]> git.ipfire.org Git - thirdparty/mdadm.git/blame - super-intel.c
projects / thirdparty / mdadm.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Fix misspelling of 'alignment' and 'geometry'
[thirdparty/mdadm.git] / super-intel.c
CommitLineData
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
611d9529
MD		 91#define IMSM_RESERVED_SECTORS 8192
92#define NUM_BLOCKS_DIRTY_STRIPE_REGION 2048
979d38be 93#define SECT_PER_MB_SHIFT 11
f36a9ecd 94#define MAX_SECTOR_SIZE 4096
c2462068
PB		 95#define MULTIPLE_PPL_AREA_SIZE_IMSM (1024 * 1024) /* Size of the whole
96 * mutliple PPL area
97 */
cdddbdbc
DW		 98
99/* Disk configuration info. */
100#define IMSM_MAX_DEVICES 255
101struct imsm_disk {
102 __u8 serial[MAX_RAID_SERIAL_LEN];/* 0xD8 - 0xE7 ascii serial number */
5551b113 103 __u32 total_blocks_lo; /* 0xE8 - 0xEB total blocks lo */
cdddbdbc 104 __u32 scsi_id; /* 0xEC - 0xEF scsi ID */
f2f27e63
DW		 105#define SPARE_DISK __cpu_to_le32(0x01) /* Spare */
106#define CONFIGURED_DISK __cpu_to_le32(0x02) /* Member of some RaidDev */
107#define FAILED_DISK __cpu_to_le32(0x04) /* Permanent failure */
2432ce9b 108#define JOURNAL_DISK __cpu_to_le32(0x2000000) /* Device marked as Journaling Drive */
cdddbdbc 109 __u32 status; /* 0xF0 - 0xF3 */
1011e834 110 __u32 owner_cfg_num; /* which config 0,1,2... owns this disk */
5551b113
CA		 111 __u32 total_blocks_hi; /* 0xF4 - 0xF5 total blocks hi */
112#define IMSM_DISK_FILLERS 3
113 __u32 filler[IMSM_DISK_FILLERS]; /* 0xF5 - 0x107 MPB_DISK_FILLERS for future expansion */
cdddbdbc
DW		 114};
115
3b451610
AK		 116/* map selector for map managment
117 */
238c0a71
AK		 118#define MAP_0 0
119#define MAP_1 1
120#define MAP_X -1
3b451610 121
cdddbdbc
DW		 122/* RAID map configuration infos. */
123struct imsm_map {
5551b113
CA		 124 __u32 pba_of_lba0_lo; /* start address of partition */
125 __u32 blocks_per_member_lo;/* blocks per member */
126 __u32 num_data_stripes_lo; /* number of data stripes */
cdddbdbc
DW		 127 __u16 blocks_per_strip;
128 __u8 map_state; /* Normal, Uninitialized, Degraded, Failed */
129#define IMSM_T_STATE_NORMAL 0
130#define IMSM_T_STATE_UNINITIALIZED 1
e3bba0e0
DW		 131#define IMSM_T_STATE_DEGRADED 2
132#define IMSM_T_STATE_FAILED 3
cdddbdbc
DW		 133 __u8 raid_level;
134#define IMSM_T_RAID0 0
135#define IMSM_T_RAID1 1
136#define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */
137 __u8 num_members; /* number of member disks */
fe7ed8cb
DW		 138 __u8 num_domains; /* number of parity domains */
139 __u8 failed_disk_num; /* valid only when state is degraded */
252d23c0 140 __u8 ddf;
5551b113
CA		 141 __u32 pba_of_lba0_hi;
142 __u32 blocks_per_member_hi;
143 __u32 num_data_stripes_hi;
144 __u32 filler[4]; /* expansion area */
7eef0453 145#define IMSM_ORD_REBUILD (1 << 24)
cdddbdbc 146 __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members],
7eef0453
DW		 147 * top byte contains some flags
148 */
cdddbdbc
DW		 149} __attribute__ ((packed));
150
151struct imsm_vol {
f8f603f1 152 __u32 curr_migr_unit;
fe7ed8cb 153 __u32 checkpoint_id; /* id to access curr_migr_unit */
cdddbdbc 154 __u8 migr_state; /* Normal or Migrating */
e3bba0e0
DW		 155#define MIGR_INIT 0
156#define MIGR_REBUILD 1
157#define MIGR_VERIFY 2 /* analagous to echo check > sync_action */
158#define MIGR_GEN_MIGR 3
159#define MIGR_STATE_CHANGE 4
1484e727 160#define MIGR_REPAIR 5
cdddbdbc 161 __u8 migr_type; /* Initializing, Rebuilding, ... */
2432ce9b
AP		 162#define RAIDVOL_CLEAN 0
163#define RAIDVOL_DIRTY 1
164#define RAIDVOL_DSRECORD_VALID 2
cdddbdbc 165 __u8 dirty;
fe7ed8cb
DW		 166 __u8 fs_state; /* fast-sync state for CnG (0xff == disabled) */
167 __u16 verify_errors; /* number of mismatches */
168 __u16 bad_blocks; /* number of bad blocks during verify */
169 __u32 filler[4];
cdddbdbc
DW		 170 struct imsm_map map[1];
171 /* here comes another one if migr_state */
172} __attribute__ ((packed));
173
174struct imsm_dev {
fe7ed8cb 175 __u8 volume[MAX_RAID_SERIAL_LEN];
cdddbdbc
DW		 176 __u32 size_low;
177 __u32 size_high;
fe7ed8cb
DW		 178#define DEV_BOOTABLE __cpu_to_le32(0x01)
179#define DEV_BOOT_DEVICE __cpu_to_le32(0x02)
180#define DEV_READ_COALESCING __cpu_to_le32(0x04)
181#define DEV_WRITE_COALESCING __cpu_to_le32(0x08)
182#define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10)
183#define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20)
184#define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40)
185#define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80)
186#define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100)
187#define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200)
188#define DEV_CLONE_N_GO __cpu_to_le32(0x400)
189#define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800)
190#define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000)
cdddbdbc
DW		 191 __u32 status; /* Persistent RaidDev status */
192 __u32 reserved_blocks; /* Reserved blocks at beginning of volume */
fe7ed8cb
DW		 193 __u8 migr_priority;
194 __u8 num_sub_vols;
195 __u8 tid;
196 __u8 cng_master_disk;
197 __u16 cache_policy;
198 __u8 cng_state;
199 __u8 cng_sub_state;
2432ce9b
AP		 200 __u16 my_vol_raid_dev_num; /* Used in Unique volume Id for this RaidDev */
201
202 /* NVM_EN */
203 __u8 nv_cache_mode;
204 __u8 nv_cache_flags;
205
206 /* Unique Volume Id of the NvCache Volume associated with this volume */
207 __u32 nvc_vol_orig_family_num;
208 __u16 nvc_vol_raid_dev_num;
209
210#define RWH_OFF 0
211#define RWH_DISTRIBUTED 1
212#define RWH_JOURNALING_DRIVE 2
c2462068
PB		 213#define RWH_MULTIPLE_DISTRIBUTED 3
214#define RWH_MULTIPLE_PPLS_JOURNALING_DRIVE 4
215#define RWH_MULTIPLE_OFF 5
2432ce9b
AP		 216 __u8 rwh_policy; /* Raid Write Hole Policy */
217 __u8 jd_serial[MAX_RAID_SERIAL_LEN]; /* Journal Drive serial number */
218 __u8 filler1;
219
220#define IMSM_DEV_FILLERS 3
cdddbdbc
DW		 221 __u32 filler[IMSM_DEV_FILLERS];
222 struct imsm_vol vol;
223} __attribute__ ((packed));
224
225struct imsm_super {
226 __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */
227 __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */
228 __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */
229 __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */
230 __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */
604b746f
JD		 231 __u32 error_log_size; /* 0x30 - 0x33 in bytes */
232 __u32 attributes; /* 0x34 - 0x37 */
cdddbdbc
DW		 233 __u8 num_disks; /* 0x38 Number of configured disks */
234 __u8 num_raid_devs; /* 0x39 Number of configured volumes */
604b746f
JD		 235 __u8 error_log_pos; /* 0x3A */
236 __u8 fill[1]; /* 0x3B */
237 __u32 cache_size; /* 0x3c - 0x40 in mb */
238 __u32 orig_family_num; /* 0x40 - 0x43 original family num */
239 __u32 pwr_cycle_count; /* 0x44 - 0x47 simulated power cycle count for array */
240 __u32 bbm_log_size; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */
2a24dc1b
PB		 241 __u16 num_raid_devs_created; /* 0x4C - 0x4D Used for generating unique
242 * volume IDs for raid_dev created in this array
243 * (starts at 1)
244 */
245 __u16 filler1; /* 0x4E - 0x4F */
246#define IMSM_FILLERS 34
247 __u32 filler[IMSM_FILLERS]; /* 0x50 - 0xD7 RAID_MPB_FILLERS */
cdddbdbc
DW		 248 struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */
249 /* here comes imsm_dev[num_raid_devs] */
604b746f 250 /* here comes BBM logs */
cdddbdbc
DW		 251} __attribute__ ((packed));
252
604b746f 253#define BBM_LOG_MAX_ENTRIES 254
8d67477f
TM		 254#define BBM_LOG_MAX_LBA_ENTRY_VAL 256 /* Represents 256 LBAs */
255#define BBM_LOG_SIGNATURE 0xabadb10c
256
257struct bbm_log_block_addr {
258 __u16 w1;
259 __u32 dw1;
260} __attribute__ ((__packed__));
604b746f
JD		 261
262struct bbm_log_entry {
8d67477f
TM		 263 __u8 marked_count; /* Number of blocks marked - 1 */
264 __u8 disk_ordinal; /* Disk entry within the imsm_super */
265 struct bbm_log_block_addr defective_block_start;
604b746f
JD		 266} __attribute__ ((__packed__));
267
268struct bbm_log {
269 __u32 signature; /* 0xABADB10C */
270 __u32 entry_count;
8d67477f 271 struct bbm_log_entry marked_block_entries[BBM_LOG_MAX_ENTRIES];
604b746f
JD		 272} __attribute__ ((__packed__));
273
cdddbdbc 274static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" };
cdddbdbc 275
b53bfba6
TM		 276#define BLOCKS_PER_KB (1024/512)
277
8e59f3d8
AK		 278#define RAID_DISK_RESERVED_BLOCKS_IMSM_HI 2209
279
280#define GEN_MIGR_AREA_SIZE 2048 /* General Migration Copy Area size in blocks */
281
de44e46f
PB		 282#define MIGR_REC_BUF_SECTORS 1 /* size of migr_record i/o buffer in sectors */
283#define MIGR_REC_SECTOR_POSITION 1 /* migr_record position offset on disk,
284 * MIGR_REC_BUF_SECTORS <= MIGR_REC_SECTOR_POS
17a4eaf9
AK		 285 */
286
8e59f3d8
AK		 287#define UNIT_SRC_NORMAL 0 /* Source data for curr_migr_unit must
288 * be recovered using srcMap */
289#define UNIT_SRC_IN_CP_AREA 1 /* Source data for curr_migr_unit has
290 * already been migrated and must
291 * be recovered from checkpoint area */
2432ce9b 292
c2462068 293#define PPL_ENTRY_SPACE (128 * 1024) /* Size of single PPL, without the header */
2432ce9b 294
8e59f3d8
AK		 295struct migr_record {
296 __u32 rec_status; /* Status used to determine how to restart
297 * migration in case it aborts
298 * in some fashion */
299 __u32 curr_migr_unit; /* 0..numMigrUnits-1 */
300 __u32 family_num; /* Family number of MPB
301 * containing the RaidDev
302 * that is migrating */
303 __u32 ascending_migr; /* True if migrating in increasing
304 * order of lbas */
305 __u32 blocks_per_unit; /* Num disk blocks per unit of operation */
306 __u32 dest_depth_per_unit; /* Num member blocks each destMap
307 * member disk
308 * advances per unit-of-operation */
309 __u32 ckpt_area_pba; /* Pba of first block of ckpt copy area */
310 __u32 dest_1st_member_lba; /* First member lba on first
311 * stripe of destination */
312 __u32 num_migr_units; /* Total num migration units-of-op */
313 __u32 post_migr_vol_cap; /* Size of volume after
314 * migration completes */
315 __u32 post_migr_vol_cap_hi; /* Expansion space for LBA64 */
316 __u32 ckpt_read_disk_num; /* Which member disk in destSubMap[0] the
317 * migration ckpt record was read from
318 * (for recovered migrations) */
319} __attribute__ ((__packed__));
320
ec50f7b6
LM		 321struct md_list {
322 /* usage marker:
323 * 1: load metadata
324 * 2: metadata does not match
325 * 4: already checked
326 */
327 int used;
328 char *devname;
329 int found;
330 int container;
331 dev_t st_rdev;
332 struct md_list *next;
333};
334
e7b84f9d 335#define pr_vrb(fmt, arg...) (void) (verbose && pr_err(fmt, ##arg))
ec50f7b6 336
1484e727
DW		 337static __u8 migr_type(struct imsm_dev *dev)
338{
339 if (dev->vol.migr_type == MIGR_VERIFY &&
340 dev->status & DEV_VERIFY_AND_FIX)
341 return MIGR_REPAIR;
342 else
343 return dev->vol.migr_type;
344}
345
346static void set_migr_type(struct imsm_dev *dev, __u8 migr_type)
347{
348 /* for compatibility with older oroms convert MIGR_REPAIR, into
349 * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status
350 */
351 if (migr_type == MIGR_REPAIR) {
352 dev->vol.migr_type = MIGR_VERIFY;
353 dev->status |= DEV_VERIFY_AND_FIX;
354 } else {
355 dev->vol.migr_type = migr_type;
356 dev->status &= ~DEV_VERIFY_AND_FIX;
357 }
358}
359
f36a9ecd 360static unsigned int sector_count(__u32 bytes, unsigned int sector_size)
cdddbdbc 361{
f36a9ecd 362 return ROUND_UP(bytes, sector_size) / sector_size;
87eb16df 363}
cdddbdbc 364
f36a9ecd
PB		 365static unsigned int mpb_sectors(struct imsm_super *mpb,
366 unsigned int sector_size)
87eb16df 367{
f36a9ecd 368 return sector_count(__le32_to_cpu(mpb->mpb_size), sector_size);
cdddbdbc
DW		 369}
370
ba2de7ba
DW		 371struct intel_dev {
372 struct imsm_dev *dev;
373 struct intel_dev *next;
f21e18ca 374 unsigned index;
ba2de7ba
DW		 375};
376
88654014
LM		 377struct intel_hba {
378 enum sys_dev_type type;
379 char *path;
380 char *pci_id;
381 struct intel_hba *next;
382};
383
1a64be56
LM		 384enum action {
385 DISK_REMOVE = 1,
386 DISK_ADD
387};
cdddbdbc
DW		 388/* internal representation of IMSM metadata */
389struct intel_super {
390 union {
949c47a0
DW		 391 void *buf; /* O_DIRECT buffer for reading/writing metadata */
392 struct imsm_super *anchor; /* immovable parameters */
cdddbdbc 393 };
8e59f3d8
AK		 394 union {
395 void *migr_rec_buf; /* buffer for I/O operations */
396 struct migr_record *migr_rec; /* migration record */
397 };
51d83f5d
AK		 398 int clean_migration_record_by_mdmon; /* when reshape is switched to next
399 array, it indicates that mdmon is allowed to clean migration
400 record */
949c47a0 401 size_t len; /* size of the 'buf' allocation */
bbab0940 402 size_t extra_space; /* extra space in 'buf' that is not used yet */
4d7b1503
DW		 403 void *next_buf; /* for realloc'ing buf from the manager */
404 size_t next_len;
c2c087e6 405 int updates_pending; /* count of pending updates for mdmon */
bf5a934a 406 int current_vol; /* index of raid device undergoing creation */
5551b113 407 unsigned long long create_offset; /* common start for 'current_vol' */
148acb7b 408 __u32 random; /* random data for seeding new family numbers */
ba2de7ba 409 struct intel_dev *devlist;
fa7bb6f8 410 unsigned int sector_size; /* sector size of used member drives */
cdddbdbc
DW		 411 struct dl {
412 struct dl *next;
413 int index;
414 __u8 serial[MAX_RAID_SERIAL_LEN];
415 int major, minor;
416 char *devname;
b9f594fe 417 struct imsm_disk disk;
cdddbdbc 418 int fd;
0dcecb2e
DW		 419 int extent_cnt;
420 struct extent *e; /* for determining freespace @ create */
efb30e7f 421 int raiddisk; /* slot to fill in autolayout */
1a64be56 422 enum action action;
ca0748fa 423 } *disks, *current_disk;
1a64be56
LM		 424 struct dl *disk_mgmt_list; /* list of disks to add/remove while mdmon
425 active */
47ee5a45 426 struct dl *missing; /* disks removed while we weren't looking */
43dad3d6 427 struct bbm_log *bbm_log;
88654014 428 struct intel_hba *hba; /* device path of the raid controller for this metadata */
88c32bb1 429 const struct imsm_orom *orom; /* platform firmware support */
a2b97981 430 struct intel_super *next; /* (temp) list for disambiguating family_num */
928f1424 431 struct md_bb bb; /* memory for get_bad_blocks call */
a2b97981
DW		 432};
433
434struct intel_disk {
435 struct imsm_disk disk;
436 #define IMSM_UNKNOWN_OWNER (-1)
437 int owner;
438 struct intel_disk *next;
cdddbdbc
DW		 439};
440
c2c087e6
DW		 441struct extent {
442 unsigned long long start, size;
443};
444
694575e7
KW		 445/* definitions of reshape process types */
446enum imsm_reshape_type {
447 CH_TAKEOVER,
b5347799 448 CH_MIGRATION,
7abc9871 449 CH_ARRAY_SIZE,
694575e7
KW		 450};
451
88758e9d
DW		 452/* definition of messages passed to imsm_process_update */
453enum imsm_update_type {
454 update_activate_spare,
8273f55e 455 update_create_array,
33414a01 456 update_kill_array,
aa534678 457 update_rename_array,
1a64be56 458 update_add_remove_disk,
78b10e66 459 update_reshape_container_disks,
48c5303a 460 update_reshape_migration,
2d40f3a1
AK		 461 update_takeover,
462 update_general_migration_checkpoint,
f3871fdc 463 update_size_change,
bbab0940 464 update_prealloc_badblocks_mem,
e6e9dd3f 465 update_rwh_policy,
88758e9d
DW		 466};
467
468struct imsm_update_activate_spare {
469 enum imsm_update_type type;
d23fe947 470 struct dl *dl;
88758e9d
DW		 471 int slot;
472 int array;
473 struct imsm_update_activate_spare *next;
474};
475
78b10e66 476struct geo_params {
4dd2df09 477 char devnm[32];
78b10e66 478 char *dev_name;
d04f65f4 479 unsigned long long size;
78b10e66
N		 480 int level;
481 int layout;
482 int chunksize;
483 int raid_disks;
484};
485
bb025c2f
KW		 486enum takeover_direction {
487 R10_TO_R0,
488 R0_TO_R10
489};
490struct imsm_update_takeover {
491 enum imsm_update_type type;
492 int subarray;
493 enum takeover_direction direction;
494};
78b10e66
N		 495
496struct imsm_update_reshape {
497 enum imsm_update_type type;
498 int old_raid_disks;
499 int new_raid_disks;
48c5303a
PC		 500
501 int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */
502};
503
504struct imsm_update_reshape_migration {
505 enum imsm_update_type type;
506 int old_raid_disks;
507 int new_raid_disks;
508 /* fields for array migration changes
509 */
510 int subdev;
511 int new_level;
512 int new_layout;
4bba0439 513 int new_chunksize;
48c5303a 514
d195167d 515 int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */
78b10e66
N		 516};
517
f3871fdc
AK		 518struct imsm_update_size_change {
519 enum imsm_update_type type;
520 int subdev;
521 long long new_size;
522};
523
2d40f3a1
AK		 524struct imsm_update_general_migration_checkpoint {
525 enum imsm_update_type type;
526 __u32 curr_migr_unit;
527};
528
54c2c1ea
DW		 529struct disk_info {
530 __u8 serial[MAX_RAID_SERIAL_LEN];
531};
532
8273f55e
DW		 533struct imsm_update_create_array {
534 enum imsm_update_type type;
8273f55e 535 int dev_idx;
6a3e913e 536 struct imsm_dev dev;
8273f55e
DW		 537};
538
33414a01
DW		 539struct imsm_update_kill_array {
540 enum imsm_update_type type;
541 int dev_idx;
542};
543
aa534678
DW		 544struct imsm_update_rename_array {
545 enum imsm_update_type type;
546 __u8 name[MAX_RAID_SERIAL_LEN];
547 int dev_idx;
548};
549
1a64be56 550struct imsm_update_add_remove_disk {
43dad3d6
DW		 551 enum imsm_update_type type;
552};
553
bbab0940
TM		 554struct imsm_update_prealloc_bb_mem {
555 enum imsm_update_type type;
556};
557
e6e9dd3f
AP		 558struct imsm_update_rwh_policy {
559 enum imsm_update_type type;
560 int new_policy;
561 int dev_idx;
562};
563
88654014
LM		 564static const char *_sys_dev_type[] = {
565 [SYS_DEV_UNKNOWN] = "Unknown",
566 [SYS_DEV_SAS] = "SAS",
614902f6 567 [SYS_DEV_SATA] = "SATA",
60f0f54d
PB		 568 [SYS_DEV_NVME] = "NVMe",
569 [SYS_DEV_VMD] = "VMD"
88654014
LM		 570};
571
572const char *get_sys_dev_type(enum sys_dev_type type)
573{
574 if (type >= SYS_DEV_MAX)
575 type = SYS_DEV_UNKNOWN;
576
577 return _sys_dev_type[type];
578}
579
580static struct intel_hba * alloc_intel_hba(struct sys_dev *device)
581{
503975b9
N		 582 struct intel_hba *result = xmalloc(sizeof(*result));
583
584 result->type = device->type;
585 result->path = xstrdup(device->path);
586 result->next = NULL;
587 if (result->path && (result->pci_id = strrchr(result->path, '/')) != NULL)
588 result->pci_id++;
589
88654014
LM		 590 return result;
591}
592
593static struct intel_hba * find_intel_hba(struct intel_hba *hba, struct sys_dev *device)
594{
594dc1b8
JS		 595 struct intel_hba *result;
596
88654014
LM		 597 for (result = hba; result; result = result->next) {
598 if (result->type == device->type && strcmp(result->path, device->path) == 0)
599 break;
600 }
601 return result;
602}
603
b4cf4cba 604static int attach_hba_to_super(struct intel_super *super, struct sys_dev *device)
88654014
LM		 605{
606 struct intel_hba *hba;
607
608 /* check if disk attached to Intel HBA */
609 hba = find_intel_hba(super->hba, device);
610 if (hba != NULL)
611 return 1;
612 /* Check if HBA is already attached to super */
613 if (super->hba == NULL) {
614 super->hba = alloc_intel_hba(device);
615 return 1;
6b781d33
AP		 616 }
617
618 hba = super->hba;
619 /* Intel metadata allows for all disks attached to the same type HBA.
614902f6 620 * Do not support HBA types mixing
6b781d33
AP		 621 */
622 if (device->type != hba->type)
88654014 623 return 2;
6b781d33
AP		 624
625 /* Multiple same type HBAs can be used if they share the same OROM */
626 const struct imsm_orom *device_orom = get_orom_by_device_id(device->dev_id);
627
628 if (device_orom != super->orom)
629 return 2;
630
631 while (hba->next)
632 hba = hba->next;
633
634 hba->next = alloc_intel_hba(device);
635 return 1;
88654014
LM		 636}
637
638static struct sys_dev* find_disk_attached_hba(int fd, const char *devname)
639{
9bc4ae77 640 struct sys_dev *list, *elem;
88654014
LM		 641 char *disk_path;
642
643 if ((list = find_intel_devices()) == NULL)
644 return 0;
645
646 if (fd < 0)
647 disk_path = (char *) devname;
648 else
649 disk_path = diskfd_to_devpath(fd);
650
9bc4ae77 651 if (!disk_path)
88654014 652 return 0;
88654014 653
9bc4ae77
N		 654 for (elem = list; elem; elem = elem->next)
655 if (path_attached_to_hba(disk_path, elem->path))
88654014 656 return elem;
9bc4ae77 657
88654014
LM		 658 if (disk_path != devname)
659 free(disk_path);
88654014
LM		 660
661 return NULL;
662}
663
d424212e
N		 664static int find_intel_hba_capability(int fd, struct intel_super *super,
665 char *devname);
f2f5c343 666
cdddbdbc
DW		 667static struct supertype *match_metadata_desc_imsm(char *arg)
668{
669 struct supertype *st;
670
671 if (strcmp(arg, "imsm") != 0 &&
672 strcmp(arg, "default") != 0
673 )
674 return NULL;
675
503975b9 676 st = xcalloc(1, sizeof(*st));
cdddbdbc
DW		 677 st->ss = &super_imsm;
678 st->max_devs = IMSM_MAX_DEVICES;
679 st->minor_version = 0;
680 st->sb = NULL;
681 return st;
682}
683
cdddbdbc
DW		 684static __u8 *get_imsm_version(struct imsm_super *mpb)
685{
686 return &mpb->sig[MPB_SIG_LEN];
687}
688
949c47a0
DW		 689/* retrieve a disk directly from the anchor when the anchor is known to be
690 * up-to-date, currently only at load time
691 */
692static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index)
cdddbdbc 693{
949c47a0 694 if (index >= mpb->num_disks)
cdddbdbc
DW		 695 return NULL;
696 return &mpb->disk[index];
697}
698
95d07a2c
LM		 699/* retrieve the disk description based on a index of the disk
700 * in the sub-array
701 */
702static struct dl *get_imsm_dl_disk(struct intel_super *super, __u8 index)
949c47a0 703{
b9f594fe
DW		 704 struct dl *d;
705
706 for (d = super->disks; d; d = d->next)
707 if (d->index == index)
95d07a2c
LM		 708 return d;
709
710 return NULL;
711}
712/* retrieve a disk from the parsed metadata */
713static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index)
714{
715 struct dl *dl;
716
717 dl = get_imsm_dl_disk(super, index);
718 if (dl)
719 return &dl->disk;
720
b9f594fe 721 return NULL;
949c47a0
DW		 722}
723
724/* generate a checksum directly from the anchor when the anchor is known to be
725 * up-to-date, currently only at load or write_super after coalescing
726 */
727static __u32 __gen_imsm_checksum(struct imsm_super *mpb)
cdddbdbc
DW		 728{
729 __u32 end = mpb->mpb_size / sizeof(end);
730 __u32 *p = (__u32 *) mpb;
731 __u32 sum = 0;
732
5d500228
N		 733 while (end--) {
734 sum += __le32_to_cpu(*p);
97f734fd
N		 735 p++;
736 }
cdddbdbc 737
5d500228 738 return sum - __le32_to_cpu(mpb->check_sum);
cdddbdbc
DW		 739}
740
a965f303
DW		 741static size_t sizeof_imsm_map(struct imsm_map *map)
742{
743 return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1);
744}
745
746struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map)
cdddbdbc 747{
5e7b0330
AK		 748 /* A device can have 2 maps if it is in the middle of a migration.
749 * If second_map is:
238c0a71
AK		 750 * MAP_0 - we return the first map
751 * MAP_1 - we return the second map if it exists, else NULL
752 * MAP_X - we return the second map if it exists, else the first
5e7b0330 753 */
a965f303 754 struct imsm_map *map = &dev->vol.map[0];
9535fc47 755 struct imsm_map *map2 = NULL;
a965f303 756
9535fc47
AK		 757 if (dev->vol.migr_state)
758 map2 = (void *)map + sizeof_imsm_map(map);
a965f303 759
9535fc47 760 switch (second_map) {
3b451610 761 case MAP_0:
9535fc47 762 break;
3b451610 763 case MAP_1:
9535fc47
AK		 764 map = map2;
765 break;
238c0a71 766 case MAP_X:
9535fc47
AK		 767 if (map2)
768 map = map2;
769 break;
9535fc47
AK		 770 default:
771 map = NULL;
772 }
773 return map;
5e7b0330 774
a965f303 775}
cdddbdbc 776
3393c6af
DW		 777/* return the size of the device.
778 * migr_state increases the returned size if map[0] were to be duplicated
779 */
780static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state)
a965f303
DW		 781{
782 size_t size = sizeof(*dev) - sizeof(struct imsm_map) +
238c0a71 783 sizeof_imsm_map(get_imsm_map(dev, MAP_0));
cdddbdbc
DW		 784
785 /* migrating means an additional map */
a965f303 786 if (dev->vol.migr_state)
238c0a71 787 size += sizeof_imsm_map(get_imsm_map(dev, MAP_1));
3393c6af 788 else if (migr_state)
238c0a71 789 size += sizeof_imsm_map(get_imsm_map(dev, MAP_0));
cdddbdbc
DW		 790
791 return size;
792}
793
54c2c1ea
DW		 794/* retrieve disk serial number list from a metadata update */
795static struct disk_info *get_disk_info(struct imsm_update_create_array *update)
796{
797 void *u = update;
798 struct disk_info *inf;
799
800 inf = u + sizeof(*update) - sizeof(struct imsm_dev) +
801 sizeof_imsm_dev(&update->dev, 0);
802
803 return inf;
804}
54c2c1ea 805
949c47a0 806static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index)
cdddbdbc
DW		 807{
808 int offset;
809 int i;
810 void *_mpb = mpb;
811
949c47a0 812 if (index >= mpb->num_raid_devs)
cdddbdbc
DW		 813 return NULL;
814
815 /* devices start after all disks */
816 offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb;
817
818 for (i = 0; i <= index; i++)
819 if (i == index)
820 return _mpb + offset;
821 else
3393c6af 822 offset += sizeof_imsm_dev(_mpb + offset, 0);
cdddbdbc
DW		 823
824 return NULL;
825}
826
949c47a0
DW		 827static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index)
828{
ba2de7ba
DW		 829 struct intel_dev *dv;
830
949c47a0
DW		 831 if (index >= super->anchor->num_raid_devs)
832 return NULL;
ba2de7ba
DW		 833 for (dv = super->devlist; dv; dv = dv->next)
834 if (dv->index == index)
835 return dv->dev;
836 return NULL;
949c47a0
DW		 837}
838
8d67477f
TM		 839static inline unsigned long long __le48_to_cpu(const struct bbm_log_block_addr
840 *addr)
841{
842 return ((((__u64)__le32_to_cpu(addr->dw1)) << 16) |
843 __le16_to_cpu(addr->w1));
844}
845
846static inline struct bbm_log_block_addr __cpu_to_le48(unsigned long long sec)
847{
848 struct bbm_log_block_addr addr;
849
850 addr.w1 = __cpu_to_le16((__u16)(sec & 0xffff));
851 addr.dw1 = __cpu_to_le32((__u32)(sec >> 16) & 0xffffffff);
852 return addr;
853}
854
8d67477f
TM		 855/* get size of the bbm log */
856static __u32 get_imsm_bbm_log_size(struct bbm_log *log)
857{
858 if (!log || log->entry_count == 0)
859 return 0;
860
861 return sizeof(log->signature) +
862 sizeof(log->entry_count) +
863 log->entry_count * sizeof(struct bbm_log_entry);
864}
6f50473f
TM		 865
866/* check if bad block is not partially stored in bbm log */
867static int is_stored_in_bbm(struct bbm_log *log, const __u8 idx, const unsigned
868 long long sector, const int length, __u32 *pos)
869{
870 __u32 i;
871
872 for (i = *pos; i < log->entry_count; i++) {
873 struct bbm_log_entry *entry = &log->marked_block_entries[i];
874 unsigned long long bb_start;
875 unsigned long long bb_end;
876
877 bb_start = __le48_to_cpu(&entry->defective_block_start);
878 bb_end = bb_start + (entry->marked_count + 1);
879
880 if ((entry->disk_ordinal == idx) && (bb_start >= sector) &&
881 (bb_end <= sector + length)) {
882 *pos = i;
883 return 1;
884 }
885 }
886 return 0;
887}
888
889/* record new bad block in bbm log */
890static int record_new_badblock(struct bbm_log *log, const __u8 idx, unsigned
891 long long sector, int length)
892{
893 int new_bb = 0;
894 __u32 pos = 0;
895 struct bbm_log_entry *entry = NULL;
896
897 while (is_stored_in_bbm(log, idx, sector, length, &pos)) {
898 struct bbm_log_entry *e = &log->marked_block_entries[pos];
899
900 if ((e->marked_count + 1 == BBM_LOG_MAX_LBA_ENTRY_VAL) &&
901 (__le48_to_cpu(&e->defective_block_start) == sector)) {
902 sector += BBM_LOG_MAX_LBA_ENTRY_VAL;
903 length -= BBM_LOG_MAX_LBA_ENTRY_VAL;
904 pos = pos + 1;
905 continue;
906 }
907 entry = e;
908 break;
909 }
910
911 if (entry) {
912 int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length :
913 BBM_LOG_MAX_LBA_ENTRY_VAL;
914 entry->defective_block_start = __cpu_to_le48(sector);
915 entry->marked_count = cnt - 1;
916 if (cnt == length)
917 return 1;
918 sector += cnt;
919 length -= cnt;
920 }
921
922 new_bb = ROUND_UP(length, BBM_LOG_MAX_LBA_ENTRY_VAL) /
923 BBM_LOG_MAX_LBA_ENTRY_VAL;
924 if (log->entry_count + new_bb > BBM_LOG_MAX_ENTRIES)
925 return 0;
926
927 while (length > 0) {
928 int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length :
929 BBM_LOG_MAX_LBA_ENTRY_VAL;
930 struct bbm_log_entry *entry =
931 &log->marked_block_entries[log->entry_count];
932
933 entry->defective_block_start = __cpu_to_le48(sector);
934 entry->marked_count = cnt - 1;
935 entry->disk_ordinal = idx;
936
937 sector += cnt;
938 length -= cnt;
939
940 log->entry_count++;
941 }
942
943 return new_bb;
944}
c07a5a4f 945
4c9e8c1e
TM		 946/* clear all bad blocks for given disk */
947static void clear_disk_badblocks(struct bbm_log *log, const __u8 idx)
948{
949 __u32 i = 0;
950
951 while (i < log->entry_count) {
952 struct bbm_log_entry *entries = log->marked_block_entries;
953
954 if (entries[i].disk_ordinal == idx) {
955 if (i < log->entry_count - 1)
956 entries[i] = entries[log->entry_count - 1];
957 log->entry_count--;
958 } else {
959 i++;
960 }
961 }
962}
963
c07a5a4f
TM		 964/* clear given bad block */
965static int clear_badblock(struct bbm_log *log, const __u8 idx, const unsigned
966 long long sector, const int length) {
967 __u32 i = 0;
968
969 while (i < log->entry_count) {
970 struct bbm_log_entry *entries = log->marked_block_entries;
971
972 if ((entries[i].disk_ordinal == idx) &&
973 (__le48_to_cpu(&entries[i].defective_block_start) ==
974 sector) && (entries[i].marked_count + 1 == length)) {
975 if (i < log->entry_count - 1)
976 entries[i] = entries[log->entry_count - 1];
977 log->entry_count--;
978 break;
979 }
980 i++;
981 }
982
983 return 1;
984}
8d67477f
TM		 985
986/* allocate and load BBM log from metadata */
987static int load_bbm_log(struct intel_super *super)
988{
989 struct imsm_super *mpb = super->anchor;
990 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
991
992 super->bbm_log = xcalloc(1, sizeof(struct bbm_log));
993 if (!super->bbm_log)
994 return 1;
995
996 if (bbm_log_size) {
997 struct bbm_log *log = (void *)mpb +
998 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
999
1000 __u32 entry_count;
1001
1002 if (bbm_log_size < sizeof(log->signature) +
1003 sizeof(log->entry_count))
1004 return 2;
1005
1006 entry_count = __le32_to_cpu(log->entry_count);
1007 if ((__le32_to_cpu(log->signature) != BBM_LOG_SIGNATURE) ||
1008 (entry_count > BBM_LOG_MAX_ENTRIES))
1009 return 3;
1010
1011 if (bbm_log_size !=
1012 sizeof(log->signature) + sizeof(log->entry_count) +
1013 entry_count * sizeof(struct bbm_log_entry))
1014 return 4;
1015
1016 memcpy(super->bbm_log, log, bbm_log_size);
1017 } else {
1018 super->bbm_log->signature = __cpu_to_le32(BBM_LOG_SIGNATURE);
1019 super->bbm_log->entry_count = 0;
1020 }
1021
1022 return 0;
1023}
1024
b12796be
TM		 1025/* checks if bad block is within volume boundaries */
1026static int is_bad_block_in_volume(const struct bbm_log_entry *entry,
1027 const unsigned long long start_sector,
1028 const unsigned long long size)
1029{
1030 unsigned long long bb_start;
1031 unsigned long long bb_end;
1032
1033 bb_start = __le48_to_cpu(&entry->defective_block_start);
1034 bb_end = bb_start + (entry->marked_count + 1);
1035
1036 if (((bb_start >= start_sector) && (bb_start < start_sector + size)) ||
1037 ((bb_end >= start_sector) && (bb_end <= start_sector + size)))
1038 return 1;
1039
1040 return 0;
1041}
1042
1043/* get list of bad blocks on a drive for a volume */
1044static void get_volume_badblocks(const struct bbm_log *log, const __u8 idx,
1045 const unsigned long long start_sector,
1046 const unsigned long long size,
1047 struct md_bb *bbs)
1048{
1049 __u32 count = 0;
1050 __u32 i;
1051
1052 for (i = 0; i < log->entry_count; i++) {
1053 const struct bbm_log_entry *ent =
1054 &log->marked_block_entries[i];
1055 struct md_bb_entry *bb;
1056
1057 if ((ent->disk_ordinal == idx) &&
1058 is_bad_block_in_volume(ent, start_sector, size)) {
1059
1060 if (!bbs->entries) {
1061 bbs->entries = xmalloc(BBM_LOG_MAX_ENTRIES *
1062 sizeof(*bb));
1063 if (!bbs->entries)
1064 break;
1065 }
1066
1067 bb = &bbs->entries[count++];
1068 bb->sector = __le48_to_cpu(&ent->defective_block_start);
1069 bb->length = ent->marked_count + 1;
1070 }
1071 }
1072 bbs->count = count;
1073}
1074
98130f40
AK		 1075/*
1076 * for second_map:
238c0a71
AK		 1077 * == MAP_0 get first map
1078 * == MAP_1 get second map
1079 * == MAP_X than get map according to the current migr_state
98130f40
AK		 1080 */
1081static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev,
1082 int slot,
1083 int second_map)
7eef0453
DW		 1084{
1085 struct imsm_map *map;
1086
5e7b0330 1087 map = get_imsm_map(dev, second_map);
7eef0453 1088
ff077194
DW		 1089 /* top byte identifies disk under rebuild */
1090 return __le32_to_cpu(map->disk_ord_tbl[slot]);
1091}
1092
1093#define ord_to_idx(ord) (((ord) << 8) >> 8)
98130f40 1094static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot, int second_map)
ff077194 1095{
98130f40 1096 __u32 ord = get_imsm_ord_tbl_ent(dev, slot, second_map);
ff077194
DW		 1097
1098 return ord_to_idx(ord);
7eef0453
DW		 1099}
1100
be73972f
DW		 1101static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord)
1102{
1103 map->disk_ord_tbl[slot] = __cpu_to_le32(ord);
1104}
1105
f21e18ca 1106static int get_imsm_disk_slot(struct imsm_map *map, unsigned idx)
620b1713
DW		 1107{
1108 int slot;
1109 __u32 ord;
1110
1111 for (slot = 0; slot < map->num_members; slot++) {
1112 ord = __le32_to_cpu(map->disk_ord_tbl[slot]);
1113 if (ord_to_idx(ord) == idx)
1114 return slot;
1115 }
1116
1117 return -1;
1118}
1119
cdddbdbc
DW		 1120static int get_imsm_raid_level(struct imsm_map *map)
1121{
1122 if (map->raid_level == 1) {
1123 if (map->num_members == 2)
1124 return 1;
1125 else
1126 return 10;
1127 }
1128
1129 return map->raid_level;
1130}
1131
c2c087e6
DW		 1132static int cmp_extent(const void *av, const void *bv)
1133{
1134 const struct extent *a = av;
1135 const struct extent *b = bv;
1136 if (a->start < b->start)
1137 return -1;
1138 if (a->start > b->start)
1139 return 1;
1140 return 0;
1141}
1142
0dcecb2e 1143static int count_memberships(struct dl *dl, struct intel_super *super)
c2c087e6 1144{
c2c087e6 1145 int memberships = 0;
620b1713 1146 int i;
c2c087e6 1147
949c47a0
DW		 1148 for (i = 0; i < super->anchor->num_raid_devs; i++) {
1149 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 1150 struct imsm_map *map = get_imsm_map(dev, MAP_0);
c2c087e6 1151
620b1713
DW		 1152 if (get_imsm_disk_slot(map, dl->index) >= 0)
1153 memberships++;
c2c087e6 1154 }
0dcecb2e
DW		 1155
1156 return memberships;
1157}
1158
b81221b7
CA		 1159static __u32 imsm_min_reserved_sectors(struct intel_super *super);
1160
5551b113
CA		 1161static int split_ull(unsigned long long n, __u32 *lo, __u32 *hi)
1162{
1163 if (lo == 0 || hi == 0)
1164 return 1;
1165 *lo = __le32_to_cpu((unsigned)n);
1166 *hi = __le32_to_cpu((unsigned)(n >> 32));
1167 return 0;
1168}
1169
1170static unsigned long long join_u32(__u32 lo, __u32 hi)
1171{
1172 return (unsigned long long)__le32_to_cpu(lo) |
1173 (((unsigned long long)__le32_to_cpu(hi)) << 32);
1174}
1175
1176static unsigned long long total_blocks(struct imsm_disk *disk)
1177{
1178 if (disk == NULL)
1179 return 0;
1180 return join_u32(disk->total_blocks_lo, disk->total_blocks_hi);
1181}
1182
1183static unsigned long long pba_of_lba0(struct imsm_map *map)
1184{
1185 if (map == NULL)
1186 return 0;
1187 return join_u32(map->pba_of_lba0_lo, map->pba_of_lba0_hi);
1188}
1189
1190static unsigned long long blocks_per_member(struct imsm_map *map)
1191{
1192 if (map == NULL)
1193 return 0;
1194 return join_u32(map->blocks_per_member_lo, map->blocks_per_member_hi);
1195}
1196
1197static unsigned long long num_data_stripes(struct imsm_map *map)
1198{
1199 if (map == NULL)
1200 return 0;
1201 return join_u32(map->num_data_stripes_lo, map->num_data_stripes_hi);
1202}
1203
fcc2c9da
MD		 1204static unsigned long long imsm_dev_size(struct imsm_dev *dev)
1205{
1206 if (dev == NULL)
1207 return 0;
1208 return join_u32(dev->size_low, dev->size_high);
1209}
1210
5551b113
CA		 1211static void set_total_blocks(struct imsm_disk *disk, unsigned long long n)
1212{
1213 split_ull(n, &disk->total_blocks_lo, &disk->total_blocks_hi);
1214}
1215
1216static void set_pba_of_lba0(struct imsm_map *map, unsigned long long n)
1217{
1218 split_ull(n, &map->pba_of_lba0_lo, &map->pba_of_lba0_hi);
1219}
1220
1221static void set_blocks_per_member(struct imsm_map *map, unsigned long long n)
1222{
1223 split_ull(n, &map->blocks_per_member_lo, &map->blocks_per_member_hi);
1224}
1225
1226static void set_num_data_stripes(struct imsm_map *map, unsigned long long n)
1227{
1228 split_ull(n, &map->num_data_stripes_lo, &map->num_data_stripes_hi);
1229}
1230
fcc2c9da
MD		 1231static void set_imsm_dev_size(struct imsm_dev *dev, unsigned long long n)
1232{
1233 split_ull(n, &dev->size_low, &dev->size_high);
1234}
1235
44490938
MD		 1236static unsigned long long per_dev_array_size(struct imsm_map *map)
1237{
1238 unsigned long long array_size = 0;
1239
1240 if (map == NULL)
1241 return array_size;
1242
1243 array_size = num_data_stripes(map) * map->blocks_per_strip;
1244 if (get_imsm_raid_level(map) == 1 || get_imsm_raid_level(map) == 10)
1245 array_size *= 2;
1246
1247 return array_size;
1248}
1249
0dcecb2e
DW		 1250static struct extent *get_extents(struct intel_super *super, struct dl *dl)
1251{
1252 /* find a list of used extents on the given physical device */
1253 struct extent *rv, *e;
620b1713 1254 int i;
0dcecb2e 1255 int memberships = count_memberships(dl, super);
b276dd33
DW		 1256 __u32 reservation;
1257
1258 /* trim the reserved area for spares, so they can join any array
1259 * regardless of whether the OROM has assigned sectors from the
1260 * IMSM_RESERVED_SECTORS region
1261 */
1262 if (dl->index == -1)
b81221b7 1263 reservation = imsm_min_reserved_sectors(super);
b276dd33
DW		 1264 else
1265 reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
0dcecb2e 1266
503975b9 1267 rv = xcalloc(sizeof(struct extent), (memberships + 1));
c2c087e6
DW		 1268 e = rv;
1269
949c47a0
DW		 1270 for (i = 0; i < super->anchor->num_raid_devs; i++) {
1271 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 1272 struct imsm_map *map = get_imsm_map(dev, MAP_0);
c2c087e6 1273
620b1713 1274 if (get_imsm_disk_slot(map, dl->index) >= 0) {
5551b113 1275 e->start = pba_of_lba0(map);
44490938 1276 e->size = per_dev_array_size(map);
620b1713 1277 e++;
c2c087e6
DW		 1278 }
1279 }
1280 qsort(rv, memberships, sizeof(*rv), cmp_extent);
1281
1011e834 1282 /* determine the start of the metadata
14e8215b
DW		 1283 * when no raid devices are defined use the default
1284 * ...otherwise allow the metadata to truncate the value
1285 * as is the case with older versions of imsm
1286 */
1287 if (memberships) {
1288 struct extent *last = &rv[memberships - 1];
5551b113 1289 unsigned long long remainder;
14e8215b 1290
5551b113 1291 remainder = total_blocks(&dl->disk) - (last->start + last->size);
dda5855f
DW		 1292 /* round down to 1k block to satisfy precision of the kernel
1293 * 'size' interface
1294 */
1295 remainder &= ~1UL;
1296 /* make sure remainder is still sane */
f21e18ca 1297 if (remainder < (unsigned)ROUND_UP(super->len, 512) >> 9)
dda5855f 1298 remainder = ROUND_UP(super->len, 512) >> 9;
14e8215b
DW		 1299 if (reservation > remainder)
1300 reservation = remainder;
1301 }
5551b113 1302 e->start = total_blocks(&dl->disk) - reservation;
c2c087e6
DW		 1303 e->size = 0;
1304 return rv;
1305}
1306
14e8215b
DW		 1307/* try to determine how much space is reserved for metadata from
1308 * the last get_extents() entry, otherwise fallback to the
1309 * default
1310 */
1311static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl)
1312{
1313 struct extent *e;
1314 int i;
1315 __u32 rv;
1316
1317 /* for spares just return a minimal reservation which will grow
1318 * once the spare is picked up by an array
1319 */
1320 if (dl->index == -1)
1321 return MPB_SECTOR_CNT;
1322
1323 e = get_extents(super, dl);
1324 if (!e)
1325 return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
1326
1327 /* scroll to last entry */
1328 for (i = 0; e[i].size; i++)
1329 continue;
1330
5551b113 1331 rv = total_blocks(&dl->disk) - e[i].start;
14e8215b
DW		 1332
1333 free(e);
1334
1335 return rv;
1336}
1337
25ed7e59
DW		 1338static int is_spare(struct imsm_disk *disk)
1339{
1340 return (disk->status & SPARE_DISK) == SPARE_DISK;
1341}
1342
1343static int is_configured(struct imsm_disk *disk)
1344{
1345 return (disk->status & CONFIGURED_DISK) == CONFIGURED_DISK;
1346}
1347
1348static int is_failed(struct imsm_disk *disk)
1349{
1350 return (disk->status & FAILED_DISK) == FAILED_DISK;
1351}
1352
2432ce9b
AP		 1353static int is_journal(struct imsm_disk *disk)
1354{
1355 return (disk->status & JOURNAL_DISK) == JOURNAL_DISK;
1356}
1357
b53bfba6
TM		 1358/* round array size down to closest MB and ensure it splits evenly
1359 * between members
1360 */
1361static unsigned long long round_size_to_mb(unsigned long long size, unsigned int
1362 disk_count)
1363{
1364 size /= disk_count;
1365 size = (size >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT;
1366 size *= disk_count;
1367
1368 return size;
1369}
1370
8b9cd157
MK		 1371static int able_to_resync(int raid_level, int missing_disks)
1372{
1373 int max_missing_disks = 0;
1374
1375 switch (raid_level) {
1376 case 10:
1377 max_missing_disks = 1;
1378 break;
1379 default:
1380 max_missing_disks = 0;
1381 }
1382 return missing_disks <= max_missing_disks;
1383}
1384
b81221b7
CA		 1385/* try to determine how much space is reserved for metadata from
1386 * the last get_extents() entry on the smallest active disk,
1387 * otherwise fallback to the default
1388 */
1389static __u32 imsm_min_reserved_sectors(struct intel_super *super)
1390{
1391 struct extent *e;
1392 int i;
5551b113
CA		 1393 unsigned long long min_active;
1394 __u32 remainder;
b81221b7
CA		 1395 __u32 rv = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
1396 struct dl *dl, *dl_min = NULL;
1397
1398 if (!super)
1399 return rv;
1400
1401 min_active = 0;
1402 for (dl = super->disks; dl; dl = dl->next) {
1403 if (dl->index < 0)
1404 continue;
5551b113
CA		 1405 unsigned long long blocks = total_blocks(&dl->disk);
1406 if (blocks < min_active || min_active == 0) {
b81221b7 1407 dl_min = dl;
5551b113 1408 min_active = blocks;
b81221b7
CA		 1409 }
1410 }
1411 if (!dl_min)
1412 return rv;
1413
1414 /* find last lba used by subarrays on the smallest active disk */
1415 e = get_extents(super, dl_min);
1416 if (!e)
1417 return rv;
1418 for (i = 0; e[i].size; i++)
1419 continue;
1420
1421 remainder = min_active - e[i].start;
1422 free(e);
1423
1424 /* to give priority to recovery we should not require full
1425 IMSM_RESERVED_SECTORS from the spare */
1426 rv = MPB_SECTOR_CNT + NUM_BLOCKS_DIRTY_STRIPE_REGION;
1427
1428 /* if real reservation is smaller use that value */
1429 return (remainder < rv) ? remainder : rv;
1430}
1431
fbfdcb06
AO		 1432/*
1433 * Return minimum size of a spare and sector size
1434 * that can be used in this array
1435 */
1436int get_spare_criteria_imsm(struct supertype *st, struct spare_criteria *c)
80e7f8c3
AC		 1437{
1438 struct intel_super *super = st->sb;
1439 struct dl *dl;
1440 struct extent *e;
1441 int i;
fbfdcb06
AO		 1442 unsigned long long size = 0;
1443
1444 c->min_size = 0;
4b57ecf6 1445 c->sector_size = 0;
80e7f8c3
AC		 1446
1447 if (!super)
fbfdcb06 1448 return -EINVAL;
80e7f8c3
AC		 1449 /* find first active disk in array */
1450 dl = super->disks;
1451 while (dl && (is_failed(&dl->disk) || dl->index == -1))
1452 dl = dl->next;
1453 if (!dl)
fbfdcb06 1454 return -EINVAL;
80e7f8c3
AC		 1455 /* find last lba used by subarrays */
1456 e = get_extents(super, dl);
1457 if (!e)
fbfdcb06 1458 return -EINVAL;
80e7f8c3
AC		 1459 for (i = 0; e[i].size; i++)
1460 continue;
1461 if (i > 0)
fbfdcb06 1462 size = e[i-1].start + e[i-1].size;
80e7f8c3 1463 free(e);
b81221b7 1464
80e7f8c3 1465 /* add the amount of space needed for metadata */
fbfdcb06
AO		 1466 size += imsm_min_reserved_sectors(super);
1467
1468 c->min_size = size * 512;
4b57ecf6 1469 c->sector_size = super->sector_size;
b81221b7 1470
fbfdcb06 1471 return 0;
80e7f8c3
AC		 1472}
1473
d1e02575
AK		 1474static int is_gen_migration(struct imsm_dev *dev);
1475
f36a9ecd
PB		 1476#define IMSM_4K_DIV 8
1477
c47b0ff6
AK		 1478static __u64 blocks_per_migr_unit(struct intel_super *super,
1479 struct imsm_dev *dev);
1e5c6983 1480
c47b0ff6
AK		 1481static void print_imsm_dev(struct intel_super *super,
1482 struct imsm_dev *dev,
1483 char *uuid,
1484 int disk_idx)
cdddbdbc
DW		 1485{
1486 __u64 sz;
0d80bb2f 1487 int slot, i;
238c0a71
AK		 1488 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1489 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
b10b37b8 1490 __u32 ord;
cdddbdbc
DW		 1491
1492 printf("\n");
1e7bc0ed 1493 printf("[%.16s]:\n", dev->volume);
44470971 1494 printf(" UUID : %s\n", uuid);
dd8bcb3b
AK		 1495 printf(" RAID Level : %d", get_imsm_raid_level(map));
1496 if (map2)
1497 printf(" <-- %d", get_imsm_raid_level(map2));
1498 printf("\n");
1499 printf(" Members : %d", map->num_members);
1500 if (map2)
1501 printf(" <-- %d", map2->num_members);
1502 printf("\n");
0d80bb2f
DW		 1503 printf(" Slots : [");
1504 for (i = 0; i < map->num_members; i++) {
238c0a71 1505 ord = get_imsm_ord_tbl_ent(dev, i, MAP_0);
0d80bb2f
DW		 1506 printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U");
1507 }
dd8bcb3b
AK		 1508 printf("]");
1509 if (map2) {
1510 printf(" <-- [");
1511 for (i = 0; i < map2->num_members; i++) {
238c0a71 1512 ord = get_imsm_ord_tbl_ent(dev, i, MAP_1);
dd8bcb3b
AK		 1513 printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U");
1514 }
1515 printf("]");
1516 }
1517 printf("\n");
7095bccb
AK		 1518 printf(" Failed disk : ");
1519 if (map->failed_disk_num == 0xff)
1520 printf("none");
1521 else
1522 printf("%i", map->failed_disk_num);
1523 printf("\n");
620b1713
DW		 1524 slot = get_imsm_disk_slot(map, disk_idx);
1525 if (slot >= 0) {
238c0a71 1526 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
b10b37b8
DW		 1527 printf(" This Slot : %d%s\n", slot,
1528 ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : "");
1529 } else
cdddbdbc 1530 printf(" This Slot : ?\n");
84918897 1531 printf(" Sector Size : %u\n", super->sector_size);
fcc2c9da 1532 sz = imsm_dev_size(dev);
84918897
MK		 1533 printf(" Array Size : %llu%s\n",
1534 (unsigned long long)sz * 512 / super->sector_size,
cdddbdbc 1535 human_size(sz * 512));
5551b113 1536 sz = blocks_per_member(map);
84918897
MK		 1537 printf(" Per Dev Size : %llu%s\n",
1538 (unsigned long long)sz * 512 / super->sector_size,
cdddbdbc 1539 human_size(sz * 512));
5551b113
CA		 1540 printf(" Sector Offset : %llu\n",
1541 pba_of_lba0(map));
1542 printf(" Num Stripes : %llu\n",
1543 num_data_stripes(map));
dd8bcb3b 1544 printf(" Chunk Size : %u KiB",
cdddbdbc 1545 __le16_to_cpu(map->blocks_per_strip) / 2);
dd8bcb3b
AK		 1546 if (map2)
1547 printf(" <-- %u KiB",
1548 __le16_to_cpu(map2->blocks_per_strip) / 2);
1549 printf("\n");
cdddbdbc 1550 printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks));
8655a7b1 1551 printf(" Migrate State : ");
1484e727
DW		 1552 if (dev->vol.migr_state) {
1553 if (migr_type(dev) == MIGR_INIT)
8655a7b1 1554 printf("initialize\n");
1484e727 1555 else if (migr_type(dev) == MIGR_REBUILD)
8655a7b1 1556 printf("rebuild\n");
1484e727 1557 else if (migr_type(dev) == MIGR_VERIFY)
8655a7b1 1558 printf("check\n");
1484e727 1559 else if (migr_type(dev) == MIGR_GEN_MIGR)
8655a7b1 1560 printf("general migration\n");
1484e727 1561 else if (migr_type(dev) == MIGR_STATE_CHANGE)
8655a7b1 1562 printf("state change\n");
1484e727 1563 else if (migr_type(dev) == MIGR_REPAIR)
8655a7b1 1564 printf("repair\n");
1484e727 1565 else
8655a7b1
DW		 1566 printf("<unknown:%d>\n", migr_type(dev));
1567 } else
1568 printf("idle\n");
3393c6af
DW		 1569 printf(" Map State : %s", map_state_str[map->map_state]);
1570 if (dev->vol.migr_state) {
238c0a71 1571 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983 1572
b10b37b8 1573 printf(" <-- %s", map_state_str[map->map_state]);
464d40e8
LD		 1574 printf("\n Checkpoint : %u ",
1575 __le32_to_cpu(dev->vol.curr_migr_unit));
089f9d79 1576 if (is_gen_migration(dev) && (slot > 1 || slot < 0))
464d40e8
LD		 1577 printf("(N/A)");
1578 else
1579 printf("(%llu)", (unsigned long long)
1580 blocks_per_migr_unit(super, dev));
3393c6af
DW		 1581 }
1582 printf("\n");
2432ce9b
AP		 1583 printf(" Dirty State : %s\n", (dev->vol.dirty & RAIDVOL_DIRTY) ?
