]> 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
Enable bitmap support for external metadata
[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 98
761e3bd9
N		 99/*
100 * This macro let's us ensure that no-one accidentally
101 * changes the size of a struct
102 */
103#define ASSERT_SIZE(_struct, size) \
104static inline void __assert_size_##_struct(void) \
105{ \
106 switch (0) { \
107 case 0: break; \
108 case (sizeof(struct _struct) == size): break; \
109 } \
110}
111
cdddbdbc
DW		 112/* Disk configuration info. */
113#define IMSM_MAX_DEVICES 255
114struct imsm_disk {
115 __u8 serial[MAX_RAID_SERIAL_LEN];/* 0xD8 - 0xE7 ascii serial number */
5551b113 116 __u32 total_blocks_lo; /* 0xE8 - 0xEB total blocks lo */
cdddbdbc 117 __u32 scsi_id; /* 0xEC - 0xEF scsi ID */
f2f27e63
DW		 118#define SPARE_DISK __cpu_to_le32(0x01) /* Spare */
119#define CONFIGURED_DISK __cpu_to_le32(0x02) /* Member of some RaidDev */
120#define FAILED_DISK __cpu_to_le32(0x04) /* Permanent failure */
2432ce9b 121#define JOURNAL_DISK __cpu_to_le32(0x2000000) /* Device marked as Journaling Drive */
cdddbdbc 122 __u32 status; /* 0xF0 - 0xF3 */
1011e834 123 __u32 owner_cfg_num; /* which config 0,1,2... owns this disk */
5551b113
CA		 124 __u32 total_blocks_hi; /* 0xF4 - 0xF5 total blocks hi */
125#define IMSM_DISK_FILLERS 3
126 __u32 filler[IMSM_DISK_FILLERS]; /* 0xF5 - 0x107 MPB_DISK_FILLERS for future expansion */
cdddbdbc 127};
761e3bd9 128ASSERT_SIZE(imsm_disk, 48)
cdddbdbc 129
3b451610
AK		 130/* map selector for map managment
131 */
238c0a71
AK		 132#define MAP_0 0
133#define MAP_1 1
134#define MAP_X -1
3b451610 135
cdddbdbc
DW		 136/* RAID map configuration infos. */
137struct imsm_map {
5551b113
CA		 138 __u32 pba_of_lba0_lo; /* start address of partition */
139 __u32 blocks_per_member_lo;/* blocks per member */
140 __u32 num_data_stripes_lo; /* number of data stripes */
cdddbdbc
DW		 141 __u16 blocks_per_strip;
142 __u8 map_state; /* Normal, Uninitialized, Degraded, Failed */
143#define IMSM_T_STATE_NORMAL 0
144#define IMSM_T_STATE_UNINITIALIZED 1
e3bba0e0
DW		 145#define IMSM_T_STATE_DEGRADED 2
146#define IMSM_T_STATE_FAILED 3
cdddbdbc
DW		 147 __u8 raid_level;
148#define IMSM_T_RAID0 0
149#define IMSM_T_RAID1 1
150#define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */
151 __u8 num_members; /* number of member disks */
fe7ed8cb
DW		 152 __u8 num_domains; /* number of parity domains */
153 __u8 failed_disk_num; /* valid only when state is degraded */
252d23c0 154 __u8 ddf;
5551b113
CA		 155 __u32 pba_of_lba0_hi;
156 __u32 blocks_per_member_hi;
157 __u32 num_data_stripes_hi;
158 __u32 filler[4]; /* expansion area */
7eef0453 159#define IMSM_ORD_REBUILD (1 << 24)
cdddbdbc 160 __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members],
7eef0453
DW		 161 * top byte contains some flags
162 */
761e3bd9
N		 163};
164ASSERT_SIZE(imsm_map, 52)
cdddbdbc
DW		 165
166struct imsm_vol {
f8f603f1 167 __u32 curr_migr_unit;
fe7ed8cb 168 __u32 checkpoint_id; /* id to access curr_migr_unit */
cdddbdbc 169 __u8 migr_state; /* Normal or Migrating */
e3bba0e0
DW		 170#define MIGR_INIT 0
171#define MIGR_REBUILD 1
172#define MIGR_VERIFY 2 /* analagous to echo check > sync_action */
173#define MIGR_GEN_MIGR 3
174#define MIGR_STATE_CHANGE 4
1484e727 175#define MIGR_REPAIR 5
cdddbdbc 176 __u8 migr_type; /* Initializing, Rebuilding, ... */
2432ce9b
AP		 177#define RAIDVOL_CLEAN 0
178#define RAIDVOL_DIRTY 1
179#define RAIDVOL_DSRECORD_VALID 2
cdddbdbc 180 __u8 dirty;
fe7ed8cb
DW		 181 __u8 fs_state; /* fast-sync state for CnG (0xff == disabled) */
182 __u16 verify_errors; /* number of mismatches */
183 __u16 bad_blocks; /* number of bad blocks during verify */
184 __u32 filler[4];
cdddbdbc
DW		 185 struct imsm_map map[1];
186 /* here comes another one if migr_state */
761e3bd9
N		 187};
188ASSERT_SIZE(imsm_vol, 84)
cdddbdbc
DW		 189
190struct imsm_dev {
fe7ed8cb 191 __u8 volume[MAX_RAID_SERIAL_LEN];
cdddbdbc
DW		 192 __u32 size_low;
193 __u32 size_high;
fe7ed8cb
DW		 194#define DEV_BOOTABLE __cpu_to_le32(0x01)
195#define DEV_BOOT_DEVICE __cpu_to_le32(0x02)
196#define DEV_READ_COALESCING __cpu_to_le32(0x04)
197#define DEV_WRITE_COALESCING __cpu_to_le32(0x08)
198#define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10)
199#define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20)
200#define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40)
201#define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80)
202#define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100)
203#define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200)
204#define DEV_CLONE_N_GO __cpu_to_le32(0x400)
205#define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800)
206#define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000)
cdddbdbc
DW		 207 __u32 status; /* Persistent RaidDev status */
208 __u32 reserved_blocks; /* Reserved blocks at beginning of volume */
fe7ed8cb
DW		 209 __u8 migr_priority;
210 __u8 num_sub_vols;
211 __u8 tid;
212 __u8 cng_master_disk;
213 __u16 cache_policy;
214 __u8 cng_state;
215 __u8 cng_sub_state;
2432ce9b
AP		 216 __u16 my_vol_raid_dev_num; /* Used in Unique volume Id for this RaidDev */
217
218 /* NVM_EN */
219 __u8 nv_cache_mode;
220 __u8 nv_cache_flags;
221
222 /* Unique Volume Id of the NvCache Volume associated with this volume */
223 __u32 nvc_vol_orig_family_num;
224 __u16 nvc_vol_raid_dev_num;
225
226#define RWH_OFF 0
227#define RWH_DISTRIBUTED 1
228#define RWH_JOURNALING_DRIVE 2
c2462068
PB		 229#define RWH_MULTIPLE_DISTRIBUTED 3
230#define RWH_MULTIPLE_PPLS_JOURNALING_DRIVE 4
231#define RWH_MULTIPLE_OFF 5
2432ce9b
AP		 232 __u8 rwh_policy; /* Raid Write Hole Policy */
233 __u8 jd_serial[MAX_RAID_SERIAL_LEN]; /* Journal Drive serial number */
234 __u8 filler1;
235
236#define IMSM_DEV_FILLERS 3
cdddbdbc
DW		 237 __u32 filler[IMSM_DEV_FILLERS];
238 struct imsm_vol vol;
761e3bd9
N		 239};
240ASSERT_SIZE(imsm_dev, 164)
cdddbdbc
DW		 241
242struct imsm_super {
243 __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */
244 __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */
245 __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */
246 __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */
247 __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */
604b746f
JD		 248 __u32 error_log_size; /* 0x30 - 0x33 in bytes */
249 __u32 attributes; /* 0x34 - 0x37 */
cdddbdbc
DW		 250 __u8 num_disks; /* 0x38 Number of configured disks */
251 __u8 num_raid_devs; /* 0x39 Number of configured volumes */
604b746f
JD		 252 __u8 error_log_pos; /* 0x3A */
253 __u8 fill[1]; /* 0x3B */
254 __u32 cache_size; /* 0x3c - 0x40 in mb */
255 __u32 orig_family_num; /* 0x40 - 0x43 original family num */
256 __u32 pwr_cycle_count; /* 0x44 - 0x47 simulated power cycle count for array */
257 __u32 bbm_log_size; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */
2a24dc1b
PB		 258 __u16 num_raid_devs_created; /* 0x4C - 0x4D Used for generating unique
259 * volume IDs for raid_dev created in this array
260 * (starts at 1)
261 */
262 __u16 filler1; /* 0x4E - 0x4F */
e48aed3c
AP		 263 __u64 creation_time; /* 0x50 - 0x57 Array creation time */
264#define IMSM_FILLERS 32
265 __u32 filler[IMSM_FILLERS]; /* 0x58 - 0xD7 RAID_MPB_FILLERS */
cdddbdbc
DW		 266 struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */
267 /* here comes imsm_dev[num_raid_devs] */
604b746f 268 /* here comes BBM logs */
761e3bd9
N		 269};
270ASSERT_SIZE(imsm_super, 264)
cdddbdbc 271
604b746f 272#define BBM_LOG_MAX_ENTRIES 254
8d67477f
TM		 273#define BBM_LOG_MAX_LBA_ENTRY_VAL 256 /* Represents 256 LBAs */
274#define BBM_LOG_SIGNATURE 0xabadb10c
275
276struct bbm_log_block_addr {
277 __u16 w1;
278 __u32 dw1;
279} __attribute__ ((__packed__));
604b746f
JD		 280
281struct bbm_log_entry {
8d67477f
TM		 282 __u8 marked_count; /* Number of blocks marked - 1 */
283 __u8 disk_ordinal; /* Disk entry within the imsm_super */
284 struct bbm_log_block_addr defective_block_start;
604b746f
JD		 285} __attribute__ ((__packed__));
286
287struct bbm_log {
288 __u32 signature; /* 0xABADB10C */
289 __u32 entry_count;
8d67477f 290 struct bbm_log_entry marked_block_entries[BBM_LOG_MAX_ENTRIES];
761e3bd9
N		 291};
292ASSERT_SIZE(bbm_log, 2040)
604b746f 293
cdddbdbc 294static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" };
cdddbdbc 295
b53bfba6
TM		 296#define BLOCKS_PER_KB (1024/512)
297
8e59f3d8
AK		 298#define RAID_DISK_RESERVED_BLOCKS_IMSM_HI 2209
299
300#define GEN_MIGR_AREA_SIZE 2048 /* General Migration Copy Area size in blocks */
301
de44e46f
PB		 302#define MIGR_REC_BUF_SECTORS 1 /* size of migr_record i/o buffer in sectors */
303#define MIGR_REC_SECTOR_POSITION 1 /* migr_record position offset on disk,
304 * MIGR_REC_BUF_SECTORS <= MIGR_REC_SECTOR_POS
17a4eaf9
AK		 305 */
306
8e59f3d8
AK		 307#define UNIT_SRC_NORMAL 0 /* Source data for curr_migr_unit must
308 * be recovered using srcMap */
309#define UNIT_SRC_IN_CP_AREA 1 /* Source data for curr_migr_unit has
310 * already been migrated and must
311 * be recovered from checkpoint area */
2432ce9b 312
c2462068 313#define PPL_ENTRY_SPACE (128 * 1024) /* Size of single PPL, without the header */
2432ce9b 314
8e59f3d8
AK		 315struct migr_record {
316 __u32 rec_status; /* Status used to determine how to restart
317 * migration in case it aborts
318 * in some fashion */
9f421827 319 __u32 curr_migr_unit_lo; /* 0..numMigrUnits-1 */
8e59f3d8
AK		 320 __u32 family_num; /* Family number of MPB
321 * containing the RaidDev
322 * that is migrating */
323 __u32 ascending_migr; /* True if migrating in increasing
324 * order of lbas */
325 __u32 blocks_per_unit; /* Num disk blocks per unit of operation */
326 __u32 dest_depth_per_unit; /* Num member blocks each destMap
327 * member disk
328 * advances per unit-of-operation */
9f421827
PB		 329 __u32 ckpt_area_pba_lo; /* Pba of first block of ckpt copy area */
330 __u32 dest_1st_member_lba_lo; /* First member lba on first
331 * stripe of destination */
332 __u32 num_migr_units_lo; /* Total num migration units-of-op */
8e59f3d8
AK		 333 __u32 post_migr_vol_cap; /* Size of volume after
334 * migration completes */
335 __u32 post_migr_vol_cap_hi; /* Expansion space for LBA64 */
336 __u32 ckpt_read_disk_num; /* Which member disk in destSubMap[0] the
337 * migration ckpt record was read from
338 * (for recovered migrations) */
9f421827
PB		 339 __u32 curr_migr_unit_hi; /* 0..numMigrUnits-1 high order 32 bits */
340 __u32 ckpt_area_pba_hi; /* Pba of first block of ckpt copy area
341 * high order 32 bits */
342 __u32 dest_1st_member_lba_hi; /* First member lba on first stripe of
343 * destination - high order 32 bits */
344 __u32 num_migr_units_hi; /* Total num migration units-of-op
345 * high order 32 bits */
761e3bd9
N		 346};
347ASSERT_SIZE(migr_record, 64)
8e59f3d8 348
ec50f7b6
LM		 349struct md_list {
350 /* usage marker:
351 * 1: load metadata
352 * 2: metadata does not match
353 * 4: already checked
354 */
355 int used;
356 char *devname;
357 int found;
358 int container;
359 dev_t st_rdev;
360 struct md_list *next;
361};
362
e7b84f9d 363#define pr_vrb(fmt, arg...) (void) (verbose && pr_err(fmt, ##arg))
ec50f7b6 364
1484e727
DW		 365static __u8 migr_type(struct imsm_dev *dev)
366{
367 if (dev->vol.migr_type == MIGR_VERIFY &&
368 dev->status & DEV_VERIFY_AND_FIX)
369 return MIGR_REPAIR;
370 else
371 return dev->vol.migr_type;
372}
373
374static void set_migr_type(struct imsm_dev *dev, __u8 migr_type)
375{
376 /* for compatibility with older oroms convert MIGR_REPAIR, into
377 * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status
378 */
379 if (migr_type == MIGR_REPAIR) {
380 dev->vol.migr_type = MIGR_VERIFY;
381 dev->status |= DEV_VERIFY_AND_FIX;
382 } else {
383 dev->vol.migr_type = migr_type;
384 dev->status &= ~DEV_VERIFY_AND_FIX;
385 }
386}
387
f36a9ecd 388static unsigned int sector_count(__u32 bytes, unsigned int sector_size)
cdddbdbc 389{
f36a9ecd 390 return ROUND_UP(bytes, sector_size) / sector_size;
87eb16df 391}
cdddbdbc 392
f36a9ecd
PB		 393static unsigned int mpb_sectors(struct imsm_super *mpb,
394 unsigned int sector_size)
87eb16df 395{
f36a9ecd 396 return sector_count(__le32_to_cpu(mpb->mpb_size), sector_size);
cdddbdbc
DW		 397}
398
ba2de7ba
DW		 399struct intel_dev {
400 struct imsm_dev *dev;
401 struct intel_dev *next;
f21e18ca 402 unsigned index;
ba2de7ba
DW		 403};
404
88654014
LM		 405struct intel_hba {
406 enum sys_dev_type type;
407 char *path;
408 char *pci_id;
409 struct intel_hba *next;
410};
411
1a64be56
LM		 412enum action {
413 DISK_REMOVE = 1,
414 DISK_ADD
415};
cdddbdbc
DW		 416/* internal representation of IMSM metadata */
417struct intel_super {
418 union {
949c47a0
DW		 419 void *buf; /* O_DIRECT buffer for reading/writing metadata */
420 struct imsm_super *anchor; /* immovable parameters */
cdddbdbc 421 };
8e59f3d8
AK		 422 union {
423 void *migr_rec_buf; /* buffer for I/O operations */
424 struct migr_record *migr_rec; /* migration record */
425 };
51d83f5d
AK		 426 int clean_migration_record_by_mdmon; /* when reshape is switched to next
427 array, it indicates that mdmon is allowed to clean migration
428 record */
949c47a0 429 size_t len; /* size of the 'buf' allocation */
bbab0940 430 size_t extra_space; /* extra space in 'buf' that is not used yet */
4d7b1503
DW		 431 void *next_buf; /* for realloc'ing buf from the manager */
432 size_t next_len;
c2c087e6 433 int updates_pending; /* count of pending updates for mdmon */
bf5a934a 434 int current_vol; /* index of raid device undergoing creation */
5551b113 435 unsigned long long create_offset; /* common start for 'current_vol' */
148acb7b 436 __u32 random; /* random data for seeding new family numbers */
ba2de7ba 437 struct intel_dev *devlist;
fa7bb6f8 438 unsigned int sector_size; /* sector size of used member drives */
cdddbdbc
DW		 439 struct dl {
440 struct dl *next;
441 int index;
442 __u8 serial[MAX_RAID_SERIAL_LEN];
443 int major, minor;
444 char *devname;
b9f594fe 445 struct imsm_disk disk;
cdddbdbc 446 int fd;
0dcecb2e
DW		 447 int extent_cnt;
448 struct extent *e; /* for determining freespace @ create */
efb30e7f 449 int raiddisk; /* slot to fill in autolayout */
1a64be56 450 enum action action;
ca0748fa 451 } *disks, *current_disk;
1a64be56
LM		 452 struct dl *disk_mgmt_list; /* list of disks to add/remove while mdmon
453 active */
47ee5a45 454 struct dl *missing; /* disks removed while we weren't looking */
43dad3d6 455 struct bbm_log *bbm_log;
88654014 456 struct intel_hba *hba; /* device path of the raid controller for this metadata */
88c32bb1 457 const struct imsm_orom *orom; /* platform firmware support */
a2b97981 458 struct intel_super *next; /* (temp) list for disambiguating family_num */
928f1424 459 struct md_bb bb; /* memory for get_bad_blocks call */
a2b97981
DW		 460};
461
462struct intel_disk {
463 struct imsm_disk disk;
464 #define IMSM_UNKNOWN_OWNER (-1)
465 int owner;
466 struct intel_disk *next;
cdddbdbc
DW		 467};
468
c2c087e6
DW		 469struct extent {
470 unsigned long long start, size;
471};
472
694575e7
KW		 473/* definitions of reshape process types */
474enum imsm_reshape_type {
475 CH_TAKEOVER,
b5347799 476 CH_MIGRATION,
7abc9871 477 CH_ARRAY_SIZE,
694575e7
KW		 478};
479
88758e9d
DW		 480/* definition of messages passed to imsm_process_update */
481enum imsm_update_type {
482 update_activate_spare,
8273f55e 483 update_create_array,
33414a01 484 update_kill_array,
aa534678 485 update_rename_array,
1a64be56 486 update_add_remove_disk,
78b10e66 487 update_reshape_container_disks,
48c5303a 488 update_reshape_migration,
2d40f3a1
AK		 489 update_takeover,
490 update_general_migration_checkpoint,
f3871fdc 491 update_size_change,
bbab0940 492 update_prealloc_badblocks_mem,
e6e9dd3f 493 update_rwh_policy,
88758e9d
DW		 494};
495
496struct imsm_update_activate_spare {
497 enum imsm_update_type type;
d23fe947 498 struct dl *dl;
88758e9d
DW		 499 int slot;
500 int array;
501 struct imsm_update_activate_spare *next;
502};
503
78b10e66 504struct geo_params {
4dd2df09 505 char devnm[32];
78b10e66 506 char *dev_name;
d04f65f4 507 unsigned long long size;
78b10e66
N		 508 int level;
509 int layout;
510 int chunksize;
511 int raid_disks;
512};
513
bb025c2f
KW		 514enum takeover_direction {
515 R10_TO_R0,
516 R0_TO_R10
517};
518struct imsm_update_takeover {
519 enum imsm_update_type type;
520 int subarray;
521 enum takeover_direction direction;
522};
78b10e66
N		 523
524struct imsm_update_reshape {
525 enum imsm_update_type type;
526 int old_raid_disks;
527 int new_raid_disks;
48c5303a
PC		 528
529 int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */
530};
531
532struct imsm_update_reshape_migration {
533 enum imsm_update_type type;
534 int old_raid_disks;
535 int new_raid_disks;
536 /* fields for array migration changes
537 */
538 int subdev;
539 int new_level;
540 int new_layout;
4bba0439 541 int new_chunksize;
48c5303a 542
d195167d 543 int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */
78b10e66
N		 544};
545
f3871fdc
AK		 546struct imsm_update_size_change {
547 enum imsm_update_type type;
548 int subdev;
549 long long new_size;
550};
551
2d40f3a1
AK		 552struct imsm_update_general_migration_checkpoint {
553 enum imsm_update_type type;
554 __u32 curr_migr_unit;
555};
556
54c2c1ea
DW		 557struct disk_info {
558 __u8 serial[MAX_RAID_SERIAL_LEN];
559};
560
8273f55e
DW		 561struct imsm_update_create_array {
562 enum imsm_update_type type;
8273f55e 563 int dev_idx;
6a3e913e 564 struct imsm_dev dev;
8273f55e
DW		 565};
566
33414a01
DW		 567struct imsm_update_kill_array {
568 enum imsm_update_type type;
569 int dev_idx;
570};
571
aa534678
DW		 572struct imsm_update_rename_array {
573 enum imsm_update_type type;
574 __u8 name[MAX_RAID_SERIAL_LEN];
575 int dev_idx;
576};
577
1a64be56 578struct imsm_update_add_remove_disk {
43dad3d6
DW		 579 enum imsm_update_type type;
580};
581
bbab0940
TM		 582struct imsm_update_prealloc_bb_mem {
583 enum imsm_update_type type;
584};
585
e6e9dd3f
AP		 586struct imsm_update_rwh_policy {
587 enum imsm_update_type type;
588 int new_policy;
589 int dev_idx;
590};
591
88654014
LM		 592static const char *_sys_dev_type[] = {
593 [SYS_DEV_UNKNOWN] = "Unknown",
594 [SYS_DEV_SAS] = "SAS",
614902f6 595 [SYS_DEV_SATA] = "SATA",
60f0f54d
PB		 596 [SYS_DEV_NVME] = "NVMe",
597 [SYS_DEV_VMD] = "VMD"
88654014
LM		 598};
599
600const char *get_sys_dev_type(enum sys_dev_type type)
601{
602 if (type >= SYS_DEV_MAX)
603 type = SYS_DEV_UNKNOWN;
604
605 return _sys_dev_type[type];
606}
607
608static struct intel_hba * alloc_intel_hba(struct sys_dev *device)
609{
503975b9
N		 610 struct intel_hba *result = xmalloc(sizeof(*result));
611
612 result->type = device->type;
613 result->path = xstrdup(device->path);
614 result->next = NULL;
615 if (result->path && (result->pci_id = strrchr(result->path, '/')) != NULL)
616 result->pci_id++;
617
88654014
LM		 618 return result;
619}
620
621static struct intel_hba * find_intel_hba(struct intel_hba *hba, struct sys_dev *device)
622{
594dc1b8
JS		 623 struct intel_hba *result;
624
88654014
LM		 625 for (result = hba; result; result = result->next) {
626 if (result->type == device->type && strcmp(result->path, device->path) == 0)
627 break;
628 }
629 return result;
630}
631
b4cf4cba 632static int attach_hba_to_super(struct intel_super *super, struct sys_dev *device)
88654014
LM		 633{
634 struct intel_hba *hba;
635
636 /* check if disk attached to Intel HBA */
637 hba = find_intel_hba(super->hba, device);
638 if (hba != NULL)
639 return 1;
640 /* Check if HBA is already attached to super */
641 if (super->hba == NULL) {
642 super->hba = alloc_intel_hba(device);
643 return 1;
6b781d33
AP		 644 }
645
646 hba = super->hba;
647 /* Intel metadata allows for all disks attached to the same type HBA.
614902f6 648 * Do not support HBA types mixing
6b781d33
AP		 649 */
650 if (device->type != hba->type)
88654014 651 return 2;
6b781d33
AP		 652
653 /* Multiple same type HBAs can be used if they share the same OROM */
654 const struct imsm_orom *device_orom = get_orom_by_device_id(device->dev_id);
655
656 if (device_orom != super->orom)
657 return 2;
658
659 while (hba->next)
660 hba = hba->next;
661
662 hba->next = alloc_intel_hba(device);
663 return 1;
88654014
LM		 664}
665
666static struct sys_dev* find_disk_attached_hba(int fd, const char *devname)
667{
9bc4ae77 668 struct sys_dev *list, *elem;
88654014
LM		 669 char *disk_path;
670
671 if ((list = find_intel_devices()) == NULL)
672 return 0;
673
674 if (fd < 0)
675 disk_path = (char *) devname;
676 else
677 disk_path = diskfd_to_devpath(fd);
678
9bc4ae77 679 if (!disk_path)
88654014 680 return 0;
88654014 681
9bc4ae77
N		 682 for (elem = list; elem; elem = elem->next)
683 if (path_attached_to_hba(disk_path, elem->path))
88654014 684 return elem;
9bc4ae77 685
88654014
LM		 686 if (disk_path != devname)
687 free(disk_path);
88654014
LM		 688
689 return NULL;
690}
691
d424212e
N		 692static int find_intel_hba_capability(int fd, struct intel_super *super,
693 char *devname);
f2f5c343 694
cdddbdbc
DW		 695static struct supertype *match_metadata_desc_imsm(char *arg)
696{
697 struct supertype *st;
698
699 if (strcmp(arg, "imsm") != 0 &&
700 strcmp(arg, "default") != 0
701 )
702 return NULL;
703
503975b9 704 st = xcalloc(1, sizeof(*st));
cdddbdbc
DW		 705 st->ss = &super_imsm;
706 st->max_devs = IMSM_MAX_DEVICES;
707 st->minor_version = 0;
708 st->sb = NULL;
709 return st;
710}
711
cdddbdbc
DW		 712static __u8 *get_imsm_version(struct imsm_super *mpb)
713{
714 return &mpb->sig[MPB_SIG_LEN];
715}
716
949c47a0
DW		 717/* retrieve a disk directly from the anchor when the anchor is known to be
718 * up-to-date, currently only at load time
719 */
720static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index)
cdddbdbc 721{
949c47a0 722 if (index >= mpb->num_disks)
cdddbdbc
DW		 723 return NULL;
724 return &mpb->disk[index];
725}
726
95d07a2c
LM		 727/* retrieve the disk description based on a index of the disk
728 * in the sub-array
729 */
730static struct dl *get_imsm_dl_disk(struct intel_super *super, __u8 index)
949c47a0 731{
b9f594fe
DW		 732 struct dl *d;
733
734 for (d = super->disks; d; d = d->next)
735 if (d->index == index)
95d07a2c
LM		 736 return d;
737
738 return NULL;
739}
740/* retrieve a disk from the parsed metadata */
741static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index)
742{
743 struct dl *dl;
744
745 dl = get_imsm_dl_disk(super, index);
746 if (dl)
747 return &dl->disk;
748
b9f594fe 749 return NULL;
949c47a0
DW		 750}
751
752/* generate a checksum directly from the anchor when the anchor is known to be
753 * up-to-date, currently only at load or write_super after coalescing
754 */
755static __u32 __gen_imsm_checksum(struct imsm_super *mpb)
cdddbdbc
DW		 756{
757 __u32 end = mpb->mpb_size / sizeof(end);
758 __u32 *p = (__u32 *) mpb;
759 __u32 sum = 0;
760
5d500228
N		 761 while (end--) {
762 sum += __le32_to_cpu(*p);
97f734fd
N		 763 p++;
764 }
cdddbdbc 765
5d500228 766 return sum - __le32_to_cpu(mpb->check_sum);
cdddbdbc
DW		 767}
768
a965f303
DW		 769static size_t sizeof_imsm_map(struct imsm_map *map)
770{
771 return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1);
772}
773
774struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map)
cdddbdbc 775{
5e7b0330
AK		 776 /* A device can have 2 maps if it is in the middle of a migration.
777 * If second_map is:
238c0a71
AK		 778 * MAP_0 - we return the first map
779 * MAP_1 - we return the second map if it exists, else NULL
780 * MAP_X - we return the second map if it exists, else the first
5e7b0330 781 */
a965f303 782 struct imsm_map *map = &dev->vol.map[0];
9535fc47 783 struct imsm_map *map2 = NULL;
a965f303 784
9535fc47
AK		 785 if (dev->vol.migr_state)
786 map2 = (void *)map + sizeof_imsm_map(map);
a965f303 787
9535fc47 788 switch (second_map) {
3b451610 789 case MAP_0:
9535fc47 790 break;
3b451610 791 case MAP_1:
9535fc47
AK		 792 map = map2;
793 break;
238c0a71 794 case MAP_X:
9535fc47
AK		 795 if (map2)
796 map = map2;
797 break;
9535fc47
AK		 798 default:
799 map = NULL;
800 }
801 return map;
5e7b0330 802
a965f303 803}
cdddbdbc 804
3393c6af
DW		 805/* return the size of the device.
806 * migr_state increases the returned size if map[0] were to be duplicated
807 */
808static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state)
a965f303
DW		 809{
810 size_t size = sizeof(*dev) - sizeof(struct imsm_map) +
238c0a71 811 sizeof_imsm_map(get_imsm_map(dev, MAP_0));
cdddbdbc
DW		 812
813 /* migrating means an additional map */
a965f303 814 if (dev->vol.migr_state)
238c0a71 815 size += sizeof_imsm_map(get_imsm_map(dev, MAP_1));
3393c6af 816 else if (migr_state)
238c0a71 817 size += sizeof_imsm_map(get_imsm_map(dev, MAP_0));
cdddbdbc
DW		 818
819 return size;
820}
821
54c2c1ea
DW		 822/* retrieve disk serial number list from a metadata update */
823static struct disk_info *get_disk_info(struct imsm_update_create_array *update)
824{
825 void *u = update;
826 struct disk_info *inf;
827
828 inf = u + sizeof(*update) - sizeof(struct imsm_dev) +
829 sizeof_imsm_dev(&update->dev, 0);
830
831 return inf;
832}
54c2c1ea 833
949c47a0 834static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index)
cdddbdbc
DW		 835{
836 int offset;
837 int i;
838 void *_mpb = mpb;
839
949c47a0 840 if (index >= mpb->num_raid_devs)
cdddbdbc
DW		 841 return NULL;
842
843 /* devices start after all disks */
844 offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb;
845
846 for (i = 0; i <= index; i++)
847 if (i == index)
848 return _mpb + offset;
849 else
3393c6af 850 offset += sizeof_imsm_dev(_mpb + offset, 0);
cdddbdbc
DW		 851
852 return NULL;
853}
854
949c47a0
DW		 855static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index)
856{
ba2de7ba
DW		 857 struct intel_dev *dv;
858
949c47a0
DW		 859 if (index >= super->anchor->num_raid_devs)
860 return NULL;
ba2de7ba
DW		 861 for (dv = super->devlist; dv; dv = dv->next)
862 if (dv->index == index)
863 return dv->dev;
864 return NULL;
949c47a0
DW		 865}
866
8d67477f
TM		 867static inline unsigned long long __le48_to_cpu(const struct bbm_log_block_addr
868 *addr)
869{
870 return ((((__u64)__le32_to_cpu(addr->dw1)) << 16) |
871 __le16_to_cpu(addr->w1));
872}
873
874static inline struct bbm_log_block_addr __cpu_to_le48(unsigned long long sec)
875{
876 struct bbm_log_block_addr addr;
877
878 addr.w1 = __cpu_to_le16((__u16)(sec & 0xffff));
879 addr.dw1 = __cpu_to_le32((__u32)(sec >> 16) & 0xffffffff);
880 return addr;
881}
882
8d67477f
TM		 883/* get size of the bbm log */
884static __u32 get_imsm_bbm_log_size(struct bbm_log *log)
885{
886 if (!log || log->entry_count == 0)
887 return 0;
888
889 return sizeof(log->signature) +
890 sizeof(log->entry_count) +
891 log->entry_count * sizeof(struct bbm_log_entry);
892}
6f50473f
TM		 893
894/* check if bad block is not partially stored in bbm log */
895static int is_stored_in_bbm(struct bbm_log *log, const __u8 idx, const unsigned
896 long long sector, const int length, __u32 *pos)
897{
898 __u32 i;
899
900 for (i = *pos; i < log->entry_count; i++) {
901 struct bbm_log_entry *entry = &log->marked_block_entries[i];
902 unsigned long long bb_start;
903 unsigned long long bb_end;
904
905 bb_start = __le48_to_cpu(&entry->defective_block_start);
906 bb_end = bb_start + (entry->marked_count + 1);
907
908 if ((entry->disk_ordinal == idx) && (bb_start >= sector) &&
909 (bb_end <= sector + length)) {
910 *pos = i;
911 return 1;
912 }
913 }
914 return 0;
915}
916
917/* record new bad block in bbm log */
918static int record_new_badblock(struct bbm_log *log, const __u8 idx, unsigned
919 long long sector, int length)
920{
921 int new_bb = 0;
922 __u32 pos = 0;
923 struct bbm_log_entry *entry = NULL;
924
925 while (is_stored_in_bbm(log, idx, sector, length, &pos)) {
926 struct bbm_log_entry *e = &log->marked_block_entries[pos];
927
928 if ((e->marked_count + 1 == BBM_LOG_MAX_LBA_ENTRY_VAL) &&
929 (__le48_to_cpu(&e->defective_block_start) == sector)) {
930 sector += BBM_LOG_MAX_LBA_ENTRY_VAL;
931 length -= BBM_LOG_MAX_LBA_ENTRY_VAL;
932 pos = pos + 1;
933 continue;
934 }
935 entry = e;
936 break;
937 }
938
939 if (entry) {
940 int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length :
941 BBM_LOG_MAX_LBA_ENTRY_VAL;
942 entry->defective_block_start = __cpu_to_le48(sector);
943 entry->marked_count = cnt - 1;
944 if (cnt == length)
945 return 1;
946 sector += cnt;
947 length -= cnt;
948 }
949
950 new_bb = ROUND_UP(length, BBM_LOG_MAX_LBA_ENTRY_VAL) /
951 BBM_LOG_MAX_LBA_ENTRY_VAL;
952 if (log->entry_count + new_bb > BBM_LOG_MAX_ENTRIES)
953 return 0;
954
955 while (length > 0) {
956 int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length :
957 BBM_LOG_MAX_LBA_ENTRY_VAL;
958 struct bbm_log_entry *entry =
959 &log->marked_block_entries[log->entry_count];
960
961 entry->defective_block_start = __cpu_to_le48(sector);
962 entry->marked_count = cnt - 1;
963 entry->disk_ordinal = idx;
964
965 sector += cnt;
966 length -= cnt;
967
968 log->entry_count++;
969 }
970
971 return new_bb;
972}
c07a5a4f 973
4c9e8c1e
TM		 974/* clear all bad blocks for given disk */
975static void clear_disk_badblocks(struct bbm_log *log, const __u8 idx)
976{
977 __u32 i = 0;
978
979 while (i < log->entry_count) {
980 struct bbm_log_entry *entries = log->marked_block_entries;
981
982 if (entries[i].disk_ordinal == idx) {
983 if (i < log->entry_count - 1)
984 entries[i] = entries[log->entry_count - 1];
985 log->entry_count--;
986 } else {
987 i++;
988 }
989 }
990}
991
c07a5a4f
TM		 992/* clear given bad block */
993static int clear_badblock(struct bbm_log *log, const __u8 idx, const unsigned
994 long long sector, const int length) {
995 __u32 i = 0;
996
997 while (i < log->entry_count) {
998 struct bbm_log_entry *entries = log->marked_block_entries;
999
1000 if ((entries[i].disk_ordinal == idx) &&
1001 (__le48_to_cpu(&entries[i].defective_block_start) ==
1002 sector) && (entries[i].marked_count + 1 == length)) {
1003 if (i < log->entry_count - 1)
1004 entries[i] = entries[log->entry_count - 1];
1005 log->entry_count--;
1006 break;
1007 }
1008 i++;
1009 }
1010
1011 return 1;
1012}
8d67477f
TM		 1013
1014/* allocate and load BBM log from metadata */
1015static int load_bbm_log(struct intel_super *super)
1016{
1017 struct imsm_super *mpb = super->anchor;
1018 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
1019
1020 super->bbm_log = xcalloc(1, sizeof(struct bbm_log));
1021 if (!super->bbm_log)
1022 return 1;
1023
1024 if (bbm_log_size) {
1025 struct bbm_log *log = (void *)mpb +
1026 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
1027
1028 __u32 entry_count;
1029
1030 if (bbm_log_size < sizeof(log->signature) +
1031 sizeof(log->entry_count))
1032 return 2;
1033
1034 entry_count = __le32_to_cpu(log->entry_count);
1035 if ((__le32_to_cpu(log->signature) != BBM_LOG_SIGNATURE) ||
1036 (entry_count > BBM_LOG_MAX_ENTRIES))
1037 return 3;
1038
1039 if (bbm_log_size !=
1040 sizeof(log->signature) + sizeof(log->entry_count) +
1041 entry_count * sizeof(struct bbm_log_entry))
1042 return 4;
1043
1044 memcpy(super->bbm_log, log, bbm_log_size);
1045 } else {
1046 super->bbm_log->signature = __cpu_to_le32(BBM_LOG_SIGNATURE);
1047 super->bbm_log->entry_count = 0;
1048 }
1049
1050 return 0;
1051}
1052
b12796be
TM		 1053/* checks if bad block is within volume boundaries */
1054static int is_bad_block_in_volume(const struct bbm_log_entry *entry,
1055 const unsigned long long start_sector,
1056 const unsigned long long size)
1057{
1058 unsigned long long bb_start;
1059 unsigned long long bb_end;
1060
1061 bb_start = __le48_to_cpu(&entry->defective_block_start);
1062 bb_end = bb_start + (entry->marked_count + 1);
1063
1064 if (((bb_start >= start_sector) && (bb_start < start_sector + size)) ||
1065 ((bb_end >= start_sector) && (bb_end <= start_sector + size)))
1066 return 1;
1067
1068 return 0;
1069}
1070
1071/* get list of bad blocks on a drive for a volume */
1072static void get_volume_badblocks(const struct bbm_log *log, const __u8 idx,
1073 const unsigned long long start_sector,
1074 const unsigned long long size,
1075 struct md_bb *bbs)
1076{
1077 __u32 count = 0;
1078 __u32 i;
1079
1080 for (i = 0; i < log->entry_count; i++) {
1081 const struct bbm_log_entry *ent =
1082 &log->marked_block_entries[i];
1083 struct md_bb_entry *bb;
1084
1085 if ((ent->disk_ordinal == idx) &&
1086 is_bad_block_in_volume(ent, start_sector, size)) {
1087
1088 if (!bbs->entries) {
1089 bbs->entries = xmalloc(BBM_LOG_MAX_ENTRIES *
1090 sizeof(*bb));
1091 if (!bbs->entries)
1092 break;
1093 }
1094
1095 bb = &bbs->entries[count++];
1096 bb->sector = __le48_to_cpu(&ent->defective_block_start);
1097 bb->length = ent->marked_count + 1;
1098 }
1099 }
1100 bbs->count = count;
1101}
1102
98130f40
AK		 1103/*
1104 * for second_map:
238c0a71
AK		 1105 * == MAP_0 get first map
1106 * == MAP_1 get second map
1107 * == MAP_X than get map according to the current migr_state
98130f40
AK		 1108 */
1109static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev,
1110 int slot,
1111 int second_map)
7eef0453
DW		 1112{
1113 struct imsm_map *map;
1114
5e7b0330 1115 map = get_imsm_map(dev, second_map);
7eef0453 1116
ff077194
DW		 1117 /* top byte identifies disk under rebuild */
1118 return __le32_to_cpu(map->disk_ord_tbl[slot]);
1119}
1120
1121#define ord_to_idx(ord) (((ord) << 8) >> 8)
98130f40 1122static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot, int second_map)
ff077194 1123{
98130f40 1124 __u32 ord = get_imsm_ord_tbl_ent(dev, slot, second_map);
ff077194
DW		 1125
1126 return ord_to_idx(ord);
7eef0453
DW		 1127}
1128
be73972f
DW		 1129static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord)
1130{
1131 map->disk_ord_tbl[slot] = __cpu_to_le32(ord);
1132}
1133
f21e18ca 1134static int get_imsm_disk_slot(struct imsm_map *map, unsigned idx)
620b1713
DW		 1135{
1136 int slot;
1137 __u32 ord;
1138
1139 for (slot = 0; slot < map->num_members; slot++) {
1140 ord = __le32_to_cpu(map->disk_ord_tbl[slot]);
1141 if (ord_to_idx(ord) == idx)
1142 return slot;
1143 }
1144
1145 return -1;
1146}
1147
cdddbdbc
DW		 1148static int get_imsm_raid_level(struct imsm_map *map)
1149{
1150 if (map->raid_level == 1) {
1151 if (map->num_members == 2)
1152 return 1;
1153 else
1154 return 10;
1155 }
1156
1157 return map->raid_level;
1158}
1159
c2c087e6
DW		 1160static int cmp_extent(const void *av, const void *bv)
1161{
1162 const struct extent *a = av;
1163 const struct extent *b = bv;
1164 if (a->start < b->start)
1165 return -1;
1166 if (a->start > b->start)
1167 return 1;
1168 return 0;
1169}
1170
0dcecb2e 1171static int count_memberships(struct dl *dl, struct intel_super *super)
c2c087e6 1172{
c2c087e6 1173 int memberships = 0;
620b1713 1174 int i;
c2c087e6 1175
949c47a0
DW		 1176 for (i = 0; i < super->anchor->num_raid_devs; i++) {
1177 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 1178 struct imsm_map *map = get_imsm_map(dev, MAP_0);
c2c087e6 1179
620b1713
DW		 1180 if (get_imsm_disk_slot(map, dl->index) >= 0)
1181 memberships++;
c2c087e6 1182 }
0dcecb2e
DW		 1183
1184 return memberships;
1185}
1186
b81221b7
CA		 1187static __u32 imsm_min_reserved_sectors(struct intel_super *super);
1188
486720e0 1189static int split_ull(unsigned long long n, void *lo, void *hi)
5551b113
CA		 1190{
1191 if (lo == 0 || hi == 0)
1192 return 1;
486720e0
JS		 1193 __put_unaligned32(__cpu_to_le32((__u32)n), lo);
1194 __put_unaligned32(__cpu_to_le32((n >> 32)), hi);
5551b113
CA		 1195 return 0;
1196}
1197
1198static unsigned long long join_u32(__u32 lo, __u32 hi)
1199{
1200 return (unsigned long long)__le32_to_cpu(lo) |
1201 (((unsigned long long)__le32_to_cpu(hi)) << 32);
1202}
1203
1204static unsigned long long total_blocks(struct imsm_disk *disk)
1205{
1206 if (disk == NULL)
1207 return 0;
1208 return join_u32(disk->total_blocks_lo, disk->total_blocks_hi);
1209}
1210
1211static unsigned long long pba_of_lba0(struct imsm_map *map)
1212{
1213 if (map == NULL)
1214 return 0;
1215 return join_u32(map->pba_of_lba0_lo, map->pba_of_lba0_hi);
1216}
1217
1218static unsigned long long blocks_per_member(struct imsm_map *map)
1219{
1220 if (map == NULL)
1221 return 0;
1222 return join_u32(map->blocks_per_member_lo, map->blocks_per_member_hi);
1223}
1224
1225static unsigned long long num_data_stripes(struct imsm_map *map)
1226{
1227 if (map == NULL)
1228 return 0;
1229 return join_u32(map->num_data_stripes_lo, map->num_data_stripes_hi);
1230}
1231
fcc2c9da
MD		 1232static unsigned long long imsm_dev_size(struct imsm_dev *dev)
1233{
1234 if (dev == NULL)
1235 return 0;
1236 return join_u32(dev->size_low, dev->size_high);
1237}
1238
9f421827
PB		 1239static unsigned long long migr_chkp_area_pba(struct migr_record *migr_rec)
1240{
1241 if (migr_rec == NULL)
1242 return 0;
1243 return join_u32(migr_rec->ckpt_area_pba_lo,
1244 migr_rec->ckpt_area_pba_hi);
1245}
1246
1247static unsigned long long current_migr_unit(struct migr_record *migr_rec)
1248{
1249 if (migr_rec == NULL)
1250 return 0;
1251 return join_u32(migr_rec->curr_migr_unit_lo,
1252 migr_rec->curr_migr_unit_hi);
1253}
1254
1255static unsigned long long migr_dest_1st_member_lba(struct migr_record *migr_rec)
1256{
1257 if (migr_rec == NULL)
1258 return 0;
1259 return join_u32(migr_rec->dest_1st_member_lba_lo,
1260 migr_rec->dest_1st_member_lba_hi);
1261}
1262
1263static unsigned long long get_num_migr_units(struct migr_record *migr_rec)
1264{
1265 if (migr_rec == NULL)
1266 return 0;
1267 return join_u32(migr_rec->num_migr_units_lo,
1268 migr_rec->num_migr_units_hi);
1269}
1270
5551b113
CA		 1271static void set_total_blocks(struct imsm_disk *disk, unsigned long long n)
1272{
1273 split_ull(n, &disk->total_blocks_lo, &disk->total_blocks_hi);
1274}
1275
1276static void set_pba_of_lba0(struct imsm_map *map, unsigned long long n)
1277{
1278 split_ull(n, &map->pba_of_lba0_lo, &map->pba_of_lba0_hi);
1279}
1280
1281static void set_blocks_per_member(struct imsm_map *map, unsigned long long n)
1282{
1283 split_ull(n, &map->blocks_per_member_lo, &map->blocks_per_member_hi);
1284}
1285
1286static void set_num_data_stripes(struct imsm_map *map, unsigned long long n)
1287{
1288 split_ull(n, &map->num_data_stripes_lo, &map->num_data_stripes_hi);
1289}
1290
fcc2c9da
MD		 1291static void set_imsm_dev_size(struct imsm_dev *dev, unsigned long long n)
1292{
1293 split_ull(n, &dev->size_low, &dev->size_high);
1294}
1295
9f421827
PB		 1296static void set_migr_chkp_area_pba(struct migr_record *migr_rec,
1297 unsigned long long n)
1298{
1299 split_ull(n, &migr_rec->ckpt_area_pba_lo, &migr_rec->ckpt_area_pba_hi);
1300}
1301
1302static void set_current_migr_unit(struct migr_record *migr_rec,
1303 unsigned long long n)
1304{
1305 split_ull(n, &migr_rec->curr_migr_unit_lo,
1306 &migr_rec->curr_migr_unit_hi);
1307}
1308
1309static void set_migr_dest_1st_member_lba(struct migr_record *migr_rec,
1310 unsigned long long n)
1311{
1312 split_ull(n, &migr_rec->dest_1st_member_lba_lo,
1313 &migr_rec->dest_1st_member_lba_hi);
1314}
1315
1316static void set_num_migr_units(struct migr_record *migr_rec,
1317 unsigned long long n)
1318{
1319 split_ull(n, &migr_rec->num_migr_units_lo,
1320 &migr_rec->num_migr_units_hi);
1321}
1322
44490938
MD		 1323static unsigned long long per_dev_array_size(struct imsm_map *map)
1324{
1325 unsigned long long array_size = 0;
1326
1327 if (map == NULL)
1328 return array_size;
1329
1330 array_size = num_data_stripes(map) * map->blocks_per_strip;
1331 if (get_imsm_raid_level(map) == 1 || get_imsm_raid_level(map) == 10)
1332 array_size *= 2;
1333
1334 return array_size;
1335}
1336
05501181
PB		 1337static struct extent *get_extents(struct intel_super *super, struct dl *dl,
1338 int get_minimal_reservation)
0dcecb2e
DW		 1339{
1340 /* find a list of used extents on the given physical device */
1341 struct extent *rv, *e;
620b1713 1342 int i;
0dcecb2e 1343 int memberships = count_memberships(dl, super);
b276dd33
DW		 1344 __u32 reservation;
1345
1346 /* trim the reserved area for spares, so they can join any array
1347 * regardless of whether the OROM has assigned sectors from the
1348 * IMSM_RESERVED_SECTORS region
1349 */
05501181 1350 if (dl->index == -1 || get_minimal_reservation)
b81221b7 1351 reservation = imsm_min_reserved_sectors(super);
b276dd33
DW		 1352 else
1353 reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
0dcecb2e 1354
503975b9 1355 rv = xcalloc(sizeof(struct extent), (memberships + 1));
c2c087e6
DW		 1356 e = rv;
1357
949c47a0
DW		 1358 for (i = 0; i < super->anchor->num_raid_devs; i++) {
1359 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 1360 struct imsm_map *map = get_imsm_map(dev, MAP_0);
c2c087e6 1361
620b1713 1362 if (get_imsm_disk_slot(map, dl->index) >= 0) {
5551b113 1363 e->start = pba_of_lba0(map);
44490938 1364 e->size = per_dev_array_size(map);
620b1713 1365 e++;
c2c087e6
DW		 1366 }
1367 }
1368 qsort(rv, memberships, sizeof(*rv), cmp_extent);
1369
1011e834 1370 /* determine the start of the metadata
14e8215b
DW		 1371 * when no raid devices are defined use the default
1372 * ...otherwise allow the metadata to truncate the value
1373 * as is the case with older versions of imsm
1374 */
1375 if (memberships) {
1376 struct extent *last = &rv[memberships - 1];
5551b113 1377 unsigned long long remainder;
14e8215b 1378
5551b113 1379 remainder = total_blocks(&dl->disk) - (last->start + last->size);
dda5855f
DW		 1380 /* round down to 1k block to satisfy precision of the kernel
1381 * 'size' interface
1382 */
1383 remainder &= ~1UL;
1384 /* make sure remainder is still sane */
f21e18ca 1385 if (remainder < (unsigned)ROUND_UP(super->len, 512) >> 9)
dda5855f 1386 remainder = ROUND_UP(super->len, 512) >> 9;
14e8215b
DW		 1387 if (reservation > remainder)
1388 reservation = remainder;
1389 }
5551b113 1390 e->start = total_blocks(&dl->disk) - reservation;
c2c087e6
DW		 1391 e->size = 0;
1392 return rv;
1393}
1394
14e8215b
DW		 1395/* try to determine how much space is reserved for metadata from
1396 * the last get_extents() entry, otherwise fallback to the
1397 * default
1398 */
1399static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl)
1400{
1401 struct extent *e;
1402 int i;
1403 __u32 rv;
1404
1405 /* for spares just return a minimal reservation which will grow
1406 * once the spare is picked up by an array
1407 */
1408 if (dl->index == -1)
1409 return MPB_SECTOR_CNT;
1410
05501181 1411 e = get_extents(super, dl, 0);
14e8215b
DW		 1412 if (!e)
1413 return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
1414
1415 /* scroll to last entry */
1416 for (i = 0; e[i].size; i++)
1417 continue;
1418
5551b113 1419 rv = total_blocks(&dl->disk) - e[i].start;
14e8215b
DW		 1420
1421 free(e);
1422
1423 return rv;
1424}
1425
25ed7e59
DW		 1426static int is_spare(struct imsm_disk *disk)
1427{
1428 return (disk->status & SPARE_DISK) == SPARE_DISK;
1429}
1430
1431static int is_configured(struct imsm_disk *disk)
1432{
1433 return (disk->status & CONFIGURED_DISK) == CONFIGURED_DISK;
1434}
1435
1436static int is_failed(struct imsm_disk *disk)
1437{
1438 return (disk->status & FAILED_DISK) == FAILED_DISK;
1439}
1440
2432ce9b
AP		 1441static int is_journal(struct imsm_disk *disk)
1442{
1443 return (disk->status & JOURNAL_DISK) == JOURNAL_DISK;
1444}
1445
b53bfba6
TM		 1446/* round array size down to closest MB and ensure it splits evenly
1447 * between members
1448 */
1449static unsigned long long round_size_to_mb(unsigned long long size, unsigned int
1450 disk_count)
1451{
1452 size /= disk_count;
1453 size = (size >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT;
1454 size *= disk_count;
1455
1456 return size;
1457}
1458
8b9cd157
MK		 1459static int able_to_resync(int raid_level, int missing_disks)
1460{
1461 int max_missing_disks = 0;
1462
1463 switch (raid_level) {
1464 case 10:
1465 max_missing_disks = 1;
1466 break;
1467 default:
1468 max_missing_disks = 0;
1469 }
1470 return missing_disks <= max_missing_disks;
1471}
1472
b81221b7
CA		 1473/* try to determine how much space is reserved for metadata from
1474 * the last get_extents() entry on the smallest active disk,
1475 * otherwise fallback to the default
1476 */
1477static __u32 imsm_min_reserved_sectors(struct intel_super *super)
1478{
1479 struct extent *e;
1480 int i;
5551b113
CA		 1481 unsigned long long min_active;
1482 __u32 remainder;
b81221b7
CA		 1483 __u32 rv = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
1484 struct dl *dl, *dl_min = NULL;
1485
1486 if (!super)
1487 return rv;
1488
1489 min_active = 0;
1490 for (dl = super->disks; dl; dl = dl->next) {
1491 if (dl->index < 0)
1492 continue;
5551b113
CA		 1493 unsigned long long blocks = total_blocks(&dl->disk);
1494 if (blocks < min_active || min_active == 0) {
b81221b7 1495 dl_min = dl;
5551b113 1496 min_active = blocks;
b81221b7
CA		 1497 }
1498 }
1499 if (!dl_min)
1500 return rv;
1501
1502 /* find last lba used by subarrays on the smallest active disk */
05501181 1503 e = get_extents(super, dl_min, 0);
b81221b7
CA		 1504 if (!e)
1505 return rv;
1506 for (i = 0; e[i].size; i++)
1507 continue;
1508
1509 remainder = min_active - e[i].start;
1510 free(e);
1511
1512 /* to give priority to recovery we should not require full
1513 IMSM_RESERVED_SECTORS from the spare */
1514 rv = MPB_SECTOR_CNT + NUM_BLOCKS_DIRTY_STRIPE_REGION;
1515
1516 /* if real reservation is smaller use that value */
1517 return (remainder < rv) ? remainder : rv;
1518}
1519
fbfdcb06
AO		 1520/*
1521 * Return minimum size of a spare and sector size
1522 * that can be used in this array
1523 */
1524int get_spare_criteria_imsm(struct supertype *st, struct spare_criteria *c)
80e7f8c3
AC		 1525{
1526 struct intel_super *super = st->sb;
1527 struct dl *dl;
1528 struct extent *e;
1529 int i;
fbfdcb06
AO		 1530 unsigned long long size = 0;
1531
1532 c->min_size = 0;
4b57ecf6 1533 c->sector_size = 0;
80e7f8c3
AC		 1534
1535 if (!super)
fbfdcb06 1536 return -EINVAL;
80e7f8c3
AC		 1537 /* find first active disk in array */
1538 dl = super->disks;
1539 while (dl && (is_failed(&dl->disk) || dl->index == -1))
1540 dl = dl->next;
1541 if (!dl)
fbfdcb06 1542 return -EINVAL;
80e7f8c3 1543 /* find last lba used by subarrays */
05501181 1544 e = get_extents(super, dl, 0);
80e7f8c3 1545 if (!e)
fbfdcb06 1546 return -EINVAL;
80e7f8c3
AC		 1547 for (i = 0; e[i].size; i++)
1548 continue;
1549 if (i > 0)
fbfdcb06 1550 size = e[i-1].start + e[i-1].size;
80e7f8c3 1551 free(e);
b81221b7 1552
80e7f8c3 1553 /* add the amount of space needed for metadata */
fbfdcb06
AO		 1554 size += imsm_min_reserved_sectors(super);
1555
1556 c->min_size = size * 512;
4b57ecf6 1557 c->sector_size = super->sector_size;
b81221b7 1558
fbfdcb06 1559 return 0;
80e7f8c3
AC		 1560}
1561
d1e02575
AK		 1562static int is_gen_migration(struct imsm_dev *dev);
1563
f36a9ecd
PB		 1564#define IMSM_4K_DIV 8
1565
c47b0ff6
AK		 1566static __u64 blocks_per_migr_unit(struct intel_super *super,
1567 struct imsm_dev *dev);
1e5c6983 1568
c47b0ff6
AK		 1569static void print_imsm_dev(struct intel_super *super,
1570 struct imsm_dev *dev,
1571 char *uuid,
1572 int disk_idx)
cdddbdbc
DW		 1573{
1574 __u64 sz;
0d80bb2f 1575 int slot, i;
238c0a71
AK		 1576 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1577 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
b10b37b8 1578 __u32 ord;
cdddbdbc
DW		 1579
1580 printf("\n");
1e7bc0ed 1581 printf("[%.16s]:\n", dev->volume);
ba1b3bc8 1582 printf(" Subarray : %d\n", super->current_vol);
44470971 1583 printf(" UUID : %s\n", uuid);
dd8bcb3b
AK		 1584 printf(" RAID Level : %d", get_imsm_raid_level(map));
1585 if (map2)
1586 printf(" <-- %d", get_imsm_raid_level(map2));
1587 printf("\n");
1588 printf(" Members : %d", map->num_members);
1589 if (map2)
1590 printf(" <-- %d", map2->num_members);
1591 printf("\n");
0d80bb2f
DW		 1592 printf(" Slots : [");
1593 for (i = 0; i < map->num_members; i++) {
238c0a71 1594 ord = get_imsm_ord_tbl_ent(dev, i, MAP_0);
0d80bb2f
DW		 1595 printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U");
1596 }
dd8bcb3b
AK		 1597 printf("]");
1598 if (map2) {
1599 printf(" <-- [");
1600 for (i = 0; i < map2->num_members; i++) {
238c0a71 1601 ord = get_imsm_ord_tbl_ent(dev, i, MAP_1);
dd8bcb3b
AK		 1602 printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U");
1603 }
1604 printf("]");
1605 }
1606 printf("\n");
7095bccb
AK		 1607 printf(" Failed disk : ");
1608 if (map->failed_disk_num == 0xff)
1609 printf("none");
1610 else
1611 printf("%i", map->failed_disk_num);
1612 printf("\n");
620b1713
DW		 1613 slot = get_imsm_disk_slot(map, disk_idx);
1614 if (slot >= 0) {
238c0a71 1615 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
b10b37b8
DW		 1616 printf(" This Slot : %d%s\n", slot,
1617 ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : "");
1618 } else
cdddbdbc 1619 printf(" This Slot : ?\n");
84918897 1620 printf(" Sector Size : %u\n", super->sector_size);
fcc2c9da 1621 sz = imsm_dev_size(dev);
84918897
MK		 1622 printf(" Array Size : %llu%s\n",
1623 (unsigned long long)sz * 512 / super->sector_size,
cdddbdbc 1624 human_size(sz * 512));
5551b113 1625 sz = blocks_per_member(map);
84918897
MK		 1626 printf(" Per Dev Size : %llu%s\n",
1627 (unsigned long long)sz * 512 / super->sector_size,
cdddbdbc 1628 human_size(sz * 512));
5551b113
CA		 1629 printf(" Sector Offset : %llu\n",
1630 pba_of_lba0(map));
1631 printf(" Num Stripes : %llu\n",
1632 num_data_stripes(map));
dd8bcb3b 1633 printf(" Chunk Size : %u KiB",
cdddbdbc 1634 __le16_to_cpu(map->blocks_per_strip) / 2);
dd8bcb3b
AK		 1635 if (map2)
1636 printf(" <-- %u KiB",
1637 __le16_to_cpu(map2->blocks_per_strip) / 2);
1638 printf("\n");
cdddbdbc 1639 printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks));
8655a7b1 1640 printf(" Migrate State : ");
1484e727
DW		 1641 if (dev->vol.migr_state) {
1642 if (migr_type(dev) == MIGR_INIT)
8655a7b1 1643 printf("initialize\n");
1484e727 1644 else if (migr_type(dev) == MIGR_REBUILD)
8655a7b1 1645 printf("rebuild\n");
1484e727 1646 else if (migr_type(dev) == MIGR_VERIFY)
8655a7b1 1647 printf("check\n");
1484e727 1648 else if (migr_type(dev) == MIGR_GEN_MIGR)
8655a7b1 1649 printf("general migration\n");
1484e727 1650 else if (migr_type(dev) == MIGR_STATE_CHANGE)
8655a7b1 1651 printf("state change\n");
1484e727 1652 else if (migr_type(dev) == MIGR_REPAIR)
8655a7b1 1653 printf("repair\n");
1484e727 1654 else
8655a7b1
DW		 1655 printf("<unknown:%d>\n", migr_type(dev));
1656 } else
1657 printf("idle\n");
3393c6af
DW		 1658 printf(" Map State : %s", map_state_str[map->map_state]);
1659 if (dev->vol.migr_state) {
238c0a71 1660 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983 1661
b10b37b8 1662 printf(" <-- %s", map_state_str[map->map_state]);
464d40e8
LD		 1663 printf("\n Checkpoint : %u ",
1664 __le32_to_cpu(dev->vol.curr_migr_unit));
089f9d79 1665 if (is_gen_migration(dev) && (slot > 1 || slot < 0))
464d40e8
LD		 1666 printf("(N/A)");
1667 else
1668 printf("(%llu)", (unsigned long long)
1669 blocks_per_migr_unit(super, dev));
3393c6af
DW		 1670 }
1671 printf("\n");
2432ce9b
AP		 1672 printf(" Dirty State : %s\n", (dev->vol.dirty & RAIDVOL_DIRTY) ?
