecd | 5765 | unsigned int sector_size = super->sector_size; |
949c47a0 | 5766 | struct imsm_super *mpb = super->anchor; |
c2c087e6 DW |
5767 | struct dl *d; |
5768 | __u32 generation; | |
5769 | __u32 sum; | |
d23fe947 | 5770 | int spares = 0; |
949c47a0 | 5771 | int i; |
a48ac0a8 | 5772 | __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk); |
36988a3d | 5773 | int num_disks = 0; |
146c6260 | 5774 | int clear_migration_record = 1; |
bbab0940 | 5775 | __u32 bbm_log_size; |
cdddbdbc | 5776 | |
c2c087e6 DW |
5777 | /* 'generation' is incremented everytime the metadata is written */ |
5778 | generation = __le32_to_cpu(mpb->generation_num); | |
5779 | generation++; | |
5780 | mpb->generation_num = __cpu_to_le32(generation); | |
5781 | ||
148acb7b DW |
5782 | /* fix up cases where previous mdadm releases failed to set |
5783 | * orig_family_num | |
5784 | */ | |
5785 | if (mpb->orig_family_num == 0) | |
5786 | mpb->orig_family_num = mpb->family_num; | |
5787 | ||
d23fe947 | 5788 | for (d = super->disks; d; d = d->next) { |
8796fdc4 | 5789 | if (d->index == -1) |
d23fe947 | 5790 | spares++; |
36988a3d | 5791 | else { |
d23fe947 | 5792 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
5793 | num_disks++; |
5794 | } | |
d23fe947 | 5795 | } |
36988a3d | 5796 | for (d = super->missing; d; d = d->next) { |
47ee5a45 | 5797 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
5798 | num_disks++; |
5799 | } | |
5800 | mpb->num_disks = num_disks; | |
5801 | mpb_size += sizeof(struct imsm_disk) * mpb->num_disks; | |
b9f594fe | 5802 | |
949c47a0 DW |
5803 | for (i = 0; i < mpb->num_raid_devs; i++) { |
5804 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
36988a3d AK |
5805 | struct imsm_dev *dev2 = get_imsm_dev(super, i); |
5806 | if (dev && dev2) { | |
5807 | imsm_copy_dev(dev, dev2); | |
5808 | mpb_size += sizeof_imsm_dev(dev, 0); | |
5809 | } | |
146c6260 AK |
5810 | if (is_gen_migration(dev2)) |
5811 | clear_migration_record = 0; | |
949c47a0 | 5812 | } |
bbab0940 TM |
5813 | |
5814 | bbm_log_size = get_imsm_bbm_log_size(super->bbm_log); | |
5815 | ||
5816 | if (bbm_log_size) { | |
5817 | memcpy((void *)mpb + mpb_size, super->bbm_log, bbm_log_size); | |
5818 | mpb->attributes |= MPB_ATTRIB_BBM; | |
5819 | } else | |
5820 | mpb->attributes &= ~MPB_ATTRIB_BBM; | |
5821 | ||
5822 | super->anchor->bbm_log_size = __cpu_to_le32(bbm_log_size); | |
5823 | mpb_size += bbm_log_size; | |
a48ac0a8 | 5824 | mpb->mpb_size = __cpu_to_le32(mpb_size); |
949c47a0 | 5825 | |
bbab0940 TM |
5826 | #ifdef DEBUG |
5827 | assert(super->len == 0 || mpb_size <= super->len); | |
5828 | #endif | |
5829 | ||
c2c087e6 | 5830 | /* recalculate checksum */ |
949c47a0 | 5831 | sum = __gen_imsm_checksum(mpb); |
c2c087e6 DW |
5832 | mpb->check_sum = __cpu_to_le32(sum); |
5833 | ||
51d83f5d AK |
5834 | if (super->clean_migration_record_by_mdmon) { |
5835 | clear_migration_record = 1; | |
5836 | super->clean_migration_record_by_mdmon = 0; | |
5837 | } | |
146c6260 | 5838 | if (clear_migration_record) |
de44e46f PB |
5839 | memset(super->migr_rec_buf, 0, |
5840 | MIGR_REC_BUF_SECTORS*sector_size); | |
146c6260 | 5841 | |
f36a9ecd PB |
5842 | if (sector_size == 4096) |
5843 | convert_to_4k(super); | |
5844 | ||
d23fe947 | 5845 | /* write the mpb for disks that compose raid devices */ |
c2c087e6 | 5846 | for (d = super->disks; d ; d = d->next) { |
86c54047 | 5847 | if (d->index < 0 || is_failed(&d->disk)) |
d23fe947 | 5848 | continue; |
30602f53 | 5849 | |
146c6260 AK |
5850 | if (clear_migration_record) { |
5851 | unsigned long long dsize; | |
5852 | ||
5853 | get_dev_size(d->fd, NULL, &dsize); | |
de44e46f PB |
5854 | if (lseek64(d->fd, dsize - sector_size, |
5855 | SEEK_SET) >= 0) { | |
466070ad PB |
5856 | if ((unsigned int)write(d->fd, |
5857 | super->migr_rec_buf, | |
de44e46f PB |
5858 | MIGR_REC_BUF_SECTORS*sector_size) != |
5859 | MIGR_REC_BUF_SECTORS*sector_size) | |
9e2d750d | 5860 | perror("Write migr_rec failed"); |
146c6260 AK |
5861 | } |
5862 | } | |
51d83f5d AK |
5863 | |
5864 | if (store_imsm_mpb(d->fd, mpb)) | |
5865 | fprintf(stderr, | |
1ade5cc1 N |
5866 | "failed for device %d:%d (fd: %d)%s\n", |
5867 | d->major, d->minor, | |
51d83f5d AK |
5868 | d->fd, strerror(errno)); |
5869 | ||
c2c087e6 DW |
5870 | if (doclose) { |
5871 | close(d->fd); | |
5872 | d->fd = -1; | |
5873 | } | |
5874 | } | |
5875 | ||
d23fe947 DW |
5876 | if (spares) |
5877 | return write_super_imsm_spares(super, doclose); | |
5878 | ||
e74255d9 | 5879 | return 0; |
c2c087e6 DW |
5880 | } |
5881 | ||
9b1fb677 | 5882 | static int create_array(struct supertype *st, int dev_idx) |
43dad3d6 DW |
5883 | { |
5884 | size_t len; | |
5885 | struct imsm_update_create_array *u; | |
5886 | struct intel_super *super = st->sb; | |
9b1fb677 | 5887 | struct imsm_dev *dev = get_imsm_dev(super, dev_idx); |
238c0a71 | 5888 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
54c2c1ea DW |
5889 | struct disk_info *inf; |
5890 | struct imsm_disk *disk; | |
5891 | int i; | |
43dad3d6 | 5892 | |
54c2c1ea DW |
5893 | len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) + |
5894 | sizeof(*inf) * map->num_members; | |
503975b9 | 5895 | u = xmalloc(len); |
43dad3d6 | 5896 | u->type = update_create_array; |
9b1fb677 | 5897 | u->dev_idx = dev_idx; |
43dad3d6 | 5898 | imsm_copy_dev(&u->dev, dev); |
54c2c1ea DW |
5899 | inf = get_disk_info(u); |
5900 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 5901 | int idx = get_imsm_disk_idx(dev, i, MAP_X); |
9b1fb677 | 5902 | |
54c2c1ea | 5903 | disk = get_imsm_disk(super, idx); |
1ca5c8e0 N |
5904 | if (!disk) |
5905 | disk = get_imsm_missing(super, idx); | |
54c2c1ea DW |
5906 | serialcpy(inf[i].serial, disk->serial); |
5907 | } | |
43dad3d6 DW |
5908 | append_metadata_update(st, u, len); |
5909 | ||
5910 | return 0; | |
5911 | } | |
5912 | ||
1a64be56 | 5913 | static int mgmt_disk(struct supertype *st) |
43dad3d6 DW |
5914 | { |
5915 | struct intel_super *super = st->sb; | |
5916 | size_t len; | |
1a64be56 | 5917 | struct imsm_update_add_remove_disk *u; |
43dad3d6 | 5918 | |
1a64be56 | 5919 | if (!super->disk_mgmt_list) |
43dad3d6 DW |
5920 | return 0; |
5921 | ||
5922 | len = sizeof(*u); | |
503975b9 | 5923 | u = xmalloc(len); |
1a64be56 | 5924 | u->type = update_add_remove_disk; |
43dad3d6 DW |
5925 | append_metadata_update(st, u, len); |
5926 | ||
5927 | return 0; | |
5928 | } | |
5929 | ||
c2c087e6 DW |
5930 | static int write_init_super_imsm(struct supertype *st) |
5931 | { | |
9b1fb677 DW |
5932 | struct intel_super *super = st->sb; |
5933 | int current_vol = super->current_vol; | |
5934 | ||
5935 | /* we are done with current_vol reset it to point st at the container */ | |
5936 | super->current_vol = -1; | |
5937 | ||
8273f55e | 5938 | if (st->update_tail) { |
43dad3d6 DW |
5939 | /* queue the recently created array / added disk |
5940 | * as a metadata update */ | |
43dad3d6 | 5941 | int rv; |
8273f55e | 5942 | |
43dad3d6 | 5943 | /* determine if we are creating a volume or adding a disk */ |
9b1fb677 | 5944 | if (current_vol < 0) { |
1a64be56 LM |
5945 | /* in the mgmt (add/remove) disk case we are running |
5946 | * in mdmon context, so don't close fd's | |
43dad3d6 | 5947 | */ |
1a64be56 | 5948 | return mgmt_disk(st); |
43dad3d6 | 5949 | } else |
9b1fb677 | 5950 | rv = create_array(st, current_vol); |
8273f55e | 5951 | |
43dad3d6 | 5952 | return rv; |
d682f344 N |
5953 | } else { |
5954 | struct dl *d; | |
5955 | for (d = super->disks; d; d = d->next) | |
ba728be7 | 5956 | Kill(d->devname, NULL, 0, -1, 1); |
36988a3d | 5957 | return write_super_imsm(st, 1); |
d682f344 | 5958 | } |
cdddbdbc | 5959 | } |
0e600426 | 5960 | #endif |
cdddbdbc | 5961 | |
e683ca88 | 5962 | static int store_super_imsm(struct supertype *st, int fd) |
cdddbdbc | 5963 | { |
e683ca88 DW |
5964 | struct intel_super *super = st->sb; |
5965 | struct imsm_super *mpb = super ? super->anchor : NULL; | |
551c80c1 | 5966 | |
e683ca88 | 5967 | if (!mpb) |
ad97895e DW |
5968 | return 1; |
5969 | ||
1799c9e8 | 5970 | #ifndef MDASSEMBLE |
f36a9ecd PB |
5971 | if (super->sector_size == 4096) |
5972 | convert_to_4k(super); | |
e683ca88 | 5973 | return store_imsm_mpb(fd, mpb); |
1799c9e8 N |
5974 | #else |
5975 | return 1; | |
5976 | #endif | |
cdddbdbc DW |
5977 | } |
5978 | ||
0e600426 | 5979 | #ifndef MDASSEMBLE |
cdddbdbc DW |
5980 | static int validate_geometry_imsm_container(struct supertype *st, int level, |
5981 | int layout, int raiddisks, int chunk, | |
af4348dd N |
5982 | unsigned long long size, |
5983 | unsigned long long data_offset, | |
5984 | char *dev, | |
2c514b71 NB |
5985 | unsigned long long *freesize, |
5986 | int verbose) | |
cdddbdbc | 5987 | { |
c2c087e6 DW |
5988 | int fd; |
5989 | unsigned long long ldsize; | |
594dc1b8 | 5990 | struct intel_super *super; |
f2f5c343 | 5991 | int rv = 0; |
cdddbdbc | 5992 | |
c2c087e6 DW |
5993 | if (level != LEVEL_CONTAINER) |
5994 | return 0; | |
5995 | if (!dev) | |
5996 | return 1; | |
5997 | ||
5998 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
5999 | if (fd < 0) { | |
ba728be7 | 6000 | if (verbose > 0) |
e7b84f9d | 6001 | pr_err("imsm: Cannot open %s: %s\n", |
2c514b71 | 6002 | dev, strerror(errno)); |
c2c087e6 DW |
6003 | return 0; |
6004 | } | |
6005 | if (!get_dev_size(fd, dev, &ldsize)) { | |
6006 | close(fd); | |
6007 | return 0; | |
6008 | } | |
f2f5c343 LM |
6009 | |
6010 | /* capabilities retrieve could be possible | |
6011 | * note that there is no fd for the disks in array. | |
6012 | */ | |
6013 | super = alloc_super(); | |
8d67477f TM |
6014 | if (!super) { |
6015 | close(fd); | |
6016 | return 0; | |
6017 | } | |
fa7bb6f8 PB |
6018 | if (!get_dev_sector_size(fd, NULL, &super->sector_size)) { |
6019 | close(fd); | |
6020 | free_imsm(super); | |
6021 | return 0; | |
6022 | } | |
6023 | ||
ba728be7 | 6024 | rv = find_intel_hba_capability(fd, super, verbose > 0 ? dev : NULL); |
f2f5c343 LM |
6025 | if (rv != 0) { |
6026 | #if DEBUG | |
6027 | char str[256]; | |
6028 | fd2devname(fd, str); | |
1ade5cc1 | 6029 | dprintf("fd: %d %s orom: %p rv: %d raiddisk: %d\n", |
f2f5c343 LM |
6030 | fd, str, super->orom, rv, raiddisks); |
6031 | #endif | |
6032 | /* no orom/efi or non-intel hba of the disk */ | |
6033 | close(fd); | |
6034 | free_imsm(super); | |
6035 | return 0; | |
6036 | } | |
c2c087e6 | 6037 | close(fd); |
9126b9a8 CA |
6038 | if (super->orom) { |
6039 | if (raiddisks > super->orom->tds) { | |
6040 | if (verbose) | |
7a862a02 | 6041 | pr_err("%d exceeds maximum number of platform supported disks: %d\n", |
9126b9a8 CA |
6042 | raiddisks, super->orom->tds); |
6043 | free_imsm(super); | |
6044 | return 0; | |
6045 | } | |
6046 | if ((super->orom->attr & IMSM_OROM_ATTR_2TB_DISK) == 0 && | |
6047 | (ldsize >> 9) >> 32 > 0) { | |
6048 | if (verbose) | |
e7b84f9d | 6049 | pr_err("%s exceeds maximum platform supported size\n", dev); |
9126b9a8 CA |
6050 | free_imsm(super); |
6051 | return 0; | |
6052 | } | |
f2f5c343 | 6053 | } |
c2c087e6 | 6054 | |
af4348dd | 6055 | *freesize = avail_size_imsm(st, ldsize >> 9, data_offset); |
f2f5c343 | 6056 | free_imsm(super); |
c2c087e6 DW |
6057 | |
6058 | return 1; | |
cdddbdbc DW |
6059 | } |
6060 | ||
0dcecb2e DW |
6061 | static unsigned long long find_size(struct extent *e, int *idx, int num_extents) |
6062 | { | |
6063 | const unsigned long long base_start = e[*idx].start; | |
6064 | unsigned long long end = base_start + e[*idx].size; | |
6065 | int i; | |
6066 | ||
6067 | if (base_start == end) | |
6068 | return 0; | |
6069 | ||
6070 | *idx = *idx + 1; | |
6071 | for (i = *idx; i < num_extents; i++) { | |
6072 | /* extend overlapping extents */ | |
6073 | if (e[i].start >= base_start && | |
6074 | e[i].start <= end) { | |
6075 | if (e[i].size == 0) | |
6076 | return 0; | |
6077 | if (e[i].start + e[i].size > end) | |
6078 | end = e[i].start + e[i].size; | |
6079 | } else if (e[i].start > end) { | |
6080 | *idx = i; | |
6081 | break; | |
6082 | } | |
6083 | } | |
6084 | ||
6085 | return end - base_start; | |
6086 | } | |
6087 | ||
6088 | static unsigned long long merge_extents(struct intel_super *super, int sum_extents) | |
6089 | { | |
6090 | /* build a composite disk with all known extents and generate a new | |
6091 | * 'maxsize' given the "all disks in an array must share a common start | |
6092 | * offset" constraint | |
6093 | */ | |
503975b9 | 6094 | struct extent *e = xcalloc(sum_extents, sizeof(*e)); |
0dcecb2e DW |
6095 | struct dl *dl; |
6096 | int i, j; | |
6097 | int start_extent; | |
6098 | unsigned long long pos; | |
b9d77223 | 6099 | unsigned long long start = 0; |
0dcecb2e DW |
6100 | unsigned long long maxsize; |
6101 | unsigned long reserve; | |
6102 | ||
0dcecb2e DW |
6103 | /* coalesce and sort all extents. also, check to see if we need to |
6104 | * reserve space between member arrays | |
6105 | */ | |
6106 | j = 0; | |
6107 | for (dl = super->disks; dl; dl = dl->next) { | |
6108 | if (!dl->e) | |
6109 | continue; | |
6110 | for (i = 0; i < dl->extent_cnt; i++) | |
6111 | e[j++] = dl->e[i]; | |
6112 | } | |
6113 | qsort(e, sum_extents, sizeof(*e), cmp_extent); | |
6114 | ||
6115 | /* merge extents */ | |
6116 | i = 0; | |
6117 | j = 0; | |
6118 | while (i < sum_extents) { | |
6119 | e[j].start = e[i].start; | |
6120 | e[j].size = find_size(e, &i, sum_extents); | |
6121 | j++; | |
6122 | if (e[j-1].size == 0) | |
6123 | break; | |
6124 | } | |
6125 | ||
6126 | pos = 0; | |
6127 | maxsize = 0; | |
6128 | start_extent = 0; | |
6129 | i = 0; | |
6130 | do { | |
6131 | unsigned long long esize; | |
6132 | ||
6133 | esize = e[i].start - pos; | |
6134 | if (esize >= maxsize) { | |
6135 | maxsize = esize; | |
6136 | start = pos; | |
6137 | start_extent = i; | |
6138 | } | |
6139 | pos = e[i].start + e[i].size; | |
6140 | i++; | |
6141 | } while (e[i-1].size); | |
6142 | free(e); | |
6143 | ||
a7dd165b DW |
6144 | if (maxsize == 0) |
6145 | return 0; | |
6146 | ||
6147 | /* FIXME assumes volume at offset 0 is the first volume in a | |
6148 | * container | |
6149 | */ | |
0dcecb2e DW |
6150 | if (start_extent > 0) |
6151 | reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */ | |
6152 | else | |
6153 | reserve = 0; | |
6154 | ||
6155 | if (maxsize < reserve) | |
a7dd165b | 6156 | return 0; |
0dcecb2e | 6157 | |
5551b113 | 6158 | super->create_offset = ~((unsigned long long) 0); |
0dcecb2e | 6159 | if (start + reserve > super->create_offset) |
a7dd165b | 6160 | return 0; /* start overflows create_offset */ |
0dcecb2e DW |
6161 | super->create_offset = start + reserve; |
6162 | ||
6163 | return maxsize - reserve; | |
6164 | } | |
6165 | ||
88c32bb1 DW |
6166 | static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks) |
6167 | { | |
6168 | if (level < 0 || level == 6 || level == 4) | |
6169 | return 0; | |
6170 | ||
6171 | /* if we have an orom prevent invalid raid levels */ | |
6172 | if (orom) | |
6173 | switch (level) { | |
6174 | case 0: return imsm_orom_has_raid0(orom); | |
6175 | case 1: | |
6176 | if (raiddisks > 2) | |
6177 | return imsm_orom_has_raid1e(orom); | |
1c556e92 DW |
6178 | return imsm_orom_has_raid1(orom) && raiddisks == 2; |
6179 | case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4; | |
6180 | case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2; | |
88c32bb1 DW |
6181 | } |
6182 | else | |
6183 | return 1; /* not on an Intel RAID platform so anything goes */ | |
6184 | ||
6185 | return 0; | |
6186 | } | |
6187 | ||
ca9de185 LM |
6188 | static int |
6189 | active_arrays_by_format(char *name, char* hba, struct md_list **devlist, | |
6190 | int dpa, int verbose) | |
6191 | { | |
6192 | struct mdstat_ent *mdstat = mdstat_read(0, 0); | |
594dc1b8 | 6193 | struct mdstat_ent *memb; |
ca9de185 LM |
6194 | int count = 0; |
6195 | int num = 0; | |
594dc1b8 | 6196 | struct md_list *dv; |
ca9de185 LM |
6197 | int found; |
6198 | ||
6199 | for (memb = mdstat ; memb ; memb = memb->next) { | |
6200 | if (memb->metadata_version && | |
6201 | (strncmp(memb->metadata_version, "external:", 9) == 0) && | |
6202 | (strcmp(&memb->metadata_version[9], name) == 0) && | |
6203 | !is_subarray(memb->metadata_version+9) && | |
6204 | memb->members) { | |
6205 | struct dev_member *dev = memb->members; | |
6206 | int fd = -1; | |
6207 | while(dev && (fd < 0)) { | |
503975b9 N |
6208 | char *path = xmalloc(strlen(dev->name) + strlen("/dev/") + 1); |
6209 | num = sprintf(path, "%s%s", "/dev/", dev->name); | |
6210 | if (num > 0) | |
6211 | fd = open(path, O_RDONLY, 0); | |
089f9d79 | 6212 | if (num <= 0 || fd < 0) { |
676e87a8 | 6213 | pr_vrb("Cannot open %s: %s\n", |
503975b9 | 6214 | dev->name, strerror(errno)); |
ca9de185 | 6215 | } |
503975b9 | 6216 | free(path); |
ca9de185 LM |
6217 | dev = dev->next; |
6218 | } | |
6219 | found = 0; | |
089f9d79 | 6220 | if (fd >= 0 && disk_attached_to_hba(fd, hba)) { |
ca9de185 LM |
6221 | struct mdstat_ent *vol; |
6222 | for (vol = mdstat ; vol ; vol = vol->next) { | |
089f9d79 | 6223 | if (vol->active > 0 && |
ca9de185 | 6224 | vol->metadata_version && |
9581efb1 | 6225 | is_container_member(vol, memb->devnm)) { |
ca9de185 LM |
6226 | found++; |
6227 | count++; | |
6228 | } | |
6229 | } | |
6230 | if (*devlist && (found < dpa)) { | |
503975b9 | 6231 | dv = xcalloc(1, sizeof(*dv)); |
9581efb1 N |
6232 | dv->devname = xmalloc(strlen(memb->devnm) + strlen("/dev/") + 1); |
6233 | sprintf(dv->devname, "%s%s", "/dev/", memb->devnm); | |
503975b9 N |
6234 | dv->found = found; |
6235 | dv->used = 0; | |
6236 | dv->next = *devlist; | |
6237 | *devlist = dv; | |
ca9de185 LM |
6238 | } |
6239 | } | |
6240 | if (fd >= 0) | |
6241 | close(fd); | |
6242 | } | |
6243 | } | |
6244 | free_mdstat(mdstat); | |
6245 | return count; | |
6246 | } | |
6247 | ||
6248 | #ifdef DEBUG_LOOP | |
6249 | static struct md_list* | |
6250 | get_loop_devices(void) | |
6251 | { | |
6252 | int i; | |
6253 | struct md_list *devlist = NULL; | |
594dc1b8 | 6254 | struct md_list *dv; |
ca9de185 LM |
6255 | |
6256 | for(i = 0; i < 12; i++) { | |
503975b9 N |
6257 | dv = xcalloc(1, sizeof(*dv)); |
6258 | dv->devname = xmalloc(40); | |
ca9de185 LM |
6259 | sprintf(dv->devname, "/dev/loop%d", i); |
6260 | dv->next = devlist; | |
6261 | devlist = dv; | |
6262 | } | |
6263 | return devlist; | |
6264 | } | |
6265 | #endif | |
6266 | ||
6267 | static struct md_list* | |
6268 | get_devices(const char *hba_path) | |
6269 | { | |
6270 | struct md_list *devlist = NULL; | |
594dc1b8 | 6271 | struct md_list *dv; |
ca9de185 LM |
6272 | struct dirent *ent; |
6273 | DIR *dir; | |
6274 | int err = 0; | |
6275 | ||
6276 | #if DEBUG_LOOP | |
6277 | devlist = get_loop_devices(); | |
6278 | return devlist; | |
6279 | #endif | |
6280 | /* scroll through /sys/dev/block looking for devices attached to | |
6281 | * this hba | |
6282 | */ | |
6283 | dir = opendir("/sys/dev/block"); | |
6284 | for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) { | |
6285 | int fd; | |
6286 | char buf[1024]; | |
6287 | int major, minor; | |
6288 | char *path = NULL; | |
6289 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
6290 | continue; | |
6291 | path = devt_to_devpath(makedev(major, minor)); | |
6292 | if (!path) | |
6293 | continue; | |
6294 | if (!path_attached_to_hba(path, hba_path)) { | |
6295 | free(path); | |
6296 | path = NULL; | |
6297 | continue; | |
6298 | } | |
6299 | free(path); | |
6300 | path = NULL; | |
6301 | fd = dev_open(ent->d_name, O_RDONLY); | |
6302 | if (fd >= 0) { | |
6303 | fd2devname(fd, buf); | |
6304 | close(fd); | |
6305 | } else { | |
e7b84f9d | 6306 | pr_err("cannot open device: %s\n", |
ca9de185 LM |
6307 | ent->d_name); |
6308 | continue; | |
6309 | } | |
6310 | ||
503975b9 N |
6311 | dv = xcalloc(1, sizeof(*dv)); |
6312 | dv->devname = xstrdup(buf); | |
ca9de185 LM |
6313 | dv->next = devlist; |
6314 | devlist = dv; | |
6315 | } | |
6316 | if (err) { | |
6317 | while(devlist) { | |
6318 | dv = devlist; | |
6319 | devlist = devlist->next; | |
6320 | free(dv->devname); | |
6321 | free(dv); | |
6322 | } | |
6323 | } | |
562aa102 | 6324 | closedir(dir); |
ca9de185 LM |
6325 | return devlist; |
6326 | } | |
6327 | ||
6328 | static int | |
6329 | count_volumes_list(struct md_list *devlist, char *homehost, | |
6330 | int verbose, int *found) | |
6331 | { | |
6332 | struct md_list *tmpdev; | |
6333 | int count = 0; | |
594dc1b8 | 6334 | struct supertype *st; |
ca9de185 LM |
6335 | |
6336 | /* first walk the list of devices to find a consistent set | |
6337 | * that match the criterea, if that is possible. | |
6338 | * We flag the ones we like with 'used'. | |
6339 | */ | |
6340 | *found = 0; | |
6341 | st = match_metadata_desc_imsm("imsm"); | |
6342 | if (st == NULL) { | |
676e87a8 | 6343 | pr_vrb("cannot allocate memory for imsm supertype\n"); |
ca9de185 LM |
6344 | return 0; |
6345 | } | |
6346 | ||
6347 | for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { | |
6348 | char *devname = tmpdev->devname; | |
6349 | struct stat stb; | |
6350 | struct supertype *tst; | |
6351 | int dfd; | |
6352 | if (tmpdev->used > 1) | |
6353 | continue; | |
6354 | tst = dup_super(st); | |
6355 | if (tst == NULL) { | |
676e87a8 | 6356 | pr_vrb("cannot allocate memory for imsm supertype\n"); |
ca9de185 LM |
6357 | goto err_1; |
6358 | } | |
6359 | tmpdev->container = 0; | |
6360 | dfd = dev_open(devname, O_RDONLY|O_EXCL); | |
6361 | if (dfd < 0) { | |
1ade5cc1 | 6362 | dprintf("cannot open device %s: %s\n", |
ca9de185 LM |
6363 | devname, strerror(errno)); |
6364 | tmpdev->used = 2; | |
6365 | } else if (fstat(dfd, &stb)< 0) { | |
6366 | /* Impossible! */ | |
1ade5cc1 | 6367 | dprintf("fstat failed for %s: %s\n", |
ca9de185 LM |
6368 | devname, strerror(errno)); |
6369 | tmpdev->used = 2; | |
6370 | } else if ((stb.st_mode & S_IFMT) != S_IFBLK) { | |
1ade5cc1 | 6371 | dprintf("%s is not a block device.\n", |
ca9de185 LM |
6372 | devname); |
6373 | tmpdev->used = 2; | |
6374 | } else if (must_be_container(dfd)) { | |
6375 | struct supertype *cst; | |
6376 | cst = super_by_fd(dfd, NULL); | |
6377 | if (cst == NULL) { | |
1ade5cc1 | 6378 | dprintf("cannot recognize container type %s\n", |
ca9de185 LM |
6379 | devname); |
6380 | tmpdev->used = 2; | |
6381 | } else if (tst->ss != st->ss) { | |
1ade5cc1 | 6382 | dprintf("non-imsm container - ignore it: %s\n", |
ca9de185 LM |
6383 | devname); |
6384 | tmpdev->used = 2; | |
6385 | } else if (!tst->ss->load_container || | |
6386 | tst->ss->load_container(tst, dfd, NULL)) | |
6387 | tmpdev->used = 2; | |
6388 | else { | |
6389 | tmpdev->container = 1; | |
6390 | } | |
6391 | if (cst) | |
6392 | cst->ss->free_super(cst); | |
6393 | } else { | |
6394 | tmpdev->st_rdev = stb.st_rdev; | |
6395 | if (tst->ss->load_super(tst,dfd, NULL)) { | |
1ade5cc1 | 6396 | dprintf("no RAID superblock on %s\n", |
ca9de185 LM |
6397 | devname); |
6398 | tmpdev->used = 2; | |
6399 | } else if (tst->ss->compare_super == NULL) { | |
1ade5cc1 | 6400 | dprintf("Cannot assemble %s metadata on %s\n", |
ca9de185 LM |
6401 | tst->ss->name, devname); |
6402 | tmpdev->used = 2; | |
6403 | } | |
6404 | } | |
6405 | if (dfd >= 0) | |
6406 | close(dfd); | |
6407 | if (tmpdev->used == 2 || tmpdev->used == 4) { | |
6408 | /* Ignore unrecognised devices during auto-assembly */ | |
6409 | goto loop; | |
6410 | } | |
6411 | else { | |
6412 | struct mdinfo info; | |
6413 | tst->ss->getinfo_super(tst, &info, NULL); | |
6414 | ||
6415 | if (st->minor_version == -1) | |
6416 | st->minor_version = tst->minor_version; | |
6417 | ||
6418 | if (memcmp(info.uuid, uuid_zero, | |
6419 | sizeof(int[4])) == 0) { | |
6420 | /* this is a floating spare. It cannot define | |
6421 | * an array unless there are no more arrays of | |
6422 | * this type to be found. It can be included | |
6423 | * in an array of this type though. | |
6424 | */ | |
6425 | tmpdev->used = 3; | |
6426 | goto loop; | |
6427 | } | |
6428 | ||
6429 | if (st->ss != tst->ss || | |
6430 | st->minor_version != tst->minor_version || | |
6431 | st->ss->compare_super(st, tst) != 0) { | |
6432 | /* Some mismatch. If exactly one array matches this host, | |
6433 | * we can resolve on that one. | |
6434 | * Or, if we are auto assembling, we just ignore the second | |
6435 | * for now. | |
6436 | */ | |
1ade5cc1 | 6437 | dprintf("superblock on %s doesn't match others - assembly aborted\n", |
ca9de185 LM |
6438 | devname); |
6439 | goto loop; | |
6440 | } | |
6441 | tmpdev->used = 1; | |
6442 | *found = 1; | |
6443 | dprintf("found: devname: %s\n", devname); | |
6444 | } | |
6445 | loop: | |
6446 | if (tst) | |
6447 | tst->ss->free_super(tst); | |
6448 | } | |
6449 | if (*found != 0) { | |
6450 | int err; | |
6451 | if ((err = load_super_imsm_all(st, -1, &st->sb, NULL, devlist, 0)) == 0) { | |
6452 | struct mdinfo *iter, *head = st->ss->container_content(st, NULL); | |
6453 | for (iter = head; iter; iter = iter->next) { | |
6454 | dprintf("content->text_version: %s vol\n", | |
6455 | iter->text_version); | |
6456 | if (iter->array.state & (1<<MD_SB_BLOCK_VOLUME)) { | |
6457 | /* do not assemble arrays with unsupported | |
6458 | configurations */ | |
1ade5cc1 | 6459 | dprintf("Cannot activate member %s.\n", |
ca9de185 LM |
6460 | iter->text_version); |
6461 | } else | |
6462 | count++; | |
6463 | } | |
6464 | sysfs_free(head); | |
6465 | ||
6466 | } else { | |
1ade5cc1 | 6467 | dprintf("No valid super block on device list: err: %d %p\n", |
ca9de185 LM |
6468 | err, st->sb); |
6469 | } | |
6470 | } else { | |
1ade5cc1 | 6471 | dprintf("no more devices to examine\n"); |
ca9de185 LM |
6472 | } |
6473 | ||
6474 | for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { | |
089f9d79 | 6475 | if (tmpdev->used == 1 && tmpdev->found) { |
ca9de185 LM |
6476 | if (count) { |
6477 | if (count < tmpdev->found) | |
6478 | count = 0; | |
6479 | else | |
6480 | count -= tmpdev->found; | |
6481 | } | |
6482 | } | |
6483 | if (tmpdev->used == 1) | |
6484 | tmpdev->used = 4; | |
6485 | } | |
6486 | err_1: | |
6487 | if (st) | |
6488 | st->ss->free_super(st); | |
6489 | return count; | |
6490 | } | |
6491 | ||
d3c11416 AO |
6492 | static int __count_volumes(char *hba_path, int dpa, int verbose, |
6493 | int cmp_hba_path) | |
ca9de185 | 6494 | { |
72a45777 | 6495 | struct sys_dev *idev, *intel_devices = find_intel_devices(); |
ca9de185 | 6496 | int count = 0; |
72a45777 PB |
6497 | const struct orom_entry *entry; |
6498 | struct devid_list *dv, *devid_list; | |
ca9de185 | 6499 | |
d3c11416 | 6500 | if (!hba_path) |
ca9de185 LM |
6501 | return 0; |
6502 | ||
72a45777 | 6503 | for (idev = intel_devices; idev; idev = idev->next) { |
d3c11416 AO |
6504 | if (strstr(idev->path, hba_path)) |
6505 | break; | |
72a45777 PB |
6506 | } |
6507 | ||
6508 | if (!idev || !idev->dev_id) | |
ca9de185 | 6509 | return 0; |
72a45777 PB |
6510 | |
6511 | entry = get_orom_entry_by_device_id(idev->dev_id); | |
6512 | ||
6513 | if (!entry || !entry->devid_list) | |
6514 | return 0; | |
6515 | ||
6516 | devid_list = entry->devid_list; | |
6517 | for (dv = devid_list; dv; dv = dv->next) { | |
594dc1b8 | 6518 | struct md_list *devlist; |
d3c11416 AO |
6519 | struct sys_dev *device = NULL; |
6520 | char *hpath; | |
72a45777 PB |
6521 | int found = 0; |
6522 | ||
d3c11416 AO |
6523 | if (cmp_hba_path) |
6524 | device = device_by_id_and_path(dv->devid, hba_path); | |
6525 | else | |
6526 | device = device_by_id(dv->devid); | |
6527 | ||
72a45777 | 6528 | if (device) |
d3c11416 | 6529 | hpath = device->path; |
72a45777 PB |
6530 | else |
6531 | return 0; | |
6532 | ||
d3c11416 | 6533 | devlist = get_devices(hpath); |
72a45777 PB |
6534 | /* if no intel devices return zero volumes */ |
6535 | if (devlist == NULL) | |
6536 | return 0; | |
6537 | ||
d3c11416 AO |
6538 | count += active_arrays_by_format("imsm", hpath, &devlist, dpa, |
6539 | verbose); | |
6540 | dprintf("path: %s active arrays: %d\n", hpath, count); | |
72a45777 PB |
6541 | if (devlist == NULL) |
6542 | return 0; | |
6543 | do { | |
6544 | found = 0; | |
6545 | count += count_volumes_list(devlist, | |
6546 | NULL, | |
6547 | verbose, | |
6548 | &found); | |
6549 | dprintf("found %d count: %d\n", found, count); | |
6550 | } while (found); | |
6551 | ||
d3c11416 | 6552 | dprintf("path: %s total number of volumes: %d\n", hpath, count); |
72a45777 PB |
6553 | |
6554 | while (devlist) { | |
6555 | struct md_list *dv = devlist; | |
6556 | devlist = devlist->next; | |
6557 | free(dv->devname); | |
6558 | free(dv); | |
6559 | } | |
ca9de185 LM |
6560 | } |
6561 | return count; | |
6562 | } | |
6563 | ||
d3c11416 AO |
6564 | static int count_volumes(struct intel_hba *hba, int dpa, int verbose) |
6565 | { | |
