]>
Commit | Line | Data |
---|---|---|
cdddbdbc DW |
1 | /* |
2 | * mdadm - Intel(R) Matrix Storage Manager Support | |
3 | * | |
a54d5262 | 4 | * Copyright (C) 2002-2008 Intel Corporation |
cdddbdbc DW |
5 | * |
6 | * This program is free software; you can redistribute it and/or modify it | |
7 | * under the terms and conditions of the GNU General Public License, | |
8 | * version 2, as published by the Free Software Foundation. | |
9 | * | |
10 | * This program is distributed in the hope it will be useful, but WITHOUT | |
11 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
12 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for | |
13 | * more details. | |
14 | * | |
15 | * You should have received a copy of the GNU General Public License along with | |
16 | * this program; if not, write to the Free Software Foundation, Inc., | |
17 | * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA. | |
18 | */ | |
19 | ||
51006d85 | 20 | #define HAVE_STDINT_H 1 |
cdddbdbc | 21 | #include "mdadm.h" |
c2a1e7da | 22 | #include "mdmon.h" |
51006d85 | 23 | #include "sha1.h" |
88c32bb1 | 24 | #include "platform-intel.h" |
cdddbdbc DW |
25 | #include <values.h> |
26 | #include <scsi/sg.h> | |
27 | #include <ctype.h> | |
d665cc31 | 28 | #include <dirent.h> |
cdddbdbc DW |
29 | |
30 | /* MPB == Metadata Parameter Block */ | |
31 | #define MPB_SIGNATURE "Intel Raid ISM Cfg Sig. " | |
32 | #define MPB_SIG_LEN (strlen(MPB_SIGNATURE)) | |
33 | #define MPB_VERSION_RAID0 "1.0.00" | |
34 | #define MPB_VERSION_RAID1 "1.1.00" | |
fe7ed8cb DW |
35 | #define MPB_VERSION_MANY_VOLUMES_PER_ARRAY "1.2.00" |
36 | #define MPB_VERSION_3OR4_DISK_ARRAY "1.2.01" | |
cdddbdbc | 37 | #define MPB_VERSION_RAID5 "1.2.02" |
fe7ed8cb DW |
38 | #define MPB_VERSION_5OR6_DISK_ARRAY "1.2.04" |
39 | #define MPB_VERSION_CNG "1.2.06" | |
40 | #define MPB_VERSION_ATTRIBS "1.3.00" | |
cdddbdbc DW |
41 | #define MAX_SIGNATURE_LENGTH 32 |
42 | #define MAX_RAID_SERIAL_LEN 16 | |
fe7ed8cb | 43 | |
19482bcc AK |
44 | /* supports RAID0 */ |
45 | #define MPB_ATTRIB_RAID0 __cpu_to_le32(0x00000001) | |
46 | /* supports RAID1 */ | |
47 | #define MPB_ATTRIB_RAID1 __cpu_to_le32(0x00000002) | |
48 | /* supports RAID10 */ | |
49 | #define MPB_ATTRIB_RAID10 __cpu_to_le32(0x00000004) | |
50 | /* supports RAID1E */ | |
51 | #define MPB_ATTRIB_RAID1E __cpu_to_le32(0x00000008) | |
52 | /* supports RAID5 */ | |
53 | #define MPB_ATTRIB_RAID5 __cpu_to_le32(0x00000010) | |
54 | /* supports RAID CNG */ | |
55 | #define MPB_ATTRIB_RAIDCNG __cpu_to_le32(0x00000020) | |
56 | /* supports expanded stripe sizes of 256K, 512K and 1MB */ | |
57 | #define MPB_ATTRIB_EXP_STRIPE_SIZE __cpu_to_le32(0x00000040) | |
58 | ||
59 | /* The OROM Support RST Caching of Volumes */ | |
60 | #define MPB_ATTRIB_NVM __cpu_to_le32(0x02000000) | |
61 | /* The OROM supports creating disks greater than 2TB */ | |
62 | #define MPB_ATTRIB_2TB_DISK __cpu_to_le32(0x04000000) | |
63 | /* The OROM supports Bad Block Management */ | |
64 | #define MPB_ATTRIB_BBM __cpu_to_le32(0x08000000) | |
65 | ||
66 | /* THe OROM Supports NVM Caching of Volumes */ | |
67 | #define MPB_ATTRIB_NEVER_USE2 __cpu_to_le32(0x10000000) | |
68 | /* The OROM supports creating volumes greater than 2TB */ | |
69 | #define MPB_ATTRIB_2TB __cpu_to_le32(0x20000000) | |
70 | /* originally for PMP, now it's wasted b/c. Never use this bit! */ | |
71 | #define MPB_ATTRIB_NEVER_USE __cpu_to_le32(0x40000000) | |
72 | /* Verify MPB contents against checksum after reading MPB */ | |
73 | #define MPB_ATTRIB_CHECKSUM_VERIFY __cpu_to_le32(0x80000000) | |
74 | ||
75 | /* Define all supported attributes that have to be accepted by mdadm | |
76 | */ | |
418f9b36 | 77 | #define MPB_ATTRIB_SUPPORTED (MPB_ATTRIB_CHECKSUM_VERIFY | \ |
19482bcc AK |
78 | MPB_ATTRIB_2TB | \ |
79 | MPB_ATTRIB_2TB_DISK | \ | |
80 | MPB_ATTRIB_RAID0 | \ | |
81 | MPB_ATTRIB_RAID1 | \ | |
82 | MPB_ATTRIB_RAID10 | \ | |
83 | MPB_ATTRIB_RAID5 | \ | |
418f9b36 N |
84 | MPB_ATTRIB_EXP_STRIPE_SIZE) |
85 | ||
86 | /* Define attributes that are unused but not harmful */ | |
87 | #define MPB_ATTRIB_IGNORED (MPB_ATTRIB_NEVER_USE) | |
fe7ed8cb | 88 | |
8e59f3d8 | 89 | #define MPB_SECTOR_CNT 2210 |
c2c087e6 | 90 | #define IMSM_RESERVED_SECTORS 4096 |
b81221b7 | 91 | #define NUM_BLOCKS_DIRTY_STRIPE_REGION 2056 |
979d38be | 92 | #define SECT_PER_MB_SHIFT 11 |
cdddbdbc DW |
93 | |
94 | /* Disk configuration info. */ | |
95 | #define IMSM_MAX_DEVICES 255 | |
96 | struct imsm_disk { | |
97 | __u8 serial[MAX_RAID_SERIAL_LEN];/* 0xD8 - 0xE7 ascii serial number */ | |
5551b113 | 98 | __u32 total_blocks_lo; /* 0xE8 - 0xEB total blocks lo */ |
cdddbdbc | 99 | __u32 scsi_id; /* 0xEC - 0xEF scsi ID */ |
f2f27e63 DW |
100 | #define SPARE_DISK __cpu_to_le32(0x01) /* Spare */ |
101 | #define CONFIGURED_DISK __cpu_to_le32(0x02) /* Member of some RaidDev */ | |
102 | #define FAILED_DISK __cpu_to_le32(0x04) /* Permanent failure */ | |
cdddbdbc | 103 | __u32 status; /* 0xF0 - 0xF3 */ |
1011e834 | 104 | __u32 owner_cfg_num; /* which config 0,1,2... owns this disk */ |
5551b113 CA |
105 | __u32 total_blocks_hi; /* 0xF4 - 0xF5 total blocks hi */ |
106 | #define IMSM_DISK_FILLERS 3 | |
107 | __u32 filler[IMSM_DISK_FILLERS]; /* 0xF5 - 0x107 MPB_DISK_FILLERS for future expansion */ | |
cdddbdbc DW |
108 | }; |
109 | ||
3b451610 AK |
110 | /* map selector for map managment |
111 | */ | |
238c0a71 AK |
112 | #define MAP_0 0 |
113 | #define MAP_1 1 | |
114 | #define MAP_X -1 | |
3b451610 | 115 | |
cdddbdbc DW |
116 | /* RAID map configuration infos. */ |
117 | struct imsm_map { | |
5551b113 CA |
118 | __u32 pba_of_lba0_lo; /* start address of partition */ |
119 | __u32 blocks_per_member_lo;/* blocks per member */ | |
120 | __u32 num_data_stripes_lo; /* number of data stripes */ | |
cdddbdbc DW |
121 | __u16 blocks_per_strip; |
122 | __u8 map_state; /* Normal, Uninitialized, Degraded, Failed */ | |
123 | #define IMSM_T_STATE_NORMAL 0 | |
124 | #define IMSM_T_STATE_UNINITIALIZED 1 | |
e3bba0e0 DW |
125 | #define IMSM_T_STATE_DEGRADED 2 |
126 | #define IMSM_T_STATE_FAILED 3 | |
cdddbdbc DW |
127 | __u8 raid_level; |
128 | #define IMSM_T_RAID0 0 | |
129 | #define IMSM_T_RAID1 1 | |
130 | #define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */ | |
131 | __u8 num_members; /* number of member disks */ | |
fe7ed8cb DW |
132 | __u8 num_domains; /* number of parity domains */ |
133 | __u8 failed_disk_num; /* valid only when state is degraded */ | |
252d23c0 | 134 | __u8 ddf; |
5551b113 CA |
135 | __u32 pba_of_lba0_hi; |
136 | __u32 blocks_per_member_hi; | |
137 | __u32 num_data_stripes_hi; | |
138 | __u32 filler[4]; /* expansion area */ | |
7eef0453 | 139 | #define IMSM_ORD_REBUILD (1 << 24) |
cdddbdbc | 140 | __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members], |
7eef0453 DW |
141 | * top byte contains some flags |
142 | */ | |
cdddbdbc DW |
143 | } __attribute__ ((packed)); |
144 | ||
145 | struct imsm_vol { | |
f8f603f1 | 146 | __u32 curr_migr_unit; |
fe7ed8cb | 147 | __u32 checkpoint_id; /* id to access curr_migr_unit */ |
cdddbdbc | 148 | __u8 migr_state; /* Normal or Migrating */ |
e3bba0e0 DW |
149 | #define MIGR_INIT 0 |
150 | #define MIGR_REBUILD 1 | |
151 | #define MIGR_VERIFY 2 /* analagous to echo check > sync_action */ | |
152 | #define MIGR_GEN_MIGR 3 | |
153 | #define MIGR_STATE_CHANGE 4 | |
1484e727 | 154 | #define MIGR_REPAIR 5 |
cdddbdbc DW |
155 | __u8 migr_type; /* Initializing, Rebuilding, ... */ |
156 | __u8 dirty; | |
fe7ed8cb DW |
157 | __u8 fs_state; /* fast-sync state for CnG (0xff == disabled) */ |
158 | __u16 verify_errors; /* number of mismatches */ | |
159 | __u16 bad_blocks; /* number of bad blocks during verify */ | |
160 | __u32 filler[4]; | |
cdddbdbc DW |
161 | struct imsm_map map[1]; |
162 | /* here comes another one if migr_state */ | |
163 | } __attribute__ ((packed)); | |
164 | ||
165 | struct imsm_dev { | |
fe7ed8cb | 166 | __u8 volume[MAX_RAID_SERIAL_LEN]; |
cdddbdbc DW |
167 | __u32 size_low; |
168 | __u32 size_high; | |
fe7ed8cb DW |
169 | #define DEV_BOOTABLE __cpu_to_le32(0x01) |
170 | #define DEV_BOOT_DEVICE __cpu_to_le32(0x02) | |
171 | #define DEV_READ_COALESCING __cpu_to_le32(0x04) | |
172 | #define DEV_WRITE_COALESCING __cpu_to_le32(0x08) | |
173 | #define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10) | |
174 | #define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20) | |
175 | #define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40) | |
176 | #define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80) | |
177 | #define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100) | |
178 | #define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200) | |
179 | #define DEV_CLONE_N_GO __cpu_to_le32(0x400) | |
180 | #define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800) | |
181 | #define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000) | |
cdddbdbc DW |
182 | __u32 status; /* Persistent RaidDev status */ |
183 | __u32 reserved_blocks; /* Reserved blocks at beginning of volume */ | |
fe7ed8cb DW |
184 | __u8 migr_priority; |
185 | __u8 num_sub_vols; | |
186 | __u8 tid; | |
187 | __u8 cng_master_disk; | |
188 | __u16 cache_policy; | |
189 | __u8 cng_state; | |
190 | __u8 cng_sub_state; | |
191 | #define IMSM_DEV_FILLERS 10 | |
cdddbdbc DW |
192 | __u32 filler[IMSM_DEV_FILLERS]; |
193 | struct imsm_vol vol; | |
194 | } __attribute__ ((packed)); | |
195 | ||
196 | struct imsm_super { | |
197 | __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */ | |
198 | __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */ | |
199 | __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */ | |
200 | __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */ | |
201 | __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */ | |
604b746f JD |
202 | __u32 error_log_size; /* 0x30 - 0x33 in bytes */ |
203 | __u32 attributes; /* 0x34 - 0x37 */ | |
cdddbdbc DW |
204 | __u8 num_disks; /* 0x38 Number of configured disks */ |
205 | __u8 num_raid_devs; /* 0x39 Number of configured volumes */ | |
604b746f JD |
206 | __u8 error_log_pos; /* 0x3A */ |
207 | __u8 fill[1]; /* 0x3B */ | |
208 | __u32 cache_size; /* 0x3c - 0x40 in mb */ | |
209 | __u32 orig_family_num; /* 0x40 - 0x43 original family num */ | |
210 | __u32 pwr_cycle_count; /* 0x44 - 0x47 simulated power cycle count for array */ | |
211 | __u32 bbm_log_size; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */ | |
212 | #define IMSM_FILLERS 35 | |
213 | __u32 filler[IMSM_FILLERS]; /* 0x4C - 0xD7 RAID_MPB_FILLERS */ | |
cdddbdbc DW |
214 | struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */ |
215 | /* here comes imsm_dev[num_raid_devs] */ | |
604b746f | 216 | /* here comes BBM logs */ |
cdddbdbc DW |
217 | } __attribute__ ((packed)); |
218 | ||
604b746f JD |
219 | #define BBM_LOG_MAX_ENTRIES 254 |
220 | ||
221 | struct bbm_log_entry { | |
222 | __u64 defective_block_start; | |
223 | #define UNREADABLE 0xFFFFFFFF | |
224 | __u32 spare_block_offset; | |
225 | __u16 remapped_marked_count; | |
226 | __u16 disk_ordinal; | |
227 | } __attribute__ ((__packed__)); | |
228 | ||
229 | struct bbm_log { | |
230 | __u32 signature; /* 0xABADB10C */ | |
231 | __u32 entry_count; | |
232 | __u32 reserved_spare_block_count; /* 0 */ | |
233 | __u32 reserved; /* 0xFFFF */ | |
234 | __u64 first_spare_lba; | |
235 | struct bbm_log_entry mapped_block_entries[BBM_LOG_MAX_ENTRIES]; | |
236 | } __attribute__ ((__packed__)); | |
237 | ||
cdddbdbc DW |
238 | #ifndef MDASSEMBLE |
239 | static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" }; | |
240 | #endif | |
241 | ||
8e59f3d8 AK |
242 | #define RAID_DISK_RESERVED_BLOCKS_IMSM_HI 2209 |
243 | ||
244 | #define GEN_MIGR_AREA_SIZE 2048 /* General Migration Copy Area size in blocks */ | |
245 | ||
17a4eaf9 AK |
246 | #define MIGR_REC_BUF_SIZE 512 /* size of migr_record i/o buffer */ |
247 | #define MIGR_REC_POSITION 512 /* migr_record position offset on disk, | |
248 | * MIGR_REC_BUF_SIZE <= MIGR_REC_POSITION | |
249 | */ | |
250 | ||
8e59f3d8 AK |
251 | #define UNIT_SRC_NORMAL 0 /* Source data for curr_migr_unit must |
252 | * be recovered using srcMap */ | |
253 | #define UNIT_SRC_IN_CP_AREA 1 /* Source data for curr_migr_unit has | |
254 | * already been migrated and must | |
255 | * be recovered from checkpoint area */ | |
256 | struct migr_record { | |
257 | __u32 rec_status; /* Status used to determine how to restart | |
258 | * migration in case it aborts | |
259 | * in some fashion */ | |
260 | __u32 curr_migr_unit; /* 0..numMigrUnits-1 */ | |
261 | __u32 family_num; /* Family number of MPB | |
262 | * containing the RaidDev | |
263 | * that is migrating */ | |
264 | __u32 ascending_migr; /* True if migrating in increasing | |
265 | * order of lbas */ | |
266 | __u32 blocks_per_unit; /* Num disk blocks per unit of operation */ | |
267 | __u32 dest_depth_per_unit; /* Num member blocks each destMap | |
268 | * member disk | |
269 | * advances per unit-of-operation */ | |
270 | __u32 ckpt_area_pba; /* Pba of first block of ckpt copy area */ | |
271 | __u32 dest_1st_member_lba; /* First member lba on first | |
272 | * stripe of destination */ | |
273 | __u32 num_migr_units; /* Total num migration units-of-op */ | |
274 | __u32 post_migr_vol_cap; /* Size of volume after | |
275 | * migration completes */ | |
276 | __u32 post_migr_vol_cap_hi; /* Expansion space for LBA64 */ | |
277 | __u32 ckpt_read_disk_num; /* Which member disk in destSubMap[0] the | |
278 | * migration ckpt record was read from | |
279 | * (for recovered migrations) */ | |
280 | } __attribute__ ((__packed__)); | |
281 | ||
ec50f7b6 LM |
282 | struct md_list { |
283 | /* usage marker: | |
284 | * 1: load metadata | |
285 | * 2: metadata does not match | |
286 | * 4: already checked | |
287 | */ | |
288 | int used; | |
289 | char *devname; | |
290 | int found; | |
291 | int container; | |
292 | dev_t st_rdev; | |
293 | struct md_list *next; | |
294 | }; | |
295 | ||
e7b84f9d | 296 | #define pr_vrb(fmt, arg...) (void) (verbose && pr_err(fmt, ##arg)) |
ec50f7b6 | 297 | |
1484e727 DW |
298 | static __u8 migr_type(struct imsm_dev *dev) |
299 | { | |
300 | if (dev->vol.migr_type == MIGR_VERIFY && | |
301 | dev->status & DEV_VERIFY_AND_FIX) | |
302 | return MIGR_REPAIR; | |
303 | else | |
304 | return dev->vol.migr_type; | |
305 | } | |
306 | ||
307 | static void set_migr_type(struct imsm_dev *dev, __u8 migr_type) | |
308 | { | |
309 | /* for compatibility with older oroms convert MIGR_REPAIR, into | |
310 | * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status | |
311 | */ | |
312 | if (migr_type == MIGR_REPAIR) { | |
313 | dev->vol.migr_type = MIGR_VERIFY; | |
314 | dev->status |= DEV_VERIFY_AND_FIX; | |
315 | } else { | |
316 | dev->vol.migr_type = migr_type; | |
317 | dev->status &= ~DEV_VERIFY_AND_FIX; | |
318 | } | |
319 | } | |
320 | ||
87eb16df | 321 | static unsigned int sector_count(__u32 bytes) |
cdddbdbc | 322 | { |
654a3817 | 323 | return ROUND_UP(bytes, 512) / 512; |
87eb16df | 324 | } |
cdddbdbc | 325 | |
87eb16df DW |
326 | static unsigned int mpb_sectors(struct imsm_super *mpb) |
327 | { | |
328 | return sector_count(__le32_to_cpu(mpb->mpb_size)); | |
cdddbdbc DW |
329 | } |
330 | ||
ba2de7ba DW |
331 | struct intel_dev { |
332 | struct imsm_dev *dev; | |
333 | struct intel_dev *next; | |
f21e18ca | 334 | unsigned index; |
ba2de7ba DW |
335 | }; |
336 | ||
88654014 LM |
337 | struct intel_hba { |
338 | enum sys_dev_type type; | |
339 | char *path; | |
340 | char *pci_id; | |
341 | struct intel_hba *next; | |
342 | }; | |
343 | ||
1a64be56 LM |
344 | enum action { |
345 | DISK_REMOVE = 1, | |
346 | DISK_ADD | |
347 | }; | |
cdddbdbc DW |
348 | /* internal representation of IMSM metadata */ |
349 | struct intel_super { | |
350 | union { | |
949c47a0 DW |
351 | void *buf; /* O_DIRECT buffer for reading/writing metadata */ |
352 | struct imsm_super *anchor; /* immovable parameters */ | |
cdddbdbc | 353 | }; |
8e59f3d8 AK |
354 | union { |
355 | void *migr_rec_buf; /* buffer for I/O operations */ | |
356 | struct migr_record *migr_rec; /* migration record */ | |
357 | }; | |
51d83f5d AK |
358 | int clean_migration_record_by_mdmon; /* when reshape is switched to next |
359 | array, it indicates that mdmon is allowed to clean migration | |
360 | record */ | |
949c47a0 | 361 | size_t len; /* size of the 'buf' allocation */ |
4d7b1503 DW |
362 | void *next_buf; /* for realloc'ing buf from the manager */ |
363 | size_t next_len; | |
c2c087e6 | 364 | int updates_pending; /* count of pending updates for mdmon */ |
bf5a934a | 365 | int current_vol; /* index of raid device undergoing creation */ |
5551b113 | 366 | unsigned long long create_offset; /* common start for 'current_vol' */ |
148acb7b | 367 | __u32 random; /* random data for seeding new family numbers */ |
ba2de7ba | 368 | struct intel_dev *devlist; |
cdddbdbc DW |
369 | struct dl { |
370 | struct dl *next; | |
371 | int index; | |
372 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
373 | int major, minor; | |
374 | char *devname; | |
b9f594fe | 375 | struct imsm_disk disk; |
cdddbdbc | 376 | int fd; |
0dcecb2e DW |
377 | int extent_cnt; |
378 | struct extent *e; /* for determining freespace @ create */ | |
efb30e7f | 379 | int raiddisk; /* slot to fill in autolayout */ |
1a64be56 | 380 | enum action action; |
ca0748fa | 381 | } *disks, *current_disk; |
1a64be56 LM |
382 | struct dl *disk_mgmt_list; /* list of disks to add/remove while mdmon |
383 | active */ | |
47ee5a45 | 384 | struct dl *missing; /* disks removed while we weren't looking */ |
43dad3d6 | 385 | struct bbm_log *bbm_log; |
88654014 | 386 | struct intel_hba *hba; /* device path of the raid controller for this metadata */ |
88c32bb1 | 387 | const struct imsm_orom *orom; /* platform firmware support */ |
a2b97981 DW |
388 | struct intel_super *next; /* (temp) list for disambiguating family_num */ |
389 | }; | |
390 | ||
391 | struct intel_disk { | |
392 | struct imsm_disk disk; | |
393 | #define IMSM_UNKNOWN_OWNER (-1) | |
394 | int owner; | |
395 | struct intel_disk *next; | |
cdddbdbc DW |
396 | }; |
397 | ||
c2c087e6 DW |
398 | struct extent { |
399 | unsigned long long start, size; | |
400 | }; | |
401 | ||
694575e7 KW |
402 | /* definitions of reshape process types */ |
403 | enum imsm_reshape_type { | |
404 | CH_TAKEOVER, | |
b5347799 | 405 | CH_MIGRATION, |
7abc9871 | 406 | CH_ARRAY_SIZE, |
694575e7 KW |
407 | }; |
408 | ||
88758e9d DW |
409 | /* definition of messages passed to imsm_process_update */ |
410 | enum imsm_update_type { | |
411 | update_activate_spare, | |
8273f55e | 412 | update_create_array, |
33414a01 | 413 | update_kill_array, |
aa534678 | 414 | update_rename_array, |
1a64be56 | 415 | update_add_remove_disk, |
78b10e66 | 416 | update_reshape_container_disks, |
48c5303a | 417 | update_reshape_migration, |
2d40f3a1 AK |
418 | update_takeover, |
419 | update_general_migration_checkpoint, | |
f3871fdc | 420 | update_size_change, |
88758e9d DW |
421 | }; |
422 | ||
423 | struct imsm_update_activate_spare { | |
424 | enum imsm_update_type type; | |
d23fe947 | 425 | struct dl *dl; |
88758e9d DW |
426 | int slot; |
427 | int array; | |
428 | struct imsm_update_activate_spare *next; | |
429 | }; | |
430 | ||
78b10e66 | 431 | struct geo_params { |
4dd2df09 | 432 | char devnm[32]; |
78b10e66 | 433 | char *dev_name; |
d04f65f4 | 434 | unsigned long long size; |
78b10e66 N |
435 | int level; |
436 | int layout; | |
437 | int chunksize; | |
438 | int raid_disks; | |
439 | }; | |
440 | ||
bb025c2f KW |
441 | enum takeover_direction { |
442 | R10_TO_R0, | |
443 | R0_TO_R10 | |
444 | }; | |
445 | struct imsm_update_takeover { | |
446 | enum imsm_update_type type; | |
447 | int subarray; | |
448 | enum takeover_direction direction; | |
449 | }; | |
78b10e66 N |
450 | |
451 | struct imsm_update_reshape { | |
452 | enum imsm_update_type type; | |
453 | int old_raid_disks; | |
454 | int new_raid_disks; | |
48c5303a PC |
455 | |
456 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ | |
457 | }; | |
458 | ||
459 | struct imsm_update_reshape_migration { | |
460 | enum imsm_update_type type; | |
461 | int old_raid_disks; | |
462 | int new_raid_disks; | |
463 | /* fields for array migration changes | |
464 | */ | |
465 | int subdev; | |
466 | int new_level; | |
467 | int new_layout; | |
4bba0439 | 468 | int new_chunksize; |
48c5303a | 469 | |
d195167d | 470 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ |
78b10e66 N |
471 | }; |
472 | ||
f3871fdc AK |
473 | struct imsm_update_size_change { |
474 | enum imsm_update_type type; | |
475 | int subdev; | |
476 | long long new_size; | |
477 | }; | |
478 | ||
2d40f3a1 AK |
479 | struct imsm_update_general_migration_checkpoint { |
480 | enum imsm_update_type type; | |
481 | __u32 curr_migr_unit; | |
482 | }; | |
483 | ||
54c2c1ea DW |
484 | struct disk_info { |
485 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
486 | }; | |
487 | ||
8273f55e DW |
488 | struct imsm_update_create_array { |
489 | enum imsm_update_type type; | |
8273f55e | 490 | int dev_idx; |
6a3e913e | 491 | struct imsm_dev dev; |
8273f55e DW |
492 | }; |
493 | ||
33414a01 DW |
494 | struct imsm_update_kill_array { |
495 | enum imsm_update_type type; | |
496 | int dev_idx; | |
497 | }; | |
498 | ||
aa534678 DW |
499 | struct imsm_update_rename_array { |
500 | enum imsm_update_type type; | |
501 | __u8 name[MAX_RAID_SERIAL_LEN]; | |
502 | int dev_idx; | |
503 | }; | |
504 | ||
1a64be56 | 505 | struct imsm_update_add_remove_disk { |
43dad3d6 DW |
506 | enum imsm_update_type type; |
507 | }; | |
508 | ||
88654014 LM |
509 | static const char *_sys_dev_type[] = { |
510 | [SYS_DEV_UNKNOWN] = "Unknown", | |
511 | [SYS_DEV_SAS] = "SAS", | |
614902f6 | 512 | [SYS_DEV_SATA] = "SATA", |
60f0f54d PB |
513 | [SYS_DEV_NVME] = "NVMe", |
514 | [SYS_DEV_VMD] = "VMD" | |
88654014 LM |
515 | }; |
516 | ||
517 | const char *get_sys_dev_type(enum sys_dev_type type) | |
518 | { | |
519 | if (type >= SYS_DEV_MAX) | |
520 | type = SYS_DEV_UNKNOWN; | |
521 | ||
522 | return _sys_dev_type[type]; | |
523 | } | |
524 | ||
525 | static struct intel_hba * alloc_intel_hba(struct sys_dev *device) | |
526 | { | |
503975b9 N |
527 | struct intel_hba *result = xmalloc(sizeof(*result)); |
528 | ||
529 | result->type = device->type; | |
530 | result->path = xstrdup(device->path); | |
531 | result->next = NULL; | |
532 | if (result->path && (result->pci_id = strrchr(result->path, '/')) != NULL) | |
533 | result->pci_id++; | |
534 | ||
88654014 LM |
535 | return result; |
536 | } | |
537 | ||
538 | static struct intel_hba * find_intel_hba(struct intel_hba *hba, struct sys_dev *device) | |
539 | { | |
540 | struct intel_hba *result=NULL; | |
541 | for (result = hba; result; result = result->next) { | |
542 | if (result->type == device->type && strcmp(result->path, device->path) == 0) | |
543 | break; | |
544 | } | |
545 | return result; | |
546 | } | |
547 | ||
b4cf4cba | 548 | static int attach_hba_to_super(struct intel_super *super, struct sys_dev *device) |
88654014 LM |
549 | { |
550 | struct intel_hba *hba; | |
551 | ||
552 | /* check if disk attached to Intel HBA */ | |
553 | hba = find_intel_hba(super->hba, device); | |
554 | if (hba != NULL) | |
555 | return 1; | |
556 | /* Check if HBA is already attached to super */ | |
557 | if (super->hba == NULL) { | |
558 | super->hba = alloc_intel_hba(device); | |
559 | return 1; | |
6b781d33 AP |
560 | } |
561 | ||
562 | hba = super->hba; | |
563 | /* Intel metadata allows for all disks attached to the same type HBA. | |
614902f6 | 564 | * Do not support HBA types mixing |
6b781d33 AP |
565 | */ |
566 | if (device->type != hba->type) | |
88654014 | 567 | return 2; |
6b781d33 | 568 | |
60f0f54d PB |
569 | /* Always forbid spanning between VMD domains (seen as different controllers by mdadm) */ |
570 | if (device->type == SYS_DEV_VMD && !path_attached_to_hba(device->path, hba->path)) | |
571 | return 2; | |
572 | ||
6b781d33 AP |
573 | /* Multiple same type HBAs can be used if they share the same OROM */ |
574 | const struct imsm_orom *device_orom = get_orom_by_device_id(device->dev_id); | |
575 | ||
576 | if (device_orom != super->orom) | |
577 | return 2; | |
578 | ||
579 | while (hba->next) | |
580 | hba = hba->next; | |
581 | ||
582 | hba->next = alloc_intel_hba(device); | |
583 | return 1; | |
88654014 LM |
584 | } |
585 | ||
586 | static struct sys_dev* find_disk_attached_hba(int fd, const char *devname) | |
587 | { | |
9bc4ae77 | 588 | struct sys_dev *list, *elem; |
88654014 LM |
589 | char *disk_path; |
590 | ||
591 | if ((list = find_intel_devices()) == NULL) | |
592 | return 0; | |
593 | ||
594 | if (fd < 0) | |
595 | disk_path = (char *) devname; | |
596 | else | |
597 | disk_path = diskfd_to_devpath(fd); | |
598 | ||
9bc4ae77 | 599 | if (!disk_path) |
88654014 | 600 | return 0; |
88654014 | 601 | |
9bc4ae77 N |
602 | for (elem = list; elem; elem = elem->next) |
603 | if (path_attached_to_hba(disk_path, elem->path)) | |
88654014 | 604 | return elem; |
9bc4ae77 | 605 | |
88654014 LM |
606 | if (disk_path != devname) |
607 | free(disk_path); | |
88654014 LM |
608 | |
609 | return NULL; | |
610 | } | |
611 | ||
d424212e N |
612 | static int find_intel_hba_capability(int fd, struct intel_super *super, |
613 | char *devname); | |
f2f5c343 | 614 | |
cdddbdbc DW |
615 | static struct supertype *match_metadata_desc_imsm(char *arg) |
616 | { | |
617 | struct supertype *st; | |
618 | ||
619 | if (strcmp(arg, "imsm") != 0 && | |
620 | strcmp(arg, "default") != 0 | |
621 | ) | |
622 | return NULL; | |
623 | ||
503975b9 | 624 | st = xcalloc(1, sizeof(*st)); |
cdddbdbc DW |
625 | st->ss = &super_imsm; |
626 | st->max_devs = IMSM_MAX_DEVICES; | |
627 | st->minor_version = 0; | |
628 | st->sb = NULL; | |
629 | return st; | |
630 | } | |
631 | ||
0e600426 | 632 | #ifndef MDASSEMBLE |
cdddbdbc DW |
633 | static __u8 *get_imsm_version(struct imsm_super *mpb) |
634 | { | |
635 | return &mpb->sig[MPB_SIG_LEN]; | |
636 | } | |
9e2d750d | 637 | #endif |
cdddbdbc | 638 | |
949c47a0 DW |
639 | /* retrieve a disk directly from the anchor when the anchor is known to be |
640 | * up-to-date, currently only at load time | |
641 | */ | |
642 | static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index) | |
cdddbdbc | 643 | { |
949c47a0 | 644 | if (index >= mpb->num_disks) |
cdddbdbc DW |
645 | return NULL; |
646 | return &mpb->disk[index]; | |
647 | } | |
648 | ||
95d07a2c LM |
649 | /* retrieve the disk description based on a index of the disk |
650 | * in the sub-array | |
651 | */ | |
652 | static struct dl *get_imsm_dl_disk(struct intel_super *super, __u8 index) | |
949c47a0 | 653 | { |
b9f594fe DW |
654 | struct dl *d; |
655 | ||
656 | for (d = super->disks; d; d = d->next) | |
657 | if (d->index == index) | |
95d07a2c LM |
658 | return d; |
659 | ||
660 | return NULL; | |
661 | } | |
662 | /* retrieve a disk from the parsed metadata */ | |
663 | static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index) | |
664 | { | |
665 | struct dl *dl; | |
666 | ||
667 | dl = get_imsm_dl_disk(super, index); | |
668 | if (dl) | |
669 | return &dl->disk; | |
670 | ||
b9f594fe | 671 | return NULL; |
949c47a0 DW |
672 | } |
673 | ||
674 | /* generate a checksum directly from the anchor when the anchor is known to be | |
675 | * up-to-date, currently only at load or write_super after coalescing | |
676 | */ | |
677 | static __u32 __gen_imsm_checksum(struct imsm_super *mpb) | |
cdddbdbc DW |
678 | { |
679 | __u32 end = mpb->mpb_size / sizeof(end); | |
680 | __u32 *p = (__u32 *) mpb; | |
681 | __u32 sum = 0; | |
682 | ||
5d500228 N |
683 | while (end--) { |
684 | sum += __le32_to_cpu(*p); | |
97f734fd N |
685 | p++; |
686 | } | |
cdddbdbc | 687 | |
5d500228 | 688 | return sum - __le32_to_cpu(mpb->check_sum); |
cdddbdbc DW |
689 | } |
690 | ||
a965f303 DW |
691 | static size_t sizeof_imsm_map(struct imsm_map *map) |
692 | { | |
693 | return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1); | |
694 | } | |
695 | ||
696 | struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map) | |
cdddbdbc | 697 | { |
5e7b0330 AK |
698 | /* A device can have 2 maps if it is in the middle of a migration. |
699 | * If second_map is: | |
238c0a71 AK |
700 | * MAP_0 - we return the first map |
701 | * MAP_1 - we return the second map if it exists, else NULL | |
702 | * MAP_X - we return the second map if it exists, else the first | |
5e7b0330 | 703 | */ |
a965f303 | 704 | struct imsm_map *map = &dev->vol.map[0]; |
9535fc47 | 705 | struct imsm_map *map2 = NULL; |
a965f303 | 706 | |
9535fc47 AK |
707 | if (dev->vol.migr_state) |
708 | map2 = (void *)map + sizeof_imsm_map(map); | |
a965f303 | 709 | |
9535fc47 | 710 | switch (second_map) { |
3b451610 | 711 | case MAP_0: |
9535fc47 | 712 | break; |
3b451610 | 713 | case MAP_1: |
9535fc47 AK |
714 | map = map2; |
715 | break; | |
238c0a71 | 716 | case MAP_X: |
9535fc47 AK |
717 | if (map2) |
718 | map = map2; | |
719 | break; | |
9535fc47 AK |
720 | default: |
721 | map = NULL; | |
722 | } | |
723 | return map; | |
5e7b0330 | 724 | |
a965f303 | 725 | } |
cdddbdbc | 726 | |
3393c6af DW |
727 | /* return the size of the device. |
728 | * migr_state increases the returned size if map[0] were to be duplicated | |
729 | */ | |
730 | static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state) | |
a965f303 DW |
731 | { |
732 | size_t size = sizeof(*dev) - sizeof(struct imsm_map) + | |
238c0a71 | 733 | sizeof_imsm_map(get_imsm_map(dev, MAP_0)); |
cdddbdbc DW |
734 | |
735 | /* migrating means an additional map */ | |
a965f303 | 736 | if (dev->vol.migr_state) |
238c0a71 | 737 | size += sizeof_imsm_map(get_imsm_map(dev, MAP_1)); |
3393c6af | 738 | else if (migr_state) |
238c0a71 | 739 | size += sizeof_imsm_map(get_imsm_map(dev, MAP_0)); |
cdddbdbc DW |
740 | |
741 | return size; | |
742 | } | |
743 | ||
54c2c1ea DW |
744 | #ifndef MDASSEMBLE |
745 | /* retrieve disk serial number list from a metadata update */ | |
746 | static struct disk_info *get_disk_info(struct imsm_update_create_array *update) | |
747 | { | |
748 | void *u = update; | |
749 | struct disk_info *inf; | |
750 | ||
751 | inf = u + sizeof(*update) - sizeof(struct imsm_dev) + | |
752 | sizeof_imsm_dev(&update->dev, 0); | |
753 | ||
754 | return inf; | |
755 | } | |
756 | #endif | |
757 | ||
949c47a0 | 758 | static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index) |
cdddbdbc DW |
759 | { |
760 | int offset; | |
761 | int i; | |
762 | void *_mpb = mpb; | |
763 | ||
949c47a0 | 764 | if (index >= mpb->num_raid_devs) |
cdddbdbc DW |
765 | return NULL; |
766 | ||
767 | /* devices start after all disks */ | |
768 | offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb; | |
769 | ||
770 | for (i = 0; i <= index; i++) | |
771 | if (i == index) | |
772 | return _mpb + offset; | |
773 | else | |
3393c6af | 774 | offset += sizeof_imsm_dev(_mpb + offset, 0); |
cdddbdbc DW |
775 | |
776 | return NULL; | |
777 | } | |
778 | ||
949c47a0 DW |
779 | static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index) |
780 | { | |
ba2de7ba DW |
781 | struct intel_dev *dv; |
782 | ||
949c47a0 DW |
783 | if (index >= super->anchor->num_raid_devs) |
784 | return NULL; | |
ba2de7ba DW |
785 | for (dv = super->devlist; dv; dv = dv->next) |
786 | if (dv->index == index) | |
787 | return dv->dev; | |
788 | return NULL; | |
949c47a0 DW |
789 | } |
790 | ||
98130f40 AK |
791 | /* |
792 | * for second_map: | |
238c0a71 AK |
793 | * == MAP_0 get first map |
794 | * == MAP_1 get second map | |
795 | * == MAP_X than get map according to the current migr_state | |
98130f40 AK |
796 | */ |
797 | static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev, | |
798 | int slot, | |
799 | int second_map) | |
7eef0453 DW |
800 | { |
801 | struct imsm_map *map; | |
802 | ||
5e7b0330 | 803 | map = get_imsm_map(dev, second_map); |
7eef0453 | 804 | |
ff077194 DW |
805 | /* top byte identifies disk under rebuild */ |
806 | return __le32_to_cpu(map->disk_ord_tbl[slot]); | |
807 | } | |
808 | ||
809 | #define ord_to_idx(ord) (((ord) << 8) >> 8) | |
98130f40 | 810 | static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot, int second_map) |
ff077194 | 811 | { |
98130f40 | 812 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, second_map); |
ff077194 DW |
813 | |
814 | return ord_to_idx(ord); | |
7eef0453 DW |
815 | } |
816 | ||
be73972f DW |
817 | static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord) |
818 | { | |
819 | map->disk_ord_tbl[slot] = __cpu_to_le32(ord); | |
820 | } | |
821 | ||
f21e18ca | 822 | static int get_imsm_disk_slot(struct imsm_map *map, unsigned idx) |
620b1713 DW |
823 | { |
824 | int slot; | |
825 | __u32 ord; | |
826 | ||
827 | for (slot = 0; slot < map->num_members; slot++) { | |
828 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
829 | if (ord_to_idx(ord) == idx) | |
830 | return slot; | |
831 | } | |
832 | ||
833 | return -1; | |
834 | } | |
835 | ||
cdddbdbc DW |
836 | static int get_imsm_raid_level(struct imsm_map *map) |
837 | { | |
838 | if (map->raid_level == 1) { | |
839 | if (map->num_members == 2) | |
840 | return 1; | |
841 | else | |
842 | return 10; | |
843 | } | |
844 | ||
845 | return map->raid_level; | |
846 | } | |
847 | ||
c2c087e6 DW |
848 | static int cmp_extent(const void *av, const void *bv) |
849 | { | |
850 | const struct extent *a = av; | |
851 | const struct extent *b = bv; | |
852 | if (a->start < b->start) | |
853 | return -1; | |
854 | if (a->start > b->start) | |
855 | return 1; | |
856 | return 0; | |
857 | } | |
858 | ||
0dcecb2e | 859 | static int count_memberships(struct dl *dl, struct intel_super *super) |
c2c087e6 | 860 | { |
c2c087e6 | 861 | int memberships = 0; |
620b1713 | 862 | int i; |
c2c087e6 | 863 | |
949c47a0 DW |
864 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
865 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 866 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
c2c087e6 | 867 | |
620b1713 DW |
868 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
869 | memberships++; | |
c2c087e6 | 870 | } |
0dcecb2e DW |
871 | |
872 | return memberships; | |
873 | } | |
874 | ||
b81221b7 CA |
875 | static __u32 imsm_min_reserved_sectors(struct intel_super *super); |
876 | ||
5551b113 CA |
877 | static int split_ull(unsigned long long n, __u32 *lo, __u32 *hi) |
878 | { | |
879 | if (lo == 0 || hi == 0) | |
880 | return 1; | |
881 | *lo = __le32_to_cpu((unsigned)n); | |
882 | *hi = __le32_to_cpu((unsigned)(n >> 32)); | |
883 | return 0; | |
884 | } | |
885 | ||
886 | static unsigned long long join_u32(__u32 lo, __u32 hi) | |
887 | { | |
888 | return (unsigned long long)__le32_to_cpu(lo) | | |
889 | (((unsigned long long)__le32_to_cpu(hi)) << 32); | |
890 | } | |
891 | ||
892 | static unsigned long long total_blocks(struct imsm_disk *disk) | |
893 | { | |
894 | if (disk == NULL) | |
895 | return 0; | |
896 | return join_u32(disk->total_blocks_lo, disk->total_blocks_hi); | |
897 | } | |
898 | ||
899 | static unsigned long long pba_of_lba0(struct imsm_map *map) | |
900 | { | |
901 | if (map == NULL) | |
902 | return 0; | |
903 | return join_u32(map->pba_of_lba0_lo, map->pba_of_lba0_hi); | |
904 | } | |
905 | ||
906 | static unsigned long long blocks_per_member(struct imsm_map *map) | |
907 | { | |
908 | if (map == NULL) | |
909 | return 0; | |
910 | return join_u32(map->blocks_per_member_lo, map->blocks_per_member_hi); | |
911 | } | |
912 | ||
ae2416e7 | 913 | #ifndef MDASSEMBLE |
5551b113 CA |
914 | static unsigned long long num_data_stripes(struct imsm_map *map) |
915 | { | |
916 | if (map == NULL) | |
917 | return 0; | |
918 | return join_u32(map->num_data_stripes_lo, map->num_data_stripes_hi); | |
919 | } | |
920 | ||
921 | static void set_total_blocks(struct imsm_disk *disk, unsigned long long n) | |
922 | { | |
923 | split_ull(n, &disk->total_blocks_lo, &disk->total_blocks_hi); | |
924 | } | |
ae2416e7 | 925 | #endif |
5551b113 CA |
926 | |
927 | static void set_pba_of_lba0(struct imsm_map *map, unsigned long long n) | |
928 | { | |
929 | split_ull(n, &map->pba_of_lba0_lo, &map->pba_of_lba0_hi); | |
930 | } | |
931 | ||
932 | static void set_blocks_per_member(struct imsm_map *map, unsigned long long n) | |
933 | { | |
934 | split_ull(n, &map->blocks_per_member_lo, &map->blocks_per_member_hi); | |
935 | } | |
936 | ||
937 | static void set_num_data_stripes(struct imsm_map *map, unsigned long long n) | |
938 | { | |
939 | split_ull(n, &map->num_data_stripes_lo, &map->num_data_stripes_hi); | |
940 | } | |
941 | ||
0dcecb2e DW |
942 | static struct extent *get_extents(struct intel_super *super, struct dl *dl) |
943 | { | |
944 | /* find a list of used extents on the given physical device */ | |
945 | struct extent *rv, *e; | |
620b1713 | 946 | int i; |
0dcecb2e | 947 | int memberships = count_memberships(dl, super); |
b276dd33 DW |
948 | __u32 reservation; |
949 | ||
950 | /* trim the reserved area for spares, so they can join any array | |
951 | * regardless of whether the OROM has assigned sectors from the | |
952 | * IMSM_RESERVED_SECTORS region | |
953 | */ | |
954 | if (dl->index == -1) | |
b81221b7 | 955 | reservation = imsm_min_reserved_sectors(super); |
b276dd33 DW |
956 | else |
957 | reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
0dcecb2e | 958 | |
503975b9 | 959 | rv = xcalloc(sizeof(struct extent), (memberships + 1)); |
c2c087e6 DW |
960 | e = rv; |
961 | ||
949c47a0 DW |
962 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
963 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 964 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
c2c087e6 | 965 | |
620b1713 | 966 | if (get_imsm_disk_slot(map, dl->index) >= 0) { |
5551b113 CA |
967 | e->start = pba_of_lba0(map); |
968 | e->size = blocks_per_member(map); | |
620b1713 | 969 | e++; |
c2c087e6 DW |
970 | } |
971 | } | |
972 | qsort(rv, memberships, sizeof(*rv), cmp_extent); | |
973 | ||
1011e834 | 974 | /* determine the start of the metadata |
14e8215b DW |
975 | * when no raid devices are defined use the default |
976 | * ...otherwise allow the metadata to truncate the value | |
977 | * as is the case with older versions of imsm | |
978 | */ | |
979 | if (memberships) { | |
980 | struct extent *last = &rv[memberships - 1]; | |
5551b113 | 981 | unsigned long long remainder; |
14e8215b | 982 | |
5551b113 | 983 | remainder = total_blocks(&dl->disk) - (last->start + last->size); |
dda5855f DW |
984 | /* round down to 1k block to satisfy precision of the kernel |
985 | * 'size' interface | |
986 | */ | |
987 | remainder &= ~1UL; | |
988 | /* make sure remainder is still sane */ | |
f21e18ca | 989 | if (remainder < (unsigned)ROUND_UP(super->len, 512) >> 9) |
dda5855f | 990 | remainder = ROUND_UP(super->len, 512) >> 9; |
14e8215b DW |
991 | if (reservation > remainder) |
992 | reservation = remainder; | |
993 | } | |
5551b113 | 994 | e->start = total_blocks(&dl->disk) - reservation; |
c2c087e6 DW |
995 | e->size = 0; |
996 | return rv; | |
997 | } | |
998 | ||
14e8215b DW |
999 | /* try to determine how much space is reserved for metadata from |
1000 | * the last get_extents() entry, otherwise fallback to the | |
1001 | * default | |
1002 | */ | |
1003 | static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl) | |
1004 | { | |
1005 | struct extent *e; | |
1006 | int i; | |
1007 | __u32 rv; | |
1008 | ||
1009 | /* for spares just return a minimal reservation which will grow | |
1010 | * once the spare is picked up by an array | |
1011 | */ | |
1012 | if (dl->index == -1) | |
1013 | return MPB_SECTOR_CNT; | |
1014 | ||
1015 | e = get_extents(super, dl); | |
1016 | if (!e) | |
1017 | return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
1018 | ||
1019 | /* scroll to last entry */ | |
1020 | for (i = 0; e[i].size; i++) | |
1021 | continue; | |
1022 | ||
5551b113 | 1023 | rv = total_blocks(&dl->disk) - e[i].start; |
14e8215b DW |
1024 | |
1025 | free(e); | |
1026 | ||
1027 | return rv; | |
1028 | } | |
1029 | ||
25ed7e59 DW |
1030 | static int is_spare(struct imsm_disk *disk) |
1031 | { | |
1032 | return (disk->status & SPARE_DISK) == SPARE_DISK; | |
1033 | } | |
1034 | ||
1035 | static int is_configured(struct imsm_disk *disk) | |
1036 | { | |
1037 | return (disk->status & CONFIGURED_DISK) == CONFIGURED_DISK; | |
1038 | } | |
1039 | ||
1040 | static int is_failed(struct imsm_disk *disk) | |
1041 | { | |
1042 | return (disk->status & FAILED_DISK) == FAILED_DISK; | |
1043 | } | |
1044 | ||
b81221b7 CA |
1045 | /* try to determine how much space is reserved for metadata from |
1046 | * the last get_extents() entry on the smallest active disk, | |
1047 | * otherwise fallback to the default | |
1048 | */ | |
1049 | static __u32 imsm_min_reserved_sectors(struct intel_super *super) | |
1050 | { | |
1051 | struct extent *e; | |
1052 | int i; | |
5551b113 CA |
1053 | unsigned long long min_active; |
1054 | __u32 remainder; | |
b81221b7 CA |
1055 | __u32 rv = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; |
1056 | struct dl *dl, *dl_min = NULL; | |
1057 | ||
1058 | if (!super) | |
1059 | return rv; | |
1060 | ||
1061 | min_active = 0; | |
1062 | for (dl = super->disks; dl; dl = dl->next) { | |
1063 | if (dl->index < 0) | |
1064 | continue; | |
5551b113 CA |
1065 | unsigned long long blocks = total_blocks(&dl->disk); |
1066 | if (blocks < min_active || min_active == 0) { | |
b81221b7 | 1067 | dl_min = dl; |
5551b113 | 1068 | min_active = blocks; |
b81221b7 CA |
1069 | } |
1070 | } | |
1071 | if (!dl_min) | |
1072 | return rv; | |
1073 | ||
1074 | /* find last lba used by subarrays on the smallest active disk */ | |
1075 | e = get_extents(super, dl_min); | |
1076 | if (!e) | |
1077 | return rv; | |
1078 | for (i = 0; e[i].size; i++) | |
1079 | continue; | |
1080 | ||
1081 | remainder = min_active - e[i].start; | |
1082 | free(e); | |
1083 | ||
1084 | /* to give priority to recovery we should not require full | |
1085 | IMSM_RESERVED_SECTORS from the spare */ | |
1086 | rv = MPB_SECTOR_CNT + NUM_BLOCKS_DIRTY_STRIPE_REGION; | |
1087 | ||
1088 | /* if real reservation is smaller use that value */ | |
1089 | return (remainder < rv) ? remainder : rv; | |
1090 | } | |
1091 | ||
80e7f8c3 AC |
1092 | /* Return minimum size of a spare that can be used in this array*/ |
1093 | static unsigned long long min_acceptable_spare_size_imsm(struct supertype *st) | |
1094 | { | |
1095 | struct intel_super *super = st->sb; | |
1096 | struct dl *dl; | |
1097 | struct extent *e; | |
1098 | int i; | |
1099 | unsigned long long rv = 0; | |
1100 | ||
1101 | if (!super) | |
1102 | return rv; | |
1103 | /* find first active disk in array */ | |
1104 | dl = super->disks; | |
1105 | while (dl && (is_failed(&dl->disk) || dl->index == -1)) | |
1106 | dl = dl->next; | |
1107 | if (!dl) | |
1108 | return rv; | |
1109 | /* find last lba used by subarrays */ | |
1110 | e = get_extents(super, dl); | |
1111 | if (!e) | |
1112 | return rv; | |
1113 | for (i = 0; e[i].size; i++) | |
1114 | continue; | |
1115 | if (i > 0) | |
1116 | rv = e[i-1].start + e[i-1].size; | |
1117 | free(e); | |
b81221b7 | 1118 | |
80e7f8c3 | 1119 | /* add the amount of space needed for metadata */ |
b81221b7 CA |
1120 | rv = rv + imsm_min_reserved_sectors(super); |
1121 | ||
80e7f8c3 AC |
1122 | return rv * 512; |
1123 | } | |
1124 | ||
d1e02575 AK |
1125 | static int is_gen_migration(struct imsm_dev *dev); |
1126 | ||
1799c9e8 | 1127 | #ifndef MDASSEMBLE |
c47b0ff6 AK |
1128 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
1129 | struct imsm_dev *dev); | |
1e5c6983 | 1130 | |
c47b0ff6 AK |
1131 | static void print_imsm_dev(struct intel_super *super, |
1132 | struct imsm_dev *dev, | |
1133 | char *uuid, | |
1134 | int disk_idx) | |
cdddbdbc DW |
1135 | { |
1136 | __u64 sz; | |
0d80bb2f | 1137 | int slot, i; |
238c0a71 AK |
1138 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1139 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); | |
b10b37b8 | 1140 | __u32 ord; |
cdddbdbc DW |
1141 | |
1142 | printf("\n"); | |
1e7bc0ed | 1143 | printf("[%.16s]:\n", dev->volume); |
44470971 | 1144 | printf(" UUID : %s\n", uuid); |
dd8bcb3b AK |
1145 | printf(" RAID Level : %d", get_imsm_raid_level(map)); |
1146 | if (map2) | |
1147 | printf(" <-- %d", get_imsm_raid_level(map2)); | |
1148 | printf("\n"); | |
1149 | printf(" Members : %d", map->num_members); | |
1150 | if (map2) | |
1151 | printf(" <-- %d", map2->num_members); | |
1152 | printf("\n"); | |
0d80bb2f DW |
1153 | printf(" Slots : ["); |
1154 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 1155 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_0); |
0d80bb2f DW |
1156 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
1157 | } | |
dd8bcb3b AK |
1158 | printf("]"); |
1159 | if (map2) { | |
1160 | printf(" <-- ["); | |
1161 | for (i = 0; i < map2->num_members; i++) { | |
238c0a71 | 1162 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_1); |
dd8bcb3b AK |
1163 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
1164 | } | |
1165 | printf("]"); | |
1166 | } | |
1167 | printf("\n"); | |
7095bccb AK |
1168 | printf(" Failed disk : "); |
1169 | if (map->failed_disk_num == 0xff) | |
1170 | printf("none"); | |
1171 | else | |
1172 | printf("%i", map->failed_disk_num); | |
1173 | printf("\n"); | |
620b1713 DW |
1174 | slot = get_imsm_disk_slot(map, disk_idx); |
1175 | if (slot >= 0) { | |
238c0a71 | 1176 | ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); |
b10b37b8 DW |
1177 | printf(" This Slot : %d%s\n", slot, |
1178 | ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : ""); | |
1179 | } else | |
cdddbdbc DW |
1180 | printf(" This Slot : ?\n"); |
1181 | sz = __le32_to_cpu(dev->size_high); | |
1182 | sz <<= 32; | |
1183 | sz += __le32_to_cpu(dev->size_low); | |
1184 | printf(" Array Size : %llu%s\n", (unsigned long long)sz, | |
1185 | human_size(sz * 512)); | |
5551b113 | 1186 | sz = blocks_per_member(map); |
cdddbdbc DW |
1187 | printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz, |
1188 | human_size(sz * 512)); | |
5551b113 CA |
1189 | printf(" Sector Offset : %llu\n", |
1190 | pba_of_lba0(map)); | |
1191 | printf(" Num Stripes : %llu\n", | |
1192 | num_data_stripes(map)); | |
dd8bcb3b | 1193 | printf(" Chunk Size : %u KiB", |
cdddbdbc | 1194 | __le16_to_cpu(map->blocks_per_strip) / 2); |
dd8bcb3b AK |
1195 | if (map2) |
1196 | printf(" <-- %u KiB", | |
1197 | __le16_to_cpu(map2->blocks_per_strip) / 2); | |
1198 | printf("\n"); | |
cdddbdbc | 1199 | printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks)); |
8655a7b1 | 1200 | printf(" Migrate State : "); |
1484e727 DW |
1201 | if (dev->vol.migr_state) { |
1202 | if (migr_type(dev) == MIGR_INIT) | |
8655a7b1 | 1203 | printf("initialize\n"); |
1484e727 | 1204 | else if (migr_type(dev) == MIGR_REBUILD) |
8655a7b1 | 1205 | printf("rebuild\n"); |
1484e727 | 1206 | else if (migr_type(dev) == MIGR_VERIFY) |
8655a7b1 | 1207 | printf("check\n"); |
1484e727 | 1208 | else if (migr_type(dev) == MIGR_GEN_MIGR) |
8655a7b1 | 1209 | printf("general migration\n"); |
1484e727 | 1210 | else if (migr_type(dev) == MIGR_STATE_CHANGE) |
8655a7b1 | 1211 | printf("state change\n"); |
1484e727 | 1212 | else if (migr_type(dev) == MIGR_REPAIR) |
8655a7b1 | 1213 | printf("repair\n"); |
1484e727 | 1214 | else |
8655a7b1 DW |
1215 | printf("<unknown:%d>\n", migr_type(dev)); |
1216 | } else | |
1217 | printf("idle\n"); | |
3393c6af DW |
1218 | printf(" Map State : %s", map_state_str[map->map_state]); |
1219 | if (dev->vol.migr_state) { | |
238c0a71 | 1220 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 | 1221 | |
b10b37b8 | 1222 | printf(" <-- %s", map_state_str[map->map_state]); |
464d40e8 LD |
1223 | printf("\n Checkpoint : %u ", |
1224 | __le32_to_cpu(dev->vol.curr_migr_unit)); | |
3136abe5 | 1225 | if ((is_gen_migration(dev)) && ((slot > 1) || (slot < 0))) |
464d40e8 LD |
1226 | printf("(N/A)"); |
1227 | else | |
1228 | printf("(%llu)", (unsigned long long) | |
1229 | blocks_per_migr_unit(super, dev)); | |
3393c6af DW |
1230 | } |
1231 | printf("\n"); | |
cdddbdbc | 1232 | printf(" Dirty State : %s\n", dev->vol.dirty ? "dirty" : "clean"); |
cdddbdbc DW |
1233 | } |
1234 | ||
0ec1f4e8 | 1235 | static void print_imsm_disk(struct imsm_disk *disk, int index, __u32 reserved) |
cdddbdbc | 1236 | { |
1f24f035 | 1237 | char str[MAX_RAID_SERIAL_LEN + 1]; |
cdddbdbc DW |
1238 | __u64 sz; |
1239 | ||
0ec1f4e8 | 1240 | if (index < -1 || !disk) |
e9d82038 DW |
1241 | return; |
1242 | ||
cdddbdbc | 1243 | printf("\n"); |
1f24f035 | 1244 | snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial); |
0ec1f4e8 DW |
1245 | if (index >= 0) |
1246 | printf(" Disk%02d Serial : %s\n", index, str); | |
1247 | else | |
1248 | printf(" Disk Serial : %s\n", str); | |
25ed7e59 DW |
1249 | printf(" State :%s%s%s\n", is_spare(disk) ? " spare" : "", |
1250 | is_configured(disk) ? " active" : "", | |
1251 | is_failed(disk) ? " failed" : ""); | |
cdddbdbc | 1252 | printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id)); |
5551b113 | 1253 | sz = total_blocks(disk) - reserved; |
cdddbdbc DW |
1254 | printf(" Usable Size : %llu%s\n", (unsigned long long)sz, |
1255 | human_size(sz * 512)); | |
1256 | } | |
1257 | ||
520e69e2 AK |
1258 | void examine_migr_rec_imsm(struct intel_super *super) |
1259 | { | |
1260 | struct migr_record *migr_rec = super->migr_rec; | |
1261 | struct imsm_super *mpb = super->anchor; | |
1262 | int i; | |
1263 | ||
1264 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
1265 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
3136abe5 | 1266 | struct imsm_map *map; |
b4ab44d8 | 1267 | int slot = -1; |
3136abe5 | 1268 | |
520e69e2 AK |
1269 | if (is_gen_migration(dev) == 0) |
1270 | continue; | |
1271 | ||
1272 | printf("\nMigration Record Information:"); | |
3136abe5 | 1273 | |
44bfe6df AK |
1274 | /* first map under migration */ |
1275 | map = get_imsm_map(dev, MAP_0); | |
3136abe5 AK |
1276 | if (map) |
1277 | slot = get_imsm_disk_slot(map, super->disks->index); | |
1278 | if ((map == NULL) || (slot > 1) || (slot < 0)) { | |
520e69e2 AK |
1279 | printf(" Empty\n "); |
1280 | printf("Examine one of first two disks in array\n"); | |
1281 | break; | |
1282 | } | |
1283 | printf("\n Status : "); | |
1284 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
1285 | printf("Normal\n"); | |
1286 | else | |
1287 | printf("Contains Data\n"); | |
1288 | printf(" Current Unit : %u\n", | |
1289 | __le32_to_cpu(migr_rec->curr_migr_unit)); | |
1290 | printf(" Family : %u\n", | |
1291 | __le32_to_cpu(migr_rec->family_num)); | |
1292 | printf(" Ascending : %u\n", | |
1293 | __le32_to_cpu(migr_rec->ascending_migr)); | |
1294 | printf(" Blocks Per Unit : %u\n", | |
1295 | __le32_to_cpu(migr_rec->blocks_per_unit)); | |
1296 | printf(" Dest. Depth Per Unit : %u\n", | |
1297 | __le32_to_cpu(migr_rec->dest_depth_per_unit)); | |
1298 | printf(" Checkpoint Area pba : %u\n", | |
1299 | __le32_to_cpu(migr_rec->ckpt_area_pba)); | |
1300 | printf(" First member lba : %u\n", | |
1301 | __le32_to_cpu(migr_rec->dest_1st_member_lba)); | |
1302 | printf(" Total Number of Units : %u\n", | |
1303 | __le32_to_cpu(migr_rec->num_migr_units)); | |
1304 | printf(" Size of volume : %u\n", | |
1305 | __le32_to_cpu(migr_rec->post_migr_vol_cap)); | |
1306 | printf(" Expansion space for LBA64 : %u\n", | |
1307 | __le32_to_cpu(migr_rec->post_migr_vol_cap_hi)); | |
1308 | printf(" Record was read from : %u\n", | |
1309 | __le32_to_cpu(migr_rec->ckpt_read_disk_num)); | |
1310 | ||
1311 | break; | |
1312 | } | |
1313 | } | |
9e2d750d | 1314 | #endif /* MDASSEMBLE */ |
19482bcc AK |
1315 | /******************************************************************************* |
1316 | * function: imsm_check_attributes | |
1317 | * Description: Function checks if features represented by attributes flags | |
1011e834 | 1318 | * are supported by mdadm. |
19482bcc AK |
1319 | * Parameters: |
1320 | * attributes - Attributes read from metadata | |
1321 | * Returns: | |
1011e834 N |
1322 | * 0 - passed attributes contains unsupported features flags |
1323 | * 1 - all features are supported | |
19482bcc AK |
1324 | ******************************************************************************/ |
1325 | static int imsm_check_attributes(__u32 attributes) | |