1584 "dirty" : "clean");
1585 printf(" RWH Policy : ");
c2462068 1586 if (dev->rwh_policy == RWH_OFF || dev->rwh_policy == RWH_MULTIPLE_OFF)
2432ce9b
AP		 1587 printf("off\n");
1588 else if (dev->rwh_policy == RWH_DISTRIBUTED)
1589 printf("PPL distributed\n");
1590 else if (dev->rwh_policy == RWH_JOURNALING_DRIVE)
1591 printf("PPL journaling drive\n");
c2462068
PB		 1592 else if (dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
1593 printf("Multiple distributed PPLs\n");
1594 else if (dev->rwh_policy == RWH_MULTIPLE_PPLS_JOURNALING_DRIVE)
1595 printf("Multiple PPLs on journaling drive\n");
2432ce9b
AP		 1596 else
1597 printf("<unknown:%d>\n", dev->rwh_policy);
cdddbdbc
DW		 1598}
1599
ef5c214e
MK		 1600static void print_imsm_disk(struct imsm_disk *disk,
1601 int index,
1602 __u32 reserved,
1603 unsigned int sector_size) {
1f24f035 1604 char str[MAX_RAID_SERIAL_LEN + 1];
cdddbdbc
DW		 1605 __u64 sz;
1606
0ec1f4e8 1607 if (index < -1 || !disk)
e9d82038
DW		 1608 return;
1609
cdddbdbc 1610 printf("\n");
1f24f035 1611 snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial);
0ec1f4e8
DW		 1612 if (index >= 0)
1613 printf(" Disk%02d Serial : %s\n", index, str);
1614 else
1615 printf(" Disk Serial : %s\n", str);
2432ce9b
AP		 1616 printf(" State :%s%s%s%s\n", is_spare(disk) ? " spare" : "",
1617 is_configured(disk) ? " active" : "",
1618 is_failed(disk) ? " failed" : "",
1619 is_journal(disk) ? " journal" : "");
cdddbdbc 1620 printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id));
5551b113 1621 sz = total_blocks(disk) - reserved;
ef5c214e
MK		 1622 printf(" Usable Size : %llu%s\n",
1623 (unsigned long long)sz * 512 / sector_size,
cdddbdbc
DW		 1624 human_size(sz * 512));
1625}
1626
de44e46f
PB		 1627void convert_to_4k_imsm_migr_rec(struct intel_super *super)
1628{
1629 struct migr_record *migr_rec = super->migr_rec;
1630
1631 migr_rec->blocks_per_unit /= IMSM_4K_DIV;
1632 migr_rec->ckpt_area_pba /= IMSM_4K_DIV;
1633 migr_rec->dest_1st_member_lba /= IMSM_4K_DIV;
1634 migr_rec->dest_depth_per_unit /= IMSM_4K_DIV;
1635 split_ull((join_u32(migr_rec->post_migr_vol_cap,
1636 migr_rec->post_migr_vol_cap_hi) / IMSM_4K_DIV),
1637 &migr_rec->post_migr_vol_cap, &migr_rec->post_migr_vol_cap_hi);
1638}
1639
f36a9ecd
PB		 1640void convert_to_4k_imsm_disk(struct imsm_disk *disk)
1641{
1642 set_total_blocks(disk, (total_blocks(disk)/IMSM_4K_DIV));
1643}
1644
1645void convert_to_4k(struct intel_super *super)
1646{
1647 struct imsm_super *mpb = super->anchor;
1648 struct imsm_disk *disk;
1649 int i;
e4467bc7 1650 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
f36a9ecd
PB		 1651
1652 for (i = 0; i < mpb->num_disks ; i++) {
1653 disk = __get_imsm_disk(mpb, i);
1654 /* disk */
1655 convert_to_4k_imsm_disk(disk);
1656 }
1657 for (i = 0; i < mpb->num_raid_devs; i++) {
1658 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
1659 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1660 /* dev */
fcc2c9da 1661 set_imsm_dev_size(dev, imsm_dev_size(dev)/IMSM_4K_DIV);
f36a9ecd
PB		 1662 dev->vol.curr_migr_unit /= IMSM_4K_DIV;
1663
1664 /* map0 */
1665 set_blocks_per_member(map, blocks_per_member(map)/IMSM_4K_DIV);
1666 map->blocks_per_strip /= IMSM_4K_DIV;
1667 set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV);
1668
1669 if (dev->vol.migr_state) {
1670 /* map1 */
1671 map = get_imsm_map(dev, MAP_1);
1672 set_blocks_per_member(map,
1673 blocks_per_member(map)/IMSM_4K_DIV);
1674 map->blocks_per_strip /= IMSM_4K_DIV;
1675 set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV);
1676 }
1677 }
e4467bc7
TM		 1678 if (bbm_log_size) {
1679 struct bbm_log *log = (void *)mpb +
1680 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
1681 __u32 i;
1682
1683 for (i = 0; i < log->entry_count; i++) {
1684 struct bbm_log_entry *entry =
1685 &log->marked_block_entries[i];
1686
1687 __u8 count = entry->marked_count + 1;
1688 unsigned long long sector =
1689 __le48_to_cpu(&entry->defective_block_start);
1690
1691 entry->defective_block_start =
1692 __cpu_to_le48(sector/IMSM_4K_DIV);
1693 entry->marked_count = max(count/IMSM_4K_DIV, 1) - 1;
1694 }
1695 }
f36a9ecd
PB		 1696
1697 mpb->check_sum = __gen_imsm_checksum(mpb);
1698}
1699
520e69e2
AK		 1700void examine_migr_rec_imsm(struct intel_super *super)
1701{
1702 struct migr_record *migr_rec = super->migr_rec;
1703 struct imsm_super *mpb = super->anchor;
1704 int i;
1705
1706 for (i = 0; i < mpb->num_raid_devs; i++) {
1707 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
3136abe5 1708 struct imsm_map *map;
b4ab44d8 1709 int slot = -1;
3136abe5 1710
520e69e2
AK		 1711 if (is_gen_migration(dev) == 0)
1712 continue;
1713
1714 printf("\nMigration Record Information:");
3136abe5 1715
44bfe6df
AK		 1716 /* first map under migration */
1717 map = get_imsm_map(dev, MAP_0);
3136abe5
AK		 1718 if (map)
1719 slot = get_imsm_disk_slot(map, super->disks->index);
089f9d79 1720 if (map == NULL || slot > 1 || slot < 0) {
520e69e2
AK		 1721 printf(" Empty\n ");
1722 printf("Examine one of first two disks in array\n");
1723 break;
1724 }
1725 printf("\n Status : ");
1726 if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL)
1727 printf("Normal\n");
1728 else
1729 printf("Contains Data\n");
1730 printf(" Current Unit : %u\n",
1731 __le32_to_cpu(migr_rec->curr_migr_unit));
1732 printf(" Family : %u\n",
1733 __le32_to_cpu(migr_rec->family_num));
1734 printf(" Ascending : %u\n",
1735 __le32_to_cpu(migr_rec->ascending_migr));
1736 printf(" Blocks Per Unit : %u\n",
1737 __le32_to_cpu(migr_rec->blocks_per_unit));
1738 printf(" Dest. Depth Per Unit : %u\n",
1739 __le32_to_cpu(migr_rec->dest_depth_per_unit));
1740 printf(" Checkpoint Area pba : %u\n",
1741 __le32_to_cpu(migr_rec->ckpt_area_pba));
1742 printf(" First member lba : %u\n",
1743 __le32_to_cpu(migr_rec->dest_1st_member_lba));
1744 printf(" Total Number of Units : %u\n",
1745 __le32_to_cpu(migr_rec->num_migr_units));
1746 printf(" Size of volume : %u\n",
1747 __le32_to_cpu(migr_rec->post_migr_vol_cap));
1748 printf(" Expansion space for LBA64 : %u\n",
1749 __le32_to_cpu(migr_rec->post_migr_vol_cap_hi));
1750 printf(" Record was read from : %u\n",
1751 __le32_to_cpu(migr_rec->ckpt_read_disk_num));
1752
1753 break;
1754 }
1755}
f36a9ecd 1756
de44e46f
PB		 1757void convert_from_4k_imsm_migr_rec(struct intel_super *super)
1758{
1759 struct migr_record *migr_rec = super->migr_rec;
1760
1761 migr_rec->blocks_per_unit *= IMSM_4K_DIV;
1762 migr_rec->ckpt_area_pba *= IMSM_4K_DIV;
1763 migr_rec->dest_1st_member_lba *= IMSM_4K_DIV;
1764 migr_rec->dest_depth_per_unit *= IMSM_4K_DIV;
1765 split_ull((join_u32(migr_rec->post_migr_vol_cap,
1766 migr_rec->post_migr_vol_cap_hi) * IMSM_4K_DIV),
1767 &migr_rec->post_migr_vol_cap,
1768 &migr_rec->post_migr_vol_cap_hi);
1769}
1770
f36a9ecd
PB		 1771void convert_from_4k(struct intel_super *super)
1772{
1773 struct imsm_super *mpb = super->anchor;
1774 struct imsm_disk *disk;
1775 int i;
e4467bc7 1776 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
f36a9ecd
PB		 1777
1778 for (i = 0; i < mpb->num_disks ; i++) {
1779 disk = __get_imsm_disk(mpb, i);
1780 /* disk */
1781 set_total_blocks(disk, (total_blocks(disk)*IMSM_4K_DIV));
1782 }
1783
1784 for (i = 0; i < mpb->num_raid_devs; i++) {
1785 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
1786 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1787 /* dev */
fcc2c9da 1788 set_imsm_dev_size(dev, imsm_dev_size(dev)*IMSM_4K_DIV);
f36a9ecd
PB		 1789 dev->vol.curr_migr_unit *= IMSM_4K_DIV;
1790
1791 /* map0 */
1792 set_blocks_per_member(map, blocks_per_member(map)*IMSM_4K_DIV);
1793 map->blocks_per_strip *= IMSM_4K_DIV;
1794 set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV);
1795
1796 if (dev->vol.migr_state) {
1797 /* map1 */
1798 map = get_imsm_map(dev, MAP_1);
1799 set_blocks_per_member(map,
1800 blocks_per_member(map)*IMSM_4K_DIV);
1801 map->blocks_per_strip *= IMSM_4K_DIV;
1802 set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV);
1803 }
1804 }
e4467bc7
TM		 1805 if (bbm_log_size) {
1806 struct bbm_log *log = (void *)mpb +
1807 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
1808 __u32 i;
1809
1810 for (i = 0; i < log->entry_count; i++) {
1811 struct bbm_log_entry *entry =
1812 &log->marked_block_entries[i];
1813
1814 __u8 count = entry->marked_count + 1;
1815 unsigned long long sector =
1816 __le48_to_cpu(&entry->defective_block_start);
1817
1818 entry->defective_block_start =
1819 __cpu_to_le48(sector*IMSM_4K_DIV);
1820 entry->marked_count = count*IMSM_4K_DIV - 1;
1821 }
1822 }
f36a9ecd
PB		 1823
1824 mpb->check_sum = __gen_imsm_checksum(mpb);
1825}
1826
19482bcc
AK		 1827/*******************************************************************************
1828 * function: imsm_check_attributes
1829 * Description: Function checks if features represented by attributes flags
1011e834 1830 * are supported by mdadm.
19482bcc
AK		 1831 * Parameters:
1832 * attributes - Attributes read from metadata
1833 * Returns:
1011e834
N		 1834 * 0 - passed attributes contains unsupported features flags
1835 * 1 - all features are supported
19482bcc
AK		 1836 ******************************************************************************/
1837static int imsm_check_attributes(__u32 attributes)
1838{
1839 int ret_val = 1;
418f9b36
N		 1840 __u32 not_supported = MPB_ATTRIB_SUPPORTED^0xffffffff;
1841
1842 not_supported &= ~MPB_ATTRIB_IGNORED;
19482bcc
AK		 1843
1844 not_supported &= attributes;
1845 if (not_supported) {
e7b84f9d 1846 pr_err("(IMSM): Unsupported attributes : %x\n",
418f9b36 1847 (unsigned)__le32_to_cpu(not_supported));
19482bcc
AK		 1848 if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) {
1849 dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY \n");
1850 not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY;
1851 }
1852 if (not_supported & MPB_ATTRIB_2TB) {
1853 dprintf("\t\tMPB_ATTRIB_2TB\n");
1854 not_supported ^= MPB_ATTRIB_2TB;
1855 }
1856 if (not_supported & MPB_ATTRIB_RAID0) {
1857 dprintf("\t\tMPB_ATTRIB_RAID0\n");
1858 not_supported ^= MPB_ATTRIB_RAID0;
1859 }
1860 if (not_supported & MPB_ATTRIB_RAID1) {
1861 dprintf("\t\tMPB_ATTRIB_RAID1\n");
1862 not_supported ^= MPB_ATTRIB_RAID1;
1863 }
1864 if (not_supported & MPB_ATTRIB_RAID10) {
1865 dprintf("\t\tMPB_ATTRIB_RAID10\n");
1866 not_supported ^= MPB_ATTRIB_RAID10;
1867 }
1868 if (not_supported & MPB_ATTRIB_RAID1E) {
1869 dprintf("\t\tMPB_ATTRIB_RAID1E\n");
1870 not_supported ^= MPB_ATTRIB_RAID1E;
1871 }
1872 if (not_supported & MPB_ATTRIB_RAID5) {
1873 dprintf("\t\tMPB_ATTRIB_RAID5\n");
1874 not_supported ^= MPB_ATTRIB_RAID5;
1875 }
1876 if (not_supported & MPB_ATTRIB_RAIDCNG) {
1877 dprintf("\t\tMPB_ATTRIB_RAIDCNG\n");
1878 not_supported ^= MPB_ATTRIB_RAIDCNG;
1879 }
1880 if (not_supported & MPB_ATTRIB_BBM) {
1881 dprintf("\t\tMPB_ATTRIB_BBM\n");
1882 not_supported ^= MPB_ATTRIB_BBM;
1883 }
1884 if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) {
1885 dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY (== MPB_ATTRIB_LEGACY)\n");
1886 not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY;
1887 }
1888 if (not_supported & MPB_ATTRIB_EXP_STRIPE_SIZE) {
1889 dprintf("\t\tMPB_ATTRIB_EXP_STRIP_SIZE\n");
1890 not_supported ^= MPB_ATTRIB_EXP_STRIPE_SIZE;
1891 }
1892 if (not_supported & MPB_ATTRIB_2TB_DISK) {
1893 dprintf("\t\tMPB_ATTRIB_2TB_DISK\n");
1894 not_supported ^= MPB_ATTRIB_2TB_DISK;
1895 }
1896 if (not_supported & MPB_ATTRIB_NEVER_USE2) {
1897 dprintf("\t\tMPB_ATTRIB_NEVER_USE2\n");
1898 not_supported ^= MPB_ATTRIB_NEVER_USE2;
1899 }
1900 if (not_supported & MPB_ATTRIB_NEVER_USE) {
1901 dprintf("\t\tMPB_ATTRIB_NEVER_USE\n");
1902 not_supported ^= MPB_ATTRIB_NEVER_USE;
1903 }
1904
1905 if (not_supported)
1ade5cc1 1906 dprintf("(IMSM): Unknown attributes : %x\n", not_supported);
19482bcc
AK		 1907
1908 ret_val = 0;
1909 }
1910
1911 return ret_val;
1912}
1913
a5d85af7 1914static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map);
44470971 1915
cdddbdbc
DW		 1916static void examine_super_imsm(struct supertype *st, char *homehost)
1917{
1918 struct intel_super *super = st->sb;
949c47a0 1919 struct imsm_super *mpb = super->anchor;
cdddbdbc
DW		 1920 char str[MAX_SIGNATURE_LENGTH];
1921 int i;
27fd6274
DW		 1922 struct mdinfo info;
1923 char nbuf[64];
cdddbdbc 1924 __u32 sum;
14e8215b 1925 __u32 reserved = imsm_reserved_sectors(super, super->disks);
94827db3 1926 struct dl *dl;
27fd6274 1927
618f4e6d
XN		 1928 strncpy(str, (char *)mpb->sig, MPB_SIG_LEN);
1929 str[MPB_SIG_LEN-1] = '\0';
cdddbdbc
DW		 1930 printf(" Magic : %s\n", str);
1931 snprintf(str, strlen(MPB_VERSION_RAID0), "%s", get_imsm_version(mpb));
1932 printf(" Version : %s\n", get_imsm_version(mpb));
148acb7b 1933 printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num));
cdddbdbc
DW		 1934 printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num));
1935 printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num));
19482bcc
AK		 1936 printf(" Attributes : ");
1937 if (imsm_check_attributes(mpb->attributes))
1938 printf("All supported\n");
1939 else
1940 printf("not supported\n");
a5d85af7 1941 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 1942 fname_from_uuid(st, &info, nbuf, ':');
27fd6274 1943 printf(" UUID : %s\n", nbuf + 5);
cdddbdbc
DW		 1944 sum = __le32_to_cpu(mpb->check_sum);
1945 printf(" Checksum : %08x %s\n", sum,
949c47a0 1946 __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect");
f36a9ecd 1947 printf(" MPB Sectors : %d\n", mpb_sectors(mpb, super->sector_size));
cdddbdbc
DW		 1948 printf(" Disks : %d\n", mpb->num_disks);
1949 printf(" RAID Devices : %d\n", mpb->num_raid_devs);
ef5c214e
MK		 1950 print_imsm_disk(__get_imsm_disk(mpb, super->disks->index),
1951 super->disks->index, reserved, super->sector_size);
8d67477f 1952 if (get_imsm_bbm_log_size(super->bbm_log)) {
604b746f
JD		 1953 struct bbm_log *log = super->bbm_log;
1954
1955 printf("\n");
1956 printf("Bad Block Management Log:\n");
1957 printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size));
1958 printf(" Signature : %x\n", __le32_to_cpu(log->signature));
1959 printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count));
604b746f 1960 }
44470971
DW		 1961 for (i = 0; i < mpb->num_raid_devs; i++) {
1962 struct mdinfo info;
1963 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
1964
1965 super->current_vol = i;
a5d85af7 1966 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 1967 fname_from_uuid(st, &info, nbuf, ':');
c47b0ff6 1968 print_imsm_dev(super, dev, nbuf + 5, super->disks->index);
44470971 1969 }
cdddbdbc
DW		 1970 for (i = 0; i < mpb->num_disks; i++) {
1971 if (i == super->disks->index)
1972 continue;
ef5c214e
MK		 1973 print_imsm_disk(__get_imsm_disk(mpb, i), i, reserved,
1974 super->sector_size);
cdddbdbc 1975 }
94827db3 1976
0ec1f4e8
DW		 1977 for (dl = super->disks; dl; dl = dl->next)
1978 if (dl->index == -1)
ef5c214e
MK		 1979 print_imsm_disk(&dl->disk, -1, reserved,
1980 super->sector_size);
520e69e2
AK		 1981
1982 examine_migr_rec_imsm(super);
cdddbdbc
DW		 1983}
1984
061f2c6a 1985static void brief_examine_super_imsm(struct supertype *st, int verbose)
cdddbdbc 1986{
27fd6274 1987 /* We just write a generic IMSM ARRAY entry */
ff54de6e
N		 1988 struct mdinfo info;
1989 char nbuf[64];
1e7bc0ed 1990 struct intel_super *super = st->sb;
1e7bc0ed 1991
0d5a423f
DW		 1992 if (!super->anchor->num_raid_devs) {
1993 printf("ARRAY metadata=imsm\n");
1e7bc0ed 1994 return;
0d5a423f 1995 }
ff54de6e 1996
a5d85af7 1997 getinfo_super_imsm(st, &info, NULL);
4737ae25
N		 1998 fname_from_uuid(st, &info, nbuf, ':');
1999 printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5);
2000}
2001
2002static void brief_examine_subarrays_imsm(struct supertype *st, int verbose)
2003{
2004 /* We just write a generic IMSM ARRAY entry */
2005 struct mdinfo info;
2006 char nbuf[64];
2007 char nbuf1[64];
2008 struct intel_super *super = st->sb;
2009 int i;
2010
2011 if (!super->anchor->num_raid_devs)
2012 return;
2013
a5d85af7 2014 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2015 fname_from_uuid(st, &info, nbuf, ':');
1e7bc0ed
DW		 2016 for (i = 0; i < super->anchor->num_raid_devs; i++) {
2017 struct imsm_dev *dev = get_imsm_dev(super, i);
2018
2019 super->current_vol = i;
a5d85af7 2020 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2021 fname_from_uuid(st, &info, nbuf1, ':');
1124b3cf 2022 printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n",
cf8de691 2023 dev->volume, nbuf + 5, i, nbuf1 + 5);
1e7bc0ed 2024 }
cdddbdbc
DW		 2025}
2026
9d84c8ea
DW		 2027static void export_examine_super_imsm(struct supertype *st)
2028{
2029 struct intel_super *super = st->sb;
2030 struct imsm_super *mpb = super->anchor;
2031 struct mdinfo info;
2032 char nbuf[64];
2033
a5d85af7 2034 getinfo_super_imsm(st, &info, NULL);
9d84c8ea
DW		 2035 fname_from_uuid(st, &info, nbuf, ':');
2036 printf("MD_METADATA=imsm\n");
2037 printf("MD_LEVEL=container\n");
2038 printf("MD_UUID=%s\n", nbuf+5);
2039 printf("MD_DEVICES=%u\n", mpb->num_disks);
2040}
2041
74db60b0
N		 2042static int copy_metadata_imsm(struct supertype *st, int from, int to)
2043{
f36a9ecd 2044 /* The second last sector of the device contains
74db60b0
N		 2045 * the "struct imsm_super" metadata.
2046 * This contains mpb_size which is the size in bytes of the
2047 * extended metadata. This is located immediately before
2048 * the imsm_super.
2049 * We want to read all that, plus the last sector which
2050 * may contain a migration record, and write it all
2051 * to the target.
2052 */
2053 void *buf;
2054 unsigned long long dsize, offset;
2055 int sectors;
2056 struct imsm_super *sb;
f36a9ecd
PB		 2057 struct intel_super *super = st->sb;
2058 unsigned int sector_size = super->sector_size;
2059 unsigned int written = 0;
74db60b0 2060
de44e46f 2061 if (posix_memalign(&buf, MAX_SECTOR_SIZE, MAX_SECTOR_SIZE) != 0)
74db60b0
N		 2062 return 1;
2063
2064 if (!get_dev_size(from, NULL, &dsize))
2065 goto err;
2066
f36a9ecd 2067 if (lseek64(from, dsize-(2*sector_size), 0) < 0)
74db60b0 2068 goto err;
466070ad 2069 if ((unsigned int)read(from, buf, sector_size) != sector_size)
74db60b0
N		 2070 goto err;
2071 sb = buf;
2072 if (strncmp((char*)sb->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0)
2073 goto err;
2074
f36a9ecd
PB		 2075 sectors = mpb_sectors(sb, sector_size) + 2;
2076 offset = dsize - sectors * sector_size;
74db60b0
N		 2077 if (lseek64(from, offset, 0) < 0 ||
2078 lseek64(to, offset, 0) < 0)
2079 goto err;
f36a9ecd
PB		 2080 while (written < sectors * sector_size) {
2081 int n = sectors*sector_size - written;
74db60b0
N		 2082 if (n > 4096)
2083 n = 4096;
2084 if (read(from, buf, n) != n)
2085 goto err;
2086 if (write(to, buf, n) != n)
2087 goto err;
2088 written += n;
2089 }
2090 free(buf);
2091 return 0;
2092err:
2093 free(buf);
2094 return 1;
2095}
2096
cdddbdbc
DW		 2097static void detail_super_imsm(struct supertype *st, char *homehost)
2098{
3ebe00a1
DW		 2099 struct mdinfo info;
2100 char nbuf[64];
2101
a5d85af7 2102 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2103 fname_from_uuid(st, &info, nbuf, ':');
65884368 2104 printf("\n UUID : %s\n", nbuf + 5);
cdddbdbc
DW		 2105}
2106
2107static void brief_detail_super_imsm(struct supertype *st)
2108{
ff54de6e
N		 2109 struct mdinfo info;
2110 char nbuf[64];
a5d85af7 2111 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2112 fname_from_uuid(st, &info, nbuf, ':');
ff54de6e 2113 printf(" UUID=%s", nbuf + 5);
cdddbdbc 2114}
d665cc31
DW		 2115
2116static int imsm_read_serial(int fd, char *devname, __u8 *serial);
2117static void fd2devname(int fd, char *name);
2118
120dc887 2119static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose)
d665cc31 2120{
120dc887
LM		 2121 /* dump an unsorted list of devices attached to AHCI Intel storage
2122 * controller, as well as non-connected ports
d665cc31
DW		 2123 */
2124 int hba_len = strlen(hba_path) + 1;
2125 struct dirent *ent;
2126 DIR *dir;
2127 char *path = NULL;
2128 int err = 0;
2129 unsigned long port_mask = (1 << port_count) - 1;
2130
f21e18ca 2131 if (port_count > (int)sizeof(port_mask) * 8) {
ba728be7 2132 if (verbose > 0)
e7b84f9d 2133 pr_err("port_count %d out of range\n", port_count);
d665cc31
DW		 2134 return 2;
2135 }
2136
2137 /* scroll through /sys/dev/block looking for devices attached to
2138 * this hba
2139 */
2140 dir = opendir("/sys/dev/block");
1a6dd6b9
PB		 2141 if (!dir)
2142 return 1;
2143
2144 for (ent = readdir(dir); ent; ent = readdir(dir)) {
d665cc31
DW		 2145 int fd;
2146 char model[64];
2147 char vendor[64];
2148 char buf[1024];
2149 int major, minor;
2150 char *device;
2151 char *c;
2152 int port;
2153 int type;
2154
2155 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
2156 continue;
2157 path = devt_to_devpath(makedev(major, minor));
2158 if (!path)
2159 continue;
2160 if (!path_attached_to_hba(path, hba_path)) {
2161 free(path);
2162 path = NULL;
2163 continue;
2164 }
2165
2166 /* retrieve the scsi device type */
2167 if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) {
ba728be7 2168 if (verbose > 0)
e7b84f9d 2169 pr_err("failed to allocate 'device'\n");
d665cc31
DW		 2170 err = 2;
2171 break;
2172 }
2173 sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor);
193b6c0b 2174 if (load_sys(device, buf, sizeof(buf)) != 0) {
ba728be7 2175 if (verbose > 0)
e7b84f9d 2176 pr_err("failed to read device type for %s\n",
d665cc31
DW		 2177 path);
2178 err = 2;
2179 free(device);
2180 break;
2181 }
2182 type = strtoul(buf, NULL, 10);
2183
2184 /* if it's not a disk print the vendor and model */
2185 if (!(type == 0 || type == 7 || type == 14)) {
2186 vendor[0] = '\0';
2187 model[0] = '\0';
2188 sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor);
193b6c0b 2189 if (load_sys(device, buf, sizeof(buf)) == 0) {
d665cc31
DW		 2190 strncpy(vendor, buf, sizeof(vendor));
2191 vendor[sizeof(vendor) - 1] = '\0';
2192 c = (char *) &vendor[sizeof(vendor) - 1];
2193 while (isspace(*c) || *c == '\0')
2194 *c-- = '\0';
2195
2196 }
2197 sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor);
193b6c0b 2198 if (load_sys(device, buf, sizeof(buf)) == 0) {
d665cc31
DW		 2199 strncpy(model, buf, sizeof(model));
2200 model[sizeof(model) - 1] = '\0';
2201 c = (char *) &model[sizeof(model) - 1];
2202 while (isspace(*c) || *c == '\0')
2203 *c-- = '\0';
2204 }
2205
2206 if (vendor[0] && model[0])
2207 sprintf(buf, "%.64s %.64s", vendor, model);
2208 else
2209 switch (type) { /* numbers from hald/linux/device.c */
2210 case 1: sprintf(buf, "tape"); break;
2211 case 2: sprintf(buf, "printer"); break;
2212 case 3: sprintf(buf, "processor"); break;
2213 case 4:
2214 case 5: sprintf(buf, "cdrom"); break;
2215 case 6: sprintf(buf, "scanner"); break;
2216 case 8: sprintf(buf, "media_changer"); break;
2217 case 9: sprintf(buf, "comm"); break;
2218 case 12: sprintf(buf, "raid"); break;
2219 default: sprintf(buf, "unknown");
2220 }
2221 } else
2222 buf[0] = '\0';
2223 free(device);
2224
2225 /* chop device path to 'host%d' and calculate the port number */
2226 c = strchr(&path[hba_len], '/');
4e5e717d 2227 if (!c) {
ba728be7 2228 if (verbose > 0)
e7b84f9d 2229 pr_err("%s - invalid path name\n", path + hba_len);
4e5e717d
AW		 2230 err = 2;
2231 break;
2232 }
d665cc31 2233 *c = '\0';
0858eccf
AP		 2234 if ((sscanf(&path[hba_len], "ata%d", &port) == 1) ||
2235 ((sscanf(&path[hba_len], "host%d", &port) == 1)))
d665cc31
DW		 2236 port -= host_base;
2237 else {
ba728be7 2238 if (verbose > 0) {
d665cc31 2239 *c = '/'; /* repair the full string */
e7b84f9d 2240 pr_err("failed to determine port number for %s\n",
d665cc31
DW		 2241 path);
2242 }
2243 err = 2;
2244 break;
2245 }
2246
2247 /* mark this port as used */
2248 port_mask &= ~(1 << port);
2249
2250 /* print out the device information */
2251 if (buf[0]) {
2252 printf(" Port%d : - non-disk device (%s) -\n", port, buf);
2253 continue;
2254 }
2255
2256 fd = dev_open(ent->d_name, O_RDONLY);
2257 if (fd < 0)
2258 printf(" Port%d : - disk info unavailable -\n", port);
2259 else {
2260 fd2devname(fd, buf);
2261 printf(" Port%d : %s", port, buf);
2262 if (imsm_read_serial(fd, NULL, (__u8 *) buf) == 0)
664d5325 2263 printf(" (%.*s)\n", MAX_RAID_SERIAL_LEN, buf);
d665cc31 2264 else
664d5325 2265 printf(" ()\n");
4dab422a 2266 close(fd);
d665cc31 2267 }
d665cc31
DW		 2268 free(path);
2269 path = NULL;
2270 }
2271 if (path)
2272 free(path);
2273 if (dir)
2274 closedir(dir);
2275 if (err == 0) {
2276 int i;
2277
2278 for (i = 0; i < port_count; i++)
2279 if (port_mask & (1 << i))
2280 printf(" Port%d : - no device attached -\n", i);
2281 }
2282
2283 return err;
2284}
2285
b5eece69 2286static int print_vmd_attached_devs(struct sys_dev *hba)
60f0f54d
PB		 2287{
2288 struct dirent *ent;
2289 DIR *dir;
2290 char path[292];
2291 char link[256];
2292 char *c, *rp;
2293
2294 if (hba->type != SYS_DEV_VMD)
b5eece69 2295 return 1;
60f0f54d
PB		 2296
2297 /* scroll through /sys/dev/block looking for devices attached to
2298 * this hba
2299 */
2300 dir = opendir("/sys/bus/pci/drivers/nvme");
b9135011 2301 if (!dir)
b5eece69 2302 return 1;
b9135011
JS		 2303
2304 for (ent = readdir(dir); ent; ent = readdir(dir)) {
60f0f54d
PB		 2305 int n;
2306
2307 /* is 'ent' a device? check that the 'subsystem' link exists and
2308 * that its target matches 'bus'
2309 */
2310 sprintf(path, "/sys/bus/pci/drivers/nvme/%s/subsystem",
2311 ent->d_name);
2312 n = readlink(path, link, sizeof(link));
2313 if (n < 0 || n >= (int)sizeof(link))
2314 continue;
2315 link[n] = '\0';
2316 c = strrchr(link, '/');
2317 if (!c)
2318 continue;
2319 if (strncmp("pci", c+1, strlen("pci")) != 0)
2320 continue;
2321
2322 sprintf(path, "/sys/bus/pci/drivers/nvme/%s", ent->d_name);
60f0f54d
PB		 2323
2324 rp = realpath(path, NULL);
2325 if (!rp)
2326 continue;
2327
2328 if (path_attached_to_hba(rp, hba->path)) {
2329 printf(" NVMe under VMD : %s\n", rp);
2330 }
2331 free(rp);
2332 }
2333
b9135011 2334 closedir(dir);
b5eece69 2335 return 0;
60f0f54d
PB		 2336}
2337
120dc887
LM		 2338static void print_found_intel_controllers(struct sys_dev *elem)
2339{
2340 for (; elem; elem = elem->next) {
e7b84f9d 2341 pr_err("found Intel(R) ");
120dc887
LM		 2342 if (elem->type == SYS_DEV_SATA)
2343 fprintf(stderr, "SATA ");
155cbb4c
LM		 2344 else if (elem->type == SYS_DEV_SAS)
2345 fprintf(stderr, "SAS ");
0858eccf
AP		 2346 else if (elem->type == SYS_DEV_NVME)
2347 fprintf(stderr, "NVMe ");
60f0f54d
PB		 2348
2349 if (elem->type == SYS_DEV_VMD)
2350 fprintf(stderr, "VMD domain");
2351 else
2352 fprintf(stderr, "RAID controller");
2353
120dc887
LM		 2354 if (elem->pci_id)
2355 fprintf(stderr, " at %s", elem->pci_id);
2356 fprintf(stderr, ".\n");
2357 }
2358 fflush(stderr);
2359}
2360
120dc887
LM		 2361static int ahci_get_port_count(const char *hba_path, int *port_count)
2362{
2363 struct dirent *ent;
2364 DIR *dir;
2365 int host_base = -1;
2366
2367 *port_count = 0;
2368 if ((dir = opendir(hba_path)) == NULL)
2369 return -1;
2370
2371 for (ent = readdir(dir); ent; ent = readdir(dir)) {
2372 int host;
2373
0858eccf
AP		 2374 if ((sscanf(ent->d_name, "ata%d", &host) != 1) &&
2375 ((sscanf(ent->d_name, "host%d", &host) != 1)))
120dc887
LM		 2376 continue;
2377 if (*port_count == 0)
2378 host_base = host;
2379 else if (host < host_base)
2380 host_base = host;
2381
2382 if (host + 1 > *port_count + host_base)
2383 *port_count = host + 1 - host_base;
2384 }
2385 closedir(dir);
2386 return host_base;
2387}
2388
a891a3c2
LM		 2389static void print_imsm_capability(const struct imsm_orom *orom)
2390{
0858eccf
AP		 2391 printf(" Platform : Intel(R) ");
2392 if (orom->capabilities == 0 && orom->driver_features == 0)
2393 printf("Matrix Storage Manager\n");
ab0c6bb9
AP		 2394 else if (imsm_orom_is_enterprise(orom) && orom->major_ver >= 6)
2395 printf("Virtual RAID on CPU\n");
0858eccf
AP		 2396 else
2397 printf("Rapid Storage Technology%s\n",
2398 imsm_orom_is_enterprise(orom) ? " enterprise" : "");
2399 if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build)
2400 printf(" Version : %d.%d.%d.%d\n", orom->major_ver,
2401 orom->minor_ver, orom->hotfix_ver, orom->build);
a891a3c2
LM		 2402 printf(" RAID Levels :%s%s%s%s%s\n",
2403 imsm_orom_has_raid0(orom) ? " raid0" : "",
2404 imsm_orom_has_raid1(orom) ? " raid1" : "",
2405 imsm_orom_has_raid1e(orom) ? " raid1e" : "",
2406 imsm_orom_has_raid10(orom) ? " raid10" : "",
2407 imsm_orom_has_raid5(orom) ? " raid5" : "");
2408 printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
2409 imsm_orom_has_chunk(orom, 2) ? " 2k" : "",
2410 imsm_orom_has_chunk(orom, 4) ? " 4k" : "",
2411 imsm_orom_has_chunk(orom, 8) ? " 8k" : "",
2412 imsm_orom_has_chunk(orom, 16) ? " 16k" : "",
2413 imsm_orom_has_chunk(orom, 32) ? " 32k" : "",
2414 imsm_orom_has_chunk(orom, 64) ? " 64k" : "",
2415 imsm_orom_has_chunk(orom, 128) ? " 128k" : "",
2416 imsm_orom_has_chunk(orom, 256) ? " 256k" : "",
2417 imsm_orom_has_chunk(orom, 512) ? " 512k" : "",
2418 imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "",
2419 imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "",
2420 imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "",
2421 imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "",
2422 imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "",
2423 imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "",
2424 imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : "");
29cd0821
CA		 2425 printf(" 2TB volumes :%s supported\n",
2426 (orom->attr & IMSM_OROM_ATTR_2TB)?"":" not");
2427 printf(" 2TB disks :%s supported\n",
2428 (orom->attr & IMSM_OROM_ATTR_2TB_DISK)?"":" not");
0e7f69a8 2429 printf(" Max Disks : %d\n", orom->tds);
0858eccf
AP		 2430 printf(" Max Volumes : %d per array, %d per %s\n",
2431 orom->vpa, orom->vphba,
2432 imsm_orom_is_nvme(orom) ? "platform" : "controller");
a891a3c2
LM		 2433 return;
2434}
2435
e50cf220
MN		 2436static void print_imsm_capability_export(const struct imsm_orom *orom)
2437{
2438 printf("MD_FIRMWARE_TYPE=imsm\n");
0858eccf
AP		 2439 if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build)
2440 printf("IMSM_VERSION=%d.%d.%d.%d\n", orom->major_ver, orom->minor_ver,
2441 orom->hotfix_ver, orom->build);
e50cf220
MN		 2442 printf("IMSM_SUPPORTED_RAID_LEVELS=%s%s%s%s%s\n",
2443 imsm_orom_has_raid0(orom) ? "raid0 " : "",
2444 imsm_orom_has_raid1(orom) ? "raid1 " : "",
2445 imsm_orom_has_raid1e(orom) ? "raid1e " : "",
2446 imsm_orom_has_raid5(orom) ? "raid10 " : "",
2447 imsm_orom_has_raid10(orom) ? "raid5 " : "");
2448 printf("IMSM_SUPPORTED_CHUNK_SIZES=%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
2449 imsm_orom_has_chunk(orom, 2) ? "2k " : "",
2450 imsm_orom_has_chunk(orom, 4) ? "4k " : "",
2451 imsm_orom_has_chunk(orom, 8) ? "8k " : "",
2452 imsm_orom_has_chunk(orom, 16) ? "16k " : "",
2453 imsm_orom_has_chunk(orom, 32) ? "32k " : "",
2454 imsm_orom_has_chunk(orom, 64) ? "64k " : "",
2455 imsm_orom_has_chunk(orom, 128) ? "128k " : "",
2456 imsm_orom_has_chunk(orom, 256) ? "256k " : "",
2457 imsm_orom_has_chunk(orom, 512) ? "512k " : "",
2458 imsm_orom_has_chunk(orom, 1024*1) ? "1M " : "",
2459 imsm_orom_has_chunk(orom, 1024*2) ? "2M " : "",
2460 imsm_orom_has_chunk(orom, 1024*4) ? "4M " : "",
2461 imsm_orom_has_chunk(orom, 1024*8) ? "8M " : "",
2462 imsm_orom_has_chunk(orom, 1024*16) ? "16M " : "",
2463 imsm_orom_has_chunk(orom, 1024*32) ? "32M " : "",
2464 imsm_orom_has_chunk(orom, 1024*64) ? "64M " : "");
2465 printf("IMSM_2TB_VOLUMES=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB) ? "yes" : "no");
2466 printf("IMSM_2TB_DISKS=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB_DISK) ? "yes" : "no");
2467 printf("IMSM_MAX_DISKS=%d\n",orom->tds);
2468 printf("IMSM_MAX_VOLUMES_PER_ARRAY=%d\n",orom->vpa);
2469 printf("IMSM_MAX_VOLUMES_PER_CONTROLLER=%d\n",orom->vphba);
2470}
2471
9eafa1de 2472static int detail_platform_imsm(int verbose, int enumerate_only, char *controller_path)
d665cc31
DW		 2473{
2474 /* There are two components to imsm platform support, the ahci SATA
2475 * controller and the option-rom. To find the SATA controller we
2476 * simply look in /sys/bus/pci/drivers/ahci to see if an ahci
2477 * controller with the Intel vendor id is present. This approach
2478 * allows mdadm to leverage the kernel's ahci detection logic, with the
2479 * caveat that if ahci.ko is not loaded mdadm will not be able to
2480 * detect platform raid capabilities. The option-rom resides in a
2481 * platform "Adapter ROM". We scan for its signature to retrieve the
2482 * platform capabilities. If raid support is disabled in the BIOS the
2483 * option-rom capability structure will not be available.
2484 */
d665cc31 2485 struct sys_dev *list, *hba;
d665cc31
DW		 2486 int host_base = 0;
2487 int port_count = 0;
9eafa1de 2488 int result=1;
d665cc31 2489
5615172f 2490 if (enumerate_only) {
a891a3c2 2491 if (check_env("IMSM_NO_PLATFORM"))
5615172f 2492 return 0;
a891a3c2
LM		 2493 list = find_intel_devices();
2494 if (!list)
2495 return 2;
2496 for (hba = list; hba; hba = hba->next) {
6b781d33
AP		 2497 if (find_imsm_capability(hba)) {
2498 result = 0;
a891a3c2
LM		 2499 break;
2500 }
9eafa1de 2501 else
6b781d33 2502 result = 2;
a891a3c2 2503 }
a891a3c2 2504 return result;
5615172f
DW		 2505 }
2506
155cbb4c
LM		 2507 list = find_intel_devices();
2508 if (!list) {
ba728be7 2509 if (verbose > 0)
7a862a02 2510 pr_err("no active Intel(R) RAID controller found.\n");
d665cc31 2511 return 2;
ba728be7 2512 } else if (verbose > 0)
155cbb4c 2513 print_found_intel_controllers(list);
d665cc31 2514
a891a3c2 2515 for (hba = list; hba; hba = hba->next) {
0858eccf 2516 if (controller_path && (compare_paths(hba->path, controller_path) != 0))
9eafa1de 2517 continue;
0858eccf 2518 if (!find_imsm_capability(hba)) {
60f0f54d 2519 char buf[PATH_MAX];
e7b84f9d 2520 pr_err("imsm capabilities not found for controller: %s (type %s)\n",
60f0f54d
PB		 2521 hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path,
2522 get_sys_dev_type(hba->type));
0858eccf
AP		 2523 continue;
2524 }
2525 result = 0;
2526 }
2527
2528 if (controller_path && result == 1) {
2529 pr_err("no active Intel(R) RAID controller found under %s\n",
2530 controller_path);
2531 return result;
2532 }
2533
5e1d6128 2534 const struct orom_entry *entry;
0858eccf 2535
5e1d6128 2536 for (entry = orom_entries; entry; entry = entry->next) {
60f0f54d 2537 if (entry->type == SYS_DEV_VMD) {
07cb1e57 2538 print_imsm_capability(&entry->orom);
32716c51
PB		 2539 printf(" 3rd party NVMe :%s supported\n",
2540 imsm_orom_has_tpv_support(&entry->orom)?"":" not");
60f0f54d
PB		 2541 for (hba = list; hba; hba = hba->next) {
2542 if (hba->type == SYS_DEV_VMD) {
2543 char buf[PATH_MAX];
60f0f54d
PB		 2544 printf(" I/O Controller : %s (%s)\n",
2545 vmd_domain_to_controller(hba, buf), get_sys_dev_type(hba->type));
b5eece69
PB		 2546 if (print_vmd_attached_devs(hba)) {
2547 if (verbose > 0)
2548 pr_err("failed to get devices attached to VMD domain.\n");
2549 result |= 2;
2550 }
60f0f54d
PB		 2551 }
2552 }
07cb1e57 2553 printf("\n");
60f0f54d
PB		 2554 continue;
2555 }
0858eccf 2556
60f0f54d
PB		 2557 print_imsm_capability(&entry->orom);
2558 if (entry->type == SYS_DEV_NVME) {
0858eccf
AP		 2559 for (hba = list; hba; hba = hba->next) {
2560 if (hba->type == SYS_DEV_NVME)
2561 printf(" NVMe Device : %s\n", hba->path);
2562 }
60f0f54d 2563 printf("\n");
0858eccf
AP		 2564 continue;
2565 }
2566
2567 struct devid_list *devid;
5e1d6128 2568 for (devid = entry->devid_list; devid; devid = devid->next) {
0858eccf
AP		 2569 hba = device_by_id(devid->devid);
2570 if (!hba)
2571 continue;
2572
9eafa1de
MN		 2573 printf(" I/O Controller : %s (%s)\n",
2574 hba->path, get_sys_dev_type(hba->type));
2575 if (hba->type == SYS_DEV_SATA) {
2576 host_base = ahci_get_port_count(hba->path, &port_count);
2577 if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) {
2578 if (verbose > 0)
7a862a02 2579 pr_err("failed to enumerate ports on SATA controller at %s.\n", hba->pci_id);
9eafa1de
MN		 2580 result |= 2;
2581 }
120dc887
LM		 2582 }
2583 }
0858eccf 2584 printf("\n");
d665cc31 2585 }
155cbb4c 2586
120dc887 2587 return result;
d665cc31 2588}
e50cf220 2589
9eafa1de 2590static int export_detail_platform_imsm(int verbose, char *controller_path)
e50cf220 2591{
e50cf220
MN		 2592 struct sys_dev *list, *hba;
2593 int result=1;
2594
2595 list = find_intel_devices();
2596 if (!list) {
2597 if (verbose > 0)
2598 pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_INTEL_DEVICES\n");
2599 result = 2;
e50cf220
MN		 2600 return result;
2601 }
2602
2603 for (hba = list; hba; hba = hba->next) {
9eafa1de
MN		 2604 if (controller_path && (compare_paths(hba->path,controller_path) != 0))
2605 continue;
60f0f54d
PB		 2606 if (!find_imsm_capability(hba) && verbose > 0) {
2607 char buf[PATH_MAX];
2608 pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_IMSM_CAPABLE_DEVICE_UNDER_%s\n",
2609 hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path);
2610 }
0858eccf 2611 else
e50cf220 2612 result = 0;
e50cf220
MN		 2613 }
2614
5e1d6128 2615 const struct orom_entry *entry;
0858eccf 2616
60f0f54d
PB		 2617 for (entry = orom_entries; entry; entry = entry->next) {
2618 if (entry->type == SYS_DEV_VMD) {
2619 for (hba = list; hba; hba = hba->next)
2620 print_imsm_capability_export(&entry->orom);
2621 continue;
2622 }
5e1d6128 2623 print_imsm_capability_export(&entry->orom);
60f0f54d 2624 }
0858eccf 2625
e50cf220
MN		 2626 return result;
2627}
2628
cdddbdbc
DW		 2629static int match_home_imsm(struct supertype *st, char *homehost)
2630{
5115ca67
DW		 2631 /* the imsm metadata format does not specify any host
2632 * identification information. We return -1 since we can never
2633 * confirm nor deny whether a given array is "meant" for this
148acb7b 2634 * host. We rely on compare_super and the 'family_num' fields to
5115ca67
DW		 2635 * exclude member disks that do not belong, and we rely on
2636 * mdadm.conf to specify the arrays that should be assembled.
2637 * Auto-assembly may still pick up "foreign" arrays.
2638 */
cdddbdbc 2639
9362c1c8 2640 return -1;
cdddbdbc
DW		 2641}
2642
2643static void uuid_from_super_imsm(struct supertype *st, int uuid[4])
2644{
51006d85
N		 2645 /* The uuid returned here is used for:
2646 * uuid to put into bitmap file (Create, Grow)
2647 * uuid for backup header when saving critical section (Grow)
2648 * comparing uuids when re-adding a device into an array
2649 * In these cases the uuid required is that of the data-array,
2650 * not the device-set.
2651 * uuid to recognise same set when adding a missing device back
2652 * to an array. This is a uuid for the device-set.
1011e834 2653 *
51006d85
N		 2654 * For each of these we can make do with a truncated
2655 * or hashed uuid rather than the original, as long as
2656 * everyone agrees.
2657 * In each case the uuid required is that of the data-array,
2658 * not the device-set.
43dad3d6 2659 */
51006d85
N		 2660 /* imsm does not track uuid's so we synthesis one using sha1 on
2661 * - The signature (Which is constant for all imsm array, but no matter)
148acb7b 2662 * - the orig_family_num of the container
51006d85
N		 2663 * - the index number of the volume
2664 * - the 'serial' number of the volume.
2665 * Hopefully these are all constant.
2666 */
2667 struct intel_super *super = st->sb;
43dad3d6 2668
51006d85
N		 2669 char buf[20];
2670 struct sha1_ctx ctx;
2671 struct imsm_dev *dev = NULL;
148acb7b 2672 __u32 family_num;
51006d85 2673
148acb7b
DW		 2674 /* some mdadm versions failed to set ->orig_family_num, in which
2675 * case fall back to ->family_num. orig_family_num will be
2676 * fixed up with the first metadata update.
2677 */
2678 family_num = super->anchor->orig_family_num;
2679 if (family_num == 0)
2680 family_num = super->anchor->family_num;
51006d85 2681 sha1_init_ctx(&ctx);
92bd8f8d 2682 sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx);
148acb7b 2683 sha1_process_bytes(&family_num, sizeof(__u32), &ctx);
51006d85
N		 2684 if (super->current_vol >= 0)
2685 dev = get_imsm_dev(super, super->current_vol);
2686 if (dev) {
2687 __u32 vol = super->current_vol;
2688 sha1_process_bytes(&vol, sizeof(vol), &ctx);
2689 sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx);
2690 }
2691 sha1_finish_ctx(&ctx, buf);
2692 memcpy(uuid, buf, 4*4);
cdddbdbc
DW		 2693}
2694
0d481d37 2695#if 0
4f5bc454
DW		 2696static void
2697get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p)
cdddbdbc 2698{
cdddbdbc
DW		 2699 __u8 *v = get_imsm_version(mpb);
2700 __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH;
2701 char major[] = { 0, 0, 0 };
2702 char minor[] = { 0 ,0, 0 };
2703 char patch[] = { 0, 0, 0 };
2704 char *ver_parse[] = { major, minor, patch };
2705 int i, j;
2706
2707 i = j = 0;
2708 while (*v != '\0' && v < end) {
2709 if (*v != '.' && j < 2)
2710 ver_parse[i][j++] = *v;
2711 else {
2712 i++;
2713 j = 0;
2714 }
2715 v++;
2716 }
2717
4f5bc454
DW		 2718 *m = strtol(minor, NULL, 0);
2719 *p = strtol(patch, NULL, 0);
2720}
0d481d37 2721#endif
4f5bc454 2722
1e5c6983
DW		 2723static __u32 migr_strip_blocks_resync(struct imsm_dev *dev)
2724{
2725 /* migr_strip_size when repairing or initializing parity */
238c0a71 2726 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2727 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2728
2729 switch (get_imsm_raid_level(map)) {
2730 case 5:
2731 case 10:
2732 return chunk;
2733 default:
2734 return 128*1024 >> 9;
2735 }
2736}
2737
2738static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev)
2739{
2740 /* migr_strip_size when rebuilding a degraded disk, no idea why
2741 * this is different than migr_strip_size_resync(), but it's good
2742 * to be compatible
2743 */
238c0a71 2744 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2745 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2746
2747 switch (get_imsm_raid_level(map)) {
2748 case 1:
2749 case 10:
2750 if (map->num_members % map->num_domains == 0)
2751 return 128*1024 >> 9;
2752 else
2753 return chunk;
2754 case 5:
2755 return max((__u32) 64*1024 >> 9, chunk);
2756 default:
2757 return 128*1024 >> 9;
2758 }
2759}
2760
2761static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev)
2762{
238c0a71
AK		 2763 struct imsm_map *lo = get_imsm_map(dev, MAP_0);
2764 struct imsm_map *hi = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2765 __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip);
2766 __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip);
2767
2768 return max((__u32) 1, hi_chunk / lo_chunk);
2769}
2770
2771static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev)
2772{
238c0a71 2773 struct imsm_map *lo = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2774 int level = get_imsm_raid_level(lo);
2775
2776 if (level == 1 || level == 10) {
238c0a71 2777 struct imsm_map *hi = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2778
2779 return hi->num_domains;
2780 } else
2781 return num_stripes_per_unit_resync(dev);
2782}
2783
9529d343 2784static __u8 imsm_num_data_members(struct imsm_map *map)
1e5c6983
DW		 2785{
2786 /* named 'imsm_' because raid0, raid1 and raid10
2787 * counter-intuitively have the same number of data disks
2788 */
1e5c6983
DW		 2789 switch (get_imsm_raid_level(map)) {
2790 case 0:
36fd8ccc
AK		 2791 return map->num_members;
2792 break;
1e5c6983
DW		 2793 case 1:
2794 case 10:
36fd8ccc 2795 return map->num_members/2;
1e5c6983
DW		 2796 case 5:
2797 return map->num_members - 1;
2798 default:
1ade5cc1 2799 dprintf("unsupported raid level\n");
1e5c6983
DW		 2800 return 0;
2801 }
2802}
2803
44490938
MD		 2804static unsigned long long calc_component_size(struct imsm_map *map,
2805 struct imsm_dev *dev)
2806{
2807 unsigned long long component_size;
2808 unsigned long long dev_size = imsm_dev_size(dev);
2809 unsigned long long calc_dev_size = 0;
2810 unsigned int member_disks = imsm_num_data_members(map);
2811
2812 if (member_disks == 0)
2813 return 0;
2814
2815 component_size = per_dev_array_size(map);
2816 calc_dev_size = component_size * member_disks;
2817
2818 /* Component size is rounded to 1MB so difference between size from
2819 * metadata and size calculated from num_data_stripes equals up to
2820 * 2048 blocks per each device. If the difference is higher it means
2821 * that array size was expanded and num_data_stripes was not updated.
2822 */
2823 if ((unsigned int)abs(calc_dev_size - dev_size) >
2824 (1 << SECT_PER_MB_SHIFT) * member_disks) {
2825 component_size = dev_size / member_disks;
2826 dprintf("Invalid num_data_stripes in metadata; expected=%llu, found=%llu\n",
2827 component_size / map->blocks_per_strip,
2828 num_data_stripes(map));
2829 }
2830
2831 return component_size;
2832}
2833
1e5c6983
DW		 2834static __u32 parity_segment_depth(struct imsm_dev *dev)
2835{
238c0a71 2836 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2837 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2838
2839 switch(get_imsm_raid_level(map)) {
2840 case 1:
2841 case 10:
2842 return chunk * map->num_domains;
2843 case 5:
2844 return chunk * map->num_members;
2845 default:
2846 return chunk;
2847 }
2848}
2849
2850static __u32 map_migr_block(struct imsm_dev *dev, __u32 block)
2851{
238c0a71 2852 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2853 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2854 __u32 strip = block / chunk;
2855
2856 switch (get_imsm_raid_level(map)) {
2857 case 1:
2858 case 10: {
2859 __u32 vol_strip = (strip * map->num_domains) + 1;
2860 __u32 vol_stripe = vol_strip / map->num_members;
2861
2862 return vol_stripe * chunk + block % chunk;
2863 } case 5: {
2864 __u32 stripe = strip / (map->num_members - 1);
2865
2866 return stripe * chunk + block % chunk;
2867 }
2868 default:
2869 return 0;
2870 }
2871}
2872
c47b0ff6
AK		 2873static __u64 blocks_per_migr_unit(struct intel_super *super,
2874 struct imsm_dev *dev)
1e5c6983
DW		 2875{
2876 /* calculate the conversion factor between per member 'blocks'
2877 * (md/{resync,rebuild}_start) and imsm migration units, return
2878 * 0 for the 'not migrating' and 'unsupported migration' cases
2879 */
2880 if (!dev->vol.migr_state)
2881 return 0;
2882
2883 switch (migr_type(dev)) {
c47b0ff6
AK		 2884 case MIGR_GEN_MIGR: {
2885 struct migr_record *migr_rec = super->migr_rec;
2886 return __le32_to_cpu(migr_rec->blocks_per_unit);
2887 }
1e5c6983
DW		 2888 case MIGR_VERIFY:
2889 case MIGR_REPAIR:
2890 case MIGR_INIT: {
238c0a71 2891 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2892 __u32 stripes_per_unit;
2893 __u32 blocks_per_unit;
2894 __u32 parity_depth;
2895 __u32 migr_chunk;
2896 __u32 block_map;
2897 __u32 block_rel;
2898 __u32 segment;
2899 __u32 stripe;
2900 __u8 disks;
2901
2902 /* yes, this is really the translation of migr_units to
2903 * per-member blocks in the 'resync' case
2904 */
2905 stripes_per_unit = num_stripes_per_unit_resync(dev);
2906 migr_chunk = migr_strip_blocks_resync(dev);
9529d343 2907 disks = imsm_num_data_members(map);
1e5c6983 2908 blocks_per_unit = stripes_per_unit * migr_chunk * disks;
7b1ab482 2909 stripe = __le16_to_cpu(map->blocks_per_strip) * disks;
1e5c6983
DW		 2910 segment = blocks_per_unit / stripe;
2911 block_rel = blocks_per_unit - segment * stripe;
2912 parity_depth = parity_segment_depth(dev);
2913 block_map = map_migr_block(dev, block_rel);
2914 return block_map + parity_depth * segment;
2915 }
2916 case MIGR_REBUILD: {
2917 __u32 stripes_per_unit;
2918 __u32 migr_chunk;
2919
2920 stripes_per_unit = num_stripes_per_unit_rebuild(dev);
2921 migr_chunk = migr_strip_blocks_rebuild(dev);
2922 return migr_chunk * stripes_per_unit;
2923 }
1e5c6983
DW		 2924 case MIGR_STATE_CHANGE:
2925 default:
2926 return 0;
2927 }
2928}
2929
c2c087e6
DW		 2930static int imsm_level_to_layout(int level)
2931{
2932 switch (level) {
2933 case 0:
2934 case 1:
2935 return 0;
2936 case 5:
2937 case 6:
a380c027 2938 return ALGORITHM_LEFT_ASYMMETRIC;
c2c087e6 2939 case 10:
c92a2527 2940 return 0x102;
c2c087e6 2941 }
a18a888e 2942 return UnSet;
c2c087e6
DW		 2943}
2944
8e59f3d8
AK		 2945/*******************************************************************************
2946 * Function: read_imsm_migr_rec
2947 * Description: Function reads imsm migration record from last sector of disk
2948 * Parameters:
2949 * fd : disk descriptor
2950 * super : metadata info
2951 * Returns:
2952 * 0 : success,
2953 * -1 : fail
2954 ******************************************************************************/
2955static int read_imsm_migr_rec(int fd, struct intel_super *super)
2956{
2957 int ret_val = -1;
de44e46f 2958 unsigned int sector_size = super->sector_size;
8e59f3d8
AK		 2959 unsigned long long dsize;
2960
2961 get_dev_size(fd, NULL, &dsize);
de44e46f
PB		 2962 if (lseek64(fd, dsize - (sector_size*MIGR_REC_SECTOR_POSITION),
2963 SEEK_SET) < 0) {
e7b84f9d
N		 2964 pr_err("Cannot seek to anchor block: %s\n",
2965 strerror(errno));
8e59f3d8
AK		 2966 goto out;
2967 }
466070ad 2968 if ((unsigned int)read(fd, super->migr_rec_buf,
de44e46f
PB		 2969 MIGR_REC_BUF_SECTORS*sector_size) !=
2970 MIGR_REC_BUF_SECTORS*sector_size) {
e7b84f9d
N		 2971 pr_err("Cannot read migr record block: %s\n",
2972 strerror(errno));
8e59f3d8
AK		 2973 goto out;
2974 }
2975 ret_val = 0;
de44e46f
PB		 2976 if (sector_size == 4096)
2977 convert_from_4k_imsm_migr_rec(super);
8e59f3d8
AK		 2978
2979out:
2980 return ret_val;
2981}
2982
3136abe5
AK		 2983static struct imsm_dev *imsm_get_device_during_migration(
2984 struct intel_super *super)
2985{
2986
2987 struct intel_dev *dv;
2988
2989 for (dv = super->devlist; dv; dv = dv->next) {
2990 if (is_gen_migration(dv->dev))
2991 return dv->dev;
2992 }
2993 return NULL;
2994}
2995
8e59f3d8
AK		 2996/*******************************************************************************
2997 * Function: load_imsm_migr_rec
2998 * Description: Function reads imsm migration record (it is stored at the last
2999 * sector of disk)
3000 * Parameters:
3001 * super : imsm internal array info
3002 * info : general array info
3003 * Returns:
3004 * 0 : success
3005 * -1 : fail
4c965cc9 3006 * -2 : no migration in progress
8e59f3d8
AK		 3007 ******************************************************************************/
3008static int load_imsm_migr_rec(struct intel_super *super, struct mdinfo *info)
3009{
3010 struct mdinfo *sd;
594dc1b8 3011 struct dl *dl;
8e59f3d8
AK		 3012 char nm[30];
3013 int retval = -1;
3014 int fd = -1;
3136abe5 3015 struct imsm_dev *dev;
594dc1b8 3016 struct imsm_map *map;
b4ab44d8 3017 int slot = -1;
3136abe5
AK		 3018
3019 /* find map under migration */
3020 dev = imsm_get_device_during_migration(super);
3021 /* nothing to load,no migration in progress?
3022 */
3023 if (dev == NULL)
4c965cc9 3024 return -2;
8e59f3d8
AK		 3025
3026 if (info) {
3027 for (sd = info->devs ; sd ; sd = sd->next) {
3028 /* read only from one of the first two slots */
12fe93e9
TM		 3029 if ((sd->disk.raid_disk < 0) ||
3030 (sd->disk.raid_disk > 1))
8e59f3d8 3031 continue;
3136abe5 3032
8e59f3d8
AK		 3033 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
3034 fd = dev_open(nm, O_RDONLY);
3035 if (fd >= 0)
3036 break;
3037 }
3038 }
3039 if (fd < 0) {
12fe93e9 3040 map = get_imsm_map(dev, MAP_0);
8e59f3d8 3041 for (dl = super->disks; dl; dl = dl->next) {
3136abe5
AK		 3042 /* skip spare and failed disks
3043 */
3044 if (dl->index < 0)
3045 continue;
8e59f3d8 3046 /* read only from one of the first two slots */
3136abe5
AK		 3047 if (map)
3048 slot = get_imsm_disk_slot(map, dl->index);
089f9d79 3049 if (map == NULL || slot > 1 || slot < 0)
8e59f3d8
AK		 3050 continue;
3051 sprintf(nm, "%d:%d", dl->major, dl->minor);
3052 fd = dev_open(nm, O_RDONLY);
3053 if (fd >= 0)
3054 break;
3055 }
3056 }
3057 if (fd < 0)
3058 goto out;
3059 retval = read_imsm_migr_rec(fd, super);
3060
3061out:
3062 if (fd >= 0)
3063 close(fd);
3064 return retval;
3065}
3066
c17608ea
AK		 3067/*******************************************************************************
3068 * function: imsm_create_metadata_checkpoint_update
3069 * Description: It creates update for checkpoint change.
3070 * Parameters:
3071 * super : imsm internal array info
3072 * u : pointer to prepared update
3073 * Returns:
3074 * Uptate length.
