]>
Commit | Line | Data |
---|---|---|
cdddbdbc DW |
1 | /* |
2 | * mdadm - Intel(R) Matrix Storage Manager Support | |
3 | * | |
a54d5262 | 4 | * Copyright (C) 2002-2008 Intel Corporation |
cdddbdbc DW |
5 | * |
6 | * This program is free software; you can redistribute it and/or modify it | |
7 | * under the terms and conditions of the GNU General Public License, | |
8 | * version 2, as published by the Free Software Foundation. | |
9 | * | |
10 | * This program is distributed in the hope it will be useful, but WITHOUT | |
11 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
12 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for | |
13 | * more details. | |
14 | * | |
15 | * You should have received a copy of the GNU General Public License along with | |
16 | * this program; if not, write to the Free Software Foundation, Inc., | |
17 | * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA. | |
18 | */ | |
19 | ||
51006d85 | 20 | #define HAVE_STDINT_H 1 |
cdddbdbc | 21 | #include "mdadm.h" |
c2a1e7da | 22 | #include "mdmon.h" |
51006d85 | 23 | #include "sha1.h" |
88c32bb1 | 24 | #include "platform-intel.h" |
cdddbdbc DW |
25 | #include <values.h> |
26 | #include <scsi/sg.h> | |
27 | #include <ctype.h> | |
d665cc31 | 28 | #include <dirent.h> |
cdddbdbc DW |
29 | |
30 | /* MPB == Metadata Parameter Block */ | |
31 | #define MPB_SIGNATURE "Intel Raid ISM Cfg Sig. " | |
32 | #define MPB_SIG_LEN (strlen(MPB_SIGNATURE)) | |
33 | #define MPB_VERSION_RAID0 "1.0.00" | |
34 | #define MPB_VERSION_RAID1 "1.1.00" | |
fe7ed8cb DW |
35 | #define MPB_VERSION_MANY_VOLUMES_PER_ARRAY "1.2.00" |
36 | #define MPB_VERSION_3OR4_DISK_ARRAY "1.2.01" | |
cdddbdbc | 37 | #define MPB_VERSION_RAID5 "1.2.02" |
fe7ed8cb DW |
38 | #define MPB_VERSION_5OR6_DISK_ARRAY "1.2.04" |
39 | #define MPB_VERSION_CNG "1.2.06" | |
40 | #define MPB_VERSION_ATTRIBS "1.3.00" | |
cdddbdbc DW |
41 | #define MAX_SIGNATURE_LENGTH 32 |
42 | #define MAX_RAID_SERIAL_LEN 16 | |
fe7ed8cb | 43 | |
19482bcc AK |
44 | /* supports RAID0 */ |
45 | #define MPB_ATTRIB_RAID0 __cpu_to_le32(0x00000001) | |
46 | /* supports RAID1 */ | |
47 | #define MPB_ATTRIB_RAID1 __cpu_to_le32(0x00000002) | |
48 | /* supports RAID10 */ | |
49 | #define MPB_ATTRIB_RAID10 __cpu_to_le32(0x00000004) | |
50 | /* supports RAID1E */ | |
51 | #define MPB_ATTRIB_RAID1E __cpu_to_le32(0x00000008) | |
52 | /* supports RAID5 */ | |
53 | #define MPB_ATTRIB_RAID5 __cpu_to_le32(0x00000010) | |
54 | /* supports RAID CNG */ | |
55 | #define MPB_ATTRIB_RAIDCNG __cpu_to_le32(0x00000020) | |
56 | /* supports expanded stripe sizes of 256K, 512K and 1MB */ | |
57 | #define MPB_ATTRIB_EXP_STRIPE_SIZE __cpu_to_le32(0x00000040) | |
58 | ||
59 | /* The OROM Support RST Caching of Volumes */ | |
60 | #define MPB_ATTRIB_NVM __cpu_to_le32(0x02000000) | |
61 | /* The OROM supports creating disks greater than 2TB */ | |
62 | #define MPB_ATTRIB_2TB_DISK __cpu_to_le32(0x04000000) | |
63 | /* The OROM supports Bad Block Management */ | |
64 | #define MPB_ATTRIB_BBM __cpu_to_le32(0x08000000) | |
65 | ||
66 | /* THe OROM Supports NVM Caching of Volumes */ | |
67 | #define MPB_ATTRIB_NEVER_USE2 __cpu_to_le32(0x10000000) | |
68 | /* The OROM supports creating volumes greater than 2TB */ | |
69 | #define MPB_ATTRIB_2TB __cpu_to_le32(0x20000000) | |
70 | /* originally for PMP, now it's wasted b/c. Never use this bit! */ | |
71 | #define MPB_ATTRIB_NEVER_USE __cpu_to_le32(0x40000000) | |
72 | /* Verify MPB contents against checksum after reading MPB */ | |
73 | #define MPB_ATTRIB_CHECKSUM_VERIFY __cpu_to_le32(0x80000000) | |
74 | ||
75 | /* Define all supported attributes that have to be accepted by mdadm | |
76 | */ | |
418f9b36 | 77 | #define MPB_ATTRIB_SUPPORTED (MPB_ATTRIB_CHECKSUM_VERIFY | \ |
19482bcc AK |
78 | MPB_ATTRIB_2TB | \ |
79 | MPB_ATTRIB_2TB_DISK | \ | |
80 | MPB_ATTRIB_RAID0 | \ | |
81 | MPB_ATTRIB_RAID1 | \ | |
82 | MPB_ATTRIB_RAID10 | \ | |
83 | MPB_ATTRIB_RAID5 | \ | |
bbab0940 TM |
84 | MPB_ATTRIB_EXP_STRIPE_SIZE | \ |
85 | MPB_ATTRIB_BBM) | |
418f9b36 N |
86 | |
87 | /* Define attributes that are unused but not harmful */ | |
88 | #define MPB_ATTRIB_IGNORED (MPB_ATTRIB_NEVER_USE) | |
fe7ed8cb | 89 | |
8e59f3d8 | 90 | #define MPB_SECTOR_CNT 2210 |
c2c087e6 | 91 | #define IMSM_RESERVED_SECTORS 4096 |
b81221b7 | 92 | #define NUM_BLOCKS_DIRTY_STRIPE_REGION 2056 |
979d38be | 93 | #define SECT_PER_MB_SHIFT 11 |
f36a9ecd | 94 | #define MAX_SECTOR_SIZE 4096 |
c2462068 PB |
95 | #define MULTIPLE_PPL_AREA_SIZE_IMSM (1024 * 1024) /* Size of the whole |
96 | * mutliple PPL area | |
97 | */ | |
cdddbdbc DW |
98 | |
99 | /* Disk configuration info. */ | |
100 | #define IMSM_MAX_DEVICES 255 | |
101 | struct imsm_disk { | |
102 | __u8 serial[MAX_RAID_SERIAL_LEN];/* 0xD8 - 0xE7 ascii serial number */ | |
5551b113 | 103 | __u32 total_blocks_lo; /* 0xE8 - 0xEB total blocks lo */ |
cdddbdbc | 104 | __u32 scsi_id; /* 0xEC - 0xEF scsi ID */ |
f2f27e63 DW |
105 | #define SPARE_DISK __cpu_to_le32(0x01) /* Spare */ |
106 | #define CONFIGURED_DISK __cpu_to_le32(0x02) /* Member of some RaidDev */ | |
107 | #define FAILED_DISK __cpu_to_le32(0x04) /* Permanent failure */ | |
2432ce9b | 108 | #define JOURNAL_DISK __cpu_to_le32(0x2000000) /* Device marked as Journaling Drive */ |
cdddbdbc | 109 | __u32 status; /* 0xF0 - 0xF3 */ |
1011e834 | 110 | __u32 owner_cfg_num; /* which config 0,1,2... owns this disk */ |
5551b113 CA |
111 | __u32 total_blocks_hi; /* 0xF4 - 0xF5 total blocks hi */ |
112 | #define IMSM_DISK_FILLERS 3 | |
113 | __u32 filler[IMSM_DISK_FILLERS]; /* 0xF5 - 0x107 MPB_DISK_FILLERS for future expansion */ | |
cdddbdbc DW |
114 | }; |
115 | ||
3b451610 AK |
116 | /* map selector for map managment |
117 | */ | |
238c0a71 AK |
118 | #define MAP_0 0 |
119 | #define MAP_1 1 | |
120 | #define MAP_X -1 | |
3b451610 | 121 | |
cdddbdbc DW |
122 | /* RAID map configuration infos. */ |
123 | struct imsm_map { | |
5551b113 CA |
124 | __u32 pba_of_lba0_lo; /* start address of partition */ |
125 | __u32 blocks_per_member_lo;/* blocks per member */ | |
126 | __u32 num_data_stripes_lo; /* number of data stripes */ | |
cdddbdbc DW |
127 | __u16 blocks_per_strip; |
128 | __u8 map_state; /* Normal, Uninitialized, Degraded, Failed */ | |
129 | #define IMSM_T_STATE_NORMAL 0 | |
130 | #define IMSM_T_STATE_UNINITIALIZED 1 | |
e3bba0e0 DW |
131 | #define IMSM_T_STATE_DEGRADED 2 |
132 | #define IMSM_T_STATE_FAILED 3 | |
cdddbdbc DW |
133 | __u8 raid_level; |
134 | #define IMSM_T_RAID0 0 | |
135 | #define IMSM_T_RAID1 1 | |
136 | #define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */ | |
137 | __u8 num_members; /* number of member disks */ | |
fe7ed8cb DW |
138 | __u8 num_domains; /* number of parity domains */ |
139 | __u8 failed_disk_num; /* valid only when state is degraded */ | |
252d23c0 | 140 | __u8 ddf; |
5551b113 CA |
141 | __u32 pba_of_lba0_hi; |
142 | __u32 blocks_per_member_hi; | |
143 | __u32 num_data_stripes_hi; | |
144 | __u32 filler[4]; /* expansion area */ | |
7eef0453 | 145 | #define IMSM_ORD_REBUILD (1 << 24) |
cdddbdbc | 146 | __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members], |
7eef0453 DW |
147 | * top byte contains some flags |
148 | */ | |
cdddbdbc DW |
149 | } __attribute__ ((packed)); |
150 | ||
151 | struct imsm_vol { | |
f8f603f1 | 152 | __u32 curr_migr_unit; |
fe7ed8cb | 153 | __u32 checkpoint_id; /* id to access curr_migr_unit */ |
cdddbdbc | 154 | __u8 migr_state; /* Normal or Migrating */ |
e3bba0e0 DW |
155 | #define MIGR_INIT 0 |
156 | #define MIGR_REBUILD 1 | |
157 | #define MIGR_VERIFY 2 /* analagous to echo check > sync_action */ | |
158 | #define MIGR_GEN_MIGR 3 | |
159 | #define MIGR_STATE_CHANGE 4 | |
1484e727 | 160 | #define MIGR_REPAIR 5 |
cdddbdbc | 161 | __u8 migr_type; /* Initializing, Rebuilding, ... */ |
2432ce9b AP |
162 | #define RAIDVOL_CLEAN 0 |
163 | #define RAIDVOL_DIRTY 1 | |
164 | #define RAIDVOL_DSRECORD_VALID 2 | |
cdddbdbc | 165 | __u8 dirty; |
fe7ed8cb DW |
166 | __u8 fs_state; /* fast-sync state for CnG (0xff == disabled) */ |
167 | __u16 verify_errors; /* number of mismatches */ | |
168 | __u16 bad_blocks; /* number of bad blocks during verify */ | |
169 | __u32 filler[4]; | |
cdddbdbc DW |
170 | struct imsm_map map[1]; |
171 | /* here comes another one if migr_state */ | |
172 | } __attribute__ ((packed)); | |
173 | ||
174 | struct imsm_dev { | |
fe7ed8cb | 175 | __u8 volume[MAX_RAID_SERIAL_LEN]; |
cdddbdbc DW |
176 | __u32 size_low; |
177 | __u32 size_high; | |
fe7ed8cb DW |
178 | #define DEV_BOOTABLE __cpu_to_le32(0x01) |
179 | #define DEV_BOOT_DEVICE __cpu_to_le32(0x02) | |
180 | #define DEV_READ_COALESCING __cpu_to_le32(0x04) | |
181 | #define DEV_WRITE_COALESCING __cpu_to_le32(0x08) | |
182 | #define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10) | |
183 | #define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20) | |
184 | #define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40) | |
185 | #define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80) | |
186 | #define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100) | |
187 | #define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200) | |
188 | #define DEV_CLONE_N_GO __cpu_to_le32(0x400) | |
189 | #define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800) | |
190 | #define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000) | |
cdddbdbc DW |
191 | __u32 status; /* Persistent RaidDev status */ |
192 | __u32 reserved_blocks; /* Reserved blocks at beginning of volume */ | |
fe7ed8cb DW |
193 | __u8 migr_priority; |
194 | __u8 num_sub_vols; | |
195 | __u8 tid; | |
196 | __u8 cng_master_disk; | |
197 | __u16 cache_policy; | |
198 | __u8 cng_state; | |
199 | __u8 cng_sub_state; | |
2432ce9b AP |
200 | __u16 my_vol_raid_dev_num; /* Used in Unique volume Id for this RaidDev */ |
201 | ||
202 | /* NVM_EN */ | |
203 | __u8 nv_cache_mode; | |
204 | __u8 nv_cache_flags; | |
205 | ||
206 | /* Unique Volume Id of the NvCache Volume associated with this volume */ | |
207 | __u32 nvc_vol_orig_family_num; | |
208 | __u16 nvc_vol_raid_dev_num; | |
209 | ||
210 | #define RWH_OFF 0 | |
211 | #define RWH_DISTRIBUTED 1 | |
212 | #define RWH_JOURNALING_DRIVE 2 | |
c2462068 PB |
213 | #define RWH_MULTIPLE_DISTRIBUTED 3 |
214 | #define RWH_MULTIPLE_PPLS_JOURNALING_DRIVE 4 | |
215 | #define RWH_MULTIPLE_OFF 5 | |
2432ce9b AP |
216 | __u8 rwh_policy; /* Raid Write Hole Policy */ |
217 | __u8 jd_serial[MAX_RAID_SERIAL_LEN]; /* Journal Drive serial number */ | |
218 | __u8 filler1; | |
219 | ||
220 | #define IMSM_DEV_FILLERS 3 | |
cdddbdbc DW |
221 | __u32 filler[IMSM_DEV_FILLERS]; |
222 | struct imsm_vol vol; | |
223 | } __attribute__ ((packed)); | |
224 | ||
225 | struct imsm_super { | |
226 | __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */ | |
227 | __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */ | |
228 | __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */ | |
229 | __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */ | |
230 | __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */ | |
604b746f JD |
231 | __u32 error_log_size; /* 0x30 - 0x33 in bytes */ |
232 | __u32 attributes; /* 0x34 - 0x37 */ | |
cdddbdbc DW |
233 | __u8 num_disks; /* 0x38 Number of configured disks */ |
234 | __u8 num_raid_devs; /* 0x39 Number of configured volumes */ | |
604b746f JD |
235 | __u8 error_log_pos; /* 0x3A */ |
236 | __u8 fill[1]; /* 0x3B */ | |
237 | __u32 cache_size; /* 0x3c - 0x40 in mb */ | |
238 | __u32 orig_family_num; /* 0x40 - 0x43 original family num */ | |
239 | __u32 pwr_cycle_count; /* 0x44 - 0x47 simulated power cycle count for array */ | |
240 | __u32 bbm_log_size; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */ | |
2a24dc1b PB |
241 | __u16 num_raid_devs_created; /* 0x4C - 0x4D Used for generating unique |
242 | * volume IDs for raid_dev created in this array | |
243 | * (starts at 1) | |
244 | */ | |
245 | __u16 filler1; /* 0x4E - 0x4F */ | |
246 | #define IMSM_FILLERS 34 | |
247 | __u32 filler[IMSM_FILLERS]; /* 0x50 - 0xD7 RAID_MPB_FILLERS */ | |
cdddbdbc DW |
248 | struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */ |
249 | /* here comes imsm_dev[num_raid_devs] */ | |
604b746f | 250 | /* here comes BBM logs */ |
cdddbdbc DW |
251 | } __attribute__ ((packed)); |
252 | ||
604b746f | 253 | #define BBM_LOG_MAX_ENTRIES 254 |
8d67477f TM |
254 | #define BBM_LOG_MAX_LBA_ENTRY_VAL 256 /* Represents 256 LBAs */ |
255 | #define BBM_LOG_SIGNATURE 0xabadb10c | |
256 | ||
257 | struct bbm_log_block_addr { | |
258 | __u16 w1; | |
259 | __u32 dw1; | |
260 | } __attribute__ ((__packed__)); | |
604b746f JD |
261 | |
262 | struct bbm_log_entry { | |
8d67477f TM |
263 | __u8 marked_count; /* Number of blocks marked - 1 */ |
264 | __u8 disk_ordinal; /* Disk entry within the imsm_super */ | |
265 | struct bbm_log_block_addr defective_block_start; | |
604b746f JD |
266 | } __attribute__ ((__packed__)); |
267 | ||
268 | struct bbm_log { | |
269 | __u32 signature; /* 0xABADB10C */ | |
270 | __u32 entry_count; | |
8d67477f | 271 | struct bbm_log_entry marked_block_entries[BBM_LOG_MAX_ENTRIES]; |
604b746f JD |
272 | } __attribute__ ((__packed__)); |
273 | ||
cdddbdbc | 274 | static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" }; |
cdddbdbc | 275 | |
b53bfba6 TM |
276 | #define BLOCKS_PER_KB (1024/512) |
277 | ||
8e59f3d8 AK |
278 | #define RAID_DISK_RESERVED_BLOCKS_IMSM_HI 2209 |
279 | ||
280 | #define GEN_MIGR_AREA_SIZE 2048 /* General Migration Copy Area size in blocks */ | |
281 | ||
de44e46f PB |
282 | #define MIGR_REC_BUF_SECTORS 1 /* size of migr_record i/o buffer in sectors */ |
283 | #define MIGR_REC_SECTOR_POSITION 1 /* migr_record position offset on disk, | |
284 | * MIGR_REC_BUF_SECTORS <= MIGR_REC_SECTOR_POS | |
17a4eaf9 AK |
285 | */ |
286 | ||
8e59f3d8 AK |
287 | #define UNIT_SRC_NORMAL 0 /* Source data for curr_migr_unit must |
288 | * be recovered using srcMap */ | |
289 | #define UNIT_SRC_IN_CP_AREA 1 /* Source data for curr_migr_unit has | |
290 | * already been migrated and must | |
291 | * be recovered from checkpoint area */ | |
2432ce9b | 292 | |
c2462068 | 293 | #define PPL_ENTRY_SPACE (128 * 1024) /* Size of single PPL, without the header */ |
2432ce9b | 294 | |
8e59f3d8 AK |
295 | struct migr_record { |
296 | __u32 rec_status; /* Status used to determine how to restart | |
297 | * migration in case it aborts | |
298 | * in some fashion */ | |
299 | __u32 curr_migr_unit; /* 0..numMigrUnits-1 */ | |
300 | __u32 family_num; /* Family number of MPB | |
301 | * containing the RaidDev | |
302 | * that is migrating */ | |
303 | __u32 ascending_migr; /* True if migrating in increasing | |
304 | * order of lbas */ | |
305 | __u32 blocks_per_unit; /* Num disk blocks per unit of operation */ | |
306 | __u32 dest_depth_per_unit; /* Num member blocks each destMap | |
307 | * member disk | |
308 | * advances per unit-of-operation */ | |
309 | __u32 ckpt_area_pba; /* Pba of first block of ckpt copy area */ | |
310 | __u32 dest_1st_member_lba; /* First member lba on first | |
311 | * stripe of destination */ | |
312 | __u32 num_migr_units; /* Total num migration units-of-op */ | |
313 | __u32 post_migr_vol_cap; /* Size of volume after | |
314 | * migration completes */ | |
315 | __u32 post_migr_vol_cap_hi; /* Expansion space for LBA64 */ | |
316 | __u32 ckpt_read_disk_num; /* Which member disk in destSubMap[0] the | |
317 | * migration ckpt record was read from | |
318 | * (for recovered migrations) */ | |
319 | } __attribute__ ((__packed__)); | |
320 | ||
ec50f7b6 LM |
321 | struct md_list { |
322 | /* usage marker: | |
323 | * 1: load metadata | |
324 | * 2: metadata does not match | |
325 | * 4: already checked | |
326 | */ | |
327 | int used; | |
328 | char *devname; | |
329 | int found; | |
330 | int container; | |
331 | dev_t st_rdev; | |
332 | struct md_list *next; | |
333 | }; | |
334 | ||
e7b84f9d | 335 | #define pr_vrb(fmt, arg...) (void) (verbose && pr_err(fmt, ##arg)) |
ec50f7b6 | 336 | |
1484e727 DW |
337 | static __u8 migr_type(struct imsm_dev *dev) |
338 | { | |
339 | if (dev->vol.migr_type == MIGR_VERIFY && | |
340 | dev->status & DEV_VERIFY_AND_FIX) | |
341 | return MIGR_REPAIR; | |
342 | else | |
343 | return dev->vol.migr_type; | |
344 | } | |
345 | ||
346 | static void set_migr_type(struct imsm_dev *dev, __u8 migr_type) | |
347 | { | |
348 | /* for compatibility with older oroms convert MIGR_REPAIR, into | |
349 | * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status | |
350 | */ | |
351 | if (migr_type == MIGR_REPAIR) { | |
352 | dev->vol.migr_type = MIGR_VERIFY; | |
353 | dev->status |= DEV_VERIFY_AND_FIX; | |
354 | } else { | |
355 | dev->vol.migr_type = migr_type; | |
356 | dev->status &= ~DEV_VERIFY_AND_FIX; | |
357 | } | |
358 | } | |
359 | ||
f36a9ecd | 360 | static unsigned int sector_count(__u32 bytes, unsigned int sector_size) |
cdddbdbc | 361 | { |
f36a9ecd | 362 | return ROUND_UP(bytes, sector_size) / sector_size; |
87eb16df | 363 | } |
cdddbdbc | 364 | |
f36a9ecd PB |
365 | static unsigned int mpb_sectors(struct imsm_super *mpb, |
366 | unsigned int sector_size) | |
87eb16df | 367 | { |
f36a9ecd | 368 | return sector_count(__le32_to_cpu(mpb->mpb_size), sector_size); |
cdddbdbc DW |
369 | } |
370 | ||
ba2de7ba DW |
371 | struct intel_dev { |
372 | struct imsm_dev *dev; | |
373 | struct intel_dev *next; | |
f21e18ca | 374 | unsigned index; |
ba2de7ba DW |
375 | }; |
376 | ||
88654014 LM |
377 | struct intel_hba { |
378 | enum sys_dev_type type; | |
379 | char *path; | |
380 | char *pci_id; | |
381 | struct intel_hba *next; | |
382 | }; | |
383 | ||
1a64be56 LM |
384 | enum action { |
385 | DISK_REMOVE = 1, | |
386 | DISK_ADD | |
387 | }; | |
cdddbdbc DW |
388 | /* internal representation of IMSM metadata */ |
389 | struct intel_super { | |
390 | union { | |
949c47a0 DW |
391 | void *buf; /* O_DIRECT buffer for reading/writing metadata */ |
392 | struct imsm_super *anchor; /* immovable parameters */ | |
cdddbdbc | 393 | }; |
8e59f3d8 AK |
394 | union { |
395 | void *migr_rec_buf; /* buffer for I/O operations */ | |
396 | struct migr_record *migr_rec; /* migration record */ | |
397 | }; | |
51d83f5d AK |
398 | int clean_migration_record_by_mdmon; /* when reshape is switched to next |
399 | array, it indicates that mdmon is allowed to clean migration | |
400 | record */ | |
949c47a0 | 401 | size_t len; /* size of the 'buf' allocation */ |
bbab0940 | 402 | size_t extra_space; /* extra space in 'buf' that is not used yet */ |
4d7b1503 DW |
403 | void *next_buf; /* for realloc'ing buf from the manager */ |
404 | size_t next_len; | |
c2c087e6 | 405 | int updates_pending; /* count of pending updates for mdmon */ |
bf5a934a | 406 | int current_vol; /* index of raid device undergoing creation */ |
5551b113 | 407 | unsigned long long create_offset; /* common start for 'current_vol' */ |
148acb7b | 408 | __u32 random; /* random data for seeding new family numbers */ |
ba2de7ba | 409 | struct intel_dev *devlist; |
fa7bb6f8 | 410 | unsigned int sector_size; /* sector size of used member drives */ |
cdddbdbc DW |
411 | struct dl { |
412 | struct dl *next; | |
413 | int index; | |
414 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
415 | int major, minor; | |
416 | char *devname; | |
b9f594fe | 417 | struct imsm_disk disk; |
cdddbdbc | 418 | int fd; |
0dcecb2e DW |
419 | int extent_cnt; |
420 | struct extent *e; /* for determining freespace @ create */ | |
efb30e7f | 421 | int raiddisk; /* slot to fill in autolayout */ |
1a64be56 | 422 | enum action action; |
ca0748fa | 423 | } *disks, *current_disk; |
1a64be56 LM |
424 | struct dl *disk_mgmt_list; /* list of disks to add/remove while mdmon |
425 | active */ | |
47ee5a45 | 426 | struct dl *missing; /* disks removed while we weren't looking */ |
43dad3d6 | 427 | struct bbm_log *bbm_log; |
88654014 | 428 | struct intel_hba *hba; /* device path of the raid controller for this metadata */ |
88c32bb1 | 429 | const struct imsm_orom *orom; /* platform firmware support */ |
a2b97981 | 430 | struct intel_super *next; /* (temp) list for disambiguating family_num */ |
928f1424 | 431 | struct md_bb bb; /* memory for get_bad_blocks call */ |
a2b97981 DW |
432 | }; |
433 | ||
434 | struct intel_disk { | |
435 | struct imsm_disk disk; | |
436 | #define IMSM_UNKNOWN_OWNER (-1) | |
437 | int owner; | |
438 | struct intel_disk *next; | |
cdddbdbc DW |
439 | }; |
440 | ||
c2c087e6 DW |
441 | struct extent { |
442 | unsigned long long start, size; | |
443 | }; | |
444 | ||
694575e7 KW |
445 | /* definitions of reshape process types */ |
446 | enum imsm_reshape_type { | |
447 | CH_TAKEOVER, | |
b5347799 | 448 | CH_MIGRATION, |
7abc9871 | 449 | CH_ARRAY_SIZE, |
694575e7 KW |
450 | }; |
451 | ||
88758e9d DW |
452 | /* definition of messages passed to imsm_process_update */ |
453 | enum imsm_update_type { | |
454 | update_activate_spare, | |
8273f55e | 455 | update_create_array, |
33414a01 | 456 | update_kill_array, |
aa534678 | 457 | update_rename_array, |
1a64be56 | 458 | update_add_remove_disk, |
78b10e66 | 459 | update_reshape_container_disks, |
48c5303a | 460 | update_reshape_migration, |
2d40f3a1 AK |
461 | update_takeover, |
462 | update_general_migration_checkpoint, | |
f3871fdc | 463 | update_size_change, |
bbab0940 | 464 | update_prealloc_badblocks_mem, |
e6e9dd3f | 465 | update_rwh_policy, |
88758e9d DW |
466 | }; |
467 | ||
468 | struct imsm_update_activate_spare { | |
469 | enum imsm_update_type type; | |
d23fe947 | 470 | struct dl *dl; |
88758e9d DW |
471 | int slot; |
472 | int array; | |
473 | struct imsm_update_activate_spare *next; | |
474 | }; | |
475 | ||
78b10e66 | 476 | struct geo_params { |
4dd2df09 | 477 | char devnm[32]; |
78b10e66 | 478 | char *dev_name; |
d04f65f4 | 479 | unsigned long long size; |
78b10e66 N |
480 | int level; |
481 | int layout; | |
482 | int chunksize; | |
483 | int raid_disks; | |
484 | }; | |
485 | ||
bb025c2f KW |
486 | enum takeover_direction { |
487 | R10_TO_R0, | |
488 | R0_TO_R10 | |
489 | }; | |
490 | struct imsm_update_takeover { | |
491 | enum imsm_update_type type; | |
492 | int subarray; | |
493 | enum takeover_direction direction; | |
494 | }; | |
78b10e66 N |
495 | |
496 | struct imsm_update_reshape { | |
497 | enum imsm_update_type type; | |
498 | int old_raid_disks; | |
499 | int new_raid_disks; | |
48c5303a PC |
500 | |
501 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ | |
502 | }; | |
503 | ||
504 | struct imsm_update_reshape_migration { | |
505 | enum imsm_update_type type; | |
506 | int old_raid_disks; | |
507 | int new_raid_disks; | |
508 | /* fields for array migration changes | |
509 | */ | |
510 | int subdev; | |
511 | int new_level; | |
512 | int new_layout; | |
4bba0439 | 513 | int new_chunksize; |
48c5303a | 514 | |
d195167d | 515 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ |
78b10e66 N |
516 | }; |
517 | ||
f3871fdc AK |
518 | struct imsm_update_size_change { |
519 | enum imsm_update_type type; | |
520 | int subdev; | |
521 | long long new_size; | |
522 | }; | |
523 | ||
2d40f3a1 AK |
524 | struct imsm_update_general_migration_checkpoint { |
525 | enum imsm_update_type type; | |
526 | __u32 curr_migr_unit; | |
527 | }; | |
528 | ||
54c2c1ea DW |
529 | struct disk_info { |
530 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
531 | }; | |
532 | ||
8273f55e DW |
533 | struct imsm_update_create_array { |
534 | enum imsm_update_type type; | |
8273f55e | 535 | int dev_idx; |
6a3e913e | 536 | struct imsm_dev dev; |
8273f55e DW |
537 | }; |
538 | ||
33414a01 DW |
539 | struct imsm_update_kill_array { |
540 | enum imsm_update_type type; | |
541 | int dev_idx; | |
542 | }; | |
543 | ||
aa534678 DW |
544 | struct imsm_update_rename_array { |
545 | enum imsm_update_type type; | |
546 | __u8 name[MAX_RAID_SERIAL_LEN]; | |
547 | int dev_idx; | |
548 | }; | |
549 | ||
1a64be56 | 550 | struct imsm_update_add_remove_disk { |
43dad3d6 DW |
551 | enum imsm_update_type type; |
552 | }; | |
553 | ||
bbab0940 TM |
554 | struct imsm_update_prealloc_bb_mem { |
555 | enum imsm_update_type type; | |
556 | }; | |
557 | ||
e6e9dd3f AP |
558 | struct imsm_update_rwh_policy { |
559 | enum imsm_update_type type; | |
560 | int new_policy; | |
561 | int dev_idx; | |
562 | }; | |
563 | ||
88654014 LM |
564 | static const char *_sys_dev_type[] = { |
565 | [SYS_DEV_UNKNOWN] = "Unknown", | |
566 | [SYS_DEV_SAS] = "SAS", | |
614902f6 | 567 | [SYS_DEV_SATA] = "SATA", |
60f0f54d PB |
568 | [SYS_DEV_NVME] = "NVMe", |
569 | [SYS_DEV_VMD] = "VMD" | |
88654014 LM |
570 | }; |
571 | ||
572 | const char *get_sys_dev_type(enum sys_dev_type type) | |
573 | { | |
574 | if (type >= SYS_DEV_MAX) | |
575 | type = SYS_DEV_UNKNOWN; | |
576 | ||
577 | return _sys_dev_type[type]; | |
578 | } | |
579 | ||
580 | static struct intel_hba * alloc_intel_hba(struct sys_dev *device) | |
581 | { | |
503975b9 N |
582 | struct intel_hba *result = xmalloc(sizeof(*result)); |
583 | ||
584 | result->type = device->type; | |
585 | result->path = xstrdup(device->path); | |
586 | result->next = NULL; | |
587 | if (result->path && (result->pci_id = strrchr(result->path, '/')) != NULL) | |
588 | result->pci_id++; | |
589 | ||
88654014 LM |
590 | return result; |
591 | } | |
592 | ||
593 | static struct intel_hba * find_intel_hba(struct intel_hba *hba, struct sys_dev *device) | |
594 | { | |
594dc1b8 JS |
595 | struct intel_hba *result; |
596 | ||
88654014 LM |
597 | for (result = hba; result; result = result->next) { |
598 | if (result->type == device->type && strcmp(result->path, device->path) == 0) | |
599 | break; | |
600 | } | |
601 | return result; | |
602 | } | |
603 | ||
b4cf4cba | 604 | static int attach_hba_to_super(struct intel_super *super, struct sys_dev *device) |
88654014 LM |
605 | { |
606 | struct intel_hba *hba; | |
607 | ||
608 | /* check if disk attached to Intel HBA */ | |
609 | hba = find_intel_hba(super->hba, device); | |
610 | if (hba != NULL) | |
611 | return 1; | |
612 | /* Check if HBA is already attached to super */ | |
613 | if (super->hba == NULL) { | |
614 | super->hba = alloc_intel_hba(device); | |
615 | return 1; | |
6b781d33 AP |
616 | } |
617 | ||
618 | hba = super->hba; | |
619 | /* Intel metadata allows for all disks attached to the same type HBA. | |
614902f6 | 620 | * Do not support HBA types mixing |
6b781d33 AP |
621 | */ |
622 | if (device->type != hba->type) | |
88654014 | 623 | return 2; |
6b781d33 AP |
624 | |
625 | /* Multiple same type HBAs can be used if they share the same OROM */ | |
626 | const struct imsm_orom *device_orom = get_orom_by_device_id(device->dev_id); | |
627 | ||
628 | if (device_orom != super->orom) | |
629 | return 2; | |
630 | ||
631 | while (hba->next) | |
632 | hba = hba->next; | |
633 | ||
634 | hba->next = alloc_intel_hba(device); | |
635 | return 1; | |
88654014 LM |
636 | } |
637 | ||
638 | static struct sys_dev* find_disk_attached_hba(int fd, const char *devname) | |
639 | { | |
9bc4ae77 | 640 | struct sys_dev *list, *elem; |
88654014 LM |
641 | char *disk_path; |
642 | ||
643 | if ((list = find_intel_devices()) == NULL) | |
644 | return 0; | |
645 | ||
646 | if (fd < 0) | |
647 | disk_path = (char *) devname; | |
648 | else | |
649 | disk_path = diskfd_to_devpath(fd); | |
650 | ||
9bc4ae77 | 651 | if (!disk_path) |
88654014 | 652 | return 0; |
88654014 | 653 | |
9bc4ae77 N |
654 | for (elem = list; elem; elem = elem->next) |
655 | if (path_attached_to_hba(disk_path, elem->path)) | |
88654014 | 656 | return elem; |
9bc4ae77 | 657 | |
88654014 LM |
658 | if (disk_path != devname) |
659 | free(disk_path); | |
88654014 LM |
660 | |
661 | return NULL; | |
662 | } | |
663 | ||
d424212e N |
664 | static int find_intel_hba_capability(int fd, struct intel_super *super, |
665 | char *devname); | |
f2f5c343 | 666 | |
cdddbdbc DW |
667 | static struct supertype *match_metadata_desc_imsm(char *arg) |
668 | { | |
669 | struct supertype *st; | |
670 | ||
671 | if (strcmp(arg, "imsm") != 0 && | |
672 | strcmp(arg, "default") != 0 | |
673 | ) | |
674 | return NULL; | |
675 | ||
503975b9 | 676 | st = xcalloc(1, sizeof(*st)); |
cdddbdbc DW |
677 | st->ss = &super_imsm; |
678 | st->max_devs = IMSM_MAX_DEVICES; | |
679 | st->minor_version = 0; | |
680 | st->sb = NULL; | |
681 | return st; | |
682 | } | |
683 | ||
cdddbdbc DW |
684 | static __u8 *get_imsm_version(struct imsm_super *mpb) |
685 | { | |
686 | return &mpb->sig[MPB_SIG_LEN]; | |
687 | } | |
688 | ||
949c47a0 DW |
689 | /* retrieve a disk directly from the anchor when the anchor is known to be |
690 | * up-to-date, currently only at load time | |
691 | */ | |
692 | static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index) | |
cdddbdbc | 693 | { |
949c47a0 | 694 | if (index >= mpb->num_disks) |
cdddbdbc DW |
695 | return NULL; |
696 | return &mpb->disk[index]; | |
697 | } | |
698 | ||
95d07a2c LM |
699 | /* retrieve the disk description based on a index of the disk |
700 | * in the sub-array | |
701 | */ | |
702 | static struct dl *get_imsm_dl_disk(struct intel_super *super, __u8 index) | |
949c47a0 | 703 | { |
b9f594fe DW |
704 | struct dl *d; |
705 | ||
706 | for (d = super->disks; d; d = d->next) | |
707 | if (d->index == index) | |
95d07a2c LM |
708 | return d; |
709 | ||
710 | return NULL; | |
711 | } | |
712 | /* retrieve a disk from the parsed metadata */ | |
713 | static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index) | |
714 | { | |
715 | struct dl *dl; | |
716 | ||
717 | dl = get_imsm_dl_disk(super, index); | |
718 | if (dl) | |
719 | return &dl->disk; | |
720 | ||
b9f594fe | 721 | return NULL; |
949c47a0 DW |
722 | } |
723 | ||
724 | /* generate a checksum directly from the anchor when the anchor is known to be | |
725 | * up-to-date, currently only at load or write_super after coalescing | |
726 | */ | |
727 | static __u32 __gen_imsm_checksum(struct imsm_super *mpb) | |
cdddbdbc DW |
728 | { |
729 | __u32 end = mpb->mpb_size / sizeof(end); | |
730 | __u32 *p = (__u32 *) mpb; | |
731 | __u32 sum = 0; | |
732 | ||
5d500228 N |
733 | while (end--) { |
734 | sum += __le32_to_cpu(*p); | |
97f734fd N |
735 | p++; |
736 | } | |
cdddbdbc | 737 | |
5d500228 | 738 | return sum - __le32_to_cpu(mpb->check_sum); |
cdddbdbc DW |
739 | } |
740 | ||
a965f303 DW |
741 | static size_t sizeof_imsm_map(struct imsm_map *map) |
742 | { | |
743 | return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1); | |
744 | } | |
745 | ||
746 | struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map) | |
cdddbdbc | 747 | { |
5e7b0330 AK |
748 | /* A device can have 2 maps if it is in the middle of a migration. |
749 | * If second_map is: | |
238c0a71 AK |
750 | * MAP_0 - we return the first map |
751 | * MAP_1 - we return the second map if it exists, else NULL | |
752 | * MAP_X - we return the second map if it exists, else the first | |
5e7b0330 | 753 | */ |
a965f303 | 754 | struct imsm_map *map = &dev->vol.map[0]; |
9535fc47 | 755 | struct imsm_map *map2 = NULL; |
a965f303 | 756 | |
9535fc47 AK |
757 | if (dev->vol.migr_state) |
758 | map2 = (void *)map + sizeof_imsm_map(map); | |
a965f303 | 759 | |
9535fc47 | 760 | switch (second_map) { |
3b451610 | 761 | case MAP_0: |
9535fc47 | 762 | break; |
3b451610 | 763 | case MAP_1: |
9535fc47 AK |
764 | map = map2; |
765 | break; | |
238c0a71 | 766 | case MAP_X: |
9535fc47 AK |
767 | if (map2) |
768 | map = map2; | |
769 | break; | |
9535fc47 AK |
770 | default: |
771 | map = NULL; | |
772 | } | |
773 | return map; | |
5e7b0330 | 774 | |
a965f303 | 775 | } |
cdddbdbc | 776 | |
3393c6af DW |
777 | /* return the size of the device. |
778 | * migr_state increases the returned size if map[0] were to be duplicated | |
779 | */ | |
780 | static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state) | |
a965f303 DW |
781 | { |
782 | size_t size = sizeof(*dev) - sizeof(struct imsm_map) + | |
238c0a71 | 783 | sizeof_imsm_map(get_imsm_map(dev, MAP_0)); |
cdddbdbc DW |
784 | |
785 | /* migrating means an additional map */ | |
a965f303 | 786 | if (dev->vol.migr_state) |
238c0a71 | 787 | size += sizeof_imsm_map(get_imsm_map(dev, MAP_1)); |
3393c6af | 788 | else if (migr_state) |
238c0a71 | 789 | size += sizeof_imsm_map(get_imsm_map(dev, MAP_0)); |
cdddbdbc DW |
790 | |
791 | return size; | |
792 | } | |
793 | ||
54c2c1ea DW |
794 | /* retrieve disk serial number list from a metadata update */ |
795 | static struct disk_info *get_disk_info(struct imsm_update_create_array *update) | |
796 | { | |
797 | void *u = update; | |
798 | struct disk_info *inf; | |
799 | ||
800 | inf = u + sizeof(*update) - sizeof(struct imsm_dev) + | |
801 | sizeof_imsm_dev(&update->dev, 0); | |
802 | ||
803 | return inf; | |
804 | } | |
54c2c1ea | 805 | |
949c47a0 | 806 | static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index) |
cdddbdbc DW |
807 | { |
808 | int offset; | |
809 | int i; | |
810 | void *_mpb = mpb; | |
811 | ||
949c47a0 | 812 | if (index >= mpb->num_raid_devs) |
cdddbdbc DW |
813 | return NULL; |
814 | ||
815 | /* devices start after all disks */ | |
816 | offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb; | |
817 | ||
818 | for (i = 0; i <= index; i++) | |
819 | if (i == index) | |
820 | return _mpb + offset; | |
821 | else | |
3393c6af | 822 | offset += sizeof_imsm_dev(_mpb + offset, 0); |
cdddbdbc DW |
823 | |
824 | return NULL; | |
825 | } | |
826 | ||
949c47a0 DW |
827 | static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index) |
828 | { | |
ba2de7ba DW |
829 | struct intel_dev *dv; |
830 | ||
949c47a0 DW |
831 | if (index >= super->anchor->num_raid_devs) |
832 | return NULL; | |
ba2de7ba DW |
833 | for (dv = super->devlist; dv; dv = dv->next) |
834 | if (dv->index == index) | |
835 | return dv->dev; | |
836 | return NULL; | |
949c47a0 DW |
837 | } |
838 | ||
8d67477f TM |
839 | static inline unsigned long long __le48_to_cpu(const struct bbm_log_block_addr |
840 | *addr) | |
841 | { | |
842 | return ((((__u64)__le32_to_cpu(addr->dw1)) << 16) | | |
843 | __le16_to_cpu(addr->w1)); | |
844 | } | |
845 | ||
846 | static inline struct bbm_log_block_addr __cpu_to_le48(unsigned long long sec) | |
847 | { | |
848 | struct bbm_log_block_addr addr; | |
849 | ||
850 | addr.w1 = __cpu_to_le16((__u16)(sec & 0xffff)); | |
851 | addr.dw1 = __cpu_to_le32((__u32)(sec >> 16) & 0xffffffff); | |
852 | return addr; | |
853 | } | |
854 | ||
8d67477f TM |
855 | /* get size of the bbm log */ |
856 | static __u32 get_imsm_bbm_log_size(struct bbm_log *log) | |
857 | { | |
858 | if (!log || log->entry_count == 0) | |
859 | return 0; | |
860 | ||
861 | return sizeof(log->signature) + | |
862 | sizeof(log->entry_count) + | |
863 | log->entry_count * sizeof(struct bbm_log_entry); | |
864 | } | |
6f50473f TM |
865 | |
866 | /* check if bad block is not partially stored in bbm log */ | |
867 | static int is_stored_in_bbm(struct bbm_log *log, const __u8 idx, const unsigned | |
868 | long long sector, const int length, __u32 *pos) | |
869 | { | |
870 | __u32 i; | |
871 | ||
872 | for (i = *pos; i < log->entry_count; i++) { | |
873 | struct bbm_log_entry *entry = &log->marked_block_entries[i]; | |
874 | unsigned long long bb_start; | |
875 | unsigned long long bb_end; | |
876 | ||
877 | bb_start = __le48_to_cpu(&entry->defective_block_start); | |
878 | bb_end = bb_start + (entry->marked_count + 1); | |
879 | ||
880 | if ((entry->disk_ordinal == idx) && (bb_start >= sector) && | |
881 | (bb_end <= sector + length)) { | |
882 | *pos = i; | |
883 | return 1; | |
884 | } | |
885 | } | |
886 | return 0; | |
887 | } | |
888 | ||
889 | /* record new bad block in bbm log */ | |
890 | static int record_new_badblock(struct bbm_log *log, const __u8 idx, unsigned | |
891 | long long sector, int length) | |
892 | { | |
893 | int new_bb = 0; | |
894 | __u32 pos = 0; | |
895 | struct bbm_log_entry *entry = NULL; | |
896 | ||
897 | while (is_stored_in_bbm(log, idx, sector, length, &pos)) { | |
898 | struct bbm_log_entry *e = &log->marked_block_entries[pos]; | |
899 | ||
900 | if ((e->marked_count + 1 == BBM_LOG_MAX_LBA_ENTRY_VAL) && | |
901 | (__le48_to_cpu(&e->defective_block_start) == sector)) { | |
902 | sector += BBM_LOG_MAX_LBA_ENTRY_VAL; | |
903 | length -= BBM_LOG_MAX_LBA_ENTRY_VAL; | |
904 | pos = pos + 1; | |
905 | continue; | |
906 | } | |
907 | entry = e; | |
908 | break; | |
909 | } | |
910 | ||
911 | if (entry) { | |
912 | int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length : | |
913 | BBM_LOG_MAX_LBA_ENTRY_VAL; | |
914 | entry->defective_block_start = __cpu_to_le48(sector); | |
915 | entry->marked_count = cnt - 1; | |
916 | if (cnt == length) | |
917 | return 1; | |
918 | sector += cnt; | |
919 | length -= cnt; | |
920 | } | |
921 | ||
922 | new_bb = ROUND_UP(length, BBM_LOG_MAX_LBA_ENTRY_VAL) / | |
923 | BBM_LOG_MAX_LBA_ENTRY_VAL; | |
924 | if (log->entry_count + new_bb > BBM_LOG_MAX_ENTRIES) | |
925 | return 0; | |
926 | ||
927 | while (length > 0) { | |
928 | int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length : | |
929 | BBM_LOG_MAX_LBA_ENTRY_VAL; | |
930 | struct bbm_log_entry *entry = | |
931 | &log->marked_block_entries[log->entry_count]; | |
932 | ||
933 | entry->defective_block_start = __cpu_to_le48(sector); | |
934 | entry->marked_count = cnt - 1; | |
935 | entry->disk_ordinal = idx; | |
936 | ||
937 | sector += cnt; | |
938 | length -= cnt; | |
939 | ||
940 | log->entry_count++; | |
941 | } | |
942 | ||
943 | return new_bb; | |
944 | } | |
c07a5a4f | 945 | |
4c9e8c1e TM |
946 | /* clear all bad blocks for given disk */ |
947 | static void clear_disk_badblocks(struct bbm_log *log, const __u8 idx) | |
948 | { | |
949 | __u32 i = 0; | |
950 | ||
951 | while (i < log->entry_count) { | |
952 | struct bbm_log_entry *entries = log->marked_block_entries; | |
953 | ||
954 | if (entries[i].disk_ordinal == idx) { | |
955 | if (i < log->entry_count - 1) | |
956 | entries[i] = entries[log->entry_count - 1]; | |
957 | log->entry_count--; | |
958 | } else { | |
959 | i++; | |
960 | } | |
961 | } | |
962 | } | |
963 | ||
c07a5a4f TM |
964 | /* clear given bad block */ |
965 | static int clear_badblock(struct bbm_log *log, const __u8 idx, const unsigned | |
966 | long long sector, const int length) { | |
967 | __u32 i = 0; | |
968 | ||
969 | while (i < log->entry_count) { | |
970 | struct bbm_log_entry *entries = log->marked_block_entries; | |
971 | ||
972 | if ((entries[i].disk_ordinal == idx) && | |
973 | (__le48_to_cpu(&entries[i].defective_block_start) == | |
974 | sector) && (entries[i].marked_count + 1 == length)) { | |
975 | if (i < log->entry_count - 1) | |
976 | entries[i] = entries[log->entry_count - 1]; | |
977 | log->entry_count--; | |
978 | break; | |
979 | } | |
980 | i++; | |
981 | } | |
982 | ||
983 | return 1; | |
984 | } | |
8d67477f TM |
985 | |
986 | /* allocate and load BBM log from metadata */ | |
987 | static int load_bbm_log(struct intel_super *super) | |
988 | { | |
989 | struct imsm_super *mpb = super->anchor; | |
990 | __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size); | |
991 | ||
992 | super->bbm_log = xcalloc(1, sizeof(struct bbm_log)); | |
993 | if (!super->bbm_log) | |
994 | return 1; | |
995 | ||
996 | if (bbm_log_size) { | |
997 | struct bbm_log *log = (void *)mpb + | |
998 | __le32_to_cpu(mpb->mpb_size) - bbm_log_size; | |
999 | ||
1000 | __u32 entry_count; | |
1001 | ||
1002 | if (bbm_log_size < sizeof(log->signature) + | |
1003 | sizeof(log->entry_count)) | |
1004 | return 2; | |
1005 | ||
1006 | entry_count = __le32_to_cpu(log->entry_count); | |
1007 | if ((__le32_to_cpu(log->signature) != BBM_LOG_SIGNATURE) || | |
1008 | (entry_count > BBM_LOG_MAX_ENTRIES)) | |
1009 | return 3; | |
1010 | ||
1011 | if (bbm_log_size != | |
1012 | sizeof(log->signature) + sizeof(log->entry_count) + | |
1013 | entry_count * sizeof(struct bbm_log_entry)) | |
1014 | return 4; | |
1015 | ||
1016 | memcpy(super->bbm_log, log, bbm_log_size); | |
1017 | } else { | |
1018 | super->bbm_log->signature = __cpu_to_le32(BBM_LOG_SIGNATURE); | |
1019 | super->bbm_log->entry_count = 0; | |
1020 | } | |
1021 | ||
1022 | return 0; | |
1023 | } | |
1024 | ||
b12796be TM |
1025 | /* checks if bad block is within volume boundaries */ |
1026 | static int is_bad_block_in_volume(const struct bbm_log_entry *entry, | |
1027 | const unsigned long long start_sector, | |
1028 | const unsigned long long size) | |
1029 | { | |
1030 | unsigned long long bb_start; | |
1031 | unsigned long long bb_end; | |
1032 | ||
1033 | bb_start = __le48_to_cpu(&entry->defective_block_start); | |
1034 | bb_end = bb_start + (entry->marked_count + 1); | |
1035 | ||
1036 | if (((bb_start >= start_sector) && (bb_start < start_sector + size)) || | |
1037 | ((bb_end >= start_sector) && (bb_end <= start_sector + size))) | |
1038 | return 1; | |
1039 | ||
1040 | return 0; | |
1041 | } | |
1042 | ||
1043 | /* get list of bad blocks on a drive for a volume */ | |
1044 | static void get_volume_badblocks(const struct bbm_log *log, const __u8 idx, | |
1045 | const unsigned long long start_sector, | |
1046 | const unsigned long long size, | |
1047 | struct md_bb *bbs) | |
1048 | { | |
1049 | __u32 count = 0; | |
1050 | __u32 i; | |
1051 | ||
1052 | for (i = 0; i < log->entry_count; i++) { | |
1053 | const struct bbm_log_entry *ent = | |
1054 | &log->marked_block_entries[i]; | |
1055 | struct md_bb_entry *bb; | |
1056 | ||
1057 | if ((ent->disk_ordinal == idx) && | |
1058 | is_bad_block_in_volume(ent, start_sector, size)) { | |
1059 | ||
1060 | if (!bbs->entries) { | |
1061 | bbs->entries = xmalloc(BBM_LOG_MAX_ENTRIES * | |
1062 | sizeof(*bb)); | |
1063 | if (!bbs->entries) | |
1064 | break; | |
1065 | } | |
1066 | ||
1067 | bb = &bbs->entries[count++]; | |
1068 | bb->sector = __le48_to_cpu(&ent->defective_block_start); | |
1069 | bb->length = ent->marked_count + 1; | |
1070 | } | |
1071 | } | |
1072 | bbs->count = count; | |
1073 | } | |
1074 | ||
98130f40 AK |
1075 | /* |
1076 | * for second_map: | |
238c0a71 AK |
1077 | * == MAP_0 get first map |
1078 | * == MAP_1 get second map | |
1079 | * == MAP_X than get map according to the current migr_state | |
98130f40 AK |
1080 | */ |
1081 | static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev, | |
1082 | int slot, | |
1083 | int second_map) | |
7eef0453 DW |
1084 | { |
1085 | struct imsm_map *map; | |
1086 | ||
5e7b0330 | 1087 | map = get_imsm_map(dev, second_map); |
7eef0453 | 1088 | |
ff077194 DW |
1089 | /* top byte identifies disk under rebuild */ |
1090 | return __le32_to_cpu(map->disk_ord_tbl[slot]); | |
1091 | } | |
1092 | ||
1093 | #define ord_to_idx(ord) (((ord) << 8) >> 8) | |
98130f40 | 1094 | static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot, int second_map) |
ff077194 | 1095 | { |
98130f40 | 1096 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, second_map); |
ff077194 DW |
1097 | |
1098 | return ord_to_idx(ord); | |
7eef0453 DW |
1099 | } |
1100 | ||
be73972f DW |
1101 | static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord) |
1102 | { | |
1103 | map->disk_ord_tbl[slot] = __cpu_to_le32(ord); | |
1104 | } | |
1105 | ||
f21e18ca | 1106 | static int get_imsm_disk_slot(struct imsm_map *map, unsigned idx) |
620b1713 DW |
1107 | { |
1108 | int slot; | |
1109 | __u32 ord; | |
1110 | ||
1111 | for (slot = 0; slot < map->num_members; slot++) { | |
1112 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
1113 | if (ord_to_idx(ord) == idx) | |
1114 | return slot; | |
1115 | } | |
1116 | ||
1117 | return -1; | |
1118 | } | |
1119 | ||
cdddbdbc DW |
1120 | static int get_imsm_raid_level(struct imsm_map *map) |
1121 | { | |
1122 | if (map->raid_level == 1) { | |
1123 | if (map->num_members == 2) | |
1124 | return 1; | |
1125 | else | |
1126 | return 10; | |
1127 | } | |
1128 | ||
1129 | return map->raid_level; | |
1130 | } | |
1131 | ||
c2c087e6 DW |
1132 | static int cmp_extent(const void *av, const void *bv) |
1133 | { | |
1134 | const struct extent *a = av; | |
1135 | const struct extent *b = bv; | |
1136 | if (a->start < b->start) | |
1137 | return -1; | |
1138 | if (a->start > b->start) | |
1139 | return 1; | |
1140 | return 0; | |
1141 | } | |
1142 | ||
0dcecb2e | 1143 | static int count_memberships(struct dl *dl, struct intel_super *super) |
c2c087e6 | 1144 | { |
c2c087e6 | 1145 | int memberships = 0; |
620b1713 | 1146 | int i; |
c2c087e6 | 1147 | |
949c47a0 DW |
1148 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
1149 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 1150 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
c2c087e6 | 1151 | |
620b1713 DW |
1152 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
1153 | memberships++; | |
c2c087e6 | 1154 | } |
0dcecb2e DW |
1155 | |
1156 | return memberships; | |
1157 | } | |
1158 | ||
b81221b7 CA |
1159 | static __u32 imsm_min_reserved_sectors(struct intel_super *super); |
1160 | ||
5551b113 CA |
1161 | static int split_ull(unsigned long long n, __u32 *lo, __u32 *hi) |
1162 | { | |
1163 | if (lo == 0 || hi == 0) | |
1164 | return 1; | |
1165 | *lo = __le32_to_cpu((unsigned)n); | |
1166 | *hi = __le32_to_cpu((unsigned)(n >> 32)); | |
1167 | return 0; | |
1168 | } | |
1169 | ||
1170 | static unsigned long long join_u32(__u32 lo, __u32 hi) | |
1171 | { | |
1172 | return (unsigned long long)__le32_to_cpu(lo) | | |
1173 | (((unsigned long long)__le32_to_cpu(hi)) << 32); | |
1174 | } | |
1175 | ||
1176 | static unsigned long long total_blocks(struct imsm_disk *disk) | |
1177 | { | |
1178 | if (disk == NULL) | |
1179 | return 0; | |
1180 | return join_u32(disk->total_blocks_lo, disk->total_blocks_hi); | |
1181 | } | |
1182 | ||
1183 | static unsigned long long pba_of_lba0(struct imsm_map *map) | |
1184 | { | |
1185 | if (map == NULL) | |
1186 | return 0; | |
1187 | return join_u32(map->pba_of_lba0_lo, map->pba_of_lba0_hi); | |
1188 | } | |
1189 | ||
1190 | static unsigned long long blocks_per_member(struct imsm_map *map) | |
1191 | { | |
1192 | if (map == NULL) | |
1193 | return 0; | |
1194 | return join_u32(map->blocks_per_member_lo, map->blocks_per_member_hi); | |
1195 | } | |
1196 | ||
1197 | static unsigned long long num_data_stripes(struct imsm_map *map) | |
1198 | { | |
1199 | if (map == NULL) | |
1200 | return 0; | |
1201 | return join_u32(map->num_data_stripes_lo, map->num_data_stripes_hi); | |
1202 | } | |
1203 | ||
1204 | static void set_total_blocks(struct imsm_disk *disk, unsigned long long n) | |
1205 | { | |
1206 | split_ull(n, &disk->total_blocks_lo, &disk->total_blocks_hi); | |
1207 | } | |
1208 | ||
1209 | static void set_pba_of_lba0(struct imsm_map *map, unsigned long long n) | |
1210 | { | |
1211 | split_ull(n, &map->pba_of_lba0_lo, &map->pba_of_lba0_hi); | |
1212 | } | |
1213 | ||
1214 | static void set_blocks_per_member(struct imsm_map *map, unsigned long long n) | |
1215 | { | |
1216 | split_ull(n, &map->blocks_per_member_lo, &map->blocks_per_member_hi); | |
1217 | } | |
1218 | ||
1219 | static void set_num_data_stripes(struct imsm_map *map, unsigned long long n) | |
1220 | { | |
1221 | split_ull(n, &map->num_data_stripes_lo, &map->num_data_stripes_hi); | |
1222 | } | |
1223 | ||
0dcecb2e DW |
1224 | static struct extent *get_extents(struct intel_super *super, struct dl *dl) |
1225 | { | |
1226 | /* find a list of used extents on the given physical device */ | |
1227 | struct extent *rv, *e; | |
620b1713 | 1228 | int i; |
0dcecb2e | 1229 | int memberships = count_memberships(dl, super); |
b276dd33 DW |
1230 | __u32 reservation; |
1231 | ||
1232 | /* trim the reserved area for spares, so they can join any array | |
1233 | * regardless of whether the OROM has assigned sectors from the | |
1234 | * IMSM_RESERVED_SECTORS region | |
1235 | */ | |
1236 | if (dl->index == -1) | |
b81221b7 | 1237 | reservation = imsm_min_reserved_sectors(super); |
b276dd33 DW |
1238 | else |
1239 | reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
0dcecb2e | 1240 | |
503975b9 | 1241 | rv = xcalloc(sizeof(struct extent), (memberships + 1)); |
c2c087e6 DW |
1242 | e = rv; |
1243 | ||
949c47a0 DW |
1244 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
1245 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 1246 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
c2c087e6 | 1247 | |
620b1713 | 1248 | if (get_imsm_disk_slot(map, dl->index) >= 0) { |
5551b113 CA |
1249 | e->start = pba_of_lba0(map); |
1250 | e->size = blocks_per_member(map); | |
620b1713 | 1251 | e++; |
c2c087e6 DW |
1252 | } |
1253 | } | |
1254 | qsort(rv, memberships, sizeof(*rv), cmp_extent); | |
1255 | ||
1011e834 | 1256 | /* determine the start of the metadata |
14e8215b DW |
1257 | * when no raid devices are defined use the default |
1258 | * ...otherwise allow the metadata to truncate the value | |
1259 | * as is the case with older versions of imsm | |
1260 | */ | |
1261 | if (memberships) { | |
1262 | struct extent *last = &rv[memberships - 1]; | |
5551b113 | 1263 | unsigned long long remainder; |
14e8215b | 1264 | |
5551b113 | 1265 | remainder = total_blocks(&dl->disk) - (last->start + last->size); |
dda5855f DW |
1266 | /* round down to 1k block to satisfy precision of the kernel |
1267 | * 'size' interface | |
1268 | */ | |
1269 | remainder &= ~1UL; | |
1270 | /* make sure remainder is still sane */ | |
f21e18ca | 1271 | if (remainder < (unsigned)ROUND_UP(super->len, 512) >> 9) |
dda5855f | 1272 | remainder = ROUND_UP(super->len, 512) >> 9; |
14e8215b DW |
1273 | if (reservation > remainder) |
1274 | reservation = remainder; | |
1275 | } | |
5551b113 | 1276 | e->start = total_blocks(&dl->disk) - reservation; |
c2c087e6 DW |
1277 | e->size = 0; |
1278 | return rv; | |
1279 | } | |
1280 | ||
14e8215b DW |
1281 | /* try to determine how much space is reserved for metadata from |
1282 | * the last get_extents() entry, otherwise fallback to the | |
1283 | * default | |
1284 | */ | |
1285 | static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl) | |
1286 | { | |
1287 | struct extent *e; | |
1288 | int i; | |
1289 | __u32 rv; | |
1290 | ||
1291 | /* for spares just return a minimal reservation which will grow | |
1292 | * once the spare is picked up by an array | |
1293 | */ | |
1294 | if (dl->index == -1) | |
1295 | return MPB_SECTOR_CNT; | |
1296 | ||
1297 | e = get_extents(super, dl); | |
1298 | if (!e) | |
1299 | return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
1300 | ||
1301 | /* scroll to last entry */ | |
1302 | for (i = 0; e[i].size; i++) | |
1303 | continue; | |
1304 | ||
5551b113 | 1305 | rv = total_blocks(&dl->disk) - e[i].start; |
14e8215b DW |
1306 | |
1307 | free(e); | |
1308 | ||
1309 | return rv; | |
1310 | } | |
1311 | ||
25ed7e59 DW |
1312 | static int is_spare(struct imsm_disk *disk) |
1313 | { | |
1314 | return (disk->status & SPARE_DISK) == SPARE_DISK; | |
1315 | } | |
1316 | ||
1317 | static int is_configured(struct imsm_disk *disk) | |
1318 | { | |
1319 | return (disk->status & CONFIGURED_DISK) == CONFIGURED_DISK; | |
1320 | } | |
1321 | ||
1322 | static int is_failed(struct imsm_disk *disk) | |
1323 | { | |
1324 | return (disk->status & FAILED_DISK) == FAILED_DISK; | |
1325 | } | |
1326 | ||
2432ce9b AP |
1327 | static int is_journal(struct imsm_disk *disk) |
1328 | { | |
1329 | return (disk->status & JOURNAL_DISK) == JOURNAL_DISK; | |
1330 | } | |
1331 | ||
b53bfba6 TM |
1332 | /* round array size down to closest MB and ensure it splits evenly |
1333 | * between members | |
1334 | */ | |
1335 | static unsigned long long round_size_to_mb(unsigned long long size, unsigned int | |
1336 | disk_count) | |
1337 | { | |
1338 | size /= disk_count; | |
1339 | size = (size >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; | |
1340 | size *= disk_count; | |
1341 | ||
1342 | return size; | |
1343 | } | |
1344 | ||
8b9cd157 MK |
1345 | static int able_to_resync(int raid_level, int missing_disks) |
1346 | { | |
1347 | int max_missing_disks = 0; | |
1348 | ||
1349 | switch (raid_level) { | |
1350 | case 10: | |
1351 | max_missing_disks = 1; | |
1352 | break; | |
1353 | default: | |
1354 | max_missing_disks = 0; | |
1355 | } | |
1356 | return missing_disks <= max_missing_disks; | |
1357 | } | |
1358 | ||
b81221b7 CA |
1359 | /* try to determine how much space is reserved for metadata from |
1360 | * the last get_extents() entry on the smallest active disk, | |
1361 | * otherwise fallback to the default | |
1362 | */ | |
1363 | static __u32 imsm_min_reserved_sectors(struct intel_super *super) | |
1364 | { | |
1365 | struct extent *e; | |
1366 | int i; | |
5551b113 CA |
1367 | unsigned long long min_active; |
1368 | __u32 remainder; | |
b81221b7 CA |
1369 | __u32 rv = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; |
1370 | struct dl *dl, *dl_min = NULL; | |
1371 | ||
1372 | if (!super) | |
1373 | return rv; | |
1374 | ||
1375 | min_active = 0; | |
1376 | for (dl = super->disks; dl; dl = dl->next) { | |
1377 | if (dl->index < 0) | |
1378 | continue; | |
5551b113 CA |
1379 | unsigned long long blocks = total_blocks(&dl->disk); |
1380 | if (blocks < min_active || min_active == 0) { | |
b81221b7 | 1381 | dl_min = dl; |
5551b113 | 1382 | min_active = blocks; |
b81221b7 CA |
1383 | } |
1384 | } | |
1385 | if (!dl_min) | |
1386 | return rv; | |
1387 | ||
1388 | /* find last lba used by subarrays on the smallest active disk */ | |
1389 | e = get_extents(super, dl_min); | |
1390 | if (!e) | |
1391 | return rv; | |
1392 | for (i = 0; e[i].size; i++) | |
1393 | continue; | |
1394 | ||
1395 | remainder = min_active - e[i].start; | |
1396 | free(e); | |
1397 | ||
1398 | /* to give priority to recovery we should not require full | |
1399 | IMSM_RESERVED_SECTORS from the spare */ | |
1400 | rv = MPB_SECTOR_CNT + NUM_BLOCKS_DIRTY_STRIPE_REGION; | |
1401 | ||
1402 | /* if real reservation is smaller use that value */ | |
1403 | return (remainder < rv) ? remainder : rv; | |
1404 | } | |
1405 | ||
fbfdcb06 AO |
1406 | /* |
1407 | * Return minimum size of a spare and sector size | |
1408 | * that can be used in this array | |
1409 | */ | |
1410 | int get_spare_criteria_imsm(struct supertype *st, struct spare_criteria *c) | |
80e7f8c3 AC |
1411 | { |
1412 | struct intel_super *super = st->sb; | |
1413 | struct dl *dl; | |
1414 | struct extent *e; | |
1415 | int i; | |
fbfdcb06 AO |
1416 | unsigned long long size = 0; |
1417 | ||
1418 | c->min_size = 0; | |
4b57ecf6 | 1419 | c->sector_size = 0; |
80e7f8c3 AC |
1420 | |
1421 | if (!super) | |
fbfdcb06 | 1422 | return -EINVAL; |
80e7f8c3 AC |
1423 | /* find first active disk in array */ |
1424 | dl = super->disks; | |
1425 | while (dl && (is_failed(&dl->disk) || dl->index == -1)) | |
1426 | dl = dl->next; | |
1427 | if (!dl) | |
fbfdcb06 | 1428 | return -EINVAL; |
80e7f8c3 AC |
1429 | /* find last lba used by subarrays */ |
1430 | e = get_extents(super, dl); | |
1431 | if (!e) | |
fbfdcb06 | 1432 | return -EINVAL; |
80e7f8c3 AC |
1433 | for (i = 0; e[i].size; i++) |
1434 | continue; | |
1435 | if (i > 0) | |
fbfdcb06 | 1436 | size = e[i-1].start + e[i-1].size; |
80e7f8c3 | 1437 | free(e); |
b81221b7 | 1438 | |
80e7f8c3 | 1439 | /* add the amount of space needed for metadata */ |
fbfdcb06 AO |
1440 | size += imsm_min_reserved_sectors(super); |
1441 | ||
1442 | c->min_size = size * 512; | |
4b57ecf6 | 1443 | c->sector_size = super->sector_size; |
b81221b7 | 1444 | |
fbfdcb06 | 1445 | return 0; |
80e7f8c3 AC |
1446 | } |
1447 | ||
d1e02575 AK |
1448 | static int is_gen_migration(struct imsm_dev *dev); |
1449 | ||
f36a9ecd PB |
1450 | #define IMSM_4K_DIV 8 |
1451 | ||
c47b0ff6 AK |
1452 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
1453 | struct imsm_dev *dev); | |
1e5c6983 | 1454 | |
c47b0ff6 AK |
1455 | static void print_imsm_dev(struct intel_super *super, |
1456 | struct imsm_dev *dev, | |
1457 | char *uuid, | |
1458 | int disk_idx) | |
cdddbdbc DW |
1459 | { |
1460 | __u64 sz; | |
0d80bb2f | 1461 | int slot, i; |
238c0a71 AK |
1462 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1463 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); | |
b10b37b8 | 1464 | __u32 ord; |
cdddbdbc DW |
1465 | |
1466 | printf("\n"); | |
1e7bc0ed | 1467 | printf("[%.16s]:\n", dev->volume); |
44470971 | 1468 | printf(" UUID : %s\n", uuid); |
dd8bcb3b AK |
1469 | printf(" RAID Level : %d", get_imsm_raid_level(map)); |
1470 | if (map2) | |
1471 | printf(" <-- %d", get_imsm_raid_level(map2)); | |
1472 | printf("\n"); | |
1473 | printf(" Members : %d", map->num_members); | |
1474 | if (map2) | |
1475 | printf(" <-- %d", map2->num_members); | |
1476 | printf("\n"); | |
0d80bb2f DW |
1477 | printf(" Slots : ["); |
1478 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 1479 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_0); |
0d80bb2f DW |
1480 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
1481 | } | |
dd8bcb3b AK |
1482 | printf("]"); |
1483 | if (map2) { | |
1484 | printf(" <-- ["); | |
1485 | for (i = 0; i < map2->num_members; i++) { | |
238c0a71 | 1486 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_1); |
dd8bcb3b AK |
1487 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
1488 | } | |
1489 | printf("]"); | |
1490 | } | |
1491 | printf("\n"); | |
7095bccb AK |
1492 | printf(" Failed disk : "); |
1493 | if (map->failed_disk_num == 0xff) | |
1494 | printf("none"); | |
1495 | else | |
1496 | printf("%i", map->failed_disk_num); | |
1497 | printf("\n"); | |
620b1713 DW |
1498 | slot = get_imsm_disk_slot(map, disk_idx); |
1499 | if (slot >= 0) { | |
238c0a71 | 1500 | ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); |
b10b37b8 DW |
1501 | printf(" This Slot : %d%s\n", slot, |
1502 | ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : ""); | |
1503 | } else | |
cdddbdbc | 1504 | printf(" This Slot : ?\n"); |
84918897 | 1505 | printf(" Sector Size : %u\n", super->sector_size); |
cdddbdbc DW |
1506 | sz = __le32_to_cpu(dev->size_high); |
1507 | sz <<= 32; | |
1508 | sz += __le32_to_cpu(dev->size_low); | |
84918897 MK |
1509 | printf(" Array Size : %llu%s\n", |
1510 | (unsigned long long)sz * 512 / super->sector_size, | |
cdddbdbc | 1511 | human_size(sz * 512)); |
5551b113 | 1512 | sz = blocks_per_member(map); |
84918897 MK |
1513 | printf(" Per Dev Size : %llu%s\n", |
1514 | (unsigned long long)sz * 512 / super->sector_size, | |
cdddbdbc | 1515 | human_size(sz * 512)); |
5551b113 CA |
1516 | printf(" Sector Offset : %llu\n", |
1517 | pba_of_lba0(map)); | |
1518 | printf(" Num Stripes : %llu\n", | |
1519 | num_data_stripes(map)); | |
dd8bcb3b | 1520 | printf(" Chunk Size : %u KiB", |
cdddbdbc | 1521 | __le16_to_cpu(map->blocks_per_strip) / 2); |
dd8bcb3b AK |
1522 | if (map2) |
1523 | printf(" <-- %u KiB", | |
1524 | __le16_to_cpu(map2->blocks_per_strip) / 2); | |
1525 | printf("\n"); | |
cdddbdbc | 1526 | printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks)); |
8655a7b1 | 1527 | printf(" Migrate State : "); |
1484e727 DW |
1528 | if (dev->vol.migr_state) { |
1529 | if (migr_type(dev) == MIGR_INIT) | |
8655a7b1 | 1530 | printf("initialize\n"); |
1484e727 | 1531 | else if (migr_type(dev) == MIGR_REBUILD) |
8655a7b1 | 1532 | printf("rebuild\n"); |
1484e727 | 1533 | else if (migr_type(dev) == MIGR_VERIFY) |