6566 | if (!hba) | |
6567 | return 0; | |
6568 | if (hba->type == SYS_DEV_VMD) { | |
6569 | struct sys_dev *dev; | |
6570 | int count = 0; | |
6571 | ||
6572 | for (dev = find_intel_devices(); dev; dev = dev->next) { | |
6573 | if (dev->type == SYS_DEV_VMD) | |
6574 | count += __count_volumes(dev->path, dpa, | |
6575 | verbose, 1); | |
6576 | } | |
6577 | return count; | |
6578 | } | |
6579 | return __count_volumes(hba->path, dpa, verbose, 0); | |
6580 | } | |
6581 | ||
cd9d1ac7 DW |
6582 | static int imsm_default_chunk(const struct imsm_orom *orom) |
6583 | { | |
6584 | /* up to 512 if the plaform supports it, otherwise the platform max. | |
6585 | * 128 if no platform detected | |
6586 | */ | |
6587 | int fs = max(7, orom ? fls(orom->sss) : 0); | |
6588 | ||
6589 | return min(512, (1 << fs)); | |
6590 | } | |
73408129 | 6591 | |
6592ce37 DW |
6592 | static int |
6593 | validate_geometry_imsm_orom(struct intel_super *super, int level, int layout, | |
2cc699af | 6594 | int raiddisks, int *chunk, unsigned long long size, int verbose) |
6592ce37 | 6595 | { |
660260d0 DW |
6596 | /* check/set platform and metadata limits/defaults */ |
6597 | if (super->orom && raiddisks > super->orom->dpa) { | |
676e87a8 | 6598 | pr_vrb("platform supports a maximum of %d disks per array\n", |
660260d0 | 6599 | super->orom->dpa); |
73408129 LM |
6600 | return 0; |
6601 | } | |
6602 | ||
5d500228 | 6603 | /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */ |
660260d0 | 6604 | if (!is_raid_level_supported(super->orom, level, raiddisks)) { |
676e87a8 | 6605 | pr_vrb("platform does not support raid%d with %d disk%s\n", |
6592ce37 DW |
6606 | level, raiddisks, raiddisks > 1 ? "s" : ""); |
6607 | return 0; | |
6608 | } | |
cd9d1ac7 | 6609 | |
7ccc4cc4 | 6610 | if (*chunk == 0 || *chunk == UnSet) |
cd9d1ac7 DW |
6611 | *chunk = imsm_default_chunk(super->orom); |
6612 | ||
7ccc4cc4 | 6613 | if (super->orom && !imsm_orom_has_chunk(super->orom, *chunk)) { |
676e87a8 | 6614 | pr_vrb("platform does not support a chunk size of: %d\n", *chunk); |
cd9d1ac7 | 6615 | return 0; |
6592ce37 | 6616 | } |
cd9d1ac7 | 6617 | |
6592ce37 DW |
6618 | if (layout != imsm_level_to_layout(level)) { |
6619 | if (level == 5) | |
676e87a8 | 6620 | pr_vrb("imsm raid 5 only supports the left-asymmetric layout\n"); |
6592ce37 | 6621 | else if (level == 10) |
676e87a8 | 6622 | pr_vrb("imsm raid 10 only supports the n2 layout\n"); |
6592ce37 | 6623 | else |
676e87a8 | 6624 | pr_vrb("imsm unknown layout %#x for this raid level %d\n", |
6592ce37 DW |
6625 | layout, level); |
6626 | return 0; | |
6627 | } | |
2cc699af | 6628 | |
7ccc4cc4 | 6629 | if (super->orom && (super->orom->attr & IMSM_OROM_ATTR_2TB) == 0 && |
2cc699af | 6630 | (calc_array_size(level, raiddisks, layout, *chunk, size) >> 32) > 0) { |
676e87a8 | 6631 | pr_vrb("platform does not support a volume size over 2TB\n"); |
2cc699af CA |
6632 | return 0; |
6633 | } | |
614902f6 | 6634 | |
6592ce37 DW |
6635 | return 1; |
6636 | } | |
6637 | ||
1011e834 | 6638 | /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd |
c2c087e6 DW |
6639 | * FIX ME add ahci details |
6640 | */ | |
8b353278 | 6641 | static int validate_geometry_imsm_volume(struct supertype *st, int level, |
c21e737b | 6642 | int layout, int raiddisks, int *chunk, |
af4348dd N |
6643 | unsigned long long size, |
6644 | unsigned long long data_offset, | |
6645 | char *dev, | |
2c514b71 NB |
6646 | unsigned long long *freesize, |
6647 | int verbose) | |
cdddbdbc | 6648 | { |
c2c087e6 DW |
6649 | struct stat stb; |
6650 | struct intel_super *super = st->sb; | |
b2916f25 | 6651 | struct imsm_super *mpb; |
c2c087e6 DW |
6652 | struct dl *dl; |
6653 | unsigned long long pos = 0; | |
6654 | unsigned long long maxsize; | |
6655 | struct extent *e; | |
6656 | int i; | |
cdddbdbc | 6657 | |
88c32bb1 DW |
6658 | /* We must have the container info already read in. */ |
6659 | if (!super) | |
c2c087e6 DW |
6660 | return 0; |
6661 | ||
b2916f25 JS |
6662 | mpb = super->anchor; |
6663 | ||
2cc699af | 6664 | if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, size, verbose)) { |
7a862a02 | 6665 | pr_err("RAID gemetry validation failed. Cannot proceed with the action(s).\n"); |
c2c087e6 | 6666 | return 0; |
d54559f0 | 6667 | } |
c2c087e6 DW |
6668 | if (!dev) { |
6669 | /* General test: make sure there is space for | |
2da8544a DW |
6670 | * 'raiddisks' device extents of size 'size' at a given |
6671 | * offset | |
c2c087e6 | 6672 | */ |
e46273eb | 6673 | unsigned long long minsize = size; |
b7528a20 | 6674 | unsigned long long start_offset = MaxSector; |
c2c087e6 DW |
6675 | int dcnt = 0; |
6676 | if (minsize == 0) | |
6677 | minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
6678 | for (dl = super->disks; dl ; dl = dl->next) { | |
6679 | int found = 0; | |
6680 | ||
bf5a934a | 6681 | pos = 0; |
c2c087e6 DW |
6682 | i = 0; |
6683 | e = get_extents(super, dl); | |
6684 | if (!e) continue; | |
6685 | do { | |
6686 | unsigned long long esize; | |
6687 | esize = e[i].start - pos; | |
6688 | if (esize >= minsize) | |
6689 | found = 1; | |
b7528a20 | 6690 | if (found && start_offset == MaxSector) { |
2da8544a DW |
6691 | start_offset = pos; |
6692 | break; | |
6693 | } else if (found && pos != start_offset) { | |
6694 | found = 0; | |
6695 | break; | |
6696 | } | |
c2c087e6 DW |
6697 | pos = e[i].start + e[i].size; |
6698 | i++; | |
6699 | } while (e[i-1].size); | |
6700 | if (found) | |
6701 | dcnt++; | |
6702 | free(e); | |
6703 | } | |
6704 | if (dcnt < raiddisks) { | |
2c514b71 | 6705 | if (verbose) |
7a862a02 | 6706 | pr_err("imsm: Not enough devices with space for this array (%d < %d)\n", |
2c514b71 | 6707 | dcnt, raiddisks); |
c2c087e6 DW |
6708 | return 0; |
6709 | } | |
6710 | return 1; | |
6711 | } | |
0dcecb2e | 6712 | |
c2c087e6 DW |
6713 | /* This device must be a member of the set */ |
6714 | if (stat(dev, &stb) < 0) | |
6715 | return 0; | |
6716 | if ((S_IFMT & stb.st_mode) != S_IFBLK) | |
6717 | return 0; | |
6718 | for (dl = super->disks ; dl ; dl = dl->next) { | |
f21e18ca N |
6719 | if (dl->major == (int)major(stb.st_rdev) && |
6720 | dl->minor == (int)minor(stb.st_rdev)) | |
c2c087e6 DW |
6721 | break; |
6722 | } | |
6723 | if (!dl) { | |
2c514b71 | 6724 | if (verbose) |
7a862a02 | 6725 | pr_err("%s is not in the same imsm set\n", dev); |
c2c087e6 | 6726 | return 0; |
a20d2ba5 DW |
6727 | } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) { |
6728 | /* If a volume is present then the current creation attempt | |
6729 | * cannot incorporate new spares because the orom may not | |
6730 | * understand this configuration (all member disks must be | |
6731 | * members of each array in the container). | |
6732 | */ | |
7a862a02 N |
6733 | pr_err("%s is a spare and a volume is already defined for this container\n", dev); |
6734 | pr_err("The option-rom requires all member disks to be a member of all volumes\n"); | |
a20d2ba5 | 6735 | return 0; |
5fe62b94 WD |
6736 | } else if (super->orom && mpb->num_raid_devs > 0 && |
6737 | mpb->num_disks != raiddisks) { | |
7a862a02 | 6738 | pr_err("The option-rom requires all member disks to be a member of all volumes\n"); |
5fe62b94 | 6739 | return 0; |
c2c087e6 | 6740 | } |
0dcecb2e DW |
6741 | |
6742 | /* retrieve the largest free space block */ | |
c2c087e6 DW |
6743 | e = get_extents(super, dl); |
6744 | maxsize = 0; | |
6745 | i = 0; | |
0dcecb2e DW |
6746 | if (e) { |
6747 | do { | |
6748 | unsigned long long esize; | |
6749 | ||
6750 | esize = e[i].start - pos; | |
6751 | if (esize >= maxsize) | |
6752 | maxsize = esize; | |
6753 | pos = e[i].start + e[i].size; | |
6754 | i++; | |
6755 | } while (e[i-1].size); | |
6756 | dl->e = e; | |
6757 | dl->extent_cnt = i; | |
6758 | } else { | |
6759 | if (verbose) | |
e7b84f9d | 6760 | pr_err("unable to determine free space for: %s\n", |
0dcecb2e DW |
6761 | dev); |
6762 | return 0; | |
6763 | } | |
6764 | if (maxsize < size) { | |
6765 | if (verbose) | |
e7b84f9d | 6766 | pr_err("%s not enough space (%llu < %llu)\n", |
0dcecb2e DW |
6767 | dev, maxsize, size); |
6768 | return 0; | |
6769 | } | |
6770 | ||
6771 | /* count total number of extents for merge */ | |
6772 | i = 0; | |
6773 | for (dl = super->disks; dl; dl = dl->next) | |
6774 | if (dl->e) | |
6775 | i += dl->extent_cnt; | |
6776 | ||
6777 | maxsize = merge_extents(super, i); | |
3baa56ab LO |
6778 | |
6779 | if (!check_env("IMSM_NO_PLATFORM") && | |
6780 | mpb->num_raid_devs > 0 && size && size != maxsize) { | |
7a862a02 | 6781 | pr_err("attempting to create a second volume with size less then remaining space. Aborting...\n"); |
3baa56ab LO |
6782 | return 0; |
6783 | } | |
6784 | ||
a7dd165b | 6785 | if (maxsize < size || maxsize == 0) { |
b3071342 LD |
6786 | if (verbose) { |
6787 | if (maxsize == 0) | |
7a862a02 | 6788 | pr_err("no free space left on device. Aborting...\n"); |
b3071342 | 6789 | else |
7a862a02 | 6790 | pr_err("not enough space to create volume of given size (%llu < %llu). Aborting...\n", |
b3071342 LD |
6791 | maxsize, size); |
6792 | } | |
0dcecb2e | 6793 | return 0; |
0dcecb2e DW |
6794 | } |
6795 | ||
c2c087e6 DW |
6796 | *freesize = maxsize; |
6797 | ||
ca9de185 | 6798 | if (super->orom) { |
72a45777 | 6799 | int count = count_volumes(super->hba, |
ca9de185 LM |
6800 | super->orom->dpa, verbose); |
6801 | if (super->orom->vphba <= count) { | |
676e87a8 | 6802 | pr_vrb("platform does not support more than %d raid volumes.\n", |
ca9de185 LM |
6803 | super->orom->vphba); |
6804 | return 0; | |
6805 | } | |
6806 | } | |
c2c087e6 | 6807 | return 1; |
cdddbdbc DW |
6808 | } |
6809 | ||
13bcac90 | 6810 | static int imsm_get_free_size(struct supertype *st, int raiddisks, |
efb30e7f DW |
6811 | unsigned long long size, int chunk, |
6812 | unsigned long long *freesize) | |
6813 | { | |
6814 | struct intel_super *super = st->sb; | |
6815 | struct imsm_super *mpb = super->anchor; | |
6816 | struct dl *dl; | |
6817 | int i; | |
6818 | int extent_cnt; | |
6819 | struct extent *e; | |
6820 | unsigned long long maxsize; | |
6821 | unsigned long long minsize; | |
6822 | int cnt; | |
6823 | int used; | |
6824 | ||
6825 | /* find the largest common start free region of the possible disks */ | |
6826 | used = 0; | |
6827 | extent_cnt = 0; | |
6828 | cnt = 0; | |
6829 | for (dl = super->disks; dl; dl = dl->next) { | |
6830 | dl->raiddisk = -1; | |
6831 | ||
6832 | if (dl->index >= 0) | |
6833 | used++; | |
6834 | ||
6835 | /* don't activate new spares if we are orom constrained | |
6836 | * and there is already a volume active in the container | |
6837 | */ | |
6838 | if (super->orom && dl->index < 0 && mpb->num_raid_devs) | |
6839 | continue; | |
6840 | ||
6841 | e = get_extents(super, dl); | |
6842 | if (!e) | |
6843 | continue; | |
6844 | for (i = 1; e[i-1].size; i++) | |
6845 | ; | |
6846 | dl->e = e; | |
6847 | dl->extent_cnt = i; | |
6848 | extent_cnt += i; | |
6849 | cnt++; | |
6850 | } | |
6851 | ||
6852 | maxsize = merge_extents(super, extent_cnt); | |
6853 | minsize = size; | |
6854 | if (size == 0) | |
612e59d8 CA |
6855 | /* chunk is in K */ |
6856 | minsize = chunk * 2; | |
efb30e7f DW |
6857 | |
6858 | if (cnt < raiddisks || | |
6859 | (super->orom && used && used != raiddisks) || | |
a7dd165b DW |
6860 | maxsize < minsize || |
6861 | maxsize == 0) { | |
e7b84f9d | 6862 | pr_err("not enough devices with space to create array.\n"); |
efb30e7f DW |
6863 | return 0; /* No enough free spaces large enough */ |
6864 | } | |
6865 | ||
6866 | if (size == 0) { | |
6867 | size = maxsize; | |
6868 | if (chunk) { | |
612e59d8 CA |
6869 | size /= 2 * chunk; |
6870 | size *= 2 * chunk; | |
efb30e7f | 6871 | } |
f878b242 LM |
6872 | maxsize = size; |
6873 | } | |
6874 | if (!check_env("IMSM_NO_PLATFORM") && | |
6875 | mpb->num_raid_devs > 0 && size && size != maxsize) { | |
7a862a02 | 6876 | pr_err("attempting to create a second volume with size less then remaining space. Aborting...\n"); |
f878b242 | 6877 | return 0; |
efb30e7f | 6878 | } |
efb30e7f DW |
6879 | cnt = 0; |
6880 | for (dl = super->disks; dl; dl = dl->next) | |
6881 | if (dl->e) | |
6882 | dl->raiddisk = cnt++; | |
6883 | ||
6884 | *freesize = size; | |
6885 | ||
13bcac90 AK |
6886 | dprintf("imsm: imsm_get_free_size() returns : %llu\n", size); |
6887 | ||
efb30e7f DW |
6888 | return 1; |
6889 | } | |
6890 | ||
13bcac90 AK |
6891 | static int reserve_space(struct supertype *st, int raiddisks, |
6892 | unsigned long long size, int chunk, | |
6893 | unsigned long long *freesize) | |
6894 | { | |
6895 | struct intel_super *super = st->sb; | |
6896 | struct dl *dl; | |
6897 | int cnt; | |
6898 | int rv = 0; | |
6899 | ||
6900 | rv = imsm_get_free_size(st, raiddisks, size, chunk, freesize); | |
6901 | if (rv) { | |
6902 | cnt = 0; | |
6903 | for (dl = super->disks; dl; dl = dl->next) | |
6904 | if (dl->e) | |
6905 | dl->raiddisk = cnt++; | |
6906 | rv = 1; | |
6907 | } | |
6908 | ||
6909 | return rv; | |
6910 | } | |
6911 | ||
bf5a934a | 6912 | static int validate_geometry_imsm(struct supertype *st, int level, int layout, |
c21e737b | 6913 | int raiddisks, int *chunk, unsigned long long size, |
af4348dd | 6914 | unsigned long long data_offset, |
bf5a934a DW |
6915 | char *dev, unsigned long long *freesize, |
6916 | int verbose) | |
6917 | { | |
6918 | int fd, cfd; | |
6919 | struct mdinfo *sra; | |
20cbe8d2 | 6920 | int is_member = 0; |
bf5a934a | 6921 | |
d54559f0 LM |
6922 | /* load capability |
6923 | * if given unused devices create a container | |
bf5a934a DW |
6924 | * if given given devices in a container create a member volume |
6925 | */ | |
6926 | if (level == LEVEL_CONTAINER) { | |
6927 | /* Must be a fresh device to add to a container */ | |
6928 | return validate_geometry_imsm_container(st, level, layout, | |
c21e737b | 6929 | raiddisks, |
7ccc4cc4 | 6930 | *chunk, |
af4348dd | 6931 | size, data_offset, |
bf5a934a DW |
6932 | dev, freesize, |
6933 | verbose); | |
6934 | } | |
9587c373 | 6935 | |
8592f29d | 6936 | if (!dev) { |
e91a3bad | 6937 | if (st->sb) { |
ca9de185 | 6938 | struct intel_super *super = st->sb; |
e91a3bad | 6939 | if (!validate_geometry_imsm_orom(st->sb, level, layout, |
2cc699af | 6940 | raiddisks, chunk, size, |
e91a3bad LM |
6941 | verbose)) |
6942 | return 0; | |
efb30e7f DW |
6943 | /* we are being asked to automatically layout a |
6944 | * new volume based on the current contents of | |
6945 | * the container. If the the parameters can be | |
6946 | * satisfied reserve_space will record the disks, | |
6947 | * start offset, and size of the volume to be | |
6948 | * created. add_to_super and getinfo_super | |
6949 | * detect when autolayout is in progress. | |
6950 | */ | |
ca9de185 LM |
6951 | /* assuming that freesize is always given when array is |
6952 | created */ | |
6953 | if (super->orom && freesize) { | |
6954 | int count; | |
72a45777 | 6955 | count = count_volumes(super->hba, |
ca9de185 LM |
6956 | super->orom->dpa, verbose); |
6957 | if (super->orom->vphba <= count) { | |
676e87a8 | 6958 | pr_vrb("platform does not support more than %d raid volumes.\n", |
ca9de185 LM |
6959 | super->orom->vphba); |
6960 | return 0; | |
6961 | } | |
6962 | } | |
e91a3bad LM |
6963 | if (freesize) |
6964 | return reserve_space(st, raiddisks, size, | |
7ccc4cc4 | 6965 | *chunk, freesize); |
8592f29d N |
6966 | } |
6967 | return 1; | |
6968 | } | |
bf5a934a DW |
6969 | if (st->sb) { |
6970 | /* creating in a given container */ | |
6971 | return validate_geometry_imsm_volume(st, level, layout, | |
6972 | raiddisks, chunk, size, | |
af4348dd | 6973 | data_offset, |
bf5a934a DW |
6974 | dev, freesize, verbose); |
6975 | } | |
6976 | ||
bf5a934a DW |
6977 | /* This device needs to be a device in an 'imsm' container */ |
6978 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
6979 | if (fd >= 0) { | |
6980 | if (verbose) | |
e7b84f9d N |
6981 | pr_err("Cannot create this array on device %s\n", |
6982 | dev); | |
bf5a934a DW |
6983 | close(fd); |
6984 | return 0; | |
6985 | } | |
6986 | if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) { | |
6987 | if (verbose) | |
e7b84f9d | 6988 | pr_err("Cannot open %s: %s\n", |
bf5a934a DW |
6989 | dev, strerror(errno)); |
6990 | return 0; | |
6991 | } | |
6992 | /* Well, it is in use by someone, maybe an 'imsm' container. */ | |
6993 | cfd = open_container(fd); | |
20cbe8d2 | 6994 | close(fd); |
bf5a934a | 6995 | if (cfd < 0) { |
bf5a934a | 6996 | if (verbose) |
e7b84f9d | 6997 | pr_err("Cannot use %s: It is busy\n", |
bf5a934a DW |
6998 | dev); |
6999 | return 0; | |
7000 | } | |
4dd2df09 | 7001 | sra = sysfs_read(cfd, NULL, GET_VERSION); |
bf5a934a | 7002 | if (sra && sra->array.major_version == -1 && |
20cbe8d2 AW |
7003 | strcmp(sra->text_version, "imsm") == 0) |
7004 | is_member = 1; | |
7005 | sysfs_free(sra); | |
7006 | if (is_member) { | |
bf5a934a DW |
7007 | /* This is a member of a imsm container. Load the container |
7008 | * and try to create a volume | |
7009 | */ | |
7010 | struct intel_super *super; | |
7011 | ||
ec50f7b6 | 7012 | if (load_super_imsm_all(st, cfd, (void **) &super, NULL, NULL, 1) == 0) { |
bf5a934a | 7013 | st->sb = super; |
4dd2df09 | 7014 | strcpy(st->container_devnm, fd2devnm(cfd)); |
bf5a934a DW |
7015 | close(cfd); |
7016 | return validate_geometry_imsm_volume(st, level, layout, | |
7017 | raiddisks, chunk, | |
af4348dd | 7018 | size, data_offset, dev, |
ecbd9e81 N |
7019 | freesize, 1) |
7020 | ? 1 : -1; | |
bf5a934a | 7021 | } |
20cbe8d2 | 7022 | } |
bf5a934a | 7023 | |
20cbe8d2 | 7024 | if (verbose) |
e7b84f9d | 7025 | pr_err("failed container membership check\n"); |
20cbe8d2 AW |
7026 | |
7027 | close(cfd); | |
7028 | return 0; | |
bf5a934a | 7029 | } |
0bd16cf2 | 7030 | |
30f58b22 | 7031 | static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk) |
0bd16cf2 DJ |
7032 | { |
7033 | struct intel_super *super = st->sb; | |
7034 | ||
30f58b22 DW |
7035 | if (level && *level == UnSet) |
7036 | *level = LEVEL_CONTAINER; | |
7037 | ||
7038 | if (level && layout && *layout == UnSet) | |
7039 | *layout = imsm_level_to_layout(*level); | |
0bd16cf2 | 7040 | |
cd9d1ac7 DW |
7041 | if (chunk && (*chunk == UnSet || *chunk == 0)) |
7042 | *chunk = imsm_default_chunk(super->orom); | |
0bd16cf2 DJ |
7043 | } |
7044 | ||
33414a01 DW |
7045 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev); |
7046 | ||
7047 | static int kill_subarray_imsm(struct supertype *st) | |
7048 | { | |
7049 | /* remove the subarray currently referenced by ->current_vol */ | |
7050 | __u8 i; | |
7051 | struct intel_dev **dp; | |
7052 | struct intel_super *super = st->sb; | |
7053 | __u8 current_vol = super->current_vol; | |
7054 | struct imsm_super *mpb = super->anchor; | |
7055 | ||
7056 | if (super->current_vol < 0) | |
7057 | return 2; | |
7058 | super->current_vol = -1; /* invalidate subarray cursor */ | |
7059 | ||
7060 | /* block deletions that would change the uuid of active subarrays | |
7061 | * | |
7062 | * FIXME when immutable ids are available, but note that we'll | |
7063 | * also need to fixup the invalidated/active subarray indexes in | |
7064 | * mdstat | |
7065 | */ | |
7066 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
7067 | char subarray[4]; | |
7068 | ||
7069 | if (i < current_vol) | |
7070 | continue; | |
7071 | sprintf(subarray, "%u", i); | |
4dd2df09 | 7072 | if (is_subarray_active(subarray, st->devnm)) { |
e7b84f9d N |
7073 | pr_err("deleting subarray-%d would change the UUID of active subarray-%d, aborting\n", |
7074 | current_vol, i); | |
33414a01 DW |
7075 | |
7076 | return 2; | |
7077 | } | |
7078 | } | |
7079 | ||
7080 | if (st->update_tail) { | |
503975b9 | 7081 | struct imsm_update_kill_array *u = xmalloc(sizeof(*u)); |
33414a01 | 7082 | |
33414a01 DW |
7083 | u->type = update_kill_array; |
7084 | u->dev_idx = current_vol; | |
7085 | append_metadata_update(st, u, sizeof(*u)); | |
7086 | ||
7087 | return 0; | |
7088 | } | |
7089 | ||
7090 | for (dp = &super->devlist; *dp;) | |
7091 | if ((*dp)->index == current_vol) { | |
7092 | *dp = (*dp)->next; | |
7093 | } else { | |
7094 | handle_missing(super, (*dp)->dev); | |
7095 | if ((*dp)->index > current_vol) | |
7096 | (*dp)->index--; | |
7097 | dp = &(*dp)->next; | |
7098 | } | |
7099 | ||
7100 | /* no more raid devices, all active components are now spares, | |
7101 | * but of course failed are still failed | |
7102 | */ | |
7103 | if (--mpb->num_raid_devs == 0) { | |
7104 | struct dl *d; | |
7105 | ||
7106 | for (d = super->disks; d; d = d->next) | |
a8619d23 AK |
7107 | if (d->index > -2) |
7108 | mark_spare(d); | |
33414a01 DW |
7109 | } |
7110 | ||
7111 | super->updates_pending++; | |
7112 | ||
7113 | return 0; | |
7114 | } | |
aa534678 | 7115 | |
a951a4f7 | 7116 | static int update_subarray_imsm(struct supertype *st, char *subarray, |
fa56eddb | 7117 | char *update, struct mddev_ident *ident) |
aa534678 DW |
7118 | { |
7119 | /* update the subarray currently referenced by ->current_vol */ | |
7120 | struct intel_super *super = st->sb; | |
7121 | struct imsm_super *mpb = super->anchor; | |
7122 | ||
aa534678 DW |
7123 | if (strcmp(update, "name") == 0) { |
7124 | char *name = ident->name; | |
a951a4f7 N |
7125 | char *ep; |
7126 | int vol; | |
aa534678 | 7127 | |
4dd2df09 | 7128 | if (is_subarray_active(subarray, st->devnm)) { |
e7b84f9d | 7129 | pr_err("Unable to update name of active subarray\n"); |
aa534678 DW |
7130 | return 2; |
7131 | } | |
7132 | ||
7133 | if (!check_name(super, name, 0)) | |
7134 | return 2; | |
7135 | ||
a951a4f7 N |
7136 | vol = strtoul(subarray, &ep, 10); |
7137 | if (*ep != '\0' || vol >= super->anchor->num_raid_devs) | |
7138 | return 2; | |
7139 | ||
aa534678 | 7140 | if (st->update_tail) { |
503975b9 | 7141 | struct imsm_update_rename_array *u = xmalloc(sizeof(*u)); |
aa534678 | 7142 | |
aa534678 | 7143 | u->type = update_rename_array; |
a951a4f7 | 7144 | u->dev_idx = vol; |
aa534678 DW |
7145 | snprintf((char *) u->name, MAX_RAID_SERIAL_LEN, "%s", name); |
7146 | append_metadata_update(st, u, sizeof(*u)); | |
7147 | } else { | |
7148 | struct imsm_dev *dev; | |
7149 | int i; | |
7150 | ||
a951a4f7 | 7151 | dev = get_imsm_dev(super, vol); |
aa534678 DW |
7152 | snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name); |
7153 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
7154 | dev = get_imsm_dev(super, i); | |
7155 | handle_missing(super, dev); | |
7156 | } | |
7157 | super->updates_pending++; | |
7158 | } | |
7159 | } else | |
7160 | return 2; | |
7161 | ||
7162 | return 0; | |
7163 | } | |
d1e02575 | 7164 | #endif /* MDASSEMBLE */ |
bf5a934a | 7165 | |
28bce06f AK |
7166 | static int is_gen_migration(struct imsm_dev *dev) |
7167 | { | |
7534230b AK |
7168 | if (dev == NULL) |
7169 | return 0; | |
7170 | ||
28bce06f AK |
7171 | if (!dev->vol.migr_state) |
7172 | return 0; | |
7173 | ||
7174 | if (migr_type(dev) == MIGR_GEN_MIGR) | |
7175 | return 1; | |
7176 | ||
7177 | return 0; | |
7178 | } | |
7179 | ||
1e5c6983 DW |
7180 | static int is_rebuilding(struct imsm_dev *dev) |
7181 | { | |
7182 | struct imsm_map *migr_map; | |
7183 | ||
7184 | if (!dev->vol.migr_state) | |
7185 | return 0; | |
7186 | ||
7187 | if (migr_type(dev) != MIGR_REBUILD) | |
7188 | return 0; | |
7189 | ||
238c0a71 | 7190 | migr_map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
7191 | |
7192 | if (migr_map->map_state == IMSM_T_STATE_DEGRADED) | |
7193 | return 1; | |
7194 | else | |
7195 | return 0; | |
7196 | } | |
7197 | ||
b4ab44d8 | 7198 | #ifndef MDASSEMBLE |
6ce1fbf1 AK |
7199 | static int is_initializing(struct imsm_dev *dev) |
7200 | { | |
7201 | struct imsm_map *migr_map; | |
7202 | ||
7203 | if (!dev->vol.migr_state) | |
7204 | return 0; | |
7205 | ||
7206 | if (migr_type(dev) != MIGR_INIT) | |
7207 | return 0; | |
7208 | ||
238c0a71 | 7209 | migr_map = get_imsm_map(dev, MAP_1); |
6ce1fbf1 AK |
7210 | |
7211 | if (migr_map->map_state == IMSM_T_STATE_UNINITIALIZED) | |
7212 | return 1; | |
7213 | ||
7214 | return 0; | |
6ce1fbf1 | 7215 | } |
b4ab44d8 | 7216 | #endif |
6ce1fbf1 | 7217 | |
c47b0ff6 AK |
7218 | static void update_recovery_start(struct intel_super *super, |
7219 | struct imsm_dev *dev, | |
7220 | struct mdinfo *array) | |
1e5c6983 DW |
7221 | { |
7222 | struct mdinfo *rebuild = NULL; | |
7223 | struct mdinfo *d; | |
7224 | __u32 units; | |
7225 | ||
7226 | if (!is_rebuilding(dev)) | |
7227 | return; | |
7228 | ||
7229 | /* Find the rebuild target, but punt on the dual rebuild case */ | |
7230 | for (d = array->devs; d; d = d->next) | |
7231 | if (d->recovery_start == 0) { | |
7232 | if (rebuild) | |
7233 | return; | |
7234 | rebuild = d; | |
7235 | } | |
7236 | ||
4363fd80 DW |
7237 | if (!rebuild) { |
7238 | /* (?) none of the disks are marked with | |
7239 | * IMSM_ORD_REBUILD, so assume they are missing and the | |
7240 | * disk_ord_tbl was not correctly updated | |
7241 | */ | |
1ade5cc1 | 7242 | dprintf("failed to locate out-of-sync disk\n"); |
4363fd80 DW |
7243 | return; |
7244 | } | |
7245 | ||
1e5c6983 | 7246 | units = __le32_to_cpu(dev->vol.curr_migr_unit); |
c47b0ff6 | 7247 | rebuild->recovery_start = units * blocks_per_migr_unit(super, dev); |
1e5c6983 DW |
7248 | } |
7249 | ||
9e2d750d | 7250 | #ifndef MDASSEMBLE |
276d77db | 7251 | static int recover_backup_imsm(struct supertype *st, struct mdinfo *info); |
9e2d750d | 7252 | #endif |
1e5c6983 | 7253 | |
00bbdbda | 7254 | static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray) |
cdddbdbc | 7255 | { |
4f5bc454 DW |
7256 | /* Given a container loaded by load_super_imsm_all, |
7257 | * extract information about all the arrays into | |
7258 | * an mdinfo tree. | |
00bbdbda | 7259 | * If 'subarray' is given, just extract info about that array. |
4f5bc454 DW |
7260 | * |
7261 | * For each imsm_dev create an mdinfo, fill it in, | |
7262 | * then look for matching devices in super->disks | |
7263 | * and create appropriate device mdinfo. | |
7264 | */ | |
7265 | struct intel_super *super = st->sb; | |
949c47a0 | 7266 | struct imsm_super *mpb = super->anchor; |
4f5bc454 | 7267 | struct mdinfo *rest = NULL; |
00bbdbda | 7268 | unsigned int i; |
81219e70 | 7269 | int sb_errors = 0; |
abef11a3 AK |
7270 | struct dl *d; |
7271 | int spare_disks = 0; | |
cdddbdbc | 7272 | |
19482bcc AK |
7273 | /* do not assemble arrays when not all attributes are supported */ |
7274 | if (imsm_check_attributes(mpb->attributes) == 0) { | |
81219e70 | 7275 | sb_errors = 1; |
7a862a02 | 7276 | pr_err("Unsupported attributes in IMSM metadata.Arrays activation is blocked.\n"); |
19482bcc AK |
7277 | } |
7278 | ||
abef11a3 AK |
7279 | /* count spare devices, not used in maps |
7280 | */ | |
7281 | for (d = super->disks; d; d = d->next) | |
7282 | if (d->index == -1) | |
7283 | spare_disks++; | |
7284 | ||
4f5bc454 | 7285 | for (i = 0; i < mpb->num_raid_devs; i++) { |
00bbdbda N |
7286 | struct imsm_dev *dev; |
7287 | struct imsm_map *map; | |
86e3692b | 7288 | struct imsm_map *map2; |
4f5bc454 | 7289 | struct mdinfo *this; |
a6482415 N |
7290 | int slot; |
7291 | #ifndef MDASSEMBLE | |
7292 | int chunk; | |
7293 | #endif | |
00bbdbda N |
7294 | char *ep; |
7295 | ||
7296 | if (subarray && | |
7297 | (i != strtoul(subarray, &ep, 10) || *ep != '\0')) | |
7298 | continue; | |
7299 | ||
7300 | dev = get_imsm_dev(super, i); | |
238c0a71 AK |
7301 | map = get_imsm_map(dev, MAP_0); |
7302 | map2 = get_imsm_map(dev, MAP_1); | |
4f5bc454 | 7303 | |
1ce0101c DW |
7304 | /* do not publish arrays that are in the middle of an |
7305 | * unsupported migration | |
7306 | */ | |
7307 | if (dev->vol.migr_state && | |
28bce06f | 7308 | (migr_type(dev) == MIGR_STATE_CHANGE)) { |
7a862a02 | 7309 | pr_err("cannot assemble volume '%.16s': unsupported migration in progress\n", |
1ce0101c DW |
7310 | dev->volume); |
7311 | continue; | |
7312 | } | |
2db86302 LM |
7313 | /* do not publish arrays that are not support by controller's |
7314 | * OROM/EFI | |
7315 | */ | |
1ce0101c | 7316 | |
503975b9 | 7317 | this = xmalloc(sizeof(*this)); |
4f5bc454 | 7318 | |
301406c9 | 7319 | super->current_vol = i; |
a5d85af7 | 7320 | getinfo_super_imsm_volume(st, this, NULL); |
9894ec0d | 7321 | this->next = rest; |
81219e70 | 7322 | #ifndef MDASSEMBLE |
a6482415 | 7323 | chunk = __le16_to_cpu(map->blocks_per_strip) >> 1; |
81219e70 LM |
7324 | /* mdadm does not support all metadata features- set the bit in all arrays state */ |
7325 | if (!validate_geometry_imsm_orom(super, | |
7326 | get_imsm_raid_level(map), /* RAID level */ | |
7327 | imsm_level_to_layout(get_imsm_raid_level(map)), | |
7328 | map->num_members, /* raid disks */ | |
