]>
Commit | Line | Data |
---|---|---|
cdddbdbc DW |
1 | /* |
2 | * mdadm - Intel(R) Matrix Storage Manager Support | |
3 | * | |
a54d5262 | 4 | * Copyright (C) 2002-2008 Intel Corporation |
cdddbdbc DW |
5 | * |
6 | * This program is free software; you can redistribute it and/or modify it | |
7 | * under the terms and conditions of the GNU General Public License, | |
8 | * version 2, as published by the Free Software Foundation. | |
9 | * | |
10 | * This program is distributed in the hope it will be useful, but WITHOUT | |
11 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
12 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for | |
13 | * more details. | |
14 | * | |
15 | * You should have received a copy of the GNU General Public License along with | |
16 | * this program; if not, write to the Free Software Foundation, Inc., | |
17 | * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA. | |
18 | */ | |
19 | ||
51006d85 | 20 | #define HAVE_STDINT_H 1 |
cdddbdbc | 21 | #include "mdadm.h" |
c2a1e7da | 22 | #include "mdmon.h" |
51006d85 | 23 | #include "sha1.h" |
88c32bb1 | 24 | #include "platform-intel.h" |
cdddbdbc DW |
25 | #include <values.h> |
26 | #include <scsi/sg.h> | |
27 | #include <ctype.h> | |
d665cc31 | 28 | #include <dirent.h> |
cdddbdbc DW |
29 | |
30 | /* MPB == Metadata Parameter Block */ | |
31 | #define MPB_SIGNATURE "Intel Raid ISM Cfg Sig. " | |
32 | #define MPB_SIG_LEN (strlen(MPB_SIGNATURE)) | |
33 | #define MPB_VERSION_RAID0 "1.0.00" | |
34 | #define MPB_VERSION_RAID1 "1.1.00" | |
fe7ed8cb DW |
35 | #define MPB_VERSION_MANY_VOLUMES_PER_ARRAY "1.2.00" |
36 | #define MPB_VERSION_3OR4_DISK_ARRAY "1.2.01" | |
cdddbdbc | 37 | #define MPB_VERSION_RAID5 "1.2.02" |
fe7ed8cb DW |
38 | #define MPB_VERSION_5OR6_DISK_ARRAY "1.2.04" |
39 | #define MPB_VERSION_CNG "1.2.06" | |
40 | #define MPB_VERSION_ATTRIBS "1.3.00" | |
cdddbdbc DW |
41 | #define MAX_SIGNATURE_LENGTH 32 |
42 | #define MAX_RAID_SERIAL_LEN 16 | |
fe7ed8cb | 43 | |
19482bcc AK |
44 | /* supports RAID0 */ |
45 | #define MPB_ATTRIB_RAID0 __cpu_to_le32(0x00000001) | |
46 | /* supports RAID1 */ | |
47 | #define MPB_ATTRIB_RAID1 __cpu_to_le32(0x00000002) | |
48 | /* supports RAID10 */ | |
49 | #define MPB_ATTRIB_RAID10 __cpu_to_le32(0x00000004) | |
50 | /* supports RAID1E */ | |
51 | #define MPB_ATTRIB_RAID1E __cpu_to_le32(0x00000008) | |
52 | /* supports RAID5 */ | |
53 | #define MPB_ATTRIB_RAID5 __cpu_to_le32(0x00000010) | |
54 | /* supports RAID CNG */ | |
55 | #define MPB_ATTRIB_RAIDCNG __cpu_to_le32(0x00000020) | |
56 | /* supports expanded stripe sizes of 256K, 512K and 1MB */ | |
57 | #define MPB_ATTRIB_EXP_STRIPE_SIZE __cpu_to_le32(0x00000040) | |
58 | ||
59 | /* The OROM Support RST Caching of Volumes */ | |
60 | #define MPB_ATTRIB_NVM __cpu_to_le32(0x02000000) | |
61 | /* The OROM supports creating disks greater than 2TB */ | |
62 | #define MPB_ATTRIB_2TB_DISK __cpu_to_le32(0x04000000) | |
63 | /* The OROM supports Bad Block Management */ | |
64 | #define MPB_ATTRIB_BBM __cpu_to_le32(0x08000000) | |
65 | ||
66 | /* THe OROM Supports NVM Caching of Volumes */ | |
67 | #define MPB_ATTRIB_NEVER_USE2 __cpu_to_le32(0x10000000) | |
68 | /* The OROM supports creating volumes greater than 2TB */ | |
69 | #define MPB_ATTRIB_2TB __cpu_to_le32(0x20000000) | |
70 | /* originally for PMP, now it's wasted b/c. Never use this bit! */ | |
71 | #define MPB_ATTRIB_NEVER_USE __cpu_to_le32(0x40000000) | |
72 | /* Verify MPB contents against checksum after reading MPB */ | |
73 | #define MPB_ATTRIB_CHECKSUM_VERIFY __cpu_to_le32(0x80000000) | |
74 | ||
75 | /* Define all supported attributes that have to be accepted by mdadm | |
76 | */ | |
418f9b36 | 77 | #define MPB_ATTRIB_SUPPORTED (MPB_ATTRIB_CHECKSUM_VERIFY | \ |
19482bcc AK |
78 | MPB_ATTRIB_2TB | \ |
79 | MPB_ATTRIB_2TB_DISK | \ | |
80 | MPB_ATTRIB_RAID0 | \ | |
81 | MPB_ATTRIB_RAID1 | \ | |
82 | MPB_ATTRIB_RAID10 | \ | |
83 | MPB_ATTRIB_RAID5 | \ | |
bbab0940 TM |
84 | MPB_ATTRIB_EXP_STRIPE_SIZE | \ |
85 | MPB_ATTRIB_BBM) | |
418f9b36 N |
86 | |
87 | /* Define attributes that are unused but not harmful */ | |
88 | #define MPB_ATTRIB_IGNORED (MPB_ATTRIB_NEVER_USE) | |
fe7ed8cb | 89 | |
8e59f3d8 | 90 | #define MPB_SECTOR_CNT 2210 |
c2c087e6 | 91 | #define IMSM_RESERVED_SECTORS 4096 |
b81221b7 | 92 | #define NUM_BLOCKS_DIRTY_STRIPE_REGION 2056 |
979d38be | 93 | #define SECT_PER_MB_SHIFT 11 |
f36a9ecd | 94 | #define MAX_SECTOR_SIZE 4096 |
cdddbdbc DW |
95 | |
96 | /* Disk configuration info. */ | |
97 | #define IMSM_MAX_DEVICES 255 | |
98 | struct imsm_disk { | |
99 | __u8 serial[MAX_RAID_SERIAL_LEN];/* 0xD8 - 0xE7 ascii serial number */ | |
5551b113 | 100 | __u32 total_blocks_lo; /* 0xE8 - 0xEB total blocks lo */ |
cdddbdbc | 101 | __u32 scsi_id; /* 0xEC - 0xEF scsi ID */ |
f2f27e63 DW |
102 | #define SPARE_DISK __cpu_to_le32(0x01) /* Spare */ |
103 | #define CONFIGURED_DISK __cpu_to_le32(0x02) /* Member of some RaidDev */ | |
104 | #define FAILED_DISK __cpu_to_le32(0x04) /* Permanent failure */ | |
2432ce9b | 105 | #define JOURNAL_DISK __cpu_to_le32(0x2000000) /* Device marked as Journaling Drive */ |
cdddbdbc | 106 | __u32 status; /* 0xF0 - 0xF3 */ |
1011e834 | 107 | __u32 owner_cfg_num; /* which config 0,1,2... owns this disk */ |
5551b113 CA |
108 | __u32 total_blocks_hi; /* 0xF4 - 0xF5 total blocks hi */ |
109 | #define IMSM_DISK_FILLERS 3 | |
110 | __u32 filler[IMSM_DISK_FILLERS]; /* 0xF5 - 0x107 MPB_DISK_FILLERS for future expansion */ | |
cdddbdbc DW |
111 | }; |
112 | ||
3b451610 AK |
113 | /* map selector for map managment |
114 | */ | |
238c0a71 AK |
115 | #define MAP_0 0 |
116 | #define MAP_1 1 | |
117 | #define MAP_X -1 | |
3b451610 | 118 | |
cdddbdbc DW |
119 | /* RAID map configuration infos. */ |
120 | struct imsm_map { | |
5551b113 CA |
121 | __u32 pba_of_lba0_lo; /* start address of partition */ |
122 | __u32 blocks_per_member_lo;/* blocks per member */ | |
123 | __u32 num_data_stripes_lo; /* number of data stripes */ | |
cdddbdbc DW |
124 | __u16 blocks_per_strip; |
125 | __u8 map_state; /* Normal, Uninitialized, Degraded, Failed */ | |
126 | #define IMSM_T_STATE_NORMAL 0 | |
127 | #define IMSM_T_STATE_UNINITIALIZED 1 | |
e3bba0e0 DW |
128 | #define IMSM_T_STATE_DEGRADED 2 |
129 | #define IMSM_T_STATE_FAILED 3 | |
cdddbdbc DW |
130 | __u8 raid_level; |
131 | #define IMSM_T_RAID0 0 | |
132 | #define IMSM_T_RAID1 1 | |
133 | #define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */ | |
134 | __u8 num_members; /* number of member disks */ | |
fe7ed8cb DW |
135 | __u8 num_domains; /* number of parity domains */ |
136 | __u8 failed_disk_num; /* valid only when state is degraded */ | |
252d23c0 | 137 | __u8 ddf; |
5551b113 CA |
138 | __u32 pba_of_lba0_hi; |
139 | __u32 blocks_per_member_hi; | |
140 | __u32 num_data_stripes_hi; | |
141 | __u32 filler[4]; /* expansion area */ | |
7eef0453 | 142 | #define IMSM_ORD_REBUILD (1 << 24) |
cdddbdbc | 143 | __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members], |
7eef0453 DW |
144 | * top byte contains some flags |
145 | */ | |
cdddbdbc DW |
146 | } __attribute__ ((packed)); |
147 | ||
148 | struct imsm_vol { | |
f8f603f1 | 149 | __u32 curr_migr_unit; |
fe7ed8cb | 150 | __u32 checkpoint_id; /* id to access curr_migr_unit */ |
cdddbdbc | 151 | __u8 migr_state; /* Normal or Migrating */ |
e3bba0e0 DW |
152 | #define MIGR_INIT 0 |
153 | #define MIGR_REBUILD 1 | |
154 | #define MIGR_VERIFY 2 /* analagous to echo check > sync_action */ | |
155 | #define MIGR_GEN_MIGR 3 | |
156 | #define MIGR_STATE_CHANGE 4 | |
1484e727 | 157 | #define MIGR_REPAIR 5 |
cdddbdbc | 158 | __u8 migr_type; /* Initializing, Rebuilding, ... */ |
2432ce9b AP |
159 | #define RAIDVOL_CLEAN 0 |
160 | #define RAIDVOL_DIRTY 1 | |
161 | #define RAIDVOL_DSRECORD_VALID 2 | |
cdddbdbc | 162 | __u8 dirty; |
fe7ed8cb DW |
163 | __u8 fs_state; /* fast-sync state for CnG (0xff == disabled) */ |
164 | __u16 verify_errors; /* number of mismatches */ | |
165 | __u16 bad_blocks; /* number of bad blocks during verify */ | |
166 | __u32 filler[4]; | |
cdddbdbc DW |
167 | struct imsm_map map[1]; |
168 | /* here comes another one if migr_state */ | |
169 | } __attribute__ ((packed)); | |
170 | ||
171 | struct imsm_dev { | |
fe7ed8cb | 172 | __u8 volume[MAX_RAID_SERIAL_LEN]; |
cdddbdbc DW |
173 | __u32 size_low; |
174 | __u32 size_high; | |
fe7ed8cb DW |
175 | #define DEV_BOOTABLE __cpu_to_le32(0x01) |
176 | #define DEV_BOOT_DEVICE __cpu_to_le32(0x02) | |
177 | #define DEV_READ_COALESCING __cpu_to_le32(0x04) | |
178 | #define DEV_WRITE_COALESCING __cpu_to_le32(0x08) | |
179 | #define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10) | |
180 | #define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20) | |
181 | #define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40) | |
182 | #define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80) | |
183 | #define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100) | |
184 | #define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200) | |
185 | #define DEV_CLONE_N_GO __cpu_to_le32(0x400) | |
186 | #define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800) | |
187 | #define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000) | |
cdddbdbc DW |
188 | __u32 status; /* Persistent RaidDev status */ |
189 | __u32 reserved_blocks; /* Reserved blocks at beginning of volume */ | |
fe7ed8cb DW |
190 | __u8 migr_priority; |
191 | __u8 num_sub_vols; | |
192 | __u8 tid; | |
193 | __u8 cng_master_disk; | |
194 | __u16 cache_policy; | |
195 | __u8 cng_state; | |
196 | __u8 cng_sub_state; | |
2432ce9b AP |
197 | __u16 my_vol_raid_dev_num; /* Used in Unique volume Id for this RaidDev */ |
198 | ||
199 | /* NVM_EN */ | |
200 | __u8 nv_cache_mode; | |
201 | __u8 nv_cache_flags; | |
202 | ||
203 | /* Unique Volume Id of the NvCache Volume associated with this volume */ | |
204 | __u32 nvc_vol_orig_family_num; | |
205 | __u16 nvc_vol_raid_dev_num; | |
206 | ||
207 | #define RWH_OFF 0 | |
208 | #define RWH_DISTRIBUTED 1 | |
209 | #define RWH_JOURNALING_DRIVE 2 | |
210 | __u8 rwh_policy; /* Raid Write Hole Policy */ | |
211 | __u8 jd_serial[MAX_RAID_SERIAL_LEN]; /* Journal Drive serial number */ | |
212 | __u8 filler1; | |
213 | ||
214 | #define IMSM_DEV_FILLERS 3 | |
cdddbdbc DW |
215 | __u32 filler[IMSM_DEV_FILLERS]; |
216 | struct imsm_vol vol; | |
217 | } __attribute__ ((packed)); | |
218 | ||
219 | struct imsm_super { | |
220 | __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */ | |
221 | __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */ | |
222 | __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */ | |
223 | __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */ | |
224 | __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */ | |
604b746f JD |
225 | __u32 error_log_size; /* 0x30 - 0x33 in bytes */ |
226 | __u32 attributes; /* 0x34 - 0x37 */ | |
cdddbdbc DW |
227 | __u8 num_disks; /* 0x38 Number of configured disks */ |
228 | __u8 num_raid_devs; /* 0x39 Number of configured volumes */ | |
604b746f JD |
229 | __u8 error_log_pos; /* 0x3A */ |
230 | __u8 fill[1]; /* 0x3B */ | |
231 | __u32 cache_size; /* 0x3c - 0x40 in mb */ | |
232 | __u32 orig_family_num; /* 0x40 - 0x43 original family num */ | |
233 | __u32 pwr_cycle_count; /* 0x44 - 0x47 simulated power cycle count for array */ | |
234 | __u32 bbm_log_size; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */ | |
2a24dc1b PB |
235 | __u16 num_raid_devs_created; /* 0x4C - 0x4D Used for generating unique |
236 | * volume IDs for raid_dev created in this array | |
237 | * (starts at 1) | |
238 | */ | |
239 | __u16 filler1; /* 0x4E - 0x4F */ | |
240 | #define IMSM_FILLERS 34 | |
241 | __u32 filler[IMSM_FILLERS]; /* 0x50 - 0xD7 RAID_MPB_FILLERS */ | |
cdddbdbc DW |
242 | struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */ |
243 | /* here comes imsm_dev[num_raid_devs] */ | |
604b746f | 244 | /* here comes BBM logs */ |
cdddbdbc DW |
245 | } __attribute__ ((packed)); |
246 | ||
604b746f | 247 | #define BBM_LOG_MAX_ENTRIES 254 |
8d67477f TM |
248 | #define BBM_LOG_MAX_LBA_ENTRY_VAL 256 /* Represents 256 LBAs */ |
249 | #define BBM_LOG_SIGNATURE 0xabadb10c | |
250 | ||
251 | struct bbm_log_block_addr { | |
252 | __u16 w1; | |
253 | __u32 dw1; | |
254 | } __attribute__ ((__packed__)); | |
604b746f JD |
255 | |
256 | struct bbm_log_entry { | |
8d67477f TM |
257 | __u8 marked_count; /* Number of blocks marked - 1 */ |
258 | __u8 disk_ordinal; /* Disk entry within the imsm_super */ | |
259 | struct bbm_log_block_addr defective_block_start; | |
604b746f JD |
260 | } __attribute__ ((__packed__)); |
261 | ||
262 | struct bbm_log { | |
263 | __u32 signature; /* 0xABADB10C */ | |
264 | __u32 entry_count; | |
8d67477f | 265 | struct bbm_log_entry marked_block_entries[BBM_LOG_MAX_ENTRIES]; |
604b746f JD |
266 | } __attribute__ ((__packed__)); |
267 | ||
cdddbdbc | 268 | static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" }; |
cdddbdbc | 269 | |
b53bfba6 TM |
270 | #define BLOCKS_PER_KB (1024/512) |
271 | ||
8e59f3d8 AK |
272 | #define RAID_DISK_RESERVED_BLOCKS_IMSM_HI 2209 |
273 | ||
274 | #define GEN_MIGR_AREA_SIZE 2048 /* General Migration Copy Area size in blocks */ | |
275 | ||
de44e46f PB |
276 | #define MIGR_REC_BUF_SECTORS 1 /* size of migr_record i/o buffer in sectors */ |
277 | #define MIGR_REC_SECTOR_POSITION 1 /* migr_record position offset on disk, | |
278 | * MIGR_REC_BUF_SECTORS <= MIGR_REC_SECTOR_POS | |
17a4eaf9 AK |
279 | */ |
280 | ||
8e59f3d8 AK |
281 | #define UNIT_SRC_NORMAL 0 /* Source data for curr_migr_unit must |
282 | * be recovered using srcMap */ | |
283 | #define UNIT_SRC_IN_CP_AREA 1 /* Source data for curr_migr_unit has | |
284 | * already been migrated and must | |
285 | * be recovered from checkpoint area */ | |
2432ce9b AP |
286 | |
287 | #define PPL_ENTRY_SPACE (128 * 1024) /* Size of the PPL, without the header */ | |
288 | ||
8e59f3d8 AK |
289 | struct migr_record { |
290 | __u32 rec_status; /* Status used to determine how to restart | |
291 | * migration in case it aborts | |
292 | * in some fashion */ | |
293 | __u32 curr_migr_unit; /* 0..numMigrUnits-1 */ | |
294 | __u32 family_num; /* Family number of MPB | |
295 | * containing the RaidDev | |
296 | * that is migrating */ | |
297 | __u32 ascending_migr; /* True if migrating in increasing | |
298 | * order of lbas */ | |
299 | __u32 blocks_per_unit; /* Num disk blocks per unit of operation */ | |
300 | __u32 dest_depth_per_unit; /* Num member blocks each destMap | |
301 | * member disk | |
302 | * advances per unit-of-operation */ | |
303 | __u32 ckpt_area_pba; /* Pba of first block of ckpt copy area */ | |
304 | __u32 dest_1st_member_lba; /* First member lba on first | |
305 | * stripe of destination */ | |
306 | __u32 num_migr_units; /* Total num migration units-of-op */ | |
307 | __u32 post_migr_vol_cap; /* Size of volume after | |
308 | * migration completes */ | |
309 | __u32 post_migr_vol_cap_hi; /* Expansion space for LBA64 */ | |
310 | __u32 ckpt_read_disk_num; /* Which member disk in destSubMap[0] the | |
311 | * migration ckpt record was read from | |
312 | * (for recovered migrations) */ | |
313 | } __attribute__ ((__packed__)); | |
314 | ||
ec50f7b6 LM |
315 | struct md_list { |
316 | /* usage marker: | |
317 | * 1: load metadata | |
318 | * 2: metadata does not match | |
319 | * 4: already checked | |
320 | */ | |
321 | int used; | |
322 | char *devname; | |
323 | int found; | |
324 | int container; | |
325 | dev_t st_rdev; | |
326 | struct md_list *next; | |
327 | }; | |
328 | ||
e7b84f9d | 329 | #define pr_vrb(fmt, arg...) (void) (verbose && pr_err(fmt, ##arg)) |
ec50f7b6 | 330 | |
1484e727 DW |
331 | static __u8 migr_type(struct imsm_dev *dev) |
332 | { | |
333 | if (dev->vol.migr_type == MIGR_VERIFY && | |
334 | dev->status & DEV_VERIFY_AND_FIX) | |
335 | return MIGR_REPAIR; | |
336 | else | |
337 | return dev->vol.migr_type; | |
338 | } | |
339 | ||
340 | static void set_migr_type(struct imsm_dev *dev, __u8 migr_type) | |
341 | { | |
342 | /* for compatibility with older oroms convert MIGR_REPAIR, into | |
343 | * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status | |
344 | */ | |
345 | if (migr_type == MIGR_REPAIR) { | |
346 | dev->vol.migr_type = MIGR_VERIFY; | |
347 | dev->status |= DEV_VERIFY_AND_FIX; | |
348 | } else { | |
349 | dev->vol.migr_type = migr_type; | |
350 | dev->status &= ~DEV_VERIFY_AND_FIX; | |
351 | } | |
352 | } | |
353 | ||
f36a9ecd | 354 | static unsigned int sector_count(__u32 bytes, unsigned int sector_size) |
cdddbdbc | 355 | { |
f36a9ecd | 356 | return ROUND_UP(bytes, sector_size) / sector_size; |
87eb16df | 357 | } |
cdddbdbc | 358 | |
f36a9ecd PB |
359 | static unsigned int mpb_sectors(struct imsm_super *mpb, |
360 | unsigned int sector_size) | |
87eb16df | 361 | { |
f36a9ecd | 362 | return sector_count(__le32_to_cpu(mpb->mpb_size), sector_size); |
cdddbdbc DW |
363 | } |
364 | ||
ba2de7ba DW |
365 | struct intel_dev { |
366 | struct imsm_dev *dev; | |
367 | struct intel_dev *next; | |
f21e18ca | 368 | unsigned index; |
ba2de7ba DW |
369 | }; |
370 | ||
88654014 LM |
371 | struct intel_hba { |
372 | enum sys_dev_type type; | |
373 | char *path; | |
374 | char *pci_id; | |
375 | struct intel_hba *next; | |
376 | }; | |
377 | ||
1a64be56 LM |
378 | enum action { |
379 | DISK_REMOVE = 1, | |
380 | DISK_ADD | |
381 | }; | |
cdddbdbc DW |
382 | /* internal representation of IMSM metadata */ |
383 | struct intel_super { | |
384 | union { | |
949c47a0 DW |
385 | void *buf; /* O_DIRECT buffer for reading/writing metadata */ |
386 | struct imsm_super *anchor; /* immovable parameters */ | |
cdddbdbc | 387 | }; |
8e59f3d8 AK |
388 | union { |
389 | void *migr_rec_buf; /* buffer for I/O operations */ | |
390 | struct migr_record *migr_rec; /* migration record */ | |
391 | }; | |
51d83f5d AK |
392 | int clean_migration_record_by_mdmon; /* when reshape is switched to next |
393 | array, it indicates that mdmon is allowed to clean migration | |
394 | record */ | |
949c47a0 | 395 | size_t len; /* size of the 'buf' allocation */ |
bbab0940 | 396 | size_t extra_space; /* extra space in 'buf' that is not used yet */ |
4d7b1503 DW |
397 | void *next_buf; /* for realloc'ing buf from the manager */ |
398 | size_t next_len; | |
c2c087e6 | 399 | int updates_pending; /* count of pending updates for mdmon */ |
bf5a934a | 400 | int current_vol; /* index of raid device undergoing creation */ |
5551b113 | 401 | unsigned long long create_offset; /* common start for 'current_vol' */ |
148acb7b | 402 | __u32 random; /* random data for seeding new family numbers */ |
ba2de7ba | 403 | struct intel_dev *devlist; |
fa7bb6f8 | 404 | unsigned int sector_size; /* sector size of used member drives */ |
cdddbdbc DW |
405 | struct dl { |
406 | struct dl *next; | |
407 | int index; | |
408 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
409 | int major, minor; | |
410 | char *devname; | |
b9f594fe | 411 | struct imsm_disk disk; |
cdddbdbc | 412 | int fd; |
0dcecb2e DW |
413 | int extent_cnt; |
414 | struct extent *e; /* for determining freespace @ create */ | |
efb30e7f | 415 | int raiddisk; /* slot to fill in autolayout */ |
1a64be56 | 416 | enum action action; |
ca0748fa | 417 | } *disks, *current_disk; |
1a64be56 LM |
418 | struct dl *disk_mgmt_list; /* list of disks to add/remove while mdmon |
419 | active */ | |
47ee5a45 | 420 | struct dl *missing; /* disks removed while we weren't looking */ |
43dad3d6 | 421 | struct bbm_log *bbm_log; |
88654014 | 422 | struct intel_hba *hba; /* device path of the raid controller for this metadata */ |
88c32bb1 | 423 | const struct imsm_orom *orom; /* platform firmware support */ |
a2b97981 | 424 | struct intel_super *next; /* (temp) list for disambiguating family_num */ |
928f1424 | 425 | struct md_bb bb; /* memory for get_bad_blocks call */ |
a2b97981 DW |
426 | }; |
427 | ||
428 | struct intel_disk { | |
429 | struct imsm_disk disk; | |
430 | #define IMSM_UNKNOWN_OWNER (-1) | |
431 | int owner; | |
432 | struct intel_disk *next; | |
cdddbdbc DW |
433 | }; |
434 | ||
c2c087e6 DW |
435 | struct extent { |
436 | unsigned long long start, size; | |
437 | }; | |
438 | ||
694575e7 KW |
439 | /* definitions of reshape process types */ |
440 | enum imsm_reshape_type { | |
441 | CH_TAKEOVER, | |
b5347799 | 442 | CH_MIGRATION, |
7abc9871 | 443 | CH_ARRAY_SIZE, |
694575e7 KW |
444 | }; |
445 | ||
88758e9d DW |
446 | /* definition of messages passed to imsm_process_update */ |
447 | enum imsm_update_type { | |
448 | update_activate_spare, | |
8273f55e | 449 | update_create_array, |
33414a01 | 450 | update_kill_array, |
aa534678 | 451 | update_rename_array, |
1a64be56 | 452 | update_add_remove_disk, |
78b10e66 | 453 | update_reshape_container_disks, |
48c5303a | 454 | update_reshape_migration, |
2d40f3a1 AK |
455 | update_takeover, |
456 | update_general_migration_checkpoint, | |
f3871fdc | 457 | update_size_change, |
bbab0940 | 458 | update_prealloc_badblocks_mem, |
e6e9dd3f | 459 | update_rwh_policy, |
88758e9d DW |
460 | }; |
461 | ||
462 | struct imsm_update_activate_spare { | |
463 | enum imsm_update_type type; | |
d23fe947 | 464 | struct dl *dl; |
88758e9d DW |
465 | int slot; |
466 | int array; | |
467 | struct imsm_update_activate_spare *next; | |
468 | }; | |
469 | ||
78b10e66 | 470 | struct geo_params { |
4dd2df09 | 471 | char devnm[32]; |
78b10e66 | 472 | char *dev_name; |
d04f65f4 | 473 | unsigned long long size; |
78b10e66 N |
474 | int level; |
475 | int layout; | |
476 | int chunksize; | |
477 | int raid_disks; | |
478 | }; | |
479 | ||
bb025c2f KW |
480 | enum takeover_direction { |
481 | R10_TO_R0, | |
482 | R0_TO_R10 | |
483 | }; | |
484 | struct imsm_update_takeover { | |
485 | enum imsm_update_type type; | |
486 | int subarray; | |
487 | enum takeover_direction direction; | |
488 | }; | |
78b10e66 N |
489 | |
490 | struct imsm_update_reshape { | |
491 | enum imsm_update_type type; | |
492 | int old_raid_disks; | |
493 | int new_raid_disks; | |
48c5303a PC |
494 | |
495 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ | |
496 | }; | |
497 | ||
498 | struct imsm_update_reshape_migration { | |
499 | enum imsm_update_type type; | |
500 | int old_raid_disks; | |
501 | int new_raid_disks; | |
502 | /* fields for array migration changes | |
503 | */ | |
504 | int subdev; | |
505 | int new_level; | |
506 | int new_layout; | |
4bba0439 | 507 | int new_chunksize; |
48c5303a | 508 | |
d195167d | 509 | int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */ |
78b10e66 N |
510 | }; |
511 | ||
f3871fdc AK |
512 | struct imsm_update_size_change { |
513 | enum imsm_update_type type; | |
514 | int subdev; | |
515 | long long new_size; | |
516 | }; | |
517 | ||
2d40f3a1 AK |
518 | struct imsm_update_general_migration_checkpoint { |
519 | enum imsm_update_type type; | |
520 | __u32 curr_migr_unit; | |
521 | }; | |
522 | ||
54c2c1ea DW |
523 | struct disk_info { |
524 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
525 | }; | |
526 | ||
8273f55e DW |
527 | struct imsm_update_create_array { |
528 | enum imsm_update_type type; | |
8273f55e | 529 | int dev_idx; |
6a3e913e | 530 | struct imsm_dev dev; |
8273f55e DW |
531 | }; |
532 | ||
33414a01 DW |
533 | struct imsm_update_kill_array { |
534 | enum imsm_update_type type; | |
535 | int dev_idx; | |
536 | }; | |
537 | ||
aa534678 DW |
538 | struct imsm_update_rename_array { |
539 | enum imsm_update_type type; | |
540 | __u8 name[MAX_RAID_SERIAL_LEN]; | |
541 | int dev_idx; | |
542 | }; | |
543 | ||
1a64be56 | 544 | struct imsm_update_add_remove_disk { |
43dad3d6 DW |
545 | enum imsm_update_type type; |
546 | }; | |
547 | ||
bbab0940 TM |
548 | struct imsm_update_prealloc_bb_mem { |
549 | enum imsm_update_type type; | |
550 | }; | |
551 | ||
e6e9dd3f AP |
552 | struct imsm_update_rwh_policy { |
553 | enum imsm_update_type type; | |
554 | int new_policy; | |
555 | int dev_idx; | |
556 | }; | |
557 | ||
88654014 LM |
558 | static const char *_sys_dev_type[] = { |
559 | [SYS_DEV_UNKNOWN] = "Unknown", | |
560 | [SYS_DEV_SAS] = "SAS", | |
614902f6 | 561 | [SYS_DEV_SATA] = "SATA", |
60f0f54d PB |
562 | [SYS_DEV_NVME] = "NVMe", |
563 | [SYS_DEV_VMD] = "VMD" | |
88654014 LM |
564 | }; |
565 | ||
566 | const char *get_sys_dev_type(enum sys_dev_type type) | |
567 | { | |
568 | if (type >= SYS_DEV_MAX) | |
569 | type = SYS_DEV_UNKNOWN; | |
570 | ||
571 | return _sys_dev_type[type]; | |
572 | } | |
573 | ||
574 | static struct intel_hba * alloc_intel_hba(struct sys_dev *device) | |
575 | { | |
503975b9 N |
576 | struct intel_hba *result = xmalloc(sizeof(*result)); |
577 | ||
578 | result->type = device->type; | |
579 | result->path = xstrdup(device->path); | |
580 | result->next = NULL; | |
581 | if (result->path && (result->pci_id = strrchr(result->path, '/')) != NULL) | |
582 | result->pci_id++; | |
583 | ||
88654014 LM |
584 | return result; |
585 | } | |
586 | ||
587 | static struct intel_hba * find_intel_hba(struct intel_hba *hba, struct sys_dev *device) | |
588 | { | |
594dc1b8 JS |
589 | struct intel_hba *result; |
590 | ||
88654014 LM |
591 | for (result = hba; result; result = result->next) { |
592 | if (result->type == device->type && strcmp(result->path, device->path) == 0) | |
593 | break; | |
594 | } | |
595 | return result; | |
596 | } | |
597 | ||
b4cf4cba | 598 | static int attach_hba_to_super(struct intel_super *super, struct sys_dev *device) |
88654014 LM |
599 | { |
600 | struct intel_hba *hba; | |
601 | ||
602 | /* check if disk attached to Intel HBA */ | |
603 | hba = find_intel_hba(super->hba, device); | |
604 | if (hba != NULL) | |
605 | return 1; | |
606 | /* Check if HBA is already attached to super */ | |
607 | if (super->hba == NULL) { | |
608 | super->hba = alloc_intel_hba(device); | |
609 | return 1; | |
6b781d33 AP |
610 | } |
611 | ||
612 | hba = super->hba; | |
613 | /* Intel metadata allows for all disks attached to the same type HBA. | |
614902f6 | 614 | * Do not support HBA types mixing |
6b781d33 AP |
615 | */ |
616 | if (device->type != hba->type) | |
88654014 | 617 | return 2; |
6b781d33 AP |
618 | |
619 | /* Multiple same type HBAs can be used if they share the same OROM */ | |
620 | const struct imsm_orom *device_orom = get_orom_by_device_id(device->dev_id); | |
621 | ||
622 | if (device_orom != super->orom) | |
623 | return 2; | |
624 | ||
625 | while (hba->next) | |
626 | hba = hba->next; | |
627 | ||
628 | hba->next = alloc_intel_hba(device); | |
629 | return 1; | |
88654014 LM |
630 | } |
631 | ||
632 | static struct sys_dev* find_disk_attached_hba(int fd, const char *devname) | |
633 | { | |
9bc4ae77 | 634 | struct sys_dev *list, *elem; |
88654014 LM |
635 | char *disk_path; |
636 | ||
637 | if ((list = find_intel_devices()) == NULL) | |
638 | return 0; | |
639 | ||
640 | if (fd < 0) | |
641 | disk_path = (char *) devname; | |
642 | else | |
643 | disk_path = diskfd_to_devpath(fd); | |
644 | ||
9bc4ae77 | 645 | if (!disk_path) |
88654014 | 646 | return 0; |
88654014 | 647 | |
9bc4ae77 N |
648 | for (elem = list; elem; elem = elem->next) |
649 | if (path_attached_to_hba(disk_path, elem->path)) | |
88654014 | 650 | return elem; |
9bc4ae77 | 651 | |
88654014 LM |
652 | if (disk_path != devname) |
653 | free(disk_path); | |
88654014 LM |
654 | |
655 | return NULL; | |
656 | } | |
657 | ||
d424212e N |
658 | static int find_intel_hba_capability(int fd, struct intel_super *super, |
659 | char *devname); | |
f2f5c343 | 660 | |
cdddbdbc DW |
661 | static struct supertype *match_metadata_desc_imsm(char *arg) |
662 | { | |
663 | struct supertype *st; | |
664 | ||
665 | if (strcmp(arg, "imsm") != 0 && | |
666 | strcmp(arg, "default") != 0 | |
667 | ) | |
668 | return NULL; | |
669 | ||
503975b9 | 670 | st = xcalloc(1, sizeof(*st)); |
cdddbdbc DW |
671 | st->ss = &super_imsm; |
672 | st->max_devs = IMSM_MAX_DEVICES; | |
673 | st->minor_version = 0; | |
674 | st->sb = NULL; | |
675 | return st; | |
676 | } | |
677 | ||
cdddbdbc DW |
678 | static __u8 *get_imsm_version(struct imsm_super *mpb) |
679 | { | |
680 | return &mpb->sig[MPB_SIG_LEN]; | |
681 | } | |
682 | ||
949c47a0 DW |
683 | /* retrieve a disk directly from the anchor when the anchor is known to be |
684 | * up-to-date, currently only at load time | |
685 | */ | |
686 | static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index) | |
cdddbdbc | 687 | { |
949c47a0 | 688 | if (index >= mpb->num_disks) |
cdddbdbc DW |
689 | return NULL; |
690 | return &mpb->disk[index]; | |
691 | } | |
692 | ||
95d07a2c LM |
693 | /* retrieve the disk description based on a index of the disk |
694 | * in the sub-array | |
695 | */ | |
696 | static struct dl *get_imsm_dl_disk(struct intel_super *super, __u8 index) | |
949c47a0 | 697 | { |
b9f594fe DW |
698 | struct dl *d; |
699 | ||
700 | for (d = super->disks; d; d = d->next) | |
701 | if (d->index == index) | |
95d07a2c LM |
702 | return d; |
703 | ||
704 | return NULL; | |
705 | } | |
706 | /* retrieve a disk from the parsed metadata */ | |
707 | static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index) | |
708 | { | |
709 | struct dl *dl; | |
710 | ||
711 | dl = get_imsm_dl_disk(super, index); | |
712 | if (dl) | |
713 | return &dl->disk; | |
714 | ||
b9f594fe | 715 | return NULL; |
949c47a0 DW |
716 | } |
717 | ||
718 | /* generate a checksum directly from the anchor when the anchor is known to be | |
719 | * up-to-date, currently only at load or write_super after coalescing | |
720 | */ | |
721 | static __u32 __gen_imsm_checksum(struct imsm_super *mpb) | |
cdddbdbc DW |
722 | { |
723 | __u32 end = mpb->mpb_size / sizeof(end); | |
724 | __u32 *p = (__u32 *) mpb; | |
725 | __u32 sum = 0; | |
726 | ||
5d500228 N |
727 | while (end--) { |
728 | sum += __le32_to_cpu(*p); | |
97f734fd N |
729 | p++; |
730 | } | |
cdddbdbc | 731 | |
5d500228 | 732 | return sum - __le32_to_cpu(mpb->check_sum); |
cdddbdbc DW |
733 | } |
734 | ||
a965f303 DW |
735 | static size_t sizeof_imsm_map(struct imsm_map *map) |
736 | { | |
737 | return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1); | |
738 | } | |
739 | ||
740 | struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map) | |
cdddbdbc | 741 | { |
5e7b0330 AK |
742 | /* A device can have 2 maps if it is in the middle of a migration. |
743 | * If second_map is: | |
238c0a71 AK |
744 | * MAP_0 - we return the first map |
745 | * MAP_1 - we return the second map if it exists, else NULL | |
746 | * MAP_X - we return the second map if it exists, else the first | |
5e7b0330 | 747 | */ |
a965f303 | 748 | struct imsm_map *map = &dev->vol.map[0]; |
9535fc47 | 749 | struct imsm_map *map2 = NULL; |
a965f303 | 750 | |
9535fc47 AK |
751 | if (dev->vol.migr_state) |
752 | map2 = (void *)map + sizeof_imsm_map(map); | |
a965f303 | 753 | |
9535fc47 | 754 | switch (second_map) { |
3b451610 | 755 | case MAP_0: |
9535fc47 | 756 | break; |
3b451610 | 757 | case MAP_1: |
9535fc47 AK |
758 | map = map2; |
759 | break; | |
238c0a71 | 760 | case MAP_X: |
9535fc47 AK |
761 | if (map2) |
762 | map = map2; | |
763 | break; | |
9535fc47 AK |
764 | default: |
765 | map = NULL; | |
766 | } | |
767 | return map; | |
5e7b0330 | 768 | |
a965f303 | 769 | } |
cdddbdbc | 770 | |
3393c6af DW |
771 | /* return the size of the device. |
772 | * migr_state increases the returned size if map[0] were to be duplicated | |
773 | */ | |
774 | static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state) | |
a965f303 DW |
775 | { |
776 | size_t size = sizeof(*dev) - sizeof(struct imsm_map) + | |
238c0a71 | 777 | sizeof_imsm_map(get_imsm_map(dev, MAP_0)); |
cdddbdbc DW |
778 | |
779 | /* migrating means an additional map */ | |
a965f303 | 780 | if (dev->vol.migr_state) |
238c0a71 | 781 | size += sizeof_imsm_map(get_imsm_map(dev, MAP_1)); |
3393c6af | 782 | else if (migr_state) |
238c0a71 | 783 | size += sizeof_imsm_map(get_imsm_map(dev, MAP_0)); |
cdddbdbc DW |
784 | |
785 | return size; | |
786 | } | |
787 | ||
54c2c1ea DW |
788 | /* retrieve disk serial number list from a metadata update */ |
789 | static struct disk_info *get_disk_info(struct imsm_update_create_array *update) | |
790 | { | |
791 | void *u = update; | |
792 | struct disk_info *inf; | |
793 | ||
794 | inf = u + sizeof(*update) - sizeof(struct imsm_dev) + | |
795 | sizeof_imsm_dev(&update->dev, 0); | |
796 | ||
797 | return inf; | |
798 | } | |
54c2c1ea | 799 | |
949c47a0 | 800 | static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index) |
cdddbdbc DW |
801 | { |
802 | int offset; | |
803 | int i; | |
804 | void *_mpb = mpb; | |
805 | ||
949c47a0 | 806 | if (index >= mpb->num_raid_devs) |
cdddbdbc DW |
807 | return NULL; |
808 | ||
809 | /* devices start after all disks */ | |
810 | offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb; | |
811 | ||
812 | for (i = 0; i <= index; i++) | |
813 | if (i == index) | |
814 | return _mpb + offset; | |
815 | else | |
3393c6af | 816 | offset += sizeof_imsm_dev(_mpb + offset, 0); |
cdddbdbc DW |
817 | |
818 | return NULL; | |
819 | } | |
820 | ||
949c47a0 DW |
821 | static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index) |
822 | { | |
ba2de7ba DW |
823 | struct intel_dev *dv; |
824 | ||
949c47a0 DW |
825 | if (index >= super->anchor->num_raid_devs) |
826 | return NULL; | |
ba2de7ba DW |
827 | for (dv = super->devlist; dv; dv = dv->next) |
828 | if (dv->index == index) | |
829 | return dv->dev; | |
830 | return NULL; | |
949c47a0 DW |
831 | } |
832 | ||
8d67477f TM |
833 | static inline unsigned long long __le48_to_cpu(const struct bbm_log_block_addr |
834 | *addr) | |
835 | { | |
836 | return ((((__u64)__le32_to_cpu(addr->dw1)) << 16) | | |
837 | __le16_to_cpu(addr->w1)); | |
838 | } | |
839 | ||
840 | static inline struct bbm_log_block_addr __cpu_to_le48(unsigned long long sec) | |
841 | { | |
842 | struct bbm_log_block_addr addr; | |
843 | ||
844 | addr.w1 = __cpu_to_le16((__u16)(sec & 0xffff)); | |
845 | addr.dw1 = __cpu_to_le32((__u32)(sec >> 16) & 0xffffffff); | |
846 | return addr; | |
847 | } | |
848 | ||
8d67477f TM |
849 | /* get size of the bbm log */ |
850 | static __u32 get_imsm_bbm_log_size(struct bbm_log *log) | |
851 | { | |
852 | if (!log || log->entry_count == 0) | |
853 | return 0; | |
854 | ||
855 | return sizeof(log->signature) + | |
856 | sizeof(log->entry_count) + | |
857 | log->entry_count * sizeof(struct bbm_log_entry); | |
858 | } | |
6f50473f TM |
859 | |
860 | /* check if bad block is not partially stored in bbm log */ | |
861 | static int is_stored_in_bbm(struct bbm_log *log, const __u8 idx, const unsigned | |
862 | long long sector, const int length, __u32 *pos) | |
863 | { | |
864 | __u32 i; | |
865 | ||
866 | for (i = *pos; i < log->entry_count; i++) { | |
867 | struct bbm_log_entry *entry = &log->marked_block_entries[i]; | |
868 | unsigned long long bb_start; | |
869 | unsigned long long bb_end; | |
870 | ||
871 | bb_start = __le48_to_cpu(&entry->defective_block_start); | |
872 | bb_end = bb_start + (entry->marked_count + 1); | |
873 | ||
874 | if ((entry->disk_ordinal == idx) && (bb_start >= sector) && | |
875 | (bb_end <= sector + length)) { | |
876 | *pos = i; | |
877 | return 1; | |
878 | } | |
879 | } | |
880 | return 0; | |
881 | } | |
882 | ||
883 | /* record new bad block in bbm log */ | |
884 | static int record_new_badblock(struct bbm_log *log, const __u8 idx, unsigned | |
885 | long long sector, int length) | |
886 | { | |
887 | int new_bb = 0; | |
888 | __u32 pos = 0; | |
889 | struct bbm_log_entry *entry = NULL; | |
890 | ||
891 | while (is_stored_in_bbm(log, idx, sector, length, &pos)) { | |
892 | struct bbm_log_entry *e = &log->marked_block_entries[pos]; | |
893 | ||
894 | if ((e->marked_count + 1 == BBM_LOG_MAX_LBA_ENTRY_VAL) && | |
895 | (__le48_to_cpu(&e->defective_block_start) == sector)) { | |
896 | sector += BBM_LOG_MAX_LBA_ENTRY_VAL; | |
897 | length -= BBM_LOG_MAX_LBA_ENTRY_VAL; | |
898 | pos = pos + 1; | |
899 | continue; | |
900 | } | |
901 | entry = e; | |
902 | break; | |
903 | } | |
904 | ||
905 | if (entry) { | |
906 | int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length : | |
907 | BBM_LOG_MAX_LBA_ENTRY_VAL; | |
908 | entry->defective_block_start = __cpu_to_le48(sector); | |
909 | entry->marked_count = cnt - 1; | |
910 | if (cnt == length) | |
911 | return 1; | |
912 | sector += cnt; | |
913 | length -= cnt; | |
914 | } | |
915 | ||
916 | new_bb = ROUND_UP(length, BBM_LOG_MAX_LBA_ENTRY_VAL) / | |
917 | BBM_LOG_MAX_LBA_ENTRY_VAL; | |
918 | if (log->entry_count + new_bb > BBM_LOG_MAX_ENTRIES) | |
919 | return 0; | |
920 | ||
921 | while (length > 0) { | |
922 | int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length : | |
923 | BBM_LOG_MAX_LBA_ENTRY_VAL; | |
924 | struct bbm_log_entry *entry = | |
925 | &log->marked_block_entries[log->entry_count]; | |
926 | ||
927 | entry->defective_block_start = __cpu_to_le48(sector); | |
928 | entry->marked_count = cnt - 1; | |
929 | entry->disk_ordinal = idx; | |
930 | ||
931 | sector += cnt; | |
932 | length -= cnt; | |
933 | ||
934 | log->entry_count++; | |
935 | } | |
936 | ||
937 | return new_bb; | |
938 | } | |
c07a5a4f | 939 | |
4c9e8c1e TM |
940 | /* clear all bad blocks for given disk */ |
941 | static void clear_disk_badblocks(struct bbm_log *log, const __u8 idx) | |
942 | { | |
943 | __u32 i = 0; | |
944 | ||
945 | while (i < log->entry_count) { | |
946 | struct bbm_log_entry *entries = log->marked_block_entries; | |
947 | ||
948 | if (entries[i].disk_ordinal == idx) { | |
949 | if (i < log->entry_count - 1) | |
950 | entries[i] = entries[log->entry_count - 1]; | |
951 | log->entry_count--; | |
952 | } else { | |
953 | i++; | |
954 | } | |
955 | } | |
956 | } | |
957 | ||
c07a5a4f TM |
958 | /* clear given bad block */ |
959 | static int clear_badblock(struct bbm_log *log, const __u8 idx, const unsigned | |
960 | long long sector, const int length) { | |
961 | __u32 i = 0; | |
962 | ||
963 | while (i < log->entry_count) { | |
964 | struct bbm_log_entry *entries = log->marked_block_entries; | |
965 | ||
966 | if ((entries[i].disk_ordinal == idx) && | |
967 | (__le48_to_cpu(&entries[i].defective_block_start) == | |
968 | sector) && (entries[i].marked_count + 1 == length)) { | |
969 | if (i < log->entry_count - 1) | |
970 | entries[i] = entries[log->entry_count - 1]; | |
971 | log->entry_count--; | |
972 | break; | |
973 | } | |
974 | i++; | |
975 | } | |
976 | ||
977 | return 1; | |
978 | } | |
8d67477f TM |
979 | |
980 | /* allocate and load BBM log from metadata */ | |
981 | static int load_bbm_log(struct intel_super *super) | |
982 | { | |
983 | struct imsm_super *mpb = super->anchor; | |
984 | __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size); | |
985 | ||
986 | super->bbm_log = xcalloc(1, sizeof(struct bbm_log)); | |
987 | if (!super->bbm_log) | |
988 | return 1; | |
989 | ||
990 | if (bbm_log_size) { | |
991 | struct bbm_log *log = (void *)mpb + | |
992 | __le32_to_cpu(mpb->mpb_size) - bbm_log_size; | |
993 | ||
994 | __u32 entry_count; | |
995 | ||
996 | if (bbm_log_size < sizeof(log->signature) + | |
997 | sizeof(log->entry_count)) | |
998 | return 2; | |
999 | ||
1000 | entry_count = __le32_to_cpu(log->entry_count); | |
1001 | if ((__le32_to_cpu(log->signature) != BBM_LOG_SIGNATURE) || | |
1002 | (entry_count > BBM_LOG_MAX_ENTRIES)) | |
1003 | return 3; | |
1004 | ||
1005 | if (bbm_log_size != | |
1006 | sizeof(log->signature) + sizeof(log->entry_count) + | |
1007 | entry_count * sizeof(struct bbm_log_entry)) | |
1008 | return 4; | |
1009 | ||
1010 | memcpy(super->bbm_log, log, bbm_log_size); | |
1011 | } else { | |
1012 | super->bbm_log->signature = __cpu_to_le32(BBM_LOG_SIGNATURE); | |
1013 | super->bbm_log->entry_count = 0; | |
1014 | } | |
1015 | ||
1016 | return 0; | |
1017 | } | |
1018 | ||
b12796be TM |
1019 | /* checks if bad block is within volume boundaries */ |
1020 | static int is_bad_block_in_volume(const struct bbm_log_entry *entry, | |
1021 | const unsigned long long start_sector, | |
1022 | const unsigned long long size) | |
1023 | { | |
1024 | unsigned long long bb_start; | |
1025 | unsigned long long bb_end; | |
1026 | ||
1027 | bb_start = __le48_to_cpu(&entry->defective_block_start); | |
1028 | bb_end = bb_start + (entry->marked_count + 1); | |
1029 | ||
1030 | if (((bb_start >= start_sector) && (bb_start < start_sector + size)) || | |
1031 | ((bb_end >= start_sector) && (bb_end <= start_sector + size))) | |
1032 | return 1; | |
1033 | ||
1034 | return 0; | |
1035 | } | |
1036 | ||
1037 | /* get list of bad blocks on a drive for a volume */ | |
1038 | static void get_volume_badblocks(const struct bbm_log *log, const __u8 idx, | |
1039 | const unsigned long long start_sector, | |
1040 | const unsigned long long size, | |
1041 | struct md_bb *bbs) | |
1042 | { | |
1043 | __u32 count = 0; | |
1044 | __u32 i; | |
1045 | ||
1046 | for (i = 0; i < log->entry_count; i++) { | |
1047 | const struct bbm_log_entry *ent = | |
1048 | &log->marked_block_entries[i]; | |
1049 | struct md_bb_entry *bb; | |
1050 | ||
1051 | if ((ent->disk_ordinal == idx) && | |
1052 | is_bad_block_in_volume(ent, start_sector, size)) { | |
1053 | ||
1054 | if (!bbs->entries) { | |
1055 | bbs->entries = xmalloc(BBM_LOG_MAX_ENTRIES * | |
1056 | sizeof(*bb)); | |
1057 | if (!bbs->entries) | |
1058 | break; | |
1059 | } | |
1060 | ||
1061 | bb = &bbs->entries[count++]; | |
1062 | bb->sector = __le48_to_cpu(&ent->defective_block_start); | |
1063 | bb->length = ent->marked_count + 1; | |
1064 | } | |
1065 | } | |
1066 | bbs->count = count; | |
1067 | } | |
1068 | ||
98130f40 AK |
1069 | /* |
1070 | * for second_map: | |
238c0a71 AK |
1071 | * == MAP_0 get first map |
1072 | * == MAP_1 get second map | |
1073 | * == MAP_X than get map according to the current migr_state | |
98130f40 AK |
1074 | */ |
1075 | static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev, | |
1076 | int slot, | |
1077 | int second_map) | |
7eef0453 DW |
1078 | { |
1079 | struct imsm_map *map; | |
1080 | ||
5e7b0330 | 1081 | map = get_imsm_map(dev, second_map); |
7eef0453 | 1082 | |
ff077194 DW |
1083 | /* top byte identifies disk under rebuild */ |
1084 | return __le32_to_cpu(map->disk_ord_tbl[slot]); | |
1085 | } | |
1086 | ||
1087 | #define ord_to_idx(ord) (((ord) << 8) >> 8) | |
98130f40 | 1088 | static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot, int second_map) |
ff077194 | 1089 | { |
98130f40 | 1090 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, second_map); |
ff077194 DW |
1091 | |
1092 | return ord_to_idx(ord); | |
7eef0453 DW |
1093 | } |
1094 | ||
be73972f DW |
1095 | static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord) |
1096 | { | |
1097 | map->disk_ord_tbl[slot] = __cpu_to_le32(ord); | |
1098 | } | |
1099 | ||
f21e18ca | 1100 | static int get_imsm_disk_slot(struct imsm_map *map, unsigned idx) |
620b1713 DW |
1101 | { |
1102 | int slot; | |
1103 | __u32 ord; | |
1104 | ||
1105 | for (slot = 0; slot < map->num_members; slot++) { | |
1106 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
1107 | if (ord_to_idx(ord) == idx) | |
1108 | return slot; | |
1109 | } | |
1110 | ||
1111 | return -1; | |
1112 | } | |
1113 | ||
cdddbdbc DW |
1114 | static int get_imsm_raid_level(struct imsm_map *map) |
1115 | { | |
1116 | if (map->raid_level == 1) { | |
1117 | if (map->num_members == 2) | |
1118 | return 1; | |
1119 | else | |
1120 | return 10; | |
1121 | } | |
1122 | ||
1123 | return map->raid_level; | |
1124 | } | |
1125 | ||
c2c087e6 DW |
1126 | static int cmp_extent(const void *av, const void *bv) |
1127 | { | |
1128 | const struct extent *a = av; | |
1129 | const struct extent *b = bv; | |
1130 | if (a->start < b->start) | |
1131 | return -1; | |
1132 | if (a->start > b->start) | |
1133 | return 1; | |
1134 | return 0; | |
1135 | } | |
1136 | ||
0dcecb2e | 1137 | static int count_memberships(struct dl *dl, struct intel_super *super) |
c2c087e6 | 1138 | { |
c2c087e6 | 1139 | int memberships = 0; |
620b1713 | 1140 | int i; |
c2c087e6 | 1141 | |
949c47a0 DW |
1142 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
1143 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 1144 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
c2c087e6 | 1145 | |
620b1713 DW |
1146 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
1147 | memberships++; | |
c2c087e6 | 1148 | } |
0dcecb2e DW |
1149 | |
1150 | return memberships; | |
1151 | } | |
1152 | ||
b81221b7 CA |
1153 | static __u32 imsm_min_reserved_sectors(struct intel_super *super); |
1154 | ||
5551b113 CA |
1155 | static int split_ull(unsigned long long n, __u32 *lo, __u32 *hi) |
1156 | { | |
1157 | if (lo == 0 || hi == 0) | |
1158 | return 1; | |
1159 | *lo = __le32_to_cpu((unsigned)n); | |
1160 | *hi = __le32_to_cpu((unsigned)(n >> 32)); | |
1161 | return 0; | |
1162 | } | |
1163 | ||
1164 | static unsigned long long join_u32(__u32 lo, __u32 hi) | |
1165 | { | |
1166 | return (unsigned long long)__le32_to_cpu(lo) | | |
1167 | (((unsigned long long)__le32_to_cpu(hi)) << 32); | |
1168 | } | |
1169 | ||
1170 | static unsigned long long total_blocks(struct imsm_disk *disk) | |
1171 | { | |
1172 | if (disk == NULL) | |
1173 | return 0; | |
1174 | return join_u32(disk->total_blocks_lo, disk->total_blocks_hi); | |
1175 | } | |
1176 | ||
1177 | static unsigned long long pba_of_lba0(struct imsm_map *map) | |
1178 | { | |
1179 | if (map == NULL) | |
1180 | return 0; | |
1181 | return join_u32(map->pba_of_lba0_lo, map->pba_of_lba0_hi); | |
1182 | } | |
1183 | ||
1184 | static unsigned long long blocks_per_member(struct imsm_map *map) | |
1185 | { | |
1186 | if (map == NULL) | |
1187 | return 0; | |
1188 | return join_u32(map->blocks_per_member_lo, map->blocks_per_member_hi); | |
1189 | } | |
1190 | ||
1191 | static unsigned long long num_data_stripes(struct imsm_map *map) | |
1192 | { | |
1193 | if (map == NULL) | |
1194 | return 0; | |
1195 | return join_u32(map->num_data_stripes_lo, map->num_data_stripes_hi); | |
1196 | } | |
1197 | ||
1198 | static void set_total_blocks(struct imsm_disk *disk, unsigned long long n) | |
1199 | { | |
1200 | split_ull(n, &disk->total_blocks_lo, &disk->total_blocks_hi); | |
1201 | } | |
1202 | ||
1203 | static void set_pba_of_lba0(struct imsm_map *map, unsigned long long n) | |
1204 | { | |
1205 | split_ull(n, &map->pba_of_lba0_lo, &map->pba_of_lba0_hi); | |
1206 | } | |
1207 | ||
1208 | static void set_blocks_per_member(struct imsm_map *map, unsigned long long n) | |
1209 | { | |
1210 | split_ull(n, &map->blocks_per_member_lo, &map->blocks_per_member_hi); | |
1211 | } | |
1212 | ||
1213 | static void set_num_data_stripes(struct imsm_map *map, unsigned long long n) | |
1214 | { | |
1215 | split_ull(n, &map->num_data_stripes_lo, &map->num_data_stripes_hi); | |
1216 | } | |
1217 | ||
0dcecb2e DW |
1218 | static struct extent *get_extents(struct intel_super *super, struct dl *dl) |
1219 | { | |
1220 | /* find a list of used extents on the given physical device */ | |
1221 | struct extent *rv, *e; | |
620b1713 | 1222 | int i; |
0dcecb2e | 1223 | int memberships = count_memberships(dl, super); |
b276dd33 DW |
1224 | __u32 reservation; |
1225 | ||
1226 | /* trim the reserved area for spares, so they can join any array | |
1227 | * regardless of whether the OROM has assigned sectors from the | |
1228 | * IMSM_RESERVED_SECTORS region | |
1229 | */ | |
1230 | if (dl->index == -1) | |
b81221b7 | 1231 | reservation = imsm_min_reserved_sectors(super); |
b276dd33 DW |
1232 | else |
1233 | reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
0dcecb2e | 1234 | |
503975b9 | 1235 | rv = xcalloc(sizeof(struct extent), (memberships + 1)); |
c2c087e6 DW |
1236 | e = rv; |
1237 | ||
949c47a0 DW |
1238 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
1239 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 1240 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
c2c087e6 | 1241 | |
620b1713 | 1242 | if (get_imsm_disk_slot(map, dl->index) >= 0) { |
5551b113 CA |
1243 | e->start = pba_of_lba0(map); |
1244 | e->size = blocks_per_member(map); | |
620b1713 | 1245 | e++; |
c2c087e6 DW |
1246 | } |
1247 | } | |
1248 | qsort(rv, memberships, sizeof(*rv), cmp_extent); | |
1249 | ||
1011e834 | 1250 | /* determine the start of the metadata |
14e8215b DW |
1251 | * when no raid devices are defined use the default |
1252 | * ...otherwise allow the metadata to truncate the value | |
1253 | * as is the case with older versions of imsm | |
1254 | */ | |
1255 | if (memberships) { | |
1256 | struct extent *last = &rv[memberships - 1]; | |
5551b113 | 1257 | unsigned long long remainder; |
14e8215b | 1258 | |
5551b113 | 1259 | remainder = total_blocks(&dl->disk) - (last->start + last->size); |
dda5855f DW |
1260 | /* round down to 1k block to satisfy precision of the kernel |
1261 | * 'size' interface | |
1262 | */ | |
1263 | remainder &= ~1UL; | |
1264 | /* make sure remainder is still sane */ | |
f21e18ca | 1265 | if (remainder < (unsigned)ROUND_UP(super->len, 512) >> 9) |
dda5855f | 1266 | remainder = ROUND_UP(super->len, 512) >> 9; |
14e8215b DW |
1267 | if (reservation > remainder) |
1268 | reservation = remainder; | |
1269 | } | |
5551b113 | 1270 | e->start = total_blocks(&dl->disk) - reservation; |
c2c087e6 DW |
1271 | e->size = 0; |
1272 | return rv; | |
1273 | } | |
1274 | ||
14e8215b DW |
1275 | /* try to determine how much space is reserved for metadata from |
1276 | * the last get_extents() entry, otherwise fallback to the | |
1277 | * default | |
1278 | */ | |
1279 | static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl) | |
1280 | { | |
1281 | struct extent *e; | |
1282 | int i; | |
1283 | __u32 rv; | |
1284 | ||
1285 | /* for spares just return a minimal reservation which will grow | |
1286 | * once the spare is picked up by an array | |
1287 | */ | |
1288 | if (dl->index == -1) | |
1289 | return MPB_SECTOR_CNT; | |
1290 | ||
1291 | e = get_extents(super, dl); | |
1292 | if (!e) | |
1293 | return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
1294 | ||
1295 | /* scroll to last entry */ | |
1296 | for (i = 0; e[i].size; i++) | |
1297 | continue; | |
1298 | ||
5551b113 | 1299 | rv = total_blocks(&dl->disk) - e[i].start; |
14e8215b DW |
1300 | |
1301 | free(e); | |
1302 | ||
1303 | return rv; | |
1304 | } | |
1305 | ||
25ed7e59 DW |
1306 | static int is_spare(struct imsm_disk *disk) |
1307 | { | |
1308 | return (disk->status & SPARE_DISK) == SPARE_DISK; | |
1309 | } | |
1310 | ||
1311 | static int is_configured(struct imsm_disk *disk) | |
1312 | { | |
1313 | return (disk->status & CONFIGURED_DISK) == CONFIGURED_DISK; | |
1314 | } | |
1315 | ||
1316 | static int is_failed(struct imsm_disk *disk) | |
1317 | { | |
1318 | return (disk->status & FAILED_DISK) == FAILED_DISK; | |
1319 | } | |
1320 | ||
2432ce9b AP |
1321 | static int is_journal(struct imsm_disk *disk) |
1322 | { | |
1323 | return (disk->status & JOURNAL_DISK) == JOURNAL_DISK; | |
1324 | } | |
1325 | ||
b53bfba6 TM |
1326 | /* round array size down to closest MB and ensure it splits evenly |
1327 | * between members | |
1328 | */ | |
1329 | static unsigned long long round_size_to_mb(unsigned long long size, unsigned int | |
1330 | disk_count) | |
1331 | { | |
1332 | size /= disk_count; | |
1333 | size = (size >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT; | |
1334 | size *= disk_count; | |
1335 | ||
1336 | return size; | |
1337 | } | |
1338 | ||
b81221b7 CA |
1339 | /* try to determine how much space is reserved for metadata from |
1340 | * the last get_extents() entry on the smallest active disk, | |
1341 | * otherwise fallback to the default | |
1342 | */ | |
1343 | static __u32 imsm_min_reserved_sectors(struct intel_super *super) | |
1344 | { | |
1345 | struct extent *e; | |
1346 | int i; | |
5551b113 CA |
1347 | unsigned long long min_active; |
1348 | __u32 remainder; | |
b81221b7 CA |
1349 | __u32 rv = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; |
1350 | struct dl *dl, *dl_min = NULL; | |
1351 | ||
1352 | if (!super) | |
1353 | return rv; | |
1354 | ||
1355 | min_active = 0; | |
1356 | for (dl = super->disks; dl; dl = dl->next) { | |
1357 | if (dl->index < 0) | |
1358 | continue; | |
5551b113 CA |
1359 | unsigned long long blocks = total_blocks(&dl->disk); |
1360 | if (blocks < min_active || min_active == 0) { | |
b81221b7 | 1361 | dl_min = dl; |
5551b113 | 1362 | min_active = blocks; |
b81221b7 CA |
1363 | } |
1364 | } | |
1365 | if (!dl_min) | |
1366 | return rv; | |
1367 | ||
1368 | /* find last lba used by subarrays on the smallest active disk */ | |
1369 | e = get_extents(super, dl_min); | |
1370 | if (!e) | |
1371 | return rv; | |
1372 | for (i = 0; e[i].size; i++) | |
1373 | continue; | |
1374 | ||
1375 | remainder = min_active - e[i].start; | |
1376 | free(e); | |
1377 | ||
1378 | /* to give priority to recovery we should not require full | |
1379 | IMSM_RESERVED_SECTORS from the spare */ | |
1380 | rv = MPB_SECTOR_CNT + NUM_BLOCKS_DIRTY_STRIPE_REGION; | |
1381 | ||
1382 | /* if real reservation is smaller use that value */ | |
1383 | return (remainder < rv) ? remainder : rv; | |
1384 | } | |
1385 | ||
fbfdcb06 AO |
1386 | /* |
1387 | * Return minimum size of a spare and sector size | |
1388 | * that can be used in this array | |
1389 | */ | |
1390 | int get_spare_criteria_imsm(struct supertype *st, struct spare_criteria *c) | |
80e7f8c3 AC |
1391 | { |
1392 | struct intel_super *super = st->sb; | |
1393 | struct dl *dl; | |
1394 | struct extent *e; | |
1395 | int i; | |
fbfdcb06 AO |
1396 | unsigned long long size = 0; |
1397 | ||
1398 | c->min_size = 0; | |
4b57ecf6 | 1399 | c->sector_size = 0; |
80e7f8c3 AC |
1400 | |
1401 | if (!super) | |
fbfdcb06 | 1402 | return -EINVAL; |
80e7f8c3 AC |
1403 | /* find first active disk in array */ |
1404 | dl = super->disks; | |
1405 | while (dl && (is_failed(&dl->disk) || dl->index == -1)) | |
1406 | dl = dl->next; | |
1407 | if (!dl) | |
fbfdcb06 | 1408 | return -EINVAL; |
80e7f8c3 AC |
1409 | /* find last lba used by subarrays */ |
1410 | e = get_extents(super, dl); | |
1411 | if (!e) | |
fbfdcb06 | 1412 | return -EINVAL; |
80e7f8c3 AC |
1413 | for (i = 0; e[i].size; i++) |
1414 | continue; | |
1415 | if (i > 0) | |
fbfdcb06 | 1416 | size = e[i-1].start + e[i-1].size; |
80e7f8c3 | 1417 | free(e); |
b81221b7 | 1418 | |
80e7f8c3 | 1419 | /* add the amount of space needed for metadata */ |
fbfdcb06 AO |
1420 | size += imsm_min_reserved_sectors(super); |
1421 | ||
1422 | c->min_size = size * 512; | |
4b57ecf6 | 1423 | c->sector_size = super->sector_size; |
b81221b7 | 1424 | |
fbfdcb06 | 1425 | return 0; |
80e7f8c3 AC |
1426 | } |
1427 | ||
d1e02575 AK |
1428 | static int is_gen_migration(struct imsm_dev *dev); |
1429 | ||
f36a9ecd PB |
1430 | #define IMSM_4K_DIV 8 |
1431 | ||
c47b0ff6 AK |
1432 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
1433 | struct imsm_dev *dev); | |
1e5c6983 | 1434 | |
c47b0ff6 AK |
1435 | static void print_imsm_dev(struct intel_super *super, |
1436 | struct imsm_dev *dev, | |
1437 | char *uuid, | |
1438 | int disk_idx) | |
cdddbdbc DW |
1439 | { |
1440 | __u64 sz; | |
0d80bb2f | 1441 | int slot, i; |
238c0a71 AK |
1442 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1443 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); | |
b10b37b8 | 1444 | __u32 ord; |
cdddbdbc DW |
1445 | |
1446 | printf("\n"); | |
1e7bc0ed | 1447 | printf("[%.16s]:\n", dev->volume); |
44470971 | 1448 | printf(" UUID : %s\n", uuid); |
dd8bcb3b AK |
1449 | printf(" RAID Level : %d", get_imsm_raid_level(map)); |
1450 | if (map2) | |
1451 | printf(" <-- %d", get_imsm_raid_level(map2)); | |
1452 | printf("\n"); | |
1453 | printf(" Members : %d", map->num_members); | |
1454 | if (map2) | |
1455 | printf(" <-- %d", map2->num_members); | |
1456 | printf("\n"); | |
0d80bb2f DW |
1457 | printf(" Slots : ["); |
1458 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 1459 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_0); |
0d80bb2f DW |
1460 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
1461 | } | |
dd8bcb3b AK |
1462 | printf("]"); |
1463 | if (map2) { | |
1464 | printf(" <-- ["); | |
1465 | for (i = 0; i < map2->num_members; i++) { | |
238c0a71 | 1466 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_1); |
dd8bcb3b AK |
1467 | printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U"); |
1468 | } | |
1469 | printf("]"); | |
1470 | } | |
1471 | printf("\n"); | |
7095bccb AK |
1472 | printf(" Failed disk : "); |
1473 | if (map->failed_disk_num == 0xff) | |
1474 | printf("none"); | |
1475 | else | |
1476 | printf("%i", map->failed_disk_num); | |
1477 | printf("\n"); | |
620b1713 DW |
1478 | slot = get_imsm_disk_slot(map, disk_idx); |
1479 | if (slot >= 0) { | |
238c0a71 | 1480 | ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); |
b10b37b8 DW |
1481 | printf(" This Slot : %d%s\n", slot, |
1482 | ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : ""); | |
1483 | } else | |
cdddbdbc | 1484 | printf(" This Slot : ?\n"); |
84918897 | 1485 | printf(" Sector Size : %u\n", super->sector_size); |
cdddbdbc DW |
1486 | sz = __le32_to_cpu(dev->size_high); |
1487 | sz <<= 32; | |
1488 | sz += __le32_to_cpu(dev->size_low); | |
84918897 MK |
1489 | printf(" Array Size : %llu%s\n", |
1490 | (unsigned long long)sz * 512 / super->sector_size, | |
cdddbdbc | 1491 | human_size(sz * 512)); |
5551b113 | 1492 | sz = blocks_per_member(map); |
84918897 MK |
1493 | printf(" Per Dev Size : %llu%s\n", |
1494 | (unsigned long long)sz * 512 / super->sector_size, | |
cdddbdbc | 1495 | human_size(sz * 512)); |
5551b113 CA |
1496 | printf(" Sector Offset : %llu\n", |
1497 | pba_of_lba0(map)); | |
1498 | printf(" Num Stripes : %llu\n", | |
1499 | num_data_stripes(map)); | |
dd8bcb3b | 1500 | printf(" Chunk Size : %u KiB", |
cdddbdbc | 1501 | __le16_to_cpu(map->blocks_per_strip) / 2); |
dd8bcb3b AK |
1502 | if (map2) |
1503 | printf(" <-- %u KiB", | |
1504 | __le16_to_cpu(map2->blocks_per_strip) / 2); | |
1505 | printf("\n"); | |
cdddbdbc | 1506 | printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks)); |
8655a7b1 | 1507 | printf(" Migrate State : "); |
1484e727 DW |
1508 | if (dev->vol.migr_state) { |
1509 | if (migr_type(dev) == MIGR_INIT) | |
8655a7b1 | 1510 | printf("initialize\n"); |
1484e727 | 1511 | else if (migr_type(dev) == MIGR_REBUILD) |
8655a7b1 | 1512 | printf("rebuild\n"); |
1484e727 | 1513 | else if (migr_type(dev) == MIGR_VERIFY) |
