]> 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
Do not confuse gcc
[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;
760365f9 5326 int namelen;
c2c087e6 5327 unsigned long long array_blocks;
2c092cad 5328 size_t size_old, size_new;
5551b113 5329 unsigned long long num_data_stripes;
b53bfba6
TM		 5330 unsigned int data_disks;
5331 unsigned long long size_per_member;
cdddbdbc 5332
88c32bb1 5333 if (super->orom && mpb->num_raid_devs >= super->orom->vpa) {
7a862a02 5334 pr_err("This imsm-container already has the maximum of %d volumes\n", super->orom->vpa);
c2c087e6
DW		 5335 return 0;
5336 }
5337
2c092cad
DW		 5338 /* ensure the mpb is large enough for the new data */
5339 size_old = __le32_to_cpu(mpb->mpb_size);
5340 size_new = disks_to_mpb_size(info->nr_disks);
5341 if (size_new > size_old) {
5342 void *mpb_new;
f36a9ecd 5343 size_t size_round = ROUND_UP(size_new, sector_size);
2c092cad 5344
f36a9ecd 5345 if (posix_memalign(&mpb_new, sector_size, size_round) != 0) {
e7b84f9d 5346 pr_err("could not allocate new mpb\n");
2c092cad
DW		 5347 return 0;
5348 }
85337573
AO		 5349 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
5350 MIGR_REC_BUF_SECTORS*
5351 MAX_SECTOR_SIZE) != 0) {
1ade5cc1 5352 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8
AK		 5353 free(super->buf);
5354 free(super);
ea944c8f 5355 free(mpb_new);
8e59f3d8
AK		 5356 return 0;
5357 }
2c092cad
DW		 5358 memcpy(mpb_new, mpb, size_old);
5359 free(mpb);
5360 mpb = mpb_new;
949c47a0 5361 super->anchor = mpb_new;
2c092cad
DW		 5362 mpb->mpb_size = __cpu_to_le32(size_new);
5363 memset(mpb_new + size_old, 0, size_round - size_old);
bbab0940 5364 super->len = size_round;
2c092cad 5365 }
bf5a934a 5366 super->current_vol = idx;
3960e579
DW		 5367
5368 /* handle 'failed_disks' by either:
5369 * a) create dummy disk entries in the table if this the first
5370 * volume in the array. We add them here as this is the only
5371 * opportunity to add them. add_to_super_imsm_volume()
5372 * handles the non-failed disks and continues incrementing
5373 * mpb->num_disks.
5374 * b) validate that 'failed_disks' matches the current number
5375 * of missing disks if the container is populated
d23fe947 5376 */
3960e579 5377 if (super->current_vol == 0) {
d23fe947 5378 mpb->num_disks = 0;
3960e579
DW		 5379 for (i = 0; i < info->failed_disks; i++) {
5380 struct imsm_disk *disk;
5381
5382 mpb->num_disks++;
5383 disk = __get_imsm_disk(mpb, i);
5384 disk->status = CONFIGURED_DISK | FAILED_DISK;
5385 disk->scsi_id = __cpu_to_le32(~(__u32)0);
5386 snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN,
5b975129 5387 "missing:%d", (__u8)i);
3960e579
DW		 5388 }
5389 find_missing(super);
5390 } else {
5391 int missing = 0;
5392 struct dl *d;
5393
5394 for (d = super->missing; d; d = d->next)
5395 missing++;
5396 if (info->failed_disks > missing) {
e7b84f9d 5397 pr_err("unable to add 'missing' disk to container\n");
3960e579
DW		 5398 return 0;
5399 }
5400 }
5a038140 5401
aa534678
DW		 5402 if (!check_name(super, name, 0))
5403 return 0;
503975b9
N		 5404 dv = xmalloc(sizeof(*dv));
5405 dev = xcalloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1));
760365f9
JS		 5406 /*
5407 * Explicitly allow truncating to not confuse gcc's
5408 * -Werror=stringop-truncation
5409 */
5410 namelen = min((int) strlen(name), MAX_RAID_SERIAL_LEN);
5411 memcpy(dev->volume, name, namelen);
e03640bd 5412 array_blocks = calc_array_size(info->level, info->raid_disks,
03bcbc65 5413 info->layout, info->chunk_size,
b53bfba6
TM		 5414 s->size * BLOCKS_PER_KB);
5415 data_disks = get_data_disks(info->level, info->layout,
5416 info->raid_disks);
5417 array_blocks = round_size_to_mb(array_blocks, data_disks);
5418 size_per_member = array_blocks / data_disks;
979d38be 5419
fcc2c9da 5420 set_imsm_dev_size(dev, array_blocks);
1a2487c2 5421 dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING);
c2c087e6
DW		 5422 vol = &dev->vol;
5423 vol->migr_state = 0;
1484e727 5424 set_migr_type(dev, MIGR_INIT);
3960e579 5425 vol->dirty = !info->state;
f8f603f1 5426 vol->curr_migr_unit = 0;
238c0a71 5427 map = get_imsm_map(dev, MAP_0);
5551b113 5428 set_pba_of_lba0(map, super->create_offset);
ef6ffade 5429 map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info));
0556e1a2 5430 map->failed_disk_num = ~0;
bf4442ab 5431 if (info->level > 0)
fffaf1ff
N		 5432 map->map_state = (info->state ? IMSM_T_STATE_NORMAL
5433 : IMSM_T_STATE_UNINITIALIZED);
bf4442ab
AK		 5434 else
5435 map->map_state = info->failed_disks ? IMSM_T_STATE_FAILED :
5436 IMSM_T_STATE_NORMAL;
252d23c0 5437 map->ddf = 1;
ef6ffade
DW		 5438
5439 if (info->level == 1 && info->raid_disks > 2) {
38950822
AW		 5440 free(dev);
5441 free(dv);
7a862a02 5442 pr_err("imsm does not support more than 2 disksin a raid1 volume\n");
ef6ffade
DW		 5443 return 0;
5444 }
81062a36
DW		 5445
5446 map->raid_level = info->level;
4d1313e9 5447 if (info->level == 10) {
c2c087e6 5448 map->raid_level = 1;
4d1313e9 5449 map->num_domains = info->raid_disks / 2;
81062a36
DW		 5450 } else if (info->level == 1)
5451 map->num_domains = info->raid_disks;
5452 else
ff596308 5453 map->num_domains = 1;
81062a36 5454
5551b113 5455 /* info->size is only int so use the 'size' parameter instead */
b53bfba6 5456 num_data_stripes = size_per_member / info_to_blocks_per_strip(info);
5551b113
CA		 5457 num_data_stripes /= map->num_domains;
5458 set_num_data_stripes(map, num_data_stripes);
ef6ffade 5459
44490938
MD		 5460 size_per_member += NUM_BLOCKS_DIRTY_STRIPE_REGION;
5461 set_blocks_per_member(map, info_to_blocks_per_member(info,
5462 size_per_member /
5463 BLOCKS_PER_KB));
5464
c2c087e6
DW		 5465 map->num_members = info->raid_disks;
5466 for (i = 0; i < map->num_members; i++) {
5467 /* initialized in add_to_super */
4eb26970 5468 set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD);
c2c087e6 5469 }
949c47a0 5470 mpb->num_raid_devs++;
2a24dc1b
PB		 5471 mpb->num_raid_devs_created++;
5472 dev->my_vol_raid_dev_num = mpb->num_raid_devs_created;
ba2de7ba 5473
b7580566 5474 if (s->consistency_policy <= CONSISTENCY_POLICY_RESYNC) {
c2462068 5475 dev->rwh_policy = RWH_MULTIPLE_OFF;
2432ce9b 5476 } else if (s->consistency_policy == CONSISTENCY_POLICY_PPL) {
c2462068 5477 dev->rwh_policy = RWH_MULTIPLE_DISTRIBUTED;
2432ce9b
AP		 5478 } else {
5479 free(dev);
5480 free(dv);
5481 pr_err("imsm does not support consistency policy %s\n",
5482 map_num(consistency_policies, s->consistency_policy));
5483 return 0;
5484 }
5485
ba2de7ba
DW		 5486 dv->dev = dev;
5487 dv->index = super->current_vol;
5488 dv->next = super->devlist;
5489 super->devlist = dv;
c2c087e6 5490
4d1313e9
DW		 5491 imsm_update_version_info(super);
5492
c2c087e6 5493 return 1;
cdddbdbc
DW		 5494}
5495
bf5a934a 5496static int init_super_imsm(struct supertype *st, mdu_array_info_t *info,
5308f117 5497 struct shape *s, char *name,
83cd1e97
N		 5498 char *homehost, int *uuid,
5499 unsigned long long data_offset)
bf5a934a
DW		 5500{
5501 /* This is primarily called by Create when creating a new array.
5502 * We will then get add_to_super called for each component, and then
5503 * write_init_super called to write it out to each device.
5504 * For IMSM, Create can create on fresh devices or on a pre-existing
5505 * array.
5506 * To create on a pre-existing array a different method will be called.
5507 * This one is just for fresh drives.
5508 */
5509 struct intel_super *super;
5510 struct imsm_super *mpb;
5511 size_t mpb_size;
4d1313e9 5512 char *version;
bf5a934a 5513
83cd1e97 5514 if (data_offset != INVALID_SECTORS) {
ed503f89 5515 pr_err("data-offset not supported by imsm\n");
83cd1e97
N		 5516 return 0;
5517 }
5518
bf5a934a 5519 if (st->sb)
5308f117 5520 return init_super_imsm_volume(st, info, s, name, homehost, uuid,
83cd1e97 5521 data_offset);
e683ca88
DW		 5522
5523 if (info)
5524 mpb_size = disks_to_mpb_size(info->nr_disks);
5525 else
f36a9ecd 5526 mpb_size = MAX_SECTOR_SIZE;
bf5a934a 5527
49133e57 5528 super = alloc_super();
f36a9ecd
PB		 5529 if (super &&
5530 posix_memalign(&super->buf, MAX_SECTOR_SIZE, mpb_size) != 0) {
8d67477f 5531 free_imsm(super);
e683ca88
DW		 5532 super = NULL;
5533 }
5534 if (!super) {
1ade5cc1 5535 pr_err("could not allocate superblock\n");
bf5a934a
DW		 5536 return 0;
5537 }
de44e46f
PB		 5538 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
5539 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) {
1ade5cc1 5540 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8 5541 free(super->buf);
8d67477f 5542 free_imsm(super);
8e59f3d8
AK		 5543 return 0;
5544 }
e683ca88 5545 memset(super->buf, 0, mpb_size);
ef649044 5546 mpb = super->buf;
e683ca88
DW		 5547 mpb->mpb_size = __cpu_to_le32(mpb_size);
5548 st->sb = super;
5549
5550 if (info == NULL) {
5551 /* zeroing superblock */
5552 return 0;
5553 }
bf5a934a 5554
4d1313e9
DW		 5555 mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
5556
5557 version = (char *) mpb->sig;
5558 strcpy(version, MPB_SIGNATURE);
5559 version += strlen(MPB_SIGNATURE);
5560 strcpy(version, MPB_VERSION_RAID0);
bf5a934a 5561
bf5a934a
DW		 5562 return 1;
5563}
5564
f2cc4f7d
AO		 5565static int drive_validate_sector_size(struct intel_super *super, struct dl *dl)
5566{
5567 unsigned int member_sector_size;
5568
5569 if (dl->fd < 0) {
5570 pr_err("Invalid file descriptor for %s\n", dl->devname);
5571 return 0;
5572 }
5573
5574 if (!get_dev_sector_size(dl->fd, dl->devname, &member_sector_size))
5575 return 0;
5576 if (member_sector_size != super->sector_size)
5577 return 0;
5578 return 1;
5579}
5580
f20c3968 5581static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk,
bf5a934a
DW		 5582 int fd, char *devname)
5583{
5584 struct intel_super *super = st->sb;
d23fe947 5585 struct imsm_super *mpb = super->anchor;
3960e579 5586 struct imsm_disk *_disk;
bf5a934a
DW		 5587 struct imsm_dev *dev;
5588 struct imsm_map *map;
3960e579 5589 struct dl *dl, *df;
4eb26970 5590 int slot;
bf5a934a 5591
949c47a0 5592 dev = get_imsm_dev(super, super->current_vol);
238c0a71 5593 map = get_imsm_map(dev, MAP_0);
bf5a934a 5594
208933a7 5595 if (! (dk->state & (1<<MD_DISK_SYNC))) {
e7b84f9d 5596 pr_err("%s: Cannot add spare devices to IMSM volume\n",
208933a7
N		 5597 devname);
5598 return 1;
5599 }
5600
efb30e7f
DW		 5601 if (fd == -1) {
5602 /* we're doing autolayout so grab the pre-marked (in
5603 * validate_geometry) raid_disk
5604 */
5605 for (dl = super->disks; dl; dl = dl->next)
5606 if (dl->raiddisk == dk->raid_disk)
5607 break;
5608 } else {
5609 for (dl = super->disks; dl ; dl = dl->next)
5610 if (dl->major == dk->major &&
5611 dl->minor == dk->minor)
5612 break;
5613 }
d23fe947 5614
208933a7 5615 if (!dl) {
e7b84f9d 5616 pr_err("%s is not a member of the same container\n", devname);
f20c3968 5617 return 1;
208933a7 5618 }
bf5a934a 5619
f2cc4f7d
AO		 5620 if (!drive_validate_sector_size(super, dl)) {
5621 pr_err("Combining drives of different sector size in one volume is not allowed\n");
5622 return 1;
5623 }
5624
d23fe947
DW		 5625 /* add a pristine spare to the metadata */
5626 if (dl->index < 0) {
5627 dl->index = super->anchor->num_disks;
5628 super->anchor->num_disks++;
5629 }
4eb26970
DW		 5630 /* Check the device has not already been added */
5631 slot = get_imsm_disk_slot(map, dl->index);
5632 if (slot >= 0 &&
238c0a71 5633 (get_imsm_ord_tbl_ent(dev, slot, MAP_X) & IMSM_ORD_REBUILD) == 0) {
e7b84f9d 5634 pr_err("%s has been included in this array twice\n",
4eb26970
DW		 5635 devname);
5636 return 1;
5637 }
656b6b5a 5638 set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index);
ee5aad5a 5639 dl->disk.status = CONFIGURED_DISK;
d23fe947 5640
3960e579
DW		 5641 /* update size of 'missing' disks to be at least as large as the
5642 * largest acitve member (we only have dummy missing disks when
5643 * creating the first volume)
5644 */
5645 if (super->current_vol == 0) {
5646 for (df = super->missing; df; df = df->next) {
5551b113
CA		 5647 if (total_blocks(&dl->disk) > total_blocks(&df->disk))
5648 set_total_blocks(&df->disk, total_blocks(&dl->disk));
3960e579
DW		 5649 _disk = __get_imsm_disk(mpb, df->index);
5650 *_disk = df->disk;
5651 }
5652 }
5653
5654 /* refresh unset/failed slots to point to valid 'missing' entries */
5655 for (df = super->missing; df; df = df->next)
5656 for (slot = 0; slot < mpb->num_disks; slot++) {
238c0a71 5657 __u32 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
3960e579
DW		 5658
5659 if ((ord & IMSM_ORD_REBUILD) == 0)
5660 continue;
5661 set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD);
1ace8403 5662 if (is_gen_migration(dev)) {
238c0a71
AK		 5663 struct imsm_map *map2 = get_imsm_map(dev,
5664 MAP_1);
0a108d63 5665 int slot2 = get_imsm_disk_slot(map2, df->index);
089f9d79 5666 if (slot2 < map2->num_members && slot2 >= 0) {
1ace8403 5667 __u32 ord2 = get_imsm_ord_tbl_ent(dev,
238c0a71
AK		 5668 slot2,
5669 MAP_1);
1ace8403
AK		 5670 if ((unsigned)df->index ==
5671 ord_to_idx(ord2))
5672 set_imsm_ord_tbl_ent(map2,
0a108d63 5673 slot2,
1ace8403
AK		 5674 df->index |
5675 IMSM_ORD_REBUILD);
5676 }
5677 }
3960e579
DW		 5678 dprintf("set slot:%d to missing disk:%d\n", slot, df->index);
5679 break;
5680 }
5681
d23fe947
DW		 5682 /* if we are creating the first raid device update the family number */
5683 if (super->current_vol == 0) {
5684 __u32 sum;
5685 struct imsm_dev *_dev = __get_imsm_dev(mpb, 0);
d23fe947 5686
3960e579 5687 _disk = __get_imsm_disk(mpb, dl->index);
791b666a 5688 if (!_dev || !_disk) {
e7b84f9d 5689 pr_err("BUG mpb setup error\n");
791b666a
AW		 5690 return 1;
5691 }
d23fe947
DW		 5692 *_dev = *dev;
5693 *_disk = dl->disk;
148acb7b
DW		 5694 sum = random32();
5695 sum += __gen_imsm_checksum(mpb);
d23fe947 5696 mpb->family_num = __cpu_to_le32(sum);
148acb7b 5697 mpb->orig_family_num = mpb->family_num;
d23fe947 5698 }
ca0748fa 5699 super->current_disk = dl;
f20c3968 5700 return 0;
bf5a934a
DW		 5701}
5702
a8619d23
AK		 5703/* mark_spare()
5704 * Function marks disk as spare and restores disk serial
5705 * in case it was previously marked as failed by takeover operation
5706 * reruns:
5707 * -1 : critical error
5708 * 0 : disk is marked as spare but serial is not set
5709 * 1 : success
5710 */
5711int mark_spare(struct dl *disk)
5712{
5713 __u8 serial[MAX_RAID_SERIAL_LEN];
5714 int ret_val = -1;
5715
5716 if (!disk)
5717 return ret_val;
5718
5719 ret_val = 0;
5720 if (!imsm_read_serial(disk->fd, NULL, serial)) {
5721 /* Restore disk serial number, because takeover marks disk
5722 * as failed and adds to serial ':0' before it becomes
5723 * a spare disk.
5724 */
5725 serialcpy(disk->serial, serial);
5726 serialcpy(disk->disk.serial, serial);
5727 ret_val = 1;
5728 }
5729 disk->disk.status = SPARE_DISK;
5730 disk->index = -1;
5731
5732 return ret_val;
5733}
88654014 5734
f20c3968 5735static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk,
72ca9bcf
N		 5736 int fd, char *devname,
5737 unsigned long long data_offset)
cdddbdbc 5738{
c2c087e6 5739 struct intel_super *super = st->sb;
c2c087e6
DW		 5740 struct dl *dd;
5741 unsigned long long size;
fa7bb6f8 5742 unsigned int member_sector_size;
f2f27e63 5743 __u32 id;
c2c087e6
DW		 5744 int rv;
5745 struct stat stb;
5746
88654014
LM		 5747 /* If we are on an RAID enabled platform check that the disk is
5748 * attached to the raid controller.
5749 * We do not need to test disks attachment for container based additions,
5750 * they shall be already tested when container was created/assembled.
88c32bb1 5751 */
d424212e 5752 rv = find_intel_hba_capability(fd, super, devname);
f2f5c343 5753 /* no orom/efi or non-intel hba of the disk */
f0f5a016
LM		 5754 if (rv != 0) {
5755 dprintf("capability: %p fd: %d ret: %d\n",
5756 super->orom, fd, rv);
5757 return 1;
88c32bb1
DW		 5758 }
5759
f20c3968
DW		 5760 if (super->current_vol >= 0)
5761 return add_to_super_imsm_volume(st, dk, fd, devname);
bf5a934a 5762
c2c087e6 5763 fstat(fd, &stb);
503975b9 5764 dd = xcalloc(sizeof(*dd), 1);
c2c087e6
DW		 5765 dd->major = major(stb.st_rdev);
5766 dd->minor = minor(stb.st_rdev);
503975b9 5767 dd->devname = devname ? xstrdup(devname) : NULL;
c2c087e6 5768 dd->fd = fd;
689c9bf3 5769 dd->e = NULL;
1a64be56 5770 dd->action = DISK_ADD;
c2c087e6 5771 rv = imsm_read_serial(fd, devname, dd->serial);
32ba9157 5772 if (rv) {
e7b84f9d 5773 pr_err("failed to retrieve scsi serial, aborting\n");
20bee0f8
PB		 5774 if (dd->devname)
5775 free(dd->devname);
949c47a0 5776 free(dd);
0030e8d6 5777 abort();
c2c087e6 5778 }
20bee0f8
PB		 5779 if (super->hba && ((super->hba->type == SYS_DEV_NVME) ||
5780 (super->hba->type == SYS_DEV_VMD))) {
5781 int i;
5782 char *devpath = diskfd_to_devpath(fd);
5783 char controller_path[PATH_MAX];
5784
5785 if (!devpath) {
5786 pr_err("failed to get devpath, aborting\n");
5787 if (dd->devname)
5788 free(dd->devname);
5789 free(dd);
5790 return 1;
5791 }
5792
5793 snprintf(controller_path, PATH_MAX-1, "%s/device", devpath);
5794 free(devpath);
5795
5796 if (devpath_to_vendor(controller_path) == 0x8086) {
5797 /*
5798 * If Intel's NVMe drive has serial ended with
5799 * "-A","-B","-1" or "-2" it means that this is "x8"
5800 * device (double drive on single PCIe card).
5801 * User should be warned about potential data loss.
5802 */
5803 for (i = MAX_RAID_SERIAL_LEN-1; i > 0; i--) {
5804 /* Skip empty character at the end */
5805 if (dd->serial[i] == 0)
5806 continue;
5807
5808 if (((dd->serial[i] == 'A') ||
5809 (dd->serial[i] == 'B') ||
5810 (dd->serial[i] == '1') ||
5811 (dd->serial[i] == '2')) &&
5812 (dd->serial[i-1] == '-'))
5813 pr_err("\tThe action you are about to take may put your data at risk.\n"
5814 "\tPlease note that x8 devices may consist of two separate x4 devices "
5815 "located on a single PCIe port.\n"
5816 "\tRAID 0 is the only supported configuration for this type of x8 device.\n");
5817 break;
5818 }
32716c51
PB		 5819 } else if (super->hba->type == SYS_DEV_VMD && super->orom &&
5820 !imsm_orom_has_tpv_support(super->orom)) {
5821 pr_err("\tPlatform configuration does not support non-Intel NVMe drives.\n"
8b751247 5822 "\tPlease refer to Intel(R) RSTe/VROC user guide.\n");
32716c51
PB		 5823 free(dd->devname);
5824 free(dd);
5825 return 1;
20bee0f8
PB		 5826 }
5827 }
c2c087e6 5828
c2c087e6 5829 get_dev_size(fd, NULL, &size);
fa7bb6f8
PB		 5830 get_dev_sector_size(fd, NULL, &member_sector_size);
5831
5832 if (super->sector_size == 0) {
5833 /* this a first device, so sector_size is not set yet */
5834 super->sector_size = member_sector_size;
fa7bb6f8
PB		 5835 }
5836
71e5411e 5837 /* clear migr_rec when adding disk to container */
85337573
AO		 5838 memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
5839 if (lseek64(fd, size - MIGR_REC_SECTOR_POSITION*member_sector_size,
de44e46f 5840 SEEK_SET) >= 0) {
466070ad 5841 if ((unsigned int)write(fd, super->migr_rec_buf,
85337573
AO		 5842 MIGR_REC_BUF_SECTORS*member_sector_size) !=
5843 MIGR_REC_BUF_SECTORS*member_sector_size)
71e5411e
PB		 5844 perror("Write migr_rec failed");
5845 }
5846
c2c087e6 5847 size /= 512;
1f24f035 5848 serialcpy(dd->disk.serial, dd->serial);
5551b113
CA		 5849 set_total_blocks(&dd->disk, size);
5850 if (__le32_to_cpu(dd->disk.total_blocks_hi) > 0) {
5851 struct imsm_super *mpb = super->anchor;
5852 mpb->attributes |= MPB_ATTRIB_2TB_DISK;
5853 }
a8619d23 5854 mark_spare(dd);
c2c087e6 5855 if (sysfs_disk_to_scsi_id(fd, &id) == 0)
b9f594fe 5856 dd->disk.scsi_id = __cpu_to_le32(id);
c2c087e6 5857 else
b9f594fe 5858 dd->disk.scsi_id = __cpu_to_le32(0);
43dad3d6
DW		 5859
5860 if (st->update_tail) {
1a64be56
LM		 5861 dd->next = super->disk_mgmt_list;
5862 super->disk_mgmt_list = dd;
43dad3d6
DW		 5863 } else {
5864 dd->next = super->disks;
5865 super->disks = dd;
ceaf0ee1 5866 super->updates_pending++;
43dad3d6 5867 }
f20c3968
DW		 5868
5869 return 0;
cdddbdbc
DW		 5870}
5871
1a64be56
LM		 5872static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk)
5873{
5874 struct intel_super *super = st->sb;
5875 struct dl *dd;
5876
5877 /* remove from super works only in mdmon - for communication
5878 * manager - monitor. Check if communication memory buffer
5879 * is prepared.
5880 */
5881 if (!st->update_tail) {
1ade5cc1 5882 pr_err("shall be used in mdmon context only\n");
1a64be56
LM		 5883 return 1;
5884 }
503975b9 5885 dd = xcalloc(1, sizeof(*dd));
1a64be56
LM		 5886 dd->major = dk->major;
5887 dd->minor = dk->minor;
1a64be56 5888 dd->fd = -1;
a8619d23 5889 mark_spare(dd);
1a64be56
LM		 5890 dd->action = DISK_REMOVE;
5891
5892 dd->next = super->disk_mgmt_list;
5893 super->disk_mgmt_list = dd;
5894
1a64be56
LM		 5895 return 0;
5896}
5897
f796af5d
DW		 5898static int store_imsm_mpb(int fd, struct imsm_super *mpb);
5899
5900static union {
f36a9ecd 5901 char buf[MAX_SECTOR_SIZE];
f796af5d 5902 struct imsm_super anchor;
f36a9ecd 5903} spare_record __attribute__ ((aligned(MAX_SECTOR_SIZE)));
c2c087e6 5904
d23fe947
DW		 5905/* spare records have their own family number and do not have any defined raid
5906 * devices
5907 */
5908static int write_super_imsm_spares(struct intel_super *super, int doclose)
5909{
d23fe947 5910 struct imsm_super *mpb = super->anchor;
f796af5d 5911 struct imsm_super *spare = &spare_record.anchor;
d23fe947
DW		 5912 __u32 sum;
5913 struct dl *d;
5914
68641cdb
JS		 5915 spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super));
5916 spare->generation_num = __cpu_to_le32(1UL);
f796af5d 5917 spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
68641cdb
JS		 5918 spare->num_disks = 1;
5919 spare->num_raid_devs = 0;
5920 spare->cache_size = mpb->cache_size;
5921 spare->pwr_cycle_count = __cpu_to_le32(1);
f796af5d
DW		 5922
5923 snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH,
5924 MPB_SIGNATURE MPB_VERSION_RAID0);
d23fe947
DW		 5925
5926 for (d = super->disks; d; d = d->next) {
8796fdc4 5927 if (d->index != -1)
d23fe947
DW		 5928 continue;
5929
f796af5d 5930 spare->disk[0] = d->disk;
027c374f
CA		 5931 if (__le32_to_cpu(d->disk.total_blocks_hi) > 0)
5932 spare->attributes |= MPB_ATTRIB_2TB_DISK;
5933
f36a9ecd
PB		 5934 if (super->sector_size == 4096)
5935 convert_to_4k_imsm_disk(&spare->disk[0]);
5936
f796af5d
DW		 5937 sum = __gen_imsm_checksum(spare);
5938 spare->family_num = __cpu_to_le32(sum);
5939 spare->orig_family_num = 0;
5940 sum = __gen_imsm_checksum(spare);
5941 spare->check_sum = __cpu_to_le32(sum);
d23fe947 5942
f796af5d 5943 if (store_imsm_mpb(d->fd, spare)) {
1ade5cc1
N		 5944 pr_err("failed for device %d:%d %s\n",
5945 d->major, d->minor, strerror(errno));
e74255d9 5946 return 1;
d23fe947
DW		 5947 }
5948 if (doclose) {
5949 close(d->fd);
5950 d->fd = -1;
5951 }
5952 }
5953
e74255d9 5954 return 0;
d23fe947
DW		 5955}
5956
36988a3d 5957static int write_super_imsm(struct supertype *st, int doclose)
cdddbdbc 5958{
36988a3d 5959 struct intel_super *super = st->sb;
f36a9ecd 5960 unsigned int sector_size = super->sector_size;
949c47a0 5961 struct imsm_super *mpb = super->anchor;
c2c087e6
DW		 5962 struct dl *d;
5963 __u32 generation;
5964 __u32 sum;
d23fe947 5965 int spares = 0;
949c47a0 5966 int i;
a48ac0a8 5967 __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk);
36988a3d 5968 int num_disks = 0;
146c6260 5969 int clear_migration_record = 1;
bbab0940 5970 __u32 bbm_log_size;
cdddbdbc 5971
c2c087e6
DW		 5972 /* 'generation' is incremented everytime the metadata is written */
5973 generation = __le32_to_cpu(mpb->generation_num);
5974 generation++;
5975 mpb->generation_num = __cpu_to_le32(generation);
5976
148acb7b
DW		 5977 /* fix up cases where previous mdadm releases failed to set
5978 * orig_family_num
5979 */
5980 if (mpb->orig_family_num == 0)
5981 mpb->orig_family_num = mpb->family_num;
5982
d23fe947 5983 for (d = super->disks; d; d = d->next) {
8796fdc4 5984 if (d->index == -1)
d23fe947 5985 spares++;
36988a3d 5986 else {
d23fe947 5987 mpb->disk[d->index] = d->disk;
36988a3d
AK		 5988 num_disks++;
5989 }
d23fe947 5990 }
36988a3d 5991 for (d = super->missing; d; d = d->next) {
47ee5a45 5992 mpb->disk[d->index] = d->disk;
36988a3d
AK		 5993 num_disks++;
5994 }
5995 mpb->num_disks = num_disks;
5996 mpb_size += sizeof(struct imsm_disk) * mpb->num_disks;
b9f594fe 5997
949c47a0
DW		 5998 for (i = 0; i < mpb->num_raid_devs; i++) {
5999 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
36988a3d
AK		 6000 struct imsm_dev *dev2 = get_imsm_dev(super, i);
6001 if (dev && dev2) {
6002 imsm_copy_dev(dev, dev2);
6003 mpb_size += sizeof_imsm_dev(dev, 0);
6004 }
146c6260
AK		 6005 if (is_gen_migration(dev2))
6006 clear_migration_record = 0;
949c47a0 6007 }
bbab0940
TM		 6008
6009 bbm_log_size = get_imsm_bbm_log_size(super->bbm_log);
6010
6011 if (bbm_log_size) {
6012 memcpy((void *)mpb + mpb_size, super->bbm_log, bbm_log_size);
6013 mpb->attributes |= MPB_ATTRIB_BBM;
6014 } else
6015 mpb->attributes &= ~MPB_ATTRIB_BBM;
6016
6017 super->anchor->bbm_log_size = __cpu_to_le32(bbm_log_size);
6018 mpb_size += bbm_log_size;
a48ac0a8 6019 mpb->mpb_size = __cpu_to_le32(mpb_size);
949c47a0 6020
bbab0940
TM		 6021#ifdef DEBUG
6022 assert(super->len == 0 || mpb_size <= super->len);
6023#endif
6024
c2c087e6 6025 /* recalculate checksum */
949c47a0 6026 sum = __gen_imsm_checksum(mpb);
c2c087e6
DW		 6027 mpb->check_sum = __cpu_to_le32(sum);
6028
51d83f5d
AK		 6029 if (super->clean_migration_record_by_mdmon) {
6030 clear_migration_record = 1;
6031 super->clean_migration_record_by_mdmon = 0;
6032 }
146c6260 6033 if (clear_migration_record)
de44e46f 6034 memset(super->migr_rec_buf, 0,
85337573 6035 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
146c6260 6036
f36a9ecd
PB		 6037 if (sector_size == 4096)
6038 convert_to_4k(super);
6039
d23fe947 6040 /* write the mpb for disks that compose raid devices */
c2c087e6 6041 for (d = super->disks; d ; d = d->next) {
86c54047 6042 if (d->index < 0 || is_failed(&d->disk))
d23fe947 6043 continue;
30602f53 6044
146c6260
AK		 6045 if (clear_migration_record) {
6046 unsigned long long dsize;
6047
6048 get_dev_size(d->fd, NULL, &dsize);
de44e46f
PB		 6049 if (lseek64(d->fd, dsize - sector_size,
6050 SEEK_SET) >= 0) {
466070ad
PB		 6051 if ((unsigned int)write(d->fd,
6052 super->migr_rec_buf,
de44e46f
PB		 6053 MIGR_REC_BUF_SECTORS*sector_size) !=
6054 MIGR_REC_BUF_SECTORS*sector_size)
9e2d750d 6055 perror("Write migr_rec failed");
146c6260
AK		 6056 }
6057 }
51d83f5d
AK		 6058
6059 if (store_imsm_mpb(d->fd, mpb))
6060 fprintf(stderr,
1ade5cc1
N		 6061 "failed for device %d:%d (fd: %d)%s\n",
6062 d->major, d->minor,
51d83f5d
AK		 6063 d->fd, strerror(errno));
6064
c2c087e6
DW		 6065 if (doclose) {
6066 close(d->fd);
6067 d->fd = -1;
6068 }
6069 }
6070
d23fe947
DW		 6071 if (spares)
6072 return write_super_imsm_spares(super, doclose);
6073
e74255d9 6074 return 0;
c2c087e6
DW		 6075}
6076
9b1fb677 6077static int create_array(struct supertype *st, int dev_idx)
43dad3d6
DW		 6078{
6079 size_t len;
6080 struct imsm_update_create_array *u;
6081 struct intel_super *super = st->sb;
9b1fb677 6082 struct imsm_dev *dev = get_imsm_dev(super, dev_idx);
238c0a71 6083 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW		 6084 struct disk_info *inf;
6085 struct imsm_disk *disk;
6086 int i;
43dad3d6 6087
54c2c1ea
DW		 6088 len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) +
6089 sizeof(*inf) * map->num_members;
503975b9 6090 u = xmalloc(len);
43dad3d6 6091 u->type = update_create_array;
9b1fb677 6092 u->dev_idx = dev_idx;
43dad3d6 6093 imsm_copy_dev(&u->dev, dev);
54c2c1ea
DW		 6094 inf = get_disk_info(u);
6095 for (i = 0; i < map->num_members; i++) {
238c0a71 6096 int idx = get_imsm_disk_idx(dev, i, MAP_X);
9b1fb677 6097
54c2c1ea 6098 disk = get_imsm_disk(super, idx);
1ca5c8e0
N		 6099 if (!disk)
6100 disk = get_imsm_missing(super, idx);
54c2c1ea
DW		 6101 serialcpy(inf[i].serial, disk->serial);
6102 }
43dad3d6
DW		 6103 append_metadata_update(st, u, len);
6104
6105 return 0;
6106}
6107
1a64be56 6108static int mgmt_disk(struct supertype *st)
43dad3d6
DW		 6109{
6110 struct intel_super *super = st->sb;
6111 size_t len;
1a64be56 6112 struct imsm_update_add_remove_disk *u;
43dad3d6 6113
1a64be56 6114 if (!super->disk_mgmt_list)
43dad3d6
DW		 6115 return 0;
6116
6117 len = sizeof(*u);
503975b9 6118 u = xmalloc(len);
1a64be56 6119 u->type = update_add_remove_disk;
43dad3d6
DW		 6120 append_metadata_update(st, u, len);
6121
6122 return 0;
6123}
2432ce9b
AP		 6124
6125__u32 crc32c_le(__u32 crc, unsigned char const *p, size_t len);
6126
e397cefe
AP		 6127static int write_ppl_header(unsigned long long ppl_sector, int fd, void *buf)
6128{
6129 struct ppl_header *ppl_hdr = buf;
6130 int ret;
6131
6132 ppl_hdr->checksum = __cpu_to_le32(~crc32c_le(~0, buf, PPL_HEADER_SIZE));
6133
6134 if (lseek64(fd, ppl_sector * 512, SEEK_SET) < 0) {
6135 ret = -errno;
6136 perror("Failed to seek to PPL header location");
6137 return ret;
6138 }
6139
6140 if (write(fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) {
6141 ret = -errno;
6142 perror("Write PPL header failed");
6143 return ret;
6144 }
6145
6146 fsync(fd);
6147
6148 return 0;
6149}
6150
2432ce9b
AP		 6151static int write_init_ppl_imsm(struct supertype *st, struct mdinfo *info, int fd)
6152{
6153 struct intel_super *super = st->sb;
6154 void *buf;
6155 struct ppl_header *ppl_hdr;
6156 int ret;
6157
b2514242
PB		 6158 /* first clear entire ppl space */
6159 ret = zero_disk_range(fd, info->ppl_sector, info->ppl_size);
6160 if (ret)
6161 return ret;
6162
6163 ret = posix_memalign(&buf, MAX_SECTOR_SIZE, PPL_HEADER_SIZE);
2432ce9b
AP		 6164 if (ret) {
6165 pr_err("Failed to allocate PPL header buffer\n");
e397cefe 6166 return -ret;
2432ce9b
AP		 6167 }
6168
6169 memset(buf, 0, PPL_HEADER_SIZE);
6170 ppl_hdr = buf;
6171 memset(ppl_hdr->reserved, 0xff, PPL_HDR_RESERVED);
6172 ppl_hdr->signature = __cpu_to_le32(super->anchor->orig_family_num);
b23d0750
AP		 6173
6174 if (info->mismatch_cnt) {
6175 /*
6176 * We are overwriting an invalid ppl. Make one entry with wrong
6177 * checksum to prevent the kernel from skipping resync.