1326 | { | |
1327 | int ret_val = 1; | |
418f9b36 N |
1328 | __u32 not_supported = MPB_ATTRIB_SUPPORTED^0xffffffff; |
1329 | ||
1330 | not_supported &= ~MPB_ATTRIB_IGNORED; | |
19482bcc AK |
1331 | |
1332 | not_supported &= attributes; | |
1333 | if (not_supported) { | |
e7b84f9d | 1334 | pr_err("(IMSM): Unsupported attributes : %x\n", |
418f9b36 | 1335 | (unsigned)__le32_to_cpu(not_supported)); |
19482bcc AK |
1336 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { |
1337 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY \n"); | |
1338 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
1339 | } | |
1340 | if (not_supported & MPB_ATTRIB_2TB) { | |
1341 | dprintf("\t\tMPB_ATTRIB_2TB\n"); | |
1342 | not_supported ^= MPB_ATTRIB_2TB; | |
1343 | } | |
1344 | if (not_supported & MPB_ATTRIB_RAID0) { | |
1345 | dprintf("\t\tMPB_ATTRIB_RAID0\n"); | |
1346 | not_supported ^= MPB_ATTRIB_RAID0; | |
1347 | } | |
1348 | if (not_supported & MPB_ATTRIB_RAID1) { | |
1349 | dprintf("\t\tMPB_ATTRIB_RAID1\n"); | |
1350 | not_supported ^= MPB_ATTRIB_RAID1; | |
1351 | } | |
1352 | if (not_supported & MPB_ATTRIB_RAID10) { | |
1353 | dprintf("\t\tMPB_ATTRIB_RAID10\n"); | |
1354 | not_supported ^= MPB_ATTRIB_RAID10; | |
1355 | } | |
1356 | if (not_supported & MPB_ATTRIB_RAID1E) { | |
1357 | dprintf("\t\tMPB_ATTRIB_RAID1E\n"); | |
1358 | not_supported ^= MPB_ATTRIB_RAID1E; | |
1359 | } | |
1360 | if (not_supported & MPB_ATTRIB_RAID5) { | |
1361 | dprintf("\t\tMPB_ATTRIB_RAID5\n"); | |
1362 | not_supported ^= MPB_ATTRIB_RAID5; | |
1363 | } | |
1364 | if (not_supported & MPB_ATTRIB_RAIDCNG) { | |
1365 | dprintf("\t\tMPB_ATTRIB_RAIDCNG\n"); | |
1366 | not_supported ^= MPB_ATTRIB_RAIDCNG; | |
1367 | } | |
1368 | if (not_supported & MPB_ATTRIB_BBM) { | |
1369 | dprintf("\t\tMPB_ATTRIB_BBM\n"); | |
1370 | not_supported ^= MPB_ATTRIB_BBM; | |
1371 | } | |
1372 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { | |
1373 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY (== MPB_ATTRIB_LEGACY)\n"); | |
1374 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
1375 | } | |
1376 | if (not_supported & MPB_ATTRIB_EXP_STRIPE_SIZE) { | |
1377 | dprintf("\t\tMPB_ATTRIB_EXP_STRIP_SIZE\n"); | |
1378 | not_supported ^= MPB_ATTRIB_EXP_STRIPE_SIZE; | |
1379 | } | |
1380 | if (not_supported & MPB_ATTRIB_2TB_DISK) { | |
1381 | dprintf("\t\tMPB_ATTRIB_2TB_DISK\n"); | |
1382 | not_supported ^= MPB_ATTRIB_2TB_DISK; | |
1383 | } | |
1384 | if (not_supported & MPB_ATTRIB_NEVER_USE2) { | |
1385 | dprintf("\t\tMPB_ATTRIB_NEVER_USE2\n"); | |
1386 | not_supported ^= MPB_ATTRIB_NEVER_USE2; | |
1387 | } | |
1388 | if (not_supported & MPB_ATTRIB_NEVER_USE) { | |
1389 | dprintf("\t\tMPB_ATTRIB_NEVER_USE\n"); | |
1390 | not_supported ^= MPB_ATTRIB_NEVER_USE; | |
1391 | } | |
1392 | ||
1393 | if (not_supported) | |
1ade5cc1 | 1394 | dprintf("(IMSM): Unknown attributes : %x\n", not_supported); |
19482bcc AK |
1395 | |
1396 | ret_val = 0; | |
1397 | } | |
1398 | ||
1399 | return ret_val; | |
1400 | } | |
1401 | ||
9e2d750d | 1402 | #ifndef MDASSEMBLE |
a5d85af7 | 1403 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map); |
44470971 | 1404 | |
cdddbdbc DW |
1405 | static void examine_super_imsm(struct supertype *st, char *homehost) |
1406 | { | |
1407 | struct intel_super *super = st->sb; | |
949c47a0 | 1408 | struct imsm_super *mpb = super->anchor; |
cdddbdbc DW |
1409 | char str[MAX_SIGNATURE_LENGTH]; |
1410 | int i; | |
27fd6274 DW |
1411 | struct mdinfo info; |
1412 | char nbuf[64]; | |
cdddbdbc | 1413 | __u32 sum; |
14e8215b | 1414 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
94827db3 | 1415 | struct dl *dl; |
27fd6274 | 1416 | |
cdddbdbc DW |
1417 | snprintf(str, MPB_SIG_LEN, "%s", mpb->sig); |
1418 | printf(" Magic : %s\n", str); | |
1419 | snprintf(str, strlen(MPB_VERSION_RAID0), "%s", get_imsm_version(mpb)); | |
1420 | printf(" Version : %s\n", get_imsm_version(mpb)); | |
148acb7b | 1421 | printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num)); |
cdddbdbc DW |
1422 | printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num)); |
1423 | printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num)); | |
19482bcc AK |
1424 | printf(" Attributes : "); |
1425 | if (imsm_check_attributes(mpb->attributes)) | |
1426 | printf("All supported\n"); | |
1427 | else | |
1428 | printf("not supported\n"); | |
a5d85af7 | 1429 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1430 | fname_from_uuid(st, &info, nbuf, ':'); |
27fd6274 | 1431 | printf(" UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
1432 | sum = __le32_to_cpu(mpb->check_sum); |
1433 | printf(" Checksum : %08x %s\n", sum, | |
949c47a0 | 1434 | __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect"); |
87eb16df | 1435 | printf(" MPB Sectors : %d\n", mpb_sectors(mpb)); |
cdddbdbc DW |
1436 | printf(" Disks : %d\n", mpb->num_disks); |
1437 | printf(" RAID Devices : %d\n", mpb->num_raid_devs); | |
0ec1f4e8 | 1438 | print_imsm_disk(__get_imsm_disk(mpb, super->disks->index), super->disks->index, reserved); |
604b746f JD |
1439 | if (super->bbm_log) { |
1440 | struct bbm_log *log = super->bbm_log; | |
1441 | ||
1442 | printf("\n"); | |
1443 | printf("Bad Block Management Log:\n"); | |
1444 | printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size)); | |
1445 | printf(" Signature : %x\n", __le32_to_cpu(log->signature)); | |
1446 | printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count)); | |
1447 | printf(" Spare Blocks : %d\n", __le32_to_cpu(log->reserved_spare_block_count)); | |
13a3b65d N |
1448 | printf(" First Spare : %llx\n", |
1449 | (unsigned long long) __le64_to_cpu(log->first_spare_lba)); | |
604b746f | 1450 | } |
44470971 DW |
1451 | for (i = 0; i < mpb->num_raid_devs; i++) { |
1452 | struct mdinfo info; | |
1453 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1454 | ||
1455 | super->current_vol = i; | |
a5d85af7 | 1456 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1457 | fname_from_uuid(st, &info, nbuf, ':'); |
c47b0ff6 | 1458 | print_imsm_dev(super, dev, nbuf + 5, super->disks->index); |
44470971 | 1459 | } |
cdddbdbc DW |
1460 | for (i = 0; i < mpb->num_disks; i++) { |
1461 | if (i == super->disks->index) | |
1462 | continue; | |
0ec1f4e8 | 1463 | print_imsm_disk(__get_imsm_disk(mpb, i), i, reserved); |
cdddbdbc | 1464 | } |
94827db3 | 1465 | |
0ec1f4e8 DW |
1466 | for (dl = super->disks; dl; dl = dl->next) |
1467 | if (dl->index == -1) | |
1468 | print_imsm_disk(&dl->disk, -1, reserved); | |
520e69e2 AK |
1469 | |
1470 | examine_migr_rec_imsm(super); | |
cdddbdbc DW |
1471 | } |
1472 | ||
061f2c6a | 1473 | static void brief_examine_super_imsm(struct supertype *st, int verbose) |
cdddbdbc | 1474 | { |
27fd6274 | 1475 | /* We just write a generic IMSM ARRAY entry */ |
ff54de6e N |
1476 | struct mdinfo info; |
1477 | char nbuf[64]; | |
1e7bc0ed | 1478 | struct intel_super *super = st->sb; |
1e7bc0ed | 1479 | |
0d5a423f DW |
1480 | if (!super->anchor->num_raid_devs) { |
1481 | printf("ARRAY metadata=imsm\n"); | |
1e7bc0ed | 1482 | return; |
0d5a423f | 1483 | } |
ff54de6e | 1484 | |
a5d85af7 | 1485 | getinfo_super_imsm(st, &info, NULL); |
4737ae25 N |
1486 | fname_from_uuid(st, &info, nbuf, ':'); |
1487 | printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5); | |
1488 | } | |
1489 | ||
1490 | static void brief_examine_subarrays_imsm(struct supertype *st, int verbose) | |
1491 | { | |
1492 | /* We just write a generic IMSM ARRAY entry */ | |
1493 | struct mdinfo info; | |
1494 | char nbuf[64]; | |
1495 | char nbuf1[64]; | |
1496 | struct intel_super *super = st->sb; | |
1497 | int i; | |
1498 | ||
1499 | if (!super->anchor->num_raid_devs) | |
1500 | return; | |
1501 | ||
a5d85af7 | 1502 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1503 | fname_from_uuid(st, &info, nbuf, ':'); |
1e7bc0ed DW |
1504 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
1505 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
1506 | ||
1507 | super->current_vol = i; | |
a5d85af7 | 1508 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1509 | fname_from_uuid(st, &info, nbuf1, ':'); |
1124b3cf | 1510 | printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n", |
cf8de691 | 1511 | dev->volume, nbuf + 5, i, nbuf1 + 5); |
1e7bc0ed | 1512 | } |
cdddbdbc DW |
1513 | } |
1514 | ||
9d84c8ea DW |
1515 | static void export_examine_super_imsm(struct supertype *st) |
1516 | { | |
1517 | struct intel_super *super = st->sb; | |
1518 | struct imsm_super *mpb = super->anchor; | |
1519 | struct mdinfo info; | |
1520 | char nbuf[64]; | |
1521 | ||
a5d85af7 | 1522 | getinfo_super_imsm(st, &info, NULL); |
9d84c8ea DW |
1523 | fname_from_uuid(st, &info, nbuf, ':'); |
1524 | printf("MD_METADATA=imsm\n"); | |
1525 | printf("MD_LEVEL=container\n"); | |
1526 | printf("MD_UUID=%s\n", nbuf+5); | |
1527 | printf("MD_DEVICES=%u\n", mpb->num_disks); | |
1528 | } | |
1529 | ||
74db60b0 N |
1530 | static int copy_metadata_imsm(struct supertype *st, int from, int to) |
1531 | { | |
1532 | /* The second last 512byte sector of the device contains | |
1533 | * the "struct imsm_super" metadata. | |
1534 | * This contains mpb_size which is the size in bytes of the | |
1535 | * extended metadata. This is located immediately before | |
1536 | * the imsm_super. | |
1537 | * We want to read all that, plus the last sector which | |
1538 | * may contain a migration record, and write it all | |
1539 | * to the target. | |
1540 | */ | |
1541 | void *buf; | |
1542 | unsigned long long dsize, offset; | |
1543 | int sectors; | |
1544 | struct imsm_super *sb; | |
1545 | int written = 0; | |
1546 | ||
1547 | if (posix_memalign(&buf, 4096, 4096) != 0) | |
1548 | return 1; | |
1549 | ||
1550 | if (!get_dev_size(from, NULL, &dsize)) | |
1551 | goto err; | |
1552 | ||
1553 | if (lseek64(from, dsize-1024, 0) < 0) | |
1554 | goto err; | |
1555 | if (read(from, buf, 512) != 512) | |
1556 | goto err; | |
1557 | sb = buf; | |
1558 | if (strncmp((char*)sb->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) | |
1559 | goto err; | |
1560 | ||
1561 | sectors = mpb_sectors(sb) + 2; | |
1562 | offset = dsize - sectors * 512; | |
1563 | if (lseek64(from, offset, 0) < 0 || | |
1564 | lseek64(to, offset, 0) < 0) | |
1565 | goto err; | |
1566 | while (written < sectors * 512) { | |
1567 | int n = sectors*512 - written; | |
1568 | if (n > 4096) | |
1569 | n = 4096; | |
1570 | if (read(from, buf, n) != n) | |
1571 | goto err; | |
1572 | if (write(to, buf, n) != n) | |
1573 | goto err; | |
1574 | written += n; | |
1575 | } | |
1576 | free(buf); | |
1577 | return 0; | |
1578 | err: | |
1579 | free(buf); | |
1580 | return 1; | |
1581 | } | |
1582 | ||
cdddbdbc DW |
1583 | static void detail_super_imsm(struct supertype *st, char *homehost) |
1584 | { | |
3ebe00a1 DW |
1585 | struct mdinfo info; |
1586 | char nbuf[64]; | |
1587 | ||
a5d85af7 | 1588 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1589 | fname_from_uuid(st, &info, nbuf, ':'); |
3ebe00a1 | 1590 | printf("\n UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
1591 | } |
1592 | ||
1593 | static void brief_detail_super_imsm(struct supertype *st) | |
1594 | { | |
ff54de6e N |
1595 | struct mdinfo info; |
1596 | char nbuf[64]; | |
a5d85af7 | 1597 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1598 | fname_from_uuid(st, &info, nbuf, ':'); |
ff54de6e | 1599 | printf(" UUID=%s", nbuf + 5); |
cdddbdbc | 1600 | } |
d665cc31 DW |
1601 | |
1602 | static int imsm_read_serial(int fd, char *devname, __u8 *serial); | |
1603 | static void fd2devname(int fd, char *name); | |
1604 | ||
120dc887 | 1605 | static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose) |
d665cc31 | 1606 | { |
120dc887 LM |
1607 | /* dump an unsorted list of devices attached to AHCI Intel storage |
1608 | * controller, as well as non-connected ports | |
d665cc31 DW |
1609 | */ |
1610 | int hba_len = strlen(hba_path) + 1; | |
1611 | struct dirent *ent; | |
1612 | DIR *dir; | |
1613 | char *path = NULL; | |
1614 | int err = 0; | |
1615 | unsigned long port_mask = (1 << port_count) - 1; | |
1616 | ||
f21e18ca | 1617 | if (port_count > (int)sizeof(port_mask) * 8) { |
ba728be7 | 1618 | if (verbose > 0) |
e7b84f9d | 1619 | pr_err("port_count %d out of range\n", port_count); |
d665cc31 DW |
1620 | return 2; |
1621 | } | |
1622 | ||
1623 | /* scroll through /sys/dev/block looking for devices attached to | |
1624 | * this hba | |
1625 | */ | |
1626 | dir = opendir("/sys/dev/block"); | |
1627 | for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) { | |
1628 | int fd; | |
1629 | char model[64]; | |
1630 | char vendor[64]; | |
1631 | char buf[1024]; | |
1632 | int major, minor; | |
1633 | char *device; | |
1634 | char *c; | |
1635 | int port; | |
1636 | int type; | |
1637 | ||
1638 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
1639 | continue; | |
1640 | path = devt_to_devpath(makedev(major, minor)); | |
1641 | if (!path) | |
1642 | continue; | |
1643 | if (!path_attached_to_hba(path, hba_path)) { | |
1644 | free(path); | |
1645 | path = NULL; | |
1646 | continue; | |
1647 | } | |
1648 | ||
1649 | /* retrieve the scsi device type */ | |
1650 | if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) { | |
ba728be7 | 1651 | if (verbose > 0) |
e7b84f9d | 1652 | pr_err("failed to allocate 'device'\n"); |
d665cc31 DW |
1653 | err = 2; |
1654 | break; | |
1655 | } | |
1656 | sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor); | |
193b6c0b | 1657 | if (load_sys(device, buf, sizeof(buf)) != 0) { |
ba728be7 | 1658 | if (verbose > 0) |
e7b84f9d | 1659 | pr_err("failed to read device type for %s\n", |
d665cc31 DW |
1660 | path); |
1661 | err = 2; | |
1662 | free(device); | |
1663 | break; | |
1664 | } | |
1665 | type = strtoul(buf, NULL, 10); | |
1666 | ||
1667 | /* if it's not a disk print the vendor and model */ | |
1668 | if (!(type == 0 || type == 7 || type == 14)) { | |
1669 | vendor[0] = '\0'; | |
1670 | model[0] = '\0'; | |
1671 | sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor); | |
193b6c0b | 1672 | if (load_sys(device, buf, sizeof(buf)) == 0) { |
d665cc31 DW |
1673 | strncpy(vendor, buf, sizeof(vendor)); |
1674 | vendor[sizeof(vendor) - 1] = '\0'; | |
1675 | c = (char *) &vendor[sizeof(vendor) - 1]; | |
1676 | while (isspace(*c) || *c == '\0') | |
1677 | *c-- = '\0'; | |
1678 | ||
1679 | } | |
1680 | sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor); | |
193b6c0b | 1681 | if (load_sys(device, buf, sizeof(buf)) == 0) { |
d665cc31 DW |
1682 | strncpy(model, buf, sizeof(model)); |
1683 | model[sizeof(model) - 1] = '\0'; | |
1684 | c = (char *) &model[sizeof(model) - 1]; | |
1685 | while (isspace(*c) || *c == '\0') | |
1686 | *c-- = '\0'; | |
1687 | } | |
1688 | ||
1689 | if (vendor[0] && model[0]) | |
1690 | sprintf(buf, "%.64s %.64s", vendor, model); | |
1691 | else | |
1692 | switch (type) { /* numbers from hald/linux/device.c */ | |
1693 | case 1: sprintf(buf, "tape"); break; | |
1694 | case 2: sprintf(buf, "printer"); break; | |
1695 | case 3: sprintf(buf, "processor"); break; | |
1696 | case 4: | |
1697 | case 5: sprintf(buf, "cdrom"); break; | |
1698 | case 6: sprintf(buf, "scanner"); break; | |
1699 | case 8: sprintf(buf, "media_changer"); break; | |
1700 | case 9: sprintf(buf, "comm"); break; | |
1701 | case 12: sprintf(buf, "raid"); break; | |
1702 | default: sprintf(buf, "unknown"); | |
1703 | } | |
1704 | } else | |
1705 | buf[0] = '\0'; | |
1706 | free(device); | |
1707 | ||
1708 | /* chop device path to 'host%d' and calculate the port number */ | |
1709 | c = strchr(&path[hba_len], '/'); | |
4e5e717d | 1710 | if (!c) { |
ba728be7 | 1711 | if (verbose > 0) |
e7b84f9d | 1712 | pr_err("%s - invalid path name\n", path + hba_len); |
4e5e717d AW |
1713 | err = 2; |
1714 | break; | |
1715 | } | |
d665cc31 | 1716 | *c = '\0'; |
0858eccf AP |
1717 | if ((sscanf(&path[hba_len], "ata%d", &port) == 1) || |
1718 | ((sscanf(&path[hba_len], "host%d", &port) == 1))) | |
d665cc31 DW |
1719 | port -= host_base; |
1720 | else { | |
ba728be7 | 1721 | if (verbose > 0) { |
d665cc31 | 1722 | *c = '/'; /* repair the full string */ |
e7b84f9d | 1723 | pr_err("failed to determine port number for %s\n", |
d665cc31 DW |
1724 | path); |
1725 | } | |
1726 | err = 2; | |
1727 | break; | |
1728 | } | |
1729 | ||
1730 | /* mark this port as used */ | |
1731 | port_mask &= ~(1 << port); | |
1732 | ||
1733 | /* print out the device information */ | |
1734 | if (buf[0]) { | |
1735 | printf(" Port%d : - non-disk device (%s) -\n", port, buf); | |
1736 | continue; | |
1737 | } | |
1738 | ||
1739 | fd = dev_open(ent->d_name, O_RDONLY); | |
1740 | if (fd < 0) | |
1741 | printf(" Port%d : - disk info unavailable -\n", port); | |
1742 | else { | |
1743 | fd2devname(fd, buf); | |
1744 | printf(" Port%d : %s", port, buf); | |
1745 | if (imsm_read_serial(fd, NULL, (__u8 *) buf) == 0) | |
664d5325 | 1746 | printf(" (%.*s)\n", MAX_RAID_SERIAL_LEN, buf); |
d665cc31 | 1747 | else |
664d5325 | 1748 | printf(" ()\n"); |
4dab422a | 1749 | close(fd); |
d665cc31 | 1750 | } |
d665cc31 DW |
1751 | free(path); |
1752 | path = NULL; | |
1753 | } | |
1754 | if (path) | |
1755 | free(path); | |
1756 | if (dir) | |
1757 | closedir(dir); | |
1758 | if (err == 0) { | |
1759 | int i; | |
1760 | ||
1761 | for (i = 0; i < port_count; i++) | |
1762 | if (port_mask & (1 << i)) | |
1763 | printf(" Port%d : - no device attached -\n", i); | |
1764 | } | |
1765 | ||
1766 | return err; | |
1767 | } | |
1768 | ||
a0abe1e6 | 1769 | static void print_vmd_attached_devs(struct sys_dev *hba) |
60f0f54d PB |
1770 | { |
1771 | struct dirent *ent; | |
1772 | DIR *dir; | |
1773 | char path[292]; | |
1774 | char link[256]; | |
1775 | char *c, *rp; | |
1776 | ||
1777 | if (hba->type != SYS_DEV_VMD) | |
a0abe1e6 | 1778 | return; |
60f0f54d PB |
1779 | |
1780 | /* scroll through /sys/dev/block looking for devices attached to | |
1781 | * this hba | |
1782 | */ | |
1783 | dir = opendir("/sys/bus/pci/drivers/nvme"); | |
b9135011 | 1784 | if (!dir) |
a0abe1e6 | 1785 | return; |
b9135011 JS |
1786 | |
1787 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
60f0f54d PB |
1788 | int n; |
1789 | ||
1790 | /* is 'ent' a device? check that the 'subsystem' link exists and | |
1791 | * that its target matches 'bus' | |
1792 | */ | |
1793 | sprintf(path, "/sys/bus/pci/drivers/nvme/%s/subsystem", | |
1794 | ent->d_name); | |
1795 | n = readlink(path, link, sizeof(link)); | |
1796 | if (n < 0 || n >= (int)sizeof(link)) | |
1797 | continue; | |
1798 | link[n] = '\0'; | |
1799 | c = strrchr(link, '/'); | |
1800 | if (!c) | |
1801 | continue; | |
1802 | if (strncmp("pci", c+1, strlen("pci")) != 0) | |
1803 | continue; | |
1804 | ||
1805 | sprintf(path, "/sys/bus/pci/drivers/nvme/%s", ent->d_name); | |
1806 | /* if not a intel NVMe - skip it*/ | |
1807 | if (devpath_to_vendor(path) != 0x8086) | |
1808 | continue; | |
1809 | ||
1810 | rp = realpath(path, NULL); | |
1811 | if (!rp) | |
1812 | continue; | |
1813 | ||
1814 | if (path_attached_to_hba(rp, hba->path)) { | |
1815 | printf(" NVMe under VMD : %s\n", rp); | |
1816 | } | |
1817 | free(rp); | |
1818 | } | |
1819 | ||
b9135011 | 1820 | closedir(dir); |
60f0f54d PB |
1821 | } |
1822 | ||
120dc887 LM |
1823 | static void print_found_intel_controllers(struct sys_dev *elem) |
1824 | { | |
1825 | for (; elem; elem = elem->next) { | |
e7b84f9d | 1826 | pr_err("found Intel(R) "); |
120dc887 LM |
1827 | if (elem->type == SYS_DEV_SATA) |
1828 | fprintf(stderr, "SATA "); | |
155cbb4c LM |
1829 | else if (elem->type == SYS_DEV_SAS) |
1830 | fprintf(stderr, "SAS "); | |
0858eccf AP |
1831 | else if (elem->type == SYS_DEV_NVME) |
1832 | fprintf(stderr, "NVMe "); | |
60f0f54d PB |
1833 | |
1834 | if (elem->type == SYS_DEV_VMD) | |
1835 | fprintf(stderr, "VMD domain"); | |
1836 | else | |
1837 | fprintf(stderr, "RAID controller"); | |
1838 | ||
120dc887 LM |
1839 | if (elem->pci_id) |
1840 | fprintf(stderr, " at %s", elem->pci_id); | |
1841 | fprintf(stderr, ".\n"); | |
1842 | } | |
1843 | fflush(stderr); | |
1844 | } | |
1845 | ||
120dc887 LM |
1846 | static int ahci_get_port_count(const char *hba_path, int *port_count) |
1847 | { | |
1848 | struct dirent *ent; | |
1849 | DIR *dir; | |
1850 | int host_base = -1; | |
1851 | ||
1852 | *port_count = 0; | |
1853 | if ((dir = opendir(hba_path)) == NULL) | |
1854 | return -1; | |
1855 | ||
1856 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
1857 | int host; | |
1858 | ||
0858eccf AP |
1859 | if ((sscanf(ent->d_name, "ata%d", &host) != 1) && |
1860 | ((sscanf(ent->d_name, "host%d", &host) != 1))) | |
120dc887 LM |
1861 | continue; |
1862 | if (*port_count == 0) | |
1863 | host_base = host; | |
1864 | else if (host < host_base) | |
1865 | host_base = host; | |
1866 | ||
1867 | if (host + 1 > *port_count + host_base) | |
1868 | *port_count = host + 1 - host_base; | |
1869 | } | |
1870 | closedir(dir); | |
1871 | return host_base; | |
1872 | } | |
1873 | ||
a891a3c2 LM |
1874 | static void print_imsm_capability(const struct imsm_orom *orom) |
1875 | { | |
0858eccf AP |
1876 | printf(" Platform : Intel(R) "); |
1877 | if (orom->capabilities == 0 && orom->driver_features == 0) | |
1878 | printf("Matrix Storage Manager\n"); | |
1879 | else | |
1880 | printf("Rapid Storage Technology%s\n", | |
1881 | imsm_orom_is_enterprise(orom) ? " enterprise" : ""); | |
1882 | if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build) | |
1883 | printf(" Version : %d.%d.%d.%d\n", orom->major_ver, | |
1884 | orom->minor_ver, orom->hotfix_ver, orom->build); | |
a891a3c2 LM |
1885 | printf(" RAID Levels :%s%s%s%s%s\n", |
1886 | imsm_orom_has_raid0(orom) ? " raid0" : "", | |
1887 | imsm_orom_has_raid1(orom) ? " raid1" : "", | |
1888 | imsm_orom_has_raid1e(orom) ? " raid1e" : "", | |
1889 | imsm_orom_has_raid10(orom) ? " raid10" : "", | |
1890 | imsm_orom_has_raid5(orom) ? " raid5" : ""); | |
1891 | printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", | |
1892 | imsm_orom_has_chunk(orom, 2) ? " 2k" : "", | |
1893 | imsm_orom_has_chunk(orom, 4) ? " 4k" : "", | |
1894 | imsm_orom_has_chunk(orom, 8) ? " 8k" : "", | |
1895 | imsm_orom_has_chunk(orom, 16) ? " 16k" : "", | |
1896 | imsm_orom_has_chunk(orom, 32) ? " 32k" : "", | |
1897 | imsm_orom_has_chunk(orom, 64) ? " 64k" : "", | |
1898 | imsm_orom_has_chunk(orom, 128) ? " 128k" : "", | |
1899 | imsm_orom_has_chunk(orom, 256) ? " 256k" : "", | |
1900 | imsm_orom_has_chunk(orom, 512) ? " 512k" : "", | |
1901 | imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "", | |
1902 | imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "", | |
1903 | imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "", | |
1904 | imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "", | |
1905 | imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "", | |
1906 | imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "", | |
1907 | imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : ""); | |
29cd0821 CA |
1908 | printf(" 2TB volumes :%s supported\n", |
1909 | (orom->attr & IMSM_OROM_ATTR_2TB)?"":" not"); | |
1910 | printf(" 2TB disks :%s supported\n", | |
1911 | (orom->attr & IMSM_OROM_ATTR_2TB_DISK)?"":" not"); | |
0e7f69a8 | 1912 | printf(" Max Disks : %d\n", orom->tds); |
0858eccf AP |
1913 | printf(" Max Volumes : %d per array, %d per %s\n", |
1914 | orom->vpa, orom->vphba, | |
1915 | imsm_orom_is_nvme(orom) ? "platform" : "controller"); | |
a891a3c2 LM |
1916 | return; |
1917 | } | |
1918 | ||
e50cf220 MN |
1919 | static void print_imsm_capability_export(const struct imsm_orom *orom) |
1920 | { | |
1921 | printf("MD_FIRMWARE_TYPE=imsm\n"); | |
0858eccf AP |
1922 | if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build) |
1923 | printf("IMSM_VERSION=%d.%d.%d.%d\n", orom->major_ver, orom->minor_ver, | |
1924 | orom->hotfix_ver, orom->build); | |
e50cf220 MN |
1925 | printf("IMSM_SUPPORTED_RAID_LEVELS=%s%s%s%s%s\n", |
1926 | imsm_orom_has_raid0(orom) ? "raid0 " : "", | |
1927 | imsm_orom_has_raid1(orom) ? "raid1 " : "", | |
1928 | imsm_orom_has_raid1e(orom) ? "raid1e " : "", | |
1929 | imsm_orom_has_raid5(orom) ? "raid10 " : "", | |
1930 | imsm_orom_has_raid10(orom) ? "raid5 " : ""); | |
1931 | printf("IMSM_SUPPORTED_CHUNK_SIZES=%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", | |
1932 | imsm_orom_has_chunk(orom, 2) ? "2k " : "", | |
1933 | imsm_orom_has_chunk(orom, 4) ? "4k " : "", | |
1934 | imsm_orom_has_chunk(orom, 8) ? "8k " : "", | |
1935 | imsm_orom_has_chunk(orom, 16) ? "16k " : "", | |
1936 | imsm_orom_has_chunk(orom, 32) ? "32k " : "", | |
1937 | imsm_orom_has_chunk(orom, 64) ? "64k " : "", | |
1938 | imsm_orom_has_chunk(orom, 128) ? "128k " : "", | |
1939 | imsm_orom_has_chunk(orom, 256) ? "256k " : "", | |
1940 | imsm_orom_has_chunk(orom, 512) ? "512k " : "", | |
1941 | imsm_orom_has_chunk(orom, 1024*1) ? "1M " : "", | |
1942 | imsm_orom_has_chunk(orom, 1024*2) ? "2M " : "", | |
1943 | imsm_orom_has_chunk(orom, 1024*4) ? "4M " : "", | |
1944 | imsm_orom_has_chunk(orom, 1024*8) ? "8M " : "", | |
1945 | imsm_orom_has_chunk(orom, 1024*16) ? "16M " : "", | |
1946 | imsm_orom_has_chunk(orom, 1024*32) ? "32M " : "", | |
1947 | imsm_orom_has_chunk(orom, 1024*64) ? "64M " : ""); | |
1948 | printf("IMSM_2TB_VOLUMES=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB) ? "yes" : "no"); | |
1949 | printf("IMSM_2TB_DISKS=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB_DISK) ? "yes" : "no"); | |
1950 | printf("IMSM_MAX_DISKS=%d\n",orom->tds); | |
1951 | printf("IMSM_MAX_VOLUMES_PER_ARRAY=%d\n",orom->vpa); | |
1952 | printf("IMSM_MAX_VOLUMES_PER_CONTROLLER=%d\n",orom->vphba); | |
1953 | } | |
1954 | ||
9eafa1de | 1955 | static int detail_platform_imsm(int verbose, int enumerate_only, char *controller_path) |
d665cc31 DW |
1956 | { |
1957 | /* There are two components to imsm platform support, the ahci SATA | |
1958 | * controller and the option-rom. To find the SATA controller we | |
1959 | * simply look in /sys/bus/pci/drivers/ahci to see if an ahci | |
1960 | * controller with the Intel vendor id is present. This approach | |
1961 | * allows mdadm to leverage the kernel's ahci detection logic, with the | |
1962 | * caveat that if ahci.ko is not loaded mdadm will not be able to | |
1963 | * detect platform raid capabilities. The option-rom resides in a | |
1964 | * platform "Adapter ROM". We scan for its signature to retrieve the | |
1965 | * platform capabilities. If raid support is disabled in the BIOS the | |
1966 | * option-rom capability structure will not be available. | |
1967 | */ | |
d665cc31 | 1968 | struct sys_dev *list, *hba; |
d665cc31 DW |
1969 | int host_base = 0; |
1970 | int port_count = 0; | |
9eafa1de | 1971 | int result=1; |
d665cc31 | 1972 | |
5615172f | 1973 | if (enumerate_only) { |
a891a3c2 | 1974 | if (check_env("IMSM_NO_PLATFORM")) |
5615172f | 1975 | return 0; |
a891a3c2 LM |
1976 | list = find_intel_devices(); |
1977 | if (!list) | |
1978 | return 2; | |
1979 | for (hba = list; hba; hba = hba->next) { | |
6b781d33 AP |
1980 | if (find_imsm_capability(hba)) { |
1981 | result = 0; | |
a891a3c2 LM |
1982 | break; |
1983 | } | |
9eafa1de | 1984 | else |
6b781d33 | 1985 | result = 2; |
a891a3c2 | 1986 | } |
a891a3c2 | 1987 | return result; |
5615172f DW |
1988 | } |
1989 | ||
155cbb4c LM |
1990 | list = find_intel_devices(); |
1991 | if (!list) { | |
ba728be7 | 1992 | if (verbose > 0) |
7a862a02 | 1993 | pr_err("no active Intel(R) RAID controller found.\n"); |
d665cc31 | 1994 | return 2; |
ba728be7 | 1995 | } else if (verbose > 0) |
155cbb4c | 1996 | print_found_intel_controllers(list); |
d665cc31 | 1997 | |
a891a3c2 | 1998 | for (hba = list; hba; hba = hba->next) { |
0858eccf | 1999 | if (controller_path && (compare_paths(hba->path, controller_path) != 0)) |
9eafa1de | 2000 | continue; |
0858eccf | 2001 | if (!find_imsm_capability(hba)) { |
60f0f54d | 2002 | char buf[PATH_MAX]; |
e7b84f9d | 2003 | pr_err("imsm capabilities not found for controller: %s (type %s)\n", |
60f0f54d PB |
2004 | hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path, |
2005 | get_sys_dev_type(hba->type)); | |
0858eccf AP |
2006 | continue; |
2007 | } | |
2008 | result = 0; | |
2009 | } | |
2010 | ||
2011 | if (controller_path && result == 1) { | |
2012 | pr_err("no active Intel(R) RAID controller found under %s\n", | |
2013 | controller_path); | |
2014 | return result; | |
2015 | } | |
2016 | ||
5e1d6128 | 2017 | const struct orom_entry *entry; |
0858eccf | 2018 | |
5e1d6128 | 2019 | for (entry = orom_entries; entry; entry = entry->next) { |
60f0f54d PB |
2020 | if (entry->type == SYS_DEV_VMD) { |
2021 | for (hba = list; hba; hba = hba->next) { | |
2022 | if (hba->type == SYS_DEV_VMD) { | |
2023 | char buf[PATH_MAX]; | |
2024 | print_imsm_capability(&entry->orom); | |
2025 | printf(" I/O Controller : %s (%s)\n", | |
2026 | vmd_domain_to_controller(hba, buf), get_sys_dev_type(hba->type)); | |
2027 | print_vmd_attached_devs(hba); | |
2028 | printf("\n"); | |
2029 | } | |
2030 | } | |
2031 | continue; | |
2032 | } | |
0858eccf | 2033 | |
60f0f54d PB |
2034 | print_imsm_capability(&entry->orom); |
2035 | if (entry->type == SYS_DEV_NVME) { | |
0858eccf AP |
2036 | for (hba = list; hba; hba = hba->next) { |
2037 | if (hba->type == SYS_DEV_NVME) | |
2038 | printf(" NVMe Device : %s\n", hba->path); | |
2039 | } | |
60f0f54d | 2040 | printf("\n"); |
0858eccf AP |
2041 | continue; |
2042 | } | |
2043 | ||
2044 | struct devid_list *devid; | |
5e1d6128 | 2045 | for (devid = entry->devid_list; devid; devid = devid->next) { |
0858eccf AP |
2046 | hba = device_by_id(devid->devid); |
2047 | if (!hba) | |
2048 | continue; | |
2049 | ||
9eafa1de MN |
2050 | printf(" I/O Controller : %s (%s)\n", |
2051 | hba->path, get_sys_dev_type(hba->type)); | |
2052 | if (hba->type == SYS_DEV_SATA) { | |
2053 | host_base = ahci_get_port_count(hba->path, &port_count); | |
2054 | if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) { | |
2055 | if (verbose > 0) | |
7a862a02 | 2056 | pr_err("failed to enumerate ports on SATA controller at %s.\n", hba->pci_id); |
9eafa1de MN |
2057 | result |= 2; |
2058 | } | |
120dc887 LM |
2059 | } |
2060 | } | |
0858eccf | 2061 | printf("\n"); |
d665cc31 | 2062 | } |
155cbb4c | 2063 | |
120dc887 | 2064 | return result; |
d665cc31 | 2065 | } |
e50cf220 | 2066 | |
9eafa1de | 2067 | static int export_detail_platform_imsm(int verbose, char *controller_path) |
e50cf220 | 2068 | { |
e50cf220 MN |
2069 | struct sys_dev *list, *hba; |
2070 | int result=1; | |
2071 | ||
2072 | list = find_intel_devices(); | |
2073 | if (!list) { | |
2074 | if (verbose > 0) | |
2075 | pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_INTEL_DEVICES\n"); | |
2076 | result = 2; | |
e50cf220 MN |
2077 | return result; |
2078 | } | |
2079 | ||
2080 | for (hba = list; hba; hba = hba->next) { | |
9eafa1de MN |
2081 | if (controller_path && (compare_paths(hba->path,controller_path) != 0)) |
2082 | continue; | |
60f0f54d PB |
2083 | if (!find_imsm_capability(hba) && verbose > 0) { |
2084 | char buf[PATH_MAX]; | |
2085 | pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_IMSM_CAPABLE_DEVICE_UNDER_%s\n", | |
2086 | hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path); | |
2087 | } | |
0858eccf | 2088 | else |
e50cf220 | 2089 | result = 0; |
e50cf220 MN |
2090 | } |
2091 | ||
5e1d6128 | 2092 | const struct orom_entry *entry; |
0858eccf | 2093 | |
60f0f54d PB |
2094 | for (entry = orom_entries; entry; entry = entry->next) { |
2095 | if (entry->type == SYS_DEV_VMD) { | |
2096 | for (hba = list; hba; hba = hba->next) | |
2097 | print_imsm_capability_export(&entry->orom); | |
2098 | continue; | |
2099 | } | |
5e1d6128 | 2100 | print_imsm_capability_export(&entry->orom); |
60f0f54d | 2101 | } |
0858eccf | 2102 | |
e50cf220 MN |
2103 | return result; |
2104 | } | |
2105 | ||
cdddbdbc DW |
2106 | #endif |
2107 | ||
2108 | static int match_home_imsm(struct supertype *st, char *homehost) | |
2109 | { | |
5115ca67 DW |
2110 | /* the imsm metadata format does not specify any host |
2111 | * identification information. We return -1 since we can never | |
2112 | * confirm nor deny whether a given array is "meant" for this | |
148acb7b | 2113 | * host. We rely on compare_super and the 'family_num' fields to |
5115ca67 DW |
2114 | * exclude member disks that do not belong, and we rely on |
2115 | * mdadm.conf to specify the arrays that should be assembled. | |
2116 | * Auto-assembly may still pick up "foreign" arrays. | |
2117 | */ | |
cdddbdbc | 2118 | |
9362c1c8 | 2119 | return -1; |
cdddbdbc DW |
2120 | } |
2121 | ||
2122 | static void uuid_from_super_imsm(struct supertype *st, int uuid[4]) | |
2123 | { | |
51006d85 N |
2124 | /* The uuid returned here is used for: |
2125 | * uuid to put into bitmap file (Create, Grow) | |
2126 | * uuid for backup header when saving critical section (Grow) | |
2127 | * comparing uuids when re-adding a device into an array | |
2128 | * In these cases the uuid required is that of the data-array, | |
2129 | * not the device-set. | |
2130 | * uuid to recognise same set when adding a missing device back | |
2131 | * to an array. This is a uuid for the device-set. | |
1011e834 | 2132 | * |
51006d85 N |
2133 | * For each of these we can make do with a truncated |
2134 | * or hashed uuid rather than the original, as long as | |
2135 | * everyone agrees. | |
2136 | * In each case the uuid required is that of the data-array, | |
2137 | * not the device-set. | |
43dad3d6 | 2138 | */ |
51006d85 N |
2139 | /* imsm does not track uuid's so we synthesis one using sha1 on |
2140 | * - The signature (Which is constant for all imsm array, but no matter) | |
148acb7b | 2141 | * - the orig_family_num of the container |
51006d85 N |
2142 | * - the index number of the volume |
2143 | * - the 'serial' number of the volume. | |
2144 | * Hopefully these are all constant. | |
2145 | */ | |
2146 | struct intel_super *super = st->sb; | |
43dad3d6 | 2147 | |
51006d85 N |
2148 | char buf[20]; |
2149 | struct sha1_ctx ctx; | |
2150 | struct imsm_dev *dev = NULL; | |
148acb7b | 2151 | __u32 family_num; |
51006d85 | 2152 | |
148acb7b DW |
2153 | /* some mdadm versions failed to set ->orig_family_num, in which |
2154 | * case fall back to ->family_num. orig_family_num will be | |
2155 | * fixed up with the first metadata update. | |
2156 | */ | |
2157 | family_num = super->anchor->orig_family_num; | |
2158 | if (family_num == 0) | |
2159 | family_num = super->anchor->family_num; | |
51006d85 | 2160 | sha1_init_ctx(&ctx); |
92bd8f8d | 2161 | sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx); |
148acb7b | 2162 | sha1_process_bytes(&family_num, sizeof(__u32), &ctx); |
51006d85 N |
2163 | if (super->current_vol >= 0) |
2164 | dev = get_imsm_dev(super, super->current_vol); | |
2165 | if (dev) { | |
2166 | __u32 vol = super->current_vol; | |
2167 | sha1_process_bytes(&vol, sizeof(vol), &ctx); | |
2168 | sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx); | |
2169 | } | |
2170 | sha1_finish_ctx(&ctx, buf); | |
2171 | memcpy(uuid, buf, 4*4); | |
cdddbdbc DW |
2172 | } |
2173 | ||
0d481d37 | 2174 | #if 0 |
4f5bc454 DW |
2175 | static void |
2176 | get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p) | |
cdddbdbc | 2177 | { |
cdddbdbc DW |
2178 | __u8 *v = get_imsm_version(mpb); |
2179 | __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH; | |
2180 | char major[] = { 0, 0, 0 }; | |
2181 | char minor[] = { 0 ,0, 0 }; | |
2182 | char patch[] = { 0, 0, 0 }; | |
2183 | char *ver_parse[] = { major, minor, patch }; | |
2184 | int i, j; | |
2185 | ||
2186 | i = j = 0; | |
2187 | while (*v != '\0' && v < end) { | |
2188 | if (*v != '.' && j < 2) | |
2189 | ver_parse[i][j++] = *v; | |
2190 | else { | |
2191 | i++; | |
2192 | j = 0; | |
2193 | } | |
2194 | v++; | |
2195 | } | |
2196 | ||
4f5bc454 DW |
2197 | *m = strtol(minor, NULL, 0); |
2198 | *p = strtol(patch, NULL, 0); | |
2199 | } | |
0d481d37 | 2200 | #endif |
4f5bc454 | 2201 | |
1e5c6983 DW |
2202 | static __u32 migr_strip_blocks_resync(struct imsm_dev *dev) |
2203 | { | |
2204 | /* migr_strip_size when repairing or initializing parity */ | |
238c0a71 | 2205 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2206 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2207 | ||
2208 | switch (get_imsm_raid_level(map)) { | |
2209 | case 5: | |
2210 | case 10: | |
2211 | return chunk; | |
2212 | default: | |
2213 | return 128*1024 >> 9; | |
2214 | } | |
2215 | } | |
2216 | ||
2217 | static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev) | |
2218 | { | |
2219 | /* migr_strip_size when rebuilding a degraded disk, no idea why | |
2220 | * this is different than migr_strip_size_resync(), but it's good | |
2221 | * to be compatible | |
2222 | */ | |
238c0a71 | 2223 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2224 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2225 | ||
2226 | switch (get_imsm_raid_level(map)) { | |
2227 | case 1: | |
2228 | case 10: | |
2229 | if (map->num_members % map->num_domains == 0) | |
2230 | return 128*1024 >> 9; | |
2231 | else | |
2232 | return chunk; | |
2233 | case 5: | |
2234 | return max((__u32) 64*1024 >> 9, chunk); | |
2235 | default: | |
2236 | return 128*1024 >> 9; | |
2237 | } | |
2238 | } | |
2239 | ||
2240 | static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev) | |
2241 | { | |
238c0a71 AK |
2242 | struct imsm_map *lo = get_imsm_map(dev, MAP_0); |
2243 | struct imsm_map *hi = get_imsm_map(dev, MAP_1); | |
1e5c6983 DW |
2244 | __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip); |
2245 | __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip); | |
2246 | ||
2247 | return max((__u32) 1, hi_chunk / lo_chunk); | |
2248 | } | |
2249 | ||
2250 | static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev) | |
2251 | { | |
238c0a71 | 2252 | struct imsm_map *lo = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2253 | int level = get_imsm_raid_level(lo); |
2254 | ||
2255 | if (level == 1 || level == 10) { | |
238c0a71 | 2256 | struct imsm_map *hi = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2257 | |
2258 | return hi->num_domains; | |
2259 | } else | |
2260 | return num_stripes_per_unit_resync(dev); | |
2261 | } | |
2262 | ||
98130f40 | 2263 | static __u8 imsm_num_data_members(struct imsm_dev *dev, int second_map) |
1e5c6983 DW |
2264 | { |
2265 | /* named 'imsm_' because raid0, raid1 and raid10 | |
2266 | * counter-intuitively have the same number of data disks | |
2267 | */ | |
98130f40 | 2268 | struct imsm_map *map = get_imsm_map(dev, second_map); |
1e5c6983 DW |
2269 | |
2270 | switch (get_imsm_raid_level(map)) { | |
2271 | case 0: | |
36fd8ccc AK |
2272 | return map->num_members; |
2273 | break; | |
1e5c6983 DW |
2274 | case 1: |
2275 | case 10: | |
36fd8ccc | 2276 | return map->num_members/2; |
1e5c6983 DW |
2277 | case 5: |
2278 | return map->num_members - 1; | |
2279 | default: | |
1ade5cc1 | 2280 | dprintf("unsupported raid level\n"); |
1e5c6983 DW |
2281 | return 0; |
2282 | } | |
2283 | } | |
2284 | ||
2285 | static __u32 parity_segment_depth(struct imsm_dev *dev) | |
2286 | { | |
238c0a71 | 2287 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2288 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2289 | ||
2290 | switch(get_imsm_raid_level(map)) { | |
2291 | case 1: | |
2292 | case 10: | |
2293 | return chunk * map->num_domains; | |
2294 | case 5: | |
2295 | return chunk * map->num_members; | |
2296 | default: | |
2297 | return chunk; | |
2298 | } | |
2299 | } | |
2300 | ||
2301 | static __u32 map_migr_block(struct imsm_dev *dev, __u32 block) | |
2302 | { | |
238c0a71 | 2303 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2304 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2305 | __u32 strip = block / chunk; | |
2306 | ||
2307 | switch (get_imsm_raid_level(map)) { | |
2308 | case 1: | |
2309 | case 10: { | |
2310 | __u32 vol_strip = (strip * map->num_domains) + 1; | |
2311 | __u32 vol_stripe = vol_strip / map->num_members; | |
2312 | ||
2313 | return vol_stripe * chunk + block % chunk; | |
2314 | } case 5: { | |
2315 | __u32 stripe = strip / (map->num_members - 1); | |
2316 | ||
2317 | return stripe * chunk + block % chunk; | |
2318 | } | |
2319 | default: | |
2320 | return 0; | |
2321 | } | |
2322 | } | |
2323 | ||
c47b0ff6 AK |
2324 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
2325 | struct imsm_dev *dev) | |
1e5c6983 DW |
2326 | { |
2327 | /* calculate the conversion factor between per member 'blocks' | |
2328 | * (md/{resync,rebuild}_start) and imsm migration units, return | |
2329 | * 0 for the 'not migrating' and 'unsupported migration' cases | |
2330 | */ | |
2331 | if (!dev->vol.migr_state) | |
2332 | return 0; | |
2333 | ||
2334 | switch (migr_type(dev)) { | |
c47b0ff6 AK |
2335 | case MIGR_GEN_MIGR: { |
2336 | struct migr_record *migr_rec = super->migr_rec; | |
2337 | return __le32_to_cpu(migr_rec->blocks_per_unit); | |
2338 | } | |
1e5c6983 DW |
2339 | case MIGR_VERIFY: |
2340 | case MIGR_REPAIR: | |
2341 | case MIGR_INIT: { | |
238c0a71 | 2342 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2343 | __u32 stripes_per_unit; |
2344 | __u32 blocks_per_unit; | |
2345 | __u32 parity_depth; | |
2346 | __u32 migr_chunk; | |
2347 | __u32 block_map; | |
2348 | __u32 block_rel; | |
2349 | __u32 segment; | |
2350 | __u32 stripe; | |
2351 | __u8 disks; | |
2352 | ||
2353 | /* yes, this is really the translation of migr_units to | |
2354 | * per-member blocks in the 'resync' case | |
2355 | */ | |
2356 | stripes_per_unit = num_stripes_per_unit_resync(dev); | |
2357 | migr_chunk = migr_strip_blocks_resync(dev); | |
238c0a71 | 2358 | disks = imsm_num_data_members(dev, MAP_0); |
1e5c6983 | 2359 | blocks_per_unit = stripes_per_unit * migr_chunk * disks; |
7b1ab482 | 2360 | stripe = __le16_to_cpu(map->blocks_per_strip) * disks; |
1e5c6983 DW |
2361 | segment = blocks_per_unit / stripe; |
2362 | block_rel = blocks_per_unit - segment * stripe; | |
2363 | parity_depth = parity_segment_depth(dev); | |
2364 | block_map = map_migr_block(dev, block_rel); | |
2365 | return block_map + parity_depth * segment; | |
2366 | } | |
2367 | case MIGR_REBUILD: { | |
2368 | __u32 stripes_per_unit; | |
2369 | __u32 migr_chunk; | |
2370 | ||
2371 | stripes_per_unit = num_stripes_per_unit_rebuild(dev); | |
2372 | migr_chunk = migr_strip_blocks_rebuild(dev); | |
2373 | return migr_chunk * stripes_per_unit; | |
2374 | } | |
1e5c6983 DW |
2375 | case MIGR_STATE_CHANGE: |
2376 | default: | |
2377 | return 0; | |
2378 | } | |
2379 | } | |
2380 | ||
c2c087e6 DW |
2381 | static int imsm_level_to_layout(int level) |
2382 | { | |
2383 | switch (level) { | |
2384 | case 0: | |
2385 | case 1: | |
2386 | return 0; | |
2387 | case 5: | |
2388 | case 6: | |
a380c027 | 2389 | return ALGORITHM_LEFT_ASYMMETRIC; |
c2c087e6 | 2390 | case 10: |
c92a2527 | 2391 | return 0x102; |
c2c087e6 | 2392 | } |
a18a888e | 2393 | return UnSet; |
c2c087e6 DW |
2394 | } |
2395 | ||
8e59f3d8 AK |
2396 | /******************************************************************************* |
2397 | * Function: read_imsm_migr_rec | |
2398 | * Description: Function reads imsm migration record from last sector of disk | |
2399 | * Parameters: | |
2400 | * fd : disk descriptor | |
2401 | * super : metadata info | |
2402 | * Returns: | |
2403 | * 0 : success, | |
2404 | * -1 : fail | |
2405 | ******************************************************************************/ | |
2406 | static int read_imsm_migr_rec(int fd, struct intel_super *super) | |
2407 | { | |
2408 | int ret_val = -1; | |
2409 | unsigned long long dsize; | |
2410 | ||
2411 | get_dev_size(fd, NULL, &dsize); | |
17a4eaf9 | 2412 | if (lseek64(fd, dsize - MIGR_REC_POSITION, SEEK_SET) < 0) { |
e7b84f9d N |
2413 | pr_err("Cannot seek to anchor block: %s\n", |
2414 | strerror(errno)); | |
8e59f3d8 AK |
2415 | goto out; |
2416 | } | |
17a4eaf9 AK |
2417 | if (read(fd, super->migr_rec_buf, MIGR_REC_BUF_SIZE) != |
2418 | MIGR_REC_BUF_SIZE) { | |
e7b84f9d N |
2419 | pr_err("Cannot read migr record block: %s\n", |
2420 | strerror(errno)); | |
8e59f3d8 AK |
2421 | goto out; |
2422 | } | |
2423 | ret_val = 0; | |
2424 | ||
2425 | out: | |
2426 | return ret_val; | |
2427 | } | |
2428 | ||
3136abe5 AK |
2429 | static struct imsm_dev *imsm_get_device_during_migration( |
2430 | struct intel_super *super) | |
2431 | { | |
2432 | ||
2433 | struct intel_dev *dv; | |
2434 | ||
2435 | for (dv = super->devlist; dv; dv = dv->next) { | |
2436 | if (is_gen_migration(dv->dev)) | |
2437 | return dv->dev; | |
2438 | } | |
2439 | return NULL; | |
2440 | } | |
2441 | ||
8e59f3d8 AK |
2442 | /******************************************************************************* |
2443 | * Function: load_imsm_migr_rec | |
2444 | * Description: Function reads imsm migration record (it is stored at the last | |
2445 | * sector of disk) | |
2446 | * Parameters: | |
2447 | * super : imsm internal array info | |
2448 | * info : general array info | |
2449 | * Returns: | |
2450 | * 0 : success | |
2451 | * -1 : fail | |
4c965cc9 | 2452 | * -2 : no migration in progress |
8e59f3d8 AK |
2453 | ******************************************************************************/ |
2454 | static int load_imsm_migr_rec(struct intel_super *super, struct mdinfo *info) | |
2455 | { | |
2456 | struct mdinfo *sd; | |
2457 | struct dl *dl = NULL; | |
2458 | char nm[30]; | |
2459 | int retval = -1; | |
2460 | int fd = -1; | |
3136abe5 AK |
2461 | struct imsm_dev *dev; |
2462 | struct imsm_map *map = NULL; | |
b4ab44d8 | 2463 | int slot = -1; |
3136abe5 AK |
2464 | |
2465 | /* find map under migration */ | |
2466 | dev = imsm_get_device_during_migration(super); | |
2467 | /* nothing to load,no migration in progress? | |
2468 | */ | |
2469 | if (dev == NULL) | |
4c965cc9 | 2470 | return -2; |
44bfe6df | 2471 | map = get_imsm_map(dev, MAP_0); |
8e59f3d8 AK |
2472 | |
2473 | if (info) { | |
2474 | for (sd = info->devs ; sd ; sd = sd->next) { | |
3136abe5 AK |
2475 | /* skip spare and failed disks |
2476 | */ | |
2477 | if (sd->disk.raid_disk < 0) | |
2478 | continue; | |
8e59f3d8 | 2479 | /* read only from one of the first two slots */ |
3136abe5 AK |
2480 | if (map) |
2481 | slot = get_imsm_disk_slot(map, | |
2482 | sd->disk.raid_disk); | |
2483 | if ((map == NULL) || (slot > 1) || (slot < 0)) | |
8e59f3d8 | 2484 | continue; |
3136abe5 | 2485 | |
8e59f3d8 AK |
2486 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); |
2487 | fd = dev_open(nm, O_RDONLY); | |
2488 | if (fd >= 0) | |
2489 | break; | |
2490 | } | |
2491 | } | |
2492 | if (fd < 0) { | |
2493 | for (dl = super->disks; dl; dl = dl->next) { | |
3136abe5 AK |
2494 | /* skip spare and failed disks |
2495 | */ | |
2496 | if (dl->index < 0) | |
2497 | continue; | |
8e59f3d8 | 2498 | /* read only from one of the first two slots */ |
3136abe5 AK |
2499 | if (map) |
2500 | slot = get_imsm_disk_slot(map, dl->index); | |
2501 | if ((map == NULL) || (slot > 1) || (slot < 0)) | |
8e59f3d8 AK |
2502 | continue; |
2503 | sprintf(nm, "%d:%d", dl->major, dl->minor); | |
2504 | fd = dev_open(nm, O_RDONLY); | |
2505 | if (fd >= 0) | |
2506 | break; | |
2507 | } | |
2508 | } | |
2509 | if (fd < 0) | |
2510 | goto out; | |
2511 | retval = read_imsm_migr_rec(fd, super); | |
2512 | ||
2513 | out: | |
2514 | if (fd >= 0) | |
2515 | close(fd); | |
2516 | return retval; | |
2517 | } | |
2518 | ||
9e2d750d | 2519 | #ifndef MDASSEMBLE |
c17608ea AK |
2520 | /******************************************************************************* |
2521 | * function: imsm_create_metadata_checkpoint_update | |
2522 | * Description: It creates update for checkpoint change. | |
2523 | * Parameters: | |
2524 | * super : imsm internal array info | |
2525 | * u : pointer to prepared update | |
2526 | * Returns: | |
2527 | * Uptate length. | |
2528 | * If length is equal to 0, input pointer u contains no update | |
2529 | ******************************************************************************/ | |
2530 | static int imsm_create_metadata_checkpoint_update( | |
2531 | struct intel_super *super, | |
2532 | struct imsm_update_general_migration_checkpoint **u) | |
2533 | { | |
2534 | ||
2535 | int update_memory_size = 0; | |
2536 | ||
1ade5cc1 | 2537 | dprintf("(enter)\n"); |
c17608ea AK |
2538 | |
2539 | if (u == NULL) | |
2540 | return 0; | |
2541 | *u = NULL; | |
2542 | ||
2543 | /* size of all update data without anchor */ | |
2544 | update_memory_size = | |
2545 | sizeof(struct imsm_update_general_migration_checkpoint); | |
2546 | ||
503975b9 | 2547 | *u = xcalloc(1, update_memory_size); |
c17608ea | 2548 | if (*u == NULL) { |
1ade5cc1 | 2549 | dprintf("error: cannot get memory\n"); |
c17608ea AK |
2550 | return 0; |
2551 | } | |
2552 | (*u)->type = update_general_migration_checkpoint; | |
2553 | (*u)->curr_migr_unit = __le32_to_cpu(super->migr_rec->curr_migr_unit); | |
1ade5cc1 | 2554 | dprintf("prepared for %u\n", (*u)->curr_migr_unit); |
c17608ea AK |
2555 | |
2556 | return update_memory_size; | |
2557 | } | |
2558 | ||
c17608ea AK |
2559 | static void imsm_update_metadata_locally(struct supertype *st, |
2560 | void *buf, int len); | |
2561 | ||
687629c2 AK |
2562 | /******************************************************************************* |
2563 | * Function: write_imsm_migr_rec | |
2564 | * Description: Function writes imsm migration record | |
2565 | * (at the last sector of disk) | |
2566 | * Parameters: | |
2567 | * super : imsm internal array info | |
2568 | * Returns: | |
2569 | * 0 : success | |
2570 | * -1 : if fail | |
2571 | ******************************************************************************/ | |
2572 | static int write_imsm_migr_rec(struct supertype *st) | |
2573 | { | |
2574 | struct intel_super *super = st->sb; | |
2575 | unsigned long long dsize; | |
2576 | char nm[30]; | |
2577 | int fd = -1; | |
2578 | int retval = -1; | |
2579 | struct dl *sd; | |
c17608ea AK |
2580 | int len; |
2581 | struct imsm_update_general_migration_checkpoint *u; | |
3136abe5 AK |
2582 | struct imsm_dev *dev; |
2583 | struct imsm_map *map = NULL; | |
2584 | ||
2585 | /* find map under migration */ | |
2586 | dev = imsm_get_device_during_migration(super); | |
2587 | /* if no migration, write buffer anyway to clear migr_record | |
2588 | * on disk based on first available device | |
2589 | */ | |
2590 | if (dev == NULL) | |
2591 | dev = get_imsm_dev(super, super->current_vol < 0 ? 0 : | |
2592 | super->current_vol); | |
2593 | ||
44bfe6df | 2594 | map = get_imsm_map(dev, MAP_0); |
687629c2 AK |
2595 | |
2596 | for (sd = super->disks ; sd ; sd = sd->next) { | |
b4ab44d8 | 2597 | int slot = -1; |
3136abe5 AK |
2598 | |
2599 | /* skip failed and spare devices */ | |
2600 | if (sd->index < 0) | |
2601 | continue; | |
687629c2 | 2602 | /* write to 2 first slots only */ |
3136abe5 AK |
2603 | if (map) |
2604 | slot = get_imsm_disk_slot(map, sd->index); | |
2605 | if ((map == NULL) || (slot > 1) || (slot < 0)) | |
687629c2 | 2606 | continue; |
3136abe5 | 2607 | |
687629c2 AK |
2608 | sprintf(nm, "%d:%d", sd->major, sd->minor); |
2609 | fd = dev_open(nm, O_RDWR); | |
2610 | if (fd < 0) | |
2611 | continue; | |
2612 | get_dev_size(fd, NULL, &dsize); | |
17a4eaf9 | 2613 | if (lseek64(fd, dsize - MIGR_REC_POSITION, SEEK_SET) < 0) { |
e7b84f9d N |
2614 | pr_err("Cannot seek to anchor block: %s\n", |
2615 | strerror(errno)); | |
687629c2 AK |
2616 | goto out; |
2617 | } | |
17a4eaf9 AK |
2618 | if (write(fd, super->migr_rec_buf, MIGR_REC_BUF_SIZE) != |
2619 | MIGR_REC_BUF_SIZE) { | |
e7b84f9d N |
2620 | pr_err("Cannot write migr record block: %s\n", |
2621 | strerror(errno)); | |
687629c2 AK |
2622 | goto out; |
2623 | } | |
2624 | close(fd); | |
2625 | fd = -1; | |
2626 | } | |
c17608ea AK |
2627 | /* update checkpoint information in metadata */ |
2628 | len = imsm_create_metadata_checkpoint_update(super, &u); | |
2629 | ||
2630 | if (len <= 0) { | |
2631 | dprintf("imsm: Cannot prepare update\n"); | |
2632 | goto out; | |
2633 | } | |
2634 | /* update metadata locally */ | |
2635 | imsm_update_metadata_locally(st, u, len); | |
2636 | /* and possibly remotely */ | |
2637 | if (st->update_tail) { | |
2638 | append_metadata_update(st, u, len); | |
2639 | /* during reshape we do all work inside metadata handler | |