8655a7b1 | 1514 | printf("check\n"); |
1484e727 | 1515 | else if (migr_type(dev) == MIGR_GEN_MIGR) |
8655a7b1 | 1516 | printf("general migration\n"); |
1484e727 | 1517 | else if (migr_type(dev) == MIGR_STATE_CHANGE) |
8655a7b1 | 1518 | printf("state change\n"); |
1484e727 | 1519 | else if (migr_type(dev) == MIGR_REPAIR) |
8655a7b1 | 1520 | printf("repair\n"); |
1484e727 | 1521 | else |
8655a7b1 DW |
1522 | printf("<unknown:%d>\n", migr_type(dev)); |
1523 | } else | |
1524 | printf("idle\n"); | |
3393c6af DW |
1525 | printf(" Map State : %s", map_state_str[map->map_state]); |
1526 | if (dev->vol.migr_state) { | |
238c0a71 | 1527 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 | 1528 | |
b10b37b8 | 1529 | printf(" <-- %s", map_state_str[map->map_state]); |
464d40e8 LD |
1530 | printf("\n Checkpoint : %u ", |
1531 | __le32_to_cpu(dev->vol.curr_migr_unit)); | |
089f9d79 | 1532 | if (is_gen_migration(dev) && (slot > 1 || slot < 0)) |
464d40e8 LD |
1533 | printf("(N/A)"); |
1534 | else | |
1535 | printf("(%llu)", (unsigned long long) | |
1536 | blocks_per_migr_unit(super, dev)); | |
3393c6af DW |
1537 | } |
1538 | printf("\n"); | |
2432ce9b AP |
1539 | printf(" Dirty State : %s\n", (dev->vol.dirty & RAIDVOL_DIRTY) ? |
1540 | "dirty" : "clean"); | |
1541 | printf(" RWH Policy : "); | |
1542 | if (dev->rwh_policy == RWH_OFF) | |
1543 | printf("off\n"); | |
1544 | else if (dev->rwh_policy == RWH_DISTRIBUTED) | |
1545 | printf("PPL distributed\n"); | |
1546 | else if (dev->rwh_policy == RWH_JOURNALING_DRIVE) | |
1547 | printf("PPL journaling drive\n"); | |
1548 | else | |
1549 | printf("<unknown:%d>\n", dev->rwh_policy); | |
cdddbdbc DW |
1550 | } |
1551 | ||
ef5c214e MK |
1552 | static void print_imsm_disk(struct imsm_disk *disk, |
1553 | int index, | |
1554 | __u32 reserved, | |
1555 | unsigned int sector_size) { | |
1f24f035 | 1556 | char str[MAX_RAID_SERIAL_LEN + 1]; |
cdddbdbc DW |
1557 | __u64 sz; |
1558 | ||
0ec1f4e8 | 1559 | if (index < -1 || !disk) |
e9d82038 DW |
1560 | return; |
1561 | ||
cdddbdbc | 1562 | printf("\n"); |
1f24f035 | 1563 | snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial); |
0ec1f4e8 DW |
1564 | if (index >= 0) |
1565 | printf(" Disk%02d Serial : %s\n", index, str); | |
1566 | else | |
1567 | printf(" Disk Serial : %s\n", str); | |
2432ce9b AP |
1568 | printf(" State :%s%s%s%s\n", is_spare(disk) ? " spare" : "", |
1569 | is_configured(disk) ? " active" : "", | |
1570 | is_failed(disk) ? " failed" : "", | |
1571 | is_journal(disk) ? " journal" : ""); | |
cdddbdbc | 1572 | printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id)); |
5551b113 | 1573 | sz = total_blocks(disk) - reserved; |
ef5c214e MK |
1574 | printf(" Usable Size : %llu%s\n", |
1575 | (unsigned long long)sz * 512 / sector_size, | |
cdddbdbc DW |
1576 | human_size(sz * 512)); |
1577 | } | |
1578 | ||
de44e46f PB |
1579 | void convert_to_4k_imsm_migr_rec(struct intel_super *super) |
1580 | { | |
1581 | struct migr_record *migr_rec = super->migr_rec; | |
1582 | ||
1583 | migr_rec->blocks_per_unit /= IMSM_4K_DIV; | |
1584 | migr_rec->ckpt_area_pba /= IMSM_4K_DIV; | |
1585 | migr_rec->dest_1st_member_lba /= IMSM_4K_DIV; | |
1586 | migr_rec->dest_depth_per_unit /= IMSM_4K_DIV; | |
1587 | split_ull((join_u32(migr_rec->post_migr_vol_cap, | |
1588 | migr_rec->post_migr_vol_cap_hi) / IMSM_4K_DIV), | |
1589 | &migr_rec->post_migr_vol_cap, &migr_rec->post_migr_vol_cap_hi); | |
1590 | } | |
1591 | ||
f36a9ecd PB |
1592 | void convert_to_4k_imsm_disk(struct imsm_disk *disk) |
1593 | { | |
1594 | set_total_blocks(disk, (total_blocks(disk)/IMSM_4K_DIV)); | |
1595 | } | |
1596 | ||
1597 | void convert_to_4k(struct intel_super *super) | |
1598 | { | |
1599 | struct imsm_super *mpb = super->anchor; | |
1600 | struct imsm_disk *disk; | |
1601 | int i; | |
e4467bc7 | 1602 | __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size); |
f36a9ecd PB |
1603 | |
1604 | for (i = 0; i < mpb->num_disks ; i++) { | |
1605 | disk = __get_imsm_disk(mpb, i); | |
1606 | /* disk */ | |
1607 | convert_to_4k_imsm_disk(disk); | |
1608 | } | |
1609 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
1610 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1611 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
1612 | /* dev */ | |
1613 | split_ull((join_u32(dev->size_low, dev->size_high)/IMSM_4K_DIV), | |
1614 | &dev->size_low, &dev->size_high); | |
1615 | dev->vol.curr_migr_unit /= IMSM_4K_DIV; | |
1616 | ||
1617 | /* map0 */ | |
1618 | set_blocks_per_member(map, blocks_per_member(map)/IMSM_4K_DIV); | |
1619 | map->blocks_per_strip /= IMSM_4K_DIV; | |
1620 | set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV); | |
1621 | ||
1622 | if (dev->vol.migr_state) { | |
1623 | /* map1 */ | |
1624 | map = get_imsm_map(dev, MAP_1); | |
1625 | set_blocks_per_member(map, | |
1626 | blocks_per_member(map)/IMSM_4K_DIV); | |
1627 | map->blocks_per_strip /= IMSM_4K_DIV; | |
1628 | set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV); | |
1629 | } | |
1630 | } | |
e4467bc7 TM |
1631 | if (bbm_log_size) { |
1632 | struct bbm_log *log = (void *)mpb + | |
1633 | __le32_to_cpu(mpb->mpb_size) - bbm_log_size; | |
1634 | __u32 i; | |
1635 | ||
1636 | for (i = 0; i < log->entry_count; i++) { | |
1637 | struct bbm_log_entry *entry = | |
1638 | &log->marked_block_entries[i]; | |
1639 | ||
1640 | __u8 count = entry->marked_count + 1; | |
1641 | unsigned long long sector = | |
1642 | __le48_to_cpu(&entry->defective_block_start); | |
1643 | ||
1644 | entry->defective_block_start = | |
1645 | __cpu_to_le48(sector/IMSM_4K_DIV); | |
1646 | entry->marked_count = max(count/IMSM_4K_DIV, 1) - 1; | |
1647 | } | |
1648 | } | |
f36a9ecd PB |
1649 | |
1650 | mpb->check_sum = __gen_imsm_checksum(mpb); | |
1651 | } | |
1652 | ||
520e69e2 AK |
1653 | void examine_migr_rec_imsm(struct intel_super *super) |
1654 | { | |
1655 | struct migr_record *migr_rec = super->migr_rec; | |
1656 | struct imsm_super *mpb = super->anchor; | |
1657 | int i; | |
1658 | ||
1659 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
1660 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
3136abe5 | 1661 | struct imsm_map *map; |
b4ab44d8 | 1662 | int slot = -1; |
3136abe5 | 1663 | |
520e69e2 AK |
1664 | if (is_gen_migration(dev) == 0) |
1665 | continue; | |
1666 | ||
1667 | printf("\nMigration Record Information:"); | |
3136abe5 | 1668 | |
44bfe6df AK |
1669 | /* first map under migration */ |
1670 | map = get_imsm_map(dev, MAP_0); | |
3136abe5 AK |
1671 | if (map) |
1672 | slot = get_imsm_disk_slot(map, super->disks->index); | |
089f9d79 | 1673 | if (map == NULL || slot > 1 || slot < 0) { |
520e69e2 AK |
1674 | printf(" Empty\n "); |
1675 | printf("Examine one of first two disks in array\n"); | |
1676 | break; | |
1677 | } | |
1678 | printf("\n Status : "); | |
1679 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
1680 | printf("Normal\n"); | |
1681 | else | |
1682 | printf("Contains Data\n"); | |
1683 | printf(" Current Unit : %u\n", | |
1684 | __le32_to_cpu(migr_rec->curr_migr_unit)); | |
1685 | printf(" Family : %u\n", | |
1686 | __le32_to_cpu(migr_rec->family_num)); | |
1687 | printf(" Ascending : %u\n", | |
1688 | __le32_to_cpu(migr_rec->ascending_migr)); | |
1689 | printf(" Blocks Per Unit : %u\n", | |
1690 | __le32_to_cpu(migr_rec->blocks_per_unit)); | |
1691 | printf(" Dest. Depth Per Unit : %u\n", | |
1692 | __le32_to_cpu(migr_rec->dest_depth_per_unit)); | |
1693 | printf(" Checkpoint Area pba : %u\n", | |
1694 | __le32_to_cpu(migr_rec->ckpt_area_pba)); | |
1695 | printf(" First member lba : %u\n", | |
1696 | __le32_to_cpu(migr_rec->dest_1st_member_lba)); | |
1697 | printf(" Total Number of Units : %u\n", | |
1698 | __le32_to_cpu(migr_rec->num_migr_units)); | |
1699 | printf(" Size of volume : %u\n", | |
1700 | __le32_to_cpu(migr_rec->post_migr_vol_cap)); | |
1701 | printf(" Expansion space for LBA64 : %u\n", | |
1702 | __le32_to_cpu(migr_rec->post_migr_vol_cap_hi)); | |
1703 | printf(" Record was read from : %u\n", | |
1704 | __le32_to_cpu(migr_rec->ckpt_read_disk_num)); | |
1705 | ||
1706 | break; | |
1707 | } | |
1708 | } | |
f36a9ecd | 1709 | |
de44e46f PB |
1710 | void convert_from_4k_imsm_migr_rec(struct intel_super *super) |
1711 | { | |
1712 | struct migr_record *migr_rec = super->migr_rec; | |
1713 | ||
1714 | migr_rec->blocks_per_unit *= IMSM_4K_DIV; | |
1715 | migr_rec->ckpt_area_pba *= IMSM_4K_DIV; | |
1716 | migr_rec->dest_1st_member_lba *= IMSM_4K_DIV; | |
1717 | migr_rec->dest_depth_per_unit *= IMSM_4K_DIV; | |
1718 | split_ull((join_u32(migr_rec->post_migr_vol_cap, | |
1719 | migr_rec->post_migr_vol_cap_hi) * IMSM_4K_DIV), | |
1720 | &migr_rec->post_migr_vol_cap, | |
1721 | &migr_rec->post_migr_vol_cap_hi); | |
1722 | } | |
1723 | ||
f36a9ecd PB |
1724 | void convert_from_4k(struct intel_super *super) |
1725 | { | |
1726 | struct imsm_super *mpb = super->anchor; | |
1727 | struct imsm_disk *disk; | |
1728 | int i; | |
e4467bc7 | 1729 | __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size); |
f36a9ecd PB |
1730 | |
1731 | for (i = 0; i < mpb->num_disks ; i++) { | |
1732 | disk = __get_imsm_disk(mpb, i); | |
1733 | /* disk */ | |
1734 | set_total_blocks(disk, (total_blocks(disk)*IMSM_4K_DIV)); | |
1735 | } | |
1736 | ||
1737 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
1738 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1739 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
1740 | /* dev */ | |
1741 | split_ull((join_u32(dev->size_low, dev->size_high)*IMSM_4K_DIV), | |
1742 | &dev->size_low, &dev->size_high); | |
1743 | dev->vol.curr_migr_unit *= IMSM_4K_DIV; | |
1744 | ||
1745 | /* map0 */ | |
1746 | set_blocks_per_member(map, blocks_per_member(map)*IMSM_4K_DIV); | |
1747 | map->blocks_per_strip *= IMSM_4K_DIV; | |
1748 | set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV); | |
1749 | ||
1750 | if (dev->vol.migr_state) { | |
1751 | /* map1 */ | |
1752 | map = get_imsm_map(dev, MAP_1); | |
1753 | set_blocks_per_member(map, | |
1754 | blocks_per_member(map)*IMSM_4K_DIV); | |
1755 | map->blocks_per_strip *= IMSM_4K_DIV; | |
1756 | set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV); | |
1757 | } | |
1758 | } | |
e4467bc7 TM |
1759 | if (bbm_log_size) { |
1760 | struct bbm_log *log = (void *)mpb + | |
1761 | __le32_to_cpu(mpb->mpb_size) - bbm_log_size; | |
1762 | __u32 i; | |
1763 | ||
1764 | for (i = 0; i < log->entry_count; i++) { | |
1765 | struct bbm_log_entry *entry = | |
1766 | &log->marked_block_entries[i]; | |
1767 | ||
1768 | __u8 count = entry->marked_count + 1; | |
1769 | unsigned long long sector = | |
1770 | __le48_to_cpu(&entry->defective_block_start); | |
1771 | ||
1772 | entry->defective_block_start = | |
1773 | __cpu_to_le48(sector*IMSM_4K_DIV); | |
1774 | entry->marked_count = count*IMSM_4K_DIV - 1; | |
1775 | } | |
1776 | } | |
f36a9ecd PB |
1777 | |
1778 | mpb->check_sum = __gen_imsm_checksum(mpb); | |
1779 | } | |
1780 | ||
19482bcc AK |
1781 | /******************************************************************************* |
1782 | * function: imsm_check_attributes | |
1783 | * Description: Function checks if features represented by attributes flags | |
1011e834 | 1784 | * are supported by mdadm. |
19482bcc AK |
1785 | * Parameters: |
1786 | * attributes - Attributes read from metadata | |
1787 | * Returns: | |
1011e834 N |
1788 | * 0 - passed attributes contains unsupported features flags |
1789 | * 1 - all features are supported | |
19482bcc AK |
1790 | ******************************************************************************/ |
1791 | static int imsm_check_attributes(__u32 attributes) | |
1792 | { | |
1793 | int ret_val = 1; | |
418f9b36 N |
1794 | __u32 not_supported = MPB_ATTRIB_SUPPORTED^0xffffffff; |
1795 | ||
1796 | not_supported &= ~MPB_ATTRIB_IGNORED; | |
19482bcc AK |
1797 | |
1798 | not_supported &= attributes; | |
1799 | if (not_supported) { | |
e7b84f9d | 1800 | pr_err("(IMSM): Unsupported attributes : %x\n", |
418f9b36 | 1801 | (unsigned)__le32_to_cpu(not_supported)); |
19482bcc AK |
1802 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { |
1803 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY \n"); | |
1804 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
1805 | } | |
1806 | if (not_supported & MPB_ATTRIB_2TB) { | |
1807 | dprintf("\t\tMPB_ATTRIB_2TB\n"); | |
1808 | not_supported ^= MPB_ATTRIB_2TB; | |
1809 | } | |
1810 | if (not_supported & MPB_ATTRIB_RAID0) { | |
1811 | dprintf("\t\tMPB_ATTRIB_RAID0\n"); | |
1812 | not_supported ^= MPB_ATTRIB_RAID0; | |
1813 | } | |
1814 | if (not_supported & MPB_ATTRIB_RAID1) { | |
1815 | dprintf("\t\tMPB_ATTRIB_RAID1\n"); | |
1816 | not_supported ^= MPB_ATTRIB_RAID1; | |
1817 | } | |
1818 | if (not_supported & MPB_ATTRIB_RAID10) { | |
1819 | dprintf("\t\tMPB_ATTRIB_RAID10\n"); | |
1820 | not_supported ^= MPB_ATTRIB_RAID10; | |
1821 | } | |
1822 | if (not_supported & MPB_ATTRIB_RAID1E) { | |
1823 | dprintf("\t\tMPB_ATTRIB_RAID1E\n"); | |
1824 | not_supported ^= MPB_ATTRIB_RAID1E; | |
1825 | } | |
1826 | if (not_supported & MPB_ATTRIB_RAID5) { | |
1827 | dprintf("\t\tMPB_ATTRIB_RAID5\n"); | |
1828 | not_supported ^= MPB_ATTRIB_RAID5; | |
1829 | } | |
1830 | if (not_supported & MPB_ATTRIB_RAIDCNG) { | |
1831 | dprintf("\t\tMPB_ATTRIB_RAIDCNG\n"); | |
1832 | not_supported ^= MPB_ATTRIB_RAIDCNG; | |
1833 | } | |
1834 | if (not_supported & MPB_ATTRIB_BBM) { | |
1835 | dprintf("\t\tMPB_ATTRIB_BBM\n"); | |
1836 | not_supported ^= MPB_ATTRIB_BBM; | |
1837 | } | |
1838 | if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) { | |
1839 | dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY (== MPB_ATTRIB_LEGACY)\n"); | |
1840 | not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY; | |
1841 | } | |
1842 | if (not_supported & MPB_ATTRIB_EXP_STRIPE_SIZE) { | |
1843 | dprintf("\t\tMPB_ATTRIB_EXP_STRIP_SIZE\n"); | |
1844 | not_supported ^= MPB_ATTRIB_EXP_STRIPE_SIZE; | |
1845 | } | |
1846 | if (not_supported & MPB_ATTRIB_2TB_DISK) { | |
1847 | dprintf("\t\tMPB_ATTRIB_2TB_DISK\n"); | |
1848 | not_supported ^= MPB_ATTRIB_2TB_DISK; | |
1849 | } | |
1850 | if (not_supported & MPB_ATTRIB_NEVER_USE2) { | |
1851 | dprintf("\t\tMPB_ATTRIB_NEVER_USE2\n"); | |
1852 | not_supported ^= MPB_ATTRIB_NEVER_USE2; | |
1853 | } | |
1854 | if (not_supported & MPB_ATTRIB_NEVER_USE) { | |
1855 | dprintf("\t\tMPB_ATTRIB_NEVER_USE\n"); | |
1856 | not_supported ^= MPB_ATTRIB_NEVER_USE; | |
1857 | } | |
1858 | ||
1859 | if (not_supported) | |
1ade5cc1 | 1860 | dprintf("(IMSM): Unknown attributes : %x\n", not_supported); |
19482bcc AK |
1861 | |
1862 | ret_val = 0; | |
1863 | } | |
1864 | ||
1865 | return ret_val; | |
1866 | } | |
1867 | ||
a5d85af7 | 1868 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map); |
44470971 | 1869 | |
cdddbdbc DW |
1870 | static void examine_super_imsm(struct supertype *st, char *homehost) |
1871 | { | |
1872 | struct intel_super *super = st->sb; | |
949c47a0 | 1873 | struct imsm_super *mpb = super->anchor; |
cdddbdbc DW |
1874 | char str[MAX_SIGNATURE_LENGTH]; |
1875 | int i; | |
27fd6274 DW |
1876 | struct mdinfo info; |
1877 | char nbuf[64]; | |
cdddbdbc | 1878 | __u32 sum; |
14e8215b | 1879 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
94827db3 | 1880 | struct dl *dl; |
27fd6274 | 1881 | |
618f4e6d XN |
1882 | strncpy(str, (char *)mpb->sig, MPB_SIG_LEN); |
1883 | str[MPB_SIG_LEN-1] = '\0'; | |
cdddbdbc DW |
1884 | printf(" Magic : %s\n", str); |
1885 | snprintf(str, strlen(MPB_VERSION_RAID0), "%s", get_imsm_version(mpb)); | |
1886 | printf(" Version : %s\n", get_imsm_version(mpb)); | |
148acb7b | 1887 | printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num)); |
cdddbdbc DW |
1888 | printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num)); |
1889 | printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num)); | |
19482bcc AK |
1890 | printf(" Attributes : "); |
1891 | if (imsm_check_attributes(mpb->attributes)) | |
1892 | printf("All supported\n"); | |
1893 | else | |
1894 | printf("not supported\n"); | |
a5d85af7 | 1895 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1896 | fname_from_uuid(st, &info, nbuf, ':'); |
27fd6274 | 1897 | printf(" UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
1898 | sum = __le32_to_cpu(mpb->check_sum); |
1899 | printf(" Checksum : %08x %s\n", sum, | |
949c47a0 | 1900 | __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect"); |
f36a9ecd | 1901 | printf(" MPB Sectors : %d\n", mpb_sectors(mpb, super->sector_size)); |
cdddbdbc DW |
1902 | printf(" Disks : %d\n", mpb->num_disks); |
1903 | printf(" RAID Devices : %d\n", mpb->num_raid_devs); | |
ef5c214e MK |
1904 | print_imsm_disk(__get_imsm_disk(mpb, super->disks->index), |
1905 | super->disks->index, reserved, super->sector_size); | |
8d67477f | 1906 | if (get_imsm_bbm_log_size(super->bbm_log)) { |
604b746f JD |
1907 | struct bbm_log *log = super->bbm_log; |
1908 | ||
1909 | printf("\n"); | |
1910 | printf("Bad Block Management Log:\n"); | |
1911 | printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size)); | |
1912 | printf(" Signature : %x\n", __le32_to_cpu(log->signature)); | |
1913 | printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count)); | |
604b746f | 1914 | } |
44470971 DW |
1915 | for (i = 0; i < mpb->num_raid_devs; i++) { |
1916 | struct mdinfo info; | |
1917 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
1918 | ||
1919 | super->current_vol = i; | |
a5d85af7 | 1920 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1921 | fname_from_uuid(st, &info, nbuf, ':'); |
c47b0ff6 | 1922 | print_imsm_dev(super, dev, nbuf + 5, super->disks->index); |
44470971 | 1923 | } |
cdddbdbc DW |
1924 | for (i = 0; i < mpb->num_disks; i++) { |
1925 | if (i == super->disks->index) | |
1926 | continue; | |
ef5c214e MK |
1927 | print_imsm_disk(__get_imsm_disk(mpb, i), i, reserved, |
1928 | super->sector_size); | |
cdddbdbc | 1929 | } |
94827db3 | 1930 | |
0ec1f4e8 DW |
1931 | for (dl = super->disks; dl; dl = dl->next) |
1932 | if (dl->index == -1) | |
ef5c214e MK |
1933 | print_imsm_disk(&dl->disk, -1, reserved, |
1934 | super->sector_size); | |
520e69e2 AK |
1935 | |
1936 | examine_migr_rec_imsm(super); | |
cdddbdbc DW |
1937 | } |
1938 | ||
061f2c6a | 1939 | static void brief_examine_super_imsm(struct supertype *st, int verbose) |
cdddbdbc | 1940 | { |
27fd6274 | 1941 | /* We just write a generic IMSM ARRAY entry */ |
ff54de6e N |
1942 | struct mdinfo info; |
1943 | char nbuf[64]; | |
1e7bc0ed | 1944 | struct intel_super *super = st->sb; |
1e7bc0ed | 1945 | |
0d5a423f DW |
1946 | if (!super->anchor->num_raid_devs) { |
1947 | printf("ARRAY metadata=imsm\n"); | |
1e7bc0ed | 1948 | return; |
0d5a423f | 1949 | } |
ff54de6e | 1950 | |
a5d85af7 | 1951 | getinfo_super_imsm(st, &info, NULL); |
4737ae25 N |
1952 | fname_from_uuid(st, &info, nbuf, ':'); |
1953 | printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5); | |
1954 | } | |
1955 | ||
1956 | static void brief_examine_subarrays_imsm(struct supertype *st, int verbose) | |
1957 | { | |
1958 | /* We just write a generic IMSM ARRAY entry */ | |
1959 | struct mdinfo info; | |
1960 | char nbuf[64]; | |
1961 | char nbuf1[64]; | |
1962 | struct intel_super *super = st->sb; | |
1963 | int i; | |
1964 | ||
1965 | if (!super->anchor->num_raid_devs) | |
1966 | return; | |
1967 | ||
a5d85af7 | 1968 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1969 | fname_from_uuid(st, &info, nbuf, ':'); |
1e7bc0ed DW |
1970 | for (i = 0; i < super->anchor->num_raid_devs; i++) { |
1971 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
1972 | ||
1973 | super->current_vol = i; | |
a5d85af7 | 1974 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 1975 | fname_from_uuid(st, &info, nbuf1, ':'); |
1124b3cf | 1976 | printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n", |
cf8de691 | 1977 | dev->volume, nbuf + 5, i, nbuf1 + 5); |
1e7bc0ed | 1978 | } |
cdddbdbc DW |
1979 | } |
1980 | ||
9d84c8ea DW |
1981 | static void export_examine_super_imsm(struct supertype *st) |
1982 | { | |
1983 | struct intel_super *super = st->sb; | |
1984 | struct imsm_super *mpb = super->anchor; | |
1985 | struct mdinfo info; | |
1986 | char nbuf[64]; | |
1987 | ||
a5d85af7 | 1988 | getinfo_super_imsm(st, &info, NULL); |
9d84c8ea DW |
1989 | fname_from_uuid(st, &info, nbuf, ':'); |
1990 | printf("MD_METADATA=imsm\n"); | |
1991 | printf("MD_LEVEL=container\n"); | |
1992 | printf("MD_UUID=%s\n", nbuf+5); | |
1993 | printf("MD_DEVICES=%u\n", mpb->num_disks); | |
1994 | } | |
1995 | ||
74db60b0 N |
1996 | static int copy_metadata_imsm(struct supertype *st, int from, int to) |
1997 | { | |
f36a9ecd | 1998 | /* The second last sector of the device contains |
74db60b0 N |
1999 | * the "struct imsm_super" metadata. |
2000 | * This contains mpb_size which is the size in bytes of the | |
2001 | * extended metadata. This is located immediately before | |
2002 | * the imsm_super. | |
2003 | * We want to read all that, plus the last sector which | |
2004 | * may contain a migration record, and write it all | |
2005 | * to the target. | |
2006 | */ | |
2007 | void *buf; | |
2008 | unsigned long long dsize, offset; | |
2009 | int sectors; | |
2010 | struct imsm_super *sb; | |
f36a9ecd PB |
2011 | struct intel_super *super = st->sb; |
2012 | unsigned int sector_size = super->sector_size; | |
2013 | unsigned int written = 0; | |
74db60b0 | 2014 | |
de44e46f | 2015 | if (posix_memalign(&buf, MAX_SECTOR_SIZE, MAX_SECTOR_SIZE) != 0) |
74db60b0 N |
2016 | return 1; |
2017 | ||
2018 | if (!get_dev_size(from, NULL, &dsize)) | |
2019 | goto err; | |
2020 | ||
f36a9ecd | 2021 | if (lseek64(from, dsize-(2*sector_size), 0) < 0) |
74db60b0 | 2022 | goto err; |
466070ad | 2023 | if ((unsigned int)read(from, buf, sector_size) != sector_size) |
74db60b0 N |
2024 | goto err; |
2025 | sb = buf; | |
2026 | if (strncmp((char*)sb->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) | |
2027 | goto err; | |
2028 | ||
f36a9ecd PB |
2029 | sectors = mpb_sectors(sb, sector_size) + 2; |
2030 | offset = dsize - sectors * sector_size; | |
74db60b0 N |
2031 | if (lseek64(from, offset, 0) < 0 || |
2032 | lseek64(to, offset, 0) < 0) | |
2033 | goto err; | |
f36a9ecd PB |
2034 | while (written < sectors * sector_size) { |
2035 | int n = sectors*sector_size - written; | |
74db60b0 N |
2036 | if (n > 4096) |
2037 | n = 4096; | |
2038 | if (read(from, buf, n) != n) | |
2039 | goto err; | |
2040 | if (write(to, buf, n) != n) | |
2041 | goto err; | |
2042 | written += n; | |
2043 | } | |
2044 | free(buf); | |
2045 | return 0; | |
2046 | err: | |
2047 | free(buf); | |
2048 | return 1; | |
2049 | } | |
2050 | ||
cdddbdbc DW |
2051 | static void detail_super_imsm(struct supertype *st, char *homehost) |
2052 | { | |
3ebe00a1 DW |
2053 | struct mdinfo info; |
2054 | char nbuf[64]; | |
2055 | ||
a5d85af7 | 2056 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 2057 | fname_from_uuid(st, &info, nbuf, ':'); |
65884368 | 2058 | printf("\n UUID : %s\n", nbuf + 5); |
cdddbdbc DW |
2059 | } |
2060 | ||
2061 | static void brief_detail_super_imsm(struct supertype *st) | |
2062 | { | |
ff54de6e N |
2063 | struct mdinfo info; |
2064 | char nbuf[64]; | |
a5d85af7 | 2065 | getinfo_super_imsm(st, &info, NULL); |
ae2bfd4e | 2066 | fname_from_uuid(st, &info, nbuf, ':'); |
ff54de6e | 2067 | printf(" UUID=%s", nbuf + 5); |
cdddbdbc | 2068 | } |
d665cc31 DW |
2069 | |
2070 | static int imsm_read_serial(int fd, char *devname, __u8 *serial); | |
2071 | static void fd2devname(int fd, char *name); | |
2072 | ||
120dc887 | 2073 | static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose) |
d665cc31 | 2074 | { |
120dc887 LM |
2075 | /* dump an unsorted list of devices attached to AHCI Intel storage |
2076 | * controller, as well as non-connected ports | |
d665cc31 DW |
2077 | */ |
2078 | int hba_len = strlen(hba_path) + 1; | |
2079 | struct dirent *ent; | |
2080 | DIR *dir; | |
2081 | char *path = NULL; | |
2082 | int err = 0; | |
2083 | unsigned long port_mask = (1 << port_count) - 1; | |
2084 | ||
f21e18ca | 2085 | if (port_count > (int)sizeof(port_mask) * 8) { |
ba728be7 | 2086 | if (verbose > 0) |
e7b84f9d | 2087 | pr_err("port_count %d out of range\n", port_count); |
d665cc31 DW |
2088 | return 2; |
2089 | } | |
2090 | ||
2091 | /* scroll through /sys/dev/block looking for devices attached to | |
2092 | * this hba | |
2093 | */ | |
2094 | dir = opendir("/sys/dev/block"); | |
1a6dd6b9 PB |
2095 | if (!dir) |
2096 | return 1; | |
2097 | ||
2098 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
d665cc31 DW |
2099 | int fd; |
2100 | char model[64]; | |
2101 | char vendor[64]; | |
2102 | char buf[1024]; | |
2103 | int major, minor; | |
2104 | char *device; | |
2105 | char *c; | |
2106 | int port; | |
2107 | int type; | |
2108 | ||
2109 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
2110 | continue; | |
2111 | path = devt_to_devpath(makedev(major, minor)); | |
2112 | if (!path) | |
2113 | continue; | |
2114 | if (!path_attached_to_hba(path, hba_path)) { | |
2115 | free(path); | |
2116 | path = NULL; | |
2117 | continue; | |
2118 | } | |
2119 | ||
2120 | /* retrieve the scsi device type */ | |
2121 | if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) { | |
ba728be7 | 2122 | if (verbose > 0) |
e7b84f9d | 2123 | pr_err("failed to allocate 'device'\n"); |
d665cc31 DW |
2124 | err = 2; |
2125 | break; | |
2126 | } | |
2127 | sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor); | |
193b6c0b | 2128 | if (load_sys(device, buf, sizeof(buf)) != 0) { |
ba728be7 | 2129 | if (verbose > 0) |
e7b84f9d | 2130 | pr_err("failed to read device type for %s\n", |
d665cc31 DW |
2131 | path); |
2132 | err = 2; | |
2133 | free(device); | |
2134 | break; | |
2135 | } | |
2136 | type = strtoul(buf, NULL, 10); | |
2137 | ||
2138 | /* if it's not a disk print the vendor and model */ | |
2139 | if (!(type == 0 || type == 7 || type == 14)) { | |
2140 | vendor[0] = '\0'; | |
2141 | model[0] = '\0'; | |
2142 | sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor); | |
193b6c0b | 2143 | if (load_sys(device, buf, sizeof(buf)) == 0) { |
d665cc31 DW |
2144 | strncpy(vendor, buf, sizeof(vendor)); |
2145 | vendor[sizeof(vendor) - 1] = '\0'; | |
2146 | c = (char *) &vendor[sizeof(vendor) - 1]; | |
2147 | while (isspace(*c) || *c == '\0') | |
2148 | *c-- = '\0'; | |
2149 | ||
2150 | } | |
2151 | sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor); | |
193b6c0b | 2152 | if (load_sys(device, buf, sizeof(buf)) == 0) { |
d665cc31 DW |
2153 | strncpy(model, buf, sizeof(model)); |
2154 | model[sizeof(model) - 1] = '\0'; | |
2155 | c = (char *) &model[sizeof(model) - 1]; | |
2156 | while (isspace(*c) || *c == '\0') | |
2157 | *c-- = '\0'; | |
2158 | } | |
2159 | ||
2160 | if (vendor[0] && model[0]) | |
2161 | sprintf(buf, "%.64s %.64s", vendor, model); | |
2162 | else | |
2163 | switch (type) { /* numbers from hald/linux/device.c */ | |
2164 | case 1: sprintf(buf, "tape"); break; | |
2165 | case 2: sprintf(buf, "printer"); break; | |
2166 | case 3: sprintf(buf, "processor"); break; | |
2167 | case 4: | |
2168 | case 5: sprintf(buf, "cdrom"); break; | |
2169 | case 6: sprintf(buf, "scanner"); break; | |
2170 | case 8: sprintf(buf, "media_changer"); break; | |
2171 | case 9: sprintf(buf, "comm"); break; | |
2172 | case 12: sprintf(buf, "raid"); break; | |
2173 | default: sprintf(buf, "unknown"); | |
2174 | } | |
2175 | } else | |
2176 | buf[0] = '\0'; | |
2177 | free(device); | |
2178 | ||
2179 | /* chop device path to 'host%d' and calculate the port number */ | |
2180 | c = strchr(&path[hba_len], '/'); | |
4e5e717d | 2181 | if (!c) { |
ba728be7 | 2182 | if (verbose > 0) |
e7b84f9d | 2183 | pr_err("%s - invalid path name\n", path + hba_len); |
4e5e717d AW |
2184 | err = 2; |
2185 | break; | |
2186 | } | |
d665cc31 | 2187 | *c = '\0'; |
0858eccf AP |
2188 | if ((sscanf(&path[hba_len], "ata%d", &port) == 1) || |
2189 | ((sscanf(&path[hba_len], "host%d", &port) == 1))) | |
d665cc31 DW |
2190 | port -= host_base; |
2191 | else { | |
ba728be7 | 2192 | if (verbose > 0) { |
d665cc31 | 2193 | *c = '/'; /* repair the full string */ |
e7b84f9d | 2194 | pr_err("failed to determine port number for %s\n", |
d665cc31 DW |
2195 | path); |
2196 | } | |
2197 | err = 2; | |
2198 | break; | |
2199 | } | |
2200 | ||
2201 | /* mark this port as used */ | |
2202 | port_mask &= ~(1 << port); | |
2203 | ||
2204 | /* print out the device information */ | |
2205 | if (buf[0]) { | |
2206 | printf(" Port%d : - non-disk device (%s) -\n", port, buf); | |
2207 | continue; | |
2208 | } | |
2209 | ||
2210 | fd = dev_open(ent->d_name, O_RDONLY); | |
2211 | if (fd < 0) | |
2212 | printf(" Port%d : - disk info unavailable -\n", port); | |
2213 | else { | |
2214 | fd2devname(fd, buf); | |
2215 | printf(" Port%d : %s", port, buf); | |
2216 | if (imsm_read_serial(fd, NULL, (__u8 *) buf) == 0) | |
664d5325 | 2217 | printf(" (%.*s)\n", MAX_RAID_SERIAL_LEN, buf); |
d665cc31 | 2218 | else |
664d5325 | 2219 | printf(" ()\n"); |
4dab422a | 2220 | close(fd); |
d665cc31 | 2221 | } |
d665cc31 DW |
2222 | free(path); |
2223 | path = NULL; | |
2224 | } | |
2225 | if (path) | |
2226 | free(path); | |
2227 | if (dir) | |
2228 | closedir(dir); | |
2229 | if (err == 0) { | |
2230 | int i; | |
2231 | ||
2232 | for (i = 0; i < port_count; i++) | |
2233 | if (port_mask & (1 << i)) | |
2234 | printf(" Port%d : - no device attached -\n", i); | |
2235 | } | |
2236 | ||
2237 | return err; | |
2238 | } | |
2239 | ||
b5eece69 | 2240 | static int print_vmd_attached_devs(struct sys_dev *hba) |
60f0f54d PB |
2241 | { |
2242 | struct dirent *ent; | |
2243 | DIR *dir; | |
2244 | char path[292]; | |
2245 | char link[256]; | |
2246 | char *c, *rp; | |
2247 | ||
2248 | if (hba->type != SYS_DEV_VMD) | |
b5eece69 | 2249 | return 1; |
60f0f54d PB |
2250 | |
2251 | /* scroll through /sys/dev/block looking for devices attached to | |
2252 | * this hba | |
2253 | */ | |
2254 | dir = opendir("/sys/bus/pci/drivers/nvme"); | |
b9135011 | 2255 | if (!dir) |
b5eece69 | 2256 | return 1; |
b9135011 JS |
2257 | |
2258 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
60f0f54d PB |
2259 | int n; |
2260 | ||
2261 | /* is 'ent' a device? check that the 'subsystem' link exists and | |
2262 | * that its target matches 'bus' | |
2263 | */ | |
2264 | sprintf(path, "/sys/bus/pci/drivers/nvme/%s/subsystem", | |
2265 | ent->d_name); | |
2266 | n = readlink(path, link, sizeof(link)); | |
2267 | if (n < 0 || n >= (int)sizeof(link)) | |
2268 | continue; | |
2269 | link[n] = '\0'; | |
2270 | c = strrchr(link, '/'); | |
2271 | if (!c) | |
2272 | continue; | |
2273 | if (strncmp("pci", c+1, strlen("pci")) != 0) | |
2274 | continue; | |
2275 | ||
2276 | sprintf(path, "/sys/bus/pci/drivers/nvme/%s", ent->d_name); | |
60f0f54d PB |
2277 | |
2278 | rp = realpath(path, NULL); | |
2279 | if (!rp) | |
2280 | continue; | |
2281 | ||
2282 | if (path_attached_to_hba(rp, hba->path)) { | |
2283 | printf(" NVMe under VMD : %s\n", rp); | |
2284 | } | |
2285 | free(rp); | |
2286 | } | |
2287 | ||
b9135011 | 2288 | closedir(dir); |
b5eece69 | 2289 | return 0; |
60f0f54d PB |
2290 | } |
2291 | ||
120dc887 LM |
2292 | static void print_found_intel_controllers(struct sys_dev *elem) |
2293 | { | |
2294 | for (; elem; elem = elem->next) { | |
e7b84f9d | 2295 | pr_err("found Intel(R) "); |
120dc887 LM |
2296 | if (elem->type == SYS_DEV_SATA) |
2297 | fprintf(stderr, "SATA "); | |
155cbb4c LM |
2298 | else if (elem->type == SYS_DEV_SAS) |
2299 | fprintf(stderr, "SAS "); | |
0858eccf AP |
2300 | else if (elem->type == SYS_DEV_NVME) |
2301 | fprintf(stderr, "NVMe "); | |
60f0f54d PB |
2302 | |
2303 | if (elem->type == SYS_DEV_VMD) | |
2304 | fprintf(stderr, "VMD domain"); | |
2305 | else | |
2306 | fprintf(stderr, "RAID controller"); | |
2307 | ||
120dc887 LM |
2308 | if (elem->pci_id) |
2309 | fprintf(stderr, " at %s", elem->pci_id); | |
2310 | fprintf(stderr, ".\n"); | |
2311 | } | |
2312 | fflush(stderr); | |
2313 | } | |
2314 | ||
120dc887 LM |
2315 | static int ahci_get_port_count(const char *hba_path, int *port_count) |
2316 | { | |
2317 | struct dirent *ent; | |
2318 | DIR *dir; | |
2319 | int host_base = -1; | |
2320 | ||
2321 | *port_count = 0; | |
2322 | if ((dir = opendir(hba_path)) == NULL) | |
2323 | return -1; | |
2324 | ||
2325 | for (ent = readdir(dir); ent; ent = readdir(dir)) { | |
2326 | int host; | |
2327 | ||
0858eccf AP |
2328 | if ((sscanf(ent->d_name, "ata%d", &host) != 1) && |
2329 | ((sscanf(ent->d_name, "host%d", &host) != 1))) | |
120dc887 LM |
2330 | continue; |
2331 | if (*port_count == 0) | |
2332 | host_base = host; | |
2333 | else if (host < host_base) | |
2334 | host_base = host; | |
2335 | ||
2336 | if (host + 1 > *port_count + host_base) | |
2337 | *port_count = host + 1 - host_base; | |
2338 | } | |
2339 | closedir(dir); | |
2340 | return host_base; | |
2341 | } | |
2342 | ||
a891a3c2 LM |
2343 | static void print_imsm_capability(const struct imsm_orom *orom) |
2344 | { | |
0858eccf AP |
2345 | printf(" Platform : Intel(R) "); |
2346 | if (orom->capabilities == 0 && orom->driver_features == 0) | |
2347 | printf("Matrix Storage Manager\n"); | |
2348 | else | |
2349 | printf("Rapid Storage Technology%s\n", | |
2350 | imsm_orom_is_enterprise(orom) ? " enterprise" : ""); | |
2351 | if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build) | |
2352 | printf(" Version : %d.%d.%d.%d\n", orom->major_ver, | |
2353 | orom->minor_ver, orom->hotfix_ver, orom->build); | |
a891a3c2 LM |
2354 | printf(" RAID Levels :%s%s%s%s%s\n", |
2355 | imsm_orom_has_raid0(orom) ? " raid0" : "", | |
2356 | imsm_orom_has_raid1(orom) ? " raid1" : "", | |
2357 | imsm_orom_has_raid1e(orom) ? " raid1e" : "", | |
2358 | imsm_orom_has_raid10(orom) ? " raid10" : "", | |
2359 | imsm_orom_has_raid5(orom) ? " raid5" : ""); | |
2360 | printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", | |
2361 | imsm_orom_has_chunk(orom, 2) ? " 2k" : "", | |
2362 | imsm_orom_has_chunk(orom, 4) ? " 4k" : "", | |
2363 | imsm_orom_has_chunk(orom, 8) ? " 8k" : "", | |
2364 | imsm_orom_has_chunk(orom, 16) ? " 16k" : "", | |
2365 | imsm_orom_has_chunk(orom, 32) ? " 32k" : "", | |
2366 | imsm_orom_has_chunk(orom, 64) ? " 64k" : "", | |
2367 | imsm_orom_has_chunk(orom, 128) ? " 128k" : "", | |
2368 | imsm_orom_has_chunk(orom, 256) ? " 256k" : "", | |
2369 | imsm_orom_has_chunk(orom, 512) ? " 512k" : "", | |
2370 | imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "", | |
2371 | imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "", | |
2372 | imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "", | |
2373 | imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "", | |
2374 | imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "", | |
2375 | imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "", | |
2376 | imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : ""); | |
29cd0821 CA |
2377 | printf(" 2TB volumes :%s supported\n", |
2378 | (orom->attr & IMSM_OROM_ATTR_2TB)?"":" not"); | |
2379 | printf(" 2TB disks :%s supported\n", | |
2380 | (orom->attr & IMSM_OROM_ATTR_2TB_DISK)?"":" not"); | |
0e7f69a8 | 2381 | printf(" Max Disks : %d\n", orom->tds); |
0858eccf AP |
2382 | printf(" Max Volumes : %d per array, %d per %s\n", |
2383 | orom->vpa, orom->vphba, | |
2384 | imsm_orom_is_nvme(orom) ? "platform" : "controller"); | |
a891a3c2 LM |
2385 | return; |
2386 | } | |
2387 | ||
e50cf220 MN |
2388 | static void print_imsm_capability_export(const struct imsm_orom *orom) |
2389 | { | |
2390 | printf("MD_FIRMWARE_TYPE=imsm\n"); | |
0858eccf AP |
2391 | if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build) |
2392 | printf("IMSM_VERSION=%d.%d.%d.%d\n", orom->major_ver, orom->minor_ver, | |
2393 | orom->hotfix_ver, orom->build); | |
e50cf220 MN |
2394 | printf("IMSM_SUPPORTED_RAID_LEVELS=%s%s%s%s%s\n", |
2395 | imsm_orom_has_raid0(orom) ? "raid0 " : "", | |
2396 | imsm_orom_has_raid1(orom) ? "raid1 " : "", | |
2397 | imsm_orom_has_raid1e(orom) ? "raid1e " : "", | |
2398 | imsm_orom_has_raid5(orom) ? "raid10 " : "", | |
2399 | imsm_orom_has_raid10(orom) ? "raid5 " : ""); | |
2400 | printf("IMSM_SUPPORTED_CHUNK_SIZES=%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", | |
2401 | imsm_orom_has_chunk(orom, 2) ? "2k " : "", | |
2402 | imsm_orom_has_chunk(orom, 4) ? "4k " : "", | |
2403 | imsm_orom_has_chunk(orom, 8) ? "8k " : "", | |
2404 | imsm_orom_has_chunk(orom, 16) ? "16k " : "", | |
2405 | imsm_orom_has_chunk(orom, 32) ? "32k " : "", | |
2406 | imsm_orom_has_chunk(orom, 64) ? "64k " : "", | |
2407 | imsm_orom_has_chunk(orom, 128) ? "128k " : "", | |
2408 | imsm_orom_has_chunk(orom, 256) ? "256k " : "", | |
2409 | imsm_orom_has_chunk(orom, 512) ? "512k " : "", | |
2410 | imsm_orom_has_chunk(orom, 1024*1) ? "1M " : "", | |
2411 | imsm_orom_has_chunk(orom, 1024*2) ? "2M " : "", | |
2412 | imsm_orom_has_chunk(orom, 1024*4) ? "4M " : "", | |
2413 | imsm_orom_has_chunk(orom, 1024*8) ? "8M " : "", | |
2414 | imsm_orom_has_chunk(orom, 1024*16) ? "16M " : "", | |
2415 | imsm_orom_has_chunk(orom, 1024*32) ? "32M " : "", | |
2416 | imsm_orom_has_chunk(orom, 1024*64) ? "64M " : ""); | |
2417 | printf("IMSM_2TB_VOLUMES=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB) ? "yes" : "no"); | |
2418 | printf("IMSM_2TB_DISKS=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB_DISK) ? "yes" : "no"); | |
2419 | printf("IMSM_MAX_DISKS=%d\n",orom->tds); | |
2420 | printf("IMSM_MAX_VOLUMES_PER_ARRAY=%d\n",orom->vpa); | |
2421 | printf("IMSM_MAX_VOLUMES_PER_CONTROLLER=%d\n",orom->vphba); | |
2422 | } | |
2423 | ||
9eafa1de | 2424 | static int detail_platform_imsm(int verbose, int enumerate_only, char *controller_path) |
d665cc31 DW |
2425 | { |
2426 | /* There are two components to imsm platform support, the ahci SATA | |
2427 | * controller and the option-rom. To find the SATA controller we | |
2428 | * simply look in /sys/bus/pci/drivers/ahci to see if an ahci | |
2429 | * controller with the Intel vendor id is present. This approach | |
2430 | * allows mdadm to leverage the kernel's ahci detection logic, with the | |
2431 | * caveat that if ahci.ko is not loaded mdadm will not be able to | |
2432 | * detect platform raid capabilities. The option-rom resides in a | |
2433 | * platform "Adapter ROM". We scan for its signature to retrieve the | |
2434 | * platform capabilities. If raid support is disabled in the BIOS the | |
2435 | * option-rom capability structure will not be available. | |
2436 | */ | |
d665cc31 | 2437 | struct sys_dev *list, *hba; |
d665cc31 DW |
2438 | int host_base = 0; |
2439 | int port_count = 0; | |
9eafa1de | 2440 | int result=1; |
d665cc31 | 2441 | |
5615172f | 2442 | if (enumerate_only) { |
a891a3c2 | 2443 | if (check_env("IMSM_NO_PLATFORM")) |
5615172f | 2444 | return 0; |
a891a3c2 LM |
2445 | list = find_intel_devices(); |
2446 | if (!list) | |
2447 | return 2; | |
2448 | for (hba = list; hba; hba = hba->next) { | |
6b781d33 AP |
2449 | if (find_imsm_capability(hba)) { |
2450 | result = 0; | |
a891a3c2 LM |
2451 | break; |
2452 | } | |
9eafa1de | 2453 | else |
6b781d33 | 2454 | result = 2; |
a891a3c2 | 2455 | } |
a891a3c2 | 2456 | return result; |
5615172f DW |
2457 | } |
2458 | ||
155cbb4c LM |
2459 | list = find_intel_devices(); |
2460 | if (!list) { | |
ba728be7 | 2461 | if (verbose > 0) |
7a862a02 | 2462 | pr_err("no active Intel(R) RAID controller found.\n"); |
d665cc31 | 2463 | return 2; |
ba728be7 | 2464 | } else if (verbose > 0) |
155cbb4c | 2465 | print_found_intel_controllers(list); |
d665cc31 | 2466 | |
a891a3c2 | 2467 | for (hba = list; hba; hba = hba->next) { |
0858eccf | 2468 | if (controller_path && (compare_paths(hba->path, controller_path) != 0)) |
9eafa1de | 2469 | continue; |
0858eccf | 2470 | if (!find_imsm_capability(hba)) { |
60f0f54d | 2471 | char buf[PATH_MAX]; |
e7b84f9d | 2472 | pr_err("imsm capabilities not found for controller: %s (type %s)\n", |
60f0f54d PB |
2473 | hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path, |
2474 | get_sys_dev_type(hba->type)); | |
0858eccf AP |
2475 | continue; |
2476 | } | |
2477 | result = 0; | |
2478 | } | |
2479 | ||
2480 | if (controller_path && result == 1) { | |
2481 | pr_err("no active Intel(R) RAID controller found under %s\n", | |
2482 | controller_path); | |
2483 | return result; | |
2484 | } | |
2485 | ||
5e1d6128 | 2486 | const struct orom_entry *entry; |
0858eccf | 2487 | |
5e1d6128 | 2488 | for (entry = orom_entries; entry; entry = entry->next) { |
60f0f54d | 2489 | if (entry->type == SYS_DEV_VMD) { |
07cb1e57 | 2490 | print_imsm_capability(&entry->orom); |
32716c51 PB |
2491 | printf(" 3rd party NVMe :%s supported\n", |
2492 | imsm_orom_has_tpv_support(&entry->orom)?"":" not"); | |
60f0f54d PB |
2493 | for (hba = list; hba; hba = hba->next) { |
2494 | if (hba->type == SYS_DEV_VMD) { | |
2495 | char buf[PATH_MAX]; | |
60f0f54d PB |
2496 | printf(" I/O Controller : %s (%s)\n", |
2497 | vmd_domain_to_controller(hba, buf), get_sys_dev_type(hba->type)); | |
b5eece69 PB |
2498 | if (print_vmd_attached_devs(hba)) { |
2499 | if (verbose > 0) | |
2500 | pr_err("failed to get devices attached to VMD domain.\n"); | |
2501 | result |= 2; | |
2502 | } | |
60f0f54d PB |
2503 | } |
2504 | } | |
07cb1e57 | 2505 | printf("\n"); |
60f0f54d PB |
2506 | continue; |
2507 | } | |
0858eccf | 2508 | |
60f0f54d PB |
2509 | print_imsm_capability(&entry->orom); |
2510 | if (entry->type == SYS_DEV_NVME) { | |
0858eccf AP |
2511 | for (hba = list; hba; hba = hba->next) { |
2512 | if (hba->type == SYS_DEV_NVME) | |
2513 | printf(" NVMe Device : %s\n", hba->path); | |
2514 | } | |
60f0f54d | 2515 | printf("\n"); |
0858eccf AP |
2516 | continue; |
2517 | } | |
2518 | ||
2519 | struct devid_list *devid; | |
5e1d6128 | 2520 | for (devid = entry->devid_list; devid; devid = devid->next) { |
0858eccf AP |
2521 | hba = device_by_id(devid->devid); |
2522 | if (!hba) | |
2523 | continue; | |
2524 | ||
9eafa1de MN |
2525 | printf(" I/O Controller : %s (%s)\n", |
2526 | hba->path, get_sys_dev_type(hba->type)); | |
2527 | if (hba->type == SYS_DEV_SATA) { | |
2528 | host_base = ahci_get_port_count(hba->path, &port_count); | |
2529 | if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) { | |
2530 | if (verbose > 0) | |
7a862a02 | 2531 | pr_err("failed to enumerate ports on SATA controller at %s.\n", hba->pci_id); |
9eafa1de MN |
2532 | result |= 2; |
2533 | } | |
120dc887 LM |
2534 | } |
2535 | } | |
0858eccf | 2536 | printf("\n"); |
d665cc31 | 2537 | } |
155cbb4c | 2538 | |
120dc887 | 2539 | return result; |
d665cc31 | 2540 | } |
e50cf220 | 2541 | |
9eafa1de | 2542 | static int export_detail_platform_imsm(int verbose, char *controller_path) |
e50cf220 | 2543 | { |
e50cf220 MN |
2544 | struct sys_dev *list, *hba; |
2545 | int result=1; | |
2546 | ||
2547 | list = find_intel_devices(); | |
2548 | if (!list) { | |
2549 | if (verbose > 0) | |
2550 | pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_INTEL_DEVICES\n"); | |
2551 | result = 2; | |
e50cf220 MN |
2552 | return result; |
2553 | } | |
2554 | ||
2555 | for (hba = list; hba; hba = hba->next) { | |
9eafa1de MN |
2556 | if (controller_path && (compare_paths(hba->path,controller_path) != 0)) |
2557 | continue; | |
60f0f54d PB |
2558 | if (!find_imsm_capability(hba) && verbose > 0) { |
2559 | char buf[PATH_MAX]; | |
2560 | pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_IMSM_CAPABLE_DEVICE_UNDER_%s\n", | |
2561 | hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path); | |
2562 | } | |
0858eccf | 2563 | else |
e50cf220 | 2564 | result = 0; |
e50cf220 MN |
2565 | } |
2566 | ||
5e1d6128 | 2567 | const struct orom_entry *entry; |
0858eccf | 2568 | |
60f0f54d PB |
2569 | for (entry = orom_entries; entry; entry = entry->next) { |
2570 | if (entry->type == SYS_DEV_VMD) { | |
2571 | for (hba = list; hba; hba = hba->next) | |
2572 | print_imsm_capability_export(&entry->orom); | |
2573 | continue; | |
2574 | } | |
5e1d6128 | 2575 | print_imsm_capability_export(&entry->orom); |
60f0f54d | 2576 | } |
0858eccf | 2577 | |
e50cf220 MN |
2578 | return result; |
2579 | } | |
2580 | ||
cdddbdbc DW |
2581 | static int match_home_imsm(struct supertype *st, char *homehost) |
2582 | { | |
5115ca67 DW |
2583 | /* the imsm metadata format does not specify any host |
2584 | * identification information. We return -1 since we can never | |
2585 | * confirm nor deny whether a given array is "meant" for this | |
148acb7b | 2586 | * host. We rely on compare_super and the 'family_num' fields to |
5115ca67 DW |
2587 | * exclude member disks that do not belong, and we rely on |
2588 | * mdadm.conf to specify the arrays that should be assembled. | |
2589 | * Auto-assembly may still pick up "foreign" arrays. | |
2590 | */ | |
cdddbdbc | 2591 | |
9362c1c8 | 2592 | return -1; |
cdddbdbc DW |
2593 | } |
2594 | ||
2595 | static void uuid_from_super_imsm(struct supertype *st, int uuid[4]) | |
2596 | { | |
51006d85 N |
2597 | /* The uuid returned here is used for: |
2598 | * uuid to put into bitmap file (Create, Grow) | |
2599 | * uuid for backup header when saving critical section (Grow) | |
2600 | * comparing uuids when re-adding a device into an array | |
2601 | * In these cases the uuid required is that of the data-array, | |
2602 | * not the device-set. | |
2603 | * uuid to recognise same set when adding a missing device back | |
2604 | * to an array. This is a uuid for the device-set. | |
1011e834 | 2605 | * |
51006d85 N |
2606 | * For each of these we can make do with a truncated |
2607 | * or hashed uuid rather than the original, as long as | |
2608 | * everyone agrees. | |
2609 | * In each case the uuid required is that of the data-array, | |
2610 | * not the device-set. | |
43dad3d6 | 2611 | */ |
51006d85 N |
2612 | /* imsm does not track uuid's so we synthesis one using sha1 on |
2613 | * - The signature (Which is constant for all imsm array, but no matter) | |
148acb7b | 2614 | * - the orig_family_num of the container |
51006d85 N |
2615 | * - the index number of the volume |
2616 | * - the 'serial' number of the volume. | |
2617 | * Hopefully these are all constant. | |
2618 | */ | |
2619 | struct intel_super *super = st->sb; | |
43dad3d6 | 2620 | |
51006d85 N |
2621 | char buf[20]; |
2622 | struct sha1_ctx ctx; | |
2623 | struct imsm_dev *dev = NULL; | |
148acb7b | 2624 | __u32 family_num; |
51006d85 | 2625 | |
148acb7b DW |
2626 | /* some mdadm versions failed to set ->orig_family_num, in which |
2627 | * case fall back to ->family_num. orig_family_num will be | |
2628 | * fixed up with the first metadata update. | |
2629 | */ | |
2630 | family_num = super->anchor->orig_family_num; | |
2631 | if (family_num == 0) | |
2632 | family_num = super->anchor->family_num; | |
51006d85 | 2633 | sha1_init_ctx(&ctx); |
92bd8f8d | 2634 | sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx); |
148acb7b | 2635 | sha1_process_bytes(&family_num, sizeof(__u32), &ctx); |
51006d85 N |
2636 | if (super->current_vol >= 0) |
2637 | dev = get_imsm_dev(super, super->current_vol); | |
2638 | if (dev) { | |
2639 | __u32 vol = super->current_vol; | |
2640 | sha1_process_bytes(&vol, sizeof(vol), &ctx); | |
2641 | sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx); | |
2642 | } | |
2643 | sha1_finish_ctx(&ctx, buf); | |
2644 | memcpy(uuid, buf, 4*4); | |
cdddbdbc DW |
2645 | } |
2646 | ||
0d481d37 | 2647 | #if 0 |
4f5bc454 DW |
2648 | static void |
2649 | get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p) | |
cdddbdbc | 2650 | { |
cdddbdbc DW |
2651 | __u8 *v = get_imsm_version(mpb); |
2652 | __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH; | |
2653 | char major[] = { 0, 0, 0 }; | |
2654 | char minor[] = { 0 ,0, 0 }; | |
2655 | char patch[] = { 0, 0, 0 }; | |
2656 | char *ver_parse[] = { major, minor, patch }; | |
2657 | int i, j; | |
2658 | ||
2659 | i = j = 0; | |
2660 | while (*v != '\0' && v < end) { | |
2661 | if (*v != '.' && j < 2) | |
2662 | ver_parse[i][j++] = *v; | |
2663 | else { | |
2664 | i++; | |
2665 | j = 0; | |
2666 | } | |
2667 | v++; | |
2668 | } | |
2669 | ||
4f5bc454 DW |
2670 | *m = strtol(minor, NULL, 0); |
2671 | *p = strtol(patch, NULL, 0); | |
2672 | } | |
0d481d37 | 2673 | #endif |
4f5bc454 | 2674 | |
1e5c6983 DW |
2675 | static __u32 migr_strip_blocks_resync(struct imsm_dev *dev) |
2676 | { | |
2677 | /* migr_strip_size when repairing or initializing parity */ | |
238c0a71 | 2678 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2679 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2680 | ||
2681 | switch (get_imsm_raid_level(map)) { | |
2682 | case 5: | |
2683 | case 10: | |
2684 | return chunk; | |
2685 | default: | |
2686 | return 128*1024 >> 9; | |
2687 | } | |
2688 | } | |
2689 | ||
2690 | static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev) | |
2691 | { | |
2692 | /* migr_strip_size when rebuilding a degraded disk, no idea why | |
2693 | * this is different than migr_strip_size_resync(), but it's good | |
2694 | * to be compatible | |
2695 | */ | |
238c0a71 | 2696 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2697 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2698 | ||
2699 | switch (get_imsm_raid_level(map)) { | |
2700 | case 1: | |
2701 | case 10: | |
2702 | if (map->num_members % map->num_domains == 0) | |
2703 | return 128*1024 >> 9; | |
2704 | else | |
2705 | return chunk; | |
2706 | case 5: | |
2707 | return max((__u32) 64*1024 >> 9, chunk); | |
2708 | default: | |
2709 | return 128*1024 >> 9; | |
2710 | } | |
2711 | } | |
2712 | ||
2713 | static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev) | |
2714 | { | |
238c0a71 AK |
2715 | struct imsm_map *lo = get_imsm_map(dev, MAP_0); |
2716 | struct imsm_map *hi = get_imsm_map(dev, MAP_1); | |
1e5c6983 DW |
2717 | __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip); |
2718 | __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip); | |
2719 | ||
2720 | return max((__u32) 1, hi_chunk / lo_chunk); | |
2721 | } | |
2722 | ||
2723 | static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev) | |
2724 | { | |
238c0a71 | 2725 | struct imsm_map *lo = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2726 | int level = get_imsm_raid_level(lo); |
2727 | ||
2728 | if (level == 1 || level == 10) { | |
238c0a71 | 2729 | struct imsm_map *hi = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2730 | |
2731 | return hi->num_domains; | |
2732 | } else | |
2733 | return num_stripes_per_unit_resync(dev); | |
2734 | } | |
2735 | ||
98130f40 | 2736 | static __u8 imsm_num_data_members(struct imsm_dev *dev, int second_map) |
1e5c6983 DW |
2737 | { |
2738 | /* named 'imsm_' because raid0, raid1 and raid10 | |
2739 | * counter-intuitively have the same number of data disks | |
2740 | */ | |
98130f40 | 2741 | struct imsm_map *map = get_imsm_map(dev, second_map); |
1e5c6983 DW |
2742 | |
2743 | switch (get_imsm_raid_level(map)) { | |
2744 | case 0: | |
36fd8ccc AK |
2745 | return map->num_members; |
2746 | break; | |
1e5c6983 DW |
2747 | case 1: |
2748 | case 10: | |
36fd8ccc | 2749 | return map->num_members/2; |
1e5c6983 DW |
2750 | case 5: |
2751 | return map->num_members - 1; | |
2752 | default: | |
1ade5cc1 | 2753 | dprintf("unsupported raid level\n"); |
1e5c6983 DW |
2754 | return 0; |
2755 | } | |
2756 | } | |
2757 | ||
2758 | static __u32 parity_segment_depth(struct imsm_dev *dev) | |
2759 | { | |
238c0a71 | 2760 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2761 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2762 | ||
2763 | switch(get_imsm_raid_level(map)) { | |
2764 | case 1: | |
2765 | case 10: | |
2766 | return chunk * map->num_domains; | |
2767 | case 5: | |
2768 | return chunk * map->num_members; | |
2769 | default: | |
2770 | return chunk; | |
2771 | } | |
2772 | } | |
2773 | ||
2774 | static __u32 map_migr_block(struct imsm_dev *dev, __u32 block) | |
2775 | { | |
238c0a71 | 2776 | struct imsm_map *map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
2777 | __u32 chunk = __le32_to_cpu(map->blocks_per_strip); |
2778 | __u32 strip = block / chunk; | |
2779 | ||
2780 | switch (get_imsm_raid_level(map)) { | |
2781 | case 1: | |
2782 | case 10: { | |
2783 | __u32 vol_strip = (strip * map->num_domains) + 1; | |
2784 | __u32 vol_stripe = vol_strip / map->num_members; | |
2785 | ||
2786 | return vol_stripe * chunk + block % chunk; | |
2787 | } case 5: { | |
2788 | __u32 stripe = strip / (map->num_members - 1); | |
2789 | ||
2790 | return stripe * chunk + block % chunk; | |
2791 | } | |
2792 | default: | |
2793 | return 0; | |
2794 | } | |
2795 | } | |
2796 | ||
c47b0ff6 AK |
2797 | static __u64 blocks_per_migr_unit(struct intel_super *super, |
2798 | struct imsm_dev *dev) | |
1e5c6983 DW |
2799 | { |
2800 | /* calculate the conversion factor between per member 'blocks' | |
2801 | * (md/{resync,rebuild}_start) and imsm migration units, return | |
2802 | * 0 for the 'not migrating' and 'unsupported migration' cases | |
2803 | */ | |
2804 | if (!dev->vol.migr_state) | |
2805 | return 0; | |
2806 | ||
2807 | switch (migr_type(dev)) { | |
c47b0ff6 AK |
2808 | case MIGR_GEN_MIGR: { |
2809 | struct migr_record *migr_rec = super->migr_rec; | |
2810 | return __le32_to_cpu(migr_rec->blocks_per_unit); | |
2811 | } | |
1e5c6983 DW |
2812 | case MIGR_VERIFY: |
2813 | case MIGR_REPAIR: | |
2814 | case MIGR_INIT: { | |
238c0a71 | 2815 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
1e5c6983 DW |
2816 | __u32 stripes_per_unit; |
2817 | __u32 blocks_per_unit; | |
2818 | __u32 parity_depth; | |
2819 | __u32 migr_chunk; | |
2820 | __u32 block_map; | |
2821 | __u32 block_rel; | |
2822 | __u32 segment; | |
2823 | __u32 stripe; | |
2824 | __u8 disks; | |
2825 | ||
2826 | /* yes, this is really the translation of migr_units to | |
2827 | * per-member blocks in the 'resync' case | |
2828 | */ | |
2829 | stripes_per_unit = num_stripes_per_unit_resync(dev); | |
2830 | migr_chunk = migr_strip_blocks_resync(dev); | |
238c0a71 | 2831 | disks = imsm_num_data_members(dev, MAP_0); |
1e5c6983 | 2832 | blocks_per_unit = stripes_per_unit * migr_chunk * disks; |
7b1ab482 | 2833 | stripe = __le16_to_cpu(map->blocks_per_strip) * disks; |
1e5c6983 DW |
2834 | segment = blocks_per_unit / stripe; |
2835 | block_rel = blocks_per_unit - segment * stripe; | |
2836 | parity_depth = parity_segment_depth(dev); | |
2837 | block_map = map_migr_block(dev, block_rel); | |
2838 | return block_map + parity_depth * segment; | |
2839 | } | |
2840 | case MIGR_REBUILD: { | |
2841 | __u32 stripes_per_unit; | |
2842 | __u32 migr_chunk; | |
2843 | ||
2844 | stripes_per_unit = num_stripes_per_unit_rebuild(dev); | |
2845 | migr_chunk = migr_strip_blocks_rebuild(dev); | |
2846 | return migr_chunk * stripes_per_unit; | |
2847 | } | |
1e5c6983 DW |
2848 | case MIGR_STATE_CHANGE: |
2849 | default: | |
2850 | return 0; | |
2851 | } | |
2852 | } | |
2853 | ||
c2c087e6 DW |
2854 | static int imsm_level_to_layout(int level) |
2855 | { | |
2856 | switch (level) { | |
2857 | case 0: | |
2858 | case 1: | |
2859 | return 0; | |
2860 | case 5: | |
2861 | case 6: | |
a380c027 | 2862 | return ALGORITHM_LEFT_ASYMMETRIC; |
c2c087e6 | 2863 | case 10: |
c92a2527 | 2864 | return 0x102; |
c2c087e6 | 2865 | } |
a18a888e | 2866 | return UnSet; |
c2c087e6 DW |
2867 | } |
2868 | ||
8e59f3d8 AK |
2869 | /******************************************************************************* |
2870 | * Function: read_imsm_migr_rec | |
2871 | * Description: Function reads imsm migration record from last sector of disk | |
2872 | * Parameters: | |
2873 | * fd : disk descriptor | |
2874 | * super : metadata info | |
2875 | * Returns: | |
2876 | * 0 : success, | |
2877 | * -1 : fail | |
2878 | ******************************************************************************/ | |
2879 | static int read_imsm_migr_rec(int fd, struct intel_super *super) | |
2880 | { | |
2881 | int ret_val = -1; | |
de44e46f | 2882 | unsigned int sector_size = super->sector_size; |
8e59f3d8 AK |
2883 | unsigned long long dsize; |
2884 | ||
2885 | get_dev_size(fd, NULL, &dsize); | |
de44e46f PB |
2886 | if (lseek64(fd, dsize - (sector_size*MIGR_REC_SECTOR_POSITION), |
2887 | SEEK_SET) < 0) { | |
e7b84f9d N |
2888 | pr_err("Cannot seek to anchor block: %s\n", |
2889 | strerror(errno)); | |
8e59f3d8 AK |
2890 | goto out; |
2891 | } | |
466070ad | 2892 | if ((unsigned int)read(fd, super->migr_rec_buf, |
de44e46f PB |
2893 | MIGR_REC_BUF_SECTORS*sector_size) != |
2894 | MIGR_REC_BUF_SECTORS*sector_size) { | |
e7b84f9d N |
2895 | pr_err("Cannot read migr record block: %s\n", |
2896 | strerror(errno)); | |
8e59f3d8 AK |
2897 | goto out; |
2898 | } | |
2899 | ret_val = 0; | |
de44e46f PB |
2900 | if (sector_size == 4096) |
2901 | convert_from_4k_imsm_migr_rec(super); | |
8e59f3d8 AK |
2902 | |
2903 | out: | |
2904 | return ret_val; | |
2905 | } | |
2906 | ||
3136abe5 AK |
2907 | static struct imsm_dev *imsm_get_device_during_migration( |
2908 | struct intel_super *super) | |
2909 | { | |
2910 | ||
2911 | struct intel_dev *dv; | |
2912 | ||
2913 | for (dv = super->devlist; dv; dv = dv->next) { | |
2914 | if (is_gen_migration(dv->dev)) | |
2915 | return dv->dev; | |
2916 | } | |
2917 | return NULL; | |
2918 | } | |
2919 | ||
8e59f3d8 AK |
2920 | /******************************************************************************* |
2921 | * Function: load_imsm_migr_rec | |
2922 | * Description: Function reads imsm migration record (it is stored at the last | |
2923 | * sector of disk) | |
2924 | * Parameters: | |
2925 | * super : imsm internal array info | |
2926 | * info : general array info | |
2927 | * Returns: | |
2928 | * 0 : success | |
2929 | * -1 : fail | |
4c965cc9 | 2930 | * -2 : no migration in progress |
8e59f3d8 AK |
2931 | ******************************************************************************/ |
2932 | static int load_imsm_migr_rec(struct intel_super *super, struct mdinfo *info) | |
2933 | { | |
2934 | struct mdinfo *sd; | |
594dc1b8 | 2935 | struct dl *dl; |
8e59f3d8 AK |
2936 | char nm[30]; |
2937 | int retval = -1; | |
2938 | int fd = -1; | |
3136abe5 | 2939 | struct imsm_dev *dev; |
594dc1b8 | 2940 | struct imsm_map *map; |
b4ab44d8 | 2941 | int slot = -1; |
3136abe5 AK |
2942 | |
2943 | /* find map under migration */ | |
2944 | dev = imsm_get_device_during_migration(super); | |
2945 | /* nothing to load,no migration in progress? | |
2946 | */ | |
2947 | if (dev == NULL) | |