6178 */
6179 ppl_hdr->entries_count = __cpu_to_le32(1);
6180 ppl_hdr->entries[0].checksum = ~0;
6181 }
6182
e397cefe 6183 ret = write_ppl_header(info->ppl_sector, fd, buf);
2432ce9b
AP		 6184
6185 free(buf);
6186 return ret;
6187}
6188
e397cefe
AP		 6189static int is_rebuilding(struct imsm_dev *dev);
6190
2432ce9b
AP		 6191static int validate_ppl_imsm(struct supertype *st, struct mdinfo *info,
6192 struct mdinfo *disk)
6193{
6194 struct intel_super *super = st->sb;
6195 struct dl *d;
e397cefe 6196 void *buf_orig, *buf, *buf_prev = NULL;
2432ce9b 6197 int ret = 0;
e397cefe 6198 struct ppl_header *ppl_hdr = NULL;
2432ce9b
AP		 6199 __u32 crc;
6200 struct imsm_dev *dev;
2432ce9b 6201 __u32 idx;
44b6b876
PB		 6202 unsigned int i;
6203 unsigned long long ppl_offset = 0;
6204 unsigned long long prev_gen_num = 0;
2432ce9b
AP		 6205
6206 if (disk->disk.raid_disk < 0)
6207 return 0;
6208
2432ce9b 6209 dev = get_imsm_dev(super, info->container_member);
2fc0fc63 6210 idx = get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_0);
2432ce9b
AP		 6211 d = get_imsm_dl_disk(super, idx);
6212
6213 if (!d || d->index < 0 || is_failed(&d->disk))
e397cefe
AP		 6214 return 0;
6215
6216 if (posix_memalign(&buf_orig, MAX_SECTOR_SIZE, PPL_HEADER_SIZE * 2)) {
6217 pr_err("Failed to allocate PPL header buffer\n");
6218 return -1;
6219 }
6220 buf = buf_orig;
2432ce9b 6221
44b6b876
PB		 6222 ret = 1;
6223 while (ppl_offset < MULTIPLE_PPL_AREA_SIZE_IMSM) {
e397cefe
AP		 6224 void *tmp;
6225
44b6b876 6226 dprintf("Checking potential PPL at offset: %llu\n", ppl_offset);
2432ce9b 6227
44b6b876
PB		 6228 if (lseek64(d->fd, info->ppl_sector * 512 + ppl_offset,
6229 SEEK_SET) < 0) {
6230 perror("Failed to seek to PPL header location");
6231 ret = -1;
e397cefe 6232 break;
44b6b876 6233 }
2432ce9b 6234
44b6b876
PB		 6235 if (read(d->fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) {
6236 perror("Read PPL header failed");
6237 ret = -1;
e397cefe 6238 break;
44b6b876 6239 }
2432ce9b 6240
44b6b876 6241 ppl_hdr = buf;
2432ce9b 6242
44b6b876
PB		 6243 crc = __le32_to_cpu(ppl_hdr->checksum);
6244 ppl_hdr->checksum = 0;
6245
6246 if (crc != ~crc32c_le(~0, buf, PPL_HEADER_SIZE)) {
6247 dprintf("Wrong PPL header checksum on %s\n",
6248 d->devname);
e397cefe 6249 break;
44b6b876
PB		 6250 }
6251
6252 if (prev_gen_num > __le64_to_cpu(ppl_hdr->generation)) {
6253 /* previous was newest, it was already checked */
e397cefe 6254 break;
44b6b876
PB		 6255 }
6256
6257 if ((__le32_to_cpu(ppl_hdr->signature) !=
6258 super->anchor->orig_family_num)) {
6259 dprintf("Wrong PPL header signature on %s\n",
6260 d->devname);
6261 ret = 1;
e397cefe 6262 break;
44b6b876
PB		 6263 }
6264
6265 ret = 0;
6266 prev_gen_num = __le64_to_cpu(ppl_hdr->generation);
2432ce9b 6267
44b6b876
PB		 6268 ppl_offset += PPL_HEADER_SIZE;
6269 for (i = 0; i < __le32_to_cpu(ppl_hdr->entries_count); i++)
6270 ppl_offset +=
6271 __le32_to_cpu(ppl_hdr->entries[i].pp_size);
e397cefe
AP		 6272
6273 if (!buf_prev)
6274 buf_prev = buf + PPL_HEADER_SIZE;
6275 tmp = buf_prev;
6276 buf_prev = buf;
6277 buf = tmp;
2432ce9b
AP		 6278 }
6279
e397cefe
AP		 6280 if (buf_prev) {
6281 buf = buf_prev;
6282 ppl_hdr = buf_prev;
6283 }
2432ce9b 6284
54148aba
PB		 6285 /*
6286 * Update metadata to use mutliple PPLs area (1MB).
6287 * This is done once for all RAID members
6288 */
6289 if (info->consistency_policy == CONSISTENCY_POLICY_PPL &&
6290 info->ppl_size != (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9)) {
6291 char subarray[20];
6292 struct mdinfo *member_dev;
6293
6294 sprintf(subarray, "%d", info->container_member);
6295
6296 if (mdmon_running(st->container_devnm))
6297 st->update_tail = &st->updates;
6298
6299 if (st->ss->update_subarray(st, subarray, "ppl", NULL)) {
6300 pr_err("Failed to update subarray %s\n",
6301 subarray);
6302 } else {
6303 if (st->update_tail)
6304 flush_metadata_updates(st);
6305 else
6306 st->ss->sync_metadata(st);
6307 info->ppl_size = (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9);
6308 for (member_dev = info->devs; member_dev;
6309 member_dev = member_dev->next)
6310 member_dev->ppl_size =
6311 (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9);
6312 }
6313 }
6314
b23d0750 6315 if (ret == 1) {
2fc0fc63
AP		 6316 struct imsm_map *map = get_imsm_map(dev, MAP_X);
6317
50b9c10d
PB		 6318 if (map->map_state == IMSM_T_STATE_UNINITIALIZED ||
6319 (map->map_state == IMSM_T_STATE_NORMAL &&
2ec9d182 6320 !(dev->vol.dirty & RAIDVOL_DIRTY)) ||
e397cefe 6321 (is_rebuilding(dev) &&
2ec9d182
AP		 6322 dev->vol.curr_migr_unit == 0 &&
6323 get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_1) != idx))
b23d0750
AP		 6324 ret = st->ss->write_init_ppl(st, info, d->fd);
6325 else
6326 info->mismatch_cnt++;
e397cefe
AP		 6327 } else if (ret == 0 &&
6328 ppl_hdr->entries_count == 0 &&
6329 is_rebuilding(dev) &&
6330 info->resync_start == 0) {
6331 /*
6332 * The header has no entries - add a single empty entry and
6333 * rewrite the header to prevent the kernel from going into
6334 * resync after an interrupted rebuild.
6335 */
6336 ppl_hdr->entries_count = __cpu_to_le32(1);
6337 ret = write_ppl_header(info->ppl_sector, d->fd, buf);
b23d0750 6338 }
2432ce9b 6339
e397cefe
AP		 6340 free(buf_orig);
6341
2432ce9b
AP		 6342 return ret;
6343}
6344
2432ce9b
AP		 6345static int write_init_ppl_imsm_all(struct supertype *st, struct mdinfo *info)
6346{
6347 struct intel_super *super = st->sb;
6348 struct dl *d;
6349 int ret = 0;
6350
6351 if (info->consistency_policy != CONSISTENCY_POLICY_PPL ||
6352 info->array.level != 5)
6353 return 0;
6354
6355 for (d = super->disks; d ; d = d->next) {
6356 if (d->index < 0 || is_failed(&d->disk))
6357 continue;
6358
6359 ret = st->ss->write_init_ppl(st, info, d->fd);
6360 if (ret)
6361 break;
6362 }
6363
6364 return ret;
6365}
43dad3d6 6366
c2c087e6
DW		 6367static int write_init_super_imsm(struct supertype *st)
6368{
9b1fb677
DW		 6369 struct intel_super *super = st->sb;
6370 int current_vol = super->current_vol;
2432ce9b
AP		 6371 int rv = 0;
6372 struct mdinfo info;
6373
6374 getinfo_super_imsm(st, &info, NULL);
9b1fb677
DW		 6375
6376 /* we are done with current_vol reset it to point st at the container */
6377 super->current_vol = -1;
6378
8273f55e 6379 if (st->update_tail) {
43dad3d6
DW		 6380 /* queue the recently created array / added disk
6381 * as a metadata update */
8273f55e 6382
43dad3d6 6383 /* determine if we are creating a volume or adding a disk */
9b1fb677 6384 if (current_vol < 0) {
1a64be56
LM		 6385 /* in the mgmt (add/remove) disk case we are running
6386 * in mdmon context, so don't close fd's
43dad3d6 6387 */
2432ce9b
AP		 6388 rv = mgmt_disk(st);
6389 } else {
6390 rv = write_init_ppl_imsm_all(st, &info);
6391 if (!rv)
6392 rv = create_array(st, current_vol);
6393 }
d682f344
N		 6394 } else {
6395 struct dl *d;
6396 for (d = super->disks; d; d = d->next)
ba728be7 6397 Kill(d->devname, NULL, 0, -1, 1);
2432ce9b
AP		 6398 if (current_vol >= 0)
6399 rv = write_init_ppl_imsm_all(st, &info);
6400 if (!rv)
6401 rv = write_super_imsm(st, 1);
d682f344 6402 }
2432ce9b
AP		 6403
6404 return rv;
cdddbdbc
DW		 6405}
6406
e683ca88 6407static int store_super_imsm(struct supertype *st, int fd)
cdddbdbc 6408{
e683ca88
DW		 6409 struct intel_super *super = st->sb;
6410 struct imsm_super *mpb = super ? super->anchor : NULL;
551c80c1 6411
e683ca88 6412 if (!mpb)
ad97895e
DW		 6413 return 1;
6414
f36a9ecd
PB		 6415 if (super->sector_size == 4096)
6416 convert_to_4k(super);
e683ca88 6417 return store_imsm_mpb(fd, mpb);
cdddbdbc
DW		 6418}
6419
cdddbdbc
DW		 6420static int validate_geometry_imsm_container(struct supertype *st, int level,
6421 int layout, int raiddisks, int chunk,
af4348dd
N		 6422 unsigned long long size,
6423 unsigned long long data_offset,
6424 char *dev,
2c514b71
NB		 6425 unsigned long long *freesize,
6426 int verbose)
cdddbdbc 6427{
c2c087e6
DW		 6428 int fd;
6429 unsigned long long ldsize;
594dc1b8 6430 struct intel_super *super;
f2f5c343 6431 int rv = 0;
cdddbdbc 6432
c2c087e6
DW		 6433 if (level != LEVEL_CONTAINER)
6434 return 0;
6435 if (!dev)
6436 return 1;
6437
6438 fd = open(dev, O_RDONLY|O_EXCL, 0);
6439 if (fd < 0) {
ba728be7 6440 if (verbose > 0)
e7b84f9d 6441 pr_err("imsm: Cannot open %s: %s\n",
2c514b71 6442 dev, strerror(errno));
c2c087e6
DW		 6443 return 0;
6444 }
6445 if (!get_dev_size(fd, dev, &ldsize)) {
6446 close(fd);
6447 return 0;
6448 }
f2f5c343
LM		 6449
6450 /* capabilities retrieve could be possible
6451 * note that there is no fd for the disks in array.
6452 */
6453 super = alloc_super();
8d67477f
TM		 6454 if (!super) {
6455 close(fd);
6456 return 0;
6457 }
fa7bb6f8
PB		 6458 if (!get_dev_sector_size(fd, NULL, &super->sector_size)) {
6459 close(fd);
6460 free_imsm(super);
6461 return 0;
6462 }
6463
ba728be7 6464 rv = find_intel_hba_capability(fd, super, verbose > 0 ? dev : NULL);
f2f5c343
LM		 6465 if (rv != 0) {
6466#if DEBUG
6467 char str[256];
6468 fd2devname(fd, str);
1ade5cc1 6469 dprintf("fd: %d %s orom: %p rv: %d raiddisk: %d\n",
f2f5c343
LM		 6470 fd, str, super->orom, rv, raiddisks);
6471#endif
6472 /* no orom/efi or non-intel hba of the disk */
6473 close(fd);
6474 free_imsm(super);
6475 return 0;
6476 }
c2c087e6 6477 close(fd);
9126b9a8
CA		 6478 if (super->orom) {
6479 if (raiddisks > super->orom->tds) {
6480 if (verbose)
7a862a02 6481 pr_err("%d exceeds maximum number of platform supported disks: %d\n",
9126b9a8
CA		 6482 raiddisks, super->orom->tds);
6483 free_imsm(super);
6484 return 0;
6485 }
6486 if ((super->orom->attr & IMSM_OROM_ATTR_2TB_DISK) == 0 &&
6487 (ldsize >> 9) >> 32 > 0) {
6488 if (verbose)
e7b84f9d 6489 pr_err("%s exceeds maximum platform supported size\n", dev);
9126b9a8
CA		 6490 free_imsm(super);
6491 return 0;
6492 }
f2f5c343 6493 }
c2c087e6 6494
af4348dd 6495 *freesize = avail_size_imsm(st, ldsize >> 9, data_offset);
f2f5c343 6496 free_imsm(super);
c2c087e6
DW		 6497
6498 return 1;
cdddbdbc
DW		 6499}
6500
0dcecb2e
DW		 6501static unsigned long long find_size(struct extent *e, int *idx, int num_extents)
6502{
6503 const unsigned long long base_start = e[*idx].start;
6504 unsigned long long end = base_start + e[*idx].size;
6505 int i;
6506
6507 if (base_start == end)
6508 return 0;
6509
6510 *idx = *idx + 1;
6511 for (i = *idx; i < num_extents; i++) {
6512 /* extend overlapping extents */
6513 if (e[i].start >= base_start &&
6514 e[i].start <= end) {
6515 if (e[i].size == 0)
6516 return 0;
6517 if (e[i].start + e[i].size > end)
6518 end = e[i].start + e[i].size;
6519 } else if (e[i].start > end) {
6520 *idx = i;
6521 break;
6522 }
6523 }
6524
6525 return end - base_start;
6526}
6527
6528static unsigned long long merge_extents(struct intel_super *super, int sum_extents)
6529{
6530 /* build a composite disk with all known extents and generate a new
6531 * 'maxsize' given the "all disks in an array must share a common start
6532 * offset" constraint
6533 */
503975b9 6534 struct extent *e = xcalloc(sum_extents, sizeof(*e));
0dcecb2e
DW		 6535 struct dl *dl;
6536 int i, j;
6537 int start_extent;
6538 unsigned long long pos;
b9d77223 6539 unsigned long long start = 0;
0dcecb2e
DW		 6540 unsigned long long maxsize;
6541 unsigned long reserve;
6542
0dcecb2e
DW		 6543 /* coalesce and sort all extents. also, check to see if we need to
6544 * reserve space between member arrays
6545 */
6546 j = 0;
6547 for (dl = super->disks; dl; dl = dl->next) {
6548 if (!dl->e)
6549 continue;
6550 for (i = 0; i < dl->extent_cnt; i++)
6551 e[j++] = dl->e[i];
6552 }
6553 qsort(e, sum_extents, sizeof(*e), cmp_extent);
6554
6555 /* merge extents */
6556 i = 0;
6557 j = 0;
6558 while (i < sum_extents) {
6559 e[j].start = e[i].start;
6560 e[j].size = find_size(e, &i, sum_extents);
6561 j++;
6562 if (e[j-1].size == 0)
6563 break;
6564 }
6565
6566 pos = 0;
6567 maxsize = 0;
6568 start_extent = 0;
6569 i = 0;
6570 do {
6571 unsigned long long esize;
6572
6573 esize = e[i].start - pos;
6574 if (esize >= maxsize) {
6575 maxsize = esize;
6576 start = pos;
6577 start_extent = i;
6578 }
6579 pos = e[i].start + e[i].size;
6580 i++;
6581 } while (e[i-1].size);
6582 free(e);
6583
a7dd165b
DW		 6584 if (maxsize == 0)
6585 return 0;
6586
6587 /* FIXME assumes volume at offset 0 is the first volume in a
6588 * container
6589 */
0dcecb2e
DW		 6590 if (start_extent > 0)
6591 reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */
6592 else
6593 reserve = 0;
6594
6595 if (maxsize < reserve)
a7dd165b 6596 return 0;
0dcecb2e 6597
5551b113 6598 super->create_offset = ~((unsigned long long) 0);
0dcecb2e 6599 if (start + reserve > super->create_offset)
a7dd165b 6600 return 0; /* start overflows create_offset */
0dcecb2e
DW		 6601 super->create_offset = start + reserve;
6602
6603 return maxsize - reserve;
6604}
6605
88c32bb1
DW		 6606static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks)
6607{
6608 if (level < 0 || level == 6 || level == 4)
6609 return 0;
6610
6611 /* if we have an orom prevent invalid raid levels */
6612 if (orom)
6613 switch (level) {
6614 case 0: return imsm_orom_has_raid0(orom);
6615 case 1:
6616 if (raiddisks > 2)
6617 return imsm_orom_has_raid1e(orom);
1c556e92
DW		 6618 return imsm_orom_has_raid1(orom) && raiddisks == 2;
6619 case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4;
6620 case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2;
88c32bb1
DW		 6621 }
6622 else
6623 return 1; /* not on an Intel RAID platform so anything goes */
6624
6625 return 0;
6626}
6627
ca9de185
LM		 6628static int
6629active_arrays_by_format(char *name, char* hba, struct md_list **devlist,
6630 int dpa, int verbose)
6631{
6632 struct mdstat_ent *mdstat = mdstat_read(0, 0);
594dc1b8 6633 struct mdstat_ent *memb;
ca9de185
LM		 6634 int count = 0;
6635 int num = 0;
594dc1b8 6636 struct md_list *dv;
ca9de185
LM		 6637 int found;
6638
6639 for (memb = mdstat ; memb ; memb = memb->next) {
6640 if (memb->metadata_version &&
fc54fe7a 6641 (strncmp(memb->metadata_version, "external:", 9) == 0) &&
ca9de185
LM		 6642 (strcmp(&memb->metadata_version[9], name) == 0) &&
6643 !is_subarray(memb->metadata_version+9) &&
6644 memb->members) {
6645 struct dev_member *dev = memb->members;
6646 int fd = -1;
6647 while(dev && (fd < 0)) {
503975b9
N		 6648 char *path = xmalloc(strlen(dev->name) + strlen("/dev/") + 1);
6649 num = sprintf(path, "%s%s", "/dev/", dev->name);
6650 if (num > 0)
6651 fd = open(path, O_RDONLY, 0);
089f9d79 6652 if (num <= 0 || fd < 0) {
676e87a8 6653 pr_vrb("Cannot open %s: %s\n",
503975b9 6654 dev->name, strerror(errno));
ca9de185 6655 }
503975b9 6656 free(path);
ca9de185
LM		 6657 dev = dev->next;
6658 }
6659 found = 0;
089f9d79 6660 if (fd >= 0 && disk_attached_to_hba(fd, hba)) {
ca9de185
LM		 6661 struct mdstat_ent *vol;
6662 for (vol = mdstat ; vol ; vol = vol->next) {
089f9d79 6663 if (vol->active > 0 &&
ca9de185 6664 vol->metadata_version &&
9581efb1 6665 is_container_member(vol, memb->devnm)) {
ca9de185
LM		 6666 found++;
6667 count++;
6668 }
6669 }
6670 if (*devlist && (found < dpa)) {
503975b9 6671 dv = xcalloc(1, sizeof(*dv));
9581efb1
N		 6672 dv->devname = xmalloc(strlen(memb->devnm) + strlen("/dev/") + 1);
6673 sprintf(dv->devname, "%s%s", "/dev/", memb->devnm);
503975b9
N		 6674 dv->found = found;
6675 dv->used = 0;
6676 dv->next = *devlist;
6677 *devlist = dv;
ca9de185
LM		 6678 }
6679 }
6680 if (fd >= 0)
6681 close(fd);
6682 }
6683 }
6684 free_mdstat(mdstat);
6685 return count;
6686}
6687
6688#ifdef DEBUG_LOOP
6689static struct md_list*
6690get_loop_devices(void)
6691{
6692 int i;
6693 struct md_list *devlist = NULL;
594dc1b8 6694 struct md_list *dv;
ca9de185
LM		 6695
6696 for(i = 0; i < 12; i++) {
503975b9
N		 6697 dv = xcalloc(1, sizeof(*dv));
6698 dv->devname = xmalloc(40);
ca9de185
LM		 6699 sprintf(dv->devname, "/dev/loop%d", i);
6700 dv->next = devlist;
6701 devlist = dv;
6702 }
6703 return devlist;
6704}
6705#endif
6706
6707static struct md_list*
6708get_devices(const char *hba_path)
6709{
6710 struct md_list *devlist = NULL;
594dc1b8 6711 struct md_list *dv;
ca9de185
LM		 6712 struct dirent *ent;
6713 DIR *dir;
6714 int err = 0;
6715
6716#if DEBUG_LOOP
6717 devlist = get_loop_devices();
6718 return devlist;
6719#endif
6720 /* scroll through /sys/dev/block looking for devices attached to
6721 * this hba
6722 */
6723 dir = opendir("/sys/dev/block");
6724 for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) {
6725 int fd;
6726 char buf[1024];
6727 int major, minor;
6728 char *path = NULL;
6729 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
6730 continue;
6731 path = devt_to_devpath(makedev(major, minor));
6732 if (!path)
6733 continue;
6734 if (!path_attached_to_hba(path, hba_path)) {
6735 free(path);
6736 path = NULL;
6737 continue;
6738 }
6739 free(path);
6740 path = NULL;
6741 fd = dev_open(ent->d_name, O_RDONLY);
6742 if (fd >= 0) {
6743 fd2devname(fd, buf);
6744 close(fd);
6745 } else {
e7b84f9d 6746 pr_err("cannot open device: %s\n",
ca9de185
LM		 6747 ent->d_name);
6748 continue;
6749 }
6750
503975b9
N		 6751 dv = xcalloc(1, sizeof(*dv));
6752 dv->devname = xstrdup(buf);
ca9de185
LM		 6753 dv->next = devlist;
6754 devlist = dv;
6755 }
6756 if (err) {
6757 while(devlist) {
6758 dv = devlist;
6759 devlist = devlist->next;
6760 free(dv->devname);
6761 free(dv);
6762 }
6763 }
562aa102 6764 closedir(dir);
ca9de185
LM		 6765 return devlist;
6766}
6767
6768static int
6769count_volumes_list(struct md_list *devlist, char *homehost,
6770 int verbose, int *found)
6771{
6772 struct md_list *tmpdev;
6773 int count = 0;
594dc1b8 6774 struct supertype *st;
ca9de185
LM		 6775
6776 /* first walk the list of devices to find a consistent set
6777 * that match the criterea, if that is possible.
6778 * We flag the ones we like with 'used'.
6779 */
6780 *found = 0;
6781 st = match_metadata_desc_imsm("imsm");
6782 if (st == NULL) {
676e87a8 6783 pr_vrb("cannot allocate memory for imsm supertype\n");
ca9de185
LM		 6784 return 0;
6785 }
6786
6787 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
6788 char *devname = tmpdev->devname;
0a6bff09 6789 dev_t rdev;
ca9de185
LM		 6790 struct supertype *tst;
6791 int dfd;
6792 if (tmpdev->used > 1)
6793 continue;
6794 tst = dup_super(st);
6795 if (tst == NULL) {
676e87a8 6796 pr_vrb("cannot allocate memory for imsm supertype\n");
ca9de185
LM		 6797 goto err_1;
6798 }
6799 tmpdev->container = 0;
6800 dfd = dev_open(devname, O_RDONLY|O_EXCL);
6801 if (dfd < 0) {
1ade5cc1 6802 dprintf("cannot open device %s: %s\n",
ca9de185
LM		 6803 devname, strerror(errno));
6804 tmpdev->used = 2;
0a6bff09 6805 } else if (!fstat_is_blkdev(dfd, devname, &rdev)) {
ca9de185
LM		 6806 tmpdev->used = 2;
6807 } else if (must_be_container(dfd)) {
6808 struct supertype *cst;
6809 cst = super_by_fd(dfd, NULL);
6810 if (cst == NULL) {
1ade5cc1 6811 dprintf("cannot recognize container type %s\n",
ca9de185
LM		 6812 devname);
6813 tmpdev->used = 2;
6814 } else if (tst->ss != st->ss) {
1ade5cc1 6815 dprintf("non-imsm container - ignore it: %s\n",
ca9de185
LM		 6816 devname);
6817 tmpdev->used = 2;
6818 } else if (!tst->ss->load_container ||
6819 tst->ss->load_container(tst, dfd, NULL))
6820 tmpdev->used = 2;
6821 else {
6822 tmpdev->container = 1;
6823 }
6824 if (cst)
6825 cst->ss->free_super(cst);
6826 } else {
0a6bff09 6827 tmpdev->st_rdev = rdev;
ca9de185 6828 if (tst->ss->load_super(tst,dfd, NULL)) {
1ade5cc1 6829 dprintf("no RAID superblock on %s\n",
ca9de185
LM		 6830 devname);
6831 tmpdev->used = 2;
6832 } else if (tst->ss->compare_super == NULL) {
1ade5cc1 6833 dprintf("Cannot assemble %s metadata on %s\n",
ca9de185
LM		 6834 tst->ss->name, devname);
6835 tmpdev->used = 2;
6836 }
6837 }
6838 if (dfd >= 0)
6839 close(dfd);
6840 if (tmpdev->used == 2 || tmpdev->used == 4) {
6841 /* Ignore unrecognised devices during auto-assembly */
6842 goto loop;
6843 }
6844 else {
6845 struct mdinfo info;
6846 tst->ss->getinfo_super(tst, &info, NULL);
6847
6848 if (st->minor_version == -1)
6849 st->minor_version = tst->minor_version;
6850
6851 if (memcmp(info.uuid, uuid_zero,
6852 sizeof(int[4])) == 0) {
6853 /* this is a floating spare. It cannot define
6854 * an array unless there are no more arrays of
6855 * this type to be found. It can be included
6856 * in an array of this type though.
6857 */
6858 tmpdev->used = 3;
6859 goto loop;
6860 }
6861
6862 if (st->ss != tst->ss ||
6863 st->minor_version != tst->minor_version ||
6864 st->ss->compare_super(st, tst) != 0) {
6865 /* Some mismatch. If exactly one array matches this host,
6866 * we can resolve on that one.
6867 * Or, if we are auto assembling, we just ignore the second
6868 * for now.
6869 */
1ade5cc1 6870 dprintf("superblock on %s doesn't match others - assembly aborted\n",
ca9de185
LM		 6871 devname);
6872 goto loop;
6873 }
6874 tmpdev->used = 1;
6875 *found = 1;
6876 dprintf("found: devname: %s\n", devname);
6877 }
6878 loop:
6879 if (tst)
6880 tst->ss->free_super(tst);
6881 }
6882 if (*found != 0) {
6883 int err;
6884 if ((err = load_super_imsm_all(st, -1, &st->sb, NULL, devlist, 0)) == 0) {
6885 struct mdinfo *iter, *head = st->ss->container_content(st, NULL);
6886 for (iter = head; iter; iter = iter->next) {
6887 dprintf("content->text_version: %s vol\n",
6888 iter->text_version);
6889 if (iter->array.state & (1<<MD_SB_BLOCK_VOLUME)) {
6890 /* do not assemble arrays with unsupported
6891 configurations */
1ade5cc1 6892 dprintf("Cannot activate member %s.\n",
ca9de185
LM		 6893 iter->text_version);
6894 } else
6895 count++;
6896 }
6897 sysfs_free(head);
6898
6899 } else {
1ade5cc1 6900 dprintf("No valid super block on device list: err: %d %p\n",
ca9de185
LM		 6901 err, st->sb);
6902 }
6903 } else {
1ade5cc1 6904 dprintf("no more devices to examine\n");
ca9de185
LM		 6905 }
6906
6907 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
089f9d79 6908 if (tmpdev->used == 1 && tmpdev->found) {
ca9de185
LM		 6909 if (count) {
6910 if (count < tmpdev->found)
6911 count = 0;
6912 else
6913 count -= tmpdev->found;
6914 }
6915 }
6916 if (tmpdev->used == 1)
6917 tmpdev->used = 4;
6918 }
6919 err_1:
6920 if (st)
6921 st->ss->free_super(st);
6922 return count;
6923}
6924
d3c11416
AO		 6925static int __count_volumes(char *hba_path, int dpa, int verbose,
6926 int cmp_hba_path)
ca9de185 6927{
72a45777 6928 struct sys_dev *idev, *intel_devices = find_intel_devices();
ca9de185 6929 int count = 0;
72a45777
PB		 6930 const struct orom_entry *entry;
6931 struct devid_list *dv, *devid_list;
ca9de185 6932
d3c11416 6933 if (!hba_path)
ca9de185
LM		 6934 return 0;
6935
72a45777 6936 for (idev = intel_devices; idev; idev = idev->next) {
d3c11416
AO		 6937 if (strstr(idev->path, hba_path))
6938 break;
72a45777
PB		 6939 }
6940
6941 if (!idev || !idev->dev_id)
ca9de185 6942 return 0;
72a45777
PB		 6943
6944 entry = get_orom_entry_by_device_id(idev->dev_id);
6945
6946 if (!entry || !entry->devid_list)
6947 return 0;
6948
6949 devid_list = entry->devid_list;
6950 for (dv = devid_list; dv; dv = dv->next) {
594dc1b8 6951 struct md_list *devlist;
d3c11416
AO		 6952 struct sys_dev *device = NULL;
6953 char *hpath;
72a45777
PB		 6954 int found = 0;
6955
d3c11416
AO		 6956 if (cmp_hba_path)
6957 device = device_by_id_and_path(dv->devid, hba_path);
6958 else
6959 device = device_by_id(dv->devid);
6960
72a45777 6961 if (device)
d3c11416 6962 hpath = device->path;
72a45777
PB		 6963 else
6964 return 0;
6965
d3c11416 6966 devlist = get_devices(hpath);
72a45777
PB		 6967 /* if no intel devices return zero volumes */
6968 if (devlist == NULL)
6969 return 0;
6970
d3c11416
AO		 6971 count += active_arrays_by_format("imsm", hpath, &devlist, dpa,
6972 verbose);
6973 dprintf("path: %s active arrays: %d\n", hpath, count);
72a45777
PB		 6974 if (devlist == NULL)
6975 return 0;
6976 do {
6977 found = 0;
6978 count += count_volumes_list(devlist,
6979 NULL,
6980 verbose,
6981 &found);
6982 dprintf("found %d count: %d\n", found, count);
6983 } while (found);
6984
d3c11416 6985 dprintf("path: %s total number of volumes: %d\n", hpath, count);
72a45777
PB		 6986
6987 while (devlist) {
6988 struct md_list *dv = devlist;
6989 devlist = devlist->next;
6990 free(dv->devname);
6991 free(dv);
6992 }
ca9de185
LM		 6993 }
6994 return count;
6995}
6996
d3c11416
AO		 6997static int count_volumes(struct intel_hba *hba, int dpa, int verbose)
6998{
6999 if (!hba)
7000 return 0;
7001 if (hba->type == SYS_DEV_VMD) {
7002 struct sys_dev *dev;
7003 int count = 0;
7004
7005 for (dev = find_intel_devices(); dev; dev = dev->next) {
7006 if (dev->type == SYS_DEV_VMD)
7007 count += __count_volumes(dev->path, dpa,
7008 verbose, 1);
7009 }
7010 return count;
7011 }
7012 return __count_volumes(hba->path, dpa, verbose, 0);
7013}
7014
cd9d1ac7
DW		 7015static int imsm_default_chunk(const struct imsm_orom *orom)
7016{
7017 /* up to 512 if the plaform supports it, otherwise the platform max.