2640 | * manage_reshape(), so metadata update has to be triggered | |
2641 | * insida it | |
2642 | */ | |
2643 | flush_metadata_updates(st); | |
2644 | st->update_tail = &st->updates; | |
2645 | } else | |
2646 | free(u); | |
687629c2 AK |
2647 | |
2648 | retval = 0; | |
2649 | out: | |
2650 | if (fd >= 0) | |
2651 | close(fd); | |
2652 | return retval; | |
2653 | } | |
9e2d750d | 2654 | #endif /* MDASSEMBLE */ |
687629c2 | 2655 | |
e2962bfc AK |
2656 | /* spare/missing disks activations are not allowe when |
2657 | * array/container performs reshape operation, because | |
2658 | * all arrays in container works on the same disks set | |
2659 | */ | |
2660 | int imsm_reshape_blocks_arrays_changes(struct intel_super *super) | |
2661 | { | |
2662 | int rv = 0; | |
2663 | struct intel_dev *i_dev; | |
2664 | struct imsm_dev *dev; | |
2665 | ||
2666 | /* check whole container | |
2667 | */ | |
2668 | for (i_dev = super->devlist; i_dev; i_dev = i_dev->next) { | |
2669 | dev = i_dev->dev; | |
3ad25638 | 2670 | if (is_gen_migration(dev)) { |
e2962bfc AK |
2671 | /* No repair during any migration in container |
2672 | */ | |
2673 | rv = 1; | |
2674 | break; | |
2675 | } | |
2676 | } | |
2677 | return rv; | |
2678 | } | |
c41e00b2 AK |
2679 | static unsigned long long imsm_component_size_aligment_check(int level, |
2680 | int chunk_size, | |
2681 | unsigned long long component_size) | |
2682 | { | |
2683 | unsigned int component_size_alligment; | |
2684 | ||
2685 | /* check component size aligment | |
2686 | */ | |
2687 | component_size_alligment = component_size % (chunk_size/512); | |
2688 | ||
1ade5cc1 | 2689 | dprintf("(Level: %i, chunk_size = %i, component_size = %llu), component_size_alligment = %u\n", |
c41e00b2 AK |
2690 | level, chunk_size, component_size, |
2691 | component_size_alligment); | |
2692 | ||
2693 | if (component_size_alligment && (level != 1) && (level != UnSet)) { | |
2694 | dprintf("imsm: reported component size alligned from %llu ", | |
2695 | component_size); | |
2696 | component_size -= component_size_alligment; | |
1ade5cc1 | 2697 | dprintf_cont("to %llu (%i).\n", |
c41e00b2 AK |
2698 | component_size, component_size_alligment); |
2699 | } | |
2700 | ||
2701 | return component_size; | |
2702 | } | |
e2962bfc | 2703 | |
a5d85af7 | 2704 | static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap) |
bf5a934a DW |
2705 | { |
2706 | struct intel_super *super = st->sb; | |
c47b0ff6 | 2707 | struct migr_record *migr_rec = super->migr_rec; |
949c47a0 | 2708 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); |
238c0a71 AK |
2709 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
2710 | struct imsm_map *prev_map = get_imsm_map(dev, MAP_1); | |
b335e593 | 2711 | struct imsm_map *map_to_analyse = map; |
efb30e7f | 2712 | struct dl *dl; |
a5d85af7 | 2713 | int map_disks = info->array.raid_disks; |
bf5a934a | 2714 | |
95eeceeb | 2715 | memset(info, 0, sizeof(*info)); |
b335e593 AK |
2716 | if (prev_map) |
2717 | map_to_analyse = prev_map; | |
2718 | ||
ca0748fa | 2719 | dl = super->current_disk; |
9894ec0d | 2720 | |
bf5a934a | 2721 | info->container_member = super->current_vol; |
cd0430a1 | 2722 | info->array.raid_disks = map->num_members; |
b335e593 | 2723 | info->array.level = get_imsm_raid_level(map_to_analyse); |
bf5a934a DW |
2724 | info->array.layout = imsm_level_to_layout(info->array.level); |
2725 | info->array.md_minor = -1; | |
2726 | info->array.ctime = 0; | |
2727 | info->array.utime = 0; | |
b335e593 AK |
2728 | info->array.chunk_size = |
2729 | __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9; | |
301406c9 | 2730 | info->array.state = !dev->vol.dirty; |
da9b4a62 DW |
2731 | info->custom_array_size = __le32_to_cpu(dev->size_high); |
2732 | info->custom_array_size <<= 32; | |
2733 | info->custom_array_size |= __le32_to_cpu(dev->size_low); | |
3ad25638 AK |
2734 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
2735 | ||
3f510843 | 2736 | if (is_gen_migration(dev)) { |
3f83228a | 2737 | info->reshape_active = 1; |
b335e593 AK |
2738 | info->new_level = get_imsm_raid_level(map); |
2739 | info->new_layout = imsm_level_to_layout(info->new_level); | |
2740 | info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9; | |
3f83228a | 2741 | info->delta_disks = map->num_members - prev_map->num_members; |
493f5dd6 N |
2742 | if (info->delta_disks) { |
2743 | /* this needs to be applied to every array | |
2744 | * in the container. | |
2745 | */ | |
81219e70 | 2746 | info->reshape_active = CONTAINER_RESHAPE; |
493f5dd6 | 2747 | } |
3f83228a N |
2748 | /* We shape information that we give to md might have to be |
2749 | * modify to cope with md's requirement for reshaping arrays. | |
2750 | * For example, when reshaping a RAID0, md requires it to be | |
2751 | * presented as a degraded RAID4. | |
2752 | * Also if a RAID0 is migrating to a RAID5 we need to specify | |
2753 | * the array as already being RAID5, but the 'before' layout | |
2754 | * is a RAID4-like layout. | |
2755 | */ | |
2756 | switch (info->array.level) { | |
2757 | case 0: | |
2758 | switch(info->new_level) { | |
2759 | case 0: | |
2760 | /* conversion is happening as RAID4 */ | |
2761 | info->array.level = 4; | |
2762 | info->array.raid_disks += 1; | |
2763 | break; | |
2764 | case 5: | |
2765 | /* conversion is happening as RAID5 */ | |
2766 | info->array.level = 5; | |
2767 | info->array.layout = ALGORITHM_PARITY_N; | |
3f83228a N |
2768 | info->delta_disks -= 1; |
2769 | break; | |
2770 | default: | |
2771 | /* FIXME error message */ | |
2772 | info->array.level = UnSet; | |
2773 | break; | |
2774 | } | |
2775 | break; | |
2776 | } | |
b335e593 AK |
2777 | } else { |
2778 | info->new_level = UnSet; | |
2779 | info->new_layout = UnSet; | |
2780 | info->new_chunk = info->array.chunk_size; | |
3f83228a | 2781 | info->delta_disks = 0; |
b335e593 | 2782 | } |
ca0748fa | 2783 | |
efb30e7f DW |
2784 | if (dl) { |
2785 | info->disk.major = dl->major; | |
2786 | info->disk.minor = dl->minor; | |
ca0748fa | 2787 | info->disk.number = dl->index; |
656b6b5a N |
2788 | info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse, |
2789 | dl->index); | |
efb30e7f | 2790 | } |
bf5a934a | 2791 | |
5551b113 CA |
2792 | info->data_offset = pba_of_lba0(map_to_analyse); |
2793 | info->component_size = blocks_per_member(map_to_analyse); | |
139dae11 | 2794 | |
c41e00b2 AK |
2795 | info->component_size = imsm_component_size_aligment_check( |
2796 | info->array.level, | |
2797 | info->array.chunk_size, | |
2798 | info->component_size); | |
139dae11 | 2799 | |
301406c9 | 2800 | memset(info->uuid, 0, sizeof(info->uuid)); |
921d9e16 | 2801 | info->recovery_start = MaxSector; |
bf5a934a | 2802 | |
d2e6d5d6 | 2803 | info->reshape_progress = 0; |
b6796ce1 | 2804 | info->resync_start = MaxSector; |
b9172665 AK |
2805 | if ((map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED || |
2806 | dev->vol.dirty) && | |
2807 | imsm_reshape_blocks_arrays_changes(super) == 0) { | |
301406c9 | 2808 | info->resync_start = 0; |
b6796ce1 AK |
2809 | } |
2810 | if (dev->vol.migr_state) { | |
1e5c6983 DW |
2811 | switch (migr_type(dev)) { |
2812 | case MIGR_REPAIR: | |
2813 | case MIGR_INIT: { | |
c47b0ff6 AK |
2814 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
2815 | dev); | |
1e5c6983 DW |
2816 | __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit); |
2817 | ||
2818 | info->resync_start = blocks_per_unit * units; | |
2819 | break; | |
2820 | } | |
d2e6d5d6 | 2821 | case MIGR_GEN_MIGR: { |
c47b0ff6 AK |
2822 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
2823 | dev); | |
2824 | __u64 units = __le32_to_cpu(migr_rec->curr_migr_unit); | |
04fa9523 AK |
2825 | unsigned long long array_blocks; |
2826 | int used_disks; | |
d2e6d5d6 | 2827 | |
befb629b AK |
2828 | if (__le32_to_cpu(migr_rec->ascending_migr) && |
2829 | (units < | |
2830 | (__le32_to_cpu(migr_rec->num_migr_units)-1)) && | |
2831 | (super->migr_rec->rec_status == | |
2832 | __cpu_to_le32(UNIT_SRC_IN_CP_AREA))) | |
2833 | units++; | |
2834 | ||
d2e6d5d6 | 2835 | info->reshape_progress = blocks_per_unit * units; |
6289d1e0 | 2836 | |
7a862a02 | 2837 | dprintf("IMSM: General Migration checkpoint : %llu (%llu) -> read reshape progress : %llu\n", |
19986c72 MB |
2838 | (unsigned long long)units, |
2839 | (unsigned long long)blocks_per_unit, | |
2840 | info->reshape_progress); | |
75156c46 | 2841 | |
238c0a71 | 2842 | used_disks = imsm_num_data_members(dev, MAP_1); |
75156c46 | 2843 | if (used_disks > 0) { |
5551b113 | 2844 | array_blocks = blocks_per_member(map) * |
75156c46 AK |
2845 | used_disks; |
2846 | /* round array size down to closest MB | |
2847 | */ | |
2848 | info->custom_array_size = (array_blocks | |
2849 | >> SECT_PER_MB_SHIFT) | |
2850 | << SECT_PER_MB_SHIFT; | |
2851 | } | |
d2e6d5d6 | 2852 | } |
1e5c6983 DW |
2853 | case MIGR_VERIFY: |
2854 | /* we could emulate the checkpointing of | |
2855 | * 'sync_action=check' migrations, but for now | |
2856 | * we just immediately complete them | |
2857 | */ | |
2858 | case MIGR_REBUILD: | |
2859 | /* this is handled by container_content_imsm() */ | |
1e5c6983 DW |
2860 | case MIGR_STATE_CHANGE: |
2861 | /* FIXME handle other migrations */ | |
2862 | default: | |
2863 | /* we are not dirty, so... */ | |
2864 | info->resync_start = MaxSector; | |
2865 | } | |
b6796ce1 | 2866 | } |
301406c9 DW |
2867 | |
2868 | strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN); | |
2869 | info->name[MAX_RAID_SERIAL_LEN] = 0; | |
bf5a934a | 2870 | |
f35f2525 N |
2871 | info->array.major_version = -1; |
2872 | info->array.minor_version = -2; | |
4dd2df09 | 2873 | sprintf(info->text_version, "/%s/%d", st->container_devnm, info->container_member); |
a67dd8cc | 2874 | info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */ |
51006d85 | 2875 | uuid_from_super_imsm(st, info->uuid); |
a5d85af7 N |
2876 | |
2877 | if (dmap) { | |
2878 | int i, j; | |
2879 | for (i=0; i<map_disks; i++) { | |
2880 | dmap[i] = 0; | |
2881 | if (i < info->array.raid_disks) { | |
2882 | struct imsm_disk *dsk; | |
238c0a71 | 2883 | j = get_imsm_disk_idx(dev, i, MAP_X); |
a5d85af7 N |
2884 | dsk = get_imsm_disk(super, j); |
2885 | if (dsk && (dsk->status & CONFIGURED_DISK)) | |
2886 | dmap[i] = 1; | |
2887 | } | |
2888 | } | |
2889 | } | |
81ac8b4d | 2890 | } |
bf5a934a | 2891 | |
3b451610 AK |
2892 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, |
2893 | int failed, int look_in_map); | |
2894 | ||
2895 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev, | |
2896 | int look_in_map); | |
2897 | ||
b4ab44d8 | 2898 | #ifndef MDASSEMBLE |
3b451610 AK |
2899 | static void manage_second_map(struct intel_super *super, struct imsm_dev *dev) |
2900 | { | |
2901 | if (is_gen_migration(dev)) { | |
2902 | int failed; | |
2903 | __u8 map_state; | |
2904 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); | |
2905 | ||
2906 | failed = imsm_count_failed(super, dev, MAP_1); | |
238c0a71 | 2907 | map_state = imsm_check_degraded(super, dev, failed, MAP_1); |
3b451610 AK |
2908 | if (map2->map_state != map_state) { |
2909 | map2->map_state = map_state; | |
2910 | super->updates_pending++; | |
2911 | } | |
2912 | } | |
2913 | } | |
b4ab44d8 | 2914 | #endif |
97b4d0e9 DW |
2915 | |
2916 | static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index) | |
2917 | { | |
2918 | struct dl *d; | |
2919 | ||
2920 | for (d = super->missing; d; d = d->next) | |
2921 | if (d->index == index) | |
2922 | return &d->disk; | |
2923 | return NULL; | |
2924 | } | |
2925 | ||
a5d85af7 | 2926 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map) |
4f5bc454 DW |
2927 | { |
2928 | struct intel_super *super = st->sb; | |
4f5bc454 | 2929 | struct imsm_disk *disk; |
a5d85af7 | 2930 | int map_disks = info->array.raid_disks; |
ab3cb6b3 N |
2931 | int max_enough = -1; |
2932 | int i; | |
2933 | struct imsm_super *mpb; | |
4f5bc454 | 2934 | |
bf5a934a | 2935 | if (super->current_vol >= 0) { |
a5d85af7 | 2936 | getinfo_super_imsm_volume(st, info, map); |
bf5a934a DW |
2937 | return; |
2938 | } | |
95eeceeb | 2939 | memset(info, 0, sizeof(*info)); |
d23fe947 DW |
2940 | |
2941 | /* Set raid_disks to zero so that Assemble will always pull in valid | |
2942 | * spares | |
2943 | */ | |
2944 | info->array.raid_disks = 0; | |
cdddbdbc DW |
2945 | info->array.level = LEVEL_CONTAINER; |
2946 | info->array.layout = 0; | |
2947 | info->array.md_minor = -1; | |
1011e834 | 2948 | info->array.ctime = 0; /* N/A for imsm */ |
cdddbdbc DW |
2949 | info->array.utime = 0; |
2950 | info->array.chunk_size = 0; | |
2951 | ||
2952 | info->disk.major = 0; | |
2953 | info->disk.minor = 0; | |
cdddbdbc | 2954 | info->disk.raid_disk = -1; |
c2c087e6 | 2955 | info->reshape_active = 0; |
f35f2525 N |
2956 | info->array.major_version = -1; |
2957 | info->array.minor_version = -2; | |
c2c087e6 | 2958 | strcpy(info->text_version, "imsm"); |
a67dd8cc | 2959 | info->safe_mode_delay = 0; |
c2c087e6 DW |
2960 | info->disk.number = -1; |
2961 | info->disk.state = 0; | |
c5afc314 | 2962 | info->name[0] = 0; |
921d9e16 | 2963 | info->recovery_start = MaxSector; |
3ad25638 | 2964 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
c2c087e6 | 2965 | |
97b4d0e9 | 2966 | /* do we have the all the insync disks that we expect? */ |
ab3cb6b3 | 2967 | mpb = super->anchor; |
97b4d0e9 | 2968 | |
ab3cb6b3 N |
2969 | for (i = 0; i < mpb->num_raid_devs; i++) { |
2970 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
2971 | int failed, enough, j, missing = 0; | |
2972 | struct imsm_map *map; | |
2973 | __u8 state; | |
97b4d0e9 | 2974 | |
3b451610 AK |
2975 | failed = imsm_count_failed(super, dev, MAP_0); |
2976 | state = imsm_check_degraded(super, dev, failed, MAP_0); | |
238c0a71 | 2977 | map = get_imsm_map(dev, MAP_0); |
ab3cb6b3 N |
2978 | |
2979 | /* any newly missing disks? | |
2980 | * (catches single-degraded vs double-degraded) | |
2981 | */ | |
2982 | for (j = 0; j < map->num_members; j++) { | |
238c0a71 | 2983 | __u32 ord = get_imsm_ord_tbl_ent(dev, j, MAP_0); |
ab3cb6b3 N |
2984 | __u32 idx = ord_to_idx(ord); |
2985 | ||
2986 | if (!(ord & IMSM_ORD_REBUILD) && | |
2987 | get_imsm_missing(super, idx)) { | |
2988 | missing = 1; | |
2989 | break; | |
2990 | } | |
97b4d0e9 | 2991 | } |
ab3cb6b3 N |
2992 | |
2993 | if (state == IMSM_T_STATE_FAILED) | |
2994 | enough = -1; | |
2995 | else if (state == IMSM_T_STATE_DEGRADED && | |
2996 | (state != map->map_state || missing)) | |
2997 | enough = 0; | |
2998 | else /* we're normal, or already degraded */ | |
2999 | enough = 1; | |
d2bde6d3 AK |
3000 | if (is_gen_migration(dev) && missing) { |
3001 | /* during general migration we need all disks | |
3002 | * that process is running on. | |
3003 | * No new missing disk is allowed. | |
3004 | */ | |
3005 | max_enough = -1; | |
3006 | enough = -1; | |
3007 | /* no more checks necessary | |
3008 | */ | |
3009 | break; | |
3010 | } | |
ab3cb6b3 N |
3011 | /* in the missing/failed disk case check to see |
3012 | * if at least one array is runnable | |
3013 | */ | |
3014 | max_enough = max(max_enough, enough); | |
3015 | } | |
1ade5cc1 | 3016 | dprintf("enough: %d\n", max_enough); |
ab3cb6b3 | 3017 | info->container_enough = max_enough; |
97b4d0e9 | 3018 | |
4a04ec6c | 3019 | if (super->disks) { |
14e8215b DW |
3020 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
3021 | ||
b9f594fe | 3022 | disk = &super->disks->disk; |
5551b113 | 3023 | info->data_offset = total_blocks(&super->disks->disk) - reserved; |
14e8215b | 3024 | info->component_size = reserved; |
25ed7e59 | 3025 | info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0; |
df474657 DW |
3026 | /* we don't change info->disk.raid_disk here because |
3027 | * this state will be finalized in mdmon after we have | |
3028 | * found the 'most fresh' version of the metadata | |
3029 | */ | |
25ed7e59 DW |
3030 | info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; |
3031 | info->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
cdddbdbc | 3032 | } |
a575e2a7 DW |
3033 | |
3034 | /* only call uuid_from_super_imsm when this disk is part of a populated container, | |
3035 | * ->compare_super may have updated the 'num_raid_devs' field for spares | |
3036 | */ | |
3037 | if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs) | |
36ba7d48 | 3038 | uuid_from_super_imsm(st, info->uuid); |
22e263f6 AC |
3039 | else |
3040 | memcpy(info->uuid, uuid_zero, sizeof(uuid_zero)); | |
a5d85af7 N |
3041 | |
3042 | /* I don't know how to compute 'map' on imsm, so use safe default */ | |
3043 | if (map) { | |
3044 | int i; | |
3045 | for (i = 0; i < map_disks; i++) | |
3046 | map[i] = 1; | |
3047 | } | |
3048 | ||
cdddbdbc DW |
3049 | } |
3050 | ||
5c4cd5da AC |
3051 | /* allocates memory and fills disk in mdinfo structure |
3052 | * for each disk in array */ | |
3053 | struct mdinfo *getinfo_super_disks_imsm(struct supertype *st) | |
3054 | { | |
3055 | struct mdinfo *mddev = NULL; | |
3056 | struct intel_super *super = st->sb; | |
3057 | struct imsm_disk *disk; | |
3058 | int count = 0; | |
3059 | struct dl *dl; | |
3060 | if (!super || !super->disks) | |
3061 | return NULL; | |
3062 | dl = super->disks; | |
503975b9 | 3063 | mddev = xcalloc(1, sizeof(*mddev)); |
5c4cd5da AC |
3064 | while (dl) { |
3065 | struct mdinfo *tmp; | |
3066 | disk = &dl->disk; | |
503975b9 | 3067 | tmp = xcalloc(1, sizeof(*tmp)); |
5c4cd5da AC |
3068 | if (mddev->devs) |
3069 | tmp->next = mddev->devs; | |
3070 | mddev->devs = tmp; | |
3071 | tmp->disk.number = count++; | |
3072 | tmp->disk.major = dl->major; | |
3073 | tmp->disk.minor = dl->minor; | |
3074 | tmp->disk.state = is_configured(disk) ? | |
3075 | (1 << MD_DISK_ACTIVE) : 0; | |
3076 | tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; | |
3077 | tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
3078 | tmp->disk.raid_disk = -1; | |
3079 | dl = dl->next; | |
3080 | } | |
3081 | return mddev; | |
3082 | } | |
3083 | ||
cdddbdbc DW |
3084 | static int update_super_imsm(struct supertype *st, struct mdinfo *info, |
3085 | char *update, char *devname, int verbose, | |
3086 | int uuid_set, char *homehost) | |
3087 | { | |
f352c545 DW |
3088 | /* For 'assemble' and 'force' we need to return non-zero if any |
3089 | * change was made. For others, the return value is ignored. | |
3090 | * Update options are: | |
3091 | * force-one : This device looks a bit old but needs to be included, | |
3092 | * update age info appropriately. | |
3093 | * assemble: clear any 'faulty' flag to allow this device to | |
3094 | * be assembled. | |
3095 | * force-array: Array is degraded but being forced, mark it clean | |
3096 | * if that will be needed to assemble it. | |
3097 | * | |
3098 | * newdev: not used ???? | |
3099 | * grow: Array has gained a new device - this is currently for | |
3100 | * linear only | |
3101 | * resync: mark as dirty so a resync will happen. | |
3102 | * name: update the name - preserving the homehost | |
6e46bf34 | 3103 | * uuid: Change the uuid of the array to match watch is given |
f352c545 DW |
3104 | * |
3105 | * Following are not relevant for this imsm: | |
3106 | * sparc2.2 : update from old dodgey metadata | |
3107 | * super-minor: change the preferred_minor number | |
3108 | * summaries: update redundant counters. | |
f352c545 DW |
3109 | * homehost: update the recorded homehost |
3110 | * _reshape_progress: record new reshape_progress position. | |
3111 | */ | |
6e46bf34 DW |
3112 | int rv = 1; |
3113 | struct intel_super *super = st->sb; | |
3114 | struct imsm_super *mpb; | |
f352c545 | 3115 | |
6e46bf34 DW |
3116 | /* we can only update container info */ |
3117 | if (!super || super->current_vol >= 0 || !super->anchor) | |
3118 | return 1; | |
3119 | ||
3120 | mpb = super->anchor; | |
3121 | ||
81a5b4f5 N |
3122 | if (strcmp(update, "uuid") == 0) { |
3123 | /* We take this to mean that the family_num should be updated. | |
3124 | * However that is much smaller than the uuid so we cannot really | |
3125 | * allow an explicit uuid to be given. And it is hard to reliably | |
3126 | * know if one was. | |
3127 | * So if !uuid_set we know the current uuid is random and just used | |
3128 | * the first 'int' and copy it to the other 3 positions. | |
3129 | * Otherwise we require the 4 'int's to be the same as would be the | |
3130 | * case if we are using a random uuid. So an explicit uuid will be | |
3131 | * accepted as long as all for ints are the same... which shouldn't hurt | |
6e46bf34 | 3132 | */ |
81a5b4f5 N |
3133 | if (!uuid_set) { |
3134 | info->uuid[1] = info->uuid[2] = info->uuid[3] = info->uuid[0]; | |
6e46bf34 | 3135 | rv = 0; |
81a5b4f5 N |
3136 | } else { |
3137 | if (info->uuid[0] != info->uuid[1] || | |
3138 | info->uuid[1] != info->uuid[2] || | |
3139 | info->uuid[2] != info->uuid[3]) | |
3140 | rv = -1; | |
3141 | else | |
3142 | rv = 0; | |
6e46bf34 | 3143 | } |
81a5b4f5 N |
3144 | if (rv == 0) |
3145 | mpb->orig_family_num = info->uuid[0]; | |
6e46bf34 DW |
3146 | } else if (strcmp(update, "assemble") == 0) |
3147 | rv = 0; | |
3148 | else | |
1e2b2765 | 3149 | rv = -1; |
f352c545 | 3150 | |
6e46bf34 DW |
3151 | /* successful update? recompute checksum */ |
3152 | if (rv == 0) | |
3153 | mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb)); | |
f352c545 DW |
3154 | |
3155 | return rv; | |
cdddbdbc DW |
3156 | } |
3157 | ||
c2c087e6 | 3158 | static size_t disks_to_mpb_size(int disks) |
cdddbdbc | 3159 | { |
c2c087e6 | 3160 | size_t size; |
cdddbdbc | 3161 | |
c2c087e6 DW |
3162 | size = sizeof(struct imsm_super); |
3163 | size += (disks - 1) * sizeof(struct imsm_disk); | |
3164 | size += 2 * sizeof(struct imsm_dev); | |
3165 | /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */ | |
3166 | size += (4 - 2) * sizeof(struct imsm_map); | |
3167 | /* 4 possible disk_ord_tbl's */ | |
3168 | size += 4 * (disks - 1) * sizeof(__u32); | |
3169 | ||
3170 | return size; | |
3171 | } | |
3172 | ||
387fcd59 N |
3173 | static __u64 avail_size_imsm(struct supertype *st, __u64 devsize, |
3174 | unsigned long long data_offset) | |
c2c087e6 DW |
3175 | { |
3176 | if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS)) | |
3177 | return 0; | |
3178 | ||
3179 | return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); | |
cdddbdbc DW |
3180 | } |
3181 | ||
ba2de7ba DW |
3182 | static void free_devlist(struct intel_super *super) |
3183 | { | |
3184 | struct intel_dev *dv; | |
3185 | ||
3186 | while (super->devlist) { | |
3187 | dv = super->devlist->next; | |
3188 | free(super->devlist->dev); | |
3189 | free(super->devlist); | |
3190 | super->devlist = dv; | |
3191 | } | |
3192 | } | |
3193 | ||
3194 | static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src) | |
3195 | { | |
3196 | memcpy(dest, src, sizeof_imsm_dev(src, 0)); | |
3197 | } | |
3198 | ||
cdddbdbc DW |
3199 | static int compare_super_imsm(struct supertype *st, struct supertype *tst) |
3200 | { | |
3201 | /* | |
3202 | * return: | |
3203 | * 0 same, or first was empty, and second was copied | |
3204 | * 1 second had wrong number | |
3205 | * 2 wrong uuid | |
3206 | * 3 wrong other info | |
3207 | */ | |
3208 | struct intel_super *first = st->sb; | |
3209 | struct intel_super *sec = tst->sb; | |
3210 | ||
5d500228 N |
3211 | if (!first) { |
3212 | st->sb = tst->sb; | |
3213 | tst->sb = NULL; | |
3214 | return 0; | |
3215 | } | |
8603ea6f LM |
3216 | /* in platform dependent environment test if the disks |
3217 | * use the same Intel hba | |
cb8f6859 | 3218 | * If not on Intel hba at all, allow anything. |
8603ea6f | 3219 | */ |
6b781d33 AP |
3220 | if (!check_env("IMSM_NO_PLATFORM") && first->hba && sec->hba) { |
3221 | if (first->hba->type != sec->hba->type) { | |
8603ea6f | 3222 | fprintf(stderr, |
6b781d33 AP |
3223 | "HBAs of devices do not match %s != %s\n", |
3224 | get_sys_dev_type(first->hba->type), | |
3225 | get_sys_dev_type(sec->hba->type)); | |
3226 | return 3; | |
3227 | } | |
3228 | if (first->orom != sec->orom) { | |
3229 | fprintf(stderr, | |
3230 | "HBAs of devices do not match %s != %s\n", | |
3231 | first->hba->pci_id, sec->hba->pci_id); | |
8603ea6f LM |
3232 | return 3; |
3233 | } | |
3234 | } | |
cdddbdbc | 3235 | |
d23fe947 DW |
3236 | /* if an anchor does not have num_raid_devs set then it is a free |
3237 | * floating spare | |
3238 | */ | |
3239 | if (first->anchor->num_raid_devs > 0 && | |
3240 | sec->anchor->num_raid_devs > 0) { | |
a2b97981 DW |
3241 | /* Determine if these disks might ever have been |
3242 | * related. Further disambiguation can only take place | |
3243 | * in load_super_imsm_all | |
3244 | */ | |
3245 | __u32 first_family = first->anchor->orig_family_num; | |
3246 | __u32 sec_family = sec->anchor->orig_family_num; | |
3247 | ||
f796af5d DW |
3248 | if (memcmp(first->anchor->sig, sec->anchor->sig, |
3249 | MAX_SIGNATURE_LENGTH) != 0) | |
3250 | return 3; | |
3251 | ||
a2b97981 DW |
3252 | if (first_family == 0) |
3253 | first_family = first->anchor->family_num; | |
3254 | if (sec_family == 0) | |
3255 | sec_family = sec->anchor->family_num; | |
3256 | ||
3257 | if (first_family != sec_family) | |
d23fe947 | 3258 | return 3; |
f796af5d | 3259 | |
d23fe947 | 3260 | } |
cdddbdbc | 3261 | |
3e372e5a DW |
3262 | /* if 'first' is a spare promote it to a populated mpb with sec's |
3263 | * family number | |
3264 | */ | |
3265 | if (first->anchor->num_raid_devs == 0 && | |
3266 | sec->anchor->num_raid_devs > 0) { | |
78d30f94 | 3267 | int i; |
ba2de7ba DW |
3268 | struct intel_dev *dv; |
3269 | struct imsm_dev *dev; | |
78d30f94 DW |
3270 | |
3271 | /* we need to copy raid device info from sec if an allocation | |
3272 | * fails here we don't associate the spare | |
3273 | */ | |
3274 | for (i = 0; i < sec->anchor->num_raid_devs; i++) { | |
503975b9 N |
3275 | dv = xmalloc(sizeof(*dv)); |
3276 | dev = xmalloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1)); | |
ba2de7ba DW |
3277 | dv->dev = dev; |
3278 | dv->index = i; | |
3279 | dv->next = first->devlist; | |
3280 | first->devlist = dv; | |
78d30f94 | 3281 | } |
709743c5 | 3282 | if (i < sec->anchor->num_raid_devs) { |
ba2de7ba DW |
3283 | /* allocation failure */ |
3284 | free_devlist(first); | |
e12b3daa | 3285 | pr_err("imsm: failed to associate spare\n"); |
ba2de7ba | 3286 | return 3; |
78d30f94 | 3287 | } |
3e372e5a | 3288 | first->anchor->num_raid_devs = sec->anchor->num_raid_devs; |
148acb7b | 3289 | first->anchor->orig_family_num = sec->anchor->orig_family_num; |
3e372e5a | 3290 | first->anchor->family_num = sec->anchor->family_num; |
ac6449be | 3291 | memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH); |
709743c5 DW |
3292 | for (i = 0; i < sec->anchor->num_raid_devs; i++) |
3293 | imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i)); | |
3e372e5a DW |
3294 | } |
3295 | ||
cdddbdbc DW |
3296 | return 0; |
3297 | } | |
3298 | ||
0030e8d6 DW |
3299 | static void fd2devname(int fd, char *name) |
3300 | { | |
3301 | struct stat st; | |
3302 | char path[256]; | |
33a6535d | 3303 | char dname[PATH_MAX]; |
0030e8d6 DW |
3304 | char *nm; |
3305 | int rv; | |
3306 | ||
3307 | name[0] = '\0'; | |
3308 | if (fstat(fd, &st) != 0) | |
3309 | return; | |
3310 | sprintf(path, "/sys/dev/block/%d:%d", | |
3311 | major(st.st_rdev), minor(st.st_rdev)); | |
3312 | ||
9cf014ec | 3313 | rv = readlink(path, dname, sizeof(dname)-1); |
0030e8d6 DW |
3314 | if (rv <= 0) |
3315 | return; | |
9587c373 | 3316 | |
0030e8d6 DW |
3317 | dname[rv] = '\0'; |
3318 | nm = strrchr(dname, '/'); | |
7897de29 JS |
3319 | if (nm) { |
3320 | nm++; | |
3321 | snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm); | |
3322 | } | |
0030e8d6 DW |
3323 | } |
3324 | ||
cdddbdbc DW |
3325 | extern int scsi_get_serial(int fd, void *buf, size_t buf_len); |
3326 | ||
3327 | static int imsm_read_serial(int fd, char *devname, | |
3328 | __u8 serial[MAX_RAID_SERIAL_LEN]) | |
3329 | { | |
3330 | unsigned char scsi_serial[255]; | |
cdddbdbc DW |
3331 | int rv; |
3332 | int rsp_len; | |
1f24f035 | 3333 | int len; |
316e2bf4 DW |
3334 | char *dest; |
3335 | char *src; | |
3336 | char *rsp_buf; | |
3337 | int i; | |
cdddbdbc DW |
3338 | |
3339 | memset(scsi_serial, 0, sizeof(scsi_serial)); | |
cdddbdbc | 3340 | |
f9ba0ff1 DW |
3341 | rv = scsi_get_serial(fd, scsi_serial, sizeof(scsi_serial)); |
3342 | ||
40ebbb9c | 3343 | if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) { |
f9ba0ff1 DW |
3344 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
3345 | fd2devname(fd, (char *) serial); | |
0030e8d6 DW |
3346 | return 0; |
3347 | } | |
3348 | ||
cdddbdbc DW |
3349 | if (rv != 0) { |
3350 | if (devname) | |
e7b84f9d N |
3351 | pr_err("Failed to retrieve serial for %s\n", |
3352 | devname); | |
cdddbdbc DW |
3353 | return rv; |
3354 | } | |
3355 | ||
3356 | rsp_len = scsi_serial[3]; | |
03cd4cc8 DW |
3357 | if (!rsp_len) { |
3358 | if (devname) | |
e7b84f9d N |
3359 | pr_err("Failed to retrieve serial for %s\n", |
3360 | devname); | |
03cd4cc8 DW |
3361 | return 2; |
3362 | } | |
1f24f035 | 3363 | rsp_buf = (char *) &scsi_serial[4]; |
5c3db629 | 3364 | |
316e2bf4 DW |
3365 | /* trim all whitespace and non-printable characters and convert |
3366 | * ':' to ';' | |
3367 | */ | |
3368 | for (i = 0, dest = rsp_buf; i < rsp_len; i++) { | |
3369 | src = &rsp_buf[i]; | |
3370 | if (*src > 0x20) { | |
3371 | /* ':' is reserved for use in placeholder serial | |
3372 | * numbers for missing disks | |
3373 | */ | |
3374 | if (*src == ':') | |
3375 | *dest++ = ';'; | |
3376 | else | |
3377 | *dest++ = *src; | |
3378 | } | |
3379 | } | |
3380 | len = dest - rsp_buf; | |
3381 | dest = rsp_buf; | |
3382 | ||
3383 | /* truncate leading characters */ | |
3384 | if (len > MAX_RAID_SERIAL_LEN) { | |
3385 | dest += len - MAX_RAID_SERIAL_LEN; | |
1f24f035 | 3386 | len = MAX_RAID_SERIAL_LEN; |
316e2bf4 | 3387 | } |
5c3db629 | 3388 | |
5c3db629 | 3389 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
316e2bf4 | 3390 | memcpy(serial, dest, len); |
cdddbdbc DW |
3391 | |
3392 | return 0; | |
3393 | } | |
3394 | ||
1f24f035 DW |
3395 | static int serialcmp(__u8 *s1, __u8 *s2) |
3396 | { | |
3397 | return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN); | |
3398 | } | |
3399 | ||
3400 | static void serialcpy(__u8 *dest, __u8 *src) | |
3401 | { | |
3402 | strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN); | |
3403 | } | |
3404 | ||
54c2c1ea DW |
3405 | static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super) |
3406 | { | |
3407 | struct dl *dl; | |
3408 | ||
3409 | for (dl = super->disks; dl; dl = dl->next) | |
3410 | if (serialcmp(dl->serial, serial) == 0) | |
3411 | break; | |
3412 | ||
3413 | return dl; | |
3414 | } | |
3415 | ||
a2b97981 DW |
3416 | static struct imsm_disk * |
3417 | __serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx) | |
3418 | { | |
3419 | int i; | |
3420 | ||
3421 | for (i = 0; i < mpb->num_disks; i++) { | |
3422 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
3423 | ||
3424 | if (serialcmp(disk->serial, serial) == 0) { | |
3425 | if (idx) | |
3426 | *idx = i; | |
3427 | return disk; | |
3428 | } | |
3429 | } | |
3430 | ||
3431 | return NULL; | |
3432 | } | |
3433 | ||
cdddbdbc DW |
3434 | static int |
3435 | load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd) | |
3436 | { | |
a2b97981 | 3437 | struct imsm_disk *disk; |
cdddbdbc DW |
3438 | struct dl *dl; |
3439 | struct stat stb; | |
cdddbdbc | 3440 | int rv; |
a2b97981 | 3441 | char name[40]; |
d23fe947 DW |
3442 | __u8 serial[MAX_RAID_SERIAL_LEN]; |
3443 | ||
3444 | rv = imsm_read_serial(fd, devname, serial); | |
3445 | ||
3446 | if (rv != 0) | |
3447 | return 2; | |
3448 | ||
503975b9 | 3449 | dl = xcalloc(1, sizeof(*dl)); |
cdddbdbc | 3450 | |
a2b97981 DW |
3451 | fstat(fd, &stb); |
3452 | dl->major = major(stb.st_rdev); | |
3453 | dl->minor = minor(stb.st_rdev); | |
3454 | dl->next = super->disks; | |
3455 | dl->fd = keep_fd ? fd : -1; | |
3456 | assert(super->disks == NULL); | |
3457 | super->disks = dl; | |
3458 | serialcpy(dl->serial, serial); | |
3459 | dl->index = -2; | |
3460 | dl->e = NULL; | |
3461 | fd2devname(fd, name); | |
3462 | if (devname) | |
503975b9 | 3463 | dl->devname = xstrdup(devname); |
a2b97981 | 3464 | else |
503975b9 | 3465 | dl->devname = xstrdup(name); |
cdddbdbc | 3466 | |
d23fe947 | 3467 | /* look up this disk's index in the current anchor */ |
a2b97981 DW |
3468 | disk = __serial_to_disk(dl->serial, super->anchor, &dl->index); |
3469 | if (disk) { | |
3470 | dl->disk = *disk; | |
3471 | /* only set index on disks that are a member of a | |
3472 | * populated contianer, i.e. one with raid_devs | |
3473 | */ | |
3474 | if (is_failed(&dl->disk)) | |
3f6efecc | 3475 | dl->index = -2; |
a2b97981 DW |
3476 | else if (is_spare(&dl->disk)) |
3477 | dl->index = -1; | |
3f6efecc DW |
3478 | } |
3479 | ||
949c47a0 DW |
3480 | return 0; |
3481 | } | |
3482 | ||
0e600426 | 3483 | #ifndef MDASSEMBLE |
0c046afd DW |
3484 | /* When migrating map0 contains the 'destination' state while map1 |
3485 | * contains the current state. When not migrating map0 contains the | |
3486 | * current state. This routine assumes that map[0].map_state is set to | |
3487 | * the current array state before being called. | |
3488 | * | |
3489 | * Migration is indicated by one of the following states | |
3490 | * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed) | |
e3bba0e0 | 3491 | * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal |
0c046afd | 3492 | * map1state=unitialized) |
1484e727 | 3493 | * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal |
0c046afd | 3494 | * map1state=normal) |
e3bba0e0 | 3495 | * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal |
0c046afd | 3496 | * map1state=degraded) |
8e59f3d8 AK |
3497 | * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal |
3498 | * map1state=normal) | |
0c046afd | 3499 | */ |
8e59f3d8 AK |
3500 | static void migrate(struct imsm_dev *dev, struct intel_super *super, |
3501 | __u8 to_state, int migr_type) | |
3393c6af | 3502 | { |
0c046afd | 3503 | struct imsm_map *dest; |
238c0a71 | 3504 | struct imsm_map *src = get_imsm_map(dev, MAP_0); |
3393c6af | 3505 | |
0c046afd | 3506 | dev->vol.migr_state = 1; |
1484e727 | 3507 | set_migr_type(dev, migr_type); |
f8f603f1 | 3508 | dev->vol.curr_migr_unit = 0; |
238c0a71 | 3509 | dest = get_imsm_map(dev, MAP_1); |
0c046afd | 3510 | |
0556e1a2 | 3511 | /* duplicate and then set the target end state in map[0] */ |
3393c6af | 3512 | memcpy(dest, src, sizeof_imsm_map(src)); |
28bce06f AK |
3513 | if ((migr_type == MIGR_REBUILD) || |
3514 | (migr_type == MIGR_GEN_MIGR)) { | |
0556e1a2 DW |
3515 | __u32 ord; |
3516 | int i; | |
3517 | ||
3518 | for (i = 0; i < src->num_members; i++) { | |
3519 | ord = __le32_to_cpu(src->disk_ord_tbl[i]); | |
3520 | set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord)); | |
3521 | } | |
3522 | } | |
3523 | ||
8e59f3d8 AK |
3524 | if (migr_type == MIGR_GEN_MIGR) |
3525 | /* Clear migration record */ | |
3526 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
3527 | ||
0c046afd | 3528 | src->map_state = to_state; |
949c47a0 | 3529 | } |
f8f603f1 | 3530 | |
809da78e AK |
3531 | static void end_migration(struct imsm_dev *dev, struct intel_super *super, |
3532 | __u8 map_state) | |
f8f603f1 | 3533 | { |
238c0a71 AK |
3534 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
3535 | struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state == 0 ? | |
3536 | MAP_0 : MAP_1); | |
28bce06f | 3537 | int i, j; |
0556e1a2 DW |
3538 | |
3539 | /* merge any IMSM_ORD_REBUILD bits that were not successfully | |
3540 | * completed in the last migration. | |
3541 | * | |
28bce06f | 3542 | * FIXME add support for raid-level-migration |
0556e1a2 | 3543 | */ |
809da78e AK |
3544 | if ((map_state != map->map_state) && (is_gen_migration(dev) == 0) && |
3545 | (prev->map_state != IMSM_T_STATE_UNINITIALIZED)) { | |
3546 | /* when final map state is other than expected | |
3547 | * merge maps (not for migration) | |
3548 | */ | |
3549 | int failed; | |
3550 | ||
3551 | for (i = 0; i < prev->num_members; i++) | |
3552 | for (j = 0; j < map->num_members; j++) | |
3553 | /* during online capacity expansion | |
3554 | * disks position can be changed | |
3555 | * if takeover is used | |
3556 | */ | |
3557 | if (ord_to_idx(map->disk_ord_tbl[j]) == | |
3558 | ord_to_idx(prev->disk_ord_tbl[i])) { | |
3559 | map->disk_ord_tbl[j] |= | |
3560 | prev->disk_ord_tbl[i]; | |
3561 | break; | |
3562 | } | |
3563 | failed = imsm_count_failed(super, dev, MAP_0); | |
3564 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
3565 | } | |
f8f603f1 DW |
3566 | |
3567 | dev->vol.migr_state = 0; | |
ea672ee1 | 3568 | set_migr_type(dev, 0); |
f8f603f1 DW |
3569 | dev->vol.curr_migr_unit = 0; |
3570 | map->map_state = map_state; | |
3571 | } | |
0e600426 | 3572 | #endif |
949c47a0 DW |
3573 | |
3574 | static int parse_raid_devices(struct intel_super *super) | |
3575 | { | |
3576 | int i; | |
3577 | struct imsm_dev *dev_new; | |
4d7b1503 | 3578 | size_t len, len_migr; |
401d313b | 3579 | size_t max_len = 0; |
4d7b1503 DW |
3580 | size_t space_needed = 0; |
3581 | struct imsm_super *mpb = super->anchor; | |
949c47a0 DW |
3582 | |
3583 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
3584 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
ba2de7ba | 3585 | struct intel_dev *dv; |
949c47a0 | 3586 | |
4d7b1503 DW |
3587 | len = sizeof_imsm_dev(dev_iter, 0); |
3588 | len_migr = sizeof_imsm_dev(dev_iter, 1); | |
3589 | if (len_migr > len) | |
3590 | space_needed += len_migr - len; | |
ca9de185 | 3591 | |
503975b9 | 3592 | dv = xmalloc(sizeof(*dv)); |
401d313b AK |
3593 | if (max_len < len_migr) |
3594 | max_len = len_migr; | |
3595 | if (max_len > len_migr) | |
3596 | space_needed += max_len - len_migr; | |
503975b9 | 3597 | dev_new = xmalloc(max_len); |
949c47a0 | 3598 | imsm_copy_dev(dev_new, dev_iter); |
ba2de7ba DW |
3599 | dv->dev = dev_new; |
3600 | dv->index = i; | |
3601 | dv->next = super->devlist; | |
3602 | super->devlist = dv; | |
949c47a0 | 3603 | } |
cdddbdbc | 3604 | |
4d7b1503 DW |
3605 | /* ensure that super->buf is large enough when all raid devices |
3606 | * are migrating | |
3607 | */ | |
3608 | if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) { | |
3609 | void *buf; | |
3610 | ||
3611 | len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed, 512); | |
3612 | if (posix_memalign(&buf, 512, len) != 0) | |
3613 | return 1; | |
3614 | ||
1f45a8ad DW |
3615 | memcpy(buf, super->buf, super->len); |
3616 | memset(buf + super->len, 0, len - super->len); | |
4d7b1503 DW |
3617 | free(super->buf); |
3618 | super->buf = buf; | |
3619 | super->len = len; | |
3620 | } | |
ca9de185 | 3621 | |
cdddbdbc DW |
3622 | return 0; |
3623 | } | |
3624 | ||
604b746f JD |
3625 | /* retrieve a pointer to the bbm log which starts after all raid devices */ |
3626 | struct bbm_log *__get_imsm_bbm_log(struct imsm_super *mpb) | |
3627 | { | |
3628 | void *ptr = NULL; | |
3629 | ||
3630 | if (__le32_to_cpu(mpb->bbm_log_size)) { | |
3631 | ptr = mpb; | |
3632 | ptr += mpb->mpb_size - __le32_to_cpu(mpb->bbm_log_size); | |
1011e834 | 3633 | } |
604b746f JD |
3634 | |
3635 | return ptr; | |
3636 | } | |
3637 | ||
e2f41b2c AK |
3638 | /******************************************************************************* |
3639 | * Function: check_mpb_migr_compatibility | |
3640 | * Description: Function checks for unsupported migration features: | |
3641 | * - migration optimization area (pba_of_lba0) | |
3642 | * - descending reshape (ascending_migr) | |
3643 | * Parameters: | |
3644 | * super : imsm metadata information | |
3645 | * Returns: | |
3646 | * 0 : migration is compatible | |
3647 | * -1 : migration is not compatible | |
3648 | ******************************************************************************/ | |
3649 | int check_mpb_migr_compatibility(struct intel_super *super) | |
3650 | { | |
3651 | struct imsm_map *map0, *map1; | |
3652 | struct migr_record *migr_rec = super->migr_rec; | |
3653 | int i; | |
3654 | ||
3655 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
3656 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
3657 | ||
3658 | if (dev_iter && | |
3659 | dev_iter->vol.migr_state == 1 && | |
3660 | dev_iter->vol.migr_type == MIGR_GEN_MIGR) { | |
3661 | /* This device is migrating */ | |
238c0a71 AK |
3662 | map0 = get_imsm_map(dev_iter, MAP_0); |
3663 | map1 = get_imsm_map(dev_iter, MAP_1); | |
5551b113 | 3664 | if (pba_of_lba0(map0) != pba_of_lba0(map1)) |
e2f41b2c AK |
3665 | /* migration optimization area was used */ |
3666 | return -1; | |
3667 | if (migr_rec->ascending_migr == 0 | |
3668 | && migr_rec->dest_depth_per_unit > 0) | |
3669 | /* descending reshape not supported yet */ | |
3670 | return -1; | |
3671 | } | |
3672 | } | |
3673 | return 0; | |
3674 | } | |
3675 | ||
d23fe947 | 3676 | static void __free_imsm(struct intel_super *super, int free_disks); |
9ca2c81c | 3677 | |
cdddbdbc | 3678 | /* load_imsm_mpb - read matrix metadata |
f2f5c343 | 3679 | * allocates super->mpb to be freed by free_imsm |
cdddbdbc DW |
3680 | */ |
3681 | static int load_imsm_mpb(int fd, struct intel_super *super, char *devname) | |
3682 | { | |
3683 | unsigned long long dsize; | |
cdddbdbc DW |
3684 | unsigned long long sectors; |
3685 | struct stat; | |
6416d527 | 3686 | struct imsm_super *anchor; |
cdddbdbc DW |
3687 | __u32 check_sum; |
3688 | ||
cdddbdbc | 3689 | get_dev_size(fd, NULL, &dsize); |
64436f06 N |
3690 | if (dsize < 1024) { |
3691 | if (devname) | |
e7b84f9d N |
3692 | pr_err("%s: device to small for imsm\n", |
3693 | devname); | |
64436f06 N |
3694 | return 1; |
3695 | } | |
cdddbdbc DW |
3696 | |
3697 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) { | |
3698 | if (devname) | |
e7b84f9d N |
3699 | pr_err("Cannot seek to anchor block on %s: %s\n", |
3700 | devname, strerror(errno)); | |
cdddbdbc DW |
3701 | return 1; |
3702 | } | |
3703 | ||
949c47a0 | 3704 | if (posix_memalign((void**)&anchor, 512, 512) != 0) { |
ad97895e | 3705 | if (devname) |
7a862a02 | 3706 | pr_err("Failed to allocate imsm anchor buffer on %s\n", devname); |
ad97895e DW |
3707 | return 1; |
3708 | } | |
949c47a0 | 3709 | if (read(fd, anchor, 512) != 512) { |
cdddbdbc | 3710 | if (devname) |
e7b84f9d N |
3711 | pr_err("Cannot read anchor block on %s: %s\n", |
3712 | devname, strerror(errno)); | |
6416d527 | 3713 | free(anchor); |
cdddbdbc DW |
3714 | return 1; |
3715 | } | |
3716 | ||
6416d527 | 3717 | if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) { |
cdddbdbc | 3718 | if (devname) |
e7b84f9d | 3719 | pr_err("no IMSM anchor on %s\n", devname); |
6416d527 | 3720 | free(anchor); |
cdddbdbc DW |
3721 | return 2; |
3722 | } | |
3723 | ||
d23fe947 | 3724 | __free_imsm(super, 0); |
f2f5c343 LM |
3725 | /* reload capability and hba */ |
3726 | ||
3727 | /* capability and hba must be updated with new super allocation */ | |
d424212e | 3728 | find_intel_hba_capability(fd, super, devname); |
949c47a0 DW |
3729 | super->len = ROUND_UP(anchor->mpb_size, 512); |
3730 | if (posix_memalign(&super->buf, 512, super->len) != 0) { | |
cdddbdbc | 3731 | if (devname) |
e7b84f9d N |
3732 | pr_err("unable to allocate %zu byte mpb buffer\n", |
3733 | super->len); | |
6416d527 | 3734 | free(anchor); |
cdddbdbc DW |
3735 | return 2; |
3736 | } | |
949c47a0 | 3737 | memcpy(super->buf, anchor, 512); |
cdddbdbc | 3738 | |
6416d527 NB |
3739 | sectors = mpb_sectors(anchor) - 1; |
3740 | free(anchor); | |
8e59f3d8 | 3741 | |
17a4eaf9 | 3742 | if (posix_memalign(&super->migr_rec_buf, 512, MIGR_REC_BUF_SIZE) != 0) { |
1ade5cc1 | 3743 | pr_err("could not allocate migr_rec buffer\n"); |
8e59f3d8 AK |
3744 | free(super->buf); |
3745 | return 2; | |
3746 | } | |
51d83f5d | 3747 | super->clean_migration_record_by_mdmon = 0; |
8e59f3d8 | 3748 | |
949c47a0 | 3749 | if (!sectors) { |
ecf45690 DW |
3750 | check_sum = __gen_imsm_checksum(super->anchor); |
3751 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
3752 | if (devname) | |
e7b84f9d N |
3753 | pr_err("IMSM checksum %x != %x on %s\n", |
3754 | check_sum, | |
3755 | __le32_to_cpu(super->anchor->check_sum), | |
3756 | devname); | |
ecf45690 DW |
3757 | return 2; |
3758 | } | |
3759 | ||
a2b97981 | 3760 | return 0; |
949c47a0 | 3761 | } |
cdddbdbc DW |
3762 | |
3763 | /* read the extended mpb */ | |
3764 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) { | |
3765 | if (devname) | |
e7b84f9d N |
3766 | pr_err("Cannot seek to extended mpb on %s: %s\n", |
3767 | devname, strerror(errno)); | |
cdddbdbc DW |
3768 | return 1; |
3769 | } | |
3770 | ||
f21e18ca | 3771 | if ((unsigned)read(fd, super->buf + 512, super->len - 512) != super->len - 512) { |
cdddbdbc | 3772 | if (devname) |
e7b84f9d N |
3773 | pr_err("Cannot read extended mpb on %s: %s\n", |
3774 | devname, strerror(errno)); | |
cdddbdbc DW |
3775 | return 2; |
3776 | } | |
3777 | ||
949c47a0 DW |
3778 | check_sum = __gen_imsm_checksum(super->anchor); |
3779 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
cdddbdbc | 3780 | if (devname) |
e7b84f9d N |
3781 | pr_err("IMSM checksum %x != %x on %s\n", |
3782 | check_sum, __le32_to_cpu(super->anchor->check_sum), | |
3783 | devname); | |
db575f3b | 3784 | return 3; |
cdddbdbc DW |
3785 | } |
3786 | ||
604b746f JD |
3787 | /* FIXME the BBM log is disk specific so we cannot use this global |
3788 | * buffer for all disks. Ok for now since we only look at the global | |
3789 | * bbm_log_size parameter to gate assembly | |
3790 | */ | |
3791 | super->bbm_log = __get_imsm_bbm_log(super->anchor); | |
3792 | ||
a2b97981 DW |
3793 | return 0; |
3794 | } | |
3795 | ||
8e59f3d8 AK |
3796 | static int read_imsm_migr_rec(int fd, struct intel_super *super); |
3797 | ||
97f81ee2 CA |
3798 | /* clears hi bits in metadata if MPB_ATTRIB_2TB_DISK not set */ |
3799 | static void clear_hi(struct intel_super *super) | |
3800 | { | |
3801 | struct imsm_super *mpb = super->anchor; | |
3802 | int i, n; | |
3803 | if (mpb->attributes & MPB_ATTRIB_2TB_DISK) | |
3804 | return; | |
3805 | for (i = 0; i < mpb->num_disks; ++i) { | |
3806 | struct imsm_disk *disk = &mpb->disk[i]; | |
3807 | disk->total_blocks_hi = 0; | |
3808 | } | |
3809 | for (i = 0; i < mpb->num_raid_devs; ++i) { | |
3810 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
3811 | if (!dev) | |
3812 | return; | |
3813 | for (n = 0; n < 2; ++n) { | |
3814 | struct imsm_map *map = get_imsm_map(dev, n); | |
3815 | if (!map) | |
3816 | continue; | |
3817 | map->pba_of_lba0_hi = 0; | |
3818 | map->blocks_per_member_hi = 0; | |
3819 | map->num_data_stripes_hi = 0; | |
3820 | } | |
3821 | } | |
3822 | } | |
3823 | ||
a2b97981 DW |
3824 | static int |
3825 | load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd) | |
3826 | { | |
3827 | int err; | |
3828 | ||
3829 | err = load_imsm_mpb(fd, super, devname); | |
3830 | if (err) | |
3831 | return err; | |
3832 | err = load_imsm_disk(fd, super, devname, keep_fd); | |
3833 | if (err) | |
3834 | return err; | |
3835 | err = parse_raid_devices(super); | |
97f81ee2 | 3836 | clear_hi(super); |
a2b97981 | 3837 | return err; |
cdddbdbc DW |
3838 | } |
3839 | ||
ae6aad82 DW |
3840 | static void __free_imsm_disk(struct dl *d) |
3841 | { | |
3842 | if (d->fd >= 0) | |
3843 | close(d->fd); | |
3844 | if (d->devname) | |
3845 | free(d->devname); | |
0dcecb2e DW |
3846 | if (d->e) |
3847 | free(d->e); | |
ae6aad82 DW |
3848 | free(d); |
3849 | ||
3850 | } | |
1a64be56 | 3851 | |
cdddbdbc DW |
3852 | static void free_imsm_disks(struct intel_super *super) |
3853 | { | |
47ee5a45 | 3854 | struct dl *d; |
cdddbdbc | 3855 | |
47ee5a45 DW |
3856 | while (super->disks) { |
3857 | d = super->disks; | |
cdddbdbc | 3858 | super->disks = d->next; |
ae6aad82 | 3859 | __free_imsm_disk(d); |
cdddbdbc | 3860 | } |
cb82edca AK |
3861 | while (super->disk_mgmt_list) { |
3862 | d = super->disk_mgmt_list; | |
3863 | super->disk_mgmt_list = d->next; | |
3864 | __free_imsm_disk(d); | |
3865 | } | |
47ee5a45 DW |
3866 | while (super->missing) { |
3867 | d = super->missing; | |
3868 | super->missing = d->next; | |
3869 | __free_imsm_disk(d); | |
3870 | } | |
3871 | ||
cdddbdbc DW |
3872 | } |
3873 | ||
9ca2c81c | 3874 | /* free all the pieces hanging off of a super pointer */ |
d23fe947 | 3875 | static void __free_imsm(struct intel_super *super, int free_disks) |
cdddbdbc | 3876 | { |
88654014 LM |
3877 | struct intel_hba *elem, *next; |
3878 | ||
9ca2c81c | 3879 | if (super->buf) { |
949c47a0 | 3880 | free(super->buf); |
9ca2c81c DW |
3881 | super->buf = NULL; |
3882 | } | |
f2f5c343 LM |
3883 | /* unlink capability description */ |
3884 | super->orom = NULL; | |
8e59f3d8 AK |
3885 | if (super->migr_rec_buf) { |
3886 | free(super->migr_rec_buf); | |
3887 | super->migr_rec_buf = NULL; | |
3888 | } | |
d23fe947 DW |
3889 | if (free_disks) |
3890 | free_imsm_disks(super); | |
ba2de7ba | 3891 | free_devlist(super); |
88654014 LM |
3892 | elem = super->hba; |
3893 | while (elem) { | |
3894 | if (elem->path) | |
3895 | free((void *)elem->path); | |
3896 | next = elem->next; | |
3897 | free(elem); | |
3898 | elem = next; | |
88c32bb1 | 3899 | } |
88654014 | 3900 | super->hba = NULL; |
cdddbdbc DW |
3901 | } |
3902 | ||
9ca2c81c DW |
3903 | static void free_imsm(struct intel_super *super) |
3904 | { | |
d23fe947 | 3905 | __free_imsm(super, 1); |
9ca2c81c DW |
3906 | free(super); |
3907 | } | |
cdddbdbc DW |
3908 | |
3909 | static void free_super_imsm(struct supertype *st) | |
3910 | { | |
3911 | struct intel_super *super = st->sb; | |
3912 | ||
3913 | if (!super) | |
3914 | return; | |
3915 | ||
3916 | free_imsm(super); | |
3917 | st->sb = NULL; | |
3918 | } | |
3919 | ||
49133e57 | 3920 | static struct intel_super *alloc_super(void) |
c2c087e6 | 3921 | { |
503975b9 | 3922 | struct intel_super *super = xcalloc(1, sizeof(*super)); |
c2c087e6 | 3923 | |
503975b9 N |
3924 | super->current_vol = -1; |
3925 | super->create_offset = ~((unsigned long long) 0); | |
c2c087e6 DW |
3926 | return super; |
3927 | } | |
3928 | ||
f0f5a016 LM |
3929 | /* |
3930 | * find and allocate hba and OROM/EFI based on valid fd of RAID component device | |
3931 | */ | |
d424212e | 3932 | static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname) |
f0f5a016 LM |
3933 | { |
3934 | struct sys_dev *hba_name; | |
3935 | int rv = 0; | |
3936 | ||
3937 | if ((fd < 0) || check_env("IMSM_NO_PLATFORM")) { | |
f2f5c343 | 3938 | super->orom = NULL; |
f0f5a016 LM |
3939 | super->hba = NULL; |
3940 | return 0; | |
3941 | } | |
3942 | hba_name = find_disk_attached_hba(fd, NULL); | |
3943 | if (!hba_name) { | |
d424212e | 3944 | if (devname) |
e7b84f9d N |
3945 | pr_err("%s is not attached to Intel(R) RAID controller.\n", |
3946 | devname); | |
f0f5a016 LM |
3947 | return 1; |
3948 | } | |
3949 | rv = attach_hba_to_super(super, hba_name); | |
3950 | if (rv == 2) { | |
d424212e N |
3951 | if (devname) { |
3952 | struct intel_hba *hba = super->hba; | |
f0f5a016 | 3953 | |
60f0f54d PB |
3954 | pr_err("%s is attached to Intel(R) %s %s (%s),\n" |
3955 | " but the container is assigned to Intel(R) %s %s (", | |
d424212e | 3956 | devname, |
614902f6 | 3957 | get_sys_dev_type(hba_name->type), |
60f0f54d | 3958 | hba_name->type == SYS_DEV_VMD ? "domain" : "RAID controller", |
f0f5a016 | 3959 | hba_name->pci_id ? : "Err!", |
60f0f54d PB |
3960 | get_sys_dev_type(super->hba->type), |
3961 | hba->type == SYS_DEV_VMD ? "domain" : "RAID controller"); | |
f0f5a016 | 3962 | |
f0f5a016 LM |
3963 | while (hba) { |
3964 | fprintf(stderr, "%s", hba->pci_id ? : "Err!"); | |
3965 | if (hba->next) | |
3966 | fprintf(stderr, ", "); | |
3967 | hba = hba->next; | |
3968 | } | |
6b781d33 | 3969 | fprintf(stderr, ").\n" |
60f0f54d PB |
3970 | " Mixing devices attached to different %s is not allowed.\n", |
3971 | hba_name->type == SYS_DEV_VMD ? "VMD domains" : "controllers"); | |
f0f5a016 | 3972 | } |
f0f5a016 LM |
3973 | return 2; |
3974 | } | |
6b781d33 | 3975 | super->orom = find_imsm_capability(hba_name); |
f2f5c343 LM |
3976 | if (!super->orom) |
3977 | return 3; | |
614902f6 | 3978 | |
f0f5a016 LM |
3979 | return 0; |
3980 | } | |
3981 | ||
47ee5a45 DW |
3982 | /* find_missing - helper routine for load_super_imsm_all that identifies |
3983 | * disks that have disappeared from the system. This routine relies on | |