8655a7b1 | 1534 | printf("check\n"); |
1484e727 | 1535 | else if (migr_type(dev) == MIGR_GEN_MIGR) |
8655a7b1 | 1536 | printf("general migration\n"); |
1484e727 | 1537 | else if (migr_type(dev) == MIGR_STATE_CHANGE) |
8655a7b1 | 1538 | printf("state change\n"); |
1484e727 | 1539 | else if (migr_type(dev) == MIGR_REPAIR) |
8655a7b1 | 1540 | printf("repair\n"); |
1484e727 | 1541 | else |
8655a7b1 DW |
1542 | printf("<unknown:%d>\n", migr_type(dev)); |
1543 | } else | |
1544 | printf("idle\n"); | |
3393c6af DW |
1545 | printf(" Map State : %s", map_state_str[map->map_state]); |
1546 | if (dev->vol.migr_state) { | |
238c0a71 | 1547 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 | 1548 | |
b10b37b8 | 1549 | printf(" <-- %s", map_state_str[map->map_state]); |
464d40e8 LD |
1550 | printf("\n Checkpoint : %u ", |
1551 | __le32_to_cpu(dev->vol.curr_migr_unit)); | |
089f9d79 | 1552 | if (is_gen_migration(dev) && (slot > 1 || slot < 0)) |
464d40e8 LD |
1553 | printf("(N/A)"); |
1554 | else | |
1555 | printf("(%llu)", (unsigned long long) | |
1556 | blocks_per_migr_unit(super, dev)); | |
3393c6af DW |
1557 | } |
1558 | printf("\n"); | |
2432ce9b AP |
1559 | printf(" Dirty State : %s\n", (dev->vol.dirty & RAIDVOL_DIRTY) ? |
1560 | "dirty" : "clean"); | |
1561 | printf(" RWH Policy : "); | |
c2462068 | 1562 | if (dev->rwh_policy == RWH_OFF || dev->rwh_policy == RWH_MULTIPLE_OFF) |
2432ce9b AP |
1563 | printf("off\n"); |
1564 | else if (dev->rwh_policy == RWH_DISTRIBUTED) | |
1565 | printf("PPL distributed\n"); | |
1566 | else if (dev->rwh_policy == RWH_JOURNALING_DRIVE) | |
1567 | printf("PPL journaling drive\n"); | |
c2462068 PB |
1568 | else if (dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED) |
1569 | printf("Multiple distributed PPLs\n"); | |
1570 | else if (dev->rwh_policy == RWH_MULTIPLE_PPLS_JOURNALING_DRIVE) | |
1571 | printf("Multiple PPLs on journaling drive\n"); | |
2432ce9b AP |
1572 | else |
1573 | printf("<unknown:%d>\n", dev->rwh_policy); | |
cdddbdbc DW |
1574 | } |
1575 | ||
ef5c214e MK |
1576 | static void print_imsm_disk(struct imsm_disk *disk, |
1577 | int index, | |
1578 | __u32 reserved, | |
1579 | unsigned int sector_size) { | |
1f24f035 | 1580 | char str[MAX_RAID_SERIAL_LEN + 1]; |
cdddbdbc DW |
1581 | __u64 sz; |
1582 | ||
0ec1f4e8 | 1583 | if (index < -1 || !disk) |
e9d82038 DW |
1584 | return; |
1585 | ||
cdddbdbc | 1586 | printf("\n"); |
1f24f035 | 1587 | snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial); |
0ec1f4e8 DW |
1588 | if (index >= 0) |
1589 | printf(" Disk%02d Serial : %s\n", index, str); | |
1590 | else | |
1591 | printf(" Disk Serial : %s\n", str); | |
2432ce9b AP |
1592 | printf(" State :%s%s%s%s\n", is_spare(disk) ? " spare" : "", |
1593 | is_configured(disk) ? " active" : "", | |
1594 | is_failed(disk) ? " failed" : "", | |
1595 | is_journal(disk) ? " journal" : ""); | |
cdddbdbc | 1596 | printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id)); |
5551b113 | 1597 | sz = total_blocks(disk) - reserved; |
ef5c214e MK |
1598 | printf(" Usable Size : %llu%s\n", |
1599 | (unsigned long long)sz * 512 / sector_size, | |
cdddbdbc DW |
1600 | human_size(sz * 512)); |
1601 | } | |
1602 | ||
de44e46f PB |
1603 | void convert_to_4k_imsm_migr_rec(struct intel_super *super) |
1604 | { | |
1605 | struct migr_record *migr_rec = super->migr_rec; | |
1606 | ||
1607 | migr_rec->blocks_per_unit /= IMSM_4K_DIV; | |
1608 | migr_rec->ckpt_area_pba /= IMSM_4K_DIV; | |
1609 | migr_rec->dest_1st_member_lba /= IMSM_4K_DIV; | |
1610 | migr_rec->dest_depth_per_unit /= IMSM_4K_DIV; | |
1611 | split_ull((join_u32(migr_rec->post_migr_vol_cap, | |
1612 | migr_rec->post_migr_vol_cap_hi) / IMSM_4K_DIV), | |
1613 | &migr_rec->post_migr_vol_cap, &migr_rec->post_migr_vol_cap_hi); | |
1614 | } | |
1615 | ||
f36a9ecd PB |
1616 | void convert_to_4k_imsm_disk(struct imsm_disk *disk) |
1617 | { | |
1618 | set_total_blocks(disk, (total_blocks(disk)/IMSM_4K_DIV)); | |
1619 | } | |
1620 | ||
1621 | void convert_to_4k(struct intel_super *super) | |
1622 | { | |
1623 | struct imsm_super *mpb = super->anchor; | |
1624 | struct imsm_disk *disk; | |
1625 | int i; | |
e4467bc7 | 1626 | __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size); |
f36a9ecd PB |
1627 | |
1628 | for (i = 0; i < mpb->num_disks ; i++) { | |
1629 | disk = __get_imsm_disk(mpb, i); | |
1630 | /* disk */ | |
1631 | convert_to_4k_imsm_disk(disk); | |
1632 | } | |
1633 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
1634 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1635 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
1636 | /* dev */ | |
1637 | split_ull((join_u32(dev->size_low, dev->size_high)/IMSM_4K_DIV), | |
1638 | &dev->size_low, &dev->size_high); | |
1639 | dev->vol.curr_migr_unit /= IMSM_4K_DIV; | |
1640 | ||
1641 | /* map0 */ | |
1642 | set_blocks_per_member(map, blocks_per_member(map)/IMSM_4K_DIV); | |
1643 | map->blocks_per_strip /= IMSM_4K_DIV; | |
1644 | set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV); | |
1645 | ||
1646 | if (dev->vol.migr_state) { | |
1647 | /* map1 */ | |
1648 | map = get_imsm_map(dev, MAP_1); | |
1649 | set_blocks_per_member(map, | |
1650 | blocks_per_member(map)/IMSM_4K_DIV); | |
1651 | map->blocks_per_strip /= IMSM_4K_DIV; | |
1652 | set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV); | |
1653 | } | |
1654 | } | |
e4467bc7 TM |
1655 | if (bbm_log_size) { |
1656 | struct bbm_log *log = (void *)mpb + | |
1657 | __le32_to_cpu(mpb->mpb_size) - bbm_log_size; | |
1658 | __u32 i; | |
1659 | ||
1660 | for (i = 0; i < log->entry_count; i++) { | |
1661 | struct bbm_log_entry *entry = | |
1662 | &log->marked_block_entries[i]; | |
1663 | ||
1664 | __u8 count = entry->marked_count + 1; | |
1665 | unsigned long long sector = | |
1666 | __le48_to_cpu(&entry->defective_block_start); | |
1667 | ||
1668 | entry->defective_block_start = | |
1669 | __cpu_to_le48(sector/IMSM_4K_DIV); | |
1670 | entry->marked_count = max(count/IMSM_4K_DIV, 1) - 1; | |
1671 | } | |
1672 | } | |
f36a9ecd PB |
1673 | |
1674 | mpb->check_sum = __gen_imsm_checksum(mpb); | |
1675 | } | |
1676 | ||
520e69e2 AK |
1677 | void examine_migr_rec_imsm(struct intel_super *super) |
1678 | { | |
1679 | struct migr_record *migr_rec = super->migr_rec; | |
1680 | struct imsm_super *mpb = super->anchor; | |
1681 | int i; | |
1682 | ||
1683 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
1684 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
3136abe5 | 1685 | struct imsm_map *map; |
b4ab44d8 | 1686 | int slot = -1; |
3136abe5 | 1687 | |
520e69e2 AK |
1688 | if (is_gen_migration(dev) == 0) |
1689 | continue; | |
1690 | ||
1691 | printf("\nMigration Record Information:"); | |
3136abe5 | 1692 | |
44bfe6df AK |
1693 | /* first map under migration */ |
1694 | map = get_imsm_map(dev, MAP_0); | |
3136abe5 AK |
1695 | if (map) |
1696 | slot = get_imsm_disk_slot(map, super->disks->index); | |
089f9d79 | 1697 | if (map == NULL || slot > 1 || slot < 0) { |
520e69e2 AK |
1698 | printf(" Empty\n "); |
1699 | printf("Examine one of first two disks in array\n"); | |
1700 | break; | |
1701 | } | |
1702 | printf("\n Status : "); | |
1703 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
1704 | printf("Normal\n"); | |
1705 | else | |
1706 | printf("Contains Data\n"); | |
1707 | printf(" Current Unit : %u\n", | |
1708 | __le32_to_cpu(migr_rec->curr_migr_unit)); | |
1709 | printf(" Family : %u\n", | |
1710 | __le32_to_cpu(migr_rec->family_num)); | |
1711 | printf(" Ascending : %u\n", | |
1712 | __le32_to_cpu(migr_rec->ascending_migr)); | |
1713 | printf(" Blocks Per Unit : %u\n", | |
1714 | __le32_to_cpu(migr_rec->blocks_per_unit)); | |
1715 | printf(" Dest. Depth Per Unit : %u\n", | |
1716 | __le32_to_cpu(migr_rec->dest_depth_per_unit)); | |
1717 | printf(" Checkpoint Area pba : %u\n", | |
1718 | __le32_to_cpu(migr_rec->ckpt_area_pba)); | |
1719 | printf(" First member lba : %u\n", | |
1720 | __le32_to_cpu(migr_rec->dest_1st_member_lba)); | |
1721 | printf(" Total Number of Units : %u\n", | |
1722 | __le32_to_cpu(migr_rec->num_migr_units)); | |
1723 | printf(" Size of volume : %u\n", | |
1724 | __le32_to_cpu(migr_rec->post_migr_vol_cap)); | |
1725 | printf(" Expansion space for LBA64 : %u\n", | |
1726 | __le32_to_cpu(migr_rec->post_migr_vol_cap_hi)); | |
1727 | printf(" Record was read from : %u\n", | |
1728 | __le32_to_cpu(migr_rec->ckpt_read_disk_num)); | |
1729 | ||
1730 | break; | |
1731 | } | |
1732 | } | |
f36a9ecd | 1733 | |
de44e46f PB |
1734 | void convert_from_4k_imsm_migr_rec(struct intel_super *super) |
1735 | { | |
1736 | struct migr_record *migr_rec = super->migr_rec; | |
1737 | ||
1738 | migr_rec->blocks_per_unit *= IMSM_4K_DIV; | |
1739 | migr_rec->ckpt_area_pba *= IMSM_4K_DIV; | |
1740 | migr_rec->dest_1st_member_lba *= IMSM_4K_DIV; | |
1741 | migr_rec->dest_depth_per_unit *= IMSM_4K_DIV; | |
1742 | split_ull((join_u32(migr_rec->post_migr_vol_cap, | |
1743 | migr_rec->post_migr_vol_cap_hi) * IMSM_4K_DIV), | |
1744 | &migr_rec->post_migr_vol_cap, | |
1745 | &migr_rec->post_migr_vol_cap_hi); | |
1746 | } | |
1747 | ||
f36a9ecd PB |
1748 | void convert_from_4k(struct intel_super *super) |
1749 | { | |
1750 | struct imsm_super *mpb = super->anchor; | |
1751 | struct imsm_disk *disk; | |
1752 | int i; | |
e4467bc7 | 1753 | __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size); |
f36a9ecd PB |
1754 | |
1755 | for (i = 0; i < mpb->num_disks ; i++) { | |
1756 | disk = __get_imsm_disk(mpb, i); | |
1757 | /* disk */ | |
1758 | set_total_blocks(disk, (total_blocks(disk)*IMSM_4K_DIV)); | |
1759 | } | |
1760 | ||
1761 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
1762 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1763 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
1764 | /* dev */ | |
1765 | split_ull((join_u32(dev->size_low, dev->size_high)*IMSM_4K_DIV), | |
1766 | &dev->size_low, &dev->size_high); | |
1767 | dev->vol.curr_migr_unit *= IMSM_4K_DIV; | |
1768 | ||
1769 | /* map0 */ | |
1770 | set_blocks_per_member(map, blocks_per_member(map)*IMSM_4K_DIV); | |
1771 | map->blocks_per_strip *= IMSM_4K_DIV; | |
1772 | set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV); | |
1773 | ||
1774 | if (dev->vol.migr_state) { | |
1775 | /* map1 */ | |
1776 | map = get_imsm_map(dev, MAP_1); | |
1777 | set_blocks_per_member(map, | |
1778 | blocks_per_member(map)*IMSM_4K_DIV); | |
1779 | map->blocks_per_strip *= IMSM_4K_DIV; | |
1780 | set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV); | |
1781 | } | |
1782 | } | |
e4467bc7 TM |
1783 | if (bbm_log_size) { |
1784 | struct bbm_log *log = (void *)mpb + | |
1785 | __le32_to_cpu(mpb->mpb_size) - bbm_log_size; | |
1786 | __u32 i; | |
1787 | ||
1788 | for (i = 0; i < log->entry_count; i++) { | |
1789 | struct bbm_log_entry *entry = | |
1790 | &log->marked_block_entries[i]; | |
1791 | ||
1792 | __u8 count = entry->marked_count + 1; | |
1793 | unsigned long long sector = | |
1794 | __le48_to_cpu(&entry->defective_block_start); | |
1795 | ||
1796 | entry->defective_block_start = | |
1797 | __cpu_to_le48(sector*IMSM_4K_DIV); | |
1798 | entry->marked_count = count*IMSM_4K_DIV - 1; | |
1799 | } | |
1800 | } | |
f36a9ecd PB |
1801 | |
1802 | mpb->check_sum = __gen_imsm_checksum(mpb); | |
1803 | } | |
1804 | ||
19482bcc AK |
1805 | /******************************************************************************* |
1806 | * function: imsm_check_attributes | |
1807 | * Description: Function checks if features represented by attributes flags | |
1011e834 | 1808 | * are supported by mdadm. |
19482bcc AK |
1809 | * Parameters: |
1810 | * attributes - Attributes read from metadata | |
1811 | * Returns: | |
1011e834 N |
1812 | * 0 - passed attributes contains unsupported features flags |
1813 | * 1 - all features are supported | |
19482bcc AK |
1814 | ******************************************************************************/ |
1815 | static int imsm_check_attributes(__u32 attributes) | |
1816 | { | |
1817 | int ret_val = 1; | |
418f9b36 N |
1818 | __u32 not_supported = MPB_ATTRIB_SUPPORTED^0xffffffff; |
1819 | ||
1820 | not_supported &= ~MPB_ATTRIB_IGNORED; | |
19482bcc AK |
1821 | |
1822 | not_supported &= attributes; | |
1823 | if (not_supported) { | |
e7b84f9d | 1824 | pr_err("(IMSM): Unsupported attributes : %x\n", |
418f9b36 | 1825 | (unsigned)__le32_to_cpu(not_supported)); |
19482bcc AK |
1826 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { |
1827 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY \n"); | |
1828 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
1829 | } | |
1830 | if (not_supported & MPB_ATTRIB_2TB) { | |
1831 | dprintf("\t\tMPB_ATTRIB_2TB\n"); | |
1832 | not_supported ^= MPB_ATTRIB_2TB; | |
1833 | } | |
1834 | if (not_supported & MPB_ATTRIB_RAID0) { | |
1835 | dprintf("\t\tMPB_ATTRIB_RAID0\n"); | |
1836 | not_supported ^= MPB_ATTRIB_RAID0; | |
1837 | } | |
1838 | if (not_supported & MPB_ATTRIB_RAID1) { | |
1839 | dprintf("\t\tMPB_ATTRIB_RAID1\n"); | |
1840 | not_supported ^= MPB_ATTRIB_RAID1; | |
1841 | } | |
1842 | if (not_supported & MPB_ATTRIB_RAID10) { | |
1843 | dprintf("\t\tMPB_ATTRIB_RAID10\n"); | |
1844 | not_supported ^= MPB_ATTRIB_RAID10; | |
1845 | } | |
1846 | if (not_supported & MPB_ATTRIB_RAID1E) { | |
1847 | dprintf("\t\tMPB_ATTRIB_RAID1E\n"); | |
1848 | not_supported ^= MPB_ATTRIB_RAID1E; | |
1849 | } | |
1850 | if (not_supported & MPB_ATTRIB_RAID5) { | |
1851 | dprintf("\t\tMPB_ATTRIB_RAID5\n"); | |
1852 | not_supported ^= MPB_ATTRIB_RAID5; | |
1853 | } | |
1854 | if (not_supported & MPB_ATTRIB_RAIDCNG) { | |
1855 | dprintf("\t\tMPB_ATTRIB_RAIDCNG\n"); | |
1856 | not_supported ^= MPB_ATTRIB_RAIDCNG; | |
1857 | } | |
1858 | if (not_supported & MPB_ATTRIB_BBM) { | |
1859 | dprintf("\t\tMPB_ATTRIB_BBM\n"); | |
1860 | not_supported ^= MPB_ATTRIB_BBM; | |
1861 | } | |
1862 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { | |
1863 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY (== MPB_ATTRIB_LEGACY)\n"); | |
1864 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
1865 | } | |
1866 | if (not_supported & MPB_ATTRIB_EXP_STRIPE_SIZE) { | |
1867 | dprintf("\t\tMPB_ATTRIB_EXP_STRIP_SIZE\n"); | |
1868 | not_supported ^= MPB_ATTRIB_EXP_STRIPE_SIZE; | |
1869 | } | |
1870 | if (not_supported & MPB_ATTRIB_2TB_DISK) { | |
1871 | dprintf("\t\tMPB_ATTRIB_2TB_DISK\n"); | |
1872 | not_supported ^= MPB_ATTRIB_2TB_DISK; | |
1873 | } | |
1874 | if (not_supported & MPB_ATTRIB_NEVER_USE2) { | |
1875 | dprintf("\t\tMPB_ATTRIB_NEVER_USE2\n"); | |
1876 | not_supported ^= MPB_ATTRIB_NEVER_USE2; | |
1877 | } | |
1878 | if (not_supported & MPB_ATTRIB_NEVER_USE) { | |
1879 | dprintf("\t\tMPB_ATTRIB_NEVER_USE\n"); | |
1880 | not_supported ^= MPB_ATTRIB_NEVER_USE; | |
1881 | } | |
1882 | ||
1883 | if (not_supported) | |
1ade5cc1 | 1884 | dprintf("(IMSM): Unknown attributes : %x\n", not_supported); |
19482bcc AK |
1885 | |
1886 | ret_val = 0; | |
1887 | } | |
1888 | ||
1889 | return ret_val; | |
1890 | } | |
1891 | ||
a5d85af7 | 1892 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map); |
44470971 | 1893 | |
cdddbdbc DW |
1894 | static void examine_super_imsm(struct supertype *st, char *homehost) |
1895 | { | |
1896 | struct intel_super *super = st->sb; | |
949c47a0 | 1897 | struct imsm_super *mpb = super->anchor; |
cdddbdbc DW |
1898 | char str[MAX_SIGNATURE_LENGTH]; |
1899 | int i; | |
27fd6274 DW |
1900 | struct mdinfo info; |
1901 | char nbuf[64]; | |
cdddbdbc | 1902 | __u32 sum; |
14e8215b | 1903 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
94827db3 | 1904 | struct dl *dl; |
27fd6274 | 1905 | |
618f4e6d XN |
1906 | strncpy(str, (char *)mpb->sig, MPB_SIG_LEN); |
1907 | str[MPB_SIG_LEN-1] = '\0'; | |
cdddbdbc DW |
1908 | printf(" Magic : %s\n", str); |
1909 | snprintf(str, strlen(MPB_VERSION_RAID0), "%s", get_imsm_version(mpb)); | |
1910 | printf(" Version : %s\n", get_imsm_version(mpb)); | |
148acb7b | 1911 | printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num)); |
cdddbdbc DW |
1912 | printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num)); |
1913 | printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num)); | |
19482bcc AK |
1914 | printf(" Attributes : "); |
1915 | if (imsm_check_attributes(mpb->attributes)) | |
1916 | printf("All supported\n"); | |
1917 | else | |
1918 | printf("not supported\n"); | |
a5d85af7 | 1919 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1920 | fname_from_uuid(st, &info, nbuf, ':'); |
27fd6274 | 1921 | printf(" UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
1922 | sum = __le32_to_cpu(mpb->check_sum); |
1923 | printf(" Checksum : %08x %s\n", sum, | |
949c47a0 | 1924 | __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect"); |
f36a9ecd | 1925 | printf(" MPB Sectors : %d\n", mpb_sectors(mpb, super->sector_size)); |
cdddbdbc DW |
1926 | printf(" Disks : %d\n", mpb->num_disks); |
1927 | printf(" RAID Devices : %d\n", mpb->num_raid_devs); | |
ef5c214e MK |
1928 | print_imsm_disk(__get_imsm_disk(mpb, super->disks->index), |
1929 | super->disks->index, reserved, super->sector_size); | |
8d67477f | 1930 | if (get_imsm_bbm_log_size(super->bbm_log)) { |
604b746f JD |
1931 | struct bbm_log *log = super->bbm_log; |
1932 | ||
1933 | printf("\n"); | |
1934 | printf("Bad Block Management Log:\n"); | |
1935 | printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size)); | |
1936 | printf(" Signature : %x\n", __le32_to_cpu(log->signature)); | |
1937 | printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count)); | |
604b746f | 1938 | } |
44470971 DW |
1939 | for (i = 0; i < mpb->num_raid_devs; i++) { |
1940 | struct mdinfo info; | |
1941 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1942 | ||
1943 | super->current_vol = i; | |
a5d85af7 | 1944 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1945 | fname_from_uuid(st, &info, nbuf, ':'); |
c47b0ff6 | 1946 | print_imsm_dev(super, dev, nbuf + 5, super->disks->index); |
44470971 | 1947 | } |
cdddbdbc DW |
1948 | for (i = 0; i < mpb->num_disks; i++) { |
1949 | if (i == super->disks->index) | |
1950 | continue; | |
ef5c214e MK |
1951 | print_imsm_disk(__get_imsm_disk(mpb, i), i, reserved, |
1952 | super->sector_size); | |
cdddbdbc | 1953 | } |
94827db3 | 1954 | |
0ec1f4e8 DW |
1955 | for (dl = super->disks; dl; dl = dl->next) |
1956 | if (dl->index == -1) | |
ef5c214e MK |
1957 | print_imsm_disk(&dl->disk, -1, reserved, |
1958 | super->sector_size); | |
520e69e2 AK |
1959 | |
1960 | examine_migr_rec_imsm(super); | |
cdddbdbc DW |
1961 | } |
1962 | ||
061f2c6a | 1963 | static void brief_examine_super_imsm(struct supertype *st, int verbose) |
cdddbdbc | 1964 | { |
27fd6274 | 1965 | /* We just write a generic IMSM ARRAY entry */ |
ff54de6e N |
1966 | struct mdinfo info; |
1967 | char nbuf[64]; | |
1e7bc0ed | 1968 | struct intel_super *super = st->sb; |
1e7bc0ed | 1969 | |
0d5a423f DW |
1970 | if (!super->anchor->num_raid_devs) { |
1971 | printf("ARRAY metadata=imsm\n"); | |
1e7bc0ed | 1972 | return; |
0d5a423f | 1973 | } |
ff54de6e | 1974 | |
a5d85af7 | 1975 | getinfo_super_imsm(st, &info, NULL); |
4737ae25 N |
1976 | fname_from_uuid(st, &info, nbuf, ':'); |
1977 | printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5); | |
1978 | } | |
1979 | ||
1980 | static void brief_examine_subarrays_imsm(struct supertype *st, int verbose) | |
1981 | { | |
1982 | /* We just write a generic IMSM ARRAY entry */ | |
1983 | struct mdinfo info; | |
1984 | char nbuf[64]; | |
1985 | char nbuf1[64]; | |
1986 | struct intel_super *super = st->sb; | |
1987 | int i; | |
1988 | ||
1989 | if (!super->anchor->num_raid_devs) | |
1990 | return; | |
1991 | ||
a5d85af7 | 1992 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1993 | fname_from_uuid(st, &info, nbuf, ':'); |
1e7bc0ed DW |
1994 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
1995 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
1996 | ||
1997 | super->current_vol = i; | |
a5d85af7 | 1998 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1999 | fname_from_uuid(st, &info, nbuf1, ':'); |
1124b3cf | 2000 | printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n", |
cf8de691 | 2001 | dev->volume, nbuf + 5, i, nbuf1 + 5); |
1e7bc0ed | 2002 | } |
cdddbdbc DW |
2003 | } |
2004 | ||
9d84c8ea DW |
2005 | static void export_examine_super_imsm(struct supertype *st) |
2006 | { | |
2007 | struct intel_super *super = st->sb; | |
2008 | struct imsm_super *mpb = super->anchor; | |
2009 | struct mdinfo info; | |
2010 | char nbuf[64]; | |
2011 | ||
a5d85af7 | 2012 | getinfo_super_imsm(st, &info, NULL); |
9d84c8ea DW |
2013 | fname_from_uuid(st, &info, nbuf, ':'); |
2014 | printf("MD_METADATA=imsm\n"); | |
2015 | printf("MD_LEVEL=container\n"); | |
2016 | printf("MD_UUID=%s\n", nbuf+5); | |
2017 | printf("MD_DEVICES=%u\n", mpb->num_disks); | |
2018 | } | |
2019 | ||
74db60b0 N |
2020 | static int copy_metadata_imsm(struct supertype *st, int from, int to) |
2021 | { | |
f36a9ecd | 2022 | /* The second last sector of the device contains |
74db60b0 N |
2023 | * the "struct imsm_super" metadata. |
2024 | * This contains mpb_size which is the size in bytes of the | |
2025 | * extended metadata. This is located immediately before | |
2026 | * the imsm_super. | |
2027 | * We want to read all that, plus the last sector which | |
2028 | * may contain a migration record, and write it all | |
2029 | * to the target. | |
2030 | */ | |
2031 | void *buf; | |
2032 | unsigned long long dsize, offset; | |
2033 | int sectors; | |
2034 | struct imsm_super *sb; | |
f36a9ecd PB |
2035 | struct intel_super *super = st->sb; |
2036 | unsigned int sector_size = super->sector_size; | |
2037 | unsigned int written = 0; | |
74db60b0 | 2038 | |
de44e46f | 2039 | if (posix_memalign(&buf, MAX_SECTOR_SIZE, MAX_SECTOR_SIZE) != 0) |
74db60b0 N |
2040 | return 1; |
2041 | ||
2042 | if (!get_dev_size(from, NULL, &dsize)) | |
2043 | goto err; | |
2044 | ||
f36a9ecd | 2045 | if (lseek64(from, dsize-(2*sector_size), 0) < 0) |
74db60b0 | 2046 | goto err; |
466070ad | 2047 | if ((unsigned int)read(from, buf, sector_size) != sector_size) |
74db60b0 N |
2048 | goto err; |
2049 | sb = buf; | |
2050 | if (strncmp((char*)sb->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) | |
2051 | goto err; | |
2052 | ||
f36a9ecd PB |
2053 | sectors = mpb_sectors(sb, sector_size) + 2; |
2054 | offset = dsize - sectors * sector_size; | |
74db60b0 N |
2055 | if (lseek64(from, offset, 0) < 0 || |
2056 | lseek64(to, offset, 0) < 0) | |
2057 | goto err; | |
f36a9ecd PB |
2058 | while (written < sectors * sector_size) { |
2059 | int n = sectors*sector_size - written; | |
74db60b0 N |
2060 | if (n > 4096) |
2061 | n = 4096; | |
2062 | if (read(from, buf, n) != n) | |
2063 | goto err; | |
2064 | if (write(to, buf, n) != n) | |
2065 | goto err; | |
2066 | written += n; | |
2067 | } | |
2068 | free(buf); | |
2069 | return 0; | |
2070 | err: | |
2071 | free(buf); | |
2072 | return 1; | |
2073 | } | |
2074 | ||
cdddbdbc DW |
2075 | static void detail_super_imsm(struct supertype *st, char *homehost) |
2076 | { | |
3ebe00a1 DW |
2077 | struct mdinfo info; |
2078 | char nbuf[64]; | |
2079 | ||
a5d85af7 | 2080 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 2081 | fname_from_uuid(st, &info, nbuf, ':'); |
65884368 | 2082 | printf("\n UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
2083 | } |
2084 | ||
2085 | static void brief_detail_super_imsm(struct supertype *st) | |
2086 | { | |
ff54de6e N |
2087 | struct mdinfo info; |
2088 | char nbuf[64]; | |
a5d85af7 | 2089 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 2090 | fname_from_uuid(st, &info, nbuf, ':'); |
ff54de6e | 2091 | printf(" UUID=%s", nbuf + 5); |
cdddbdbc | 2092 | } |
d665cc31 DW |
2093 | |
2094 | static int imsm_read_serial(int fd, char *devname, __u8 *serial); | |
2095 | static void fd2devname(int fd, char *name); | |
2096 | ||
120dc887 | 2097 | static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose) |
d665cc31 | 2098 | { |
120dc887 LM |
2099 | /* dump an unsorted list of devices attached to AHCI Intel storage |
2100 | * controller, as well as non-connected ports | |
d665cc31 DW |
2101 | */ |
2102 | int hba_len = strlen(hba_path) + 1; | |
2103 | struct dirent *ent; | |
2104 | DIR *dir; | |
2105 | char *path = NULL; | |
2106 | int err = 0; | |
2107 | unsigned long port_mask = (1 << port_count) - 1; | |
2108 | ||
f21e18ca | 2109 | if (port_count > (int)sizeof(port_mask) * 8) { |
ba728be7 | 2110 | if (verbose > 0) |
e7b84f9d | 2111 | pr_err("port_count %d out of range\n", port_count); |
d665cc31 DW |
2112 | return 2; |
2113 | } | |
2114 | ||
2115 | /* scroll through /sys/dev/block looking for devices attached to | |
2116 | * this hba | |
2117 | */ | |
2118 | dir = opendir("/sys/dev/block"); | |
1a6dd6b9 PB |
2119 | if (!dir) |
2120 | return 1; | |
2121 | ||
2122 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
d665cc31 DW |
2123 | int fd; |
2124 | char model[64]; | |
2125 | char vendor[64]; | |
2126 | char buf[1024]; | |
2127 | int major, minor; | |
2128 | char *device; | |
2129 | char *c; | |
2130 | int port; | |
2131 | int type; | |
2132 | ||
2133 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
2134 | continue; | |
2135 | path = devt_to_devpath(makedev(major, minor)); | |
2136 | if (!path) | |
2137 | continue; | |
2138 | if (!path_attached_to_hba(path, hba_path)) { | |
2139 | free(path); | |
2140 | path = NULL; | |
2141 | continue; | |
2142 | } | |
2143 | ||
2144 | /* retrieve the scsi device type */ | |
2145 | if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) { | |
ba728be7 | 2146 | if (verbose > 0) |
e7b84f9d | 2147 | pr_err("failed to allocate 'device'\n"); |
d665cc31 DW |
2148 | err = 2; |
2149 | break; | |
2150 | } | |
2151 | sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor); | |
193b6c0b | 2152 | if (load_sys(device, buf, sizeof(buf)) != 0) { |
ba728be7 | 2153 | if (verbose > 0) |
e7b84f9d | 2154 | pr_err("failed to read device type for %s\n", |
d665cc31 DW |
2155 | path); |
2156 | err = 2; | |
2157 | free(device); | |
2158 | break; | |
2159 | } | |
2160 | type = strtoul(buf, NULL, 10); | |
2161 | ||
2162 | /* if it's not a disk print the vendor and model */ | |
2163 | if (!(type == 0 || type == 7 || type == 14)) { | |
2164 | vendor[0] = '\0'; | |
2165 | model[0] = '\0'; | |
2166 | sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor); | |
193b6c0b | 2167 | if (load_sys(device, buf, sizeof(buf)) == 0) { |
d665cc31 DW |
2168 | strncpy(vendor, buf, sizeof(vendor)); |
2169 | vendor[sizeof(vendor) - 1] = '\0'; | |
2170 | c = (char *) &vendor[sizeof(vendor) - 1]; | |
2171 | while (isspace(*c) || *c == '\0') | |
2172 | *c-- = '\0'; | |
2173 | ||
2174 | } | |
2175 | sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor); | |
193b6c0b | 2176 | if (load_sys(device, buf, sizeof(buf)) == 0) { |
d665cc31 DW |
2177 | strncpy(model, buf, sizeof(model)); |
2178 | model[sizeof(model) - 1] = '\0'; | |
2179 | c = (char *) &model[sizeof(model) - 1]; | |
2180 | while (isspace(*c) || *c == '\0') | |
2181 | *c-- = '\0'; | |
2182 | } | |
2183 | ||
2184 | if (vendor[0] && model[0]) | |
2185 | sprintf(buf, "%.64s %.64s", vendor, model); | |
2186 | else | |
2187 | switch (type) { /* numbers from hald/linux/device.c */ | |
2188 | case 1: sprintf(buf, "tape"); break; | |
2189 | case 2: sprintf(buf, "printer"); break; | |
2190 | case 3: sprintf(buf, "processor"); break; | |
2191 | case 4: | |
2192 | case 5: sprintf(buf, "cdrom"); break; | |
2193 | case 6: sprintf(buf, "scanner"); break; | |
2194 | case 8: sprintf(buf, "media_changer"); break; | |
2195 | case 9: sprintf(buf, "comm"); break; | |
2196 | case 12: sprintf(buf, "raid"); break; | |
2197 | default: sprintf(buf, "unknown"); | |
2198 | } | |
2199 | } else | |
2200 | buf[0] = '\0'; | |
2201 | free(device); | |
2202 | ||
2203 | /* chop device path to 'host%d' and calculate the port number */ | |
2204 | c = strchr(&path[hba_len], '/'); | |
4e5e717d | 2205 | if (!c) { |
ba728be7 | 2206 | if (verbose > 0) |
e7b84f9d | 2207 | pr_err("%s - invalid path name\n", path + hba_len); |
4e5e717d AW |
2208 | err = 2; |
2209 | break; | |
2210 | } | |
d665cc31 | 2211 | *c = '\0'; |
0858eccf AP |
2212 | if ((sscanf(&path[hba_len], "ata%d", &port) == 1) || |
2213 | ((sscanf(&path[hba_len], "host%d", &port) == 1))) | |
d665cc31 DW |
2214 | port -= host_base; |
2215 | else { | |
ba728be7 | 2216 | if (verbose > 0) { |
d665cc31 | 2217 | *c = '/'; /* repair the full string */ |
e7b84f9d | 2218 | pr_err("failed to determine port number for %s\n", |
d665cc31 DW |
2219 | path); |
2220 | } | |
2221 | err = 2; | |
2222 | break; | |
2223 | } | |
2224 | ||
2225 | /* mark this port as used */ | |
2226 | port_mask &= ~(1 << port); | |
2227 | ||
2228 | /* print out the device information */ | |
2229 | if (buf[0]) { | |
2230 | printf(" Port%d : - non-disk device (%s) -\n", port, buf); | |
2231 | continue; | |
2232 | } | |
2233 | ||
2234 | fd = dev_open(ent->d_name, O_RDONLY); | |
2235 | if (fd < 0) | |
2236 | printf(" Port%d : - disk info unavailable -\n", port); | |
2237 | else { | |
2238 | fd2devname(fd, buf); | |
2239 | printf(" Port%d : %s", port, buf); | |
2240 | if (imsm_read_serial(fd, NULL, (__u8 *) buf) == 0) | |
664d5325 | 2241 | printf(" (%.*s)\n", MAX_RAID_SERIAL_LEN, buf); |
d665cc31 | 2242 | else |
664d5325 | 2243 | printf(" ()\n"); |
4dab422a | 2244 | close(fd); |
d665cc31 | 2245 | } |
d665cc31 DW |
2246 | free(path); |
2247 | path = NULL; | |
2248 | } | |
2249 | if (path) | |
2250 | free(path); | |
2251 | if (dir) | |
2252 | closedir(dir); | |
2253 | if (err == 0) { | |
2254 | int i; | |
2255 | ||
2256 | for (i = 0; i < port_count; i++) | |
2257 | if (port_mask & (1 << i)) | |
2258 | printf(" Port%d : - no device attached -\n", i); | |
2259 | } | |
2260 | ||
2261 | return err; | |
2262 | } | |
2263 | ||
b5eece69 | 2264 | static int print_vmd_attached_devs(struct sys_dev *hba) |
60f0f54d PB |
2265 | { |
2266 | struct dirent *ent; | |
2267 | DIR *dir; | |
2268 | char path[292]; | |
2269 | char link[256]; | |
2270 | char *c, *rp; | |
2271 | ||
2272 | if (hba->type != SYS_DEV_VMD) | |
b5eece69 | 2273 | return 1; |
60f0f54d PB |
2274 | |
2275 | /* scroll through /sys/dev/block looking for devices attached to | |
2276 | * this hba | |
2277 | */ | |
2278 | dir = opendir("/sys/bus/pci/drivers/nvme"); | |
b9135011 | 2279 | if (!dir) |
b5eece69 | 2280 | return 1; |
b9135011 JS |
2281 | |
2282 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
60f0f54d PB |
2283 | int n; |
2284 | ||
2285 | /* is 'ent' a device? check that the 'subsystem' link exists and | |
2286 | * that its target matches 'bus' | |
2287 | */ | |
2288 | sprintf(path, "/sys/bus/pci/drivers/nvme/%s/subsystem", | |
2289 | ent->d_name); | |
2290 | n = readlink(path, link, sizeof(link)); | |
2291 | if (n < 0 || n >= (int)sizeof(link)) | |
2292 | continue; | |
2293 | link[n] = '\0'; | |
2294 | c = strrchr(link, '/'); | |
2295 | if (!c) | |
2296 | continue; | |
2297 | if (strncmp("pci", c+1, strlen("pci")) != 0) | |
2298 | continue; | |
2299 | ||
2300 | sprintf(path, "/sys/bus/pci/drivers/nvme/%s", ent->d_name); | |
60f0f54d PB |
2301 | |
2302 | rp = realpath(path, NULL); | |
2303 | if (!rp) | |
2304 | continue; | |
2305 | ||
2306 | if (path_attached_to_hba(rp, hba->path)) { | |
2307 | printf(" NVMe under VMD : %s\n", rp); | |
2308 | } | |
2309 | free(rp); | |
2310 | } | |
2311 | ||
b9135011 | 2312 | closedir(dir); |
b5eece69 | 2313 | return 0; |
60f0f54d PB |
2314 | } |
2315 | ||
120dc887 LM |
2316 | static void print_found_intel_controllers(struct sys_dev *elem) |
2317 | { | |
2318 | for (; elem; elem = elem->next) { | |
e7b84f9d | 2319 | pr_err("found Intel(R) "); |
120dc887 LM |
2320 | if (elem->type == SYS_DEV_SATA) |
2321 | fprintf(stderr, "SATA "); | |
155cbb4c LM |
2322 | else if (elem->type == SYS_DEV_SAS) |
2323 | fprintf(stderr, "SAS "); | |
0858eccf AP |
2324 | else if (elem->type == SYS_DEV_NVME) |
2325 | fprintf(stderr, "NVMe "); | |
60f0f54d PB |
2326 | |
2327 | if (elem->type == SYS_DEV_VMD) | |
2328 | fprintf(stderr, "VMD domain"); | |
2329 | else | |
2330 | fprintf(stderr, "RAID controller"); | |
2331 | ||
120dc887 LM |
2332 | if (elem->pci_id) |
2333 | fprintf(stderr, " at %s", elem->pci_id); | |
2334 | fprintf(stderr, ".\n"); | |
2335 | } | |
2336 | fflush(stderr); | |
2337 | } | |
2338 | ||
120dc887 LM |
2339 | static int ahci_get_port_count(const char *hba_path, int *port_count) |
2340 | { | |
2341 | struct dirent *ent; | |
2342 | DIR *dir; | |
2343 | int host_base = -1; | |
2344 | ||
2345 | *port_count = 0; | |
2346 | if ((dir = opendir(hba_path)) == NULL) | |
2347 | return -1; | |
2348 | ||
2349 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
2350 | int host; | |
2351 | ||
0858eccf AP |
2352 | if ((sscanf(ent->d_name, "ata%d", &host) != 1) && |
2353 | ((sscanf(ent->d_name, "host%d", &host) != 1))) | |
120dc887 LM |
2354 | continue; |
2355 | if (*port_count == 0) | |
2356 | host_base = host; | |
2357 | else if (host < host_base) | |
2358 | host_base = host; | |
2359 | ||
2360 | if (host + 1 > *port_count + host_base) | |
2361 | *port_count = host + 1 - host_base; | |
2362 | } | |
2363 | closedir(dir); | |
2364 | return host_base; | |
2365 | } | |
2366 | ||
a891a3c2 LM |
2367 | static void print_imsm_capability(const struct imsm_orom *orom) |
2368 | { | |
0858eccf AP |
2369 | printf(" Platform : Intel(R) "); |
2370 | if (orom->capabilities == 0 && orom->driver_features == 0) | |
2371 | printf("Matrix Storage Manager\n"); | |
ab0c6bb9 AP |
2372 | else if (imsm_orom_is_enterprise(orom) && orom->major_ver >= 6) |
2373 | printf("Virtual RAID on CPU\n"); | |
0858eccf AP |
2374 | else |
2375 | printf("Rapid Storage Technology%s\n", | |
2376 | imsm_orom_is_enterprise(orom) ? " enterprise" : ""); | |
2377 | if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build) | |
2378 | printf(" Version : %d.%d.%d.%d\n", orom->major_ver, | |
2379 | orom->minor_ver, orom->hotfix_ver, orom->build); | |
a891a3c2 LM |
2380 | printf(" RAID Levels :%s%s%s%s%s\n", |
2381 | imsm_orom_has_raid0(orom) ? " raid0" : "", | |
2382 | imsm_orom_has_raid1(orom) ? " raid1" : "", | |
2383 | imsm_orom_has_raid1e(orom) ? " raid1e" : "", | |
2384 | imsm_orom_has_raid10(orom) ? " raid10" : "", | |
2385 | imsm_orom_has_raid5(orom) ? " raid5" : ""); | |
2386 | printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", | |
2387 | imsm_orom_has_chunk(orom, 2) ? " 2k" : "", | |
2388 | imsm_orom_has_chunk(orom, 4) ? " 4k" : "", | |
2389 | imsm_orom_has_chunk(orom, 8) ? " 8k" : "", | |
2390 | imsm_orom_has_chunk(orom, 16) ? " 16k" : "", | |
2391 | imsm_orom_has_chunk(orom, 32) ? " 32k" : "", | |
2392 | imsm_orom_has_chunk(orom, 64) ? " 64k" : "", | |
2393 | imsm_orom_has_chunk(orom, 128) ? " 128k" : "", | |
2394 | imsm_orom_has_chunk(orom, 256) ? " 256k" : "", | |
2395 | imsm_orom_has_chunk(orom, 512) ? " 512k" : "", | |
2396 | imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "", | |
2397 | imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "", | |
2398 | imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "", | |
2399 | imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "", | |
2400 | imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "", | |
2401 | imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "", | |
2402 | imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : ""); | |
29cd0821 CA |
2403 | printf(" 2TB volumes :%s supported\n", |
2404 | (orom->attr & IMSM_OROM_ATTR_2TB)?"":" not"); | |
2405 | printf(" 2TB disks :%s supported\n", | |
2406 | (orom->attr & IMSM_OROM_ATTR_2TB_DISK)?"":" not"); | |
0e7f69a8 | 2407 | printf(" Max Disks : %d\n", orom->tds); |
0858eccf AP |
2408 | printf(" Max Volumes : %d per array, %d per %s\n", |
2409 | orom->vpa, orom->vphba, | |
2410 | imsm_orom_is_nvme(orom) ? "platform" : "controller"); | |
a891a3c2 LM |
2411 | return; |
2412 | } | |
2413 | ||
e50cf220 MN |
2414 | static void print_imsm_capability_export(const struct imsm_orom *orom) |
2415 | { | |
2416 | printf("MD_FIRMWARE_TYPE=imsm\n"); | |
0858eccf AP |
2417 | if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build) |
2418 | printf("IMSM_VERSION=%d.%d.%d.%d\n", orom->major_ver, orom->minor_ver, | |
2419 | orom->hotfix_ver, orom->build); | |
e50cf220 MN |
2420 | printf("IMSM_SUPPORTED_RAID_LEVELS=%s%s%s%s%s\n", |
2421 | imsm_orom_has_raid0(orom) ? "raid0 " : "", | |
2422 | imsm_orom_has_raid1(orom) ? "raid1 " : "", | |
2423 | imsm_orom_has_raid1e(orom) ? "raid1e " : "", | |
2424 | imsm_orom_has_raid5(orom) ? "raid10 " : "", | |
2425 | imsm_orom_has_raid10(orom) ? "raid5 " : ""); | |
2426 | printf("IMSM_SUPPORTED_CHUNK_SIZES=%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", | |
2427 | imsm_orom_has_chunk(orom, 2) ? "2k " : "", | |
2428 | imsm_orom_has_chunk(orom, 4) ? "4k " : "", | |
2429 | imsm_orom_has_chunk(orom, 8) ? "8k " : "", | |
2430 | imsm_orom_has_chunk(orom, 16) ? "16k " : "", | |
2431 | imsm_orom_has_chunk(orom, 32) ? "32k " : "", | |
2432 | imsm_orom_has_chunk(orom, 64) ? "64k " : "", | |
2433 | imsm_orom_has_chunk(orom, 128) ? "128k " : "", | |
2434 | imsm_orom_has_chunk(orom, 256) ? "256k " : "", | |
2435 | imsm_orom_has_chunk(orom, 512) ? "512k " : "", | |
2436 | imsm_orom_has_chunk(orom, 1024*1) ? "1M " : "", | |
2437 | imsm_orom_has_chunk(orom, 1024*2) ? "2M " : "", | |
2438 | imsm_orom_has_chunk(orom, 1024*4) ? "4M " : "", | |
2439 | imsm_orom_has_chunk(orom, 1024*8) ? "8M " : "", | |
2440 | imsm_orom_has_chunk(orom, 1024*16) ? "16M " : "", | |
2441 | imsm_orom_has_chunk(orom, 1024*32) ? "32M " : "", | |
2442 | imsm_orom_has_chunk(orom, 1024*64) ? "64M " : ""); | |
2443 | printf("IMSM_2TB_VOLUMES=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB) ? "yes" : "no"); | |
2444 | printf("IMSM_2TB_DISKS=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB_DISK) ? "yes" : "no"); | |
2445 | printf("IMSM_MAX_DISKS=%d\n",orom->tds); | |
2446 | printf("IMSM_MAX_VOLUMES_PER_ARRAY=%d\n",orom->vpa); | |
2447 | printf("IMSM_MAX_VOLUMES_PER_CONTROLLER=%d\n",orom->vphba); | |
2448 | } | |
2449 | ||
9eafa1de | 2450 | static int detail_platform_imsm(int verbose, int enumerate_only, char *controller_path) |
d665cc31 DW |
2451 | { |
2452 | /* There are two components to imsm platform support, the ahci SATA | |
2453 | * controller and the option-rom. To find the SATA controller we | |
2454 | * simply look in /sys/bus/pci/drivers/ahci to see if an ahci | |
2455 | * controller with the Intel vendor id is present. This approach | |
2456 | * allows mdadm to leverage the kernel's ahci detection logic, with the | |
2457 | * caveat that if ahci.ko is not loaded mdadm will not be able to | |
2458 | * detect platform raid capabilities. The option-rom resides in a | |
2459 | * platform "Adapter ROM". We scan for its signature to retrieve the | |
2460 | * platform capabilities. If raid support is disabled in the BIOS the | |
2461 | * option-rom capability structure will not be available. | |
2462 | */ | |
d665cc31 | 2463 | struct sys_dev *list, *hba; |
d665cc31 DW |
2464 | int host_base = 0; |
2465 | int port_count = 0; | |
9eafa1de | 2466 | int result=1; |
d665cc31 | 2467 | |
5615172f | 2468 | if (enumerate_only) { |
a891a3c2 | 2469 | if (check_env("IMSM_NO_PLATFORM")) |
5615172f | 2470 | return 0; |
a891a3c2 LM |
2471 | list = find_intel_devices(); |
2472 | if (!list) | |
2473 | return 2; | |
2474 | for (hba = list; hba; hba = hba->next) { | |
6b781d33 AP |
2475 | if (find_imsm_capability(hba)) { |
2476 | result = 0; | |
a891a3c2 LM |
2477 | break; |
2478 | } | |
9eafa1de | 2479 | else |
6b781d33 | 2480 | result = 2; |
a891a3c2 | 2481 | } |
a891a3c2 | 2482 | return result; |
5615172f DW |
2483 | } |
2484 | ||
155cbb4c LM |
2485 | list = find_intel_devices(); |
2486 | if (!list) { | |
ba728be7 | 2487 | if (verbose > 0) |
7a862a02 | 2488 | pr_err("no active Intel(R) RAID controller found.\n"); |
d665cc31 | 2489 | return 2; |
ba728be7 | 2490 | } else if (verbose > 0) |
155cbb4c | 2491 | print_found_intel_controllers(list); |
d665cc31 | 2492 | |
a891a3c2 | 2493 | for (hba = list; hba; hba = hba->next) { |
0858eccf | 2494 | if (controller_path && (compare_paths(hba->path, controller_path) != 0)) |
9eafa1de | 2495 | continue; |
0858eccf | 2496 | if (!find_imsm_capability(hba)) { |
60f0f54d | 2497 | char buf[PATH_MAX]; |
e7b84f9d | 2498 | pr_err("imsm capabilities not found for controller: %s (type %s)\n", |
60f0f54d PB |
2499 | hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path, |
2500 | get_sys_dev_type(hba->type)); | |
0858eccf AP |
2501 | continue; |
2502 | } | |
2503 | result = 0; | |
2504 | } | |
2505 | ||
2506 | if (controller_path && result == 1) { | |
2507 | pr_err("no active Intel(R) RAID controller found under %s\n", | |
2508 | controller_path); | |
2509 | return result; | |
2510 | } | |
2511 | ||
5e1d6128 | 2512 | const struct orom_entry *entry; |
0858eccf | 2513 | |
5e1d6128 | 2514 | for (entry = orom_entries; entry; entry = entry->next) { |
60f0f54d | 2515 | if (entry->type == SYS_DEV_VMD) { |
07cb1e57 | 2516 | print_imsm_capability(&entry->orom); |
32716c51 PB |
2517 | printf(" 3rd party NVMe :%s supported\n", |
2518 | imsm_orom_has_tpv_support(&entry->orom)?"":" not"); | |
60f0f54d PB |
2519 | for (hba = list; hba; hba = hba->next) { |
2520 | if (hba->type == SYS_DEV_VMD) { | |
2521 | char buf[PATH_MAX]; | |
60f0f54d PB |
2522 | printf(" I/O Controller : %s (%s)\n", |
2523 | vmd_domain_to_controller(hba, buf), get_sys_dev_type(hba->type)); | |
b5eece69 PB |
2524 | if (print_vmd_attached_devs(hba)) { |
2525 | if (verbose > 0) | |
2526 | pr_err("failed to get devices attached to VMD domain.\n"); | |
2527 | result |= 2; | |
2528 | } | |
60f0f54d PB |
2529 | } |
2530 | } | |
07cb1e57 | 2531 | printf("\n"); |
60f0f54d PB |
2532 | continue; |
2533 | } | |
0858eccf | 2534 | |
60f0f54d PB |
2535 | print_imsm_capability(&entry->orom); |
2536 | if (entry->type == SYS_DEV_NVME) { | |
0858eccf AP |
2537 | for (hba = list; hba; hba = hba->next) { |
2538 | if (hba->type == SYS_DEV_NVME) | |
2539 | printf(" NVMe Device : %s\n", hba->path); | |
2540 | } | |
60f0f54d | 2541 | printf("\n"); |
0858eccf AP |
2542 | continue; |
2543 | } | |
2544 | ||
2545 | struct devid_list *devid; | |
5e1d6128 | 2546 | for (devid = entry->devid_list; devid; devid = devid->next) { |
0858eccf AP |
2547 | hba = device_by_id(devid->devid); |
2548 | if (!hba) | |
2549 | continue; | |
2550 | ||
9eafa1de MN |
2551 | printf(" I/O Controller : %s (%s)\n", |
2552 | hba->path, get_sys_dev_type(hba->type)); | |
2553 | if (hba->type == SYS_DEV_SATA) { | |
2554 | host_base = ahci_get_port_count(hba->path, &port_count); | |
2555 | if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) { | |
2556 | if (verbose > 0) | |
7a862a02 | 2557 | pr_err("failed to enumerate ports on SATA controller at %s.\n", hba->pci_id); |
9eafa1de MN |
2558 | result |= 2; |
2559 | } | |
120dc887 LM |
2560 | } |
2561 | } | |
0858eccf | 2562 | printf("\n"); |
d665cc31 | 2563 | } |
155cbb4c | 2564 | |
120dc887 | 2565 | return result; |
d665cc31 | 2566 | } |
e50cf220 | 2567 | |
9eafa1de | 2568 | static int export_detail_platform_imsm(int verbose, char *controller_path) |
e50cf220 | 2569 | { |
e50cf220 MN |
2570 | struct sys_dev *list, *hba; |
2571 | int result=1; | |
2572 | ||
2573 | list = find_intel_devices(); | |
2574 | if (!list) { | |
2575 | if (verbose > 0) | |
2576 | pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_INTEL_DEVICES\n"); | |
2577 | result = 2; | |
e50cf220 MN |
2578 | return result; |
2579 | } | |
2580 | ||
2581 | for (hba = list; hba; hba = hba->next) { | |
9eafa1de MN |
2582 | if (controller_path && (compare_paths(hba->path,controller_path) != 0)) |
2583 | continue; | |
60f0f54d PB |
2584 | if (!find_imsm_capability(hba) && verbose > 0) { |
2585 | char buf[PATH_MAX]; | |
2586 | pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_IMSM_CAPABLE_DEVICE_UNDER_%s\n", | |
2587 | hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path); | |
2588 | } | |
0858eccf | 2589 | else |
e50cf220 | 2590 | result = 0; |
e50cf220 MN |
2591 | } |
2592 | ||
5e1d6128 | 2593 | const struct orom_entry *entry; |
0858eccf | 2594 | |
60f0f54d PB |
2595 | for (entry = orom_entries; entry; entry = entry->next) { |
2596 | if (entry->type == SYS_DEV_VMD) { | |
2597 | for (hba = list; hba; hba = hba->next) | |
2598 | print_imsm_capability_export(&entry->orom); | |
2599 | continue; | |
2600 | } | |
5e1d6128 | 2601 | print_imsm_capability_export(&entry->orom); |
60f0f54d | 2602 | } |
0858eccf | 2603 | |
e50cf220 MN |
2604 | return result; |
2605 | } | |
2606 | ||
cdddbdbc DW |
2607 | static int match_home_imsm(struct supertype *st, char *homehost) |
2608 | { | |
5115ca67 DW |
2609 | /* the imsm metadata format does not specify any host |
2610 | * identification information. We return -1 since we can never | |
2611 | * confirm nor deny whether a given array is "meant" for this | |
148acb7b | 2612 | * host. We rely on compare_super and the 'family_num' fields to |
5115ca67 DW |
2613 | * exclude member disks that do not belong, and we rely on |
2614 | * mdadm.conf to specify the arrays that should be assembled. | |
2615 | * Auto-assembly may still pick up "foreign" arrays. | |
2616 | */ | |
cdddbdbc | 2617 | |
9362c1c8 | 2618 | return -1; |
cdddbdbc DW |
2619 | } |
2620 | ||
2621 | static void uuid_from_super_imsm(struct supertype *st, int uuid[4]) | |
2622 | { | |
51006d85 N |
2623 | /* The uuid returned here is used for: |
2624 | * uuid to put into bitmap file (Create, Grow) | |
2625 | * uuid for backup header when saving critical section (Grow) | |
2626 | * comparing uuids when re-adding a device into an array | |
2627 | * In these cases the uuid required is that of the data-array, | |
2628 | * not the device-set. | |
2629 | * uuid to recognise same set when adding a missing device back | |
2630 | * to an array. This is a uuid for the device-set. | |
1011e834 | 2631 | * |
51006d85 N |
2632 | * For each of these we can make do with a truncated |
2633 | * or hashed uuid rather than the original, as long as | |
2634 | * everyone agrees. | |
2635 | * In each case the uuid required is that of the data-array, | |
2636 | * not the device-set. | |
43dad3d6 | 2637 | */ |
51006d85 N |
2638 | /* imsm does not track uuid's so we synthesis one using sha1 on |
2639 | * - The signature (Which is constant for all imsm array, but no matter) | |
148acb7b | 2640 | * - the orig_family_num of the container |
51006d85 N |
2641 | * - the index number of the volume |
2642 | * - the 'serial' number of the volume. | |
2643 | * Hopefully these are all constant. | |
2644 | */ | |
2645 | struct intel_super *super = st->sb; | |
43dad3d6 | 2646 | |
51006d85 N |
2647 | char buf[20]; |
2648 | struct sha1_ctx ctx; | |
2649 | struct imsm_dev *dev = NULL; | |
148acb7b | 2650 | __u32 family_num; |
51006d85 | 2651 | |
148acb7b DW |
2652 | /* some mdadm versions failed to set ->orig_family_num, in which |
2653 | * case fall back to ->family_num. orig_family_num will be | |
2654 | * fixed up with the first metadata update. | |
2655 | */ | |
2656 | family_num = super->anchor->orig_family_num; | |
2657 | if (family_num == 0) | |
2658 | family_num = super->anchor->family_num; | |
51006d85 | 2659 | sha1_init_ctx(&ctx); |
92bd8f8d | 2660 | sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx); |
148acb7b | 2661 | sha1_process_bytes(&family_num, sizeof(__u32), &ctx); |
51006d85 N |
2662 | if (super->current_vol >= 0) |
2663 | dev = get_imsm_dev(super, super->current_vol); | |
2664 | if (dev) { | |
2665 | __u32 vol = super->current_vol; | |
2666 | sha1_process_bytes(&vol, sizeof(vol), &ctx); | |
2667 | sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx); | |
2668 | } | |
2669 | sha1_finish_ctx(&ctx, buf); | |
2670 | memcpy(uuid, buf, 4*4); | |
cdddbdbc DW |
2671 | } |
2672 | ||
0d481d37 | 2673 | #if 0 |
4f5bc454 DW |
2674 | static void |
2675 | get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p) | |
cdddbdbc | 2676 | { |
cdddbdbc DW |
2677 | __u8 *v = get_imsm_version(mpb); |
2678 | __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH; | |
2679 | char major[] = { 0, 0, 0 }; | |
2680 | char minor[] = { 0 ,0, 0 }; | |
2681 | char patch[] = { 0, 0, 0 }; | |
2682 | char *ver_parse[] = { major, minor, patch }; | |
2683 | int i, j; | |
2684 | ||
2685 | i = j = 0; | |
2686 | while (*v != '\0' && v < end) { | |
2687 | if (*v != '.' && j < 2) | |
2688 | ver_parse[i][j++] = *v; | |
2689 | else { | |
2690 | i++; | |
2691 | j = 0; | |
2692 | } | |
2693 | v++; | |
2694 | } | |
2695 | ||
4f5bc454 DW |
2696 | *m = strtol(minor, NULL, 0); |
2697 | *p = strtol(patch, NULL, 0); | |
2698 | } | |
0d481d37 | 2699 | #endif |
4f5bc454 | 2700 | |
1e5c6983 DW |
2701 | static __u32 migr_strip_blocks_resync(struct imsm_dev *dev) |
2702 | { | |
2703 | /* migr_strip_size when repairing or initializing parity */ | |
238c0a71 | 2704 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2705 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2706 | ||
2707 | switch (get_imsm_raid_level(map)) { | |
2708 | case 5: | |
2709 | case 10: | |
2710 | return chunk; | |
2711 | default: | |
2712 | return 128*1024 >> 9; | |
2713 | } | |
2714 | } | |
2715 | ||
2716 | static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev) | |
2717 | { | |
2718 | /* migr_strip_size when rebuilding a degraded disk, no idea why | |
2719 | * this is different than migr_strip_size_resync(), but it's good | |
2720 | * to be compatible | |
2721 | */ | |
238c0a71 | 2722 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2723 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2724 | ||
2725 | switch (get_imsm_raid_level(map)) { | |
2726 | case 1: | |
2727 | case 10: | |
2728 | if (map->num_members % map->num_domains == 0) | |
2729 | return 128*1024 >> 9; | |
2730 | else | |
2731 | return chunk; | |
2732 | case 5: | |
2733 | return max((__u32) 64*1024 >> 9, chunk); | |
2734 | default: | |
2735 | return 128*1024 >> 9; | |
2736 | } | |
2737 | } | |
2738 | ||
2739 | static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev) | |
2740 | { | |
238c0a71 AK |
2741 | struct imsm_map *lo = get_imsm_map(dev, MAP_0); |
2742 | struct imsm_map *hi = get_imsm_map(dev, MAP_1); | |
1e5c6983 DW |
2743 | __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip); |
2744 | __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip); | |
2745 | ||
2746 | return max((__u32) 1, hi_chunk / lo_chunk); | |
2747 | } | |
2748 | ||
2749 | static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev) | |
2750 | { | |
238c0a71 | 2751 | struct imsm_map *lo = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2752 | int level = get_imsm_raid_level(lo); |
2753 | ||
2754 | if (level == 1 || level == 10) { | |
238c0a71 | 2755 | struct imsm_map *hi = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2756 | |
2757 | return hi->num_domains; | |
2758 | } else | |
2759 | return num_stripes_per_unit_resync(dev); | |
2760 | } | |
2761 | ||
98130f40 | 2762 | static __u8 imsm_num_data_members(struct imsm_dev *dev, int second_map) |
1e5c6983 DW |
2763 | { |
2764 | /* named 'imsm_' because raid0, raid1 and raid10 | |
2765 | * counter-intuitively have the same number of data disks | |
2766 | */ | |
98130f40 | 2767 | struct imsm_map *map = get_imsm_map(dev, second_map); |
1e5c6983 DW |
2768 | |
2769 | switch (get_imsm_raid_level(map)) { | |
2770 | case 0: | |
36fd8ccc AK |
2771 | return map->num_members; |
2772 | break; | |
1e5c6983 DW |
2773 | case 1: |
2774 | case 10: | |
36fd8ccc | 2775 | return map->num_members/2; |
1e5c6983 DW |
2776 | case 5: |
2777 | return map->num_members - 1; | |
2778 | default: | |
1ade5cc1 | 2779 | dprintf("unsupported raid level\n"); |
1e5c6983 DW |
2780 | return 0; |
2781 | } | |
2782 | } | |
2783 | ||
2784 | static __u32 parity_segment_depth(struct imsm_dev *dev) | |
2785 | { | |
238c0a71 | 2786 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2787 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2788 | ||
2789 | switch(get_imsm_raid_level(map)) { | |
2790 | case 1: | |
2791 | case 10: | |
2792 | return chunk * map->num_domains; | |
2793 | case 5: | |
2794 | return chunk * map->num_members; | |
2795 | default: | |
2796 | return chunk; | |
2797 | } | |
2798 | } | |
2799 | ||
2800 | static __u32 map_migr_block(struct imsm_dev *dev, __u32 block) | |
2801 | { | |
238c0a71 | 2802 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2803 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2804 | __u32 strip = block / chunk; | |
2805 | ||
2806 | switch (get_imsm_raid_level(map)) { | |
2807 | case 1: | |
2808 | case 10: { | |
2809 | __u32 vol_strip = (strip * map->num_domains) + 1; | |
2810 | __u32 vol_stripe = vol_strip / map->num_members; | |
2811 | ||
2812 | return vol_stripe * chunk + block % chunk; | |
2813 | } case 5: { | |
2814 | __u32 stripe = strip / (map->num_members - 1); | |
2815 | ||
2816 | return stripe * chunk + block % chunk; | |
2817 | } | |
2818 | default: | |
2819 | return 0; | |
2820 | } | |
2821 | } | |
2822 | ||
c47b0ff6 AK |
2823 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
2824 | struct imsm_dev *dev) | |
1e5c6983 DW |
2825 | { |
2826 | /* calculate the conversion factor between per member 'blocks' | |
2827 | * (md/{resync,rebuild}_start) and imsm migration units, return | |
2828 | * 0 for the 'not migrating' and 'unsupported migration' cases | |
2829 | */ | |
2830 | if (!dev->vol.migr_state) | |
2831 | return 0; | |
2832 | ||
2833 | switch (migr_type(dev)) { | |
c47b0ff6 AK |
2834 | case MIGR_GEN_MIGR: { |
2835 | struct migr_record *migr_rec = super->migr_rec; | |
2836 | return __le32_to_cpu(migr_rec->blocks_per_unit); | |
2837 | } | |
1e5c6983 DW |
2838 | case MIGR_VERIFY: |
2839 | case MIGR_REPAIR: | |
2840 | case MIGR_INIT: { | |
238c0a71 | 2841 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2842 | __u32 stripes_per_unit; |
2843 | __u32 blocks_per_unit; | |
2844 | __u32 parity_depth; | |
2845 | __u32 migr_chunk; | |
2846 | __u32 block_map; | |
2847 | __u32 block_rel; | |
2848 | __u32 segment; | |
2849 | __u32 stripe; | |
2850 | __u8 disks; | |
2851 | ||
2852 | /* yes, this is really the translation of migr_units to | |
2853 | * per-member blocks in the 'resync' case | |
2854 | */ | |
2855 | stripes_per_unit = num_stripes_per_unit_resync(dev); | |
2856 | migr_chunk = migr_strip_blocks_resync(dev); | |
238c0a71 | 2857 | disks = imsm_num_data_members(dev, MAP_0); |
1e5c6983 | 2858 | blocks_per_unit = stripes_per_unit * migr_chunk * disks; |
7b1ab482 | 2859 | stripe = __le16_to_cpu(map->blocks_per_strip) * disks; |
1e5c6983 DW |
2860 | segment = blocks_per_unit / stripe; |
2861 | block_rel = blocks_per_unit - segment * stripe; | |
2862 | parity_depth = parity_segment_depth(dev); | |
2863 | block_map = map_migr_block(dev, block_rel); | |
2864 | return block_map + parity_depth * segment; | |
2865 | } | |
2866 | case MIGR_REBUILD: { | |
2867 | __u32 stripes_per_unit; | |
2868 | __u32 migr_chunk; | |
2869 | ||
2870 | stripes_per_unit = num_stripes_per_unit_rebuild(dev); | |
2871 | migr_chunk = migr_strip_blocks_rebuild(dev); | |
2872 | return migr_chunk * stripes_per_unit; | |
2873 | } | |
1e5c6983 DW |
2874 | case MIGR_STATE_CHANGE: |
2875 | default: | |
2876 | return 0; | |
2877 | } | |
2878 | } | |
2879 | ||
c2c087e6 DW |
2880 | static int imsm_level_to_layout(int level) |
2881 | { | |
2882 | switch (level) { | |
2883 | case 0: | |
2884 | case 1: | |
2885 | return 0; | |
2886 | case 5: | |
2887 | case 6: | |
a380c027 | 2888 | return ALGORITHM_LEFT_ASYMMETRIC; |
c2c087e6 | 2889 | case 10: |
c92a2527 | 2890 | return 0x102; |
c2c087e6 | 2891 | } |
a18a888e | 2892 | return UnSet; |
c2c087e6 DW |
2893 | } |
2894 | ||
8e59f3d8 AK |
2895 | /******************************************************************************* |
2896 | * Function: read_imsm_migr_rec | |
2897 | * Description: Function reads imsm migration record from last sector of disk | |
2898 | * Parameters: | |
2899 | * fd : disk descriptor | |
2900 | * super : metadata info | |
2901 | * Returns: | |
2902 | * 0 : success, | |
2903 | * -1 : fail | |
2904 | ******************************************************************************/ | |
2905 | static int read_imsm_migr_rec(int fd, struct intel_super *super) | |
2906 | { | |
2907 | int ret_val = -1; | |
de44e46f | 2908 | unsigned int sector_size = super->sector_size; |
8e59f3d8 AK |
2909 | unsigned long long dsize; |
2910 | ||
2911 | get_dev_size(fd, NULL, &dsize); | |
de44e46f PB |
2912 | if (lseek64(fd, dsize - (sector_size*MIGR_REC_SECTOR_POSITION), |
2913 | SEEK_SET) < 0) { | |
e7b84f9d N |
2914 | pr_err("Cannot seek to anchor block: %s\n", |
2915 | strerror(errno)); | |
8e59f3d8 AK |
2916 | goto out; |
2917 | } | |
466070ad | 2918 | if ((unsigned int)read(fd, super->migr_rec_buf, |
de44e46f PB |
2919 | MIGR_REC_BUF_SECTORS*sector_size) != |
2920 | MIGR_REC_BUF_SECTORS*sector_size) { | |
e7b84f9d N |
2921 | pr_err("Cannot read migr record block: %s\n", |
2922 | strerror(errno)); | |
8e59f3d8 AK |
2923 | goto out; |
2924 | } | |
2925 | ret_val = 0; | |
de44e46f PB |
2926 | if (sector_size == 4096) |
2927 | convert_from_4k_imsm_migr_rec(super); | |
8e59f3d8 AK |
2928 | |
2929 | out: | |
2930 | return ret_val; | |
2931 | } | |
2932 | ||
3136abe5 AK |
2933 | static struct imsm_dev *imsm_get_device_during_migration( |
2934 | struct intel_super *super) | |
2935 | { | |
2936 | ||
2937 | struct intel_dev *dv; | |
2938 | ||
2939 | for (dv = super->devlist; dv; dv = dv->next) { | |
2940 | if (is_gen_migration(dv->dev)) | |
2941 | return dv->dev; | |
2942 | } | |
2943 | return NULL; | |
2944 | } | |
2945 | ||
8e59f3d8 AK |
2946 | /******************************************************************************* |
2947 | * Function: load_imsm_migr_rec | |
2948 | * Description: Function reads imsm migration record (it is stored at the last | |
2949 | * sector of disk) | |
2950 | * Parameters: | |
2951 | * super : imsm internal array info | |
2952 | * info : general array info | |
2953 | * Returns: | |
2954 | * 0 : success | |
2955 | * -1 : fail | |
4c965cc9 | 2956 | * -2 : no migration in progress |
8e59f3d8 AK |
2957 | ******************************************************************************/ |
2958 | static int load_imsm_migr_rec(struct intel_super *super, struct mdinfo *info) | |
2959 | { | |
2960 | struct mdinfo *sd; | |
594dc1b8 | 2961 | struct dl *dl; |
8e59f3d8 AK |
2962 | char nm[30]; |
2963 | int retval = -1; | |
2964 | int fd = -1; | |
3136abe5 | 2965 | struct imsm_dev *dev; |
594dc1b8 | 2966 | struct imsm_map *map; |
b4ab44d8 | 2967 | int slot = -1; |
3136abe5 AK |
2968 | |
2969 | /* find map under migration */ | |
2970 | dev = imsm_get_device_during_migration(super); | |
2971 | /* nothing to load,no migration in progress? | |
2972 | */ | |
2973 | if (dev == NULL) | |
4c965cc9 | 2974 | return -2; |
8e59f3d8 AK |
2975 | |
2976 | if (info) { | |
2977 | for (sd = info->devs ; sd ; sd = sd->next) { | |
2978 | /* read only from one of the first two slots */ | |