2cc699af | 7329 | &chunk, join_u32(dev->size_low, dev->size_high), |
81219e70 | 7330 | 1 /* verbose */)) { |
7a862a02 | 7331 | pr_err("IMSM RAID geometry validation failed. Array %s activation is blocked.\n", |
81219e70 LM |
7332 | dev->volume); |
7333 | this->array.state |= | |
7334 | (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) | | |
7335 | (1<<MD_SB_BLOCK_VOLUME); | |
7336 | } | |
7337 | #endif | |
7338 | ||
7339 | /* if array has bad blocks, set suitable bit in all arrays state */ | |
7340 | if (sb_errors) | |
7341 | this->array.state |= | |
7342 | (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) | | |
7343 | (1<<MD_SB_BLOCK_VOLUME); | |
7344 | ||
4f5bc454 | 7345 | for (slot = 0 ; slot < map->num_members; slot++) { |
1e5c6983 | 7346 | unsigned long long recovery_start; |
4f5bc454 DW |
7347 | struct mdinfo *info_d; |
7348 | struct dl *d; | |
7349 | int idx; | |
9a1608e5 | 7350 | int skip; |
7eef0453 | 7351 | __u32 ord; |
4f5bc454 | 7352 | |
9a1608e5 | 7353 | skip = 0; |
238c0a71 AK |
7354 | idx = get_imsm_disk_idx(dev, slot, MAP_0); |
7355 | ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); | |
4f5bc454 DW |
7356 | for (d = super->disks; d ; d = d->next) |
7357 | if (d->index == idx) | |
0fbd635c | 7358 | break; |
4f5bc454 | 7359 | |
1e5c6983 | 7360 | recovery_start = MaxSector; |
4f5bc454 | 7361 | if (d == NULL) |
9a1608e5 | 7362 | skip = 1; |
25ed7e59 | 7363 | if (d && is_failed(&d->disk)) |
9a1608e5 | 7364 | skip = 1; |
7eef0453 | 7365 | if (ord & IMSM_ORD_REBUILD) |
1e5c6983 | 7366 | recovery_start = 0; |
9a1608e5 | 7367 | |
1011e834 | 7368 | /* |
9a1608e5 | 7369 | * if we skip some disks the array will be assmebled degraded; |
1e5c6983 DW |
7370 | * reset resync start to avoid a dirty-degraded |
7371 | * situation when performing the intial sync | |
9a1608e5 DW |
7372 | * |
7373 | * FIXME handle dirty degraded | |
7374 | */ | |
1e5c6983 | 7375 | if ((skip || recovery_start == 0) && !dev->vol.dirty) |
b7528a20 | 7376 | this->resync_start = MaxSector; |
9a1608e5 DW |
7377 | if (skip) |
7378 | continue; | |
4f5bc454 | 7379 | |
503975b9 | 7380 | info_d = xcalloc(1, sizeof(*info_d)); |
4f5bc454 DW |
7381 | info_d->next = this->devs; |
7382 | this->devs = info_d; | |
7383 | ||
4f5bc454 DW |
7384 | info_d->disk.number = d->index; |
7385 | info_d->disk.major = d->major; | |
7386 | info_d->disk.minor = d->minor; | |
7387 | info_d->disk.raid_disk = slot; | |
1e5c6983 | 7388 | info_d->recovery_start = recovery_start; |
86e3692b AK |
7389 | if (map2) { |
7390 | if (slot < map2->num_members) | |
7391 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
7392 | else |
7393 | this->array.spare_disks++; | |
86e3692b AK |
7394 | } else { |
7395 | if (slot < map->num_members) | |
7396 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
7397 | else |
7398 | this->array.spare_disks++; | |
86e3692b | 7399 | } |
1e5c6983 DW |
7400 | if (info_d->recovery_start == MaxSector) |
7401 | this->array.working_disks++; | |
4f5bc454 DW |
7402 | |
7403 | info_d->events = __le32_to_cpu(mpb->generation_num); | |
5551b113 | 7404 | info_d->data_offset = pba_of_lba0(map); |
06fb291a PB |
7405 | |
7406 | if (map->raid_level == 5) { | |
7407 | info_d->component_size = | |
7408 | num_data_stripes(map) * | |
7409 | map->blocks_per_strip; | |
7410 | } else { | |
7411 | info_d->component_size = blocks_per_member(map); | |
7412 | } | |
b12796be | 7413 | |
5e46202e | 7414 | info_d->bb.supported = 1; |
b12796be TM |
7415 | get_volume_badblocks(super->bbm_log, ord_to_idx(ord), |
7416 | info_d->data_offset, | |
7417 | info_d->component_size, | |
7418 | &info_d->bb); | |
4f5bc454 | 7419 | } |
1e5c6983 | 7420 | /* now that the disk list is up-to-date fixup recovery_start */ |
c47b0ff6 | 7421 | update_recovery_start(super, dev, this); |
abef11a3 | 7422 | this->array.spare_disks += spare_disks; |
276d77db | 7423 | |
9e2d750d | 7424 | #ifndef MDASSEMBLE |
276d77db AK |
7425 | /* check for reshape */ |
7426 | if (this->reshape_active == 1) | |
7427 | recover_backup_imsm(st, this); | |
9e2d750d | 7428 | #endif |
9a1608e5 | 7429 | rest = this; |
4f5bc454 DW |
7430 | } |
7431 | ||
7432 | return rest; | |
cdddbdbc DW |
7433 | } |
7434 | ||
3b451610 AK |
7435 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, |
7436 | int failed, int look_in_map) | |
c2a1e7da | 7437 | { |
3b451610 AK |
7438 | struct imsm_map *map; |
7439 | ||
7440 | map = get_imsm_map(dev, look_in_map); | |
c2a1e7da DW |
7441 | |
7442 | if (!failed) | |
1011e834 | 7443 | return map->map_state == IMSM_T_STATE_UNINITIALIZED ? |
3393c6af | 7444 | IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL; |
c2a1e7da DW |
7445 | |
7446 | switch (get_imsm_raid_level(map)) { | |
7447 | case 0: | |
7448 | return IMSM_T_STATE_FAILED; | |
7449 | break; | |
7450 | case 1: | |
7451 | if (failed < map->num_members) | |
7452 | return IMSM_T_STATE_DEGRADED; | |
7453 | else | |
7454 | return IMSM_T_STATE_FAILED; | |
7455 | break; | |
7456 | case 10: | |
7457 | { | |
7458 | /** | |
c92a2527 DW |
7459 | * check to see if any mirrors have failed, otherwise we |
7460 | * are degraded. Even numbered slots are mirrored on | |
7461 | * slot+1 | |
c2a1e7da | 7462 | */ |
c2a1e7da | 7463 | int i; |
d9b420a5 N |
7464 | /* gcc -Os complains that this is unused */ |
7465 | int insync = insync; | |
c2a1e7da DW |
7466 | |
7467 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 7468 | __u32 ord = get_imsm_ord_tbl_ent(dev, i, MAP_X); |
c92a2527 DW |
7469 | int idx = ord_to_idx(ord); |
7470 | struct imsm_disk *disk; | |
c2a1e7da | 7471 | |
c92a2527 | 7472 | /* reset the potential in-sync count on even-numbered |
1011e834 | 7473 | * slots. num_copies is always 2 for imsm raid10 |
c92a2527 DW |
7474 | */ |
7475 | if ((i & 1) == 0) | |
7476 | insync = 2; | |
c2a1e7da | 7477 | |
c92a2527 | 7478 | disk = get_imsm_disk(super, idx); |
25ed7e59 | 7479 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) |
c92a2527 | 7480 | insync--; |
c2a1e7da | 7481 | |
c92a2527 DW |
7482 | /* no in-sync disks left in this mirror the |
7483 | * array has failed | |
7484 | */ | |
7485 | if (insync == 0) | |
7486 | return IMSM_T_STATE_FAILED; | |
c2a1e7da DW |
7487 | } |
7488 | ||
7489 | return IMSM_T_STATE_DEGRADED; | |
7490 | } | |
7491 | case 5: | |
7492 | if (failed < 2) | |
7493 | return IMSM_T_STATE_DEGRADED; | |
7494 | else | |
7495 | return IMSM_T_STATE_FAILED; | |
7496 | break; | |
7497 | default: | |
7498 | break; | |
7499 | } | |
7500 | ||
7501 | return map->map_state; | |
7502 | } | |
7503 | ||
3b451610 AK |
7504 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev, |
7505 | int look_in_map) | |
c2a1e7da DW |
7506 | { |
7507 | int i; | |
7508 | int failed = 0; | |
7509 | struct imsm_disk *disk; | |
d5985138 AK |
7510 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
7511 | struct imsm_map *prev = get_imsm_map(dev, MAP_1); | |
68fe4598 | 7512 | struct imsm_map *map_for_loop; |
0556e1a2 DW |
7513 | __u32 ord; |
7514 | int idx; | |
d5985138 | 7515 | int idx_1; |
c2a1e7da | 7516 | |
0556e1a2 DW |
7517 | /* at the beginning of migration we set IMSM_ORD_REBUILD on |
7518 | * disks that are being rebuilt. New failures are recorded to | |
7519 | * map[0]. So we look through all the disks we started with and | |
7520 | * see if any failures are still present, or if any new ones | |
7521 | * have arrived | |
0556e1a2 | 7522 | */ |
d5985138 AK |
7523 | map_for_loop = map; |
7524 | if (prev && (map->num_members < prev->num_members)) | |
7525 | map_for_loop = prev; | |
68fe4598 LD |
7526 | |
7527 | for (i = 0; i < map_for_loop->num_members; i++) { | |
d5985138 | 7528 | idx_1 = -255; |
238c0a71 AK |
7529 | /* when MAP_X is passed both maps failures are counted |
7530 | */ | |
d5985138 | 7531 | if (prev && |
089f9d79 JS |
7532 | (look_in_map == MAP_1 || look_in_map == MAP_X) && |
7533 | i < prev->num_members) { | |
d5985138 AK |
7534 | ord = __le32_to_cpu(prev->disk_ord_tbl[i]); |
7535 | idx_1 = ord_to_idx(ord); | |
c2a1e7da | 7536 | |
d5985138 AK |
7537 | disk = get_imsm_disk(super, idx_1); |
7538 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) | |
7539 | failed++; | |
7540 | } | |
089f9d79 JS |
7541 | if ((look_in_map == MAP_0 || look_in_map == MAP_X) && |
7542 | i < map->num_members) { | |
d5985138 AK |
7543 | ord = __le32_to_cpu(map->disk_ord_tbl[i]); |
7544 | idx = ord_to_idx(ord); | |
7545 | ||
7546 | if (idx != idx_1) { | |
7547 | disk = get_imsm_disk(super, idx); | |
7548 | if (!disk || is_failed(disk) || | |
7549 | ord & IMSM_ORD_REBUILD) | |
7550 | failed++; | |
7551 | } | |
7552 | } | |
c2a1e7da DW |
7553 | } |
7554 | ||
7555 | return failed; | |
845dea95 NB |
7556 | } |
7557 | ||
97b4d0e9 DW |
7558 | #ifndef MDASSEMBLE |
7559 | static int imsm_open_new(struct supertype *c, struct active_array *a, | |
7560 | char *inst) | |
7561 | { | |
7562 | struct intel_super *super = c->sb; | |
7563 | struct imsm_super *mpb = super->anchor; | |
bbab0940 | 7564 | struct imsm_update_prealloc_bb_mem u; |
9587c373 | 7565 | |
97b4d0e9 | 7566 | if (atoi(inst) >= mpb->num_raid_devs) { |
1ade5cc1 | 7567 | pr_err("subarry index %d, out of range\n", atoi(inst)); |
97b4d0e9 DW |
7568 | return -ENODEV; |
7569 | } | |
7570 | ||
7571 | dprintf("imsm: open_new %s\n", inst); | |
7572 | a->info.container_member = atoi(inst); | |
bbab0940 TM |
7573 | |
7574 | u.type = update_prealloc_badblocks_mem; | |
7575 | imsm_update_metadata_locally(c, &u, sizeof(u)); | |
7576 | ||
97b4d0e9 DW |
7577 | return 0; |
7578 | } | |
7579 | ||
0c046afd DW |
7580 | static int is_resyncing(struct imsm_dev *dev) |
7581 | { | |
7582 | struct imsm_map *migr_map; | |
7583 | ||
7584 | if (!dev->vol.migr_state) | |
7585 | return 0; | |
7586 | ||
1484e727 DW |
7587 | if (migr_type(dev) == MIGR_INIT || |
7588 | migr_type(dev) == MIGR_REPAIR) | |
0c046afd DW |
7589 | return 1; |
7590 | ||
4c9bc37b AK |
7591 | if (migr_type(dev) == MIGR_GEN_MIGR) |
7592 | return 0; | |
7593 | ||
238c0a71 | 7594 | migr_map = get_imsm_map(dev, MAP_1); |
0c046afd | 7595 | |
089f9d79 JS |
7596 | if (migr_map->map_state == IMSM_T_STATE_NORMAL && |
7597 | dev->vol.migr_type != MIGR_GEN_MIGR) | |
0c046afd DW |
7598 | return 1; |
7599 | else | |
7600 | return 0; | |
7601 | } | |
7602 | ||
0556e1a2 | 7603 | /* return true if we recorded new information */ |
4c9e8c1e TM |
7604 | static int mark_failure(struct intel_super *super, |
7605 | struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
47ee5a45 | 7606 | { |
0556e1a2 DW |
7607 | __u32 ord; |
7608 | int slot; | |
7609 | struct imsm_map *map; | |
86c54047 DW |
7610 | char buf[MAX_RAID_SERIAL_LEN+3]; |
7611 | unsigned int len, shift = 0; | |
0556e1a2 DW |
7612 | |
7613 | /* new failures are always set in map[0] */ | |
238c0a71 | 7614 | map = get_imsm_map(dev, MAP_0); |
0556e1a2 DW |
7615 | |
7616 | slot = get_imsm_disk_slot(map, idx); | |
7617 | if (slot < 0) | |
7618 | return 0; | |
7619 | ||
7620 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
25ed7e59 | 7621 | if (is_failed(disk) && (ord & IMSM_ORD_REBUILD)) |
0556e1a2 DW |
7622 | return 0; |
7623 | ||
7d0c5e24 LD |
7624 | memcpy(buf, disk->serial, MAX_RAID_SERIAL_LEN); |
7625 | buf[MAX_RAID_SERIAL_LEN] = '\000'; | |
7626 | strcat(buf, ":0"); | |
86c54047 DW |
7627 | if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN) |
7628 | shift = len - MAX_RAID_SERIAL_LEN + 1; | |
7629 | strncpy((char *)disk->serial, &buf[shift], MAX_RAID_SERIAL_LEN); | |
7630 | ||
f2f27e63 | 7631 | disk->status |= FAILED_DISK; |
0556e1a2 | 7632 | set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD); |
17788994 AK |
7633 | /* mark failures in second map if second map exists and this disk |
7634 | * in this slot. | |
7635 | * This is valid for migration, initialization and rebuild | |
7636 | */ | |
7637 | if (dev->vol.migr_state) { | |
238c0a71 | 7638 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); |
0a108d63 AK |
7639 | int slot2 = get_imsm_disk_slot(map2, idx); |
7640 | ||
089f9d79 | 7641 | if (slot2 < map2->num_members && slot2 >= 0) |
0a108d63 | 7642 | set_imsm_ord_tbl_ent(map2, slot2, |
1ace8403 AK |
7643 | idx | IMSM_ORD_REBUILD); |
7644 | } | |
f21e18ca | 7645 | if (map->failed_disk_num == 0xff) |
0556e1a2 | 7646 | map->failed_disk_num = slot; |
4c9e8c1e TM |
7647 | |
7648 | clear_disk_badblocks(super->bbm_log, ord_to_idx(ord)); | |
7649 | ||
0556e1a2 DW |
7650 | return 1; |
7651 | } | |
7652 | ||
4c9e8c1e TM |
7653 | static void mark_missing(struct intel_super *super, |
7654 | struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
0556e1a2 | 7655 | { |
4c9e8c1e | 7656 | mark_failure(super, dev, disk, idx); |
0556e1a2 DW |
7657 | |
7658 | if (disk->scsi_id == __cpu_to_le32(~(__u32)0)) | |
7659 | return; | |
7660 | ||
47ee5a45 DW |
7661 | disk->scsi_id = __cpu_to_le32(~(__u32)0); |
7662 | memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1); | |
7663 | } | |
7664 | ||
33414a01 DW |
7665 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev) |
7666 | { | |
33414a01 | 7667 | struct dl *dl; |
33414a01 DW |
7668 | |
7669 | if (!super->missing) | |
7670 | return; | |
33414a01 | 7671 | |
79b68f1b PC |
7672 | /* When orom adds replacement for missing disk it does |
7673 | * not remove entry of missing disk, but just updates map with | |
7674 | * new added disk. So it is not enough just to test if there is | |
7675 | * any missing disk, we have to look if there are any failed disks | |
7676 | * in map to stop migration */ | |
7677 | ||
33414a01 | 7678 | dprintf("imsm: mark missing\n"); |
3d59f0c0 AK |
7679 | /* end process for initialization and rebuild only |
7680 | */ | |
7681 | if (is_gen_migration(dev) == 0) { | |
7682 | __u8 map_state; | |
7683 | int failed; | |
7684 | ||
7685 | failed = imsm_count_failed(super, dev, MAP_0); | |
7686 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
7687 | ||
79b68f1b PC |
7688 | if (failed) |
7689 | end_migration(dev, super, map_state); | |
3d59f0c0 | 7690 | } |
33414a01 | 7691 | for (dl = super->missing; dl; dl = dl->next) |
4c9e8c1e | 7692 | mark_missing(super, dev, &dl->disk, dl->index); |
33414a01 DW |
7693 | super->updates_pending++; |
7694 | } | |
7695 | ||
f3871fdc AK |
7696 | static unsigned long long imsm_set_array_size(struct imsm_dev *dev, |
7697 | long long new_size) | |
70bdf0dc | 7698 | { |
238c0a71 | 7699 | int used_disks = imsm_num_data_members(dev, MAP_0); |
70bdf0dc AK |
7700 | unsigned long long array_blocks; |
7701 | struct imsm_map *map; | |
7702 | ||
7703 | if (used_disks == 0) { | |
7704 | /* when problems occures | |
7705 | * return current array_blocks value | |
7706 | */ | |
7707 | array_blocks = __le32_to_cpu(dev->size_high); | |
7708 | array_blocks = array_blocks << 32; | |
7709 | array_blocks += __le32_to_cpu(dev->size_low); | |
7710 | ||
7711 | return array_blocks; | |
7712 | } | |
7713 | ||
7714 | /* set array size in metadata | |
7715 | */ | |
f3871fdc AK |
7716 | if (new_size <= 0) { |
7717 | /* OLCE size change is caused by added disks | |
7718 | */ | |
7719 | map = get_imsm_map(dev, MAP_0); | |
7720 | array_blocks = blocks_per_member(map) * used_disks; | |
7721 | } else { | |
7722 | /* Online Volume Size Change | |
7723 | * Using available free space | |
7724 | */ | |
7725 | array_blocks = new_size; | |
7726 | } | |
70bdf0dc AK |
7727 | |
7728 | /* round array size down to closest MB | |
7729 | */ | |
7730 | array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; | |
7731 | dev->size_low = __cpu_to_le32((__u32)array_blocks); | |
7732 | dev->size_high = __cpu_to_le32((__u32)(array_blocks >> 32)); | |
7733 | ||
7734 | return array_blocks; | |
7735 | } | |
7736 | ||
28bce06f AK |
7737 | static void imsm_set_disk(struct active_array *a, int n, int state); |
7738 | ||
0e2d1a4e AK |
7739 | static void imsm_progress_container_reshape(struct intel_super *super) |
7740 | { | |
7741 | /* if no device has a migr_state, but some device has a | |
7742 | * different number of members than the previous device, start | |
7743 | * changing the number of devices in this device to match | |
7744 | * previous. | |
7745 | */ | |
7746 | struct imsm_super *mpb = super->anchor; | |
7747 | int prev_disks = -1; | |
7748 | int i; | |
1dfaa380 | 7749 | int copy_map_size; |
0e2d1a4e AK |
7750 | |
7751 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
7752 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 7753 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
0e2d1a4e AK |
7754 | struct imsm_map *map2; |
7755 | int prev_num_members; | |
0e2d1a4e AK |
7756 | |
7757 | if (dev->vol.migr_state) | |
7758 | return; | |
7759 | ||
7760 | if (prev_disks == -1) | |
7761 | prev_disks = map->num_members; | |
7762 | if (prev_disks == map->num_members) | |
7763 | continue; | |
7764 | ||
7765 | /* OK, this array needs to enter reshape mode. | |
7766 | * i.e it needs a migr_state | |
7767 | */ | |
7768 | ||
1dfaa380 | 7769 | copy_map_size = sizeof_imsm_map(map); |
0e2d1a4e AK |
7770 | prev_num_members = map->num_members; |
7771 | map->num_members = prev_disks; | |
7772 | dev->vol.migr_state = 1; | |
7773 | dev->vol.curr_migr_unit = 0; | |
ea672ee1 | 7774 | set_migr_type(dev, MIGR_GEN_MIGR); |
0e2d1a4e AK |
7775 | for (i = prev_num_members; |
7776 | i < map->num_members; i++) | |
7777 | set_imsm_ord_tbl_ent(map, i, i); | |
238c0a71 | 7778 | map2 = get_imsm_map(dev, MAP_1); |
0e2d1a4e | 7779 | /* Copy the current map */ |
1dfaa380 | 7780 | memcpy(map2, map, copy_map_size); |
0e2d1a4e AK |
7781 | map2->num_members = prev_num_members; |
7782 | ||
f3871fdc | 7783 | imsm_set_array_size(dev, -1); |
51d83f5d | 7784 | super->clean_migration_record_by_mdmon = 1; |
0e2d1a4e AK |
7785 | super->updates_pending++; |
7786 | } | |
7787 | } | |
7788 | ||
aad6f216 | 7789 | /* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild |
0c046afd DW |
7790 | * states are handled in imsm_set_disk() with one exception, when a |
7791 | * resync is stopped due to a new failure this routine will set the | |
7792 | * 'degraded' state for the array. | |
7793 | */ | |
01f157d7 | 7794 | static int imsm_set_array_state(struct active_array *a, int consistent) |
a862209d DW |
7795 | { |
7796 | int inst = a->info.container_member; | |
7797 | struct intel_super *super = a->container->sb; | |
949c47a0 | 7798 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 7799 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
3b451610 AK |
7800 | int failed = imsm_count_failed(super, dev, MAP_0); |
7801 | __u8 map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
1e5c6983 | 7802 | __u32 blocks_per_unit; |
a862209d | 7803 | |
1af97990 AK |
7804 | if (dev->vol.migr_state && |
7805 | dev->vol.migr_type == MIGR_GEN_MIGR) { | |
7806 | /* array state change is blocked due to reshape action | |
aad6f216 N |
7807 | * We might need to |
7808 | * - abort the reshape (if last_checkpoint is 0 and action!= reshape) | |
7809 | * - finish the reshape (if last_checkpoint is big and action != reshape) | |
7810 | * - update curr_migr_unit | |
1af97990 | 7811 | */ |
aad6f216 N |
7812 | if (a->curr_action == reshape) { |
7813 | /* still reshaping, maybe update curr_migr_unit */ | |
633b5610 | 7814 | goto mark_checkpoint; |
aad6f216 N |
7815 | } else { |
7816 | if (a->last_checkpoint == 0 && a->prev_action == reshape) { | |
7817 | /* for some reason we aborted the reshape. | |
b66e591b AK |
7818 | * |
7819 | * disable automatic metadata rollback | |
7820 | * user action is required to recover process | |
aad6f216 | 7821 | */ |
b66e591b | 7822 | if (0) { |
238c0a71 AK |
7823 | struct imsm_map *map2 = |
7824 | get_imsm_map(dev, MAP_1); | |
7825 | dev->vol.migr_state = 0; | |
7826 | set_migr_type(dev, 0); | |
7827 | dev->vol.curr_migr_unit = 0; | |
7828 | memcpy(map, map2, | |
7829 | sizeof_imsm_map(map2)); | |
7830 | super->updates_pending++; | |
b66e591b | 7831 | } |
aad6f216 N |
7832 | } |
7833 | if (a->last_checkpoint >= a->info.component_size) { | |
7834 | unsigned long long array_blocks; | |
7835 | int used_disks; | |
e154ced3 | 7836 | struct mdinfo *mdi; |
aad6f216 | 7837 | |
238c0a71 | 7838 | used_disks = imsm_num_data_members(dev, MAP_0); |
d55adef9 AK |
7839 | if (used_disks > 0) { |
7840 | array_blocks = | |
5551b113 | 7841 | blocks_per_member(map) * |
d55adef9 AK |
7842 | used_disks; |
7843 | /* round array size down to closest MB | |
7844 | */ | |
7845 | array_blocks = (array_blocks | |
7846 | >> SECT_PER_MB_SHIFT) | |
7847 | << SECT_PER_MB_SHIFT; | |
d55adef9 AK |
7848 | a->info.custom_array_size = array_blocks; |
7849 | /* encourage manager to update array | |
7850 | * size | |
7851 | */ | |
e154ced3 | 7852 | |
d55adef9 | 7853 | a->check_reshape = 1; |
633b5610 | 7854 | } |
e154ced3 AK |
7855 | /* finalize online capacity expansion/reshape */ |
7856 | for (mdi = a->info.devs; mdi; mdi = mdi->next) | |
7857 | imsm_set_disk(a, | |
7858 | mdi->disk.raid_disk, | |
7859 | mdi->curr_state); | |
7860 | ||
0e2d1a4e | 7861 | imsm_progress_container_reshape(super); |
e154ced3 | 7862 | } |
aad6f216 | 7863 | } |
1af97990 AK |
7864 | } |
7865 | ||
47ee5a45 | 7866 | /* before we activate this array handle any missing disks */ |
33414a01 DW |
7867 | if (consistent == 2) |
7868 | handle_missing(super, dev); | |
1e5c6983 | 7869 | |
0c046afd | 7870 | if (consistent == 2 && |
b7941fd6 | 7871 | (!is_resync_complete(&a->info) || |
0c046afd DW |
7872 | map_state != IMSM_T_STATE_NORMAL || |
7873 | dev->vol.migr_state)) | |
01f157d7 | 7874 | consistent = 0; |
272906ef | 7875 | |
b7941fd6 | 7876 | if (is_resync_complete(&a->info)) { |
0c046afd | 7877 | /* complete intialization / resync, |
0556e1a2 DW |
7878 | * recovery and interrupted recovery is completed in |
7879 | * ->set_disk | |
0c046afd DW |
7880 | */ |
7881 | if (is_resyncing(dev)) { | |
7882 | dprintf("imsm: mark resync done\n"); | |
809da78e | 7883 | end_migration(dev, super, map_state); |
115c3803 | 7884 | super->updates_pending++; |
484240d8 | 7885 | a->last_checkpoint = 0; |
115c3803 | 7886 | } |
b9172665 AK |
7887 | } else if ((!is_resyncing(dev) && !failed) && |
7888 | (imsm_reshape_blocks_arrays_changes(super) == 0)) { | |
0c046afd | 7889 | /* mark the start of the init process if nothing is failed */ |
b7941fd6 | 7890 | dprintf("imsm: mark resync start\n"); |
1484e727 | 7891 | if (map->map_state == IMSM_T_STATE_UNINITIALIZED) |
8e59f3d8 | 7892 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT); |
1484e727 | 7893 | else |
8e59f3d8 | 7894 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR); |
3393c6af | 7895 | super->updates_pending++; |
115c3803 | 7896 | } |
a862209d | 7897 | |
633b5610 | 7898 | mark_checkpoint: |
5b83bacf AK |
7899 | /* skip checkpointing for general migration, |
7900 | * it is controlled in mdadm | |
7901 | */ | |
7902 | if (is_gen_migration(dev)) | |
7903 | goto skip_mark_checkpoint; | |
7904 | ||
1e5c6983 | 7905 | /* check if we can update curr_migr_unit from resync_start, recovery_start */ |
c47b0ff6 | 7906 | blocks_per_unit = blocks_per_migr_unit(super, dev); |
4f0a7acc | 7907 | if (blocks_per_unit) { |
1e5c6983 DW |
7908 | __u32 units32; |
7909 | __u64 units; | |
7910 | ||
4f0a7acc | 7911 | units = a->last_checkpoint / blocks_per_unit; |
1e5c6983 DW |
7912 | units32 = units; |
7913 | ||
7914 | /* check that we did not overflow 32-bits, and that | |
7915 | * curr_migr_unit needs updating | |
7916 | */ | |
7917 | if (units32 == units && | |
bfd80a56 | 7918 | units32 != 0 && |
1e5c6983 DW |
7919 | __le32_to_cpu(dev->vol.curr_migr_unit) != units32) { |
7920 | dprintf("imsm: mark checkpoint (%u)\n", units32); | |
7921 | dev->vol.curr_migr_unit = __cpu_to_le32(units32); | |
7922 | super->updates_pending++; | |
7923 | } | |
7924 | } | |
f8f603f1 | 7925 | |
5b83bacf | 7926 | skip_mark_checkpoint: |
3393c6af | 7927 | /* mark dirty / clean */ |
0c046afd | 7928 | if (dev->vol.dirty != !consistent) { |
b7941fd6 | 7929 | dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty"); |
0c046afd DW |
7930 | if (consistent) |
7931 | dev->vol.dirty = 0; | |
7932 | else | |
7933 | dev->vol.dirty = 1; | |
a862209d DW |
7934 | super->updates_pending++; |
7935 | } | |
28bce06f | 7936 | |
01f157d7 | 7937 | return consistent; |
a862209d DW |
7938 | } |
7939 | ||
6f50473f TM |
7940 | static int imsm_disk_slot_to_ord(struct active_array *a, int slot) |
7941 | { | |
7942 | int inst = a->info.container_member; | |
7943 | struct intel_super *super = a->container->sb; | |
7944 | struct imsm_dev *dev = get_imsm_dev(super, inst); | |
7945 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
7946 | ||
7947 | if (slot > map->num_members) { | |
7948 | pr_err("imsm: imsm_disk_slot_to_ord %d out of range 0..%d\n", | |
7949 | slot, map->num_members - 1); | |
7950 | return -1; | |
7951 | } | |
7952 | ||
7953 | if (slot < 0) | |
7954 | return -1; | |
7955 | ||
7956 | return get_imsm_ord_tbl_ent(dev, slot, MAP_0); | |
7957 | } | |
7958 | ||
8d45d196 | 7959 | static void imsm_set_disk(struct active_array *a, int n, int state) |
845dea95 | 7960 | { |
8d45d196 DW |
7961 | int inst = a->info.container_member; |
7962 | struct intel_super *super = a->container->sb; | |
949c47a0 | 7963 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 7964 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
8d45d196 | 7965 | struct imsm_disk *disk; |
7ce05701 LD |
7966 | struct mdinfo *mdi; |
7967 | int recovery_not_finished = 0; | |
0c046afd | 7968 | int failed; |
6f50473f | 7969 | int ord; |
0c046afd | 7970 | __u8 map_state; |
8d45d196 | 7971 | |
6f50473f TM |
7972 | ord = imsm_disk_slot_to_ord(a, n); |
7973 | if (ord < 0) | |
8d45d196 DW |
7974 | return; |
7975 | ||
4e6e574a | 7976 | dprintf("imsm: set_disk %d:%x\n", n, state); |
b10b37b8 | 7977 | disk = get_imsm_disk(super, ord_to_idx(ord)); |
8d45d196 | 7978 | |
5802a811 | 7979 | /* check for new failures */ |
0556e1a2 | 7980 | if (state & DS_FAULTY) { |
4c9e8c1e | 7981 | if (mark_failure(super, dev, disk, ord_to_idx(ord))) |
0556e1a2 | 7982 | super->updates_pending++; |
8d45d196 | 7983 | } |
47ee5a45 | 7984 | |
19859edc | 7985 | /* check if in_sync */ |
0556e1a2 | 7986 | if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) { |
238c0a71 | 7987 | struct imsm_map *migr_map = get_imsm_map(dev, MAP_1); |
b10b37b8 DW |
7988 | |
7989 | set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord)); | |
19859edc DW |
7990 | super->updates_pending++; |
7991 | } | |
8d45d196 | 7992 | |
3b451610 AK |
7993 | failed = imsm_count_failed(super, dev, MAP_0); |
7994 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
5802a811 | 7995 | |
0c046afd | 7996 | /* check if recovery complete, newly degraded, or failed */ |
94002678 AK |
7997 | dprintf("imsm: Detected transition to state "); |
7998 | switch (map_state) { | |
7999 | case IMSM_T_STATE_NORMAL: /* transition to normal state */ | |
8000 | dprintf("normal: "); | |
8001 | if (is_rebuilding(dev)) { | |
1ade5cc1 | 8002 | dprintf_cont("while rebuilding"); |
7ce05701 LD |
8003 | /* check if recovery is really finished */ |
8004 | for (mdi = a->info.devs; mdi ; mdi = mdi->next) | |
8005 | if (mdi->recovery_start != MaxSector) { | |
8006 | recovery_not_finished = 1; | |
8007 | break; | |
8008 | } | |
8009 | if (recovery_not_finished) { | |
1ade5cc1 N |
8010 | dprintf_cont("\n"); |
8011 | dprintf("Rebuild has not finished yet, state not changed"); | |
7ce05701 LD |
8012 | if (a->last_checkpoint < mdi->recovery_start) { |
8013 | a->last_checkpoint = mdi->recovery_start; | |
8014 | super->updates_pending++; | |
8015 | } | |
8016 | break; | |
8017 | } | |
94002678 | 8018 | end_migration(dev, super, map_state); |
238c0a71 | 8019 | map = get_imsm_map(dev, MAP_0); |
94002678 AK |
8020 | map->failed_disk_num = ~0; |
8021 | super->updates_pending++; | |
8022 | a->last_checkpoint = 0; | |
8023 | break; | |
8024 | } | |
8025 | if (is_gen_migration(dev)) { | |
1ade5cc1 | 8026 | dprintf_cont("while general migration"); |
bf2f0071 | 8027 | if (a->last_checkpoint >= a->info.component_size) |
809da78e | 8028 | end_migration(dev, super, map_state); |
94002678 AK |
8029 | else |
8030 | map->map_state = map_state; | |
238c0a71 | 8031 | map = get_imsm_map(dev, MAP_0); |
28bce06f | 8032 | map->failed_disk_num = ~0; |
94002678 | 8033 | super->updates_pending++; |
bf2f0071 | 8034 | break; |
94002678 AK |
8035 | } |
8036 | break; | |
8037 | case IMSM_T_STATE_DEGRADED: /* transition to degraded state */ | |
1ade5cc1 | 8038 | dprintf_cont("degraded: "); |
089f9d79 | 8039 | if (map->map_state != map_state && !dev->vol.migr_state) { |
1ade5cc1 | 8040 | dprintf_cont("mark degraded"); |
94002678 AK |
8041 | map->map_state = map_state; |
8042 | super->updates_pending++; | |
8043 | a->last_checkpoint = 0; | |
8044 | break; | |
8045 | } | |
8046 | if (is_rebuilding(dev)) { | |