3984 | * the mpb being uptodate, which it is at load time. | |
3985 | */ | |
3986 | static int find_missing(struct intel_super *super) | |
3987 | { | |
3988 | int i; | |
3989 | struct imsm_super *mpb = super->anchor; | |
3990 | struct dl *dl; | |
3991 | struct imsm_disk *disk; | |
47ee5a45 DW |
3992 | |
3993 | for (i = 0; i < mpb->num_disks; i++) { | |
3994 | disk = __get_imsm_disk(mpb, i); | |
54c2c1ea | 3995 | dl = serial_to_dl(disk->serial, super); |
47ee5a45 DW |
3996 | if (dl) |
3997 | continue; | |
47ee5a45 | 3998 | |
503975b9 | 3999 | dl = xmalloc(sizeof(*dl)); |
47ee5a45 DW |
4000 | dl->major = 0; |
4001 | dl->minor = 0; | |
4002 | dl->fd = -1; | |
503975b9 | 4003 | dl->devname = xstrdup("missing"); |
47ee5a45 DW |
4004 | dl->index = i; |
4005 | serialcpy(dl->serial, disk->serial); | |
4006 | dl->disk = *disk; | |
689c9bf3 | 4007 | dl->e = NULL; |
47ee5a45 DW |
4008 | dl->next = super->missing; |
4009 | super->missing = dl; | |
4010 | } | |
4011 | ||
4012 | return 0; | |
4013 | } | |
4014 | ||
3960e579 | 4015 | #ifndef MDASSEMBLE |
a2b97981 DW |
4016 | static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list) |
4017 | { | |
4018 | struct intel_disk *idisk = disk_list; | |
4019 | ||
4020 | while (idisk) { | |
4021 | if (serialcmp(idisk->disk.serial, serial) == 0) | |
4022 | break; | |
4023 | idisk = idisk->next; | |
4024 | } | |
4025 | ||
4026 | return idisk; | |
4027 | } | |
4028 | ||
4029 | static int __prep_thunderdome(struct intel_super **table, int tbl_size, | |
4030 | struct intel_super *super, | |
4031 | struct intel_disk **disk_list) | |
4032 | { | |
4033 | struct imsm_disk *d = &super->disks->disk; | |
4034 | struct imsm_super *mpb = super->anchor; | |
4035 | int i, j; | |
4036 | ||
4037 | for (i = 0; i < tbl_size; i++) { | |
4038 | struct imsm_super *tbl_mpb = table[i]->anchor; | |
4039 | struct imsm_disk *tbl_d = &table[i]->disks->disk; | |
4040 | ||
4041 | if (tbl_mpb->family_num == mpb->family_num) { | |
4042 | if (tbl_mpb->check_sum == mpb->check_sum) { | |
1ade5cc1 N |
4043 | dprintf("mpb from %d:%d matches %d:%d\n", |
4044 | super->disks->major, | |
a2b97981 DW |
4045 | super->disks->minor, |
4046 | table[i]->disks->major, | |
4047 | table[i]->disks->minor); | |
4048 | break; | |
4049 | } | |
4050 | ||
4051 | if (((is_configured(d) && !is_configured(tbl_d)) || | |
4052 | is_configured(d) == is_configured(tbl_d)) && | |
4053 | tbl_mpb->generation_num < mpb->generation_num) { | |
4054 | /* current version of the mpb is a | |
4055 | * better candidate than the one in | |
4056 | * super_table, but copy over "cross | |
4057 | * generational" status | |
4058 | */ | |
4059 | struct intel_disk *idisk; | |
4060 | ||
1ade5cc1 N |
4061 | dprintf("mpb from %d:%d replaces %d:%d\n", |
4062 | super->disks->major, | |
a2b97981 DW |
4063 | super->disks->minor, |
4064 | table[i]->disks->major, | |
4065 | table[i]->disks->minor); | |
4066 | ||
4067 | idisk = disk_list_get(tbl_d->serial, *disk_list); | |
4068 | if (idisk && is_failed(&idisk->disk)) | |
4069 | tbl_d->status |= FAILED_DISK; | |
4070 | break; | |
4071 | } else { | |
4072 | struct intel_disk *idisk; | |
4073 | struct imsm_disk *disk; | |
4074 | ||
4075 | /* tbl_mpb is more up to date, but copy | |
4076 | * over cross generational status before | |
4077 | * returning | |
4078 | */ | |
4079 | disk = __serial_to_disk(d->serial, mpb, NULL); | |
4080 | if (disk && is_failed(disk)) | |
4081 | d->status |= FAILED_DISK; | |
4082 | ||
4083 | idisk = disk_list_get(d->serial, *disk_list); | |
4084 | if (idisk) { | |
4085 | idisk->owner = i; | |
4086 | if (disk && is_configured(disk)) | |
4087 | idisk->disk.status |= CONFIGURED_DISK; | |
4088 | } | |
4089 | ||
1ade5cc1 N |
4090 | dprintf("mpb from %d:%d prefer %d:%d\n", |
4091 | super->disks->major, | |
a2b97981 DW |
4092 | super->disks->minor, |
4093 | table[i]->disks->major, | |
4094 | table[i]->disks->minor); | |
4095 | ||
4096 | return tbl_size; | |
4097 | } | |
4098 | } | |
4099 | } | |
4100 | ||
4101 | if (i >= tbl_size) | |
4102 | table[tbl_size++] = super; | |
4103 | else | |
4104 | table[i] = super; | |
4105 | ||
4106 | /* update/extend the merged list of imsm_disk records */ | |
4107 | for (j = 0; j < mpb->num_disks; j++) { | |
4108 | struct imsm_disk *disk = __get_imsm_disk(mpb, j); | |
4109 | struct intel_disk *idisk; | |
4110 | ||
4111 | idisk = disk_list_get(disk->serial, *disk_list); | |
4112 | if (idisk) { | |
4113 | idisk->disk.status |= disk->status; | |
4114 | if (is_configured(&idisk->disk) || | |
4115 | is_failed(&idisk->disk)) | |
4116 | idisk->disk.status &= ~(SPARE_DISK); | |
4117 | } else { | |
503975b9 | 4118 | idisk = xcalloc(1, sizeof(*idisk)); |
a2b97981 DW |
4119 | idisk->owner = IMSM_UNKNOWN_OWNER; |
4120 | idisk->disk = *disk; | |
4121 | idisk->next = *disk_list; | |
4122 | *disk_list = idisk; | |
4123 | } | |
4124 | ||
4125 | if (serialcmp(idisk->disk.serial, d->serial) == 0) | |
4126 | idisk->owner = i; | |
4127 | } | |
4128 | ||
4129 | return tbl_size; | |
4130 | } | |
4131 | ||
4132 | static struct intel_super * | |
4133 | validate_members(struct intel_super *super, struct intel_disk *disk_list, | |
4134 | const int owner) | |
4135 | { | |
4136 | struct imsm_super *mpb = super->anchor; | |
4137 | int ok_count = 0; | |
4138 | int i; | |
4139 | ||
4140 | for (i = 0; i < mpb->num_disks; i++) { | |
4141 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
4142 | struct intel_disk *idisk; | |
4143 | ||
4144 | idisk = disk_list_get(disk->serial, disk_list); | |
4145 | if (idisk) { | |
4146 | if (idisk->owner == owner || | |
4147 | idisk->owner == IMSM_UNKNOWN_OWNER) | |
4148 | ok_count++; | |
4149 | else | |
1ade5cc1 N |
4150 | dprintf("'%.16s' owner %d != %d\n", |
4151 | disk->serial, idisk->owner, | |
a2b97981 DW |
4152 | owner); |
4153 | } else { | |
1ade5cc1 N |
4154 | dprintf("unknown disk %x [%d]: %.16s\n", |
4155 | __le32_to_cpu(mpb->family_num), i, | |
a2b97981 DW |
4156 | disk->serial); |
4157 | break; | |
4158 | } | |
4159 | } | |
4160 | ||
4161 | if (ok_count == mpb->num_disks) | |
4162 | return super; | |
4163 | return NULL; | |
4164 | } | |
4165 | ||
4166 | static void show_conflicts(__u32 family_num, struct intel_super *super_list) | |
4167 | { | |
4168 | struct intel_super *s; | |
4169 | ||
4170 | for (s = super_list; s; s = s->next) { | |
4171 | if (family_num != s->anchor->family_num) | |
4172 | continue; | |
e12b3daa | 4173 | pr_err("Conflict, offlining family %#x on '%s'\n", |
a2b97981 DW |
4174 | __le32_to_cpu(family_num), s->disks->devname); |
4175 | } | |
4176 | } | |
4177 | ||
4178 | static struct intel_super * | |
4179 | imsm_thunderdome(struct intel_super **super_list, int len) | |
4180 | { | |
4181 | struct intel_super *super_table[len]; | |
4182 | struct intel_disk *disk_list = NULL; | |
4183 | struct intel_super *champion, *spare; | |
4184 | struct intel_super *s, **del; | |
4185 | int tbl_size = 0; | |
4186 | int conflict; | |
4187 | int i; | |
4188 | ||
4189 | memset(super_table, 0, sizeof(super_table)); | |
4190 | for (s = *super_list; s; s = s->next) | |
4191 | tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list); | |
4192 | ||
4193 | for (i = 0; i < tbl_size; i++) { | |
4194 | struct imsm_disk *d; | |
4195 | struct intel_disk *idisk; | |
4196 | struct imsm_super *mpb = super_table[i]->anchor; | |
4197 | ||
4198 | s = super_table[i]; | |
4199 | d = &s->disks->disk; | |
4200 | ||
4201 | /* 'd' must appear in merged disk list for its | |
4202 | * configuration to be valid | |
4203 | */ | |
4204 | idisk = disk_list_get(d->serial, disk_list); | |
4205 | if (idisk && idisk->owner == i) | |
4206 | s = validate_members(s, disk_list, i); | |
4207 | else | |
4208 | s = NULL; | |
4209 | ||
4210 | if (!s) | |
1ade5cc1 N |
4211 | dprintf("marking family: %#x from %d:%d offline\n", |
4212 | mpb->family_num, | |
a2b97981 DW |
4213 | super_table[i]->disks->major, |
4214 | super_table[i]->disks->minor); | |
4215 | super_table[i] = s; | |
4216 | } | |
4217 | ||
4218 | /* This is where the mdadm implementation differs from the Windows | |
4219 | * driver which has no strict concept of a container. We can only | |
4220 | * assemble one family from a container, so when returning a prodigal | |
4221 | * array member to this system the code will not be able to disambiguate | |
4222 | * the container contents that should be assembled ("foreign" versus | |
4223 | * "local"). It requires user intervention to set the orig_family_num | |
4224 | * to a new value to establish a new container. The Windows driver in | |
4225 | * this situation fixes up the volume name in place and manages the | |
4226 | * foreign array as an independent entity. | |
4227 | */ | |
4228 | s = NULL; | |
4229 | spare = NULL; | |
4230 | conflict = 0; | |
4231 | for (i = 0; i < tbl_size; i++) { | |
4232 | struct intel_super *tbl_ent = super_table[i]; | |
4233 | int is_spare = 0; | |
4234 | ||
4235 | if (!tbl_ent) | |
4236 | continue; | |
4237 | ||
4238 | if (tbl_ent->anchor->num_raid_devs == 0) { | |
4239 | spare = tbl_ent; | |
4240 | is_spare = 1; | |
4241 | } | |
4242 | ||
4243 | if (s && !is_spare) { | |
4244 | show_conflicts(tbl_ent->anchor->family_num, *super_list); | |
4245 | conflict++; | |
4246 | } else if (!s && !is_spare) | |
4247 | s = tbl_ent; | |
4248 | } | |
4249 | ||
4250 | if (!s) | |
4251 | s = spare; | |
4252 | if (!s) { | |
4253 | champion = NULL; | |
4254 | goto out; | |
4255 | } | |
4256 | champion = s; | |
4257 | ||
4258 | if (conflict) | |
7a862a02 | 4259 | pr_err("Chose family %#x on '%s', assemble conflicts to new container with '--update=uuid'\n", |
a2b97981 DW |
4260 | __le32_to_cpu(s->anchor->family_num), s->disks->devname); |
4261 | ||
4262 | /* collect all dl's onto 'champion', and update them to | |
4263 | * champion's version of the status | |
4264 | */ | |
4265 | for (s = *super_list; s; s = s->next) { | |
4266 | struct imsm_super *mpb = champion->anchor; | |
4267 | struct dl *dl = s->disks; | |
4268 | ||
4269 | if (s == champion) | |
4270 | continue; | |
4271 | ||
5d7b407a CA |
4272 | mpb->attributes |= s->anchor->attributes & MPB_ATTRIB_2TB_DISK; |
4273 | ||
a2b97981 DW |
4274 | for (i = 0; i < mpb->num_disks; i++) { |
4275 | struct imsm_disk *disk; | |
4276 | ||
4277 | disk = __serial_to_disk(dl->serial, mpb, &dl->index); | |
4278 | if (disk) { | |
4279 | dl->disk = *disk; | |
4280 | /* only set index on disks that are a member of | |
4281 | * a populated contianer, i.e. one with | |
4282 | * raid_devs | |
4283 | */ | |
4284 | if (is_failed(&dl->disk)) | |
4285 | dl->index = -2; | |
4286 | else if (is_spare(&dl->disk)) | |
4287 | dl->index = -1; | |
4288 | break; | |
4289 | } | |
4290 | } | |
4291 | ||
4292 | if (i >= mpb->num_disks) { | |
4293 | struct intel_disk *idisk; | |
4294 | ||
4295 | idisk = disk_list_get(dl->serial, disk_list); | |
ecf408e9 | 4296 | if (idisk && is_spare(&idisk->disk) && |
a2b97981 DW |
4297 | !is_failed(&idisk->disk) && !is_configured(&idisk->disk)) |
4298 | dl->index = -1; | |
4299 | else { | |
4300 | dl->index = -2; | |
4301 | continue; | |
4302 | } | |
4303 | } | |
4304 | ||
4305 | dl->next = champion->disks; | |
4306 | champion->disks = dl; | |
4307 | s->disks = NULL; | |
4308 | } | |
4309 | ||
4310 | /* delete 'champion' from super_list */ | |
4311 | for (del = super_list; *del; ) { | |
4312 | if (*del == champion) { | |
4313 | *del = (*del)->next; | |
4314 | break; | |
4315 | } else | |
4316 | del = &(*del)->next; | |
4317 | } | |
4318 | champion->next = NULL; | |
4319 | ||
4320 | out: | |
4321 | while (disk_list) { | |
4322 | struct intel_disk *idisk = disk_list; | |
4323 | ||
4324 | disk_list = disk_list->next; | |
4325 | free(idisk); | |
4326 | } | |
4327 | ||
4328 | return champion; | |
4329 | } | |
4330 | ||
9587c373 LM |
4331 | static int |
4332 | get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd); | |
4dd2df09 | 4333 | static int get_super_block(struct intel_super **super_list, char *devnm, char *devname, |
9587c373 | 4334 | int major, int minor, int keep_fd); |
ec50f7b6 LM |
4335 | static int |
4336 | get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list, | |
4337 | int *max, int keep_fd); | |
4338 | ||
cdddbdbc | 4339 | static int load_super_imsm_all(struct supertype *st, int fd, void **sbp, |
ec50f7b6 LM |
4340 | char *devname, struct md_list *devlist, |
4341 | int keep_fd) | |
cdddbdbc | 4342 | { |
a2b97981 DW |
4343 | struct intel_super *super_list = NULL; |
4344 | struct intel_super *super = NULL; | |
a2b97981 | 4345 | int err = 0; |
9587c373 | 4346 | int i = 0; |
dab4a513 | 4347 | |
9587c373 LM |
4348 | if (fd >= 0) |
4349 | /* 'fd' is an opened container */ | |
4350 | err = get_sra_super_block(fd, &super_list, devname, &i, keep_fd); | |
4351 | else | |
ec50f7b6 LM |
4352 | /* get super block from devlist devices */ |
4353 | err = get_devlist_super_block(devlist, &super_list, &i, keep_fd); | |
9587c373 | 4354 | if (err) |
1602d52c | 4355 | goto error; |
a2b97981 DW |
4356 | /* all mpbs enter, maybe one leaves */ |
4357 | super = imsm_thunderdome(&super_list, i); | |
4358 | if (!super) { | |
4359 | err = 1; | |
4360 | goto error; | |
cdddbdbc DW |
4361 | } |
4362 | ||
47ee5a45 DW |
4363 | if (find_missing(super) != 0) { |
4364 | free_imsm(super); | |
a2b97981 DW |
4365 | err = 2; |
4366 | goto error; | |
47ee5a45 | 4367 | } |
8e59f3d8 AK |
4368 | |
4369 | /* load migration record */ | |
4370 | err = load_imsm_migr_rec(super, NULL); | |
4c965cc9 AK |
4371 | if (err == -1) { |
4372 | /* migration is in progress, | |
4373 | * but migr_rec cannot be loaded, | |
4374 | */ | |
8e59f3d8 AK |
4375 | err = 4; |
4376 | goto error; | |
4377 | } | |
e2f41b2c AK |
4378 | |
4379 | /* Check migration compatibility */ | |
4c965cc9 | 4380 | if ((err == 0) && (check_mpb_migr_compatibility(super) != 0)) { |
e7b84f9d | 4381 | pr_err("Unsupported migration detected"); |
e2f41b2c AK |
4382 | if (devname) |
4383 | fprintf(stderr, " on %s\n", devname); | |
4384 | else | |
4385 | fprintf(stderr, " (IMSM).\n"); | |
4386 | ||
4387 | err = 5; | |
4388 | goto error; | |
4389 | } | |
4390 | ||
a2b97981 DW |
4391 | err = 0; |
4392 | ||
4393 | error: | |
4394 | while (super_list) { | |
4395 | struct intel_super *s = super_list; | |
4396 | ||
4397 | super_list = super_list->next; | |
4398 | free_imsm(s); | |
4399 | } | |
9587c373 | 4400 | |
a2b97981 DW |
4401 | if (err) |
4402 | return err; | |
f7e7067b | 4403 | |
cdddbdbc | 4404 | *sbp = super; |
9587c373 | 4405 | if (fd >= 0) |
4dd2df09 | 4406 | strcpy(st->container_devnm, fd2devnm(fd)); |
9587c373 | 4407 | else |
4dd2df09 | 4408 | st->container_devnm[0] = 0; |
a2b97981 | 4409 | if (err == 0 && st->ss == NULL) { |
bf5a934a | 4410 | st->ss = &super_imsm; |
cdddbdbc DW |
4411 | st->minor_version = 0; |
4412 | st->max_devs = IMSM_MAX_DEVICES; | |
4413 | } | |
cdddbdbc DW |
4414 | return 0; |
4415 | } | |
2b959fbf | 4416 | |
ec50f7b6 LM |
4417 | static int |
4418 | get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list, | |
4419 | int *max, int keep_fd) | |
4420 | { | |
4421 | struct md_list *tmpdev; | |
4422 | int err = 0; | |
4423 | int i = 0; | |
9587c373 | 4424 | |
ec50f7b6 LM |
4425 | for (i = 0, tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { |
4426 | if (tmpdev->used != 1) | |
4427 | continue; | |
4428 | if (tmpdev->container == 1) { | |
ca9de185 | 4429 | int lmax = 0; |
ec50f7b6 LM |
4430 | int fd = dev_open(tmpdev->devname, O_RDONLY|O_EXCL); |
4431 | if (fd < 0) { | |
e7b84f9d | 4432 | pr_err("cannot open device %s: %s\n", |
ec50f7b6 LM |
4433 | tmpdev->devname, strerror(errno)); |
4434 | err = 8; | |
4435 | goto error; | |
4436 | } | |
4437 | err = get_sra_super_block(fd, super_list, | |
4438 | tmpdev->devname, &lmax, | |
4439 | keep_fd); | |
4440 | i += lmax; | |
4441 | close(fd); | |
4442 | if (err) { | |
4443 | err = 7; | |
4444 | goto error; | |
4445 | } | |
4446 | } else { | |
4447 | int major = major(tmpdev->st_rdev); | |
4448 | int minor = minor(tmpdev->st_rdev); | |
4449 | err = get_super_block(super_list, | |
4dd2df09 | 4450 | NULL, |
ec50f7b6 LM |
4451 | tmpdev->devname, |
4452 | major, minor, | |
4453 | keep_fd); | |
4454 | i++; | |
4455 | if (err) { | |
4456 | err = 6; | |
4457 | goto error; | |
4458 | } | |
4459 | } | |
4460 | } | |
4461 | error: | |
4462 | *max = i; | |
4463 | return err; | |
4464 | } | |
9587c373 | 4465 | |
4dd2df09 | 4466 | static int get_super_block(struct intel_super **super_list, char *devnm, char *devname, |
9587c373 LM |
4467 | int major, int minor, int keep_fd) |
4468 | { | |
4469 | struct intel_super*s = NULL; | |
4470 | char nm[32]; | |
4471 | int dfd = -1; | |
9587c373 LM |
4472 | int err = 0; |
4473 | int retry; | |
4474 | ||
4475 | s = alloc_super(); | |
4476 | if (!s) { | |
4477 | err = 1; | |
4478 | goto error; | |
4479 | } | |
4480 | ||
4481 | sprintf(nm, "%d:%d", major, minor); | |
4482 | dfd = dev_open(nm, O_RDWR); | |
4483 | if (dfd < 0) { | |
4484 | err = 2; | |
4485 | goto error; | |
4486 | } | |
4487 | ||
cb8f6859 | 4488 | find_intel_hba_capability(dfd, s, devname); |
9587c373 LM |
4489 | err = load_and_parse_mpb(dfd, s, NULL, keep_fd); |
4490 | ||
4491 | /* retry the load if we might have raced against mdmon */ | |
4dd2df09 | 4492 | if (err == 3 && devnm && mdmon_running(devnm)) |
9587c373 LM |
4493 | for (retry = 0; retry < 3; retry++) { |
4494 | usleep(3000); | |
4495 | err = load_and_parse_mpb(dfd, s, NULL, keep_fd); | |
4496 | if (err != 3) | |
4497 | break; | |
4498 | } | |
4499 | error: | |
4500 | if (!err) { | |
4501 | s->next = *super_list; | |
4502 | *super_list = s; | |
4503 | } else { | |
4504 | if (s) | |
4505 | free(s); | |
36614e95 | 4506 | if (dfd >= 0) |
9587c373 LM |
4507 | close(dfd); |
4508 | } | |
4509 | if ((dfd >= 0) && (!keep_fd)) | |
4510 | close(dfd); | |
4511 | return err; | |
4512 | ||
4513 | } | |
4514 | ||
4515 | static int | |
4516 | get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd) | |
4517 | { | |
4518 | struct mdinfo *sra; | |
4dd2df09 | 4519 | char *devnm; |
9587c373 LM |
4520 | struct mdinfo *sd; |
4521 | int err = 0; | |
4522 | int i = 0; | |
4dd2df09 | 4523 | sra = sysfs_read(fd, NULL, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE); |
9587c373 LM |
4524 | if (!sra) |
4525 | return 1; | |
4526 | ||
4527 | if (sra->array.major_version != -1 || | |
4528 | sra->array.minor_version != -2 || | |
4529 | strcmp(sra->text_version, "imsm") != 0) { | |
4530 | err = 1; | |
4531 | goto error; | |
4532 | } | |
4533 | /* load all mpbs */ | |
4dd2df09 | 4534 | devnm = fd2devnm(fd); |
9587c373 | 4535 | for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) { |
4dd2df09 | 4536 | if (get_super_block(super_list, devnm, devname, |
9587c373 LM |
4537 | sd->disk.major, sd->disk.minor, keep_fd) != 0) { |
4538 | err = 7; | |
4539 | goto error; | |
4540 | } | |
4541 | } | |
4542 | error: | |
4543 | sysfs_free(sra); | |
4544 | *max = i; | |
4545 | return err; | |
4546 | } | |
4547 | ||
2b959fbf N |
4548 | static int load_container_imsm(struct supertype *st, int fd, char *devname) |
4549 | { | |
ec50f7b6 | 4550 | return load_super_imsm_all(st, fd, &st->sb, devname, NULL, 1); |
2b959fbf | 4551 | } |
cdddbdbc DW |
4552 | #endif |
4553 | ||
4554 | static int load_super_imsm(struct supertype *st, int fd, char *devname) | |
4555 | { | |
4556 | struct intel_super *super; | |
4557 | int rv; | |
8a3544f8 | 4558 | int retry; |
cdddbdbc | 4559 | |
357ac106 | 4560 | if (test_partition(fd)) |
691c6ee1 N |
4561 | /* IMSM not allowed on partitions */ |
4562 | return 1; | |
4563 | ||
37424f13 DW |
4564 | free_super_imsm(st); |
4565 | ||
49133e57 | 4566 | super = alloc_super(); |
ea2bc72b LM |
4567 | /* Load hba and capabilities if they exist. |
4568 | * But do not preclude loading metadata in case capabilities or hba are | |
4569 | * non-compliant and ignore_hw_compat is set. | |
4570 | */ | |
d424212e | 4571 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 4572 | /* no orom/efi or non-intel hba of the disk */ |
ea2bc72b | 4573 | if ((rv != 0) && (st->ignore_hw_compat == 0)) { |
f2f5c343 | 4574 | if (devname) |
e7b84f9d | 4575 | pr_err("No OROM/EFI properties for %s\n", devname); |
f2f5c343 LM |
4576 | free_imsm(super); |
4577 | return 2; | |
4578 | } | |
a2b97981 | 4579 | rv = load_and_parse_mpb(fd, super, devname, 0); |
cdddbdbc | 4580 | |
8a3544f8 AP |
4581 | /* retry the load if we might have raced against mdmon */ |
4582 | if (rv == 3) { | |
4583 | struct mdstat_ent *mdstat = mdstat_by_component(fd2devnm(fd)); | |
4584 | ||
4585 | if (mdstat && mdmon_running(mdstat->devnm) && getpid() != mdmon_pid(mdstat->devnm)) { | |
4586 | for (retry = 0; retry < 3; retry++) { | |
4587 | usleep(3000); | |
4588 | rv = load_and_parse_mpb(fd, super, devname, 0); | |
4589 | if (rv != 3) | |
4590 | break; | |
4591 | } | |
4592 | } | |
4593 | ||
4594 | free_mdstat(mdstat); | |
4595 | } | |
4596 | ||
cdddbdbc DW |
4597 | if (rv) { |
4598 | if (devname) | |
7a862a02 | 4599 | pr_err("Failed to load all information sections on %s\n", devname); |
cdddbdbc DW |
4600 | free_imsm(super); |
4601 | return rv; | |
4602 | } | |
4603 | ||
4604 | st->sb = super; | |
4605 | if (st->ss == NULL) { | |
4606 | st->ss = &super_imsm; | |
4607 | st->minor_version = 0; | |
4608 | st->max_devs = IMSM_MAX_DEVICES; | |
4609 | } | |
8e59f3d8 AK |
4610 | |
4611 | /* load migration record */ | |
2e062e82 AK |
4612 | if (load_imsm_migr_rec(super, NULL) == 0) { |
4613 | /* Check for unsupported migration features */ | |
4614 | if (check_mpb_migr_compatibility(super) != 0) { | |
e7b84f9d | 4615 | pr_err("Unsupported migration detected"); |
2e062e82 AK |
4616 | if (devname) |
4617 | fprintf(stderr, " on %s\n", devname); | |
4618 | else | |
4619 | fprintf(stderr, " (IMSM).\n"); | |
4620 | return 3; | |
4621 | } | |
e2f41b2c AK |
4622 | } |
4623 | ||
cdddbdbc DW |
4624 | return 0; |
4625 | } | |
4626 | ||
ef6ffade DW |
4627 | static __u16 info_to_blocks_per_strip(mdu_array_info_t *info) |
4628 | { | |
4629 | if (info->level == 1) | |
4630 | return 128; | |
4631 | return info->chunk_size >> 9; | |
4632 | } | |
4633 | ||
5551b113 CA |
4634 | static unsigned long long info_to_blocks_per_member(mdu_array_info_t *info, |
4635 | unsigned long long size) | |
fcfd9599 | 4636 | { |
4025c288 | 4637 | if (info->level == 1) |
5551b113 | 4638 | return size * 2; |
4025c288 | 4639 | else |
5551b113 | 4640 | return (size * 2) & ~(info_to_blocks_per_strip(info) - 1); |
fcfd9599 DW |
4641 | } |
4642 | ||
4d1313e9 DW |
4643 | static void imsm_update_version_info(struct intel_super *super) |
4644 | { | |
4645 | /* update the version and attributes */ | |
4646 | struct imsm_super *mpb = super->anchor; | |
4647 | char *version; | |
4648 | struct imsm_dev *dev; | |
4649 | struct imsm_map *map; | |
4650 | int i; | |
4651 | ||
4652 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
4653 | dev = get_imsm_dev(super, i); | |
238c0a71 | 4654 | map = get_imsm_map(dev, MAP_0); |
4d1313e9 DW |
4655 | if (__le32_to_cpu(dev->size_high) > 0) |
4656 | mpb->attributes |= MPB_ATTRIB_2TB; | |
4657 | ||
4658 | /* FIXME detect when an array spans a port multiplier */ | |
4659 | #if 0 | |
4660 | mpb->attributes |= MPB_ATTRIB_PM; | |
4661 | #endif | |
4662 | ||
4663 | if (mpb->num_raid_devs > 1 || | |
4664 | mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) { | |
4665 | version = MPB_VERSION_ATTRIBS; | |
4666 | switch (get_imsm_raid_level(map)) { | |
4667 | case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break; | |
4668 | case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break; | |
4669 | case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break; | |
4670 | case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break; | |
4671 | } | |
4672 | } else { | |
4673 | if (map->num_members >= 5) | |
4674 | version = MPB_VERSION_5OR6_DISK_ARRAY; | |
4675 | else if (dev->status == DEV_CLONE_N_GO) | |
4676 | version = MPB_VERSION_CNG; | |
4677 | else if (get_imsm_raid_level(map) == 5) | |
4678 | version = MPB_VERSION_RAID5; | |
4679 | else if (map->num_members >= 3) | |
4680 | version = MPB_VERSION_3OR4_DISK_ARRAY; | |
4681 | else if (get_imsm_raid_level(map) == 1) | |
4682 | version = MPB_VERSION_RAID1; | |
4683 | else | |
4684 | version = MPB_VERSION_RAID0; | |
4685 | } | |
4686 | strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version); | |
4687 | } | |
4688 | } | |
4689 | ||
aa534678 DW |
4690 | static int check_name(struct intel_super *super, char *name, int quiet) |
4691 | { | |
4692 | struct imsm_super *mpb = super->anchor; | |
4693 | char *reason = NULL; | |
4694 | int i; | |
4695 | ||
4696 | if (strlen(name) > MAX_RAID_SERIAL_LEN) | |
4697 | reason = "must be 16 characters or less"; | |
4698 | ||
4699 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
4700 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
4701 | ||
4702 | if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) { | |
4703 | reason = "already exists"; | |
4704 | break; | |
4705 | } | |
4706 | } | |
4707 | ||
4708 | if (reason && !quiet) | |
e7b84f9d | 4709 | pr_err("imsm volume name %s\n", reason); |
aa534678 DW |
4710 | |
4711 | return !reason; | |
4712 | } | |
4713 | ||
8b353278 DW |
4714 | static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info, |
4715 | unsigned long long size, char *name, | |
83cd1e97 N |
4716 | char *homehost, int *uuid, |
4717 | long long data_offset) | |
cdddbdbc | 4718 | { |
c2c087e6 DW |
4719 | /* We are creating a volume inside a pre-existing container. |
4720 | * so st->sb is already set. | |
4721 | */ | |
4722 | struct intel_super *super = st->sb; | |
949c47a0 | 4723 | struct imsm_super *mpb = super->anchor; |
ba2de7ba | 4724 | struct intel_dev *dv; |
c2c087e6 DW |
4725 | struct imsm_dev *dev; |
4726 | struct imsm_vol *vol; | |
4727 | struct imsm_map *map; | |
4728 | int idx = mpb->num_raid_devs; | |
4729 | int i; | |
4730 | unsigned long long array_blocks; | |
2c092cad | 4731 | size_t size_old, size_new; |
5551b113 | 4732 | unsigned long long num_data_stripes; |
cdddbdbc | 4733 | |
88c32bb1 | 4734 | if (super->orom && mpb->num_raid_devs >= super->orom->vpa) { |
7a862a02 | 4735 | pr_err("This imsm-container already has the maximum of %d volumes\n", super->orom->vpa); |
c2c087e6 DW |
4736 | return 0; |
4737 | } | |
4738 | ||
2c092cad DW |
4739 | /* ensure the mpb is large enough for the new data */ |
4740 | size_old = __le32_to_cpu(mpb->mpb_size); | |
4741 | size_new = disks_to_mpb_size(info->nr_disks); | |
4742 | if (size_new > size_old) { | |
4743 | void *mpb_new; | |
4744 | size_t size_round = ROUND_UP(size_new, 512); | |
4745 | ||
4746 | if (posix_memalign(&mpb_new, 512, size_round) != 0) { | |
e7b84f9d | 4747 | pr_err("could not allocate new mpb\n"); |
2c092cad DW |
4748 | return 0; |
4749 | } | |
17a4eaf9 AK |
4750 | if (posix_memalign(&super->migr_rec_buf, 512, |
4751 | MIGR_REC_BUF_SIZE) != 0) { | |
1ade5cc1 | 4752 | pr_err("could not allocate migr_rec buffer\n"); |
8e59f3d8 AK |
4753 | free(super->buf); |
4754 | free(super); | |
ea944c8f | 4755 | free(mpb_new); |
8e59f3d8 AK |
4756 | return 0; |
4757 | } | |
2c092cad DW |
4758 | memcpy(mpb_new, mpb, size_old); |
4759 | free(mpb); | |
4760 | mpb = mpb_new; | |
949c47a0 | 4761 | super->anchor = mpb_new; |
2c092cad DW |
4762 | mpb->mpb_size = __cpu_to_le32(size_new); |
4763 | memset(mpb_new + size_old, 0, size_round - size_old); | |
4764 | } | |
bf5a934a | 4765 | super->current_vol = idx; |
3960e579 DW |
4766 | |
4767 | /* handle 'failed_disks' by either: | |
4768 | * a) create dummy disk entries in the table if this the first | |
4769 | * volume in the array. We add them here as this is the only | |
4770 | * opportunity to add them. add_to_super_imsm_volume() | |
4771 | * handles the non-failed disks and continues incrementing | |
4772 | * mpb->num_disks. | |
4773 | * b) validate that 'failed_disks' matches the current number | |
4774 | * of missing disks if the container is populated | |
d23fe947 | 4775 | */ |
3960e579 | 4776 | if (super->current_vol == 0) { |
d23fe947 | 4777 | mpb->num_disks = 0; |
3960e579 DW |
4778 | for (i = 0; i < info->failed_disks; i++) { |
4779 | struct imsm_disk *disk; | |
4780 | ||
4781 | mpb->num_disks++; | |
4782 | disk = __get_imsm_disk(mpb, i); | |
4783 | disk->status = CONFIGURED_DISK | FAILED_DISK; | |
4784 | disk->scsi_id = __cpu_to_le32(~(__u32)0); | |
4785 | snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN, | |
4786 | "missing:%d", i); | |
4787 | } | |
4788 | find_missing(super); | |
4789 | } else { | |
4790 | int missing = 0; | |
4791 | struct dl *d; | |
4792 | ||
4793 | for (d = super->missing; d; d = d->next) | |
4794 | missing++; | |
4795 | if (info->failed_disks > missing) { | |
e7b84f9d | 4796 | pr_err("unable to add 'missing' disk to container\n"); |
3960e579 DW |
4797 | return 0; |
4798 | } | |
4799 | } | |
5a038140 | 4800 | |
aa534678 DW |
4801 | if (!check_name(super, name, 0)) |
4802 | return 0; | |
503975b9 N |
4803 | dv = xmalloc(sizeof(*dv)); |
4804 | dev = xcalloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1)); | |
c2c087e6 | 4805 | strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN); |
e03640bd | 4806 | array_blocks = calc_array_size(info->level, info->raid_disks, |
03bcbc65 | 4807 | info->layout, info->chunk_size, |
5551b113 | 4808 | size * 2); |
979d38be DW |
4809 | /* round array size down to closest MB */ |
4810 | array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; | |
4811 | ||
c2c087e6 DW |
4812 | dev->size_low = __cpu_to_le32((__u32) array_blocks); |
4813 | dev->size_high = __cpu_to_le32((__u32) (array_blocks >> 32)); | |
1a2487c2 | 4814 | dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING); |
c2c087e6 DW |
4815 | vol = &dev->vol; |
4816 | vol->migr_state = 0; | |
1484e727 | 4817 | set_migr_type(dev, MIGR_INIT); |
3960e579 | 4818 | vol->dirty = !info->state; |
f8f603f1 | 4819 | vol->curr_migr_unit = 0; |
238c0a71 | 4820 | map = get_imsm_map(dev, MAP_0); |
5551b113 CA |
4821 | set_pba_of_lba0(map, super->create_offset); |
4822 | set_blocks_per_member(map, info_to_blocks_per_member(info, size)); | |
ef6ffade | 4823 | map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info)); |
0556e1a2 | 4824 | map->failed_disk_num = ~0; |
bf4442ab | 4825 | if (info->level > 0) |
fffaf1ff N |
4826 | map->map_state = (info->state ? IMSM_T_STATE_NORMAL |
4827 | : IMSM_T_STATE_UNINITIALIZED); | |
bf4442ab AK |
4828 | else |
4829 | map->map_state = info->failed_disks ? IMSM_T_STATE_FAILED : | |
4830 | IMSM_T_STATE_NORMAL; | |
252d23c0 | 4831 | map->ddf = 1; |
ef6ffade DW |
4832 | |
4833 | if (info->level == 1 && info->raid_disks > 2) { | |
38950822 AW |
4834 | free(dev); |
4835 | free(dv); | |
7a862a02 | 4836 | pr_err("imsm does not support more than 2 disksin a raid1 volume\n"); |
ef6ffade DW |
4837 | return 0; |
4838 | } | |
81062a36 DW |
4839 | |
4840 | map->raid_level = info->level; | |
4d1313e9 | 4841 | if (info->level == 10) { |
c2c087e6 | 4842 | map->raid_level = 1; |
4d1313e9 | 4843 | map->num_domains = info->raid_disks / 2; |
81062a36 DW |
4844 | } else if (info->level == 1) |
4845 | map->num_domains = info->raid_disks; | |
4846 | else | |
ff596308 | 4847 | map->num_domains = 1; |
81062a36 | 4848 | |
5551b113 CA |
4849 | /* info->size is only int so use the 'size' parameter instead */ |
4850 | num_data_stripes = (size * 2) / info_to_blocks_per_strip(info); | |
4851 | num_data_stripes /= map->num_domains; | |
4852 | set_num_data_stripes(map, num_data_stripes); | |
ef6ffade | 4853 | |
c2c087e6 DW |
4854 | map->num_members = info->raid_disks; |
4855 | for (i = 0; i < map->num_members; i++) { | |
4856 | /* initialized in add_to_super */ | |
4eb26970 | 4857 | set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD); |
c2c087e6 | 4858 | } |
949c47a0 | 4859 | mpb->num_raid_devs++; |
ba2de7ba DW |
4860 | |
4861 | dv->dev = dev; | |
4862 | dv->index = super->current_vol; | |
4863 | dv->next = super->devlist; | |
4864 | super->devlist = dv; | |
c2c087e6 | 4865 | |
4d1313e9 DW |
4866 | imsm_update_version_info(super); |
4867 | ||
c2c087e6 | 4868 | return 1; |
cdddbdbc DW |
4869 | } |
4870 | ||
bf5a934a DW |
4871 | static int init_super_imsm(struct supertype *st, mdu_array_info_t *info, |
4872 | unsigned long long size, char *name, | |
83cd1e97 N |
4873 | char *homehost, int *uuid, |
4874 | unsigned long long data_offset) | |
bf5a934a DW |
4875 | { |
4876 | /* This is primarily called by Create when creating a new array. | |
4877 | * We will then get add_to_super called for each component, and then | |
4878 | * write_init_super called to write it out to each device. | |
4879 | * For IMSM, Create can create on fresh devices or on a pre-existing | |
4880 | * array. | |
4881 | * To create on a pre-existing array a different method will be called. | |
4882 | * This one is just for fresh drives. | |
4883 | */ | |
4884 | struct intel_super *super; | |
4885 | struct imsm_super *mpb; | |
4886 | size_t mpb_size; | |
4d1313e9 | 4887 | char *version; |
bf5a934a | 4888 | |
83cd1e97 | 4889 | if (data_offset != INVALID_SECTORS) { |
ed503f89 | 4890 | pr_err("data-offset not supported by imsm\n"); |
83cd1e97 N |
4891 | return 0; |
4892 | } | |
4893 | ||
bf5a934a | 4894 | if (st->sb) |
83cd1e97 N |
4895 | return init_super_imsm_volume(st, info, size, name, homehost, uuid, |
4896 | data_offset); | |
e683ca88 DW |
4897 | |
4898 | if (info) | |
4899 | mpb_size = disks_to_mpb_size(info->nr_disks); | |
4900 | else | |
4901 | mpb_size = 512; | |
bf5a934a | 4902 | |
49133e57 | 4903 | super = alloc_super(); |
e683ca88 | 4904 | if (super && posix_memalign(&super->buf, 512, mpb_size) != 0) { |
bf5a934a | 4905 | free(super); |
e683ca88 DW |
4906 | super = NULL; |
4907 | } | |
4908 | if (!super) { | |
1ade5cc1 | 4909 | pr_err("could not allocate superblock\n"); |
bf5a934a DW |
4910 | return 0; |
4911 | } | |
17a4eaf9 | 4912 | if (posix_memalign(&super->migr_rec_buf, 512, MIGR_REC_BUF_SIZE) != 0) { |
1ade5cc1 | 4913 | pr_err("could not allocate migr_rec buffer\n"); |
8e59f3d8 AK |
4914 | free(super->buf); |
4915 | free(super); | |
4916 | return 0; | |
4917 | } | |
e683ca88 | 4918 | memset(super->buf, 0, mpb_size); |
ef649044 | 4919 | mpb = super->buf; |
e683ca88 DW |
4920 | mpb->mpb_size = __cpu_to_le32(mpb_size); |
4921 | st->sb = super; | |
4922 | ||
4923 | if (info == NULL) { | |
4924 | /* zeroing superblock */ | |
4925 | return 0; | |
4926 | } | |
bf5a934a | 4927 | |
4d1313e9 DW |
4928 | mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; |
4929 | ||
4930 | version = (char *) mpb->sig; | |
4931 | strcpy(version, MPB_SIGNATURE); | |
4932 | version += strlen(MPB_SIGNATURE); | |
4933 | strcpy(version, MPB_VERSION_RAID0); | |
bf5a934a | 4934 | |
bf5a934a DW |
4935 | return 1; |
4936 | } | |
4937 | ||
0e600426 | 4938 | #ifndef MDASSEMBLE |
f20c3968 | 4939 | static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk, |
bf5a934a DW |
4940 | int fd, char *devname) |
4941 | { | |
4942 | struct intel_super *super = st->sb; | |
d23fe947 | 4943 | struct imsm_super *mpb = super->anchor; |
3960e579 | 4944 | struct imsm_disk *_disk; |
bf5a934a DW |
4945 | struct imsm_dev *dev; |
4946 | struct imsm_map *map; | |
3960e579 | 4947 | struct dl *dl, *df; |
4eb26970 | 4948 | int slot; |
bf5a934a | 4949 | |
949c47a0 | 4950 | dev = get_imsm_dev(super, super->current_vol); |
238c0a71 | 4951 | map = get_imsm_map(dev, MAP_0); |
bf5a934a | 4952 | |
208933a7 | 4953 | if (! (dk->state & (1<<MD_DISK_SYNC))) { |
e7b84f9d | 4954 | pr_err("%s: Cannot add spare devices to IMSM volume\n", |
208933a7 N |
4955 | devname); |
4956 | return 1; | |
4957 | } | |
4958 | ||
efb30e7f DW |
4959 | if (fd == -1) { |
4960 | /* we're doing autolayout so grab the pre-marked (in | |
4961 | * validate_geometry) raid_disk | |
4962 | */ | |
4963 | for (dl = super->disks; dl; dl = dl->next) | |
4964 | if (dl->raiddisk == dk->raid_disk) | |
4965 | break; | |
4966 | } else { | |
4967 | for (dl = super->disks; dl ; dl = dl->next) | |
4968 | if (dl->major == dk->major && | |
4969 | dl->minor == dk->minor) | |
4970 | break; | |
4971 | } | |
d23fe947 | 4972 | |
208933a7 | 4973 | if (!dl) { |
e7b84f9d | 4974 | pr_err("%s is not a member of the same container\n", devname); |
f20c3968 | 4975 | return 1; |
208933a7 | 4976 | } |
bf5a934a | 4977 | |
d23fe947 DW |
4978 | /* add a pristine spare to the metadata */ |
4979 | if (dl->index < 0) { | |
4980 | dl->index = super->anchor->num_disks; | |
4981 | super->anchor->num_disks++; | |
4982 | } | |
4eb26970 DW |
4983 | /* Check the device has not already been added */ |
4984 | slot = get_imsm_disk_slot(map, dl->index); | |
4985 | if (slot >= 0 && | |
238c0a71 | 4986 | (get_imsm_ord_tbl_ent(dev, slot, MAP_X) & IMSM_ORD_REBUILD) == 0) { |
e7b84f9d | 4987 | pr_err("%s has been included in this array twice\n", |
4eb26970 DW |
4988 | devname); |
4989 | return 1; | |
4990 | } | |
656b6b5a | 4991 | set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index); |
ee5aad5a | 4992 | dl->disk.status = CONFIGURED_DISK; |
d23fe947 | 4993 | |
3960e579 DW |
4994 | /* update size of 'missing' disks to be at least as large as the |
4995 | * largest acitve member (we only have dummy missing disks when | |
4996 | * creating the first volume) | |
4997 | */ | |
4998 | if (super->current_vol == 0) { | |
4999 | for (df = super->missing; df; df = df->next) { | |
5551b113 CA |
5000 | if (total_blocks(&dl->disk) > total_blocks(&df->disk)) |
5001 | set_total_blocks(&df->disk, total_blocks(&dl->disk)); | |
3960e579 DW |
5002 | _disk = __get_imsm_disk(mpb, df->index); |
5003 | *_disk = df->disk; | |
5004 | } | |
5005 | } | |
5006 | ||
5007 | /* refresh unset/failed slots to point to valid 'missing' entries */ | |
5008 | for (df = super->missing; df; df = df->next) | |
5009 | for (slot = 0; slot < mpb->num_disks; slot++) { | |
238c0a71 | 5010 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); |
3960e579 DW |
5011 | |
5012 | if ((ord & IMSM_ORD_REBUILD) == 0) | |
5013 | continue; | |
5014 | set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD); | |
1ace8403 | 5015 | if (is_gen_migration(dev)) { |
238c0a71 AK |
5016 | struct imsm_map *map2 = get_imsm_map(dev, |
5017 | MAP_1); | |
0a108d63 AK |
5018 | int slot2 = get_imsm_disk_slot(map2, df->index); |
5019 | if ((slot2 < map2->num_members) && | |
5020 | (slot2 >= 0)) { | |
1ace8403 | 5021 | __u32 ord2 = get_imsm_ord_tbl_ent(dev, |
238c0a71 AK |
5022 | slot2, |
5023 | MAP_1); | |
1ace8403 AK |
5024 | if ((unsigned)df->index == |
5025 | ord_to_idx(ord2)) | |
5026 | set_imsm_ord_tbl_ent(map2, | |
0a108d63 | 5027 | slot2, |
1ace8403 AK |
5028 | df->index | |
5029 | IMSM_ORD_REBUILD); | |
5030 | } | |
5031 | } | |
3960e579 DW |
5032 | dprintf("set slot:%d to missing disk:%d\n", slot, df->index); |
5033 | break; | |
5034 | } | |
5035 | ||
d23fe947 DW |
5036 | /* if we are creating the first raid device update the family number */ |
5037 | if (super->current_vol == 0) { | |
5038 | __u32 sum; | |
5039 | struct imsm_dev *_dev = __get_imsm_dev(mpb, 0); | |
d23fe947 | 5040 | |
3960e579 | 5041 | _disk = __get_imsm_disk(mpb, dl->index); |
791b666a | 5042 | if (!_dev || !_disk) { |
e7b84f9d | 5043 | pr_err("BUG mpb setup error\n"); |
791b666a AW |
5044 | return 1; |
5045 | } | |
d23fe947 DW |
5046 | *_dev = *dev; |
5047 | *_disk = dl->disk; | |
148acb7b DW |
5048 | sum = random32(); |
5049 | sum += __gen_imsm_checksum(mpb); | |
d23fe947 | 5050 | mpb->family_num = __cpu_to_le32(sum); |
148acb7b | 5051 | mpb->orig_family_num = mpb->family_num; |
d23fe947 | 5052 | } |
ca0748fa | 5053 | super->current_disk = dl; |
f20c3968 | 5054 | return 0; |
bf5a934a DW |
5055 | } |
5056 | ||
a8619d23 AK |
5057 | /* mark_spare() |
5058 | * Function marks disk as spare and restores disk serial | |
5059 | * in case it was previously marked as failed by takeover operation | |
5060 | * reruns: | |
5061 | * -1 : critical error | |
5062 | * 0 : disk is marked as spare but serial is not set | |
5063 | * 1 : success | |
5064 | */ | |
5065 | int mark_spare(struct dl *disk) | |
5066 | { | |
5067 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
5068 | int ret_val = -1; | |
5069 | ||
5070 | if (!disk) | |
5071 | return ret_val; | |
5072 | ||
5073 | ret_val = 0; | |
5074 | if (!imsm_read_serial(disk->fd, NULL, serial)) { | |
5075 | /* Restore disk serial number, because takeover marks disk | |
5076 | * as failed and adds to serial ':0' before it becomes | |
5077 | * a spare disk. | |
5078 | */ | |
5079 | serialcpy(disk->serial, serial); | |
5080 | serialcpy(disk->disk.serial, serial); | |
5081 | ret_val = 1; | |
5082 | } | |
5083 | disk->disk.status = SPARE_DISK; | |
5084 | disk->index = -1; | |
5085 | ||
5086 | return ret_val; | |
5087 | } | |
88654014 | 5088 | |
f20c3968 | 5089 | static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk, |
72ca9bcf N |
5090 | int fd, char *devname, |
5091 | unsigned long long data_offset) | |
cdddbdbc | 5092 | { |
c2c087e6 | 5093 | struct intel_super *super = st->sb; |
c2c087e6 DW |
5094 | struct dl *dd; |
5095 | unsigned long long size; | |
f2f27e63 | 5096 | __u32 id; |
c2c087e6 DW |
5097 | int rv; |
5098 | struct stat stb; | |
5099 | ||
88654014 LM |
5100 | /* If we are on an RAID enabled platform check that the disk is |
5101 | * attached to the raid controller. | |
5102 | * We do not need to test disks attachment for container based additions, | |
5103 | * they shall be already tested when container was created/assembled. | |
88c32bb1 | 5104 | */ |
d424212e | 5105 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 5106 | /* no orom/efi or non-intel hba of the disk */ |
f0f5a016 LM |
5107 | if (rv != 0) { |
5108 | dprintf("capability: %p fd: %d ret: %d\n", | |
5109 | super->orom, fd, rv); | |
5110 | return 1; | |
88c32bb1 DW |
5111 | } |
5112 | ||
f20c3968 DW |
5113 | if (super->current_vol >= 0) |
5114 | return add_to_super_imsm_volume(st, dk, fd, devname); | |
bf5a934a | 5115 | |
c2c087e6 | 5116 | fstat(fd, &stb); |
503975b9 | 5117 | dd = xcalloc(sizeof(*dd), 1); |
c2c087e6 DW |
5118 | dd->major = major(stb.st_rdev); |
5119 | dd->minor = minor(stb.st_rdev); | |
503975b9 | 5120 | dd->devname = devname ? xstrdup(devname) : NULL; |
c2c087e6 | 5121 | dd->fd = fd; |
689c9bf3 | 5122 | dd->e = NULL; |
1a64be56 | 5123 | dd->action = DISK_ADD; |
c2c087e6 | 5124 | rv = imsm_read_serial(fd, devname, dd->serial); |
32ba9157 | 5125 | if (rv) { |
e7b84f9d | 5126 | pr_err("failed to retrieve scsi serial, aborting\n"); |
949c47a0 | 5127 | free(dd); |
0030e8d6 | 5128 | abort(); |
c2c087e6 DW |
5129 | } |
5130 | ||
c2c087e6 | 5131 | get_dev_size(fd, NULL, &size); |
71e5411e PB |
5132 | /* clear migr_rec when adding disk to container */ |
5133 | memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SIZE); | |
5134 | if (lseek64(fd, size - MIGR_REC_POSITION, SEEK_SET) >= 0) { | |
5135 | if (write(fd, super->migr_rec_buf, | |
5136 | MIGR_REC_BUF_SIZE) != MIGR_REC_BUF_SIZE) | |
5137 | perror("Write migr_rec failed"); | |
5138 | } | |
5139 | ||
c2c087e6 | 5140 | size /= 512; |
1f24f035 | 5141 | serialcpy(dd->disk.serial, dd->serial); |
5551b113 CA |
5142 | set_total_blocks(&dd->disk, size); |
5143 | if (__le32_to_cpu(dd->disk.total_blocks_hi) > 0) { | |
5144 | struct imsm_super *mpb = super->anchor; | |
5145 | mpb->attributes |= MPB_ATTRIB_2TB_DISK; | |
5146 | } | |
a8619d23 | 5147 | mark_spare(dd); |
c2c087e6 | 5148 | if (sysfs_disk_to_scsi_id(fd, &id) == 0) |
b9f594fe | 5149 | dd->disk.scsi_id = __cpu_to_le32(id); |
c2c087e6 | 5150 | else |
b9f594fe | 5151 | dd->disk.scsi_id = __cpu_to_le32(0); |
43dad3d6 DW |
5152 | |
5153 | if (st->update_tail) { | |
1a64be56 LM |
5154 | dd->next = super->disk_mgmt_list; |
5155 | super->disk_mgmt_list = dd; | |
43dad3d6 DW |
5156 | } else { |
5157 | dd->next = super->disks; | |
5158 | super->disks = dd; | |
ceaf0ee1 | 5159 | super->updates_pending++; |
43dad3d6 | 5160 | } |
f20c3968 DW |
5161 | |
5162 | return 0; | |
cdddbdbc DW |
5163 | } |
5164 | ||
1a64be56 LM |
5165 | static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk) |
5166 | { | |
5167 | struct intel_super *super = st->sb; | |
5168 | struct dl *dd; | |
5169 | ||
5170 | /* remove from super works only in mdmon - for communication | |
5171 | * manager - monitor. Check if communication memory buffer | |
5172 | * is prepared. | |
5173 | */ | |
5174 | if (!st->update_tail) { | |
1ade5cc1 | 5175 | pr_err("shall be used in mdmon context only\n"); |
1a64be56 LM |
5176 | return 1; |
5177 | } | |
503975b9 | 5178 | dd = xcalloc(1, sizeof(*dd)); |
1a64be56 LM |
5179 | dd->major = dk->major; |
5180 | dd->minor = dk->minor; | |
1a64be56 | 5181 | dd->fd = -1; |
a8619d23 | 5182 | mark_spare(dd); |
1a64be56 LM |
5183 | dd->action = DISK_REMOVE; |
5184 | ||
5185 | dd->next = super->disk_mgmt_list; | |
5186 | super->disk_mgmt_list = dd; | |
5187 | ||
1a64be56 LM |
5188 | return 0; |
5189 | } | |
5190 | ||
f796af5d DW |
5191 | static int store_imsm_mpb(int fd, struct imsm_super *mpb); |
5192 | ||
5193 | static union { | |
5194 | char buf[512]; | |
5195 | struct imsm_super anchor; | |
5196 | } spare_record __attribute__ ((aligned(512))); | |
c2c087e6 | 5197 | |
d23fe947 DW |
5198 | /* spare records have their own family number and do not have any defined raid |
5199 | * devices | |
5200 | */ | |
5201 | static int write_super_imsm_spares(struct intel_super *super, int doclose) | |
5202 | { | |
d23fe947 | 5203 | struct imsm_super *mpb = super->anchor; |
f796af5d | 5204 | struct imsm_super *spare = &spare_record.anchor; |
d23fe947 DW |
5205 | __u32 sum; |
5206 | struct dl *d; | |
5207 | ||
68641cdb JS |
5208 | spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super)); |
5209 | spare->generation_num = __cpu_to_le32(1UL); | |
f796af5d | 5210 | spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; |
68641cdb JS |
5211 | spare->num_disks = 1; |
5212 | spare->num_raid_devs = 0; | |
5213 | spare->cache_size = mpb->cache_size; | |
5214 | spare->pwr_cycle_count = __cpu_to_le32(1); | |
f796af5d DW |
5215 | |
5216 | snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH, | |
5217 | MPB_SIGNATURE MPB_VERSION_RAID0); | |
d23fe947 DW |
5218 | |
5219 | for (d = super->disks; d; d = d->next) { | |
8796fdc4 | 5220 | if (d->index != -1) |
d23fe947 DW |
5221 | continue; |
5222 | ||
f796af5d | 5223 | spare->disk[0] = d->disk; |
027c374f CA |
5224 | if (__le32_to_cpu(d->disk.total_blocks_hi) > 0) |
5225 | spare->attributes |= MPB_ATTRIB_2TB_DISK; | |
5226 | ||
f796af5d DW |
5227 | sum = __gen_imsm_checksum(spare); |
5228 | spare->family_num = __cpu_to_le32(sum); | |
5229 | spare->orig_family_num = 0; | |
5230 | sum = __gen_imsm_checksum(spare); | |
5231 | spare->check_sum = __cpu_to_le32(sum); | |
d23fe947 | 5232 | |
f796af5d | 5233 | if (store_imsm_mpb(d->fd, spare)) { |
1ade5cc1 N |
5234 | pr_err("failed for device %d:%d %s\n", |
5235 | d->major, d->minor, strerror(errno)); | |
e74255d9 | 5236 | return 1; |
d23fe947 DW |
5237 | } |
5238 | if (doclose) { | |
5239 | close(d->fd); | |
5240 | d->fd = -1; | |
5241 | } | |
5242 | } | |
5243 | ||
e74255d9 | 5244 | return 0; |
d23fe947 DW |
5245 | } |
5246 | ||
36988a3d | 5247 | static int write_super_imsm(struct supertype *st, int doclose) |
cdddbdbc | 5248 | { |
36988a3d | 5249 | struct intel_super *super = st->sb; |
949c47a0 | 5250 | struct imsm_super *mpb = super->anchor; |
c2c087e6 DW |
5251 | struct dl *d; |
5252 | __u32 generation; | |
5253 | __u32 sum; | |
d23fe947 | 5254 | int spares = 0; |
949c47a0 | 5255 | int i; |
a48ac0a8 | 5256 | __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk); |
36988a3d | 5257 | int num_disks = 0; |
146c6260 | 5258 | int clear_migration_record = 1; |
cdddbdbc | 5259 | |
c2c087e6 DW |
5260 | /* 'generation' is incremented everytime the metadata is written */ |
5261 | generation = __le32_to_cpu(mpb->generation_num); | |
5262 | generation++; | |
5263 | mpb->generation_num = __cpu_to_le32(generation); | |
5264 | ||
148acb7b DW |
5265 | /* fix up cases where previous mdadm releases failed to set |
5266 | * orig_family_num | |
5267 | */ | |
5268 | if (mpb->orig_family_num == 0) | |
5269 | mpb->orig_family_num = mpb->family_num; | |
5270 | ||
d23fe947 | 5271 | for (d = super->disks; d; d = d->next) { |
8796fdc4 | 5272 | if (d->index == -1) |
d23fe947 | 5273 | spares++; |
36988a3d | 5274 | else { |
d23fe947 | 5275 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
5276 | num_disks++; |
5277 | } | |
d23fe947 | 5278 | } |
36988a3d | 5279 | for (d = super->missing; d; d = d->next) { |
47ee5a45 | 5280 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
5281 | num_disks++; |
5282 | } | |
5283 | mpb->num_disks = num_disks; | |
5284 | mpb_size += sizeof(struct imsm_disk) * mpb->num_disks; | |
b9f594fe | 5285 | |
949c47a0 DW |
5286 | for (i = 0; i < mpb->num_raid_devs; i++) { |
5287 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
36988a3d AK |
5288 | struct imsm_dev *dev2 = get_imsm_dev(super, i); |
5289 | if (dev && dev2) { | |
5290 | imsm_copy_dev(dev, dev2); | |
5291 | mpb_size += sizeof_imsm_dev(dev, 0); | |
5292 | } | |
146c6260 AK |
5293 | if (is_gen_migration(dev2)) |
5294 | clear_migration_record = 0; | |
949c47a0 | 5295 | } |
a48ac0a8 DW |
5296 | mpb_size += __le32_to_cpu(mpb->bbm_log_size); |
5297 | mpb->mpb_size = __cpu_to_le32(mpb_size); | |
949c47a0 | 5298 | |
c2c087e6 | 5299 | /* recalculate checksum */ |
949c47a0 | 5300 | sum = __gen_imsm_checksum(mpb); |
c2c087e6 DW |
5301 | mpb->check_sum = __cpu_to_le32(sum); |
5302 | ||
51d83f5d AK |
5303 | if (super->clean_migration_record_by_mdmon) { |
5304 | clear_migration_record = 1; | |
5305 | super->clean_migration_record_by_mdmon = 0; | |
5306 | } | |
146c6260 | 5307 | if (clear_migration_record) |
17a4eaf9 | 5308 | memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SIZE); |
146c6260 | 5309 | |
d23fe947 | 5310 | /* write the mpb for disks that compose raid devices */ |
c2c087e6 | 5311 | for (d = super->disks; d ; d = d->next) { |
86c54047 | 5312 | if (d->index < 0 || is_failed(&d->disk)) |
d23fe947 | 5313 | continue; |
30602f53 | 5314 | |
146c6260 AK |
5315 | if (clear_migration_record) { |
5316 | unsigned long long dsize; | |
5317 | ||
5318 | get_dev_size(d->fd, NULL, &dsize); | |
5319 | if (lseek64(d->fd, dsize - 512, SEEK_SET) >= 0) { | |
17a4eaf9 AK |
5320 | if (write(d->fd, super->migr_rec_buf, |
5321 | MIGR_REC_BUF_SIZE) != MIGR_REC_BUF_SIZE) | |
9e2d750d | 5322 | perror("Write migr_rec failed"); |
146c6260 AK |
5323 | } |
5324 | } | |
51d83f5d AK |
5325 | |
5326 | if (store_imsm_mpb(d->fd, mpb)) | |
5327 | fprintf(stderr, | |
1ade5cc1 N |
5328 | "failed for device %d:%d (fd: %d)%s\n", |
5329 | d->major, d->minor, | |
51d83f5d AK |
5330 | d->fd, strerror(errno)); |
5331 | ||
c2c087e6 DW |
5332 | if (doclose) { |
5333 | close(d->fd); | |
5334 | d->fd = -1; | |
5335 | } | |
5336 | } | |
5337 | ||
d23fe947 DW |
5338 | if (spares) |
5339 | return write_super_imsm_spares(super, doclose); | |
5340 | ||
e74255d9 | 5341 | return 0; |
c2c087e6 DW |
5342 | } |
5343 | ||
9b1fb677 | 5344 | static int create_array(struct supertype *st, int dev_idx) |
43dad3d6 DW |
5345 | { |
5346 | size_t len; | |
5347 | struct imsm_update_create_array *u; | |
5348 | struct intel_super *super = st->sb; | |
9b1fb677 | 5349 | struct imsm_dev *dev = get_imsm_dev(super, dev_idx); |
238c0a71 | 5350 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
54c2c1ea DW |
5351 | struct disk_info *inf; |
5352 | struct imsm_disk *disk; | |
5353 | int i; | |
43dad3d6 | 5354 | |
54c2c1ea DW |
5355 | len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) + |
5356 | sizeof(*inf) * map->num_members; | |
503975b9 | 5357 | u = xmalloc(len); |
43dad3d6 | 5358 | u->type = update_create_array; |
9b1fb677 | 5359 | u->dev_idx = dev_idx; |
43dad3d6 | 5360 | imsm_copy_dev(&u->dev, dev); |
54c2c1ea DW |
5361 | inf = get_disk_info(u); |
5362 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 5363 | int idx = get_imsm_disk_idx(dev, i, MAP_X); |
9b1fb677 | 5364 | |
54c2c1ea | 5365 | disk = get_imsm_disk(super, idx); |
1ca5c8e0 N |
5366 | if (!disk) |
5367 | disk = get_imsm_missing(super, idx); | |
54c2c1ea DW |
5368 | serialcpy(inf[i].serial, disk->serial); |
5369 | } | |
43dad3d6 DW |
5370 | append_metadata_update(st, u, len); |
5371 | ||
5372 | return 0; | |
5373 | } | |
5374 | ||
1a64be56 | 5375 | static int mgmt_disk(struct supertype *st) |
43dad3d6 DW |
5376 | { |