1673 "dirty" : "clean");
1674 printf(" RWH Policy : ");
c2462068 1675 if (dev->rwh_policy == RWH_OFF || dev->rwh_policy == RWH_MULTIPLE_OFF)
2432ce9b
AP		 1676 printf("off\n");
1677 else if (dev->rwh_policy == RWH_DISTRIBUTED)
1678 printf("PPL distributed\n");
1679 else if (dev->rwh_policy == RWH_JOURNALING_DRIVE)
1680 printf("PPL journaling drive\n");
c2462068
PB		 1681 else if (dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
1682 printf("Multiple distributed PPLs\n");
1683 else if (dev->rwh_policy == RWH_MULTIPLE_PPLS_JOURNALING_DRIVE)
1684 printf("Multiple PPLs on journaling drive\n");
2432ce9b
AP		 1685 else
1686 printf("<unknown:%d>\n", dev->rwh_policy);
ba1b3bc8
AP		 1687
1688 printf(" Volume ID : %u\n", dev->my_vol_raid_dev_num);
cdddbdbc
DW		 1689}
1690
ef5c214e
MK		 1691static void print_imsm_disk(struct imsm_disk *disk,
1692 int index,
1693 __u32 reserved,
1694 unsigned int sector_size) {
1f24f035 1695 char str[MAX_RAID_SERIAL_LEN + 1];
cdddbdbc
DW		 1696 __u64 sz;
1697
0ec1f4e8 1698 if (index < -1 || !disk)
e9d82038
DW		 1699 return;
1700
cdddbdbc 1701 printf("\n");
1f24f035 1702 snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial);
0ec1f4e8
DW		 1703 if (index >= 0)
1704 printf(" Disk%02d Serial : %s\n", index, str);
1705 else
1706 printf(" Disk Serial : %s\n", str);
2432ce9b
AP		 1707 printf(" State :%s%s%s%s\n", is_spare(disk) ? " spare" : "",
1708 is_configured(disk) ? " active" : "",
1709 is_failed(disk) ? " failed" : "",
1710 is_journal(disk) ? " journal" : "");
cdddbdbc 1711 printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id));
5551b113 1712 sz = total_blocks(disk) - reserved;
ef5c214e
MK		 1713 printf(" Usable Size : %llu%s\n",
1714 (unsigned long long)sz * 512 / sector_size,
cdddbdbc
DW		 1715 human_size(sz * 512));
1716}
1717
de44e46f
PB		 1718void convert_to_4k_imsm_migr_rec(struct intel_super *super)
1719{
1720 struct migr_record *migr_rec = super->migr_rec;
1721
1722 migr_rec->blocks_per_unit /= IMSM_4K_DIV;
de44e46f
PB		 1723 migr_rec->dest_depth_per_unit /= IMSM_4K_DIV;
1724 split_ull((join_u32(migr_rec->post_migr_vol_cap,
1725 migr_rec->post_migr_vol_cap_hi) / IMSM_4K_DIV),
1726 &migr_rec->post_migr_vol_cap, &migr_rec->post_migr_vol_cap_hi);
9f421827
PB		 1727 set_migr_chkp_area_pba(migr_rec,
1728 migr_chkp_area_pba(migr_rec) / IMSM_4K_DIV);
1729 set_migr_dest_1st_member_lba(migr_rec,
1730 migr_dest_1st_member_lba(migr_rec) / IMSM_4K_DIV);
de44e46f
PB		 1731}
1732
f36a9ecd
PB		 1733void convert_to_4k_imsm_disk(struct imsm_disk *disk)
1734{
1735 set_total_blocks(disk, (total_blocks(disk)/IMSM_4K_DIV));
1736}
1737
1738void convert_to_4k(struct intel_super *super)
1739{
1740 struct imsm_super *mpb = super->anchor;
1741 struct imsm_disk *disk;
1742 int i;
e4467bc7 1743 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
f36a9ecd
PB		 1744
1745 for (i = 0; i < mpb->num_disks ; i++) {
1746 disk = __get_imsm_disk(mpb, i);
1747 /* disk */
1748 convert_to_4k_imsm_disk(disk);
1749 }
1750 for (i = 0; i < mpb->num_raid_devs; i++) {
1751 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
1752 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1753 /* dev */
fcc2c9da 1754 set_imsm_dev_size(dev, imsm_dev_size(dev)/IMSM_4K_DIV);
f36a9ecd
PB		 1755 dev->vol.curr_migr_unit /= IMSM_4K_DIV;
1756
1757 /* map0 */
1758 set_blocks_per_member(map, blocks_per_member(map)/IMSM_4K_DIV);
1759 map->blocks_per_strip /= IMSM_4K_DIV;
1760 set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV);
1761
1762 if (dev->vol.migr_state) {
1763 /* map1 */
1764 map = get_imsm_map(dev, MAP_1);
1765 set_blocks_per_member(map,
1766 blocks_per_member(map)/IMSM_4K_DIV);
1767 map->blocks_per_strip /= IMSM_4K_DIV;
1768 set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV);
1769 }
1770 }
e4467bc7
TM		 1771 if (bbm_log_size) {
1772 struct bbm_log *log = (void *)mpb +
1773 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
1774 __u32 i;
1775
1776 for (i = 0; i < log->entry_count; i++) {
1777 struct bbm_log_entry *entry =
1778 &log->marked_block_entries[i];
1779
1780 __u8 count = entry->marked_count + 1;
1781 unsigned long long sector =
1782 __le48_to_cpu(&entry->defective_block_start);
1783
1784 entry->defective_block_start =
1785 __cpu_to_le48(sector/IMSM_4K_DIV);
1786 entry->marked_count = max(count/IMSM_4K_DIV, 1) - 1;
1787 }
1788 }
f36a9ecd
PB		 1789
1790 mpb->check_sum = __gen_imsm_checksum(mpb);
1791}
1792
520e69e2
AK		 1793void examine_migr_rec_imsm(struct intel_super *super)
1794{
1795 struct migr_record *migr_rec = super->migr_rec;
1796 struct imsm_super *mpb = super->anchor;
1797 int i;
1798
1799 for (i = 0; i < mpb->num_raid_devs; i++) {
1800 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
3136abe5 1801 struct imsm_map *map;
b4ab44d8 1802 int slot = -1;
3136abe5 1803
520e69e2
AK		 1804 if (is_gen_migration(dev) == 0)
1805 continue;
1806
1807 printf("\nMigration Record Information:");
3136abe5 1808
44bfe6df
AK		 1809 /* first map under migration */
1810 map = get_imsm_map(dev, MAP_0);
3136abe5
AK		 1811 if (map)
1812 slot = get_imsm_disk_slot(map, super->disks->index);
089f9d79 1813 if (map == NULL || slot > 1 || slot < 0) {
520e69e2
AK		 1814 printf(" Empty\n ");
1815 printf("Examine one of first two disks in array\n");
1816 break;
1817 }
1818 printf("\n Status : ");
1819 if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL)
1820 printf("Normal\n");
1821 else
1822 printf("Contains Data\n");
9f421827
PB		 1823 printf(" Current Unit : %llu\n",
1824 current_migr_unit(migr_rec));
520e69e2
AK		 1825 printf(" Family : %u\n",
1826 __le32_to_cpu(migr_rec->family_num));
1827 printf(" Ascending : %u\n",
1828 __le32_to_cpu(migr_rec->ascending_migr));
1829 printf(" Blocks Per Unit : %u\n",
1830 __le32_to_cpu(migr_rec->blocks_per_unit));
1831 printf(" Dest. Depth Per Unit : %u\n",
1832 __le32_to_cpu(migr_rec->dest_depth_per_unit));
9f421827
PB		 1833 printf(" Checkpoint Area pba : %llu\n",
1834 migr_chkp_area_pba(migr_rec));
1835 printf(" First member lba : %llu\n",
1836 migr_dest_1st_member_lba(migr_rec));
1837 printf(" Total Number of Units : %llu\n",
1838 get_num_migr_units(migr_rec));
1839 printf(" Size of volume : %llu\n",
1840 join_u32(migr_rec->post_migr_vol_cap,
1841 migr_rec->post_migr_vol_cap_hi));
520e69e2
AK		 1842 printf(" Record was read from : %u\n",
1843 __le32_to_cpu(migr_rec->ckpt_read_disk_num));
1844
1845 break;
1846 }
1847}
f36a9ecd 1848
de44e46f
PB		 1849void convert_from_4k_imsm_migr_rec(struct intel_super *super)
1850{
1851 struct migr_record *migr_rec = super->migr_rec;
1852
1853 migr_rec->blocks_per_unit *= IMSM_4K_DIV;
de44e46f
PB		 1854 migr_rec->dest_depth_per_unit *= IMSM_4K_DIV;
1855 split_ull((join_u32(migr_rec->post_migr_vol_cap,
1856 migr_rec->post_migr_vol_cap_hi) * IMSM_4K_DIV),
1857 &migr_rec->post_migr_vol_cap,
1858 &migr_rec->post_migr_vol_cap_hi);
9f421827
PB		 1859 set_migr_chkp_area_pba(migr_rec,
1860 migr_chkp_area_pba(migr_rec) * IMSM_4K_DIV);
1861 set_migr_dest_1st_member_lba(migr_rec,
1862 migr_dest_1st_member_lba(migr_rec) * IMSM_4K_DIV);
de44e46f
PB		 1863}
1864
f36a9ecd
PB		 1865void convert_from_4k(struct intel_super *super)
1866{
1867 struct imsm_super *mpb = super->anchor;
1868 struct imsm_disk *disk;
1869 int i;
e4467bc7 1870 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
f36a9ecd
PB		 1871
1872 for (i = 0; i < mpb->num_disks ; i++) {
1873 disk = __get_imsm_disk(mpb, i);
1874 /* disk */
1875 set_total_blocks(disk, (total_blocks(disk)*IMSM_4K_DIV));
1876 }
1877
1878 for (i = 0; i < mpb->num_raid_devs; i++) {
1879 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
1880 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1881 /* dev */
fcc2c9da 1882 set_imsm_dev_size(dev, imsm_dev_size(dev)*IMSM_4K_DIV);
f36a9ecd
PB		 1883 dev->vol.curr_migr_unit *= IMSM_4K_DIV;
1884
1885 /* map0 */
1886 set_blocks_per_member(map, blocks_per_member(map)*IMSM_4K_DIV);
1887 map->blocks_per_strip *= IMSM_4K_DIV;
1888 set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV);
1889
1890 if (dev->vol.migr_state) {
1891 /* map1 */
1892 map = get_imsm_map(dev, MAP_1);
1893 set_blocks_per_member(map,
1894 blocks_per_member(map)*IMSM_4K_DIV);
1895 map->blocks_per_strip *= IMSM_4K_DIV;
1896 set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV);
1897 }
1898 }
e4467bc7
TM		 1899 if (bbm_log_size) {
1900 struct bbm_log *log = (void *)mpb +
1901 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
1902 __u32 i;
1903
1904 for (i = 0; i < log->entry_count; i++) {
1905 struct bbm_log_entry *entry =
1906 &log->marked_block_entries[i];
1907
1908 __u8 count = entry->marked_count + 1;
1909 unsigned long long sector =
1910 __le48_to_cpu(&entry->defective_block_start);
1911
1912 entry->defective_block_start =
1913 __cpu_to_le48(sector*IMSM_4K_DIV);
1914 entry->marked_count = count*IMSM_4K_DIV - 1;
1915 }
1916 }
f36a9ecd
PB		 1917
1918 mpb->check_sum = __gen_imsm_checksum(mpb);
1919}
1920
19482bcc
AK		 1921/*******************************************************************************
1922 * function: imsm_check_attributes
1923 * Description: Function checks if features represented by attributes flags
1011e834 1924 * are supported by mdadm.
19482bcc
AK		 1925 * Parameters:
1926 * attributes - Attributes read from metadata
1927 * Returns:
1011e834
N		 1928 * 0 - passed attributes contains unsupported features flags
1929 * 1 - all features are supported
19482bcc
AK		 1930 ******************************************************************************/
1931static int imsm_check_attributes(__u32 attributes)
1932{
1933 int ret_val = 1;
418f9b36
N		 1934 __u32 not_supported = MPB_ATTRIB_SUPPORTED^0xffffffff;
1935
1936 not_supported &= ~MPB_ATTRIB_IGNORED;
19482bcc
AK		 1937
1938 not_supported &= attributes;
1939 if (not_supported) {
e7b84f9d 1940 pr_err("(IMSM): Unsupported attributes : %x\n",
418f9b36 1941 (unsigned)__le32_to_cpu(not_supported));
19482bcc
AK		 1942 if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) {
1943 dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY \n");
1944 not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY;
1945 }
1946 if (not_supported & MPB_ATTRIB_2TB) {
1947 dprintf("\t\tMPB_ATTRIB_2TB\n");
1948 not_supported ^= MPB_ATTRIB_2TB;
1949 }
1950 if (not_supported & MPB_ATTRIB_RAID0) {
1951 dprintf("\t\tMPB_ATTRIB_RAID0\n");
1952 not_supported ^= MPB_ATTRIB_RAID0;
1953 }
1954 if (not_supported & MPB_ATTRIB_RAID1) {
1955 dprintf("\t\tMPB_ATTRIB_RAID1\n");
1956 not_supported ^= MPB_ATTRIB_RAID1;
1957 }
1958 if (not_supported & MPB_ATTRIB_RAID10) {
1959 dprintf("\t\tMPB_ATTRIB_RAID10\n");
1960 not_supported ^= MPB_ATTRIB_RAID10;
1961 }
1962 if (not_supported & MPB_ATTRIB_RAID1E) {
1963 dprintf("\t\tMPB_ATTRIB_RAID1E\n");
1964 not_supported ^= MPB_ATTRIB_RAID1E;
1965 }
1966 if (not_supported & MPB_ATTRIB_RAID5) {
1967 dprintf("\t\tMPB_ATTRIB_RAID5\n");
1968 not_supported ^= MPB_ATTRIB_RAID5;
1969 }
1970 if (not_supported & MPB_ATTRIB_RAIDCNG) {
1971 dprintf("\t\tMPB_ATTRIB_RAIDCNG\n");
1972 not_supported ^= MPB_ATTRIB_RAIDCNG;
1973 }
1974 if (not_supported & MPB_ATTRIB_BBM) {
1975 dprintf("\t\tMPB_ATTRIB_BBM\n");
1976 not_supported ^= MPB_ATTRIB_BBM;
1977 }
1978 if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) {
1979 dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY (== MPB_ATTRIB_LEGACY)\n");
1980 not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY;
1981 }
1982 if (not_supported & MPB_ATTRIB_EXP_STRIPE_SIZE) {
1983 dprintf("\t\tMPB_ATTRIB_EXP_STRIP_SIZE\n");
1984 not_supported ^= MPB_ATTRIB_EXP_STRIPE_SIZE;
1985 }
1986 if (not_supported & MPB_ATTRIB_2TB_DISK) {
1987 dprintf("\t\tMPB_ATTRIB_2TB_DISK\n");
1988 not_supported ^= MPB_ATTRIB_2TB_DISK;
1989 }
1990 if (not_supported & MPB_ATTRIB_NEVER_USE2) {
1991 dprintf("\t\tMPB_ATTRIB_NEVER_USE2\n");
1992 not_supported ^= MPB_ATTRIB_NEVER_USE2;
1993 }
1994 if (not_supported & MPB_ATTRIB_NEVER_USE) {
1995 dprintf("\t\tMPB_ATTRIB_NEVER_USE\n");
1996 not_supported ^= MPB_ATTRIB_NEVER_USE;
1997 }
1998
1999 if (not_supported)
1ade5cc1 2000 dprintf("(IMSM): Unknown attributes : %x\n", not_supported);
19482bcc
AK		 2001
2002 ret_val = 0;
2003 }
2004
2005 return ret_val;
2006}
2007
a5d85af7 2008static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map);
44470971 2009
cdddbdbc
DW		 2010static void examine_super_imsm(struct supertype *st, char *homehost)
2011{
2012 struct intel_super *super = st->sb;
949c47a0 2013 struct imsm_super *mpb = super->anchor;
cdddbdbc
DW		 2014 char str[MAX_SIGNATURE_LENGTH];
2015 int i;
27fd6274
DW		 2016 struct mdinfo info;
2017 char nbuf[64];
cdddbdbc 2018 __u32 sum;
14e8215b 2019 __u32 reserved = imsm_reserved_sectors(super, super->disks);
94827db3 2020 struct dl *dl;
e48aed3c 2021 time_t creation_time;
27fd6274 2022
618f4e6d
XN		 2023 strncpy(str, (char *)mpb->sig, MPB_SIG_LEN);
2024 str[MPB_SIG_LEN-1] = '\0';
cdddbdbc 2025 printf(" Magic : %s\n", str);
cdddbdbc 2026 printf(" Version : %s\n", get_imsm_version(mpb));
148acb7b 2027 printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num));
cdddbdbc
DW		 2028 printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num));
2029 printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num));
e48aed3c
AP		 2030 creation_time = __le64_to_cpu(mpb->creation_time);
2031 printf(" Creation Time : %.24s\n",
2032 creation_time ? ctime(&creation_time) : "Unknown");
19482bcc
AK		 2033 printf(" Attributes : ");
2034 if (imsm_check_attributes(mpb->attributes))
2035 printf("All supported\n");
2036 else
2037 printf("not supported\n");
a5d85af7 2038 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2039 fname_from_uuid(st, &info, nbuf, ':');
27fd6274 2040 printf(" UUID : %s\n", nbuf + 5);
cdddbdbc
DW		 2041 sum = __le32_to_cpu(mpb->check_sum);
2042 printf(" Checksum : %08x %s\n", sum,
949c47a0 2043 __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect");
f36a9ecd 2044 printf(" MPB Sectors : %d\n", mpb_sectors(mpb, super->sector_size));
cdddbdbc
DW		 2045 printf(" Disks : %d\n", mpb->num_disks);
2046 printf(" RAID Devices : %d\n", mpb->num_raid_devs);
ef5c214e
MK		 2047 print_imsm_disk(__get_imsm_disk(mpb, super->disks->index),
2048 super->disks->index, reserved, super->sector_size);
8d67477f 2049 if (get_imsm_bbm_log_size(super->bbm_log)) {
604b746f
JD		 2050 struct bbm_log *log = super->bbm_log;
2051
2052 printf("\n");
2053 printf("Bad Block Management Log:\n");
2054 printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size));
2055 printf(" Signature : %x\n", __le32_to_cpu(log->signature));
2056 printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count));
604b746f 2057 }
44470971
DW		 2058 for (i = 0; i < mpb->num_raid_devs; i++) {
2059 struct mdinfo info;
2060 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
2061
2062 super->current_vol = i;
a5d85af7 2063 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2064 fname_from_uuid(st, &info, nbuf, ':');
c47b0ff6 2065 print_imsm_dev(super, dev, nbuf + 5, super->disks->index);
44470971 2066 }
cdddbdbc
DW		 2067 for (i = 0; i < mpb->num_disks; i++) {
2068 if (i == super->disks->index)
2069 continue;
ef5c214e
MK		 2070 print_imsm_disk(__get_imsm_disk(mpb, i), i, reserved,
2071 super->sector_size);
cdddbdbc 2072 }
94827db3 2073
0ec1f4e8
DW		 2074 for (dl = super->disks; dl; dl = dl->next)
2075 if (dl->index == -1)
ef5c214e
MK		 2076 print_imsm_disk(&dl->disk, -1, reserved,
2077 super->sector_size);
520e69e2
AK		 2078
2079 examine_migr_rec_imsm(super);
cdddbdbc
DW		 2080}
2081
061f2c6a 2082static void brief_examine_super_imsm(struct supertype *st, int verbose)
cdddbdbc 2083{
27fd6274 2084 /* We just write a generic IMSM ARRAY entry */
ff54de6e
N		 2085 struct mdinfo info;
2086 char nbuf[64];
1e7bc0ed 2087 struct intel_super *super = st->sb;
1e7bc0ed 2088
0d5a423f
DW		 2089 if (!super->anchor->num_raid_devs) {
2090 printf("ARRAY metadata=imsm\n");
1e7bc0ed 2091 return;
0d5a423f 2092 }
ff54de6e 2093
a5d85af7 2094 getinfo_super_imsm(st, &info, NULL);
4737ae25
N		 2095 fname_from_uuid(st, &info, nbuf, ':');
2096 printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5);
2097}
2098
2099static void brief_examine_subarrays_imsm(struct supertype *st, int verbose)
2100{
2101 /* We just write a generic IMSM ARRAY entry */
2102 struct mdinfo info;
2103 char nbuf[64];
2104 char nbuf1[64];
2105 struct intel_super *super = st->sb;
2106 int i;
2107
2108 if (!super->anchor->num_raid_devs)
2109 return;
2110
a5d85af7 2111 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2112 fname_from_uuid(st, &info, nbuf, ':');
1e7bc0ed
DW		 2113 for (i = 0; i < super->anchor->num_raid_devs; i++) {
2114 struct imsm_dev *dev = get_imsm_dev(super, i);
2115
2116 super->current_vol = i;
a5d85af7 2117 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2118 fname_from_uuid(st, &info, nbuf1, ':');
1124b3cf 2119 printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n",
cf8de691 2120 dev->volume, nbuf + 5, i, nbuf1 + 5);
1e7bc0ed 2121 }
cdddbdbc
DW		 2122}
2123
9d84c8ea
DW		 2124static void export_examine_super_imsm(struct supertype *st)
2125{
2126 struct intel_super *super = st->sb;
2127 struct imsm_super *mpb = super->anchor;
2128 struct mdinfo info;
2129 char nbuf[64];
2130
a5d85af7 2131 getinfo_super_imsm(st, &info, NULL);
9d84c8ea
DW		 2132 fname_from_uuid(st, &info, nbuf, ':');
2133 printf("MD_METADATA=imsm\n");
2134 printf("MD_LEVEL=container\n");
2135 printf("MD_UUID=%s\n", nbuf+5);
2136 printf("MD_DEVICES=%u\n", mpb->num_disks);
e48aed3c 2137 printf("MD_CREATION_TIME=%llu\n", __le64_to_cpu(mpb->creation_time));
9d84c8ea
DW		 2138}
2139
b771faef
BK		 2140static void detail_super_imsm(struct supertype *st, char *homehost,
2141 char *subarray)
cdddbdbc 2142{
3ebe00a1
DW		 2143 struct mdinfo info;
2144 char nbuf[64];
b771faef
BK		 2145 struct intel_super *super = st->sb;
2146 int temp_vol = super->current_vol;
2147
2148 if (subarray)
2149 super->current_vol = strtoul(subarray, NULL, 10);
3ebe00a1 2150
a5d85af7 2151 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2152 fname_from_uuid(st, &info, nbuf, ':');
65884368 2153 printf("\n UUID : %s\n", nbuf + 5);
b771faef
BK		 2154
2155 super->current_vol = temp_vol;
cdddbdbc
DW		 2156}
2157
b771faef 2158static void brief_detail_super_imsm(struct supertype *st, char *subarray)
cdddbdbc 2159{
ff54de6e
N		 2160 struct mdinfo info;
2161 char nbuf[64];
b771faef
BK		 2162 struct intel_super *super = st->sb;
2163 int temp_vol = super->current_vol;
2164
2165 if (subarray)
2166 super->current_vol = strtoul(subarray, NULL, 10);
2167
a5d85af7 2168 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2169 fname_from_uuid(st, &info, nbuf, ':');
ff54de6e 2170 printf(" UUID=%s", nbuf + 5);
b771faef
BK		 2171
2172 super->current_vol = temp_vol;
cdddbdbc 2173}
d665cc31 2174
6da53c0e
BK		 2175static int imsm_read_serial(int fd, char *devname, __u8 *serial,
2176 size_t serial_buf_len);
d665cc31
DW		 2177static void fd2devname(int fd, char *name);
2178
120dc887 2179static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose)
d665cc31 2180{
120dc887
LM		 2181 /* dump an unsorted list of devices attached to AHCI Intel storage
2182 * controller, as well as non-connected ports
d665cc31
DW		 2183 */
2184 int hba_len = strlen(hba_path) + 1;
2185 struct dirent *ent;
2186 DIR *dir;
2187 char *path = NULL;
2188 int err = 0;
2189 unsigned long port_mask = (1 << port_count) - 1;
2190
f21e18ca 2191 if (port_count > (int)sizeof(port_mask) * 8) {
ba728be7 2192 if (verbose > 0)
e7b84f9d 2193 pr_err("port_count %d out of range\n", port_count);
d665cc31
DW		 2194 return 2;
2195 }
2196
2197 /* scroll through /sys/dev/block looking for devices attached to
2198 * this hba
2199 */
2200 dir = opendir("/sys/dev/block");
1a6dd6b9
PB		 2201 if (!dir)
2202 return 1;
2203
2204 for (ent = readdir(dir); ent; ent = readdir(dir)) {
d665cc31
DW		 2205 int fd;
2206 char model[64];
2207 char vendor[64];
2208 char buf[1024];
2209 int major, minor;
2210 char *device;
2211 char *c;
2212 int port;
2213 int type;
2214
2215 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
2216 continue;
2217 path = devt_to_devpath(makedev(major, minor));
2218 if (!path)
2219 continue;
2220 if (!path_attached_to_hba(path, hba_path)) {
2221 free(path);
2222 path = NULL;
2223 continue;
2224 }
2225
2226 /* retrieve the scsi device type */
2227 if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) {
ba728be7 2228 if (verbose > 0)
e7b84f9d 2229 pr_err("failed to allocate 'device'\n");
d665cc31
DW		 2230 err = 2;
2231 break;
2232 }
2233 sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor);
193b6c0b 2234 if (load_sys(device, buf, sizeof(buf)) != 0) {
ba728be7 2235 if (verbose > 0)
e7b84f9d 2236 pr_err("failed to read device type for %s\n",
d665cc31
DW		 2237 path);
2238 err = 2;
2239 free(device);
2240 break;
2241 }
2242 type = strtoul(buf, NULL, 10);
2243
2244 /* if it's not a disk print the vendor and model */
2245 if (!(type == 0 || type == 7 || type == 14)) {
2246 vendor[0] = '\0';
2247 model[0] = '\0';
2248 sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor);
193b6c0b 2249 if (load_sys(device, buf, sizeof(buf)) == 0) {
d665cc31
DW		 2250 strncpy(vendor, buf, sizeof(vendor));
2251 vendor[sizeof(vendor) - 1] = '\0';
2252 c = (char *) &vendor[sizeof(vendor) - 1];
2253 while (isspace(*c) || *c == '\0')
2254 *c-- = '\0';
2255
2256 }
2257 sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor);
193b6c0b 2258 if (load_sys(device, buf, sizeof(buf)) == 0) {
d665cc31
DW		 2259 strncpy(model, buf, sizeof(model));
2260 model[sizeof(model) - 1] = '\0';
2261 c = (char *) &model[sizeof(model) - 1];
2262 while (isspace(*c) || *c == '\0')
2263 *c-- = '\0';
2264 }
2265
2266 if (vendor[0] && model[0])
2267 sprintf(buf, "%.64s %.64s", vendor, model);
2268 else
2269 switch (type) { /* numbers from hald/linux/device.c */
2270 case 1: sprintf(buf, "tape"); break;
2271 case 2: sprintf(buf, "printer"); break;
2272 case 3: sprintf(buf, "processor"); break;
2273 case 4:
2274 case 5: sprintf(buf, "cdrom"); break;
2275 case 6: sprintf(buf, "scanner"); break;
2276 case 8: sprintf(buf, "media_changer"); break;
2277 case 9: sprintf(buf, "comm"); break;
2278 case 12: sprintf(buf, "raid"); break;
2279 default: sprintf(buf, "unknown");
2280 }
2281 } else
2282 buf[0] = '\0';
2283 free(device);
2284
2285 /* chop device path to 'host%d' and calculate the port number */
2286 c = strchr(&path[hba_len], '/');
4e5e717d 2287 if (!c) {
ba728be7 2288 if (verbose > 0)
e7b84f9d 2289 pr_err("%s - invalid path name\n", path + hba_len);
4e5e717d
AW		 2290 err = 2;
2291 break;
2292 }
d665cc31 2293 *c = '\0';
0858eccf
AP		 2294 if ((sscanf(&path[hba_len], "ata%d", &port) == 1) ||
2295 ((sscanf(&path[hba_len], "host%d", &port) == 1)))
d665cc31
DW		 2296 port -= host_base;
2297 else {
ba728be7 2298 if (verbose > 0) {
d665cc31 2299 *c = '/'; /* repair the full string */
e7b84f9d 2300 pr_err("failed to determine port number for %s\n",
d665cc31
DW		 2301 path);
2302 }
2303 err = 2;
2304 break;
2305 }
2306
2307 /* mark this port as used */
2308 port_mask &= ~(1 << port);
2309
2310 /* print out the device information */
2311 if (buf[0]) {
2312 printf(" Port%d : - non-disk device (%s) -\n", port, buf);
2313 continue;
2314 }
2315
2316 fd = dev_open(ent->d_name, O_RDONLY);
2317 if (fd < 0)
2318 printf(" Port%d : - disk info unavailable -\n", port);
2319 else {
2320 fd2devname(fd, buf);
2321 printf(" Port%d : %s", port, buf);
6da53c0e
BK		 2322 if (imsm_read_serial(fd, NULL, (__u8 *)buf,
2323 sizeof(buf)) == 0)
2324 printf(" (%s)\n", buf);
d665cc31 2325 else
664d5325 2326 printf(" ()\n");
4dab422a 2327 close(fd);
d665cc31 2328 }
d665cc31
DW		 2329 free(path);
2330 path = NULL;
2331 }
2332 if (path)
2333 free(path);
2334 if (dir)
2335 closedir(dir);
2336 if (err == 0) {
2337 int i;
2338
2339 for (i = 0; i < port_count; i++)
2340 if (port_mask & (1 << i))
2341 printf(" Port%d : - no device attached -\n", i);
2342 }
2343
2344 return err;
2345}
2346
6da53c0e 2347static int print_nvme_info(struct sys_dev *hba)
60f0f54d 2348{
6da53c0e 2349 char buf[1024];
60f0f54d
PB		 2350 struct dirent *ent;
2351 DIR *dir;
6da53c0e
BK		 2352 char *rp;
2353 int fd;
60f0f54d 2354
6da53c0e 2355 dir = opendir("/sys/block/");
b9135011 2356 if (!dir)
b5eece69 2357 return 1;
b9135011
JS		 2358
2359 for (ent = readdir(dir); ent; ent = readdir(dir)) {
6da53c0e
BK		 2360 if (strstr(ent->d_name, "nvme")) {
2361 sprintf(buf, "/sys/block/%s", ent->d_name);
2362 rp = realpath(buf, NULL);
2363 if (!rp)
2364 continue;
2365 if (path_attached_to_hba(rp, hba->path)) {
2366 fd = open_dev(ent->d_name);
a8f3cfd5
MT		 2367 if (!imsm_is_nvme_supported(fd, 0)) {
2368 if (fd >= 0)
2369 close(fd);
6da53c0e
BK		 2370 free(rp);
2371 continue;
2372 }
60f0f54d 2373
6da53c0e
BK		 2374 fd2devname(fd, buf);
2375 if (hba->type == SYS_DEV_VMD)
2376 printf(" NVMe under VMD : %s", buf);
2377 else if (hba->type == SYS_DEV_NVME)
2378 printf(" NVMe Device : %s", buf);
2379 if (!imsm_read_serial(fd, NULL, (__u8 *)buf,
2380 sizeof(buf)))
2381 printf(" (%s)\n", buf);
2382 else
2383 printf("()\n");
2384 close(fd);
2385 }
2386 free(rp);
60f0f54d 2387 }
60f0f54d
PB		 2388 }
2389
b9135011 2390 closedir(dir);
b5eece69 2391 return 0;
60f0f54d
PB		 2392}
2393
120dc887
LM		 2394static void print_found_intel_controllers(struct sys_dev *elem)
2395{
2396 for (; elem; elem = elem->next) {
e7b84f9d 2397 pr_err("found Intel(R) ");
120dc887
LM		 2398 if (elem->type == SYS_DEV_SATA)
2399 fprintf(stderr, "SATA ");
155cbb4c
LM		 2400 else if (elem->type == SYS_DEV_SAS)
2401 fprintf(stderr, "SAS ");
0858eccf
AP		 2402 else if (elem->type == SYS_DEV_NVME)
2403 fprintf(stderr, "NVMe ");
60f0f54d
PB		 2404
2405 if (elem->type == SYS_DEV_VMD)
2406 fprintf(stderr, "VMD domain");
2407 else
2408 fprintf(stderr, "RAID controller");
2409
120dc887
LM		 2410 if (elem->pci_id)
2411 fprintf(stderr, " at %s", elem->pci_id);
2412 fprintf(stderr, ".\n");
2413 }
2414 fflush(stderr);
2415}
2416
120dc887
LM		 2417static int ahci_get_port_count(const char *hba_path, int *port_count)
2418{
2419 struct dirent *ent;
2420 DIR *dir;
2421 int host_base = -1;
2422
2423 *port_count = 0;
2424 if ((dir = opendir(hba_path)) == NULL)
2425 return -1;
2426
2427 for (ent = readdir(dir); ent; ent = readdir(dir)) {
2428 int host;
2429
0858eccf
AP		 2430 if ((sscanf(ent->d_name, "ata%d", &host) != 1) &&
2431 ((sscanf(ent->d_name, "host%d", &host) != 1)))
120dc887
LM		 2432 continue;
2433 if (*port_count == 0)
2434 host_base = host;
2435 else if (host < host_base)
2436 host_base = host;
2437
2438 if (host + 1 > *port_count + host_base)
2439 *port_count = host + 1 - host_base;
2440 }
2441 closedir(dir);
2442 return host_base;
2443}
2444
a891a3c2
LM		 2445static void print_imsm_capability(const struct imsm_orom *orom)
2446{
0858eccf
AP		 2447 printf(" Platform : Intel(R) ");
2448 if (orom->capabilities == 0 && orom->driver_features == 0)
2449 printf("Matrix Storage Manager\n");
ab0c6bb9
AP		 2450 else if (imsm_orom_is_enterprise(orom) && orom->major_ver >= 6)
2451 printf("Virtual RAID on CPU\n");
0858eccf
AP		 2452 else
2453 printf("Rapid Storage Technology%s\n",
2454 imsm_orom_is_enterprise(orom) ? " enterprise" : "");
2455 if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build)
2456 printf(" Version : %d.%d.%d.%d\n", orom->major_ver,
2457 orom->minor_ver, orom->hotfix_ver, orom->build);
a891a3c2
LM		 2458 printf(" RAID Levels :%s%s%s%s%s\n",
2459 imsm_orom_has_raid0(orom) ? " raid0" : "",
2460 imsm_orom_has_raid1(orom) ? " raid1" : "",
2461 imsm_orom_has_raid1e(orom) ? " raid1e" : "",
2462 imsm_orom_has_raid10(orom) ? " raid10" : "",
2463 imsm_orom_has_raid5(orom) ? " raid5" : "");
2464 printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
2465 imsm_orom_has_chunk(orom, 2) ? " 2k" : "",
2466 imsm_orom_has_chunk(orom, 4) ? " 4k" : "",
2467 imsm_orom_has_chunk(orom, 8) ? " 8k" : "",
2468 imsm_orom_has_chunk(orom, 16) ? " 16k" : "",
2469 imsm_orom_has_chunk(orom, 32) ? " 32k" : "",
2470 imsm_orom_has_chunk(orom, 64) ? " 64k" : "",
2471 imsm_orom_has_chunk(orom, 128) ? " 128k" : "",
2472 imsm_orom_has_chunk(orom, 256) ? " 256k" : "",
2473 imsm_orom_has_chunk(orom, 512) ? " 512k" : "",
2474 imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "",
2475 imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "",
2476 imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "",
2477 imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "",
2478 imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "",
2479 imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "",
2480 imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : "");
29cd0821
CA		 2481 printf(" 2TB volumes :%s supported\n",
2482 (orom->attr & IMSM_OROM_ATTR_2TB)?"":" not");
2483 printf(" 2TB disks :%s supported\n",
2484 (orom->attr & IMSM_OROM_ATTR_2TB_DISK)?"":" not");
0e7f69a8 2485 printf(" Max Disks : %d\n", orom->tds);
0858eccf
AP		 2486 printf(" Max Volumes : %d per array, %d per %s\n",
2487 orom->vpa, orom->vphba,
2488 imsm_orom_is_nvme(orom) ? "platform" : "controller");
a891a3c2
LM		 2489 return;
2490}
2491
e50cf220
MN		 2492static void print_imsm_capability_export(const struct imsm_orom *orom)
2493{
2494 printf("MD_FIRMWARE_TYPE=imsm\n");
0858eccf
AP		 2495 if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build)
2496 printf("IMSM_VERSION=%d.%d.%d.%d\n", orom->major_ver, orom->minor_ver,
2497 orom->hotfix_ver, orom->build);
e50cf220
MN		 2498 printf("IMSM_SUPPORTED_RAID_LEVELS=%s%s%s%s%s\n",
2499 imsm_orom_has_raid0(orom) ? "raid0 " : "",
2500 imsm_orom_has_raid1(orom) ? "raid1 " : "",
2501 imsm_orom_has_raid1e(orom) ? "raid1e " : "",
2502 imsm_orom_has_raid5(orom) ? "raid10 " : "",
2503 imsm_orom_has_raid10(orom) ? "raid5 " : "");
2504 printf("IMSM_SUPPORTED_CHUNK_SIZES=%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
2505 imsm_orom_has_chunk(orom, 2) ? "2k " : "",
2506 imsm_orom_has_chunk(orom, 4) ? "4k " : "",
2507 imsm_orom_has_chunk(orom, 8) ? "8k " : "",
2508 imsm_orom_has_chunk(orom, 16) ? "16k " : "",
2509 imsm_orom_has_chunk(orom, 32) ? "32k " : "",
2510 imsm_orom_has_chunk(orom, 64) ? "64k " : "",
2511 imsm_orom_has_chunk(orom, 128) ? "128k " : "",
2512 imsm_orom_has_chunk(orom, 256) ? "256k " : "",
2513 imsm_orom_has_chunk(orom, 512) ? "512k " : "",
2514 imsm_orom_has_chunk(orom, 1024*1) ? "1M " : "",
2515 imsm_orom_has_chunk(orom, 1024*2) ? "2M " : "",
2516 imsm_orom_has_chunk(orom, 1024*4) ? "4M " : "",
2517 imsm_orom_has_chunk(orom, 1024*8) ? "8M " : "",
2518 imsm_orom_has_chunk(orom, 1024*16) ? "16M " : "",
2519 imsm_orom_has_chunk(orom, 1024*32) ? "32M " : "",
2520 imsm_orom_has_chunk(orom, 1024*64) ? "64M " : "");
2521 printf("IMSM_2TB_VOLUMES=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB) ? "yes" : "no");
2522 printf("IMSM_2TB_DISKS=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB_DISK) ? "yes" : "no");
2523 printf("IMSM_MAX_DISKS=%d\n",orom->tds);
2524 printf("IMSM_MAX_VOLUMES_PER_ARRAY=%d\n",orom->vpa);
2525 printf("IMSM_MAX_VOLUMES_PER_CONTROLLER=%d\n",orom->vphba);
2526}
2527
9eafa1de 2528static int detail_platform_imsm(int verbose, int enumerate_only, char *controller_path)
d665cc31
DW		 2529{
2530 /* There are two components to imsm platform support, the ahci SATA
2531 * controller and the option-rom. To find the SATA controller we
2532 * simply look in /sys/bus/pci/drivers/ahci to see if an ahci
2533 * controller with the Intel vendor id is present. This approach
2534 * allows mdadm to leverage the kernel's ahci detection logic, with the
2535 * caveat that if ahci.ko is not loaded mdadm will not be able to
2536 * detect platform raid capabilities. The option-rom resides in a
2537 * platform "Adapter ROM". We scan for its signature to retrieve the
2538 * platform capabilities. If raid support is disabled in the BIOS the
2539 * option-rom capability structure will not be available.