4c965cc9 | 2948 | return -2; |
8e59f3d8 AK |
2949 | |
2950 | if (info) { | |
2951 | for (sd = info->devs ; sd ; sd = sd->next) { | |
2952 | /* read only from one of the first two slots */ | |
12fe93e9 TM |
2953 | if ((sd->disk.raid_disk < 0) || |
2954 | (sd->disk.raid_disk > 1)) | |
8e59f3d8 | 2955 | continue; |
3136abe5 | 2956 | |
8e59f3d8 AK |
2957 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); |
2958 | fd = dev_open(nm, O_RDONLY); | |
2959 | if (fd >= 0) | |
2960 | break; | |
2961 | } | |
2962 | } | |
2963 | if (fd < 0) { | |
12fe93e9 | 2964 | map = get_imsm_map(dev, MAP_0); |
8e59f3d8 | 2965 | for (dl = super->disks; dl; dl = dl->next) { |
3136abe5 AK |
2966 | /* skip spare and failed disks |
2967 | */ | |
2968 | if (dl->index < 0) | |
2969 | continue; | |
8e59f3d8 | 2970 | /* read only from one of the first two slots */ |
3136abe5 AK |
2971 | if (map) |
2972 | slot = get_imsm_disk_slot(map, dl->index); | |
089f9d79 | 2973 | if (map == NULL || slot > 1 || slot < 0) |
8e59f3d8 AK |
2974 | continue; |
2975 | sprintf(nm, "%d:%d", dl->major, dl->minor); | |
2976 | fd = dev_open(nm, O_RDONLY); | |
2977 | if (fd >= 0) | |
2978 | break; | |
2979 | } | |
2980 | } | |
2981 | if (fd < 0) | |
2982 | goto out; | |
2983 | retval = read_imsm_migr_rec(fd, super); | |
2984 | ||
2985 | out: | |
2986 | if (fd >= 0) | |
2987 | close(fd); | |
2988 | return retval; | |
2989 | } | |
2990 | ||
c17608ea AK |
2991 | /******************************************************************************* |
2992 | * function: imsm_create_metadata_checkpoint_update | |
2993 | * Description: It creates update for checkpoint change. | |
2994 | * Parameters: | |
2995 | * super : imsm internal array info | |
2996 | * u : pointer to prepared update | |
2997 | * Returns: | |
2998 | * Uptate length. | |
2999 | * If length is equal to 0, input pointer u contains no update | |
3000 | ******************************************************************************/ | |
3001 | static int imsm_create_metadata_checkpoint_update( | |
3002 | struct intel_super *super, | |
3003 | struct imsm_update_general_migration_checkpoint **u) | |
3004 | { | |
3005 | ||
3006 | int update_memory_size = 0; | |
3007 | ||
1ade5cc1 | 3008 | dprintf("(enter)\n"); |
c17608ea AK |
3009 | |
3010 | if (u == NULL) | |
3011 | return 0; | |
3012 | *u = NULL; | |
3013 | ||
3014 | /* size of all update data without anchor */ | |
3015 | update_memory_size = | |
3016 | sizeof(struct imsm_update_general_migration_checkpoint); | |
3017 | ||
503975b9 | 3018 | *u = xcalloc(1, update_memory_size); |
c17608ea | 3019 | if (*u == NULL) { |
1ade5cc1 | 3020 | dprintf("error: cannot get memory\n"); |
c17608ea AK |
3021 | return 0; |
3022 | } | |
3023 | (*u)->type = update_general_migration_checkpoint; | |
3024 | (*u)->curr_migr_unit = __le32_to_cpu(super->migr_rec->curr_migr_unit); | |
1ade5cc1 | 3025 | dprintf("prepared for %u\n", (*u)->curr_migr_unit); |
c17608ea AK |
3026 | |
3027 | return update_memory_size; | |
3028 | } | |
3029 | ||
c17608ea AK |
3030 | static void imsm_update_metadata_locally(struct supertype *st, |
3031 | void *buf, int len); | |
3032 | ||
687629c2 AK |
3033 | /******************************************************************************* |
3034 | * Function: write_imsm_migr_rec | |
3035 | * Description: Function writes imsm migration record | |
3036 | * (at the last sector of disk) | |
3037 | * Parameters: | |
3038 | * super : imsm internal array info | |
3039 | * Returns: | |
3040 | * 0 : success | |
3041 | * -1 : if fail | |
3042 | ******************************************************************************/ | |
3043 | static int write_imsm_migr_rec(struct supertype *st) | |
3044 | { | |
3045 | struct intel_super *super = st->sb; | |
de44e46f | 3046 | unsigned int sector_size = super->sector_size; |
687629c2 AK |
3047 | unsigned long long dsize; |
3048 | char nm[30]; | |
3049 | int fd = -1; | |
3050 | int retval = -1; | |
3051 | struct dl *sd; | |
c17608ea AK |
3052 | int len; |
3053 | struct imsm_update_general_migration_checkpoint *u; | |
3136abe5 | 3054 | struct imsm_dev *dev; |
594dc1b8 | 3055 | struct imsm_map *map; |
3136abe5 AK |
3056 | |
3057 | /* find map under migration */ | |
3058 | dev = imsm_get_device_during_migration(super); | |
3059 | /* if no migration, write buffer anyway to clear migr_record | |
3060 | * on disk based on first available device | |
3061 | */ | |
3062 | if (dev == NULL) | |
3063 | dev = get_imsm_dev(super, super->current_vol < 0 ? 0 : | |
3064 | super->current_vol); | |
3065 | ||
44bfe6df | 3066 | map = get_imsm_map(dev, MAP_0); |
687629c2 | 3067 | |
de44e46f PB |
3068 | if (sector_size == 4096) |
3069 | convert_to_4k_imsm_migr_rec(super); | |
687629c2 | 3070 | for (sd = super->disks ; sd ; sd = sd->next) { |
b4ab44d8 | 3071 | int slot = -1; |
3136abe5 AK |
3072 | |
3073 | /* skip failed and spare devices */ | |
3074 | if (sd->index < 0) | |
3075 | continue; | |
687629c2 | 3076 | /* write to 2 first slots only */ |
3136abe5 AK |
3077 | if (map) |
3078 | slot = get_imsm_disk_slot(map, sd->index); | |
089f9d79 | 3079 | if (map == NULL || slot > 1 || slot < 0) |
687629c2 | 3080 | continue; |
3136abe5 | 3081 | |
687629c2 AK |
3082 | sprintf(nm, "%d:%d", sd->major, sd->minor); |
3083 | fd = dev_open(nm, O_RDWR); | |
3084 | if (fd < 0) | |
3085 | continue; | |
3086 | get_dev_size(fd, NULL, &dsize); | |
de44e46f PB |
3087 | if (lseek64(fd, dsize - (MIGR_REC_SECTOR_POSITION*sector_size), |
3088 | SEEK_SET) < 0) { | |
e7b84f9d N |
3089 | pr_err("Cannot seek to anchor block: %s\n", |
3090 | strerror(errno)); | |
687629c2 AK |
3091 | goto out; |
3092 | } | |
466070ad | 3093 | if ((unsigned int)write(fd, super->migr_rec_buf, |
de44e46f PB |
3094 | MIGR_REC_BUF_SECTORS*sector_size) != |
3095 | MIGR_REC_BUF_SECTORS*sector_size) { | |
e7b84f9d N |
3096 | pr_err("Cannot write migr record block: %s\n", |
3097 | strerror(errno)); | |
687629c2 AK |
3098 | goto out; |
3099 | } | |
3100 | close(fd); | |
3101 | fd = -1; | |
3102 | } | |
de44e46f PB |
3103 | if (sector_size == 4096) |
3104 | convert_from_4k_imsm_migr_rec(super); | |
c17608ea AK |
3105 | /* update checkpoint information in metadata */ |
3106 | len = imsm_create_metadata_checkpoint_update(super, &u); | |
c17608ea AK |
3107 | if (len <= 0) { |
3108 | dprintf("imsm: Cannot prepare update\n"); | |
3109 | goto out; | |
3110 | } | |
3111 | /* update metadata locally */ | |
3112 | imsm_update_metadata_locally(st, u, len); | |
3113 | /* and possibly remotely */ | |
3114 | if (st->update_tail) { | |
3115 | append_metadata_update(st, u, len); | |
3116 | /* during reshape we do all work inside metadata handler | |
3117 | * manage_reshape(), so metadata update has to be triggered | |
3118 | * insida it | |
3119 | */ | |
3120 | flush_metadata_updates(st); | |
3121 | st->update_tail = &st->updates; | |
3122 | } else | |
3123 | free(u); | |
687629c2 AK |
3124 | |
3125 | retval = 0; | |
3126 | out: | |
3127 | if (fd >= 0) | |
3128 | close(fd); | |
3129 | return retval; | |
3130 | } | |
3131 | ||
e2962bfc AK |
3132 | /* spare/missing disks activations are not allowe when |
3133 | * array/container performs reshape operation, because | |
3134 | * all arrays in container works on the same disks set | |
3135 | */ | |
3136 | int imsm_reshape_blocks_arrays_changes(struct intel_super *super) | |
3137 | { | |
3138 | int rv = 0; | |
3139 | struct intel_dev *i_dev; | |
3140 | struct imsm_dev *dev; | |
3141 | ||
3142 | /* check whole container | |
3143 | */ | |
3144 | for (i_dev = super->devlist; i_dev; i_dev = i_dev->next) { | |
3145 | dev = i_dev->dev; | |
3ad25638 | 3146 | if (is_gen_migration(dev)) { |
e2962bfc AK |
3147 | /* No repair during any migration in container |
3148 | */ | |
3149 | rv = 1; | |
3150 | break; | |
3151 | } | |
3152 | } | |
3153 | return rv; | |
3154 | } | |
c41e00b2 AK |
3155 | static unsigned long long imsm_component_size_aligment_check(int level, |
3156 | int chunk_size, | |
f36a9ecd | 3157 | unsigned int sector_size, |
c41e00b2 AK |
3158 | unsigned long long component_size) |
3159 | { | |
3160 | unsigned int component_size_alligment; | |
3161 | ||
3162 | /* check component size aligment | |
3163 | */ | |
f36a9ecd | 3164 | component_size_alligment = component_size % (chunk_size/sector_size); |
c41e00b2 | 3165 | |
1ade5cc1 | 3166 | dprintf("(Level: %i, chunk_size = %i, component_size = %llu), component_size_alligment = %u\n", |
c41e00b2 AK |
3167 | level, chunk_size, component_size, |
3168 | component_size_alligment); | |
3169 | ||
3170 | if (component_size_alligment && (level != 1) && (level != UnSet)) { | |
3171 | dprintf("imsm: reported component size alligned from %llu ", | |
3172 | component_size); | |
3173 | component_size -= component_size_alligment; | |
1ade5cc1 | 3174 | dprintf_cont("to %llu (%i).\n", |
c41e00b2 AK |
3175 | component_size, component_size_alligment); |
3176 | } | |
3177 | ||
3178 | return component_size; | |
3179 | } | |
e2962bfc | 3180 | |
2432ce9b AP |
3181 | static unsigned long long get_ppl_sector(struct intel_super *super, int dev_idx) |
3182 | { | |
3183 | struct imsm_dev *dev = get_imsm_dev(super, dev_idx); | |
3184 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
3185 | ||
3186 | return pba_of_lba0(map) + | |
3187 | (num_data_stripes(map) * map->blocks_per_strip); | |
3188 | } | |
3189 | ||
a5d85af7 | 3190 | static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap) |
bf5a934a DW |
3191 | { |
3192 | struct intel_super *super = st->sb; | |
c47b0ff6 | 3193 | struct migr_record *migr_rec = super->migr_rec; |
949c47a0 | 3194 | struct imsm_dev *dev = get_imsm_dev(super, super->current_vol); |
238c0a71 AK |
3195 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
3196 | struct imsm_map *prev_map = get_imsm_map(dev, MAP_1); | |
b335e593 | 3197 | struct imsm_map *map_to_analyse = map; |
efb30e7f | 3198 | struct dl *dl; |
a5d85af7 | 3199 | int map_disks = info->array.raid_disks; |
bf5a934a | 3200 | |
95eeceeb | 3201 | memset(info, 0, sizeof(*info)); |
b335e593 AK |
3202 | if (prev_map) |
3203 | map_to_analyse = prev_map; | |
3204 | ||
ca0748fa | 3205 | dl = super->current_disk; |
9894ec0d | 3206 | |
bf5a934a | 3207 | info->container_member = super->current_vol; |
cd0430a1 | 3208 | info->array.raid_disks = map->num_members; |
b335e593 | 3209 | info->array.level = get_imsm_raid_level(map_to_analyse); |
bf5a934a DW |
3210 | info->array.layout = imsm_level_to_layout(info->array.level); |
3211 | info->array.md_minor = -1; | |
3212 | info->array.ctime = 0; | |
3213 | info->array.utime = 0; | |
b335e593 AK |
3214 | info->array.chunk_size = |
3215 | __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9; | |
2432ce9b | 3216 | info->array.state = !(dev->vol.dirty & RAIDVOL_DIRTY); |
da9b4a62 DW |
3217 | info->custom_array_size = __le32_to_cpu(dev->size_high); |
3218 | info->custom_array_size <<= 32; | |
3219 | info->custom_array_size |= __le32_to_cpu(dev->size_low); | |
3ad25638 AK |
3220 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
3221 | ||
3f510843 | 3222 | if (is_gen_migration(dev)) { |
3f83228a | 3223 | info->reshape_active = 1; |
b335e593 AK |
3224 | info->new_level = get_imsm_raid_level(map); |
3225 | info->new_layout = imsm_level_to_layout(info->new_level); | |
3226 | info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9; | |
3f83228a | 3227 | info->delta_disks = map->num_members - prev_map->num_members; |
493f5dd6 N |
3228 | if (info->delta_disks) { |
3229 | /* this needs to be applied to every array | |
3230 | * in the container. | |
3231 | */ | |
81219e70 | 3232 | info->reshape_active = CONTAINER_RESHAPE; |
493f5dd6 | 3233 | } |
3f83228a N |
3234 | /* We shape information that we give to md might have to be |
3235 | * modify to cope with md's requirement for reshaping arrays. | |
3236 | * For example, when reshaping a RAID0, md requires it to be | |
3237 | * presented as a degraded RAID4. | |
3238 | * Also if a RAID0 is migrating to a RAID5 we need to specify | |
3239 | * the array as already being RAID5, but the 'before' layout | |
3240 | * is a RAID4-like layout. | |
3241 | */ | |
3242 | switch (info->array.level) { | |
3243 | case 0: | |
3244 | switch(info->new_level) { | |
3245 | case 0: | |
3246 | /* conversion is happening as RAID4 */ | |
3247 | info->array.level = 4; | |
3248 | info->array.raid_disks += 1; | |
3249 | break; | |
3250 | case 5: | |
3251 | /* conversion is happening as RAID5 */ | |
3252 | info->array.level = 5; | |
3253 | info->array.layout = ALGORITHM_PARITY_N; | |
3f83228a N |
3254 | info->delta_disks -= 1; |
3255 | break; | |
3256 | default: | |
3257 | /* FIXME error message */ | |
3258 | info->array.level = UnSet; | |
3259 | break; | |
3260 | } | |
3261 | break; | |
3262 | } | |
b335e593 AK |
3263 | } else { |
3264 | info->new_level = UnSet; | |
3265 | info->new_layout = UnSet; | |
3266 | info->new_chunk = info->array.chunk_size; | |
3f83228a | 3267 | info->delta_disks = 0; |
b335e593 | 3268 | } |
ca0748fa | 3269 | |
efb30e7f DW |
3270 | if (dl) { |
3271 | info->disk.major = dl->major; | |
3272 | info->disk.minor = dl->minor; | |
ca0748fa | 3273 | info->disk.number = dl->index; |
656b6b5a N |
3274 | info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse, |
3275 | dl->index); | |
efb30e7f | 3276 | } |
bf5a934a | 3277 | |
5551b113 | 3278 | info->data_offset = pba_of_lba0(map_to_analyse); |
06fb291a PB |
3279 | |
3280 | if (info->array.level == 5) { | |
3281 | info->component_size = num_data_stripes(map_to_analyse) * | |
3282 | map_to_analyse->blocks_per_strip; | |
3283 | } else { | |
3284 | info->component_size = blocks_per_member(map_to_analyse); | |
3285 | } | |
139dae11 | 3286 | |
c41e00b2 AK |
3287 | info->component_size = imsm_component_size_aligment_check( |
3288 | info->array.level, | |
3289 | info->array.chunk_size, | |
f36a9ecd | 3290 | super->sector_size, |
c41e00b2 | 3291 | info->component_size); |
5e46202e | 3292 | info->bb.supported = 1; |
139dae11 | 3293 | |
301406c9 | 3294 | memset(info->uuid, 0, sizeof(info->uuid)); |
921d9e16 | 3295 | info->recovery_start = MaxSector; |
bf5a934a | 3296 | |
2432ce9b AP |
3297 | if (info->array.level == 5 && dev->rwh_policy == RWH_DISTRIBUTED) { |
3298 | info->consistency_policy = CONSISTENCY_POLICY_PPL; | |
3299 | info->ppl_sector = get_ppl_sector(super, super->current_vol); | |
3300 | info->ppl_size = (PPL_HEADER_SIZE + PPL_ENTRY_SPACE) >> 9; | |
3301 | } else if (info->array.level <= 0) { | |
3302 | info->consistency_policy = CONSISTENCY_POLICY_NONE; | |
3303 | } else { | |
3304 | info->consistency_policy = CONSISTENCY_POLICY_RESYNC; | |
3305 | } | |
3306 | ||
d2e6d5d6 | 3307 | info->reshape_progress = 0; |
b6796ce1 | 3308 | info->resync_start = MaxSector; |
b9172665 | 3309 | if ((map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED || |
2432ce9b | 3310 | !(info->array.state & 1)) && |
b9172665 | 3311 | imsm_reshape_blocks_arrays_changes(super) == 0) { |
301406c9 | 3312 | info->resync_start = 0; |
b6796ce1 AK |
3313 | } |
3314 | if (dev->vol.migr_state) { | |
1e5c6983 DW |
3315 | switch (migr_type(dev)) { |
3316 | case MIGR_REPAIR: | |
3317 | case MIGR_INIT: { | |
c47b0ff6 AK |
3318 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
3319 | dev); | |
1e5c6983 DW |
3320 | __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit); |
3321 | ||
3322 | info->resync_start = blocks_per_unit * units; | |
3323 | break; | |
3324 | } | |
d2e6d5d6 | 3325 | case MIGR_GEN_MIGR: { |
c47b0ff6 AK |
3326 | __u64 blocks_per_unit = blocks_per_migr_unit(super, |
3327 | dev); | |
3328 | __u64 units = __le32_to_cpu(migr_rec->curr_migr_unit); | |
04fa9523 AK |
3329 | unsigned long long array_blocks; |
3330 | int used_disks; | |
d2e6d5d6 | 3331 | |
befb629b AK |
3332 | if (__le32_to_cpu(migr_rec->ascending_migr) && |
3333 | (units < | |
3334 | (__le32_to_cpu(migr_rec->num_migr_units)-1)) && | |
3335 | (super->migr_rec->rec_status == | |
3336 | __cpu_to_le32(UNIT_SRC_IN_CP_AREA))) | |
3337 | units++; | |
3338 | ||
d2e6d5d6 | 3339 | info->reshape_progress = blocks_per_unit * units; |
6289d1e0 | 3340 | |
7a862a02 | 3341 | dprintf("IMSM: General Migration checkpoint : %llu (%llu) -> read reshape progress : %llu\n", |
19986c72 MB |
3342 | (unsigned long long)units, |
3343 | (unsigned long long)blocks_per_unit, | |
3344 | info->reshape_progress); | |
75156c46 | 3345 | |
238c0a71 | 3346 | used_disks = imsm_num_data_members(dev, MAP_1); |
75156c46 | 3347 | if (used_disks > 0) { |
5551b113 | 3348 | array_blocks = blocks_per_member(map) * |
75156c46 | 3349 | used_disks; |
b53bfba6 TM |
3350 | info->custom_array_size = |
3351 | round_size_to_mb(array_blocks, | |
3352 | used_disks); | |
3353 | ||
75156c46 | 3354 | } |
d2e6d5d6 | 3355 | } |
1e5c6983 DW |
3356 | case MIGR_VERIFY: |
3357 | /* we could emulate the checkpointing of | |
3358 | * 'sync_action=check' migrations, but for now | |
3359 | * we just immediately complete them | |
3360 | */ | |
3361 | case MIGR_REBUILD: | |
3362 | /* this is handled by container_content_imsm() */ | |
1e5c6983 DW |
3363 | case MIGR_STATE_CHANGE: |
3364 | /* FIXME handle other migrations */ | |
3365 | default: | |
3366 | /* we are not dirty, so... */ | |
3367 | info->resync_start = MaxSector; | |
3368 | } | |
b6796ce1 | 3369 | } |
301406c9 DW |
3370 | |
3371 | strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN); | |
3372 | info->name[MAX_RAID_SERIAL_LEN] = 0; | |
bf5a934a | 3373 | |
f35f2525 N |
3374 | info->array.major_version = -1; |
3375 | info->array.minor_version = -2; | |
4dd2df09 | 3376 | sprintf(info->text_version, "/%s/%d", st->container_devnm, info->container_member); |
a67dd8cc | 3377 | info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */ |
51006d85 | 3378 | uuid_from_super_imsm(st, info->uuid); |
a5d85af7 N |
3379 | |
3380 | if (dmap) { | |
3381 | int i, j; | |
3382 | for (i=0; i<map_disks; i++) { | |
3383 | dmap[i] = 0; | |
3384 | if (i < info->array.raid_disks) { | |
3385 | struct imsm_disk *dsk; | |
238c0a71 | 3386 | j = get_imsm_disk_idx(dev, i, MAP_X); |
a5d85af7 N |
3387 | dsk = get_imsm_disk(super, j); |
3388 | if (dsk && (dsk->status & CONFIGURED_DISK)) | |
3389 | dmap[i] = 1; | |
3390 | } | |
3391 | } | |
3392 | } | |
81ac8b4d | 3393 | } |
bf5a934a | 3394 | |
3b451610 AK |
3395 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, |
3396 | int failed, int look_in_map); | |
3397 | ||
3398 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev, | |
3399 | int look_in_map); | |
3400 | ||
3401 | static void manage_second_map(struct intel_super *super, struct imsm_dev *dev) | |
3402 | { | |
3403 | if (is_gen_migration(dev)) { | |
3404 | int failed; | |
3405 | __u8 map_state; | |
3406 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); | |
3407 | ||
3408 | failed = imsm_count_failed(super, dev, MAP_1); | |
238c0a71 | 3409 | map_state = imsm_check_degraded(super, dev, failed, MAP_1); |
3b451610 AK |
3410 | if (map2->map_state != map_state) { |
3411 | map2->map_state = map_state; | |
3412 | super->updates_pending++; | |
3413 | } | |
3414 | } | |
3415 | } | |
97b4d0e9 DW |
3416 | |
3417 | static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index) | |
3418 | { | |
3419 | struct dl *d; | |
3420 | ||
3421 | for (d = super->missing; d; d = d->next) | |
3422 | if (d->index == index) | |
3423 | return &d->disk; | |
3424 | return NULL; | |
3425 | } | |
3426 | ||
a5d85af7 | 3427 | static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map) |
4f5bc454 DW |
3428 | { |
3429 | struct intel_super *super = st->sb; | |
4f5bc454 | 3430 | struct imsm_disk *disk; |
a5d85af7 | 3431 | int map_disks = info->array.raid_disks; |
ab3cb6b3 N |
3432 | int max_enough = -1; |
3433 | int i; | |
3434 | struct imsm_super *mpb; | |
4f5bc454 | 3435 | |
bf5a934a | 3436 | if (super->current_vol >= 0) { |
a5d85af7 | 3437 | getinfo_super_imsm_volume(st, info, map); |
bf5a934a DW |
3438 | return; |
3439 | } | |
95eeceeb | 3440 | memset(info, 0, sizeof(*info)); |
d23fe947 DW |
3441 | |
3442 | /* Set raid_disks to zero so that Assemble will always pull in valid | |
3443 | * spares | |
3444 | */ | |
3445 | info->array.raid_disks = 0; | |
cdddbdbc DW |
3446 | info->array.level = LEVEL_CONTAINER; |
3447 | info->array.layout = 0; | |
3448 | info->array.md_minor = -1; | |
1011e834 | 3449 | info->array.ctime = 0; /* N/A for imsm */ |
cdddbdbc DW |
3450 | info->array.utime = 0; |
3451 | info->array.chunk_size = 0; | |
3452 | ||
3453 | info->disk.major = 0; | |
3454 | info->disk.minor = 0; | |
cdddbdbc | 3455 | info->disk.raid_disk = -1; |
c2c087e6 | 3456 | info->reshape_active = 0; |
f35f2525 N |
3457 | info->array.major_version = -1; |
3458 | info->array.minor_version = -2; | |
c2c087e6 | 3459 | strcpy(info->text_version, "imsm"); |
a67dd8cc | 3460 | info->safe_mode_delay = 0; |
c2c087e6 DW |
3461 | info->disk.number = -1; |
3462 | info->disk.state = 0; | |
c5afc314 | 3463 | info->name[0] = 0; |
921d9e16 | 3464 | info->recovery_start = MaxSector; |
3ad25638 | 3465 | info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb); |
5e46202e | 3466 | info->bb.supported = 1; |
c2c087e6 | 3467 | |
97b4d0e9 | 3468 | /* do we have the all the insync disks that we expect? */ |
ab3cb6b3 | 3469 | mpb = super->anchor; |
b7d81a38 | 3470 | info->events = __le32_to_cpu(mpb->generation_num); |
97b4d0e9 | 3471 | |
ab3cb6b3 N |
3472 | for (i = 0; i < mpb->num_raid_devs; i++) { |
3473 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
3474 | int failed, enough, j, missing = 0; | |
3475 | struct imsm_map *map; | |
3476 | __u8 state; | |
97b4d0e9 | 3477 | |
3b451610 AK |
3478 | failed = imsm_count_failed(super, dev, MAP_0); |
3479 | state = imsm_check_degraded(super, dev, failed, MAP_0); | |
238c0a71 | 3480 | map = get_imsm_map(dev, MAP_0); |
ab3cb6b3 N |
3481 | |
3482 | /* any newly missing disks? | |
3483 | * (catches single-degraded vs double-degraded) | |
3484 | */ | |
3485 | for (j = 0; j < map->num_members; j++) { | |
238c0a71 | 3486 | __u32 ord = get_imsm_ord_tbl_ent(dev, j, MAP_0); |
ab3cb6b3 N |
3487 | __u32 idx = ord_to_idx(ord); |
3488 | ||
3489 | if (!(ord & IMSM_ORD_REBUILD) && | |
3490 | get_imsm_missing(super, idx)) { | |
3491 | missing = 1; | |
3492 | break; | |
3493 | } | |
97b4d0e9 | 3494 | } |
ab3cb6b3 N |
3495 | |
3496 | if (state == IMSM_T_STATE_FAILED) | |
3497 | enough = -1; | |
3498 | else if (state == IMSM_T_STATE_DEGRADED && | |
3499 | (state != map->map_state || missing)) | |
3500 | enough = 0; | |
3501 | else /* we're normal, or already degraded */ | |
3502 | enough = 1; | |
d2bde6d3 AK |
3503 | if (is_gen_migration(dev) && missing) { |
3504 | /* during general migration we need all disks | |
3505 | * that process is running on. | |
3506 | * No new missing disk is allowed. | |
3507 | */ | |
3508 | max_enough = -1; | |
3509 | enough = -1; | |
3510 | /* no more checks necessary | |
3511 | */ | |
3512 | break; | |
3513 | } | |
ab3cb6b3 N |
3514 | /* in the missing/failed disk case check to see |
3515 | * if at least one array is runnable | |
3516 | */ | |
3517 | max_enough = max(max_enough, enough); | |
3518 | } | |
1ade5cc1 | 3519 | dprintf("enough: %d\n", max_enough); |
ab3cb6b3 | 3520 | info->container_enough = max_enough; |
97b4d0e9 | 3521 | |
4a04ec6c | 3522 | if (super->disks) { |
14e8215b DW |
3523 | __u32 reserved = imsm_reserved_sectors(super, super->disks); |
3524 | ||
b9f594fe | 3525 | disk = &super->disks->disk; |
5551b113 | 3526 | info->data_offset = total_blocks(&super->disks->disk) - reserved; |
14e8215b | 3527 | info->component_size = reserved; |
25ed7e59 | 3528 | info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0; |
df474657 DW |
3529 | /* we don't change info->disk.raid_disk here because |
3530 | * this state will be finalized in mdmon after we have | |
3531 | * found the 'most fresh' version of the metadata | |
3532 | */ | |
25ed7e59 | 3533 | info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; |
2432ce9b AP |
3534 | info->disk.state |= (is_spare(disk) || is_journal(disk)) ? |
3535 | 0 : (1 << MD_DISK_SYNC); | |
cdddbdbc | 3536 | } |
a575e2a7 DW |
3537 | |
3538 | /* only call uuid_from_super_imsm when this disk is part of a populated container, | |
3539 | * ->compare_super may have updated the 'num_raid_devs' field for spares | |
3540 | */ | |
3541 | if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs) | |
36ba7d48 | 3542 | uuid_from_super_imsm(st, info->uuid); |
22e263f6 AC |
3543 | else |
3544 | memcpy(info->uuid, uuid_zero, sizeof(uuid_zero)); | |
a5d85af7 N |
3545 | |
3546 | /* I don't know how to compute 'map' on imsm, so use safe default */ | |
3547 | if (map) { | |
3548 | int i; | |
3549 | for (i = 0; i < map_disks; i++) | |
3550 | map[i] = 1; | |
3551 | } | |
3552 | ||
cdddbdbc DW |
3553 | } |
3554 | ||
5c4cd5da AC |
3555 | /* allocates memory and fills disk in mdinfo structure |
3556 | * for each disk in array */ | |
3557 | struct mdinfo *getinfo_super_disks_imsm(struct supertype *st) | |
3558 | { | |
594dc1b8 | 3559 | struct mdinfo *mddev; |
5c4cd5da AC |
3560 | struct intel_super *super = st->sb; |
3561 | struct imsm_disk *disk; | |
3562 | int count = 0; | |
3563 | struct dl *dl; | |
3564 | if (!super || !super->disks) | |
3565 | return NULL; | |
3566 | dl = super->disks; | |
503975b9 | 3567 | mddev = xcalloc(1, sizeof(*mddev)); |
5c4cd5da AC |
3568 | while (dl) { |
3569 | struct mdinfo *tmp; | |
3570 | disk = &dl->disk; | |
503975b9 | 3571 | tmp = xcalloc(1, sizeof(*tmp)); |
5c4cd5da AC |
3572 | if (mddev->devs) |
3573 | tmp->next = mddev->devs; | |
3574 | mddev->devs = tmp; | |
3575 | tmp->disk.number = count++; | |
3576 | tmp->disk.major = dl->major; | |
3577 | tmp->disk.minor = dl->minor; | |
3578 | tmp->disk.state = is_configured(disk) ? | |
3579 | (1 << MD_DISK_ACTIVE) : 0; | |
3580 | tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0; | |
3581 | tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC); | |
3582 | tmp->disk.raid_disk = -1; | |
3583 | dl = dl->next; | |
3584 | } | |
3585 | return mddev; | |
3586 | } | |
3587 | ||
cdddbdbc DW |
3588 | static int update_super_imsm(struct supertype *st, struct mdinfo *info, |
3589 | char *update, char *devname, int verbose, | |
3590 | int uuid_set, char *homehost) | |
3591 | { | |
f352c545 DW |
3592 | /* For 'assemble' and 'force' we need to return non-zero if any |
3593 | * change was made. For others, the return value is ignored. | |
3594 | * Update options are: | |
3595 | * force-one : This device looks a bit old but needs to be included, | |
3596 | * update age info appropriately. | |
3597 | * assemble: clear any 'faulty' flag to allow this device to | |
3598 | * be assembled. | |
3599 | * force-array: Array is degraded but being forced, mark it clean | |
3600 | * if that will be needed to assemble it. | |
3601 | * | |
3602 | * newdev: not used ???? | |
3603 | * grow: Array has gained a new device - this is currently for | |
3604 | * linear only | |
3605 | * resync: mark as dirty so a resync will happen. | |
3606 | * name: update the name - preserving the homehost | |
6e46bf34 | 3607 | * uuid: Change the uuid of the array to match watch is given |
f352c545 DW |
3608 | * |
3609 | * Following are not relevant for this imsm: | |
3610 | * sparc2.2 : update from old dodgey metadata | |
3611 | * super-minor: change the preferred_minor number | |
3612 | * summaries: update redundant counters. | |
f352c545 DW |
3613 | * homehost: update the recorded homehost |
3614 | * _reshape_progress: record new reshape_progress position. | |
3615 | */ | |
6e46bf34 DW |
3616 | int rv = 1; |
3617 | struct intel_super *super = st->sb; | |
3618 | struct imsm_super *mpb; | |
f352c545 | 3619 | |
6e46bf34 DW |
3620 | /* we can only update container info */ |
3621 | if (!super || super->current_vol >= 0 || !super->anchor) | |
3622 | return 1; | |
3623 | ||
3624 | mpb = super->anchor; | |
3625 | ||
81a5b4f5 N |
3626 | if (strcmp(update, "uuid") == 0) { |
3627 | /* We take this to mean that the family_num should be updated. | |
3628 | * However that is much smaller than the uuid so we cannot really | |
3629 | * allow an explicit uuid to be given. And it is hard to reliably | |
3630 | * know if one was. | |
3631 | * So if !uuid_set we know the current uuid is random and just used | |
3632 | * the first 'int' and copy it to the other 3 positions. | |
3633 | * Otherwise we require the 4 'int's to be the same as would be the | |
3634 | * case if we are using a random uuid. So an explicit uuid will be | |
3635 | * accepted as long as all for ints are the same... which shouldn't hurt | |
6e46bf34 | 3636 | */ |
81a5b4f5 N |
3637 | if (!uuid_set) { |
3638 | info->uuid[1] = info->uuid[2] = info->uuid[3] = info->uuid[0]; | |
6e46bf34 | 3639 | rv = 0; |
81a5b4f5 N |
3640 | } else { |
3641 | if (info->uuid[0] != info->uuid[1] || | |
3642 | info->uuid[1] != info->uuid[2] || | |
3643 | info->uuid[2] != info->uuid[3]) | |
3644 | rv = -1; | |
3645 | else | |
3646 | rv = 0; | |
6e46bf34 | 3647 | } |
81a5b4f5 N |
3648 | if (rv == 0) |
3649 | mpb->orig_family_num = info->uuid[0]; | |
6e46bf34 DW |
3650 | } else if (strcmp(update, "assemble") == 0) |
3651 | rv = 0; | |
3652 | else | |
1e2b2765 | 3653 | rv = -1; |
f352c545 | 3654 | |
6e46bf34 DW |
3655 | /* successful update? recompute checksum */ |
3656 | if (rv == 0) | |
3657 | mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb)); | |
f352c545 DW |
3658 | |
3659 | return rv; | |
cdddbdbc DW |
3660 | } |
3661 | ||
c2c087e6 | 3662 | static size_t disks_to_mpb_size(int disks) |
cdddbdbc | 3663 | { |
c2c087e6 | 3664 | size_t size; |
cdddbdbc | 3665 | |
c2c087e6 DW |
3666 | size = sizeof(struct imsm_super); |
3667 | size += (disks - 1) * sizeof(struct imsm_disk); | |
3668 | size += 2 * sizeof(struct imsm_dev); | |
3669 | /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */ | |
3670 | size += (4 - 2) * sizeof(struct imsm_map); | |
3671 | /* 4 possible disk_ord_tbl's */ | |
3672 | size += 4 * (disks - 1) * sizeof(__u32); | |
bbab0940 TM |
3673 | /* maximum bbm log */ |
3674 | size += sizeof(struct bbm_log); | |
c2c087e6 DW |
3675 | |
3676 | return size; | |
3677 | } | |
3678 | ||
387fcd59 N |
3679 | static __u64 avail_size_imsm(struct supertype *st, __u64 devsize, |
3680 | unsigned long long data_offset) | |
c2c087e6 DW |
3681 | { |
3682 | if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS)) | |
3683 | return 0; | |
3684 | ||
3685 | return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS); | |
cdddbdbc DW |
3686 | } |
3687 | ||
ba2de7ba DW |
3688 | static void free_devlist(struct intel_super *super) |
3689 | { | |
3690 | struct intel_dev *dv; | |
3691 | ||
3692 | while (super->devlist) { | |
3693 | dv = super->devlist->next; | |
3694 | free(super->devlist->dev); | |
3695 | free(super->devlist); | |
3696 | super->devlist = dv; | |
3697 | } | |
3698 | } | |
3699 | ||
3700 | static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src) | |
3701 | { | |
3702 | memcpy(dest, src, sizeof_imsm_dev(src, 0)); | |
3703 | } | |
3704 | ||
cdddbdbc DW |
3705 | static int compare_super_imsm(struct supertype *st, struct supertype *tst) |
3706 | { | |
3707 | /* | |
3708 | * return: | |
3709 | * 0 same, or first was empty, and second was copied | |
3710 | * 1 second had wrong number | |
3711 | * 2 wrong uuid | |
3712 | * 3 wrong other info | |
3713 | */ | |
3714 | struct intel_super *first = st->sb; | |
3715 | struct intel_super *sec = tst->sb; | |
3716 | ||
5d500228 N |
3717 | if (!first) { |
3718 | st->sb = tst->sb; | |
3719 | tst->sb = NULL; | |
3720 | return 0; | |
3721 | } | |
8603ea6f LM |
3722 | /* in platform dependent environment test if the disks |
3723 | * use the same Intel hba | |
cb8f6859 | 3724 | * If not on Intel hba at all, allow anything. |
8603ea6f | 3725 | */ |
6b781d33 AP |
3726 | if (!check_env("IMSM_NO_PLATFORM") && first->hba && sec->hba) { |
3727 | if (first->hba->type != sec->hba->type) { | |
8603ea6f | 3728 | fprintf(stderr, |
6b781d33 AP |
3729 | "HBAs of devices do not match %s != %s\n", |
3730 | get_sys_dev_type(first->hba->type), | |
3731 | get_sys_dev_type(sec->hba->type)); | |
3732 | return 3; | |
3733 | } | |
3734 | if (first->orom != sec->orom) { | |
3735 | fprintf(stderr, | |
3736 | "HBAs of devices do not match %s != %s\n", | |
3737 | first->hba->pci_id, sec->hba->pci_id); | |
8603ea6f LM |
3738 | return 3; |
3739 | } | |
3740 | } | |
cdddbdbc | 3741 | |
d23fe947 DW |
3742 | /* if an anchor does not have num_raid_devs set then it is a free |
3743 | * floating spare | |
3744 | */ | |
3745 | if (first->anchor->num_raid_devs > 0 && | |
3746 | sec->anchor->num_raid_devs > 0) { | |
a2b97981 DW |
3747 | /* Determine if these disks might ever have been |
3748 | * related. Further disambiguation can only take place | |
3749 | * in load_super_imsm_all | |
3750 | */ | |
3751 | __u32 first_family = first->anchor->orig_family_num; | |
3752 | __u32 sec_family = sec->anchor->orig_family_num; | |
3753 | ||
f796af5d DW |
3754 | if (memcmp(first->anchor->sig, sec->anchor->sig, |
3755 | MAX_SIGNATURE_LENGTH) != 0) | |
3756 | return 3; | |
3757 | ||
a2b97981 DW |
3758 | if (first_family == 0) |
3759 | first_family = first->anchor->family_num; | |
3760 | if (sec_family == 0) | |
3761 | sec_family = sec->anchor->family_num; | |
3762 | ||
3763 | if (first_family != sec_family) | |
d23fe947 | 3764 | return 3; |
f796af5d | 3765 | |
d23fe947 | 3766 | } |
cdddbdbc | 3767 | |
3e372e5a DW |
3768 | /* if 'first' is a spare promote it to a populated mpb with sec's |
3769 | * family number | |
3770 | */ | |
3771 | if (first->anchor->num_raid_devs == 0 && | |
3772 | sec->anchor->num_raid_devs > 0) { | |
78d30f94 | 3773 | int i; |
ba2de7ba DW |
3774 | struct intel_dev *dv; |
3775 | struct imsm_dev *dev; | |
78d30f94 DW |
3776 | |
3777 | /* we need to copy raid device info from sec if an allocation | |
3778 | * fails here we don't associate the spare | |
3779 | */ | |
3780 | for (i = 0; i < sec->anchor->num_raid_devs; i++) { | |
503975b9 N |
3781 | dv = xmalloc(sizeof(*dv)); |
3782 | dev = xmalloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1)); | |
ba2de7ba DW |
3783 | dv->dev = dev; |
3784 | dv->index = i; | |
3785 | dv->next = first->devlist; | |
3786 | first->devlist = dv; | |
78d30f94 | 3787 | } |
709743c5 | 3788 | if (i < sec->anchor->num_raid_devs) { |
ba2de7ba DW |
3789 | /* allocation failure */ |
3790 | free_devlist(first); | |
e12b3daa | 3791 | pr_err("imsm: failed to associate spare\n"); |
ba2de7ba | 3792 | return 3; |
78d30f94 | 3793 | } |
3e372e5a | 3794 | first->anchor->num_raid_devs = sec->anchor->num_raid_devs; |
148acb7b | 3795 | first->anchor->orig_family_num = sec->anchor->orig_family_num; |
3e372e5a | 3796 | first->anchor->family_num = sec->anchor->family_num; |
ac6449be | 3797 | memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH); |
709743c5 DW |
3798 | for (i = 0; i < sec->anchor->num_raid_devs; i++) |
3799 | imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i)); | |
3e372e5a DW |
3800 | } |
3801 | ||
cdddbdbc DW |
3802 | return 0; |
3803 | } | |
3804 | ||
0030e8d6 DW |
3805 | static void fd2devname(int fd, char *name) |
3806 | { | |
3807 | struct stat st; | |
3808 | char path[256]; | |
33a6535d | 3809 | char dname[PATH_MAX]; |
0030e8d6 DW |
3810 | char *nm; |
3811 | int rv; | |
3812 | ||
3813 | name[0] = '\0'; | |
3814 | if (fstat(fd, &st) != 0) | |
3815 | return; | |
3816 | sprintf(path, "/sys/dev/block/%d:%d", | |
3817 | major(st.st_rdev), minor(st.st_rdev)); | |
3818 | ||
9cf014ec | 3819 | rv = readlink(path, dname, sizeof(dname)-1); |
0030e8d6 DW |
3820 | if (rv <= 0) |
3821 | return; | |
9587c373 | 3822 | |
0030e8d6 DW |
3823 | dname[rv] = '\0'; |
3824 | nm = strrchr(dname, '/'); | |
7897de29 JS |
3825 | if (nm) { |
3826 | nm++; | |
3827 | snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm); | |
3828 | } | |
0030e8d6 DW |
3829 | } |
3830 | ||
21e9380b AP |
3831 | static int nvme_get_serial(int fd, void *buf, size_t buf_len) |
3832 | { | |
3833 | char path[60]; | |
3834 | char *name = fd2kname(fd); | |
3835 | ||
3836 | if (!name) | |
3837 | return 1; | |
3838 | ||
3839 | if (strncmp(name, "nvme", 4) != 0) | |
3840 | return 1; | |
3841 | ||
3842 | snprintf(path, sizeof(path) - 1, "/sys/block/%s/device/serial", name); | |
3843 | ||
3844 | return load_sys(path, buf, buf_len); | |
3845 | } | |
3846 | ||
cdddbdbc DW |
3847 | extern int scsi_get_serial(int fd, void *buf, size_t buf_len); |
3848 | ||
3849 | static int imsm_read_serial(int fd, char *devname, | |
3850 | __u8 serial[MAX_RAID_SERIAL_LEN]) | |
3851 | { | |
21e9380b | 3852 | char buf[50]; |
cdddbdbc | 3853 | int rv; |
1f24f035 | 3854 | int len; |
316e2bf4 DW |
3855 | char *dest; |
3856 | char *src; | |
21e9380b AP |
3857 | unsigned int i; |
3858 | ||
3859 | memset(buf, 0, sizeof(buf)); | |
cdddbdbc | 3860 | |
21e9380b | 3861 | rv = nvme_get_serial(fd, buf, sizeof(buf)); |
cdddbdbc | 3862 | |
21e9380b AP |
3863 | if (rv) |
3864 | rv = scsi_get_serial(fd, buf, sizeof(buf)); | |
f9ba0ff1 | 3865 | |
40ebbb9c | 3866 | if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) { |
f9ba0ff1 DW |
3867 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
3868 | fd2devname(fd, (char *) serial); | |
0030e8d6 DW |
3869 | return 0; |
3870 | } | |
3871 | ||
cdddbdbc DW |
3872 | if (rv != 0) { |
3873 | if (devname) | |
e7b84f9d N |
3874 | pr_err("Failed to retrieve serial for %s\n", |
3875 | devname); | |
cdddbdbc DW |
3876 | return rv; |
3877 | } | |
3878 | ||
316e2bf4 DW |
3879 | /* trim all whitespace and non-printable characters and convert |
3880 | * ':' to ';' | |
3881 | */ | |
21e9380b AP |
3882 | for (i = 0, dest = buf; i < sizeof(buf) && buf[i]; i++) { |
3883 | src = &buf[i]; | |
316e2bf4 DW |
3884 | if (*src > 0x20) { |
3885 | /* ':' is reserved for use in placeholder serial | |
3886 | * numbers for missing disks | |
3887 | */ | |
3888 | if (*src == ':') | |
3889 | *dest++ = ';'; | |
3890 | else | |
3891 | *dest++ = *src; | |
3892 | } | |
3893 | } | |
21e9380b AP |
3894 | len = dest - buf; |
3895 | dest = buf; | |
316e2bf4 DW |
3896 | |
3897 | /* truncate leading characters */ | |
3898 | if (len > MAX_RAID_SERIAL_LEN) { | |
3899 | dest += len - MAX_RAID_SERIAL_LEN; | |
1f24f035 | 3900 | len = MAX_RAID_SERIAL_LEN; |
316e2bf4 | 3901 | } |
5c3db629 | 3902 | |
5c3db629 | 3903 | memset(serial, 0, MAX_RAID_SERIAL_LEN); |
316e2bf4 | 3904 | memcpy(serial, dest, len); |
cdddbdbc DW |
3905 | |
3906 | return 0; | |
3907 | } | |
3908 | ||
1f24f035 DW |
3909 | static int serialcmp(__u8 *s1, __u8 *s2) |
3910 | { | |
3911 | return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN); | |
3912 | } | |
3913 | ||
3914 | static void serialcpy(__u8 *dest, __u8 *src) | |
3915 | { | |
3916 | strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN); | |
3917 | } | |
3918 | ||
54c2c1ea DW |
3919 | static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super) |
3920 | { | |
3921 | struct dl *dl; | |
3922 | ||
3923 | for (dl = super->disks; dl; dl = dl->next) | |
3924 | if (serialcmp(dl->serial, serial) == 0) | |
3925 | break; | |
3926 | ||
3927 | return dl; | |
3928 | } | |
3929 | ||
a2b97981 DW |
3930 | static struct imsm_disk * |
3931 | __serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx) | |
3932 | { | |
3933 | int i; | |
3934 | ||
3935 | for (i = 0; i < mpb->num_disks; i++) { | |
3936 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
3937 | ||
3938 | if (serialcmp(disk->serial, serial) == 0) { | |
3939 | if (idx) | |
3940 | *idx = i; | |
3941 | return disk; | |
3942 | } | |
3943 | } | |
3944 | ||
3945 | return NULL; | |
3946 | } | |
3947 | ||
cdddbdbc DW |
3948 | static int |
3949 | load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd) | |
3950 | { | |
a2b97981 | 3951 | struct imsm_disk *disk; |
cdddbdbc DW |
3952 | struct dl *dl; |
3953 | struct stat stb; | |
cdddbdbc | 3954 | int rv; |
a2b97981 | 3955 | char name[40]; |
d23fe947 DW |
3956 | __u8 serial[MAX_RAID_SERIAL_LEN]; |
3957 | ||
3958 | rv = imsm_read_serial(fd, devname, serial); | |
3959 | ||
3960 | if (rv != 0) | |
3961 | return 2; | |
3962 | ||
503975b9 | 3963 | dl = xcalloc(1, sizeof(*dl)); |
cdddbdbc | 3964 | |
a2b97981 DW |
3965 | fstat(fd, &stb); |
3966 | dl->major = major(stb.st_rdev); | |
3967 | dl->minor = minor(stb.st_rdev); | |
3968 | dl->next = super->disks; | |
3969 | dl->fd = keep_fd ? fd : -1; | |
3970 | assert(super->disks == NULL); | |
3971 | super->disks = dl; | |
3972 | serialcpy(dl->serial, serial); | |
3973 | dl->index = -2; | |
3974 | dl->e = NULL; | |
3975 | fd2devname(fd, name); | |
3976 | if (devname) | |
503975b9 | 3977 | dl->devname = xstrdup(devname); |
a2b97981 | 3978 | else |
503975b9 | 3979 | dl->devname = xstrdup(name); |
cdddbdbc | 3980 | |
d23fe947 | 3981 | /* look up this disk's index in the current anchor */ |
a2b97981 DW |
3982 | disk = __serial_to_disk(dl->serial, super->anchor, &dl->index); |
3983 | if (disk) { | |
3984 | dl->disk = *disk; | |
3985 | /* only set index on disks that are a member of a | |
3986 | * populated contianer, i.e. one with raid_devs | |
3987 | */ | |
3988 | if (is_failed(&dl->disk)) | |
3f6efecc | 3989 | dl->index = -2; |
2432ce9b | 3990 | else if (is_spare(&dl->disk) || is_journal(&dl->disk)) |
a2b97981 | 3991 | dl->index = -1; |
3f6efecc DW |
3992 | } |
3993 | ||
949c47a0 DW |
3994 | return 0; |
3995 | } | |
3996 | ||
0c046afd DW |
3997 | /* When migrating map0 contains the 'destination' state while map1 |
3998 | * contains the current state. When not migrating map0 contains the | |
3999 | * current state. This routine assumes that map[0].map_state is set to | |
4000 | * the current array state before being called. | |
4001 | * | |
4002 | * Migration is indicated by one of the following states | |
4003 | * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed) | |
e3bba0e0 | 4004 | * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal |
0c046afd | 4005 | * map1state=unitialized) |
1484e727 | 4006 | * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal |
0c046afd | 4007 | * map1state=normal) |
e3bba0e0 | 4008 | * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal |
0c046afd | 4009 | * map1state=degraded) |
8e59f3d8 AK |
4010 | * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal |
4011 | * map1state=normal) | |
0c046afd | 4012 | */ |
8e59f3d8 AK |
4013 | static void migrate(struct imsm_dev *dev, struct intel_super *super, |
4014 | __u8 to_state, int migr_type) | |
3393c6af | 4015 | { |
0c046afd | 4016 | struct imsm_map *dest; |
238c0a71 | 4017 | struct imsm_map *src = get_imsm_map(dev, MAP_0); |
3393c6af | 4018 | |
0c046afd | 4019 | dev->vol.migr_state = 1; |
1484e727 | 4020 | set_migr_type(dev, migr_type); |
f8f603f1 | 4021 | dev->vol.curr_migr_unit = 0; |
238c0a71 | 4022 | dest = get_imsm_map(dev, MAP_1); |
0c046afd | 4023 | |
0556e1a2 | 4024 | /* duplicate and then set the target end state in map[0] */ |
3393c6af | 4025 | memcpy(dest, src, sizeof_imsm_map(src)); |
fb12a745 | 4026 | if (migr_type == MIGR_GEN_MIGR) { |
0556e1a2 DW |
4027 | __u32 ord; |
4028 | int i; | |
4029 | ||
4030 | for (i = 0; i < src->num_members; i++) { | |
4031 | ord = __le32_to_cpu(src->disk_ord_tbl[i]); | |
4032 | set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord)); | |
4033 | } | |
4034 | } | |
4035 | ||
8e59f3d8 AK |
4036 | if (migr_type == MIGR_GEN_MIGR) |
4037 | /* Clear migration record */ | |
4038 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
4039 | ||
0c046afd | 4040 | src->map_state = to_state; |
949c47a0 | 4041 | } |
f8f603f1 | 4042 | |
809da78e AK |
4043 | static void end_migration(struct imsm_dev *dev, struct intel_super *super, |
4044 | __u8 map_state) | |
f8f603f1 | 4045 | { |
238c0a71 AK |
4046 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
4047 | struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state == 0 ? | |
4048 | MAP_0 : MAP_1); | |
28bce06f | 4049 | int i, j; |
0556e1a2 DW |
4050 | |
4051 | /* merge any IMSM_ORD_REBUILD bits that were not successfully | |
4052 | * completed in the last migration. | |
4053 | * | |
28bce06f | 4054 | * FIXME add support for raid-level-migration |
0556e1a2 | 4055 | */ |
089f9d79 JS |
4056 | if (map_state != map->map_state && (is_gen_migration(dev) == 0) && |
4057 | prev->map_state != IMSM_T_STATE_UNINITIALIZED) { | |
809da78e AK |
4058 | /* when final map state is other than expected |
4059 | * merge maps (not for migration) | |
4060 | */ | |
4061 | int failed; | |
4062 | ||
4063 | for (i = 0; i < prev->num_members; i++) | |
4064 | for (j = 0; j < map->num_members; j++) | |
4065 | /* during online capacity expansion | |
4066 | * disks position can be changed | |
4067 | * if takeover is used | |
4068 | */ | |
4069 | if (ord_to_idx(map->disk_ord_tbl[j]) == | |
4070 | ord_to_idx(prev->disk_ord_tbl[i])) { | |
4071 | map->disk_ord_tbl[j] |= | |
4072 | prev->disk_ord_tbl[i]; | |
4073 | break; | |
4074 | } | |
4075 | failed = imsm_count_failed(super, dev, MAP_0); | |
4076 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
4077 | } | |
f8f603f1 DW |
4078 | |
4079 | dev->vol.migr_state = 0; | |
ea672ee1 | 4080 | set_migr_type(dev, 0); |
f8f603f1 DW |
4081 | dev->vol.curr_migr_unit = 0; |
4082 | map->map_state = map_state; | |
4083 | } | |
949c47a0 DW |
4084 | |
4085 | static int parse_raid_devices(struct intel_super *super) | |
4086 | { | |
4087 | int i; | |
4088 | struct imsm_dev *dev_new; | |
4d7b1503 | 4089 | size_t len, len_migr; |
401d313b | 4090 | size_t max_len = 0; |
4d7b1503 DW |
4091 | size_t space_needed = 0; |
4092 | struct imsm_super *mpb = super->anchor; | |
949c47a0 DW |
4093 | |
4094 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
4095 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
ba2de7ba | 4096 | struct intel_dev *dv; |
949c47a0 | 4097 | |
4d7b1503 DW |
4098 | len = sizeof_imsm_dev(dev_iter, 0); |
4099 | len_migr = sizeof_imsm_dev(dev_iter, 1); | |
4100 | if (len_migr > len) | |
4101 | space_needed += len_migr - len; | |
ca9de185 | 4102 | |
503975b9 | 4103 | dv = xmalloc(sizeof(*dv)); |
401d313b AK |
4104 | if (max_len < len_migr) |
4105 | max_len = len_migr; | |
4106 | if (max_len > len_migr) | |
4107 | space_needed += max_len - len_migr; | |
503975b9 | 4108 | dev_new = xmalloc(max_len); |
949c47a0 | 4109 | imsm_copy_dev(dev_new, dev_iter); |
ba2de7ba DW |
4110 | dv->dev = dev_new; |
4111 | dv->index = i; | |
4112 | dv->next = super->devlist; | |
4113 | super->devlist = dv; | |
949c47a0 | 4114 | } |
cdddbdbc | 4115 | |
4d7b1503 DW |
4116 | /* ensure that super->buf is large enough when all raid devices |
4117 | * are migrating | |
4118 | */ | |
4119 | if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) { | |
4120 | void *buf; | |
4121 | ||
f36a9ecd PB |
4122 | len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed, |
4123 | super->sector_size); | |
4124 | if (posix_memalign(&buf, MAX_SECTOR_SIZE, len) != 0) | |
4d7b1503 DW |
4125 | return 1; |
4126 | ||
1f45a8ad DW |
4127 | memcpy(buf, super->buf, super->len); |
4128 | memset(buf + super->len, 0, len - super->len); | |
4d7b1503 DW |
4129 | free(super->buf); |
4130 | super->buf = buf; | |
4131 | super->len = len; | |
4132 | } | |
ca9de185 | 4133 | |
bbab0940 TM |
4134 | super->extra_space += space_needed; |
4135 | ||
cdddbdbc DW |
4136 | return 0; |
4137 | } | |
4138 | ||
e2f41b2c AK |
4139 | /******************************************************************************* |
4140 | * Function: check_mpb_migr_compatibility | |
4141 | * Description: Function checks for unsupported migration features: | |
4142 | * - migration optimization area (pba_of_lba0) | |
4143 | * - descending reshape (ascending_migr) | |
4144 | * Parameters: | |
4145 | * super : imsm metadata information | |
4146 | * Returns: | |
4147 | * 0 : migration is compatible | |
4148 | * -1 : migration is not compatible | |
4149 | ******************************************************************************/ | |
4150 | int check_mpb_migr_compatibility(struct intel_super *super) | |
4151 | { | |
4152 | struct imsm_map *map0, *map1; | |
4153 | struct migr_record *migr_rec = super->migr_rec; | |
4154 | int i; | |
4155 | ||
4156 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
4157 | struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i); | |
4158 | ||
4159 | if (dev_iter && | |
4160 | dev_iter->vol.migr_state == 1 && | |
4161 | dev_iter->vol.migr_type == MIGR_GEN_MIGR) { | |
4162 | /* This device is migrating */ | |
238c0a71 AK |
4163 | map0 = get_imsm_map(dev_iter, MAP_0); |
4164 | map1 = get_imsm_map(dev_iter, MAP_1); | |
5551b113 | 4165 | if (pba_of_lba0(map0) != pba_of_lba0(map1)) |
e2f41b2c AK |
4166 | /* migration optimization area was used */ |
4167 | return -1; | |
fc54fe7a JS |
4168 | if (migr_rec->ascending_migr == 0 && |
4169 | migr_rec->dest_depth_per_unit > 0) | |
e2f41b2c AK |
4170 | /* descending reshape not supported yet */ |
4171 | return -1; | |
4172 | } | |
4173 | } | |
4174 | return 0; | |
4175 | } | |
4176 | ||
d23fe947 | 4177 | static void __free_imsm(struct intel_super *super, int free_disks); |
9ca2c81c | 4178 | |
cdddbdbc | 4179 | /* load_imsm_mpb - read matrix metadata |
f2f5c343 | 4180 | * allocates super->mpb to be freed by free_imsm |
cdddbdbc DW |
4181 | */ |
4182 | static int load_imsm_mpb(int fd, struct intel_super *super, char *devname) | |
4183 | { | |
4184 | unsigned long long dsize; | |
cdddbdbc | 4185 | unsigned long long sectors; |
f36a9ecd | 4186 | unsigned int sector_size = super->sector_size; |
cdddbdbc | 4187 | struct stat; |
6416d527 | 4188 | struct imsm_super *anchor; |
cdddbdbc DW |
4189 | __u32 check_sum; |
4190 | ||
cdddbdbc | 4191 | get_dev_size(fd, NULL, &dsize); |
f36a9ecd | 4192 | if (dsize < 2*sector_size) { |
64436f06 | 4193 | if (devname) |
e7b84f9d N |
4194 | pr_err("%s: device to small for imsm\n", |
4195 | devname); | |
64436f06 N |
4196 | return 1; |
4197 | } | |
cdddbdbc | 4198 | |
f36a9ecd | 4199 | if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0) { |
cdddbdbc | 4200 | if (devname) |
e7b84f9d N |
4201 | pr_err("Cannot seek to anchor block on %s: %s\n", |
4202 | devname, strerror(errno)); | |
cdddbdbc DW |
4203 | return 1; |
4204 | } | |
4205 | ||
f36a9ecd | 4206 | if (posix_memalign((void **)&anchor, sector_size, sector_size) != 0) { |
ad97895e | 4207 | if (devname) |
7a862a02 | 4208 | pr_err("Failed to allocate imsm anchor buffer on %s\n", devname); |
ad97895e DW |
4209 | return 1; |
4210 | } | |
466070ad | 4211 | if ((unsigned int)read(fd, anchor, sector_size) != sector_size) { |
cdddbdbc | 4212 | if (devname) |
e7b84f9d N |
4213 | pr_err("Cannot read anchor block on %s: %s\n", |
4214 | devname, strerror(errno)); | |
6416d527 | 4215 | free(anchor); |
cdddbdbc DW |
4216 | return 1; |
4217 | } | |
4218 | ||
6416d527 | 4219 | if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) { |
cdddbdbc | 4220 | if (devname) |
e7b84f9d | 4221 | pr_err("no IMSM anchor on %s\n", devname); |
6416d527 | 4222 | free(anchor); |
cdddbdbc DW |
4223 | return 2; |
4224 | } | |
4225 | ||
d23fe947 | 4226 | __free_imsm(super, 0); |
f2f5c343 LM |
4227 | /* reload capability and hba */ |
4228 | ||
4229 | /* capability and hba must be updated with new super allocation */ | |
d424212e | 4230 | find_intel_hba_capability(fd, super, devname); |
f36a9ecd PB |
4231 | super->len = ROUND_UP(anchor->mpb_size, sector_size); |
4232 | if (posix_memalign(&super->buf, MAX_SECTOR_SIZE, super->len) != 0) { | |
cdddbdbc | 4233 | if (devname) |
e7b84f9d N |
4234 | pr_err("unable to allocate %zu byte mpb buffer\n", |
4235 | super->len); | |
6416d527 | 4236 | free(anchor); |
cdddbdbc DW |
4237 | return 2; |
4238 | } | |
f36a9ecd | 4239 | memcpy(super->buf, anchor, sector_size); |
cdddbdbc | 4240 | |
f36a9ecd | 4241 | sectors = mpb_sectors(anchor, sector_size) - 1; |
6416d527 | 4242 | free(anchor); |
8e59f3d8 | 4243 | |
85337573 AO |
4244 | if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE, |
4245 | MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) { | |
1ade5cc1 | 4246 | pr_err("could not allocate migr_rec buffer\n"); |
8e59f3d8 AK |
4247 | free(super->buf); |
4248 | return 2; | |
4249 | } | |
51d83f5d | 4250 | super->clean_migration_record_by_mdmon = 0; |
8e59f3d8 | 4251 | |
949c47a0 | 4252 | if (!sectors) { |
ecf45690 DW |
4253 | check_sum = __gen_imsm_checksum(super->anchor); |
4254 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
4255 | if (devname) | |
e7b84f9d N |
4256 | pr_err("IMSM checksum %x != %x on %s\n", |
4257 | check_sum, | |
4258 | __le32_to_cpu(super->anchor->check_sum), | |
4259 | devname); | |
ecf45690 DW |
4260 | return 2; |
4261 | } | |
4262 | ||
a2b97981 | 4263 | return 0; |
949c47a0 | 4264 | } |
cdddbdbc DW |
4265 | |
4266 | /* read the extended mpb */ | |
f36a9ecd | 4267 | if (lseek64(fd, dsize - (sector_size * (2 + sectors)), SEEK_SET) < 0) { |
cdddbdbc | 4268 | if (devname) |
e7b84f9d N |
4269 | pr_err("Cannot seek to extended mpb on %s: %s\n", |
4270 | devname, strerror(errno)); | |
cdddbdbc DW |
4271 | return 1; |
4272 | } | |
4273 | ||
f36a9ecd PB |
4274 | if ((unsigned int)read(fd, super->buf + sector_size, |
4275 | super->len - sector_size) != super->len - sector_size) { | |
cdddbdbc | 4276 | if (devname) |
e7b84f9d N |
4277 | pr_err("Cannot read extended mpb on %s: %s\n", |
4278 | devname, strerror(errno)); | |
cdddbdbc DW |
4279 | return 2; |
4280 | } | |
4281 | ||
949c47a0 DW |
4282 | check_sum = __gen_imsm_checksum(super->anchor); |
4283 | if (check_sum != __le32_to_cpu(super->anchor->check_sum)) { | |
cdddbdbc | 4284 | if (devname) |
e7b84f9d N |
4285 | pr_err("IMSM checksum %x != %x on %s\n", |
4286 | check_sum, __le32_to_cpu(super->anchor->check_sum), | |
4287 | devname); | |
db575f3b | 4288 | return 3; |
cdddbdbc DW |
4289 | } |
4290 | ||
a2b97981 DW |
4291 | return 0; |
4292 | } | |
4293 | ||
8e59f3d8 AK |
4294 | static int read_imsm_migr_rec(int fd, struct intel_super *super); |
4295 | ||
97f81ee2 CA |
4296 | /* clears hi bits in metadata if MPB_ATTRIB_2TB_DISK not set */ |
4297 | static void clear_hi(struct intel_super *super) | |
4298 | { | |
4299 | struct imsm_super *mpb = super->anchor; | |
4300 | int i, n; | |
4301 | if (mpb->attributes & MPB_ATTRIB_2TB_DISK) | |
4302 | return; | |
4303 | for (i = 0; i < mpb->num_disks; ++i) { | |
4304 | struct imsm_disk *disk = &mpb->disk[i]; | |
4305 | disk->total_blocks_hi = 0; | |
4306 | } | |
4307 | for (i = 0; i < mpb->num_raid_devs; ++i) { | |
4308 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
4309 | if (!dev) | |
4310 | return; | |
4311 | for (n = 0; n < 2; ++n) { | |
4312 | struct imsm_map *map = get_imsm_map(dev, n); | |
4313 | if (!map) | |
4314 | continue; | |
4315 | map->pba_of_lba0_hi = 0; | |
4316 | map->blocks_per_member_hi = 0; | |
4317 | map->num_data_stripes_hi = 0; | |
4318 | } | |
4319 | } | |
4320 | } | |
4321 | ||
a2b97981 DW |
4322 | static int |
4323 | load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd) | |
4324 | { | |
4325 | int err; | |
4326 | ||
4327 | err = load_imsm_mpb(fd, super, devname); | |
4328 | if (err) | |
4329 | return err; | |
f36a9ecd PB |
4330 | if (super->sector_size == 4096) |
4331 | convert_from_4k(super); | |
a2b97981 DW |
4332 | err = load_imsm_disk(fd, super, devname, keep_fd); |
4333 | if (err) | |
4334 | return err; | |
4335 | err = parse_raid_devices(super); | |
8d67477f TM |
4336 | if (err) |
4337 | return err; | |
4338 | err = load_bbm_log(super); | |
97f81ee2 | 4339 | clear_hi(super); |
a2b97981 | 4340 | return err; |
cdddbdbc DW |
4341 | } |
4342 | ||
ae6aad82 DW |
4343 | static void __free_imsm_disk(struct dl *d) |
4344 | { | |
4345 | if (d->fd >= 0) | |
4346 | close(d->fd); | |
4347 | if (d->devname) | |
4348 | free(d->devname); | |
0dcecb2e DW |
4349 | if (d->e) |
4350 | free(d->e); | |
ae6aad82 DW |
4351 | free(d); |
4352 | ||
4353 | } | |
1a64be56 | 4354 | |
cdddbdbc DW |
4355 | static void free_imsm_disks(struct intel_super *super) |
4356 | { | |
47ee5a45 | 4357 | struct dl *d; |
cdddbdbc | 4358 | |
47ee5a45 DW |
4359 | while (super->disks) { |
4360 | d = super->disks; | |
cdddbdbc | 4361 | super->disks = d->next; |
ae6aad82 | 4362 | __free_imsm_disk(d); |
cdddbdbc | 4363 | } |
cb82edca AK |
4364 | while (super->disk_mgmt_list) { |
4365 | d = super->disk_mgmt_list; | |
4366 | super->disk_mgmt_list = d->next; | |
4367 | __free_imsm_disk(d); | |
4368 | } | |
47ee5a45 DW |
4369 | while (super->missing) { |
4370 | d = super->missing; | |
4371 | super->missing = d->next; | |
4372 | __free_imsm_disk(d); | |
4373 | } | |
4374 | ||
cdddbdbc DW |
4375 | } |
4376 | ||
9ca2c81c | 4377 | /* free all the pieces hanging off of a super pointer */ |
d23fe947 | 4378 | static void __free_imsm(struct intel_super *super, int free_disks) |
cdddbdbc | 4379 | { |
88654014 LM |
4380 | struct intel_hba *elem, *next; |
4381 | ||
9ca2c81c | 4382 | if (super->buf) { |
949c47a0 | 4383 | free(super->buf); |
9ca2c81c DW |
4384 | super->buf = NULL; |
4385 | } | |
f2f5c343 LM |
4386 | /* unlink capability description */ |
4387 | super->orom = NULL; | |
8e59f3d8 AK |
4388 | if (super->migr_rec_buf) { |
4389 | free(super->migr_rec_buf); | |
4390 | super->migr_rec_buf = NULL; | |
4391 | } | |
d23fe947 DW |
4392 | if (free_disks) |
4393 | free_imsm_disks(super); | |
ba2de7ba | 4394 | free_devlist(super); |
88654014 LM |
4395 | elem = super->hba; |
4396 | while (elem) { | |
4397 | if (elem->path) | |
4398 | free((void *)elem->path); | |
4399 | next = elem->next; | |
4400 | free(elem); | |
4401 | elem = next; | |
88c32bb1 | 4402 | } |
8d67477f TM |
4403 | if (super->bbm_log) |
4404 | free(super->bbm_log); | |
88654014 | 4405 | super->hba = NULL; |
cdddbdbc DW |
4406 | } |
4407 | ||
9ca2c81c DW |
4408 | static void free_imsm(struct intel_super *super) |
4409 | { | |
d23fe947 | 4410 | __free_imsm(super, 1); |
928f1424 | 4411 | free(super->bb.entries); |
9ca2c81c DW |
4412 | free(super); |
4413 | } | |
cdddbdbc DW |
4414 | |
4415 | static void free_super_imsm(struct supertype *st) | |
4416 | { | |
4417 | struct intel_super *super = st->sb; | |
4418 | ||
4419 | if (!super) | |
4420 | return; | |
4421 | ||
4422 | free_imsm(super); | |
4423 | st->sb = NULL; | |
4424 | } | |
4425 | ||
49133e57 | 4426 | static struct intel_super *alloc_super(void) |
c2c087e6 | 4427 | { |
503975b9 | 4428 | struct intel_super *super = xcalloc(1, sizeof(*super)); |
c2c087e6 | 4429 | |
503975b9 N |
4430 | super->current_vol = -1; |
4431 | super->create_offset = ~((unsigned long long) 0); | |
928f1424 TM |
4432 | |
4433 | super->bb.entries = xmalloc(BBM_LOG_MAX_ENTRIES * | |