7018 * 128 if no platform detected
7019 */
7020 int fs = max(7, orom ? fls(orom->sss) : 0);
7021
7022 return min(512, (1 << fs));
7023}
73408129 7024
6592ce37
DW		 7025static int
7026validate_geometry_imsm_orom(struct intel_super *super, int level, int layout,
2cc699af 7027 int raiddisks, int *chunk, unsigned long long size, int verbose)
6592ce37 7028{
660260d0
DW		 7029 /* check/set platform and metadata limits/defaults */
7030 if (super->orom && raiddisks > super->orom->dpa) {
676e87a8 7031 pr_vrb("platform supports a maximum of %d disks per array\n",
660260d0 7032 super->orom->dpa);
73408129
LM		 7033 return 0;
7034 }
7035
5d500228 7036 /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */
660260d0 7037 if (!is_raid_level_supported(super->orom, level, raiddisks)) {
676e87a8 7038 pr_vrb("platform does not support raid%d with %d disk%s\n",
6592ce37
DW		 7039 level, raiddisks, raiddisks > 1 ? "s" : "");
7040 return 0;
7041 }
cd9d1ac7 7042
7ccc4cc4 7043 if (*chunk == 0 || *chunk == UnSet)
cd9d1ac7
DW		 7044 *chunk = imsm_default_chunk(super->orom);
7045
7ccc4cc4 7046 if (super->orom && !imsm_orom_has_chunk(super->orom, *chunk)) {
676e87a8 7047 pr_vrb("platform does not support a chunk size of: %d\n", *chunk);
cd9d1ac7 7048 return 0;
6592ce37 7049 }
cd9d1ac7 7050
6592ce37
DW		 7051 if (layout != imsm_level_to_layout(level)) {
7052 if (level == 5)
676e87a8 7053 pr_vrb("imsm raid 5 only supports the left-asymmetric layout\n");
6592ce37 7054 else if (level == 10)
676e87a8 7055 pr_vrb("imsm raid 10 only supports the n2 layout\n");
6592ce37 7056 else
676e87a8 7057 pr_vrb("imsm unknown layout %#x for this raid level %d\n",
6592ce37
DW		 7058 layout, level);
7059 return 0;
7060 }
2cc699af 7061
7ccc4cc4 7062 if (super->orom && (super->orom->attr & IMSM_OROM_ATTR_2TB) == 0 &&
2cc699af 7063 (calc_array_size(level, raiddisks, layout, *chunk, size) >> 32) > 0) {
676e87a8 7064 pr_vrb("platform does not support a volume size over 2TB\n");
2cc699af
CA		 7065 return 0;
7066 }
614902f6 7067
6592ce37
DW		 7068 return 1;
7069}
7070
1011e834 7071/* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd
c2c087e6
DW		 7072 * FIX ME add ahci details
7073 */
8b353278 7074static int validate_geometry_imsm_volume(struct supertype *st, int level,
c21e737b 7075 int layout, int raiddisks, int *chunk,
af4348dd
N		 7076 unsigned long long size,
7077 unsigned long long data_offset,
7078 char *dev,
2c514b71
NB		 7079 unsigned long long *freesize,
7080 int verbose)
cdddbdbc 7081{
9e04ac1c 7082 dev_t rdev;
c2c087e6 7083 struct intel_super *super = st->sb;
b2916f25 7084 struct imsm_super *mpb;
c2c087e6
DW		 7085 struct dl *dl;
7086 unsigned long long pos = 0;
7087 unsigned long long maxsize;
7088 struct extent *e;
7089 int i;
cdddbdbc 7090
88c32bb1
DW		 7091 /* We must have the container info already read in. */
7092 if (!super)
c2c087e6
DW		 7093 return 0;
7094
b2916f25
JS		 7095 mpb = super->anchor;
7096
2cc699af 7097 if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, size, verbose)) {
3e684231 7098 pr_err("RAID geometry validation failed. Cannot proceed with the action(s).\n");
c2c087e6 7099 return 0;
d54559f0 7100 }
c2c087e6
DW		 7101 if (!dev) {
7102 /* General test: make sure there is space for
2da8544a
DW		 7103 * 'raiddisks' device extents of size 'size' at a given
7104 * offset
c2c087e6 7105 */
e46273eb 7106 unsigned long long minsize = size;
b7528a20 7107 unsigned long long start_offset = MaxSector;
c2c087e6
DW		 7108 int dcnt = 0;
7109 if (minsize == 0)
7110 minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
7111 for (dl = super->disks; dl ; dl = dl->next) {
7112 int found = 0;
7113
bf5a934a 7114 pos = 0;
c2c087e6
DW		 7115 i = 0;
7116 e = get_extents(super, dl);
7117 if (!e) continue;
7118 do {
7119 unsigned long long esize;
7120 esize = e[i].start - pos;
7121 if (esize >= minsize)
7122 found = 1;
b7528a20 7123 if (found && start_offset == MaxSector) {
2da8544a
DW		 7124 start_offset = pos;
7125 break;
7126 } else if (found && pos != start_offset) {
7127 found = 0;
7128 break;
7129 }
c2c087e6
DW		 7130 pos = e[i].start + e[i].size;
7131 i++;
7132 } while (e[i-1].size);
7133 if (found)
7134 dcnt++;
7135 free(e);
7136 }
7137 if (dcnt < raiddisks) {
2c514b71 7138 if (verbose)
7a862a02 7139 pr_err("imsm: Not enough devices with space for this array (%d < %d)\n",
2c514b71 7140 dcnt, raiddisks);
c2c087e6
DW		 7141 return 0;
7142 }
7143 return 1;
7144 }
0dcecb2e 7145
c2c087e6 7146 /* This device must be a member of the set */
9e04ac1c 7147 if (!stat_is_blkdev(dev, &rdev))
c2c087e6
DW		 7148 return 0;
7149 for (dl = super->disks ; dl ; dl = dl->next) {
9e04ac1c
ZL		 7150 if (dl->major == (int)major(rdev) &&
7151 dl->minor == (int)minor(rdev))
c2c087e6
DW		 7152 break;
7153 }
7154 if (!dl) {
2c514b71 7155 if (verbose)
7a862a02 7156 pr_err("%s is not in the same imsm set\n", dev);
c2c087e6 7157 return 0;
a20d2ba5
DW		 7158 } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) {
7159 /* If a volume is present then the current creation attempt
7160 * cannot incorporate new spares because the orom may not
7161 * understand this configuration (all member disks must be
7162 * members of each array in the container).
7163 */
7a862a02
N		 7164 pr_err("%s is a spare and a volume is already defined for this container\n", dev);
7165 pr_err("The option-rom requires all member disks to be a member of all volumes\n");
a20d2ba5 7166 return 0;
5fe62b94
WD		 7167 } else if (super->orom && mpb->num_raid_devs > 0 &&
7168 mpb->num_disks != raiddisks) {
7a862a02 7169 pr_err("The option-rom requires all member disks to be a member of all volumes\n");
5fe62b94 7170 return 0;
c2c087e6 7171 }
0dcecb2e
DW		 7172
7173 /* retrieve the largest free space block */
c2c087e6
DW		 7174 e = get_extents(super, dl);
7175 maxsize = 0;
7176 i = 0;
0dcecb2e
DW		 7177 if (e) {
7178 do {
7179 unsigned long long esize;
7180
7181 esize = e[i].start - pos;
7182 if (esize >= maxsize)
7183 maxsize = esize;
7184 pos = e[i].start + e[i].size;
7185 i++;
7186 } while (e[i-1].size);
7187 dl->e = e;
7188 dl->extent_cnt = i;
7189 } else {
7190 if (verbose)
e7b84f9d 7191 pr_err("unable to determine free space for: %s\n",
0dcecb2e
DW		 7192 dev);
7193 return 0;
7194 }
7195 if (maxsize < size) {
7196 if (verbose)
e7b84f9d 7197 pr_err("%s not enough space (%llu < %llu)\n",
0dcecb2e
DW		 7198 dev, maxsize, size);
7199 return 0;
7200 }
7201
7202 /* count total number of extents for merge */
7203 i = 0;
7204 for (dl = super->disks; dl; dl = dl->next)
7205 if (dl->e)
7206 i += dl->extent_cnt;
7207
7208 maxsize = merge_extents(super, i);
3baa56ab
LO		 7209
7210 if (!check_env("IMSM_NO_PLATFORM") &&
7211 mpb->num_raid_devs > 0 && size && size != maxsize) {
7a862a02 7212 pr_err("attempting to create a second volume with size less then remaining space. Aborting...\n");
3baa56ab
LO		 7213 return 0;
7214 }
7215
a7dd165b 7216 if (maxsize < size || maxsize == 0) {
b3071342
LD		 7217 if (verbose) {
7218 if (maxsize == 0)
7a862a02 7219 pr_err("no free space left on device. Aborting...\n");
b3071342 7220 else
7a862a02 7221 pr_err("not enough space to create volume of given size (%llu < %llu). Aborting...\n",
b3071342
LD		 7222 maxsize, size);
7223 }
0dcecb2e 7224 return 0;
0dcecb2e
DW		 7225 }
7226
c2c087e6
DW		 7227 *freesize = maxsize;
7228
ca9de185 7229 if (super->orom) {
72a45777 7230 int count = count_volumes(super->hba,
ca9de185
LM		 7231 super->orom->dpa, verbose);
7232 if (super->orom->vphba <= count) {
676e87a8 7233 pr_vrb("platform does not support more than %d raid volumes.\n",
ca9de185
LM		 7234 super->orom->vphba);
7235 return 0;
7236 }
7237 }
c2c087e6 7238 return 1;
cdddbdbc
DW		 7239}
7240
13bcac90 7241static int imsm_get_free_size(struct supertype *st, int raiddisks,
efb30e7f
DW		 7242 unsigned long long size, int chunk,
7243 unsigned long long *freesize)
7244{
7245 struct intel_super *super = st->sb;
7246 struct imsm_super *mpb = super->anchor;
7247 struct dl *dl;
7248 int i;
7249 int extent_cnt;
7250 struct extent *e;
7251 unsigned long long maxsize;
7252 unsigned long long minsize;
7253 int cnt;
7254 int used;
7255
7256 /* find the largest common start free region of the possible disks */
7257 used = 0;
7258 extent_cnt = 0;
7259 cnt = 0;
7260 for (dl = super->disks; dl; dl = dl->next) {
7261 dl->raiddisk = -1;
7262
7263 if (dl->index >= 0)
7264 used++;
7265
7266 /* don't activate new spares if we are orom constrained
7267 * and there is already a volume active in the container
7268 */
7269 if (super->orom && dl->index < 0 && mpb->num_raid_devs)
7270 continue;
7271
7272 e = get_extents(super, dl);
7273 if (!e)
7274 continue;
7275 for (i = 1; e[i-1].size; i++)
7276 ;
7277 dl->e = e;
7278 dl->extent_cnt = i;
7279 extent_cnt += i;
7280 cnt++;
7281 }
7282
7283 maxsize = merge_extents(super, extent_cnt);
7284 minsize = size;
7285 if (size == 0)
612e59d8
CA		 7286 /* chunk is in K */
7287 minsize = chunk * 2;
efb30e7f
DW		 7288
7289 if (cnt < raiddisks ||
7290 (super->orom && used && used != raiddisks) ||
a7dd165b
DW		 7291 maxsize < minsize ||
7292 maxsize == 0) {
e7b84f9d 7293 pr_err("not enough devices with space to create array.\n");
efb30e7f
DW		 7294 return 0; /* No enough free spaces large enough */
7295 }
7296
7297 if (size == 0) {
7298 size = maxsize;
7299 if (chunk) {
612e59d8
CA		 7300 size /= 2 * chunk;
7301 size *= 2 * chunk;
efb30e7f 7302 }
f878b242
LM		 7303 maxsize = size;
7304 }
7305 if (!check_env("IMSM_NO_PLATFORM") &&
7306 mpb->num_raid_devs > 0 && size && size != maxsize) {
7a862a02 7307 pr_err("attempting to create a second volume with size less then remaining space. Aborting...\n");
f878b242 7308 return 0;
efb30e7f 7309 }
efb30e7f
DW		 7310 cnt = 0;
7311 for (dl = super->disks; dl; dl = dl->next)
7312 if (dl->e)
7313 dl->raiddisk = cnt++;
7314
7315 *freesize = size;
7316
13bcac90
AK		 7317 dprintf("imsm: imsm_get_free_size() returns : %llu\n", size);
7318
efb30e7f
DW		 7319 return 1;
7320}
7321
13bcac90
AK		 7322static int reserve_space(struct supertype *st, int raiddisks,
7323 unsigned long long size, int chunk,
7324 unsigned long long *freesize)
7325{
7326 struct intel_super *super = st->sb;
7327 struct dl *dl;
7328 int cnt;
7329 int rv = 0;
7330
7331 rv = imsm_get_free_size(st, raiddisks, size, chunk, freesize);
7332 if (rv) {
7333 cnt = 0;
7334 for (dl = super->disks; dl; dl = dl->next)
7335 if (dl->e)
7336 dl->raiddisk = cnt++;
7337 rv = 1;
7338 }
7339
7340 return rv;
7341}
7342
bf5a934a 7343static int validate_geometry_imsm(struct supertype *st, int level, int layout,
c21e737b 7344 int raiddisks, int *chunk, unsigned long long size,
af4348dd 7345 unsigned long long data_offset,
bf5a934a 7346 char *dev, unsigned long long *freesize,
5308f117 7347 int consistency_policy, int verbose)
bf5a934a
DW		 7348{
7349 int fd, cfd;
7350 struct mdinfo *sra;
20cbe8d2 7351 int is_member = 0;
bf5a934a 7352
d54559f0
LM		 7353 /* load capability
7354 * if given unused devices create a container
bf5a934a
DW		 7355 * if given given devices in a container create a member volume
7356 */
7357 if (level == LEVEL_CONTAINER) {
7358 /* Must be a fresh device to add to a container */
7359 return validate_geometry_imsm_container(st, level, layout,
c21e737b 7360 raiddisks,
7ccc4cc4 7361 *chunk,
af4348dd 7362 size, data_offset,
bf5a934a
DW		 7363 dev, freesize,
7364 verbose);
7365 }
9587c373 7366
54865c30
RS		 7367 if (size && ((size < 1024) || (*chunk != UnSet &&
7368 size < (unsigned long long) *chunk))) {
7369 pr_err("Given size must be greater than 1M and chunk size.\n");
7370 /* Depends on algorithm in Create.c :
7371 * if container was given (dev == NULL) return -1,
7372 * if block device was given ( dev != NULL) return 0.
7373 */
7374 return dev ? -1 : 0;
7375 }
7376
8592f29d 7377 if (!dev) {
e91a3bad 7378 if (st->sb) {
ca9de185 7379 struct intel_super *super = st->sb;
e91a3bad 7380 if (!validate_geometry_imsm_orom(st->sb, level, layout,
2cc699af 7381 raiddisks, chunk, size,
e91a3bad
LM		 7382 verbose))
7383 return 0;
efb30e7f
DW		 7384 /* we are being asked to automatically layout a
7385 * new volume based on the current contents of
7386 * the container. If the the parameters can be
7387 * satisfied reserve_space will record the disks,
7388 * start offset, and size of the volume to be
7389 * created. add_to_super and getinfo_super
7390 * detect when autolayout is in progress.
7391 */
ca9de185
LM		 7392 /* assuming that freesize is always given when array is
7393 created */
7394 if (super->orom && freesize) {
7395 int count;
72a45777 7396 count = count_volumes(super->hba,
ca9de185
LM		 7397 super->orom->dpa, verbose);
7398 if (super->orom->vphba <= count) {
676e87a8 7399 pr_vrb("platform does not support more than %d raid volumes.\n",
ca9de185
LM		 7400 super->orom->vphba);
7401 return 0;
7402 }
7403 }
e91a3bad
LM		 7404 if (freesize)
7405 return reserve_space(st, raiddisks, size,
7ccc4cc4 7406 *chunk, freesize);
8592f29d
N		 7407 }
7408 return 1;
7409 }
bf5a934a
DW		 7410 if (st->sb) {
7411 /* creating in a given container */
7412 return validate_geometry_imsm_volume(st, level, layout,
7413 raiddisks, chunk, size,
af4348dd 7414 data_offset,
bf5a934a
DW		 7415 dev, freesize, verbose);
7416 }
7417
bf5a934a
DW		 7418 /* This device needs to be a device in an 'imsm' container */
7419 fd = open(dev, O_RDONLY|O_EXCL, 0);
7420 if (fd >= 0) {
7421 if (verbose)
e7b84f9d
N		 7422 pr_err("Cannot create this array on device %s\n",
7423 dev);
bf5a934a
DW		 7424 close(fd);
7425 return 0;
7426 }
7427 if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) {
7428 if (verbose)
e7b84f9d 7429 pr_err("Cannot open %s: %s\n",
bf5a934a
DW		 7430 dev, strerror(errno));
7431 return 0;
7432 }
7433 /* Well, it is in use by someone, maybe an 'imsm' container. */
7434 cfd = open_container(fd);
20cbe8d2 7435 close(fd);
bf5a934a 7436 if (cfd < 0) {
bf5a934a 7437 if (verbose)
e7b84f9d 7438 pr_err("Cannot use %s: It is busy\n",
bf5a934a
DW		 7439 dev);
7440 return 0;
7441 }
4dd2df09 7442 sra = sysfs_read(cfd, NULL, GET_VERSION);
bf5a934a 7443 if (sra && sra->array.major_version == -1 &&
20cbe8d2
AW		 7444 strcmp(sra->text_version, "imsm") == 0)
7445 is_member = 1;
7446 sysfs_free(sra);
7447 if (is_member) {
bf5a934a
DW		 7448 /* This is a member of a imsm container. Load the container
7449 * and try to create a volume
7450 */
7451 struct intel_super *super;
7452
ec50f7b6 7453 if (load_super_imsm_all(st, cfd, (void **) &super, NULL, NULL, 1) == 0) {
bf5a934a 7454 st->sb = super;
4dd2df09 7455 strcpy(st->container_devnm, fd2devnm(cfd));
bf5a934a
DW		 7456 close(cfd);
7457 return validate_geometry_imsm_volume(st, level, layout,
7458 raiddisks, chunk,
af4348dd 7459 size, data_offset, dev,
ecbd9e81
N		 7460 freesize, 1)
7461 ? 1 : -1;
bf5a934a 7462 }
20cbe8d2 7463 }
bf5a934a 7464
20cbe8d2 7465 if (verbose)
e7b84f9d 7466 pr_err("failed container membership check\n");
20cbe8d2
AW		 7467
7468 close(cfd);
7469 return 0;
bf5a934a 7470}
0bd16cf2 7471
30f58b22 7472static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk)
0bd16cf2
DJ		 7473{
7474 struct intel_super *super = st->sb;
7475
30f58b22
DW		 7476 if (level && *level == UnSet)
7477 *level = LEVEL_CONTAINER;
7478
7479 if (level && layout && *layout == UnSet)
7480 *layout = imsm_level_to_layout(*level);
0bd16cf2 7481
cd9d1ac7
DW		 7482 if (chunk && (*chunk == UnSet || *chunk == 0))
7483 *chunk = imsm_default_chunk(super->orom);
0bd16cf2
DJ		 7484}
7485
33414a01
DW		 7486static void handle_missing(struct intel_super *super, struct imsm_dev *dev);
7487
7488static int kill_subarray_imsm(struct supertype *st)
7489{
7490 /* remove the subarray currently referenced by ->current_vol */
7491 __u8 i;
7492 struct intel_dev **dp;
7493 struct intel_super *super = st->sb;
7494 __u8 current_vol = super->current_vol;
7495 struct imsm_super *mpb = super->anchor;
7496
7497 if (super->current_vol < 0)
7498 return 2;
7499 super->current_vol = -1; /* invalidate subarray cursor */
7500
7501 /* block deletions that would change the uuid of active subarrays
7502 *
7503 * FIXME when immutable ids are available, but note that we'll
7504 * also need to fixup the invalidated/active subarray indexes in
7505 * mdstat
7506 */
7507 for (i = 0; i < mpb->num_raid_devs; i++) {
7508 char subarray[4];
7509
7510 if (i < current_vol)
7511 continue;
7512 sprintf(subarray, "%u", i);
4dd2df09 7513 if (is_subarray_active(subarray, st->devnm)) {
e7b84f9d
N		 7514 pr_err("deleting subarray-%d would change the UUID of active subarray-%d, aborting\n",
7515 current_vol, i);
33414a01
DW		 7516
7517 return 2;
7518 }
7519 }
7520
7521 if (st->update_tail) {
503975b9 7522 struct imsm_update_kill_array *u = xmalloc(sizeof(*u));
33414a01 7523
33414a01
DW		 7524 u->type = update_kill_array;
7525 u->dev_idx = current_vol;
7526 append_metadata_update(st, u, sizeof(*u));
7527
7528 return 0;
7529 }
7530
7531 for (dp = &super->devlist; *dp;)
7532 if ((*dp)->index == current_vol) {
7533 *dp = (*dp)->next;
7534 } else {
7535 handle_missing(super, (*dp)->dev);
7536 if ((*dp)->index > current_vol)
7537 (*dp)->index--;
7538 dp = &(*dp)->next;
7539 }
7540
7541 /* no more raid devices, all active components are now spares,
7542 * but of course failed are still failed
7543 */
7544 if (--mpb->num_raid_devs == 0) {
7545 struct dl *d;
7546
7547 for (d = super->disks; d; d = d->next)
a8619d23
AK		 7548 if (d->index > -2)
7549 mark_spare(d);
33414a01
DW		 7550 }
7551
7552 super->updates_pending++;
7553
7554 return 0;
7555}
aa534678 7556
a951a4f7 7557static int update_subarray_imsm(struct supertype *st, char *subarray,
fa56eddb 7558 char *update, struct mddev_ident *ident)
aa534678
DW		 7559{
7560 /* update the subarray currently referenced by ->current_vol */
7561 struct intel_super *super = st->sb;
7562 struct imsm_super *mpb = super->anchor;
7563
aa534678
DW		 7564 if (strcmp(update, "name") == 0) {
7565 char *name = ident->name;
a951a4f7
N		 7566 char *ep;
7567 int vol;
aa534678 7568
4dd2df09 7569 if (is_subarray_active(subarray, st->devnm)) {
e7b84f9d 7570 pr_err("Unable to update name of active subarray\n");
aa534678
DW		 7571 return 2;
7572 }
7573
7574 if (!check_name(super, name, 0))
7575 return 2;
7576
a951a4f7
N		 7577 vol = strtoul(subarray, &ep, 10);
7578 if (*ep != '\0' || vol >= super->anchor->num_raid_devs)
7579 return 2;
7580
aa534678 7581 if (st->update_tail) {
503975b9 7582 struct imsm_update_rename_array *u = xmalloc(sizeof(*u));
aa534678 7583
aa534678 7584 u->type = update_rename_array;
a951a4f7 7585 u->dev_idx = vol;
618f4e6d
XN		 7586 strncpy((char *) u->name, name, MAX_RAID_SERIAL_LEN);
7587 u->name[MAX_RAID_SERIAL_LEN-1] = '\0';
aa534678
DW		 7588 append_metadata_update(st, u, sizeof(*u));
7589 } else {
7590 struct imsm_dev *dev;
7591 int i;
7592
a951a4f7 7593 dev = get_imsm_dev(super, vol);
618f4e6d
XN		 7594 strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN);
7595 dev->volume[MAX_RAID_SERIAL_LEN-1] = '\0';
aa534678
DW		 7596 for (i = 0; i < mpb->num_raid_devs; i++) {
7597 dev = get_imsm_dev(super, i);
7598 handle_missing(super, dev);
7599 }
7600 super->updates_pending++;
7601 }
e6e9dd3f
AP		 7602 } else if (strcmp(update, "ppl") == 0 ||
7603 strcmp(update, "no-ppl") == 0) {
7604 int new_policy;
7605 char *ep;
7606 int vol = strtoul(subarray, &ep, 10);
7607
7608 if (*ep != '\0' || vol >= super->anchor->num_raid_devs)
7609 return 2;
7610
7611 if (strcmp(update, "ppl") == 0)
c2462068 7612 new_policy = RWH_MULTIPLE_DISTRIBUTED;
e6e9dd3f 7613 else
c2462068 7614 new_policy = RWH_MULTIPLE_OFF;
e6e9dd3f
AP		 7615
7616 if (st->update_tail) {
7617 struct imsm_update_rwh_policy *u = xmalloc(sizeof(*u));
7618
7619 u->type = update_rwh_policy;
7620 u->dev_idx = vol;
7621 u->new_policy = new_policy;
7622 append_metadata_update(st, u, sizeof(*u));
7623 } else {
7624 struct imsm_dev *dev;
7625
7626 dev = get_imsm_dev(super, vol);
7627 dev->rwh_policy = new_policy;
7628 super->updates_pending++;
7629 }
aa534678
DW		 7630 } else
7631 return 2;
7632
7633 return 0;
7634}
bf5a934a 7635
28bce06f
AK		 7636static int is_gen_migration(struct imsm_dev *dev)
7637{
7534230b
AK		 7638 if (dev == NULL)
7639 return 0;
7640
28bce06f
AK		 7641 if (!dev->vol.migr_state)
7642 return 0;
7643
7644 if (migr_type(dev) == MIGR_GEN_MIGR)
7645 return 1;
7646
7647 return 0;
7648}
7649
1e5c6983
DW		 7650static int is_rebuilding(struct imsm_dev *dev)
7651{
7652 struct imsm_map *migr_map;
7653
7654 if (!dev->vol.migr_state)
7655 return 0;
7656
7657 if (migr_type(dev) != MIGR_REBUILD)
7658 return 0;
7659
238c0a71 7660 migr_map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 7661
7662 if (migr_map->map_state == IMSM_T_STATE_DEGRADED)
7663 return 1;
7664 else
7665 return 0;
7666}
7667
6ce1fbf1
AK		 7668static int is_initializing(struct imsm_dev *dev)
7669{
7670 struct imsm_map *migr_map;
7671
7672 if (!dev->vol.migr_state)
7673 return 0;
7674
7675 if (migr_type(dev) != MIGR_INIT)
7676 return 0;
7677
238c0a71 7678 migr_map = get_imsm_map(dev, MAP_1);
6ce1fbf1
AK		 7679
7680 if (migr_map->map_state == IMSM_T_STATE_UNINITIALIZED)
7681 return 1;
7682
7683 return 0;
6ce1fbf1
AK		 7684}
7685
c47b0ff6
AK		 7686static void update_recovery_start(struct intel_super *super,
7687 struct imsm_dev *dev,
7688 struct mdinfo *array)
1e5c6983
DW		 7689{
7690 struct mdinfo *rebuild = NULL;
7691 struct mdinfo *d;
7692 __u32 units;
7693
7694 if (!is_rebuilding(dev))
7695 return;
7696
7697 /* Find the rebuild target, but punt on the dual rebuild case */
7698 for (d = array->devs; d; d = d->next)
7699 if (d->recovery_start == 0) {
7700 if (rebuild)
7701 return;
7702 rebuild = d;
7703 }
7704
4363fd80
DW		 7705 if (!rebuild) {
7706 /* (?) none of the disks are marked with
7707 * IMSM_ORD_REBUILD, so assume they are missing and the
7708 * disk_ord_tbl was not correctly updated
7709 */
1ade5cc1 7710 dprintf("failed to locate out-of-sync disk\n");
4363fd80
DW		 7711 return;
7712 }
7713
1e5c6983 7714 units = __le32_to_cpu(dev->vol.curr_migr_unit);
c47b0ff6 7715 rebuild->recovery_start = units * blocks_per_migr_unit(super, dev);
1e5c6983
DW		 7716}
7717
276d77db 7718static int recover_backup_imsm(struct supertype *st, struct mdinfo *info);
1e5c6983 7719
00bbdbda 7720static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray)
cdddbdbc 7721{
4f5bc454
DW		 7722 /* Given a container loaded by load_super_imsm_all,
7723 * extract information about all the arrays into
7724 * an mdinfo tree.
00bbdbda 7725 * If 'subarray' is given, just extract info about that array.
4f5bc454
DW		 7726 *
7727 * For each imsm_dev create an mdinfo, fill it in,
7728 * then look for matching devices in super->disks
7729 * and create appropriate device mdinfo.
7730 */
7731 struct intel_super *super = st->sb;
949c47a0 7732 struct imsm_super *mpb = super->anchor;
4f5bc454 7733 struct mdinfo *rest = NULL;
00bbdbda 7734 unsigned int i;
81219e70 7735 int sb_errors = 0;
abef11a3
AK		 7736 struct dl *d;
7737 int spare_disks = 0;
cdddbdbc 7738
19482bcc
AK		 7739 /* do not assemble arrays when not all attributes are supported */
7740 if (imsm_check_attributes(mpb->attributes) == 0) {
81219e70 7741 sb_errors = 1;
7a862a02 7742 pr_err("Unsupported attributes in IMSM metadata.Arrays activation is blocked.\n");
19482bcc
AK		 7743 }
7744
abef11a3
AK		 7745 /* count spare devices, not used in maps
7746 */
7747 for (d = super->disks; d; d = d->next)
7748 if (d->index == -1)
7749 spare_disks++;
7750
4f5bc454 7751 for (i = 0; i < mpb->num_raid_devs; i++) {
00bbdbda
N		 7752 struct imsm_dev *dev;
7753 struct imsm_map *map;
86e3692b 7754 struct imsm_map *map2;
4f5bc454 7755 struct mdinfo *this;
a6482415 7756 int slot;
a6482415 7757 int chunk;
00bbdbda 7758 char *ep;
8b9cd157 7759 int level;
00bbdbda
N		 7760
7761 if (subarray &&
7762 (i != strtoul(subarray, &ep, 10) || *ep != '\0'))
7763 continue;
7764
7765 dev = get_imsm_dev(super, i);
238c0a71
AK		 7766 map = get_imsm_map(dev, MAP_0);
7767 map2 = get_imsm_map(dev, MAP_1);
8b9cd157 7768 level = get_imsm_raid_level(map);
4f5bc454 7769
1ce0101c
DW		 7770 /* do not publish arrays that are in the middle of an
7771 * unsupported migration
7772 */
7773 if (dev->vol.migr_state &&
28bce06f 7774 (migr_type(dev) == MIGR_STATE_CHANGE)) {
7a862a02 7775 pr_err("cannot assemble volume '%.16s': unsupported migration in progress\n",
1ce0101c
DW		 7776 dev->volume);
7777 continue;
7778 }
2db86302
LM		 7779 /* do not publish arrays that are not support by controller's
7780 * OROM/EFI
7781 */
1ce0101c 7782
503975b9 7783 this = xmalloc(sizeof(*this));
4f5bc454 7784
301406c9 7785 super->current_vol = i;
a5d85af7 7786 getinfo_super_imsm_volume(st, this, NULL);
9894ec0d 7787 this->next = rest;
a6482415 7788 chunk = __le16_to_cpu(map->blocks_per_strip) >> 1;
81219e70
LM		 7789 /* mdadm does not support all metadata features- set the bit in all arrays state */
7790 if (!validate_geometry_imsm_orom(super,
8b9cd157
MK		 7791 level, /* RAID level */
7792 imsm_level_to_layout(level),
81219e70 7793 map->num_members, /* raid disks */
fcc2c9da 7794 &chunk, imsm_dev_size(dev),
81219e70 7795 1 /* verbose */)) {
7a862a02 7796 pr_err("IMSM RAID geometry validation failed. Array %s activation is blocked.\n",
81219e70
LM		 7797 dev->volume);
7798 this->array.state |=
7799 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
7800 (1<<MD_SB_BLOCK_VOLUME);
7801 }
81219e70
LM		 7802
7803 /* if array has bad blocks, set suitable bit in all arrays state */
7804 if (sb_errors)
7805 this->array.state |=
7806 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
7807 (1<<MD_SB_BLOCK_VOLUME);
7808
4f5bc454 7809 for (slot = 0 ; slot < map->num_members; slot++) {
1e5c6983 7810 unsigned long long recovery_start;
4f5bc454
DW		 7811 struct mdinfo *info_d;
7812 struct dl *d;
7813 int idx;
9a1608e5 7814 int skip;
7eef0453 7815 __u32 ord;
8b9cd157 7816 int missing = 0;
4f5bc454 7817
9a1608e5 7818 skip = 0;
238c0a71
AK		 7819 idx = get_imsm_disk_idx(dev, slot, MAP_0);
7820 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
4f5bc454
DW		 7821 for (d = super->disks; d ; d = d->next)
7822 if (d->index == idx)
0fbd635c 7823 break;
4f5bc454 7824
1e5c6983 7825 recovery_start = MaxSector;
4f5bc454 7826 if (d == NULL)
9a1608e5 7827 skip = 1;
25ed7e59 7828 if (d && is_failed(&d->disk))
9a1608e5 7829 skip = 1;
8b9cd157 7830 if (!skip && (ord & IMSM_ORD_REBUILD))
1e5c6983 7831 recovery_start = 0;
9a1608e5 7832
1011e834 7833 /*
9a1608e5 7834 * if we skip some disks the array will be assmebled degraded;
1e5c6983
DW		 7835 * reset resync start to avoid a dirty-degraded
7836 * situation when performing the intial sync
9a1608e5 7837 */
8b9cd157
MK		 7838 if (skip)
7839 missing++;
7840
7841 if (!(dev->vol.dirty & RAIDVOL_DIRTY)) {
7842 if ((!able_to_resync(level, missing) ||
7843 recovery_start == 0))
7844 this->resync_start = MaxSector;
7845 } else {
7846 /*
7847 * FIXME handle dirty degraded
7848 */
7849 }
7850
9a1608e5
DW		 7851 if (skip)
7852 continue;
4f5bc454 7853
503975b9 7854 info_d = xcalloc(1, sizeof(*info_d));
4f5bc454
DW		 7855 info_d->next = this->devs;
7856 this->devs = info_d;
7857
4f5bc454
DW		 7858 info_d->disk.number = d->index;
7859 info_d->disk.major = d->major;
7860 info_d->disk.minor = d->minor;
7861 info_d->disk.raid_disk = slot;
1e5c6983 7862 info_d->recovery_start = recovery_start;
86e3692b
AK		 7863 if (map2) {
7864 if (slot < map2->num_members)
7865 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK		 7866 else
7867 this->array.spare_disks++;
86e3692b
AK		 7868 } else {
7869 if (slot < map->num_members)
7870 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK		 7871 else
7872 this->array.spare_disks++;
86e3692b 7873 }
1e5c6983
DW		 7874 if (info_d->recovery_start == MaxSector)
7875 this->array.working_disks++;
4f5bc454
DW		 7876
7877 info_d->events = __le32_to_cpu(mpb->generation_num);
5551b113 7878 info_d->data_offset = pba_of_lba0(map);
44490938 7879 info_d->component_size = calc_component_size(map, dev);
06fb291a
PB		 7880
7881 if (map->raid_level == 5) {
2432ce9b
AP		 7882 info_d->ppl_sector = this->ppl_sector;
7883 info_d->ppl_size = this->ppl_size;
98e96bdb
AP		 7884 if (this->consistency_policy == CONSISTENCY_POLICY_PPL &&
7885 recovery_start == 0)
7886 this->resync_start = 0;
06fb291a 7887 }
b12796be 7888
5e46202e 7889 info_d->bb.supported = 1;
b12796be
TM		 7890 get_volume_badblocks(super->bbm_log, ord_to_idx(ord),
7891 info_d->data_offset,
7892 info_d->component_size,
7893 &info_d->bb);
4f5bc454 7894 }
1e5c6983 7895 /* now that the disk list is up-to-date fixup recovery_start */
c47b0ff6 7896 update_recovery_start(super, dev, this);
abef11a3 7897 this->array.spare_disks += spare_disks;
276d77db
AK		 7898
7899 /* check for reshape */
7900 if (this->reshape_active == 1)
7901 recover_backup_imsm(st, this);
9a1608e5 7902 rest = this;
4f5bc454
DW		 7903 }
7904
7905 return rest;
cdddbdbc
DW		 7906}
7907
3b451610
AK		 7908static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev,
7909 int failed, int look_in_map)
c2a1e7da 7910{
3b451610
AK		 7911 struct imsm_map *map;
7912
7913 map = get_imsm_map(dev, look_in_map);
c2a1e7da
DW		 7914
7915 if (!failed)
1011e834 7916 return map->map_state == IMSM_T_STATE_UNINITIALIZED ?