3075 * If length is equal to 0, input pointer u contains no update
3076 ******************************************************************************/
3077static int imsm_create_metadata_checkpoint_update(
3078 struct intel_super *super,
3079 struct imsm_update_general_migration_checkpoint **u)
3080{
3081
3082 int update_memory_size = 0;
3083
1ade5cc1 3084 dprintf("(enter)\n");
c17608ea
AK		 3085
3086 if (u == NULL)
3087 return 0;
3088 *u = NULL;
3089
3090 /* size of all update data without anchor */
3091 update_memory_size =
3092 sizeof(struct imsm_update_general_migration_checkpoint);
3093
503975b9 3094 *u = xcalloc(1, update_memory_size);
c17608ea 3095 if (*u == NULL) {
1ade5cc1 3096 dprintf("error: cannot get memory\n");
c17608ea
AK		 3097 return 0;
3098 }
3099 (*u)->type = update_general_migration_checkpoint;
3100 (*u)->curr_migr_unit = __le32_to_cpu(super->migr_rec->curr_migr_unit);
1ade5cc1 3101 dprintf("prepared for %u\n", (*u)->curr_migr_unit);
c17608ea
AK		 3102
3103 return update_memory_size;
3104}
3105
c17608ea
AK		 3106static void imsm_update_metadata_locally(struct supertype *st,
3107 void *buf, int len);
3108
687629c2
AK		 3109/*******************************************************************************
3110 * Function: write_imsm_migr_rec
3111 * Description: Function writes imsm migration record
3112 * (at the last sector of disk)
3113 * Parameters:
3114 * super : imsm internal array info
3115 * Returns:
3116 * 0 : success
3117 * -1 : if fail
3118 ******************************************************************************/
3119static int write_imsm_migr_rec(struct supertype *st)
3120{
3121 struct intel_super *super = st->sb;
de44e46f 3122 unsigned int sector_size = super->sector_size;
687629c2
AK		 3123 unsigned long long dsize;
3124 char nm[30];
3125 int fd = -1;
3126 int retval = -1;
3127 struct dl *sd;
c17608ea
AK		 3128 int len;
3129 struct imsm_update_general_migration_checkpoint *u;
3136abe5 3130 struct imsm_dev *dev;
594dc1b8 3131 struct imsm_map *map;
3136abe5
AK		 3132
3133 /* find map under migration */
3134 dev = imsm_get_device_during_migration(super);
3135 /* if no migration, write buffer anyway to clear migr_record
3136 * on disk based on first available device
3137 */
3138 if (dev == NULL)
3139 dev = get_imsm_dev(super, super->current_vol < 0 ? 0 :
3140 super->current_vol);
3141
44bfe6df 3142 map = get_imsm_map(dev, MAP_0);
687629c2 3143
de44e46f
PB		 3144 if (sector_size == 4096)
3145 convert_to_4k_imsm_migr_rec(super);
687629c2 3146 for (sd = super->disks ; sd ; sd = sd->next) {
b4ab44d8 3147 int slot = -1;
3136abe5
AK		 3148
3149 /* skip failed and spare devices */
3150 if (sd->index < 0)
3151 continue;
687629c2 3152 /* write to 2 first slots only */
3136abe5
AK		 3153 if (map)
3154 slot = get_imsm_disk_slot(map, sd->index);
089f9d79 3155 if (map == NULL || slot > 1 || slot < 0)
687629c2 3156 continue;
3136abe5 3157
687629c2
AK		 3158 sprintf(nm, "%d:%d", sd->major, sd->minor);
3159 fd = dev_open(nm, O_RDWR);
3160 if (fd < 0)
3161 continue;
3162 get_dev_size(fd, NULL, &dsize);
de44e46f
PB		 3163 if (lseek64(fd, dsize - (MIGR_REC_SECTOR_POSITION*sector_size),
3164 SEEK_SET) < 0) {
e7b84f9d
N		 3165 pr_err("Cannot seek to anchor block: %s\n",
3166 strerror(errno));
687629c2
AK		 3167 goto out;
3168 }
466070ad 3169 if ((unsigned int)write(fd, super->migr_rec_buf,
de44e46f
PB		 3170 MIGR_REC_BUF_SECTORS*sector_size) !=
3171 MIGR_REC_BUF_SECTORS*sector_size) {
e7b84f9d
N		 3172 pr_err("Cannot write migr record block: %s\n",
3173 strerror(errno));
687629c2
AK		 3174 goto out;
3175 }
3176 close(fd);
3177 fd = -1;
3178 }
de44e46f
PB		 3179 if (sector_size == 4096)
3180 convert_from_4k_imsm_migr_rec(super);
c17608ea
AK		 3181 /* update checkpoint information in metadata */
3182 len = imsm_create_metadata_checkpoint_update(super, &u);
c17608ea
AK		 3183 if (len <= 0) {
3184 dprintf("imsm: Cannot prepare update\n");
3185 goto out;
3186 }
3187 /* update metadata locally */
3188 imsm_update_metadata_locally(st, u, len);
3189 /* and possibly remotely */
3190 if (st->update_tail) {
3191 append_metadata_update(st, u, len);
3192 /* during reshape we do all work inside metadata handler
3193 * manage_reshape(), so metadata update has to be triggered
3194 * insida it
3195 */
3196 flush_metadata_updates(st);
3197 st->update_tail = &st->updates;
3198 } else
3199 free(u);
687629c2
AK		 3200
3201 retval = 0;
3202 out:
3203 if (fd >= 0)
3204 close(fd);
3205 return retval;
3206}
3207
e2962bfc
AK		 3208/* spare/missing disks activations are not allowe when
3209 * array/container performs reshape operation, because
3210 * all arrays in container works on the same disks set
3211 */
3212int imsm_reshape_blocks_arrays_changes(struct intel_super *super)
3213{
3214 int rv = 0;
3215 struct intel_dev *i_dev;
3216 struct imsm_dev *dev;
3217
3218 /* check whole container
3219 */
3220 for (i_dev = super->devlist; i_dev; i_dev = i_dev->next) {
3221 dev = i_dev->dev;
3ad25638 3222 if (is_gen_migration(dev)) {
e2962bfc
AK		 3223 /* No repair during any migration in container
3224 */
3225 rv = 1;
3226 break;
3227 }
3228 }
3229 return rv;
3230}
3e684231 3231static unsigned long long imsm_component_size_alignment_check(int level,
c41e00b2 3232 int chunk_size,
f36a9ecd 3233 unsigned int sector_size,
c41e00b2
AK		 3234 unsigned long long component_size)
3235{
3e684231 3236 unsigned int component_size_alignment;
c41e00b2 3237
3e684231 3238 /* check component size alignment
c41e00b2 3239 */
3e684231 3240 component_size_alignment = component_size % (chunk_size/sector_size);
c41e00b2 3241
3e684231 3242 dprintf("(Level: %i, chunk_size = %i, component_size = %llu), component_size_alignment = %u\n",
c41e00b2 3243 level, chunk_size, component_size,
3e684231 3244 component_size_alignment);
c41e00b2 3245
3e684231
MZ		 3246 if (component_size_alignment && (level != 1) && (level != UnSet)) {
3247 dprintf("imsm: reported component size aligned from %llu ",
c41e00b2 3248 component_size);
3e684231 3249 component_size -= component_size_alignment;
1ade5cc1 3250 dprintf_cont("to %llu (%i).\n",
3e684231 3251 component_size, component_size_alignment);
c41e00b2
AK		 3252 }
3253
3254 return component_size;
3255}
e2962bfc 3256
2432ce9b
AP		 3257static unsigned long long get_ppl_sector(struct intel_super *super, int dev_idx)
3258{
3259 struct imsm_dev *dev = get_imsm_dev(super, dev_idx);
3260 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3261
3262 return pba_of_lba0(map) +
3263 (num_data_stripes(map) * map->blocks_per_strip);
3264}
3265
a5d85af7 3266static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap)
bf5a934a
DW		 3267{
3268 struct intel_super *super = st->sb;
c47b0ff6 3269 struct migr_record *migr_rec = super->migr_rec;
949c47a0 3270 struct imsm_dev *dev = get_imsm_dev(super, super->current_vol);
238c0a71
AK		 3271 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3272 struct imsm_map *prev_map = get_imsm_map(dev, MAP_1);
b335e593 3273 struct imsm_map *map_to_analyse = map;
efb30e7f 3274 struct dl *dl;
a5d85af7 3275 int map_disks = info->array.raid_disks;
bf5a934a 3276
95eeceeb 3277 memset(info, 0, sizeof(*info));
b335e593
AK		 3278 if (prev_map)
3279 map_to_analyse = prev_map;
3280
ca0748fa 3281 dl = super->current_disk;
9894ec0d 3282
bf5a934a 3283 info->container_member = super->current_vol;
cd0430a1 3284 info->array.raid_disks = map->num_members;
b335e593 3285 info->array.level = get_imsm_raid_level(map_to_analyse);
bf5a934a
DW		 3286 info->array.layout = imsm_level_to_layout(info->array.level);
3287 info->array.md_minor = -1;
3288 info->array.ctime = 0;
3289 info->array.utime = 0;
b335e593
AK		 3290 info->array.chunk_size =
3291 __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9;
2432ce9b 3292 info->array.state = !(dev->vol.dirty & RAIDVOL_DIRTY);
fcc2c9da 3293 info->custom_array_size = imsm_dev_size(dev);
3ad25638
AK		 3294 info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb);
3295
3f510843 3296 if (is_gen_migration(dev)) {
3f83228a 3297 info->reshape_active = 1;
b335e593
AK		 3298 info->new_level = get_imsm_raid_level(map);
3299 info->new_layout = imsm_level_to_layout(info->new_level);
3300 info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9;
3f83228a 3301 info->delta_disks = map->num_members - prev_map->num_members;
493f5dd6
N		 3302 if (info->delta_disks) {
3303 /* this needs to be applied to every array
3304 * in the container.
3305 */
81219e70 3306 info->reshape_active = CONTAINER_RESHAPE;
493f5dd6 3307 }
3f83228a
N		 3308 /* We shape information that we give to md might have to be
3309 * modify to cope with md's requirement for reshaping arrays.
3310 * For example, when reshaping a RAID0, md requires it to be
3311 * presented as a degraded RAID4.
3312 * Also if a RAID0 is migrating to a RAID5 we need to specify
3313 * the array as already being RAID5, but the 'before' layout
3314 * is a RAID4-like layout.
3315 */
3316 switch (info->array.level) {
3317 case 0:
3318 switch(info->new_level) {
3319 case 0:
3320 /* conversion is happening as RAID4 */
3321 info->array.level = 4;
3322 info->array.raid_disks += 1;
3323 break;
3324 case 5:
3325 /* conversion is happening as RAID5 */
3326 info->array.level = 5;
3327 info->array.layout = ALGORITHM_PARITY_N;
3f83228a
N		 3328 info->delta_disks -= 1;
3329 break;
3330 default:
3331 /* FIXME error message */
3332 info->array.level = UnSet;
3333 break;
3334 }
3335 break;
3336 }
b335e593
AK		 3337 } else {
3338 info->new_level = UnSet;
3339 info->new_layout = UnSet;
3340 info->new_chunk = info->array.chunk_size;
3f83228a 3341 info->delta_disks = 0;
b335e593 3342 }
ca0748fa 3343
efb30e7f
DW		 3344 if (dl) {
3345 info->disk.major = dl->major;
3346 info->disk.minor = dl->minor;
ca0748fa 3347 info->disk.number = dl->index;
656b6b5a
N		 3348 info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse,
3349 dl->index);
efb30e7f 3350 }
bf5a934a 3351
5551b113 3352 info->data_offset = pba_of_lba0(map_to_analyse);
44490938 3353 info->component_size = calc_component_size(map, dev);
3e684231 3354 info->component_size = imsm_component_size_alignment_check(
c41e00b2
AK		 3355 info->array.level,
3356 info->array.chunk_size,
f36a9ecd 3357 super->sector_size,
c41e00b2 3358 info->component_size);
5e46202e 3359 info->bb.supported = 1;
139dae11 3360
301406c9 3361 memset(info->uuid, 0, sizeof(info->uuid));
921d9e16 3362 info->recovery_start = MaxSector;
bf5a934a 3363
c2462068
PB		 3364 if (info->array.level == 5 &&
3365 (dev->rwh_policy == RWH_DISTRIBUTED ||
3366 dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)) {
2432ce9b
AP		 3367 info->consistency_policy = CONSISTENCY_POLICY_PPL;
3368 info->ppl_sector = get_ppl_sector(super, super->current_vol);
c2462068
PB		 3369 if (dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
3370 info->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9;
3371 else
3372 info->ppl_size = (PPL_HEADER_SIZE + PPL_ENTRY_SPACE)
3373 >> 9;
2432ce9b
AP		 3374 } else if (info->array.level <= 0) {
3375 info->consistency_policy = CONSISTENCY_POLICY_NONE;
3376 } else {
3377 info->consistency_policy = CONSISTENCY_POLICY_RESYNC;
3378 }
3379
d2e6d5d6 3380 info->reshape_progress = 0;
b6796ce1 3381 info->resync_start = MaxSector;
b9172665 3382 if ((map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED ||
2432ce9b 3383 !(info->array.state & 1)) &&
b9172665 3384 imsm_reshape_blocks_arrays_changes(super) == 0) {
301406c9 3385 info->resync_start = 0;
b6796ce1
AK		 3386 }
3387 if (dev->vol.migr_state) {
1e5c6983
DW		 3388 switch (migr_type(dev)) {
3389 case MIGR_REPAIR:
3390 case MIGR_INIT: {
c47b0ff6
AK		 3391 __u64 blocks_per_unit = blocks_per_migr_unit(super,
3392 dev);
1e5c6983
DW		 3393 __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit);
3394
3395 info->resync_start = blocks_per_unit * units;
3396 break;
3397 }
d2e6d5d6 3398 case MIGR_GEN_MIGR: {
c47b0ff6
AK		 3399 __u64 blocks_per_unit = blocks_per_migr_unit(super,
3400 dev);
3401 __u64 units = __le32_to_cpu(migr_rec->curr_migr_unit);
04fa9523
AK		 3402 unsigned long long array_blocks;
3403 int used_disks;
d2e6d5d6 3404
befb629b
AK		 3405 if (__le32_to_cpu(migr_rec->ascending_migr) &&
3406 (units <
3407 (__le32_to_cpu(migr_rec->num_migr_units)-1)) &&
3408 (super->migr_rec->rec_status ==
3409 __cpu_to_le32(UNIT_SRC_IN_CP_AREA)))
3410 units++;
3411
d2e6d5d6 3412 info->reshape_progress = blocks_per_unit * units;
6289d1e0 3413
7a862a02 3414 dprintf("IMSM: General Migration checkpoint : %llu (%llu) -> read reshape progress : %llu\n",
19986c72
MB		 3415 (unsigned long long)units,
3416 (unsigned long long)blocks_per_unit,
3417 info->reshape_progress);
75156c46 3418
9529d343 3419 used_disks = imsm_num_data_members(prev_map);
75156c46 3420 if (used_disks > 0) {
44490938 3421 array_blocks = per_dev_array_size(map) *
75156c46 3422 used_disks;
b53bfba6
TM		 3423 info->custom_array_size =
3424 round_size_to_mb(array_blocks,
3425 used_disks);
3426
75156c46 3427 }
d2e6d5d6 3428 }
1e5c6983
DW		 3429 case MIGR_VERIFY:
3430 /* we could emulate the checkpointing of
3431 * 'sync_action=check' migrations, but for now
3432 * we just immediately complete them
3433 */
3434 case MIGR_REBUILD:
3435 /* this is handled by container_content_imsm() */
1e5c6983
DW		 3436 case MIGR_STATE_CHANGE:
3437 /* FIXME handle other migrations */
3438 default:
3439 /* we are not dirty, so... */
3440 info->resync_start = MaxSector;
3441 }
b6796ce1 3442 }
301406c9
DW		 3443
3444 strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN);
3445 info->name[MAX_RAID_SERIAL_LEN] = 0;
bf5a934a 3446
f35f2525
N		 3447 info->array.major_version = -1;
3448 info->array.minor_version = -2;
4dd2df09 3449 sprintf(info->text_version, "/%s/%d", st->container_devnm, info->container_member);
a67dd8cc 3450 info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */
51006d85 3451 uuid_from_super_imsm(st, info->uuid);
a5d85af7
N		 3452
3453 if (dmap) {
3454 int i, j;
3455 for (i=0; i<map_disks; i++) {
3456 dmap[i] = 0;
3457 if (i < info->array.raid_disks) {
3458 struct imsm_disk *dsk;
238c0a71 3459 j = get_imsm_disk_idx(dev, i, MAP_X);
a5d85af7
N		 3460 dsk = get_imsm_disk(super, j);
3461 if (dsk && (dsk->status & CONFIGURED_DISK))
3462 dmap[i] = 1;
3463 }
3464 }
3465 }
81ac8b4d 3466}
bf5a934a 3467
3b451610
AK		 3468static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev,
3469 int failed, int look_in_map);
3470
3471static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev,
3472 int look_in_map);
3473
3474static void manage_second_map(struct intel_super *super, struct imsm_dev *dev)
3475{
3476 if (is_gen_migration(dev)) {
3477 int failed;
3478 __u8 map_state;
3479 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
3480
3481 failed = imsm_count_failed(super, dev, MAP_1);
238c0a71 3482 map_state = imsm_check_degraded(super, dev, failed, MAP_1);
3b451610
AK		 3483 if (map2->map_state != map_state) {
3484 map2->map_state = map_state;
3485 super->updates_pending++;
3486 }
3487 }
3488}
97b4d0e9
DW		 3489
3490static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index)
3491{
3492 struct dl *d;
3493
3494 for (d = super->missing; d; d = d->next)
3495 if (d->index == index)
3496 return &d->disk;
3497 return NULL;
3498}
3499
a5d85af7 3500static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map)
4f5bc454
DW		 3501{
3502 struct intel_super *super = st->sb;
4f5bc454 3503 struct imsm_disk *disk;
a5d85af7 3504 int map_disks = info->array.raid_disks;
ab3cb6b3
N		 3505 int max_enough = -1;
3506 int i;
3507 struct imsm_super *mpb;
4f5bc454 3508
bf5a934a 3509 if (super->current_vol >= 0) {
a5d85af7 3510 getinfo_super_imsm_volume(st, info, map);
bf5a934a
DW		 3511 return;
3512 }
95eeceeb 3513 memset(info, 0, sizeof(*info));
d23fe947
DW		 3514
3515 /* Set raid_disks to zero so that Assemble will always pull in valid
3516 * spares
3517 */
3518 info->array.raid_disks = 0;
cdddbdbc
DW		 3519 info->array.level = LEVEL_CONTAINER;
3520 info->array.layout = 0;
3521 info->array.md_minor = -1;
1011e834 3522 info->array.ctime = 0; /* N/A for imsm */
cdddbdbc
DW		 3523 info->array.utime = 0;
3524 info->array.chunk_size = 0;
3525
3526 info->disk.major = 0;
3527 info->disk.minor = 0;
cdddbdbc 3528 info->disk.raid_disk = -1;
c2c087e6 3529 info->reshape_active = 0;
f35f2525
N		 3530 info->array.major_version = -1;
3531 info->array.minor_version = -2;
c2c087e6 3532 strcpy(info->text_version, "imsm");
a67dd8cc 3533 info->safe_mode_delay = 0;
c2c087e6
DW		 3534 info->disk.number = -1;
3535 info->disk.state = 0;
c5afc314 3536 info->name[0] = 0;
921d9e16 3537 info->recovery_start = MaxSector;
3ad25638 3538 info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb);
5e46202e 3539 info->bb.supported = 1;
c2c087e6 3540
97b4d0e9 3541 /* do we have the all the insync disks that we expect? */
ab3cb6b3 3542 mpb = super->anchor;
b7d81a38 3543 info->events = __le32_to_cpu(mpb->generation_num);
97b4d0e9 3544
ab3cb6b3
N		 3545 for (i = 0; i < mpb->num_raid_devs; i++) {
3546 struct imsm_dev *dev = get_imsm_dev(super, i);
3547 int failed, enough, j, missing = 0;
3548 struct imsm_map *map;
3549 __u8 state;
97b4d0e9 3550
3b451610
AK		 3551 failed = imsm_count_failed(super, dev, MAP_0);
3552 state = imsm_check_degraded(super, dev, failed, MAP_0);
238c0a71 3553 map = get_imsm_map(dev, MAP_0);
ab3cb6b3
N		 3554
3555 /* any newly missing disks?
3556 * (catches single-degraded vs double-degraded)
3557 */
3558 for (j = 0; j < map->num_members; j++) {
238c0a71 3559 __u32 ord = get_imsm_ord_tbl_ent(dev, j, MAP_0);
ab3cb6b3
N		 3560 __u32 idx = ord_to_idx(ord);
3561
20dc76d1
MT		 3562 if (super->disks && super->disks->index == (int)idx)
3563 info->disk.raid_disk = j;
3564
ab3cb6b3
N		 3565 if (!(ord & IMSM_ORD_REBUILD) &&
3566 get_imsm_missing(super, idx)) {
3567 missing = 1;
3568 break;
3569 }
97b4d0e9 3570 }
ab3cb6b3
N		 3571
3572 if (state == IMSM_T_STATE_FAILED)
3573 enough = -1;
3574 else if (state == IMSM_T_STATE_DEGRADED &&
3575 (state != map->map_state || missing))
3576 enough = 0;
3577 else /* we're normal, or already degraded */
3578 enough = 1;
d2bde6d3
AK		 3579 if (is_gen_migration(dev) && missing) {
3580 /* during general migration we need all disks
3581 * that process is running on.
3582 * No new missing disk is allowed.
3583 */
3584 max_enough = -1;
3585 enough = -1;
3586 /* no more checks necessary
3587 */
3588 break;
3589 }
ab3cb6b3
N		 3590 /* in the missing/failed disk case check to see
3591 * if at least one array is runnable
3592 */
3593 max_enough = max(max_enough, enough);
3594 }
1ade5cc1 3595 dprintf("enough: %d\n", max_enough);
ab3cb6b3 3596 info->container_enough = max_enough;
97b4d0e9 3597
4a04ec6c 3598 if (super->disks) {
14e8215b
DW		 3599 __u32 reserved = imsm_reserved_sectors(super, super->disks);
3600
b9f594fe 3601 disk = &super->disks->disk;
5551b113 3602 info->data_offset = total_blocks(&super->disks->disk) - reserved;
14e8215b 3603 info->component_size = reserved;
25ed7e59 3604 info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0;
df474657
DW		 3605 /* we don't change info->disk.raid_disk here because
3606 * this state will be finalized in mdmon after we have
3607 * found the 'most fresh' version of the metadata
3608 */
25ed7e59 3609 info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0;
2432ce9b
AP		 3610 info->disk.state |= (is_spare(disk) || is_journal(disk)) ?
3611 0 : (1 << MD_DISK_SYNC);
cdddbdbc 3612 }
a575e2a7
DW		 3613
3614 /* only call uuid_from_super_imsm when this disk is part of a populated container,
3615 * ->compare_super may have updated the 'num_raid_devs' field for spares
3616 */
3617 if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs)
36ba7d48 3618 uuid_from_super_imsm(st, info->uuid);
22e263f6
AC		 3619 else
3620 memcpy(info->uuid, uuid_zero, sizeof(uuid_zero));
a5d85af7
N		 3621
3622 /* I don't know how to compute 'map' on imsm, so use safe default */
3623 if (map) {
3624 int i;
3625 for (i = 0; i < map_disks; i++)
3626 map[i] = 1;
3627 }
3628
cdddbdbc
DW		 3629}
3630
5c4cd5da
AC		 3631/* allocates memory and fills disk in mdinfo structure
3632 * for each disk in array */
3633struct mdinfo *getinfo_super_disks_imsm(struct supertype *st)
3634{
594dc1b8 3635 struct mdinfo *mddev;
5c4cd5da
AC		 3636 struct intel_super *super = st->sb;
3637 struct imsm_disk *disk;
3638 int count = 0;
3639 struct dl *dl;
3640 if (!super || !super->disks)
3641 return NULL;
3642 dl = super->disks;
503975b9 3643 mddev = xcalloc(1, sizeof(*mddev));
5c4cd5da
AC		 3644 while (dl) {
3645 struct mdinfo *tmp;
3646 disk = &dl->disk;
503975b9 3647 tmp = xcalloc(1, sizeof(*tmp));
5c4cd5da
AC		 3648 if (mddev->devs)
3649 tmp->next = mddev->devs;
3650 mddev->devs = tmp;
3651 tmp->disk.number = count++;
3652 tmp->disk.major = dl->major;
3653 tmp->disk.minor = dl->minor;
3654 tmp->disk.state = is_configured(disk) ?
3655 (1 << MD_DISK_ACTIVE) : 0;
3656 tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0;
3657 tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC);
3658 tmp->disk.raid_disk = -1;
3659 dl = dl->next;
3660 }
3661 return mddev;
3662}
3663
cdddbdbc
DW		 3664static int update_super_imsm(struct supertype *st, struct mdinfo *info,
3665 char *update, char *devname, int verbose,
3666 int uuid_set, char *homehost)
3667{
f352c545
DW		 3668 /* For 'assemble' and 'force' we need to return non-zero if any
3669 * change was made. For others, the return value is ignored.
3670 * Update options are:
3671 * force-one : This device looks a bit old but needs to be included,
3672 * update age info appropriately.
3673 * assemble: clear any 'faulty' flag to allow this device to
3674 * be assembled.
3675 * force-array: Array is degraded but being forced, mark it clean
3676 * if that will be needed to assemble it.
3677 *
3678 * newdev: not used ????
3679 * grow: Array has gained a new device - this is currently for
3680 * linear only
3681 * resync: mark as dirty so a resync will happen.
3682 * name: update the name - preserving the homehost
6e46bf34 3683 * uuid: Change the uuid of the array to match watch is given
f352c545
DW		 3684 *
3685 * Following are not relevant for this imsm:
3686 * sparc2.2 : update from old dodgey metadata
3687 * super-minor: change the preferred_minor number
3688 * summaries: update redundant counters.
f352c545
DW		 3689 * homehost: update the recorded homehost
3690 * _reshape_progress: record new reshape_progress position.
3691 */
6e46bf34
DW		 3692 int rv = 1;
3693 struct intel_super *super = st->sb;
3694 struct imsm_super *mpb;
f352c545 3695
6e46bf34
DW		 3696 /* we can only update container info */
3697 if (!super || super->current_vol >= 0 || !super->anchor)
3698 return 1;
3699
3700 mpb = super->anchor;
3701
81a5b4f5
N		 3702 if (strcmp(update, "uuid") == 0) {
3703 /* We take this to mean that the family_num should be updated.
3704 * However that is much smaller than the uuid so we cannot really
3705 * allow an explicit uuid to be given. And it is hard to reliably
3706 * know if one was.
3707 * So if !uuid_set we know the current uuid is random and just used
3708 * the first 'int' and copy it to the other 3 positions.
3709 * Otherwise we require the 4 'int's to be the same as would be the
3710 * case if we are using a random uuid. So an explicit uuid will be
3711 * accepted as long as all for ints are the same... which shouldn't hurt
6e46bf34 3712 */
81a5b4f5
N		 3713 if (!uuid_set) {
3714 info->uuid[1] = info->uuid[2] = info->uuid[3] = info->uuid[0];
6e46bf34 3715 rv = 0;
81a5b4f5
N		 3716 } else {
3717 if (info->uuid[0] != info->uuid[1] ||
3718 info->uuid[1] != info->uuid[2] ||
3719 info->uuid[2] != info->uuid[3])
3720 rv = -1;
3721 else
3722 rv = 0;
6e46bf34 3723 }
81a5b4f5
N		 3724 if (rv == 0)
3725 mpb->orig_family_num = info->uuid[0];
6e46bf34
DW		 3726 } else if (strcmp(update, "assemble") == 0)
3727 rv = 0;
3728 else
1e2b2765 3729 rv = -1;
f352c545 3730
6e46bf34
DW		 3731 /* successful update? recompute checksum */
3732 if (rv == 0)
3733 mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb));
f352c545
DW		 3734
3735 return rv;
cdddbdbc
DW		 3736}
3737
c2c087e6 3738static size_t disks_to_mpb_size(int disks)
cdddbdbc 3739{
c2c087e6 3740 size_t size;
cdddbdbc 3741
c2c087e6
DW		 3742 size = sizeof(struct imsm_super);
3743 size += (disks - 1) * sizeof(struct imsm_disk);
3744 size += 2 * sizeof(struct imsm_dev);
3745 /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */
3746 size += (4 - 2) * sizeof(struct imsm_map);
3747 /* 4 possible disk_ord_tbl's */
3748 size += 4 * (disks - 1) * sizeof(__u32);
bbab0940
TM		 3749 /* maximum bbm log */
3750 size += sizeof(struct bbm_log);
c2c087e6
DW		 3751
3752 return size;
3753}
3754
387fcd59
N		 3755static __u64 avail_size_imsm(struct supertype *st, __u64 devsize,
3756 unsigned long long data_offset)
c2c087e6
DW		 3757{
3758 if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS))
3759 return 0;
3760
3761 return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS);
cdddbdbc
DW		 3762}
3763
ba2de7ba
DW		 3764static void free_devlist(struct intel_super *super)
3765{
3766 struct intel_dev *dv;
3767
3768 while (super->devlist) {
3769 dv = super->devlist->next;
3770 free(super->devlist->dev);
3771 free(super->devlist);
3772 super->devlist = dv;
3773 }
3774}
3775
3776static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src)
3777{
3778 memcpy(dest, src, sizeof_imsm_dev(src, 0));
3779}
3780
cdddbdbc
DW		 3781static int compare_super_imsm(struct supertype *st, struct supertype *tst)
3782{
3783 /*
3784 * return:
3785 * 0 same, or first was empty, and second was copied
3786 * 1 second had wrong number
3787 * 2 wrong uuid
3788 * 3 wrong other info
3789 */
3790 struct intel_super *first = st->sb;
3791 struct intel_super *sec = tst->sb;
3792
5d500228
N		 3793 if (!first) {
3794 st->sb = tst->sb;
3795 tst->sb = NULL;
3796 return 0;
3797 }
8603ea6f
LM		 3798 /* in platform dependent environment test if the disks
3799 * use the same Intel hba
cb8f6859 3800 * If not on Intel hba at all, allow anything.
8603ea6f 3801 */
6b781d33
AP		 3802 if (!check_env("IMSM_NO_PLATFORM") && first->hba && sec->hba) {
3803 if (first->hba->type != sec->hba->type) {
8603ea6f 3804 fprintf(stderr,
6b781d33
AP		 3805 "HBAs of devices do not match %s != %s\n",
3806 get_sys_dev_type(first->hba->type),
3807 get_sys_dev_type(sec->hba->type));
3808 return 3;
3809 }
3810 if (first->orom != sec->orom) {
3811 fprintf(stderr,
3812 "HBAs of devices do not match %s != %s\n",
3813 first->hba->pci_id, sec->hba->pci_id);
8603ea6f
LM		 3814 return 3;
3815 }
3816 }
cdddbdbc 3817
d23fe947
DW		 3818 /* if an anchor does not have num_raid_devs set then it is a free
3819 * floating spare
3820 */
3821 if (first->anchor->num_raid_devs > 0 &&
3822 sec->anchor->num_raid_devs > 0) {
a2b97981
DW		 3823 /* Determine if these disks might ever have been
3824 * related. Further disambiguation can only take place
3825 * in load_super_imsm_all
3826 */
3827 __u32 first_family = first->anchor->orig_family_num;
3828 __u32 sec_family = sec->anchor->orig_family_num;
3829
f796af5d
DW		 3830 if (memcmp(first->anchor->sig, sec->anchor->sig,
3831 MAX_SIGNATURE_LENGTH) != 0)
3832 return 3;
3833
a2b97981
DW		 3834 if (first_family == 0)
3835 first_family = first->anchor->family_num;
3836 if (sec_family == 0)
3837 sec_family = sec->anchor->family_num;
3838
3839 if (first_family != sec_family)
d23fe947 3840 return 3;
f796af5d 3841
d23fe947 3842 }
cdddbdbc 3843
3e372e5a
DW		 3844 /* if 'first' is a spare promote it to a populated mpb with sec's
3845 * family number
3846 */
3847 if (first->anchor->num_raid_devs == 0 &&
3848 sec->anchor->num_raid_devs > 0) {
78d30f94 3849 int i;
ba2de7ba
DW		 3850 struct intel_dev *dv;
3851 struct imsm_dev *dev;
78d30f94
DW		 3852
3853 /* we need to copy raid device info from sec if an allocation
3854 * fails here we don't associate the spare
3855 */
3856 for (i = 0; i < sec->anchor->num_raid_devs; i++) {
503975b9
N		 3857 dv = xmalloc(sizeof(*dv));
3858 dev = xmalloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1));
ba2de7ba
DW		 3859 dv->dev = dev;
3860 dv->index = i;
3861 dv->next = first->devlist;
3862 first->devlist = dv;
78d30f94 3863 }
709743c5 3864 if (i < sec->anchor->num_raid_devs) {
ba2de7ba
DW		 3865 /* allocation failure */
3866 free_devlist(first);
e12b3daa 3867 pr_err("imsm: failed to associate spare\n");
ba2de7ba 3868 return 3;
78d30f94 3869 }
3e372e5a 3870 first->anchor->num_raid_devs = sec->anchor->num_raid_devs;
148acb7b 3871 first->anchor->orig_family_num = sec->anchor->orig_family_num;
3e372e5a 3872 first->anchor->family_num = sec->anchor->family_num;
ac6449be 3873 memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH);
709743c5
DW		 3874 for (i = 0; i < sec->anchor->num_raid_devs; i++)
3875 imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i));
3e372e5a
DW		 3876 }
3877
cdddbdbc
DW		 3878 return 0;
3879}
3880
0030e8d6
DW		 3881static void fd2devname(int fd, char *name)
3882{
3883 struct stat st;
3884 char path[256];
33a6535d 3885 char dname[PATH_MAX];
0030e8d6
DW		 3886 char *nm;
3887 int rv;
3888
3889 name[0] = '\0';
3890 if (fstat(fd, &st) != 0)
3891 return;
3892 sprintf(path, "/sys/dev/block/%d:%d",
3893 major(st.st_rdev), minor(st.st_rdev));
3894
9cf014ec 3895 rv = readlink(path, dname, sizeof(dname)-1);
0030e8d6
DW		 3896 if (rv <= 0)
3897 return;
9587c373 3898
0030e8d6
DW		 3899 dname[rv] = '\0';
3900 nm = strrchr(dname, '/');
7897de29
JS		 3901 if (nm) {
3902 nm++;
3903 snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm);
3904 }
0030e8d6
DW		 3905}
3906
21e9380b
AP		 3907static int nvme_get_serial(int fd, void *buf, size_t buf_len)
3908{
3909 char path[60];
3910 char *name = fd2kname(fd);
3911
3912 if (!name)
3913 return 1;
3914
3915 if (strncmp(name, "nvme", 4) != 0)
3916 return 1;
3917
3918 snprintf(path, sizeof(path) - 1, "/sys/block/%s/device/serial", name);
3919
3920 return load_sys(path, buf, buf_len);
3921}
3922
cdddbdbc
DW		 3923extern int scsi_get_serial(int fd, void *buf, size_t buf_len);
3924
3925static int imsm_read_serial(int fd, char *devname,
3926 __u8 serial[MAX_RAID_SERIAL_LEN])
3927{
21e9380b 3928 char buf[50];
cdddbdbc 3929 int rv;
1f24f035 3930 int len;
316e2bf4
DW		 3931 char *dest;
3932 char *src;
21e9380b
AP		 3933 unsigned int i;
3934
3935 memset(buf, 0, sizeof(buf));
cdddbdbc 3936
21e9380b 3937 rv = nvme_get_serial(fd, buf, sizeof(buf));
cdddbdbc 3938
21e9380b
AP		 3939 if (rv)
3940 rv = scsi_get_serial(fd, buf, sizeof(buf));
f9ba0ff1 3941
40ebbb9c 3942 if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) {
f9ba0ff1
DW		 3943 memset(serial, 0, MAX_RAID_SERIAL_LEN);
3944 fd2devname(fd, (char *) serial);
0030e8d6
DW		 3945 return 0;
3946 }
3947
cdddbdbc
DW		 3948 if (rv != 0) {
3949 if (devname)
e7b84f9d
N		 3950 pr_err("Failed to retrieve serial for %s\n",
3951 devname);
cdddbdbc
DW		 3952 return rv;
3953 }
3954
316e2bf4
DW		 3955 /* trim all whitespace and non-printable characters and convert
3956 * ':' to ';'
3957 */
21e9380b
AP		 3958 for (i = 0, dest = buf; i < sizeof(buf) && buf[i]; i++) {
3959 src = &buf[i];
316e2bf4
DW		 3960 if (*src > 0x20) {
3961 /* ':' is reserved for use in placeholder serial
3962 * numbers for missing disks
3963 */
3964 if (*src == ':')
3965 *dest++ = ';';
3966 else
3967 *dest++ = *src;
3968 }
3969 }
21e9380b
AP		 3970 len = dest - buf;
3971 dest = buf;
316e2bf4
DW		 3972
3973 /* truncate leading characters */
3974 if (len > MAX_RAID_SERIAL_LEN) {
3975 dest += len - MAX_RAID_SERIAL_LEN;
1f24f035 3976 len = MAX_RAID_SERIAL_LEN;
316e2bf4 3977 }
5c3db629 3978
5c3db629 3979 memset(serial, 0, MAX_RAID_SERIAL_LEN);
316e2bf4 3980 memcpy(serial, dest, len);
cdddbdbc
DW		 3981
3982 return 0;
3983}
3984
1f24f035
DW		 3985static int serialcmp(__u8 *s1, __u8 *s2)
3986{
3987 return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN);
3988}
3989
3990static void serialcpy(__u8 *dest, __u8 *src)
3991{
3992 strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN);
3993}
3994
54c2c1ea
DW		 3995static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super)
3996{
3997 struct dl *dl;
3998
3999 for (dl = super->disks; dl; dl = dl->next)
4000 if (serialcmp(dl->serial, serial) == 0)
4001 break;
4002
4003 return dl;
4004}
4005
a2b97981
DW		 4006static struct imsm_disk *
4007__serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx)
4008{
4009 int i;
4010
4011 for (i = 0; i < mpb->num_disks; i++) {
4012 struct imsm_disk *disk = __get_imsm_disk(mpb, i);
4013
4014 if (serialcmp(disk->serial, serial) == 0) {
4015 if (idx)
4016 *idx = i;
4017 return disk;
4018 }
4019 }
4020
4021 return NULL;
4022}
4023
cdddbdbc
DW		 4024static int
4025load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd)
4026{
a2b97981 4027 struct imsm_disk *disk;
cdddbdbc
DW		 4028 struct dl *dl;
4029 struct stat stb;
cdddbdbc 4030 int rv;
a2b97981 4031 char name[40];
d23fe947
DW		 4032 __u8 serial[MAX_RAID_SERIAL_LEN];
4033
4034 rv = imsm_read_serial(fd, devname, serial);
4035
4036 if (rv != 0)
4037 return 2;
4038
503975b9 4039 dl = xcalloc(1, sizeof(*dl));
cdddbdbc 4040
a2b97981
DW		 4041 fstat(fd, &stb);
4042 dl->major = major(stb.st_rdev);
4043 dl->minor = minor(stb.st_rdev);
4044 dl->next = super->disks;
4045 dl->fd = keep_fd ? fd : -1;
4046 assert(super->disks == NULL);
4047 super->disks = dl;
4048 serialcpy(dl->serial, serial);
4049 dl->index = -2;
4050 dl->e = NULL;
4051 fd2devname(fd, name);
4052 if (devname)
503975b9 4053 dl->devname = xstrdup(devname);
a2b97981 4054 else
503975b9 4055 dl->devname = xstrdup(name);
cdddbdbc 4056
d23fe947 4057 /* look up this disk's index in the current anchor */
a2b97981
DW		 4058 disk = __serial_to_disk(dl->serial, super->anchor, &dl->index);
4059 if (disk) {
4060 dl->disk = *disk;
4061 /* only set index on disks that are a member of a
4062 * populated contianer, i.e. one with raid_devs
4063 */
4064 if (is_failed(&dl->disk))
3f6efecc 4065 dl->index = -2;
2432ce9b 4066 else if (is_spare(&dl->disk) || is_journal(&dl->disk))
a2b97981 4067 dl->index = -1;
3f6efecc
DW		 4068 }
4069
949c47a0
DW		 4070 return 0;
4071}
4072
0c046afd
DW		 4073/* When migrating map0 contains the 'destination' state while map1
4074 * contains the current state. When not migrating map0 contains the
4075 * current state. This routine assumes that map[0].map_state is set to
4076 * the current array state before being called.
4077 *
4078 * Migration is indicated by one of the following states
4079 * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed)
e3bba0e0 4080 * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal
0c046afd 4081 * map1state=unitialized)
1484e727 4082 * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal
0c046afd 4083 * map1state=normal)
e3bba0e0 4084 * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal
0c046afd 4085 * map1state=degraded)
8e59f3d8
AK		 4086 * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal
4087 * map1state=normal)
0c046afd 4088 */
8e59f3d8
AK		 4089static void migrate(struct imsm_dev *dev, struct intel_super *super,
4090 __u8 to_state, int migr_type)
3393c6af 4091{
0c046afd 4092 struct imsm_map *dest;
238c0a71 4093 struct imsm_map *src = get_imsm_map(dev, MAP_0);
3393c6af 4094
0c046afd 4095 dev->vol.migr_state = 1;
1484e727 4096 set_migr_type(dev, migr_type);
f8f603f1 4097 dev->vol.curr_migr_unit = 0;
238c0a71 4098 dest = get_imsm_map(dev, MAP_1);
0c046afd 4099
0556e1a2 4100 /* duplicate and then set the target end state in map[0] */
3393c6af 4101 memcpy(dest, src, sizeof_imsm_map(src));
fb12a745 4102 if (migr_type == MIGR_GEN_MIGR) {
0556e1a2
DW		 4103 __u32 ord;
4104 int i;
4105
4106 for (i = 0; i < src->num_members; i++) {
4107 ord = __le32_to_cpu(src->disk_ord_tbl[i]);
4108 set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord));
4109 }
4110 }
4111
8e59f3d8
AK		 4112 if (migr_type == MIGR_GEN_MIGR)
4113 /* Clear migration record */
4114 memset(super->migr_rec, 0, sizeof(struct migr_record));
4115
0c046afd 4116 src->map_state = to_state;
949c47a0 4117}
f8f603f1 4118
809da78e
AK		 4119static void end_migration(struct imsm_dev *dev, struct intel_super *super,
4120 __u8 map_state)
f8f603f1 4121{
238c0a71
AK		 4122 struct imsm_map *map = get_imsm_map(dev, MAP_0);
4123 struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state == 0 ?
4124 MAP_0 : MAP_1);
28bce06f 4125 int i, j;
0556e1a2
DW		 4126
4127 /* merge any IMSM_ORD_REBUILD bits that were not successfully
4128 * completed in the last migration.
4129 *
28bce06f 4130 * FIXME add support for raid-level-migration
0556e1a2 4131 */
089f9d79
JS		 4132 if (map_state != map->map_state && (is_gen_migration(dev) == 0) &&
4133 prev->map_state != IMSM_T_STATE_UNINITIALIZED) {
809da78e
AK		 4134 /* when final map state is other than expected
4135 * merge maps (not for migration)
4136 */
4137 int failed;
4138
4139 for (i = 0; i < prev->num_members; i++)
4140 for (j = 0; j < map->num_members; j++)
4141 /* during online capacity expansion
4142 * disks position can be changed
4143 * if takeover is used
4144 */
4145 if (ord_to_idx(map->disk_ord_tbl[j]) ==
4146 ord_to_idx(prev->disk_ord_tbl[i])) {
4147 map->disk_ord_tbl[j] |=
4148 prev->disk_ord_tbl[i];
4149 break;
4150 }
4151 failed = imsm_count_failed(super, dev, MAP_0);
4152 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
4153 }
f8f603f1
DW		 4154
4155 dev->vol.migr_state = 0;
ea672ee1 4156 set_migr_type(dev, 0);
f8f603f1
DW		 4157 dev->vol.curr_migr_unit = 0;
4158 map->map_state = map_state;
4159}
949c47a0
DW		 4160
4161static int parse_raid_devices(struct intel_super *super)
4162{
4163 int i;
4164 struct imsm_dev *dev_new;
4d7b1503 4165 size_t len, len_migr;
401d313b 4166 size_t max_len = 0;
4d7b1503
DW		 4167 size_t space_needed = 0;
4168 struct imsm_super *mpb = super->anchor;
949c47a0
DW		 4169
4170 for (i = 0; i < super->anchor->num_raid_devs; i++) {
4171 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
ba2de7ba 4172 struct intel_dev *dv;
949c47a0 4173
4d7b1503
DW		 4174 len = sizeof_imsm_dev(dev_iter, 0);
4175 len_migr = sizeof_imsm_dev(dev_iter, 1);
4176 if (len_migr > len)
4177 space_needed += len_migr - len;
ca9de185 4178
503975b9 4179 dv = xmalloc(sizeof(*dv));
401d313b
AK		 4180 if (max_len < len_migr)
4181 max_len = len_migr;
4182 if (max_len > len_migr)
4183 space_needed += max_len - len_migr;
503975b9 4184 dev_new = xmalloc(max_len);
949c47a0 4185 imsm_copy_dev(dev_new, dev_iter);
ba2de7ba
DW		 4186 dv->dev = dev_new;
4187 dv->index = i;
4188 dv->next = super->devlist;
4189 super->devlist = dv;
949c47a0 4190 }
cdddbdbc 4191
4d7b1503
DW		 4192 /* ensure that super->buf is large enough when all raid devices
4193 * are migrating
4194 */
4195 if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) {
4196 void *buf;
4197
f36a9ecd
PB		 4198 len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed,
4199 super->sector_size);
4200 if (posix_memalign(&buf, MAX_SECTOR_SIZE, len) != 0)
4d7b1503
DW		 4201 return 1;
4202
1f45a8ad
DW		 4203 memcpy(buf, super->buf, super->len);
4204 memset(buf + super->len, 0, len - super->len);
4d7b1503
DW		 4205 free(super->buf);
4206 super->buf = buf;
4207 super->len = len;
4208 }
ca9de185 4209
bbab0940
TM		 4210 super->extra_space += space_needed;
4211
cdddbdbc
DW		 4212 return 0;
4213}
4214
e2f41b2c
AK		 4215/*******************************************************************************
4216 * Function: check_mpb_migr_compatibility
4217 * Description: Function checks for unsupported migration features:
4218 * - migration optimization area (pba_of_lba0)
4219 * - descending reshape (ascending_migr)
4220 * Parameters:
4221 * super : imsm metadata information
4222 * Returns:
4223 * 0 : migration is compatible
4224 * -1 : migration is not compatible
4225 ******************************************************************************/
4226int check_mpb_migr_compatibility(struct intel_super *super)
4227{
4228 struct imsm_map *map0, *map1;
4229 struct migr_record *migr_rec = super->migr_rec;
4230 int i;
4231
4232 for (i = 0; i < super->anchor->num_raid_devs; i++) {
4233 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
4234
4235 if (dev_iter &&
4236 dev_iter->vol.migr_state == 1 &&
4237 dev_iter->vol.migr_type == MIGR_GEN_MIGR) {
4238 /* This device is migrating */
238c0a71
AK		 4239 map0 = get_imsm_map(dev_iter, MAP_0);
4240 map1 = get_imsm_map(dev_iter, MAP_1);
5551b113 4241 if (pba_of_lba0(map0) != pba_of_lba0(map1))
e2f41b2c
AK		 4242 /* migration optimization area was used */
4243 return -1;
fc54fe7a
JS		 4244 if (migr_rec->ascending_migr == 0 &&
4245 migr_rec->dest_depth_per_unit > 0)
e2f41b2c
AK		 4246 /* descending reshape not supported yet */
4247 return -1;
4248 }
4249 }
4250 return 0;
4251}
4252
d23fe947 4253static void __free_imsm(struct intel_super *super, int free_disks);
9ca2c81c 4254
cdddbdbc 4255/* load_imsm_mpb - read matrix metadata
f2f5c343 4256 * allocates super->mpb to be freed by free_imsm
cdddbdbc
DW		 4257 */
4258static int load_imsm_mpb(int fd, struct intel_super *super, char *devname)
4259{
4260 unsigned long long dsize;
cdddbdbc 4261 unsigned long long sectors;
f36a9ecd 4262 unsigned int sector_size = super->sector_size;
cdddbdbc 4263 struct stat;
6416d527 4264 struct imsm_super *anchor;
cdddbdbc
DW		 4265 __u32 check_sum;
4266
cdddbdbc 4267 get_dev_size(fd, NULL, &dsize);
f36a9ecd 4268 if (dsize < 2*sector_size) {
64436f06 4269 if (devname)
e7b84f9d
N		 4270 pr_err("%s: device to small for imsm\n",
4271 devname);
64436f06
N		 4272 return 1;
4273 }
cdddbdbc 4274
f36a9ecd 4275 if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0) {
cdddbdbc 4276 if (devname)
e7b84f9d
N		 4277 pr_err("Cannot seek to anchor block on %s: %s\n",
4278 devname, strerror(errno));
cdddbdbc
DW		 4279 return 1;
4280 }
4281
f36a9ecd 4282 if (posix_memalign((void **)&anchor, sector_size, sector_size) != 0) {
ad97895e 4283 if (devname)
7a862a02 4284 pr_err("Failed to allocate imsm anchor buffer on %s\n", devname);
ad97895e
DW		 4285 return 1;
4286 }
466070ad 4287 if ((unsigned int)read(fd, anchor, sector_size) != sector_size) {
cdddbdbc 4288 if (devname)
e7b84f9d
N		 4289 pr_err("Cannot read anchor block on %s: %s\n",
4290 devname, strerror(errno));
6416d527 4291 free(anchor);
cdddbdbc
DW		 4292 return 1;
4293 }
4294
6416d527 4295 if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) {
cdddbdbc 4296 if (devname)
e7b84f9d 4297 pr_err("no IMSM anchor on %s\n", devname);
6416d527 4298 free(anchor);
cdddbdbc
DW		 4299 return 2;
4300 }
4301
d23fe947 4302 __free_imsm(super, 0);
f2f5c343
LM		 4303 /* reload capability and hba */
4304
4305 /* capability and hba must be updated with new super allocation */
d424212e 4306 find_intel_hba_capability(fd, super, devname);
f36a9ecd
PB		 4307 super->len = ROUND_UP(anchor->mpb_size, sector_size);
4308 if (posix_memalign(&super->buf, MAX_SECTOR_SIZE, super->len) != 0) {
cdddbdbc 4309 if (devname)
e7b84f9d
N		 4310 pr_err("unable to allocate %zu byte mpb buffer\n",
4311 super->len);
6416d527 4312 free(anchor);
cdddbdbc
DW		 4313 return 2;
4314 }
f36a9ecd 4315 memcpy(super->buf, anchor, sector_size);
cdddbdbc 4316
f36a9ecd 4317 sectors = mpb_sectors(anchor, sector_size) - 1;
6416d527 4318 free(anchor);
8e59f3d8 4319
85337573
AO		 4320 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
4321 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) {
1ade5cc1 4322 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8
AK		 4323 free(super->buf);
4324 return 2;
4325 }
51d83f5d 4326 super->clean_migration_record_by_mdmon = 0;
8e59f3d8 4327
949c47a0 4328 if (!sectors) {
ecf45690
DW		 4329 check_sum = __gen_imsm_checksum(super->anchor);
4330 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
4331 if (devname)
e7b84f9d
N		 4332 pr_err("IMSM checksum %x != %x on %s\n",
4333 check_sum,
4334 __le32_to_cpu(super->anchor->check_sum),
4335 devname);
ecf45690
DW		 4336 return 2;
4337 }
4338
a2b97981 4339 return 0;
949c47a0 4340 }
cdddbdbc
DW		 4341
4342 /* read the extended mpb */
f36a9ecd 4343 if (lseek64(fd, dsize - (sector_size * (2 + sectors)), SEEK_SET) < 0) {
cdddbdbc 4344 if (devname)
e7b84f9d
N		 4345 pr_err("Cannot seek to extended mpb on %s: %s\n",
4346 devname, strerror(errno));
cdddbdbc
DW		 4347 return 1;
4348 }
4349
f36a9ecd
PB		 4350 if ((unsigned int)read(fd, super->buf + sector_size,
4351 super->len - sector_size) != super->len - sector_size) {
cdddbdbc 4352 if (devname)
e7b84f9d
N		 4353 pr_err("Cannot read extended mpb on %s: %s\n",
4354 devname, strerror(errno));
cdddbdbc
DW		 4355 return 2;
4356 }
4357
949c47a0
DW		 4358 check_sum = __gen_imsm_checksum(super->anchor);
4359 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
cdddbdbc 4360 if (devname)
e7b84f9d
N		 4361 pr_err("IMSM checksum %x != %x on %s\n",
4362 check_sum, __le32_to_cpu(super->anchor->check_sum),
4363 devname);
db575f3b 4364 return 3;
cdddbdbc
DW		 4365 }
4366
a2b97981
DW		 4367 return 0;
4368}
4369
8e59f3d8
AK		 4370static int read_imsm_migr_rec(int fd, struct intel_super *super);
4371
97f81ee2
CA		 4372/* clears hi bits in metadata if MPB_ATTRIB_2TB_DISK not set */
4373static void clear_hi(struct intel_super *super)
4374{
4375 struct imsm_super *mpb = super->anchor;
4376 int i, n;
4377 if (mpb->attributes & MPB_ATTRIB_2TB_DISK)
4378 return;
4379 for (i = 0; i < mpb->num_disks; ++i) {
4380 struct imsm_disk *disk = &mpb->disk[i];
4381 disk->total_blocks_hi = 0;
4382 }
4383 for (i = 0; i < mpb->num_raid_devs; ++i) {
4384 struct imsm_dev *dev = get_imsm_dev(super, i);
4385 if (!dev)
4386 return;
4387 for (n = 0; n < 2; ++n) {
4388 struct imsm_map *map = get_imsm_map(dev, n);
4389 if (!map)
4390 continue;
4391 map->pba_of_lba0_hi = 0;
4392 map->blocks_per_member_hi = 0;
4393 map->num_data_stripes_hi = 0;
4394 }
4395 }
4396}
4397
a2b97981
DW		 4398static int
4399load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd)
4400{
4401 int err;
4402
4403 err = load_imsm_mpb(fd, super, devname);
4404 if (err)
4405 return err;
f36a9ecd
PB		 4406 if (super->sector_size == 4096)
4407 convert_from_4k(super);
a2b97981
DW		 4408 err = load_imsm_disk(fd, super, devname, keep_fd);
4409 if (err)
4410 return err;
4411 err = parse_raid_devices(super);
8d67477f
TM		 4412 if (err)
4413 return err;
4414 err = load_bbm_log(super);
97f81ee2 4415 clear_hi(super);
a2b97981 4416 return err;
cdddbdbc
DW		 4417}
4418
ae6aad82
DW		 4419static void __free_imsm_disk(struct dl *d)
4420{
4421 if (d->fd >= 0)
4422 close(d->fd);
4423 if (d->devname)
4424 free(d->devname);
0dcecb2e
DW		 4425 if (d->e)
4426 free(d->e);
ae6aad82
DW		 4427 free(d);
4428
4429}
1a64be56 4430
cdddbdbc
DW		 4431static void free_imsm_disks(struct intel_super *super)
4432{
47ee5a45 4433 struct dl *d;
cdddbdbc 4434
47ee5a45
DW		 4435 while (super->disks) {
4436 d = super->disks;
cdddbdbc 4437 super->disks = d->next;
ae6aad82 4438 __free_imsm_disk(d);
cdddbdbc 4439 }
cb82edca
AK		 4440 while (super->disk_mgmt_list) {
4441 d = super->disk_mgmt_list;
4442 super->disk_mgmt_list = d->next;
4443 __free_imsm_disk(d);
4444 }
47ee5a45
DW		 4445 while (super->missing) {
4446 d = super->missing;
4447 super->missing = d->next;
4448 __free_imsm_disk(d);
4449 }
4450
cdddbdbc
DW		 4451}
4452
9ca2c81c 4453/* free all the pieces hanging off of a super pointer */
d23fe947 4454static void __free_imsm(struct intel_super *super, int free_disks)
cdddbdbc 4455{
88654014
LM		 4456 struct intel_hba *elem, *next;
4457
9ca2c81c 4458 if (super->buf) {
949c47a0 4459 free(super->buf);
9ca2c81c
DW		 4460 super->buf = NULL;
4461 }
f2f5c343
LM		 4462 /* unlink capability description */
4463 super->orom = NULL;
8e59f3d8
AK		 4464 if (super->migr_rec_buf) {
4465 free(super->migr_rec_buf);
4466 super->migr_rec_buf = NULL;
4467 }
d23fe947
DW		 4468 if (free_disks)
4469 free_imsm_disks(super);
ba2de7ba 4470 free_devlist(super);
88654014
LM		 4471 elem = super->hba;
4472 while (elem) {
4473 if (elem->path)
4474 free((void *)elem->path);
4475 next = elem->next;
4476 free(elem);
4477 elem = next;
88c32bb1 4478 }
8d67477f
TM		 4479 if (super->bbm_log)
4480 free(super->bbm_log);
88654014 4481 super->hba = NULL;
cdddbdbc
DW		 4482}
4483
9ca2c81c
DW		 4484static void free_imsm(struct intel_super *super)
4485{
d23fe947 4486 __free_imsm(super, 1);
928f1424 4487 free(super->bb.entries);
9ca2c81c
DW		 4488 free(super);
4489}
cdddbdbc
DW		 4490
4491static void free_super_imsm(struct supertype *st)
4492{
4493 struct intel_super *super = st->sb;
4494
4495 if (!super)
4496 return;
4497
4498 free_imsm(super);
4499 st->sb = NULL;
4500}
4501
49133e57 4502static struct intel_super *alloc_super(void)
c2c087e6 4503{
503975b9 4504 struct intel_super *super = xcalloc(1, sizeof(*super));
c2c087e6 4505
503975b9
N		 4506 super->current_vol = -1;
4507 super->create_offset = ~((unsigned long long) 0);
928f1424
TM		 4508
4509 super->bb.entries = xmalloc(BBM_LOG_MAX_ENTRIES *
4510 sizeof(struct md_bb_entry));
4511 if (!super->bb.entries) {
4512 free(super);
4513 return NULL;
4514 }
4515
c2c087e6
DW		 4516 return super;
4517}
4518
f0f5a016
LM		 4519/*
4520 * find and allocate hba and OROM/EFI based on valid fd of RAID component device
4521 */
d424212e 4522static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname)
f0f5a016
LM		 4523{
4524 struct sys_dev *hba_name;
4525 int rv = 0;
4526
089f9d79 4527 if (fd < 0 || check_env("IMSM_NO_PLATFORM")) {
f2f5c343 4528 super->orom = NULL;
f0f5a016
LM		 4529 super->hba = NULL;
4530 return 0;
4531 }
4532 hba_name = find_disk_attached_hba(fd, NULL);
4533 if (!hba_name) {
d424212e 4534 if (devname)
e7b84f9d
N		 4535 pr_err("%s is not attached to Intel(R) RAID controller.\n",
4536 devname);
f0f5a016
LM		 4537 return 1;
4538 }
4539 rv = attach_hba_to_super(super, hba_name);
4540 if (rv == 2) {
d424212e
N		 4541 if (devname) {
4542 struct intel_hba *hba = super->hba;
f0f5a016 4543
60f0f54d
PB		 4544 pr_err("%s is attached to Intel(R) %s %s (%s),\n"
4545 " but the container is assigned to Intel(R) %s %s (",
d424212e 4546 devname,
614902f6 4547 get_sys_dev_type(hba_name->type),
60f0f54d 4548 hba_name->type == SYS_DEV_VMD ? "domain" : "RAID controller",
f0f5a016 4549 hba_name->pci_id ? : "Err!",
60f0f54d
PB		 4550 get_sys_dev_type(super->hba->type),
4551 hba->type == SYS_DEV_VMD ? "domain" : "RAID controller");
f0f5a016 4552
f0f5a016
LM		 4553 while (hba) {
4554 fprintf(stderr, "%s", hba->pci_id ? : "Err!");
4555 if (hba->next)
4556 fprintf(stderr, ", ");
4557 hba = hba->next;
4558 }
6b781d33 4559 fprintf(stderr, ").\n"
cca67208 4560 " Mixing devices attached to different controllers is not allowed.\n");
f0f5a016 4561 }
f0f5a016
LM		 4562 return 2;
4563 }
6b781d33 4564 super->orom = find_imsm_capability(hba_name);
f2f5c343
LM		 4565 if (!super->orom)
4566 return 3;
614902f6 4567
f0f5a016
LM		 4568 return 0;
4569}
4570
47ee5a45
DW		 4571/* find_missing - helper routine for load_super_imsm_all that identifies
4572 * disks that have disappeared from the system. This routine relies on
4573 * the mpb being uptodate, which it is at load time.