12fe93e9 TM |
2979 | if ((sd->disk.raid_disk < 0) || |
2980 | (sd->disk.raid_disk > 1)) | |
8e59f3d8 | 2981 | continue; |
3136abe5 | 2982 | |
8e59f3d8 AK |
2983 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); |
2984 | fd = dev_open(nm, O_RDONLY); | |
2985 | if (fd >= 0) | |
2986 | break; | |
2987 | } | |
2988 | } | |
2989 | if (fd < 0) { | |
12fe93e9 | 2990 | map = get_imsm_map(dev, MAP_0); |
8e59f3d8 | 2991 | for (dl = super->disks; dl; dl = dl->next) { |
3136abe5 AK |
2992 | /* skip spare and failed disks |
2993 | */ | |
2994 | if (dl->index < 0) | |
2995 | continue; | |
8e59f3d8 | 2996 | /* read only from one of the first two slots */ |
3136abe5 AK |
2997 | if (map) |
2998 | slot = get_imsm_disk_slot(map, dl->index); | |
089f9d79 | 2999 | if (map == NULL || slot > 1 || slot < 0) |
8e59f3d8 AK |
3000 | continue; |
3001 | sprintf(nm, "%d:%d", dl->major, dl->minor); | |
3002 | fd = dev_open(nm, O_RDONLY); | |
3003 | if (fd >= 0) | |
3004 | break; | |
3005 | } | |
3006 | } | |
3007 | if (fd < 0) | |
3008 | goto out; | |
3009 | retval = read_imsm_migr_rec(fd, super); | |
3010 | ||
3011 | out: | |
3012 | if (fd >= 0) | |
3013 | close(fd); | |
3014 | return retval; | |
3015 | } | |
3016 | ||
c17608ea AK |
3017 | /******************************************************************************* |
3018 | * function: imsm_create_metadata_checkpoint_update | |
3019 | * Description: It creates update for checkpoint change. | |
3020 | * Parameters: | |
3021 | * super : imsm internal array info | |
3022 | * u : pointer to prepared update | |
3023 | * Returns: | |
3024 | * Uptate length. | |
3025 | * If length is equal to 0, input pointer u contains no update | |
3026 | ******************************************************************************/ | |
3027 | static int imsm_create_metadata_checkpoint_update( | |
3028 | struct intel_super *super, | |
3029 | struct imsm_update_general_migration_checkpoint **u) | |
3030 | { | |
3031 | ||
3032 | int update_memory_size = 0; | |
3033 | ||
1ade5cc1 | 3034 | dprintf("(enter)\n"); |
c17608ea AK |
3035 | |
3036 | if (u == NULL) | |
3037 | return 0; | |
3038 | *u = NULL; | |
3039 | ||
3040 | /* size of all update data without anchor */ | |
3041 | update_memory_size = | |
3042 | sizeof(struct imsm_update_general_migration_checkpoint); | |
3043 | ||
503975b9 | 3044 | *u = xcalloc(1, update_memory_size); |
c17608ea | 3045 | if (*u == NULL) { |
1ade5cc1 | 3046 | dprintf("error: cannot get memory\n"); |
c17608ea AK |
3047 | return 0; |
3048 | } | |
3049 | (*u)->type = update_general_migration_checkpoint; | |
3050 | (*u)->curr_migr_unit = __le32_to_cpu(super->migr_rec->curr_migr_unit); | |
1ade5cc1 | 3051 | dprintf("prepared for %u\n", (*u)->curr_migr_unit); |
c17608ea AK |
3052 | |
3053 | return update_memory_size; | |
3054 | } | |
3055 | ||
c17608ea AK |
3056 | static void imsm_update_metadata_locally(struct supertype *st, |
3057 | void *buf, int len); | |
3058 | ||
687629c2 AK |
3059 | /******************************************************************************* |
3060 | * Function: write_imsm_migr_rec | |
3061 | * Description: Function writes imsm migration record | |
3062 | * (at the last sector of disk) | |
3063 | * Parameters: | |
3064 | * super : imsm internal array info | |
3065 | * Returns: | |
3066 | * 0 : success | |
3067 | * -1 : if fail | |
3068 | ******************************************************************************/ | |
3069 | static int write_imsm_migr_rec(struct supertype *st) | |
3070 | { | |
3071 | struct intel_super *super = st->sb; | |
de44e46f | 3072 | unsigned int sector_size = super->sector_size; |
687629c2 AK |
3073 | unsigned long long dsize; |
3074 | char nm[30]; | |
3075 | int fd = -1; | |
3076 | int retval = -1; | |
3077 | struct dl *sd; | |
c17608ea AK |
3078 | int len; |
3079 | struct imsm_update_general_migration_checkpoint *u; | |
3136abe5 | 3080 | struct imsm_dev *dev; |
594dc1b8 | 3081 | struct imsm_map *map; |
3136abe5 AK |
3082 | |
3083 | /* find map under migration */ | |
3084 | dev = imsm_get_device_during_migration(super); | |
3085 | /* if no migration, write buffer anyway to clear migr_record | |
3086 | * on disk based on first available device | |
3087 | */ | |
3088 | if (dev == NULL) | |
3089 | dev = get_imsm_dev(super, super->current_vol < 0 ? 0 : | |
3090 | super->current_vol); | |
3091 | ||
44bfe6df | 3092 | map = get_imsm_map(dev, MAP_0); |
687629c2 | 3093 | |
de44e46f PB |
3094 | if (sector_size == 4096) |
3095 | convert_to_4k_imsm_migr_rec(super); | |
687629c2 | 3096 | for (sd = super->disks ; sd ; sd = sd->next) { |
b4ab44d8 | 3097 | int slot = -1; |
3136abe5 AK |
3098 | |
3099 | /* skip failed and spare devices */ | |
3100 | if (sd->index < 0) | |
3101 | continue; | |
687629c2 | 3102 | /* write to 2 first slots only */ |
3136abe5 AK |
3103 | if (map) |
3104 | slot = get_imsm_disk_slot(map, sd->index); | |
089f9d79 | 3105 | if (map == NULL || slot > 1 || slot < 0) |
687629c2 | 3106 | continue; |
3136abe5 | 3107 | |
687629c2 AK |
3108 | sprintf(nm, "%d:%d", sd->major, sd->minor); |
3109 | fd = dev_open(nm, O_RDWR); | |
3110 | if (fd < 0) | |
3111 | continue; | |
3112 | get_dev_size(fd, NULL, &dsize); | |
de44e46f PB |
3113 | if (lseek64(fd, dsize - (MIGR_REC_SECTOR_POSITION*sector_size), |
3114 | SEEK_SET) < 0) { | |
e7b84f9d N |
3115 | pr_err("Cannot seek to anchor block: %s\n", |
3116 | strerror(errno)); | |
687629c2 AK |
3117 | goto out; |
3118 | } | |
466070ad | 3119 | if ((unsigned int)write(fd, super->migr_rec_buf, |
de44e46f PB |
3120 | MIGR_REC_BUF_SECTORS*sector_size) != |
3121 | MIGR_REC_BUF_SECTORS*sector_size) { | |
e7b84f9d N |
3122 | pr_err("Cannot write migr record block: %s\n", |
3123 | strerror(errno)); | |
687629c2 AK |
3124 | goto out; |
3125 | } | |
3126 | close(fd); | |
3127 | fd = -1; | |
3128 | } | |
de44e46f PB |
3129 | if (sector_size == 4096) |
3130 | convert_from_4k_imsm_migr_rec(super); | |
c17608ea AK |
3131 | /* update checkpoint information in metadata */ |
3132 | len = imsm_create_metadata_checkpoint_update(super, &u); | |
c17608ea AK |
3133 | if (len <= 0) { |
3134 | dprintf("imsm: Cannot prepare update\n"); | |
3135 | goto out; | |
3136 | } | |
3137 | /* update metadata locally */ | |
3138 | imsm_update_metadata_locally(st, u, len); | |
3139 | /* and possibly remotely */ | |
3140 | if (st->update_tail) { | |
3141 | append_metadata_update(st, u, len); | |
3142 | /* during reshape we do all work inside metadata handler | |
3143 | * manage_reshape(), so metadata update has to be triggered | |
3144 | * insida it | |
3145 | */ | |
3146 | flush_metadata_updates(st); | |
3147 | st->update_tail = &st->updates; | |
3148 | } else | |
3149 | free(u); | |
687629c2 AK |
3150 | |
3151 | retval = 0; | |
3152 | out: | |
3153 | if (fd >= 0) | |
3154 | close(fd); | |
3155 | return retval; | |
3156 | } | |
3157 | ||
e2962bfc AK |
3158 | /* spare/missing disks activations are not allowe when |
3159 | * array/container performs reshape operation, because | |
3160 | * all arrays in container works on the same disks set | |
3161 | */ | |
3162 | int imsm_reshape_blocks_arrays_changes(struct intel_super *super) | |
3163 | { | |
3164 | int rv = 0; | |
3165 | struct intel_dev *i_dev; | |
3166 | struct imsm_dev *dev; | |
3167 | ||
3168 | /* check whole container | |
3169 | */ | |
3170 | for (i_dev = super->devlist; i_dev; i_dev = i_dev->next) { | |
3171 | dev = i_dev->dev; | |
3ad25638 | 3172 | if (is_gen_migration(dev)) { |
e2962bfc AK |
3173 | /* No repair during any migration in container |
3174 | */ | |
3175 | rv = 1; | |
3176 | break; | |
3177 | } | |
3178 | } | |
3179 | return rv; | |
3180 | } | |
c41e00b2 AK |
3181 | static unsigned long long imsm_component_size_aligment_check(int level, |
3182 | int chunk_size, | |
f36a9ecd | 3183 | unsigned int sector_size, |
c41e00b2 AK |
3184 | unsigned long long component_size) |
3185 | { | |
3186 | unsigned int component_size_alligment; | |
3187 | ||
3188 | /* check component size aligment | |
3189 | */ | |
f36a9ecd | 3190 | component_size_alligment = component_size % (chunk_size/sector_size); |
c41e00b2 | 3191 | |
1ade5cc1 | 3192 | dprintf("(Level: %i, chunk_size = %i, component_size = %llu), component_size_alligment = %u\n", |
c41e00b2 AK |
3193 | level, chunk_size, component_size, |
3194 | component_size_alligment); | |
3195 | ||
3196 | if (component_size_alligment && (level != 1) && (level != UnSet)) { | |
3197 | dprintf("imsm: reported component size alligned from %llu ", | |
3198 | component_size); | |
3199 | component_size -= component_size_alligment; | |
1ade5cc1 | 3200 | dprintf_cont("to %llu (%i).\n", |
c41e00b2 AK |
3201 | component_size, component_size_alligment); |
3202 | } | |
3203 | ||
3204 | return component_size; | |
3205 | } | |
e2962bfc | 3206 | |
2432ce9b AP |
3207 | static unsigned long long get_ppl_sector(struct intel_super *super, int dev_idx) |
3208 | { | |
3209 | struct imsm_dev *dev = get_imsm_dev(super, dev_idx); | |
3210 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
3211 | ||
3212 | return pba_of_lba0(map) + | |
3213 | (num_data_stripes(map) * map->blocks_per_strip); | |
3214 | } | |
3215 | ||
a5d85af7 | 3216 | static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap) |
bf5a934a DW |
3217 | { |
3218 | struct intel_super *super = st->sb; | |
c47b0ff6 | 3219 | struct migr_record *migr_rec = super->migr_rec; |
949c47a0 | 3220 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); |
238c0a71 AK |
3221 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
3222 | struct imsm_map *prev_map = get_imsm_map(dev, MAP_1); | |
b335e593 | 3223 | struct imsm_map *map_to_analyse = map; |
efb30e7f | 3224 | struct dl *dl; |
a5d85af7 | 3225 | int map_disks = info->array.raid_disks; |
bf5a934a | 3226 | |
95eeceeb | 3227 | memset(info, 0, sizeof(*info)); |
b335e593 AK |
3228 | if (prev_map) |
3229 | map_to_analyse = prev_map; | |
3230 | ||
ca0748fa | 3231 | dl = super->current_disk; |
9894ec0d | 3232 | |
bf5a934a | 3233 | info->container_member = super->current_vol; |
cd0430a1 | 3234 | info->array.raid_disks = map->num_members; |
b335e593 | 3235 | info->array.level = get_imsm_raid_level(map_to_analyse); |
bf5a934a DW |
3236 | info->array.layout = imsm_level_to_layout(info->array.level); |
3237 | info->array.md_minor = -1; | |
3238 | info->array.ctime = 0; | |
3239 | info->array.utime = 0; | |
b335e593 AK |
3240 | info->array.chunk_size = |
3241 | __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9; | |
2432ce9b | 3242 | info->array.state = !(dev->vol.dirty & RAIDVOL_DIRTY); |
da9b4a62 DW |
3243 | info->custom_array_size = __le32_to_cpu(dev->size_high); |
3244 | info->custom_array_size <<= 32; | |
3245 | info->custom_array_size |= __le32_to_cpu(dev->size_low); | |
3ad25638 AK |
3246 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
3247 | ||
3f510843 | 3248 | if (is_gen_migration(dev)) { |
3f83228a | 3249 | info->reshape_active = 1; |
b335e593 AK |
3250 | info->new_level = get_imsm_raid_level(map); |
3251 | info->new_layout = imsm_level_to_layout(info->new_level); | |
3252 | info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9; | |
3f83228a | 3253 | info->delta_disks = map->num_members - prev_map->num_members; |
493f5dd6 N |
3254 | if (info->delta_disks) { |
3255 | /* this needs to be applied to every array | |
3256 | * in the container. | |
3257 | */ | |
81219e70 | 3258 | info->reshape_active = CONTAINER_RESHAPE; |
493f5dd6 | 3259 | } |
3f83228a N |
3260 | /* We shape information that we give to md might have to be |
3261 | * modify to cope with md's requirement for reshaping arrays. | |
3262 | * For example, when reshaping a RAID0, md requires it to be | |
3263 | * presented as a degraded RAID4. | |
3264 | * Also if a RAID0 is migrating to a RAID5 we need to specify | |
3265 | * the array as already being RAID5, but the 'before' layout | |
3266 | * is a RAID4-like layout. | |
3267 | */ | |
3268 | switch (info->array.level) { | |
3269 | case 0: | |
3270 | switch(info->new_level) { | |
3271 | case 0: | |
3272 | /* conversion is happening as RAID4 */ | |
3273 | info->array.level = 4; | |
3274 | info->array.raid_disks += 1; | |
3275 | break; | |
3276 | case 5: | |
3277 | /* conversion is happening as RAID5 */ | |
3278 | info->array.level = 5; | |
3279 | info->array.layout = ALGORITHM_PARITY_N; | |
3f83228a N |
3280 | info->delta_disks -= 1; |
3281 | break; | |
3282 | default: | |
3283 | /* FIXME error message */ | |
3284 | info->array.level = UnSet; | |
3285 | break; | |
3286 | } | |
3287 | break; | |
3288 | } | |
b335e593 AK |
3289 | } else { |
3290 | info->new_level = UnSet; | |
3291 | info->new_layout = UnSet; | |
3292 | info->new_chunk = info->array.chunk_size; | |
3f83228a | 3293 | info->delta_disks = 0; |
b335e593 | 3294 | } |
ca0748fa | 3295 | |
efb30e7f DW |
3296 | if (dl) { |
3297 | info->disk.major = dl->major; | |
3298 | info->disk.minor = dl->minor; | |
ca0748fa | 3299 | info->disk.number = dl->index; |
656b6b5a N |
3300 | info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse, |
3301 | dl->index); | |
efb30e7f | 3302 | } |
bf5a934a | 3303 | |
5551b113 | 3304 | info->data_offset = pba_of_lba0(map_to_analyse); |
06fb291a PB |
3305 | |
3306 | if (info->array.level == 5) { | |
3307 | info->component_size = num_data_stripes(map_to_analyse) * | |
3308 | map_to_analyse->blocks_per_strip; | |
3309 | } else { | |
3310 | info->component_size = blocks_per_member(map_to_analyse); | |
3311 | } | |
139dae11 | 3312 | |
c41e00b2 AK |
3313 | info->component_size = imsm_component_size_aligment_check( |
3314 | info->array.level, | |
3315 | info->array.chunk_size, | |
f36a9ecd | 3316 | super->sector_size, |
c41e00b2 | 3317 | info->component_size); |
5e46202e | 3318 | info->bb.supported = 1; |
139dae11 | 3319 | |
301406c9 | 3320 | memset(info->uuid, 0, sizeof(info->uuid)); |
921d9e16 | 3321 | info->recovery_start = MaxSector; |
bf5a934a | 3322 | |
c2462068 PB |
3323 | if (info->array.level == 5 && |
3324 | (dev->rwh_policy == RWH_DISTRIBUTED || | |
3325 | dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)) { | |
2432ce9b AP |
3326 | info->consistency_policy = CONSISTENCY_POLICY_PPL; |
3327 | info->ppl_sector = get_ppl_sector(super, super->current_vol); | |
c2462068 PB |
3328 | if (dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED) |
3329 | info->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9; | |
3330 | else | |
3331 | info->ppl_size = (PPL_HEADER_SIZE + PPL_ENTRY_SPACE) | |
3332 | >> 9; | |
2432ce9b AP |
3333 | } else if (info->array.level <= 0) { |
3334 | info->consistency_policy = CONSISTENCY_POLICY_NONE; | |
3335 | } else { | |
3336 | info->consistency_policy = CONSISTENCY_POLICY_RESYNC; | |
3337 | } | |
3338 | ||
d2e6d5d6 | 3339 | info->reshape_progress = 0; |
b6796ce1 | 3340 | info->resync_start = MaxSector; |
b9172665 | 3341 | if ((map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED || |
2432ce9b | 3342 | !(info->array.state & 1)) && |
b9172665 | 3343 | imsm_reshape_blocks_arrays_changes(super) == 0) { |
301406c9 | 3344 | info->resync_start = 0; |
b6796ce1 AK |
3345 | } |
3346 | if (dev->vol.migr_state) { | |
1e5c6983 DW |
3347 | switch (migr_type(dev)) { |
3348 | case MIGR_REPAIR: | |
3349 | case MIGR_INIT: { | |
c47b0ff6 AK |
3350 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
3351 | dev); | |
1e5c6983 DW |
3352 | __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit); |
3353 | ||
3354 | info->resync_start = blocks_per_unit * units; | |
3355 | break; | |
3356 | } | |
d2e6d5d6 | 3357 | case MIGR_GEN_MIGR: { |
c47b0ff6 AK |
3358 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
3359 | dev); | |
3360 | __u64 units = __le32_to_cpu(migr_rec->curr_migr_unit); | |
04fa9523 AK |
3361 | unsigned long long array_blocks; |
3362 | int used_disks; | |
d2e6d5d6 | 3363 | |
befb629b AK |
3364 | if (__le32_to_cpu(migr_rec->ascending_migr) && |
3365 | (units < | |
3366 | (__le32_to_cpu(migr_rec->num_migr_units)-1)) && | |
3367 | (super->migr_rec->rec_status == | |
3368 | __cpu_to_le32(UNIT_SRC_IN_CP_AREA))) | |
3369 | units++; | |
3370 | ||
d2e6d5d6 | 3371 | info->reshape_progress = blocks_per_unit * units; |
6289d1e0 | 3372 | |
7a862a02 | 3373 | dprintf("IMSM: General Migration checkpoint : %llu (%llu) -> read reshape progress : %llu\n", |
19986c72 MB |
3374 | (unsigned long long)units, |
3375 | (unsigned long long)blocks_per_unit, | |
3376 | info->reshape_progress); | |
75156c46 | 3377 | |
238c0a71 | 3378 | used_disks = imsm_num_data_members(dev, MAP_1); |
75156c46 | 3379 | if (used_disks > 0) { |
5551b113 | 3380 | array_blocks = blocks_per_member(map) * |
75156c46 | 3381 | used_disks; |
b53bfba6 TM |
3382 | info->custom_array_size = |
3383 | round_size_to_mb(array_blocks, | |
3384 | used_disks); | |
3385 | ||
75156c46 | 3386 | } |
d2e6d5d6 | 3387 | } |
1e5c6983 DW |
3388 | case MIGR_VERIFY: |
3389 | /* we could emulate the checkpointing of | |
3390 | * 'sync_action=check' migrations, but for now | |
3391 | * we just immediately complete them | |
3392 | */ | |
3393 | case MIGR_REBUILD: | |
3394 | /* this is handled by container_content_imsm() */ | |
1e5c6983 DW |
3395 | case MIGR_STATE_CHANGE: |
3396 | /* FIXME handle other migrations */ | |
3397 | default: | |
3398 | /* we are not dirty, so... */ | |
3399 | info->resync_start = MaxSector; | |
3400 | } | |
b6796ce1 | 3401 | } |
301406c9 DW |
3402 | |
3403 | strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN); | |
3404 | info->name[MAX_RAID_SERIAL_LEN] = 0; | |
bf5a934a | 3405 | |
f35f2525 N |
3406 | info->array.major_version = -1; |
3407 | info->array.minor_version = -2; | |
4dd2df09 | 3408 | sprintf(info->text_version, "/%s/%d", st->container_devnm, info->container_member); |
a67dd8cc | 3409 | info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */ |
51006d85 | 3410 | uuid_from_super_imsm(st, info->uuid); |
a5d85af7 N |
3411 | |
3412 | if (dmap) { | |
3413 | int i, j; | |
3414 | for (i=0; i<map_disks; i++) { | |
3415 | dmap[i] = 0; | |
3416 | if (i < info->array.raid_disks) { | |
3417 | struct imsm_disk *dsk; | |
238c0a71 | 3418 | j = get_imsm_disk_idx(dev, i, MAP_X); |
a5d85af7 N |
3419 | dsk = get_imsm_disk(super, j); |
3420 | if (dsk && (dsk->status & CONFIGURED_DISK)) | |
3421 | dmap[i] = 1; | |
3422 | } | |
3423 | } | |
3424 | } | |
81ac8b4d | 3425 | } |
bf5a934a | 3426 | |
3b451610 AK |
3427 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, |
3428 | int failed, int look_in_map); | |
3429 | ||
3430 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev, | |
3431 | int look_in_map); | |
3432 | ||
3433 | static void manage_second_map(struct intel_super *super, struct imsm_dev *dev) | |
3434 | { | |
3435 | if (is_gen_migration(dev)) { | |
3436 | int failed; | |
3437 | __u8 map_state; | |
3438 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); | |
3439 | ||
3440 | failed = imsm_count_failed(super, dev, MAP_1); | |
238c0a71 | 3441 | map_state = imsm_check_degraded(super, dev, failed, MAP_1); |
3b451610 AK |
3442 | if (map2->map_state != map_state) { |
3443 | map2->map_state = map_state; | |
3444 | super->updates_pending++; | |
3445 | } | |
3446 | } | |
3447 | } | |
97b4d0e9 DW |
3448 | |
3449 | static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index) | |
3450 | { | |
3451 | struct dl *d; | |
3452 | ||
3453 | for (d = super->missing; d; d = d->next) | |
3454 | if (d->index == index) | |
3455 | return &d->disk; | |
3456 | return NULL; | |
3457 | } | |
3458 | ||
a5d85af7 | 3459 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map) |
4f5bc454 DW |
3460 | { |
3461 | struct intel_super *super = st->sb; | |
4f5bc454 | 3462 | struct imsm_disk *disk; |
a5d85af7 | 3463 | int map_disks = info->array.raid_disks; |
ab3cb6b3 N |
3464 | int max_enough = -1; |
3465 | int i; | |
3466 | struct imsm_super *mpb; | |
4f5bc454 | 3467 | |
bf5a934a | 3468 | if (super->current_vol >= 0) { |
a5d85af7 | 3469 | getinfo_super_imsm_volume(st, info, map); |
bf5a934a DW |
3470 | return; |
3471 | } | |
95eeceeb | 3472 | memset(info, 0, sizeof(*info)); |
d23fe947 DW |
3473 | |
3474 | /* Set raid_disks to zero so that Assemble will always pull in valid | |
3475 | * spares | |
3476 | */ | |
3477 | info->array.raid_disks = 0; | |
cdddbdbc DW |
3478 | info->array.level = LEVEL_CONTAINER; |
3479 | info->array.layout = 0; | |
3480 | info->array.md_minor = -1; | |
1011e834 | 3481 | info->array.ctime = 0; /* N/A for imsm */ |
cdddbdbc DW |
3482 | info->array.utime = 0; |
3483 | info->array.chunk_size = 0; | |
3484 | ||
3485 | info->disk.major = 0; | |
3486 | info->disk.minor = 0; | |
cdddbdbc | 3487 | info->disk.raid_disk = -1; |
c2c087e6 | 3488 | info->reshape_active = 0; |
f35f2525 N |
3489 | info->array.major_version = -1; |
3490 | info->array.minor_version = -2; | |
c2c087e6 | 3491 | strcpy(info->text_version, "imsm"); |
a67dd8cc | 3492 | info->safe_mode_delay = 0; |
c2c087e6 DW |
3493 | info->disk.number = -1; |
3494 | info->disk.state = 0; | |
c5afc314 | 3495 | info->name[0] = 0; |
921d9e16 | 3496 | info->recovery_start = MaxSector; |
3ad25638 | 3497 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
5e46202e | 3498 | info->bb.supported = 1; |
c2c087e6 | 3499 | |
97b4d0e9 | 3500 | /* do we have the all the insync disks that we expect? */ |
ab3cb6b3 | 3501 | mpb = super->anchor; |
b7d81a38 | 3502 | info->events = __le32_to_cpu(mpb->generation_num); |
97b4d0e9 | 3503 | |
ab3cb6b3 N |
3504 | for (i = 0; i < mpb->num_raid_devs; i++) { |
3505 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
3506 | int failed, enough, j, missing = 0; | |
3507 | struct imsm_map *map; | |
3508 | __u8 state; | |
97b4d0e9 | 3509 | |
3b451610 AK |
3510 | failed = imsm_count_failed(super, dev, MAP_0); |
3511 | state = imsm_check_degraded(super, dev, failed, MAP_0); | |
238c0a71 | 3512 | map = get_imsm_map(dev, MAP_0); |
ab3cb6b3 N |
3513 | |
3514 | /* any newly missing disks? | |
3515 | * (catches single-degraded vs double-degraded) | |
3516 | */ | |
3517 | for (j = 0; j < map->num_members; j++) { | |
238c0a71 | 3518 | __u32 ord = get_imsm_ord_tbl_ent(dev, j, MAP_0); |
ab3cb6b3 N |
3519 | __u32 idx = ord_to_idx(ord); |
3520 | ||
20dc76d1 MT |
3521 | if (super->disks && super->disks->index == (int)idx) |
3522 | info->disk.raid_disk = j; | |
3523 | ||
ab3cb6b3 N |
3524 | if (!(ord & IMSM_ORD_REBUILD) && |
3525 | get_imsm_missing(super, idx)) { | |
3526 | missing = 1; | |
3527 | break; | |
3528 | } | |
97b4d0e9 | 3529 | } |
ab3cb6b3 N |
3530 | |
3531 | if (state == IMSM_T_STATE_FAILED) | |
3532 | enough = -1; | |
3533 | else if (state == IMSM_T_STATE_DEGRADED && | |
3534 | (state != map->map_state || missing)) | |
3535 | enough = 0; | |
3536 | else /* we're normal, or already degraded */ | |
3537 | enough = 1; | |
d2bde6d3 AK |
3538 | if (is_gen_migration(dev) && missing) { |
3539 | /* during general migration we need all disks | |
3540 | * that process is running on. | |
3541 | * No new missing disk is allowed. | |
3542 | */ | |
3543 | max_enough = -1; | |
3544 | enough = -1; | |
3545 | /* no more checks necessary | |
3546 | */ | |
3547 | break; | |
3548 | } | |
ab3cb6b3 N |
3549 | /* in the missing/failed disk case check to see |
3550 | * if at least one array is runnable | |
3551 | */ | |
3552 | max_enough = max(max_enough, enough); | |
3553 | } | |
1ade5cc1 | 3554 | dprintf("enough: %d\n", max_enough); |
ab3cb6b3 | 3555 | info->container_enough = max_enough; |
97b4d0e9 | 3556 | |
4a04ec6c | 3557 | if (super->disks) { |
14e8215b DW |
3558 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
3559 | ||
b9f594fe | 3560 | disk = &super->disks->disk; |
5551b113 | 3561 | info->data_offset = total_blocks(&super->disks->disk) - reserved; |
14e8215b | 3562 | info->component_size = reserved; |
25ed7e59 | 3563 | info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0; |
df474657 DW |
3564 | /* we don't change info->disk.raid_disk here because |
3565 | * this state will be finalized in mdmon after we have | |
3566 | * found the 'most fresh' version of the metadata | |
3567 | */ | |
25ed7e59 | 3568 | info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; |
2432ce9b AP |
3569 | info->disk.state |= (is_spare(disk) || is_journal(disk)) ? |
3570 | 0 : (1 << MD_DISK_SYNC); | |
cdddbdbc | 3571 | } |
a575e2a7 DW |
3572 | |
3573 | /* only call uuid_from_super_imsm when this disk is part of a populated container, | |
3574 | * ->compare_super may have updated the 'num_raid_devs' field for spares | |
3575 | */ | |
3576 | if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs) | |
36ba7d48 | 3577 | uuid_from_super_imsm(st, info->uuid); |
22e263f6 AC |
3578 | else |
3579 | memcpy(info->uuid, uuid_zero, sizeof(uuid_zero)); | |
a5d85af7 N |
3580 | |
3581 | /* I don't know how to compute 'map' on imsm, so use safe default */ | |
3582 | if (map) { | |
3583 | int i; | |
3584 | for (i = 0; i < map_disks; i++) | |
3585 | map[i] = 1; | |
3586 | } | |
3587 | ||
cdddbdbc DW |
3588 | } |
3589 | ||
5c4cd5da AC |
3590 | /* allocates memory and fills disk in mdinfo structure |
3591 | * for each disk in array */ | |
3592 | struct mdinfo *getinfo_super_disks_imsm(struct supertype *st) | |
3593 | { | |
594dc1b8 | 3594 | struct mdinfo *mddev; |
5c4cd5da AC |
3595 | struct intel_super *super = st->sb; |
3596 | struct imsm_disk *disk; | |
3597 | int count = 0; | |
3598 | struct dl *dl; | |
3599 | if (!super || !super->disks) | |
3600 | return NULL; | |
3601 | dl = super->disks; | |
503975b9 | 3602 | mddev = xcalloc(1, sizeof(*mddev)); |
5c4cd5da AC |
3603 | while (dl) { |
3604 | struct mdinfo *tmp; | |
3605 | disk = &dl->disk; | |
503975b9 | 3606 | tmp = xcalloc(1, sizeof(*tmp)); |
5c4cd5da AC |
3607 | if (mddev->devs) |
3608 | tmp->next = mddev->devs; | |
3609 | mddev->devs = tmp; | |
3610 | tmp->disk.number = count++; | |
3611 | tmp->disk.major = dl->major; | |
3612 | tmp->disk.minor = dl->minor; | |
3613 | tmp->disk.state = is_configured(disk) ? | |
3614 | (1 << MD_DISK_ACTIVE) : 0; | |
3615 | tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; | |
3616 | tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
3617 | tmp->disk.raid_disk = -1; | |
3618 | dl = dl->next; | |
3619 | } | |
3620 | return mddev; | |
3621 | } | |
3622 | ||
cdddbdbc DW |
3623 | static int update_super_imsm(struct supertype *st, struct mdinfo *info, |
3624 | char *update, char *devname, int verbose, | |
3625 | int uuid_set, char *homehost) | |
3626 | { | |
f352c545 DW |
3627 | /* For 'assemble' and 'force' we need to return non-zero if any |
3628 | * change was made. For others, the return value is ignored. | |
3629 | * Update options are: | |
3630 | * force-one : This device looks a bit old but needs to be included, | |
3631 | * update age info appropriately. | |
3632 | * assemble: clear any 'faulty' flag to allow this device to | |
3633 | * be assembled. | |
3634 | * force-array: Array is degraded but being forced, mark it clean | |
3635 | * if that will be needed to assemble it. | |
3636 | * | |
3637 | * newdev: not used ???? | |
3638 | * grow: Array has gained a new device - this is currently for | |
3639 | * linear only | |
3640 | * resync: mark as dirty so a resync will happen. | |
3641 | * name: update the name - preserving the homehost | |
6e46bf34 | 3642 | * uuid: Change the uuid of the array to match watch is given |
f352c545 DW |
3643 | * |
3644 | * Following are not relevant for this imsm: | |
3645 | * sparc2.2 : update from old dodgey metadata | |
3646 | * super-minor: change the preferred_minor number | |
3647 | * summaries: update redundant counters. | |
f352c545 DW |
3648 | * homehost: update the recorded homehost |
3649 | * _reshape_progress: record new reshape_progress position. | |
3650 | */ | |
6e46bf34 DW |
3651 | int rv = 1; |
3652 | struct intel_super *super = st->sb; | |
3653 | struct imsm_super *mpb; | |
f352c545 | 3654 | |
6e46bf34 DW |
3655 | /* we can only update container info */ |
3656 | if (!super || super->current_vol >= 0 || !super->anchor) | |
3657 | return 1; | |
3658 | ||
3659 | mpb = super->anchor; | |
3660 | ||
81a5b4f5 N |
3661 | if (strcmp(update, "uuid") == 0) { |
3662 | /* We take this to mean that the family_num should be updated. | |
3663 | * However that is much smaller than the uuid so we cannot really | |
3664 | * allow an explicit uuid to be given. And it is hard to reliably | |
3665 | * know if one was. | |
3666 | * So if !uuid_set we know the current uuid is random and just used | |
3667 | * the first 'int' and copy it to the other 3 positions. | |
3668 | * Otherwise we require the 4 'int's to be the same as would be the | |
3669 | * case if we are using a random uuid. So an explicit uuid will be | |
3670 | * accepted as long as all for ints are the same... which shouldn't hurt | |
6e46bf34 | 3671 | */ |
81a5b4f5 N |
3672 | if (!uuid_set) { |
3673 | info->uuid[1] = info->uuid[2] = info->uuid[3] = info->uuid[0]; | |
6e46bf34 | 3674 | rv = 0; |
81a5b4f5 N |
3675 | } else { |
3676 | if (info->uuid[0] != info->uuid[1] || | |
3677 | info->uuid[1] != info->uuid[2] || | |
3678 | info->uuid[2] != info->uuid[3]) | |
3679 | rv = -1; | |
3680 | else | |
3681 | rv = 0; | |
6e46bf34 | 3682 | } |
81a5b4f5 N |
3683 | if (rv == 0) |
3684 | mpb->orig_family_num = info->uuid[0]; | |
6e46bf34 DW |
3685 | } else if (strcmp(update, "assemble") == 0) |
3686 | rv = 0; | |
3687 | else | |
1e2b2765 | 3688 | rv = -1; |
f352c545 | 3689 | |
6e46bf34 DW |
3690 | /* successful update? recompute checksum */ |
3691 | if (rv == 0) | |
3692 | mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb)); | |
f352c545 DW |
3693 | |
3694 | return rv; | |
cdddbdbc DW |
3695 | } |
3696 | ||
c2c087e6 | 3697 | static size_t disks_to_mpb_size(int disks) |
cdddbdbc | 3698 | { |
c2c087e6 | 3699 | size_t size; |
cdddbdbc | 3700 | |
c2c087e6 DW |
3701 | size = sizeof(struct imsm_super); |
3702 | size += (disks - 1) * sizeof(struct imsm_disk); | |
3703 | size += 2 * sizeof(struct imsm_dev); | |
3704 | /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */ | |
3705 | size += (4 - 2) * sizeof(struct imsm_map); | |
3706 | /* 4 possible disk_ord_tbl's */ | |
3707 | size += 4 * (disks - 1) * sizeof(__u32); | |
bbab0940 TM |
3708 | /* maximum bbm log */ |
3709 | size += sizeof(struct bbm_log); | |
c2c087e6 DW |
3710 | |
3711 | return size; | |
3712 | } | |
3713 | ||
387fcd59 N |
3714 | static __u64 avail_size_imsm(struct supertype *st, __u64 devsize, |
3715 | unsigned long long data_offset) | |
c2c087e6 DW |
3716 | { |
3717 | if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS)) | |
3718 | return 0; | |
3719 | ||
3720 | return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); | |
cdddbdbc DW |
3721 | } |
3722 | ||
ba2de7ba DW |
3723 | static void free_devlist(struct intel_super *super) |
3724 | { | |
3725 | struct intel_dev *dv; | |
3726 | ||
3727 | while (super->devlist) { | |
3728 | dv = super->devlist->next; | |
3729 | free(super->devlist->dev); | |
3730 | free(super->devlist); | |
3731 | super->devlist = dv; | |
3732 | } | |
3733 | } | |
3734 | ||
3735 | static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src) | |
3736 | { | |
3737 | memcpy(dest, src, sizeof_imsm_dev(src, 0)); | |
3738 | } | |
3739 | ||
cdddbdbc DW |
3740 | static int compare_super_imsm(struct supertype *st, struct supertype *tst) |
3741 | { | |
3742 | /* | |
3743 | * return: | |
3744 | * 0 same, or first was empty, and second was copied | |
3745 | * 1 second had wrong number | |
3746 | * 2 wrong uuid | |
3747 | * 3 wrong other info | |
3748 | */ | |
3749 | struct intel_super *first = st->sb; | |
3750 | struct intel_super *sec = tst->sb; | |
3751 | ||
5d500228 N |
3752 | if (!first) { |
3753 | st->sb = tst->sb; | |
3754 | tst->sb = NULL; | |
3755 | return 0; | |
3756 | } | |
8603ea6f LM |
3757 | /* in platform dependent environment test if the disks |
3758 | * use the same Intel hba | |
cb8f6859 | 3759 | * If not on Intel hba at all, allow anything. |
8603ea6f | 3760 | */ |
6b781d33 AP |
3761 | if (!check_env("IMSM_NO_PLATFORM") && first->hba && sec->hba) { |
3762 | if (first->hba->type != sec->hba->type) { | |
8603ea6f | 3763 | fprintf(stderr, |
6b781d33 AP |
3764 | "HBAs of devices do not match %s != %s\n", |
3765 | get_sys_dev_type(first->hba->type), | |
3766 | get_sys_dev_type(sec->hba->type)); | |
3767 | return 3; | |
3768 | } | |
3769 | if (first->orom != sec->orom) { | |
3770 | fprintf(stderr, | |
3771 | "HBAs of devices do not match %s != %s\n", | |
3772 | first->hba->pci_id, sec->hba->pci_id); | |
8603ea6f LM |
3773 | return 3; |
3774 | } | |
3775 | } | |
cdddbdbc | 3776 | |
d23fe947 DW |
3777 | /* if an anchor does not have num_raid_devs set then it is a free |
3778 | * floating spare | |
3779 | */ | |
3780 | if (first->anchor->num_raid_devs > 0 && | |
3781 | sec->anchor->num_raid_devs > 0) { | |
a2b97981 DW |
3782 | /* Determine if these disks might ever have been |
3783 | * related. Further disambiguation can only take place | |
3784 | * in load_super_imsm_all | |
3785 | */ | |
3786 | __u32 first_family = first->anchor->orig_family_num; | |
3787 | __u32 sec_family = sec->anchor->orig_family_num; | |
3788 | ||
f796af5d DW |
3789 | if (memcmp(first->anchor->sig, sec->anchor->sig, |
3790 | MAX_SIGNATURE_LENGTH) != 0) | |
3791 | return 3; | |
3792 | ||
a2b97981 DW |
3793 | if (first_family == 0) |
3794 | first_family = first->anchor->family_num; | |
3795 | if (sec_family == 0) | |
3796 | sec_family = sec->anchor->family_num; | |
3797 | ||
3798 | if (first_family != sec_family) | |
d23fe947 | 3799 | return 3; |
f796af5d | 3800 | |
d23fe947 | 3801 | } |
cdddbdbc | 3802 | |
3e372e5a DW |
3803 | /* if 'first' is a spare promote it to a populated mpb with sec's |
3804 | * family number | |
3805 | */ | |
3806 | if (first->anchor->num_raid_devs == 0 && | |
3807 | sec->anchor->num_raid_devs > 0) { | |
78d30f94 | 3808 | int i; |
ba2de7ba DW |
3809 | struct intel_dev *dv; |
3810 | struct imsm_dev *dev; | |
78d30f94 DW |
3811 | |
3812 | /* we need to copy raid device info from sec if an allocation | |
3813 | * fails here we don't associate the spare | |
3814 | */ | |
3815 | for (i = 0; i < sec->anchor->num_raid_devs; i++) { | |
503975b9 N |
3816 | dv = xmalloc(sizeof(*dv)); |
3817 | dev = xmalloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1)); | |
ba2de7ba DW |
3818 | dv->dev = dev; |
3819 | dv->index = i; | |
3820 | dv->next = first->devlist; | |
3821 | first->devlist = dv; | |
78d30f94 | 3822 | } |
709743c5 | 3823 | if (i < sec->anchor->num_raid_devs) { |
ba2de7ba DW |
3824 | /* allocation failure */ |
3825 | free_devlist(first); | |
e12b3daa | 3826 | pr_err("imsm: failed to associate spare\n"); |
ba2de7ba | 3827 | return 3; |
78d30f94 | 3828 | } |
3e372e5a | 3829 | first->anchor->num_raid_devs = sec->anchor->num_raid_devs; |
148acb7b | 3830 | first->anchor->orig_family_num = sec->anchor->orig_family_num; |
3e372e5a | 3831 | first->anchor->family_num = sec->anchor->family_num; |
ac6449be | 3832 | memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH); |
709743c5 DW |
3833 | for (i = 0; i < sec->anchor->num_raid_devs; i++) |
3834 | imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i)); | |
3e372e5a DW |
3835 | } |
3836 | ||
cdddbdbc DW |
3837 | return 0; |
3838 | } | |
3839 | ||
0030e8d6 DW |
3840 | static void fd2devname(int fd, char *name) |
3841 | { | |
3842 | struct stat st; | |
3843 | char path[256]; | |
33a6535d | 3844 | char dname[PATH_MAX]; |
0030e8d6 DW |
3845 | char *nm; |
3846 | int rv; | |
3847 | ||
3848 | name[0] = '\0'; | |
3849 | if (fstat(fd, &st) != 0) | |
3850 | return; | |
3851 | sprintf(path, "/sys/dev/block/%d:%d", | |
3852 | major(st.st_rdev), minor(st.st_rdev)); | |
3853 | ||
9cf014ec | 3854 | rv = readlink(path, dname, sizeof(dname)-1); |
0030e8d6 DW |
3855 | if (rv <= 0) |
3856 | return; | |
9587c373 | 3857 | |
0030e8d6 DW |
3858 | dname[rv] = '\0'; |
3859 | nm = strrchr(dname, '/'); | |
7897de29 JS |
3860 | if (nm) { |
3861 | nm++; | |
3862 | snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm); | |
3863 | } | |
0030e8d6 DW |
3864 | } |
3865 | ||
21e9380b AP |
3866 | static int nvme_get_serial(int fd, void *buf, size_t buf_len) |
3867 | { | |
3868 | char path[60]; | |
3869 | char *name = fd2kname(fd); | |
3870 | ||
3871 | if (!name) | |
3872 | return 1; | |
3873 | ||
3874 | if (strncmp(name, "nvme", 4) != 0) | |
3875 | return 1; | |
3876 | ||
3877 | snprintf(path, sizeof(path) - 1, "/sys/block/%s/device/serial", name); | |
3878 | ||
3879 | return load_sys(path, buf, buf_len); | |
3880 | } | |
3881 | ||
cdddbdbc DW |
3882 | extern int scsi_get_serial(int fd, void *buf, size_t buf_len); |
3883 | ||
3884 | static int imsm_read_serial(int fd, char *devname, | |
3885 | __u8 serial[MAX_RAID_SERIAL_LEN]) | |
3886 | { | |
21e9380b | 3887 | char buf[50]; |
cdddbdbc | 3888 | int rv; |
1f24f035 | 3889 | int len; |
316e2bf4 DW |
3890 | char *dest; |
3891 | char *src; | |
21e9380b AP |
3892 | unsigned int i; |
3893 | ||
3894 | memset(buf, 0, sizeof(buf)); | |
cdddbdbc | 3895 | |
21e9380b | 3896 | rv = nvme_get_serial(fd, buf, sizeof(buf)); |
cdddbdbc | 3897 | |
21e9380b AP |
3898 | if (rv) |
3899 | rv = scsi_get_serial(fd, buf, sizeof(buf)); | |
f9ba0ff1 | 3900 | |
40ebbb9c | 3901 | if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) { |
f9ba0ff1 DW |
3902 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
3903 | fd2devname(fd, (char *) serial); | |
0030e8d6 DW |
3904 | return 0; |
3905 | } | |
3906 | ||
cdddbdbc DW |
3907 | if (rv != 0) { |
3908 | if (devname) | |
e7b84f9d N |
3909 | pr_err("Failed to retrieve serial for %s\n", |
3910 | devname); | |
cdddbdbc DW |
3911 | return rv; |
3912 | } | |
3913 | ||
316e2bf4 DW |
3914 | /* trim all whitespace and non-printable characters and convert |
3915 | * ':' to ';' | |
3916 | */ | |
21e9380b AP |
3917 | for (i = 0, dest = buf; i < sizeof(buf) && buf[i]; i++) { |
3918 | src = &buf[i]; | |
316e2bf4 DW |
3919 | if (*src > 0x20) { |
3920 | /* ':' is reserved for use in placeholder serial | |
3921 | * numbers for missing disks | |
3922 | */ | |
3923 | if (*src == ':') | |
3924 | *dest++ = ';'; | |
3925 | else | |
3926 | *dest++ = *src; | |
3927 | } | |
3928 | } | |
21e9380b AP |
3929 | len = dest - buf; |
3930 | dest = buf; | |
316e2bf4 DW |
3931 | |
3932 | /* truncate leading characters */ | |
3933 | if (len > MAX_RAID_SERIAL_LEN) { | |
3934 | dest += len - MAX_RAID_SERIAL_LEN; | |
1f24f035 | 3935 | len = MAX_RAID_SERIAL_LEN; |
316e2bf4 | 3936 | } |
5c3db629 | 3937 | |
5c3db629 | 3938 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
316e2bf4 | 3939 | memcpy(serial, dest, len); |
cdddbdbc DW |
3940 | |
3941 | return 0; | |
3942 | } | |
3943 | ||
1f24f035 DW |
3944 | static int serialcmp(__u8 *s1, __u8 *s2) |
3945 | { | |
3946 | return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN); | |
3947 | } | |
3948 | ||
3949 | static void serialcpy(__u8 *dest, __u8 *src) | |
3950 | { | |
3951 | strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN); | |
3952 | } | |
3953 | ||
54c2c1ea DW |
3954 | static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super) |
3955 | { | |
3956 | struct dl *dl; | |
3957 | ||
3958 | for (dl = super->disks; dl; dl = dl->next) | |
3959 | if (serialcmp(dl->serial, serial) == 0) | |
3960 | break; | |
3961 | ||
3962 | return dl; | |
3963 | } | |
3964 | ||
a2b97981 DW |
3965 | static struct imsm_disk * |
3966 | __serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx) | |
3967 | { | |
3968 | int i; | |
3969 | ||
3970 | for (i = 0; i < mpb->num_disks; i++) { | |
3971 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
3972 | ||
3973 | if (serialcmp(disk->serial, serial) == 0) { | |
3974 | if (idx) | |
3975 | *idx = i; | |
3976 | return disk; | |
3977 | } | |
3978 | } | |
3979 | ||
3980 | return NULL; | |
3981 | } | |
3982 | ||
cdddbdbc DW |
3983 | static int |
3984 | load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd) | |
3985 | { | |
a2b97981 | 3986 | struct imsm_disk *disk; |
cdddbdbc DW |
3987 | struct dl *dl; |
3988 | struct stat stb; | |
cdddbdbc | 3989 | int rv; |
a2b97981 | 3990 | char name[40]; |
d23fe947 DW |
3991 | __u8 serial[MAX_RAID_SERIAL_LEN]; |
3992 | ||
3993 | rv = imsm_read_serial(fd, devname, serial); | |
3994 | ||
3995 | if (rv != 0) | |
3996 | return 2; | |
3997 | ||
503975b9 | 3998 | dl = xcalloc(1, sizeof(*dl)); |
cdddbdbc | 3999 | |
a2b97981 DW |
4000 | fstat(fd, &stb); |
4001 | dl->major = major(stb.st_rdev); | |
4002 | dl->minor = minor(stb.st_rdev); | |
4003 | dl->next = super->disks; | |
4004 | dl->fd = keep_fd ? fd : -1; | |
4005 | assert(super->disks == NULL); | |
4006 | super->disks = dl; | |
4007 | serialcpy(dl->serial, serial); | |
4008 | dl->index = -2; | |
4009 | dl->e = NULL; | |
4010 | fd2devname(fd, name); | |
4011 | if (devname) | |
503975b9 | 4012 | dl->devname = xstrdup(devname); |
a2b97981 | 4013 | else |
503975b9 | 4014 | dl->devname = xstrdup(name); |
cdddbdbc | 4015 | |
d23fe947 | 4016 | /* look up this disk's index in the current anchor */ |
a2b97981 DW |
4017 | disk = __serial_to_disk(dl->serial, super->anchor, &dl->index); |
4018 | if (disk) { | |
4019 | dl->disk = *disk; | |
4020 | /* only set index on disks that are a member of a | |
4021 | * populated contianer, i.e. one with raid_devs | |
4022 | */ | |
4023 | if (is_failed(&dl->disk)) | |
3f6efecc | 4024 | dl->index = -2; |
2432ce9b | 4025 | else if (is_spare(&dl->disk) || is_journal(&dl->disk)) |
a2b97981 | 4026 | dl->index = -1; |
3f6efecc DW |
4027 | } |
4028 | ||
949c47a0 DW |
4029 | return 0; |
4030 | } | |
4031 | ||
0c046afd DW |
4032 | /* When migrating map0 contains the 'destination' state while map1 |
4033 | * contains the current state. When not migrating map0 contains the | |
4034 | * current state. This routine assumes that map[0].map_state is set to | |
4035 | * the current array state before being called. | |
4036 | * | |
4037 | * Migration is indicated by one of the following states | |
4038 | * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed) | |
e3bba0e0 | 4039 | * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal |
0c046afd | 4040 | * map1state=unitialized) |
1484e727 | 4041 | * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal |
0c046afd | 4042 | * map1state=normal) |
e3bba0e0 | 4043 | * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal |
0c046afd | 4044 | * map1state=degraded) |
8e59f3d8 AK |
4045 | * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal |
4046 | * map1state=normal) | |
0c046afd | 4047 | */ |
8e59f3d8 AK |
4048 | static void migrate(struct imsm_dev *dev, struct intel_super *super, |
4049 | __u8 to_state, int migr_type) | |
3393c6af | 4050 | { |
0c046afd | 4051 | struct imsm_map *dest; |
238c0a71 | 4052 | struct imsm_map *src = get_imsm_map(dev, MAP_0); |
3393c6af | 4053 | |
0c046afd | 4054 | dev->vol.migr_state = 1; |
1484e727 | 4055 | set_migr_type(dev, migr_type); |
f8f603f1 | 4056 | dev->vol.curr_migr_unit = 0; |
238c0a71 | 4057 | dest = get_imsm_map(dev, MAP_1); |
0c046afd | 4058 | |
0556e1a2 | 4059 | /* duplicate and then set the target end state in map[0] */ |
3393c6af | 4060 | memcpy(dest, src, sizeof_imsm_map(src)); |
fb12a745 | 4061 | if (migr_type == MIGR_GEN_MIGR) { |
0556e1a2 DW |
4062 | __u32 ord; |
4063 | int i; | |
4064 | ||
4065 | for (i = 0; i < src->num_members; i++) { | |
4066 | ord = __le32_to_cpu(src->disk_ord_tbl[i]); | |
4067 | set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord)); | |
4068 | } | |
4069 | } | |
4070 | ||
8e59f3d8 AK |
4071 | if (migr_type == MIGR_GEN_MIGR) |
4072 | /* Clear migration record */ | |
4073 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
4074 | ||
0c046afd | 4075 | src->map_state = to_state; |
949c47a0 | 4076 | } |
f8f603f1 | 4077 | |
809da78e AK |
4078 | static void end_migration(struct imsm_dev *dev, struct intel_super *super, |
4079 | __u8 map_state) | |
f8f603f1 | 4080 | { |
238c0a71 AK |
4081 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
4082 | struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state == 0 ? | |
4083 | MAP_0 : MAP_1); | |
28bce06f | 4084 | int i, j; |
0556e1a2 DW |
4085 | |
4086 | /* merge any IMSM_ORD_REBUILD bits that were not successfully | |
4087 | * completed in the last migration. | |
4088 | * | |
28bce06f | 4089 | * FIXME add support for raid-level-migration |
0556e1a2 | 4090 | */ |
089f9d79 JS |
4091 | if (map_state != map->map_state && (is_gen_migration(dev) == 0) && |
4092 | prev->map_state != IMSM_T_STATE_UNINITIALIZED) { | |
809da78e AK |
4093 | /* when final map state is other than expected |
4094 | * merge maps (not for migration) | |
4095 | */ | |
4096 | int failed; | |
4097 | ||
4098 | for (i = 0; i < prev->num_members; i++) | |
4099 | for (j = 0; j < map->num_members; j++) | |
4100 | /* during online capacity expansion | |
4101 | * disks position can be changed | |
4102 | * if takeover is used | |
4103 | */ | |
4104 | if (ord_to_idx(map->disk_ord_tbl[j]) == | |
4105 | ord_to_idx(prev->disk_ord_tbl[i])) { | |
4106 | map->disk_ord_tbl[j] |= | |
4107 | prev->disk_ord_tbl[i]; | |
4108 | break; | |
4109 | } | |
4110 | failed = imsm_count_failed(super, dev, MAP_0); | |
4111 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
4112 | } | |
f8f603f1 DW |
4113 | |
4114 | dev->vol.migr_state = 0; | |
ea672ee1 | 4115 | set_migr_type(dev, 0); |
f8f603f1 DW |
4116 | dev->vol.curr_migr_unit = 0; |
4117 | map->map_state = map_state; | |
4118 | } | |
949c47a0 DW |
4119 | |
4120 | static int parse_raid_devices(struct intel_super *super) | |
4121 | { | |
4122 | int i; | |
4123 | struct imsm_dev *dev_new; | |
4d7b1503 | 4124 | size_t len, len_migr; |
401d313b | 4125 | size_t max_len = 0; |
4d7b1503 DW |
4126 | size_t space_needed = 0; |
4127 | struct imsm_super *mpb = super->anchor; | |
949c47a0 DW |
4128 | |
4129 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
4130 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
ba2de7ba | 4131 | struct intel_dev *dv; |
949c47a0 | 4132 | |
4d7b1503 DW |
4133 | len = sizeof_imsm_dev(dev_iter, 0); |
4134 | len_migr = sizeof_imsm_dev(dev_iter, 1); | |
4135 | if (len_migr > len) | |
4136 | space_needed += len_migr - len; | |
ca9de185 | 4137 | |
503975b9 | 4138 | dv = xmalloc(sizeof(*dv)); |
401d313b AK |
4139 | if (max_len < len_migr) |
4140 | max_len = len_migr; | |
4141 | if (max_len > len_migr) | |
4142 | space_needed += max_len - len_migr; | |
503975b9 | 4143 | dev_new = xmalloc(max_len); |
949c47a0 | 4144 | imsm_copy_dev(dev_new, dev_iter); |
ba2de7ba DW |
4145 | dv->dev = dev_new; |
4146 | dv->index = i; | |
4147 | dv->next = super->devlist; | |
4148 | super->devlist = dv; | |
949c47a0 | 4149 | } |
cdddbdbc | 4150 | |
4d7b1503 DW |
4151 | /* ensure that super->buf is large enough when all raid devices |
4152 | * are migrating | |
4153 | */ | |
4154 | if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) { | |
4155 | void *buf; | |
4156 | ||
f36a9ecd PB |
4157 | len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed, |
4158 | super->sector_size); | |
4159 | if (posix_memalign(&buf, MAX_SECTOR_SIZE, len) != 0) | |
4d7b1503 DW |
4160 | return 1; |
4161 | ||
1f45a8ad DW |
4162 | memcpy(buf, super->buf, super->len); |
4163 | memset(buf + super->len, 0, len - super->len); | |
4d7b1503 DW |
4164 | free(super->buf); |
4165 | super->buf = buf; | |
4166 | super->len = len; | |
4167 | } | |
ca9de185 | 4168 | |
bbab0940 TM |
4169 | super->extra_space += space_needed; |
4170 | ||
cdddbdbc DW |
4171 | return 0; |
4172 | } | |
4173 | ||
e2f41b2c AK |
4174 | /******************************************************************************* |
4175 | * Function: check_mpb_migr_compatibility | |
4176 | * Description: Function checks for unsupported migration features: | |
4177 | * - migration optimization area (pba_of_lba0) | |
4178 | * - descending reshape (ascending_migr) | |
4179 | * Parameters: | |
4180 | * super : imsm metadata information | |
4181 | * Returns: | |
4182 | * 0 : migration is compatible | |
4183 | * -1 : migration is not compatible | |
4184 | ******************************************************************************/ | |
4185 | int check_mpb_migr_compatibility(struct intel_super *super) | |
4186 | { | |
4187 | struct imsm_map *map0, *map1; | |
4188 | struct migr_record *migr_rec = super->migr_rec; | |
4189 | int i; | |
4190 | ||
4191 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
4192 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
4193 | ||
4194 | if (dev_iter && | |
4195 | dev_iter->vol.migr_state == 1 && | |
4196 | dev_iter->vol.migr_type == MIGR_GEN_MIGR) { | |
4197 | /* This device is migrating */ | |
238c0a71 AK |
4198 | map0 = get_imsm_map(dev_iter, MAP_0); |
4199 | map1 = get_imsm_map(dev_iter, MAP_1); | |
5551b113 | 4200 | if (pba_of_lba0(map0) != pba_of_lba0(map1)) |
e2f41b2c AK |
4201 | /* migration optimization area was used */ |
4202 | return -1; | |
fc54fe7a JS |
4203 | if (migr_rec->ascending_migr == 0 && |
4204 | migr_rec->dest_depth_per_unit > 0) | |
e2f41b2c AK |
4205 | /* descending reshape not supported yet */ |
4206 | return -1; | |
4207 | } | |
4208 | } | |
4209 | return 0; | |
4210 | } | |
4211 | ||
d23fe947 | 4212 | static void __free_imsm(struct intel_super *super, int free_disks); |
9ca2c81c | 4213 | |
cdddbdbc | 4214 | /* load_imsm_mpb - read matrix metadata |
f2f5c343 | 4215 | * allocates super->mpb to be freed by free_imsm |
cdddbdbc DW |
4216 | */ |
4217 | static int load_imsm_mpb(int fd, struct intel_super *super, char *devname) | |
4218 | { | |
4219 | unsigned long long dsize; | |
cdddbdbc | 4220 | unsigned long long sectors; |
f36a9ecd | 4221 | unsigned int sector_size = super->sector_size; |
cdddbdbc | 4222 | struct stat; |
6416d527 | 4223 | struct imsm_super *anchor; |
cdddbdbc DW |
4224 | __u32 check_sum; |
4225 | ||
cdddbdbc | 4226 | get_dev_size(fd, NULL, &dsize); |
f36a9ecd | 4227 | if (dsize < 2*sector_size) { |
64436f06 | 4228 | if (devname) |
e7b84f9d N |
4229 | pr_err("%s: device to small for imsm\n", |
4230 | devname); | |
64436f06 N |
4231 | return 1; |
4232 | } | |
cdddbdbc | 4233 | |
f36a9ecd | 4234 | if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0) { |
cdddbdbc | 4235 | if (devname) |
e7b84f9d N |
4236 | pr_err("Cannot seek to anchor block on %s: %s\n", |
4237 | devname, strerror(errno)); | |
cdddbdbc DW |
4238 | return 1; |
4239 | } | |
4240 | ||
f36a9ecd | 4241 | if (posix_memalign((void **)&anchor, sector_size, sector_size) != 0) { |
ad97895e | 4242 | if (devname) |
7a862a02 | 4243 | pr_err("Failed to allocate imsm anchor buffer on %s\n", devname); |
ad97895e DW |
4244 | return 1; |
4245 | } | |
466070ad | 4246 | if ((unsigned int)read(fd, anchor, sector_size) != sector_size) { |
cdddbdbc | 4247 | if (devname) |
e7b84f9d N |
4248 | pr_err("Cannot read anchor block on %s: %s\n", |
4249 | devname, strerror(errno)); | |
6416d527 | 4250 | free(anchor); |
cdddbdbc DW |
4251 | return 1; |
4252 | } | |
4253 | ||
6416d527 | 4254 | if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) { |
cdddbdbc | 4255 | if (devname) |
e7b84f9d | 4256 | pr_err("no IMSM anchor on %s\n", devname); |
6416d527 | 4257 | free(anchor); |
cdddbdbc DW |
4258 | return 2; |
4259 | } | |
4260 | ||
d23fe947 | 4261 | __free_imsm(super, 0); |
f2f5c343 LM |
4262 | /* reload capability and hba */ |
4263 | ||
4264 | /* capability and hba must be updated with new super allocation */ | |
d424212e | 4265 | find_intel_hba_capability(fd, super, devname); |
f36a9ecd PB |
4266 | super->len = ROUND_UP(anchor->mpb_size, sector_size); |
4267 | if (posix_memalign(&super->buf, MAX_SECTOR_SIZE, super->len) != 0) { | |
cdddbdbc | 4268 | if (devname) |
e7b84f9d N |
4269 | pr_err("unable to allocate %zu byte mpb buffer\n", |
4270 | super->len); | |
6416d527 | 4271 | free(anchor); |
cdddbdbc DW |
4272 | return 2; |
4273 | } | |
f36a9ecd | 4274 | memcpy(super->buf, anchor, sector_size); |
cdddbdbc | 4275 | |
f36a9ecd | 4276 | sectors = mpb_sectors(anchor, sector_size) - 1; |
6416d527 | 4277 | free(anchor); |
8e59f3d8 | 4278 | |
85337573 AO |
4279 | if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE, |
4280 | MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) { | |
1ade5cc1 | 4281 | pr_err("could not allocate migr_rec buffer\n"); |
8e59f3d8 AK |
4282 | free(super->buf); |
4283 | return 2; | |
4284 | } | |
51d83f5d | 4285 | super->clean_migration_record_by_mdmon = 0; |
8e59f3d8 | 4286 | |
949c47a0 | 4287 | if (!sectors) { |
ecf45690 DW |
4288 | check_sum = __gen_imsm_checksum(super->anchor); |
4289 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
4290 | if (devname) | |
e7b84f9d N |
4291 | pr_err("IMSM checksum %x != %x on %s\n", |
4292 | check_sum, | |
4293 | __le32_to_cpu(super->anchor->check_sum), | |
4294 | devname); | |
ecf45690 DW |
4295 | return 2; |
4296 | } | |
4297 | ||
a2b97981 | 4298 | return 0; |
949c47a0 | 4299 | } |
cdddbdbc DW |
4300 | |
4301 | /* read the extended mpb */ | |
f36a9ecd | 4302 | if (lseek64(fd, dsize - (sector_size * (2 + sectors)), SEEK_SET) < 0) { |
cdddbdbc | 4303 | if (devname) |
e7b84f9d N |
4304 | pr_err("Cannot seek to extended mpb on %s: %s\n", |
4305 | devname, strerror(errno)); | |
cdddbdbc DW |
4306 | return 1; |
4307 | } | |
4308 | ||
f36a9ecd PB |
4309 | if ((unsigned int)read(fd, super->buf + sector_size, |
4310 | super->len - sector_size) != super->len - sector_size) { | |
cdddbdbc | 4311 | if (devname) |
e7b84f9d N |
4312 | pr_err("Cannot read extended mpb on %s: %s\n", |
4313 | devname, strerror(errno)); | |
cdddbdbc DW |
4314 | return 2; |
4315 | } | |
4316 | ||
949c47a0 DW |
4317 | check_sum = __gen_imsm_checksum(super->anchor); |
4318 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
cdddbdbc | 4319 | if (devname) |
e7b84f9d N |
4320 | pr_err("IMSM checksum %x != %x on %s\n", |
4321 | check_sum, __le32_to_cpu(super->anchor->check_sum), | |
4322 | devname); | |
db575f3b | 4323 | return 3; |
cdddbdbc DW |
4324 | } |
4325 | ||
a2b97981 DW |
4326 | return 0; |
4327 | } | |
4328 | ||
8e59f3d8 AK |
4329 | static int read_imsm_migr_rec(int fd, struct intel_super *super); |
4330 | ||
97f81ee2 CA |
4331 | /* clears hi bits in metadata if MPB_ATTRIB_2TB_DISK not set */ |
4332 | static void clear_hi(struct intel_super *super) | |
4333 | { | |
4334 | struct imsm_super *mpb = super->anchor; | |
4335 | int i, n; | |
4336 | if (mpb->attributes & MPB_ATTRIB_2TB_DISK) | |
4337 | return; | |
4338 | for (i = 0; i < mpb->num_disks; ++i) { | |
4339 | struct imsm_disk *disk = &mpb->disk[i]; | |
4340 | disk->total_blocks_hi = 0; | |
4341 | } | |
4342 | for (i = 0; i < mpb->num_raid_devs; ++i) { | |
4343 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
4344 | if (!dev) | |
4345 | return; | |
4346 | for (n = 0; n < 2; ++n) { | |
4347 | struct imsm_map *map = get_imsm_map(dev, n); | |
4348 | if (!map) | |
4349 | continue; | |
4350 | map->pba_of_lba0_hi = 0; | |
4351 | map->blocks_per_member_hi = 0; | |
4352 | map->num_data_stripes_hi = 0; | |
4353 | } | |
4354 | } | |
4355 | } | |
4356 | ||
a2b97981 DW |
4357 | static int |
4358 | load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd) | |
4359 | { | |
4360 | int err; | |
4361 | ||
4362 | err = load_imsm_mpb(fd, super, devname); | |
4363 | if (err) | |
4364 | return err; | |
f36a9ecd PB |
4365 | if (super->sector_size == 4096) |
4366 | convert_from_4k(super); | |
a2b97981 DW |
4367 | err = load_imsm_disk(fd, super, devname, keep_fd); |
4368 | if (err) | |
4369 | return err; | |
4370 | err = parse_raid_devices(super); | |
8d67477f TM |
4371 | if (err) |
4372 | return err; | |
4373 | err = load_bbm_log(super); | |
97f81ee2 | 4374 | clear_hi(super); |
a2b97981 | 4375 | return err; |
cdddbdbc DW |
4376 | } |
4377 | ||
ae6aad82 DW |
4378 | static void __free_imsm_disk(struct dl *d) |
4379 | { | |
4380 | if (d->fd >= 0) | |
4381 | close(d->fd); | |
4382 | if (d->devname) | |
4383 | free(d->devname); | |
0dcecb2e DW |
4384 | if (d->e) |
4385 | free(d->e); | |
ae6aad82 DW |
4386 | free(d); |
4387 | ||
4388 | } | |
1a64be56 | 4389 | |
cdddbdbc DW |
4390 | static void free_imsm_disks(struct intel_super *super) |
4391 | { | |
47ee5a45 | 4392 | struct dl *d; |
cdddbdbc | 4393 | |
47ee5a45 DW |
4394 | while (super->disks) { |
4395 | d = super->disks; | |
cdddbdbc | 4396 | super->disks = d->next; |
ae6aad82 | 4397 | __free_imsm_disk(d); |
cdddbdbc | 4398 | } |
cb82edca AK |
4399 | while (super->disk_mgmt_list) { |
4400 | d = super->disk_mgmt_list; | |
4401 | super->disk_mgmt_list = d->next; | |
4402 | __free_imsm_disk(d); | |
4403 | } | |
47ee5a45 DW |
4404 | while (super->missing) { |
4405 | d = super->missing; | |
4406 | super->missing = d->next; | |
4407 | __free_imsm_disk(d); | |
4408 | } | |
4409 | ||
cdddbdbc DW |
4410 | } |
4411 | ||
9ca2c81c | 4412 | /* free all the pieces hanging off of a super pointer */ |
d23fe947 | 4413 | static void __free_imsm(struct intel_super *super, int free_disks) |
cdddbdbc | 4414 | { |
88654014 LM |
4415 | struct intel_hba *elem, *next; |
4416 | ||
9ca2c81c | 4417 | if (super->buf) { |
949c47a0 | 4418 | free(super->buf); |
9ca2c81c DW |
4419 | super->buf = NULL; |
4420 | } | |
f2f5c343 LM |
4421 | /* unlink capability description */ |
4422 | super->orom = NULL; | |
8e59f3d8 AK |
4423 | if (super->migr_rec_buf) { |
4424 | free(super->migr_rec_buf); | |
4425 | super->migr_rec_buf = NULL; | |
4426 | } | |
d23fe947 DW |
4427 | if (free_disks) |
4428 | free_imsm_disks(super); | |
ba2de7ba | 4429 | free_devlist(super); |
88654014 LM |
4430 | elem = super->hba; |
4431 | while (elem) { | |
4432 | if (elem->path) | |
4433 | free((void *)elem->path); | |
4434 | next = elem->next; | |
4435 | free(elem); | |
4436 | elem = next; | |
88c32bb1 | 4437 | } |
8d67477f TM |
4438 | if (super->bbm_log) |
4439 | free(super->bbm_log); | |
88654014 | 4440 | super->hba = NULL; |
cdddbdbc DW |
4441 | } |
4442 | ||
9ca2c81c DW |
4443 | static void free_imsm(struct intel_super *super) |
4444 | { | |
d23fe947 | 4445 | __free_imsm(super, 1); |
928f1424 | 4446 | free(super->bb.entries); |
9ca2c81c DW |
4447 | free(super); |
4448 | } | |
cdddbdbc DW |
4449 | |
4450 | static void free_super_imsm(struct supertype *st) | |
4451 | { | |
4452 | struct intel_super *super = st->sb; | |
4453 | ||
4454 | if (!super) | |
4455 | return; | |
4456 | ||
4457 | free_imsm(super); | |
4458 | st->sb = NULL; | |
4459 | } | |
4460 | ||
49133e57 | 4461 | static struct intel_super *alloc_super(void) |
c2c087e6 | 4462 | { |
503975b9 | 4463 | struct intel_super *super = xcalloc(1, sizeof(*super)); |
c2c087e6 | 4464 | |
503975b9 N |
4465 | super->current_vol = -1; |
4466 | super->create_offset = ~((unsigned long long) 0); | |
928f1424 TM |
4467 | |
4468 | super->bb.entries = xmalloc(BBM_LOG_MAX_ENTRIES * | |
4469 | sizeof(struct md_bb_entry)); | |
4470 | if (!super->bb.entries) { | |
4471 | free(super); | |
4472 | return NULL; | |
4473 | } | |
4474 | ||
c2c087e6 DW |
4475 | return super; |
4476 | } | |
4477 | ||
f0f5a016 LM |
4478 | /* |
4479 | * find and allocate hba and OROM/EFI based on valid fd of RAID component device | |
4480 | */ | |
d424212e | 4481 | static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname) |
f0f5a016 LM |
4482 | { |
4483 | struct sys_dev *hba_name; | |