3393c6af 7917 IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL;
c2a1e7da
DW		 7918
7919 switch (get_imsm_raid_level(map)) {
7920 case 0:
7921 return IMSM_T_STATE_FAILED;
7922 break;
7923 case 1:
7924 if (failed < map->num_members)
7925 return IMSM_T_STATE_DEGRADED;
7926 else
7927 return IMSM_T_STATE_FAILED;
7928 break;
7929 case 10:
7930 {
7931 /**
c92a2527
DW		 7932 * check to see if any mirrors have failed, otherwise we
7933 * are degraded. Even numbered slots are mirrored on
7934 * slot+1
c2a1e7da 7935 */
c2a1e7da 7936 int i;
d9b420a5
N		 7937 /* gcc -Os complains that this is unused */
7938 int insync = insync;
c2a1e7da
DW		 7939
7940 for (i = 0; i < map->num_members; i++) {
238c0a71 7941 __u32 ord = get_imsm_ord_tbl_ent(dev, i, MAP_X);
c92a2527
DW		 7942 int idx = ord_to_idx(ord);
7943 struct imsm_disk *disk;
c2a1e7da 7944
c92a2527 7945 /* reset the potential in-sync count on even-numbered
1011e834 7946 * slots. num_copies is always 2 for imsm raid10
c92a2527
DW		 7947 */
7948 if ((i & 1) == 0)
7949 insync = 2;
c2a1e7da 7950
c92a2527 7951 disk = get_imsm_disk(super, idx);
25ed7e59 7952 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
c92a2527 7953 insync--;
c2a1e7da 7954
c92a2527
DW		 7955 /* no in-sync disks left in this mirror the
7956 * array has failed
7957 */
7958 if (insync == 0)
7959 return IMSM_T_STATE_FAILED;
c2a1e7da
DW		 7960 }
7961
7962 return IMSM_T_STATE_DEGRADED;
7963 }
7964 case 5:
7965 if (failed < 2)
7966 return IMSM_T_STATE_DEGRADED;
7967 else
7968 return IMSM_T_STATE_FAILED;
7969 break;
7970 default:
7971 break;
7972 }
7973
7974 return map->map_state;
7975}
7976
3b451610
AK		 7977static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev,
7978 int look_in_map)
c2a1e7da
DW		 7979{
7980 int i;
7981 int failed = 0;
7982 struct imsm_disk *disk;
d5985138
AK		 7983 struct imsm_map *map = get_imsm_map(dev, MAP_0);
7984 struct imsm_map *prev = get_imsm_map(dev, MAP_1);
68fe4598 7985 struct imsm_map *map_for_loop;
0556e1a2
DW		 7986 __u32 ord;
7987 int idx;
d5985138 7988 int idx_1;
c2a1e7da 7989
0556e1a2
DW		 7990 /* at the beginning of migration we set IMSM_ORD_REBUILD on
7991 * disks that are being rebuilt. New failures are recorded to
7992 * map[0]. So we look through all the disks we started with and
7993 * see if any failures are still present, or if any new ones
7994 * have arrived
0556e1a2 7995 */
d5985138
AK		 7996 map_for_loop = map;
7997 if (prev && (map->num_members < prev->num_members))
7998 map_for_loop = prev;
68fe4598
LD		 7999
8000 for (i = 0; i < map_for_loop->num_members; i++) {
d5985138 8001 idx_1 = -255;
238c0a71
AK		 8002 /* when MAP_X is passed both maps failures are counted
8003 */
d5985138 8004 if (prev &&
089f9d79
JS		 8005 (look_in_map == MAP_1 || look_in_map == MAP_X) &&
8006 i < prev->num_members) {
d5985138
AK		 8007 ord = __le32_to_cpu(prev->disk_ord_tbl[i]);
8008 idx_1 = ord_to_idx(ord);
c2a1e7da 8009
d5985138
AK		 8010 disk = get_imsm_disk(super, idx_1);
8011 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
8012 failed++;
8013 }
089f9d79
JS		 8014 if ((look_in_map == MAP_0 || look_in_map == MAP_X) &&
8015 i < map->num_members) {
d5985138
AK		 8016 ord = __le32_to_cpu(map->disk_ord_tbl[i]);
8017 idx = ord_to_idx(ord);
8018
8019 if (idx != idx_1) {
8020 disk = get_imsm_disk(super, idx);
8021 if (!disk || is_failed(disk) ||
8022 ord & IMSM_ORD_REBUILD)
8023 failed++;
8024 }
8025 }
c2a1e7da
DW		 8026 }
8027
8028 return failed;
845dea95
NB		 8029}
8030
97b4d0e9
DW		 8031static int imsm_open_new(struct supertype *c, struct active_array *a,
8032 char *inst)
8033{
8034 struct intel_super *super = c->sb;
8035 struct imsm_super *mpb = super->anchor;
bbab0940 8036 struct imsm_update_prealloc_bb_mem u;
9587c373 8037
97b4d0e9 8038 if (atoi(inst) >= mpb->num_raid_devs) {
1ade5cc1 8039 pr_err("subarry index %d, out of range\n", atoi(inst));
97b4d0e9
DW		 8040 return -ENODEV;
8041 }
8042
8043 dprintf("imsm: open_new %s\n", inst);
8044 a->info.container_member = atoi(inst);
bbab0940
TM		 8045
8046 u.type = update_prealloc_badblocks_mem;
8047 imsm_update_metadata_locally(c, &u, sizeof(u));
8048
97b4d0e9
DW		 8049 return 0;
8050}
8051
0c046afd
DW		 8052static int is_resyncing(struct imsm_dev *dev)
8053{
8054 struct imsm_map *migr_map;
8055
8056 if (!dev->vol.migr_state)
8057 return 0;
8058
1484e727
DW		 8059 if (migr_type(dev) == MIGR_INIT ||
8060 migr_type(dev) == MIGR_REPAIR)
0c046afd
DW		 8061 return 1;
8062
4c9bc37b
AK		 8063 if (migr_type(dev) == MIGR_GEN_MIGR)
8064 return 0;
8065
238c0a71 8066 migr_map = get_imsm_map(dev, MAP_1);
0c046afd 8067
089f9d79
JS		 8068 if (migr_map->map_state == IMSM_T_STATE_NORMAL &&
8069 dev->vol.migr_type != MIGR_GEN_MIGR)
0c046afd
DW		 8070 return 1;
8071 else
8072 return 0;
8073}
8074
0556e1a2 8075/* return true if we recorded new information */
4c9e8c1e
TM		 8076static int mark_failure(struct intel_super *super,
8077 struct imsm_dev *dev, struct imsm_disk *disk, int idx)
47ee5a45 8078{
0556e1a2
DW		 8079 __u32 ord;
8080 int slot;
8081 struct imsm_map *map;
86c54047
DW		 8082 char buf[MAX_RAID_SERIAL_LEN+3];
8083 unsigned int len, shift = 0;
0556e1a2
DW		 8084
8085 /* new failures are always set in map[0] */
238c0a71 8086 map = get_imsm_map(dev, MAP_0);
0556e1a2
DW		 8087
8088 slot = get_imsm_disk_slot(map, idx);
8089 if (slot < 0)
8090 return 0;
8091
8092 ord = __le32_to_cpu(map->disk_ord_tbl[slot]);
25ed7e59 8093 if (is_failed(disk) && (ord & IMSM_ORD_REBUILD))
0556e1a2
DW		 8094 return 0;
8095
7d0c5e24
LD		 8096 memcpy(buf, disk->serial, MAX_RAID_SERIAL_LEN);
8097 buf[MAX_RAID_SERIAL_LEN] = '\000';
8098 strcat(buf, ":0");
86c54047
DW		 8099 if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN)
8100 shift = len - MAX_RAID_SERIAL_LEN + 1;
8101 strncpy((char *)disk->serial, &buf[shift], MAX_RAID_SERIAL_LEN);
8102
f2f27e63 8103 disk->status |= FAILED_DISK;
0556e1a2 8104 set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD);
17788994
AK		 8105 /* mark failures in second map if second map exists and this disk
8106 * in this slot.
8107 * This is valid for migration, initialization and rebuild
8108 */
8109 if (dev->vol.migr_state) {
238c0a71 8110 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
0a108d63
AK		 8111 int slot2 = get_imsm_disk_slot(map2, idx);
8112
089f9d79 8113 if (slot2 < map2->num_members && slot2 >= 0)
0a108d63 8114 set_imsm_ord_tbl_ent(map2, slot2,
1ace8403
AK		 8115 idx | IMSM_ORD_REBUILD);
8116 }
f21e18ca 8117 if (map->failed_disk_num == 0xff)
0556e1a2 8118 map->failed_disk_num = slot;
4c9e8c1e
TM		 8119
8120 clear_disk_badblocks(super->bbm_log, ord_to_idx(ord));
8121
0556e1a2
DW		 8122 return 1;
8123}
8124
4c9e8c1e
TM		 8125static void mark_missing(struct intel_super *super,
8126 struct imsm_dev *dev, struct imsm_disk *disk, int idx)
0556e1a2 8127{
4c9e8c1e 8128 mark_failure(super, dev, disk, idx);
0556e1a2
DW		 8129
8130 if (disk->scsi_id == __cpu_to_le32(~(__u32)0))
8131 return;
8132
47ee5a45
DW		 8133 disk->scsi_id = __cpu_to_le32(~(__u32)0);
8134 memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1);
8135}
8136
33414a01
DW		 8137static void handle_missing(struct intel_super *super, struct imsm_dev *dev)
8138{
33414a01 8139 struct dl *dl;
33414a01
DW		 8140
8141 if (!super->missing)
8142 return;
33414a01 8143
79b68f1b
PC		 8144 /* When orom adds replacement for missing disk it does
8145 * not remove entry of missing disk, but just updates map with
8146 * new added disk. So it is not enough just to test if there is
8147 * any missing disk, we have to look if there are any failed disks
8148 * in map to stop migration */
8149
33414a01 8150 dprintf("imsm: mark missing\n");
3d59f0c0
AK		 8151 /* end process for initialization and rebuild only
8152 */
8153 if (is_gen_migration(dev) == 0) {
fb12a745 8154 int failed = imsm_count_failed(super, dev, MAP_0);
3d59f0c0 8155
fb12a745
TM		 8156 if (failed) {
8157 __u8 map_state;
8158 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8159 struct imsm_map *map1;
8160 int i, ord, ord_map1;
8161 int rebuilt = 1;
3d59f0c0 8162
fb12a745
TM		 8163 for (i = 0; i < map->num_members; i++) {
8164 ord = get_imsm_ord_tbl_ent(dev, i, MAP_0);
8165 if (!(ord & IMSM_ORD_REBUILD))
8166 continue;
8167
8168 map1 = get_imsm_map(dev, MAP_1);
8169 if (!map1)
8170 continue;
8171
8172 ord_map1 = __le32_to_cpu(map1->disk_ord_tbl[i]);
8173 if (ord_map1 & IMSM_ORD_REBUILD)
8174 rebuilt = 0;
8175 }
8176
8177 if (rebuilt) {
8178 map_state = imsm_check_degraded(super, dev,
8179 failed, MAP_0);
8180 end_migration(dev, super, map_state);
8181 }
8182 }
3d59f0c0 8183 }
33414a01 8184 for (dl = super->missing; dl; dl = dl->next)
4c9e8c1e 8185 mark_missing(super, dev, &dl->disk, dl->index);
33414a01
DW		 8186 super->updates_pending++;
8187}
8188
f3871fdc
AK		 8189static unsigned long long imsm_set_array_size(struct imsm_dev *dev,
8190 long long new_size)
70bdf0dc 8191{
70bdf0dc 8192 unsigned long long array_blocks;
9529d343
MD		 8193 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8194 int used_disks = imsm_num_data_members(map);
70bdf0dc
AK		 8195
8196 if (used_disks == 0) {
8197 /* when problems occures
8198 * return current array_blocks value
8199 */
fcc2c9da 8200 array_blocks = imsm_dev_size(dev);
70bdf0dc
AK		 8201
8202 return array_blocks;
8203 }
8204
8205 /* set array size in metadata
8206 */
9529d343 8207 if (new_size <= 0)
f3871fdc
AK		 8208 /* OLCE size change is caused by added disks
8209 */
44490938 8210 array_blocks = per_dev_array_size(map) * used_disks;
9529d343 8211 else
f3871fdc
AK		 8212 /* Online Volume Size Change
8213 * Using available free space
8214 */
8215 array_blocks = new_size;
70bdf0dc 8216
b53bfba6 8217 array_blocks = round_size_to_mb(array_blocks, used_disks);
fcc2c9da 8218 set_imsm_dev_size(dev, array_blocks);
70bdf0dc
AK		 8219
8220 return array_blocks;
8221}
8222
28bce06f
AK		 8223static void imsm_set_disk(struct active_array *a, int n, int state);
8224
0e2d1a4e
AK		 8225static void imsm_progress_container_reshape(struct intel_super *super)
8226{
8227 /* if no device has a migr_state, but some device has a
8228 * different number of members than the previous device, start
8229 * changing the number of devices in this device to match
8230 * previous.
8231 */
8232 struct imsm_super *mpb = super->anchor;
8233 int prev_disks = -1;
8234 int i;
1dfaa380 8235 int copy_map_size;
0e2d1a4e
AK		 8236
8237 for (i = 0; i < mpb->num_raid_devs; i++) {
8238 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 8239 struct imsm_map *map = get_imsm_map(dev, MAP_0);
0e2d1a4e
AK		 8240 struct imsm_map *map2;
8241 int prev_num_members;
0e2d1a4e
AK		 8242
8243 if (dev->vol.migr_state)
8244 return;
8245
8246 if (prev_disks == -1)
8247 prev_disks = map->num_members;
8248 if (prev_disks == map->num_members)
8249 continue;
8250
8251 /* OK, this array needs to enter reshape mode.
8252 * i.e it needs a migr_state
8253 */
8254
1dfaa380 8255 copy_map_size = sizeof_imsm_map(map);
0e2d1a4e
AK		 8256 prev_num_members = map->num_members;
8257 map->num_members = prev_disks;
8258 dev->vol.migr_state = 1;
8259 dev->vol.curr_migr_unit = 0;
ea672ee1 8260 set_migr_type(dev, MIGR_GEN_MIGR);
0e2d1a4e
AK		 8261 for (i = prev_num_members;
8262 i < map->num_members; i++)
8263 set_imsm_ord_tbl_ent(map, i, i);
238c0a71 8264 map2 = get_imsm_map(dev, MAP_1);
0e2d1a4e 8265 /* Copy the current map */
1dfaa380 8266 memcpy(map2, map, copy_map_size);
0e2d1a4e
AK		 8267 map2->num_members = prev_num_members;
8268
f3871fdc 8269 imsm_set_array_size(dev, -1);
51d83f5d 8270 super->clean_migration_record_by_mdmon = 1;
0e2d1a4e
AK		 8271 super->updates_pending++;
8272 }
8273}
8274
aad6f216 8275/* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild
0c046afd
DW		 8276 * states are handled in imsm_set_disk() with one exception, when a
8277 * resync is stopped due to a new failure this routine will set the
8278 * 'degraded' state for the array.
8279 */
01f157d7 8280static int imsm_set_array_state(struct active_array *a, int consistent)
a862209d
DW		 8281{
8282 int inst = a->info.container_member;
8283 struct intel_super *super = a->container->sb;
949c47a0 8284 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8285 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3b451610
AK		 8286 int failed = imsm_count_failed(super, dev, MAP_0);
8287 __u8 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
1e5c6983 8288 __u32 blocks_per_unit;
a862209d 8289
1af97990
AK		 8290 if (dev->vol.migr_state &&
8291 dev->vol.migr_type == MIGR_GEN_MIGR) {
8292 /* array state change is blocked due to reshape action
aad6f216
N		 8293 * We might need to
8294 * - abort the reshape (if last_checkpoint is 0 and action!= reshape)
8295 * - finish the reshape (if last_checkpoint is big and action != reshape)
8296 * - update curr_migr_unit
1af97990 8297 */
aad6f216
N		 8298 if (a->curr_action == reshape) {
8299 /* still reshaping, maybe update curr_migr_unit */
633b5610 8300 goto mark_checkpoint;
aad6f216
N		 8301 } else {
8302 if (a->last_checkpoint == 0 && a->prev_action == reshape) {
8303 /* for some reason we aborted the reshape.
b66e591b
AK		 8304 *
8305 * disable automatic metadata rollback
8306 * user action is required to recover process
aad6f216 8307 */
b66e591b 8308 if (0) {
238c0a71
AK		 8309 struct imsm_map *map2 =
8310 get_imsm_map(dev, MAP_1);
8311 dev->vol.migr_state = 0;
8312 set_migr_type(dev, 0);
8313 dev->vol.curr_migr_unit = 0;
8314 memcpy(map, map2,
8315 sizeof_imsm_map(map2));
8316 super->updates_pending++;
b66e591b 8317 }
aad6f216
N		 8318 }
8319 if (a->last_checkpoint >= a->info.component_size) {
8320 unsigned long long array_blocks;
8321 int used_disks;
e154ced3 8322 struct mdinfo *mdi;
aad6f216 8323
9529d343 8324 used_disks = imsm_num_data_members(map);
d55adef9
AK		 8325 if (used_disks > 0) {
8326 array_blocks =
44490938 8327 per_dev_array_size(map) *
d55adef9 8328 used_disks;
b53bfba6
TM		 8329 array_blocks =
8330 round_size_to_mb(array_blocks,
8331 used_disks);
d55adef9
AK		 8332 a->info.custom_array_size = array_blocks;
8333 /* encourage manager to update array
8334 * size
8335 */
e154ced3 8336
d55adef9 8337 a->check_reshape = 1;
633b5610 8338 }
e154ced3
AK		 8339 /* finalize online capacity expansion/reshape */
8340 for (mdi = a->info.devs; mdi; mdi = mdi->next)
8341 imsm_set_disk(a,
8342 mdi->disk.raid_disk,
8343 mdi->curr_state);
8344
0e2d1a4e 8345 imsm_progress_container_reshape(super);
e154ced3 8346 }
aad6f216 8347 }
1af97990
AK		 8348 }
8349
47ee5a45 8350 /* before we activate this array handle any missing disks */
33414a01
DW		 8351 if (consistent == 2)
8352 handle_missing(super, dev);
1e5c6983 8353
0c046afd 8354 if (consistent == 2 &&
b7941fd6 8355 (!is_resync_complete(&a->info) ||
0c046afd
DW		 8356 map_state != IMSM_T_STATE_NORMAL ||
8357 dev->vol.migr_state))
01f157d7 8358 consistent = 0;
272906ef 8359
b7941fd6 8360 if (is_resync_complete(&a->info)) {
0c046afd 8361 /* complete intialization / resync,
0556e1a2
DW		 8362 * recovery and interrupted recovery is completed in
8363 * ->set_disk
0c046afd
DW		 8364 */
8365 if (is_resyncing(dev)) {
8366 dprintf("imsm: mark resync done\n");
809da78e 8367 end_migration(dev, super, map_state);
115c3803 8368 super->updates_pending++;
484240d8 8369 a->last_checkpoint = 0;
115c3803 8370 }
b9172665
AK		 8371 } else if ((!is_resyncing(dev) && !failed) &&
8372 (imsm_reshape_blocks_arrays_changes(super) == 0)) {
0c046afd 8373 /* mark the start of the init process if nothing is failed */
b7941fd6 8374 dprintf("imsm: mark resync start\n");
1484e727 8375 if (map->map_state == IMSM_T_STATE_UNINITIALIZED)
8e59f3d8 8376 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT);
1484e727 8377 else
8e59f3d8 8378 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR);
3393c6af 8379 super->updates_pending++;
115c3803 8380 }
a862209d 8381
633b5610 8382mark_checkpoint:
5b83bacf
AK		 8383 /* skip checkpointing for general migration,
8384 * it is controlled in mdadm
8385 */
8386 if (is_gen_migration(dev))
8387 goto skip_mark_checkpoint;
8388
1e5c6983 8389 /* check if we can update curr_migr_unit from resync_start, recovery_start */
c47b0ff6 8390 blocks_per_unit = blocks_per_migr_unit(super, dev);
4f0a7acc 8391 if (blocks_per_unit) {
1e5c6983
DW		 8392 __u32 units32;
8393 __u64 units;
8394
4f0a7acc 8395 units = a->last_checkpoint / blocks_per_unit;
1e5c6983
DW		 8396 units32 = units;
8397
8398 /* check that we did not overflow 32-bits, and that
8399 * curr_migr_unit needs updating
8400 */
8401 if (units32 == units &&
bfd80a56 8402 units32 != 0 &&
1e5c6983
DW		 8403 __le32_to_cpu(dev->vol.curr_migr_unit) != units32) {
8404 dprintf("imsm: mark checkpoint (%u)\n", units32);
8405 dev->vol.curr_migr_unit = __cpu_to_le32(units32);
8406 super->updates_pending++;
8407 }
8408 }
f8f603f1 8409
5b83bacf 8410skip_mark_checkpoint:
3393c6af 8411 /* mark dirty / clean */
2432ce9b
AP		 8412 if (((dev->vol.dirty & RAIDVOL_DIRTY) && consistent) ||
8413 (!(dev->vol.dirty & RAIDVOL_DIRTY) && !consistent)) {
b7941fd6 8414 dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty");
2432ce9b
AP		 8415 if (consistent) {
8416 dev->vol.dirty = RAIDVOL_CLEAN;
8417 } else {
8418 dev->vol.dirty = RAIDVOL_DIRTY;
c2462068
PB		 8419 if (dev->rwh_policy == RWH_DISTRIBUTED ||
8420 dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
2432ce9b
AP		 8421 dev->vol.dirty |= RAIDVOL_DSRECORD_VALID;
8422 }
a862209d
DW		 8423 super->updates_pending++;
8424 }
28bce06f 8425
01f157d7 8426 return consistent;
a862209d
DW		 8427}
8428
6f50473f
TM		 8429static int imsm_disk_slot_to_ord(struct active_array *a, int slot)
8430{
8431 int inst = a->info.container_member;
8432 struct intel_super *super = a->container->sb;
8433 struct imsm_dev *dev = get_imsm_dev(super, inst);
8434 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8435
8436 if (slot > map->num_members) {
8437 pr_err("imsm: imsm_disk_slot_to_ord %d out of range 0..%d\n",
8438 slot, map->num_members - 1);
8439 return -1;
8440 }
8441
8442 if (slot < 0)
8443 return -1;
8444
8445 return get_imsm_ord_tbl_ent(dev, slot, MAP_0);
8446}
8447
8d45d196 8448static void imsm_set_disk(struct active_array *a, int n, int state)
845dea95 8449{
8d45d196
DW		 8450 int inst = a->info.container_member;
8451 struct intel_super *super = a->container->sb;
949c47a0 8452 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8453 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8d45d196 8454 struct imsm_disk *disk;
7ce05701
LD		 8455 struct mdinfo *mdi;
8456 int recovery_not_finished = 0;
0c046afd 8457 int failed;
6f50473f 8458 int ord;
0c046afd 8459 __u8 map_state;
fb12a745
TM		 8460 int rebuild_done = 0;
8461 int i;
8d45d196 8462
fb12a745 8463 ord = get_imsm_ord_tbl_ent(dev, n, MAP_X);
6f50473f 8464 if (ord < 0)
8d45d196
DW		 8465 return;
8466
4e6e574a 8467 dprintf("imsm: set_disk %d:%x\n", n, state);
b10b37b8 8468 disk = get_imsm_disk(super, ord_to_idx(ord));
8d45d196 8469
5802a811 8470 /* check for new failures */
0556e1a2 8471 if (state & DS_FAULTY) {
4c9e8c1e 8472 if (mark_failure(super, dev, disk, ord_to_idx(ord)))
0556e1a2 8473 super->updates_pending++;
8d45d196 8474 }
47ee5a45 8475
19859edc 8476 /* check if in_sync */
0556e1a2 8477 if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) {
238c0a71 8478 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
b10b37b8
DW		 8479
8480 set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord));
fb12a745 8481 rebuild_done = 1;
19859edc
DW		 8482 super->updates_pending++;
8483 }
8d45d196 8484
3b451610
AK		 8485 failed = imsm_count_failed(super, dev, MAP_0);
8486 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
5802a811 8487
0c046afd 8488 /* check if recovery complete, newly degraded, or failed */
94002678
AK		 8489 dprintf("imsm: Detected transition to state ");
8490 switch (map_state) {
8491 case IMSM_T_STATE_NORMAL: /* transition to normal state */
8492 dprintf("normal: ");
8493 if (is_rebuilding(dev)) {
1ade5cc1 8494 dprintf_cont("while rebuilding");
7ce05701
LD		 8495 /* check if recovery is really finished */
8496 for (mdi = a->info.devs; mdi ; mdi = mdi->next)
8497 if (mdi->recovery_start != MaxSector) {
8498 recovery_not_finished = 1;
8499 break;
8500 }
8501 if (recovery_not_finished) {
1ade5cc1
N		 8502 dprintf_cont("\n");
8503 dprintf("Rebuild has not finished yet, state not changed");
7ce05701
LD		 8504 if (a->last_checkpoint < mdi->recovery_start) {
8505 a->last_checkpoint = mdi->recovery_start;
8506 super->updates_pending++;
8507 }
8508 break;
8509 }
94002678 8510 end_migration(dev, super, map_state);
238c0a71 8511 map = get_imsm_map(dev, MAP_0);
94002678
AK		 8512 map->failed_disk_num = ~0;
8513 super->updates_pending++;
8514 a->last_checkpoint = 0;
8515 break;
8516 }
8517 if (is_gen_migration(dev)) {
1ade5cc1 8518 dprintf_cont("while general migration");
bf2f0071 8519 if (a->last_checkpoint >= a->info.component_size)
809da78e 8520 end_migration(dev, super, map_state);
94002678
AK		 8521 else
8522 map->map_state = map_state;
238c0a71 8523 map = get_imsm_map(dev, MAP_0);
28bce06f 8524 map->failed_disk_num = ~0;
94002678 8525 super->updates_pending++;
bf2f0071 8526 break;
94002678
AK		 8527 }
8528 break;
8529 case IMSM_T_STATE_DEGRADED: /* transition to degraded state */
1ade5cc1 8530 dprintf_cont("degraded: ");
089f9d79 8531 if (map->map_state != map_state && !dev->vol.migr_state) {
1ade5cc1 8532 dprintf_cont("mark degraded");
94002678
AK		 8533 map->map_state = map_state;
8534 super->updates_pending++;
8535 a->last_checkpoint = 0;
8536 break;
8537 }
8538 if (is_rebuilding(dev)) {
1ade5cc1 8539 dprintf_cont("while rebuilding.");
94002678 8540 if (map->map_state != map_state) {
1ade5cc1 8541 dprintf_cont(" Map state change");
94002678
AK		 8542 end_migration(dev, super, map_state);
8543 super->updates_pending++;
fb12a745
TM		 8544 } else if (!rebuild_done) {
8545 break;
8546 }
8547
8548 /* check if recovery is really finished */
8549 for (mdi = a->info.devs; mdi ; mdi = mdi->next)
8550 if (mdi->recovery_start != MaxSector) {
8551 recovery_not_finished = 1;
8552 break;
8553 }
8554 if (recovery_not_finished) {
8555 dprintf_cont("\n");
8556 dprintf("Rebuild has not finished yet, state not changed");
8557 if (a->last_checkpoint < mdi->recovery_start) {
8558 a->last_checkpoint =
8559 mdi->recovery_start;
8560 super->updates_pending++;
8561 }
8562 break;
94002678 8563 }
fb12a745
TM		 8564
8565 dprintf_cont(" Rebuild done, still degraded");
8566 dev->vol.migr_state = 0;
8567 set_migr_type(dev, 0);
8568 dev->vol.curr_migr_unit = 0;
8569
8570 for (i = 0; i < map->num_members; i++) {
8571 int idx = get_imsm_ord_tbl_ent(dev, i, MAP_0);
8572
8573 if (idx & IMSM_ORD_REBUILD)
8574 map->failed_disk_num = i;
8575 }
8576 super->updates_pending++;
94002678
AK		 8577 break;
8578 }
8579 if (is_gen_migration(dev)) {
1ade5cc1 8580 dprintf_cont("while general migration");
bf2f0071 8581 if (a->last_checkpoint >= a->info.component_size)
809da78e 8582 end_migration(dev, super, map_state);
94002678
AK		 8583 else {
8584 map->map_state = map_state;
3b451610 8585 manage_second_map(super, dev);
94002678
AK		 8586 }
8587 super->updates_pending++;
bf2f0071 8588 break;
28bce06f 8589 }
6ce1fbf1 8590 if (is_initializing(dev)) {
1ade5cc1 8591 dprintf_cont("while initialization.");
6ce1fbf1
AK		 8592 map->map_state = map_state;
8593 super->updates_pending++;
8594 break;
8595 }
94002678
AK		 8596 break;
8597 case IMSM_T_STATE_FAILED: /* transition to failed state */
1ade5cc1 8598 dprintf_cont("failed: ");
94002678 8599 if (is_gen_migration(dev)) {
1ade5cc1 8600 dprintf_cont("while general migration");
94002678
AK		 8601 map->map_state = map_state;
8602 super->updates_pending++;
8603 break;
8604 }
8605 if (map->map_state != map_state) {
1ade5cc1 8606 dprintf_cont("mark failed");
94002678
AK		 8607 end_migration(dev, super, map_state);
8608 super->updates_pending++;
8609 a->last_checkpoint = 0;
8610 break;
8611 }
8612 break;
8613 default:
1ade5cc1 8614 dprintf_cont("state %i\n", map_state);
5802a811 8615 }
1ade5cc1 8616 dprintf_cont("\n");
845dea95
NB		 8617}
8618
f796af5d 8619static int store_imsm_mpb(int fd, struct imsm_super *mpb)
c2a1e7da 8620{
f796af5d 8621 void *buf = mpb;
c2a1e7da
DW		 8622 __u32 mpb_size = __le32_to_cpu(mpb->mpb_size);
8623 unsigned long long dsize;
8624 unsigned long long sectors;
f36a9ecd 8625 unsigned int sector_size;
c2a1e7da 8626
f36a9ecd 8627 get_dev_sector_size(fd, NULL, &sector_size);
c2a1e7da
DW		 8628 get_dev_size(fd, NULL, &dsize);
8629
f36a9ecd 8630 if (mpb_size > sector_size) {
272f648f 8631 /* -1 to account for anchor */
f36a9ecd 8632 sectors = mpb_sectors(mpb, sector_size) - 1;
c2a1e7da 8633
272f648f 8634 /* write the extended mpb to the sectors preceeding the anchor */
f36a9ecd
PB		 8635 if (lseek64(fd, dsize - (sector_size * (2 + sectors)),
8636 SEEK_SET) < 0)
272f648f 8637 return 1;
c2a1e7da 8638
f36a9ecd
PB		 8639 if ((unsigned long long)write(fd, buf + sector_size,
8640 sector_size * sectors) != sector_size * sectors)
272f648f
DW		 8641 return 1;
8642 }
c2a1e7da 8643
272f648f 8644 /* first block is stored on second to last sector of the disk */
f36a9ecd 8645 if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0)
c2a1e7da
DW		 8646 return 1;
8647
466070ad 8648 if ((unsigned int)write(fd, buf, sector_size) != sector_size)
c2a1e7da
DW		 8649 return 1;
8650
c2a1e7da
DW		 8651 return 0;
8652}
8653
2e735d19 8654static void imsm_sync_metadata(struct supertype *container)
845dea95 8655{
2e735d19 8656 struct intel_super *super = container->sb;
c2a1e7da 8657
1a64be56 8658 dprintf("sync metadata: %d\n", super->updates_pending);
c2a1e7da
DW		 8659 if (!super->updates_pending)
8660 return;
8661
36988a3d 8662 write_super_imsm(container, 0);
c2a1e7da
DW		 8663
8664 super->updates_pending = 0;
845dea95
NB		 8665}
8666
272906ef
DW		 8667static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a)
8668{
8669 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8670 int i = get_imsm_disk_idx(dev, idx, MAP_X);
272906ef
DW		 8671 struct dl *dl;
8672
8673 for (dl = super->disks; dl; dl = dl->next)
8674 if (dl->index == i)
8675 break;
8676
25ed7e59 8677 if (dl && is_failed(&dl->disk))
272906ef
DW		 8678 dl = NULL;
8679
8680 if (dl)
1ade5cc1 8681 dprintf("found %x:%x\n", dl->major, dl->minor);
272906ef
DW		 8682
8683 return dl;
8684}
8685
a20d2ba5 8686static struct dl *imsm_add_spare(struct intel_super *super, int slot,
8ba77d32
AK		 8687 struct active_array *a, int activate_new,
8688 struct mdinfo *additional_test_list)
272906ef
DW		 8689{
8690 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8691 int idx = get_imsm_disk_idx(dev, slot, MAP_X);
a20d2ba5
DW		 8692 struct imsm_super *mpb = super->anchor;
8693 struct imsm_map *map;
272906ef
DW		 8694 unsigned long long pos;
8695 struct mdinfo *d;
8696 struct extent *ex;
a20d2ba5 8697 int i, j;
272906ef 8698 int found;
569cc43f
DW		 8699 __u32 array_start = 0;
8700 __u32 array_end = 0;
272906ef 8701 struct dl *dl;
6c932028 8702 struct mdinfo *test_list;
272906ef
DW		 8703
8704 for (dl = super->disks; dl; dl = dl->next) {
8705 /* If in this array, skip */
8706 for (d = a->info.devs ; d ; d = d->next)
e553d2a4
DW		 8707 if (d->state_fd >= 0 &&
8708 d->disk.major == dl->major &&
272906ef 8709 d->disk.minor == dl->minor) {
8ba77d32
AK		 8710 dprintf("%x:%x already in array\n",
8711 dl->major, dl->minor);
272906ef
DW		 8712 break;
8713 }
8714 if (d)
8715 continue;
6c932028
AK		 8716 test_list = additional_test_list;
8717 while (test_list) {
8718 if (test_list->disk.major == dl->major &&
8719 test_list->disk.minor == dl->minor) {
8ba77d32
AK		 8720 dprintf("%x:%x already in additional test list\n",
8721 dl->major, dl->minor);
8722 break;
8723 }
6c932028 8724 test_list = test_list->next;
8ba77d32 8725 }
6c932028 8726 if (test_list)
8ba77d32 8727 continue;
272906ef 8728
e553d2a4 8729 /* skip in use or failed drives */
25ed7e59 8730 if (is_failed(&dl->disk) || idx == dl->index ||
df474657
DW		 8731 dl->index == -2) {
8732 dprintf("%x:%x status (failed: %d index: %d)\n",
25ed7e59 8733 dl->major, dl->minor, is_failed(&dl->disk), idx);
9a1608e5
DW		 8734 continue;
8735 }
8736
a20d2ba5
DW		 8737 /* skip pure spares when we are looking for partially
8738 * assimilated drives
8739 */
8740 if (dl->index == -1 && !activate_new)
8741 continue;
8742
f2cc4f7d
AO		 8743 if (!drive_validate_sector_size(super, dl))
8744 continue;
8745
272906ef 8746 /* Does this unused device have the requisite free space?
a20d2ba5 8747 * It needs to be able to cover all member volumes
272906ef
DW		 8748 */
8749 ex = get_extents(super, dl);
8750 if (!ex) {
8751 dprintf("cannot get extents\n");
8752 continue;
8753 }
a20d2ba5
DW		 8754 for (i = 0; i < mpb->num_raid_devs; i++) {
8755 dev = get_imsm_dev(super, i);
238c0a71 8756 map = get_imsm_map(dev, MAP_0);
272906ef 8757
a20d2ba5
DW		 8758 /* check if this disk is already a member of
8759 * this array
272906ef 8760 */
620b1713 8761 if (get_imsm_disk_slot(map, dl->index) >= 0)
a20d2ba5
DW		 8762 continue;
8763
8764 found = 0;
8765 j = 0;
8766 pos = 0;
5551b113 8767 array_start = pba_of_lba0(map);
329c8278 8768 array_end = array_start +
44490938 8769 per_dev_array_size(map) - 1;
a20d2ba5
DW		 8770
8771 do {
8772 /* check that we can start at pba_of_lba0 with
44490938 8773 * num_data_stripes*blocks_per_stripe of space
a20d2ba5 8774 */
329c8278 8775 if (array_start >= pos && array_end < ex[j].start) {
a20d2ba5
DW		 8776 found = 1;
8777 break;
8778 }
8779 pos = ex[j].start + ex[j].size;
8780 j++;
8781 } while (ex[j-1].size);
8782
8783 if (!found)
272906ef 8784 break;
a20d2ba5 8785 }
272906ef
DW		 8786
8787 free(ex);
a20d2ba5 8788 if (i < mpb->num_raid_devs) {
329c8278
DW		 8789 dprintf("%x:%x does not have %u to %u available\n",
8790 dl->major, dl->minor, array_start, array_end);
272906ef
DW		 8791 /* No room */
8792 continue;
a20d2ba5
DW		 8793 }
8794 return dl;
272906ef
DW		 8795 }
8796
8797 return dl;
8798}
8799
95d07a2c
LM		 8800static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed)
8801{
8802 struct imsm_dev *dev2;
8803 struct imsm_map *map;
8804 struct dl *idisk;
8805 int slot;
8806 int idx;
8807 __u8 state;
8808
8809 dev2 = get_imsm_dev(cont->sb, dev_idx);
8810 if (dev2) {
238c0a71 8811 state = imsm_check_degraded(cont->sb, dev2, failed, MAP_0);
95d07a2c 8812 if (state == IMSM_T_STATE_FAILED) {
238c0a71 8813 map = get_imsm_map(dev2, MAP_0);
95d07a2c
LM		 8814 if (!map)
8815 return 1;
8816 for (slot = 0; slot < map->num_members; slot++) {
8817 /*
8818 * Check if failed disks are deleted from intel
8819 * disk list or are marked to be deleted
8820 */
238c0a71 8821 idx = get_imsm_disk_idx(dev2, slot, MAP_X);
95d07a2c
LM		 8822 idisk = get_imsm_dl_disk(cont->sb, idx);
8823 /*
8824 * Do not rebuild the array if failed disks
8825 * from failed sub-array are not removed from
8826 * container.