4574 */
4575static int find_missing(struct intel_super *super)
4576{
4577 int i;
4578 struct imsm_super *mpb = super->anchor;
4579 struct dl *dl;
4580 struct imsm_disk *disk;
47ee5a45
DW		 4581
4582 for (i = 0; i < mpb->num_disks; i++) {
4583 disk = __get_imsm_disk(mpb, i);
54c2c1ea 4584 dl = serial_to_dl(disk->serial, super);
47ee5a45
DW		 4585 if (dl)
4586 continue;
47ee5a45 4587
503975b9 4588 dl = xmalloc(sizeof(*dl));
47ee5a45
DW		 4589 dl->major = 0;
4590 dl->minor = 0;
4591 dl->fd = -1;
503975b9 4592 dl->devname = xstrdup("missing");
47ee5a45
DW		 4593 dl->index = i;
4594 serialcpy(dl->serial, disk->serial);
4595 dl->disk = *disk;
689c9bf3 4596 dl->e = NULL;
47ee5a45
DW		 4597 dl->next = super->missing;
4598 super->missing = dl;
4599 }
4600
4601 return 0;
4602}
4603
a2b97981
DW		 4604static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list)
4605{
4606 struct intel_disk *idisk = disk_list;
4607
4608 while (idisk) {
4609 if (serialcmp(idisk->disk.serial, serial) == 0)
4610 break;
4611 idisk = idisk->next;
4612 }
4613
4614 return idisk;
4615}
4616
4617static int __prep_thunderdome(struct intel_super **table, int tbl_size,
4618 struct intel_super *super,
4619 struct intel_disk **disk_list)
4620{
4621 struct imsm_disk *d = &super->disks->disk;
4622 struct imsm_super *mpb = super->anchor;
4623 int i, j;
4624
4625 for (i = 0; i < tbl_size; i++) {
4626 struct imsm_super *tbl_mpb = table[i]->anchor;
4627 struct imsm_disk *tbl_d = &table[i]->disks->disk;
4628
4629 if (tbl_mpb->family_num == mpb->family_num) {
4630 if (tbl_mpb->check_sum == mpb->check_sum) {
1ade5cc1
N		 4631 dprintf("mpb from %d:%d matches %d:%d\n",
4632 super->disks->major,
a2b97981
DW		 4633 super->disks->minor,
4634 table[i]->disks->major,
4635 table[i]->disks->minor);
4636 break;
4637 }
4638
4639 if (((is_configured(d) && !is_configured(tbl_d)) ||
4640 is_configured(d) == is_configured(tbl_d)) &&
4641 tbl_mpb->generation_num < mpb->generation_num) {
4642 /* current version of the mpb is a
4643 * better candidate than the one in
4644 * super_table, but copy over "cross
4645 * generational" status
4646 */
4647 struct intel_disk *idisk;
4648
1ade5cc1
N		 4649 dprintf("mpb from %d:%d replaces %d:%d\n",
4650 super->disks->major,
a2b97981
DW		 4651 super->disks->minor,
4652 table[i]->disks->major,
4653 table[i]->disks->minor);
4654
4655 idisk = disk_list_get(tbl_d->serial, *disk_list);
4656 if (idisk && is_failed(&idisk->disk))
4657 tbl_d->status |= FAILED_DISK;
4658 break;
4659 } else {
4660 struct intel_disk *idisk;
4661 struct imsm_disk *disk;
4662
4663 /* tbl_mpb is more up to date, but copy
4664 * over cross generational status before
4665 * returning
4666 */
4667 disk = __serial_to_disk(d->serial, mpb, NULL);
4668 if (disk && is_failed(disk))
4669 d->status |= FAILED_DISK;
4670
4671 idisk = disk_list_get(d->serial, *disk_list);
4672 if (idisk) {
4673 idisk->owner = i;
4674 if (disk && is_configured(disk))
4675 idisk->disk.status |= CONFIGURED_DISK;
4676 }
4677
1ade5cc1
N		 4678 dprintf("mpb from %d:%d prefer %d:%d\n",
4679 super->disks->major,
a2b97981
DW		 4680 super->disks->minor,
4681 table[i]->disks->major,
4682 table[i]->disks->minor);
4683
4684 return tbl_size;
4685 }
4686 }
4687 }
4688
4689 if (i >= tbl_size)
4690 table[tbl_size++] = super;
4691 else
4692 table[i] = super;
4693
4694 /* update/extend the merged list of imsm_disk records */
4695 for (j = 0; j < mpb->num_disks; j++) {
4696 struct imsm_disk *disk = __get_imsm_disk(mpb, j);
4697 struct intel_disk *idisk;
4698
4699 idisk = disk_list_get(disk->serial, *disk_list);
4700 if (idisk) {
4701 idisk->disk.status |= disk->status;
4702 if (is_configured(&idisk->disk) ||
4703 is_failed(&idisk->disk))
4704 idisk->disk.status &= ~(SPARE_DISK);
4705 } else {
503975b9 4706 idisk = xcalloc(1, sizeof(*idisk));
a2b97981
DW		 4707 idisk->owner = IMSM_UNKNOWN_OWNER;
4708 idisk->disk = *disk;
4709 idisk->next = *disk_list;
4710 *disk_list = idisk;
4711 }
4712
4713 if (serialcmp(idisk->disk.serial, d->serial) == 0)
4714 idisk->owner = i;
4715 }
4716
4717 return tbl_size;
4718}
4719
4720static struct intel_super *
4721validate_members(struct intel_super *super, struct intel_disk *disk_list,
4722 const int owner)
4723{
4724 struct imsm_super *mpb = super->anchor;
4725 int ok_count = 0;
4726 int i;
4727
4728 for (i = 0; i < mpb->num_disks; i++) {
4729 struct imsm_disk *disk = __get_imsm_disk(mpb, i);
4730 struct intel_disk *idisk;
4731
4732 idisk = disk_list_get(disk->serial, disk_list);
4733 if (idisk) {
4734 if (idisk->owner == owner ||
4735 idisk->owner == IMSM_UNKNOWN_OWNER)
4736 ok_count++;
4737 else
1ade5cc1
N		 4738 dprintf("'%.16s' owner %d != %d\n",
4739 disk->serial, idisk->owner,
a2b97981
DW		 4740 owner);
4741 } else {
1ade5cc1
N		 4742 dprintf("unknown disk %x [%d]: %.16s\n",
4743 __le32_to_cpu(mpb->family_num), i,
a2b97981
DW		 4744 disk->serial);
4745 break;
4746 }
4747 }
4748
4749 if (ok_count == mpb->num_disks)
4750 return super;
4751 return NULL;
4752}
4753
4754static void show_conflicts(__u32 family_num, struct intel_super *super_list)
4755{
4756 struct intel_super *s;
4757
4758 for (s = super_list; s; s = s->next) {
4759 if (family_num != s->anchor->family_num)
4760 continue;
e12b3daa 4761 pr_err("Conflict, offlining family %#x on '%s'\n",
a2b97981
DW		 4762 __le32_to_cpu(family_num), s->disks->devname);
4763 }
4764}
4765
4766static struct intel_super *
4767imsm_thunderdome(struct intel_super **super_list, int len)
4768{
4769 struct intel_super *super_table[len];
4770 struct intel_disk *disk_list = NULL;
4771 struct intel_super *champion, *spare;
4772 struct intel_super *s, **del;
4773 int tbl_size = 0;
4774 int conflict;
4775 int i;
4776
4777 memset(super_table, 0, sizeof(super_table));
4778 for (s = *super_list; s; s = s->next)
4779 tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list);
4780
4781 for (i = 0; i < tbl_size; i++) {
4782 struct imsm_disk *d;
4783 struct intel_disk *idisk;
4784 struct imsm_super *mpb = super_table[i]->anchor;
4785
4786 s = super_table[i];
4787 d = &s->disks->disk;
4788
4789 /* 'd' must appear in merged disk list for its
4790 * configuration to be valid
4791 */
4792 idisk = disk_list_get(d->serial, disk_list);
4793 if (idisk && idisk->owner == i)
4794 s = validate_members(s, disk_list, i);
4795 else
4796 s = NULL;
4797
4798 if (!s)
1ade5cc1
N		 4799 dprintf("marking family: %#x from %d:%d offline\n",
4800 mpb->family_num,
a2b97981
DW		 4801 super_table[i]->disks->major,
4802 super_table[i]->disks->minor);
4803 super_table[i] = s;
4804 }
4805
4806 /* This is where the mdadm implementation differs from the Windows
4807 * driver which has no strict concept of a container. We can only
4808 * assemble one family from a container, so when returning a prodigal
4809 * array member to this system the code will not be able to disambiguate
4810 * the container contents that should be assembled ("foreign" versus
4811 * "local"). It requires user intervention to set the orig_family_num
4812 * to a new value to establish a new container. The Windows driver in
4813 * this situation fixes up the volume name in place and manages the
4814 * foreign array as an independent entity.
4815 */
4816 s = NULL;
4817 spare = NULL;
4818 conflict = 0;
4819 for (i = 0; i < tbl_size; i++) {
4820 struct intel_super *tbl_ent = super_table[i];
4821 int is_spare = 0;
4822
4823 if (!tbl_ent)
4824 continue;
4825
4826 if (tbl_ent->anchor->num_raid_devs == 0) {
4827 spare = tbl_ent;
4828 is_spare = 1;
4829 }
4830
4831 if (s && !is_spare) {
4832 show_conflicts(tbl_ent->anchor->family_num, *super_list);
4833 conflict++;
4834 } else if (!s && !is_spare)
4835 s = tbl_ent;
4836 }
4837
4838 if (!s)
4839 s = spare;
4840 if (!s) {
4841 champion = NULL;
4842 goto out;
4843 }
4844 champion = s;
4845
4846 if (conflict)
7a862a02 4847 pr_err("Chose family %#x on '%s', assemble conflicts to new container with '--update=uuid'\n",
a2b97981
DW		 4848 __le32_to_cpu(s->anchor->family_num), s->disks->devname);
4849
4850 /* collect all dl's onto 'champion', and update them to
4851 * champion's version of the status
4852 */
4853 for (s = *super_list; s; s = s->next) {
4854 struct imsm_super *mpb = champion->anchor;
4855 struct dl *dl = s->disks;
4856
4857 if (s == champion)
4858 continue;
4859
5d7b407a
CA		 4860 mpb->attributes |= s->anchor->attributes & MPB_ATTRIB_2TB_DISK;
4861
a2b97981
DW		 4862 for (i = 0; i < mpb->num_disks; i++) {
4863 struct imsm_disk *disk;
4864
4865 disk = __serial_to_disk(dl->serial, mpb, &dl->index);
4866 if (disk) {
4867 dl->disk = *disk;
4868 /* only set index on disks that are a member of
4869 * a populated contianer, i.e. one with
4870 * raid_devs
4871 */
4872 if (is_failed(&dl->disk))
4873 dl->index = -2;
4874 else if (is_spare(&dl->disk))
4875 dl->index = -1;
4876 break;
4877 }
4878 }
4879
4880 if (i >= mpb->num_disks) {
4881 struct intel_disk *idisk;
4882
4883 idisk = disk_list_get(dl->serial, disk_list);
ecf408e9 4884 if (idisk && is_spare(&idisk->disk) &&
a2b97981
DW		 4885 !is_failed(&idisk->disk) && !is_configured(&idisk->disk))
4886 dl->index = -1;
4887 else {
4888 dl->index = -2;
4889 continue;
4890 }
4891 }
4892
4893 dl->next = champion->disks;
4894 champion->disks = dl;
4895 s->disks = NULL;
4896 }
4897
4898 /* delete 'champion' from super_list */
4899 for (del = super_list; *del; ) {
4900 if (*del == champion) {
4901 *del = (*del)->next;
4902 break;
4903 } else
4904 del = &(*del)->next;
4905 }
4906 champion->next = NULL;
4907
4908 out:
4909 while (disk_list) {
4910 struct intel_disk *idisk = disk_list;
4911
4912 disk_list = disk_list->next;
4913 free(idisk);
4914 }
4915
4916 return champion;
4917}
4918
9587c373
LM		 4919static int
4920get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd);
4dd2df09 4921static int get_super_block(struct intel_super **super_list, char *devnm, char *devname,
9587c373 4922 int major, int minor, int keep_fd);
ec50f7b6
LM		 4923static int
4924get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list,
4925 int *max, int keep_fd);
4926
cdddbdbc 4927static int load_super_imsm_all(struct supertype *st, int fd, void **sbp,
ec50f7b6
LM		 4928 char *devname, struct md_list *devlist,
4929 int keep_fd)
cdddbdbc 4930{
a2b97981
DW		 4931 struct intel_super *super_list = NULL;
4932 struct intel_super *super = NULL;
a2b97981 4933 int err = 0;
9587c373 4934 int i = 0;
dab4a513 4935
9587c373
LM		 4936 if (fd >= 0)
4937 /* 'fd' is an opened container */
4938 err = get_sra_super_block(fd, &super_list, devname, &i, keep_fd);
4939 else
ec50f7b6
LM		 4940 /* get super block from devlist devices */
4941 err = get_devlist_super_block(devlist, &super_list, &i, keep_fd);
9587c373 4942 if (err)
1602d52c 4943 goto error;
a2b97981
DW		 4944 /* all mpbs enter, maybe one leaves */
4945 super = imsm_thunderdome(&super_list, i);
4946 if (!super) {
4947 err = 1;
4948 goto error;
cdddbdbc
DW		 4949 }
4950
47ee5a45
DW		 4951 if (find_missing(super) != 0) {
4952 free_imsm(super);
a2b97981
DW		 4953 err = 2;
4954 goto error;
47ee5a45 4955 }
8e59f3d8
AK		 4956
4957 /* load migration record */
4958 err = load_imsm_migr_rec(super, NULL);
4c965cc9
AK		 4959 if (err == -1) {
4960 /* migration is in progress,
4961 * but migr_rec cannot be loaded,
4962 */
8e59f3d8
AK		 4963 err = 4;
4964 goto error;
4965 }
e2f41b2c
AK		 4966
4967 /* Check migration compatibility */
089f9d79 4968 if (err == 0 && check_mpb_migr_compatibility(super) != 0) {
e7b84f9d 4969 pr_err("Unsupported migration detected");
e2f41b2c
AK		 4970 if (devname)
4971 fprintf(stderr, " on %s\n", devname);
4972 else
4973 fprintf(stderr, " (IMSM).\n");
4974
4975 err = 5;
4976 goto error;
4977 }
4978
a2b97981
DW		 4979 err = 0;
4980
4981 error:
4982 while (super_list) {
4983 struct intel_super *s = super_list;
4984
4985 super_list = super_list->next;
4986 free_imsm(s);
4987 }
9587c373 4988
a2b97981
DW		 4989 if (err)
4990 return err;
f7e7067b 4991
cdddbdbc 4992 *sbp = super;
9587c373 4993 if (fd >= 0)
4dd2df09 4994 strcpy(st->container_devnm, fd2devnm(fd));
9587c373 4995 else
4dd2df09 4996 st->container_devnm[0] = 0;
a2b97981 4997 if (err == 0 && st->ss == NULL) {
bf5a934a 4998 st->ss = &super_imsm;
cdddbdbc
DW		 4999 st->minor_version = 0;
5000 st->max_devs = IMSM_MAX_DEVICES;
5001 }
cdddbdbc
DW		 5002 return 0;
5003}
2b959fbf 5004
ec50f7b6
LM		 5005static int
5006get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list,
5007 int *max, int keep_fd)
5008{
5009 struct md_list *tmpdev;
5010 int err = 0;
5011 int i = 0;
9587c373 5012
ec50f7b6
LM		 5013 for (i = 0, tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
5014 if (tmpdev->used != 1)
5015 continue;
5016 if (tmpdev->container == 1) {
ca9de185 5017 int lmax = 0;
ec50f7b6
LM		 5018 int fd = dev_open(tmpdev->devname, O_RDONLY|O_EXCL);
5019 if (fd < 0) {
e7b84f9d 5020 pr_err("cannot open device %s: %s\n",
ec50f7b6
LM		 5021 tmpdev->devname, strerror(errno));
5022 err = 8;
5023 goto error;
5024 }
5025 err = get_sra_super_block(fd, super_list,
5026 tmpdev->devname, &lmax,
5027 keep_fd);
5028 i += lmax;
5029 close(fd);
5030 if (err) {
5031 err = 7;
5032 goto error;
5033 }
5034 } else {
5035 int major = major(tmpdev->st_rdev);
5036 int minor = minor(tmpdev->st_rdev);
5037 err = get_super_block(super_list,
4dd2df09 5038 NULL,
ec50f7b6
LM		 5039 tmpdev->devname,
5040 major, minor,
5041 keep_fd);
5042 i++;
5043 if (err) {
5044 err = 6;
5045 goto error;
5046 }
5047 }
5048 }
5049 error:
5050 *max = i;
5051 return err;
5052}
9587c373 5053
4dd2df09 5054static int get_super_block(struct intel_super **super_list, char *devnm, char *devname,
9587c373
LM		 5055 int major, int minor, int keep_fd)
5056{
594dc1b8 5057 struct intel_super *s;
9587c373
LM		 5058 char nm[32];
5059 int dfd = -1;
9587c373
LM		 5060 int err = 0;
5061 int retry;
5062
5063 s = alloc_super();
5064 if (!s) {
5065 err = 1;
5066 goto error;
5067 }
5068
5069 sprintf(nm, "%d:%d", major, minor);
5070 dfd = dev_open(nm, O_RDWR);
5071 if (dfd < 0) {
5072 err = 2;
5073 goto error;
5074 }
5075
fa7bb6f8 5076 get_dev_sector_size(dfd, NULL, &s->sector_size);
cb8f6859 5077 find_intel_hba_capability(dfd, s, devname);
9587c373
LM		 5078 err = load_and_parse_mpb(dfd, s, NULL, keep_fd);
5079
5080 /* retry the load if we might have raced against mdmon */
4dd2df09 5081 if (err == 3 && devnm && mdmon_running(devnm))
9587c373
LM		 5082 for (retry = 0; retry < 3; retry++) {
5083 usleep(3000);
5084 err = load_and_parse_mpb(dfd, s, NULL, keep_fd);
5085 if (err != 3)
5086 break;
5087 }
5088 error:
5089 if (!err) {
5090 s->next = *super_list;
5091 *super_list = s;
5092 } else {
5093 if (s)
8d67477f 5094 free_imsm(s);
36614e95 5095 if (dfd >= 0)
9587c373
LM		 5096 close(dfd);
5097 }
089f9d79 5098 if (dfd >= 0 && !keep_fd)
9587c373
LM		 5099 close(dfd);
5100 return err;
5101
5102}
5103
5104static int
5105get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd)
5106{
5107 struct mdinfo *sra;
4dd2df09 5108 char *devnm;
9587c373
LM		 5109 struct mdinfo *sd;
5110 int err = 0;
5111 int i = 0;
4dd2df09 5112 sra = sysfs_read(fd, NULL, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE);
9587c373
LM		 5113 if (!sra)
5114 return 1;
5115
5116 if (sra->array.major_version != -1 ||
5117 sra->array.minor_version != -2 ||
5118 strcmp(sra->text_version, "imsm") != 0) {
5119 err = 1;
5120 goto error;
5121 }
5122 /* load all mpbs */
4dd2df09 5123 devnm = fd2devnm(fd);
9587c373 5124 for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) {
4dd2df09 5125 if (get_super_block(super_list, devnm, devname,
9587c373
LM		 5126 sd->disk.major, sd->disk.minor, keep_fd) != 0) {
5127 err = 7;
5128 goto error;
5129 }
5130 }
5131 error:
5132 sysfs_free(sra);
5133 *max = i;
5134 return err;
5135}
5136
2b959fbf
N		 5137static int load_container_imsm(struct supertype *st, int fd, char *devname)
5138{
ec50f7b6 5139 return load_super_imsm_all(st, fd, &st->sb, devname, NULL, 1);
2b959fbf 5140}
cdddbdbc
DW		 5141
5142static int load_super_imsm(struct supertype *st, int fd, char *devname)
5143{
5144 struct intel_super *super;
5145 int rv;
8a3544f8 5146 int retry;
cdddbdbc 5147
357ac106 5148 if (test_partition(fd))
691c6ee1
N		 5149 /* IMSM not allowed on partitions */
5150 return 1;
5151
37424f13
DW		 5152 free_super_imsm(st);
5153
49133e57 5154 super = alloc_super();
fa7bb6f8 5155 get_dev_sector_size(fd, NULL, &super->sector_size);
8d67477f
TM		 5156 if (!super)
5157 return 1;
ea2bc72b
LM		 5158 /* Load hba and capabilities if they exist.
5159 * But do not preclude loading metadata in case capabilities or hba are
5160 * non-compliant and ignore_hw_compat is set.
5161 */
d424212e 5162 rv = find_intel_hba_capability(fd, super, devname);
f2f5c343 5163 /* no orom/efi or non-intel hba of the disk */
089f9d79 5164 if (rv != 0 && st->ignore_hw_compat == 0) {
f2f5c343 5165 if (devname)
e7b84f9d 5166 pr_err("No OROM/EFI properties for %s\n", devname);
f2f5c343
LM		 5167 free_imsm(super);
5168 return 2;
5169 }
a2b97981 5170 rv = load_and_parse_mpb(fd, super, devname, 0);
cdddbdbc 5171
8a3544f8
AP		 5172 /* retry the load if we might have raced against mdmon */
5173 if (rv == 3) {
f96b1302
AP		 5174 struct mdstat_ent *mdstat = NULL;
5175 char *name = fd2kname(fd);
5176
5177 if (name)
5178 mdstat = mdstat_by_component(name);
8a3544f8
AP		 5179
5180 if (mdstat && mdmon_running(mdstat->devnm) && getpid() != mdmon_pid(mdstat->devnm)) {
5181 for (retry = 0; retry < 3; retry++) {
5182 usleep(3000);
5183 rv = load_and_parse_mpb(fd, super, devname, 0);
5184 if (rv != 3)
5185 break;
5186 }
5187 }
5188
5189 free_mdstat(mdstat);
5190 }
5191
cdddbdbc
DW		 5192 if (rv) {
5193 if (devname)
7a862a02 5194 pr_err("Failed to load all information sections on %s\n", devname);
cdddbdbc
DW		 5195 free_imsm(super);
5196 return rv;
5197 }
5198
5199 st->sb = super;
5200 if (st->ss == NULL) {
5201 st->ss = &super_imsm;
5202 st->minor_version = 0;
5203 st->max_devs = IMSM_MAX_DEVICES;
5204 }
8e59f3d8
AK		 5205
5206 /* load migration record */
2e062e82
AK		 5207 if (load_imsm_migr_rec(super, NULL) == 0) {
5208 /* Check for unsupported migration features */
5209 if (check_mpb_migr_compatibility(super) != 0) {
e7b84f9d 5210 pr_err("Unsupported migration detected");
2e062e82
AK		 5211 if (devname)
5212 fprintf(stderr, " on %s\n", devname);
5213 else
5214 fprintf(stderr, " (IMSM).\n");
5215 return 3;
5216 }
e2f41b2c
AK		 5217 }
5218
cdddbdbc
DW		 5219 return 0;
5220}
5221
ef6ffade
DW		 5222static __u16 info_to_blocks_per_strip(mdu_array_info_t *info)
5223{
5224 if (info->level == 1)
5225 return 128;
5226 return info->chunk_size >> 9;
5227}
5228
5551b113
CA		 5229static unsigned long long info_to_blocks_per_member(mdu_array_info_t *info,
5230 unsigned long long size)
fcfd9599 5231{
4025c288 5232 if (info->level == 1)
5551b113 5233 return size * 2;
4025c288 5234 else
5551b113 5235 return (size * 2) & ~(info_to_blocks_per_strip(info) - 1);
fcfd9599
DW		 5236}
5237
4d1313e9
DW		 5238static void imsm_update_version_info(struct intel_super *super)
5239{
5240 /* update the version and attributes */
5241 struct imsm_super *mpb = super->anchor;
5242 char *version;
5243 struct imsm_dev *dev;
5244 struct imsm_map *map;
5245 int i;
5246
5247 for (i = 0; i < mpb->num_raid_devs; i++) {
5248 dev = get_imsm_dev(super, i);
238c0a71 5249 map = get_imsm_map(dev, MAP_0);
4d1313e9
DW		 5250 if (__le32_to_cpu(dev->size_high) > 0)
5251 mpb->attributes |= MPB_ATTRIB_2TB;
5252
5253 /* FIXME detect when an array spans a port multiplier */
5254 #if 0
5255 mpb->attributes |= MPB_ATTRIB_PM;
5256 #endif
5257
5258 if (mpb->num_raid_devs > 1 ||
5259 mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) {
5260 version = MPB_VERSION_ATTRIBS;
5261 switch (get_imsm_raid_level(map)) {
5262 case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break;
5263 case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break;
5264 case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break;
5265 case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break;
5266 }
5267 } else {
5268 if (map->num_members >= 5)
5269 version = MPB_VERSION_5OR6_DISK_ARRAY;
5270 else if (dev->status == DEV_CLONE_N_GO)
5271 version = MPB_VERSION_CNG;
5272 else if (get_imsm_raid_level(map) == 5)
5273 version = MPB_VERSION_RAID5;
5274 else if (map->num_members >= 3)
5275 version = MPB_VERSION_3OR4_DISK_ARRAY;
5276 else if (get_imsm_raid_level(map) == 1)
5277 version = MPB_VERSION_RAID1;
5278 else
5279 version = MPB_VERSION_RAID0;
5280 }
5281 strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version);
5282 }
5283}
5284
aa534678
DW		 5285static int check_name(struct intel_super *super, char *name, int quiet)
5286{
5287 struct imsm_super *mpb = super->anchor;
5288 char *reason = NULL;
5289 int i;
5290
5291 if (strlen(name) > MAX_RAID_SERIAL_LEN)
5292 reason = "must be 16 characters or less";
5293
5294 for (i = 0; i < mpb->num_raid_devs; i++) {
5295 struct imsm_dev *dev = get_imsm_dev(super, i);
5296
5297 if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) {
5298 reason = "already exists";
5299 break;
5300 }
5301 }
5302
5303 if (reason && !quiet)
e7b84f9d 5304 pr_err("imsm volume name %s\n", reason);
aa534678
DW		 5305
5306 return !reason;
5307}
5308
8b353278 5309static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info,
5308f117 5310 struct shape *s, char *name,
83cd1e97
N		 5311 char *homehost, int *uuid,
5312 long long data_offset)
cdddbdbc 5313{
c2c087e6
DW		 5314 /* We are creating a volume inside a pre-existing container.
5315 * so st->sb is already set.
5316 */
5317 struct intel_super *super = st->sb;
f36a9ecd 5318 unsigned int sector_size = super->sector_size;
949c47a0 5319 struct imsm_super *mpb = super->anchor;
ba2de7ba 5320 struct intel_dev *dv;
c2c087e6
DW		 5321 struct imsm_dev *dev;
5322 struct imsm_vol *vol;
5323 struct imsm_map *map;
5324 int idx = mpb->num_raid_devs;
5325 int i;
5326 unsigned long long array_blocks;
2c092cad 5327 size_t size_old, size_new;
5551b113 5328 unsigned long long num_data_stripes;
b53bfba6
TM		 5329 unsigned int data_disks;
5330 unsigned long long size_per_member;
cdddbdbc 5331
88c32bb1 5332 if (super->orom && mpb->num_raid_devs >= super->orom->vpa) {
7a862a02 5333 pr_err("This imsm-container already has the maximum of %d volumes\n", super->orom->vpa);
c2c087e6
DW		 5334 return 0;
5335 }
5336
2c092cad
DW		 5337 /* ensure the mpb is large enough for the new data */
5338 size_old = __le32_to_cpu(mpb->mpb_size);
5339 size_new = disks_to_mpb_size(info->nr_disks);
5340 if (size_new > size_old) {
5341 void *mpb_new;
f36a9ecd 5342 size_t size_round = ROUND_UP(size_new, sector_size);
2c092cad 5343
f36a9ecd 5344 if (posix_memalign(&mpb_new, sector_size, size_round) != 0) {
e7b84f9d 5345 pr_err("could not allocate new mpb\n");
2c092cad
DW		 5346 return 0;
5347 }
85337573
AO		 5348 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
5349 MIGR_REC_BUF_SECTORS*
5350 MAX_SECTOR_SIZE) != 0) {
1ade5cc1 5351 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8
AK		 5352 free(super->buf);
5353 free(super);
ea944c8f 5354 free(mpb_new);
8e59f3d8
AK		 5355 return 0;
5356 }
2c092cad
DW		 5357 memcpy(mpb_new, mpb, size_old);
5358 free(mpb);
5359 mpb = mpb_new;
949c47a0 5360 super->anchor = mpb_new;
2c092cad
DW		 5361 mpb->mpb_size = __cpu_to_le32(size_new);
5362 memset(mpb_new + size_old, 0, size_round - size_old);
bbab0940 5363 super->len = size_round;
2c092cad 5364 }
bf5a934a 5365 super->current_vol = idx;
3960e579
DW		 5366
5367 /* handle 'failed_disks' by either:
5368 * a) create dummy disk entries in the table if this the first
5369 * volume in the array. We add them here as this is the only
5370 * opportunity to add them. add_to_super_imsm_volume()
5371 * handles the non-failed disks and continues incrementing
5372 * mpb->num_disks.
5373 * b) validate that 'failed_disks' matches the current number
5374 * of missing disks if the container is populated
d23fe947 5375 */
3960e579 5376 if (super->current_vol == 0) {
d23fe947 5377 mpb->num_disks = 0;
3960e579
DW		 5378 for (i = 0; i < info->failed_disks; i++) {
5379 struct imsm_disk *disk;
5380
5381 mpb->num_disks++;
5382 disk = __get_imsm_disk(mpb, i);
5383 disk->status = CONFIGURED_DISK | FAILED_DISK;
5384 disk->scsi_id = __cpu_to_le32(~(__u32)0);
5385 snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN,
5b975129 5386 "missing:%d", (__u8)i);
3960e579
DW		 5387 }
5388 find_missing(super);
5389 } else {
5390 int missing = 0;
5391 struct dl *d;
5392
5393 for (d = super->missing; d; d = d->next)
5394 missing++;
5395 if (info->failed_disks > missing) {
e7b84f9d 5396 pr_err("unable to add 'missing' disk to container\n");
3960e579
DW		 5397 return 0;
5398 }
5399 }
5a038140 5400
aa534678
DW		 5401 if (!check_name(super, name, 0))
5402 return 0;
503975b9
N		 5403 dv = xmalloc(sizeof(*dv));
5404 dev = xcalloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1));
c2c087e6 5405 strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN);
e03640bd 5406 array_blocks = calc_array_size(info->level, info->raid_disks,
03bcbc65 5407 info->layout, info->chunk_size,
b53bfba6
TM		 5408 s->size * BLOCKS_PER_KB);
5409 data_disks = get_data_disks(info->level, info->layout,
5410 info->raid_disks);
5411 array_blocks = round_size_to_mb(array_blocks, data_disks);
5412 size_per_member = array_blocks / data_disks;
979d38be 5413
fcc2c9da 5414 set_imsm_dev_size(dev, array_blocks);
1a2487c2 5415 dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING);
c2c087e6
DW		 5416 vol = &dev->vol;
5417 vol->migr_state = 0;
1484e727 5418 set_migr_type(dev, MIGR_INIT);
3960e579 5419 vol->dirty = !info->state;
f8f603f1 5420 vol->curr_migr_unit = 0;
238c0a71 5421 map = get_imsm_map(dev, MAP_0);
5551b113 5422 set_pba_of_lba0(map, super->create_offset);
ef6ffade 5423 map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info));
0556e1a2 5424 map->failed_disk_num = ~0;
bf4442ab 5425 if (info->level > 0)
fffaf1ff
N		 5426 map->map_state = (info->state ? IMSM_T_STATE_NORMAL
5427 : IMSM_T_STATE_UNINITIALIZED);
bf4442ab
AK		 5428 else
5429 map->map_state = info->failed_disks ? IMSM_T_STATE_FAILED :
5430 IMSM_T_STATE_NORMAL;
252d23c0 5431 map->ddf = 1;
ef6ffade
DW		 5432
5433 if (info->level == 1 && info->raid_disks > 2) {
38950822
AW		 5434 free(dev);
5435 free(dv);
7a862a02 5436 pr_err("imsm does not support more than 2 disksin a raid1 volume\n");
ef6ffade
DW		 5437 return 0;
5438 }
81062a36
DW		 5439
5440 map->raid_level = info->level;
4d1313e9 5441 if (info->level == 10) {
c2c087e6 5442 map->raid_level = 1;
4d1313e9 5443 map->num_domains = info->raid_disks / 2;
81062a36
DW		 5444 } else if (info->level == 1)
5445 map->num_domains = info->raid_disks;
5446 else
ff596308 5447 map->num_domains = 1;
81062a36 5448
5551b113 5449 /* info->size is only int so use the 'size' parameter instead */
b53bfba6 5450 num_data_stripes = size_per_member / info_to_blocks_per_strip(info);
5551b113
CA		 5451 num_data_stripes /= map->num_domains;
5452 set_num_data_stripes(map, num_data_stripes);
ef6ffade 5453
44490938
MD		 5454 size_per_member += NUM_BLOCKS_DIRTY_STRIPE_REGION;
5455 set_blocks_per_member(map, info_to_blocks_per_member(info,
5456 size_per_member /
5457 BLOCKS_PER_KB));
5458
c2c087e6
DW		 5459 map->num_members = info->raid_disks;
5460 for (i = 0; i < map->num_members; i++) {
5461 /* initialized in add_to_super */
4eb26970 5462 set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD);
c2c087e6 5463 }
949c47a0 5464 mpb->num_raid_devs++;
2a24dc1b
PB		 5465 mpb->num_raid_devs_created++;
5466 dev->my_vol_raid_dev_num = mpb->num_raid_devs_created;
ba2de7ba 5467
b7580566 5468 if (s->consistency_policy <= CONSISTENCY_POLICY_RESYNC) {
c2462068 5469 dev->rwh_policy = RWH_MULTIPLE_OFF;
2432ce9b 5470 } else if (s->consistency_policy == CONSISTENCY_POLICY_PPL) {
c2462068 5471 dev->rwh_policy = RWH_MULTIPLE_DISTRIBUTED;
2432ce9b
AP		 5472 } else {
5473 free(dev);
5474 free(dv);
5475 pr_err("imsm does not support consistency policy %s\n",
5476 map_num(consistency_policies, s->consistency_policy));
5477 return 0;
5478 }
5479
ba2de7ba
DW		 5480 dv->dev = dev;
5481 dv->index = super->current_vol;
5482 dv->next = super->devlist;
5483 super->devlist = dv;
c2c087e6 5484
4d1313e9
DW		 5485 imsm_update_version_info(super);
5486
c2c087e6 5487 return 1;
cdddbdbc
DW		 5488}
5489
bf5a934a 5490static int init_super_imsm(struct supertype *st, mdu_array_info_t *info,
5308f117 5491 struct shape *s, char *name,
83cd1e97
N		 5492 char *homehost, int *uuid,
5493 unsigned long long data_offset)
bf5a934a
DW		 5494{
5495 /* This is primarily called by Create when creating a new array.
5496 * We will then get add_to_super called for each component, and then
5497 * write_init_super called to write it out to each device.
5498 * For IMSM, Create can create on fresh devices or on a pre-existing
5499 * array.
5500 * To create on a pre-existing array a different method will be called.
5501 * This one is just for fresh drives.
5502 */
5503 struct intel_super *super;
5504 struct imsm_super *mpb;
5505 size_t mpb_size;
4d1313e9 5506 char *version;
bf5a934a 5507
83cd1e97 5508 if (data_offset != INVALID_SECTORS) {
ed503f89 5509 pr_err("data-offset not supported by imsm\n");
83cd1e97
N		 5510 return 0;
5511 }
5512
bf5a934a 5513 if (st->sb)
5308f117 5514 return init_super_imsm_volume(st, info, s, name, homehost, uuid,
83cd1e97 5515 data_offset);
e683ca88
DW		 5516
5517 if (info)
5518 mpb_size = disks_to_mpb_size(info->nr_disks);
5519 else
f36a9ecd 5520 mpb_size = MAX_SECTOR_SIZE;
bf5a934a 5521
49133e57 5522 super = alloc_super();
f36a9ecd
PB		 5523 if (super &&
5524 posix_memalign(&super->buf, MAX_SECTOR_SIZE, mpb_size) != 0) {
8d67477f 5525 free_imsm(super);
e683ca88
DW		 5526 super = NULL;
5527 }
5528 if (!super) {
1ade5cc1 5529 pr_err("could not allocate superblock\n");
bf5a934a
DW		 5530 return 0;
5531 }
de44e46f
PB		 5532 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
5533 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) {
1ade5cc1 5534 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8 5535 free(super->buf);
8d67477f 5536 free_imsm(super);
8e59f3d8
AK		 5537 return 0;
5538 }
e683ca88 5539 memset(super->buf, 0, mpb_size);
ef649044 5540 mpb = super->buf;
e683ca88
DW		 5541 mpb->mpb_size = __cpu_to_le32(mpb_size);
5542 st->sb = super;
5543
5544 if (info == NULL) {
5545 /* zeroing superblock */
5546 return 0;
5547 }
bf5a934a 5548
4d1313e9
DW		 5549 mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
5550
5551 version = (char *) mpb->sig;
5552 strcpy(version, MPB_SIGNATURE);
5553 version += strlen(MPB_SIGNATURE);
5554 strcpy(version, MPB_VERSION_RAID0);
bf5a934a 5555
bf5a934a
DW		 5556 return 1;
5557}
5558
f2cc4f7d
AO		 5559static int drive_validate_sector_size(struct intel_super *super, struct dl *dl)
5560{
5561 unsigned int member_sector_size;
5562
5563 if (dl->fd < 0) {
5564 pr_err("Invalid file descriptor for %s\n", dl->devname);
5565 return 0;
5566 }
5567
5568 if (!get_dev_sector_size(dl->fd, dl->devname, &member_sector_size))
5569 return 0;
5570 if (member_sector_size != super->sector_size)
5571 return 0;
5572 return 1;
5573}
5574
f20c3968 5575static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk,
bf5a934a
DW		 5576 int fd, char *devname)
5577{
5578 struct intel_super *super = st->sb;
d23fe947 5579 struct imsm_super *mpb = super->anchor;
3960e579 5580 struct imsm_disk *_disk;
bf5a934a
DW		 5581 struct imsm_dev *dev;
5582 struct imsm_map *map;
3960e579 5583 struct dl *dl, *df;
4eb26970 5584 int slot;
bf5a934a 5585
949c47a0 5586 dev = get_imsm_dev(super, super->current_vol);
238c0a71 5587 map = get_imsm_map(dev, MAP_0);
bf5a934a 5588
208933a7 5589 if (! (dk->state & (1<<MD_DISK_SYNC))) {
e7b84f9d 5590 pr_err("%s: Cannot add spare devices to IMSM volume\n",
208933a7
N		 5591 devname);
5592 return 1;
5593 }
5594
efb30e7f
DW		 5595 if (fd == -1) {
5596 /* we're doing autolayout so grab the pre-marked (in
5597 * validate_geometry) raid_disk
5598 */
5599 for (dl = super->disks; dl; dl = dl->next)
5600 if (dl->raiddisk == dk->raid_disk)
5601 break;
5602 } else {
5603 for (dl = super->disks; dl ; dl = dl->next)
5604 if (dl->major == dk->major &&
5605 dl->minor == dk->minor)
5606 break;
5607 }
d23fe947 5608
208933a7 5609 if (!dl) {
e7b84f9d 5610 pr_err("%s is not a member of the same container\n", devname);
f20c3968 5611 return 1;
208933a7 5612 }
bf5a934a 5613
f2cc4f7d
AO		 5614 if (!drive_validate_sector_size(super, dl)) {
5615 pr_err("Combining drives of different sector size in one volume is not allowed\n");
5616 return 1;
5617 }
5618
d23fe947
DW		 5619 /* add a pristine spare to the metadata */
5620 if (dl->index < 0) {
5621 dl->index = super->anchor->num_disks;
5622 super->anchor->num_disks++;
5623 }
4eb26970
DW		 5624 /* Check the device has not already been added */
5625 slot = get_imsm_disk_slot(map, dl->index);
5626 if (slot >= 0 &&
238c0a71 5627 (get_imsm_ord_tbl_ent(dev, slot, MAP_X) & IMSM_ORD_REBUILD) == 0) {
e7b84f9d 5628 pr_err("%s has been included in this array twice\n",
4eb26970
DW		 5629 devname);
5630 return 1;
5631 }
656b6b5a 5632 set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index);
ee5aad5a 5633 dl->disk.status = CONFIGURED_DISK;
d23fe947 5634
3960e579
DW		 5635 /* update size of 'missing' disks to be at least as large as the
5636 * largest acitve member (we only have dummy missing disks when
5637 * creating the first volume)
5638 */
5639 if (super->current_vol == 0) {
5640 for (df = super->missing; df; df = df->next) {
5551b113
CA		 5641 if (total_blocks(&dl->disk) > total_blocks(&df->disk))
5642 set_total_blocks(&df->disk, total_blocks(&dl->disk));
3960e579
DW		 5643 _disk = __get_imsm_disk(mpb, df->index);
5644 *_disk = df->disk;
5645 }
5646 }
5647
5648 /* refresh unset/failed slots to point to valid 'missing' entries */
5649 for (df = super->missing; df; df = df->next)
5650 for (slot = 0; slot < mpb->num_disks; slot++) {
238c0a71 5651 __u32 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
3960e579
DW		 5652
5653 if ((ord & IMSM_ORD_REBUILD) == 0)
5654 continue;
5655 set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD);
1ace8403 5656 if (is_gen_migration(dev)) {
238c0a71
AK		 5657 struct imsm_map *map2 = get_imsm_map(dev,
5658 MAP_1);
0a108d63 5659 int slot2 = get_imsm_disk_slot(map2, df->index);
089f9d79 5660 if (slot2 < map2->num_members && slot2 >= 0) {
1ace8403 5661 __u32 ord2 = get_imsm_ord_tbl_ent(dev,
238c0a71
AK		 5662 slot2,
5663 MAP_1);
1ace8403
AK		 5664 if ((unsigned)df->index ==
5665 ord_to_idx(ord2))
5666 set_imsm_ord_tbl_ent(map2,
0a108d63 5667 slot2,
1ace8403
AK		 5668 df->index |
5669 IMSM_ORD_REBUILD);
5670 }
5671 }
3960e579
DW		 5672 dprintf("set slot:%d to missing disk:%d\n", slot, df->index);
5673 break;
5674 }
5675
d23fe947
DW		 5676 /* if we are creating the first raid device update the family number */
5677 if (super->current_vol == 0) {
5678 __u32 sum;
5679 struct imsm_dev *_dev = __get_imsm_dev(mpb, 0);
d23fe947 5680
3960e579 5681 _disk = __get_imsm_disk(mpb, dl->index);
791b666a 5682 if (!_dev || !_disk) {
e7b84f9d 5683 pr_err("BUG mpb setup error\n");
791b666a
AW		 5684 return 1;
5685 }
d23fe947
DW		 5686 *_dev = *dev;
5687 *_disk = dl->disk;
148acb7b
DW		 5688 sum = random32();
5689 sum += __gen_imsm_checksum(mpb);
d23fe947 5690 mpb->family_num = __cpu_to_le32(sum);
148acb7b 5691 mpb->orig_family_num = mpb->family_num;
d23fe947 5692 }
ca0748fa 5693 super->current_disk = dl;
f20c3968 5694 return 0;
bf5a934a
DW		 5695}
5696
a8619d23
AK		 5697/* mark_spare()
5698 * Function marks disk as spare and restores disk serial
5699 * in case it was previously marked as failed by takeover operation
5700 * reruns:
5701 * -1 : critical error
5702 * 0 : disk is marked as spare but serial is not set
5703 * 1 : success
5704 */
5705int mark_spare(struct dl *disk)
5706{
5707 __u8 serial[MAX_RAID_SERIAL_LEN];
5708 int ret_val = -1;
5709
5710 if (!disk)
5711 return ret_val;
5712
5713 ret_val = 0;
5714 if (!imsm_read_serial(disk->fd, NULL, serial)) {
5715 /* Restore disk serial number, because takeover marks disk
5716 * as failed and adds to serial ':0' before it becomes
5717 * a spare disk.
5718 */
5719 serialcpy(disk->serial, serial);
5720 serialcpy(disk->disk.serial, serial);
5721 ret_val = 1;
5722 }
5723 disk->disk.status = SPARE_DISK;
5724 disk->index = -1;
5725
5726 return ret_val;
5727}
88654014 5728
f20c3968 5729static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk,
72ca9bcf
N		 5730 int fd, char *devname,
5731 unsigned long long data_offset)
cdddbdbc 5732{
c2c087e6 5733 struct intel_super *super = st->sb;
c2c087e6
DW		 5734 struct dl *dd;
5735 unsigned long long size;
fa7bb6f8 5736 unsigned int member_sector_size;
f2f27e63 5737 __u32 id;
c2c087e6
DW		 5738 int rv;
5739 struct stat stb;
5740
88654014
LM		 5741 /* If we are on an RAID enabled platform check that the disk is
5742 * attached to the raid controller.
5743 * We do not need to test disks attachment for container based additions,
5744 * they shall be already tested when container was created/assembled.
88c32bb1 5745 */
d424212e 5746 rv = find_intel_hba_capability(fd, super, devname);
f2f5c343 5747 /* no orom/efi or non-intel hba of the disk */
f0f5a016
LM		 5748 if (rv != 0) {
5749 dprintf("capability: %p fd: %d ret: %d\n",
5750 super->orom, fd, rv);
5751 return 1;
88c32bb1
DW		 5752 }
5753
f20c3968
DW		 5754 if (super->current_vol >= 0)
5755 return add_to_super_imsm_volume(st, dk, fd, devname);
bf5a934a 5756
c2c087e6 5757 fstat(fd, &stb);
503975b9 5758 dd = xcalloc(sizeof(*dd), 1);
c2c087e6
DW		 5759 dd->major = major(stb.st_rdev);
5760 dd->minor = minor(stb.st_rdev);
503975b9 5761 dd->devname = devname ? xstrdup(devname) : NULL;
c2c087e6 5762 dd->fd = fd;
689c9bf3 5763 dd->e = NULL;
1a64be56 5764 dd->action = DISK_ADD;
c2c087e6 5765 rv = imsm_read_serial(fd, devname, dd->serial);
32ba9157 5766 if (rv) {
e7b84f9d 5767 pr_err("failed to retrieve scsi serial, aborting\n");
20bee0f8
PB		 5768 if (dd->devname)
5769 free(dd->devname);
949c47a0 5770 free(dd);
0030e8d6 5771 abort();
c2c087e6 5772 }
20bee0f8
PB		 5773 if (super->hba && ((super->hba->type == SYS_DEV_NVME) ||
5774 (super->hba->type == SYS_DEV_VMD))) {
5775 int i;
5776 char *devpath = diskfd_to_devpath(fd);
5777 char controller_path[PATH_MAX];
5778
5779 if (!devpath) {
5780 pr_err("failed to get devpath, aborting\n");
5781 if (dd->devname)
5782 free(dd->devname);
5783 free(dd);
5784 return 1;
5785 }
5786
5787 snprintf(controller_path, PATH_MAX-1, "%s/device", devpath);
5788 free(devpath);
5789
5790 if (devpath_to_vendor(controller_path) == 0x8086) {
5791 /*
5792 * If Intel's NVMe drive has serial ended with
5793 * "-A","-B","-1" or "-2" it means that this is "x8"
5794 * device (double drive on single PCIe card).
5795 * User should be warned about potential data loss.
5796 */
5797 for (i = MAX_RAID_SERIAL_LEN-1; i > 0; i--) {
5798 /* Skip empty character at the end */
5799 if (dd->serial[i] == 0)
5800 continue;
5801
5802 if (((dd->serial[i] == 'A') ||
5803 (dd->serial[i] == 'B') ||
5804 (dd->serial[i] == '1') ||
5805 (dd->serial[i] == '2')) &&
5806 (dd->serial[i-1] == '-'))
5807 pr_err("\tThe action you are about to take may put your data at risk.\n"
5808 "\tPlease note that x8 devices may consist of two separate x4 devices "
5809 "located on a single PCIe port.\n"
5810 "\tRAID 0 is the only supported configuration for this type of x8 device.\n");
5811 break;
5812 }
32716c51
PB		 5813 } else if (super->hba->type == SYS_DEV_VMD && super->orom &&
5814 !imsm_orom_has_tpv_support(super->orom)) {
5815 pr_err("\tPlatform configuration does not support non-Intel NVMe drives.\n"
8b751247 5816 "\tPlease refer to Intel(R) RSTe/VROC user guide.\n");
32716c51
PB		 5817 free(dd->devname);
5818 free(dd);
5819 return 1;
20bee0f8
PB		 5820 }
5821 }
c2c087e6 5822
c2c087e6 5823 get_dev_size(fd, NULL, &size);
fa7bb6f8
PB		 5824 get_dev_sector_size(fd, NULL, &member_sector_size);
5825
5826 if (super->sector_size == 0) {
5827 /* this a first device, so sector_size is not set yet */
5828 super->sector_size = member_sector_size;
fa7bb6f8
PB		 5829 }
5830
71e5411e 5831 /* clear migr_rec when adding disk to container */
85337573
AO		 5832 memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
5833 if (lseek64(fd, size - MIGR_REC_SECTOR_POSITION*member_sector_size,
de44e46f 5834 SEEK_SET) >= 0) {
466070ad 5835 if ((unsigned int)write(fd, super->migr_rec_buf,
85337573
AO		 5836 MIGR_REC_BUF_SECTORS*member_sector_size) !=
5837 MIGR_REC_BUF_SECTORS*member_sector_size)
71e5411e
PB		 5838 perror("Write migr_rec failed");
5839 }
5840
c2c087e6 5841 size /= 512;
1f24f035 5842 serialcpy(dd->disk.serial, dd->serial);
5551b113
CA		 5843 set_total_blocks(&dd->disk, size);
5844 if (__le32_to_cpu(dd->disk.total_blocks_hi) > 0) {
5845 struct imsm_super *mpb = super->anchor;
5846 mpb->attributes |= MPB_ATTRIB_2TB_DISK;
5847 }
a8619d23 5848 mark_spare(dd);
c2c087e6 5849 if (sysfs_disk_to_scsi_id(fd, &id) == 0)
b9f594fe 5850 dd->disk.scsi_id = __cpu_to_le32(id);
c2c087e6 5851 else
b9f594fe 5852 dd->disk.scsi_id = __cpu_to_le32(0);
43dad3d6
DW		 5853
5854 if (st->update_tail) {
1a64be56
LM		 5855 dd->next = super->disk_mgmt_list;
5856 super->disk_mgmt_list = dd;
43dad3d6
DW		 5857 } else {
5858 dd->next = super->disks;
5859 super->disks = dd;
ceaf0ee1 5860 super->updates_pending++;
43dad3d6 5861 }
f20c3968
DW		 5862
5863 return 0;
cdddbdbc
DW		 5864}
5865
1a64be56
LM		 5866static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk)
5867{
5868 struct intel_super *super = st->sb;
5869 struct dl *dd;
5870
5871 /* remove from super works only in mdmon - for communication
5872 * manager - monitor. Check if communication memory buffer
5873 * is prepared.
5874 */
5875 if (!st->update_tail) {
1ade5cc1 5876 pr_err("shall be used in mdmon context only\n");
1a64be56
LM		 5877 return 1;
5878 }
503975b9 5879 dd = xcalloc(1, sizeof(*dd));
1a64be56
LM		 5880 dd->major = dk->major;
5881 dd->minor = dk->minor;
1a64be56 5882 dd->fd = -1;
a8619d23 5883 mark_spare(dd);
1a64be56
LM		 5884 dd->action = DISK_REMOVE;
5885
5886 dd->next = super->disk_mgmt_list;
5887 super->disk_mgmt_list = dd;
5888
1a64be56
LM		 5889 return 0;
5890}
5891
f796af5d
DW		 5892static int store_imsm_mpb(int fd, struct imsm_super *mpb);
5893
5894static union {
f36a9ecd 5895 char buf[MAX_SECTOR_SIZE];
f796af5d 5896 struct imsm_super anchor;
f36a9ecd 5897} spare_record __attribute__ ((aligned(MAX_SECTOR_SIZE)));
c2c087e6 5898
d23fe947
DW		 5899/* spare records have their own family number and do not have any defined raid
5900 * devices
5901 */
5902static int write_super_imsm_spares(struct intel_super *super, int doclose)
5903{
d23fe947 5904 struct imsm_super *mpb = super->anchor;
f796af5d 5905 struct imsm_super *spare = &spare_record.anchor;
d23fe947
DW		 5906 __u32 sum;
5907 struct dl *d;
5908
68641cdb
JS		 5909 spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super));
5910 spare->generation_num = __cpu_to_le32(1UL);
f796af5d 5911 spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
68641cdb
JS		 5912 spare->num_disks = 1;
5913 spare->num_raid_devs = 0;
5914 spare->cache_size = mpb->cache_size;
5915 spare->pwr_cycle_count = __cpu_to_le32(1);
f796af5d
DW		 5916
5917 snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH,
5918 MPB_SIGNATURE MPB_VERSION_RAID0);
d23fe947
DW		 5919
5920 for (d = super->disks; d; d = d->next) {
8796fdc4 5921 if (d->index != -1)
d23fe947
DW		 5922 continue;
5923
f796af5d 5924 spare->disk[0] = d->disk;
027c374f
CA		 5925 if (__le32_to_cpu(d->disk.total_blocks_hi) > 0)
5926 spare->attributes |= MPB_ATTRIB_2TB_DISK;
5927
f36a9ecd
PB		 5928 if (super->sector_size == 4096)
5929 convert_to_4k_imsm_disk(&spare->disk[0]);
5930
f796af5d
DW		 5931 sum = __gen_imsm_checksum(spare);
5932 spare->family_num = __cpu_to_le32(sum);
5933 spare->orig_family_num = 0;
5934 sum = __gen_imsm_checksum(spare);
5935 spare->check_sum = __cpu_to_le32(sum);
d23fe947 5936
f796af5d 5937 if (store_imsm_mpb(d->fd, spare)) {
1ade5cc1
N		 5938 pr_err("failed for device %d:%d %s\n",
5939 d->major, d->minor, strerror(errno));
e74255d9 5940 return 1;
d23fe947
DW		 5941 }
5942 if (doclose) {
5943 close(d->fd);
5944 d->fd = -1;
5945 }
5946 }
5947
e74255d9 5948 return 0;
d23fe947
DW		 5949}
5950
36988a3d 5951static int write_super_imsm(struct supertype *st, int doclose)
cdddbdbc 5952{
36988a3d 5953 struct intel_super *super = st->sb;
f36a9ecd 5954 unsigned int sector_size = super->sector_size;
949c47a0 5955 struct imsm_super *mpb = super->anchor;
c2c087e6
DW		 5956 struct dl *d;
5957 __u32 generation;
5958 __u32 sum;
d23fe947 5959 int spares = 0;
949c47a0 5960 int i;
a48ac0a8 5961 __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk);
36988a3d 5962 int num_disks = 0;
146c6260 5963 int clear_migration_record = 1;
bbab0940 5964 __u32 bbm_log_size;
cdddbdbc 5965
c2c087e6
DW		 5966 /* 'generation' is incremented everytime the metadata is written */
5967 generation = __le32_to_cpu(mpb->generation_num);
5968 generation++;
5969 mpb->generation_num = __cpu_to_le32(generation);
5970
148acb7b
DW		 5971 /* fix up cases where previous mdadm releases failed to set
5972 * orig_family_num
5973 */
5974 if (mpb->orig_family_num == 0)
5975 mpb->orig_family_num = mpb->family_num;
5976
d23fe947 5977 for (d = super->disks; d; d = d->next) {
8796fdc4 5978 if (d->index == -1)
d23fe947 5979 spares++;
36988a3d 5980 else {
d23fe947 5981 mpb->disk[d->index] = d->disk;
36988a3d
AK		 5982 num_disks++;
5983 }
d23fe947 5984 }
36988a3d 5985 for (d = super->missing; d; d = d->next) {
47ee5a45 5986 mpb->disk[d->index] = d->disk;
36988a3d
AK		 5987 num_disks++;
5988 }
5989 mpb->num_disks = num_disks;
5990 mpb_size += sizeof(struct imsm_disk) * mpb->num_disks;
b9f594fe 5991
949c47a0
DW		 5992 for (i = 0; i < mpb->num_raid_devs; i++) {
5993 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
36988a3d
AK		 5994 struct imsm_dev *dev2 = get_imsm_dev(super, i);
5995 if (dev && dev2) {
5996 imsm_copy_dev(dev, dev2);
5997 mpb_size += sizeof_imsm_dev(dev, 0);
5998 }
146c6260
AK		 5999 if (is_gen_migration(dev2))
6000 clear_migration_record = 0;
949c47a0 6001 }
bbab0940
TM		 6002
6003 bbm_log_size = get_imsm_bbm_log_size(super->bbm_log);
6004
6005 if (bbm_log_size) {
6006 memcpy((void *)mpb + mpb_size, super->bbm_log, bbm_log_size);
6007 mpb->attributes |= MPB_ATTRIB_BBM;
6008 } else
6009 mpb->attributes &= ~MPB_ATTRIB_BBM;
6010
6011 super->anchor->bbm_log_size = __cpu_to_le32(bbm_log_size);
6012 mpb_size += bbm_log_size;
a48ac0a8 6013 mpb->mpb_size = __cpu_to_le32(mpb_size);
949c47a0 6014
bbab0940
TM		 6015#ifdef DEBUG
6016 assert(super->len == 0 || mpb_size <= super->len);
6017#endif
6018
c2c087e6 6019 /* recalculate checksum */
949c47a0 6020 sum = __gen_imsm_checksum(mpb);
c2c087e6
DW		 6021 mpb->check_sum = __cpu_to_le32(sum);
6022
51d83f5d
AK		 6023 if (super->clean_migration_record_by_mdmon) {
6024 clear_migration_record = 1;
6025 super->clean_migration_record_by_mdmon = 0;
6026 }
146c6260 6027 if (clear_migration_record)
de44e46f 6028 memset(super->migr_rec_buf, 0,
85337573 6029 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
146c6260 6030
f36a9ecd
PB		 6031 if (sector_size == 4096)
6032 convert_to_4k(super);
6033
d23fe947 6034 /* write the mpb for disks that compose raid devices */
c2c087e6 6035 for (d = super->disks; d ; d = d->next) {
86c54047 6036 if (d->index < 0 || is_failed(&d->disk))
d23fe947 6037 continue;
30602f53 6038
146c6260
AK		 6039 if (clear_migration_record) {
6040 unsigned long long dsize;
6041
6042 get_dev_size(d->fd, NULL, &dsize);
de44e46f
PB		 6043 if (lseek64(d->fd, dsize - sector_size,
6044 SEEK_SET) >= 0) {
466070ad
PB		 6045 if ((unsigned int)write(d->fd,
6046 super->migr_rec_buf,
de44e46f
PB		 6047 MIGR_REC_BUF_SECTORS*sector_size) !=
6048 MIGR_REC_BUF_SECTORS*sector_size)
9e2d750d 6049 perror("Write migr_rec failed");
146c6260
AK		 6050 }
6051 }
51d83f5d
AK		 6052
6053 if (store_imsm_mpb(d->fd, mpb))
6054 fprintf(stderr,
1ade5cc1
N		 6055 "failed for device %d:%d (fd: %d)%s\n",
6056 d->major, d->minor,
51d83f5d
AK		 6057 d->fd, strerror(errno));
6058
c2c087e6
DW		 6059 if (doclose) {
6060 close(d->fd);
6061 d->fd = -1;
6062 }
6063 }
6064
d23fe947
DW		 6065 if (spares)
6066 return write_super_imsm_spares(super, doclose);
6067
e74255d9 6068 return 0;
c2c087e6
DW		 6069}
6070
9b1fb677 6071static int create_array(struct supertype *st, int dev_idx)
43dad3d6
DW		 6072{
6073 size_t len;
6074 struct imsm_update_create_array *u;
6075 struct intel_super *super = st->sb;
9b1fb677 6076 struct imsm_dev *dev = get_imsm_dev(super, dev_idx);
238c0a71 6077 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW		 6078 struct disk_info *inf;
6079 struct imsm_disk *disk;
6080 int i;
43dad3d6 6081
54c2c1ea
DW		 6082 len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) +
6083 sizeof(*inf) * map->num_members;
503975b9 6084 u = xmalloc(len);
43dad3d6 6085 u->type = update_create_array;
9b1fb677 6086 u->dev_idx = dev_idx;
43dad3d6 6087 imsm_copy_dev(&u->dev, dev);
54c2c1ea
DW		 6088 inf = get_disk_info(u);
6089 for (i = 0; i < map->num_members; i++) {
238c0a71 6090 int idx = get_imsm_disk_idx(dev, i, MAP_X);
9b1fb677 6091
54c2c1ea 6092 disk = get_imsm_disk(super, idx);
1ca5c8e0
N		 6093 if (!disk)
6094 disk = get_imsm_missing(super, idx);
54c2c1ea
DW		 6095 serialcpy(inf[i].serial, disk->serial);
6096 }
43dad3d6
DW		 6097 append_metadata_update(st, u, len);
6098
6099 return 0;
6100}
6101
1a64be56 6102static int mgmt_disk(struct supertype *st)
43dad3d6
DW		 6103{
6104 struct intel_super *super = st->sb;
6105 size_t len;
1a64be56 6106 struct imsm_update_add_remove_disk *u;
43dad3d6 6107
1a64be56 6108 if (!super->disk_mgmt_list)
43dad3d6
DW		 6109 return 0;
6110
6111 len = sizeof(*u);
503975b9 6112 u = xmalloc(len);
1a64be56 6113 u->type = update_add_remove_disk;
43dad3d6
DW		 6114 append_metadata_update(st, u, len);
6115
6116 return 0;
6117}
2432ce9b
AP		 6118
6119__u32 crc32c_le(__u32 crc, unsigned char const *p, size_t len);
6120
e397cefe
AP		 6121static int write_ppl_header(unsigned long long ppl_sector, int fd, void *buf)
6122{
6123 struct ppl_header *ppl_hdr = buf;
6124 int ret;
6125
6126 ppl_hdr->checksum = __cpu_to_le32(~crc32c_le(~0, buf, PPL_HEADER_SIZE));
6127
6128 if (lseek64(fd, ppl_sector * 512, SEEK_SET) < 0) {
6129 ret = -errno;
6130 perror("Failed to seek to PPL header location");
6131 return ret;
6132 }
6133
6134 if (write(fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) {
6135 ret = -errno;
6136 perror("Write PPL header failed");
6137 return ret;
6138 }
6139
6140 fsync(fd);
6141
6142 return 0;
6143}
6144
2432ce9b
AP		 6145static int write_init_ppl_imsm(struct supertype *st, struct mdinfo *info, int fd)
6146{
6147 struct intel_super *super = st->sb;
6148 void *buf;
6149 struct ppl_header *ppl_hdr;
6150 int ret;
6151
b2514242
PB		 6152 /* first clear entire ppl space */
6153 ret = zero_disk_range(fd, info->ppl_sector, info->ppl_size);
6154 if (ret)
6155 return ret;
6156
6157 ret = posix_memalign(&buf, MAX_SECTOR_SIZE, PPL_HEADER_SIZE);
2432ce9b
AP		 6158 if (ret) {
6159 pr_err("Failed to allocate PPL header buffer\n");
e397cefe 6160 return -ret;
2432ce9b
AP		 6161 }
6162
6163 memset(buf, 0, PPL_HEADER_SIZE);
6164 ppl_hdr = buf;
6165 memset(ppl_hdr->reserved, 0xff, PPL_HDR_RESERVED);
6166 ppl_hdr->signature = __cpu_to_le32(super->anchor->orig_family_num);
b23d0750
AP		 6167
6168 if (info->mismatch_cnt) {
6169 /*
6170 * We are overwriting an invalid ppl. Make one entry with wrong
6171 * checksum to prevent the kernel from skipping resync.