1ade5cc1 | 8047 | dprintf_cont("while rebuilding."); |
94002678 | 8048 | if (map->map_state != map_state) { |
1ade5cc1 | 8049 | dprintf_cont(" Map state change"); |
94002678 AK |
8050 | end_migration(dev, super, map_state); |
8051 | super->updates_pending++; | |
8052 | } | |
8053 | break; | |
8054 | } | |
8055 | if (is_gen_migration(dev)) { | |
1ade5cc1 | 8056 | dprintf_cont("while general migration"); |
bf2f0071 | 8057 | if (a->last_checkpoint >= a->info.component_size) |
809da78e | 8058 | end_migration(dev, super, map_state); |
94002678 AK |
8059 | else { |
8060 | map->map_state = map_state; | |
3b451610 | 8061 | manage_second_map(super, dev); |
94002678 AK |
8062 | } |
8063 | super->updates_pending++; | |
bf2f0071 | 8064 | break; |
28bce06f | 8065 | } |
6ce1fbf1 | 8066 | if (is_initializing(dev)) { |
1ade5cc1 | 8067 | dprintf_cont("while initialization."); |
6ce1fbf1 AK |
8068 | map->map_state = map_state; |
8069 | super->updates_pending++; | |
8070 | break; | |
8071 | } | |
94002678 AK |
8072 | break; |
8073 | case IMSM_T_STATE_FAILED: /* transition to failed state */ | |
1ade5cc1 | 8074 | dprintf_cont("failed: "); |
94002678 | 8075 | if (is_gen_migration(dev)) { |
1ade5cc1 | 8076 | dprintf_cont("while general migration"); |
94002678 AK |
8077 | map->map_state = map_state; |
8078 | super->updates_pending++; | |
8079 | break; | |
8080 | } | |
8081 | if (map->map_state != map_state) { | |
1ade5cc1 | 8082 | dprintf_cont("mark failed"); |
94002678 AK |
8083 | end_migration(dev, super, map_state); |
8084 | super->updates_pending++; | |
8085 | a->last_checkpoint = 0; | |
8086 | break; | |
8087 | } | |
8088 | break; | |
8089 | default: | |
1ade5cc1 | 8090 | dprintf_cont("state %i\n", map_state); |
5802a811 | 8091 | } |
1ade5cc1 | 8092 | dprintf_cont("\n"); |
845dea95 NB |
8093 | } |
8094 | ||
f796af5d | 8095 | static int store_imsm_mpb(int fd, struct imsm_super *mpb) |
c2a1e7da | 8096 | { |
f796af5d | 8097 | void *buf = mpb; |
c2a1e7da DW |
8098 | __u32 mpb_size = __le32_to_cpu(mpb->mpb_size); |
8099 | unsigned long long dsize; | |
8100 | unsigned long long sectors; | |
f36a9ecd | 8101 | unsigned int sector_size; |
c2a1e7da | 8102 | |
f36a9ecd | 8103 | get_dev_sector_size(fd, NULL, §or_size); |
c2a1e7da DW |
8104 | get_dev_size(fd, NULL, &dsize); |
8105 | ||
f36a9ecd | 8106 | if (mpb_size > sector_size) { |
272f648f | 8107 | /* -1 to account for anchor */ |
f36a9ecd | 8108 | sectors = mpb_sectors(mpb, sector_size) - 1; |
c2a1e7da | 8109 | |
272f648f | 8110 | /* write the extended mpb to the sectors preceeding the anchor */ |
f36a9ecd PB |
8111 | if (lseek64(fd, dsize - (sector_size * (2 + sectors)), |
8112 | SEEK_SET) < 0) | |
272f648f | 8113 | return 1; |
c2a1e7da | 8114 | |
f36a9ecd PB |
8115 | if ((unsigned long long)write(fd, buf + sector_size, |
8116 | sector_size * sectors) != sector_size * sectors) | |
272f648f DW |
8117 | return 1; |
8118 | } | |
c2a1e7da | 8119 | |
272f648f | 8120 | /* first block is stored on second to last sector of the disk */ |
f36a9ecd | 8121 | if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0) |
c2a1e7da DW |
8122 | return 1; |
8123 | ||
466070ad | 8124 | if ((unsigned int)write(fd, buf, sector_size) != sector_size) |
c2a1e7da DW |
8125 | return 1; |
8126 | ||
c2a1e7da DW |
8127 | return 0; |
8128 | } | |
8129 | ||
2e735d19 | 8130 | static void imsm_sync_metadata(struct supertype *container) |
845dea95 | 8131 | { |
2e735d19 | 8132 | struct intel_super *super = container->sb; |
c2a1e7da | 8133 | |
1a64be56 | 8134 | dprintf("sync metadata: %d\n", super->updates_pending); |
c2a1e7da DW |
8135 | if (!super->updates_pending) |
8136 | return; | |
8137 | ||
36988a3d | 8138 | write_super_imsm(container, 0); |
c2a1e7da DW |
8139 | |
8140 | super->updates_pending = 0; | |
845dea95 NB |
8141 | } |
8142 | ||
272906ef DW |
8143 | static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a) |
8144 | { | |
8145 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 8146 | int i = get_imsm_disk_idx(dev, idx, MAP_X); |
272906ef DW |
8147 | struct dl *dl; |
8148 | ||
8149 | for (dl = super->disks; dl; dl = dl->next) | |
8150 | if (dl->index == i) | |
8151 | break; | |
8152 | ||
25ed7e59 | 8153 | if (dl && is_failed(&dl->disk)) |
272906ef DW |
8154 | dl = NULL; |
8155 | ||
8156 | if (dl) | |
1ade5cc1 | 8157 | dprintf("found %x:%x\n", dl->major, dl->minor); |
272906ef DW |
8158 | |
8159 | return dl; | |
8160 | } | |
8161 | ||
a20d2ba5 | 8162 | static struct dl *imsm_add_spare(struct intel_super *super, int slot, |
8ba77d32 AK |
8163 | struct active_array *a, int activate_new, |
8164 | struct mdinfo *additional_test_list) | |
272906ef DW |
8165 | { |
8166 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 8167 | int idx = get_imsm_disk_idx(dev, slot, MAP_X); |
a20d2ba5 DW |
8168 | struct imsm_super *mpb = super->anchor; |
8169 | struct imsm_map *map; | |
272906ef DW |
8170 | unsigned long long pos; |
8171 | struct mdinfo *d; | |
8172 | struct extent *ex; | |
a20d2ba5 | 8173 | int i, j; |
272906ef | 8174 | int found; |
569cc43f DW |
8175 | __u32 array_start = 0; |
8176 | __u32 array_end = 0; | |
272906ef | 8177 | struct dl *dl; |
6c932028 | 8178 | struct mdinfo *test_list; |
272906ef DW |
8179 | |
8180 | for (dl = super->disks; dl; dl = dl->next) { | |
8181 | /* If in this array, skip */ | |
8182 | for (d = a->info.devs ; d ; d = d->next) | |
e553d2a4 DW |
8183 | if (d->state_fd >= 0 && |
8184 | d->disk.major == dl->major && | |
272906ef | 8185 | d->disk.minor == dl->minor) { |
8ba77d32 AK |
8186 | dprintf("%x:%x already in array\n", |
8187 | dl->major, dl->minor); | |
272906ef DW |
8188 | break; |
8189 | } | |
8190 | if (d) | |
8191 | continue; | |
6c932028 AK |
8192 | test_list = additional_test_list; |
8193 | while (test_list) { | |
8194 | if (test_list->disk.major == dl->major && | |
8195 | test_list->disk.minor == dl->minor) { | |
8ba77d32 AK |
8196 | dprintf("%x:%x already in additional test list\n", |
8197 | dl->major, dl->minor); | |
8198 | break; | |
8199 | } | |
6c932028 | 8200 | test_list = test_list->next; |
8ba77d32 | 8201 | } |
6c932028 | 8202 | if (test_list) |
8ba77d32 | 8203 | continue; |
272906ef | 8204 | |
e553d2a4 | 8205 | /* skip in use or failed drives */ |
25ed7e59 | 8206 | if (is_failed(&dl->disk) || idx == dl->index || |
df474657 DW |
8207 | dl->index == -2) { |
8208 | dprintf("%x:%x status (failed: %d index: %d)\n", | |
25ed7e59 | 8209 | dl->major, dl->minor, is_failed(&dl->disk), idx); |
9a1608e5 DW |
8210 | continue; |
8211 | } | |
8212 | ||
a20d2ba5 DW |
8213 | /* skip pure spares when we are looking for partially |
8214 | * assimilated drives | |
8215 | */ | |
8216 | if (dl->index == -1 && !activate_new) | |
8217 | continue; | |
8218 | ||
272906ef | 8219 | /* Does this unused device have the requisite free space? |
a20d2ba5 | 8220 | * It needs to be able to cover all member volumes |
272906ef DW |
8221 | */ |
8222 | ex = get_extents(super, dl); | |
8223 | if (!ex) { | |
8224 | dprintf("cannot get extents\n"); | |
8225 | continue; | |
8226 | } | |
a20d2ba5 DW |
8227 | for (i = 0; i < mpb->num_raid_devs; i++) { |
8228 | dev = get_imsm_dev(super, i); | |
238c0a71 | 8229 | map = get_imsm_map(dev, MAP_0); |
272906ef | 8230 | |
a20d2ba5 DW |
8231 | /* check if this disk is already a member of |
8232 | * this array | |
272906ef | 8233 | */ |
620b1713 | 8234 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
a20d2ba5 DW |
8235 | continue; |
8236 | ||
8237 | found = 0; | |
8238 | j = 0; | |
8239 | pos = 0; | |
5551b113 | 8240 | array_start = pba_of_lba0(map); |
329c8278 | 8241 | array_end = array_start + |
5551b113 | 8242 | blocks_per_member(map) - 1; |
a20d2ba5 DW |
8243 | |
8244 | do { | |
8245 | /* check that we can start at pba_of_lba0 with | |
8246 | * blocks_per_member of space | |
8247 | */ | |
329c8278 | 8248 | if (array_start >= pos && array_end < ex[j].start) { |
a20d2ba5 DW |
8249 | found = 1; |
8250 | break; | |
8251 | } | |
8252 | pos = ex[j].start + ex[j].size; | |
8253 | j++; | |
8254 | } while (ex[j-1].size); | |
8255 | ||
8256 | if (!found) | |
272906ef | 8257 | break; |
a20d2ba5 | 8258 | } |
272906ef DW |
8259 | |
8260 | free(ex); | |
a20d2ba5 | 8261 | if (i < mpb->num_raid_devs) { |
329c8278 DW |
8262 | dprintf("%x:%x does not have %u to %u available\n", |
8263 | dl->major, dl->minor, array_start, array_end); | |
272906ef DW |
8264 | /* No room */ |
8265 | continue; | |
a20d2ba5 DW |
8266 | } |
8267 | return dl; | |
272906ef DW |
8268 | } |
8269 | ||
8270 | return dl; | |
8271 | } | |
8272 | ||
95d07a2c LM |
8273 | static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed) |
8274 | { | |
8275 | struct imsm_dev *dev2; | |
8276 | struct imsm_map *map; | |
8277 | struct dl *idisk; | |
8278 | int slot; | |
8279 | int idx; | |
8280 | __u8 state; | |
8281 | ||
8282 | dev2 = get_imsm_dev(cont->sb, dev_idx); | |
8283 | if (dev2) { | |
238c0a71 | 8284 | state = imsm_check_degraded(cont->sb, dev2, failed, MAP_0); |
95d07a2c | 8285 | if (state == IMSM_T_STATE_FAILED) { |
238c0a71 | 8286 | map = get_imsm_map(dev2, MAP_0); |
95d07a2c LM |
8287 | if (!map) |
8288 | return 1; | |
8289 | for (slot = 0; slot < map->num_members; slot++) { | |
8290 | /* | |
8291 | * Check if failed disks are deleted from intel | |
8292 | * disk list or are marked to be deleted | |
8293 | */ | |
238c0a71 | 8294 | idx = get_imsm_disk_idx(dev2, slot, MAP_X); |
95d07a2c LM |
8295 | idisk = get_imsm_dl_disk(cont->sb, idx); |
8296 | /* | |
8297 | * Do not rebuild the array if failed disks | |
8298 | * from failed sub-array are not removed from | |
8299 | * container. | |
8300 | */ | |
8301 | if (idisk && | |
8302 | is_failed(&idisk->disk) && | |
8303 | (idisk->action != DISK_REMOVE)) | |
8304 | return 0; | |
8305 | } | |
8306 | } | |
8307 | } | |
8308 | return 1; | |
8309 | } | |
8310 | ||
88758e9d DW |
8311 | static struct mdinfo *imsm_activate_spare(struct active_array *a, |
8312 | struct metadata_update **updates) | |
8313 | { | |
8314 | /** | |
d23fe947 DW |
8315 | * Find a device with unused free space and use it to replace a |
8316 | * failed/vacant region in an array. We replace failed regions one a | |
8317 | * array at a time. The result is that a new spare disk will be added | |
8318 | * to the first failed array and after the monitor has finished | |
8319 | * propagating failures the remainder will be consumed. | |
88758e9d | 8320 | * |
d23fe947 DW |
8321 | * FIXME add a capability for mdmon to request spares from another |
8322 | * container. | |
88758e9d DW |
8323 | */ |
8324 | ||
8325 | struct intel_super *super = a->container->sb; | |
88758e9d | 8326 | int inst = a->info.container_member; |
949c47a0 | 8327 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 8328 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
88758e9d DW |
8329 | int failed = a->info.array.raid_disks; |
8330 | struct mdinfo *rv = NULL; | |
8331 | struct mdinfo *d; | |
8332 | struct mdinfo *di; | |
8333 | struct metadata_update *mu; | |
8334 | struct dl *dl; | |
8335 | struct imsm_update_activate_spare *u; | |
8336 | int num_spares = 0; | |
8337 | int i; | |
95d07a2c | 8338 | int allowed; |
88758e9d DW |
8339 | |
8340 | for (d = a->info.devs ; d ; d = d->next) { | |
8341 | if ((d->curr_state & DS_FAULTY) && | |
8342 | d->state_fd >= 0) | |
8343 | /* wait for Removal to happen */ | |
8344 | return NULL; | |
8345 | if (d->state_fd >= 0) | |
8346 | failed--; | |
8347 | } | |
8348 | ||
8349 | dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n", | |
8350 | inst, failed, a->info.array.raid_disks, a->info.array.level); | |
1af97990 | 8351 | |
e2962bfc AK |
8352 | if (imsm_reshape_blocks_arrays_changes(super)) |
8353 | return NULL; | |
1af97990 | 8354 | |
fc8ca064 AK |
8355 | /* Cannot activate another spare if rebuild is in progress already |
8356 | */ | |
8357 | if (is_rebuilding(dev)) { | |
7a862a02 | 8358 | dprintf("imsm: No spare activation allowed. Rebuild in progress already.\n"); |
fc8ca064 AK |
8359 | return NULL; |
8360 | } | |
8361 | ||
89c67882 AK |
8362 | if (a->info.array.level == 4) |
8363 | /* No repair for takeovered array | |
8364 | * imsm doesn't support raid4 | |
8365 | */ | |
8366 | return NULL; | |
8367 | ||
3b451610 AK |
8368 | if (imsm_check_degraded(super, dev, failed, MAP_0) != |
8369 | IMSM_T_STATE_DEGRADED) | |
88758e9d DW |
8370 | return NULL; |
8371 | ||
83ca7d45 AP |
8372 | if (get_imsm_map(dev, MAP_0)->map_state == IMSM_T_STATE_UNINITIALIZED) { |
8373 | dprintf("imsm: No spare activation allowed. Volume is not initialized.\n"); | |
8374 | return NULL; | |
8375 | } | |
8376 | ||
95d07a2c LM |
8377 | /* |
8378 | * If there are any failed disks check state of the other volume. | |
8379 | * Block rebuild if the another one is failed until failed disks | |
8380 | * are removed from container. | |
8381 | */ | |
8382 | if (failed) { | |
7a862a02 | 8383 | dprintf("found failed disks in %.*s, check if there anotherfailed sub-array.\n", |
c4acd1e5 | 8384 | MAX_RAID_SERIAL_LEN, dev->volume); |
95d07a2c LM |
8385 | /* check if states of the other volumes allow for rebuild */ |
8386 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
8387 | if (i != inst) { | |
8388 | allowed = imsm_rebuild_allowed(a->container, | |
8389 | i, failed); | |
8390 | if (!allowed) | |
8391 | return NULL; | |
8392 | } | |
8393 | } | |
8394 | } | |
8395 | ||
88758e9d | 8396 | /* For each slot, if it is not working, find a spare */ |
88758e9d DW |
8397 | for (i = 0; i < a->info.array.raid_disks; i++) { |
8398 | for (d = a->info.devs ; d ; d = d->next) | |
8399 | if (d->disk.raid_disk == i) | |
8400 | break; | |
8401 | dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0); | |
8402 | if (d && (d->state_fd >= 0)) | |
8403 | continue; | |
8404 | ||
272906ef | 8405 | /* |
a20d2ba5 DW |
8406 | * OK, this device needs recovery. Try to re-add the |
8407 | * previous occupant of this slot, if this fails see if | |
8408 | * we can continue the assimilation of a spare that was | |
8409 | * partially assimilated, finally try to activate a new | |
8410 | * spare. | |
272906ef DW |
8411 | */ |
8412 | dl = imsm_readd(super, i, a); | |
8413 | if (!dl) | |
b303fe21 | 8414 | dl = imsm_add_spare(super, i, a, 0, rv); |
a20d2ba5 | 8415 | if (!dl) |
b303fe21 | 8416 | dl = imsm_add_spare(super, i, a, 1, rv); |
272906ef DW |
8417 | if (!dl) |
8418 | continue; | |
1011e834 | 8419 | |
272906ef | 8420 | /* found a usable disk with enough space */ |
503975b9 | 8421 | di = xcalloc(1, sizeof(*di)); |
272906ef DW |
8422 | |
8423 | /* dl->index will be -1 in the case we are activating a | |
8424 | * pristine spare. imsm_process_update() will create a | |
8425 | * new index in this case. Once a disk is found to be | |
8426 | * failed in all member arrays it is kicked from the | |
8427 | * metadata | |
8428 | */ | |
8429 | di->disk.number = dl->index; | |
d23fe947 | 8430 | |
272906ef DW |
8431 | /* (ab)use di->devs to store a pointer to the device |
8432 | * we chose | |
8433 | */ | |
8434 | di->devs = (struct mdinfo *) dl; | |
8435 | ||
8436 | di->disk.raid_disk = i; | |
8437 | di->disk.major = dl->major; | |
8438 | di->disk.minor = dl->minor; | |
8439 | di->disk.state = 0; | |
d23534e4 | 8440 | di->recovery_start = 0; |
5551b113 | 8441 | di->data_offset = pba_of_lba0(map); |
272906ef DW |
8442 | di->component_size = a->info.component_size; |
8443 | di->container_member = inst; | |
5e46202e | 8444 | di->bb.supported = 1; |
148acb7b | 8445 | super->random = random32(); |
272906ef DW |
8446 | di->next = rv; |
8447 | rv = di; | |
8448 | num_spares++; | |
8449 | dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor, | |
8450 | i, di->data_offset); | |
88758e9d DW |
8451 | } |
8452 | ||
8453 | if (!rv) | |
8454 | /* No spares found */ | |
8455 | return rv; | |
8456 | /* Now 'rv' has a list of devices to return. | |
8457 | * Create a metadata_update record to update the | |
8458 | * disk_ord_tbl for the array | |
8459 | */ | |
503975b9 | 8460 | mu = xmalloc(sizeof(*mu)); |
1011e834 | 8461 | mu->buf = xcalloc(num_spares, |
503975b9 | 8462 | sizeof(struct imsm_update_activate_spare)); |
88758e9d | 8463 | mu->space = NULL; |
cb23f1f4 | 8464 | mu->space_list = NULL; |
88758e9d DW |
8465 | mu->len = sizeof(struct imsm_update_activate_spare) * num_spares; |
8466 | mu->next = *updates; | |
8467 | u = (struct imsm_update_activate_spare *) mu->buf; | |
8468 | ||
8469 | for (di = rv ; di ; di = di->next) { | |
8470 | u->type = update_activate_spare; | |
d23fe947 DW |
8471 | u->dl = (struct dl *) di->devs; |
8472 | di->devs = NULL; | |
88758e9d DW |
8473 | u->slot = di->disk.raid_disk; |
8474 | u->array = inst; | |
8475 | u->next = u + 1; | |
8476 | u++; | |
8477 | } | |
8478 | (u-1)->next = NULL; | |
8479 | *updates = mu; | |
8480 | ||
8481 | return rv; | |
8482 | } | |
8483 | ||
54c2c1ea | 8484 | static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u) |
8273f55e | 8485 | { |
54c2c1ea | 8486 | struct imsm_dev *dev = get_imsm_dev(super, idx); |
238c0a71 AK |
8487 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
8488 | struct imsm_map *new_map = get_imsm_map(&u->dev, MAP_0); | |
54c2c1ea DW |
8489 | struct disk_info *inf = get_disk_info(u); |
8490 | struct imsm_disk *disk; | |
8273f55e DW |
8491 | int i; |
8492 | int j; | |
8273f55e | 8493 | |
54c2c1ea | 8494 | for (i = 0; i < map->num_members; i++) { |
238c0a71 | 8495 | disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, MAP_X)); |
54c2c1ea DW |
8496 | for (j = 0; j < new_map->num_members; j++) |
8497 | if (serialcmp(disk->serial, inf[j].serial) == 0) | |
8273f55e DW |
8498 | return 1; |
8499 | } | |
8500 | ||
8501 | return 0; | |
8502 | } | |
8503 | ||
1a64be56 LM |
8504 | static struct dl *get_disk_super(struct intel_super *super, int major, int minor) |
8505 | { | |
594dc1b8 JS |
8506 | struct dl *dl; |
8507 | ||
1a64be56 | 8508 | for (dl = super->disks; dl; dl = dl->next) |
089f9d79 | 8509 | if (dl->major == major && dl->minor == minor) |
1a64be56 LM |
8510 | return dl; |
8511 | return NULL; | |
8512 | } | |
8513 | ||
8514 | static int remove_disk_super(struct intel_super *super, int major, int minor) | |
8515 | { | |
594dc1b8 | 8516 | struct dl *prev; |
1a64be56 LM |
8517 | struct dl *dl; |
8518 | ||
8519 | prev = NULL; | |
8520 | for (dl = super->disks; dl; dl = dl->next) { | |
089f9d79 | 8521 | if (dl->major == major && dl->minor == minor) { |
1a64be56 LM |
8522 | /* remove */ |
8523 | if (prev) | |
8524 | prev->next = dl->next; | |
8525 | else | |
8526 | super->disks = dl->next; | |
8527 | dl->next = NULL; | |
8528 | __free_imsm_disk(dl); | |
1ade5cc1 | 8529 | dprintf("removed %x:%x\n", major, minor); |
1a64be56 LM |
8530 | break; |
8531 | } | |
8532 | prev = dl; | |
8533 | } | |
8534 | return 0; | |
8535 | } | |
8536 | ||
f21e18ca | 8537 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index); |
ae6aad82 | 8538 | |
1a64be56 LM |
8539 | static int add_remove_disk_update(struct intel_super *super) |
8540 | { | |
8541 | int check_degraded = 0; | |
594dc1b8 JS |
8542 | struct dl *disk; |
8543 | ||
1a64be56 LM |
8544 | /* add/remove some spares to/from the metadata/contrainer */ |
8545 | while (super->disk_mgmt_list) { | |
8546 | struct dl *disk_cfg; | |
8547 | ||
8548 | disk_cfg = super->disk_mgmt_list; | |
8549 | super->disk_mgmt_list = disk_cfg->next; | |
8550 | disk_cfg->next = NULL; | |
8551 | ||
8552 | if (disk_cfg->action == DISK_ADD) { | |
8553 | disk_cfg->next = super->disks; | |
8554 | super->disks = disk_cfg; | |
8555 | check_degraded = 1; | |
1ade5cc1 N |
8556 | dprintf("added %x:%x\n", |
8557 | disk_cfg->major, disk_cfg->minor); | |
1a64be56 LM |
8558 | } else if (disk_cfg->action == DISK_REMOVE) { |
8559 | dprintf("Disk remove action processed: %x.%x\n", | |
8560 | disk_cfg->major, disk_cfg->minor); | |
8561 | disk = get_disk_super(super, | |
8562 | disk_cfg->major, | |
8563 | disk_cfg->minor); | |
8564 | if (disk) { | |
8565 | /* store action status */ | |
8566 | disk->action = DISK_REMOVE; | |
8567 | /* remove spare disks only */ | |
8568 | if (disk->index == -1) { | |
8569 | remove_disk_super(super, | |
8570 | disk_cfg->major, | |
8571 | disk_cfg->minor); | |
8572 | } | |
8573 | } | |
8574 | /* release allocate disk structure */ | |
8575 | __free_imsm_disk(disk_cfg); | |
8576 | } | |
8577 | } | |
8578 | return check_degraded; | |
8579 | } | |
8580 | ||
a29911da PC |
8581 | static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u, |
8582 | struct intel_super *super, | |
8583 | void ***space_list) | |
8584 | { | |
8585 | struct intel_dev *id; | |
8586 | void **tofree = NULL; | |
8587 | int ret_val = 0; | |
8588 | ||
1ade5cc1 | 8589 | dprintf("(enter)\n"); |
089f9d79 | 8590 | if (u->subdev < 0 || u->subdev > 1) { |
a29911da PC |
8591 | dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev); |
8592 | return ret_val; | |
8593 | } | |
089f9d79 | 8594 | if (space_list == NULL || *space_list == NULL) { |
a29911da PC |
8595 | dprintf("imsm: Error: Memory is not allocated\n"); |
8596 | return ret_val; | |
8597 | } | |
8598 | ||
8599 | for (id = super->devlist ; id; id = id->next) { | |
8600 | if (id->index == (unsigned)u->subdev) { | |
8601 | struct imsm_dev *dev = get_imsm_dev(super, u->subdev); | |
8602 | struct imsm_map *map; | |
8603 | struct imsm_dev *new_dev = | |
8604 | (struct imsm_dev *)*space_list; | |
238c0a71 | 8605 | struct imsm_map *migr_map = get_imsm_map(dev, MAP_1); |
a29911da PC |
8606 | int to_state; |
8607 | struct dl *new_disk; | |
8608 | ||
8609 | if (new_dev == NULL) | |
8610 | return ret_val; | |
8611 | *space_list = **space_list; | |
8612 | memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0)); | |
238c0a71 | 8613 | map = get_imsm_map(new_dev, MAP_0); |
a29911da PC |
8614 | if (migr_map) { |
8615 | dprintf("imsm: Error: migration in progress"); | |
8616 | return ret_val; | |
8617 | } | |
8618 | ||
8619 | to_state = map->map_state; | |
8620 | if ((u->new_level == 5) && (map->raid_level == 0)) { | |
8621 | map->num_members++; | |
8622 | /* this should not happen */ | |
8623 | if (u->new_disks[0] < 0) { | |
8624 | map->failed_disk_num = | |
8625 | map->num_members - 1; | |
8626 | to_state = IMSM_T_STATE_DEGRADED; | |
8627 | } else | |
8628 | to_state = IMSM_T_STATE_NORMAL; | |
8629 | } | |
8e59f3d8 | 8630 | migrate(new_dev, super, to_state, MIGR_GEN_MIGR); |
a29911da PC |
8631 | if (u->new_level > -1) |
8632 | map->raid_level = u->new_level; | |
238c0a71 | 8633 | migr_map = get_imsm_map(new_dev, MAP_1); |
a29911da PC |
8634 | if ((u->new_level == 5) && |
8635 | (migr_map->raid_level == 0)) { | |
8636 | int ord = map->num_members - 1; | |
8637 | migr_map->num_members--; | |
8638 | if (u->new_disks[0] < 0) | |
8639 | ord |= IMSM_ORD_REBUILD; | |
8640 | set_imsm_ord_tbl_ent(map, | |
8641 | map->num_members - 1, | |
8642 | ord); | |
8643 | } | |
8644 | id->dev = new_dev; | |
8645 | tofree = (void **)dev; | |
8646 | ||
4bba0439 PC |
8647 | /* update chunk size |
8648 | */ | |
06fb291a PB |
8649 | if (u->new_chunksize > 0) { |
8650 | unsigned long long num_data_stripes; | |
8651 | int used_disks = | |
8652 | imsm_num_data_members(dev, MAP_0); | |
8653 | ||
8654 | if (used_disks == 0) | |
8655 | return ret_val; | |
8656 | ||
4bba0439 PC |
8657 | map->blocks_per_strip = |
8658 | __cpu_to_le16(u->new_chunksize * 2); | |
06fb291a PB |
8659 | num_data_stripes = |
8660 | (join_u32(dev->size_low, dev->size_high) | |
8661 | / used_disks); | |
8662 | num_data_stripes /= map->blocks_per_strip; | |
8663 | num_data_stripes /= map->num_domains; | |
8664 | set_num_data_stripes(map, num_data_stripes); | |
8665 | } | |
4bba0439 | 8666 | |
a29911da PC |
8667 | /* add disk |
8668 | */ | |
089f9d79 JS |
8669 | if (u->new_level != 5 || migr_map->raid_level != 0 || |
8670 | migr_map->raid_level == map->raid_level) | |
a29911da PC |
8671 | goto skip_disk_add; |
8672 | ||
8673 | if (u->new_disks[0] >= 0) { | |
8674 | /* use passes spare | |
8675 | */ | |
8676 | new_disk = get_disk_super(super, | |
8677 | major(u->new_disks[0]), | |
8678 | minor(u->new_disks[0])); | |
7a862a02 | 8679 | dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n", |
a29911da PC |
8680 | major(u->new_disks[0]), |
8681 | minor(u->new_disks[0]), | |
8682 | new_disk, new_disk->index); | |
8683 | if (new_disk == NULL) | |
8684 | goto error_disk_add; | |
8685 | ||
8686 | new_disk->index = map->num_members - 1; | |
8687 | /* slot to fill in autolayout | |
8688 | */ | |
8689 | new_disk->raiddisk = new_disk->index; | |
8690 | new_disk->disk.status |= CONFIGURED_DISK; | |
8691 | new_disk->disk.status &= ~SPARE_DISK; | |
8692 | } else | |
8693 | goto error_disk_add; | |
8694 | ||
8695 | skip_disk_add: | |
8696 | *tofree = *space_list; | |
8697 | /* calculate new size | |
8698 | */ | |
f3871fdc | 8699 | imsm_set_array_size(new_dev, -1); |
a29911da PC |
8700 | |
8701 | ret_val = 1; | |
8702 | } | |
8703 | } | |
8704 | ||
8705 | if (tofree) | |
8706 | *space_list = tofree; | |
8707 | return ret_val; | |
8708 | ||
8709 | error_disk_add: | |
8710 | dprintf("Error: imsm: Cannot find disk.\n"); | |
8711 | return ret_val; | |
8712 | } | |
8713 | ||
f3871fdc AK |
8714 | static int apply_size_change_update(struct imsm_update_size_change *u, |
8715 | struct intel_super *super) | |
8716 | { | |
8717 | struct intel_dev *id; | |
8718 | int ret_val = 0; | |
8719 | ||
1ade5cc1 | 8720 | dprintf("(enter)\n"); |
089f9d79 | 8721 | if (u->subdev < 0 || u->subdev > 1) { |
f3871fdc AK |
8722 | dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev); |
8723 | return ret_val; | |
8724 | } | |
8725 | ||
8726 | for (id = super->devlist ; id; id = id->next) { | |
8727 | if (id->index == (unsigned)u->subdev) { | |
8728 | struct imsm_dev *dev = get_imsm_dev(super, u->subdev); | |
8729 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
8730 | int used_disks = imsm_num_data_members(dev, MAP_0); | |
8731 | unsigned long long blocks_per_member; | |
06fb291a | 8732 | unsigned long long num_data_stripes; |
f3871fdc AK |
8733 | |
8734 | /* calculate new size | |
8735 | */ | |
8736 | blocks_per_member = u->new_size / used_disks; | |
06fb291a PB |
8737 | num_data_stripes = blocks_per_member / |
8738 | map->blocks_per_strip; | |
8739 | num_data_stripes /= map->num_domains; | |
8740 | dprintf("(size: %llu, blocks per member: %llu, num_data_stipes: %llu)\n", | |
8741 | u->new_size, blocks_per_member, | |
8742 | num_data_stripes); | |
f3871fdc | 8743 | set_blocks_per_member(map, blocks_per_member); |
06fb291a | 8744 | set_num_data_stripes(map, num_data_stripes); |
f3871fdc AK |
8745 | imsm_set_array_size(dev, u->new_size); |
8746 | ||
8747 | ret_val = 1; | |
8748 | break; | |
8749 | } | |
8750 | } | |
8751 | ||
8752 | return ret_val; | |
8753 | } | |
8754 | ||
061d7da3 | 8755 | static int apply_update_activate_spare(struct imsm_update_activate_spare *u, |
ca9de185 | 8756 | struct intel_super *super, |
061d7da3 LO |
8757 | struct active_array *active_array) |
8758 | { | |
8759 | struct imsm_super *mpb = super->anchor; | |
8760 | struct imsm_dev *dev = get_imsm_dev(super, u->array); | |
238c0a71 | 8761 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
061d7da3 LO |
8762 | struct imsm_map *migr_map; |
8763 | struct active_array *a; | |
8764 | struct imsm_disk *disk; | |
8765 | __u8 to_state; | |
8766 | struct dl *dl; | |
8767 | unsigned int found; | |
8768 | int failed; | |
5961eeec | 8769 | int victim; |
061d7da3 | 8770 | int i; |
5961eeec | 8771 | int second_map_created = 0; |
061d7da3 | 8772 | |
5961eeec | 8773 | for (; u; u = u->next) { |
238c0a71 | 8774 | victim = get_imsm_disk_idx(dev, u->slot, MAP_X); |
061d7da3 | 8775 | |
5961eeec | 8776 | if (victim < 0) |
8777 | return 0; | |
061d7da3 | 8778 | |
5961eeec | 8779 | for (dl = super->disks; dl; dl = dl->next) |
8780 | if (dl == u->dl) | |
8781 | break; | |
061d7da3 | 8782 | |
5961eeec | 8783 | if (!dl) { |
7a862a02 | 8784 | pr_err("error: imsm_activate_spare passed an unknown disk (index: %d)\n", |
5961eeec | 8785 | u->dl->index); |
8786 | return 0; | |
8787 | } | |
061d7da3 | 8788 | |
5961eeec | 8789 | /* count failures (excluding rebuilds and the victim) |
8790 | * to determine map[0] state | |
8791 | */ | |
8792 | failed = 0; | |
8793 | for (i = 0; i < map->num_members; i++) { | |
8794 | if (i == u->slot) | |
8795 | continue; | |
8796 | disk = get_imsm_disk(super, | |
238c0a71 | 8797 | get_imsm_disk_idx(dev, i, MAP_X)); |
5961eeec | 8798 | if (!disk || is_failed(disk)) |
8799 | failed++; | |
8800 | } | |
061d7da3 | 8801 | |
5961eeec | 8802 | /* adding a pristine spare, assign a new index */ |
8803 | if (dl->index < 0) { | |
8804 | dl->index = super->anchor->num_disks; | |