8827 */
8828 if (idisk &&
8829 is_failed(&idisk->disk) &&
8830 (idisk->action != DISK_REMOVE))
8831 return 0;
8832 }
8833 }
8834 }
8835 return 1;
8836}
8837
88758e9d
DW		 8838static struct mdinfo *imsm_activate_spare(struct active_array *a,
8839 struct metadata_update **updates)
8840{
8841 /**
d23fe947
DW		 8842 * Find a device with unused free space and use it to replace a
8843 * failed/vacant region in an array. We replace failed regions one a
8844 * array at a time. The result is that a new spare disk will be added
8845 * to the first failed array and after the monitor has finished
8846 * propagating failures the remainder will be consumed.
88758e9d 8847 *
d23fe947
DW		 8848 * FIXME add a capability for mdmon to request spares from another
8849 * container.
88758e9d
DW		 8850 */
8851
8852 struct intel_super *super = a->container->sb;
88758e9d 8853 int inst = a->info.container_member;
949c47a0 8854 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8855 struct imsm_map *map = get_imsm_map(dev, MAP_0);
88758e9d
DW		 8856 int failed = a->info.array.raid_disks;
8857 struct mdinfo *rv = NULL;
8858 struct mdinfo *d;
8859 struct mdinfo *di;
8860 struct metadata_update *mu;
8861 struct dl *dl;
8862 struct imsm_update_activate_spare *u;
8863 int num_spares = 0;
8864 int i;
95d07a2c 8865 int allowed;
88758e9d
DW		 8866
8867 for (d = a->info.devs ; d ; d = d->next) {
8868 if ((d->curr_state & DS_FAULTY) &&
8869 d->state_fd >= 0)
8870 /* wait for Removal to happen */
8871 return NULL;
8872 if (d->state_fd >= 0)
8873 failed--;
8874 }
8875
8876 dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n",
8877 inst, failed, a->info.array.raid_disks, a->info.array.level);
1af97990 8878
e2962bfc
AK		 8879 if (imsm_reshape_blocks_arrays_changes(super))
8880 return NULL;
1af97990 8881
fc8ca064
AK		 8882 /* Cannot activate another spare if rebuild is in progress already
8883 */
8884 if (is_rebuilding(dev)) {
7a862a02 8885 dprintf("imsm: No spare activation allowed. Rebuild in progress already.\n");
fc8ca064
AK		 8886 return NULL;
8887 }
8888
89c67882
AK		 8889 if (a->info.array.level == 4)
8890 /* No repair for takeovered array
8891 * imsm doesn't support raid4
8892 */
8893 return NULL;
8894
3b451610
AK		 8895 if (imsm_check_degraded(super, dev, failed, MAP_0) !=
8896 IMSM_T_STATE_DEGRADED)
88758e9d
DW		 8897 return NULL;
8898
83ca7d45
AP		 8899 if (get_imsm_map(dev, MAP_0)->map_state == IMSM_T_STATE_UNINITIALIZED) {
8900 dprintf("imsm: No spare activation allowed. Volume is not initialized.\n");
8901 return NULL;
8902 }
8903
95d07a2c
LM		 8904 /*
8905 * If there are any failed disks check state of the other volume.
8906 * Block rebuild if the another one is failed until failed disks
8907 * are removed from container.
8908 */
8909 if (failed) {
7a862a02 8910 dprintf("found failed disks in %.*s, check if there anotherfailed sub-array.\n",
c4acd1e5 8911 MAX_RAID_SERIAL_LEN, dev->volume);
95d07a2c
LM		 8912 /* check if states of the other volumes allow for rebuild */
8913 for (i = 0; i < super->anchor->num_raid_devs; i++) {
8914 if (i != inst) {
8915 allowed = imsm_rebuild_allowed(a->container,
8916 i, failed);
8917 if (!allowed)
8918 return NULL;
8919 }
8920 }
8921 }
8922
88758e9d 8923 /* For each slot, if it is not working, find a spare */
88758e9d
DW		 8924 for (i = 0; i < a->info.array.raid_disks; i++) {
8925 for (d = a->info.devs ; d ; d = d->next)
8926 if (d->disk.raid_disk == i)
8927 break;
8928 dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0);
8929 if (d && (d->state_fd >= 0))
8930 continue;
8931
272906ef 8932 /*
a20d2ba5
DW		 8933 * OK, this device needs recovery. Try to re-add the
8934 * previous occupant of this slot, if this fails see if
8935 * we can continue the assimilation of a spare that was
8936 * partially assimilated, finally try to activate a new
8937 * spare.
272906ef
DW		 8938 */
8939 dl = imsm_readd(super, i, a);
8940 if (!dl)
b303fe21 8941 dl = imsm_add_spare(super, i, a, 0, rv);
a20d2ba5 8942 if (!dl)
b303fe21 8943 dl = imsm_add_spare(super, i, a, 1, rv);
272906ef
DW		 8944 if (!dl)
8945 continue;
1011e834 8946
272906ef 8947 /* found a usable disk with enough space */
503975b9 8948 di = xcalloc(1, sizeof(*di));
272906ef
DW		 8949
8950 /* dl->index will be -1 in the case we are activating a
8951 * pristine spare. imsm_process_update() will create a
8952 * new index in this case. Once a disk is found to be
8953 * failed in all member arrays it is kicked from the
8954 * metadata
8955 */
8956 di->disk.number = dl->index;
d23fe947 8957
272906ef
DW		 8958 /* (ab)use di->devs to store a pointer to the device
8959 * we chose
8960 */
8961 di->devs = (struct mdinfo *) dl;
8962
8963 di->disk.raid_disk = i;
8964 di->disk.major = dl->major;
8965 di->disk.minor = dl->minor;
8966 di->disk.state = 0;
d23534e4 8967 di->recovery_start = 0;
5551b113 8968 di->data_offset = pba_of_lba0(map);
272906ef
DW		 8969 di->component_size = a->info.component_size;
8970 di->container_member = inst;
5e46202e 8971 di->bb.supported = 1;
2c8890e9 8972 if (a->info.consistency_policy == CONSISTENCY_POLICY_PPL) {
2432ce9b 8973 di->ppl_sector = get_ppl_sector(super, inst);
c2462068 8974 di->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9;
2432ce9b 8975 }
148acb7b 8976 super->random = random32();
272906ef
DW		 8977 di->next = rv;
8978 rv = di;
8979 num_spares++;
8980 dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor,
8981 i, di->data_offset);
88758e9d
DW		 8982 }
8983
8984 if (!rv)
8985 /* No spares found */
8986 return rv;
8987 /* Now 'rv' has a list of devices to return.
8988 * Create a metadata_update record to update the
8989 * disk_ord_tbl for the array
8990 */
503975b9 8991 mu = xmalloc(sizeof(*mu));
1011e834 8992 mu->buf = xcalloc(num_spares,
503975b9 8993 sizeof(struct imsm_update_activate_spare));
88758e9d 8994 mu->space = NULL;
cb23f1f4 8995 mu->space_list = NULL;
88758e9d
DW		 8996 mu->len = sizeof(struct imsm_update_activate_spare) * num_spares;
8997 mu->next = *updates;
8998 u = (struct imsm_update_activate_spare *) mu->buf;
8999
9000 for (di = rv ; di ; di = di->next) {
9001 u->type = update_activate_spare;
d23fe947
DW		 9002 u->dl = (struct dl *) di->devs;
9003 di->devs = NULL;
88758e9d
DW		 9004 u->slot = di->disk.raid_disk;
9005 u->array = inst;
9006 u->next = u + 1;
9007 u++;
9008 }
9009 (u-1)->next = NULL;
9010 *updates = mu;
9011
9012 return rv;
9013}
9014
54c2c1ea 9015static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u)
8273f55e 9016{
54c2c1ea 9017 struct imsm_dev *dev = get_imsm_dev(super, idx);
238c0a71
AK		 9018 struct imsm_map *map = get_imsm_map(dev, MAP_0);
9019 struct imsm_map *new_map = get_imsm_map(&u->dev, MAP_0);
54c2c1ea
DW		 9020 struct disk_info *inf = get_disk_info(u);
9021 struct imsm_disk *disk;
8273f55e
DW		 9022 int i;
9023 int j;
8273f55e 9024
54c2c1ea 9025 for (i = 0; i < map->num_members; i++) {
238c0a71 9026 disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, MAP_X));
54c2c1ea
DW		 9027 for (j = 0; j < new_map->num_members; j++)
9028 if (serialcmp(disk->serial, inf[j].serial) == 0)
8273f55e
DW		 9029 return 1;
9030 }
9031
9032 return 0;
9033}
9034
1a64be56
LM		 9035static struct dl *get_disk_super(struct intel_super *super, int major, int minor)
9036{
594dc1b8
JS		 9037 struct dl *dl;
9038
1a64be56 9039 for (dl = super->disks; dl; dl = dl->next)
089f9d79 9040 if (dl->major == major && dl->minor == minor)
1a64be56
LM		 9041 return dl;
9042 return NULL;
9043}
9044
9045static int remove_disk_super(struct intel_super *super, int major, int minor)
9046{
594dc1b8 9047 struct dl *prev;
1a64be56
LM		 9048 struct dl *dl;
9049
9050 prev = NULL;
9051 for (dl = super->disks; dl; dl = dl->next) {
089f9d79 9052 if (dl->major == major && dl->minor == minor) {
1a64be56
LM		 9053 /* remove */
9054 if (prev)
9055 prev->next = dl->next;
9056 else
9057 super->disks = dl->next;
9058 dl->next = NULL;
9059 __free_imsm_disk(dl);
1ade5cc1 9060 dprintf("removed %x:%x\n", major, minor);
1a64be56
LM		 9061 break;
9062 }
9063 prev = dl;
9064 }
9065 return 0;
9066}
9067
f21e18ca 9068static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index);
ae6aad82 9069
1a64be56
LM		 9070static int add_remove_disk_update(struct intel_super *super)
9071{
9072 int check_degraded = 0;
594dc1b8
JS		 9073 struct dl *disk;
9074
1a64be56
LM		 9075 /* add/remove some spares to/from the metadata/contrainer */
9076 while (super->disk_mgmt_list) {
9077 struct dl *disk_cfg;
9078
9079 disk_cfg = super->disk_mgmt_list;
9080 super->disk_mgmt_list = disk_cfg->next;
9081 disk_cfg->next = NULL;
9082
9083 if (disk_cfg->action == DISK_ADD) {
9084 disk_cfg->next = super->disks;
9085 super->disks = disk_cfg;
9086 check_degraded = 1;
1ade5cc1
N		 9087 dprintf("added %x:%x\n",
9088 disk_cfg->major, disk_cfg->minor);
1a64be56
LM		 9089 } else if (disk_cfg->action == DISK_REMOVE) {
9090 dprintf("Disk remove action processed: %x.%x\n",
9091 disk_cfg->major, disk_cfg->minor);
9092 disk = get_disk_super(super,
9093 disk_cfg->major,
9094 disk_cfg->minor);
9095 if (disk) {
9096 /* store action status */
9097 disk->action = DISK_REMOVE;
9098 /* remove spare disks only */
9099 if (disk->index == -1) {
9100 remove_disk_super(super,
9101 disk_cfg->major,
9102 disk_cfg->minor);
9103 }
9104 }
9105 /* release allocate disk structure */
9106 __free_imsm_disk(disk_cfg);
9107 }
9108 }
9109 return check_degraded;
9110}
9111
a29911da
PC		 9112static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u,
9113 struct intel_super *super,
9114 void ***space_list)
9115{
9116 struct intel_dev *id;
9117 void **tofree = NULL;
9118 int ret_val = 0;
9119
1ade5cc1 9120 dprintf("(enter)\n");
089f9d79 9121 if (u->subdev < 0 || u->subdev > 1) {
a29911da
PC		 9122 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
9123 return ret_val;
9124 }
089f9d79 9125 if (space_list == NULL || *space_list == NULL) {
a29911da
PC		 9126 dprintf("imsm: Error: Memory is not allocated\n");
9127 return ret_val;
9128 }
9129
9130 for (id = super->devlist ; id; id = id->next) {
9131 if (id->index == (unsigned)u->subdev) {
9132 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
9133 struct imsm_map *map;
9134 struct imsm_dev *new_dev =
9135 (struct imsm_dev *)*space_list;
238c0a71 9136 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
a29911da
PC		 9137 int to_state;
9138 struct dl *new_disk;
9139
9140 if (new_dev == NULL)
9141 return ret_val;
9142 *space_list = **space_list;
9143 memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0));
238c0a71 9144 map = get_imsm_map(new_dev, MAP_0);
a29911da
PC		 9145 if (migr_map) {
9146 dprintf("imsm: Error: migration in progress");
9147 return ret_val;
9148 }
9149
9150 to_state = map->map_state;
9151 if ((u->new_level == 5) && (map->raid_level == 0)) {
9152 map->num_members++;
9153 /* this should not happen */
9154 if (u->new_disks[0] < 0) {
9155 map->failed_disk_num =
9156 map->num_members - 1;
9157 to_state = IMSM_T_STATE_DEGRADED;
9158 } else
9159 to_state = IMSM_T_STATE_NORMAL;
9160 }
8e59f3d8 9161 migrate(new_dev, super, to_state, MIGR_GEN_MIGR);
a29911da
PC		 9162 if (u->new_level > -1)
9163 map->raid_level = u->new_level;
238c0a71 9164 migr_map = get_imsm_map(new_dev, MAP_1);
a29911da
PC		 9165 if ((u->new_level == 5) &&
9166 (migr_map->raid_level == 0)) {
9167 int ord = map->num_members - 1;
9168 migr_map->num_members--;
9169 if (u->new_disks[0] < 0)
9170 ord |= IMSM_ORD_REBUILD;
9171 set_imsm_ord_tbl_ent(map,
9172 map->num_members - 1,
9173 ord);
9174 }
9175 id->dev = new_dev;
9176 tofree = (void **)dev;
9177
4bba0439
PC		 9178 /* update chunk size
9179 */
06fb291a
PB		 9180 if (u->new_chunksize > 0) {
9181 unsigned long long num_data_stripes;
9529d343
MD		 9182 struct imsm_map *dest_map =
9183 get_imsm_map(dev, MAP_0);
06fb291a 9184 int used_disks =
9529d343 9185 imsm_num_data_members(dest_map);
06fb291a
PB		 9186
9187 if (used_disks == 0)
9188 return ret_val;
9189
4bba0439
PC		 9190 map->blocks_per_strip =
9191 __cpu_to_le16(u->new_chunksize * 2);
06fb291a 9192 num_data_stripes =
fcc2c9da 9193 imsm_dev_size(dev) / used_disks;
06fb291a
PB		 9194 num_data_stripes /= map->blocks_per_strip;
9195 num_data_stripes /= map->num_domains;
9196 set_num_data_stripes(map, num_data_stripes);
9197 }
4bba0439 9198
44490938
MD		 9199 /* ensure blocks_per_member has valid value
9200 */
9201 set_blocks_per_member(map,
9202 per_dev_array_size(map) +
9203 NUM_BLOCKS_DIRTY_STRIPE_REGION);
9204
a29911da
PC		 9205 /* add disk
9206 */
089f9d79
JS		 9207 if (u->new_level != 5 || migr_map->raid_level != 0 ||
9208 migr_map->raid_level == map->raid_level)
a29911da
PC		 9209 goto skip_disk_add;
9210
9211 if (u->new_disks[0] >= 0) {
9212 /* use passes spare
9213 */
9214 new_disk = get_disk_super(super,
9215 major(u->new_disks[0]),
9216 minor(u->new_disks[0]));
7a862a02 9217 dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n",
a29911da
PC		 9218 major(u->new_disks[0]),
9219 minor(u->new_disks[0]),
9220 new_disk, new_disk->index);
9221 if (new_disk == NULL)
9222 goto error_disk_add;
9223
9224 new_disk->index = map->num_members - 1;
9225 /* slot to fill in autolayout
9226 */
9227 new_disk->raiddisk = new_disk->index;
9228 new_disk->disk.status |= CONFIGURED_DISK;
9229 new_disk->disk.status &= ~SPARE_DISK;
9230 } else
9231 goto error_disk_add;
9232
9233skip_disk_add:
9234 *tofree = *space_list;
9235 /* calculate new size
9236 */
f3871fdc 9237 imsm_set_array_size(new_dev, -1);
a29911da
PC		 9238
9239 ret_val = 1;
9240 }
9241 }
9242
9243 if (tofree)
9244 *space_list = tofree;
9245 return ret_val;
9246
9247error_disk_add:
9248 dprintf("Error: imsm: Cannot find disk.\n");
9249 return ret_val;
9250}
9251
f3871fdc
AK		 9252static int apply_size_change_update(struct imsm_update_size_change *u,
9253 struct intel_super *super)
9254{
9255 struct intel_dev *id;
9256 int ret_val = 0;
9257
1ade5cc1 9258 dprintf("(enter)\n");
089f9d79 9259 if (u->subdev < 0 || u->subdev > 1) {
f3871fdc
AK		 9260 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
9261 return ret_val;
9262 }
9263
9264 for (id = super->devlist ; id; id = id->next) {
9265 if (id->index == (unsigned)u->subdev) {
9266 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
9267 struct imsm_map *map = get_imsm_map(dev, MAP_0);
9529d343 9268 int used_disks = imsm_num_data_members(map);
f3871fdc 9269 unsigned long long blocks_per_member;
06fb291a 9270 unsigned long long num_data_stripes;
44490938
MD		 9271 unsigned long long new_size_per_disk;
9272
9273 if (used_disks == 0)
9274 return 0;
f3871fdc
AK		 9275
9276 /* calculate new size
9277 */
44490938
MD		 9278 new_size_per_disk = u->new_size / used_disks;
9279 blocks_per_member = new_size_per_disk +
9280 NUM_BLOCKS_DIRTY_STRIPE_REGION;
9281 num_data_stripes = new_size_per_disk /
06fb291a
PB		 9282 map->blocks_per_strip;
9283 num_data_stripes /= map->num_domains;
9284 dprintf("(size: %llu, blocks per member: %llu, num_data_stipes: %llu)\n",
44490938 9285 u->new_size, new_size_per_disk,
06fb291a 9286 num_data_stripes);
f3871fdc 9287 set_blocks_per_member(map, blocks_per_member);
06fb291a 9288 set_num_data_stripes(map, num_data_stripes);
f3871fdc
AK		 9289 imsm_set_array_size(dev, u->new_size);
9290
9291 ret_val = 1;
9292 break;
9293 }
9294 }
9295
9296 return ret_val;
9297}
9298
061d7da3 9299static int apply_update_activate_spare(struct imsm_update_activate_spare *u,
ca9de185 9300 struct intel_super *super,
061d7da3
LO		 9301 struct active_array *active_array)
9302{
9303 struct imsm_super *mpb = super->anchor;
9304 struct imsm_dev *dev = get_imsm_dev(super, u->array);
238c0a71 9305 struct imsm_map *map = get_imsm_map(dev, MAP_0);
061d7da3
LO		 9306 struct imsm_map *migr_map;
9307 struct active_array *a;
9308 struct imsm_disk *disk;
9309 __u8 to_state;
9310 struct dl *dl;
9311 unsigned int found;
9312 int failed;
5961eeec 9313 int victim;
061d7da3 9314 int i;
5961eeec 9315 int second_map_created = 0;
061d7da3 9316
5961eeec 9317 for (; u; u = u->next) {
238c0a71 9318 victim = get_imsm_disk_idx(dev, u->slot, MAP_X);
061d7da3 9319
5961eeec 9320 if (victim < 0)
9321 return 0;
061d7da3 9322
5961eeec 9323 for (dl = super->disks; dl; dl = dl->next)
9324 if (dl == u->dl)
9325 break;
061d7da3 9326
5961eeec 9327 if (!dl) {
7a862a02 9328 pr_err("error: imsm_activate_spare passed an unknown disk (index: %d)\n",
5961eeec 9329 u->dl->index);
9330 return 0;
9331 }
061d7da3 9332
5961eeec 9333 /* count failures (excluding rebuilds and the victim)
9334 * to determine map[0] state
9335 */
9336 failed = 0;
9337 for (i = 0; i < map->num_members; i++) {
9338 if (i == u->slot)
9339 continue;
9340 disk = get_imsm_disk(super,
238c0a71 9341 get_imsm_disk_idx(dev, i, MAP_X));
5961eeec 9342 if (!disk || is_failed(disk))
9343 failed++;
9344 }
061d7da3 9345
5961eeec 9346 /* adding a pristine spare, assign a new index */
9347 if (dl->index < 0) {
9348 dl->index = super->anchor->num_disks;
9349 super->anchor->num_disks++;
9350 }
9351 disk = &dl->disk;
9352 disk->status |= CONFIGURED_DISK;
9353 disk->status &= ~SPARE_DISK;
9354
9355 /* mark rebuild */
238c0a71 9356 to_state = imsm_check_degraded(super, dev, failed, MAP_0);
5961eeec 9357 if (!second_map_created) {
9358 second_map_created = 1;
9359 map->map_state = IMSM_T_STATE_DEGRADED;
9360 migrate(dev, super, to_state, MIGR_REBUILD);
9361 } else
9362 map->map_state = to_state;
238c0a71 9363 migr_map = get_imsm_map(dev, MAP_1);
5961eeec 9364 set_imsm_ord_tbl_ent(map, u->slot, dl->index);
9365 set_imsm_ord_tbl_ent(migr_map, u->slot,
9366 dl->index | IMSM_ORD_REBUILD);
9367
9368 /* update the family_num to mark a new container
9369 * generation, being careful to record the existing
9370 * family_num in orig_family_num to clean up after
9371 * earlier mdadm versions that neglected to set it.
9372 */
9373 if (mpb->orig_family_num == 0)
9374 mpb->orig_family_num = mpb->family_num;
9375 mpb->family_num += super->random;
9376
9377 /* count arrays using the victim in the metadata */
9378 found = 0;
9379 for (a = active_array; a ; a = a->next) {
9380 dev = get_imsm_dev(super, a->info.container_member);
238c0a71 9381 map = get_imsm_map(dev, MAP_0);
061d7da3 9382
5961eeec 9383 if (get_imsm_disk_slot(map, victim) >= 0)
9384 found++;
9385 }
061d7da3 9386
5961eeec 9387 /* delete the victim if it is no longer being
9388 * utilized anywhere
061d7da3 9389 */
5961eeec 9390 if (!found) {
9391 struct dl **dlp;
061d7da3 9392
5961eeec 9393 /* We know that 'manager' isn't touching anything,
9394 * so it is safe to delete
9395 */
9396 for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next)
061d7da3
LO		 9397 if ((*dlp)->index == victim)
9398 break;
5961eeec 9399
9400 /* victim may be on the missing list */
9401 if (!*dlp)
9402 for (dlp = &super->missing; *dlp;
9403 dlp = &(*dlp)->next)
9404 if ((*dlp)->index == victim)
9405 break;
9406 imsm_delete(super, dlp, victim);
9407 }
061d7da3
LO		 9408 }
9409
9410 return 1;
9411}
a29911da 9412
2e5dc010
N		 9413static int apply_reshape_container_disks_update(struct imsm_update_reshape *u,
9414 struct intel_super *super,
9415 void ***space_list)
9416{
9417 struct dl *new_disk;
9418 struct intel_dev *id;
9419 int i;
9420 int delta_disks = u->new_raid_disks - u->old_raid_disks;
ee4beede 9421 int disk_count = u->old_raid_disks;
2e5dc010
N		 9422 void **tofree = NULL;
9423 int devices_to_reshape = 1;
9424 struct imsm_super *mpb = super->anchor;
9425 int ret_val = 0;
d098291a 9426 unsigned int dev_id;
2e5dc010 9427
1ade5cc1 9428 dprintf("(enter)\n");
2e5dc010
N		 9429
9430 /* enable spares to use in array */
9431 for (i = 0; i < delta_disks; i++) {
9432 new_disk = get_disk_super(super,
9433 major(u->new_disks[i]),
9434 minor(u->new_disks[i]));
7a862a02 9435 dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n",
2e5dc010
N		 9436 major(u->new_disks[i]), minor(u->new_disks[i]),
9437 new_disk, new_disk->index);
089f9d79
JS		 9438 if (new_disk == NULL ||
9439 (new_disk->index >= 0 &&
9440 new_disk->index < u->old_raid_disks))
2e5dc010 9441 goto update_reshape_exit;
ee4beede 9442 new_disk->index = disk_count++;
2e5dc010
N		 9443 /* slot to fill in autolayout
9444 */
9445 new_disk->raiddisk = new_disk->index;
9446 new_disk->disk.status |=
9447 CONFIGURED_DISK;
9448 new_disk->disk.status &= ~SPARE_DISK;
9449 }
9450
ed7333bd
AK		 9451 dprintf("imsm: volume set mpb->num_raid_devs = %i\n",
9452 mpb->num_raid_devs);
2e5dc010
N		 9453 /* manage changes in volume
9454 */
d098291a 9455 for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) {
2e5dc010
N		 9456 void **sp = *space_list;
9457 struct imsm_dev *newdev;
9458 struct imsm_map *newmap, *oldmap;
9459
d098291a
AK		 9460 for (id = super->devlist ; id; id = id->next) {
9461 if (id->index == dev_id)
9462 break;
9463 }
9464 if (id == NULL)
9465 break;
2e5dc010
N		 9466 if (!sp)
9467 continue;
9468 *space_list = *sp;
9469 newdev = (void*)sp;
9470 /* Copy the dev, but not (all of) the map */
9471 memcpy(newdev, id->dev, sizeof(*newdev));
238c0a71
AK		 9472 oldmap = get_imsm_map(id->dev, MAP_0);
9473 newmap = get_imsm_map(newdev, MAP_0);
2e5dc010
N		 9474 /* Copy the current map */
9475 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
9476 /* update one device only
9477 */
9478 if (devices_to_reshape) {
ed7333bd
AK		 9479 dprintf("imsm: modifying subdev: %i\n",
9480 id->index);
2e5dc010
N		 9481 devices_to_reshape--;
9482 newdev->vol.migr_state = 1;
9483 newdev->vol.curr_migr_unit = 0;
ea672ee1 9484 set_migr_type(newdev, MIGR_GEN_MIGR);
2e5dc010
N		 9485 newmap->num_members = u->new_raid_disks;
9486 for (i = 0; i < delta_disks; i++) {
9487 set_imsm_ord_tbl_ent(newmap,
9488 u->old_raid_disks + i,
9489 u->old_raid_disks + i);
9490 }
9491 /* New map is correct, now need to save old map
9492 */
238c0a71 9493 newmap = get_imsm_map(newdev, MAP_1);
2e5dc010
N		 9494 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
9495
f3871fdc 9496 imsm_set_array_size(newdev, -1);
2e5dc010
N		 9497 }
9498
9499 sp = (void **)id->dev;
9500 id->dev = newdev;
9501 *sp = tofree;
9502 tofree = sp;
8e59f3d8
AK		 9503
9504 /* Clear migration record */
9505 memset(super->migr_rec, 0, sizeof(struct migr_record));
2e5dc010 9506 }
819bc634
AK		 9507 if (tofree)
9508 *space_list = tofree;
2e5dc010
N		 9509 ret_val = 1;
9510
9511update_reshape_exit:
9512
9513 return ret_val;
9514}
9515
bb025c2f 9516static int apply_takeover_update(struct imsm_update_takeover *u,
8ca6df95
KW		 9517 struct intel_super *super,
9518 void ***space_list)
bb025c2f
KW		 9519{
9520 struct imsm_dev *dev = NULL;
8ca6df95
KW		 9521 struct intel_dev *dv;
9522 struct imsm_dev *dev_new;
bb025c2f
KW		 9523 struct imsm_map *map;
9524 struct dl *dm, *du;
8ca6df95 9525 int i;
bb025c2f
KW		 9526
9527 for (dv = super->devlist; dv; dv = dv->next)
9528 if (dv->index == (unsigned int)u->subarray) {
9529 dev = dv->dev;
9530 break;
9531 }
9532
9533 if (dev == NULL)
9534 return 0;
9535
238c0a71 9536 map = get_imsm_map(dev, MAP_0);
bb025c2f
KW		 9537
9538 if (u->direction == R10_TO_R0) {
06fb291a
PB		 9539 unsigned long long num_data_stripes;
9540
9541 map->num_domains = 1;
44490938 9542 num_data_stripes = imsm_dev_size(dev) / 2;
06fb291a
PB		 9543 num_data_stripes /= map->blocks_per_strip;
9544 num_data_stripes /= map->num_domains;
9545 set_num_data_stripes(map, num_data_stripes);
9546
43d5ec18 9547 /* Number of failed disks must be half of initial disk number */
3b451610
AK		 9548 if (imsm_count_failed(super, dev, MAP_0) !=
9549 (map->num_members / 2))
43d5ec18
KW		 9550 return 0;
9551
bb025c2f
KW		 9552 /* iterate through devices to mark removed disks as spare */
9553 for (dm = super->disks; dm; dm = dm->next) {
9554 if (dm->disk.status & FAILED_DISK) {
9555 int idx = dm->index;
9556 /* update indexes on the disk list */
9557/* FIXME this loop-with-the-loop looks wrong, I'm not convinced
9558 the index values will end up being correct.... NB */
9559 for (du = super->disks; du; du = du->next)
9560 if (du->index > idx)
9561 du->index--;
9562 /* mark as spare disk */
a8619d23 9563 mark_spare(dm);
bb025c2f
KW		 9564 }
9565 }
bb025c2f
KW		 9566 /* update map */
9567 map->num_members = map->num_members / 2;
9568 map->map_state = IMSM_T_STATE_NORMAL;
9569 map->num_domains = 1;
9570 map->raid_level = 0;
9571 map->failed_disk_num = -1;
9572 }
9573
8ca6df95
KW		 9574 if (u->direction == R0_TO_R10) {
9575 void **space;
9576 /* update slots in current disk list */
9577 for (dm = super->disks; dm; dm = dm->next) {
9578 if (dm->index >= 0)
9579 dm->index *= 2;
9580 }
9581 /* create new *missing* disks */
9582 for (i = 0; i < map->num_members; i++) {
9583 space = *space_list;
9584 if (!space)
9585 continue;
9586 *space_list = *space;
9587 du = (void *)space;
9588 memcpy(du, super->disks, sizeof(*du));
8ca6df95
KW		 9589 du->fd = -1;
9590 du->minor = 0;
9591 du->major = 0;
9592 du->index = (i * 2) + 1;
9593 sprintf((char *)du->disk.serial,
9594 " MISSING_%d", du->index);
9595 sprintf((char *)du->serial,
9596 "MISSING_%d", du->index);
9597 du->next = super->missing;
9598 super->missing = du;
9599 }
9600 /* create new dev and map */
9601 space = *space_list;
9602 if (!space)
9603 return 0;
9604 *space_list = *space;
9605 dev_new = (void *)space;
9606 memcpy(dev_new, dev, sizeof(*dev));
9607 /* update new map */
238c0a71 9608 map = get_imsm_map(dev_new, MAP_0);
8ca6df95 9609 map->num_members = map->num_members * 2;
1a2487c2 9610 map->map_state = IMSM_T_STATE_DEGRADED;
8ca6df95
KW		 9611 map->num_domains = 2;
9612 map->raid_level = 1;
9613 /* replace dev<->dev_new */
9614 dv->dev = dev_new;
9615 }
bb025c2f
KW		 9616 /* update disk order table */
9617 for (du = super->disks; du; du = du->next)
9618 if (du->index >= 0)
9619 set_imsm_ord_tbl_ent(map, du->index, du->index);
8ca6df95 9620 for (du = super->missing; du; du = du->next)
1a2487c2
KW		 9621 if (du->index >= 0) {
9622 set_imsm_ord_tbl_ent(map, du->index, du->index);
4c9e8c1e 9623 mark_missing(super, dv->dev, &du->disk, du->index);
1a2487c2 9624 }
bb025c2f
KW		 9625
9626 return 1;
9627}
9628
e8319a19
DW		 9629static void imsm_process_update(struct supertype *st,
9630 struct metadata_update *update)
9631{
9632 /**
9633 * crack open the metadata_update envelope to find the update record
9634 * update can be one of:
d195167d
AK		 9635 * update_reshape_container_disks - all the arrays in the container
9636 * are being reshaped to have more devices. We need to mark
9637 * the arrays for general migration and convert selected spares
9638 * into active devices.