6172 */
6173 ppl_hdr->entries_count = __cpu_to_le32(1);
6174 ppl_hdr->entries[0].checksum = ~0;
6175 }
6176
e397cefe 6177 ret = write_ppl_header(info->ppl_sector, fd, buf);
2432ce9b
AP		 6178
6179 free(buf);
6180 return ret;
6181}
6182
e397cefe
AP		 6183static int is_rebuilding(struct imsm_dev *dev);
6184
2432ce9b
AP		 6185static int validate_ppl_imsm(struct supertype *st, struct mdinfo *info,
6186 struct mdinfo *disk)
6187{
6188 struct intel_super *super = st->sb;
6189 struct dl *d;
e397cefe 6190 void *buf_orig, *buf, *buf_prev = NULL;
2432ce9b 6191 int ret = 0;
e397cefe 6192 struct ppl_header *ppl_hdr = NULL;
2432ce9b
AP		 6193 __u32 crc;
6194 struct imsm_dev *dev;
2432ce9b 6195 __u32 idx;
44b6b876
PB		 6196 unsigned int i;
6197 unsigned long long ppl_offset = 0;
6198 unsigned long long prev_gen_num = 0;
2432ce9b
AP		 6199
6200 if (disk->disk.raid_disk < 0)
6201 return 0;
6202
2432ce9b 6203 dev = get_imsm_dev(super, info->container_member);
2fc0fc63 6204 idx = get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_0);
2432ce9b
AP		 6205 d = get_imsm_dl_disk(super, idx);
6206
6207 if (!d || d->index < 0 || is_failed(&d->disk))
e397cefe
AP		 6208 return 0;
6209
6210 if (posix_memalign(&buf_orig, MAX_SECTOR_SIZE, PPL_HEADER_SIZE * 2)) {
6211 pr_err("Failed to allocate PPL header buffer\n");
6212 return -1;
6213 }
6214 buf = buf_orig;
2432ce9b 6215
44b6b876
PB		 6216 ret = 1;
6217 while (ppl_offset < MULTIPLE_PPL_AREA_SIZE_IMSM) {
e397cefe
AP		 6218 void *tmp;
6219
44b6b876 6220 dprintf("Checking potential PPL at offset: %llu\n", ppl_offset);
2432ce9b 6221
44b6b876
PB		 6222 if (lseek64(d->fd, info->ppl_sector * 512 + ppl_offset,
6223 SEEK_SET) < 0) {
6224 perror("Failed to seek to PPL header location");
6225 ret = -1;
e397cefe 6226 break;
44b6b876 6227 }
2432ce9b 6228
44b6b876
PB		 6229 if (read(d->fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) {
6230 perror("Read PPL header failed");
6231 ret = -1;
e397cefe 6232 break;
44b6b876 6233 }
2432ce9b 6234
44b6b876 6235 ppl_hdr = buf;
2432ce9b 6236
44b6b876
PB		 6237 crc = __le32_to_cpu(ppl_hdr->checksum);
6238 ppl_hdr->checksum = 0;
6239
6240 if (crc != ~crc32c_le(~0, buf, PPL_HEADER_SIZE)) {
6241 dprintf("Wrong PPL header checksum on %s\n",
6242 d->devname);
e397cefe 6243 break;
44b6b876
PB		 6244 }
6245
6246 if (prev_gen_num > __le64_to_cpu(ppl_hdr->generation)) {
6247 /* previous was newest, it was already checked */
e397cefe 6248 break;
44b6b876
PB		 6249 }
6250
6251 if ((__le32_to_cpu(ppl_hdr->signature) !=
6252 super->anchor->orig_family_num)) {
6253 dprintf("Wrong PPL header signature on %s\n",
6254 d->devname);
6255 ret = 1;
e397cefe 6256 break;
44b6b876
PB		 6257 }
6258
6259 ret = 0;
6260 prev_gen_num = __le64_to_cpu(ppl_hdr->generation);
2432ce9b 6261
44b6b876
PB		 6262 ppl_offset += PPL_HEADER_SIZE;
6263 for (i = 0; i < __le32_to_cpu(ppl_hdr->entries_count); i++)
6264 ppl_offset +=
6265 __le32_to_cpu(ppl_hdr->entries[i].pp_size);
e397cefe
AP		 6266
6267 if (!buf_prev)
6268 buf_prev = buf + PPL_HEADER_SIZE;
6269 tmp = buf_prev;
6270 buf_prev = buf;
6271 buf = tmp;
2432ce9b
AP		 6272 }
6273
e397cefe
AP		 6274 if (buf_prev) {
6275 buf = buf_prev;
6276 ppl_hdr = buf_prev;
6277 }
2432ce9b 6278
54148aba
PB		 6279 /*
6280 * Update metadata to use mutliple PPLs area (1MB).
6281 * This is done once for all RAID members
6282 */
6283 if (info->consistency_policy == CONSISTENCY_POLICY_PPL &&
6284 info->ppl_size != (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9)) {
6285 char subarray[20];
6286 struct mdinfo *member_dev;
6287
6288 sprintf(subarray, "%d", info->container_member);
6289
6290 if (mdmon_running(st->container_devnm))
6291 st->update_tail = &st->updates;
6292
6293 if (st->ss->update_subarray(st, subarray, "ppl", NULL)) {
6294 pr_err("Failed to update subarray %s\n",
6295 subarray);
6296 } else {
6297 if (st->update_tail)
6298 flush_metadata_updates(st);
6299 else
6300 st->ss->sync_metadata(st);
6301 info->ppl_size = (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9);
6302 for (member_dev = info->devs; member_dev;
6303 member_dev = member_dev->next)
6304 member_dev->ppl_size =
6305 (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9);
6306 }
6307 }
6308
b23d0750 6309 if (ret == 1) {
2fc0fc63
AP		 6310 struct imsm_map *map = get_imsm_map(dev, MAP_X);
6311
50b9c10d
PB		 6312 if (map->map_state == IMSM_T_STATE_UNINITIALIZED ||
6313 (map->map_state == IMSM_T_STATE_NORMAL &&
2ec9d182 6314 !(dev->vol.dirty & RAIDVOL_DIRTY)) ||
e397cefe 6315 (is_rebuilding(dev) &&
2ec9d182
AP		 6316 dev->vol.curr_migr_unit == 0 &&
6317 get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_1) != idx))
b23d0750
AP		 6318 ret = st->ss->write_init_ppl(st, info, d->fd);
6319 else
6320 info->mismatch_cnt++;
e397cefe
AP		 6321 } else if (ret == 0 &&
6322 ppl_hdr->entries_count == 0 &&
6323 is_rebuilding(dev) &&
6324 info->resync_start == 0) {
6325 /*
6326 * The header has no entries - add a single empty entry and
6327 * rewrite the header to prevent the kernel from going into
6328 * resync after an interrupted rebuild.
6329 */
6330 ppl_hdr->entries_count = __cpu_to_le32(1);
6331 ret = write_ppl_header(info->ppl_sector, d->fd, buf);
b23d0750 6332 }
2432ce9b 6333
e397cefe
AP		 6334 free(buf_orig);
6335
2432ce9b
AP		 6336 return ret;
6337}
6338
2432ce9b
AP		 6339static int write_init_ppl_imsm_all(struct supertype *st, struct mdinfo *info)
6340{
6341 struct intel_super *super = st->sb;
6342 struct dl *d;
6343 int ret = 0;
6344
6345 if (info->consistency_policy != CONSISTENCY_POLICY_PPL ||
6346 info->array.level != 5)
6347 return 0;
6348
6349 for (d = super->disks; d ; d = d->next) {
6350 if (d->index < 0 || is_failed(&d->disk))
6351 continue;
6352
6353 ret = st->ss->write_init_ppl(st, info, d->fd);
6354 if (ret)
6355 break;
6356 }
6357
6358 return ret;
6359}
43dad3d6 6360
c2c087e6
DW		 6361static int write_init_super_imsm(struct supertype *st)
6362{
9b1fb677
DW		 6363 struct intel_super *super = st->sb;
6364 int current_vol = super->current_vol;
2432ce9b
AP		 6365 int rv = 0;
6366 struct mdinfo info;
6367
6368 getinfo_super_imsm(st, &info, NULL);
9b1fb677
DW		 6369
6370 /* we are done with current_vol reset it to point st at the container */
6371 super->current_vol = -1;
6372
8273f55e 6373 if (st->update_tail) {
43dad3d6
DW		 6374 /* queue the recently created array / added disk
6375 * as a metadata update */
8273f55e 6376
43dad3d6 6377 /* determine if we are creating a volume or adding a disk */
9b1fb677 6378 if (current_vol < 0) {
1a64be56
LM		 6379 /* in the mgmt (add/remove) disk case we are running
6380 * in mdmon context, so don't close fd's
43dad3d6 6381 */
2432ce9b
AP		 6382 rv = mgmt_disk(st);
6383 } else {
6384 rv = write_init_ppl_imsm_all(st, &info);
6385 if (!rv)
6386 rv = create_array(st, current_vol);
6387 }
d682f344
N		 6388 } else {
6389 struct dl *d;
6390 for (d = super->disks; d; d = d->next)
ba728be7 6391 Kill(d->devname, NULL, 0, -1, 1);
2432ce9b
AP		 6392 if (current_vol >= 0)
6393 rv = write_init_ppl_imsm_all(st, &info);
6394 if (!rv)
6395 rv = write_super_imsm(st, 1);
d682f344 6396 }
2432ce9b
AP		 6397
6398 return rv;
cdddbdbc
DW		 6399}
6400
e683ca88 6401static int store_super_imsm(struct supertype *st, int fd)
cdddbdbc 6402{
e683ca88
DW		 6403 struct intel_super *super = st->sb;
6404 struct imsm_super *mpb = super ? super->anchor : NULL;
551c80c1 6405
e683ca88 6406 if (!mpb)
ad97895e
DW		 6407 return 1;
6408
f36a9ecd
PB		 6409 if (super->sector_size == 4096)
6410 convert_to_4k(super);
e683ca88 6411 return store_imsm_mpb(fd, mpb);
cdddbdbc
DW		 6412}
6413
cdddbdbc
DW		 6414static int validate_geometry_imsm_container(struct supertype *st, int level,
6415 int layout, int raiddisks, int chunk,
af4348dd
N		 6416 unsigned long long size,
6417 unsigned long long data_offset,
6418 char *dev,
2c514b71
NB		 6419 unsigned long long *freesize,
6420 int verbose)
cdddbdbc 6421{
c2c087e6
DW		 6422 int fd;
6423 unsigned long long ldsize;
594dc1b8 6424 struct intel_super *super;
f2f5c343 6425 int rv = 0;
cdddbdbc 6426
c2c087e6
DW		 6427 if (level != LEVEL_CONTAINER)
6428 return 0;
6429 if (!dev)
6430 return 1;
6431
6432 fd = open(dev, O_RDONLY|O_EXCL, 0);
6433 if (fd < 0) {
ba728be7 6434 if (verbose > 0)
e7b84f9d 6435 pr_err("imsm: Cannot open %s: %s\n",
2c514b71 6436 dev, strerror(errno));
c2c087e6
DW		 6437 return 0;
6438 }
6439 if (!get_dev_size(fd, dev, &ldsize)) {
6440 close(fd);
6441 return 0;
6442 }
f2f5c343
LM		 6443
6444 /* capabilities retrieve could be possible
6445 * note that there is no fd for the disks in array.
6446 */
6447 super = alloc_super();
8d67477f
TM		 6448 if (!super) {
6449 close(fd);
6450 return 0;
6451 }
fa7bb6f8
PB		 6452 if (!get_dev_sector_size(fd, NULL, &super->sector_size)) {
6453 close(fd);
6454 free_imsm(super);
6455 return 0;
6456 }
6457
ba728be7 6458 rv = find_intel_hba_capability(fd, super, verbose > 0 ? dev : NULL);
f2f5c343
LM		 6459 if (rv != 0) {
6460#if DEBUG
6461 char str[256];
6462 fd2devname(fd, str);
1ade5cc1 6463 dprintf("fd: %d %s orom: %p rv: %d raiddisk: %d\n",
f2f5c343
LM		 6464 fd, str, super->orom, rv, raiddisks);
6465#endif
6466 /* no orom/efi or non-intel hba of the disk */
6467 close(fd);
6468 free_imsm(super);
6469 return 0;
6470 }
c2c087e6 6471 close(fd);
9126b9a8
CA		 6472 if (super->orom) {
6473 if (raiddisks > super->orom->tds) {
6474 if (verbose)
7a862a02 6475 pr_err("%d exceeds maximum number of platform supported disks: %d\n",
9126b9a8
CA		 6476 raiddisks, super->orom->tds);
6477 free_imsm(super);
6478 return 0;
6479 }
6480 if ((super->orom->attr & IMSM_OROM_ATTR_2TB_DISK) == 0 &&
6481 (ldsize >> 9) >> 32 > 0) {
6482 if (verbose)
e7b84f9d 6483 pr_err("%s exceeds maximum platform supported size\n", dev);
9126b9a8
CA		 6484 free_imsm(super);
6485 return 0;
6486 }
f2f5c343 6487 }
c2c087e6 6488
af4348dd 6489 *freesize = avail_size_imsm(st, ldsize >> 9, data_offset);
f2f5c343 6490 free_imsm(super);
c2c087e6
DW		 6491
6492 return 1;
cdddbdbc
DW		 6493}
6494
0dcecb2e
DW		 6495static unsigned long long find_size(struct extent *e, int *idx, int num_extents)
6496{
6497 const unsigned long long base_start = e[*idx].start;
6498 unsigned long long end = base_start + e[*idx].size;
6499 int i;
6500
6501 if (base_start == end)
6502 return 0;
6503
6504 *idx = *idx + 1;
6505 for (i = *idx; i < num_extents; i++) {
6506 /* extend overlapping extents */
6507 if (e[i].start >= base_start &&
6508 e[i].start <= end) {
6509 if (e[i].size == 0)
6510 return 0;
6511 if (e[i].start + e[i].size > end)
6512 end = e[i].start + e[i].size;
6513 } else if (e[i].start > end) {
6514 *idx = i;
6515 break;
6516 }
6517 }
6518
6519 return end - base_start;
6520}
6521
6522static unsigned long long merge_extents(struct intel_super *super, int sum_extents)
6523{
6524 /* build a composite disk with all known extents and generate a new
6525 * 'maxsize' given the "all disks in an array must share a common start
6526 * offset" constraint
6527 */
503975b9 6528 struct extent *e = xcalloc(sum_extents, sizeof(*e));
0dcecb2e
DW		 6529 struct dl *dl;
6530 int i, j;
6531 int start_extent;
6532 unsigned long long pos;
b9d77223 6533 unsigned long long start = 0;
0dcecb2e
DW		 6534 unsigned long long maxsize;
6535 unsigned long reserve;
6536
0dcecb2e
DW		 6537 /* coalesce and sort all extents. also, check to see if we need to
6538 * reserve space between member arrays
6539 */
6540 j = 0;
6541 for (dl = super->disks; dl; dl = dl->next) {
6542 if (!dl->e)
6543 continue;
6544 for (i = 0; i < dl->extent_cnt; i++)
6545 e[j++] = dl->e[i];
6546 }
6547 qsort(e, sum_extents, sizeof(*e), cmp_extent);
6548
6549 /* merge extents */
6550 i = 0;
6551 j = 0;
6552 while (i < sum_extents) {
6553 e[j].start = e[i].start;
6554 e[j].size = find_size(e, &i, sum_extents);
6555 j++;
6556 if (e[j-1].size == 0)
6557 break;
6558 }
6559
6560 pos = 0;
6561 maxsize = 0;
6562 start_extent = 0;
6563 i = 0;
6564 do {
6565 unsigned long long esize;
6566
6567 esize = e[i].start - pos;
6568 if (esize >= maxsize) {
6569 maxsize = esize;
6570 start = pos;
6571 start_extent = i;
6572 }
6573 pos = e[i].start + e[i].size;
6574 i++;
6575 } while (e[i-1].size);
6576 free(e);
6577
a7dd165b
DW		 6578 if (maxsize == 0)
6579 return 0;
6580
6581 /* FIXME assumes volume at offset 0 is the first volume in a
6582 * container
6583 */
0dcecb2e
DW		 6584 if (start_extent > 0)
6585 reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */
6586 else
6587 reserve = 0;
6588
6589 if (maxsize < reserve)
a7dd165b 6590 return 0;
0dcecb2e 6591
5551b113 6592 super->create_offset = ~((unsigned long long) 0);
0dcecb2e 6593 if (start + reserve > super->create_offset)
a7dd165b 6594 return 0; /* start overflows create_offset */
0dcecb2e
DW		 6595 super->create_offset = start + reserve;
6596
6597 return maxsize - reserve;
6598}
6599
88c32bb1
DW		 6600static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks)
6601{
6602 if (level < 0 || level == 6 || level == 4)
6603 return 0;
6604
6605 /* if we have an orom prevent invalid raid levels */
6606 if (orom)
6607 switch (level) {
6608 case 0: return imsm_orom_has_raid0(orom);
6609 case 1:
6610 if (raiddisks > 2)
6611 return imsm_orom_has_raid1e(orom);
1c556e92
DW		 6612 return imsm_orom_has_raid1(orom) && raiddisks == 2;
6613 case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4;
6614 case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2;
88c32bb1
DW		 6615 }
6616 else
6617 return 1; /* not on an Intel RAID platform so anything goes */
6618
6619 return 0;
6620}
6621
ca9de185
LM		 6622static int
6623active_arrays_by_format(char *name, char* hba, struct md_list **devlist,
6624 int dpa, int verbose)
6625{
6626 struct mdstat_ent *mdstat = mdstat_read(0, 0);
594dc1b8 6627 struct mdstat_ent *memb;
ca9de185
LM		 6628 int count = 0;
6629 int num = 0;
594dc1b8 6630 struct md_list *dv;
ca9de185
LM		 6631 int found;
6632
6633 for (memb = mdstat ; memb ; memb = memb->next) {
6634 if (memb->metadata_version &&
fc54fe7a 6635 (strncmp(memb->metadata_version, "external:", 9) == 0) &&
ca9de185
LM		 6636 (strcmp(&memb->metadata_version[9], name) == 0) &&
6637 !is_subarray(memb->metadata_version+9) &&
6638 memb->members) {
6639 struct dev_member *dev = memb->members;
6640 int fd = -1;
6641 while(dev && (fd < 0)) {
503975b9
N		 6642 char *path = xmalloc(strlen(dev->name) + strlen("/dev/") + 1);
6643 num = sprintf(path, "%s%s", "/dev/", dev->name);
6644 if (num > 0)
6645 fd = open(path, O_RDONLY, 0);
089f9d79 6646 if (num <= 0 || fd < 0) {
676e87a8 6647 pr_vrb("Cannot open %s: %s\n",
503975b9 6648 dev->name, strerror(errno));
ca9de185 6649 }
503975b9 6650 free(path);
ca9de185
LM		 6651 dev = dev->next;
6652 }
6653 found = 0;
089f9d79 6654 if (fd >= 0 && disk_attached_to_hba(fd, hba)) {
ca9de185
LM		 6655 struct mdstat_ent *vol;
6656 for (vol = mdstat ; vol ; vol = vol->next) {
089f9d79 6657 if (vol->active > 0 &&
ca9de185 6658 vol->metadata_version &&
9581efb1 6659 is_container_member(vol, memb->devnm)) {
ca9de185
LM		 6660 found++;
6661 count++;
6662 }
6663 }
6664 if (*devlist && (found < dpa)) {
503975b9 6665 dv = xcalloc(1, sizeof(*dv));
9581efb1
N		 6666 dv->devname = xmalloc(strlen(memb->devnm) + strlen("/dev/") + 1);
6667 sprintf(dv->devname, "%s%s", "/dev/", memb->devnm);
503975b9
N		 6668 dv->found = found;
6669 dv->used = 0;
6670 dv->next = *devlist;
6671 *devlist = dv;
ca9de185
LM		 6672 }
6673 }
6674 if (fd >= 0)
6675 close(fd);
6676 }
6677 }
6678 free_mdstat(mdstat);
6679 return count;
6680}
6681
6682#ifdef DEBUG_LOOP
6683static struct md_list*
6684get_loop_devices(void)
6685{
6686 int i;
6687 struct md_list *devlist = NULL;
594dc1b8 6688 struct md_list *dv;
ca9de185
LM		 6689
6690 for(i = 0; i < 12; i++) {
503975b9
N		 6691 dv = xcalloc(1, sizeof(*dv));
6692 dv->devname = xmalloc(40);
ca9de185
LM		 6693 sprintf(dv->devname, "/dev/loop%d", i);
6694 dv->next = devlist;
6695 devlist = dv;
6696 }
6697 return devlist;
6698}
6699#endif
6700
6701static struct md_list*
6702get_devices(const char *hba_path)
6703{
6704 struct md_list *devlist = NULL;
594dc1b8 6705 struct md_list *dv;
ca9de185
LM		 6706 struct dirent *ent;
6707 DIR *dir;
6708 int err = 0;
6709
6710#if DEBUG_LOOP
6711 devlist = get_loop_devices();
6712 return devlist;
6713#endif
6714 /* scroll through /sys/dev/block looking for devices attached to
6715 * this hba
6716 */
6717 dir = opendir("/sys/dev/block");
6718 for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) {
6719 int fd;
6720 char buf[1024];
6721 int major, minor;
6722 char *path = NULL;
6723 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
6724 continue;
6725 path = devt_to_devpath(makedev(major, minor));
6726 if (!path)
6727 continue;
6728 if (!path_attached_to_hba(path, hba_path)) {
6729 free(path);
6730 path = NULL;
6731 continue;
6732 }
6733 free(path);
6734 path = NULL;
6735 fd = dev_open(ent->d_name, O_RDONLY);
6736 if (fd >= 0) {
6737 fd2devname(fd, buf);
6738 close(fd);
6739 } else {
e7b84f9d 6740 pr_err("cannot open device: %s\n",
ca9de185
LM		 6741 ent->d_name);
6742 continue;
6743 }
6744
503975b9
N		 6745 dv = xcalloc(1, sizeof(*dv));
6746 dv->devname = xstrdup(buf);
ca9de185
LM		 6747 dv->next = devlist;
6748 devlist = dv;
6749 }
6750 if (err) {
6751 while(devlist) {
6752 dv = devlist;
6753 devlist = devlist->next;
6754 free(dv->devname);
6755 free(dv);
6756 }
6757 }
562aa102 6758 closedir(dir);
ca9de185
LM		 6759 return devlist;
6760}
6761
6762static int
6763count_volumes_list(struct md_list *devlist, char *homehost,
6764 int verbose, int *found)
6765{
6766 struct md_list *tmpdev;
6767 int count = 0;
594dc1b8 6768 struct supertype *st;
ca9de185
LM		 6769
6770 /* first walk the list of devices to find a consistent set
6771 * that match the criterea, if that is possible.
6772 * We flag the ones we like with 'used'.
6773 */
6774 *found = 0;
6775 st = match_metadata_desc_imsm("imsm");
6776 if (st == NULL) {
676e87a8 6777 pr_vrb("cannot allocate memory for imsm supertype\n");
ca9de185
LM		 6778 return 0;
6779 }
6780
6781 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
6782 char *devname = tmpdev->devname;
0a6bff09 6783 dev_t rdev;
ca9de185
LM		 6784 struct supertype *tst;
6785 int dfd;
6786 if (tmpdev->used > 1)
6787 continue;
6788 tst = dup_super(st);
6789 if (tst == NULL) {
676e87a8 6790 pr_vrb("cannot allocate memory for imsm supertype\n");
ca9de185
LM		 6791 goto err_1;
6792 }
6793 tmpdev->container = 0;
6794 dfd = dev_open(devname, O_RDONLY|O_EXCL);
6795 if (dfd < 0) {
1ade5cc1 6796 dprintf("cannot open device %s: %s\n",
ca9de185
LM		 6797 devname, strerror(errno));
6798 tmpdev->used = 2;
0a6bff09 6799 } else if (!fstat_is_blkdev(dfd, devname, &rdev)) {
ca9de185
LM		 6800 tmpdev->used = 2;
6801 } else if (must_be_container(dfd)) {
6802 struct supertype *cst;
6803 cst = super_by_fd(dfd, NULL);
6804 if (cst == NULL) {
1ade5cc1 6805 dprintf("cannot recognize container type %s\n",
ca9de185
LM		 6806 devname);
6807 tmpdev->used = 2;
6808 } else if (tst->ss != st->ss) {
1ade5cc1 6809 dprintf("non-imsm container - ignore it: %s\n",
ca9de185
LM		 6810 devname);
6811 tmpdev->used = 2;
6812 } else if (!tst->ss->load_container ||
6813 tst->ss->load_container(tst, dfd, NULL))
6814 tmpdev->used = 2;
6815 else {
6816 tmpdev->container = 1;
6817 }
6818 if (cst)
6819 cst->ss->free_super(cst);
6820 } else {
0a6bff09 6821 tmpdev->st_rdev = rdev;
ca9de185 6822 if (tst->ss->load_super(tst,dfd, NULL)) {
1ade5cc1 6823 dprintf("no RAID superblock on %s\n",
ca9de185
LM		 6824 devname);
6825 tmpdev->used = 2;
6826 } else if (tst->ss->compare_super == NULL) {
1ade5cc1 6827 dprintf("Cannot assemble %s metadata on %s\n",
ca9de185
LM		 6828 tst->ss->name, devname);
6829 tmpdev->used = 2;
6830 }
6831 }
6832 if (dfd >= 0)
6833 close(dfd);
6834 if (tmpdev->used == 2 || tmpdev->used == 4) {
6835 /* Ignore unrecognised devices during auto-assembly */
6836 goto loop;
6837 }
6838 else {
6839 struct mdinfo info;
6840 tst->ss->getinfo_super(tst, &info, NULL);
6841
6842 if (st->minor_version == -1)
6843 st->minor_version = tst->minor_version;
6844
6845 if (memcmp(info.uuid, uuid_zero,
6846 sizeof(int[4])) == 0) {
6847 /* this is a floating spare. It cannot define
6848 * an array unless there are no more arrays of
6849 * this type to be found. It can be included
6850 * in an array of this type though.
6851 */
6852 tmpdev->used = 3;
6853 goto loop;
6854 }
6855
6856 if (st->ss != tst->ss ||
6857 st->minor_version != tst->minor_version ||
6858 st->ss->compare_super(st, tst) != 0) {
6859 /* Some mismatch. If exactly one array matches this host,
6860 * we can resolve on that one.
6861 * Or, if we are auto assembling, we just ignore the second
6862 * for now.
6863 */
1ade5cc1 6864 dprintf("superblock on %s doesn't match others - assembly aborted\n",
ca9de185
LM		 6865 devname);
6866 goto loop;
6867 }
6868 tmpdev->used = 1;
6869 *found = 1;
6870 dprintf("found: devname: %s\n", devname);
6871 }
6872 loop:
6873 if (tst)
6874 tst->ss->free_super(tst);
6875 }
6876 if (*found != 0) {
6877 int err;
6878 if ((err = load_super_imsm_all(st, -1, &st->sb, NULL, devlist, 0)) == 0) {
6879 struct mdinfo *iter, *head = st->ss->container_content(st, NULL);
6880 for (iter = head; iter; iter = iter->next) {
6881 dprintf("content->text_version: %s vol\n",
6882 iter->text_version);
6883 if (iter->array.state & (1<<MD_SB_BLOCK_VOLUME)) {
6884 /* do not assemble arrays with unsupported
6885 configurations */
1ade5cc1 6886 dprintf("Cannot activate member %s.\n",
ca9de185
LM		 6887 iter->text_version);
6888 } else
6889 count++;
6890 }
6891 sysfs_free(head);
6892
6893 } else {
1ade5cc1 6894 dprintf("No valid super block on device list: err: %d %p\n",
ca9de185
LM		 6895 err, st->sb);
6896 }
6897 } else {
1ade5cc1 6898 dprintf("no more devices to examine\n");
ca9de185
LM		 6899 }
6900
6901 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
089f9d79 6902 if (tmpdev->used == 1 && tmpdev->found) {
ca9de185
LM		 6903 if (count) {
6904 if (count < tmpdev->found)
6905 count = 0;
6906 else
6907 count -= tmpdev->found;
6908 }
6909 }
6910 if (tmpdev->used == 1)
6911 tmpdev->used = 4;
6912 }
6913 err_1:
6914 if (st)
6915 st->ss->free_super(st);
6916 return count;
6917}
6918
d3c11416
AO		 6919static int __count_volumes(char *hba_path, int dpa, int verbose,
6920 int cmp_hba_path)
ca9de185 6921{
72a45777 6922 struct sys_dev *idev, *intel_devices = find_intel_devices();
ca9de185 6923 int count = 0;
72a45777
PB		 6924 const struct orom_entry *entry;
6925 struct devid_list *dv, *devid_list;
ca9de185 6926
d3c11416 6927 if (!hba_path)
ca9de185
LM		 6928 return 0;
6929
72a45777 6930 for (idev = intel_devices; idev; idev = idev->next) {
d3c11416
AO		 6931 if (strstr(idev->path, hba_path))
6932 break;
72a45777
PB		 6933 }
6934
6935 if (!idev || !idev->dev_id)
ca9de185 6936 return 0;
72a45777
PB		 6937
6938 entry = get_orom_entry_by_device_id(idev->dev_id);
6939
6940 if (!entry || !entry->devid_list)
6941 return 0;
6942
6943 devid_list = entry->devid_list;
6944 for (dv = devid_list; dv; dv = dv->next) {
594dc1b8 6945 struct md_list *devlist;
d3c11416
AO		 6946 struct sys_dev *device = NULL;
6947 char *hpath;
72a45777
PB		 6948 int found = 0;
6949
d3c11416
AO		 6950 if (cmp_hba_path)
6951 device = device_by_id_and_path(dv->devid, hba_path);
6952 else
6953 device = device_by_id(dv->devid);
6954
72a45777 6955 if (device)
d3c11416 6956 hpath = device->path;
72a45777
PB		 6957 else
6958 return 0;
6959
d3c11416 6960 devlist = get_devices(hpath);
72a45777
PB		 6961 /* if no intel devices return zero volumes */
6962 if (devlist == NULL)
6963 return 0;
6964
d3c11416
AO		 6965 count += active_arrays_by_format("imsm", hpath, &devlist, dpa,
6966 verbose);
6967 dprintf("path: %s active arrays: %d\n", hpath, count);
72a45777
PB		 6968 if (devlist == NULL)
6969 return 0;
6970 do {
6971 found = 0;
6972 count += count_volumes_list(devlist,
6973 NULL,
6974 verbose,
6975 &found);
6976 dprintf("found %d count: %d\n", found, count);
6977 } while (found);
6978
d3c11416 6979 dprintf("path: %s total number of volumes: %d\n", hpath, count);
72a45777
PB		 6980
6981 while (devlist) {
6982 struct md_list *dv = devlist;
6983 devlist = devlist->next;
6984 free(dv->devname);
6985 free(dv);
6986 }
ca9de185
LM		 6987 }
6988 return count;
6989}
6990
d3c11416
AO		 6991static int count_volumes(struct intel_hba *hba, int dpa, int verbose)
6992{
6993 if (!hba)
6994 return 0;
6995 if (hba->type == SYS_DEV_VMD) {
6996 struct sys_dev *dev;
6997 int count = 0;
6998
6999 for (dev = find_intel_devices(); dev; dev = dev->next) {
7000 if (dev->type == SYS_DEV_VMD)
7001 count += __count_volumes(dev->path, dpa,
7002 verbose, 1);
7003 }
7004 return count;
7005 }
7006 return __count_volumes(hba->path, dpa, verbose, 0);
7007}
7008
cd9d1ac7
DW		 7009static int imsm_default_chunk(const struct imsm_orom *orom)
7010{
7011 /* up to 512 if the plaform supports it, otherwise the platform max.
7012 * 128 if no platform detected
7013 */
7014 int fs = max(7, orom ? fls(orom->sss) : 0);
7015
7016 return min(512, (1 << fs));
7017}
73408129 7018
6592ce37
DW		 7019static int
7020validate_geometry_imsm_orom(struct intel_super *super, int level, int layout,
2cc699af 7021 int raiddisks, int *chunk, unsigned long long size, int verbose)
6592ce37 7022{
660260d0
DW		 7023 /* check/set platform and metadata limits/defaults */
7024 if (super->orom && raiddisks > super->orom->dpa) {
676e87a8 7025 pr_vrb("platform supports a maximum of %d disks per array\n",
660260d0 7026 super->orom->dpa);
73408129
LM		 7027 return 0;
7028 }
7029
5d500228 7030 /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */
660260d0 7031 if (!is_raid_level_supported(super->orom, level, raiddisks)) {
676e87a8 7032 pr_vrb("platform does not support raid%d with %d disk%s\n",
6592ce37
DW		 7033 level, raiddisks, raiddisks > 1 ? "s" : "");
7034 return 0;
7035 }
cd9d1ac7 7036
7ccc4cc4 7037 if (*chunk == 0 || *chunk == UnSet)
cd9d1ac7
DW		 7038 *chunk = imsm_default_chunk(super->orom);
7039
7ccc4cc4 7040 if (super->orom && !imsm_orom_has_chunk(super->orom, *chunk)) {
676e87a8 7041 pr_vrb("platform does not support a chunk size of: %d\n", *chunk);
cd9d1ac7 7042 return 0;
6592ce37 7043 }
cd9d1ac7 7044
6592ce37
DW		 7045 if (layout != imsm_level_to_layout(level)) {
7046 if (level == 5)
676e87a8 7047 pr_vrb("imsm raid 5 only supports the left-asymmetric layout\n");
6592ce37 7048 else if (level == 10)
676e87a8 7049 pr_vrb("imsm raid 10 only supports the n2 layout\n");
6592ce37 7050 else
676e87a8 7051 pr_vrb("imsm unknown layout %#x for this raid level %d\n",
6592ce37
DW		 7052 layout, level);
7053 return 0;
7054 }
2cc699af 7055
7ccc4cc4 7056 if (super->orom && (super->orom->attr & IMSM_OROM_ATTR_2TB) == 0 &&
2cc699af 7057 (calc_array_size(level, raiddisks, layout, *chunk, size) >> 32) > 0) {
676e87a8 7058 pr_vrb("platform does not support a volume size over 2TB\n");
2cc699af
CA		 7059 return 0;
7060 }
614902f6 7061
6592ce37
DW		 7062 return 1;
7063}
7064
1011e834 7065/* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd
c2c087e6
DW		 7066 * FIX ME add ahci details
7067 */
8b353278 7068static int validate_geometry_imsm_volume(struct supertype *st, int level,
c21e737b 7069 int layout, int raiddisks, int *chunk,
af4348dd
N		 7070 unsigned long long size,
7071 unsigned long long data_offset,
7072 char *dev,
2c514b71
NB		 7073 unsigned long long *freesize,
7074 int verbose)
cdddbdbc 7075{
9e04ac1c 7076 dev_t rdev;
c2c087e6 7077 struct intel_super *super = st->sb;
b2916f25 7078 struct imsm_super *mpb;
c2c087e6
DW		 7079 struct dl *dl;
7080 unsigned long long pos = 0;
7081 unsigned long long maxsize;
7082 struct extent *e;
7083 int i;
cdddbdbc 7084
88c32bb1
DW		 7085 /* We must have the container info already read in. */
7086 if (!super)
c2c087e6
DW		 7087 return 0;
7088
b2916f25
JS		 7089 mpb = super->anchor;
7090
2cc699af 7091 if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, size, verbose)) {
3e684231 7092 pr_err("RAID geometry validation failed. Cannot proceed with the action(s).\n");
c2c087e6 7093 return 0;
d54559f0 7094 }
c2c087e6
DW		 7095 if (!dev) {
7096 /* General test: make sure there is space for
2da8544a
DW		 7097 * 'raiddisks' device extents of size 'size' at a given
7098 * offset
c2c087e6 7099 */
e46273eb 7100 unsigned long long minsize = size;
b7528a20 7101 unsigned long long start_offset = MaxSector;
c2c087e6
DW		 7102 int dcnt = 0;
7103 if (minsize == 0)
7104 minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
7105 for (dl = super->disks; dl ; dl = dl->next) {
7106 int found = 0;
7107
bf5a934a 7108 pos = 0;
c2c087e6
DW		 7109 i = 0;
7110 e = get_extents(super, dl);
7111 if (!e) continue;
7112 do {
7113 unsigned long long esize;
7114 esize = e[i].start - pos;
7115 if (esize >= minsize)
7116 found = 1;
b7528a20 7117 if (found && start_offset == MaxSector) {
2da8544a
DW		 7118 start_offset = pos;
7119 break;
7120 } else if (found && pos != start_offset) {
7121 found = 0;
7122 break;
7123 }
c2c087e6
DW		 7124 pos = e[i].start + e[i].size;
7125 i++;
7126 } while (e[i-1].size);
7127 if (found)
7128 dcnt++;
7129 free(e);
7130 }
7131 if (dcnt < raiddisks) {
2c514b71 7132 if (verbose)
7a862a02 7133 pr_err("imsm: Not enough devices with space for this array (%d < %d)\n",
2c514b71 7134 dcnt, raiddisks);
c2c087e6
DW		 7135 return 0;
7136 }
7137 return 1;
7138 }
0dcecb2e 7139
c2c087e6 7140 /* This device must be a member of the set */
9e04ac1c 7141 if (!stat_is_blkdev(dev, &rdev))
c2c087e6
DW		 7142 return 0;
7143 for (dl = super->disks ; dl ; dl = dl->next) {
9e04ac1c
ZL		 7144 if (dl->major == (int)major(rdev) &&
7145 dl->minor == (int)minor(rdev))
c2c087e6
DW		 7146 break;
7147 }
7148 if (!dl) {
2c514b71 7149 if (verbose)
7a862a02 7150 pr_err("%s is not in the same imsm set\n", dev);
c2c087e6 7151 return 0;
a20d2ba5
DW		 7152 } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) {
7153 /* If a volume is present then the current creation attempt
7154 * cannot incorporate new spares because the orom may not
7155 * understand this configuration (all member disks must be
7156 * members of each array in the container).
7157 */
7a862a02
N		 7158 pr_err("%s is a spare and a volume is already defined for this container\n", dev);
7159 pr_err("The option-rom requires all member disks to be a member of all volumes\n");
a20d2ba5 7160 return 0;
5fe62b94
WD		 7161 } else if (super->orom && mpb->num_raid_devs > 0 &&
7162 mpb->num_disks != raiddisks) {
7a862a02 7163 pr_err("The option-rom requires all member disks to be a member of all volumes\n");
5fe62b94 7164 return 0;
c2c087e6 7165 }
0dcecb2e
DW		 7166
7167 /* retrieve the largest free space block */
c2c087e6
DW		 7168 e = get_extents(super, dl);
7169 maxsize = 0;
7170 i = 0;
0dcecb2e
DW		 7171 if (e) {
7172 do {
7173 unsigned long long esize;
7174
7175 esize = e[i].start - pos;
7176 if (esize >= maxsize)
7177 maxsize = esize;
7178 pos = e[i].start + e[i].size;
7179 i++;
7180 } while (e[i-1].size);
7181 dl->e = e;
7182 dl->extent_cnt = i;
7183 } else {
7184 if (verbose)
e7b84f9d 7185 pr_err("unable to determine free space for: %s\n",
0dcecb2e
DW		 7186 dev);
7187 return 0;
7188 }
7189 if (maxsize < size) {
7190 if (verbose)
e7b84f9d 7191 pr_err("%s not enough space (%llu < %llu)\n",
0dcecb2e
DW		 7192 dev, maxsize, size);
7193 return 0;
7194 }
7195
7196 /* count total number of extents for merge */
7197 i = 0;
7198 for (dl = super->disks; dl; dl = dl->next)
7199 if (dl->e)
7200 i += dl->extent_cnt;
7201
7202 maxsize = merge_extents(super, i);
3baa56ab
LO		 7203
7204 if (!check_env("IMSM_NO_PLATFORM") &&
7205 mpb->num_raid_devs > 0 && size && size != maxsize) {
7a862a02 7206 pr_err("attempting to create a second volume with size less then remaining space. Aborting...\n");
3baa56ab
LO		 7207 return 0;
7208 }
7209
a7dd165b 7210 if (maxsize < size || maxsize == 0) {
b3071342
LD		 7211 if (verbose) {
7212 if (maxsize == 0)
7a862a02 7213 pr_err("no free space left on device. Aborting...\n");
b3071342 7214 else
7a862a02 7215 pr_err("not enough space to create volume of given size (%llu < %llu). Aborting...\n",
b3071342
LD		 7216 maxsize, size);
7217 }
0dcecb2e 7218 return 0;
0dcecb2e
DW		 7219 }
7220
c2c087e6
DW		 7221 *freesize = maxsize;
7222
ca9de185 7223 if (super->orom) {
72a45777 7224 int count = count_volumes(super->hba,
ca9de185
LM		 7225 super->orom->dpa, verbose);
7226 if (super->orom->vphba <= count) {
676e87a8 7227 pr_vrb("platform does not support more than %d raid volumes.\n",
ca9de185
LM		 7228 super->orom->vphba);
7229 return 0;
7230 }
7231 }
c2c087e6 7232 return 1;
cdddbdbc
DW		 7233}
7234
13bcac90 7235static int imsm_get_free_size(struct supertype *st, int raiddisks,
efb30e7f
DW		 7236 unsigned long long size, int chunk,
7237 unsigned long long *freesize)
7238{
7239 struct intel_super *super = st->sb;
7240 struct imsm_super *mpb = super->anchor;
7241 struct dl *dl;
7242 int i;
7243 int extent_cnt;
7244 struct extent *e;
7245 unsigned long long maxsize;
7246 unsigned long long minsize;
7247 int cnt;
7248 int used;
7249
7250 /* find the largest common start free region of the possible disks */
7251 used = 0;
7252 extent_cnt = 0;
7253 cnt = 0;
7254 for (dl = super->disks; dl; dl = dl->next) {
7255 dl->raiddisk = -1;
7256
7257 if (dl->index >= 0)
7258 used++;
7259
7260 /* don't activate new spares if we are orom constrained
7261 * and there is already a volume active in the container
7262 */
7263 if (super->orom && dl->index < 0 && mpb->num_raid_devs)
7264 continue;
7265
7266 e = get_extents(super, dl);
7267 if (!e)
7268 continue;
7269 for (i = 1; e[i-1].size; i++)
7270 ;
7271 dl->e = e;
7272 dl->extent_cnt = i;
7273 extent_cnt += i;
7274 cnt++;
7275 }
7276
7277 maxsize = merge_extents(super, extent_cnt);
7278 minsize = size;
7279 if (size == 0)
612e59d8
CA		 7280 /* chunk is in K */
7281 minsize = chunk * 2;
efb30e7f
DW		 7282
7283 if (cnt < raiddisks ||
7284 (super->orom && used && used != raiddisks) ||
a7dd165b
DW		 7285 maxsize < minsize ||
7286 maxsize == 0) {
e7b84f9d 7287 pr_err("not enough devices with space to create array.\n");
efb30e7f
DW		 7288 return 0; /* No enough free spaces large enough */
7289 }
7290
7291 if (size == 0) {
7292 size = maxsize;
7293 if (chunk) {
612e59d8
CA		 7294 size /= 2 * chunk;
7295 size *= 2 * chunk;
efb30e7f 7296 }
f878b242
LM		 7297 maxsize = size;
7298 }
7299 if (!check_env("IMSM_NO_PLATFORM") &&
7300 mpb->num_raid_devs > 0 && size && size != maxsize) {
7a862a02 7301 pr_err("attempting to create a second volume with size less then remaining space. Aborting...\n");
f878b242 7302 return 0;
efb30e7f 7303 }
efb30e7f
DW		 7304 cnt = 0;
7305 for (dl = super->disks; dl; dl = dl->next)
7306 if (dl->e)
7307 dl->raiddisk = cnt++;
7308
7309 *freesize = size;
7310
13bcac90
AK		 7311 dprintf("imsm: imsm_get_free_size() returns : %llu\n", size);
7312
efb30e7f
DW		 7313 return 1;
7314}
7315
13bcac90
AK		 7316static int reserve_space(struct supertype *st, int raiddisks,
7317 unsigned long long size, int chunk,
7318 unsigned long long *freesize)
7319{
7320 struct intel_super *super = st->sb;
7321 struct dl *dl;
7322 int cnt;
7323 int rv = 0;
7324
7325 rv = imsm_get_free_size(st, raiddisks, size, chunk, freesize);
7326 if (rv) {
7327 cnt = 0;
7328 for (dl = super->disks; dl; dl = dl->next)
7329 if (dl->e)
7330 dl->raiddisk = cnt++;
7331 rv = 1;
7332 }
7333
7334 return rv;
7335}
7336
bf5a934a 7337static int validate_geometry_imsm(struct supertype *st, int level, int layout,
c21e737b 7338 int raiddisks, int *chunk, unsigned long long size,
af4348dd 7339 unsigned long long data_offset,
bf5a934a 7340 char *dev, unsigned long long *freesize,
5308f117 7341 int consistency_policy, int verbose)
bf5a934a
DW		 7342{
7343 int fd, cfd;
7344 struct mdinfo *sra;
20cbe8d2 7345 int is_member = 0;
bf5a934a 7346
d54559f0
LM		 7347 /* load capability
7348 * if given unused devices create a container
bf5a934a
DW		 7349 * if given given devices in a container create a member volume
7350 */
7351 if (level == LEVEL_CONTAINER) {
7352 /* Must be a fresh device to add to a container */
7353 return validate_geometry_imsm_container(st, level, layout,
c21e737b 7354 raiddisks,
7ccc4cc4 7355 *chunk,
af4348dd 7356 size, data_offset,
bf5a934a
DW		 7357 dev, freesize,
7358 verbose);
7359 }
9587c373 7360
54865c30
RS		 7361 if (size && ((size < 1024) || (*chunk != UnSet &&
7362 size < (unsigned long long) *chunk))) {
7363 pr_err("Given size must be greater than 1M and chunk size.\n");
7364 /* Depends on algorithm in Create.c :
7365 * if container was given (dev == NULL) return -1,
7366 * if block device was given ( dev != NULL) return 0.
7367 */
7368 return dev ? -1 : 0;
7369 }
7370
8592f29d 7371 if (!dev) {
e91a3bad 7372 if (st->sb) {
ca9de185 7373 struct intel_super *super = st->sb;
e91a3bad 7374 if (!validate_geometry_imsm_orom(st->sb, level, layout,
2cc699af 7375 raiddisks, chunk, size,
e91a3bad
LM		 7376 verbose))
7377 return 0;
efb30e7f
DW		 7378 /* we are being asked to automatically layout a
7379 * new volume based on the current contents of
7380 * the container. If the the parameters can be
7381 * satisfied reserve_space will record the disks,
7382 * start offset, and size of the volume to be
7383 * created. add_to_super and getinfo_super
7384 * detect when autolayout is in progress.
7385 */
ca9de185
LM		 7386 /* assuming that freesize is always given when array is
7387 created */
7388 if (super->orom && freesize) {
7389 int count;
72a45777 7390 count = count_volumes(super->hba,
ca9de185
LM		 7391 super->orom->dpa, verbose);
7392 if (super->orom->vphba <= count) {
676e87a8 7393 pr_vrb("platform does not support more than %d raid volumes.\n",
ca9de185
LM		 7394 super->orom->vphba);
7395 return 0;
7396 }
7397 }
e91a3bad
LM		 7398 if (freesize)
7399 return reserve_space(st, raiddisks, size,
7ccc4cc4 7400 *chunk, freesize);
8592f29d
N		 7401 }
7402 return 1;
7403 }
bf5a934a
DW		 7404 if (st->sb) {
7405 /* creating in a given container */
7406 return validate_geometry_imsm_volume(st, level, layout,
7407 raiddisks, chunk, size,
af4348dd 7408 data_offset,
bf5a934a
DW		 7409 dev, freesize, verbose);
7410 }
7411
bf5a934a
DW		 7412 /* This device needs to be a device in an 'imsm' container */
7413 fd = open(dev, O_RDONLY|O_EXCL, 0);
7414 if (fd >= 0) {
7415 if (verbose)
e7b84f9d
N		 7416 pr_err("Cannot create this array on device %s\n",
7417 dev);
bf5a934a
DW		 7418 close(fd);
7419 return 0;
7420 }
7421 if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) {
7422 if (verbose)
e7b84f9d 7423 pr_err("Cannot open %s: %s\n",
bf5a934a
DW		 7424 dev, strerror(errno));
7425 return 0;
7426 }
7427 /* Well, it is in use by someone, maybe an 'imsm' container. */
7428 cfd = open_container(fd);
20cbe8d2 7429 close(fd);
bf5a934a 7430 if (cfd < 0) {
bf5a934a 7431 if (verbose)
e7b84f9d 7432 pr_err("Cannot use %s: It is busy\n",
bf5a934a
DW		 7433 dev);
7434 return 0;
7435 }
4dd2df09 7436 sra = sysfs_read(cfd, NULL, GET_VERSION);
bf5a934a 7437 if (sra && sra->array.major_version == -1 &&
20cbe8d2
AW		 7438 strcmp(sra->text_version, "imsm") == 0)
7439 is_member = 1;
7440 sysfs_free(sra);
7441 if (is_member) {
bf5a934a
DW		 7442 /* This is a member of a imsm container. Load the container
7443 * and try to create a volume
7444 */
7445 struct intel_super *super;
7446
ec50f7b6 7447 if (load_super_imsm_all(st, cfd, (void **) &super, NULL, NULL, 1) == 0) {
bf5a934a 7448 st->sb = super;
4dd2df09 7449 strcpy(st->container_devnm, fd2devnm(cfd));
bf5a934a
DW		 7450 close(cfd);
7451 return validate_geometry_imsm_volume(st, level, layout,
7452 raiddisks, chunk,
af4348dd 7453 size, data_offset, dev,
ecbd9e81
N		 7454 freesize, 1)
7455 ? 1 : -1;
bf5a934a 7456 }
20cbe8d2 7457 }
bf5a934a 7458
20cbe8d2 7459 if (verbose)
e7b84f9d 7460 pr_err("failed container membership check\n");
20cbe8d2
AW		 7461
7462 close(cfd);
7463 return 0;
bf5a934a 7464}
0bd16cf2 7465
30f58b22 7466static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk)
0bd16cf2
DJ		 7467{
7468 struct intel_super *super = st->sb;
7469
30f58b22
DW		 7470 if (level && *level == UnSet)
7471 *level = LEVEL_CONTAINER;
7472
7473 if (level && layout && *layout == UnSet)
7474 *layout = imsm_level_to_layout(*level);
0bd16cf2 7475
cd9d1ac7
DW		 7476 if (chunk && (*chunk == UnSet || *chunk == 0))
7477 *chunk = imsm_default_chunk(super->orom);
0bd16cf2
DJ		 7478}
7479
33414a01
DW		 7480static void handle_missing(struct intel_super *super, struct imsm_dev *dev);
7481
7482static int kill_subarray_imsm(struct supertype *st)
7483{
7484 /* remove the subarray currently referenced by ->current_vol */
7485 __u8 i;
7486 struct intel_dev **dp;
7487 struct intel_super *super = st->sb;
7488 __u8 current_vol = super->current_vol;
7489 struct imsm_super *mpb = super->anchor;
7490
7491 if (super->current_vol < 0)
7492 return 2;
7493 super->current_vol = -1; /* invalidate subarray cursor */
7494
7495 /* block deletions that would change the uuid of active subarrays
7496 *
7497 * FIXME when immutable ids are available, but note that we'll
7498 * also need to fixup the invalidated/active subarray indexes in
7499 * mdstat
7500 */
7501 for (i = 0; i < mpb->num_raid_devs; i++) {
7502 char subarray[4];
7503
7504 if (i < current_vol)
7505 continue;
7506 sprintf(subarray, "%u", i);
4dd2df09 7507 if (is_subarray_active(subarray, st->devnm)) {
e7b84f9d
N		 7508 pr_err("deleting subarray-%d would change the UUID of active subarray-%d, aborting\n",
7509 current_vol, i);
33414a01
DW		 7510
7511 return 2;
7512 }
7513 }
7514
7515 if (st->update_tail) {
503975b9 7516 struct imsm_update_kill_array *u = xmalloc(sizeof(*u));
33414a01 7517
33414a01
DW		 7518 u->type = update_kill_array;
7519 u->dev_idx = current_vol;
7520 append_metadata_update(st, u, sizeof(*u));
7521
7522 return 0;
7523 }
7524
7525 for (dp = &super->devlist; *dp;)
7526 if ((*dp)->index == current_vol) {
7527 *dp = (*dp)->next;
7528 } else {
7529 handle_missing(super, (*dp)->dev);
7530 if ((*dp)->index > current_vol)
7531 (*dp)->index--;
7532 dp = &(*dp)->next;
7533 }
7534
7535 /* no more raid devices, all active components are now spares,
7536 * but of course failed are still failed
7537 */
7538 if (--mpb->num_raid_devs == 0) {
7539 struct dl *d;
7540
7541 for (d = super->disks; d; d = d->next)
a8619d23
AK		 7542 if (d->index > -2)
7543 mark_spare(d);
33414a01
DW		 7544 }
7545
7546 super->updates_pending++;
7547
7548 return 0;
7549}
aa534678 7550
a951a4f7 7551static int update_subarray_imsm(struct supertype *st, char *subarray,
fa56eddb 7552 char *update, struct mddev_ident *ident)
aa534678
DW		 7553{
7554 /* update the subarray currently referenced by ->current_vol */
7555 struct intel_super *super = st->sb;
7556 struct imsm_super *mpb = super->anchor;
7557
aa534678
DW		 7558 if (strcmp(update, "name") == 0) {
7559 char *name = ident->name;
a951a4f7
N		 7560 char *ep;
7561 int vol;
aa534678 7562
4dd2df09 7563 if (is_subarray_active(subarray, st->devnm)) {
e7b84f9d 7564 pr_err("Unable to update name of active subarray\n");
aa534678
DW		 7565 return 2;
7566 }
7567
7568 if (!check_name(super, name, 0))
7569 return 2;
7570
a951a4f7
N		 7571 vol = strtoul(subarray, &ep, 10);
7572 if (*ep != '\0' || vol >= super->anchor->num_raid_devs)
7573 return 2;
7574
aa534678 7575 if (st->update_tail) {
503975b9 7576 struct imsm_update_rename_array *u = xmalloc(sizeof(*u));
aa534678 7577
aa534678 7578 u->type = update_rename_array;
a951a4f7 7579 u->dev_idx = vol;
618f4e6d
XN		 7580 strncpy((char *) u->name, name, MAX_RAID_SERIAL_LEN);
7581 u->name[MAX_RAID_SERIAL_LEN-1] = '\0';
aa534678
DW		 7582 append_metadata_update(st, u, sizeof(*u));
7583 } else {
7584 struct imsm_dev *dev;
7585 int i;
7586
a951a4f7 7587 dev = get_imsm_dev(super, vol);
618f4e6d
XN		 7588 strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN);
7589 dev->volume[MAX_RAID_SERIAL_LEN-1] = '\0';
aa534678
DW		 7590 for (i = 0; i < mpb->num_raid_devs; i++) {
7591 dev = get_imsm_dev(super, i);
7592 handle_missing(super, dev);
7593 }
7594 super->updates_pending++;
7595 }
e6e9dd3f
AP		 7596 } else if (strcmp(update, "ppl") == 0 ||
7597 strcmp(update, "no-ppl") == 0) {
7598 int new_policy;
7599 char *ep;
7600 int vol = strtoul(subarray, &ep, 10);
7601
7602 if (*ep != '\0' || vol >= super->anchor->num_raid_devs)
7603 return 2;
7604
7605 if (strcmp(update, "ppl") == 0)
c2462068 7606 new_policy = RWH_MULTIPLE_DISTRIBUTED;
e6e9dd3f 7607 else
c2462068 7608 new_policy = RWH_MULTIPLE_OFF;
e6e9dd3f
AP		 7609
7610 if (st->update_tail) {
7611 struct imsm_update_rwh_policy *u = xmalloc(sizeof(*u));
7612
7613 u->type = update_rwh_policy;
7614 u->dev_idx = vol;
7615 u->new_policy = new_policy;
7616 append_metadata_update(st, u, sizeof(*u));
7617 } else {
7618 struct imsm_dev *dev;
7619
7620 dev = get_imsm_dev(super, vol);
7621 dev->rwh_policy = new_policy;
7622 super->updates_pending++;
7623 }
aa534678
DW		 7624 } else
7625 return 2;
7626
7627 return 0;
7628}
bf5a934a 7629
28bce06f
AK		 7630static int is_gen_migration(struct imsm_dev *dev)
7631{
7534230b
AK		 7632 if (dev == NULL)
7633 return 0;
7634
28bce06f
AK		 7635 if (!dev->vol.migr_state)
7636 return 0;
7637
7638 if (migr_type(dev) == MIGR_GEN_MIGR)
7639 return 1;
7640
7641 return 0;
7642}
7643
1e5c6983
DW		 7644static int is_rebuilding(struct imsm_dev *dev)
7645{
7646 struct imsm_map *migr_map;
7647
7648 if (!dev->vol.migr_state)
7649 return 0;
7650
7651 if (migr_type(dev) != MIGR_REBUILD)
7652 return 0;
7653
238c0a71 7654 migr_map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 7655
7656 if (migr_map->map_state == IMSM_T_STATE_DEGRADED)
7657 return 1;
7658 else
7659 return 0;
7660}
7661
6ce1fbf1
AK		 7662static int is_initializing(struct imsm_dev *dev)
7663{
7664 struct imsm_map *migr_map;
7665
7666 if (!dev->vol.migr_state)
7667 return 0;
7668
7669 if (migr_type(dev) != MIGR_INIT)
7670 return 0;
7671
238c0a71 7672 migr_map = get_imsm_map(dev, MAP_1);
6ce1fbf1
AK		 7673
7674 if (migr_map->map_state == IMSM_T_STATE_UNINITIALIZED)
7675 return 1;
7676
7677 return 0;
6ce1fbf1
AK		 7678}
7679
c47b0ff6
AK		 7680static void update_recovery_start(struct intel_super *super,
7681 struct imsm_dev *dev,
7682 struct mdinfo *array)
1e5c6983
DW		 7683{
7684 struct mdinfo *rebuild = NULL;
7685 struct mdinfo *d;
7686 __u32 units;
7687
7688 if (!is_rebuilding(dev))
7689 return;
7690
7691 /* Find the rebuild target, but punt on the dual rebuild case */
7692 for (d = array->devs; d; d = d->next)
7693 if (d->recovery_start == 0) {
7694 if (rebuild)
7695 return;
7696 rebuild = d;
7697 }
7698
4363fd80
DW		 7699 if (!rebuild) {
7700 /* (?) none of the disks are marked with
7701 * IMSM_ORD_REBUILD, so assume they are missing and the
7702 * disk_ord_tbl was not correctly updated
7703 */
1ade5cc1 7704 dprintf("failed to locate out-of-sync disk\n");
4363fd80
DW		 7705 return;
7706 }
7707
1e5c6983 7708 units = __le32_to_cpu(dev->vol.curr_migr_unit);
c47b0ff6 7709 rebuild->recovery_start = units * blocks_per_migr_unit(super, dev);
1e5c6983
DW		 7710}
7711
276d77db 7712static int recover_backup_imsm(struct supertype *st, struct mdinfo *info);
1e5c6983 7713
00bbdbda 7714static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray)
cdddbdbc 7715{
4f5bc454
DW		 7716 /* Given a container loaded by load_super_imsm_all,
7717 * extract information about all the arrays into
7718 * an mdinfo tree.