4484 | int rv = 0; | |
4485 | ||
089f9d79 | 4486 | if (fd < 0 || check_env("IMSM_NO_PLATFORM")) { |
f2f5c343 | 4487 | super->orom = NULL; |
f0f5a016 LM |
4488 | super->hba = NULL; |
4489 | return 0; | |
4490 | } | |
4491 | hba_name = find_disk_attached_hba(fd, NULL); | |
4492 | if (!hba_name) { | |
d424212e | 4493 | if (devname) |
e7b84f9d N |
4494 | pr_err("%s is not attached to Intel(R) RAID controller.\n", |
4495 | devname); | |
f0f5a016 LM |
4496 | return 1; |
4497 | } | |
4498 | rv = attach_hba_to_super(super, hba_name); | |
4499 | if (rv == 2) { | |
d424212e N |
4500 | if (devname) { |
4501 | struct intel_hba *hba = super->hba; | |
f0f5a016 | 4502 | |
60f0f54d PB |
4503 | pr_err("%s is attached to Intel(R) %s %s (%s),\n" |
4504 | " but the container is assigned to Intel(R) %s %s (", | |
d424212e | 4505 | devname, |
614902f6 | 4506 | get_sys_dev_type(hba_name->type), |
60f0f54d | 4507 | hba_name->type == SYS_DEV_VMD ? "domain" : "RAID controller", |
f0f5a016 | 4508 | hba_name->pci_id ? : "Err!", |
60f0f54d PB |
4509 | get_sys_dev_type(super->hba->type), |
4510 | hba->type == SYS_DEV_VMD ? "domain" : "RAID controller"); | |
f0f5a016 | 4511 | |
f0f5a016 LM |
4512 | while (hba) { |
4513 | fprintf(stderr, "%s", hba->pci_id ? : "Err!"); | |
4514 | if (hba->next) | |
4515 | fprintf(stderr, ", "); | |
4516 | hba = hba->next; | |
4517 | } | |
6b781d33 | 4518 | fprintf(stderr, ").\n" |
cca67208 | 4519 | " Mixing devices attached to different controllers is not allowed.\n"); |
f0f5a016 | 4520 | } |
f0f5a016 LM |
4521 | return 2; |
4522 | } | |
6b781d33 | 4523 | super->orom = find_imsm_capability(hba_name); |
f2f5c343 LM |
4524 | if (!super->orom) |
4525 | return 3; | |
614902f6 | 4526 | |
f0f5a016 LM |
4527 | return 0; |
4528 | } | |
4529 | ||
47ee5a45 DW |
4530 | /* find_missing - helper routine for load_super_imsm_all that identifies |
4531 | * disks that have disappeared from the system. This routine relies on | |
4532 | * the mpb being uptodate, which it is at load time. | |
4533 | */ | |
4534 | static int find_missing(struct intel_super *super) | |
4535 | { | |
4536 | int i; | |
4537 | struct imsm_super *mpb = super->anchor; | |
4538 | struct dl *dl; | |
4539 | struct imsm_disk *disk; | |
47ee5a45 DW |
4540 | |
4541 | for (i = 0; i < mpb->num_disks; i++) { | |
4542 | disk = __get_imsm_disk(mpb, i); | |
54c2c1ea | 4543 | dl = serial_to_dl(disk->serial, super); |
47ee5a45 DW |
4544 | if (dl) |
4545 | continue; | |
47ee5a45 | 4546 | |
503975b9 | 4547 | dl = xmalloc(sizeof(*dl)); |
47ee5a45 DW |
4548 | dl->major = 0; |
4549 | dl->minor = 0; | |
4550 | dl->fd = -1; | |
503975b9 | 4551 | dl->devname = xstrdup("missing"); |
47ee5a45 DW |
4552 | dl->index = i; |
4553 | serialcpy(dl->serial, disk->serial); | |
4554 | dl->disk = *disk; | |
689c9bf3 | 4555 | dl->e = NULL; |
47ee5a45 DW |
4556 | dl->next = super->missing; |
4557 | super->missing = dl; | |
4558 | } | |
4559 | ||
4560 | return 0; | |
4561 | } | |
4562 | ||
a2b97981 DW |
4563 | static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list) |
4564 | { | |
4565 | struct intel_disk *idisk = disk_list; | |
4566 | ||
4567 | while (idisk) { | |
4568 | if (serialcmp(idisk->disk.serial, serial) == 0) | |
4569 | break; | |
4570 | idisk = idisk->next; | |
4571 | } | |
4572 | ||
4573 | return idisk; | |
4574 | } | |
4575 | ||
4576 | static int __prep_thunderdome(struct intel_super **table, int tbl_size, | |
4577 | struct intel_super *super, | |
4578 | struct intel_disk **disk_list) | |
4579 | { | |
4580 | struct imsm_disk *d = &super->disks->disk; | |
4581 | struct imsm_super *mpb = super->anchor; | |
4582 | int i, j; | |
4583 | ||
4584 | for (i = 0; i < tbl_size; i++) { | |
4585 | struct imsm_super *tbl_mpb = table[i]->anchor; | |
4586 | struct imsm_disk *tbl_d = &table[i]->disks->disk; | |
4587 | ||
4588 | if (tbl_mpb->family_num == mpb->family_num) { | |
4589 | if (tbl_mpb->check_sum == mpb->check_sum) { | |
1ade5cc1 N |
4590 | dprintf("mpb from %d:%d matches %d:%d\n", |
4591 | super->disks->major, | |
a2b97981 DW |
4592 | super->disks->minor, |
4593 | table[i]->disks->major, | |
4594 | table[i]->disks->minor); | |
4595 | break; | |
4596 | } | |
4597 | ||
4598 | if (((is_configured(d) && !is_configured(tbl_d)) || | |
4599 | is_configured(d) == is_configured(tbl_d)) && | |
4600 | tbl_mpb->generation_num < mpb->generation_num) { | |
4601 | /* current version of the mpb is a | |
4602 | * better candidate than the one in | |
4603 | * super_table, but copy over "cross | |
4604 | * generational" status | |
4605 | */ | |
4606 | struct intel_disk *idisk; | |
4607 | ||
1ade5cc1 N |
4608 | dprintf("mpb from %d:%d replaces %d:%d\n", |
4609 | super->disks->major, | |
a2b97981 DW |
4610 | super->disks->minor, |
4611 | table[i]->disks->major, | |
4612 | table[i]->disks->minor); | |
4613 | ||
4614 | idisk = disk_list_get(tbl_d->serial, *disk_list); | |
4615 | if (idisk && is_failed(&idisk->disk)) | |
4616 | tbl_d->status |= FAILED_DISK; | |
4617 | break; | |
4618 | } else { | |
4619 | struct intel_disk *idisk; | |
4620 | struct imsm_disk *disk; | |
4621 | ||
4622 | /* tbl_mpb is more up to date, but copy | |
4623 | * over cross generational status before | |
4624 | * returning | |
4625 | */ | |
4626 | disk = __serial_to_disk(d->serial, mpb, NULL); | |
4627 | if (disk && is_failed(disk)) | |
4628 | d->status |= FAILED_DISK; | |
4629 | ||
4630 | idisk = disk_list_get(d->serial, *disk_list); | |
4631 | if (idisk) { | |
4632 | idisk->owner = i; | |
4633 | if (disk && is_configured(disk)) | |
4634 | idisk->disk.status |= CONFIGURED_DISK; | |
4635 | } | |
4636 | ||
1ade5cc1 N |
4637 | dprintf("mpb from %d:%d prefer %d:%d\n", |
4638 | super->disks->major, | |
a2b97981 DW |
4639 | super->disks->minor, |
4640 | table[i]->disks->major, | |
4641 | table[i]->disks->minor); | |
4642 | ||
4643 | return tbl_size; | |
4644 | } | |
4645 | } | |
4646 | } | |
4647 | ||
4648 | if (i >= tbl_size) | |
4649 | table[tbl_size++] = super; | |
4650 | else | |
4651 | table[i] = super; | |
4652 | ||
4653 | /* update/extend the merged list of imsm_disk records */ | |
4654 | for (j = 0; j < mpb->num_disks; j++) { | |
4655 | struct imsm_disk *disk = __get_imsm_disk(mpb, j); | |
4656 | struct intel_disk *idisk; | |
4657 | ||
4658 | idisk = disk_list_get(disk->serial, *disk_list); | |
4659 | if (idisk) { | |
4660 | idisk->disk.status |= disk->status; | |
4661 | if (is_configured(&idisk->disk) || | |
4662 | is_failed(&idisk->disk)) | |
4663 | idisk->disk.status &= ~(SPARE_DISK); | |
4664 | } else { | |
503975b9 | 4665 | idisk = xcalloc(1, sizeof(*idisk)); |
a2b97981 DW |
4666 | idisk->owner = IMSM_UNKNOWN_OWNER; |
4667 | idisk->disk = *disk; | |
4668 | idisk->next = *disk_list; | |
4669 | *disk_list = idisk; | |
4670 | } | |
4671 | ||
4672 | if (serialcmp(idisk->disk.serial, d->serial) == 0) | |
4673 | idisk->owner = i; | |
4674 | } | |
4675 | ||
4676 | return tbl_size; | |
4677 | } | |
4678 | ||
4679 | static struct intel_super * | |
4680 | validate_members(struct intel_super *super, struct intel_disk *disk_list, | |
4681 | const int owner) | |
4682 | { | |
4683 | struct imsm_super *mpb = super->anchor; | |
4684 | int ok_count = 0; | |
4685 | int i; | |
4686 | ||
4687 | for (i = 0; i < mpb->num_disks; i++) { | |
4688 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
4689 | struct intel_disk *idisk; | |
4690 | ||
4691 | idisk = disk_list_get(disk->serial, disk_list); | |
4692 | if (idisk) { | |
4693 | if (idisk->owner == owner || | |
4694 | idisk->owner == IMSM_UNKNOWN_OWNER) | |
4695 | ok_count++; | |
4696 | else | |
1ade5cc1 N |
4697 | dprintf("'%.16s' owner %d != %d\n", |
4698 | disk->serial, idisk->owner, | |
a2b97981 DW |
4699 | owner); |
4700 | } else { | |
1ade5cc1 N |
4701 | dprintf("unknown disk %x [%d]: %.16s\n", |
4702 | __le32_to_cpu(mpb->family_num), i, | |
a2b97981 DW |
4703 | disk->serial); |
4704 | break; | |
4705 | } | |
4706 | } | |
4707 | ||
4708 | if (ok_count == mpb->num_disks) | |
4709 | return super; | |
4710 | return NULL; | |
4711 | } | |
4712 | ||
4713 | static void show_conflicts(__u32 family_num, struct intel_super *super_list) | |
4714 | { | |
4715 | struct intel_super *s; | |
4716 | ||
4717 | for (s = super_list; s; s = s->next) { | |
4718 | if (family_num != s->anchor->family_num) | |
4719 | continue; | |
e12b3daa | 4720 | pr_err("Conflict, offlining family %#x on '%s'\n", |
a2b97981 DW |
4721 | __le32_to_cpu(family_num), s->disks->devname); |
4722 | } | |
4723 | } | |
4724 | ||
4725 | static struct intel_super * | |
4726 | imsm_thunderdome(struct intel_super **super_list, int len) | |
4727 | { | |
4728 | struct intel_super *super_table[len]; | |
4729 | struct intel_disk *disk_list = NULL; | |
4730 | struct intel_super *champion, *spare; | |
4731 | struct intel_super *s, **del; | |
4732 | int tbl_size = 0; | |
4733 | int conflict; | |
4734 | int i; | |
4735 | ||
4736 | memset(super_table, 0, sizeof(super_table)); | |
4737 | for (s = *super_list; s; s = s->next) | |
4738 | tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list); | |
4739 | ||
4740 | for (i = 0; i < tbl_size; i++) { | |
4741 | struct imsm_disk *d; | |
4742 | struct intel_disk *idisk; | |
4743 | struct imsm_super *mpb = super_table[i]->anchor; | |
4744 | ||
4745 | s = super_table[i]; | |
4746 | d = &s->disks->disk; | |
4747 | ||
4748 | /* 'd' must appear in merged disk list for its | |
4749 | * configuration to be valid | |
4750 | */ | |
4751 | idisk = disk_list_get(d->serial, disk_list); | |
4752 | if (idisk && idisk->owner == i) | |
4753 | s = validate_members(s, disk_list, i); | |
4754 | else | |
4755 | s = NULL; | |
4756 | ||
4757 | if (!s) | |
1ade5cc1 N |
4758 | dprintf("marking family: %#x from %d:%d offline\n", |
4759 | mpb->family_num, | |
a2b97981 DW |
4760 | super_table[i]->disks->major, |
4761 | super_table[i]->disks->minor); | |
4762 | super_table[i] = s; | |
4763 | } | |
4764 | ||
4765 | /* This is where the mdadm implementation differs from the Windows | |
4766 | * driver which has no strict concept of a container. We can only | |
4767 | * assemble one family from a container, so when returning a prodigal | |
4768 | * array member to this system the code will not be able to disambiguate | |
4769 | * the container contents that should be assembled ("foreign" versus | |
4770 | * "local"). It requires user intervention to set the orig_family_num | |
4771 | * to a new value to establish a new container. The Windows driver in | |
4772 | * this situation fixes up the volume name in place and manages the | |
4773 | * foreign array as an independent entity. | |
4774 | */ | |
4775 | s = NULL; | |
4776 | spare = NULL; | |
4777 | conflict = 0; | |
4778 | for (i = 0; i < tbl_size; i++) { | |
4779 | struct intel_super *tbl_ent = super_table[i]; | |
4780 | int is_spare = 0; | |
4781 | ||
4782 | if (!tbl_ent) | |
4783 | continue; | |
4784 | ||
4785 | if (tbl_ent->anchor->num_raid_devs == 0) { | |
4786 | spare = tbl_ent; | |
4787 | is_spare = 1; | |
4788 | } | |
4789 | ||
4790 | if (s && !is_spare) { | |
4791 | show_conflicts(tbl_ent->anchor->family_num, *super_list); | |
4792 | conflict++; | |
4793 | } else if (!s && !is_spare) | |
4794 | s = tbl_ent; | |
4795 | } | |
4796 | ||
4797 | if (!s) | |
4798 | s = spare; | |
4799 | if (!s) { | |
4800 | champion = NULL; | |
4801 | goto out; | |
4802 | } | |
4803 | champion = s; | |
4804 | ||
4805 | if (conflict) | |
7a862a02 | 4806 | pr_err("Chose family %#x on '%s', assemble conflicts to new container with '--update=uuid'\n", |
a2b97981 DW |
4807 | __le32_to_cpu(s->anchor->family_num), s->disks->devname); |
4808 | ||
4809 | /* collect all dl's onto 'champion', and update them to | |
4810 | * champion's version of the status | |
4811 | */ | |
4812 | for (s = *super_list; s; s = s->next) { | |
4813 | struct imsm_super *mpb = champion->anchor; | |
4814 | struct dl *dl = s->disks; | |
4815 | ||
4816 | if (s == champion) | |
4817 | continue; | |
4818 | ||
5d7b407a CA |
4819 | mpb->attributes |= s->anchor->attributes & MPB_ATTRIB_2TB_DISK; |
4820 | ||
a2b97981 DW |
4821 | for (i = 0; i < mpb->num_disks; i++) { |
4822 | struct imsm_disk *disk; | |
4823 | ||
4824 | disk = __serial_to_disk(dl->serial, mpb, &dl->index); | |
4825 | if (disk) { | |
4826 | dl->disk = *disk; | |
4827 | /* only set index on disks that are a member of | |
4828 | * a populated contianer, i.e. one with | |
4829 | * raid_devs | |
4830 | */ | |
4831 | if (is_failed(&dl->disk)) | |
4832 | dl->index = -2; | |
4833 | else if (is_spare(&dl->disk)) | |
4834 | dl->index = -1; | |
4835 | break; | |
4836 | } | |
4837 | } | |
4838 | ||
4839 | if (i >= mpb->num_disks) { | |
4840 | struct intel_disk *idisk; | |
4841 | ||
4842 | idisk = disk_list_get(dl->serial, disk_list); | |
ecf408e9 | 4843 | if (idisk && is_spare(&idisk->disk) && |
a2b97981 DW |
4844 | !is_failed(&idisk->disk) && !is_configured(&idisk->disk)) |
4845 | dl->index = -1; | |
4846 | else { | |
4847 | dl->index = -2; | |
4848 | continue; | |
4849 | } | |
4850 | } | |
4851 | ||
4852 | dl->next = champion->disks; | |
4853 | champion->disks = dl; | |
4854 | s->disks = NULL; | |
4855 | } | |
4856 | ||
4857 | /* delete 'champion' from super_list */ | |
4858 | for (del = super_list; *del; ) { | |
4859 | if (*del == champion) { | |
4860 | *del = (*del)->next; | |
4861 | break; | |
4862 | } else | |
4863 | del = &(*del)->next; | |
4864 | } | |
4865 | champion->next = NULL; | |
4866 | ||
4867 | out: | |
4868 | while (disk_list) { | |
4869 | struct intel_disk *idisk = disk_list; | |
4870 | ||
4871 | disk_list = disk_list->next; | |
4872 | free(idisk); | |
4873 | } | |
4874 | ||
4875 | return champion; | |
4876 | } | |
4877 | ||
9587c373 LM |
4878 | static int |
4879 | get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd); | |
4dd2df09 | 4880 | static int get_super_block(struct intel_super **super_list, char *devnm, char *devname, |
9587c373 | 4881 | int major, int minor, int keep_fd); |
ec50f7b6 LM |
4882 | static int |
4883 | get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list, | |
4884 | int *max, int keep_fd); | |
4885 | ||
cdddbdbc | 4886 | static int load_super_imsm_all(struct supertype *st, int fd, void **sbp, |
ec50f7b6 LM |
4887 | char *devname, struct md_list *devlist, |
4888 | int keep_fd) | |
cdddbdbc | 4889 | { |
a2b97981 DW |
4890 | struct intel_super *super_list = NULL; |
4891 | struct intel_super *super = NULL; | |
a2b97981 | 4892 | int err = 0; |
9587c373 | 4893 | int i = 0; |
dab4a513 | 4894 | |
9587c373 LM |
4895 | if (fd >= 0) |
4896 | /* 'fd' is an opened container */ | |
4897 | err = get_sra_super_block(fd, &super_list, devname, &i, keep_fd); | |
4898 | else | |
ec50f7b6 LM |
4899 | /* get super block from devlist devices */ |
4900 | err = get_devlist_super_block(devlist, &super_list, &i, keep_fd); | |
9587c373 | 4901 | if (err) |
1602d52c | 4902 | goto error; |
a2b97981 DW |
4903 | /* all mpbs enter, maybe one leaves */ |
4904 | super = imsm_thunderdome(&super_list, i); | |
4905 | if (!super) { | |
4906 | err = 1; | |
4907 | goto error; | |
cdddbdbc DW |
4908 | } |
4909 | ||
47ee5a45 DW |
4910 | if (find_missing(super) != 0) { |
4911 | free_imsm(super); | |
a2b97981 DW |
4912 | err = 2; |
4913 | goto error; | |
47ee5a45 | 4914 | } |
8e59f3d8 AK |
4915 | |
4916 | /* load migration record */ | |
4917 | err = load_imsm_migr_rec(super, NULL); | |
4c965cc9 AK |
4918 | if (err == -1) { |
4919 | /* migration is in progress, | |
4920 | * but migr_rec cannot be loaded, | |
4921 | */ | |
8e59f3d8 AK |
4922 | err = 4; |
4923 | goto error; | |
4924 | } | |
e2f41b2c AK |
4925 | |
4926 | /* Check migration compatibility */ | |
089f9d79 | 4927 | if (err == 0 && check_mpb_migr_compatibility(super) != 0) { |
e7b84f9d | 4928 | pr_err("Unsupported migration detected"); |
e2f41b2c AK |
4929 | if (devname) |
4930 | fprintf(stderr, " on %s\n", devname); | |
4931 | else | |
4932 | fprintf(stderr, " (IMSM).\n"); | |
4933 | ||
4934 | err = 5; | |
4935 | goto error; | |
4936 | } | |
4937 | ||
a2b97981 DW |
4938 | err = 0; |
4939 | ||
4940 | error: | |
4941 | while (super_list) { | |
4942 | struct intel_super *s = super_list; | |
4943 | ||
4944 | super_list = super_list->next; | |
4945 | free_imsm(s); | |
4946 | } | |
9587c373 | 4947 | |
a2b97981 DW |
4948 | if (err) |
4949 | return err; | |
f7e7067b | 4950 | |
cdddbdbc | 4951 | *sbp = super; |
9587c373 | 4952 | if (fd >= 0) |
4dd2df09 | 4953 | strcpy(st->container_devnm, fd2devnm(fd)); |
9587c373 | 4954 | else |
4dd2df09 | 4955 | st->container_devnm[0] = 0; |
a2b97981 | 4956 | if (err == 0 && st->ss == NULL) { |
bf5a934a | 4957 | st->ss = &super_imsm; |
cdddbdbc DW |
4958 | st->minor_version = 0; |
4959 | st->max_devs = IMSM_MAX_DEVICES; | |
4960 | } | |
cdddbdbc DW |
4961 | return 0; |
4962 | } | |
2b959fbf | 4963 | |
ec50f7b6 LM |
4964 | static int |
4965 | get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list, | |
4966 | int *max, int keep_fd) | |
4967 | { | |
4968 | struct md_list *tmpdev; | |
4969 | int err = 0; | |
4970 | int i = 0; | |
9587c373 | 4971 | |
ec50f7b6 LM |
4972 | for (i = 0, tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { |
4973 | if (tmpdev->used != 1) | |
4974 | continue; | |
4975 | if (tmpdev->container == 1) { | |
ca9de185 | 4976 | int lmax = 0; |
ec50f7b6 LM |
4977 | int fd = dev_open(tmpdev->devname, O_RDONLY|O_EXCL); |
4978 | if (fd < 0) { | |
e7b84f9d | 4979 | pr_err("cannot open device %s: %s\n", |
ec50f7b6 LM |
4980 | tmpdev->devname, strerror(errno)); |
4981 | err = 8; | |
4982 | goto error; | |
4983 | } | |
4984 | err = get_sra_super_block(fd, super_list, | |
4985 | tmpdev->devname, &lmax, | |
4986 | keep_fd); | |
4987 | i += lmax; | |
4988 | close(fd); | |
4989 | if (err) { | |
4990 | err = 7; | |
4991 | goto error; | |
4992 | } | |
4993 | } else { | |
4994 | int major = major(tmpdev->st_rdev); | |
4995 | int minor = minor(tmpdev->st_rdev); | |
4996 | err = get_super_block(super_list, | |
4dd2df09 | 4997 | NULL, |
ec50f7b6 LM |
4998 | tmpdev->devname, |
4999 | major, minor, | |
5000 | keep_fd); | |
5001 | i++; | |
5002 | if (err) { | |
5003 | err = 6; | |
5004 | goto error; | |
5005 | } | |
5006 | } | |
5007 | } | |
5008 | error: | |
5009 | *max = i; | |
5010 | return err; | |
5011 | } | |
9587c373 | 5012 | |
4dd2df09 | 5013 | static int get_super_block(struct intel_super **super_list, char *devnm, char *devname, |
9587c373 LM |
5014 | int major, int minor, int keep_fd) |
5015 | { | |
594dc1b8 | 5016 | struct intel_super *s; |
9587c373 LM |
5017 | char nm[32]; |
5018 | int dfd = -1; | |
9587c373 LM |
5019 | int err = 0; |
5020 | int retry; | |
5021 | ||
5022 | s = alloc_super(); | |
5023 | if (!s) { | |
5024 | err = 1; | |
5025 | goto error; | |
5026 | } | |
5027 | ||
5028 | sprintf(nm, "%d:%d", major, minor); | |
5029 | dfd = dev_open(nm, O_RDWR); | |
5030 | if (dfd < 0) { | |
5031 | err = 2; | |
5032 | goto error; | |
5033 | } | |
5034 | ||
fa7bb6f8 | 5035 | get_dev_sector_size(dfd, NULL, &s->sector_size); |
cb8f6859 | 5036 | find_intel_hba_capability(dfd, s, devname); |
9587c373 LM |
5037 | err = load_and_parse_mpb(dfd, s, NULL, keep_fd); |
5038 | ||
5039 | /* retry the load if we might have raced against mdmon */ | |
4dd2df09 | 5040 | if (err == 3 && devnm && mdmon_running(devnm)) |
9587c373 LM |
5041 | for (retry = 0; retry < 3; retry++) { |
5042 | usleep(3000); | |
5043 | err = load_and_parse_mpb(dfd, s, NULL, keep_fd); | |
5044 | if (err != 3) | |
5045 | break; | |
5046 | } | |
5047 | error: | |
5048 | if (!err) { | |
5049 | s->next = *super_list; | |
5050 | *super_list = s; | |
5051 | } else { | |
5052 | if (s) | |
8d67477f | 5053 | free_imsm(s); |
36614e95 | 5054 | if (dfd >= 0) |
9587c373 LM |
5055 | close(dfd); |
5056 | } | |
089f9d79 | 5057 | if (dfd >= 0 && !keep_fd) |
9587c373 LM |
5058 | close(dfd); |
5059 | return err; | |
5060 | ||
5061 | } | |
5062 | ||
5063 | static int | |
5064 | get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd) | |
5065 | { | |
5066 | struct mdinfo *sra; | |
4dd2df09 | 5067 | char *devnm; |
9587c373 LM |
5068 | struct mdinfo *sd; |
5069 | int err = 0; | |
5070 | int i = 0; | |
4dd2df09 | 5071 | sra = sysfs_read(fd, NULL, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE); |
9587c373 LM |
5072 | if (!sra) |
5073 | return 1; | |
5074 | ||
5075 | if (sra->array.major_version != -1 || | |
5076 | sra->array.minor_version != -2 || | |
5077 | strcmp(sra->text_version, "imsm") != 0) { | |
5078 | err = 1; | |
5079 | goto error; | |
5080 | } | |
5081 | /* load all mpbs */ | |
4dd2df09 | 5082 | devnm = fd2devnm(fd); |
9587c373 | 5083 | for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) { |
4dd2df09 | 5084 | if (get_super_block(super_list, devnm, devname, |
9587c373 LM |
5085 | sd->disk.major, sd->disk.minor, keep_fd) != 0) { |
5086 | err = 7; | |
5087 | goto error; | |
5088 | } | |
5089 | } | |
5090 | error: | |
5091 | sysfs_free(sra); | |
5092 | *max = i; | |
5093 | return err; | |
5094 | } | |
5095 | ||
2b959fbf N |
5096 | static int load_container_imsm(struct supertype *st, int fd, char *devname) |
5097 | { | |
ec50f7b6 | 5098 | return load_super_imsm_all(st, fd, &st->sb, devname, NULL, 1); |
2b959fbf | 5099 | } |
cdddbdbc DW |
5100 | |
5101 | static int load_super_imsm(struct supertype *st, int fd, char *devname) | |
5102 | { | |
5103 | struct intel_super *super; | |
5104 | int rv; | |
8a3544f8 | 5105 | int retry; |
cdddbdbc | 5106 | |
357ac106 | 5107 | if (test_partition(fd)) |
691c6ee1 N |
5108 | /* IMSM not allowed on partitions */ |
5109 | return 1; | |
5110 | ||
37424f13 DW |
5111 | free_super_imsm(st); |
5112 | ||
49133e57 | 5113 | super = alloc_super(); |
fa7bb6f8 | 5114 | get_dev_sector_size(fd, NULL, &super->sector_size); |
8d67477f TM |
5115 | if (!super) |
5116 | return 1; | |
ea2bc72b LM |
5117 | /* Load hba and capabilities if they exist. |
5118 | * But do not preclude loading metadata in case capabilities or hba are | |
5119 | * non-compliant and ignore_hw_compat is set. | |
5120 | */ | |
d424212e | 5121 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 5122 | /* no orom/efi or non-intel hba of the disk */ |
089f9d79 | 5123 | if (rv != 0 && st->ignore_hw_compat == 0) { |
f2f5c343 | 5124 | if (devname) |
e7b84f9d | 5125 | pr_err("No OROM/EFI properties for %s\n", devname); |
f2f5c343 LM |
5126 | free_imsm(super); |
5127 | return 2; | |
5128 | } | |
a2b97981 | 5129 | rv = load_and_parse_mpb(fd, super, devname, 0); |
cdddbdbc | 5130 | |
8a3544f8 AP |
5131 | /* retry the load if we might have raced against mdmon */ |
5132 | if (rv == 3) { | |
f96b1302 AP |
5133 | struct mdstat_ent *mdstat = NULL; |
5134 | char *name = fd2kname(fd); | |
5135 | ||
5136 | if (name) | |
5137 | mdstat = mdstat_by_component(name); | |
8a3544f8 AP |
5138 | |
5139 | if (mdstat && mdmon_running(mdstat->devnm) && getpid() != mdmon_pid(mdstat->devnm)) { | |
5140 | for (retry = 0; retry < 3; retry++) { | |
5141 | usleep(3000); | |
5142 | rv = load_and_parse_mpb(fd, super, devname, 0); | |
5143 | if (rv != 3) | |
5144 | break; | |
5145 | } | |
5146 | } | |
5147 | ||
5148 | free_mdstat(mdstat); | |
5149 | } | |
5150 | ||
cdddbdbc DW |
5151 | if (rv) { |
5152 | if (devname) | |
7a862a02 | 5153 | pr_err("Failed to load all information sections on %s\n", devname); |
cdddbdbc DW |
5154 | free_imsm(super); |
5155 | return rv; | |
5156 | } | |
5157 | ||
5158 | st->sb = super; | |
5159 | if (st->ss == NULL) { | |
5160 | st->ss = &super_imsm; | |
5161 | st->minor_version = 0; | |
5162 | st->max_devs = IMSM_MAX_DEVICES; | |
5163 | } | |
8e59f3d8 AK |
5164 | |
5165 | /* load migration record */ | |
2e062e82 AK |
5166 | if (load_imsm_migr_rec(super, NULL) == 0) { |
5167 | /* Check for unsupported migration features */ | |
5168 | if (check_mpb_migr_compatibility(super) != 0) { | |
e7b84f9d | 5169 | pr_err("Unsupported migration detected"); |
2e062e82 AK |
5170 | if (devname) |
5171 | fprintf(stderr, " on %s\n", devname); | |
5172 | else | |
5173 | fprintf(stderr, " (IMSM).\n"); | |
5174 | return 3; | |
5175 | } | |
e2f41b2c AK |
5176 | } |
5177 | ||
cdddbdbc DW |
5178 | return 0; |
5179 | } | |
5180 | ||
ef6ffade DW |
5181 | static __u16 info_to_blocks_per_strip(mdu_array_info_t *info) |
5182 | { | |
5183 | if (info->level == 1) | |
5184 | return 128; | |
5185 | return info->chunk_size >> 9; | |
5186 | } | |
5187 | ||
5551b113 CA |
5188 | static unsigned long long info_to_blocks_per_member(mdu_array_info_t *info, |
5189 | unsigned long long size) | |
fcfd9599 | 5190 | { |
4025c288 | 5191 | if (info->level == 1) |
5551b113 | 5192 | return size * 2; |
4025c288 | 5193 | else |
5551b113 | 5194 | return (size * 2) & ~(info_to_blocks_per_strip(info) - 1); |
fcfd9599 DW |
5195 | } |
5196 | ||
4d1313e9 DW |
5197 | static void imsm_update_version_info(struct intel_super *super) |
5198 | { | |
5199 | /* update the version and attributes */ | |
5200 | struct imsm_super *mpb = super->anchor; | |
5201 | char *version; | |
5202 | struct imsm_dev *dev; | |
5203 | struct imsm_map *map; | |
5204 | int i; | |
5205 | ||
5206 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
5207 | dev = get_imsm_dev(super, i); | |
238c0a71 | 5208 | map = get_imsm_map(dev, MAP_0); |
4d1313e9 DW |
5209 | if (__le32_to_cpu(dev->size_high) > 0) |
5210 | mpb->attributes |= MPB_ATTRIB_2TB; | |
5211 | ||
5212 | /* FIXME detect when an array spans a port multiplier */ | |
5213 | #if 0 | |
5214 | mpb->attributes |= MPB_ATTRIB_PM; | |
5215 | #endif | |
5216 | ||
5217 | if (mpb->num_raid_devs > 1 || | |
5218 | mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) { | |
5219 | version = MPB_VERSION_ATTRIBS; | |
5220 | switch (get_imsm_raid_level(map)) { | |
5221 | case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break; | |
5222 | case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break; | |
5223 | case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break; | |
5224 | case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break; | |
5225 | } | |
5226 | } else { | |
5227 | if (map->num_members >= 5) | |
5228 | version = MPB_VERSION_5OR6_DISK_ARRAY; | |
5229 | else if (dev->status == DEV_CLONE_N_GO) | |
5230 | version = MPB_VERSION_CNG; | |
5231 | else if (get_imsm_raid_level(map) == 5) | |
5232 | version = MPB_VERSION_RAID5; | |
5233 | else if (map->num_members >= 3) | |
5234 | version = MPB_VERSION_3OR4_DISK_ARRAY; | |
5235 | else if (get_imsm_raid_level(map) == 1) | |
5236 | version = MPB_VERSION_RAID1; | |
5237 | else | |
5238 | version = MPB_VERSION_RAID0; | |
5239 | } | |
5240 | strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version); | |
5241 | } | |
5242 | } | |
5243 | ||
aa534678 DW |
5244 | static int check_name(struct intel_super *super, char *name, int quiet) |
5245 | { | |
5246 | struct imsm_super *mpb = super->anchor; | |
5247 | char *reason = NULL; | |
5248 | int i; | |
5249 | ||
5250 | if (strlen(name) > MAX_RAID_SERIAL_LEN) | |
5251 | reason = "must be 16 characters or less"; | |
5252 | ||
5253 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
5254 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
5255 | ||
5256 | if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) { | |
5257 | reason = "already exists"; | |
5258 | break; | |
5259 | } | |
5260 | } | |
5261 | ||
5262 | if (reason && !quiet) | |
e7b84f9d | 5263 | pr_err("imsm volume name %s\n", reason); |
aa534678 DW |
5264 | |
5265 | return !reason; | |
5266 | } | |
5267 | ||
8b353278 | 5268 | static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info, |
5308f117 | 5269 | struct shape *s, char *name, |
83cd1e97 N |
5270 | char *homehost, int *uuid, |
5271 | long long data_offset) | |
cdddbdbc | 5272 | { |
c2c087e6 DW |
5273 | /* We are creating a volume inside a pre-existing container. |
5274 | * so st->sb is already set. | |
5275 | */ | |
5276 | struct intel_super *super = st->sb; | |
f36a9ecd | 5277 | unsigned int sector_size = super->sector_size; |
949c47a0 | 5278 | struct imsm_super *mpb = super->anchor; |
ba2de7ba | 5279 | struct intel_dev *dv; |
c2c087e6 DW |
5280 | struct imsm_dev *dev; |
5281 | struct imsm_vol *vol; | |
5282 | struct imsm_map *map; | |
5283 | int idx = mpb->num_raid_devs; | |
5284 | int i; | |
5285 | unsigned long long array_blocks; | |
2c092cad | 5286 | size_t size_old, size_new; |
5551b113 | 5287 | unsigned long long num_data_stripes; |
b53bfba6 TM |
5288 | unsigned int data_disks; |
5289 | unsigned long long size_per_member; | |
cdddbdbc | 5290 | |
88c32bb1 | 5291 | if (super->orom && mpb->num_raid_devs >= super->orom->vpa) { |
7a862a02 | 5292 | pr_err("This imsm-container already has the maximum of %d volumes\n", super->orom->vpa); |
c2c087e6 DW |
5293 | return 0; |
5294 | } | |
5295 | ||
2c092cad DW |
5296 | /* ensure the mpb is large enough for the new data */ |
5297 | size_old = __le32_to_cpu(mpb->mpb_size); | |
5298 | size_new = disks_to_mpb_size(info->nr_disks); | |
5299 | if (size_new > size_old) { | |
5300 | void *mpb_new; | |
f36a9ecd | 5301 | size_t size_round = ROUND_UP(size_new, sector_size); |
2c092cad | 5302 | |
f36a9ecd | 5303 | if (posix_memalign(&mpb_new, sector_size, size_round) != 0) { |
e7b84f9d | 5304 | pr_err("could not allocate new mpb\n"); |
2c092cad DW |
5305 | return 0; |
5306 | } | |
85337573 AO |
5307 | if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE, |
5308 | MIGR_REC_BUF_SECTORS* | |
5309 | MAX_SECTOR_SIZE) != 0) { | |
1ade5cc1 | 5310 | pr_err("could not allocate migr_rec buffer\n"); |
8e59f3d8 AK |
5311 | free(super->buf); |
5312 | free(super); | |
ea944c8f | 5313 | free(mpb_new); |
8e59f3d8 AK |
5314 | return 0; |
5315 | } | |
2c092cad DW |
5316 | memcpy(mpb_new, mpb, size_old); |
5317 | free(mpb); | |
5318 | mpb = mpb_new; | |
949c47a0 | 5319 | super->anchor = mpb_new; |
2c092cad DW |
5320 | mpb->mpb_size = __cpu_to_le32(size_new); |
5321 | memset(mpb_new + size_old, 0, size_round - size_old); | |
bbab0940 | 5322 | super->len = size_round; |
2c092cad | 5323 | } |
bf5a934a | 5324 | super->current_vol = idx; |
3960e579 DW |
5325 | |
5326 | /* handle 'failed_disks' by either: | |
5327 | * a) create dummy disk entries in the table if this the first | |
5328 | * volume in the array. We add them here as this is the only | |
5329 | * opportunity to add them. add_to_super_imsm_volume() | |
5330 | * handles the non-failed disks and continues incrementing | |
5331 | * mpb->num_disks. | |
5332 | * b) validate that 'failed_disks' matches the current number | |
5333 | * of missing disks if the container is populated | |
d23fe947 | 5334 | */ |
3960e579 | 5335 | if (super->current_vol == 0) { |
d23fe947 | 5336 | mpb->num_disks = 0; |
3960e579 DW |
5337 | for (i = 0; i < info->failed_disks; i++) { |
5338 | struct imsm_disk *disk; | |
5339 | ||
5340 | mpb->num_disks++; | |
5341 | disk = __get_imsm_disk(mpb, i); | |
5342 | disk->status = CONFIGURED_DISK | FAILED_DISK; | |
5343 | disk->scsi_id = __cpu_to_le32(~(__u32)0); | |
5344 | snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN, | |
5b975129 | 5345 | "missing:%d", (__u8)i); |
3960e579 DW |
5346 | } |
5347 | find_missing(super); | |
5348 | } else { | |
5349 | int missing = 0; | |
5350 | struct dl *d; | |
5351 | ||
5352 | for (d = super->missing; d; d = d->next) | |
5353 | missing++; | |
5354 | if (info->failed_disks > missing) { | |
e7b84f9d | 5355 | pr_err("unable to add 'missing' disk to container\n"); |
3960e579 DW |
5356 | return 0; |
5357 | } | |
5358 | } | |
5a038140 | 5359 | |
aa534678 DW |
5360 | if (!check_name(super, name, 0)) |
5361 | return 0; | |
503975b9 N |
5362 | dv = xmalloc(sizeof(*dv)); |
5363 | dev = xcalloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1)); | |
c2c087e6 | 5364 | strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN); |
e03640bd | 5365 | array_blocks = calc_array_size(info->level, info->raid_disks, |
03bcbc65 | 5366 | info->layout, info->chunk_size, |
b53bfba6 TM |
5367 | s->size * BLOCKS_PER_KB); |
5368 | data_disks = get_data_disks(info->level, info->layout, | |
5369 | info->raid_disks); | |
5370 | array_blocks = round_size_to_mb(array_blocks, data_disks); | |
5371 | size_per_member = array_blocks / data_disks; | |
979d38be | 5372 | |
c2c087e6 DW |
5373 | dev->size_low = __cpu_to_le32((__u32) array_blocks); |
5374 | dev->size_high = __cpu_to_le32((__u32) (array_blocks >> 32)); | |
1a2487c2 | 5375 | dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING); |
c2c087e6 DW |
5376 | vol = &dev->vol; |
5377 | vol->migr_state = 0; | |
1484e727 | 5378 | set_migr_type(dev, MIGR_INIT); |
3960e579 | 5379 | vol->dirty = !info->state; |
f8f603f1 | 5380 | vol->curr_migr_unit = 0; |
238c0a71 | 5381 | map = get_imsm_map(dev, MAP_0); |
5551b113 | 5382 | set_pba_of_lba0(map, super->create_offset); |
b53bfba6 TM |
5383 | set_blocks_per_member(map, info_to_blocks_per_member(info, |
5384 | size_per_member / | |
5385 | BLOCKS_PER_KB)); | |
ef6ffade | 5386 | map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info)); |
0556e1a2 | 5387 | map->failed_disk_num = ~0; |
bf4442ab | 5388 | if (info->level > 0) |
fffaf1ff N |
5389 | map->map_state = (info->state ? IMSM_T_STATE_NORMAL |
5390 | : IMSM_T_STATE_UNINITIALIZED); | |
bf4442ab AK |
5391 | else |
5392 | map->map_state = info->failed_disks ? IMSM_T_STATE_FAILED : | |
5393 | IMSM_T_STATE_NORMAL; | |
252d23c0 | 5394 | map->ddf = 1; |
ef6ffade DW |
5395 | |
5396 | if (info->level == 1 && info->raid_disks > 2) { | |
38950822 AW |
5397 | free(dev); |
5398 | free(dv); | |
7a862a02 | 5399 | pr_err("imsm does not support more than 2 disksin a raid1 volume\n"); |
ef6ffade DW |
5400 | return 0; |
5401 | } | |
81062a36 DW |
5402 | |
5403 | map->raid_level = info->level; | |
4d1313e9 | 5404 | if (info->level == 10) { |
c2c087e6 | 5405 | map->raid_level = 1; |
4d1313e9 | 5406 | map->num_domains = info->raid_disks / 2; |
81062a36 DW |
5407 | } else if (info->level == 1) |
5408 | map->num_domains = info->raid_disks; | |
5409 | else | |
ff596308 | 5410 | map->num_domains = 1; |
81062a36 | 5411 | |
5551b113 | 5412 | /* info->size is only int so use the 'size' parameter instead */ |
b53bfba6 | 5413 | num_data_stripes = size_per_member / info_to_blocks_per_strip(info); |
5551b113 CA |
5414 | num_data_stripes /= map->num_domains; |
5415 | set_num_data_stripes(map, num_data_stripes); | |
ef6ffade | 5416 | |
c2c087e6 DW |
5417 | map->num_members = info->raid_disks; |
5418 | for (i = 0; i < map->num_members; i++) { | |
5419 | /* initialized in add_to_super */ | |
4eb26970 | 5420 | set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD); |
c2c087e6 | 5421 | } |
949c47a0 | 5422 | mpb->num_raid_devs++; |
2a24dc1b PB |
5423 | mpb->num_raid_devs_created++; |
5424 | dev->my_vol_raid_dev_num = mpb->num_raid_devs_created; | |
ba2de7ba | 5425 | |
b7580566 | 5426 | if (s->consistency_policy <= CONSISTENCY_POLICY_RESYNC) { |
c2462068 | 5427 | dev->rwh_policy = RWH_MULTIPLE_OFF; |
2432ce9b | 5428 | } else if (s->consistency_policy == CONSISTENCY_POLICY_PPL) { |
c2462068 | 5429 | dev->rwh_policy = RWH_MULTIPLE_DISTRIBUTED; |
2432ce9b AP |
5430 | } else { |
5431 | free(dev); | |
5432 | free(dv); | |
5433 | pr_err("imsm does not support consistency policy %s\n", | |
5434 | map_num(consistency_policies, s->consistency_policy)); | |
5435 | return 0; | |
5436 | } | |
5437 | ||
ba2de7ba DW |
5438 | dv->dev = dev; |
5439 | dv->index = super->current_vol; | |
5440 | dv->next = super->devlist; | |
5441 | super->devlist = dv; | |
c2c087e6 | 5442 | |
4d1313e9 DW |
5443 | imsm_update_version_info(super); |
5444 | ||
c2c087e6 | 5445 | return 1; |
cdddbdbc DW |
5446 | } |
5447 | ||
bf5a934a | 5448 | static int init_super_imsm(struct supertype *st, mdu_array_info_t *info, |
5308f117 | 5449 | struct shape *s, char *name, |
83cd1e97 N |
5450 | char *homehost, int *uuid, |
5451 | unsigned long long data_offset) | |
bf5a934a DW |
5452 | { |
5453 | /* This is primarily called by Create when creating a new array. | |
5454 | * We will then get add_to_super called for each component, and then | |
5455 | * write_init_super called to write it out to each device. | |
5456 | * For IMSM, Create can create on fresh devices or on a pre-existing | |
5457 | * array. | |
5458 | * To create on a pre-existing array a different method will be called. | |
5459 | * This one is just for fresh drives. | |
5460 | */ | |
5461 | struct intel_super *super; | |
5462 | struct imsm_super *mpb; | |
5463 | size_t mpb_size; | |
4d1313e9 | 5464 | char *version; |
bf5a934a | 5465 | |
83cd1e97 | 5466 | if (data_offset != INVALID_SECTORS) { |
ed503f89 | 5467 | pr_err("data-offset not supported by imsm\n"); |
83cd1e97 N |
5468 | return 0; |
5469 | } | |
5470 | ||
bf5a934a | 5471 | if (st->sb) |
5308f117 | 5472 | return init_super_imsm_volume(st, info, s, name, homehost, uuid, |
83cd1e97 | 5473 | data_offset); |
e683ca88 DW |
5474 | |
5475 | if (info) | |
5476 | mpb_size = disks_to_mpb_size(info->nr_disks); | |
5477 | else | |
f36a9ecd | 5478 | mpb_size = MAX_SECTOR_SIZE; |
bf5a934a | 5479 | |
49133e57 | 5480 | super = alloc_super(); |
f36a9ecd PB |
5481 | if (super && |
5482 | posix_memalign(&super->buf, MAX_SECTOR_SIZE, mpb_size) != 0) { | |
8d67477f | 5483 | free_imsm(super); |
e683ca88 DW |
5484 | super = NULL; |
5485 | } | |
5486 | if (!super) { | |
1ade5cc1 | 5487 | pr_err("could not allocate superblock\n"); |
bf5a934a DW |
5488 | return 0; |
5489 | } | |
de44e46f PB |
5490 | if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE, |
5491 | MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) { | |
1ade5cc1 | 5492 | pr_err("could not allocate migr_rec buffer\n"); |
8e59f3d8 | 5493 | free(super->buf); |
8d67477f | 5494 | free_imsm(super); |
8e59f3d8 AK |
5495 | return 0; |
5496 | } | |
e683ca88 | 5497 | memset(super->buf, 0, mpb_size); |
ef649044 | 5498 | mpb = super->buf; |
e683ca88 DW |
5499 | mpb->mpb_size = __cpu_to_le32(mpb_size); |
5500 | st->sb = super; | |
5501 | ||
5502 | if (info == NULL) { | |
5503 | /* zeroing superblock */ | |
5504 | return 0; | |
5505 | } | |
bf5a934a | 5506 | |
4d1313e9 DW |
5507 | mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; |
5508 | ||
5509 | version = (char *) mpb->sig; | |
5510 | strcpy(version, MPB_SIGNATURE); | |
5511 | version += strlen(MPB_SIGNATURE); | |
5512 | strcpy(version, MPB_VERSION_RAID0); | |
bf5a934a | 5513 | |
bf5a934a DW |
5514 | return 1; |
5515 | } | |
5516 | ||
f2cc4f7d AO |
5517 | static int drive_validate_sector_size(struct intel_super *super, struct dl *dl) |
5518 | { | |
5519 | unsigned int member_sector_size; | |
5520 | ||
5521 | if (dl->fd < 0) { | |
5522 | pr_err("Invalid file descriptor for %s\n", dl->devname); | |
5523 | return 0; | |
5524 | } | |
5525 | ||
5526 | if (!get_dev_sector_size(dl->fd, dl->devname, &member_sector_size)) | |
5527 | return 0; | |
5528 | if (member_sector_size != super->sector_size) | |
5529 | return 0; | |
5530 | return 1; | |
5531 | } | |
5532 | ||
f20c3968 | 5533 | static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk, |
bf5a934a DW |
5534 | int fd, char *devname) |
5535 | { | |
5536 | struct intel_super *super = st->sb; | |
d23fe947 | 5537 | struct imsm_super *mpb = super->anchor; |
3960e579 | 5538 | struct imsm_disk *_disk; |
bf5a934a DW |
5539 | struct imsm_dev *dev; |
5540 | struct imsm_map *map; | |
3960e579 | 5541 | struct dl *dl, *df; |
4eb26970 | 5542 | int slot; |
bf5a934a | 5543 | |
949c47a0 | 5544 | dev = get_imsm_dev(super, super->current_vol); |
238c0a71 | 5545 | map = get_imsm_map(dev, MAP_0); |
bf5a934a | 5546 | |
208933a7 | 5547 | if (! (dk->state & (1<<MD_DISK_SYNC))) { |
e7b84f9d | 5548 | pr_err("%s: Cannot add spare devices to IMSM volume\n", |
208933a7 N |
5549 | devname); |
5550 | return 1; | |
5551 | } | |
5552 | ||
efb30e7f DW |
5553 | if (fd == -1) { |
5554 | /* we're doing autolayout so grab the pre-marked (in | |
5555 | * validate_geometry) raid_disk | |
5556 | */ | |
5557 | for (dl = super->disks; dl; dl = dl->next) | |
5558 | if (dl->raiddisk == dk->raid_disk) | |
5559 | break; | |
5560 | } else { | |
5561 | for (dl = super->disks; dl ; dl = dl->next) | |
5562 | if (dl->major == dk->major && | |
5563 | dl->minor == dk->minor) | |
5564 | break; | |
5565 | } | |
d23fe947 | 5566 | |
208933a7 | 5567 | if (!dl) { |
e7b84f9d | 5568 | pr_err("%s is not a member of the same container\n", devname); |
f20c3968 | 5569 | return 1; |
208933a7 | 5570 | } |
bf5a934a | 5571 | |
f2cc4f7d AO |
5572 | if (!drive_validate_sector_size(super, dl)) { |
5573 | pr_err("Combining drives of different sector size in one volume is not allowed\n"); | |
5574 | return 1; | |
5575 | } | |
5576 | ||
d23fe947 DW |
5577 | /* add a pristine spare to the metadata */ |
5578 | if (dl->index < 0) { | |
5579 | dl->index = super->anchor->num_disks; | |
5580 | super->anchor->num_disks++; | |
5581 | } | |
4eb26970 DW |
5582 | /* Check the device has not already been added */ |
5583 | slot = get_imsm_disk_slot(map, dl->index); | |
5584 | if (slot >= 0 && | |
238c0a71 | 5585 | (get_imsm_ord_tbl_ent(dev, slot, MAP_X) & IMSM_ORD_REBUILD) == 0) { |
e7b84f9d | 5586 | pr_err("%s has been included in this array twice\n", |
4eb26970 DW |
5587 | devname); |
5588 | return 1; | |
5589 | } | |
656b6b5a | 5590 | set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index); |
ee5aad5a | 5591 | dl->disk.status = CONFIGURED_DISK; |
d23fe947 | 5592 | |
3960e579 DW |
5593 | /* update size of 'missing' disks to be at least as large as the |
5594 | * largest acitve member (we only have dummy missing disks when | |
5595 | * creating the first volume) | |
5596 | */ | |
5597 | if (super->current_vol == 0) { | |
5598 | for (df = super->missing; df; df = df->next) { | |
5551b113 CA |
5599 | if (total_blocks(&dl->disk) > total_blocks(&df->disk)) |
5600 | set_total_blocks(&df->disk, total_blocks(&dl->disk)); | |
3960e579 DW |
5601 | _disk = __get_imsm_disk(mpb, df->index); |
5602 | *_disk = df->disk; | |
5603 | } | |
5604 | } | |
5605 | ||
5606 | /* refresh unset/failed slots to point to valid 'missing' entries */ | |
5607 | for (df = super->missing; df; df = df->next) | |
5608 | for (slot = 0; slot < mpb->num_disks; slot++) { | |
238c0a71 | 5609 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); |
3960e579 DW |
5610 | |
5611 | if ((ord & IMSM_ORD_REBUILD) == 0) | |
5612 | continue; | |
5613 | set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD); | |
1ace8403 | 5614 | if (is_gen_migration(dev)) { |
238c0a71 AK |
5615 | struct imsm_map *map2 = get_imsm_map(dev, |
5616 | MAP_1); | |
0a108d63 | 5617 | int slot2 = get_imsm_disk_slot(map2, df->index); |
089f9d79 | 5618 | if (slot2 < map2->num_members && slot2 >= 0) { |
1ace8403 | 5619 | __u32 ord2 = get_imsm_ord_tbl_ent(dev, |
238c0a71 AK |
5620 | slot2, |
5621 | MAP_1); | |
1ace8403 AK |
5622 | if ((unsigned)df->index == |
5623 | ord_to_idx(ord2)) | |
5624 | set_imsm_ord_tbl_ent(map2, | |
0a108d63 | 5625 | slot2, |
1ace8403 AK |
5626 | df->index | |
5627 | IMSM_ORD_REBUILD); | |
5628 | } | |
5629 | } | |
3960e579 DW |
5630 | dprintf("set slot:%d to missing disk:%d\n", slot, df->index); |
5631 | break; | |
5632 | } | |
5633 | ||
d23fe947 DW |
5634 | /* if we are creating the first raid device update the family number */ |
5635 | if (super->current_vol == 0) { | |
5636 | __u32 sum; | |
5637 | struct imsm_dev *_dev = __get_imsm_dev(mpb, 0); | |
d23fe947 | 5638 | |
3960e579 | 5639 | _disk = __get_imsm_disk(mpb, dl->index); |
791b666a | 5640 | if (!_dev || !_disk) { |
e7b84f9d | 5641 | pr_err("BUG mpb setup error\n"); |
791b666a AW |
5642 | return 1; |
5643 | } | |
d23fe947 DW |
5644 | *_dev = *dev; |
5645 | *_disk = dl->disk; | |
148acb7b DW |
5646 | sum = random32(); |
5647 | sum += __gen_imsm_checksum(mpb); | |
d23fe947 | 5648 | mpb->family_num = __cpu_to_le32(sum); |
148acb7b | 5649 | mpb->orig_family_num = mpb->family_num; |
d23fe947 | 5650 | } |
ca0748fa | 5651 | super->current_disk = dl; |
f20c3968 | 5652 | return 0; |
bf5a934a DW |
5653 | } |
5654 | ||
a8619d23 AK |
5655 | /* mark_spare() |
5656 | * Function marks disk as spare and restores disk serial | |
5657 | * in case it was previously marked as failed by takeover operation | |
5658 | * reruns: | |
5659 | * -1 : critical error | |
5660 | * 0 : disk is marked as spare but serial is not set | |
5661 | * 1 : success | |
5662 | */ | |
5663 | int mark_spare(struct dl *disk) | |
5664 | { | |
5665 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
5666 | int ret_val = -1; | |
5667 | ||
5668 | if (!disk) | |
5669 | return ret_val; | |
5670 | ||
5671 | ret_val = 0; | |
5672 | if (!imsm_read_serial(disk->fd, NULL, serial)) { | |
5673 | /* Restore disk serial number, because takeover marks disk | |
5674 | * as failed and adds to serial ':0' before it becomes | |
5675 | * a spare disk. | |
5676 | */ | |
5677 | serialcpy(disk->serial, serial); | |
5678 | serialcpy(disk->disk.serial, serial); | |
5679 | ret_val = 1; | |
5680 | } | |
5681 | disk->disk.status = SPARE_DISK; | |
5682 | disk->index = -1; | |
5683 | ||
5684 | return ret_val; | |
5685 | } | |
88654014 | 5686 | |
f20c3968 | 5687 | static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk, |
72ca9bcf N |
5688 | int fd, char *devname, |
5689 | unsigned long long data_offset) | |
cdddbdbc | 5690 | { |
c2c087e6 | 5691 | struct intel_super *super = st->sb; |
c2c087e6 DW |
5692 | struct dl *dd; |
5693 | unsigned long long size; | |
fa7bb6f8 | 5694 | unsigned int member_sector_size; |
f2f27e63 | 5695 | __u32 id; |
c2c087e6 DW |
5696 | int rv; |
5697 | struct stat stb; | |
5698 | ||
88654014 LM |
5699 | /* If we are on an RAID enabled platform check that the disk is |
5700 | * attached to the raid controller. | |
5701 | * We do not need to test disks attachment for container based additions, | |
5702 | * they shall be already tested when container was created/assembled. | |
88c32bb1 | 5703 | */ |
d424212e | 5704 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 5705 | /* no orom/efi or non-intel hba of the disk */ |
f0f5a016 LM |
5706 | if (rv != 0) { |
5707 | dprintf("capability: %p fd: %d ret: %d\n", | |
5708 | super->orom, fd, rv); | |
5709 | return 1; | |
88c32bb1 DW |
5710 | } |
5711 | ||
f20c3968 DW |
5712 | if (super->current_vol >= 0) |
5713 | return add_to_super_imsm_volume(st, dk, fd, devname); | |
bf5a934a | 5714 | |
c2c087e6 | 5715 | fstat(fd, &stb); |
503975b9 | 5716 | dd = xcalloc(sizeof(*dd), 1); |
c2c087e6 DW |
5717 | dd->major = major(stb.st_rdev); |
5718 | dd->minor = minor(stb.st_rdev); | |
503975b9 | 5719 | dd->devname = devname ? xstrdup(devname) : NULL; |
c2c087e6 | 5720 | dd->fd = fd; |
689c9bf3 | 5721 | dd->e = NULL; |
1a64be56 | 5722 | dd->action = DISK_ADD; |
c2c087e6 | 5723 | rv = imsm_read_serial(fd, devname, dd->serial); |
32ba9157 | 5724 | if (rv) { |
e7b84f9d | 5725 | pr_err("failed to retrieve scsi serial, aborting\n"); |
20bee0f8 PB |
5726 | if (dd->devname) |
5727 | free(dd->devname); | |
949c47a0 | 5728 | free(dd); |
0030e8d6 | 5729 | abort(); |
c2c087e6 | 5730 | } |
20bee0f8 PB |
5731 | if (super->hba && ((super->hba->type == SYS_DEV_NVME) || |
5732 | (super->hba->type == SYS_DEV_VMD))) { | |
5733 | int i; | |
5734 | char *devpath = diskfd_to_devpath(fd); | |
5735 | char controller_path[PATH_MAX]; | |
5736 | ||
5737 | if (!devpath) { | |
5738 | pr_err("failed to get devpath, aborting\n"); | |
5739 | if (dd->devname) | |
5740 | free(dd->devname); | |
5741 | free(dd); | |
5742 | return 1; | |
5743 | } | |
5744 | ||
5745 | snprintf(controller_path, PATH_MAX-1, "%s/device", devpath); | |
5746 | free(devpath); | |
5747 | ||
5748 | if (devpath_to_vendor(controller_path) == 0x8086) { | |
5749 | /* | |
5750 | * If Intel's NVMe drive has serial ended with | |
5751 | * "-A","-B","-1" or "-2" it means that this is "x8" | |
5752 | * device (double drive on single PCIe card). | |
5753 | * User should be warned about potential data loss. | |
5754 | */ | |
5755 | for (i = MAX_RAID_SERIAL_LEN-1; i > 0; i--) { | |
5756 | /* Skip empty character at the end */ | |
5757 | if (dd->serial[i] == 0) | |
5758 | continue; | |
5759 | ||
5760 | if (((dd->serial[i] == 'A') || | |
5761 | (dd->serial[i] == 'B') || | |
5762 | (dd->serial[i] == '1') || | |
5763 | (dd->serial[i] == '2')) && | |
5764 | (dd->serial[i-1] == '-')) | |
5765 | pr_err("\tThe action you are about to take may put your data at risk.\n" | |
5766 | "\tPlease note that x8 devices may consist of two separate x4 devices " | |
5767 | "located on a single PCIe port.\n" | |
5768 | "\tRAID 0 is the only supported configuration for this type of x8 device.\n"); | |
5769 | break; | |
5770 | } | |
32716c51 PB |
5771 | } else if (super->hba->type == SYS_DEV_VMD && super->orom && |
5772 | !imsm_orom_has_tpv_support(super->orom)) { | |
5773 | pr_err("\tPlatform configuration does not support non-Intel NVMe drives.\n" | |
5774 | "\tPlease refer to Intel(R) RSTe user guide.\n"); | |
5775 | free(dd->devname); | |
5776 | free(dd); | |
5777 | return 1; | |
20bee0f8 PB |
5778 | } |
5779 | } | |
c2c087e6 | 5780 | |
c2c087e6 | 5781 | get_dev_size(fd, NULL, &size); |
fa7bb6f8 PB |
5782 | get_dev_sector_size(fd, NULL, &member_sector_size); |
5783 | ||
5784 | if (super->sector_size == 0) { | |
5785 | /* this a first device, so sector_size is not set yet */ | |
5786 | super->sector_size = member_sector_size; | |
fa7bb6f8 PB |
5787 | } |
5788 | ||
71e5411e | 5789 | /* clear migr_rec when adding disk to container */ |
85337573 AO |
5790 | memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE); |
5791 | if (lseek64(fd, size - MIGR_REC_SECTOR_POSITION*member_sector_size, | |
de44e46f | 5792 | SEEK_SET) >= 0) { |
466070ad | 5793 | if ((unsigned int)write(fd, super->migr_rec_buf, |
85337573 AO |
5794 | MIGR_REC_BUF_SECTORS*member_sector_size) != |
5795 | MIGR_REC_BUF_SECTORS*member_sector_size) | |
71e5411e PB |
5796 | perror("Write migr_rec failed"); |
5797 | } | |
5798 | ||
c2c087e6 | 5799 | size /= 512; |
1f24f035 | 5800 | serialcpy(dd->disk.serial, dd->serial); |
5551b113 CA |
5801 | set_total_blocks(&dd->disk, size); |
5802 | if (__le32_to_cpu(dd->disk.total_blocks_hi) > 0) { | |
5803 | struct imsm_super *mpb = super->anchor; | |
5804 | mpb->attributes |= MPB_ATTRIB_2TB_DISK; | |
5805 | } | |
a8619d23 | 5806 | mark_spare(dd); |
c2c087e6 | 5807 | if (sysfs_disk_to_scsi_id(fd, &id) == 0) |
b9f594fe | 5808 | dd->disk.scsi_id = __cpu_to_le32(id); |
c2c087e6 | 5809 | else |
b9f594fe | 5810 | dd->disk.scsi_id = __cpu_to_le32(0); |
43dad3d6 DW |
5811 | |
5812 | if (st->update_tail) { | |
1a64be56 LM |
5813 | dd->next = super->disk_mgmt_list; |
5814 | super->disk_mgmt_list = dd; | |
43dad3d6 DW |
5815 | } else { |
5816 | dd->next = super->disks; | |
5817 | super->disks = dd; | |
ceaf0ee1 | 5818 | super->updates_pending++; |
43dad3d6 | 5819 | } |
f20c3968 DW |
5820 | |
5821 | return 0; | |
cdddbdbc DW |
5822 | } |
5823 | ||
1a64be56 LM |
5824 | static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk) |
5825 | { | |
5826 | struct intel_super *super = st->sb; | |
5827 | struct dl *dd; | |
5828 | ||
5829 | /* remove from super works only in mdmon - for communication | |
5830 | * manager - monitor. Check if communication memory buffer | |
5831 | * is prepared. | |
5832 | */ | |
5833 | if (!st->update_tail) { | |
1ade5cc1 | 5834 | pr_err("shall be used in mdmon context only\n"); |
1a64be56 LM |
5835 | return 1; |
5836 | } | |
503975b9 | 5837 | dd = xcalloc(1, sizeof(*dd)); |
1a64be56 LM |
5838 | dd->major = dk->major; |
5839 | dd->minor = dk->minor; | |
1a64be56 | 5840 | dd->fd = -1; |
a8619d23 | 5841 | mark_spare(dd); |
1a64be56 LM |
5842 | dd->action = DISK_REMOVE; |
5843 | ||
5844 | dd->next = super->disk_mgmt_list; | |
5845 | super->disk_mgmt_list = dd; | |
5846 | ||
1a64be56 LM |
5847 | return 0; |
5848 | } | |
5849 | ||
f796af5d DW |
5850 | static int store_imsm_mpb(int fd, struct imsm_super *mpb); |
5851 | ||
5852 | static union { | |
f36a9ecd | 5853 | char buf[MAX_SECTOR_SIZE]; |
f796af5d | 5854 | struct imsm_super anchor; |
f36a9ecd | 5855 | } spare_record __attribute__ ((aligned(MAX_SECTOR_SIZE))); |
c2c087e6 | 5856 | |
d23fe947 DW |
5857 | /* spare records have their own family number and do not have any defined raid |
5858 | * devices | |
5859 | */ | |
5860 | static int write_super_imsm_spares(struct intel_super *super, int doclose) | |
5861 | { | |
d23fe947 | 5862 | struct imsm_super *mpb = super->anchor; |
f796af5d | 5863 | struct imsm_super *spare = &spare_record.anchor; |
d23fe947 DW |
5864 | __u32 sum; |
5865 | struct dl *d; | |
5866 | ||
68641cdb JS |
5867 | spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super)); |
5868 | spare->generation_num = __cpu_to_le32(1UL); | |
f796af5d | 5869 | spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; |
68641cdb JS |
5870 | spare->num_disks = 1; |
5871 | spare->num_raid_devs = 0; | |
5872 | spare->cache_size = mpb->cache_size; | |
5873 | spare->pwr_cycle_count = __cpu_to_le32(1); | |
f796af5d DW |
5874 | |
5875 | snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH, | |
5876 | MPB_SIGNATURE MPB_VERSION_RAID0); | |
d23fe947 DW |
5877 | |
5878 | for (d = super->disks; d; d = d->next) { | |
8796fdc4 | 5879 | if (d->index != -1) |
d23fe947 DW |
5880 | continue; |
5881 | ||
f796af5d | 5882 | spare->disk[0] = d->disk; |
027c374f CA |
5883 | if (__le32_to_cpu(d->disk.total_blocks_hi) > 0) |
5884 | spare->attributes |= MPB_ATTRIB_2TB_DISK; | |
5885 | ||
f36a9ecd PB |
5886 | if (super->sector_size == 4096) |
5887 | convert_to_4k_imsm_disk(&spare->disk[0]); | |
5888 | ||
f796af5d DW |
5889 | sum = __gen_imsm_checksum(spare); |
5890 | spare->family_num = __cpu_to_le32(sum); | |
5891 | spare->orig_family_num = 0; | |
5892 | sum = __gen_imsm_checksum(spare); | |
5893 | spare->check_sum = __cpu_to_le32(sum); | |
d23fe947 | 5894 | |
f796af5d | 5895 | if (store_imsm_mpb(d->fd, spare)) { |
1ade5cc1 N |
5896 | pr_err("failed for device %d:%d %s\n", |
5897 | d->major, d->minor, strerror(errno)); | |
e74255d9 | 5898 | return 1; |
d23fe947 DW |
5899 | } |
5900 | if (doclose) { | |
5901 | close(d->fd); | |
5902 | d->fd = -1; | |
5903 | } | |
5904 | } | |
5905 | ||
e74255d9 | 5906 | return 0; |
d23fe947 DW |
5907 | } |
5908 | ||
36988a3d | 5909 | static int write_super_imsm(struct supertype *st, int doclose) |
cdddbdbc | 5910 | { |
36988a3d | 5911 | struct intel_super *super = st->sb; |
f36a9ecd | 5912 | unsigned int sector_size = super->sector_size; |
949c47a0 | 5913 | struct imsm_super *mpb = super->anchor; |
c2c087e6 DW |
5914 | struct dl *d; |
5915 | __u32 generation; | |
5916 | __u32 sum; | |
d23fe947 | 5917 | int spares = 0; |
949c47a0 | 5918 | int i; |
a48ac0a8 | 5919 | __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk); |
36988a3d | 5920 | int num_disks = 0; |
146c6260 | 5921 | int clear_migration_record = 1; |
bbab0940 | 5922 | __u32 bbm_log_size; |
cdddbdbc | 5923 | |
c2c087e6 DW |
5924 | /* 'generation' is incremented everytime the metadata is written */ |
5925 | generation = __le32_to_cpu(mpb->generation_num); | |
5926 | generation++; | |
5927 | mpb->generation_num = __cpu_to_le32(generation); | |
5928 | ||
148acb7b DW |
5929 | /* fix up cases where previous mdadm releases failed to set |
5930 | * orig_family_num | |
5931 | */ | |
5932 | if (mpb->orig_family_num == 0) | |
5933 | mpb->orig_family_num = mpb->family_num; | |
5934 | ||
d23fe947 | 5935 | for (d = super->disks; d; d = d->next) { |
8796fdc4 | 5936 | if (d->index == -1) |
d23fe947 | 5937 | spares++; |
36988a3d | 5938 | else { |
d23fe947 | 5939 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
5940 | num_disks++; |
5941 | } | |
d23fe947 | 5942 | } |
36988a3d | 5943 | for (d = super->missing; d; d = d->next) { |
47ee5a45 | 5944 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
5945 | num_disks++; |
5946 | } | |
5947 | mpb->num_disks = num_disks; | |
5948 | mpb_size += sizeof(struct imsm_disk) * mpb->num_disks; | |
b9f594fe | 5949 | |
949c47a0 DW |
5950 | for (i = 0; i < mpb->num_raid_devs; i++) { |
5951 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
36988a3d AK |
5952 | struct imsm_dev *dev2 = get_imsm_dev(super, i); |
5953 | if (dev && dev2) { | |
5954 | imsm_copy_dev(dev, dev2); | |
5955 | mpb_size += sizeof_imsm_dev(dev, 0); | |
5956 | } | |
146c6260 AK |
5957 | if (is_gen_migration(dev2)) |
5958 | clear_migration_record = 0; | |
949c47a0 | 5959 | } |
bbab0940 TM |
5960 | |
5961 | bbm_log_size = get_imsm_bbm_log_size(super->bbm_log); | |
5962 | ||
5963 | if (bbm_log_size) { | |
5964 | memcpy((void *)mpb + mpb_size, super->bbm_log, bbm_log_size); | |
5965 | mpb->attributes |= MPB_ATTRIB_BBM; | |
5966 | } else | |
5967 | mpb->attributes &= ~MPB_ATTRIB_BBM; | |
5968 | ||
5969 | super->anchor->bbm_log_size = __cpu_to_le32(bbm_log_size); | |
5970 | mpb_size += bbm_log_size; | |
a48ac0a8 | 5971 | mpb->mpb_size = __cpu_to_le32(mpb_size); |
949c47a0 | 5972 | |
bbab0940 TM |
5973 | #ifdef DEBUG |
5974 | assert(super->len == 0 || mpb_size <= super->len); | |
5975 | #endif | |
5976 | ||
c2c087e6 | 5977 | /* recalculate checksum */ |
949c47a0 | 5978 | sum = __gen_imsm_checksum(mpb); |
c2c087e6 DW |
5979 | mpb->check_sum = __cpu_to_le32(sum); |
5980 | ||
51d83f5d AK |
5981 | if (super->clean_migration_record_by_mdmon) { |
5982 | clear_migration_record = 1; | |
5983 | super->clean_migration_record_by_mdmon = 0; | |
5984 | } | |
146c6260 | 5985 | if (clear_migration_record) |
de44e46f | 5986 | memset(super->migr_rec_buf, 0, |
85337573 | 5987 | MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE); |
146c6260 | 5988 | |
f36a9ecd PB |
5989 | if (sector_size == 4096) |
5990 | convert_to_4k(super); | |
5991 | ||
d23fe947 | 5992 | /* write the mpb for disks that compose raid devices */ |
c2c087e6 | 5993 | for (d = super->disks; d ; d = d->next) { |
86c54047 | 5994 | if (d->index < 0 || is_failed(&d->disk)) |
d23fe947 | 5995 | continue; |
30602f53 | 5996 | |
146c6260 AK |
5997 | if (clear_migration_record) { |
5998 | unsigned long long dsize; | |
5999 | ||
6000 | get_dev_size(d->fd, NULL, &dsize); | |
de44e46f PB |
6001 | if (lseek64(d->fd, dsize - sector_size, |
6002 | SEEK_SET) >= 0) { | |
466070ad PB |
6003 | if ((unsigned int)write(d->fd, |
6004 | super->migr_rec_buf, | |
de44e46f PB |
6005 | MIGR_REC_BUF_SECTORS*sector_size) != |
6006 | MIGR_REC_BUF_SECTORS*sector_size) | |
9e2d750d | 6007 | perror("Write migr_rec failed"); |
146c6260 AK |
6008 | } |
6009 | } | |
51d83f5d AK |
6010 | |
6011 | if (store_imsm_mpb(d->fd, mpb)) | |
6012 | fprintf(stderr, | |
1ade5cc1 N |
6013 | "failed for device %d:%d (fd: %d)%s\n", |
6014 | d->major, d->minor, | |
51d83f5d AK |
6015 | d->fd, strerror(errno)); |
6016 | ||
c2c087e6 DW |
6017 | if (doclose) { |
6018 | close(d->fd); | |
6019 | d->fd = -1; | |
6020 | } | |
6021 | } | |
6022 | ||
d23fe947 DW |
6023 | if (spares) |
6024 | return write_super_imsm_spares(super, doclose); | |
6025 | ||
e74255d9 | 6026 | return 0; |
c2c087e6 DW |
6027 | } |
6028 | ||
9b1fb677 | 6029 | static int create_array(struct supertype *st, int dev_idx) |
43dad3d6 DW |
6030 | { |
6031 | size_t len; | |
6032 | struct imsm_update_create_array *u; | |
6033 | struct intel_super *super = st->sb; | |
9b1fb677 | 6034 | struct imsm_dev *dev = get_imsm_dev(super, dev_idx); |