5377 | struct intel_super *super = st->sb; | |
5378 | size_t len; | |
1a64be56 | 5379 | struct imsm_update_add_remove_disk *u; |
43dad3d6 | 5380 | |
1a64be56 | 5381 | if (!super->disk_mgmt_list) |
43dad3d6 DW |
5382 | return 0; |
5383 | ||
5384 | len = sizeof(*u); | |
503975b9 | 5385 | u = xmalloc(len); |
1a64be56 | 5386 | u->type = update_add_remove_disk; |
43dad3d6 DW |
5387 | append_metadata_update(st, u, len); |
5388 | ||
5389 | return 0; | |
5390 | } | |
5391 | ||
c2c087e6 DW |
5392 | static int write_init_super_imsm(struct supertype *st) |
5393 | { | |
9b1fb677 DW |
5394 | struct intel_super *super = st->sb; |
5395 | int current_vol = super->current_vol; | |
5396 | ||
5397 | /* we are done with current_vol reset it to point st at the container */ | |
5398 | super->current_vol = -1; | |
5399 | ||
8273f55e | 5400 | if (st->update_tail) { |
43dad3d6 DW |
5401 | /* queue the recently created array / added disk |
5402 | * as a metadata update */ | |
43dad3d6 | 5403 | int rv; |
8273f55e | 5404 | |
43dad3d6 | 5405 | /* determine if we are creating a volume or adding a disk */ |
9b1fb677 | 5406 | if (current_vol < 0) { |
1a64be56 LM |
5407 | /* in the mgmt (add/remove) disk case we are running |
5408 | * in mdmon context, so don't close fd's | |
43dad3d6 | 5409 | */ |
1a64be56 | 5410 | return mgmt_disk(st); |
43dad3d6 | 5411 | } else |
9b1fb677 | 5412 | rv = create_array(st, current_vol); |
8273f55e | 5413 | |
43dad3d6 | 5414 | return rv; |
d682f344 N |
5415 | } else { |
5416 | struct dl *d; | |
5417 | for (d = super->disks; d; d = d->next) | |
ba728be7 | 5418 | Kill(d->devname, NULL, 0, -1, 1); |
36988a3d | 5419 | return write_super_imsm(st, 1); |
d682f344 | 5420 | } |
cdddbdbc | 5421 | } |
0e600426 | 5422 | #endif |
cdddbdbc | 5423 | |
e683ca88 | 5424 | static int store_super_imsm(struct supertype *st, int fd) |
cdddbdbc | 5425 | { |
e683ca88 DW |
5426 | struct intel_super *super = st->sb; |
5427 | struct imsm_super *mpb = super ? super->anchor : NULL; | |
551c80c1 | 5428 | |
e683ca88 | 5429 | if (!mpb) |
ad97895e DW |
5430 | return 1; |
5431 | ||
1799c9e8 | 5432 | #ifndef MDASSEMBLE |
e683ca88 | 5433 | return store_imsm_mpb(fd, mpb); |
1799c9e8 N |
5434 | #else |
5435 | return 1; | |
5436 | #endif | |
cdddbdbc DW |
5437 | } |
5438 | ||
0e600426 N |
5439 | static int imsm_bbm_log_size(struct imsm_super *mpb) |
5440 | { | |
5441 | return __le32_to_cpu(mpb->bbm_log_size); | |
5442 | } | |
5443 | ||
5444 | #ifndef MDASSEMBLE | |
cdddbdbc DW |
5445 | static int validate_geometry_imsm_container(struct supertype *st, int level, |
5446 | int layout, int raiddisks, int chunk, | |
af4348dd N |
5447 | unsigned long long size, |
5448 | unsigned long long data_offset, | |
5449 | char *dev, | |
2c514b71 NB |
5450 | unsigned long long *freesize, |
5451 | int verbose) | |
cdddbdbc | 5452 | { |
c2c087e6 DW |
5453 | int fd; |
5454 | unsigned long long ldsize; | |
f2f5c343 LM |
5455 | struct intel_super *super=NULL; |
5456 | int rv = 0; | |
cdddbdbc | 5457 | |
c2c087e6 DW |
5458 | if (level != LEVEL_CONTAINER) |
5459 | return 0; | |
5460 | if (!dev) | |
5461 | return 1; | |
5462 | ||
5463 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
5464 | if (fd < 0) { | |
ba728be7 | 5465 | if (verbose > 0) |
e7b84f9d | 5466 | pr_err("imsm: Cannot open %s: %s\n", |
2c514b71 | 5467 | dev, strerror(errno)); |
c2c087e6 DW |
5468 | return 0; |
5469 | } | |
5470 | if (!get_dev_size(fd, dev, &ldsize)) { | |
5471 | close(fd); | |
5472 | return 0; | |
5473 | } | |
f2f5c343 LM |
5474 | |
5475 | /* capabilities retrieve could be possible | |
5476 | * note that there is no fd for the disks in array. | |
5477 | */ | |
5478 | super = alloc_super(); | |
ba728be7 | 5479 | rv = find_intel_hba_capability(fd, super, verbose > 0 ? dev : NULL); |
f2f5c343 LM |
5480 | if (rv != 0) { |
5481 | #if DEBUG | |
5482 | char str[256]; | |
5483 | fd2devname(fd, str); | |
1ade5cc1 | 5484 | dprintf("fd: %d %s orom: %p rv: %d raiddisk: %d\n", |
f2f5c343 LM |
5485 | fd, str, super->orom, rv, raiddisks); |
5486 | #endif | |
5487 | /* no orom/efi or non-intel hba of the disk */ | |
5488 | close(fd); | |
5489 | free_imsm(super); | |
5490 | return 0; | |
5491 | } | |
c2c087e6 | 5492 | close(fd); |
9126b9a8 CA |
5493 | if (super->orom) { |
5494 | if (raiddisks > super->orom->tds) { | |
5495 | if (verbose) | |
7a862a02 | 5496 | pr_err("%d exceeds maximum number of platform supported disks: %d\n", |
9126b9a8 CA |
5497 | raiddisks, super->orom->tds); |
5498 | free_imsm(super); | |
5499 | return 0; | |
5500 | } | |
5501 | if ((super->orom->attr & IMSM_OROM_ATTR_2TB_DISK) == 0 && | |
5502 | (ldsize >> 9) >> 32 > 0) { | |
5503 | if (verbose) | |
e7b84f9d | 5504 | pr_err("%s exceeds maximum platform supported size\n", dev); |
9126b9a8 CA |
5505 | free_imsm(super); |
5506 | return 0; | |
5507 | } | |
f2f5c343 | 5508 | } |
c2c087e6 | 5509 | |
af4348dd | 5510 | *freesize = avail_size_imsm(st, ldsize >> 9, data_offset); |
f2f5c343 | 5511 | free_imsm(super); |
c2c087e6 DW |
5512 | |
5513 | return 1; | |
cdddbdbc DW |
5514 | } |
5515 | ||
0dcecb2e DW |
5516 | static unsigned long long find_size(struct extent *e, int *idx, int num_extents) |
5517 | { | |
5518 | const unsigned long long base_start = e[*idx].start; | |
5519 | unsigned long long end = base_start + e[*idx].size; | |
5520 | int i; | |
5521 | ||
5522 | if (base_start == end) | |
5523 | return 0; | |
5524 | ||
5525 | *idx = *idx + 1; | |
5526 | for (i = *idx; i < num_extents; i++) { | |
5527 | /* extend overlapping extents */ | |
5528 | if (e[i].start >= base_start && | |
5529 | e[i].start <= end) { | |
5530 | if (e[i].size == 0) | |
5531 | return 0; | |
5532 | if (e[i].start + e[i].size > end) | |
5533 | end = e[i].start + e[i].size; | |
5534 | } else if (e[i].start > end) { | |
5535 | *idx = i; | |
5536 | break; | |
5537 | } | |
5538 | } | |
5539 | ||
5540 | return end - base_start; | |
5541 | } | |
5542 | ||
5543 | static unsigned long long merge_extents(struct intel_super *super, int sum_extents) | |
5544 | { | |
5545 | /* build a composite disk with all known extents and generate a new | |
5546 | * 'maxsize' given the "all disks in an array must share a common start | |
5547 | * offset" constraint | |
5548 | */ | |
503975b9 | 5549 | struct extent *e = xcalloc(sum_extents, sizeof(*e)); |
0dcecb2e DW |
5550 | struct dl *dl; |
5551 | int i, j; | |
5552 | int start_extent; | |
5553 | unsigned long long pos; | |
b9d77223 | 5554 | unsigned long long start = 0; |
0dcecb2e DW |
5555 | unsigned long long maxsize; |
5556 | unsigned long reserve; | |
5557 | ||
0dcecb2e DW |
5558 | /* coalesce and sort all extents. also, check to see if we need to |
5559 | * reserve space between member arrays | |
5560 | */ | |
5561 | j = 0; | |
5562 | for (dl = super->disks; dl; dl = dl->next) { | |
5563 | if (!dl->e) | |
5564 | continue; | |
5565 | for (i = 0; i < dl->extent_cnt; i++) | |
5566 | e[j++] = dl->e[i]; | |
5567 | } | |
5568 | qsort(e, sum_extents, sizeof(*e), cmp_extent); | |
5569 | ||
5570 | /* merge extents */ | |
5571 | i = 0; | |
5572 | j = 0; | |
5573 | while (i < sum_extents) { | |
5574 | e[j].start = e[i].start; | |
5575 | e[j].size = find_size(e, &i, sum_extents); | |
5576 | j++; | |
5577 | if (e[j-1].size == 0) | |
5578 | break; | |
5579 | } | |
5580 | ||
5581 | pos = 0; | |
5582 | maxsize = 0; | |
5583 | start_extent = 0; | |
5584 | i = 0; | |
5585 | do { | |
5586 | unsigned long long esize; | |
5587 | ||
5588 | esize = e[i].start - pos; | |
5589 | if (esize >= maxsize) { | |
5590 | maxsize = esize; | |
5591 | start = pos; | |
5592 | start_extent = i; | |
5593 | } | |
5594 | pos = e[i].start + e[i].size; | |
5595 | i++; | |
5596 | } while (e[i-1].size); | |
5597 | free(e); | |
5598 | ||
a7dd165b DW |
5599 | if (maxsize == 0) |
5600 | return 0; | |
5601 | ||
5602 | /* FIXME assumes volume at offset 0 is the first volume in a | |
5603 | * container | |
5604 | */ | |
0dcecb2e DW |
5605 | if (start_extent > 0) |
5606 | reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */ | |
5607 | else | |
5608 | reserve = 0; | |
5609 | ||
5610 | if (maxsize < reserve) | |
a7dd165b | 5611 | return 0; |
0dcecb2e | 5612 | |
5551b113 | 5613 | super->create_offset = ~((unsigned long long) 0); |
0dcecb2e | 5614 | if (start + reserve > super->create_offset) |
a7dd165b | 5615 | return 0; /* start overflows create_offset */ |
0dcecb2e DW |
5616 | super->create_offset = start + reserve; |
5617 | ||
5618 | return maxsize - reserve; | |
5619 | } | |
5620 | ||
88c32bb1 DW |
5621 | static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks) |
5622 | { | |
5623 | if (level < 0 || level == 6 || level == 4) | |
5624 | return 0; | |
5625 | ||
5626 | /* if we have an orom prevent invalid raid levels */ | |
5627 | if (orom) | |
5628 | switch (level) { | |
5629 | case 0: return imsm_orom_has_raid0(orom); | |
5630 | case 1: | |
5631 | if (raiddisks > 2) | |
5632 | return imsm_orom_has_raid1e(orom); | |
1c556e92 DW |
5633 | return imsm_orom_has_raid1(orom) && raiddisks == 2; |
5634 | case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4; | |
5635 | case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2; | |
88c32bb1 DW |
5636 | } |
5637 | else | |
5638 | return 1; /* not on an Intel RAID platform so anything goes */ | |
5639 | ||
5640 | return 0; | |
5641 | } | |
5642 | ||
ca9de185 LM |
5643 | static int |
5644 | active_arrays_by_format(char *name, char* hba, struct md_list **devlist, | |
5645 | int dpa, int verbose) | |
5646 | { | |
5647 | struct mdstat_ent *mdstat = mdstat_read(0, 0); | |
5648 | struct mdstat_ent *memb = NULL; | |
5649 | int count = 0; | |
5650 | int num = 0; | |
5651 | struct md_list *dv = NULL; | |
5652 | int found; | |
5653 | ||
5654 | for (memb = mdstat ; memb ; memb = memb->next) { | |
5655 | if (memb->metadata_version && | |
5656 | (strncmp(memb->metadata_version, "external:", 9) == 0) && | |
5657 | (strcmp(&memb->metadata_version[9], name) == 0) && | |
5658 | !is_subarray(memb->metadata_version+9) && | |
5659 | memb->members) { | |
5660 | struct dev_member *dev = memb->members; | |
5661 | int fd = -1; | |
5662 | while(dev && (fd < 0)) { | |
503975b9 N |
5663 | char *path = xmalloc(strlen(dev->name) + strlen("/dev/") + 1); |
5664 | num = sprintf(path, "%s%s", "/dev/", dev->name); | |
5665 | if (num > 0) | |
5666 | fd = open(path, O_RDONLY, 0); | |
5667 | if ((num <= 0) || (fd < 0)) { | |
5668 | pr_vrb(": Cannot open %s: %s\n", | |
5669 | dev->name, strerror(errno)); | |
ca9de185 | 5670 | } |
503975b9 | 5671 | free(path); |
ca9de185 LM |
5672 | dev = dev->next; |
5673 | } | |
5674 | found = 0; | |
5675 | if ((fd >= 0) && disk_attached_to_hba(fd, hba)) { | |
5676 | struct mdstat_ent *vol; | |
5677 | for (vol = mdstat ; vol ; vol = vol->next) { | |
5678 | if ((vol->active > 0) && | |
5679 | vol->metadata_version && | |
9581efb1 | 5680 | is_container_member(vol, memb->devnm)) { |
ca9de185 LM |
5681 | found++; |
5682 | count++; | |
5683 | } | |
5684 | } | |
5685 | if (*devlist && (found < dpa)) { | |
503975b9 | 5686 | dv = xcalloc(1, sizeof(*dv)); |
9581efb1 N |
5687 | dv->devname = xmalloc(strlen(memb->devnm) + strlen("/dev/") + 1); |
5688 | sprintf(dv->devname, "%s%s", "/dev/", memb->devnm); | |
503975b9 N |
5689 | dv->found = found; |
5690 | dv->used = 0; | |
5691 | dv->next = *devlist; | |
5692 | *devlist = dv; | |
ca9de185 LM |
5693 | } |
5694 | } | |
5695 | if (fd >= 0) | |
5696 | close(fd); | |
5697 | } | |
5698 | } | |
5699 | free_mdstat(mdstat); | |
5700 | return count; | |
5701 | } | |
5702 | ||
5703 | #ifdef DEBUG_LOOP | |
5704 | static struct md_list* | |
5705 | get_loop_devices(void) | |
5706 | { | |
5707 | int i; | |
5708 | struct md_list *devlist = NULL; | |
5709 | struct md_list *dv = NULL; | |
5710 | ||
5711 | for(i = 0; i < 12; i++) { | |
503975b9 N |
5712 | dv = xcalloc(1, sizeof(*dv)); |
5713 | dv->devname = xmalloc(40); | |
ca9de185 LM |
5714 | sprintf(dv->devname, "/dev/loop%d", i); |
5715 | dv->next = devlist; | |
5716 | devlist = dv; | |
5717 | } | |
5718 | return devlist; | |
5719 | } | |
5720 | #endif | |
5721 | ||
5722 | static struct md_list* | |
5723 | get_devices(const char *hba_path) | |
5724 | { | |
5725 | struct md_list *devlist = NULL; | |
5726 | struct md_list *dv = NULL; | |
5727 | struct dirent *ent; | |
5728 | DIR *dir; | |
5729 | int err = 0; | |
5730 | ||
5731 | #if DEBUG_LOOP | |
5732 | devlist = get_loop_devices(); | |
5733 | return devlist; | |
5734 | #endif | |
5735 | /* scroll through /sys/dev/block looking for devices attached to | |
5736 | * this hba | |
5737 | */ | |
5738 | dir = opendir("/sys/dev/block"); | |
5739 | for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) { | |
5740 | int fd; | |
5741 | char buf[1024]; | |
5742 | int major, minor; | |
5743 | char *path = NULL; | |
5744 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
5745 | continue; | |
5746 | path = devt_to_devpath(makedev(major, minor)); | |
5747 | if (!path) | |
5748 | continue; | |
5749 | if (!path_attached_to_hba(path, hba_path)) { | |
5750 | free(path); | |
5751 | path = NULL; | |
5752 | continue; | |
5753 | } | |
5754 | free(path); | |
5755 | path = NULL; | |
5756 | fd = dev_open(ent->d_name, O_RDONLY); | |
5757 | if (fd >= 0) { | |
5758 | fd2devname(fd, buf); | |
5759 | close(fd); | |
5760 | } else { | |
e7b84f9d | 5761 | pr_err("cannot open device: %s\n", |
ca9de185 LM |
5762 | ent->d_name); |
5763 | continue; | |
5764 | } | |
5765 | ||
503975b9 N |
5766 | dv = xcalloc(1, sizeof(*dv)); |
5767 | dv->devname = xstrdup(buf); | |
ca9de185 LM |
5768 | dv->next = devlist; |
5769 | devlist = dv; | |
5770 | } | |
5771 | if (err) { | |
5772 | while(devlist) { | |
5773 | dv = devlist; | |
5774 | devlist = devlist->next; | |
5775 | free(dv->devname); | |
5776 | free(dv); | |
5777 | } | |
5778 | } | |
562aa102 | 5779 | closedir(dir); |
ca9de185 LM |
5780 | return devlist; |
5781 | } | |
5782 | ||
5783 | static int | |
5784 | count_volumes_list(struct md_list *devlist, char *homehost, | |
5785 | int verbose, int *found) | |
5786 | { | |
5787 | struct md_list *tmpdev; | |
5788 | int count = 0; | |
5789 | struct supertype *st = NULL; | |
5790 | ||
5791 | /* first walk the list of devices to find a consistent set | |
5792 | * that match the criterea, if that is possible. | |
5793 | * We flag the ones we like with 'used'. | |
5794 | */ | |
5795 | *found = 0; | |
5796 | st = match_metadata_desc_imsm("imsm"); | |
5797 | if (st == NULL) { | |
5798 | pr_vrb(": cannot allocate memory for imsm supertype\n"); | |
5799 | return 0; | |
5800 | } | |
5801 | ||
5802 | for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { | |
5803 | char *devname = tmpdev->devname; | |
5804 | struct stat stb; | |
5805 | struct supertype *tst; | |
5806 | int dfd; | |
5807 | if (tmpdev->used > 1) | |
5808 | continue; | |
5809 | tst = dup_super(st); | |
5810 | if (tst == NULL) { | |
5811 | pr_vrb(": cannot allocate memory for imsm supertype\n"); | |
5812 | goto err_1; | |
5813 | } | |
5814 | tmpdev->container = 0; | |
5815 | dfd = dev_open(devname, O_RDONLY|O_EXCL); | |
5816 | if (dfd < 0) { | |
1ade5cc1 | 5817 | dprintf("cannot open device %s: %s\n", |
ca9de185 LM |
5818 | devname, strerror(errno)); |
5819 | tmpdev->used = 2; | |
5820 | } else if (fstat(dfd, &stb)< 0) { | |
5821 | /* Impossible! */ | |
1ade5cc1 | 5822 | dprintf("fstat failed for %s: %s\n", |
ca9de185 LM |
5823 | devname, strerror(errno)); |
5824 | tmpdev->used = 2; | |
5825 | } else if ((stb.st_mode & S_IFMT) != S_IFBLK) { | |
1ade5cc1 | 5826 | dprintf("%s is not a block device.\n", |
ca9de185 LM |
5827 | devname); |
5828 | tmpdev->used = 2; | |
5829 | } else if (must_be_container(dfd)) { | |
5830 | struct supertype *cst; | |
5831 | cst = super_by_fd(dfd, NULL); | |
5832 | if (cst == NULL) { | |
1ade5cc1 | 5833 | dprintf("cannot recognize container type %s\n", |
ca9de185 LM |
5834 | devname); |
5835 | tmpdev->used = 2; | |
5836 | } else if (tst->ss != st->ss) { | |
1ade5cc1 | 5837 | dprintf("non-imsm container - ignore it: %s\n", |
ca9de185 LM |
5838 | devname); |
5839 | tmpdev->used = 2; | |
5840 | } else if (!tst->ss->load_container || | |
5841 | tst->ss->load_container(tst, dfd, NULL)) | |
5842 | tmpdev->used = 2; | |
5843 | else { | |
5844 | tmpdev->container = 1; | |
5845 | } | |
5846 | if (cst) | |
5847 | cst->ss->free_super(cst); | |
5848 | } else { | |
5849 | tmpdev->st_rdev = stb.st_rdev; | |
5850 | if (tst->ss->load_super(tst,dfd, NULL)) { | |
1ade5cc1 | 5851 | dprintf("no RAID superblock on %s\n", |
ca9de185 LM |
5852 | devname); |
5853 | tmpdev->used = 2; | |
5854 | } else if (tst->ss->compare_super == NULL) { | |
1ade5cc1 | 5855 | dprintf("Cannot assemble %s metadata on %s\n", |
ca9de185 LM |
5856 | tst->ss->name, devname); |
5857 | tmpdev->used = 2; | |
5858 | } | |
5859 | } | |
5860 | if (dfd >= 0) | |
5861 | close(dfd); | |
5862 | if (tmpdev->used == 2 || tmpdev->used == 4) { | |
5863 | /* Ignore unrecognised devices during auto-assembly */ | |
5864 | goto loop; | |
5865 | } | |
5866 | else { | |
5867 | struct mdinfo info; | |
5868 | tst->ss->getinfo_super(tst, &info, NULL); | |
5869 | ||
5870 | if (st->minor_version == -1) | |
5871 | st->minor_version = tst->minor_version; | |
5872 | ||
5873 | if (memcmp(info.uuid, uuid_zero, | |
5874 | sizeof(int[4])) == 0) { | |
5875 | /* this is a floating spare. It cannot define | |
5876 | * an array unless there are no more arrays of | |
5877 | * this type to be found. It can be included | |
5878 | * in an array of this type though. | |
5879 | */ | |
5880 | tmpdev->used = 3; | |
5881 | goto loop; | |
5882 | } | |
5883 | ||
5884 | if (st->ss != tst->ss || | |
5885 | st->minor_version != tst->minor_version || | |
5886 | st->ss->compare_super(st, tst) != 0) { | |
5887 | /* Some mismatch. If exactly one array matches this host, | |
5888 | * we can resolve on that one. | |
5889 | * Or, if we are auto assembling, we just ignore the second | |
5890 | * for now. | |
5891 | */ | |
1ade5cc1 | 5892 | dprintf("superblock on %s doesn't match others - assembly aborted\n", |
ca9de185 LM |
5893 | devname); |
5894 | goto loop; | |
5895 | } | |
5896 | tmpdev->used = 1; | |
5897 | *found = 1; | |
5898 | dprintf("found: devname: %s\n", devname); | |
5899 | } | |
5900 | loop: | |
5901 | if (tst) | |
5902 | tst->ss->free_super(tst); | |
5903 | } | |
5904 | if (*found != 0) { | |
5905 | int err; | |
5906 | if ((err = load_super_imsm_all(st, -1, &st->sb, NULL, devlist, 0)) == 0) { | |
5907 | struct mdinfo *iter, *head = st->ss->container_content(st, NULL); | |
5908 | for (iter = head; iter; iter = iter->next) { | |
5909 | dprintf("content->text_version: %s vol\n", | |
5910 | iter->text_version); | |
5911 | if (iter->array.state & (1<<MD_SB_BLOCK_VOLUME)) { | |
5912 | /* do not assemble arrays with unsupported | |
5913 | configurations */ | |
1ade5cc1 | 5914 | dprintf("Cannot activate member %s.\n", |
ca9de185 LM |
5915 | iter->text_version); |
5916 | } else | |
5917 | count++; | |
5918 | } | |
5919 | sysfs_free(head); | |
5920 | ||
5921 | } else { | |
1ade5cc1 | 5922 | dprintf("No valid super block on device list: err: %d %p\n", |
ca9de185 LM |
5923 | err, st->sb); |
5924 | } | |
5925 | } else { | |
1ade5cc1 | 5926 | dprintf("no more devices to examine\n"); |
ca9de185 LM |
5927 | } |
5928 | ||
5929 | for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { | |
5930 | if ((tmpdev->used == 1) && (tmpdev->found)) { | |
5931 | if (count) { | |
5932 | if (count < tmpdev->found) | |
5933 | count = 0; | |
5934 | else | |
5935 | count -= tmpdev->found; | |
5936 | } | |
5937 | } | |
5938 | if (tmpdev->used == 1) | |
5939 | tmpdev->used = 4; | |
5940 | } | |
5941 | err_1: | |
5942 | if (st) | |
5943 | st->ss->free_super(st); | |
5944 | return count; | |
5945 | } | |
5946 | ||
ca9de185 | 5947 | static int |
72a45777 | 5948 | count_volumes(struct intel_hba *hba, int dpa, int verbose) |
ca9de185 | 5949 | { |
72a45777 | 5950 | struct sys_dev *idev, *intel_devices = find_intel_devices(); |
ca9de185 | 5951 | int count = 0; |
72a45777 PB |
5952 | const struct orom_entry *entry; |
5953 | struct devid_list *dv, *devid_list; | |
ca9de185 | 5954 | |
72a45777 | 5955 | if (!hba || !hba->path) |
ca9de185 LM |
5956 | return 0; |
5957 | ||
72a45777 PB |
5958 | for (idev = intel_devices; idev; idev = idev->next) { |
5959 | if (strstr(idev->path, hba->path)) | |
5960 | break; | |
5961 | } | |
5962 | ||
5963 | if (!idev || !idev->dev_id) | |
ca9de185 | 5964 | return 0; |
72a45777 PB |
5965 | |
5966 | entry = get_orom_entry_by_device_id(idev->dev_id); | |
5967 | ||
5968 | if (!entry || !entry->devid_list) | |
5969 | return 0; | |
5970 | ||
5971 | devid_list = entry->devid_list; | |
5972 | for (dv = devid_list; dv; dv = dv->next) { | |
72a45777 PB |
5973 | struct md_list *devlist = NULL; |
5974 | struct sys_dev *device = device_by_id(dv->devid); | |
5975 | char *hba_path; | |
5976 | int found = 0; | |
5977 | ||
5978 | if (device) | |
5979 | hba_path = device->path; | |
5980 | else | |
5981 | return 0; | |
5982 | ||
60f0f54d PB |
5983 | /* VMD has one orom entry for all domain, but spanning is not allowed. |
5984 | * VMD arrays should be counted per domain (controller), so skip | |
5985 | * domains that are not the given one. | |
5986 | */ | |
5987 | if ((hba->type == SYS_DEV_VMD) && | |
5988 | (strncmp(device->path, hba->path, strlen(device->path)) != 0)) | |
5989 | continue; | |
5990 | ||
72a45777 PB |
5991 | devlist = get_devices(hba_path); |
5992 | /* if no intel devices return zero volumes */ | |
5993 | if (devlist == NULL) | |
5994 | return 0; | |
5995 | ||
5996 | count += active_arrays_by_format("imsm", hba_path, &devlist, dpa, verbose); | |
5997 | dprintf("path: %s active arrays: %d\n", hba_path, count); | |
5998 | if (devlist == NULL) | |
5999 | return 0; | |
6000 | do { | |
6001 | found = 0; | |
6002 | count += count_volumes_list(devlist, | |
6003 | NULL, | |
6004 | verbose, | |
6005 | &found); | |
6006 | dprintf("found %d count: %d\n", found, count); | |
6007 | } while (found); | |
6008 | ||
6009 | dprintf("path: %s total number of volumes: %d\n", hba_path, count); | |
6010 | ||
6011 | while (devlist) { | |
6012 | struct md_list *dv = devlist; | |
6013 | devlist = devlist->next; | |
6014 | free(dv->devname); | |
6015 | free(dv); | |
6016 | } | |
ca9de185 LM |
6017 | } |
6018 | return count; | |
6019 | } | |
6020 | ||
cd9d1ac7 DW |
6021 | static int imsm_default_chunk(const struct imsm_orom *orom) |
6022 | { | |
6023 | /* up to 512 if the plaform supports it, otherwise the platform max. | |
6024 | * 128 if no platform detected | |
6025 | */ | |
6026 | int fs = max(7, orom ? fls(orom->sss) : 0); | |
6027 | ||
6028 | return min(512, (1 << fs)); | |
6029 | } | |
73408129 | 6030 | |
6592ce37 DW |
6031 | static int |
6032 | validate_geometry_imsm_orom(struct intel_super *super, int level, int layout, | |
2cc699af | 6033 | int raiddisks, int *chunk, unsigned long long size, int verbose) |
6592ce37 | 6034 | { |
660260d0 DW |
6035 | /* check/set platform and metadata limits/defaults */ |
6036 | if (super->orom && raiddisks > super->orom->dpa) { | |
6037 | pr_vrb(": platform supports a maximum of %d disks per array\n", | |
6038 | super->orom->dpa); | |
73408129 LM |
6039 | return 0; |
6040 | } | |
6041 | ||
5d500228 | 6042 | /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */ |
660260d0 | 6043 | if (!is_raid_level_supported(super->orom, level, raiddisks)) { |
6592ce37 DW |
6044 | pr_vrb(": platform does not support raid%d with %d disk%s\n", |
6045 | level, raiddisks, raiddisks > 1 ? "s" : ""); | |
6046 | return 0; | |
6047 | } | |
cd9d1ac7 | 6048 | |
7ccc4cc4 | 6049 | if (*chunk == 0 || *chunk == UnSet) |
cd9d1ac7 DW |
6050 | *chunk = imsm_default_chunk(super->orom); |
6051 | ||
7ccc4cc4 | 6052 | if (super->orom && !imsm_orom_has_chunk(super->orom, *chunk)) { |
7a862a02 | 6053 | pr_vrb(": platform does not support a chunk size of: %d\n", *chunk); |
cd9d1ac7 | 6054 | return 0; |
6592ce37 | 6055 | } |
cd9d1ac7 | 6056 | |
6592ce37 DW |
6057 | if (layout != imsm_level_to_layout(level)) { |
6058 | if (level == 5) | |
6059 | pr_vrb(": imsm raid 5 only supports the left-asymmetric layout\n"); | |
6060 | else if (level == 10) | |
6061 | pr_vrb(": imsm raid 10 only supports the n2 layout\n"); | |
6062 | else | |
6063 | pr_vrb(": imsm unknown layout %#x for this raid level %d\n", | |
6064 | layout, level); | |
6065 | return 0; | |
6066 | } | |
2cc699af | 6067 | |
7ccc4cc4 | 6068 | if (super->orom && (super->orom->attr & IMSM_OROM_ATTR_2TB) == 0 && |
2cc699af CA |
6069 | (calc_array_size(level, raiddisks, layout, *chunk, size) >> 32) > 0) { |
6070 | pr_vrb(": platform does not support a volume size over 2TB\n"); | |
6071 | return 0; | |
6072 | } | |
614902f6 | 6073 | |
6592ce37 DW |
6074 | return 1; |
6075 | } | |
6076 | ||
1011e834 | 6077 | /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd |
c2c087e6 DW |
6078 | * FIX ME add ahci details |
6079 | */ | |
8b353278 | 6080 | static int validate_geometry_imsm_volume(struct supertype *st, int level, |
c21e737b | 6081 | int layout, int raiddisks, int *chunk, |
af4348dd N |
6082 | unsigned long long size, |
6083 | unsigned long long data_offset, | |
6084 | char *dev, | |
2c514b71 NB |
6085 | unsigned long long *freesize, |
6086 | int verbose) | |
cdddbdbc | 6087 | { |
c2c087e6 DW |
6088 | struct stat stb; |
6089 | struct intel_super *super = st->sb; | |
b2916f25 | 6090 | struct imsm_super *mpb; |
c2c087e6 DW |
6091 | struct dl *dl; |
6092 | unsigned long long pos = 0; | |
6093 | unsigned long long maxsize; | |
6094 | struct extent *e; | |
6095 | int i; | |
cdddbdbc | 6096 | |
88c32bb1 DW |
6097 | /* We must have the container info already read in. */ |
6098 | if (!super) | |
c2c087e6 DW |
6099 | return 0; |
6100 | ||
b2916f25 JS |
6101 | mpb = super->anchor; |
6102 | ||
2cc699af | 6103 | if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, size, verbose)) { |
7a862a02 | 6104 | pr_err("RAID gemetry validation failed. Cannot proceed with the action(s).\n"); |
c2c087e6 | 6105 | return 0; |
d54559f0 | 6106 | } |
c2c087e6 DW |
6107 | if (!dev) { |
6108 | /* General test: make sure there is space for | |
2da8544a DW |
6109 | * 'raiddisks' device extents of size 'size' at a given |
6110 | * offset | |
c2c087e6 | 6111 | */ |
e46273eb | 6112 | unsigned long long minsize = size; |
b7528a20 | 6113 | unsigned long long start_offset = MaxSector; |
c2c087e6 DW |
6114 | int dcnt = 0; |
6115 | if (minsize == 0) | |
6116 | minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
6117 | for (dl = super->disks; dl ; dl = dl->next) { | |
6118 | int found = 0; | |
6119 | ||
bf5a934a | 6120 | pos = 0; |
c2c087e6 DW |
6121 | i = 0; |
6122 | e = get_extents(super, dl); | |
6123 | if (!e) continue; | |
6124 | do { | |
6125 | unsigned long long esize; | |
6126 | esize = e[i].start - pos; | |
6127 | if (esize >= minsize) | |
6128 | found = 1; | |
b7528a20 | 6129 | if (found && start_offset == MaxSector) { |
2da8544a DW |
6130 | start_offset = pos; |
6131 | break; | |
6132 | } else if (found && pos != start_offset) { | |
6133 | found = 0; | |
6134 | break; | |
6135 | } | |
c2c087e6 DW |
6136 | pos = e[i].start + e[i].size; |
6137 | i++; | |
6138 | } while (e[i-1].size); | |
6139 | if (found) | |
6140 | dcnt++; | |
6141 | free(e); | |
6142 | } | |
6143 | if (dcnt < raiddisks) { | |
2c514b71 | 6144 | if (verbose) |
7a862a02 | 6145 | pr_err("imsm: Not enough devices with space for this array (%d < %d)\n", |
2c514b71 | 6146 | dcnt, raiddisks); |
c2c087e6 DW |
6147 | return 0; |
6148 | } | |
6149 | return 1; | |
6150 | } | |
0dcecb2e | 6151 | |
c2c087e6 DW |
6152 | /* This device must be a member of the set */ |
6153 | if (stat(dev, &stb) < 0) | |
6154 | return 0; | |
6155 | if ((S_IFMT & stb.st_mode) != S_IFBLK) | |
6156 | return 0; | |
6157 | for (dl = super->disks ; dl ; dl = dl->next) { | |
f21e18ca N |
6158 | if (dl->major == (int)major(stb.st_rdev) && |
6159 | dl->minor == (int)minor(stb.st_rdev)) | |
c2c087e6 DW |
6160 | break; |
6161 | } | |
6162 | if (!dl) { | |
2c514b71 | 6163 | if (verbose) |
7a862a02 | 6164 | pr_err("%s is not in the same imsm set\n", dev); |
c2c087e6 | 6165 | return 0; |
a20d2ba5 DW |
6166 | } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) { |
6167 | /* If a volume is present then the current creation attempt | |
6168 | * cannot incorporate new spares because the orom may not | |
6169 | * understand this configuration (all member disks must be | |
6170 | * members of each array in the container). | |
6171 | */ | |
7a862a02 N |
6172 | pr_err("%s is a spare and a volume is already defined for this container\n", dev); |
6173 | pr_err("The option-rom requires all member disks to be a member of all volumes\n"); | |
a20d2ba5 | 6174 | return 0; |
5fe62b94 WD |
6175 | } else if (super->orom && mpb->num_raid_devs > 0 && |
6176 | mpb->num_disks != raiddisks) { | |
7a862a02 | 6177 | pr_err("The option-rom requires all member disks to be a member of all volumes\n"); |
5fe62b94 | 6178 | return 0; |
c2c087e6 | 6179 | } |
0dcecb2e DW |
6180 | |
6181 | /* retrieve the largest free space block */ | |
c2c087e6 DW |
6182 | e = get_extents(super, dl); |
6183 | maxsize = 0; | |
6184 | i = 0; | |
0dcecb2e DW |
6185 | if (e) { |
6186 | do { | |
6187 | unsigned long long esize; | |
6188 | ||
6189 | esize = e[i].start - pos; | |
6190 | if (esize >= maxsize) | |
6191 | maxsize = esize; | |
6192 | pos = e[i].start + e[i].size; | |
6193 | i++; | |
6194 | } while (e[i-1].size); | |
6195 | dl->e = e; | |
6196 | dl->extent_cnt = i; | |
6197 | } else { | |
6198 | if (verbose) | |
e7b84f9d | 6199 | pr_err("unable to determine free space for: %s\n", |
0dcecb2e DW |
6200 | dev); |
6201 | return 0; | |
6202 | } | |
6203 | if (maxsize < size) { | |
6204 | if (verbose) | |
e7b84f9d | 6205 | pr_err("%s not enough space (%llu < %llu)\n", |
0dcecb2e DW |
6206 | dev, maxsize, size); |
6207 | return 0; | |
6208 | } | |
6209 | ||
6210 | /* count total number of extents for merge */ | |
6211 | i = 0; | |
6212 | for (dl = super->disks; dl; dl = dl->next) | |
6213 | if (dl->e) | |
6214 | i += dl->extent_cnt; | |
6215 | ||
6216 | maxsize = merge_extents(super, i); | |
3baa56ab LO |
6217 | |
6218 | if (!check_env("IMSM_NO_PLATFORM") && | |
6219 | mpb->num_raid_devs > 0 && size && size != maxsize) { | |
7a862a02 | 6220 | pr_err("attempting to create a second volume with size less then remaining space. Aborting...\n"); |
3baa56ab LO |
6221 | return 0; |
6222 | } | |
6223 | ||
a7dd165b | 6224 | if (maxsize < size || maxsize == 0) { |
b3071342 LD |
6225 | if (verbose) { |
6226 | if (maxsize == 0) | |
7a862a02 | 6227 | pr_err("no free space left on device. Aborting...\n"); |
b3071342 | 6228 | else |
7a862a02 | 6229 | pr_err("not enough space to create volume of given size (%llu < %llu). Aborting...\n", |
b3071342 LD |
6230 | maxsize, size); |
6231 | } | |
0dcecb2e | 6232 | return 0; |
0dcecb2e DW |
6233 | } |
6234 | ||
c2c087e6 DW |
6235 | *freesize = maxsize; |
6236 | ||
ca9de185 | 6237 | if (super->orom) { |
72a45777 | 6238 | int count = count_volumes(super->hba, |
ca9de185 LM |
6239 | super->orom->dpa, verbose); |
6240 | if (super->orom->vphba <= count) { | |
40110b90 | 6241 | pr_vrb(": platform does not support more than %d raid volumes.\n", |
ca9de185 LM |
6242 | super->orom->vphba); |
6243 | return 0; | |
6244 | } | |
6245 | } | |
c2c087e6 | 6246 | return 1; |
cdddbdbc DW |
6247 | } |
6248 | ||
13bcac90 | 6249 | static int imsm_get_free_size(struct supertype *st, int raiddisks, |
efb30e7f DW |
6250 | unsigned long long size, int chunk, |
6251 | unsigned long long *freesize) | |
6252 | { | |
6253 | struct intel_super *super = st->sb; | |
6254 | struct imsm_super *mpb = super->anchor; | |
6255 | struct dl *dl; | |
6256 | int i; | |
6257 | int extent_cnt; | |
6258 | struct extent *e; | |
6259 | unsigned long long maxsize; | |
6260 | unsigned long long minsize; | |
6261 | int cnt; | |
6262 | int used; | |
6263 | ||
6264 | /* find the largest common start free region of the possible disks */ | |
6265 | used = 0; | |
6266 | extent_cnt = 0; | |
6267 | cnt = 0; | |
6268 | for (dl = super->disks; dl; dl = dl->next) { | |
6269 | dl->raiddisk = -1; | |
6270 | ||
6271 | if (dl->index >= 0) | |
6272 | used++; | |
6273 | ||
6274 | /* don't activate new spares if we are orom constrained | |
6275 | * and there is already a volume active in the container | |
6276 | */ | |
6277 | if (super->orom && dl->index < 0 && mpb->num_raid_devs) | |
6278 | continue; | |
6279 | ||
6280 | e = get_extents(super, dl); | |
6281 | if (!e) | |
6282 | continue; | |
6283 | for (i = 1; e[i-1].size; i++) | |
6284 | ; | |
6285 | dl->e = e; | |
6286 | dl->extent_cnt = i; | |
6287 | extent_cnt += i; | |
6288 | cnt++; | |
6289 | } | |
6290 | ||
6291 | maxsize = merge_extents(super, extent_cnt); | |
6292 | minsize = size; | |
6293 | if (size == 0) | |
612e59d8 CA |
6294 | /* chunk is in K */ |
6295 | minsize = chunk * 2; | |
efb30e7f DW |
6296 | |
6297 | if (cnt < raiddisks || | |
6298 | (super->orom && used && used != raiddisks) || | |
a7dd165b DW |
6299 | maxsize < minsize || |
6300 | maxsize == 0) { | |
e7b84f9d | 6301 | pr_err("not enough devices with space to create array.\n"); |
efb30e7f DW |
6302 | return 0; /* No enough free spaces large enough */ |
6303 | } | |
6304 | ||
6305 | if (size == 0) { | |
6306 | size = maxsize; | |
6307 | if (chunk) { | |
612e59d8 CA |
6308 | size /= 2 * chunk; |
6309 | size *= 2 * chunk; | |
efb30e7f | 6310 | } |
f878b242 LM |
6311 | maxsize = size; |
6312 | } | |
6313 | if (!check_env("IMSM_NO_PLATFORM") && | |
6314 | mpb->num_raid_devs > 0 && size && size != maxsize) { | |
7a862a02 | 6315 | pr_err("attempting to create a second volume with size less then remaining space. Aborting...\n"); |
f878b242 | 6316 | return 0; |
efb30e7f | 6317 | } |
efb30e7f DW |
6318 | cnt = 0; |
6319 | for (dl = super->disks; dl; dl = dl->next) | |
6320 | if (dl->e) | |
6321 | dl->raiddisk = cnt++; | |
6322 | ||
6323 | *freesize = size; | |
6324 | ||
13bcac90 AK |
6325 | dprintf("imsm: imsm_get_free_size() returns : %llu\n", size); |
6326 | ||
efb30e7f DW |
6327 | return 1; |
6328 | } | |
6329 | ||
13bcac90 AK |
6330 | static int reserve_space(struct supertype *st, int raiddisks, |
6331 | unsigned long long size, int chunk, | |
6332 | unsigned long long *freesize) | |
6333 | { | |
6334 | struct intel_super *super = st->sb; | |
6335 | struct dl *dl; | |
6336 | int cnt; | |
6337 | int rv = 0; | |
6338 | ||
6339 | rv = imsm_get_free_size(st, raiddisks, size, chunk, freesize); | |
6340 | if (rv) { | |
6341 | cnt = 0; | |
6342 | for (dl = super->disks; dl; dl = dl->next) | |
6343 | if (dl->e) | |
6344 | dl->raiddisk = cnt++; | |
6345 | rv = 1; | |
6346 | } | |
6347 | ||
6348 | return rv; | |
6349 | } | |
6350 | ||
bf5a934a | 6351 | static int validate_geometry_imsm(struct supertype *st, int level, int layout, |
c21e737b | 6352 | int raiddisks, int *chunk, unsigned long long size, |
af4348dd | 6353 | unsigned long long data_offset, |
bf5a934a DW |
6354 | char *dev, unsigned long long *freesize, |
6355 | int verbose) | |
6356 | { | |
6357 | int fd, cfd; | |
6358 | struct mdinfo *sra; | |
20cbe8d2 | 6359 | int is_member = 0; |
bf5a934a | 6360 | |
d54559f0 LM |
6361 | /* load capability |
6362 | * if given unused devices create a container | |
bf5a934a DW |
6363 | * if given given devices in a container create a member volume |
6364 | */ | |
6365 | if (level == LEVEL_CONTAINER) { | |
6366 | /* Must be a fresh device to add to a container */ | |
6367 | return validate_geometry_imsm_container(st, level, layout, | |
c21e737b | 6368 | raiddisks, |
7ccc4cc4 | 6369 | *chunk, |
af4348dd | 6370 | size, data_offset, |
bf5a934a DW |
6371 | dev, freesize, |
6372 | verbose); | |
6373 | } | |
9587c373 | 6374 | |
8592f29d | 6375 | if (!dev) { |
e91a3bad | 6376 | if (st->sb) { |
ca9de185 | 6377 | struct intel_super *super = st->sb; |
e91a3bad | 6378 | if (!validate_geometry_imsm_orom(st->sb, level, layout, |
2cc699af | 6379 | raiddisks, chunk, size, |
e91a3bad LM |
6380 | verbose)) |
6381 | return 0; | |
efb30e7f DW |
6382 | /* we are being asked to automatically layout a |
6383 | * new volume based on the current contents of | |
6384 | * the container. If the the parameters can be | |
6385 | * satisfied reserve_space will record the disks, | |
6386 | * start offset, and size of the volume to be | |
6387 | * created. add_to_super and getinfo_super | |
6388 | * detect when autolayout is in progress. | |
6389 | */ | |
ca9de185 LM |
6390 | /* assuming that freesize is always given when array is |
6391 | created */ | |
6392 | if (super->orom && freesize) { | |
6393 | int count; | |
72a45777 | 6394 | count = count_volumes(super->hba, |
ca9de185 LM |
6395 | super->orom->dpa, verbose); |
6396 | if (super->orom->vphba <= count) { | |
7a862a02 | 6397 | pr_vrb(": platform does not support more than %d raid volumes.\n", |
ca9de185 LM |
6398 | super->orom->vphba); |
6399 | return 0; | |
6400 | } | |
6401 | } | |
e91a3bad LM |
6402 | if (freesize) |
6403 | return reserve_space(st, raiddisks, size, | |
7ccc4cc4 | 6404 | *chunk, freesize); |
8592f29d N |
6405 | } |
6406 | return 1; | |
6407 | } | |
bf5a934a DW |
6408 | if (st->sb) { |
6409 | /* creating in a given container */ | |
6410 | return validate_geometry_imsm_volume(st, level, layout, | |
6411 | raiddisks, chunk, size, | |
af4348dd | 6412 | data_offset, |
bf5a934a DW |
6413 | dev, freesize, verbose); |
6414 | } | |
6415 | ||
bf5a934a DW |
6416 | /* This device needs to be a device in an 'imsm' container */ |
6417 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
6418 | if (fd >= 0) { | |
6419 | if (verbose) | |
e7b84f9d N |
6420 | pr_err("Cannot create this array on device %s\n", |
6421 | dev); | |
bf5a934a DW |
6422 | close(fd); |
6423 | return 0; | |
6424 | } | |
6425 | if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) { | |
6426 | if (verbose) | |
e7b84f9d | 6427 | pr_err("Cannot open %s: %s\n", |
bf5a934a DW |
6428 | dev, strerror(errno)); |
6429 | return 0; | |
6430 | } | |
6431 | /* Well, it is in use by someone, maybe an 'imsm' container. */ | |
6432 | cfd = open_container(fd); | |
20cbe8d2 | 6433 | close(fd); |
bf5a934a | 6434 | if (cfd < 0) { |
bf5a934a | 6435 | if (verbose) |
e7b84f9d | 6436 | pr_err("Cannot use %s: It is busy\n", |
bf5a934a DW |
6437 | dev); |
6438 | return 0; | |
6439 | } | |
4dd2df09 | 6440 | sra = sysfs_read(cfd, NULL, GET_VERSION); |
bf5a934a | 6441 | if (sra && sra->array.major_version == -1 && |
20cbe8d2 AW |
6442 | strcmp(sra->text_version, "imsm") == 0) |
6443 | is_member = 1; | |
6444 | sysfs_free(sra); | |
6445 | if (is_member) { | |
bf5a934a DW |
6446 | /* This is a member of a imsm container. Load the container |
6447 | * and try to create a volume | |
6448 | */ | |
6449 | struct intel_super *super; | |
6450 | ||
ec50f7b6 | 6451 | if (load_super_imsm_all(st, cfd, (void **) &super, NULL, NULL, 1) == 0) { |
bf5a934a | 6452 | st->sb = super; |
4dd2df09 | 6453 | strcpy(st->container_devnm, fd2devnm(cfd)); |
bf5a934a DW |
6454 | close(cfd); |
6455 | return validate_geometry_imsm_volume(st, level, layout, | |
6456 | raiddisks, chunk, | |
af4348dd | 6457 | size, data_offset, dev, |
ecbd9e81 N |
6458 | freesize, 1) |
6459 | ? 1 : -1; | |
bf5a934a | 6460 | } |
20cbe8d2 | 6461 | } |
bf5a934a | 6462 | |
20cbe8d2 | 6463 | if (verbose) |
e7b84f9d | 6464 | pr_err("failed container membership check\n"); |
20cbe8d2 AW |
6465 | |
6466 | close(cfd); | |
6467 | return 0; | |
bf5a934a | 6468 | } |
0bd16cf2 | 6469 | |
30f58b22 | 6470 | static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk) |
0bd16cf2 DJ |
6471 | { |
6472 | struct intel_super *super = st->sb; | |
6473 | ||
30f58b22 DW |
6474 | if (level && *level == UnSet) |
6475 | *level = LEVEL_CONTAINER; | |
6476 | ||
6477 | if (level && layout && *layout == UnSet) | |
6478 | *layout = imsm_level_to_layout(*level); | |
0bd16cf2 | 6479 | |
cd9d1ac7 DW |
6480 | if (chunk && (*chunk == UnSet || *chunk == 0)) |
6481 | *chunk = imsm_default_chunk(super->orom); | |
0bd16cf2 DJ |
6482 | } |
6483 | ||
33414a01 DW |
6484 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev); |
6485 | ||
6486 | static int kill_subarray_imsm(struct supertype *st) | |
6487 | { | |
6488 | /* remove the subarray currently referenced by ->current_vol */ | |
6489 | __u8 i; | |
6490 | struct intel_dev **dp; | |
6491 | struct intel_super *super = st->sb; | |
6492 | __u8 current_vol = super->current_vol; | |
6493 | struct imsm_super *mpb = super->anchor; | |
6494 | ||
6495 | if (super->current_vol < 0) | |
6496 | return 2; | |
6497 | super->current_vol = -1; /* invalidate subarray cursor */ | |
6498 | ||
6499 | /* block deletions that would change the uuid of active subarrays | |
6500 | * | |
6501 | * FIXME when immutable ids are available, but note that we'll | |
6502 | * also need to fixup the invalidated/active subarray indexes in | |
6503 | * mdstat | |
6504 | */ | |
6505 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
6506 | char subarray[4]; | |
6507 | ||
6508 | if (i < current_vol) | |
6509 | continue; | |
6510 | sprintf(subarray, "%u", i); | |
4dd2df09 | 6511 | if (is_subarray_active(subarray, st->devnm)) { |
e7b84f9d N |
6512 | pr_err("deleting subarray-%d would change the UUID of active subarray-%d, aborting\n", |
6513 | current_vol, i); | |
33414a01 DW |
6514 | |
6515 | return 2; | |
6516 | } | |
6517 | } | |
6518 | ||
6519 | if (st->update_tail) { | |
503975b9 | 6520 | struct imsm_update_kill_array *u = xmalloc(sizeof(*u)); |
33414a01 | 6521 | |
33414a01 DW |
6522 | u->type = update_kill_array; |
6523 | u->dev_idx = current_vol; | |
6524 | append_metadata_update(st, u, sizeof(*u)); | |
6525 | ||
6526 | return 0; | |
6527 | } | |
6528 | ||
6529 | for (dp = &super->devlist; *dp;) | |
6530 | if ((*dp)->index == current_vol) { | |
6531 | *dp = (*dp)->next; | |
6532 | } else { | |
6533 | handle_missing(super, (*dp)->dev); | |
6534 | if ((*dp)->index > current_vol) | |
6535 | (*dp)->index--; | |
6536 | dp = &(*dp)->next; | |
6537 | } | |
6538 | ||
6539 | /* no more raid devices, all active components are now spares, | |
6540 | * but of course failed are still failed | |
6541 | */ | |
6542 | if (--mpb->num_raid_devs == 0) { | |
6543 | struct dl *d; | |
6544 | ||
6545 | for (d = super->disks; d; d = d->next) | |
a8619d23 AK |
6546 | if (d->index > -2) |
6547 | mark_spare(d); | |
33414a01 DW |
6548 | } |
6549 | ||
6550 | super->updates_pending++; | |
6551 | ||
6552 | return 0; | |
6553 | } | |
aa534678 | 6554 | |
a951a4f7 | 6555 | static int update_subarray_imsm(struct supertype *st, char *subarray, |
fa56eddb | 6556 | char *update, struct mddev_ident *ident) |
aa534678 DW |
6557 | { |
6558 | /* update the subarray currently referenced by ->current_vol */ | |
6559 | struct intel_super *super = st->sb; | |
6560 | struct imsm_super *mpb = super->anchor; | |
6561 | ||
aa534678 DW |
6562 | if (strcmp(update, "name") == 0) { |
6563 | char *name = ident->name; | |
a951a4f7 N |
6564 | char *ep; |
6565 | int vol; | |
aa534678 | 6566 | |
4dd2df09 | 6567 | if (is_subarray_active(subarray, st->devnm)) { |
e7b84f9d | 6568 | pr_err("Unable to update name of active subarray\n"); |
aa534678 DW |
6569 | return 2; |
6570 | } | |
6571 | ||
6572 | if (!check_name(super, name, 0)) | |
6573 | return 2; | |
6574 | ||
a951a4f7 N |
6575 | vol = strtoul(subarray, &ep, 10); |
6576 | if (*ep != '\0' || vol >= super->anchor->num_raid_devs) | |
6577 | return 2; | |
6578 | ||
aa534678 | 6579 | if (st->update_tail) { |
503975b9 | 6580 | struct imsm_update_rename_array *u = xmalloc(sizeof(*u)); |
aa534678 | 6581 | |
aa534678 | 6582 | u->type = update_rename_array; |
a951a4f7 | 6583 | u->dev_idx = vol; |
aa534678 DW |
6584 | snprintf((char *) u->name, MAX_RAID_SERIAL_LEN, "%s", name); |
6585 | append_metadata_update(st, u, sizeof(*u)); | |
6586 | } else { | |
6587 | struct imsm_dev *dev; | |
6588 | int i; | |
6589 | ||
a951a4f7 | 6590 | dev = get_imsm_dev(super, vol); |
aa534678 DW |
6591 | snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name); |
6592 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
6593 | dev = get_imsm_dev(super, i); | |
6594 | handle_missing(super, dev); | |
6595 | } | |
6596 | super->updates_pending++; | |
6597 | } | |
6598 | } else | |
6599 | return 2; | |
6600 | ||
6601 | return 0; | |
6602 | } | |
d1e02575 | 6603 | #endif /* MDASSEMBLE */ |
bf5a934a | 6604 | |
28bce06f AK |
6605 | static int is_gen_migration(struct imsm_dev *dev) |
6606 | { | |
7534230b AK |
6607 | if (dev == NULL) |
6608 | return 0; | |
6609 | ||
28bce06f AK |
6610 | if (!dev->vol.migr_state) |
6611 | return 0; | |
6612 | ||
6613 | if (migr_type(dev) == MIGR_GEN_MIGR) | |
6614 | return 1; | |
6615 | ||
6616 | return 0; | |
6617 | } | |
6618 | ||
1e5c6983 DW |
6619 | static int is_rebuilding(struct imsm_dev *dev) |
6620 | { | |
6621 | struct imsm_map *migr_map; | |
6622 | ||
6623 | if (!dev->vol.migr_state) | |
6624 | return 0; | |
6625 | ||
6626 | if (migr_type(dev) != MIGR_REBUILD) | |
6627 | return 0; | |
6628 | ||
238c0a71 | 6629 | migr_map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
6630 | |
6631 | if (migr_map->map_state == IMSM_T_STATE_DEGRADED) | |
6632 | return 1; | |
6633 | else | |
6634 | return 0; | |
6635 | } | |
6636 | ||
b4ab44d8 | 6637 | #ifndef MDASSEMBLE |
6ce1fbf1 AK |
6638 | static int is_initializing(struct imsm_dev *dev) |
6639 | { | |
6640 | struct imsm_map *migr_map; | |
6641 | ||
6642 | if (!dev->vol.migr_state) | |
6643 | return 0; | |
6644 | ||
6645 | if (migr_type(dev) != MIGR_INIT) | |
6646 | return 0; | |
6647 | ||
238c0a71 | 6648 | migr_map = get_imsm_map(dev, MAP_1); |
6ce1fbf1 AK |
6649 | |
6650 | if (migr_map->map_state == IMSM_T_STATE_UNINITIALIZED) | |
6651 | return 1; | |
6652 | ||
6653 | return 0; | |
6ce1fbf1 | 6654 | } |
b4ab44d8 | 6655 | #endif |
6ce1fbf1 | 6656 | |
c47b0ff6 AK |
6657 | static void update_recovery_start(struct intel_super *super, |
6658 | struct imsm_dev *dev, | |
6659 | struct mdinfo *array) | |
1e5c6983 DW |
6660 | { |
6661 | struct mdinfo *rebuild = NULL; | |
6662 | struct mdinfo *d; | |
6663 | __u32 units; | |
6664 | ||
6665 | if (!is_rebuilding(dev)) | |
6666 | return; | |
6667 | ||
6668 | /* Find the rebuild target, but punt on the dual rebuild case */ | |
6669 | for (d = array->devs; d; d = d->next) | |
6670 | if (d->recovery_start == 0) { | |
6671 | if (rebuild) | |
6672 | return; | |
6673 | rebuild = d; | |
6674 | } | |
6675 | ||
4363fd80 DW |
6676 | if (!rebuild) { |
6677 | /* (?) none of the disks are marked with | |
6678 | * IMSM_ORD_REBUILD, so assume they are missing and the | |
6679 | * disk_ord_tbl was not correctly updated | |
6680 | */ | |
1ade5cc1 | 6681 | dprintf("failed to locate out-of-sync disk\n"); |
4363fd80 DW |
6682 | return; |
6683 | } | |
6684 | ||
1e5c6983 | 6685 | units = __le32_to_cpu(dev->vol.curr_migr_unit); |
c47b0ff6 | 6686 | rebuild->recovery_start = units * blocks_per_migr_unit(super, dev); |
1e5c6983 DW |
6687 | } |
6688 | ||
9e2d750d | 6689 | #ifndef MDASSEMBLE |
276d77db | 6690 | static int recover_backup_imsm(struct supertype *st, struct mdinfo *info); |
9e2d750d | 6691 | #endif |
1e5c6983 | 6692 | |
00bbdbda | 6693 | static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray) |
cdddbdbc | 6694 | { |
4f5bc454 DW |
6695 | /* Given a container loaded by load_super_imsm_all, |
6696 | * extract information about all the arrays into | |
6697 | * an mdinfo tree. | |
00bbdbda | 6698 | * If 'subarray' is given, just extract info about that array. |
4f5bc454 DW |
6699 | * |
6700 | * For each imsm_dev create an mdinfo, fill it in, | |
6701 | * then look for matching devices in super->disks | |
6702 | * and create appropriate device mdinfo. | |
6703 | */ | |
6704 | struct intel_super *super = st->sb; | |
949c47a0 | 6705 | struct imsm_super *mpb = super->anchor; |
4f5bc454 | 6706 | struct mdinfo *rest = NULL; |
00bbdbda | 6707 | unsigned int i; |
81219e70 | 6708 | int sb_errors = 0; |
abef11a3 AK |
6709 | struct dl *d; |
6710 | int spare_disks = 0; | |
cdddbdbc | 6711 | |
19482bcc AK |
6712 | /* do not assemble arrays when not all attributes are supported */ |
6713 | if (imsm_check_attributes(mpb->attributes) == 0) { | |
81219e70 | 6714 | sb_errors = 1; |
7a862a02 | 6715 | pr_err("Unsupported attributes in IMSM metadata.Arrays activation is blocked.\n"); |
19482bcc AK |
6716 | } |
6717 | ||
a06d022d | 6718 | /* check for bad blocks */ |
81219e70 | 6719 | if (imsm_bbm_log_size(super->anchor)) { |
7a862a02 | 6720 | pr_err("BBM log found in IMSM metadata.Arrays activation is blocked.\n"); |
81219e70 LM |
6721 | sb_errors = 1; |
6722 | } | |
6723 | ||
abef11a3 AK |
6724 | /* count spare devices, not used in maps |
6725 | */ | |
6726 | for (d = super->disks; d; d = d->next) | |
6727 | if (d->index == -1) | |
6728 | spare_disks++; | |
6729 | ||
4f5bc454 | 6730 | for (i = 0; i < mpb->num_raid_devs; i++) { |
00bbdbda N |
6731 | struct imsm_dev *dev; |
6732 | struct imsm_map *map; | |
86e3692b | 6733 | struct imsm_map *map2; |
4f5bc454 | 6734 | struct mdinfo *this; |
a6482415 N |
6735 | int slot; |
6736 | #ifndef MDASSEMBLE | |
6737 | int chunk; | |
6738 | #endif | |
00bbdbda N |
6739 | char *ep; |
6740 | ||
6741 | if (subarray && | |
6742 | (i != strtoul(subarray, &ep, 10) || *ep != '\0')) | |
6743 | continue; | |
6744 | ||
6745 | dev = get_imsm_dev(super, i); | |
238c0a71 AK |
6746 | map = get_imsm_map(dev, MAP_0); |
6747 | map2 = get_imsm_map(dev, MAP_1); | |
4f5bc454 | 6748 | |
1ce0101c DW |
6749 | /* do not publish arrays that are in the middle of an |
6750 | * unsupported migration | |
6751 | */ | |
6752 | if (dev->vol.migr_state && | |
28bce06f | 6753 | (migr_type(dev) == MIGR_STATE_CHANGE)) { |
7a862a02 | 6754 | pr_err("cannot assemble volume '%.16s': unsupported migration in progress\n", |
1ce0101c DW |
6755 | dev->volume); |
6756 | continue; | |
6757 | } | |
2db86302 LM |
6758 | /* do not publish arrays that are not support by controller's |
6759 | * OROM/EFI | |
6760 | */ | |
1ce0101c | 6761 | |
503975b9 | 6762 | this = xmalloc(sizeof(*this)); |
4f5bc454 | 6763 | |
301406c9 | 6764 | super->current_vol = i; |
a5d85af7 | 6765 | getinfo_super_imsm_volume(st, this, NULL); |
9894ec0d | 6766 | this->next = rest; |
81219e70 | 6767 | #ifndef MDASSEMBLE |
a6482415 | 6768 | chunk = __le16_to_cpu(map->blocks_per_strip) >> 1; |
81219e70 LM |
6769 | /* mdadm does not support all metadata features- set the bit in all arrays state */ |
6770 | if (!validate_geometry_imsm_orom(super, | |
6771 | get_imsm_raid_level(map), /* RAID level */ | |
6772 | imsm_level_to_layout(get_imsm_raid_level(map)), | |
6773 | map->num_members, /* raid disks */ | |
2cc699af | 6774 | &chunk, join_u32(dev->size_low, dev->size_high), |
81219e70 | 6775 | 1 /* verbose */)) { |
7a862a02 | 6776 | pr_err("IMSM RAID geometry validation failed. Array %s activation is blocked.\n", |
81219e70 LM |
6777 | dev->volume); |
6778 | this->array.state |= | |
6779 | (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) | | |
6780 | (1<<MD_SB_BLOCK_VOLUME); | |
6781 | } | |
6782 | #endif | |
6783 | ||
6784 | /* if array has bad blocks, set suitable bit in all arrays state */ | |
6785 | if (sb_errors) | |
6786 | this->array.state |= | |
6787 | (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) | | |
6788 | (1<<MD_SB_BLOCK_VOLUME); | |
6789 | ||
4f5bc454 | 6790 | for (slot = 0 ; slot < map->num_members; slot++) { |
1e5c6983 | 6791 | unsigned long long recovery_start; |
4f5bc454 DW |
6792 | struct mdinfo *info_d; |
6793 | struct dl *d; | |
6794 | int idx; | |
9a1608e5 | 6795 | int skip; |
7eef0453 | 6796 | __u32 ord; |
4f5bc454 | 6797 | |
9a1608e5 | 6798 | skip = 0; |
238c0a71 AK |
6799 | idx = get_imsm_disk_idx(dev, slot, MAP_0); |
6800 | ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); | |
4f5bc454 DW |
6801 | for (d = super->disks; d ; d = d->next) |
6802 | if (d->index == idx) | |
0fbd635c | 6803 | break; |
4f5bc454 | 6804 | |
1e5c6983 | 6805 | recovery_start = MaxSector; |
4f5bc454 | 6806 | if (d == NULL) |
9a1608e5 | 6807 | skip = 1; |
25ed7e59 | 6808 | if (d && is_failed(&d->disk)) |
9a1608e5 | 6809 | skip = 1; |
7eef0453 | 6810 | if (ord & IMSM_ORD_REBUILD) |
1e5c6983 | 6811 | recovery_start = 0; |
9a1608e5 | 6812 | |
1011e834 | 6813 | /* |
9a1608e5 | 6814 | * if we skip some disks the array will be assmebled degraded; |
1e5c6983 DW |