00bbdbda 7719 * If 'subarray' is given, just extract info about that array.
4f5bc454
DW		 7720 *
7721 * For each imsm_dev create an mdinfo, fill it in,
7722 * then look for matching devices in super->disks
7723 * and create appropriate device mdinfo.
7724 */
7725 struct intel_super *super = st->sb;
949c47a0 7726 struct imsm_super *mpb = super->anchor;
4f5bc454 7727 struct mdinfo *rest = NULL;
00bbdbda 7728 unsigned int i;
81219e70 7729 int sb_errors = 0;
abef11a3
AK		 7730 struct dl *d;
7731 int spare_disks = 0;
cdddbdbc 7732
19482bcc
AK		 7733 /* do not assemble arrays when not all attributes are supported */
7734 if (imsm_check_attributes(mpb->attributes) == 0) {
81219e70 7735 sb_errors = 1;
7a862a02 7736 pr_err("Unsupported attributes in IMSM metadata.Arrays activation is blocked.\n");
19482bcc
AK		 7737 }
7738
abef11a3
AK		 7739 /* count spare devices, not used in maps
7740 */
7741 for (d = super->disks; d; d = d->next)
7742 if (d->index == -1)
7743 spare_disks++;
7744
4f5bc454 7745 for (i = 0; i < mpb->num_raid_devs; i++) {
00bbdbda
N		 7746 struct imsm_dev *dev;
7747 struct imsm_map *map;
86e3692b 7748 struct imsm_map *map2;
4f5bc454 7749 struct mdinfo *this;
a6482415 7750 int slot;
a6482415 7751 int chunk;
00bbdbda 7752 char *ep;
8b9cd157 7753 int level;
00bbdbda
N		 7754
7755 if (subarray &&
7756 (i != strtoul(subarray, &ep, 10) || *ep != '\0'))
7757 continue;
7758
7759 dev = get_imsm_dev(super, i);
238c0a71
AK		 7760 map = get_imsm_map(dev, MAP_0);
7761 map2 = get_imsm_map(dev, MAP_1);
8b9cd157 7762 level = get_imsm_raid_level(map);
4f5bc454 7763
1ce0101c
DW		 7764 /* do not publish arrays that are in the middle of an
7765 * unsupported migration
7766 */
7767 if (dev->vol.migr_state &&
28bce06f 7768 (migr_type(dev) == MIGR_STATE_CHANGE)) {
7a862a02 7769 pr_err("cannot assemble volume '%.16s': unsupported migration in progress\n",
1ce0101c
DW		 7770 dev->volume);
7771 continue;
7772 }
2db86302
LM		 7773 /* do not publish arrays that are not support by controller's
7774 * OROM/EFI
7775 */
1ce0101c 7776
503975b9 7777 this = xmalloc(sizeof(*this));
4f5bc454 7778
301406c9 7779 super->current_vol = i;
a5d85af7 7780 getinfo_super_imsm_volume(st, this, NULL);
9894ec0d 7781 this->next = rest;
a6482415 7782 chunk = __le16_to_cpu(map->blocks_per_strip) >> 1;
81219e70
LM		 7783 /* mdadm does not support all metadata features- set the bit in all arrays state */
7784 if (!validate_geometry_imsm_orom(super,
8b9cd157
MK		 7785 level, /* RAID level */
7786 imsm_level_to_layout(level),
81219e70 7787 map->num_members, /* raid disks */
fcc2c9da 7788 &chunk, imsm_dev_size(dev),
81219e70 7789 1 /* verbose */)) {
7a862a02 7790 pr_err("IMSM RAID geometry validation failed. Array %s activation is blocked.\n",
81219e70
LM		 7791 dev->volume);
7792 this->array.state |=
7793 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
7794 (1<<MD_SB_BLOCK_VOLUME);
7795 }
81219e70
LM		 7796
7797 /* if array has bad blocks, set suitable bit in all arrays state */
7798 if (sb_errors)
7799 this->array.state |=
7800 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
7801 (1<<MD_SB_BLOCK_VOLUME);
7802
4f5bc454 7803 for (slot = 0 ; slot < map->num_members; slot++) {
1e5c6983 7804 unsigned long long recovery_start;
4f5bc454
DW		 7805 struct mdinfo *info_d;
7806 struct dl *d;
7807 int idx;
9a1608e5 7808 int skip;
7eef0453 7809 __u32 ord;
8b9cd157 7810 int missing = 0;
4f5bc454 7811
9a1608e5 7812 skip = 0;
238c0a71
AK		 7813 idx = get_imsm_disk_idx(dev, slot, MAP_0);
7814 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
4f5bc454
DW		 7815 for (d = super->disks; d ; d = d->next)
7816 if (d->index == idx)
0fbd635c 7817 break;
4f5bc454 7818
1e5c6983 7819 recovery_start = MaxSector;
4f5bc454 7820 if (d == NULL)
9a1608e5 7821 skip = 1;
25ed7e59 7822 if (d && is_failed(&d->disk))
9a1608e5 7823 skip = 1;
8b9cd157 7824 if (!skip && (ord & IMSM_ORD_REBUILD))
1e5c6983 7825 recovery_start = 0;
9a1608e5 7826
1011e834 7827 /*
9a1608e5 7828 * if we skip some disks the array will be assmebled degraded;
1e5c6983
DW		 7829 * reset resync start to avoid a dirty-degraded
7830 * situation when performing the intial sync
9a1608e5 7831 */
8b9cd157
MK		 7832 if (skip)
7833 missing++;
7834
7835 if (!(dev->vol.dirty & RAIDVOL_DIRTY)) {
7836 if ((!able_to_resync(level, missing) ||
7837 recovery_start == 0))
7838 this->resync_start = MaxSector;
7839 } else {
7840 /*
7841 * FIXME handle dirty degraded
7842 */
7843 }
7844
9a1608e5
DW		 7845 if (skip)
7846 continue;
4f5bc454 7847
503975b9 7848 info_d = xcalloc(1, sizeof(*info_d));
4f5bc454
DW		 7849 info_d->next = this->devs;
7850 this->devs = info_d;
7851
4f5bc454
DW		 7852 info_d->disk.number = d->index;
7853 info_d->disk.major = d->major;
7854 info_d->disk.minor = d->minor;
7855 info_d->disk.raid_disk = slot;
1e5c6983 7856 info_d->recovery_start = recovery_start;
86e3692b
AK		 7857 if (map2) {
7858 if (slot < map2->num_members)
7859 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK		 7860 else
7861 this->array.spare_disks++;
86e3692b
AK		 7862 } else {
7863 if (slot < map->num_members)
7864 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK		 7865 else
7866 this->array.spare_disks++;
86e3692b 7867 }
1e5c6983
DW		 7868 if (info_d->recovery_start == MaxSector)
7869 this->array.working_disks++;
4f5bc454
DW		 7870
7871 info_d->events = __le32_to_cpu(mpb->generation_num);
5551b113 7872 info_d->data_offset = pba_of_lba0(map);
44490938 7873 info_d->component_size = calc_component_size(map, dev);
06fb291a
PB		 7874
7875 if (map->raid_level == 5) {
2432ce9b
AP		 7876 info_d->ppl_sector = this->ppl_sector;
7877 info_d->ppl_size = this->ppl_size;
98e96bdb
AP		 7878 if (this->consistency_policy == CONSISTENCY_POLICY_PPL &&
7879 recovery_start == 0)
7880 this->resync_start = 0;
06fb291a 7881 }
b12796be 7882
5e46202e 7883 info_d->bb.supported = 1;
b12796be
TM		 7884 get_volume_badblocks(super->bbm_log, ord_to_idx(ord),
7885 info_d->data_offset,
7886 info_d->component_size,
7887 &info_d->bb);
4f5bc454 7888 }
1e5c6983 7889 /* now that the disk list is up-to-date fixup recovery_start */
c47b0ff6 7890 update_recovery_start(super, dev, this);
abef11a3 7891 this->array.spare_disks += spare_disks;
276d77db
AK		 7892
7893 /* check for reshape */
7894 if (this->reshape_active == 1)
7895 recover_backup_imsm(st, this);
9a1608e5 7896 rest = this;
4f5bc454
DW		 7897 }
7898
7899 return rest;
cdddbdbc
DW		 7900}
7901
3b451610
AK		 7902static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev,
7903 int failed, int look_in_map)
c2a1e7da 7904{
3b451610
AK		 7905 struct imsm_map *map;
7906
7907 map = get_imsm_map(dev, look_in_map);
c2a1e7da
DW		 7908
7909 if (!failed)
1011e834 7910 return map->map_state == IMSM_T_STATE_UNINITIALIZED ?
3393c6af 7911 IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL;
c2a1e7da
DW		 7912
7913 switch (get_imsm_raid_level(map)) {
7914 case 0:
7915 return IMSM_T_STATE_FAILED;
7916 break;
7917 case 1:
7918 if (failed < map->num_members)
7919 return IMSM_T_STATE_DEGRADED;
7920 else
7921 return IMSM_T_STATE_FAILED;
7922 break;
7923 case 10:
7924 {
7925 /**
c92a2527
DW		 7926 * check to see if any mirrors have failed, otherwise we
7927 * are degraded. Even numbered slots are mirrored on
7928 * slot+1
c2a1e7da 7929 */
c2a1e7da 7930 int i;
d9b420a5
N		 7931 /* gcc -Os complains that this is unused */
7932 int insync = insync;
c2a1e7da
DW		 7933
7934 for (i = 0; i < map->num_members; i++) {
238c0a71 7935 __u32 ord = get_imsm_ord_tbl_ent(dev, i, MAP_X);
c92a2527
DW		 7936 int idx = ord_to_idx(ord);
7937 struct imsm_disk *disk;
c2a1e7da 7938
c92a2527 7939 /* reset the potential in-sync count on even-numbered
1011e834 7940 * slots. num_copies is always 2 for imsm raid10
c92a2527
DW		 7941 */
7942 if ((i & 1) == 0)
7943 insync = 2;
c2a1e7da 7944
c92a2527 7945 disk = get_imsm_disk(super, idx);
25ed7e59 7946 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
c92a2527 7947 insync--;
c2a1e7da 7948
c92a2527
DW		 7949 /* no in-sync disks left in this mirror the
7950 * array has failed
7951 */
7952 if (insync == 0)
7953 return IMSM_T_STATE_FAILED;
c2a1e7da
DW		 7954 }
7955
7956 return IMSM_T_STATE_DEGRADED;
7957 }
7958 case 5:
7959 if (failed < 2)
7960 return IMSM_T_STATE_DEGRADED;
7961 else
7962 return IMSM_T_STATE_FAILED;
7963 break;
7964 default:
7965 break;
7966 }
7967
7968 return map->map_state;
7969}
7970
3b451610
AK		 7971static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev,
7972 int look_in_map)
c2a1e7da
DW		 7973{
7974 int i;
7975 int failed = 0;
7976 struct imsm_disk *disk;
d5985138
AK		 7977 struct imsm_map *map = get_imsm_map(dev, MAP_0);
7978 struct imsm_map *prev = get_imsm_map(dev, MAP_1);
68fe4598 7979 struct imsm_map *map_for_loop;
0556e1a2
DW		 7980 __u32 ord;
7981 int idx;
d5985138 7982 int idx_1;
c2a1e7da 7983
0556e1a2
DW		 7984 /* at the beginning of migration we set IMSM_ORD_REBUILD on
7985 * disks that are being rebuilt. New failures are recorded to
7986 * map[0]. So we look through all the disks we started with and
7987 * see if any failures are still present, or if any new ones
7988 * have arrived
0556e1a2 7989 */
d5985138
AK		 7990 map_for_loop = map;
7991 if (prev && (map->num_members < prev->num_members))
7992 map_for_loop = prev;
68fe4598
LD		 7993
7994 for (i = 0; i < map_for_loop->num_members; i++) {
d5985138 7995 idx_1 = -255;
238c0a71
AK		 7996 /* when MAP_X is passed both maps failures are counted
7997 */
d5985138 7998 if (prev &&
089f9d79
JS		 7999 (look_in_map == MAP_1 || look_in_map == MAP_X) &&
8000 i < prev->num_members) {
d5985138
AK		 8001 ord = __le32_to_cpu(prev->disk_ord_tbl[i]);
8002 idx_1 = ord_to_idx(ord);
c2a1e7da 8003
d5985138
AK		 8004 disk = get_imsm_disk(super, idx_1);
8005 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
8006 failed++;
8007 }
089f9d79
JS		 8008 if ((look_in_map == MAP_0 || look_in_map == MAP_X) &&
8009 i < map->num_members) {
d5985138
AK		 8010 ord = __le32_to_cpu(map->disk_ord_tbl[i]);
8011 idx = ord_to_idx(ord);
8012
8013 if (idx != idx_1) {
8014 disk = get_imsm_disk(super, idx);
8015 if (!disk || is_failed(disk) ||
8016 ord & IMSM_ORD_REBUILD)
8017 failed++;
8018 }
8019 }
c2a1e7da
DW		 8020 }
8021
8022 return failed;
845dea95
NB		 8023}
8024
97b4d0e9
DW		 8025static int imsm_open_new(struct supertype *c, struct active_array *a,
8026 char *inst)
8027{
8028 struct intel_super *super = c->sb;
8029 struct imsm_super *mpb = super->anchor;
bbab0940 8030 struct imsm_update_prealloc_bb_mem u;
9587c373 8031
97b4d0e9 8032 if (atoi(inst) >= mpb->num_raid_devs) {
1ade5cc1 8033 pr_err("subarry index %d, out of range\n", atoi(inst));
97b4d0e9
DW		 8034 return -ENODEV;
8035 }
8036
8037 dprintf("imsm: open_new %s\n", inst);
8038 a->info.container_member = atoi(inst);
bbab0940
TM		 8039
8040 u.type = update_prealloc_badblocks_mem;
8041 imsm_update_metadata_locally(c, &u, sizeof(u));
8042
97b4d0e9
DW		 8043 return 0;
8044}
8045
0c046afd
DW		 8046static int is_resyncing(struct imsm_dev *dev)
8047{
8048 struct imsm_map *migr_map;
8049
8050 if (!dev->vol.migr_state)
8051 return 0;
8052
1484e727
DW		 8053 if (migr_type(dev) == MIGR_INIT ||
8054 migr_type(dev) == MIGR_REPAIR)
0c046afd
DW		 8055 return 1;
8056
4c9bc37b
AK		 8057 if (migr_type(dev) == MIGR_GEN_MIGR)
8058 return 0;
8059
238c0a71 8060 migr_map = get_imsm_map(dev, MAP_1);
0c046afd 8061
089f9d79
JS		 8062 if (migr_map->map_state == IMSM_T_STATE_NORMAL &&
8063 dev->vol.migr_type != MIGR_GEN_MIGR)
0c046afd
DW		 8064 return 1;
8065 else
8066 return 0;
8067}
8068
0556e1a2 8069/* return true if we recorded new information */
4c9e8c1e
TM		 8070static int mark_failure(struct intel_super *super,
8071 struct imsm_dev *dev, struct imsm_disk *disk, int idx)
47ee5a45 8072{
0556e1a2
DW		 8073 __u32 ord;
8074 int slot;
8075 struct imsm_map *map;
86c54047
DW		 8076 char buf[MAX_RAID_SERIAL_LEN+3];
8077 unsigned int len, shift = 0;
0556e1a2
DW		 8078
8079 /* new failures are always set in map[0] */
238c0a71 8080 map = get_imsm_map(dev, MAP_0);
0556e1a2
DW		 8081
8082 slot = get_imsm_disk_slot(map, idx);
8083 if (slot < 0)
8084 return 0;
8085
8086 ord = __le32_to_cpu(map->disk_ord_tbl[slot]);
25ed7e59 8087 if (is_failed(disk) && (ord & IMSM_ORD_REBUILD))
0556e1a2
DW		 8088 return 0;
8089
7d0c5e24
LD		 8090 memcpy(buf, disk->serial, MAX_RAID_SERIAL_LEN);
8091 buf[MAX_RAID_SERIAL_LEN] = '\000';
8092 strcat(buf, ":0");
86c54047
DW		 8093 if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN)
8094 shift = len - MAX_RAID_SERIAL_LEN + 1;
8095 strncpy((char *)disk->serial, &buf[shift], MAX_RAID_SERIAL_LEN);
8096
f2f27e63 8097 disk->status |= FAILED_DISK;
0556e1a2 8098 set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD);
17788994
AK		 8099 /* mark failures in second map if second map exists and this disk
8100 * in this slot.
8101 * This is valid for migration, initialization and rebuild
8102 */
8103 if (dev->vol.migr_state) {
238c0a71 8104 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
0a108d63
AK		 8105 int slot2 = get_imsm_disk_slot(map2, idx);
8106
089f9d79 8107 if (slot2 < map2->num_members && slot2 >= 0)
0a108d63 8108 set_imsm_ord_tbl_ent(map2, slot2,
1ace8403
AK		 8109 idx | IMSM_ORD_REBUILD);
8110 }
f21e18ca 8111 if (map->failed_disk_num == 0xff)
0556e1a2 8112 map->failed_disk_num = slot;
4c9e8c1e
TM		 8113
8114 clear_disk_badblocks(super->bbm_log, ord_to_idx(ord));
8115
0556e1a2
DW		 8116 return 1;
8117}
8118
4c9e8c1e
TM		 8119static void mark_missing(struct intel_super *super,
8120 struct imsm_dev *dev, struct imsm_disk *disk, int idx)
0556e1a2 8121{
4c9e8c1e 8122 mark_failure(super, dev, disk, idx);
0556e1a2
DW		 8123
8124 if (disk->scsi_id == __cpu_to_le32(~(__u32)0))
8125 return;
8126
47ee5a45
DW		 8127 disk->scsi_id = __cpu_to_le32(~(__u32)0);
8128 memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1);
8129}
8130
33414a01
DW		 8131static void handle_missing(struct intel_super *super, struct imsm_dev *dev)
8132{
33414a01 8133 struct dl *dl;
33414a01
DW		 8134
8135 if (!super->missing)
8136 return;
33414a01 8137
79b68f1b
PC		 8138 /* When orom adds replacement for missing disk it does
8139 * not remove entry of missing disk, but just updates map with
8140 * new added disk. So it is not enough just to test if there is
8141 * any missing disk, we have to look if there are any failed disks
8142 * in map to stop migration */
8143
33414a01 8144 dprintf("imsm: mark missing\n");
3d59f0c0
AK		 8145 /* end process for initialization and rebuild only
8146 */
8147 if (is_gen_migration(dev) == 0) {
fb12a745 8148 int failed = imsm_count_failed(super, dev, MAP_0);
3d59f0c0 8149
fb12a745
TM		 8150 if (failed) {
8151 __u8 map_state;
8152 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8153 struct imsm_map *map1;
8154 int i, ord, ord_map1;
8155 int rebuilt = 1;
3d59f0c0 8156
fb12a745
TM		 8157 for (i = 0; i < map->num_members; i++) {
8158 ord = get_imsm_ord_tbl_ent(dev, i, MAP_0);
8159 if (!(ord & IMSM_ORD_REBUILD))
8160 continue;
8161
8162 map1 = get_imsm_map(dev, MAP_1);
8163 if (!map1)
8164 continue;
8165
8166 ord_map1 = __le32_to_cpu(map1->disk_ord_tbl[i]);
8167 if (ord_map1 & IMSM_ORD_REBUILD)
8168 rebuilt = 0;
8169 }
8170
8171 if (rebuilt) {
8172 map_state = imsm_check_degraded(super, dev,
8173 failed, MAP_0);
8174 end_migration(dev, super, map_state);
8175 }
8176 }
3d59f0c0 8177 }
33414a01 8178 for (dl = super->missing; dl; dl = dl->next)
4c9e8c1e 8179 mark_missing(super, dev, &dl->disk, dl->index);
33414a01
DW		 8180 super->updates_pending++;
8181}
8182
f3871fdc
AK		 8183static unsigned long long imsm_set_array_size(struct imsm_dev *dev,
8184 long long new_size)
70bdf0dc 8185{
70bdf0dc 8186 unsigned long long array_blocks;
9529d343
MD		 8187 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8188 int used_disks = imsm_num_data_members(map);
70bdf0dc
AK		 8189
8190 if (used_disks == 0) {
8191 /* when problems occures
8192 * return current array_blocks value
8193 */
fcc2c9da 8194 array_blocks = imsm_dev_size(dev);
70bdf0dc
AK		 8195
8196 return array_blocks;
8197 }
8198
8199 /* set array size in metadata
8200 */
9529d343 8201 if (new_size <= 0)
f3871fdc
AK		 8202 /* OLCE size change is caused by added disks
8203 */
44490938 8204 array_blocks = per_dev_array_size(map) * used_disks;
9529d343 8205 else
f3871fdc
AK		 8206 /* Online Volume Size Change
8207 * Using available free space
8208 */
8209 array_blocks = new_size;
70bdf0dc 8210
b53bfba6 8211 array_blocks = round_size_to_mb(array_blocks, used_disks);
fcc2c9da 8212 set_imsm_dev_size(dev, array_blocks);
70bdf0dc
AK		 8213
8214 return array_blocks;
8215}
8216
28bce06f
AK		 8217static void imsm_set_disk(struct active_array *a, int n, int state);
8218
0e2d1a4e
AK		 8219static void imsm_progress_container_reshape(struct intel_super *super)
8220{
8221 /* if no device has a migr_state, but some device has a
8222 * different number of members than the previous device, start
8223 * changing the number of devices in this device to match
8224 * previous.
8225 */
8226 struct imsm_super *mpb = super->anchor;
8227 int prev_disks = -1;
8228 int i;
1dfaa380 8229 int copy_map_size;
0e2d1a4e
AK		 8230
8231 for (i = 0; i < mpb->num_raid_devs; i++) {
8232 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 8233 struct imsm_map *map = get_imsm_map(dev, MAP_0);
0e2d1a4e
AK		 8234 struct imsm_map *map2;
8235 int prev_num_members;
0e2d1a4e
AK		 8236
8237 if (dev->vol.migr_state)
8238 return;
8239
8240 if (prev_disks == -1)
8241 prev_disks = map->num_members;
8242 if (prev_disks == map->num_members)
8243 continue;
8244
8245 /* OK, this array needs to enter reshape mode.
8246 * i.e it needs a migr_state
8247 */
8248
1dfaa380 8249 copy_map_size = sizeof_imsm_map(map);
0e2d1a4e
AK		 8250 prev_num_members = map->num_members;
8251 map->num_members = prev_disks;
8252 dev->vol.migr_state = 1;
8253 dev->vol.curr_migr_unit = 0;
ea672ee1 8254 set_migr_type(dev, MIGR_GEN_MIGR);
0e2d1a4e
AK		 8255 for (i = prev_num_members;
8256 i < map->num_members; i++)
8257 set_imsm_ord_tbl_ent(map, i, i);
238c0a71 8258 map2 = get_imsm_map(dev, MAP_1);
0e2d1a4e 8259 /* Copy the current map */
1dfaa380 8260 memcpy(map2, map, copy_map_size);
0e2d1a4e
AK		 8261 map2->num_members = prev_num_members;
8262
f3871fdc 8263 imsm_set_array_size(dev, -1);
51d83f5d 8264 super->clean_migration_record_by_mdmon = 1;
0e2d1a4e
AK		 8265 super->updates_pending++;
8266 }
8267}
8268
aad6f216 8269/* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild
0c046afd
DW		 8270 * states are handled in imsm_set_disk() with one exception, when a
8271 * resync is stopped due to a new failure this routine will set the
8272 * 'degraded' state for the array.
8273 */
01f157d7 8274static int imsm_set_array_state(struct active_array *a, int consistent)
a862209d
DW		 8275{
8276 int inst = a->info.container_member;
8277 struct intel_super *super = a->container->sb;
949c47a0 8278 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8279 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3b451610
AK		 8280 int failed = imsm_count_failed(super, dev, MAP_0);
8281 __u8 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
1e5c6983 8282 __u32 blocks_per_unit;
a862209d 8283
1af97990
AK		 8284 if (dev->vol.migr_state &&
8285 dev->vol.migr_type == MIGR_GEN_MIGR) {
8286 /* array state change is blocked due to reshape action
aad6f216
N		 8287 * We might need to
8288 * - abort the reshape (if last_checkpoint is 0 and action!= reshape)
8289 * - finish the reshape (if last_checkpoint is big and action != reshape)
8290 * - update curr_migr_unit
1af97990 8291 */
aad6f216
N		 8292 if (a->curr_action == reshape) {
8293 /* still reshaping, maybe update curr_migr_unit */
633b5610 8294 goto mark_checkpoint;
aad6f216
N		 8295 } else {
8296 if (a->last_checkpoint == 0 && a->prev_action == reshape) {
8297 /* for some reason we aborted the reshape.
b66e591b
AK		 8298 *
8299 * disable automatic metadata rollback
8300 * user action is required to recover process
aad6f216 8301 */
b66e591b 8302 if (0) {
238c0a71
AK		 8303 struct imsm_map *map2 =
8304 get_imsm_map(dev, MAP_1);
8305 dev->vol.migr_state = 0;
8306 set_migr_type(dev, 0);
8307 dev->vol.curr_migr_unit = 0;
8308 memcpy(map, map2,
8309 sizeof_imsm_map(map2));
8310 super->updates_pending++;
b66e591b 8311 }
aad6f216
N		 8312 }
8313 if (a->last_checkpoint >= a->info.component_size) {
8314 unsigned long long array_blocks;
8315 int used_disks;
e154ced3 8316 struct mdinfo *mdi;
aad6f216 8317
9529d343 8318 used_disks = imsm_num_data_members(map);
d55adef9
AK		 8319 if (used_disks > 0) {
8320 array_blocks =
44490938 8321 per_dev_array_size(map) *
d55adef9 8322 used_disks;
b53bfba6
TM		 8323 array_blocks =
8324 round_size_to_mb(array_blocks,
8325 used_disks);
d55adef9
AK		 8326 a->info.custom_array_size = array_blocks;
8327 /* encourage manager to update array
8328 * size
8329 */
e154ced3 8330
d55adef9 8331 a->check_reshape = 1;
633b5610 8332 }
e154ced3
AK		 8333 /* finalize online capacity expansion/reshape */
8334 for (mdi = a->info.devs; mdi; mdi = mdi->next)
8335 imsm_set_disk(a,
8336 mdi->disk.raid_disk,
8337 mdi->curr_state);
8338
0e2d1a4e 8339 imsm_progress_container_reshape(super);
e154ced3 8340 }
aad6f216 8341 }
1af97990
AK		 8342 }
8343
47ee5a45 8344 /* before we activate this array handle any missing disks */
33414a01
DW		 8345 if (consistent == 2)
8346 handle_missing(super, dev);
1e5c6983 8347
0c046afd 8348 if (consistent == 2 &&
b7941fd6 8349 (!is_resync_complete(&a->info) ||
0c046afd
DW		 8350 map_state != IMSM_T_STATE_NORMAL ||
8351 dev->vol.migr_state))
01f157d7 8352 consistent = 0;
272906ef 8353
b7941fd6 8354 if (is_resync_complete(&a->info)) {
0c046afd 8355 /* complete intialization / resync,
0556e1a2
DW		 8356 * recovery and interrupted recovery is completed in
8357 * ->set_disk
0c046afd
DW		 8358 */
8359 if (is_resyncing(dev)) {
8360 dprintf("imsm: mark resync done\n");
809da78e 8361 end_migration(dev, super, map_state);
115c3803 8362 super->updates_pending++;
484240d8 8363 a->last_checkpoint = 0;
115c3803 8364 }
b9172665
AK		 8365 } else if ((!is_resyncing(dev) && !failed) &&
8366 (imsm_reshape_blocks_arrays_changes(super) == 0)) {
0c046afd 8367 /* mark the start of the init process if nothing is failed */
b7941fd6 8368 dprintf("imsm: mark resync start\n");
1484e727 8369 if (map->map_state == IMSM_T_STATE_UNINITIALIZED)
8e59f3d8 8370 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT);
1484e727 8371 else
8e59f3d8 8372 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR);
3393c6af 8373 super->updates_pending++;
115c3803 8374 }
a862209d 8375
633b5610 8376mark_checkpoint:
5b83bacf
AK		 8377 /* skip checkpointing for general migration,
8378 * it is controlled in mdadm
8379 */
8380 if (is_gen_migration(dev))
8381 goto skip_mark_checkpoint;
8382
1e5c6983 8383 /* check if we can update curr_migr_unit from resync_start, recovery_start */
c47b0ff6 8384 blocks_per_unit = blocks_per_migr_unit(super, dev);
4f0a7acc 8385 if (blocks_per_unit) {
1e5c6983
DW		 8386 __u32 units32;
8387 __u64 units;
8388
4f0a7acc 8389 units = a->last_checkpoint / blocks_per_unit;
1e5c6983
DW		 8390 units32 = units;
8391
8392 /* check that we did not overflow 32-bits, and that
8393 * curr_migr_unit needs updating
8394 */
8395 if (units32 == units &&
bfd80a56 8396 units32 != 0 &&
1e5c6983
DW		 8397 __le32_to_cpu(dev->vol.curr_migr_unit) != units32) {
8398 dprintf("imsm: mark checkpoint (%u)\n", units32);
8399 dev->vol.curr_migr_unit = __cpu_to_le32(units32);
8400 super->updates_pending++;
8401 }
8402 }
f8f603f1 8403
5b83bacf 8404skip_mark_checkpoint:
3393c6af 8405 /* mark dirty / clean */
2432ce9b
AP		 8406 if (((dev->vol.dirty & RAIDVOL_DIRTY) && consistent) ||
8407 (!(dev->vol.dirty & RAIDVOL_DIRTY) && !consistent)) {
b7941fd6 8408 dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty");
2432ce9b
AP		 8409 if (consistent) {
8410 dev->vol.dirty = RAIDVOL_CLEAN;
8411 } else {
8412 dev->vol.dirty = RAIDVOL_DIRTY;
c2462068
PB		 8413 if (dev->rwh_policy == RWH_DISTRIBUTED ||
8414 dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
2432ce9b
AP		 8415 dev->vol.dirty |= RAIDVOL_DSRECORD_VALID;
8416 }
a862209d
DW		 8417 super->updates_pending++;
8418 }
28bce06f 8419
01f157d7 8420 return consistent;
a862209d
DW		 8421}
8422
6f50473f
TM		 8423static int imsm_disk_slot_to_ord(struct active_array *a, int slot)
8424{
8425 int inst = a->info.container_member;
8426 struct intel_super *super = a->container->sb;
8427 struct imsm_dev *dev = get_imsm_dev(super, inst);
8428 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8429
8430 if (slot > map->num_members) {
8431 pr_err("imsm: imsm_disk_slot_to_ord %d out of range 0..%d\n",
8432 slot, map->num_members - 1);
8433 return -1;
8434 }
8435
8436 if (slot < 0)
8437 return -1;
8438
8439 return get_imsm_ord_tbl_ent(dev, slot, MAP_0);
8440}
8441
8d45d196 8442static void imsm_set_disk(struct active_array *a, int n, int state)
845dea95 8443{
8d45d196
DW		 8444 int inst = a->info.container_member;
8445 struct intel_super *super = a->container->sb;
949c47a0 8446 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8447 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8d45d196 8448 struct imsm_disk *disk;
7ce05701
LD		 8449 struct mdinfo *mdi;
8450 int recovery_not_finished = 0;
0c046afd 8451 int failed;
6f50473f 8452 int ord;
0c046afd 8453 __u8 map_state;
fb12a745
TM		 8454 int rebuild_done = 0;
8455 int i;
8d45d196 8456
fb12a745 8457 ord = get_imsm_ord_tbl_ent(dev, n, MAP_X);
6f50473f 8458 if (ord < 0)
8d45d196
DW		 8459 return;
8460
4e6e574a 8461 dprintf("imsm: set_disk %d:%x\n", n, state);
b10b37b8 8462 disk = get_imsm_disk(super, ord_to_idx(ord));
8d45d196 8463
5802a811 8464 /* check for new failures */
0556e1a2 8465 if (state & DS_FAULTY) {
4c9e8c1e 8466 if (mark_failure(super, dev, disk, ord_to_idx(ord)))
0556e1a2 8467 super->updates_pending++;
8d45d196 8468 }
47ee5a45 8469
19859edc 8470 /* check if in_sync */
0556e1a2 8471 if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) {
238c0a71 8472 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
b10b37b8
DW		 8473
8474 set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord));
fb12a745 8475 rebuild_done = 1;
19859edc
DW		 8476 super->updates_pending++;
8477 }
8d45d196 8478
3b451610
AK		 8479 failed = imsm_count_failed(super, dev, MAP_0);
8480 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
5802a811 8481
0c046afd 8482 /* check if recovery complete, newly degraded, or failed */
94002678
AK		 8483 dprintf("imsm: Detected transition to state ");
8484 switch (map_state) {
8485 case IMSM_T_STATE_NORMAL: /* transition to normal state */
8486 dprintf("normal: ");
8487 if (is_rebuilding(dev)) {
1ade5cc1 8488 dprintf_cont("while rebuilding");
7ce05701
LD		 8489 /* check if recovery is really finished */
8490 for (mdi = a->info.devs; mdi ; mdi = mdi->next)
8491 if (mdi->recovery_start != MaxSector) {
8492 recovery_not_finished = 1;
8493 break;
8494 }
8495 if (recovery_not_finished) {
1ade5cc1
N		 8496 dprintf_cont("\n");
8497 dprintf("Rebuild has not finished yet, state not changed");
7ce05701
LD		 8498 if (a->last_checkpoint < mdi->recovery_start) {
8499 a->last_checkpoint = mdi->recovery_start;
8500 super->updates_pending++;
8501 }
8502 break;
8503 }
94002678 8504 end_migration(dev, super, map_state);
238c0a71 8505 map = get_imsm_map(dev, MAP_0);
94002678
AK		 8506 map->failed_disk_num = ~0;
8507 super->updates_pending++;
8508 a->last_checkpoint = 0;
8509 break;
8510 }
8511 if (is_gen_migration(dev)) {
1ade5cc1 8512 dprintf_cont("while general migration");
bf2f0071 8513 if (a->last_checkpoint >= a->info.component_size)
809da78e 8514 end_migration(dev, super, map_state);
94002678
AK		 8515 else
8516 map->map_state = map_state;
238c0a71 8517 map = get_imsm_map(dev, MAP_0);
28bce06f 8518 map->failed_disk_num = ~0;
94002678 8519 super->updates_pending++;
bf2f0071 8520 break;
94002678
AK		 8521 }
8522 break;
8523 case IMSM_T_STATE_DEGRADED: /* transition to degraded state */
1ade5cc1 8524 dprintf_cont("degraded: ");
089f9d79 8525 if (map->map_state != map_state && !dev->vol.migr_state) {
1ade5cc1 8526 dprintf_cont("mark degraded");
94002678
AK		 8527 map->map_state = map_state;
8528 super->updates_pending++;
8529 a->last_checkpoint = 0;
8530 break;
8531 }
8532 if (is_rebuilding(dev)) {
1ade5cc1 8533 dprintf_cont("while rebuilding.");
94002678 8534 if (map->map_state != map_state) {
1ade5cc1 8535 dprintf_cont(" Map state change");
94002678
AK		 8536 end_migration(dev, super, map_state);
8537 super->updates_pending++;
fb12a745
TM		 8538 } else if (!rebuild_done) {
8539 break;
8540 }
8541
8542 /* check if recovery is really finished */
8543 for (mdi = a->info.devs; mdi ; mdi = mdi->next)
8544 if (mdi->recovery_start != MaxSector) {
8545 recovery_not_finished = 1;
8546 break;
8547 }
8548 if (recovery_not_finished) {
8549 dprintf_cont("\n");
8550 dprintf("Rebuild has not finished yet, state not changed");
8551 if (a->last_checkpoint < mdi->recovery_start) {
8552 a->last_checkpoint =
8553 mdi->recovery_start;
8554 super->updates_pending++;
8555 }
8556 break;
94002678 8557 }
fb12a745
TM		 8558
8559 dprintf_cont(" Rebuild done, still degraded");
8560 dev->vol.migr_state = 0;
8561 set_migr_type(dev, 0);
8562 dev->vol.curr_migr_unit = 0;
8563
8564 for (i = 0; i < map->num_members; i++) {
8565 int idx = get_imsm_ord_tbl_ent(dev, i, MAP_0);
8566
8567 if (idx & IMSM_ORD_REBUILD)
8568 map->failed_disk_num = i;
8569 }
8570 super->updates_pending++;
94002678
AK		 8571 break;
8572 }
8573 if (is_gen_migration(dev)) {
1ade5cc1 8574 dprintf_cont("while general migration");
bf2f0071 8575 if (a->last_checkpoint >= a->info.component_size)
809da78e 8576 end_migration(dev, super, map_state);
94002678
AK		 8577 else {
8578 map->map_state = map_state;
3b451610 8579 manage_second_map(super, dev);
94002678
AK		 8580 }
8581 super->updates_pending++;
bf2f0071 8582 break;
28bce06f 8583 }
6ce1fbf1 8584 if (is_initializing(dev)) {
1ade5cc1 8585 dprintf_cont("while initialization.");
6ce1fbf1
AK		 8586 map->map_state = map_state;
8587 super->updates_pending++;
8588 break;
8589 }
94002678
AK		 8590 break;
8591 case IMSM_T_STATE_FAILED: /* transition to failed state */
1ade5cc1 8592 dprintf_cont("failed: ");
94002678 8593 if (is_gen_migration(dev)) {
1ade5cc1 8594 dprintf_cont("while general migration");
94002678
AK		 8595 map->map_state = map_state;
8596 super->updates_pending++;
8597 break;
8598 }
8599 if (map->map_state != map_state) {
1ade5cc1 8600 dprintf_cont("mark failed");
94002678
AK		 8601 end_migration(dev, super, map_state);
8602 super->updates_pending++;
8603 a->last_checkpoint = 0;
8604 break;
8605 }
8606 break;
8607 default:
1ade5cc1 8608 dprintf_cont("state %i\n", map_state);
5802a811 8609 }
1ade5cc1 8610 dprintf_cont("\n");
845dea95
NB		 8611}
8612
f796af5d 8613static int store_imsm_mpb(int fd, struct imsm_super *mpb)
c2a1e7da 8614{
f796af5d 8615 void *buf = mpb;
c2a1e7da
DW		 8616 __u32 mpb_size = __le32_to_cpu(mpb->mpb_size);
8617 unsigned long long dsize;
8618 unsigned long long sectors;
f36a9ecd 8619 unsigned int sector_size;
c2a1e7da 8620
f36a9ecd 8621 get_dev_sector_size(fd, NULL, &sector_size);
c2a1e7da
DW		 8622 get_dev_size(fd, NULL, &dsize);
8623
f36a9ecd 8624 if (mpb_size > sector_size) {
272f648f 8625 /* -1 to account for anchor */
f36a9ecd 8626 sectors = mpb_sectors(mpb, sector_size) - 1;
c2a1e7da 8627
272f648f 8628 /* write the extended mpb to the sectors preceeding the anchor */
f36a9ecd
PB		 8629 if (lseek64(fd, dsize - (sector_size * (2 + sectors)),
8630 SEEK_SET) < 0)
272f648f 8631 return 1;
c2a1e7da 8632
f36a9ecd
PB		 8633 if ((unsigned long long)write(fd, buf + sector_size,
8634 sector_size * sectors) != sector_size * sectors)
272f648f
DW		 8635 return 1;
8636 }
c2a1e7da 8637
272f648f 8638 /* first block is stored on second to last sector of the disk */
f36a9ecd 8639 if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0)
c2a1e7da
DW		 8640 return 1;
8641
466070ad 8642 if ((unsigned int)write(fd, buf, sector_size) != sector_size)
c2a1e7da
DW		 8643 return 1;
8644
c2a1e7da
DW		 8645 return 0;
8646}
8647
2e735d19 8648static void imsm_sync_metadata(struct supertype *container)
845dea95 8649{
2e735d19 8650 struct intel_super *super = container->sb;
c2a1e7da 8651
1a64be56 8652 dprintf("sync metadata: %d\n", super->updates_pending);
c2a1e7da
DW		 8653 if (!super->updates_pending)
8654 return;
8655
36988a3d 8656 write_super_imsm(container, 0);
c2a1e7da
DW		 8657
8658 super->updates_pending = 0;
845dea95
NB		 8659}
8660
272906ef
DW		 8661static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a)
8662{
8663 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8664 int i = get_imsm_disk_idx(dev, idx, MAP_X);
272906ef
DW		 8665 struct dl *dl;
8666
8667 for (dl = super->disks; dl; dl = dl->next)
8668 if (dl->index == i)
8669 break;
8670
25ed7e59 8671 if (dl && is_failed(&dl->disk))
272906ef
DW		 8672 dl = NULL;
8673
8674 if (dl)
1ade5cc1 8675 dprintf("found %x:%x\n", dl->major, dl->minor);
272906ef
DW		 8676
8677 return dl;
8678}
8679
a20d2ba5 8680static struct dl *imsm_add_spare(struct intel_super *super, int slot,
8ba77d32
AK		 8681 struct active_array *a, int activate_new,
8682 struct mdinfo *additional_test_list)
272906ef
DW		 8683{
8684 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8685 int idx = get_imsm_disk_idx(dev, slot, MAP_X);
a20d2ba5
DW		 8686 struct imsm_super *mpb = super->anchor;
8687 struct imsm_map *map;
272906ef
DW		 8688 unsigned long long pos;
8689 struct mdinfo *d;
8690 struct extent *ex;
a20d2ba5 8691 int i, j;
272906ef 8692 int found;
569cc43f
DW		 8693 __u32 array_start = 0;
8694 __u32 array_end = 0;
272906ef 8695 struct dl *dl;
6c932028 8696 struct mdinfo *test_list;
272906ef
DW		 8697
8698 for (dl = super->disks; dl; dl = dl->next) {
8699 /* If in this array, skip */
8700 for (d = a->info.devs ; d ; d = d->next)
e553d2a4
DW		 8701 if (d->state_fd >= 0 &&
8702 d->disk.major == dl->major &&
272906ef 8703 d->disk.minor == dl->minor) {
8ba77d32
AK		 8704 dprintf("%x:%x already in array\n",
8705 dl->major, dl->minor);
272906ef
DW		 8706 break;
8707 }
8708 if (d)
8709 continue;
6c932028
AK		 8710 test_list = additional_test_list;
8711 while (test_list) {
8712 if (test_list->disk.major == dl->major &&
8713 test_list->disk.minor == dl->minor) {
8ba77d32
AK		 8714 dprintf("%x:%x already in additional test list\n",
8715 dl->major, dl->minor);
8716 break;
8717 }
6c932028 8718 test_list = test_list->next;
8ba77d32 8719 }
6c932028 8720 if (test_list)
8ba77d32 8721 continue;
272906ef 8722
e553d2a4 8723 /* skip in use or failed drives */
25ed7e59 8724 if (is_failed(&dl->disk) || idx == dl->index ||
df474657
DW		 8725 dl->index == -2) {
8726 dprintf("%x:%x status (failed: %d index: %d)\n",
25ed7e59 8727 dl->major, dl->minor, is_failed(&dl->disk), idx);
9a1608e5
DW		 8728 continue;
8729 }
8730
a20d2ba5
DW		 8731 /* skip pure spares when we are looking for partially
8732 * assimilated drives
8733 */
8734 if (dl->index == -1 && !activate_new)
8735 continue;
8736
f2cc4f7d
AO		 8737 if (!drive_validate_sector_size(super, dl))
8738 continue;
8739
272906ef 8740 /* Does this unused device have the requisite free space?
a20d2ba5 8741 * It needs to be able to cover all member volumes
272906ef
DW		 8742 */
8743 ex = get_extents(super, dl);
8744 if (!ex) {
8745 dprintf("cannot get extents\n");
8746 continue;
8747 }
a20d2ba5
DW		 8748 for (i = 0; i < mpb->num_raid_devs; i++) {
8749 dev = get_imsm_dev(super, i);
238c0a71 8750 map = get_imsm_map(dev, MAP_0);
272906ef 8751
a20d2ba5
DW		 8752 /* check if this disk is already a member of
8753 * this array
272906ef 8754 */
620b1713 8755 if (get_imsm_disk_slot(map, dl->index) >= 0)
a20d2ba5
DW		 8756 continue;
8757
8758 found = 0;
8759 j = 0;
8760 pos = 0;
5551b113 8761 array_start = pba_of_lba0(map);
329c8278 8762 array_end = array_start +
44490938 8763 per_dev_array_size(map) - 1;
a20d2ba5
DW		 8764
8765 do {
8766 /* check that we can start at pba_of_lba0 with
44490938 8767 * num_data_stripes*blocks_per_stripe of space
a20d2ba5 8768 */
329c8278 8769 if (array_start >= pos && array_end < ex[j].start) {
a20d2ba5
DW		 8770 found = 1;
8771 break;
8772 }
8773 pos = ex[j].start + ex[j].size;
8774 j++;
8775 } while (ex[j-1].size);
8776
8777 if (!found)
272906ef 8778 break;
a20d2ba5 8779 }
272906ef
DW		 8780
8781 free(ex);
a20d2ba5 8782 if (i < mpb->num_raid_devs) {
329c8278
DW		 8783 dprintf("%x:%x does not have %u to %u available\n",
8784 dl->major, dl->minor, array_start, array_end);
272906ef
DW		 8785 /* No room */
8786 continue;
a20d2ba5
DW		 8787 }
8788 return dl;
272906ef
DW		 8789 }
8790
8791 return dl;
8792}
8793
95d07a2c
LM		 8794static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed)
8795{
8796 struct imsm_dev *dev2;
8797 struct imsm_map *map;
8798 struct dl *idisk;
8799 int slot;
8800 int idx;
8801 __u8 state;
8802
8803 dev2 = get_imsm_dev(cont->sb, dev_idx);
8804 if (dev2) {
238c0a71 8805 state = imsm_check_degraded(cont->sb, dev2, failed, MAP_0);
95d07a2c 8806 if (state == IMSM_T_STATE_FAILED) {
238c0a71 8807 map = get_imsm_map(dev2, MAP_0);
95d07a2c
LM		 8808 if (!map)
8809 return 1;
8810 for (slot = 0; slot < map->num_members; slot++) {
8811 /*
8812 * Check if failed disks are deleted from intel
8813 * disk list or are marked to be deleted
8814 */
238c0a71 8815 idx = get_imsm_disk_idx(dev2, slot, MAP_X);
95d07a2c
LM		 8816 idisk = get_imsm_dl_disk(cont->sb, idx);
8817 /*
8818 * Do not rebuild the array if failed disks
8819 * from failed sub-array are not removed from
8820 * container.
8821 */
8822 if (idisk &&
8823 is_failed(&idisk->disk) &&
8824 (idisk->action != DISK_REMOVE))
8825 return 0;
8826 }
8827 }
8828 }
8829 return 1;
8830}
8831
88758e9d
DW		 8832static struct mdinfo *imsm_activate_spare(struct active_array *a,
8833 struct metadata_update **updates)
8834{
8835 /**
d23fe947
DW		 8836 * Find a device with unused free space and use it to replace a
8837 * failed/vacant region in an array. We replace failed regions one a
8838 * array at a time. The result is that a new spare disk will be added
8839 * to the first failed array and after the monitor has finished
8840 * propagating failures the remainder will be consumed.
88758e9d 8841 *
d23fe947
DW		 8842 * FIXME add a capability for mdmon to request spares from another
8843 * container.
88758e9d
DW		 8844 */
8845
8846 struct intel_super *super = a->container->sb;
88758e9d 8847 int inst = a->info.container_member;
949c47a0 8848 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8849 struct imsm_map *map = get_imsm_map(dev, MAP_0);
88758e9d
DW		 8850 int failed = a->info.array.raid_disks;
8851 struct mdinfo *rv = NULL;
8852 struct mdinfo *d;
8853 struct mdinfo *di;
8854 struct metadata_update *mu;
8855 struct dl *dl;
8856 struct imsm_update_activate_spare *u;
8857 int num_spares = 0;
8858 int i;
95d07a2c 8859 int allowed;
88758e9d
DW		 8860
8861 for (d = a->info.devs ; d ; d = d->next) {
8862 if ((d->curr_state & DS_FAULTY) &&
8863 d->state_fd >= 0)
8864 /* wait for Removal to happen */
8865 return NULL;
8866 if (d->state_fd >= 0)
8867 failed--;
8868 }
8869
8870 dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n",
8871 inst, failed, a->info.array.raid_disks, a->info.array.level);
1af97990 8872
e2962bfc
AK		 8873 if (imsm_reshape_blocks_arrays_changes(super))
8874 return NULL;
1af97990 8875
fc8ca064
AK		 8876 /* Cannot activate another spare if rebuild is in progress already
8877 */
8878 if (is_rebuilding(dev)) {
7a862a02 8879 dprintf("imsm: No spare activation allowed. Rebuild in progress already.\n");
fc8ca064
AK		 8880 return NULL;
8881 }
8882
89c67882
AK		 8883 if (a->info.array.level == 4)
8884 /* No repair for takeovered array
8885 * imsm doesn't support raid4
8886 */
8887 return NULL;
8888
3b451610
AK		 8889 if (imsm_check_degraded(super, dev, failed, MAP_0) !=
8890 IMSM_T_STATE_DEGRADED)
88758e9d
DW		 8891 return NULL;
8892
83ca7d45
AP		 8893 if (get_imsm_map(dev, MAP_0)->map_state == IMSM_T_STATE_UNINITIALIZED) {
8894 dprintf("imsm: No spare activation allowed. Volume is not initialized.\n");
8895 return NULL;
8896 }
8897
95d07a2c
LM		 8898 /*
8899 * If there are any failed disks check state of the other volume.
8900 * Block rebuild if the another one is failed until failed disks
8901 * are removed from container.
8902 */
8903 if (failed) {
7a862a02 8904 dprintf("found failed disks in %.*s, check if there anotherfailed sub-array.\n",
c4acd1e5 8905 MAX_RAID_SERIAL_LEN, dev->volume);
95d07a2c
LM		 8906 /* check if states of the other volumes allow for rebuild */
8907 for (i = 0; i < super->anchor->num_raid_devs; i++) {
8908 if (i != inst) {
8909 allowed = imsm_rebuild_allowed(a->container,
8910 i, failed);
8911 if (!allowed)
8912 return NULL;
8913 }
8914 }
8915 }
8916
88758e9d 8917 /* For each slot, if it is not working, find a spare */
88758e9d
DW		 8918 for (i = 0; i < a->info.array.raid_disks; i++) {
8919 for (d = a->info.devs ; d ; d = d->next)
8920 if (d->disk.raid_disk == i)
8921 break;
8922 dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0);
8923 if (d && (d->state_fd >= 0))
8924 continue;
8925
272906ef 8926 /*
a20d2ba5
DW		 8927 * OK, this device needs recovery. Try to re-add the
8928 * previous occupant of this slot, if this fails see if
8929 * we can continue the assimilation of a spare that was
8930 * partially assimilated, finally try to activate a new
8931 * spare.
272906ef
DW		 8932 */
8933 dl = imsm_readd(super, i, a);
8934 if (!dl)
b303fe21 8935 dl = imsm_add_spare(super, i, a, 0, rv);
a20d2ba5 8936 if (!dl)
b303fe21 8937 dl = imsm_add_spare(super, i, a, 1, rv);
272906ef
DW		 8938 if (!dl)
8939 continue;
1011e834 8940
272906ef 8941 /* found a usable disk with enough space */
503975b9 8942 di = xcalloc(1, sizeof(*di));
272906ef
DW		 8943
8944 /* dl->index will be -1 in the case we are activating a
8945 * pristine spare. imsm_process_update() will create a
8946 * new index in this case. Once a disk is found to be
8947 * failed in all member arrays it is kicked from the
8948 * metadata
8949 */
8950 di->disk.number = dl->index;
d23fe947 8951
272906ef
DW		 8952 /* (ab)use di->devs to store a pointer to the device
8953 * we chose
8954 */
8955 di->devs = (struct mdinfo *) dl;
8956
8957 di->disk.raid_disk = i;
8958 di->disk.major = dl->major;
8959 di->disk.minor = dl->minor;
8960 di->disk.state = 0;
d23534e4 8961 di->recovery_start = 0;
5551b113 8962 di->data_offset = pba_of_lba0(map);
272906ef
DW		 8963 di->component_size = a->info.component_size;
8964 di->container_member = inst;
5e46202e 8965 di->bb.supported = 1;
2c8890e9 8966 if (a->info.consistency_policy == CONSISTENCY_POLICY_PPL) {
2432ce9b 8967 di->ppl_sector = get_ppl_sector(super, inst);
c2462068 8968 di->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9;
2432ce9b 8969 }
148acb7b 8970 super->random = random32();
272906ef
DW		 8971 di->next = rv;
8972 rv = di;
8973 num_spares++;
8974 dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor,
8975 i, di->data_offset);
88758e9d
DW		 8976 }
8977
8978 if (!rv)
8979 /* No spares found */
8980 return rv;
8981 /* Now 'rv' has a list of devices to return.