238c0a71 | 6035 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
54c2c1ea DW |
6036 | struct disk_info *inf; |
6037 | struct imsm_disk *disk; | |
6038 | int i; | |
43dad3d6 | 6039 | |
54c2c1ea DW |
6040 | len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) + |
6041 | sizeof(*inf) * map->num_members; | |
503975b9 | 6042 | u = xmalloc(len); |
43dad3d6 | 6043 | u->type = update_create_array; |
9b1fb677 | 6044 | u->dev_idx = dev_idx; |
43dad3d6 | 6045 | imsm_copy_dev(&u->dev, dev); |
54c2c1ea DW |
6046 | inf = get_disk_info(u); |
6047 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 6048 | int idx = get_imsm_disk_idx(dev, i, MAP_X); |
9b1fb677 | 6049 | |
54c2c1ea | 6050 | disk = get_imsm_disk(super, idx); |
1ca5c8e0 N |
6051 | if (!disk) |
6052 | disk = get_imsm_missing(super, idx); | |
54c2c1ea DW |
6053 | serialcpy(inf[i].serial, disk->serial); |
6054 | } | |
43dad3d6 DW |
6055 | append_metadata_update(st, u, len); |
6056 | ||
6057 | return 0; | |
6058 | } | |
6059 | ||
1a64be56 | 6060 | static int mgmt_disk(struct supertype *st) |
43dad3d6 DW |
6061 | { |
6062 | struct intel_super *super = st->sb; | |
6063 | size_t len; | |
1a64be56 | 6064 | struct imsm_update_add_remove_disk *u; |
43dad3d6 | 6065 | |
1a64be56 | 6066 | if (!super->disk_mgmt_list) |
43dad3d6 DW |
6067 | return 0; |
6068 | ||
6069 | len = sizeof(*u); | |
503975b9 | 6070 | u = xmalloc(len); |
1a64be56 | 6071 | u->type = update_add_remove_disk; |
43dad3d6 DW |
6072 | append_metadata_update(st, u, len); |
6073 | ||
6074 | return 0; | |
6075 | } | |
2432ce9b AP |
6076 | |
6077 | __u32 crc32c_le(__u32 crc, unsigned char const *p, size_t len); | |
6078 | ||
6079 | static int write_init_ppl_imsm(struct supertype *st, struct mdinfo *info, int fd) | |
6080 | { | |
6081 | struct intel_super *super = st->sb; | |
6082 | void *buf; | |
6083 | struct ppl_header *ppl_hdr; | |
6084 | int ret; | |
6085 | ||
b2514242 PB |
6086 | /* first clear entire ppl space */ |
6087 | ret = zero_disk_range(fd, info->ppl_sector, info->ppl_size); | |
6088 | if (ret) | |
6089 | return ret; | |
6090 | ||
6091 | ret = posix_memalign(&buf, MAX_SECTOR_SIZE, PPL_HEADER_SIZE); | |
2432ce9b AP |
6092 | if (ret) { |
6093 | pr_err("Failed to allocate PPL header buffer\n"); | |
6094 | return ret; | |
6095 | } | |
6096 | ||
6097 | memset(buf, 0, PPL_HEADER_SIZE); | |
6098 | ppl_hdr = buf; | |
6099 | memset(ppl_hdr->reserved, 0xff, PPL_HDR_RESERVED); | |
6100 | ppl_hdr->signature = __cpu_to_le32(super->anchor->orig_family_num); | |
b23d0750 AP |
6101 | |
6102 | if (info->mismatch_cnt) { | |
6103 | /* | |
6104 | * We are overwriting an invalid ppl. Make one entry with wrong | |
6105 | * checksum to prevent the kernel from skipping resync. | |
6106 | */ | |
6107 | ppl_hdr->entries_count = __cpu_to_le32(1); | |
6108 | ppl_hdr->entries[0].checksum = ~0; | |
6109 | } | |
6110 | ||
2432ce9b AP |
6111 | ppl_hdr->checksum = __cpu_to_le32(~crc32c_le(~0, buf, PPL_HEADER_SIZE)); |
6112 | ||
6113 | if (lseek64(fd, info->ppl_sector * 512, SEEK_SET) < 0) { | |
6114 | ret = errno; | |
6115 | perror("Failed to seek to PPL header location"); | |
6116 | } | |
6117 | ||
6118 | if (!ret && write(fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) { | |
6119 | ret = errno; | |
6120 | perror("Write PPL header failed"); | |
6121 | } | |
6122 | ||
6123 | if (!ret) | |
6124 | fsync(fd); | |
6125 | ||
6126 | free(buf); | |
6127 | return ret; | |
6128 | } | |
6129 | ||
6130 | static int validate_ppl_imsm(struct supertype *st, struct mdinfo *info, | |
6131 | struct mdinfo *disk) | |
6132 | { | |
6133 | struct intel_super *super = st->sb; | |
6134 | struct dl *d; | |
6135 | void *buf; | |
6136 | int ret = 0; | |
6137 | struct ppl_header *ppl_hdr; | |
6138 | __u32 crc; | |
6139 | struct imsm_dev *dev; | |
2432ce9b | 6140 | __u32 idx; |
44b6b876 PB |
6141 | unsigned int i; |
6142 | unsigned long long ppl_offset = 0; | |
6143 | unsigned long long prev_gen_num = 0; | |
2432ce9b AP |
6144 | |
6145 | if (disk->disk.raid_disk < 0) | |
6146 | return 0; | |
6147 | ||
44b6b876 | 6148 | if (posix_memalign(&buf, MAX_SECTOR_SIZE, PPL_HEADER_SIZE)) { |
2432ce9b AP |
6149 | pr_err("Failed to allocate PPL header buffer\n"); |
6150 | return -1; | |
6151 | } | |
6152 | ||
6153 | dev = get_imsm_dev(super, info->container_member); | |
2fc0fc63 | 6154 | idx = get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_0); |
2432ce9b AP |
6155 | d = get_imsm_dl_disk(super, idx); |
6156 | ||
6157 | if (!d || d->index < 0 || is_failed(&d->disk)) | |
6158 | goto out; | |
6159 | ||
44b6b876 PB |
6160 | ret = 1; |
6161 | while (ppl_offset < MULTIPLE_PPL_AREA_SIZE_IMSM) { | |
6162 | dprintf("Checking potential PPL at offset: %llu\n", ppl_offset); | |
2432ce9b | 6163 | |
44b6b876 PB |
6164 | if (lseek64(d->fd, info->ppl_sector * 512 + ppl_offset, |
6165 | SEEK_SET) < 0) { | |
6166 | perror("Failed to seek to PPL header location"); | |
6167 | ret = -1; | |
6168 | goto out; | |
6169 | } | |
2432ce9b | 6170 | |
44b6b876 PB |
6171 | if (read(d->fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) { |
6172 | perror("Read PPL header failed"); | |
6173 | ret = -1; | |
6174 | goto out; | |
6175 | } | |
2432ce9b | 6176 | |
44b6b876 | 6177 | ppl_hdr = buf; |
2432ce9b | 6178 | |
44b6b876 PB |
6179 | crc = __le32_to_cpu(ppl_hdr->checksum); |
6180 | ppl_hdr->checksum = 0; | |
6181 | ||
6182 | if (crc != ~crc32c_le(~0, buf, PPL_HEADER_SIZE)) { | |
6183 | dprintf("Wrong PPL header checksum on %s\n", | |
6184 | d->devname); | |
6185 | goto out; | |
6186 | } | |
6187 | ||
6188 | if (prev_gen_num > __le64_to_cpu(ppl_hdr->generation)) { | |
6189 | /* previous was newest, it was already checked */ | |
6190 | goto out; | |
6191 | } | |
6192 | ||
6193 | if ((__le32_to_cpu(ppl_hdr->signature) != | |
6194 | super->anchor->orig_family_num)) { | |
6195 | dprintf("Wrong PPL header signature on %s\n", | |
6196 | d->devname); | |
6197 | ret = 1; | |
6198 | goto out; | |
6199 | } | |
6200 | ||
6201 | ret = 0; | |
6202 | prev_gen_num = __le64_to_cpu(ppl_hdr->generation); | |
2432ce9b | 6203 | |
44b6b876 PB |
6204 | ppl_offset += PPL_HEADER_SIZE; |
6205 | for (i = 0; i < __le32_to_cpu(ppl_hdr->entries_count); i++) | |
6206 | ppl_offset += | |
6207 | __le32_to_cpu(ppl_hdr->entries[i].pp_size); | |
2432ce9b AP |
6208 | } |
6209 | ||
6210 | out: | |
6211 | free(buf); | |
6212 | ||
54148aba PB |
6213 | /* |
6214 | * Update metadata to use mutliple PPLs area (1MB). | |
6215 | * This is done once for all RAID members | |
6216 | */ | |
6217 | if (info->consistency_policy == CONSISTENCY_POLICY_PPL && | |
6218 | info->ppl_size != (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9)) { | |
6219 | char subarray[20]; | |
6220 | struct mdinfo *member_dev; | |
6221 | ||
6222 | sprintf(subarray, "%d", info->container_member); | |
6223 | ||
6224 | if (mdmon_running(st->container_devnm)) | |
6225 | st->update_tail = &st->updates; | |
6226 | ||
6227 | if (st->ss->update_subarray(st, subarray, "ppl", NULL)) { | |
6228 | pr_err("Failed to update subarray %s\n", | |
6229 | subarray); | |
6230 | } else { | |
6231 | if (st->update_tail) | |
6232 | flush_metadata_updates(st); | |
6233 | else | |
6234 | st->ss->sync_metadata(st); | |
6235 | info->ppl_size = (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9); | |
6236 | for (member_dev = info->devs; member_dev; | |
6237 | member_dev = member_dev->next) | |
6238 | member_dev->ppl_size = | |
6239 | (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9); | |
6240 | } | |
6241 | } | |
6242 | ||
b23d0750 | 6243 | if (ret == 1) { |
2fc0fc63 AP |
6244 | struct imsm_map *map = get_imsm_map(dev, MAP_X); |
6245 | ||
50b9c10d PB |
6246 | if (map->map_state == IMSM_T_STATE_UNINITIALIZED || |
6247 | (map->map_state == IMSM_T_STATE_NORMAL && | |
2ec9d182 AP |
6248 | !(dev->vol.dirty & RAIDVOL_DIRTY)) || |
6249 | (dev->vol.migr_state == MIGR_REBUILD && | |
6250 | dev->vol.curr_migr_unit == 0 && | |
6251 | get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_1) != idx)) | |
b23d0750 AP |
6252 | ret = st->ss->write_init_ppl(st, info, d->fd); |
6253 | else | |
6254 | info->mismatch_cnt++; | |
6255 | } | |
2432ce9b AP |
6256 | |
6257 | return ret; | |
6258 | } | |
6259 | ||
2432ce9b AP |
6260 | static int write_init_ppl_imsm_all(struct supertype *st, struct mdinfo *info) |
6261 | { | |
6262 | struct intel_super *super = st->sb; | |
6263 | struct dl *d; | |
6264 | int ret = 0; | |
6265 | ||
6266 | if (info->consistency_policy != CONSISTENCY_POLICY_PPL || | |
6267 | info->array.level != 5) | |
6268 | return 0; | |
6269 | ||
6270 | for (d = super->disks; d ; d = d->next) { | |
6271 | if (d->index < 0 || is_failed(&d->disk)) | |
6272 | continue; | |
6273 | ||
6274 | ret = st->ss->write_init_ppl(st, info, d->fd); | |
6275 | if (ret) | |
6276 | break; | |
6277 | } | |
6278 | ||
6279 | return ret; | |
6280 | } | |
43dad3d6 | 6281 | |
c2c087e6 DW |
6282 | static int write_init_super_imsm(struct supertype *st) |
6283 | { | |
9b1fb677 DW |
6284 | struct intel_super *super = st->sb; |
6285 | int current_vol = super->current_vol; | |
2432ce9b AP |
6286 | int rv = 0; |
6287 | struct mdinfo info; | |
6288 | ||
6289 | getinfo_super_imsm(st, &info, NULL); | |
9b1fb677 DW |
6290 | |
6291 | /* we are done with current_vol reset it to point st at the container */ | |
6292 | super->current_vol = -1; | |
6293 | ||
8273f55e | 6294 | if (st->update_tail) { |
43dad3d6 DW |
6295 | /* queue the recently created array / added disk |
6296 | * as a metadata update */ | |
8273f55e | 6297 | |
43dad3d6 | 6298 | /* determine if we are creating a volume or adding a disk */ |
9b1fb677 | 6299 | if (current_vol < 0) { |
1a64be56 LM |
6300 | /* in the mgmt (add/remove) disk case we are running |
6301 | * in mdmon context, so don't close fd's | |
43dad3d6 | 6302 | */ |
2432ce9b AP |
6303 | rv = mgmt_disk(st); |
6304 | } else { | |
6305 | rv = write_init_ppl_imsm_all(st, &info); | |
6306 | if (!rv) | |
6307 | rv = create_array(st, current_vol); | |
6308 | } | |
d682f344 N |
6309 | } else { |
6310 | struct dl *d; | |
6311 | for (d = super->disks; d; d = d->next) | |
ba728be7 | 6312 | Kill(d->devname, NULL, 0, -1, 1); |
2432ce9b AP |
6313 | if (current_vol >= 0) |
6314 | rv = write_init_ppl_imsm_all(st, &info); | |
6315 | if (!rv) | |
6316 | rv = write_super_imsm(st, 1); | |
d682f344 | 6317 | } |
2432ce9b AP |
6318 | |
6319 | return rv; | |
cdddbdbc DW |
6320 | } |
6321 | ||
e683ca88 | 6322 | static int store_super_imsm(struct supertype *st, int fd) |
cdddbdbc | 6323 | { |
e683ca88 DW |
6324 | struct intel_super *super = st->sb; |
6325 | struct imsm_super *mpb = super ? super->anchor : NULL; | |
551c80c1 | 6326 | |
e683ca88 | 6327 | if (!mpb) |
ad97895e DW |
6328 | return 1; |
6329 | ||
f36a9ecd PB |
6330 | if (super->sector_size == 4096) |
6331 | convert_to_4k(super); | |
e683ca88 | 6332 | return store_imsm_mpb(fd, mpb); |
cdddbdbc DW |
6333 | } |
6334 | ||
cdddbdbc DW |
6335 | static int validate_geometry_imsm_container(struct supertype *st, int level, |
6336 | int layout, int raiddisks, int chunk, | |
af4348dd N |
6337 | unsigned long long size, |
6338 | unsigned long long data_offset, | |
6339 | char *dev, | |
2c514b71 NB |
6340 | unsigned long long *freesize, |
6341 | int verbose) | |
cdddbdbc | 6342 | { |
c2c087e6 DW |
6343 | int fd; |
6344 | unsigned long long ldsize; | |
594dc1b8 | 6345 | struct intel_super *super; |
f2f5c343 | 6346 | int rv = 0; |
cdddbdbc | 6347 | |
c2c087e6 DW |
6348 | if (level != LEVEL_CONTAINER) |
6349 | return 0; | |
6350 | if (!dev) | |
6351 | return 1; | |
6352 | ||
6353 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
6354 | if (fd < 0) { | |
ba728be7 | 6355 | if (verbose > 0) |
e7b84f9d | 6356 | pr_err("imsm: Cannot open %s: %s\n", |
2c514b71 | 6357 | dev, strerror(errno)); |
c2c087e6 DW |
6358 | return 0; |
6359 | } | |
6360 | if (!get_dev_size(fd, dev, &ldsize)) { | |
6361 | close(fd); | |
6362 | return 0; | |
6363 | } | |
f2f5c343 LM |
6364 | |
6365 | /* capabilities retrieve could be possible | |
6366 | * note that there is no fd for the disks in array. | |
6367 | */ | |
6368 | super = alloc_super(); | |
8d67477f TM |
6369 | if (!super) { |
6370 | close(fd); | |
6371 | return 0; | |
6372 | } | |
fa7bb6f8 PB |
6373 | if (!get_dev_sector_size(fd, NULL, &super->sector_size)) { |
6374 | close(fd); | |
6375 | free_imsm(super); | |
6376 | return 0; | |
6377 | } | |
6378 | ||
ba728be7 | 6379 | rv = find_intel_hba_capability(fd, super, verbose > 0 ? dev : NULL); |
f2f5c343 LM |
6380 | if (rv != 0) { |
6381 | #if DEBUG | |
6382 | char str[256]; | |
6383 | fd2devname(fd, str); | |
1ade5cc1 | 6384 | dprintf("fd: %d %s orom: %p rv: %d raiddisk: %d\n", |
f2f5c343 LM |
6385 | fd, str, super->orom, rv, raiddisks); |
6386 | #endif | |
6387 | /* no orom/efi or non-intel hba of the disk */ | |
6388 | close(fd); | |
6389 | free_imsm(super); | |
6390 | return 0; | |
6391 | } | |
c2c087e6 | 6392 | close(fd); |
9126b9a8 CA |
6393 | if (super->orom) { |
6394 | if (raiddisks > super->orom->tds) { | |
6395 | if (verbose) | |
7a862a02 | 6396 | pr_err("%d exceeds maximum number of platform supported disks: %d\n", |
9126b9a8 CA |
6397 | raiddisks, super->orom->tds); |
6398 | free_imsm(super); | |
6399 | return 0; | |
6400 | } | |
6401 | if ((super->orom->attr & IMSM_OROM_ATTR_2TB_DISK) == 0 && | |
6402 | (ldsize >> 9) >> 32 > 0) { | |
6403 | if (verbose) | |
e7b84f9d | 6404 | pr_err("%s exceeds maximum platform supported size\n", dev); |
9126b9a8 CA |
6405 | free_imsm(super); |
6406 | return 0; | |
6407 | } | |
f2f5c343 | 6408 | } |
c2c087e6 | 6409 | |
af4348dd | 6410 | *freesize = avail_size_imsm(st, ldsize >> 9, data_offset); |
f2f5c343 | 6411 | free_imsm(super); |
c2c087e6 DW |
6412 | |
6413 | return 1; | |
cdddbdbc DW |
6414 | } |
6415 | ||
0dcecb2e DW |
6416 | static unsigned long long find_size(struct extent *e, int *idx, int num_extents) |
6417 | { | |
6418 | const unsigned long long base_start = e[*idx].start; | |
6419 | unsigned long long end = base_start + e[*idx].size; | |
6420 | int i; | |
6421 | ||
6422 | if (base_start == end) | |
6423 | return 0; | |
6424 | ||
6425 | *idx = *idx + 1; | |
6426 | for (i = *idx; i < num_extents; i++) { | |
6427 | /* extend overlapping extents */ | |
6428 | if (e[i].start >= base_start && | |
6429 | e[i].start <= end) { | |
6430 | if (e[i].size == 0) | |
6431 | return 0; | |
6432 | if (e[i].start + e[i].size > end) | |
6433 | end = e[i].start + e[i].size; | |
6434 | } else if (e[i].start > end) { | |
6435 | *idx = i; | |
6436 | break; | |
6437 | } | |
6438 | } | |
6439 | ||
6440 | return end - base_start; | |
6441 | } | |
6442 | ||
6443 | static unsigned long long merge_extents(struct intel_super *super, int sum_extents) | |
6444 | { | |
6445 | /* build a composite disk with all known extents and generate a new | |
6446 | * 'maxsize' given the "all disks in an array must share a common start | |
6447 | * offset" constraint | |
6448 | */ | |
503975b9 | 6449 | struct extent *e = xcalloc(sum_extents, sizeof(*e)); |
0dcecb2e DW |
6450 | struct dl *dl; |
6451 | int i, j; | |
6452 | int start_extent; | |
6453 | unsigned long long pos; | |
b9d77223 | 6454 | unsigned long long start = 0; |
0dcecb2e DW |
6455 | unsigned long long maxsize; |
6456 | unsigned long reserve; | |
6457 | ||
0dcecb2e DW |
6458 | /* coalesce and sort all extents. also, check to see if we need to |
6459 | * reserve space between member arrays | |
6460 | */ | |
6461 | j = 0; | |
6462 | for (dl = super->disks; dl; dl = dl->next) { | |
6463 | if (!dl->e) | |
6464 | continue; | |
6465 | for (i = 0; i < dl->extent_cnt; i++) | |
6466 | e[j++] = dl->e[i]; | |
6467 | } | |
6468 | qsort(e, sum_extents, sizeof(*e), cmp_extent); | |
6469 | ||
6470 | /* merge extents */ | |
6471 | i = 0; | |
6472 | j = 0; | |
6473 | while (i < sum_extents) { | |
6474 | e[j].start = e[i].start; | |
6475 | e[j].size = find_size(e, &i, sum_extents); | |
6476 | j++; | |
6477 | if (e[j-1].size == 0) | |
6478 | break; | |
6479 | } | |
6480 | ||
6481 | pos = 0; | |
6482 | maxsize = 0; | |
6483 | start_extent = 0; | |
6484 | i = 0; | |
6485 | do { | |
6486 | unsigned long long esize; | |
6487 | ||
6488 | esize = e[i].start - pos; | |
6489 | if (esize >= maxsize) { | |
6490 | maxsize = esize; | |
6491 | start = pos; | |
6492 | start_extent = i; | |
6493 | } | |
6494 | pos = e[i].start + e[i].size; | |
6495 | i++; | |
6496 | } while (e[i-1].size); | |
6497 | free(e); | |
6498 | ||
a7dd165b DW |
6499 | if (maxsize == 0) |
6500 | return 0; | |
6501 | ||
6502 | /* FIXME assumes volume at offset 0 is the first volume in a | |
6503 | * container | |
6504 | */ | |
0dcecb2e DW |
6505 | if (start_extent > 0) |
6506 | reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */ | |
6507 | else | |
6508 | reserve = 0; | |
6509 | ||
6510 | if (maxsize < reserve) | |
a7dd165b | 6511 | return 0; |
0dcecb2e | 6512 | |
5551b113 | 6513 | super->create_offset = ~((unsigned long long) 0); |
0dcecb2e | 6514 | if (start + reserve > super->create_offset) |
a7dd165b | 6515 | return 0; /* start overflows create_offset */ |
0dcecb2e DW |
6516 | super->create_offset = start + reserve; |
6517 | ||
6518 | return maxsize - reserve; | |
6519 | } | |
6520 | ||
88c32bb1 DW |
6521 | static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks) |
6522 | { | |
6523 | if (level < 0 || level == 6 || level == 4) | |
6524 | return 0; | |
6525 | ||
6526 | /* if we have an orom prevent invalid raid levels */ | |
6527 | if (orom) | |
6528 | switch (level) { | |
6529 | case 0: return imsm_orom_has_raid0(orom); | |
6530 | case 1: | |
6531 | if (raiddisks > 2) | |
6532 | return imsm_orom_has_raid1e(orom); | |
1c556e92 DW |
6533 | return imsm_orom_has_raid1(orom) && raiddisks == 2; |
6534 | case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4; | |
6535 | case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2; | |
88c32bb1 DW |
6536 | } |
6537 | else | |
6538 | return 1; /* not on an Intel RAID platform so anything goes */ | |
6539 | ||
6540 | return 0; | |
6541 | } | |
6542 | ||
ca9de185 LM |
6543 | static int |
6544 | active_arrays_by_format(char *name, char* hba, struct md_list **devlist, | |
6545 | int dpa, int verbose) | |
6546 | { | |
6547 | struct mdstat_ent *mdstat = mdstat_read(0, 0); | |
594dc1b8 | 6548 | struct mdstat_ent *memb; |
ca9de185 LM |
6549 | int count = 0; |
6550 | int num = 0; | |
594dc1b8 | 6551 | struct md_list *dv; |
ca9de185 LM |
6552 | int found; |
6553 | ||
6554 | for (memb = mdstat ; memb ; memb = memb->next) { | |
6555 | if (memb->metadata_version && | |
fc54fe7a | 6556 | (strncmp(memb->metadata_version, "external:", 9) == 0) && |
ca9de185 LM |
6557 | (strcmp(&memb->metadata_version[9], name) == 0) && |
6558 | !is_subarray(memb->metadata_version+9) && | |
6559 | memb->members) { | |
6560 | struct dev_member *dev = memb->members; | |
6561 | int fd = -1; | |
6562 | while(dev && (fd < 0)) { | |
503975b9 N |
6563 | char *path = xmalloc(strlen(dev->name) + strlen("/dev/") + 1); |
6564 | num = sprintf(path, "%s%s", "/dev/", dev->name); | |
6565 | if (num > 0) | |
6566 | fd = open(path, O_RDONLY, 0); | |
089f9d79 | 6567 | if (num <= 0 || fd < 0) { |
676e87a8 | 6568 | pr_vrb("Cannot open %s: %s\n", |
503975b9 | 6569 | dev->name, strerror(errno)); |
ca9de185 | 6570 | } |
503975b9 | 6571 | free(path); |
ca9de185 LM |
6572 | dev = dev->next; |
6573 | } | |
6574 | found = 0; | |
089f9d79 | 6575 | if (fd >= 0 && disk_attached_to_hba(fd, hba)) { |
ca9de185 LM |
6576 | struct mdstat_ent *vol; |
6577 | for (vol = mdstat ; vol ; vol = vol->next) { | |
089f9d79 | 6578 | if (vol->active > 0 && |
ca9de185 | 6579 | vol->metadata_version && |
9581efb1 | 6580 | is_container_member(vol, memb->devnm)) { |
ca9de185 LM |
6581 | found++; |
6582 | count++; | |
6583 | } | |
6584 | } | |
6585 | if (*devlist && (found < dpa)) { | |
503975b9 | 6586 | dv = xcalloc(1, sizeof(*dv)); |
9581efb1 N |
6587 | dv->devname = xmalloc(strlen(memb->devnm) + strlen("/dev/") + 1); |
6588 | sprintf(dv->devname, "%s%s", "/dev/", memb->devnm); | |
503975b9 N |
6589 | dv->found = found; |
6590 | dv->used = 0; | |
6591 | dv->next = *devlist; | |
6592 | *devlist = dv; | |
ca9de185 LM |
6593 | } |
6594 | } | |
6595 | if (fd >= 0) | |
6596 | close(fd); | |
6597 | } | |
6598 | } | |
6599 | free_mdstat(mdstat); | |
6600 | return count; | |
6601 | } | |
6602 | ||
6603 | #ifdef DEBUG_LOOP | |
6604 | static struct md_list* | |
6605 | get_loop_devices(void) | |
6606 | { | |
6607 | int i; | |
6608 | struct md_list *devlist = NULL; | |
594dc1b8 | 6609 | struct md_list *dv; |
ca9de185 LM |
6610 | |
6611 | for(i = 0; i < 12; i++) { | |
503975b9 N |
6612 | dv = xcalloc(1, sizeof(*dv)); |
6613 | dv->devname = xmalloc(40); | |
ca9de185 LM |
6614 | sprintf(dv->devname, "/dev/loop%d", i); |
6615 | dv->next = devlist; | |
6616 | devlist = dv; | |
6617 | } | |
6618 | return devlist; | |
6619 | } | |
6620 | #endif | |
6621 | ||
6622 | static struct md_list* | |
6623 | get_devices(const char *hba_path) | |
6624 | { | |
6625 | struct md_list *devlist = NULL; | |
594dc1b8 | 6626 | struct md_list *dv; |
ca9de185 LM |
6627 | struct dirent *ent; |
6628 | DIR *dir; | |
6629 | int err = 0; | |
6630 | ||
6631 | #if DEBUG_LOOP | |
6632 | devlist = get_loop_devices(); | |
6633 | return devlist; | |
6634 | #endif | |
6635 | /* scroll through /sys/dev/block looking for devices attached to | |
6636 | * this hba | |
6637 | */ | |
6638 | dir = opendir("/sys/dev/block"); | |
6639 | for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) { | |
6640 | int fd; | |
6641 | char buf[1024]; | |
6642 | int major, minor; | |
6643 | char *path = NULL; | |
6644 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
6645 | continue; | |
6646 | path = devt_to_devpath(makedev(major, minor)); | |
6647 | if (!path) | |
6648 | continue; | |
6649 | if (!path_attached_to_hba(path, hba_path)) { | |
6650 | free(path); | |
6651 | path = NULL; | |
6652 | continue; | |
6653 | } | |
6654 | free(path); | |
6655 | path = NULL; | |
6656 | fd = dev_open(ent->d_name, O_RDONLY); | |
6657 | if (fd >= 0) { | |
6658 | fd2devname(fd, buf); | |
6659 | close(fd); | |
6660 | } else { | |
e7b84f9d | 6661 | pr_err("cannot open device: %s\n", |
ca9de185 LM |
6662 | ent->d_name); |
6663 | continue; | |
6664 | } | |
6665 | ||
503975b9 N |
6666 | dv = xcalloc(1, sizeof(*dv)); |
6667 | dv->devname = xstrdup(buf); | |
ca9de185 LM |
6668 | dv->next = devlist; |
6669 | devlist = dv; | |
6670 | } | |
6671 | if (err) { | |
6672 | while(devlist) { | |
6673 | dv = devlist; | |
6674 | devlist = devlist->next; | |
6675 | free(dv->devname); | |
6676 | free(dv); | |
6677 | } | |
6678 | } | |
562aa102 | 6679 | closedir(dir); |
ca9de185 LM |
6680 | return devlist; |
6681 | } | |
6682 | ||
6683 | static int | |
6684 | count_volumes_list(struct md_list *devlist, char *homehost, | |
6685 | int verbose, int *found) | |
6686 | { | |
6687 | struct md_list *tmpdev; | |
6688 | int count = 0; | |
594dc1b8 | 6689 | struct supertype *st; |
ca9de185 LM |
6690 | |
6691 | /* first walk the list of devices to find a consistent set | |
6692 | * that match the criterea, if that is possible. | |
6693 | * We flag the ones we like with 'used'. | |
6694 | */ | |
6695 | *found = 0; | |
6696 | st = match_metadata_desc_imsm("imsm"); | |
6697 | if (st == NULL) { | |
676e87a8 | 6698 | pr_vrb("cannot allocate memory for imsm supertype\n"); |
ca9de185 LM |
6699 | return 0; |
6700 | } | |
6701 | ||
6702 | for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { | |
6703 | char *devname = tmpdev->devname; | |
0a6bff09 | 6704 | dev_t rdev; |
ca9de185 LM |
6705 | struct supertype *tst; |
6706 | int dfd; | |
6707 | if (tmpdev->used > 1) | |
6708 | continue; | |
6709 | tst = dup_super(st); | |
6710 | if (tst == NULL) { | |
676e87a8 | 6711 | pr_vrb("cannot allocate memory for imsm supertype\n"); |
ca9de185 LM |
6712 | goto err_1; |
6713 | } | |
6714 | tmpdev->container = 0; | |
6715 | dfd = dev_open(devname, O_RDONLY|O_EXCL); | |
6716 | if (dfd < 0) { | |
1ade5cc1 | 6717 | dprintf("cannot open device %s: %s\n", |
ca9de185 LM |
6718 | devname, strerror(errno)); |
6719 | tmpdev->used = 2; | |
0a6bff09 | 6720 | } else if (!fstat_is_blkdev(dfd, devname, &rdev)) { |
ca9de185 LM |
6721 | tmpdev->used = 2; |
6722 | } else if (must_be_container(dfd)) { | |
6723 | struct supertype *cst; | |
6724 | cst = super_by_fd(dfd, NULL); | |
6725 | if (cst == NULL) { | |
1ade5cc1 | 6726 | dprintf("cannot recognize container type %s\n", |
ca9de185 LM |
6727 | devname); |
6728 | tmpdev->used = 2; | |
6729 | } else if (tst->ss != st->ss) { | |
1ade5cc1 | 6730 | dprintf("non-imsm container - ignore it: %s\n", |
ca9de185 LM |
6731 | devname); |
6732 | tmpdev->used = 2; | |
6733 | } else if (!tst->ss->load_container || | |
6734 | tst->ss->load_container(tst, dfd, NULL)) | |
6735 | tmpdev->used = 2; | |
6736 | else { | |
6737 | tmpdev->container = 1; | |
6738 | } | |
6739 | if (cst) | |
6740 | cst->ss->free_super(cst); | |
6741 | } else { | |
0a6bff09 | 6742 | tmpdev->st_rdev = rdev; |
ca9de185 | 6743 | if (tst->ss->load_super(tst,dfd, NULL)) { |
1ade5cc1 | 6744 | dprintf("no RAID superblock on %s\n", |
ca9de185 LM |
6745 | devname); |
6746 | tmpdev->used = 2; | |
6747 | } else if (tst->ss->compare_super == NULL) { | |
1ade5cc1 | 6748 | dprintf("Cannot assemble %s metadata on %s\n", |
ca9de185 LM |
6749 | tst->ss->name, devname); |
6750 | tmpdev->used = 2; | |
6751 | } | |
6752 | } | |
6753 | if (dfd >= 0) | |
6754 | close(dfd); | |
6755 | if (tmpdev->used == 2 || tmpdev->used == 4) { | |
6756 | /* Ignore unrecognised devices during auto-assembly */ | |
6757 | goto loop; | |
6758 | } | |
6759 | else { | |
6760 | struct mdinfo info; | |
6761 | tst->ss->getinfo_super(tst, &info, NULL); | |
6762 | ||
6763 | if (st->minor_version == -1) | |
6764 | st->minor_version = tst->minor_version; | |
6765 | ||
6766 | if (memcmp(info.uuid, uuid_zero, | |
6767 | sizeof(int[4])) == 0) { | |
6768 | /* this is a floating spare. It cannot define | |
6769 | * an array unless there are no more arrays of | |
6770 | * this type to be found. It can be included | |
6771 | * in an array of this type though. | |
6772 | */ | |
6773 | tmpdev->used = 3; | |
6774 | goto loop; | |
6775 | } | |
6776 | ||
6777 | if (st->ss != tst->ss || | |
6778 | st->minor_version != tst->minor_version || | |
6779 | st->ss->compare_super(st, tst) != 0) { | |
6780 | /* Some mismatch. If exactly one array matches this host, | |
6781 | * we can resolve on that one. | |
6782 | * Or, if we are auto assembling, we just ignore the second | |
6783 | * for now. | |
6784 | */ | |
1ade5cc1 | 6785 | dprintf("superblock on %s doesn't match others - assembly aborted\n", |
ca9de185 LM |
6786 | devname); |
6787 | goto loop; | |
6788 | } | |
6789 | tmpdev->used = 1; | |
6790 | *found = 1; | |
6791 | dprintf("found: devname: %s\n", devname); | |
6792 | } | |
6793 | loop: | |
6794 | if (tst) | |
6795 | tst->ss->free_super(tst); | |
6796 | } | |
6797 | if (*found != 0) { | |
6798 | int err; | |
6799 | if ((err = load_super_imsm_all(st, -1, &st->sb, NULL, devlist, 0)) == 0) { | |
6800 | struct mdinfo *iter, *head = st->ss->container_content(st, NULL); | |
6801 | for (iter = head; iter; iter = iter->next) { | |
6802 | dprintf("content->text_version: %s vol\n", | |
6803 | iter->text_version); | |
6804 | if (iter->array.state & (1<<MD_SB_BLOCK_VOLUME)) { | |
6805 | /* do not assemble arrays with unsupported | |
6806 | configurations */ | |
1ade5cc1 | 6807 | dprintf("Cannot activate member %s.\n", |
ca9de185 LM |
6808 | iter->text_version); |
6809 | } else | |
6810 | count++; | |
6811 | } | |
6812 | sysfs_free(head); | |
6813 | ||
6814 | } else { | |
1ade5cc1 | 6815 | dprintf("No valid super block on device list: err: %d %p\n", |
ca9de185 LM |
6816 | err, st->sb); |
6817 | } | |
6818 | } else { | |
1ade5cc1 | 6819 | dprintf("no more devices to examine\n"); |
ca9de185 LM |
6820 | } |
6821 | ||
6822 | for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { | |
089f9d79 | 6823 | if (tmpdev->used == 1 && tmpdev->found) { |
ca9de185 LM |
6824 | if (count) { |
6825 | if (count < tmpdev->found) | |
6826 | count = 0; | |
6827 | else | |
6828 | count -= tmpdev->found; | |
6829 | } | |
6830 | } | |
6831 | if (tmpdev->used == 1) | |
6832 | tmpdev->used = 4; | |
6833 | } | |
6834 | err_1: | |
6835 | if (st) | |
6836 | st->ss->free_super(st); | |
6837 | return count; | |
6838 | } | |
6839 | ||
d3c11416 AO |
6840 | static int __count_volumes(char *hba_path, int dpa, int verbose, |
6841 | int cmp_hba_path) | |
ca9de185 | 6842 | { |
72a45777 | 6843 | struct sys_dev *idev, *intel_devices = find_intel_devices(); |
ca9de185 | 6844 | int count = 0; |
72a45777 PB |
6845 | const struct orom_entry *entry; |
6846 | struct devid_list *dv, *devid_list; | |
ca9de185 | 6847 | |
d3c11416 | 6848 | if (!hba_path) |
ca9de185 LM |
6849 | return 0; |
6850 | ||
72a45777 | 6851 | for (idev = intel_devices; idev; idev = idev->next) { |
d3c11416 AO |
6852 | if (strstr(idev->path, hba_path)) |
6853 | break; | |
72a45777 PB |
6854 | } |
6855 | ||
6856 | if (!idev || !idev->dev_id) | |
ca9de185 | 6857 | return 0; |
72a45777 PB |
6858 | |
6859 | entry = get_orom_entry_by_device_id(idev->dev_id); | |
6860 | ||
6861 | if (!entry || !entry->devid_list) | |
6862 | return 0; | |
6863 | ||
6864 | devid_list = entry->devid_list; | |
6865 | for (dv = devid_list; dv; dv = dv->next) { | |
594dc1b8 | 6866 | struct md_list *devlist; |
d3c11416 AO |
6867 | struct sys_dev *device = NULL; |
6868 | char *hpath; | |
72a45777 PB |
6869 | int found = 0; |
6870 | ||
d3c11416 AO |
6871 | if (cmp_hba_path) |
6872 | device = device_by_id_and_path(dv->devid, hba_path); | |
6873 | else | |
6874 | device = device_by_id(dv->devid); | |
6875 | ||
72a45777 | 6876 | if (device) |
d3c11416 | 6877 | hpath = device->path; |
72a45777 PB |
6878 | else |
6879 | return 0; | |
6880 | ||
d3c11416 | 6881 | devlist = get_devices(hpath); |
72a45777 PB |
6882 | /* if no intel devices return zero volumes */ |
6883 | if (devlist == NULL) | |
6884 | return 0; | |
6885 | ||
d3c11416 AO |
6886 | count += active_arrays_by_format("imsm", hpath, &devlist, dpa, |
6887 | verbose); | |
6888 | dprintf("path: %s active arrays: %d\n", hpath, count); | |
72a45777 PB |
6889 | if (devlist == NULL) |
6890 | return 0; | |
6891 | do { | |
6892 | found = 0; | |
6893 | count += count_volumes_list(devlist, | |
6894 | NULL, | |
6895 | verbose, | |
6896 | &found); | |
6897 | dprintf("found %d count: %d\n", found, count); | |
6898 | } while (found); | |
6899 | ||
d3c11416 | 6900 | dprintf("path: %s total number of volumes: %d\n", hpath, count); |
72a45777 PB |
6901 | |
6902 | while (devlist) { | |
6903 | struct md_list *dv = devlist; | |
6904 | devlist = devlist->next; | |
6905 | free(dv->devname); | |
6906 | free(dv); | |
6907 | } | |
ca9de185 LM |
6908 | } |
6909 | return count; | |
6910 | } | |
6911 | ||
d3c11416 AO |
6912 | static int count_volumes(struct intel_hba *hba, int dpa, int verbose) |
6913 | { | |
6914 | if (!hba) | |
6915 | return 0; | |
6916 | if (hba->type == SYS_DEV_VMD) { | |
6917 | struct sys_dev *dev; | |
6918 | int count = 0; | |
6919 | ||
6920 | for (dev = find_intel_devices(); dev; dev = dev->next) { | |
6921 | if (dev->type == SYS_DEV_VMD) | |
6922 | count += __count_volumes(dev->path, dpa, | |
6923 | verbose, 1); | |
6924 | } | |
6925 | return count; | |
6926 | } | |
6927 | return __count_volumes(hba->path, dpa, verbose, 0); | |
6928 | } | |
6929 | ||
cd9d1ac7 DW |
6930 | static int imsm_default_chunk(const struct imsm_orom *orom) |
6931 | { | |
6932 | /* up to 512 if the plaform supports it, otherwise the platform max. | |
6933 | * 128 if no platform detected | |
6934 | */ | |
6935 | int fs = max(7, orom ? fls(orom->sss) : 0); | |
6936 | ||
6937 | return min(512, (1 << fs)); | |
6938 | } | |
73408129 | 6939 | |
6592ce37 DW |
6940 | static int |
6941 | validate_geometry_imsm_orom(struct intel_super *super, int level, int layout, | |
2cc699af | 6942 | int raiddisks, int *chunk, unsigned long long size, int verbose) |
6592ce37 | 6943 | { |
660260d0 DW |
6944 | /* check/set platform and metadata limits/defaults */ |
6945 | if (super->orom && raiddisks > super->orom->dpa) { | |
676e87a8 | 6946 | pr_vrb("platform supports a maximum of %d disks per array\n", |
660260d0 | 6947 | super->orom->dpa); |
73408129 LM |
6948 | return 0; |
6949 | } | |
6950 | ||
5d500228 | 6951 | /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */ |
660260d0 | 6952 | if (!is_raid_level_supported(super->orom, level, raiddisks)) { |
676e87a8 | 6953 | pr_vrb("platform does not support raid%d with %d disk%s\n", |
6592ce37 DW |
6954 | level, raiddisks, raiddisks > 1 ? "s" : ""); |
6955 | return 0; | |
6956 | } | |
cd9d1ac7 | 6957 | |
7ccc4cc4 | 6958 | if (*chunk == 0 || *chunk == UnSet) |
cd9d1ac7 DW |
6959 | *chunk = imsm_default_chunk(super->orom); |
6960 | ||
7ccc4cc4 | 6961 | if (super->orom && !imsm_orom_has_chunk(super->orom, *chunk)) { |
676e87a8 | 6962 | pr_vrb("platform does not support a chunk size of: %d\n", *chunk); |
cd9d1ac7 | 6963 | return 0; |
6592ce37 | 6964 | } |
cd9d1ac7 | 6965 | |
6592ce37 DW |
6966 | if (layout != imsm_level_to_layout(level)) { |
6967 | if (level == 5) | |
676e87a8 | 6968 | pr_vrb("imsm raid 5 only supports the left-asymmetric layout\n"); |
6592ce37 | 6969 | else if (level == 10) |
676e87a8 | 6970 | pr_vrb("imsm raid 10 only supports the n2 layout\n"); |
6592ce37 | 6971 | else |
676e87a8 | 6972 | pr_vrb("imsm unknown layout %#x for this raid level %d\n", |
6592ce37 DW |
6973 | layout, level); |
6974 | return 0; | |
6975 | } | |
2cc699af | 6976 | |
7ccc4cc4 | 6977 | if (super->orom && (super->orom->attr & IMSM_OROM_ATTR_2TB) == 0 && |
2cc699af | 6978 | (calc_array_size(level, raiddisks, layout, *chunk, size) >> 32) > 0) { |
676e87a8 | 6979 | pr_vrb("platform does not support a volume size over 2TB\n"); |
2cc699af CA |
6980 | return 0; |
6981 | } | |
614902f6 | 6982 | |
6592ce37 DW |
6983 | return 1; |
6984 | } | |
6985 | ||
1011e834 | 6986 | /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd |
c2c087e6 DW |
6987 | * FIX ME add ahci details |
6988 | */ | |
8b353278 | 6989 | static int validate_geometry_imsm_volume(struct supertype *st, int level, |
c21e737b | 6990 | int layout, int raiddisks, int *chunk, |
af4348dd N |
6991 | unsigned long long size, |
6992 | unsigned long long data_offset, | |
6993 | char *dev, | |
2c514b71 NB |
6994 | unsigned long long *freesize, |
6995 | int verbose) | |
cdddbdbc | 6996 | { |
9e04ac1c | 6997 | dev_t rdev; |
c2c087e6 | 6998 | struct intel_super *super = st->sb; |
b2916f25 | 6999 | struct imsm_super *mpb; |
c2c087e6 DW |
7000 | struct dl *dl; |
7001 | unsigned long long pos = 0; | |
7002 | unsigned long long maxsize; | |
7003 | struct extent *e; | |
7004 | int i; | |
cdddbdbc | 7005 | |
88c32bb1 DW |
7006 | /* We must have the container info already read in. */ |
7007 | if (!super) | |
c2c087e6 DW |
7008 | return 0; |
7009 | ||
b2916f25 JS |
7010 | mpb = super->anchor; |
7011 | ||
2cc699af | 7012 | if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, size, verbose)) { |
7a862a02 | 7013 | pr_err("RAID gemetry validation failed. Cannot proceed with the action(s).\n"); |
c2c087e6 | 7014 | return 0; |
d54559f0 | 7015 | } |
c2c087e6 DW |
7016 | if (!dev) { |
7017 | /* General test: make sure there is space for | |
2da8544a DW |
7018 | * 'raiddisks' device extents of size 'size' at a given |
7019 | * offset | |
c2c087e6 | 7020 | */ |
e46273eb | 7021 | unsigned long long minsize = size; |
b7528a20 | 7022 | unsigned long long start_offset = MaxSector; |
c2c087e6 DW |
7023 | int dcnt = 0; |
7024 | if (minsize == 0) | |
7025 | minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
7026 | for (dl = super->disks; dl ; dl = dl->next) { | |
7027 | int found = 0; | |
7028 | ||
bf5a934a | 7029 | pos = 0; |
c2c087e6 DW |
7030 | i = 0; |
7031 | e = get_extents(super, dl); | |
7032 | if (!e) continue; | |
7033 | do { | |
7034 | unsigned long long esize; | |
7035 | esize = e[i].start - pos; | |
7036 | if (esize >= minsize) | |
7037 | found = 1; | |
b7528a20 | 7038 | if (found && start_offset == MaxSector) { |
2da8544a DW |
7039 | start_offset = pos; |
7040 | break; | |
7041 | } else if (found && pos != start_offset) { | |
7042 | found = 0; | |
7043 | break; | |
7044 | } | |
c2c087e6 DW |
7045 | pos = e[i].start + e[i].size; |
7046 | i++; | |
7047 | } while (e[i-1].size); | |
7048 | if (found) | |
7049 | dcnt++; | |
7050 | free(e); | |
7051 | } | |
7052 | if (dcnt < raiddisks) { | |
2c514b71 | 7053 | if (verbose) |
7a862a02 | 7054 | pr_err("imsm: Not enough devices with space for this array (%d < %d)\n", |
2c514b71 | 7055 | dcnt, raiddisks); |
c2c087e6 DW |
7056 | return 0; |
7057 | } | |
7058 | return 1; | |
7059 | } | |
0dcecb2e | 7060 | |
c2c087e6 | 7061 | /* This device must be a member of the set */ |
9e04ac1c | 7062 | if (!stat_is_blkdev(dev, &rdev)) |
c2c087e6 DW |
7063 | return 0; |
7064 | for (dl = super->disks ; dl ; dl = dl->next) { | |
9e04ac1c ZL |
7065 | if (dl->major == (int)major(rdev) && |
7066 | dl->minor == (int)minor(rdev)) | |
c2c087e6 DW |
7067 | break; |
7068 | } | |
7069 | if (!dl) { | |
2c514b71 | 7070 | if (verbose) |
7a862a02 | 7071 | pr_err("%s is not in the same imsm set\n", dev); |
c2c087e6 | 7072 | return 0; |
a20d2ba5 DW |
7073 | } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) { |
7074 | /* If a volume is present then the current creation attempt | |
7075 | * cannot incorporate new spares because the orom may not | |
7076 | * understand this configuration (all member disks must be | |
7077 | * members of each array in the container). | |
7078 | */ | |
7a862a02 N |
7079 | pr_err("%s is a spare and a volume is already defined for this container\n", dev); |
7080 | pr_err("The option-rom requires all member disks to be a member of all volumes\n"); | |
a20d2ba5 | 7081 | return 0; |
5fe62b94 WD |
7082 | } else if (super->orom && mpb->num_raid_devs > 0 && |
7083 | mpb->num_disks != raiddisks) { | |
7a862a02 | 7084 | pr_err("The option-rom requires all member disks to be a member of all volumes\n"); |
5fe62b94 | 7085 | return 0; |
c2c087e6 | 7086 | } |
0dcecb2e DW |
7087 | |
7088 | /* retrieve the largest free space block */ | |
c2c087e6 DW |
7089 | e = get_extents(super, dl); |
7090 | maxsize = 0; | |
7091 | i = 0; | |
0dcecb2e DW |
7092 | if (e) { |
7093 | do { | |
7094 | unsigned long long esize; | |
7095 | ||
7096 | esize = e[i].start - pos; | |
7097 | if (esize >= maxsize) | |
7098 | maxsize = esize; | |
7099 | pos = e[i].start + e[i].size; | |
7100 | i++; | |
7101 | } while (e[i-1].size); | |
7102 | dl->e = e; | |
7103 | dl->extent_cnt = i; | |
7104 | } else { | |
7105 | if (verbose) | |
e7b84f9d | 7106 | pr_err("unable to determine free space for: %s\n", |
0dcecb2e DW |
7107 | dev); |
7108 | return 0; | |
7109 | } | |
7110 | if (maxsize < size) { | |
7111 | if (verbose) | |
e7b84f9d | 7112 | pr_err("%s not enough space (%llu < %llu)\n", |
0dcecb2e DW |
7113 | dev, maxsize, size); |
7114 | return 0; | |
7115 | } | |
7116 | ||
7117 | /* count total number of extents for merge */ | |
7118 | i = 0; | |
7119 | for (dl = super->disks; dl; dl = dl->next) | |
7120 | if (dl->e) | |
7121 | i += dl->extent_cnt; | |
7122 | ||
7123 | maxsize = merge_extents(super, i); | |
3baa56ab LO |
7124 | |
7125 | if (!check_env("IMSM_NO_PLATFORM") && | |
7126 | mpb->num_raid_devs > 0 && size && size != maxsize) { | |
7a862a02 | 7127 | pr_err("attempting to create a second volume with size less then remaining space. Aborting...\n"); |
3baa56ab LO |
7128 | return 0; |
7129 | } | |
7130 | ||
a7dd165b | 7131 | if (maxsize < size || maxsize == 0) { |
b3071342 LD |
7132 | if (verbose) { |
7133 | if (maxsize == 0) | |
7a862a02 | 7134 | pr_err("no free space left on device. Aborting...\n"); |
b3071342 | 7135 | else |
7a862a02 | 7136 | pr_err("not enough space to create volume of given size (%llu < %llu). Aborting...\n", |
b3071342 LD |
7137 | maxsize, size); |
7138 | } | |
0dcecb2e | 7139 | return 0; |
0dcecb2e DW |
7140 | } |
7141 | ||
c2c087e6 DW |
7142 | *freesize = maxsize; |
7143 | ||
ca9de185 | 7144 | if (super->orom) { |
72a45777 | 7145 | int count = count_volumes(super->hba, |
ca9de185 LM |
7146 | super->orom->dpa, verbose); |
7147 | if (super->orom->vphba <= count) { | |
676e87a8 | 7148 | pr_vrb("platform does not support more than %d raid volumes.\n", |
ca9de185 LM |
7149 | super->orom->vphba); |
7150 | return 0; | |
7151 | } | |
7152 | } | |
c2c087e6 | 7153 | return 1; |
cdddbdbc DW |
7154 | } |
7155 | ||
13bcac90 | 7156 | static int imsm_get_free_size(struct supertype *st, int raiddisks, |
efb30e7f DW |
7157 | unsigned long long size, int chunk, |
7158 | unsigned long long *freesize) | |
7159 | { | |
7160 | struct intel_super *super = st->sb; | |
7161 | struct imsm_super *mpb = super->anchor; | |
7162 | struct dl *dl; | |
7163 | int i; | |
7164 | int extent_cnt; | |
7165 | struct extent *e; | |
7166 | unsigned long long maxsize; | |
7167 | unsigned long long minsize; | |
7168 | int cnt; | |
7169 | int used; | |
7170 | ||
7171 | /* find the largest common start free region of the possible disks */ | |
7172 | used = 0; | |
7173 | extent_cnt = 0; | |
7174 | cnt = 0; | |
7175 | for (dl = super->disks; dl; dl = dl->next) { | |
7176 | dl->raiddisk = -1; | |
7177 | ||
7178 | if (dl->index >= 0) | |
7179 | used++; | |
7180 | ||
7181 | /* don't activate new spares if we are orom constrained | |
7182 | * and there is already a volume active in the container | |
7183 | */ | |
7184 | if (super->orom && dl->index < 0 && mpb->num_raid_devs) | |
7185 | continue; | |
7186 | ||
7187 | e = get_extents(super, dl); | |
7188 | if (!e) | |
7189 | continue; | |
7190 | for (i = 1; e[i-1].size; i++) | |
7191 | ; | |
7192 | dl->e = e; | |
7193 | dl->extent_cnt = i; | |
7194 | extent_cnt += i; | |
7195 | cnt++; | |
7196 | } | |
7197 | ||
7198 | maxsize = merge_extents(super, extent_cnt); | |
7199 | minsize = size; | |
7200 | if (size == 0) | |
612e59d8 CA |
7201 | /* chunk is in K */ |
7202 | minsize = chunk * 2; | |
efb30e7f DW |
7203 | |
7204 | if (cnt < raiddisks || | |
7205 | (super->orom && used && used != raiddisks) || | |
a7dd165b DW |
7206 | maxsize < minsize || |
7207 | maxsize == 0) { | |
e7b84f9d | 7208 | pr_err("not enough devices with space to create array.\n"); |
efb30e7f DW |
7209 | return 0; /* No enough free spaces large enough */ |
7210 | } | |
7211 | ||
7212 | if (size == 0) { | |
7213 | size = maxsize; | |
7214 | if (chunk) { | |
612e59d8 CA |
7215 | size /= 2 * chunk; |
7216 | size *= 2 * chunk; | |
efb30e7f | 7217 | } |
f878b242 LM |
7218 | maxsize = size; |
7219 | } | |
7220 | if (!check_env("IMSM_NO_PLATFORM") && | |
7221 | mpb->num_raid_devs > 0 && size && size != maxsize) { | |
7a862a02 | 7222 | pr_err("attempting to create a second volume with size less then remaining space. Aborting...\n"); |
f878b242 | 7223 | return 0; |
efb30e7f | 7224 | } |
efb30e7f DW |
7225 | cnt = 0; |
7226 | for (dl = super->disks; dl; dl = dl->next) | |
7227 | if (dl->e) | |
7228 | dl->raiddisk = cnt++; | |
7229 | ||
7230 | *freesize = size; | |
7231 | ||
13bcac90 AK |
7232 | dprintf("imsm: imsm_get_free_size() returns : %llu\n", size); |
7233 | ||
efb30e7f DW |
7234 | return 1; |
7235 | } | |
7236 | ||
13bcac90 AK |
7237 | static int reserve_space(struct supertype *st, int raiddisks, |
7238 | unsigned long long size, int chunk, | |
7239 | unsigned long long *freesize) | |
7240 | { | |
7241 | struct intel_super *super = st->sb; | |
7242 | struct dl *dl; | |
7243 | int cnt; | |
7244 | int rv = 0; | |
7245 | ||
7246 | rv = imsm_get_free_size(st, raiddisks, size, chunk, freesize); | |
7247 | if (rv) { | |
7248 | cnt = 0; | |
7249 | for (dl = super->disks; dl; dl = dl->next) | |
7250 | if (dl->e) | |
7251 | dl->raiddisk = cnt++; | |
7252 | rv = 1; | |
7253 | } | |
7254 | ||
7255 | return rv; | |
7256 | } | |
7257 | ||
bf5a934a | 7258 | static int validate_geometry_imsm(struct supertype *st, int level, int layout, |
c21e737b | 7259 | int raiddisks, int *chunk, unsigned long long size, |
af4348dd | 7260 | unsigned long long data_offset, |
bf5a934a | 7261 | char *dev, unsigned long long *freesize, |
5308f117 | 7262 | int consistency_policy, int verbose) |
bf5a934a DW |
7263 | { |
7264 | int fd, cfd; | |
7265 | struct mdinfo *sra; | |
20cbe8d2 | 7266 | int is_member = 0; |
bf5a934a | 7267 | |
d54559f0 LM |
7268 | /* load capability |
7269 | * if given unused devices create a container | |
bf5a934a DW |
7270 | * if given given devices in a container create a member volume |
7271 | */ | |
7272 | if (level == LEVEL_CONTAINER) { | |
7273 | /* Must be a fresh device to add to a container */ | |
7274 | return validate_geometry_imsm_container(st, level, layout, | |
c21e737b | 7275 | raiddisks, |
7ccc4cc4 | 7276 | *chunk, |
af4348dd | 7277 | size, data_offset, |
bf5a934a DW |
7278 | dev, freesize, |
7279 | verbose); | |
7280 | } | |
9587c373 | 7281 | |
8592f29d | 7282 | if (!dev) { |
e91a3bad | 7283 | if (st->sb) { |
ca9de185 | 7284 | struct intel_super *super = st->sb; |
e91a3bad | 7285 | if (!validate_geometry_imsm_orom(st->sb, level, layout, |
2cc699af | 7286 | raiddisks, chunk, size, |
e91a3bad LM |
7287 | verbose)) |
7288 | return 0; | |
efb30e7f DW |
7289 | /* we are being asked to automatically layout a |
7290 | * new volume based on the current contents of | |
7291 | * the container. If the the parameters can be | |
7292 | * satisfied reserve_space will record the disks, | |
7293 | * start offset, and size of the volume to be | |
7294 | * created. add_to_super and getinfo_super | |
7295 | * detect when autolayout is in progress. | |
7296 | */ | |
ca9de185 LM |
7297 | /* assuming that freesize is always given when array is |
7298 | created */ | |
7299 | if (super->orom && freesize) { | |
7300 | int count; | |
72a45777 | 7301 | count = count_volumes(super->hba, |
ca9de185 LM |
7302 | super->orom->dpa, verbose); |
7303 | if (super->orom->vphba <= count) { | |
676e87a8 | 7304 | pr_vrb("platform does not support more than %d raid volumes.\n", |
ca9de185 LM |
7305 | super->orom->vphba); |
7306 | return 0; | |
7307 | } | |
7308 | } | |
e91a3bad LM |
7309 | if (freesize) |
7310 | return reserve_space(st, raiddisks, size, | |
7ccc4cc4 | 7311 | *chunk, freesize); |
8592f29d N |
7312 | } |
7313 | return 1; | |
7314 | } | |
bf5a934a DW |
7315 | if (st->sb) { |
7316 | /* creating in a given container */ | |
7317 | return validate_geometry_imsm_volume(st, level, layout, | |
7318 | raiddisks, chunk, size, | |
af4348dd | 7319 | data_offset, |
bf5a934a DW |
7320 | dev, freesize, verbose); |
7321 | } | |
7322 | ||
bf5a934a DW |
7323 | /* This device needs to be a device in an 'imsm' container */ |
7324 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
7325 | if (fd >= 0) { | |
7326 | if (verbose) | |
e7b84f9d N |
7327 | pr_err("Cannot create this array on device %s\n", |
7328 | dev); | |
bf5a934a DW |
7329 | close(fd); |
7330 | return 0; | |
7331 | } | |
7332 | if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) { | |
7333 | if (verbose) | |
e7b84f9d | 7334 | pr_err("Cannot open %s: %s\n", |
bf5a934a DW |
7335 | dev, strerror(errno)); |
7336 | return 0; | |
7337 | } | |
7338 | /* Well, it is in use by someone, maybe an 'imsm' container. */ | |
7339 | cfd = open_container(fd); | |
20cbe8d2 | 7340 | close(fd); |
bf5a934a | 7341 | if (cfd < 0) { |
bf5a934a | 7342 | if (verbose) |
e7b84f9d | 7343 | pr_err("Cannot use %s: It is busy\n", |
bf5a934a DW |
7344 | dev); |
7345 | return 0; | |
7346 | } | |
4dd2df09 | 7347 | sra = sysfs_read(cfd, NULL, GET_VERSION); |
bf5a934a | 7348 | if (sra && sra->array.major_version == -1 && |
20cbe8d2 AW |
7349 | strcmp(sra->text_version, "imsm") == 0) |
7350 | is_member = 1; | |
7351 | sysfs_free(sra); | |
7352 | if (is_member) { | |
bf5a934a DW |
7353 | /* This is a member of a imsm container. Load the container |
7354 | * and try to create a volume | |
7355 | */ | |
7356 | struct intel_super *super; | |
7357 | ||
ec50f7b6 | 7358 | if (load_super_imsm_all(st, cfd, (void **) &super, NULL, NULL, 1) == 0) { |
bf5a934a | 7359 | st->sb = super; |
4dd2df09 | 7360 | strcpy(st->container_devnm, fd2devnm(cfd)); |
bf5a934a DW |
7361 | close(cfd); |
7362 | return validate_geometry_imsm_volume(st, level, layout, | |
7363 | raiddisks, chunk, | |
af4348dd | 7364 | size, data_offset, dev, |
ecbd9e81 N |
7365 | freesize, 1) |
7366 | ? 1 : -1; | |
bf5a934a | 7367 | } |
20cbe8d2 | 7368 | } |
bf5a934a | 7369 | |
20cbe8d2 | 7370 | if (verbose) |
e7b84f9d | 7371 | pr_err("failed container membership check\n"); |
20cbe8d2 AW |
7372 | |
7373 | close(cfd); | |
7374 | return 0; | |
bf5a934a | 7375 | } |
0bd16cf2 | 7376 | |
30f58b22 | 7377 | static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk) |
0bd16cf2 DJ |
7378 | { |
7379 | struct intel_super *super = st->sb; | |
7380 | ||
30f58b22 DW |
7381 | if (level && *level == UnSet) |
7382 | *level = LEVEL_CONTAINER; | |
7383 | ||
7384 | if (level && layout && *layout == UnSet) | |
7385 | *layout = imsm_level_to_layout(*level); | |
0bd16cf2 | 7386 | |
cd9d1ac7 DW |
7387 | if (chunk && (*chunk == UnSet || *chunk == 0)) |
7388 | *chunk = imsm_default_chunk(super->orom); | |
0bd16cf2 DJ |
7389 | } |
7390 | ||
33414a01 DW |
7391 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev); |
7392 | ||
7393 | static int kill_subarray_imsm(struct supertype *st) | |
7394 | { | |
7395 | /* remove the subarray currently referenced by ->current_vol */ | |
7396 | __u8 i; | |
7397 | struct intel_dev **dp; | |
7398 | struct intel_super *super = st->sb; | |
7399 | __u8 current_vol = super->current_vol; | |
7400 | struct imsm_super *mpb = super->anchor; | |
7401 | ||
7402 | if (super->current_vol < 0) | |
7403 | return 2; | |
7404 | super->current_vol = -1; /* invalidate subarray cursor */ | |
7405 | ||
7406 | /* block deletions that would change the uuid of active subarrays | |
7407 | * | |
7408 | * FIXME when immutable ids are available, but note that we'll | |
7409 | * also need to fixup the invalidated/active subarray indexes in | |
7410 | * mdstat | |
7411 | */ | |
7412 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
7413 | char subarray[4]; | |
7414 | ||
7415 | if (i < current_vol) | |
7416 | continue; | |
7417 | sprintf(subarray, "%u", i); | |
4dd2df09 | 7418 | if (is_subarray_active(subarray, st->devnm)) { |
e7b84f9d N |
7419 | pr_err("deleting subarray-%d would change the UUID of active subarray-%d, aborting\n", |
7420 | current_vol, i); | |
33414a01 DW |
7421 | |
7422 | return 2; | |
7423 | } | |
7424 | } | |
7425 | ||
7426 | if (st->update_tail) { | |
503975b9 | 7427 | struct imsm_update_kill_array *u = xmalloc(sizeof(*u)); |
33414a01 | 7428 | |
33414a01 DW |
7429 | u->type = update_kill_array; |
7430 | u->dev_idx = current_vol; | |
7431 | append_metadata_update(st, u, sizeof(*u)); | |
7432 | ||
7433 | return 0; | |
7434 | } | |
7435 | ||
7436 | for (dp = &super->devlist; *dp;) | |
7437 | if ((*dp)->index == current_vol) { | |
7438 | *dp = (*dp)->next; | |
7439 | } else { | |
7440 | handle_missing(super, (*dp)->dev); | |
7441 | if ((*dp)->index > current_vol) | |
7442 | (*dp)->index--; | |
7443 | dp = &(*dp)->next; | |
7444 | } | |
7445 | ||
7446 | /* no more raid devices, all active components are now spares, | |
7447 | * but of course failed are still failed | |
7448 | */ | |
7449 | if (--mpb->num_raid_devs == 0) { | |
7450 | struct dl *d; | |
7451 | ||
7452 | for (d = super->disks; d; d = d->next) | |
a8619d23 AK |
7453 | if (d->index > -2) |
7454 | mark_spare(d); | |
33414a01 DW |
7455 | } |
7456 | ||
7457 | super->updates_pending++; | |
7458 | ||
7459 | return 0; | |
7460 | } | |
aa534678 | 7461 | |
a951a4f7 | 7462 | static int update_subarray_imsm(struct supertype *st, char *subarray, |
fa56eddb | 7463 | char *update, struct mddev_ident *ident) |
aa534678 DW |
7464 | { |
7465 | /* update the subarray currently referenced by ->current_vol */ | |
7466 | struct intel_super *super = st->sb; | |
7467 | struct imsm_super *mpb = super->anchor; | |
7468 | ||
aa534678 DW |
7469 | if (strcmp(update, "name") == 0) { |
7470 | char *name = ident->name; | |
a951a4f7 N |
7471 | char *ep; |
7472 | int vol; | |
aa534678 | 7473 | |
4dd2df09 | 7474 | if (is_subarray_active(subarray, st->devnm)) { |
e7b84f9d | 7475 | pr_err("Unable to update name of active subarray\n"); |
aa534678 DW |
7476 | return 2; |
7477 | } | |
7478 | ||
7479 | if (!check_name(super, name, 0)) | |
7480 | return 2; | |
7481 | ||
a951a4f7 N |
7482 | vol = strtoul(subarray, &ep, 10); |
7483 | if (*ep != '\0' || vol >= super->anchor->num_raid_devs) | |
7484 | return 2; | |
7485 | ||
aa534678 | 7486 | if (st->update_tail) { |
503975b9 | 7487 | struct imsm_update_rename_array *u = xmalloc(sizeof(*u)); |
aa534678 | 7488 | |
aa534678 | 7489 | u->type = update_rename_array; |
a951a4f7 | 7490 | u->dev_idx = vol; |
618f4e6d XN |
7491 | strncpy((char *) u->name, name, MAX_RAID_SERIAL_LEN); |
7492 | u->name[MAX_RAID_SERIAL_LEN-1] = '\0'; | |
aa534678 DW |
7493 | append_metadata_update(st, u, sizeof(*u)); |
7494 | } else { | |
7495 | struct imsm_dev *dev; | |
7496 | int i; | |
7497 | ||
a951a4f7 | 7498 | dev = get_imsm_dev(super, vol); |
618f4e6d XN |
7499 | strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN); |
7500 | dev->volume[MAX_RAID_SERIAL_LEN-1] = '\0'; | |
aa534678 DW |
7501 | for (i = 0; i < mpb->num_raid_devs; i++) { |
7502 | dev = get_imsm_dev(super, i); | |
7503 | handle_missing(super, dev); | |
7504 | } | |
7505 | super->updates_pending++; | |
7506 | } | |
e6e9dd3f AP |
7507 | } else if (strcmp(update, "ppl") == 0 || |
7508 | strcmp(update, "no-ppl") == 0) { | |
7509 | int new_policy; | |
7510 | char *ep; | |
7511 | int vol = strtoul(subarray, &ep, 10); | |
7512 | ||
7513 | if (*ep != '\0' || vol >= super->anchor->num_raid_devs) | |
7514 | return 2; | |
7515 | ||
7516 | if (strcmp(update, "ppl") == 0) | |
c2462068 | 7517 | new_policy = RWH_MULTIPLE_DISTRIBUTED; |
e6e9dd3f | 7518 | else |
c2462068 | 7519 | new_policy = RWH_MULTIPLE_OFF; |
e6e9dd3f AP |
7520 | |
7521 | if (st->update_tail) { | |
7522 | struct imsm_update_rwh_policy *u = xmalloc(sizeof(*u)); | |
7523 | ||
7524 | u->type = update_rwh_policy; | |
7525 | u->dev_idx = vol; | |
7526 | u->new_policy = new_policy; | |
7527 | append_metadata_update(st, u, sizeof(*u)); | |
7528 | } else { | |
7529 | struct imsm_dev *dev; | |
7530 | ||
7531 | dev = get_imsm_dev(super, vol); | |
7532 | dev->rwh_policy = new_policy; | |
7533 | super->updates_pending++; | |
7534 | } | |
aa534678 DW |
7535 | } else |
7536 | return 2; | |
7537 | ||
7538 | return 0; | |
7539 | } | |
bf5a934a | 7540 | |
28bce06f AK |
7541 | static int is_gen_migration(struct imsm_dev *dev) |
7542 | { | |
7534230b AK |
7543 | if (dev == NULL) |
7544 | return 0; | |
7545 | ||
28bce06f AK |
7546 | if (!dev->vol.migr_state) |
7547 | return 0; | |
7548 | ||
7549 | if (migr_type(dev) == MIGR_GEN_MIGR) | |
7550 | return 1; | |
7551 | ||
7552 | return 0; | |
7553 | } | |
7554 | ||
1e5c6983 DW |
7555 | static int is_rebuilding(struct imsm_dev *dev) |
7556 | { | |
7557 | struct imsm_map *migr_map; | |
7558 | ||
7559 | if (!dev->vol.migr_state) | |
7560 | return 0; | |
7561 | ||
7562 | if (migr_type(dev) != MIGR_REBUILD) | |
7563 | return 0; | |
7564 | ||
238c0a71 | 7565 | migr_map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
7566 | |
7567 | if (migr_map->map_state == IMSM_T_STATE_DEGRADED) | |
7568 | return 1; | |
7569 | else | |
7570 | return 0; | |
7571 | } | |
7572 | ||
6ce1fbf1 AK |
7573 | static int is_initializing(struct imsm_dev *dev) |
7574 | { | |
7575 | struct imsm_map *migr_map; | |
7576 | ||
7577 | if (!dev->vol.migr_state) | |
7578 | return 0; | |
7579 | ||
7580 | if (migr_type(dev) != MIGR_INIT) | |
7581 | return 0; | |
7582 | ||
238c0a71 | 7583 | migr_map = get_imsm_map(dev, MAP_1); |
6ce1fbf1 AK |
7584 | |
7585 | if (migr_map->map_state == IMSM_T_STATE_UNINITIALIZED) | |
7586 | return 1; | |
7587 | ||
7588 | return 0; | |
6ce1fbf1 AK |
7589 | } |
7590 | ||
c47b0ff6 AK |
7591 | static void update_recovery_start(struct intel_super *super, |
7592 | struct imsm_dev *dev, | |
7593 | struct mdinfo *array) | |
1e5c6983 DW |
7594 | { |
7595 | struct mdinfo *rebuild = NULL; | |
7596 | struct mdinfo *d; | |
7597 | __u32 units; | |
7598 | ||
7599 | if (!is_rebuilding(dev)) | |
7600 | return; | |
7601 | ||
7602 | /* Find the rebuild target, but punt on the dual rebuild case */ | |
7603 | for (d = array->devs; d; d = d->next) | |
7604 | if (d->recovery_start == 0) { | |
7605 | if (rebuild) | |
7606 | return; | |
7607 | rebuild = d; | |
7608 | } | |
7609 | ||
4363fd80 DW |
7610 | if (!rebuild) { |
7611 | /* (?) none of the disks are marked with | |
7612 | * IMSM_ORD_REBUILD, so assume they are missing and the | |
7613 | * disk_ord_tbl was not correctly updated | |
7614 | */ | |
1ade5cc1 | 7615 | dprintf("failed to locate out-of-sync disk\n"); |
4363fd80 DW |
7616 | return; |
7617 | } | |
7618 | ||
1e5c6983 | 7619 | units = __le32_to_cpu(dev->vol.curr_migr_unit); |
c47b0ff6 | 7620 | rebuild->recovery_start = units * blocks_per_migr_unit(super, dev); |
1e5c6983 DW |
7621 | } |
7622 | ||
276d77db | 7623 | static int recover_backup_imsm(struct supertype *st, struct mdinfo *info); |
1e5c6983 | 7624 | |
00bbdbda | 7625 | static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray) |
cdddbdbc | 7626 | { |
4f5bc454 DW |
7627 | /* Given a container loaded by load_super_imsm_all, |
7628 | * extract information about all the arrays into | |
7629 | * an mdinfo tree. | |
00bbdbda | 7630 | * If 'subarray' is given, just extract info about that array. |
4f5bc454 DW |
7631 | * |
7632 | * For each imsm_dev create an mdinfo, fill it in, | |
7633 | * then look for matching devices in super->disks | |
7634 | * and create appropriate device mdinfo. | |
7635 | */ | |
7636 | struct intel_super *super = st->sb; | |
949c47a0 | 7637 | struct imsm_super *mpb = super->anchor; |
4f5bc454 | 7638 | struct mdinfo *rest = NULL; |
00bbdbda | 7639 | unsigned int i; |
81219e70 | 7640 | int sb_errors = 0; |
abef11a3 AK |
7641 | struct dl *d; |
7642 | int spare_disks = 0; | |
cdddbdbc | 7643 | |
19482bcc AK |
7644 | /* do not assemble arrays when not all attributes are supported */ |
7645 | if (imsm_check_attributes(mpb->attributes) == 0) { | |
81219e70 | 7646 | sb_errors = 1; |
7a862a02 | 7647 | pr_err("Unsupported attributes in IMSM metadata.Arrays activation is blocked.\n"); |
19482bcc AK |
7648 | } |
7649 | ||
abef11a3 AK |
7650 | /* count spare devices, not used in maps |
7651 | */ | |
7652 | for (d = super->disks; d; d = d->next) | |
7653 | if (d->index == -1) | |
7654 | spare_disks++; | |
7655 | ||
4f5bc454 | 7656 | for (i = 0; i < mpb->num_raid_devs; i++) { |
00bbdbda N |
7657 | struct imsm_dev *dev; |
7658 | struct imsm_map *map; | |
86e3692b | 7659 | struct imsm_map *map2; |
4f5bc454 | 7660 | struct mdinfo *this; |
a6482415 | 7661 | int slot; |
a6482415 | 7662 | int chunk; |
00bbdbda | 7663 | char *ep; |
8b9cd157 | 7664 | int level; |
00bbdbda N |
7665 | |
7666 | if (subarray && | |
7667 | (i != strtoul(subarray, &ep, 10) || *ep != '\0')) | |