6815 | * reset resync start to avoid a dirty-degraded |
6816 | * situation when performing the intial sync | |
9a1608e5 DW |
6817 | * |
6818 | * FIXME handle dirty degraded | |
6819 | */ | |
1e5c6983 | 6820 | if ((skip || recovery_start == 0) && !dev->vol.dirty) |
b7528a20 | 6821 | this->resync_start = MaxSector; |
9a1608e5 DW |
6822 | if (skip) |
6823 | continue; | |
4f5bc454 | 6824 | |
503975b9 | 6825 | info_d = xcalloc(1, sizeof(*info_d)); |
4f5bc454 DW |
6826 | info_d->next = this->devs; |
6827 | this->devs = info_d; | |
6828 | ||
4f5bc454 DW |
6829 | info_d->disk.number = d->index; |
6830 | info_d->disk.major = d->major; | |
6831 | info_d->disk.minor = d->minor; | |
6832 | info_d->disk.raid_disk = slot; | |
1e5c6983 | 6833 | info_d->recovery_start = recovery_start; |
86e3692b AK |
6834 | if (map2) { |
6835 | if (slot < map2->num_members) | |
6836 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
6837 | else |
6838 | this->array.spare_disks++; | |
86e3692b AK |
6839 | } else { |
6840 | if (slot < map->num_members) | |
6841 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
6842 | else |
6843 | this->array.spare_disks++; | |
86e3692b | 6844 | } |
1e5c6983 DW |
6845 | if (info_d->recovery_start == MaxSector) |
6846 | this->array.working_disks++; | |
4f5bc454 DW |
6847 | |
6848 | info_d->events = __le32_to_cpu(mpb->generation_num); | |
5551b113 CA |
6849 | info_d->data_offset = pba_of_lba0(map); |
6850 | info_d->component_size = blocks_per_member(map); | |
4f5bc454 | 6851 | } |
1e5c6983 | 6852 | /* now that the disk list is up-to-date fixup recovery_start */ |
c47b0ff6 | 6853 | update_recovery_start(super, dev, this); |
abef11a3 | 6854 | this->array.spare_disks += spare_disks; |
276d77db | 6855 | |
9e2d750d | 6856 | #ifndef MDASSEMBLE |
276d77db AK |
6857 | /* check for reshape */ |
6858 | if (this->reshape_active == 1) | |
6859 | recover_backup_imsm(st, this); | |
9e2d750d | 6860 | #endif |
9a1608e5 | 6861 | rest = this; |
4f5bc454 DW |
6862 | } |
6863 | ||
6864 | return rest; | |
cdddbdbc DW |
6865 | } |
6866 | ||
3b451610 AK |
6867 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, |
6868 | int failed, int look_in_map) | |
c2a1e7da | 6869 | { |
3b451610 AK |
6870 | struct imsm_map *map; |
6871 | ||
6872 | map = get_imsm_map(dev, look_in_map); | |
c2a1e7da DW |
6873 | |
6874 | if (!failed) | |
1011e834 | 6875 | return map->map_state == IMSM_T_STATE_UNINITIALIZED ? |
3393c6af | 6876 | IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL; |
c2a1e7da DW |
6877 | |
6878 | switch (get_imsm_raid_level(map)) { | |
6879 | case 0: | |
6880 | return IMSM_T_STATE_FAILED; | |
6881 | break; | |
6882 | case 1: | |
6883 | if (failed < map->num_members) | |
6884 | return IMSM_T_STATE_DEGRADED; | |
6885 | else | |
6886 | return IMSM_T_STATE_FAILED; | |
6887 | break; | |
6888 | case 10: | |
6889 | { | |
6890 | /** | |
c92a2527 DW |
6891 | * check to see if any mirrors have failed, otherwise we |
6892 | * are degraded. Even numbered slots are mirrored on | |
6893 | * slot+1 | |
c2a1e7da | 6894 | */ |
c2a1e7da | 6895 | int i; |
d9b420a5 N |
6896 | /* gcc -Os complains that this is unused */ |
6897 | int insync = insync; | |
c2a1e7da DW |
6898 | |
6899 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 6900 | __u32 ord = get_imsm_ord_tbl_ent(dev, i, MAP_X); |
c92a2527 DW |
6901 | int idx = ord_to_idx(ord); |
6902 | struct imsm_disk *disk; | |
c2a1e7da | 6903 | |
c92a2527 | 6904 | /* reset the potential in-sync count on even-numbered |
1011e834 | 6905 | * slots. num_copies is always 2 for imsm raid10 |
c92a2527 DW |
6906 | */ |
6907 | if ((i & 1) == 0) | |
6908 | insync = 2; | |
c2a1e7da | 6909 | |
c92a2527 | 6910 | disk = get_imsm_disk(super, idx); |
25ed7e59 | 6911 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) |
c92a2527 | 6912 | insync--; |
c2a1e7da | 6913 | |
c92a2527 DW |
6914 | /* no in-sync disks left in this mirror the |
6915 | * array has failed | |
6916 | */ | |
6917 | if (insync == 0) | |
6918 | return IMSM_T_STATE_FAILED; | |
c2a1e7da DW |
6919 | } |
6920 | ||
6921 | return IMSM_T_STATE_DEGRADED; | |
6922 | } | |
6923 | case 5: | |
6924 | if (failed < 2) | |
6925 | return IMSM_T_STATE_DEGRADED; | |
6926 | else | |
6927 | return IMSM_T_STATE_FAILED; | |
6928 | break; | |
6929 | default: | |
6930 | break; | |
6931 | } | |
6932 | ||
6933 | return map->map_state; | |
6934 | } | |
6935 | ||
3b451610 AK |
6936 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev, |
6937 | int look_in_map) | |
c2a1e7da DW |
6938 | { |
6939 | int i; | |
6940 | int failed = 0; | |
6941 | struct imsm_disk *disk; | |
d5985138 AK |
6942 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
6943 | struct imsm_map *prev = get_imsm_map(dev, MAP_1); | |
68fe4598 | 6944 | struct imsm_map *map_for_loop; |
0556e1a2 DW |
6945 | __u32 ord; |
6946 | int idx; | |
d5985138 | 6947 | int idx_1; |
c2a1e7da | 6948 | |
0556e1a2 DW |
6949 | /* at the beginning of migration we set IMSM_ORD_REBUILD on |
6950 | * disks that are being rebuilt. New failures are recorded to | |
6951 | * map[0]. So we look through all the disks we started with and | |
6952 | * see if any failures are still present, or if any new ones | |
6953 | * have arrived | |
0556e1a2 | 6954 | */ |
d5985138 AK |
6955 | map_for_loop = map; |
6956 | if (prev && (map->num_members < prev->num_members)) | |
6957 | map_for_loop = prev; | |
68fe4598 LD |
6958 | |
6959 | for (i = 0; i < map_for_loop->num_members; i++) { | |
d5985138 | 6960 | idx_1 = -255; |
238c0a71 AK |
6961 | /* when MAP_X is passed both maps failures are counted |
6962 | */ | |
d5985138 | 6963 | if (prev && |
238c0a71 AK |
6964 | ((look_in_map == MAP_1) || (look_in_map == MAP_X)) && |
6965 | (i < prev->num_members)) { | |
d5985138 AK |
6966 | ord = __le32_to_cpu(prev->disk_ord_tbl[i]); |
6967 | idx_1 = ord_to_idx(ord); | |
c2a1e7da | 6968 | |
d5985138 AK |
6969 | disk = get_imsm_disk(super, idx_1); |
6970 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) | |
6971 | failed++; | |
6972 | } | |
238c0a71 AK |
6973 | if (((look_in_map == MAP_0) || (look_in_map == MAP_X)) && |
6974 | (i < map->num_members)) { | |
d5985138 AK |
6975 | ord = __le32_to_cpu(map->disk_ord_tbl[i]); |
6976 | idx = ord_to_idx(ord); | |
6977 | ||
6978 | if (idx != idx_1) { | |
6979 | disk = get_imsm_disk(super, idx); | |
6980 | if (!disk || is_failed(disk) || | |
6981 | ord & IMSM_ORD_REBUILD) | |
6982 | failed++; | |
6983 | } | |
6984 | } | |
c2a1e7da DW |
6985 | } |
6986 | ||
6987 | return failed; | |
845dea95 NB |
6988 | } |
6989 | ||
97b4d0e9 DW |
6990 | #ifndef MDASSEMBLE |
6991 | static int imsm_open_new(struct supertype *c, struct active_array *a, | |
6992 | char *inst) | |
6993 | { | |
6994 | struct intel_super *super = c->sb; | |
6995 | struct imsm_super *mpb = super->anchor; | |
9587c373 | 6996 | |
97b4d0e9 | 6997 | if (atoi(inst) >= mpb->num_raid_devs) { |
1ade5cc1 | 6998 | pr_err("subarry index %d, out of range\n", atoi(inst)); |
97b4d0e9 DW |
6999 | return -ENODEV; |
7000 | } | |
7001 | ||
7002 | dprintf("imsm: open_new %s\n", inst); | |
7003 | a->info.container_member = atoi(inst); | |
7004 | return 0; | |
7005 | } | |
7006 | ||
0c046afd DW |
7007 | static int is_resyncing(struct imsm_dev *dev) |
7008 | { | |
7009 | struct imsm_map *migr_map; | |
7010 | ||
7011 | if (!dev->vol.migr_state) | |
7012 | return 0; | |
7013 | ||
1484e727 DW |
7014 | if (migr_type(dev) == MIGR_INIT || |
7015 | migr_type(dev) == MIGR_REPAIR) | |
0c046afd DW |
7016 | return 1; |
7017 | ||
4c9bc37b AK |
7018 | if (migr_type(dev) == MIGR_GEN_MIGR) |
7019 | return 0; | |
7020 | ||
238c0a71 | 7021 | migr_map = get_imsm_map(dev, MAP_1); |
0c046afd | 7022 | |
4c9bc37b AK |
7023 | if ((migr_map->map_state == IMSM_T_STATE_NORMAL) && |
7024 | (dev->vol.migr_type != MIGR_GEN_MIGR)) | |
0c046afd DW |
7025 | return 1; |
7026 | else | |
7027 | return 0; | |
7028 | } | |
7029 | ||
0556e1a2 DW |
7030 | /* return true if we recorded new information */ |
7031 | static int mark_failure(struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
47ee5a45 | 7032 | { |
0556e1a2 DW |
7033 | __u32 ord; |
7034 | int slot; | |
7035 | struct imsm_map *map; | |
86c54047 DW |
7036 | char buf[MAX_RAID_SERIAL_LEN+3]; |
7037 | unsigned int len, shift = 0; | |
0556e1a2 DW |
7038 | |
7039 | /* new failures are always set in map[0] */ | |
238c0a71 | 7040 | map = get_imsm_map(dev, MAP_0); |
0556e1a2 DW |
7041 | |
7042 | slot = get_imsm_disk_slot(map, idx); | |
7043 | if (slot < 0) | |
7044 | return 0; | |
7045 | ||
7046 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
25ed7e59 | 7047 | if (is_failed(disk) && (ord & IMSM_ORD_REBUILD)) |
0556e1a2 DW |
7048 | return 0; |
7049 | ||
7d0c5e24 LD |
7050 | memcpy(buf, disk->serial, MAX_RAID_SERIAL_LEN); |
7051 | buf[MAX_RAID_SERIAL_LEN] = '\000'; | |
7052 | strcat(buf, ":0"); | |
86c54047 DW |
7053 | if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN) |
7054 | shift = len - MAX_RAID_SERIAL_LEN + 1; | |
7055 | strncpy((char *)disk->serial, &buf[shift], MAX_RAID_SERIAL_LEN); | |
7056 | ||
f2f27e63 | 7057 | disk->status |= FAILED_DISK; |
0556e1a2 | 7058 | set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD); |
17788994 AK |
7059 | /* mark failures in second map if second map exists and this disk |
7060 | * in this slot. | |
7061 | * This is valid for migration, initialization and rebuild | |
7062 | */ | |
7063 | if (dev->vol.migr_state) { | |
238c0a71 | 7064 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); |
0a108d63 AK |
7065 | int slot2 = get_imsm_disk_slot(map2, idx); |
7066 | ||
7067 | if ((slot2 < map2->num_members) && | |
7068 | (slot2 >= 0)) | |
7069 | set_imsm_ord_tbl_ent(map2, slot2, | |
1ace8403 AK |
7070 | idx | IMSM_ORD_REBUILD); |
7071 | } | |
f21e18ca | 7072 | if (map->failed_disk_num == 0xff) |
0556e1a2 DW |
7073 | map->failed_disk_num = slot; |
7074 | return 1; | |
7075 | } | |
7076 | ||
7077 | static void mark_missing(struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
7078 | { | |
7079 | mark_failure(dev, disk, idx); | |
7080 | ||
7081 | if (disk->scsi_id == __cpu_to_le32(~(__u32)0)) | |
7082 | return; | |
7083 | ||
47ee5a45 DW |
7084 | disk->scsi_id = __cpu_to_le32(~(__u32)0); |
7085 | memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1); | |
7086 | } | |
7087 | ||
33414a01 DW |
7088 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev) |
7089 | { | |
33414a01 | 7090 | struct dl *dl; |
33414a01 DW |
7091 | |
7092 | if (!super->missing) | |
7093 | return; | |
33414a01 | 7094 | |
79b68f1b PC |
7095 | /* When orom adds replacement for missing disk it does |
7096 | * not remove entry of missing disk, but just updates map with | |
7097 | * new added disk. So it is not enough just to test if there is | |
7098 | * any missing disk, we have to look if there are any failed disks | |
7099 | * in map to stop migration */ | |
7100 | ||
33414a01 | 7101 | dprintf("imsm: mark missing\n"); |
3d59f0c0 AK |
7102 | /* end process for initialization and rebuild only |
7103 | */ | |
7104 | if (is_gen_migration(dev) == 0) { | |
7105 | __u8 map_state; | |
7106 | int failed; | |
7107 | ||
7108 | failed = imsm_count_failed(super, dev, MAP_0); | |
7109 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
7110 | ||
79b68f1b PC |
7111 | if (failed) |
7112 | end_migration(dev, super, map_state); | |
3d59f0c0 | 7113 | } |
33414a01 DW |
7114 | for (dl = super->missing; dl; dl = dl->next) |
7115 | mark_missing(dev, &dl->disk, dl->index); | |
7116 | super->updates_pending++; | |
7117 | } | |
7118 | ||
f3871fdc AK |
7119 | static unsigned long long imsm_set_array_size(struct imsm_dev *dev, |
7120 | long long new_size) | |
70bdf0dc | 7121 | { |
238c0a71 | 7122 | int used_disks = imsm_num_data_members(dev, MAP_0); |
70bdf0dc AK |
7123 | unsigned long long array_blocks; |
7124 | struct imsm_map *map; | |
7125 | ||
7126 | if (used_disks == 0) { | |
7127 | /* when problems occures | |
7128 | * return current array_blocks value | |
7129 | */ | |
7130 | array_blocks = __le32_to_cpu(dev->size_high); | |
7131 | array_blocks = array_blocks << 32; | |
7132 | array_blocks += __le32_to_cpu(dev->size_low); | |
7133 | ||
7134 | return array_blocks; | |
7135 | } | |
7136 | ||
7137 | /* set array size in metadata | |
7138 | */ | |
f3871fdc AK |
7139 | if (new_size <= 0) { |
7140 | /* OLCE size change is caused by added disks | |
7141 | */ | |
7142 | map = get_imsm_map(dev, MAP_0); | |
7143 | array_blocks = blocks_per_member(map) * used_disks; | |
7144 | } else { | |
7145 | /* Online Volume Size Change | |
7146 | * Using available free space | |
7147 | */ | |
7148 | array_blocks = new_size; | |
7149 | } | |
70bdf0dc AK |
7150 | |
7151 | /* round array size down to closest MB | |
7152 | */ | |
7153 | array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; | |
7154 | dev->size_low = __cpu_to_le32((__u32)array_blocks); | |
7155 | dev->size_high = __cpu_to_le32((__u32)(array_blocks >> 32)); | |
7156 | ||
7157 | return array_blocks; | |
7158 | } | |
7159 | ||
28bce06f AK |
7160 | static void imsm_set_disk(struct active_array *a, int n, int state); |
7161 | ||
0e2d1a4e AK |
7162 | static void imsm_progress_container_reshape(struct intel_super *super) |
7163 | { | |
7164 | /* if no device has a migr_state, but some device has a | |
7165 | * different number of members than the previous device, start | |
7166 | * changing the number of devices in this device to match | |
7167 | * previous. | |
7168 | */ | |
7169 | struct imsm_super *mpb = super->anchor; | |
7170 | int prev_disks = -1; | |
7171 | int i; | |
1dfaa380 | 7172 | int copy_map_size; |
0e2d1a4e AK |
7173 | |
7174 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
7175 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 7176 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
0e2d1a4e AK |
7177 | struct imsm_map *map2; |
7178 | int prev_num_members; | |
0e2d1a4e AK |
7179 | |
7180 | if (dev->vol.migr_state) | |
7181 | return; | |
7182 | ||
7183 | if (prev_disks == -1) | |
7184 | prev_disks = map->num_members; | |
7185 | if (prev_disks == map->num_members) | |
7186 | continue; | |
7187 | ||
7188 | /* OK, this array needs to enter reshape mode. | |
7189 | * i.e it needs a migr_state | |
7190 | */ | |
7191 | ||
1dfaa380 | 7192 | copy_map_size = sizeof_imsm_map(map); |
0e2d1a4e AK |
7193 | prev_num_members = map->num_members; |
7194 | map->num_members = prev_disks; | |
7195 | dev->vol.migr_state = 1; | |
7196 | dev->vol.curr_migr_unit = 0; | |
ea672ee1 | 7197 | set_migr_type(dev, MIGR_GEN_MIGR); |
0e2d1a4e AK |
7198 | for (i = prev_num_members; |
7199 | i < map->num_members; i++) | |
7200 | set_imsm_ord_tbl_ent(map, i, i); | |
238c0a71 | 7201 | map2 = get_imsm_map(dev, MAP_1); |
0e2d1a4e | 7202 | /* Copy the current map */ |
1dfaa380 | 7203 | memcpy(map2, map, copy_map_size); |
0e2d1a4e AK |
7204 | map2->num_members = prev_num_members; |
7205 | ||
f3871fdc | 7206 | imsm_set_array_size(dev, -1); |
51d83f5d | 7207 | super->clean_migration_record_by_mdmon = 1; |
0e2d1a4e AK |
7208 | super->updates_pending++; |
7209 | } | |
7210 | } | |
7211 | ||
aad6f216 | 7212 | /* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild |
0c046afd DW |
7213 | * states are handled in imsm_set_disk() with one exception, when a |
7214 | * resync is stopped due to a new failure this routine will set the | |
7215 | * 'degraded' state for the array. | |
7216 | */ | |
01f157d7 | 7217 | static int imsm_set_array_state(struct active_array *a, int consistent) |
a862209d DW |
7218 | { |
7219 | int inst = a->info.container_member; | |
7220 | struct intel_super *super = a->container->sb; | |
949c47a0 | 7221 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 7222 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
3b451610 AK |
7223 | int failed = imsm_count_failed(super, dev, MAP_0); |
7224 | __u8 map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
1e5c6983 | 7225 | __u32 blocks_per_unit; |
a862209d | 7226 | |
1af97990 AK |
7227 | if (dev->vol.migr_state && |
7228 | dev->vol.migr_type == MIGR_GEN_MIGR) { | |
7229 | /* array state change is blocked due to reshape action | |
aad6f216 N |
7230 | * We might need to |
7231 | * - abort the reshape (if last_checkpoint is 0 and action!= reshape) | |
7232 | * - finish the reshape (if last_checkpoint is big and action != reshape) | |
7233 | * - update curr_migr_unit | |
1af97990 | 7234 | */ |
aad6f216 N |
7235 | if (a->curr_action == reshape) { |
7236 | /* still reshaping, maybe update curr_migr_unit */ | |
633b5610 | 7237 | goto mark_checkpoint; |
aad6f216 N |
7238 | } else { |
7239 | if (a->last_checkpoint == 0 && a->prev_action == reshape) { | |
7240 | /* for some reason we aborted the reshape. | |
b66e591b AK |
7241 | * |
7242 | * disable automatic metadata rollback | |
7243 | * user action is required to recover process | |
aad6f216 | 7244 | */ |
b66e591b | 7245 | if (0) { |
238c0a71 AK |
7246 | struct imsm_map *map2 = |
7247 | get_imsm_map(dev, MAP_1); | |
7248 | dev->vol.migr_state = 0; | |
7249 | set_migr_type(dev, 0); | |
7250 | dev->vol.curr_migr_unit = 0; | |
7251 | memcpy(map, map2, | |
7252 | sizeof_imsm_map(map2)); | |
7253 | super->updates_pending++; | |
b66e591b | 7254 | } |
aad6f216 N |
7255 | } |
7256 | if (a->last_checkpoint >= a->info.component_size) { | |
7257 | unsigned long long array_blocks; | |
7258 | int used_disks; | |
e154ced3 | 7259 | struct mdinfo *mdi; |
aad6f216 | 7260 | |
238c0a71 | 7261 | used_disks = imsm_num_data_members(dev, MAP_0); |
d55adef9 AK |
7262 | if (used_disks > 0) { |
7263 | array_blocks = | |
5551b113 | 7264 | blocks_per_member(map) * |
d55adef9 AK |
7265 | used_disks; |
7266 | /* round array size down to closest MB | |
7267 | */ | |
7268 | array_blocks = (array_blocks | |
7269 | >> SECT_PER_MB_SHIFT) | |
7270 | << SECT_PER_MB_SHIFT; | |
d55adef9 AK |
7271 | a->info.custom_array_size = array_blocks; |
7272 | /* encourage manager to update array | |
7273 | * size | |
7274 | */ | |
e154ced3 | 7275 | |
d55adef9 | 7276 | a->check_reshape = 1; |
633b5610 | 7277 | } |
e154ced3 AK |
7278 | /* finalize online capacity expansion/reshape */ |
7279 | for (mdi = a->info.devs; mdi; mdi = mdi->next) | |
7280 | imsm_set_disk(a, | |
7281 | mdi->disk.raid_disk, | |
7282 | mdi->curr_state); | |
7283 | ||
0e2d1a4e | 7284 | imsm_progress_container_reshape(super); |
e154ced3 | 7285 | } |
aad6f216 | 7286 | } |
1af97990 AK |
7287 | } |
7288 | ||
47ee5a45 | 7289 | /* before we activate this array handle any missing disks */ |
33414a01 DW |
7290 | if (consistent == 2) |
7291 | handle_missing(super, dev); | |
1e5c6983 | 7292 | |
0c046afd | 7293 | if (consistent == 2 && |
b7941fd6 | 7294 | (!is_resync_complete(&a->info) || |
0c046afd DW |
7295 | map_state != IMSM_T_STATE_NORMAL || |
7296 | dev->vol.migr_state)) | |
01f157d7 | 7297 | consistent = 0; |
272906ef | 7298 | |
b7941fd6 | 7299 | if (is_resync_complete(&a->info)) { |
0c046afd | 7300 | /* complete intialization / resync, |
0556e1a2 DW |
7301 | * recovery and interrupted recovery is completed in |
7302 | * ->set_disk | |
0c046afd DW |
7303 | */ |
7304 | if (is_resyncing(dev)) { | |
7305 | dprintf("imsm: mark resync done\n"); | |
809da78e | 7306 | end_migration(dev, super, map_state); |
115c3803 | 7307 | super->updates_pending++; |
484240d8 | 7308 | a->last_checkpoint = 0; |
115c3803 | 7309 | } |
b9172665 AK |
7310 | } else if ((!is_resyncing(dev) && !failed) && |
7311 | (imsm_reshape_blocks_arrays_changes(super) == 0)) { | |
0c046afd | 7312 | /* mark the start of the init process if nothing is failed */ |
b7941fd6 | 7313 | dprintf("imsm: mark resync start\n"); |
1484e727 | 7314 | if (map->map_state == IMSM_T_STATE_UNINITIALIZED) |
8e59f3d8 | 7315 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT); |
1484e727 | 7316 | else |
8e59f3d8 | 7317 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR); |
3393c6af | 7318 | super->updates_pending++; |
115c3803 | 7319 | } |
a862209d | 7320 | |
633b5610 | 7321 | mark_checkpoint: |
5b83bacf AK |
7322 | /* skip checkpointing for general migration, |
7323 | * it is controlled in mdadm | |
7324 | */ | |
7325 | if (is_gen_migration(dev)) | |
7326 | goto skip_mark_checkpoint; | |
7327 | ||
1e5c6983 | 7328 | /* check if we can update curr_migr_unit from resync_start, recovery_start */ |
c47b0ff6 | 7329 | blocks_per_unit = blocks_per_migr_unit(super, dev); |
4f0a7acc | 7330 | if (blocks_per_unit) { |
1e5c6983 DW |
7331 | __u32 units32; |
7332 | __u64 units; | |
7333 | ||
4f0a7acc | 7334 | units = a->last_checkpoint / blocks_per_unit; |
1e5c6983 DW |
7335 | units32 = units; |
7336 | ||
7337 | /* check that we did not overflow 32-bits, and that | |
7338 | * curr_migr_unit needs updating | |
7339 | */ | |
7340 | if (units32 == units && | |
bfd80a56 | 7341 | units32 != 0 && |
1e5c6983 DW |
7342 | __le32_to_cpu(dev->vol.curr_migr_unit) != units32) { |
7343 | dprintf("imsm: mark checkpoint (%u)\n", units32); | |
7344 | dev->vol.curr_migr_unit = __cpu_to_le32(units32); | |
7345 | super->updates_pending++; | |
7346 | } | |
7347 | } | |
f8f603f1 | 7348 | |
5b83bacf | 7349 | skip_mark_checkpoint: |
3393c6af | 7350 | /* mark dirty / clean */ |
0c046afd | 7351 | if (dev->vol.dirty != !consistent) { |
b7941fd6 | 7352 | dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty"); |
0c046afd DW |
7353 | if (consistent) |
7354 | dev->vol.dirty = 0; | |
7355 | else | |
7356 | dev->vol.dirty = 1; | |
a862209d DW |
7357 | super->updates_pending++; |
7358 | } | |
28bce06f | 7359 | |
01f157d7 | 7360 | return consistent; |
a862209d DW |
7361 | } |
7362 | ||
8d45d196 | 7363 | static void imsm_set_disk(struct active_array *a, int n, int state) |
845dea95 | 7364 | { |
8d45d196 DW |
7365 | int inst = a->info.container_member; |
7366 | struct intel_super *super = a->container->sb; | |
949c47a0 | 7367 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 7368 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
8d45d196 | 7369 | struct imsm_disk *disk; |
7ce05701 LD |
7370 | struct mdinfo *mdi; |
7371 | int recovery_not_finished = 0; | |
0c046afd | 7372 | int failed; |
b10b37b8 | 7373 | __u32 ord; |
0c046afd | 7374 | __u8 map_state; |
8d45d196 DW |
7375 | |
7376 | if (n > map->num_members) | |
e12b3daa | 7377 | pr_err("imsm: set_disk %d out of range 0..%d\n", |
8d45d196 DW |
7378 | n, map->num_members - 1); |
7379 | ||
7380 | if (n < 0) | |
7381 | return; | |
7382 | ||
4e6e574a | 7383 | dprintf("imsm: set_disk %d:%x\n", n, state); |
8d45d196 | 7384 | |
238c0a71 | 7385 | ord = get_imsm_ord_tbl_ent(dev, n, MAP_0); |
b10b37b8 | 7386 | disk = get_imsm_disk(super, ord_to_idx(ord)); |
8d45d196 | 7387 | |
5802a811 | 7388 | /* check for new failures */ |
0556e1a2 DW |
7389 | if (state & DS_FAULTY) { |
7390 | if (mark_failure(dev, disk, ord_to_idx(ord))) | |
7391 | super->updates_pending++; | |
8d45d196 | 7392 | } |
47ee5a45 | 7393 | |
19859edc | 7394 | /* check if in_sync */ |
0556e1a2 | 7395 | if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) { |
238c0a71 | 7396 | struct imsm_map *migr_map = get_imsm_map(dev, MAP_1); |
b10b37b8 DW |
7397 | |
7398 | set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord)); | |
19859edc DW |
7399 | super->updates_pending++; |
7400 | } | |
8d45d196 | 7401 | |
3b451610 AK |
7402 | failed = imsm_count_failed(super, dev, MAP_0); |
7403 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
5802a811 | 7404 | |
0c046afd | 7405 | /* check if recovery complete, newly degraded, or failed */ |
94002678 AK |
7406 | dprintf("imsm: Detected transition to state "); |
7407 | switch (map_state) { | |
7408 | case IMSM_T_STATE_NORMAL: /* transition to normal state */ | |
7409 | dprintf("normal: "); | |
7410 | if (is_rebuilding(dev)) { | |
1ade5cc1 | 7411 | dprintf_cont("while rebuilding"); |
7ce05701 LD |
7412 | /* check if recovery is really finished */ |
7413 | for (mdi = a->info.devs; mdi ; mdi = mdi->next) | |
7414 | if (mdi->recovery_start != MaxSector) { | |
7415 | recovery_not_finished = 1; | |
7416 | break; | |
7417 | } | |
7418 | if (recovery_not_finished) { | |
1ade5cc1 N |
7419 | dprintf_cont("\n"); |
7420 | dprintf("Rebuild has not finished yet, state not changed"); | |
7ce05701 LD |
7421 | if (a->last_checkpoint < mdi->recovery_start) { |
7422 | a->last_checkpoint = mdi->recovery_start; | |
7423 | super->updates_pending++; | |
7424 | } | |
7425 | break; | |
7426 | } | |
94002678 | 7427 | end_migration(dev, super, map_state); |
238c0a71 | 7428 | map = get_imsm_map(dev, MAP_0); |
94002678 AK |
7429 | map->failed_disk_num = ~0; |
7430 | super->updates_pending++; | |
7431 | a->last_checkpoint = 0; | |
7432 | break; | |
7433 | } | |
7434 | if (is_gen_migration(dev)) { | |
1ade5cc1 | 7435 | dprintf_cont("while general migration"); |
bf2f0071 | 7436 | if (a->last_checkpoint >= a->info.component_size) |
809da78e | 7437 | end_migration(dev, super, map_state); |
94002678 AK |
7438 | else |
7439 | map->map_state = map_state; | |
238c0a71 | 7440 | map = get_imsm_map(dev, MAP_0); |
28bce06f | 7441 | map->failed_disk_num = ~0; |
94002678 | 7442 | super->updates_pending++; |
bf2f0071 | 7443 | break; |
94002678 AK |
7444 | } |
7445 | break; | |
7446 | case IMSM_T_STATE_DEGRADED: /* transition to degraded state */ | |
1ade5cc1 | 7447 | dprintf_cont("degraded: "); |
94002678 AK |
7448 | if ((map->map_state != map_state) && |
7449 | !dev->vol.migr_state) { | |
1ade5cc1 | 7450 | dprintf_cont("mark degraded"); |
94002678 AK |
7451 | map->map_state = map_state; |
7452 | super->updates_pending++; | |
7453 | a->last_checkpoint = 0; | |
7454 | break; | |
7455 | } | |
7456 | if (is_rebuilding(dev)) { | |
1ade5cc1 | 7457 | dprintf_cont("while rebuilding."); |
94002678 | 7458 | if (map->map_state != map_state) { |
1ade5cc1 | 7459 | dprintf_cont(" Map state change"); |
94002678 AK |
7460 | end_migration(dev, super, map_state); |
7461 | super->updates_pending++; | |
7462 | } | |
7463 | break; | |
7464 | } | |
7465 | if (is_gen_migration(dev)) { | |
1ade5cc1 | 7466 | dprintf_cont("while general migration"); |
bf2f0071 | 7467 | if (a->last_checkpoint >= a->info.component_size) |
809da78e | 7468 | end_migration(dev, super, map_state); |
94002678 AK |
7469 | else { |
7470 | map->map_state = map_state; | |
3b451610 | 7471 | manage_second_map(super, dev); |
94002678 AK |
7472 | } |
7473 | super->updates_pending++; | |
bf2f0071 | 7474 | break; |
28bce06f | 7475 | } |
6ce1fbf1 | 7476 | if (is_initializing(dev)) { |
1ade5cc1 | 7477 | dprintf_cont("while initialization."); |
6ce1fbf1 AK |
7478 | map->map_state = map_state; |
7479 | super->updates_pending++; | |
7480 | break; | |
7481 | } | |
94002678 AK |
7482 | break; |
7483 | case IMSM_T_STATE_FAILED: /* transition to failed state */ | |
1ade5cc1 | 7484 | dprintf_cont("failed: "); |
94002678 | 7485 | if (is_gen_migration(dev)) { |
1ade5cc1 | 7486 | dprintf_cont("while general migration"); |
94002678 AK |
7487 | map->map_state = map_state; |
7488 | super->updates_pending++; | |
7489 | break; | |
7490 | } | |
7491 | if (map->map_state != map_state) { | |
1ade5cc1 | 7492 | dprintf_cont("mark failed"); |
94002678 AK |
7493 | end_migration(dev, super, map_state); |
7494 | super->updates_pending++; | |
7495 | a->last_checkpoint = 0; | |
7496 | break; | |
7497 | } | |
7498 | break; | |
7499 | default: | |
1ade5cc1 | 7500 | dprintf_cont("state %i\n", map_state); |
5802a811 | 7501 | } |
1ade5cc1 | 7502 | dprintf_cont("\n"); |
845dea95 NB |
7503 | } |
7504 | ||
f796af5d | 7505 | static int store_imsm_mpb(int fd, struct imsm_super *mpb) |
c2a1e7da | 7506 | { |
f796af5d | 7507 | void *buf = mpb; |
c2a1e7da DW |
7508 | __u32 mpb_size = __le32_to_cpu(mpb->mpb_size); |
7509 | unsigned long long dsize; | |
7510 | unsigned long long sectors; | |
7511 | ||
7512 | get_dev_size(fd, NULL, &dsize); | |
7513 | ||
272f648f DW |
7514 | if (mpb_size > 512) { |
7515 | /* -1 to account for anchor */ | |
7516 | sectors = mpb_sectors(mpb) - 1; | |
c2a1e7da | 7517 | |
272f648f DW |
7518 | /* write the extended mpb to the sectors preceeding the anchor */ |
7519 | if (lseek64(fd, dsize - (512 * (2 + sectors)), SEEK_SET) < 0) | |
7520 | return 1; | |
c2a1e7da | 7521 | |
f21e18ca N |
7522 | if ((unsigned long long)write(fd, buf + 512, 512 * sectors) |
7523 | != 512 * sectors) | |
272f648f DW |
7524 | return 1; |
7525 | } | |
c2a1e7da | 7526 | |
272f648f DW |
7527 | /* first block is stored on second to last sector of the disk */ |
7528 | if (lseek64(fd, dsize - (512 * 2), SEEK_SET) < 0) | |
c2a1e7da DW |
7529 | return 1; |
7530 | ||
f796af5d | 7531 | if (write(fd, buf, 512) != 512) |
c2a1e7da DW |
7532 | return 1; |
7533 | ||
c2a1e7da DW |
7534 | return 0; |
7535 | } | |
7536 | ||
2e735d19 | 7537 | static void imsm_sync_metadata(struct supertype *container) |
845dea95 | 7538 | { |
2e735d19 | 7539 | struct intel_super *super = container->sb; |
c2a1e7da | 7540 | |
1a64be56 | 7541 | dprintf("sync metadata: %d\n", super->updates_pending); |
c2a1e7da DW |
7542 | if (!super->updates_pending) |
7543 | return; | |
7544 | ||
36988a3d | 7545 | write_super_imsm(container, 0); |
c2a1e7da DW |
7546 | |
7547 | super->updates_pending = 0; | |
845dea95 NB |
7548 | } |
7549 | ||
272906ef DW |
7550 | static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a) |
7551 | { | |
7552 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 7553 | int i = get_imsm_disk_idx(dev, idx, MAP_X); |
272906ef DW |
7554 | struct dl *dl; |
7555 | ||
7556 | for (dl = super->disks; dl; dl = dl->next) | |
7557 | if (dl->index == i) | |
7558 | break; | |
7559 | ||
25ed7e59 | 7560 | if (dl && is_failed(&dl->disk)) |
272906ef DW |
7561 | dl = NULL; |
7562 | ||
7563 | if (dl) | |
1ade5cc1 | 7564 | dprintf("found %x:%x\n", dl->major, dl->minor); |
272906ef DW |
7565 | |
7566 | return dl; | |
7567 | } | |
7568 | ||
a20d2ba5 | 7569 | static struct dl *imsm_add_spare(struct intel_super *super, int slot, |
8ba77d32 AK |
7570 | struct active_array *a, int activate_new, |
7571 | struct mdinfo *additional_test_list) | |
272906ef DW |
7572 | { |
7573 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 7574 | int idx = get_imsm_disk_idx(dev, slot, MAP_X); |
a20d2ba5 DW |
7575 | struct imsm_super *mpb = super->anchor; |
7576 | struct imsm_map *map; | |
272906ef DW |
7577 | unsigned long long pos; |
7578 | struct mdinfo *d; | |
7579 | struct extent *ex; | |
a20d2ba5 | 7580 | int i, j; |
272906ef | 7581 | int found; |
569cc43f DW |
7582 | __u32 array_start = 0; |
7583 | __u32 array_end = 0; | |
272906ef | 7584 | struct dl *dl; |
6c932028 | 7585 | struct mdinfo *test_list; |
272906ef DW |
7586 | |
7587 | for (dl = super->disks; dl; dl = dl->next) { | |
7588 | /* If in this array, skip */ | |
7589 | for (d = a->info.devs ; d ; d = d->next) | |
e553d2a4 DW |
7590 | if (d->state_fd >= 0 && |
7591 | d->disk.major == dl->major && | |
272906ef | 7592 | d->disk.minor == dl->minor) { |
8ba77d32 AK |
7593 | dprintf("%x:%x already in array\n", |
7594 | dl->major, dl->minor); | |
272906ef DW |
7595 | break; |
7596 | } | |
7597 | if (d) | |
7598 | continue; | |
6c932028 AK |
7599 | test_list = additional_test_list; |
7600 | while (test_list) { | |
7601 | if (test_list->disk.major == dl->major && | |
7602 | test_list->disk.minor == dl->minor) { | |
8ba77d32 AK |
7603 | dprintf("%x:%x already in additional test list\n", |
7604 | dl->major, dl->minor); | |
7605 | break; | |
7606 | } | |
6c932028 | 7607 | test_list = test_list->next; |
8ba77d32 | 7608 | } |
6c932028 | 7609 | if (test_list) |
8ba77d32 | 7610 | continue; |
272906ef | 7611 | |
e553d2a4 | 7612 | /* skip in use or failed drives */ |
25ed7e59 | 7613 | if (is_failed(&dl->disk) || idx == dl->index || |
df474657 DW |
7614 | dl->index == -2) { |
7615 | dprintf("%x:%x status (failed: %d index: %d)\n", | |
25ed7e59 | 7616 | dl->major, dl->minor, is_failed(&dl->disk), idx); |
9a1608e5 DW |
7617 | continue; |
7618 | } | |
7619 | ||
a20d2ba5 DW |
7620 | /* skip pure spares when we are looking for partially |
7621 | * assimilated drives | |
7622 | */ | |
7623 | if (dl->index == -1 && !activate_new) | |
7624 | continue; | |
7625 | ||
272906ef | 7626 | /* Does this unused device have the requisite free space? |
a20d2ba5 | 7627 | * It needs to be able to cover all member volumes |
272906ef DW |
7628 | */ |
7629 | ex = get_extents(super, dl); | |
7630 | if (!ex) { | |
7631 | dprintf("cannot get extents\n"); | |
7632 | continue; | |
7633 | } | |
a20d2ba5 DW |
7634 | for (i = 0; i < mpb->num_raid_devs; i++) { |
7635 | dev = get_imsm_dev(super, i); | |
238c0a71 | 7636 | map = get_imsm_map(dev, MAP_0); |
272906ef | 7637 | |
a20d2ba5 DW |
7638 | /* check if this disk is already a member of |
7639 | * this array | |
272906ef | 7640 | */ |
620b1713 | 7641 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
a20d2ba5 DW |
7642 | continue; |
7643 | ||
7644 | found = 0; | |
7645 | j = 0; | |
7646 | pos = 0; | |
5551b113 | 7647 | array_start = pba_of_lba0(map); |
329c8278 | 7648 | array_end = array_start + |
5551b113 | 7649 | blocks_per_member(map) - 1; |
a20d2ba5 DW |
7650 | |
7651 | do { | |
7652 | /* check that we can start at pba_of_lba0 with | |
7653 | * blocks_per_member of space | |
7654 | */ | |
329c8278 | 7655 | if (array_start >= pos && array_end < ex[j].start) { |
a20d2ba5 DW |
7656 | found = 1; |
7657 | break; | |
7658 | } | |
7659 | pos = ex[j].start + ex[j].size; | |
7660 | j++; | |
7661 | } while (ex[j-1].size); | |
7662 | ||
7663 | if (!found) | |
272906ef | 7664 | break; |
a20d2ba5 | 7665 | } |
272906ef DW |
7666 | |
7667 | free(ex); | |
a20d2ba5 | 7668 | if (i < mpb->num_raid_devs) { |
329c8278 DW |
7669 | dprintf("%x:%x does not have %u to %u available\n", |
7670 | dl->major, dl->minor, array_start, array_end); | |
272906ef DW |
7671 | /* No room */ |
7672 | continue; | |
a20d2ba5 DW |
7673 | } |
7674 | return dl; | |
272906ef DW |
7675 | } |
7676 | ||
7677 | return dl; | |
7678 | } | |
7679 | ||
95d07a2c LM |
7680 | static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed) |
7681 | { | |
7682 | struct imsm_dev *dev2; | |
7683 | struct imsm_map *map; | |
7684 | struct dl *idisk; | |
7685 | int slot; | |
7686 | int idx; | |
7687 | __u8 state; | |
7688 | ||
7689 | dev2 = get_imsm_dev(cont->sb, dev_idx); | |
7690 | if (dev2) { | |
238c0a71 | 7691 | state = imsm_check_degraded(cont->sb, dev2, failed, MAP_0); |
95d07a2c | 7692 | if (state == IMSM_T_STATE_FAILED) { |
238c0a71 | 7693 | map = get_imsm_map(dev2, MAP_0); |
95d07a2c LM |
7694 | if (!map) |
7695 | return 1; | |
7696 | for (slot = 0; slot < map->num_members; slot++) { | |
7697 | /* | |
7698 | * Check if failed disks are deleted from intel | |
7699 | * disk list or are marked to be deleted | |
7700 | */ | |
238c0a71 | 7701 | idx = get_imsm_disk_idx(dev2, slot, MAP_X); |
95d07a2c LM |
7702 | idisk = get_imsm_dl_disk(cont->sb, idx); |
7703 | /* | |
7704 | * Do not rebuild the array if failed disks | |
7705 | * from failed sub-array are not removed from | |
7706 | * container. | |
7707 | */ | |
7708 | if (idisk && | |
7709 | is_failed(&idisk->disk) && | |
7710 | (idisk->action != DISK_REMOVE)) | |
7711 | return 0; | |
7712 | } | |
7713 | } | |
7714 | } | |
7715 | return 1; | |
7716 | } | |
7717 | ||
88758e9d DW |
7718 | static struct mdinfo *imsm_activate_spare(struct active_array *a, |
7719 | struct metadata_update **updates) | |
7720 | { | |
7721 | /** | |
d23fe947 DW |
7722 | * Find a device with unused free space and use it to replace a |
7723 | * failed/vacant region in an array. We replace failed regions one a | |
7724 | * array at a time. The result is that a new spare disk will be added | |
7725 | * to the first failed array and after the monitor has finished | |
7726 | * propagating failures the remainder will be consumed. | |
88758e9d | 7727 | * |
d23fe947 DW |
7728 | * FIXME add a capability for mdmon to request spares from another |
7729 | * container. | |
88758e9d DW |
7730 | */ |
7731 | ||
7732 | struct intel_super *super = a->container->sb; | |
88758e9d | 7733 | int inst = a->info.container_member; |
949c47a0 | 7734 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 7735 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
88758e9d DW |
7736 | int failed = a->info.array.raid_disks; |
7737 | struct mdinfo *rv = NULL; | |
7738 | struct mdinfo *d; | |
7739 | struct mdinfo *di; | |
7740 | struct metadata_update *mu; | |
7741 | struct dl *dl; | |
7742 | struct imsm_update_activate_spare *u; | |
7743 | int num_spares = 0; | |
7744 | int i; | |
95d07a2c | 7745 | int allowed; |
88758e9d DW |
7746 | |
7747 | for (d = a->info.devs ; d ; d = d->next) { | |
7748 | if ((d->curr_state & DS_FAULTY) && | |
7749 | d->state_fd >= 0) | |
7750 | /* wait for Removal to happen */ | |
7751 | return NULL; | |
7752 | if (d->state_fd >= 0) | |
7753 | failed--; | |
7754 | } | |
7755 | ||
7756 | dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n", | |
7757 | inst, failed, a->info.array.raid_disks, a->info.array.level); | |
1af97990 | 7758 | |
e2962bfc AK |
7759 | if (imsm_reshape_blocks_arrays_changes(super)) |
7760 | return NULL; | |
1af97990 | 7761 | |
fc8ca064 AK |
7762 | /* Cannot activate another spare if rebuild is in progress already |
7763 | */ | |
7764 | if (is_rebuilding(dev)) { | |
7a862a02 | 7765 | dprintf("imsm: No spare activation allowed. Rebuild in progress already.\n"); |
fc8ca064 AK |
7766 | return NULL; |
7767 | } | |
7768 | ||
89c67882 AK |
7769 | if (a->info.array.level == 4) |
7770 | /* No repair for takeovered array | |
7771 | * imsm doesn't support raid4 | |
7772 | */ | |
7773 | return NULL; | |
7774 | ||
3b451610 AK |
7775 | if (imsm_check_degraded(super, dev, failed, MAP_0) != |
7776 | IMSM_T_STATE_DEGRADED) | |
88758e9d DW |
7777 | return NULL; |
7778 | ||
95d07a2c LM |
7779 | /* |
7780 | * If there are any failed disks check state of the other volume. | |
7781 | * Block rebuild if the another one is failed until failed disks | |
7782 | * are removed from container. | |
7783 | */ | |
7784 | if (failed) { | |
7a862a02 | 7785 | dprintf("found failed disks in %.*s, check if there anotherfailed sub-array.\n", |
c4acd1e5 | 7786 | MAX_RAID_SERIAL_LEN, dev->volume); |
95d07a2c LM |
7787 | /* check if states of the other volumes allow for rebuild */ |
7788 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
7789 | if (i != inst) { | |
7790 | allowed = imsm_rebuild_allowed(a->container, | |
7791 | i, failed); | |
7792 | if (!allowed) | |
7793 | return NULL; | |
7794 | } | |
7795 | } | |
7796 | } | |
7797 | ||
88758e9d | 7798 | /* For each slot, if it is not working, find a spare */ |
88758e9d DW |
7799 | for (i = 0; i < a->info.array.raid_disks; i++) { |
7800 | for (d = a->info.devs ; d ; d = d->next) | |
7801 | if (d->disk.raid_disk == i) | |
7802 | break; | |
7803 | dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0); | |
7804 | if (d && (d->state_fd >= 0)) | |
7805 | continue; | |
7806 | ||
272906ef | 7807 | /* |
a20d2ba5 DW |
7808 | * OK, this device needs recovery. Try to re-add the |
7809 | * previous occupant of this slot, if this fails see if | |
7810 | * we can continue the assimilation of a spare that was | |
7811 | * partially assimilated, finally try to activate a new | |
7812 | * spare. | |
272906ef DW |
7813 | */ |
7814 | dl = imsm_readd(super, i, a); | |
7815 | if (!dl) | |
b303fe21 | 7816 | dl = imsm_add_spare(super, i, a, 0, rv); |
a20d2ba5 | 7817 | if (!dl) |
b303fe21 | 7818 | dl = imsm_add_spare(super, i, a, 1, rv); |
272906ef DW |
7819 | if (!dl) |
7820 | continue; | |
1011e834 | 7821 | |
272906ef | 7822 | /* found a usable disk with enough space */ |
503975b9 | 7823 | di = xcalloc(1, sizeof(*di)); |
272906ef DW |
7824 | |
7825 | /* dl->index will be -1 in the case we are activating a | |
7826 | * pristine spare. imsm_process_update() will create a | |
7827 | * new index in this case. Once a disk is found to be | |
7828 | * failed in all member arrays it is kicked from the | |
7829 | * metadata | |
7830 | */ | |
7831 | di->disk.number = dl->index; | |
d23fe947 | 7832 | |
272906ef DW |
7833 | /* (ab)use di->devs to store a pointer to the device |
7834 | * we chose | |
7835 | */ | |
7836 | di->devs = (struct mdinfo *) dl; | |
7837 | ||
7838 | di->disk.raid_disk = i; | |
7839 | di->disk.major = dl->major; | |
7840 | di->disk.minor = dl->minor; | |
7841 | di->disk.state = 0; | |
d23534e4 | 7842 | di->recovery_start = 0; |
5551b113 | 7843 | di->data_offset = pba_of_lba0(map); |
272906ef DW |
7844 | di->component_size = a->info.component_size; |
7845 | di->container_member = inst; | |
148acb7b | 7846 | super->random = random32(); |
272906ef DW |
7847 | di->next = rv; |
7848 | rv = di; | |
7849 | num_spares++; | |
7850 | dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor, | |
7851 | i, di->data_offset); | |
88758e9d DW |
7852 | } |
7853 | ||
7854 | if (!rv) | |
7855 | /* No spares found */ | |
7856 | return rv; | |
7857 | /* Now 'rv' has a list of devices to return. | |
7858 | * Create a metadata_update record to update the | |
7859 | * disk_ord_tbl for the array | |
7860 | */ | |
503975b9 | 7861 | mu = xmalloc(sizeof(*mu)); |
1011e834 | 7862 | mu->buf = xcalloc(num_spares, |
503975b9 | 7863 | sizeof(struct imsm_update_activate_spare)); |
88758e9d | 7864 | mu->space = NULL; |
cb23f1f4 | 7865 | mu->space_list = NULL; |
88758e9d DW |
7866 | mu->len = sizeof(struct imsm_update_activate_spare) * num_spares; |
7867 | mu->next = *updates; | |
7868 | u = (struct imsm_update_activate_spare *) mu->buf; | |
7869 | ||
7870 | for (di = rv ; di ; di = di->next) { | |
7871 | u->type = update_activate_spare; | |
d23fe947 DW |
7872 | u->dl = (struct dl *) di->devs; |
7873 | di->devs = NULL; | |
88758e9d DW |
7874 | u->slot = di->disk.raid_disk; |
7875 | u->array = inst; | |
7876 | u->next = u + 1; | |
7877 | u++; | |
7878 | } | |
7879 | (u-1)->next = NULL; | |
7880 | *updates = mu; | |
7881 | ||
7882 | return rv; | |
7883 | } | |
7884 | ||
54c2c1ea | 7885 | static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u) |
8273f55e | 7886 | { |
54c2c1ea | 7887 | struct imsm_dev *dev = get_imsm_dev(super, idx); |
238c0a71 AK |
7888 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
7889 | struct imsm_map *new_map = get_imsm_map(&u->dev, MAP_0); | |
54c2c1ea DW |
7890 | struct disk_info *inf = get_disk_info(u); |
7891 | struct imsm_disk *disk; | |
8273f55e DW |
7892 | int i; |
7893 | int j; | |
8273f55e | 7894 | |
54c2c1ea | 7895 | for (i = 0; i < map->num_members; i++) { |
238c0a71 | 7896 | disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, MAP_X)); |
54c2c1ea DW |
7897 | for (j = 0; j < new_map->num_members; j++) |
7898 | if (serialcmp(disk->serial, inf[j].serial) == 0) | |
8273f55e DW |
7899 | return 1; |
7900 | } | |
7901 | ||
7902 | return 0; | |
7903 | } | |
7904 | ||
1a64be56 LM |
7905 | static struct dl *get_disk_super(struct intel_super *super, int major, int minor) |
7906 | { | |
7907 | struct dl *dl = NULL; | |
7908 | for (dl = super->disks; dl; dl = dl->next) | |
7909 | if ((dl->major == major) && (dl->minor == minor)) | |
7910 | return dl; | |
7911 | return NULL; | |
7912 | } | |
7913 | ||
7914 | static int remove_disk_super(struct intel_super *super, int major, int minor) | |
7915 | { | |
7916 | struct dl *prev = NULL; | |
7917 | struct dl *dl; | |
7918 | ||
7919 | prev = NULL; | |
7920 | for (dl = super->disks; dl; dl = dl->next) { | |
7921 | if ((dl->major == major) && (dl->minor == minor)) { | |
7922 | /* remove */ | |
7923 | if (prev) | |
7924 | prev->next = dl->next; | |
7925 | else | |
7926 | super->disks = dl->next; | |
7927 | dl->next = NULL; | |
7928 | __free_imsm_disk(dl); | |
1ade5cc1 | 7929 | dprintf("removed %x:%x\n", major, minor); |
1a64be56 LM |
7930 | break; |
7931 | } | |
7932 | prev = dl; | |
7933 | } | |
7934 | return 0; | |
7935 | } | |
7936 | ||
f21e18ca | 7937 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index); |
ae6aad82 | 7938 | |
1a64be56 LM |
7939 | static int add_remove_disk_update(struct intel_super *super) |
7940 | { | |
7941 | int check_degraded = 0; | |
7942 | struct dl *disk = NULL; | |
7943 | /* add/remove some spares to/from the metadata/contrainer */ | |
7944 | while (super->disk_mgmt_list) { | |
7945 | struct dl *disk_cfg; | |
7946 | ||
7947 | disk_cfg = super->disk_mgmt_list; | |
7948 | super->disk_mgmt_list = disk_cfg->next; | |
7949 | disk_cfg->next = NULL; | |
7950 | ||
7951 | if (disk_cfg->action == DISK_ADD) { | |
7952 | disk_cfg->next = super->disks; | |
7953 | super->disks = disk_cfg; | |
7954 | check_degraded = 1; | |
1ade5cc1 N |
7955 | dprintf("added %x:%x\n", |
7956 | disk_cfg->major, disk_cfg->minor); | |
1a64be56 LM |
7957 | } else if (disk_cfg->action == DISK_REMOVE) { |
7958 | dprintf("Disk remove action processed: %x.%x\n", | |
7959 | disk_cfg->major, disk_cfg->minor); | |
7960 | disk = get_disk_super(super, | |
7961 | disk_cfg->major, | |
7962 | disk_cfg->minor); | |
7963 | if (disk) { | |
7964 | /* store action status */ | |
7965 | disk->action = DISK_REMOVE; | |
7966 | /* remove spare disks only */ | |
7967 | if (disk->index == -1) { | |
7968 | remove_disk_super(super, | |
7969 | disk_cfg->major, | |
7970 | disk_cfg->minor); | |
7971 | } | |
7972 | } | |
7973 | /* release allocate disk structure */ | |
7974 | __free_imsm_disk(disk_cfg); | |
7975 | } | |
7976 | } | |
7977 | return check_degraded; | |
7978 | } | |
7979 | ||
a29911da PC |
7980 | static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u, |
7981 | struct intel_super *super, | |
7982 | void ***space_list) | |
7983 | { | |
7984 | struct intel_dev *id; | |
7985 | void **tofree = NULL; | |
7986 | int ret_val = 0; | |
7987 | ||
1ade5cc1 | 7988 | dprintf("(enter)\n"); |
a29911da PC |
7989 | if ((u->subdev < 0) || |
7990 | (u->subdev > 1)) { | |
7991 | dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev); | |
7992 | return ret_val; | |
7993 | } | |
7994 | if ((space_list == NULL) || (*space_list == NULL)) { | |
7995 | dprintf("imsm: Error: Memory is not allocated\n"); | |
7996 | return ret_val; | |
7997 | } | |
7998 | ||
7999 | for (id = super->devlist ; id; id = id->next) { | |
8000 | if (id->index == (unsigned)u->subdev) { | |
8001 | struct imsm_dev *dev = get_imsm_dev(super, u->subdev); | |
8002 | struct imsm_map *map; | |
8003 | struct imsm_dev *new_dev = | |
8004 | (struct imsm_dev *)*space_list; | |
238c0a71 | 8005 | struct imsm_map *migr_map = get_imsm_map(dev, MAP_1); |
a29911da PC |
8006 | int to_state; |
8007 | struct dl *new_disk; | |
8008 | ||
8009 | if (new_dev == NULL) | |
8010 | return ret_val; | |
8011 | *space_list = **space_list; | |
8012 | memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0)); | |
238c0a71 | 8013 | map = get_imsm_map(new_dev, MAP_0); |
a29911da PC |
8014 | if (migr_map) { |
8015 | dprintf("imsm: Error: migration in progress"); | |
8016 | return ret_val; | |
8017 | } | |
8018 | ||
8019 | to_state = map->map_state; | |
8020 | if ((u->new_level == 5) && (map->raid_level == 0)) { | |
8021 | map->num_members++; | |
8022 | /* this should not happen */ | |
8023 | if (u->new_disks[0] < 0) { | |
8024 | map->failed_disk_num = | |
8025 | map->num_members - 1; | |
8026 | to_state = IMSM_T_STATE_DEGRADED; | |
8027 | } else | |
8028 | to_state = IMSM_T_STATE_NORMAL; | |
8029 | } | |
8e59f3d8 | 8030 | migrate(new_dev, super, to_state, MIGR_GEN_MIGR); |
a29911da PC |
8031 | if (u->new_level > -1) |
8032 | map->raid_level = u->new_level; | |
238c0a71 | 8033 | migr_map = get_imsm_map(new_dev, MAP_1); |
a29911da PC |
8034 | if ((u->new_level == 5) && |
8035 | (migr_map->raid_level == 0)) { | |
8036 | int ord = map->num_members - 1; | |
8037 | migr_map->num_members--; | |
8038 | if (u->new_disks[0] < 0) | |
8039 | ord |= IMSM_ORD_REBUILD; | |
8040 | set_imsm_ord_tbl_ent(map, | |
8041 | map->num_members - 1, | |
8042 | ord); | |
8043 | } | |
8044 | id->dev = new_dev; | |
8045 | tofree = (void **)dev; | |
8046 | ||
4bba0439 PC |
8047 | /* update chunk size |
8048 | */ | |
8049 | if (u->new_chunksize > 0) | |
8050 | map->blocks_per_strip = | |
8051 | __cpu_to_le16(u->new_chunksize * 2); | |
8052 | ||
a29911da PC |
8053 | /* add disk |
8054 | */ | |
8055 | if ((u->new_level != 5) || | |
8056 | (migr_map->raid_level != 0) || | |
8057 | (migr_map->raid_level == map->raid_level)) | |
8058 | goto skip_disk_add; | |
8059 | ||
8060 | if (u->new_disks[0] >= 0) { | |
8061 | /* use passes spare | |
8062 | */ | |
8063 | new_disk = get_disk_super(super, | |
8064 | major(u->new_disks[0]), | |
8065 | minor(u->new_disks[0])); | |
7a862a02 | 8066 | dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n", |
a29911da PC |
8067 | major(u->new_disks[0]), |
8068 | minor(u->new_disks[0]), | |
8069 | new_disk, new_disk->index); | |
8070 | if (new_disk == NULL) | |
8071 | goto error_disk_add; | |
8072 | ||
8073 | new_disk->index = map->num_members - 1; | |
8074 | /* slot to fill in autolayout | |
8075 | */ | |
8076 | new_disk->raiddisk = new_disk->index; | |
8077 | new_disk->disk.status |= CONFIGURED_DISK; | |
8078 | new_disk->disk.status &= ~SPARE_DISK; | |
8079 | } else | |
8080 | goto error_disk_add; | |
8081 | ||
8082 | skip_disk_add: | |
8083 | *tofree = *space_list; | |
8084 | /* calculate new size | |
8085 | */ | |
f3871fdc | 8086 | imsm_set_array_size(new_dev, -1); |
a29911da PC |
8087 | |
8088 | ret_val = 1; | |
8089 | } | |
8090 | } | |
8091 | ||
8092 | if (tofree) | |
8093 | *space_list = tofree; | |
8094 | return ret_val; | |
8095 | ||
8096 | error_disk_add: | |
8097 | dprintf("Error: imsm: Cannot find disk.\n"); | |
8098 | return ret_val; | |
8099 | } | |
8100 | ||
f3871fdc AK |
8101 | static int apply_size_change_update(struct imsm_update_size_change *u, |
8102 | struct intel_super *super) | |
8103 | { | |
8104 | struct intel_dev *id; | |
8105 | int ret_val = 0; | |
8106 | ||
1ade5cc1 | 8107 | dprintf("(enter)\n"); |
f3871fdc AK |
8108 | if ((u->subdev < 0) || |
8109 | (u->subdev > 1)) { | |
8110 | dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev); | |
8111 | return ret_val; | |
8112 | } | |
8113 | ||
8114 | for (id = super->devlist ; id; id = id->next) { | |
8115 | if (id->index == (unsigned)u->subdev) { | |
8116 | struct imsm_dev *dev = get_imsm_dev(super, u->subdev); | |
8117 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
8118 | int used_disks = imsm_num_data_members(dev, MAP_0); | |
8119 | unsigned long long blocks_per_member; | |
8120 | ||
8121 | /* calculate new size | |
8122 | */ | |
8123 | blocks_per_member = u->new_size / used_disks; | |
1ade5cc1 | 8124 | dprintf("(size: %llu, blocks per member: %llu)\n", |
f3871fdc AK |
8125 | u->new_size, blocks_per_member); |
8126 | set_blocks_per_member(map, blocks_per_member); | |
8127 | imsm_set_array_size(dev, u->new_size); | |
8128 | ||
8129 | ret_val = 1; | |
8130 | break; | |
8131 | } | |
8132 | } | |
8133 | ||
8134 | return ret_val; | |
8135 | } | |
8136 | ||
061d7da3 | 8137 | static int apply_update_activate_spare(struct imsm_update_activate_spare *u, |
ca9de185 | 8138 | struct intel_super *super, |
061d7da3 LO |
8139 | struct active_array *active_array) |
8140 | { | |
8141 | struct imsm_super *mpb = super->anchor; | |
8142 | struct imsm_dev *dev = get_imsm_dev(super, u->array); | |
238c0a71 | 8143 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
061d7da3 LO |
8144 | struct imsm_map *migr_map; |
8145 | struct active_array *a; | |
8146 | struct imsm_disk *disk; | |
8147 | __u8 to_state; | |
8148 | struct dl *dl; | |
8149 | unsigned int found; | |
8150 | int failed; | |
5961eeec | 8151 | int victim; |
061d7da3 | 8152 | int i; |
5961eeec | 8153 | int second_map_created = 0; |
061d7da3 | 8154 | |
5961eeec | 8155 | for (; u; u = u->next) { |
238c0a71 | 8156 | victim = get_imsm_disk_idx(dev, u->slot, MAP_X); |
061d7da3 | 8157 | |
5961eeec | 8158 | if (victim < 0) |
8159 | return 0; | |
061d7da3 | 8160 | |
5961eeec | 8161 | for (dl = super->disks; dl; dl = dl->next) |
8162 | if (dl == u->dl) | |
8163 | break; | |
061d7da3 | 8164 | |
5961eeec | 8165 | if (!dl) { |
7a862a02 | 8166 | pr_err("error: imsm_activate_spare passed an unknown disk (index: %d)\n", |
5961eeec | 8167 | u->dl->index); |
8168 | return 0; | |
8169 | } | |
061d7da3 | 8170 | |
5961eeec | 8171 | /* count failures (excluding rebuilds and the victim) |
8172 | * to determine map[0] state | |
8173 | */ | |
8174 | failed = 0; | |
8175 | for (i = 0; i < map->num_members; i++) { | |
8176 | if (i == u->slot) | |
8177 | continue; | |
8178 | disk = get_imsm_disk(super, | |