8805 | super->anchor->num_disks++; | |
8806 | } | |
8807 | disk = &dl->disk; | |
8808 | disk->status |= CONFIGURED_DISK; | |
8809 | disk->status &= ~SPARE_DISK; | |
8810 | ||
8811 | /* mark rebuild */ | |
238c0a71 | 8812 | to_state = imsm_check_degraded(super, dev, failed, MAP_0); |
5961eeec | 8813 | if (!second_map_created) { |
8814 | second_map_created = 1; | |
8815 | map->map_state = IMSM_T_STATE_DEGRADED; | |
8816 | migrate(dev, super, to_state, MIGR_REBUILD); | |
8817 | } else | |
8818 | map->map_state = to_state; | |
238c0a71 | 8819 | migr_map = get_imsm_map(dev, MAP_1); |
5961eeec | 8820 | set_imsm_ord_tbl_ent(map, u->slot, dl->index); |
8821 | set_imsm_ord_tbl_ent(migr_map, u->slot, | |
8822 | dl->index | IMSM_ORD_REBUILD); | |
8823 | ||
8824 | /* update the family_num to mark a new container | |
8825 | * generation, being careful to record the existing | |
8826 | * family_num in orig_family_num to clean up after | |
8827 | * earlier mdadm versions that neglected to set it. | |
8828 | */ | |
8829 | if (mpb->orig_family_num == 0) | |
8830 | mpb->orig_family_num = mpb->family_num; | |
8831 | mpb->family_num += super->random; | |
8832 | ||
8833 | /* count arrays using the victim in the metadata */ | |
8834 | found = 0; | |
8835 | for (a = active_array; a ; a = a->next) { | |
8836 | dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 8837 | map = get_imsm_map(dev, MAP_0); |
061d7da3 | 8838 | |
5961eeec | 8839 | if (get_imsm_disk_slot(map, victim) >= 0) |
8840 | found++; | |
8841 | } | |
061d7da3 | 8842 | |
5961eeec | 8843 | /* delete the victim if it is no longer being |
8844 | * utilized anywhere | |
061d7da3 | 8845 | */ |
5961eeec | 8846 | if (!found) { |
8847 | struct dl **dlp; | |
061d7da3 | 8848 | |
5961eeec | 8849 | /* We know that 'manager' isn't touching anything, |
8850 | * so it is safe to delete | |
8851 | */ | |
8852 | for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next) | |
061d7da3 LO |
8853 | if ((*dlp)->index == victim) |
8854 | break; | |
5961eeec | 8855 | |
8856 | /* victim may be on the missing list */ | |
8857 | if (!*dlp) | |
8858 | for (dlp = &super->missing; *dlp; | |
8859 | dlp = &(*dlp)->next) | |
8860 | if ((*dlp)->index == victim) | |
8861 | break; | |
8862 | imsm_delete(super, dlp, victim); | |
8863 | } | |
061d7da3 LO |
8864 | } |
8865 | ||
8866 | return 1; | |
8867 | } | |
a29911da | 8868 | |
2e5dc010 N |
8869 | static int apply_reshape_container_disks_update(struct imsm_update_reshape *u, |
8870 | struct intel_super *super, | |
8871 | void ***space_list) | |
8872 | { | |
8873 | struct dl *new_disk; | |
8874 | struct intel_dev *id; | |
8875 | int i; | |
8876 | int delta_disks = u->new_raid_disks - u->old_raid_disks; | |
ee4beede | 8877 | int disk_count = u->old_raid_disks; |
2e5dc010 N |
8878 | void **tofree = NULL; |
8879 | int devices_to_reshape = 1; | |
8880 | struct imsm_super *mpb = super->anchor; | |
8881 | int ret_val = 0; | |
d098291a | 8882 | unsigned int dev_id; |
2e5dc010 | 8883 | |
1ade5cc1 | 8884 | dprintf("(enter)\n"); |
2e5dc010 N |
8885 | |
8886 | /* enable spares to use in array */ | |
8887 | for (i = 0; i < delta_disks; i++) { | |
8888 | new_disk = get_disk_super(super, | |
8889 | major(u->new_disks[i]), | |
8890 | minor(u->new_disks[i])); | |
7a862a02 | 8891 | dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n", |
2e5dc010 N |
8892 | major(u->new_disks[i]), minor(u->new_disks[i]), |
8893 | new_disk, new_disk->index); | |
089f9d79 JS |
8894 | if (new_disk == NULL || |
8895 | (new_disk->index >= 0 && | |
8896 | new_disk->index < u->old_raid_disks)) | |
2e5dc010 | 8897 | goto update_reshape_exit; |
ee4beede | 8898 | new_disk->index = disk_count++; |
2e5dc010 N |
8899 | /* slot to fill in autolayout |
8900 | */ | |
8901 | new_disk->raiddisk = new_disk->index; | |
8902 | new_disk->disk.status |= | |
8903 | CONFIGURED_DISK; | |
8904 | new_disk->disk.status &= ~SPARE_DISK; | |
8905 | } | |
8906 | ||
ed7333bd AK |
8907 | dprintf("imsm: volume set mpb->num_raid_devs = %i\n", |
8908 | mpb->num_raid_devs); | |
2e5dc010 N |
8909 | /* manage changes in volume |
8910 | */ | |
d098291a | 8911 | for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) { |
2e5dc010 N |
8912 | void **sp = *space_list; |
8913 | struct imsm_dev *newdev; | |
8914 | struct imsm_map *newmap, *oldmap; | |
8915 | ||
d098291a AK |
8916 | for (id = super->devlist ; id; id = id->next) { |
8917 | if (id->index == dev_id) | |
8918 | break; | |
8919 | } | |
8920 | if (id == NULL) | |
8921 | break; | |
2e5dc010 N |
8922 | if (!sp) |
8923 | continue; | |
8924 | *space_list = *sp; | |
8925 | newdev = (void*)sp; | |
8926 | /* Copy the dev, but not (all of) the map */ | |
8927 | memcpy(newdev, id->dev, sizeof(*newdev)); | |
238c0a71 AK |
8928 | oldmap = get_imsm_map(id->dev, MAP_0); |
8929 | newmap = get_imsm_map(newdev, MAP_0); | |
2e5dc010 N |
8930 | /* Copy the current map */ |
8931 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); | |
8932 | /* update one device only | |
8933 | */ | |
8934 | if (devices_to_reshape) { | |
ed7333bd AK |
8935 | dprintf("imsm: modifying subdev: %i\n", |
8936 | id->index); | |
2e5dc010 N |
8937 | devices_to_reshape--; |
8938 | newdev->vol.migr_state = 1; | |
8939 | newdev->vol.curr_migr_unit = 0; | |
ea672ee1 | 8940 | set_migr_type(newdev, MIGR_GEN_MIGR); |
2e5dc010 N |
8941 | newmap->num_members = u->new_raid_disks; |
8942 | for (i = 0; i < delta_disks; i++) { | |
8943 | set_imsm_ord_tbl_ent(newmap, | |
8944 | u->old_raid_disks + i, | |
8945 | u->old_raid_disks + i); | |
8946 | } | |
8947 | /* New map is correct, now need to save old map | |
8948 | */ | |
238c0a71 | 8949 | newmap = get_imsm_map(newdev, MAP_1); |
2e5dc010 N |
8950 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); |
8951 | ||
f3871fdc | 8952 | imsm_set_array_size(newdev, -1); |
2e5dc010 N |
8953 | } |
8954 | ||
8955 | sp = (void **)id->dev; | |
8956 | id->dev = newdev; | |
8957 | *sp = tofree; | |
8958 | tofree = sp; | |
8e59f3d8 AK |
8959 | |
8960 | /* Clear migration record */ | |
8961 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
2e5dc010 | 8962 | } |
819bc634 AK |
8963 | if (tofree) |
8964 | *space_list = tofree; | |
2e5dc010 N |
8965 | ret_val = 1; |
8966 | ||
8967 | update_reshape_exit: | |
8968 | ||
8969 | return ret_val; | |
8970 | } | |
8971 | ||
bb025c2f | 8972 | static int apply_takeover_update(struct imsm_update_takeover *u, |
8ca6df95 KW |
8973 | struct intel_super *super, |
8974 | void ***space_list) | |
bb025c2f KW |
8975 | { |
8976 | struct imsm_dev *dev = NULL; | |
8ca6df95 KW |
8977 | struct intel_dev *dv; |
8978 | struct imsm_dev *dev_new; | |
bb025c2f KW |
8979 | struct imsm_map *map; |
8980 | struct dl *dm, *du; | |
8ca6df95 | 8981 | int i; |
bb025c2f KW |
8982 | |
8983 | for (dv = super->devlist; dv; dv = dv->next) | |
8984 | if (dv->index == (unsigned int)u->subarray) { | |
8985 | dev = dv->dev; | |
8986 | break; | |
8987 | } | |
8988 | ||
8989 | if (dev == NULL) | |
8990 | return 0; | |
8991 | ||
238c0a71 | 8992 | map = get_imsm_map(dev, MAP_0); |
bb025c2f KW |
8993 | |
8994 | if (u->direction == R10_TO_R0) { | |
06fb291a PB |
8995 | unsigned long long num_data_stripes; |
8996 | ||
8997 | map->num_domains = 1; | |
8998 | num_data_stripes = blocks_per_member(map); | |
8999 | num_data_stripes /= map->blocks_per_strip; | |
9000 | num_data_stripes /= map->num_domains; | |
9001 | set_num_data_stripes(map, num_data_stripes); | |
9002 | ||
43d5ec18 | 9003 | /* Number of failed disks must be half of initial disk number */ |
3b451610 AK |
9004 | if (imsm_count_failed(super, dev, MAP_0) != |
9005 | (map->num_members / 2)) | |
43d5ec18 KW |
9006 | return 0; |
9007 | ||
bb025c2f KW |
9008 | /* iterate through devices to mark removed disks as spare */ |
9009 | for (dm = super->disks; dm; dm = dm->next) { | |
9010 | if (dm->disk.status & FAILED_DISK) { | |
9011 | int idx = dm->index; | |
9012 | /* update indexes on the disk list */ | |
9013 | /* FIXME this loop-with-the-loop looks wrong, I'm not convinced | |
9014 | the index values will end up being correct.... NB */ | |
9015 | for (du = super->disks; du; du = du->next) | |
9016 | if (du->index > idx) | |
9017 | du->index--; | |
9018 | /* mark as spare disk */ | |
a8619d23 | 9019 | mark_spare(dm); |
bb025c2f KW |
9020 | } |
9021 | } | |
bb025c2f KW |
9022 | /* update map */ |
9023 | map->num_members = map->num_members / 2; | |
9024 | map->map_state = IMSM_T_STATE_NORMAL; | |
9025 | map->num_domains = 1; | |
9026 | map->raid_level = 0; | |
9027 | map->failed_disk_num = -1; | |
9028 | } | |
9029 | ||
8ca6df95 KW |
9030 | if (u->direction == R0_TO_R10) { |
9031 | void **space; | |
9032 | /* update slots in current disk list */ | |
9033 | for (dm = super->disks; dm; dm = dm->next) { | |
9034 | if (dm->index >= 0) | |
9035 | dm->index *= 2; | |
9036 | } | |
9037 | /* create new *missing* disks */ | |
9038 | for (i = 0; i < map->num_members; i++) { | |
9039 | space = *space_list; | |
9040 | if (!space) | |
9041 | continue; | |
9042 | *space_list = *space; | |
9043 | du = (void *)space; | |
9044 | memcpy(du, super->disks, sizeof(*du)); | |
8ca6df95 KW |
9045 | du->fd = -1; |
9046 | du->minor = 0; | |
9047 | du->major = 0; | |
9048 | du->index = (i * 2) + 1; | |
9049 | sprintf((char *)du->disk.serial, | |
9050 | " MISSING_%d", du->index); | |
9051 | sprintf((char *)du->serial, | |
9052 | "MISSING_%d", du->index); | |
9053 | du->next = super->missing; | |
9054 | super->missing = du; | |
9055 | } | |
9056 | /* create new dev and map */ | |
9057 | space = *space_list; | |
9058 | if (!space) | |
9059 | return 0; | |
9060 | *space_list = *space; | |
9061 | dev_new = (void *)space; | |
9062 | memcpy(dev_new, dev, sizeof(*dev)); | |
9063 | /* update new map */ | |
238c0a71 | 9064 | map = get_imsm_map(dev_new, MAP_0); |
8ca6df95 | 9065 | map->num_members = map->num_members * 2; |
1a2487c2 | 9066 | map->map_state = IMSM_T_STATE_DEGRADED; |
8ca6df95 KW |
9067 | map->num_domains = 2; |
9068 | map->raid_level = 1; | |
9069 | /* replace dev<->dev_new */ | |
9070 | dv->dev = dev_new; | |
9071 | } | |
bb025c2f KW |
9072 | /* update disk order table */ |
9073 | for (du = super->disks; du; du = du->next) | |
9074 | if (du->index >= 0) | |
9075 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
8ca6df95 | 9076 | for (du = super->missing; du; du = du->next) |
1a2487c2 KW |
9077 | if (du->index >= 0) { |
9078 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
4c9e8c1e | 9079 | mark_missing(super, dv->dev, &du->disk, du->index); |
1a2487c2 | 9080 | } |
bb025c2f KW |
9081 | |
9082 | return 1; | |
9083 | } | |
9084 | ||
e8319a19 DW |
9085 | static void imsm_process_update(struct supertype *st, |
9086 | struct metadata_update *update) | |
9087 | { | |
9088 | /** | |
9089 | * crack open the metadata_update envelope to find the update record | |
9090 | * update can be one of: | |
d195167d AK |
9091 | * update_reshape_container_disks - all the arrays in the container |
9092 | * are being reshaped to have more devices. We need to mark | |
9093 | * the arrays for general migration and convert selected spares | |
9094 | * into active devices. | |
9095 | * update_activate_spare - a spare device has replaced a failed | |
1011e834 N |
9096 | * device in an array, update the disk_ord_tbl. If this disk is |
9097 | * present in all member arrays then also clear the SPARE_DISK | |
9098 | * flag | |
d195167d AK |
9099 | * update_create_array |
9100 | * update_kill_array | |
9101 | * update_rename_array | |
9102 | * update_add_remove_disk | |
e8319a19 DW |
9103 | */ |
9104 | struct intel_super *super = st->sb; | |
4d7b1503 | 9105 | struct imsm_super *mpb; |
e8319a19 DW |
9106 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
9107 | ||
4d7b1503 DW |
9108 | /* update requires a larger buf but the allocation failed */ |
9109 | if (super->next_len && !super->next_buf) { | |
9110 | super->next_len = 0; | |
9111 | return; | |
9112 | } | |
9113 | ||
9114 | if (super->next_buf) { | |
9115 | memcpy(super->next_buf, super->buf, super->len); | |
9116 | free(super->buf); | |
9117 | super->len = super->next_len; | |
9118 | super->buf = super->next_buf; | |
9119 | ||
9120 | super->next_len = 0; | |
9121 | super->next_buf = NULL; | |
9122 | } | |
9123 | ||
9124 | mpb = super->anchor; | |
9125 | ||
e8319a19 | 9126 | switch (type) { |
0ec5d470 AK |
9127 | case update_general_migration_checkpoint: { |
9128 | struct intel_dev *id; | |
9129 | struct imsm_update_general_migration_checkpoint *u = | |
9130 | (void *)update->buf; | |
9131 | ||
1ade5cc1 | 9132 | dprintf("called for update_general_migration_checkpoint\n"); |
0ec5d470 AK |
9133 | |
9134 | /* find device under general migration */ | |
9135 | for (id = super->devlist ; id; id = id->next) { | |
9136 | if (is_gen_migration(id->dev)) { | |
9137 | id->dev->vol.curr_migr_unit = | |
9138 | __cpu_to_le32(u->curr_migr_unit); | |
9139 | super->updates_pending++; | |
9140 | } | |
9141 | } | |
9142 | break; | |
9143 | } | |
bb025c2f KW |
9144 | case update_takeover: { |
9145 | struct imsm_update_takeover *u = (void *)update->buf; | |
1a2487c2 KW |
9146 | if (apply_takeover_update(u, super, &update->space_list)) { |
9147 | imsm_update_version_info(super); | |
bb025c2f | 9148 | super->updates_pending++; |
1a2487c2 | 9149 | } |
bb025c2f KW |
9150 | break; |
9151 | } | |
9152 | ||
78b10e66 | 9153 | case update_reshape_container_disks: { |
d195167d | 9154 | struct imsm_update_reshape *u = (void *)update->buf; |
2e5dc010 N |
9155 | if (apply_reshape_container_disks_update( |
9156 | u, super, &update->space_list)) | |
9157 | super->updates_pending++; | |
78b10e66 N |
9158 | break; |
9159 | } | |
48c5303a | 9160 | case update_reshape_migration: { |
a29911da PC |
9161 | struct imsm_update_reshape_migration *u = (void *)update->buf; |
9162 | if (apply_reshape_migration_update( | |
9163 | u, super, &update->space_list)) | |
9164 | super->updates_pending++; | |
48c5303a PC |
9165 | break; |
9166 | } | |
f3871fdc AK |
9167 | case update_size_change: { |
9168 | struct imsm_update_size_change *u = (void *)update->buf; | |
9169 | if (apply_size_change_update(u, super)) | |
9170 | super->updates_pending++; | |
9171 | break; | |
9172 | } | |
e8319a19 | 9173 | case update_activate_spare: { |
1011e834 | 9174 | struct imsm_update_activate_spare *u = (void *) update->buf; |
061d7da3 LO |
9175 | if (apply_update_activate_spare(u, super, st->arrays)) |
9176 | super->updates_pending++; | |
8273f55e DW |
9177 | break; |
9178 | } | |
9179 | case update_create_array: { | |
9180 | /* someone wants to create a new array, we need to be aware of | |
9181 | * a few races/collisions: | |
9182 | * 1/ 'Create' called by two separate instances of mdadm | |
9183 | * 2/ 'Create' versus 'activate_spare': mdadm has chosen | |
9184 | * devices that have since been assimilated via | |
9185 | * activate_spare. | |
9186 | * In the event this update can not be carried out mdadm will | |
9187 | * (FIX ME) notice that its update did not take hold. | |
9188 | */ | |
9189 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 9190 | struct intel_dev *dv; |
8273f55e DW |
9191 | struct imsm_dev *dev; |
9192 | struct imsm_map *map, *new_map; | |
9193 | unsigned long long start, end; | |
9194 | unsigned long long new_start, new_end; | |
9195 | int i; | |
54c2c1ea DW |
9196 | struct disk_info *inf; |
9197 | struct dl *dl; | |
8273f55e DW |
9198 | |
9199 | /* handle racing creates: first come first serve */ | |
9200 | if (u->dev_idx < mpb->num_raid_devs) { | |
1ade5cc1 | 9201 | dprintf("subarray %d already defined\n", u->dev_idx); |
ba2de7ba | 9202 | goto create_error; |
8273f55e DW |
9203 | } |
9204 | ||
9205 | /* check update is next in sequence */ | |
9206 | if (u->dev_idx != mpb->num_raid_devs) { | |
1ade5cc1 N |
9207 | dprintf("can not create array %d expected index %d\n", |
9208 | u->dev_idx, mpb->num_raid_devs); | |
ba2de7ba | 9209 | goto create_error; |
8273f55e DW |
9210 | } |
9211 | ||
238c0a71 | 9212 | new_map = get_imsm_map(&u->dev, MAP_0); |
5551b113 CA |
9213 | new_start = pba_of_lba0(new_map); |
9214 | new_end = new_start + blocks_per_member(new_map); | |
54c2c1ea | 9215 | inf = get_disk_info(u); |
8273f55e DW |
9216 | |
9217 | /* handle activate_spare versus create race: | |
9218 | * check to make sure that overlapping arrays do not include | |
9219 | * overalpping disks | |
9220 | */ | |
9221 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
949c47a0 | 9222 | dev = get_imsm_dev(super, i); |
238c0a71 | 9223 | map = get_imsm_map(dev, MAP_0); |
5551b113 CA |
9224 | start = pba_of_lba0(map); |
9225 | end = start + blocks_per_member(map); | |
8273f55e DW |
9226 | if ((new_start >= start && new_start <= end) || |
9227 | (start >= new_start && start <= new_end)) | |
54c2c1ea DW |
9228 | /* overlap */; |
9229 | else | |
9230 | continue; | |
9231 | ||
9232 | if (disks_overlap(super, i, u)) { | |
1ade5cc1 | 9233 | dprintf("arrays overlap\n"); |
ba2de7ba | 9234 | goto create_error; |
8273f55e DW |
9235 | } |
9236 | } | |
8273f55e | 9237 | |
949c47a0 DW |
9238 | /* check that prepare update was successful */ |
9239 | if (!update->space) { | |
1ade5cc1 | 9240 | dprintf("prepare update failed\n"); |
ba2de7ba | 9241 | goto create_error; |
949c47a0 DW |
9242 | } |
9243 | ||
54c2c1ea DW |
9244 | /* check that all disks are still active before committing |
9245 | * changes. FIXME: could we instead handle this by creating a | |
9246 | * degraded array? That's probably not what the user expects, | |
9247 | * so better to drop this update on the floor. | |
9248 | */ | |
9249 | for (i = 0; i < new_map->num_members; i++) { | |
9250 | dl = serial_to_dl(inf[i].serial, super); | |
9251 | if (!dl) { | |
1ade5cc1 | 9252 | dprintf("disk disappeared\n"); |
ba2de7ba | 9253 | goto create_error; |
54c2c1ea | 9254 | } |
949c47a0 DW |
9255 | } |
9256 | ||
8273f55e | 9257 | super->updates_pending++; |
54c2c1ea DW |
9258 | |
9259 | /* convert spares to members and fixup ord_tbl */ | |
9260 | for (i = 0; i < new_map->num_members; i++) { | |
9261 | dl = serial_to_dl(inf[i].serial, super); | |
9262 | if (dl->index == -1) { | |
9263 | dl->index = mpb->num_disks; | |
9264 | mpb->num_disks++; | |
9265 | dl->disk.status |= CONFIGURED_DISK; | |
9266 | dl->disk.status &= ~SPARE_DISK; | |
9267 | } | |
9268 | set_imsm_ord_tbl_ent(new_map, i, dl->index); | |
9269 | } | |
9270 | ||
ba2de7ba DW |
9271 | dv = update->space; |
9272 | dev = dv->dev; | |
949c47a0 DW |
9273 | update->space = NULL; |
9274 | imsm_copy_dev(dev, &u->dev); | |
ba2de7ba DW |
9275 | dv->index = u->dev_idx; |
9276 | dv->next = super->devlist; | |
9277 | super->devlist = dv; | |
8273f55e | 9278 | mpb->num_raid_devs++; |
8273f55e | 9279 | |
4d1313e9 | 9280 | imsm_update_version_info(super); |
8273f55e | 9281 | break; |
ba2de7ba DW |
9282 | create_error: |
9283 | /* mdmon knows how to release update->space, but not | |
9284 | * ((struct intel_dev *) update->space)->dev | |
9285 | */ | |
9286 | if (update->space) { | |
9287 | dv = update->space; | |
9288 | free(dv->dev); | |
9289 | } | |
8273f55e | 9290 | break; |
e8319a19 | 9291 | } |
33414a01 DW |
9292 | case update_kill_array: { |
9293 | struct imsm_update_kill_array *u = (void *) update->buf; | |
9294 | int victim = u->dev_idx; | |
9295 | struct active_array *a; | |
9296 | struct intel_dev **dp; | |
9297 | struct imsm_dev *dev; | |
9298 | ||
9299 | /* sanity check that we are not affecting the uuid of | |
9300 | * active arrays, or deleting an active array | |
9301 | * | |
9302 | * FIXME when immutable ids are available, but note that | |
9303 | * we'll also need to fixup the invalidated/active | |
9304 | * subarray indexes in mdstat | |
9305 | */ | |
9306 | for (a = st->arrays; a; a = a->next) | |
9307 | if (a->info.container_member >= victim) | |
9308 | break; | |
9309 | /* by definition if mdmon is running at least one array | |
9310 | * is active in the container, so checking | |
9311 | * mpb->num_raid_devs is just extra paranoia | |
9312 | */ | |
9313 | dev = get_imsm_dev(super, victim); | |
9314 | if (a || !dev || mpb->num_raid_devs == 1) { | |
9315 | dprintf("failed to delete subarray-%d\n", victim); | |
9316 | break; | |
9317 | } | |
9318 | ||
9319 | for (dp = &super->devlist; *dp;) | |
f21e18ca | 9320 | if ((*dp)->index == (unsigned)super->current_vol) { |
33414a01 DW |
9321 | *dp = (*dp)->next; |
9322 | } else { | |
f21e18ca | 9323 | if ((*dp)->index > (unsigned)victim) |
33414a01 DW |
9324 | (*dp)->index--; |
9325 | dp = &(*dp)->next; | |
9326 | } | |
9327 | mpb->num_raid_devs--; | |
9328 | super->updates_pending++; | |
9329 | break; | |
9330 | } | |
aa534678 DW |
9331 | case update_rename_array: { |
9332 | struct imsm_update_rename_array *u = (void *) update->buf; | |
9333 | char name[MAX_RAID_SERIAL_LEN+1]; | |
9334 | int target = u->dev_idx; | |
9335 | struct active_array *a; | |
9336 | struct imsm_dev *dev; | |
9337 | ||
9338 | /* sanity check that we are not affecting the uuid of | |
9339 | * an active array | |
9340 | */ | |
9341 | snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name); | |
9342 | name[MAX_RAID_SERIAL_LEN] = '\0'; | |
9343 | for (a = st->arrays; a; a = a->next) | |
9344 | if (a->info.container_member == target) | |
9345 | break; | |
9346 | dev = get_imsm_dev(super, u->dev_idx); | |
9347 | if (a || !dev || !check_name(super, name, 1)) { | |
9348 | dprintf("failed to rename subarray-%d\n", target); | |
9349 | break; | |
9350 | } | |
9351 | ||
cdbe98cd | 9352 | snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name); |
aa534678 DW |
9353 | super->updates_pending++; |
9354 | break; | |
9355 | } | |
1a64be56 | 9356 | case update_add_remove_disk: { |
43dad3d6 | 9357 | /* we may be able to repair some arrays if disks are |
095b8088 | 9358 | * being added, check the status of add_remove_disk |
1a64be56 LM |
9359 | * if discs has been added. |
9360 | */ | |
9361 | if (add_remove_disk_update(super)) { | |
43dad3d6 | 9362 | struct active_array *a; |
072b727f DW |
9363 | |
9364 | super->updates_pending++; | |
1a64be56 | 9365 | for (a = st->arrays; a; a = a->next) |
43dad3d6 DW |
9366 | a->check_degraded = 1; |
9367 | } | |
43dad3d6 | 9368 | break; |
e8319a19 | 9369 | } |
bbab0940 TM |
9370 | case update_prealloc_badblocks_mem: |
9371 | break; | |
1a64be56 | 9372 | default: |
7a862a02 | 9373 | pr_err("error: unsuported process update type:(type: %d)\n", type); |
1a64be56 | 9374 | } |
e8319a19 | 9375 | } |
88758e9d | 9376 | |
bc0b9d34 PC |
9377 | static struct mdinfo *get_spares_for_grow(struct supertype *st); |
9378 | ||
5fe6f031 N |
9379 | static int imsm_prepare_update(struct supertype *st, |
9380 | struct metadata_update *update) | |
8273f55e | 9381 | { |
949c47a0 | 9382 | /** |
4d7b1503 DW |
9383 | * Allocate space to hold new disk entries, raid-device entries or a new |
9384 | * mpb if necessary. The manager synchronously waits for updates to | |
9385 | * complete in the monitor, so new mpb buffers allocated here can be | |
9386 | * integrated by the monitor thread without worrying about live pointers | |
9387 | * in the manager thread. | |
8273f55e | 9388 | */ |
095b8088 | 9389 | enum imsm_update_type type; |
4d7b1503 | 9390 | struct intel_super *super = st->sb; |
f36a9ecd | 9391 | unsigned int sector_size = super->sector_size; |
4d7b1503 DW |
9392 | struct imsm_super *mpb = super->anchor; |
9393 | size_t buf_len; | |
9394 | size_t len = 0; | |
949c47a0 | 9395 | |
095b8088 N |
9396 | if (update->len < (int)sizeof(type)) |
9397 | return 0; | |
9398 | ||
9399 | type = *(enum imsm_update_type *) update->buf; | |
9400 | ||
949c47a0 | 9401 | switch (type) { |
0ec5d470 | 9402 | case update_general_migration_checkpoint: |
095b8088 N |
9403 | if (update->len < (int)sizeof(struct imsm_update_general_migration_checkpoint)) |
9404 | return 0; | |
1ade5cc1 | 9405 | dprintf("called for update_general_migration_checkpoint\n"); |
0ec5d470 | 9406 | break; |
abedf5fc KW |
9407 | case update_takeover: { |
9408 | struct imsm_update_takeover *u = (void *)update->buf; | |
095b8088 N |
9409 | if (update->len < (int)sizeof(*u)) |
9410 | return 0; | |
abedf5fc KW |
9411 | if (u->direction == R0_TO_R10) { |
9412 | void **tail = (void **)&update->space_list; | |
9413 | struct imsm_dev *dev = get_imsm_dev(super, u->subarray); | |
238c0a71 | 9414 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
abedf5fc KW |
9415 | int num_members = map->num_members; |
9416 | void *space; | |
9417 | int size, i; | |
abedf5fc KW |
9418 | /* allocate memory for added disks */ |
9419 | for (i = 0; i < num_members; i++) { | |
9420 | size = sizeof(struct dl); | |
503975b9 | 9421 | space = xmalloc(size); |
abedf5fc KW |
9422 | *tail = space; |
9423 | tail = space; | |
9424 | *tail = NULL; | |
9425 | } | |
9426 | /* allocate memory for new device */ | |
9427 | size = sizeof_imsm_dev(super->devlist->dev, 0) + | |
9428 | (num_members * sizeof(__u32)); | |
503975b9 N |
9429 | space = xmalloc(size); |
9430 | *tail = space; | |
9431 | tail = space; | |
9432 | *tail = NULL; | |
9433 | len = disks_to_mpb_size(num_members * 2); | |
abedf5fc KW |
9434 | } |
9435 | ||
9436 | break; | |
9437 | } | |
78b10e66 | 9438 | case update_reshape_container_disks: { |
d195167d AK |
9439 | /* Every raid device in the container is about to |
9440 | * gain some more devices, and we will enter a | |
9441 | * reconfiguration. | |
9442 | * So each 'imsm_map' will be bigger, and the imsm_vol | |
9443 | * will now hold 2 of them. | |
9444 | * Thus we need new 'struct imsm_dev' allocations sized | |
9445 | * as sizeof_imsm_dev but with more devices in both maps. | |
9446 | */ | |
9447 | struct imsm_update_reshape *u = (void *)update->buf; | |
9448 | struct intel_dev *dl; | |
9449 | void **space_tail = (void**)&update->space_list; | |
9450 | ||
095b8088 N |
9451 | if (update->len < (int)sizeof(*u)) |
9452 | return 0; | |
9453 | ||
1ade5cc1 | 9454 | dprintf("for update_reshape\n"); |
d195167d AK |
9455 | |
9456 | for (dl = super->devlist; dl; dl = dl->next) { | |
9457 | int size = sizeof_imsm_dev(dl->dev, 1); | |
9458 | void *s; | |
d677e0b8 AK |
9459 | if (u->new_raid_disks > u->old_raid_disks) |
9460 | size += sizeof(__u32)*2* | |
9461 | (u->new_raid_disks - u->old_raid_disks); | |
503975b9 | 9462 | s = xmalloc(size); |
d195167d AK |
9463 | *space_tail = s; |
9464 | space_tail = s; | |
9465 | *space_tail = NULL; | |
9466 | } | |
9467 | ||
9468 | len = disks_to_mpb_size(u->new_raid_disks); | |
9469 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
78b10e66 N |
9470 | break; |
9471 | } | |
48c5303a | 9472 | case update_reshape_migration: { |
bc0b9d34 PC |
9473 | /* for migration level 0->5 we need to add disks |
9474 | * so the same as for container operation we will copy | |
9475 | * device to the bigger location. | |
9476 | * in memory prepared device and new disk area are prepared | |
9477 | * for usage in process update | |
9478 | */ | |
9479 | struct imsm_update_reshape_migration *u = (void *)update->buf; | |
9480 | struct intel_dev *id; | |
9481 | void **space_tail = (void **)&update->space_list; | |
9482 | int size; | |
9483 | void *s; | |
9484 | int current_level = -1; | |
9485 | ||
095b8088 N |
9486 | if (update->len < (int)sizeof(*u)) |
9487 | return 0; | |
9488 | ||
1ade5cc1 | 9489 | dprintf("for update_reshape\n"); |
bc0b9d34 PC |
9490 | |
9491 | /* add space for bigger array in update | |
9492 | */ | |
9493 | for (id = super->devlist; id; id = id->next) { | |
9494 | if (id->index == (unsigned)u->subdev) { | |
9495 | size = sizeof_imsm_dev(id->dev, 1); | |
9496 | if (u->new_raid_disks > u->old_raid_disks) | |
9497 | size += sizeof(__u32)*2* | |
9498 | (u->new_raid_disks - u->old_raid_disks); | |
503975b9 | 9499 | s = xmalloc(size); |
bc0b9d34 PC |
9500 | *space_tail = s; |
9501 | space_tail = s; | |
9502 | *space_tail = NULL; | |
9503 | break; | |
9504 | } | |
9505 | } | |
9506 | if (update->space_list == NULL) | |
9507 | break; | |
9508 | ||
9509 | /* add space for disk in update | |
9510 | */ | |
9511 | size = sizeof(struct dl); | |
503975b9 | 9512 | s = xmalloc(size); |
bc0b9d34 PC |
9513 | *space_tail = s; |
9514 | space_tail = s; | |
9515 | *space_tail = NULL; | |
9516 | ||
9517 | /* add spare device to update | |
9518 | */ | |
9519 | for (id = super->devlist ; id; id = id->next) | |
9520 | if (id->index == (unsigned)u->subdev) { | |
9521 | struct imsm_dev *dev; | |
9522 | struct imsm_map *map; | |
9523 | ||
9524 | dev = get_imsm_dev(super, u->subdev); | |
238c0a71 | 9525 | map = get_imsm_map(dev, MAP_0); |
bc0b9d34 PC |
9526 | current_level = map->raid_level; |
9527 | break; | |
9528 | } | |
089f9d79 | 9529 | if (u->new_level == 5 && u->new_level != current_level) { |