4434 | sizeof(struct md_bb_entry)); | |
4435 | if (!super->bb.entries) { | |
4436 | free(super); | |
4437 | return NULL; | |
4438 | } | |
4439 | ||
c2c087e6 DW |
4440 | return super; |
4441 | } | |
4442 | ||
f0f5a016 LM |
4443 | /* |
4444 | * find and allocate hba and OROM/EFI based on valid fd of RAID component device | |
4445 | */ | |
d424212e | 4446 | static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname) |
f0f5a016 LM |
4447 | { |
4448 | struct sys_dev *hba_name; | |
4449 | int rv = 0; | |
4450 | ||
089f9d79 | 4451 | if (fd < 0 || check_env("IMSM_NO_PLATFORM")) { |
f2f5c343 | 4452 | super->orom = NULL; |
f0f5a016 LM |
4453 | super->hba = NULL; |
4454 | return 0; | |
4455 | } | |
4456 | hba_name = find_disk_attached_hba(fd, NULL); | |
4457 | if (!hba_name) { | |
d424212e | 4458 | if (devname) |
e7b84f9d N |
4459 | pr_err("%s is not attached to Intel(R) RAID controller.\n", |
4460 | devname); | |
f0f5a016 LM |
4461 | return 1; |
4462 | } | |
4463 | rv = attach_hba_to_super(super, hba_name); | |
4464 | if (rv == 2) { | |
d424212e N |
4465 | if (devname) { |
4466 | struct intel_hba *hba = super->hba; | |
f0f5a016 | 4467 | |
60f0f54d PB |
4468 | pr_err("%s is attached to Intel(R) %s %s (%s),\n" |
4469 | " but the container is assigned to Intel(R) %s %s (", | |
d424212e | 4470 | devname, |
614902f6 | 4471 | get_sys_dev_type(hba_name->type), |
60f0f54d | 4472 | hba_name->type == SYS_DEV_VMD ? "domain" : "RAID controller", |
f0f5a016 | 4473 | hba_name->pci_id ? : "Err!", |
60f0f54d PB |
4474 | get_sys_dev_type(super->hba->type), |
4475 | hba->type == SYS_DEV_VMD ? "domain" : "RAID controller"); | |
f0f5a016 | 4476 | |
f0f5a016 LM |
4477 | while (hba) { |
4478 | fprintf(stderr, "%s", hba->pci_id ? : "Err!"); | |
4479 | if (hba->next) | |
4480 | fprintf(stderr, ", "); | |
4481 | hba = hba->next; | |
4482 | } | |
6b781d33 | 4483 | fprintf(stderr, ").\n" |
60f0f54d PB |
4484 | " Mixing devices attached to different %s is not allowed.\n", |
4485 | hba_name->type == SYS_DEV_VMD ? "VMD domains" : "controllers"); | |
f0f5a016 | 4486 | } |
f0f5a016 LM |
4487 | return 2; |
4488 | } | |
6b781d33 | 4489 | super->orom = find_imsm_capability(hba_name); |
f2f5c343 LM |
4490 | if (!super->orom) |
4491 | return 3; | |
614902f6 | 4492 | |
f0f5a016 LM |
4493 | return 0; |
4494 | } | |
4495 | ||
47ee5a45 DW |
4496 | /* find_missing - helper routine for load_super_imsm_all that identifies |
4497 | * disks that have disappeared from the system. This routine relies on | |
4498 | * the mpb being uptodate, which it is at load time. | |
4499 | */ | |
4500 | static int find_missing(struct intel_super *super) | |
4501 | { | |
4502 | int i; | |
4503 | struct imsm_super *mpb = super->anchor; | |
4504 | struct dl *dl; | |
4505 | struct imsm_disk *disk; | |
47ee5a45 DW |
4506 | |
4507 | for (i = 0; i < mpb->num_disks; i++) { | |
4508 | disk = __get_imsm_disk(mpb, i); | |
54c2c1ea | 4509 | dl = serial_to_dl(disk->serial, super); |
47ee5a45 DW |
4510 | if (dl) |
4511 | continue; | |
47ee5a45 | 4512 | |
503975b9 | 4513 | dl = xmalloc(sizeof(*dl)); |
47ee5a45 DW |
4514 | dl->major = 0; |
4515 | dl->minor = 0; | |
4516 | dl->fd = -1; | |
503975b9 | 4517 | dl->devname = xstrdup("missing"); |
47ee5a45 DW |
4518 | dl->index = i; |
4519 | serialcpy(dl->serial, disk->serial); | |
4520 | dl->disk = *disk; | |
689c9bf3 | 4521 | dl->e = NULL; |
47ee5a45 DW |
4522 | dl->next = super->missing; |
4523 | super->missing = dl; | |
4524 | } | |
4525 | ||
4526 | return 0; | |
4527 | } | |
4528 | ||
a2b97981 DW |
4529 | static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list) |
4530 | { | |
4531 | struct intel_disk *idisk = disk_list; | |
4532 | ||
4533 | while (idisk) { | |
4534 | if (serialcmp(idisk->disk.serial, serial) == 0) | |
4535 | break; | |
4536 | idisk = idisk->next; | |
4537 | } | |
4538 | ||
4539 | return idisk; | |
4540 | } | |
4541 | ||
4542 | static int __prep_thunderdome(struct intel_super **table, int tbl_size, | |
4543 | struct intel_super *super, | |
4544 | struct intel_disk **disk_list) | |
4545 | { | |
4546 | struct imsm_disk *d = &super->disks->disk; | |
4547 | struct imsm_super *mpb = super->anchor; | |
4548 | int i, j; | |
4549 | ||
4550 | for (i = 0; i < tbl_size; i++) { | |
4551 | struct imsm_super *tbl_mpb = table[i]->anchor; | |
4552 | struct imsm_disk *tbl_d = &table[i]->disks->disk; | |
4553 | ||
4554 | if (tbl_mpb->family_num == mpb->family_num) { | |
4555 | if (tbl_mpb->check_sum == mpb->check_sum) { | |
1ade5cc1 N |
4556 | dprintf("mpb from %d:%d matches %d:%d\n", |
4557 | super->disks->major, | |
a2b97981 DW |
4558 | super->disks->minor, |
4559 | table[i]->disks->major, | |
4560 | table[i]->disks->minor); | |
4561 | break; | |
4562 | } | |
4563 | ||
4564 | if (((is_configured(d) && !is_configured(tbl_d)) || | |
4565 | is_configured(d) == is_configured(tbl_d)) && | |
4566 | tbl_mpb->generation_num < mpb->generation_num) { | |
4567 | /* current version of the mpb is a | |
4568 | * better candidate than the one in | |
4569 | * super_table, but copy over "cross | |
4570 | * generational" status | |
4571 | */ | |
4572 | struct intel_disk *idisk; | |
4573 | ||
1ade5cc1 N |
4574 | dprintf("mpb from %d:%d replaces %d:%d\n", |
4575 | super->disks->major, | |
a2b97981 DW |
4576 | super->disks->minor, |
4577 | table[i]->disks->major, | |
4578 | table[i]->disks->minor); | |
4579 | ||
4580 | idisk = disk_list_get(tbl_d->serial, *disk_list); | |
4581 | if (idisk && is_failed(&idisk->disk)) | |
4582 | tbl_d->status |= FAILED_DISK; | |
4583 | break; | |
4584 | } else { | |
4585 | struct intel_disk *idisk; | |
4586 | struct imsm_disk *disk; | |
4587 | ||
4588 | /* tbl_mpb is more up to date, but copy | |
4589 | * over cross generational status before | |
4590 | * returning | |
4591 | */ | |
4592 | disk = __serial_to_disk(d->serial, mpb, NULL); | |
4593 | if (disk && is_failed(disk)) | |
4594 | d->status |= FAILED_DISK; | |
4595 | ||
4596 | idisk = disk_list_get(d->serial, *disk_list); | |
4597 | if (idisk) { | |
4598 | idisk->owner = i; | |
4599 | if (disk && is_configured(disk)) | |
4600 | idisk->disk.status |= CONFIGURED_DISK; | |
4601 | } | |
4602 | ||
1ade5cc1 N |
4603 | dprintf("mpb from %d:%d prefer %d:%d\n", |
4604 | super->disks->major, | |
a2b97981 DW |
4605 | super->disks->minor, |
4606 | table[i]->disks->major, | |
4607 | table[i]->disks->minor); | |
4608 | ||
4609 | return tbl_size; | |
4610 | } | |
4611 | } | |
4612 | } | |
4613 | ||
4614 | if (i >= tbl_size) | |
4615 | table[tbl_size++] = super; | |
4616 | else | |
4617 | table[i] = super; | |
4618 | ||
4619 | /* update/extend the merged list of imsm_disk records */ | |
4620 | for (j = 0; j < mpb->num_disks; j++) { | |
4621 | struct imsm_disk *disk = __get_imsm_disk(mpb, j); | |
4622 | struct intel_disk *idisk; | |
4623 | ||
4624 | idisk = disk_list_get(disk->serial, *disk_list); | |
4625 | if (idisk) { | |
4626 | idisk->disk.status |= disk->status; | |
4627 | if (is_configured(&idisk->disk) || | |
4628 | is_failed(&idisk->disk)) | |
4629 | idisk->disk.status &= ~(SPARE_DISK); | |
4630 | } else { | |
503975b9 | 4631 | idisk = xcalloc(1, sizeof(*idisk)); |
a2b97981 DW |
4632 | idisk->owner = IMSM_UNKNOWN_OWNER; |
4633 | idisk->disk = *disk; | |
4634 | idisk->next = *disk_list; | |
4635 | *disk_list = idisk; | |
4636 | } | |
4637 | ||
4638 | if (serialcmp(idisk->disk.serial, d->serial) == 0) | |
4639 | idisk->owner = i; | |
4640 | } | |
4641 | ||
4642 | return tbl_size; | |
4643 | } | |
4644 | ||
4645 | static struct intel_super * | |
4646 | validate_members(struct intel_super *super, struct intel_disk *disk_list, | |
4647 | const int owner) | |
4648 | { | |
4649 | struct imsm_super *mpb = super->anchor; | |
4650 | int ok_count = 0; | |
4651 | int i; | |
4652 | ||
4653 | for (i = 0; i < mpb->num_disks; i++) { | |
4654 | struct imsm_disk *disk = __get_imsm_disk(mpb, i); | |
4655 | struct intel_disk *idisk; | |
4656 | ||
4657 | idisk = disk_list_get(disk->serial, disk_list); | |
4658 | if (idisk) { | |
4659 | if (idisk->owner == owner || | |
4660 | idisk->owner == IMSM_UNKNOWN_OWNER) | |
4661 | ok_count++; | |
4662 | else | |
1ade5cc1 N |
4663 | dprintf("'%.16s' owner %d != %d\n", |
4664 | disk->serial, idisk->owner, | |
a2b97981 DW |
4665 | owner); |
4666 | } else { | |
1ade5cc1 N |
4667 | dprintf("unknown disk %x [%d]: %.16s\n", |
4668 | __le32_to_cpu(mpb->family_num), i, | |
a2b97981 DW |
4669 | disk->serial); |
4670 | break; | |
4671 | } | |
4672 | } | |
4673 | ||
4674 | if (ok_count == mpb->num_disks) | |
4675 | return super; | |
4676 | return NULL; | |
4677 | } | |
4678 | ||
4679 | static void show_conflicts(__u32 family_num, struct intel_super *super_list) | |
4680 | { | |
4681 | struct intel_super *s; | |
4682 | ||
4683 | for (s = super_list; s; s = s->next) { | |
4684 | if (family_num != s->anchor->family_num) | |
4685 | continue; | |
e12b3daa | 4686 | pr_err("Conflict, offlining family %#x on '%s'\n", |
a2b97981 DW |
4687 | __le32_to_cpu(family_num), s->disks->devname); |
4688 | } | |
4689 | } | |
4690 | ||
4691 | static struct intel_super * | |
4692 | imsm_thunderdome(struct intel_super **super_list, int len) | |
4693 | { | |
4694 | struct intel_super *super_table[len]; | |
4695 | struct intel_disk *disk_list = NULL; | |
4696 | struct intel_super *champion, *spare; | |
4697 | struct intel_super *s, **del; | |
4698 | int tbl_size = 0; | |
4699 | int conflict; | |
4700 | int i; | |
4701 | ||
4702 | memset(super_table, 0, sizeof(super_table)); | |
4703 | for (s = *super_list; s; s = s->next) | |
4704 | tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list); | |
4705 | ||
4706 | for (i = 0; i < tbl_size; i++) { | |
4707 | struct imsm_disk *d; | |
4708 | struct intel_disk *idisk; | |
4709 | struct imsm_super *mpb = super_table[i]->anchor; | |
4710 | ||
4711 | s = super_table[i]; | |
4712 | d = &s->disks->disk; | |
4713 | ||
4714 | /* 'd' must appear in merged disk list for its | |
4715 | * configuration to be valid | |
4716 | */ | |
4717 | idisk = disk_list_get(d->serial, disk_list); | |
4718 | if (idisk && idisk->owner == i) | |
4719 | s = validate_members(s, disk_list, i); | |
4720 | else | |
4721 | s = NULL; | |
4722 | ||
4723 | if (!s) | |
1ade5cc1 N |
4724 | dprintf("marking family: %#x from %d:%d offline\n", |
4725 | mpb->family_num, | |
a2b97981 DW |
4726 | super_table[i]->disks->major, |
4727 | super_table[i]->disks->minor); | |
4728 | super_table[i] = s; | |
4729 | } | |
4730 | ||
4731 | /* This is where the mdadm implementation differs from the Windows | |
4732 | * driver which has no strict concept of a container. We can only | |
4733 | * assemble one family from a container, so when returning a prodigal | |
4734 | * array member to this system the code will not be able to disambiguate | |
4735 | * the container contents that should be assembled ("foreign" versus | |
4736 | * "local"). It requires user intervention to set the orig_family_num | |
4737 | * to a new value to establish a new container. The Windows driver in | |
4738 | * this situation fixes up the volume name in place and manages the | |
4739 | * foreign array as an independent entity. | |
4740 | */ | |
4741 | s = NULL; | |
4742 | spare = NULL; | |
4743 | conflict = 0; | |
4744 | for (i = 0; i < tbl_size; i++) { | |
4745 | struct intel_super *tbl_ent = super_table[i]; | |
4746 | int is_spare = 0; | |
4747 | ||
4748 | if (!tbl_ent) | |
4749 | continue; | |
4750 | ||
4751 | if (tbl_ent->anchor->num_raid_devs == 0) { | |
4752 | spare = tbl_ent; | |
4753 | is_spare = 1; | |
4754 | } | |
4755 | ||
4756 | if (s && !is_spare) { | |
4757 | show_conflicts(tbl_ent->anchor->family_num, *super_list); | |
4758 | conflict++; | |
4759 | } else if (!s && !is_spare) | |
4760 | s = tbl_ent; | |
4761 | } | |
4762 | ||
4763 | if (!s) | |
4764 | s = spare; | |
4765 | if (!s) { | |
4766 | champion = NULL; | |
4767 | goto out; | |
4768 | } | |
4769 | champion = s; | |
4770 | ||
4771 | if (conflict) | |
7a862a02 | 4772 | pr_err("Chose family %#x on '%s', assemble conflicts to new container with '--update=uuid'\n", |
a2b97981 DW |
4773 | __le32_to_cpu(s->anchor->family_num), s->disks->devname); |
4774 | ||
4775 | /* collect all dl's onto 'champion', and update them to | |
4776 | * champion's version of the status | |
4777 | */ | |
4778 | for (s = *super_list; s; s = s->next) { | |
4779 | struct imsm_super *mpb = champion->anchor; | |
4780 | struct dl *dl = s->disks; | |
4781 | ||
4782 | if (s == champion) | |
4783 | continue; | |
4784 | ||
5d7b407a CA |
4785 | mpb->attributes |= s->anchor->attributes & MPB_ATTRIB_2TB_DISK; |
4786 | ||
a2b97981 DW |
4787 | for (i = 0; i < mpb->num_disks; i++) { |
4788 | struct imsm_disk *disk; | |
4789 | ||
4790 | disk = __serial_to_disk(dl->serial, mpb, &dl->index); | |
4791 | if (disk) { | |
4792 | dl->disk = *disk; | |
4793 | /* only set index on disks that are a member of | |
4794 | * a populated contianer, i.e. one with | |
4795 | * raid_devs | |
4796 | */ | |
4797 | if (is_failed(&dl->disk)) | |
4798 | dl->index = -2; | |
4799 | else if (is_spare(&dl->disk)) | |
4800 | dl->index = -1; | |
4801 | break; | |
4802 | } | |
4803 | } | |
4804 | ||
4805 | if (i >= mpb->num_disks) { | |
4806 | struct intel_disk *idisk; | |
4807 | ||
4808 | idisk = disk_list_get(dl->serial, disk_list); | |
ecf408e9 | 4809 | if (idisk && is_spare(&idisk->disk) && |
a2b97981 DW |
4810 | !is_failed(&idisk->disk) && !is_configured(&idisk->disk)) |
4811 | dl->index = -1; | |
4812 | else { | |
4813 | dl->index = -2; | |
4814 | continue; | |
4815 | } | |
4816 | } | |
4817 | ||
4818 | dl->next = champion->disks; | |
4819 | champion->disks = dl; | |
4820 | s->disks = NULL; | |
4821 | } | |
4822 | ||
4823 | /* delete 'champion' from super_list */ | |
4824 | for (del = super_list; *del; ) { | |
4825 | if (*del == champion) { | |
4826 | *del = (*del)->next; | |
4827 | break; | |
4828 | } else | |
4829 | del = &(*del)->next; | |
4830 | } | |
4831 | champion->next = NULL; | |
4832 | ||
4833 | out: | |
4834 | while (disk_list) { | |
4835 | struct intel_disk *idisk = disk_list; | |
4836 | ||
4837 | disk_list = disk_list->next; | |
4838 | free(idisk); | |
4839 | } | |
4840 | ||
4841 | return champion; | |
4842 | } | |
4843 | ||
9587c373 LM |
4844 | static int |
4845 | get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd); | |
4dd2df09 | 4846 | static int get_super_block(struct intel_super **super_list, char *devnm, char *devname, |
9587c373 | 4847 | int major, int minor, int keep_fd); |
ec50f7b6 LM |
4848 | static int |
4849 | get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list, | |
4850 | int *max, int keep_fd); | |
4851 | ||
cdddbdbc | 4852 | static int load_super_imsm_all(struct supertype *st, int fd, void **sbp, |
ec50f7b6 LM |
4853 | char *devname, struct md_list *devlist, |
4854 | int keep_fd) | |
cdddbdbc | 4855 | { |
a2b97981 DW |
4856 | struct intel_super *super_list = NULL; |
4857 | struct intel_super *super = NULL; | |
a2b97981 | 4858 | int err = 0; |
9587c373 | 4859 | int i = 0; |
dab4a513 | 4860 | |
9587c373 LM |
4861 | if (fd >= 0) |
4862 | /* 'fd' is an opened container */ | |
4863 | err = get_sra_super_block(fd, &super_list, devname, &i, keep_fd); | |
4864 | else | |
ec50f7b6 LM |
4865 | /* get super block from devlist devices */ |
4866 | err = get_devlist_super_block(devlist, &super_list, &i, keep_fd); | |
9587c373 | 4867 | if (err) |
1602d52c | 4868 | goto error; |
a2b97981 DW |
4869 | /* all mpbs enter, maybe one leaves */ |
4870 | super = imsm_thunderdome(&super_list, i); | |
4871 | if (!super) { | |
4872 | err = 1; | |
4873 | goto error; | |
cdddbdbc DW |
4874 | } |
4875 | ||
47ee5a45 DW |
4876 | if (find_missing(super) != 0) { |
4877 | free_imsm(super); | |
a2b97981 DW |
4878 | err = 2; |
4879 | goto error; | |
47ee5a45 | 4880 | } |
8e59f3d8 AK |
4881 | |
4882 | /* load migration record */ | |
4883 | err = load_imsm_migr_rec(super, NULL); | |
4c965cc9 AK |
4884 | if (err == -1) { |
4885 | /* migration is in progress, | |
4886 | * but migr_rec cannot be loaded, | |
4887 | */ | |
8e59f3d8 AK |
4888 | err = 4; |
4889 | goto error; | |
4890 | } | |
e2f41b2c AK |
4891 | |
4892 | /* Check migration compatibility */ | |
089f9d79 | 4893 | if (err == 0 && check_mpb_migr_compatibility(super) != 0) { |
e7b84f9d | 4894 | pr_err("Unsupported migration detected"); |
e2f41b2c AK |
4895 | if (devname) |
4896 | fprintf(stderr, " on %s\n", devname); | |
4897 | else | |
4898 | fprintf(stderr, " (IMSM).\n"); | |
4899 | ||
4900 | err = 5; | |
4901 | goto error; | |
4902 | } | |
4903 | ||
a2b97981 DW |
4904 | err = 0; |
4905 | ||
4906 | error: | |
4907 | while (super_list) { | |
4908 | struct intel_super *s = super_list; | |
4909 | ||
4910 | super_list = super_list->next; | |
4911 | free_imsm(s); | |
4912 | } | |
9587c373 | 4913 | |
a2b97981 DW |
4914 | if (err) |
4915 | return err; | |
f7e7067b | 4916 | |
cdddbdbc | 4917 | *sbp = super; |
9587c373 | 4918 | if (fd >= 0) |
4dd2df09 | 4919 | strcpy(st->container_devnm, fd2devnm(fd)); |
9587c373 | 4920 | else |
4dd2df09 | 4921 | st->container_devnm[0] = 0; |
a2b97981 | 4922 | if (err == 0 && st->ss == NULL) { |
bf5a934a | 4923 | st->ss = &super_imsm; |
cdddbdbc DW |
4924 | st->minor_version = 0; |
4925 | st->max_devs = IMSM_MAX_DEVICES; | |
4926 | } | |
cdddbdbc DW |
4927 | return 0; |
4928 | } | |
2b959fbf | 4929 | |
ec50f7b6 LM |
4930 | static int |
4931 | get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list, | |
4932 | int *max, int keep_fd) | |
4933 | { | |
4934 | struct md_list *tmpdev; | |
4935 | int err = 0; | |
4936 | int i = 0; | |
9587c373 | 4937 | |
ec50f7b6 LM |
4938 | for (i = 0, tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { |
4939 | if (tmpdev->used != 1) | |
4940 | continue; | |
4941 | if (tmpdev->container == 1) { | |
ca9de185 | 4942 | int lmax = 0; |
ec50f7b6 LM |
4943 | int fd = dev_open(tmpdev->devname, O_RDONLY|O_EXCL); |
4944 | if (fd < 0) { | |
e7b84f9d | 4945 | pr_err("cannot open device %s: %s\n", |
ec50f7b6 LM |
4946 | tmpdev->devname, strerror(errno)); |
4947 | err = 8; | |
4948 | goto error; | |
4949 | } | |
4950 | err = get_sra_super_block(fd, super_list, | |
4951 | tmpdev->devname, &lmax, | |
4952 | keep_fd); | |
4953 | i += lmax; | |
4954 | close(fd); | |
4955 | if (err) { | |
4956 | err = 7; | |
4957 | goto error; | |
4958 | } | |
4959 | } else { | |
4960 | int major = major(tmpdev->st_rdev); | |
4961 | int minor = minor(tmpdev->st_rdev); | |
4962 | err = get_super_block(super_list, | |
4dd2df09 | 4963 | NULL, |
ec50f7b6 LM |
4964 | tmpdev->devname, |
4965 | major, minor, | |
4966 | keep_fd); | |
4967 | i++; | |
4968 | if (err) { | |
4969 | err = 6; | |
4970 | goto error; | |
4971 | } | |
4972 | } | |
4973 | } | |
4974 | error: | |
4975 | *max = i; | |
4976 | return err; | |
4977 | } | |
9587c373 | 4978 | |
4dd2df09 | 4979 | static int get_super_block(struct intel_super **super_list, char *devnm, char *devname, |
9587c373 LM |
4980 | int major, int minor, int keep_fd) |
4981 | { | |
594dc1b8 | 4982 | struct intel_super *s; |
9587c373 LM |
4983 | char nm[32]; |
4984 | int dfd = -1; | |
9587c373 LM |
4985 | int err = 0; |
4986 | int retry; | |
4987 | ||
4988 | s = alloc_super(); | |
4989 | if (!s) { | |
4990 | err = 1; | |
4991 | goto error; | |
4992 | } | |
4993 | ||
4994 | sprintf(nm, "%d:%d", major, minor); | |
4995 | dfd = dev_open(nm, O_RDWR); | |
4996 | if (dfd < 0) { | |
4997 | err = 2; | |
4998 | goto error; | |
4999 | } | |
5000 | ||
fa7bb6f8 | 5001 | get_dev_sector_size(dfd, NULL, &s->sector_size); |
cb8f6859 | 5002 | find_intel_hba_capability(dfd, s, devname); |
9587c373 LM |
5003 | err = load_and_parse_mpb(dfd, s, NULL, keep_fd); |
5004 | ||
5005 | /* retry the load if we might have raced against mdmon */ | |
4dd2df09 | 5006 | if (err == 3 && devnm && mdmon_running(devnm)) |
9587c373 LM |
5007 | for (retry = 0; retry < 3; retry++) { |
5008 | usleep(3000); | |
5009 | err = load_and_parse_mpb(dfd, s, NULL, keep_fd); | |
5010 | if (err != 3) | |
5011 | break; | |
5012 | } | |
5013 | error: | |
5014 | if (!err) { | |
5015 | s->next = *super_list; | |
5016 | *super_list = s; | |
5017 | } else { | |
5018 | if (s) | |
8d67477f | 5019 | free_imsm(s); |
36614e95 | 5020 | if (dfd >= 0) |
9587c373 LM |
5021 | close(dfd); |
5022 | } | |
089f9d79 | 5023 | if (dfd >= 0 && !keep_fd) |
9587c373 LM |
5024 | close(dfd); |
5025 | return err; | |
5026 | ||
5027 | } | |
5028 | ||
5029 | static int | |
5030 | get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd) | |
5031 | { | |
5032 | struct mdinfo *sra; | |
4dd2df09 | 5033 | char *devnm; |
9587c373 LM |
5034 | struct mdinfo *sd; |
5035 | int err = 0; | |
5036 | int i = 0; | |
4dd2df09 | 5037 | sra = sysfs_read(fd, NULL, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE); |
9587c373 LM |
5038 | if (!sra) |
5039 | return 1; | |
5040 | ||
5041 | if (sra->array.major_version != -1 || | |
5042 | sra->array.minor_version != -2 || | |
5043 | strcmp(sra->text_version, "imsm") != 0) { | |
5044 | err = 1; | |
5045 | goto error; | |
5046 | } | |
5047 | /* load all mpbs */ | |
4dd2df09 | 5048 | devnm = fd2devnm(fd); |
9587c373 | 5049 | for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) { |
4dd2df09 | 5050 | if (get_super_block(super_list, devnm, devname, |
9587c373 LM |
5051 | sd->disk.major, sd->disk.minor, keep_fd) != 0) { |
5052 | err = 7; | |
5053 | goto error; | |
5054 | } | |
5055 | } | |
5056 | error: | |
5057 | sysfs_free(sra); | |
5058 | *max = i; | |
5059 | return err; | |
5060 | } | |
5061 | ||
2b959fbf N |
5062 | static int load_container_imsm(struct supertype *st, int fd, char *devname) |
5063 | { | |
ec50f7b6 | 5064 | return load_super_imsm_all(st, fd, &st->sb, devname, NULL, 1); |
2b959fbf | 5065 | } |
cdddbdbc DW |
5066 | |
5067 | static int load_super_imsm(struct supertype *st, int fd, char *devname) | |
5068 | { | |
5069 | struct intel_super *super; | |
5070 | int rv; | |
8a3544f8 | 5071 | int retry; |
cdddbdbc | 5072 | |
357ac106 | 5073 | if (test_partition(fd)) |
691c6ee1 N |
5074 | /* IMSM not allowed on partitions */ |
5075 | return 1; | |
5076 | ||
37424f13 DW |
5077 | free_super_imsm(st); |
5078 | ||
49133e57 | 5079 | super = alloc_super(); |
fa7bb6f8 | 5080 | get_dev_sector_size(fd, NULL, &super->sector_size); |
8d67477f TM |
5081 | if (!super) |
5082 | return 1; | |
ea2bc72b LM |
5083 | /* Load hba and capabilities if they exist. |
5084 | * But do not preclude loading metadata in case capabilities or hba are | |
5085 | * non-compliant and ignore_hw_compat is set. | |
5086 | */ | |
d424212e | 5087 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 5088 | /* no orom/efi or non-intel hba of the disk */ |
089f9d79 | 5089 | if (rv != 0 && st->ignore_hw_compat == 0) { |
f2f5c343 | 5090 | if (devname) |
e7b84f9d | 5091 | pr_err("No OROM/EFI properties for %s\n", devname); |
f2f5c343 LM |
5092 | free_imsm(super); |
5093 | return 2; | |
5094 | } | |
a2b97981 | 5095 | rv = load_and_parse_mpb(fd, super, devname, 0); |
cdddbdbc | 5096 | |
8a3544f8 AP |
5097 | /* retry the load if we might have raced against mdmon */ |
5098 | if (rv == 3) { | |
f96b1302 AP |
5099 | struct mdstat_ent *mdstat = NULL; |
5100 | char *name = fd2kname(fd); | |
5101 | ||
5102 | if (name) | |
5103 | mdstat = mdstat_by_component(name); | |
8a3544f8 AP |
5104 | |
5105 | if (mdstat && mdmon_running(mdstat->devnm) && getpid() != mdmon_pid(mdstat->devnm)) { | |
5106 | for (retry = 0; retry < 3; retry++) { | |
5107 | usleep(3000); | |
5108 | rv = load_and_parse_mpb(fd, super, devname, 0); | |
5109 | if (rv != 3) | |
5110 | break; | |
5111 | } | |
5112 | } | |
5113 | ||
5114 | free_mdstat(mdstat); | |
5115 | } | |
5116 | ||
cdddbdbc DW |
5117 | if (rv) { |
5118 | if (devname) | |
7a862a02 | 5119 | pr_err("Failed to load all information sections on %s\n", devname); |
cdddbdbc DW |
5120 | free_imsm(super); |
5121 | return rv; | |
5122 | } | |
5123 | ||
5124 | st->sb = super; | |
5125 | if (st->ss == NULL) { | |
5126 | st->ss = &super_imsm; | |
5127 | st->minor_version = 0; | |
5128 | st->max_devs = IMSM_MAX_DEVICES; | |
5129 | } | |
8e59f3d8 AK |
5130 | |
5131 | /* load migration record */ | |
2e062e82 AK |
5132 | if (load_imsm_migr_rec(super, NULL) == 0) { |
5133 | /* Check for unsupported migration features */ | |
5134 | if (check_mpb_migr_compatibility(super) != 0) { | |
e7b84f9d | 5135 | pr_err("Unsupported migration detected"); |
2e062e82 AK |
5136 | if (devname) |
5137 | fprintf(stderr, " on %s\n", devname); | |
5138 | else | |
5139 | fprintf(stderr, " (IMSM).\n"); | |
5140 | return 3; | |
5141 | } | |
e2f41b2c AK |
5142 | } |
5143 | ||
cdddbdbc DW |
5144 | return 0; |
5145 | } | |
5146 | ||
ef6ffade DW |
5147 | static __u16 info_to_blocks_per_strip(mdu_array_info_t *info) |
5148 | { | |
5149 | if (info->level == 1) | |
5150 | return 128; | |
5151 | return info->chunk_size >> 9; | |
5152 | } | |
5153 | ||
5551b113 CA |
5154 | static unsigned long long info_to_blocks_per_member(mdu_array_info_t *info, |
5155 | unsigned long long size) | |
fcfd9599 | 5156 | { |
4025c288 | 5157 | if (info->level == 1) |
5551b113 | 5158 | return size * 2; |
4025c288 | 5159 | else |
5551b113 | 5160 | return (size * 2) & ~(info_to_blocks_per_strip(info) - 1); |
fcfd9599 DW |
5161 | } |
5162 | ||
4d1313e9 DW |
5163 | static void imsm_update_version_info(struct intel_super *super) |
5164 | { | |
5165 | /* update the version and attributes */ | |
5166 | struct imsm_super *mpb = super->anchor; | |
5167 | char *version; | |
5168 | struct imsm_dev *dev; | |
5169 | struct imsm_map *map; | |
5170 | int i; | |
5171 | ||
5172 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
5173 | dev = get_imsm_dev(super, i); | |
238c0a71 | 5174 | map = get_imsm_map(dev, MAP_0); |
4d1313e9 DW |
5175 | if (__le32_to_cpu(dev->size_high) > 0) |
5176 | mpb->attributes |= MPB_ATTRIB_2TB; | |
5177 | ||
5178 | /* FIXME detect when an array spans a port multiplier */ | |
5179 | #if 0 | |
5180 | mpb->attributes |= MPB_ATTRIB_PM; | |
5181 | #endif | |
5182 | ||
5183 | if (mpb->num_raid_devs > 1 || | |
5184 | mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) { | |
5185 | version = MPB_VERSION_ATTRIBS; | |
5186 | switch (get_imsm_raid_level(map)) { | |
5187 | case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break; | |
5188 | case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break; | |
5189 | case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break; | |
5190 | case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break; | |
5191 | } | |
5192 | } else { | |
5193 | if (map->num_members >= 5) | |
5194 | version = MPB_VERSION_5OR6_DISK_ARRAY; | |
5195 | else if (dev->status == DEV_CLONE_N_GO) | |
5196 | version = MPB_VERSION_CNG; | |
5197 | else if (get_imsm_raid_level(map) == 5) | |
5198 | version = MPB_VERSION_RAID5; | |
5199 | else if (map->num_members >= 3) | |
5200 | version = MPB_VERSION_3OR4_DISK_ARRAY; | |
5201 | else if (get_imsm_raid_level(map) == 1) | |
5202 | version = MPB_VERSION_RAID1; | |
5203 | else | |
5204 | version = MPB_VERSION_RAID0; | |
5205 | } | |
5206 | strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version); | |
5207 | } | |
5208 | } | |
5209 | ||
aa534678 DW |
5210 | static int check_name(struct intel_super *super, char *name, int quiet) |
5211 | { | |
5212 | struct imsm_super *mpb = super->anchor; | |
5213 | char *reason = NULL; | |
5214 | int i; | |
5215 | ||
5216 | if (strlen(name) > MAX_RAID_SERIAL_LEN) | |
5217 | reason = "must be 16 characters or less"; | |
5218 | ||
5219 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
5220 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
5221 | ||
5222 | if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) { | |
5223 | reason = "already exists"; | |
5224 | break; | |
5225 | } | |
5226 | } | |
5227 | ||
5228 | if (reason && !quiet) | |
e7b84f9d | 5229 | pr_err("imsm volume name %s\n", reason); |
aa534678 DW |
5230 | |
5231 | return !reason; | |
5232 | } | |
5233 | ||
8b353278 | 5234 | static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info, |
5308f117 | 5235 | struct shape *s, char *name, |
83cd1e97 N |
5236 | char *homehost, int *uuid, |
5237 | long long data_offset) | |
cdddbdbc | 5238 | { |
c2c087e6 DW |
5239 | /* We are creating a volume inside a pre-existing container. |
5240 | * so st->sb is already set. | |
5241 | */ | |
5242 | struct intel_super *super = st->sb; | |
f36a9ecd | 5243 | unsigned int sector_size = super->sector_size; |
949c47a0 | 5244 | struct imsm_super *mpb = super->anchor; |
ba2de7ba | 5245 | struct intel_dev *dv; |
c2c087e6 DW |
5246 | struct imsm_dev *dev; |
5247 | struct imsm_vol *vol; | |
5248 | struct imsm_map *map; | |
5249 | int idx = mpb->num_raid_devs; | |
5250 | int i; | |
5251 | unsigned long long array_blocks; | |
2c092cad | 5252 | size_t size_old, size_new; |
5551b113 | 5253 | unsigned long long num_data_stripes; |
b53bfba6 TM |
5254 | unsigned int data_disks; |
5255 | unsigned long long size_per_member; | |
cdddbdbc | 5256 | |
88c32bb1 | 5257 | if (super->orom && mpb->num_raid_devs >= super->orom->vpa) { |
7a862a02 | 5258 | pr_err("This imsm-container already has the maximum of %d volumes\n", super->orom->vpa); |
c2c087e6 DW |
5259 | return 0; |
5260 | } | |
5261 | ||
2c092cad DW |
5262 | /* ensure the mpb is large enough for the new data */ |
5263 | size_old = __le32_to_cpu(mpb->mpb_size); | |
5264 | size_new = disks_to_mpb_size(info->nr_disks); | |
5265 | if (size_new > size_old) { | |
5266 | void *mpb_new; | |
f36a9ecd | 5267 | size_t size_round = ROUND_UP(size_new, sector_size); |
2c092cad | 5268 | |
f36a9ecd | 5269 | if (posix_memalign(&mpb_new, sector_size, size_round) != 0) { |
e7b84f9d | 5270 | pr_err("could not allocate new mpb\n"); |
2c092cad DW |
5271 | return 0; |
5272 | } | |
85337573 AO |
5273 | if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE, |
5274 | MIGR_REC_BUF_SECTORS* | |
5275 | MAX_SECTOR_SIZE) != 0) { | |
1ade5cc1 | 5276 | pr_err("could not allocate migr_rec buffer\n"); |
8e59f3d8 AK |
5277 | free(super->buf); |
5278 | free(super); | |
ea944c8f | 5279 | free(mpb_new); |
8e59f3d8 AK |
5280 | return 0; |
5281 | } | |
2c092cad DW |
5282 | memcpy(mpb_new, mpb, size_old); |
5283 | free(mpb); | |
5284 | mpb = mpb_new; | |
949c47a0 | 5285 | super->anchor = mpb_new; |
2c092cad DW |
5286 | mpb->mpb_size = __cpu_to_le32(size_new); |
5287 | memset(mpb_new + size_old, 0, size_round - size_old); | |
bbab0940 | 5288 | super->len = size_round; |
2c092cad | 5289 | } |
bf5a934a | 5290 | super->current_vol = idx; |
3960e579 DW |
5291 | |
5292 | /* handle 'failed_disks' by either: | |
5293 | * a) create dummy disk entries in the table if this the first | |
5294 | * volume in the array. We add them here as this is the only | |
5295 | * opportunity to add them. add_to_super_imsm_volume() | |
5296 | * handles the non-failed disks and continues incrementing | |
5297 | * mpb->num_disks. | |
5298 | * b) validate that 'failed_disks' matches the current number | |
5299 | * of missing disks if the container is populated | |
d23fe947 | 5300 | */ |
3960e579 | 5301 | if (super->current_vol == 0) { |
d23fe947 | 5302 | mpb->num_disks = 0; |
3960e579 DW |
5303 | for (i = 0; i < info->failed_disks; i++) { |
5304 | struct imsm_disk *disk; | |
5305 | ||
5306 | mpb->num_disks++; | |
5307 | disk = __get_imsm_disk(mpb, i); | |
5308 | disk->status = CONFIGURED_DISK | FAILED_DISK; | |
5309 | disk->scsi_id = __cpu_to_le32(~(__u32)0); | |
5310 | snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN, | |
5b975129 | 5311 | "missing:%d", (__u8)i); |
3960e579 DW |
5312 | } |
5313 | find_missing(super); | |
5314 | } else { | |
5315 | int missing = 0; | |
5316 | struct dl *d; | |
5317 | ||
5318 | for (d = super->missing; d; d = d->next) | |
5319 | missing++; | |
5320 | if (info->failed_disks > missing) { | |
e7b84f9d | 5321 | pr_err("unable to add 'missing' disk to container\n"); |
3960e579 DW |
5322 | return 0; |
5323 | } | |
5324 | } | |
5a038140 | 5325 | |
aa534678 DW |
5326 | if (!check_name(super, name, 0)) |
5327 | return 0; | |
503975b9 N |
5328 | dv = xmalloc(sizeof(*dv)); |
5329 | dev = xcalloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1)); | |
c2c087e6 | 5330 | strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN); |
e03640bd | 5331 | array_blocks = calc_array_size(info->level, info->raid_disks, |
03bcbc65 | 5332 | info->layout, info->chunk_size, |
b53bfba6 TM |
5333 | s->size * BLOCKS_PER_KB); |
5334 | data_disks = get_data_disks(info->level, info->layout, | |
5335 | info->raid_disks); | |
5336 | array_blocks = round_size_to_mb(array_blocks, data_disks); | |
5337 | size_per_member = array_blocks / data_disks; | |
979d38be | 5338 | |
c2c087e6 DW |
5339 | dev->size_low = __cpu_to_le32((__u32) array_blocks); |
5340 | dev->size_high = __cpu_to_le32((__u32) (array_blocks >> 32)); | |
1a2487c2 | 5341 | dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING); |
c2c087e6 DW |
5342 | vol = &dev->vol; |
5343 | vol->migr_state = 0; | |
1484e727 | 5344 | set_migr_type(dev, MIGR_INIT); |
3960e579 | 5345 | vol->dirty = !info->state; |
f8f603f1 | 5346 | vol->curr_migr_unit = 0; |
238c0a71 | 5347 | map = get_imsm_map(dev, MAP_0); |
5551b113 | 5348 | set_pba_of_lba0(map, super->create_offset); |
b53bfba6 TM |
5349 | set_blocks_per_member(map, info_to_blocks_per_member(info, |
5350 | size_per_member / | |
5351 | BLOCKS_PER_KB)); | |
ef6ffade | 5352 | map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info)); |
0556e1a2 | 5353 | map->failed_disk_num = ~0; |
bf4442ab | 5354 | if (info->level > 0) |
fffaf1ff N |
5355 | map->map_state = (info->state ? IMSM_T_STATE_NORMAL |
5356 | : IMSM_T_STATE_UNINITIALIZED); | |
bf4442ab AK |
5357 | else |
5358 | map->map_state = info->failed_disks ? IMSM_T_STATE_FAILED : | |
5359 | IMSM_T_STATE_NORMAL; | |
252d23c0 | 5360 | map->ddf = 1; |
ef6ffade DW |
5361 | |
5362 | if (info->level == 1 && info->raid_disks > 2) { | |
38950822 AW |
5363 | free(dev); |
5364 | free(dv); | |
7a862a02 | 5365 | pr_err("imsm does not support more than 2 disksin a raid1 volume\n"); |
ef6ffade DW |
5366 | return 0; |
5367 | } | |
81062a36 DW |
5368 | |
5369 | map->raid_level = info->level; | |
4d1313e9 | 5370 | if (info->level == 10) { |
c2c087e6 | 5371 | map->raid_level = 1; |
4d1313e9 | 5372 | map->num_domains = info->raid_disks / 2; |
81062a36 DW |
5373 | } else if (info->level == 1) |
5374 | map->num_domains = info->raid_disks; | |
5375 | else | |
ff596308 | 5376 | map->num_domains = 1; |
81062a36 | 5377 | |
5551b113 | 5378 | /* info->size is only int so use the 'size' parameter instead */ |
b53bfba6 | 5379 | num_data_stripes = size_per_member / info_to_blocks_per_strip(info); |
5551b113 CA |
5380 | num_data_stripes /= map->num_domains; |
5381 | set_num_data_stripes(map, num_data_stripes); | |
ef6ffade | 5382 | |
c2c087e6 DW |
5383 | map->num_members = info->raid_disks; |
5384 | for (i = 0; i < map->num_members; i++) { | |
5385 | /* initialized in add_to_super */ | |
4eb26970 | 5386 | set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD); |
c2c087e6 | 5387 | } |
949c47a0 | 5388 | mpb->num_raid_devs++; |
2a24dc1b PB |
5389 | mpb->num_raid_devs_created++; |
5390 | dev->my_vol_raid_dev_num = mpb->num_raid_devs_created; | |
ba2de7ba | 5391 | |
b7580566 | 5392 | if (s->consistency_policy <= CONSISTENCY_POLICY_RESYNC) { |
2432ce9b AP |
5393 | dev->rwh_policy = RWH_OFF; |
5394 | } else if (s->consistency_policy == CONSISTENCY_POLICY_PPL) { | |
5395 | dev->rwh_policy = RWH_DISTRIBUTED; | |
5396 | } else { | |
5397 | free(dev); | |
5398 | free(dv); | |
5399 | pr_err("imsm does not support consistency policy %s\n", | |
5400 | map_num(consistency_policies, s->consistency_policy)); | |
5401 | return 0; | |
5402 | } | |
5403 | ||
ba2de7ba DW |
5404 | dv->dev = dev; |
5405 | dv->index = super->current_vol; | |
5406 | dv->next = super->devlist; | |
5407 | super->devlist = dv; | |
c2c087e6 | 5408 | |
4d1313e9 DW |
5409 | imsm_update_version_info(super); |
5410 | ||
c2c087e6 | 5411 | return 1; |
cdddbdbc DW |
5412 | } |
5413 | ||
bf5a934a | 5414 | static int init_super_imsm(struct supertype *st, mdu_array_info_t *info, |
5308f117 | 5415 | struct shape *s, char *name, |
83cd1e97 N |
5416 | char *homehost, int *uuid, |
5417 | unsigned long long data_offset) | |
bf5a934a DW |
5418 | { |
5419 | /* This is primarily called by Create when creating a new array. | |
5420 | * We will then get add_to_super called for each component, and then | |
5421 | * write_init_super called to write it out to each device. | |
5422 | * For IMSM, Create can create on fresh devices or on a pre-existing | |
5423 | * array. | |
5424 | * To create on a pre-existing array a different method will be called. | |
5425 | * This one is just for fresh drives. | |
5426 | */ | |
5427 | struct intel_super *super; | |
5428 | struct imsm_super *mpb; | |
5429 | size_t mpb_size; | |
4d1313e9 | 5430 | char *version; |
bf5a934a | 5431 | |
83cd1e97 | 5432 | if (data_offset != INVALID_SECTORS) { |
ed503f89 | 5433 | pr_err("data-offset not supported by imsm\n"); |
83cd1e97 N |
5434 | return 0; |
5435 | } | |
5436 | ||
bf5a934a | 5437 | if (st->sb) |
5308f117 | 5438 | return init_super_imsm_volume(st, info, s, name, homehost, uuid, |
83cd1e97 | 5439 | data_offset); |
e683ca88 DW |
5440 | |
5441 | if (info) | |
5442 | mpb_size = disks_to_mpb_size(info->nr_disks); | |
5443 | else | |
f36a9ecd | 5444 | mpb_size = MAX_SECTOR_SIZE; |
bf5a934a | 5445 | |
49133e57 | 5446 | super = alloc_super(); |
f36a9ecd PB |
5447 | if (super && |
5448 | posix_memalign(&super->buf, MAX_SECTOR_SIZE, mpb_size) != 0) { | |
8d67477f | 5449 | free_imsm(super); |
e683ca88 DW |
5450 | super = NULL; |
5451 | } | |
5452 | if (!super) { | |
1ade5cc1 | 5453 | pr_err("could not allocate superblock\n"); |
bf5a934a DW |
5454 | return 0; |
5455 | } | |
de44e46f PB |
5456 | if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE, |
5457 | MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) { | |
1ade5cc1 | 5458 | pr_err("could not allocate migr_rec buffer\n"); |
8e59f3d8 | 5459 | free(super->buf); |
8d67477f | 5460 | free_imsm(super); |
8e59f3d8 AK |
5461 | return 0; |
5462 | } | |
e683ca88 | 5463 | memset(super->buf, 0, mpb_size); |
ef649044 | 5464 | mpb = super->buf; |
e683ca88 DW |
5465 | mpb->mpb_size = __cpu_to_le32(mpb_size); |
5466 | st->sb = super; | |
5467 | ||
5468 | if (info == NULL) { | |
5469 | /* zeroing superblock */ | |
5470 | return 0; | |
5471 | } | |
bf5a934a | 5472 | |
4d1313e9 DW |
5473 | mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; |
5474 | ||
5475 | version = (char *) mpb->sig; | |
5476 | strcpy(version, MPB_SIGNATURE); | |
5477 | version += strlen(MPB_SIGNATURE); | |
5478 | strcpy(version, MPB_VERSION_RAID0); | |
bf5a934a | 5479 | |
bf5a934a DW |
5480 | return 1; |
5481 | } | |
5482 | ||
f2cc4f7d AO |
5483 | static int drive_validate_sector_size(struct intel_super *super, struct dl *dl) |
5484 | { | |
5485 | unsigned int member_sector_size; | |
5486 | ||
5487 | if (dl->fd < 0) { | |
5488 | pr_err("Invalid file descriptor for %s\n", dl->devname); | |
5489 | return 0; | |
5490 | } | |
5491 | ||
5492 | if (!get_dev_sector_size(dl->fd, dl->devname, &member_sector_size)) | |
5493 | return 0; | |
5494 | if (member_sector_size != super->sector_size) | |
5495 | return 0; | |
5496 | return 1; | |
5497 | } | |
5498 | ||
f20c3968 | 5499 | static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk, |
bf5a934a DW |
5500 | int fd, char *devname) |
5501 | { | |
5502 | struct intel_super *super = st->sb; | |
d23fe947 | 5503 | struct imsm_super *mpb = super->anchor; |
3960e579 | 5504 | struct imsm_disk *_disk; |
bf5a934a DW |
5505 | struct imsm_dev *dev; |
5506 | struct imsm_map *map; | |
3960e579 | 5507 | struct dl *dl, *df; |
4eb26970 | 5508 | int slot; |
bf5a934a | 5509 | |
949c47a0 | 5510 | dev = get_imsm_dev(super, super->current_vol); |
238c0a71 | 5511 | map = get_imsm_map(dev, MAP_0); |
bf5a934a | 5512 | |
208933a7 | 5513 | if (! (dk->state & (1<<MD_DISK_SYNC))) { |
e7b84f9d | 5514 | pr_err("%s: Cannot add spare devices to IMSM volume\n", |
208933a7 N |
5515 | devname); |
5516 | return 1; | |
5517 | } | |
5518 | ||
efb30e7f DW |
5519 | if (fd == -1) { |
5520 | /* we're doing autolayout so grab the pre-marked (in | |
5521 | * validate_geometry) raid_disk | |
5522 | */ | |
5523 | for (dl = super->disks; dl; dl = dl->next) | |
5524 | if (dl->raiddisk == dk->raid_disk) | |
5525 | break; | |
5526 | } else { | |
5527 | for (dl = super->disks; dl ; dl = dl->next) | |
5528 | if (dl->major == dk->major && | |
5529 | dl->minor == dk->minor) | |
5530 | break; | |
5531 | } | |
d23fe947 | 5532 | |
208933a7 | 5533 | if (!dl) { |
e7b84f9d | 5534 | pr_err("%s is not a member of the same container\n", devname); |
f20c3968 | 5535 | return 1; |
208933a7 | 5536 | } |
bf5a934a | 5537 | |
f2cc4f7d AO |
5538 | if (!drive_validate_sector_size(super, dl)) { |
5539 | pr_err("Combining drives of different sector size in one volume is not allowed\n"); | |
5540 | return 1; | |
5541 | } | |
5542 | ||
d23fe947 DW |
5543 | /* add a pristine spare to the metadata */ |
5544 | if (dl->index < 0) { | |
5545 | dl->index = super->anchor->num_disks; | |
5546 | super->anchor->num_disks++; | |
5547 | } | |
4eb26970 DW |
5548 | /* Check the device has not already been added */ |
5549 | slot = get_imsm_disk_slot(map, dl->index); | |
5550 | if (slot >= 0 && | |
238c0a71 | 5551 | (get_imsm_ord_tbl_ent(dev, slot, MAP_X) & IMSM_ORD_REBUILD) == 0) { |
e7b84f9d | 5552 | pr_err("%s has been included in this array twice\n", |
4eb26970 DW |
5553 | devname); |
5554 | return 1; | |
5555 | } | |
656b6b5a | 5556 | set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index); |
ee5aad5a | 5557 | dl->disk.status = CONFIGURED_DISK; |
d23fe947 | 5558 | |
3960e579 DW |
5559 | /* update size of 'missing' disks to be at least as large as the |
5560 | * largest acitve member (we only have dummy missing disks when | |
5561 | * creating the first volume) | |
5562 | */ | |
5563 | if (super->current_vol == 0) { | |
5564 | for (df = super->missing; df; df = df->next) { | |
5551b113 CA |
5565 | if (total_blocks(&dl->disk) > total_blocks(&df->disk)) |
5566 | set_total_blocks(&df->disk, total_blocks(&dl->disk)); | |
3960e579 DW |
5567 | _disk = __get_imsm_disk(mpb, df->index); |
5568 | *_disk = df->disk; | |
5569 | } | |
5570 | } | |
5571 | ||
5572 | /* refresh unset/failed slots to point to valid 'missing' entries */ | |
5573 | for (df = super->missing; df; df = df->next) | |
5574 | for (slot = 0; slot < mpb->num_disks; slot++) { | |
238c0a71 | 5575 | __u32 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); |
3960e579 DW |
5576 | |
5577 | if ((ord & IMSM_ORD_REBUILD) == 0) | |
5578 | continue; | |
5579 | set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD); | |
1ace8403 | 5580 | if (is_gen_migration(dev)) { |
238c0a71 AK |
5581 | struct imsm_map *map2 = get_imsm_map(dev, |
5582 | MAP_1); | |
0a108d63 | 5583 | int slot2 = get_imsm_disk_slot(map2, df->index); |
089f9d79 | 5584 | if (slot2 < map2->num_members && slot2 >= 0) { |
1ace8403 | 5585 | __u32 ord2 = get_imsm_ord_tbl_ent(dev, |
238c0a71 AK |
5586 | slot2, |
5587 | MAP_1); | |
1ace8403 AK |
5588 | if ((unsigned)df->index == |
5589 | ord_to_idx(ord2)) | |
5590 | set_imsm_ord_tbl_ent(map2, | |
0a108d63 | 5591 | slot2, |
1ace8403 AK |
5592 | df->index | |
5593 | IMSM_ORD_REBUILD); | |
5594 | } | |
5595 | } | |
3960e579 DW |
5596 | dprintf("set slot:%d to missing disk:%d\n", slot, df->index); |
5597 | break; | |
5598 | } | |
5599 | ||
d23fe947 DW |
5600 | /* if we are creating the first raid device update the family number */ |
5601 | if (super->current_vol == 0) { | |
5602 | __u32 sum; | |
5603 | struct imsm_dev *_dev = __get_imsm_dev(mpb, 0); | |
d23fe947 | 5604 | |
3960e579 | 5605 | _disk = __get_imsm_disk(mpb, dl->index); |
791b666a | 5606 | if (!_dev || !_disk) { |
e7b84f9d | 5607 | pr_err("BUG mpb setup error\n"); |
791b666a AW |
5608 | return 1; |
5609 | } | |
d23fe947 DW |
5610 | *_dev = *dev; |
5611 | *_disk = dl->disk; | |
148acb7b DW |
5612 | sum = random32(); |
5613 | sum += __gen_imsm_checksum(mpb); | |
d23fe947 | 5614 | mpb->family_num = __cpu_to_le32(sum); |
148acb7b | 5615 | mpb->orig_family_num = mpb->family_num; |
d23fe947 | 5616 | } |
ca0748fa | 5617 | super->current_disk = dl; |
f20c3968 | 5618 | return 0; |
bf5a934a DW |
5619 | } |
5620 | ||
a8619d23 AK |
5621 | /* mark_spare() |
5622 | * Function marks disk as spare and restores disk serial | |
5623 | * in case it was previously marked as failed by takeover operation | |
5624 | * reruns: | |
5625 | * -1 : critical error | |
5626 | * 0 : disk is marked as spare but serial is not set | |
5627 | * 1 : success | |
5628 | */ | |
5629 | int mark_spare(struct dl *disk) | |
5630 | { | |
5631 | __u8 serial[MAX_RAID_SERIAL_LEN]; | |
5632 | int ret_val = -1; | |
5633 | ||
5634 | if (!disk) | |
5635 | return ret_val; | |
5636 | ||
5637 | ret_val = 0; | |
5638 | if (!imsm_read_serial(disk->fd, NULL, serial)) { | |
5639 | /* Restore disk serial number, because takeover marks disk | |
5640 | * as failed and adds to serial ':0' before it becomes | |
5641 | * a spare disk. | |
5642 | */ | |
5643 | serialcpy(disk->serial, serial); | |
5644 | serialcpy(disk->disk.serial, serial); | |
5645 | ret_val = 1; | |
5646 | } | |
5647 | disk->disk.status = SPARE_DISK; | |
5648 | disk->index = -1; | |
5649 | ||
5650 | return ret_val; | |
5651 | } | |
88654014 | 5652 | |
f20c3968 | 5653 | static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk, |
72ca9bcf N |
5654 | int fd, char *devname, |
5655 | unsigned long long data_offset) | |
cdddbdbc | 5656 | { |
c2c087e6 | 5657 | struct intel_super *super = st->sb; |
c2c087e6 DW |
5658 | struct dl *dd; |
5659 | unsigned long long size; | |
fa7bb6f8 | 5660 | unsigned int member_sector_size; |
f2f27e63 | 5661 | __u32 id; |
c2c087e6 DW |
5662 | int rv; |
5663 | struct stat stb; | |
5664 | ||
88654014 LM |
5665 | /* If we are on an RAID enabled platform check that the disk is |
5666 | * attached to the raid controller. | |
5667 | * We do not need to test disks attachment for container based additions, | |
5668 | * they shall be already tested when container was created/assembled. | |
88c32bb1 | 5669 | */ |
d424212e | 5670 | rv = find_intel_hba_capability(fd, super, devname); |
f2f5c343 | 5671 | /* no orom/efi or non-intel hba of the disk */ |
f0f5a016 LM |
5672 | if (rv != 0) { |
5673 | dprintf("capability: %p fd: %d ret: %d\n", | |
5674 | super->orom, fd, rv); | |
5675 | return 1; | |
88c32bb1 DW |
5676 | } |
5677 | ||
f20c3968 DW |
5678 | if (super->current_vol >= 0) |
5679 | return add_to_super_imsm_volume(st, dk, fd, devname); | |
bf5a934a | 5680 | |
c2c087e6 | 5681 | fstat(fd, &stb); |
503975b9 | 5682 | dd = xcalloc(sizeof(*dd), 1); |
c2c087e6 DW |
5683 | dd->major = major(stb.st_rdev); |
5684 | dd->minor = minor(stb.st_rdev); | |
503975b9 | 5685 | dd->devname = devname ? xstrdup(devname) : NULL; |
c2c087e6 | 5686 | dd->fd = fd; |
689c9bf3 | 5687 | dd->e = NULL; |
1a64be56 | 5688 | dd->action = DISK_ADD; |
c2c087e6 | 5689 | rv = imsm_read_serial(fd, devname, dd->serial); |
32ba9157 | 5690 | if (rv) { |
e7b84f9d | 5691 | pr_err("failed to retrieve scsi serial, aborting\n"); |
20bee0f8 PB |
5692 | if (dd->devname) |
5693 | free(dd->devname); | |
949c47a0 | 5694 | free(dd); |
0030e8d6 | 5695 | abort(); |
c2c087e6 | 5696 | } |
20bee0f8 PB |
5697 | if (super->hba && ((super->hba->type == SYS_DEV_NVME) || |
5698 | (super->hba->type == SYS_DEV_VMD))) { | |
5699 | int i; | |
5700 | char *devpath = diskfd_to_devpath(fd); | |
5701 | char controller_path[PATH_MAX]; | |
5702 | ||
5703 | if (!devpath) { | |
5704 | pr_err("failed to get devpath, aborting\n"); | |
5705 | if (dd->devname) | |
5706 | free(dd->devname); | |
5707 | free(dd); | |
5708 | return 1; | |
5709 | } | |
5710 | ||
5711 | snprintf(controller_path, PATH_MAX-1, "%s/device", devpath); | |
5712 | free(devpath); | |
5713 | ||
5714 | if (devpath_to_vendor(controller_path) == 0x8086) { | |
5715 | /* | |
5716 | * If Intel's NVMe drive has serial ended with | |
5717 | * "-A","-B","-1" or "-2" it means that this is "x8" | |
5718 | * device (double drive on single PCIe card). | |
5719 | * User should be warned about potential data loss. | |
5720 | */ | |
5721 | for (i = MAX_RAID_SERIAL_LEN-1; i > 0; i--) { | |
5722 | /* Skip empty character at the end */ | |
5723 | if (dd->serial[i] == 0) | |
5724 | continue; | |
5725 | ||
5726 | if (((dd->serial[i] == 'A') || | |
5727 | (dd->serial[i] == 'B') || | |
5728 | (dd->serial[i] == '1') || | |
5729 | (dd->serial[i] == '2')) && | |
5730 | (dd->serial[i-1] == '-')) | |
5731 | pr_err("\tThe action you are about to take may put your data at risk.\n" | |
5732 | "\tPlease note that x8 devices may consist of two separate x4 devices " | |
5733 | "located on a single PCIe port.\n" | |
5734 | "\tRAID 0 is the only supported configuration for this type of x8 device.\n"); | |
5735 | break; | |
5736 | } | |
32716c51 PB |
5737 | } else if (super->hba->type == SYS_DEV_VMD && super->orom && |
5738 | !imsm_orom_has_tpv_support(super->orom)) { | |
5739 | pr_err("\tPlatform configuration does not support non-Intel NVMe drives.\n" | |
5740 | "\tPlease refer to Intel(R) RSTe user guide.\n"); | |
5741 | free(dd->devname); | |
5742 | free(dd); | |
5743 | return 1; | |
20bee0f8 PB |
5744 | } |
5745 | } | |
c2c087e6 | 5746 | |
c2c087e6 | 5747 | get_dev_size(fd, NULL, &size); |
fa7bb6f8 PB |
5748 | get_dev_sector_size(fd, NULL, &member_sector_size); |
5749 | ||
5750 | if (super->sector_size == 0) { | |
5751 | /* this a first device, so sector_size is not set yet */ | |
5752 | super->sector_size = member_sector_size; | |
fa7bb6f8 PB |
5753 | } |
5754 | ||
71e5411e | 5755 | /* clear migr_rec when adding disk to container */ |
85337573 AO |
5756 | memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE); |
5757 | if (lseek64(fd, size - MIGR_REC_SECTOR_POSITION*member_sector_size, | |
de44e46f | 5758 | SEEK_SET) >= 0) { |
466070ad | 5759 | if ((unsigned int)write(fd, super->migr_rec_buf, |
85337573 AO |
5760 | MIGR_REC_BUF_SECTORS*member_sector_size) != |
5761 | MIGR_REC_BUF_SECTORS*member_sector_size) | |
71e5411e PB |
5762 | perror("Write migr_rec failed"); |
5763 | } | |
5764 | ||
c2c087e6 | 5765 | size /= 512; |
1f24f035 | 5766 | serialcpy(dd->disk.serial, dd->serial); |
5551b113 CA |
5767 | set_total_blocks(&dd->disk, size); |
5768 | if (__le32_to_cpu(dd->disk.total_blocks_hi) > 0) { | |
5769 | struct imsm_super *mpb = super->anchor; | |
5770 | mpb->attributes |= MPB_ATTRIB_2TB_DISK; | |
5771 | } | |
a8619d23 | 5772 | mark_spare(dd); |
c2c087e6 | 5773 | if (sysfs_disk_to_scsi_id(fd, &id) == 0) |
b9f594fe | 5774 | dd->disk.scsi_id = __cpu_to_le32(id); |
c2c087e6 | 5775 | else |
b9f594fe | 5776 | dd->disk.scsi_id = __cpu_to_le32(0); |
43dad3d6 DW |
5777 | |
5778 | if (st->update_tail) { | |
1a64be56 LM |
5779 | dd->next = super->disk_mgmt_list; |
5780 | super->disk_mgmt_list = dd; | |
43dad3d6 DW |
5781 | } else { |
5782 | dd->next = super->disks; | |
5783 | super->disks = dd; | |
ceaf0ee1 | 5784 | super->updates_pending++; |
43dad3d6 | 5785 | } |
f20c3968 DW |
5786 | |
5787 | return 0; | |
cdddbdbc DW |
5788 | } |
5789 | ||
1a64be56 LM |
5790 | static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk) |
5791 | { | |
5792 | struct intel_super *super = st->sb; | |
5793 | struct dl *dd; | |
5794 | ||
5795 | /* remove from super works only in mdmon - for communication | |
5796 | * manager - monitor. Check if communication memory buffer | |
5797 | * is prepared. | |
5798 | */ | |
5799 | if (!st->update_tail) { | |
1ade5cc1 | 5800 | pr_err("shall be used in mdmon context only\n"); |
1a64be56 LM |
5801 | return 1; |
5802 | } | |
503975b9 | 5803 | dd = xcalloc(1, sizeof(*dd)); |
1a64be56 LM |
5804 | dd->major = dk->major; |
5805 | dd->minor = dk->minor; | |
1a64be56 | 5806 | dd->fd = -1; |
a8619d23 | 5807 | mark_spare(dd); |
1a64be56 LM |
5808 | dd->action = DISK_REMOVE; |
5809 | ||
5810 | dd->next = super->disk_mgmt_list; | |
5811 | super->disk_mgmt_list = dd; | |
5812 | ||
1a64be56 LM |
5813 | return 0; |
5814 | } | |
5815 | ||
f796af5d DW |
5816 | static int store_imsm_mpb(int fd, struct imsm_super *mpb); |
5817 | ||
5818 | static union { | |
f36a9ecd | 5819 | char buf[MAX_SECTOR_SIZE]; |
f796af5d | 5820 | struct imsm_super anchor; |
f36a9ecd | 5821 | } spare_record __attribute__ ((aligned(MAX_SECTOR_SIZE))); |
c2c087e6 | 5822 | |
d23fe947 DW |
5823 | /* spare records have their own family number and do not have any defined raid |
5824 | * devices | |
5825 | */ | |
5826 | static int write_super_imsm_spares(struct intel_super *super, int doclose) | |
5827 | { | |
d23fe947 | 5828 | struct imsm_super *mpb = super->anchor; |
f796af5d | 5829 | struct imsm_super *spare = &spare_record.anchor; |
d23fe947 DW |
5830 | __u32 sum; |
5831 | struct dl *d; | |
5832 | ||
68641cdb JS |
5833 | spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super)); |
5834 | spare->generation_num = __cpu_to_le32(1UL); | |
f796af5d | 5835 | spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY; |
68641cdb JS |
5836 | spare->num_disks = 1; |
5837 | spare->num_raid_devs = 0; | |
5838 | spare->cache_size = mpb->cache_size; | |
5839 | spare->pwr_cycle_count = __cpu_to_le32(1); | |
f796af5d DW |
5840 | |
5841 | snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH, | |
5842 | MPB_SIGNATURE MPB_VERSION_RAID0); | |
d23fe947 DW |
5843 | |
5844 | for (d = super->disks; d; d = d->next) { | |
8796fdc4 | 5845 | if (d->index != -1) |
d23fe947 DW |
5846 | continue; |
5847 | ||
f796af5d | 5848 | spare->disk[0] = d->disk; |
027c374f CA |
5849 | if (__le32_to_cpu(d->disk.total_blocks_hi) > 0) |
5850 | spare->attributes |= MPB_ATTRIB_2TB_DISK; | |
5851 | ||
f36a9ecd PB |
5852 | if (super->sector_size == 4096) |
5853 | convert_to_4k_imsm_disk(&spare->disk[0]); | |
5854 | ||
f796af5d DW |
5855 | sum = __gen_imsm_checksum(spare); |
5856 | spare->family_num = __cpu_to_le32(sum); | |
5857 | spare->orig_family_num = 0; | |
5858 | sum = __gen_imsm_checksum(spare); | |
5859 | spare->check_sum = __cpu_to_le32(sum); | |
d23fe947 | 5860 | |
f796af5d | 5861 | if (store_imsm_mpb(d->fd, spare)) { |
1ade5cc1 N |
5862 | pr_err("failed for device %d:%d %s\n", |
5863 | d->major, d->minor, strerror(errno)); | |
e74255d9 | 5864 | return 1; |
d23fe947 DW |
5865 | } |
5866 | if (doclose) { | |
5867 | close(d->fd); | |
5868 | d->fd = -1; | |
5869 | } | |
5870 | } | |
5871 | ||
e74255d9 | 5872 | return 0; |
d23fe947 DW |
5873 | } |
5874 | ||
36988a3d | 5875 | static int write_super_imsm(struct supertype *st, int doclose) |
cdddbdbc | 5876 | { |
36988a3d | 5877 | struct intel_super *super = st->sb; |
f36a9ecd | 5878 | unsigned int sector_size = super->sector_size; |
949c47a0 | 5879 | struct imsm_super *mpb = super->anchor; |
c2c087e6 DW |
5880 | struct dl *d; |
5881 | __u32 generation; | |
5882 | __u32 sum; | |
d23fe947 | 5883 | int spares = 0; |
949c47a0 | 5884 | int i; |
a48ac0a8 | 5885 | __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk); |
36988a3d | 5886 | int num_disks = 0; |
146c6260 | 5887 | int clear_migration_record = 1; |
bbab0940 | 5888 | __u32 bbm_log_size; |
cdddbdbc | 5889 | |
c2c087e6 DW |
5890 | /* 'generation' is incremented everytime the metadata is written */ |
5891 | generation = __le32_to_cpu(mpb->generation_num); | |
5892 | generation++; | |
5893 | mpb->generation_num = __cpu_to_le32(generation); | |
5894 | ||
148acb7b DW |
5895 | /* fix up cases where previous mdadm releases failed to set |
5896 | * orig_family_num | |
5897 | */ | |
5898 | if (mpb->orig_family_num == 0) | |
5899 | mpb->orig_family_num = mpb->family_num; | |
5900 | ||
d23fe947 | 5901 | for (d = super->disks; d; d = d->next) { |
8796fdc4 | 5902 | if (d->index == -1) |
d23fe947 | 5903 | spares++; |
36988a3d | 5904 | else { |
d23fe947 | 5905 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
5906 | num_disks++; |
5907 | } | |
d23fe947 | 5908 | } |
36988a3d | 5909 | for (d = super->missing; d; d = d->next) { |
47ee5a45 | 5910 | mpb->disk[d->index] = d->disk; |
36988a3d AK |
5911 | num_disks++; |
5912 | } | |
5913 | mpb->num_disks = num_disks; | |
5914 | mpb_size += sizeof(struct imsm_disk) * mpb->num_disks; | |
b9f594fe | 5915 | |
949c47a0 DW |
5916 | for (i = 0; i < mpb->num_raid_devs; i++) { |
5917 | struct imsm_dev *dev = __get_imsm_dev(mpb, i); | |
36988a3d AK |
5918 | struct imsm_dev *dev2 = get_imsm_dev(super, i); |
5919 | if (dev && dev2) { | |
5920 | imsm_copy_dev(dev, dev2); | |
5921 | mpb_size += sizeof_imsm_dev(dev, 0); | |
5922 | } | |
146c6260 AK |
5923 | if (is_gen_migration(dev2)) |
5924 | clear_migration_record = 0; | |
949c47a0 | 5925 | } |
bbab0940 TM |
5926 | |
5927 | bbm_log_size = get_imsm_bbm_log_size(super->bbm_log); | |
5928 | ||
5929 | if (bbm_log_size) { | |
5930 | memcpy((void *)mpb + mpb_size, super->bbm_log, bbm_log_size); | |
5931 | mpb->attributes |= MPB_ATTRIB_BBM; | |
5932 | } else | |
5933 | mpb->attributes &= ~MPB_ATTRIB_BBM; | |
5934 | ||
5935 | super->anchor->bbm_log_size = __cpu_to_le32(bbm_log_size); | |
5936 | mpb_size += bbm_log_size; | |
a48ac0a8 | 5937 | mpb->mpb_size = __cpu_to_le32(mpb_size); |
949c47a0 | 5938 | |
bbab0940 TM |
5939 | #ifdef DEBUG |
5940 | assert(super->len == 0 || mpb_size <= super->len); | |
5941 | #endif | |
5942 | ||
c2c087e6 | 5943 | /* recalculate checksum */ |
949c47a0 | 5944 | sum = __gen_imsm_checksum(mpb); |
c2c087e6 DW |
5945 | mpb->check_sum = __cpu_to_le32(sum); |
5946 | ||
51d83f5d AK |
5947 | if (super->clean_migration_record_by_mdmon) { |
5948 | clear_migration_record = 1; | |
5949 | super->clean_migration_record_by_mdmon = 0; | |
5950 | } | |
146c6260 | 5951 | if (clear_migration_record) |
de44e46f | 5952 | memset(super->migr_rec_buf, 0, |
85337573 | 5953 | MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE); |
146c6260 | 5954 | |
f36a9ecd PB |
5955 | if (sector_size == 4096) |
5956 | convert_to_4k(super); | |
5957 | ||
d23fe947 | 5958 | /* write the mpb for disks that compose raid devices */ |
c2c087e6 | 5959 | for (d = super->disks; d ; d = d->next) { |
86c54047 | 5960 | if (d->index < 0 || is_failed(&d->disk)) |
d23fe947 | 5961 | continue; |
30602f53 | 5962 | |
146c6260 AK |
5963 | if (clear_migration_record) { |
5964 | unsigned long long dsize; | |
5965 | ||
5966 | get_dev_size(d->fd, NULL, &dsize); | |
de44e46f PB |
5967 | if (lseek64(d->fd, dsize - sector_size, |