9639 * update_activate_spare - a spare device has replaced a failed
1011e834
N		 9640 * device in an array, update the disk_ord_tbl. If this disk is
9641 * present in all member arrays then also clear the SPARE_DISK
9642 * flag
d195167d
AK		 9643 * update_create_array
9644 * update_kill_array
9645 * update_rename_array
9646 * update_add_remove_disk
e8319a19
DW		 9647 */
9648 struct intel_super *super = st->sb;
4d7b1503 9649 struct imsm_super *mpb;
e8319a19
DW		 9650 enum imsm_update_type type = *(enum imsm_update_type *) update->buf;
9651
4d7b1503
DW		 9652 /* update requires a larger buf but the allocation failed */
9653 if (super->next_len && !super->next_buf) {
9654 super->next_len = 0;
9655 return;
9656 }
9657
9658 if (super->next_buf) {
9659 memcpy(super->next_buf, super->buf, super->len);
9660 free(super->buf);
9661 super->len = super->next_len;
9662 super->buf = super->next_buf;
9663
9664 super->next_len = 0;
9665 super->next_buf = NULL;
9666 }
9667
9668 mpb = super->anchor;
9669
e8319a19 9670 switch (type) {
0ec5d470
AK		 9671 case update_general_migration_checkpoint: {
9672 struct intel_dev *id;
9673 struct imsm_update_general_migration_checkpoint *u =
9674 (void *)update->buf;
9675
1ade5cc1 9676 dprintf("called for update_general_migration_checkpoint\n");
0ec5d470
AK		 9677
9678 /* find device under general migration */
9679 for (id = super->devlist ; id; id = id->next) {
9680 if (is_gen_migration(id->dev)) {
9681 id->dev->vol.curr_migr_unit =
9682 __cpu_to_le32(u->curr_migr_unit);
9683 super->updates_pending++;
9684 }
9685 }
9686 break;
9687 }
bb025c2f
KW		 9688 case update_takeover: {
9689 struct imsm_update_takeover *u = (void *)update->buf;
1a2487c2
KW		 9690 if (apply_takeover_update(u, super, &update->space_list)) {
9691 imsm_update_version_info(super);
bb025c2f 9692 super->updates_pending++;
1a2487c2 9693 }
bb025c2f
KW		 9694 break;
9695 }
9696
78b10e66 9697 case update_reshape_container_disks: {
d195167d 9698 struct imsm_update_reshape *u = (void *)update->buf;
2e5dc010
N		 9699 if (apply_reshape_container_disks_update(
9700 u, super, &update->space_list))
9701 super->updates_pending++;
78b10e66
N		 9702 break;
9703 }
48c5303a 9704 case update_reshape_migration: {
a29911da
PC		 9705 struct imsm_update_reshape_migration *u = (void *)update->buf;
9706 if (apply_reshape_migration_update(
9707 u, super, &update->space_list))
9708 super->updates_pending++;
48c5303a
PC		 9709 break;
9710 }
f3871fdc
AK		 9711 case update_size_change: {
9712 struct imsm_update_size_change *u = (void *)update->buf;
9713 if (apply_size_change_update(u, super))
9714 super->updates_pending++;
9715 break;
9716 }
e8319a19 9717 case update_activate_spare: {
1011e834 9718 struct imsm_update_activate_spare *u = (void *) update->buf;
061d7da3
LO		 9719 if (apply_update_activate_spare(u, super, st->arrays))
9720 super->updates_pending++;
8273f55e
DW		 9721 break;
9722 }
9723 case update_create_array: {
9724 /* someone wants to create a new array, we need to be aware of
9725 * a few races/collisions:
9726 * 1/ 'Create' called by two separate instances of mdadm
9727 * 2/ 'Create' versus 'activate_spare': mdadm has chosen
9728 * devices that have since been assimilated via
9729 * activate_spare.
9730 * In the event this update can not be carried out mdadm will
9731 * (FIX ME) notice that its update did not take hold.
9732 */
9733 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 9734 struct intel_dev *dv;
8273f55e
DW		 9735 struct imsm_dev *dev;
9736 struct imsm_map *map, *new_map;
9737 unsigned long long start, end;
9738 unsigned long long new_start, new_end;
9739 int i;
54c2c1ea
DW		 9740 struct disk_info *inf;
9741 struct dl *dl;
8273f55e
DW		 9742
9743 /* handle racing creates: first come first serve */
9744 if (u->dev_idx < mpb->num_raid_devs) {
1ade5cc1 9745 dprintf("subarray %d already defined\n", u->dev_idx);
ba2de7ba 9746 goto create_error;
8273f55e
DW		 9747 }
9748
9749 /* check update is next in sequence */
9750 if (u->dev_idx != mpb->num_raid_devs) {
1ade5cc1
N		 9751 dprintf("can not create array %d expected index %d\n",
9752 u->dev_idx, mpb->num_raid_devs);
ba2de7ba 9753 goto create_error;
8273f55e
DW		 9754 }
9755
238c0a71 9756 new_map = get_imsm_map(&u->dev, MAP_0);
5551b113 9757 new_start = pba_of_lba0(new_map);
44490938 9758 new_end = new_start + per_dev_array_size(new_map);
54c2c1ea 9759 inf = get_disk_info(u);
8273f55e
DW		 9760
9761 /* handle activate_spare versus create race:
9762 * check to make sure that overlapping arrays do not include
9763 * overalpping disks
9764 */
9765 for (i = 0; i < mpb->num_raid_devs; i++) {
949c47a0 9766 dev = get_imsm_dev(super, i);
238c0a71 9767 map = get_imsm_map(dev, MAP_0);
5551b113 9768 start = pba_of_lba0(map);
44490938 9769 end = start + per_dev_array_size(map);
8273f55e
DW		 9770 if ((new_start >= start && new_start <= end) ||
9771 (start >= new_start && start <= new_end))
54c2c1ea
DW		 9772 /* overlap */;
9773 else
9774 continue;
9775
9776 if (disks_overlap(super, i, u)) {
1ade5cc1 9777 dprintf("arrays overlap\n");
ba2de7ba 9778 goto create_error;
8273f55e
DW		 9779 }
9780 }
8273f55e 9781
949c47a0
DW		 9782 /* check that prepare update was successful */
9783 if (!update->space) {
1ade5cc1 9784 dprintf("prepare update failed\n");
ba2de7ba 9785 goto create_error;
949c47a0
DW		 9786 }
9787
54c2c1ea
DW		 9788 /* check that all disks are still active before committing
9789 * changes. FIXME: could we instead handle this by creating a
9790 * degraded array? That's probably not what the user expects,
9791 * so better to drop this update on the floor.
9792 */
9793 for (i = 0; i < new_map->num_members; i++) {
9794 dl = serial_to_dl(inf[i].serial, super);
9795 if (!dl) {
1ade5cc1 9796 dprintf("disk disappeared\n");
ba2de7ba 9797 goto create_error;
54c2c1ea 9798 }
949c47a0
DW		 9799 }
9800
8273f55e 9801 super->updates_pending++;
54c2c1ea
DW		 9802
9803 /* convert spares to members and fixup ord_tbl */
9804 for (i = 0; i < new_map->num_members; i++) {
9805 dl = serial_to_dl(inf[i].serial, super);
9806 if (dl->index == -1) {
9807 dl->index = mpb->num_disks;
9808 mpb->num_disks++;
9809 dl->disk.status |= CONFIGURED_DISK;
9810 dl->disk.status &= ~SPARE_DISK;
9811 }
9812 set_imsm_ord_tbl_ent(new_map, i, dl->index);
9813 }
9814
ba2de7ba
DW		 9815 dv = update->space;
9816 dev = dv->dev;
949c47a0
DW		 9817 update->space = NULL;
9818 imsm_copy_dev(dev, &u->dev);
ba2de7ba
DW		 9819 dv->index = u->dev_idx;
9820 dv->next = super->devlist;
9821 super->devlist = dv;
8273f55e 9822 mpb->num_raid_devs++;
8273f55e 9823
4d1313e9 9824 imsm_update_version_info(super);
8273f55e 9825 break;
ba2de7ba
DW		 9826 create_error:
9827 /* mdmon knows how to release update->space, but not
9828 * ((struct intel_dev *) update->space)->dev
9829 */
9830 if (update->space) {
9831 dv = update->space;
9832 free(dv->dev);
9833 }
8273f55e 9834 break;
e8319a19 9835 }
33414a01
DW		 9836 case update_kill_array: {
9837 struct imsm_update_kill_array *u = (void *) update->buf;
9838 int victim = u->dev_idx;
9839 struct active_array *a;
9840 struct intel_dev **dp;
9841 struct imsm_dev *dev;
9842
9843 /* sanity check that we are not affecting the uuid of
9844 * active arrays, or deleting an active array
9845 *
9846 * FIXME when immutable ids are available, but note that
9847 * we'll also need to fixup the invalidated/active
9848 * subarray indexes in mdstat
9849 */
9850 for (a = st->arrays; a; a = a->next)
9851 if (a->info.container_member >= victim)
9852 break;
9853 /* by definition if mdmon is running at least one array
9854 * is active in the container, so checking
9855 * mpb->num_raid_devs is just extra paranoia
9856 */
9857 dev = get_imsm_dev(super, victim);
9858 if (a || !dev || mpb->num_raid_devs == 1) {
9859 dprintf("failed to delete subarray-%d\n", victim);
9860 break;
9861 }
9862
9863 for (dp = &super->devlist; *dp;)
f21e18ca 9864 if ((*dp)->index == (unsigned)super->current_vol) {
33414a01
DW		 9865 *dp = (*dp)->next;
9866 } else {
f21e18ca 9867 if ((*dp)->index > (unsigned)victim)
33414a01
DW		 9868 (*dp)->index--;
9869 dp = &(*dp)->next;
9870 }
9871 mpb->num_raid_devs--;
9872 super->updates_pending++;
9873 break;
9874 }
aa534678
DW		 9875 case update_rename_array: {
9876 struct imsm_update_rename_array *u = (void *) update->buf;
9877 char name[MAX_RAID_SERIAL_LEN+1];
9878 int target = u->dev_idx;
9879 struct active_array *a;
9880 struct imsm_dev *dev;
9881
9882 /* sanity check that we are not affecting the uuid of
9883 * an active array
9884 */
9885 snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name);
9886 name[MAX_RAID_SERIAL_LEN] = '\0';
9887 for (a = st->arrays; a; a = a->next)
9888 if (a->info.container_member == target)
9889 break;
9890 dev = get_imsm_dev(super, u->dev_idx);
9891 if (a || !dev || !check_name(super, name, 1)) {
9892 dprintf("failed to rename subarray-%d\n", target);
9893 break;
9894 }
9895
cdbe98cd 9896 snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name);
aa534678
DW		 9897 super->updates_pending++;
9898 break;
9899 }
1a64be56 9900 case update_add_remove_disk: {
43dad3d6 9901 /* we may be able to repair some arrays if disks are
095b8088 9902 * being added, check the status of add_remove_disk
1a64be56
LM		 9903 * if discs has been added.
9904 */
9905 if (add_remove_disk_update(super)) {
43dad3d6 9906 struct active_array *a;
072b727f
DW		 9907
9908 super->updates_pending++;
1a64be56 9909 for (a = st->arrays; a; a = a->next)
43dad3d6
DW		 9910 a->check_degraded = 1;
9911 }
43dad3d6 9912 break;
e8319a19 9913 }
bbab0940
TM		 9914 case update_prealloc_badblocks_mem:
9915 break;
e6e9dd3f
AP		 9916 case update_rwh_policy: {
9917 struct imsm_update_rwh_policy *u = (void *)update->buf;
9918 int target = u->dev_idx;
9919 struct imsm_dev *dev = get_imsm_dev(super, target);
9920 if (!dev) {
9921 dprintf("could not find subarray-%d\n", target);
9922 break;
9923 }
9924
9925 if (dev->rwh_policy != u->new_policy) {
9926 dev->rwh_policy = u->new_policy;
9927 super->updates_pending++;
9928 }
9929 break;
9930 }
1a64be56 9931 default:
7a862a02 9932 pr_err("error: unsuported process update type:(type: %d)\n", type);
1a64be56 9933 }
e8319a19 9934}
88758e9d 9935
bc0b9d34
PC		 9936static struct mdinfo *get_spares_for_grow(struct supertype *st);
9937
5fe6f031
N		 9938static int imsm_prepare_update(struct supertype *st,
9939 struct metadata_update *update)
8273f55e 9940{
949c47a0 9941 /**
4d7b1503
DW		 9942 * Allocate space to hold new disk entries, raid-device entries or a new
9943 * mpb if necessary. The manager synchronously waits for updates to
9944 * complete in the monitor, so new mpb buffers allocated here can be
9945 * integrated by the monitor thread without worrying about live pointers
9946 * in the manager thread.
8273f55e 9947 */
095b8088 9948 enum imsm_update_type type;
4d7b1503 9949 struct intel_super *super = st->sb;
f36a9ecd 9950 unsigned int sector_size = super->sector_size;
4d7b1503
DW		 9951 struct imsm_super *mpb = super->anchor;
9952 size_t buf_len;
9953 size_t len = 0;
949c47a0 9954
095b8088
N		 9955 if (update->len < (int)sizeof(type))
9956 return 0;
9957
9958 type = *(enum imsm_update_type *) update->buf;
9959
949c47a0 9960 switch (type) {
0ec5d470 9961 case update_general_migration_checkpoint:
095b8088
N		 9962 if (update->len < (int)sizeof(struct imsm_update_general_migration_checkpoint))
9963 return 0;
1ade5cc1 9964 dprintf("called for update_general_migration_checkpoint\n");
0ec5d470 9965 break;
abedf5fc
KW		 9966 case update_takeover: {
9967 struct imsm_update_takeover *u = (void *)update->buf;
095b8088
N		 9968 if (update->len < (int)sizeof(*u))
9969 return 0;
abedf5fc
KW		 9970 if (u->direction == R0_TO_R10) {
9971 void **tail = (void **)&update->space_list;
9972 struct imsm_dev *dev = get_imsm_dev(super, u->subarray);
238c0a71 9973 struct imsm_map *map = get_imsm_map(dev, MAP_0);
abedf5fc
KW		 9974 int num_members = map->num_members;
9975 void *space;
9976 int size, i;
abedf5fc
KW		 9977 /* allocate memory for added disks */
9978 for (i = 0; i < num_members; i++) {
9979 size = sizeof(struct dl);
503975b9 9980 space = xmalloc(size);
abedf5fc
KW		 9981 *tail = space;
9982 tail = space;
9983 *tail = NULL;
9984 }
9985 /* allocate memory for new device */
9986 size = sizeof_imsm_dev(super->devlist->dev, 0) +
9987 (num_members * sizeof(__u32));
503975b9
N		 9988 space = xmalloc(size);
9989 *tail = space;
9990 tail = space;
9991 *tail = NULL;
9992 len = disks_to_mpb_size(num_members * 2);
abedf5fc
KW		 9993 }
9994
9995 break;
9996 }
78b10e66 9997 case update_reshape_container_disks: {
d195167d
AK		 9998 /* Every raid device in the container is about to
9999 * gain some more devices, and we will enter a
10000 * reconfiguration.
10001 * So each 'imsm_map' will be bigger, and the imsm_vol
10002 * will now hold 2 of them.
10003 * Thus we need new 'struct imsm_dev' allocations sized
10004 * as sizeof_imsm_dev but with more devices in both maps.
10005 */
10006 struct imsm_update_reshape *u = (void *)update->buf;
10007 struct intel_dev *dl;
10008 void **space_tail = (void**)&update->space_list;
10009
095b8088
N		 10010 if (update->len < (int)sizeof(*u))
10011 return 0;
10012
1ade5cc1 10013 dprintf("for update_reshape\n");
d195167d
AK		 10014
10015 for (dl = super->devlist; dl; dl = dl->next) {
10016 int size = sizeof_imsm_dev(dl->dev, 1);
10017 void *s;
d677e0b8
AK		 10018 if (u->new_raid_disks > u->old_raid_disks)
10019 size += sizeof(__u32)*2*
10020 (u->new_raid_disks - u->old_raid_disks);
503975b9 10021 s = xmalloc(size);
d195167d
AK		 10022 *space_tail = s;
10023 space_tail = s;
10024 *space_tail = NULL;
10025 }
10026
10027 len = disks_to_mpb_size(u->new_raid_disks);
10028 dprintf("New anchor length is %llu\n", (unsigned long long)len);
78b10e66
N		 10029 break;
10030 }
48c5303a 10031 case update_reshape_migration: {
bc0b9d34
PC		 10032 /* for migration level 0->5 we need to add disks
10033 * so the same as for container operation we will copy
10034 * device to the bigger location.
10035 * in memory prepared device and new disk area are prepared
10036 * for usage in process update
10037 */
10038 struct imsm_update_reshape_migration *u = (void *)update->buf;
10039 struct intel_dev *id;
10040 void **space_tail = (void **)&update->space_list;
10041 int size;
10042 void *s;
10043 int current_level = -1;
10044
095b8088
N		 10045 if (update->len < (int)sizeof(*u))
10046 return 0;
10047
1ade5cc1 10048 dprintf("for update_reshape\n");
bc0b9d34
PC		 10049
10050 /* add space for bigger array in update
10051 */
10052 for (id = super->devlist; id; id = id->next) {
10053 if (id->index == (unsigned)u->subdev) {
10054 size = sizeof_imsm_dev(id->dev, 1);
10055 if (u->new_raid_disks > u->old_raid_disks)
10056 size += sizeof(__u32)*2*
10057 (u->new_raid_disks - u->old_raid_disks);
503975b9 10058 s = xmalloc(size);
bc0b9d34
PC		 10059 *space_tail = s;
10060 space_tail = s;
10061 *space_tail = NULL;
10062 break;
10063 }
10064 }
10065 if (update->space_list == NULL)
10066 break;
10067
10068 /* add space for disk in update
10069 */
10070 size = sizeof(struct dl);
503975b9 10071 s = xmalloc(size);
bc0b9d34
PC		 10072 *space_tail = s;
10073 space_tail = s;
10074 *space_tail = NULL;
10075
10076 /* add spare device to update
10077 */
10078 for (id = super->devlist ; id; id = id->next)
10079 if (id->index == (unsigned)u->subdev) {
10080 struct imsm_dev *dev;
10081 struct imsm_map *map;
10082
10083 dev = get_imsm_dev(super, u->subdev);
238c0a71 10084 map = get_imsm_map(dev, MAP_0);
bc0b9d34
PC		 10085 current_level = map->raid_level;
10086 break;
10087 }
089f9d79 10088 if (u->new_level == 5 && u->new_level != current_level) {
bc0b9d34
PC		 10089 struct mdinfo *spares;
10090
10091 spares = get_spares_for_grow(st);
10092 if (spares) {
10093 struct dl *dl;
10094 struct mdinfo *dev;
10095
10096 dev = spares->devs;
10097 if (dev) {
10098 u->new_disks[0] =
10099 makedev(dev->disk.major,
10100 dev->disk.minor);
10101 dl = get_disk_super(super,
10102 dev->disk.major,
10103 dev->disk.minor);
10104 dl->index = u->old_raid_disks;
10105 dev = dev->next;
10106 }
10107 sysfs_free(spares);
10108 }
10109 }
10110 len = disks_to_mpb_size(u->new_raid_disks);
10111 dprintf("New anchor length is %llu\n", (unsigned long long)len);
48c5303a
PC		 10112 break;
10113 }
f3871fdc 10114 case update_size_change: {
095b8088
N		 10115 if (update->len < (int)sizeof(struct imsm_update_size_change))
10116 return 0;
10117 break;
10118 }
10119 case update_activate_spare: {
10120 if (update->len < (int)sizeof(struct imsm_update_activate_spare))
10121 return 0;
f3871fdc
AK		 10122 break;
10123 }
949c47a0
DW		 10124 case update_create_array: {
10125 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 10126 struct intel_dev *dv;
54c2c1ea 10127 struct imsm_dev *dev = &u->dev;
238c0a71 10128 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW		 10129 struct dl *dl;
10130 struct disk_info *inf;
10131 int i;
10132 int activate = 0;
949c47a0 10133
095b8088
N		 10134 if (update->len < (int)sizeof(*u))
10135 return 0;
10136
54c2c1ea
DW		 10137 inf = get_disk_info(u);
10138 len = sizeof_imsm_dev(dev, 1);
ba2de7ba 10139 /* allocate a new super->devlist entry */
503975b9
N		 10140 dv = xmalloc(sizeof(*dv));
10141 dv->dev = xmalloc(len);
10142 update->space = dv;
949c47a0 10143
54c2c1ea
DW		 10144 /* count how many spares will be converted to members */
10145 for (i = 0; i < map->num_members; i++) {
10146 dl = serial_to_dl(inf[i].serial, super);
10147 if (!dl) {
10148 /* hmm maybe it failed?, nothing we can do about
10149 * it here
10150 */
10151 continue;
10152 }
10153 if (count_memberships(dl, super) == 0)
10154 activate++;
10155 }
10156 len += activate * sizeof(struct imsm_disk);
949c47a0 10157 break;
095b8088
N		 10158 }
10159 case update_kill_array: {
10160 if (update->len < (int)sizeof(struct imsm_update_kill_array))
10161 return 0;
949c47a0
DW		 10162 break;
10163 }
095b8088
N		 10164 case update_rename_array: {
10165 if (update->len < (int)sizeof(struct imsm_update_rename_array))
10166 return 0;
10167 break;
10168 }
10169 case update_add_remove_disk:
10170 /* no update->len needed */
10171 break;
bbab0940
TM		 10172 case update_prealloc_badblocks_mem:
10173 super->extra_space += sizeof(struct bbm_log) -
10174 get_imsm_bbm_log_size(super->bbm_log);
10175 break;
e6e9dd3f
AP		 10176 case update_rwh_policy: {
10177 if (update->len < (int)sizeof(struct imsm_update_rwh_policy))
10178 return 0;
10179 break;
10180 }
095b8088
N		 10181 default:
10182 return 0;
949c47a0 10183 }
8273f55e 10184
4d7b1503
DW		 10185 /* check if we need a larger metadata buffer */
10186 if (super->next_buf)
10187 buf_len = super->next_len;
10188 else
10189 buf_len = super->len;
10190
bbab0940 10191 if (__le32_to_cpu(mpb->mpb_size) + super->extra_space + len > buf_len) {
4d7b1503
DW		 10192 /* ok we need a larger buf than what is currently allocated
10193 * if this allocation fails process_update will notice that
10194 * ->next_len is set and ->next_buf is NULL
10195 */
bbab0940
TM		 10196 buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) +
10197 super->extra_space + len, sector_size);
4d7b1503
DW		 10198 if (super->next_buf)
10199 free(super->next_buf);
10200
10201 super->next_len = buf_len;
f36a9ecd 10202 if (posix_memalign(&super->next_buf, sector_size, buf_len) == 0)
1f45a8ad
DW		 10203 memset(super->next_buf, 0, buf_len);
10204 else
4d7b1503
DW		 10205 super->next_buf = NULL;
10206 }
5fe6f031 10207 return 1;
8273f55e
DW		 10208}
10209
ae6aad82 10210/* must be called while manager is quiesced */
f21e18ca 10211static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index)
ae6aad82
DW		 10212{
10213 struct imsm_super *mpb = super->anchor;
ae6aad82
DW		 10214 struct dl *iter;
10215 struct imsm_dev *dev;
10216 struct imsm_map *map;
4c9e8c1e 10217 unsigned int i, j, num_members;
fb12a745 10218 __u32 ord, ord_map0;
4c9e8c1e 10219 struct bbm_log *log = super->bbm_log;
ae6aad82 10220
1ade5cc1 10221 dprintf("deleting device[%d] from imsm_super\n", index);
ae6aad82
DW		 10222
10223 /* shift all indexes down one */
10224 for (iter = super->disks; iter; iter = iter->next)
f21e18ca 10225 if (iter->index > (int)index)
ae6aad82 10226 iter->index--;
47ee5a45 10227 for (iter = super->missing; iter; iter = iter->next)
f21e18ca 10228 if (iter->index > (int)index)
47ee5a45 10229 iter->index--;
ae6aad82
DW		 10230
10231 for (i = 0; i < mpb->num_raid_devs; i++) {
10232 dev = get_imsm_dev(super, i);
238c0a71 10233 map = get_imsm_map(dev, MAP_0);
24565c9a
DW		 10234 num_members = map->num_members;
10235 for (j = 0; j < num_members; j++) {
10236 /* update ord entries being careful not to propagate
10237 * ord-flags to the first map
10238 */
238c0a71 10239 ord = get_imsm_ord_tbl_ent(dev, j, MAP_X);
fb12a745 10240 ord_map0 = get_imsm_ord_tbl_ent(dev, j, MAP_0);
ae6aad82 10241
24565c9a
DW		 10242 if (ord_to_idx(ord) <= index)
10243 continue;
ae6aad82 10244
238c0a71 10245 map = get_imsm_map(dev, MAP_0);
fb12a745 10246 set_imsm_ord_tbl_ent(map, j, ord_map0 - 1);
238c0a71 10247 map = get_imsm_map(dev, MAP_1);
24565c9a
DW		 10248 if (map)
10249 set_imsm_ord_tbl_ent(map, j, ord - 1);
ae6aad82
DW		 10250 }
10251 }
10252
4c9e8c1e
TM		 10253 for (i = 0; i < log->entry_count; i++) {
10254 struct bbm_log_entry *entry = &log->marked_block_entries[i];
10255
10256 if (entry->disk_ordinal <= index)
10257 continue;
10258 entry->disk_ordinal--;
10259 }
10260
ae6aad82
DW		 10261 mpb->num_disks--;
10262 super->updates_pending++;
24565c9a
DW		 10263 if (*dlp) {
10264 struct dl *dl = *dlp;
10265
10266 *dlp = (*dlp)->next;
10267 __free_imsm_disk(dl);
10268 }
ae6aad82 10269}
9a717282
AK		 10270
10271static void close_targets(int *targets, int new_disks)
10272{
10273 int i;
10274
10275 if (!targets)
10276 return;
10277
10278 for (i = 0; i < new_disks; i++) {
10279 if (targets[i] >= 0) {
10280 close(targets[i]);
10281 targets[i] = -1;
10282 }
10283 }
10284}
10285
10286static int imsm_get_allowed_degradation(int level, int raid_disks,
10287 struct intel_super *super,
10288 struct imsm_dev *dev)
10289{
10290 switch (level) {
bf5cf7c7 10291 case 1:
9a717282
AK		 10292 case 10:{
10293 int ret_val = 0;
10294 struct imsm_map *map;
10295 int i;
10296
10297 ret_val = raid_disks/2;
10298 /* check map if all disks pairs not failed
10299 * in both maps
10300 */
238c0a71 10301 map = get_imsm_map(dev, MAP_0);
9a717282
AK		 10302 for (i = 0; i < ret_val; i++) {
10303 int degradation = 0;
10304 if (get_imsm_disk(super, i) == NULL)
10305 degradation++;
10306 if (get_imsm_disk(super, i + 1) == NULL)
10307 degradation++;
10308 if (degradation == 2)
10309 return 0;
10310 }
238c0a71 10311 map = get_imsm_map(dev, MAP_1);
9a717282
AK		 10312 /* if there is no second map
10313 * result can be returned
10314 */
10315 if (map == NULL)
10316 return ret_val;
10317 /* check degradation in second map
10318 */
10319 for (i = 0; i < ret_val; i++) {
10320 int degradation = 0;
10321 if (get_imsm_disk(super, i) == NULL)
10322 degradation++;
10323 if (get_imsm_disk(super, i + 1) == NULL)
10324 degradation++;
10325 if (degradation == 2)
10326 return 0;
10327 }
10328 return ret_val;
10329 }
10330 case 5:
10331 return 1;
10332 case 6:
10333 return 2;
10334 default:
10335 return 0;
10336 }
10337}
10338
687629c2
AK		 10339/*******************************************************************************
10340 * Function: open_backup_targets
10341 * Description: Function opens file descriptors for all devices given in
10342 * info->devs
10343 * Parameters:
10344 * info : general array info
10345 * raid_disks : number of disks
10346 * raid_fds : table of device's file descriptors
9a717282
AK		 10347 * super : intel super for raid10 degradation check
10348 * dev : intel device for raid10 degradation check
687629c2
AK		 10349 * Returns:
10350 * 0 : success
10351 * -1 : fail
10352 ******************************************************************************/
9a717282
AK		 10353int open_backup_targets(struct mdinfo *info, int raid_disks, int *raid_fds,
10354 struct intel_super *super, struct imsm_dev *dev)
687629c2
AK		 10355{
10356 struct mdinfo *sd;
f627f5ad 10357 int i;
9a717282 10358 int opened = 0;
f627f5ad
AK		 10359
10360 for (i = 0; i < raid_disks; i++)
10361 raid_fds[i] = -1;
687629c2
AK		 10362
10363 for (sd = info->devs ; sd ; sd = sd->next) {
10364 char *dn;
10365
10366 if (sd->disk.state & (1<<MD_DISK_FAULTY)) {
10367 dprintf("disk is faulty!!\n");
10368 continue;
10369 }
10370
089f9d79 10371 if (sd->disk.raid_disk >= raid_disks || sd->disk.raid_disk < 0)
687629c2
AK		 10372 continue;
10373
10374 dn = map_dev(sd->disk.major,
10375 sd->disk.minor, 1);
10376 raid_fds[sd->disk.raid_disk] = dev_open(dn, O_RDWR);
10377 if (raid_fds[sd->disk.raid_disk] < 0) {
e12b3daa 10378 pr_err("cannot open component\n");
9a717282 10379 continue;
687629c2 10380 }
9a717282
AK		 10381 opened++;
10382 }
10383 /* check if maximum array degradation level is not exceeded
10384 */
10385 if ((raid_disks - opened) >
089f9d79
JS		 10386 imsm_get_allowed_degradation(info->new_level, raid_disks,
10387 super, dev)) {
e12b3daa 10388 pr_err("Not enough disks can be opened.\n");
9a717282
AK		 10389 close_targets(raid_fds, raid_disks);
10390 return -2;
687629c2
AK		 10391 }
10392 return 0;
10393}
10394
d31ad643
PB		 10395/*******************************************************************************
10396 * Function: validate_container_imsm
10397 * Description: This routine validates container after assemble,
10398 * eg. if devices in container are under the same controller.