bc0b9d34 PC |
9530 | struct mdinfo *spares; |
9531 | ||
9532 | spares = get_spares_for_grow(st); | |
9533 | if (spares) { | |
9534 | struct dl *dl; | |
9535 | struct mdinfo *dev; | |
9536 | ||
9537 | dev = spares->devs; | |
9538 | if (dev) { | |
9539 | u->new_disks[0] = | |
9540 | makedev(dev->disk.major, | |
9541 | dev->disk.minor); | |
9542 | dl = get_disk_super(super, | |
9543 | dev->disk.major, | |
9544 | dev->disk.minor); | |
9545 | dl->index = u->old_raid_disks; | |
9546 | dev = dev->next; | |
9547 | } | |
9548 | sysfs_free(spares); | |
9549 | } | |
9550 | } | |
9551 | len = disks_to_mpb_size(u->new_raid_disks); | |
9552 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
48c5303a PC |
9553 | break; |
9554 | } | |
f3871fdc | 9555 | case update_size_change: { |
095b8088 N |
9556 | if (update->len < (int)sizeof(struct imsm_update_size_change)) |
9557 | return 0; | |
9558 | break; | |
9559 | } | |
9560 | case update_activate_spare: { | |
9561 | if (update->len < (int)sizeof(struct imsm_update_activate_spare)) | |
9562 | return 0; | |
f3871fdc AK |
9563 | break; |
9564 | } | |
949c47a0 DW |
9565 | case update_create_array: { |
9566 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 9567 | struct intel_dev *dv; |
54c2c1ea | 9568 | struct imsm_dev *dev = &u->dev; |
238c0a71 | 9569 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
54c2c1ea DW |
9570 | struct dl *dl; |
9571 | struct disk_info *inf; | |
9572 | int i; | |
9573 | int activate = 0; | |
949c47a0 | 9574 | |
095b8088 N |
9575 | if (update->len < (int)sizeof(*u)) |
9576 | return 0; | |
9577 | ||
54c2c1ea DW |
9578 | inf = get_disk_info(u); |
9579 | len = sizeof_imsm_dev(dev, 1); | |
ba2de7ba | 9580 | /* allocate a new super->devlist entry */ |
503975b9 N |
9581 | dv = xmalloc(sizeof(*dv)); |
9582 | dv->dev = xmalloc(len); | |
9583 | update->space = dv; | |
949c47a0 | 9584 | |
54c2c1ea DW |
9585 | /* count how many spares will be converted to members */ |
9586 | for (i = 0; i < map->num_members; i++) { | |
9587 | dl = serial_to_dl(inf[i].serial, super); | |
9588 | if (!dl) { | |
9589 | /* hmm maybe it failed?, nothing we can do about | |
9590 | * it here | |
9591 | */ | |
9592 | continue; | |
9593 | } | |
9594 | if (count_memberships(dl, super) == 0) | |
9595 | activate++; | |
9596 | } | |
9597 | len += activate * sizeof(struct imsm_disk); | |
949c47a0 | 9598 | break; |
095b8088 N |
9599 | } |
9600 | case update_kill_array: { | |
9601 | if (update->len < (int)sizeof(struct imsm_update_kill_array)) | |
9602 | return 0; | |
949c47a0 DW |
9603 | break; |
9604 | } | |
095b8088 N |
9605 | case update_rename_array: { |
9606 | if (update->len < (int)sizeof(struct imsm_update_rename_array)) | |
9607 | return 0; | |
9608 | break; | |
9609 | } | |
9610 | case update_add_remove_disk: | |
9611 | /* no update->len needed */ | |
9612 | break; | |
bbab0940 TM |
9613 | case update_prealloc_badblocks_mem: |
9614 | super->extra_space += sizeof(struct bbm_log) - | |
9615 | get_imsm_bbm_log_size(super->bbm_log); | |
9616 | break; | |
095b8088 N |
9617 | default: |
9618 | return 0; | |
949c47a0 | 9619 | } |
8273f55e | 9620 | |
4d7b1503 DW |
9621 | /* check if we need a larger metadata buffer */ |
9622 | if (super->next_buf) | |
9623 | buf_len = super->next_len; | |
9624 | else | |
9625 | buf_len = super->len; | |
9626 | ||
bbab0940 | 9627 | if (__le32_to_cpu(mpb->mpb_size) + super->extra_space + len > buf_len) { |
4d7b1503 DW |
9628 | /* ok we need a larger buf than what is currently allocated |
9629 | * if this allocation fails process_update will notice that | |
9630 | * ->next_len is set and ->next_buf is NULL | |
9631 | */ | |
bbab0940 TM |
9632 | buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + |
9633 | super->extra_space + len, sector_size); | |
4d7b1503 DW |
9634 | if (super->next_buf) |
9635 | free(super->next_buf); | |
9636 | ||
9637 | super->next_len = buf_len; | |
f36a9ecd | 9638 | if (posix_memalign(&super->next_buf, sector_size, buf_len) == 0) |
1f45a8ad DW |
9639 | memset(super->next_buf, 0, buf_len); |
9640 | else | |
4d7b1503 DW |
9641 | super->next_buf = NULL; |
9642 | } | |
5fe6f031 | 9643 | return 1; |
8273f55e DW |
9644 | } |
9645 | ||
ae6aad82 | 9646 | /* must be called while manager is quiesced */ |
f21e18ca | 9647 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index) |
ae6aad82 DW |
9648 | { |
9649 | struct imsm_super *mpb = super->anchor; | |
ae6aad82 DW |
9650 | struct dl *iter; |
9651 | struct imsm_dev *dev; | |
9652 | struct imsm_map *map; | |
4c9e8c1e | 9653 | unsigned int i, j, num_members; |
24565c9a | 9654 | __u32 ord; |
4c9e8c1e | 9655 | struct bbm_log *log = super->bbm_log; |
ae6aad82 | 9656 | |
1ade5cc1 | 9657 | dprintf("deleting device[%d] from imsm_super\n", index); |
ae6aad82 DW |
9658 | |
9659 | /* shift all indexes down one */ | |
9660 | for (iter = super->disks; iter; iter = iter->next) | |
f21e18ca | 9661 | if (iter->index > (int)index) |
ae6aad82 | 9662 | iter->index--; |
47ee5a45 | 9663 | for (iter = super->missing; iter; iter = iter->next) |
f21e18ca | 9664 | if (iter->index > (int)index) |
47ee5a45 | 9665 | iter->index--; |
ae6aad82 DW |
9666 | |
9667 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
9668 | dev = get_imsm_dev(super, i); | |
238c0a71 | 9669 | map = get_imsm_map(dev, MAP_0); |
24565c9a DW |
9670 | num_members = map->num_members; |
9671 | for (j = 0; j < num_members; j++) { | |
9672 | /* update ord entries being careful not to propagate | |
9673 | * ord-flags to the first map | |
9674 | */ | |
238c0a71 | 9675 | ord = get_imsm_ord_tbl_ent(dev, j, MAP_X); |
ae6aad82 | 9676 | |
24565c9a DW |
9677 | if (ord_to_idx(ord) <= index) |
9678 | continue; | |
ae6aad82 | 9679 | |
238c0a71 | 9680 | map = get_imsm_map(dev, MAP_0); |
24565c9a | 9681 | set_imsm_ord_tbl_ent(map, j, ord_to_idx(ord - 1)); |
238c0a71 | 9682 | map = get_imsm_map(dev, MAP_1); |
24565c9a DW |
9683 | if (map) |
9684 | set_imsm_ord_tbl_ent(map, j, ord - 1); | |
ae6aad82 DW |
9685 | } |
9686 | } | |
9687 | ||
4c9e8c1e TM |
9688 | for (i = 0; i < log->entry_count; i++) { |
9689 | struct bbm_log_entry *entry = &log->marked_block_entries[i]; | |
9690 | ||
9691 | if (entry->disk_ordinal <= index) | |
9692 | continue; | |
9693 | entry->disk_ordinal--; | |
9694 | } | |
9695 | ||
ae6aad82 DW |
9696 | mpb->num_disks--; |
9697 | super->updates_pending++; | |
24565c9a DW |
9698 | if (*dlp) { |
9699 | struct dl *dl = *dlp; | |
9700 | ||
9701 | *dlp = (*dlp)->next; | |
9702 | __free_imsm_disk(dl); | |
9703 | } | |
ae6aad82 | 9704 | } |
9e2d750d | 9705 | #endif /* MDASSEMBLE */ |
9a717282 AK |
9706 | |
9707 | static void close_targets(int *targets, int new_disks) | |
9708 | { | |
9709 | int i; | |
9710 | ||
9711 | if (!targets) | |
9712 | return; | |
9713 | ||
9714 | for (i = 0; i < new_disks; i++) { | |
9715 | if (targets[i] >= 0) { | |
9716 | close(targets[i]); | |
9717 | targets[i] = -1; | |
9718 | } | |
9719 | } | |
9720 | } | |
9721 | ||
9722 | static int imsm_get_allowed_degradation(int level, int raid_disks, | |
9723 | struct intel_super *super, | |
9724 | struct imsm_dev *dev) | |
9725 | { | |
9726 | switch (level) { | |
bf5cf7c7 | 9727 | case 1: |
9a717282 AK |
9728 | case 10:{ |
9729 | int ret_val = 0; | |
9730 | struct imsm_map *map; | |
9731 | int i; | |
9732 | ||
9733 | ret_val = raid_disks/2; | |
9734 | /* check map if all disks pairs not failed | |
9735 | * in both maps | |
9736 | */ | |
238c0a71 | 9737 | map = get_imsm_map(dev, MAP_0); |
9a717282 AK |
9738 | for (i = 0; i < ret_val; i++) { |
9739 | int degradation = 0; | |
9740 | if (get_imsm_disk(super, i) == NULL) | |
9741 | degradation++; | |
9742 | if (get_imsm_disk(super, i + 1) == NULL) | |
9743 | degradation++; | |
9744 | if (degradation == 2) | |
9745 | return 0; | |
9746 | } | |
238c0a71 | 9747 | map = get_imsm_map(dev, MAP_1); |
9a717282 AK |
9748 | /* if there is no second map |
9749 | * result can be returned | |
9750 | */ | |
9751 | if (map == NULL) | |
9752 | return ret_val; | |
9753 | /* check degradation in second map | |
9754 | */ | |
9755 | for (i = 0; i < ret_val; i++) { | |
9756 | int degradation = 0; | |
9757 | if (get_imsm_disk(super, i) == NULL) | |
9758 | degradation++; | |
9759 | if (get_imsm_disk(super, i + 1) == NULL) | |
9760 | degradation++; | |
9761 | if (degradation == 2) | |
9762 | return 0; | |
9763 | } | |
9764 | return ret_val; | |
9765 | } | |
9766 | case 5: | |
9767 | return 1; | |
9768 | case 6: | |
9769 | return 2; | |
9770 | default: | |
9771 | return 0; | |
9772 | } | |
9773 | } | |
9774 | ||
687629c2 AK |
9775 | /******************************************************************************* |
9776 | * Function: open_backup_targets | |
9777 | * Description: Function opens file descriptors for all devices given in | |
9778 | * info->devs | |
9779 | * Parameters: | |
9780 | * info : general array info | |
9781 | * raid_disks : number of disks | |
9782 | * raid_fds : table of device's file descriptors | |
9a717282 AK |
9783 | * super : intel super for raid10 degradation check |
9784 | * dev : intel device for raid10 degradation check | |
687629c2 AK |
9785 | * Returns: |
9786 | * 0 : success | |
9787 | * -1 : fail | |
9788 | ******************************************************************************/ | |
9a717282 AK |
9789 | int open_backup_targets(struct mdinfo *info, int raid_disks, int *raid_fds, |
9790 | struct intel_super *super, struct imsm_dev *dev) | |
687629c2 AK |
9791 | { |
9792 | struct mdinfo *sd; | |
f627f5ad | 9793 | int i; |
9a717282 | 9794 | int opened = 0; |
f627f5ad AK |
9795 | |
9796 | for (i = 0; i < raid_disks; i++) | |
9797 | raid_fds[i] = -1; | |
687629c2 AK |
9798 | |
9799 | for (sd = info->devs ; sd ; sd = sd->next) { | |
9800 | char *dn; | |
9801 | ||
9802 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) { | |
9803 | dprintf("disk is faulty!!\n"); | |
9804 | continue; | |
9805 | } | |
9806 | ||
089f9d79 | 9807 | if (sd->disk.raid_disk >= raid_disks || sd->disk.raid_disk < 0) |
687629c2 AK |
9808 | continue; |
9809 | ||
9810 | dn = map_dev(sd->disk.major, | |
9811 | sd->disk.minor, 1); | |
9812 | raid_fds[sd->disk.raid_disk] = dev_open(dn, O_RDWR); | |
9813 | if (raid_fds[sd->disk.raid_disk] < 0) { | |
e12b3daa | 9814 | pr_err("cannot open component\n"); |
9a717282 | 9815 | continue; |
687629c2 | 9816 | } |
9a717282 AK |
9817 | opened++; |
9818 | } | |
9819 | /* check if maximum array degradation level is not exceeded | |
9820 | */ | |
9821 | if ((raid_disks - opened) > | |
089f9d79 JS |
9822 | imsm_get_allowed_degradation(info->new_level, raid_disks, |
9823 | super, dev)) { | |
e12b3daa | 9824 | pr_err("Not enough disks can be opened.\n"); |
9a717282 AK |
9825 | close_targets(raid_fds, raid_disks); |
9826 | return -2; | |
687629c2 AK |
9827 | } |
9828 | return 0; | |
9829 | } | |
9830 | ||
d31ad643 PB |
9831 | /******************************************************************************* |
9832 | * Function: validate_container_imsm | |
9833 | * Description: This routine validates container after assemble, | |
9834 | * eg. if devices in container are under the same controller. | |
9835 | * | |
9836 | * Parameters: | |
9837 | * info : linked list with info about devices used in array | |
9838 | * Returns: | |
9839 | * 1 : HBA mismatch | |
9840 | * 0 : Success | |
9841 | ******************************************************************************/ | |
9842 | int validate_container_imsm(struct mdinfo *info) | |
9843 | { | |
6b781d33 AP |
9844 | if (check_env("IMSM_NO_PLATFORM")) |
9845 | return 0; | |
d31ad643 | 9846 | |
6b781d33 AP |
9847 | struct sys_dev *idev; |
9848 | struct sys_dev *hba = NULL; | |
9849 | struct sys_dev *intel_devices = find_intel_devices(); | |
9850 | char *dev_path = devt_to_devpath(makedev(info->disk.major, | |
9851 | info->disk.minor)); | |
9852 | ||
9853 | for (idev = intel_devices; idev; idev = idev->next) { | |
9854 | if (dev_path && strstr(dev_path, idev->path)) { | |
9855 | hba = idev; | |
9856 | break; | |
d31ad643 | 9857 | } |
6b781d33 AP |
9858 | } |
9859 | if (dev_path) | |
d31ad643 PB |
9860 | free(dev_path); |
9861 | ||
6b781d33 AP |
9862 | if (!hba) { |
9863 | pr_err("WARNING - Cannot detect HBA for device %s!\n", | |
9864 | devid2kname(makedev(info->disk.major, info->disk.minor))); | |
9865 | return 1; | |
9866 | } | |
9867 | ||
9868 | const struct imsm_orom *orom = get_orom_by_device_id(hba->dev_id); | |
9869 | struct mdinfo *dev; | |
9870 | ||
9871 | for (dev = info->next; dev; dev = dev->next) { | |
9872 | dev_path = devt_to_devpath(makedev(dev->disk.major, dev->disk.minor)); | |
9873 | ||
9874 | struct sys_dev *hba2 = NULL; | |
9875 | for (idev = intel_devices; idev; idev = idev->next) { | |
9876 | if (dev_path && strstr(dev_path, idev->path)) { | |
9877 | hba2 = idev; | |
9878 | break; | |
d31ad643 PB |
9879 | } |
9880 | } | |
6b781d33 AP |
9881 | if (dev_path) |
9882 | free(dev_path); | |
9883 | ||
9884 | const struct imsm_orom *orom2 = hba2 == NULL ? NULL : | |
9885 | get_orom_by_device_id(hba2->dev_id); | |
9886 | ||
9887 | if (hba2 && hba->type != hba2->type) { | |
9888 | pr_err("WARNING - HBAs of devices do not match %s != %s\n", | |
9889 | get_sys_dev_type(hba->type), get_sys_dev_type(hba2->type)); | |
9890 | return 1; | |
9891 | } | |
9892 | ||
07cb1e57 | 9893 | if (orom != orom2) { |
6b781d33 AP |
9894 | pr_err("WARNING - IMSM container assembled with disks under different HBAs!\n" |
9895 | " This operation is not supported and can lead to data loss.\n"); | |
9896 | return 1; | |
9897 | } | |
9898 | ||
9899 | if (!orom) { | |
9900 | pr_err("WARNING - IMSM container assembled with disks under HBAs without IMSM platform support!\n" | |
9901 | " This operation is not supported and can lead to data loss.\n"); | |
9902 | return 1; | |
9903 | } | |
d31ad643 | 9904 | } |
6b781d33 | 9905 | |
d31ad643 PB |
9906 | return 0; |
9907 | } | |
9e2d750d | 9908 | #ifndef MDASSEMBLE |
6f50473f TM |
9909 | /******************************************************************************* |
9910 | * Function: imsm_record_badblock | |
9911 | * Description: This routine stores new bad block record in BBM log | |
9912 | * | |
9913 | * Parameters: | |
9914 | * a : array containing a bad block | |
9915 | * slot : disk number containing a bad block | |
9916 | * sector : bad block sector | |
9917 | * length : bad block sectors range | |
9918 | * Returns: | |
9919 | * 1 : Success | |
9920 | * 0 : Error | |
9921 | ******************************************************************************/ | |
9922 | static int imsm_record_badblock(struct active_array *a, int slot, | |
9923 | unsigned long long sector, int length) | |
9924 | { | |
9925 | struct intel_super *super = a->container->sb; | |
9926 | int ord; | |
9927 | int ret; | |
9928 | ||
9929 | ord = imsm_disk_slot_to_ord(a, slot); | |
9930 | if (ord < 0) | |
9931 | return 0; | |
9932 | ||
9933 | ret = record_new_badblock(super->bbm_log, ord_to_idx(ord), sector, | |
9934 | length); | |
9935 | if (ret) | |
9936 | super->updates_pending++; | |
9937 | ||
9938 | return ret; | |
9939 | } | |
c07a5a4f TM |
9940 | /******************************************************************************* |
9941 | * Function: imsm_clear_badblock | |
9942 | * Description: This routine clears bad block record from BBM log | |
9943 | * | |
9944 | * Parameters: | |
9945 | * a : array containing a bad block | |
9946 | * slot : disk number containing a bad block | |
9947 | * sector : bad block sector | |
9948 | * length : bad block sectors range | |
9949 | * Returns: | |
9950 | * 1 : Success | |
9951 | * 0 : Error | |
9952 | ******************************************************************************/ | |
9953 | static int imsm_clear_badblock(struct active_array *a, int slot, | |
9954 | unsigned long long sector, int length) | |
9955 | { | |
9956 | struct intel_super *super = a->container->sb; | |
9957 | int ord; | |
9958 | int ret; | |
9959 | ||
9960 | ord = imsm_disk_slot_to_ord(a, slot); | |
9961 | if (ord < 0) | |
9962 | return 0; | |
9963 | ||
9964 | ret = clear_badblock(super->bbm_log, ord_to_idx(ord), sector, length); | |
9965 | if (ret) | |
9966 | super->updates_pending++; | |
9967 | ||
9968 | return ret; | |
9969 | } | |
928f1424 TM |
9970 | /******************************************************************************* |
9971 | * Function: imsm_get_badblocks | |
9972 | * Description: This routine get list of bad blocks for an array | |
9973 | * | |
9974 | * Parameters: | |
9975 | * a : array | |
9976 | * slot : disk number | |
9977 | * Returns: | |
9978 | * bb : structure containing bad blocks | |
9979 | * NULL : error | |
9980 | ******************************************************************************/ | |
9981 | static struct md_bb *imsm_get_badblocks(struct active_array *a, int slot) | |
9982 | { | |
9983 | int inst = a->info.container_member; | |
9984 | struct intel_super *super = a->container->sb; | |
9985 | struct imsm_dev *dev = get_imsm_dev(super, inst); | |
9986 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
9987 | int ord; | |
9988 | ||
9989 | ord = imsm_disk_slot_to_ord(a, slot); | |
9990 | if (ord < 0) | |
9991 | return NULL; | |
9992 | ||
9993 | get_volume_badblocks(super->bbm_log, ord_to_idx(ord), pba_of_lba0(map), | |
9994 | blocks_per_member(map), &super->bb); | |
9995 | ||
9996 | return &super->bb; | |
9997 | } | |
27156a57 TM |
9998 | /******************************************************************************* |
9999 | * Function: examine_badblocks_imsm | |
10000 | * Description: Prints list of bad blocks on a disk to the standard output | |
10001 | * | |
10002 | * Parameters: | |
10003 | * st : metadata handler | |
10004 | * fd : open file descriptor for device | |
10005 | * devname : device name | |
10006 | * Returns: | |
10007 | * 0 : Success | |
10008 | * 1 : Error | |
10009 | ******************************************************************************/ | |
10010 | static int examine_badblocks_imsm(struct supertype *st, int fd, char *devname) | |
10011 | { | |
10012 | struct intel_super *super = st->sb; | |
10013 | struct bbm_log *log = super->bbm_log; | |
10014 | struct dl *d = NULL; | |
10015 | int any = 0; | |
10016 | ||
10017 | for (d = super->disks; d ; d = d->next) { | |
10018 | if (strcmp(d->devname, devname) == 0) | |
10019 | break; | |
10020 | } | |
10021 | ||
10022 | if ((d == NULL) || (d->index < 0)) { /* serial mismatch probably */ | |
10023 | pr_err("%s doesn't appear to be part of a raid array\n", | |
10024 | devname); | |
10025 | return 1; | |
10026 | } | |
10027 | ||
10028 | if (log != NULL) { | |
10029 | unsigned int i; | |
10030 | struct bbm_log_entry *entry = &log->marked_block_entries[0]; | |
10031 | ||
10032 | for (i = 0; i < log->entry_count; i++) { | |
10033 | if (entry[i].disk_ordinal == d->index) { | |
10034 | unsigned long long sector = __le48_to_cpu( | |
10035 | &entry[i].defective_block_start); | |
10036 | int cnt = entry[i].marked_count + 1; | |
10037 | ||
10038 | if (!any) { | |
10039 | printf("Bad-blocks on %s:\n", devname); | |
10040 | any = 1; | |
10041 | } | |
10042 | ||
10043 | printf("%20llu for %d sectors\n", sector, cnt); | |
10044 | } | |
10045 | } | |
10046 | } | |
10047 | ||
10048 | if (!any) | |
10049 | printf("No bad-blocks list configured on %s\n", devname); | |
10050 | ||
10051 | return 0; | |
10052 | } | |
687629c2 AK |
10053 | /******************************************************************************* |
10054 | * Function: init_migr_record_imsm | |
10055 | * Description: Function inits imsm migration record | |
10056 | * Parameters: | |
10057 | * super : imsm internal array info | |
10058 | * dev : device under migration | |
10059 | * info : general array info to find the smallest device | |
10060 | * Returns: | |
10061 | * none | |
10062 | ******************************************************************************/ | |
10063 | void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev, | |
10064 | struct mdinfo *info) | |
10065 | { | |
10066 | struct intel_super *super = st->sb; | |
10067 | struct migr_record *migr_rec = super->migr_rec; | |
10068 | int new_data_disks; | |
10069 | unsigned long long dsize, dev_sectors; | |
10070 | long long unsigned min_dev_sectors = -1LLU; | |
10071 | struct mdinfo *sd; | |
10072 | char nm[30]; | |
10073 | int fd; | |
238c0a71 AK |
10074 | struct imsm_map *map_dest = get_imsm_map(dev, MAP_0); |
10075 | struct imsm_map *map_src = get_imsm_map(dev, MAP_1); | |
687629c2 | 10076 | unsigned long long num_migr_units; |
3ef4403c | 10077 | unsigned long long array_blocks; |
687629c2 AK |
10078 | |
10079 | memset(migr_rec, 0, sizeof(struct migr_record)); | |
10080 | migr_rec->family_num = __cpu_to_le32(super->anchor->family_num); | |
10081 | ||
10082 | /* only ascending reshape supported now */ | |
10083 | migr_rec->ascending_migr = __cpu_to_le32(1); | |
10084 | ||
10085 | migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE / | |
10086 | max(map_dest->blocks_per_strip, map_src->blocks_per_strip); | |
e1742195 AK |
10087 | migr_rec->dest_depth_per_unit *= |
10088 | max(map_dest->blocks_per_strip, map_src->blocks_per_strip); | |
238c0a71 | 10089 | new_data_disks = imsm_num_data_members(dev, MAP_0); |
687629c2 AK |
10090 | migr_rec->blocks_per_unit = |
10091 | __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks); | |
10092 | migr_rec->dest_depth_per_unit = | |
10093 | __cpu_to_le32(migr_rec->dest_depth_per_unit); | |
3ef4403c | 10094 | array_blocks = info->component_size * new_data_disks; |
687629c2 AK |
10095 | num_migr_units = |
10096 | array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit); | |
10097 | ||
10098 | if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit)) | |
10099 | num_migr_units++; | |
10100 | migr_rec->num_migr_units = __cpu_to_le32(num_migr_units); | |
10101 | ||
10102 | migr_rec->post_migr_vol_cap = dev->size_low; | |
10103 | migr_rec->post_migr_vol_cap_hi = dev->size_high; | |
10104 | ||
687629c2 AK |
10105 | /* Find the smallest dev */ |
10106 | for (sd = info->devs ; sd ; sd = sd->next) { | |
10107 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
10108 | fd = dev_open(nm, O_RDONLY); | |
10109 | if (fd < 0) | |
10110 | continue; | |
10111 | get_dev_size(fd, NULL, &dsize); | |
10112 | dev_sectors = dsize / 512; | |
10113 | if (dev_sectors < min_dev_sectors) | |
10114 | min_dev_sectors = dev_sectors; | |
10115 | close(fd); | |
10116 | } | |
10117 | migr_rec->ckpt_area_pba = __cpu_to_le32(min_dev_sectors - | |
10118 | RAID_DISK_RESERVED_BLOCKS_IMSM_HI); | |
10119 | ||
10120 | write_imsm_migr_rec(st); | |
10121 | ||
10122 | return; | |
10123 | } | |
10124 | ||
10125 | /******************************************************************************* | |
10126 | * Function: save_backup_imsm | |
10127 | * Description: Function saves critical data stripes to Migration Copy Area | |
10128 | * and updates the current migration unit status. | |
10129 | * Use restore_stripes() to form a destination stripe, | |
10130 | * and to write it to the Copy Area. | |
10131 | * Parameters: | |
10132 | * st : supertype information | |
aea93171 | 10133 | * dev : imsm device that backup is saved for |
687629c2 AK |
10134 | * info : general array info |
10135 | * buf : input buffer | |
687629c2 AK |
10136 | * length : length of data to backup (blocks_per_unit) |
10137 | * Returns: | |
10138 | * 0 : success | |
10139 | *, -1 : fail | |
10140 | ******************************************************************************/ | |
10141 | int save_backup_imsm(struct supertype *st, | |
10142 | struct imsm_dev *dev, | |
10143 | struct mdinfo *info, | |
10144 | void *buf, | |
687629c2 AK |
10145 | int length) |
10146 | { | |
10147 | int rv = -1; | |
10148 | struct intel_super *super = st->sb; | |
594dc1b8 JS |
10149 | unsigned long long *target_offsets; |
10150 | int *targets; | |
687629c2 | 10151 | int i; |
238c0a71 | 10152 | struct imsm_map *map_dest = get_imsm_map(dev, MAP_0); |
687629c2 | 10153 | int new_disks = map_dest->num_members; |
ab724b98 AK |
10154 | int dest_layout = 0; |
10155 | int dest_chunk; | |
d1877f69 | 10156 | unsigned long long start; |
238c0a71 | 10157 | int data_disks = imsm_num_data_members(dev, MAP_0); |
687629c2 | 10158 | |
503975b9 | 10159 | targets = xmalloc(new_disks * sizeof(int)); |
687629c2 | 10160 | |
7e45b550 AK |
10161 | for (i = 0; i < new_disks; i++) |
10162 | targets[i] = -1; | |
10163 | ||
503975b9 | 10164 | target_offsets = xcalloc(new_disks, sizeof(unsigned long long)); |
687629c2 | 10165 | |
d1877f69 | 10166 | start = info->reshape_progress * 512; |
687629c2 | 10167 | for (i = 0; i < new_disks; i++) { |
687629c2 AK |
10168 | target_offsets[i] = (unsigned long long) |
10169 | __le32_to_cpu(super->migr_rec->ckpt_area_pba) * 512; | |
d1877f69 AK |
10170 | /* move back copy area adderss, it will be moved forward |
10171 | * in restore_stripes() using start input variable | |
10172 | */ | |
10173 | target_offsets[i] -= start/data_disks; | |
687629c2 AK |
10174 | } |
10175 | ||
9a717282 AK |
10176 | if (open_backup_targets(info, new_disks, targets, |
10177 | super, dev)) | |
687629c2 AK |
10178 | goto abort; |
10179 | ||
68eb8bc6 | 10180 | dest_layout = imsm_level_to_layout(map_dest->raid_level); |
ab724b98 AK |
10181 | dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512; |
10182 | ||
687629c2 AK |
10183 | if (restore_stripes(targets, /* list of dest devices */ |
10184 | target_offsets, /* migration record offsets */ | |
10185 | new_disks, | |
ab724b98 AK |
10186 | dest_chunk, |
10187 | map_dest->raid_level, | |
10188 | dest_layout, | |
10189 | -1, /* source backup file descriptor */ | |
10190 | 0, /* input buf offset | |
10191 | * always 0 buf is already offseted */ | |
d1877f69 | 10192 | start, |
687629c2 AK |
10193 | length, |
10194 | buf) != 0) { | |
e7b84f9d | 10195 | pr_err("Error restoring stripes\n"); |
687629c2 AK |
10196 | goto abort; |
10197 | } | |
10198 | ||
10199 | rv = 0; | |
10200 | ||
10201 | abort: | |
10202 | if (targets) { | |
9a717282 | 10203 | close_targets(targets, new_disks); |
687629c2 AK |
10204 | free(targets); |
10205 | } | |
10206 | free(target_offsets); | |
10207 | ||
10208 | return rv; | |
10209 | } | |
10210 | ||
10211 | /******************************************************************************* | |
10212 | * Function: save_checkpoint_imsm | |
10213 | * Description: Function called for current unit status update | |
10214 | * in the migration record. It writes it to disk. | |
10215 | * Parameters: | |
10216 | * super : imsm internal array info | |
10217 | * info : general array info | |
10218 | * Returns: | |
10219 | * 0: success | |
10220 | * 1: failure | |
0228d92c AK |
10221 | * 2: failure, means no valid migration record |
10222 | * / no general migration in progress / | |
687629c2 AK |
10223 | ******************************************************************************/ |
10224 | int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state) | |
10225 | { | |
10226 | struct intel_super *super = st->sb; | |
f8b72ef5 AK |
10227 | unsigned long long blocks_per_unit; |
10228 | unsigned long long curr_migr_unit; | |
10229 | ||
2e062e82 | 10230 | if (load_imsm_migr_rec(super, info) != 0) { |
7a862a02 | 10231 | dprintf("imsm: ERROR: Cannot read migration record for checkpoint save.\n"); |
2e062e82 AK |
10232 | return 1; |
10233 | } | |
10234 | ||
f8b72ef5 AK |
10235 | blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit); |
10236 | if (blocks_per_unit == 0) { | |
0228d92c AK |
10237 | dprintf("imsm: no migration in progress.\n"); |
10238 | return 2; | |
687629c2 | 10239 | } |
f8b72ef5 AK |
10240 | curr_migr_unit = info->reshape_progress / blocks_per_unit; |
10241 | /* check if array is alligned to copy area | |
10242 | * if it is not alligned, add one to current migration unit value | |
10243 | * this can happend on array reshape finish only | |
10244 | */ | |
10245 | if (info->reshape_progress % blocks_per_unit) | |
10246 | curr_migr_unit++; | |
687629c2 AK |
10247 | |
10248 | super->migr_rec->curr_migr_unit = | |
f8b72ef5 | 10249 | __cpu_to_le32(curr_migr_unit); |
687629c2 AK |
10250 | super->migr_rec->rec_status = __cpu_to_le32(state); |
10251 | super->migr_rec->dest_1st_member_lba = | |
f8b72ef5 AK |
10252 | __cpu_to_le32(curr_migr_unit * |
10253 | __le32_to_cpu(super->migr_rec->dest_depth_per_unit)); | |
687629c2 | 10254 | if (write_imsm_migr_rec(st) < 0) { |
7a862a02 | 10255 | dprintf("imsm: Cannot write migration record outside backup area\n"); |
687629c2 AK |
10256 | return 1; |
10257 | } | |
10258 | ||
10259 | return 0; | |
10260 | } | |
10261 | ||