5968 | SEEK_SET) >= 0) { | |
466070ad PB |
5969 | if ((unsigned int)write(d->fd, |
5970 | super->migr_rec_buf, | |
de44e46f PB |
5971 | MIGR_REC_BUF_SECTORS*sector_size) != |
5972 | MIGR_REC_BUF_SECTORS*sector_size) | |
9e2d750d | 5973 | perror("Write migr_rec failed"); |
146c6260 AK |
5974 | } |
5975 | } | |
51d83f5d AK |
5976 | |
5977 | if (store_imsm_mpb(d->fd, mpb)) | |
5978 | fprintf(stderr, | |
1ade5cc1 N |
5979 | "failed for device %d:%d (fd: %d)%s\n", |
5980 | d->major, d->minor, | |
51d83f5d AK |
5981 | d->fd, strerror(errno)); |
5982 | ||
c2c087e6 DW |
5983 | if (doclose) { |
5984 | close(d->fd); | |
5985 | d->fd = -1; | |
5986 | } | |
5987 | } | |
5988 | ||
d23fe947 DW |
5989 | if (spares) |
5990 | return write_super_imsm_spares(super, doclose); | |
5991 | ||
e74255d9 | 5992 | return 0; |
c2c087e6 DW |
5993 | } |
5994 | ||
9b1fb677 | 5995 | static int create_array(struct supertype *st, int dev_idx) |
43dad3d6 DW |
5996 | { |
5997 | size_t len; | |
5998 | struct imsm_update_create_array *u; | |
5999 | struct intel_super *super = st->sb; | |
9b1fb677 | 6000 | struct imsm_dev *dev = get_imsm_dev(super, dev_idx); |
238c0a71 | 6001 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
54c2c1ea DW |
6002 | struct disk_info *inf; |
6003 | struct imsm_disk *disk; | |
6004 | int i; | |
43dad3d6 | 6005 | |
54c2c1ea DW |
6006 | len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) + |
6007 | sizeof(*inf) * map->num_members; | |
503975b9 | 6008 | u = xmalloc(len); |
43dad3d6 | 6009 | u->type = update_create_array; |
9b1fb677 | 6010 | u->dev_idx = dev_idx; |
43dad3d6 | 6011 | imsm_copy_dev(&u->dev, dev); |
54c2c1ea DW |
6012 | inf = get_disk_info(u); |
6013 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 6014 | int idx = get_imsm_disk_idx(dev, i, MAP_X); |
9b1fb677 | 6015 | |
54c2c1ea | 6016 | disk = get_imsm_disk(super, idx); |
1ca5c8e0 N |
6017 | if (!disk) |
6018 | disk = get_imsm_missing(super, idx); | |
54c2c1ea DW |
6019 | serialcpy(inf[i].serial, disk->serial); |
6020 | } | |
43dad3d6 DW |
6021 | append_metadata_update(st, u, len); |
6022 | ||
6023 | return 0; | |
6024 | } | |
6025 | ||
1a64be56 | 6026 | static int mgmt_disk(struct supertype *st) |
43dad3d6 DW |
6027 | { |
6028 | struct intel_super *super = st->sb; | |
6029 | size_t len; | |
1a64be56 | 6030 | struct imsm_update_add_remove_disk *u; |
43dad3d6 | 6031 | |
1a64be56 | 6032 | if (!super->disk_mgmt_list) |
43dad3d6 DW |
6033 | return 0; |
6034 | ||
6035 | len = sizeof(*u); | |
503975b9 | 6036 | u = xmalloc(len); |
1a64be56 | 6037 | u->type = update_add_remove_disk; |
43dad3d6 DW |
6038 | append_metadata_update(st, u, len); |
6039 | ||
6040 | return 0; | |
6041 | } | |
2432ce9b AP |
6042 | |
6043 | __u32 crc32c_le(__u32 crc, unsigned char const *p, size_t len); | |
6044 | ||
6045 | static int write_init_ppl_imsm(struct supertype *st, struct mdinfo *info, int fd) | |
6046 | { | |
6047 | struct intel_super *super = st->sb; | |
6048 | void *buf; | |
6049 | struct ppl_header *ppl_hdr; | |
6050 | int ret; | |
6051 | ||
6052 | ret = posix_memalign(&buf, 4096, PPL_HEADER_SIZE); | |
6053 | if (ret) { | |
6054 | pr_err("Failed to allocate PPL header buffer\n"); | |
6055 | return ret; | |
6056 | } | |
6057 | ||
6058 | memset(buf, 0, PPL_HEADER_SIZE); | |
6059 | ppl_hdr = buf; | |
6060 | memset(ppl_hdr->reserved, 0xff, PPL_HDR_RESERVED); | |
6061 | ppl_hdr->signature = __cpu_to_le32(super->anchor->orig_family_num); | |
6062 | ppl_hdr->checksum = __cpu_to_le32(~crc32c_le(~0, buf, PPL_HEADER_SIZE)); | |
6063 | ||
6064 | if (lseek64(fd, info->ppl_sector * 512, SEEK_SET) < 0) { | |
6065 | ret = errno; | |
6066 | perror("Failed to seek to PPL header location"); | |
6067 | } | |
6068 | ||
6069 | if (!ret && write(fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) { | |
6070 | ret = errno; | |
6071 | perror("Write PPL header failed"); | |
6072 | } | |
6073 | ||
6074 | if (!ret) | |
6075 | fsync(fd); | |
6076 | ||
6077 | free(buf); | |
6078 | return ret; | |
6079 | } | |
6080 | ||
6081 | static int validate_ppl_imsm(struct supertype *st, struct mdinfo *info, | |
6082 | struct mdinfo *disk) | |
6083 | { | |
6084 | struct intel_super *super = st->sb; | |
6085 | struct dl *d; | |
6086 | void *buf; | |
6087 | int ret = 0; | |
6088 | struct ppl_header *ppl_hdr; | |
6089 | __u32 crc; | |
6090 | struct imsm_dev *dev; | |
6091 | struct imsm_map *map; | |
6092 | __u32 idx; | |
6093 | ||
6094 | if (disk->disk.raid_disk < 0) | |
6095 | return 0; | |
6096 | ||
6097 | if (posix_memalign(&buf, 4096, PPL_HEADER_SIZE)) { | |
6098 | pr_err("Failed to allocate PPL header buffer\n"); | |
6099 | return -1; | |
6100 | } | |
6101 | ||
6102 | dev = get_imsm_dev(super, info->container_member); | |
6103 | map = get_imsm_map(dev, MAP_X); | |
6104 | idx = get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_X); | |
6105 | d = get_imsm_dl_disk(super, idx); | |
6106 | ||
6107 | if (!d || d->index < 0 || is_failed(&d->disk)) | |
6108 | goto out; | |
6109 | ||
6110 | if (lseek64(d->fd, info->ppl_sector * 512, SEEK_SET) < 0) { | |
6111 | perror("Failed to seek to PPL header location"); | |
6112 | ret = -1; | |
6113 | goto out; | |
6114 | } | |
6115 | ||
6116 | if (read(d->fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) { | |
6117 | perror("Read PPL header failed"); | |
6118 | ret = -1; | |
6119 | goto out; | |
6120 | } | |
6121 | ||
6122 | ppl_hdr = buf; | |
6123 | ||
6124 | crc = __le32_to_cpu(ppl_hdr->checksum); | |
6125 | ppl_hdr->checksum = 0; | |
6126 | ||
6127 | if (crc != ~crc32c_le(~0, buf, PPL_HEADER_SIZE)) { | |
6128 | dprintf("Wrong PPL header checksum on %s\n", | |
6129 | d->devname); | |
6130 | ret = 1; | |
6131 | } | |
6132 | ||
6133 | if (!ret && (__le32_to_cpu(ppl_hdr->signature) != | |
6134 | super->anchor->orig_family_num)) { | |
6135 | dprintf("Wrong PPL header signature on %s\n", | |
6136 | d->devname); | |
6137 | ret = 1; | |
6138 | } | |
6139 | ||
6140 | out: | |
6141 | free(buf); | |
6142 | ||
6143 | if (ret == 1 && map->map_state == IMSM_T_STATE_UNINITIALIZED) | |
6144 | return st->ss->write_init_ppl(st, info, d->fd); | |
6145 | ||
6146 | return ret; | |
6147 | } | |
6148 | ||
2432ce9b AP |
6149 | static int write_init_ppl_imsm_all(struct supertype *st, struct mdinfo *info) |
6150 | { | |
6151 | struct intel_super *super = st->sb; | |
6152 | struct dl *d; | |
6153 | int ret = 0; | |
6154 | ||
6155 | if (info->consistency_policy != CONSISTENCY_POLICY_PPL || | |
6156 | info->array.level != 5) | |
6157 | return 0; | |
6158 | ||
6159 | for (d = super->disks; d ; d = d->next) { | |
6160 | if (d->index < 0 || is_failed(&d->disk)) | |
6161 | continue; | |
6162 | ||
6163 | ret = st->ss->write_init_ppl(st, info, d->fd); | |
6164 | if (ret) | |
6165 | break; | |
6166 | } | |
6167 | ||
6168 | return ret; | |
6169 | } | |
43dad3d6 | 6170 | |
c2c087e6 DW |
6171 | static int write_init_super_imsm(struct supertype *st) |
6172 | { | |
9b1fb677 DW |
6173 | struct intel_super *super = st->sb; |
6174 | int current_vol = super->current_vol; | |
2432ce9b AP |
6175 | int rv = 0; |
6176 | struct mdinfo info; | |
6177 | ||
6178 | getinfo_super_imsm(st, &info, NULL); | |
9b1fb677 DW |
6179 | |
6180 | /* we are done with current_vol reset it to point st at the container */ | |
6181 | super->current_vol = -1; | |
6182 | ||
8273f55e | 6183 | if (st->update_tail) { |
43dad3d6 DW |
6184 | /* queue the recently created array / added disk |
6185 | * as a metadata update */ | |
8273f55e | 6186 | |
43dad3d6 | 6187 | /* determine if we are creating a volume or adding a disk */ |
9b1fb677 | 6188 | if (current_vol < 0) { |
1a64be56 LM |
6189 | /* in the mgmt (add/remove) disk case we are running |
6190 | * in mdmon context, so don't close fd's | |
43dad3d6 | 6191 | */ |
2432ce9b AP |
6192 | rv = mgmt_disk(st); |
6193 | } else { | |
6194 | rv = write_init_ppl_imsm_all(st, &info); | |
6195 | if (!rv) | |
6196 | rv = create_array(st, current_vol); | |
6197 | } | |
d682f344 N |
6198 | } else { |
6199 | struct dl *d; | |
6200 | for (d = super->disks; d; d = d->next) | |
ba728be7 | 6201 | Kill(d->devname, NULL, 0, -1, 1); |
2432ce9b AP |
6202 | if (current_vol >= 0) |
6203 | rv = write_init_ppl_imsm_all(st, &info); | |
6204 | if (!rv) | |
6205 | rv = write_super_imsm(st, 1); | |
d682f344 | 6206 | } |
2432ce9b AP |
6207 | |
6208 | return rv; | |
cdddbdbc DW |
6209 | } |
6210 | ||
e683ca88 | 6211 | static int store_super_imsm(struct supertype *st, int fd) |
cdddbdbc | 6212 | { |
e683ca88 DW |
6213 | struct intel_super *super = st->sb; |
6214 | struct imsm_super *mpb = super ? super->anchor : NULL; | |
551c80c1 | 6215 | |
e683ca88 | 6216 | if (!mpb) |
ad97895e DW |
6217 | return 1; |
6218 | ||
f36a9ecd PB |
6219 | if (super->sector_size == 4096) |
6220 | convert_to_4k(super); | |
e683ca88 | 6221 | return store_imsm_mpb(fd, mpb); |
cdddbdbc DW |
6222 | } |
6223 | ||
cdddbdbc DW |
6224 | static int validate_geometry_imsm_container(struct supertype *st, int level, |
6225 | int layout, int raiddisks, int chunk, | |
af4348dd N |
6226 | unsigned long long size, |
6227 | unsigned long long data_offset, | |
6228 | char *dev, | |
2c514b71 NB |
6229 | unsigned long long *freesize, |
6230 | int verbose) | |
cdddbdbc | 6231 | { |
c2c087e6 DW |
6232 | int fd; |
6233 | unsigned long long ldsize; | |
594dc1b8 | 6234 | struct intel_super *super; |
f2f5c343 | 6235 | int rv = 0; |
cdddbdbc | 6236 | |
c2c087e6 DW |
6237 | if (level != LEVEL_CONTAINER) |
6238 | return 0; | |
6239 | if (!dev) | |
6240 | return 1; | |
6241 | ||
6242 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
6243 | if (fd < 0) { | |
ba728be7 | 6244 | if (verbose > 0) |
e7b84f9d | 6245 | pr_err("imsm: Cannot open %s: %s\n", |
2c514b71 | 6246 | dev, strerror(errno)); |
c2c087e6 DW |
6247 | return 0; |
6248 | } | |
6249 | if (!get_dev_size(fd, dev, &ldsize)) { | |
6250 | close(fd); | |
6251 | return 0; | |
6252 | } | |
f2f5c343 LM |
6253 | |
6254 | /* capabilities retrieve could be possible | |
6255 | * note that there is no fd for the disks in array. | |
6256 | */ | |
6257 | super = alloc_super(); | |
8d67477f TM |
6258 | if (!super) { |
6259 | close(fd); | |
6260 | return 0; | |
6261 | } | |
fa7bb6f8 PB |
6262 | if (!get_dev_sector_size(fd, NULL, &super->sector_size)) { |
6263 | close(fd); | |
6264 | free_imsm(super); | |
6265 | return 0; | |
6266 | } | |
6267 | ||
ba728be7 | 6268 | rv = find_intel_hba_capability(fd, super, verbose > 0 ? dev : NULL); |
f2f5c343 LM |
6269 | if (rv != 0) { |
6270 | #if DEBUG | |
6271 | char str[256]; | |
6272 | fd2devname(fd, str); | |
1ade5cc1 | 6273 | dprintf("fd: %d %s orom: %p rv: %d raiddisk: %d\n", |
f2f5c343 LM |
6274 | fd, str, super->orom, rv, raiddisks); |
6275 | #endif | |
6276 | /* no orom/efi or non-intel hba of the disk */ | |
6277 | close(fd); | |
6278 | free_imsm(super); | |
6279 | return 0; | |
6280 | } | |
c2c087e6 | 6281 | close(fd); |
9126b9a8 CA |
6282 | if (super->orom) { |
6283 | if (raiddisks > super->orom->tds) { | |
6284 | if (verbose) | |
7a862a02 | 6285 | pr_err("%d exceeds maximum number of platform supported disks: %d\n", |
9126b9a8 CA |
6286 | raiddisks, super->orom->tds); |
6287 | free_imsm(super); | |
6288 | return 0; | |
6289 | } | |
6290 | if ((super->orom->attr & IMSM_OROM_ATTR_2TB_DISK) == 0 && | |
6291 | (ldsize >> 9) >> 32 > 0) { | |
6292 | if (verbose) | |
e7b84f9d | 6293 | pr_err("%s exceeds maximum platform supported size\n", dev); |
9126b9a8 CA |
6294 | free_imsm(super); |
6295 | return 0; | |
6296 | } | |
f2f5c343 | 6297 | } |
c2c087e6 | 6298 | |
af4348dd | 6299 | *freesize = avail_size_imsm(st, ldsize >> 9, data_offset); |
f2f5c343 | 6300 | free_imsm(super); |
c2c087e6 DW |
6301 | |
6302 | return 1; | |
cdddbdbc DW |
6303 | } |
6304 | ||
0dcecb2e DW |
6305 | static unsigned long long find_size(struct extent *e, int *idx, int num_extents) |
6306 | { | |
6307 | const unsigned long long base_start = e[*idx].start; | |
6308 | unsigned long long end = base_start + e[*idx].size; | |
6309 | int i; | |
6310 | ||
6311 | if (base_start == end) | |
6312 | return 0; | |
6313 | ||
6314 | *idx = *idx + 1; | |
6315 | for (i = *idx; i < num_extents; i++) { | |
6316 | /* extend overlapping extents */ | |
6317 | if (e[i].start >= base_start && | |
6318 | e[i].start <= end) { | |
6319 | if (e[i].size == 0) | |
6320 | return 0; | |
6321 | if (e[i].start + e[i].size > end) | |
6322 | end = e[i].start + e[i].size; | |
6323 | } else if (e[i].start > end) { | |
6324 | *idx = i; | |
6325 | break; | |
6326 | } | |
6327 | } | |
6328 | ||
6329 | return end - base_start; | |
6330 | } | |
6331 | ||
6332 | static unsigned long long merge_extents(struct intel_super *super, int sum_extents) | |
6333 | { | |
6334 | /* build a composite disk with all known extents and generate a new | |
6335 | * 'maxsize' given the "all disks in an array must share a common start | |
6336 | * offset" constraint | |
6337 | */ | |
503975b9 | 6338 | struct extent *e = xcalloc(sum_extents, sizeof(*e)); |
0dcecb2e DW |
6339 | struct dl *dl; |
6340 | int i, j; | |
6341 | int start_extent; | |
6342 | unsigned long long pos; | |
b9d77223 | 6343 | unsigned long long start = 0; |
0dcecb2e DW |
6344 | unsigned long long maxsize; |
6345 | unsigned long reserve; | |
6346 | ||
0dcecb2e DW |
6347 | /* coalesce and sort all extents. also, check to see if we need to |
6348 | * reserve space between member arrays | |
6349 | */ | |
6350 | j = 0; | |
6351 | for (dl = super->disks; dl; dl = dl->next) { | |
6352 | if (!dl->e) | |
6353 | continue; | |
6354 | for (i = 0; i < dl->extent_cnt; i++) | |
6355 | e[j++] = dl->e[i]; | |
6356 | } | |
6357 | qsort(e, sum_extents, sizeof(*e), cmp_extent); | |
6358 | ||
6359 | /* merge extents */ | |
6360 | i = 0; | |
6361 | j = 0; | |
6362 | while (i < sum_extents) { | |
6363 | e[j].start = e[i].start; | |
6364 | e[j].size = find_size(e, &i, sum_extents); | |
6365 | j++; | |
6366 | if (e[j-1].size == 0) | |
6367 | break; | |
6368 | } | |
6369 | ||
6370 | pos = 0; | |
6371 | maxsize = 0; | |
6372 | start_extent = 0; | |
6373 | i = 0; | |
6374 | do { | |
6375 | unsigned long long esize; | |
6376 | ||
6377 | esize = e[i].start - pos; | |
6378 | if (esize >= maxsize) { | |
6379 | maxsize = esize; | |
6380 | start = pos; | |
6381 | start_extent = i; | |
6382 | } | |
6383 | pos = e[i].start + e[i].size; | |
6384 | i++; | |
6385 | } while (e[i-1].size); | |
6386 | free(e); | |
6387 | ||
a7dd165b DW |
6388 | if (maxsize == 0) |
6389 | return 0; | |
6390 | ||
6391 | /* FIXME assumes volume at offset 0 is the first volume in a | |
6392 | * container | |
6393 | */ | |
0dcecb2e DW |
6394 | if (start_extent > 0) |
6395 | reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */ | |
6396 | else | |
6397 | reserve = 0; | |
6398 | ||
6399 | if (maxsize < reserve) | |
a7dd165b | 6400 | return 0; |
0dcecb2e | 6401 | |
5551b113 | 6402 | super->create_offset = ~((unsigned long long) 0); |
0dcecb2e | 6403 | if (start + reserve > super->create_offset) |
a7dd165b | 6404 | return 0; /* start overflows create_offset */ |
0dcecb2e DW |
6405 | super->create_offset = start + reserve; |
6406 | ||
6407 | return maxsize - reserve; | |
6408 | } | |
6409 | ||
88c32bb1 DW |
6410 | static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks) |
6411 | { | |
6412 | if (level < 0 || level == 6 || level == 4) | |
6413 | return 0; | |
6414 | ||
6415 | /* if we have an orom prevent invalid raid levels */ | |
6416 | if (orom) | |
6417 | switch (level) { | |
6418 | case 0: return imsm_orom_has_raid0(orom); | |
6419 | case 1: | |
6420 | if (raiddisks > 2) | |
6421 | return imsm_orom_has_raid1e(orom); | |
1c556e92 DW |
6422 | return imsm_orom_has_raid1(orom) && raiddisks == 2; |
6423 | case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4; | |
6424 | case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2; | |
88c32bb1 DW |
6425 | } |
6426 | else | |
6427 | return 1; /* not on an Intel RAID platform so anything goes */ | |
6428 | ||
6429 | return 0; | |
6430 | } | |
6431 | ||
ca9de185 LM |
6432 | static int |
6433 | active_arrays_by_format(char *name, char* hba, struct md_list **devlist, | |
6434 | int dpa, int verbose) | |
6435 | { | |
6436 | struct mdstat_ent *mdstat = mdstat_read(0, 0); | |
594dc1b8 | 6437 | struct mdstat_ent *memb; |
ca9de185 LM |
6438 | int count = 0; |
6439 | int num = 0; | |
594dc1b8 | 6440 | struct md_list *dv; |
ca9de185 LM |
6441 | int found; |
6442 | ||
6443 | for (memb = mdstat ; memb ; memb = memb->next) { | |
6444 | if (memb->metadata_version && | |
fc54fe7a | 6445 | (strncmp(memb->metadata_version, "external:", 9) == 0) && |
ca9de185 LM |
6446 | (strcmp(&memb->metadata_version[9], name) == 0) && |
6447 | !is_subarray(memb->metadata_version+9) && | |
6448 | memb->members) { | |
6449 | struct dev_member *dev = memb->members; | |
6450 | int fd = -1; | |
6451 | while(dev && (fd < 0)) { | |
503975b9 N |
6452 | char *path = xmalloc(strlen(dev->name) + strlen("/dev/") + 1); |
6453 | num = sprintf(path, "%s%s", "/dev/", dev->name); | |
6454 | if (num > 0) | |
6455 | fd = open(path, O_RDONLY, 0); | |
089f9d79 | 6456 | if (num <= 0 || fd < 0) { |
676e87a8 | 6457 | pr_vrb("Cannot open %s: %s\n", |
503975b9 | 6458 | dev->name, strerror(errno)); |
ca9de185 | 6459 | } |
503975b9 | 6460 | free(path); |
ca9de185 LM |
6461 | dev = dev->next; |
6462 | } | |
6463 | found = 0; | |
089f9d79 | 6464 | if (fd >= 0 && disk_attached_to_hba(fd, hba)) { |
ca9de185 LM |
6465 | struct mdstat_ent *vol; |
6466 | for (vol = mdstat ; vol ; vol = vol->next) { | |
089f9d79 | 6467 | if (vol->active > 0 && |
ca9de185 | 6468 | vol->metadata_version && |
9581efb1 | 6469 | is_container_member(vol, memb->devnm)) { |
ca9de185 LM |
6470 | found++; |
6471 | count++; | |
6472 | } | |
6473 | } | |
6474 | if (*devlist && (found < dpa)) { | |
503975b9 | 6475 | dv = xcalloc(1, sizeof(*dv)); |
9581efb1 N |
6476 | dv->devname = xmalloc(strlen(memb->devnm) + strlen("/dev/") + 1); |
6477 | sprintf(dv->devname, "%s%s", "/dev/", memb->devnm); | |
503975b9 N |
6478 | dv->found = found; |
6479 | dv->used = 0; | |
6480 | dv->next = *devlist; | |
6481 | *devlist = dv; | |
ca9de185 LM |
6482 | } |
6483 | } | |
6484 | if (fd >= 0) | |
6485 | close(fd); | |
6486 | } | |
6487 | } | |
6488 | free_mdstat(mdstat); | |
6489 | return count; | |
6490 | } | |
6491 | ||
6492 | #ifdef DEBUG_LOOP | |
6493 | static struct md_list* | |
6494 | get_loop_devices(void) | |
6495 | { | |
6496 | int i; | |
6497 | struct md_list *devlist = NULL; | |
594dc1b8 | 6498 | struct md_list *dv; |
ca9de185 LM |
6499 | |
6500 | for(i = 0; i < 12; i++) { | |
503975b9 N |
6501 | dv = xcalloc(1, sizeof(*dv)); |
6502 | dv->devname = xmalloc(40); | |
ca9de185 LM |
6503 | sprintf(dv->devname, "/dev/loop%d", i); |
6504 | dv->next = devlist; | |
6505 | devlist = dv; | |
6506 | } | |
6507 | return devlist; | |
6508 | } | |
6509 | #endif | |
6510 | ||
6511 | static struct md_list* | |
6512 | get_devices(const char *hba_path) | |
6513 | { | |
6514 | struct md_list *devlist = NULL; | |
594dc1b8 | 6515 | struct md_list *dv; |
ca9de185 LM |
6516 | struct dirent *ent; |
6517 | DIR *dir; | |
6518 | int err = 0; | |
6519 | ||
6520 | #if DEBUG_LOOP | |
6521 | devlist = get_loop_devices(); | |
6522 | return devlist; | |
6523 | #endif | |
6524 | /* scroll through /sys/dev/block looking for devices attached to | |
6525 | * this hba | |
6526 | */ | |
6527 | dir = opendir("/sys/dev/block"); | |
6528 | for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) { | |
6529 | int fd; | |
6530 | char buf[1024]; | |
6531 | int major, minor; | |
6532 | char *path = NULL; | |
6533 | if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2) | |
6534 | continue; | |
6535 | path = devt_to_devpath(makedev(major, minor)); | |
6536 | if (!path) | |
6537 | continue; | |
6538 | if (!path_attached_to_hba(path, hba_path)) { | |
6539 | free(path); | |
6540 | path = NULL; | |
6541 | continue; | |
6542 | } | |
6543 | free(path); | |
6544 | path = NULL; | |
6545 | fd = dev_open(ent->d_name, O_RDONLY); | |
6546 | if (fd >= 0) { | |
6547 | fd2devname(fd, buf); | |
6548 | close(fd); | |
6549 | } else { | |
e7b84f9d | 6550 | pr_err("cannot open device: %s\n", |
ca9de185 LM |
6551 | ent->d_name); |
6552 | continue; | |
6553 | } | |
6554 | ||
503975b9 N |
6555 | dv = xcalloc(1, sizeof(*dv)); |
6556 | dv->devname = xstrdup(buf); | |
ca9de185 LM |
6557 | dv->next = devlist; |
6558 | devlist = dv; | |
6559 | } | |
6560 | if (err) { | |
6561 | while(devlist) { | |
6562 | dv = devlist; | |
6563 | devlist = devlist->next; | |
6564 | free(dv->devname); | |
6565 | free(dv); | |
6566 | } | |
6567 | } | |
562aa102 | 6568 | closedir(dir); |
ca9de185 LM |
6569 | return devlist; |
6570 | } | |
6571 | ||
6572 | static int | |
6573 | count_volumes_list(struct md_list *devlist, char *homehost, | |
6574 | int verbose, int *found) | |
6575 | { | |
6576 | struct md_list *tmpdev; | |
6577 | int count = 0; | |
594dc1b8 | 6578 | struct supertype *st; |
ca9de185 LM |
6579 | |
6580 | /* first walk the list of devices to find a consistent set | |
6581 | * that match the criterea, if that is possible. | |
6582 | * We flag the ones we like with 'used'. | |
6583 | */ | |
6584 | *found = 0; | |
6585 | st = match_metadata_desc_imsm("imsm"); | |
6586 | if (st == NULL) { | |
676e87a8 | 6587 | pr_vrb("cannot allocate memory for imsm supertype\n"); |
ca9de185 LM |
6588 | return 0; |
6589 | } | |
6590 | ||
6591 | for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { | |
6592 | char *devname = tmpdev->devname; | |
0a6bff09 | 6593 | dev_t rdev; |
ca9de185 LM |
6594 | struct supertype *tst; |
6595 | int dfd; | |
6596 | if (tmpdev->used > 1) | |
6597 | continue; | |
6598 | tst = dup_super(st); | |
6599 | if (tst == NULL) { | |
676e87a8 | 6600 | pr_vrb("cannot allocate memory for imsm supertype\n"); |
ca9de185 LM |
6601 | goto err_1; |
6602 | } | |
6603 | tmpdev->container = 0; | |
6604 | dfd = dev_open(devname, O_RDONLY|O_EXCL); | |
6605 | if (dfd < 0) { | |
1ade5cc1 | 6606 | dprintf("cannot open device %s: %s\n", |
ca9de185 LM |
6607 | devname, strerror(errno)); |
6608 | tmpdev->used = 2; | |
0a6bff09 | 6609 | } else if (!fstat_is_blkdev(dfd, devname, &rdev)) { |
ca9de185 LM |
6610 | tmpdev->used = 2; |
6611 | } else if (must_be_container(dfd)) { | |
6612 | struct supertype *cst; | |
6613 | cst = super_by_fd(dfd, NULL); | |
6614 | if (cst == NULL) { | |
1ade5cc1 | 6615 | dprintf("cannot recognize container type %s\n", |
ca9de185 LM |
6616 | devname); |
6617 | tmpdev->used = 2; | |
6618 | } else if (tst->ss != st->ss) { | |
1ade5cc1 | 6619 | dprintf("non-imsm container - ignore it: %s\n", |
ca9de185 LM |
6620 | devname); |
6621 | tmpdev->used = 2; | |
6622 | } else if (!tst->ss->load_container || | |
6623 | tst->ss->load_container(tst, dfd, NULL)) | |
6624 | tmpdev->used = 2; | |
6625 | else { | |
6626 | tmpdev->container = 1; | |
6627 | } | |
6628 | if (cst) | |
6629 | cst->ss->free_super(cst); | |
6630 | } else { | |
0a6bff09 | 6631 | tmpdev->st_rdev = rdev; |
ca9de185 | 6632 | if (tst->ss->load_super(tst,dfd, NULL)) { |
1ade5cc1 | 6633 | dprintf("no RAID superblock on %s\n", |
ca9de185 LM |
6634 | devname); |
6635 | tmpdev->used = 2; | |
6636 | } else if (tst->ss->compare_super == NULL) { | |
1ade5cc1 | 6637 | dprintf("Cannot assemble %s metadata on %s\n", |
ca9de185 LM |
6638 | tst->ss->name, devname); |
6639 | tmpdev->used = 2; | |
6640 | } | |
6641 | } | |
6642 | if (dfd >= 0) | |
6643 | close(dfd); | |
6644 | if (tmpdev->used == 2 || tmpdev->used == 4) { | |
6645 | /* Ignore unrecognised devices during auto-assembly */ | |
6646 | goto loop; | |
6647 | } | |
6648 | else { | |
6649 | struct mdinfo info; | |
6650 | tst->ss->getinfo_super(tst, &info, NULL); | |
6651 | ||
6652 | if (st->minor_version == -1) | |
6653 | st->minor_version = tst->minor_version; | |
6654 | ||
6655 | if (memcmp(info.uuid, uuid_zero, | |
6656 | sizeof(int[4])) == 0) { | |
6657 | /* this is a floating spare. It cannot define | |
6658 | * an array unless there are no more arrays of | |
6659 | * this type to be found. It can be included | |
6660 | * in an array of this type though. | |
6661 | */ | |
6662 | tmpdev->used = 3; | |
6663 | goto loop; | |
6664 | } | |
6665 | ||
6666 | if (st->ss != tst->ss || | |
6667 | st->minor_version != tst->minor_version || | |
6668 | st->ss->compare_super(st, tst) != 0) { | |
6669 | /* Some mismatch. If exactly one array matches this host, | |
6670 | * we can resolve on that one. | |
6671 | * Or, if we are auto assembling, we just ignore the second | |
6672 | * for now. | |
6673 | */ | |
1ade5cc1 | 6674 | dprintf("superblock on %s doesn't match others - assembly aborted\n", |
ca9de185 LM |
6675 | devname); |
6676 | goto loop; | |
6677 | } | |
6678 | tmpdev->used = 1; | |
6679 | *found = 1; | |
6680 | dprintf("found: devname: %s\n", devname); | |
6681 | } | |
6682 | loop: | |
6683 | if (tst) | |
6684 | tst->ss->free_super(tst); | |
6685 | } | |
6686 | if (*found != 0) { | |
6687 | int err; | |
6688 | if ((err = load_super_imsm_all(st, -1, &st->sb, NULL, devlist, 0)) == 0) { | |
6689 | struct mdinfo *iter, *head = st->ss->container_content(st, NULL); | |
6690 | for (iter = head; iter; iter = iter->next) { | |
6691 | dprintf("content->text_version: %s vol\n", | |
6692 | iter->text_version); | |
6693 | if (iter->array.state & (1<<MD_SB_BLOCK_VOLUME)) { | |
6694 | /* do not assemble arrays with unsupported | |
6695 | configurations */ | |
1ade5cc1 | 6696 | dprintf("Cannot activate member %s.\n", |
ca9de185 LM |
6697 | iter->text_version); |
6698 | } else | |
6699 | count++; | |
6700 | } | |
6701 | sysfs_free(head); | |
6702 | ||
6703 | } else { | |
1ade5cc1 | 6704 | dprintf("No valid super block on device list: err: %d %p\n", |
ca9de185 LM |
6705 | err, st->sb); |
6706 | } | |
6707 | } else { | |
1ade5cc1 | 6708 | dprintf("no more devices to examine\n"); |
ca9de185 LM |
6709 | } |
6710 | ||
6711 | for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) { | |
089f9d79 | 6712 | if (tmpdev->used == 1 && tmpdev->found) { |
ca9de185 LM |
6713 | if (count) { |
6714 | if (count < tmpdev->found) | |
6715 | count = 0; | |
6716 | else | |
6717 | count -= tmpdev->found; | |
6718 | } | |
6719 | } | |
6720 | if (tmpdev->used == 1) | |
6721 | tmpdev->used = 4; | |
6722 | } | |
6723 | err_1: | |
6724 | if (st) | |
6725 | st->ss->free_super(st); | |
6726 | return count; | |
6727 | } | |
6728 | ||
d3c11416 AO |
6729 | static int __count_volumes(char *hba_path, int dpa, int verbose, |
6730 | int cmp_hba_path) | |
ca9de185 | 6731 | { |
72a45777 | 6732 | struct sys_dev *idev, *intel_devices = find_intel_devices(); |
ca9de185 | 6733 | int count = 0; |
72a45777 PB |
6734 | const struct orom_entry *entry; |
6735 | struct devid_list *dv, *devid_list; | |
ca9de185 | 6736 | |
d3c11416 | 6737 | if (!hba_path) |
ca9de185 LM |
6738 | return 0; |
6739 | ||
72a45777 | 6740 | for (idev = intel_devices; idev; idev = idev->next) { |
d3c11416 AO |
6741 | if (strstr(idev->path, hba_path)) |
6742 | break; | |
72a45777 PB |
6743 | } |
6744 | ||
6745 | if (!idev || !idev->dev_id) | |
ca9de185 | 6746 | return 0; |
72a45777 PB |
6747 | |
6748 | entry = get_orom_entry_by_device_id(idev->dev_id); | |
6749 | ||
6750 | if (!entry || !entry->devid_list) | |
6751 | return 0; | |
6752 | ||
6753 | devid_list = entry->devid_list; | |
6754 | for (dv = devid_list; dv; dv = dv->next) { | |
594dc1b8 | 6755 | struct md_list *devlist; |
d3c11416 AO |
6756 | struct sys_dev *device = NULL; |
6757 | char *hpath; | |
72a45777 PB |
6758 | int found = 0; |
6759 | ||
d3c11416 AO |
6760 | if (cmp_hba_path) |
6761 | device = device_by_id_and_path(dv->devid, hba_path); | |
6762 | else | |
6763 | device = device_by_id(dv->devid); | |
6764 | ||
72a45777 | 6765 | if (device) |
d3c11416 | 6766 | hpath = device->path; |
72a45777 PB |
6767 | else |
6768 | return 0; | |
6769 | ||
d3c11416 | 6770 | devlist = get_devices(hpath); |
72a45777 PB |
6771 | /* if no intel devices return zero volumes */ |
6772 | if (devlist == NULL) | |
6773 | return 0; | |
6774 | ||
d3c11416 AO |
6775 | count += active_arrays_by_format("imsm", hpath, &devlist, dpa, |
6776 | verbose); | |
6777 | dprintf("path: %s active arrays: %d\n", hpath, count); | |
72a45777 PB |
6778 | if (devlist == NULL) |
6779 | return 0; | |
6780 | do { | |
6781 | found = 0; | |
6782 | count += count_volumes_list(devlist, | |
6783 | NULL, | |
6784 | verbose, | |
6785 | &found); | |
6786 | dprintf("found %d count: %d\n", found, count); | |
6787 | } while (found); | |
6788 | ||
d3c11416 | 6789 | dprintf("path: %s total number of volumes: %d\n", hpath, count); |
72a45777 PB |
6790 | |
6791 | while (devlist) { | |
6792 | struct md_list *dv = devlist; | |
6793 | devlist = devlist->next; | |
6794 | free(dv->devname); | |
6795 | free(dv); | |
6796 | } | |
ca9de185 LM |
6797 | } |
6798 | return count; | |
6799 | } | |
6800 | ||
d3c11416 AO |
6801 | static int count_volumes(struct intel_hba *hba, int dpa, int verbose) |
6802 | { | |
6803 | if (!hba) | |
6804 | return 0; | |
6805 | if (hba->type == SYS_DEV_VMD) { | |
6806 | struct sys_dev *dev; | |
6807 | int count = 0; | |
6808 | ||
6809 | for (dev = find_intel_devices(); dev; dev = dev->next) { | |
6810 | if (dev->type == SYS_DEV_VMD) | |
6811 | count += __count_volumes(dev->path, dpa, | |
6812 | verbose, 1); | |
6813 | } | |
6814 | return count; | |
6815 | } | |
6816 | return __count_volumes(hba->path, dpa, verbose, 0); | |
6817 | } | |
6818 | ||
cd9d1ac7 DW |
6819 | static int imsm_default_chunk(const struct imsm_orom *orom) |
6820 | { | |
6821 | /* up to 512 if the plaform supports it, otherwise the platform max. | |
6822 | * 128 if no platform detected | |
6823 | */ | |
6824 | int fs = max(7, orom ? fls(orom->sss) : 0); | |
6825 | ||
6826 | return min(512, (1 << fs)); | |
6827 | } | |
73408129 | 6828 | |
6592ce37 DW |
6829 | static int |
6830 | validate_geometry_imsm_orom(struct intel_super *super, int level, int layout, | |
2cc699af | 6831 | int raiddisks, int *chunk, unsigned long long size, int verbose) |
6592ce37 | 6832 | { |
660260d0 DW |
6833 | /* check/set platform and metadata limits/defaults */ |
6834 | if (super->orom && raiddisks > super->orom->dpa) { | |
676e87a8 | 6835 | pr_vrb("platform supports a maximum of %d disks per array\n", |
660260d0 | 6836 | super->orom->dpa); |
73408129 LM |
6837 | return 0; |
6838 | } | |
6839 | ||
5d500228 | 6840 | /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */ |
660260d0 | 6841 | if (!is_raid_level_supported(super->orom, level, raiddisks)) { |
676e87a8 | 6842 | pr_vrb("platform does not support raid%d with %d disk%s\n", |
6592ce37 DW |
6843 | level, raiddisks, raiddisks > 1 ? "s" : ""); |
6844 | return 0; | |
6845 | } | |
cd9d1ac7 | 6846 | |
7ccc4cc4 | 6847 | if (*chunk == 0 || *chunk == UnSet) |
cd9d1ac7 DW |
6848 | *chunk = imsm_default_chunk(super->orom); |
6849 | ||
7ccc4cc4 | 6850 | if (super->orom && !imsm_orom_has_chunk(super->orom, *chunk)) { |
676e87a8 | 6851 | pr_vrb("platform does not support a chunk size of: %d\n", *chunk); |
cd9d1ac7 | 6852 | return 0; |
6592ce37 | 6853 | } |
cd9d1ac7 | 6854 | |
6592ce37 DW |
6855 | if (layout != imsm_level_to_layout(level)) { |
6856 | if (level == 5) | |
676e87a8 | 6857 | pr_vrb("imsm raid 5 only supports the left-asymmetric layout\n"); |
6592ce37 | 6858 | else if (level == 10) |
676e87a8 | 6859 | pr_vrb("imsm raid 10 only supports the n2 layout\n"); |
6592ce37 | 6860 | else |
676e87a8 | 6861 | pr_vrb("imsm unknown layout %#x for this raid level %d\n", |
6592ce37 DW |
6862 | layout, level); |
6863 | return 0; | |
6864 | } | |
2cc699af | 6865 | |
7ccc4cc4 | 6866 | if (super->orom && (super->orom->attr & IMSM_OROM_ATTR_2TB) == 0 && |
2cc699af | 6867 | (calc_array_size(level, raiddisks, layout, *chunk, size) >> 32) > 0) { |
676e87a8 | 6868 | pr_vrb("platform does not support a volume size over 2TB\n"); |
2cc699af CA |
6869 | return 0; |
6870 | } | |
614902f6 | 6871 | |
6592ce37 DW |
6872 | return 1; |
6873 | } | |
6874 | ||
1011e834 | 6875 | /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd |
c2c087e6 DW |
6876 | * FIX ME add ahci details |
6877 | */ | |
8b353278 | 6878 | static int validate_geometry_imsm_volume(struct supertype *st, int level, |
c21e737b | 6879 | int layout, int raiddisks, int *chunk, |
af4348dd N |
6880 | unsigned long long size, |
6881 | unsigned long long data_offset, | |
6882 | char *dev, | |
2c514b71 NB |
6883 | unsigned long long *freesize, |
6884 | int verbose) | |
cdddbdbc | 6885 | { |
9e04ac1c | 6886 | dev_t rdev; |
c2c087e6 | 6887 | struct intel_super *super = st->sb; |
b2916f25 | 6888 | struct imsm_super *mpb; |
c2c087e6 DW |
6889 | struct dl *dl; |
6890 | unsigned long long pos = 0; | |
6891 | unsigned long long maxsize; | |
6892 | struct extent *e; | |
6893 | int i; | |
cdddbdbc | 6894 | |
88c32bb1 DW |
6895 | /* We must have the container info already read in. */ |
6896 | if (!super) | |
c2c087e6 DW |
6897 | return 0; |
6898 | ||
b2916f25 JS |
6899 | mpb = super->anchor; |
6900 | ||
2cc699af | 6901 | if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, size, verbose)) { |
7a862a02 | 6902 | pr_err("RAID gemetry validation failed. Cannot proceed with the action(s).\n"); |
c2c087e6 | 6903 | return 0; |
d54559f0 | 6904 | } |
c2c087e6 DW |
6905 | if (!dev) { |
6906 | /* General test: make sure there is space for | |
2da8544a DW |
6907 | * 'raiddisks' device extents of size 'size' at a given |
6908 | * offset | |
c2c087e6 | 6909 | */ |
e46273eb | 6910 | unsigned long long minsize = size; |
b7528a20 | 6911 | unsigned long long start_offset = MaxSector; |
c2c087e6 DW |
6912 | int dcnt = 0; |
6913 | if (minsize == 0) | |
6914 | minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS; | |
6915 | for (dl = super->disks; dl ; dl = dl->next) { | |
6916 | int found = 0; | |
6917 | ||
bf5a934a | 6918 | pos = 0; |
c2c087e6 DW |
6919 | i = 0; |
6920 | e = get_extents(super, dl); | |
6921 | if (!e) continue; | |
6922 | do { | |
6923 | unsigned long long esize; | |
6924 | esize = e[i].start - pos; | |
6925 | if (esize >= minsize) | |
6926 | found = 1; | |
b7528a20 | 6927 | if (found && start_offset == MaxSector) { |
2da8544a DW |
6928 | start_offset = pos; |
6929 | break; | |
6930 | } else if (found && pos != start_offset) { | |
6931 | found = 0; | |
6932 | break; | |
6933 | } | |
c2c087e6 DW |
6934 | pos = e[i].start + e[i].size; |
6935 | i++; | |
6936 | } while (e[i-1].size); | |
6937 | if (found) | |
6938 | dcnt++; | |
6939 | free(e); | |
6940 | } | |
6941 | if (dcnt < raiddisks) { | |
2c514b71 | 6942 | if (verbose) |
7a862a02 | 6943 | pr_err("imsm: Not enough devices with space for this array (%d < %d)\n", |
2c514b71 | 6944 | dcnt, raiddisks); |
c2c087e6 DW |
6945 | return 0; |
6946 | } | |
6947 | return 1; | |
6948 | } | |
0dcecb2e | 6949 | |
c2c087e6 | 6950 | /* This device must be a member of the set */ |
9e04ac1c | 6951 | if (!stat_is_blkdev(dev, &rdev)) |
c2c087e6 DW |
6952 | return 0; |
6953 | for (dl = super->disks ; dl ; dl = dl->next) { | |
9e04ac1c ZL |
6954 | if (dl->major == (int)major(rdev) && |
6955 | dl->minor == (int)minor(rdev)) | |
c2c087e6 DW |
6956 | break; |
6957 | } | |
6958 | if (!dl) { | |
2c514b71 | 6959 | if (verbose) |
7a862a02 | 6960 | pr_err("%s is not in the same imsm set\n", dev); |
c2c087e6 | 6961 | return 0; |
a20d2ba5 DW |
6962 | } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) { |
6963 | /* If a volume is present then the current creation attempt | |
6964 | * cannot incorporate new spares because the orom may not | |
6965 | * understand this configuration (all member disks must be | |
6966 | * members of each array in the container). | |
6967 | */ | |
7a862a02 N |
6968 | pr_err("%s is a spare and a volume is already defined for this container\n", dev); |
6969 | pr_err("The option-rom requires all member disks to be a member of all volumes\n"); | |
a20d2ba5 | 6970 | return 0; |
5fe62b94 WD |
6971 | } else if (super->orom && mpb->num_raid_devs > 0 && |
6972 | mpb->num_disks != raiddisks) { | |
7a862a02 | 6973 | pr_err("The option-rom requires all member disks to be a member of all volumes\n"); |
5fe62b94 | 6974 | return 0; |
c2c087e6 | 6975 | } |
0dcecb2e DW |
6976 | |
6977 | /* retrieve the largest free space block */ | |
c2c087e6 DW |
6978 | e = get_extents(super, dl); |
6979 | maxsize = 0; | |
6980 | i = 0; | |
0dcecb2e DW |
6981 | if (e) { |
6982 | do { | |
6983 | unsigned long long esize; | |
6984 | ||
6985 | esize = e[i].start - pos; | |
6986 | if (esize >= maxsize) | |
6987 | maxsize = esize; | |
6988 | pos = e[i].start + e[i].size; | |
6989 | i++; | |
6990 | } while (e[i-1].size); | |
6991 | dl->e = e; | |
6992 | dl->extent_cnt = i; | |
6993 | } else { | |
6994 | if (verbose) | |
e7b84f9d | 6995 | pr_err("unable to determine free space for: %s\n", |
0dcecb2e DW |
6996 | dev); |
6997 | return 0; | |
6998 | } | |
6999 | if (maxsize < size) { | |
7000 | if (verbose) | |
e7b84f9d | 7001 | pr_err("%s not enough space (%llu < %llu)\n", |
0dcecb2e DW |
7002 | dev, maxsize, size); |
7003 | return 0; | |
7004 | } | |
7005 | ||
7006 | /* count total number of extents for merge */ | |
7007 | i = 0; | |
7008 | for (dl = super->disks; dl; dl = dl->next) | |
7009 | if (dl->e) | |
7010 | i += dl->extent_cnt; | |
7011 | ||
7012 | maxsize = merge_extents(super, i); | |
3baa56ab LO |
7013 | |
7014 | if (!check_env("IMSM_NO_PLATFORM") && | |
7015 | mpb->num_raid_devs > 0 && size && size != maxsize) { | |
7a862a02 | 7016 | pr_err("attempting to create a second volume with size less then remaining space. Aborting...\n"); |
3baa56ab LO |
7017 | return 0; |
7018 | } | |
7019 | ||
a7dd165b | 7020 | if (maxsize < size || maxsize == 0) { |
b3071342 LD |
7021 | if (verbose) { |
7022 | if (maxsize == 0) | |
7a862a02 | 7023 | pr_err("no free space left on device. Aborting...\n"); |
b3071342 | 7024 | else |
7a862a02 | 7025 | pr_err("not enough space to create volume of given size (%llu < %llu). Aborting...\n", |
b3071342 LD |
7026 | maxsize, size); |
7027 | } | |
0dcecb2e | 7028 | return 0; |
0dcecb2e DW |
7029 | } |
7030 | ||
c2c087e6 DW |
7031 | *freesize = maxsize; |
7032 | ||
ca9de185 | 7033 | if (super->orom) { |
72a45777 | 7034 | int count = count_volumes(super->hba, |
ca9de185 LM |
7035 | super->orom->dpa, verbose); |
7036 | if (super->orom->vphba <= count) { | |
676e87a8 | 7037 | pr_vrb("platform does not support more than %d raid volumes.\n", |
ca9de185 LM |
7038 | super->orom->vphba); |
7039 | return 0; | |
7040 | } | |
7041 | } | |
c2c087e6 | 7042 | return 1; |
cdddbdbc DW |
7043 | } |
7044 | ||
13bcac90 | 7045 | static int imsm_get_free_size(struct supertype *st, int raiddisks, |
efb30e7f DW |
7046 | unsigned long long size, int chunk, |
7047 | unsigned long long *freesize) | |
7048 | { | |
7049 | struct intel_super *super = st->sb; | |
7050 | struct imsm_super *mpb = super->anchor; | |
7051 | struct dl *dl; | |
7052 | int i; | |
7053 | int extent_cnt; | |
7054 | struct extent *e; | |
7055 | unsigned long long maxsize; | |
7056 | unsigned long long minsize; | |
7057 | int cnt; | |
7058 | int used; | |
7059 | ||
7060 | /* find the largest common start free region of the possible disks */ | |
7061 | used = 0; | |
7062 | extent_cnt = 0; | |
7063 | cnt = 0; | |
7064 | for (dl = super->disks; dl; dl = dl->next) { | |
7065 | dl->raiddisk = -1; | |
7066 | ||
7067 | if (dl->index >= 0) | |
7068 | used++; | |
7069 | ||
7070 | /* don't activate new spares if we are orom constrained | |
7071 | * and there is already a volume active in the container | |
7072 | */ | |
7073 | if (super->orom && dl->index < 0 && mpb->num_raid_devs) | |
7074 | continue; | |
7075 | ||
7076 | e = get_extents(super, dl); | |
7077 | if (!e) | |
7078 | continue; | |
7079 | for (i = 1; e[i-1].size; i++) | |
7080 | ; | |
7081 | dl->e = e; | |
7082 | dl->extent_cnt = i; | |
7083 | extent_cnt += i; | |
7084 | cnt++; | |
7085 | } | |
7086 | ||
7087 | maxsize = merge_extents(super, extent_cnt); | |
7088 | minsize = size; | |
7089 | if (size == 0) | |
612e59d8 CA |
7090 | /* chunk is in K */ |
7091 | minsize = chunk * 2; | |
efb30e7f DW |
7092 | |
7093 | if (cnt < raiddisks || | |
7094 | (super->orom && used && used != raiddisks) || | |
a7dd165b DW |
7095 | maxsize < minsize || |
7096 | maxsize == 0) { | |
e7b84f9d | 7097 | pr_err("not enough devices with space to create array.\n"); |
efb30e7f DW |
7098 | return 0; /* No enough free spaces large enough */ |
7099 | } | |
7100 | ||
7101 | if (size == 0) { | |
7102 | size = maxsize; | |
7103 | if (chunk) { | |
612e59d8 CA |
7104 | size /= 2 * chunk; |
7105 | size *= 2 * chunk; | |
efb30e7f | 7106 | } |
f878b242 LM |
7107 | maxsize = size; |
7108 | } | |
7109 | if (!check_env("IMSM_NO_PLATFORM") && | |
7110 | mpb->num_raid_devs > 0 && size && size != maxsize) { | |
7a862a02 | 7111 | pr_err("attempting to create a second volume with size less then remaining space. Aborting...\n"); |
f878b242 | 7112 | return 0; |
efb30e7f | 7113 | } |
efb30e7f DW |
7114 | cnt = 0; |
7115 | for (dl = super->disks; dl; dl = dl->next) | |
7116 | if (dl->e) | |
7117 | dl->raiddisk = cnt++; | |
7118 | ||
7119 | *freesize = size; | |
7120 | ||
13bcac90 AK |
7121 | dprintf("imsm: imsm_get_free_size() returns : %llu\n", size); |
7122 | ||
efb30e7f DW |
7123 | return 1; |
7124 | } | |
7125 | ||
13bcac90 AK |
7126 | static int reserve_space(struct supertype *st, int raiddisks, |
7127 | unsigned long long size, int chunk, | |
7128 | unsigned long long *freesize) | |
7129 | { | |
7130 | struct intel_super *super = st->sb; | |
7131 | struct dl *dl; | |
7132 | int cnt; | |
7133 | int rv = 0; | |
7134 | ||
7135 | rv = imsm_get_free_size(st, raiddisks, size, chunk, freesize); | |
7136 | if (rv) { | |
7137 | cnt = 0; | |
7138 | for (dl = super->disks; dl; dl = dl->next) | |
7139 | if (dl->e) | |
7140 | dl->raiddisk = cnt++; | |
7141 | rv = 1; | |
7142 | } | |
7143 | ||
7144 | return rv; | |
7145 | } | |
7146 | ||
bf5a934a | 7147 | static int validate_geometry_imsm(struct supertype *st, int level, int layout, |
c21e737b | 7148 | int raiddisks, int *chunk, unsigned long long size, |
af4348dd | 7149 | unsigned long long data_offset, |
bf5a934a | 7150 | char *dev, unsigned long long *freesize, |
5308f117 | 7151 | int consistency_policy, int verbose) |
bf5a934a DW |
7152 | { |
7153 | int fd, cfd; | |
7154 | struct mdinfo *sra; | |
20cbe8d2 | 7155 | int is_member = 0; |
bf5a934a | 7156 | |
d54559f0 LM |
7157 | /* load capability |
7158 | * if given unused devices create a container | |
bf5a934a DW |
7159 | * if given given devices in a container create a member volume |
7160 | */ | |
7161 | if (level == LEVEL_CONTAINER) { | |
7162 | /* Must be a fresh device to add to a container */ | |
7163 | return validate_geometry_imsm_container(st, level, layout, | |
c21e737b | 7164 | raiddisks, |
7ccc4cc4 | 7165 | *chunk, |
af4348dd | 7166 | size, data_offset, |
bf5a934a DW |
7167 | dev, freesize, |
7168 | verbose); | |
7169 | } | |
9587c373 | 7170 | |
8592f29d | 7171 | if (!dev) { |
e91a3bad | 7172 | if (st->sb) { |
ca9de185 | 7173 | struct intel_super *super = st->sb; |
e91a3bad | 7174 | if (!validate_geometry_imsm_orom(st->sb, level, layout, |
2cc699af | 7175 | raiddisks, chunk, size, |
e91a3bad LM |
7176 | verbose)) |
7177 | return 0; | |
efb30e7f DW |
7178 | /* we are being asked to automatically layout a |
7179 | * new volume based on the current contents of | |
7180 | * the container. If the the parameters can be | |
7181 | * satisfied reserve_space will record the disks, | |
7182 | * start offset, and size of the volume to be | |
7183 | * created. add_to_super and getinfo_super | |
7184 | * detect when autolayout is in progress. | |
7185 | */ | |
ca9de185 LM |
7186 | /* assuming that freesize is always given when array is |
7187 | created */ | |
7188 | if (super->orom && freesize) { | |
7189 | int count; | |
72a45777 | 7190 | count = count_volumes(super->hba, |
ca9de185 LM |
7191 | super->orom->dpa, verbose); |
7192 | if (super->orom->vphba <= count) { | |
676e87a8 | 7193 | pr_vrb("platform does not support more than %d raid volumes.\n", |
ca9de185 LM |
7194 | super->orom->vphba); |
7195 | return 0; | |
7196 | } | |
7197 | } | |
e91a3bad LM |
7198 | if (freesize) |
7199 | return reserve_space(st, raiddisks, size, | |
7ccc4cc4 | 7200 | *chunk, freesize); |
8592f29d N |
7201 | } |
7202 | return 1; | |
7203 | } | |
bf5a934a DW |
7204 | if (st->sb) { |
7205 | /* creating in a given container */ | |
7206 | return validate_geometry_imsm_volume(st, level, layout, | |
7207 | raiddisks, chunk, size, | |
af4348dd | 7208 | data_offset, |
bf5a934a DW |
7209 | dev, freesize, verbose); |
7210 | } | |
7211 | ||
bf5a934a DW |
7212 | /* This device needs to be a device in an 'imsm' container */ |
7213 | fd = open(dev, O_RDONLY|O_EXCL, 0); | |
7214 | if (fd >= 0) { | |
7215 | if (verbose) | |
e7b84f9d N |
7216 | pr_err("Cannot create this array on device %s\n", |
7217 | dev); | |
bf5a934a DW |
7218 | close(fd); |
7219 | return 0; | |
7220 | } | |
7221 | if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) { | |
7222 | if (verbose) | |
e7b84f9d | 7223 | pr_err("Cannot open %s: %s\n", |
bf5a934a DW |
7224 | dev, strerror(errno)); |
7225 | return 0; | |
7226 | } | |
7227 | /* Well, it is in use by someone, maybe an 'imsm' container. */ | |
7228 | cfd = open_container(fd); | |
20cbe8d2 | 7229 | close(fd); |
bf5a934a | 7230 | if (cfd < 0) { |
bf5a934a | 7231 | if (verbose) |
e7b84f9d | 7232 | pr_err("Cannot use %s: It is busy\n", |
bf5a934a DW |
7233 | dev); |
7234 | return 0; | |
7235 | } | |
4dd2df09 | 7236 | sra = sysfs_read(cfd, NULL, GET_VERSION); |
bf5a934a | 7237 | if (sra && sra->array.major_version == -1 && |
20cbe8d2 AW |
7238 | strcmp(sra->text_version, "imsm") == 0) |
7239 | is_member = 1; | |
7240 | sysfs_free(sra); | |
7241 | if (is_member) { | |
bf5a934a DW |
7242 | /* This is a member of a imsm container. Load the container |
7243 | * and try to create a volume | |
7244 | */ | |
7245 | struct intel_super *super; | |
7246 | ||
ec50f7b6 | 7247 | if (load_super_imsm_all(st, cfd, (void **) &super, NULL, NULL, 1) == 0) { |
bf5a934a | 7248 | st->sb = super; |
4dd2df09 | 7249 | strcpy(st->container_devnm, fd2devnm(cfd)); |
bf5a934a DW |
7250 | close(cfd); |
7251 | return validate_geometry_imsm_volume(st, level, layout, | |
7252 | raiddisks, chunk, | |
af4348dd | 7253 | size, data_offset, dev, |
ecbd9e81 N |
7254 | freesize, 1) |
7255 | ? 1 : -1; | |
bf5a934a | 7256 | } |
20cbe8d2 | 7257 | } |
bf5a934a | 7258 | |
20cbe8d2 | 7259 | if (verbose) |
e7b84f9d | 7260 | pr_err("failed container membership check\n"); |
20cbe8d2 AW |
7261 | |
7262 | close(cfd); | |
7263 | return 0; | |
bf5a934a | 7264 | } |
0bd16cf2 | 7265 | |
30f58b22 | 7266 | static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk) |
0bd16cf2 DJ |
7267 | { |
7268 | struct intel_super *super = st->sb; | |
7269 | ||
30f58b22 DW |
7270 | if (level && *level == UnSet) |
7271 | *level = LEVEL_CONTAINER; | |
7272 | ||
7273 | if (level && layout && *layout == UnSet) | |
7274 | *layout = imsm_level_to_layout(*level); | |
0bd16cf2 | 7275 | |
cd9d1ac7 DW |
7276 | if (chunk && (*chunk == UnSet || *chunk == 0)) |
7277 | *chunk = imsm_default_chunk(super->orom); | |
0bd16cf2 DJ |
7278 | } |
7279 | ||
33414a01 DW |
7280 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev); |
7281 | ||
7282 | static int kill_subarray_imsm(struct supertype *st) | |
7283 | { | |
7284 | /* remove the subarray currently referenced by ->current_vol */ | |
7285 | __u8 i; | |
7286 | struct intel_dev **dp; | |
7287 | struct intel_super *super = st->sb; | |
7288 | __u8 current_vol = super->current_vol; | |
7289 | struct imsm_super *mpb = super->anchor; | |
7290 | ||
7291 | if (super->current_vol < 0) | |
7292 | return 2; | |
7293 | super->current_vol = -1; /* invalidate subarray cursor */ | |
7294 | ||
7295 | /* block deletions that would change the uuid of active subarrays | |
7296 | * | |
7297 | * FIXME when immutable ids are available, but note that we'll | |
7298 | * also need to fixup the invalidated/active subarray indexes in | |
7299 | * mdstat | |
7300 | */ | |
7301 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
7302 | char subarray[4]; | |
7303 | ||
7304 | if (i < current_vol) | |
7305 | continue; | |
7306 | sprintf(subarray, "%u", i); | |
4dd2df09 | 7307 | if (is_subarray_active(subarray, st->devnm)) { |
e7b84f9d N |
7308 | pr_err("deleting subarray-%d would change the UUID of active subarray-%d, aborting\n", |
7309 | current_vol, i); | |
33414a01 DW |
7310 | |
7311 | return 2; | |
7312 | } | |
7313 | } | |
7314 | ||
7315 | if (st->update_tail) { | |
503975b9 | 7316 | struct imsm_update_kill_array *u = xmalloc(sizeof(*u)); |
33414a01 | 7317 | |
33414a01 DW |
7318 | u->type = update_kill_array; |
7319 | u->dev_idx = current_vol; | |
7320 | append_metadata_update(st, u, sizeof(*u)); | |
7321 | ||
7322 | return 0; | |
7323 | } | |
7324 | ||
7325 | for (dp = &super->devlist; *dp;) | |
7326 | if ((*dp)->index == current_vol) { | |
7327 | *dp = (*dp)->next; | |
7328 | } else { | |
7329 | handle_missing(super, (*dp)->dev); | |
7330 | if ((*dp)->index > current_vol) | |
7331 | (*dp)->index--; | |
7332 | dp = &(*dp)->next; | |
7333 | } | |
7334 | ||
7335 | /* no more raid devices, all active components are now spares, | |
7336 | * but of course failed are still failed | |
7337 | */ | |
7338 | if (--mpb->num_raid_devs == 0) { | |
7339 | struct dl *d; | |
7340 | ||
7341 | for (d = super->disks; d; d = d->next) | |
a8619d23 AK |
7342 | if (d->index > -2) |
7343 | mark_spare(d); | |
33414a01 DW |
7344 | } |
7345 | ||
7346 | super->updates_pending++; | |
7347 | ||
7348 | return 0; | |
7349 | } | |
aa534678 | 7350 | |
a951a4f7 | 7351 | static int update_subarray_imsm(struct supertype *st, char *subarray, |
fa56eddb | 7352 | char *update, struct mddev_ident *ident) |
aa534678 DW |
7353 | { |
7354 | /* update the subarray currently referenced by ->current_vol */ | |
7355 | struct intel_super *super = st->sb; | |
7356 | struct imsm_super *mpb = super->anchor; | |
7357 | ||
aa534678 DW |
7358 | if (strcmp(update, "name") == 0) { |
7359 | char *name = ident->name; | |
a951a4f7 N |
7360 | char *ep; |
7361 | int vol; | |
aa534678 | 7362 | |
4dd2df09 | 7363 | if (is_subarray_active(subarray, st->devnm)) { |
e7b84f9d | 7364 | pr_err("Unable to update name of active subarray\n"); |
aa534678 DW |
7365 | return 2; |
7366 | } | |
7367 | ||
7368 | if (!check_name(super, name, 0)) | |
7369 | return 2; | |
7370 | ||
a951a4f7 N |
7371 | vol = strtoul(subarray, &ep, 10); |
7372 | if (*ep != '\0' || vol >= super->anchor->num_raid_devs) | |
7373 | return 2; | |
7374 | ||
aa534678 | 7375 | if (st->update_tail) { |
503975b9 | 7376 | struct imsm_update_rename_array *u = xmalloc(sizeof(*u)); |
aa534678 | 7377 | |
aa534678 | 7378 | u->type = update_rename_array; |
a951a4f7 | 7379 | u->dev_idx = vol; |
618f4e6d XN |
7380 | strncpy((char *) u->name, name, MAX_RAID_SERIAL_LEN); |
7381 | u->name[MAX_RAID_SERIAL_LEN-1] = '\0'; | |
aa534678 DW |
7382 | append_metadata_update(st, u, sizeof(*u)); |
7383 | } else { | |
7384 | struct imsm_dev *dev; | |
7385 | int i; | |
7386 | ||
a951a4f7 | 7387 | dev = get_imsm_dev(super, vol); |
618f4e6d XN |
7388 | strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN); |
7389 | dev->volume[MAX_RAID_SERIAL_LEN-1] = '\0'; | |
aa534678 DW |
7390 | for (i = 0; i < mpb->num_raid_devs; i++) { |
7391 | dev = get_imsm_dev(super, i); | |
7392 | handle_missing(super, dev); | |
7393 | } | |
7394 | super->updates_pending++; | |
7395 | } | |
e6e9dd3f AP |
7396 | } else if (strcmp(update, "ppl") == 0 || |
7397 | strcmp(update, "no-ppl") == 0) { | |
7398 | int new_policy; | |
7399 | char *ep; | |
7400 | int vol = strtoul(subarray, &ep, 10); | |
7401 | ||
7402 | if (*ep != '\0' || vol >= super->anchor->num_raid_devs) | |
7403 | return 2; | |
7404 | ||
7405 | if (strcmp(update, "ppl") == 0) | |
7406 | new_policy = RWH_DISTRIBUTED; | |
7407 | else | |
7408 | new_policy = RWH_OFF; | |
7409 | ||
7410 | if (st->update_tail) { | |
7411 | struct imsm_update_rwh_policy *u = xmalloc(sizeof(*u)); | |
7412 | ||
7413 | u->type = update_rwh_policy; | |
7414 | u->dev_idx = vol; | |
7415 | u->new_policy = new_policy; | |
7416 | append_metadata_update(st, u, sizeof(*u)); | |
7417 | } else { | |
7418 | struct imsm_dev *dev; | |
7419 | ||
7420 | dev = get_imsm_dev(super, vol); | |
7421 | dev->rwh_policy = new_policy; | |
7422 | super->updates_pending++; | |
7423 | } | |
aa534678 DW |
7424 | } else |
7425 | return 2; | |
7426 | ||
7427 | return 0; | |
7428 | } | |
bf5a934a | 7429 | |
28bce06f AK |
7430 | static int is_gen_migration(struct imsm_dev *dev) |
7431 | { | |
7534230b AK |
7432 | if (dev == NULL) |
7433 | return 0; | |
7434 | ||
28bce06f AK |
7435 | if (!dev->vol.migr_state) |
7436 | return 0; | |
7437 | ||
7438 | if (migr_type(dev) == MIGR_GEN_MIGR) | |
7439 | return 1; | |
7440 | ||
7441 | return 0; | |
7442 | } | |
7443 | ||
1e5c6983 DW |
7444 | static int is_rebuilding(struct imsm_dev *dev) |
7445 | { | |
7446 | struct imsm_map *migr_map; | |
7447 | ||
7448 | if (!dev->vol.migr_state) | |
7449 | return 0; | |
7450 | ||
7451 | if (migr_type(dev) != MIGR_REBUILD) | |
7452 | return 0; | |
7453 | ||
238c0a71 | 7454 | migr_map = get_imsm_map(dev, MAP_1); |
1e5c6983 DW |
7455 | |
7456 | if (migr_map->map_state == IMSM_T_STATE_DEGRADED) | |
7457 | return 1; | |
7458 | else | |
7459 | return 0; | |
7460 | } | |
7461 | ||
6ce1fbf1 AK |
7462 | static int is_initializing(struct imsm_dev *dev) |
7463 | { | |
7464 | struct imsm_map *migr_map; | |
7465 | ||
7466 | if (!dev->vol.migr_state) | |
7467 | return 0; | |
7468 | ||
7469 | if (migr_type(dev) != MIGR_INIT) | |
7470 | return 0; | |
7471 | ||
238c0a71 | 7472 | migr_map = get_imsm_map(dev, MAP_1); |
6ce1fbf1 AK |
7473 | |
7474 | if (migr_map->map_state == IMSM_T_STATE_UNINITIALIZED) | |
7475 | return 1; | |
7476 | ||
7477 | return 0; | |
6ce1fbf1 AK |
7478 | } |
7479 | ||
c47b0ff6 AK |
7480 | static void update_recovery_start(struct intel_super *super, |
7481 | struct imsm_dev *dev, | |
7482 | struct mdinfo *array) | |
1e5c6983 DW |
7483 | { |
7484 | struct mdinfo *rebuild = NULL; | |
7485 | struct mdinfo *d; | |
7486 | __u32 units; | |
7487 | ||
7488 | if (!is_rebuilding(dev)) | |
7489 | return; | |
7490 | ||
7491 | /* Find the rebuild target, but punt on the dual rebuild case */ | |
7492 | for (d = array->devs; d; d = d->next) | |
7493 | if (d->recovery_start == 0) { | |
7494 | if (rebuild) | |
7495 | return; | |
7496 | rebuild = d; | |
7497 | } | |
7498 | ||
4363fd80 DW |
7499 | if (!rebuild) { |
7500 | /* (?) none of the disks are marked with | |
7501 | * IMSM_ORD_REBUILD, so assume they are missing and the | |
7502 | * disk_ord_tbl was not correctly updated | |
7503 | */ | |
1ade5cc1 | 7504 | dprintf("failed to locate out-of-sync disk\n"); |
4363fd80 DW |
7505 | return; |
7506 | } | |
7507 | ||
1e5c6983 | 7508 | units = __le32_to_cpu(dev->vol.curr_migr_unit); |
c47b0ff6 | 7509 | rebuild->recovery_start = units * blocks_per_migr_unit(super, dev); |
1e5c6983 DW |
7510 | } |
7511 | ||
276d77db | 7512 | static int recover_backup_imsm(struct supertype *st, struct mdinfo *info); |
1e5c6983 | 7513 | |
00bbdbda | 7514 | static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray) |
cdddbdbc | 7515 | { |
4f5bc454 DW |
7516 | /* Given a container loaded by load_super_imsm_all, |
7517 | * extract information about all the arrays into | |
7518 | * an mdinfo tree. | |
00bbdbda | 7519 | * If 'subarray' is given, just extract info about that array. |
4f5bc454 DW |
7520 | * |
7521 | * For each imsm_dev create an mdinfo, fill it in, | |
7522 | * then look for matching devices in super->disks | |
7523 | * and create appropriate device mdinfo. | |
7524 | */ | |
7525 | struct intel_super *super = st->sb; | |
949c47a0 | 7526 | struct imsm_super *mpb = super->anchor; |
4f5bc454 | 7527 | struct mdinfo *rest = NULL; |
00bbdbda | 7528 | unsigned int i; |
81219e70 | 7529 | int sb_errors = 0; |
abef11a3 AK |
7530 | struct dl *d; |
7531 | int spare_disks = 0; | |
cdddbdbc | 7532 | |
19482bcc AK |
7533 | /* do not assemble arrays when not all attributes are supported */ |
7534 | if (imsm_check_attributes(mpb->attributes) == 0) { | |
81219e70 | 7535 | sb_errors = 1; |
7a862a02 | 7536 | pr_err("Unsupported attributes in IMSM metadata.Arrays activation is blocked.\n"); |
19482bcc AK |
7537 | } |
7538 | ||
abef11a3 AK |
7539 | /* count spare devices, not used in maps |
7540 | */ | |
7541 | for (d = super->disks; d; d = d->next) | |
7542 | if (d->index == -1) | |
7543 | spare_disks++; | |
7544 | ||
4f5bc454 | 7545 | for (i = 0; i < mpb->num_raid_devs; i++) { |
00bbdbda N |
7546 | struct imsm_dev *dev; |
7547 | struct imsm_map *map; | |
86e3692b | 7548 | struct imsm_map *map2; |
4f5bc454 | 7549 | struct mdinfo *this; |
a6482415 | 7550 | int slot; |
a6482415 | 7551 | int chunk; |
00bbdbda N |
7552 | char *ep; |
7553 | ||
7554 | if (subarray && | |
7555 | (i != strtoul(subarray, &ep, 10) || *ep != '\0')) | |
7556 | continue; | |
7557 | ||
7558 | dev = get_imsm_dev(super, i); | |
238c0a71 AK |
7559 | map = get_imsm_map(dev, MAP_0); |
7560 | map2 = get_imsm_map(dev, MAP_1); | |
4f5bc454 | 7561 | |
1ce0101c DW |
7562 | /* do not publish arrays that are in the middle of an |
7563 | * unsupported migration | |
7564 | */ | |
7565 | if (dev->vol.migr_state && | |
28bce06f | 7566 | (migr_type(dev) == MIGR_STATE_CHANGE)) { |
7a862a02 | 7567 | pr_err("cannot assemble volume '%.16s': unsupported migration in progress\n", |
1ce0101c DW |
7568 | dev->volume); |
7569 | continue; | |
7570 | } | |
2db86302 LM |
7571 | /* do not publish arrays that are not support by controller's |
7572 | * OROM/EFI | |
7573 | */ | |
1ce0101c | 7574 | |
503975b9 | 7575 | this = xmalloc(sizeof(*this)); |
4f5bc454 | 7576 | |
301406c9 | 7577 | super->current_vol = i; |
a5d85af7 | 7578 | getinfo_super_imsm_volume(st, this, NULL); |
9894ec0d | 7579 | this->next = rest; |
a6482415 | 7580 | chunk = __le16_to_cpu(map->blocks_per_strip) >> 1; |
81219e70 LM |
7581 | /* mdadm does not support all metadata features- set the bit in all arrays state */ |
7582 | if (!validate_geometry_imsm_orom(super, | |
7583 | get_imsm_raid_level(map), /* RAID level */ | |
7584 | imsm_level_to_layout(get_imsm_raid_level(map)), | |
7585 | map->num_members, /* raid disks */ | |
2cc699af | 7586 | &chunk, join_u32(dev->size_low, dev->size_high), |
81219e70 | 7587 | 1 /* verbose */)) { |