7668 | continue; | |
7669 | ||
7670 | dev = get_imsm_dev(super, i); | |
238c0a71 AK |
7671 | map = get_imsm_map(dev, MAP_0); |
7672 | map2 = get_imsm_map(dev, MAP_1); | |
8b9cd157 | 7673 | level = get_imsm_raid_level(map); |
4f5bc454 | 7674 | |
1ce0101c DW |
7675 | /* do not publish arrays that are in the middle of an |
7676 | * unsupported migration | |
7677 | */ | |
7678 | if (dev->vol.migr_state && | |
28bce06f | 7679 | (migr_type(dev) == MIGR_STATE_CHANGE)) { |
7a862a02 | 7680 | pr_err("cannot assemble volume '%.16s': unsupported migration in progress\n", |
1ce0101c DW |
7681 | dev->volume); |
7682 | continue; | |
7683 | } | |
2db86302 LM |
7684 | /* do not publish arrays that are not support by controller's |
7685 | * OROM/EFI | |
7686 | */ | |
1ce0101c | 7687 | |
503975b9 | 7688 | this = xmalloc(sizeof(*this)); |
4f5bc454 | 7689 | |
301406c9 | 7690 | super->current_vol = i; |
a5d85af7 | 7691 | getinfo_super_imsm_volume(st, this, NULL); |
9894ec0d | 7692 | this->next = rest; |
a6482415 | 7693 | chunk = __le16_to_cpu(map->blocks_per_strip) >> 1; |
81219e70 LM |
7694 | /* mdadm does not support all metadata features- set the bit in all arrays state */ |
7695 | if (!validate_geometry_imsm_orom(super, | |
8b9cd157 MK |
7696 | level, /* RAID level */ |
7697 | imsm_level_to_layout(level), | |
81219e70 | 7698 | map->num_members, /* raid disks */ |
2cc699af | 7699 | &chunk, join_u32(dev->size_low, dev->size_high), |
81219e70 | 7700 | 1 /* verbose */)) { |
7a862a02 | 7701 | pr_err("IMSM RAID geometry validation failed. Array %s activation is blocked.\n", |
81219e70 LM |
7702 | dev->volume); |
7703 | this->array.state |= | |
7704 | (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) | | |
7705 | (1<<MD_SB_BLOCK_VOLUME); | |
7706 | } | |
81219e70 LM |
7707 | |
7708 | /* if array has bad blocks, set suitable bit in all arrays state */ | |
7709 | if (sb_errors) | |
7710 | this->array.state |= | |
7711 | (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) | | |
7712 | (1<<MD_SB_BLOCK_VOLUME); | |
7713 | ||
4f5bc454 | 7714 | for (slot = 0 ; slot < map->num_members; slot++) { |
1e5c6983 | 7715 | unsigned long long recovery_start; |
4f5bc454 DW |
7716 | struct mdinfo *info_d; |
7717 | struct dl *d; | |
7718 | int idx; | |
9a1608e5 | 7719 | int skip; |
7eef0453 | 7720 | __u32 ord; |
8b9cd157 | 7721 | int missing = 0; |
4f5bc454 | 7722 | |
9a1608e5 | 7723 | skip = 0; |
238c0a71 AK |
7724 | idx = get_imsm_disk_idx(dev, slot, MAP_0); |
7725 | ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); | |
4f5bc454 DW |
7726 | for (d = super->disks; d ; d = d->next) |
7727 | if (d->index == idx) | |
0fbd635c | 7728 | break; |
4f5bc454 | 7729 | |
1e5c6983 | 7730 | recovery_start = MaxSector; |
4f5bc454 | 7731 | if (d == NULL) |
9a1608e5 | 7732 | skip = 1; |
25ed7e59 | 7733 | if (d && is_failed(&d->disk)) |
9a1608e5 | 7734 | skip = 1; |
8b9cd157 | 7735 | if (!skip && (ord & IMSM_ORD_REBUILD)) |
1e5c6983 | 7736 | recovery_start = 0; |
9a1608e5 | 7737 | |
1011e834 | 7738 | /* |
9a1608e5 | 7739 | * if we skip some disks the array will be assmebled degraded; |
1e5c6983 DW |
7740 | * reset resync start to avoid a dirty-degraded |
7741 | * situation when performing the intial sync | |
9a1608e5 | 7742 | */ |
8b9cd157 MK |
7743 | if (skip) |
7744 | missing++; | |
7745 | ||
7746 | if (!(dev->vol.dirty & RAIDVOL_DIRTY)) { | |
7747 | if ((!able_to_resync(level, missing) || | |
7748 | recovery_start == 0)) | |
7749 | this->resync_start = MaxSector; | |
7750 | } else { | |
7751 | /* | |
7752 | * FIXME handle dirty degraded | |
7753 | */ | |
7754 | } | |
7755 | ||
9a1608e5 DW |
7756 | if (skip) |
7757 | continue; | |
4f5bc454 | 7758 | |
503975b9 | 7759 | info_d = xcalloc(1, sizeof(*info_d)); |
4f5bc454 DW |
7760 | info_d->next = this->devs; |
7761 | this->devs = info_d; | |
7762 | ||
4f5bc454 DW |
7763 | info_d->disk.number = d->index; |
7764 | info_d->disk.major = d->major; | |
7765 | info_d->disk.minor = d->minor; | |
7766 | info_d->disk.raid_disk = slot; | |
1e5c6983 | 7767 | info_d->recovery_start = recovery_start; |
86e3692b AK |
7768 | if (map2) { |
7769 | if (slot < map2->num_members) | |
7770 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
7771 | else |
7772 | this->array.spare_disks++; | |
86e3692b AK |
7773 | } else { |
7774 | if (slot < map->num_members) | |
7775 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
7776 | else |
7777 | this->array.spare_disks++; | |
86e3692b | 7778 | } |
1e5c6983 DW |
7779 | if (info_d->recovery_start == MaxSector) |
7780 | this->array.working_disks++; | |
4f5bc454 DW |
7781 | |
7782 | info_d->events = __le32_to_cpu(mpb->generation_num); | |
5551b113 | 7783 | info_d->data_offset = pba_of_lba0(map); |
06fb291a PB |
7784 | |
7785 | if (map->raid_level == 5) { | |
7786 | info_d->component_size = | |
7787 | num_data_stripes(map) * | |
7788 | map->blocks_per_strip; | |
2432ce9b AP |
7789 | info_d->ppl_sector = this->ppl_sector; |
7790 | info_d->ppl_size = this->ppl_size; | |
98e96bdb AP |
7791 | if (this->consistency_policy == CONSISTENCY_POLICY_PPL && |
7792 | recovery_start == 0) | |
7793 | this->resync_start = 0; | |
06fb291a PB |
7794 | } else { |
7795 | info_d->component_size = blocks_per_member(map); | |
7796 | } | |
b12796be | 7797 | |
5e46202e | 7798 | info_d->bb.supported = 1; |
b12796be TM |
7799 | get_volume_badblocks(super->bbm_log, ord_to_idx(ord), |
7800 | info_d->data_offset, | |
7801 | info_d->component_size, | |
7802 | &info_d->bb); | |
4f5bc454 | 7803 | } |
1e5c6983 | 7804 | /* now that the disk list is up-to-date fixup recovery_start */ |
c47b0ff6 | 7805 | update_recovery_start(super, dev, this); |
abef11a3 | 7806 | this->array.spare_disks += spare_disks; |
276d77db AK |
7807 | |
7808 | /* check for reshape */ | |
7809 | if (this->reshape_active == 1) | |
7810 | recover_backup_imsm(st, this); | |
9a1608e5 | 7811 | rest = this; |
4f5bc454 DW |
7812 | } |
7813 | ||
7814 | return rest; | |
cdddbdbc DW |
7815 | } |
7816 | ||
3b451610 AK |
7817 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, |
7818 | int failed, int look_in_map) | |
c2a1e7da | 7819 | { |
3b451610 AK |
7820 | struct imsm_map *map; |
7821 | ||
7822 | map = get_imsm_map(dev, look_in_map); | |
c2a1e7da DW |
7823 | |
7824 | if (!failed) | |
1011e834 | 7825 | return map->map_state == IMSM_T_STATE_UNINITIALIZED ? |
3393c6af | 7826 | IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL; |
c2a1e7da DW |
7827 | |
7828 | switch (get_imsm_raid_level(map)) { | |
7829 | case 0: | |
7830 | return IMSM_T_STATE_FAILED; | |
7831 | break; | |
7832 | case 1: | |
7833 | if (failed < map->num_members) | |
7834 | return IMSM_T_STATE_DEGRADED; | |
7835 | else | |
7836 | return IMSM_T_STATE_FAILED; | |
7837 | break; | |
7838 | case 10: | |
7839 | { | |
7840 | /** | |
c92a2527 DW |
7841 | * check to see if any mirrors have failed, otherwise we |
7842 | * are degraded. Even numbered slots are mirrored on | |
7843 | * slot+1 | |
c2a1e7da | 7844 | */ |
c2a1e7da | 7845 | int i; |
d9b420a5 N |
7846 | /* gcc -Os complains that this is unused */ |
7847 | int insync = insync; | |
c2a1e7da DW |
7848 | |
7849 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 7850 | __u32 ord = get_imsm_ord_tbl_ent(dev, i, MAP_X); |
c92a2527 DW |
7851 | int idx = ord_to_idx(ord); |
7852 | struct imsm_disk *disk; | |
c2a1e7da | 7853 | |
c92a2527 | 7854 | /* reset the potential in-sync count on even-numbered |
1011e834 | 7855 | * slots. num_copies is always 2 for imsm raid10 |
c92a2527 DW |
7856 | */ |
7857 | if ((i & 1) == 0) | |
7858 | insync = 2; | |
c2a1e7da | 7859 | |
c92a2527 | 7860 | disk = get_imsm_disk(super, idx); |
25ed7e59 | 7861 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) |
c92a2527 | 7862 | insync--; |
c2a1e7da | 7863 | |
c92a2527 DW |
7864 | /* no in-sync disks left in this mirror the |
7865 | * array has failed | |
7866 | */ | |
7867 | if (insync == 0) | |
7868 | return IMSM_T_STATE_FAILED; | |
c2a1e7da DW |
7869 | } |
7870 | ||
7871 | return IMSM_T_STATE_DEGRADED; | |
7872 | } | |
7873 | case 5: | |
7874 | if (failed < 2) | |
7875 | return IMSM_T_STATE_DEGRADED; | |
7876 | else | |
7877 | return IMSM_T_STATE_FAILED; | |
7878 | break; | |
7879 | default: | |
7880 | break; | |
7881 | } | |
7882 | ||
7883 | return map->map_state; | |
7884 | } | |
7885 | ||
3b451610 AK |
7886 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev, |
7887 | int look_in_map) | |
c2a1e7da DW |
7888 | { |
7889 | int i; | |
7890 | int failed = 0; | |
7891 | struct imsm_disk *disk; | |
d5985138 AK |
7892 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
7893 | struct imsm_map *prev = get_imsm_map(dev, MAP_1); | |
68fe4598 | 7894 | struct imsm_map *map_for_loop; |
0556e1a2 DW |
7895 | __u32 ord; |
7896 | int idx; | |
d5985138 | 7897 | int idx_1; |
c2a1e7da | 7898 | |
0556e1a2 DW |
7899 | /* at the beginning of migration we set IMSM_ORD_REBUILD on |
7900 | * disks that are being rebuilt. New failures are recorded to | |
7901 | * map[0]. So we look through all the disks we started with and | |
7902 | * see if any failures are still present, or if any new ones | |
7903 | * have arrived | |
0556e1a2 | 7904 | */ |
d5985138 AK |
7905 | map_for_loop = map; |
7906 | if (prev && (map->num_members < prev->num_members)) | |
7907 | map_for_loop = prev; | |
68fe4598 LD |
7908 | |
7909 | for (i = 0; i < map_for_loop->num_members; i++) { | |
d5985138 | 7910 | idx_1 = -255; |
238c0a71 AK |
7911 | /* when MAP_X is passed both maps failures are counted |
7912 | */ | |
d5985138 | 7913 | if (prev && |
089f9d79 JS |
7914 | (look_in_map == MAP_1 || look_in_map == MAP_X) && |
7915 | i < prev->num_members) { | |
d5985138 AK |
7916 | ord = __le32_to_cpu(prev->disk_ord_tbl[i]); |
7917 | idx_1 = ord_to_idx(ord); | |
c2a1e7da | 7918 | |
d5985138 AK |
7919 | disk = get_imsm_disk(super, idx_1); |
7920 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) | |
7921 | failed++; | |
7922 | } | |
089f9d79 JS |
7923 | if ((look_in_map == MAP_0 || look_in_map == MAP_X) && |
7924 | i < map->num_members) { | |
d5985138 AK |
7925 | ord = __le32_to_cpu(map->disk_ord_tbl[i]); |
7926 | idx = ord_to_idx(ord); | |
7927 | ||
7928 | if (idx != idx_1) { | |
7929 | disk = get_imsm_disk(super, idx); | |
7930 | if (!disk || is_failed(disk) || | |
7931 | ord & IMSM_ORD_REBUILD) | |
7932 | failed++; | |
7933 | } | |
7934 | } | |
c2a1e7da DW |
7935 | } |
7936 | ||
7937 | return failed; | |
845dea95 NB |
7938 | } |
7939 | ||
97b4d0e9 DW |
7940 | static int imsm_open_new(struct supertype *c, struct active_array *a, |
7941 | char *inst) | |
7942 | { | |
7943 | struct intel_super *super = c->sb; | |
7944 | struct imsm_super *mpb = super->anchor; | |
bbab0940 | 7945 | struct imsm_update_prealloc_bb_mem u; |
9587c373 | 7946 | |
97b4d0e9 | 7947 | if (atoi(inst) >= mpb->num_raid_devs) { |
1ade5cc1 | 7948 | pr_err("subarry index %d, out of range\n", atoi(inst)); |
97b4d0e9 DW |
7949 | return -ENODEV; |
7950 | } | |
7951 | ||
7952 | dprintf("imsm: open_new %s\n", inst); | |
7953 | a->info.container_member = atoi(inst); | |
bbab0940 TM |
7954 | |
7955 | u.type = update_prealloc_badblocks_mem; | |
7956 | imsm_update_metadata_locally(c, &u, sizeof(u)); | |
7957 | ||
97b4d0e9 DW |
7958 | return 0; |
7959 | } | |
7960 | ||
0c046afd DW |
7961 | static int is_resyncing(struct imsm_dev *dev) |
7962 | { | |
7963 | struct imsm_map *migr_map; | |
7964 | ||
7965 | if (!dev->vol.migr_state) | |
7966 | return 0; | |
7967 | ||
1484e727 DW |
7968 | if (migr_type(dev) == MIGR_INIT || |
7969 | migr_type(dev) == MIGR_REPAIR) | |
0c046afd DW |
7970 | return 1; |
7971 | ||
4c9bc37b AK |
7972 | if (migr_type(dev) == MIGR_GEN_MIGR) |
7973 | return 0; | |
7974 | ||
238c0a71 | 7975 | migr_map = get_imsm_map(dev, MAP_1); |
0c046afd | 7976 | |
089f9d79 JS |
7977 | if (migr_map->map_state == IMSM_T_STATE_NORMAL && |
7978 | dev->vol.migr_type != MIGR_GEN_MIGR) | |
0c046afd DW |
7979 | return 1; |
7980 | else | |
7981 | return 0; | |
7982 | } | |
7983 | ||
0556e1a2 | 7984 | /* return true if we recorded new information */ |
4c9e8c1e TM |
7985 | static int mark_failure(struct intel_super *super, |
7986 | struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
47ee5a45 | 7987 | { |
0556e1a2 DW |
7988 | __u32 ord; |
7989 | int slot; | |
7990 | struct imsm_map *map; | |
86c54047 DW |
7991 | char buf[MAX_RAID_SERIAL_LEN+3]; |
7992 | unsigned int len, shift = 0; | |
0556e1a2 DW |
7993 | |
7994 | /* new failures are always set in map[0] */ | |
238c0a71 | 7995 | map = get_imsm_map(dev, MAP_0); |
0556e1a2 DW |
7996 | |
7997 | slot = get_imsm_disk_slot(map, idx); | |
7998 | if (slot < 0) | |
7999 | return 0; | |
8000 | ||
8001 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
25ed7e59 | 8002 | if (is_failed(disk) && (ord & IMSM_ORD_REBUILD)) |
0556e1a2 DW |
8003 | return 0; |
8004 | ||
7d0c5e24 LD |
8005 | memcpy(buf, disk->serial, MAX_RAID_SERIAL_LEN); |
8006 | buf[MAX_RAID_SERIAL_LEN] = '\000'; | |
8007 | strcat(buf, ":0"); | |
86c54047 DW |
8008 | if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN) |
8009 | shift = len - MAX_RAID_SERIAL_LEN + 1; | |
8010 | strncpy((char *)disk->serial, &buf[shift], MAX_RAID_SERIAL_LEN); | |
8011 | ||
f2f27e63 | 8012 | disk->status |= FAILED_DISK; |
0556e1a2 | 8013 | set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD); |
17788994 AK |
8014 | /* mark failures in second map if second map exists and this disk |
8015 | * in this slot. | |
8016 | * This is valid for migration, initialization and rebuild | |
8017 | */ | |
8018 | if (dev->vol.migr_state) { | |
238c0a71 | 8019 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); |
0a108d63 AK |
8020 | int slot2 = get_imsm_disk_slot(map2, idx); |
8021 | ||
089f9d79 | 8022 | if (slot2 < map2->num_members && slot2 >= 0) |
0a108d63 | 8023 | set_imsm_ord_tbl_ent(map2, slot2, |
1ace8403 AK |
8024 | idx | IMSM_ORD_REBUILD); |
8025 | } | |
f21e18ca | 8026 | if (map->failed_disk_num == 0xff) |
0556e1a2 | 8027 | map->failed_disk_num = slot; |
4c9e8c1e TM |
8028 | |
8029 | clear_disk_badblocks(super->bbm_log, ord_to_idx(ord)); | |
8030 | ||
0556e1a2 DW |
8031 | return 1; |
8032 | } | |
8033 | ||
4c9e8c1e TM |
8034 | static void mark_missing(struct intel_super *super, |
8035 | struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
0556e1a2 | 8036 | { |
4c9e8c1e | 8037 | mark_failure(super, dev, disk, idx); |
0556e1a2 DW |
8038 | |
8039 | if (disk->scsi_id == __cpu_to_le32(~(__u32)0)) | |
8040 | return; | |
8041 | ||
47ee5a45 DW |
8042 | disk->scsi_id = __cpu_to_le32(~(__u32)0); |
8043 | memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1); | |
8044 | } | |
8045 | ||
33414a01 DW |
8046 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev) |
8047 | { | |
33414a01 | 8048 | struct dl *dl; |
33414a01 DW |
8049 | |
8050 | if (!super->missing) | |
8051 | return; | |
33414a01 | 8052 | |
79b68f1b PC |
8053 | /* When orom adds replacement for missing disk it does |
8054 | * not remove entry of missing disk, but just updates map with | |
8055 | * new added disk. So it is not enough just to test if there is | |
8056 | * any missing disk, we have to look if there are any failed disks | |
8057 | * in map to stop migration */ | |
8058 | ||
33414a01 | 8059 | dprintf("imsm: mark missing\n"); |
3d59f0c0 AK |
8060 | /* end process for initialization and rebuild only |
8061 | */ | |
8062 | if (is_gen_migration(dev) == 0) { | |
fb12a745 | 8063 | int failed = imsm_count_failed(super, dev, MAP_0); |
3d59f0c0 | 8064 | |
fb12a745 TM |
8065 | if (failed) { |
8066 | __u8 map_state; | |
8067 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
8068 | struct imsm_map *map1; | |
8069 | int i, ord, ord_map1; | |
8070 | int rebuilt = 1; | |
3d59f0c0 | 8071 | |
fb12a745 TM |
8072 | for (i = 0; i < map->num_members; i++) { |
8073 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_0); | |
8074 | if (!(ord & IMSM_ORD_REBUILD)) | |
8075 | continue; | |
8076 | ||
8077 | map1 = get_imsm_map(dev, MAP_1); | |
8078 | if (!map1) | |
8079 | continue; | |
8080 | ||
8081 | ord_map1 = __le32_to_cpu(map1->disk_ord_tbl[i]); | |
8082 | if (ord_map1 & IMSM_ORD_REBUILD) | |
8083 | rebuilt = 0; | |
8084 | } | |
8085 | ||
8086 | if (rebuilt) { | |
8087 | map_state = imsm_check_degraded(super, dev, | |
8088 | failed, MAP_0); | |
8089 | end_migration(dev, super, map_state); | |
8090 | } | |
8091 | } | |
3d59f0c0 | 8092 | } |
33414a01 | 8093 | for (dl = super->missing; dl; dl = dl->next) |
4c9e8c1e | 8094 | mark_missing(super, dev, &dl->disk, dl->index); |
33414a01 DW |
8095 | super->updates_pending++; |
8096 | } | |
8097 | ||
f3871fdc AK |
8098 | static unsigned long long imsm_set_array_size(struct imsm_dev *dev, |
8099 | long long new_size) | |
70bdf0dc | 8100 | { |
238c0a71 | 8101 | int used_disks = imsm_num_data_members(dev, MAP_0); |
70bdf0dc AK |
8102 | unsigned long long array_blocks; |
8103 | struct imsm_map *map; | |
8104 | ||
8105 | if (used_disks == 0) { | |
8106 | /* when problems occures | |
8107 | * return current array_blocks value | |
8108 | */ | |
8109 | array_blocks = __le32_to_cpu(dev->size_high); | |
8110 | array_blocks = array_blocks << 32; | |
8111 | array_blocks += __le32_to_cpu(dev->size_low); | |
8112 | ||
8113 | return array_blocks; | |
8114 | } | |
8115 | ||
8116 | /* set array size in metadata | |
8117 | */ | |
f3871fdc AK |
8118 | if (new_size <= 0) { |
8119 | /* OLCE size change is caused by added disks | |
8120 | */ | |
8121 | map = get_imsm_map(dev, MAP_0); | |
8122 | array_blocks = blocks_per_member(map) * used_disks; | |
8123 | } else { | |
8124 | /* Online Volume Size Change | |
8125 | * Using available free space | |
8126 | */ | |
8127 | array_blocks = new_size; | |
8128 | } | |
70bdf0dc | 8129 | |
b53bfba6 | 8130 | array_blocks = round_size_to_mb(array_blocks, used_disks); |
70bdf0dc AK |
8131 | dev->size_low = __cpu_to_le32((__u32)array_blocks); |
8132 | dev->size_high = __cpu_to_le32((__u32)(array_blocks >> 32)); | |
8133 | ||
8134 | return array_blocks; | |
8135 | } | |
8136 | ||
28bce06f AK |
8137 | static void imsm_set_disk(struct active_array *a, int n, int state); |
8138 | ||
0e2d1a4e AK |
8139 | static void imsm_progress_container_reshape(struct intel_super *super) |
8140 | { | |
8141 | /* if no device has a migr_state, but some device has a | |
8142 | * different number of members than the previous device, start | |
8143 | * changing the number of devices in this device to match | |
8144 | * previous. | |
8145 | */ | |
8146 | struct imsm_super *mpb = super->anchor; | |
8147 | int prev_disks = -1; | |
8148 | int i; | |
1dfaa380 | 8149 | int copy_map_size; |
0e2d1a4e AK |
8150 | |
8151 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
8152 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 8153 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
0e2d1a4e AK |
8154 | struct imsm_map *map2; |
8155 | int prev_num_members; | |
0e2d1a4e AK |
8156 | |
8157 | if (dev->vol.migr_state) | |
8158 | return; | |
8159 | ||
8160 | if (prev_disks == -1) | |
8161 | prev_disks = map->num_members; | |
8162 | if (prev_disks == map->num_members) | |
8163 | continue; | |
8164 | ||
8165 | /* OK, this array needs to enter reshape mode. | |
8166 | * i.e it needs a migr_state | |
8167 | */ | |
8168 | ||
1dfaa380 | 8169 | copy_map_size = sizeof_imsm_map(map); |
0e2d1a4e AK |
8170 | prev_num_members = map->num_members; |
8171 | map->num_members = prev_disks; | |
8172 | dev->vol.migr_state = 1; | |
8173 | dev->vol.curr_migr_unit = 0; | |
ea672ee1 | 8174 | set_migr_type(dev, MIGR_GEN_MIGR); |
0e2d1a4e AK |
8175 | for (i = prev_num_members; |
8176 | i < map->num_members; i++) | |
8177 | set_imsm_ord_tbl_ent(map, i, i); | |
238c0a71 | 8178 | map2 = get_imsm_map(dev, MAP_1); |
0e2d1a4e | 8179 | /* Copy the current map */ |
1dfaa380 | 8180 | memcpy(map2, map, copy_map_size); |
0e2d1a4e AK |
8181 | map2->num_members = prev_num_members; |
8182 | ||
f3871fdc | 8183 | imsm_set_array_size(dev, -1); |
51d83f5d | 8184 | super->clean_migration_record_by_mdmon = 1; |
0e2d1a4e AK |
8185 | super->updates_pending++; |
8186 | } | |
8187 | } | |
8188 | ||
aad6f216 | 8189 | /* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild |
0c046afd DW |
8190 | * states are handled in imsm_set_disk() with one exception, when a |
8191 | * resync is stopped due to a new failure this routine will set the | |
8192 | * 'degraded' state for the array. | |
8193 | */ | |
01f157d7 | 8194 | static int imsm_set_array_state(struct active_array *a, int consistent) |
a862209d DW |
8195 | { |
8196 | int inst = a->info.container_member; | |
8197 | struct intel_super *super = a->container->sb; | |
949c47a0 | 8198 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 8199 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
3b451610 AK |
8200 | int failed = imsm_count_failed(super, dev, MAP_0); |
8201 | __u8 map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
1e5c6983 | 8202 | __u32 blocks_per_unit; |
a862209d | 8203 | |
1af97990 AK |
8204 | if (dev->vol.migr_state && |
8205 | dev->vol.migr_type == MIGR_GEN_MIGR) { | |
8206 | /* array state change is blocked due to reshape action | |
aad6f216 N |
8207 | * We might need to |
8208 | * - abort the reshape (if last_checkpoint is 0 and action!= reshape) | |
8209 | * - finish the reshape (if last_checkpoint is big and action != reshape) | |
8210 | * - update curr_migr_unit | |
1af97990 | 8211 | */ |
aad6f216 N |
8212 | if (a->curr_action == reshape) { |
8213 | /* still reshaping, maybe update curr_migr_unit */ | |
633b5610 | 8214 | goto mark_checkpoint; |
aad6f216 N |
8215 | } else { |
8216 | if (a->last_checkpoint == 0 && a->prev_action == reshape) { | |
8217 | /* for some reason we aborted the reshape. | |
b66e591b AK |
8218 | * |
8219 | * disable automatic metadata rollback | |
8220 | * user action is required to recover process | |
aad6f216 | 8221 | */ |
b66e591b | 8222 | if (0) { |
238c0a71 AK |
8223 | struct imsm_map *map2 = |
8224 | get_imsm_map(dev, MAP_1); | |
8225 | dev->vol.migr_state = 0; | |
8226 | set_migr_type(dev, 0); | |
8227 | dev->vol.curr_migr_unit = 0; | |
8228 | memcpy(map, map2, | |
8229 | sizeof_imsm_map(map2)); | |
8230 | super->updates_pending++; | |
b66e591b | 8231 | } |
aad6f216 N |
8232 | } |
8233 | if (a->last_checkpoint >= a->info.component_size) { | |
8234 | unsigned long long array_blocks; | |
8235 | int used_disks; | |
e154ced3 | 8236 | struct mdinfo *mdi; |
aad6f216 | 8237 | |
238c0a71 | 8238 | used_disks = imsm_num_data_members(dev, MAP_0); |
d55adef9 AK |
8239 | if (used_disks > 0) { |
8240 | array_blocks = | |
5551b113 | 8241 | blocks_per_member(map) * |
d55adef9 | 8242 | used_disks; |
b53bfba6 TM |
8243 | array_blocks = |
8244 | round_size_to_mb(array_blocks, | |
8245 | used_disks); | |
d55adef9 AK |
8246 | a->info.custom_array_size = array_blocks; |
8247 | /* encourage manager to update array | |
8248 | * size | |
8249 | */ | |
e154ced3 | 8250 | |
d55adef9 | 8251 | a->check_reshape = 1; |
633b5610 | 8252 | } |
e154ced3 AK |
8253 | /* finalize online capacity expansion/reshape */ |
8254 | for (mdi = a->info.devs; mdi; mdi = mdi->next) | |
8255 | imsm_set_disk(a, | |
8256 | mdi->disk.raid_disk, | |
8257 | mdi->curr_state); | |
8258 | ||
0e2d1a4e | 8259 | imsm_progress_container_reshape(super); |
e154ced3 | 8260 | } |
aad6f216 | 8261 | } |
1af97990 AK |
8262 | } |
8263 | ||
47ee5a45 | 8264 | /* before we activate this array handle any missing disks */ |
33414a01 DW |
8265 | if (consistent == 2) |
8266 | handle_missing(super, dev); | |
1e5c6983 | 8267 | |
0c046afd | 8268 | if (consistent == 2 && |
b7941fd6 | 8269 | (!is_resync_complete(&a->info) || |
0c046afd DW |
8270 | map_state != IMSM_T_STATE_NORMAL || |
8271 | dev->vol.migr_state)) | |
01f157d7 | 8272 | consistent = 0; |
272906ef | 8273 | |
b7941fd6 | 8274 | if (is_resync_complete(&a->info)) { |
0c046afd | 8275 | /* complete intialization / resync, |
0556e1a2 DW |
8276 | * recovery and interrupted recovery is completed in |
8277 | * ->set_disk | |
0c046afd DW |
8278 | */ |
8279 | if (is_resyncing(dev)) { | |
8280 | dprintf("imsm: mark resync done\n"); | |
809da78e | 8281 | end_migration(dev, super, map_state); |
115c3803 | 8282 | super->updates_pending++; |
484240d8 | 8283 | a->last_checkpoint = 0; |
115c3803 | 8284 | } |
b9172665 AK |
8285 | } else if ((!is_resyncing(dev) && !failed) && |
8286 | (imsm_reshape_blocks_arrays_changes(super) == 0)) { | |
0c046afd | 8287 | /* mark the start of the init process if nothing is failed */ |
b7941fd6 | 8288 | dprintf("imsm: mark resync start\n"); |
1484e727 | 8289 | if (map->map_state == IMSM_T_STATE_UNINITIALIZED) |
8e59f3d8 | 8290 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT); |
1484e727 | 8291 | else |
8e59f3d8 | 8292 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR); |
3393c6af | 8293 | super->updates_pending++; |
115c3803 | 8294 | } |
a862209d | 8295 | |
633b5610 | 8296 | mark_checkpoint: |
5b83bacf AK |
8297 | /* skip checkpointing for general migration, |
8298 | * it is controlled in mdadm | |
8299 | */ | |
8300 | if (is_gen_migration(dev)) | |
8301 | goto skip_mark_checkpoint; | |
8302 | ||
1e5c6983 | 8303 | /* check if we can update curr_migr_unit from resync_start, recovery_start */ |
c47b0ff6 | 8304 | blocks_per_unit = blocks_per_migr_unit(super, dev); |
4f0a7acc | 8305 | if (blocks_per_unit) { |
1e5c6983 DW |
8306 | __u32 units32; |
8307 | __u64 units; | |
8308 | ||
4f0a7acc | 8309 | units = a->last_checkpoint / blocks_per_unit; |
1e5c6983 DW |
8310 | units32 = units; |
8311 | ||
8312 | /* check that we did not overflow 32-bits, and that | |
8313 | * curr_migr_unit needs updating | |
8314 | */ | |
8315 | if (units32 == units && | |
bfd80a56 | 8316 | units32 != 0 && |
1e5c6983 DW |
8317 | __le32_to_cpu(dev->vol.curr_migr_unit) != units32) { |
8318 | dprintf("imsm: mark checkpoint (%u)\n", units32); | |
8319 | dev->vol.curr_migr_unit = __cpu_to_le32(units32); | |
8320 | super->updates_pending++; | |
8321 | } | |
8322 | } | |
f8f603f1 | 8323 | |
5b83bacf | 8324 | skip_mark_checkpoint: |
3393c6af | 8325 | /* mark dirty / clean */ |
2432ce9b AP |
8326 | if (((dev->vol.dirty & RAIDVOL_DIRTY) && consistent) || |
8327 | (!(dev->vol.dirty & RAIDVOL_DIRTY) && !consistent)) { | |
b7941fd6 | 8328 | dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty"); |
2432ce9b AP |
8329 | if (consistent) { |
8330 | dev->vol.dirty = RAIDVOL_CLEAN; | |
8331 | } else { | |
8332 | dev->vol.dirty = RAIDVOL_DIRTY; | |
c2462068 PB |
8333 | if (dev->rwh_policy == RWH_DISTRIBUTED || |
8334 | dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED) | |
2432ce9b AP |
8335 | dev->vol.dirty |= RAIDVOL_DSRECORD_VALID; |
8336 | } | |
a862209d DW |
8337 | super->updates_pending++; |
8338 | } | |
28bce06f | 8339 | |
01f157d7 | 8340 | return consistent; |
a862209d DW |
8341 | } |
8342 | ||
6f50473f TM |
8343 | static int imsm_disk_slot_to_ord(struct active_array *a, int slot) |
8344 | { | |
8345 | int inst = a->info.container_member; | |
8346 | struct intel_super *super = a->container->sb; | |
8347 | struct imsm_dev *dev = get_imsm_dev(super, inst); | |
8348 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
8349 | ||
8350 | if (slot > map->num_members) { | |
8351 | pr_err("imsm: imsm_disk_slot_to_ord %d out of range 0..%d\n", | |
8352 | slot, map->num_members - 1); | |
8353 | return -1; | |
8354 | } | |
8355 | ||
8356 | if (slot < 0) | |
8357 | return -1; | |
8358 | ||
8359 | return get_imsm_ord_tbl_ent(dev, slot, MAP_0); | |
8360 | } | |
8361 | ||
8d45d196 | 8362 | static void imsm_set_disk(struct active_array *a, int n, int state) |
845dea95 | 8363 | { |
8d45d196 DW |
8364 | int inst = a->info.container_member; |
8365 | struct intel_super *super = a->container->sb; | |
949c47a0 | 8366 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 8367 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
8d45d196 | 8368 | struct imsm_disk *disk; |
7ce05701 LD |
8369 | struct mdinfo *mdi; |
8370 | int recovery_not_finished = 0; | |
0c046afd | 8371 | int failed; |
6f50473f | 8372 | int ord; |
0c046afd | 8373 | __u8 map_state; |
fb12a745 TM |
8374 | int rebuild_done = 0; |
8375 | int i; | |
8d45d196 | 8376 | |
fb12a745 | 8377 | ord = get_imsm_ord_tbl_ent(dev, n, MAP_X); |
6f50473f | 8378 | if (ord < 0) |
8d45d196 DW |
8379 | return; |
8380 | ||
4e6e574a | 8381 | dprintf("imsm: set_disk %d:%x\n", n, state); |
b10b37b8 | 8382 | disk = get_imsm_disk(super, ord_to_idx(ord)); |
8d45d196 | 8383 | |
5802a811 | 8384 | /* check for new failures */ |
0556e1a2 | 8385 | if (state & DS_FAULTY) { |
4c9e8c1e | 8386 | if (mark_failure(super, dev, disk, ord_to_idx(ord))) |
0556e1a2 | 8387 | super->updates_pending++; |
8d45d196 | 8388 | } |
47ee5a45 | 8389 | |
19859edc | 8390 | /* check if in_sync */ |
0556e1a2 | 8391 | if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) { |
238c0a71 | 8392 | struct imsm_map *migr_map = get_imsm_map(dev, MAP_1); |
b10b37b8 DW |
8393 | |
8394 | set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord)); | |
fb12a745 | 8395 | rebuild_done = 1; |
19859edc DW |
8396 | super->updates_pending++; |
8397 | } | |
8d45d196 | 8398 | |
3b451610 AK |
8399 | failed = imsm_count_failed(super, dev, MAP_0); |
8400 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
5802a811 | 8401 | |
0c046afd | 8402 | /* check if recovery complete, newly degraded, or failed */ |
94002678 AK |
8403 | dprintf("imsm: Detected transition to state "); |
8404 | switch (map_state) { | |
8405 | case IMSM_T_STATE_NORMAL: /* transition to normal state */ | |
8406 | dprintf("normal: "); | |
8407 | if (is_rebuilding(dev)) { | |
1ade5cc1 | 8408 | dprintf_cont("while rebuilding"); |
7ce05701 LD |
8409 | /* check if recovery is really finished */ |
8410 | for (mdi = a->info.devs; mdi ; mdi = mdi->next) | |
8411 | if (mdi->recovery_start != MaxSector) { | |
8412 | recovery_not_finished = 1; | |
8413 | break; | |
8414 | } | |
8415 | if (recovery_not_finished) { | |
1ade5cc1 N |
8416 | dprintf_cont("\n"); |
8417 | dprintf("Rebuild has not finished yet, state not changed"); | |
7ce05701 LD |
8418 | if (a->last_checkpoint < mdi->recovery_start) { |
8419 | a->last_checkpoint = mdi->recovery_start; | |
8420 | super->updates_pending++; | |
8421 | } | |
8422 | break; | |
8423 | } | |
94002678 | 8424 | end_migration(dev, super, map_state); |
238c0a71 | 8425 | map = get_imsm_map(dev, MAP_0); |
94002678 AK |
8426 | map->failed_disk_num = ~0; |
8427 | super->updates_pending++; | |
8428 | a->last_checkpoint = 0; | |
8429 | break; | |
8430 | } | |
8431 | if (is_gen_migration(dev)) { | |
1ade5cc1 | 8432 | dprintf_cont("while general migration"); |
bf2f0071 | 8433 | if (a->last_checkpoint >= a->info.component_size) |
809da78e | 8434 | end_migration(dev, super, map_state); |
94002678 AK |
8435 | else |
8436 | map->map_state = map_state; | |
238c0a71 | 8437 | map = get_imsm_map(dev, MAP_0); |
28bce06f | 8438 | map->failed_disk_num = ~0; |
94002678 | 8439 | super->updates_pending++; |
bf2f0071 | 8440 | break; |
94002678 AK |
8441 | } |
8442 | break; | |
8443 | case IMSM_T_STATE_DEGRADED: /* transition to degraded state */ | |
1ade5cc1 | 8444 | dprintf_cont("degraded: "); |
089f9d79 | 8445 | if (map->map_state != map_state && !dev->vol.migr_state) { |
1ade5cc1 | 8446 | dprintf_cont("mark degraded"); |
94002678 AK |
8447 | map->map_state = map_state; |
8448 | super->updates_pending++; | |
8449 | a->last_checkpoint = 0; | |
8450 | break; | |
8451 | } | |
8452 | if (is_rebuilding(dev)) { | |
1ade5cc1 | 8453 | dprintf_cont("while rebuilding."); |
94002678 | 8454 | if (map->map_state != map_state) { |
1ade5cc1 | 8455 | dprintf_cont(" Map state change"); |
94002678 AK |
8456 | end_migration(dev, super, map_state); |
8457 | super->updates_pending++; | |
fb12a745 TM |
8458 | } else if (!rebuild_done) { |
8459 | break; | |
8460 | } | |
8461 | ||
8462 | /* check if recovery is really finished */ | |
8463 | for (mdi = a->info.devs; mdi ; mdi = mdi->next) | |
8464 | if (mdi->recovery_start != MaxSector) { | |
8465 | recovery_not_finished = 1; | |
8466 | break; | |
8467 | } | |
8468 | if (recovery_not_finished) { | |
8469 | dprintf_cont("\n"); | |
8470 | dprintf("Rebuild has not finished yet, state not changed"); | |
8471 | if (a->last_checkpoint < mdi->recovery_start) { | |
8472 | a->last_checkpoint = | |
8473 | mdi->recovery_start; | |
8474 | super->updates_pending++; | |
8475 | } | |
8476 | break; | |
94002678 | 8477 | } |
fb12a745 TM |
8478 | |
8479 | dprintf_cont(" Rebuild done, still degraded"); | |
8480 | dev->vol.migr_state = 0; | |
8481 | set_migr_type(dev, 0); | |
8482 | dev->vol.curr_migr_unit = 0; | |
8483 | ||
8484 | for (i = 0; i < map->num_members; i++) { | |
8485 | int idx = get_imsm_ord_tbl_ent(dev, i, MAP_0); | |
8486 | ||
8487 | if (idx & IMSM_ORD_REBUILD) | |
8488 | map->failed_disk_num = i; | |
8489 | } | |
8490 | super->updates_pending++; | |
94002678 AK |
8491 | break; |
8492 | } | |
8493 | if (is_gen_migration(dev)) { | |
1ade5cc1 | 8494 | dprintf_cont("while general migration"); |
bf2f0071 | 8495 | if (a->last_checkpoint >= a->info.component_size) |
809da78e | 8496 | end_migration(dev, super, map_state); |
94002678 AK |
8497 | else { |
8498 | map->map_state = map_state; | |
3b451610 | 8499 | manage_second_map(super, dev); |
94002678 AK |
8500 | } |
8501 | super->updates_pending++; | |
bf2f0071 | 8502 | break; |
28bce06f | 8503 | } |
6ce1fbf1 | 8504 | if (is_initializing(dev)) { |
1ade5cc1 | 8505 | dprintf_cont("while initialization."); |
6ce1fbf1 AK |
8506 | map->map_state = map_state; |
8507 | super->updates_pending++; | |
8508 | break; | |
8509 | } | |
94002678 AK |
8510 | break; |
8511 | case IMSM_T_STATE_FAILED: /* transition to failed state */ | |
1ade5cc1 | 8512 | dprintf_cont("failed: "); |
94002678 | 8513 | if (is_gen_migration(dev)) { |
1ade5cc1 | 8514 | dprintf_cont("while general migration"); |
94002678 AK |
8515 | map->map_state = map_state; |
8516 | super->updates_pending++; | |
8517 | break; | |
8518 | } | |
8519 | if (map->map_state != map_state) { | |
1ade5cc1 | 8520 | dprintf_cont("mark failed"); |
94002678 AK |
8521 | end_migration(dev, super, map_state); |
8522 | super->updates_pending++; | |
8523 | a->last_checkpoint = 0; | |
8524 | break; | |
8525 | } | |
8526 | break; | |
8527 | default: | |
1ade5cc1 | 8528 | dprintf_cont("state %i\n", map_state); |
5802a811 | 8529 | } |
1ade5cc1 | 8530 | dprintf_cont("\n"); |
845dea95 NB |
8531 | } |
8532 | ||
f796af5d | 8533 | static int store_imsm_mpb(int fd, struct imsm_super *mpb) |
c2a1e7da | 8534 | { |
f796af5d | 8535 | void *buf = mpb; |
c2a1e7da DW |
8536 | __u32 mpb_size = __le32_to_cpu(mpb->mpb_size); |
8537 | unsigned long long dsize; | |
8538 | unsigned long long sectors; | |
f36a9ecd | 8539 | unsigned int sector_size; |
c2a1e7da | 8540 | |
f36a9ecd | 8541 | get_dev_sector_size(fd, NULL, §or_size); |
c2a1e7da DW |
8542 | get_dev_size(fd, NULL, &dsize); |
8543 | ||
f36a9ecd | 8544 | if (mpb_size > sector_size) { |
272f648f | 8545 | /* -1 to account for anchor */ |
f36a9ecd | 8546 | sectors = mpb_sectors(mpb, sector_size) - 1; |
c2a1e7da | 8547 | |
272f648f | 8548 | /* write the extended mpb to the sectors preceeding the anchor */ |
f36a9ecd PB |
8549 | if (lseek64(fd, dsize - (sector_size * (2 + sectors)), |
8550 | SEEK_SET) < 0) | |
272f648f | 8551 | return 1; |
c2a1e7da | 8552 | |
f36a9ecd PB |
8553 | if ((unsigned long long)write(fd, buf + sector_size, |
8554 | sector_size * sectors) != sector_size * sectors) | |
272f648f DW |
8555 | return 1; |
8556 | } | |
c2a1e7da | 8557 | |
272f648f | 8558 | /* first block is stored on second to last sector of the disk */ |
f36a9ecd | 8559 | if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0) |
c2a1e7da DW |
8560 | return 1; |
8561 | ||
466070ad | 8562 | if ((unsigned int)write(fd, buf, sector_size) != sector_size) |
c2a1e7da DW |
8563 | return 1; |
8564 | ||
c2a1e7da DW |
8565 | return 0; |
8566 | } | |
8567 | ||
2e735d19 | 8568 | static void imsm_sync_metadata(struct supertype *container) |
845dea95 | 8569 | { |
2e735d19 | 8570 | struct intel_super *super = container->sb; |
c2a1e7da | 8571 | |
1a64be56 | 8572 | dprintf("sync metadata: %d\n", super->updates_pending); |
c2a1e7da DW |
8573 | if (!super->updates_pending) |
8574 | return; | |
8575 | ||
36988a3d | 8576 | write_super_imsm(container, 0); |
c2a1e7da DW |
8577 | |
8578 | super->updates_pending = 0; | |
845dea95 NB |
8579 | } |
8580 | ||
272906ef DW |
8581 | static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a) |
8582 | { | |
8583 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 8584 | int i = get_imsm_disk_idx(dev, idx, MAP_X); |
272906ef DW |
8585 | struct dl *dl; |
8586 | ||
8587 | for (dl = super->disks; dl; dl = dl->next) | |
8588 | if (dl->index == i) | |
8589 | break; | |
8590 | ||
25ed7e59 | 8591 | if (dl && is_failed(&dl->disk)) |
272906ef DW |
8592 | dl = NULL; |
8593 | ||
8594 | if (dl) | |
1ade5cc1 | 8595 | dprintf("found %x:%x\n", dl->major, dl->minor); |
272906ef DW |
8596 | |
8597 | return dl; | |
8598 | } | |
8599 | ||
a20d2ba5 | 8600 | static struct dl *imsm_add_spare(struct intel_super *super, int slot, |
8ba77d32 AK |
8601 | struct active_array *a, int activate_new, |
8602 | struct mdinfo *additional_test_list) | |
272906ef DW |
8603 | { |
8604 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 8605 | int idx = get_imsm_disk_idx(dev, slot, MAP_X); |
a20d2ba5 DW |
8606 | struct imsm_super *mpb = super->anchor; |
8607 | struct imsm_map *map; | |
272906ef DW |
8608 | unsigned long long pos; |
8609 | struct mdinfo *d; | |
8610 | struct extent *ex; | |
a20d2ba5 | 8611 | int i, j; |
272906ef | 8612 | int found; |
569cc43f DW |
8613 | __u32 array_start = 0; |
8614 | __u32 array_end = 0; | |
272906ef | 8615 | struct dl *dl; |
6c932028 | 8616 | struct mdinfo *test_list; |
272906ef DW |
8617 | |
8618 | for (dl = super->disks; dl; dl = dl->next) { | |
8619 | /* If in this array, skip */ | |
8620 | for (d = a->info.devs ; d ; d = d->next) | |
e553d2a4 DW |
8621 | if (d->state_fd >= 0 && |
8622 | d->disk.major == dl->major && | |
272906ef | 8623 | d->disk.minor == dl->minor) { |
8ba77d32 AK |
8624 | dprintf("%x:%x already in array\n", |
8625 | dl->major, dl->minor); | |
272906ef DW |
8626 | break; |
8627 | } | |
8628 | if (d) | |
8629 | continue; | |
6c932028 AK |
8630 | test_list = additional_test_list; |
8631 | while (test_list) { | |
8632 | if (test_list->disk.major == dl->major && | |
8633 | test_list->disk.minor == dl->minor) { | |
8ba77d32 AK |
8634 | dprintf("%x:%x already in additional test list\n", |
8635 | dl->major, dl->minor); | |
8636 | break; | |
8637 | } | |
6c932028 | 8638 | test_list = test_list->next; |
8ba77d32 | 8639 | } |
6c932028 | 8640 | if (test_list) |
8ba77d32 | 8641 | continue; |
272906ef | 8642 | |
e553d2a4 | 8643 | /* skip in use or failed drives */ |
25ed7e59 | 8644 | if (is_failed(&dl->disk) || idx == dl->index || |
df474657 DW |
8645 | dl->index == -2) { |
8646 | dprintf("%x:%x status (failed: %d index: %d)\n", | |
25ed7e59 | 8647 | dl->major, dl->minor, is_failed(&dl->disk), idx); |
9a1608e5 DW |
8648 | continue; |
8649 | } | |
8650 | ||
a20d2ba5 DW |
8651 | /* skip pure spares when we are looking for partially |
8652 | * assimilated drives | |
8653 | */ | |
8654 | if (dl->index == -1 && !activate_new) | |
8655 | continue; | |
8656 | ||
f2cc4f7d AO |
8657 | if (!drive_validate_sector_size(super, dl)) |
8658 | continue; | |
8659 | ||
272906ef | 8660 | /* Does this unused device have the requisite free space? |
a20d2ba5 | 8661 | * It needs to be able to cover all member volumes |
272906ef DW |
8662 | */ |
8663 | ex = get_extents(super, dl); | |
8664 | if (!ex) { | |
8665 | dprintf("cannot get extents\n"); | |
8666 | continue; | |
8667 | } | |
a20d2ba5 DW |
8668 | for (i = 0; i < mpb->num_raid_devs; i++) { |
8669 | dev = get_imsm_dev(super, i); | |
238c0a71 | 8670 | map = get_imsm_map(dev, MAP_0); |
272906ef | 8671 | |
a20d2ba5 DW |
8672 | /* check if this disk is already a member of |
8673 | * this array | |
272906ef | 8674 | */ |
620b1713 | 8675 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
a20d2ba5 DW |
8676 | continue; |
8677 | ||
8678 | found = 0; | |
8679 | j = 0; | |
8680 | pos = 0; | |
5551b113 | 8681 | array_start = pba_of_lba0(map); |
329c8278 | 8682 | array_end = array_start + |
5551b113 | 8683 | blocks_per_member(map) - 1; |
a20d2ba5 DW |
8684 | |
8685 | do { | |
8686 | /* check that we can start at pba_of_lba0 with | |
8687 | * blocks_per_member of space | |
8688 | */ | |
329c8278 | 8689 | if (array_start >= pos && array_end < ex[j].start) { |
a20d2ba5 DW |
8690 | found = 1; |
8691 | break; | |
8692 | } | |
8693 | pos = ex[j].start + ex[j].size; | |
8694 | j++; | |
8695 | } while (ex[j-1].size); | |
8696 | ||
8697 | if (!found) | |
272906ef | 8698 | break; |
a20d2ba5 | 8699 | } |
272906ef DW |
8700 | |
8701 | free(ex); | |
a20d2ba5 | 8702 | if (i < mpb->num_raid_devs) { |
329c8278 DW |
8703 | dprintf("%x:%x does not have %u to %u available\n", |
8704 | dl->major, dl->minor, array_start, array_end); | |
272906ef DW |
8705 | /* No room */ |
8706 | continue; | |
a20d2ba5 DW |
8707 | } |
8708 | return dl; | |
272906ef DW |
8709 | } |
8710 | ||
8711 | return dl; | |
8712 | } | |
8713 | ||
95d07a2c LM |
8714 | static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed) |
8715 | { | |
8716 | struct imsm_dev *dev2; | |
8717 | struct imsm_map *map; | |
8718 | struct dl *idisk; | |
8719 | int slot; | |
8720 | int idx; | |
8721 | __u8 state; | |
8722 | ||
8723 | dev2 = get_imsm_dev(cont->sb, dev_idx); | |
8724 | if (dev2) { | |
238c0a71 | 8725 | state = imsm_check_degraded(cont->sb, dev2, failed, MAP_0); |
95d07a2c | 8726 | if (state == IMSM_T_STATE_FAILED) { |
238c0a71 | 8727 | map = get_imsm_map(dev2, MAP_0); |
95d07a2c LM |
8728 | if (!map) |
8729 | return 1; | |
8730 | for (slot = 0; slot < map->num_members; slot++) { | |
8731 | /* | |
8732 | * Check if failed disks are deleted from intel | |
8733 | * disk list or are marked to be deleted | |
8734 | */ | |
238c0a71 | 8735 | idx = get_imsm_disk_idx(dev2, slot, MAP_X); |
95d07a2c LM |
8736 | idisk = get_imsm_dl_disk(cont->sb, idx); |
8737 | /* | |
8738 | * Do not rebuild the array if failed disks | |
8739 | * from failed sub-array are not removed from | |
8740 | * container. | |
8741 | */ | |
8742 | if (idisk && | |
8743 | is_failed(&idisk->disk) && | |
8744 | (idisk->action != DISK_REMOVE)) | |
8745 | return 0; | |
8746 | } | |
8747 | } | |
8748 | } | |
8749 | return 1; | |
8750 | } | |
8751 | ||
88758e9d DW |
8752 | static struct mdinfo *imsm_activate_spare(struct active_array *a, |
8753 | struct metadata_update **updates) | |
8754 | { | |
8755 | /** | |
d23fe947 DW |
8756 | * Find a device with unused free space and use it to replace a |
8757 | * failed/vacant region in an array. We replace failed regions one a | |
8758 | * array at a time. The result is that a new spare disk will be added | |
8759 | * to the first failed array and after the monitor has finished | |
8760 | * propagating failures the remainder will be consumed. | |
88758e9d | 8761 | * |
d23fe947 DW |
8762 | * FIXME add a capability for mdmon to request spares from another |
8763 | * container. | |
88758e9d DW |
8764 | */ |
8765 | ||
8766 | struct intel_super *super = a->container->sb; | |
88758e9d | 8767 | int inst = a->info.container_member; |
949c47a0 | 8768 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 8769 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
88758e9d DW |
8770 | int failed = a->info.array.raid_disks; |
8771 | struct mdinfo *rv = NULL; | |
8772 | struct mdinfo *d; | |
8773 | struct mdinfo *di; | |
8774 | struct metadata_update *mu; | |
8775 | struct dl *dl; | |
8776 | struct imsm_update_activate_spare *u; | |
8777 | int num_spares = 0; | |
8778 | int i; | |
95d07a2c | 8779 | int allowed; |
88758e9d DW |
8780 | |
8781 | for (d = a->info.devs ; d ; d = d->next) { | |
8782 | if ((d->curr_state & DS_FAULTY) && | |
8783 | d->state_fd >= 0) | |
8784 | /* wait for Removal to happen */ | |
8785 | return NULL; | |
8786 | if (d->state_fd >= 0) | |
8787 | failed--; | |
8788 | } | |
8789 | ||
8790 | dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n", | |
8791 | inst, failed, a->info.array.raid_disks, a->info.array.level); | |
1af97990 | 8792 | |
e2962bfc AK |
8793 | if (imsm_reshape_blocks_arrays_changes(super)) |
8794 | return NULL; | |
1af97990 | 8795 | |
fc8ca064 AK |
8796 | /* Cannot activate another spare if rebuild is in progress already |
8797 | */ | |
8798 | if (is_rebuilding(dev)) { | |
7a862a02 | 8799 | dprintf("imsm: No spare activation allowed. Rebuild in progress already.\n"); |
fc8ca064 AK |
8800 | return NULL; |
8801 | } | |
8802 | ||
89c67882 AK |
8803 | if (a->info.array.level == 4) |
8804 | /* No repair for takeovered array | |
8805 | * imsm doesn't support raid4 | |
8806 | */ | |
8807 | return NULL; | |
8808 | ||
3b451610 AK |
8809 | if (imsm_check_degraded(super, dev, failed, MAP_0) != |
8810 | IMSM_T_STATE_DEGRADED) | |
88758e9d DW |
8811 | return NULL; |
8812 | ||
83ca7d45 AP |
8813 | if (get_imsm_map(dev, MAP_0)->map_state == IMSM_T_STATE_UNINITIALIZED) { |
8814 | dprintf("imsm: No spare activation allowed. Volume is not initialized.\n"); | |
8815 | return NULL; | |
8816 | } | |
8817 | ||
95d07a2c LM |
8818 | /* |
8819 | * If there are any failed disks check state of the other volume. | |
8820 | * Block rebuild if the another one is failed until failed disks | |
8821 | * are removed from container. | |
8822 | */ | |
8823 | if (failed) { | |
7a862a02 | 8824 | dprintf("found failed disks in %.*s, check if there anotherfailed sub-array.\n", |
c4acd1e5 | 8825 | MAX_RAID_SERIAL_LEN, dev->volume); |
95d07a2c LM |
8826 | /* check if states of the other volumes allow for rebuild */ |
8827 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
8828 | if (i != inst) { | |
8829 | allowed = imsm_rebuild_allowed(a->container, | |
8830 | i, failed); | |
8831 | if (!allowed) | |
8832 | return NULL; | |
8833 | } | |
8834 | } | |
8835 | } | |
8836 | ||
88758e9d | 8837 | /* For each slot, if it is not working, find a spare */ |
88758e9d DW |
8838 | for (i = 0; i < a->info.array.raid_disks; i++) { |
8839 | for (d = a->info.devs ; d ; d = d->next) | |
8840 | if (d->disk.raid_disk == i) | |
8841 | break; | |
8842 | dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0); | |
8843 | if (d && (d->state_fd >= 0)) | |
8844 | continue; | |
8845 | ||
272906ef | 8846 | /* |
a20d2ba5 DW |
8847 | * OK, this device needs recovery. Try to re-add the |
8848 | * previous occupant of this slot, if this fails see if | |
8849 | * we can continue the assimilation of a spare that was | |
8850 | * partially assimilated, finally try to activate a new | |
8851 | * spare. | |
272906ef DW |
8852 | */ |
8853 | dl = imsm_readd(super, i, a); | |
8854 | if (!dl) | |
b303fe21 | 8855 | dl = imsm_add_spare(super, i, a, 0, rv); |
a20d2ba5 | 8856 | if (!dl) |
b303fe21 | 8857 | dl = imsm_add_spare(super, i, a, 1, rv); |
272906ef DW |
8858 | if (!dl) |
8859 | continue; | |
1011e834 | 8860 | |
272906ef | 8861 | /* found a usable disk with enough space */ |
503975b9 | 8862 | di = xcalloc(1, sizeof(*di)); |
272906ef DW |
8863 | |
8864 | /* dl->index will be -1 in the case we are activating a | |
8865 | * pristine spare. imsm_process_update() will create a | |
8866 | * new index in this case. Once a disk is found to be | |
8867 | * failed in all member arrays it is kicked from the | |
8868 | * metadata | |
8869 | */ | |
8870 | di->disk.number = dl->index; | |
d23fe947 | 8871 | |
272906ef DW |
8872 | /* (ab)use di->devs to store a pointer to the device |
8873 | * we chose | |
8874 | */ | |
8875 | di->devs = (struct mdinfo *) dl; | |
8876 | ||
8877 | di->disk.raid_disk = i; | |
8878 | di->disk.major = dl->major; | |
8879 | di->disk.minor = dl->minor; | |
8880 | di->disk.state = 0; | |
d23534e4 | 8881 | di->recovery_start = 0; |
5551b113 | 8882 | di->data_offset = pba_of_lba0(map); |
272906ef DW |
8883 | di->component_size = a->info.component_size; |
8884 | di->container_member = inst; | |
5e46202e | 8885 | di->bb.supported = 1; |
2c8890e9 | 8886 | if (a->info.consistency_policy == CONSISTENCY_POLICY_PPL) { |
2432ce9b | 8887 | di->ppl_sector = get_ppl_sector(super, inst); |
c2462068 | 8888 | di->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9; |
2432ce9b | 8889 | } |
148acb7b | 8890 | super->random = random32(); |
272906ef DW |
8891 | di->next = rv; |
8892 | rv = di; | |
8893 | num_spares++; | |
8894 | dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor, | |
8895 | i, di->data_offset); | |
88758e9d DW |
8896 | } |
8897 | ||
8898 | if (!rv) | |
8899 | /* No spares found */ | |
8900 | return rv; | |
8901 | /* Now 'rv' has a list of devices to return. | |
8902 | * Create a metadata_update record to update the | |
8903 | * disk_ord_tbl for the array | |
8904 | */ | |
503975b9 | 8905 | mu = xmalloc(sizeof(*mu)); |
1011e834 | 8906 | mu->buf = xcalloc(num_spares, |
503975b9 | 8907 | sizeof(struct imsm_update_activate_spare)); |
88758e9d | 8908 | mu->space = NULL; |
cb23f1f4 | 8909 | mu->space_list = NULL; |
88758e9d DW |
8910 | mu->len = sizeof(struct imsm_update_activate_spare) * num_spares; |
8911 | mu->next = *updates; | |
8912 | u = (struct imsm_update_activate_spare *) mu->buf; | |
8913 | ||
8914 | for (di = rv ; di ; di = di->next) { | |
8915 | u->type = update_activate_spare; | |
d23fe947 DW |
8916 | u->dl = (struct dl *) di->devs; |
8917 | di->devs = NULL; | |
88758e9d DW |
8918 | u->slot = di->disk.raid_disk; |
8919 | u->array = inst; | |
8920 | u->next = u + 1; | |
8921 | u++; | |
8922 | } | |
8923 | (u-1)->next = NULL; | |
8924 | *updates = mu; | |
8925 | ||
8926 | return rv; | |
8927 | } | |
8928 | ||
54c2c1ea | 8929 | static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u) |
8273f55e | 8930 | { |
54c2c1ea | 8931 | struct imsm_dev *dev = get_imsm_dev(super, idx); |
238c0a71 AK |
8932 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
8933 | struct imsm_map *new_map = get_imsm_map(&u->dev, MAP_0); | |
54c2c1ea DW |
8934 | struct disk_info *inf = get_disk_info(u); |
8935 | struct imsm_disk *disk; | |
8273f55e DW |
8936 | int i; |
8937 | int j; | |
8273f55e | 8938 | |
54c2c1ea | 8939 | for (i = 0; i < map->num_members; i++) { |
238c0a71 | 8940 | disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, MAP_X)); |
54c2c1ea DW |
8941 | for (j = 0; j < new_map->num_members; j++) |
8942 | if (serialcmp(disk->serial, inf[j].serial) == 0) | |
8273f55e DW |
8943 | return 1; |
8944 | } | |
8945 | ||
8946 | return 0; | |
8947 | } | |
8948 | ||
1a64be56 LM |
8949 | static struct dl *get_disk_super(struct intel_super *super, int major, int minor) |
8950 | { | |
594dc1b8 JS |
8951 | struct dl *dl; |
8952 | ||
1a64be56 | 8953 | for (dl = super->disks; dl; dl = dl->next) |
089f9d79 | 8954 | if (dl->major == major && dl->minor == minor) |
1a64be56 LM |
8955 | return dl; |
8956 | return NULL; | |
8957 | } | |
8958 | ||
8959 | static int remove_disk_super(struct intel_super *super, int major, int minor) | |
8960 | { | |
594dc1b8 | 8961 | struct dl *prev; |
1a64be56 LM |
8962 | struct dl *dl; |
8963 | ||
8964 | prev = NULL; | |
8965 | for (dl = super->disks; dl; dl = dl->next) { | |
089f9d79 | 8966 | if (dl->major == major && dl->minor == minor) { |
1a64be56 LM |
8967 | /* remove */ |
8968 | if (prev) | |
8969 | prev->next = dl->next; | |
8970 | else | |
8971 | super->disks = dl->next; | |
8972 | dl->next = NULL; | |
8973 | __free_imsm_disk(dl); | |
1ade5cc1 | 8974 | dprintf("removed %x:%x\n", major, minor); |
1a64be56 LM |
8975 | break; |
8976 | } | |
8977 | prev = dl; | |
8978 | } | |
8979 | return 0; | |
8980 | } | |
8981 | ||
f21e18ca | 8982 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index); |
ae6aad82 | 8983 | |
1a64be56 LM |
8984 | static int add_remove_disk_update(struct intel_super *super) |
8985 | { | |
8986 | int check_degraded = 0; | |
594dc1b8 JS |
8987 | struct dl *disk; |
8988 | ||
1a64be56 LM |
8989 | /* add/remove some spares to/from the metadata/contrainer */ |
8990 | while (super->disk_mgmt_list) { | |
8991 | struct dl *disk_cfg; | |
8992 | ||
8993 | disk_cfg = super->disk_mgmt_list; | |
8994 | super->disk_mgmt_list = disk_cfg->next; | |
8995 | disk_cfg->next = NULL; | |
8996 | ||
8997 | if (disk_cfg->action == DISK_ADD) { | |
8998 | disk_cfg->next = super->disks; | |
8999 | super->disks = disk_cfg; | |
9000 | check_degraded = 1; | |
1ade5cc1 N |
9001 | dprintf("added %x:%x\n", |
9002 | disk_cfg->major, disk_cfg->minor); | |
1a64be56 LM |
9003 | } else if (disk_cfg->action == DISK_REMOVE) { |
9004 | dprintf("Disk remove action processed: %x.%x\n", | |
9005 | disk_cfg->major, disk_cfg->minor); | |
9006 | disk = get_disk_super(super, | |
9007 | disk_cfg->major, | |
9008 | disk_cfg->minor); | |
9009 | if (disk) { | |
9010 | /* store action status */ | |
9011 | disk->action = DISK_REMOVE; | |
9012 | /* remove spare disks only */ | |
9013 | if (disk->index == -1) { | |
9014 | remove_disk_super(super, | |
9015 | disk_cfg->major, | |
9016 | disk_cfg->minor); | |
9017 | } | |
9018 | } | |
9019 | /* release allocate disk structure */ | |
9020 | __free_imsm_disk(disk_cfg); | |
9021 | } | |
9022 | } | |
9023 | return check_degraded; | |
9024 | } | |
9025 | ||
a29911da PC |
9026 | static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u, |
9027 | struct intel_super *super, | |
9028 | void ***space_list) | |
9029 | { | |
9030 | struct intel_dev *id; | |
9031 | void **tofree = NULL; | |
9032 | int ret_val = 0; | |
9033 | ||
1ade5cc1 | 9034 | dprintf("(enter)\n"); |
089f9d79 | 9035 | if (u->subdev < 0 || u->subdev > 1) { |
a29911da PC |
9036 | dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev); |
9037 | return ret_val; | |
9038 | } | |
089f9d79 | 9039 | if (space_list == NULL || *space_list == NULL) { |
a29911da PC |
9040 | dprintf("imsm: Error: Memory is not allocated\n"); |
9041 | return ret_val; | |
9042 | } | |
9043 | ||
9044 | for (id = super->devlist ; id; id = id->next) { | |
9045 | if (id->index == (unsigned)u->subdev) { | |
9046 | struct imsm_dev *dev = get_imsm_dev(super, u->subdev); | |
9047 | struct imsm_map *map; | |
9048 | struct imsm_dev *new_dev = | |
9049 | (struct imsm_dev *)*space_list; | |
238c0a71 | 9050 | struct imsm_map *migr_map = get_imsm_map(dev, MAP_1); |
a29911da PC |
9051 | int to_state; |
9052 | struct dl *new_disk; | |
9053 | ||
9054 | if (new_dev == NULL) | |
9055 | return ret_val; | |
9056 | *space_list = **space_list; | |
9057 | memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0)); | |
238c0a71 | 9058 | map = get_imsm_map(new_dev, MAP_0); |
a29911da PC |
9059 | if (migr_map) { |
9060 | dprintf("imsm: Error: migration in progress"); | |
9061 | return ret_val; | |
9062 | } | |
9063 | ||
9064 | to_state = map->map_state; | |
9065 | if ((u->new_level == 5) && (map->raid_level == 0)) { | |
9066 | map->num_members++; | |
9067 | /* this should not happen */ | |
9068 | if (u->new_disks[0] < 0) { | |
9069 | map->failed_disk_num = | |
9070 | map->num_members - 1; | |
9071 | to_state = IMSM_T_STATE_DEGRADED; | |
9072 | } else | |
9073 | to_state = IMSM_T_STATE_NORMAL; | |
9074 | } | |
8e59f3d8 | 9075 | migrate(new_dev, super, to_state, MIGR_GEN_MIGR); |
a29911da PC |
9076 | if (u->new_level > -1) |
9077 | map->raid_level = u->new_level; | |
238c0a71 | 9078 | migr_map = get_imsm_map(new_dev, MAP_1); |
a29911da PC |
9079 | if ((u->new_level == 5) && |
9080 | (migr_map->raid_level == 0)) { | |
9081 | int ord = map->num_members - 1; | |
9082 | migr_map->num_members--; | |
9083 | if (u->new_disks[0] < 0) | |
9084 | ord |= IMSM_ORD_REBUILD; | |
9085 | set_imsm_ord_tbl_ent(map, | |
9086 | map->num_members - 1, | |
9087 | ord); | |
9088 | } | |
9089 | id->dev = new_dev; | |
9090 | tofree = (void **)dev; | |
9091 | ||
4bba0439 PC |
9092 | /* update chunk size |
9093 | */ | |
06fb291a PB |
9094 | if (u->new_chunksize > 0) { |
9095 | unsigned long long num_data_stripes; | |
9096 | int used_disks = | |
9097 | imsm_num_data_members(dev, MAP_0); | |
9098 | ||
9099 | if (used_disks == 0) | |
9100 | return ret_val; | |
9101 | ||
4bba0439 PC |
9102 | map->blocks_per_strip = |
9103 | __cpu_to_le16(u->new_chunksize * 2); | |
06fb291a PB |
9104 | num_data_stripes = |
9105 | (join_u32(dev->size_low, dev->size_high) | |
9106 | / used_disks); | |
9107 | num_data_stripes /= map->blocks_per_strip; | |
9108 | num_data_stripes /= map->num_domains; | |
9109 | set_num_data_stripes(map, num_data_stripes); | |
9110 | } | |
4bba0439 | 9111 | |
a29911da PC |
9112 | /* add disk |
9113 | */ | |
089f9d79 JS |
9114 | if (u->new_level != 5 || migr_map->raid_level != 0 || |
9115 | migr_map->raid_level == map->raid_level) | |
a29911da PC |
9116 | goto skip_disk_add; |
9117 | ||
9118 | if (u->new_disks[0] >= 0) { | |
9119 | /* use passes spare | |
9120 | */ | |
9121 | new_disk = get_disk_super(super, | |
9122 | major(u->new_disks[0]), | |
9123 | minor(u->new_disks[0])); | |
7a862a02 | 9124 | dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n", |
a29911da PC |
9125 | major(u->new_disks[0]), |
9126 | minor(u->new_disks[0]), | |
9127 | new_disk, new_disk->index); | |
9128 | if (new_disk == NULL) | |
9129 | goto error_disk_add; | |
9130 | ||
9131 | new_disk->index = map->num_members - 1; | |
9132 | /* slot to fill in autolayout | |
9133 | */ | |
9134 | new_disk->raiddisk = new_disk->index; | |
9135 | new_disk->disk.status |= CONFIGURED_DISK; | |
9136 | new_disk->disk.status &= ~SPARE_DISK; | |
9137 | } else | |
9138 | goto error_disk_add; | |
9139 | ||
9140 | skip_disk_add: | |
9141 | *tofree = *space_list; | |
9142 | /* calculate new size | |
9143 | */ | |
f3871fdc | 9144 | imsm_set_array_size(new_dev, -1); |
a29911da PC |
9145 | |
9146 | ret_val = 1; | |
9147 | } | |
9148 | } | |
9149 | ||
9150 | if (tofree) | |
9151 | *space_list = tofree; | |
9152 | return ret_val; | |
9153 | ||
9154 | error_disk_add: | |
9155 | dprintf("Error: imsm: Cannot find disk.\n"); | |
9156 | return ret_val; | |
9157 | } | |
9158 | ||
f3871fdc AK |
9159 | static int apply_size_change_update(struct imsm_update_size_change *u, |
9160 | struct intel_super *super) | |
9161 | { | |
9162 | struct intel_dev *id; | |
9163 | int ret_val = 0; | |
9164 | ||
1ade5cc1 | 9165 | dprintf("(enter)\n"); |
089f9d79 | 9166 | if (u->subdev < 0 || u->subdev > 1) { |
f3871fdc AK |
9167 | dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev); |
9168 | return ret_val; | |
9169 | } | |
9170 | ||
9171 | for (id = super->devlist ; id; id = id->next) { | |
9172 | if (id->index == (unsigned)u->subdev) { | |
9173 | struct imsm_dev *dev = get_imsm_dev(super, u->subdev); | |
9174 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
9175 | int used_disks = imsm_num_data_members(dev, MAP_0); | |
9176 | unsigned long long blocks_per_member; | |
06fb291a | 9177 | unsigned long long num_data_stripes; |
f3871fdc AK |
9178 | |
9179 | /* calculate new size | |
9180 | */ | |
9181 | blocks_per_member = u->new_size / used_disks; | |
06fb291a PB |
9182 | num_data_stripes = blocks_per_member / |
9183 | map->blocks_per_strip; | |