2540 */
d665cc31 2541 struct sys_dev *list, *hba;
d665cc31
DW		 2542 int host_base = 0;
2543 int port_count = 0;
9eafa1de 2544 int result=1;
d665cc31 2545
5615172f 2546 if (enumerate_only) {
a891a3c2 2547 if (check_env("IMSM_NO_PLATFORM"))
5615172f 2548 return 0;
a891a3c2
LM		 2549 list = find_intel_devices();
2550 if (!list)
2551 return 2;
2552 for (hba = list; hba; hba = hba->next) {
6b781d33
AP		 2553 if (find_imsm_capability(hba)) {
2554 result = 0;
a891a3c2
LM		 2555 break;
2556 }
9eafa1de 2557 else
6b781d33 2558 result = 2;
a891a3c2 2559 }
a891a3c2 2560 return result;
5615172f
DW		 2561 }
2562
155cbb4c
LM		 2563 list = find_intel_devices();
2564 if (!list) {
ba728be7 2565 if (verbose > 0)
7a862a02 2566 pr_err("no active Intel(R) RAID controller found.\n");
d665cc31 2567 return 2;
ba728be7 2568 } else if (verbose > 0)
155cbb4c 2569 print_found_intel_controllers(list);
d665cc31 2570
a891a3c2 2571 for (hba = list; hba; hba = hba->next) {
0858eccf 2572 if (controller_path && (compare_paths(hba->path, controller_path) != 0))
9eafa1de 2573 continue;
0858eccf 2574 if (!find_imsm_capability(hba)) {
60f0f54d 2575 char buf[PATH_MAX];
e7b84f9d 2576 pr_err("imsm capabilities not found for controller: %s (type %s)\n",
60f0f54d
PB		 2577 hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path,
2578 get_sys_dev_type(hba->type));
0858eccf
AP		 2579 continue;
2580 }
2581 result = 0;
2582 }
2583
2584 if (controller_path && result == 1) {
2585 pr_err("no active Intel(R) RAID controller found under %s\n",
2586 controller_path);
2587 return result;
2588 }
2589
5e1d6128 2590 const struct orom_entry *entry;
0858eccf 2591
5e1d6128 2592 for (entry = orom_entries; entry; entry = entry->next) {
60f0f54d 2593 if (entry->type == SYS_DEV_VMD) {
07cb1e57 2594 print_imsm_capability(&entry->orom);
32716c51
PB		 2595 printf(" 3rd party NVMe :%s supported\n",
2596 imsm_orom_has_tpv_support(&entry->orom)?"":" not");
60f0f54d
PB		 2597 for (hba = list; hba; hba = hba->next) {
2598 if (hba->type == SYS_DEV_VMD) {
2599 char buf[PATH_MAX];
60f0f54d
PB		 2600 printf(" I/O Controller : %s (%s)\n",
2601 vmd_domain_to_controller(hba, buf), get_sys_dev_type(hba->type));
6da53c0e 2602 if (print_nvme_info(hba)) {
b5eece69
PB		 2603 if (verbose > 0)
2604 pr_err("failed to get devices attached to VMD domain.\n");
2605 result |= 2;
2606 }
60f0f54d
PB		 2607 }
2608 }
07cb1e57 2609 printf("\n");
60f0f54d
PB		 2610 continue;
2611 }
0858eccf 2612
60f0f54d
PB		 2613 print_imsm_capability(&entry->orom);
2614 if (entry->type == SYS_DEV_NVME) {
0858eccf
AP		 2615 for (hba = list; hba; hba = hba->next) {
2616 if (hba->type == SYS_DEV_NVME)
6da53c0e 2617 print_nvme_info(hba);
0858eccf 2618 }
60f0f54d 2619 printf("\n");
0858eccf
AP		 2620 continue;
2621 }
2622
2623 struct devid_list *devid;
5e1d6128 2624 for (devid = entry->devid_list; devid; devid = devid->next) {
0858eccf
AP		 2625 hba = device_by_id(devid->devid);
2626 if (!hba)
2627 continue;
2628
9eafa1de
MN		 2629 printf(" I/O Controller : %s (%s)\n",
2630 hba->path, get_sys_dev_type(hba->type));
2631 if (hba->type == SYS_DEV_SATA) {
2632 host_base = ahci_get_port_count(hba->path, &port_count);
2633 if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) {
2634 if (verbose > 0)
7a862a02 2635 pr_err("failed to enumerate ports on SATA controller at %s.\n", hba->pci_id);
9eafa1de
MN		 2636 result |= 2;
2637 }
120dc887
LM		 2638 }
2639 }
0858eccf 2640 printf("\n");
d665cc31 2641 }
155cbb4c 2642
120dc887 2643 return result;
d665cc31 2644}
e50cf220 2645
9eafa1de 2646static int export_detail_platform_imsm(int verbose, char *controller_path)
e50cf220 2647{
e50cf220
MN		 2648 struct sys_dev *list, *hba;
2649 int result=1;
2650
2651 list = find_intel_devices();
2652 if (!list) {
2653 if (verbose > 0)
2654 pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_INTEL_DEVICES\n");
2655 result = 2;
e50cf220
MN		 2656 return result;
2657 }
2658
2659 for (hba = list; hba; hba = hba->next) {
9eafa1de
MN		 2660 if (controller_path && (compare_paths(hba->path,controller_path) != 0))
2661 continue;
60f0f54d
PB		 2662 if (!find_imsm_capability(hba) && verbose > 0) {
2663 char buf[PATH_MAX];
2664 pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_IMSM_CAPABLE_DEVICE_UNDER_%s\n",
2665 hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path);
2666 }
0858eccf 2667 else
e50cf220 2668 result = 0;
e50cf220
MN		 2669 }
2670
5e1d6128 2671 const struct orom_entry *entry;
0858eccf 2672
60f0f54d
PB		 2673 for (entry = orom_entries; entry; entry = entry->next) {
2674 if (entry->type == SYS_DEV_VMD) {
2675 for (hba = list; hba; hba = hba->next)
2676 print_imsm_capability_export(&entry->orom);
2677 continue;
2678 }
5e1d6128 2679 print_imsm_capability_export(&entry->orom);
60f0f54d 2680 }
0858eccf 2681
e50cf220
MN		 2682 return result;
2683}
2684
cdddbdbc
DW		 2685static int match_home_imsm(struct supertype *st, char *homehost)
2686{
5115ca67
DW		 2687 /* the imsm metadata format does not specify any host
2688 * identification information. We return -1 since we can never
2689 * confirm nor deny whether a given array is "meant" for this
148acb7b 2690 * host. We rely on compare_super and the 'family_num' fields to
5115ca67
DW		 2691 * exclude member disks that do not belong, and we rely on
2692 * mdadm.conf to specify the arrays that should be assembled.
2693 * Auto-assembly may still pick up "foreign" arrays.
2694 */
cdddbdbc 2695
9362c1c8 2696 return -1;
cdddbdbc
DW		 2697}
2698
2699static void uuid_from_super_imsm(struct supertype *st, int uuid[4])
2700{
51006d85
N		 2701 /* The uuid returned here is used for:
2702 * uuid to put into bitmap file (Create, Grow)
2703 * uuid for backup header when saving critical section (Grow)
2704 * comparing uuids when re-adding a device into an array
2705 * In these cases the uuid required is that of the data-array,
2706 * not the device-set.
2707 * uuid to recognise same set when adding a missing device back
2708 * to an array. This is a uuid for the device-set.
1011e834 2709 *
51006d85
N		 2710 * For each of these we can make do with a truncated
2711 * or hashed uuid rather than the original, as long as
2712 * everyone agrees.
2713 * In each case the uuid required is that of the data-array,
2714 * not the device-set.
43dad3d6 2715 */
51006d85
N		 2716 /* imsm does not track uuid's so we synthesis one using sha1 on
2717 * - The signature (Which is constant for all imsm array, but no matter)
148acb7b 2718 * - the orig_family_num of the container
51006d85
N		 2719 * - the index number of the volume
2720 * - the 'serial' number of the volume.
2721 * Hopefully these are all constant.
2722 */
2723 struct intel_super *super = st->sb;
43dad3d6 2724
51006d85
N		 2725 char buf[20];
2726 struct sha1_ctx ctx;
2727 struct imsm_dev *dev = NULL;
148acb7b 2728 __u32 family_num;
51006d85 2729
148acb7b
DW		 2730 /* some mdadm versions failed to set ->orig_family_num, in which
2731 * case fall back to ->family_num. orig_family_num will be
2732 * fixed up with the first metadata update.
2733 */
2734 family_num = super->anchor->orig_family_num;
2735 if (family_num == 0)
2736 family_num = super->anchor->family_num;
51006d85 2737 sha1_init_ctx(&ctx);
92bd8f8d 2738 sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx);
148acb7b 2739 sha1_process_bytes(&family_num, sizeof(__u32), &ctx);
51006d85
N		 2740 if (super->current_vol >= 0)
2741 dev = get_imsm_dev(super, super->current_vol);
2742 if (dev) {
2743 __u32 vol = super->current_vol;
2744 sha1_process_bytes(&vol, sizeof(vol), &ctx);
2745 sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx);
2746 }
2747 sha1_finish_ctx(&ctx, buf);
2748 memcpy(uuid, buf, 4*4);
cdddbdbc
DW		 2749}
2750
0d481d37 2751#if 0
4f5bc454
DW		 2752static void
2753get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p)
cdddbdbc 2754{
cdddbdbc
DW		 2755 __u8 *v = get_imsm_version(mpb);
2756 __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH;
2757 char major[] = { 0, 0, 0 };
2758 char minor[] = { 0 ,0, 0 };
2759 char patch[] = { 0, 0, 0 };
2760 char *ver_parse[] = { major, minor, patch };
2761 int i, j;
2762
2763 i = j = 0;
2764 while (*v != '\0' && v < end) {
2765 if (*v != '.' && j < 2)
2766 ver_parse[i][j++] = *v;
2767 else {
2768 i++;
2769 j = 0;
2770 }
2771 v++;
2772 }
2773
4f5bc454
DW		 2774 *m = strtol(minor, NULL, 0);
2775 *p = strtol(patch, NULL, 0);
2776}
0d481d37 2777#endif
4f5bc454 2778
1e5c6983
DW		 2779static __u32 migr_strip_blocks_resync(struct imsm_dev *dev)
2780{
2781 /* migr_strip_size when repairing or initializing parity */
238c0a71 2782 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2783 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2784
2785 switch (get_imsm_raid_level(map)) {
2786 case 5:
2787 case 10:
2788 return chunk;
2789 default:
2790 return 128*1024 >> 9;
2791 }
2792}
2793
2794static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev)
2795{
2796 /* migr_strip_size when rebuilding a degraded disk, no idea why
2797 * this is different than migr_strip_size_resync(), but it's good
2798 * to be compatible
2799 */
238c0a71 2800 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2801 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2802
2803 switch (get_imsm_raid_level(map)) {
2804 case 1:
2805 case 10:
2806 if (map->num_members % map->num_domains == 0)
2807 return 128*1024 >> 9;
2808 else
2809 return chunk;
2810 case 5:
2811 return max((__u32) 64*1024 >> 9, chunk);
2812 default:
2813 return 128*1024 >> 9;
2814 }
2815}
2816
2817static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev)
2818{
238c0a71
AK		 2819 struct imsm_map *lo = get_imsm_map(dev, MAP_0);
2820 struct imsm_map *hi = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2821 __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip);
2822 __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip);
2823
2824 return max((__u32) 1, hi_chunk / lo_chunk);
2825}
2826
2827static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev)
2828{
238c0a71 2829 struct imsm_map *lo = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2830 int level = get_imsm_raid_level(lo);
2831
2832 if (level == 1 || level == 10) {
238c0a71 2833 struct imsm_map *hi = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2834
2835 return hi->num_domains;
2836 } else
2837 return num_stripes_per_unit_resync(dev);
2838}
2839
9529d343 2840static __u8 imsm_num_data_members(struct imsm_map *map)
1e5c6983
DW		 2841{
2842 /* named 'imsm_' because raid0, raid1 and raid10
2843 * counter-intuitively have the same number of data disks
2844 */
1e5c6983
DW		 2845 switch (get_imsm_raid_level(map)) {
2846 case 0:
36fd8ccc
AK		 2847 return map->num_members;
2848 break;
1e5c6983
DW		 2849 case 1:
2850 case 10:
36fd8ccc 2851 return map->num_members/2;
1e5c6983
DW		 2852 case 5:
2853 return map->num_members - 1;
2854 default:
1ade5cc1 2855 dprintf("unsupported raid level\n");
1e5c6983
DW		 2856 return 0;
2857 }
2858}
2859
44490938
MD		 2860static unsigned long long calc_component_size(struct imsm_map *map,
2861 struct imsm_dev *dev)
2862{
2863 unsigned long long component_size;
2864 unsigned long long dev_size = imsm_dev_size(dev);
a4f7290c 2865 long long calc_dev_size = 0;
44490938
MD		 2866 unsigned int member_disks = imsm_num_data_members(map);
2867
2868 if (member_disks == 0)
2869 return 0;
2870
2871 component_size = per_dev_array_size(map);
2872 calc_dev_size = component_size * member_disks;
2873
2874 /* Component size is rounded to 1MB so difference between size from
2875 * metadata and size calculated from num_data_stripes equals up to
2876 * 2048 blocks per each device. If the difference is higher it means
2877 * that array size was expanded and num_data_stripes was not updated.
2878 */
a4f7290c 2879 if (llabs(calc_dev_size - (long long)dev_size) >
44490938
MD		 2880 (1 << SECT_PER_MB_SHIFT) * member_disks) {
2881 component_size = dev_size / member_disks;
2882 dprintf("Invalid num_data_stripes in metadata; expected=%llu, found=%llu\n",
2883 component_size / map->blocks_per_strip,
2884 num_data_stripes(map));
2885 }
2886
2887 return component_size;
2888}
2889
1e5c6983
DW		 2890static __u32 parity_segment_depth(struct imsm_dev *dev)
2891{
238c0a71 2892 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2893 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2894
2895 switch(get_imsm_raid_level(map)) {
2896 case 1:
2897 case 10:
2898 return chunk * map->num_domains;
2899 case 5:
2900 return chunk * map->num_members;
2901 default:
2902 return chunk;
2903 }
2904}
2905
2906static __u32 map_migr_block(struct imsm_dev *dev, __u32 block)
2907{
238c0a71 2908 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2909 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2910 __u32 strip = block / chunk;
2911
2912 switch (get_imsm_raid_level(map)) {
2913 case 1:
2914 case 10: {
2915 __u32 vol_strip = (strip * map->num_domains) + 1;
2916 __u32 vol_stripe = vol_strip / map->num_members;
2917
2918 return vol_stripe * chunk + block % chunk;
2919 } case 5: {
2920 __u32 stripe = strip / (map->num_members - 1);
2921
2922 return stripe * chunk + block % chunk;
2923 }
2924 default:
2925 return 0;
2926 }
2927}
2928
c47b0ff6
AK		 2929static __u64 blocks_per_migr_unit(struct intel_super *super,
2930 struct imsm_dev *dev)
1e5c6983
DW		 2931{
2932 /* calculate the conversion factor between per member 'blocks'
2933 * (md/{resync,rebuild}_start) and imsm migration units, return
2934 * 0 for the 'not migrating' and 'unsupported migration' cases
2935 */
2936 if (!dev->vol.migr_state)
2937 return 0;
2938
2939 switch (migr_type(dev)) {
c47b0ff6
AK		 2940 case MIGR_GEN_MIGR: {
2941 struct migr_record *migr_rec = super->migr_rec;
2942 return __le32_to_cpu(migr_rec->blocks_per_unit);
2943 }
1e5c6983
DW		 2944 case MIGR_VERIFY:
2945 case MIGR_REPAIR:
2946 case MIGR_INIT: {
238c0a71 2947 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2948 __u32 stripes_per_unit;
2949 __u32 blocks_per_unit;
2950 __u32 parity_depth;
2951 __u32 migr_chunk;
2952 __u32 block_map;
2953 __u32 block_rel;
2954 __u32 segment;
2955 __u32 stripe;
2956 __u8 disks;
2957
2958 /* yes, this is really the translation of migr_units to
2959 * per-member blocks in the 'resync' case
2960 */
2961 stripes_per_unit = num_stripes_per_unit_resync(dev);
2962 migr_chunk = migr_strip_blocks_resync(dev);
9529d343 2963 disks = imsm_num_data_members(map);
1e5c6983 2964 blocks_per_unit = stripes_per_unit * migr_chunk * disks;
7b1ab482 2965 stripe = __le16_to_cpu(map->blocks_per_strip) * disks;
1e5c6983
DW		 2966 segment = blocks_per_unit / stripe;
2967 block_rel = blocks_per_unit - segment * stripe;
2968 parity_depth = parity_segment_depth(dev);
2969 block_map = map_migr_block(dev, block_rel);
2970 return block_map + parity_depth * segment;
2971 }
2972 case MIGR_REBUILD: {
2973 __u32 stripes_per_unit;
2974 __u32 migr_chunk;
2975
2976 stripes_per_unit = num_stripes_per_unit_rebuild(dev);
2977 migr_chunk = migr_strip_blocks_rebuild(dev);
2978 return migr_chunk * stripes_per_unit;
2979 }
1e5c6983
DW		 2980 case MIGR_STATE_CHANGE:
2981 default:
2982 return 0;
2983 }
2984}
2985
c2c087e6
DW		 2986static int imsm_level_to_layout(int level)
2987{
2988 switch (level) {
2989 case 0:
2990 case 1:
2991 return 0;
2992 case 5:
2993 case 6:
a380c027 2994 return ALGORITHM_LEFT_ASYMMETRIC;
c2c087e6 2995 case 10:
c92a2527 2996 return 0x102;
c2c087e6 2997 }
a18a888e 2998 return UnSet;
c2c087e6
DW		 2999}
3000
8e59f3d8
AK		 3001/*******************************************************************************
3002 * Function: read_imsm_migr_rec
3003 * Description: Function reads imsm migration record from last sector of disk
3004 * Parameters:
3005 * fd : disk descriptor
3006 * super : metadata info
3007 * Returns:
3008 * 0 : success,
3009 * -1 : fail
3010 ******************************************************************************/
3011static int read_imsm_migr_rec(int fd, struct intel_super *super)
3012{
3013 int ret_val = -1;
de44e46f 3014 unsigned int sector_size = super->sector_size;
8e59f3d8
AK		 3015 unsigned long long dsize;
3016
3017 get_dev_size(fd, NULL, &dsize);
de44e46f
PB		 3018 if (lseek64(fd, dsize - (sector_size*MIGR_REC_SECTOR_POSITION),
3019 SEEK_SET) < 0) {
e7b84f9d
N		 3020 pr_err("Cannot seek to anchor block: %s\n",
3021 strerror(errno));
8e59f3d8
AK		 3022 goto out;
3023 }
466070ad 3024 if ((unsigned int)read(fd, super->migr_rec_buf,
de44e46f
PB		 3025 MIGR_REC_BUF_SECTORS*sector_size) !=
3026 MIGR_REC_BUF_SECTORS*sector_size) {
e7b84f9d
N		 3027 pr_err("Cannot read migr record block: %s\n",
3028 strerror(errno));
8e59f3d8
AK		 3029 goto out;
3030 }
3031 ret_val = 0;
de44e46f
PB		 3032 if (sector_size == 4096)
3033 convert_from_4k_imsm_migr_rec(super);
8e59f3d8
AK		 3034
3035out:
3036 return ret_val;
3037}
3038
3136abe5
AK		 3039static struct imsm_dev *imsm_get_device_during_migration(
3040 struct intel_super *super)
3041{
3042
3043 struct intel_dev *dv;
3044
3045 for (dv = super->devlist; dv; dv = dv->next) {
3046 if (is_gen_migration(dv->dev))
3047 return dv->dev;
3048 }
3049 return NULL;
3050}
3051
8e59f3d8
AK		 3052/*******************************************************************************
3053 * Function: load_imsm_migr_rec
3054 * Description: Function reads imsm migration record (it is stored at the last
3055 * sector of disk)
3056 * Parameters:
3057 * super : imsm internal array info
8e59f3d8
AK		 3058 * Returns:
3059 * 0 : success
3060 * -1 : fail
4c965cc9 3061 * -2 : no migration in progress
8e59f3d8 3062 ******************************************************************************/
2f86fda3 3063static int load_imsm_migr_rec(struct intel_super *super)
8e59f3d8 3064{
594dc1b8 3065 struct dl *dl;
8e59f3d8
AK		 3066 char nm[30];
3067 int retval = -1;
3068 int fd = -1;
3136abe5 3069 struct imsm_dev *dev;
594dc1b8 3070 struct imsm_map *map;
b4ab44d8 3071 int slot = -1;
2f86fda3 3072 int keep_fd = 1;
3136abe5
AK		 3073
3074 /* find map under migration */
3075 dev = imsm_get_device_during_migration(super);
3076 /* nothing to load,no migration in progress?
3077 */
3078 if (dev == NULL)
4c965cc9 3079 return -2;
8e59f3d8 3080
2f86fda3
MT		 3081 map = get_imsm_map(dev, MAP_0);
3082 if (!map)
3083 return -1;
3136abe5 3084
2f86fda3
MT		 3085 for (dl = super->disks; dl; dl = dl->next) {
3086 /* skip spare and failed disks
3087 */
3088 if (dl->index < 0)
3089 continue;
3090 /* read only from one of the first two slots
3091 */
3092 slot = get_imsm_disk_slot(map, dl->index);
3093 if (slot > 1 || slot < 0)
3094 continue;
3095
3096 if (dl->fd < 0) {
8e59f3d8
AK		 3097 sprintf(nm, "%d:%d", dl->major, dl->minor);
3098 fd = dev_open(nm, O_RDONLY);
2f86fda3
MT		 3099 if (fd >= 0) {
3100 keep_fd = 0;
8e59f3d8 3101 break;
2f86fda3
MT		 3102 }
3103 } else {
3104 fd = dl->fd;
3105 break;
8e59f3d8
AK		 3106 }
3107 }
2f86fda3 3108
8e59f3d8 3109 if (fd < 0)
2f86fda3 3110 return retval;
8e59f3d8 3111 retval = read_imsm_migr_rec(fd, super);
2f86fda3 3112 if (!keep_fd)
8e59f3d8 3113 close(fd);
2f86fda3 3114
8e59f3d8
AK		 3115 return retval;
3116}
3117
c17608ea
AK		 3118/*******************************************************************************
3119 * function: imsm_create_metadata_checkpoint_update
3120 * Description: It creates update for checkpoint change.
3121 * Parameters:
3122 * super : imsm internal array info
3123 * u : pointer to prepared update
3124 * Returns:
3125 * Uptate length.
3126 * If length is equal to 0, input pointer u contains no update
3127 ******************************************************************************/
3128static int imsm_create_metadata_checkpoint_update(
3129 struct intel_super *super,
3130 struct imsm_update_general_migration_checkpoint **u)
3131{
3132
3133 int update_memory_size = 0;
3134
1ade5cc1 3135 dprintf("(enter)\n");
c17608ea
AK		 3136
3137 if (u == NULL)
3138 return 0;
3139 *u = NULL;
3140
3141 /* size of all update data without anchor */
3142 update_memory_size =
3143 sizeof(struct imsm_update_general_migration_checkpoint);
3144
503975b9 3145 *u = xcalloc(1, update_memory_size);
c17608ea 3146 if (*u == NULL) {
1ade5cc1 3147 dprintf("error: cannot get memory\n");
c17608ea
AK		 3148 return 0;
3149 }
3150 (*u)->type = update_general_migration_checkpoint;
9f421827 3151 (*u)->curr_migr_unit = current_migr_unit(super->migr_rec);
1ade5cc1 3152 dprintf("prepared for %u\n", (*u)->curr_migr_unit);
c17608ea
AK		 3153
3154 return update_memory_size;
3155}
3156
c17608ea
AK		 3157static void imsm_update_metadata_locally(struct supertype *st,
3158 void *buf, int len);
3159
687629c2
AK		 3160/*******************************************************************************
3161 * Function: write_imsm_migr_rec
3162 * Description: Function writes imsm migration record
3163 * (at the last sector of disk)
3164 * Parameters:
3165 * super : imsm internal array info
3166 * Returns:
3167 * 0 : success
3168 * -1 : if fail
3169 ******************************************************************************/
3170static int write_imsm_migr_rec(struct supertype *st)
3171{
3172 struct intel_super *super = st->sb;
de44e46f 3173 unsigned int sector_size = super->sector_size;
687629c2 3174 unsigned long long dsize;
687629c2
AK		 3175 int retval = -1;
3176 struct dl *sd;
c17608ea
AK		 3177 int len;
3178 struct imsm_update_general_migration_checkpoint *u;
3136abe5 3179 struct imsm_dev *dev;
594dc1b8 3180 struct imsm_map *map;
3136abe5
AK		 3181
3182 /* find map under migration */
3183 dev = imsm_get_device_during_migration(super);
3184 /* if no migration, write buffer anyway to clear migr_record
3185 * on disk based on first available device
3186 */
3187 if (dev == NULL)
3188 dev = get_imsm_dev(super, super->current_vol < 0 ? 0 :
3189 super->current_vol);
3190
44bfe6df 3191 map = get_imsm_map(dev, MAP_0);
687629c2 3192
de44e46f
PB		 3193 if (sector_size == 4096)
3194 convert_to_4k_imsm_migr_rec(super);
687629c2 3195 for (sd = super->disks ; sd ; sd = sd->next) {
b4ab44d8 3196 int slot = -1;
3136abe5
AK		 3197
3198 /* skip failed and spare devices */
3199 if (sd->index < 0)
3200 continue;
687629c2 3201 /* write to 2 first slots only */
3136abe5
AK		 3202 if (map)
3203 slot = get_imsm_disk_slot(map, sd->index);
089f9d79 3204 if (map == NULL || slot > 1 || slot < 0)
687629c2 3205 continue;
3136abe5 3206
2f86fda3
MT		 3207 get_dev_size(sd->fd, NULL, &dsize);
3208 if (lseek64(sd->fd, dsize - (MIGR_REC_SECTOR_POSITION *
3209 sector_size),
de44e46f 3210 SEEK_SET) < 0) {
e7b84f9d
N		 3211 pr_err("Cannot seek to anchor block: %s\n",
3212 strerror(errno));
687629c2
AK		 3213 goto out;
3214 }
2f86fda3 3215 if ((unsigned int)write(sd->fd, super->migr_rec_buf,
de44e46f
PB		 3216 MIGR_REC_BUF_SECTORS*sector_size) !=
3217 MIGR_REC_BUF_SECTORS*sector_size) {
e7b84f9d
N		 3218 pr_err("Cannot write migr record block: %s\n",
3219 strerror(errno));
687629c2
AK		 3220 goto out;
3221 }
687629c2 3222 }
de44e46f
PB		 3223 if (sector_size == 4096)
3224 convert_from_4k_imsm_migr_rec(super);
c17608ea
AK		 3225 /* update checkpoint information in metadata */
3226 len = imsm_create_metadata_checkpoint_update(super, &u);
c17608ea
AK		 3227 if (len <= 0) {
3228 dprintf("imsm: Cannot prepare update\n");
3229 goto out;
3230 }
3231 /* update metadata locally */
3232 imsm_update_metadata_locally(st, u, len);
3233 /* and possibly remotely */
3234 if (st->update_tail) {
3235 append_metadata_update(st, u, len);
3236 /* during reshape we do all work inside metadata handler
3237 * manage_reshape(), so metadata update has to be triggered
3238 * insida it
3239 */
3240 flush_metadata_updates(st);
3241 st->update_tail = &st->updates;
3242 } else
3243 free(u);
687629c2
AK		 3244
3245 retval = 0;
3246 out:
687629c2
AK		 3247 return retval;
3248}
3249
e2962bfc
AK		 3250/* spare/missing disks activations are not allowe when
3251 * array/container performs reshape operation, because
3252 * all arrays in container works on the same disks set
3253 */
3254int imsm_reshape_blocks_arrays_changes(struct intel_super *super)
3255{
3256 int rv = 0;
3257 struct intel_dev *i_dev;
3258 struct imsm_dev *dev;
3259
3260 /* check whole container
3261 */
3262 for (i_dev = super->devlist; i_dev; i_dev = i_dev->next) {
3263 dev = i_dev->dev;
3ad25638 3264 if (is_gen_migration(dev)) {
e2962bfc
AK		 3265 /* No repair during any migration in container
3266 */
3267 rv = 1;
3268 break;
3269 }
3270 }
3271 return rv;
3272}
3e684231 3273static unsigned long long imsm_component_size_alignment_check(int level,
c41e00b2 3274 int chunk_size,
f36a9ecd 3275 unsigned int sector_size,
c41e00b2
AK		 3276 unsigned long long component_size)
3277{
3e684231 3278 unsigned int component_size_alignment;
c41e00b2 3279
3e684231 3280 /* check component size alignment
c41e00b2 3281 */
3e684231 3282 component_size_alignment = component_size % (chunk_size/sector_size);
c41e00b2 3283
3e684231 3284 dprintf("(Level: %i, chunk_size = %i, component_size = %llu), component_size_alignment = %u\n",
c41e00b2 3285 level, chunk_size, component_size,
3e684231 3286 component_size_alignment);
c41e00b2 3287
3e684231
MZ		 3288 if (component_size_alignment && (level != 1) && (level != UnSet)) {
3289 dprintf("imsm: reported component size aligned from %llu ",
c41e00b2 3290 component_size);
3e684231 3291 component_size -= component_size_alignment;
1ade5cc1 3292 dprintf_cont("to %llu (%i).\n",
3e684231 3293 component_size, component_size_alignment);
c41e00b2
AK		 3294 }
3295
3296 return component_size;
3297}
e2962bfc 3298
2432ce9b
AP		 3299static unsigned long long get_ppl_sector(struct intel_super *super, int dev_idx)
3300{
3301 struct imsm_dev *dev = get_imsm_dev(super, dev_idx);
3302 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3303
3304 return pba_of_lba0(map) +
3305 (num_data_stripes(map) * map->blocks_per_strip);
3306}
3307
a5d85af7 3308static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap)
bf5a934a
DW		 3309{
3310 struct intel_super *super = st->sb;
c47b0ff6 3311 struct migr_record *migr_rec = super->migr_rec;
949c47a0 3312 struct imsm_dev *dev = get_imsm_dev(super, super->current_vol);
238c0a71
AK		 3313 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3314 struct imsm_map *prev_map = get_imsm_map(dev, MAP_1);
b335e593 3315 struct imsm_map *map_to_analyse = map;
efb30e7f 3316 struct dl *dl;
a5d85af7 3317 int map_disks = info->array.raid_disks;
bf5a934a 3318
95eeceeb 3319 memset(info, 0, sizeof(*info));
b335e593
AK		 3320 if (prev_map)
3321 map_to_analyse = prev_map;
3322
ca0748fa 3323 dl = super->current_disk;
9894ec0d 3324
bf5a934a 3325 info->container_member = super->current_vol;
cd0430a1 3326 info->array.raid_disks = map->num_members;
b335e593 3327 info->array.level = get_imsm_raid_level(map_to_analyse);
bf5a934a
DW		 3328 info->array.layout = imsm_level_to_layout(info->array.level);
3329 info->array.md_minor = -1;
3330 info->array.ctime = 0;
3331 info->array.utime = 0;
b335e593
AK		 3332 info->array.chunk_size =
3333 __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9;
2432ce9b 3334 info->array.state = !(dev->vol.dirty & RAIDVOL_DIRTY);
fcc2c9da 3335 info->custom_array_size = imsm_dev_size(dev);
3ad25638
AK		 3336 info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb);
3337
3f510843 3338 if (is_gen_migration(dev)) {
3f83228a 3339 info->reshape_active = 1;
b335e593
AK		 3340 info->new_level = get_imsm_raid_level(map);
3341 info->new_layout = imsm_level_to_layout(info->new_level);
3342 info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9;
3f83228a 3343 info->delta_disks = map->num_members - prev_map->num_members;
493f5dd6
N		 3344 if (info->delta_disks) {
3345 /* this needs to be applied to every array
3346 * in the container.
3347 */
81219e70 3348 info->reshape_active = CONTAINER_RESHAPE;
493f5dd6 3349 }
3f83228a
N		 3350 /* We shape information that we give to md might have to be
3351 * modify to cope with md's requirement for reshaping arrays.
3352 * For example, when reshaping a RAID0, md requires it to be
3353 * presented as a degraded RAID4.
3354 * Also if a RAID0 is migrating to a RAID5 we need to specify
3355 * the array as already being RAID5, but the 'before' layout
3356 * is a RAID4-like layout.
3357 */
3358 switch (info->array.level) {
3359 case 0:
3360 switch(info->new_level) {
3361 case 0:
3362 /* conversion is happening as RAID4 */
3363 info->array.level = 4;
3364 info->array.raid_disks += 1;
3365 break;
3366 case 5:
3367 /* conversion is happening as RAID5 */
3368 info->array.level = 5;
3369 info->array.layout = ALGORITHM_PARITY_N;
3f83228a
N		 3370 info->delta_disks -= 1;
3371 break;
3372 default:
3373 /* FIXME error message */
3374 info->array.level = UnSet;
3375 break;
3376 }
3377 break;
3378 }
b335e593
AK		 3379 } else {
3380 info->new_level = UnSet;
3381 info->new_layout = UnSet;
3382 info->new_chunk = info->array.chunk_size;
3f83228a 3383 info->delta_disks = 0;
b335e593 3384 }
ca0748fa 3385
efb30e7f
DW		 3386 if (dl) {
3387 info->disk.major = dl->major;
3388 info->disk.minor = dl->minor;
ca0748fa 3389 info->disk.number = dl->index;
656b6b5a
N		 3390 info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse,
3391 dl->index);
efb30e7f 3392 }
bf5a934a 3393
5551b113 3394 info->data_offset = pba_of_lba0(map_to_analyse);
44490938 3395 info->component_size = calc_component_size(map, dev);
3e684231 3396 info->component_size = imsm_component_size_alignment_check(
c41e00b2
AK		 3397 info->array.level,
3398 info->array.chunk_size,
f36a9ecd 3399 super->sector_size,
c41e00b2 3400 info->component_size);
5e46202e 3401 info->bb.supported = 1;
139dae11 3402
301406c9 3403 memset(info->uuid, 0, sizeof(info->uuid));
921d9e16 3404 info->recovery_start = MaxSector;
bf5a934a 3405
c2462068
PB		 3406 if (info->array.level == 5 &&
3407 (dev->rwh_policy == RWH_DISTRIBUTED ||
3408 dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)) {
2432ce9b
AP		 3409 info->consistency_policy = CONSISTENCY_POLICY_PPL;
3410 info->ppl_sector = get_ppl_sector(super, super->current_vol);
c2462068
PB		 3411 if (dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
3412 info->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9;
3413 else
3414 info->ppl_size = (PPL_HEADER_SIZE + PPL_ENTRY_SPACE)
3415 >> 9;
2432ce9b
AP		 3416 } else if (info->array.level <= 0) {
3417 info->consistency_policy = CONSISTENCY_POLICY_NONE;
3418 } else {
3419 info->consistency_policy = CONSISTENCY_POLICY_RESYNC;
3420 }
3421
d2e6d5d6 3422 info->reshape_progress = 0;
b6796ce1 3423 info->resync_start = MaxSector;
b9172665 3424 if ((map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED ||
2432ce9b 3425 !(info->array.state & 1)) &&
b9172665 3426 imsm_reshape_blocks_arrays_changes(super) == 0) {
301406c9 3427 info->resync_start = 0;
b6796ce1
AK		 3428 }
3429 if (dev->vol.migr_state) {
1e5c6983
DW		 3430 switch (migr_type(dev)) {
3431 case MIGR_REPAIR:
3432 case MIGR_INIT: {
c47b0ff6
AK		 3433 __u64 blocks_per_unit = blocks_per_migr_unit(super,
3434 dev);
1e5c6983
DW		 3435 __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit);
3436
3437 info->resync_start = blocks_per_unit * units;
3438 break;
3439 }
d2e6d5d6 3440 case MIGR_GEN_MIGR: {
c47b0ff6
AK		 3441 __u64 blocks_per_unit = blocks_per_migr_unit(super,
3442 dev);
9f421827 3443 __u64 units = current_migr_unit(migr_rec);
04fa9523 3444 int used_disks;
d2e6d5d6 3445
befb629b
AK		 3446 if (__le32_to_cpu(migr_rec->ascending_migr) &&
3447 (units <
9f421827 3448 (get_num_migr_units(migr_rec)-1)) &&
befb629b
AK		 3449 (super->migr_rec->rec_status ==
3450 __cpu_to_le32(UNIT_SRC_IN_CP_AREA)))
3451 units++;
3452
d2e6d5d6 3453 info->reshape_progress = blocks_per_unit * units;
6289d1e0 3454
7a862a02 3455 dprintf("IMSM: General Migration checkpoint : %llu (%llu) -> read reshape progress : %llu\n",
19986c72
MB		 3456 (unsigned long long)units,
3457 (unsigned long long)blocks_per_unit,
3458 info->reshape_progress);
75156c46 3459
9529d343 3460 used_disks = imsm_num_data_members(prev_map);
75156c46 3461 if (used_disks > 0) {
895ffd99 3462 info->custom_array_size = per_dev_array_size(map) *
75156c46 3463 used_disks;
75156c46 3464 }
d2e6d5d6 3465 }
1e5c6983
DW		 3466 case MIGR_VERIFY:
3467 /* we could emulate the checkpointing of
3468 * 'sync_action=check' migrations, but for now
3469 * we just immediately complete them
3470 */
3471 case MIGR_REBUILD:
3472 /* this is handled by container_content_imsm() */
1e5c6983
DW		 3473 case MIGR_STATE_CHANGE:
3474 /* FIXME handle other migrations */
3475 default:
3476 /* we are not dirty, so... */
3477 info->resync_start = MaxSector;
3478 }
b6796ce1 3479 }
301406c9
DW		 3480
3481 strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN);
3482 info->name[MAX_RAID_SERIAL_LEN] = 0;
bf5a934a 3483
f35f2525
N		 3484 info->array.major_version = -1;
3485 info->array.minor_version = -2;
4dd2df09 3486 sprintf(info->text_version, "/%s/%d", st->container_devnm, info->container_member);
a67dd8cc 3487 info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */
51006d85 3488 uuid_from_super_imsm(st, info->uuid);
a5d85af7
N		 3489
3490 if (dmap) {
3491 int i, j;
3492 for (i=0; i<map_disks; i++) {
3493 dmap[i] = 0;
3494 if (i < info->array.raid_disks) {
3495 struct imsm_disk *dsk;
238c0a71 3496 j = get_imsm_disk_idx(dev, i, MAP_X);
a5d85af7
N		 3497 dsk = get_imsm_disk(super, j);
3498 if (dsk && (dsk->status & CONFIGURED_DISK))
3499 dmap[i] = 1;
3500 }
3501 }
3502 }
81ac8b4d 3503}
bf5a934a 3504
3b451610
AK		 3505static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev,
3506 int failed, int look_in_map);
3507
3508static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev,
3509 int look_in_map);
3510
3511static void manage_second_map(struct intel_super *super, struct imsm_dev *dev)
3512{
3513 if (is_gen_migration(dev)) {
3514 int failed;
3515 __u8 map_state;
3516 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
3517
3518 failed = imsm_count_failed(super, dev, MAP_1);
238c0a71 3519 map_state = imsm_check_degraded(super, dev, failed, MAP_1);
3b451610
AK		 3520 if (map2->map_state != map_state) {
3521 map2->map_state = map_state;
3522 super->updates_pending++;
3523 }
3524 }
3525}
97b4d0e9
DW		 3526
3527static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index)
3528{
3529 struct dl *d;
3530
3531 for (d = super->missing; d; d = d->next)
3532 if (d->index == index)
3533 return &d->disk;
3534 return NULL;
3535}
3536
a5d85af7 3537static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map)
4f5bc454
DW		 3538{
3539 struct intel_super *super = st->sb;
4f5bc454 3540 struct imsm_disk *disk;
a5d85af7 3541 int map_disks = info->array.raid_disks;
ab3cb6b3
N		 3542 int max_enough = -1;
3543 int i;
3544 struct imsm_super *mpb;
4f5bc454 3545
bf5a934a 3546 if (super->current_vol >= 0) {
a5d85af7 3547 getinfo_super_imsm_volume(st, info, map);
bf5a934a
DW		 3548 return;
3549 }
95eeceeb 3550 memset(info, 0, sizeof(*info));
d23fe947
DW		 3551
3552 /* Set raid_disks to zero so that Assemble will always pull in valid
3553 * spares
3554 */
3555 info->array.raid_disks = 0;
cdddbdbc
DW		 3556 info->array.level = LEVEL_CONTAINER;
3557 info->array.layout = 0;
3558 info->array.md_minor = -1;
1011e834 3559 info->array.ctime = 0; /* N/A for imsm */
cdddbdbc
DW		 3560 info->array.utime = 0;
3561 info->array.chunk_size = 0;
3562
3563 info->disk.major = 0;
3564 info->disk.minor = 0;
cdddbdbc 3565 info->disk.raid_disk = -1;
c2c087e6 3566 info->reshape_active = 0;
f35f2525
N		 3567 info->array.major_version = -1;
3568 info->array.minor_version = -2;
c2c087e6 3569 strcpy(info->text_version, "imsm");
a67dd8cc 3570 info->safe_mode_delay = 0;
c2c087e6
DW		 3571 info->disk.number = -1;
3572 info->disk.state = 0;
c5afc314 3573 info->name[0] = 0;
921d9e16 3574 info->recovery_start = MaxSector;
3ad25638 3575 info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb);
5e46202e 3576 info->bb.supported = 1;
c2c087e6 3577
97b4d0e9 3578 /* do we have the all the insync disks that we expect? */
ab3cb6b3 3579 mpb = super->anchor;
b7d81a38 3580 info->events = __le32_to_cpu(mpb->generation_num);
97b4d0e9 3581
ab3cb6b3
N		 3582 for (i = 0; i < mpb->num_raid_devs; i++) {
3583 struct imsm_dev *dev = get_imsm_dev(super, i);
3584 int failed, enough, j, missing = 0;
3585 struct imsm_map *map;
3586 __u8 state;
97b4d0e9 3587
3b451610
AK		 3588 failed = imsm_count_failed(super, dev, MAP_0);
3589 state = imsm_check_degraded(super, dev, failed, MAP_0);
238c0a71 3590 map = get_imsm_map(dev, MAP_0);
ab3cb6b3
N		 3591
3592 /* any newly missing disks?
3593 * (catches single-degraded vs double-degraded)
3594 */
3595 for (j = 0; j < map->num_members; j++) {
238c0a71 3596 __u32 ord = get_imsm_ord_tbl_ent(dev, j, MAP_0);
ab3cb6b3
N		 3597 __u32 idx = ord_to_idx(ord);
3598
20dc76d1
MT		 3599 if (super->disks && super->disks->index == (int)idx)
3600 info->disk.raid_disk = j;
3601
ab3cb6b3
N		 3602 if (!(ord & IMSM_ORD_REBUILD) &&
3603 get_imsm_missing(super, idx)) {
3604 missing = 1;
3605 break;
3606 }
97b4d0e9 3607 }
ab3cb6b3
N		 3608
3609 if (state == IMSM_T_STATE_FAILED)
3610 enough = -1;
3611 else if (state == IMSM_T_STATE_DEGRADED &&
3612 (state != map->map_state || missing))
3613 enough = 0;
3614 else /* we're normal, or already degraded */
3615 enough = 1;
d2bde6d3
AK		 3616 if (is_gen_migration(dev) && missing) {
3617 /* during general migration we need all disks
3618 * that process is running on.
3619 * No new missing disk is allowed.
3620 */
3621 max_enough = -1;
3622 enough = -1;
3623 /* no more checks necessary
3624 */
3625 break;
3626 }
ab3cb6b3
N		 3627 /* in the missing/failed disk case check to see
3628 * if at least one array is runnable
3629 */
3630 max_enough = max(max_enough, enough);
3631 }
1ade5cc1 3632 dprintf("enough: %d\n", max_enough);
ab3cb6b3 3633 info->container_enough = max_enough;
97b4d0e9 3634
4a04ec6c 3635 if (super->disks) {
14e8215b
DW		 3636 __u32 reserved = imsm_reserved_sectors(super, super->disks);
3637
b9f594fe 3638 disk = &super->disks->disk;
5551b113 3639 info->data_offset = total_blocks(&super->disks->disk) - reserved;
14e8215b 3640 info->component_size = reserved;
25ed7e59 3641 info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0;
df474657
DW		 3642 /* we don't change info->disk.raid_disk here because
3643 * this state will be finalized in mdmon after we have
3644 * found the 'most fresh' version of the metadata
3645 */
25ed7e59 3646 info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0;
2432ce9b
AP		 3647 info->disk.state |= (is_spare(disk) || is_journal(disk)) ?
3648 0 : (1 << MD_DISK_SYNC);
cdddbdbc 3649 }
a575e2a7
DW		 3650
3651 /* only call uuid_from_super_imsm when this disk is part of a populated container,
3652 * ->compare_super may have updated the 'num_raid_devs' field for spares
3653 */
3654 if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs)
36ba7d48 3655 uuid_from_super_imsm(st, info->uuid);
22e263f6
AC		 3656 else
3657 memcpy(info->uuid, uuid_zero, sizeof(uuid_zero));
a5d85af7
N		 3658
3659 /* I don't know how to compute 'map' on imsm, so use safe default */
3660 if (map) {
3661 int i;
3662 for (i = 0; i < map_disks; i++)
3663 map[i] = 1;
3664 }
3665
cdddbdbc
DW		 3666}
3667
5c4cd5da
AC		 3668/* allocates memory and fills disk in mdinfo structure
3669 * for each disk in array */
3670struct mdinfo *getinfo_super_disks_imsm(struct supertype *st)
3671{
594dc1b8 3672 struct mdinfo *mddev;
5c4cd5da
AC		 3673 struct intel_super *super = st->sb;
3674 struct imsm_disk *disk;
3675 int count = 0;
3676 struct dl *dl;
3677 if (!super || !super->disks)
3678 return NULL;
3679 dl = super->disks;
503975b9 3680 mddev = xcalloc(1, sizeof(*mddev));
5c4cd5da
AC		 3681 while (dl) {
3682 struct mdinfo *tmp;
3683 disk = &dl->disk;
503975b9 3684 tmp = xcalloc(1, sizeof(*tmp));
5c4cd5da
AC		 3685 if (mddev->devs)
3686 tmp->next = mddev->devs;
3687 mddev->devs = tmp;
3688 tmp->disk.number = count++;
3689 tmp->disk.major = dl->major;
3690 tmp->disk.minor = dl->minor;
3691 tmp->disk.state = is_configured(disk) ?