7a862a02 | 7588 | pr_err("IMSM RAID geometry validation failed. Array %s activation is blocked.\n", |
81219e70 LM |
7589 | dev->volume); |
7590 | this->array.state |= | |
7591 | (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) | | |
7592 | (1<<MD_SB_BLOCK_VOLUME); | |
7593 | } | |
81219e70 LM |
7594 | |
7595 | /* if array has bad blocks, set suitable bit in all arrays state */ | |
7596 | if (sb_errors) | |
7597 | this->array.state |= | |
7598 | (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) | | |
7599 | (1<<MD_SB_BLOCK_VOLUME); | |
7600 | ||
4f5bc454 | 7601 | for (slot = 0 ; slot < map->num_members; slot++) { |
1e5c6983 | 7602 | unsigned long long recovery_start; |
4f5bc454 DW |
7603 | struct mdinfo *info_d; |
7604 | struct dl *d; | |
7605 | int idx; | |
9a1608e5 | 7606 | int skip; |
7eef0453 | 7607 | __u32 ord; |
4f5bc454 | 7608 | |
9a1608e5 | 7609 | skip = 0; |
238c0a71 AK |
7610 | idx = get_imsm_disk_idx(dev, slot, MAP_0); |
7611 | ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X); | |
4f5bc454 DW |
7612 | for (d = super->disks; d ; d = d->next) |
7613 | if (d->index == idx) | |
0fbd635c | 7614 | break; |
4f5bc454 | 7615 | |
1e5c6983 | 7616 | recovery_start = MaxSector; |
4f5bc454 | 7617 | if (d == NULL) |
9a1608e5 | 7618 | skip = 1; |
25ed7e59 | 7619 | if (d && is_failed(&d->disk)) |
9a1608e5 | 7620 | skip = 1; |
7eef0453 | 7621 | if (ord & IMSM_ORD_REBUILD) |
1e5c6983 | 7622 | recovery_start = 0; |
9a1608e5 | 7623 | |
1011e834 | 7624 | /* |
9a1608e5 | 7625 | * if we skip some disks the array will be assmebled degraded; |
1e5c6983 DW |
7626 | * reset resync start to avoid a dirty-degraded |
7627 | * situation when performing the intial sync | |
9a1608e5 DW |
7628 | * |
7629 | * FIXME handle dirty degraded | |
7630 | */ | |
2432ce9b AP |
7631 | if ((skip || recovery_start == 0) && |
7632 | !(dev->vol.dirty & RAIDVOL_DIRTY)) | |
b7528a20 | 7633 | this->resync_start = MaxSector; |
9a1608e5 DW |
7634 | if (skip) |
7635 | continue; | |
4f5bc454 | 7636 | |
503975b9 | 7637 | info_d = xcalloc(1, sizeof(*info_d)); |
4f5bc454 DW |
7638 | info_d->next = this->devs; |
7639 | this->devs = info_d; | |
7640 | ||
4f5bc454 DW |
7641 | info_d->disk.number = d->index; |
7642 | info_d->disk.major = d->major; | |
7643 | info_d->disk.minor = d->minor; | |
7644 | info_d->disk.raid_disk = slot; | |
1e5c6983 | 7645 | info_d->recovery_start = recovery_start; |
86e3692b AK |
7646 | if (map2) { |
7647 | if (slot < map2->num_members) | |
7648 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
7649 | else |
7650 | this->array.spare_disks++; | |
86e3692b AK |
7651 | } else { |
7652 | if (slot < map->num_members) | |
7653 | info_d->disk.state = (1 << MD_DISK_ACTIVE); | |
04c3c514 AK |
7654 | else |
7655 | this->array.spare_disks++; | |
86e3692b | 7656 | } |
1e5c6983 DW |
7657 | if (info_d->recovery_start == MaxSector) |
7658 | this->array.working_disks++; | |
4f5bc454 DW |
7659 | |
7660 | info_d->events = __le32_to_cpu(mpb->generation_num); | |
5551b113 | 7661 | info_d->data_offset = pba_of_lba0(map); |
06fb291a PB |
7662 | |
7663 | if (map->raid_level == 5) { | |
7664 | info_d->component_size = | |
7665 | num_data_stripes(map) * | |
7666 | map->blocks_per_strip; | |
2432ce9b AP |
7667 | info_d->ppl_sector = this->ppl_sector; |
7668 | info_d->ppl_size = this->ppl_size; | |
06fb291a PB |
7669 | } else { |
7670 | info_d->component_size = blocks_per_member(map); | |
7671 | } | |
2432ce9b | 7672 | info_d->consistency_policy = this->consistency_policy; |
b12796be | 7673 | |
5e46202e | 7674 | info_d->bb.supported = 1; |
b12796be TM |
7675 | get_volume_badblocks(super->bbm_log, ord_to_idx(ord), |
7676 | info_d->data_offset, | |
7677 | info_d->component_size, | |
7678 | &info_d->bb); | |
4f5bc454 | 7679 | } |
1e5c6983 | 7680 | /* now that the disk list is up-to-date fixup recovery_start */ |
c47b0ff6 | 7681 | update_recovery_start(super, dev, this); |
abef11a3 | 7682 | this->array.spare_disks += spare_disks; |
276d77db AK |
7683 | |
7684 | /* check for reshape */ | |
7685 | if (this->reshape_active == 1) | |
7686 | recover_backup_imsm(st, this); | |
9a1608e5 | 7687 | rest = this; |
4f5bc454 DW |
7688 | } |
7689 | ||
7690 | return rest; | |
cdddbdbc DW |
7691 | } |
7692 | ||
3b451610 AK |
7693 | static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, |
7694 | int failed, int look_in_map) | |
c2a1e7da | 7695 | { |
3b451610 AK |
7696 | struct imsm_map *map; |
7697 | ||
7698 | map = get_imsm_map(dev, look_in_map); | |
c2a1e7da DW |
7699 | |
7700 | if (!failed) | |
1011e834 | 7701 | return map->map_state == IMSM_T_STATE_UNINITIALIZED ? |
3393c6af | 7702 | IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL; |
c2a1e7da DW |
7703 | |
7704 | switch (get_imsm_raid_level(map)) { | |
7705 | case 0: | |
7706 | return IMSM_T_STATE_FAILED; | |
7707 | break; | |
7708 | case 1: | |
7709 | if (failed < map->num_members) | |
7710 | return IMSM_T_STATE_DEGRADED; | |
7711 | else | |
7712 | return IMSM_T_STATE_FAILED; | |
7713 | break; | |
7714 | case 10: | |
7715 | { | |
7716 | /** | |
c92a2527 DW |
7717 | * check to see if any mirrors have failed, otherwise we |
7718 | * are degraded. Even numbered slots are mirrored on | |
7719 | * slot+1 | |
c2a1e7da | 7720 | */ |
c2a1e7da | 7721 | int i; |
d9b420a5 N |
7722 | /* gcc -Os complains that this is unused */ |
7723 | int insync = insync; | |
c2a1e7da DW |
7724 | |
7725 | for (i = 0; i < map->num_members; i++) { | |
238c0a71 | 7726 | __u32 ord = get_imsm_ord_tbl_ent(dev, i, MAP_X); |
c92a2527 DW |
7727 | int idx = ord_to_idx(ord); |
7728 | struct imsm_disk *disk; | |
c2a1e7da | 7729 | |
c92a2527 | 7730 | /* reset the potential in-sync count on even-numbered |
1011e834 | 7731 | * slots. num_copies is always 2 for imsm raid10 |
c92a2527 DW |
7732 | */ |
7733 | if ((i & 1) == 0) | |
7734 | insync = 2; | |
c2a1e7da | 7735 | |
c92a2527 | 7736 | disk = get_imsm_disk(super, idx); |
25ed7e59 | 7737 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) |
c92a2527 | 7738 | insync--; |
c2a1e7da | 7739 | |
c92a2527 DW |
7740 | /* no in-sync disks left in this mirror the |
7741 | * array has failed | |
7742 | */ | |
7743 | if (insync == 0) | |
7744 | return IMSM_T_STATE_FAILED; | |
c2a1e7da DW |
7745 | } |
7746 | ||
7747 | return IMSM_T_STATE_DEGRADED; | |
7748 | } | |
7749 | case 5: | |
7750 | if (failed < 2) | |
7751 | return IMSM_T_STATE_DEGRADED; | |
7752 | else | |
7753 | return IMSM_T_STATE_FAILED; | |
7754 | break; | |
7755 | default: | |
7756 | break; | |
7757 | } | |
7758 | ||
7759 | return map->map_state; | |
7760 | } | |
7761 | ||
3b451610 AK |
7762 | static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev, |
7763 | int look_in_map) | |
c2a1e7da DW |
7764 | { |
7765 | int i; | |
7766 | int failed = 0; | |
7767 | struct imsm_disk *disk; | |
d5985138 AK |
7768 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
7769 | struct imsm_map *prev = get_imsm_map(dev, MAP_1); | |
68fe4598 | 7770 | struct imsm_map *map_for_loop; |
0556e1a2 DW |
7771 | __u32 ord; |
7772 | int idx; | |
d5985138 | 7773 | int idx_1; |
c2a1e7da | 7774 | |
0556e1a2 DW |
7775 | /* at the beginning of migration we set IMSM_ORD_REBUILD on |
7776 | * disks that are being rebuilt. New failures are recorded to | |
7777 | * map[0]. So we look through all the disks we started with and | |
7778 | * see if any failures are still present, or if any new ones | |
7779 | * have arrived | |
0556e1a2 | 7780 | */ |
d5985138 AK |
7781 | map_for_loop = map; |
7782 | if (prev && (map->num_members < prev->num_members)) | |
7783 | map_for_loop = prev; | |
68fe4598 LD |
7784 | |
7785 | for (i = 0; i < map_for_loop->num_members; i++) { | |
d5985138 | 7786 | idx_1 = -255; |
238c0a71 AK |
7787 | /* when MAP_X is passed both maps failures are counted |
7788 | */ | |
d5985138 | 7789 | if (prev && |
089f9d79 JS |
7790 | (look_in_map == MAP_1 || look_in_map == MAP_X) && |
7791 | i < prev->num_members) { | |
d5985138 AK |
7792 | ord = __le32_to_cpu(prev->disk_ord_tbl[i]); |
7793 | idx_1 = ord_to_idx(ord); | |
c2a1e7da | 7794 | |
d5985138 AK |
7795 | disk = get_imsm_disk(super, idx_1); |
7796 | if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD) | |
7797 | failed++; | |
7798 | } | |
089f9d79 JS |
7799 | if ((look_in_map == MAP_0 || look_in_map == MAP_X) && |
7800 | i < map->num_members) { | |
d5985138 AK |
7801 | ord = __le32_to_cpu(map->disk_ord_tbl[i]); |
7802 | idx = ord_to_idx(ord); | |
7803 | ||
7804 | if (idx != idx_1) { | |
7805 | disk = get_imsm_disk(super, idx); | |
7806 | if (!disk || is_failed(disk) || | |
7807 | ord & IMSM_ORD_REBUILD) | |
7808 | failed++; | |
7809 | } | |
7810 | } | |
c2a1e7da DW |
7811 | } |
7812 | ||
7813 | return failed; | |
845dea95 NB |
7814 | } |
7815 | ||
97b4d0e9 DW |
7816 | static int imsm_open_new(struct supertype *c, struct active_array *a, |
7817 | char *inst) | |
7818 | { | |
7819 | struct intel_super *super = c->sb; | |
7820 | struct imsm_super *mpb = super->anchor; | |
bbab0940 | 7821 | struct imsm_update_prealloc_bb_mem u; |
9587c373 | 7822 | |
97b4d0e9 | 7823 | if (atoi(inst) >= mpb->num_raid_devs) { |
1ade5cc1 | 7824 | pr_err("subarry index %d, out of range\n", atoi(inst)); |
97b4d0e9 DW |
7825 | return -ENODEV; |
7826 | } | |
7827 | ||
7828 | dprintf("imsm: open_new %s\n", inst); | |
7829 | a->info.container_member = atoi(inst); | |
bbab0940 TM |
7830 | |
7831 | u.type = update_prealloc_badblocks_mem; | |
7832 | imsm_update_metadata_locally(c, &u, sizeof(u)); | |
7833 | ||
97b4d0e9 DW |
7834 | return 0; |
7835 | } | |
7836 | ||
0c046afd DW |
7837 | static int is_resyncing(struct imsm_dev *dev) |
7838 | { | |
7839 | struct imsm_map *migr_map; | |
7840 | ||
7841 | if (!dev->vol.migr_state) | |
7842 | return 0; | |
7843 | ||
1484e727 DW |
7844 | if (migr_type(dev) == MIGR_INIT || |
7845 | migr_type(dev) == MIGR_REPAIR) | |
0c046afd DW |
7846 | return 1; |
7847 | ||
4c9bc37b AK |
7848 | if (migr_type(dev) == MIGR_GEN_MIGR) |
7849 | return 0; | |
7850 | ||
238c0a71 | 7851 | migr_map = get_imsm_map(dev, MAP_1); |
0c046afd | 7852 | |
089f9d79 JS |
7853 | if (migr_map->map_state == IMSM_T_STATE_NORMAL && |
7854 | dev->vol.migr_type != MIGR_GEN_MIGR) | |
0c046afd DW |
7855 | return 1; |
7856 | else | |
7857 | return 0; | |
7858 | } | |
7859 | ||
0556e1a2 | 7860 | /* return true if we recorded new information */ |
4c9e8c1e TM |
7861 | static int mark_failure(struct intel_super *super, |
7862 | struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
47ee5a45 | 7863 | { |
0556e1a2 DW |
7864 | __u32 ord; |
7865 | int slot; | |
7866 | struct imsm_map *map; | |
86c54047 DW |
7867 | char buf[MAX_RAID_SERIAL_LEN+3]; |
7868 | unsigned int len, shift = 0; | |
0556e1a2 DW |
7869 | |
7870 | /* new failures are always set in map[0] */ | |
238c0a71 | 7871 | map = get_imsm_map(dev, MAP_0); |
0556e1a2 DW |
7872 | |
7873 | slot = get_imsm_disk_slot(map, idx); | |
7874 | if (slot < 0) | |
7875 | return 0; | |
7876 | ||
7877 | ord = __le32_to_cpu(map->disk_ord_tbl[slot]); | |
25ed7e59 | 7878 | if (is_failed(disk) && (ord & IMSM_ORD_REBUILD)) |
0556e1a2 DW |
7879 | return 0; |
7880 | ||
7d0c5e24 LD |
7881 | memcpy(buf, disk->serial, MAX_RAID_SERIAL_LEN); |
7882 | buf[MAX_RAID_SERIAL_LEN] = '\000'; | |
7883 | strcat(buf, ":0"); | |
86c54047 DW |
7884 | if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN) |
7885 | shift = len - MAX_RAID_SERIAL_LEN + 1; | |
7886 | strncpy((char *)disk->serial, &buf[shift], MAX_RAID_SERIAL_LEN); | |
7887 | ||
f2f27e63 | 7888 | disk->status |= FAILED_DISK; |
0556e1a2 | 7889 | set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD); |
17788994 AK |
7890 | /* mark failures in second map if second map exists and this disk |
7891 | * in this slot. | |
7892 | * This is valid for migration, initialization and rebuild | |
7893 | */ | |
7894 | if (dev->vol.migr_state) { | |
238c0a71 | 7895 | struct imsm_map *map2 = get_imsm_map(dev, MAP_1); |
0a108d63 AK |
7896 | int slot2 = get_imsm_disk_slot(map2, idx); |
7897 | ||
089f9d79 | 7898 | if (slot2 < map2->num_members && slot2 >= 0) |
0a108d63 | 7899 | set_imsm_ord_tbl_ent(map2, slot2, |
1ace8403 AK |
7900 | idx | IMSM_ORD_REBUILD); |
7901 | } | |
f21e18ca | 7902 | if (map->failed_disk_num == 0xff) |
0556e1a2 | 7903 | map->failed_disk_num = slot; |
4c9e8c1e TM |
7904 | |
7905 | clear_disk_badblocks(super->bbm_log, ord_to_idx(ord)); | |
7906 | ||
0556e1a2 DW |
7907 | return 1; |
7908 | } | |
7909 | ||
4c9e8c1e TM |
7910 | static void mark_missing(struct intel_super *super, |
7911 | struct imsm_dev *dev, struct imsm_disk *disk, int idx) | |
0556e1a2 | 7912 | { |
4c9e8c1e | 7913 | mark_failure(super, dev, disk, idx); |
0556e1a2 DW |
7914 | |
7915 | if (disk->scsi_id == __cpu_to_le32(~(__u32)0)) | |
7916 | return; | |
7917 | ||
47ee5a45 DW |
7918 | disk->scsi_id = __cpu_to_le32(~(__u32)0); |
7919 | memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1); | |
7920 | } | |
7921 | ||
33414a01 DW |
7922 | static void handle_missing(struct intel_super *super, struct imsm_dev *dev) |
7923 | { | |
33414a01 | 7924 | struct dl *dl; |
33414a01 DW |
7925 | |
7926 | if (!super->missing) | |
7927 | return; | |
33414a01 | 7928 | |
79b68f1b PC |
7929 | /* When orom adds replacement for missing disk it does |
7930 | * not remove entry of missing disk, but just updates map with | |
7931 | * new added disk. So it is not enough just to test if there is | |
7932 | * any missing disk, we have to look if there are any failed disks | |
7933 | * in map to stop migration */ | |
7934 | ||
33414a01 | 7935 | dprintf("imsm: mark missing\n"); |
3d59f0c0 AK |
7936 | /* end process for initialization and rebuild only |
7937 | */ | |
7938 | if (is_gen_migration(dev) == 0) { | |
fb12a745 | 7939 | int failed = imsm_count_failed(super, dev, MAP_0); |
3d59f0c0 | 7940 | |
fb12a745 TM |
7941 | if (failed) { |
7942 | __u8 map_state; | |
7943 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
7944 | struct imsm_map *map1; | |
7945 | int i, ord, ord_map1; | |
7946 | int rebuilt = 1; | |
3d59f0c0 | 7947 | |
fb12a745 TM |
7948 | for (i = 0; i < map->num_members; i++) { |
7949 | ord = get_imsm_ord_tbl_ent(dev, i, MAP_0); | |
7950 | if (!(ord & IMSM_ORD_REBUILD)) | |
7951 | continue; | |
7952 | ||
7953 | map1 = get_imsm_map(dev, MAP_1); | |
7954 | if (!map1) | |
7955 | continue; | |
7956 | ||
7957 | ord_map1 = __le32_to_cpu(map1->disk_ord_tbl[i]); | |
7958 | if (ord_map1 & IMSM_ORD_REBUILD) | |
7959 | rebuilt = 0; | |
7960 | } | |
7961 | ||
7962 | if (rebuilt) { | |
7963 | map_state = imsm_check_degraded(super, dev, | |
7964 | failed, MAP_0); | |
7965 | end_migration(dev, super, map_state); | |
7966 | } | |
7967 | } | |
3d59f0c0 | 7968 | } |
33414a01 | 7969 | for (dl = super->missing; dl; dl = dl->next) |
4c9e8c1e | 7970 | mark_missing(super, dev, &dl->disk, dl->index); |
33414a01 DW |
7971 | super->updates_pending++; |
7972 | } | |
7973 | ||
f3871fdc AK |
7974 | static unsigned long long imsm_set_array_size(struct imsm_dev *dev, |
7975 | long long new_size) | |
70bdf0dc | 7976 | { |
238c0a71 | 7977 | int used_disks = imsm_num_data_members(dev, MAP_0); |
70bdf0dc AK |
7978 | unsigned long long array_blocks; |
7979 | struct imsm_map *map; | |
7980 | ||
7981 | if (used_disks == 0) { | |
7982 | /* when problems occures | |
7983 | * return current array_blocks value | |
7984 | */ | |
7985 | array_blocks = __le32_to_cpu(dev->size_high); | |
7986 | array_blocks = array_blocks << 32; | |
7987 | array_blocks += __le32_to_cpu(dev->size_low); | |
7988 | ||
7989 | return array_blocks; | |
7990 | } | |
7991 | ||
7992 | /* set array size in metadata | |
7993 | */ | |
f3871fdc AK |
7994 | if (new_size <= 0) { |
7995 | /* OLCE size change is caused by added disks | |
7996 | */ | |
7997 | map = get_imsm_map(dev, MAP_0); | |
7998 | array_blocks = blocks_per_member(map) * used_disks; | |
7999 | } else { | |
8000 | /* Online Volume Size Change | |
8001 | * Using available free space | |
8002 | */ | |
8003 | array_blocks = new_size; | |
8004 | } | |
70bdf0dc | 8005 | |
b53bfba6 | 8006 | array_blocks = round_size_to_mb(array_blocks, used_disks); |
70bdf0dc AK |
8007 | dev->size_low = __cpu_to_le32((__u32)array_blocks); |
8008 | dev->size_high = __cpu_to_le32((__u32)(array_blocks >> 32)); | |
8009 | ||
8010 | return array_blocks; | |
8011 | } | |
8012 | ||
28bce06f AK |
8013 | static void imsm_set_disk(struct active_array *a, int n, int state); |
8014 | ||
0e2d1a4e AK |
8015 | static void imsm_progress_container_reshape(struct intel_super *super) |
8016 | { | |
8017 | /* if no device has a migr_state, but some device has a | |
8018 | * different number of members than the previous device, start | |
8019 | * changing the number of devices in this device to match | |
8020 | * previous. | |
8021 | */ | |
8022 | struct imsm_super *mpb = super->anchor; | |
8023 | int prev_disks = -1; | |
8024 | int i; | |
1dfaa380 | 8025 | int copy_map_size; |
0e2d1a4e AK |
8026 | |
8027 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
8028 | struct imsm_dev *dev = get_imsm_dev(super, i); | |
238c0a71 | 8029 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
0e2d1a4e AK |
8030 | struct imsm_map *map2; |
8031 | int prev_num_members; | |
0e2d1a4e AK |
8032 | |
8033 | if (dev->vol.migr_state) | |
8034 | return; | |
8035 | ||
8036 | if (prev_disks == -1) | |
8037 | prev_disks = map->num_members; | |
8038 | if (prev_disks == map->num_members) | |
8039 | continue; | |
8040 | ||
8041 | /* OK, this array needs to enter reshape mode. | |
8042 | * i.e it needs a migr_state | |
8043 | */ | |
8044 | ||
1dfaa380 | 8045 | copy_map_size = sizeof_imsm_map(map); |
0e2d1a4e AK |
8046 | prev_num_members = map->num_members; |
8047 | map->num_members = prev_disks; | |
8048 | dev->vol.migr_state = 1; | |
8049 | dev->vol.curr_migr_unit = 0; | |
ea672ee1 | 8050 | set_migr_type(dev, MIGR_GEN_MIGR); |
0e2d1a4e AK |
8051 | for (i = prev_num_members; |
8052 | i < map->num_members; i++) | |
8053 | set_imsm_ord_tbl_ent(map, i, i); | |
238c0a71 | 8054 | map2 = get_imsm_map(dev, MAP_1); |
0e2d1a4e | 8055 | /* Copy the current map */ |
1dfaa380 | 8056 | memcpy(map2, map, copy_map_size); |
0e2d1a4e AK |
8057 | map2->num_members = prev_num_members; |
8058 | ||
f3871fdc | 8059 | imsm_set_array_size(dev, -1); |
51d83f5d | 8060 | super->clean_migration_record_by_mdmon = 1; |
0e2d1a4e AK |
8061 | super->updates_pending++; |
8062 | } | |
8063 | } | |
8064 | ||
aad6f216 | 8065 | /* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild |
0c046afd DW |
8066 | * states are handled in imsm_set_disk() with one exception, when a |
8067 | * resync is stopped due to a new failure this routine will set the | |
8068 | * 'degraded' state for the array. | |
8069 | */ | |
01f157d7 | 8070 | static int imsm_set_array_state(struct active_array *a, int consistent) |
a862209d DW |
8071 | { |
8072 | int inst = a->info.container_member; | |
8073 | struct intel_super *super = a->container->sb; | |
949c47a0 | 8074 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 8075 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
3b451610 AK |
8076 | int failed = imsm_count_failed(super, dev, MAP_0); |
8077 | __u8 map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
1e5c6983 | 8078 | __u32 blocks_per_unit; |
a862209d | 8079 | |
1af97990 AK |
8080 | if (dev->vol.migr_state && |
8081 | dev->vol.migr_type == MIGR_GEN_MIGR) { | |
8082 | /* array state change is blocked due to reshape action | |
aad6f216 N |
8083 | * We might need to |
8084 | * - abort the reshape (if last_checkpoint is 0 and action!= reshape) | |
8085 | * - finish the reshape (if last_checkpoint is big and action != reshape) | |
8086 | * - update curr_migr_unit | |
1af97990 | 8087 | */ |
aad6f216 N |
8088 | if (a->curr_action == reshape) { |
8089 | /* still reshaping, maybe update curr_migr_unit */ | |
633b5610 | 8090 | goto mark_checkpoint; |
aad6f216 N |
8091 | } else { |
8092 | if (a->last_checkpoint == 0 && a->prev_action == reshape) { | |
8093 | /* for some reason we aborted the reshape. | |
b66e591b AK |
8094 | * |
8095 | * disable automatic metadata rollback | |
8096 | * user action is required to recover process | |
aad6f216 | 8097 | */ |
b66e591b | 8098 | if (0) { |
238c0a71 AK |
8099 | struct imsm_map *map2 = |
8100 | get_imsm_map(dev, MAP_1); | |
8101 | dev->vol.migr_state = 0; | |
8102 | set_migr_type(dev, 0); | |
8103 | dev->vol.curr_migr_unit = 0; | |
8104 | memcpy(map, map2, | |
8105 | sizeof_imsm_map(map2)); | |
8106 | super->updates_pending++; | |
b66e591b | 8107 | } |
aad6f216 N |
8108 | } |
8109 | if (a->last_checkpoint >= a->info.component_size) { | |
8110 | unsigned long long array_blocks; | |
8111 | int used_disks; | |
e154ced3 | 8112 | struct mdinfo *mdi; |
aad6f216 | 8113 | |
238c0a71 | 8114 | used_disks = imsm_num_data_members(dev, MAP_0); |
d55adef9 AK |
8115 | if (used_disks > 0) { |
8116 | array_blocks = | |
5551b113 | 8117 | blocks_per_member(map) * |
d55adef9 | 8118 | used_disks; |
b53bfba6 TM |
8119 | array_blocks = |
8120 | round_size_to_mb(array_blocks, | |
8121 | used_disks); | |
d55adef9 AK |
8122 | a->info.custom_array_size = array_blocks; |
8123 | /* encourage manager to update array | |
8124 | * size | |
8125 | */ | |
e154ced3 | 8126 | |
d55adef9 | 8127 | a->check_reshape = 1; |
633b5610 | 8128 | } |
e154ced3 AK |
8129 | /* finalize online capacity expansion/reshape */ |
8130 | for (mdi = a->info.devs; mdi; mdi = mdi->next) | |
8131 | imsm_set_disk(a, | |
8132 | mdi->disk.raid_disk, | |
8133 | mdi->curr_state); | |
8134 | ||
0e2d1a4e | 8135 | imsm_progress_container_reshape(super); |
e154ced3 | 8136 | } |
aad6f216 | 8137 | } |
1af97990 AK |
8138 | } |
8139 | ||
47ee5a45 | 8140 | /* before we activate this array handle any missing disks */ |
33414a01 DW |
8141 | if (consistent == 2) |
8142 | handle_missing(super, dev); | |
1e5c6983 | 8143 | |
0c046afd | 8144 | if (consistent == 2 && |
b7941fd6 | 8145 | (!is_resync_complete(&a->info) || |
0c046afd DW |
8146 | map_state != IMSM_T_STATE_NORMAL || |
8147 | dev->vol.migr_state)) | |
01f157d7 | 8148 | consistent = 0; |
272906ef | 8149 | |
b7941fd6 | 8150 | if (is_resync_complete(&a->info)) { |
0c046afd | 8151 | /* complete intialization / resync, |
0556e1a2 DW |
8152 | * recovery and interrupted recovery is completed in |
8153 | * ->set_disk | |
0c046afd DW |
8154 | */ |
8155 | if (is_resyncing(dev)) { | |
8156 | dprintf("imsm: mark resync done\n"); | |
809da78e | 8157 | end_migration(dev, super, map_state); |
115c3803 | 8158 | super->updates_pending++; |
484240d8 | 8159 | a->last_checkpoint = 0; |
115c3803 | 8160 | } |
b9172665 AK |
8161 | } else if ((!is_resyncing(dev) && !failed) && |
8162 | (imsm_reshape_blocks_arrays_changes(super) == 0)) { | |
0c046afd | 8163 | /* mark the start of the init process if nothing is failed */ |
b7941fd6 | 8164 | dprintf("imsm: mark resync start\n"); |
1484e727 | 8165 | if (map->map_state == IMSM_T_STATE_UNINITIALIZED) |
8e59f3d8 | 8166 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT); |
1484e727 | 8167 | else |
8e59f3d8 | 8168 | migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR); |
3393c6af | 8169 | super->updates_pending++; |
115c3803 | 8170 | } |
a862209d | 8171 | |
633b5610 | 8172 | mark_checkpoint: |
5b83bacf AK |
8173 | /* skip checkpointing for general migration, |
8174 | * it is controlled in mdadm | |
8175 | */ | |
8176 | if (is_gen_migration(dev)) | |
8177 | goto skip_mark_checkpoint; | |
8178 | ||
1e5c6983 | 8179 | /* check if we can update curr_migr_unit from resync_start, recovery_start */ |
c47b0ff6 | 8180 | blocks_per_unit = blocks_per_migr_unit(super, dev); |
4f0a7acc | 8181 | if (blocks_per_unit) { |
1e5c6983 DW |
8182 | __u32 units32; |
8183 | __u64 units; | |
8184 | ||
4f0a7acc | 8185 | units = a->last_checkpoint / blocks_per_unit; |
1e5c6983 DW |
8186 | units32 = units; |
8187 | ||
8188 | /* check that we did not overflow 32-bits, and that | |
8189 | * curr_migr_unit needs updating | |
8190 | */ | |
8191 | if (units32 == units && | |
bfd80a56 | 8192 | units32 != 0 && |
1e5c6983 DW |
8193 | __le32_to_cpu(dev->vol.curr_migr_unit) != units32) { |
8194 | dprintf("imsm: mark checkpoint (%u)\n", units32); | |
8195 | dev->vol.curr_migr_unit = __cpu_to_le32(units32); | |
8196 | super->updates_pending++; | |
8197 | } | |
8198 | } | |
f8f603f1 | 8199 | |
5b83bacf | 8200 | skip_mark_checkpoint: |
3393c6af | 8201 | /* mark dirty / clean */ |
2432ce9b AP |
8202 | if (((dev->vol.dirty & RAIDVOL_DIRTY) && consistent) || |
8203 | (!(dev->vol.dirty & RAIDVOL_DIRTY) && !consistent)) { | |
b7941fd6 | 8204 | dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty"); |
2432ce9b AP |
8205 | if (consistent) { |
8206 | dev->vol.dirty = RAIDVOL_CLEAN; | |
8207 | } else { | |
8208 | dev->vol.dirty = RAIDVOL_DIRTY; | |
8209 | if (dev->rwh_policy == RWH_DISTRIBUTED) | |
8210 | dev->vol.dirty |= RAIDVOL_DSRECORD_VALID; | |
8211 | } | |
a862209d DW |
8212 | super->updates_pending++; |
8213 | } | |
28bce06f | 8214 | |
01f157d7 | 8215 | return consistent; |
a862209d DW |
8216 | } |
8217 | ||
6f50473f TM |
8218 | static int imsm_disk_slot_to_ord(struct active_array *a, int slot) |
8219 | { | |
8220 | int inst = a->info.container_member; | |
8221 | struct intel_super *super = a->container->sb; | |
8222 | struct imsm_dev *dev = get_imsm_dev(super, inst); | |
8223 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
8224 | ||
8225 | if (slot > map->num_members) { | |
8226 | pr_err("imsm: imsm_disk_slot_to_ord %d out of range 0..%d\n", | |
8227 | slot, map->num_members - 1); | |
8228 | return -1; | |
8229 | } | |
8230 | ||
8231 | if (slot < 0) | |
8232 | return -1; | |
8233 | ||
8234 | return get_imsm_ord_tbl_ent(dev, slot, MAP_0); | |
8235 | } | |
8236 | ||
8d45d196 | 8237 | static void imsm_set_disk(struct active_array *a, int n, int state) |
845dea95 | 8238 | { |
8d45d196 DW |
8239 | int inst = a->info.container_member; |
8240 | struct intel_super *super = a->container->sb; | |
949c47a0 | 8241 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 8242 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
8d45d196 | 8243 | struct imsm_disk *disk; |
7ce05701 LD |
8244 | struct mdinfo *mdi; |
8245 | int recovery_not_finished = 0; | |
0c046afd | 8246 | int failed; |
6f50473f | 8247 | int ord; |
0c046afd | 8248 | __u8 map_state; |
fb12a745 TM |
8249 | int rebuild_done = 0; |
8250 | int i; | |
8d45d196 | 8251 | |
fb12a745 | 8252 | ord = get_imsm_ord_tbl_ent(dev, n, MAP_X); |
6f50473f | 8253 | if (ord < 0) |
8d45d196 DW |
8254 | return; |
8255 | ||
4e6e574a | 8256 | dprintf("imsm: set_disk %d:%x\n", n, state); |
b10b37b8 | 8257 | disk = get_imsm_disk(super, ord_to_idx(ord)); |
8d45d196 | 8258 | |
5802a811 | 8259 | /* check for new failures */ |
0556e1a2 | 8260 | if (state & DS_FAULTY) { |
4c9e8c1e | 8261 | if (mark_failure(super, dev, disk, ord_to_idx(ord))) |
0556e1a2 | 8262 | super->updates_pending++; |
8d45d196 | 8263 | } |
47ee5a45 | 8264 | |
19859edc | 8265 | /* check if in_sync */ |
0556e1a2 | 8266 | if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) { |
238c0a71 | 8267 | struct imsm_map *migr_map = get_imsm_map(dev, MAP_1); |
b10b37b8 DW |
8268 | |
8269 | set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord)); | |
fb12a745 | 8270 | rebuild_done = 1; |
19859edc DW |
8271 | super->updates_pending++; |
8272 | } | |
8d45d196 | 8273 | |
3b451610 AK |
8274 | failed = imsm_count_failed(super, dev, MAP_0); |
8275 | map_state = imsm_check_degraded(super, dev, failed, MAP_0); | |
5802a811 | 8276 | |
0c046afd | 8277 | /* check if recovery complete, newly degraded, or failed */ |
94002678 AK |
8278 | dprintf("imsm: Detected transition to state "); |
8279 | switch (map_state) { | |
8280 | case IMSM_T_STATE_NORMAL: /* transition to normal state */ | |
8281 | dprintf("normal: "); | |
8282 | if (is_rebuilding(dev)) { | |
1ade5cc1 | 8283 | dprintf_cont("while rebuilding"); |
7ce05701 LD |
8284 | /* check if recovery is really finished */ |
8285 | for (mdi = a->info.devs; mdi ; mdi = mdi->next) | |
8286 | if (mdi->recovery_start != MaxSector) { | |
8287 | recovery_not_finished = 1; | |
8288 | break; | |
8289 | } | |
8290 | if (recovery_not_finished) { | |
1ade5cc1 N |
8291 | dprintf_cont("\n"); |
8292 | dprintf("Rebuild has not finished yet, state not changed"); | |
7ce05701 LD |
8293 | if (a->last_checkpoint < mdi->recovery_start) { |
8294 | a->last_checkpoint = mdi->recovery_start; | |
8295 | super->updates_pending++; | |
8296 | } | |
8297 | break; | |
8298 | } | |
94002678 | 8299 | end_migration(dev, super, map_state); |
238c0a71 | 8300 | map = get_imsm_map(dev, MAP_0); |
94002678 AK |
8301 | map->failed_disk_num = ~0; |
8302 | super->updates_pending++; | |
8303 | a->last_checkpoint = 0; | |
8304 | break; | |
8305 | } | |
8306 | if (is_gen_migration(dev)) { | |
1ade5cc1 | 8307 | dprintf_cont("while general migration"); |
bf2f0071 | 8308 | if (a->last_checkpoint >= a->info.component_size) |
809da78e | 8309 | end_migration(dev, super, map_state); |
94002678 AK |
8310 | else |
8311 | map->map_state = map_state; | |
238c0a71 | 8312 | map = get_imsm_map(dev, MAP_0); |
28bce06f | 8313 | map->failed_disk_num = ~0; |
94002678 | 8314 | super->updates_pending++; |
bf2f0071 | 8315 | break; |
94002678 AK |
8316 | } |
8317 | break; | |
8318 | case IMSM_T_STATE_DEGRADED: /* transition to degraded state */ | |
1ade5cc1 | 8319 | dprintf_cont("degraded: "); |
089f9d79 | 8320 | if (map->map_state != map_state && !dev->vol.migr_state) { |
1ade5cc1 | 8321 | dprintf_cont("mark degraded"); |
94002678 AK |
8322 | map->map_state = map_state; |
8323 | super->updates_pending++; | |
8324 | a->last_checkpoint = 0; | |
8325 | break; | |
8326 | } | |
8327 | if (is_rebuilding(dev)) { | |
1ade5cc1 | 8328 | dprintf_cont("while rebuilding."); |
94002678 | 8329 | if (map->map_state != map_state) { |
1ade5cc1 | 8330 | dprintf_cont(" Map state change"); |
94002678 AK |
8331 | end_migration(dev, super, map_state); |
8332 | super->updates_pending++; | |
fb12a745 TM |
8333 | } else if (!rebuild_done) { |
8334 | break; | |
8335 | } | |
8336 | ||
8337 | /* check if recovery is really finished */ | |
8338 | for (mdi = a->info.devs; mdi ; mdi = mdi->next) | |
8339 | if (mdi->recovery_start != MaxSector) { | |
8340 | recovery_not_finished = 1; | |
8341 | break; | |
8342 | } | |
8343 | if (recovery_not_finished) { | |
8344 | dprintf_cont("\n"); | |
8345 | dprintf("Rebuild has not finished yet, state not changed"); | |
8346 | if (a->last_checkpoint < mdi->recovery_start) { | |
8347 | a->last_checkpoint = | |
8348 | mdi->recovery_start; | |
8349 | super->updates_pending++; | |
8350 | } | |
8351 | break; | |
94002678 | 8352 | } |
fb12a745 TM |
8353 | |
8354 | dprintf_cont(" Rebuild done, still degraded"); | |
8355 | dev->vol.migr_state = 0; | |
8356 | set_migr_type(dev, 0); | |
8357 | dev->vol.curr_migr_unit = 0; | |
8358 | ||
8359 | for (i = 0; i < map->num_members; i++) { | |
8360 | int idx = get_imsm_ord_tbl_ent(dev, i, MAP_0); | |
8361 | ||
8362 | if (idx & IMSM_ORD_REBUILD) | |
8363 | map->failed_disk_num = i; | |
8364 | } | |
8365 | super->updates_pending++; | |
94002678 AK |
8366 | break; |
8367 | } | |
8368 | if (is_gen_migration(dev)) { | |
1ade5cc1 | 8369 | dprintf_cont("while general migration"); |
bf2f0071 | 8370 | if (a->last_checkpoint >= a->info.component_size) |
809da78e | 8371 | end_migration(dev, super, map_state); |
94002678 AK |
8372 | else { |
8373 | map->map_state = map_state; | |
3b451610 | 8374 | manage_second_map(super, dev); |
94002678 AK |
8375 | } |
8376 | super->updates_pending++; | |
bf2f0071 | 8377 | break; |
28bce06f | 8378 | } |
6ce1fbf1 | 8379 | if (is_initializing(dev)) { |
1ade5cc1 | 8380 | dprintf_cont("while initialization."); |
6ce1fbf1 AK |
8381 | map->map_state = map_state; |
8382 | super->updates_pending++; | |
8383 | break; | |
8384 | } | |
94002678 AK |
8385 | break; |
8386 | case IMSM_T_STATE_FAILED: /* transition to failed state */ | |
1ade5cc1 | 8387 | dprintf_cont("failed: "); |
94002678 | 8388 | if (is_gen_migration(dev)) { |
1ade5cc1 | 8389 | dprintf_cont("while general migration"); |
94002678 AK |
8390 | map->map_state = map_state; |
8391 | super->updates_pending++; | |
8392 | break; | |
8393 | } | |
8394 | if (map->map_state != map_state) { | |
1ade5cc1 | 8395 | dprintf_cont("mark failed"); |
94002678 AK |
8396 | end_migration(dev, super, map_state); |
8397 | super->updates_pending++; | |
8398 | a->last_checkpoint = 0; | |
8399 | break; | |
8400 | } | |
8401 | break; | |
8402 | default: | |
1ade5cc1 | 8403 | dprintf_cont("state %i\n", map_state); |
5802a811 | 8404 | } |
1ade5cc1 | 8405 | dprintf_cont("\n"); |
845dea95 NB |
8406 | } |
8407 | ||
f796af5d | 8408 | static int store_imsm_mpb(int fd, struct imsm_super *mpb) |
c2a1e7da | 8409 | { |
f796af5d | 8410 | void *buf = mpb; |
c2a1e7da DW |
8411 | __u32 mpb_size = __le32_to_cpu(mpb->mpb_size); |
8412 | unsigned long long dsize; | |
8413 | unsigned long long sectors; | |
f36a9ecd | 8414 | unsigned int sector_size; |
c2a1e7da | 8415 | |
f36a9ecd | 8416 | get_dev_sector_size(fd, NULL, §or_size); |
c2a1e7da DW |
8417 | get_dev_size(fd, NULL, &dsize); |
8418 | ||
f36a9ecd | 8419 | if (mpb_size > sector_size) { |
272f648f | 8420 | /* -1 to account for anchor */ |
f36a9ecd | 8421 | sectors = mpb_sectors(mpb, sector_size) - 1; |
c2a1e7da | 8422 | |
272f648f | 8423 | /* write the extended mpb to the sectors preceeding the anchor */ |
f36a9ecd PB |
8424 | if (lseek64(fd, dsize - (sector_size * (2 + sectors)), |
8425 | SEEK_SET) < 0) | |
272f648f | 8426 | return 1; |
c2a1e7da | 8427 | |
f36a9ecd PB |
8428 | if ((unsigned long long)write(fd, buf + sector_size, |
8429 | sector_size * sectors) != sector_size * sectors) | |
272f648f DW |
8430 | return 1; |
8431 | } | |
c2a1e7da | 8432 | |
272f648f | 8433 | /* first block is stored on second to last sector of the disk */ |
f36a9ecd | 8434 | if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0) |
c2a1e7da DW |
8435 | return 1; |
8436 | ||
466070ad | 8437 | if ((unsigned int)write(fd, buf, sector_size) != sector_size) |
c2a1e7da DW |
8438 | return 1; |
8439 | ||
c2a1e7da DW |
8440 | return 0; |
8441 | } | |
8442 | ||
2e735d19 | 8443 | static void imsm_sync_metadata(struct supertype *container) |
845dea95 | 8444 | { |
2e735d19 | 8445 | struct intel_super *super = container->sb; |
c2a1e7da | 8446 | |
1a64be56 | 8447 | dprintf("sync metadata: %d\n", super->updates_pending); |
c2a1e7da DW |
8448 | if (!super->updates_pending) |
8449 | return; | |
8450 | ||
36988a3d | 8451 | write_super_imsm(container, 0); |
c2a1e7da DW |
8452 | |
8453 | super->updates_pending = 0; | |
845dea95 NB |
8454 | } |
8455 | ||
272906ef DW |
8456 | static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a) |
8457 | { | |
8458 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 8459 | int i = get_imsm_disk_idx(dev, idx, MAP_X); |
272906ef DW |
8460 | struct dl *dl; |
8461 | ||
8462 | for (dl = super->disks; dl; dl = dl->next) | |
8463 | if (dl->index == i) | |
8464 | break; | |
8465 | ||
25ed7e59 | 8466 | if (dl && is_failed(&dl->disk)) |
272906ef DW |
8467 | dl = NULL; |
8468 | ||
8469 | if (dl) | |
1ade5cc1 | 8470 | dprintf("found %x:%x\n", dl->major, dl->minor); |
272906ef DW |
8471 | |
8472 | return dl; | |
8473 | } | |
8474 | ||
a20d2ba5 | 8475 | static struct dl *imsm_add_spare(struct intel_super *super, int slot, |
8ba77d32 AK |
8476 | struct active_array *a, int activate_new, |
8477 | struct mdinfo *additional_test_list) | |
272906ef DW |
8478 | { |
8479 | struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 8480 | int idx = get_imsm_disk_idx(dev, slot, MAP_X); |
a20d2ba5 DW |
8481 | struct imsm_super *mpb = super->anchor; |
8482 | struct imsm_map *map; | |
272906ef DW |
8483 | unsigned long long pos; |
8484 | struct mdinfo *d; | |
8485 | struct extent *ex; | |
a20d2ba5 | 8486 | int i, j; |
272906ef | 8487 | int found; |
569cc43f DW |
8488 | __u32 array_start = 0; |
8489 | __u32 array_end = 0; | |
272906ef | 8490 | struct dl *dl; |
6c932028 | 8491 | struct mdinfo *test_list; |
272906ef DW |
8492 | |
8493 | for (dl = super->disks; dl; dl = dl->next) { | |
8494 | /* If in this array, skip */ | |
8495 | for (d = a->info.devs ; d ; d = d->next) | |
e553d2a4 DW |
8496 | if (d->state_fd >= 0 && |
8497 | d->disk.major == dl->major && | |
272906ef | 8498 | d->disk.minor == dl->minor) { |
8ba77d32 AK |
8499 | dprintf("%x:%x already in array\n", |
8500 | dl->major, dl->minor); | |
272906ef DW |
8501 | break; |
8502 | } | |
8503 | if (d) | |
8504 | continue; | |
6c932028 AK |
8505 | test_list = additional_test_list; |
8506 | while (test_list) { | |
8507 | if (test_list->disk.major == dl->major && | |
8508 | test_list->disk.minor == dl->minor) { | |
8ba77d32 AK |
8509 | dprintf("%x:%x already in additional test list\n", |
8510 | dl->major, dl->minor); | |
8511 | break; | |
8512 | } | |
6c932028 | 8513 | test_list = test_list->next; |
8ba77d32 | 8514 | } |
6c932028 | 8515 | if (test_list) |
8ba77d32 | 8516 | continue; |
272906ef | 8517 | |
e553d2a4 | 8518 | /* skip in use or failed drives */ |
25ed7e59 | 8519 | if (is_failed(&dl->disk) || idx == dl->index || |
df474657 DW |
8520 | dl->index == -2) { |
8521 | dprintf("%x:%x status (failed: %d index: %d)\n", | |
25ed7e59 | 8522 | dl->major, dl->minor, is_failed(&dl->disk), idx); |
9a1608e5 DW |
8523 | continue; |
8524 | } | |
8525 | ||
a20d2ba5 DW |
8526 | /* skip pure spares when we are looking for partially |
8527 | * assimilated drives | |
8528 | */ | |
8529 | if (dl->index == -1 && !activate_new) | |
8530 | continue; | |
8531 | ||
f2cc4f7d AO |
8532 | if (!drive_validate_sector_size(super, dl)) |
8533 | continue; | |
8534 | ||
272906ef | 8535 | /* Does this unused device have the requisite free space? |
a20d2ba5 | 8536 | * It needs to be able to cover all member volumes |
272906ef DW |
8537 | */ |
8538 | ex = get_extents(super, dl); | |
8539 | if (!ex) { | |
8540 | dprintf("cannot get extents\n"); | |
8541 | continue; | |
8542 | } | |
a20d2ba5 DW |
8543 | for (i = 0; i < mpb->num_raid_devs; i++) { |
8544 | dev = get_imsm_dev(super, i); | |
238c0a71 | 8545 | map = get_imsm_map(dev, MAP_0); |
272906ef | 8546 | |
a20d2ba5 DW |
8547 | /* check if this disk is already a member of |
8548 | * this array | |
272906ef | 8549 | */ |
620b1713 | 8550 | if (get_imsm_disk_slot(map, dl->index) >= 0) |
a20d2ba5 DW |
8551 | continue; |
8552 | ||
8553 | found = 0; | |
8554 | j = 0; | |
8555 | pos = 0; | |
5551b113 | 8556 | array_start = pba_of_lba0(map); |
329c8278 | 8557 | array_end = array_start + |
5551b113 | 8558 | blocks_per_member(map) - 1; |
a20d2ba5 DW |
8559 | |
8560 | do { | |
8561 | /* check that we can start at pba_of_lba0 with | |
8562 | * blocks_per_member of space | |
8563 | */ | |
329c8278 | 8564 | if (array_start >= pos && array_end < ex[j].start) { |
a20d2ba5 DW |
8565 | found = 1; |
8566 | break; | |
8567 | } | |
8568 | pos = ex[j].start + ex[j].size; | |
8569 | j++; | |
8570 | } while (ex[j-1].size); | |
8571 | ||
8572 | if (!found) | |
272906ef | 8573 | break; |
a20d2ba5 | 8574 | } |
272906ef DW |
8575 | |
8576 | free(ex); | |
a20d2ba5 | 8577 | if (i < mpb->num_raid_devs) { |
329c8278 DW |
8578 | dprintf("%x:%x does not have %u to %u available\n", |
8579 | dl->major, dl->minor, array_start, array_end); | |
272906ef DW |
8580 | /* No room */ |
8581 | continue; | |
a20d2ba5 DW |
8582 | } |
8583 | return dl; | |
272906ef DW |
8584 | } |
8585 | ||
8586 | return dl; | |
8587 | } | |
8588 | ||
95d07a2c LM |
8589 | static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed) |
8590 | { | |
8591 | struct imsm_dev *dev2; | |
8592 | struct imsm_map *map; | |
8593 | struct dl *idisk; | |
8594 | int slot; | |
8595 | int idx; | |
8596 | __u8 state; | |
8597 | ||
8598 | dev2 = get_imsm_dev(cont->sb, dev_idx); | |
8599 | if (dev2) { | |
238c0a71 | 8600 | state = imsm_check_degraded(cont->sb, dev2, failed, MAP_0); |
95d07a2c | 8601 | if (state == IMSM_T_STATE_FAILED) { |
238c0a71 | 8602 | map = get_imsm_map(dev2, MAP_0); |
95d07a2c LM |
8603 | if (!map) |
8604 | return 1; | |
8605 | for (slot = 0; slot < map->num_members; slot++) { | |
8606 | /* | |
8607 | * Check if failed disks are deleted from intel | |
8608 | * disk list or are marked to be deleted | |
8609 | */ | |
238c0a71 | 8610 | idx = get_imsm_disk_idx(dev2, slot, MAP_X); |
95d07a2c LM |
8611 | idisk = get_imsm_dl_disk(cont->sb, idx); |
8612 | /* | |
8613 | * Do not rebuild the array if failed disks | |
8614 | * from failed sub-array are not removed from | |
8615 | * container. | |
8616 | */ | |
8617 | if (idisk && | |
8618 | is_failed(&idisk->disk) && | |
8619 | (idisk->action != DISK_REMOVE)) | |
8620 | return 0; | |
8621 | } | |
8622 | } | |
8623 | } | |
8624 | return 1; | |
8625 | } | |
8626 | ||
88758e9d DW |
8627 | static struct mdinfo *imsm_activate_spare(struct active_array *a, |
8628 | struct metadata_update **updates) | |
8629 | { | |
8630 | /** | |
d23fe947 DW |
8631 | * Find a device with unused free space and use it to replace a |
8632 | * failed/vacant region in an array. We replace failed regions one a | |
8633 | * array at a time. The result is that a new spare disk will be added | |
8634 | * to the first failed array and after the monitor has finished | |
8635 | * propagating failures the remainder will be consumed. | |
88758e9d | 8636 | * |
d23fe947 DW |
8637 | * FIXME add a capability for mdmon to request spares from another |
8638 | * container. | |
88758e9d DW |
8639 | */ |
8640 | ||
8641 | struct intel_super *super = a->container->sb; | |
88758e9d | 8642 | int inst = a->info.container_member; |
949c47a0 | 8643 | struct imsm_dev *dev = get_imsm_dev(super, inst); |
238c0a71 | 8644 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
88758e9d DW |
8645 | int failed = a->info.array.raid_disks; |
8646 | struct mdinfo *rv = NULL; | |
8647 | struct mdinfo *d; | |
8648 | struct mdinfo *di; | |
8649 | struct metadata_update *mu; | |
8650 | struct dl *dl; | |
8651 | struct imsm_update_activate_spare *u; | |
8652 | int num_spares = 0; | |
8653 | int i; | |
95d07a2c | 8654 | int allowed; |
88758e9d DW |
8655 | |
8656 | for (d = a->info.devs ; d ; d = d->next) { | |
8657 | if ((d->curr_state & DS_FAULTY) && | |
8658 | d->state_fd >= 0) | |
8659 | /* wait for Removal to happen */ | |
8660 | return NULL; | |
8661 | if (d->state_fd >= 0) | |
8662 | failed--; | |
8663 | } | |
8664 | ||
8665 | dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n", | |
8666 | inst, failed, a->info.array.raid_disks, a->info.array.level); | |
1af97990 | 8667 | |
e2962bfc AK |
8668 | if (imsm_reshape_blocks_arrays_changes(super)) |
8669 | return NULL; | |
1af97990 | 8670 | |
fc8ca064 AK |
8671 | /* Cannot activate another spare if rebuild is in progress already |
8672 | */ | |
8673 | if (is_rebuilding(dev)) { | |
7a862a02 | 8674 | dprintf("imsm: No spare activation allowed. Rebuild in progress already.\n"); |
fc8ca064 AK |
8675 | return NULL; |
8676 | } | |
8677 | ||
89c67882 AK |
8678 | if (a->info.array.level == 4) |
8679 | /* No repair for takeovered array | |
8680 | * imsm doesn't support raid4 | |
8681 | */ | |
8682 | return NULL; | |
8683 | ||
3b451610 AK |
8684 | if (imsm_check_degraded(super, dev, failed, MAP_0) != |
8685 | IMSM_T_STATE_DEGRADED) | |
88758e9d DW |
8686 | return NULL; |
8687 | ||
83ca7d45 AP |
8688 | if (get_imsm_map(dev, MAP_0)->map_state == IMSM_T_STATE_UNINITIALIZED) { |
8689 | dprintf("imsm: No spare activation allowed. Volume is not initialized.\n"); | |
8690 | return NULL; | |
8691 | } | |
8692 | ||
95d07a2c LM |
8693 | /* |
8694 | * If there are any failed disks check state of the other volume. | |
8695 | * Block rebuild if the another one is failed until failed disks | |
8696 | * are removed from container. | |
8697 | */ | |
8698 | if (failed) { | |
7a862a02 | 8699 | dprintf("found failed disks in %.*s, check if there anotherfailed sub-array.\n", |
c4acd1e5 | 8700 | MAX_RAID_SERIAL_LEN, dev->volume); |
95d07a2c LM |
8701 | /* check if states of the other volumes allow for rebuild */ |
8702 | for (i = 0; i < super->anchor->num_raid_devs; i++) { | |
8703 | if (i != inst) { | |
8704 | allowed = imsm_rebuild_allowed(a->container, | |
8705 | i, failed); | |
8706 | if (!allowed) | |
8707 | return NULL; | |
8708 | } | |
8709 | } | |
8710 | } | |
8711 | ||
88758e9d | 8712 | /* For each slot, if it is not working, find a spare */ |
88758e9d DW |
8713 | for (i = 0; i < a->info.array.raid_disks; i++) { |
8714 | for (d = a->info.devs ; d ; d = d->next) | |
8715 | if (d->disk.raid_disk == i) | |
8716 | break; | |
8717 | dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0); | |
8718 | if (d && (d->state_fd >= 0)) | |
8719 | continue; | |
8720 | ||
272906ef | 8721 | /* |
a20d2ba5 DW |
8722 | * OK, this device needs recovery. Try to re-add the |
8723 | * previous occupant of this slot, if this fails see if | |
8724 | * we can continue the assimilation of a spare that was | |
8725 | * partially assimilated, finally try to activate a new | |
8726 | * spare. | |
272906ef DW |
8727 | */ |
8728 | dl = imsm_readd(super, i, a); | |
8729 | if (!dl) | |
b303fe21 | 8730 | dl = imsm_add_spare(super, i, a, 0, rv); |
a20d2ba5 | 8731 | if (!dl) |
b303fe21 | 8732 | dl = imsm_add_spare(super, i, a, 1, rv); |
272906ef DW |
8733 | if (!dl) |
8734 | continue; | |
1011e834 | 8735 | |
272906ef | 8736 | /* found a usable disk with enough space */ |
503975b9 | 8737 | di = xcalloc(1, sizeof(*di)); |
272906ef DW |
8738 | |
8739 | /* dl->index will be -1 in the case we are activating a | |
8740 | * pristine spare. imsm_process_update() will create a | |
8741 | * new index in this case. Once a disk is found to be | |
8742 | * failed in all member arrays it is kicked from the | |
8743 | * metadata | |
8744 | */ | |
8745 | di->disk.number = dl->index; | |
d23fe947 | 8746 | |
272906ef DW |
8747 | /* (ab)use di->devs to store a pointer to the device |
8748 | * we chose | |
8749 | */ | |
8750 | di->devs = (struct mdinfo *) dl; | |
8751 | ||
8752 | di->disk.raid_disk = i; | |
8753 | di->disk.major = dl->major; | |
8754 | di->disk.minor = dl->minor; | |
8755 | di->disk.state = 0; | |
d23534e4 | 8756 | di->recovery_start = 0; |
5551b113 | 8757 | di->data_offset = pba_of_lba0(map); |
272906ef DW |
8758 | di->component_size = a->info.component_size; |
8759 | di->container_member = inst; | |
5e46202e | 8760 | di->bb.supported = 1; |
2432ce9b AP |
8761 | if (dev->rwh_policy == RWH_DISTRIBUTED) { |
8762 | di->consistency_policy = CONSISTENCY_POLICY_PPL; | |
8763 | di->ppl_sector = get_ppl_sector(super, inst); | |
8764 | di->ppl_size = (PPL_HEADER_SIZE + PPL_ENTRY_SPACE) >> 9; | |
8765 | } | |
148acb7b | 8766 | super->random = random32(); |
272906ef DW |
8767 | di->next = rv; |
8768 | rv = di; | |
8769 | num_spares++; | |
8770 | dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor, | |
8771 | i, di->data_offset); | |
88758e9d DW |
8772 | } |
8773 | ||
8774 | if (!rv) | |
8775 | /* No spares found */ | |
8776 | return rv; | |
8777 | /* Now 'rv' has a list of devices to return. | |
8778 | * Create a metadata_update record to update the | |
8779 | * disk_ord_tbl for the array | |
8780 | */ | |
503975b9 | 8781 | mu = xmalloc(sizeof(*mu)); |
1011e834 | 8782 | mu->buf = xcalloc(num_spares, |
503975b9 | 8783 | sizeof(struct imsm_update_activate_spare)); |
88758e9d | 8784 | mu->space = NULL; |
cb23f1f4 | 8785 | mu->space_list = NULL; |
88758e9d DW |
8786 | mu->len = sizeof(struct imsm_update_activate_spare) * num_spares; |
8787 | mu->next = *updates; | |
8788 | u = (struct imsm_update_activate_spare *) mu->buf; | |
8789 | ||
8790 | for (di = rv ; di ; di = di->next) { | |
8791 | u->type = update_activate_spare; | |
d23fe947 DW |
8792 | u->dl = (struct dl *) di->devs; |
8793 | di->devs = NULL; | |
88758e9d DW |
8794 | u->slot = di->disk.raid_disk; |
8795 | u->array = inst; | |
8796 | u->next = u + 1; | |
8797 | u++; | |
8798 | } | |
8799 | (u-1)->next = NULL; | |
8800 | *updates = mu; | |
8801 | ||
8802 | return rv; | |
8803 | } | |
8804 | ||
54c2c1ea | 8805 | static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u) |
8273f55e | 8806 | { |
54c2c1ea | 8807 | struct imsm_dev *dev = get_imsm_dev(super, idx); |
238c0a71 AK |
8808 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
8809 | struct imsm_map *new_map = get_imsm_map(&u->dev, MAP_0); | |
54c2c1ea DW |
8810 | struct disk_info *inf = get_disk_info(u); |
8811 | struct imsm_disk *disk; | |
8273f55e DW |
8812 | int i; |
8813 | int j; | |
8273f55e | 8814 | |
54c2c1ea | 8815 | for (i = 0; i < map->num_members; i++) { |
238c0a71 | 8816 | disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, MAP_X)); |
54c2c1ea DW |
8817 | for (j = 0; j < new_map->num_members; j++) |
8818 | if (serialcmp(disk->serial, inf[j].serial) == 0) | |
8273f55e DW |
8819 | return 1; |
8820 | } | |
8821 | ||
8822 | return 0; | |
8823 | } | |
8824 | ||
1a64be56 LM |
8825 | static struct dl *get_disk_super(struct intel_super *super, int major, int minor) |
8826 | { | |
594dc1b8 JS |
8827 | struct dl *dl; |
8828 | ||
1a64be56 | 8829 | for (dl = super->disks; dl; dl = dl->next) |
089f9d79 | 8830 | if (dl->major == major && dl->minor == minor) |
1a64be56 LM |
8831 | return dl; |
8832 | return NULL; | |
8833 | } | |
8834 | ||
8835 | static int remove_disk_super(struct intel_super *super, int major, int minor) | |
8836 | { | |
594dc1b8 | 8837 | struct dl *prev; |
1a64be56 LM |
8838 | struct dl *dl; |
8839 | ||
8840 | prev = NULL; | |
8841 | for (dl = super->disks; dl; dl = dl->next) { | |
089f9d79 | 8842 | if (dl->major == major && dl->minor == minor) { |
1a64be56 LM |
8843 | /* remove */ |
8844 | if (prev) | |
8845 | prev->next = dl->next; | |
8846 | else | |
8847 | super->disks = dl->next; | |
8848 | dl->next = NULL; | |
8849 | __free_imsm_disk(dl); | |
1ade5cc1 | 8850 | dprintf("removed %x:%x\n", major, minor); |
1a64be56 LM |
8851 | break; |
8852 | } | |
8853 | prev = dl; | |
8854 | } | |
8855 | return 0; | |
8856 | } | |
8857 | ||
f21e18ca | 8858 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index); |
ae6aad82 | 8859 | |
1a64be56 LM |
8860 | static int add_remove_disk_update(struct intel_super *super) |
8861 | { | |
8862 | int check_degraded = 0; | |
594dc1b8 JS |
8863 | struct dl *disk; |
8864 | ||
1a64be56 LM |
8865 | /* add/remove some spares to/from the metadata/contrainer */ |
8866 | while (super->disk_mgmt_list) { | |
8867 | struct dl *disk_cfg; | |
8868 | ||
8869 | disk_cfg = super->disk_mgmt_list; | |
8870 | super->disk_mgmt_list = disk_cfg->next; | |
8871 | disk_cfg->next = NULL; | |
8872 | ||
8873 | if (disk_cfg->action == DISK_ADD) { | |
8874 | disk_cfg->next = super->disks; | |
8875 | super->disks = disk_cfg; | |
8876 | check_degraded = 1; | |
1ade5cc1 N |
8877 | dprintf("added %x:%x\n", |
8878 | disk_cfg->major, disk_cfg->minor); | |
1a64be56 LM |
8879 | } else if (disk_cfg->action == DISK_REMOVE) { |
8880 | dprintf("Disk remove action processed: %x.%x\n", | |
8881 | disk_cfg->major, disk_cfg->minor); | |
8882 | disk = get_disk_super(super, | |
8883 | disk_cfg->major, | |
8884 | disk_cfg->minor); | |
8885 | if (disk) { | |
8886 | /* store action status */ | |
8887 | disk->action = DISK_REMOVE; | |
8888 | /* remove spare disks only */ | |
8889 | if (disk->index == -1) { | |
8890 | remove_disk_super(super, | |
8891 | disk_cfg->major, | |
8892 | disk_cfg->minor); | |
8893 | } | |
8894 | } | |
8895 | /* release allocate disk structure */ | |
8896 | __free_imsm_disk(disk_cfg); | |
8897 | } | |
8898 | } | |
8899 | return check_degraded; | |
8900 | } | |
8901 | ||
a29911da PC |
8902 | static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u, |
8903 | struct intel_super *super, | |
8904 | void ***space_list) | |
8905 | { | |
8906 | struct intel_dev *id; | |
8907 | void **tofree = NULL; | |
8908 | int ret_val = 0; | |
8909 | ||
1ade5cc1 | 8910 | dprintf("(enter)\n"); |
089f9d79 | 8911 | if (u->subdev < 0 || u->subdev > 1) { |
a29911da PC |
8912 | dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev); |
8913 | return ret_val; | |
8914 | } | |
089f9d79 | 8915 | if (space_list == NULL || *space_list == NULL) { |
a29911da PC |
8916 | dprintf("imsm: Error: Memory is not allocated\n"); |
8917 | return ret_val; | |
8918 | } | |
8919 | ||
8920 | for (id = super->devlist ; id; id = id->next) { | |
8921 | if (id->index == (unsigned)u->subdev) { | |
8922 | struct imsm_dev *dev = get_imsm_dev(super, u->subdev); | |
8923 | struct imsm_map *map; | |
8924 | struct imsm_dev *new_dev = | |
8925 | (struct imsm_dev *)*space_list; | |
238c0a71 | 8926 | struct imsm_map *migr_map = get_imsm_map(dev, MAP_1); |
a29911da PC |
8927 | int to_state; |
8928 | struct dl *new_disk; | |
8929 | ||
8930 | if (new_dev == NULL) | |
8931 | return ret_val; | |
8932 | *space_list = **space_list; | |
8933 | memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0)); | |
238c0a71 | 8934 | map = get_imsm_map(new_dev, MAP_0); |
a29911da PC |
8935 | if (migr_map) { |
8936 | dprintf("imsm: Error: migration in progress"); | |
8937 | return ret_val; | |
8938 | } | |
8939 | ||
8940 | to_state = map->map_state; | |
8941 | if ((u->new_level == 5) && (map->raid_level == 0)) { | |
8942 | map->num_members++; | |
8943 | /* this should not happen */ | |
8944 | if (u->new_disks[0] < 0) { | |
8945 | map->failed_disk_num = | |
8946 | map->num_members - 1; | |
8947 | to_state = IMSM_T_STATE_DEGRADED; | |
8948 | } else | |
8949 | to_state = IMSM_T_STATE_NORMAL; | |
8950 | } | |
8e59f3d8 | 8951 | migrate(new_dev, super, to_state, MIGR_GEN_MIGR); |
a29911da PC |
8952 | if (u->new_level > -1) |
8953 | map->raid_level = u->new_level; | |
238c0a71 | 8954 | migr_map = get_imsm_map(new_dev, MAP_1); |
a29911da PC |
8955 | if ((u->new_level == 5) && |
8956 | (migr_map->raid_level == 0)) { | |
8957 | int ord = map->num_members - 1; | |
8958 | migr_map->num_members--; | |
8959 | if (u->new_disks[0] < 0) | |
8960 | ord |= IMSM_ORD_REBUILD; | |
8961 | set_imsm_ord_tbl_ent(map, | |
8962 | map->num_members - 1, | |
8963 | ord); | |
8964 | } | |
8965 | id->dev = new_dev; | |
8966 | tofree = (void **)dev; | |
8967 | ||
4bba0439 PC |
8968 | /* update chunk size |
8969 | */ | |
06fb291a PB |
8970 | if (u->new_chunksize > 0) { |
8971 | unsigned long long num_data_stripes; | |
8972 | int used_disks = | |
8973 | imsm_num_data_members(dev, MAP_0); | |
8974 | ||
8975 | if (used_disks == 0) | |
8976 | return ret_val; | |
8977 | ||
4bba0439 PC |
8978 | map->blocks_per_strip = |
8979 | __cpu_to_le16(u->new_chunksize * 2); | |
06fb291a PB |
8980 | num_data_stripes = |
8981 | (join_u32(dev->size_low, dev->size_high) | |
8982 | / used_disks); | |
8983 | num_data_stripes /= map->blocks_per_strip; | |
8984 | num_data_stripes /= map->num_domains; | |
8985 | set_num_data_stripes(map, num_data_stripes); | |
8986 | } | |
4bba0439 | 8987 | |
a29911da PC |
8988 | /* add disk |
8989 | */ | |
089f9d79 JS |
8990 | if (u->new_level != 5 || migr_map->raid_level != 0 || |
8991 | migr_map->raid_level == map->raid_level) | |
a29911da PC |
8992 | goto skip_disk_add; |
8993 | ||
8994 | if (u->new_disks[0] >= 0) { | |
8995 | /* use passes spare | |
8996 | */ | |
8997 | new_disk = get_disk_super(super, | |
8998 | major(u->new_disks[0]), | |
8999 | minor(u->new_disks[0])); | |
7a862a02 | 9000 | dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n", |
a29911da PC |
9001 | major(u->new_disks[0]), |
9002 | minor(u->new_disks[0]), | |
9003 | new_disk, new_disk->index); | |
9004 | if (new_disk == NULL) | |
9005 | goto error_disk_add; | |
9006 | ||
9007 | new_disk->index = map->num_members - 1; | |
9008 | /* slot to fill in autolayout | |
9009 | */ | |
9010 | new_disk->raiddisk = new_disk->index; | |
9011 | new_disk->disk.status |= CONFIGURED_DISK; | |
9012 | new_disk->disk.status &= ~SPARE_DISK; | |
9013 | } else | |
9014 | goto error_disk_add; | |
9015 | ||
9016 | skip_disk_add: | |
9017 | *tofree = *space_list; | |
9018 | /* calculate new size | |
9019 | */ | |
f3871fdc | 9020 | imsm_set_array_size(new_dev, -1); |
a29911da PC |
9021 | |
9022 | ret_val = 1; | |
9023 | } | |
9024 | } | |
9025 | ||
9026 | if (tofree) | |
9027 | *space_list = tofree; | |
9028 | return ret_val; | |
9029 | ||
9030 | error_disk_add: | |
9031 | dprintf("Error: imsm: Cannot find disk.\n"); | |
9032 | return ret_val; | |
9033 | } | |
9034 | ||
f3871fdc AK |
9035 | static int apply_size_change_update(struct imsm_update_size_change *u, |
9036 | struct intel_super *super) | |
9037 | { | |
9038 | struct intel_dev *id; | |
9039 | int ret_val = 0; | |
9040 | ||
1ade5cc1 | 9041 | dprintf("(enter)\n"); |
089f9d79 | 9042 | if (u->subdev < 0 || u->subdev > 1) { |
f3871fdc AK |
9043 | dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev); |
9044 | return ret_val; | |
9045 | } | |
9046 | ||
9047 | for (id = super->devlist ; id; id = id->next) { | |
9048 | if (id->index == (unsigned)u->subdev) { | |
9049 | struct imsm_dev *dev = get_imsm_dev(super, u->subdev); | |
9050 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
9051 | int used_disks = imsm_num_data_members(dev, MAP_0); | |
9052 | unsigned long long blocks_per_member; | |
06fb291a | 9053 | unsigned long long num_data_stripes; |
f3871fdc AK |
9054 | |
9055 | /* calculate new size | |
9056 | */ | |
9057 | blocks_per_member = u->new_size / used_disks; | |
06fb291a PB |
9058 | num_data_stripes = blocks_per_member / |
9059 | map->blocks_per_strip; | |
9060 | num_data_stripes /= map->num_domains; | |
9061 | dprintf("(size: %llu, blocks per member: %llu, num_data_stipes: %llu)\n", | |
9062 | u->new_size, blocks_per_member, | |
9063 | num_data_stripes); | |
f3871fdc | 9064 | set_blocks_per_member(map, blocks_per_member); |
06fb291a | 9065 | set_num_data_stripes(map, num_data_stripes); |
f3871fdc AK |
9066 | imsm_set_array_size(dev, u->new_size); |
9067 | ||
9068 | ret_val = 1; | |
9069 | break; | |
9070 | } | |
9071 | } | |
9072 | ||
9073 | return ret_val; | |
9074 | } | |
9075 | ||
061d7da3 | 9076 | static int apply_update_activate_spare(struct imsm_update_activate_spare *u, |
ca9de185 | 9077 | struct intel_super *super, |
061d7da3 LO |
9078 | struct active_array *active_array) |
9079 | { | |
9080 | struct imsm_super *mpb = super->anchor; | |
9081 | struct imsm_dev *dev = get_imsm_dev(super, u->array); | |
238c0a71 | 9082 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
061d7da3 LO |
9083 | struct imsm_map *migr_map; |
9084 | struct active_array *a; | |
9085 | struct imsm_disk *disk; | |
9086 | __u8 to_state; | |
9087 | struct dl *dl; | |