276d77db AK |
10262 | /******************************************************************************* |
10263 | * Function: recover_backup_imsm | |
10264 | * Description: Function recovers critical data from the Migration Copy Area | |
10265 | * while assembling an array. | |
10266 | * Parameters: | |
10267 | * super : imsm internal array info | |
10268 | * info : general array info | |
10269 | * Returns: | |
10270 | * 0 : success (or there is no data to recover) | |
10271 | * 1 : fail | |
10272 | ******************************************************************************/ | |
10273 | int recover_backup_imsm(struct supertype *st, struct mdinfo *info) | |
10274 | { | |
10275 | struct intel_super *super = st->sb; | |
10276 | struct migr_record *migr_rec = super->migr_rec; | |
594dc1b8 | 10277 | struct imsm_map *map_dest; |
276d77db AK |
10278 | struct intel_dev *id = NULL; |
10279 | unsigned long long read_offset; | |
10280 | unsigned long long write_offset; | |
10281 | unsigned unit_len; | |
10282 | int *targets = NULL; | |
10283 | int new_disks, i, err; | |
10284 | char *buf = NULL; | |
10285 | int retval = 1; | |
f36a9ecd | 10286 | unsigned int sector_size = super->sector_size; |
276d77db AK |
10287 | unsigned long curr_migr_unit = __le32_to_cpu(migr_rec->curr_migr_unit); |
10288 | unsigned long num_migr_units = __le32_to_cpu(migr_rec->num_migr_units); | |
276d77db | 10289 | char buffer[20]; |
6c3560c0 | 10290 | int skipped_disks = 0; |
276d77db AK |
10291 | |
10292 | err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20); | |
10293 | if (err < 1) | |
10294 | return 1; | |
10295 | ||
10296 | /* recover data only during assemblation */ | |
10297 | if (strncmp(buffer, "inactive", 8) != 0) | |
10298 | return 0; | |
10299 | /* no data to recover */ | |
10300 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
10301 | return 0; | |
10302 | if (curr_migr_unit >= num_migr_units) | |
10303 | return 1; | |
10304 | ||
10305 | /* find device during reshape */ | |
10306 | for (id = super->devlist; id; id = id->next) | |
10307 | if (is_gen_migration(id->dev)) | |
10308 | break; | |
10309 | if (id == NULL) | |
10310 | return 1; | |
10311 | ||
238c0a71 | 10312 | map_dest = get_imsm_map(id->dev, MAP_0); |
276d77db AK |
10313 | new_disks = map_dest->num_members; |
10314 | ||
10315 | read_offset = (unsigned long long) | |
10316 | __le32_to_cpu(migr_rec->ckpt_area_pba) * 512; | |
10317 | ||
10318 | write_offset = ((unsigned long long) | |
10319 | __le32_to_cpu(migr_rec->dest_1st_member_lba) + | |
5551b113 | 10320 | pba_of_lba0(map_dest)) * 512; |
276d77db AK |
10321 | |
10322 | unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
f36a9ecd | 10323 | if (posix_memalign((void **)&buf, sector_size, unit_len) != 0) |
276d77db | 10324 | goto abort; |
503975b9 | 10325 | targets = xcalloc(new_disks, sizeof(int)); |
276d77db | 10326 | |
9a717282 | 10327 | if (open_backup_targets(info, new_disks, targets, super, id->dev)) { |
e7b84f9d | 10328 | pr_err("Cannot open some devices belonging to array.\n"); |
f627f5ad AK |
10329 | goto abort; |
10330 | } | |
276d77db AK |
10331 | |
10332 | for (i = 0; i < new_disks; i++) { | |
6c3560c0 AK |
10333 | if (targets[i] < 0) { |
10334 | skipped_disks++; | |
10335 | continue; | |
10336 | } | |
276d77db | 10337 | if (lseek64(targets[i], read_offset, SEEK_SET) < 0) { |
e7b84f9d N |
10338 | pr_err("Cannot seek to block: %s\n", |
10339 | strerror(errno)); | |
137debce AK |
10340 | skipped_disks++; |
10341 | continue; | |
276d77db | 10342 | } |
9ec11d1a | 10343 | if ((unsigned)read(targets[i], buf, unit_len) != unit_len) { |
e7b84f9d N |
10344 | pr_err("Cannot read copy area block: %s\n", |
10345 | strerror(errno)); | |
137debce AK |
10346 | skipped_disks++; |
10347 | continue; | |
276d77db AK |
10348 | } |
10349 | if (lseek64(targets[i], write_offset, SEEK_SET) < 0) { | |
e7b84f9d N |
10350 | pr_err("Cannot seek to block: %s\n", |
10351 | strerror(errno)); | |
137debce AK |
10352 | skipped_disks++; |
10353 | continue; | |
276d77db | 10354 | } |
9ec11d1a | 10355 | if ((unsigned)write(targets[i], buf, unit_len) != unit_len) { |
e7b84f9d N |
10356 | pr_err("Cannot restore block: %s\n", |
10357 | strerror(errno)); | |
137debce AK |
10358 | skipped_disks++; |
10359 | continue; | |
276d77db AK |
10360 | } |
10361 | } | |
10362 | ||
137debce AK |
10363 | if (skipped_disks > imsm_get_allowed_degradation(info->new_level, |
10364 | new_disks, | |
10365 | super, | |
10366 | id->dev)) { | |
7a862a02 | 10367 | pr_err("Cannot restore data from backup. Too many failed disks\n"); |
6c3560c0 AK |
10368 | goto abort; |
10369 | } | |
10370 | ||
befb629b AK |
10371 | if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) { |
10372 | /* ignore error == 2, this can mean end of reshape here | |
10373 | */ | |
7a862a02 | 10374 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL) during restart\n"); |
befb629b | 10375 | } else |
276d77db | 10376 | retval = 0; |
276d77db AK |
10377 | |
10378 | abort: | |
10379 | if (targets) { | |
10380 | for (i = 0; i < new_disks; i++) | |
10381 | if (targets[i]) | |
10382 | close(targets[i]); | |
10383 | free(targets); | |
10384 | } | |
10385 | free(buf); | |
10386 | return retval; | |
10387 | } | |
10388 | ||
2cda7640 ML |
10389 | static char disk_by_path[] = "/dev/disk/by-path/"; |
10390 | ||
10391 | static const char *imsm_get_disk_controller_domain(const char *path) | |
10392 | { | |
2cda7640 | 10393 | char disk_path[PATH_MAX]; |
96234762 LM |
10394 | char *drv=NULL; |
10395 | struct stat st; | |
2cda7640 | 10396 | |
6d8d290a | 10397 | strcpy(disk_path, disk_by_path); |
96234762 LM |
10398 | strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1); |
10399 | if (stat(disk_path, &st) == 0) { | |
10400 | struct sys_dev* hba; | |
594dc1b8 | 10401 | char *path; |
96234762 LM |
10402 | |
10403 | path = devt_to_devpath(st.st_rdev); | |
10404 | if (path == NULL) | |
10405 | return "unknown"; | |
10406 | hba = find_disk_attached_hba(-1, path); | |
10407 | if (hba && hba->type == SYS_DEV_SAS) | |
10408 | drv = "isci"; | |
10409 | else if (hba && hba->type == SYS_DEV_SATA) | |
10410 | drv = "ahci"; | |
1011e834 | 10411 | else |
96234762 LM |
10412 | drv = "unknown"; |
10413 | dprintf("path: %s hba: %s attached: %s\n", | |
10414 | path, (hba) ? hba->path : "NULL", drv); | |
10415 | free(path); | |
2cda7640 | 10416 | } |
96234762 | 10417 | return drv; |
2cda7640 ML |
10418 | } |
10419 | ||
4dd2df09 | 10420 | static char *imsm_find_array_devnm_by_subdev(int subdev, char *container) |
78b10e66 | 10421 | { |
4dd2df09 | 10422 | static char devnm[32]; |
78b10e66 N |
10423 | char subdev_name[20]; |
10424 | struct mdstat_ent *mdstat; | |
10425 | ||
10426 | sprintf(subdev_name, "%d", subdev); | |
10427 | mdstat = mdstat_by_subdev(subdev_name, container); | |
10428 | if (!mdstat) | |
4dd2df09 | 10429 | return NULL; |
78b10e66 | 10430 | |
4dd2df09 | 10431 | strcpy(devnm, mdstat->devnm); |
78b10e66 | 10432 | free_mdstat(mdstat); |
4dd2df09 | 10433 | return devnm; |
78b10e66 N |
10434 | } |
10435 | ||
10436 | static int imsm_reshape_is_allowed_on_container(struct supertype *st, | |
10437 | struct geo_params *geo, | |
fbf3d202 AK |
10438 | int *old_raid_disks, |
10439 | int direction) | |
78b10e66 | 10440 | { |
694575e7 KW |
10441 | /* currently we only support increasing the number of devices |
10442 | * for a container. This increases the number of device for each | |
10443 | * member array. They must all be RAID0 or RAID5. | |
10444 | */ | |
78b10e66 N |
10445 | int ret_val = 0; |
10446 | struct mdinfo *info, *member; | |
10447 | int devices_that_can_grow = 0; | |
10448 | ||
7a862a02 | 10449 | dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): st->devnm = (%s)\n", st->devnm); |
78b10e66 | 10450 | |
d04f65f4 | 10451 | if (geo->size > 0 || |
78b10e66 N |
10452 | geo->level != UnSet || |
10453 | geo->layout != UnSet || | |
10454 | geo->chunksize != 0 || | |
10455 | geo->raid_disks == UnSet) { | |
7a862a02 | 10456 | dprintf("imsm: Container operation is allowed for raid disks number change only.\n"); |
78b10e66 N |
10457 | return ret_val; |
10458 | } | |
10459 | ||
fbf3d202 | 10460 | if (direction == ROLLBACK_METADATA_CHANGES) { |
7a862a02 | 10461 | dprintf("imsm: Metadata changes rollback is not supported for container operation.\n"); |
fbf3d202 AK |
10462 | return ret_val; |
10463 | } | |
10464 | ||
78b10e66 N |
10465 | info = container_content_imsm(st, NULL); |
10466 | for (member = info; member; member = member->next) { | |
4dd2df09 | 10467 | char *result; |
78b10e66 N |
10468 | |
10469 | dprintf("imsm: checking device_num: %i\n", | |
10470 | member->container_member); | |
10471 | ||
d7d205bd | 10472 | if (geo->raid_disks <= member->array.raid_disks) { |
78b10e66 N |
10473 | /* we work on container for Online Capacity Expansion |
10474 | * only so raid_disks has to grow | |
10475 | */ | |
7a862a02 | 10476 | dprintf("imsm: for container operation raid disks increase is required\n"); |
78b10e66 N |
10477 | break; |
10478 | } | |
10479 | ||
089f9d79 | 10480 | if (info->array.level != 0 && info->array.level != 5) { |
78b10e66 N |
10481 | /* we cannot use this container with other raid level |
10482 | */ | |
7a862a02 | 10483 | dprintf("imsm: for container operation wrong raid level (%i) detected\n", |
78b10e66 N |
10484 | info->array.level); |
10485 | break; | |
10486 | } else { | |
10487 | /* check for platform support | |
10488 | * for this raid level configuration | |
10489 | */ | |
10490 | struct intel_super *super = st->sb; | |
10491 | if (!is_raid_level_supported(super->orom, | |
10492 | member->array.level, | |
10493 | geo->raid_disks)) { | |
7a862a02 | 10494 | dprintf("platform does not support raid%d with %d disk%s\n", |
78b10e66 N |
10495 | info->array.level, |
10496 | geo->raid_disks, | |
10497 | geo->raid_disks > 1 ? "s" : ""); | |
10498 | break; | |
10499 | } | |
2a4a08e7 AK |
10500 | /* check if component size is aligned to chunk size |
10501 | */ | |
10502 | if (info->component_size % | |
10503 | (info->array.chunk_size/512)) { | |
7a862a02 | 10504 | dprintf("Component size is not aligned to chunk size\n"); |
2a4a08e7 AK |
10505 | break; |
10506 | } | |
78b10e66 N |
10507 | } |
10508 | ||
10509 | if (*old_raid_disks && | |
10510 | info->array.raid_disks != *old_raid_disks) | |
10511 | break; | |
10512 | *old_raid_disks = info->array.raid_disks; | |
10513 | ||
10514 | /* All raid5 and raid0 volumes in container | |
10515 | * have to be ready for Online Capacity Expansion | |
10516 | * so they need to be assembled. We have already | |
10517 | * checked that no recovery etc is happening. | |
10518 | */ | |
4dd2df09 N |
10519 | result = imsm_find_array_devnm_by_subdev(member->container_member, |
10520 | st->container_devnm); | |
10521 | if (result == NULL) { | |
78b10e66 N |
10522 | dprintf("imsm: cannot find array\n"); |
10523 | break; | |
10524 | } | |
10525 | devices_that_can_grow++; | |
10526 | } | |
10527 | sysfs_free(info); | |
10528 | if (!member && devices_that_can_grow) | |
10529 | ret_val = 1; | |
10530 | ||
10531 | if (ret_val) | |
1ade5cc1 | 10532 | dprintf("Container operation allowed\n"); |
78b10e66 | 10533 | else |
1ade5cc1 | 10534 | dprintf("Error: %i\n", ret_val); |
78b10e66 N |
10535 | |
10536 | return ret_val; | |
10537 | } | |
10538 | ||
10539 | /* Function: get_spares_for_grow | |
10540 | * Description: Allocates memory and creates list of spare devices | |
1011e834 | 10541 | * avaliable in container. Checks if spare drive size is acceptable. |
78b10e66 N |
10542 | * Parameters: Pointer to the supertype structure |
10543 | * Returns: Pointer to the list of spare devices (mdinfo structure) on success, | |
1011e834 | 10544 | * NULL if fail |
78b10e66 N |
10545 | */ |
10546 | static struct mdinfo *get_spares_for_grow(struct supertype *st) | |
10547 | { | |
78b10e66 | 10548 | unsigned long long min_size = min_acceptable_spare_size_imsm(st); |
326727d9 | 10549 | return container_choose_spares(st, min_size, NULL, NULL, NULL, 0); |
78b10e66 N |
10550 | } |
10551 | ||
10552 | /****************************************************************************** | |
10553 | * function: imsm_create_metadata_update_for_reshape | |
10554 | * Function creates update for whole IMSM container. | |
10555 | * | |
10556 | ******************************************************************************/ | |
10557 | static int imsm_create_metadata_update_for_reshape( | |
10558 | struct supertype *st, | |
10559 | struct geo_params *geo, | |
10560 | int old_raid_disks, | |
10561 | struct imsm_update_reshape **updatep) | |
10562 | { | |
10563 | struct intel_super *super = st->sb; | |
10564 | struct imsm_super *mpb = super->anchor; | |
594dc1b8 JS |
10565 | int update_memory_size; |
10566 | struct imsm_update_reshape *u; | |
10567 | struct mdinfo *spares; | |
78b10e66 | 10568 | int i; |
594dc1b8 | 10569 | int delta_disks; |
bbd24d86 | 10570 | struct mdinfo *dev; |
78b10e66 | 10571 | |
1ade5cc1 | 10572 | dprintf("(enter) raid_disks = %i\n", geo->raid_disks); |
78b10e66 N |
10573 | |
10574 | delta_disks = geo->raid_disks - old_raid_disks; | |
10575 | ||
10576 | /* size of all update data without anchor */ | |
10577 | update_memory_size = sizeof(struct imsm_update_reshape); | |
10578 | ||
10579 | /* now add space for spare disks that we need to add. */ | |
10580 | update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1); | |
10581 | ||
503975b9 | 10582 | u = xcalloc(1, update_memory_size); |
78b10e66 N |
10583 | u->type = update_reshape_container_disks; |
10584 | u->old_raid_disks = old_raid_disks; | |
10585 | u->new_raid_disks = geo->raid_disks; | |
10586 | ||
10587 | /* now get spare disks list | |
10588 | */ | |
10589 | spares = get_spares_for_grow(st); | |
10590 | ||
10591 | if (spares == NULL | |
10592 | || delta_disks > spares->array.spare_disks) { | |
7a862a02 | 10593 | pr_err("imsm: ERROR: Cannot get spare devices for %s.\n", geo->dev_name); |
e4c72d1d | 10594 | i = -1; |
78b10e66 N |
10595 | goto abort; |
10596 | } | |
10597 | ||
10598 | /* we have got spares | |
10599 | * update disk list in imsm_disk list table in anchor | |
10600 | */ | |
10601 | dprintf("imsm: %i spares are available.\n\n", | |
10602 | spares->array.spare_disks); | |
10603 | ||
bbd24d86 | 10604 | dev = spares->devs; |
78b10e66 | 10605 | for (i = 0; i < delta_disks; i++) { |
78b10e66 N |
10606 | struct dl *dl; |
10607 | ||
bbd24d86 AK |
10608 | if (dev == NULL) |
10609 | break; | |
78b10e66 N |
10610 | u->new_disks[i] = makedev(dev->disk.major, |
10611 | dev->disk.minor); | |
10612 | dl = get_disk_super(super, dev->disk.major, dev->disk.minor); | |
ee4beede AK |
10613 | dl->index = mpb->num_disks; |
10614 | mpb->num_disks++; | |
bbd24d86 | 10615 | dev = dev->next; |
78b10e66 | 10616 | } |
78b10e66 N |
10617 | |
10618 | abort: | |
10619 | /* free spares | |
10620 | */ | |
10621 | sysfs_free(spares); | |
10622 | ||
d677e0b8 | 10623 | dprintf("imsm: reshape update preparation :"); |
78b10e66 | 10624 | if (i == delta_disks) { |
1ade5cc1 | 10625 | dprintf_cont(" OK\n"); |
78b10e66 N |
10626 | *updatep = u; |
10627 | return update_memory_size; | |
10628 | } | |
10629 | free(u); | |
1ade5cc1 | 10630 | dprintf_cont(" Error\n"); |
78b10e66 N |
10631 | |
10632 | return 0; | |
10633 | } | |
10634 | ||
f3871fdc AK |
10635 | /****************************************************************************** |
10636 | * function: imsm_create_metadata_update_for_size_change() | |
10637 | * Creates update for IMSM array for array size change. | |
10638 | * | |
10639 | ******************************************************************************/ | |
10640 | static int imsm_create_metadata_update_for_size_change( | |
10641 | struct supertype *st, | |
10642 | struct geo_params *geo, | |
10643 | struct imsm_update_size_change **updatep) | |
10644 | { | |
10645 | struct intel_super *super = st->sb; | |
594dc1b8 JS |
10646 | int update_memory_size; |
10647 | struct imsm_update_size_change *u; | |
f3871fdc | 10648 | |
1ade5cc1 | 10649 | dprintf("(enter) New size = %llu\n", geo->size); |
f3871fdc AK |
10650 | |
10651 | /* size of all update data without anchor */ | |
10652 | update_memory_size = sizeof(struct imsm_update_size_change); | |
10653 | ||
503975b9 | 10654 | u = xcalloc(1, update_memory_size); |
f3871fdc AK |
10655 | u->type = update_size_change; |
10656 | u->subdev = super->current_vol; | |
10657 | u->new_size = geo->size; | |
10658 | ||
10659 | dprintf("imsm: reshape update preparation : OK\n"); | |
10660 | *updatep = u; | |
10661 | ||
10662 | return update_memory_size; | |
10663 | } | |
10664 | ||
48c5303a PC |
10665 | /****************************************************************************** |
10666 | * function: imsm_create_metadata_update_for_migration() | |
10667 | * Creates update for IMSM array. | |
10668 | * | |
10669 | ******************************************************************************/ | |
10670 | static int imsm_create_metadata_update_for_migration( | |
10671 | struct supertype *st, | |
10672 | struct geo_params *geo, | |
10673 | struct imsm_update_reshape_migration **updatep) | |
10674 | { | |
10675 | struct intel_super *super = st->sb; | |
594dc1b8 JS |
10676 | int update_memory_size; |
10677 | struct imsm_update_reshape_migration *u; | |
48c5303a PC |
10678 | struct imsm_dev *dev; |
10679 | int previous_level = -1; | |
10680 | ||
1ade5cc1 | 10681 | dprintf("(enter) New Level = %i\n", geo->level); |
48c5303a PC |
10682 | |
10683 | /* size of all update data without anchor */ | |
10684 | update_memory_size = sizeof(struct imsm_update_reshape_migration); | |
10685 | ||
503975b9 | 10686 | u = xcalloc(1, update_memory_size); |
48c5303a PC |
10687 | u->type = update_reshape_migration; |
10688 | u->subdev = super->current_vol; | |
10689 | u->new_level = geo->level; | |
10690 | u->new_layout = geo->layout; | |
10691 | u->new_raid_disks = u->old_raid_disks = geo->raid_disks; | |
10692 | u->new_disks[0] = -1; | |
4bba0439 | 10693 | u->new_chunksize = -1; |
48c5303a PC |
10694 | |
10695 | dev = get_imsm_dev(super, u->subdev); | |
10696 | if (dev) { | |
10697 | struct imsm_map *map; | |
10698 | ||
238c0a71 | 10699 | map = get_imsm_map(dev, MAP_0); |
4bba0439 PC |
10700 | if (map) { |
10701 | int current_chunk_size = | |
10702 | __le16_to_cpu(map->blocks_per_strip) / 2; | |
10703 | ||
10704 | if (geo->chunksize != current_chunk_size) { | |
10705 | u->new_chunksize = geo->chunksize / 1024; | |
7a862a02 | 10706 | dprintf("imsm: chunk size change from %i to %i\n", |
4bba0439 PC |
10707 | current_chunk_size, u->new_chunksize); |
10708 | } | |
48c5303a | 10709 | previous_level = map->raid_level; |
4bba0439 | 10710 | } |
48c5303a | 10711 | } |
089f9d79 | 10712 | if (geo->level == 5 && previous_level == 0) { |
48c5303a PC |
10713 | struct mdinfo *spares = NULL; |
10714 | ||
10715 | u->new_raid_disks++; | |
10716 | spares = get_spares_for_grow(st); | |
089f9d79 | 10717 | if (spares == NULL || spares->array.spare_disks < 1) { |
48c5303a PC |
10718 | free(u); |
10719 | sysfs_free(spares); | |
10720 | update_memory_size = 0; | |
565cc99e | 10721 | pr_err("cannot get spare device for requested migration\n"); |
48c5303a PC |
10722 | return 0; |
10723 | } | |
10724 | sysfs_free(spares); | |
10725 | } | |
10726 | dprintf("imsm: reshape update preparation : OK\n"); | |
10727 | *updatep = u; | |
10728 | ||
10729 | return update_memory_size; | |
10730 | } | |
10731 | ||
8dd70bce AK |
10732 | static void imsm_update_metadata_locally(struct supertype *st, |
10733 | void *buf, int len) | |
10734 | { | |
10735 | struct metadata_update mu; | |
10736 | ||
10737 | mu.buf = buf; | |
10738 | mu.len = len; | |
10739 | mu.space = NULL; | |
10740 | mu.space_list = NULL; | |
10741 | mu.next = NULL; | |
5fe6f031 N |
10742 | if (imsm_prepare_update(st, &mu)) |
10743 | imsm_process_update(st, &mu); | |
8dd70bce AK |
10744 | |
10745 | while (mu.space_list) { | |
10746 | void **space = mu.space_list; | |
10747 | mu.space_list = *space; | |
10748 | free(space); | |
10749 | } | |
10750 | } | |
78b10e66 | 10751 | |
471bceb6 | 10752 | /*************************************************************************** |
694575e7 | 10753 | * Function: imsm_analyze_change |
471bceb6 | 10754 | * Description: Function analyze change for single volume |
1011e834 | 10755 | * and validate if transition is supported |
fbf3d202 AK |
10756 | * Parameters: Geometry parameters, supertype structure, |
10757 | * metadata change direction (apply/rollback) | |
694575e7 | 10758 | * Returns: Operation type code on success, -1 if fail |
471bceb6 KW |
10759 | ****************************************************************************/ |
10760 | enum imsm_reshape_type imsm_analyze_change(struct supertype *st, | |
fbf3d202 AK |
10761 | struct geo_params *geo, |
10762 | int direction) | |
694575e7 | 10763 | { |
471bceb6 KW |
10764 | struct mdinfo info; |
10765 | int change = -1; | |
10766 | int check_devs = 0; | |
c21e737b | 10767 | int chunk; |
67a2db32 AK |
10768 | /* number of added/removed disks in operation result */ |
10769 | int devNumChange = 0; | |
10770 | /* imsm compatible layout value for array geometry verification */ | |
10771 | int imsm_layout = -1; | |
7abc9871 AK |
10772 | int data_disks; |
10773 | struct imsm_dev *dev; | |
10774 | struct intel_super *super; | |
d04f65f4 | 10775 | unsigned long long current_size; |
65d38cca | 10776 | unsigned long long free_size; |
d04f65f4 | 10777 | unsigned long long max_size; |
65d38cca | 10778 | int rv; |
471bceb6 KW |
10779 | |
10780 | getinfo_super_imsm_volume(st, &info, NULL); | |
089f9d79 JS |
10781 | if (geo->level != info.array.level && geo->level >= 0 && |
10782 | geo->level != UnSet) { | |
471bceb6 KW |
10783 | switch (info.array.level) { |
10784 | case 0: | |
10785 | if (geo->level == 5) { | |
b5347799 | 10786 | change = CH_MIGRATION; |
e13ce846 | 10787 | if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) { |
7a862a02 | 10788 | pr_err("Error. Requested Layout not supported (left-asymmetric layout is supported only)!\n"); |
e13ce846 AK |
10789 | change = -1; |
10790 | goto analyse_change_exit; | |
10791 | } | |
67a2db32 | 10792 | imsm_layout = geo->layout; |
471bceb6 | 10793 | check_devs = 1; |
e91a3bad LM |
10794 | devNumChange = 1; /* parity disk added */ |
10795 | } else if (geo->level == 10) { | |
471bceb6 KW |
10796 | change = CH_TAKEOVER; |
10797 | check_devs = 1; | |
e91a3bad | 10798 | devNumChange = 2; /* two mirrors added */ |
67a2db32 | 10799 | imsm_layout = 0x102; /* imsm supported layout */ |
471bceb6 | 10800 | } |
dfe77a9e KW |
10801 | break; |
10802 | case 1: | |
471bceb6 KW |
10803 | case 10: |
10804 | if (geo->level == 0) { | |
10805 | change = CH_TAKEOVER; | |
10806 | check_devs = 1; | |
e91a3bad | 10807 | devNumChange = -(geo->raid_disks/2); |
67a2db32 | 10808 | imsm_layout = 0; /* imsm raid0 layout */ |
471bceb6 KW |
10809 | } |
10810 | break; | |
10811 | } | |
10812 | if (change == -1) { | |
7a862a02 | 10813 | pr_err("Error. Level Migration from %d to %d not supported!\n", |
e7b84f9d | 10814 | info.array.level, geo->level); |
471bceb6 KW |
10815 | goto analyse_change_exit; |
10816 | } | |
10817 | } else | |
10818 | geo->level = info.array.level; | |
10819 | ||
089f9d79 JS |
10820 | if (geo->layout != info.array.layout && |
10821 | (geo->layout != UnSet && geo->layout != -1)) { | |
b5347799 | 10822 | change = CH_MIGRATION; |
089f9d79 JS |
10823 | if (info.array.layout == 0 && info.array.level == 5 && |
10824 | geo->layout == 5) { | |
471bceb6 | 10825 | /* reshape 5 -> 4 */ |
089f9d79 JS |
10826 | } else if (info.array.layout == 5 && info.array.level == 5 && |
10827 | geo->layout == 0) { | |
471bceb6 KW |
10828 | /* reshape 4 -> 5 */ |
10829 | geo->layout = 0; | |
10830 | geo->level = 5; | |
10831 | } else { | |
7a862a02 | 10832 | pr_err("Error. Layout Migration from %d to %d not supported!\n", |
e7b84f9d | 10833 | info.array.layout, geo->layout); |
471bceb6 KW |
10834 | change = -1; |
10835 | goto analyse_change_exit; | |
10836 | } | |
67a2db32 | 10837 | } else { |
471bceb6 | 10838 | geo->layout = info.array.layout; |
67a2db32 AK |
10839 | if (imsm_layout == -1) |
10840 | imsm_layout = info.array.layout; | |
10841 | } | |
471bceb6 | 10842 | |
089f9d79 JS |
10843 | if (geo->chunksize > 0 && geo->chunksize != UnSet && |
10844 | geo->chunksize != info.array.chunk_size) { | |
2d2b0eb7 MD |
10845 | if (info.array.level == 10) { |
10846 | pr_err("Error. Chunk size change for RAID 10 is not supported.\n"); | |
10847 | change = -1; | |
10848 | goto analyse_change_exit; | |
1e9b2c3f PB |
10849 | } else if (info.component_size % (geo->chunksize/512)) { |
10850 | pr_err("New chunk size (%dK) does not evenly divide device size (%lluk). Aborting...\n", | |
10851 | geo->chunksize/1024, info.component_size/2); | |
10852 | change = -1; | |
10853 | goto analyse_change_exit; | |
2d2b0eb7 | 10854 | } |
b5347799 | 10855 | change = CH_MIGRATION; |
2d2b0eb7 | 10856 | } else { |
471bceb6 | 10857 | geo->chunksize = info.array.chunk_size; |
2d2b0eb7 | 10858 | } |
471bceb6 | 10859 | |
c21e737b | 10860 | chunk = geo->chunksize / 1024; |
7abc9871 AK |
10861 | |
10862 | super = st->sb; | |
10863 | dev = get_imsm_dev(super, super->current_vol); | |
10864 | data_disks = imsm_num_data_members(dev , MAP_0); | |
c41e00b2 | 10865 | /* compute current size per disk member |
7abc9871 | 10866 | */ |
c41e00b2 AK |
10867 | current_size = info.custom_array_size / data_disks; |
10868 | ||
089f9d79 | 10869 | if (geo->size > 0 && geo->size != MAX_SIZE) { |
c41e00b2 AK |
10870 | /* align component size |
10871 | */ | |
10872 | geo->size = imsm_component_size_aligment_check( | |
10873 | get_imsm_raid_level(dev->vol.map), | |
f36a9ecd | 10874 | chunk * 1024, super->sector_size, |
c41e00b2 | 10875 | geo->size * 2); |
65d0b4ce | 10876 | if (geo->size == 0) { |
7a862a02 | 10877 | pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is 0).\n", |
65d0b4ce LD |
10878 | current_size); |
10879 | goto analyse_change_exit; | |
10880 | } | |
c41e00b2 | 10881 | } |
7abc9871 | 10882 | |
089f9d79 | 10883 | if (current_size != geo->size && geo->size > 0) { |
7abc9871 | 10884 | if (change != -1) { |
7a862a02 | 10885 | pr_err("Error. Size change should be the only one at a time.\n"); |
7abc9871 AK |
10886 | change = -1; |
10887 | goto analyse_change_exit; | |
10888 | } | |
10889 | if ((super->current_vol + 1) != super->anchor->num_raid_devs) { | |
7a862a02 | 10890 | pr_err("Error. The last volume in container can be expanded only (%i/%s).\n", |
4dd2df09 | 10891 | super->current_vol, st->devnm); |
7abc9871 AK |
10892 | goto analyse_change_exit; |
10893 | } | |
65d38cca LD |
10894 | /* check the maximum available size |
10895 | */ | |
10896 | rv = imsm_get_free_size(st, dev->vol.map->num_members, | |
10897 | 0, chunk, &free_size); | |
10898 | if (rv == 0) | |
10899 | /* Cannot find maximum available space | |
10900 | */ | |
10901 | max_size = 0; | |
10902 | else { | |
10903 | max_size = free_size + current_size; | |
10904 | /* align component size | |
10905 | */ | |
10906 | max_size = imsm_component_size_aligment_check( | |
10907 | get_imsm_raid_level(dev->vol.map), | |
f36a9ecd | 10908 | chunk * 1024, super->sector_size, |
65d38cca LD |
10909 | max_size); |
10910 | } | |
d04f65f4 | 10911 | if (geo->size == MAX_SIZE) { |
b130333f AK |
10912 | /* requested size change to the maximum available size |
10913 | */ | |
65d38cca | 10914 | if (max_size == 0) { |
7a862a02 | 10915 | pr_err("Error. Cannot find maximum available space.\n"); |
b130333f AK |
10916 | change = -1; |
10917 | goto analyse_change_exit; | |
65d38cca LD |
10918 | } else |
10919 | geo->size = max_size; | |
c41e00b2 | 10920 | } |
b130333f | 10921 | |
681b7ae2 | 10922 | if (direction == ROLLBACK_METADATA_CHANGES) { |
fbf3d202 AK |
10923 | /* accept size for rollback only |
10924 | */ | |
10925 | } else { | |
10926 | /* round size due to metadata compatibility | |
10927 | */ | |
10928 | geo->size = (geo->size >> SECT_PER_MB_SHIFT) | |
10929 | << SECT_PER_MB_SHIFT; | |
10930 | dprintf("Prepare update for size change to %llu\n", | |
10931 | geo->size ); | |
10932 | if (current_size >= geo->size) { | |
7a862a02 | 10933 | pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is %llu).\n", |
e7b84f9d | 10934 | current_size, geo->size); |
fbf3d202 AK |
10935 | goto analyse_change_exit; |
10936 | } | |
65d38cca | 10937 | if (max_size && geo->size > max_size) { |
7a862a02 | 10938 | pr_err("Error. Requested size is larger than maximum available size (maximum available size is %llu, requested size /rounded/ is %llu).\n", |