3692 (1 << MD_DISK_ACTIVE) : 0;
3693 tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0;
3694 tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC);
3695 tmp->disk.raid_disk = -1;
3696 dl = dl->next;
3697 }
3698 return mddev;
3699}
3700
cdddbdbc
DW		 3701static int update_super_imsm(struct supertype *st, struct mdinfo *info,
3702 char *update, char *devname, int verbose,
3703 int uuid_set, char *homehost)
3704{
f352c545
DW		 3705 /* For 'assemble' and 'force' we need to return non-zero if any
3706 * change was made. For others, the return value is ignored.
3707 * Update options are:
3708 * force-one : This device looks a bit old but needs to be included,
3709 * update age info appropriately.
3710 * assemble: clear any 'faulty' flag to allow this device to
3711 * be assembled.
3712 * force-array: Array is degraded but being forced, mark it clean
3713 * if that will be needed to assemble it.
3714 *
3715 * newdev: not used ????
3716 * grow: Array has gained a new device - this is currently for
3717 * linear only
3718 * resync: mark as dirty so a resync will happen.
3719 * name: update the name - preserving the homehost
6e46bf34 3720 * uuid: Change the uuid of the array to match watch is given
f352c545
DW		 3721 *
3722 * Following are not relevant for this imsm:
3723 * sparc2.2 : update from old dodgey metadata
3724 * super-minor: change the preferred_minor number
3725 * summaries: update redundant counters.
f352c545
DW		 3726 * homehost: update the recorded homehost
3727 * _reshape_progress: record new reshape_progress position.
3728 */
6e46bf34
DW		 3729 int rv = 1;
3730 struct intel_super *super = st->sb;
3731 struct imsm_super *mpb;
f352c545 3732
6e46bf34
DW		 3733 /* we can only update container info */
3734 if (!super || super->current_vol >= 0 || !super->anchor)
3735 return 1;
3736
3737 mpb = super->anchor;
3738
81a5b4f5
N		 3739 if (strcmp(update, "uuid") == 0) {
3740 /* We take this to mean that the family_num should be updated.
3741 * However that is much smaller than the uuid so we cannot really
3742 * allow an explicit uuid to be given. And it is hard to reliably
3743 * know if one was.
3744 * So if !uuid_set we know the current uuid is random and just used
3745 * the first 'int' and copy it to the other 3 positions.
3746 * Otherwise we require the 4 'int's to be the same as would be the
3747 * case if we are using a random uuid. So an explicit uuid will be
3748 * accepted as long as all for ints are the same... which shouldn't hurt
6e46bf34 3749 */
81a5b4f5
N		 3750 if (!uuid_set) {
3751 info->uuid[1] = info->uuid[2] = info->uuid[3] = info->uuid[0];
6e46bf34 3752 rv = 0;
81a5b4f5
N		 3753 } else {
3754 if (info->uuid[0] != info->uuid[1] ||
3755 info->uuid[1] != info->uuid[2] ||
3756 info->uuid[2] != info->uuid[3])
3757 rv = -1;
3758 else
3759 rv = 0;
6e46bf34 3760 }
81a5b4f5
N		 3761 if (rv == 0)
3762 mpb->orig_family_num = info->uuid[0];
6e46bf34
DW		 3763 } else if (strcmp(update, "assemble") == 0)
3764 rv = 0;
3765 else
1e2b2765 3766 rv = -1;
f352c545 3767
6e46bf34
DW		 3768 /* successful update? recompute checksum */
3769 if (rv == 0)
3770 mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb));
f352c545
DW		 3771
3772 return rv;
cdddbdbc
DW		 3773}
3774
c2c087e6 3775static size_t disks_to_mpb_size(int disks)
cdddbdbc 3776{
c2c087e6 3777 size_t size;
cdddbdbc 3778
c2c087e6
DW		 3779 size = sizeof(struct imsm_super);
3780 size += (disks - 1) * sizeof(struct imsm_disk);
3781 size += 2 * sizeof(struct imsm_dev);
3782 /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */
3783 size += (4 - 2) * sizeof(struct imsm_map);
3784 /* 4 possible disk_ord_tbl's */
3785 size += 4 * (disks - 1) * sizeof(__u32);
bbab0940
TM		 3786 /* maximum bbm log */
3787 size += sizeof(struct bbm_log);
c2c087e6
DW		 3788
3789 return size;
3790}
3791
387fcd59
N		 3792static __u64 avail_size_imsm(struct supertype *st, __u64 devsize,
3793 unsigned long long data_offset)
c2c087e6
DW		 3794{
3795 if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS))
3796 return 0;
3797
3798 return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS);
cdddbdbc
DW		 3799}
3800
ba2de7ba
DW		 3801static void free_devlist(struct intel_super *super)
3802{
3803 struct intel_dev *dv;
3804
3805 while (super->devlist) {
3806 dv = super->devlist->next;
3807 free(super->devlist->dev);
3808 free(super->devlist);
3809 super->devlist = dv;
3810 }
3811}
3812
3813static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src)
3814{
3815 memcpy(dest, src, sizeof_imsm_dev(src, 0));
3816}
3817
c7b8547c
MT		 3818static int compare_super_imsm(struct supertype *st, struct supertype *tst,
3819 int verbose)
cdddbdbc
DW		 3820{
3821 /*
3822 * return:
3823 * 0 same, or first was empty, and second was copied
3824 * 1 second had wrong number
3825 * 2 wrong uuid
3826 * 3 wrong other info
3827 */
3828 struct intel_super *first = st->sb;
3829 struct intel_super *sec = tst->sb;
3830
5d500228
N		 3831 if (!first) {
3832 st->sb = tst->sb;
3833 tst->sb = NULL;
3834 return 0;
3835 }
8603ea6f
LM		 3836 /* in platform dependent environment test if the disks
3837 * use the same Intel hba
cb8f6859 3838 * If not on Intel hba at all, allow anything.
8603ea6f 3839 */
6b781d33
AP		 3840 if (!check_env("IMSM_NO_PLATFORM") && first->hba && sec->hba) {
3841 if (first->hba->type != sec->hba->type) {
c7b8547c
MT		 3842 if (verbose)
3843 pr_err("HBAs of devices do not match %s != %s\n",
3844 get_sys_dev_type(first->hba->type),
3845 get_sys_dev_type(sec->hba->type));
6b781d33
AP		 3846 return 3;
3847 }
c7b8547c 3848
6b781d33 3849 if (first->orom != sec->orom) {
c7b8547c
MT		 3850 if (verbose)
3851 pr_err("HBAs of devices do not match %s != %s\n",
3852 first->hba->pci_id, sec->hba->pci_id);
8603ea6f
LM		 3853 return 3;
3854 }
c7b8547c 3855
8603ea6f 3856 }
cdddbdbc 3857
d23fe947
DW		 3858 /* if an anchor does not have num_raid_devs set then it is a free
3859 * floating spare
3860 */
3861 if (first->anchor->num_raid_devs > 0 &&
3862 sec->anchor->num_raid_devs > 0) {
a2b97981
DW		 3863 /* Determine if these disks might ever have been
3864 * related. Further disambiguation can only take place
3865 * in load_super_imsm_all
3866 */
3867 __u32 first_family = first->anchor->orig_family_num;
3868 __u32 sec_family = sec->anchor->orig_family_num;
3869
f796af5d
DW		 3870 if (memcmp(first->anchor->sig, sec->anchor->sig,
3871 MAX_SIGNATURE_LENGTH) != 0)
3872 return 3;
3873
a2b97981
DW		 3874 if (first_family == 0)
3875 first_family = first->anchor->family_num;
3876 if (sec_family == 0)
3877 sec_family = sec->anchor->family_num;
3878
3879 if (first_family != sec_family)
d23fe947 3880 return 3;
f796af5d 3881
d23fe947 3882 }
cdddbdbc 3883
3e372e5a
DW		 3884 /* if 'first' is a spare promote it to a populated mpb with sec's
3885 * family number
3886 */
3887 if (first->anchor->num_raid_devs == 0 &&
3888 sec->anchor->num_raid_devs > 0) {
78d30f94 3889 int i;
ba2de7ba
DW		 3890 struct intel_dev *dv;
3891 struct imsm_dev *dev;
78d30f94
DW		 3892
3893 /* we need to copy raid device info from sec if an allocation
3894 * fails here we don't associate the spare
3895 */
3896 for (i = 0; i < sec->anchor->num_raid_devs; i++) {
503975b9
N		 3897 dv = xmalloc(sizeof(*dv));
3898 dev = xmalloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1));
ba2de7ba
DW		 3899 dv->dev = dev;
3900 dv->index = i;
3901 dv->next = first->devlist;
3902 first->devlist = dv;
78d30f94 3903 }
709743c5 3904 if (i < sec->anchor->num_raid_devs) {
ba2de7ba
DW		 3905 /* allocation failure */
3906 free_devlist(first);
e12b3daa 3907 pr_err("imsm: failed to associate spare\n");
ba2de7ba 3908 return 3;
78d30f94 3909 }
3e372e5a 3910 first->anchor->num_raid_devs = sec->anchor->num_raid_devs;
148acb7b 3911 first->anchor->orig_family_num = sec->anchor->orig_family_num;
3e372e5a 3912 first->anchor->family_num = sec->anchor->family_num;
ac6449be 3913 memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH);
709743c5
DW		 3914 for (i = 0; i < sec->anchor->num_raid_devs; i++)
3915 imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i));
3e372e5a
DW		 3916 }
3917
cdddbdbc
DW		 3918 return 0;
3919}
3920
0030e8d6
DW		 3921static void fd2devname(int fd, char *name)
3922{
3923 struct stat st;
3924 char path[256];
33a6535d 3925 char dname[PATH_MAX];
0030e8d6
DW		 3926 char *nm;
3927 int rv;
3928
3929 name[0] = '\0';
3930 if (fstat(fd, &st) != 0)
3931 return;
3932 sprintf(path, "/sys/dev/block/%d:%d",
3933 major(st.st_rdev), minor(st.st_rdev));
3934
9cf014ec 3935 rv = readlink(path, dname, sizeof(dname)-1);
0030e8d6
DW		 3936 if (rv <= 0)
3937 return;
9587c373 3938
0030e8d6
DW		 3939 dname[rv] = '\0';
3940 nm = strrchr(dname, '/');
7897de29
JS		 3941 if (nm) {
3942 nm++;
3943 snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm);
3944 }
0030e8d6
DW		 3945}
3946
21e9380b
AP		 3947static int nvme_get_serial(int fd, void *buf, size_t buf_len)
3948{
3949 char path[60];
3950 char *name = fd2kname(fd);
3951
3952 if (!name)
3953 return 1;
3954
3955 if (strncmp(name, "nvme", 4) != 0)
3956 return 1;
3957
3958 snprintf(path, sizeof(path) - 1, "/sys/block/%s/device/serial", name);
3959
3960 return load_sys(path, buf, buf_len);
3961}
3962
cdddbdbc
DW		 3963extern int scsi_get_serial(int fd, void *buf, size_t buf_len);
3964
3965static int imsm_read_serial(int fd, char *devname,
6da53c0e 3966 __u8 *serial, size_t serial_buf_len)
cdddbdbc 3967{
21e9380b 3968 char buf[50];
cdddbdbc 3969 int rv;
6da53c0e 3970 size_t len;
316e2bf4
DW		 3971 char *dest;
3972 char *src;
21e9380b
AP		 3973 unsigned int i;
3974
3975 memset(buf, 0, sizeof(buf));
cdddbdbc 3976
21e9380b 3977 rv = nvme_get_serial(fd, buf, sizeof(buf));
cdddbdbc 3978
21e9380b
AP		 3979 if (rv)
3980 rv = scsi_get_serial(fd, buf, sizeof(buf));
f9ba0ff1 3981
40ebbb9c 3982 if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) {
f9ba0ff1
DW		 3983 memset(serial, 0, MAX_RAID_SERIAL_LEN);
3984 fd2devname(fd, (char *) serial);
0030e8d6
DW		 3985 return 0;
3986 }
3987
cdddbdbc
DW		 3988 if (rv != 0) {
3989 if (devname)
e7b84f9d
N		 3990 pr_err("Failed to retrieve serial for %s\n",
3991 devname);
cdddbdbc
DW		 3992 return rv;
3993 }
3994
316e2bf4
DW		 3995 /* trim all whitespace and non-printable characters and convert
3996 * ':' to ';'
3997 */
21e9380b
AP		 3998 for (i = 0, dest = buf; i < sizeof(buf) && buf[i]; i++) {
3999 src = &buf[i];
316e2bf4
DW		 4000 if (*src > 0x20) {
4001 /* ':' is reserved for use in placeholder serial
4002 * numbers for missing disks
4003 */
4004 if (*src == ':')
4005 *dest++ = ';';
4006 else
4007 *dest++ = *src;
4008 }
4009 }
21e9380b
AP		 4010 len = dest - buf;
4011 dest = buf;
316e2bf4 4012
6da53c0e
BK		 4013 if (len > serial_buf_len) {
4014 /* truncate leading characters */
4015 dest += len - serial_buf_len;
4016 len = serial_buf_len;
316e2bf4 4017 }
5c3db629 4018
6da53c0e 4019 memset(serial, 0, serial_buf_len);
316e2bf4 4020 memcpy(serial, dest, len);
cdddbdbc
DW		 4021
4022 return 0;
4023}
4024
1f24f035
DW		 4025static int serialcmp(__u8 *s1, __u8 *s2)
4026{
4027 return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN);
4028}
4029
4030static void serialcpy(__u8 *dest, __u8 *src)
4031{
4032 strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN);
4033}
4034
54c2c1ea
DW		 4035static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super)
4036{
4037 struct dl *dl;
4038
4039 for (dl = super->disks; dl; dl = dl->next)
4040 if (serialcmp(dl->serial, serial) == 0)
4041 break;
4042
4043 return dl;
4044}
4045
a2b97981
DW		 4046static struct imsm_disk *
4047__serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx)
4048{
4049 int i;
4050
4051 for (i = 0; i < mpb->num_disks; i++) {
4052 struct imsm_disk *disk = __get_imsm_disk(mpb, i);
4053
4054 if (serialcmp(disk->serial, serial) == 0) {
4055 if (idx)
4056 *idx = i;
4057 return disk;
4058 }
4059 }
4060
4061 return NULL;
4062}
4063
cdddbdbc
DW		 4064static int
4065load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd)
4066{
a2b97981 4067 struct imsm_disk *disk;
cdddbdbc
DW		 4068 struct dl *dl;
4069 struct stat stb;
cdddbdbc 4070 int rv;
a2b97981 4071 char name[40];
d23fe947
DW		 4072 __u8 serial[MAX_RAID_SERIAL_LEN];
4073
6da53c0e 4074 rv = imsm_read_serial(fd, devname, serial, MAX_RAID_SERIAL_LEN);
d23fe947
DW		 4075
4076 if (rv != 0)
4077 return 2;
4078
503975b9 4079 dl = xcalloc(1, sizeof(*dl));
cdddbdbc 4080
a2b97981
DW		 4081 fstat(fd, &stb);
4082 dl->major = major(stb.st_rdev);
4083 dl->minor = minor(stb.st_rdev);
4084 dl->next = super->disks;
4085 dl->fd = keep_fd ? fd : -1;
4086 assert(super->disks == NULL);
4087 super->disks = dl;
4088 serialcpy(dl->serial, serial);
4089 dl->index = -2;
4090 dl->e = NULL;
4091 fd2devname(fd, name);
4092 if (devname)
503975b9 4093 dl->devname = xstrdup(devname);
a2b97981 4094 else
503975b9 4095 dl->devname = xstrdup(name);
cdddbdbc 4096
d23fe947 4097 /* look up this disk's index in the current anchor */
a2b97981
DW		 4098 disk = __serial_to_disk(dl->serial, super->anchor, &dl->index);
4099 if (disk) {
4100 dl->disk = *disk;
4101 /* only set index on disks that are a member of a
4102 * populated contianer, i.e. one with raid_devs
4103 */
4104 if (is_failed(&dl->disk))
3f6efecc 4105 dl->index = -2;
2432ce9b 4106 else if (is_spare(&dl->disk) || is_journal(&dl->disk))
a2b97981 4107 dl->index = -1;
3f6efecc
DW		 4108 }
4109
949c47a0
DW		 4110 return 0;
4111}
4112
0c046afd
DW		 4113/* When migrating map0 contains the 'destination' state while map1
4114 * contains the current state. When not migrating map0 contains the
4115 * current state. This routine assumes that map[0].map_state is set to
4116 * the current array state before being called.
4117 *
4118 * Migration is indicated by one of the following states
4119 * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed)
e3bba0e0 4120 * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal
0c046afd 4121 * map1state=unitialized)
1484e727 4122 * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal
0c046afd 4123 * map1state=normal)
e3bba0e0 4124 * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal
0c046afd 4125 * map1state=degraded)
8e59f3d8
AK		 4126 * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal
4127 * map1state=normal)
0c046afd 4128 */
8e59f3d8
AK		 4129static void migrate(struct imsm_dev *dev, struct intel_super *super,
4130 __u8 to_state, int migr_type)
3393c6af 4131{
0c046afd 4132 struct imsm_map *dest;
238c0a71 4133 struct imsm_map *src = get_imsm_map(dev, MAP_0);
3393c6af 4134
0c046afd 4135 dev->vol.migr_state = 1;
1484e727 4136 set_migr_type(dev, migr_type);
f8f603f1 4137 dev->vol.curr_migr_unit = 0;
238c0a71 4138 dest = get_imsm_map(dev, MAP_1);
0c046afd 4139
0556e1a2 4140 /* duplicate and then set the target end state in map[0] */
3393c6af 4141 memcpy(dest, src, sizeof_imsm_map(src));
fb12a745 4142 if (migr_type == MIGR_GEN_MIGR) {
0556e1a2
DW		 4143 __u32 ord;
4144 int i;
4145
4146 for (i = 0; i < src->num_members; i++) {
4147 ord = __le32_to_cpu(src->disk_ord_tbl[i]);
4148 set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord));
4149 }
4150 }
4151
8e59f3d8
AK		 4152 if (migr_type == MIGR_GEN_MIGR)
4153 /* Clear migration record */
4154 memset(super->migr_rec, 0, sizeof(struct migr_record));
4155
0c046afd 4156 src->map_state = to_state;
949c47a0 4157}
f8f603f1 4158
809da78e
AK		 4159static void end_migration(struct imsm_dev *dev, struct intel_super *super,
4160 __u8 map_state)
f8f603f1 4161{
238c0a71
AK		 4162 struct imsm_map *map = get_imsm_map(dev, MAP_0);
4163 struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state == 0 ?
4164 MAP_0 : MAP_1);
28bce06f 4165 int i, j;
0556e1a2
DW		 4166
4167 /* merge any IMSM_ORD_REBUILD bits that were not successfully
4168 * completed in the last migration.
4169 *
28bce06f 4170 * FIXME add support for raid-level-migration
0556e1a2 4171 */
089f9d79
JS		 4172 if (map_state != map->map_state && (is_gen_migration(dev) == 0) &&
4173 prev->map_state != IMSM_T_STATE_UNINITIALIZED) {
809da78e
AK		 4174 /* when final map state is other than expected
4175 * merge maps (not for migration)
4176 */
4177 int failed;
4178
4179 for (i = 0; i < prev->num_members; i++)
4180 for (j = 0; j < map->num_members; j++)
4181 /* during online capacity expansion
4182 * disks position can be changed
4183 * if takeover is used
4184 */
4185 if (ord_to_idx(map->disk_ord_tbl[j]) ==
4186 ord_to_idx(prev->disk_ord_tbl[i])) {
4187 map->disk_ord_tbl[j] |=
4188 prev->disk_ord_tbl[i];
4189 break;
4190 }
4191 failed = imsm_count_failed(super, dev, MAP_0);
4192 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
4193 }
f8f603f1
DW		 4194
4195 dev->vol.migr_state = 0;
ea672ee1 4196 set_migr_type(dev, 0);
f8f603f1
DW		 4197 dev->vol.curr_migr_unit = 0;
4198 map->map_state = map_state;
4199}
949c47a0
DW		 4200
4201static int parse_raid_devices(struct intel_super *super)
4202{
4203 int i;
4204 struct imsm_dev *dev_new;
4d7b1503 4205 size_t len, len_migr;
401d313b 4206 size_t max_len = 0;
4d7b1503
DW		 4207 size_t space_needed = 0;
4208 struct imsm_super *mpb = super->anchor;
949c47a0
DW		 4209
4210 for (i = 0; i < super->anchor->num_raid_devs; i++) {
4211 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
ba2de7ba 4212 struct intel_dev *dv;
949c47a0 4213
4d7b1503
DW		 4214 len = sizeof_imsm_dev(dev_iter, 0);
4215 len_migr = sizeof_imsm_dev(dev_iter, 1);
4216 if (len_migr > len)
4217 space_needed += len_migr - len;
ca9de185 4218
503975b9 4219 dv = xmalloc(sizeof(*dv));
401d313b
AK		 4220 if (max_len < len_migr)
4221 max_len = len_migr;
4222 if (max_len > len_migr)
4223 space_needed += max_len - len_migr;
503975b9 4224 dev_new = xmalloc(max_len);
949c47a0 4225 imsm_copy_dev(dev_new, dev_iter);
ba2de7ba
DW		 4226 dv->dev = dev_new;
4227 dv->index = i;
4228 dv->next = super->devlist;
4229 super->devlist = dv;
949c47a0 4230 }
cdddbdbc 4231
4d7b1503
DW		 4232 /* ensure that super->buf is large enough when all raid devices
4233 * are migrating
4234 */
4235 if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) {
4236 void *buf;
4237
f36a9ecd
PB		 4238 len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed,
4239 super->sector_size);
4240 if (posix_memalign(&buf, MAX_SECTOR_SIZE, len) != 0)
4d7b1503
DW		 4241 return 1;
4242
1f45a8ad
DW		 4243 memcpy(buf, super->buf, super->len);
4244 memset(buf + super->len, 0, len - super->len);
4d7b1503
DW		 4245 free(super->buf);
4246 super->buf = buf;
4247 super->len = len;
4248 }
ca9de185 4249
bbab0940
TM		 4250 super->extra_space += space_needed;
4251
cdddbdbc
DW		 4252 return 0;
4253}
4254
e2f41b2c
AK		 4255/*******************************************************************************
4256 * Function: check_mpb_migr_compatibility
4257 * Description: Function checks for unsupported migration features:
4258 * - migration optimization area (pba_of_lba0)
4259 * - descending reshape (ascending_migr)
4260 * Parameters:
4261 * super : imsm metadata information
4262 * Returns:
4263 * 0 : migration is compatible
4264 * -1 : migration is not compatible
4265 ******************************************************************************/
4266int check_mpb_migr_compatibility(struct intel_super *super)
4267{
4268 struct imsm_map *map0, *map1;
4269 struct migr_record *migr_rec = super->migr_rec;
4270 int i;
4271
4272 for (i = 0; i < super->anchor->num_raid_devs; i++) {
4273 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
4274
4275 if (dev_iter &&
4276 dev_iter->vol.migr_state == 1 &&
4277 dev_iter->vol.migr_type == MIGR_GEN_MIGR) {
4278 /* This device is migrating */
238c0a71
AK		 4279 map0 = get_imsm_map(dev_iter, MAP_0);
4280 map1 = get_imsm_map(dev_iter, MAP_1);
5551b113 4281 if (pba_of_lba0(map0) != pba_of_lba0(map1))
e2f41b2c
AK		 4282 /* migration optimization area was used */
4283 return -1;
fc54fe7a
JS		 4284 if (migr_rec->ascending_migr == 0 &&
4285 migr_rec->dest_depth_per_unit > 0)
e2f41b2c
AK		 4286 /* descending reshape not supported yet */
4287 return -1;
4288 }
4289 }
4290 return 0;
4291}
4292
d23fe947 4293static void __free_imsm(struct intel_super *super, int free_disks);
9ca2c81c 4294
cdddbdbc 4295/* load_imsm_mpb - read matrix metadata
f2f5c343 4296 * allocates super->mpb to be freed by free_imsm
cdddbdbc
DW		 4297 */
4298static int load_imsm_mpb(int fd, struct intel_super *super, char *devname)
4299{
4300 unsigned long long dsize;
cdddbdbc 4301 unsigned long long sectors;
f36a9ecd 4302 unsigned int sector_size = super->sector_size;
cdddbdbc 4303 struct stat;
6416d527 4304 struct imsm_super *anchor;
cdddbdbc
DW		 4305 __u32 check_sum;
4306
cdddbdbc 4307 get_dev_size(fd, NULL, &dsize);
f36a9ecd 4308 if (dsize < 2*sector_size) {
64436f06 4309 if (devname)
e7b84f9d
N		 4310 pr_err("%s: device to small for imsm\n",
4311 devname);
64436f06
N		 4312 return 1;
4313 }
cdddbdbc 4314
f36a9ecd 4315 if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0) {
cdddbdbc 4316 if (devname)
e7b84f9d
N		 4317 pr_err("Cannot seek to anchor block on %s: %s\n",
4318 devname, strerror(errno));
cdddbdbc
DW		 4319 return 1;
4320 }
4321
f36a9ecd 4322 if (posix_memalign((void **)&anchor, sector_size, sector_size) != 0) {
ad97895e 4323 if (devname)
7a862a02 4324 pr_err("Failed to allocate imsm anchor buffer on %s\n", devname);
ad97895e
DW		 4325 return 1;
4326 }
466070ad 4327 if ((unsigned int)read(fd, anchor, sector_size) != sector_size) {
cdddbdbc 4328 if (devname)
e7b84f9d
N		 4329 pr_err("Cannot read anchor block on %s: %s\n",
4330 devname, strerror(errno));
6416d527 4331 free(anchor);
cdddbdbc
DW		 4332 return 1;
4333 }
4334
6416d527 4335 if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) {
cdddbdbc 4336 if (devname)
e7b84f9d 4337 pr_err("no IMSM anchor on %s\n", devname);
6416d527 4338 free(anchor);
cdddbdbc
DW		 4339 return 2;
4340 }
4341
d23fe947 4342 __free_imsm(super, 0);
f2f5c343
LM		 4343 /* reload capability and hba */
4344
4345 /* capability and hba must be updated with new super allocation */
d424212e 4346 find_intel_hba_capability(fd, super, devname);
f36a9ecd
PB		 4347 super->len = ROUND_UP(anchor->mpb_size, sector_size);
4348 if (posix_memalign(&super->buf, MAX_SECTOR_SIZE, super->len) != 0) {
cdddbdbc 4349 if (devname)
e7b84f9d
N		 4350 pr_err("unable to allocate %zu byte mpb buffer\n",
4351 super->len);
6416d527 4352 free(anchor);
cdddbdbc
DW		 4353 return 2;
4354 }
f36a9ecd 4355 memcpy(super->buf, anchor, sector_size);
cdddbdbc 4356
f36a9ecd 4357 sectors = mpb_sectors(anchor, sector_size) - 1;
6416d527 4358 free(anchor);
8e59f3d8 4359
85337573
AO		 4360 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
4361 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) {
1ade5cc1 4362 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8
AK		 4363 free(super->buf);
4364 return 2;
4365 }
51d83f5d 4366 super->clean_migration_record_by_mdmon = 0;
8e59f3d8 4367
949c47a0 4368 if (!sectors) {
ecf45690
DW		 4369 check_sum = __gen_imsm_checksum(super->anchor);
4370 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
4371 if (devname)
e7b84f9d
N		 4372 pr_err("IMSM checksum %x != %x on %s\n",
4373 check_sum,
4374 __le32_to_cpu(super->anchor->check_sum),
4375 devname);
ecf45690
DW		 4376 return 2;
4377 }
4378
a2b97981 4379 return 0;
949c47a0 4380 }
cdddbdbc
DW		 4381
4382 /* read the extended mpb */
f36a9ecd 4383 if (lseek64(fd, dsize - (sector_size * (2 + sectors)), SEEK_SET) < 0) {
cdddbdbc 4384 if (devname)
e7b84f9d
N		 4385 pr_err("Cannot seek to extended mpb on %s: %s\n",
4386 devname, strerror(errno));
cdddbdbc
DW		 4387 return 1;
4388 }
4389
f36a9ecd
PB		 4390 if ((unsigned int)read(fd, super->buf + sector_size,
4391 super->len - sector_size) != super->len - sector_size) {
cdddbdbc 4392 if (devname)
e7b84f9d
N		 4393 pr_err("Cannot read extended mpb on %s: %s\n",
4394 devname, strerror(errno));
cdddbdbc
DW		 4395 return 2;
4396 }
4397
949c47a0
DW		 4398 check_sum = __gen_imsm_checksum(super->anchor);
4399 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
cdddbdbc 4400 if (devname)
e7b84f9d
N		 4401 pr_err("IMSM checksum %x != %x on %s\n",
4402 check_sum, __le32_to_cpu(super->anchor->check_sum),
4403 devname);
db575f3b 4404 return 3;
cdddbdbc
DW		 4405 }
4406
a2b97981
DW		 4407 return 0;
4408}
4409
8e59f3d8
AK		 4410static int read_imsm_migr_rec(int fd, struct intel_super *super);
4411
97f81ee2
CA		 4412/* clears hi bits in metadata if MPB_ATTRIB_2TB_DISK not set */
4413static void clear_hi(struct intel_super *super)
4414{
4415 struct imsm_super *mpb = super->anchor;
4416 int i, n;
4417 if (mpb->attributes & MPB_ATTRIB_2TB_DISK)
4418 return;
4419 for (i = 0; i < mpb->num_disks; ++i) {
4420 struct imsm_disk *disk = &mpb->disk[i];
4421 disk->total_blocks_hi = 0;
4422 }
4423 for (i = 0; i < mpb->num_raid_devs; ++i) {
4424 struct imsm_dev *dev = get_imsm_dev(super, i);
4425 if (!dev)
4426 return;
4427 for (n = 0; n < 2; ++n) {
4428 struct imsm_map *map = get_imsm_map(dev, n);
4429 if (!map)
4430 continue;
4431 map->pba_of_lba0_hi = 0;
4432 map->blocks_per_member_hi = 0;
4433 map->num_data_stripes_hi = 0;
4434 }
4435 }
4436}
4437
a2b97981
DW		 4438static int
4439load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd)
4440{
4441 int err;
4442
4443 err = load_imsm_mpb(fd, super, devname);
4444 if (err)
4445 return err;
f36a9ecd
PB		 4446 if (super->sector_size == 4096)
4447 convert_from_4k(super);
a2b97981
DW		 4448 err = load_imsm_disk(fd, super, devname, keep_fd);
4449 if (err)
4450 return err;
4451 err = parse_raid_devices(super);
8d67477f
TM		 4452 if (err)
4453 return err;
4454 err = load_bbm_log(super);
97f81ee2 4455 clear_hi(super);
a2b97981 4456 return err;
cdddbdbc
DW		 4457}
4458
ae6aad82
DW		 4459static void __free_imsm_disk(struct dl *d)
4460{
4461 if (d->fd >= 0)
4462 close(d->fd);
4463 if (d->devname)
4464 free(d->devname);
0dcecb2e
DW		 4465 if (d->e)
4466 free(d->e);
ae6aad82
DW		 4467 free(d);
4468
4469}
1a64be56 4470
cdddbdbc
DW		 4471static void free_imsm_disks(struct intel_super *super)
4472{
47ee5a45 4473 struct dl *d;
cdddbdbc 4474
47ee5a45
DW		 4475 while (super->disks) {
4476 d = super->disks;
cdddbdbc 4477 super->disks = d->next;
ae6aad82 4478 __free_imsm_disk(d);
cdddbdbc 4479 }
cb82edca
AK		 4480 while (super->disk_mgmt_list) {
4481 d = super->disk_mgmt_list;
4482 super->disk_mgmt_list = d->next;
4483 __free_imsm_disk(d);
4484 }
47ee5a45
DW		 4485 while (super->missing) {
4486 d = super->missing;
4487 super->missing = d->next;
4488 __free_imsm_disk(d);
4489 }
4490
cdddbdbc
DW		 4491}
4492
9ca2c81c 4493/* free all the pieces hanging off of a super pointer */
d23fe947 4494static void __free_imsm(struct intel_super *super, int free_disks)
cdddbdbc 4495{
88654014
LM		 4496 struct intel_hba *elem, *next;
4497
9ca2c81c 4498 if (super->buf) {
949c47a0 4499 free(super->buf);
9ca2c81c
DW		 4500 super->buf = NULL;
4501 }
f2f5c343
LM		 4502 /* unlink capability description */
4503 super->orom = NULL;
8e59f3d8
AK		 4504 if (super->migr_rec_buf) {
4505 free(super->migr_rec_buf);
4506 super->migr_rec_buf = NULL;
4507 }
d23fe947
DW		 4508 if (free_disks)
4509 free_imsm_disks(super);
ba2de7ba 4510 free_devlist(super);
88654014
LM		 4511 elem = super->hba;
4512 while (elem) {
4513 if (elem->path)
4514 free((void *)elem->path);
4515 next = elem->next;
4516 free(elem);
4517 elem = next;
88c32bb1 4518 }
8d67477f
TM		 4519 if (super->bbm_log)
4520 free(super->bbm_log);
88654014 4521 super->hba = NULL;
cdddbdbc
DW		 4522}
4523
9ca2c81c
DW		 4524static void free_imsm(struct intel_super *super)
4525{
d23fe947 4526 __free_imsm(super, 1);
928f1424 4527 free(super->bb.entries);
9ca2c81c
DW		 4528 free(super);
4529}
cdddbdbc
DW		 4530
4531static void free_super_imsm(struct supertype *st)
4532{
4533 struct intel_super *super = st->sb;
4534
4535 if (!super)
4536 return;
4537
4538 free_imsm(super);
4539 st->sb = NULL;
4540}
4541
49133e57 4542static struct intel_super *alloc_super(void)
c2c087e6 4543{
503975b9 4544 struct intel_super *super = xcalloc(1, sizeof(*super));
c2c087e6 4545
503975b9
N		 4546 super->current_vol = -1;
4547 super->create_offset = ~((unsigned long long) 0);
928f1424
TM		 4548
4549 super->bb.entries = xmalloc(BBM_LOG_MAX_ENTRIES *
4550 sizeof(struct md_bb_entry));
4551 if (!super->bb.entries) {
4552 free(super);
4553 return NULL;
4554 }
4555
c2c087e6
DW		 4556 return super;
4557}
4558
f0f5a016
LM		 4559/*
4560 * find and allocate hba and OROM/EFI based on valid fd of RAID component device
4561 */
d424212e 4562static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname)
f0f5a016
LM		 4563{
4564 struct sys_dev *hba_name;
4565 int rv = 0;
4566
3a30e28e
MT		 4567 if (fd >= 0 && test_partition(fd)) {
4568 pr_err("imsm: %s is a partition, cannot be used in IMSM\n",
4569 devname);
4570 return 1;
4571 }
089f9d79 4572 if (fd < 0 || check_env("IMSM_NO_PLATFORM")) {
f2f5c343 4573 super->orom = NULL;
f0f5a016
LM		 4574 super->hba = NULL;
4575 return 0;
4576 }
4577 hba_name = find_disk_attached_hba(fd, NULL);
4578 if (!hba_name) {
d424212e 4579 if (devname)
e7b84f9d
N		 4580 pr_err("%s is not attached to Intel(R) RAID controller.\n",
4581 devname);
f0f5a016
LM		 4582 return 1;
4583 }
4584 rv = attach_hba_to_super(super, hba_name);
4585 if (rv == 2) {
d424212e
N		 4586 if (devname) {
4587 struct intel_hba *hba = super->hba;
f0f5a016 4588
60f0f54d
PB		 4589 pr_err("%s is attached to Intel(R) %s %s (%s),\n"
4590 " but the container is assigned to Intel(R) %s %s (",
d424212e 4591 devname,
614902f6 4592 get_sys_dev_type(hba_name->type),
60f0f54d 4593 hba_name->type == SYS_DEV_VMD ? "domain" : "RAID controller",
f0f5a016 4594 hba_name->pci_id ? : "Err!",
60f0f54d
PB		 4595 get_sys_dev_type(super->hba->type),
4596 hba->type == SYS_DEV_VMD ? "domain" : "RAID controller");
f0f5a016 4597
f0f5a016
LM		 4598 while (hba) {
4599 fprintf(stderr, "%s", hba->pci_id ? : "Err!");
4600 if (hba->next)
4601 fprintf(stderr, ", ");
4602 hba = hba->next;
4603 }
6b781d33 4604 fprintf(stderr, ").\n"
cca67208 4605 " Mixing devices attached to different controllers is not allowed.\n");
f0f5a016 4606 }
f0f5a016
LM		 4607 return 2;
4608 }
6b781d33 4609 super->orom = find_imsm_capability(hba_name);
f2f5c343
LM		 4610 if (!super->orom)
4611 return 3;
614902f6 4612
f0f5a016
LM		 4613 return 0;
4614}
4615
47ee5a45
DW		 4616/* find_missing - helper routine for load_super_imsm_all that identifies
4617 * disks that have disappeared from the system. This routine relies on
4618 * the mpb being uptodate, which it is at load time.
4619 */
4620static int find_missing(struct intel_super *super)
4621{
4622 int i;
4623 struct imsm_super *mpb = super->anchor;
4624 struct dl *dl;
4625 struct imsm_disk *disk;
47ee5a45
DW		 4626
4627 for (i = 0; i < mpb->num_disks; i++) {
4628 disk = __get_imsm_disk(mpb, i);
54c2c1ea 4629 dl = serial_to_dl(disk->serial, super);
47ee5a45
DW		 4630 if (dl)
4631 continue;
47ee5a45 4632
503975b9 4633 dl = xmalloc(sizeof(*dl));
47ee5a45
DW		 4634 dl->major = 0;
4635 dl->minor = 0;
4636 dl->fd = -1;
503975b9 4637 dl->devname = xstrdup("missing");
47ee5a45
DW		 4638 dl->index = i;
4639 serialcpy(dl->serial, disk->serial);
4640 dl->disk = *disk;
689c9bf3 4641 dl->e = NULL;
47ee5a45
DW		 4642 dl->next = super->missing;
4643 super->missing = dl;
4644 }
4645
4646 return 0;
4647}
4648
a2b97981
DW		 4649static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list)
4650{
4651 struct intel_disk *idisk = disk_list;
4652
4653 while (idisk) {
4654 if (serialcmp(idisk->disk.serial, serial) == 0)
4655 break;
4656 idisk = idisk->next;
4657 }
4658
4659 return idisk;
4660}
4661
4662static int __prep_thunderdome(struct intel_super **table, int tbl_size,
4663 struct intel_super *super,
4664 struct intel_disk **disk_list)
4665{
4666 struct imsm_disk *d = &super->disks->disk;
4667 struct imsm_super *mpb = super->anchor;
4668 int i, j;
4669
4670 for (i = 0; i < tbl_size; i++) {
4671 struct imsm_super *tbl_mpb = table[i]->anchor;
4672 struct imsm_disk *tbl_d = &table[i]->disks->disk;
4673
4674 if (tbl_mpb->family_num == mpb->family_num) {
4675 if (tbl_mpb->check_sum == mpb->check_sum) {
1ade5cc1
N		 4676 dprintf("mpb from %d:%d matches %d:%d\n",
4677 super->disks->major,
a2b97981
DW		 4678 super->disks->minor,
4679 table[i]->disks->major,
4680 table[i]->disks->minor);
4681 break;
4682 }
4683
4684 if (((is_configured(d) && !is_configured(tbl_d)) ||
4685 is_configured(d) == is_configured(tbl_d)) &&
4686 tbl_mpb->generation_num < mpb->generation_num) {
4687 /* current version of the mpb is a
4688 * better candidate than the one in
4689 * super_table, but copy over "cross
4690 * generational" status
4691 */
4692 struct intel_disk *idisk;
4693
1ade5cc1
N		 4694 dprintf("mpb from %d:%d replaces %d:%d\n",
4695 super->disks->major,
a2b97981
DW		 4696 super->disks->minor,
4697 table[i]->disks->major,
4698 table[i]->disks->minor);
4699
4700 idisk = disk_list_get(tbl_d->serial, *disk_list);
4701 if (idisk && is_failed(&idisk->disk))
4702 tbl_d->status |= FAILED_DISK;
4703 break;
4704 } else {
4705 struct intel_disk *idisk;
4706 struct imsm_disk *disk;
4707
4708 /* tbl_mpb is more up to date, but copy
4709 * over cross generational status before
4710 * returning
4711 */
4712 disk = __serial_to_disk(d->serial, mpb, NULL);
4713 if (disk && is_failed(disk))
4714 d->status |= FAILED_DISK;
4715
4716 idisk = disk_list_get(d->serial, *disk_list);
4717 if (idisk) {
4718 idisk->owner = i;
4719 if (disk && is_configured(disk))
4720 idisk->disk.status |= CONFIGURED_DISK;
4721 }
4722
1ade5cc1
N		 4723 dprintf("mpb from %d:%d prefer %d:%d\n",
4724 super->disks->major,
a2b97981
DW		 4725 super->disks->minor,
4726 table[i]->disks->major,
4727 table[i]->disks->minor);
4728
4729 return tbl_size;
4730 }
4731 }
4732 }
4733
4734 if (i >= tbl_size)
4735 table[tbl_size++] = super;
4736 else
4737 table[i] = super;
4738
4739 /* update/extend the merged list of imsm_disk records */
4740 for (j = 0; j < mpb->num_disks; j++) {
4741 struct imsm_disk *disk = __get_imsm_disk(mpb, j);
4742 struct intel_disk *idisk;
4743
4744 idisk = disk_list_get(disk->serial, *disk_list);
4745 if (idisk) {
4746 idisk->disk.status |= disk->status;
4747 if (is_configured(&idisk->disk) ||
4748 is_failed(&idisk->disk))
4749 idisk->disk.status &= ~(SPARE_DISK);
4750 } else {
503975b9 4751 idisk = xcalloc(1, sizeof(*idisk));
a2b97981
DW		 4752 idisk->owner = IMSM_UNKNOWN_OWNER;
4753 idisk->disk = *disk;
4754 idisk->next = *disk_list;
4755 *disk_list = idisk;
4756 }
4757
4758 if (serialcmp(idisk->disk.serial, d->serial) == 0)
4759 idisk->owner = i;
4760 }
4761
4762 return tbl_size;
4763}
4764
4765static struct intel_super *
4766validate_members(struct intel_super *super, struct intel_disk *disk_list,
4767 const int owner)
4768{
4769 struct imsm_super *mpb = super->anchor;
4770 int ok_count = 0;
4771 int i;
4772
4773 for (i = 0; i < mpb->num_disks; i++) {
4774 struct imsm_disk *disk = __get_imsm_disk(mpb, i);
4775 struct intel_disk *idisk;
4776
4777 idisk = disk_list_get(disk->serial, disk_list);
4778 if (idisk) {
4779 if (idisk->owner == owner ||
4780 idisk->owner == IMSM_UNKNOWN_OWNER)
4781 ok_count++;
4782 else
1ade5cc1
N		 4783 dprintf("'%.16s' owner %d != %d\n",
4784 disk->serial, idisk->owner,
a2b97981
DW		 4785 owner);
4786 } else {
1ade5cc1
N		 4787 dprintf("unknown disk %x [%d]: %.16s\n",
4788 __le32_to_cpu(mpb->family_num), i,
a2b97981
DW		 4789 disk->serial);
4790 break;
4791 }
4792 }
4793
4794 if (ok_count == mpb->num_disks)
4795 return super;
4796 return NULL;
4797}
4798
4799static void show_conflicts(__u32 family_num, struct intel_super *super_list)
4800{
4801 struct intel_super *s;
4802
4803 for (s = super_list; s; s = s->next) {
4804 if (family_num != s->anchor->family_num)
4805 continue;
e12b3daa 4806 pr_err("Conflict, offlining family %#x on '%s'\n",
a2b97981
DW		 4807 __le32_to_cpu(family_num), s->disks->devname);
4808 }
4809}
4810
4811static struct intel_super *
4812imsm_thunderdome(struct intel_super **super_list, int len)
4813{
4814 struct intel_super *super_table[len];
4815 struct intel_disk *disk_list = NULL;
4816 struct intel_super *champion, *spare;
4817 struct intel_super *s, **del;
4818 int tbl_size = 0;
4819 int conflict;
4820 int i;
4821
4822 memset(super_table, 0, sizeof(super_table));
4823 for (s = *super_list; s; s = s->next)
4824 tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list);
4825
4826 for (i = 0; i < tbl_size; i++) {
4827 struct imsm_disk *d;
4828 struct intel_disk *idisk;
4829 struct imsm_super *mpb = super_table[i]->anchor;
4830
4831 s = super_table[i];
4832 d = &s->disks->disk;
4833
4834 /* 'd' must appear in merged disk list for its
4835 * configuration to be valid
4836 */
4837 idisk = disk_list_get(d->serial, disk_list);
4838 if (idisk && idisk->owner == i)
4839 s = validate_members(s, disk_list, i);
4840 else
4841 s = NULL;
4842
4843 if (!s)
1ade5cc1
N		 4844 dprintf("marking family: %#x from %d:%d offline\n",
4845 mpb->family_num,
a2b97981
DW		 4846 super_table[i]->disks->major,
4847 super_table[i]->disks->minor);
4848 super_table[i] = s;
4849 }
4850
4851 /* This is where the mdadm implementation differs from the Windows
4852 * driver which has no strict concept of a container. We can only
4853 * assemble one family from a container, so when returning a prodigal
4854 * array member to this system the code will not be able to disambiguate
4855 * the container contents that should be assembled ("foreign" versus
4856 * "local"). It requires user intervention to set the orig_family_num
4857 * to a new value to establish a new container. The Windows driver in
4858 * this situation fixes up the volume name in place and manages the
4859 * foreign array as an independent entity.