8982 * Create a metadata_update record to update the
8983 * disk_ord_tbl for the array
8984 */
503975b9 8985 mu = xmalloc(sizeof(*mu));
1011e834 8986 mu->buf = xcalloc(num_spares,
503975b9 8987 sizeof(struct imsm_update_activate_spare));
88758e9d 8988 mu->space = NULL;
cb23f1f4 8989 mu->space_list = NULL;
88758e9d
DW		 8990 mu->len = sizeof(struct imsm_update_activate_spare) * num_spares;
8991 mu->next = *updates;
8992 u = (struct imsm_update_activate_spare *) mu->buf;
8993
8994 for (di = rv ; di ; di = di->next) {
8995 u->type = update_activate_spare;
d23fe947
DW		 8996 u->dl = (struct dl *) di->devs;
8997 di->devs = NULL;
88758e9d
DW		 8998 u->slot = di->disk.raid_disk;
8999 u->array = inst;
9000 u->next = u + 1;
9001 u++;
9002 }
9003 (u-1)->next = NULL;
9004 *updates = mu;
9005
9006 return rv;
9007}
9008
54c2c1ea 9009static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u)
8273f55e 9010{
54c2c1ea 9011 struct imsm_dev *dev = get_imsm_dev(super, idx);
238c0a71
AK		 9012 struct imsm_map *map = get_imsm_map(dev, MAP_0);
9013 struct imsm_map *new_map = get_imsm_map(&u->dev, MAP_0);
54c2c1ea
DW		 9014 struct disk_info *inf = get_disk_info(u);
9015 struct imsm_disk *disk;
8273f55e
DW		 9016 int i;
9017 int j;
8273f55e 9018
54c2c1ea 9019 for (i = 0; i < map->num_members; i++) {
238c0a71 9020 disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, MAP_X));
54c2c1ea
DW		 9021 for (j = 0; j < new_map->num_members; j++)
9022 if (serialcmp(disk->serial, inf[j].serial) == 0)
8273f55e
DW		 9023 return 1;
9024 }
9025
9026 return 0;
9027}
9028
1a64be56
LM		 9029static struct dl *get_disk_super(struct intel_super *super, int major, int minor)
9030{
594dc1b8
JS		 9031 struct dl *dl;
9032
1a64be56 9033 for (dl = super->disks; dl; dl = dl->next)
089f9d79 9034 if (dl->major == major && dl->minor == minor)
1a64be56
LM		 9035 return dl;
9036 return NULL;
9037}
9038
9039static int remove_disk_super(struct intel_super *super, int major, int minor)
9040{
594dc1b8 9041 struct dl *prev;
1a64be56
LM		 9042 struct dl *dl;
9043
9044 prev = NULL;
9045 for (dl = super->disks; dl; dl = dl->next) {
089f9d79 9046 if (dl->major == major && dl->minor == minor) {
1a64be56
LM		 9047 /* remove */
9048 if (prev)
9049 prev->next = dl->next;
9050 else
9051 super->disks = dl->next;
9052 dl->next = NULL;
9053 __free_imsm_disk(dl);
1ade5cc1 9054 dprintf("removed %x:%x\n", major, minor);
1a64be56
LM		 9055 break;
9056 }
9057 prev = dl;
9058 }
9059 return 0;
9060}
9061
f21e18ca 9062static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index);
ae6aad82 9063
1a64be56
LM		 9064static int add_remove_disk_update(struct intel_super *super)
9065{
9066 int check_degraded = 0;
594dc1b8
JS		 9067 struct dl *disk;
9068
1a64be56
LM		 9069 /* add/remove some spares to/from the metadata/contrainer */
9070 while (super->disk_mgmt_list) {
9071 struct dl *disk_cfg;
9072
9073 disk_cfg = super->disk_mgmt_list;
9074 super->disk_mgmt_list = disk_cfg->next;
9075 disk_cfg->next = NULL;
9076
9077 if (disk_cfg->action == DISK_ADD) {
9078 disk_cfg->next = super->disks;
9079 super->disks = disk_cfg;
9080 check_degraded = 1;
1ade5cc1
N		 9081 dprintf("added %x:%x\n",
9082 disk_cfg->major, disk_cfg->minor);
1a64be56
LM		 9083 } else if (disk_cfg->action == DISK_REMOVE) {
9084 dprintf("Disk remove action processed: %x.%x\n",
9085 disk_cfg->major, disk_cfg->minor);
9086 disk = get_disk_super(super,
9087 disk_cfg->major,
9088 disk_cfg->minor);
9089 if (disk) {
9090 /* store action status */
9091 disk->action = DISK_REMOVE;
9092 /* remove spare disks only */
9093 if (disk->index == -1) {
9094 remove_disk_super(super,
9095 disk_cfg->major,
9096 disk_cfg->minor);
9097 }
9098 }
9099 /* release allocate disk structure */
9100 __free_imsm_disk(disk_cfg);
9101 }
9102 }
9103 return check_degraded;
9104}
9105
a29911da
PC		 9106static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u,
9107 struct intel_super *super,
9108 void ***space_list)
9109{
9110 struct intel_dev *id;
9111 void **tofree = NULL;
9112 int ret_val = 0;
9113
1ade5cc1 9114 dprintf("(enter)\n");
089f9d79 9115 if (u->subdev < 0 || u->subdev > 1) {
a29911da
PC		 9116 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
9117 return ret_val;
9118 }
089f9d79 9119 if (space_list == NULL || *space_list == NULL) {
a29911da
PC		 9120 dprintf("imsm: Error: Memory is not allocated\n");
9121 return ret_val;
9122 }
9123
9124 for (id = super->devlist ; id; id = id->next) {
9125 if (id->index == (unsigned)u->subdev) {
9126 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
9127 struct imsm_map *map;
9128 struct imsm_dev *new_dev =
9129 (struct imsm_dev *)*space_list;
238c0a71 9130 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
a29911da
PC		 9131 int to_state;
9132 struct dl *new_disk;
9133
9134 if (new_dev == NULL)
9135 return ret_val;
9136 *space_list = **space_list;
9137 memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0));
238c0a71 9138 map = get_imsm_map(new_dev, MAP_0);
a29911da
PC		 9139 if (migr_map) {
9140 dprintf("imsm: Error: migration in progress");
9141 return ret_val;
9142 }
9143
9144 to_state = map->map_state;
9145 if ((u->new_level == 5) && (map->raid_level == 0)) {
9146 map->num_members++;
9147 /* this should not happen */
9148 if (u->new_disks[0] < 0) {
9149 map->failed_disk_num =
9150 map->num_members - 1;
9151 to_state = IMSM_T_STATE_DEGRADED;
9152 } else
9153 to_state = IMSM_T_STATE_NORMAL;
9154 }
8e59f3d8 9155 migrate(new_dev, super, to_state, MIGR_GEN_MIGR);
a29911da
PC		 9156 if (u->new_level > -1)
9157 map->raid_level = u->new_level;
238c0a71 9158 migr_map = get_imsm_map(new_dev, MAP_1);
a29911da
PC		 9159 if ((u->new_level == 5) &&
9160 (migr_map->raid_level == 0)) {
9161 int ord = map->num_members - 1;
9162 migr_map->num_members--;
9163 if (u->new_disks[0] < 0)
9164 ord |= IMSM_ORD_REBUILD;
9165 set_imsm_ord_tbl_ent(map,
9166 map->num_members - 1,
9167 ord);
9168 }
9169 id->dev = new_dev;
9170 tofree = (void **)dev;
9171
4bba0439
PC		 9172 /* update chunk size
9173 */
06fb291a
PB		 9174 if (u->new_chunksize > 0) {
9175 unsigned long long num_data_stripes;
9529d343
MD		 9176 struct imsm_map *dest_map =
9177 get_imsm_map(dev, MAP_0);
06fb291a 9178 int used_disks =
9529d343 9179 imsm_num_data_members(dest_map);
06fb291a
PB		 9180
9181 if (used_disks == 0)
9182 return ret_val;
9183
4bba0439
PC		 9184 map->blocks_per_strip =
9185 __cpu_to_le16(u->new_chunksize * 2);
06fb291a 9186 num_data_stripes =
fcc2c9da 9187 imsm_dev_size(dev) / used_disks;
06fb291a
PB		 9188 num_data_stripes /= map->blocks_per_strip;
9189 num_data_stripes /= map->num_domains;
9190 set_num_data_stripes(map, num_data_stripes);
9191 }
4bba0439 9192
44490938
MD		 9193 /* ensure blocks_per_member has valid value
9194 */
9195 set_blocks_per_member(map,
9196 per_dev_array_size(map) +
9197 NUM_BLOCKS_DIRTY_STRIPE_REGION);
9198
a29911da
PC		 9199 /* add disk
9200 */
089f9d79
JS		 9201 if (u->new_level != 5 || migr_map->raid_level != 0 ||
9202 migr_map->raid_level == map->raid_level)
a29911da
PC		 9203 goto skip_disk_add;
9204
9205 if (u->new_disks[0] >= 0) {
9206 /* use passes spare
9207 */
9208 new_disk = get_disk_super(super,
9209 major(u->new_disks[0]),
9210 minor(u->new_disks[0]));
7a862a02 9211 dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n",
a29911da
PC		 9212 major(u->new_disks[0]),
9213 minor(u->new_disks[0]),
9214 new_disk, new_disk->index);
9215 if (new_disk == NULL)
9216 goto error_disk_add;
9217
9218 new_disk->index = map->num_members - 1;
9219 /* slot to fill in autolayout
9220 */
9221 new_disk->raiddisk = new_disk->index;
9222 new_disk->disk.status |= CONFIGURED_DISK;
9223 new_disk->disk.status &= ~SPARE_DISK;
9224 } else
9225 goto error_disk_add;
9226
9227skip_disk_add:
9228 *tofree = *space_list;
9229 /* calculate new size
9230 */
f3871fdc 9231 imsm_set_array_size(new_dev, -1);
a29911da
PC		 9232
9233 ret_val = 1;
9234 }
9235 }
9236
9237 if (tofree)
9238 *space_list = tofree;
9239 return ret_val;
9240
9241error_disk_add:
9242 dprintf("Error: imsm: Cannot find disk.\n");
9243 return ret_val;
9244}
9245
f3871fdc
AK		 9246static int apply_size_change_update(struct imsm_update_size_change *u,
9247 struct intel_super *super)
9248{
9249 struct intel_dev *id;
9250 int ret_val = 0;
9251
1ade5cc1 9252 dprintf("(enter)\n");
089f9d79 9253 if (u->subdev < 0 || u->subdev > 1) {
f3871fdc
AK		 9254 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
9255 return ret_val;
9256 }
9257
9258 for (id = super->devlist ; id; id = id->next) {
9259 if (id->index == (unsigned)u->subdev) {
9260 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
9261 struct imsm_map *map = get_imsm_map(dev, MAP_0);
9529d343 9262 int used_disks = imsm_num_data_members(map);
f3871fdc 9263 unsigned long long blocks_per_member;
06fb291a 9264 unsigned long long num_data_stripes;
44490938
MD		 9265 unsigned long long new_size_per_disk;
9266
9267 if (used_disks == 0)
9268 return 0;
f3871fdc
AK		 9269
9270 /* calculate new size
9271 */
44490938
MD		 9272 new_size_per_disk = u->new_size / used_disks;
9273 blocks_per_member = new_size_per_disk +
9274 NUM_BLOCKS_DIRTY_STRIPE_REGION;
9275 num_data_stripes = new_size_per_disk /
06fb291a
PB		 9276 map->blocks_per_strip;
9277 num_data_stripes /= map->num_domains;
9278 dprintf("(size: %llu, blocks per member: %llu, num_data_stipes: %llu)\n",
44490938 9279 u->new_size, new_size_per_disk,
06fb291a 9280 num_data_stripes);
f3871fdc 9281 set_blocks_per_member(map, blocks_per_member);
06fb291a 9282 set_num_data_stripes(map, num_data_stripes);
f3871fdc
AK		 9283 imsm_set_array_size(dev, u->new_size);
9284
9285 ret_val = 1;
9286 break;
9287 }
9288 }
9289
9290 return ret_val;
9291}
9292
061d7da3 9293static int apply_update_activate_spare(struct imsm_update_activate_spare *u,
ca9de185 9294 struct intel_super *super,
061d7da3
LO		 9295 struct active_array *active_array)
9296{
9297 struct imsm_super *mpb = super->anchor;
9298 struct imsm_dev *dev = get_imsm_dev(super, u->array);
238c0a71 9299 struct imsm_map *map = get_imsm_map(dev, MAP_0);
061d7da3
LO		 9300 struct imsm_map *migr_map;
9301 struct active_array *a;
9302 struct imsm_disk *disk;
9303 __u8 to_state;
9304 struct dl *dl;
9305 unsigned int found;
9306 int failed;
5961eeec 9307 int victim;
061d7da3 9308 int i;
5961eeec 9309 int second_map_created = 0;
061d7da3 9310
5961eeec 9311 for (; u; u = u->next) {
238c0a71 9312 victim = get_imsm_disk_idx(dev, u->slot, MAP_X);
061d7da3 9313
5961eeec 9314 if (victim < 0)
9315 return 0;
061d7da3 9316
5961eeec 9317 for (dl = super->disks; dl; dl = dl->next)
9318 if (dl == u->dl)
9319 break;
061d7da3 9320
5961eeec 9321 if (!dl) {
7a862a02 9322 pr_err("error: imsm_activate_spare passed an unknown disk (index: %d)\n",
5961eeec 9323 u->dl->index);
9324 return 0;
9325 }
061d7da3 9326
5961eeec 9327 /* count failures (excluding rebuilds and the victim)
9328 * to determine map[0] state
9329 */
9330 failed = 0;
9331 for (i = 0; i < map->num_members; i++) {
9332 if (i == u->slot)
9333 continue;
9334 disk = get_imsm_disk(super,
238c0a71 9335 get_imsm_disk_idx(dev, i, MAP_X));
5961eeec 9336 if (!disk || is_failed(disk))
9337 failed++;
9338 }
061d7da3 9339
5961eeec 9340 /* adding a pristine spare, assign a new index */
9341 if (dl->index < 0) {
9342 dl->index = super->anchor->num_disks;
9343 super->anchor->num_disks++;
9344 }
9345 disk = &dl->disk;
9346 disk->status |= CONFIGURED_DISK;
9347 disk->status &= ~SPARE_DISK;
9348
9349 /* mark rebuild */
238c0a71 9350 to_state = imsm_check_degraded(super, dev, failed, MAP_0);
5961eeec 9351 if (!second_map_created) {
9352 second_map_created = 1;
9353 map->map_state = IMSM_T_STATE_DEGRADED;
9354 migrate(dev, super, to_state, MIGR_REBUILD);
9355 } else
9356 map->map_state = to_state;
238c0a71 9357 migr_map = get_imsm_map(dev, MAP_1);
5961eeec 9358 set_imsm_ord_tbl_ent(map, u->slot, dl->index);
9359 set_imsm_ord_tbl_ent(migr_map, u->slot,
9360 dl->index | IMSM_ORD_REBUILD);
9361
9362 /* update the family_num to mark a new container
9363 * generation, being careful to record the existing
9364 * family_num in orig_family_num to clean up after
9365 * earlier mdadm versions that neglected to set it.
9366 */
9367 if (mpb->orig_family_num == 0)
9368 mpb->orig_family_num = mpb->family_num;
9369 mpb->family_num += super->random;
9370
9371 /* count arrays using the victim in the metadata */
9372 found = 0;
9373 for (a = active_array; a ; a = a->next) {
9374 dev = get_imsm_dev(super, a->info.container_member);
238c0a71 9375 map = get_imsm_map(dev, MAP_0);
061d7da3 9376
5961eeec 9377 if (get_imsm_disk_slot(map, victim) >= 0)
9378 found++;
9379 }
061d7da3 9380
5961eeec 9381 /* delete the victim if it is no longer being
9382 * utilized anywhere
061d7da3 9383 */
5961eeec 9384 if (!found) {
9385 struct dl **dlp;
061d7da3 9386
5961eeec 9387 /* We know that 'manager' isn't touching anything,
9388 * so it is safe to delete
9389 */
9390 for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next)
061d7da3
LO		 9391 if ((*dlp)->index == victim)
9392 break;
5961eeec 9393
9394 /* victim may be on the missing list */
9395 if (!*dlp)
9396 for (dlp = &super->missing; *dlp;
9397 dlp = &(*dlp)->next)
9398 if ((*dlp)->index == victim)
9399 break;
9400 imsm_delete(super, dlp, victim);
9401 }
061d7da3
LO		 9402 }
9403
9404 return 1;
9405}
a29911da 9406
2e5dc010
N		 9407static int apply_reshape_container_disks_update(struct imsm_update_reshape *u,
9408 struct intel_super *super,
9409 void ***space_list)
9410{
9411 struct dl *new_disk;
9412 struct intel_dev *id;
9413 int i;
9414 int delta_disks = u->new_raid_disks - u->old_raid_disks;
ee4beede 9415 int disk_count = u->old_raid_disks;
2e5dc010
N		 9416 void **tofree = NULL;
9417 int devices_to_reshape = 1;
9418 struct imsm_super *mpb = super->anchor;
9419 int ret_val = 0;
d098291a 9420 unsigned int dev_id;
2e5dc010 9421
1ade5cc1 9422 dprintf("(enter)\n");
2e5dc010
N		 9423
9424 /* enable spares to use in array */
9425 for (i = 0; i < delta_disks; i++) {
9426 new_disk = get_disk_super(super,
9427 major(u->new_disks[i]),
9428 minor(u->new_disks[i]));
7a862a02 9429 dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n",
2e5dc010
N		 9430 major(u->new_disks[i]), minor(u->new_disks[i]),
9431 new_disk, new_disk->index);
089f9d79
JS		 9432 if (new_disk == NULL ||
9433 (new_disk->index >= 0 &&
9434 new_disk->index < u->old_raid_disks))
2e5dc010 9435 goto update_reshape_exit;
ee4beede 9436 new_disk->index = disk_count++;
2e5dc010
N		 9437 /* slot to fill in autolayout
9438 */
9439 new_disk->raiddisk = new_disk->index;
9440 new_disk->disk.status |=
9441 CONFIGURED_DISK;
9442 new_disk->disk.status &= ~SPARE_DISK;
9443 }
9444
ed7333bd
AK		 9445 dprintf("imsm: volume set mpb->num_raid_devs = %i\n",
9446 mpb->num_raid_devs);
2e5dc010
N		 9447 /* manage changes in volume
9448 */
d098291a 9449 for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) {
2e5dc010
N		 9450 void **sp = *space_list;
9451 struct imsm_dev *newdev;
9452 struct imsm_map *newmap, *oldmap;
9453
d098291a
AK		 9454 for (id = super->devlist ; id; id = id->next) {
9455 if (id->index == dev_id)
9456 break;
9457 }
9458 if (id == NULL)
9459 break;
2e5dc010
N		 9460 if (!sp)
9461 continue;
9462 *space_list = *sp;
9463 newdev = (void*)sp;
9464 /* Copy the dev, but not (all of) the map */
9465 memcpy(newdev, id->dev, sizeof(*newdev));
238c0a71
AK		 9466 oldmap = get_imsm_map(id->dev, MAP_0);
9467 newmap = get_imsm_map(newdev, MAP_0);
2e5dc010
N		 9468 /* Copy the current map */
9469 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
9470 /* update one device only
9471 */
9472 if (devices_to_reshape) {
ed7333bd
AK		 9473 dprintf("imsm: modifying subdev: %i\n",
9474 id->index);
2e5dc010
N		 9475 devices_to_reshape--;
9476 newdev->vol.migr_state = 1;
9477 newdev->vol.curr_migr_unit = 0;
ea672ee1 9478 set_migr_type(newdev, MIGR_GEN_MIGR);
2e5dc010
N		 9479 newmap->num_members = u->new_raid_disks;
9480 for (i = 0; i < delta_disks; i++) {
9481 set_imsm_ord_tbl_ent(newmap,
9482 u->old_raid_disks + i,
9483 u->old_raid_disks + i);
9484 }
9485 /* New map is correct, now need to save old map
9486 */
238c0a71 9487 newmap = get_imsm_map(newdev, MAP_1);
2e5dc010
N		 9488 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
9489
f3871fdc 9490 imsm_set_array_size(newdev, -1);
2e5dc010
N		 9491 }
9492
9493 sp = (void **)id->dev;
9494 id->dev = newdev;
9495 *sp = tofree;
9496 tofree = sp;
8e59f3d8
AK		 9497
9498 /* Clear migration record */
9499 memset(super->migr_rec, 0, sizeof(struct migr_record));
2e5dc010 9500 }
819bc634
AK		 9501 if (tofree)
9502 *space_list = tofree;
2e5dc010
N		 9503 ret_val = 1;
9504
9505update_reshape_exit:
9506
9507 return ret_val;
9508}
9509
bb025c2f 9510static int apply_takeover_update(struct imsm_update_takeover *u,
8ca6df95
KW		 9511 struct intel_super *super,
9512 void ***space_list)
bb025c2f
KW		 9513{
9514 struct imsm_dev *dev = NULL;
8ca6df95
KW		 9515 struct intel_dev *dv;
9516 struct imsm_dev *dev_new;
bb025c2f
KW		 9517 struct imsm_map *map;
9518 struct dl *dm, *du;
8ca6df95 9519 int i;
bb025c2f
KW		 9520
9521 for (dv = super->devlist; dv; dv = dv->next)
9522 if (dv->index == (unsigned int)u->subarray) {
9523 dev = dv->dev;
9524 break;
9525 }
9526
9527 if (dev == NULL)
9528 return 0;
9529
238c0a71 9530 map = get_imsm_map(dev, MAP_0);
bb025c2f
KW		 9531
9532 if (u->direction == R10_TO_R0) {
06fb291a
PB		 9533 unsigned long long num_data_stripes;
9534
9535 map->num_domains = 1;
44490938 9536 num_data_stripes = imsm_dev_size(dev) / 2;
06fb291a
PB		 9537 num_data_stripes /= map->blocks_per_strip;
9538 num_data_stripes /= map->num_domains;
9539 set_num_data_stripes(map, num_data_stripes);
9540
43d5ec18 9541 /* Number of failed disks must be half of initial disk number */
3b451610
AK		 9542 if (imsm_count_failed(super, dev, MAP_0) !=
9543 (map->num_members / 2))
43d5ec18
KW		 9544 return 0;
9545
bb025c2f
KW		 9546 /* iterate through devices to mark removed disks as spare */
9547 for (dm = super->disks; dm; dm = dm->next) {
9548 if (dm->disk.status & FAILED_DISK) {
9549 int idx = dm->index;
9550 /* update indexes on the disk list */
9551/* FIXME this loop-with-the-loop looks wrong, I'm not convinced
9552 the index values will end up being correct.... NB */
9553 for (du = super->disks; du; du = du->next)
9554 if (du->index > idx)
9555 du->index--;
9556 /* mark as spare disk */
a8619d23 9557 mark_spare(dm);
bb025c2f
KW		 9558 }
9559 }
bb025c2f
KW		 9560 /* update map */
9561 map->num_members = map->num_members / 2;
9562 map->map_state = IMSM_T_STATE_NORMAL;
9563 map->num_domains = 1;
9564 map->raid_level = 0;
9565 map->failed_disk_num = -1;
9566 }
9567
8ca6df95
KW		 9568 if (u->direction == R0_TO_R10) {
9569 void **space;
9570 /* update slots in current disk list */
9571 for (dm = super->disks; dm; dm = dm->next) {
9572 if (dm->index >= 0)
9573 dm->index *= 2;
9574 }
9575 /* create new *missing* disks */
9576 for (i = 0; i < map->num_members; i++) {
9577 space = *space_list;
9578 if (!space)
9579 continue;
9580 *space_list = *space;
9581 du = (void *)space;
9582 memcpy(du, super->disks, sizeof(*du));
8ca6df95
KW		 9583 du->fd = -1;
9584 du->minor = 0;
9585 du->major = 0;
9586 du->index = (i * 2) + 1;
9587 sprintf((char *)du->disk.serial,
9588 " MISSING_%d", du->index);
9589 sprintf((char *)du->serial,
9590 "MISSING_%d", du->index);
9591 du->next = super->missing;
9592 super->missing = du;
9593 }
9594 /* create new dev and map */
9595 space = *space_list;
9596 if (!space)
9597 return 0;
9598 *space_list = *space;
9599 dev_new = (void *)space;
9600 memcpy(dev_new, dev, sizeof(*dev));
9601 /* update new map */
238c0a71 9602 map = get_imsm_map(dev_new, MAP_0);
8ca6df95 9603 map->num_members = map->num_members * 2;
1a2487c2 9604 map->map_state = IMSM_T_STATE_DEGRADED;
8ca6df95
KW		 9605 map->num_domains = 2;
9606 map->raid_level = 1;
9607 /* replace dev<->dev_new */
9608 dv->dev = dev_new;
9609 }
bb025c2f
KW		 9610 /* update disk order table */
9611 for (du = super->disks; du; du = du->next)
9612 if (du->index >= 0)
9613 set_imsm_ord_tbl_ent(map, du->index, du->index);
8ca6df95 9614 for (du = super->missing; du; du = du->next)
1a2487c2
KW		 9615 if (du->index >= 0) {
9616 set_imsm_ord_tbl_ent(map, du->index, du->index);
4c9e8c1e 9617 mark_missing(super, dv->dev, &du->disk, du->index);
1a2487c2 9618 }
bb025c2f
KW		 9619
9620 return 1;
9621}
9622
e8319a19
DW		 9623static void imsm_process_update(struct supertype *st,
9624 struct metadata_update *update)
9625{
9626 /**
9627 * crack open the metadata_update envelope to find the update record
9628 * update can be one of:
d195167d
AK		 9629 * update_reshape_container_disks - all the arrays in the container
9630 * are being reshaped to have more devices. We need to mark
9631 * the arrays for general migration and convert selected spares
9632 * into active devices.
9633 * update_activate_spare - a spare device has replaced a failed
1011e834
N		 9634 * device in an array, update the disk_ord_tbl. If this disk is
9635 * present in all member arrays then also clear the SPARE_DISK
9636 * flag
d195167d
AK		 9637 * update_create_array
9638 * update_kill_array
9639 * update_rename_array
9640 * update_add_remove_disk
e8319a19
DW		 9641 */
9642 struct intel_super *super = st->sb;
4d7b1503 9643 struct imsm_super *mpb;
e8319a19
DW		 9644 enum imsm_update_type type = *(enum imsm_update_type *) update->buf;
9645
4d7b1503
DW		 9646 /* update requires a larger buf but the allocation failed */
9647 if (super->next_len && !super->next_buf) {
9648 super->next_len = 0;
9649 return;
9650 }
9651
9652 if (super->next_buf) {
9653 memcpy(super->next_buf, super->buf, super->len);
9654 free(super->buf);
9655 super->len = super->next_len;
9656 super->buf = super->next_buf;
9657
9658 super->next_len = 0;
9659 super->next_buf = NULL;
9660 }
9661
9662 mpb = super->anchor;
9663
e8319a19 9664 switch (type) {
0ec5d470
AK		 9665 case update_general_migration_checkpoint: {
9666 struct intel_dev *id;
9667 struct imsm_update_general_migration_checkpoint *u =
9668 (void *)update->buf;
9669
1ade5cc1 9670 dprintf("called for update_general_migration_checkpoint\n");
0ec5d470
AK		 9671
9672 /* find device under general migration */
9673 for (id = super->devlist ; id; id = id->next) {
9674 if (is_gen_migration(id->dev)) {
9675 id->dev->vol.curr_migr_unit =
9676 __cpu_to_le32(u->curr_migr_unit);
9677 super->updates_pending++;
9678 }
9679 }
9680 break;
9681 }
bb025c2f
KW		 9682 case update_takeover: {
9683 struct imsm_update_takeover *u = (void *)update->buf;
1a2487c2
KW		 9684 if (apply_takeover_update(u, super, &update->space_list)) {
9685 imsm_update_version_info(super);
bb025c2f 9686 super->updates_pending++;
1a2487c2 9687 }
bb025c2f
KW		 9688 break;
9689 }
9690
78b10e66 9691 case update_reshape_container_disks: {
d195167d 9692 struct imsm_update_reshape *u = (void *)update->buf;
2e5dc010
N		 9693 if (apply_reshape_container_disks_update(
9694 u, super, &update->space_list))
9695 super->updates_pending++;
78b10e66
N		 9696 break;
9697 }
48c5303a 9698 case update_reshape_migration: {
a29911da
PC		 9699 struct imsm_update_reshape_migration *u = (void *)update->buf;
9700 if (apply_reshape_migration_update(
9701 u, super, &update->space_list))
9702 super->updates_pending++;
48c5303a
PC		 9703 break;
9704 }
f3871fdc
AK		 9705 case update_size_change: {
9706 struct imsm_update_size_change *u = (void *)update->buf;
9707 if (apply_size_change_update(u, super))
9708 super->updates_pending++;
9709 break;
9710 }
e8319a19 9711 case update_activate_spare: {
1011e834 9712 struct imsm_update_activate_spare *u = (void *) update->buf;
061d7da3
LO		 9713 if (apply_update_activate_spare(u, super, st->arrays))
9714 super->updates_pending++;
8273f55e
DW		 9715 break;
9716 }
9717 case update_create_array: {
9718 /* someone wants to create a new array, we need to be aware of
9719 * a few races/collisions:
9720 * 1/ 'Create' called by two separate instances of mdadm
9721 * 2/ 'Create' versus 'activate_spare': mdadm has chosen
9722 * devices that have since been assimilated via
9723 * activate_spare.
9724 * In the event this update can not be carried out mdadm will
9725 * (FIX ME) notice that its update did not take hold.
9726 */
9727 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 9728 struct intel_dev *dv;
8273f55e
DW		 9729 struct imsm_dev *dev;
9730 struct imsm_map *map, *new_map;
9731 unsigned long long start, end;
9732 unsigned long long new_start, new_end;
9733 int i;
54c2c1ea
DW		 9734 struct disk_info *inf;
9735 struct dl *dl;
8273f55e
DW		 9736
9737 /* handle racing creates: first come first serve */
9738 if (u->dev_idx < mpb->num_raid_devs) {
1ade5cc1 9739 dprintf("subarray %d already defined\n", u->dev_idx);
ba2de7ba 9740 goto create_error;
8273f55e
DW		 9741 }
9742
9743 /* check update is next in sequence */
9744 if (u->dev_idx != mpb->num_raid_devs) {
1ade5cc1
N		 9745 dprintf("can not create array %d expected index %d\n",
9746 u->dev_idx, mpb->num_raid_devs);
ba2de7ba 9747 goto create_error;
8273f55e
DW		 9748 }
9749
238c0a71 9750 new_map = get_imsm_map(&u->dev, MAP_0);
5551b113 9751 new_start = pba_of_lba0(new_map);
44490938 9752 new_end = new_start + per_dev_array_size(new_map);
54c2c1ea 9753 inf = get_disk_info(u);
8273f55e
DW		 9754
9755 /* handle activate_spare versus create race:
9756 * check to make sure that overlapping arrays do not include
9757 * overalpping disks
9758 */
9759 for (i = 0; i < mpb->num_raid_devs; i++) {
949c47a0 9760 dev = get_imsm_dev(super, i);
238c0a71 9761 map = get_imsm_map(dev, MAP_0);
5551b113 9762 start = pba_of_lba0(map);
44490938 9763 end = start + per_dev_array_size(map);
8273f55e
DW		 9764 if ((new_start >= start && new_start <= end) ||
9765 (start >= new_start && start <= new_end))
54c2c1ea
DW		 9766 /* overlap */;
9767 else
9768 continue;
9769
9770 if (disks_overlap(super, i, u)) {
1ade5cc1 9771 dprintf("arrays overlap\n");
ba2de7ba 9772 goto create_error;
8273f55e
DW		 9773 }
9774 }
8273f55e 9775
949c47a0
DW		 9776 /* check that prepare update was successful */
9777 if (!update->space) {
1ade5cc1 9778 dprintf("prepare update failed\n");
ba2de7ba 9779 goto create_error;
949c47a0
DW		 9780 }
9781
54c2c1ea
DW		 9782 /* check that all disks are still active before committing
9783 * changes. FIXME: could we instead handle this by creating a
9784 * degraded array? That's probably not what the user expects,
9785 * so better to drop this update on the floor.
9786 */
9787 for (i = 0; i < new_map->num_members; i++) {
9788 dl = serial_to_dl(inf[i].serial, super);
9789 if (!dl) {
1ade5cc1 9790 dprintf("disk disappeared\n");
ba2de7ba 9791 goto create_error;
54c2c1ea 9792 }
949c47a0
DW		 9793 }
9794
8273f55e 9795 super->updates_pending++;
54c2c1ea
DW		 9796
9797 /* convert spares to members and fixup ord_tbl */
9798 for (i = 0; i < new_map->num_members; i++) {
9799 dl = serial_to_dl(inf[i].serial, super);
9800 if (dl->index == -1) {
9801 dl->index = mpb->num_disks;
9802 mpb->num_disks++;
9803 dl->disk.status |= CONFIGURED_DISK;
9804 dl->disk.status &= ~SPARE_DISK;
9805 }
9806 set_imsm_ord_tbl_ent(new_map, i, dl->index);
9807 }
9808
ba2de7ba
DW		 9809 dv = update->space;
9810 dev = dv->dev;
949c47a0
DW		 9811 update->space = NULL;
9812 imsm_copy_dev(dev, &u->dev);
ba2de7ba
DW		 9813 dv->index = u->dev_idx;
9814 dv->next = super->devlist;
9815 super->devlist = dv;
8273f55e 9816 mpb->num_raid_devs++;
8273f55e 9817
4d1313e9 9818 imsm_update_version_info(super);
8273f55e 9819 break;
ba2de7ba
DW		 9820 create_error:
9821 /* mdmon knows how to release update->space, but not
9822 * ((struct intel_dev *) update->space)->dev
9823 */
9824 if (update->space) {
9825 dv = update->space;
9826 free(dv->dev);
9827 }
8273f55e 9828 break;
e8319a19 9829 }
33414a01
DW		 9830 case update_kill_array: {
9831 struct imsm_update_kill_array *u = (void *) update->buf;
9832 int victim = u->dev_idx;
9833 struct active_array *a;
9834 struct intel_dev **dp;
9835 struct imsm_dev *dev;
9836
9837 /* sanity check that we are not affecting the uuid of
9838 * active arrays, or deleting an active array
9839 *
9840 * FIXME when immutable ids are available, but note that
9841 * we'll also need to fixup the invalidated/active
9842 * subarray indexes in mdstat
9843 */
9844 for (a = st->arrays; a; a = a->next)
9845 if (a->info.container_member >= victim)
9846 break;
9847 /* by definition if mdmon is running at least one array
9848 * is active in the container, so checking
9849 * mpb->num_raid_devs is just extra paranoia
9850 */
9851 dev = get_imsm_dev(super, victim);
9852 if (a || !dev || mpb->num_raid_devs == 1) {
9853 dprintf("failed to delete subarray-%d\n", victim);
9854 break;
9855 }
9856
9857 for (dp = &super->devlist; *dp;)
f21e18ca 9858 if ((*dp)->index == (unsigned)super->current_vol) {
33414a01
DW		 9859 *dp = (*dp)->next;
9860 } else {
f21e18ca 9861 if ((*dp)->index > (unsigned)victim)
33414a01
DW		 9862 (*dp)->index--;
9863 dp = &(*dp)->next;
9864 }
9865 mpb->num_raid_devs--;
9866 super->updates_pending++;
9867 break;
9868 }
aa534678
DW		 9869 case update_rename_array: {
9870 struct imsm_update_rename_array *u = (void *) update->buf;
9871 char name[MAX_RAID_SERIAL_LEN+1];
9872 int target = u->dev_idx;
9873 struct active_array *a;
9874 struct imsm_dev *dev;
9875
9876 /* sanity check that we are not affecting the uuid of
9877 * an active array
9878 */
9879 snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name);
9880 name[MAX_RAID_SERIAL_LEN] = '\0';
9881 for (a = st->arrays; a; a = a->next)
9882 if (a->info.container_member == target)
9883 break;
9884 dev = get_imsm_dev(super, u->dev_idx);
9885 if (a || !dev || !check_name(super, name, 1)) {
9886 dprintf("failed to rename subarray-%d\n", target);
9887 break;
9888 }
9889
cdbe98cd 9890 snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name);
aa534678
DW		 9891 super->updates_pending++;
9892 break;
9893 }
1a64be56 9894 case update_add_remove_disk: {
43dad3d6 9895 /* we may be able to repair some arrays if disks are
095b8088 9896 * being added, check the status of add_remove_disk
1a64be56
LM		 9897 * if discs has been added.
9898 */
9899 if (add_remove_disk_update(super)) {
43dad3d6 9900 struct active_array *a;
072b727f
DW		 9901
9902 super->updates_pending++;
1a64be56 9903 for (a = st->arrays; a; a = a->next)
43dad3d6
DW		 9904 a->check_degraded = 1;
9905 }
43dad3d6 9906 break;
e8319a19 9907 }
bbab0940
TM		 9908 case update_prealloc_badblocks_mem:
9909 break;
e6e9dd3f
AP		 9910 case update_rwh_policy: {
9911 struct imsm_update_rwh_policy *u = (void *)update->buf;
9912 int target = u->dev_idx;
9913 struct imsm_dev *dev = get_imsm_dev(super, target);
9914 if (!dev) {
9915 dprintf("could not find subarray-%d\n", target);
9916 break;
9917 }
9918
9919 if (dev->rwh_policy != u->new_policy) {
9920 dev->rwh_policy = u->new_policy;
9921 super->updates_pending++;
9922 }
9923 break;
9924 }
1a64be56 9925 default:
7a862a02 9926 pr_err("error: unsuported process update type:(type: %d)\n", type);
1a64be56 9927 }
e8319a19 9928}
88758e9d 9929
bc0b9d34
PC		 9930static struct mdinfo *get_spares_for_grow(struct supertype *st);
9931
5fe6f031
N		 9932static int imsm_prepare_update(struct supertype *st,
9933 struct metadata_update *update)
8273f55e 9934{
949c47a0 9935 /**
4d7b1503
DW		 9936 * Allocate space to hold new disk entries, raid-device entries or a new
9937 * mpb if necessary. The manager synchronously waits for updates to
9938 * complete in the monitor, so new mpb buffers allocated here can be
9939 * integrated by the monitor thread without worrying about live pointers
9940 * in the manager thread.
8273f55e 9941 */
095b8088 9942 enum imsm_update_type type;
4d7b1503 9943 struct intel_super *super = st->sb;
f36a9ecd 9944 unsigned int sector_size = super->sector_size;
4d7b1503
DW		 9945 struct imsm_super *mpb = super->anchor;
9946 size_t buf_len;
9947 size_t len = 0;
949c47a0 9948
095b8088
N		 9949 if (update->len < (int)sizeof(type))
9950 return 0;
9951
9952 type = *(enum imsm_update_type *) update->buf;
9953
949c47a0 9954 switch (type) {
0ec5d470 9955 case update_general_migration_checkpoint:
095b8088
N		 9956 if (update->len < (int)sizeof(struct imsm_update_general_migration_checkpoint))
9957 return 0;
1ade5cc1 9958 dprintf("called for update_general_migration_checkpoint\n");
0ec5d470 9959 break;
abedf5fc
KW		 9960 case update_takeover: {
9961 struct imsm_update_takeover *u = (void *)update->buf;
095b8088
N		 9962 if (update->len < (int)sizeof(*u))
9963 return 0;
abedf5fc
KW		 9964 if (u->direction == R0_TO_R10) {
9965 void **tail = (void **)&update->space_list;
9966 struct imsm_dev *dev = get_imsm_dev(super, u->subarray);
238c0a71 9967 struct imsm_map *map = get_imsm_map(dev, MAP_0);
abedf5fc
KW		 9968 int num_members = map->num_members;
9969 void *space;
9970 int size, i;
abedf5fc
KW		 9971 /* allocate memory for added disks */
9972 for (i = 0; i < num_members; i++) {
9973 size = sizeof(struct dl);
503975b9 9974 space = xmalloc(size);
abedf5fc
KW		 9975 *tail = space;
9976 tail = space;
9977 *tail = NULL;
9978 }
9979 /* allocate memory for new device */
9980 size = sizeof_imsm_dev(super->devlist->dev, 0) +
9981 (num_members * sizeof(__u32));
503975b9
N		 9982 space = xmalloc(size);
9983 *tail = space;
9984 tail = space;
9985 *tail = NULL;
9986 len = disks_to_mpb_size(num_members * 2);
abedf5fc
KW		 9987 }
9988
9989 break;
9990 }
78b10e66 9991 case update_reshape_container_disks: {
d195167d
AK		 9992 /* Every raid device in the container is about to
9993 * gain some more devices, and we will enter a
9994 * reconfiguration.
9995 * So each 'imsm_map' will be bigger, and the imsm_vol
9996 * will now hold 2 of them.
9997 * Thus we need new 'struct imsm_dev' allocations sized
9998 * as sizeof_imsm_dev but with more devices in both maps.
9999 */
10000 struct imsm_update_reshape *u = (void *)update->buf;
10001 struct intel_dev *dl;
10002 void **space_tail = (void**)&update->space_list;
10003
095b8088
N		 10004 if (update->len < (int)sizeof(*u))
10005 return 0;
10006
1ade5cc1 10007 dprintf("for update_reshape\n");
d195167d
AK		 10008
10009 for (dl = super->devlist; dl; dl = dl->next) {
10010 int size = sizeof_imsm_dev(dl->dev, 1);
10011 void *s;
d677e0b8
AK		 10012 if (u->new_raid_disks > u->old_raid_disks)
10013 size += sizeof(__u32)*2*
10014 (u->new_raid_disks - u->old_raid_disks);
503975b9 10015 s = xmalloc(size);
d195167d
AK		 10016 *space_tail = s;
10017 space_tail = s;
10018 *space_tail = NULL;
10019 }
10020
10021 len = disks_to_mpb_size(u->new_raid_disks);
10022 dprintf("New anchor length is %llu\n", (unsigned long long)len);
78b10e66
N		 10023 break;
10024 }
48c5303a 10025 case update_reshape_migration: {
bc0b9d34
PC		 10026 /* for migration level 0->5 we need to add disks
10027 * so the same as for container operation we will copy
10028 * device to the bigger location.
10029 * in memory prepared device and new disk area are prepared
10030 * for usage in process update
10031 */
10032 struct imsm_update_reshape_migration *u = (void *)update->buf;
10033 struct intel_dev *id;
10034 void **space_tail = (void **)&update->space_list;
10035 int size;
10036 void *s;
10037 int current_level = -1;
10038
095b8088
N		 10039 if (update->len < (int)sizeof(*u))
10040 return 0;
10041
1ade5cc1 10042 dprintf("for update_reshape\n");
bc0b9d34
PC		 10043
10044 /* add space for bigger array in update
10045 */
10046 for (id = super->devlist; id; id = id->next) {
10047 if (id->index == (unsigned)u->subdev) {
10048 size = sizeof_imsm_dev(id->dev, 1);
10049 if (u->new_raid_disks > u->old_raid_disks)
10050 size += sizeof(__u32)*2*
10051 (u->new_raid_disks - u->old_raid_disks);
503975b9 10052 s = xmalloc(size);
bc0b9d34
PC		 10053 *space_tail = s;
10054 space_tail = s;
10055 *space_tail = NULL;
10056 break;
10057 }
10058 }
10059 if (update->space_list == NULL)
10060 break;
10061
10062 /* add space for disk in update
10063 */
10064 size = sizeof(struct dl);
503975b9 10065 s = xmalloc(size);
bc0b9d34
PC		 10066 *space_tail = s;
10067 space_tail = s;
10068 *space_tail = NULL;
10069
10070 /* add spare device to update
10071 */
10072 for (id = super->devlist ; id; id = id->next)
10073 if (id->index == (unsigned)u->subdev) {
10074 struct imsm_dev *dev;
10075 struct imsm_map *map;
10076
10077 dev = get_imsm_dev(super, u->subdev);
238c0a71 10078 map = get_imsm_map(dev, MAP_0);
bc0b9d34
PC		 10079 current_level = map->raid_level;
10080 break;
10081 }
089f9d79 10082 if (u->new_level == 5 && u->new_level != current_level) {
bc0b9d34
PC		 10083 struct mdinfo *spares;
10084
10085 spares = get_spares_for_grow(st);
10086 if (spares) {
10087 struct dl *dl;
10088 struct mdinfo *dev;
10089
10090 dev = spares->devs;
10091 if (dev) {
10092 u->new_disks[0] =
10093 makedev(dev->disk.major,
10094 dev->disk.minor);
10095 dl = get_disk_super(super,
10096 dev->disk.major,
10097 dev->disk.minor);
10098 dl->index = u->old_raid_disks;
10099 dev = dev->next;
10100 }
10101 sysfs_free(spares);
10102 }
10103 }
10104 len = disks_to_mpb_size(u->new_raid_disks);
10105 dprintf("New anchor length is %llu\n", (unsigned long long)len);
48c5303a
PC		 10106 break;
10107 }
f3871fdc 10108 case update_size_change: {
095b8088
N		 10109 if (update->len < (int)sizeof(struct imsm_update_size_change))
10110 return 0;
10111 break;
10112 }
10113 case update_activate_spare: {
10114 if (update->len < (int)sizeof(struct imsm_update_activate_spare))
10115 return 0;
f3871fdc
AK		 10116 break;
10117 }
949c47a0
DW		 10118 case update_create_array: {
10119 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 10120 struct intel_dev *dv;
54c2c1ea 10121 struct imsm_dev *dev = &u->dev;
238c0a71 10122 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW		 10123 struct dl *dl;
10124 struct disk_info *inf;
10125 int i;
10126 int activate = 0;
949c47a0 10127
095b8088
N		 10128 if (update->len < (int)sizeof(*u))
10129 return 0;
10130
54c2c1ea
DW		 10131 inf = get_disk_info(u);
10132 len = sizeof_imsm_dev(dev, 1);
ba2de7ba 10133 /* allocate a new super->devlist entry */
503975b9
N		 10134 dv = xmalloc(sizeof(*dv));
10135 dv->dev = xmalloc(len);
10136 update->space = dv;
949c47a0 10137
54c2c1ea
DW		 10138 /* count how many spares will be converted to members */
10139 for (i = 0; i < map->num_members; i++) {
10140 dl = serial_to_dl(inf[i].serial, super);
10141 if (!dl) {
10142 /* hmm maybe it failed?, nothing we can do about
10143 * it here
10144 */
10145 continue;
10146 }
10147 if (count_memberships(dl, super) == 0)
10148 activate++;
10149 }
10150 len += activate * sizeof(struct imsm_disk);
949c47a0 10151 break;
095b8088
N		 10152 }
10153 case update_kill_array: {
10154 if (update->len < (int)sizeof(struct imsm_update_kill_array))
10155 return 0;
949c47a0
DW		 10156 break;
10157 }
095b8088
N		 10158 case update_rename_array: {
10159 if (update->len < (int)sizeof(struct imsm_update_rename_array))
10160 return 0;
10161 break;
10162 }
10163 case update_add_remove_disk:
10164 /* no update->len needed */
10165 break;
bbab0940
TM		 10166 case update_prealloc_badblocks_mem:
10167 super->extra_space += sizeof(struct bbm_log) -
10168 get_imsm_bbm_log_size(super->bbm_log);
10169 break;
e6e9dd3f
AP		 10170 case update_rwh_policy: {
10171 if (update->len < (int)sizeof(struct imsm_update_rwh_policy))
10172 return 0;
10173 break;
10174 }
095b8088
N		 10175 default:
10176 return 0;
949c47a0 10177 }
8273f55e 10178
4d7b1503
DW		 10179 /* check if we need a larger metadata buffer */
10180 if (super->next_buf)
10181 buf_len = super->next_len;
10182 else
10183 buf_len = super->len;
10184
bbab0940 10185 if (__le32_to_cpu(mpb->mpb_size) + super->extra_space + len > buf_len) {
4d7b1503
DW		 10186 /* ok we need a larger buf than what is currently allocated
10187 * if this allocation fails process_update will notice that
10188 * ->next_len is set and ->next_buf is NULL
10189 */
bbab0940
TM		 10190 buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) +
10191 super->extra_space + len, sector_size);
4d7b1503
DW		 10192 if (super->next_buf)
10193 free(super->next_buf);
10194
10195 super->next_len = buf_len;
f36a9ecd 10196 if (posix_memalign(&super->next_buf, sector_size, buf_len) == 0)
1f45a8ad
DW		 10197 memset(super->next_buf, 0, buf_len);
10198 else
4d7b1503
DW		 10199 super->next_buf = NULL;
10200 }
5fe6f031 10201 return 1;
8273f55e
DW		 10202}
10203
ae6aad82 10204/* must be called while manager is quiesced */
f21e18ca 10205static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index)
ae6aad82
DW		 10206{
10207 struct imsm_super *mpb = super->anchor;
ae6aad82
DW		 10208 struct dl *iter;
10209 struct imsm_dev *dev;
10210 struct imsm_map *map;
4c9e8c1e 10211 unsigned int i, j, num_members;
fb12a745 10212 __u32 ord, ord_map0;
4c9e8c1e 10213 struct bbm_log *log = super->bbm_log;
ae6aad82 10214
1ade5cc1 10215 dprintf("deleting device[%d] from imsm_super\n", index);
ae6aad82
DW		 10216
10217 /* shift all indexes down one */
10218 for (iter = super->disks; iter; iter = iter->next)
f21e18ca 10219 if (iter->index > (int)index)
ae6aad82 10220 iter->index--;
47ee5a45 10221 for (iter = super->missing; iter; iter = iter->next)
f21e18ca 10222 if (iter->index > (int)index)
47ee5a45 10223 iter->index--;
ae6aad82
DW		 10224
10225 for (i = 0; i < mpb->num_raid_devs; i++) {
10226 dev = get_imsm_dev(super, i);
238c0a71 10227 map = get_imsm_map(dev, MAP_0);
24565c9a
DW		 10228 num_members = map->num_members;
10229 for (j = 0; j < num_members; j++) {
10230 /* update ord entries being careful not to propagate
10231 * ord-flags to the first map
10232 */
238c0a71 10233 ord = get_imsm_ord_tbl_ent(dev, j, MAP_X);
fb12a745 10234 ord_map0 = get_imsm_ord_tbl_ent(dev, j, MAP_0);
ae6aad82 10235
24565c9a
DW		 10236 if (ord_to_idx(ord) <= index)
10237 continue;
ae6aad82 10238
238c0a71 10239 map = get_imsm_map(dev, MAP_0);
fb12a745 10240 set_imsm_ord_tbl_ent(map, j, ord_map0 - 1);
238c0a71 10241 map = get_imsm_map(dev, MAP_1);
24565c9a
DW		 10242 if (map)
10243 set_imsm_ord_tbl_ent(map, j, ord - 1);
ae6aad82
DW		 10244 }
10245 }
10246
4c9e8c1e
TM		 10247 for (i = 0; i < log->entry_count; i++) {
10248 struct bbm_log_entry *entry = &log->marked_block_entries[i];
10249
10250 if (entry->disk_ordinal <= index)
10251 continue;
10252 entry->disk_ordinal--;
10253 }
10254
ae6aad82
DW		 10255 mpb->num_disks--;
10256 super->updates_pending++;
24565c9a
DW		 10257 if (*dlp) {
10258 struct dl *dl = *dlp;
10259
10260 *dlp = (*dlp)->next;
10261 __free_imsm_disk(dl);
10262 }
ae6aad82 10263}
9a717282
AK		 10264
10265static void close_targets(int *targets, int new_disks)
10266{
10267 int i;
10268
10269 if (!targets)
10270 return;
10271
10272 for (i = 0; i < new_disks; i++) {
10273 if (targets[i] >= 0) {
10274 close(targets[i]);
10275 targets[i] = -1;
10276 }
10277 }
10278}
10279
10280static int imsm_get_allowed_degradation(int level, int raid_disks,
10281 struct intel_super *super,
10282 struct imsm_dev *dev)
10283{
10284 switch (level) {
bf5cf7c7 10285 case 1:
9a717282
AK		 10286 case 10:{
10287 int ret_val = 0;
10288 struct imsm_map *map;
10289 int i;
10290
10291 ret_val = raid_disks/2;
10292 /* check map if all disks pairs not failed
10293 * in both maps
10294 */
238c0a71 10295 map = get_imsm_map(dev, MAP_0);
9a717282
AK		 10296 for (i = 0; i < ret_val; i++) {
10297 int degradation = 0;
10298 if (get_imsm_disk(super, i) == NULL)
10299 degradation++;
10300 if (get_imsm_disk(super, i + 1) == NULL)
10301 degradation++;
10302 if (degradation == 2)
10303 return 0;
10304 }
238c0a71 10305 map = get_imsm_map(dev, MAP_1);
9a717282
AK		 10306 /* if there is no second map
10307 * result can be returned
10308 */
10309 if (map == NULL)
10310 return ret_val;
10311 /* check degradation in second map
10312 */
10313 for (i = 0; i < ret_val; i++) {
10314 int degradation = 0;
10315 if (get_imsm_disk(super, i) == NULL)
10316 degradation++;
10317 if (get_imsm_disk(super, i + 1) == NULL)
10318 degradation++;
10319 if (degradation == 2)
10320 return 0;
10321 }
10322 return ret_val;
10323 }
10324 case 5:
10325 return 1;
10326 case 6:
10327 return 2;
10328 default:
10329 return 0;
10330 }
10331}
10332
687629c2
AK		 10333/*******************************************************************************
10334 * Function: open_backup_targets
10335 * Description: Function opens file descriptors for all devices given in
10336 * info->devs
10337 * Parameters:
10338 * info : general array info
10339 * raid_disks : number of disks
10340 * raid_fds : table of device's file descriptors
9a717282
AK		 10341 * super : intel super for raid10 degradation check
10342 * dev : intel device for raid10 degradation check
687629c2
AK		 10343 * Returns:
10344 * 0 : success
10345 * -1 : fail
10346 ******************************************************************************/
9a717282
AK		 10347int open_backup_targets(struct mdinfo *info, int raid_disks, int *raid_fds,
10348 struct intel_super *super, struct imsm_dev *dev)
687629c2
AK		 10349{
10350 struct mdinfo *sd;
f627f5ad 10351 int i;
9a717282 10352 int opened = 0;
f627f5ad
AK		 10353
10354 for (i = 0; i < raid_disks; i++)
10355 raid_fds[i] = -1;
687629c2
AK		 10356
10357 for (sd = info->devs ; sd ; sd = sd->next) {
10358 char *dn;
10359
10360 if (sd->disk.state & (1<<MD_DISK_FAULTY)) {
10361 dprintf("disk is faulty!!\n");
10362 continue;
10363 }
10364
089f9d79 10365 if (sd->disk.raid_disk >= raid_disks || sd->disk.raid_disk < 0)
687629c2
AK		 10366 continue;
10367
10368 dn = map_dev(sd->disk.major,
10369 sd->disk.minor, 1);
10370 raid_fds[sd->disk.raid_disk] = dev_open(dn, O_RDWR);
10371 if (raid_fds[sd->disk.raid_disk] < 0) {
e12b3daa 10372 pr_err("cannot open component\n");
9a717282 10373 continue;
687629c2 10374 }
9a717282
AK		 10375 opened++;
10376 }
10377 /* check if maximum array degradation level is not exceeded
10378 */
10379 if ((raid_disks - opened) >
089f9d79
JS		 10380 imsm_get_allowed_degradation(info->new_level, raid_disks,
10381 super, dev)) {
e12b3daa 10382 pr_err("Not enough disks can be opened.\n");
9a717282
AK		 10383 close_targets(raid_fds, raid_disks);
10384 return -2;
687629c2
AK		 10385 }
10386 return 0;
10387}
10388
d31ad643
PB		 10389/*******************************************************************************
10390 * Function: validate_container_imsm
10391 * Description: This routine validates container after assemble,
10392 * eg. if devices in container are under the same controller.
10393 *
10394 * Parameters:
10395 * info : linked list with info about devices used in array
10396 * Returns:
10397 * 1 : HBA mismatch
10398 * 0 : Success
10399 ******************************************************************************/
10400int validate_container_imsm(struct mdinfo *info)
10401{
6b781d33
AP		 10402 if (check_env("IMSM_NO_PLATFORM"))
10403 return 0;
d31ad643 10404
6b781d33
AP		 10405 struct sys_dev *idev;
10406 struct sys_dev *hba = NULL;
10407 struct sys_dev *intel_devices = find_intel_devices();
10408 char *dev_path = devt_to_devpath(makedev(info->disk.major,
10409 info->disk.minor));
10410
10411 for (idev = intel_devices; idev; idev = idev->next) {
10412 if (dev_path && strstr(dev_path, idev->path)) {
10413 hba = idev;
10414 break;
d31ad643 10415 }
6b781d33
AP		 10416 }
10417 if (dev_path)
d31ad643
PB		 10418 free(dev_path);
10419
6b781d33
AP		 10420 if (!hba) {
10421 pr_err("WARNING - Cannot detect HBA for device %s!\n",
10422 devid2kname(makedev(info->disk.major, info->disk.minor)));
10423 return 1;
10424 }
10425
10426 const struct imsm_orom *orom = get_orom_by_device_id(hba->dev_id);
10427 struct mdinfo *dev;
10428
10429 for (dev = info->next; dev; dev = dev->next) {
10430 dev_path = devt_to_devpath(makedev(dev->disk.major, dev->disk.minor));
10431
10432 struct sys_dev *hba2 = NULL;
10433 for (idev = intel_devices; idev; idev = idev->next) {
10434 if (dev_path && strstr(dev_path, idev->path)) {
10435 hba2 = idev;
10436 break;
d31ad643
PB		 10437 }
10438 }
6b781d33
AP		 10439 if (dev_path)
10440 free(dev_path);
10441
10442 const struct imsm_orom *orom2 = hba2 == NULL ? NULL :
10443 get_orom_by_device_id(hba2->dev_id);
10444
10445 if (hba2 && hba->type != hba2->type) {
10446 pr_err("WARNING - HBAs of devices do not match %s != %s\n",
10447 get_sys_dev_type(hba->type), get_sys_dev_type(hba2->type));
10448 return 1;
10449 }
10450
07cb1e57 10451 if (orom != orom2) {
6b781d33
AP		 10452 pr_err("WARNING - IMSM container assembled with disks under different HBAs!\n"
10453 " This operation is not supported and can lead to data loss.\n");
10454 return 1;
10455 }
10456
10457 if (!orom) {
10458 pr_err("WARNING - IMSM container assembled with disks under HBAs without IMSM platform support!\n"
10459 " This operation is not supported and can lead to data loss.\n");
10460 return 1;
10461 }
d31ad643 10462 }
6b781d33 10463
d31ad643
PB		 10464 return 0;
10465}
32141c17 10466
6f50473f
TM		 10467/*******************************************************************************
10468* Function: imsm_record_badblock
10469* Description: This routine stores new bad block record in BBM log
10470*
10471* Parameters:
10472* a : array containing a bad block
10473* slot : disk number containing a bad block
10474* sector : bad block sector
10475* length : bad block sectors range
10476* Returns:
10477* 1 : Success
10478* 0 : Error
10479******************************************************************************/
10480static int imsm_record_badblock(struct active_array *a, int slot,
10481 unsigned long long sector, int length)
10482{
10483 struct intel_super *super = a->container->sb;
10484 int ord;
10485 int ret;
10486
10487 ord = imsm_disk_slot_to_ord(a, slot);
10488 if (ord < 0)
10489 return 0;
10490
10491 ret = record_new_badblock(super->bbm_log, ord_to_idx(ord), sector,
10492 length);
10493 if (ret)
10494 super->updates_pending++;
10495
10496 return ret;
10497}
c07a5a4f
TM		 10498/*******************************************************************************
10499* Function: imsm_clear_badblock
10500* Description: This routine clears bad block record from BBM log
10501*
10502* Parameters:
10503* a : array containing a bad block
10504* slot : disk number containing a bad block
10505* sector : bad block sector
10506* length : bad block sectors range
10507* Returns:
10508* 1 : Success
10509* 0 : Error
10510******************************************************************************/
10511static int imsm_clear_badblock(struct active_array *a, int slot,
10512 unsigned long long sector, int length)
10513{
10514 struct intel_super *super = a->container->sb;
10515 int ord;
10516 int ret;
10517
10518 ord = imsm_disk_slot_to_ord(a, slot);
10519 if (ord < 0)
10520 return 0;
10521
10522 ret = clear_badblock(super->bbm_log, ord_to_idx(ord), sector, length);
10523 if (ret)
10524 super->updates_pending++;
10525
10526 return ret;
10527}
928f1424
TM		 10528/*******************************************************************************
10529* Function: imsm_get_badblocks
10530* Description: This routine get list of bad blocks for an array
10531*
10532* Parameters:
10533* a : array
10534* slot : disk number
10535* Returns:
10536* bb : structure containing bad blocks
10537* NULL : error
10538******************************************************************************/
10539static struct md_bb *imsm_get_badblocks(struct active_array *a, int slot)
10540{
10541 int inst = a->info.container_member;
10542 struct intel_super *super = a->container->sb;
10543 struct imsm_dev *dev = get_imsm_dev(super, inst);
10544 struct imsm_map *map = get_imsm_map(dev, MAP_0);
10545 int ord;
10546
10547 ord = imsm_disk_slot_to_ord(a, slot);
10548 if (ord < 0)
10549 return NULL;
10550
10551 get_volume_badblocks(super->bbm_log, ord_to_idx(ord), pba_of_lba0(map),
44490938 10552 per_dev_array_size(map), &super->bb);
928f1424
TM		 10553
10554 return &super->bb;
10555}
27156a57
TM		 10556/*******************************************************************************
10557* Function: examine_badblocks_imsm
10558* Description: Prints list of bad blocks on a disk to the standard output
10559*
10560* Parameters:
10561* st : metadata handler
10562* fd : open file descriptor for device
10563* devname : device name
10564* Returns:
10565* 0 : Success
10566* 1 : Error
10567******************************************************************************/
10568static int examine_badblocks_imsm(struct supertype *st, int fd, char *devname)
10569{
10570 struct intel_super *super = st->sb;
10571 struct bbm_log *log = super->bbm_log;
10572 struct dl *d = NULL;
10573 int any = 0;
10574
10575 for (d = super->disks; d ; d = d->next) {
10576 if (strcmp(d->devname, devname) == 0)
10577 break;
10578 }
10579
10580 if ((d == NULL) || (d->index < 0)) { /* serial mismatch probably */
10581 pr_err("%s doesn't appear to be part of a raid array\n",
10582 devname);
10583 return 1;
10584 }
10585
10586 if (log != NULL) {
10587 unsigned int i;
10588 struct bbm_log_entry *entry = &log->marked_block_entries[0];
10589
10590 for (i = 0; i < log->entry_count; i++) {
10591 if (entry[i].disk_ordinal == d->index) {
10592 unsigned long long sector = __le48_to_cpu(
10593 &entry[i].defective_block_start);
10594 int cnt = entry[i].marked_count + 1;
10595
10596 if (!any) {
10597 printf("Bad-blocks on %s:\n", devname);
10598 any = 1;
10599 }
10600
10601 printf("%20llu for %d sectors\n", sector, cnt);
10602 }
10603 }
10604 }
10605
10606 if (!any)
10607 printf("No bad-blocks list configured on %s\n", devname);
10608
10609 return 0;
10610}
687629c2
AK		 10611/*******************************************************************************
10612 * Function: init_migr_record_imsm
10613 * Description: Function inits imsm migration record
10614 * Parameters:
10615 * super : imsm internal array info
10616 * dev : device under migration
10617 * info : general array info to find the smallest device
10618 * Returns:
10619 * none
10620 ******************************************************************************/
10621void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev,
10622 struct mdinfo *info)
10623{
10624 struct intel_super *super = st->sb;
10625 struct migr_record *migr_rec = super->migr_rec;
10626 int new_data_disks;
10627 unsigned long long dsize, dev_sectors;
10628 long long unsigned min_dev_sectors = -1LLU;
10629 struct mdinfo *sd;
10630 char nm[30];
10631 int fd;
238c0a71
AK		 10632 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
10633 struct imsm_map *map_src = get_imsm_map(dev, MAP_1);
687629c2 10634 unsigned long long num_migr_units;
3ef4403c 10635 unsigned long long array_blocks;
687629c2
AK		 10636
10637 memset(migr_rec, 0, sizeof(struct migr_record));
10638 migr_rec->family_num = __cpu_to_le32(super->anchor->family_num);
10639
10640 /* only ascending reshape supported now */
10641 migr_rec->ascending_migr = __cpu_to_le32(1);
10642
10643 migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE /
10644 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
e1742195
AK		 10645 migr_rec->dest_depth_per_unit *=
10646 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
9529d343 10647 new_data_disks = imsm_num_data_members(map_dest);
687629c2
AK		 10648 migr_rec->blocks_per_unit =
10649 __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks);
10650 migr_rec->dest_depth_per_unit =
10651 __cpu_to_le32(migr_rec->dest_depth_per_unit);
3ef4403c 10652 array_blocks = info->component_size * new_data_disks;
687629c2
AK		 10653 num_migr_units =
10654 array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit);
10655
10656 if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit))
10657 num_migr_units++;
10658 migr_rec->num_migr_units = __cpu_to_le32(num_migr_units);
10659
10660 migr_rec->post_migr_vol_cap = dev->size_low;
10661 migr_rec->post_migr_vol_cap_hi = dev->size_high;
10662
687629c2
AK		 10663 /* Find the smallest dev */
10664 for (sd = info->devs ; sd ; sd = sd->next) {
10665 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
10666 fd = dev_open(nm, O_RDONLY);
10667 if (fd < 0)
10668 continue;
10669 get_dev_size(fd, NULL, &dsize);
10670 dev_sectors = dsize / 512;
10671 if (dev_sectors < min_dev_sectors)
10672 min_dev_sectors = dev_sectors;
10673 close(fd);
10674 }
10675 migr_rec->ckpt_area_pba = __cpu_to_le32(min_dev_sectors -
10676 RAID_DISK_RESERVED_BLOCKS_IMSM_HI);
10677
10678 write_imsm_migr_rec(st);
10679
10680 return;
10681}
10682
10683/*******************************************************************************
10684 * Function: save_backup_imsm
10685 * Description: Function saves critical data stripes to Migration Copy Area
10686 * and updates the current migration unit status.
10687 * Use restore_stripes() to form a destination stripe,
10688 * and to write it to the Copy Area.
10689 * Parameters:
10690 * st : supertype information
aea93171 10691 * dev : imsm device that backup is saved for
687629c2
AK		 10692 * info : general array info
10693 * buf : input buffer
687629c2
AK		 10694 * length : length of data to backup (blocks_per_unit)
10695 * Returns:
10696 * 0 : success
10697 *, -1 : fail
10698 ******************************************************************************/
10699int save_backup_imsm(struct supertype *st,
10700 struct imsm_dev *dev,
10701 struct mdinfo *info,
10702 void *buf,
687629c2
AK		 10703 int length)
10704{
10705 int rv = -1;
10706 struct intel_super *super = st->sb;
594dc1b8
JS		 10707 unsigned long long *target_offsets;
10708 int *targets;
687629c2 10709 int i;
238c0a71 10710 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
687629c2 10711 int new_disks = map_dest->num_members;
ab724b98
AK		 10712 int dest_layout = 0;
10713 int dest_chunk;
d1877f69 10714 unsigned long long start;
9529d343 10715 int data_disks = imsm_num_data_members(map_dest);
687629c2 10716
503975b9 10717 targets = xmalloc(new_disks * sizeof(int));
687629c2 10718
7e45b550
AK		 10719 for (i = 0; i < new_disks; i++)
10720 targets[i] = -1;
10721
503975b9 10722 target_offsets = xcalloc(new_disks, sizeof(unsigned long long));
687629c2 10723
d1877f69 10724 start = info->reshape_progress * 512;
687629c2 10725 for (i = 0; i < new_disks; i++) {
687629c2
AK		 10726 target_offsets[i] = (unsigned long long)
10727 __le32_to_cpu(super->migr_rec->ckpt_area_pba) * 512;
d1877f69
AK		 10728 /* move back copy area adderss, it will be moved forward
10729 * in restore_stripes() using start input variable
10730 */
10731 target_offsets[i] -= start/data_disks;
687629c2
AK		 10732 }
10733
9a717282
AK		 10734 if (open_backup_targets(info, new_disks, targets,
10735 super, dev))
687629c2
AK		 10736 goto abort;
10737
68eb8bc6 10738 dest_layout = imsm_level_to_layout(map_dest->raid_level);
ab724b98
AK		 10739 dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512;
10740
687629c2
AK		 10741 if (restore_stripes(targets, /* list of dest devices */
10742 target_offsets, /* migration record offsets */
10743 new_disks,
ab724b98
AK		 10744 dest_chunk,
10745 map_dest->raid_level,
10746 dest_layout,
10747 -1, /* source backup file descriptor */
10748 0, /* input buf offset
10749 * always 0 buf is already offseted */
d1877f69 10750 start,
687629c2
AK		 10751 length,
10752 buf) != 0) {
e7b84f9d 10753 pr_err("Error restoring stripes\n");
687629c2
AK		 10754 goto abort;
10755 }
10756
10757 rv = 0;
10758
10759abort:
10760 if (targets) {
9a717282 10761 close_targets(targets, new_disks);
687629c2
AK		 10762 free(targets);
10763 }
10764 free(target_offsets);
10765
10766 return rv;
10767}
10768
10769/*******************************************************************************
10770 * Function: save_checkpoint_imsm
10771 * Description: Function called for current unit status update
10772 * in the migration record. It writes it to disk.