9184 | num_data_stripes /= map->num_domains; | |
9185 | dprintf("(size: %llu, blocks per member: %llu, num_data_stipes: %llu)\n", | |
9186 | u->new_size, blocks_per_member, | |
9187 | num_data_stripes); | |
f3871fdc | 9188 | set_blocks_per_member(map, blocks_per_member); |
06fb291a | 9189 | set_num_data_stripes(map, num_data_stripes); |
f3871fdc AK |
9190 | imsm_set_array_size(dev, u->new_size); |
9191 | ||
9192 | ret_val = 1; | |
9193 | break; | |
9194 | } | |
9195 | } | |
9196 | ||
9197 | return ret_val; | |
9198 | } | |
9199 | ||
061d7da3 | 9200 | static int apply_update_activate_spare(struct imsm_update_activate_spare *u, |
ca9de185 | 9201 | struct intel_super *super, |
061d7da3 LO |
9202 | struct active_array *active_array) |
9203 | { | |
9204 | struct imsm_super *mpb = super->anchor; | |
9205 | struct imsm_dev *dev = get_imsm_dev(super, u->array); | |
238c0a71 | 9206 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
061d7da3 LO |
9207 | struct imsm_map *migr_map; |
9208 | struct active_array *a; | |
9209 | struct imsm_disk *disk; | |
9210 | __u8 to_state; | |
9211 | struct dl *dl; | |
9212 | unsigned int found; | |
9213 | int failed; | |
5961eeec | 9214 | int victim; |
061d7da3 | 9215 | int i; |
5961eeec | 9216 | int second_map_created = 0; |
061d7da3 | 9217 | |
5961eeec | 9218 | for (; u; u = u->next) { |
238c0a71 | 9219 | victim = get_imsm_disk_idx(dev, u->slot, MAP_X); |
061d7da3 | 9220 | |
5961eeec | 9221 | if (victim < 0) |
9222 | return 0; | |
061d7da3 | 9223 | |
5961eeec | 9224 | for (dl = super->disks; dl; dl = dl->next) |
9225 | if (dl == u->dl) | |
9226 | break; | |
061d7da3 | 9227 | |
5961eeec | 9228 | if (!dl) { |
7a862a02 | 9229 | pr_err("error: imsm_activate_spare passed an unknown disk (index: %d)\n", |
5961eeec | 9230 | u->dl->index); |
9231 | return 0; | |
9232 | } | |
061d7da3 | 9233 | |
5961eeec | 9234 | /* count failures (excluding rebuilds and the victim) |
9235 | * to determine map[0] state | |
9236 | */ | |
9237 | failed = 0; | |
9238 | for (i = 0; i < map->num_members; i++) { | |
9239 | if (i == u->slot) | |
9240 | continue; | |
9241 | disk = get_imsm_disk(super, | |
238c0a71 | 9242 | get_imsm_disk_idx(dev, i, MAP_X)); |
5961eeec | 9243 | if (!disk || is_failed(disk)) |
9244 | failed++; | |
9245 | } | |
061d7da3 | 9246 | |
5961eeec | 9247 | /* adding a pristine spare, assign a new index */ |
9248 | if (dl->index < 0) { | |
9249 | dl->index = super->anchor->num_disks; | |
9250 | super->anchor->num_disks++; | |
9251 | } | |
9252 | disk = &dl->disk; | |
9253 | disk->status |= CONFIGURED_DISK; | |
9254 | disk->status &= ~SPARE_DISK; | |
9255 | ||
9256 | /* mark rebuild */ | |
238c0a71 | 9257 | to_state = imsm_check_degraded(super, dev, failed, MAP_0); |
5961eeec | 9258 | if (!second_map_created) { |
9259 | second_map_created = 1; | |
9260 | map->map_state = IMSM_T_STATE_DEGRADED; | |
9261 | migrate(dev, super, to_state, MIGR_REBUILD); | |
9262 | } else | |
9263 | map->map_state = to_state; | |
238c0a71 | 9264 | migr_map = get_imsm_map(dev, MAP_1); |
5961eeec | 9265 | set_imsm_ord_tbl_ent(map, u->slot, dl->index); |
9266 | set_imsm_ord_tbl_ent(migr_map, u->slot, | |
9267 | dl->index | IMSM_ORD_REBUILD); | |
9268 | ||
9269 | /* update the family_num to mark a new container | |
9270 | * generation, being careful to record the existing | |
9271 | * family_num in orig_family_num to clean up after | |
9272 | * earlier mdadm versions that neglected to set it. | |
9273 | */ | |
9274 | if (mpb->orig_family_num == 0) | |
9275 | mpb->orig_family_num = mpb->family_num; | |
9276 | mpb->family_num += super->random; | |
9277 | ||
9278 | /* count arrays using the victim in the metadata */ | |
9279 | found = 0; | |
9280 | for (a = active_array; a ; a = a->next) { | |
9281 | dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 9282 | map = get_imsm_map(dev, MAP_0); |
061d7da3 | 9283 | |
5961eeec | 9284 | if (get_imsm_disk_slot(map, victim) >= 0) |
9285 | found++; | |
9286 | } | |
061d7da3 | 9287 | |
5961eeec | 9288 | /* delete the victim if it is no longer being |
9289 | * utilized anywhere | |
061d7da3 | 9290 | */ |
5961eeec | 9291 | if (!found) { |
9292 | struct dl **dlp; | |
061d7da3 | 9293 | |
5961eeec | 9294 | /* We know that 'manager' isn't touching anything, |
9295 | * so it is safe to delete | |
9296 | */ | |
9297 | for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next) | |
061d7da3 LO |
9298 | if ((*dlp)->index == victim) |
9299 | break; | |
5961eeec | 9300 | |
9301 | /* victim may be on the missing list */ | |
9302 | if (!*dlp) | |
9303 | for (dlp = &super->missing; *dlp; | |
9304 | dlp = &(*dlp)->next) | |
9305 | if ((*dlp)->index == victim) | |
9306 | break; | |
9307 | imsm_delete(super, dlp, victim); | |
9308 | } | |
061d7da3 LO |
9309 | } |
9310 | ||
9311 | return 1; | |
9312 | } | |
a29911da | 9313 | |
2e5dc010 N |
9314 | static int apply_reshape_container_disks_update(struct imsm_update_reshape *u, |
9315 | struct intel_super *super, | |
9316 | void ***space_list) | |
9317 | { | |
9318 | struct dl *new_disk; | |
9319 | struct intel_dev *id; | |
9320 | int i; | |
9321 | int delta_disks = u->new_raid_disks - u->old_raid_disks; | |
ee4beede | 9322 | int disk_count = u->old_raid_disks; |
2e5dc010 N |
9323 | void **tofree = NULL; |
9324 | int devices_to_reshape = 1; | |
9325 | struct imsm_super *mpb = super->anchor; | |
9326 | int ret_val = 0; | |
d098291a | 9327 | unsigned int dev_id; |
2e5dc010 | 9328 | |
1ade5cc1 | 9329 | dprintf("(enter)\n"); |
2e5dc010 N |
9330 | |
9331 | /* enable spares to use in array */ | |
9332 | for (i = 0; i < delta_disks; i++) { | |
9333 | new_disk = get_disk_super(super, | |
9334 | major(u->new_disks[i]), | |
9335 | minor(u->new_disks[i])); | |
7a862a02 | 9336 | dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n", |
2e5dc010 N |
9337 | major(u->new_disks[i]), minor(u->new_disks[i]), |
9338 | new_disk, new_disk->index); | |
089f9d79 JS |
9339 | if (new_disk == NULL || |
9340 | (new_disk->index >= 0 && | |
9341 | new_disk->index < u->old_raid_disks)) | |
2e5dc010 | 9342 | goto update_reshape_exit; |
ee4beede | 9343 | new_disk->index = disk_count++; |
2e5dc010 N |
9344 | /* slot to fill in autolayout |
9345 | */ | |
9346 | new_disk->raiddisk = new_disk->index; | |
9347 | new_disk->disk.status |= | |
9348 | CONFIGURED_DISK; | |
9349 | new_disk->disk.status &= ~SPARE_DISK; | |
9350 | } | |
9351 | ||
ed7333bd AK |
9352 | dprintf("imsm: volume set mpb->num_raid_devs = %i\n", |
9353 | mpb->num_raid_devs); | |
2e5dc010 N |
9354 | /* manage changes in volume |
9355 | */ | |
d098291a | 9356 | for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) { |
2e5dc010 N |
9357 | void **sp = *space_list; |
9358 | struct imsm_dev *newdev; | |
9359 | struct imsm_map *newmap, *oldmap; | |
9360 | ||
d098291a AK |
9361 | for (id = super->devlist ; id; id = id->next) { |
9362 | if (id->index == dev_id) | |
9363 | break; | |
9364 | } | |
9365 | if (id == NULL) | |
9366 | break; | |
2e5dc010 N |
9367 | if (!sp) |
9368 | continue; | |
9369 | *space_list = *sp; | |
9370 | newdev = (void*)sp; | |
9371 | /* Copy the dev, but not (all of) the map */ | |
9372 | memcpy(newdev, id->dev, sizeof(*newdev)); | |
238c0a71 AK |
9373 | oldmap = get_imsm_map(id->dev, MAP_0); |
9374 | newmap = get_imsm_map(newdev, MAP_0); | |
2e5dc010 N |
9375 | /* Copy the current map */ |
9376 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); | |
9377 | /* update one device only | |
9378 | */ | |
9379 | if (devices_to_reshape) { | |
ed7333bd AK |
9380 | dprintf("imsm: modifying subdev: %i\n", |
9381 | id->index); | |
2e5dc010 N |
9382 | devices_to_reshape--; |
9383 | newdev->vol.migr_state = 1; | |
9384 | newdev->vol.curr_migr_unit = 0; | |
ea672ee1 | 9385 | set_migr_type(newdev, MIGR_GEN_MIGR); |
2e5dc010 N |
9386 | newmap->num_members = u->new_raid_disks; |
9387 | for (i = 0; i < delta_disks; i++) { | |
9388 | set_imsm_ord_tbl_ent(newmap, | |
9389 | u->old_raid_disks + i, | |
9390 | u->old_raid_disks + i); | |
9391 | } | |
9392 | /* New map is correct, now need to save old map | |
9393 | */ | |
238c0a71 | 9394 | newmap = get_imsm_map(newdev, MAP_1); |
2e5dc010 N |
9395 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); |
9396 | ||
f3871fdc | 9397 | imsm_set_array_size(newdev, -1); |
2e5dc010 N |
9398 | } |
9399 | ||
9400 | sp = (void **)id->dev; | |
9401 | id->dev = newdev; | |
9402 | *sp = tofree; | |
9403 | tofree = sp; | |
8e59f3d8 AK |
9404 | |
9405 | /* Clear migration record */ | |
9406 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
2e5dc010 | 9407 | } |
819bc634 AK |
9408 | if (tofree) |
9409 | *space_list = tofree; | |
2e5dc010 N |
9410 | ret_val = 1; |
9411 | ||
9412 | update_reshape_exit: | |
9413 | ||
9414 | return ret_val; | |
9415 | } | |
9416 | ||
bb025c2f | 9417 | static int apply_takeover_update(struct imsm_update_takeover *u, |
8ca6df95 KW |
9418 | struct intel_super *super, |
9419 | void ***space_list) | |
bb025c2f KW |
9420 | { |
9421 | struct imsm_dev *dev = NULL; | |
8ca6df95 KW |
9422 | struct intel_dev *dv; |
9423 | struct imsm_dev *dev_new; | |
bb025c2f KW |
9424 | struct imsm_map *map; |
9425 | struct dl *dm, *du; | |
8ca6df95 | 9426 | int i; |
bb025c2f KW |
9427 | |
9428 | for (dv = super->devlist; dv; dv = dv->next) | |
9429 | if (dv->index == (unsigned int)u->subarray) { | |
9430 | dev = dv->dev; | |
9431 | break; | |
9432 | } | |
9433 | ||
9434 | if (dev == NULL) | |
9435 | return 0; | |
9436 | ||
238c0a71 | 9437 | map = get_imsm_map(dev, MAP_0); |
bb025c2f KW |
9438 | |
9439 | if (u->direction == R10_TO_R0) { | |
06fb291a PB |
9440 | unsigned long long num_data_stripes; |
9441 | ||
9442 | map->num_domains = 1; | |
9443 | num_data_stripes = blocks_per_member(map); | |
9444 | num_data_stripes /= map->blocks_per_strip; | |
9445 | num_data_stripes /= map->num_domains; | |
9446 | set_num_data_stripes(map, num_data_stripes); | |
9447 | ||
43d5ec18 | 9448 | /* Number of failed disks must be half of initial disk number */ |
3b451610 AK |
9449 | if (imsm_count_failed(super, dev, MAP_0) != |
9450 | (map->num_members / 2)) | |
43d5ec18 KW |
9451 | return 0; |
9452 | ||
bb025c2f KW |
9453 | /* iterate through devices to mark removed disks as spare */ |
9454 | for (dm = super->disks; dm; dm = dm->next) { | |
9455 | if (dm->disk.status & FAILED_DISK) { | |
9456 | int idx = dm->index; | |
9457 | /* update indexes on the disk list */ | |
9458 | /* FIXME this loop-with-the-loop looks wrong, I'm not convinced | |
9459 | the index values will end up being correct.... NB */ | |
9460 | for (du = super->disks; du; du = du->next) | |
9461 | if (du->index > idx) | |
9462 | du->index--; | |
9463 | /* mark as spare disk */ | |
a8619d23 | 9464 | mark_spare(dm); |
bb025c2f KW |
9465 | } |
9466 | } | |
bb025c2f KW |
9467 | /* update map */ |
9468 | map->num_members = map->num_members / 2; | |
9469 | map->map_state = IMSM_T_STATE_NORMAL; | |
9470 | map->num_domains = 1; | |
9471 | map->raid_level = 0; | |
9472 | map->failed_disk_num = -1; | |
9473 | } | |
9474 | ||
8ca6df95 KW |
9475 | if (u->direction == R0_TO_R10) { |
9476 | void **space; | |
9477 | /* update slots in current disk list */ | |
9478 | for (dm = super->disks; dm; dm = dm->next) { | |
9479 | if (dm->index >= 0) | |
9480 | dm->index *= 2; | |
9481 | } | |
9482 | /* create new *missing* disks */ | |
9483 | for (i = 0; i < map->num_members; i++) { | |
9484 | space = *space_list; | |
9485 | if (!space) | |
9486 | continue; | |
9487 | *space_list = *space; | |
9488 | du = (void *)space; | |
9489 | memcpy(du, super->disks, sizeof(*du)); | |
8ca6df95 KW |
9490 | du->fd = -1; |
9491 | du->minor = 0; | |
9492 | du->major = 0; | |
9493 | du->index = (i * 2) + 1; | |
9494 | sprintf((char *)du->disk.serial, | |
9495 | " MISSING_%d", du->index); | |
9496 | sprintf((char *)du->serial, | |
9497 | "MISSING_%d", du->index); | |
9498 | du->next = super->missing; | |
9499 | super->missing = du; | |
9500 | } | |
9501 | /* create new dev and map */ | |
9502 | space = *space_list; | |
9503 | if (!space) | |
9504 | return 0; | |
9505 | *space_list = *space; | |
9506 | dev_new = (void *)space; | |
9507 | memcpy(dev_new, dev, sizeof(*dev)); | |
9508 | /* update new map */ | |
238c0a71 | 9509 | map = get_imsm_map(dev_new, MAP_0); |
8ca6df95 | 9510 | map->num_members = map->num_members * 2; |
1a2487c2 | 9511 | map->map_state = IMSM_T_STATE_DEGRADED; |
8ca6df95 KW |
9512 | map->num_domains = 2; |
9513 | map->raid_level = 1; | |
9514 | /* replace dev<->dev_new */ | |
9515 | dv->dev = dev_new; | |
9516 | } | |
bb025c2f KW |
9517 | /* update disk order table */ |
9518 | for (du = super->disks; du; du = du->next) | |
9519 | if (du->index >= 0) | |
9520 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
8ca6df95 | 9521 | for (du = super->missing; du; du = du->next) |
1a2487c2 KW |
9522 | if (du->index >= 0) { |
9523 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
4c9e8c1e | 9524 | mark_missing(super, dv->dev, &du->disk, du->index); |
1a2487c2 | 9525 | } |
bb025c2f KW |
9526 | |
9527 | return 1; | |
9528 | } | |
9529 | ||
e8319a19 DW |
9530 | static void imsm_process_update(struct supertype *st, |
9531 | struct metadata_update *update) | |
9532 | { | |
9533 | /** | |
9534 | * crack open the metadata_update envelope to find the update record | |
9535 | * update can be one of: | |
d195167d AK |
9536 | * update_reshape_container_disks - all the arrays in the container |
9537 | * are being reshaped to have more devices. We need to mark | |
9538 | * the arrays for general migration and convert selected spares | |
9539 | * into active devices. | |
9540 | * update_activate_spare - a spare device has replaced a failed | |
1011e834 N |
9541 | * device in an array, update the disk_ord_tbl. If this disk is |
9542 | * present in all member arrays then also clear the SPARE_DISK | |
9543 | * flag | |
d195167d AK |
9544 | * update_create_array |
9545 | * update_kill_array | |
9546 | * update_rename_array | |
9547 | * update_add_remove_disk | |
e8319a19 DW |
9548 | */ |
9549 | struct intel_super *super = st->sb; | |
4d7b1503 | 9550 | struct imsm_super *mpb; |
e8319a19 DW |
9551 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
9552 | ||
4d7b1503 DW |
9553 | /* update requires a larger buf but the allocation failed */ |
9554 | if (super->next_len && !super->next_buf) { | |
9555 | super->next_len = 0; | |
9556 | return; | |
9557 | } | |
9558 | ||
9559 | if (super->next_buf) { | |
9560 | memcpy(super->next_buf, super->buf, super->len); | |
9561 | free(super->buf); | |
9562 | super->len = super->next_len; | |
9563 | super->buf = super->next_buf; | |
9564 | ||
9565 | super->next_len = 0; | |
9566 | super->next_buf = NULL; | |
9567 | } | |
9568 | ||
9569 | mpb = super->anchor; | |
9570 | ||
e8319a19 | 9571 | switch (type) { |
0ec5d470 AK |
9572 | case update_general_migration_checkpoint: { |
9573 | struct intel_dev *id; | |
9574 | struct imsm_update_general_migration_checkpoint *u = | |
9575 | (void *)update->buf; | |
9576 | ||
1ade5cc1 | 9577 | dprintf("called for update_general_migration_checkpoint\n"); |
0ec5d470 AK |
9578 | |
9579 | /* find device under general migration */ | |
9580 | for (id = super->devlist ; id; id = id->next) { | |
9581 | if (is_gen_migration(id->dev)) { | |
9582 | id->dev->vol.curr_migr_unit = | |
9583 | __cpu_to_le32(u->curr_migr_unit); | |
9584 | super->updates_pending++; | |
9585 | } | |
9586 | } | |
9587 | break; | |
9588 | } | |
bb025c2f KW |
9589 | case update_takeover: { |
9590 | struct imsm_update_takeover *u = (void *)update->buf; | |
1a2487c2 KW |
9591 | if (apply_takeover_update(u, super, &update->space_list)) { |
9592 | imsm_update_version_info(super); | |
bb025c2f | 9593 | super->updates_pending++; |
1a2487c2 | 9594 | } |
bb025c2f KW |
9595 | break; |
9596 | } | |
9597 | ||
78b10e66 | 9598 | case update_reshape_container_disks: { |
d195167d | 9599 | struct imsm_update_reshape *u = (void *)update->buf; |
2e5dc010 N |
9600 | if (apply_reshape_container_disks_update( |
9601 | u, super, &update->space_list)) | |
9602 | super->updates_pending++; | |
78b10e66 N |
9603 | break; |
9604 | } | |
48c5303a | 9605 | case update_reshape_migration: { |
a29911da PC |
9606 | struct imsm_update_reshape_migration *u = (void *)update->buf; |
9607 | if (apply_reshape_migration_update( | |
9608 | u, super, &update->space_list)) | |
9609 | super->updates_pending++; | |
48c5303a PC |
9610 | break; |
9611 | } | |
f3871fdc AK |
9612 | case update_size_change: { |
9613 | struct imsm_update_size_change *u = (void *)update->buf; | |
9614 | if (apply_size_change_update(u, super)) | |
9615 | super->updates_pending++; | |
9616 | break; | |
9617 | } | |
e8319a19 | 9618 | case update_activate_spare: { |
1011e834 | 9619 | struct imsm_update_activate_spare *u = (void *) update->buf; |
061d7da3 LO |
9620 | if (apply_update_activate_spare(u, super, st->arrays)) |
9621 | super->updates_pending++; | |
8273f55e DW |
9622 | break; |
9623 | } | |
9624 | case update_create_array: { | |
9625 | /* someone wants to create a new array, we need to be aware of | |
9626 | * a few races/collisions: | |
9627 | * 1/ 'Create' called by two separate instances of mdadm | |
9628 | * 2/ 'Create' versus 'activate_spare': mdadm has chosen | |
9629 | * devices that have since been assimilated via | |
9630 | * activate_spare. | |
9631 | * In the event this update can not be carried out mdadm will | |
9632 | * (FIX ME) notice that its update did not take hold. | |
9633 | */ | |
9634 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 9635 | struct intel_dev *dv; |
8273f55e DW |
9636 | struct imsm_dev *dev; |
9637 | struct imsm_map *map, *new_map; | |
9638 | unsigned long long start, end; | |
9639 | unsigned long long new_start, new_end; | |
9640 | int i; | |
54c2c1ea DW |
9641 | struct disk_info *inf; |
9642 | struct dl *dl; | |
8273f55e DW |
9643 | |
9644 | /* handle racing creates: first come first serve */ | |
9645 | if (u->dev_idx < mpb->num_raid_devs) { | |
1ade5cc1 | 9646 | dprintf("subarray %d already defined\n", u->dev_idx); |
ba2de7ba | 9647 | goto create_error; |
8273f55e DW |
9648 | } |
9649 | ||
9650 | /* check update is next in sequence */ | |
9651 | if (u->dev_idx != mpb->num_raid_devs) { | |
1ade5cc1 N |
9652 | dprintf("can not create array %d expected index %d\n", |
9653 | u->dev_idx, mpb->num_raid_devs); | |
ba2de7ba | 9654 | goto create_error; |
8273f55e DW |
9655 | } |
9656 | ||
238c0a71 | 9657 | new_map = get_imsm_map(&u->dev, MAP_0); |
5551b113 CA |
9658 | new_start = pba_of_lba0(new_map); |
9659 | new_end = new_start + blocks_per_member(new_map); | |
54c2c1ea | 9660 | inf = get_disk_info(u); |
8273f55e DW |
9661 | |
9662 | /* handle activate_spare versus create race: | |
9663 | * check to make sure that overlapping arrays do not include | |
9664 | * overalpping disks | |
9665 | */ | |
9666 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
949c47a0 | 9667 | dev = get_imsm_dev(super, i); |
238c0a71 | 9668 | map = get_imsm_map(dev, MAP_0); |
5551b113 CA |
9669 | start = pba_of_lba0(map); |
9670 | end = start + blocks_per_member(map); | |
8273f55e DW |
9671 | if ((new_start >= start && new_start <= end) || |
9672 | (start >= new_start && start <= new_end)) | |
54c2c1ea DW |
9673 | /* overlap */; |
9674 | else | |
9675 | continue; | |
9676 | ||
9677 | if (disks_overlap(super, i, u)) { | |
1ade5cc1 | 9678 | dprintf("arrays overlap\n"); |
ba2de7ba | 9679 | goto create_error; |
8273f55e DW |
9680 | } |
9681 | } | |
8273f55e | 9682 | |
949c47a0 DW |
9683 | /* check that prepare update was successful */ |
9684 | if (!update->space) { | |
1ade5cc1 | 9685 | dprintf("prepare update failed\n"); |
ba2de7ba | 9686 | goto create_error; |
949c47a0 DW |
9687 | } |
9688 | ||
54c2c1ea DW |
9689 | /* check that all disks are still active before committing |
9690 | * changes. FIXME: could we instead handle this by creating a | |
9691 | * degraded array? That's probably not what the user expects, | |
9692 | * so better to drop this update on the floor. | |
9693 | */ | |
9694 | for (i = 0; i < new_map->num_members; i++) { | |
9695 | dl = serial_to_dl(inf[i].serial, super); | |
9696 | if (!dl) { | |
1ade5cc1 | 9697 | dprintf("disk disappeared\n"); |
ba2de7ba | 9698 | goto create_error; |
54c2c1ea | 9699 | } |
949c47a0 DW |
9700 | } |
9701 | ||
8273f55e | 9702 | super->updates_pending++; |
54c2c1ea DW |
9703 | |
9704 | /* convert spares to members and fixup ord_tbl */ | |
9705 | for (i = 0; i < new_map->num_members; i++) { | |
9706 | dl = serial_to_dl(inf[i].serial, super); | |
9707 | if (dl->index == -1) { | |
9708 | dl->index = mpb->num_disks; | |
9709 | mpb->num_disks++; | |
9710 | dl->disk.status |= CONFIGURED_DISK; | |
9711 | dl->disk.status &= ~SPARE_DISK; | |
9712 | } | |
9713 | set_imsm_ord_tbl_ent(new_map, i, dl->index); | |
9714 | } | |
9715 | ||
ba2de7ba DW |
9716 | dv = update->space; |
9717 | dev = dv->dev; | |
949c47a0 DW |
9718 | update->space = NULL; |
9719 | imsm_copy_dev(dev, &u->dev); | |
ba2de7ba DW |
9720 | dv->index = u->dev_idx; |
9721 | dv->next = super->devlist; | |
9722 | super->devlist = dv; | |
8273f55e | 9723 | mpb->num_raid_devs++; |
8273f55e | 9724 | |
4d1313e9 | 9725 | imsm_update_version_info(super); |
8273f55e | 9726 | break; |
ba2de7ba DW |
9727 | create_error: |
9728 | /* mdmon knows how to release update->space, but not | |
9729 | * ((struct intel_dev *) update->space)->dev | |
9730 | */ | |
9731 | if (update->space) { | |
9732 | dv = update->space; | |
9733 | free(dv->dev); | |
9734 | } | |
8273f55e | 9735 | break; |
e8319a19 | 9736 | } |
33414a01 DW |
9737 | case update_kill_array: { |
9738 | struct imsm_update_kill_array *u = (void *) update->buf; | |
9739 | int victim = u->dev_idx; | |
9740 | struct active_array *a; | |
9741 | struct intel_dev **dp; | |
9742 | struct imsm_dev *dev; | |
9743 | ||
9744 | /* sanity check that we are not affecting the uuid of | |
9745 | * active arrays, or deleting an active array | |
9746 | * | |
9747 | * FIXME when immutable ids are available, but note that | |
9748 | * we'll also need to fixup the invalidated/active | |
9749 | * subarray indexes in mdstat | |
9750 | */ | |
9751 | for (a = st->arrays; a; a = a->next) | |
9752 | if (a->info.container_member >= victim) | |
9753 | break; | |
9754 | /* by definition if mdmon is running at least one array | |
9755 | * is active in the container, so checking | |
9756 | * mpb->num_raid_devs is just extra paranoia | |
9757 | */ | |
9758 | dev = get_imsm_dev(super, victim); | |
9759 | if (a || !dev || mpb->num_raid_devs == 1) { | |
9760 | dprintf("failed to delete subarray-%d\n", victim); | |
9761 | break; | |
9762 | } | |
9763 | ||
9764 | for (dp = &super->devlist; *dp;) | |
f21e18ca | 9765 | if ((*dp)->index == (unsigned)super->current_vol) { |
33414a01 DW |
9766 | *dp = (*dp)->next; |
9767 | } else { | |
f21e18ca | 9768 | if ((*dp)->index > (unsigned)victim) |
33414a01 DW |
9769 | (*dp)->index--; |
9770 | dp = &(*dp)->next; | |
9771 | } | |
9772 | mpb->num_raid_devs--; | |
9773 | super->updates_pending++; | |
9774 | break; | |
9775 | } | |
aa534678 DW |
9776 | case update_rename_array: { |
9777 | struct imsm_update_rename_array *u = (void *) update->buf; | |
9778 | char name[MAX_RAID_SERIAL_LEN+1]; | |
9779 | int target = u->dev_idx; | |
9780 | struct active_array *a; | |
9781 | struct imsm_dev *dev; | |
9782 | ||
9783 | /* sanity check that we are not affecting the uuid of | |
9784 | * an active array | |
9785 | */ | |
9786 | snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name); | |
9787 | name[MAX_RAID_SERIAL_LEN] = '\0'; | |
9788 | for (a = st->arrays; a; a = a->next) | |
9789 | if (a->info.container_member == target) | |
9790 | break; | |
9791 | dev = get_imsm_dev(super, u->dev_idx); | |
9792 | if (a || !dev || !check_name(super, name, 1)) { | |
9793 | dprintf("failed to rename subarray-%d\n", target); | |
9794 | break; | |
9795 | } | |
9796 | ||
cdbe98cd | 9797 | snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name); |
aa534678 DW |
9798 | super->updates_pending++; |
9799 | break; | |
9800 | } | |
1a64be56 | 9801 | case update_add_remove_disk: { |
43dad3d6 | 9802 | /* we may be able to repair some arrays if disks are |
095b8088 | 9803 | * being added, check the status of add_remove_disk |
1a64be56 LM |
9804 | * if discs has been added. |
9805 | */ | |
9806 | if (add_remove_disk_update(super)) { | |
43dad3d6 | 9807 | struct active_array *a; |
072b727f DW |
9808 | |
9809 | super->updates_pending++; | |
1a64be56 | 9810 | for (a = st->arrays; a; a = a->next) |
43dad3d6 DW |
9811 | a->check_degraded = 1; |
9812 | } | |
43dad3d6 | 9813 | break; |
e8319a19 | 9814 | } |
bbab0940 TM |
9815 | case update_prealloc_badblocks_mem: |
9816 | break; | |
e6e9dd3f AP |
9817 | case update_rwh_policy: { |
9818 | struct imsm_update_rwh_policy *u = (void *)update->buf; | |
9819 | int target = u->dev_idx; | |
9820 | struct imsm_dev *dev = get_imsm_dev(super, target); | |
9821 | if (!dev) { | |
9822 | dprintf("could not find subarray-%d\n", target); | |
9823 | break; | |
9824 | } | |
9825 | ||
9826 | if (dev->rwh_policy != u->new_policy) { | |
9827 | dev->rwh_policy = u->new_policy; | |
9828 | super->updates_pending++; | |
9829 | } | |
9830 | break; | |
9831 | } | |
1a64be56 | 9832 | default: |
7a862a02 | 9833 | pr_err("error: unsuported process update type:(type: %d)\n", type); |
1a64be56 | 9834 | } |
e8319a19 | 9835 | } |
88758e9d | 9836 | |
bc0b9d34 PC |
9837 | static struct mdinfo *get_spares_for_grow(struct supertype *st); |
9838 | ||
5fe6f031 N |
9839 | static int imsm_prepare_update(struct supertype *st, |
9840 | struct metadata_update *update) | |
8273f55e | 9841 | { |
949c47a0 | 9842 | /** |
4d7b1503 DW |
9843 | * Allocate space to hold new disk entries, raid-device entries or a new |
9844 | * mpb if necessary. The manager synchronously waits for updates to | |
9845 | * complete in the monitor, so new mpb buffers allocated here can be | |
9846 | * integrated by the monitor thread without worrying about live pointers | |
9847 | * in the manager thread. | |
8273f55e | 9848 | */ |
095b8088 | 9849 | enum imsm_update_type type; |
4d7b1503 | 9850 | struct intel_super *super = st->sb; |
f36a9ecd | 9851 | unsigned int sector_size = super->sector_size; |
4d7b1503 DW |
9852 | struct imsm_super *mpb = super->anchor; |
9853 | size_t buf_len; | |
9854 | size_t len = 0; | |
949c47a0 | 9855 | |
095b8088 N |
9856 | if (update->len < (int)sizeof(type)) |
9857 | return 0; | |
9858 | ||
9859 | type = *(enum imsm_update_type *) update->buf; | |
9860 | ||
949c47a0 | 9861 | switch (type) { |
0ec5d470 | 9862 | case update_general_migration_checkpoint: |
095b8088 N |
9863 | if (update->len < (int)sizeof(struct imsm_update_general_migration_checkpoint)) |
9864 | return 0; | |
1ade5cc1 | 9865 | dprintf("called for update_general_migration_checkpoint\n"); |
0ec5d470 | 9866 | break; |
abedf5fc KW |
9867 | case update_takeover: { |
9868 | struct imsm_update_takeover *u = (void *)update->buf; | |
095b8088 N |
9869 | if (update->len < (int)sizeof(*u)) |
9870 | return 0; | |
abedf5fc KW |
9871 | if (u->direction == R0_TO_R10) { |
9872 | void **tail = (void **)&update->space_list; | |
9873 | struct imsm_dev *dev = get_imsm_dev(super, u->subarray); | |
238c0a71 | 9874 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
abedf5fc KW |
9875 | int num_members = map->num_members; |
9876 | void *space; | |
9877 | int size, i; | |
abedf5fc KW |
9878 | /* allocate memory for added disks */ |
9879 | for (i = 0; i < num_members; i++) { | |
9880 | size = sizeof(struct dl); | |
503975b9 | 9881 | space = xmalloc(size); |
abedf5fc KW |
9882 | *tail = space; |
9883 | tail = space; | |
9884 | *tail = NULL; | |
9885 | } | |
9886 | /* allocate memory for new device */ | |
9887 | size = sizeof_imsm_dev(super->devlist->dev, 0) + | |
9888 | (num_members * sizeof(__u32)); | |
503975b9 N |
9889 | space = xmalloc(size); |
9890 | *tail = space; | |
9891 | tail = space; | |
9892 | *tail = NULL; | |
9893 | len = disks_to_mpb_size(num_members * 2); | |
abedf5fc KW |
9894 | } |
9895 | ||
9896 | break; | |
9897 | } | |
78b10e66 | 9898 | case update_reshape_container_disks: { |
d195167d AK |
9899 | /* Every raid device in the container is about to |
9900 | * gain some more devices, and we will enter a | |
9901 | * reconfiguration. | |
9902 | * So each 'imsm_map' will be bigger, and the imsm_vol | |
9903 | * will now hold 2 of them. | |
9904 | * Thus we need new 'struct imsm_dev' allocations sized | |
9905 | * as sizeof_imsm_dev but with more devices in both maps. | |
9906 | */ | |
9907 | struct imsm_update_reshape *u = (void *)update->buf; | |
9908 | struct intel_dev *dl; | |
9909 | void **space_tail = (void**)&update->space_list; | |
9910 | ||
095b8088 N |
9911 | if (update->len < (int)sizeof(*u)) |
9912 | return 0; | |
9913 | ||
1ade5cc1 | 9914 | dprintf("for update_reshape\n"); |
d195167d AK |
9915 | |
9916 | for (dl = super->devlist; dl; dl = dl->next) { | |
9917 | int size = sizeof_imsm_dev(dl->dev, 1); | |
9918 | void *s; | |
d677e0b8 AK |
9919 | if (u->new_raid_disks > u->old_raid_disks) |
9920 | size += sizeof(__u32)*2* | |
9921 | (u->new_raid_disks - u->old_raid_disks); | |
503975b9 | 9922 | s = xmalloc(size); |
d195167d AK |
9923 | *space_tail = s; |
9924 | space_tail = s; | |
9925 | *space_tail = NULL; | |
9926 | } | |
9927 | ||
9928 | len = disks_to_mpb_size(u->new_raid_disks); | |
9929 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
78b10e66 N |
9930 | break; |
9931 | } | |
48c5303a | 9932 | case update_reshape_migration: { |
bc0b9d34 PC |
9933 | /* for migration level 0->5 we need to add disks |
9934 | * so the same as for container operation we will copy | |
9935 | * device to the bigger location. | |
9936 | * in memory prepared device and new disk area are prepared | |
9937 | * for usage in process update | |
9938 | */ | |
9939 | struct imsm_update_reshape_migration *u = (void *)update->buf; | |
9940 | struct intel_dev *id; | |
9941 | void **space_tail = (void **)&update->space_list; | |
9942 | int size; | |
9943 | void *s; | |
9944 | int current_level = -1; | |
9945 | ||
095b8088 N |
9946 | if (update->len < (int)sizeof(*u)) |
9947 | return 0; | |
9948 | ||
1ade5cc1 | 9949 | dprintf("for update_reshape\n"); |
bc0b9d34 PC |
9950 | |
9951 | /* add space for bigger array in update | |
9952 | */ | |
9953 | for (id = super->devlist; id; id = id->next) { | |
9954 | if (id->index == (unsigned)u->subdev) { | |
9955 | size = sizeof_imsm_dev(id->dev, 1); | |
9956 | if (u->new_raid_disks > u->old_raid_disks) | |
9957 | size += sizeof(__u32)*2* | |
9958 | (u->new_raid_disks - u->old_raid_disks); | |
503975b9 | 9959 | s = xmalloc(size); |
bc0b9d34 PC |
9960 | *space_tail = s; |
9961 | space_tail = s; | |
9962 | *space_tail = NULL; | |
9963 | break; | |
9964 | } | |
9965 | } | |
9966 | if (update->space_list == NULL) | |
9967 | break; | |
9968 | ||
9969 | /* add space for disk in update | |
9970 | */ | |
9971 | size = sizeof(struct dl); | |
503975b9 | 9972 | s = xmalloc(size); |
bc0b9d34 PC |
9973 | *space_tail = s; |
9974 | space_tail = s; | |
9975 | *space_tail = NULL; | |
9976 | ||
9977 | /* add spare device to update | |
9978 | */ | |
9979 | for (id = super->devlist ; id; id = id->next) | |
9980 | if (id->index == (unsigned)u->subdev) { | |
9981 | struct imsm_dev *dev; | |
9982 | struct imsm_map *map; | |
9983 | ||
9984 | dev = get_imsm_dev(super, u->subdev); | |
238c0a71 | 9985 | map = get_imsm_map(dev, MAP_0); |
bc0b9d34 PC |
9986 | current_level = map->raid_level; |
9987 | break; | |
9988 | } | |
089f9d79 | 9989 | if (u->new_level == 5 && u->new_level != current_level) { |
bc0b9d34 PC |
9990 | struct mdinfo *spares; |
9991 | ||
9992 | spares = get_spares_for_grow(st); | |
9993 | if (spares) { | |
9994 | struct dl *dl; | |
9995 | struct mdinfo *dev; | |
9996 | ||
9997 | dev = spares->devs; | |
9998 | if (dev) { | |
9999 | u->new_disks[0] = | |
10000 | makedev(dev->disk.major, | |
10001 | dev->disk.minor); | |
10002 | dl = get_disk_super(super, | |
10003 | dev->disk.major, | |
10004 | dev->disk.minor); | |
10005 | dl->index = u->old_raid_disks; | |
10006 | dev = dev->next; | |
10007 | } | |
10008 | sysfs_free(spares); | |
10009 | } | |
10010 | } | |
10011 | len = disks_to_mpb_size(u->new_raid_disks); | |
10012 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
48c5303a PC |
10013 | break; |
10014 | } | |
f3871fdc | 10015 | case update_size_change: { |
095b8088 N |
10016 | if (update->len < (int)sizeof(struct imsm_update_size_change)) |
10017 | return 0; | |
10018 | break; | |
10019 | } | |
10020 | case update_activate_spare: { | |
10021 | if (update->len < (int)sizeof(struct imsm_update_activate_spare)) | |
10022 | return 0; | |
f3871fdc AK |
10023 | break; |
10024 | } | |
949c47a0 DW |
10025 | case update_create_array: { |
10026 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 10027 | struct intel_dev *dv; |
54c2c1ea | 10028 | struct imsm_dev *dev = &u->dev; |
238c0a71 | 10029 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
54c2c1ea DW |
10030 | struct dl *dl; |
10031 | struct disk_info *inf; | |
10032 | int i; | |
10033 | int activate = 0; | |
949c47a0 | 10034 | |
095b8088 N |
10035 | if (update->len < (int)sizeof(*u)) |
10036 | return 0; | |
10037 | ||
54c2c1ea DW |
10038 | inf = get_disk_info(u); |
10039 | len = sizeof_imsm_dev(dev, 1); | |
ba2de7ba | 10040 | /* allocate a new super->devlist entry */ |
503975b9 N |
10041 | dv = xmalloc(sizeof(*dv)); |
10042 | dv->dev = xmalloc(len); | |
10043 | update->space = dv; | |
949c47a0 | 10044 | |
54c2c1ea DW |
10045 | /* count how many spares will be converted to members */ |
10046 | for (i = 0; i < map->num_members; i++) { | |
10047 | dl = serial_to_dl(inf[i].serial, super); | |
10048 | if (!dl) { | |
10049 | /* hmm maybe it failed?, nothing we can do about | |
10050 | * it here | |
10051 | */ | |
10052 | continue; | |
10053 | } | |
10054 | if (count_memberships(dl, super) == 0) | |
10055 | activate++; | |
10056 | } | |
10057 | len += activate * sizeof(struct imsm_disk); | |
949c47a0 | 10058 | break; |
095b8088 N |
10059 | } |
10060 | case update_kill_array: { | |
10061 | if (update->len < (int)sizeof(struct imsm_update_kill_array)) | |
10062 | return 0; | |
949c47a0 DW |
10063 | break; |
10064 | } | |
095b8088 N |
10065 | case update_rename_array: { |
10066 | if (update->len < (int)sizeof(struct imsm_update_rename_array)) | |
10067 | return 0; | |
10068 | break; | |
10069 | } | |
10070 | case update_add_remove_disk: | |
10071 | /* no update->len needed */ | |
10072 | break; | |
bbab0940 TM |
10073 | case update_prealloc_badblocks_mem: |
10074 | super->extra_space += sizeof(struct bbm_log) - | |
10075 | get_imsm_bbm_log_size(super->bbm_log); | |
10076 | break; | |
e6e9dd3f AP |
10077 | case update_rwh_policy: { |
10078 | if (update->len < (int)sizeof(struct imsm_update_rwh_policy)) | |
10079 | return 0; | |
10080 | break; | |
10081 | } | |
095b8088 N |
10082 | default: |
10083 | return 0; | |
949c47a0 | 10084 | } |
8273f55e | 10085 | |
4d7b1503 DW |
10086 | /* check if we need a larger metadata buffer */ |
10087 | if (super->next_buf) | |
10088 | buf_len = super->next_len; | |
10089 | else | |
10090 | buf_len = super->len; | |
10091 | ||
bbab0940 | 10092 | if (__le32_to_cpu(mpb->mpb_size) + super->extra_space + len > buf_len) { |
4d7b1503 DW |
10093 | /* ok we need a larger buf than what is currently allocated |
10094 | * if this allocation fails process_update will notice that | |
10095 | * ->next_len is set and ->next_buf is NULL | |
10096 | */ | |
bbab0940 TM |
10097 | buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + |
10098 | super->extra_space + len, sector_size); | |
4d7b1503 DW |
10099 | if (super->next_buf) |
10100 | free(super->next_buf); | |
10101 | ||
10102 | super->next_len = buf_len; | |
f36a9ecd | 10103 | if (posix_memalign(&super->next_buf, sector_size, buf_len) == 0) |
1f45a8ad DW |
10104 | memset(super->next_buf, 0, buf_len); |
10105 | else | |
4d7b1503 DW |
10106 | super->next_buf = NULL; |
10107 | } | |
5fe6f031 | 10108 | return 1; |
8273f55e DW |
10109 | } |
10110 | ||
ae6aad82 | 10111 | /* must be called while manager is quiesced */ |
f21e18ca | 10112 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index) |
ae6aad82 DW |
10113 | { |
10114 | struct imsm_super *mpb = super->anchor; | |
ae6aad82 DW |
10115 | struct dl *iter; |
10116 | struct imsm_dev *dev; | |
10117 | struct imsm_map *map; | |
4c9e8c1e | 10118 | unsigned int i, j, num_members; |
fb12a745 | 10119 | __u32 ord, ord_map0; |
4c9e8c1e | 10120 | struct bbm_log *log = super->bbm_log; |
ae6aad82 | 10121 | |
1ade5cc1 | 10122 | dprintf("deleting device[%d] from imsm_super\n", index); |
ae6aad82 DW |
10123 | |
10124 | /* shift all indexes down one */ | |
10125 | for (iter = super->disks; iter; iter = iter->next) | |
f21e18ca | 10126 | if (iter->index > (int)index) |
ae6aad82 | 10127 | iter->index--; |
47ee5a45 | 10128 | for (iter = super->missing; iter; iter = iter->next) |
f21e18ca | 10129 | if (iter->index > (int)index) |
47ee5a45 | 10130 | iter->index--; |
ae6aad82 DW |
10131 | |
10132 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
10133 | dev = get_imsm_dev(super, i); | |
238c0a71 | 10134 | map = get_imsm_map(dev, MAP_0); |
24565c9a DW |
10135 | num_members = map->num_members; |
10136 | for (j = 0; j < num_members; j++) { | |
10137 | /* update ord entries being careful not to propagate | |
10138 | * ord-flags to the first map | |
10139 | */ | |
238c0a71 | 10140 | ord = get_imsm_ord_tbl_ent(dev, j, MAP_X); |
fb12a745 | 10141 | ord_map0 = get_imsm_ord_tbl_ent(dev, j, MAP_0); |
ae6aad82 | 10142 | |
24565c9a DW |
10143 | if (ord_to_idx(ord) <= index) |
10144 | continue; | |
ae6aad82 | 10145 | |
238c0a71 | 10146 | map = get_imsm_map(dev, MAP_0); |
fb12a745 | 10147 | set_imsm_ord_tbl_ent(map, j, ord_map0 - 1); |
238c0a71 | 10148 | map = get_imsm_map(dev, MAP_1); |
24565c9a DW |
10149 | if (map) |
10150 | set_imsm_ord_tbl_ent(map, j, ord - 1); | |
ae6aad82 DW |
10151 | } |
10152 | } | |
10153 | ||
4c9e8c1e TM |
10154 | for (i = 0; i < log->entry_count; i++) { |
10155 | struct bbm_log_entry *entry = &log->marked_block_entries[i]; | |
10156 | ||
10157 | if (entry->disk_ordinal <= index) | |
10158 | continue; | |
10159 | entry->disk_ordinal--; | |
10160 | } | |
10161 | ||
ae6aad82 DW |
10162 | mpb->num_disks--; |
10163 | super->updates_pending++; | |
24565c9a DW |
10164 | if (*dlp) { |
10165 | struct dl *dl = *dlp; | |
10166 | ||
10167 | *dlp = (*dlp)->next; | |
10168 | __free_imsm_disk(dl); | |
10169 | } | |
ae6aad82 | 10170 | } |
9a717282 AK |
10171 | |
10172 | static void close_targets(int *targets, int new_disks) | |
10173 | { | |
10174 | int i; | |
10175 | ||
10176 | if (!targets) | |
10177 | return; | |
10178 | ||
10179 | for (i = 0; i < new_disks; i++) { | |
10180 | if (targets[i] >= 0) { | |
10181 | close(targets[i]); | |
10182 | targets[i] = -1; | |
10183 | } | |
10184 | } | |
10185 | } | |
10186 | ||
10187 | static int imsm_get_allowed_degradation(int level, int raid_disks, | |
10188 | struct intel_super *super, | |
10189 | struct imsm_dev *dev) | |
10190 | { | |
10191 | switch (level) { | |
bf5cf7c7 | 10192 | case 1: |
9a717282 AK |
10193 | case 10:{ |
10194 | int ret_val = 0; | |
10195 | struct imsm_map *map; | |
10196 | int i; | |
10197 | ||
10198 | ret_val = raid_disks/2; | |
10199 | /* check map if all disks pairs not failed | |
10200 | * in both maps | |
10201 | */ | |
238c0a71 | 10202 | map = get_imsm_map(dev, MAP_0); |
9a717282 AK |
10203 | for (i = 0; i < ret_val; i++) { |
10204 | int degradation = 0; | |
10205 | if (get_imsm_disk(super, i) == NULL) | |
10206 | degradation++; | |
10207 | if (get_imsm_disk(super, i + 1) == NULL) | |
10208 | degradation++; | |
10209 | if (degradation == 2) | |
10210 | return 0; | |
10211 | } | |
238c0a71 | 10212 | map = get_imsm_map(dev, MAP_1); |
9a717282 AK |
10213 | /* if there is no second map |
10214 | * result can be returned | |
10215 | */ | |
10216 | if (map == NULL) | |
10217 | return ret_val; | |
10218 | /* check degradation in second map | |
10219 | */ | |
10220 | for (i = 0; i < ret_val; i++) { | |
10221 | int degradation = 0; | |
10222 | if (get_imsm_disk(super, i) == NULL) | |
10223 | degradation++; | |
10224 | if (get_imsm_disk(super, i + 1) == NULL) | |
10225 | degradation++; | |
10226 | if (degradation == 2) | |
10227 | return 0; | |
10228 | } | |
10229 | return ret_val; | |
10230 | } | |
10231 | case 5: | |
10232 | return 1; | |
10233 | case 6: | |
10234 | return 2; | |
10235 | default: | |
10236 | return 0; | |
10237 | } | |
10238 | } | |
10239 | ||
687629c2 AK |
10240 | /******************************************************************************* |
10241 | * Function: open_backup_targets | |
10242 | * Description: Function opens file descriptors for all devices given in | |
10243 | * info->devs | |
10244 | * Parameters: | |
10245 | * info : general array info | |
10246 | * raid_disks : number of disks | |
10247 | * raid_fds : table of device's file descriptors | |
9a717282 AK |
10248 | * super : intel super for raid10 degradation check |
10249 | * dev : intel device for raid10 degradation check | |
687629c2 AK |
10250 | * Returns: |
10251 | * 0 : success | |
10252 | * -1 : fail | |
10253 | ******************************************************************************/ | |
9a717282 AK |
10254 | int open_backup_targets(struct mdinfo *info, int raid_disks, int *raid_fds, |
10255 | struct intel_super *super, struct imsm_dev *dev) | |
687629c2 AK |
10256 | { |
10257 | struct mdinfo *sd; | |
f627f5ad | 10258 | int i; |
9a717282 | 10259 | int opened = 0; |
f627f5ad AK |
10260 | |
10261 | for (i = 0; i < raid_disks; i++) | |
10262 | raid_fds[i] = -1; | |
687629c2 AK |
10263 | |
10264 | for (sd = info->devs ; sd ; sd = sd->next) { | |
10265 | char *dn; | |
10266 | ||
10267 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) { | |
10268 | dprintf("disk is faulty!!\n"); | |
10269 | continue; | |
10270 | } | |
10271 | ||
089f9d79 | 10272 | if (sd->disk.raid_disk >= raid_disks || sd->disk.raid_disk < 0) |
687629c2 AK |
10273 | continue; |
10274 | ||
10275 | dn = map_dev(sd->disk.major, | |
10276 | sd->disk.minor, 1); | |
10277 | raid_fds[sd->disk.raid_disk] = dev_open(dn, O_RDWR); | |
10278 | if (raid_fds[sd->disk.raid_disk] < 0) { | |
e12b3daa | 10279 | pr_err("cannot open component\n"); |
9a717282 | 10280 | continue; |
687629c2 | 10281 | } |
9a717282 AK |
10282 | opened++; |
10283 | } | |
10284 | /* check if maximum array degradation level is not exceeded | |
10285 | */ | |
10286 | if ((raid_disks - opened) > | |
089f9d79 JS |
10287 | imsm_get_allowed_degradation(info->new_level, raid_disks, |
10288 | super, dev)) { | |
e12b3daa | 10289 | pr_err("Not enough disks can be opened.\n"); |
9a717282 AK |
10290 | close_targets(raid_fds, raid_disks); |
10291 | return -2; | |
687629c2 AK |
10292 | } |
10293 | return 0; | |
10294 | } | |
10295 | ||
d31ad643 PB |
10296 | /******************************************************************************* |
10297 | * Function: validate_container_imsm | |
10298 | * Description: This routine validates container after assemble, | |
10299 | * eg. if devices in container are under the same controller. | |
10300 | * | |
10301 | * Parameters: | |
10302 | * info : linked list with info about devices used in array | |
10303 | * Returns: | |
10304 | * 1 : HBA mismatch | |
10305 | * 0 : Success | |
10306 | ******************************************************************************/ | |
10307 | int validate_container_imsm(struct mdinfo *info) | |
10308 | { | |
6b781d33 AP |
10309 | if (check_env("IMSM_NO_PLATFORM")) |
10310 | return 0; | |
d31ad643 | 10311 | |
6b781d33 AP |
10312 | struct sys_dev *idev; |
10313 | struct sys_dev *hba = NULL; | |
10314 | struct sys_dev *intel_devices = find_intel_devices(); | |
10315 | char *dev_path = devt_to_devpath(makedev(info->disk.major, | |
10316 | info->disk.minor)); | |
10317 | ||
10318 | for (idev = intel_devices; idev; idev = idev->next) { | |
10319 | if (dev_path && strstr(dev_path, idev->path)) { | |
10320 | hba = idev; | |
10321 | break; | |
d31ad643 | 10322 | } |
6b781d33 AP |
10323 | } |
10324 | if (dev_path) | |
d31ad643 PB |
10325 | free(dev_path); |
10326 | ||
6b781d33 AP |
10327 | if (!hba) { |
10328 | pr_err("WARNING - Cannot detect HBA for device %s!\n", | |
10329 | devid2kname(makedev(info->disk.major, info->disk.minor))); | |
10330 | return 1; | |
10331 | } | |
10332 | ||
10333 | const struct imsm_orom *orom = get_orom_by_device_id(hba->dev_id); | |
10334 | struct mdinfo *dev; | |
10335 | ||
10336 | for (dev = info->next; dev; dev = dev->next) { | |
10337 | dev_path = devt_to_devpath(makedev(dev->disk.major, dev->disk.minor)); | |
10338 | ||
10339 | struct sys_dev *hba2 = NULL; | |
10340 | for (idev = intel_devices; idev; idev = idev->next) { | |
10341 | if (dev_path && strstr(dev_path, idev->path)) { | |
10342 | hba2 = idev; | |
10343 | break; | |
d31ad643 PB |
10344 | } |
10345 | } | |
6b781d33 AP |
10346 | if (dev_path) |
10347 | free(dev_path); | |
10348 | ||
10349 | const struct imsm_orom *orom2 = hba2 == NULL ? NULL : | |
10350 | get_orom_by_device_id(hba2->dev_id); | |
10351 | ||
10352 | if (hba2 && hba->type != hba2->type) { | |
10353 | pr_err("WARNING - HBAs of devices do not match %s != %s\n", | |
10354 | get_sys_dev_type(hba->type), get_sys_dev_type(hba2->type)); | |
10355 | return 1; | |
10356 | } | |
10357 | ||
07cb1e57 | 10358 | if (orom != orom2) { |
6b781d33 AP |
10359 | pr_err("WARNING - IMSM container assembled with disks under different HBAs!\n" |
10360 | " This operation is not supported and can lead to data loss.\n"); | |
10361 | return 1; | |
10362 | } | |
10363 | ||
10364 | if (!orom) { | |
10365 | pr_err("WARNING - IMSM container assembled with disks under HBAs without IMSM platform support!\n" | |
10366 | " This operation is not supported and can lead to data loss.\n"); | |
10367 | return 1; | |
10368 | } | |
d31ad643 | 10369 | } |
6b781d33 | 10370 | |
d31ad643 PB |
10371 | return 0; |
10372 | } | |
32141c17 | 10373 | |
6f50473f TM |
10374 | /******************************************************************************* |
10375 | * Function: imsm_record_badblock | |
10376 | * Description: This routine stores new bad block record in BBM log | |
10377 | * | |
10378 | * Parameters: | |
10379 | * a : array containing a bad block | |
10380 | * slot : disk number containing a bad block | |
10381 | * sector : bad block sector | |
10382 | * length : bad block sectors range | |
10383 | * Returns: | |
10384 | * 1 : Success | |
10385 | * 0 : Error | |
10386 | ******************************************************************************/ | |
10387 | static int imsm_record_badblock(struct active_array *a, int slot, | |
10388 | unsigned long long sector, int length) | |
10389 | { | |
10390 | struct intel_super *super = a->container->sb; | |
10391 | int ord; | |
10392 | int ret; | |
10393 | ||
10394 | ord = imsm_disk_slot_to_ord(a, slot); | |
10395 | if (ord < 0) | |
10396 | return 0; | |
10397 | ||
10398 | ret = record_new_badblock(super->bbm_log, ord_to_idx(ord), sector, | |
10399 | length); | |
10400 | if (ret) | |
10401 | super->updates_pending++; | |
10402 | ||
10403 | return ret; | |
10404 | } | |
c07a5a4f TM |
10405 | /******************************************************************************* |
10406 | * Function: imsm_clear_badblock | |
10407 | * Description: This routine clears bad block record from BBM log | |
10408 | * | |
10409 | * Parameters: | |
10410 | * a : array containing a bad block | |
10411 | * slot : disk number containing a bad block | |
10412 | * sector : bad block sector | |
10413 | * length : bad block sectors range | |
10414 | * Returns: | |
10415 | * 1 : Success | |
10416 | * 0 : Error | |
10417 | ******************************************************************************/ | |
10418 | static int imsm_clear_badblock(struct active_array *a, int slot, | |
10419 | unsigned long long sector, int length) | |
10420 | { | |
10421 | struct intel_super *super = a->container->sb; | |
10422 | int ord; | |
10423 | int ret; | |
10424 | ||
10425 | ord = imsm_disk_slot_to_ord(a, slot); | |
10426 | if (ord < 0) | |
10427 | return 0; | |
10428 | ||
10429 | ret = clear_badblock(super->bbm_log, ord_to_idx(ord), sector, length); | |
10430 | if (ret) | |
10431 | super->updates_pending++; | |
10432 | ||
10433 | return ret; | |
10434 | } | |
928f1424 TM |
10435 | /******************************************************************************* |
10436 | * Function: imsm_get_badblocks | |
10437 | * Description: This routine get list of bad blocks for an array | |
10438 | * | |
10439 | * Parameters: | |
10440 | * a : array | |
10441 | * slot : disk number | |
10442 | * Returns: | |
10443 | * bb : structure containing bad blocks | |
10444 | * NULL : error | |
10445 | ******************************************************************************/ | |
10446 | static struct md_bb *imsm_get_badblocks(struct active_array *a, int slot) | |
10447 | { | |
10448 | int inst = a->info.container_member; | |
10449 | struct intel_super *super = a->container->sb; | |
10450 | struct imsm_dev *dev = get_imsm_dev(super, inst); | |
10451 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
10452 | int ord; | |
10453 | ||
10454 | ord = imsm_disk_slot_to_ord(a, slot); | |
10455 | if (ord < 0) | |
10456 | return NULL; | |
10457 | ||
10458 | get_volume_badblocks(super->bbm_log, ord_to_idx(ord), pba_of_lba0(map), | |
10459 | blocks_per_member(map), &super->bb); | |
10460 | ||
10461 | return &super->bb; | |
10462 | } | |
27156a57 TM |
10463 | /******************************************************************************* |
10464 | * Function: examine_badblocks_imsm | |
10465 | * Description: Prints list of bad blocks on a disk to the standard output | |
10466 | * | |
10467 | * Parameters: | |
10468 | * st : metadata handler | |
10469 | * fd : open file descriptor for device | |
10470 | * devname : device name | |
10471 | * Returns: | |
10472 | * 0 : Success | |
10473 | * 1 : Error | |
10474 | ******************************************************************************/ | |
10475 | static int examine_badblocks_imsm(struct supertype *st, int fd, char *devname) | |
10476 | { | |
10477 | struct intel_super *super = st->sb; | |
10478 | struct bbm_log *log = super->bbm_log; | |
10479 | struct dl *d = NULL; | |
10480 | int any = 0; | |
10481 | ||
10482 | for (d = super->disks; d ; d = d->next) { | |
10483 | if (strcmp(d->devname, devname) == 0) | |
10484 | break; | |
10485 | } | |
10486 | ||
10487 | if ((d == NULL) || (d->index < 0)) { /* serial mismatch probably */ | |
10488 | pr_err("%s doesn't appear to be part of a raid array\n", | |
10489 | devname); | |
10490 | return 1; | |
10491 | } | |
10492 | ||
10493 | if (log != NULL) { | |
10494 | unsigned int i; | |
10495 | struct bbm_log_entry *entry = &log->marked_block_entries[0]; | |
10496 | ||
10497 | for (i = 0; i < log->entry_count; i++) { | |
10498 | if (entry[i].disk_ordinal == d->index) { | |
10499 | unsigned long long sector = __le48_to_cpu( | |
10500 | &entry[i].defective_block_start); | |
10501 | int cnt = entry[i].marked_count + 1; | |
10502 | ||
10503 | if (!any) { | |
10504 | printf("Bad-blocks on %s:\n", devname); | |
10505 | any = 1; | |
10506 | } | |
10507 | ||
10508 | printf("%20llu for %d sectors\n", sector, cnt); | |
10509 | } | |
10510 | } | |
10511 | } | |
10512 | ||
10513 | if (!any) | |
10514 | printf("No bad-blocks list configured on %s\n", devname); | |
10515 | ||
10516 | return 0; | |
10517 | } | |
687629c2 AK |
10518 | /******************************************************************************* |
10519 | * Function: init_migr_record_imsm | |
10520 | * Description: Function inits imsm migration record | |
10521 | * Parameters: | |
10522 | * super : imsm internal array info | |
10523 | * dev : device under migration | |
10524 | * info : general array info to find the smallest device | |
10525 | * Returns: | |
10526 | * none | |
10527 | ******************************************************************************/ | |
10528 | void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev, | |
10529 | struct mdinfo *info) | |
10530 | { | |
10531 | struct intel_super *super = st->sb; | |
10532 | struct migr_record *migr_rec = super->migr_rec; | |
10533 | int new_data_disks; | |
10534 | unsigned long long dsize, dev_sectors; | |
10535 | long long unsigned min_dev_sectors = -1LLU; | |
10536 | struct mdinfo *sd; | |
10537 | char nm[30]; | |
10538 | int fd; | |
238c0a71 AK |
10539 | struct imsm_map *map_dest = get_imsm_map(dev, MAP_0); |
10540 | struct imsm_map *map_src = get_imsm_map(dev, MAP_1); | |
687629c2 | 10541 | unsigned long long num_migr_units; |
3ef4403c | 10542 | unsigned long long array_blocks; |
687629c2 AK |
10543 | |
10544 | memset(migr_rec, 0, sizeof(struct migr_record)); | |
10545 | migr_rec->family_num = __cpu_to_le32(super->anchor->family_num); | |
10546 | ||
10547 | /* only ascending reshape supported now */ | |
10548 | migr_rec->ascending_migr = __cpu_to_le32(1); | |
10549 | ||
10550 | migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE / | |
10551 | max(map_dest->blocks_per_strip, map_src->blocks_per_strip); | |
e1742195 AK |
10552 | migr_rec->dest_depth_per_unit *= |
10553 | max(map_dest->blocks_per_strip, map_src->blocks_per_strip); | |
238c0a71 | 10554 | new_data_disks = imsm_num_data_members(dev, MAP_0); |
687629c2 AK |
10555 | migr_rec->blocks_per_unit = |
10556 | __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks); | |
10557 | migr_rec->dest_depth_per_unit = | |
10558 | __cpu_to_le32(migr_rec->dest_depth_per_unit); | |
3ef4403c | 10559 | array_blocks = info->component_size * new_data_disks; |
687629c2 AK |
10560 | num_migr_units = |
10561 | array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit); | |
10562 | ||
10563 | if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit)) | |
10564 | num_migr_units++; | |
10565 | migr_rec->num_migr_units = __cpu_to_le32(num_migr_units); | |
10566 | ||
10567 | migr_rec->post_migr_vol_cap = dev->size_low; | |
10568 | migr_rec->post_migr_vol_cap_hi = dev->size_high; | |
10569 | ||
687629c2 AK |
10570 | /* Find the smallest dev */ |
10571 | for (sd = info->devs ; sd ; sd = sd->next) { | |
10572 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
10573 | fd = dev_open(nm, O_RDONLY); | |
10574 | if (fd < 0) | |
10575 | continue; | |
10576 | get_dev_size(fd, NULL, &dsize); | |
10577 | dev_sectors = dsize / 512; | |
10578 | if (dev_sectors < min_dev_sectors) | |
10579 | min_dev_sectors = dev_sectors; | |
10580 | close(fd); | |
10581 | } | |
10582 | migr_rec->ckpt_area_pba = __cpu_to_le32(min_dev_sectors - | |
10583 | RAID_DISK_RESERVED_BLOCKS_IMSM_HI); | |
10584 | ||
10585 | write_imsm_migr_rec(st); | |
10586 | ||
10587 | return; | |
10588 | } | |
10589 | ||
10590 | /******************************************************************************* | |
10591 | * Function: save_backup_imsm | |
10592 | * Description: Function saves critical data stripes to Migration Copy Area | |
10593 | * and updates the current migration unit status. | |
10594 | * Use restore_stripes() to form a destination stripe, | |
10595 | * and to write it to the Copy Area. | |
10596 | * Parameters: | |
10597 | * st : supertype information | |
aea93171 | 10598 | * dev : imsm device that backup is saved for |
687629c2 AK |
10599 | * info : general array info |
10600 | * buf : input buffer | |
687629c2 AK |
10601 | * length : length of data to backup (blocks_per_unit) |
10602 | * Returns: | |
10603 | * 0 : success | |
10604 | *, -1 : fail | |
10605 | ******************************************************************************/ | |
10606 | int save_backup_imsm(struct supertype *st, | |
10607 | struct imsm_dev *dev, | |
10608 | struct mdinfo *info, | |
10609 | void *buf, | |
687629c2 AK |
10610 | int length) |
10611 | { | |
10612 | int rv = -1; | |
10613 | struct intel_super *super = st->sb; | |
594dc1b8 JS |
10614 | unsigned long long *target_offsets; |
10615 | int *targets; | |
687629c2 | 10616 | int i; |
238c0a71 | 10617 | struct imsm_map *map_dest = get_imsm_map(dev, MAP_0); |
687629c2 | 10618 | int new_disks = map_dest->num_members; |
ab724b98 AK |
10619 | int dest_layout = 0; |
10620 | int dest_chunk; | |
d1877f69 | 10621 | unsigned long long start; |
238c0a71 | 10622 | int data_disks = imsm_num_data_members(dev, MAP_0); |
687629c2 | 10623 | |
503975b9 | 10624 | targets = xmalloc(new_disks * sizeof(int)); |
687629c2 | 10625 | |
7e45b550 AK |
10626 | for (i = 0; i < new_disks; i++) |
10627 | targets[i] = -1; | |
10628 | ||
503975b9 | 10629 | target_offsets = xcalloc(new_disks, sizeof(unsigned long long)); |
687629c2 | 10630 | |
d1877f69 | 10631 | start = info->reshape_progress * 512; |
687629c2 | 10632 | for (i = 0; i < new_disks; i++) { |
687629c2 AK |
10633 | target_offsets[i] = (unsigned long long) |
10634 | __le32_to_cpu(super->migr_rec->ckpt_area_pba) * 512; | |
d1877f69 AK |
10635 | /* move back copy area adderss, it will be moved forward |
10636 | * in restore_stripes() using start input variable | |
10637 | */ | |
10638 | target_offsets[i] -= start/data_disks; | |
687629c2 AK |
10639 | } |
10640 | ||
9a717282 AK |
10641 | if (open_backup_targets(info, new_disks, targets, |
10642 | super, dev)) | |
687629c2 AK |
10643 | goto abort; |
10644 | ||
68eb8bc6 | 10645 | dest_layout = imsm_level_to_layout(map_dest->raid_level); |
ab724b98 AK |
10646 | dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512; |
10647 | ||
687629c2 AK |
10648 | if (restore_stripes(targets, /* list of dest devices */ |
10649 | target_offsets, /* migration record offsets */ | |
10650 | new_disks, | |
ab724b98 AK |
10651 | dest_chunk, |
10652 | map_dest->raid_level, | |
10653 | dest_layout, | |
10654 | -1, /* source backup file descriptor */ | |
10655 | 0, /* input buf offset | |
10656 | * always 0 buf is already offseted */ | |
d1877f69 | 10657 | start, |
687629c2 AK |
10658 | length, |
10659 | buf) != 0) { | |
e7b84f9d | 10660 | pr_err("Error restoring stripes\n"); |
687629c2 AK |
10661 | goto abort; |
10662 | } | |
10663 | ||
10664 | rv = 0; | |
10665 | ||
10666 | abort: | |
10667 | if (targets) { | |
9a717282 | 10668 | close_targets(targets, new_disks); |
687629c2 AK |
10669 | free(targets); |
10670 | } | |
10671 | free(target_offsets); | |
10672 | ||
10673 | return rv; | |
10674 | } | |
10675 | ||
10676 | /******************************************************************************* | |
10677 | * Function: save_checkpoint_imsm | |
10678 | * Description: Function called for current unit status update | |
10679 | * in the migration record. It writes it to disk. | |
10680 | * Parameters: | |
10681 | * super : imsm internal array info | |
10682 | * info : general array info | |
10683 | * Returns: | |
10684 | * 0: success | |
10685 | * 1: failure | |
0228d92c AK |
10686 | * 2: failure, means no valid migration record |
10687 | * / no general migration in progress / | |
687629c2 AK |