4860 */
4861 s = NULL;
4862 spare = NULL;
4863 conflict = 0;
4864 for (i = 0; i < tbl_size; i++) {
4865 struct intel_super *tbl_ent = super_table[i];
4866 int is_spare = 0;
4867
4868 if (!tbl_ent)
4869 continue;
4870
4871 if (tbl_ent->anchor->num_raid_devs == 0) {
4872 spare = tbl_ent;
4873 is_spare = 1;
4874 }
4875
4876 if (s && !is_spare) {
4877 show_conflicts(tbl_ent->anchor->family_num, *super_list);
4878 conflict++;
4879 } else if (!s && !is_spare)
4880 s = tbl_ent;
4881 }
4882
4883 if (!s)
4884 s = spare;
4885 if (!s) {
4886 champion = NULL;
4887 goto out;
4888 }
4889 champion = s;
4890
4891 if (conflict)
7a862a02 4892 pr_err("Chose family %#x on '%s', assemble conflicts to new container with '--update=uuid'\n",
a2b97981
DW		 4893 __le32_to_cpu(s->anchor->family_num), s->disks->devname);
4894
4895 /* collect all dl's onto 'champion', and update them to
4896 * champion's version of the status
4897 */
4898 for (s = *super_list; s; s = s->next) {
4899 struct imsm_super *mpb = champion->anchor;
4900 struct dl *dl = s->disks;
4901
4902 if (s == champion)
4903 continue;
4904
5d7b407a
CA		 4905 mpb->attributes |= s->anchor->attributes & MPB_ATTRIB_2TB_DISK;
4906
a2b97981
DW		 4907 for (i = 0; i < mpb->num_disks; i++) {
4908 struct imsm_disk *disk;
4909
4910 disk = __serial_to_disk(dl->serial, mpb, &dl->index);
4911 if (disk) {
4912 dl->disk = *disk;
4913 /* only set index on disks that are a member of
4914 * a populated contianer, i.e. one with
4915 * raid_devs
4916 */
4917 if (is_failed(&dl->disk))
4918 dl->index = -2;
4919 else if (is_spare(&dl->disk))
4920 dl->index = -1;
4921 break;
4922 }
4923 }
4924
4925 if (i >= mpb->num_disks) {
4926 struct intel_disk *idisk;
4927
4928 idisk = disk_list_get(dl->serial, disk_list);
ecf408e9 4929 if (idisk && is_spare(&idisk->disk) &&
a2b97981
DW		 4930 !is_failed(&idisk->disk) && !is_configured(&idisk->disk))
4931 dl->index = -1;
4932 else {
4933 dl->index = -2;
4934 continue;
4935 }
4936 }
4937
4938 dl->next = champion->disks;
4939 champion->disks = dl;
4940 s->disks = NULL;
4941 }
4942
4943 /* delete 'champion' from super_list */
4944 for (del = super_list; *del; ) {
4945 if (*del == champion) {
4946 *del = (*del)->next;
4947 break;
4948 } else
4949 del = &(*del)->next;
4950 }
4951 champion->next = NULL;
4952
4953 out:
4954 while (disk_list) {
4955 struct intel_disk *idisk = disk_list;
4956
4957 disk_list = disk_list->next;
4958 free(idisk);
4959 }
4960
4961 return champion;
4962}
4963
9587c373
LM		 4964static int
4965get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd);
4dd2df09 4966static int get_super_block(struct intel_super **super_list, char *devnm, char *devname,
9587c373 4967 int major, int minor, int keep_fd);
ec50f7b6
LM		 4968static int
4969get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list,
4970 int *max, int keep_fd);
4971
cdddbdbc 4972static int load_super_imsm_all(struct supertype *st, int fd, void **sbp,
ec50f7b6
LM		 4973 char *devname, struct md_list *devlist,
4974 int keep_fd)
cdddbdbc 4975{
a2b97981
DW		 4976 struct intel_super *super_list = NULL;
4977 struct intel_super *super = NULL;
a2b97981 4978 int err = 0;
9587c373 4979 int i = 0;
dab4a513 4980
9587c373
LM		 4981 if (fd >= 0)
4982 /* 'fd' is an opened container */
4983 err = get_sra_super_block(fd, &super_list, devname, &i, keep_fd);
4984 else
ec50f7b6
LM		 4985 /* get super block from devlist devices */
4986 err = get_devlist_super_block(devlist, &super_list, &i, keep_fd);
9587c373 4987 if (err)
1602d52c 4988 goto error;
a2b97981
DW		 4989 /* all mpbs enter, maybe one leaves */
4990 super = imsm_thunderdome(&super_list, i);
4991 if (!super) {
4992 err = 1;
4993 goto error;
cdddbdbc
DW		 4994 }
4995
47ee5a45
DW		 4996 if (find_missing(super) != 0) {
4997 free_imsm(super);
a2b97981
DW		 4998 err = 2;
4999 goto error;
47ee5a45 5000 }
8e59f3d8
AK		 5001
5002 /* load migration record */
2f86fda3 5003 err = load_imsm_migr_rec(super);
4c965cc9
AK		 5004 if (err == -1) {
5005 /* migration is in progress,
5006 * but migr_rec cannot be loaded,
5007 */
8e59f3d8
AK		 5008 err = 4;
5009 goto error;
5010 }
e2f41b2c
AK		 5011
5012 /* Check migration compatibility */
089f9d79 5013 if (err == 0 && check_mpb_migr_compatibility(super) != 0) {
e7b84f9d 5014 pr_err("Unsupported migration detected");
e2f41b2c
AK		 5015 if (devname)
5016 fprintf(stderr, " on %s\n", devname);
5017 else
5018 fprintf(stderr, " (IMSM).\n");
5019
5020 err = 5;
5021 goto error;
5022 }
5023
a2b97981
DW		 5024 err = 0;
5025
5026 error:
5027 while (super_list) {
5028 struct intel_super *s = super_list;
5029
5030 super_list = super_list->next;
5031 free_imsm(s);
5032 }
9587c373 5033
a2b97981
DW		 5034 if (err)
5035 return err;
f7e7067b 5036
cdddbdbc 5037 *sbp = super;
9587c373 5038 if (fd >= 0)
4dd2df09 5039 strcpy(st->container_devnm, fd2devnm(fd));
9587c373 5040 else
4dd2df09 5041 st->container_devnm[0] = 0;
a2b97981 5042 if (err == 0 && st->ss == NULL) {
bf5a934a 5043 st->ss = &super_imsm;
cdddbdbc
DW		 5044 st->minor_version = 0;
5045 st->max_devs = IMSM_MAX_DEVICES;
5046 }
cdddbdbc
DW		 5047 return 0;
5048}
2b959fbf 5049
ec50f7b6
LM		 5050static int
5051get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list,
5052 int *max, int keep_fd)
5053{
5054 struct md_list *tmpdev;
5055 int err = 0;
5056 int i = 0;
9587c373 5057
ec50f7b6
LM		 5058 for (i = 0, tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
5059 if (tmpdev->used != 1)
5060 continue;
5061 if (tmpdev->container == 1) {
ca9de185 5062 int lmax = 0;
ec50f7b6
LM		 5063 int fd = dev_open(tmpdev->devname, O_RDONLY|O_EXCL);
5064 if (fd < 0) {
e7b84f9d 5065 pr_err("cannot open device %s: %s\n",
ec50f7b6
LM		 5066 tmpdev->devname, strerror(errno));
5067 err = 8;
5068 goto error;
5069 }
5070 err = get_sra_super_block(fd, super_list,
5071 tmpdev->devname, &lmax,
5072 keep_fd);
5073 i += lmax;
5074 close(fd);
5075 if (err) {
5076 err = 7;
5077 goto error;
5078 }
5079 } else {
5080 int major = major(tmpdev->st_rdev);
5081 int minor = minor(tmpdev->st_rdev);
5082 err = get_super_block(super_list,
4dd2df09 5083 NULL,
ec50f7b6
LM		 5084 tmpdev->devname,
5085 major, minor,
5086 keep_fd);
5087 i++;
5088 if (err) {
5089 err = 6;
5090 goto error;
5091 }
5092 }
5093 }
5094 error:
5095 *max = i;
5096 return err;
5097}
9587c373 5098
4dd2df09 5099static int get_super_block(struct intel_super **super_list, char *devnm, char *devname,
9587c373
LM		 5100 int major, int minor, int keep_fd)
5101{
594dc1b8 5102 struct intel_super *s;
9587c373
LM		 5103 char nm[32];
5104 int dfd = -1;
9587c373
LM		 5105 int err = 0;
5106 int retry;
5107
5108 s = alloc_super();
5109 if (!s) {
5110 err = 1;
5111 goto error;
5112 }
5113
5114 sprintf(nm, "%d:%d", major, minor);
5115 dfd = dev_open(nm, O_RDWR);
5116 if (dfd < 0) {
5117 err = 2;
5118 goto error;
5119 }
5120
fa7bb6f8 5121 get_dev_sector_size(dfd, NULL, &s->sector_size);
cb8f6859 5122 find_intel_hba_capability(dfd, s, devname);
9587c373
LM		 5123 err = load_and_parse_mpb(dfd, s, NULL, keep_fd);
5124
5125 /* retry the load if we might have raced against mdmon */
4dd2df09 5126 if (err == 3 && devnm && mdmon_running(devnm))
9587c373
LM		 5127 for (retry = 0; retry < 3; retry++) {
5128 usleep(3000);
5129 err = load_and_parse_mpb(dfd, s, NULL, keep_fd);
5130 if (err != 3)
5131 break;
5132 }
5133 error:
5134 if (!err) {
5135 s->next = *super_list;
5136 *super_list = s;
5137 } else {
5138 if (s)
8d67477f 5139 free_imsm(s);
36614e95 5140 if (dfd >= 0)
9587c373
LM		 5141 close(dfd);
5142 }
089f9d79 5143 if (dfd >= 0 && !keep_fd)
9587c373
LM		 5144 close(dfd);
5145 return err;
5146
5147}
5148
5149static int
5150get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd)
5151{
5152 struct mdinfo *sra;
4dd2df09 5153 char *devnm;
9587c373
LM		 5154 struct mdinfo *sd;
5155 int err = 0;
5156 int i = 0;
4dd2df09 5157 sra = sysfs_read(fd, NULL, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE);
9587c373
LM		 5158 if (!sra)
5159 return 1;
5160
5161 if (sra->array.major_version != -1 ||
5162 sra->array.minor_version != -2 ||
5163 strcmp(sra->text_version, "imsm") != 0) {
5164 err = 1;
5165 goto error;
5166 }
5167 /* load all mpbs */
4dd2df09 5168 devnm = fd2devnm(fd);
9587c373 5169 for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) {
4dd2df09 5170 if (get_super_block(super_list, devnm, devname,
9587c373
LM		 5171 sd->disk.major, sd->disk.minor, keep_fd) != 0) {
5172 err = 7;
5173 goto error;
5174 }
5175 }
5176 error:
5177 sysfs_free(sra);
5178 *max = i;
5179 return err;
5180}
5181
2b959fbf
N		 5182static int load_container_imsm(struct supertype *st, int fd, char *devname)
5183{
ec50f7b6 5184 return load_super_imsm_all(st, fd, &st->sb, devname, NULL, 1);
2b959fbf 5185}
cdddbdbc
DW		 5186
5187static int load_super_imsm(struct supertype *st, int fd, char *devname)
5188{
5189 struct intel_super *super;
5190 int rv;
8a3544f8 5191 int retry;
cdddbdbc 5192
357ac106 5193 if (test_partition(fd))
691c6ee1
N		 5194 /* IMSM not allowed on partitions */
5195 return 1;
5196
37424f13
DW		 5197 free_super_imsm(st);
5198
49133e57 5199 super = alloc_super();
fa7bb6f8 5200 get_dev_sector_size(fd, NULL, &super->sector_size);
8d67477f
TM		 5201 if (!super)
5202 return 1;
ea2bc72b
LM		 5203 /* Load hba and capabilities if they exist.
5204 * But do not preclude loading metadata in case capabilities or hba are
5205 * non-compliant and ignore_hw_compat is set.
5206 */
d424212e 5207 rv = find_intel_hba_capability(fd, super, devname);
f2f5c343 5208 /* no orom/efi or non-intel hba of the disk */
089f9d79 5209 if (rv != 0 && st->ignore_hw_compat == 0) {
f2f5c343 5210 if (devname)
e7b84f9d 5211 pr_err("No OROM/EFI properties for %s\n", devname);
f2f5c343
LM		 5212 free_imsm(super);
5213 return 2;
5214 }
a2b97981 5215 rv = load_and_parse_mpb(fd, super, devname, 0);
cdddbdbc 5216
8a3544f8
AP		 5217 /* retry the load if we might have raced against mdmon */
5218 if (rv == 3) {
f96b1302
AP		 5219 struct mdstat_ent *mdstat = NULL;
5220 char *name = fd2kname(fd);
5221
5222 if (name)
5223 mdstat = mdstat_by_component(name);
8a3544f8
AP		 5224
5225 if (mdstat && mdmon_running(mdstat->devnm) && getpid() != mdmon_pid(mdstat->devnm)) {
5226 for (retry = 0; retry < 3; retry++) {
5227 usleep(3000);
5228 rv = load_and_parse_mpb(fd, super, devname, 0);
5229 if (rv != 3)
5230 break;
5231 }
5232 }
5233
5234 free_mdstat(mdstat);
5235 }
5236
cdddbdbc
DW		 5237 if (rv) {
5238 if (devname)
7a862a02 5239 pr_err("Failed to load all information sections on %s\n", devname);
cdddbdbc
DW		 5240 free_imsm(super);
5241 return rv;
5242 }
5243
5244 st->sb = super;
5245 if (st->ss == NULL) {
5246 st->ss = &super_imsm;
5247 st->minor_version = 0;
5248 st->max_devs = IMSM_MAX_DEVICES;
5249 }
8e59f3d8
AK		 5250
5251 /* load migration record */
2f86fda3 5252 if (load_imsm_migr_rec(super) == 0) {
2e062e82
AK		 5253 /* Check for unsupported migration features */
5254 if (check_mpb_migr_compatibility(super) != 0) {
e7b84f9d 5255 pr_err("Unsupported migration detected");
2e062e82
AK		 5256 if (devname)
5257 fprintf(stderr, " on %s\n", devname);
5258 else
5259 fprintf(stderr, " (IMSM).\n");
5260 return 3;
5261 }
e2f41b2c
AK		 5262 }
5263
cdddbdbc
DW		 5264 return 0;
5265}
5266
ef6ffade
DW		 5267static __u16 info_to_blocks_per_strip(mdu_array_info_t *info)
5268{
5269 if (info->level == 1)
5270 return 128;
5271 return info->chunk_size >> 9;
5272}
5273
5551b113
CA		 5274static unsigned long long info_to_blocks_per_member(mdu_array_info_t *info,
5275 unsigned long long size)
fcfd9599 5276{
4025c288 5277 if (info->level == 1)
5551b113 5278 return size * 2;
4025c288 5279 else
5551b113 5280 return (size * 2) & ~(info_to_blocks_per_strip(info) - 1);
fcfd9599
DW		 5281}
5282
4d1313e9
DW		 5283static void imsm_update_version_info(struct intel_super *super)
5284{
5285 /* update the version and attributes */
5286 struct imsm_super *mpb = super->anchor;
5287 char *version;
5288 struct imsm_dev *dev;
5289 struct imsm_map *map;
5290 int i;
5291
5292 for (i = 0; i < mpb->num_raid_devs; i++) {
5293 dev = get_imsm_dev(super, i);
238c0a71 5294 map = get_imsm_map(dev, MAP_0);
4d1313e9
DW		 5295 if (__le32_to_cpu(dev->size_high) > 0)
5296 mpb->attributes |= MPB_ATTRIB_2TB;
5297
5298 /* FIXME detect when an array spans a port multiplier */
5299 #if 0
5300 mpb->attributes |= MPB_ATTRIB_PM;
5301 #endif
5302
5303 if (mpb->num_raid_devs > 1 ||
5304 mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) {
5305 version = MPB_VERSION_ATTRIBS;
5306 switch (get_imsm_raid_level(map)) {
5307 case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break;
5308 case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break;
5309 case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break;
5310 case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break;
5311 }
5312 } else {
5313 if (map->num_members >= 5)
5314 version = MPB_VERSION_5OR6_DISK_ARRAY;
5315 else if (dev->status == DEV_CLONE_N_GO)
5316 version = MPB_VERSION_CNG;
5317 else if (get_imsm_raid_level(map) == 5)
5318 version = MPB_VERSION_RAID5;
5319 else if (map->num_members >= 3)
5320 version = MPB_VERSION_3OR4_DISK_ARRAY;
5321 else if (get_imsm_raid_level(map) == 1)
5322 version = MPB_VERSION_RAID1;
5323 else
5324 version = MPB_VERSION_RAID0;
5325 }
5326 strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version);
5327 }
5328}
5329
aa534678
DW		 5330static int check_name(struct intel_super *super, char *name, int quiet)
5331{
5332 struct imsm_super *mpb = super->anchor;
5333 char *reason = NULL;
9bd99a90
RS		 5334 char *start = name;
5335 size_t len = strlen(name);
aa534678
DW		 5336 int i;
5337
9bd99a90
RS		 5338 if (len > 0) {
5339 while (isspace(start[len - 1]))
5340 start[--len] = 0;
5341 while (*start && isspace(*start))
5342 ++start, --len;
5343 memmove(name, start, len + 1);
5344 }
5345
5346 if (len > MAX_RAID_SERIAL_LEN)
aa534678 5347 reason = "must be 16 characters or less";
9bd99a90
RS		 5348 else if (len == 0)
5349 reason = "must be a non-empty string";
aa534678
DW		 5350
5351 for (i = 0; i < mpb->num_raid_devs; i++) {
5352 struct imsm_dev *dev = get_imsm_dev(super, i);
5353
5354 if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) {
5355 reason = "already exists";
5356 break;
5357 }
5358 }
5359
5360 if (reason && !quiet)
e7b84f9d 5361 pr_err("imsm volume name %s\n", reason);
aa534678
DW		 5362
5363 return !reason;
5364}
5365
8b353278 5366static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info,
5308f117 5367 struct shape *s, char *name,
83cd1e97
N		 5368 char *homehost, int *uuid,
5369 long long data_offset)
cdddbdbc 5370{
c2c087e6
DW		 5371 /* We are creating a volume inside a pre-existing container.
5372 * so st->sb is already set.
5373 */
5374 struct intel_super *super = st->sb;
f36a9ecd 5375 unsigned int sector_size = super->sector_size;
949c47a0 5376 struct imsm_super *mpb = super->anchor;
ba2de7ba 5377 struct intel_dev *dv;
c2c087e6
DW		 5378 struct imsm_dev *dev;
5379 struct imsm_vol *vol;
5380 struct imsm_map *map;
5381 int idx = mpb->num_raid_devs;
5382 int i;
760365f9 5383 int namelen;
c2c087e6 5384 unsigned long long array_blocks;
2c092cad 5385 size_t size_old, size_new;
5551b113 5386 unsigned long long num_data_stripes;
b53bfba6
TM		 5387 unsigned int data_disks;
5388 unsigned long long size_per_member;
cdddbdbc 5389
88c32bb1 5390 if (super->orom && mpb->num_raid_devs >= super->orom->vpa) {
7a862a02 5391 pr_err("This imsm-container already has the maximum of %d volumes\n", super->orom->vpa);
c2c087e6
DW		 5392 return 0;
5393 }
5394
2c092cad
DW		 5395 /* ensure the mpb is large enough for the new data */
5396 size_old = __le32_to_cpu(mpb->mpb_size);
5397 size_new = disks_to_mpb_size(info->nr_disks);
5398 if (size_new > size_old) {
5399 void *mpb_new;
f36a9ecd 5400 size_t size_round = ROUND_UP(size_new, sector_size);
2c092cad 5401
f36a9ecd 5402 if (posix_memalign(&mpb_new, sector_size, size_round) != 0) {
e7b84f9d 5403 pr_err("could not allocate new mpb\n");
2c092cad
DW		 5404 return 0;
5405 }
85337573
AO		 5406 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
5407 MIGR_REC_BUF_SECTORS*
5408 MAX_SECTOR_SIZE) != 0) {
1ade5cc1 5409 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8
AK		 5410 free(super->buf);
5411 free(super);
ea944c8f 5412 free(mpb_new);
8e59f3d8
AK		 5413 return 0;
5414 }
2c092cad
DW		 5415 memcpy(mpb_new, mpb, size_old);
5416 free(mpb);
5417 mpb = mpb_new;
949c47a0 5418 super->anchor = mpb_new;
2c092cad
DW		 5419 mpb->mpb_size = __cpu_to_le32(size_new);
5420 memset(mpb_new + size_old, 0, size_round - size_old);
bbab0940 5421 super->len = size_round;
2c092cad 5422 }
bf5a934a 5423 super->current_vol = idx;
3960e579
DW		 5424
5425 /* handle 'failed_disks' by either:
5426 * a) create dummy disk entries in the table if this the first
5427 * volume in the array. We add them here as this is the only
5428 * opportunity to add them. add_to_super_imsm_volume()
5429 * handles the non-failed disks and continues incrementing
5430 * mpb->num_disks.
5431 * b) validate that 'failed_disks' matches the current number
5432 * of missing disks if the container is populated
d23fe947 5433 */
3960e579 5434 if (super->current_vol == 0) {
d23fe947 5435 mpb->num_disks = 0;
3960e579
DW		 5436 for (i = 0; i < info->failed_disks; i++) {
5437 struct imsm_disk *disk;
5438
5439 mpb->num_disks++;
5440 disk = __get_imsm_disk(mpb, i);
5441 disk->status = CONFIGURED_DISK | FAILED_DISK;
5442 disk->scsi_id = __cpu_to_le32(~(__u32)0);
5443 snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN,
5b975129 5444 "missing:%d", (__u8)i);
3960e579
DW		 5445 }
5446 find_missing(super);
5447 } else {
5448 int missing = 0;
5449 struct dl *d;
5450
5451 for (d = super->missing; d; d = d->next)
5452 missing++;
5453 if (info->failed_disks > missing) {
e7b84f9d 5454 pr_err("unable to add 'missing' disk to container\n");
3960e579
DW		 5455 return 0;
5456 }
5457 }
5a038140 5458
aa534678
DW		 5459 if (!check_name(super, name, 0))
5460 return 0;
503975b9
N		 5461 dv = xmalloc(sizeof(*dv));
5462 dev = xcalloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1));
760365f9
JS		 5463 /*
5464 * Explicitly allow truncating to not confuse gcc's
5465 * -Werror=stringop-truncation
5466 */
5467 namelen = min((int) strlen(name), MAX_RAID_SERIAL_LEN);
5468 memcpy(dev->volume, name, namelen);
e03640bd 5469 array_blocks = calc_array_size(info->level, info->raid_disks,
03bcbc65 5470 info->layout, info->chunk_size,
b53bfba6
TM		 5471 s->size * BLOCKS_PER_KB);
5472 data_disks = get_data_disks(info->level, info->layout,
5473 info->raid_disks);
5474 array_blocks = round_size_to_mb(array_blocks, data_disks);
5475 size_per_member = array_blocks / data_disks;
979d38be 5476
fcc2c9da 5477 set_imsm_dev_size(dev, array_blocks);
1a2487c2 5478 dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING);
c2c087e6
DW		 5479 vol = &dev->vol;
5480 vol->migr_state = 0;
1484e727 5481 set_migr_type(dev, MIGR_INIT);
3960e579 5482 vol->dirty = !info->state;
f8f603f1 5483 vol->curr_migr_unit = 0;
238c0a71 5484 map = get_imsm_map(dev, MAP_0);
5551b113 5485 set_pba_of_lba0(map, super->create_offset);
ef6ffade 5486 map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info));
0556e1a2 5487 map->failed_disk_num = ~0;
bf4442ab 5488 if (info->level > 0)
fffaf1ff
N		 5489 map->map_state = (info->state ? IMSM_T_STATE_NORMAL
5490 : IMSM_T_STATE_UNINITIALIZED);
bf4442ab
AK		 5491 else
5492 map->map_state = info->failed_disks ? IMSM_T_STATE_FAILED :
5493 IMSM_T_STATE_NORMAL;
252d23c0 5494 map->ddf = 1;
ef6ffade
DW		 5495
5496 if (info->level == 1 && info->raid_disks > 2) {
38950822
AW		 5497 free(dev);
5498 free(dv);
7a862a02 5499 pr_err("imsm does not support more than 2 disksin a raid1 volume\n");
ef6ffade
DW		 5500 return 0;
5501 }
81062a36
DW		 5502
5503 map->raid_level = info->level;
4d1313e9 5504 if (info->level == 10) {
c2c087e6 5505 map->raid_level = 1;
4d1313e9 5506 map->num_domains = info->raid_disks / 2;
81062a36
DW		 5507 } else if (info->level == 1)
5508 map->num_domains = info->raid_disks;
5509 else
ff596308 5510 map->num_domains = 1;
81062a36 5511
5551b113 5512 /* info->size is only int so use the 'size' parameter instead */
b53bfba6 5513 num_data_stripes = size_per_member / info_to_blocks_per_strip(info);
5551b113
CA		 5514 num_data_stripes /= map->num_domains;
5515 set_num_data_stripes(map, num_data_stripes);
ef6ffade 5516
44490938
MD		 5517 size_per_member += NUM_BLOCKS_DIRTY_STRIPE_REGION;
5518 set_blocks_per_member(map, info_to_blocks_per_member(info,
5519 size_per_member /
5520 BLOCKS_PER_KB));
5521
c2c087e6
DW		 5522 map->num_members = info->raid_disks;
5523 for (i = 0; i < map->num_members; i++) {
5524 /* initialized in add_to_super */
4eb26970 5525 set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD);
c2c087e6 5526 }
949c47a0 5527 mpb->num_raid_devs++;
2a24dc1b
PB		 5528 mpb->num_raid_devs_created++;
5529 dev->my_vol_raid_dev_num = mpb->num_raid_devs_created;
ba2de7ba 5530
b7580566 5531 if (s->consistency_policy <= CONSISTENCY_POLICY_RESYNC) {
c2462068 5532 dev->rwh_policy = RWH_MULTIPLE_OFF;
2432ce9b 5533 } else if (s->consistency_policy == CONSISTENCY_POLICY_PPL) {
c2462068 5534 dev->rwh_policy = RWH_MULTIPLE_DISTRIBUTED;
2432ce9b
AP		 5535 } else {
5536 free(dev);
5537 free(dv);
5538 pr_err("imsm does not support consistency policy %s\n",
5539 map_num(consistency_policies, s->consistency_policy));
5540 return 0;
5541 }
5542
ba2de7ba
DW		 5543 dv->dev = dev;
5544 dv->index = super->current_vol;
5545 dv->next = super->devlist;
5546 super->devlist = dv;
c2c087e6 5547
4d1313e9
DW		 5548 imsm_update_version_info(super);
5549
c2c087e6 5550 return 1;
cdddbdbc
DW		 5551}
5552
bf5a934a 5553static int init_super_imsm(struct supertype *st, mdu_array_info_t *info,
5308f117 5554 struct shape *s, char *name,
83cd1e97
N		 5555 char *homehost, int *uuid,
5556 unsigned long long data_offset)
bf5a934a
DW		 5557{
5558 /* This is primarily called by Create when creating a new array.
5559 * We will then get add_to_super called for each component, and then
5560 * write_init_super called to write it out to each device.
5561 * For IMSM, Create can create on fresh devices or on a pre-existing
5562 * array.
5563 * To create on a pre-existing array a different method will be called.
5564 * This one is just for fresh drives.
5565 */
5566 struct intel_super *super;
5567 struct imsm_super *mpb;
5568 size_t mpb_size;
4d1313e9 5569 char *version;
bf5a934a 5570
83cd1e97 5571 if (data_offset != INVALID_SECTORS) {
ed503f89 5572 pr_err("data-offset not supported by imsm\n");
83cd1e97
N		 5573 return 0;
5574 }
5575
bf5a934a 5576 if (st->sb)
5308f117 5577 return init_super_imsm_volume(st, info, s, name, homehost, uuid,
83cd1e97 5578 data_offset);
e683ca88
DW		 5579
5580 if (info)
5581 mpb_size = disks_to_mpb_size(info->nr_disks);
5582 else
f36a9ecd 5583 mpb_size = MAX_SECTOR_SIZE;
bf5a934a 5584
49133e57 5585 super = alloc_super();
f36a9ecd
PB		 5586 if (super &&
5587 posix_memalign(&super->buf, MAX_SECTOR_SIZE, mpb_size) != 0) {
8d67477f 5588 free_imsm(super);
e683ca88
DW		 5589 super = NULL;
5590 }
5591 if (!super) {
1ade5cc1 5592 pr_err("could not allocate superblock\n");
bf5a934a
DW		 5593 return 0;
5594 }
de44e46f
PB		 5595 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
5596 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) {
1ade5cc1 5597 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8 5598 free(super->buf);
8d67477f 5599 free_imsm(super);
8e59f3d8
AK		 5600 return 0;
5601 }
e683ca88 5602 memset(super->buf, 0, mpb_size);
ef649044 5603 mpb = super->buf;
e683ca88
DW		 5604 mpb->mpb_size = __cpu_to_le32(mpb_size);
5605 st->sb = super;
5606
5607 if (info == NULL) {
5608 /* zeroing superblock */
5609 return 0;
5610 }
bf5a934a 5611
4d1313e9
DW		 5612 mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
5613
5614 version = (char *) mpb->sig;
5615 strcpy(version, MPB_SIGNATURE);
5616 version += strlen(MPB_SIGNATURE);
5617 strcpy(version, MPB_VERSION_RAID0);
bf5a934a 5618
bf5a934a
DW		 5619 return 1;
5620}
5621
f2cc4f7d
AO		 5622static int drive_validate_sector_size(struct intel_super *super, struct dl *dl)
5623{
5624 unsigned int member_sector_size;
5625
5626 if (dl->fd < 0) {
5627 pr_err("Invalid file descriptor for %s\n", dl->devname);
5628 return 0;
5629 }
5630
5631 if (!get_dev_sector_size(dl->fd, dl->devname, &member_sector_size))
5632 return 0;
5633 if (member_sector_size != super->sector_size)
5634 return 0;
5635 return 1;
5636}
5637
f20c3968 5638static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk,
bf5a934a
DW		 5639 int fd, char *devname)
5640{
5641 struct intel_super *super = st->sb;
d23fe947 5642 struct imsm_super *mpb = super->anchor;
3960e579 5643 struct imsm_disk *_disk;
bf5a934a
DW		 5644 struct imsm_dev *dev;
5645 struct imsm_map *map;
3960e579 5646 struct dl *dl, *df;
4eb26970 5647 int slot;
bf5a934a 5648
949c47a0 5649 dev = get_imsm_dev(super, super->current_vol);
238c0a71 5650 map = get_imsm_map(dev, MAP_0);
bf5a934a 5651
208933a7 5652 if (! (dk->state & (1<<MD_DISK_SYNC))) {
e7b84f9d 5653 pr_err("%s: Cannot add spare devices to IMSM volume\n",
208933a7
N		 5654 devname);
5655 return 1;
5656 }
5657
efb30e7f
DW		 5658 if (fd == -1) {
5659 /* we're doing autolayout so grab the pre-marked (in
5660 * validate_geometry) raid_disk
5661 */
5662 for (dl = super->disks; dl; dl = dl->next)
5663 if (dl->raiddisk == dk->raid_disk)
5664 break;
5665 } else {
5666 for (dl = super->disks; dl ; dl = dl->next)
5667 if (dl->major == dk->major &&
5668 dl->minor == dk->minor)
5669 break;
5670 }
d23fe947 5671
208933a7 5672 if (!dl) {
e7b84f9d 5673 pr_err("%s is not a member of the same container\n", devname);
f20c3968 5674 return 1;
208933a7 5675 }
bf5a934a 5676
59632db9
MZ		 5677 if (mpb->num_disks == 0)
5678 if (!get_dev_sector_size(dl->fd, dl->devname,
5679 &super->sector_size))
5680 return 1;
5681
f2cc4f7d
AO		 5682 if (!drive_validate_sector_size(super, dl)) {
5683 pr_err("Combining drives of different sector size in one volume is not allowed\n");
5684 return 1;
5685 }
5686
d23fe947
DW		 5687 /* add a pristine spare to the metadata */
5688 if (dl->index < 0) {
5689 dl->index = super->anchor->num_disks;
5690 super->anchor->num_disks++;
5691 }
4eb26970
DW		 5692 /* Check the device has not already been added */
5693 slot = get_imsm_disk_slot(map, dl->index);
5694 if (slot >= 0 &&
238c0a71 5695 (get_imsm_ord_tbl_ent(dev, slot, MAP_X) & IMSM_ORD_REBUILD) == 0) {
e7b84f9d 5696 pr_err("%s has been included in this array twice\n",
4eb26970
DW		 5697 devname);
5698 return 1;
5699 }
656b6b5a 5700 set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index);
ee5aad5a 5701 dl->disk.status = CONFIGURED_DISK;
d23fe947 5702
3960e579
DW		 5703 /* update size of 'missing' disks to be at least as large as the
5704 * largest acitve member (we only have dummy missing disks when
5705 * creating the first volume)
5706 */
5707 if (super->current_vol == 0) {
5708 for (df = super->missing; df; df = df->next) {
5551b113
CA		 5709 if (total_blocks(&dl->disk) > total_blocks(&df->disk))
5710 set_total_blocks(&df->disk, total_blocks(&dl->disk));
3960e579
DW		 5711 _disk = __get_imsm_disk(mpb, df->index);
5712 *_disk = df->disk;
5713 }
5714 }
5715
5716 /* refresh unset/failed slots to point to valid 'missing' entries */
5717 for (df = super->missing; df; df = df->next)
5718 for (slot = 0; slot < mpb->num_disks; slot++) {
238c0a71 5719 __u32 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
3960e579
DW		 5720
5721 if ((ord & IMSM_ORD_REBUILD) == 0)
5722 continue;
5723 set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD);
1ace8403 5724 if (is_gen_migration(dev)) {
238c0a71
AK		 5725 struct imsm_map *map2 = get_imsm_map(dev,
5726 MAP_1);
0a108d63 5727 int slot2 = get_imsm_disk_slot(map2, df->index);
089f9d79 5728 if (slot2 < map2->num_members && slot2 >= 0) {
1ace8403 5729 __u32 ord2 = get_imsm_ord_tbl_ent(dev,
238c0a71
AK		 5730 slot2,
5731 MAP_1);
1ace8403
AK		 5732 if ((unsigned)df->index ==
5733 ord_to_idx(ord2))
5734 set_imsm_ord_tbl_ent(map2,
0a108d63 5735 slot2,
1ace8403
AK		 5736 df->index |
5737 IMSM_ORD_REBUILD);
5738 }
5739 }
3960e579
DW		 5740 dprintf("set slot:%d to missing disk:%d\n", slot, df->index);
5741 break;
5742 }
5743
d23fe947
DW		 5744 /* if we are creating the first raid device update the family number */
5745 if (super->current_vol == 0) {
5746 __u32 sum;
5747 struct imsm_dev *_dev = __get_imsm_dev(mpb, 0);
d23fe947 5748
3960e579 5749 _disk = __get_imsm_disk(mpb, dl->index);
791b666a 5750 if (!_dev || !_disk) {
e7b84f9d 5751 pr_err("BUG mpb setup error\n");
791b666a
AW		 5752 return 1;
5753 }
d23fe947
DW		 5754 *_dev = *dev;
5755 *_disk = dl->disk;
148acb7b
DW		 5756 sum = random32();
5757 sum += __gen_imsm_checksum(mpb);
d23fe947 5758 mpb->family_num = __cpu_to_le32(sum);
148acb7b 5759 mpb->orig_family_num = mpb->family_num;
e48aed3c 5760 mpb->creation_time = __cpu_to_le64((__u64)time(NULL));
d23fe947 5761 }
ca0748fa 5762 super->current_disk = dl;
f20c3968 5763 return 0;
bf5a934a
DW		 5764}
5765
a8619d23
AK		 5766/* mark_spare()
5767 * Function marks disk as spare and restores disk serial
5768 * in case it was previously marked as failed by takeover operation
5769 * reruns:
5770 * -1 : critical error
5771 * 0 : disk is marked as spare but serial is not set
5772 * 1 : success
5773 */
5774int mark_spare(struct dl *disk)
5775{
5776 __u8 serial[MAX_RAID_SERIAL_LEN];
5777 int ret_val = -1;
5778
5779 if (!disk)
5780 return ret_val;
5781
5782 ret_val = 0;
6da53c0e 5783 if (!imsm_read_serial(disk->fd, NULL, serial, MAX_RAID_SERIAL_LEN)) {
a8619d23
AK		 5784 /* Restore disk serial number, because takeover marks disk
5785 * as failed and adds to serial ':0' before it becomes
5786 * a spare disk.
5787 */
5788 serialcpy(disk->serial, serial);
5789 serialcpy(disk->disk.serial, serial);
5790 ret_val = 1;
5791 }
5792 disk->disk.status = SPARE_DISK;
5793 disk->index = -1;
5794
5795 return ret_val;
5796}
88654014 5797
12724c01
TM		 5798
5799static int write_super_imsm_spare(struct intel_super *super, struct dl *d);
5800
f20c3968 5801static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk,
72ca9bcf
N		 5802 int fd, char *devname,
5803 unsigned long long data_offset)
cdddbdbc 5804{
c2c087e6 5805 struct intel_super *super = st->sb;
c2c087e6
DW		 5806 struct dl *dd;
5807 unsigned long long size;
fa7bb6f8 5808 unsigned int member_sector_size;
f2f27e63 5809 __u32 id;
c2c087e6
DW		 5810 int rv;
5811 struct stat stb;
5812
88654014
LM		 5813 /* If we are on an RAID enabled platform check that the disk is
5814 * attached to the raid controller.
5815 * We do not need to test disks attachment for container based additions,
5816 * they shall be already tested when container was created/assembled.
88c32bb1 5817 */
d424212e 5818 rv = find_intel_hba_capability(fd, super, devname);
f2f5c343 5819 /* no orom/efi or non-intel hba of the disk */
f0f5a016
LM		 5820 if (rv != 0) {
5821 dprintf("capability: %p fd: %d ret: %d\n",
5822 super->orom, fd, rv);
5823 return 1;
88c32bb1
DW		 5824 }
5825
f20c3968
DW		 5826 if (super->current_vol >= 0)
5827 return add_to_super_imsm_volume(st, dk, fd, devname);
bf5a934a 5828
c2c087e6 5829 fstat(fd, &stb);
503975b9 5830 dd = xcalloc(sizeof(*dd), 1);
c2c087e6
DW		 5831 dd->major = major(stb.st_rdev);
5832 dd->minor = minor(stb.st_rdev);
503975b9 5833 dd->devname = devname ? xstrdup(devname) : NULL;
c2c087e6 5834 dd->fd = fd;
689c9bf3 5835 dd->e = NULL;
1a64be56 5836 dd->action = DISK_ADD;
6da53c0e 5837 rv = imsm_read_serial(fd, devname, dd->serial, MAX_RAID_SERIAL_LEN);
32ba9157 5838 if (rv) {
e7b84f9d 5839 pr_err("failed to retrieve scsi serial, aborting\n");
20bee0f8
PB		 5840 if (dd->devname)
5841 free(dd->devname);
949c47a0 5842 free(dd);
0030e8d6 5843 abort();
c2c087e6 5844 }
20bee0f8
PB		 5845 if (super->hba && ((super->hba->type == SYS_DEV_NVME) ||
5846 (super->hba->type == SYS_DEV_VMD))) {
5847 int i;
5848 char *devpath = diskfd_to_devpath(fd);
5849 char controller_path[PATH_MAX];
5850
5851 if (!devpath) {
5852 pr_err("failed to get devpath, aborting\n");
5853 if (dd->devname)
5854 free(dd->devname);
5855 free(dd);
5856 return 1;
5857 }
5858
5859 snprintf(controller_path, PATH_MAX-1, "%s/device", devpath);
5860 free(devpath);
5861
a8f3cfd5
MT		 5862 if (!imsm_is_nvme_supported(dd->fd, 1)) {
5863 if (dd->devname)
5864 free(dd->devname);
5865 free(dd);
5866 return 1;
5867 }
5868
20bee0f8
PB		 5869 if (devpath_to_vendor(controller_path) == 0x8086) {
5870 /*
5871 * If Intel's NVMe drive has serial ended with
5872 * "-A","-B","-1" or "-2" it means that this is "x8"
5873 * device (double drive on single PCIe card).
5874 * User should be warned about potential data loss.
5875 */
5876 for (i = MAX_RAID_SERIAL_LEN-1; i > 0; i--) {
5877 /* Skip empty character at the end */
5878 if (dd->serial[i] == 0)
5879 continue;
5880
5881 if (((dd->serial[i] == 'A') ||
5882 (dd->serial[i] == 'B') ||
5883 (dd->serial[i] == '1') ||
5884 (dd->serial[i] == '2')) &&
5885 (dd->serial[i-1] == '-'))
5886 pr_err("\tThe action you are about to take may put your data at risk.\n"
5887 "\tPlease note that x8 devices may consist of two separate x4 devices "
5888 "located on a single PCIe port.\n"
5889 "\tRAID 0 is the only supported configuration for this type of x8 device.\n");
5890 break;
5891 }
32716c51
PB		 5892 } else if (super->hba->type == SYS_DEV_VMD && super->orom &&
5893 !imsm_orom_has_tpv_support(super->orom)) {
5894 pr_err("\tPlatform configuration does not support non-Intel NVMe drives.\n"
8b751247 5895 "\tPlease refer to Intel(R) RSTe/VROC user guide.\n");
32716c51
PB		 5896 free(dd->devname);
5897 free(dd);
5898 return 1;
20bee0f8
PB		 5899 }
5900 }
c2c087e6 5901
c2c087e6 5902 get_dev_size(fd, NULL, &size);
fa7bb6f8
PB		 5903 get_dev_sector_size(fd, NULL, &member_sector_size);
5904
5905 if (super->sector_size == 0) {
5906 /* this a first device, so sector_size is not set yet */
5907 super->sector_size = member_sector_size;
fa7bb6f8
PB		 5908 }
5909
71e5411e 5910 /* clear migr_rec when adding disk to container */
85337573
AO		 5911 memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
5912 if (lseek64(fd, size - MIGR_REC_SECTOR_POSITION*member_sector_size,
de44e46f 5913 SEEK_SET) >= 0) {
466070ad 5914 if ((unsigned int)write(fd, super->migr_rec_buf,
85337573
AO		 5915 MIGR_REC_BUF_SECTORS*member_sector_size) !=
5916 MIGR_REC_BUF_SECTORS*member_sector_size)
71e5411e
PB		 5917 perror("Write migr_rec failed");
5918 }
5919
c2c087e6 5920 size /= 512;
1f24f035 5921 serialcpy(dd->disk.serial, dd->serial);
5551b113
CA		 5922 set_total_blocks(&dd->disk, size);
5923 if (__le32_to_cpu(dd->disk.total_blocks_hi) > 0) {
5924 struct imsm_super *mpb = super->anchor;
5925 mpb->attributes |= MPB_ATTRIB_2TB_DISK;
5926 }
a8619d23 5927 mark_spare(dd);
c2c087e6 5928 if (sysfs_disk_to_scsi_id(fd, &id) == 0)
b9f594fe 5929 dd->disk.scsi_id = __cpu_to_le32(id);
c2c087e6 5930 else
b9f594fe 5931 dd->disk.scsi_id = __cpu_to_le32(0);
43dad3d6
DW		 5932
5933 if (st->update_tail) {
1a64be56
LM		 5934 dd->next = super->disk_mgmt_list;
5935 super->disk_mgmt_list = dd;
43dad3d6 5936 } else {
12724c01
TM		 5937 /* this is called outside of mdmon
5938 * write initial spare metadata
5939 * mdmon will overwrite it.
5940 */
43dad3d6
DW		 5941 dd->next = super->disks;
5942 super->disks = dd;
12724c01 5943 write_super_imsm_spare(super, dd);
43dad3d6 5944 }
f20c3968
DW		 5945
5946 return 0;
cdddbdbc
DW		 5947}
5948
1a64be56
LM		 5949static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk)
5950{
5951 struct intel_super *super = st->sb;
5952 struct dl *dd;
5953
5954 /* remove from super works only in mdmon - for communication
5955 * manager - monitor. Check if communication memory buffer
5956 * is prepared.
5957 */
5958 if (!st->update_tail) {
1ade5cc1 5959 pr_err("shall be used in mdmon context only\n");
1a64be56
LM		 5960 return 1;
5961 }
503975b9 5962 dd = xcalloc(1, sizeof(*dd));
1a64be56
LM		 5963 dd->major = dk->major;
5964 dd->minor = dk->minor;
1a64be56 5965 dd->fd = -1;
a8619d23 5966 mark_spare(dd);
1a64be56
LM		 5967 dd->action = DISK_REMOVE;
5968
5969 dd->next = super->disk_mgmt_list;
5970 super->disk_mgmt_list = dd;
5971
1a64be56
LM		 5972 return 0;
5973}
5974
f796af5d
DW		 5975static int store_imsm_mpb(int fd, struct imsm_super *mpb);
5976
5977static union {
f36a9ecd 5978 char buf[MAX_SECTOR_SIZE];
f796af5d 5979 struct imsm_super anchor;
f36a9ecd 5980} spare_record __attribute__ ((aligned(MAX_SECTOR_SIZE)));
c2c087e6 5981
12724c01
TM		 5982
5983static int write_super_imsm_spare(struct intel_super *super, struct dl *d)
d23fe947 5984{
d23fe947 5985 struct imsm_super *mpb = super->anchor;
f796af5d 5986 struct imsm_super *spare = &spare_record.anchor;
d23fe947 5987 __u32 sum;
12724c01
TM		 5988
5989 if (d->index != -1)
5990 return 1;
d23fe947 5991
68641cdb
JS		 5992 spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super));
5993 spare->generation_num = __cpu_to_le32(1UL);
f796af5d 5994 spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
68641cdb
JS		 5995 spare->num_disks = 1;
5996 spare->num_raid_devs = 0;
5997 spare->cache_size = mpb->cache_size;
5998 spare->pwr_cycle_count = __cpu_to_le32(1);
f796af5d
DW		 5999
6000 snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH,
6001 MPB_SIGNATURE MPB_VERSION_RAID0);
d23fe947 6002
12724c01
TM		 6003 spare->disk[0] = d->disk;
6004 if (__le32_to_cpu(d->disk.total_blocks_hi) > 0)
6005 spare->attributes |= MPB_ATTRIB_2TB_DISK;
6006
6007 if (super->sector_size == 4096)
6008 convert_to_4k_imsm_disk(&spare->disk[0]);
d23fe947 6009
12724c01
TM		 6010 sum = __gen_imsm_checksum(spare);
6011 spare->family_num = __cpu_to_le32(sum);
6012 spare->orig_family_num = 0;
6013 sum = __gen_imsm_checksum(spare);
6014 spare->check_sum = __cpu_to_le32(sum);
027c374f 6015
12724c01
TM		 6016 if (store_imsm_mpb(d->fd, spare)) {
6017 pr_err("failed for device %d:%d %s\n",
6018 d->major, d->minor, strerror(errno));
6019 return 1;
6020 }
6021
6022 return 0;
6023}
6024/* spare records have their own family number and do not have any defined raid
6025 * devices
6026 */
6027static int write_super_imsm_spares(struct intel_super *super, int doclose)
6028{
6029 struct dl *d;
f36a9ecd 6030
12724c01
TM		 6031 for (d = super->disks; d; d = d->next) {
6032 if (d->index != -1)
6033 continue;
d23fe947 6034
12724c01 6035 if (write_super_imsm_spare(super, d))
e74255d9 6036 return 1;
12724c01 6037
d23fe947
DW		 6038 if (doclose) {
6039 close(d->fd);
6040 d->fd = -1;
6041 }
6042 }
6043
e74255d9 6044 return 0;
d23fe947
DW		 6045}
6046
36988a3d 6047static int write_super_imsm(struct supertype *st, int doclose)
cdddbdbc 6048{
36988a3d 6049 struct intel_super *super = st->sb;
f36a9ecd 6050 unsigned int sector_size = super->sector_size;
949c47a0 6051 struct imsm_super *mpb = super->anchor;
c2c087e6
DW		 6052 struct dl *d;
6053 __u32 generation;
6054 __u32 sum;
d23fe947 6055 int spares = 0;
949c47a0 6056 int i;
a48ac0a8 6057 __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk);
36988a3d 6058 int num_disks = 0;
146c6260 6059 int clear_migration_record = 1;
bbab0940 6060 __u32 bbm_log_size;
cdddbdbc 6061
c2c087e6
DW		 6062 /* 'generation' is incremented everytime the metadata is written */
6063 generation = __le32_to_cpu(mpb->generation_num);
6064 generation++;
6065 mpb->generation_num = __cpu_to_le32(generation);
6066
148acb7b
DW		 6067 /* fix up cases where previous mdadm releases failed to set
6068 * orig_family_num
6069 */
6070 if (mpb->orig_family_num == 0)
6071 mpb->orig_family_num = mpb->family_num;
6072
d23fe947 6073 for (d = super->disks; d; d = d->next) {
8796fdc4 6074 if (d->index == -1)
d23fe947 6075 spares++;
36988a3d 6076 else {
d23fe947 6077 mpb->disk[d->index] = d->disk;
36988a3d
AK		 6078 num_disks++;
6079 }
d23fe947 6080 }
36988a3d 6081 for (d = super->missing; d; d = d->next) {
47ee5a45 6082 mpb->disk[d->index] = d->disk;
36988a3d
AK		 6083 num_disks++;
6084 }
6085 mpb->num_disks = num_disks;
6086 mpb_size += sizeof(struct imsm_disk) * mpb->num_disks;
b9f594fe 6087
949c47a0
DW		 6088 for (i = 0; i < mpb->num_raid_devs; i++) {
6089 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
36988a3d
AK		 6090 struct imsm_dev *dev2 = get_imsm_dev(super, i);
6091 if (dev && dev2) {
6092 imsm_copy_dev(dev, dev2);
6093 mpb_size += sizeof_imsm_dev(dev, 0);
6094 }
146c6260
AK		 6095 if (is_gen_migration(dev2))
6096 clear_migration_record = 0;
949c47a0 6097 }
bbab0940
TM		 6098
6099 bbm_log_size = get_imsm_bbm_log_size(super->bbm_log);
6100
6101 if (bbm_log_size) {
6102 memcpy((void *)mpb + mpb_size, super->bbm_log, bbm_log_size);
6103 mpb->attributes |= MPB_ATTRIB_BBM;
6104 } else
6105 mpb->attributes &= ~MPB_ATTRIB_BBM;
6106
6107 super->anchor->bbm_log_size = __cpu_to_le32(bbm_log_size);
6108 mpb_size += bbm_log_size;
a48ac0a8 6109 mpb->mpb_size = __cpu_to_le32(mpb_size);
949c47a0 6110
bbab0940
TM		 6111#ifdef DEBUG
6112 assert(super->len == 0 || mpb_size <= super->len);
6113#endif
6114
c2c087e6 6115 /* recalculate checksum */
949c47a0 6116 sum = __gen_imsm_checksum(mpb);
c2c087e6
DW		 6117 mpb->check_sum = __cpu_to_le32(sum);
6118
51d83f5d
AK		 6119 if (super->clean_migration_record_by_mdmon) {
6120 clear_migration_record = 1;
6121 super->clean_migration_record_by_mdmon = 0;
6122 }
146c6260 6123 if (clear_migration_record)
de44e46f 6124 memset(super->migr_rec_buf, 0,
85337573 6125 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
146c6260 6126
f36a9ecd
PB		 6127 if (sector_size == 4096)
6128 convert_to_4k(super);
6129
d23fe947 6130 /* write the mpb for disks that compose raid devices */
c2c087e6 6131 for (d = super->disks; d ; d = d->next) {
86c54047 6132 if (d->index < 0 || is_failed(&d->disk))
d23fe947 6133 continue;
30602f53 6134
146c6260
AK		 6135 if (clear_migration_record) {
6136 unsigned long long dsize;
6137
6138 get_dev_size(d->fd, NULL, &dsize);
de44e46f
PB		 6139 if (lseek64(d->fd, dsize - sector_size,
6140 SEEK_SET) >= 0) {
466070ad
PB		 6141 if ((unsigned int)write(d->fd,
6142 super->migr_rec_buf,
de44e46f
PB		 6143 MIGR_REC_BUF_SECTORS*sector_size) !=
6144 MIGR_REC_BUF_SECTORS*sector_size)
9e2d750d 6145 perror("Write migr_rec failed");
146c6260
AK		 6146 }
6147 }
51d83f5d
AK		 6148
6149 if (store_imsm_mpb(d->fd, mpb))
6150 fprintf(stderr,
1ade5cc1
N		 6151 "failed for device %d:%d (fd: %d)%s\n",
6152 d->major, d->minor,
51d83f5d
AK		 6153 d->fd, strerror(errno));
6154
c2c087e6
DW		 6155 if (doclose) {
6156 close(d->fd);
6157 d->fd = -1;
6158 }
6159 }
6160
d23fe947
DW		 6161 if (spares)
6162 return write_super_imsm_spares(super, doclose);
6163
e74255d9 6164 return 0;
c2c087e6
DW		 6165}
6166
9b1fb677 6167static int create_array(struct supertype *st, int dev_idx)
43dad3d6
DW		 6168{
6169 size_t len;
6170 struct imsm_update_create_array *u;
6171 struct intel_super *super = st->sb;
9b1fb677 6172 struct imsm_dev *dev = get_imsm_dev(super, dev_idx);
238c0a71 6173 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW		 6174 struct disk_info *inf;
6175 struct imsm_disk *disk;
6176 int i;
43dad3d6 6177
54c2c1ea
DW		 6178 len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) +
6179 sizeof(*inf) * map->num_members;
503975b9 6180 u = xmalloc(len);
43dad3d6 6181 u->type = update_create_array;
9b1fb677 6182 u->dev_idx = dev_idx;
43dad3d6 6183 imsm_copy_dev(&u->dev, dev);
54c2c1ea
DW		 6184 inf = get_disk_info(u);
6185 for (i = 0; i < map->num_members; i++) {
238c0a71 6186 int idx = get_imsm_disk_idx(dev, i, MAP_X);
9b1fb677 6187
54c2c1ea 6188 disk = get_imsm_disk(super, idx);
1ca5c8e0
N		 6189 if (!disk)
6190 disk = get_imsm_missing(super, idx);
54c2c1ea
DW		 6191 serialcpy(inf[i].serial, disk->serial);
6192 }
43dad3d6
DW		 6193 append_metadata_update(st, u, len);
6194
6195 return 0;
6196}
6197
1a64be56 6198static int mgmt_disk(struct supertype *st)
43dad3d6
DW		 6199{
6200 struct intel_super *super = st->sb;
6201 size_t len;
1a64be56 6202 struct imsm_update_add_remove_disk *u;
43dad3d6 6203
1a64be56 6204 if (!super->disk_mgmt_list)
43dad3d6
DW		 6205 return 0;
6206
6207 len = sizeof(*u);
503975b9 6208 u = xmalloc(len);
1a64be56 6209 u->type = update_add_remove_disk;
43dad3d6
DW		 6210 append_metadata_update(st, u, len);
6211
6212 return 0;
6213}
2432ce9b
AP		 6214
6215__u32 crc32c_le(__u32 crc, unsigned char const *p, size_t len);
6216
e397cefe
AP		 6217static int write_ppl_header(unsigned long long ppl_sector, int fd, void *buf)
6218{
6219 struct ppl_header *ppl_hdr = buf;
6220 int ret;
6221
6222 ppl_hdr->checksum = __cpu_to_le32(~crc32c_le(~0, buf, PPL_HEADER_SIZE));
6223
6224 if (lseek64(fd, ppl_sector * 512, SEEK_SET) < 0) {
6225 ret = -errno;
6226 perror("Failed to seek to PPL header location");
6227 return ret;
6228 }
6229
6230 if (write(fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) {
6231 ret = -errno;
6232 perror("Write PPL header failed");
6233 return ret;
6234 }
6235
6236 fsync(fd);
6237
6238 return 0;
6239}
6240
2432ce9b
AP		 6241static int write_init_ppl_imsm(struct supertype *st, struct mdinfo *info, int fd)
6242{
6243 struct intel_super *super = st->sb;
6244 void *buf;
6245 struct ppl_header *ppl_hdr;
6246 int ret;
6247
b2514242
PB		 6248 /* first clear entire ppl space */
6249 ret = zero_disk_range(fd, info->ppl_sector, info->ppl_size);
6250 if (ret)
6251 return ret;
6252
6253 ret = posix_memalign(&buf, MAX_SECTOR_SIZE, PPL_HEADER_SIZE);
2432ce9b
AP		 6254 if (ret) {
6255 pr_err("Failed to allocate PPL header buffer\n");
e397cefe 6256 return -ret;
2432ce9b
AP		 6257 }
6258
6259 memset(buf, 0, PPL_HEADER_SIZE);
6260 ppl_hdr = buf;
6261 memset(ppl_hdr->reserved, 0xff, PPL_HDR_RESERVED);
6262 ppl_hdr->signature = __cpu_to_le32(super->anchor->orig_family_num);
b23d0750
AP		 6263
6264 if (info->mismatch_cnt) {
6265 /*
6266 * We are overwriting an invalid ppl. Make one entry with wrong
6267 * checksum to prevent the kernel from skipping resync.