9088 | unsigned int found; | |
9089 | int failed; | |
5961eeec | 9090 | int victim; |
061d7da3 | 9091 | int i; |
5961eeec | 9092 | int second_map_created = 0; |
061d7da3 | 9093 | |
5961eeec | 9094 | for (; u; u = u->next) { |
238c0a71 | 9095 | victim = get_imsm_disk_idx(dev, u->slot, MAP_X); |
061d7da3 | 9096 | |
5961eeec | 9097 | if (victim < 0) |
9098 | return 0; | |
061d7da3 | 9099 | |
5961eeec | 9100 | for (dl = super->disks; dl; dl = dl->next) |
9101 | if (dl == u->dl) | |
9102 | break; | |
061d7da3 | 9103 | |
5961eeec | 9104 | if (!dl) { |
7a862a02 | 9105 | pr_err("error: imsm_activate_spare passed an unknown disk (index: %d)\n", |
5961eeec | 9106 | u->dl->index); |
9107 | return 0; | |
9108 | } | |
061d7da3 | 9109 | |
5961eeec | 9110 | /* count failures (excluding rebuilds and the victim) |
9111 | * to determine map[0] state | |
9112 | */ | |
9113 | failed = 0; | |
9114 | for (i = 0; i < map->num_members; i++) { | |
9115 | if (i == u->slot) | |
9116 | continue; | |
9117 | disk = get_imsm_disk(super, | |
238c0a71 | 9118 | get_imsm_disk_idx(dev, i, MAP_X)); |
5961eeec | 9119 | if (!disk || is_failed(disk)) |
9120 | failed++; | |
9121 | } | |
061d7da3 | 9122 | |
5961eeec | 9123 | /* adding a pristine spare, assign a new index */ |
9124 | if (dl->index < 0) { | |
9125 | dl->index = super->anchor->num_disks; | |
9126 | super->anchor->num_disks++; | |
9127 | } | |
9128 | disk = &dl->disk; | |
9129 | disk->status |= CONFIGURED_DISK; | |
9130 | disk->status &= ~SPARE_DISK; | |
9131 | ||
9132 | /* mark rebuild */ | |
238c0a71 | 9133 | to_state = imsm_check_degraded(super, dev, failed, MAP_0); |
5961eeec | 9134 | if (!second_map_created) { |
9135 | second_map_created = 1; | |
9136 | map->map_state = IMSM_T_STATE_DEGRADED; | |
9137 | migrate(dev, super, to_state, MIGR_REBUILD); | |
9138 | } else | |
9139 | map->map_state = to_state; | |
238c0a71 | 9140 | migr_map = get_imsm_map(dev, MAP_1); |
5961eeec | 9141 | set_imsm_ord_tbl_ent(map, u->slot, dl->index); |
9142 | set_imsm_ord_tbl_ent(migr_map, u->slot, | |
9143 | dl->index | IMSM_ORD_REBUILD); | |
9144 | ||
9145 | /* update the family_num to mark a new container | |
9146 | * generation, being careful to record the existing | |
9147 | * family_num in orig_family_num to clean up after | |
9148 | * earlier mdadm versions that neglected to set it. | |
9149 | */ | |
9150 | if (mpb->orig_family_num == 0) | |
9151 | mpb->orig_family_num = mpb->family_num; | |
9152 | mpb->family_num += super->random; | |
9153 | ||
9154 | /* count arrays using the victim in the metadata */ | |
9155 | found = 0; | |
9156 | for (a = active_array; a ; a = a->next) { | |
9157 | dev = get_imsm_dev(super, a->info.container_member); | |
238c0a71 | 9158 | map = get_imsm_map(dev, MAP_0); |
061d7da3 | 9159 | |
5961eeec | 9160 | if (get_imsm_disk_slot(map, victim) >= 0) |
9161 | found++; | |
9162 | } | |
061d7da3 | 9163 | |
5961eeec | 9164 | /* delete the victim if it is no longer being |
9165 | * utilized anywhere | |
061d7da3 | 9166 | */ |
5961eeec | 9167 | if (!found) { |
9168 | struct dl **dlp; | |
061d7da3 | 9169 | |
5961eeec | 9170 | /* We know that 'manager' isn't touching anything, |
9171 | * so it is safe to delete | |
9172 | */ | |
9173 | for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next) | |
061d7da3 LO |
9174 | if ((*dlp)->index == victim) |
9175 | break; | |
5961eeec | 9176 | |
9177 | /* victim may be on the missing list */ | |
9178 | if (!*dlp) | |
9179 | for (dlp = &super->missing; *dlp; | |
9180 | dlp = &(*dlp)->next) | |
9181 | if ((*dlp)->index == victim) | |
9182 | break; | |
9183 | imsm_delete(super, dlp, victim); | |
9184 | } | |
061d7da3 LO |
9185 | } |
9186 | ||
9187 | return 1; | |
9188 | } | |
a29911da | 9189 | |
2e5dc010 N |
9190 | static int apply_reshape_container_disks_update(struct imsm_update_reshape *u, |
9191 | struct intel_super *super, | |
9192 | void ***space_list) | |
9193 | { | |
9194 | struct dl *new_disk; | |
9195 | struct intel_dev *id; | |
9196 | int i; | |
9197 | int delta_disks = u->new_raid_disks - u->old_raid_disks; | |
ee4beede | 9198 | int disk_count = u->old_raid_disks; |
2e5dc010 N |
9199 | void **tofree = NULL; |
9200 | int devices_to_reshape = 1; | |
9201 | struct imsm_super *mpb = super->anchor; | |
9202 | int ret_val = 0; | |
d098291a | 9203 | unsigned int dev_id; |
2e5dc010 | 9204 | |
1ade5cc1 | 9205 | dprintf("(enter)\n"); |
2e5dc010 N |
9206 | |
9207 | /* enable spares to use in array */ | |
9208 | for (i = 0; i < delta_disks; i++) { | |
9209 | new_disk = get_disk_super(super, | |
9210 | major(u->new_disks[i]), | |
9211 | minor(u->new_disks[i])); | |
7a862a02 | 9212 | dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n", |
2e5dc010 N |
9213 | major(u->new_disks[i]), minor(u->new_disks[i]), |
9214 | new_disk, new_disk->index); | |
089f9d79 JS |
9215 | if (new_disk == NULL || |
9216 | (new_disk->index >= 0 && | |
9217 | new_disk->index < u->old_raid_disks)) | |
2e5dc010 | 9218 | goto update_reshape_exit; |
ee4beede | 9219 | new_disk->index = disk_count++; |
2e5dc010 N |
9220 | /* slot to fill in autolayout |
9221 | */ | |
9222 | new_disk->raiddisk = new_disk->index; | |
9223 | new_disk->disk.status |= | |
9224 | CONFIGURED_DISK; | |
9225 | new_disk->disk.status &= ~SPARE_DISK; | |
9226 | } | |
9227 | ||
ed7333bd AK |
9228 | dprintf("imsm: volume set mpb->num_raid_devs = %i\n", |
9229 | mpb->num_raid_devs); | |
2e5dc010 N |
9230 | /* manage changes in volume |
9231 | */ | |
d098291a | 9232 | for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) { |
2e5dc010 N |
9233 | void **sp = *space_list; |
9234 | struct imsm_dev *newdev; | |
9235 | struct imsm_map *newmap, *oldmap; | |
9236 | ||
d098291a AK |
9237 | for (id = super->devlist ; id; id = id->next) { |
9238 | if (id->index == dev_id) | |
9239 | break; | |
9240 | } | |
9241 | if (id == NULL) | |
9242 | break; | |
2e5dc010 N |
9243 | if (!sp) |
9244 | continue; | |
9245 | *space_list = *sp; | |
9246 | newdev = (void*)sp; | |
9247 | /* Copy the dev, but not (all of) the map */ | |
9248 | memcpy(newdev, id->dev, sizeof(*newdev)); | |
238c0a71 AK |
9249 | oldmap = get_imsm_map(id->dev, MAP_0); |
9250 | newmap = get_imsm_map(newdev, MAP_0); | |
2e5dc010 N |
9251 | /* Copy the current map */ |
9252 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); | |
9253 | /* update one device only | |
9254 | */ | |
9255 | if (devices_to_reshape) { | |
ed7333bd AK |
9256 | dprintf("imsm: modifying subdev: %i\n", |
9257 | id->index); | |
2e5dc010 N |
9258 | devices_to_reshape--; |
9259 | newdev->vol.migr_state = 1; | |
9260 | newdev->vol.curr_migr_unit = 0; | |
ea672ee1 | 9261 | set_migr_type(newdev, MIGR_GEN_MIGR); |
2e5dc010 N |
9262 | newmap->num_members = u->new_raid_disks; |
9263 | for (i = 0; i < delta_disks; i++) { | |
9264 | set_imsm_ord_tbl_ent(newmap, | |
9265 | u->old_raid_disks + i, | |
9266 | u->old_raid_disks + i); | |
9267 | } | |
9268 | /* New map is correct, now need to save old map | |
9269 | */ | |
238c0a71 | 9270 | newmap = get_imsm_map(newdev, MAP_1); |
2e5dc010 N |
9271 | memcpy(newmap, oldmap, sizeof_imsm_map(oldmap)); |
9272 | ||
f3871fdc | 9273 | imsm_set_array_size(newdev, -1); |
2e5dc010 N |
9274 | } |
9275 | ||
9276 | sp = (void **)id->dev; | |
9277 | id->dev = newdev; | |
9278 | *sp = tofree; | |
9279 | tofree = sp; | |
8e59f3d8 AK |
9280 | |
9281 | /* Clear migration record */ | |
9282 | memset(super->migr_rec, 0, sizeof(struct migr_record)); | |
2e5dc010 | 9283 | } |
819bc634 AK |
9284 | if (tofree) |
9285 | *space_list = tofree; | |
2e5dc010 N |
9286 | ret_val = 1; |
9287 | ||
9288 | update_reshape_exit: | |
9289 | ||
9290 | return ret_val; | |
9291 | } | |
9292 | ||
bb025c2f | 9293 | static int apply_takeover_update(struct imsm_update_takeover *u, |
8ca6df95 KW |
9294 | struct intel_super *super, |
9295 | void ***space_list) | |
bb025c2f KW |
9296 | { |
9297 | struct imsm_dev *dev = NULL; | |
8ca6df95 KW |
9298 | struct intel_dev *dv; |
9299 | struct imsm_dev *dev_new; | |
bb025c2f KW |
9300 | struct imsm_map *map; |
9301 | struct dl *dm, *du; | |
8ca6df95 | 9302 | int i; |
bb025c2f KW |
9303 | |
9304 | for (dv = super->devlist; dv; dv = dv->next) | |
9305 | if (dv->index == (unsigned int)u->subarray) { | |
9306 | dev = dv->dev; | |
9307 | break; | |
9308 | } | |
9309 | ||
9310 | if (dev == NULL) | |
9311 | return 0; | |
9312 | ||
238c0a71 | 9313 | map = get_imsm_map(dev, MAP_0); |
bb025c2f KW |
9314 | |
9315 | if (u->direction == R10_TO_R0) { | |
06fb291a PB |
9316 | unsigned long long num_data_stripes; |
9317 | ||
9318 | map->num_domains = 1; | |
9319 | num_data_stripes = blocks_per_member(map); | |
9320 | num_data_stripes /= map->blocks_per_strip; | |
9321 | num_data_stripes /= map->num_domains; | |
9322 | set_num_data_stripes(map, num_data_stripes); | |
9323 | ||
43d5ec18 | 9324 | /* Number of failed disks must be half of initial disk number */ |
3b451610 AK |
9325 | if (imsm_count_failed(super, dev, MAP_0) != |
9326 | (map->num_members / 2)) | |
43d5ec18 KW |
9327 | return 0; |
9328 | ||
bb025c2f KW |
9329 | /* iterate through devices to mark removed disks as spare */ |
9330 | for (dm = super->disks; dm; dm = dm->next) { | |
9331 | if (dm->disk.status & FAILED_DISK) { | |
9332 | int idx = dm->index; | |
9333 | /* update indexes on the disk list */ | |
9334 | /* FIXME this loop-with-the-loop looks wrong, I'm not convinced | |
9335 | the index values will end up being correct.... NB */ | |
9336 | for (du = super->disks; du; du = du->next) | |
9337 | if (du->index > idx) | |
9338 | du->index--; | |
9339 | /* mark as spare disk */ | |
a8619d23 | 9340 | mark_spare(dm); |
bb025c2f KW |
9341 | } |
9342 | } | |
bb025c2f KW |
9343 | /* update map */ |
9344 | map->num_members = map->num_members / 2; | |
9345 | map->map_state = IMSM_T_STATE_NORMAL; | |
9346 | map->num_domains = 1; | |
9347 | map->raid_level = 0; | |
9348 | map->failed_disk_num = -1; | |
9349 | } | |
9350 | ||
8ca6df95 KW |
9351 | if (u->direction == R0_TO_R10) { |
9352 | void **space; | |
9353 | /* update slots in current disk list */ | |
9354 | for (dm = super->disks; dm; dm = dm->next) { | |
9355 | if (dm->index >= 0) | |
9356 | dm->index *= 2; | |
9357 | } | |
9358 | /* create new *missing* disks */ | |
9359 | for (i = 0; i < map->num_members; i++) { | |
9360 | space = *space_list; | |
9361 | if (!space) | |
9362 | continue; | |
9363 | *space_list = *space; | |
9364 | du = (void *)space; | |
9365 | memcpy(du, super->disks, sizeof(*du)); | |
8ca6df95 KW |
9366 | du->fd = -1; |
9367 | du->minor = 0; | |
9368 | du->major = 0; | |
9369 | du->index = (i * 2) + 1; | |
9370 | sprintf((char *)du->disk.serial, | |
9371 | " MISSING_%d", du->index); | |
9372 | sprintf((char *)du->serial, | |
9373 | "MISSING_%d", du->index); | |
9374 | du->next = super->missing; | |
9375 | super->missing = du; | |
9376 | } | |
9377 | /* create new dev and map */ | |
9378 | space = *space_list; | |
9379 | if (!space) | |
9380 | return 0; | |
9381 | *space_list = *space; | |
9382 | dev_new = (void *)space; | |
9383 | memcpy(dev_new, dev, sizeof(*dev)); | |
9384 | /* update new map */ | |
238c0a71 | 9385 | map = get_imsm_map(dev_new, MAP_0); |
8ca6df95 | 9386 | map->num_members = map->num_members * 2; |
1a2487c2 | 9387 | map->map_state = IMSM_T_STATE_DEGRADED; |
8ca6df95 KW |
9388 | map->num_domains = 2; |
9389 | map->raid_level = 1; | |
9390 | /* replace dev<->dev_new */ | |
9391 | dv->dev = dev_new; | |
9392 | } | |
bb025c2f KW |
9393 | /* update disk order table */ |
9394 | for (du = super->disks; du; du = du->next) | |
9395 | if (du->index >= 0) | |
9396 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
8ca6df95 | 9397 | for (du = super->missing; du; du = du->next) |
1a2487c2 KW |
9398 | if (du->index >= 0) { |
9399 | set_imsm_ord_tbl_ent(map, du->index, du->index); | |
4c9e8c1e | 9400 | mark_missing(super, dv->dev, &du->disk, du->index); |
1a2487c2 | 9401 | } |
bb025c2f KW |
9402 | |
9403 | return 1; | |
9404 | } | |
9405 | ||
e8319a19 DW |
9406 | static void imsm_process_update(struct supertype *st, |
9407 | struct metadata_update *update) | |
9408 | { | |
9409 | /** | |
9410 | * crack open the metadata_update envelope to find the update record | |
9411 | * update can be one of: | |
d195167d AK |
9412 | * update_reshape_container_disks - all the arrays in the container |
9413 | * are being reshaped to have more devices. We need to mark | |
9414 | * the arrays for general migration and convert selected spares | |
9415 | * into active devices. | |
9416 | * update_activate_spare - a spare device has replaced a failed | |
1011e834 N |
9417 | * device in an array, update the disk_ord_tbl. If this disk is |
9418 | * present in all member arrays then also clear the SPARE_DISK | |
9419 | * flag | |
d195167d AK |
9420 | * update_create_array |
9421 | * update_kill_array | |
9422 | * update_rename_array | |
9423 | * update_add_remove_disk | |
e8319a19 DW |
9424 | */ |
9425 | struct intel_super *super = st->sb; | |
4d7b1503 | 9426 | struct imsm_super *mpb; |
e8319a19 DW |
9427 | enum imsm_update_type type = *(enum imsm_update_type *) update->buf; |
9428 | ||
4d7b1503 DW |
9429 | /* update requires a larger buf but the allocation failed */ |
9430 | if (super->next_len && !super->next_buf) { | |
9431 | super->next_len = 0; | |
9432 | return; | |
9433 | } | |
9434 | ||
9435 | if (super->next_buf) { | |
9436 | memcpy(super->next_buf, super->buf, super->len); | |
9437 | free(super->buf); | |
9438 | super->len = super->next_len; | |
9439 | super->buf = super->next_buf; | |
9440 | ||
9441 | super->next_len = 0; | |
9442 | super->next_buf = NULL; | |
9443 | } | |
9444 | ||
9445 | mpb = super->anchor; | |
9446 | ||
e8319a19 | 9447 | switch (type) { |
0ec5d470 AK |
9448 | case update_general_migration_checkpoint: { |
9449 | struct intel_dev *id; | |
9450 | struct imsm_update_general_migration_checkpoint *u = | |
9451 | (void *)update->buf; | |
9452 | ||
1ade5cc1 | 9453 | dprintf("called for update_general_migration_checkpoint\n"); |
0ec5d470 AK |
9454 | |
9455 | /* find device under general migration */ | |
9456 | for (id = super->devlist ; id; id = id->next) { | |
9457 | if (is_gen_migration(id->dev)) { | |
9458 | id->dev->vol.curr_migr_unit = | |
9459 | __cpu_to_le32(u->curr_migr_unit); | |
9460 | super->updates_pending++; | |
9461 | } | |
9462 | } | |
9463 | break; | |
9464 | } | |
bb025c2f KW |
9465 | case update_takeover: { |
9466 | struct imsm_update_takeover *u = (void *)update->buf; | |
1a2487c2 KW |
9467 | if (apply_takeover_update(u, super, &update->space_list)) { |
9468 | imsm_update_version_info(super); | |
bb025c2f | 9469 | super->updates_pending++; |
1a2487c2 | 9470 | } |
bb025c2f KW |
9471 | break; |
9472 | } | |
9473 | ||
78b10e66 | 9474 | case update_reshape_container_disks: { |
d195167d | 9475 | struct imsm_update_reshape *u = (void *)update->buf; |
2e5dc010 N |
9476 | if (apply_reshape_container_disks_update( |
9477 | u, super, &update->space_list)) | |
9478 | super->updates_pending++; | |
78b10e66 N |
9479 | break; |
9480 | } | |
48c5303a | 9481 | case update_reshape_migration: { |
a29911da PC |
9482 | struct imsm_update_reshape_migration *u = (void *)update->buf; |
9483 | if (apply_reshape_migration_update( | |
9484 | u, super, &update->space_list)) | |
9485 | super->updates_pending++; | |
48c5303a PC |
9486 | break; |
9487 | } | |
f3871fdc AK |
9488 | case update_size_change: { |
9489 | struct imsm_update_size_change *u = (void *)update->buf; | |
9490 | if (apply_size_change_update(u, super)) | |
9491 | super->updates_pending++; | |
9492 | break; | |
9493 | } | |
e8319a19 | 9494 | case update_activate_spare: { |
1011e834 | 9495 | struct imsm_update_activate_spare *u = (void *) update->buf; |
061d7da3 LO |
9496 | if (apply_update_activate_spare(u, super, st->arrays)) |
9497 | super->updates_pending++; | |
8273f55e DW |
9498 | break; |
9499 | } | |
9500 | case update_create_array: { | |
9501 | /* someone wants to create a new array, we need to be aware of | |
9502 | * a few races/collisions: | |
9503 | * 1/ 'Create' called by two separate instances of mdadm | |
9504 | * 2/ 'Create' versus 'activate_spare': mdadm has chosen | |
9505 | * devices that have since been assimilated via | |
9506 | * activate_spare. | |
9507 | * In the event this update can not be carried out mdadm will | |
9508 | * (FIX ME) notice that its update did not take hold. | |
9509 | */ | |
9510 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 9511 | struct intel_dev *dv; |
8273f55e DW |
9512 | struct imsm_dev *dev; |
9513 | struct imsm_map *map, *new_map; | |
9514 | unsigned long long start, end; | |
9515 | unsigned long long new_start, new_end; | |
9516 | int i; | |
54c2c1ea DW |
9517 | struct disk_info *inf; |
9518 | struct dl *dl; | |
8273f55e DW |
9519 | |
9520 | /* handle racing creates: first come first serve */ | |
9521 | if (u->dev_idx < mpb->num_raid_devs) { | |
1ade5cc1 | 9522 | dprintf("subarray %d already defined\n", u->dev_idx); |
ba2de7ba | 9523 | goto create_error; |
8273f55e DW |
9524 | } |
9525 | ||
9526 | /* check update is next in sequence */ | |
9527 | if (u->dev_idx != mpb->num_raid_devs) { | |
1ade5cc1 N |
9528 | dprintf("can not create array %d expected index %d\n", |
9529 | u->dev_idx, mpb->num_raid_devs); | |
ba2de7ba | 9530 | goto create_error; |
8273f55e DW |
9531 | } |
9532 | ||
238c0a71 | 9533 | new_map = get_imsm_map(&u->dev, MAP_0); |
5551b113 CA |
9534 | new_start = pba_of_lba0(new_map); |
9535 | new_end = new_start + blocks_per_member(new_map); | |
54c2c1ea | 9536 | inf = get_disk_info(u); |
8273f55e DW |
9537 | |
9538 | /* handle activate_spare versus create race: | |
9539 | * check to make sure that overlapping arrays do not include | |
9540 | * overalpping disks | |
9541 | */ | |
9542 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
949c47a0 | 9543 | dev = get_imsm_dev(super, i); |
238c0a71 | 9544 | map = get_imsm_map(dev, MAP_0); |
5551b113 CA |
9545 | start = pba_of_lba0(map); |
9546 | end = start + blocks_per_member(map); | |
8273f55e DW |
9547 | if ((new_start >= start && new_start <= end) || |
9548 | (start >= new_start && start <= new_end)) | |
54c2c1ea DW |
9549 | /* overlap */; |
9550 | else | |
9551 | continue; | |
9552 | ||
9553 | if (disks_overlap(super, i, u)) { | |
1ade5cc1 | 9554 | dprintf("arrays overlap\n"); |
ba2de7ba | 9555 | goto create_error; |
8273f55e DW |
9556 | } |
9557 | } | |
8273f55e | 9558 | |
949c47a0 DW |
9559 | /* check that prepare update was successful */ |
9560 | if (!update->space) { | |
1ade5cc1 | 9561 | dprintf("prepare update failed\n"); |
ba2de7ba | 9562 | goto create_error; |
949c47a0 DW |
9563 | } |
9564 | ||
54c2c1ea DW |
9565 | /* check that all disks are still active before committing |
9566 | * changes. FIXME: could we instead handle this by creating a | |
9567 | * degraded array? That's probably not what the user expects, | |
9568 | * so better to drop this update on the floor. | |
9569 | */ | |
9570 | for (i = 0; i < new_map->num_members; i++) { | |
9571 | dl = serial_to_dl(inf[i].serial, super); | |
9572 | if (!dl) { | |
1ade5cc1 | 9573 | dprintf("disk disappeared\n"); |
ba2de7ba | 9574 | goto create_error; |
54c2c1ea | 9575 | } |
949c47a0 DW |
9576 | } |
9577 | ||
8273f55e | 9578 | super->updates_pending++; |
54c2c1ea DW |
9579 | |
9580 | /* convert spares to members and fixup ord_tbl */ | |
9581 | for (i = 0; i < new_map->num_members; i++) { | |
9582 | dl = serial_to_dl(inf[i].serial, super); | |
9583 | if (dl->index == -1) { | |
9584 | dl->index = mpb->num_disks; | |
9585 | mpb->num_disks++; | |
9586 | dl->disk.status |= CONFIGURED_DISK; | |
9587 | dl->disk.status &= ~SPARE_DISK; | |
9588 | } | |
9589 | set_imsm_ord_tbl_ent(new_map, i, dl->index); | |
9590 | } | |
9591 | ||
ba2de7ba DW |
9592 | dv = update->space; |
9593 | dev = dv->dev; | |
949c47a0 DW |
9594 | update->space = NULL; |
9595 | imsm_copy_dev(dev, &u->dev); | |
ba2de7ba DW |
9596 | dv->index = u->dev_idx; |
9597 | dv->next = super->devlist; | |
9598 | super->devlist = dv; | |
8273f55e | 9599 | mpb->num_raid_devs++; |
8273f55e | 9600 | |
4d1313e9 | 9601 | imsm_update_version_info(super); |
8273f55e | 9602 | break; |
ba2de7ba DW |
9603 | create_error: |
9604 | /* mdmon knows how to release update->space, but not | |
9605 | * ((struct intel_dev *) update->space)->dev | |
9606 | */ | |
9607 | if (update->space) { | |
9608 | dv = update->space; | |
9609 | free(dv->dev); | |
9610 | } | |
8273f55e | 9611 | break; |
e8319a19 | 9612 | } |
33414a01 DW |
9613 | case update_kill_array: { |
9614 | struct imsm_update_kill_array *u = (void *) update->buf; | |
9615 | int victim = u->dev_idx; | |
9616 | struct active_array *a; | |
9617 | struct intel_dev **dp; | |
9618 | struct imsm_dev *dev; | |
9619 | ||
9620 | /* sanity check that we are not affecting the uuid of | |
9621 | * active arrays, or deleting an active array | |
9622 | * | |
9623 | * FIXME when immutable ids are available, but note that | |
9624 | * we'll also need to fixup the invalidated/active | |
9625 | * subarray indexes in mdstat | |
9626 | */ | |
9627 | for (a = st->arrays; a; a = a->next) | |
9628 | if (a->info.container_member >= victim) | |
9629 | break; | |
9630 | /* by definition if mdmon is running at least one array | |
9631 | * is active in the container, so checking | |
9632 | * mpb->num_raid_devs is just extra paranoia | |
9633 | */ | |
9634 | dev = get_imsm_dev(super, victim); | |
9635 | if (a || !dev || mpb->num_raid_devs == 1) { | |
9636 | dprintf("failed to delete subarray-%d\n", victim); | |
9637 | break; | |
9638 | } | |
9639 | ||
9640 | for (dp = &super->devlist; *dp;) | |
f21e18ca | 9641 | if ((*dp)->index == (unsigned)super->current_vol) { |
33414a01 DW |
9642 | *dp = (*dp)->next; |
9643 | } else { | |
f21e18ca | 9644 | if ((*dp)->index > (unsigned)victim) |
33414a01 DW |
9645 | (*dp)->index--; |
9646 | dp = &(*dp)->next; | |
9647 | } | |
9648 | mpb->num_raid_devs--; | |
9649 | super->updates_pending++; | |
9650 | break; | |
9651 | } | |
aa534678 DW |
9652 | case update_rename_array: { |
9653 | struct imsm_update_rename_array *u = (void *) update->buf; | |
9654 | char name[MAX_RAID_SERIAL_LEN+1]; | |
9655 | int target = u->dev_idx; | |
9656 | struct active_array *a; | |
9657 | struct imsm_dev *dev; | |
9658 | ||
9659 | /* sanity check that we are not affecting the uuid of | |
9660 | * an active array | |
9661 | */ | |
9662 | snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name); | |
9663 | name[MAX_RAID_SERIAL_LEN] = '\0'; | |
9664 | for (a = st->arrays; a; a = a->next) | |
9665 | if (a->info.container_member == target) | |
9666 | break; | |
9667 | dev = get_imsm_dev(super, u->dev_idx); | |
9668 | if (a || !dev || !check_name(super, name, 1)) { | |
9669 | dprintf("failed to rename subarray-%d\n", target); | |
9670 | break; | |
9671 | } | |
9672 | ||
cdbe98cd | 9673 | snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name); |
aa534678 DW |
9674 | super->updates_pending++; |
9675 | break; | |
9676 | } | |
1a64be56 | 9677 | case update_add_remove_disk: { |
43dad3d6 | 9678 | /* we may be able to repair some arrays if disks are |
095b8088 | 9679 | * being added, check the status of add_remove_disk |
1a64be56 LM |
9680 | * if discs has been added. |
9681 | */ | |
9682 | if (add_remove_disk_update(super)) { | |
43dad3d6 | 9683 | struct active_array *a; |
072b727f DW |
9684 | |
9685 | super->updates_pending++; | |
1a64be56 | 9686 | for (a = st->arrays; a; a = a->next) |
43dad3d6 DW |
9687 | a->check_degraded = 1; |
9688 | } | |
43dad3d6 | 9689 | break; |
e8319a19 | 9690 | } |
bbab0940 TM |
9691 | case update_prealloc_badblocks_mem: |
9692 | break; | |
e6e9dd3f AP |
9693 | case update_rwh_policy: { |
9694 | struct imsm_update_rwh_policy *u = (void *)update->buf; | |
9695 | int target = u->dev_idx; | |
9696 | struct imsm_dev *dev = get_imsm_dev(super, target); | |
9697 | if (!dev) { | |
9698 | dprintf("could not find subarray-%d\n", target); | |
9699 | break; | |
9700 | } | |
9701 | ||
9702 | if (dev->rwh_policy != u->new_policy) { | |
9703 | dev->rwh_policy = u->new_policy; | |
9704 | super->updates_pending++; | |
9705 | } | |
9706 | break; | |
9707 | } | |
1a64be56 | 9708 | default: |
7a862a02 | 9709 | pr_err("error: unsuported process update type:(type: %d)\n", type); |
1a64be56 | 9710 | } |
e8319a19 | 9711 | } |
88758e9d | 9712 | |
bc0b9d34 PC |
9713 | static struct mdinfo *get_spares_for_grow(struct supertype *st); |
9714 | ||
5fe6f031 N |
9715 | static int imsm_prepare_update(struct supertype *st, |
9716 | struct metadata_update *update) | |
8273f55e | 9717 | { |
949c47a0 | 9718 | /** |
4d7b1503 DW |
9719 | * Allocate space to hold new disk entries, raid-device entries or a new |
9720 | * mpb if necessary. The manager synchronously waits for updates to | |
9721 | * complete in the monitor, so new mpb buffers allocated here can be | |
9722 | * integrated by the monitor thread without worrying about live pointers | |
9723 | * in the manager thread. | |
8273f55e | 9724 | */ |
095b8088 | 9725 | enum imsm_update_type type; |
4d7b1503 | 9726 | struct intel_super *super = st->sb; |
f36a9ecd | 9727 | unsigned int sector_size = super->sector_size; |
4d7b1503 DW |
9728 | struct imsm_super *mpb = super->anchor; |
9729 | size_t buf_len; | |
9730 | size_t len = 0; | |
949c47a0 | 9731 | |
095b8088 N |
9732 | if (update->len < (int)sizeof(type)) |
9733 | return 0; | |
9734 | ||
9735 | type = *(enum imsm_update_type *) update->buf; | |
9736 | ||
949c47a0 | 9737 | switch (type) { |
0ec5d470 | 9738 | case update_general_migration_checkpoint: |
095b8088 N |
9739 | if (update->len < (int)sizeof(struct imsm_update_general_migration_checkpoint)) |
9740 | return 0; | |
1ade5cc1 | 9741 | dprintf("called for update_general_migration_checkpoint\n"); |
0ec5d470 | 9742 | break; |
abedf5fc KW |
9743 | case update_takeover: { |
9744 | struct imsm_update_takeover *u = (void *)update->buf; | |
095b8088 N |
9745 | if (update->len < (int)sizeof(*u)) |
9746 | return 0; | |
abedf5fc KW |
9747 | if (u->direction == R0_TO_R10) { |
9748 | void **tail = (void **)&update->space_list; | |
9749 | struct imsm_dev *dev = get_imsm_dev(super, u->subarray); | |
238c0a71 | 9750 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
abedf5fc KW |
9751 | int num_members = map->num_members; |
9752 | void *space; | |
9753 | int size, i; | |
abedf5fc KW |
9754 | /* allocate memory for added disks */ |
9755 | for (i = 0; i < num_members; i++) { | |
9756 | size = sizeof(struct dl); | |
503975b9 | 9757 | space = xmalloc(size); |
abedf5fc KW |
9758 | *tail = space; |
9759 | tail = space; | |
9760 | *tail = NULL; | |
9761 | } | |
9762 | /* allocate memory for new device */ | |
9763 | size = sizeof_imsm_dev(super->devlist->dev, 0) + | |
9764 | (num_members * sizeof(__u32)); | |
503975b9 N |
9765 | space = xmalloc(size); |
9766 | *tail = space; | |
9767 | tail = space; | |
9768 | *tail = NULL; | |
9769 | len = disks_to_mpb_size(num_members * 2); | |
abedf5fc KW |
9770 | } |
9771 | ||
9772 | break; | |
9773 | } | |
78b10e66 | 9774 | case update_reshape_container_disks: { |
d195167d AK |
9775 | /* Every raid device in the container is about to |
9776 | * gain some more devices, and we will enter a | |
9777 | * reconfiguration. | |
9778 | * So each 'imsm_map' will be bigger, and the imsm_vol | |
9779 | * will now hold 2 of them. | |
9780 | * Thus we need new 'struct imsm_dev' allocations sized | |
9781 | * as sizeof_imsm_dev but with more devices in both maps. | |
9782 | */ | |
9783 | struct imsm_update_reshape *u = (void *)update->buf; | |
9784 | struct intel_dev *dl; | |
9785 | void **space_tail = (void**)&update->space_list; | |
9786 | ||
095b8088 N |
9787 | if (update->len < (int)sizeof(*u)) |
9788 | return 0; | |
9789 | ||
1ade5cc1 | 9790 | dprintf("for update_reshape\n"); |
d195167d AK |
9791 | |
9792 | for (dl = super->devlist; dl; dl = dl->next) { | |
9793 | int size = sizeof_imsm_dev(dl->dev, 1); | |
9794 | void *s; | |
d677e0b8 AK |
9795 | if (u->new_raid_disks > u->old_raid_disks) |
9796 | size += sizeof(__u32)*2* | |
9797 | (u->new_raid_disks - u->old_raid_disks); | |
503975b9 | 9798 | s = xmalloc(size); |
d195167d AK |
9799 | *space_tail = s; |
9800 | space_tail = s; | |
9801 | *space_tail = NULL; | |
9802 | } | |
9803 | ||
9804 | len = disks_to_mpb_size(u->new_raid_disks); | |
9805 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
78b10e66 N |
9806 | break; |
9807 | } | |
48c5303a | 9808 | case update_reshape_migration: { |
bc0b9d34 PC |
9809 | /* for migration level 0->5 we need to add disks |
9810 | * so the same as for container operation we will copy | |
9811 | * device to the bigger location. | |
9812 | * in memory prepared device and new disk area are prepared | |
9813 | * for usage in process update | |
9814 | */ | |
9815 | struct imsm_update_reshape_migration *u = (void *)update->buf; | |
9816 | struct intel_dev *id; | |
9817 | void **space_tail = (void **)&update->space_list; | |
9818 | int size; | |
9819 | void *s; | |
9820 | int current_level = -1; | |
9821 | ||
095b8088 N |
9822 | if (update->len < (int)sizeof(*u)) |
9823 | return 0; | |
9824 | ||
1ade5cc1 | 9825 | dprintf("for update_reshape\n"); |
bc0b9d34 PC |
9826 | |
9827 | /* add space for bigger array in update | |
9828 | */ | |
9829 | for (id = super->devlist; id; id = id->next) { | |
9830 | if (id->index == (unsigned)u->subdev) { | |
9831 | size = sizeof_imsm_dev(id->dev, 1); | |
9832 | if (u->new_raid_disks > u->old_raid_disks) | |
9833 | size += sizeof(__u32)*2* | |
9834 | (u->new_raid_disks - u->old_raid_disks); | |
503975b9 | 9835 | s = xmalloc(size); |
bc0b9d34 PC |
9836 | *space_tail = s; |
9837 | space_tail = s; | |
9838 | *space_tail = NULL; | |
9839 | break; | |
9840 | } | |
9841 | } | |
9842 | if (update->space_list == NULL) | |
9843 | break; | |
9844 | ||
9845 | /* add space for disk in update | |
9846 | */ | |
9847 | size = sizeof(struct dl); | |
503975b9 | 9848 | s = xmalloc(size); |
bc0b9d34 PC |
9849 | *space_tail = s; |
9850 | space_tail = s; | |
9851 | *space_tail = NULL; | |
9852 | ||
9853 | /* add spare device to update | |
9854 | */ | |
9855 | for (id = super->devlist ; id; id = id->next) | |
9856 | if (id->index == (unsigned)u->subdev) { | |
9857 | struct imsm_dev *dev; | |
9858 | struct imsm_map *map; | |
9859 | ||
9860 | dev = get_imsm_dev(super, u->subdev); | |
238c0a71 | 9861 | map = get_imsm_map(dev, MAP_0); |
bc0b9d34 PC |
9862 | current_level = map->raid_level; |
9863 | break; | |
9864 | } | |
089f9d79 | 9865 | if (u->new_level == 5 && u->new_level != current_level) { |
bc0b9d34 PC |
9866 | struct mdinfo *spares; |
9867 | ||
9868 | spares = get_spares_for_grow(st); | |
9869 | if (spares) { | |
9870 | struct dl *dl; | |
9871 | struct mdinfo *dev; | |
9872 | ||
9873 | dev = spares->devs; | |
9874 | if (dev) { | |
9875 | u->new_disks[0] = | |
9876 | makedev(dev->disk.major, | |
9877 | dev->disk.minor); | |
9878 | dl = get_disk_super(super, | |
9879 | dev->disk.major, | |
9880 | dev->disk.minor); | |
9881 | dl->index = u->old_raid_disks; | |
9882 | dev = dev->next; | |
9883 | } | |
9884 | sysfs_free(spares); | |
9885 | } | |
9886 | } | |
9887 | len = disks_to_mpb_size(u->new_raid_disks); | |
9888 | dprintf("New anchor length is %llu\n", (unsigned long long)len); | |
48c5303a PC |
9889 | break; |
9890 | } | |
f3871fdc | 9891 | case update_size_change: { |
095b8088 N |
9892 | if (update->len < (int)sizeof(struct imsm_update_size_change)) |
9893 | return 0; | |
9894 | break; | |
9895 | } | |
9896 | case update_activate_spare: { | |
9897 | if (update->len < (int)sizeof(struct imsm_update_activate_spare)) | |
9898 | return 0; | |
f3871fdc AK |
9899 | break; |
9900 | } | |
949c47a0 DW |
9901 | case update_create_array: { |
9902 | struct imsm_update_create_array *u = (void *) update->buf; | |
ba2de7ba | 9903 | struct intel_dev *dv; |
54c2c1ea | 9904 | struct imsm_dev *dev = &u->dev; |
238c0a71 | 9905 | struct imsm_map *map = get_imsm_map(dev, MAP_0); |
54c2c1ea DW |
9906 | struct dl *dl; |
9907 | struct disk_info *inf; | |
9908 | int i; | |
9909 | int activate = 0; | |
949c47a0 | 9910 | |
095b8088 N |
9911 | if (update->len < (int)sizeof(*u)) |
9912 | return 0; | |
9913 | ||
54c2c1ea DW |
9914 | inf = get_disk_info(u); |
9915 | len = sizeof_imsm_dev(dev, 1); | |
ba2de7ba | 9916 | /* allocate a new super->devlist entry */ |
503975b9 N |
9917 | dv = xmalloc(sizeof(*dv)); |
9918 | dv->dev = xmalloc(len); | |
9919 | update->space = dv; | |
949c47a0 | 9920 | |
54c2c1ea DW |
9921 | /* count how many spares will be converted to members */ |
9922 | for (i = 0; i < map->num_members; i++) { | |
9923 | dl = serial_to_dl(inf[i].serial, super); | |
9924 | if (!dl) { | |
9925 | /* hmm maybe it failed?, nothing we can do about | |
9926 | * it here | |
9927 | */ | |
9928 | continue; | |
9929 | } | |
9930 | if (count_memberships(dl, super) == 0) | |
9931 | activate++; | |
9932 | } | |
9933 | len += activate * sizeof(struct imsm_disk); | |
949c47a0 | 9934 | break; |
095b8088 N |
9935 | } |
9936 | case update_kill_array: { | |
9937 | if (update->len < (int)sizeof(struct imsm_update_kill_array)) | |
9938 | return 0; | |
949c47a0 DW |
9939 | break; |
9940 | } | |
095b8088 N |
9941 | case update_rename_array: { |
9942 | if (update->len < (int)sizeof(struct imsm_update_rename_array)) | |
9943 | return 0; | |
9944 | break; | |
9945 | } | |
9946 | case update_add_remove_disk: | |
9947 | /* no update->len needed */ | |
9948 | break; | |
bbab0940 TM |
9949 | case update_prealloc_badblocks_mem: |
9950 | super->extra_space += sizeof(struct bbm_log) - | |
9951 | get_imsm_bbm_log_size(super->bbm_log); | |
9952 | break; | |
e6e9dd3f AP |
9953 | case update_rwh_policy: { |
9954 | if (update->len < (int)sizeof(struct imsm_update_rwh_policy)) | |
9955 | return 0; | |
9956 | break; | |
9957 | } | |
095b8088 N |
9958 | default: |
9959 | return 0; | |
949c47a0 | 9960 | } |
8273f55e | 9961 | |
4d7b1503 DW |
9962 | /* check if we need a larger metadata buffer */ |
9963 | if (super->next_buf) | |
9964 | buf_len = super->next_len; | |
9965 | else | |
9966 | buf_len = super->len; | |
9967 | ||
bbab0940 | 9968 | if (__le32_to_cpu(mpb->mpb_size) + super->extra_space + len > buf_len) { |
4d7b1503 DW |
9969 | /* ok we need a larger buf than what is currently allocated |
9970 | * if this allocation fails process_update will notice that | |
9971 | * ->next_len is set and ->next_buf is NULL | |
9972 | */ | |
bbab0940 TM |
9973 | buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + |
9974 | super->extra_space + len, sector_size); | |
4d7b1503 DW |
9975 | if (super->next_buf) |
9976 | free(super->next_buf); | |
9977 | ||
9978 | super->next_len = buf_len; | |
f36a9ecd | 9979 | if (posix_memalign(&super->next_buf, sector_size, buf_len) == 0) |
1f45a8ad DW |
9980 | memset(super->next_buf, 0, buf_len); |
9981 | else | |
4d7b1503 DW |
9982 | super->next_buf = NULL; |
9983 | } | |
5fe6f031 | 9984 | return 1; |
8273f55e DW |
9985 | } |
9986 | ||
ae6aad82 | 9987 | /* must be called while manager is quiesced */ |
f21e18ca | 9988 | static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index) |
ae6aad82 DW |
9989 | { |
9990 | struct imsm_super *mpb = super->anchor; | |
ae6aad82 DW |
9991 | struct dl *iter; |
9992 | struct imsm_dev *dev; | |
9993 | struct imsm_map *map; | |
4c9e8c1e | 9994 | unsigned int i, j, num_members; |
fb12a745 | 9995 | __u32 ord, ord_map0; |
4c9e8c1e | 9996 | struct bbm_log *log = super->bbm_log; |
ae6aad82 | 9997 | |
1ade5cc1 | 9998 | dprintf("deleting device[%d] from imsm_super\n", index); |
ae6aad82 DW |
9999 | |
10000 | /* shift all indexes down one */ | |
10001 | for (iter = super->disks; iter; iter = iter->next) | |
f21e18ca | 10002 | if (iter->index > (int)index) |
ae6aad82 | 10003 | iter->index--; |
47ee5a45 | 10004 | for (iter = super->missing; iter; iter = iter->next) |
f21e18ca | 10005 | if (iter->index > (int)index) |
47ee5a45 | 10006 | iter->index--; |
ae6aad82 DW |
10007 | |
10008 | for (i = 0; i < mpb->num_raid_devs; i++) { | |
10009 | dev = get_imsm_dev(super, i); | |
238c0a71 | 10010 | map = get_imsm_map(dev, MAP_0); |
24565c9a DW |
10011 | num_members = map->num_members; |
10012 | for (j = 0; j < num_members; j++) { | |
10013 | /* update ord entries being careful not to propagate | |
10014 | * ord-flags to the first map | |
10015 | */ | |
238c0a71 | 10016 | ord = get_imsm_ord_tbl_ent(dev, j, MAP_X); |
fb12a745 | 10017 | ord_map0 = get_imsm_ord_tbl_ent(dev, j, MAP_0); |
ae6aad82 | 10018 | |
24565c9a DW |
10019 | if (ord_to_idx(ord) <= index) |
10020 | continue; | |
ae6aad82 | 10021 | |
238c0a71 | 10022 | map = get_imsm_map(dev, MAP_0); |
fb12a745 | 10023 | set_imsm_ord_tbl_ent(map, j, ord_map0 - 1); |
238c0a71 | 10024 | map = get_imsm_map(dev, MAP_1); |
24565c9a DW |
10025 | if (map) |
10026 | set_imsm_ord_tbl_ent(map, j, ord - 1); | |
ae6aad82 DW |
10027 | } |
10028 | } | |
10029 | ||
4c9e8c1e TM |
10030 | for (i = 0; i < log->entry_count; i++) { |
10031 | struct bbm_log_entry *entry = &log->marked_block_entries[i]; | |
10032 | ||
10033 | if (entry->disk_ordinal <= index) | |
10034 | continue; | |
10035 | entry->disk_ordinal--; | |
10036 | } | |
10037 | ||
ae6aad82 DW |
10038 | mpb->num_disks--; |
10039 | super->updates_pending++; | |
24565c9a DW |
10040 | if (*dlp) { |
10041 | struct dl *dl = *dlp; | |
10042 | ||
10043 | *dlp = (*dlp)->next; | |
10044 | __free_imsm_disk(dl); | |
10045 | } | |
ae6aad82 | 10046 | } |
9a717282 AK |
10047 | |
10048 | static void close_targets(int *targets, int new_disks) | |
10049 | { | |
10050 | int i; | |
10051 | ||
10052 | if (!targets) | |
10053 | return; | |
10054 | ||
10055 | for (i = 0; i < new_disks; i++) { | |
10056 | if (targets[i] >= 0) { | |
10057 | close(targets[i]); | |
10058 | targets[i] = -1; | |
10059 | } | |
10060 | } | |
10061 | } | |
10062 | ||
10063 | static int imsm_get_allowed_degradation(int level, int raid_disks, | |
10064 | struct intel_super *super, | |
10065 | struct imsm_dev *dev) | |
10066 | { | |
10067 | switch (level) { | |
bf5cf7c7 | 10068 | case 1: |
9a717282 AK |
10069 | case 10:{ |
10070 | int ret_val = 0; | |
10071 | struct imsm_map *map; | |
10072 | int i; | |
10073 | ||
10074 | ret_val = raid_disks/2; | |
10075 | /* check map if all disks pairs not failed | |
10076 | * in both maps | |
10077 | */ | |
238c0a71 | 10078 | map = get_imsm_map(dev, MAP_0); |
9a717282 AK |
10079 | for (i = 0; i < ret_val; i++) { |
10080 | int degradation = 0; | |
10081 | if (get_imsm_disk(super, i) == NULL) | |
10082 | degradation++; | |
10083 | if (get_imsm_disk(super, i + 1) == NULL) | |
10084 | degradation++; | |
10085 | if (degradation == 2) | |
10086 | return 0; | |
10087 | } | |
238c0a71 | 10088 | map = get_imsm_map(dev, MAP_1); |
9a717282 AK |
10089 | /* if there is no second map |
10090 | * result can be returned | |
10091 | */ | |
10092 | if (map == NULL) | |
10093 | return ret_val; | |
10094 | /* check degradation in second map | |
10095 | */ | |
10096 | for (i = 0; i < ret_val; i++) { | |
10097 | int degradation = 0; | |
10098 | if (get_imsm_disk(super, i) == NULL) | |
10099 | degradation++; | |
10100 | if (get_imsm_disk(super, i + 1) == NULL) | |
10101 | degradation++; | |
10102 | if (degradation == 2) | |
10103 | return 0; | |
10104 | } | |
10105 | return ret_val; | |
10106 | } | |
10107 | case 5: | |
10108 | return 1; | |
10109 | case 6: | |
10110 | return 2; | |
10111 | default: | |
10112 | return 0; | |
10113 | } | |
10114 | } | |
10115 | ||
687629c2 AK |
10116 | /******************************************************************************* |
10117 | * Function: open_backup_targets | |
10118 | * Description: Function opens file descriptors for all devices given in | |
10119 | * info->devs | |
10120 | * Parameters: | |
10121 | * info : general array info | |
10122 | * raid_disks : number of disks | |
10123 | * raid_fds : table of device's file descriptors | |
9a717282 AK |
10124 | * super : intel super for raid10 degradation check |
10125 | * dev : intel device for raid10 degradation check | |
687629c2 AK |
10126 | * Returns: |
10127 | * 0 : success | |
10128 | * -1 : fail | |
10129 | ******************************************************************************/ | |
9a717282 AK |
10130 | int open_backup_targets(struct mdinfo *info, int raid_disks, int *raid_fds, |
10131 | struct intel_super *super, struct imsm_dev *dev) | |
687629c2 AK |
10132 | { |
10133 | struct mdinfo *sd; | |
f627f5ad | 10134 | int i; |
9a717282 | 10135 | int opened = 0; |
f627f5ad AK |
10136 | |
10137 | for (i = 0; i < raid_disks; i++) | |
10138 | raid_fds[i] = -1; | |
687629c2 AK |
10139 | |
10140 | for (sd = info->devs ; sd ; sd = sd->next) { | |
10141 | char *dn; | |
10142 | ||
10143 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) { | |
10144 | dprintf("disk is faulty!!\n"); | |
10145 | continue; | |
10146 | } | |
10147 | ||
089f9d79 | 10148 | if (sd->disk.raid_disk >= raid_disks || sd->disk.raid_disk < 0) |
687629c2 AK |
10149 | continue; |
10150 | ||
10151 | dn = map_dev(sd->disk.major, | |
10152 | sd->disk.minor, 1); | |
10153 | raid_fds[sd->disk.raid_disk] = dev_open(dn, O_RDWR); | |
10154 | if (raid_fds[sd->disk.raid_disk] < 0) { | |
e12b3daa | 10155 | pr_err("cannot open component\n"); |
9a717282 | 10156 | continue; |
687629c2 | 10157 | } |
9a717282 AK |
10158 | opened++; |
10159 | } | |
10160 | /* check if maximum array degradation level is not exceeded | |
10161 | */ | |
10162 | if ((raid_disks - opened) > | |
089f9d79 JS |
10163 | imsm_get_allowed_degradation(info->new_level, raid_disks, |
10164 | super, dev)) { | |
e12b3daa | 10165 | pr_err("Not enough disks can be opened.\n"); |
9a717282 AK |
10166 | close_targets(raid_fds, raid_disks); |
10167 | return -2; | |
687629c2 AK |
10168 | } |
10169 | return 0; | |
10170 | } | |
10171 | ||
d31ad643 PB |
10172 | /******************************************************************************* |
10173 | * Function: validate_container_imsm | |
10174 | * Description: This routine validates container after assemble, | |
10175 | * eg. if devices in container are under the same controller. | |
10176 | * | |
10177 | * Parameters: | |
10178 | * info : linked list with info about devices used in array | |
10179 | * Returns: | |
10180 | * 1 : HBA mismatch | |
10181 | * 0 : Success | |
10182 | ******************************************************************************/ | |
10183 | int validate_container_imsm(struct mdinfo *info) | |
10184 | { | |
6b781d33 AP |
10185 | if (check_env("IMSM_NO_PLATFORM")) |
10186 | return 0; | |
d31ad643 | 10187 | |
6b781d33 AP |
10188 | struct sys_dev *idev; |
10189 | struct sys_dev *hba = NULL; | |
10190 | struct sys_dev *intel_devices = find_intel_devices(); | |
10191 | char *dev_path = devt_to_devpath(makedev(info->disk.major, | |
10192 | info->disk.minor)); | |
10193 | ||
10194 | for (idev = intel_devices; idev; idev = idev->next) { | |
10195 | if (dev_path && strstr(dev_path, idev->path)) { | |
10196 | hba = idev; | |
10197 | break; | |
d31ad643 | 10198 | } |
6b781d33 AP |
10199 | } |
10200 | if (dev_path) | |
d31ad643 PB |
10201 | free(dev_path); |
10202 | ||
6b781d33 AP |
10203 | if (!hba) { |
10204 | pr_err("WARNING - Cannot detect HBA for device %s!\n", | |
10205 | devid2kname(makedev(info->disk.major, info->disk.minor))); | |
10206 | return 1; | |
10207 | } | |
10208 | ||
10209 | const struct imsm_orom *orom = get_orom_by_device_id(hba->dev_id); | |
10210 | struct mdinfo *dev; | |
10211 | ||
10212 | for (dev = info->next; dev; dev = dev->next) { | |
10213 | dev_path = devt_to_devpath(makedev(dev->disk.major, dev->disk.minor)); | |
10214 | ||
10215 | struct sys_dev *hba2 = NULL; | |
10216 | for (idev = intel_devices; idev; idev = idev->next) { | |
10217 | if (dev_path && strstr(dev_path, idev->path)) { | |
10218 | hba2 = idev; | |
10219 | break; | |
d31ad643 PB |
10220 | } |
10221 | } | |
6b781d33 AP |
10222 | if (dev_path) |
10223 | free(dev_path); | |
10224 | ||
10225 | const struct imsm_orom *orom2 = hba2 == NULL ? NULL : | |
10226 | get_orom_by_device_id(hba2->dev_id); | |
10227 | ||
10228 | if (hba2 && hba->type != hba2->type) { | |
10229 | pr_err("WARNING - HBAs of devices do not match %s != %s\n", | |
10230 | get_sys_dev_type(hba->type), get_sys_dev_type(hba2->type)); | |
10231 | return 1; | |
10232 | } | |
10233 | ||
07cb1e57 | 10234 | if (orom != orom2) { |
6b781d33 AP |
10235 | pr_err("WARNING - IMSM container assembled with disks under different HBAs!\n" |
10236 | " This operation is not supported and can lead to data loss.\n"); | |
10237 | return 1; | |
10238 | } | |
10239 | ||
10240 | if (!orom) { | |
10241 | pr_err("WARNING - IMSM container assembled with disks under HBAs without IMSM platform support!\n" | |
10242 | " This operation is not supported and can lead to data loss.\n"); | |
10243 | return 1; | |
10244 | } | |
d31ad643 | 10245 | } |
6b781d33 | 10246 | |
d31ad643 PB |
10247 | return 0; |
10248 | } | |
32141c17 | 10249 | |
6f50473f TM |
10250 | /******************************************************************************* |
10251 | * Function: imsm_record_badblock | |
10252 | * Description: This routine stores new bad block record in BBM log | |
10253 | * | |
10254 | * Parameters: | |
10255 | * a : array containing a bad block | |
10256 | * slot : disk number containing a bad block | |
10257 | * sector : bad block sector | |
10258 | * length : bad block sectors range | |
10259 | * Returns: | |
10260 | * 1 : Success | |
10261 | * 0 : Error | |
10262 | ******************************************************************************/ | |
10263 | static int imsm_record_badblock(struct active_array *a, int slot, | |
10264 | unsigned long long sector, int length) | |
10265 | { | |
10266 | struct intel_super *super = a->container->sb; | |
10267 | int ord; | |
10268 | int ret; | |
10269 | ||
10270 | ord = imsm_disk_slot_to_ord(a, slot); | |
10271 | if (ord < 0) | |
10272 | return 0; | |
10273 | ||
10274 | ret = record_new_badblock(super->bbm_log, ord_to_idx(ord), sector, | |
10275 | length); | |
10276 | if (ret) | |
10277 | super->updates_pending++; | |
10278 | ||
10279 | return ret; | |
10280 | } | |
c07a5a4f TM |
10281 | /******************************************************************************* |
10282 | * Function: imsm_clear_badblock | |
10283 | * Description: This routine clears bad block record from BBM log | |
10284 | * | |
10285 | * Parameters: | |
10286 | * a : array containing a bad block | |
10287 | * slot : disk number containing a bad block | |
10288 | * sector : bad block sector | |
10289 | * length : bad block sectors range | |
10290 | * Returns: | |
10291 | * 1 : Success | |
10292 | * 0 : Error | |
10293 | ******************************************************************************/ | |
10294 | static int imsm_clear_badblock(struct active_array *a, int slot, | |
10295 | unsigned long long sector, int length) | |
10296 | { | |
10297 | struct intel_super *super = a->container->sb; | |
10298 | int ord; | |
10299 | int ret; | |
10300 | ||
10301 | ord = imsm_disk_slot_to_ord(a, slot); | |
10302 | if (ord < 0) | |
10303 | return 0; | |
10304 | ||
10305 | ret = clear_badblock(super->bbm_log, ord_to_idx(ord), sector, length); | |
10306 | if (ret) | |
10307 | super->updates_pending++; | |
10308 | ||
10309 | return ret; | |
10310 | } | |
928f1424 TM |
10311 | /******************************************************************************* |
10312 | * Function: imsm_get_badblocks | |
10313 | * Description: This routine get list of bad blocks for an array | |
10314 | * | |
10315 | * Parameters: | |
10316 | * a : array | |
10317 | * slot : disk number | |
10318 | * Returns: | |
10319 | * bb : structure containing bad blocks | |
10320 | * NULL : error | |
10321 | ******************************************************************************/ | |
10322 | static struct md_bb *imsm_get_badblocks(struct active_array *a, int slot) | |
10323 | { | |
10324 | int inst = a->info.container_member; | |
10325 | struct intel_super *super = a->container->sb; | |
10326 | struct imsm_dev *dev = get_imsm_dev(super, inst); | |
10327 | struct imsm_map *map = get_imsm_map(dev, MAP_0); | |
10328 | int ord; | |
10329 | ||
10330 | ord = imsm_disk_slot_to_ord(a, slot); | |
10331 | if (ord < 0) | |
10332 | return NULL; | |
10333 | ||
10334 | get_volume_badblocks(super->bbm_log, ord_to_idx(ord), pba_of_lba0(map), | |
10335 | blocks_per_member(map), &super->bb); | |
10336 | ||
10337 | return &super->bb; | |
10338 | } | |
27156a57 TM |
10339 | /******************************************************************************* |
10340 | * Function: examine_badblocks_imsm | |
10341 | * Description: Prints list of bad blocks on a disk to the standard output | |
10342 | * | |
10343 | * Parameters: | |
10344 | * st : metadata handler | |
10345 | * fd : open file descriptor for device | |
10346 | * devname : device name | |
10347 | * Returns: | |
10348 | * 0 : Success | |
10349 | * 1 : Error | |
10350 | ******************************************************************************/ | |
10351 | static int examine_badblocks_imsm(struct supertype *st, int fd, char *devname) | |
10352 | { | |
10353 | struct intel_super *super = st->sb; | |
10354 | struct bbm_log *log = super->bbm_log; | |
10355 | struct dl *d = NULL; | |
10356 | int any = 0; | |
10357 | ||
10358 | for (d = super->disks; d ; d = d->next) { | |
10359 | if (strcmp(d->devname, devname) == 0) | |
10360 | break; | |
10361 | } | |
10362 | ||
10363 | if ((d == NULL) || (d->index < 0)) { /* serial mismatch probably */ | |
10364 | pr_err("%s doesn't appear to be part of a raid array\n", | |
10365 | devname); | |
10366 | return 1; | |
10367 | } | |
10368 | ||
10369 | if (log != NULL) { | |
10370 | unsigned int i; | |
10371 | struct bbm_log_entry *entry = &log->marked_block_entries[0]; | |
10372 | ||
10373 | for (i = 0; i < log->entry_count; i++) { | |
10374 | if (entry[i].disk_ordinal == d->index) { | |
10375 | unsigned long long sector = __le48_to_cpu( | |
10376 | &entry[i].defective_block_start); | |
10377 | int cnt = entry[i].marked_count + 1; | |
10378 | ||
10379 | if (!any) { | |
10380 | printf("Bad-blocks on %s:\n", devname); | |
10381 | any = 1; | |
10382 | } | |
10383 | ||
10384 | printf("%20llu for %d sectors\n", sector, cnt); | |
10385 | } | |
10386 | } | |
10387 | } | |
10388 | ||
10389 | if (!any) | |
10390 | printf("No bad-blocks list configured on %s\n", devname); | |
10391 | ||
10392 | return 0; | |
10393 | } | |
687629c2 AK |
10394 | /******************************************************************************* |
10395 | * Function: init_migr_record_imsm | |
10396 | * Description: Function inits imsm migration record | |
10397 | * Parameters: | |
10398 | * super : imsm internal array info | |
10399 | * dev : device under migration | |
10400 | * info : general array info to find the smallest device | |
10401 | * Returns: | |
10402 | * none | |
10403 | ******************************************************************************/ | |
10404 | void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev, | |
10405 | struct mdinfo *info) | |
10406 | { | |
10407 | struct intel_super *super = st->sb; | |
10408 | struct migr_record *migr_rec = super->migr_rec; | |
10409 | int new_data_disks; | |
10410 | unsigned long long dsize, dev_sectors; | |
10411 | long long unsigned min_dev_sectors = -1LLU; | |
10412 | struct mdinfo *sd; | |
10413 | char nm[30]; | |
10414 | int fd; | |
238c0a71 AK |
10415 | struct imsm_map *map_dest = get_imsm_map(dev, MAP_0); |
10416 | struct imsm_map *map_src = get_imsm_map(dev, MAP_1); | |
687629c2 | 10417 | unsigned long long num_migr_units; |
3ef4403c | 10418 | unsigned long long array_blocks; |
687629c2 AK |
10419 | |
10420 | memset(migr_rec, 0, sizeof(struct migr_record)); | |
10421 | migr_rec->family_num = __cpu_to_le32(super->anchor->family_num); | |
10422 | ||
10423 | /* only ascending reshape supported now */ | |
10424 | migr_rec->ascending_migr = __cpu_to_le32(1); | |
10425 | ||
10426 | migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE / | |
10427 | max(map_dest->blocks_per_strip, map_src->blocks_per_strip); | |
e1742195 AK |
10428 | migr_rec->dest_depth_per_unit *= |
10429 | max(map_dest->blocks_per_strip, map_src->blocks_per_strip); | |
238c0a71 | 10430 | new_data_disks = imsm_num_data_members(dev, MAP_0); |
687629c2 AK |
10431 | migr_rec->blocks_per_unit = |
10432 | __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks); | |
10433 | migr_rec->dest_depth_per_unit = | |
10434 | __cpu_to_le32(migr_rec->dest_depth_per_unit); | |
3ef4403c | 10435 | array_blocks = info->component_size * new_data_disks; |
687629c2 AK |
10436 | num_migr_units = |
10437 | array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit); | |
10438 | ||
10439 | if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit)) | |
10440 | num_migr_units++; | |
10441 | migr_rec->num_migr_units = __cpu_to_le32(num_migr_units); | |
10442 | ||
10443 | migr_rec->post_migr_vol_cap = dev->size_low; | |
10444 | migr_rec->post_migr_vol_cap_hi = dev->size_high; | |
10445 | ||
687629c2 AK |
10446 | /* Find the smallest dev */ |
10447 | for (sd = info->devs ; sd ; sd = sd->next) { | |
10448 | sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor); | |
10449 | fd = dev_open(nm, O_RDONLY); | |
10450 | if (fd < 0) | |
10451 | continue; | |
10452 | get_dev_size(fd, NULL, &dsize); | |
10453 | dev_sectors = dsize / 512; | |
10454 | if (dev_sectors < min_dev_sectors) | |
10455 | min_dev_sectors = dev_sectors; | |
10456 | close(fd); | |
10457 | } | |
10458 | migr_rec->ckpt_area_pba = __cpu_to_le32(min_dev_sectors - | |
10459 | RAID_DISK_RESERVED_BLOCKS_IMSM_HI); | |
10460 | ||
10461 | write_imsm_migr_rec(st); | |
10462 | ||
10463 | return; | |
10464 | } | |
10465 | ||
10466 | /******************************************************************************* | |
10467 | * Function: save_backup_imsm | |
10468 | * Description: Function saves critical data stripes to Migration Copy Area | |
10469 | * and updates the current migration unit status. | |
10470 | * Use restore_stripes() to form a destination stripe, | |
10471 | * and to write it to the Copy Area. | |
10472 | * Parameters: | |
10473 | * st : supertype information | |
aea93171 | 10474 | * dev : imsm device that backup is saved for |
687629c2 AK |
10475 | * info : general array info |
10476 | * buf : input buffer | |
687629c2 AK |
10477 | * length : length of data to backup (blocks_per_unit) |
10478 | * Returns: | |
10479 | * 0 : success | |
10480 | *, -1 : fail | |
10481 | ******************************************************************************/ | |
10482 | int save_backup_imsm(struct supertype *st, | |
10483 | struct imsm_dev *dev, | |
10484 | struct mdinfo *info, | |
10485 | void *buf, | |
687629c2 AK |
10486 | int length) |
10487 | { | |
10488 | int rv = -1; | |
10489 | struct intel_super *super = st->sb; | |
594dc1b8 JS |
10490 | unsigned long long *target_offsets; |
10491 | int *targets; | |
687629c2 | 10492 | int i; |
238c0a71 | 10493 | struct imsm_map *map_dest = get_imsm_map(dev, MAP_0); |
687629c2 | 10494 | int new_disks = map_dest->num_members; |
ab724b98 AK |
10495 | int dest_layout = 0; |
10496 | int dest_chunk; | |
d1877f69 | 10497 | unsigned long long start; |
238c0a71 | 10498 | int data_disks = imsm_num_data_members(dev, MAP_0); |
687629c2 | 10499 | |
503975b9 | 10500 | targets = xmalloc(new_disks * sizeof(int)); |
687629c2 | 10501 | |
7e45b550 AK |
10502 | for (i = 0; i < new_disks; i++) |
10503 | targets[i] = -1; | |
10504 | ||
503975b9 | 10505 | target_offsets = xcalloc(new_disks, sizeof(unsigned long long)); |
687629c2 | 10506 | |
d1877f69 | 10507 | start = info->reshape_progress * 512; |
687629c2 | 10508 | for (i = 0; i < new_disks; i++) { |
687629c2 AK |
10509 | target_offsets[i] = (unsigned long long) |
10510 | __le32_to_cpu(super->migr_rec->ckpt_area_pba) * 512; | |
d1877f69 AK |
10511 | /* move back copy area adderss, it will be moved forward |
10512 | * in restore_stripes() using start input variable | |
10513 | */ | |
10514 | target_offsets[i] -= start/data_disks; | |
687629c2 AK |
10515 | } |
10516 | ||
9a717282 AK |
10517 | if (open_backup_targets(info, new_disks, targets, |
10518 | super, dev)) | |
687629c2 AK |
10519 | goto abort; |
10520 | ||
68eb8bc6 | 10521 | dest_layout = imsm_level_to_layout(map_dest->raid_level); |
ab724b98 AK |
10522 | dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512; |
10523 | ||
687629c2 AK |
10524 | if (restore_stripes(targets, /* list of dest devices */ |
10525 | target_offsets, /* migration record offsets */ | |
10526 | new_disks, | |
ab724b98 AK |
10527 | dest_chunk, |
10528 | map_dest->raid_level, | |
10529 | dest_layout, | |
10530 | -1, /* source backup file descriptor */ | |
10531 | 0, /* input buf offset | |
10532 | * always 0 buf is already offseted */ | |
d1877f69 | 10533 | start, |
687629c2 AK |
10534 | length, |
10535 | buf) != 0) { | |
e7b84f9d | 10536 | pr_err("Error restoring stripes\n"); |
687629c2 AK |
10537 | goto abort; |
10538 | } | |
10539 | ||
10540 | rv = 0; | |
10541 | ||
10542 | abort: | |
10543 | if (targets) { | |
9a717282 | 10544 | close_targets(targets, new_disks); |
687629c2 AK |
10545 | free(targets); |
10546 | } | |
10547 | free(target_offsets); | |
10548 | ||
10549 | return rv; | |
10550 | } | |
10551 | ||
10552 | /******************************************************************************* | |
10553 | * Function: save_checkpoint_imsm | |
10554 | * Description: Function called for current unit status update | |
10555 | * in the migration record. It writes it to disk. | |
10556 | * Parameters: | |
10557 | * super : imsm internal array info | |
10558 | * info : general array info | |
10559 | * Returns: | |
10560 | * 0: success | |
10561 | * 1: failure | |
0228d92c AK |
10562 | * 2: failure, means no valid migration record |
10563 | * / no general migration in progress / | |
687629c2 AK |
10564 | ******************************************************************************/ |
10565 | int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state) | |
10566 | { | |
10567 | struct intel_super *super = st->sb; | |
f8b72ef5 AK |
10568 | unsigned long long blocks_per_unit; |
10569 | unsigned long long curr_migr_unit; | |
10570 | ||
2e062e82 | 10571 | if (load_imsm_migr_rec(super, info) != 0) { |
7a862a02 | 10572 | dprintf("imsm: ERROR: Cannot read migration record for checkpoint save.\n"); |