10773 * Parameters:
10774 * super : imsm internal array info
10775 * info : general array info
10776 * Returns:
10777 * 0: success
10778 * 1: failure
0228d92c
AK		 10779 * 2: failure, means no valid migration record
10780 * / no general migration in progress /
687629c2
AK		 10781 ******************************************************************************/
10782int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state)
10783{
10784 struct intel_super *super = st->sb;
f8b72ef5
AK		 10785 unsigned long long blocks_per_unit;
10786 unsigned long long curr_migr_unit;
10787
2e062e82 10788 if (load_imsm_migr_rec(super, info) != 0) {
7a862a02 10789 dprintf("imsm: ERROR: Cannot read migration record for checkpoint save.\n");
2e062e82
AK		 10790 return 1;
10791 }
10792
f8b72ef5
AK		 10793 blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit);
10794 if (blocks_per_unit == 0) {
0228d92c
AK		 10795 dprintf("imsm: no migration in progress.\n");
10796 return 2;
687629c2 10797 }
f8b72ef5
AK		 10798 curr_migr_unit = info->reshape_progress / blocks_per_unit;
10799 /* check if array is alligned to copy area
10800 * if it is not alligned, add one to current migration unit value
10801 * this can happend on array reshape finish only
10802 */
10803 if (info->reshape_progress % blocks_per_unit)
10804 curr_migr_unit++;
687629c2
AK		 10805
10806 super->migr_rec->curr_migr_unit =
f8b72ef5 10807 __cpu_to_le32(curr_migr_unit);
687629c2
AK		 10808 super->migr_rec->rec_status = __cpu_to_le32(state);
10809 super->migr_rec->dest_1st_member_lba =
f8b72ef5
AK		 10810 __cpu_to_le32(curr_migr_unit *
10811 __le32_to_cpu(super->migr_rec->dest_depth_per_unit));
687629c2 10812 if (write_imsm_migr_rec(st) < 0) {
7a862a02 10813 dprintf("imsm: Cannot write migration record outside backup area\n");
687629c2
AK		 10814 return 1;
10815 }
10816
10817 return 0;
10818}
10819
276d77db
AK		 10820/*******************************************************************************
10821 * Function: recover_backup_imsm
10822 * Description: Function recovers critical data from the Migration Copy Area
10823 * while assembling an array.
10824 * Parameters:
10825 * super : imsm internal array info
10826 * info : general array info
10827 * Returns:
10828 * 0 : success (or there is no data to recover)
10829 * 1 : fail
10830 ******************************************************************************/
10831int recover_backup_imsm(struct supertype *st, struct mdinfo *info)
10832{
10833 struct intel_super *super = st->sb;
10834 struct migr_record *migr_rec = super->migr_rec;
594dc1b8 10835 struct imsm_map *map_dest;
276d77db
AK		 10836 struct intel_dev *id = NULL;
10837 unsigned long long read_offset;
10838 unsigned long long write_offset;
10839 unsigned unit_len;
10840 int *targets = NULL;
10841 int new_disks, i, err;
10842 char *buf = NULL;
10843 int retval = 1;
f36a9ecd 10844 unsigned int sector_size = super->sector_size;
276d77db
AK		 10845 unsigned long curr_migr_unit = __le32_to_cpu(migr_rec->curr_migr_unit);
10846 unsigned long num_migr_units = __le32_to_cpu(migr_rec->num_migr_units);
276d77db 10847 char buffer[20];
6c3560c0 10848 int skipped_disks = 0;
276d77db
AK		 10849
10850 err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20);
10851 if (err < 1)
10852 return 1;
10853
10854 /* recover data only during assemblation */
10855 if (strncmp(buffer, "inactive", 8) != 0)
10856 return 0;
10857 /* no data to recover */
10858 if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL)
10859 return 0;
10860 if (curr_migr_unit >= num_migr_units)
10861 return 1;
10862
10863 /* find device during reshape */
10864 for (id = super->devlist; id; id = id->next)
10865 if (is_gen_migration(id->dev))
10866 break;
10867 if (id == NULL)
10868 return 1;
10869
238c0a71 10870 map_dest = get_imsm_map(id->dev, MAP_0);
276d77db
AK		 10871 new_disks = map_dest->num_members;
10872
10873 read_offset = (unsigned long long)
10874 __le32_to_cpu(migr_rec->ckpt_area_pba) * 512;
10875
10876 write_offset = ((unsigned long long)
10877 __le32_to_cpu(migr_rec->dest_1st_member_lba) +
5551b113 10878 pba_of_lba0(map_dest)) * 512;
276d77db
AK		 10879
10880 unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
f36a9ecd 10881 if (posix_memalign((void **)&buf, sector_size, unit_len) != 0)
276d77db 10882 goto abort;
503975b9 10883 targets = xcalloc(new_disks, sizeof(int));
276d77db 10884
9a717282 10885 if (open_backup_targets(info, new_disks, targets, super, id->dev)) {
e7b84f9d 10886 pr_err("Cannot open some devices belonging to array.\n");
f627f5ad
AK		 10887 goto abort;
10888 }
276d77db
AK		 10889
10890 for (i = 0; i < new_disks; i++) {
6c3560c0
AK		 10891 if (targets[i] < 0) {
10892 skipped_disks++;
10893 continue;
10894 }
276d77db 10895 if (lseek64(targets[i], read_offset, SEEK_SET) < 0) {
e7b84f9d
N		 10896 pr_err("Cannot seek to block: %s\n",
10897 strerror(errno));
137debce
AK		 10898 skipped_disks++;
10899 continue;
276d77db 10900 }
9ec11d1a 10901 if ((unsigned)read(targets[i], buf, unit_len) != unit_len) {
e7b84f9d
N		 10902 pr_err("Cannot read copy area block: %s\n",
10903 strerror(errno));
137debce
AK		 10904 skipped_disks++;
10905 continue;
276d77db
AK		 10906 }
10907 if (lseek64(targets[i], write_offset, SEEK_SET) < 0) {
e7b84f9d
N		 10908 pr_err("Cannot seek to block: %s\n",
10909 strerror(errno));
137debce
AK		 10910 skipped_disks++;
10911 continue;
276d77db 10912 }
9ec11d1a 10913 if ((unsigned)write(targets[i], buf, unit_len) != unit_len) {
e7b84f9d
N		 10914 pr_err("Cannot restore block: %s\n",
10915 strerror(errno));
137debce
AK		 10916 skipped_disks++;
10917 continue;
276d77db
AK		 10918 }
10919 }
10920
137debce
AK		 10921 if (skipped_disks > imsm_get_allowed_degradation(info->new_level,
10922 new_disks,
10923 super,
10924 id->dev)) {
7a862a02 10925 pr_err("Cannot restore data from backup. Too many failed disks\n");
6c3560c0
AK		 10926 goto abort;
10927 }
10928
befb629b
AK		 10929 if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) {
10930 /* ignore error == 2, this can mean end of reshape here
10931 */
7a862a02 10932 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL) during restart\n");
befb629b 10933 } else
276d77db 10934 retval = 0;
276d77db
AK		 10935
10936abort:
10937 if (targets) {
10938 for (i = 0; i < new_disks; i++)
10939 if (targets[i])
10940 close(targets[i]);
10941 free(targets);
10942 }
10943 free(buf);
10944 return retval;
10945}
10946
2cda7640
ML		 10947static char disk_by_path[] = "/dev/disk/by-path/";
10948
10949static const char *imsm_get_disk_controller_domain(const char *path)
10950{
2cda7640 10951 char disk_path[PATH_MAX];
96234762
LM		 10952 char *drv=NULL;
10953 struct stat st;
2cda7640 10954
6d8d290a 10955 strcpy(disk_path, disk_by_path);
96234762
LM		 10956 strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1);
10957 if (stat(disk_path, &st) == 0) {
10958 struct sys_dev* hba;
594dc1b8 10959 char *path;
96234762
LM		 10960
10961 path = devt_to_devpath(st.st_rdev);
10962 if (path == NULL)
10963 return "unknown";
10964 hba = find_disk_attached_hba(-1, path);
10965 if (hba && hba->type == SYS_DEV_SAS)
10966 drv = "isci";
10967 else if (hba && hba->type == SYS_DEV_SATA)
10968 drv = "ahci";
c6839718
MT		 10969 else if (hba && hba->type == SYS_DEV_VMD)
10970 drv = "vmd";
10971 else if (hba && hba->type == SYS_DEV_NVME)
10972 drv = "nvme";
1011e834 10973 else
96234762
LM		 10974 drv = "unknown";
10975 dprintf("path: %s hba: %s attached: %s\n",
10976 path, (hba) ? hba->path : "NULL", drv);
10977 free(path);
2cda7640 10978 }
96234762 10979 return drv;
2cda7640
ML		 10980}
10981
4dd2df09 10982static char *imsm_find_array_devnm_by_subdev(int subdev, char *container)
78b10e66 10983{
4dd2df09 10984 static char devnm[32];
78b10e66
N		 10985 char subdev_name[20];
10986 struct mdstat_ent *mdstat;
10987
10988 sprintf(subdev_name, "%d", subdev);
10989 mdstat = mdstat_by_subdev(subdev_name, container);
10990 if (!mdstat)
4dd2df09 10991 return NULL;
78b10e66 10992
4dd2df09 10993 strcpy(devnm, mdstat->devnm);
78b10e66 10994 free_mdstat(mdstat);
4dd2df09 10995 return devnm;
78b10e66
N		 10996}
10997
10998static int imsm_reshape_is_allowed_on_container(struct supertype *st,
10999 struct geo_params *geo,
fbf3d202
AK		 11000 int *old_raid_disks,
11001 int direction)
78b10e66 11002{
694575e7
KW		 11003 /* currently we only support increasing the number of devices
11004 * for a container. This increases the number of device for each
11005 * member array. They must all be RAID0 or RAID5.
11006 */
78b10e66
N		 11007 int ret_val = 0;
11008 struct mdinfo *info, *member;
11009 int devices_that_can_grow = 0;
11010
7a862a02 11011 dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): st->devnm = (%s)\n", st->devnm);
78b10e66 11012
d04f65f4 11013 if (geo->size > 0 ||
78b10e66
N		 11014 geo->level != UnSet ||
11015 geo->layout != UnSet ||
11016 geo->chunksize != 0 ||
11017 geo->raid_disks == UnSet) {
7a862a02 11018 dprintf("imsm: Container operation is allowed for raid disks number change only.\n");
78b10e66
N		 11019 return ret_val;
11020 }
11021
fbf3d202 11022 if (direction == ROLLBACK_METADATA_CHANGES) {
7a862a02 11023 dprintf("imsm: Metadata changes rollback is not supported for container operation.\n");
fbf3d202
AK		 11024 return ret_val;
11025 }
11026
78b10e66
N		 11027 info = container_content_imsm(st, NULL);
11028 for (member = info; member; member = member->next) {
4dd2df09 11029 char *result;
78b10e66
N		 11030
11031 dprintf("imsm: checking device_num: %i\n",
11032 member->container_member);
11033
d7d205bd 11034 if (geo->raid_disks <= member->array.raid_disks) {
78b10e66
N		 11035 /* we work on container for Online Capacity Expansion
11036 * only so raid_disks has to grow
11037 */
7a862a02 11038 dprintf("imsm: for container operation raid disks increase is required\n");
78b10e66
N		 11039 break;
11040 }
11041
089f9d79 11042 if (info->array.level != 0 && info->array.level != 5) {
78b10e66
N		 11043 /* we cannot use this container with other raid level
11044 */
7a862a02 11045 dprintf("imsm: for container operation wrong raid level (%i) detected\n",
78b10e66
N		 11046 info->array.level);
11047 break;
11048 } else {
11049 /* check for platform support
11050 * for this raid level configuration
11051 */
11052 struct intel_super *super = st->sb;
11053 if (!is_raid_level_supported(super->orom,
11054 member->array.level,
11055 geo->raid_disks)) {
7a862a02 11056 dprintf("platform does not support raid%d with %d disk%s\n",
78b10e66
N		 11057 info->array.level,
11058 geo->raid_disks,
11059 geo->raid_disks > 1 ? "s" : "");
11060 break;
11061 }
2a4a08e7
AK		 11062 /* check if component size is aligned to chunk size
11063 */
11064 if (info->component_size %
11065 (info->array.chunk_size/512)) {
7a862a02 11066 dprintf("Component size is not aligned to chunk size\n");
2a4a08e7
AK		 11067 break;
11068 }
78b10e66
N		 11069 }
11070
11071 if (*old_raid_disks &&
11072 info->array.raid_disks != *old_raid_disks)
11073 break;
11074 *old_raid_disks = info->array.raid_disks;
11075
11076 /* All raid5 and raid0 volumes in container
11077 * have to be ready for Online Capacity Expansion
11078 * so they need to be assembled. We have already
11079 * checked that no recovery etc is happening.
11080 */
4dd2df09
N		 11081 result = imsm_find_array_devnm_by_subdev(member->container_member,
11082 st->container_devnm);
11083 if (result == NULL) {
78b10e66
N		 11084 dprintf("imsm: cannot find array\n");
11085 break;
11086 }
11087 devices_that_can_grow++;
11088 }
11089 sysfs_free(info);
11090 if (!member && devices_that_can_grow)
11091 ret_val = 1;
11092
11093 if (ret_val)
1ade5cc1 11094 dprintf("Container operation allowed\n");
78b10e66 11095 else
1ade5cc1 11096 dprintf("Error: %i\n", ret_val);
78b10e66
N		 11097
11098 return ret_val;
11099}
11100
11101/* Function: get_spares_for_grow
11102 * Description: Allocates memory and creates list of spare devices
1011e834 11103 * avaliable in container. Checks if spare drive size is acceptable.
78b10e66
N		 11104 * Parameters: Pointer to the supertype structure
11105 * Returns: Pointer to the list of spare devices (mdinfo structure) on success,
1011e834 11106 * NULL if fail
78b10e66
N		 11107 */
11108static struct mdinfo *get_spares_for_grow(struct supertype *st)
11109{
fbfdcb06
AO		 11110 struct spare_criteria sc;
11111
11112 get_spare_criteria_imsm(st, &sc);
11113 return container_choose_spares(st, &sc, NULL, NULL, NULL, 0);
78b10e66
N		 11114}
11115
11116/******************************************************************************
11117 * function: imsm_create_metadata_update_for_reshape
11118 * Function creates update for whole IMSM container.
11119 *
11120 ******************************************************************************/
11121static int imsm_create_metadata_update_for_reshape(
11122 struct supertype *st,
11123 struct geo_params *geo,
11124 int old_raid_disks,
11125 struct imsm_update_reshape **updatep)
11126{
11127 struct intel_super *super = st->sb;
11128 struct imsm_super *mpb = super->anchor;
594dc1b8
JS		 11129 int update_memory_size;
11130 struct imsm_update_reshape *u;
11131 struct mdinfo *spares;
78b10e66 11132 int i;
594dc1b8 11133 int delta_disks;
bbd24d86 11134 struct mdinfo *dev;
78b10e66 11135
1ade5cc1 11136 dprintf("(enter) raid_disks = %i\n", geo->raid_disks);
78b10e66
N		 11137
11138 delta_disks = geo->raid_disks - old_raid_disks;
11139
11140 /* size of all update data without anchor */
11141 update_memory_size = sizeof(struct imsm_update_reshape);
11142
11143 /* now add space for spare disks that we need to add. */
11144 update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1);
11145
503975b9 11146 u = xcalloc(1, update_memory_size);
78b10e66
N		 11147 u->type = update_reshape_container_disks;
11148 u->old_raid_disks = old_raid_disks;
11149 u->new_raid_disks = geo->raid_disks;
11150
11151 /* now get spare disks list
11152 */
11153 spares = get_spares_for_grow(st);
11154
d7be7d87 11155 if (spares == NULL || delta_disks > spares->array.spare_disks) {
7a862a02 11156 pr_err("imsm: ERROR: Cannot get spare devices for %s.\n", geo->dev_name);
e4c72d1d 11157 i = -1;
78b10e66
N		 11158 goto abort;
11159 }
11160
11161 /* we have got spares
11162 * update disk list in imsm_disk list table in anchor
11163 */
11164 dprintf("imsm: %i spares are available.\n\n",
11165 spares->array.spare_disks);
11166
bbd24d86 11167 dev = spares->devs;
78b10e66 11168 for (i = 0; i < delta_disks; i++) {
78b10e66
N		 11169 struct dl *dl;
11170
bbd24d86
AK		 11171 if (dev == NULL)
11172 break;
78b10e66
N		 11173 u->new_disks[i] = makedev(dev->disk.major,
11174 dev->disk.minor);
11175 dl = get_disk_super(super, dev->disk.major, dev->disk.minor);
ee4beede
AK		 11176 dl->index = mpb->num_disks;
11177 mpb->num_disks++;
bbd24d86 11178 dev = dev->next;
78b10e66 11179 }
78b10e66
N		 11180
11181abort:
11182 /* free spares
11183 */
11184 sysfs_free(spares);
11185
d677e0b8 11186 dprintf("imsm: reshape update preparation :");
78b10e66 11187 if (i == delta_disks) {
1ade5cc1 11188 dprintf_cont(" OK\n");
78b10e66
N		 11189 *updatep = u;
11190 return update_memory_size;
11191 }
11192 free(u);
1ade5cc1 11193 dprintf_cont(" Error\n");
78b10e66
N		 11194
11195 return 0;
11196}
11197
f3871fdc
AK		 11198/******************************************************************************
11199 * function: imsm_create_metadata_update_for_size_change()
11200 * Creates update for IMSM array for array size change.
11201 *
11202 ******************************************************************************/
11203static int imsm_create_metadata_update_for_size_change(
11204 struct supertype *st,
11205 struct geo_params *geo,
11206 struct imsm_update_size_change **updatep)
11207{
11208 struct intel_super *super = st->sb;
594dc1b8
JS		 11209 int update_memory_size;
11210 struct imsm_update_size_change *u;
f3871fdc 11211
1ade5cc1 11212 dprintf("(enter) New size = %llu\n", geo->size);
f3871fdc
AK		 11213
11214 /* size of all update data without anchor */
11215 update_memory_size = sizeof(struct imsm_update_size_change);
11216
503975b9 11217 u = xcalloc(1, update_memory_size);
f3871fdc
AK		 11218 u->type = update_size_change;
11219 u->subdev = super->current_vol;
11220 u->new_size = geo->size;
11221
11222 dprintf("imsm: reshape update preparation : OK\n");
11223 *updatep = u;
11224
11225 return update_memory_size;
11226}
11227
48c5303a
PC		 11228/******************************************************************************
11229 * function: imsm_create_metadata_update_for_migration()
11230 * Creates update for IMSM array.
11231 *
11232 ******************************************************************************/
11233static int imsm_create_metadata_update_for_migration(
11234 struct supertype *st,
11235 struct geo_params *geo,
11236 struct imsm_update_reshape_migration **updatep)
11237{
11238 struct intel_super *super = st->sb;
594dc1b8
JS		 11239 int update_memory_size;
11240 struct imsm_update_reshape_migration *u;
48c5303a
PC		 11241 struct imsm_dev *dev;
11242 int previous_level = -1;
11243
1ade5cc1 11244 dprintf("(enter) New Level = %i\n", geo->level);
48c5303a
PC		 11245
11246 /* size of all update data without anchor */
11247 update_memory_size = sizeof(struct imsm_update_reshape_migration);
11248
503975b9 11249 u = xcalloc(1, update_memory_size);
48c5303a
PC		 11250 u->type = update_reshape_migration;
11251 u->subdev = super->current_vol;
11252 u->new_level = geo->level;
11253 u->new_layout = geo->layout;
11254 u->new_raid_disks = u->old_raid_disks = geo->raid_disks;
11255 u->new_disks[0] = -1;
4bba0439 11256 u->new_chunksize = -1;
48c5303a
PC		 11257
11258 dev = get_imsm_dev(super, u->subdev);
11259 if (dev) {
11260 struct imsm_map *map;
11261
238c0a71 11262 map = get_imsm_map(dev, MAP_0);
4bba0439
PC		 11263 if (map) {
11264 int current_chunk_size =
11265 __le16_to_cpu(map->blocks_per_strip) / 2;
11266
11267 if (geo->chunksize != current_chunk_size) {
11268 u->new_chunksize = geo->chunksize / 1024;
7a862a02 11269 dprintf("imsm: chunk size change from %i to %i\n",
4bba0439
PC		 11270 current_chunk_size, u->new_chunksize);
11271 }
48c5303a 11272 previous_level = map->raid_level;
4bba0439 11273 }
48c5303a 11274 }
089f9d79 11275 if (geo->level == 5 && previous_level == 0) {
48c5303a
PC		 11276 struct mdinfo *spares = NULL;
11277
11278 u->new_raid_disks++;
11279 spares = get_spares_for_grow(st);
089f9d79 11280 if (spares == NULL || spares->array.spare_disks < 1) {
48c5303a
PC		 11281 free(u);
11282 sysfs_free(spares);
11283 update_memory_size = 0;
565cc99e 11284 pr_err("cannot get spare device for requested migration\n");
48c5303a
PC		 11285 return 0;
11286 }
11287 sysfs_free(spares);
11288 }
11289 dprintf("imsm: reshape update preparation : OK\n");
11290 *updatep = u;
11291
11292 return update_memory_size;
11293}
11294
8dd70bce
AK		 11295static void imsm_update_metadata_locally(struct supertype *st,
11296 void *buf, int len)
11297{
11298 struct metadata_update mu;
11299
11300 mu.buf = buf;
11301 mu.len = len;
11302 mu.space = NULL;
11303 mu.space_list = NULL;
11304 mu.next = NULL;
5fe6f031
N		 11305 if (imsm_prepare_update(st, &mu))
11306 imsm_process_update(st, &mu);
8dd70bce
AK		 11307
11308 while (mu.space_list) {
11309 void **space = mu.space_list;
11310 mu.space_list = *space;
11311 free(space);
11312 }
11313}
78b10e66 11314
471bceb6 11315/***************************************************************************
694575e7 11316* Function: imsm_analyze_change
471bceb6 11317* Description: Function analyze change for single volume
1011e834 11318* and validate if transition is supported
fbf3d202
AK		 11319* Parameters: Geometry parameters, supertype structure,
11320* metadata change direction (apply/rollback)
694575e7 11321* Returns: Operation type code on success, -1 if fail
471bceb6
KW		 11322****************************************************************************/
11323enum imsm_reshape_type imsm_analyze_change(struct supertype *st,
fbf3d202
AK		 11324 struct geo_params *geo,
11325 int direction)
694575e7 11326{
471bceb6
KW		 11327 struct mdinfo info;
11328 int change = -1;
11329 int check_devs = 0;
c21e737b 11330 int chunk;
67a2db32
AK		 11331 /* number of added/removed disks in operation result */
11332 int devNumChange = 0;
11333 /* imsm compatible layout value for array geometry verification */
11334 int imsm_layout = -1;
7abc9871
AK		 11335 int data_disks;
11336 struct imsm_dev *dev;
9529d343 11337 struct imsm_map *map;
7abc9871 11338 struct intel_super *super;
d04f65f4 11339 unsigned long long current_size;
65d38cca 11340 unsigned long long free_size;
d04f65f4 11341 unsigned long long max_size;
65d38cca 11342 int rv;
471bceb6
KW		 11343
11344 getinfo_super_imsm_volume(st, &info, NULL);
089f9d79
JS		 11345 if (geo->level != info.array.level && geo->level >= 0 &&
11346 geo->level != UnSet) {
471bceb6
KW		 11347 switch (info.array.level) {
11348 case 0:
11349 if (geo->level == 5) {
b5347799 11350 change = CH_MIGRATION;
e13ce846 11351 if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) {
7a862a02 11352 pr_err("Error. Requested Layout not supported (left-asymmetric layout is supported only)!\n");
e13ce846
AK		 11353 change = -1;
11354 goto analyse_change_exit;
11355 }
67a2db32 11356 imsm_layout = geo->layout;
471bceb6 11357 check_devs = 1;
e91a3bad
LM		 11358 devNumChange = 1; /* parity disk added */
11359 } else if (geo->level == 10) {
471bceb6
KW		 11360 change = CH_TAKEOVER;
11361 check_devs = 1;
e91a3bad 11362 devNumChange = 2; /* two mirrors added */
67a2db32 11363 imsm_layout = 0x102; /* imsm supported layout */
471bceb6 11364 }
dfe77a9e
KW		 11365 break;
11366 case 1:
471bceb6
KW		 11367 case 10:
11368 if (geo->level == 0) {
11369 change = CH_TAKEOVER;
11370 check_devs = 1;
e91a3bad 11371 devNumChange = -(geo->raid_disks/2);
67a2db32 11372 imsm_layout = 0; /* imsm raid0 layout */
471bceb6
KW		 11373 }
11374 break;
11375 }
11376 if (change == -1) {
7a862a02 11377 pr_err("Error. Level Migration from %d to %d not supported!\n",
e7b84f9d 11378 info.array.level, geo->level);
471bceb6
KW		 11379 goto analyse_change_exit;
11380 }
11381 } else
11382 geo->level = info.array.level;
11383
089f9d79
JS		 11384 if (geo->layout != info.array.layout &&
11385 (geo->layout != UnSet && geo->layout != -1)) {
b5347799 11386 change = CH_MIGRATION;
089f9d79
JS		 11387 if (info.array.layout == 0 && info.array.level == 5 &&
11388 geo->layout == 5) {
471bceb6 11389 /* reshape 5 -> 4 */
089f9d79
JS		 11390 } else if (info.array.layout == 5 && info.array.level == 5 &&
11391 geo->layout == 0) {
471bceb6
KW		 11392 /* reshape 4 -> 5 */
11393 geo->layout = 0;
11394 geo->level = 5;
11395 } else {
7a862a02 11396 pr_err("Error. Layout Migration from %d to %d not supported!\n",
e7b84f9d 11397 info.array.layout, geo->layout);
471bceb6
KW		 11398 change = -1;
11399 goto analyse_change_exit;
11400 }
67a2db32 11401 } else {
471bceb6 11402 geo->layout = info.array.layout;
67a2db32
AK		 11403 if (imsm_layout == -1)
11404 imsm_layout = info.array.layout;
11405 }
471bceb6 11406
089f9d79
JS		 11407 if (geo->chunksize > 0 && geo->chunksize != UnSet &&
11408 geo->chunksize != info.array.chunk_size) {
2d2b0eb7
MD		 11409 if (info.array.level == 10) {
11410 pr_err("Error. Chunk size change for RAID 10 is not supported.\n");
11411 change = -1;
11412 goto analyse_change_exit;
1e9b2c3f
PB		 11413 } else if (info.component_size % (geo->chunksize/512)) {
11414 pr_err("New chunk size (%dK) does not evenly divide device size (%lluk). Aborting...\n",
11415 geo->chunksize/1024, info.component_size/2);
11416 change = -1;
11417 goto analyse_change_exit;
2d2b0eb7 11418 }
b5347799 11419 change = CH_MIGRATION;
2d2b0eb7 11420 } else {
471bceb6 11421 geo->chunksize = info.array.chunk_size;
2d2b0eb7 11422 }
471bceb6 11423
c21e737b 11424 chunk = geo->chunksize / 1024;
7abc9871
AK		 11425
11426 super = st->sb;
11427 dev = get_imsm_dev(super, super->current_vol);
9529d343
MD		 11428 map = get_imsm_map(dev, MAP_0);
11429 data_disks = imsm_num_data_members(map);
c41e00b2 11430 /* compute current size per disk member
7abc9871 11431 */
c41e00b2
AK		 11432 current_size = info.custom_array_size / data_disks;
11433
089f9d79 11434 if (geo->size > 0 && geo->size != MAX_SIZE) {
c41e00b2
AK		 11435 /* align component size
11436 */
3e684231 11437 geo->size = imsm_component_size_alignment_check(
c41e00b2 11438 get_imsm_raid_level(dev->vol.map),
f36a9ecd 11439 chunk * 1024, super->sector_size,
c41e00b2 11440 geo->size * 2);
65d0b4ce 11441 if (geo->size == 0) {
7a862a02 11442 pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is 0).\n",
65d0b4ce
LD		 11443 current_size);
11444 goto analyse_change_exit;
11445 }
c41e00b2 11446 }
7abc9871 11447
089f9d79 11448 if (current_size != geo->size && geo->size > 0) {
7abc9871 11449 if (change != -1) {
7a862a02 11450 pr_err("Error. Size change should be the only one at a time.\n");
7abc9871
AK		 11451 change = -1;
11452 goto analyse_change_exit;
11453 }
11454 if ((super->current_vol + 1) != super->anchor->num_raid_devs) {
7a862a02 11455 pr_err("Error. The last volume in container can be expanded only (%i/%s).\n",
4dd2df09 11456 super->current_vol, st->devnm);
7abc9871
AK		 11457 goto analyse_change_exit;
11458 }
65d38cca
LD		 11459 /* check the maximum available size
11460 */
11461 rv = imsm_get_free_size(st, dev->vol.map->num_members,
11462 0, chunk, &free_size);
11463 if (rv == 0)
11464 /* Cannot find maximum available space
11465 */
11466 max_size = 0;
11467 else {
11468 max_size = free_size + current_size;
11469 /* align component size
11470 */
3e684231 11471 max_size = imsm_component_size_alignment_check(
65d38cca 11472 get_imsm_raid_level(dev->vol.map),
f36a9ecd 11473 chunk * 1024, super->sector_size,
65d38cca
LD		 11474 max_size);
11475 }
d04f65f4 11476 if (geo->size == MAX_SIZE) {
b130333f
AK		 11477 /* requested size change to the maximum available size
11478 */
65d38cca 11479 if (max_size == 0) {
7a862a02 11480 pr_err("Error. Cannot find maximum available space.\n");
b130333f
AK		 11481 change = -1;
11482 goto analyse_change_exit;
65d38cca
LD		 11483 } else
11484 geo->size = max_size;
c41e00b2 11485 }
b130333f 11486
681b7ae2 11487 if (direction == ROLLBACK_METADATA_CHANGES) {
fbf3d202
AK		 11488 /* accept size for rollback only
11489 */
11490 } else {
11491 /* round size due to metadata compatibility
11492 */
11493 geo->size = (geo->size >> SECT_PER_MB_SHIFT)
11494 << SECT_PER_MB_SHIFT;
11495 dprintf("Prepare update for size change to %llu\n",
11496 geo->size );
11497 if (current_size >= geo->size) {
7a862a02 11498 pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is %llu).\n",
e7b84f9d 11499 current_size, geo->size);
fbf3d202
AK		 11500 goto analyse_change_exit;
11501 }
65d38cca 11502 if (max_size && geo->size > max_size) {
7a862a02 11503 pr_err("Error. Requested size is larger than maximum available size (maximum available size is %llu, requested size /rounded/ is %llu).\n",
e7b84f9d 11504 max_size, geo->size);
65d38cca
LD		 11505 goto analyse_change_exit;
11506 }
7abc9871
AK		 11507 }
11508 geo->size *= data_disks;
11509 geo->raid_disks = dev->vol.map->num_members;
11510 change = CH_ARRAY_SIZE;
11511 }
471bceb6
KW		 11512 if (!validate_geometry_imsm(st,
11513 geo->level,
67a2db32 11514 imsm_layout,
e91a3bad 11515 geo->raid_disks + devNumChange,
c21e737b 11516 &chunk,
af4348dd 11517 geo->size, INVALID_SECTORS,
5308f117 11518 0, 0, info.consistency_policy, 1))
471bceb6
KW		 11519 change = -1;
11520
11521 if (check_devs) {
11522 struct intel_super *super = st->sb;
11523 struct imsm_super *mpb = super->anchor;
11524
11525 if (mpb->num_raid_devs > 1) {
7a862a02 11526 pr_err("Error. Cannot perform operation on %s- for this operation it MUST be single array in container\n",
e7b84f9d 11527 geo->dev_name);
471bceb6
KW		 11528 change = -1;
11529 }
11530 }
11531
11532analyse_change_exit:
089f9d79
JS		 11533 if (direction == ROLLBACK_METADATA_CHANGES &&
11534 (change == CH_MIGRATION || change == CH_TAKEOVER)) {
7a862a02 11535 dprintf("imsm: Metadata changes rollback is not supported for migration and takeover operations.\n");
fbf3d202
AK		 11536 change = -1;
11537 }
471bceb6 11538 return change;
694575e7
KW		 11539}
11540
bb025c2f
KW		 11541int imsm_takeover(struct supertype *st, struct geo_params *geo)
11542{
11543 struct intel_super *super = st->sb;
11544 struct imsm_update_takeover *u;
11545
503975b9 11546 u = xmalloc(sizeof(struct imsm_update_takeover));
bb025c2f
KW		 11547
11548 u->type = update_takeover;
11549 u->subarray = super->current_vol;
11550
11551 /* 10->0 transition */
11552 if (geo->level == 0)
11553 u->direction = R10_TO_R0;
11554
0529c688
KW		 11555 /* 0->10 transition */
11556 if (geo->level == 10)
11557 u->direction = R0_TO_R10;
11558
bb025c2f
KW		 11559 /* update metadata locally */
11560 imsm_update_metadata_locally(st, u,
11561 sizeof(struct imsm_update_takeover));
11562 /* and possibly remotely */
11563 if (st->update_tail)
11564 append_metadata_update(st, u,
11565 sizeof(struct imsm_update_takeover));
11566 else
11567 free(u);
11568
11569 return 0;
11570}
11571
d04f65f4
N		 11572static int imsm_reshape_super(struct supertype *st, unsigned long long size,
11573 int level,
78b10e66 11574 int layout, int chunksize, int raid_disks,
41784c88 11575 int delta_disks, char *backup, char *dev,
016e00f5 11576 int direction, int verbose)
78b10e66 11577{
78b10e66
N		 11578 int ret_val = 1;
11579 struct geo_params geo;
11580
1ade5cc1 11581 dprintf("(enter)\n");
78b10e66 11582
71204a50 11583 memset(&geo, 0, sizeof(struct geo_params));
78b10e66
N		 11584
11585 geo.dev_name = dev;
4dd2df09 11586 strcpy(geo.devnm, st->devnm);
78b10e66
N		 11587 geo.size = size;
11588 geo.level = level;
11589 geo.layout = layout;
11590 geo.chunksize = chunksize;
11591 geo.raid_disks = raid_disks;
41784c88
AK		 11592 if (delta_disks != UnSet)
11593 geo.raid_disks += delta_disks;
78b10e66 11594
1ade5cc1
N		 11595 dprintf("for level : %i\n", geo.level);
11596 dprintf("for raid_disks : %i\n", geo.raid_disks);
78b10e66
N		 11597
11598 if (experimental() == 0)
11599 return ret_val;
11600
4dd2df09 11601 if (strcmp(st->container_devnm, st->devnm) == 0) {
694575e7
KW		 11602 /* On container level we can only increase number of devices. */
11603 dprintf("imsm: info: Container operation\n");
78b10e66 11604 int old_raid_disks = 0;
6dc0be30 11605
78b10e66 11606 if (imsm_reshape_is_allowed_on_container(
fbf3d202 11607 st, &geo, &old_raid_disks, direction)) {
78b10e66
N		 11608 struct imsm_update_reshape *u = NULL;
11609 int len;
11610
11611 len = imsm_create_metadata_update_for_reshape(
11612 st, &geo, old_raid_disks, &u);
11613
ed08d51c
AK		 11614 if (len <= 0) {
11615 dprintf("imsm: Cannot prepare update\n");
11616 goto exit_imsm_reshape_super;
11617 }
11618
8dd70bce
AK		 11619 ret_val = 0;
11620 /* update metadata locally */
11621 imsm_update_metadata_locally(st, u, len);
11622 /* and possibly remotely */
11623 if (st->update_tail)
11624 append_metadata_update(st, u, len);
11625 else
ed08d51c 11626 free(u);
8dd70bce 11627
694575e7 11628 } else {
7a862a02 11629 pr_err("(imsm) Operation is not allowed on this container\n");
694575e7
KW		 11630 }
11631 } else {
11632 /* On volume level we support following operations
471bceb6
KW		 11633 * - takeover: raid10 -> raid0; raid0 -> raid10
11634 * - chunk size migration
11635 * - migration: raid5 -> raid0; raid0 -> raid5
11636 */
11637 struct intel_super *super = st->sb;
11638 struct intel_dev *dev = super->devlist;
4dd2df09 11639 int change;
694575e7 11640 dprintf("imsm: info: Volume operation\n");
471bceb6
KW		 11641 /* find requested device */
11642 while (dev) {
1011e834 11643 char *devnm =
4dd2df09
N		 11644 imsm_find_array_devnm_by_subdev(
11645 dev->index, st->container_devnm);
11646 if (devnm && strcmp(devnm, geo.devnm) == 0)
471bceb6
KW		 11647 break;
11648 dev = dev->next;
11649 }
11650 if (dev == NULL) {
4dd2df09
N		 11651 pr_err("Cannot find %s (%s) subarray\n",
11652 geo.dev_name, geo.devnm);
471bceb6
KW		 11653 goto exit_imsm_reshape_super;
11654 }
11655 super->current_vol = dev->index;
fbf3d202 11656 change = imsm_analyze_change(st, &geo, direction);
694575e7 11657 switch (change) {
471bceb6 11658 case CH_TAKEOVER:
bb025c2f 11659 ret_val = imsm_takeover(st, &geo);
694575e7 11660 break;
48c5303a
PC		 11661 case CH_MIGRATION: {
11662 struct imsm_update_reshape_migration *u = NULL;
11663 int len =
11664 imsm_create_metadata_update_for_migration(
11665 st, &geo, &u);
11666 if (len < 1) {
7a862a02 11667 dprintf("imsm: Cannot prepare update\n");
48c5303a
PC		 11668 break;
11669 }
471bceb6 11670 ret_val = 0;
48c5303a
PC		 11671 /* update metadata locally */
11672 imsm_update_metadata_locally(st, u, len);
11673 /* and possibly remotely */
11674 if (st->update_tail)
11675 append_metadata_update(st, u, len);
11676 else
11677 free(u);
11678 }
11679 break;
7abc9871 11680 case CH_ARRAY_SIZE: {
f3871fdc
AK		 11681 struct imsm_update_size_change *u = NULL;
11682 int len =
11683 imsm_create_metadata_update_for_size_change(
11684 st, &geo, &u);
11685 if (len < 1) {
7a862a02 11686 dprintf("imsm: Cannot prepare update\n");
f3871fdc
AK		 11687 break;
11688 }
11689 ret_val = 0;
11690 /* update metadata locally */
11691 imsm_update_metadata_locally(st, u, len);
11692 /* and possibly remotely */
11693 if (st->update_tail)
11694 append_metadata_update(st, u, len);
11695 else
11696 free(u);
7abc9871
AK		 11697 }
11698 break;
471bceb6
KW		 11699 default:
11700 ret_val = 1;
694575e7 11701 }
694575e7 11702 }
78b10e66 11703
ed08d51c 11704exit_imsm_reshape_super:
78b10e66
N		 11705 dprintf("imsm: reshape_super Exit code = %i\n", ret_val);
11706 return ret_val;
11707}
2cda7640 11708
0febb20c
AO		 11709#define COMPLETED_OK 0
11710#define COMPLETED_NONE 1
11711#define COMPLETED_DELAYED 2
11712
11713static int read_completed(int fd, unsigned long long *val)
11714{
11715 int ret;
11716 char buf[50];
11717
11718 ret = sysfs_fd_get_str(fd, buf, 50);
11719 if (ret < 0)
11720 return ret;
11721
11722 ret = COMPLETED_OK;
11723 if (strncmp(buf, "none", 4) == 0) {
11724 ret = COMPLETED_NONE;
11725 } else if (strncmp(buf, "delayed", 7) == 0) {
11726 ret = COMPLETED_DELAYED;
11727 } else {
11728 char *ep;
11729 *val = strtoull(buf, &ep, 0);
11730 if (ep == buf || (*ep != 0 && *ep != '\n' && *ep != ' '))
11731 ret = -1;
11732 }
11733 return ret;
11734}
11735
eee67a47
AK		 11736/*******************************************************************************
11737 * Function: wait_for_reshape_imsm
11738 * Description: Function writes new sync_max value and waits until
11739 * reshape process reach new position
11740 * Parameters:
11741 * sra : general array info
eee67a47
AK		 11742 * ndata : number of disks in new array's layout
11743 * Returns:
11744 * 0 : success,
11745 * 1 : there is no reshape in progress,
11746 * -1 : fail
11747 ******************************************************************************/
ae9f01f8 11748int wait_for_reshape_imsm(struct mdinfo *sra, int ndata)
eee67a47 11749{
85ca499c 11750 int fd = sysfs_get_fd(sra, NULL, "sync_completed");
df2647fa 11751 int retry = 3;
eee67a47 11752 unsigned long long completed;
ae9f01f8
AK		 11753 /* to_complete : new sync_max position */
11754 unsigned long long to_complete = sra->reshape_progress;
11755 unsigned long long position_to_set = to_complete / ndata;
eee67a47 11756
ae9f01f8 11757 if (fd < 0) {
1ade5cc1 11758 dprintf("cannot open reshape_position\n");
eee67a47 11759 return 1;
ae9f01f8 11760 }
eee67a47 11761
df2647fa
PB		 11762 do {
11763 if (sysfs_fd_get_ll(fd, &completed) < 0) {
11764 if (!retry) {
11765 dprintf("cannot read reshape_position (no reshape in progres)\n");
11766 close(fd);
11767 return 1;
11768 }
11769 usleep(30000);
11770 } else
11771 break;
11772 } while (retry--);
eee67a47 11773
85ca499c 11774 if (completed > position_to_set) {
1ade5cc1 11775 dprintf("wrong next position to set %llu (%llu)\n",
85ca499c 11776 to_complete, position_to_set);
ae9f01f8
AK		 11777 close(fd);
11778 return -1;
11779 }
11780 dprintf("Position set: %llu\n", position_to_set);
11781 if (sysfs_set_num(sra, NULL, "sync_max",
11782 position_to_set) != 0) {
1ade5cc1 11783 dprintf("cannot set reshape position to %llu\n",
ae9f01f8
AK		 11784 position_to_set);
11785 close(fd);
11786 return -1;
eee67a47
AK		 11787 }
11788
eee67a47 11789 do {
0febb20c 11790 int rc;
eee67a47 11791 char action[20];
5ff3a780 11792 int timeout = 3000;
0febb20c 11793
5ff3a780 11794 sysfs_wait(fd, &timeout);
a47e44fb
AK		 11795 if (sysfs_get_str(sra, NULL, "sync_action",
11796 action, 20) > 0 &&
d7d3809a 11797 strncmp(action, "reshape", 7) != 0) {
b2be2b62
AO		 11798 if (strncmp(action, "idle", 4) == 0)
11799 break;
d7d3809a
AP		 11800 close(fd);
11801 return -1;
11802 }
0febb20c
AO		 11803
11804 rc = read_completed(fd, &completed);
11805 if (rc < 0) {
1ade5cc1 11806 dprintf("cannot read reshape_position (in loop)\n");
eee67a47
AK		 11807 close(fd);
11808 return 1;
0febb20c
AO		 11809 } else if (rc == COMPLETED_NONE)
11810 break;
85ca499c 11811 } while (completed < position_to_set);
b2be2b62 11812
eee67a47
AK		 11813 close(fd);
11814 return 0;
eee67a47
AK		 11815}
11816
b915c95f
AK		 11817/*******************************************************************************
11818 * Function: check_degradation_change
11819 * Description: Check that array hasn't become failed.
11820 * Parameters:
11821 * info : for sysfs access
11822 * sources : source disks descriptors
11823 * degraded: previous degradation level
11824 * Returns:
11825 * degradation level
11826 ******************************************************************************/
11827int check_degradation_change(struct mdinfo *info,
11828 int *sources,
11829 int degraded)
11830{
11831 unsigned long long new_degraded;
e1993023
LD		 11832 int rv;
11833
11834 rv = sysfs_get_ll(info, NULL, "degraded", &new_degraded);
089f9d79 11835 if (rv == -1 || (new_degraded != (unsigned long long)degraded)) {
b915c95f
AK		 11836 /* check each device to ensure it is still working */
11837 struct mdinfo *sd;
11838 new_degraded = 0;
11839 for (sd = info->devs ; sd ; sd = sd->next) {
11840 if (sd->disk.state & (1<<MD_DISK_FAULTY))
11841 continue;
11842 if (sd->disk.state & (1<<MD_DISK_SYNC)) {
cf52eff5
TM		 11843 char sbuf[100];
11844
b915c95f 11845 if (sysfs_get_str(info,
cf52eff5 11846 sd, "state", sbuf, sizeof(sbuf)) < 0 ||
b915c95f
AK		 11847 strstr(sbuf, "faulty") ||
11848 strstr(sbuf, "in_sync") == NULL) {
11849 /* this device is dead */
11850 sd->disk.state = (1<<MD_DISK_FAULTY);
11851 if (sd->disk.raid_disk >= 0 &&
11852 sources[sd->disk.raid_disk] >= 0) {
11853 close(sources[
11854 sd->disk.raid_disk]);
11855 sources[sd->disk.raid_disk] =
11856 -1;
11857 }
11858 new_degraded++;
11859 }
11860 }
11861 }
11862 }
11863
11864 return new_degraded;
11865}
11866
10f22854
AK		 11867/*******************************************************************************
11868 * Function: imsm_manage_reshape
11869 * Description: Function finds array under reshape and it manages reshape
11870 * process. It creates stripes backups (if required) and sets
942e1cdb 11871 * checkpoints.
10f22854
AK		 11872 * Parameters:
11873 * afd : Backup handle (nattive) - not used
11874 * sra : general array info
11875 * reshape : reshape parameters - not used
11876 * st : supertype structure
11877 * blocks : size of critical section [blocks]
11878 * fds : table of source device descriptor
11879 * offsets : start of array (offest per devices)
11880 * dests : not used
11881 * destfd : table of destination device descriptor
11882 * destoffsets : table of destination offsets (per device)
11883 * Returns:
11884 * 1 : success, reshape is done
11885 * 0 : fail
11886 ******************************************************************************/
999b4972
N		 11887static int imsm_manage_reshape(
11888 int afd, struct mdinfo *sra, struct reshape *reshape,
10f22854 11889 struct supertype *st, unsigned long backup_blocks,
999b4972
N		 11890 int *fds, unsigned long long *offsets,
11891 int dests, int *destfd, unsigned long long *destoffsets)
11892{
10f22854
AK		 11893 int ret_val = 0;
11894 struct intel_super *super = st->sb;
594dc1b8 11895 struct intel_dev *dv;
de44e46f 11896 unsigned int sector_size = super->sector_size;
10f22854 11897 struct imsm_dev *dev = NULL;
9529d343 11898 struct imsm_map *map_src, *map_dest;
10f22854
AK		 11899 int migr_vol_qan = 0;
11900 int ndata, odata; /* [bytes] */
11901 int chunk; /* [bytes] */
11902 struct migr_record *migr_rec;
11903 char *buf = NULL;
11904 unsigned int buf_size; /* [bytes] */
11905 unsigned long long max_position; /* array size [bytes] */
11906 unsigned long long next_step; /* [blocks]/[bytes] */
11907 unsigned long long old_data_stripe_length;
10f22854
AK		 11908 unsigned long long start_src; /* [bytes] */
11909 unsigned long long start; /* [bytes] */
11910 unsigned long long start_buf_shift; /* [bytes] */
b915c95f 11911 int degraded = 0;
ab724b98 11912 int source_layout = 0;
10f22854 11913
79a16a9b
JS		 11914 if (!sra)
11915 return ret_val;
11916
11917 if (!fds || !offsets)
10f22854
AK		 11918 goto abort;
11919
11920 /* Find volume during the reshape */
11921 for (dv = super->devlist; dv; dv = dv->next) {
fc54fe7a
JS		 11922 if (dv->dev->vol.migr_type == MIGR_GEN_MIGR &&
11923 dv->dev->vol.migr_state == 1) {
10f22854
AK		 11924 dev = dv->dev;
11925 migr_vol_qan++;
11926 }
11927 }
11928 /* Only one volume can migrate at the same time */
11929 if (migr_vol_qan != 1) {
676e87a8 11930 pr_err("%s", migr_vol_qan ?
10f22854
AK		 11931 "Number of migrating volumes greater than 1\n" :
11932 "There is no volume during migrationg\n");
11933 goto abort;
11934 }
11935
9529d343 11936 map_dest = get_imsm_map(dev, MAP_0);
238c0a71 11937 map_src = get_imsm_map(dev, MAP_1);
10f22854
AK		 11938 if (map_src == NULL)
11939 goto abort;
10f22854 11940
9529d343
MD		 11941 ndata = imsm_num_data_members(map_dest);
11942 odata = imsm_num_data_members(map_src);
10f22854 11943
7b1ab482 11944 chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512;
10f22854
AK		 11945 old_data_stripe_length = odata * chunk;
11946
11947 migr_rec = super->migr_rec;
11948
10f22854
AK		 11949 /* initialize migration record for start condition */
11950 if (sra->reshape_progress == 0)
11951 init_migr_record_imsm(st, dev, sra);
b2c59438
AK		 11952 else {
11953 if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) {
7a862a02 11954 dprintf("imsm: cannot restart migration when data are present in copy area.\n");
b2c59438
AK		 11955 goto abort;
11956 }
6a75c8ca
AK		 11957 /* Save checkpoint to update migration record for current
11958 * reshape position (in md). It can be farther than current
11959 * reshape position in metadata.
11960 */
11961 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
11962 /* ignore error == 2, this can mean end of reshape here
11963 */
7a862a02 11964 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL, initial save)\n");
6a75c8ca
AK		 11965 goto abort;
11966 }
b2c59438 11967 }
10f22854
AK		 11968
11969 /* size for data */
11970 buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512;
11971 /* extend buffer size for parity disk */
11972 buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
3e684231 11973 /* add space for stripe alignment */
10f22854 11974 buf_size += old_data_stripe_length;
de44e46f
PB		 11975 if (posix_memalign((void **)&buf, MAX_SECTOR_SIZE, buf_size)) {
11976 dprintf("imsm: Cannot allocate checkpoint buffer\n");
10f22854
AK		 11977 goto abort;
11978 }
11979
3ef4403c 11980 max_position = sra->component_size * ndata;
68eb8bc6 11981 source_layout = imsm_level_to_layout(map_src->raid_level);
10f22854
AK		 11982
11983 while (__le32_to_cpu(migr_rec->curr_migr_unit) <
11984 __le32_to_cpu(migr_rec->num_migr_units)) {
11985 /* current reshape position [blocks] */
11986 unsigned long long current_position =
11987 __le32_to_cpu(migr_rec->blocks_per_unit)
11988 * __le32_to_cpu(migr_rec->curr_migr_unit);
11989 unsigned long long border;
11990
b915c95f
AK		 11991 /* Check that array hasn't become failed.
11992 */
11993 degraded = check_degradation_change(sra, fds, degraded);
11994 if (degraded > 1) {
7a862a02 11995 dprintf("imsm: Abort reshape due to degradation level (%i)\n", degraded);
b915c95f
AK		 11996 goto abort;
11997 }
11998
10f22854
AK		 11999 next_step = __le32_to_cpu(migr_rec->blocks_per_unit);
12000
12001 if ((current_position + next_step) > max_position)
12002 next_step = max_position - current_position;
12003
92144abf 12004 start = current_position * 512;
10f22854 12005
942e1cdb 12006 /* align reading start to old geometry */
10f22854
AK		 12007 start_buf_shift = start % old_data_stripe_length;
12008 start_src = start - start_buf_shift;
12009
12010 border = (start_src / odata) - (start / ndata);
12011 border /= 512;
12012 if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) {
12013 /* save critical stripes to buf
12014 * start - start address of current unit
12015 * to backup [bytes]
12016 * start_src - start address of current unit
12017 * to backup alligned to source array
12018 * [bytes]
12019 */
594dc1b8 12020 unsigned long long next_step_filler;
10f22854
AK		 12021 unsigned long long copy_length = next_step * 512;
12022
12023 /* allign copy area length to stripe in old geometry */
12024 next_step_filler = ((copy_length + start_buf_shift)
12025 % old_data_stripe_length);
12026 if (next_step_filler)
12027 next_step_filler = (old_data_stripe_length
12028 - next_step_filler);
7a862a02 12029 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		 12030 start, start_src, copy_length,
12031 start_buf_shift, next_step_filler);
12032
12033 if (save_stripes(fds, offsets, map_src->num_members,
ab724b98
AK		 12034 chunk, map_src->raid_level,
12035 source_layout, 0, NULL, start_src,
10f22854
AK		 12036 copy_length +
12037 next_step_filler + start_buf_shift,
12038 buf)) {
7a862a02 12039 dprintf("imsm: Cannot save stripes to buffer\n");
10f22854
AK		 12040 goto abort;
12041 }
12042 /* Convert data to destination format and store it
12043 * in backup general migration area
12044 */
12045 if (save_backup_imsm(st, dev, sra,
aea93171 12046 buf + start_buf_shift, copy_length)) {
7a862a02 12047 dprintf("imsm: Cannot save stripes to target devices\n");
10f22854
AK		 12048 goto abort;
12049 }
12050 if (save_checkpoint_imsm(st, sra,
12051 UNIT_SRC_IN_CP_AREA)) {
7a862a02 12052 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_IN_CP_AREA)\n");
10f22854
AK		 12053 goto abort;
12054 }
8016a6d4
AK		 12055 } else {
12056 /* set next step to use whole border area */
12057 border /= next_step;
12058 if (border > 1)
12059 next_step *= border;
10f22854
AK		 12060 }
12061 /* When data backed up, checkpoint stored,
12062 * kick the kernel to reshape unit of data
12063 */
12064 next_step = next_step + sra->reshape_progress;
8016a6d4
AK		 12065 /* limit next step to array max position */
12066 if (next_step > max_position)
12067 next_step = max_position;
10f22854
AK		 12068 sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress);
12069 sysfs_set_num(sra, NULL, "suspend_hi", next_step);
ae9f01f8 12070 sra->reshape_progress = next_step;
10f22854
AK		 12071
12072 /* wait until reshape finish */
c85338c6 12073 if (wait_for_reshape_imsm(sra, ndata)) {
c47b0ff6
AK		 12074 dprintf("wait_for_reshape_imsm returned error!\n");
12075 goto abort;
12076 }
84d11e6c
N		 12077 if (sigterm)
12078 goto abort;
10f22854 12079
0228d92c
AK		 12080 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
12081 /* ignore error == 2, this can mean end of reshape here
12082 */
7a862a02 12083 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL)\n");
10f22854
AK		 12084 goto abort;
12085 }
12086
12087 }
12088
71e5411e
PB		 12089 /* clear migr_rec on disks after successful migration */
12090 struct dl *d;
12091
85337573 12092 memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
71e5411e
PB		 12093 for (d = super->disks; d; d = d->next) {
12094 if (d->index < 0 || is_failed(&d->disk))
12095 continue;
12096 unsigned long long dsize;
12097
12098 get_dev_size(d->fd, NULL, &dsize);
de44e46f 12099 if (lseek64(d->fd, dsize - MIGR_REC_SECTOR_POSITION*sector_size,
71e5411e 12100 SEEK_SET) >= 0) {
466070ad 12101 if ((unsigned int)write(d->fd, super->migr_rec_buf,
de44e46f
PB		 12102 MIGR_REC_BUF_SECTORS*sector_size) !=
12103 MIGR_REC_BUF_SECTORS*sector_size)
71e5411e
PB		 12104 perror("Write migr_rec failed");
12105 }
12106 }
12107
10f22854
AK		 12108 /* return '1' if done */
12109 ret_val = 1;
12110abort:
12111 free(buf);
942e1cdb
N		 12112 /* See Grow.c: abort_reshape() for further explanation */
12113 sysfs_set_num(sra, NULL, "suspend_lo", 0x7FFFFFFFFFFFFFFFULL);
12114 sysfs_set_num(sra, NULL, "suspend_hi", 0);
12115 sysfs_set_num(sra, NULL, "suspend_lo", 0);
10f22854
AK		 12116
12117 return ret_val;
999b4972 12118}
0c21b485 12119
cdddbdbc 12120struct superswitch super_imsm = {
cdddbdbc
DW		 12121 .examine_super = examine_super_imsm,
12122 .brief_examine_super = brief_examine_super_imsm,
4737ae25 12123 .brief_examine_subarrays = brief_examine_subarrays_imsm,
9d84c8ea 12124 .export_examine_super = export_examine_super_imsm,
cdddbdbc
DW		 12125 .detail_super = detail_super_imsm,
12126 .brief_detail_super = brief_detail_super_imsm,
bf5a934a 12127 .write_init_super = write_init_super_imsm,
0e600426
N		 12128 .validate_geometry = validate_geometry_imsm,
12129 .add_to_super = add_to_super_imsm,
1a64be56 12130 .remove_from_super = remove_from_super_imsm,
d665cc31 12131 .detail_platform = detail_platform_imsm,
e50cf220 12132 .export_detail_platform = export_detail_platform_imsm,
33414a01 12133 .kill_subarray = kill_subarray_imsm,
aa534678 12134 .update_subarray = update_subarray_imsm,
2b959fbf 12135 .load_container = load_container_imsm,
71204a50
N		 12136 .default_geometry = default_geometry_imsm,
12137 .get_disk_controller_domain = imsm_get_disk_controller_domain,
12138 .reshape_super = imsm_reshape_super,
12139 .manage_reshape = imsm_manage_reshape,
9e2d750d 12140 .recover_backup = recover_backup_imsm,
74db60b0 12141 .copy_metadata = copy_metadata_imsm,
27156a57 12142 .examine_badblocks = examine_badblocks_imsm,
cdddbdbc
DW		 12143 .match_home = match_home_imsm,
12144 .uuid_from_super= uuid_from_super_imsm,
12145 .getinfo_super = getinfo_super_imsm,
5c4cd5da 12146 .getinfo_super_disks = getinfo_super_disks_imsm,
cdddbdbc
DW		 12147 .update_super = update_super_imsm,
12148
12149 .avail_size = avail_size_imsm,
fbfdcb06 12150 .get_spare_criteria = get_spare_criteria_imsm,
cdddbdbc
DW		 12151
12152 .compare_super = compare_super_imsm,
12153
12154 .load_super = load_super_imsm,
bf5a934a 12155 .init_super = init_super_imsm,
e683ca88 12156 .store_super = store_super_imsm,
cdddbdbc
DW		 12157 .free_super = free_super_imsm,
12158 .match_metadata_desc = match_metadata_desc_imsm,
bf5a934a 12159 .container_content = container_content_imsm,
0c21b485 12160 .validate_container = validate_container_imsm,
cdddbdbc 12161
2432ce9b
AP		 12162 .write_init_ppl = write_init_ppl_imsm,
12163 .validate_ppl = validate_ppl_imsm,
12164
cdddbdbc 12165 .external = 1,
4cce4069 12166 .name = "imsm",
845dea95
NB		 12167
12168/* for mdmon */
12169 .open_new = imsm_open_new,
ed9d66aa 12170 .set_array_state= imsm_set_array_state,
845dea95
NB		 12171 .set_disk = imsm_set_disk,
12172 .sync_metadata = imsm_sync_metadata,
88758e9d 12173 .activate_spare = imsm_activate_spare,
e8319a19 12174 .process_update = imsm_process_update,
8273f55e 12175 .prepare_update = imsm_prepare_update,
6f50473f 12176 .record_bad_block = imsm_record_badblock,
c07a5a4f 12177 .clear_bad_block = imsm_clear_badblock,
928f1424 12178 .get_bad_blocks = imsm_get_badblocks,
cdddbdbc 12179};
Unnamed repository; edit this file 'description' to name the repository.