6268 */
6269 ppl_hdr->entries_count = __cpu_to_le32(1);
6270 ppl_hdr->entries[0].checksum = ~0;
6271 }
6272
e397cefe 6273 ret = write_ppl_header(info->ppl_sector, fd, buf);
2432ce9b
AP		 6274
6275 free(buf);
6276 return ret;
6277}
6278
e397cefe
AP		 6279static int is_rebuilding(struct imsm_dev *dev);
6280
2432ce9b
AP		 6281static int validate_ppl_imsm(struct supertype *st, struct mdinfo *info,
6282 struct mdinfo *disk)
6283{
6284 struct intel_super *super = st->sb;
6285 struct dl *d;
e397cefe 6286 void *buf_orig, *buf, *buf_prev = NULL;
2432ce9b 6287 int ret = 0;
e397cefe 6288 struct ppl_header *ppl_hdr = NULL;
2432ce9b
AP		 6289 __u32 crc;
6290 struct imsm_dev *dev;
2432ce9b 6291 __u32 idx;
44b6b876
PB		 6292 unsigned int i;
6293 unsigned long long ppl_offset = 0;
6294 unsigned long long prev_gen_num = 0;
2432ce9b
AP		 6295
6296 if (disk->disk.raid_disk < 0)
6297 return 0;
6298
2432ce9b 6299 dev = get_imsm_dev(super, info->container_member);
2fc0fc63 6300 idx = get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_0);
2432ce9b
AP		 6301 d = get_imsm_dl_disk(super, idx);
6302
6303 if (!d || d->index < 0 || is_failed(&d->disk))
e397cefe
AP		 6304 return 0;
6305
6306 if (posix_memalign(&buf_orig, MAX_SECTOR_SIZE, PPL_HEADER_SIZE * 2)) {
6307 pr_err("Failed to allocate PPL header buffer\n");
6308 return -1;
6309 }
6310 buf = buf_orig;
2432ce9b 6311
44b6b876
PB		 6312 ret = 1;
6313 while (ppl_offset < MULTIPLE_PPL_AREA_SIZE_IMSM) {
e397cefe
AP		 6314 void *tmp;
6315
44b6b876 6316 dprintf("Checking potential PPL at offset: %llu\n", ppl_offset);
2432ce9b 6317
44b6b876
PB		 6318 if (lseek64(d->fd, info->ppl_sector * 512 + ppl_offset,
6319 SEEK_SET) < 0) {
6320 perror("Failed to seek to PPL header location");
6321 ret = -1;
e397cefe 6322 break;
44b6b876 6323 }
2432ce9b 6324
44b6b876
PB		 6325 if (read(d->fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) {
6326 perror("Read PPL header failed");
6327 ret = -1;
e397cefe 6328 break;
44b6b876 6329 }
2432ce9b 6330
44b6b876 6331 ppl_hdr = buf;
2432ce9b 6332
44b6b876
PB		 6333 crc = __le32_to_cpu(ppl_hdr->checksum);
6334 ppl_hdr->checksum = 0;
6335
6336 if (crc != ~crc32c_le(~0, buf, PPL_HEADER_SIZE)) {
6337 dprintf("Wrong PPL header checksum on %s\n",
6338 d->devname);
e397cefe 6339 break;
44b6b876
PB		 6340 }
6341
6342 if (prev_gen_num > __le64_to_cpu(ppl_hdr->generation)) {
6343 /* previous was newest, it was already checked */
e397cefe 6344 break;
44b6b876
PB		 6345 }
6346
6347 if ((__le32_to_cpu(ppl_hdr->signature) !=
6348 super->anchor->orig_family_num)) {
6349 dprintf("Wrong PPL header signature on %s\n",
6350 d->devname);
6351 ret = 1;
e397cefe 6352 break;
44b6b876
PB		 6353 }
6354
6355 ret = 0;
6356 prev_gen_num = __le64_to_cpu(ppl_hdr->generation);
2432ce9b 6357
44b6b876
PB		 6358 ppl_offset += PPL_HEADER_SIZE;
6359 for (i = 0; i < __le32_to_cpu(ppl_hdr->entries_count); i++)
6360 ppl_offset +=
6361 __le32_to_cpu(ppl_hdr->entries[i].pp_size);
e397cefe
AP		 6362
6363 if (!buf_prev)
6364 buf_prev = buf + PPL_HEADER_SIZE;
6365 tmp = buf_prev;
6366 buf_prev = buf;
6367 buf = tmp;
2432ce9b
AP		 6368 }
6369
e397cefe
AP		 6370 if (buf_prev) {
6371 buf = buf_prev;
6372 ppl_hdr = buf_prev;
6373 }
2432ce9b 6374
54148aba
PB		 6375 /*
6376 * Update metadata to use mutliple PPLs area (1MB).
6377 * This is done once for all RAID members
6378 */
6379 if (info->consistency_policy == CONSISTENCY_POLICY_PPL &&
6380 info->ppl_size != (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9)) {
6381 char subarray[20];
6382 struct mdinfo *member_dev;
6383
6384 sprintf(subarray, "%d", info->container_member);
6385
6386 if (mdmon_running(st->container_devnm))
6387 st->update_tail = &st->updates;
6388
6389 if (st->ss->update_subarray(st, subarray, "ppl", NULL)) {
6390 pr_err("Failed to update subarray %s\n",
6391 subarray);
6392 } else {
6393 if (st->update_tail)
6394 flush_metadata_updates(st);
6395 else
6396 st->ss->sync_metadata(st);
6397 info->ppl_size = (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9);
6398 for (member_dev = info->devs; member_dev;
6399 member_dev = member_dev->next)
6400 member_dev->ppl_size =
6401 (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9);
6402 }
6403 }
6404
b23d0750 6405 if (ret == 1) {
2fc0fc63
AP		 6406 struct imsm_map *map = get_imsm_map(dev, MAP_X);
6407
50b9c10d
PB		 6408 if (map->map_state == IMSM_T_STATE_UNINITIALIZED ||
6409 (map->map_state == IMSM_T_STATE_NORMAL &&
2ec9d182 6410 !(dev->vol.dirty & RAIDVOL_DIRTY)) ||
e397cefe 6411 (is_rebuilding(dev) &&
2ec9d182
AP		 6412 dev->vol.curr_migr_unit == 0 &&
6413 get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_1) != idx))
b23d0750
AP		 6414 ret = st->ss->write_init_ppl(st, info, d->fd);
6415 else
6416 info->mismatch_cnt++;
e397cefe
AP		 6417 } else if (ret == 0 &&
6418 ppl_hdr->entries_count == 0 &&
6419 is_rebuilding(dev) &&
6420 info->resync_start == 0) {
6421 /*
6422 * The header has no entries - add a single empty entry and
6423 * rewrite the header to prevent the kernel from going into
6424 * resync after an interrupted rebuild.
6425 */
6426 ppl_hdr->entries_count = __cpu_to_le32(1);
6427 ret = write_ppl_header(info->ppl_sector, d->fd, buf);
b23d0750 6428 }
2432ce9b 6429
e397cefe
AP		 6430 free(buf_orig);
6431
2432ce9b
AP		 6432 return ret;
6433}
6434
2432ce9b
AP		 6435static int write_init_ppl_imsm_all(struct supertype *st, struct mdinfo *info)
6436{
6437 struct intel_super *super = st->sb;
6438 struct dl *d;
6439 int ret = 0;
6440
6441 if (info->consistency_policy != CONSISTENCY_POLICY_PPL ||
6442 info->array.level != 5)
6443 return 0;
6444
6445 for (d = super->disks; d ; d = d->next) {
6446 if (d->index < 0 || is_failed(&d->disk))
6447 continue;
6448
6449 ret = st->ss->write_init_ppl(st, info, d->fd);
6450 if (ret)
6451 break;
6452 }
6453
6454 return ret;
6455}
43dad3d6 6456
c2c087e6
DW		 6457static int write_init_super_imsm(struct supertype *st)
6458{
9b1fb677
DW		 6459 struct intel_super *super = st->sb;
6460 int current_vol = super->current_vol;
2432ce9b
AP		 6461 int rv = 0;
6462 struct mdinfo info;
6463
6464 getinfo_super_imsm(st, &info, NULL);
9b1fb677
DW		 6465
6466 /* we are done with current_vol reset it to point st at the container */
6467 super->current_vol = -1;
6468
8273f55e 6469 if (st->update_tail) {
43dad3d6
DW		 6470 /* queue the recently created array / added disk
6471 * as a metadata update */
8273f55e 6472
43dad3d6 6473 /* determine if we are creating a volume or adding a disk */
9b1fb677 6474 if (current_vol < 0) {
1a64be56
LM		 6475 /* in the mgmt (add/remove) disk case we are running
6476 * in mdmon context, so don't close fd's
43dad3d6 6477 */
2432ce9b
AP		 6478 rv = mgmt_disk(st);
6479 } else {
6480 rv = write_init_ppl_imsm_all(st, &info);
6481 if (!rv)
6482 rv = create_array(st, current_vol);
6483 }
d682f344
N		 6484 } else {
6485 struct dl *d;
6486 for (d = super->disks; d; d = d->next)
ba728be7 6487 Kill(d->devname, NULL, 0, -1, 1);
2432ce9b
AP		 6488 if (current_vol >= 0)
6489 rv = write_init_ppl_imsm_all(st, &info);
6490 if (!rv)
6491 rv = write_super_imsm(st, 1);
d682f344 6492 }
2432ce9b
AP		 6493
6494 return rv;
cdddbdbc
DW		 6495}
6496
e683ca88 6497static int store_super_imsm(struct supertype *st, int fd)
cdddbdbc 6498{
e683ca88
DW		 6499 struct intel_super *super = st->sb;
6500 struct imsm_super *mpb = super ? super->anchor : NULL;
551c80c1 6501
e683ca88 6502 if (!mpb)
ad97895e
DW		 6503 return 1;
6504
f36a9ecd
PB		 6505 if (super->sector_size == 4096)
6506 convert_to_4k(super);
e683ca88 6507 return store_imsm_mpb(fd, mpb);
cdddbdbc
DW		 6508}
6509
cdddbdbc
DW		 6510static int validate_geometry_imsm_container(struct supertype *st, int level,
6511 int layout, int raiddisks, int chunk,
af4348dd
N		 6512 unsigned long long size,
6513 unsigned long long data_offset,
6514 char *dev,
2c514b71
NB		 6515 unsigned long long *freesize,
6516 int verbose)
cdddbdbc 6517{
c2c087e6
DW		 6518 int fd;
6519 unsigned long long ldsize;
594dc1b8 6520 struct intel_super *super;
f2f5c343 6521 int rv = 0;
cdddbdbc 6522
c2c087e6
DW		 6523 if (level != LEVEL_CONTAINER)
6524 return 0;
6525 if (!dev)
6526 return 1;
6527
6528 fd = open(dev, O_RDONLY|O_EXCL, 0);
6529 if (fd < 0) {
ba728be7 6530 if (verbose > 0)
e7b84f9d 6531 pr_err("imsm: Cannot open %s: %s\n",
2c514b71 6532 dev, strerror(errno));
c2c087e6
DW		 6533 return 0;
6534 }
6535 if (!get_dev_size(fd, dev, &ldsize)) {
6536 close(fd);
6537 return 0;
6538 }
f2f5c343
LM		 6539
6540 /* capabilities retrieve could be possible
6541 * note that there is no fd for the disks in array.
6542 */
6543 super = alloc_super();
8d67477f
TM		 6544 if (!super) {
6545 close(fd);
6546 return 0;
6547 }
fa7bb6f8
PB		 6548 if (!get_dev_sector_size(fd, NULL, &super->sector_size)) {
6549 close(fd);
6550 free_imsm(super);
6551 return 0;
6552 }
6553
ba728be7 6554 rv = find_intel_hba_capability(fd, super, verbose > 0 ? dev : NULL);
f2f5c343
LM		 6555 if (rv != 0) {
6556#if DEBUG
6557 char str[256];
6558 fd2devname(fd, str);
1ade5cc1 6559 dprintf("fd: %d %s orom: %p rv: %d raiddisk: %d\n",
f2f5c343
LM		 6560 fd, str, super->orom, rv, raiddisks);
6561#endif
6562 /* no orom/efi or non-intel hba of the disk */
6563 close(fd);
6564 free_imsm(super);
6565 return 0;
6566 }
c2c087e6 6567 close(fd);
9126b9a8
CA		 6568 if (super->orom) {
6569 if (raiddisks > super->orom->tds) {
6570 if (verbose)
7a862a02 6571 pr_err("%d exceeds maximum number of platform supported disks: %d\n",
9126b9a8
CA		 6572 raiddisks, super->orom->tds);
6573 free_imsm(super);
6574 return 0;
6575 }
6576 if ((super->orom->attr & IMSM_OROM_ATTR_2TB_DISK) == 0 &&
6577 (ldsize >> 9) >> 32 > 0) {
6578 if (verbose)
e7b84f9d 6579 pr_err("%s exceeds maximum platform supported size\n", dev);
9126b9a8
CA		 6580 free_imsm(super);
6581 return 0;
6582 }
f2f5c343 6583 }
c2c087e6 6584
af4348dd 6585 *freesize = avail_size_imsm(st, ldsize >> 9, data_offset);
f2f5c343 6586 free_imsm(super);
c2c087e6
DW		 6587
6588 return 1;
cdddbdbc
DW		 6589}
6590
0dcecb2e
DW		 6591static unsigned long long find_size(struct extent *e, int *idx, int num_extents)
6592{
6593 const unsigned long long base_start = e[*idx].start;
6594 unsigned long long end = base_start + e[*idx].size;
6595 int i;
6596
6597 if (base_start == end)
6598 return 0;
6599
6600 *idx = *idx + 1;
6601 for (i = *idx; i < num_extents; i++) {
6602 /* extend overlapping extents */
6603 if (e[i].start >= base_start &&
6604 e[i].start <= end) {
6605 if (e[i].size == 0)
6606 return 0;
6607 if (e[i].start + e[i].size > end)
6608 end = e[i].start + e[i].size;
6609 } else if (e[i].start > end) {
6610 *idx = i;
6611 break;
6612 }
6613 }
6614
6615 return end - base_start;
6616}
6617
6618static unsigned long long merge_extents(struct intel_super *super, int sum_extents)
6619{
6620 /* build a composite disk with all known extents and generate a new
6621 * 'maxsize' given the "all disks in an array must share a common start
6622 * offset" constraint
6623 */
503975b9 6624 struct extent *e = xcalloc(sum_extents, sizeof(*e));
0dcecb2e
DW		 6625 struct dl *dl;
6626 int i, j;
6627 int start_extent;
6628 unsigned long long pos;
b9d77223 6629 unsigned long long start = 0;
0dcecb2e
DW		 6630 unsigned long long maxsize;
6631 unsigned long reserve;
6632
0dcecb2e
DW		 6633 /* coalesce and sort all extents. also, check to see if we need to
6634 * reserve space between member arrays
6635 */
6636 j = 0;
6637 for (dl = super->disks; dl; dl = dl->next) {
6638 if (!dl->e)
6639 continue;
6640 for (i = 0; i < dl->extent_cnt; i++)
6641 e[j++] = dl->e[i];
6642 }
6643 qsort(e, sum_extents, sizeof(*e), cmp_extent);
6644
6645 /* merge extents */
6646 i = 0;
6647 j = 0;
6648 while (i < sum_extents) {
6649 e[j].start = e[i].start;
6650 e[j].size = find_size(e, &i, sum_extents);
6651 j++;
6652 if (e[j-1].size == 0)
6653 break;
6654 }
6655
6656 pos = 0;
6657 maxsize = 0;
6658 start_extent = 0;
6659 i = 0;
6660 do {
6661 unsigned long long esize;
6662
6663 esize = e[i].start - pos;
6664 if (esize >= maxsize) {
6665 maxsize = esize;
6666 start = pos;
6667 start_extent = i;
6668 }
6669 pos = e[i].start + e[i].size;
6670 i++;
6671 } while (e[i-1].size);
6672 free(e);
6673
a7dd165b
DW		 6674 if (maxsize == 0)
6675 return 0;
6676
6677 /* FIXME assumes volume at offset 0 is the first volume in a
6678 * container
6679 */
0dcecb2e
DW		 6680 if (start_extent > 0)
6681 reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */
6682 else
6683 reserve = 0;
6684
6685 if (maxsize < reserve)
a7dd165b 6686 return 0;
0dcecb2e 6687
5551b113 6688 super->create_offset = ~((unsigned long long) 0);
0dcecb2e 6689 if (start + reserve > super->create_offset)
a7dd165b 6690 return 0; /* start overflows create_offset */
0dcecb2e
DW		 6691 super->create_offset = start + reserve;
6692
6693 return maxsize - reserve;
6694}
6695
88c32bb1
DW		 6696static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks)
6697{
6698 if (level < 0 || level == 6 || level == 4)
6699 return 0;
6700
6701 /* if we have an orom prevent invalid raid levels */
6702 if (orom)
6703 switch (level) {
6704 case 0: return imsm_orom_has_raid0(orom);
6705 case 1:
6706 if (raiddisks > 2)
6707 return imsm_orom_has_raid1e(orom);
1c556e92
DW		 6708 return imsm_orom_has_raid1(orom) && raiddisks == 2;
6709 case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4;
6710 case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2;
88c32bb1
DW		 6711 }
6712 else
6713 return 1; /* not on an Intel RAID platform so anything goes */
6714
6715 return 0;
6716}
6717
ca9de185
LM		 6718static int
6719active_arrays_by_format(char *name, char* hba, struct md_list **devlist,
6720 int dpa, int verbose)
6721{
6722 struct mdstat_ent *mdstat = mdstat_read(0, 0);
594dc1b8 6723 struct mdstat_ent *memb;
ca9de185
LM		 6724 int count = 0;
6725 int num = 0;
594dc1b8 6726 struct md_list *dv;
ca9de185
LM		 6727 int found;
6728
6729 for (memb = mdstat ; memb ; memb = memb->next) {
6730 if (memb->metadata_version &&
fc54fe7a 6731 (strncmp(memb->metadata_version, "external:", 9) == 0) &&
ca9de185
LM		 6732 (strcmp(&memb->metadata_version[9], name) == 0) &&
6733 !is_subarray(memb->metadata_version+9) &&
6734 memb->members) {
6735 struct dev_member *dev = memb->members;
6736 int fd = -1;
6737 while(dev && (fd < 0)) {
503975b9
N		 6738 char *path = xmalloc(strlen(dev->name) + strlen("/dev/") + 1);
6739 num = sprintf(path, "%s%s", "/dev/", dev->name);
6740 if (num > 0)
6741 fd = open(path, O_RDONLY, 0);
089f9d79 6742 if (num <= 0 || fd < 0) {
676e87a8 6743 pr_vrb("Cannot open %s: %s\n",
503975b9 6744 dev->name, strerror(errno));
ca9de185 6745 }
503975b9 6746 free(path);
ca9de185
LM		 6747 dev = dev->next;
6748 }
6749 found = 0;
089f9d79 6750 if (fd >= 0 && disk_attached_to_hba(fd, hba)) {
ca9de185
LM		 6751 struct mdstat_ent *vol;
6752 for (vol = mdstat ; vol ; vol = vol->next) {
089f9d79 6753 if (vol->active > 0 &&
ca9de185 6754 vol->metadata_version &&
9581efb1 6755 is_container_member(vol, memb->devnm)) {
ca9de185
LM		 6756 found++;
6757 count++;
6758 }
6759 }
6760 if (*devlist && (found < dpa)) {
503975b9 6761 dv = xcalloc(1, sizeof(*dv));
9581efb1
N		 6762 dv->devname = xmalloc(strlen(memb->devnm) + strlen("/dev/") + 1);
6763 sprintf(dv->devname, "%s%s", "/dev/", memb->devnm);
503975b9
N		 6764 dv->found = found;
6765 dv->used = 0;
6766 dv->next = *devlist;
6767 *devlist = dv;
ca9de185
LM		 6768 }
6769 }
6770 if (fd >= 0)
6771 close(fd);
6772 }
6773 }
6774 free_mdstat(mdstat);
6775 return count;
6776}
6777
6778#ifdef DEBUG_LOOP
6779static struct md_list*
6780get_loop_devices(void)
6781{
6782 int i;
6783 struct md_list *devlist = NULL;
594dc1b8 6784 struct md_list *dv;
ca9de185
LM		 6785
6786 for(i = 0; i < 12; i++) {
503975b9
N		 6787 dv = xcalloc(1, sizeof(*dv));
6788 dv->devname = xmalloc(40);
ca9de185
LM		 6789 sprintf(dv->devname, "/dev/loop%d", i);
6790 dv->next = devlist;
6791 devlist = dv;
6792 }
6793 return devlist;
6794}
6795#endif
6796
6797static struct md_list*
6798get_devices(const char *hba_path)
6799{
6800 struct md_list *devlist = NULL;
594dc1b8 6801 struct md_list *dv;
ca9de185
LM		 6802 struct dirent *ent;
6803 DIR *dir;
6804 int err = 0;
6805
6806#if DEBUG_LOOP
6807 devlist = get_loop_devices();
6808 return devlist;
6809#endif
6810 /* scroll through /sys/dev/block looking for devices attached to
6811 * this hba
6812 */
6813 dir = opendir("/sys/dev/block");
6814 for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) {
6815 int fd;
6816 char buf[1024];
6817 int major, minor;
6818 char *path = NULL;
6819 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
6820 continue;
6821 path = devt_to_devpath(makedev(major, minor));
6822 if (!path)
6823 continue;
6824 if (!path_attached_to_hba(path, hba_path)) {
6825 free(path);
6826 path = NULL;
6827 continue;
6828 }
6829 free(path);
6830 path = NULL;
6831 fd = dev_open(ent->d_name, O_RDONLY);
6832 if (fd >= 0) {
6833 fd2devname(fd, buf);
6834 close(fd);
6835 } else {
e7b84f9d 6836 pr_err("cannot open device: %s\n",
ca9de185
LM		 6837 ent->d_name);
6838 continue;
6839 }
6840
503975b9
N		 6841 dv = xcalloc(1, sizeof(*dv));
6842 dv->devname = xstrdup(buf);
ca9de185
LM		 6843 dv->next = devlist;
6844 devlist = dv;
6845 }
6846 if (err) {
6847 while(devlist) {
6848 dv = devlist;
6849 devlist = devlist->next;
6850 free(dv->devname);
6851 free(dv);
6852 }
6853 }
562aa102 6854 closedir(dir);
ca9de185
LM		 6855 return devlist;
6856}
6857
6858static int
6859count_volumes_list(struct md_list *devlist, char *homehost,
6860 int verbose, int *found)
6861{
6862 struct md_list *tmpdev;
6863 int count = 0;
594dc1b8 6864 struct supertype *st;
ca9de185
LM		 6865
6866 /* first walk the list of devices to find a consistent set
6867 * that match the criterea, if that is possible.
6868 * We flag the ones we like with 'used'.
6869 */
6870 *found = 0;
6871 st = match_metadata_desc_imsm("imsm");
6872 if (st == NULL) {
676e87a8 6873 pr_vrb("cannot allocate memory for imsm supertype\n");
ca9de185
LM		 6874 return 0;
6875 }
6876
6877 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
6878 char *devname = tmpdev->devname;
0a6bff09 6879 dev_t rdev;
ca9de185
LM		 6880 struct supertype *tst;
6881 int dfd;
6882 if (tmpdev->used > 1)
6883 continue;
6884 tst = dup_super(st);
6885 if (tst == NULL) {
676e87a8 6886 pr_vrb("cannot allocate memory for imsm supertype\n");
ca9de185
LM		 6887 goto err_1;
6888 }
6889 tmpdev->container = 0;
6890 dfd = dev_open(devname, O_RDONLY|O_EXCL);
6891 if (dfd < 0) {
1ade5cc1 6892 dprintf("cannot open device %s: %s\n",
ca9de185
LM		 6893 devname, strerror(errno));
6894 tmpdev->used = 2;
0a6bff09 6895 } else if (!fstat_is_blkdev(dfd, devname, &rdev)) {
ca9de185
LM		 6896 tmpdev->used = 2;
6897 } else if (must_be_container(dfd)) {
6898 struct supertype *cst;
6899 cst = super_by_fd(dfd, NULL);
6900 if (cst == NULL) {
1ade5cc1 6901 dprintf("cannot recognize container type %s\n",
ca9de185
LM		 6902 devname);
6903 tmpdev->used = 2;
6904 } else if (tst->ss != st->ss) {
1ade5cc1 6905 dprintf("non-imsm container - ignore it: %s\n",
ca9de185
LM		 6906 devname);
6907 tmpdev->used = 2;
6908 } else if (!tst->ss->load_container ||
6909 tst->ss->load_container(tst, dfd, NULL))
6910 tmpdev->used = 2;
6911 else {
6912 tmpdev->container = 1;
6913 }
6914 if (cst)
6915 cst->ss->free_super(cst);
6916 } else {
0a6bff09 6917 tmpdev->st_rdev = rdev;
ca9de185 6918 if (tst->ss->load_super(tst,dfd, NULL)) {
1ade5cc1 6919 dprintf("no RAID superblock on %s\n",
ca9de185
LM		 6920 devname);
6921 tmpdev->used = 2;
6922 } else if (tst->ss->compare_super == NULL) {
1ade5cc1 6923 dprintf("Cannot assemble %s metadata on %s\n",
ca9de185
LM		 6924 tst->ss->name, devname);
6925 tmpdev->used = 2;
6926 }
6927 }
6928 if (dfd >= 0)
6929 close(dfd);
6930 if (tmpdev->used == 2 || tmpdev->used == 4) {
6931 /* Ignore unrecognised devices during auto-assembly */
6932 goto loop;
6933 }
6934 else {
6935 struct mdinfo info;
6936 tst->ss->getinfo_super(tst, &info, NULL);
6937
6938 if (st->minor_version == -1)
6939 st->minor_version = tst->minor_version;
6940
6941 if (memcmp(info.uuid, uuid_zero,
6942 sizeof(int[4])) == 0) {
6943 /* this is a floating spare. It cannot define
6944 * an array unless there are no more arrays of
6945 * this type to be found. It can be included
6946 * in an array of this type though.
6947 */
6948 tmpdev->used = 3;
6949 goto loop;
6950 }
6951
6952 if (st->ss != tst->ss ||
6953 st->minor_version != tst->minor_version ||
c7b8547c 6954 st->ss->compare_super(st, tst, 1) != 0) {
ca9de185
LM		 6955 /* Some mismatch. If exactly one array matches this host,
6956 * we can resolve on that one.
6957 * Or, if we are auto assembling, we just ignore the second
6958 * for now.
6959 */
1ade5cc1 6960 dprintf("superblock on %s doesn't match others - assembly aborted\n",
ca9de185
LM		 6961 devname);
6962 goto loop;
6963 }
6964 tmpdev->used = 1;
6965 *found = 1;
6966 dprintf("found: devname: %s\n", devname);
6967 }
6968 loop:
6969 if (tst)
6970 tst->ss->free_super(tst);
6971 }
6972 if (*found != 0) {
6973 int err;
6974 if ((err = load_super_imsm_all(st, -1, &st->sb, NULL, devlist, 0)) == 0) {
6975 struct mdinfo *iter, *head = st->ss->container_content(st, NULL);
6976 for (iter = head; iter; iter = iter->next) {
6977 dprintf("content->text_version: %s vol\n",
6978 iter->text_version);
6979 if (iter->array.state & (1<<MD_SB_BLOCK_VOLUME)) {
6980 /* do not assemble arrays with unsupported
6981 configurations */
1ade5cc1 6982 dprintf("Cannot activate member %s.\n",
ca9de185
LM		 6983 iter->text_version);
6984 } else
6985 count++;
6986 }
6987 sysfs_free(head);
6988
6989 } else {
1ade5cc1 6990 dprintf("No valid super block on device list: err: %d %p\n",
ca9de185
LM		 6991 err, st->sb);
6992 }
6993 } else {
1ade5cc1 6994 dprintf("no more devices to examine\n");
ca9de185
LM		 6995 }
6996
6997 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
089f9d79 6998 if (tmpdev->used == 1 && tmpdev->found) {
ca9de185
LM		 6999 if (count) {
7000 if (count < tmpdev->found)
7001 count = 0;
7002 else
7003 count -= tmpdev->found;
7004 }
7005 }
7006 if (tmpdev->used == 1)
7007 tmpdev->used = 4;
7008 }
7009 err_1:
7010 if (st)
7011 st->ss->free_super(st);
7012 return count;
7013}
7014
d3c11416
AO		 7015static int __count_volumes(char *hba_path, int dpa, int verbose,
7016 int cmp_hba_path)
ca9de185 7017{
72a45777 7018 struct sys_dev *idev, *intel_devices = find_intel_devices();
ca9de185 7019 int count = 0;
72a45777
PB		 7020 const struct orom_entry *entry;
7021 struct devid_list *dv, *devid_list;
ca9de185 7022
d3c11416 7023 if (!hba_path)
ca9de185
LM		 7024 return 0;
7025
72a45777 7026 for (idev = intel_devices; idev; idev = idev->next) {
d3c11416
AO		 7027 if (strstr(idev->path, hba_path))
7028 break;
72a45777
PB		 7029 }
7030
7031 if (!idev || !idev->dev_id)
ca9de185 7032 return 0;
72a45777
PB		 7033
7034 entry = get_orom_entry_by_device_id(idev->dev_id);
7035
7036 if (!entry || !entry->devid_list)
7037 return 0;
7038
7039 devid_list = entry->devid_list;
7040 for (dv = devid_list; dv; dv = dv->next) {
594dc1b8 7041 struct md_list *devlist;
d3c11416
AO		 7042 struct sys_dev *device = NULL;
7043 char *hpath;
72a45777
PB		 7044 int found = 0;
7045
d3c11416
AO		 7046 if (cmp_hba_path)
7047 device = device_by_id_and_path(dv->devid, hba_path);
7048 else
7049 device = device_by_id(dv->devid);
7050
72a45777 7051 if (device)
d3c11416 7052 hpath = device->path;
72a45777
PB		 7053 else
7054 return 0;
7055
d3c11416 7056 devlist = get_devices(hpath);
72a45777
PB		 7057 /* if no intel devices return zero volumes */
7058 if (devlist == NULL)
7059 return 0;
7060
d3c11416
AO		 7061 count += active_arrays_by_format("imsm", hpath, &devlist, dpa,
7062 verbose);
7063 dprintf("path: %s active arrays: %d\n", hpath, count);
72a45777
PB		 7064 if (devlist == NULL)
7065 return 0;
7066 do {
7067 found = 0;
7068 count += count_volumes_list(devlist,
7069 NULL,
7070 verbose,
7071 &found);
7072 dprintf("found %d count: %d\n", found, count);
7073 } while (found);
7074
d3c11416 7075 dprintf("path: %s total number of volumes: %d\n", hpath, count);
72a45777
PB		 7076
7077 while (devlist) {
7078 struct md_list *dv = devlist;
7079 devlist = devlist->next;
7080 free(dv->devname);
7081 free(dv);
7082 }
ca9de185
LM		 7083 }
7084 return count;
7085}
7086
d3c11416
AO		 7087static int count_volumes(struct intel_hba *hba, int dpa, int verbose)
7088{
7089 if (!hba)
7090 return 0;
7091 if (hba->type == SYS_DEV_VMD) {
7092 struct sys_dev *dev;
7093 int count = 0;
7094
7095 for (dev = find_intel_devices(); dev; dev = dev->next) {
7096 if (dev->type == SYS_DEV_VMD)
7097 count += __count_volumes(dev->path, dpa,
7098 verbose, 1);
7099 }
7100 return count;
7101 }
7102 return __count_volumes(hba->path, dpa, verbose, 0);
7103}
7104
cd9d1ac7
DW		 7105static int imsm_default_chunk(const struct imsm_orom *orom)
7106{
7107 /* up to 512 if the plaform supports it, otherwise the platform max.
7108 * 128 if no platform detected
7109 */
7110 int fs = max(7, orom ? fls(orom->sss) : 0);
7111
7112 return min(512, (1 << fs));
7113}
73408129 7114
6592ce37
DW		 7115static int
7116validate_geometry_imsm_orom(struct intel_super *super, int level, int layout,
2cc699af 7117 int raiddisks, int *chunk, unsigned long long size, int verbose)
6592ce37 7118{
660260d0
DW		 7119 /* check/set platform and metadata limits/defaults */
7120 if (super->orom && raiddisks > super->orom->dpa) {
676e87a8 7121 pr_vrb("platform supports a maximum of %d disks per array\n",
660260d0 7122 super->orom->dpa);
73408129
LM		 7123 return 0;
7124 }
7125
5d500228 7126 /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */
660260d0 7127 if (!is_raid_level_supported(super->orom, level, raiddisks)) {
676e87a8 7128 pr_vrb("platform does not support raid%d with %d disk%s\n",
6592ce37
DW		 7129 level, raiddisks, raiddisks > 1 ? "s" : "");
7130 return 0;
7131 }
cd9d1ac7 7132
7ccc4cc4 7133 if (*chunk == 0 || *chunk == UnSet)
cd9d1ac7
DW		 7134 *chunk = imsm_default_chunk(super->orom);
7135
7ccc4cc4 7136 if (super->orom && !imsm_orom_has_chunk(super->orom, *chunk)) {
676e87a8 7137 pr_vrb("platform does not support a chunk size of: %d\n", *chunk);
cd9d1ac7 7138 return 0;
6592ce37 7139 }
cd9d1ac7 7140
6592ce37
DW		 7141 if (layout != imsm_level_to_layout(level)) {
7142 if (level == 5)
676e87a8 7143 pr_vrb("imsm raid 5 only supports the left-asymmetric layout\n");
6592ce37 7144 else if (level == 10)
676e87a8 7145 pr_vrb("imsm raid 10 only supports the n2 layout\n");
6592ce37 7146 else
676e87a8 7147 pr_vrb("imsm unknown layout %#x for this raid level %d\n",
6592ce37
DW		 7148 layout, level);
7149 return 0;
7150 }
2cc699af 7151
7ccc4cc4 7152 if (super->orom && (super->orom->attr & IMSM_OROM_ATTR_2TB) == 0 &&
2cc699af 7153 (calc_array_size(level, raiddisks, layout, *chunk, size) >> 32) > 0) {
676e87a8 7154 pr_vrb("platform does not support a volume size over 2TB\n");
2cc699af
CA		 7155 return 0;
7156 }
614902f6 7157
6592ce37
DW		 7158 return 1;
7159}
7160
1011e834 7161/* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd
c2c087e6
DW		 7162 * FIX ME add ahci details
7163 */
8b353278 7164static int validate_geometry_imsm_volume(struct supertype *st, int level,
c21e737b 7165 int layout, int raiddisks, int *chunk,
af4348dd
N		 7166 unsigned long long size,
7167 unsigned long long data_offset,
7168 char *dev,
2c514b71
NB		 7169 unsigned long long *freesize,
7170 int verbose)
cdddbdbc 7171{
9e04ac1c 7172 dev_t rdev;
c2c087e6 7173 struct intel_super *super = st->sb;
b2916f25 7174 struct imsm_super *mpb;
c2c087e6
DW		 7175 struct dl *dl;
7176 unsigned long long pos = 0;
7177 unsigned long long maxsize;
7178 struct extent *e;
7179 int i;
cdddbdbc 7180
88c32bb1
DW		 7181 /* We must have the container info already read in. */
7182 if (!super)
c2c087e6
DW		 7183 return 0;
7184
b2916f25
JS		 7185 mpb = super->anchor;
7186
2cc699af 7187 if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, size, verbose)) {
3e684231 7188 pr_err("RAID geometry validation failed. Cannot proceed with the action(s).\n");
c2c087e6 7189 return 0;
d54559f0 7190 }
c2c087e6
DW		 7191 if (!dev) {
7192 /* General test: make sure there is space for
2da8544a
DW		 7193 * 'raiddisks' device extents of size 'size' at a given
7194 * offset
c2c087e6 7195 */
e46273eb 7196 unsigned long long minsize = size;
b7528a20 7197 unsigned long long start_offset = MaxSector;
c2c087e6
DW		 7198 int dcnt = 0;
7199 if (minsize == 0)
7200 minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
7201 for (dl = super->disks; dl ; dl = dl->next) {
7202 int found = 0;
7203
bf5a934a 7204 pos = 0;
c2c087e6 7205 i = 0;
05501181 7206 e = get_extents(super, dl, 0);
c2c087e6
DW		 7207 if (!e) continue;
7208 do {
7209 unsigned long long esize;
7210 esize = e[i].start - pos;
7211 if (esize >= minsize)
7212 found = 1;
b7528a20 7213 if (found && start_offset == MaxSector) {
2da8544a
DW		 7214 start_offset = pos;
7215 break;
7216 } else if (found && pos != start_offset) {
7217 found = 0;
7218 break;
7219 }
c2c087e6
DW		 7220 pos = e[i].start + e[i].size;
7221 i++;
7222 } while (e[i-1].size);
7223 if (found)
7224 dcnt++;
7225 free(e);
7226 }
7227 if (dcnt < raiddisks) {
2c514b71 7228 if (verbose)
7a862a02 7229 pr_err("imsm: Not enough devices with space for this array (%d < %d)\n",
2c514b71 7230 dcnt, raiddisks);
c2c087e6
DW		 7231 return 0;
7232 }
7233 return 1;
7234 }
0dcecb2e 7235
c2c087e6 7236 /* This device must be a member of the set */
9e04ac1c 7237 if (!stat_is_blkdev(dev, &rdev))
c2c087e6
DW		 7238 return 0;
7239 for (dl = super->disks ; dl ; dl = dl->next) {
9e04ac1c
ZL		 7240 if (dl->major == (int)major(rdev) &&
7241 dl->minor == (int)minor(rdev))
c2c087e6
DW		 7242 break;
7243 }
7244 if (!dl) {
2c514b71 7245 if (verbose)
7a862a02 7246 pr_err("%s is not in the same imsm set\n", dev);
c2c087e6 7247 return 0;
a20d2ba5
DW		 7248 } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) {
7249 /* If a volume is present then the current creation attempt
7250 * cannot incorporate new spares because the orom may not
7251 * understand this configuration (all member disks must be
7252 * members of each array in the container).
7253 */
7a862a02
N		 7254 pr_err("%s is a spare and a volume is already defined for this container\n", dev);
7255 pr_err("The option-rom requires all member disks to be a member of all volumes\n");
a20d2ba5 7256 return 0;
5fe62b94
WD		 7257 } else if (super->orom && mpb->num_raid_devs > 0 &&
7258 mpb->num_disks != raiddisks) {
7a862a02 7259 pr_err("The option-rom requires all member disks to be a member of all volumes\n");
5fe62b94 7260 return 0;
c2c087e6 7261 }
0dcecb2e
DW		 7262
7263 /* retrieve the largest free space block */
05501181 7264 e = get_extents(super, dl, 0);
c2c087e6
DW		 7265 maxsize = 0;
7266 i = 0;
0dcecb2e
DW		 7267 if (e) {
7268 do {
7269 unsigned long long esize;
7270
7271 esize = e[i].start - pos;
7272 if (esize >= maxsize)
7273 maxsize = esize;
7274 pos = e[i].start + e[i].size;
7275 i++;
7276 } while (e[i-1].size);
7277 dl->e = e;
7278 dl->extent_cnt = i;
7279 } else {
7280 if (verbose)
e7b84f9d 7281 pr_err("unable to determine free space for: %s\n",
0dcecb2e
DW		 7282 dev);
7283 return 0;
7284 }
7285 if (maxsize < size) {
7286 if (verbose)
e7b84f9d 7287 pr_err("%s not enough space (%llu < %llu)\n",
0dcecb2e
DW		 7288 dev, maxsize, size);
7289 return 0;
7290 }
7291
7292 /* count total number of extents for merge */
7293 i = 0;
7294 for (dl = super->disks; dl; dl = dl->next)
7295 if (dl->e)
7296 i += dl->extent_cnt;
7297
7298 maxsize = merge_extents(super, i);
3baa56ab 7299
1a1ced1e
KS		 7300 if (mpb->num_raid_devs > 0 && size && size != maxsize)
7301 pr_err("attempting to create a second volume with size less then remaining space.\n");
3baa56ab 7302
a7dd165b 7303 if (maxsize < size || maxsize == 0) {
b3071342
LD		 7304 if (verbose) {
7305 if (maxsize == 0)
7a862a02 7306 pr_err("no free space left on device. Aborting...\n");
b3071342 7307 else
7a862a02 7308 pr_err("not enough space to create volume of given size (%llu < %llu). Aborting...\n",
b3071342
LD		 7309 maxsize, size);
7310 }
0dcecb2e 7311 return 0;
0dcecb2e
DW		 7312 }
7313
c2c087e6
DW		 7314 *freesize = maxsize;
7315
ca9de185 7316 if (super->orom) {
72a45777 7317 int count = count_volumes(super->hba,
ca9de185
LM		 7318 super->orom->dpa, verbose);
7319 if (super->orom->vphba <= count) {
676e87a8 7320 pr_vrb("platform does not support more than %d raid volumes.\n",
ca9de185
LM		 7321 super->orom->vphba);
7322 return 0;
7323 }
7324 }
c2c087e6 7325 return 1;
cdddbdbc
DW		 7326}
7327
13bcac90 7328static int imsm_get_free_size(struct supertype *st, int raiddisks,
efb30e7f
DW		 7329 unsigned long long size, int chunk,
7330 unsigned long long *freesize)
7331{
7332 struct intel_super *super = st->sb;
7333 struct imsm_super *mpb = super->anchor;
7334 struct dl *dl;
7335 int i;
7336 int extent_cnt;
7337 struct extent *e;
7338 unsigned long long maxsize;
7339 unsigned long long minsize;
7340 int cnt;
7341 int used;
7342
7343 /* find the largest common start free region of the possible disks */
7344 used = 0;
7345 extent_cnt = 0;
7346 cnt = 0;
7347 for (dl = super->disks; dl; dl = dl->next) {
7348 dl->raiddisk = -1;
7349
7350 if (dl->index >= 0)
7351 used++;
7352
7353 /* don't activate new spares if we are orom constrained
7354 * and there is already a volume active in the container
7355 */
7356 if (super->orom && dl->index < 0 && mpb->num_raid_devs)
7357 continue;
7358
05501181 7359 e = get_extents(super, dl, 0);
efb30e7f
DW		 7360 if (!e)
7361 continue;
7362 for (i = 1; e[i-1].size; i++)
7363 ;
7364 dl->e = e;
7365 dl->extent_cnt = i;
7366 extent_cnt += i;
7367 cnt++;
7368 }
7369
7370 maxsize = merge_extents(super, extent_cnt);
7371 minsize = size;
7372 if (size == 0)
612e59d8
CA		 7373 /* chunk is in K */
7374 minsize = chunk * 2;
efb30e7f
DW		 7375
7376 if (cnt < raiddisks ||
7377 (super->orom && used && used != raiddisks) ||
a7dd165b
DW		 7378 maxsize < minsize ||
7379 maxsize == 0) {
e7b84f9d 7380 pr_err("not enough devices with space to create array.\n");
efb30e7f
DW		 7381 return 0; /* No enough free spaces large enough */
7382 }
7383
7384 if (size == 0) {
7385 size = maxsize;
7386 if (chunk) {
612e59d8
CA		 7387 size /= 2 * chunk;
7388 size *= 2 * chunk;
efb30e7f 7389 }
f878b242
LM		 7390 maxsize = size;
7391 }
1a1ced1e
KS		 7392 if (mpb->num_raid_devs > 0 && size && size != maxsize)
7393 pr_err("attempting to create a second volume with size less then remaining space.\n");
efb30e7f
DW		 7394 cnt = 0;
7395 for (dl = super->disks; dl; dl = dl->next)
7396 if (dl->e)
7397 dl->raiddisk = cnt++;
7398
7399 *freesize = size;
7400
13bcac90
AK		 7401 dprintf("imsm: imsm_get_free_size() returns : %llu\n", size);
7402
efb30e7f
DW		 7403 return 1;
7404}
7405
13bcac90
AK		 7406static int reserve_space(struct supertype *st, int raiddisks,
7407 unsigned long long size, int chunk,
7408 unsigned long long *freesize)
7409{
7410 struct intel_super *super = st->sb;
7411 struct dl *dl;
7412 int cnt;
7413 int rv = 0;
7414
7415 rv = imsm_get_free_size(st, raiddisks, size, chunk, freesize);
7416 if (rv) {
7417 cnt = 0;
7418 for (dl = super->disks; dl; dl = dl->next)
7419 if (dl->e)
7420 dl->raiddisk = cnt++;
7421 rv = 1;
7422 }
7423
7424 return rv;
7425}
7426
bf5a934a 7427static int validate_geometry_imsm(struct supertype *st, int level, int layout,
c21e737b 7428 int raiddisks, int *chunk, unsigned long long size,
af4348dd 7429 unsigned long long data_offset,
bf5a934a 7430 char *dev, unsigned long long *freesize,
5308f117 7431 int consistency_policy, int verbose)
bf5a934a
DW		 7432{
7433 int fd, cfd;
7434 struct mdinfo *sra;
20cbe8d2 7435 int is_member = 0;
bf5a934a 7436
d54559f0
LM		 7437 /* load capability
7438 * if given unused devices create a container
bf5a934a
DW		 7439 * if given given devices in a container create a member volume
7440 */
7441 if (level == LEVEL_CONTAINER) {
7442 /* Must be a fresh device to add to a container */
7443 return validate_geometry_imsm_container(st, level, layout,
c21e737b 7444 raiddisks,
7ccc4cc4 7445 *chunk,
af4348dd 7446 size, data_offset,
bf5a934a
DW		 7447 dev, freesize,
7448 verbose);
7449 }
9587c373 7450
06a6101c
BK		 7451 /*
7452 * Size is given in sectors.