238c0a71 | 8179 | get_imsm_disk_idx(dev, i, MAP_X)); |
5961eeec | 8180 | if (!disk || is_failed(disk)) |
8181 | failed++; | |
8182 | } | |
061d7da3 | 8183 | |
5961eeec | 8184 | /* adding a pristine spare, assign a new index */ |
8185 | if (dl->index < 0) { | |
8186 | dl->index = super->anchor->num_disks; | |
8187 | super->anchor->num_disks++; | |
8188 | } | |
8189 | disk = &dl->disk; | |
8190 | disk->status |= CONFIGURED_DISK; | |
8191 | disk->status &= ~SPARE_DISK; | |
8192 | ||
8193 | /* mark rebuild */ | |
238c0a71 | 8194 | to_state = imsm_check_degraded(super, dev, failed, MAP_0); |
5961eeec | 8195 | if (!second_map_created) { |
8196 | second_map_created = 1; | |
8197 | map->map_state = IMSM_T_STATE_DEGRADED; | |
8198 | migrate(dev, super, to_state, MIGR_REBUILD); | |
8199 | } else | |
8200 | map->map_state = to_state; | |
238c0a71 | 8201 | migr_map = get_imsm_map(dev, MAP_1); |
5961eeec | 8202 | set_imsm_ord_tbl_ent(map, u->slot, dl->index); |
8203 | set_imsm_ord_tbl_ent(migr_map, u->slot, | |
8204 | dl->index | IMSM_ORD_REBUILD); | |
8205 | ||
8206 | /* update the family_num to mark a new container | |
8207 | * generation, being careful to record the existing | |
8208 | * family_num in orig_family_num to clean up after | |
8209 | * earlier mdadm versions that neglected to set it. | |
8210 | */ | |
8211 | if (mpb->orig_family_num == 0) | |
8212 | mpb->orig_family_num = mpb->family_num; | |
8213 | mpb->family_num += super->random; | |
8214 | ||
8215 | /* count arrays using the victim in the metadata */ | |
8216 | found = 0; | |
8217 | for (a = active_array; a ; a = a->next) { | |
8218 | dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 8219 | map = get_imsm_map(dev, MAP_0); |
061d7da3 | 8220 | |
5961eeec | 8221 | if (get_imsm_disk_slot(map, victim) >= 0) |
8222 | found++; | |
8223 | } | |
061d7da3 | 8224 | |
5961eeec | 8225 | /* delete the victim if it is no longer being |
8226 | * utilized anywhere | |
061d7da3 | 8227 | */ |
5961eeec | 8228 | if (!found) { |
8229 | struct dl **dlp; | |
061d7da3 | 8230 | |
5961eeec | 8231 | /* We know that 'manager' isn't touching anything, |
8232 | * so it is safe to delete | |
8233 | */ | |
8234 | for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next) | |
061d7da3 LO |
8235 | if ((*dlp)->index == victim) |
8236 | break; | |
5961eeec | 8237 | |
8238 | /* victim may be on the missing list */ | |
8239 | if (!*dlp) | |
8240 | for (dlp = &super->missing; *dlp; | |
8241 | dlp = &(*dlp)->next) | |
8242 | if ((*dlp)->index == victim) | |
8243 | break; | |
8244 | imsm_delete(super, dlp, victim); | |
8245 | } | |
061d7da3 LO |
8246 | } |
8247 | ||
8248 | return 1; | |
8249 | } | |
a29911da | 8250 | |
2e5dc010 N |
8251 | static int apply_reshape_container_disks_update(struct imsm_update_reshape *u, |
8252 | struct intel_super *super, | |
8253 | void ***space_list) | |
8254 | { | |
8255 | struct dl *new_disk; | |
8256 | struct intel_dev *id; | |
8257 | int i; | |
8258 | int delta_disks = u->new_raid_disks - u->old_raid_disks; | |
ee4beede | 8259 | int disk_count = u->old_raid_disks; |
2e5dc010 N |
8260 | void **tofree = NULL; |
8261 | int devices_to_reshape = 1; | |
8262 | struct imsm_super *mpb = super->anchor; | |
8263 | int ret_val = 0; | |
d098291a | 8264 | unsigned int dev_id; |
2e5dc010 | 8265 | |
1ade5cc1 | 8266 | dprintf("(enter)\n"); |
2e5dc010 N |
8267 | |
8268 | /* enable spares to use in array */ | |
8269 | for (i = 0; i < delta_disks; i++) { | |
8270 | new_disk = get_disk_super(super, | |
8271 | major(u->new_disks[i]), | |
8272 | minor(u->new_disks[i])); | |
7a862a02 | 8273 | dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n", |
2e5dc010 N |
8274 | major(u->new_disks[i]), minor(u->new_disks[i]), |
8275 | new_disk, new_disk->index); | |
8276 | if ((new_disk == NULL) || | |
8277 | ((new_disk->index >= 0) && | |
8278 | (new_disk->index < u->old_raid_disks))) | |
8279 | goto update_reshape_exit; | |
ee4beede | 8280 | new_disk->index = disk_count++; |
2e5dc010 N |
8281 | /* slot to fill in autolayout |
8282 | */ | |
8283 | new_disk->raiddisk = new_disk->index; | |
8284 | new_disk->disk.status |= | |
8285 | CONFIGURED_DISK; | |
8286 | new_disk->disk.status &= ~SPARE_DISK; | |
8287 | } | |
8288 | ||
ed7333bd AK |
8289 | dprintf("imsm: volume set mpb->num_raid_devs = %i\n", |
8290 | mpb->num_raid_devs); | |
2e5dc010 N |
8291 | /* manage changes in volume |
8292 | */ | |
d098291a | 8293 | for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) { |
2e5dc010 N |
8294 | void **sp = *space_list; |
8295 | struct imsm_dev *newdev; | |
8296 | struct imsm_map *newmap, *oldmap; | |
8297 | ||
d098291a AK |
8298 | for (id = super->devlist ; id; id = id->next) { |
8299 | if (id->index == dev_id) | |
8300 | break; | |
8301 | } | |
8302 | if (id == NULL) | |
8303 | break; | |
2e5dc010 N |
8304 | if (!sp) |
8305 | continue; | |
8306 | *space_list = *sp; | |
8307 | newdev = (void*)sp; | |
8308 | /* Copy the dev, but not (all of) the map */ | |
8309 | memcpy(newdev, id->dev, sizeof(*newdev)); | |
238c0a71 AK |
8310 | oldmap = get_imsm_map(id->dev, MAP_0); |
8311 | newmap = get_imsm_map(newdev, MAP_0); | |
2e5dc010 N |
8312 | /* Copy the current map */ |
8313 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); | |
8314 | /* update one device only | |
8315 | */ | |
8316 | if (devices_to_reshape) { | |
ed7333bd AK |
8317 | dprintf("imsm: modifying subdev: %i\n", |
8318 | id->index); | |
2e5dc010 N |
8319 | devices_to_reshape--; |
8320 | newdev->vol.migr_state = 1; | |
8321 | newdev->vol.curr_migr_unit = 0; | |
ea672ee1 | 8322 | set_migr_type(newdev, MIGR_GEN_MIGR); |
2e5dc010 N |
8323 | newmap->num_members = u->new_raid_disks; |
8324 | for (i = 0; i < delta_disks; i++) { | |
8325 | set_imsm_ord_tbl_ent(newmap, | |
8326 | u->old_raid_disks + i, | |
8327 | u->old_raid_disks + i); | |
8328 | } | |
8329 | /* New map is correct, now need to save old map | |
8330 | */ | |
238c0a71 | 8331 | newmap = get_imsm_map(newdev, MAP_1); |
2e5dc010 N |
8332 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); |
8333 | ||
f3871fdc | 8334 | imsm_set_array_size(newdev, -1); |
2e5dc010 N |
8335 | } |
8336 | ||
8337 | sp = (void **)id->dev; | |
8338 | id->dev = newdev; | |
8339 | *sp = tofree; | |
8340 | tofree = sp; | |
8e59f3d8 AK |
8341 | |
8342 | /* Clear migration record */ | |
8343 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
2e5dc010 | 8344 | } |
819bc634 AK |
8345 | if (tofree) |
8346 | *space_list = tofree; | |
2e5dc010 N |
8347 | ret_val = 1; |
8348 | ||
8349 | update_reshape_exit: | |
8350 | ||
8351 | return ret_val; | |
8352 | } | |
8353 | ||
bb025c2f | 8354 | static int apply_takeover_update(struct imsm_update_takeover *u, |
8ca6df95 KW |
8355 | struct intel_super *super, |
8356 | void ***space_list) | |
bb025c2f KW |
8357 | { |
8358 | struct imsm_dev *dev = NULL; | |
8ca6df95 KW |
8359 | struct intel_dev *dv; |
8360 | struct imsm_dev *dev_new; | |
bb025c2f KW |
8361 | struct imsm_map *map; |
8362 | struct dl *dm, *du; | |
8ca6df95 | 8363 | int i; |
bb025c2f KW |
8364 | |
8365 | for (dv = super->devlist; dv; dv = dv->next) | |
8366 | if (dv->index == (unsigned int)u->subarray) { | |
8367 | dev = dv->dev; | |
8368 | break; | |
8369 | } | |
8370 | ||
8371 | if (dev == NULL) | |
8372 | return 0; | |
8373 | ||
238c0a71 | 8374 | map = get_imsm_map(dev, MAP_0); |
bb025c2f KW |
8375 | |
8376 | if (u->direction == R10_TO_R0) { | |
43d5ec18 | 8377 | /* Number of failed disks must be half of initial disk number */ |
3b451610 AK |
8378 | if (imsm_count_failed(super, dev, MAP_0) != |
8379 | (map->num_members / 2)) | |
43d5ec18 KW |
8380 | return 0; |
8381 | ||
bb025c2f KW |
8382 | /* iterate through devices to mark removed disks as spare */ |
8383 | for (dm = super->disks; dm; dm = dm->next) { | |
8384 | if (dm->disk.status & FAILED_DISK) { | |
8385 | int idx = dm->index; | |
8386 | /* update indexes on the disk list */ | |
8387 | /* FIXME this loop-with-the-loop looks wrong, I'm not convinced | |
8388 | the index values will end up being correct.... NB */ | |
8389 | for (du = super->disks; du; du = du->next) | |
8390 | if (du->index > idx) | |
8391 | du->index--; | |
8392 | /* mark as spare disk */ | |
a8619d23 | 8393 | mark_spare(dm); |
bb025c2f KW |
8394 | } |
8395 | } | |
bb025c2f KW |
8396 | /* update map */ |
8397 | map->num_members = map->num_members / 2; | |
8398 | map->map_state = IMSM_T_STATE_NORMAL; | |
8399 | map->num_domains = 1; | |
8400 | map->raid_level = 0; | |
8401 | map->failed_disk_num = -1; | |
8402 | } | |
8403 | ||
8ca6df95 KW |
8404 | if (u->direction == R0_TO_R10) { |
8405 | void **space; | |
8406 | /* update slots in current disk list */ | |
8407 | for (dm = super->disks; dm; dm = dm->next) { | |
8408 | if (dm->index >= 0) | |
8409 | dm->index *= 2; | |
8410 | } | |
8411 | /* create new *missing* disks */ | |
8412 | for (i = 0; i < map->num_members; i++) { | |
8413 | space = *space_list; | |
8414 | if (!space) | |
8415 | continue; | |
8416 | *space_list = *space; | |
8417 | du = (void *)space; | |
8418 | memcpy(du, super->disks, sizeof(*du)); | |
8ca6df95 KW |
8419 | du->fd = -1; |
8420 | du->minor = 0; | |
8421 | du->major = 0; | |
8422 | du->index = (i * 2) + 1; | |
8423 | sprintf((char *)du->disk.serial, | |
8424 | " MISSING_%d", du->index); | |
8425 | sprintf((char *)du->serial, | |
8426 | "MISSING_%d", du->index); | |
8427 | du->next = super->missing; | |
8428 | super->missing = du; | |
8429 | } | |
8430 | /* create new dev and map */ | |
8431 | space = *space_list; | |
8432 | if (!space) | |
8433 | return 0; | |
8434 | *space_list = *space; | |
8435 | dev_new = (void *)space; | |
8436 | memcpy(dev_new, dev, sizeof(*dev)); | |
8437 | /* update new map */ | |
238c0a71 | 8438 | map = get_imsm_map(dev_new, MAP_0); |
8ca6df95 | 8439 | map->num_members = map->num_members * 2; |
1a2487c2 | 8440 | map->map_state = IMSM_T_STATE_DEGRADED; |
8ca6df95 KW |
8441 | map->num_domains = 2; |
8442 | map->raid_level = 1; | |
8443 | /* replace dev<->dev_new */ | |
8444 | dv->dev = dev_new; | |
8445 | } | |
bb025c2f KW |
8446 | /* update disk order table */ |
8447 | for (du = super->disks; du; du = du->next) | |
8448 | if (du->index >= 0) | |
8449 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
8ca6df95 | 8450 | for (du = super->missing; du; du = du->next) |
1a2487c2 KW |
8451 | if (du->index >= 0) { |
8452 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
e4c72d1d | 8453 | mark_missing(dv->dev, &du->disk, du->index); |
1a2487c2 | 8454 | } |
bb025c2f KW |
8455 | |
8456 | return 1; | |
8457 | } | |
8458 | ||
e8319a19 DW |
8459 | static void imsm_process_update(struct supertype *st, |
8460 | struct metadata_update *update) | |
8461 | { | |
8462 | /** | |
8463 | * crack open the metadata_update envelope to find the update record | |
8464 | * update can be one of: | |
d195167d AK |
8465 | * update_reshape_container_disks - all the arrays in the container |
8466 | * are being reshaped to have more devices. We need to mark | |
8467 | * the arrays for general migration and convert selected spares | |
8468 | * into active devices. | |
8469 | * update_activate_spare - a spare device has replaced a failed | |
1011e834 N |
8470 | * device in an array, update the disk_ord_tbl. If this disk is |
8471 | * present in all member arrays then also clear the SPARE_DISK | |
8472 | * flag | |
d195167d AK |
8473 | * update_create_array |
8474 | * update_kill_array | |
8475 | * update_rename_array | |
8476 | * update_add_remove_disk | |
e8319a19 DW |
8477 | */ |
8478 | struct intel_super *super = st->sb; | |
4d7b1503 | 8479 | struct imsm_super *mpb; |
e8319a19 DW |
8480 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
8481 | ||
4d7b1503 DW |
8482 | /* update requires a larger buf but the allocation failed */ |
8483 | if (super->next_len && !super->next_buf) { | |
8484 | super->next_len = 0; | |
8485 | return; | |
8486 | } | |
8487 | ||
8488 | if (super->next_buf) { | |
8489 | memcpy(super->next_buf, super->buf, super->len); | |
8490 | free(super->buf); | |
8491 | super->len = super->next_len; | |
8492 | super->buf = super->next_buf; | |
8493 | ||
8494 | super->next_len = 0; | |
8495 | super->next_buf = NULL; | |
8496 | } | |
8497 | ||
8498 | mpb = super->anchor; | |
8499 | ||
e8319a19 | 8500 | switch (type) { |
0ec5d470 AK |
8501 | case update_general_migration_checkpoint: { |
8502 | struct intel_dev *id; | |
8503 | struct imsm_update_general_migration_checkpoint *u = | |
8504 | (void *)update->buf; | |
8505 | ||
1ade5cc1 | 8506 | dprintf("called for update_general_migration_checkpoint\n"); |
0ec5d470 AK |
8507 | |
8508 | /* find device under general migration */ | |
8509 | for (id = super->devlist ; id; id = id->next) { | |
8510 | if (is_gen_migration(id->dev)) { | |
8511 | id->dev->vol.curr_migr_unit = | |
8512 | __cpu_to_le32(u->curr_migr_unit); | |
8513 | super->updates_pending++; | |
8514 | } | |
8515 | } | |
8516 | break; | |
8517 | } | |
bb025c2f KW |
8518 | case update_takeover: { |
8519 | struct imsm_update_takeover *u = (void *)update->buf; | |
1a2487c2 KW |
8520 | if (apply_takeover_update(u, super, &update->space_list)) { |
8521 | imsm_update_version_info(super); | |
bb025c2f | 8522 | super->updates_pending++; |
1a2487c2 | 8523 | } |
bb025c2f KW |
8524 | break; |
8525 | } | |
8526 | ||
78b10e66 | 8527 | case update_reshape_container_disks: { |
d195167d | 8528 | struct imsm_update_reshape *u = (void *)update->buf; |
2e5dc010 N |
8529 | if (apply_reshape_container_disks_update( |
8530 | u, super, &update->space_list)) | |
8531 | super->updates_pending++; | |
78b10e66 N |
8532 | break; |
8533 | } | |
48c5303a | 8534 | case update_reshape_migration: { |
a29911da PC |
8535 | struct imsm_update_reshape_migration *u = (void *)update->buf; |
8536 | if (apply_reshape_migration_update( | |
8537 | u, super, &update->space_list)) | |
8538 | super->updates_pending++; | |
48c5303a PC |
8539 | break; |
8540 | } | |
f3871fdc AK |
8541 | case update_size_change: { |
8542 | struct imsm_update_size_change *u = (void *)update->buf; | |
8543 | if (apply_size_change_update(u, super)) | |
8544 | super->updates_pending++; | |
8545 | break; | |
8546 | } | |
e8319a19 | 8547 | case update_activate_spare: { |
1011e834 | 8548 | struct imsm_update_activate_spare *u = (void *) update->buf; |
061d7da3 LO |
8549 | if (apply_update_activate_spare(u, super, st->arrays)) |
8550 | super->updates_pending++; | |
8273f55e DW |
8551 | break; |
8552 | } | |
8553 | case update_create_array: { | |
8554 | /* someone wants to create a new array, we need to be aware of | |
8555 | * a few races/collisions: | |
8556 | * 1/ 'Create' called by two separate instances of mdadm | |
8557 | * 2/ 'Create' versus 'activate_spare': mdadm has chosen | |
8558 | * devices that have since been assimilated via | |
8559 | * activate_spare. | |
8560 | * In the event this update can not be carried out mdadm will | |
8561 | * (FIX ME) notice that its update did not take hold. | |
8562 | */ | |
8563 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 8564 | struct intel_dev *dv; |
8273f55e DW |
8565 | struct imsm_dev *dev; |
8566 | struct imsm_map *map, *new_map; | |
8567 | unsigned long long start, end; | |
8568 | unsigned long long new_start, new_end; | |
8569 | int i; | |
54c2c1ea DW |
8570 | struct disk_info *inf; |
8571 | struct dl *dl; | |
8273f55e DW |
8572 | |
8573 | /* handle racing creates: first come first serve */ | |
8574 | if (u->dev_idx < mpb->num_raid_devs) { | |
1ade5cc1 | 8575 | dprintf("subarray %d already defined\n", u->dev_idx); |
ba2de7ba | 8576 | goto create_error; |
8273f55e DW |
8577 | } |
8578 | ||
8579 | /* check update is next in sequence */ | |
8580 | if (u->dev_idx != mpb->num_raid_devs) { | |
1ade5cc1 N |
8581 | dprintf("can not create array %d expected index %d\n", |
8582 | u->dev_idx, mpb->num_raid_devs); | |
ba2de7ba | 8583 | goto create_error; |
8273f55e DW |
8584 | } |
8585 | ||
238c0a71 | 8586 | new_map = get_imsm_map(&u->dev, MAP_0); |
5551b113 CA |
8587 | new_start = pba_of_lba0(new_map); |
8588 | new_end = new_start + blocks_per_member(new_map); | |
54c2c1ea | 8589 | inf = get_disk_info(u); |
8273f55e DW |
8590 | |
8591 | /* handle activate_spare versus create race: | |
8592 | * check to make sure that overlapping arrays do not include | |
8593 | * overalpping disks | |
8594 | */ | |
8595 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
949c47a0 | 8596 | dev = get_imsm_dev(super, i); |
238c0a71 | 8597 | map = get_imsm_map(dev, MAP_0); |
5551b113 CA |
8598 | start = pba_of_lba0(map); |
8599 | end = start + blocks_per_member(map); | |
8273f55e DW |
8600 | if ((new_start >= start && new_start <= end) || |
8601 | (start >= new_start && start <= new_end)) | |
54c2c1ea DW |
8602 | /* overlap */; |
8603 | else | |
8604 | continue; | |
8605 | ||
8606 | if (disks_overlap(super, i, u)) { | |
1ade5cc1 | 8607 | dprintf("arrays overlap\n"); |
ba2de7ba | 8608 | goto create_error; |
8273f55e DW |
8609 | } |
8610 | } | |
8273f55e | 8611 | |
949c47a0 DW |
8612 | /* check that prepare update was successful */ |
8613 | if (!update->space) { | |
1ade5cc1 | 8614 | dprintf("prepare update failed\n"); |
ba2de7ba | 8615 | goto create_error; |
949c47a0 DW |
8616 | } |
8617 | ||
54c2c1ea DW |
8618 | /* check that all disks are still active before committing |
8619 | * changes. FIXME: could we instead handle this by creating a | |
8620 | * degraded array? That's probably not what the user expects, | |
8621 | * so better to drop this update on the floor. | |
8622 | */ | |
8623 | for (i = 0; i < new_map->num_members; i++) { | |
8624 | dl = serial_to_dl(inf[i].serial, super); | |
8625 | if (!dl) { | |
1ade5cc1 | 8626 | dprintf("disk disappeared\n"); |
ba2de7ba | 8627 | goto create_error; |
54c2c1ea | 8628 | } |
949c47a0 DW |
8629 | } |
8630 | ||
8273f55e | 8631 | super->updates_pending++; |
54c2c1ea DW |
8632 | |
8633 | /* convert spares to members and fixup ord_tbl */ | |
8634 | for (i = 0; i < new_map->num_members; i++) { | |
8635 | dl = serial_to_dl(inf[i].serial, super); | |
8636 | if (dl->index == -1) { | |
8637 | dl->index = mpb->num_disks; | |
8638 | mpb->num_disks++; | |
8639 | dl->disk.status |= CONFIGURED_DISK; | |
8640 | dl->disk.status &= ~SPARE_DISK; | |
8641 | } | |
8642 | set_imsm_ord_tbl_ent(new_map, i, dl->index); | |
8643 | } | |
8644 | ||
ba2de7ba DW |
8645 | dv = update->space; |
8646 | dev = dv->dev; | |
949c47a0 DW |
8647 | update->space = NULL; |
8648 | imsm_copy_dev(dev, &u->dev); | |
ba2de7ba DW |
8649 | dv->index = u->dev_idx; |
8650 | dv->next = super->devlist; | |
8651 | super->devlist = dv; | |
8273f55e | 8652 | mpb->num_raid_devs++; |
8273f55e | 8653 | |
4d1313e9 | 8654 | imsm_update_version_info(super); |
8273f55e | 8655 | break; |
ba2de7ba DW |
8656 | create_error: |
8657 | /* mdmon knows how to release update->space, but not | |
8658 | * ((struct intel_dev *) update->space)->dev | |
8659 | */ | |
8660 | if (update->space) { | |
8661 | dv = update->space; | |
8662 | free(dv->dev); | |
8663 | } | |
8273f55e | 8664 | break; |
e8319a19 | 8665 | } |
33414a01 DW |
8666 | case update_kill_array: { |
8667 | struct imsm_update_kill_array *u = (void *) update->buf; | |
8668 | int victim = u->dev_idx; | |
8669 | struct active_array *a; | |
8670 | struct intel_dev **dp; | |
8671 | struct imsm_dev *dev; | |
8672 | ||
8673 | /* sanity check that we are not affecting the uuid of | |
8674 | * active arrays, or deleting an active array | |
8675 | * | |
8676 | * FIXME when immutable ids are available, but note that | |
8677 | * we'll also need to fixup the invalidated/active | |
8678 | * subarray indexes in mdstat | |
8679 | */ | |
8680 | for (a = st->arrays; a; a = a->next) | |
8681 | if (a->info.container_member >= victim) | |
8682 | break; | |
8683 | /* by definition if mdmon is running at least one array | |
8684 | * is active in the container, so checking | |
8685 | * mpb->num_raid_devs is just extra paranoia | |
8686 | */ | |
8687 | dev = get_imsm_dev(super, victim); | |
8688 | if (a || !dev || mpb->num_raid_devs == 1) { | |
8689 | dprintf("failed to delete subarray-%d\n", victim); | |
8690 | break; | |
8691 | } | |
8692 | ||
8693 | for (dp = &super->devlist; *dp;) | |
f21e18ca | 8694 | if ((*dp)->index == (unsigned)super->current_vol) { |
33414a01 DW |
8695 | *dp = (*dp)->next; |
8696 | } else { | |
f21e18ca | 8697 | if ((*dp)->index > (unsigned)victim) |
33414a01 DW |
8698 | (*dp)->index--; |
8699 | dp = &(*dp)->next; | |
8700 | } | |
8701 | mpb->num_raid_devs--; | |
8702 | super->updates_pending++; | |
8703 | break; | |
8704 | } | |
aa534678 DW |
8705 | case update_rename_array: { |
8706 | struct imsm_update_rename_array *u = (void *) update->buf; | |
8707 | char name[MAX_RAID_SERIAL_LEN+1]; | |
8708 | int target = u->dev_idx; | |
8709 | struct active_array *a; | |
8710 | struct imsm_dev *dev; | |
8711 | ||
8712 | /* sanity check that we are not affecting the uuid of | |
8713 | * an active array | |
8714 | */ | |
8715 | snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name); | |
8716 | name[MAX_RAID_SERIAL_LEN] = '\0'; | |
8717 | for (a = st->arrays; a; a = a->next) | |
8718 | if (a->info.container_member == target) | |
8719 | break; | |
8720 | dev = get_imsm_dev(super, u->dev_idx); | |
8721 | if (a || !dev || !check_name(super, name, 1)) { | |
8722 | dprintf("failed to rename subarray-%d\n", target); | |
8723 | break; | |
8724 | } | |
8725 | ||
cdbe98cd | 8726 | snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name); |
aa534678 DW |
8727 | super->updates_pending++; |
8728 | break; | |
8729 | } | |
1a64be56 | 8730 | case update_add_remove_disk: { |
43dad3d6 | 8731 | /* we may be able to repair some arrays if disks are |
095b8088 | 8732 | * being added, check the status of add_remove_disk |
1a64be56 LM |
8733 | * if discs has been added. |
8734 | */ | |
8735 | if (add_remove_disk_update(super)) { | |
43dad3d6 | 8736 | struct active_array *a; |
072b727f DW |
8737 | |
8738 | super->updates_pending++; | |
1a64be56 | 8739 | for (a = st->arrays; a; a = a->next) |
43dad3d6 DW |
8740 | a->check_degraded = 1; |
8741 | } | |
43dad3d6 | 8742 | break; |
e8319a19 | 8743 | } |
1a64be56 | 8744 | default: |
7a862a02 | 8745 | pr_err("error: unsuported process update type:(type: %d)\n", type); |
1a64be56 | 8746 | } |
e8319a19 | 8747 | } |
88758e9d | 8748 | |
bc0b9d34 PC |
8749 | static struct mdinfo *get_spares_for_grow(struct supertype *st); |
8750 | ||
5fe6f031 N |
8751 | static int imsm_prepare_update(struct supertype *st, |
8752 | struct metadata_update *update) | |
8273f55e | 8753 | { |
949c47a0 | 8754 | /** |
4d7b1503 DW |
8755 | * Allocate space to hold new disk entries, raid-device entries or a new |
8756 | * mpb if necessary. The manager synchronously waits for updates to | |
8757 | * complete in the monitor, so new mpb buffers allocated here can be | |
8758 | * integrated by the monitor thread without worrying about live pointers | |
8759 | * in the manager thread. | |
8273f55e | 8760 | */ |
095b8088 | 8761 | enum imsm_update_type type; |
4d7b1503 DW |
8762 | struct intel_super *super = st->sb; |
8763 | struct imsm_super *mpb = super->anchor; | |
8764 | size_t buf_len; | |
8765 | size_t len = 0; | |
949c47a0 | 8766 | |
095b8088 N |
8767 | if (update->len < (int)sizeof(type)) |
8768 | return 0; | |
8769 | ||
8770 | type = *(enum imsm_update_type *) update->buf; | |
8771 | ||
949c47a0 | 8772 | switch (type) { |
0ec5d470 | 8773 | case update_general_migration_checkpoint: |
095b8088 N |
8774 | if (update->len < (int)sizeof(struct imsm_update_general_migration_checkpoint)) |
8775 | return 0; | |
1ade5cc1 | 8776 | dprintf("called for update_general_migration_checkpoint\n"); |
0ec5d470 | 8777 | break; |
abedf5fc KW |
8778 | case update_takeover: { |
8779 | struct imsm_update_takeover *u = (void *)update->buf; | |
095b8088 N |
8780 | if (update->len < (int)sizeof(*u)) |
8781 | return 0; | |
abedf5fc KW |
8782 | if (u->direction == R0_TO_R10) { |
8783 | void **tail = (void **)&update->space_list; | |
8784 | struct imsm_dev *dev = get_imsm_dev(super, u->subarray); | |
238c0a71 | 8785 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
abedf5fc KW |
8786 | int num_members = map->num_members; |
8787 | void *space; | |
8788 | int size, i; | |
abedf5fc KW |
8789 | /* allocate memory for added disks */ |
8790 | for (i = 0; i < num_members; i++) { | |
8791 | size = sizeof(struct dl); | |
503975b9 | 8792 | space = xmalloc(size); |
abedf5fc KW |
8793 | *tail = space; |
8794 | tail = space; | |
8795 | *tail = NULL; | |
8796 | } | |
8797 | /* allocate memory for new device */ | |
8798 | size = sizeof_imsm_dev(super->devlist->dev, 0) + | |
8799 | (num_members * sizeof(__u32)); | |
503975b9 N |
8800 | space = xmalloc(size); |
8801 | *tail = space; | |
8802 | tail = space; | |
8803 | *tail = NULL; | |
8804 | len = disks_to_mpb_size(num_members * 2); | |
abedf5fc KW |
8805 | } |
8806 | ||
8807 | break; | |
8808 | } | |
78b10e66 | 8809 | case update_reshape_container_disks: { |
d195167d AK |
8810 | /* Every raid device in the container is about to |
8811 | * gain some more devices, and we will enter a | |
8812 | * reconfiguration. | |
8813 | * So each 'imsm_map' will be bigger, and the imsm_vol | |
8814 | * will now hold 2 of them. | |
8815 | * Thus we need new 'struct imsm_dev' allocations sized | |
8816 | * as sizeof_imsm_dev but with more devices in both maps. | |
8817 | */ | |
8818 | struct imsm_update_reshape *u = (void *)update->buf; | |
8819 | struct intel_dev *dl; | |
8820 | void **space_tail = (void**)&update->space_list; | |
8821 | ||
095b8088 N |
8822 | if (update->len < (int)sizeof(*u)) |
8823 | return 0; | |
8824 | ||
1ade5cc1 | 8825 | dprintf("for update_reshape\n"); |
d195167d AK |
8826 | |
8827 | for (dl = super->devlist; dl; dl = dl->next) { | |
8828 | int size = sizeof_imsm_dev(dl->dev, 1); | |
8829 | void *s; | |
d677e0b8 AK |
8830 | if (u->new_raid_disks > u->old_raid_disks) |
8831 | size += sizeof(__u32)*2* | |
8832 | (u->new_raid_disks - u->old_raid_disks); | |
503975b9 | 8833 | s = xmalloc(size); |
d195167d AK |
8834 | *space_tail = s; |
8835 | space_tail = s; | |
8836 | *space_tail = NULL; | |
8837 | } | |
8838 | ||
8839 | len = disks_to_mpb_size(u->new_raid_disks); | |
8840 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
78b10e66 N |
8841 | break; |
8842 | } | |
48c5303a | 8843 | case update_reshape_migration: { |
bc0b9d34 PC |
8844 | /* for migration level 0->5 we need to add disks |
8845 | * so the same as for container operation we will copy | |
8846 | * device to the bigger location. | |
8847 | * in memory prepared device and new disk area are prepared | |
8848 | * for usage in process update | |
8849 | */ | |
8850 | struct imsm_update_reshape_migration *u = (void *)update->buf; | |
8851 | struct intel_dev *id; | |
8852 | void **space_tail = (void **)&update->space_list; | |
8853 | int size; | |
8854 | void *s; | |
8855 | int current_level = -1; | |
8856 | ||
095b8088 N |
8857 | if (update->len < (int)sizeof(*u)) |
8858 | return 0; | |
8859 | ||
1ade5cc1 | 8860 | dprintf("for update_reshape\n"); |
bc0b9d34 PC |
8861 | |
8862 | /* add space for bigger array in update | |
8863 | */ | |
8864 | for (id = super->devlist; id; id = id->next) { | |
8865 | if (id->index == (unsigned)u->subdev) { | |
8866 | size = sizeof_imsm_dev(id->dev, 1); | |
8867 | if (u->new_raid_disks > u->old_raid_disks) | |
8868 | size += sizeof(__u32)*2* | |
8869 | (u->new_raid_disks - u->old_raid_disks); | |
503975b9 | 8870 | s = xmalloc(size); |
bc0b9d34 PC |
8871 | *space_tail = s; |
8872 | space_tail = s; | |
8873 | *space_tail = NULL; | |
8874 | break; | |
8875 | } | |
8876 | } | |
8877 | if (update->space_list == NULL) | |
8878 | break; | |
8879 | ||
8880 | /* add space for disk in update | |
8881 | */ | |
8882 | size = sizeof(struct dl); | |
503975b9 | 8883 | s = xmalloc(size); |
bc0b9d34 PC |
8884 | *space_tail = s; |
8885 | space_tail = s; | |
8886 | *space_tail = NULL; | |
8887 | ||
8888 | /* add spare device to update | |
8889 | */ | |
8890 | for (id = super->devlist ; id; id = id->next) | |
8891 | if (id->index == (unsigned)u->subdev) { | |
8892 | struct imsm_dev *dev; | |
8893 | struct imsm_map *map; | |
8894 | ||
8895 | dev = get_imsm_dev(super, u->subdev); | |
238c0a71 | 8896 | map = get_imsm_map(dev, MAP_0); |
bc0b9d34 PC |
8897 | current_level = map->raid_level; |
8898 | break; | |
8899 | } | |
8900 | if ((u->new_level == 5) && (u->new_level != current_level)) { | |
8901 | struct mdinfo *spares; | |
8902 | ||
8903 | spares = get_spares_for_grow(st); | |
8904 | if (spares) { | |
8905 | struct dl *dl; | |
8906 | struct mdinfo *dev; | |
8907 | ||
8908 | dev = spares->devs; | |
8909 | if (dev) { | |
8910 | u->new_disks[0] = | |
8911 | makedev(dev->disk.major, | |
8912 | dev->disk.minor); | |
8913 | dl = get_disk_super(super, | |
8914 | dev->disk.major, | |
8915 | dev->disk.minor); | |
8916 | dl->index = u->old_raid_disks; | |
8917 | dev = dev->next; | |
8918 | } | |
8919 | sysfs_free(spares); | |
8920 | } | |
8921 | } | |
8922 | len = disks_to_mpb_size(u->new_raid_disks); | |
8923 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
48c5303a PC |
8924 | break; |
8925 | } | |
f3871fdc | 8926 | case update_size_change: { |
095b8088 N |
8927 | if (update->len < (int)sizeof(struct imsm_update_size_change)) |
8928 | return 0; | |
8929 | break; | |
8930 | } | |
8931 | case update_activate_spare: { | |
8932 | if (update->len < (int)sizeof(struct imsm_update_activate_spare)) | |
8933 | return 0; | |
f3871fdc AK |
8934 | break; |
8935 | } | |
949c47a0 DW |
8936 | case update_create_array: { |
8937 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 8938 | struct intel_dev *dv; |
54c2c1ea | 8939 | struct imsm_dev *dev = &u->dev; |
238c0a71 | 8940 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
54c2c1ea DW |
8941 | struct dl *dl; |
8942 | struct disk_info *inf; | |
8943 | int i; | |
8944 | int activate = 0; | |
949c47a0 | 8945 | |
095b8088 N |
8946 | if (update->len < (int)sizeof(*u)) |
8947 | return 0; | |
8948 | ||
54c2c1ea DW |
8949 | inf = get_disk_info(u); |
8950 | len = sizeof_imsm_dev(dev, 1); | |
ba2de7ba | 8951 | /* allocate a new super->devlist entry */ |
503975b9 N |
8952 | dv = xmalloc(sizeof(*dv)); |
8953 | dv->dev = xmalloc(len); | |
8954 | update->space = dv; | |
949c47a0 | 8955 | |
54c2c1ea DW |
8956 | /* count how many spares will be converted to members */ |
8957 | for (i = 0; i < map->num_members; i++) { | |
8958 | dl = serial_to_dl(inf[i].serial, super); | |
8959 | if (!dl) { | |
8960 | /* hmm maybe it failed?, nothing we can do about | |
8961 | * it here | |
8962 | */ | |
8963 | continue; | |
8964 | } | |
8965 | if (count_memberships(dl, super) == 0) | |
8966 | activate++; | |
8967 | } | |
8968 | len += activate * sizeof(struct imsm_disk); | |
949c47a0 | 8969 | break; |
095b8088 N |
8970 | } |
8971 | case update_kill_array: { | |
8972 | if (update->len < (int)sizeof(struct imsm_update_kill_array)) | |
8973 | return 0; | |
949c47a0 DW |
8974 | break; |
8975 | } | |
095b8088 N |
8976 | case update_rename_array: { |
8977 | if (update->len < (int)sizeof(struct imsm_update_rename_array)) | |
8978 | return 0; | |
8979 | break; | |
8980 | } | |
8981 | case update_add_remove_disk: | |
8982 | /* no update->len needed */ | |
8983 | break; | |
8984 | default: | |
8985 | return 0; | |
949c47a0 | 8986 | } |
8273f55e | 8987 | |
4d7b1503 DW |
8988 | /* check if we need a larger metadata buffer */ |
8989 | if (super->next_buf) | |
8990 | buf_len = super->next_len; | |
8991 | else | |
8992 | buf_len = super->len; | |
8993 | ||
8994 | if (__le32_to_cpu(mpb->mpb_size) + len > buf_len) { | |
8995 | /* ok we need a larger buf than what is currently allocated | |
8996 | * if this allocation fails process_update will notice that | |
8997 | * ->next_len is set and ->next_buf is NULL | |
8998 | */ | |
8999 | buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + len, 512); | |
9000 | if (super->next_buf) | |
9001 | free(super->next_buf); | |
9002 | ||
9003 | super->next_len = buf_len; | |
1f45a8ad DW |
9004 | if (posix_memalign(&super->next_buf, 512, buf_len) == 0) |
9005 | memset(super->next_buf, 0, buf_len); | |
9006 | else | |
4d7b1503 DW |
9007 | super->next_buf = NULL; |
9008 | } | |
5fe6f031 | 9009 | return 1; |
8273f55e DW |
9010 | } |
9011 | ||
ae6aad82 | 9012 | /* must be called while manager is quiesced */ |
f21e18ca | 9013 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index) |
ae6aad82 DW |
9014 | { |
9015 | struct imsm_super *mpb = super->anchor; | |
ae6aad82 DW |
9016 | struct dl *iter; |
9017 | struct imsm_dev *dev; | |
9018 | struct imsm_map *map; | |
24565c9a DW |
9019 | int i, j, num_members; |
9020 | __u32 ord; | |
ae6aad82 | 9021 | |
1ade5cc1 | 9022 | dprintf("deleting device[%d] from imsm_super\n", index); |
ae6aad82 DW |
9023 | |
9024 | /* shift all indexes down one */ | |
9025 | for (iter = super->disks; iter; iter = iter->next) | |
f21e18ca | 9026 | if (iter->index > (int)index) |
ae6aad82 | 9027 | iter->index--; |
47ee5a45 | 9028 | for (iter = super->missing; iter; iter = iter->next) |
f21e18ca | 9029 | if (iter->index > (int)index) |
47ee5a45 | 9030 | iter->index--; |
ae6aad82 DW |
9031 | |
9032 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
9033 | dev = get_imsm_dev(super, i); | |
238c0a71 | 9034 | map = get_imsm_map(dev, MAP_0); |
24565c9a DW |
9035 | num_members = map->num_members; |
9036 | for (j = 0; j < num_members; j++) { | |
9037 | /* update ord entries being careful not to propagate | |
9038 | * ord-flags to the first map | |
9039 | */ | |
238c0a71 | 9040 | ord = get_imsm_ord_tbl_ent(dev, j, MAP_X); |
ae6aad82 | 9041 | |
24565c9a DW |
9042 | if (ord_to_idx(ord) <= index) |
9043 | continue; | |
ae6aad82 | 9044 | |
238c0a71 | 9045 | map = get_imsm_map(dev, MAP_0); |
24565c9a | 9046 | set_imsm_ord_tbl_ent(map, j, ord_to_idx(ord - 1)); |
238c0a71 | 9047 | map = get_imsm_map(dev, MAP_1); |
24565c9a DW |
9048 | if (map) |
9049 | set_imsm_ord_tbl_ent(map, j, ord - 1); | |
ae6aad82 DW |
9050 | } |
9051 | } | |
9052 | ||
9053 | mpb->num_disks--; | |
9054 | super->updates_pending++; | |
24565c9a DW |
9055 | if (*dlp) { |
9056 | struct dl *dl = *dlp; | |
9057 | ||
9058 | *dlp = (*dlp)->next; | |
9059 | __free_imsm_disk(dl); | |
9060 | } | |
ae6aad82 | 9061 | } |
9e2d750d | 9062 | #endif /* MDASSEMBLE */ |
9a717282 AK |
9063 | |
9064 | static void close_targets(int *targets, int new_disks) | |
9065 | { | |
9066 | int i; | |
9067 | ||
9068 | if (!targets) | |
9069 | return; | |
9070 | ||
9071 | for (i = 0; i < new_disks; i++) { | |
9072 | if (targets[i] >= 0) { | |
9073 | close(targets[i]); | |
9074 | targets[i] = -1; | |
9075 | } | |
9076 | } | |
9077 | } | |
9078 | ||
9079 | static int imsm_get_allowed_degradation(int level, int raid_disks, | |
9080 | struct intel_super *super, | |
9081 | struct imsm_dev *dev) | |
9082 | { | |
9083 | switch (level) { | |
bf5cf7c7 | 9084 | case 1: |
9a717282 AK |
9085 | case 10:{ |
9086 | int ret_val = 0; | |
9087 | struct imsm_map *map; | |
9088 | int i; | |
9089 | ||
9090 | ret_val = raid_disks/2; | |
9091 | /* check map if all disks pairs not failed | |
9092 | * in both maps | |
9093 | */ | |
238c0a71 | 9094 | map = get_imsm_map(dev, MAP_0); |
9a717282 AK |
9095 | for (i = 0; i < ret_val; i++) { |
9096 | int degradation = 0; | |
9097 | if (get_imsm_disk(super, i) == NULL) | |
9098 | degradation++; | |
9099 | if (get_imsm_disk(super, i + 1) == NULL) | |
9100 | degradation++; | |
9101 | if (degradation == 2) | |
9102 | return 0; | |
9103 | } | |
238c0a71 | 9104 | map = get_imsm_map(dev, MAP_1); |
9a717282 AK |
9105 | /* if there is no second map |
9106 | * result can be returned | |
9107 | */ | |
9108 | if (map == NULL) | |
9109 | return ret_val; | |
9110 | /* check degradation in second map | |
9111 | */ | |
9112 | for (i = 0; i < ret_val; i++) { | |
9113 | int degradation = 0; | |
9114 | if (get_imsm_disk(super, i) == NULL) | |
9115 | degradation++; | |
9116 | if (get_imsm_disk(super, i + 1) == NULL) | |
9117 | degradation++; | |
9118 | if (degradation == 2) | |
9119 | return 0; | |
9120 | } | |
9121 | return ret_val; | |
9122 | } | |
9123 | case 5: | |
9124 | return 1; | |
9125 | case 6: | |
9126 | return 2; | |
9127 | default: | |
9128 | return 0; | |
9129 | } | |
9130 | } | |
9131 | ||
687629c2 AK |
9132 | /******************************************************************************* |
9133 | * Function: open_backup_targets | |
9134 | * Description: Function opens file descriptors for all devices given in | |
9135 | * info->devs | |
9136 | * Parameters: | |
9137 | * info : general array info | |
9138 | * raid_disks : number of disks | |
9139 | * raid_fds : table of device's file descriptors | |
9a717282 AK |
9140 | * super : intel super for raid10 degradation check |
9141 | * dev : intel device for raid10 degradation check | |
687629c2 AK |
9142 | * Returns: |
9143 | * 0 : success | |
9144 | * -1 : fail | |
9145 | ******************************************************************************/ | |
9a717282 AK |
9146 | int open_backup_targets(struct mdinfo *info, int raid_disks, int *raid_fds, |
9147 | struct intel_super *super, struct imsm_dev *dev) | |
687629c2 AK |
9148 | { |
9149 | struct mdinfo *sd; | |
f627f5ad | 9150 | int i; |
9a717282 | 9151 | int opened = 0; |
f627f5ad AK |
9152 | |
9153 | for (i = 0; i < raid_disks; i++) | |
9154 | raid_fds[i] = -1; | |
687629c2 AK |
9155 | |
9156 | for (sd = info->devs ; sd ; sd = sd->next) { | |
9157 | char *dn; | |
9158 | ||
9159 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) { | |
9160 | dprintf("disk is faulty!!\n"); | |
9161 | continue; | |
9162 | } | |
9163 | ||
9164 | if ((sd->disk.raid_disk >= raid_disks) || | |
9165 | (sd->disk.raid_disk < 0)) | |
9166 | continue; | |
9167 | ||
9168 | dn = map_dev(sd->disk.major, | |
9169 | sd->disk.minor, 1); | |
9170 | raid_fds[sd->disk.raid_disk] = dev_open(dn, O_RDWR); | |
9171 | if (raid_fds[sd->disk.raid_disk] < 0) { | |
e12b3daa | 9172 | pr_err("cannot open component\n"); |
9a717282 | 9173 | continue; |
687629c2 | 9174 | } |
9a717282 AK |
9175 | opened++; |
9176 | } | |
9177 | /* check if maximum array degradation level is not exceeded | |
9178 | */ | |
9179 | if ((raid_disks - opened) > | |
9180 | imsm_get_allowed_degradation(info->new_level, | |
9181 | raid_disks, | |
9182 | super, dev)) { | |
e12b3daa | 9183 | pr_err("Not enough disks can be opened.\n"); |
9a717282 AK |
9184 | close_targets(raid_fds, raid_disks); |
9185 | return -2; | |
687629c2 AK |
9186 | } |
9187 | return 0; | |
9188 | } | |
9189 | ||
d31ad643 PB |
9190 | /******************************************************************************* |
9191 | * Function: validate_container_imsm | |
9192 | * Description: This routine validates container after assemble, | |
9193 | * eg. if devices in container are under the same controller. | |
9194 | * | |
9195 | * Parameters: | |
9196 | * info : linked list with info about devices used in array | |
9197 | * Returns: | |
9198 | * 1 : HBA mismatch | |
9199 | * 0 : Success | |
9200 | ******************************************************************************/ | |
9201 | int validate_container_imsm(struct mdinfo *info) | |
9202 | { | |
6b781d33 AP |
9203 | if (check_env("IMSM_NO_PLATFORM")) |
9204 | return 0; | |
d31ad643 | 9205 | |
6b781d33 AP |
9206 | struct sys_dev *idev; |
9207 | struct sys_dev *hba = NULL; | |
9208 | struct sys_dev *intel_devices = find_intel_devices(); | |
9209 | char *dev_path = devt_to_devpath(makedev(info->disk.major, | |
9210 | info->disk.minor)); | |
9211 | ||
9212 | for (idev = intel_devices; idev; idev = idev->next) { | |
9213 | if (dev_path && strstr(dev_path, idev->path)) { | |
9214 | hba = idev; | |
9215 | break; | |
d31ad643 | 9216 | } |
6b781d33 AP |
9217 | } |
9218 | if (dev_path) | |
d31ad643 PB |
9219 | free(dev_path); |
9220 | ||
6b781d33 AP |
9221 | if (!hba) { |
9222 | pr_err("WARNING - Cannot detect HBA for device %s!\n", | |
9223 | devid2kname(makedev(info->disk.major, info->disk.minor))); | |
9224 | return 1; | |
9225 | } | |
9226 | ||
9227 | const struct imsm_orom *orom = get_orom_by_device_id(hba->dev_id); | |
9228 | struct mdinfo *dev; | |
9229 | ||
9230 | for (dev = info->next; dev; dev = dev->next) { | |
9231 | dev_path = devt_to_devpath(makedev(dev->disk.major, dev->disk.minor)); | |
9232 | ||
9233 | struct sys_dev *hba2 = NULL; | |
9234 | for (idev = intel_devices; idev; idev = idev->next) { | |
9235 | if (dev_path && strstr(dev_path, idev->path)) { | |
9236 | hba2 = idev; | |
9237 | break; | |
d31ad643 PB |
9238 | } |
9239 | } | |
6b781d33 AP |
9240 | if (dev_path) |
9241 | free(dev_path); | |
9242 | ||
9243 | const struct imsm_orom *orom2 = hba2 == NULL ? NULL : | |
9244 | get_orom_by_device_id(hba2->dev_id); | |
9245 | ||
9246 | if (hba2 && hba->type != hba2->type) { | |
9247 | pr_err("WARNING - HBAs of devices do not match %s != %s\n", | |
9248 | get_sys_dev_type(hba->type), get_sys_dev_type(hba2->type)); | |
9249 | return 1; | |
9250 | } | |
9251 | ||
60f0f54d | 9252 | if ((orom != orom2) || ((hba->type == SYS_DEV_VMD) && (hba != hba2))) { |
6b781d33 AP |
9253 | pr_err("WARNING - IMSM container assembled with disks under different HBAs!\n" |
9254 | " This operation is not supported and can lead to data loss.\n"); | |
9255 | return 1; | |
9256 | } | |
9257 | ||
9258 | if (!orom) { | |
9259 | pr_err("WARNING - IMSM container assembled with disks under HBAs without IMSM platform support!\n" | |
9260 | " This operation is not supported and can lead to data loss.\n"); | |
9261 | return 1; | |
9262 | } | |
d31ad643 | 9263 | } |
6b781d33 | 9264 | |
d31ad643 PB |
9265 | return 0; |
9266 | } | |
9e2d750d | 9267 | #ifndef MDASSEMBLE |
687629c2 AK |
9268 | /******************************************************************************* |
9269 | * Function: init_migr_record_imsm | |
9270 | * Description: Function inits imsm migration record | |
9271 | * Parameters: | |
9272 | * super : imsm internal array info | |
9273 | * dev : device under migration | |
9274 | * info : general array info to find the smallest device | |
9275 | * Returns: | |
9276 | * none | |
9277 | ******************************************************************************/ | |
9278 | void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev, | |
9279 | struct mdinfo *info) | |
9280 | { | |
9281 | struct intel_super *super = st->sb; | |
9282 | struct migr_record *migr_rec = super->migr_rec; | |
9283 | int new_data_disks; | |
9284 | unsigned long long dsize, dev_sectors; | |
9285 | long long unsigned min_dev_sectors = -1LLU; | |
9286 | struct mdinfo *sd; | |
9287 | char nm[30]; | |
9288 | int fd; | |
238c0a71 AK |
9289 | struct imsm_map *map_dest = get_imsm_map(dev, MAP_0); |
9290 | struct imsm_map *map_src = get_imsm_map(dev, MAP_1); | |
687629c2 | 9291 | unsigned long long num_migr_units; |
3ef4403c | 9292 | unsigned long long array_blocks; |
687629c2 AK |
9293 | |
9294 | memset(migr_rec, 0, sizeof(struct migr_record)); | |
9295 | migr_rec->family_num = __cpu_to_le32(super->anchor->family_num); | |
9296 | ||
9297 | /* only ascending reshape supported now */ | |
9298 | migr_rec->ascending_migr = __cpu_to_le32(1); | |
9299 | ||
9300 | migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE / | |
9301 | max(map_dest->blocks_per_strip, map_src->blocks_per_strip); | |
e1742195 AK |
9302 | migr_rec->dest_depth_per_unit *= |
9303 | max(map_dest->blocks_per_strip, map_src->blocks_per_strip); | |
238c0a71 | 9304 | new_data_disks = imsm_num_data_members(dev, MAP_0); |
687629c2 AK |
9305 | migr_rec->blocks_per_unit = |
9306 | __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks); | |
9307 | migr_rec->dest_depth_per_unit = | |
9308 | __cpu_to_le32(migr_rec->dest_depth_per_unit); | |
3ef4403c | 9309 | array_blocks = info->component_size * new_data_disks; |
687629c2 AK |
9310 | num_migr_units = |
9311 | array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit); | |
9312 | ||
9313 | if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit)) | |
9314 | num_migr_units++; | |
9315 | migr_rec->num_migr_units = __cpu_to_le32(num_migr_units); | |
9316 | ||
9317 | migr_rec->post_migr_vol_cap = dev->size_low; | |
9318 | migr_rec->post_migr_vol_cap_hi = dev->size_high; | |
9319 | ||
687629c2 AK |
9320 | /* Find the smallest dev */ |
9321 | for (sd = info->devs ; sd ; sd = sd->next) { | |
9322 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
9323 | fd = dev_open(nm, O_RDONLY); | |
9324 | if (fd < 0) | |
9325 | continue; | |
9326 | get_dev_size(fd, NULL, &dsize); | |
9327 | dev_sectors = dsize / 512; | |
9328 | if (dev_sectors < min_dev_sectors) | |
9329 | min_dev_sectors = dev_sectors; | |
9330 | close(fd); | |
9331 | } | |
9332 | migr_rec->ckpt_area_pba = __cpu_to_le32(min_dev_sectors - | |
9333 | RAID_DISK_RESERVED_BLOCKS_IMSM_HI); | |
9334 | ||
9335 | write_imsm_migr_rec(st); | |
9336 | ||
9337 | return; | |
9338 | } | |
9339 | ||
9340 | /******************************************************************************* | |
9341 | * Function: save_backup_imsm | |
9342 | * Description: Function saves critical data stripes to Migration Copy Area | |
9343 | * and updates the current migration unit status. | |
9344 | * Use restore_stripes() to form a destination stripe, | |
9345 | * and to write it to the Copy Area. | |
9346 | * Parameters: | |
9347 | * st : supertype information | |
aea93171 | 9348 | * dev : imsm device that backup is saved for |
687629c2 AK |
9349 | * info : general array info |
9350 | * buf : input buffer | |
687629c2 AK |
9351 | * length : length of data to backup (blocks_per_unit) |
9352 | * Returns: | |
9353 | * 0 : success | |
9354 | *, -1 : fail | |
9355 | ******************************************************************************/ | |
9356 | int save_backup_imsm(struct supertype *st, | |
9357 | struct imsm_dev *dev, | |
9358 | struct mdinfo *info, | |
9359 | void *buf, | |
687629c2 AK |
9360 | int length) |
9361 | { | |
9362 | int rv = -1; | |
9363 | struct intel_super *super = st->sb; | |
9364 | unsigned long long *target_offsets = NULL; | |
9365 | int *targets = NULL; | |
9366 | int i; | |
238c0a71 | 9367 | struct imsm_map *map_dest = get_imsm_map(dev, MAP_0); |
687629c2 | 9368 | int new_disks = map_dest->num_members; |
ab724b98 AK |
9369 | int dest_layout = 0; |
9370 | int dest_chunk; | |
d1877f69 | 9371 | unsigned long long start; |
238c0a71 | 9372 | int data_disks = imsm_num_data_members(dev, MAP_0); |
687629c2 | 9373 | |
503975b9 | 9374 | targets = xmalloc(new_disks * sizeof(int)); |
687629c2 | 9375 | |
7e45b550 AK |
9376 | for (i = 0; i < new_disks; i++) |
9377 | targets[i] = -1; | |
9378 | ||
503975b9 | 9379 | target_offsets = xcalloc(new_disks, sizeof(unsigned long long)); |
687629c2 | 9380 | |
d1877f69 | 9381 | start = info->reshape_progress * 512; |
687629c2 | 9382 | for (i = 0; i < new_disks; i++) { |
687629c2 AK |
9383 | target_offsets[i] = (unsigned long long) |
9384 | __le32_to_cpu(super->migr_rec->ckpt_area_pba) * 512; | |
d1877f69 AK |
9385 | /* move back copy area adderss, it will be moved forward |
9386 | * in restore_stripes() using start input variable | |
9387 | */ | |
9388 | target_offsets[i] -= start/data_disks; | |
687629c2 AK |
9389 | } |
9390 | ||
9a717282 AK |
9391 | if (open_backup_targets(info, new_disks, targets, |
9392 | super, dev)) | |
687629c2 AK |
9393 | goto abort; |
9394 | ||
68eb8bc6 | 9395 | dest_layout = imsm_level_to_layout(map_dest->raid_level); |
ab724b98 AK |
9396 | dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512; |
9397 | ||
687629c2 AK |
9398 | if (restore_stripes(targets, /* list of dest devices */ |
9399 | target_offsets, /* migration record offsets */ | |
9400 | new_disks, | |
ab724b98 AK |
9401 | dest_chunk, |
9402 | map_dest->raid_level, | |
9403 | dest_layout, | |
9404 | -1, /* source backup file descriptor */ | |
9405 | 0, /* input buf offset | |
9406 | * always 0 buf is already offseted */ | |
d1877f69 | 9407 | start, |
687629c2 AK |
9408 | length, |
9409 | buf) != 0) { | |
e7b84f9d | 9410 | pr_err("Error restoring stripes\n"); |
687629c2 AK |
9411 | goto abort; |
9412 | } | |
9413 | ||
9414 | rv = 0; | |
9415 | ||
9416 | abort: | |
9417 | if (targets) { | |
9a717282 | 9418 | close_targets(targets, new_disks); |
687629c2 AK |
9419 | free(targets); |
9420 | } | |
9421 | free(target_offsets); | |
9422 | ||
9423 | return rv; | |
9424 | } | |
9425 | ||
9426 | /******************************************************************************* | |
9427 | * Function: save_checkpoint_imsm | |
9428 | * Description: Function called for current unit status update | |
9429 | * in the migration record. It writes it to disk. | |
9430 | * Parameters: | |
9431 | * super : imsm internal array info | |
9432 | * info : general array info | |
9433 | * Returns: | |
9434 | * 0: success | |
9435 | * 1: failure | |
0228d92c AK |
9436 | * 2: failure, means no valid migration record |
9437 | * / no general migration in progress / | |
687629c2 AK |
9438 | ******************************************************************************/ |
9439 | int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state) | |
9440 | { | |
9441 | struct intel_super *super = st->sb; | |
f8b72ef5 AK |
9442 | unsigned long long blocks_per_unit; |
9443 | unsigned long long curr_migr_unit; | |
9444 | ||
2e062e82 | 9445 | if (load_imsm_migr_rec(super, info) != 0) { |
7a862a02 | 9446 | dprintf("imsm: ERROR: Cannot read migration record for checkpoint save.\n"); |
2e062e82 AK |
9447 | return 1; |
9448 | } | |
9449 | ||
f8b72ef5 AK |
9450 | blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit); |
9451 | if (blocks_per_unit == 0) { | |
0228d92c AK |
9452 | dprintf("imsm: no migration in progress.\n"); |
9453 | return 2; | |
687629c2 | 9454 | } |
f8b72ef5 AK |
9455 | curr_migr_unit = info->reshape_progress / blocks_per_unit; |
9456 | /* check if array is alligned to copy area | |
9457 | * if it is not alligned, add one to current migration unit value | |
9458 | * this can happend on array reshape finish only | |
9459 | */ | |
9460 | if (info->reshape_progress % blocks_per_unit) | |
9461 | curr_migr_unit++; | |
687629c2 AK |
9462 | |
9463 | super->migr_rec->curr_migr_unit = | |
f8b72ef5 | 9464 | __cpu_to_le32(curr_migr_unit); |
687629c2 AK |
9465 | super->migr_rec->rec_status = __cpu_to_le32(state); |
9466 | super->migr_rec->dest_1st_member_lba = | |
f8b72ef5 AK |
9467 | __cpu_to_le32(curr_migr_unit * |
9468 | __le32_to_cpu(super->migr_rec->dest_depth_per_unit)); | |
687629c2 | 9469 | if (write_imsm_migr_rec(st) < 0) { |
7a862a02 | 9470 | dprintf("imsm: Cannot write migration record outside backup area\n"); |
687629c2 AK |
9471 | return 1; |
9472 | } | |
9473 | ||
9474 | return 0; | |
9475 | } | |
9476 | ||
276d77db AK |
9477 | /******************************************************************************* |
9478 | * Function: recover_backup_imsm | |
9479 | * Description: Function recovers critical data from the Migration Copy Area | |
9480 | * while assembling an array. | |
9481 | * Parameters: | |
9482 | * super : imsm internal array info | |
9483 | * info : general array info | |
9484 | * Returns: | |
9485 | * 0 : success (or there is no data to recover) | |
9486 | * 1 : fail | |
9487 | ******************************************************************************/ | |
9488 | int recover_backup_imsm(struct supertype *st, struct mdinfo *info) | |
9489 | { | |
9490 | struct intel_super *super = st->sb; | |
9491 | struct migr_record *migr_rec = super->migr_rec; | |
9492 | struct imsm_map *map_dest = NULL; | |
9493 | struct intel_dev *id = NULL; | |
9494 | unsigned long long read_offset; | |
9495 | unsigned long long write_offset; | |
9496 | unsigned unit_len; | |
9497 | int *targets = NULL; | |
9498 | int new_disks, i, err; | |
9499 | char *buf = NULL; | |
9500 | int retval = 1; | |
9501 | unsigned long curr_migr_unit = __le32_to_cpu(migr_rec->curr_migr_unit); | |
9502 | unsigned long num_migr_units = __le32_to_cpu(migr_rec->num_migr_units); | |
276d77db | 9503 | char buffer[20]; |
6c3560c0 | 9504 | int skipped_disks = 0; |
276d77db AK |
9505 | |
9506 | err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20); | |
9507 | if (err < 1) | |
9508 | return 1; | |
9509 | ||
9510 | /* recover data only during assemblation */ | |