10688 | ******************************************************************************/ |
10689 | int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state) | |
10690 | { | |
10691 | struct intel_super *super = st->sb; | |
f8b72ef5 AK |
10692 | unsigned long long blocks_per_unit; |
10693 | unsigned long long curr_migr_unit; | |
10694 | ||
2e062e82 | 10695 | if (load_imsm_migr_rec(super, info) != 0) { |
7a862a02 | 10696 | dprintf("imsm: ERROR: Cannot read migration record for checkpoint save.\n"); |
2e062e82 AK |
10697 | return 1; |
10698 | } | |
10699 | ||
f8b72ef5 AK |
10700 | blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit); |
10701 | if (blocks_per_unit == 0) { | |
0228d92c AK |
10702 | dprintf("imsm: no migration in progress.\n"); |
10703 | return 2; | |
687629c2 | 10704 | } |
f8b72ef5 AK |
10705 | curr_migr_unit = info->reshape_progress / blocks_per_unit; |
10706 | /* check if array is alligned to copy area | |
10707 | * if it is not alligned, add one to current migration unit value | |
10708 | * this can happend on array reshape finish only | |
10709 | */ | |
10710 | if (info->reshape_progress % blocks_per_unit) | |
10711 | curr_migr_unit++; | |
687629c2 AK |
10712 | |
10713 | super->migr_rec->curr_migr_unit = | |
f8b72ef5 | 10714 | __cpu_to_le32(curr_migr_unit); |
687629c2 AK |
10715 | super->migr_rec->rec_status = __cpu_to_le32(state); |
10716 | super->migr_rec->dest_1st_member_lba = | |
f8b72ef5 AK |
10717 | __cpu_to_le32(curr_migr_unit * |
10718 | __le32_to_cpu(super->migr_rec->dest_depth_per_unit)); | |
687629c2 | 10719 | if (write_imsm_migr_rec(st) < 0) { |
7a862a02 | 10720 | dprintf("imsm: Cannot write migration record outside backup area\n"); |
687629c2 AK |
10721 | return 1; |
10722 | } | |
10723 | ||
10724 | return 0; | |
10725 | } | |
10726 | ||
276d77db AK |
10727 | /******************************************************************************* |
10728 | * Function: recover_backup_imsm | |
10729 | * Description: Function recovers critical data from the Migration Copy Area | |
10730 | * while assembling an array. | |
10731 | * Parameters: | |
10732 | * super : imsm internal array info | |
10733 | * info : general array info | |
10734 | * Returns: | |
10735 | * 0 : success (or there is no data to recover) | |
10736 | * 1 : fail | |
10737 | ******************************************************************************/ | |
10738 | int recover_backup_imsm(struct supertype *st, struct mdinfo *info) | |
10739 | { | |
10740 | struct intel_super *super = st->sb; | |
10741 | struct migr_record *migr_rec = super->migr_rec; | |
594dc1b8 | 10742 | struct imsm_map *map_dest; |
276d77db AK |
10743 | struct intel_dev *id = NULL; |
10744 | unsigned long long read_offset; | |
10745 | unsigned long long write_offset; | |
10746 | unsigned unit_len; | |
10747 | int *targets = NULL; | |
10748 | int new_disks, i, err; | |
10749 | char *buf = NULL; | |
10750 | int retval = 1; | |
f36a9ecd | 10751 | unsigned int sector_size = super->sector_size; |
276d77db AK |
10752 | unsigned long curr_migr_unit = __le32_to_cpu(migr_rec->curr_migr_unit); |
10753 | unsigned long num_migr_units = __le32_to_cpu(migr_rec->num_migr_units); | |
276d77db | 10754 | char buffer[20]; |
6c3560c0 | 10755 | int skipped_disks = 0; |
276d77db AK |
10756 | |
10757 | err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20); | |
10758 | if (err < 1) | |
10759 | return 1; | |
10760 | ||
10761 | /* recover data only during assemblation */ | |
10762 | if (strncmp(buffer, "inactive", 8) != 0) | |
10763 | return 0; | |
10764 | /* no data to recover */ | |
10765 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
10766 | return 0; | |
10767 | if (curr_migr_unit >= num_migr_units) | |
10768 | return 1; | |
10769 | ||
10770 | /* find device during reshape */ | |
10771 | for (id = super->devlist; id; id = id->next) | |
10772 | if (is_gen_migration(id->dev)) | |
10773 | break; | |
10774 | if (id == NULL) | |
10775 | return 1; | |
10776 | ||
238c0a71 | 10777 | map_dest = get_imsm_map(id->dev, MAP_0); |
276d77db AK |
10778 | new_disks = map_dest->num_members; |
10779 | ||
10780 | read_offset = (unsigned long long) | |
10781 | __le32_to_cpu(migr_rec->ckpt_area_pba) * 512; | |
10782 | ||
10783 | write_offset = ((unsigned long long) | |
10784 | __le32_to_cpu(migr_rec->dest_1st_member_lba) + | |
5551b113 | 10785 | pba_of_lba0(map_dest)) * 512; |
276d77db AK |
10786 | |
10787 | unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
f36a9ecd | 10788 | if (posix_memalign((void **)&buf, sector_size, unit_len) != 0) |
276d77db | 10789 | goto abort; |
503975b9 | 10790 | targets = xcalloc(new_disks, sizeof(int)); |
276d77db | 10791 | |
9a717282 | 10792 | if (open_backup_targets(info, new_disks, targets, super, id->dev)) { |
e7b84f9d | 10793 | pr_err("Cannot open some devices belonging to array.\n"); |
f627f5ad AK |
10794 | goto abort; |
10795 | } | |
276d77db AK |
10796 | |
10797 | for (i = 0; i < new_disks; i++) { | |
6c3560c0 AK |
10798 | if (targets[i] < 0) { |
10799 | skipped_disks++; | |
10800 | continue; | |
10801 | } | |
276d77db | 10802 | if (lseek64(targets[i], read_offset, SEEK_SET) < 0) { |
e7b84f9d N |
10803 | pr_err("Cannot seek to block: %s\n", |
10804 | strerror(errno)); | |
137debce AK |
10805 | skipped_disks++; |
10806 | continue; | |
276d77db | 10807 | } |
9ec11d1a | 10808 | if ((unsigned)read(targets[i], buf, unit_len) != unit_len) { |
e7b84f9d N |
10809 | pr_err("Cannot read copy area block: %s\n", |
10810 | strerror(errno)); | |
137debce AK |
10811 | skipped_disks++; |
10812 | continue; | |
276d77db AK |
10813 | } |
10814 | if (lseek64(targets[i], write_offset, SEEK_SET) < 0) { | |
e7b84f9d N |
10815 | pr_err("Cannot seek to block: %s\n", |
10816 | strerror(errno)); | |
137debce AK |
10817 | skipped_disks++; |
10818 | continue; | |
276d77db | 10819 | } |
9ec11d1a | 10820 | if ((unsigned)write(targets[i], buf, unit_len) != unit_len) { |
e7b84f9d N |
10821 | pr_err("Cannot restore block: %s\n", |
10822 | strerror(errno)); | |
137debce AK |
10823 | skipped_disks++; |
10824 | continue; | |
276d77db AK |
10825 | } |
10826 | } | |
10827 | ||
137debce AK |
10828 | if (skipped_disks > imsm_get_allowed_degradation(info->new_level, |
10829 | new_disks, | |
10830 | super, | |
10831 | id->dev)) { | |
7a862a02 | 10832 | pr_err("Cannot restore data from backup. Too many failed disks\n"); |
6c3560c0 AK |
10833 | goto abort; |
10834 | } | |
10835 | ||
befb629b AK |
10836 | if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) { |
10837 | /* ignore error == 2, this can mean end of reshape here | |
10838 | */ | |
7a862a02 | 10839 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL) during restart\n"); |
befb629b | 10840 | } else |
276d77db | 10841 | retval = 0; |
276d77db AK |
10842 | |
10843 | abort: | |
10844 | if (targets) { | |
10845 | for (i = 0; i < new_disks; i++) | |
10846 | if (targets[i]) | |
10847 | close(targets[i]); | |
10848 | free(targets); | |
10849 | } | |
10850 | free(buf); | |
10851 | return retval; | |
10852 | } | |
10853 | ||
2cda7640 ML |
10854 | static char disk_by_path[] = "/dev/disk/by-path/"; |
10855 | ||
10856 | static const char *imsm_get_disk_controller_domain(const char *path) | |
10857 | { | |
2cda7640 | 10858 | char disk_path[PATH_MAX]; |
96234762 LM |
10859 | char *drv=NULL; |
10860 | struct stat st; | |
2cda7640 | 10861 | |
6d8d290a | 10862 | strcpy(disk_path, disk_by_path); |
96234762 LM |
10863 | strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1); |
10864 | if (stat(disk_path, &st) == 0) { | |
10865 | struct sys_dev* hba; | |
594dc1b8 | 10866 | char *path; |
96234762 LM |
10867 | |
10868 | path = devt_to_devpath(st.st_rdev); | |
10869 | if (path == NULL) | |
10870 | return "unknown"; | |
10871 | hba = find_disk_attached_hba(-1, path); | |
10872 | if (hba && hba->type == SYS_DEV_SAS) | |
10873 | drv = "isci"; | |
10874 | else if (hba && hba->type == SYS_DEV_SATA) | |
10875 | drv = "ahci"; | |
c6839718 MT |
10876 | else if (hba && hba->type == SYS_DEV_VMD) |
10877 | drv = "vmd"; | |
10878 | else if (hba && hba->type == SYS_DEV_NVME) | |
10879 | drv = "nvme"; | |
1011e834 | 10880 | else |
96234762 LM |
10881 | drv = "unknown"; |
10882 | dprintf("path: %s hba: %s attached: %s\n", | |
10883 | path, (hba) ? hba->path : "NULL", drv); | |
10884 | free(path); | |
2cda7640 | 10885 | } |
96234762 | 10886 | return drv; |
2cda7640 ML |
10887 | } |
10888 | ||
4dd2df09 | 10889 | static char *imsm_find_array_devnm_by_subdev(int subdev, char *container) |
78b10e66 | 10890 | { |
4dd2df09 | 10891 | static char devnm[32]; |
78b10e66 N |
10892 | char subdev_name[20]; |
10893 | struct mdstat_ent *mdstat; | |
10894 | ||
10895 | sprintf(subdev_name, "%d", subdev); | |
10896 | mdstat = mdstat_by_subdev(subdev_name, container); | |
10897 | if (!mdstat) | |
4dd2df09 | 10898 | return NULL; |
78b10e66 | 10899 | |
4dd2df09 | 10900 | strcpy(devnm, mdstat->devnm); |
78b10e66 | 10901 | free_mdstat(mdstat); |
4dd2df09 | 10902 | return devnm; |
78b10e66 N |
10903 | } |
10904 | ||
10905 | static int imsm_reshape_is_allowed_on_container(struct supertype *st, | |
10906 | struct geo_params *geo, | |
fbf3d202 AK |
10907 | int *old_raid_disks, |
10908 | int direction) | |
78b10e66 | 10909 | { |
694575e7 KW |
10910 | /* currently we only support increasing the number of devices |
10911 | * for a container. This increases the number of device for each | |
10912 | * member array. They must all be RAID0 or RAID5. | |
10913 | */ | |
78b10e66 N |
10914 | int ret_val = 0; |
10915 | struct mdinfo *info, *member; | |
10916 | int devices_that_can_grow = 0; | |
10917 | ||
7a862a02 | 10918 | dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): st->devnm = (%s)\n", st->devnm); |
78b10e66 | 10919 | |
d04f65f4 | 10920 | if (geo->size > 0 || |
78b10e66 N |
10921 | geo->level != UnSet || |
10922 | geo->layout != UnSet || | |
10923 | geo->chunksize != 0 || | |
10924 | geo->raid_disks == UnSet) { | |
7a862a02 | 10925 | dprintf("imsm: Container operation is allowed for raid disks number change only.\n"); |
78b10e66 N |
10926 | return ret_val; |
10927 | } | |
10928 | ||
fbf3d202 | 10929 | if (direction == ROLLBACK_METADATA_CHANGES) { |
7a862a02 | 10930 | dprintf("imsm: Metadata changes rollback is not supported for container operation.\n"); |
fbf3d202 AK |
10931 | return ret_val; |
10932 | } | |
10933 | ||
78b10e66 N |
10934 | info = container_content_imsm(st, NULL); |
10935 | for (member = info; member; member = member->next) { | |
4dd2df09 | 10936 | char *result; |
78b10e66 N |
10937 | |
10938 | dprintf("imsm: checking device_num: %i\n", | |
10939 | member->container_member); | |
10940 | ||
d7d205bd | 10941 | if (geo->raid_disks <= member->array.raid_disks) { |
78b10e66 N |
10942 | /* we work on container for Online Capacity Expansion |
10943 | * only so raid_disks has to grow | |
10944 | */ | |
7a862a02 | 10945 | dprintf("imsm: for container operation raid disks increase is required\n"); |
78b10e66 N |
10946 | break; |
10947 | } | |
10948 | ||
089f9d79 | 10949 | if (info->array.level != 0 && info->array.level != 5) { |
78b10e66 N |
10950 | /* we cannot use this container with other raid level |
10951 | */ | |
7a862a02 | 10952 | dprintf("imsm: for container operation wrong raid level (%i) detected\n", |
78b10e66 N |
10953 | info->array.level); |
10954 | break; | |
10955 | } else { | |
10956 | /* check for platform support | |
10957 | * for this raid level configuration | |
10958 | */ | |
10959 | struct intel_super *super = st->sb; | |
10960 | if (!is_raid_level_supported(super->orom, | |
10961 | member->array.level, | |
10962 | geo->raid_disks)) { | |
7a862a02 | 10963 | dprintf("platform does not support raid%d with %d disk%s\n", |
78b10e66 N |
10964 | info->array.level, |
10965 | geo->raid_disks, | |
10966 | geo->raid_disks > 1 ? "s" : ""); | |
10967 | break; | |
10968 | } | |
2a4a08e7 AK |
10969 | /* check if component size is aligned to chunk size |
10970 | */ | |
10971 | if (info->component_size % | |
10972 | (info->array.chunk_size/512)) { | |
7a862a02 | 10973 | dprintf("Component size is not aligned to chunk size\n"); |
2a4a08e7 AK |
10974 | break; |
10975 | } | |
78b10e66 N |
10976 | } |
10977 | ||
10978 | if (*old_raid_disks && | |
10979 | info->array.raid_disks != *old_raid_disks) | |
10980 | break; | |
10981 | *old_raid_disks = info->array.raid_disks; | |
10982 | ||
10983 | /* All raid5 and raid0 volumes in container | |
10984 | * have to be ready for Online Capacity Expansion | |
10985 | * so they need to be assembled. We have already | |
10986 | * checked that no recovery etc is happening. | |
10987 | */ | |
4dd2df09 N |
10988 | result = imsm_find_array_devnm_by_subdev(member->container_member, |
10989 | st->container_devnm); | |
10990 | if (result == NULL) { | |
78b10e66 N |
10991 | dprintf("imsm: cannot find array\n"); |
10992 | break; | |
10993 | } | |
10994 | devices_that_can_grow++; | |
10995 | } | |
10996 | sysfs_free(info); | |
10997 | if (!member && devices_that_can_grow) | |
10998 | ret_val = 1; | |
10999 | ||
11000 | if (ret_val) | |
1ade5cc1 | 11001 | dprintf("Container operation allowed\n"); |
78b10e66 | 11002 | else |
1ade5cc1 | 11003 | dprintf("Error: %i\n", ret_val); |
78b10e66 N |
11004 | |
11005 | return ret_val; | |
11006 | } | |
11007 | ||
11008 | /* Function: get_spares_for_grow | |
11009 | * Description: Allocates memory and creates list of spare devices | |
1011e834 | 11010 | * avaliable in container. Checks if spare drive size is acceptable. |
78b10e66 N |
11011 | * Parameters: Pointer to the supertype structure |
11012 | * Returns: Pointer to the list of spare devices (mdinfo structure) on success, | |
1011e834 | 11013 | * NULL if fail |
78b10e66 N |
11014 | */ |
11015 | static struct mdinfo *get_spares_for_grow(struct supertype *st) | |
11016 | { | |
fbfdcb06 AO |
11017 | struct spare_criteria sc; |
11018 | ||
11019 | get_spare_criteria_imsm(st, &sc); | |
11020 | return container_choose_spares(st, &sc, NULL, NULL, NULL, 0); | |
78b10e66 N |
11021 | } |
11022 | ||
11023 | /****************************************************************************** | |
11024 | * function: imsm_create_metadata_update_for_reshape | |
11025 | * Function creates update for whole IMSM container. | |
11026 | * | |
11027 | ******************************************************************************/ | |
11028 | static int imsm_create_metadata_update_for_reshape( | |
11029 | struct supertype *st, | |
11030 | struct geo_params *geo, | |
11031 | int old_raid_disks, | |
11032 | struct imsm_update_reshape **updatep) | |
11033 | { | |
11034 | struct intel_super *super = st->sb; | |
11035 | struct imsm_super *mpb = super->anchor; | |
594dc1b8 JS |
11036 | int update_memory_size; |
11037 | struct imsm_update_reshape *u; | |
11038 | struct mdinfo *spares; | |
78b10e66 | 11039 | int i; |
594dc1b8 | 11040 | int delta_disks; |
bbd24d86 | 11041 | struct mdinfo *dev; |
78b10e66 | 11042 | |
1ade5cc1 | 11043 | dprintf("(enter) raid_disks = %i\n", geo->raid_disks); |
78b10e66 N |
11044 | |
11045 | delta_disks = geo->raid_disks - old_raid_disks; | |
11046 | ||
11047 | /* size of all update data without anchor */ | |
11048 | update_memory_size = sizeof(struct imsm_update_reshape); | |
11049 | ||
11050 | /* now add space for spare disks that we need to add. */ | |
11051 | update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1); | |
11052 | ||
503975b9 | 11053 | u = xcalloc(1, update_memory_size); |
78b10e66 N |
11054 | u->type = update_reshape_container_disks; |
11055 | u->old_raid_disks = old_raid_disks; | |
11056 | u->new_raid_disks = geo->raid_disks; | |
11057 | ||
11058 | /* now get spare disks list | |
11059 | */ | |
11060 | spares = get_spares_for_grow(st); | |
11061 | ||
d7be7d87 | 11062 | if (spares == NULL || delta_disks > spares->array.spare_disks) { |
7a862a02 | 11063 | pr_err("imsm: ERROR: Cannot get spare devices for %s.\n", geo->dev_name); |
e4c72d1d | 11064 | i = -1; |
78b10e66 N |
11065 | goto abort; |
11066 | } | |
11067 | ||
11068 | /* we have got spares | |
11069 | * update disk list in imsm_disk list table in anchor | |
11070 | */ | |
11071 | dprintf("imsm: %i spares are available.\n\n", | |
11072 | spares->array.spare_disks); | |
11073 | ||
bbd24d86 | 11074 | dev = spares->devs; |
78b10e66 | 11075 | for (i = 0; i < delta_disks; i++) { |
78b10e66 N |
11076 | struct dl *dl; |
11077 | ||
bbd24d86 AK |
11078 | if (dev == NULL) |
11079 | break; | |
78b10e66 N |
11080 | u->new_disks[i] = makedev(dev->disk.major, |
11081 | dev->disk.minor); | |
11082 | dl = get_disk_super(super, dev->disk.major, dev->disk.minor); | |
ee4beede AK |
11083 | dl->index = mpb->num_disks; |
11084 | mpb->num_disks++; | |
bbd24d86 | 11085 | dev = dev->next; |
78b10e66 | 11086 | } |
78b10e66 N |
11087 | |
11088 | abort: | |
11089 | /* free spares | |
11090 | */ | |
11091 | sysfs_free(spares); | |
11092 | ||
d677e0b8 | 11093 | dprintf("imsm: reshape update preparation :"); |
78b10e66 | 11094 | if (i == delta_disks) { |
1ade5cc1 | 11095 | dprintf_cont(" OK\n"); |
78b10e66 N |
11096 | *updatep = u; |
11097 | return update_memory_size; | |
11098 | } | |
11099 | free(u); | |
1ade5cc1 | 11100 | dprintf_cont(" Error\n"); |
78b10e66 N |
11101 | |
11102 | return 0; | |
11103 | } | |
11104 | ||
f3871fdc AK |
11105 | /****************************************************************************** |
11106 | * function: imsm_create_metadata_update_for_size_change() | |
11107 | * Creates update for IMSM array for array size change. | |
11108 | * | |
11109 | ******************************************************************************/ | |
11110 | static int imsm_create_metadata_update_for_size_change( | |
11111 | struct supertype *st, | |
11112 | struct geo_params *geo, | |
11113 | struct imsm_update_size_change **updatep) | |
11114 | { | |
11115 | struct intel_super *super = st->sb; | |
594dc1b8 JS |
11116 | int update_memory_size; |
11117 | struct imsm_update_size_change *u; | |
f3871fdc | 11118 | |
1ade5cc1 | 11119 | dprintf("(enter) New size = %llu\n", geo->size); |
f3871fdc AK |
11120 | |
11121 | /* size of all update data without anchor */ | |
11122 | update_memory_size = sizeof(struct imsm_update_size_change); | |
11123 | ||
503975b9 | 11124 | u = xcalloc(1, update_memory_size); |
f3871fdc AK |
11125 | u->type = update_size_change; |
11126 | u->subdev = super->current_vol; | |
11127 | u->new_size = geo->size; | |
11128 | ||
11129 | dprintf("imsm: reshape update preparation : OK\n"); | |
11130 | *updatep = u; | |
11131 | ||
11132 | return update_memory_size; | |
11133 | } | |
11134 | ||
48c5303a PC |
11135 | /****************************************************************************** |
11136 | * function: imsm_create_metadata_update_for_migration() | |
11137 | * Creates update for IMSM array. | |
11138 | * | |
11139 | ******************************************************************************/ | |
11140 | static int imsm_create_metadata_update_for_migration( | |
11141 | struct supertype *st, | |
11142 | struct geo_params *geo, | |
11143 | struct imsm_update_reshape_migration **updatep) | |
11144 | { | |
11145 | struct intel_super *super = st->sb; | |
594dc1b8 JS |
11146 | int update_memory_size; |
11147 | struct imsm_update_reshape_migration *u; | |
48c5303a PC |
11148 | struct imsm_dev *dev; |
11149 | int previous_level = -1; | |
11150 | ||
1ade5cc1 | 11151 | dprintf("(enter) New Level = %i\n", geo->level); |
48c5303a PC |
11152 | |
11153 | /* size of all update data without anchor */ | |
11154 | update_memory_size = sizeof(struct imsm_update_reshape_migration); | |
11155 | ||
503975b9 | 11156 | u = xcalloc(1, update_memory_size); |
48c5303a PC |
11157 | u->type = update_reshape_migration; |
11158 | u->subdev = super->current_vol; | |
11159 | u->new_level = geo->level; | |
11160 | u->new_layout = geo->layout; | |
11161 | u->new_raid_disks = u->old_raid_disks = geo->raid_disks; | |
11162 | u->new_disks[0] = -1; | |
4bba0439 | 11163 | u->new_chunksize = -1; |
48c5303a PC |
11164 | |
11165 | dev = get_imsm_dev(super, u->subdev); | |
11166 | if (dev) { | |
11167 | struct imsm_map *map; | |
11168 | ||
238c0a71 | 11169 | map = get_imsm_map(dev, MAP_0); |
4bba0439 PC |
11170 | if (map) { |
11171 | int current_chunk_size = | |
11172 | __le16_to_cpu(map->blocks_per_strip) / 2; | |
11173 | ||
11174 | if (geo->chunksize != current_chunk_size) { | |
11175 | u->new_chunksize = geo->chunksize / 1024; | |
7a862a02 | 11176 | dprintf("imsm: chunk size change from %i to %i\n", |
4bba0439 PC |
11177 | current_chunk_size, u->new_chunksize); |
11178 | } | |
48c5303a | 11179 | previous_level = map->raid_level; |
4bba0439 | 11180 | } |
48c5303a | 11181 | } |
089f9d79 | 11182 | if (geo->level == 5 && previous_level == 0) { |
48c5303a PC |
11183 | struct mdinfo *spares = NULL; |
11184 | ||
11185 | u->new_raid_disks++; | |
11186 | spares = get_spares_for_grow(st); | |
089f9d79 | 11187 | if (spares == NULL || spares->array.spare_disks < 1) { |
48c5303a PC |
11188 | free(u); |
11189 | sysfs_free(spares); | |
11190 | update_memory_size = 0; | |
565cc99e | 11191 | pr_err("cannot get spare device for requested migration\n"); |
48c5303a PC |
11192 | return 0; |
11193 | } | |
11194 | sysfs_free(spares); | |
11195 | } | |
11196 | dprintf("imsm: reshape update preparation : OK\n"); | |
11197 | *updatep = u; | |
11198 | ||
11199 | return update_memory_size; | |
11200 | } | |
11201 | ||
8dd70bce AK |
11202 | static void imsm_update_metadata_locally(struct supertype *st, |
11203 | void *buf, int len) | |
11204 | { | |
11205 | struct metadata_update mu; | |
11206 | ||
11207 | mu.buf = buf; | |
11208 | mu.len = len; | |
11209 | mu.space = NULL; | |
11210 | mu.space_list = NULL; | |
11211 | mu.next = NULL; | |
5fe6f031 N |
11212 | if (imsm_prepare_update(st, &mu)) |
11213 | imsm_process_update(st, &mu); | |
8dd70bce AK |
11214 | |
11215 | while (mu.space_list) { | |
11216 | void **space = mu.space_list; | |
11217 | mu.space_list = *space; | |
11218 | free(space); | |
11219 | } | |
11220 | } | |
78b10e66 | 11221 | |
471bceb6 | 11222 | /*************************************************************************** |
694575e7 | 11223 | * Function: imsm_analyze_change |
471bceb6 | 11224 | * Description: Function analyze change for single volume |
1011e834 | 11225 | * and validate if transition is supported |
fbf3d202 AK |
11226 | * Parameters: Geometry parameters, supertype structure, |
11227 | * metadata change direction (apply/rollback) | |
694575e7 | 11228 | * Returns: Operation type code on success, -1 if fail |
471bceb6 KW |
11229 | ****************************************************************************/ |
11230 | enum imsm_reshape_type imsm_analyze_change(struct supertype *st, | |
fbf3d202 AK |
11231 | struct geo_params *geo, |
11232 | int direction) | |
694575e7 | 11233 | { |
471bceb6 KW |
11234 | struct mdinfo info; |
11235 | int change = -1; | |
11236 | int check_devs = 0; | |
c21e737b | 11237 | int chunk; |
67a2db32 AK |
11238 | /* number of added/removed disks in operation result */ |
11239 | int devNumChange = 0; | |
11240 | /* imsm compatible layout value for array geometry verification */ | |
11241 | int imsm_layout = -1; | |
7abc9871 AK |
11242 | int data_disks; |
11243 | struct imsm_dev *dev; | |
11244 | struct intel_super *super; | |
d04f65f4 | 11245 | unsigned long long current_size; |
65d38cca | 11246 | unsigned long long free_size; |
d04f65f4 | 11247 | unsigned long long max_size; |
65d38cca | 11248 | int rv; |
471bceb6 KW |
11249 | |
11250 | getinfo_super_imsm_volume(st, &info, NULL); | |
089f9d79 JS |
11251 | if (geo->level != info.array.level && geo->level >= 0 && |
11252 | geo->level != UnSet) { | |
471bceb6 KW |
11253 | switch (info.array.level) { |
11254 | case 0: | |
11255 | if (geo->level == 5) { | |
b5347799 | 11256 | change = CH_MIGRATION; |
e13ce846 | 11257 | if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) { |
7a862a02 | 11258 | pr_err("Error. Requested Layout not supported (left-asymmetric layout is supported only)!\n"); |
e13ce846 AK |
11259 | change = -1; |
11260 | goto analyse_change_exit; | |
11261 | } | |
67a2db32 | 11262 | imsm_layout = geo->layout; |
471bceb6 | 11263 | check_devs = 1; |
e91a3bad LM |
11264 | devNumChange = 1; /* parity disk added */ |
11265 | } else if (geo->level == 10) { | |
471bceb6 KW |
11266 | change = CH_TAKEOVER; |
11267 | check_devs = 1; | |
e91a3bad | 11268 | devNumChange = 2; /* two mirrors added */ |
67a2db32 | 11269 | imsm_layout = 0x102; /* imsm supported layout */ |
471bceb6 | 11270 | } |
dfe77a9e KW |
11271 | break; |
11272 | case 1: | |
471bceb6 KW |
11273 | case 10: |
11274 | if (geo->level == 0) { | |
11275 | change = CH_TAKEOVER; | |
11276 | check_devs = 1; | |
e91a3bad | 11277 | devNumChange = -(geo->raid_disks/2); |
67a2db32 | 11278 | imsm_layout = 0; /* imsm raid0 layout */ |
471bceb6 KW |
11279 | } |
11280 | break; | |
11281 | } | |
11282 | if (change == -1) { | |
7a862a02 | 11283 | pr_err("Error. Level Migration from %d to %d not supported!\n", |
e7b84f9d | 11284 | info.array.level, geo->level); |
471bceb6 KW |
11285 | goto analyse_change_exit; |
11286 | } | |
11287 | } else | |
11288 | geo->level = info.array.level; | |
11289 | ||
089f9d79 JS |
11290 | if (geo->layout != info.array.layout && |
11291 | (geo->layout != UnSet && geo->layout != -1)) { | |
b5347799 | 11292 | change = CH_MIGRATION; |
089f9d79 JS |
11293 | if (info.array.layout == 0 && info.array.level == 5 && |
11294 | geo->layout == 5) { | |
471bceb6 | 11295 | /* reshape 5 -> 4 */ |
089f9d79 JS |
11296 | } else if (info.array.layout == 5 && info.array.level == 5 && |
11297 | geo->layout == 0) { | |
471bceb6 KW |
11298 | /* reshape 4 -> 5 */ |
11299 | geo->layout = 0; | |
11300 | geo->level = 5; | |
11301 | } else { | |
7a862a02 | 11302 | pr_err("Error. Layout Migration from %d to %d not supported!\n", |
e7b84f9d | 11303 | info.array.layout, geo->layout); |
471bceb6 KW |
11304 | change = -1; |
11305 | goto analyse_change_exit; | |
11306 | } | |
67a2db32 | 11307 | } else { |
471bceb6 | 11308 | geo->layout = info.array.layout; |
67a2db32 AK |
11309 | if (imsm_layout == -1) |
11310 | imsm_layout = info.array.layout; | |
11311 | } | |
471bceb6 | 11312 | |
089f9d79 JS |
11313 | if (geo->chunksize > 0 && geo->chunksize != UnSet && |
11314 | geo->chunksize != info.array.chunk_size) { | |
2d2b0eb7 MD |
11315 | if (info.array.level == 10) { |
11316 | pr_err("Error. Chunk size change for RAID 10 is not supported.\n"); | |
11317 | change = -1; | |
11318 | goto analyse_change_exit; | |
1e9b2c3f PB |
11319 | } else if (info.component_size % (geo->chunksize/512)) { |
11320 | pr_err("New chunk size (%dK) does not evenly divide device size (%lluk). Aborting...\n", | |
11321 | geo->chunksize/1024, info.component_size/2); | |
11322 | change = -1; | |
11323 | goto analyse_change_exit; | |
2d2b0eb7 | 11324 | } |
b5347799 | 11325 | change = CH_MIGRATION; |
2d2b0eb7 | 11326 | } else { |
471bceb6 | 11327 | geo->chunksize = info.array.chunk_size; |
2d2b0eb7 | 11328 | } |
471bceb6 | 11329 | |
c21e737b | 11330 | chunk = geo->chunksize / 1024; |
7abc9871 AK |
11331 | |
11332 | super = st->sb; | |
11333 | dev = get_imsm_dev(super, super->current_vol); | |
11334 | data_disks = imsm_num_data_members(dev , MAP_0); | |
c41e00b2 | 11335 | /* compute current size per disk member |
7abc9871 | 11336 | */ |
c41e00b2 AK |
11337 | current_size = info.custom_array_size / data_disks; |
11338 | ||
089f9d79 | 11339 | if (geo->size > 0 && geo->size != MAX_SIZE) { |
c41e00b2 AK |
11340 | /* align component size |
11341 | */ | |
11342 | geo->size = imsm_component_size_aligment_check( | |
11343 | get_imsm_raid_level(dev->vol.map), | |
f36a9ecd | 11344 | chunk * 1024, super->sector_size, |
c41e00b2 | 11345 | geo->size * 2); |
65d0b4ce | 11346 | if (geo->size == 0) { |
7a862a02 | 11347 | pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is 0).\n", |
65d0b4ce LD |
11348 | current_size); |
11349 | goto analyse_change_exit; | |
11350 | } | |
c41e00b2 | 11351 | } |
7abc9871 | 11352 | |
089f9d79 | 11353 | if (current_size != geo->size && geo->size > 0) { |
7abc9871 | 11354 | if (change != -1) { |
7a862a02 | 11355 | pr_err("Error. Size change should be the only one at a time.\n"); |
7abc9871 AK |
11356 | change = -1; |
11357 | goto analyse_change_exit; | |
11358 | } | |
11359 | if ((super->current_vol + 1) != super->anchor->num_raid_devs) { | |
7a862a02 | 11360 | pr_err("Error. The last volume in container can be expanded only (%i/%s).\n", |
4dd2df09 | 11361 | super->current_vol, st->devnm); |
7abc9871 AK |
11362 | goto analyse_change_exit; |
11363 | } | |
65d38cca LD |
11364 | /* check the maximum available size |
11365 | */ | |
11366 | rv = imsm_get_free_size(st, dev->vol.map->num_members, | |
11367 | 0, chunk, &free_size); | |
11368 | if (rv == 0) | |
11369 | /* Cannot find maximum available space | |
11370 | */ | |
11371 | max_size = 0; | |
11372 | else { | |
11373 | max_size = free_size + current_size; | |
11374 | /* align component size | |
11375 | */ | |
11376 | max_size = imsm_component_size_aligment_check( | |
11377 | get_imsm_raid_level(dev->vol.map), | |
f36a9ecd | 11378 | chunk * 1024, super->sector_size, |
65d38cca LD |
11379 | max_size); |
11380 | } | |
d04f65f4 | 11381 | if (geo->size == MAX_SIZE) { |
b130333f AK |
11382 | /* requested size change to the maximum available size |
11383 | */ | |
65d38cca | 11384 | if (max_size == 0) { |
7a862a02 | 11385 | pr_err("Error. Cannot find maximum available space.\n"); |
b130333f AK |
11386 | change = -1; |
11387 | goto analyse_change_exit; | |
65d38cca LD |
11388 | } else |
11389 | geo->size = max_size; | |
c41e00b2 | 11390 | } |
b130333f | 11391 | |
681b7ae2 | 11392 | if (direction == ROLLBACK_METADATA_CHANGES) { |
fbf3d202 AK |
11393 | /* accept size for rollback only |
11394 | */ | |
11395 | } else { | |
11396 | /* round size due to metadata compatibility | |
11397 | */ | |
11398 | geo->size = (geo->size >> SECT_PER_MB_SHIFT) | |
11399 | << SECT_PER_MB_SHIFT; | |
11400 | dprintf("Prepare update for size change to %llu\n", | |
11401 | geo->size ); | |
11402 | if (current_size >= geo->size) { | |
7a862a02 | 11403 | pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is %llu).\n", |
e7b84f9d | 11404 | current_size, geo->size); |
fbf3d202 AK |
11405 | goto analyse_change_exit; |
11406 | } | |
65d38cca | 11407 | if (max_size && geo->size > max_size) { |
7a862a02 | 11408 | pr_err("Error. Requested size is larger than maximum available size (maximum available size is %llu, requested size /rounded/ is %llu).\n", |
e7b84f9d | 11409 | max_size, geo->size); |
65d38cca LD |
11410 | goto analyse_change_exit; |
11411 | } | |
7abc9871 AK |
11412 | } |
11413 | geo->size *= data_disks; | |
11414 | geo->raid_disks = dev->vol.map->num_members; | |
11415 | change = CH_ARRAY_SIZE; | |
11416 | } | |
471bceb6 KW |
11417 | if (!validate_geometry_imsm(st, |
11418 | geo->level, | |
67a2db32 | 11419 | imsm_layout, |
e91a3bad | 11420 | geo->raid_disks + devNumChange, |
c21e737b | 11421 | &chunk, |
af4348dd | 11422 | geo->size, INVALID_SECTORS, |
5308f117 | 11423 | 0, 0, info.consistency_policy, 1)) |
471bceb6 KW |
11424 | change = -1; |
11425 | ||
11426 | if (check_devs) { | |
11427 | struct intel_super *super = st->sb; | |
11428 | struct imsm_super *mpb = super->anchor; | |
11429 | ||
11430 | if (mpb->num_raid_devs > 1) { | |
7a862a02 | 11431 | pr_err("Error. Cannot perform operation on %s- for this operation it MUST be single array in container\n", |
e7b84f9d | 11432 | geo->dev_name); |
471bceb6 KW |
11433 | change = -1; |
11434 | } | |
11435 | } | |
11436 | ||
11437 | analyse_change_exit: | |
089f9d79 JS |
11438 | if (direction == ROLLBACK_METADATA_CHANGES && |
11439 | (change == CH_MIGRATION || change == CH_TAKEOVER)) { | |
7a862a02 | 11440 | dprintf("imsm: Metadata changes rollback is not supported for migration and takeover operations.\n"); |
fbf3d202 AK |
11441 | change = -1; |
11442 | } | |
471bceb6 | 11443 | return change; |
694575e7 KW |
11444 | } |
11445 | ||
bb025c2f KW |
11446 | int imsm_takeover(struct supertype *st, struct geo_params *geo) |
11447 | { | |
11448 | struct intel_super *super = st->sb; | |
11449 | struct imsm_update_takeover *u; | |
11450 | ||
503975b9 | 11451 | u = xmalloc(sizeof(struct imsm_update_takeover)); |
bb025c2f KW |
11452 | |
11453 | u->type = update_takeover; | |
11454 | u->subarray = super->current_vol; | |
11455 | ||
11456 | /* 10->0 transition */ | |
11457 | if (geo->level == 0) | |
11458 | u->direction = R10_TO_R0; | |
11459 | ||
0529c688 KW |
11460 | /* 0->10 transition */ |
11461 | if (geo->level == 10) | |
11462 | u->direction = R0_TO_R10; | |
11463 | ||
bb025c2f KW |
11464 | /* update metadata locally */ |
11465 | imsm_update_metadata_locally(st, u, | |
11466 | sizeof(struct imsm_update_takeover)); | |
11467 | /* and possibly remotely */ | |
11468 | if (st->update_tail) | |
11469 | append_metadata_update(st, u, | |
11470 | sizeof(struct imsm_update_takeover)); | |
11471 | else | |
11472 | free(u); | |
11473 | ||
11474 | return 0; | |
11475 | } | |
11476 | ||
d04f65f4 N |
11477 | static int imsm_reshape_super(struct supertype *st, unsigned long long size, |
11478 | int level, | |
78b10e66 | 11479 | int layout, int chunksize, int raid_disks, |
41784c88 | 11480 | int delta_disks, char *backup, char *dev, |
016e00f5 | 11481 | int direction, int verbose) |
78b10e66 | 11482 | { |
78b10e66 N |
11483 | int ret_val = 1; |
11484 | struct geo_params geo; | |
11485 | ||
1ade5cc1 | 11486 | dprintf("(enter)\n"); |
78b10e66 | 11487 | |
71204a50 | 11488 | memset(&geo, 0, sizeof(struct geo_params)); |
78b10e66 N |
11489 | |
11490 | geo.dev_name = dev; | |
4dd2df09 | 11491 | strcpy(geo.devnm, st->devnm); |
78b10e66 N |
11492 | geo.size = size; |
11493 | geo.level = level; | |
11494 | geo.layout = layout; | |
11495 | geo.chunksize = chunksize; | |
11496 | geo.raid_disks = raid_disks; | |
41784c88 AK |
11497 | if (delta_disks != UnSet) |
11498 | geo.raid_disks += delta_disks; | |
78b10e66 | 11499 | |
1ade5cc1 N |
11500 | dprintf("for level : %i\n", geo.level); |
11501 | dprintf("for raid_disks : %i\n", geo.raid_disks); | |
78b10e66 N |
11502 | |
11503 | if (experimental() == 0) | |
11504 | return ret_val; | |
11505 | ||
4dd2df09 | 11506 | if (strcmp(st->container_devnm, st->devnm) == 0) { |
694575e7 KW |
11507 | /* On container level we can only increase number of devices. */ |
11508 | dprintf("imsm: info: Container operation\n"); | |
78b10e66 | 11509 | int old_raid_disks = 0; |
6dc0be30 | 11510 | |
78b10e66 | 11511 | if (imsm_reshape_is_allowed_on_container( |
fbf3d202 | 11512 | st, &geo, &old_raid_disks, direction)) { |
78b10e66 N |
11513 | struct imsm_update_reshape *u = NULL; |
11514 | int len; | |
11515 | ||
11516 | len = imsm_create_metadata_update_for_reshape( | |
11517 | st, &geo, old_raid_disks, &u); | |
11518 | ||
ed08d51c AK |
11519 | if (len <= 0) { |
11520 | dprintf("imsm: Cannot prepare update\n"); | |
11521 | goto exit_imsm_reshape_super; | |
11522 | } | |
11523 | ||
8dd70bce AK |
11524 | ret_val = 0; |
11525 | /* update metadata locally */ | |
11526 | imsm_update_metadata_locally(st, u, len); | |
11527 | /* and possibly remotely */ | |
11528 | if (st->update_tail) | |
11529 | append_metadata_update(st, u, len); | |
11530 | else | |
ed08d51c | 11531 | free(u); |
8dd70bce | 11532 | |
694575e7 | 11533 | } else { |
7a862a02 | 11534 | pr_err("(imsm) Operation is not allowed on this container\n"); |
694575e7 KW |
11535 | } |
11536 | } else { | |
11537 | /* On volume level we support following operations | |
471bceb6 KW |
11538 | * - takeover: raid10 -> raid0; raid0 -> raid10 |
11539 | * - chunk size migration | |
11540 | * - migration: raid5 -> raid0; raid0 -> raid5 | |
11541 | */ | |
11542 | struct intel_super *super = st->sb; | |
11543 | struct intel_dev *dev = super->devlist; | |
4dd2df09 | 11544 | int change; |
694575e7 | 11545 | dprintf("imsm: info: Volume operation\n"); |
471bceb6 KW |
11546 | /* find requested device */ |
11547 | while (dev) { | |
1011e834 | 11548 | char *devnm = |
4dd2df09 N |
11549 | imsm_find_array_devnm_by_subdev( |
11550 | dev->index, st->container_devnm); | |
11551 | if (devnm && strcmp(devnm, geo.devnm) == 0) | |
471bceb6 KW |
11552 | break; |
11553 | dev = dev->next; | |
11554 | } | |
11555 | if (dev == NULL) { | |
4dd2df09 N |
11556 | pr_err("Cannot find %s (%s) subarray\n", |
11557 | geo.dev_name, geo.devnm); | |
471bceb6 KW |
11558 | goto exit_imsm_reshape_super; |
11559 | } | |
11560 | super->current_vol = dev->index; | |
fbf3d202 | 11561 | change = imsm_analyze_change(st, &geo, direction); |
694575e7 | 11562 | switch (change) { |
471bceb6 | 11563 | case CH_TAKEOVER: |
bb025c2f | 11564 | ret_val = imsm_takeover(st, &geo); |
694575e7 | 11565 | break; |
48c5303a PC |
11566 | case CH_MIGRATION: { |
11567 | struct imsm_update_reshape_migration *u = NULL; | |
11568 | int len = | |
11569 | imsm_create_metadata_update_for_migration( | |
11570 | st, &geo, &u); | |
11571 | if (len < 1) { | |
7a862a02 | 11572 | dprintf("imsm: Cannot prepare update\n"); |
48c5303a PC |
11573 | break; |
11574 | } | |
471bceb6 | 11575 | ret_val = 0; |
48c5303a PC |
11576 | /* update metadata locally */ |
11577 | imsm_update_metadata_locally(st, u, len); | |
11578 | /* and possibly remotely */ | |
11579 | if (st->update_tail) | |
11580 | append_metadata_update(st, u, len); | |
11581 | else | |
11582 | free(u); | |
11583 | } | |
11584 | break; | |
7abc9871 | 11585 | case CH_ARRAY_SIZE: { |
f3871fdc AK |
11586 | struct imsm_update_size_change *u = NULL; |
11587 | int len = | |
11588 | imsm_create_metadata_update_for_size_change( | |
11589 | st, &geo, &u); | |
11590 | if (len < 1) { | |
7a862a02 | 11591 | dprintf("imsm: Cannot prepare update\n"); |
f3871fdc AK |
11592 | break; |
11593 | } | |
11594 | ret_val = 0; | |
11595 | /* update metadata locally */ | |
11596 | imsm_update_metadata_locally(st, u, len); | |
11597 | /* and possibly remotely */ | |
11598 | if (st->update_tail) | |
11599 | append_metadata_update(st, u, len); | |
11600 | else | |
11601 | free(u); | |
7abc9871 AK |
11602 | } |
11603 | break; | |
471bceb6 KW |
11604 | default: |
11605 | ret_val = 1; | |
694575e7 | 11606 | } |
694575e7 | 11607 | } |
78b10e66 | 11608 | |
ed08d51c | 11609 | exit_imsm_reshape_super: |
78b10e66 N |
11610 | dprintf("imsm: reshape_super Exit code = %i\n", ret_val); |
11611 | return ret_val; | |
11612 | } | |
2cda7640 | 11613 | |
0febb20c AO |
11614 | #define COMPLETED_OK 0 |
11615 | #define COMPLETED_NONE 1 | |
11616 | #define COMPLETED_DELAYED 2 | |
11617 | ||
11618 | static int read_completed(int fd, unsigned long long *val) | |
11619 | { | |
11620 | int ret; | |
11621 | char buf[50]; | |
11622 | ||
11623 | ret = sysfs_fd_get_str(fd, buf, 50); | |
11624 | if (ret < 0) | |
11625 | return ret; | |
11626 | ||
11627 | ret = COMPLETED_OK; | |
11628 | if (strncmp(buf, "none", 4) == 0) { | |
11629 | ret = COMPLETED_NONE; | |
11630 | } else if (strncmp(buf, "delayed", 7) == 0) { | |
11631 | ret = COMPLETED_DELAYED; | |
11632 | } else { | |
11633 | char *ep; | |
11634 | *val = strtoull(buf, &ep, 0); | |
11635 | if (ep == buf || (*ep != 0 && *ep != '\n' && *ep != ' ')) | |
11636 | ret = -1; | |
11637 | } | |
11638 | return ret; | |
11639 | } | |
11640 | ||
eee67a47 AK |
11641 | /******************************************************************************* |
11642 | * Function: wait_for_reshape_imsm | |
11643 | * Description: Function writes new sync_max value and waits until | |
11644 | * reshape process reach new position | |
11645 | * Parameters: | |
11646 | * sra : general array info | |
eee67a47 AK |
11647 | * ndata : number of disks in new array's layout |
11648 | * Returns: | |
11649 | * 0 : success, | |
11650 | * 1 : there is no reshape in progress, | |
11651 | * -1 : fail | |
11652 | ******************************************************************************/ | |
ae9f01f8 | 11653 | int wait_for_reshape_imsm(struct mdinfo *sra, int ndata) |
eee67a47 | 11654 | { |
85ca499c | 11655 | int fd = sysfs_get_fd(sra, NULL, "sync_completed"); |
df2647fa | 11656 | int retry = 3; |
eee67a47 | 11657 | unsigned long long completed; |
ae9f01f8 AK |
11658 | /* to_complete : new sync_max position */ |
11659 | unsigned long long to_complete = sra->reshape_progress; | |
11660 | unsigned long long position_to_set = to_complete / ndata; | |
eee67a47 | 11661 | |
ae9f01f8 | 11662 | if (fd < 0) { |
1ade5cc1 | 11663 | dprintf("cannot open reshape_position\n"); |
eee67a47 | 11664 | return 1; |
ae9f01f8 | 11665 | } |
eee67a47 | 11666 | |
df2647fa PB |
11667 | do { |
11668 | if (sysfs_fd_get_ll(fd, &completed) < 0) { | |
11669 | if (!retry) { | |
11670 | dprintf("cannot read reshape_position (no reshape in progres)\n"); | |
11671 | close(fd); | |
11672 | return 1; | |
11673 | } | |
11674 | usleep(30000); | |
11675 | } else | |
11676 | break; | |
11677 | } while (retry--); | |
eee67a47 | 11678 | |
85ca499c | 11679 | if (completed > position_to_set) { |
1ade5cc1 | 11680 | dprintf("wrong next position to set %llu (%llu)\n", |
85ca499c | 11681 | to_complete, position_to_set); |
ae9f01f8 AK |
11682 | close(fd); |
11683 | return -1; | |
11684 | } | |
11685 | dprintf("Position set: %llu\n", position_to_set); | |
11686 | if (sysfs_set_num(sra, NULL, "sync_max", | |
11687 | position_to_set) != 0) { | |
1ade5cc1 | 11688 | dprintf("cannot set reshape position to %llu\n", |
ae9f01f8 AK |
11689 | position_to_set); |
11690 | close(fd); | |
11691 | return -1; | |
eee67a47 AK |
11692 | } |
11693 | ||
eee67a47 | 11694 | do { |
0febb20c | 11695 | int rc; |
eee67a47 | 11696 | char action[20]; |
5ff3a780 | 11697 | int timeout = 3000; |
0febb20c | 11698 | |
5ff3a780 | 11699 | sysfs_wait(fd, &timeout); |
a47e44fb AK |
11700 | if (sysfs_get_str(sra, NULL, "sync_action", |
11701 | action, 20) > 0 && | |
d7d3809a | 11702 | strncmp(action, "reshape", 7) != 0) { |
b2be2b62 AO |
11703 | if (strncmp(action, "idle", 4) == 0) |
11704 | break; | |
d7d3809a AP |
11705 | close(fd); |
11706 | return -1; | |
11707 | } | |
0febb20c AO |
11708 | |
11709 | rc = read_completed(fd, &completed); | |
11710 | if (rc < 0) { | |
1ade5cc1 | 11711 | dprintf("cannot read reshape_position (in loop)\n"); |
eee67a47 AK |
11712 | close(fd); |
11713 | return 1; | |
0febb20c AO |
11714 | } else if (rc == COMPLETED_NONE) |
11715 | break; | |
85ca499c | 11716 | } while (completed < position_to_set); |
b2be2b62 | 11717 | |
eee67a47 AK |
11718 | close(fd); |
11719 | return 0; | |
eee67a47 AK |
11720 | } |
11721 | ||
b915c95f AK |
11722 | /******************************************************************************* |
11723 | * Function: check_degradation_change | |
11724 | * Description: Check that array hasn't become failed. | |
11725 | * Parameters: | |
11726 | * info : for sysfs access | |
11727 | * sources : source disks descriptors | |
11728 | * degraded: previous degradation level | |
11729 | * Returns: | |
11730 | * degradation level | |
11731 | ******************************************************************************/ | |
11732 | int check_degradation_change(struct mdinfo *info, | |
11733 | int *sources, | |
11734 | int degraded) | |
11735 | { | |
11736 | unsigned long long new_degraded; | |
e1993023 LD |
11737 | int rv; |
11738 | ||
11739 | rv = sysfs_get_ll(info, NULL, "degraded", &new_degraded); | |
089f9d79 | 11740 | if (rv == -1 || (new_degraded != (unsigned long long)degraded)) { |
b915c95f AK |
11741 | /* check each device to ensure it is still working */ |
11742 | struct mdinfo *sd; | |
11743 | new_degraded = 0; | |
11744 | for (sd = info->devs ; sd ; sd = sd->next) { | |
11745 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) | |
11746 | continue; | |
11747 | if (sd->disk.state & (1<<MD_DISK_SYNC)) { | |
cf52eff5 TM |
11748 | char sbuf[100]; |
11749 | ||
b915c95f | 11750 | if (sysfs_get_str(info, |
cf52eff5 | 11751 | sd, "state", sbuf, sizeof(sbuf)) < 0 || |
b915c95f AK |
11752 | strstr(sbuf, "faulty") || |
11753 | strstr(sbuf, "in_sync") == NULL) { | |
11754 | /* this device is dead */ | |
11755 | sd->disk.state = (1<<MD_DISK_FAULTY); | |
11756 | if (sd->disk.raid_disk >= 0 && | |
11757 | sources[sd->disk.raid_disk] >= 0) { | |
11758 | close(sources[ | |
11759 | sd->disk.raid_disk]); | |
11760 | sources[sd->disk.raid_disk] = | |
11761 | -1; | |
11762 | } | |
11763 | new_degraded++; | |
11764 | } | |
11765 | } | |
11766 | } | |
11767 | } | |
11768 | ||
11769 | return new_degraded; | |
11770 | } | |
11771 | ||
10f22854 AK |
11772 | /******************************************************************************* |
11773 | * Function: imsm_manage_reshape | |
11774 | * Description: Function finds array under reshape and it manages reshape | |
11775 | * process. It creates stripes backups (if required) and sets | |
942e1cdb | 11776 | * checkpoints. |
10f22854 AK |
11777 | * Parameters: |
11778 | * afd : Backup handle (nattive) - not used | |
11779 | * sra : general array info | |
11780 | * reshape : reshape parameters - not used | |
11781 | * st : supertype structure | |
11782 | * blocks : size of critical section [blocks] | |
11783 | * fds : table of source device descriptor | |
11784 | * offsets : start of array (offest per devices) | |
11785 | * dests : not used | |
11786 | * destfd : table of destination device descriptor | |
11787 | * destoffsets : table of destination offsets (per device) | |
11788 | * Returns: | |
11789 | * 1 : success, reshape is done | |
11790 | * 0 : fail | |
11791 | ******************************************************************************/ | |
999b4972 N |
11792 | static int imsm_manage_reshape( |
11793 | int afd, struct mdinfo *sra, struct reshape *reshape, | |
10f22854 | 11794 | struct supertype *st, unsigned long backup_blocks, |
999b4972 N |
11795 | int *fds, unsigned long long *offsets, |
11796 | int dests, int *destfd, unsigned long long *destoffsets) | |
11797 | { | |
10f22854 AK |
11798 | int ret_val = 0; |
11799 | struct intel_super *super = st->sb; | |
594dc1b8 | 11800 | struct intel_dev *dv; |
de44e46f | 11801 | unsigned int sector_size = super->sector_size; |
10f22854 | 11802 | struct imsm_dev *dev = NULL; |
a6b6d984 | 11803 | struct imsm_map *map_src; |
10f22854 AK |
11804 | int migr_vol_qan = 0; |
11805 | int ndata, odata; /* [bytes] */ | |
11806 | int chunk; /* [bytes] */ | |
11807 | struct migr_record *migr_rec; | |
11808 | char *buf = NULL; | |
11809 | unsigned int buf_size; /* [bytes] */ | |
11810 | unsigned long long max_position; /* array size [bytes] */ | |
11811 | unsigned long long next_step; /* [blocks]/[bytes] */ | |
11812 | unsigned long long old_data_stripe_length; | |
10f22854 AK |
11813 | unsigned long long start_src; /* [bytes] */ |
11814 | unsigned long long start; /* [bytes] */ | |
11815 | unsigned long long start_buf_shift; /* [bytes] */ | |
b915c95f | 11816 | int degraded = 0; |
ab724b98 | 11817 | int source_layout = 0; |
10f22854 | 11818 | |
79a16a9b JS |
11819 | if (!sra) |
11820 | return ret_val; | |
11821 | ||
11822 | if (!fds || !offsets) | |
10f22854 AK |
11823 | goto abort; |
11824 | ||
11825 | /* Find volume during the reshape */ | |
11826 | for (dv = super->devlist; dv; dv = dv->next) { | |
fc54fe7a JS |
11827 | if (dv->dev->vol.migr_type == MIGR_GEN_MIGR && |
11828 | dv->dev->vol.migr_state == 1) { | |
10f22854 AK |
11829 | dev = dv->dev; |
11830 | migr_vol_qan++; | |
11831 | } | |
11832 | } | |
11833 | /* Only one volume can migrate at the same time */ | |
11834 | if (migr_vol_qan != 1) { | |
676e87a8 | 11835 | pr_err("%s", migr_vol_qan ? |
10f22854 AK |
11836 | "Number of migrating volumes greater than 1\n" : |
11837 | "There is no volume during migrationg\n"); | |
11838 | goto abort; | |
11839 | } | |
11840 | ||
238c0a71 | 11841 | map_src = get_imsm_map(dev, MAP_1); |
10f22854 AK |
11842 | if (map_src == NULL) |
11843 | goto abort; | |
10f22854 | 11844 | |
238c0a71 AK |
11845 | ndata = imsm_num_data_members(dev, MAP_0); |
11846 | odata = imsm_num_data_members(dev, MAP_1); | |
10f22854 | 11847 | |
7b1ab482 | 11848 | chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512; |
10f22854 AK |
11849 | old_data_stripe_length = odata * chunk; |
11850 | ||
11851 | migr_rec = super->migr_rec; | |
11852 | ||
10f22854 AK |
11853 | /* initialize migration record for start condition */ |
11854 | if (sra->reshape_progress == 0) | |
11855 | init_migr_record_imsm(st, dev, sra); | |
b2c59438 AK |
11856 | else { |
11857 | if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) { | |
7a862a02 | 11858 | dprintf("imsm: cannot restart migration when data are present in copy area.\n"); |
b2c59438 AK |
11859 | goto abort; |
11860 | } | |
6a75c8ca AK |
11861 | /* Save checkpoint to update migration record for current |
11862 | * reshape position (in md). It can be farther than current | |
11863 | * reshape position in metadata. | |
11864 | */ | |
11865 | if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) { | |
11866 | /* ignore error == 2, this can mean end of reshape here | |
11867 | */ | |
7a862a02 | 11868 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL, initial save)\n"); |
6a75c8ca AK |
11869 | goto abort; |
11870 | } | |
b2c59438 | 11871 | } |
10f22854 AK |
11872 | |
11873 | /* size for data */ | |
11874 | buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512; | |
11875 | /* extend buffer size for parity disk */ | |
11876 | buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
11877 | /* add space for stripe aligment */ | |
11878 | buf_size += old_data_stripe_length; | |
de44e46f PB |
11879 | if (posix_memalign((void **)&buf, MAX_SECTOR_SIZE, buf_size)) { |
11880 | dprintf("imsm: Cannot allocate checkpoint buffer\n"); | |
10f22854 AK |
11881 | goto abort; |
11882 | } | |
11883 | ||
3ef4403c | 11884 | max_position = sra->component_size * ndata; |
68eb8bc6 | 11885 | source_layout = imsm_level_to_layout(map_src->raid_level); |
10f22854 AK |
11886 | |
11887 | while (__le32_to_cpu(migr_rec->curr_migr_unit) < | |
11888 | __le32_to_cpu(migr_rec->num_migr_units)) { | |
11889 | /* current reshape position [blocks] */ | |
11890 | unsigned long long current_position = | |
11891 | __le32_to_cpu(migr_rec->blocks_per_unit) | |
11892 | * __le32_to_cpu(migr_rec->curr_migr_unit); | |
11893 | unsigned long long border; | |
11894 | ||
b915c95f AK |
11895 | /* Check that array hasn't become failed. |
11896 | */ | |
11897 | degraded = check_degradation_change(sra, fds, degraded); | |
11898 | if (degraded > 1) { | |
7a862a02 | 11899 | dprintf("imsm: Abort reshape due to degradation level (%i)\n", degraded); |
b915c95f AK |
11900 | goto abort; |
11901 | } | |
11902 | ||
10f22854 AK |
11903 | next_step = __le32_to_cpu(migr_rec->blocks_per_unit); |
11904 | ||
11905 | if ((current_position + next_step) > max_position) | |
11906 | next_step = max_position - current_position; | |
11907 | ||
92144abf | 11908 | start = current_position * 512; |
10f22854 | 11909 | |
942e1cdb | 11910 | /* align reading start to old geometry */ |
10f22854 AK |
11911 | start_buf_shift = start % old_data_stripe_length; |
11912 | start_src = start - start_buf_shift; | |
11913 | ||
11914 | border = (start_src / odata) - (start / ndata); | |
11915 | border /= 512; | |
11916 | if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) { | |
11917 | /* save critical stripes to buf | |
11918 | * start - start address of current unit | |
11919 | * to backup [bytes] | |
11920 | * start_src - start address of current unit | |
11921 | * to backup alligned to source array | |
11922 | * [bytes] | |
11923 | */ | |
594dc1b8 | 11924 | unsigned long long next_step_filler; |
10f22854 AK |
11925 | unsigned long long copy_length = next_step * 512; |
11926 | ||
11927 | /* allign copy area length to stripe in old geometry */ | |
11928 | next_step_filler = ((copy_length + start_buf_shift) | |
11929 | % old_data_stripe_length); | |
11930 | if (next_step_filler) | |
11931 | next_step_filler = (old_data_stripe_length | |
11932 | - next_step_filler); | |
7a862a02 | 11933 | 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 |
11934 | start, start_src, copy_length, |
11935 | start_buf_shift, next_step_filler); | |
11936 | ||
11937 | if (save_stripes(fds, offsets, map_src->num_members, | |
ab724b98 AK |
11938 | chunk, map_src->raid_level, |
11939 | source_layout, 0, NULL, start_src, | |
10f22854 AK |
11940 | copy_length + |
11941 | next_step_filler + start_buf_shift, | |
11942 | buf)) { | |
7a862a02 | 11943 | dprintf("imsm: Cannot save stripes to buffer\n"); |
10f22854 AK |
11944 | goto abort; |
11945 | } | |
11946 | /* Convert data to destination format and store it | |
11947 | * in backup general migration area | |
11948 | */ | |
11949 | if (save_backup_imsm(st, dev, sra, | |
aea93171 | 11950 | buf + start_buf_shift, copy_length)) { |
7a862a02 | 11951 | dprintf("imsm: Cannot save stripes to target devices\n"); |
10f22854 AK |
11952 | goto abort; |
11953 | } | |
11954 | if (save_checkpoint_imsm(st, sra, | |
11955 | UNIT_SRC_IN_CP_AREA)) { | |
7a862a02 | 11956 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_IN_CP_AREA)\n"); |
10f22854 AK |
11957 | goto abort; |
11958 | } | |
8016a6d4 AK |
11959 | } else { |
11960 | /* set next step to use whole border area */ | |
11961 | border /= next_step; | |
11962 | if (border > 1) | |
11963 | next_step *= border; | |
10f22854 AK |
11964 | } |
11965 | /* When data backed up, checkpoint stored, | |
11966 | * kick the kernel to reshape unit of data | |
11967 | */ | |
11968 | next_step = next_step + sra->reshape_progress; | |
8016a6d4 AK |
11969 | /* limit next step to array max position */ |
11970 | if (next_step > max_position) | |
11971 | next_step = max_position; | |
10f22854 AK |
11972 | sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress); |
11973 | sysfs_set_num(sra, NULL, "suspend_hi", next_step); | |
ae9f01f8 | 11974 | sra->reshape_progress = next_step; |
10f22854 AK |
11975 | |
11976 | /* wait until reshape finish */ | |
c85338c6 | 11977 | if (wait_for_reshape_imsm(sra, ndata)) { |
c47b0ff6 AK |
11978 | dprintf("wait_for_reshape_imsm returned error!\n"); |
11979 | goto abort; | |
11980 | } | |
84d11e6c N |
11981 | if (sigterm) |
11982 | goto abort; | |
10f22854 | 11983 | |
0228d92c AK |
11984 | if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) { |
11985 | /* ignore error == 2, this can mean end of reshape here | |
11986 | */ | |
7a862a02 | 11987 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL)\n"); |
10f22854 AK |
11988 | goto abort; |
11989 | } | |
11990 | ||
11991 | } | |
11992 | ||
71e5411e PB |
11993 | /* clear migr_rec on disks after successful migration */ |
11994 | struct dl *d; | |
11995 | ||
85337573 | 11996 | memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE); |
71e5411e PB |
11997 | for (d = super->disks; d; d = d->next) { |
11998 | if (d->index < 0 || is_failed(&d->disk)) | |
11999 | continue; | |
12000 | unsigned long long dsize; | |
12001 | ||
12002 | get_dev_size(d->fd, NULL, &dsize); | |
de44e46f | 12003 | if (lseek64(d->fd, dsize - MIGR_REC_SECTOR_POSITION*sector_size, |
71e5411e | 12004 | SEEK_SET) >= 0) { |
466070ad | 12005 | if ((unsigned int)write(d->fd, super->migr_rec_buf, |
de44e46f PB |
12006 | MIGR_REC_BUF_SECTORS*sector_size) != |
12007 | MIGR_REC_BUF_SECTORS*sector_size) | |
71e5411e PB |
12008 | perror("Write migr_rec failed"); |
12009 | } | |
12010 | } | |
12011 | ||
10f22854 AK |
12012 | /* return '1' if done */ |
12013 | ret_val = 1; | |
12014 | abort: | |
12015 | free(buf); | |
942e1cdb N |
12016 | /* See Grow.c: abort_reshape() for further explanation */ |
12017 | sysfs_set_num(sra, NULL, "suspend_lo", 0x7FFFFFFFFFFFFFFFULL); | |
12018 | sysfs_set_num(sra, NULL, "suspend_hi", 0); | |
12019 | sysfs_set_num(sra, NULL, "suspend_lo", 0); | |
10f22854 AK |
12020 | |
12021 | return ret_val; | |
999b4972 | 12022 | } |
0c21b485 | 12023 | |
cdddbdbc | 12024 | struct superswitch super_imsm = { |
cdddbdbc DW |
12025 | .examine_super = examine_super_imsm, |
12026 | .brief_examine_super = brief_examine_super_imsm, | |
4737ae25 | 12027 | .brief_examine_subarrays = brief_examine_subarrays_imsm, |
9d84c8ea | 12028 | .export_examine_super = export_examine_super_imsm, |
cdddbdbc DW |
12029 | .detail_super = detail_super_imsm, |
12030 | .brief_detail_super = brief_detail_super_imsm, | |
bf5a934a | 12031 | .write_init_super = write_init_super_imsm, |
0e600426 N |
12032 | .validate_geometry = validate_geometry_imsm, |
12033 | .add_to_super = add_to_super_imsm, | |
1a64be56 | 12034 | .remove_from_super = remove_from_super_imsm, |
d665cc31 | 12035 | .detail_platform = detail_platform_imsm, |
e50cf220 | 12036 | .export_detail_platform = export_detail_platform_imsm, |
33414a01 | 12037 | .kill_subarray = kill_subarray_imsm, |
aa534678 | 12038 | .update_subarray = update_subarray_imsm, |
2b959fbf | 12039 | .load_container = load_container_imsm, |
71204a50 N |
12040 | .default_geometry = default_geometry_imsm, |
12041 | .get_disk_controller_domain = imsm_get_disk_controller_domain, | |
12042 | .reshape_super = imsm_reshape_super, | |
12043 | .manage_reshape = imsm_manage_reshape, | |
9e2d750d | 12044 | .recover_backup = recover_backup_imsm, |
74db60b0 | 12045 | .copy_metadata = copy_metadata_imsm, |
27156a57 | 12046 | .examine_badblocks = examine_badblocks_imsm, |
cdddbdbc DW |
12047 | .match_home = match_home_imsm, |
12048 | .uuid_from_super= uuid_from_super_imsm, | |
12049 | .getinfo_super = getinfo_super_imsm, | |
5c4cd5da | 12050 | .getinfo_super_disks = getinfo_super_disks_imsm, |
cdddbdbc DW |
12051 | .update_super = update_super_imsm, |
12052 | ||
12053 | .avail_size = avail_size_imsm, | |
fbfdcb06 | 12054 | .get_spare_criteria = get_spare_criteria_imsm, |
cdddbdbc DW |
12055 | |
12056 | .compare_super = compare_super_imsm, | |
12057 | ||
12058 | .load_super = load_super_imsm, | |
bf5a934a | 12059 | .init_super = init_super_imsm, |
e683ca88 | 12060 | .store_super = store_super_imsm, |
cdddbdbc DW |
12061 | .free_super = free_super_imsm, |
12062 | .match_metadata_desc = match_metadata_desc_imsm, | |
bf5a934a | 12063 | .container_content = container_content_imsm, |
0c21b485 | 12064 | .validate_container = validate_container_imsm, |
cdddbdbc | 12065 | |
2432ce9b AP |
12066 | .write_init_ppl = write_init_ppl_imsm, |
12067 | .validate_ppl = validate_ppl_imsm, | |
12068 | ||
cdddbdbc | 12069 | .external = 1, |
4cce4069 | 12070 | .name = "imsm", |
845dea95 NB |
12071 | |
12072 | /* for mdmon */ | |
12073 | .open_new = imsm_open_new, | |
ed9d66aa | 12074 | .set_array_state= imsm_set_array_state, |
845dea95 NB |
12075 | .set_disk = imsm_set_disk, |
12076 | .sync_metadata = imsm_sync_metadata, | |
88758e9d | 12077 | .activate_spare = imsm_activate_spare, |
e8319a19 | 12078 | .process_update = imsm_process_update, |
8273f55e | 12079 | .prepare_update = imsm_prepare_update, |
6f50473f | 12080 | .record_bad_block = imsm_record_badblock, |
c07a5a4f | 12081 | .clear_bad_block = imsm_clear_badblock, |
928f1424 | 12082 | .get_bad_blocks = imsm_get_badblocks, |
cdddbdbc | 12083 | }; |