7453 */
7454 if (size && (size < 2048)) {
22dc741f 7455 pr_err("Given size must be greater than 1M.\n");
54865c30
RS		 7456 /* Depends on algorithm in Create.c :
7457 * if container was given (dev == NULL) return -1,
7458 * if block device was given ( dev != NULL) return 0.
7459 */
7460 return dev ? -1 : 0;
7461 }
7462
8592f29d 7463 if (!dev) {
e91a3bad 7464 if (st->sb) {
ca9de185 7465 struct intel_super *super = st->sb;
e91a3bad 7466 if (!validate_geometry_imsm_orom(st->sb, level, layout,
2cc699af 7467 raiddisks, chunk, size,
e91a3bad
LM		 7468 verbose))
7469 return 0;
efb30e7f
DW		 7470 /* we are being asked to automatically layout a
7471 * new volume based on the current contents of
7472 * the container. If the the parameters can be
7473 * satisfied reserve_space will record the disks,
7474 * start offset, and size of the volume to be
7475 * created. add_to_super and getinfo_super
7476 * detect when autolayout is in progress.
7477 */
ca9de185
LM		 7478 /* assuming that freesize is always given when array is
7479 created */
7480 if (super->orom && freesize) {
7481 int count;
72a45777 7482 count = count_volumes(super->hba,
ca9de185
LM		 7483 super->orom->dpa, verbose);
7484 if (super->orom->vphba <= count) {
676e87a8 7485 pr_vrb("platform does not support more than %d raid volumes.\n",
ca9de185
LM		 7486 super->orom->vphba);
7487 return 0;
7488 }
7489 }
e91a3bad
LM		 7490 if (freesize)
7491 return reserve_space(st, raiddisks, size,
7ccc4cc4 7492 *chunk, freesize);
8592f29d
N		 7493 }
7494 return 1;
7495 }
bf5a934a
DW		 7496 if (st->sb) {
7497 /* creating in a given container */
7498 return validate_geometry_imsm_volume(st, level, layout,
7499 raiddisks, chunk, size,
af4348dd 7500 data_offset,
bf5a934a
DW		 7501 dev, freesize, verbose);
7502 }
7503
bf5a934a
DW		 7504 /* This device needs to be a device in an 'imsm' container */
7505 fd = open(dev, O_RDONLY|O_EXCL, 0);
7506 if (fd >= 0) {
7507 if (verbose)
e7b84f9d
N		 7508 pr_err("Cannot create this array on device %s\n",
7509 dev);
bf5a934a
DW		 7510 close(fd);
7511 return 0;
7512 }
7513 if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) {
7514 if (verbose)
e7b84f9d 7515 pr_err("Cannot open %s: %s\n",
bf5a934a
DW		 7516 dev, strerror(errno));
7517 return 0;
7518 }
7519 /* Well, it is in use by someone, maybe an 'imsm' container. */
7520 cfd = open_container(fd);
20cbe8d2 7521 close(fd);
bf5a934a 7522 if (cfd < 0) {
bf5a934a 7523 if (verbose)
e7b84f9d 7524 pr_err("Cannot use %s: It is busy\n",
bf5a934a
DW		 7525 dev);
7526 return 0;
7527 }
4dd2df09 7528 sra = sysfs_read(cfd, NULL, GET_VERSION);
bf5a934a 7529 if (sra && sra->array.major_version == -1 &&
20cbe8d2
AW		 7530 strcmp(sra->text_version, "imsm") == 0)
7531 is_member = 1;
7532 sysfs_free(sra);
7533 if (is_member) {
bf5a934a
DW		 7534 /* This is a member of a imsm container. Load the container
7535 * and try to create a volume
7536 */
7537 struct intel_super *super;
7538
ec50f7b6 7539 if (load_super_imsm_all(st, cfd, (void **) &super, NULL, NULL, 1) == 0) {
bf5a934a 7540 st->sb = super;
4dd2df09 7541 strcpy(st->container_devnm, fd2devnm(cfd));
bf5a934a
DW		 7542 close(cfd);
7543 return validate_geometry_imsm_volume(st, level, layout,
7544 raiddisks, chunk,
af4348dd 7545 size, data_offset, dev,
ecbd9e81
N		 7546 freesize, 1)
7547 ? 1 : -1;
bf5a934a 7548 }
20cbe8d2 7549 }
bf5a934a 7550
20cbe8d2 7551 if (verbose)
e7b84f9d 7552 pr_err("failed container membership check\n");
20cbe8d2
AW		 7553
7554 close(cfd);
7555 return 0;
bf5a934a 7556}
0bd16cf2 7557
30f58b22 7558static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk)
0bd16cf2
DJ		 7559{
7560 struct intel_super *super = st->sb;
7561
30f58b22
DW		 7562 if (level && *level == UnSet)
7563 *level = LEVEL_CONTAINER;
7564
7565 if (level && layout && *layout == UnSet)
7566 *layout = imsm_level_to_layout(*level);
0bd16cf2 7567
cd9d1ac7
DW		 7568 if (chunk && (*chunk == UnSet || *chunk == 0))
7569 *chunk = imsm_default_chunk(super->orom);
0bd16cf2
DJ		 7570}
7571
33414a01
DW		 7572static void handle_missing(struct intel_super *super, struct imsm_dev *dev);
7573
3364781b 7574static int kill_subarray_imsm(struct supertype *st, char *subarray_id)
33414a01 7575{
3364781b 7576 /* remove the subarray currently referenced by subarray_id */
33414a01
DW		 7577 __u8 i;
7578 struct intel_dev **dp;
7579 struct intel_super *super = st->sb;
3364781b 7580 __u8 current_vol = strtoul(subarray_id, NULL, 10);
33414a01
DW		 7581 struct imsm_super *mpb = super->anchor;
7582
3364781b 7583 if (mpb->num_raid_devs == 0)
33414a01 7584 return 2;
33414a01
DW		 7585
7586 /* block deletions that would change the uuid of active subarrays
7587 *
7588 * FIXME when immutable ids are available, but note that we'll
7589 * also need to fixup the invalidated/active subarray indexes in
7590 * mdstat
7591 */
7592 for (i = 0; i < mpb->num_raid_devs; i++) {
7593 char subarray[4];
7594
7595 if (i < current_vol)
7596 continue;
7597 sprintf(subarray, "%u", i);
4dd2df09 7598 if (is_subarray_active(subarray, st->devnm)) {
e7b84f9d
N		 7599 pr_err("deleting subarray-%d would change the UUID of active subarray-%d, aborting\n",
7600 current_vol, i);
33414a01
DW		 7601
7602 return 2;
7603 }
7604 }
7605
7606 if (st->update_tail) {
503975b9 7607 struct imsm_update_kill_array *u = xmalloc(sizeof(*u));
33414a01 7608
33414a01
DW		 7609 u->type = update_kill_array;
7610 u->dev_idx = current_vol;
7611 append_metadata_update(st, u, sizeof(*u));
7612
7613 return 0;
7614 }
7615
7616 for (dp = &super->devlist; *dp;)
7617 if ((*dp)->index == current_vol) {
7618 *dp = (*dp)->next;
7619 } else {
7620 handle_missing(super, (*dp)->dev);
7621 if ((*dp)->index > current_vol)
7622 (*dp)->index--;
7623 dp = &(*dp)->next;
7624 }
7625
7626 /* no more raid devices, all active components are now spares,
7627 * but of course failed are still failed
7628 */
7629 if (--mpb->num_raid_devs == 0) {
7630 struct dl *d;
7631
7632 for (d = super->disks; d; d = d->next)
a8619d23
AK		 7633 if (d->index > -2)
7634 mark_spare(d);
33414a01
DW		 7635 }
7636
7637 super->updates_pending++;
7638
7639 return 0;
7640}
aa534678 7641
a951a4f7 7642static int update_subarray_imsm(struct supertype *st, char *subarray,
fa56eddb 7643 char *update, struct mddev_ident *ident)
aa534678
DW		 7644{
7645 /* update the subarray currently referenced by ->current_vol */
7646 struct intel_super *super = st->sb;
7647 struct imsm_super *mpb = super->anchor;
7648
aa534678
DW		 7649 if (strcmp(update, "name") == 0) {
7650 char *name = ident->name;
a951a4f7
N		 7651 char *ep;
7652 int vol;
aa534678 7653
4dd2df09 7654 if (is_subarray_active(subarray, st->devnm)) {
e7b84f9d 7655 pr_err("Unable to update name of active subarray\n");
aa534678
DW		 7656 return 2;
7657 }
7658
7659 if (!check_name(super, name, 0))
7660 return 2;
7661
a951a4f7
N		 7662 vol = strtoul(subarray, &ep, 10);
7663 if (*ep != '\0' || vol >= super->anchor->num_raid_devs)
7664 return 2;
7665
aa534678 7666 if (st->update_tail) {
503975b9 7667 struct imsm_update_rename_array *u = xmalloc(sizeof(*u));
aa534678 7668
aa534678 7669 u->type = update_rename_array;
a951a4f7 7670 u->dev_idx = vol;
618f4e6d
XN		 7671 strncpy((char *) u->name, name, MAX_RAID_SERIAL_LEN);
7672 u->name[MAX_RAID_SERIAL_LEN-1] = '\0';
aa534678
DW		 7673 append_metadata_update(st, u, sizeof(*u));
7674 } else {
7675 struct imsm_dev *dev;
ebad3af2 7676 int i, namelen;
aa534678 7677
a951a4f7 7678 dev = get_imsm_dev(super, vol);
ebad3af2
JS		 7679 memset(dev->volume, '\0', MAX_RAID_SERIAL_LEN);
7680 namelen = min((int)strlen(name), MAX_RAID_SERIAL_LEN);
7681 memcpy(dev->volume, name, namelen);
aa534678
DW		 7682 for (i = 0; i < mpb->num_raid_devs; i++) {
7683 dev = get_imsm_dev(super, i);
7684 handle_missing(super, dev);
7685 }
7686 super->updates_pending++;
7687 }
e6e9dd3f
AP		 7688 } else if (strcmp(update, "ppl") == 0 ||
7689 strcmp(update, "no-ppl") == 0) {
7690 int new_policy;
7691 char *ep;
7692 int vol = strtoul(subarray, &ep, 10);
7693
7694 if (*ep != '\0' || vol >= super->anchor->num_raid_devs)
7695 return 2;
7696
7697 if (strcmp(update, "ppl") == 0)
c2462068 7698 new_policy = RWH_MULTIPLE_DISTRIBUTED;
e6e9dd3f 7699 else
c2462068 7700 new_policy = RWH_MULTIPLE_OFF;
e6e9dd3f
AP		 7701
7702 if (st->update_tail) {
7703 struct imsm_update_rwh_policy *u = xmalloc(sizeof(*u));
7704
7705 u->type = update_rwh_policy;
7706 u->dev_idx = vol;
7707 u->new_policy = new_policy;
7708 append_metadata_update(st, u, sizeof(*u));
7709 } else {
7710 struct imsm_dev *dev;
7711
7712 dev = get_imsm_dev(super, vol);
7713 dev->rwh_policy = new_policy;
7714 super->updates_pending++;
7715 }
aa534678
DW		 7716 } else
7717 return 2;
7718
7719 return 0;
7720}
bf5a934a 7721
28bce06f
AK		 7722static int is_gen_migration(struct imsm_dev *dev)
7723{
7534230b
AK		 7724 if (dev == NULL)
7725 return 0;
7726
28bce06f
AK		 7727 if (!dev->vol.migr_state)
7728 return 0;
7729
7730 if (migr_type(dev) == MIGR_GEN_MIGR)
7731 return 1;
7732
7733 return 0;
7734}
7735
1e5c6983
DW		 7736static int is_rebuilding(struct imsm_dev *dev)
7737{
7738 struct imsm_map *migr_map;
7739
7740 if (!dev->vol.migr_state)
7741 return 0;
7742
7743 if (migr_type(dev) != MIGR_REBUILD)
7744 return 0;
7745
238c0a71 7746 migr_map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 7747
7748 if (migr_map->map_state == IMSM_T_STATE_DEGRADED)
7749 return 1;
7750 else
7751 return 0;
7752}
7753
6ce1fbf1
AK		 7754static int is_initializing(struct imsm_dev *dev)
7755{
7756 struct imsm_map *migr_map;
7757
7758 if (!dev->vol.migr_state)
7759 return 0;
7760
7761 if (migr_type(dev) != MIGR_INIT)
7762 return 0;
7763
238c0a71 7764 migr_map = get_imsm_map(dev, MAP_1);
6ce1fbf1
AK		 7765
7766 if (migr_map->map_state == IMSM_T_STATE_UNINITIALIZED)
7767 return 1;
7768
7769 return 0;
6ce1fbf1
AK		 7770}
7771
c47b0ff6
AK		 7772static void update_recovery_start(struct intel_super *super,
7773 struct imsm_dev *dev,
7774 struct mdinfo *array)
1e5c6983
DW		 7775{
7776 struct mdinfo *rebuild = NULL;
7777 struct mdinfo *d;
7778 __u32 units;
7779
7780 if (!is_rebuilding(dev))
7781 return;
7782
7783 /* Find the rebuild target, but punt on the dual rebuild case */
7784 for (d = array->devs; d; d = d->next)
7785 if (d->recovery_start == 0) {
7786 if (rebuild)
7787 return;
7788 rebuild = d;
7789 }
7790
4363fd80
DW		 7791 if (!rebuild) {
7792 /* (?) none of the disks are marked with
7793 * IMSM_ORD_REBUILD, so assume they are missing and the
7794 * disk_ord_tbl was not correctly updated
7795 */
1ade5cc1 7796 dprintf("failed to locate out-of-sync disk\n");
4363fd80
DW		 7797 return;
7798 }
7799
1e5c6983 7800 units = __le32_to_cpu(dev->vol.curr_migr_unit);
c47b0ff6 7801 rebuild->recovery_start = units * blocks_per_migr_unit(super, dev);
1e5c6983
DW		 7802}
7803
276d77db 7804static int recover_backup_imsm(struct supertype *st, struct mdinfo *info);
1e5c6983 7805
00bbdbda 7806static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray)
cdddbdbc 7807{
4f5bc454
DW		 7808 /* Given a container loaded by load_super_imsm_all,
7809 * extract information about all the arrays into
7810 * an mdinfo tree.
00bbdbda 7811 * If 'subarray' is given, just extract info about that array.
4f5bc454
DW		 7812 *
7813 * For each imsm_dev create an mdinfo, fill it in,
7814 * then look for matching devices in super->disks
7815 * and create appropriate device mdinfo.
7816 */
7817 struct intel_super *super = st->sb;
949c47a0 7818 struct imsm_super *mpb = super->anchor;
4f5bc454 7819 struct mdinfo *rest = NULL;
00bbdbda 7820 unsigned int i;
81219e70 7821 int sb_errors = 0;
abef11a3
AK		 7822 struct dl *d;
7823 int spare_disks = 0;
b6180160 7824 int current_vol = super->current_vol;
cdddbdbc 7825
19482bcc
AK		 7826 /* do not assemble arrays when not all attributes are supported */
7827 if (imsm_check_attributes(mpb->attributes) == 0) {
81219e70 7828 sb_errors = 1;
7a862a02 7829 pr_err("Unsupported attributes in IMSM metadata.Arrays activation is blocked.\n");
19482bcc
AK		 7830 }
7831
abef11a3
AK		 7832 /* count spare devices, not used in maps
7833 */
7834 for (d = super->disks; d; d = d->next)
7835 if (d->index == -1)
7836 spare_disks++;
7837
4f5bc454 7838 for (i = 0; i < mpb->num_raid_devs; i++) {
00bbdbda
N		 7839 struct imsm_dev *dev;
7840 struct imsm_map *map;
86e3692b 7841 struct imsm_map *map2;
4f5bc454 7842 struct mdinfo *this;
a6482415 7843 int slot;
a6482415 7844 int chunk;
00bbdbda 7845 char *ep;
8b9cd157 7846 int level;
00bbdbda
N		 7847
7848 if (subarray &&
7849 (i != strtoul(subarray, &ep, 10) || *ep != '\0'))
7850 continue;
7851
7852 dev = get_imsm_dev(super, i);
238c0a71
AK		 7853 map = get_imsm_map(dev, MAP_0);
7854 map2 = get_imsm_map(dev, MAP_1);
8b9cd157 7855 level = get_imsm_raid_level(map);
4f5bc454 7856
1ce0101c
DW		 7857 /* do not publish arrays that are in the middle of an
7858 * unsupported migration
7859 */
7860 if (dev->vol.migr_state &&
28bce06f 7861 (migr_type(dev) == MIGR_STATE_CHANGE)) {
7a862a02 7862 pr_err("cannot assemble volume '%.16s': unsupported migration in progress\n",
1ce0101c
DW		 7863 dev->volume);
7864 continue;
7865 }
2db86302
LM		 7866 /* do not publish arrays that are not support by controller's
7867 * OROM/EFI
7868 */
1ce0101c 7869
503975b9 7870 this = xmalloc(sizeof(*this));
4f5bc454 7871
301406c9 7872 super->current_vol = i;
a5d85af7 7873 getinfo_super_imsm_volume(st, this, NULL);
9894ec0d 7874 this->next = rest;
a6482415 7875 chunk = __le16_to_cpu(map->blocks_per_strip) >> 1;
81219e70
LM		 7876 /* mdadm does not support all metadata features- set the bit in all arrays state */
7877 if (!validate_geometry_imsm_orom(super,
8b9cd157
MK		 7878 level, /* RAID level */
7879 imsm_level_to_layout(level),
81219e70 7880 map->num_members, /* raid disks */
fcc2c9da 7881 &chunk, imsm_dev_size(dev),
81219e70 7882 1 /* verbose */)) {
7a862a02 7883 pr_err("IMSM RAID geometry validation failed. Array %s activation is blocked.\n",
81219e70
LM		 7884 dev->volume);
7885 this->array.state |=
7886 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
7887 (1<<MD_SB_BLOCK_VOLUME);
7888 }
81219e70
LM		 7889
7890 /* if array has bad blocks, set suitable bit in all arrays state */
7891 if (sb_errors)
7892 this->array.state |=
7893 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
7894 (1<<MD_SB_BLOCK_VOLUME);
7895
4f5bc454 7896 for (slot = 0 ; slot < map->num_members; slot++) {
1e5c6983 7897 unsigned long long recovery_start;
4f5bc454
DW		 7898 struct mdinfo *info_d;
7899 struct dl *d;
7900 int idx;
9a1608e5 7901 int skip;
7eef0453 7902 __u32 ord;
8b9cd157 7903 int missing = 0;
4f5bc454 7904
9a1608e5 7905 skip = 0;
238c0a71
AK		 7906 idx = get_imsm_disk_idx(dev, slot, MAP_0);
7907 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
4f5bc454
DW		 7908 for (d = super->disks; d ; d = d->next)
7909 if (d->index == idx)
0fbd635c 7910 break;
4f5bc454 7911
1e5c6983 7912 recovery_start = MaxSector;
4f5bc454 7913 if (d == NULL)
9a1608e5 7914 skip = 1;
25ed7e59 7915 if (d && is_failed(&d->disk))
9a1608e5 7916 skip = 1;
8b9cd157 7917 if (!skip && (ord & IMSM_ORD_REBUILD))
1e5c6983 7918 recovery_start = 0;
1e93d0d1
BK		 7919 if (!(ord & IMSM_ORD_REBUILD))
7920 this->array.working_disks++;
1011e834 7921 /*
9a1608e5 7922 * if we skip some disks the array will be assmebled degraded;
1e5c6983
DW		 7923 * reset resync start to avoid a dirty-degraded
7924 * situation when performing the intial sync
9a1608e5 7925 */
8b9cd157
MK		 7926 if (skip)
7927 missing++;
7928
7929 if (!(dev->vol.dirty & RAIDVOL_DIRTY)) {
7930 if ((!able_to_resync(level, missing) ||
7931 recovery_start == 0))
7932 this->resync_start = MaxSector;
7933 } else {
7934 /*
7935 * FIXME handle dirty degraded
7936 */
7937 }
7938
9a1608e5
DW		 7939 if (skip)
7940 continue;
4f5bc454 7941
503975b9 7942 info_d = xcalloc(1, sizeof(*info_d));
4f5bc454
DW		 7943 info_d->next = this->devs;
7944 this->devs = info_d;
7945
4f5bc454
DW		 7946 info_d->disk.number = d->index;
7947 info_d->disk.major = d->major;
7948 info_d->disk.minor = d->minor;
7949 info_d->disk.raid_disk = slot;
1e5c6983 7950 info_d->recovery_start = recovery_start;
86e3692b
AK		 7951 if (map2) {
7952 if (slot < map2->num_members)
7953 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK		 7954 else
7955 this->array.spare_disks++;
86e3692b
AK		 7956 } else {
7957 if (slot < map->num_members)
7958 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK		 7959 else
7960 this->array.spare_disks++;
86e3692b 7961 }
4f5bc454
DW		 7962
7963 info_d->events = __le32_to_cpu(mpb->generation_num);
5551b113 7964 info_d->data_offset = pba_of_lba0(map);
44490938 7965 info_d->component_size = calc_component_size(map, dev);
06fb291a
PB		 7966
7967 if (map->raid_level == 5) {
2432ce9b
AP		 7968 info_d->ppl_sector = this->ppl_sector;
7969 info_d->ppl_size = this->ppl_size;
98e96bdb
AP		 7970 if (this->consistency_policy == CONSISTENCY_POLICY_PPL &&
7971 recovery_start == 0)
7972 this->resync_start = 0;
06fb291a 7973 }
b12796be 7974
5e46202e 7975 info_d->bb.supported = 1;
b12796be
TM		 7976 get_volume_badblocks(super->bbm_log, ord_to_idx(ord),
7977 info_d->data_offset,
7978 info_d->component_size,
7979 &info_d->bb);
4f5bc454 7980 }
1e5c6983 7981 /* now that the disk list is up-to-date fixup recovery_start */
c47b0ff6 7982 update_recovery_start(super, dev, this);
abef11a3 7983 this->array.spare_disks += spare_disks;
276d77db
AK		 7984
7985 /* check for reshape */
7986 if (this->reshape_active == 1)
7987 recover_backup_imsm(st, this);
9a1608e5 7988 rest = this;
4f5bc454
DW		 7989 }
7990
b6180160 7991 super->current_vol = current_vol;
4f5bc454 7992 return rest;
cdddbdbc
DW		 7993}
7994
3b451610
AK		 7995static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev,
7996 int failed, int look_in_map)
c2a1e7da 7997{
3b451610
AK		 7998 struct imsm_map *map;
7999
8000 map = get_imsm_map(dev, look_in_map);
c2a1e7da
DW		 8001
8002 if (!failed)
1011e834 8003 return map->map_state == IMSM_T_STATE_UNINITIALIZED ?
3393c6af 8004 IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL;
c2a1e7da
DW		 8005
8006 switch (get_imsm_raid_level(map)) {
8007 case 0:
8008 return IMSM_T_STATE_FAILED;
8009 break;
8010 case 1:
8011 if (failed < map->num_members)
8012 return IMSM_T_STATE_DEGRADED;
8013 else
8014 return IMSM_T_STATE_FAILED;
8015 break;
8016 case 10:
8017 {
8018 /**
c92a2527
DW		 8019 * check to see if any mirrors have failed, otherwise we
8020 * are degraded. Even numbered slots are mirrored on
8021 * slot+1
c2a1e7da 8022 */
c2a1e7da 8023 int i;
d9b420a5
N		 8024 /* gcc -Os complains that this is unused */
8025 int insync = insync;
c2a1e7da
DW		 8026
8027 for (i = 0; i < map->num_members; i++) {
238c0a71 8028 __u32 ord = get_imsm_ord_tbl_ent(dev, i, MAP_X);
c92a2527
DW		 8029 int idx = ord_to_idx(ord);
8030 struct imsm_disk *disk;
c2a1e7da 8031
c92a2527 8032 /* reset the potential in-sync count on even-numbered
1011e834 8033 * slots. num_copies is always 2 for imsm raid10
c92a2527
DW		 8034 */
8035 if ((i & 1) == 0)
8036 insync = 2;
c2a1e7da 8037
c92a2527 8038 disk = get_imsm_disk(super, idx);
25ed7e59 8039 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
c92a2527 8040 insync--;
c2a1e7da 8041
c92a2527
DW		 8042 /* no in-sync disks left in this mirror the
8043 * array has failed
8044 */
8045 if (insync == 0)
8046 return IMSM_T_STATE_FAILED;
c2a1e7da
DW		 8047 }
8048
8049 return IMSM_T_STATE_DEGRADED;
8050 }
8051 case 5:
8052 if (failed < 2)
8053 return IMSM_T_STATE_DEGRADED;
8054 else
8055 return IMSM_T_STATE_FAILED;
8056 break;
8057 default:
8058 break;
8059 }
8060
8061 return map->map_state;
8062}
8063
3b451610
AK		 8064static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev,
8065 int look_in_map)
c2a1e7da
DW		 8066{
8067 int i;
8068 int failed = 0;
8069 struct imsm_disk *disk;
d5985138
AK		 8070 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8071 struct imsm_map *prev = get_imsm_map(dev, MAP_1);
68fe4598 8072 struct imsm_map *map_for_loop;
0556e1a2
DW		 8073 __u32 ord;
8074 int idx;
d5985138 8075 int idx_1;
c2a1e7da 8076
0556e1a2
DW		 8077 /* at the beginning of migration we set IMSM_ORD_REBUILD on
8078 * disks that are being rebuilt. New failures are recorded to
8079 * map[0]. So we look through all the disks we started with and
8080 * see if any failures are still present, or if any new ones
8081 * have arrived
0556e1a2 8082 */
d5985138
AK		 8083 map_for_loop = map;
8084 if (prev && (map->num_members < prev->num_members))
8085 map_for_loop = prev;
68fe4598
LD		 8086
8087 for (i = 0; i < map_for_loop->num_members; i++) {
d5985138 8088 idx_1 = -255;
238c0a71
AK		 8089 /* when MAP_X is passed both maps failures are counted
8090 */
d5985138 8091 if (prev &&
089f9d79
JS		 8092 (look_in_map == MAP_1 || look_in_map == MAP_X) &&
8093 i < prev->num_members) {
d5985138
AK		 8094 ord = __le32_to_cpu(prev->disk_ord_tbl[i]);
8095 idx_1 = ord_to_idx(ord);
c2a1e7da 8096
d5985138
AK		 8097 disk = get_imsm_disk(super, idx_1);
8098 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
8099 failed++;
8100 }
089f9d79
JS		 8101 if ((look_in_map == MAP_0 || look_in_map == MAP_X) &&
8102 i < map->num_members) {
d5985138
AK		 8103 ord = __le32_to_cpu(map->disk_ord_tbl[i]);
8104 idx = ord_to_idx(ord);
8105
8106 if (idx != idx_1) {
8107 disk = get_imsm_disk(super, idx);
8108 if (!disk || is_failed(disk) ||
8109 ord & IMSM_ORD_REBUILD)
8110 failed++;
8111 }
8112 }
c2a1e7da
DW		 8113 }
8114
8115 return failed;
845dea95
NB		 8116}
8117
97b4d0e9
DW		 8118static int imsm_open_new(struct supertype *c, struct active_array *a,
8119 char *inst)
8120{
8121 struct intel_super *super = c->sb;
8122 struct imsm_super *mpb = super->anchor;
bbab0940 8123 struct imsm_update_prealloc_bb_mem u;
9587c373 8124
97b4d0e9 8125 if (atoi(inst) >= mpb->num_raid_devs) {
1ade5cc1 8126 pr_err("subarry index %d, out of range\n", atoi(inst));
97b4d0e9
DW		 8127 return -ENODEV;
8128 }
8129
8130 dprintf("imsm: open_new %s\n", inst);
8131 a->info.container_member = atoi(inst);
bbab0940
TM		 8132
8133 u.type = update_prealloc_badblocks_mem;
8134 imsm_update_metadata_locally(c, &u, sizeof(u));
8135
97b4d0e9
DW		 8136 return 0;
8137}
8138
0c046afd
DW		 8139static int is_resyncing(struct imsm_dev *dev)
8140{
8141 struct imsm_map *migr_map;
8142
8143 if (!dev->vol.migr_state)
8144 return 0;
8145
1484e727
DW		 8146 if (migr_type(dev) == MIGR_INIT ||
8147 migr_type(dev) == MIGR_REPAIR)
0c046afd
DW		 8148 return 1;
8149
4c9bc37b
AK		 8150 if (migr_type(dev) == MIGR_GEN_MIGR)
8151 return 0;
8152
238c0a71 8153 migr_map = get_imsm_map(dev, MAP_1);
0c046afd 8154
089f9d79
JS		 8155 if (migr_map->map_state == IMSM_T_STATE_NORMAL &&
8156 dev->vol.migr_type != MIGR_GEN_MIGR)
0c046afd
DW		 8157 return 1;
8158 else
8159 return 0;
8160}
8161
0556e1a2 8162/* return true if we recorded new information */
4c9e8c1e
TM		 8163static int mark_failure(struct intel_super *super,
8164 struct imsm_dev *dev, struct imsm_disk *disk, int idx)
47ee5a45 8165{
0556e1a2
DW		 8166 __u32 ord;
8167 int slot;
8168 struct imsm_map *map;
86c54047
DW		 8169 char buf[MAX_RAID_SERIAL_LEN+3];
8170 unsigned int len, shift = 0;
0556e1a2
DW		 8171
8172 /* new failures are always set in map[0] */
238c0a71 8173 map = get_imsm_map(dev, MAP_0);
0556e1a2
DW		 8174
8175 slot = get_imsm_disk_slot(map, idx);
8176 if (slot < 0)
8177 return 0;
8178
8179 ord = __le32_to_cpu(map->disk_ord_tbl[slot]);
25ed7e59 8180 if (is_failed(disk) && (ord & IMSM_ORD_REBUILD))
0556e1a2
DW		 8181 return 0;
8182
7d0c5e24
LD		 8183 memcpy(buf, disk->serial, MAX_RAID_SERIAL_LEN);
8184 buf[MAX_RAID_SERIAL_LEN] = '\000';
8185 strcat(buf, ":0");
86c54047
DW		 8186 if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN)
8187 shift = len - MAX_RAID_SERIAL_LEN + 1;
167d8bb8 8188 memcpy(disk->serial, &buf[shift], len + 1 - shift);
86c54047 8189
f2f27e63 8190 disk->status |= FAILED_DISK;
0556e1a2 8191 set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD);
17788994
AK		 8192 /* mark failures in second map if second map exists and this disk
8193 * in this slot.
8194 * This is valid for migration, initialization and rebuild
8195 */
8196 if (dev->vol.migr_state) {
238c0a71 8197 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
0a108d63
AK		 8198 int slot2 = get_imsm_disk_slot(map2, idx);
8199
089f9d79 8200 if (slot2 < map2->num_members && slot2 >= 0)
0a108d63 8201 set_imsm_ord_tbl_ent(map2, slot2,
1ace8403
AK		 8202 idx | IMSM_ORD_REBUILD);
8203 }
d7a1fda2
MT		 8204 if (map->failed_disk_num == 0xff ||
8205 (!is_rebuilding(dev) && map->failed_disk_num > slot))
0556e1a2 8206 map->failed_disk_num = slot;
4c9e8c1e
TM		 8207
8208 clear_disk_badblocks(super->bbm_log, ord_to_idx(ord));
8209
0556e1a2
DW		 8210 return 1;
8211}
8212
4c9e8c1e
TM		 8213static void mark_missing(struct intel_super *super,
8214 struct imsm_dev *dev, struct imsm_disk *disk, int idx)
0556e1a2 8215{
4c9e8c1e 8216 mark_failure(super, dev, disk, idx);
0556e1a2
DW		 8217
8218 if (disk->scsi_id == __cpu_to_le32(~(__u32)0))
8219 return;
8220
47ee5a45
DW		 8221 disk->scsi_id = __cpu_to_le32(~(__u32)0);
8222 memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1);
8223}
8224
33414a01
DW		 8225static void handle_missing(struct intel_super *super, struct imsm_dev *dev)
8226{
33414a01 8227 struct dl *dl;
33414a01
DW		 8228
8229 if (!super->missing)
8230 return;
33414a01 8231
79b68f1b
PC		 8232 /* When orom adds replacement for missing disk it does
8233 * not remove entry of missing disk, but just updates map with
8234 * new added disk. So it is not enough just to test if there is
8235 * any missing disk, we have to look if there are any failed disks
8236 * in map to stop migration */
8237
33414a01 8238 dprintf("imsm: mark missing\n");
3d59f0c0
AK		 8239 /* end process for initialization and rebuild only
8240 */
8241 if (is_gen_migration(dev) == 0) {
fb12a745 8242 int failed = imsm_count_failed(super, dev, MAP_0);
3d59f0c0 8243
fb12a745
TM		 8244 if (failed) {
8245 __u8 map_state;
8246 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8247 struct imsm_map *map1;
8248 int i, ord, ord_map1;
8249 int rebuilt = 1;
3d59f0c0 8250
fb12a745
TM		 8251 for (i = 0; i < map->num_members; i++) {
8252 ord = get_imsm_ord_tbl_ent(dev, i, MAP_0);
8253 if (!(ord & IMSM_ORD_REBUILD))
8254 continue;
8255
8256 map1 = get_imsm_map(dev, MAP_1);
8257 if (!map1)
8258 continue;
8259
8260 ord_map1 = __le32_to_cpu(map1->disk_ord_tbl[i]);
8261 if (ord_map1 & IMSM_ORD_REBUILD)
8262 rebuilt = 0;
8263 }
8264
8265 if (rebuilt) {
8266 map_state = imsm_check_degraded(super, dev,
8267 failed, MAP_0);
8268 end_migration(dev, super, map_state);
8269 }
8270 }
3d59f0c0 8271 }
33414a01 8272 for (dl = super->missing; dl; dl = dl->next)
4c9e8c1e 8273 mark_missing(super, dev, &dl->disk, dl->index);
33414a01
DW		 8274 super->updates_pending++;
8275}
8276
f3871fdc
AK		 8277static unsigned long long imsm_set_array_size(struct imsm_dev *dev,
8278 long long new_size)
70bdf0dc 8279{
70bdf0dc 8280 unsigned long long array_blocks;
9529d343
MD		 8281 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8282 int used_disks = imsm_num_data_members(map);
70bdf0dc
AK		 8283
8284 if (used_disks == 0) {
8285 /* when problems occures
8286 * return current array_blocks value
8287 */
fcc2c9da 8288 array_blocks = imsm_dev_size(dev);
70bdf0dc
AK		 8289
8290 return array_blocks;
8291 }
8292
8293 /* set array size in metadata
8294 */
9529d343 8295 if (new_size <= 0)
f3871fdc
AK		 8296 /* OLCE size change is caused by added disks
8297 */
44490938 8298 array_blocks = per_dev_array_size(map) * used_disks;
9529d343 8299 else
f3871fdc
AK		 8300 /* Online Volume Size Change
8301 * Using available free space
8302 */
8303 array_blocks = new_size;
70bdf0dc 8304
b53bfba6 8305 array_blocks = round_size_to_mb(array_blocks, used_disks);
fcc2c9da 8306 set_imsm_dev_size(dev, array_blocks);
70bdf0dc
AK		 8307
8308 return array_blocks;
8309}
8310
28bce06f
AK		 8311static void imsm_set_disk(struct active_array *a, int n, int state);
8312
0e2d1a4e
AK		 8313static void imsm_progress_container_reshape(struct intel_super *super)
8314{
8315 /* if no device has a migr_state, but some device has a
8316 * different number of members than the previous device, start
8317 * changing the number of devices in this device to match
8318 * previous.
8319 */
8320 struct imsm_super *mpb = super->anchor;
8321 int prev_disks = -1;
8322 int i;
1dfaa380 8323 int copy_map_size;
0e2d1a4e
AK		 8324
8325 for (i = 0; i < mpb->num_raid_devs; i++) {
8326 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 8327 struct imsm_map *map = get_imsm_map(dev, MAP_0);
0e2d1a4e
AK		 8328 struct imsm_map *map2;
8329 int prev_num_members;
0e2d1a4e
AK		 8330
8331 if (dev->vol.migr_state)
8332 return;
8333
8334 if (prev_disks == -1)
8335 prev_disks = map->num_members;
8336 if (prev_disks == map->num_members)
8337 continue;
8338
8339 /* OK, this array needs to enter reshape mode.
8340 * i.e it needs a migr_state
8341 */
8342
1dfaa380 8343 copy_map_size = sizeof_imsm_map(map);
0e2d1a4e
AK		 8344 prev_num_members = map->num_members;
8345 map->num_members = prev_disks;
8346 dev->vol.migr_state = 1;
8347 dev->vol.curr_migr_unit = 0;
ea672ee1 8348 set_migr_type(dev, MIGR_GEN_MIGR);
0e2d1a4e
AK		 8349 for (i = prev_num_members;
8350 i < map->num_members; i++)
8351 set_imsm_ord_tbl_ent(map, i, i);
238c0a71 8352 map2 = get_imsm_map(dev, MAP_1);
0e2d1a4e 8353 /* Copy the current map */
1dfaa380 8354 memcpy(map2, map, copy_map_size);
0e2d1a4e
AK		 8355 map2->num_members = prev_num_members;
8356
f3871fdc 8357 imsm_set_array_size(dev, -1);
51d83f5d 8358 super->clean_migration_record_by_mdmon = 1;
0e2d1a4e
AK		 8359 super->updates_pending++;
8360 }
8361}
8362
aad6f216 8363/* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild
0c046afd
DW		 8364 * states are handled in imsm_set_disk() with one exception, when a
8365 * resync is stopped due to a new failure this routine will set the
8366 * 'degraded' state for the array.
8367 */
01f157d7 8368static int imsm_set_array_state(struct active_array *a, int consistent)
a862209d
DW		 8369{
8370 int inst = a->info.container_member;
8371 struct intel_super *super = a->container->sb;
949c47a0 8372 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8373 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3b451610
AK		 8374 int failed = imsm_count_failed(super, dev, MAP_0);
8375 __u8 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
1e5c6983 8376 __u32 blocks_per_unit;
a862209d 8377
1af97990
AK		 8378 if (dev->vol.migr_state &&
8379 dev->vol.migr_type == MIGR_GEN_MIGR) {
8380 /* array state change is blocked due to reshape action
aad6f216
N		 8381 * We might need to
8382 * - abort the reshape (if last_checkpoint is 0 and action!= reshape)
8383 * - finish the reshape (if last_checkpoint is big and action != reshape)
8384 * - update curr_migr_unit
1af97990 8385 */
aad6f216
N		 8386 if (a->curr_action == reshape) {
8387 /* still reshaping, maybe update curr_migr_unit */
633b5610 8388 goto mark_checkpoint;
aad6f216
N		 8389 } else {
8390 if (a->last_checkpoint == 0 && a->prev_action == reshape) {
8391 /* for some reason we aborted the reshape.