10399 *
10400 * Parameters:
10401 * info : linked list with info about devices used in array
10402 * Returns:
10403 * 1 : HBA mismatch
10404 * 0 : Success
10405 ******************************************************************************/
10406int validate_container_imsm(struct mdinfo *info)
10407{
6b781d33
AP		 10408 if (check_env("IMSM_NO_PLATFORM"))
10409 return 0;
d31ad643 10410
6b781d33
AP		 10411 struct sys_dev *idev;
10412 struct sys_dev *hba = NULL;
10413 struct sys_dev *intel_devices = find_intel_devices();
10414 char *dev_path = devt_to_devpath(makedev(info->disk.major,
10415 info->disk.minor));
10416
10417 for (idev = intel_devices; idev; idev = idev->next) {
10418 if (dev_path && strstr(dev_path, idev->path)) {
10419 hba = idev;
10420 break;
d31ad643 10421 }
6b781d33
AP		 10422 }
10423 if (dev_path)
d31ad643
PB		 10424 free(dev_path);
10425
6b781d33
AP		 10426 if (!hba) {
10427 pr_err("WARNING - Cannot detect HBA for device %s!\n",
10428 devid2kname(makedev(info->disk.major, info->disk.minor)));
10429 return 1;
10430 }
10431
10432 const struct imsm_orom *orom = get_orom_by_device_id(hba->dev_id);
10433 struct mdinfo *dev;
10434
10435 for (dev = info->next; dev; dev = dev->next) {
10436 dev_path = devt_to_devpath(makedev(dev->disk.major, dev->disk.minor));
10437
10438 struct sys_dev *hba2 = NULL;
10439 for (idev = intel_devices; idev; idev = idev->next) {
10440 if (dev_path && strstr(dev_path, idev->path)) {
10441 hba2 = idev;
10442 break;
d31ad643
PB		 10443 }
10444 }
6b781d33
AP		 10445 if (dev_path)
10446 free(dev_path);
10447
10448 const struct imsm_orom *orom2 = hba2 == NULL ? NULL :
10449 get_orom_by_device_id(hba2->dev_id);
10450
10451 if (hba2 && hba->type != hba2->type) {
10452 pr_err("WARNING - HBAs of devices do not match %s != %s\n",
10453 get_sys_dev_type(hba->type), get_sys_dev_type(hba2->type));
10454 return 1;
10455 }
10456
07cb1e57 10457 if (orom != orom2) {
6b781d33
AP		 10458 pr_err("WARNING - IMSM container assembled with disks under different HBAs!\n"
10459 " This operation is not supported and can lead to data loss.\n");
10460 return 1;
10461 }
10462
10463 if (!orom) {
10464 pr_err("WARNING - IMSM container assembled with disks under HBAs without IMSM platform support!\n"
10465 " This operation is not supported and can lead to data loss.\n");
10466 return 1;
10467 }
d31ad643 10468 }
6b781d33 10469
d31ad643
PB		 10470 return 0;
10471}
32141c17 10472
6f50473f
TM		 10473/*******************************************************************************
10474* Function: imsm_record_badblock
10475* Description: This routine stores new bad block record in BBM log
10476*
10477* Parameters:
10478* a : array containing a bad block
10479* slot : disk number containing a bad block
10480* sector : bad block sector
10481* length : bad block sectors range
10482* Returns:
10483* 1 : Success
10484* 0 : Error
10485******************************************************************************/
10486static int imsm_record_badblock(struct active_array *a, int slot,
10487 unsigned long long sector, int length)
10488{
10489 struct intel_super *super = a->container->sb;
10490 int ord;
10491 int ret;
10492
10493 ord = imsm_disk_slot_to_ord(a, slot);
10494 if (ord < 0)
10495 return 0;
10496
10497 ret = record_new_badblock(super->bbm_log, ord_to_idx(ord), sector,
10498 length);
10499 if (ret)
10500 super->updates_pending++;
10501
10502 return ret;
10503}
c07a5a4f
TM		 10504/*******************************************************************************
10505* Function: imsm_clear_badblock
10506* Description: This routine clears bad block record from BBM log
10507*
10508* Parameters:
10509* a : array containing a bad block
10510* slot : disk number containing a bad block
10511* sector : bad block sector
10512* length : bad block sectors range
10513* Returns:
10514* 1 : Success
10515* 0 : Error
10516******************************************************************************/
10517static int imsm_clear_badblock(struct active_array *a, int slot,
10518 unsigned long long sector, int length)
10519{
10520 struct intel_super *super = a->container->sb;
10521 int ord;
10522 int ret;
10523
10524 ord = imsm_disk_slot_to_ord(a, slot);
10525 if (ord < 0)
10526 return 0;
10527
10528 ret = clear_badblock(super->bbm_log, ord_to_idx(ord), sector, length);
10529 if (ret)
10530 super->updates_pending++;
10531
10532 return ret;
10533}
928f1424
TM		 10534/*******************************************************************************
10535* Function: imsm_get_badblocks
10536* Description: This routine get list of bad blocks for an array
10537*
10538* Parameters:
10539* a : array
10540* slot : disk number
10541* Returns:
10542* bb : structure containing bad blocks
10543* NULL : error
10544******************************************************************************/
10545static struct md_bb *imsm_get_badblocks(struct active_array *a, int slot)
10546{
10547 int inst = a->info.container_member;
10548 struct intel_super *super = a->container->sb;
10549 struct imsm_dev *dev = get_imsm_dev(super, inst);
10550 struct imsm_map *map = get_imsm_map(dev, MAP_0);
10551 int ord;
10552
10553 ord = imsm_disk_slot_to_ord(a, slot);
10554 if (ord < 0)
10555 return NULL;
10556
10557 get_volume_badblocks(super->bbm_log, ord_to_idx(ord), pba_of_lba0(map),
44490938 10558 per_dev_array_size(map), &super->bb);
928f1424
TM		 10559
10560 return &super->bb;
10561}
27156a57
TM		 10562/*******************************************************************************
10563* Function: examine_badblocks_imsm
10564* Description: Prints list of bad blocks on a disk to the standard output
10565*
10566* Parameters:
10567* st : metadata handler
10568* fd : open file descriptor for device
10569* devname : device name
10570* Returns:
10571* 0 : Success
10572* 1 : Error
10573******************************************************************************/
10574static int examine_badblocks_imsm(struct supertype *st, int fd, char *devname)
10575{
10576 struct intel_super *super = st->sb;
10577 struct bbm_log *log = super->bbm_log;
10578 struct dl *d = NULL;
10579 int any = 0;
10580
10581 for (d = super->disks; d ; d = d->next) {
10582 if (strcmp(d->devname, devname) == 0)
10583 break;
10584 }
10585
10586 if ((d == NULL) || (d->index < 0)) { /* serial mismatch probably */
10587 pr_err("%s doesn't appear to be part of a raid array\n",
10588 devname);
10589 return 1;
10590 }
10591
10592 if (log != NULL) {
10593 unsigned int i;
10594 struct bbm_log_entry *entry = &log->marked_block_entries[0];
10595
10596 for (i = 0; i < log->entry_count; i++) {
10597 if (entry[i].disk_ordinal == d->index) {
10598 unsigned long long sector = __le48_to_cpu(
10599 &entry[i].defective_block_start);
10600 int cnt = entry[i].marked_count + 1;
10601
10602 if (!any) {
10603 printf("Bad-blocks on %s:\n", devname);
10604 any = 1;
10605 }
10606
10607 printf("%20llu for %d sectors\n", sector, cnt);
10608 }
10609 }
10610 }
10611
10612 if (!any)
10613 printf("No bad-blocks list configured on %s\n", devname);
10614
10615 return 0;
10616}
687629c2
AK		 10617/*******************************************************************************
10618 * Function: init_migr_record_imsm
10619 * Description: Function inits imsm migration record
10620 * Parameters:
10621 * super : imsm internal array info
10622 * dev : device under migration
10623 * info : general array info to find the smallest device
10624 * Returns:
10625 * none
10626 ******************************************************************************/
10627void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev,
10628 struct mdinfo *info)
10629{
10630 struct intel_super *super = st->sb;
10631 struct migr_record *migr_rec = super->migr_rec;
10632 int new_data_disks;
10633 unsigned long long dsize, dev_sectors;
10634 long long unsigned min_dev_sectors = -1LLU;
10635 struct mdinfo *sd;
10636 char nm[30];
10637 int fd;
238c0a71
AK		 10638 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
10639 struct imsm_map *map_src = get_imsm_map(dev, MAP_1);
687629c2 10640 unsigned long long num_migr_units;
3ef4403c 10641 unsigned long long array_blocks;
687629c2
AK		 10642
10643 memset(migr_rec, 0, sizeof(struct migr_record));
10644 migr_rec->family_num = __cpu_to_le32(super->anchor->family_num);
10645
10646 /* only ascending reshape supported now */
10647 migr_rec->ascending_migr = __cpu_to_le32(1);
10648
10649 migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE /
10650 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
e1742195
AK		 10651 migr_rec->dest_depth_per_unit *=
10652 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
9529d343 10653 new_data_disks = imsm_num_data_members(map_dest);
687629c2
AK		 10654 migr_rec->blocks_per_unit =
10655 __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks);
10656 migr_rec->dest_depth_per_unit =
10657 __cpu_to_le32(migr_rec->dest_depth_per_unit);
3ef4403c 10658 array_blocks = info->component_size * new_data_disks;
687629c2
AK		 10659 num_migr_units =
10660 array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit);
10661
10662 if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit))
10663 num_migr_units++;
10664 migr_rec->num_migr_units = __cpu_to_le32(num_migr_units);
10665
10666 migr_rec->post_migr_vol_cap = dev->size_low;
10667 migr_rec->post_migr_vol_cap_hi = dev->size_high;
10668
687629c2
AK		 10669 /* Find the smallest dev */
10670 for (sd = info->devs ; sd ; sd = sd->next) {
10671 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
10672 fd = dev_open(nm, O_RDONLY);
10673 if (fd < 0)
10674 continue;
10675 get_dev_size(fd, NULL, &dsize);
10676 dev_sectors = dsize / 512;
10677 if (dev_sectors < min_dev_sectors)
10678 min_dev_sectors = dev_sectors;
10679 close(fd);
10680 }
10681 migr_rec->ckpt_area_pba = __cpu_to_le32(min_dev_sectors -
10682 RAID_DISK_RESERVED_BLOCKS_IMSM_HI);
10683
10684 write_imsm_migr_rec(st);
10685
10686 return;
10687}
10688
10689/*******************************************************************************
10690 * Function: save_backup_imsm
10691 * Description: Function saves critical data stripes to Migration Copy Area
10692 * and updates the current migration unit status.
10693 * Use restore_stripes() to form a destination stripe,
10694 * and to write it to the Copy Area.
10695 * Parameters:
10696 * st : supertype information
aea93171 10697 * dev : imsm device that backup is saved for
687629c2
AK		 10698 * info : general array info
10699 * buf : input buffer
687629c2
AK		 10700 * length : length of data to backup (blocks_per_unit)
10701 * Returns:
10702 * 0 : success
10703 *, -1 : fail
10704 ******************************************************************************/
10705int save_backup_imsm(struct supertype *st,
10706 struct imsm_dev *dev,
10707 struct mdinfo *info,
10708 void *buf,
687629c2
AK		 10709 int length)
10710{
10711 int rv = -1;
10712 struct intel_super *super = st->sb;
594dc1b8
JS		 10713 unsigned long long *target_offsets;
10714 int *targets;
687629c2 10715 int i;
238c0a71 10716 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
687629c2 10717 int new_disks = map_dest->num_members;
ab724b98
AK		 10718 int dest_layout = 0;
10719 int dest_chunk;
d1877f69 10720 unsigned long long start;
9529d343 10721 int data_disks = imsm_num_data_members(map_dest);
687629c2 10722
503975b9 10723 targets = xmalloc(new_disks * sizeof(int));
687629c2 10724
7e45b550
AK		 10725 for (i = 0; i < new_disks; i++)
10726 targets[i] = -1;
10727
503975b9 10728 target_offsets = xcalloc(new_disks, sizeof(unsigned long long));
687629c2 10729
d1877f69 10730 start = info->reshape_progress * 512;
687629c2 10731 for (i = 0; i < new_disks; i++) {
687629c2
AK		 10732 target_offsets[i] = (unsigned long long)
10733 __le32_to_cpu(super->migr_rec->ckpt_area_pba) * 512;
d1877f69
AK		 10734 /* move back copy area adderss, it will be moved forward
10735 * in restore_stripes() using start input variable
10736 */
10737 target_offsets[i] -= start/data_disks;
687629c2
AK		 10738 }
10739
9a717282
AK		 10740 if (open_backup_targets(info, new_disks, targets,
10741 super, dev))
687629c2
AK		 10742 goto abort;
10743
68eb8bc6 10744 dest_layout = imsm_level_to_layout(map_dest->raid_level);
ab724b98
AK		 10745 dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512;
10746
687629c2
AK		 10747 if (restore_stripes(targets, /* list of dest devices */
10748 target_offsets, /* migration record offsets */
10749 new_disks,
ab724b98
AK		 10750 dest_chunk,
10751 map_dest->raid_level,
10752 dest_layout,
10753 -1, /* source backup file descriptor */
10754 0, /* input buf offset
10755 * always 0 buf is already offseted */
d1877f69 10756 start,
687629c2
AK		 10757 length,
10758 buf) != 0) {
e7b84f9d 10759 pr_err("Error restoring stripes\n");
687629c2
AK		 10760 goto abort;
10761 }
10762
10763 rv = 0;
10764
10765abort:
10766 if (targets) {
9a717282 10767 close_targets(targets, new_disks);
687629c2
AK		 10768 free(targets);
10769 }
10770 free(target_offsets);
10771
10772 return rv;
10773}
10774
10775/*******************************************************************************
10776 * Function: save_checkpoint_imsm
10777 * Description: Function called for current unit status update
10778 * in the migration record. It writes it to disk.
10779 * Parameters:
10780 * super : imsm internal array info
10781 * info : general array info
10782 * Returns:
10783 * 0: success
10784 * 1: failure
0228d92c
AK		 10785 * 2: failure, means no valid migration record
10786 * / no general migration in progress /
687629c2
AK		 10787 ******************************************************************************/
10788int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state)
10789{
10790 struct intel_super *super = st->sb;
f8b72ef5
AK		 10791 unsigned long long blocks_per_unit;
10792 unsigned long long curr_migr_unit;
10793
2e062e82 10794 if (load_imsm_migr_rec(super, info) != 0) {
7a862a02 10795 dprintf("imsm: ERROR: Cannot read migration record for checkpoint save.\n");
2e062e82
AK		 10796 return 1;
10797 }
10798
f8b72ef5
AK		 10799 blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit);
10800 if (blocks_per_unit == 0) {
0228d92c
AK		 10801 dprintf("imsm: no migration in progress.\n");
10802 return 2;
687629c2 10803 }
f8b72ef5
AK		 10804 curr_migr_unit = info->reshape_progress / blocks_per_unit;
10805 /* check if array is alligned to copy area
10806 * if it is not alligned, add one to current migration unit value
10807 * this can happend on array reshape finish only
10808 */
10809 if (info->reshape_progress % blocks_per_unit)
10810 curr_migr_unit++;
687629c2
AK		 10811
10812 super->migr_rec->curr_migr_unit =
f8b72ef5 10813 __cpu_to_le32(curr_migr_unit);
687629c2
AK		 10814 super->migr_rec->rec_status = __cpu_to_le32(state);
10815 super->migr_rec->dest_1st_member_lba =
f8b72ef5
AK		 10816 __cpu_to_le32(curr_migr_unit *
10817 __le32_to_cpu(super->migr_rec->dest_depth_per_unit));
687629c2 10818 if (write_imsm_migr_rec(st) < 0) {
7a862a02 10819 dprintf("imsm: Cannot write migration record outside backup area\n");
687629c2
AK		 10820 return 1;
10821 }
10822
10823 return 0;
10824}
10825
276d77db
AK		 10826/*******************************************************************************
10827 * Function: recover_backup_imsm
10828 * Description: Function recovers critical data from the Migration Copy Area
10829 * while assembling an array.
10830 * Parameters:
10831 * super : imsm internal array info
10832 * info : general array info
10833 * Returns:
10834 * 0 : success (or there is no data to recover)
10835 * 1 : fail
10836 ******************************************************************************/
10837int recover_backup_imsm(struct supertype *st, struct mdinfo *info)
10838{
10839 struct intel_super *super = st->sb;
10840 struct migr_record *migr_rec = super->migr_rec;
594dc1b8 10841 struct imsm_map *map_dest;
276d77db
AK		 10842 struct intel_dev *id = NULL;
10843 unsigned long long read_offset;
10844 unsigned long long write_offset;
10845 unsigned unit_len;
10846 int *targets = NULL;
10847 int new_disks, i, err;
10848 char *buf = NULL;
10849 int retval = 1;
f36a9ecd 10850 unsigned int sector_size = super->sector_size;
276d77db
AK		 10851 unsigned long curr_migr_unit = __le32_to_cpu(migr_rec->curr_migr_unit);
10852 unsigned long num_migr_units = __le32_to_cpu(migr_rec->num_migr_units);
276d77db 10853 char buffer[20];
6c3560c0 10854 int skipped_disks = 0;
276d77db
AK		 10855
10856 err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20);
10857 if (err < 1)
10858 return 1;
10859
10860 /* recover data only during assemblation */
10861 if (strncmp(buffer, "inactive", 8) != 0)
10862 return 0;
10863 /* no data to recover */
10864 if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL)
10865 return 0;
10866 if (curr_migr_unit >= num_migr_units)
10867 return 1;
10868
10869 /* find device during reshape */
10870 for (id = super->devlist; id; id = id->next)
10871 if (is_gen_migration(id->dev))
10872 break;
10873 if (id == NULL)
10874 return 1;
10875
238c0a71 10876 map_dest = get_imsm_map(id->dev, MAP_0);
276d77db
AK		 10877 new_disks = map_dest->num_members;
10878
10879 read_offset = (unsigned long long)
10880 __le32_to_cpu(migr_rec->ckpt_area_pba) * 512;
10881
10882 write_offset = ((unsigned long long)
10883 __le32_to_cpu(migr_rec->dest_1st_member_lba) +
5551b113 10884 pba_of_lba0(map_dest)) * 512;
276d77db
AK		 10885
10886 unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
f36a9ecd 10887 if (posix_memalign((void **)&buf, sector_size, unit_len) != 0)
276d77db 10888 goto abort;
503975b9 10889 targets = xcalloc(new_disks, sizeof(int));
276d77db 10890
9a717282 10891 if (open_backup_targets(info, new_disks, targets, super, id->dev)) {
e7b84f9d 10892 pr_err("Cannot open some devices belonging to array.\n");
f627f5ad
AK		 10893 goto abort;
10894 }
276d77db
AK		 10895
10896 for (i = 0; i < new_disks; i++) {
6c3560c0
AK		 10897 if (targets[i] < 0) {
10898 skipped_disks++;
10899 continue;
10900 }
276d77db 10901 if (lseek64(targets[i], read_offset, SEEK_SET) < 0) {
e7b84f9d
N		 10902 pr_err("Cannot seek to block: %s\n",
10903 strerror(errno));
137debce
AK		 10904 skipped_disks++;
10905 continue;
276d77db 10906 }
9ec11d1a 10907 if ((unsigned)read(targets[i], buf, unit_len) != unit_len) {
e7b84f9d
N		 10908 pr_err("Cannot read copy area block: %s\n",
10909 strerror(errno));
137debce
AK		 10910 skipped_disks++;
10911 continue;
276d77db
AK		 10912 }
10913 if (lseek64(targets[i], write_offset, SEEK_SET) < 0) {
e7b84f9d
N		 10914 pr_err("Cannot seek to block: %s\n",
10915 strerror(errno));
137debce
AK		 10916 skipped_disks++;
10917 continue;
276d77db 10918 }
9ec11d1a 10919 if ((unsigned)write(targets[i], buf, unit_len) != unit_len) {
e7b84f9d
N		 10920 pr_err("Cannot restore block: %s\n",
10921 strerror(errno));
137debce
AK		 10922 skipped_disks++;
10923 continue;
276d77db
AK		 10924 }
10925 }
10926
137debce
AK		 10927 if (skipped_disks > imsm_get_allowed_degradation(info->new_level,
10928 new_disks,
10929 super,
10930 id->dev)) {
7a862a02 10931 pr_err("Cannot restore data from backup. Too many failed disks\n");
6c3560c0
AK		 10932 goto abort;
10933 }
10934
befb629b
AK		 10935 if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) {
10936 /* ignore error == 2, this can mean end of reshape here
10937 */
7a862a02 10938 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL) during restart\n");
befb629b 10939 } else
276d77db 10940 retval = 0;
276d77db
AK		 10941
10942abort:
10943 if (targets) {
10944 for (i = 0; i < new_disks; i++)
10945 if (targets[i])
10946 close(targets[i]);
10947 free(targets);
10948 }
10949 free(buf);
10950 return retval;
10951}
10952
2cda7640
ML		 10953static char disk_by_path[] = "/dev/disk/by-path/";
10954
10955static const char *imsm_get_disk_controller_domain(const char *path)
10956{
2cda7640 10957 char disk_path[PATH_MAX];
96234762
LM		 10958 char *drv=NULL;
10959 struct stat st;
2cda7640 10960
6d8d290a 10961 strcpy(disk_path, disk_by_path);
96234762
LM		 10962 strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1);
10963 if (stat(disk_path, &st) == 0) {
10964 struct sys_dev* hba;
594dc1b8 10965 char *path;
96234762
LM		 10966
10967 path = devt_to_devpath(st.st_rdev);
10968 if (path == NULL)
10969 return "unknown";
10970 hba = find_disk_attached_hba(-1, path);
10971 if (hba && hba->type == SYS_DEV_SAS)
10972 drv = "isci";
10973 else if (hba && hba->type == SYS_DEV_SATA)
10974 drv = "ahci";
c6839718
MT		 10975 else if (hba && hba->type == SYS_DEV_VMD)
10976 drv = "vmd";
10977 else if (hba && hba->type == SYS_DEV_NVME)
10978 drv = "nvme";
1011e834 10979 else
96234762
LM		 10980 drv = "unknown";
10981 dprintf("path: %s hba: %s attached: %s\n",
10982 path, (hba) ? hba->path : "NULL", drv);
10983 free(path);
2cda7640 10984 }
96234762 10985 return drv;
2cda7640
ML		 10986}
10987
4dd2df09 10988static char *imsm_find_array_devnm_by_subdev(int subdev, char *container)
78b10e66 10989{
4dd2df09 10990 static char devnm[32];
78b10e66
N		 10991 char subdev_name[20];
10992 struct mdstat_ent *mdstat;
10993
10994 sprintf(subdev_name, "%d", subdev);
10995 mdstat = mdstat_by_subdev(subdev_name, container);
10996 if (!mdstat)
4dd2df09 10997 return NULL;
78b10e66 10998
4dd2df09 10999 strcpy(devnm, mdstat->devnm);
78b10e66 11000 free_mdstat(mdstat);
4dd2df09 11001 return devnm;
78b10e66
N		 11002}
11003
11004static int imsm_reshape_is_allowed_on_container(struct supertype *st,
11005 struct geo_params *geo,
fbf3d202
AK		 11006 int *old_raid_disks,
11007 int direction)
78b10e66 11008{
694575e7
KW		 11009 /* currently we only support increasing the number of devices
11010 * for a container. This increases the number of device for each
11011 * member array. They must all be RAID0 or RAID5.
11012 */
78b10e66
N		 11013 int ret_val = 0;
11014 struct mdinfo *info, *member;
11015 int devices_that_can_grow = 0;
11016
7a862a02 11017 dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): st->devnm = (%s)\n", st->devnm);
78b10e66 11018
d04f65f4 11019 if (geo->size > 0 ||
78b10e66
N		 11020 geo->level != UnSet ||
11021 geo->layout != UnSet ||
11022 geo->chunksize != 0 ||
11023 geo->raid_disks == UnSet) {
7a862a02 11024 dprintf("imsm: Container operation is allowed for raid disks number change only.\n");
78b10e66
N		 11025 return ret_val;
11026 }
11027
fbf3d202 11028 if (direction == ROLLBACK_METADATA_CHANGES) {
7a862a02 11029 dprintf("imsm: Metadata changes rollback is not supported for container operation.\n");
fbf3d202
AK		 11030 return ret_val;
11031 }
11032
78b10e66
N		 11033 info = container_content_imsm(st, NULL);
11034 for (member = info; member; member = member->next) {
4dd2df09 11035 char *result;
78b10e66
N		 11036
11037 dprintf("imsm: checking device_num: %i\n",
11038 member->container_member);
11039
d7d205bd 11040 if (geo->raid_disks <= member->array.raid_disks) {
78b10e66
N		 11041 /* we work on container for Online Capacity Expansion
11042 * only so raid_disks has to grow
11043 */
7a862a02 11044 dprintf("imsm: for container operation raid disks increase is required\n");
78b10e66
N		 11045 break;
11046 }
11047
089f9d79 11048 if (info->array.level != 0 && info->array.level != 5) {
78b10e66
N		 11049 /* we cannot use this container with other raid level
11050 */
7a862a02 11051 dprintf("imsm: for container operation wrong raid level (%i) detected\n",
78b10e66
N		 11052 info->array.level);
11053 break;
11054 } else {
11055 /* check for platform support
11056 * for this raid level configuration
11057 */
11058 struct intel_super *super = st->sb;
11059 if (!is_raid_level_supported(super->orom,
11060 member->array.level,
11061 geo->raid_disks)) {
7a862a02 11062 dprintf("platform does not support raid%d with %d disk%s\n",
78b10e66
N		 11063 info->array.level,
11064 geo->raid_disks,
11065 geo->raid_disks > 1 ? "s" : "");
11066 break;
11067 }
2a4a08e7
AK		 11068 /* check if component size is aligned to chunk size
11069 */
11070 if (info->component_size %
11071 (info->array.chunk_size/512)) {
7a862a02 11072 dprintf("Component size is not aligned to chunk size\n");
2a4a08e7
AK		 11073 break;
11074 }
78b10e66
N		 11075 }
11076
11077 if (*old_raid_disks &&
11078 info->array.raid_disks != *old_raid_disks)
11079 break;
11080 *old_raid_disks = info->array.raid_disks;
11081
11082 /* All raid5 and raid0 volumes in container
11083 * have to be ready for Online Capacity Expansion
11084 * so they need to be assembled. We have already
11085 * checked that no recovery etc is happening.
11086 */
4dd2df09
N		 11087 result = imsm_find_array_devnm_by_subdev(member->container_member,
11088 st->container_devnm);
11089 if (result == NULL) {
78b10e66
N		 11090 dprintf("imsm: cannot find array\n");
11091 break;
11092 }
11093 devices_that_can_grow++;
11094 }
11095 sysfs_free(info);
11096 if (!member && devices_that_can_grow)
11097 ret_val = 1;
11098
11099 if (ret_val)
1ade5cc1 11100 dprintf("Container operation allowed\n");
78b10e66 11101 else
1ade5cc1 11102 dprintf("Error: %i\n", ret_val);
78b10e66
N		 11103
11104 return ret_val;
11105}
11106
11107/* Function: get_spares_for_grow
11108 * Description: Allocates memory and creates list of spare devices
1011e834 11109 * avaliable in container. Checks if spare drive size is acceptable.
78b10e66
N		 11110 * Parameters: Pointer to the supertype structure
11111 * Returns: Pointer to the list of spare devices (mdinfo structure) on success,
1011e834 11112 * NULL if fail
78b10e66
N		 11113 */
11114static struct mdinfo *get_spares_for_grow(struct supertype *st)
11115{
fbfdcb06
AO		 11116 struct spare_criteria sc;
11117
11118 get_spare_criteria_imsm(st, &sc);
11119 return container_choose_spares(st, &sc, NULL, NULL, NULL, 0);
78b10e66
N		 11120}
11121
11122/******************************************************************************
11123 * function: imsm_create_metadata_update_for_reshape
11124 * Function creates update for whole IMSM container.
11125 *
11126 ******************************************************************************/
11127static int imsm_create_metadata_update_for_reshape(
11128 struct supertype *st,
11129 struct geo_params *geo,
11130 int old_raid_disks,
11131 struct imsm_update_reshape **updatep)
11132{
11133 struct intel_super *super = st->sb;
11134 struct imsm_super *mpb = super->anchor;
594dc1b8
JS		 11135 int update_memory_size;
11136 struct imsm_update_reshape *u;
11137 struct mdinfo *spares;
78b10e66 11138 int i;
594dc1b8 11139 int delta_disks;
bbd24d86 11140 struct mdinfo *dev;
78b10e66 11141
1ade5cc1 11142 dprintf("(enter) raid_disks = %i\n", geo->raid_disks);
78b10e66
N		 11143
11144 delta_disks = geo->raid_disks - old_raid_disks;
11145
11146 /* size of all update data without anchor */
11147 update_memory_size = sizeof(struct imsm_update_reshape);
11148
11149 /* now add space for spare disks that we need to add. */
11150 update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1);
11151
503975b9 11152 u = xcalloc(1, update_memory_size);
78b10e66
N		 11153 u->type = update_reshape_container_disks;
11154 u->old_raid_disks = old_raid_disks;
11155 u->new_raid_disks = geo->raid_disks;
11156
11157 /* now get spare disks list
11158 */
11159 spares = get_spares_for_grow(st);
11160
d7be7d87 11161 if (spares == NULL || delta_disks > spares->array.spare_disks) {
7a862a02 11162 pr_err("imsm: ERROR: Cannot get spare devices for %s.\n", geo->dev_name);
e4c72d1d 11163 i = -1;
78b10e66
N		 11164 goto abort;
11165 }
11166
11167 /* we have got spares
11168 * update disk list in imsm_disk list table in anchor
11169 */
11170 dprintf("imsm: %i spares are available.\n\n",
11171 spares->array.spare_disks);
11172
bbd24d86 11173 dev = spares->devs;
78b10e66 11174 for (i = 0; i < delta_disks; i++) {
78b10e66
N		 11175 struct dl *dl;
11176
bbd24d86
AK		 11177 if (dev == NULL)
11178 break;
78b10e66
N		 11179 u->new_disks[i] = makedev(dev->disk.major,
11180 dev->disk.minor);
11181 dl = get_disk_super(super, dev->disk.major, dev->disk.minor);
ee4beede
AK		 11182 dl->index = mpb->num_disks;
11183 mpb->num_disks++;
bbd24d86 11184 dev = dev->next;
78b10e66 11185 }
78b10e66
N		 11186
11187abort:
11188 /* free spares
11189 */
11190 sysfs_free(spares);
11191
d677e0b8 11192 dprintf("imsm: reshape update preparation :");
78b10e66 11193 if (i == delta_disks) {
1ade5cc1 11194 dprintf_cont(" OK\n");
78b10e66
N		 11195 *updatep = u;
11196 return update_memory_size;
11197 }
11198 free(u);
1ade5cc1 11199 dprintf_cont(" Error\n");
78b10e66
N		 11200
11201 return 0;
11202}
11203
f3871fdc
AK		 11204/******************************************************************************
11205 * function: imsm_create_metadata_update_for_size_change()
11206 * Creates update for IMSM array for array size change.
11207 *
11208 ******************************************************************************/
11209static int imsm_create_metadata_update_for_size_change(
11210 struct supertype *st,
11211 struct geo_params *geo,
11212 struct imsm_update_size_change **updatep)
11213{
11214 struct intel_super *super = st->sb;
594dc1b8
JS		 11215 int update_memory_size;
11216 struct imsm_update_size_change *u;
f3871fdc 11217
1ade5cc1 11218 dprintf("(enter) New size = %llu\n", geo->size);
f3871fdc
AK		 11219
11220 /* size of all update data without anchor */
11221 update_memory_size = sizeof(struct imsm_update_size_change);
11222
503975b9 11223 u = xcalloc(1, update_memory_size);
f3871fdc
AK		 11224 u->type = update_size_change;
11225 u->subdev = super->current_vol;
11226 u->new_size = geo->size;
11227
11228 dprintf("imsm: reshape update preparation : OK\n");
11229 *updatep = u;
11230
11231 return update_memory_size;
11232}
11233
48c5303a
PC		 11234/******************************************************************************
11235 * function: imsm_create_metadata_update_for_migration()
11236 * Creates update for IMSM array.