2e062e82 AK |
10573 | return 1; |
10574 | } | |
10575 | ||
f8b72ef5 AK |
10576 | blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit); |
10577 | if (blocks_per_unit == 0) { | |
0228d92c AK |
10578 | dprintf("imsm: no migration in progress.\n"); |
10579 | return 2; | |
687629c2 | 10580 | } |
f8b72ef5 AK |
10581 | curr_migr_unit = info->reshape_progress / blocks_per_unit; |
10582 | /* check if array is alligned to copy area | |
10583 | * if it is not alligned, add one to current migration unit value | |
10584 | * this can happend on array reshape finish only | |
10585 | */ | |
10586 | if (info->reshape_progress % blocks_per_unit) | |
10587 | curr_migr_unit++; | |
687629c2 AK |
10588 | |
10589 | super->migr_rec->curr_migr_unit = | |
f8b72ef5 | 10590 | __cpu_to_le32(curr_migr_unit); |
687629c2 AK |
10591 | super->migr_rec->rec_status = __cpu_to_le32(state); |
10592 | super->migr_rec->dest_1st_member_lba = | |
f8b72ef5 AK |
10593 | __cpu_to_le32(curr_migr_unit * |
10594 | __le32_to_cpu(super->migr_rec->dest_depth_per_unit)); | |
687629c2 | 10595 | if (write_imsm_migr_rec(st) < 0) { |
7a862a02 | 10596 | dprintf("imsm: Cannot write migration record outside backup area\n"); |
687629c2 AK |
10597 | return 1; |
10598 | } | |
10599 | ||
10600 | return 0; | |
10601 | } | |
10602 | ||
276d77db AK |
10603 | /******************************************************************************* |
10604 | * Function: recover_backup_imsm | |
10605 | * Description: Function recovers critical data from the Migration Copy Area | |
10606 | * while assembling an array. | |
10607 | * Parameters: | |
10608 | * super : imsm internal array info | |
10609 | * info : general array info | |
10610 | * Returns: | |
10611 | * 0 : success (or there is no data to recover) | |
10612 | * 1 : fail | |
10613 | ******************************************************************************/ | |
10614 | int recover_backup_imsm(struct supertype *st, struct mdinfo *info) | |
10615 | { | |
10616 | struct intel_super *super = st->sb; | |
10617 | struct migr_record *migr_rec = super->migr_rec; | |
594dc1b8 | 10618 | struct imsm_map *map_dest; |
276d77db AK |
10619 | struct intel_dev *id = NULL; |
10620 | unsigned long long read_offset; | |
10621 | unsigned long long write_offset; | |
10622 | unsigned unit_len; | |
10623 | int *targets = NULL; | |
10624 | int new_disks, i, err; | |
10625 | char *buf = NULL; | |
10626 | int retval = 1; | |
f36a9ecd | 10627 | unsigned int sector_size = super->sector_size; |
276d77db AK |
10628 | unsigned long curr_migr_unit = __le32_to_cpu(migr_rec->curr_migr_unit); |
10629 | unsigned long num_migr_units = __le32_to_cpu(migr_rec->num_migr_units); | |
276d77db | 10630 | char buffer[20]; |
6c3560c0 | 10631 | int skipped_disks = 0; |
276d77db AK |
10632 | |
10633 | err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20); | |
10634 | if (err < 1) | |
10635 | return 1; | |
10636 | ||
10637 | /* recover data only during assemblation */ | |
10638 | if (strncmp(buffer, "inactive", 8) != 0) | |
10639 | return 0; | |
10640 | /* no data to recover */ | |
10641 | if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL) | |
10642 | return 0; | |
10643 | if (curr_migr_unit >= num_migr_units) | |
10644 | return 1; | |
10645 | ||
10646 | /* find device during reshape */ | |
10647 | for (id = super->devlist; id; id = id->next) | |
10648 | if (is_gen_migration(id->dev)) | |
10649 | break; | |
10650 | if (id == NULL) | |
10651 | return 1; | |
10652 | ||
238c0a71 | 10653 | map_dest = get_imsm_map(id->dev, MAP_0); |
276d77db AK |
10654 | new_disks = map_dest->num_members; |
10655 | ||
10656 | read_offset = (unsigned long long) | |
10657 | __le32_to_cpu(migr_rec->ckpt_area_pba) * 512; | |
10658 | ||
10659 | write_offset = ((unsigned long long) | |
10660 | __le32_to_cpu(migr_rec->dest_1st_member_lba) + | |
5551b113 | 10661 | pba_of_lba0(map_dest)) * 512; |
276d77db AK |
10662 | |
10663 | unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
f36a9ecd | 10664 | if (posix_memalign((void **)&buf, sector_size, unit_len) != 0) |
276d77db | 10665 | goto abort; |
503975b9 | 10666 | targets = xcalloc(new_disks, sizeof(int)); |
276d77db | 10667 | |
9a717282 | 10668 | if (open_backup_targets(info, new_disks, targets, super, id->dev)) { |
e7b84f9d | 10669 | pr_err("Cannot open some devices belonging to array.\n"); |
f627f5ad AK |
10670 | goto abort; |
10671 | } | |
276d77db AK |
10672 | |
10673 | for (i = 0; i < new_disks; i++) { | |
6c3560c0 AK |
10674 | if (targets[i] < 0) { |
10675 | skipped_disks++; | |
10676 | continue; | |
10677 | } | |
276d77db | 10678 | if (lseek64(targets[i], read_offset, SEEK_SET) < 0) { |
e7b84f9d N |
10679 | pr_err("Cannot seek to block: %s\n", |
10680 | strerror(errno)); | |
137debce AK |
10681 | skipped_disks++; |
10682 | continue; | |
276d77db | 10683 | } |
9ec11d1a | 10684 | if ((unsigned)read(targets[i], buf, unit_len) != unit_len) { |
e7b84f9d N |
10685 | pr_err("Cannot read copy area block: %s\n", |
10686 | strerror(errno)); | |
137debce AK |
10687 | skipped_disks++; |
10688 | continue; | |
276d77db AK |
10689 | } |
10690 | if (lseek64(targets[i], write_offset, SEEK_SET) < 0) { | |
e7b84f9d N |
10691 | pr_err("Cannot seek to block: %s\n", |
10692 | strerror(errno)); | |
137debce AK |
10693 | skipped_disks++; |
10694 | continue; | |
276d77db | 10695 | } |
9ec11d1a | 10696 | if ((unsigned)write(targets[i], buf, unit_len) != unit_len) { |
e7b84f9d N |
10697 | pr_err("Cannot restore block: %s\n", |
10698 | strerror(errno)); | |
137debce AK |
10699 | skipped_disks++; |
10700 | continue; | |
276d77db AK |
10701 | } |
10702 | } | |
10703 | ||
137debce AK |
10704 | if (skipped_disks > imsm_get_allowed_degradation(info->new_level, |
10705 | new_disks, | |
10706 | super, | |
10707 | id->dev)) { | |
7a862a02 | 10708 | pr_err("Cannot restore data from backup. Too many failed disks\n"); |
6c3560c0 AK |
10709 | goto abort; |
10710 | } | |
10711 | ||
befb629b AK |
10712 | if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) { |
10713 | /* ignore error == 2, this can mean end of reshape here | |
10714 | */ | |
7a862a02 | 10715 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL) during restart\n"); |
befb629b | 10716 | } else |
276d77db | 10717 | retval = 0; |
276d77db AK |
10718 | |
10719 | abort: | |
10720 | if (targets) { | |
10721 | for (i = 0; i < new_disks; i++) | |
10722 | if (targets[i]) | |
10723 | close(targets[i]); | |
10724 | free(targets); | |
10725 | } | |
10726 | free(buf); | |
10727 | return retval; | |
10728 | } | |
10729 | ||
2cda7640 ML |
10730 | static char disk_by_path[] = "/dev/disk/by-path/"; |
10731 | ||
10732 | static const char *imsm_get_disk_controller_domain(const char *path) | |
10733 | { | |
2cda7640 | 10734 | char disk_path[PATH_MAX]; |
96234762 LM |
10735 | char *drv=NULL; |
10736 | struct stat st; | |
2cda7640 | 10737 | |
6d8d290a | 10738 | strcpy(disk_path, disk_by_path); |
96234762 LM |
10739 | strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1); |
10740 | if (stat(disk_path, &st) == 0) { | |
10741 | struct sys_dev* hba; | |
594dc1b8 | 10742 | char *path; |
96234762 LM |
10743 | |
10744 | path = devt_to_devpath(st.st_rdev); | |
10745 | if (path == NULL) | |
10746 | return "unknown"; | |
10747 | hba = find_disk_attached_hba(-1, path); | |
10748 | if (hba && hba->type == SYS_DEV_SAS) | |
10749 | drv = "isci"; | |
10750 | else if (hba && hba->type == SYS_DEV_SATA) | |
10751 | drv = "ahci"; | |
1011e834 | 10752 | else |
96234762 LM |
10753 | drv = "unknown"; |
10754 | dprintf("path: %s hba: %s attached: %s\n", | |
10755 | path, (hba) ? hba->path : "NULL", drv); | |
10756 | free(path); | |
2cda7640 | 10757 | } |
96234762 | 10758 | return drv; |
2cda7640 ML |
10759 | } |
10760 | ||
4dd2df09 | 10761 | static char *imsm_find_array_devnm_by_subdev(int subdev, char *container) |
78b10e66 | 10762 | { |
4dd2df09 | 10763 | static char devnm[32]; |
78b10e66 N |
10764 | char subdev_name[20]; |
10765 | struct mdstat_ent *mdstat; | |
10766 | ||
10767 | sprintf(subdev_name, "%d", subdev); | |
10768 | mdstat = mdstat_by_subdev(subdev_name, container); | |
10769 | if (!mdstat) | |
4dd2df09 | 10770 | return NULL; |
78b10e66 | 10771 | |
4dd2df09 | 10772 | strcpy(devnm, mdstat->devnm); |
78b10e66 | 10773 | free_mdstat(mdstat); |
4dd2df09 | 10774 | return devnm; |
78b10e66 N |
10775 | } |
10776 | ||
10777 | static int imsm_reshape_is_allowed_on_container(struct supertype *st, | |
10778 | struct geo_params *geo, | |
fbf3d202 AK |
10779 | int *old_raid_disks, |
10780 | int direction) | |
78b10e66 | 10781 | { |
694575e7 KW |
10782 | /* currently we only support increasing the number of devices |
10783 | * for a container. This increases the number of device for each | |
10784 | * member array. They must all be RAID0 or RAID5. | |
10785 | */ | |
78b10e66 N |
10786 | int ret_val = 0; |
10787 | struct mdinfo *info, *member; | |
10788 | int devices_that_can_grow = 0; | |
10789 | ||
7a862a02 | 10790 | dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): st->devnm = (%s)\n", st->devnm); |
78b10e66 | 10791 | |
d04f65f4 | 10792 | if (geo->size > 0 || |
78b10e66 N |
10793 | geo->level != UnSet || |
10794 | geo->layout != UnSet || | |
10795 | geo->chunksize != 0 || | |
10796 | geo->raid_disks == UnSet) { | |
7a862a02 | 10797 | dprintf("imsm: Container operation is allowed for raid disks number change only.\n"); |
78b10e66 N |
10798 | return ret_val; |
10799 | } | |
10800 | ||
fbf3d202 | 10801 | if (direction == ROLLBACK_METADATA_CHANGES) { |
7a862a02 | 10802 | dprintf("imsm: Metadata changes rollback is not supported for container operation.\n"); |
fbf3d202 AK |
10803 | return ret_val; |
10804 | } | |
10805 | ||
78b10e66 N |
10806 | info = container_content_imsm(st, NULL); |
10807 | for (member = info; member; member = member->next) { | |
4dd2df09 | 10808 | char *result; |
78b10e66 N |
10809 | |
10810 | dprintf("imsm: checking device_num: %i\n", | |
10811 | member->container_member); | |
10812 | ||
d7d205bd | 10813 | if (geo->raid_disks <= member->array.raid_disks) { |
78b10e66 N |
10814 | /* we work on container for Online Capacity Expansion |
10815 | * only so raid_disks has to grow | |
10816 | */ | |
7a862a02 | 10817 | dprintf("imsm: for container operation raid disks increase is required\n"); |
78b10e66 N |
10818 | break; |
10819 | } | |
10820 | ||
089f9d79 | 10821 | if (info->array.level != 0 && info->array.level != 5) { |
78b10e66 N |
10822 | /* we cannot use this container with other raid level |
10823 | */ | |
7a862a02 | 10824 | dprintf("imsm: for container operation wrong raid level (%i) detected\n", |
78b10e66 N |
10825 | info->array.level); |
10826 | break; | |
10827 | } else { | |
10828 | /* check for platform support | |
10829 | * for this raid level configuration | |
10830 | */ | |
10831 | struct intel_super *super = st->sb; | |
10832 | if (!is_raid_level_supported(super->orom, | |
10833 | member->array.level, | |
10834 | geo->raid_disks)) { | |
7a862a02 | 10835 | dprintf("platform does not support raid%d with %d disk%s\n", |
78b10e66 N |
10836 | info->array.level, |
10837 | geo->raid_disks, | |
10838 | geo->raid_disks > 1 ? "s" : ""); | |
10839 | break; | |
10840 | } | |
2a4a08e7 AK |
10841 | /* check if component size is aligned to chunk size |
10842 | */ | |
10843 | if (info->component_size % | |
10844 | (info->array.chunk_size/512)) { | |
7a862a02 | 10845 | dprintf("Component size is not aligned to chunk size\n"); |
2a4a08e7 AK |
10846 | break; |
10847 | } | |
78b10e66 N |
10848 | } |
10849 | ||
10850 | if (*old_raid_disks && | |
10851 | info->array.raid_disks != *old_raid_disks) | |
10852 | break; | |
10853 | *old_raid_disks = info->array.raid_disks; | |
10854 | ||
10855 | /* All raid5 and raid0 volumes in container | |
10856 | * have to be ready for Online Capacity Expansion | |
10857 | * so they need to be assembled. We have already | |
10858 | * checked that no recovery etc is happening. | |
10859 | */ | |
4dd2df09 N |
10860 | result = imsm_find_array_devnm_by_subdev(member->container_member, |
10861 | st->container_devnm); | |
10862 | if (result == NULL) { | |
78b10e66 N |
10863 | dprintf("imsm: cannot find array\n"); |
10864 | break; | |
10865 | } | |
10866 | devices_that_can_grow++; | |
10867 | } | |
10868 | sysfs_free(info); | |
10869 | if (!member && devices_that_can_grow) | |
10870 | ret_val = 1; | |
10871 | ||
10872 | if (ret_val) | |
1ade5cc1 | 10873 | dprintf("Container operation allowed\n"); |
78b10e66 | 10874 | else |
1ade5cc1 | 10875 | dprintf("Error: %i\n", ret_val); |
78b10e66 N |
10876 | |
10877 | return ret_val; | |
10878 | } | |
10879 | ||
10880 | /* Function: get_spares_for_grow | |
10881 | * Description: Allocates memory and creates list of spare devices | |
1011e834 | 10882 | * avaliable in container. Checks if spare drive size is acceptable. |
78b10e66 N |
10883 | * Parameters: Pointer to the supertype structure |
10884 | * Returns: Pointer to the list of spare devices (mdinfo structure) on success, | |
1011e834 | 10885 | * NULL if fail |
78b10e66 N |
10886 | */ |
10887 | static struct mdinfo *get_spares_for_grow(struct supertype *st) | |
10888 | { | |
fbfdcb06 AO |
10889 | struct spare_criteria sc; |
10890 | ||
10891 | get_spare_criteria_imsm(st, &sc); | |
10892 | return container_choose_spares(st, &sc, NULL, NULL, NULL, 0); | |
78b10e66 N |
10893 | } |
10894 | ||
10895 | /****************************************************************************** | |
10896 | * function: imsm_create_metadata_update_for_reshape | |
10897 | * Function creates update for whole IMSM container. | |
10898 | * | |
10899 | ******************************************************************************/ | |
10900 | static int imsm_create_metadata_update_for_reshape( | |
10901 | struct supertype *st, | |
10902 | struct geo_params *geo, | |
10903 | int old_raid_disks, | |
10904 | struct imsm_update_reshape **updatep) | |
10905 | { | |
10906 | struct intel_super *super = st->sb; | |
10907 | struct imsm_super *mpb = super->anchor; | |
594dc1b8 JS |
10908 | int update_memory_size; |
10909 | struct imsm_update_reshape *u; | |
10910 | struct mdinfo *spares; | |
78b10e66 | 10911 | int i; |
594dc1b8 | 10912 | int delta_disks; |
bbd24d86 | 10913 | struct mdinfo *dev; |
78b10e66 | 10914 | |
1ade5cc1 | 10915 | dprintf("(enter) raid_disks = %i\n", geo->raid_disks); |
78b10e66 N |
10916 | |
10917 | delta_disks = geo->raid_disks - old_raid_disks; | |
10918 | ||
10919 | /* size of all update data without anchor */ | |
10920 | update_memory_size = sizeof(struct imsm_update_reshape); | |
10921 | ||
10922 | /* now add space for spare disks that we need to add. */ | |
10923 | update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1); | |
10924 | ||
503975b9 | 10925 | u = xcalloc(1, update_memory_size); |
78b10e66 N |
10926 | u->type = update_reshape_container_disks; |
10927 | u->old_raid_disks = old_raid_disks; | |
10928 | u->new_raid_disks = geo->raid_disks; | |
10929 | ||
10930 | /* now get spare disks list | |
10931 | */ | |
10932 | spares = get_spares_for_grow(st); | |
10933 | ||
d7be7d87 | 10934 | if (spares == NULL || delta_disks > spares->array.spare_disks) { |
7a862a02 | 10935 | pr_err("imsm: ERROR: Cannot get spare devices for %s.\n", geo->dev_name); |
e4c72d1d | 10936 | i = -1; |
78b10e66 N |
10937 | goto abort; |
10938 | } | |
10939 | ||
10940 | /* we have got spares | |
10941 | * update disk list in imsm_disk list table in anchor | |
10942 | */ | |
10943 | dprintf("imsm: %i spares are available.\n\n", | |
10944 | spares->array.spare_disks); | |
10945 | ||
bbd24d86 | 10946 | dev = spares->devs; |
78b10e66 | 10947 | for (i = 0; i < delta_disks; i++) { |
78b10e66 N |
10948 | struct dl *dl; |
10949 | ||
bbd24d86 AK |
10950 | if (dev == NULL) |
10951 | break; | |
78b10e66 N |
10952 | u->new_disks[i] = makedev(dev->disk.major, |
10953 | dev->disk.minor); | |
10954 | dl = get_disk_super(super, dev->disk.major, dev->disk.minor); | |
ee4beede AK |
10955 | dl->index = mpb->num_disks; |
10956 | mpb->num_disks++; | |
bbd24d86 | 10957 | dev = dev->next; |
78b10e66 | 10958 | } |
78b10e66 N |
10959 | |
10960 | abort: | |
10961 | /* free spares | |
10962 | */ | |
10963 | sysfs_free(spares); | |
10964 | ||
d677e0b8 | 10965 | dprintf("imsm: reshape update preparation :"); |
78b10e66 | 10966 | if (i == delta_disks) { |
1ade5cc1 | 10967 | dprintf_cont(" OK\n"); |
78b10e66 N |
10968 | *updatep = u; |
10969 | return update_memory_size; | |
10970 | } | |
10971 | free(u); | |
1ade5cc1 | 10972 | dprintf_cont(" Error\n"); |
78b10e66 N |
10973 | |
10974 | return 0; | |
10975 | } | |
10976 | ||
f3871fdc AK |
10977 | /****************************************************************************** |
10978 | * function: imsm_create_metadata_update_for_size_change() | |
10979 | * Creates update for IMSM array for array size change. | |
10980 | * | |
10981 | ******************************************************************************/ | |
10982 | static int imsm_create_metadata_update_for_size_change( | |
10983 | struct supertype *st, | |
10984 | struct geo_params *geo, | |
10985 | struct imsm_update_size_change **updatep) | |
10986 | { | |
10987 | struct intel_super *super = st->sb; | |
594dc1b8 JS |
10988 | int update_memory_size; |
10989 | struct imsm_update_size_change *u; | |
f3871fdc | 10990 | |
1ade5cc1 | 10991 | dprintf("(enter) New size = %llu\n", geo->size); |
f3871fdc AK |
10992 | |
10993 | /* size of all update data without anchor */ | |
10994 | update_memory_size = sizeof(struct imsm_update_size_change); | |
10995 | ||
503975b9 | 10996 | u = xcalloc(1, update_memory_size); |
f3871fdc AK |
10997 | u->type = update_size_change; |
10998 | u->subdev = super->current_vol; | |
10999 | u->new_size = geo->size; | |
11000 | ||
11001 | dprintf("imsm: reshape update preparation : OK\n"); | |
11002 | *updatep = u; | |
11003 | ||
11004 | return update_memory_size; | |
11005 | } | |
11006 | ||
48c5303a PC |
11007 | /****************************************************************************** |
11008 | * function: imsm_create_metadata_update_for_migration() | |
11009 | * Creates update for IMSM array. | |
11010 | * | |
11011 | ******************************************************************************/ | |
11012 | static int imsm_create_metadata_update_for_migration( | |
11013 | struct supertype *st, | |
11014 | struct geo_params *geo, | |
11015 | struct imsm_update_reshape_migration **updatep) | |
11016 | { | |
11017 | struct intel_super *super = st->sb; | |
594dc1b8 JS |
11018 | int update_memory_size; |
11019 | struct imsm_update_reshape_migration *u; | |
48c5303a PC |
11020 | struct imsm_dev *dev; |
11021 | int previous_level = -1; | |
11022 | ||
1ade5cc1 | 11023 | dprintf("(enter) New Level = %i\n", geo->level); |
48c5303a PC |
11024 | |
11025 | /* size of all update data without anchor */ | |
11026 | update_memory_size = sizeof(struct imsm_update_reshape_migration); | |
11027 | ||
503975b9 | 11028 | u = xcalloc(1, update_memory_size); |
48c5303a PC |
11029 | u->type = update_reshape_migration; |
11030 | u->subdev = super->current_vol; | |
11031 | u->new_level = geo->level; | |
11032 | u->new_layout = geo->layout; | |
11033 | u->new_raid_disks = u->old_raid_disks = geo->raid_disks; | |
11034 | u->new_disks[0] = -1; | |
4bba0439 | 11035 | u->new_chunksize = -1; |
48c5303a PC |
11036 | |
11037 | dev = get_imsm_dev(super, u->subdev); | |
11038 | if (dev) { | |
11039 | struct imsm_map *map; | |
11040 | ||
238c0a71 | 11041 | map = get_imsm_map(dev, MAP_0); |
4bba0439 PC |
11042 | if (map) { |
11043 | int current_chunk_size = | |
11044 | __le16_to_cpu(map->blocks_per_strip) / 2; | |
11045 | ||
11046 | if (geo->chunksize != current_chunk_size) { | |
11047 | u->new_chunksize = geo->chunksize / 1024; | |
7a862a02 | 11048 | dprintf("imsm: chunk size change from %i to %i\n", |
4bba0439 PC |
11049 | current_chunk_size, u->new_chunksize); |
11050 | } | |
48c5303a | 11051 | previous_level = map->raid_level; |
4bba0439 | 11052 | } |
48c5303a | 11053 | } |
089f9d79 | 11054 | if (geo->level == 5 && previous_level == 0) { |
48c5303a PC |
11055 | struct mdinfo *spares = NULL; |
11056 | ||
11057 | u->new_raid_disks++; | |
11058 | spares = get_spares_for_grow(st); | |
089f9d79 | 11059 | if (spares == NULL || spares->array.spare_disks < 1) { |
48c5303a PC |
11060 | free(u); |
11061 | sysfs_free(spares); | |
11062 | update_memory_size = 0; | |
565cc99e | 11063 | pr_err("cannot get spare device for requested migration\n"); |
48c5303a PC |
11064 | return 0; |
11065 | } | |
11066 | sysfs_free(spares); | |
11067 | } | |
11068 | dprintf("imsm: reshape update preparation : OK\n"); | |
11069 | *updatep = u; | |
11070 | ||
11071 | return update_memory_size; | |
11072 | } | |
11073 | ||
8dd70bce AK |
11074 | static void imsm_update_metadata_locally(struct supertype *st, |
11075 | void *buf, int len) | |
11076 | { | |
11077 | struct metadata_update mu; | |
11078 | ||
11079 | mu.buf = buf; | |
11080 | mu.len = len; | |
11081 | mu.space = NULL; | |
11082 | mu.space_list = NULL; | |
11083 | mu.next = NULL; | |
5fe6f031 N |
11084 | if (imsm_prepare_update(st, &mu)) |
11085 | imsm_process_update(st, &mu); | |
8dd70bce AK |
11086 | |
11087 | while (mu.space_list) { | |
11088 | void **space = mu.space_list; | |
11089 | mu.space_list = *space; | |
11090 | free(space); | |
11091 | } | |
11092 | } | |
78b10e66 | 11093 | |
471bceb6 | 11094 | /*************************************************************************** |
694575e7 | 11095 | * Function: imsm_analyze_change |
471bceb6 | 11096 | * Description: Function analyze change for single volume |
1011e834 | 11097 | * and validate if transition is supported |
fbf3d202 AK |
11098 | * Parameters: Geometry parameters, supertype structure, |
11099 | * metadata change direction (apply/rollback) | |
694575e7 | 11100 | * Returns: Operation type code on success, -1 if fail |
471bceb6 KW |
11101 | ****************************************************************************/ |
11102 | enum imsm_reshape_type imsm_analyze_change(struct supertype *st, | |
fbf3d202 AK |
11103 | struct geo_params *geo, |
11104 | int direction) | |
694575e7 | 11105 | { |
471bceb6 KW |
11106 | struct mdinfo info; |
11107 | int change = -1; | |
11108 | int check_devs = 0; | |
c21e737b | 11109 | int chunk; |
67a2db32 AK |
11110 | /* number of added/removed disks in operation result */ |
11111 | int devNumChange = 0; | |
11112 | /* imsm compatible layout value for array geometry verification */ | |
11113 | int imsm_layout = -1; | |
7abc9871 AK |
11114 | int data_disks; |
11115 | struct imsm_dev *dev; | |
11116 | struct intel_super *super; | |
d04f65f4 | 11117 | unsigned long long current_size; |
65d38cca | 11118 | unsigned long long free_size; |
d04f65f4 | 11119 | unsigned long long max_size; |
65d38cca | 11120 | int rv; |
471bceb6 KW |
11121 | |
11122 | getinfo_super_imsm_volume(st, &info, NULL); | |
089f9d79 JS |
11123 | if (geo->level != info.array.level && geo->level >= 0 && |
11124 | geo->level != UnSet) { | |
471bceb6 KW |
11125 | switch (info.array.level) { |
11126 | case 0: | |
11127 | if (geo->level == 5) { | |
b5347799 | 11128 | change = CH_MIGRATION; |
e13ce846 | 11129 | if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) { |
7a862a02 | 11130 | pr_err("Error. Requested Layout not supported (left-asymmetric layout is supported only)!\n"); |
e13ce846 AK |
11131 | change = -1; |
11132 | goto analyse_change_exit; | |
11133 | } | |
67a2db32 | 11134 | imsm_layout = geo->layout; |
471bceb6 | 11135 | check_devs = 1; |
e91a3bad LM |
11136 | devNumChange = 1; /* parity disk added */ |
11137 | } else if (geo->level == 10) { | |
471bceb6 KW |
11138 | change = CH_TAKEOVER; |
11139 | check_devs = 1; | |
e91a3bad | 11140 | devNumChange = 2; /* two mirrors added */ |
67a2db32 | 11141 | imsm_layout = 0x102; /* imsm supported layout */ |
471bceb6 | 11142 | } |
dfe77a9e KW |
11143 | break; |
11144 | case 1: | |
471bceb6 KW |
11145 | case 10: |
11146 | if (geo->level == 0) { | |
11147 | change = CH_TAKEOVER; | |
11148 | check_devs = 1; | |
e91a3bad | 11149 | devNumChange = -(geo->raid_disks/2); |
67a2db32 | 11150 | imsm_layout = 0; /* imsm raid0 layout */ |
471bceb6 KW |
11151 | } |
11152 | break; | |
11153 | } | |
11154 | if (change == -1) { | |
7a862a02 | 11155 | pr_err("Error. Level Migration from %d to %d not supported!\n", |
e7b84f9d | 11156 | info.array.level, geo->level); |
471bceb6 KW |
11157 | goto analyse_change_exit; |
11158 | } | |
11159 | } else | |
11160 | geo->level = info.array.level; | |
11161 | ||
089f9d79 JS |
11162 | if (geo->layout != info.array.layout && |
11163 | (geo->layout != UnSet && geo->layout != -1)) { | |
b5347799 | 11164 | change = CH_MIGRATION; |
089f9d79 JS |
11165 | if (info.array.layout == 0 && info.array.level == 5 && |
11166 | geo->layout == 5) { | |
471bceb6 | 11167 | /* reshape 5 -> 4 */ |
089f9d79 JS |
11168 | } else if (info.array.layout == 5 && info.array.level == 5 && |
11169 | geo->layout == 0) { | |
471bceb6 KW |
11170 | /* reshape 4 -> 5 */ |
11171 | geo->layout = 0; | |
11172 | geo->level = 5; | |
11173 | } else { | |
7a862a02 | 11174 | pr_err("Error. Layout Migration from %d to %d not supported!\n", |
e7b84f9d | 11175 | info.array.layout, geo->layout); |
471bceb6 KW |
11176 | change = -1; |
11177 | goto analyse_change_exit; | |
11178 | } | |
67a2db32 | 11179 | } else { |
471bceb6 | 11180 | geo->layout = info.array.layout; |
67a2db32 AK |
11181 | if (imsm_layout == -1) |
11182 | imsm_layout = info.array.layout; | |
11183 | } | |
471bceb6 | 11184 | |
089f9d79 JS |
11185 | if (geo->chunksize > 0 && geo->chunksize != UnSet && |
11186 | geo->chunksize != info.array.chunk_size) { | |
2d2b0eb7 MD |
11187 | if (info.array.level == 10) { |
11188 | pr_err("Error. Chunk size change for RAID 10 is not supported.\n"); | |
11189 | change = -1; | |
11190 | goto analyse_change_exit; | |
1e9b2c3f PB |
11191 | } else if (info.component_size % (geo->chunksize/512)) { |
11192 | pr_err("New chunk size (%dK) does not evenly divide device size (%lluk). Aborting...\n", | |
11193 | geo->chunksize/1024, info.component_size/2); | |
11194 | change = -1; | |
11195 | goto analyse_change_exit; | |
2d2b0eb7 | 11196 | } |
b5347799 | 11197 | change = CH_MIGRATION; |
2d2b0eb7 | 11198 | } else { |
471bceb6 | 11199 | geo->chunksize = info.array.chunk_size; |
2d2b0eb7 | 11200 | } |
471bceb6 | 11201 | |
c21e737b | 11202 | chunk = geo->chunksize / 1024; |
7abc9871 AK |
11203 | |
11204 | super = st->sb; | |
11205 | dev = get_imsm_dev(super, super->current_vol); | |
11206 | data_disks = imsm_num_data_members(dev , MAP_0); | |
c41e00b2 | 11207 | /* compute current size per disk member |
7abc9871 | 11208 | */ |
c41e00b2 AK |
11209 | current_size = info.custom_array_size / data_disks; |
11210 | ||
089f9d79 | 11211 | if (geo->size > 0 && geo->size != MAX_SIZE) { |
c41e00b2 AK |
11212 | /* align component size |
11213 | */ | |
11214 | geo->size = imsm_component_size_aligment_check( | |
11215 | get_imsm_raid_level(dev->vol.map), | |
f36a9ecd | 11216 | chunk * 1024, super->sector_size, |
c41e00b2 | 11217 | geo->size * 2); |
65d0b4ce | 11218 | if (geo->size == 0) { |
7a862a02 | 11219 | pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is 0).\n", |
65d0b4ce LD |
11220 | current_size); |
11221 | goto analyse_change_exit; | |
11222 | } | |
c41e00b2 | 11223 | } |
7abc9871 | 11224 | |
089f9d79 | 11225 | if (current_size != geo->size && geo->size > 0) { |
7abc9871 | 11226 | if (change != -1) { |
7a862a02 | 11227 | pr_err("Error. Size change should be the only one at a time.\n"); |
7abc9871 AK |
11228 | change = -1; |
11229 | goto analyse_change_exit; | |
11230 | } | |
11231 | if ((super->current_vol + 1) != super->anchor->num_raid_devs) { | |
7a862a02 | 11232 | pr_err("Error. The last volume in container can be expanded only (%i/%s).\n", |
4dd2df09 | 11233 | super->current_vol, st->devnm); |
7abc9871 AK |
11234 | goto analyse_change_exit; |
11235 | } | |
65d38cca LD |
11236 | /* check the maximum available size |
11237 | */ | |
11238 | rv = imsm_get_free_size(st, dev->vol.map->num_members, | |
11239 | 0, chunk, &free_size); | |
11240 | if (rv == 0) | |
11241 | /* Cannot find maximum available space | |
11242 | */ | |
11243 | max_size = 0; | |
11244 | else { | |
11245 | max_size = free_size + current_size; | |
11246 | /* align component size | |
11247 | */ | |
11248 | max_size = imsm_component_size_aligment_check( | |
11249 | get_imsm_raid_level(dev->vol.map), | |
f36a9ecd | 11250 | chunk * 1024, super->sector_size, |
65d38cca LD |
11251 | max_size); |
11252 | } | |
d04f65f4 | 11253 | if (geo->size == MAX_SIZE) { |
b130333f AK |
11254 | /* requested size change to the maximum available size |
11255 | */ | |
65d38cca | 11256 | if (max_size == 0) { |
7a862a02 | 11257 | pr_err("Error. Cannot find maximum available space.\n"); |
b130333f AK |
11258 | change = -1; |
11259 | goto analyse_change_exit; | |
65d38cca LD |
11260 | } else |
11261 | geo->size = max_size; | |
c41e00b2 | 11262 | } |
b130333f | 11263 | |
681b7ae2 | 11264 | if (direction == ROLLBACK_METADATA_CHANGES) { |
fbf3d202 AK |
11265 | /* accept size for rollback only |
11266 | */ | |
11267 | } else { | |
11268 | /* round size due to metadata compatibility | |
11269 | */ | |
11270 | geo->size = (geo->size >> SECT_PER_MB_SHIFT) | |
11271 | << SECT_PER_MB_SHIFT; | |
11272 | dprintf("Prepare update for size change to %llu\n", | |
11273 | geo->size ); | |
11274 | if (current_size >= geo->size) { | |
7a862a02 | 11275 | pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is %llu).\n", |
e7b84f9d | 11276 | current_size, geo->size); |
fbf3d202 AK |
11277 | goto analyse_change_exit; |
11278 | } | |
65d38cca | 11279 | if (max_size && geo->size > max_size) { |
7a862a02 | 11280 | pr_err("Error. Requested size is larger than maximum available size (maximum available size is %llu, requested size /rounded/ is %llu).\n", |
e7b84f9d | 11281 | max_size, geo->size); |
65d38cca LD |
11282 | goto analyse_change_exit; |
11283 | } | |
7abc9871 AK |
11284 | } |
11285 | geo->size *= data_disks; | |
11286 | geo->raid_disks = dev->vol.map->num_members; | |
11287 | change = CH_ARRAY_SIZE; | |
11288 | } | |
471bceb6 KW |
11289 | if (!validate_geometry_imsm(st, |
11290 | geo->level, | |
67a2db32 | 11291 | imsm_layout, |
e91a3bad | 11292 | geo->raid_disks + devNumChange, |
c21e737b | 11293 | &chunk, |
af4348dd | 11294 | geo->size, INVALID_SECTORS, |
5308f117 | 11295 | 0, 0, info.consistency_policy, 1)) |
471bceb6 KW |
11296 | change = -1; |
11297 | ||
11298 | if (check_devs) { | |
11299 | struct intel_super *super = st->sb; | |
11300 | struct imsm_super *mpb = super->anchor; | |
11301 | ||
11302 | if (mpb->num_raid_devs > 1) { | |
7a862a02 | 11303 | pr_err("Error. Cannot perform operation on %s- for this operation it MUST be single array in container\n", |
e7b84f9d | 11304 | geo->dev_name); |
471bceb6 KW |
11305 | change = -1; |
11306 | } | |
11307 | } | |
11308 | ||
11309 | analyse_change_exit: | |
089f9d79 JS |
11310 | if (direction == ROLLBACK_METADATA_CHANGES && |
11311 | (change == CH_MIGRATION || change == CH_TAKEOVER)) { | |
7a862a02 | 11312 | dprintf("imsm: Metadata changes rollback is not supported for migration and takeover operations.\n"); |
fbf3d202 AK |
11313 | change = -1; |
11314 | } | |
471bceb6 | 11315 | return change; |
694575e7 KW |
11316 | } |
11317 | ||
bb025c2f KW |
11318 | int imsm_takeover(struct supertype *st, struct geo_params *geo) |
11319 | { | |
11320 | struct intel_super *super = st->sb; | |
11321 | struct imsm_update_takeover *u; | |
11322 | ||
503975b9 | 11323 | u = xmalloc(sizeof(struct imsm_update_takeover)); |
bb025c2f KW |
11324 | |
11325 | u->type = update_takeover; | |
11326 | u->subarray = super->current_vol; | |
11327 | ||
11328 | /* 10->0 transition */ | |
11329 | if (geo->level == 0) | |
11330 | u->direction = R10_TO_R0; | |
11331 | ||
0529c688 KW |
11332 | /* 0->10 transition */ |
11333 | if (geo->level == 10) | |
11334 | u->direction = R0_TO_R10; | |
11335 | ||
bb025c2f KW |
11336 | /* update metadata locally */ |
11337 | imsm_update_metadata_locally(st, u, | |
11338 | sizeof(struct imsm_update_takeover)); | |
11339 | /* and possibly remotely */ | |
11340 | if (st->update_tail) | |
11341 | append_metadata_update(st, u, | |
11342 | sizeof(struct imsm_update_takeover)); | |
11343 | else | |
11344 | free(u); | |
11345 | ||
11346 | return 0; | |
11347 | } | |
11348 | ||
d04f65f4 N |
11349 | static int imsm_reshape_super(struct supertype *st, unsigned long long size, |
11350 | int level, | |
78b10e66 | 11351 | int layout, int chunksize, int raid_disks, |
41784c88 | 11352 | int delta_disks, char *backup, char *dev, |
016e00f5 | 11353 | int direction, int verbose) |
78b10e66 | 11354 | { |
78b10e66 N |
11355 | int ret_val = 1; |
11356 | struct geo_params geo; | |
11357 | ||
1ade5cc1 | 11358 | dprintf("(enter)\n"); |
78b10e66 | 11359 | |
71204a50 | 11360 | memset(&geo, 0, sizeof(struct geo_params)); |
78b10e66 N |
11361 | |
11362 | geo.dev_name = dev; | |
4dd2df09 | 11363 | strcpy(geo.devnm, st->devnm); |
78b10e66 N |
11364 | geo.size = size; |
11365 | geo.level = level; | |
11366 | geo.layout = layout; | |
11367 | geo.chunksize = chunksize; | |
11368 | geo.raid_disks = raid_disks; | |
41784c88 AK |
11369 | if (delta_disks != UnSet) |
11370 | geo.raid_disks += delta_disks; | |
78b10e66 | 11371 | |
1ade5cc1 N |
11372 | dprintf("for level : %i\n", geo.level); |
11373 | dprintf("for raid_disks : %i\n", geo.raid_disks); | |
78b10e66 N |
11374 | |
11375 | if (experimental() == 0) | |
11376 | return ret_val; | |
11377 | ||
4dd2df09 | 11378 | if (strcmp(st->container_devnm, st->devnm) == 0) { |
694575e7 KW |
11379 | /* On container level we can only increase number of devices. */ |
11380 | dprintf("imsm: info: Container operation\n"); | |
78b10e66 | 11381 | int old_raid_disks = 0; |
6dc0be30 | 11382 | |
78b10e66 | 11383 | if (imsm_reshape_is_allowed_on_container( |
fbf3d202 | 11384 | st, &geo, &old_raid_disks, direction)) { |
78b10e66 N |
11385 | struct imsm_update_reshape *u = NULL; |
11386 | int len; | |
11387 | ||
11388 | len = imsm_create_metadata_update_for_reshape( | |
11389 | st, &geo, old_raid_disks, &u); | |
11390 | ||
ed08d51c AK |
11391 | if (len <= 0) { |
11392 | dprintf("imsm: Cannot prepare update\n"); | |
11393 | goto exit_imsm_reshape_super; | |
11394 | } | |
11395 | ||
8dd70bce AK |
11396 | ret_val = 0; |
11397 | /* update metadata locally */ | |
11398 | imsm_update_metadata_locally(st, u, len); | |
11399 | /* and possibly remotely */ | |
11400 | if (st->update_tail) | |
11401 | append_metadata_update(st, u, len); | |
11402 | else | |
ed08d51c | 11403 | free(u); |
8dd70bce | 11404 | |
694575e7 | 11405 | } else { |
7a862a02 | 11406 | pr_err("(imsm) Operation is not allowed on this container\n"); |
694575e7 KW |
11407 | } |
11408 | } else { | |
11409 | /* On volume level we support following operations | |
471bceb6 KW |
11410 | * - takeover: raid10 -> raid0; raid0 -> raid10 |
11411 | * - chunk size migration | |
11412 | * - migration: raid5 -> raid0; raid0 -> raid5 | |
11413 | */ | |
11414 | struct intel_super *super = st->sb; | |
11415 | struct intel_dev *dev = super->devlist; | |
4dd2df09 | 11416 | int change; |
694575e7 | 11417 | dprintf("imsm: info: Volume operation\n"); |
471bceb6 KW |
11418 | /* find requested device */ |
11419 | while (dev) { | |
1011e834 | 11420 | char *devnm = |
4dd2df09 N |
11421 | imsm_find_array_devnm_by_subdev( |
11422 | dev->index, st->container_devnm); | |
11423 | if (devnm && strcmp(devnm, geo.devnm) == 0) | |
471bceb6 KW |
11424 | break; |
11425 | dev = dev->next; | |
11426 | } | |
11427 | if (dev == NULL) { | |
4dd2df09 N |
11428 | pr_err("Cannot find %s (%s) subarray\n", |
11429 | geo.dev_name, geo.devnm); | |
471bceb6 KW |
11430 | goto exit_imsm_reshape_super; |
11431 | } | |
11432 | super->current_vol = dev->index; | |
fbf3d202 | 11433 | change = imsm_analyze_change(st, &geo, direction); |
694575e7 | 11434 | switch (change) { |
471bceb6 | 11435 | case CH_TAKEOVER: |
bb025c2f | 11436 | ret_val = imsm_takeover(st, &geo); |
694575e7 | 11437 | break; |
48c5303a PC |
11438 | case CH_MIGRATION: { |
11439 | struct imsm_update_reshape_migration *u = NULL; | |
11440 | int len = | |
11441 | imsm_create_metadata_update_for_migration( | |
11442 | st, &geo, &u); | |
11443 | if (len < 1) { | |
7a862a02 | 11444 | dprintf("imsm: Cannot prepare update\n"); |
48c5303a PC |
11445 | break; |
11446 | } | |
471bceb6 | 11447 | ret_val = 0; |
48c5303a PC |
11448 | /* update metadata locally */ |
11449 | imsm_update_metadata_locally(st, u, len); | |
11450 | /* and possibly remotely */ | |
11451 | if (st->update_tail) | |
11452 | append_metadata_update(st, u, len); | |
11453 | else | |
11454 | free(u); | |
11455 | } | |
11456 | break; | |
7abc9871 | 11457 | case CH_ARRAY_SIZE: { |
f3871fdc AK |
11458 | struct imsm_update_size_change *u = NULL; |
11459 | int len = | |
11460 | imsm_create_metadata_update_for_size_change( | |
11461 | st, &geo, &u); | |
11462 | if (len < 1) { | |
7a862a02 | 11463 | dprintf("imsm: Cannot prepare update\n"); |
f3871fdc AK |
11464 | break; |
11465 | } | |
11466 | ret_val = 0; | |
11467 | /* update metadata locally */ | |
11468 | imsm_update_metadata_locally(st, u, len); | |
11469 | /* and possibly remotely */ | |
11470 | if (st->update_tail) | |
11471 | append_metadata_update(st, u, len); | |
11472 | else | |
11473 | free(u); | |
7abc9871 AK |
11474 | } |
11475 | break; | |
471bceb6 KW |
11476 | default: |
11477 | ret_val = 1; | |
694575e7 | 11478 | } |
694575e7 | 11479 | } |
78b10e66 | 11480 | |
ed08d51c | 11481 | exit_imsm_reshape_super: |
78b10e66 N |
11482 | dprintf("imsm: reshape_super Exit code = %i\n", ret_val); |
11483 | return ret_val; | |
11484 | } | |
2cda7640 | 11485 | |
0febb20c AO |
11486 | #define COMPLETED_OK 0 |
11487 | #define COMPLETED_NONE 1 | |
11488 | #define COMPLETED_DELAYED 2 | |
11489 | ||
11490 | static int read_completed(int fd, unsigned long long *val) | |
11491 | { | |
11492 | int ret; | |
11493 | char buf[50]; | |
11494 | ||
11495 | ret = sysfs_fd_get_str(fd, buf, 50); | |
11496 | if (ret < 0) | |
11497 | return ret; | |
11498 | ||
11499 | ret = COMPLETED_OK; | |
11500 | if (strncmp(buf, "none", 4) == 0) { | |
11501 | ret = COMPLETED_NONE; | |
11502 | } else if (strncmp(buf, "delayed", 7) == 0) { | |
11503 | ret = COMPLETED_DELAYED; | |
11504 | } else { | |
11505 | char *ep; | |
11506 | *val = strtoull(buf, &ep, 0); | |
11507 | if (ep == buf || (*ep != 0 && *ep != '\n' && *ep != ' ')) | |
11508 | ret = -1; | |
11509 | } | |
11510 | return ret; | |
11511 | } | |
11512 | ||
eee67a47 AK |
11513 | /******************************************************************************* |
11514 | * Function: wait_for_reshape_imsm | |
11515 | * Description: Function writes new sync_max value and waits until | |
11516 | * reshape process reach new position | |
11517 | * Parameters: | |
11518 | * sra : general array info | |
eee67a47 AK |
11519 | * ndata : number of disks in new array's layout |
11520 | * Returns: | |
11521 | * 0 : success, | |
11522 | * 1 : there is no reshape in progress, | |
11523 | * -1 : fail | |
11524 | ******************************************************************************/ | |
ae9f01f8 | 11525 | int wait_for_reshape_imsm(struct mdinfo *sra, int ndata) |
eee67a47 | 11526 | { |
85ca499c | 11527 | int fd = sysfs_get_fd(sra, NULL, "sync_completed"); |
df2647fa | 11528 | int retry = 3; |
eee67a47 | 11529 | unsigned long long completed; |
ae9f01f8 AK |
11530 | /* to_complete : new sync_max position */ |
11531 | unsigned long long to_complete = sra->reshape_progress; | |
11532 | unsigned long long position_to_set = to_complete / ndata; | |
eee67a47 | 11533 | |
ae9f01f8 | 11534 | if (fd < 0) { |
1ade5cc1 | 11535 | dprintf("cannot open reshape_position\n"); |
eee67a47 | 11536 | return 1; |
ae9f01f8 | 11537 | } |
eee67a47 | 11538 | |
df2647fa PB |
11539 | do { |
11540 | if (sysfs_fd_get_ll(fd, &completed) < 0) { | |
11541 | if (!retry) { | |
11542 | dprintf("cannot read reshape_position (no reshape in progres)\n"); | |
11543 | close(fd); | |
11544 | return 1; | |
11545 | } | |
11546 | usleep(30000); | |
11547 | } else | |
11548 | break; | |
11549 | } while (retry--); | |
eee67a47 | 11550 | |
85ca499c | 11551 | if (completed > position_to_set) { |
1ade5cc1 | 11552 | dprintf("wrong next position to set %llu (%llu)\n", |
85ca499c | 11553 | to_complete, position_to_set); |
ae9f01f8 AK |
11554 | close(fd); |
11555 | return -1; | |
11556 | } | |
11557 | dprintf("Position set: %llu\n", position_to_set); | |
11558 | if (sysfs_set_num(sra, NULL, "sync_max", | |
11559 | position_to_set) != 0) { | |
1ade5cc1 | 11560 | dprintf("cannot set reshape position to %llu\n", |
ae9f01f8 AK |
11561 | position_to_set); |
11562 | close(fd); | |
11563 | return -1; | |
eee67a47 AK |
11564 | } |
11565 | ||
eee67a47 | 11566 | do { |
0febb20c | 11567 | int rc; |
eee67a47 | 11568 | char action[20]; |
5ff3a780 | 11569 | int timeout = 3000; |
0febb20c | 11570 | |
5ff3a780 | 11571 | sysfs_wait(fd, &timeout); |
a47e44fb AK |
11572 | if (sysfs_get_str(sra, NULL, "sync_action", |
11573 | action, 20) > 0 && | |
d7d3809a | 11574 | strncmp(action, "reshape", 7) != 0) { |
b2be2b62 AO |
11575 | if (strncmp(action, "idle", 4) == 0) |
11576 | break; | |
d7d3809a AP |
11577 | close(fd); |
11578 | return -1; | |
11579 | } | |
0febb20c AO |
11580 | |
11581 | rc = read_completed(fd, &completed); | |
11582 | if (rc < 0) { | |
1ade5cc1 | 11583 | dprintf("cannot read reshape_position (in loop)\n"); |
eee67a47 AK |
11584 | close(fd); |
11585 | return 1; | |
0febb20c AO |
11586 | } else if (rc == COMPLETED_NONE) |
11587 | break; | |
85ca499c | 11588 | } while (completed < position_to_set); |
b2be2b62 | 11589 | |
eee67a47 AK |
11590 | close(fd); |
11591 | return 0; | |
eee67a47 AK |
11592 | } |
11593 | ||
b915c95f AK |
11594 | /******************************************************************************* |
11595 | * Function: check_degradation_change | |
11596 | * Description: Check that array hasn't become failed. | |
11597 | * Parameters: | |
11598 | * info : for sysfs access | |
11599 | * sources : source disks descriptors | |
11600 | * degraded: previous degradation level | |
11601 | * Returns: | |
11602 | * degradation level | |
11603 | ******************************************************************************/ | |
11604 | int check_degradation_change(struct mdinfo *info, | |
11605 | int *sources, | |
11606 | int degraded) | |
11607 | { | |
11608 | unsigned long long new_degraded; | |
e1993023 LD |
11609 | int rv; |
11610 | ||
11611 | rv = sysfs_get_ll(info, NULL, "degraded", &new_degraded); | |
089f9d79 | 11612 | if (rv == -1 || (new_degraded != (unsigned long long)degraded)) { |
b915c95f AK |
11613 | /* check each device to ensure it is still working */ |
11614 | struct mdinfo *sd; | |
11615 | new_degraded = 0; | |
11616 | for (sd = info->devs ; sd ; sd = sd->next) { | |
11617 | if (sd->disk.state & (1<<MD_DISK_FAULTY)) | |
11618 | continue; | |
11619 | if (sd->disk.state & (1<<MD_DISK_SYNC)) { | |
cf52eff5 TM |
11620 | char sbuf[100]; |
11621 | ||
b915c95f | 11622 | if (sysfs_get_str(info, |
cf52eff5 | 11623 | sd, "state", sbuf, sizeof(sbuf)) < 0 || |
b915c95f AK |
11624 | strstr(sbuf, "faulty") || |
11625 | strstr(sbuf, "in_sync") == NULL) { | |
11626 | /* this device is dead */ | |
11627 | sd->disk.state = (1<<MD_DISK_FAULTY); | |
11628 | if (sd->disk.raid_disk >= 0 && | |
11629 | sources[sd->disk.raid_disk] >= 0) { | |
11630 | close(sources[ | |
11631 | sd->disk.raid_disk]); | |
11632 | sources[sd->disk.raid_disk] = | |
11633 | -1; | |
11634 | } | |
11635 | new_degraded++; | |
11636 | } | |
11637 | } | |
11638 | } | |
11639 | } | |
11640 | ||
11641 | return new_degraded; | |
11642 | } | |
11643 | ||
10f22854 AK |
11644 | /******************************************************************************* |
11645 | * Function: imsm_manage_reshape | |
11646 | * Description: Function finds array under reshape and it manages reshape | |
11647 | * process. It creates stripes backups (if required) and sets | |
942e1cdb | 11648 | * checkpoints. |
10f22854 AK |
11649 | * Parameters: |
11650 | * afd : Backup handle (nattive) - not used | |
11651 | * sra : general array info | |
11652 | * reshape : reshape parameters - not used | |
11653 | * st : supertype structure | |
11654 | * blocks : size of critical section [blocks] | |
11655 | * fds : table of source device descriptor | |
11656 | * offsets : start of array (offest per devices) | |
11657 | * dests : not used | |
11658 | * destfd : table of destination device descriptor | |
11659 | * destoffsets : table of destination offsets (per device) | |
11660 | * Returns: | |
11661 | * 1 : success, reshape is done | |
11662 | * 0 : fail | |
11663 | ******************************************************************************/ | |
999b4972 N |
11664 | static int imsm_manage_reshape( |
11665 | int afd, struct mdinfo *sra, struct reshape *reshape, | |
10f22854 | 11666 | struct supertype *st, unsigned long backup_blocks, |
999b4972 N |
11667 | int *fds, unsigned long long *offsets, |
11668 | int dests, int *destfd, unsigned long long *destoffsets) | |
11669 | { | |
10f22854 AK |
11670 | int ret_val = 0; |
11671 | struct intel_super *super = st->sb; | |
594dc1b8 | 11672 | struct intel_dev *dv; |
de44e46f | 11673 | unsigned int sector_size = super->sector_size; |
10f22854 | 11674 | struct imsm_dev *dev = NULL; |
a6b6d984 | 11675 | struct imsm_map *map_src; |
10f22854 AK |
11676 | int migr_vol_qan = 0; |
11677 | int ndata, odata; /* [bytes] */ | |
11678 | int chunk; /* [bytes] */ | |
11679 | struct migr_record *migr_rec; | |
11680 | char *buf = NULL; | |
11681 | unsigned int buf_size; /* [bytes] */ | |
11682 | unsigned long long max_position; /* array size [bytes] */ | |
11683 | unsigned long long next_step; /* [blocks]/[bytes] */ | |
11684 | unsigned long long old_data_stripe_length; | |
10f22854 AK |
11685 | unsigned long long start_src; /* [bytes] */ |
11686 | unsigned long long start; /* [bytes] */ | |
11687 | unsigned long long start_buf_shift; /* [bytes] */ | |
b915c95f | 11688 | int degraded = 0; |
ab724b98 | 11689 | int source_layout = 0; |
10f22854 | 11690 | |
79a16a9b JS |
11691 | if (!sra) |
11692 | return ret_val; | |
11693 | ||
11694 | if (!fds || !offsets) | |
10f22854 AK |
11695 | goto abort; |
11696 | ||
11697 | /* Find volume during the reshape */ | |
11698 | for (dv = super->devlist; dv; dv = dv->next) { | |
fc54fe7a JS |
11699 | if (dv->dev->vol.migr_type == MIGR_GEN_MIGR && |
11700 | dv->dev->vol.migr_state == 1) { | |
10f22854 AK |
11701 | dev = dv->dev; |
11702 | migr_vol_qan++; | |
11703 | } | |
11704 | } | |
11705 | /* Only one volume can migrate at the same time */ | |
11706 | if (migr_vol_qan != 1) { | |
676e87a8 | 11707 | pr_err("%s", migr_vol_qan ? |
10f22854 AK |
11708 | "Number of migrating volumes greater than 1\n" : |
11709 | "There is no volume during migrationg\n"); | |
11710 | goto abort; | |
11711 | } | |
11712 | ||
238c0a71 | 11713 | map_src = get_imsm_map(dev, MAP_1); |
10f22854 AK |
11714 | if (map_src == NULL) |
11715 | goto abort; | |
10f22854 | 11716 | |
238c0a71 AK |
11717 | ndata = imsm_num_data_members(dev, MAP_0); |
11718 | odata = imsm_num_data_members(dev, MAP_1); | |
10f22854 | 11719 | |
7b1ab482 | 11720 | chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512; |
10f22854 AK |
11721 | old_data_stripe_length = odata * chunk; |
11722 | ||
11723 | migr_rec = super->migr_rec; | |
11724 | ||
10f22854 AK |
11725 | /* initialize migration record for start condition */ |
11726 | if (sra->reshape_progress == 0) | |
11727 | init_migr_record_imsm(st, dev, sra); | |
b2c59438 AK |
11728 | else { |
11729 | if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) { | |
7a862a02 | 11730 | dprintf("imsm: cannot restart migration when data are present in copy area.\n"); |
b2c59438 AK |
11731 | goto abort; |
11732 | } | |
6a75c8ca AK |
11733 | /* Save checkpoint to update migration record for current |
11734 | * reshape position (in md). It can be farther than current | |
11735 | * reshape position in metadata. | |
11736 | */ | |
11737 | if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) { | |
11738 | /* ignore error == 2, this can mean end of reshape here | |
11739 | */ | |
7a862a02 | 11740 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL, initial save)\n"); |
6a75c8ca AK |
11741 | goto abort; |
11742 | } | |
b2c59438 | 11743 | } |
10f22854 AK |
11744 | |
11745 | /* size for data */ | |
11746 | buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512; | |
11747 | /* extend buffer size for parity disk */ | |
11748 | buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512; | |
11749 | /* add space for stripe aligment */ | |
11750 | buf_size += old_data_stripe_length; | |
de44e46f PB |
11751 | if (posix_memalign((void **)&buf, MAX_SECTOR_SIZE, buf_size)) { |
11752 | dprintf("imsm: Cannot allocate checkpoint buffer\n"); | |
10f22854 AK |
11753 | goto abort; |
11754 | } | |
11755 | ||
3ef4403c | 11756 | max_position = sra->component_size * ndata; |
68eb8bc6 | 11757 | source_layout = imsm_level_to_layout(map_src->raid_level); |
10f22854 AK |
11758 | |
11759 | while (__le32_to_cpu(migr_rec->curr_migr_unit) < | |
11760 | __le32_to_cpu(migr_rec->num_migr_units)) { | |
11761 | /* current reshape position [blocks] */ | |
11762 | unsigned long long current_position = | |
11763 | __le32_to_cpu(migr_rec->blocks_per_unit) | |
11764 | * __le32_to_cpu(migr_rec->curr_migr_unit); | |
11765 | unsigned long long border; | |
11766 | ||
b915c95f AK |
11767 | /* Check that array hasn't become failed. |
11768 | */ | |
11769 | degraded = check_degradation_change(sra, fds, degraded); | |
11770 | if (degraded > 1) { | |
7a862a02 | 11771 | dprintf("imsm: Abort reshape due to degradation level (%i)\n", degraded); |
b915c95f AK |
11772 | goto abort; |
11773 | } | |
11774 | ||
10f22854 AK |
11775 | next_step = __le32_to_cpu(migr_rec->blocks_per_unit); |
11776 | ||
11777 | if ((current_position + next_step) > max_position) | |
11778 | next_step = max_position - current_position; | |
11779 | ||
92144abf | 11780 | start = current_position * 512; |
10f22854 | 11781 | |
942e1cdb | 11782 | /* align reading start to old geometry */ |
10f22854 AK |
11783 | start_buf_shift = start % old_data_stripe_length; |
11784 | start_src = start - start_buf_shift; | |
11785 | ||
11786 | border = (start_src / odata) - (start / ndata); | |
11787 | border /= 512; | |
11788 | if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) { | |
11789 | /* save critical stripes to buf | |
11790 | * start - start address of current unit | |
11791 | * to backup [bytes] | |
11792 | * start_src - start address of current unit | |
11793 | * to backup alligned to source array | |
11794 | * [bytes] | |
11795 | */ | |
594dc1b8 | 11796 | unsigned long long next_step_filler; |
10f22854 AK |
11797 | unsigned long long copy_length = next_step * 512; |
11798 | ||
11799 | /* allign copy area length to stripe in old geometry */ | |
11800 | next_step_filler = ((copy_length + start_buf_shift) | |
11801 | % old_data_stripe_length); | |
11802 | if (next_step_filler) | |
11803 | next_step_filler = (old_data_stripe_length | |
11804 | - next_step_filler); | |
7a862a02 | 11805 | 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 |
11806 | start, start_src, copy_length, |
11807 | start_buf_shift, next_step_filler); | |
11808 | ||
11809 | if (save_stripes(fds, offsets, map_src->num_members, | |
ab724b98 AK |
11810 | chunk, map_src->raid_level, |
11811 | source_layout, 0, NULL, start_src, | |
10f22854 AK |
11812 | copy_length + |
11813 | next_step_filler + start_buf_shift, | |
11814 | buf)) { | |
7a862a02 | 11815 | dprintf("imsm: Cannot save stripes to buffer\n"); |
10f22854 AK |
11816 | goto abort; |
11817 | } | |
11818 | /* Convert data to destination format and store it | |
11819 | * in backup general migration area | |
11820 | */ | |
11821 | if (save_backup_imsm(st, dev, sra, | |
aea93171 | 11822 | buf + start_buf_shift, copy_length)) { |
7a862a02 | 11823 | dprintf("imsm: Cannot save stripes to target devices\n"); |
10f22854 AK |
11824 | goto abort; |
11825 | } | |
11826 | if (save_checkpoint_imsm(st, sra, | |
11827 | UNIT_SRC_IN_CP_AREA)) { | |
7a862a02 | 11828 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_IN_CP_AREA)\n"); |
10f22854 AK |
11829 | goto abort; |
11830 | } | |
8016a6d4 AK |
11831 | } else { |
11832 | /* set next step to use whole border area */ | |
11833 | border /= next_step; | |
11834 | if (border > 1) | |
11835 | next_step *= border; | |
10f22854 AK |
11836 | } |
11837 | /* When data backed up, checkpoint stored, | |
11838 | * kick the kernel to reshape unit of data | |
11839 | */ | |
11840 | next_step = next_step + sra->reshape_progress; | |
8016a6d4 AK |
11841 | /* limit next step to array max position */ |
11842 | if (next_step > max_position) | |
11843 | next_step = max_position; | |
10f22854 AK |
11844 | sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress); |
11845 | sysfs_set_num(sra, NULL, "suspend_hi", next_step); | |
ae9f01f8 | 11846 | sra->reshape_progress = next_step; |
10f22854 AK |
11847 | |
11848 | /* wait until reshape finish */ | |
c85338c6 | 11849 | if (wait_for_reshape_imsm(sra, ndata)) { |
c47b0ff6 AK |
11850 | dprintf("wait_for_reshape_imsm returned error!\n"); |
11851 | goto abort; | |
11852 | } | |
84d11e6c N |
11853 | if (sigterm) |
11854 | goto abort; | |
10f22854 | 11855 | |
0228d92c AK |
11856 | if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) { |
11857 | /* ignore error == 2, this can mean end of reshape here | |
11858 | */ | |
7a862a02 | 11859 | dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL)\n"); |
10f22854 AK |
11860 | goto abort; |
11861 | } | |
11862 | ||
11863 | } | |
11864 | ||
71e5411e PB |
11865 | /* clear migr_rec on disks after successful migration */ |
11866 | struct dl *d; | |
11867 | ||
85337573 | 11868 | memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE); |
71e5411e PB |
11869 | for (d = super->disks; d; d = d->next) { |
11870 | if (d->index < 0 || is_failed(&d->disk)) | |
11871 | continue; | |
11872 | unsigned long long dsize; | |
11873 | ||
11874 | get_dev_size(d->fd, NULL, &dsize); | |
de44e46f | 11875 | if (lseek64(d->fd, dsize - MIGR_REC_SECTOR_POSITION*sector_size, |
71e5411e | 11876 | SEEK_SET) >= 0) { |
466070ad | 11877 | if ((unsigned int)write(d->fd, super->migr_rec_buf, |
de44e46f PB |
11878 | MIGR_REC_BUF_SECTORS*sector_size) != |
11879 | MIGR_REC_BUF_SECTORS*sector_size) | |
71e5411e PB |
11880 | perror("Write migr_rec failed"); |
11881 | } | |
11882 | } | |
11883 | ||
10f22854 AK |
11884 | /* return '1' if done */ |
11885 | ret_val = 1; | |
11886 | abort: | |
11887 | free(buf); | |
942e1cdb N |
11888 | /* See Grow.c: abort_reshape() for further explanation */ |
11889 | sysfs_set_num(sra, NULL, "suspend_lo", 0x7FFFFFFFFFFFFFFFULL); | |
11890 | sysfs_set_num(sra, NULL, "suspend_hi", 0); | |
11891 | sysfs_set_num(sra, NULL, "suspend_lo", 0); | |
10f22854 AK |
11892 | |
11893 | return ret_val; | |
999b4972 | 11894 | } |
0c21b485 | 11895 | |
cdddbdbc | 11896 | struct superswitch super_imsm = { |
cdddbdbc DW |
11897 | .examine_super = examine_super_imsm, |
11898 | .brief_examine_super = brief_examine_super_imsm, | |
4737ae25 | 11899 | .brief_examine_subarrays = brief_examine_subarrays_imsm, |
9d84c8ea | 11900 | .export_examine_super = export_examine_super_imsm, |
cdddbdbc DW |
11901 | .detail_super = detail_super_imsm, |
11902 | .brief_detail_super = brief_detail_super_imsm, | |
bf5a934a | 11903 | .write_init_super = write_init_super_imsm, |
0e600426 N |
11904 | .validate_geometry = validate_geometry_imsm, |
11905 | .add_to_super = add_to_super_imsm, | |
1a64be56 | 11906 | .remove_from_super = remove_from_super_imsm, |
d665cc31 | 11907 | .detail_platform = detail_platform_imsm, |
e50cf220 | 11908 | .export_detail_platform = export_detail_platform_imsm, |
33414a01 | 11909 | .kill_subarray = kill_subarray_imsm, |
aa534678 | 11910 | .update_subarray = update_subarray_imsm, |
2b959fbf | 11911 | .load_container = load_container_imsm, |
71204a50 N |
11912 | .default_geometry = default_geometry_imsm, |
11913 | .get_disk_controller_domain = imsm_get_disk_controller_domain, | |
11914 | .reshape_super = imsm_reshape_super, | |
11915 | .manage_reshape = imsm_manage_reshape, | |
9e2d750d | 11916 | .recover_backup = recover_backup_imsm, |
74db60b0 | 11917 | .copy_metadata = copy_metadata_imsm, |
27156a57 | 11918 | .examine_badblocks = examine_badblocks_imsm, |
cdddbdbc DW |
11919 | .match_home = match_home_imsm, |
11920 | .uuid_from_super= uuid_from_super_imsm, | |
11921 | .getinfo_super = getinfo_super_imsm, | |
5c4cd5da | 11922 | .getinfo_super_disks = getinfo_super_disks_imsm, |
cdddbdbc DW |
11923 | .update_super = update_super_imsm, |
11924 | ||
11925 | .avail_size = avail_size_imsm, | |
fbfdcb06 | 11926 | .get_spare_criteria = get_spare_criteria_imsm, |
cdddbdbc DW |
11927 | |
11928 | .compare_super = compare_super_imsm, | |
11929 | ||
11930 | .load_super = load_super_imsm, | |
bf5a934a | 11931 | .init_super = init_super_imsm, |
e683ca88 | 11932 | .store_super = store_super_imsm, |
cdddbdbc DW |
11933 | .free_super = free_super_imsm, |
11934 | .match_metadata_desc = match_metadata_desc_imsm, | |
bf5a934a | 11935 | .container_content = container_content_imsm, |
0c21b485 | 11936 | .validate_container = validate_container_imsm, |
cdddbdbc | 11937 | |
2432ce9b AP |
11938 | .write_init_ppl = write_init_ppl_imsm, |
11939 | .validate_ppl = validate_ppl_imsm, | |
11940 | ||
cdddbdbc | 11941 | .external = 1, |
4cce4069 | 11942 | .name = "imsm", |
845dea95 NB |
11943 | |
11944 | /* for mdmon */ | |
11945 | .open_new = imsm_open_new, | |
ed9d66aa | 11946 | .set_array_state= imsm_set_array_state, |
845dea95 NB |
11947 | .set_disk = imsm_set_disk, |
11948 | .sync_metadata = imsm_sync_metadata, | |
88758e9d | 11949 | .activate_spare = imsm_activate_spare, |
e8319a19 | 11950 | .process_update = imsm_process_update, |
8273f55e | 11951 | .prepare_update = imsm_prepare_update, |
6f50473f | 11952 | .record_bad_block = imsm_record_badblock, |
c07a5a4f | 11953 | .clear_bad_block = imsm_clear_badblock, |
928f1424 | 11954 | .get_bad_blocks = imsm_get_badblocks, |
cdddbdbc | 11955 | }; |