e7b84f9d | 10939 | max_size, geo->size); |
65d38cca LD |
10940 | goto analyse_change_exit; |
10941 | } | |
7abc9871 AK |
10942 | } |
10943 | geo->size *= data_disks; | |
10944 | geo->raid_disks = dev->vol.map->num_members; | |
10945 | change = CH_ARRAY_SIZE; | |
10946 | } | |
471bceb6 KW |
10947 | if (!validate_geometry_imsm(st, |
10948 | geo->level, | |
67a2db32 | 10949 | imsm_layout, |
e91a3bad | 10950 | geo->raid_disks + devNumChange, |
c21e737b | 10951 | &chunk, |
af4348dd | 10952 | geo->size, INVALID_SECTORS, |
471bceb6 KW |
10953 | 0, 0, 1)) |
10954 | change = -1; | |
10955 | ||
10956 | if (check_devs) { | |
10957 | struct intel_super *super = st->sb; | |
10958 | struct imsm_super *mpb = super->anchor; | |
10959 | ||
10960 | if (mpb->num_raid_devs > 1) { | |
7a862a02 | 10961 | pr_err("Error. Cannot perform operation on %s- for this operation it MUST be single array in container\n", |
e7b84f9d | 10962 | geo->dev_name); |
471bceb6 KW |
10963 | change = -1; |
10964 | } | |
10965 | } | |
10966 | ||
10967 | analyse_change_exit: | |
089f9d79 JS |
10968 | if (direction == ROLLBACK_METADATA_CHANGES && |
10969 | (change == CH_MIGRATION || change == CH_TAKEOVER)) { | |
7a862a02 | 10970 | dprintf("imsm: Metadata changes rollback is not supported for migration and takeover operations.\n"); |
fbf3d202 AK |
10971 | change = -1; |
10972 | } | |
471bceb6 | 10973 | return change; |
694575e7 KW |
10974 | } |
10975 | ||
bb025c2f KW |
10976 | int imsm_takeover(struct supertype *st, struct geo_params *geo) |
10977 | { | |
10978 | struct intel_super *super = st->sb; | |
10979 | struct imsm_update_takeover *u; | |
10980 | ||
503975b9 | 10981 | u = xmalloc(sizeof(struct imsm_update_takeover)); |
bb025c2f KW |
10982 | |
10983 | u->type = update_takeover; | |
10984 | u->subarray = super->current_vol; | |
10985 | ||
10986 | /* 10->0 transition */ | |
10987 | if (geo->level == 0) | |
10988 | u->direction = R10_TO_R0; | |
10989 | ||
0529c688 KW |
10990 | /* 0->10 transition */ |
10991 | if (geo->level == 10) | |
10992 | u->direction = R0_TO_R10; | |
10993 | ||
bb025c2f KW |
10994 | /* update metadata locally */ |
10995 | imsm_update_metadata_locally(st, u, | |
10996 | sizeof(struct imsm_update_takeover)); | |
10997 | /* and possibly remotely */ | |
10998 | if (st->update_tail) | |
10999 | append_metadata_update(st, u, | |
11000 | sizeof(struct imsm_update_takeover)); | |
11001 | else | |
11002 | free(u); | |
11003 | ||
11004 | return 0; | |
11005 | } | |
11006 | ||
d04f65f4 N |
11007 | static int imsm_reshape_super(struct supertype *st, unsigned long long size, |
11008 | int level, | |
78b10e66 | 11009 | int layout, int chunksize, int raid_disks, |
41784c88 | 11010 | int delta_disks, char *backup, char *dev, |
016e00f5 | 11011 | int direction, int verbose) |
78b10e66 | 11012 | { |
78b10e66 N |
11013 | int ret_val = 1; |
11014 | struct geo_params geo; | |
11015 | ||
1ade5cc1 | 11016 | dprintf("(enter)\n"); |
78b10e66 | 11017 | |
71204a50 | 11018 | memset(&geo, 0, sizeof(struct geo_params)); |
78b10e66 N |
11019 | |
11020 | geo.dev_name = dev; | |
4dd2df09 | 11021 | strcpy(geo.devnm, st->devnm); |
78b10e66 N |
11022 | geo.size = size; |
11023 | geo.level = level; | |
11024 | geo.layout = layout; | |
11025 | geo.chunksize = chunksize; | |
11026 | geo.raid_disks = raid_disks; | |
41784c88 AK |
11027 | if (delta_disks != UnSet) |
11028 | geo.raid_disks += delta_disks; | |
78b10e66 | 11029 | |
1ade5cc1 N |
11030 | dprintf("for level : %i\n", geo.level); |
11031 | dprintf("for raid_disks : %i\n", geo.raid_disks); | |
78b10e66 N |
11032 | |
11033 | if (experimental() == 0) | |
11034 | return ret_val; | |
11035 | ||
4dd2df09 | 11036 | if (strcmp(st->container_devnm, st->devnm) == 0) { |
694575e7 KW |
11037 | /* On container level we can only increase number of devices. */ |
11038 | dprintf("imsm: info: Container operation\n"); | |
78b10e66 | 11039 | int old_raid_disks = 0; |
6dc0be30 | 11040 | |
78b10e66 | 11041 | if (imsm_reshape_is_allowed_on_container( |
fbf3d202 | 11042 | st, &geo, &old_raid_disks, direction)) { |
78b10e66 N |
11043 | struct imsm_update_reshape *u = NULL; |
11044 | int len; | |
11045 | ||
11046 | len = imsm_create_metadata_update_for_reshape( | |
11047 | st, &geo, old_raid_disks, &u); | |
11048 | ||
ed08d51c AK |
11049 | if (len <= 0) { |
11050 | dprintf("imsm: Cannot prepare update\n"); | |
11051 | goto exit_imsm_reshape_super; | |
11052 | } | |
11053 | ||
8dd70bce AK |
11054 | ret_val = 0; |
11055 | /* update metadata locally */ | |
11056 | imsm_update_metadata_locally(st, u, len); | |
11057 | /* and possibly remotely */ | |
11058 | if (st->update_tail) | |
11059 | append_metadata_update(st, u, len); | |
11060 | else | |
ed08d51c | 11061 | free(u); |
8dd70bce | 11062 | |
694575e7 | 11063 | } else { |
7a862a02 | 11064 | pr_err("(imsm) Operation is not allowed on this container\n"); |
694575e7 KW |
11065 | } |
11066 | } else { | |
11067 | /* On volume level we support following operations | |
471bceb6 KW |
11068 | * - takeover: raid10 -> raid0; raid0 -> raid10 |
11069 | * - chunk size migration | |
11070 | * - migration: raid5 -> raid0; raid0 -> raid5 | |
11071 | */ | |
11072 | struct intel_super *super = st->sb; | |
11073 | struct intel_dev *dev = super->devlist; | |
4dd2df09 | 11074 | int change; |
694575e7 | 11075 | dprintf("imsm: info: Volume operation\n"); |
471bceb6 KW |
11076 | /* find requested device */ |
11077 | while (dev) { | |
1011e834 | 11078 | char *devnm = |
4dd2df09 N |
11079 | imsm_find_array_devnm_by_subdev( |
11080 | dev->index, st->container_devnm); | |
11081 | if (devnm && strcmp(devnm, geo.devnm) == 0) | |
471bceb6 KW |
11082 | break; |
11083 | dev = dev->next; | |
11084 | } | |
11085 | if (dev == NULL) { | |
4dd2df09 N |
11086 | pr_err("Cannot find %s (%s) subarray\n", |
11087 | geo.dev_name, geo.devnm); | |
471bceb6 KW |
11088 | goto exit_imsm_reshape_super; |
11089 | } | |
11090 | super->current_vol = dev->index; | |
fbf3d202 | 11091 | change = imsm_analyze_change(st, &geo, direction); |
694575e7 | 11092 | switch (change) { |
471bceb6 | 11093 | case CH_TAKEOVER: |
bb025c2f | 11094 | ret_val = imsm_takeover(st, &geo); |
694575e7 | 11095 | break; |
48c5303a PC |
11096 | case CH_MIGRATION: { |
11097 | struct imsm_update_reshape_migration *u = NULL; | |
11098 | int len = | |
11099 | imsm_create_metadata_update_for_migration( | |
11100 | st, &geo, &u); | |
11101 | if (len < 1) { | |
7a862a02 | 11102 | dprintf("imsm: Cannot prepare update\n"); |
48c5303a PC |
11103 | break; |
11104 | } | |
471bceb6 | 11105 | ret_val = 0; |
48c5303a PC |
11106 | /* update metadata locally */ |
11107 | imsm_update_metadata_locally(st, u, len); | |
11108 | /* and possibly remotely */ | |
11109 | if (st->update_tail) | |
11110 | append_metadata_update(st, u, len); | |
11111 | else | |
11112 | free(u); | |
11113 | } | |
11114 | break; | |
7abc9871 | 11115 | case CH_ARRAY_SIZE: { |
f3871fdc AK |
11116 | struct imsm_update_size_change *u = NULL; |
11117 | int len = | |
11118 | imsm_create_metadata_update_for_size_change( | |
11119 | st, &geo, &u); | |
11120 | if (len < 1) { | |
7a862a02 | 11121 | dprintf("imsm: Cannot prepare update\n"); |
f3871fdc AK |
11122 | break; |
11123 | } | |
11124 | ret_val = 0; | |
11125 | /* update metadata locally */ | |
11126 | imsm_update_metadata_locally(st, u, len); | |
11127 | /* and possibly remotely */ | |
11128 | if (st->update_tail) | |
11129 | append_metadata_update(st, u, len); | |
11130 | else | |
11131 | free(u); | |
7abc9871 AK |
11132 | } |
11133 | break; | |
471bceb6 KW |
11134 | default: |
11135 | ret_val = 1; | |
694575e7 | 11136 | } |
694575e7 | 11137 | } |
78b10e66 | 11138 | |
ed08d51c | 11139 | exit_imsm_reshape_super: |
78b10e66 N |
11140 | dprintf("imsm: reshape_super Exit code = %i\n", ret_val); |
11141 | return ret_val; | |
11142 | } | |
2cda7640 | 11143 | |
0febb20c AO |
11144 | #define COMPLETED_OK 0 |
11145 | #define COMPLETED_NONE 1 | |
11146 | #define COMPLETED_DELAYED 2 | |
11147 | ||
11148 | static int read_completed(int fd, unsigned long long *val) | |
11149 | { | |
11150 | int ret; | |
11151 | char buf[50]; | |
11152 | ||
11153 | ret = sysfs_fd_get_str(fd, buf, 50); | |
11154 | if (ret < 0) | |
11155 | return ret; | |
11156 | ||
11157 | ret = COMPLETED_OK; | |
11158 | if (strncmp(buf, "none", 4) == 0) { | |
11159 | ret = COMPLETED_NONE; | |
11160 | } else if (strncmp(buf, "delayed", 7) == 0) { | |
11161 | ret = COMPLETED_DELAYED; | |
11162 | } else { | |
11163 | char *ep; | |
11164 | *val = strtoull(buf, &ep, 0); | |
11165 | if (ep == buf || (*ep != 0 && *ep != '\n' && *ep != ' ')) | |
11166 | ret = -1; | |
11167 | } | |
11168 | return ret; | |
11169 | } | |
11170 | ||
eee67a47 AK |
11171 | /******************************************************************************* |
11172 | * Function: wait_for_reshape_imsm | |
11173 | * Description: Function writes new sync_max value and waits until | |
11174 | * reshape process reach new position | |
11175 | * Parameters: | |
11176 | * sra : general array info | |
eee67a47 AK |
11177 | * ndata : number of disks in new array's layout |
11178 | * Returns: | |
11179 | * 0 : success, | |
11180 | * 1 : there is no reshape in progress, | |
11181 | * -1 : fail | |
11182 | ******************************************************************************/ | |
ae9f01f8 | 11183 | int wait_for_reshape_imsm(struct mdinfo *sra, int ndata) |
eee67a47 | 11184 | { |
85ca499c | 11185 | int fd = sysfs_get_fd(sra, NULL, "sync_completed"); |
df2647fa | 11186 | int retry = 3; |
eee67a47 | 11187 | unsigned long long completed; |
ae9f01f8 AK |
11188 | /* to_complete : new sync_max position */ |
11189 | unsigned long long to_complete = sra->reshape_progress; | |
11190 | unsigned long long position_to_set = to_complete / ndata; | |
eee67a47 | 11191 | |
ae9f01f8 | 11192 | if (fd < 0) { |
1ade5cc1 | 11193 | dprintf("cannot open reshape_position\n"); |
eee67a47 | 11194 | return 1; |
ae9f01f8 | 11195 | } |
eee67a47 | 11196 | |
df2647fa PB |
11197 | do { |
11198 | if (sysfs_fd_get_ll(fd, &completed) < 0) { | |
11199 | if (!retry) { | |
11200 | dprintf("cannot read reshape_position (no reshape in progres)\n"); | |
11201 | close(fd); | |
11202 | return 1; | |
11203 | } | |
11204 | usleep(30000); | |
11205 | } else | |
11206 | break; | |
11207 | } while (retry--); | |
eee67a47 | 11208 | |
85ca499c | 11209 | if (completed > position_to_set) { |
1ade5cc1 | 11210 | dprintf("wrong next position to set %llu (%llu)\n", |
85ca499c | 11211 | to_complete, position_to_set); |
ae9f01f8 AK |
11212 | close(fd); |
11213 | return -1; | |
11214 | } | |
11215 | dprintf("Position set: %llu\n", position_to_set); | |
11216 | if (sysfs_set_num(sra, NULL, "sync_max", | |
11217 | position_to_set) != 0) { | |
1ade5cc1 | 11218 | dprintf("cannot set reshape position to %llu\n", |
ae9f01f8 AK |
11219 | position_to_set); |
11220 | close(fd); | |
11221 | return -1; | |
eee67a47 AK |
11222 | } |
11223 | ||
eee67a47 | 11224 | do { |
0febb20c | 11225 | int rc; |
eee67a47 | 11226 | char action[20]; |
5ff3a780 | 11227 | int timeout = 3000; |
0febb20c | 11228 | |
5ff3a780 | 11229 | sysfs_wait(fd, &timeout); |
a47e44fb AK |
11230 | if (sysfs_get_str(sra, NULL, "sync_action", |
11231 | action, 20) > 0 && | |
d7d3809a | 11232 | strncmp(action, "reshape", 7) != 0) { |
b2be2b62 AO |
11233 | if (strncmp(action, "idle", 4) == 0) |
11234 | break; | |
d7d3809a AP |
11235 | close(fd); |
11236 | return -1; | |
11237 | } | |
0febb20c AO |
11238 | |
11239 | rc = read_completed(fd, &completed); | |
11240 | if (rc < 0) { | |
1ade5cc1 | 11241 | dprintf("cannot read reshape_position (in loop)\n"); |
eee67a47 AK |
11242 | close(fd); |
11243 | return 1; | |
0febb20c AO |
11244 | } else if (rc == COMPLETED_NONE) |
11245 | break; | |
85ca499c | 11246 | } while (completed < position_to_set); |
b2be2b62 | 11247 | |
eee67a47 AK |
11248 | close(fd); |
11249 | return 0; | |
eee67a47 AK |
11250 | } |
11251 | ||
b915c95f AK |
11252 | /******************************************************************************* |
11253 | * Function: check_degradation_change | |
11254 | * Description: Check that array hasn't become failed. | |
11255 | * Parameters: | |
11256 | * info : for sysfs access | |
11257 | * sources : source disks descriptors | |
11258 | * degraded: previous degradation level | |
11259 | * Returns: | |
11260 | * degradation level | |
11261 | ******************************************************************************/ | |
11262 | int check_degradation_change(struct mdinfo *info, | |
11263 | int *sources, | |
11264 | int degraded) | |
11265 | { | |
11266 | unsigned long long new_degraded; | |
e1993023 LD |
11267 | int rv; |
11268 | ||
11269 | rv = sysfs_get_ll(info, NULL, "degraded", &new_degraded); | |
089f9d79 | 11270 | if (rv == -1 || (new_degraded != (unsigned long long)degraded)) { |
b915c95f AK |
11271 | /* check each device to ensure it is still working */ |
11272 | struct mdinfo *sd; | |
11273 | new_degraded = 0; | |
11274 | for (sd = info->devs ; sd ; sd = sd->next) { | |
11275 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) | |
11276 | continue; | |
11277 | if (sd->disk.state & (1<<MD_DISK_SYNC)) { | |
cf52eff5 TM |
11278 | char sbuf[100]; |
11279 | ||
b915c95f | 11280 | if (sysfs_get_str(info, |
cf52eff5 | 11281 | sd, "state", sbuf, sizeof(sbuf)) < 0 || |
b915c95f AK |
11282 | strstr(sbuf, "faulty") || |
11283 | strstr(sbuf, "in_sync") == NULL) { | |
11284 | /* this device is dead */ | |
11285 | sd->disk.state = (1<<MD_DISK_FAULTY); | |
11286 | if (sd->disk.raid_disk >= 0 && | |
11287 | sources[sd->disk.raid_disk] >= 0) { | |
11288 | close(sources[ | |
11289 | sd->disk.raid_disk]); | |
11290 | sources[sd->disk.raid_disk] = | |
11291 | -1; | |
11292 | } | |
11293 | new_degraded++; | |
11294 | } | |
11295 | } | |
11296 | } | |
11297 | } | |
11298 | ||
11299 | return new_degraded; | |
11300 | } | |
11301 | ||
10f22854 AK |
11302 | /******************************************************************************* |
11303 | * Function: imsm_manage_reshape | |
11304 | * Description: Function finds array under reshape and it manages reshape | |
11305 | * process. It creates stripes backups (if required) and sets | |
942e1cdb | 11306 | * checkpoints. |
10f22854 AK |
11307 | * Parameters: |
11308 | * afd : Backup handle (nattive) - not used | |
11309 | * sra : general array info | |
11310 | * reshape : reshape parameters - not used | |
11311 | * st : supertype structure | |
11312 | * blocks : size of critical section [blocks] | |
11313 | * fds : table of source device descriptor | |
11314 | * offsets : start of array (offest per devices) | |
11315 | * dests : not used | |
11316 | * destfd : table of destination device descriptor | |
11317 | * destoffsets : table of destination offsets (per device) | |
11318 | * Returns: | |
11319 | * 1 : success, reshape is done | |
11320 | * 0 : fail | |
11321 | ******************************************************************************/ | |
999b4972 N |
11322 | static int imsm_manage_reshape( |
11323 | int afd, struct mdinfo *sra, struct reshape *reshape, | |
10f22854 | 11324 | struct supertype *st, unsigned long backup_blocks, |
999b4972 N |
11325 | int *fds, unsigned long long *offsets, |
11326 | int dests, int *destfd, unsigned long long *destoffsets) | |
11327 | { | |
10f22854 AK |
11328 | int ret_val = 0; |
11329 | struct intel_super *super = st->sb; | |
594dc1b8 | 11330 | struct intel_dev *dv; |
de44e46f | 11331 | unsigned int sector_size = super->sector_size; |
10f22854 | 11332 | struct imsm_dev *dev = NULL; |
a6b6d984 | 11333 | struct imsm_map *map_src; |
10f22854 AK |
11334 | int migr_vol_qan = 0; |
11335 | int ndata, odata; /* [bytes] */ | |
11336 | int chunk; /* [bytes] */ | |
11337 | struct migr_record *migr_rec; | |
11338 | char *buf = NULL; | |
11339 | unsigned int buf_size; /* [bytes] */ | |
11340 | unsigned long long max_position; /* array size [bytes] */ | |
11341 | unsigned long long next_step; /* [blocks]/[bytes] */ | |
11342 | unsigned long long old_data_stripe_length; | |
10f22854 AK |
11343 | unsigned long long start_src; /* [bytes] */ |
11344 | unsigned long long start; /* [bytes] */ | |
11345 | unsigned long long start_buf_shift; /* [bytes] */ | |
b915c95f | 11346 | int degraded = 0; |
ab724b98 | 11347 | int source_layout = 0; |
10f22854 | 11348 | |
79a16a9b JS |
11349 | if (!sra) |
11350 | return ret_val; | |
11351 | ||
11352 | if (!fds || !offsets) | |
10f22854 AK |
11353 | goto abort; |
11354 | ||
11355 | /* Find volume during the reshape */ | |
11356 | for (dv = super->devlist; dv; dv = dv->next) { | |
11357 | if (dv->dev->vol.migr_type == MIGR_GEN_MIGR | |
11358 | && dv->dev->vol.migr_state == 1) { | |
11359 | dev = dv->dev; | |
11360 | migr_vol_qan++; | |
11361 | } | |
11362 | } | |
11363 | /* Only one volume can migrate at the same time */ | |
11364 | if (migr_vol_qan != 1) { | |
676e87a8 | 11365 | pr_err("%s", migr_vol_qan ? |
10f22854 AK |
11366 | "Number of migrating volumes greater than 1\n" : |
11367 | "There is no volume during migrationg\n"); | |
11368 | goto abort; | |
11369 | } | |
11370 | ||
238c0a71 | 11371 | map_src = get_imsm_map(dev, MAP_1); |
10f22854 AK |
11372 | if (map_src == NULL) |
11373 | goto abort; | |
10f22854 | 11374 | |
238c0a71 AK |
11375 | ndata = imsm_num_data_members(dev, MAP_0); |
11376 | odata = imsm_num_data_members(dev, MAP_1); | |
10f22854 | 11377 | |
7b1ab482 | 11378 | chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512; |
10f22854 AK |
11379 | old_data_stripe_length = odata * chunk; |
11380 | ||
11381 | migr_rec = super->migr_rec; | |
11382 | ||
10f22854 AK |
11383 | /* initialize migration record for start condition */ |
11384 | if (sra->reshape_progress == 0) | |
11385 | init_migr_record_imsm(st, dev, sra); | |
b2c59438 AK |
11386 | else { |
11387 | if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) { | |
7a862a02 | 11388 | dprintf("imsm: cannot restart migration when data are present in copy area.\n"); |
b2c59438 AK |
11389 | goto abort; |
11390 | } | |
6a75c8ca AK |
11391 | /* Save checkpoint to update migration record for current |
11392 | * reshape position (in md). It can be farther than current | |
11393 | * reshape position in metadata. | |
11394 | */ | |
11395 | if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) { | |
11396 | /* ignore error == 2, this can mean end of reshape here | |
11397 | */ | |
7a862a02 | 11398 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL, initial save)\n"); |
6a75c8ca AK |
11399 | goto abort; |
11400 | } | |
b2c59438 | 11401 | } |
10f22854 AK |
11402 | |
11403 | /* size for data */ | |
11404 | buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512; | |
11405 | /* extend buffer size for parity disk */ | |
11406 | buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
11407 | /* add space for stripe aligment */ | |
11408 | buf_size += old_data_stripe_length; | |
de44e46f PB |
11409 | if (posix_memalign((void **)&buf, MAX_SECTOR_SIZE, buf_size)) { |
11410 | dprintf("imsm: Cannot allocate checkpoint buffer\n"); | |
10f22854 AK |
11411 | goto abort; |
11412 | } | |
11413 | ||
3ef4403c | 11414 | max_position = sra->component_size * ndata; |
68eb8bc6 | 11415 | source_layout = imsm_level_to_layout(map_src->raid_level); |
10f22854 AK |
11416 | |
11417 | while (__le32_to_cpu(migr_rec->curr_migr_unit) < | |
11418 | __le32_to_cpu(migr_rec->num_migr_units)) { | |
11419 | /* current reshape position [blocks] */ | |
11420 | unsigned long long current_position = | |
11421 | __le32_to_cpu(migr_rec->blocks_per_unit) | |
11422 | * __le32_to_cpu(migr_rec->curr_migr_unit); | |
11423 | unsigned long long border; | |
11424 | ||
b915c95f AK |
11425 | /* Check that array hasn't become failed. |
11426 | */ | |
11427 | degraded = check_degradation_change(sra, fds, degraded); | |
11428 | if (degraded > 1) { | |
7a862a02 | 11429 | dprintf("imsm: Abort reshape due to degradation level (%i)\n", degraded); |
b915c95f AK |
11430 | goto abort; |
11431 | } | |
11432 | ||
10f22854 AK |
11433 | next_step = __le32_to_cpu(migr_rec->blocks_per_unit); |
11434 | ||
11435 | if ((current_position + next_step) > max_position) | |
11436 | next_step = max_position - current_position; | |
11437 | ||
92144abf | 11438 | start = current_position * 512; |
10f22854 | 11439 | |
942e1cdb | 11440 | /* align reading start to old geometry */ |
10f22854 AK |
11441 | start_buf_shift = start % old_data_stripe_length; |
11442 | start_src = start - start_buf_shift; | |
11443 | ||
11444 | border = (start_src / odata) - (start / ndata); | |
11445 | border /= 512; | |
11446 | if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) { | |
11447 | /* save critical stripes to buf | |
11448 | * start - start address of current unit | |
11449 | * to backup [bytes] | |
11450 | * start_src - start address of current unit | |
11451 | * to backup alligned to source array | |
11452 | * [bytes] | |
11453 | */ | |
594dc1b8 | 11454 | unsigned long long next_step_filler; |
10f22854 AK |
11455 | unsigned long long copy_length = next_step * 512; |
11456 | ||
11457 | /* allign copy area length to stripe in old geometry */ | |
11458 | next_step_filler = ((copy_length + start_buf_shift) | |
11459 | % old_data_stripe_length); | |
11460 | if (next_step_filler) | |
11461 | next_step_filler = (old_data_stripe_length | |
11462 | - next_step_filler); | |
7a862a02 | 11463 | dprintf("save_stripes() parameters: start = %llu,\tstart_src = %llu,\tnext_step*512 = %llu,\tstart_in_buf_shift = %llu,\tnext_step_filler = %llu\n", |
10f22854 AK |
11464 | start, start_src, copy_length, |
11465 | start_buf_shift, next_step_filler); | |
11466 | ||
11467 | if (save_stripes(fds, offsets, map_src->num_members, | |
ab724b98 AK |
11468 | chunk, map_src->raid_level, |
11469 | source_layout, 0, NULL, start_src, | |
10f22854 AK |
11470 | copy_length + |
11471 | next_step_filler + start_buf_shift, | |
11472 | buf)) { | |
7a862a02 | 11473 | dprintf("imsm: Cannot save stripes to buffer\n"); |
10f22854 AK |
11474 | goto abort; |
11475 | } | |
11476 | /* Convert data to destination format and store it | |
11477 | * in backup general migration area | |
11478 | */ | |
11479 | if (save_backup_imsm(st, dev, sra, | |
aea93171 | 11480 | buf + start_buf_shift, copy_length)) { |
7a862a02 | 11481 | dprintf("imsm: Cannot save stripes to target devices\n"); |
10f22854 AK |
11482 | goto abort; |
11483 | } | |
11484 | if (save_checkpoint_imsm(st, sra, | |
11485 | UNIT_SRC_IN_CP_AREA)) { | |
7a862a02 | 11486 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_IN_CP_AREA)\n"); |
10f22854 AK |
11487 | goto abort; |
11488 | } | |
8016a6d4 AK |
11489 | } else { |
11490 | /* set next step to use whole border area */ | |
11491 | border /= next_step; | |
11492 | if (border > 1) | |
11493 | next_step *= border; | |
10f22854 AK |
11494 | } |
11495 | /* When data backed up, checkpoint stored, | |
11496 | * kick the kernel to reshape unit of data | |
11497 | */ | |
11498 | next_step = next_step + sra->reshape_progress; | |
8016a6d4 AK |
11499 | /* limit next step to array max position */ |
11500 | if (next_step > max_position) | |
11501 | next_step = max_position; | |
10f22854 AK |
11502 | sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress); |
11503 | sysfs_set_num(sra, NULL, "suspend_hi", next_step); | |
ae9f01f8 | 11504 | sra->reshape_progress = next_step; |
10f22854 AK |
11505 | |
11506 | /* wait until reshape finish */ | |
c85338c6 | 11507 | if (wait_for_reshape_imsm(sra, ndata)) { |
c47b0ff6 AK |
11508 | dprintf("wait_for_reshape_imsm returned error!\n"); |
11509 | goto abort; | |
11510 | } | |
84d11e6c N |
11511 | if (sigterm) |
11512 | goto abort; | |
10f22854 | 11513 | |
0228d92c AK |
11514 | if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) { |
11515 | /* ignore error == 2, this can mean end of reshape here | |
11516 | */ | |
7a862a02 | 11517 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL)\n"); |
10f22854 AK |
11518 | goto abort; |
11519 | } | |
11520 | ||
11521 | } | |
11522 | ||
71e5411e PB |
11523 | /* clear migr_rec on disks after successful migration */ |
11524 | struct dl *d; | |
11525 | ||
de44e46f | 11526 | memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*sector_size); |
71e5411e PB |
11527 | for (d = super->disks; d; d = d->next) { |
11528 | if (d->index < 0 || is_failed(&d->disk)) | |
11529 | continue; | |
11530 | unsigned long long dsize; | |
11531 | ||
11532 | get_dev_size(d->fd, NULL, &dsize); | |
de44e46f | 11533 | if (lseek64(d->fd, dsize - MIGR_REC_SECTOR_POSITION*sector_size, |
71e5411e | 11534 | SEEK_SET) >= 0) { |
466070ad | 11535 | if ((unsigned int)write(d->fd, super->migr_rec_buf, |
de44e46f PB |
11536 | MIGR_REC_BUF_SECTORS*sector_size) != |
11537 | MIGR_REC_BUF_SECTORS*sector_size) | |
71e5411e PB |
11538 | perror("Write migr_rec failed"); |
11539 | } | |
11540 | } | |
11541 | ||
10f22854 AK |
11542 | /* return '1' if done */ |
11543 | ret_val = 1; | |
11544 | abort: | |
11545 | free(buf); | |
942e1cdb N |
11546 | /* See Grow.c: abort_reshape() for further explanation */ |
11547 | sysfs_set_num(sra, NULL, "suspend_lo", 0x7FFFFFFFFFFFFFFFULL); | |
11548 | sysfs_set_num(sra, NULL, "suspend_hi", 0); | |
11549 | sysfs_set_num(sra, NULL, "suspend_lo", 0); | |
10f22854 AK |
11550 | |
11551 | return ret_val; | |
999b4972 | 11552 | } |
0c21b485 | 11553 | |
71204a50 | 11554 | #endif /* MDASSEMBLE */ |
999b4972 | 11555 | |
cdddbdbc DW |
11556 | struct superswitch super_imsm = { |
11557 | #ifndef MDASSEMBLE | |
11558 | .examine_super = examine_super_imsm, | |
11559 | .brief_examine_super = brief_examine_super_imsm, | |
4737ae25 | 11560 | .brief_examine_subarrays = brief_examine_subarrays_imsm, |
9d84c8ea | 11561 | .export_examine_super = export_examine_super_imsm, |
cdddbdbc DW |
11562 | .detail_super = detail_super_imsm, |
11563 | .brief_detail_super = brief_detail_super_imsm, | |
bf5a934a | 11564 | .write_init_super = write_init_super_imsm, |
0e600426 N |
11565 | .validate_geometry = validate_geometry_imsm, |
11566 | .add_to_super = add_to_super_imsm, | |
1a64be56 | 11567 | .remove_from_super = remove_from_super_imsm, |
d665cc31 | 11568 | .detail_platform = detail_platform_imsm, |
e50cf220 | 11569 | .export_detail_platform = export_detail_platform_imsm, |
33414a01 | 11570 | .kill_subarray = kill_subarray_imsm, |
aa534678 | 11571 | .update_subarray = update_subarray_imsm, |
2b959fbf | 11572 | .load_container = load_container_imsm, |
71204a50 N |
11573 | .default_geometry = default_geometry_imsm, |
11574 | .get_disk_controller_domain = imsm_get_disk_controller_domain, | |
11575 | .reshape_super = imsm_reshape_super, | |
11576 | .manage_reshape = imsm_manage_reshape, | |
9e2d750d | 11577 | .recover_backup = recover_backup_imsm, |
74db60b0 | 11578 | .copy_metadata = copy_metadata_imsm, |
27156a57 | 11579 | .examine_badblocks = examine_badblocks_imsm, |
cdddbdbc DW |
11580 | #endif |
11581 | .match_home = match_home_imsm, | |
11582 | .uuid_from_super= uuid_from_super_imsm, | |
11583 | .getinfo_super = getinfo_super_imsm, | |
5c4cd5da | 11584 | .getinfo_super_disks = getinfo_super_disks_imsm, |
cdddbdbc DW |
11585 | .update_super = update_super_imsm, |
11586 | ||
11587 | .avail_size = avail_size_imsm, | |
80e7f8c3 | 11588 | .min_acceptable_spare_size = min_acceptable_spare_size_imsm, |
cdddbdbc DW |
11589 | |
11590 | .compare_super = compare_super_imsm, | |
11591 | ||
11592 | .load_super = load_super_imsm, | |
bf5a934a | 11593 | .init_super = init_super_imsm, |
e683ca88 | 11594 | .store_super = store_super_imsm, |
cdddbdbc DW |
11595 | .free_super = free_super_imsm, |
11596 | .match_metadata_desc = match_metadata_desc_imsm, | |
bf5a934a | 11597 | .container_content = container_content_imsm, |
0c21b485 | 11598 | .validate_container = validate_container_imsm, |
cdddbdbc | 11599 | |
cdddbdbc | 11600 | .external = 1, |
4cce4069 | 11601 | .name = "imsm", |
845dea95 | 11602 | |
0e600426 | 11603 | #ifndef MDASSEMBLE |
845dea95 NB |
11604 | /* for mdmon */ |
11605 | .open_new = imsm_open_new, | |
ed9d66aa | 11606 | .set_array_state= imsm_set_array_state, |
845dea95 NB |
11607 | .set_disk = imsm_set_disk, |
11608 | .sync_metadata = imsm_sync_metadata, | |
88758e9d | 11609 | .activate_spare = imsm_activate_spare, |
e8319a19 | 11610 | .process_update = imsm_process_update, |
8273f55e | 11611 | .prepare_update = imsm_prepare_update, |
6f50473f | 11612 | .record_bad_block = imsm_record_badblock, |
c07a5a4f | 11613 | .clear_bad_block = imsm_clear_badblock, |
928f1424 | 11614 | .get_bad_blocks = imsm_get_badblocks, |
0e600426 | 11615 | #endif /* MDASSEMBLE */ |
cdddbdbc | 11616 | }; |