11237 *
11238 ******************************************************************************/
11239static int imsm_create_metadata_update_for_migration(
11240 struct supertype *st,
11241 struct geo_params *geo,
11242 struct imsm_update_reshape_migration **updatep)
11243{
11244 struct intel_super *super = st->sb;
594dc1b8
JS		 11245 int update_memory_size;
11246 struct imsm_update_reshape_migration *u;
48c5303a
PC		 11247 struct imsm_dev *dev;
11248 int previous_level = -1;
11249
1ade5cc1 11250 dprintf("(enter) New Level = %i\n", geo->level);
48c5303a
PC		 11251
11252 /* size of all update data without anchor */
11253 update_memory_size = sizeof(struct imsm_update_reshape_migration);
11254
503975b9 11255 u = xcalloc(1, update_memory_size);
48c5303a
PC		 11256 u->type = update_reshape_migration;
11257 u->subdev = super->current_vol;
11258 u->new_level = geo->level;
11259 u->new_layout = geo->layout;
11260 u->new_raid_disks = u->old_raid_disks = geo->raid_disks;
11261 u->new_disks[0] = -1;
4bba0439 11262 u->new_chunksize = -1;
48c5303a
PC		 11263
11264 dev = get_imsm_dev(super, u->subdev);
11265 if (dev) {
11266 struct imsm_map *map;
11267
238c0a71 11268 map = get_imsm_map(dev, MAP_0);
4bba0439
PC		 11269 if (map) {
11270 int current_chunk_size =
11271 __le16_to_cpu(map->blocks_per_strip) / 2;
11272
11273 if (geo->chunksize != current_chunk_size) {
11274 u->new_chunksize = geo->chunksize / 1024;
7a862a02 11275 dprintf("imsm: chunk size change from %i to %i\n",
4bba0439
PC		 11276 current_chunk_size, u->new_chunksize);
11277 }
48c5303a 11278 previous_level = map->raid_level;
4bba0439 11279 }
48c5303a 11280 }
089f9d79 11281 if (geo->level == 5 && previous_level == 0) {
48c5303a
PC		 11282 struct mdinfo *spares = NULL;
11283
11284 u->new_raid_disks++;
11285 spares = get_spares_for_grow(st);
089f9d79 11286 if (spares == NULL || spares->array.spare_disks < 1) {
48c5303a
PC		 11287 free(u);
11288 sysfs_free(spares);
11289 update_memory_size = 0;
565cc99e 11290 pr_err("cannot get spare device for requested migration\n");
48c5303a
PC		 11291 return 0;
11292 }
11293 sysfs_free(spares);
11294 }
11295 dprintf("imsm: reshape update preparation : OK\n");
11296 *updatep = u;
11297
11298 return update_memory_size;
11299}
11300
8dd70bce
AK		 11301static void imsm_update_metadata_locally(struct supertype *st,
11302 void *buf, int len)
11303{
11304 struct metadata_update mu;
11305
11306 mu.buf = buf;
11307 mu.len = len;
11308 mu.space = NULL;
11309 mu.space_list = NULL;
11310 mu.next = NULL;
5fe6f031
N		 11311 if (imsm_prepare_update(st, &mu))
11312 imsm_process_update(st, &mu);
8dd70bce
AK		 11313
11314 while (mu.space_list) {
11315 void **space = mu.space_list;
11316 mu.space_list = *space;
11317 free(space);
11318 }
11319}
78b10e66 11320
471bceb6 11321/***************************************************************************
694575e7 11322* Function: imsm_analyze_change
471bceb6 11323* Description: Function analyze change for single volume
1011e834 11324* and validate if transition is supported
fbf3d202
AK		 11325* Parameters: Geometry parameters, supertype structure,
11326* metadata change direction (apply/rollback)
694575e7 11327* Returns: Operation type code on success, -1 if fail
471bceb6
KW		 11328****************************************************************************/
11329enum imsm_reshape_type imsm_analyze_change(struct supertype *st,
fbf3d202
AK		 11330 struct geo_params *geo,
11331 int direction)
694575e7 11332{
471bceb6
KW		 11333 struct mdinfo info;
11334 int change = -1;
11335 int check_devs = 0;
c21e737b 11336 int chunk;
67a2db32
AK		 11337 /* number of added/removed disks in operation result */
11338 int devNumChange = 0;
11339 /* imsm compatible layout value for array geometry verification */
11340 int imsm_layout = -1;
7abc9871
AK		 11341 int data_disks;
11342 struct imsm_dev *dev;
9529d343 11343 struct imsm_map *map;
7abc9871 11344 struct intel_super *super;
d04f65f4 11345 unsigned long long current_size;
65d38cca 11346 unsigned long long free_size;
d04f65f4 11347 unsigned long long max_size;
65d38cca 11348 int rv;
471bceb6
KW		 11349
11350 getinfo_super_imsm_volume(st, &info, NULL);
089f9d79
JS		 11351 if (geo->level != info.array.level && geo->level >= 0 &&
11352 geo->level != UnSet) {
471bceb6
KW		 11353 switch (info.array.level) {
11354 case 0:
11355 if (geo->level == 5) {
b5347799 11356 change = CH_MIGRATION;
e13ce846 11357 if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) {
7a862a02 11358 pr_err("Error. Requested Layout not supported (left-asymmetric layout is supported only)!\n");
e13ce846
AK		 11359 change = -1;
11360 goto analyse_change_exit;
11361 }
67a2db32 11362 imsm_layout = geo->layout;
471bceb6 11363 check_devs = 1;
e91a3bad
LM		 11364 devNumChange = 1; /* parity disk added */
11365 } else if (geo->level == 10) {
471bceb6
KW		 11366 change = CH_TAKEOVER;
11367 check_devs = 1;
e91a3bad 11368 devNumChange = 2; /* two mirrors added */
67a2db32 11369 imsm_layout = 0x102; /* imsm supported layout */
471bceb6 11370 }
dfe77a9e
KW		 11371 break;
11372 case 1:
471bceb6
KW		 11373 case 10:
11374 if (geo->level == 0) {
11375 change = CH_TAKEOVER;
11376 check_devs = 1;
e91a3bad 11377 devNumChange = -(geo->raid_disks/2);
67a2db32 11378 imsm_layout = 0; /* imsm raid0 layout */
471bceb6
KW		 11379 }
11380 break;
11381 }
11382 if (change == -1) {
7a862a02 11383 pr_err("Error. Level Migration from %d to %d not supported!\n",
e7b84f9d 11384 info.array.level, geo->level);
471bceb6
KW		 11385 goto analyse_change_exit;
11386 }
11387 } else
11388 geo->level = info.array.level;
11389
089f9d79
JS		 11390 if (geo->layout != info.array.layout &&
11391 (geo->layout != UnSet && geo->layout != -1)) {
b5347799 11392 change = CH_MIGRATION;
089f9d79
JS		 11393 if (info.array.layout == 0 && info.array.level == 5 &&
11394 geo->layout == 5) {
471bceb6 11395 /* reshape 5 -> 4 */
089f9d79
JS		 11396 } else if (info.array.layout == 5 && info.array.level == 5 &&
11397 geo->layout == 0) {
471bceb6
KW		 11398 /* reshape 4 -> 5 */
11399 geo->layout = 0;
11400 geo->level = 5;
11401 } else {
7a862a02 11402 pr_err("Error. Layout Migration from %d to %d not supported!\n",
e7b84f9d 11403 info.array.layout, geo->layout);
471bceb6
KW		 11404 change = -1;
11405 goto analyse_change_exit;
11406 }
67a2db32 11407 } else {
471bceb6 11408 geo->layout = info.array.layout;
67a2db32
AK		 11409 if (imsm_layout == -1)
11410 imsm_layout = info.array.layout;
11411 }
471bceb6 11412
089f9d79
JS		 11413 if (geo->chunksize > 0 && geo->chunksize != UnSet &&
11414 geo->chunksize != info.array.chunk_size) {
2d2b0eb7
MD		 11415 if (info.array.level == 10) {
11416 pr_err("Error. Chunk size change for RAID 10 is not supported.\n");
11417 change = -1;
11418 goto analyse_change_exit;
1e9b2c3f
PB		 11419 } else if (info.component_size % (geo->chunksize/512)) {
11420 pr_err("New chunk size (%dK) does not evenly divide device size (%lluk). Aborting...\n",
11421 geo->chunksize/1024, info.component_size/2);
11422 change = -1;
11423 goto analyse_change_exit;
2d2b0eb7 11424 }
b5347799 11425 change = CH_MIGRATION;
2d2b0eb7 11426 } else {
471bceb6 11427 geo->chunksize = info.array.chunk_size;
2d2b0eb7 11428 }
471bceb6 11429
c21e737b 11430 chunk = geo->chunksize / 1024;
7abc9871
AK		 11431
11432 super = st->sb;
11433 dev = get_imsm_dev(super, super->current_vol);
9529d343
MD		 11434 map = get_imsm_map(dev, MAP_0);
11435 data_disks = imsm_num_data_members(map);
c41e00b2 11436 /* compute current size per disk member
7abc9871 11437 */
c41e00b2
AK		 11438 current_size = info.custom_array_size / data_disks;
11439
089f9d79 11440 if (geo->size > 0 && geo->size != MAX_SIZE) {
c41e00b2
AK		 11441 /* align component size
11442 */
3e684231 11443 geo->size = imsm_component_size_alignment_check(
c41e00b2 11444 get_imsm_raid_level(dev->vol.map),
f36a9ecd 11445 chunk * 1024, super->sector_size,
c41e00b2 11446 geo->size * 2);
65d0b4ce 11447 if (geo->size == 0) {
7a862a02 11448 pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is 0).\n",
65d0b4ce
LD		 11449 current_size);
11450 goto analyse_change_exit;
11451 }
c41e00b2 11452 }
7abc9871 11453
089f9d79 11454 if (current_size != geo->size && geo->size > 0) {
7abc9871 11455 if (change != -1) {
7a862a02 11456 pr_err("Error. Size change should be the only one at a time.\n");
7abc9871
AK		 11457 change = -1;
11458 goto analyse_change_exit;
11459 }
11460 if ((super->current_vol + 1) != super->anchor->num_raid_devs) {
7a862a02 11461 pr_err("Error. The last volume in container can be expanded only (%i/%s).\n",
4dd2df09 11462 super->current_vol, st->devnm);
7abc9871
AK		 11463 goto analyse_change_exit;
11464 }
65d38cca
LD		 11465 /* check the maximum available size
11466 */
11467 rv = imsm_get_free_size(st, dev->vol.map->num_members,
11468 0, chunk, &free_size);
11469 if (rv == 0)
11470 /* Cannot find maximum available space
11471 */
11472 max_size = 0;
11473 else {
11474 max_size = free_size + current_size;
11475 /* align component size
11476 */
3e684231 11477 max_size = imsm_component_size_alignment_check(
65d38cca 11478 get_imsm_raid_level(dev->vol.map),
f36a9ecd 11479 chunk * 1024, super->sector_size,
65d38cca
LD		 11480 max_size);
11481 }
d04f65f4 11482 if (geo->size == MAX_SIZE) {
b130333f
AK		 11483 /* requested size change to the maximum available size
11484 */
65d38cca 11485 if (max_size == 0) {
7a862a02 11486 pr_err("Error. Cannot find maximum available space.\n");
b130333f
AK		 11487 change = -1;
11488 goto analyse_change_exit;
65d38cca
LD		 11489 } else
11490 geo->size = max_size;
c41e00b2 11491 }
b130333f 11492
681b7ae2 11493 if (direction == ROLLBACK_METADATA_CHANGES) {
fbf3d202
AK		 11494 /* accept size for rollback only
11495 */
11496 } else {
11497 /* round size due to metadata compatibility
11498 */
11499 geo->size = (geo->size >> SECT_PER_MB_SHIFT)
11500 << SECT_PER_MB_SHIFT;
11501 dprintf("Prepare update for size change to %llu\n",
11502 geo->size );
11503 if (current_size >= geo->size) {
7a862a02 11504 pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is %llu).\n",
e7b84f9d 11505 current_size, geo->size);
fbf3d202
AK		 11506 goto analyse_change_exit;
11507 }
65d38cca 11508 if (max_size && geo->size > max_size) {
7a862a02 11509 pr_err("Error. Requested size is larger than maximum available size (maximum available size is %llu, requested size /rounded/ is %llu).\n",
e7b84f9d 11510 max_size, geo->size);
65d38cca
LD		 11511 goto analyse_change_exit;
11512 }
7abc9871
AK		 11513 }
11514 geo->size *= data_disks;
11515 geo->raid_disks = dev->vol.map->num_members;
11516 change = CH_ARRAY_SIZE;
11517 }
471bceb6
KW		 11518 if (!validate_geometry_imsm(st,
11519 geo->level,
67a2db32 11520 imsm_layout,
e91a3bad 11521 geo->raid_disks + devNumChange,
c21e737b 11522 &chunk,
af4348dd 11523 geo->size, INVALID_SECTORS,
5308f117 11524 0, 0, info.consistency_policy, 1))
471bceb6
KW		 11525 change = -1;
11526
11527 if (check_devs) {
11528 struct intel_super *super = st->sb;
11529 struct imsm_super *mpb = super->anchor;
11530
11531 if (mpb->num_raid_devs > 1) {
7a862a02 11532 pr_err("Error. Cannot perform operation on %s- for this operation it MUST be single array in container\n",
e7b84f9d 11533 geo->dev_name);
471bceb6
KW		 11534 change = -1;
11535 }
11536 }
11537
11538analyse_change_exit:
089f9d79
JS		 11539 if (direction == ROLLBACK_METADATA_CHANGES &&
11540 (change == CH_MIGRATION || change == CH_TAKEOVER)) {
7a862a02 11541 dprintf("imsm: Metadata changes rollback is not supported for migration and takeover operations.\n");
fbf3d202
AK		 11542 change = -1;
11543 }
471bceb6 11544 return change;
694575e7
KW		 11545}
11546
bb025c2f
KW		 11547int imsm_takeover(struct supertype *st, struct geo_params *geo)
11548{
11549 struct intel_super *super = st->sb;
11550 struct imsm_update_takeover *u;
11551
503975b9 11552 u = xmalloc(sizeof(struct imsm_update_takeover));
bb025c2f
KW		 11553
11554 u->type = update_takeover;
11555 u->subarray = super->current_vol;
11556
11557 /* 10->0 transition */
11558 if (geo->level == 0)
11559 u->direction = R10_TO_R0;
11560
0529c688
KW		 11561 /* 0->10 transition */
11562 if (geo->level == 10)
11563 u->direction = R0_TO_R10;
11564
bb025c2f
KW		 11565 /* update metadata locally */
11566 imsm_update_metadata_locally(st, u,
11567 sizeof(struct imsm_update_takeover));
11568 /* and possibly remotely */
11569 if (st->update_tail)
11570 append_metadata_update(st, u,
11571 sizeof(struct imsm_update_takeover));
11572 else
11573 free(u);
11574
11575 return 0;
11576}
11577
d04f65f4
N		 11578static int imsm_reshape_super(struct supertype *st, unsigned long long size,
11579 int level,
78b10e66 11580 int layout, int chunksize, int raid_disks,
41784c88 11581 int delta_disks, char *backup, char *dev,
016e00f5 11582 int direction, int verbose)
78b10e66 11583{
78b10e66
N		 11584 int ret_val = 1;
11585 struct geo_params geo;
11586
1ade5cc1 11587 dprintf("(enter)\n");
78b10e66 11588
71204a50 11589 memset(&geo, 0, sizeof(struct geo_params));
78b10e66
N		 11590
11591 geo.dev_name = dev;
4dd2df09 11592 strcpy(geo.devnm, st->devnm);
78b10e66
N		 11593 geo.size = size;
11594 geo.level = level;
11595 geo.layout = layout;
11596 geo.chunksize = chunksize;
11597 geo.raid_disks = raid_disks;
41784c88
AK		 11598 if (delta_disks != UnSet)
11599 geo.raid_disks += delta_disks;
78b10e66 11600
1ade5cc1
N		 11601 dprintf("for level : %i\n", geo.level);
11602 dprintf("for raid_disks : %i\n", geo.raid_disks);
78b10e66
N		 11603
11604 if (experimental() == 0)
11605 return ret_val;
11606
4dd2df09 11607 if (strcmp(st->container_devnm, st->devnm) == 0) {
694575e7
KW		 11608 /* On container level we can only increase number of devices. */
11609 dprintf("imsm: info: Container operation\n");
78b10e66 11610 int old_raid_disks = 0;
6dc0be30 11611
78b10e66 11612 if (imsm_reshape_is_allowed_on_container(
fbf3d202 11613 st, &geo, &old_raid_disks, direction)) {
78b10e66
N		 11614 struct imsm_update_reshape *u = NULL;
11615 int len;
11616
11617 len = imsm_create_metadata_update_for_reshape(
11618 st, &geo, old_raid_disks, &u);
11619
ed08d51c
AK		 11620 if (len <= 0) {
11621 dprintf("imsm: Cannot prepare update\n");
11622 goto exit_imsm_reshape_super;
11623 }
11624
8dd70bce
AK		 11625 ret_val = 0;
11626 /* update metadata locally */
11627 imsm_update_metadata_locally(st, u, len);
11628 /* and possibly remotely */
11629 if (st->update_tail)
11630 append_metadata_update(st, u, len);
11631 else
ed08d51c 11632 free(u);
8dd70bce 11633
694575e7 11634 } else {
7a862a02 11635 pr_err("(imsm) Operation is not allowed on this container\n");
694575e7
KW		 11636 }
11637 } else {
11638 /* On volume level we support following operations
471bceb6
KW		 11639 * - takeover: raid10 -> raid0; raid0 -> raid10
11640 * - chunk size migration
11641 * - migration: raid5 -> raid0; raid0 -> raid5
11642 */
11643 struct intel_super *super = st->sb;
11644 struct intel_dev *dev = super->devlist;
4dd2df09 11645 int change;
694575e7 11646 dprintf("imsm: info: Volume operation\n");
471bceb6
KW		 11647 /* find requested device */
11648 while (dev) {
1011e834 11649 char *devnm =
4dd2df09
N		 11650 imsm_find_array_devnm_by_subdev(
11651 dev->index, st->container_devnm);
11652 if (devnm && strcmp(devnm, geo.devnm) == 0)
471bceb6
KW		 11653 break;
11654 dev = dev->next;
11655 }
11656 if (dev == NULL) {
4dd2df09
N		 11657 pr_err("Cannot find %s (%s) subarray\n",
11658 geo.dev_name, geo.devnm);
471bceb6
KW		 11659 goto exit_imsm_reshape_super;
11660 }
11661 super->current_vol = dev->index;
fbf3d202 11662 change = imsm_analyze_change(st, &geo, direction);
694575e7 11663 switch (change) {
471bceb6 11664 case CH_TAKEOVER:
bb025c2f 11665 ret_val = imsm_takeover(st, &geo);
694575e7 11666 break;
48c5303a
PC		 11667 case CH_MIGRATION: {
11668 struct imsm_update_reshape_migration *u = NULL;
11669 int len =
11670 imsm_create_metadata_update_for_migration(
11671 st, &geo, &u);
11672 if (len < 1) {
7a862a02 11673 dprintf("imsm: Cannot prepare update\n");
48c5303a
PC		 11674 break;
11675 }
471bceb6 11676 ret_val = 0;
48c5303a
PC		 11677 /* update metadata locally */
11678 imsm_update_metadata_locally(st, u, len);
11679 /* and possibly remotely */
11680 if (st->update_tail)
11681 append_metadata_update(st, u, len);
11682 else
11683 free(u);
11684 }
11685 break;
7abc9871 11686 case CH_ARRAY_SIZE: {
f3871fdc
AK		 11687 struct imsm_update_size_change *u = NULL;
11688 int len =
11689 imsm_create_metadata_update_for_size_change(
11690 st, &geo, &u);
11691 if (len < 1) {
7a862a02 11692 dprintf("imsm: Cannot prepare update\n");
f3871fdc
AK		 11693 break;
11694 }
11695 ret_val = 0;
11696 /* update metadata locally */
11697 imsm_update_metadata_locally(st, u, len);
11698 /* and possibly remotely */
11699 if (st->update_tail)
11700 append_metadata_update(st, u, len);
11701 else
11702 free(u);
7abc9871
AK		 11703 }
11704 break;
471bceb6
KW		 11705 default:
11706 ret_val = 1;
694575e7 11707 }
694575e7 11708 }
78b10e66 11709
ed08d51c 11710exit_imsm_reshape_super:
78b10e66
N		 11711 dprintf("imsm: reshape_super Exit code = %i\n", ret_val);
11712 return ret_val;
11713}
2cda7640 11714
0febb20c
AO		 11715#define COMPLETED_OK 0
11716#define COMPLETED_NONE 1
11717#define COMPLETED_DELAYED 2
11718
11719static int read_completed(int fd, unsigned long long *val)
11720{
11721 int ret;
11722 char buf[50];
11723
11724 ret = sysfs_fd_get_str(fd, buf, 50);
11725 if (ret < 0)
11726 return ret;
11727
11728 ret = COMPLETED_OK;
11729 if (strncmp(buf, "none", 4) == 0) {
11730 ret = COMPLETED_NONE;
11731 } else if (strncmp(buf, "delayed", 7) == 0) {
11732 ret = COMPLETED_DELAYED;
11733 } else {
11734 char *ep;
11735 *val = strtoull(buf, &ep, 0);
11736 if (ep == buf || (*ep != 0 && *ep != '\n' && *ep != ' '))
11737 ret = -1;
11738 }
11739 return ret;
11740}
11741
eee67a47
AK		 11742/*******************************************************************************
11743 * Function: wait_for_reshape_imsm
11744 * Description: Function writes new sync_max value and waits until
11745 * reshape process reach new position
11746 * Parameters:
11747 * sra : general array info
eee67a47
AK		 11748 * ndata : number of disks in new array's layout
11749 * Returns:
11750 * 0 : success,
11751 * 1 : there is no reshape in progress,
11752 * -1 : fail
11753 ******************************************************************************/
ae9f01f8 11754int wait_for_reshape_imsm(struct mdinfo *sra, int ndata)
eee67a47 11755{
85ca499c 11756 int fd = sysfs_get_fd(sra, NULL, "sync_completed");
df2647fa 11757 int retry = 3;
eee67a47 11758 unsigned long long completed;
ae9f01f8
AK		 11759 /* to_complete : new sync_max position */
11760 unsigned long long to_complete = sra->reshape_progress;
11761 unsigned long long position_to_set = to_complete / ndata;
eee67a47 11762
ae9f01f8 11763 if (fd < 0) {
1ade5cc1 11764 dprintf("cannot open reshape_position\n");
eee67a47 11765 return 1;
ae9f01f8 11766 }
eee67a47 11767
df2647fa
PB		 11768 do {
11769 if (sysfs_fd_get_ll(fd, &completed) < 0) {
11770 if (!retry) {
11771 dprintf("cannot read reshape_position (no reshape in progres)\n");
11772 close(fd);
11773 return 1;
11774 }
11775 usleep(30000);
11776 } else
11777 break;
11778 } while (retry--);
eee67a47 11779
85ca499c 11780 if (completed > position_to_set) {
1ade5cc1 11781 dprintf("wrong next position to set %llu (%llu)\n",
85ca499c 11782 to_complete, position_to_set);
ae9f01f8
AK		 11783 close(fd);
11784 return -1;
11785 }
11786 dprintf("Position set: %llu\n", position_to_set);
11787 if (sysfs_set_num(sra, NULL, "sync_max",
11788 position_to_set) != 0) {
1ade5cc1 11789 dprintf("cannot set reshape position to %llu\n",
ae9f01f8
AK		 11790 position_to_set);
11791 close(fd);
11792 return -1;
eee67a47
AK		 11793 }
11794
eee67a47 11795 do {
0febb20c 11796 int rc;
eee67a47 11797 char action[20];
5ff3a780 11798 int timeout = 3000;
0febb20c 11799
5ff3a780 11800 sysfs_wait(fd, &timeout);
a47e44fb
AK		 11801 if (sysfs_get_str(sra, NULL, "sync_action",
11802 action, 20) > 0 &&
d7d3809a 11803 strncmp(action, "reshape", 7) != 0) {
b2be2b62
AO		 11804 if (strncmp(action, "idle", 4) == 0)
11805 break;
d7d3809a
AP		 11806 close(fd);
11807 return -1;
11808 }
0febb20c
AO		 11809
11810 rc = read_completed(fd, &completed);
11811 if (rc < 0) {
1ade5cc1 11812 dprintf("cannot read reshape_position (in loop)\n");
eee67a47
AK		 11813 close(fd);
11814 return 1;
0febb20c
AO		 11815 } else if (rc == COMPLETED_NONE)
11816 break;
85ca499c 11817 } while (completed < position_to_set);
b2be2b62 11818
eee67a47
AK		 11819 close(fd);
11820 return 0;
eee67a47
AK		 11821}
11822
b915c95f
AK		 11823/*******************************************************************************
11824 * Function: check_degradation_change
11825 * Description: Check that array hasn't become failed.
11826 * Parameters:
11827 * info : for sysfs access
11828 * sources : source disks descriptors
11829 * degraded: previous degradation level
11830 * Returns:
11831 * degradation level
11832 ******************************************************************************/
11833int check_degradation_change(struct mdinfo *info,
11834 int *sources,
11835 int degraded)
11836{
11837 unsigned long long new_degraded;
e1993023
LD		 11838 int rv;
11839
11840 rv = sysfs_get_ll(info, NULL, "degraded", &new_degraded);
089f9d79 11841 if (rv == -1 || (new_degraded != (unsigned long long)degraded)) {
b915c95f
AK		 11842 /* check each device to ensure it is still working */
11843 struct mdinfo *sd;
11844 new_degraded = 0;
11845 for (sd = info->devs ; sd ; sd = sd->next) {
11846 if (sd->disk.state & (1<<MD_DISK_FAULTY))
11847 continue;
11848 if (sd->disk.state & (1<<MD_DISK_SYNC)) {
cf52eff5
TM		 11849 char sbuf[100];
11850
b915c95f 11851 if (sysfs_get_str(info,
cf52eff5 11852 sd, "state", sbuf, sizeof(sbuf)) < 0 ||
b915c95f
AK		 11853 strstr(sbuf, "faulty") ||
11854 strstr(sbuf, "in_sync") == NULL) {
11855 /* this device is dead */
11856 sd->disk.state = (1<<MD_DISK_FAULTY);
11857 if (sd->disk.raid_disk >= 0 &&
11858 sources[sd->disk.raid_disk] >= 0) {
11859 close(sources[
11860 sd->disk.raid_disk]);
11861 sources[sd->disk.raid_disk] =
11862 -1;
11863 }
11864 new_degraded++;
11865 }
11866 }
11867 }
11868 }
11869
11870 return new_degraded;
11871}
11872
10f22854
AK		 11873/*******************************************************************************
11874 * Function: imsm_manage_reshape
11875 * Description: Function finds array under reshape and it manages reshape
11876 * process. It creates stripes backups (if required) and sets
942e1cdb 11877 * checkpoints.
10f22854
AK		 11878 * Parameters:
11879 * afd : Backup handle (nattive) - not used
11880 * sra : general array info
11881 * reshape : reshape parameters - not used
11882 * st : supertype structure
11883 * blocks : size of critical section [blocks]
11884 * fds : table of source device descriptor
11885 * offsets : start of array (offest per devices)
11886 * dests : not used
11887 * destfd : table of destination device descriptor
11888 * destoffsets : table of destination offsets (per device)
11889 * Returns:
11890 * 1 : success, reshape is done
11891 * 0 : fail
11892 ******************************************************************************/
999b4972
N		 11893static int imsm_manage_reshape(
11894 int afd, struct mdinfo *sra, struct reshape *reshape,
10f22854 11895 struct supertype *st, unsigned long backup_blocks,
999b4972
N		 11896 int *fds, unsigned long long *offsets,
11897 int dests, int *destfd, unsigned long long *destoffsets)
11898{
10f22854
AK		 11899 int ret_val = 0;
11900 struct intel_super *super = st->sb;
594dc1b8 11901 struct intel_dev *dv;
de44e46f 11902 unsigned int sector_size = super->sector_size;
10f22854 11903 struct imsm_dev *dev = NULL;
9529d343 11904 struct imsm_map *map_src, *map_dest;
10f22854
AK		 11905 int migr_vol_qan = 0;
11906 int ndata, odata; /* [bytes] */
11907 int chunk; /* [bytes] */
11908 struct migr_record *migr_rec;
11909 char *buf = NULL;
11910 unsigned int buf_size; /* [bytes] */
11911 unsigned long long max_position; /* array size [bytes] */
11912 unsigned long long next_step; /* [blocks]/[bytes] */
11913 unsigned long long old_data_stripe_length;
10f22854
AK		 11914 unsigned long long start_src; /* [bytes] */
11915 unsigned long long start; /* [bytes] */
11916 unsigned long long start_buf_shift; /* [bytes] */
b915c95f 11917 int degraded = 0;
ab724b98 11918 int source_layout = 0;
10f22854 11919
79a16a9b
JS		 11920 if (!sra)
11921 return ret_val;
11922
11923 if (!fds || !offsets)
10f22854
AK		 11924 goto abort;
11925
11926 /* Find volume during the reshape */
11927 for (dv = super->devlist; dv; dv = dv->next) {
fc54fe7a
JS		 11928 if (dv->dev->vol.migr_type == MIGR_GEN_MIGR &&
11929 dv->dev->vol.migr_state == 1) {
10f22854
AK		 11930 dev = dv->dev;
11931 migr_vol_qan++;
11932 }
11933 }
11934 /* Only one volume can migrate at the same time */
11935 if (migr_vol_qan != 1) {
676e87a8 11936 pr_err("%s", migr_vol_qan ?
10f22854
AK		 11937 "Number of migrating volumes greater than 1\n" :
11938 "There is no volume during migrationg\n");
11939 goto abort;
11940 }
11941
9529d343 11942 map_dest = get_imsm_map(dev, MAP_0);
238c0a71 11943 map_src = get_imsm_map(dev, MAP_1);
10f22854
AK		 11944 if (map_src == NULL)
11945 goto abort;
10f22854 11946
9529d343
MD		 11947 ndata = imsm_num_data_members(map_dest);
11948 odata = imsm_num_data_members(map_src);
10f22854 11949
7b1ab482 11950 chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512;
10f22854
AK		 11951 old_data_stripe_length = odata * chunk;
11952
11953 migr_rec = super->migr_rec;
11954
10f22854
AK		 11955 /* initialize migration record for start condition */
11956 if (sra->reshape_progress == 0)
11957 init_migr_record_imsm(st, dev, sra);
b2c59438
AK		 11958 else {
11959 if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) {
7a862a02 11960 dprintf("imsm: cannot restart migration when data are present in copy area.\n");
b2c59438
AK		 11961 goto abort;
11962 }
6a75c8ca
AK		 11963 /* Save checkpoint to update migration record for current
11964 * reshape position (in md). It can be farther than current
11965 * reshape position in metadata.
11966 */
11967 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
11968 /* ignore error == 2, this can mean end of reshape here
11969 */
7a862a02 11970 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL, initial save)\n");
6a75c8ca
AK		 11971 goto abort;
11972 }
b2c59438 11973 }
10f22854
AK		 11974
11975 /* size for data */
11976 buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512;
11977 /* extend buffer size for parity disk */
11978 buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
3e684231 11979 /* add space for stripe alignment */
10f22854 11980 buf_size += old_data_stripe_length;
de44e46f
PB		 11981 if (posix_memalign((void **)&buf, MAX_SECTOR_SIZE, buf_size)) {
11982 dprintf("imsm: Cannot allocate checkpoint buffer\n");
10f22854
AK		 11983 goto abort;
11984 }
11985
3ef4403c 11986 max_position = sra->component_size * ndata;
68eb8bc6 11987 source_layout = imsm_level_to_layout(map_src->raid_level);
10f22854
AK		 11988
11989 while (__le32_to_cpu(migr_rec->curr_migr_unit) <
11990 __le32_to_cpu(migr_rec->num_migr_units)) {
11991 /* current reshape position [blocks] */
11992 unsigned long long current_position =
11993 __le32_to_cpu(migr_rec->blocks_per_unit)
11994 * __le32_to_cpu(migr_rec->curr_migr_unit);
11995 unsigned long long border;
11996
b915c95f
AK		 11997 /* Check that array hasn't become failed.
11998 */
11999 degraded = check_degradation_change(sra, fds, degraded);
12000 if (degraded > 1) {
7a862a02 12001 dprintf("imsm: Abort reshape due to degradation level (%i)\n", degraded);
b915c95f
AK		 12002 goto abort;
12003 }
12004
10f22854
AK		 12005 next_step = __le32_to_cpu(migr_rec->blocks_per_unit);
12006
12007 if ((current_position + next_step) > max_position)
12008 next_step = max_position - current_position;
12009
92144abf 12010 start = current_position * 512;
10f22854 12011
942e1cdb 12012 /* align reading start to old geometry */
10f22854
AK		 12013 start_buf_shift = start % old_data_stripe_length;
12014 start_src = start - start_buf_shift;
12015
12016 border = (start_src / odata) - (start / ndata);
12017 border /= 512;
12018 if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) {
12019 /* save critical stripes to buf
12020 * start - start address of current unit
12021 * to backup [bytes]
12022 * start_src - start address of current unit
12023 * to backup alligned to source array
12024 * [bytes]
12025 */
594dc1b8 12026 unsigned long long next_step_filler;
10f22854
AK		 12027 unsigned long long copy_length = next_step * 512;
12028
12029 /* allign copy area length to stripe in old geometry */
12030 next_step_filler = ((copy_length + start_buf_shift)
12031 % old_data_stripe_length);
12032 if (next_step_filler)
12033 next_step_filler = (old_data_stripe_length
12034 - next_step_filler);
7a862a02 12035 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		 12036 start, start_src, copy_length,
12037 start_buf_shift, next_step_filler);
12038
12039 if (save_stripes(fds, offsets, map_src->num_members,
ab724b98
AK		 12040 chunk, map_src->raid_level,
12041 source_layout, 0, NULL, start_src,
10f22854
AK		 12042 copy_length +
12043 next_step_filler + start_buf_shift,
12044 buf)) {
7a862a02 12045 dprintf("imsm: Cannot save stripes to buffer\n");
10f22854
AK		 12046 goto abort;
12047 }
12048 /* Convert data to destination format and store it
12049 * in backup general migration area
12050 */
12051 if (save_backup_imsm(st, dev, sra,
aea93171 12052 buf + start_buf_shift, copy_length)) {
7a862a02 12053 dprintf("imsm: Cannot save stripes to target devices\n");
10f22854
AK		 12054 goto abort;
12055 }
12056 if (save_checkpoint_imsm(st, sra,
12057 UNIT_SRC_IN_CP_AREA)) {
7a862a02 12058 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_IN_CP_AREA)\n");
10f22854
AK		 12059 goto abort;
12060 }
8016a6d4
AK		 12061 } else {
12062 /* set next step to use whole border area */
12063 border /= next_step;
12064 if (border > 1)
12065 next_step *= border;
10f22854
AK		 12066 }
12067 /* When data backed up, checkpoint stored,
12068 * kick the kernel to reshape unit of data
12069 */
12070 next_step = next_step + sra->reshape_progress;
8016a6d4
AK		 12071 /* limit next step to array max position */
12072 if (next_step > max_position)
12073 next_step = max_position;
10f22854
AK		 12074 sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress);
12075 sysfs_set_num(sra, NULL, "suspend_hi", next_step);
ae9f01f8 12076 sra->reshape_progress = next_step;
10f22854
AK		 12077
12078 /* wait until reshape finish */
c85338c6 12079 if (wait_for_reshape_imsm(sra, ndata)) {
c47b0ff6
AK		 12080 dprintf("wait_for_reshape_imsm returned error!\n");
12081 goto abort;
12082 }
84d11e6c
N		 12083 if (sigterm)
12084 goto abort;
10f22854 12085
0228d92c
AK		 12086 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
12087 /* ignore error == 2, this can mean end of reshape here
12088 */
7a862a02 12089 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL)\n");
10f22854
AK		 12090 goto abort;
12091 }
12092
12093 }
12094
71e5411e
PB		 12095 /* clear migr_rec on disks after successful migration */
12096 struct dl *d;
12097
85337573 12098 memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
71e5411e
PB		 12099 for (d = super->disks; d; d = d->next) {
12100 if (d->index < 0 || is_failed(&d->disk))
12101 continue;
12102 unsigned long long dsize;
12103
12104 get_dev_size(d->fd, NULL, &dsize);
de44e46f 12105 if (lseek64(d->fd, dsize - MIGR_REC_SECTOR_POSITION*sector_size,
71e5411e 12106 SEEK_SET) >= 0) {
466070ad 12107 if ((unsigned int)write(d->fd, super->migr_rec_buf,
de44e46f
PB		 12108 MIGR_REC_BUF_SECTORS*sector_size) !=
12109 MIGR_REC_BUF_SECTORS*sector_size)
71e5411e
PB		 12110 perror("Write migr_rec failed");
12111 }
12112 }
12113
10f22854
AK		 12114 /* return '1' if done */
12115 ret_val = 1;
12116abort:
12117 free(buf);
942e1cdb
N		 12118 /* See Grow.c: abort_reshape() for further explanation */
12119 sysfs_set_num(sra, NULL, "suspend_lo", 0x7FFFFFFFFFFFFFFFULL);
12120 sysfs_set_num(sra, NULL, "suspend_hi", 0);
12121 sysfs_set_num(sra, NULL, "suspend_lo", 0);
10f22854
AK		 12122
12123 return ret_val;
999b4972 12124}
0c21b485 12125
cdddbdbc 12126struct superswitch super_imsm = {
cdddbdbc
DW		 12127 .examine_super = examine_super_imsm,
12128 .brief_examine_super = brief_examine_super_imsm,
4737ae25 12129 .brief_examine_subarrays = brief_examine_subarrays_imsm,
9d84c8ea 12130 .export_examine_super = export_examine_super_imsm,
cdddbdbc
DW		 12131 .detail_super = detail_super_imsm,
12132 .brief_detail_super = brief_detail_super_imsm,
bf5a934a 12133 .write_init_super = write_init_super_imsm,
0e600426
N		 12134 .validate_geometry = validate_geometry_imsm,
12135 .add_to_super = add_to_super_imsm,
1a64be56 12136 .remove_from_super = remove_from_super_imsm,
d665cc31 12137 .detail_platform = detail_platform_imsm,
e50cf220 12138 .export_detail_platform = export_detail_platform_imsm,
33414a01 12139 .kill_subarray = kill_subarray_imsm,
aa534678 12140 .update_subarray = update_subarray_imsm,
2b959fbf 12141 .load_container = load_container_imsm,
71204a50
N		 12142 .default_geometry = default_geometry_imsm,
12143 .get_disk_controller_domain = imsm_get_disk_controller_domain,
12144 .reshape_super = imsm_reshape_super,
12145 .manage_reshape = imsm_manage_reshape,
9e2d750d 12146 .recover_backup = recover_backup_imsm,
74db60b0 12147 .copy_metadata = copy_metadata_imsm,
27156a57 12148 .examine_badblocks = examine_badblocks_imsm,
cdddbdbc
DW		 12149 .match_home = match_home_imsm,
12150 .uuid_from_super= uuid_from_super_imsm,
12151 .getinfo_super = getinfo_super_imsm,
5c4cd5da 12152 .getinfo_super_disks = getinfo_super_disks_imsm,
cdddbdbc
DW		 12153 .update_super = update_super_imsm,
12154
12155 .avail_size = avail_size_imsm,
fbfdcb06 12156 .get_spare_criteria = get_spare_criteria_imsm,
cdddbdbc
DW		 12157
12158 .compare_super = compare_super_imsm,
12159
12160 .load_super = load_super_imsm,
bf5a934a 12161 .init_super = init_super_imsm,
e683ca88 12162 .store_super = store_super_imsm,
cdddbdbc
DW		 12163 .free_super = free_super_imsm,
12164 .match_metadata_desc = match_metadata_desc_imsm,
bf5a934a 12165 .container_content = container_content_imsm,
0c21b485 12166 .validate_container = validate_container_imsm,
cdddbdbc 12167
2432ce9b
AP		 12168 .write_init_ppl = write_init_ppl_imsm,
12169 .validate_ppl = validate_ppl_imsm,
12170
cdddbdbc 12171 .external = 1,
4cce4069 12172 .name = "imsm",
845dea95
NB		 12173
12174/* for mdmon */
12175 .open_new = imsm_open_new,
ed9d66aa 12176 .set_array_state= imsm_set_array_state,
845dea95
NB		 12177 .set_disk = imsm_set_disk,
12178 .sync_metadata = imsm_sync_metadata,
88758e9d 12179 .activate_spare = imsm_activate_spare,
e8319a19 12180 .process_update = imsm_process_update,
8273f55e 12181 .prepare_update = imsm_prepare_update,
6f50473f 12182 .record_bad_block = imsm_record_badblock,
c07a5a4f 12183 .clear_bad_block = imsm_clear_badblock,
928f1424 12184 .get_bad_blocks = imsm_get_badblocks,
cdddbdbc 12185};
Unnamed repository; edit this file 'description' to name the repository.