9511 | if (strncmp(buffer, "inactive", 8) != 0) | |
9512 | return 0; | |
9513 | /* no data to recover */ | |
9514 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
9515 | return 0; | |
9516 | if (curr_migr_unit >= num_migr_units) | |
9517 | return 1; | |
9518 | ||
9519 | /* find device during reshape */ | |
9520 | for (id = super->devlist; id; id = id->next) | |
9521 | if (is_gen_migration(id->dev)) | |
9522 | break; | |
9523 | if (id == NULL) | |
9524 | return 1; | |
9525 | ||
238c0a71 | 9526 | map_dest = get_imsm_map(id->dev, MAP_0); |
276d77db AK |
9527 | new_disks = map_dest->num_members; |
9528 | ||
9529 | read_offset = (unsigned long long) | |
9530 | __le32_to_cpu(migr_rec->ckpt_area_pba) * 512; | |
9531 | ||
9532 | write_offset = ((unsigned long long) | |
9533 | __le32_to_cpu(migr_rec->dest_1st_member_lba) + | |
5551b113 | 9534 | pba_of_lba0(map_dest)) * 512; |
276d77db AK |
9535 | |
9536 | unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
9537 | if (posix_memalign((void **)&buf, 512, unit_len) != 0) | |
9538 | goto abort; | |
503975b9 | 9539 | targets = xcalloc(new_disks, sizeof(int)); |
276d77db | 9540 | |
9a717282 | 9541 | if (open_backup_targets(info, new_disks, targets, super, id->dev)) { |
e7b84f9d | 9542 | pr_err("Cannot open some devices belonging to array.\n"); |
f627f5ad AK |
9543 | goto abort; |
9544 | } | |
276d77db AK |
9545 | |
9546 | for (i = 0; i < new_disks; i++) { | |
6c3560c0 AK |
9547 | if (targets[i] < 0) { |
9548 | skipped_disks++; | |
9549 | continue; | |
9550 | } | |
276d77db | 9551 | if (lseek64(targets[i], read_offset, SEEK_SET) < 0) { |
e7b84f9d N |
9552 | pr_err("Cannot seek to block: %s\n", |
9553 | strerror(errno)); | |
137debce AK |
9554 | skipped_disks++; |
9555 | continue; | |
276d77db | 9556 | } |
9ec11d1a | 9557 | if ((unsigned)read(targets[i], buf, unit_len) != unit_len) { |
e7b84f9d N |
9558 | pr_err("Cannot read copy area block: %s\n", |
9559 | strerror(errno)); | |
137debce AK |
9560 | skipped_disks++; |
9561 | continue; | |
276d77db AK |
9562 | } |
9563 | if (lseek64(targets[i], write_offset, SEEK_SET) < 0) { | |
e7b84f9d N |
9564 | pr_err("Cannot seek to block: %s\n", |
9565 | strerror(errno)); | |
137debce AK |
9566 | skipped_disks++; |
9567 | continue; | |
276d77db | 9568 | } |
9ec11d1a | 9569 | if ((unsigned)write(targets[i], buf, unit_len) != unit_len) { |
e7b84f9d N |
9570 | pr_err("Cannot restore block: %s\n", |
9571 | strerror(errno)); | |
137debce AK |
9572 | skipped_disks++; |
9573 | continue; | |
276d77db AK |
9574 | } |
9575 | } | |
9576 | ||
137debce AK |
9577 | if (skipped_disks > imsm_get_allowed_degradation(info->new_level, |
9578 | new_disks, | |
9579 | super, | |
9580 | id->dev)) { | |
7a862a02 | 9581 | pr_err("Cannot restore data from backup. Too many failed disks\n"); |
6c3560c0 AK |
9582 | goto abort; |
9583 | } | |
9584 | ||
befb629b AK |
9585 | if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) { |
9586 | /* ignore error == 2, this can mean end of reshape here | |
9587 | */ | |
7a862a02 | 9588 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL) during restart\n"); |
befb629b | 9589 | } else |
276d77db | 9590 | retval = 0; |
276d77db AK |
9591 | |
9592 | abort: | |
9593 | if (targets) { | |
9594 | for (i = 0; i < new_disks; i++) | |
9595 | if (targets[i]) | |
9596 | close(targets[i]); | |
9597 | free(targets); | |
9598 | } | |
9599 | free(buf); | |
9600 | return retval; | |
9601 | } | |
9602 | ||
2cda7640 ML |
9603 | static char disk_by_path[] = "/dev/disk/by-path/"; |
9604 | ||
9605 | static const char *imsm_get_disk_controller_domain(const char *path) | |
9606 | { | |
2cda7640 | 9607 | char disk_path[PATH_MAX]; |
96234762 LM |
9608 | char *drv=NULL; |
9609 | struct stat st; | |
2cda7640 | 9610 | |
6d8d290a | 9611 | strcpy(disk_path, disk_by_path); |
96234762 LM |
9612 | strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1); |
9613 | if (stat(disk_path, &st) == 0) { | |
9614 | struct sys_dev* hba; | |
9615 | char *path=NULL; | |
9616 | ||
9617 | path = devt_to_devpath(st.st_rdev); | |
9618 | if (path == NULL) | |
9619 | return "unknown"; | |
9620 | hba = find_disk_attached_hba(-1, path); | |
9621 | if (hba && hba->type == SYS_DEV_SAS) | |
9622 | drv = "isci"; | |
9623 | else if (hba && hba->type == SYS_DEV_SATA) | |
9624 | drv = "ahci"; | |
1011e834 | 9625 | else |
96234762 LM |
9626 | drv = "unknown"; |
9627 | dprintf("path: %s hba: %s attached: %s\n", | |
9628 | path, (hba) ? hba->path : "NULL", drv); | |
9629 | free(path); | |
2cda7640 | 9630 | } |
96234762 | 9631 | return drv; |
2cda7640 ML |
9632 | } |
9633 | ||
4dd2df09 | 9634 | static char *imsm_find_array_devnm_by_subdev(int subdev, char *container) |
78b10e66 | 9635 | { |
4dd2df09 | 9636 | static char devnm[32]; |
78b10e66 N |
9637 | char subdev_name[20]; |
9638 | struct mdstat_ent *mdstat; | |
9639 | ||
9640 | sprintf(subdev_name, "%d", subdev); | |
9641 | mdstat = mdstat_by_subdev(subdev_name, container); | |
9642 | if (!mdstat) | |
4dd2df09 | 9643 | return NULL; |
78b10e66 | 9644 | |
4dd2df09 | 9645 | strcpy(devnm, mdstat->devnm); |
78b10e66 | 9646 | free_mdstat(mdstat); |
4dd2df09 | 9647 | return devnm; |
78b10e66 N |
9648 | } |
9649 | ||
9650 | static int imsm_reshape_is_allowed_on_container(struct supertype *st, | |
9651 | struct geo_params *geo, | |
fbf3d202 AK |
9652 | int *old_raid_disks, |
9653 | int direction) | |
78b10e66 | 9654 | { |
694575e7 KW |
9655 | /* currently we only support increasing the number of devices |
9656 | * for a container. This increases the number of device for each | |
9657 | * member array. They must all be RAID0 or RAID5. | |
9658 | */ | |
78b10e66 N |
9659 | int ret_val = 0; |
9660 | struct mdinfo *info, *member; | |
9661 | int devices_that_can_grow = 0; | |
9662 | ||
7a862a02 | 9663 | dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): st->devnm = (%s)\n", st->devnm); |
78b10e66 | 9664 | |
d04f65f4 | 9665 | if (geo->size > 0 || |
78b10e66 N |
9666 | geo->level != UnSet || |
9667 | geo->layout != UnSet || | |
9668 | geo->chunksize != 0 || | |
9669 | geo->raid_disks == UnSet) { | |
7a862a02 | 9670 | dprintf("imsm: Container operation is allowed for raid disks number change only.\n"); |
78b10e66 N |
9671 | return ret_val; |
9672 | } | |
9673 | ||
fbf3d202 | 9674 | if (direction == ROLLBACK_METADATA_CHANGES) { |
7a862a02 | 9675 | dprintf("imsm: Metadata changes rollback is not supported for container operation.\n"); |
fbf3d202 AK |
9676 | return ret_val; |
9677 | } | |
9678 | ||
78b10e66 N |
9679 | info = container_content_imsm(st, NULL); |
9680 | for (member = info; member; member = member->next) { | |
4dd2df09 | 9681 | char *result; |
78b10e66 N |
9682 | |
9683 | dprintf("imsm: checking device_num: %i\n", | |
9684 | member->container_member); | |
9685 | ||
d7d205bd | 9686 | if (geo->raid_disks <= member->array.raid_disks) { |
78b10e66 N |
9687 | /* we work on container for Online Capacity Expansion |
9688 | * only so raid_disks has to grow | |
9689 | */ | |
7a862a02 | 9690 | dprintf("imsm: for container operation raid disks increase is required\n"); |
78b10e66 N |
9691 | break; |
9692 | } | |
9693 | ||
9694 | if ((info->array.level != 0) && | |
9695 | (info->array.level != 5)) { | |
9696 | /* we cannot use this container with other raid level | |
9697 | */ | |
7a862a02 | 9698 | dprintf("imsm: for container operation wrong raid level (%i) detected\n", |
78b10e66 N |
9699 | info->array.level); |
9700 | break; | |
9701 | } else { | |
9702 | /* check for platform support | |
9703 | * for this raid level configuration | |
9704 | */ | |
9705 | struct intel_super *super = st->sb; | |
9706 | if (!is_raid_level_supported(super->orom, | |
9707 | member->array.level, | |
9708 | geo->raid_disks)) { | |
7a862a02 | 9709 | dprintf("platform does not support raid%d with %d disk%s\n", |
78b10e66 N |
9710 | info->array.level, |
9711 | geo->raid_disks, | |
9712 | geo->raid_disks > 1 ? "s" : ""); | |
9713 | break; | |
9714 | } | |
2a4a08e7 AK |
9715 | /* check if component size is aligned to chunk size |
9716 | */ | |
9717 | if (info->component_size % | |
9718 | (info->array.chunk_size/512)) { | |
7a862a02 | 9719 | dprintf("Component size is not aligned to chunk size\n"); |
2a4a08e7 AK |
9720 | break; |
9721 | } | |
78b10e66 N |
9722 | } |
9723 | ||
9724 | if (*old_raid_disks && | |
9725 | info->array.raid_disks != *old_raid_disks) | |
9726 | break; | |
9727 | *old_raid_disks = info->array.raid_disks; | |
9728 | ||
9729 | /* All raid5 and raid0 volumes in container | |
9730 | * have to be ready for Online Capacity Expansion | |
9731 | * so they need to be assembled. We have already | |
9732 | * checked that no recovery etc is happening. | |
9733 | */ | |
4dd2df09 N |
9734 | result = imsm_find_array_devnm_by_subdev(member->container_member, |
9735 | st->container_devnm); | |
9736 | if (result == NULL) { | |
78b10e66 N |
9737 | dprintf("imsm: cannot find array\n"); |
9738 | break; | |
9739 | } | |
9740 | devices_that_can_grow++; | |
9741 | } | |
9742 | sysfs_free(info); | |
9743 | if (!member && devices_that_can_grow) | |
9744 | ret_val = 1; | |
9745 | ||
9746 | if (ret_val) | |
1ade5cc1 | 9747 | dprintf("Container operation allowed\n"); |
78b10e66 | 9748 | else |
1ade5cc1 | 9749 | dprintf("Error: %i\n", ret_val); |
78b10e66 N |
9750 | |
9751 | return ret_val; | |
9752 | } | |
9753 | ||
9754 | /* Function: get_spares_for_grow | |
9755 | * Description: Allocates memory and creates list of spare devices | |
1011e834 | 9756 | * avaliable in container. Checks if spare drive size is acceptable. |
78b10e66 N |
9757 | * Parameters: Pointer to the supertype structure |
9758 | * Returns: Pointer to the list of spare devices (mdinfo structure) on success, | |
1011e834 | 9759 | * NULL if fail |
78b10e66 N |
9760 | */ |
9761 | static struct mdinfo *get_spares_for_grow(struct supertype *st) | |
9762 | { | |
78b10e66 | 9763 | unsigned long long min_size = min_acceptable_spare_size_imsm(st); |
326727d9 | 9764 | return container_choose_spares(st, min_size, NULL, NULL, NULL, 0); |
78b10e66 N |
9765 | } |
9766 | ||
9767 | /****************************************************************************** | |
9768 | * function: imsm_create_metadata_update_for_reshape | |
9769 | * Function creates update for whole IMSM container. | |
9770 | * | |
9771 | ******************************************************************************/ | |
9772 | static int imsm_create_metadata_update_for_reshape( | |
9773 | struct supertype *st, | |
9774 | struct geo_params *geo, | |
9775 | int old_raid_disks, | |
9776 | struct imsm_update_reshape **updatep) | |
9777 | { | |
9778 | struct intel_super *super = st->sb; | |
9779 | struct imsm_super *mpb = super->anchor; | |
9780 | int update_memory_size = 0; | |
9781 | struct imsm_update_reshape *u = NULL; | |
9782 | struct mdinfo *spares = NULL; | |
9783 | int i; | |
9784 | int delta_disks = 0; | |
bbd24d86 | 9785 | struct mdinfo *dev; |
78b10e66 | 9786 | |
1ade5cc1 | 9787 | dprintf("(enter) raid_disks = %i\n", geo->raid_disks); |
78b10e66 N |
9788 | |
9789 | delta_disks = geo->raid_disks - old_raid_disks; | |
9790 | ||
9791 | /* size of all update data without anchor */ | |
9792 | update_memory_size = sizeof(struct imsm_update_reshape); | |
9793 | ||
9794 | /* now add space for spare disks that we need to add. */ | |
9795 | update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1); | |
9796 | ||
503975b9 | 9797 | u = xcalloc(1, update_memory_size); |
78b10e66 N |
9798 | u->type = update_reshape_container_disks; |
9799 | u->old_raid_disks = old_raid_disks; | |
9800 | u->new_raid_disks = geo->raid_disks; | |
9801 | ||
9802 | /* now get spare disks list | |
9803 | */ | |
9804 | spares = get_spares_for_grow(st); | |
9805 | ||
9806 | if (spares == NULL | |
9807 | || delta_disks > spares->array.spare_disks) { | |
7a862a02 | 9808 | pr_err("imsm: ERROR: Cannot get spare devices for %s.\n", geo->dev_name); |
e4c72d1d | 9809 | i = -1; |
78b10e66 N |
9810 | goto abort; |
9811 | } | |
9812 | ||
9813 | /* we have got spares | |
9814 | * update disk list in imsm_disk list table in anchor | |
9815 | */ | |
9816 | dprintf("imsm: %i spares are available.\n\n", | |
9817 | spares->array.spare_disks); | |
9818 | ||
bbd24d86 | 9819 | dev = spares->devs; |
78b10e66 | 9820 | for (i = 0; i < delta_disks; i++) { |
78b10e66 N |
9821 | struct dl *dl; |
9822 | ||
bbd24d86 AK |
9823 | if (dev == NULL) |
9824 | break; | |
78b10e66 N |
9825 | u->new_disks[i] = makedev(dev->disk.major, |
9826 | dev->disk.minor); | |
9827 | dl = get_disk_super(super, dev->disk.major, dev->disk.minor); | |
ee4beede AK |
9828 | dl->index = mpb->num_disks; |
9829 | mpb->num_disks++; | |
bbd24d86 | 9830 | dev = dev->next; |
78b10e66 | 9831 | } |
78b10e66 N |
9832 | |
9833 | abort: | |
9834 | /* free spares | |
9835 | */ | |
9836 | sysfs_free(spares); | |
9837 | ||
d677e0b8 | 9838 | dprintf("imsm: reshape update preparation :"); |
78b10e66 | 9839 | if (i == delta_disks) { |
1ade5cc1 | 9840 | dprintf_cont(" OK\n"); |
78b10e66 N |
9841 | *updatep = u; |
9842 | return update_memory_size; | |
9843 | } | |
9844 | free(u); | |
1ade5cc1 | 9845 | dprintf_cont(" Error\n"); |
78b10e66 N |
9846 | |
9847 | return 0; | |
9848 | } | |
9849 | ||
f3871fdc AK |
9850 | /****************************************************************************** |
9851 | * function: imsm_create_metadata_update_for_size_change() | |
9852 | * Creates update for IMSM array for array size change. | |
9853 | * | |
9854 | ******************************************************************************/ | |
9855 | static int imsm_create_metadata_update_for_size_change( | |
9856 | struct supertype *st, | |
9857 | struct geo_params *geo, | |
9858 | struct imsm_update_size_change **updatep) | |
9859 | { | |
9860 | struct intel_super *super = st->sb; | |
9861 | int update_memory_size = 0; | |
9862 | struct imsm_update_size_change *u = NULL; | |
9863 | ||
1ade5cc1 | 9864 | dprintf("(enter) New size = %llu\n", geo->size); |
f3871fdc AK |
9865 | |
9866 | /* size of all update data without anchor */ | |
9867 | update_memory_size = sizeof(struct imsm_update_size_change); | |
9868 | ||
503975b9 | 9869 | u = xcalloc(1, update_memory_size); |
f3871fdc AK |
9870 | u->type = update_size_change; |
9871 | u->subdev = super->current_vol; | |
9872 | u->new_size = geo->size; | |
9873 | ||
9874 | dprintf("imsm: reshape update preparation : OK\n"); | |
9875 | *updatep = u; | |
9876 | ||
9877 | return update_memory_size; | |
9878 | } | |
9879 | ||
48c5303a PC |
9880 | /****************************************************************************** |
9881 | * function: imsm_create_metadata_update_for_migration() | |
9882 | * Creates update for IMSM array. | |
9883 | * | |
9884 | ******************************************************************************/ | |
9885 | static int imsm_create_metadata_update_for_migration( | |
9886 | struct supertype *st, | |
9887 | struct geo_params *geo, | |
9888 | struct imsm_update_reshape_migration **updatep) | |
9889 | { | |
9890 | struct intel_super *super = st->sb; | |
9891 | int update_memory_size = 0; | |
9892 | struct imsm_update_reshape_migration *u = NULL; | |
9893 | struct imsm_dev *dev; | |
9894 | int previous_level = -1; | |
9895 | ||
1ade5cc1 | 9896 | dprintf("(enter) New Level = %i\n", geo->level); |
48c5303a PC |
9897 | |
9898 | /* size of all update data without anchor */ | |
9899 | update_memory_size = sizeof(struct imsm_update_reshape_migration); | |
9900 | ||
503975b9 | 9901 | u = xcalloc(1, update_memory_size); |
48c5303a PC |
9902 | u->type = update_reshape_migration; |
9903 | u->subdev = super->current_vol; | |
9904 | u->new_level = geo->level; | |
9905 | u->new_layout = geo->layout; | |
9906 | u->new_raid_disks = u->old_raid_disks = geo->raid_disks; | |
9907 | u->new_disks[0] = -1; | |
4bba0439 | 9908 | u->new_chunksize = -1; |
48c5303a PC |
9909 | |
9910 | dev = get_imsm_dev(super, u->subdev); | |
9911 | if (dev) { | |
9912 | struct imsm_map *map; | |
9913 | ||
238c0a71 | 9914 | map = get_imsm_map(dev, MAP_0); |
4bba0439 PC |
9915 | if (map) { |
9916 | int current_chunk_size = | |
9917 | __le16_to_cpu(map->blocks_per_strip) / 2; | |
9918 | ||
9919 | if (geo->chunksize != current_chunk_size) { | |
9920 | u->new_chunksize = geo->chunksize / 1024; | |
7a862a02 | 9921 | dprintf("imsm: chunk size change from %i to %i\n", |
4bba0439 PC |
9922 | current_chunk_size, u->new_chunksize); |
9923 | } | |
48c5303a | 9924 | previous_level = map->raid_level; |
4bba0439 | 9925 | } |
48c5303a PC |
9926 | } |
9927 | if ((geo->level == 5) && (previous_level == 0)) { | |
9928 | struct mdinfo *spares = NULL; | |
9929 | ||
9930 | u->new_raid_disks++; | |
9931 | spares = get_spares_for_grow(st); | |
9932 | if ((spares == NULL) || (spares->array.spare_disks < 1)) { | |
9933 | free(u); | |
9934 | sysfs_free(spares); | |
9935 | update_memory_size = 0; | |
7a862a02 | 9936 | dprintf("error: cannot get spare device for requested migration"); |
48c5303a PC |
9937 | return 0; |
9938 | } | |
9939 | sysfs_free(spares); | |
9940 | } | |
9941 | dprintf("imsm: reshape update preparation : OK\n"); | |
9942 | *updatep = u; | |
9943 | ||
9944 | return update_memory_size; | |
9945 | } | |
9946 | ||
8dd70bce AK |
9947 | static void imsm_update_metadata_locally(struct supertype *st, |
9948 | void *buf, int len) | |
9949 | { | |
9950 | struct metadata_update mu; | |
9951 | ||
9952 | mu.buf = buf; | |
9953 | mu.len = len; | |
9954 | mu.space = NULL; | |
9955 | mu.space_list = NULL; | |
9956 | mu.next = NULL; | |
5fe6f031 N |
9957 | if (imsm_prepare_update(st, &mu)) |
9958 | imsm_process_update(st, &mu); | |
8dd70bce AK |
9959 | |
9960 | while (mu.space_list) { | |
9961 | void **space = mu.space_list; | |
9962 | mu.space_list = *space; | |
9963 | free(space); | |
9964 | } | |
9965 | } | |
78b10e66 | 9966 | |
471bceb6 | 9967 | /*************************************************************************** |
694575e7 | 9968 | * Function: imsm_analyze_change |
471bceb6 | 9969 | * Description: Function analyze change for single volume |
1011e834 | 9970 | * and validate if transition is supported |
fbf3d202 AK |
9971 | * Parameters: Geometry parameters, supertype structure, |
9972 | * metadata change direction (apply/rollback) | |
694575e7 | 9973 | * Returns: Operation type code on success, -1 if fail |
471bceb6 KW |
9974 | ****************************************************************************/ |
9975 | enum imsm_reshape_type imsm_analyze_change(struct supertype *st, | |
fbf3d202 AK |
9976 | struct geo_params *geo, |
9977 | int direction) | |
694575e7 | 9978 | { |
471bceb6 KW |
9979 | struct mdinfo info; |
9980 | int change = -1; | |
9981 | int check_devs = 0; | |
c21e737b | 9982 | int chunk; |
67a2db32 AK |
9983 | /* number of added/removed disks in operation result */ |
9984 | int devNumChange = 0; | |
9985 | /* imsm compatible layout value for array geometry verification */ | |
9986 | int imsm_layout = -1; | |
7abc9871 AK |
9987 | int data_disks; |
9988 | struct imsm_dev *dev; | |
9989 | struct intel_super *super; | |
d04f65f4 | 9990 | unsigned long long current_size; |
65d38cca | 9991 | unsigned long long free_size; |
d04f65f4 | 9992 | unsigned long long max_size; |
65d38cca | 9993 | int rv; |
471bceb6 KW |
9994 | |
9995 | getinfo_super_imsm_volume(st, &info, NULL); | |
471bceb6 KW |
9996 | if ((geo->level != info.array.level) && |
9997 | (geo->level >= 0) && | |
9998 | (geo->level != UnSet)) { | |
9999 | switch (info.array.level) { | |
10000 | case 0: | |
10001 | if (geo->level == 5) { | |
b5347799 | 10002 | change = CH_MIGRATION; |
e13ce846 | 10003 | if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) { |
7a862a02 | 10004 | pr_err("Error. Requested Layout not supported (left-asymmetric layout is supported only)!\n"); |
e13ce846 AK |
10005 | change = -1; |
10006 | goto analyse_change_exit; | |
10007 | } | |
67a2db32 | 10008 | imsm_layout = geo->layout; |
471bceb6 | 10009 | check_devs = 1; |
e91a3bad LM |
10010 | devNumChange = 1; /* parity disk added */ |
10011 | } else if (geo->level == 10) { | |
471bceb6 KW |
10012 | change = CH_TAKEOVER; |
10013 | check_devs = 1; | |
e91a3bad | 10014 | devNumChange = 2; /* two mirrors added */ |
67a2db32 | 10015 | imsm_layout = 0x102; /* imsm supported layout */ |
471bceb6 | 10016 | } |
dfe77a9e KW |
10017 | break; |
10018 | case 1: | |
471bceb6 KW |
10019 | case 10: |
10020 | if (geo->level == 0) { | |
10021 | change = CH_TAKEOVER; | |
10022 | check_devs = 1; | |
e91a3bad | 10023 | devNumChange = -(geo->raid_disks/2); |
67a2db32 | 10024 | imsm_layout = 0; /* imsm raid0 layout */ |
471bceb6 KW |
10025 | } |
10026 | break; | |
10027 | } | |
10028 | if (change == -1) { | |
7a862a02 | 10029 | pr_err("Error. Level Migration from %d to %d not supported!\n", |
e7b84f9d | 10030 | info.array.level, geo->level); |
471bceb6 KW |
10031 | goto analyse_change_exit; |
10032 | } | |
10033 | } else | |
10034 | geo->level = info.array.level; | |
10035 | ||
10036 | if ((geo->layout != info.array.layout) | |
10037 | && ((geo->layout != UnSet) && (geo->layout != -1))) { | |
b5347799 | 10038 | change = CH_MIGRATION; |
471bceb6 KW |
10039 | if ((info.array.layout == 0) |
10040 | && (info.array.level == 5) | |
10041 | && (geo->layout == 5)) { | |
10042 | /* reshape 5 -> 4 */ | |
10043 | } else if ((info.array.layout == 5) | |
10044 | && (info.array.level == 5) | |
10045 | && (geo->layout == 0)) { | |
10046 | /* reshape 4 -> 5 */ | |
10047 | geo->layout = 0; | |
10048 | geo->level = 5; | |
10049 | } else { | |
7a862a02 | 10050 | pr_err("Error. Layout Migration from %d to %d not supported!\n", |
e7b84f9d | 10051 | info.array.layout, geo->layout); |
471bceb6 KW |
10052 | change = -1; |
10053 | goto analyse_change_exit; | |
10054 | } | |
67a2db32 | 10055 | } else { |
471bceb6 | 10056 | geo->layout = info.array.layout; |
67a2db32 AK |
10057 | if (imsm_layout == -1) |
10058 | imsm_layout = info.array.layout; | |
10059 | } | |
471bceb6 KW |
10060 | |
10061 | if ((geo->chunksize > 0) && (geo->chunksize != UnSet) | |
10062 | && (geo->chunksize != info.array.chunk_size)) | |
b5347799 | 10063 | change = CH_MIGRATION; |
471bceb6 KW |
10064 | else |
10065 | geo->chunksize = info.array.chunk_size; | |
10066 | ||
c21e737b | 10067 | chunk = geo->chunksize / 1024; |
7abc9871 AK |
10068 | |
10069 | super = st->sb; | |
10070 | dev = get_imsm_dev(super, super->current_vol); | |
10071 | data_disks = imsm_num_data_members(dev , MAP_0); | |
c41e00b2 | 10072 | /* compute current size per disk member |
7abc9871 | 10073 | */ |
c41e00b2 AK |
10074 | current_size = info.custom_array_size / data_disks; |
10075 | ||
d5d2c614 | 10076 | if ((geo->size > 0) && (geo->size != MAX_SIZE)) { |
c41e00b2 AK |
10077 | /* align component size |
10078 | */ | |
10079 | geo->size = imsm_component_size_aligment_check( | |
10080 | get_imsm_raid_level(dev->vol.map), | |
10081 | chunk * 1024, | |
10082 | geo->size * 2); | |
65d0b4ce | 10083 | if (geo->size == 0) { |
7a862a02 | 10084 | pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is 0).\n", |
65d0b4ce LD |
10085 | current_size); |
10086 | goto analyse_change_exit; | |
10087 | } | |
c41e00b2 | 10088 | } |
7abc9871 | 10089 | |
d04f65f4 | 10090 | if ((current_size != geo->size) && (geo->size > 0)) { |
7abc9871 | 10091 | if (change != -1) { |
7a862a02 | 10092 | pr_err("Error. Size change should be the only one at a time.\n"); |
7abc9871 AK |
10093 | change = -1; |
10094 | goto analyse_change_exit; | |
10095 | } | |
10096 | if ((super->current_vol + 1) != super->anchor->num_raid_devs) { | |
7a862a02 | 10097 | pr_err("Error. The last volume in container can be expanded only (%i/%s).\n", |
4dd2df09 | 10098 | super->current_vol, st->devnm); |
7abc9871 AK |
10099 | goto analyse_change_exit; |
10100 | } | |
65d38cca LD |
10101 | /* check the maximum available size |
10102 | */ | |
10103 | rv = imsm_get_free_size(st, dev->vol.map->num_members, | |
10104 | 0, chunk, &free_size); | |
10105 | if (rv == 0) | |
10106 | /* Cannot find maximum available space | |
10107 | */ | |
10108 | max_size = 0; | |
10109 | else { | |
10110 | max_size = free_size + current_size; | |
10111 | /* align component size | |
10112 | */ | |
10113 | max_size = imsm_component_size_aligment_check( | |
10114 | get_imsm_raid_level(dev->vol.map), | |
10115 | chunk * 1024, | |
10116 | max_size); | |
10117 | } | |
d04f65f4 | 10118 | if (geo->size == MAX_SIZE) { |
b130333f AK |
10119 | /* requested size change to the maximum available size |
10120 | */ | |
65d38cca | 10121 | if (max_size == 0) { |
7a862a02 | 10122 | pr_err("Error. Cannot find maximum available space.\n"); |
b130333f AK |
10123 | change = -1; |
10124 | goto analyse_change_exit; | |
65d38cca LD |
10125 | } else |
10126 | geo->size = max_size; | |
c41e00b2 | 10127 | } |
b130333f | 10128 | |
fbf3d202 AK |
10129 | if ((direction == ROLLBACK_METADATA_CHANGES)) { |
10130 | /* accept size for rollback only | |
10131 | */ | |
10132 | } else { | |
10133 | /* round size due to metadata compatibility | |
10134 | */ | |
10135 | geo->size = (geo->size >> SECT_PER_MB_SHIFT) | |
10136 | << SECT_PER_MB_SHIFT; | |
10137 | dprintf("Prepare update for size change to %llu\n", | |
10138 | geo->size ); | |
10139 | if (current_size >= geo->size) { | |
7a862a02 | 10140 | pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is %llu).\n", |
e7b84f9d | 10141 | current_size, geo->size); |
fbf3d202 AK |
10142 | goto analyse_change_exit; |
10143 | } | |
65d38cca | 10144 | if (max_size && geo->size > max_size) { |
7a862a02 | 10145 | pr_err("Error. Requested size is larger than maximum available size (maximum available size is %llu, requested size /rounded/ is %llu).\n", |
e7b84f9d | 10146 | max_size, geo->size); |
65d38cca LD |
10147 | goto analyse_change_exit; |
10148 | } | |
7abc9871 AK |
10149 | } |
10150 | geo->size *= data_disks; | |
10151 | geo->raid_disks = dev->vol.map->num_members; | |
10152 | change = CH_ARRAY_SIZE; | |
10153 | } | |
471bceb6 KW |
10154 | if (!validate_geometry_imsm(st, |
10155 | geo->level, | |
67a2db32 | 10156 | imsm_layout, |
e91a3bad | 10157 | geo->raid_disks + devNumChange, |
c21e737b | 10158 | &chunk, |
af4348dd | 10159 | geo->size, INVALID_SECTORS, |
471bceb6 KW |
10160 | 0, 0, 1)) |
10161 | change = -1; | |
10162 | ||
10163 | if (check_devs) { | |
10164 | struct intel_super *super = st->sb; | |
10165 | struct imsm_super *mpb = super->anchor; | |
10166 | ||
10167 | if (mpb->num_raid_devs > 1) { | |
7a862a02 | 10168 | pr_err("Error. Cannot perform operation on %s- for this operation it MUST be single array in container\n", |
e7b84f9d | 10169 | geo->dev_name); |
471bceb6 KW |
10170 | change = -1; |
10171 | } | |
10172 | } | |
10173 | ||
10174 | analyse_change_exit: | |
fbf3d202 AK |
10175 | if ((direction == ROLLBACK_METADATA_CHANGES) && |
10176 | ((change == CH_MIGRATION) || (change == CH_TAKEOVER))) { | |
7a862a02 | 10177 | dprintf("imsm: Metadata changes rollback is not supported for migration and takeover operations.\n"); |
fbf3d202 AK |
10178 | change = -1; |
10179 | } | |
471bceb6 | 10180 | return change; |
694575e7 KW |
10181 | } |
10182 | ||
bb025c2f KW |
10183 | int imsm_takeover(struct supertype *st, struct geo_params *geo) |
10184 | { | |
10185 | struct intel_super *super = st->sb; | |
10186 | struct imsm_update_takeover *u; | |
10187 | ||
503975b9 | 10188 | u = xmalloc(sizeof(struct imsm_update_takeover)); |
bb025c2f KW |
10189 | |
10190 | u->type = update_takeover; | |
10191 | u->subarray = super->current_vol; | |
10192 | ||
10193 | /* 10->0 transition */ | |
10194 | if (geo->level == 0) | |
10195 | u->direction = R10_TO_R0; | |
10196 | ||
0529c688 KW |
10197 | /* 0->10 transition */ |
10198 | if (geo->level == 10) | |
10199 | u->direction = R0_TO_R10; | |
10200 | ||
bb025c2f KW |
10201 | /* update metadata locally */ |
10202 | imsm_update_metadata_locally(st, u, | |
10203 | sizeof(struct imsm_update_takeover)); | |
10204 | /* and possibly remotely */ | |
10205 | if (st->update_tail) | |
10206 | append_metadata_update(st, u, | |
10207 | sizeof(struct imsm_update_takeover)); | |
10208 | else | |
10209 | free(u); | |
10210 | ||
10211 | return 0; | |
10212 | } | |
10213 | ||
d04f65f4 N |
10214 | static int imsm_reshape_super(struct supertype *st, unsigned long long size, |
10215 | int level, | |
78b10e66 | 10216 | int layout, int chunksize, int raid_disks, |
41784c88 | 10217 | int delta_disks, char *backup, char *dev, |
016e00f5 | 10218 | int direction, int verbose) |
78b10e66 | 10219 | { |
78b10e66 N |
10220 | int ret_val = 1; |
10221 | struct geo_params geo; | |
10222 | ||
1ade5cc1 | 10223 | dprintf("(enter)\n"); |
78b10e66 | 10224 | |
71204a50 | 10225 | memset(&geo, 0, sizeof(struct geo_params)); |
78b10e66 N |
10226 | |
10227 | geo.dev_name = dev; | |
4dd2df09 | 10228 | strcpy(geo.devnm, st->devnm); |
78b10e66 N |
10229 | geo.size = size; |
10230 | geo.level = level; | |
10231 | geo.layout = layout; | |
10232 | geo.chunksize = chunksize; | |
10233 | geo.raid_disks = raid_disks; | |
41784c88 AK |
10234 | if (delta_disks != UnSet) |
10235 | geo.raid_disks += delta_disks; | |
78b10e66 | 10236 | |
1ade5cc1 N |
10237 | dprintf("for level : %i\n", geo.level); |
10238 | dprintf("for raid_disks : %i\n", geo.raid_disks); | |
78b10e66 N |
10239 | |
10240 | if (experimental() == 0) | |
10241 | return ret_val; | |
10242 | ||
4dd2df09 | 10243 | if (strcmp(st->container_devnm, st->devnm) == 0) { |
694575e7 KW |
10244 | /* On container level we can only increase number of devices. */ |
10245 | dprintf("imsm: info: Container operation\n"); | |
78b10e66 | 10246 | int old_raid_disks = 0; |
6dc0be30 | 10247 | |
78b10e66 | 10248 | if (imsm_reshape_is_allowed_on_container( |
fbf3d202 | 10249 | st, &geo, &old_raid_disks, direction)) { |
78b10e66 N |
10250 | struct imsm_update_reshape *u = NULL; |
10251 | int len; | |
10252 | ||
10253 | len = imsm_create_metadata_update_for_reshape( | |
10254 | st, &geo, old_raid_disks, &u); | |
10255 | ||
ed08d51c AK |
10256 | if (len <= 0) { |
10257 | dprintf("imsm: Cannot prepare update\n"); | |
10258 | goto exit_imsm_reshape_super; | |
10259 | } | |
10260 | ||
8dd70bce AK |
10261 | ret_val = 0; |
10262 | /* update metadata locally */ | |
10263 | imsm_update_metadata_locally(st, u, len); | |
10264 | /* and possibly remotely */ | |
10265 | if (st->update_tail) | |
10266 | append_metadata_update(st, u, len); | |
10267 | else | |
ed08d51c | 10268 | free(u); |
8dd70bce | 10269 | |
694575e7 | 10270 | } else { |
7a862a02 | 10271 | pr_err("(imsm) Operation is not allowed on this container\n"); |
694575e7 KW |
10272 | } |
10273 | } else { | |
10274 | /* On volume level we support following operations | |
471bceb6 KW |
10275 | * - takeover: raid10 -> raid0; raid0 -> raid10 |
10276 | * - chunk size migration | |
10277 | * - migration: raid5 -> raid0; raid0 -> raid5 | |
10278 | */ | |
10279 | struct intel_super *super = st->sb; | |
10280 | struct intel_dev *dev = super->devlist; | |
4dd2df09 | 10281 | int change; |
694575e7 | 10282 | dprintf("imsm: info: Volume operation\n"); |
471bceb6 KW |
10283 | /* find requested device */ |
10284 | while (dev) { | |
1011e834 | 10285 | char *devnm = |
4dd2df09 N |
10286 | imsm_find_array_devnm_by_subdev( |
10287 | dev->index, st->container_devnm); | |
10288 | if (devnm && strcmp(devnm, geo.devnm) == 0) | |
471bceb6 KW |
10289 | break; |
10290 | dev = dev->next; | |
10291 | } | |
10292 | if (dev == NULL) { | |
4dd2df09 N |
10293 | pr_err("Cannot find %s (%s) subarray\n", |
10294 | geo.dev_name, geo.devnm); | |
471bceb6 KW |
10295 | goto exit_imsm_reshape_super; |
10296 | } | |
10297 | super->current_vol = dev->index; | |
fbf3d202 | 10298 | change = imsm_analyze_change(st, &geo, direction); |
694575e7 | 10299 | switch (change) { |
471bceb6 | 10300 | case CH_TAKEOVER: |
bb025c2f | 10301 | ret_val = imsm_takeover(st, &geo); |
694575e7 | 10302 | break; |
48c5303a PC |
10303 | case CH_MIGRATION: { |
10304 | struct imsm_update_reshape_migration *u = NULL; | |
10305 | int len = | |
10306 | imsm_create_metadata_update_for_migration( | |
10307 | st, &geo, &u); | |
10308 | if (len < 1) { | |
7a862a02 | 10309 | dprintf("imsm: Cannot prepare update\n"); |
48c5303a PC |
10310 | break; |
10311 | } | |
471bceb6 | 10312 | ret_val = 0; |
48c5303a PC |
10313 | /* update metadata locally */ |
10314 | imsm_update_metadata_locally(st, u, len); | |
10315 | /* and possibly remotely */ | |
10316 | if (st->update_tail) | |
10317 | append_metadata_update(st, u, len); | |
10318 | else | |
10319 | free(u); | |
10320 | } | |
10321 | break; | |
7abc9871 | 10322 | case CH_ARRAY_SIZE: { |
f3871fdc AK |
10323 | struct imsm_update_size_change *u = NULL; |
10324 | int len = | |
10325 | imsm_create_metadata_update_for_size_change( | |
10326 | st, &geo, &u); | |
10327 | if (len < 1) { | |
7a862a02 | 10328 | dprintf("imsm: Cannot prepare update\n"); |
f3871fdc AK |
10329 | break; |
10330 | } | |
10331 | ret_val = 0; | |
10332 | /* update metadata locally */ | |
10333 | imsm_update_metadata_locally(st, u, len); | |
10334 | /* and possibly remotely */ | |
10335 | if (st->update_tail) | |
10336 | append_metadata_update(st, u, len); | |
10337 | else | |
10338 | free(u); | |
7abc9871 AK |
10339 | } |
10340 | break; | |
471bceb6 KW |
10341 | default: |
10342 | ret_val = 1; | |
694575e7 | 10343 | } |
694575e7 | 10344 | } |
78b10e66 | 10345 | |
ed08d51c | 10346 | exit_imsm_reshape_super: |
78b10e66 N |
10347 | dprintf("imsm: reshape_super Exit code = %i\n", ret_val); |
10348 | return ret_val; | |
10349 | } | |
2cda7640 | 10350 | |
eee67a47 AK |
10351 | /******************************************************************************* |
10352 | * Function: wait_for_reshape_imsm | |
10353 | * Description: Function writes new sync_max value and waits until | |
10354 | * reshape process reach new position | |
10355 | * Parameters: | |
10356 | * sra : general array info | |
eee67a47 AK |
10357 | * ndata : number of disks in new array's layout |
10358 | * Returns: | |
10359 | * 0 : success, | |
10360 | * 1 : there is no reshape in progress, | |
10361 | * -1 : fail | |
10362 | ******************************************************************************/ | |
ae9f01f8 | 10363 | int wait_for_reshape_imsm(struct mdinfo *sra, int ndata) |
eee67a47 | 10364 | { |
85ca499c | 10365 | int fd = sysfs_get_fd(sra, NULL, "sync_completed"); |
eee67a47 | 10366 | unsigned long long completed; |
ae9f01f8 AK |
10367 | /* to_complete : new sync_max position */ |
10368 | unsigned long long to_complete = sra->reshape_progress; | |
10369 | unsigned long long position_to_set = to_complete / ndata; | |
eee67a47 | 10370 | |
ae9f01f8 | 10371 | if (fd < 0) { |
1ade5cc1 | 10372 | dprintf("cannot open reshape_position\n"); |
eee67a47 | 10373 | return 1; |
ae9f01f8 | 10374 | } |
eee67a47 | 10375 | |
ae9f01f8 | 10376 | if (sysfs_fd_get_ll(fd, &completed) < 0) { |
1ade5cc1 | 10377 | dprintf("cannot read reshape_position (no reshape in progres)\n"); |
ae9f01f8 | 10378 | close(fd); |
c85338c6 | 10379 | return 1; |
ae9f01f8 | 10380 | } |
eee67a47 | 10381 | |
85ca499c | 10382 | if (completed > position_to_set) { |
1ade5cc1 | 10383 | dprintf("wrong next position to set %llu (%llu)\n", |
85ca499c | 10384 | to_complete, position_to_set); |
ae9f01f8 AK |
10385 | close(fd); |
10386 | return -1; | |
10387 | } | |
10388 | dprintf("Position set: %llu\n", position_to_set); | |
10389 | if (sysfs_set_num(sra, NULL, "sync_max", | |
10390 | position_to_set) != 0) { | |
1ade5cc1 | 10391 | dprintf("cannot set reshape position to %llu\n", |
ae9f01f8 AK |
10392 | position_to_set); |
10393 | close(fd); | |
10394 | return -1; | |
eee67a47 AK |
10395 | } |
10396 | ||
eee67a47 AK |
10397 | do { |
10398 | char action[20]; | |
5ff3a780 AP |
10399 | int timeout = 3000; |
10400 | sysfs_wait(fd, &timeout); | |
a47e44fb AK |
10401 | if (sysfs_get_str(sra, NULL, "sync_action", |
10402 | action, 20) > 0 && | |
d7d3809a AP |
10403 | strncmp(action, "reshape", 7) != 0) { |
10404 | close(fd); | |
10405 | return -1; | |
10406 | } | |
eee67a47 | 10407 | if (sysfs_fd_get_ll(fd, &completed) < 0) { |
1ade5cc1 | 10408 | dprintf("cannot read reshape_position (in loop)\n"); |
eee67a47 AK |
10409 | close(fd); |
10410 | return 1; | |
10411 | } | |
85ca499c | 10412 | } while (completed < position_to_set); |
eee67a47 AK |
10413 | close(fd); |
10414 | return 0; | |
10415 | ||
10416 | } | |
10417 | ||
b915c95f AK |
10418 | /******************************************************************************* |
10419 | * Function: check_degradation_change | |
10420 | * Description: Check that array hasn't become failed. | |
10421 | * Parameters: | |
10422 | * info : for sysfs access | |
10423 | * sources : source disks descriptors | |
10424 | * degraded: previous degradation level | |
10425 | * Returns: | |
10426 | * degradation level | |
10427 | ******************************************************************************/ | |
10428 | int check_degradation_change(struct mdinfo *info, | |
10429 | int *sources, | |
10430 | int degraded) | |
10431 | { | |
10432 | unsigned long long new_degraded; | |
e1993023 LD |
10433 | int rv; |
10434 | ||
10435 | rv = sysfs_get_ll(info, NULL, "degraded", &new_degraded); | |
10436 | if ((rv == -1) || (new_degraded != (unsigned long long)degraded)) { | |
b915c95f AK |
10437 | /* check each device to ensure it is still working */ |
10438 | struct mdinfo *sd; | |
10439 | new_degraded = 0; | |
10440 | for (sd = info->devs ; sd ; sd = sd->next) { | |
10441 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) | |
10442 | continue; | |
10443 | if (sd->disk.state & (1<<MD_DISK_SYNC)) { | |
10444 | char sbuf[20]; | |
10445 | if (sysfs_get_str(info, | |
10446 | sd, "state", sbuf, 20) < 0 || | |
10447 | strstr(sbuf, "faulty") || | |
10448 | strstr(sbuf, "in_sync") == NULL) { | |
10449 | /* this device is dead */ | |
10450 | sd->disk.state = (1<<MD_DISK_FAULTY); | |
10451 | if (sd->disk.raid_disk >= 0 && | |
10452 | sources[sd->disk.raid_disk] >= 0) { | |
10453 | close(sources[ | |
10454 | sd->disk.raid_disk]); | |
10455 | sources[sd->disk.raid_disk] = | |
10456 | -1; | |
10457 | } | |
10458 | new_degraded++; | |
10459 | } | |
10460 | } | |
10461 | } | |
10462 | } | |
10463 | ||
10464 | return new_degraded; | |
10465 | } | |
10466 | ||
10f22854 AK |
10467 | /******************************************************************************* |
10468 | * Function: imsm_manage_reshape | |
10469 | * Description: Function finds array under reshape and it manages reshape | |
10470 | * process. It creates stripes backups (if required) and sets | |
942e1cdb | 10471 | * checkpoints. |
10f22854 AK |
10472 | * Parameters: |
10473 | * afd : Backup handle (nattive) - not used | |
10474 | * sra : general array info | |
10475 | * reshape : reshape parameters - not used | |
10476 | * st : supertype structure | |
10477 | * blocks : size of critical section [blocks] | |
10478 | * fds : table of source device descriptor | |
10479 | * offsets : start of array (offest per devices) | |
10480 | * dests : not used | |
10481 | * destfd : table of destination device descriptor | |
10482 | * destoffsets : table of destination offsets (per device) | |
10483 | * Returns: | |
10484 | * 1 : success, reshape is done | |
10485 | * 0 : fail | |
10486 | ******************************************************************************/ | |
999b4972 N |
10487 | static int imsm_manage_reshape( |
10488 | int afd, struct mdinfo *sra, struct reshape *reshape, | |
10f22854 | 10489 | struct supertype *st, unsigned long backup_blocks, |
999b4972 N |
10490 | int *fds, unsigned long long *offsets, |
10491 | int dests, int *destfd, unsigned long long *destoffsets) | |
10492 | { | |
10f22854 AK |
10493 | int ret_val = 0; |
10494 | struct intel_super *super = st->sb; | |
10495 | struct intel_dev *dv = NULL; | |
10496 | struct imsm_dev *dev = NULL; | |
a6b6d984 | 10497 | struct imsm_map *map_src; |
10f22854 AK |
10498 | int migr_vol_qan = 0; |
10499 | int ndata, odata; /* [bytes] */ | |
10500 | int chunk; /* [bytes] */ | |
10501 | struct migr_record *migr_rec; | |
10502 | char *buf = NULL; | |
10503 | unsigned int buf_size; /* [bytes] */ | |
10504 | unsigned long long max_position; /* array size [bytes] */ | |
10505 | unsigned long long next_step; /* [blocks]/[bytes] */ | |
10506 | unsigned long long old_data_stripe_length; | |
10f22854 AK |
10507 | unsigned long long start_src; /* [bytes] */ |
10508 | unsigned long long start; /* [bytes] */ | |
10509 | unsigned long long start_buf_shift; /* [bytes] */ | |
b915c95f | 10510 | int degraded = 0; |
ab724b98 | 10511 | int source_layout = 0; |
10f22854 | 10512 | |
1ab242d8 | 10513 | if (!fds || !offsets || !sra) |
10f22854 AK |
10514 | goto abort; |
10515 | ||
10516 | /* Find volume during the reshape */ | |
10517 | for (dv = super->devlist; dv; dv = dv->next) { | |
10518 | if (dv->dev->vol.migr_type == MIGR_GEN_MIGR | |
10519 | && dv->dev->vol.migr_state == 1) { | |
10520 | dev = dv->dev; | |
10521 | migr_vol_qan++; | |
10522 | } | |
10523 | } | |
10524 | /* Only one volume can migrate at the same time */ | |
10525 | if (migr_vol_qan != 1) { | |
e7b84f9d | 10526 | pr_err(": %s", migr_vol_qan ? |
10f22854 AK |
10527 | "Number of migrating volumes greater than 1\n" : |
10528 | "There is no volume during migrationg\n"); | |
10529 | goto abort; | |
10530 | } | |
10531 | ||
238c0a71 | 10532 | map_src = get_imsm_map(dev, MAP_1); |
10f22854 AK |
10533 | if (map_src == NULL) |
10534 | goto abort; | |
10f22854 | 10535 | |
238c0a71 AK |
10536 | ndata = imsm_num_data_members(dev, MAP_0); |
10537 | odata = imsm_num_data_members(dev, MAP_1); | |
10f22854 | 10538 | |
7b1ab482 | 10539 | chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512; |
10f22854 AK |
10540 | old_data_stripe_length = odata * chunk; |
10541 | ||
10542 | migr_rec = super->migr_rec; | |
10543 | ||
10f22854 AK |
10544 | /* initialize migration record for start condition */ |
10545 | if (sra->reshape_progress == 0) | |
10546 | init_migr_record_imsm(st, dev, sra); | |
b2c59438 AK |
10547 | else { |
10548 | if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) { | |
7a862a02 | 10549 | dprintf("imsm: cannot restart migration when data are present in copy area.\n"); |
b2c59438 AK |
10550 | goto abort; |
10551 | } | |
6a75c8ca AK |
10552 | /* Save checkpoint to update migration record for current |
10553 | * reshape position (in md). It can be farther than current | |
10554 | * reshape position in metadata. | |
10555 | */ | |
10556 | if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) { | |
10557 | /* ignore error == 2, this can mean end of reshape here | |
10558 | */ | |
7a862a02 | 10559 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL, initial save)\n"); |
6a75c8ca AK |
10560 | goto abort; |
10561 | } | |
b2c59438 | 10562 | } |
10f22854 AK |
10563 | |
10564 | /* size for data */ | |
10565 | buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512; | |
10566 | /* extend buffer size for parity disk */ | |
10567 | buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
10568 | /* add space for stripe aligment */ | |
10569 | buf_size += old_data_stripe_length; | |
10570 | if (posix_memalign((void **)&buf, 4096, buf_size)) { | |
10571 | dprintf("imsm: Cannot allocate checpoint buffer\n"); | |
10572 | goto abort; | |
10573 | } | |
10574 | ||
3ef4403c | 10575 | max_position = sra->component_size * ndata; |
68eb8bc6 | 10576 | source_layout = imsm_level_to_layout(map_src->raid_level); |
10f22854 AK |
10577 | |
10578 | while (__le32_to_cpu(migr_rec->curr_migr_unit) < | |
10579 | __le32_to_cpu(migr_rec->num_migr_units)) { | |
10580 | /* current reshape position [blocks] */ | |
10581 | unsigned long long current_position = | |
10582 | __le32_to_cpu(migr_rec->blocks_per_unit) | |
10583 | * __le32_to_cpu(migr_rec->curr_migr_unit); | |
10584 | unsigned long long border; | |
10585 | ||
b915c95f AK |
10586 | /* Check that array hasn't become failed. |
10587 | */ | |
10588 | degraded = check_degradation_change(sra, fds, degraded); | |
10589 | if (degraded > 1) { | |
7a862a02 | 10590 | dprintf("imsm: Abort reshape due to degradation level (%i)\n", degraded); |
b915c95f AK |
10591 | goto abort; |
10592 | } | |
10593 | ||
10f22854 AK |
10594 | next_step = __le32_to_cpu(migr_rec->blocks_per_unit); |
10595 | ||
10596 | if ((current_position + next_step) > max_position) | |
10597 | next_step = max_position - current_position; | |
10598 | ||
92144abf | 10599 | start = current_position * 512; |
10f22854 | 10600 | |
942e1cdb | 10601 | /* align reading start to old geometry */ |
10f22854 AK |
10602 | start_buf_shift = start % old_data_stripe_length; |
10603 | start_src = start - start_buf_shift; | |
10604 | ||
10605 | border = (start_src / odata) - (start / ndata); | |
10606 | border /= 512; | |
10607 | if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) { | |
10608 | /* save critical stripes to buf | |
10609 | * start - start address of current unit | |
10610 | * to backup [bytes] | |
10611 | * start_src - start address of current unit | |
10612 | * to backup alligned to source array | |
10613 | * [bytes] | |
10614 | */ | |
10615 | unsigned long long next_step_filler = 0; | |
10616 | unsigned long long copy_length = next_step * 512; | |
10617 | ||
10618 | /* allign copy area length to stripe in old geometry */ | |
10619 | next_step_filler = ((copy_length + start_buf_shift) | |
10620 | % old_data_stripe_length); | |
10621 | if (next_step_filler) | |
10622 | next_step_filler = (old_data_stripe_length | |
10623 | - next_step_filler); | |
7a862a02 | 10624 | 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 |
10625 | start, start_src, copy_length, |
10626 | start_buf_shift, next_step_filler); | |
10627 | ||
10628 | if (save_stripes(fds, offsets, map_src->num_members, | |
ab724b98 AK |
10629 | chunk, map_src->raid_level, |
10630 | source_layout, 0, NULL, start_src, | |
10f22854 AK |
10631 | copy_length + |
10632 | next_step_filler + start_buf_shift, | |
10633 | buf)) { | |
7a862a02 | 10634 | dprintf("imsm: Cannot save stripes to buffer\n"); |
10f22854 AK |
10635 | goto abort; |
10636 | } | |
10637 | /* Convert data to destination format and store it | |
10638 | * in backup general migration area | |
10639 | */ | |
10640 | if (save_backup_imsm(st, dev, sra, | |
aea93171 | 10641 | buf + start_buf_shift, copy_length)) { |
7a862a02 | 10642 | dprintf("imsm: Cannot save stripes to target devices\n"); |
10f22854 AK |
10643 | goto abort; |
10644 | } | |
10645 | if (save_checkpoint_imsm(st, sra, | |
10646 | UNIT_SRC_IN_CP_AREA)) { | |
7a862a02 | 10647 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_IN_CP_AREA)\n"); |
10f22854 AK |
10648 | goto abort; |
10649 | } | |
8016a6d4 AK |
10650 | } else { |
10651 | /* set next step to use whole border area */ | |
10652 | border /= next_step; | |
10653 | if (border > 1) | |
10654 | next_step *= border; | |
10f22854 AK |
10655 | } |
10656 | /* When data backed up, checkpoint stored, | |
10657 | * kick the kernel to reshape unit of data | |
10658 | */ | |
10659 | next_step = next_step + sra->reshape_progress; | |
8016a6d4 AK |
10660 | /* limit next step to array max position */ |
10661 | if (next_step > max_position) | |
10662 | next_step = max_position; | |
10f22854 AK |
10663 | sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress); |
10664 | sysfs_set_num(sra, NULL, "suspend_hi", next_step); | |
ae9f01f8 | 10665 | sra->reshape_progress = next_step; |
10f22854 AK |
10666 | |
10667 | /* wait until reshape finish */ | |
c85338c6 | 10668 | if (wait_for_reshape_imsm(sra, ndata)) { |
c47b0ff6 AK |
10669 | dprintf("wait_for_reshape_imsm returned error!\n"); |
10670 | goto abort; | |
10671 | } | |
84d11e6c N |
10672 | if (sigterm) |
10673 | goto abort; | |
10f22854 | 10674 | |
0228d92c AK |
10675 | if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) { |
10676 | /* ignore error == 2, this can mean end of reshape here | |
10677 | */ | |
7a862a02 | 10678 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL)\n"); |
10f22854 AK |
10679 | goto abort; |
10680 | } | |
10681 | ||
10682 | } | |
10683 | ||
71e5411e PB |
10684 | /* clear migr_rec on disks after successful migration */ |
10685 | struct dl *d; | |
10686 | ||
10687 | memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SIZE); | |
10688 | for (d = super->disks; d; d = d->next) { | |
10689 | if (d->index < 0 || is_failed(&d->disk)) | |
10690 | continue; | |
10691 | unsigned long long dsize; | |
10692 | ||
10693 | get_dev_size(d->fd, NULL, &dsize); | |
10694 | if (lseek64(d->fd, dsize - MIGR_REC_POSITION, | |
10695 | SEEK_SET) >= 0) { | |
10696 | if (write(d->fd, super->migr_rec_buf, | |
10697 | MIGR_REC_BUF_SIZE) != MIGR_REC_BUF_SIZE) | |
10698 | perror("Write migr_rec failed"); | |
10699 | } | |
10700 | } | |
10701 | ||
10f22854 AK |
10702 | /* return '1' if done */ |
10703 | ret_val = 1; | |
10704 | abort: | |
10705 | free(buf); | |
942e1cdb N |
10706 | /* See Grow.c: abort_reshape() for further explanation */ |
10707 | sysfs_set_num(sra, NULL, "suspend_lo", 0x7FFFFFFFFFFFFFFFULL); | |
10708 | sysfs_set_num(sra, NULL, "suspend_hi", 0); | |
10709 | sysfs_set_num(sra, NULL, "suspend_lo", 0); | |
10f22854 AK |
10710 | |
10711 | return ret_val; | |
999b4972 | 10712 | } |
0c21b485 | 10713 | |
71204a50 | 10714 | #endif /* MDASSEMBLE */ |
999b4972 | 10715 | |
cdddbdbc DW |
10716 | struct superswitch super_imsm = { |
10717 | #ifndef MDASSEMBLE | |
10718 | .examine_super = examine_super_imsm, | |
10719 | .brief_examine_super = brief_examine_super_imsm, | |
4737ae25 | 10720 | .brief_examine_subarrays = brief_examine_subarrays_imsm, |
9d84c8ea | 10721 | .export_examine_super = export_examine_super_imsm, |
cdddbdbc DW |
10722 | .detail_super = detail_super_imsm, |
10723 | .brief_detail_super = brief_detail_super_imsm, | |
bf5a934a | 10724 | .write_init_super = write_init_super_imsm, |
0e600426 N |
10725 | .validate_geometry = validate_geometry_imsm, |
10726 | .add_to_super = add_to_super_imsm, | |
1a64be56 | 10727 | .remove_from_super = remove_from_super_imsm, |
d665cc31 | 10728 | .detail_platform = detail_platform_imsm, |
e50cf220 | 10729 | .export_detail_platform = export_detail_platform_imsm, |
33414a01 | 10730 | .kill_subarray = kill_subarray_imsm, |
aa534678 | 10731 | .update_subarray = update_subarray_imsm, |
2b959fbf | 10732 | .load_container = load_container_imsm, |
71204a50 N |
10733 | .default_geometry = default_geometry_imsm, |
10734 | .get_disk_controller_domain = imsm_get_disk_controller_domain, | |
10735 | .reshape_super = imsm_reshape_super, | |
10736 | .manage_reshape = imsm_manage_reshape, | |
9e2d750d | 10737 | .recover_backup = recover_backup_imsm, |
74db60b0 | 10738 | .copy_metadata = copy_metadata_imsm, |
cdddbdbc DW |
10739 | #endif |
10740 | .match_home = match_home_imsm, | |
10741 | .uuid_from_super= uuid_from_super_imsm, | |
10742 | .getinfo_super = getinfo_super_imsm, | |
5c4cd5da | 10743 | .getinfo_super_disks = getinfo_super_disks_imsm, |
cdddbdbc DW |
10744 | .update_super = update_super_imsm, |
10745 | ||
10746 | .avail_size = avail_size_imsm, | |
80e7f8c3 | 10747 | .min_acceptable_spare_size = min_acceptable_spare_size_imsm, |
cdddbdbc DW |
10748 | |
10749 | .compare_super = compare_super_imsm, | |
10750 | ||
10751 | .load_super = load_super_imsm, | |
bf5a934a | 10752 | .init_super = init_super_imsm, |
e683ca88 | 10753 | .store_super = store_super_imsm, |
cdddbdbc DW |
10754 | .free_super = free_super_imsm, |
10755 | .match_metadata_desc = match_metadata_desc_imsm, | |
bf5a934a | 10756 | .container_content = container_content_imsm, |
0c21b485 | 10757 | .validate_container = validate_container_imsm, |
cdddbdbc | 10758 | |
cdddbdbc | 10759 | .external = 1, |
4cce4069 | 10760 | .name = "imsm", |
845dea95 | 10761 | |
0e600426 | 10762 | #ifndef MDASSEMBLE |
845dea95 NB |
10763 | /* for mdmon */ |
10764 | .open_new = imsm_open_new, | |
ed9d66aa | 10765 | .set_array_state= imsm_set_array_state, |
845dea95 NB |
10766 | .set_disk = imsm_set_disk, |
10767 | .sync_metadata = imsm_sync_metadata, | |
88758e9d | 10768 | .activate_spare = imsm_activate_spare, |
e8319a19 | 10769 | .process_update = imsm_process_update, |
8273f55e | 10770 | .prepare_update = imsm_prepare_update, |
0e600426 | 10771 | #endif /* MDASSEMBLE */ |
cdddbdbc | 10772 | }; |