b66e591b
AK		 8392 *
8393 * disable automatic metadata rollback
8394 * user action is required to recover process
aad6f216 8395 */
b66e591b 8396 if (0) {
238c0a71
AK		 8397 struct imsm_map *map2 =
8398 get_imsm_map(dev, MAP_1);
8399 dev->vol.migr_state = 0;
8400 set_migr_type(dev, 0);
8401 dev->vol.curr_migr_unit = 0;
8402 memcpy(map, map2,
8403 sizeof_imsm_map(map2));
8404 super->updates_pending++;
b66e591b 8405 }
aad6f216
N		 8406 }
8407 if (a->last_checkpoint >= a->info.component_size) {
8408 unsigned long long array_blocks;
8409 int used_disks;
e154ced3 8410 struct mdinfo *mdi;
aad6f216 8411
9529d343 8412 used_disks = imsm_num_data_members(map);
d55adef9
AK		 8413 if (used_disks > 0) {
8414 array_blocks =
44490938 8415 per_dev_array_size(map) *
d55adef9 8416 used_disks;
b53bfba6
TM		 8417 array_blocks =
8418 round_size_to_mb(array_blocks,
8419 used_disks);
d55adef9
AK		 8420 a->info.custom_array_size = array_blocks;
8421 /* encourage manager to update array
8422 * size
8423 */
e154ced3 8424
d55adef9 8425 a->check_reshape = 1;
633b5610 8426 }
e154ced3
AK		 8427 /* finalize online capacity expansion/reshape */
8428 for (mdi = a->info.devs; mdi; mdi = mdi->next)
8429 imsm_set_disk(a,
8430 mdi->disk.raid_disk,
8431 mdi->curr_state);
8432
0e2d1a4e 8433 imsm_progress_container_reshape(super);
e154ced3 8434 }
aad6f216 8435 }
1af97990
AK		 8436 }
8437
47ee5a45 8438 /* before we activate this array handle any missing disks */
33414a01
DW		 8439 if (consistent == 2)
8440 handle_missing(super, dev);
1e5c6983 8441
0c046afd 8442 if (consistent == 2 &&
b7941fd6 8443 (!is_resync_complete(&a->info) ||
0c046afd
DW		 8444 map_state != IMSM_T_STATE_NORMAL ||
8445 dev->vol.migr_state))
01f157d7 8446 consistent = 0;
272906ef 8447
b7941fd6 8448 if (is_resync_complete(&a->info)) {
0c046afd 8449 /* complete intialization / resync,
0556e1a2
DW		 8450 * recovery and interrupted recovery is completed in
8451 * ->set_disk
0c046afd
DW		 8452 */
8453 if (is_resyncing(dev)) {
8454 dprintf("imsm: mark resync done\n");
809da78e 8455 end_migration(dev, super, map_state);
115c3803 8456 super->updates_pending++;
484240d8 8457 a->last_checkpoint = 0;
115c3803 8458 }
b9172665
AK		 8459 } else if ((!is_resyncing(dev) && !failed) &&
8460 (imsm_reshape_blocks_arrays_changes(super) == 0)) {
0c046afd 8461 /* mark the start of the init process if nothing is failed */
b7941fd6 8462 dprintf("imsm: mark resync start\n");
1484e727 8463 if (map->map_state == IMSM_T_STATE_UNINITIALIZED)
8e59f3d8 8464 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT);
1484e727 8465 else
8e59f3d8 8466 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR);
3393c6af 8467 super->updates_pending++;
115c3803 8468 }
a862209d 8469
633b5610 8470mark_checkpoint:
5b83bacf
AK		 8471 /* skip checkpointing for general migration,
8472 * it is controlled in mdadm
8473 */
8474 if (is_gen_migration(dev))
8475 goto skip_mark_checkpoint;
8476
1e5c6983 8477 /* check if we can update curr_migr_unit from resync_start, recovery_start */
c47b0ff6 8478 blocks_per_unit = blocks_per_migr_unit(super, dev);
4f0a7acc 8479 if (blocks_per_unit) {
1e5c6983
DW		 8480 __u32 units32;
8481 __u64 units;
8482
4f0a7acc 8483 units = a->last_checkpoint / blocks_per_unit;
1e5c6983
DW		 8484 units32 = units;
8485
8486 /* check that we did not overflow 32-bits, and that
8487 * curr_migr_unit needs updating
8488 */
8489 if (units32 == units &&
bfd80a56 8490 units32 != 0 &&
1e5c6983
DW		 8491 __le32_to_cpu(dev->vol.curr_migr_unit) != units32) {
8492 dprintf("imsm: mark checkpoint (%u)\n", units32);
8493 dev->vol.curr_migr_unit = __cpu_to_le32(units32);
8494 super->updates_pending++;
8495 }
8496 }
f8f603f1 8497
5b83bacf 8498skip_mark_checkpoint:
3393c6af 8499 /* mark dirty / clean */
2432ce9b
AP		 8500 if (((dev->vol.dirty & RAIDVOL_DIRTY) && consistent) ||
8501 (!(dev->vol.dirty & RAIDVOL_DIRTY) && !consistent)) {
b7941fd6 8502 dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty");
2432ce9b
AP		 8503 if (consistent) {
8504 dev->vol.dirty = RAIDVOL_CLEAN;
8505 } else {
8506 dev->vol.dirty = RAIDVOL_DIRTY;
c2462068
PB		 8507 if (dev->rwh_policy == RWH_DISTRIBUTED ||
8508 dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
2432ce9b
AP		 8509 dev->vol.dirty |= RAIDVOL_DSRECORD_VALID;
8510 }
a862209d
DW		 8511 super->updates_pending++;
8512 }
28bce06f 8513
01f157d7 8514 return consistent;
a862209d
DW		 8515}
8516
6f50473f
TM		 8517static int imsm_disk_slot_to_ord(struct active_array *a, int slot)
8518{
8519 int inst = a->info.container_member;
8520 struct intel_super *super = a->container->sb;
8521 struct imsm_dev *dev = get_imsm_dev(super, inst);
8522 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8523
8524 if (slot > map->num_members) {
8525 pr_err("imsm: imsm_disk_slot_to_ord %d out of range 0..%d\n",
8526 slot, map->num_members - 1);
8527 return -1;
8528 }
8529
8530 if (slot < 0)
8531 return -1;
8532
8533 return get_imsm_ord_tbl_ent(dev, slot, MAP_0);
8534}
8535
8d45d196 8536static void imsm_set_disk(struct active_array *a, int n, int state)
845dea95 8537{
8d45d196
DW		 8538 int inst = a->info.container_member;
8539 struct intel_super *super = a->container->sb;
949c47a0 8540 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8541 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8d45d196 8542 struct imsm_disk *disk;
7ce05701
LD		 8543 struct mdinfo *mdi;
8544 int recovery_not_finished = 0;
0c046afd 8545 int failed;
6f50473f 8546 int ord;
0c046afd 8547 __u8 map_state;
fb12a745
TM		 8548 int rebuild_done = 0;
8549 int i;
8d45d196 8550
fb12a745 8551 ord = get_imsm_ord_tbl_ent(dev, n, MAP_X);
6f50473f 8552 if (ord < 0)
8d45d196
DW		 8553 return;
8554
4e6e574a 8555 dprintf("imsm: set_disk %d:%x\n", n, state);
b10b37b8 8556 disk = get_imsm_disk(super, ord_to_idx(ord));
8d45d196 8557
5802a811 8558 /* check for new failures */
ae7d61e3 8559 if (disk && (state & DS_FAULTY)) {
4c9e8c1e 8560 if (mark_failure(super, dev, disk, ord_to_idx(ord)))
0556e1a2 8561 super->updates_pending++;
8d45d196 8562 }
47ee5a45 8563
19859edc 8564 /* check if in_sync */
0556e1a2 8565 if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) {
238c0a71 8566 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
b10b37b8
DW		 8567
8568 set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord));
fb12a745 8569 rebuild_done = 1;
19859edc
DW		 8570 super->updates_pending++;
8571 }
8d45d196 8572
3b451610
AK		 8573 failed = imsm_count_failed(super, dev, MAP_0);
8574 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
5802a811 8575
0c046afd 8576 /* check if recovery complete, newly degraded, or failed */
94002678
AK		 8577 dprintf("imsm: Detected transition to state ");
8578 switch (map_state) {
8579 case IMSM_T_STATE_NORMAL: /* transition to normal state */
8580 dprintf("normal: ");
8581 if (is_rebuilding(dev)) {
1ade5cc1 8582 dprintf_cont("while rebuilding");
7ce05701
LD		 8583 /* check if recovery is really finished */
8584 for (mdi = a->info.devs; mdi ; mdi = mdi->next)
8585 if (mdi->recovery_start != MaxSector) {
8586 recovery_not_finished = 1;
8587 break;
8588 }
8589 if (recovery_not_finished) {
1ade5cc1
N		 8590 dprintf_cont("\n");
8591 dprintf("Rebuild has not finished yet, state not changed");
7ce05701
LD		 8592 if (a->last_checkpoint < mdi->recovery_start) {
8593 a->last_checkpoint = mdi->recovery_start;
8594 super->updates_pending++;
8595 }
8596 break;
8597 }
94002678 8598 end_migration(dev, super, map_state);
94002678
AK		 8599 map->failed_disk_num = ~0;
8600 super->updates_pending++;
8601 a->last_checkpoint = 0;
8602 break;
8603 }
8604 if (is_gen_migration(dev)) {
1ade5cc1 8605 dprintf_cont("while general migration");
bf2f0071 8606 if (a->last_checkpoint >= a->info.component_size)
809da78e 8607 end_migration(dev, super, map_state);
94002678
AK		 8608 else
8609 map->map_state = map_state;
28bce06f 8610 map->failed_disk_num = ~0;
94002678 8611 super->updates_pending++;
bf2f0071 8612 break;
94002678
AK		 8613 }
8614 break;
8615 case IMSM_T_STATE_DEGRADED: /* transition to degraded state */
1ade5cc1 8616 dprintf_cont("degraded: ");
089f9d79 8617 if (map->map_state != map_state && !dev->vol.migr_state) {
1ade5cc1 8618 dprintf_cont("mark degraded");
94002678
AK		 8619 map->map_state = map_state;
8620 super->updates_pending++;
8621 a->last_checkpoint = 0;
8622 break;
8623 }
8624 if (is_rebuilding(dev)) {
d7a1fda2 8625 dprintf_cont("while rebuilding ");
a4e96fd8
MT		 8626 if (state & DS_FAULTY) {
8627 dprintf_cont("removing failed drive ");
d7a1fda2
MT		 8628 if (n == map->failed_disk_num) {
8629 dprintf_cont("end migration");
8630 end_migration(dev, super, map_state);
a4e96fd8 8631 a->last_checkpoint = 0;
d7a1fda2 8632 } else {
a4e96fd8 8633 dprintf_cont("fail detected during rebuild, changing map state");
d7a1fda2
MT		 8634 map->map_state = map_state;
8635 }
94002678 8636 super->updates_pending++;
fb12a745
TM		 8637 }
8638
a4e96fd8
MT		 8639 if (!rebuild_done)
8640 break;
8641
fb12a745
TM		 8642 /* check if recovery is really finished */
8643 for (mdi = a->info.devs; mdi ; mdi = mdi->next)
8644 if (mdi->recovery_start != MaxSector) {
8645 recovery_not_finished = 1;
8646 break;
8647 }
8648 if (recovery_not_finished) {
8649 dprintf_cont("\n");
a4e96fd8 8650 dprintf_cont("Rebuild has not finished yet");
fb12a745
TM		 8651 if (a->last_checkpoint < mdi->recovery_start) {
8652 a->last_checkpoint =
8653 mdi->recovery_start;
8654 super->updates_pending++;
8655 }
8656 break;
94002678 8657 }
fb12a745
TM		 8658
8659 dprintf_cont(" Rebuild done, still degraded");
a4e96fd8
MT		 8660 end_migration(dev, super, map_state);
8661 a->last_checkpoint = 0;
8662 super->updates_pending++;
fb12a745
TM		 8663
8664 for (i = 0; i < map->num_members; i++) {
8665 int idx = get_imsm_ord_tbl_ent(dev, i, MAP_0);
8666
8667 if (idx & IMSM_ORD_REBUILD)
8668 map->failed_disk_num = i;
8669 }
8670 super->updates_pending++;
94002678
AK		 8671 break;
8672 }
8673 if (is_gen_migration(dev)) {
1ade5cc1 8674 dprintf_cont("while general migration");
bf2f0071 8675 if (a->last_checkpoint >= a->info.component_size)
809da78e 8676 end_migration(dev, super, map_state);
94002678
AK		 8677 else {
8678 map->map_state = map_state;
3b451610 8679 manage_second_map(super, dev);
94002678
AK		 8680 }
8681 super->updates_pending++;
bf2f0071 8682 break;
28bce06f 8683 }
6ce1fbf1 8684 if (is_initializing(dev)) {
1ade5cc1 8685 dprintf_cont("while initialization.");
6ce1fbf1
AK		 8686 map->map_state = map_state;
8687 super->updates_pending++;
8688 break;
8689 }
94002678
AK		 8690 break;
8691 case IMSM_T_STATE_FAILED: /* transition to failed state */
1ade5cc1 8692 dprintf_cont("failed: ");
94002678 8693 if (is_gen_migration(dev)) {
1ade5cc1 8694 dprintf_cont("while general migration");
94002678
AK		 8695 map->map_state = map_state;
8696 super->updates_pending++;
8697 break;
8698 }
8699 if (map->map_state != map_state) {
1ade5cc1 8700 dprintf_cont("mark failed");
94002678
AK		 8701 end_migration(dev, super, map_state);
8702 super->updates_pending++;
8703 a->last_checkpoint = 0;
8704 break;
8705 }
8706 break;
8707 default:
1ade5cc1 8708 dprintf_cont("state %i\n", map_state);
5802a811 8709 }
1ade5cc1 8710 dprintf_cont("\n");
845dea95
NB		 8711}
8712
f796af5d 8713static int store_imsm_mpb(int fd, struct imsm_super *mpb)
c2a1e7da 8714{
f796af5d 8715 void *buf = mpb;
c2a1e7da
DW		 8716 __u32 mpb_size = __le32_to_cpu(mpb->mpb_size);
8717 unsigned long long dsize;
8718 unsigned long long sectors;
f36a9ecd 8719 unsigned int sector_size;
c2a1e7da 8720
f36a9ecd 8721 get_dev_sector_size(fd, NULL, &sector_size);
c2a1e7da
DW		 8722 get_dev_size(fd, NULL, &dsize);
8723
f36a9ecd 8724 if (mpb_size > sector_size) {
272f648f 8725 /* -1 to account for anchor */
f36a9ecd 8726 sectors = mpb_sectors(mpb, sector_size) - 1;
c2a1e7da 8727
272f648f 8728 /* write the extended mpb to the sectors preceeding the anchor */
f36a9ecd
PB		 8729 if (lseek64(fd, dsize - (sector_size * (2 + sectors)),
8730 SEEK_SET) < 0)
272f648f 8731 return 1;
c2a1e7da 8732
f36a9ecd
PB		 8733 if ((unsigned long long)write(fd, buf + sector_size,
8734 sector_size * sectors) != sector_size * sectors)
272f648f
DW		 8735 return 1;
8736 }
c2a1e7da 8737
272f648f 8738 /* first block is stored on second to last sector of the disk */
f36a9ecd 8739 if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0)
c2a1e7da
DW		 8740 return 1;
8741
466070ad 8742 if ((unsigned int)write(fd, buf, sector_size) != sector_size)
c2a1e7da
DW		 8743 return 1;
8744
c2a1e7da
DW		 8745 return 0;
8746}
8747
2e735d19 8748static void imsm_sync_metadata(struct supertype *container)
845dea95 8749{
2e735d19 8750 struct intel_super *super = container->sb;
c2a1e7da 8751
1a64be56 8752 dprintf("sync metadata: %d\n", super->updates_pending);
c2a1e7da
DW		 8753 if (!super->updates_pending)
8754 return;
8755
36988a3d 8756 write_super_imsm(container, 0);
c2a1e7da
DW		 8757
8758 super->updates_pending = 0;
845dea95
NB		 8759}
8760
272906ef
DW		 8761static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a)
8762{
8763 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8764 int i = get_imsm_disk_idx(dev, idx, MAP_X);
272906ef
DW		 8765 struct dl *dl;
8766
8767 for (dl = super->disks; dl; dl = dl->next)
8768 if (dl->index == i)
8769 break;
8770
25ed7e59 8771 if (dl && is_failed(&dl->disk))
272906ef
DW		 8772 dl = NULL;
8773
8774 if (dl)
1ade5cc1 8775 dprintf("found %x:%x\n", dl->major, dl->minor);
272906ef
DW		 8776
8777 return dl;
8778}
8779
a20d2ba5 8780static struct dl *imsm_add_spare(struct intel_super *super, int slot,
8ba77d32
AK		 8781 struct active_array *a, int activate_new,
8782 struct mdinfo *additional_test_list)
272906ef
DW		 8783{
8784 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8785 int idx = get_imsm_disk_idx(dev, slot, MAP_X);
a20d2ba5
DW		 8786 struct imsm_super *mpb = super->anchor;
8787 struct imsm_map *map;
272906ef
DW		 8788 unsigned long long pos;
8789 struct mdinfo *d;
8790 struct extent *ex;
a20d2ba5 8791 int i, j;
272906ef 8792 int found;
569cc43f
DW		 8793 __u32 array_start = 0;
8794 __u32 array_end = 0;
272906ef 8795 struct dl *dl;
6c932028 8796 struct mdinfo *test_list;
272906ef
DW		 8797
8798 for (dl = super->disks; dl; dl = dl->next) {
8799 /* If in this array, skip */
8800 for (d = a->info.devs ; d ; d = d->next)
e553d2a4
DW		 8801 if (d->state_fd >= 0 &&
8802 d->disk.major == dl->major &&
272906ef 8803 d->disk.minor == dl->minor) {
8ba77d32
AK		 8804 dprintf("%x:%x already in array\n",
8805 dl->major, dl->minor);
272906ef
DW		 8806 break;
8807 }
8808 if (d)
8809 continue;
6c932028
AK		 8810 test_list = additional_test_list;
8811 while (test_list) {
8812 if (test_list->disk.major == dl->major &&
8813 test_list->disk.minor == dl->minor) {
8ba77d32
AK		 8814 dprintf("%x:%x already in additional test list\n",
8815 dl->major, dl->minor);
8816 break;
8817 }
6c932028 8818 test_list = test_list->next;
8ba77d32 8819 }
6c932028 8820 if (test_list)
8ba77d32 8821 continue;
272906ef 8822
e553d2a4 8823 /* skip in use or failed drives */
25ed7e59 8824 if (is_failed(&dl->disk) || idx == dl->index ||
df474657
DW		 8825 dl->index == -2) {
8826 dprintf("%x:%x status (failed: %d index: %d)\n",
25ed7e59 8827 dl->major, dl->minor, is_failed(&dl->disk), idx);
9a1608e5
DW		 8828 continue;
8829 }
8830
a20d2ba5
DW		 8831 /* skip pure spares when we are looking for partially
8832 * assimilated drives
8833 */
8834 if (dl->index == -1 && !activate_new)
8835 continue;
8836
f2cc4f7d
AO		 8837 if (!drive_validate_sector_size(super, dl))
8838 continue;
8839
272906ef 8840 /* Does this unused device have the requisite free space?
a20d2ba5 8841 * It needs to be able to cover all member volumes
272906ef 8842 */
05501181 8843 ex = get_extents(super, dl, 1);
272906ef
DW		 8844 if (!ex) {
8845 dprintf("cannot get extents\n");
8846 continue;
8847 }
a20d2ba5
DW		 8848 for (i = 0; i < mpb->num_raid_devs; i++) {
8849 dev = get_imsm_dev(super, i);
238c0a71 8850 map = get_imsm_map(dev, MAP_0);
272906ef 8851
a20d2ba5
DW		 8852 /* check if this disk is already a member of
8853 * this array
272906ef 8854 */
620b1713 8855 if (get_imsm_disk_slot(map, dl->index) >= 0)
a20d2ba5
DW		 8856 continue;
8857
8858 found = 0;
8859 j = 0;
8860 pos = 0;
5551b113 8861 array_start = pba_of_lba0(map);
329c8278 8862 array_end = array_start +
44490938 8863 per_dev_array_size(map) - 1;
a20d2ba5
DW		 8864
8865 do {
8866 /* check that we can start at pba_of_lba0 with
44490938 8867 * num_data_stripes*blocks_per_stripe of space
a20d2ba5 8868 */
329c8278 8869 if (array_start >= pos && array_end < ex[j].start) {
a20d2ba5
DW		 8870 found = 1;
8871 break;
8872 }
8873 pos = ex[j].start + ex[j].size;
8874 j++;
8875 } while (ex[j-1].size);
8876
8877 if (!found)
272906ef 8878 break;
a20d2ba5 8879 }
272906ef
DW		 8880
8881 free(ex);
a20d2ba5 8882 if (i < mpb->num_raid_devs) {
329c8278
DW		 8883 dprintf("%x:%x does not have %u to %u available\n",
8884 dl->major, dl->minor, array_start, array_end);
272906ef
DW		 8885 /* No room */
8886 continue;
a20d2ba5
DW		 8887 }
8888 return dl;
272906ef
DW		 8889 }
8890
8891 return dl;
8892}
8893
95d07a2c
LM		 8894static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed)
8895{
8896 struct imsm_dev *dev2;
8897 struct imsm_map *map;
8898 struct dl *idisk;
8899 int slot;
8900 int idx;
8901 __u8 state;
8902
8903 dev2 = get_imsm_dev(cont->sb, dev_idx);
8904 if (dev2) {
238c0a71 8905 state = imsm_check_degraded(cont->sb, dev2, failed, MAP_0);
95d07a2c 8906 if (state == IMSM_T_STATE_FAILED) {
238c0a71 8907 map = get_imsm_map(dev2, MAP_0);
95d07a2c
LM		 8908 if (!map)
8909 return 1;
8910 for (slot = 0; slot < map->num_members; slot++) {
8911 /*
8912 * Check if failed disks are deleted from intel
8913 * disk list or are marked to be deleted
8914 */
238c0a71 8915 idx = get_imsm_disk_idx(dev2, slot, MAP_X);
95d07a2c
LM		 8916 idisk = get_imsm_dl_disk(cont->sb, idx);
8917 /*
8918 * Do not rebuild the array if failed disks
8919 * from failed sub-array are not removed from
8920 * container.
8921 */
8922 if (idisk &&
8923 is_failed(&idisk->disk) &&
8924 (idisk->action != DISK_REMOVE))
8925 return 0;
8926 }
8927 }
8928 }
8929 return 1;
8930}
8931
88758e9d
DW		 8932static struct mdinfo *imsm_activate_spare(struct active_array *a,
8933 struct metadata_update **updates)
8934{
8935 /**
d23fe947
DW		 8936 * Find a device with unused free space and use it to replace a
8937 * failed/vacant region in an array. We replace failed regions one a
8938 * array at a time. The result is that a new spare disk will be added
8939 * to the first failed array and after the monitor has finished
8940 * propagating failures the remainder will be consumed.
88758e9d 8941 *
d23fe947
DW		 8942 * FIXME add a capability for mdmon to request spares from another
8943 * container.
88758e9d
DW		 8944 */
8945
8946 struct intel_super *super = a->container->sb;
88758e9d 8947 int inst = a->info.container_member;
949c47a0 8948 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8949 struct imsm_map *map = get_imsm_map(dev, MAP_0);
88758e9d
DW		 8950 int failed = a->info.array.raid_disks;
8951 struct mdinfo *rv = NULL;
8952 struct mdinfo *d;
8953 struct mdinfo *di;
8954 struct metadata_update *mu;
8955 struct dl *dl;
8956 struct imsm_update_activate_spare *u;
8957 int num_spares = 0;
8958 int i;
95d07a2c 8959 int allowed;
88758e9d
DW		 8960
8961 for (d = a->info.devs ; d ; d = d->next) {
8962 if ((d->curr_state & DS_FAULTY) &&
8963 d->state_fd >= 0)
8964 /* wait for Removal to happen */
8965 return NULL;
8966 if (d->state_fd >= 0)
8967 failed--;
8968 }
8969
8970 dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n",
8971 inst, failed, a->info.array.raid_disks, a->info.array.level);
1af97990 8972
e2962bfc
AK		 8973 if (imsm_reshape_blocks_arrays_changes(super))
8974 return NULL;
1af97990 8975
fc8ca064
AK		 8976 /* Cannot activate another spare if rebuild is in progress already
8977 */
8978 if (is_rebuilding(dev)) {
7a862a02 8979 dprintf("imsm: No spare activation allowed. Rebuild in progress already.\n");
fc8ca064
AK		 8980 return NULL;
8981 }
8982
89c67882
AK		 8983 if (a->info.array.level == 4)
8984 /* No repair for takeovered array
8985 * imsm doesn't support raid4
8986 */
8987 return NULL;
8988
3b451610
AK		 8989 if (imsm_check_degraded(super, dev, failed, MAP_0) !=
8990 IMSM_T_STATE_DEGRADED)
88758e9d
DW		 8991 return NULL;
8992
83ca7d45
AP		 8993 if (get_imsm_map(dev, MAP_0)->map_state == IMSM_T_STATE_UNINITIALIZED) {
8994 dprintf("imsm: No spare activation allowed. Volume is not initialized.\n");
8995 return NULL;
8996 }
8997
95d07a2c
LM		 8998 /*
8999 * If there are any failed disks check state of the other volume.
9000 * Block rebuild if the another one is failed until failed disks
9001 * are removed from container.
9002 */
9003 if (failed) {
7a862a02 9004 dprintf("found failed disks in %.*s, check if there anotherfailed sub-array.\n",
c4acd1e5 9005 MAX_RAID_SERIAL_LEN, dev->volume);
95d07a2c
LM		 9006 /* check if states of the other volumes allow for rebuild */
9007 for (i = 0; i < super->anchor->num_raid_devs; i++) {
9008 if (i != inst) {
9009 allowed = imsm_rebuild_allowed(a->container,
9010 i, failed);
9011 if (!allowed)
9012 return NULL;
9013 }
9014 }
9015 }
9016
88758e9d 9017 /* For each slot, if it is not working, find a spare */
88758e9d
DW		 9018 for (i = 0; i < a->info.array.raid_disks; i++) {
9019 for (d = a->info.devs ; d ; d = d->next)
9020 if (d->disk.raid_disk == i)
9021 break;
9022 dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0);
9023 if (d && (d->state_fd >= 0))
9024 continue;
9025
272906ef 9026 /*
a20d2ba5
DW		 9027 * OK, this device needs recovery. Try to re-add the
9028 * previous occupant of this slot, if this fails see if
9029 * we can continue the assimilation of a spare that was
9030 * partially assimilated, finally try to activate a new
9031 * spare.
272906ef
DW		 9032 */
9033 dl = imsm_readd(super, i, a);
9034 if (!dl)
b303fe21 9035 dl = imsm_add_spare(super, i, a, 0, rv);
a20d2ba5 9036 if (!dl)
b303fe21 9037 dl = imsm_add_spare(super, i, a, 1, rv);
272906ef
DW		 9038 if (!dl)
9039 continue;
1011e834 9040
272906ef 9041 /* found a usable disk with enough space */
503975b9 9042 di = xcalloc(1, sizeof(*di));
272906ef
DW		 9043
9044 /* dl->index will be -1 in the case we are activating a
9045 * pristine spare. imsm_process_update() will create a
9046 * new index in this case. Once a disk is found to be
9047 * failed in all member arrays it is kicked from the
9048 * metadata
9049 */
9050 di->disk.number = dl->index;
d23fe947 9051
272906ef
DW		 9052 /* (ab)use di->devs to store a pointer to the device
9053 * we chose
9054 */
9055 di->devs = (struct mdinfo *) dl;
9056
9057 di->disk.raid_disk = i;
9058 di->disk.major = dl->major;
9059 di->disk.minor = dl->minor;
9060 di->disk.state = 0;
d23534e4 9061 di->recovery_start = 0;
5551b113 9062 di->data_offset = pba_of_lba0(map);
272906ef
DW		 9063 di->component_size = a->info.component_size;
9064 di->container_member = inst;
5e46202e 9065 di->bb.supported = 1;
2c8890e9 9066 if (a->info.consistency_policy == CONSISTENCY_POLICY_PPL) {
2432ce9b 9067 di->ppl_sector = get_ppl_sector(super, inst);
c2462068 9068 di->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9;
2432ce9b 9069 }
148acb7b 9070 super->random = random32();
272906ef
DW		 9071 di->next = rv;
9072 rv = di;
9073 num_spares++;
9074 dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor,
9075 i, di->data_offset);
88758e9d
DW		 9076 }
9077
9078 if (!rv)
9079 /* No spares found */
9080 return rv;
9081 /* Now 'rv' has a list of devices to return.
9082 * Create a metadata_update record to update the
9083 * disk_ord_tbl for the array
9084 */
503975b9 9085 mu = xmalloc(sizeof(*mu));
1011e834 9086 mu->buf = xcalloc(num_spares,
503975b9 9087 sizeof(struct imsm_update_activate_spare));
88758e9d 9088 mu->space = NULL;
cb23f1f4 9089 mu->space_list = NULL;
88758e9d
DW		 9090 mu->len = sizeof(struct imsm_update_activate_spare) * num_spares;
9091 mu->next = *updates;
9092 u = (struct imsm_update_activate_spare *) mu->buf;
9093
9094 for (di = rv ; di ; di = di->next) {
9095 u->type = update_activate_spare;
d23fe947
DW		 9096 u->dl = (struct dl *) di->devs;
9097 di->devs = NULL;
88758e9d
DW		 9098 u->slot = di->disk.raid_disk;
9099 u->array = inst;
9100 u->next = u + 1;
9101 u++;
9102 }
9103 (u-1)->next = NULL;
9104 *updates = mu;
9105
9106 return rv;
9107}
9108
54c2c1ea 9109static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u)
8273f55e 9110{
54c2c1ea 9111 struct imsm_dev *dev = get_imsm_dev(super, idx);
238c0a71
AK		 9112 struct imsm_map *map = get_imsm_map(dev, MAP_0);
9113 struct imsm_map *new_map = get_imsm_map(&u->dev, MAP_0);
54c2c1ea
DW		 9114 struct disk_info *inf = get_disk_info(u);
9115 struct imsm_disk *disk;
8273f55e
DW		 9116 int i;
9117 int j;
8273f55e 9118
54c2c1ea 9119 for (i = 0; i < map->num_members; i++) {
238c0a71 9120 disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, MAP_X));
54c2c1ea
DW		 9121 for (j = 0; j < new_map->num_members; j++)
9122 if (serialcmp(disk->serial, inf[j].serial) == 0)
8273f55e
DW		 9123 return 1;
9124 }
9125
9126 return 0;
9127}
9128
1a64be56
LM		 9129static struct dl *get_disk_super(struct intel_super *super, int major, int minor)
9130{
594dc1b8
JS		 9131 struct dl *dl;
9132
1a64be56 9133 for (dl = super->disks; dl; dl = dl->next)
089f9d79 9134 if (dl->major == major && dl->minor == minor)
1a64be56
LM		 9135 return dl;
9136 return NULL;
9137}
9138
9139static int remove_disk_super(struct intel_super *super, int major, int minor)
9140{
594dc1b8 9141 struct dl *prev;
1a64be56
LM		 9142 struct dl *dl;
9143
9144 prev = NULL;
9145 for (dl = super->disks; dl; dl = dl->next) {
089f9d79 9146 if (dl->major == major && dl->minor == minor) {
1a64be56
LM		 9147 /* remove */
9148 if (prev)
9149 prev->next = dl->next;
9150 else
9151 super->disks = dl->next;
9152 dl->next = NULL;
9153 __free_imsm_disk(dl);
1ade5cc1 9154 dprintf("removed %x:%x\n", major, minor);
1a64be56
LM		 9155 break;
9156 }
9157 prev = dl;
9158 }
9159 return 0;
9160}
9161
f21e18ca 9162static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index);
ae6aad82 9163
1a64be56
LM		 9164static int add_remove_disk_update(struct intel_super *super)
9165{
9166 int check_degraded = 0;
594dc1b8
JS		 9167 struct dl *disk;
9168
1a64be56
LM		 9169 /* add/remove some spares to/from the metadata/contrainer */
9170 while (super->disk_mgmt_list) {
9171 struct dl *disk_cfg;
9172
9173 disk_cfg = super->disk_mgmt_list;
9174 super->disk_mgmt_list = disk_cfg->next;
9175 disk_cfg->next = NULL;
9176
9177 if (disk_cfg->action == DISK_ADD) {
9178 disk_cfg->next = super->disks;
9179 super->disks = disk_cfg;
9180 check_degraded = 1;
1ade5cc1
N		 9181 dprintf("added %x:%x\n",
9182 disk_cfg->major, disk_cfg->minor);
1a64be56
LM		 9183 } else if (disk_cfg->action == DISK_REMOVE) {
9184 dprintf("Disk remove action processed: %x.%x\n",
9185 disk_cfg->major, disk_cfg->minor);
9186 disk = get_disk_super(super,
9187 disk_cfg->major,
9188 disk_cfg->minor);
9189 if (disk) {
9190 /* store action status */
9191 disk->action = DISK_REMOVE;
9192 /* remove spare disks only */
9193 if (disk->index == -1) {
9194 remove_disk_super(super,
9195 disk_cfg->major,
9196 disk_cfg->minor);
91c97c54
MT		 9197 } else {
9198 disk_cfg->fd = disk->fd;
9199 disk->fd = -1;
1a64be56
LM		 9200 }
9201 }
9202 /* release allocate disk structure */
9203 __free_imsm_disk(disk_cfg);
9204 }
9205 }
9206 return check_degraded;
9207}
9208
a29911da
PC		 9209static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u,
9210 struct intel_super *super,
9211 void ***space_list)
9212{
9213 struct intel_dev *id;
9214 void **tofree = NULL;
9215 int ret_val = 0;
9216
1ade5cc1 9217 dprintf("(enter)\n");
089f9d79 9218 if (u->subdev < 0 || u->subdev > 1) {
a29911da
PC		 9219 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
9220 return ret_val;
9221 }
089f9d79 9222 if (space_list == NULL || *space_list == NULL) {
a29911da
PC		 9223 dprintf("imsm: Error: Memory is not allocated\n");
9224 return ret_val;
9225 }
9226
9227 for (id = super->devlist ; id; id = id->next) {
9228 if (id->index == (unsigned)u->subdev) {
9229 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
9230 struct imsm_map *map;
9231 struct imsm_dev *new_dev =
9232 (struct imsm_dev *)*space_list;
238c0a71 9233 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
a29911da
PC		 9234 int to_state;
9235 struct dl *new_disk;
9236
9237 if (new_dev == NULL)
9238 return ret_val;
9239 *space_list = **space_list;
9240 memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0));
238c0a71 9241 map = get_imsm_map(new_dev, MAP_0);
a29911da
PC		 9242 if (migr_map) {
9243 dprintf("imsm: Error: migration in progress");
9244 return ret_val;
9245 }
9246
9247 to_state = map->map_state;
9248 if ((u->new_level == 5) && (map->raid_level == 0)) {
9249 map->num_members++;
9250 /* this should not happen */
9251 if (u->new_disks[0] < 0) {
9252 map->failed_disk_num =
9253 map->num_members - 1;
9254 to_state = IMSM_T_STATE_DEGRADED;
9255 } else
9256 to_state = IMSM_T_STATE_NORMAL;
9257 }
8e59f3d8 9258 migrate(new_dev, super, to_state, MIGR_GEN_MIGR);
a29911da
PC		 9259 if (u->new_level > -1)
9260 map->raid_level = u->new_level;
238c0a71 9261 migr_map = get_imsm_map(new_dev, MAP_1);
a29911da
PC		 9262 if ((u->new_level == 5) &&
9263 (migr_map->raid_level == 0)) {
9264 int ord = map->num_members - 1;
9265 migr_map->num_members--;
9266 if (u->new_disks[0] < 0)
9267 ord |= IMSM_ORD_REBUILD;
9268 set_imsm_ord_tbl_ent(map,
9269 map->num_members - 1,
9270 ord);
9271 }
9272 id->dev = new_dev;
9273 tofree = (void **)dev;
9274
4bba0439
PC		 9275 /* update chunk size
9276 */
06fb291a
PB		 9277 if (u->new_chunksize > 0) {
9278 unsigned long long num_data_stripes;
9529d343
MD		 9279 struct imsm_map *dest_map =
9280 get_imsm_map(dev, MAP_0);
06fb291a 9281 int used_disks =
9529d343 9282 imsm_num_data_members(dest_map);
06fb291a
PB		 9283
9284 if (used_disks == 0)
9285 return ret_val;
9286
4bba0439
PC		 9287 map->blocks_per_strip =
9288 __cpu_to_le16(u->new_chunksize * 2);
06fb291a 9289 num_data_stripes =
fcc2c9da 9290 imsm_dev_size(dev) / used_disks;
06fb291a
PB		 9291 num_data_stripes /= map->blocks_per_strip;
9292 num_data_stripes /= map->num_domains;
9293 set_num_data_stripes(map, num_data_stripes);
9294 }
4bba0439 9295
44490938
MD		 9296 /* ensure blocks_per_member has valid value
9297 */
9298 set_blocks_per_member(map,
9299 per_dev_array_size(map) +
9300 NUM_BLOCKS_DIRTY_STRIPE_REGION);
9301
a29911da
PC		 9302 /* add disk
9303 */
089f9d79
JS		 9304 if (u->new_level != 5 || migr_map->raid_level != 0 ||
9305 migr_map->raid_level == map->raid_level)
a29911da
PC		 9306 goto skip_disk_add;
9307
9308 if (u->new_disks[0] >= 0) {
9309 /* use passes spare
9310 */
9311 new_disk = get_disk_super(super,
9312 major(u->new_disks[0]),
9313 minor(u->new_disks[0]));
7a862a02 9314 dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n",
a29911da
PC		 9315 major(u->new_disks[0]),
9316 minor(u->new_disks[0]),
9317 new_disk, new_disk->index);
9318 if (new_disk == NULL)
9319 goto error_disk_add;
9320
9321 new_disk->index = map->num_members - 1;
9322 /* slot to fill in autolayout
9323 */
9324 new_disk->raiddisk = new_disk->index;
9325 new_disk->disk.status |= CONFIGURED_DISK;
9326 new_disk->disk.status &= ~SPARE_DISK;
9327 } else
9328 goto error_disk_add;
9329
9330skip_disk_add:
9331 *tofree = *space_list;
9332 /* calculate new size
9333 */
f3871fdc 9334 imsm_set_array_size(new_dev, -1);
a29911da
PC		 9335
9336 ret_val = 1;
9337 }
9338 }
9339
9340 if (tofree)
9341 *space_list = tofree;
9342 return ret_val;
9343
9344error_disk_add:
9345 dprintf("Error: imsm: Cannot find disk.\n");
9346 return ret_val;
9347}
9348
f3871fdc
AK		 9349static int apply_size_change_update(struct imsm_update_size_change *u,
9350 struct intel_super *super)
9351{
9352 struct intel_dev *id;
9353 int ret_val = 0;
9354
1ade5cc1 9355 dprintf("(enter)\n");
089f9d79 9356 if (u->subdev < 0 || u->subdev > 1) {
f3871fdc
AK		 9357 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
9358 return ret_val;
9359 }
9360
9361 for (id = super->devlist ; id; id = id->next) {
9362 if (id->index == (unsigned)u->subdev) {
9363 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
9364 struct imsm_map *map = get_imsm_map(dev, MAP_0);
9529d343 9365 int used_disks = imsm_num_data_members(map);
f3871fdc 9366 unsigned long long blocks_per_member;
06fb291a 9367 unsigned long long num_data_stripes;
44490938
MD		 9368 unsigned long long new_size_per_disk;
9369
9370 if (used_disks == 0)
9371 return 0;
f3871fdc
AK		 9372
9373 /* calculate new size
9374 */
44490938
MD		 9375 new_size_per_disk = u->new_size / used_disks;
9376 blocks_per_member = new_size_per_disk +
9377 NUM_BLOCKS_DIRTY_STRIPE_REGION;
9378 num_data_stripes = new_size_per_disk /
06fb291a
PB		 9379 map->blocks_per_strip;
9380 num_data_stripes /= map->num_domains;
9381 dprintf("(size: %llu, blocks per member: %llu, num_data_stipes: %llu)\n",
44490938 9382 u->new_size, new_size_per_disk,
06fb291a 9383 num_data_stripes);
f3871fdc 9384 set_blocks_per_member(map, blocks_per_member);
06fb291a 9385 set_num_data_stripes(map, num_data_stripes);
f3871fdc
AK		 9386 imsm_set_array_size(dev, u->new_size);
9387
9388 ret_val = 1;
9389 break;
9390 }
9391 }
9392
9393 return ret_val;
9394}
9395
061d7da3 9396static int apply_update_activate_spare(struct imsm_update_activate_spare *u,
ca9de185 9397 struct intel_super *super,
061d7da3
LO		 9398 struct active_array *active_array)
9399{
9400 struct imsm_super *mpb = super->anchor;
9401 struct imsm_dev *dev = get_imsm_dev(super, u->array);
238c0a71 9402 struct imsm_map *map = get_imsm_map(dev, MAP_0);
061d7da3
LO		 9403 struct imsm_map *migr_map;
9404 struct active_array *a;
9405 struct imsm_disk *disk;
9406 __u8 to_state;
9407 struct dl *dl;
9408 unsigned int found;
9409 int failed;
5961eeec 9410 int victim;
061d7da3 9411 int i;
5961eeec 9412 int second_map_created = 0;
061d7da3 9413
5961eeec 9414 for (; u; u = u->next) {
238c0a71 9415 victim = get_imsm_disk_idx(dev, u->slot, MAP_X);
061d7da3 9416
5961eeec 9417 if (victim < 0)
9418 return 0;
061d7da3 9419
5961eeec 9420 for (dl = super->disks; dl; dl = dl->next)
9421 if (dl == u->dl)
9422 break;
061d7da3 9423
5961eeec 9424 if (!dl) {
7a862a02 9425 pr_err("error: imsm_activate_spare passed an unknown disk (index: %d)\n",
5961eeec 9426 u->dl->index);
9427 return 0;
9428 }
061d7da3 9429
5961eeec 9430 /* count failures (excluding rebuilds and the victim)
9431 * to determine map[0] state
9432 */
9433 failed = 0;
9434 for (i = 0; i < map->num_members; i++) {
9435 if (i == u->slot)
9436 continue;
9437 disk = get_imsm_disk(super,
238c0a71 9438 get_imsm_disk_idx(dev, i, MAP_X));
5961eeec 9439 if (!disk || is_failed(disk))
9440 failed++;
9441 }
061d7da3 9442
5961eeec 9443 /* adding a pristine spare, assign a new index */
9444 if (dl->index < 0) {
9445 dl->index = super->anchor->num_disks;
9446 super->anchor->num_disks++;
9447 }
9448 disk = &dl->disk;
9449 disk->status |= CONFIGURED_DISK;
9450 disk->status &= ~SPARE_DISK;
9451
9452 /* mark rebuild */
238c0a71 9453 to_state = imsm_check_degraded(super, dev, failed, MAP_0);
5961eeec 9454 if (!second_map_created) {
9455 second_map_created = 1;
9456 map->map_state = IMSM_T_STATE_DEGRADED;
9457 migrate(dev, super, to_state, MIGR_REBUILD);
9458 } else
9459 map->map_state = to_state;
238c0a71 9460 migr_map = get_imsm_map(dev, MAP_1);
5961eeec 9461 set_imsm_ord_tbl_ent(map, u->slot, dl->index);
9462 set_imsm_ord_tbl_ent(migr_map, u->slot,
9463 dl->index | IMSM_ORD_REBUILD);
9464
9465 /* update the family_num to mark a new container
9466 * generation, being careful to record the existing
9467 * family_num in orig_family_num to clean up after
9468 * earlier mdadm versions that neglected to set it.
9469 */
9470 if (mpb->orig_family_num == 0)
9471 mpb->orig_family_num = mpb->family_num;
9472 mpb->family_num += super->random;
9473
9474 /* count arrays using the victim in the metadata */
9475 found = 0;
9476 for (a = active_array; a ; a = a->next) {
9477 dev = get_imsm_dev(super, a->info.container_member);
238c0a71 9478 map = get_imsm_map(dev, MAP_0);
061d7da3 9479
5961eeec 9480 if (get_imsm_disk_slot(map, victim) >= 0)
9481 found++;
9482 }
061d7da3 9483
5961eeec 9484 /* delete the victim if it is no longer being
9485 * utilized anywhere
061d7da3 9486 */
5961eeec 9487 if (!found) {
9488 struct dl **dlp;
061d7da3 9489
5961eeec 9490 /* We know that 'manager' isn't touching anything,
9491 * so it is safe to delete
9492 */
9493 for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next)
061d7da3
LO		 9494 if ((*dlp)->index == victim)
9495 break;
5961eeec 9496
9497 /* victim may be on the missing list */
9498 if (!*dlp)
9499 for (dlp = &super->missing; *dlp;
9500 dlp = &(*dlp)->next)
9501 if ((*dlp)->index == victim)
9502 break;
9503 imsm_delete(super, dlp, victim);
9504 }
061d7da3
LO		 9505 }
9506
9507 return 1;
9508}
a29911da 9509
2e5dc010
N		 9510static int apply_reshape_container_disks_update(struct imsm_update_reshape *u,
9511 struct intel_super *super,
9512 void ***space_list)
9513{
9514 struct dl *new_disk;
9515 struct intel_dev *id;
9516 int i;
9517 int delta_disks = u->new_raid_disks - u->old_raid_disks;
ee4beede 9518 int disk_count = u->old_raid_disks;
2e5dc010
N		 9519 void **tofree = NULL;
9520 int devices_to_reshape = 1;
9521 struct imsm_super *mpb = super->anchor;
9522 int ret_val = 0;
d098291a 9523 unsigned int dev_id;
2e5dc010 9524
1ade5cc1 9525 dprintf("(enter)\n");
2e5dc010
N		 9526
9527 /* enable spares to use in array */
9528 for (i = 0; i < delta_disks; i++) {
9529 new_disk = get_disk_super(super,
9530 major(u->new_disks[i]),
9531 minor(u->new_disks[i]));
7a862a02 9532 dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n",
2e5dc010
N		 9533 major(u->new_disks[i]), minor(u->new_disks[i]),
9534 new_disk, new_disk->index);
089f9d79
JS		 9535 if (new_disk == NULL ||
9536 (new_disk->index >= 0 &&
9537 new_disk->index < u->old_raid_disks))
2e5dc010 9538 goto update_reshape_exit;
ee4beede 9539 new_disk->index = disk_count++;
2e5dc010
N		 9540 /* slot to fill in autolayout
9541 */
9542 new_disk->raiddisk = new_disk->index;
9543 new_disk->disk.status |=
9544 CONFIGURED_DISK;
9545 new_disk->disk.status &= ~SPARE_DISK;
9546 }
9547
ed7333bd
AK		 9548 dprintf("imsm: volume set mpb->num_raid_devs = %i\n",
9549 mpb->num_raid_devs);
2e5dc010
N		 9550 /* manage changes in volume
9551 */
d098291a 9552 for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) {
2e5dc010
N		 9553 void **sp = *space_list;
9554 struct imsm_dev *newdev;
9555 struct imsm_map *newmap, *oldmap;
9556
d098291a
AK		 9557 for (id = super->devlist ; id; id = id->next) {
9558 if (id->index == dev_id)
9559 break;
9560 }
9561 if (id == NULL)
9562 break;
2e5dc010
N		 9563 if (!sp)
9564 continue;
9565 *space_list = *sp;
9566 newdev = (void*)sp;
9567 /* Copy the dev, but not (all of) the map */
9568 memcpy(newdev, id->dev, sizeof(*newdev));
238c0a71
AK		 9569 oldmap = get_imsm_map(id->dev, MAP_0);
9570 newmap = get_imsm_map(newdev, MAP_0);
2e5dc010
N		 9571 /* Copy the current map */
9572 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
9573 /* update one device only
9574 */
9575 if (devices_to_reshape) {
ed7333bd
AK		 9576 dprintf("imsm: modifying subdev: %i\n",
9577 id->index);
2e5dc010
N		 9578 devices_to_reshape--;
9579 newdev->vol.migr_state = 1;
9580 newdev->vol.curr_migr_unit = 0;
ea672ee1 9581 set_migr_type(newdev, MIGR_GEN_MIGR);
2e5dc010
N		 9582 newmap->num_members = u->new_raid_disks;
9583 for (i = 0; i < delta_disks; i++) {
9584 set_imsm_ord_tbl_ent(newmap,
9585 u->old_raid_disks + i,
9586 u->old_raid_disks + i);
9587 }
9588 /* New map is correct, now need to save old map
9589 */
238c0a71 9590 newmap = get_imsm_map(newdev, MAP_1);
2e5dc010
N		 9591 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
9592
f3871fdc 9593 imsm_set_array_size(newdev, -1);
2e5dc010
N		 9594 }
9595
9596 sp = (void **)id->dev;
9597 id->dev = newdev;
9598 *sp = tofree;
9599 tofree = sp;
8e59f3d8
AK		 9600
9601 /* Clear migration record */
9602 memset(super->migr_rec, 0, sizeof(struct migr_record));
2e5dc010 9603 }
819bc634
AK		 9604 if (tofree)
9605 *space_list = tofree;
2e5dc010
N		 9606 ret_val = 1;
9607
9608update_reshape_exit:
9609
9610 return ret_val;
9611}
9612
bb025c2f 9613static int apply_takeover_update(struct imsm_update_takeover *u,
8ca6df95
KW		 9614 struct intel_super *super,
9615 void ***space_list)
bb025c2f
KW		 9616{
9617 struct imsm_dev *dev = NULL;
8ca6df95
KW		 9618 struct intel_dev *dv;
9619 struct imsm_dev *dev_new;
bb025c2f
KW		 9620 struct imsm_map *map;
9621 struct dl *dm, *du;
8ca6df95 9622 int i;
bb025c2f
KW		 9623
9624 for (dv = super->devlist; dv; dv = dv->next)
9625 if (dv->index == (unsigned int)u->subarray) {
9626 dev = dv->dev;
9627 break;
9628 }
9629
9630 if (dev == NULL)
9631 return 0;
9632
238c0a71 9633 map = get_imsm_map(dev, MAP_0);
bb025c2f
KW		 9634
9635 if (u->direction == R10_TO_R0) {
06fb291a
PB		 9636 unsigned long long num_data_stripes;
9637
43d5ec18 9638 /* Number of failed disks must be half of initial disk number */
3b451610
AK		 9639 if (imsm_count_failed(super, dev, MAP_0) !=
9640 (map->num_members / 2))
43d5ec18
KW		 9641 return 0;
9642
bb025c2f
KW		 9643 /* iterate through devices to mark removed disks as spare */
9644 for (dm = super->disks; dm; dm = dm->next) {
9645 if (dm->disk.status & FAILED_DISK) {
9646 int idx = dm->index;
9647 /* update indexes on the disk list */
9648/* FIXME this loop-with-the-loop looks wrong, I'm not convinced
9649 the index values will end up being correct.... NB */
9650 for (du = super->disks; du; du = du->next)
9651 if (du->index > idx)
9652 du->index--;
9653 /* mark as spare disk */
a8619d23 9654 mark_spare(dm);
bb025c2f
KW		 9655 }
9656 }
bb025c2f
KW		 9657 /* update map */
9658 map->num_members = map->num_members / 2;
9659 map->map_state = IMSM_T_STATE_NORMAL;
9660 map->num_domains = 1;
9661 map->raid_level = 0;
9662 map->failed_disk_num = -1;
4a353e6e
RS		 9663 num_data_stripes = imsm_dev_size(dev) / 2;
9664 num_data_stripes /= map->blocks_per_strip;
9665 set_num_data_stripes(map, num_data_stripes);
bb025c2f
KW		 9666 }
9667
8ca6df95
KW		 9668 if (u->direction == R0_TO_R10) {
9669 void **space;
4a353e6e
RS		 9670 unsigned long long num_data_stripes;
9671
8ca6df95
KW		 9672 /* update slots in current disk list */
9673 for (dm = super->disks; dm; dm = dm->next) {
9674 if (dm->index >= 0)
9675 dm->index *= 2;
9676 }
9677 /* create new *missing* disks */
9678 for (i = 0; i < map->num_members; i++) {
9679 space = *space_list;
9680 if (!space)
9681 continue;
9682 *space_list = *space;
9683 du = (void *)space;
9684 memcpy(du, super->disks, sizeof(*du));
8ca6df95
KW		 9685 du->fd = -1;
9686 du->minor = 0;
9687 du->major = 0;
9688 du->index = (i * 2) + 1;
9689 sprintf((char *)du->disk.serial,
9690 " MISSING_%d", du->index);
9691 sprintf((char *)du->serial,
9692 "MISSING_%d", du->index);
9693 du->next = super->missing;
9694 super->missing = du;
9695 }
9696 /* create new dev and map */
9697 space = *space_list;
9698 if (!space)
9699 return 0;
9700 *space_list = *space;
9701 dev_new = (void *)space;
9702 memcpy(dev_new, dev, sizeof(*dev));
9703 /* update new map */
238c0a71 9704 map = get_imsm_map(dev_new, MAP_0);
8ca6df95 9705 map->num_members = map->num_members * 2;
1a2487c2 9706 map->map_state = IMSM_T_STATE_DEGRADED;
8ca6df95
KW		 9707 map->num_domains = 2;
9708 map->raid_level = 1;
4a353e6e
RS		 9709 num_data_stripes = imsm_dev_size(dev) / 2;
9710 num_data_stripes /= map->blocks_per_strip;
9711 num_data_stripes /= map->num_domains;
9712 set_num_data_stripes(map, num_data_stripes);
9713
8ca6df95
KW		 9714 /* replace dev<->dev_new */
9715 dv->dev = dev_new;
9716 }
bb025c2f
KW		 9717 /* update disk order table */
9718 for (du = super->disks; du; du = du->next)
9719 if (du->index >= 0)
9720 set_imsm_ord_tbl_ent(map, du->index, du->index);
8ca6df95 9721 for (du = super->missing; du; du = du->next)
1a2487c2
KW		 9722 if (du->index >= 0) {
9723 set_imsm_ord_tbl_ent(map, du->index, du->index);
4c9e8c1e 9724 mark_missing(super, dv->dev, &du->disk, du->index);
1a2487c2 9725 }
bb025c2f
KW		 9726
9727 return 1;
9728}
9729
e8319a19
DW		 9730static void imsm_process_update(struct supertype *st,
9731 struct metadata_update *update)
9732{
9733 /**
9734 * crack open the metadata_update envelope to find the update record
9735 * update can be one of:
d195167d
AK		 9736 * update_reshape_container_disks - all the arrays in the container
9737 * are being reshaped to have more devices. We need to mark
9738 * the arrays for general migration and convert selected spares
9739 * into active devices.
9740 * update_activate_spare - a spare device has replaced a failed
1011e834
N		 9741 * device in an array, update the disk_ord_tbl. If this disk is
9742 * present in all member arrays then also clear the SPARE_DISK
9743 * flag
d195167d
AK		 9744 * update_create_array
9745 * update_kill_array
9746 * update_rename_array
9747 * update_add_remove_disk
e8319a19
DW		 9748 */
9749 struct intel_super *super = st->sb;
4d7b1503 9750 struct imsm_super *mpb;
e8319a19
DW		 9751 enum imsm_update_type type = *(enum imsm_update_type *) update->buf;
9752
4d7b1503
DW		 9753 /* update requires a larger buf but the allocation failed */
9754 if (super->next_len && !super->next_buf) {
9755 super->next_len = 0;
9756 return;
9757 }
9758
9759 if (super->next_buf) {
9760 memcpy(super->next_buf, super->buf, super->len);
9761 free(super->buf);
9762 super->len = super->next_len;
9763 super->buf = super->next_buf;
9764
9765 super->next_len = 0;
9766 super->next_buf = NULL;
9767 }
9768
9769 mpb = super->anchor;
9770
e8319a19 9771 switch (type) {
0ec5d470
AK		 9772 case update_general_migration_checkpoint: {
9773 struct intel_dev *id;
9774 struct imsm_update_general_migration_checkpoint *u =
9775 (void *)update->buf;
9776
1ade5cc1 9777 dprintf("called for update_general_migration_checkpoint\n");
0ec5d470
AK		 9778
9779 /* find device under general migration */
9780 for (id = super->devlist ; id; id = id->next) {
9781 if (is_gen_migration(id->dev)) {
9782 id->dev->vol.curr_migr_unit =
9783 __cpu_to_le32(u->curr_migr_unit);
9784 super->updates_pending++;
9785 }
9786 }
9787 break;
9788 }
bb025c2f
KW		 9789 case update_takeover: {
9790 struct imsm_update_takeover *u = (void *)update->buf;
1a2487c2
KW		 9791 if (apply_takeover_update(u, super, &update->space_list)) {
9792 imsm_update_version_info(super);
bb025c2f 9793 super->updates_pending++;
1a2487c2 9794 }
bb025c2f
KW		 9795 break;
9796 }
9797
78b10e66 9798 case update_reshape_container_disks: {
d195167d 9799 struct imsm_update_reshape *u = (void *)update->buf;
2e5dc010
N		 9800 if (apply_reshape_container_disks_update(
9801 u, super, &update->space_list))
9802 super->updates_pending++;
78b10e66
N		 9803 break;
9804 }
48c5303a 9805 case update_reshape_migration: {
a29911da
PC		 9806 struct imsm_update_reshape_migration *u = (void *)update->buf;
9807 if (apply_reshape_migration_update(
9808 u, super, &update->space_list))
9809 super->updates_pending++;
48c5303a
PC		 9810 break;
9811 }
f3871fdc
AK		 9812 case update_size_change: {
9813 struct imsm_update_size_change *u = (void *)update->buf;
9814 if (apply_size_change_update(u, super))
9815 super->updates_pending++;
9816 break;
9817 }
e8319a19 9818 case update_activate_spare: {
1011e834 9819 struct imsm_update_activate_spare *u = (void *) update->buf;
061d7da3
LO		 9820 if (apply_update_activate_spare(u, super, st->arrays))
9821 super->updates_pending++;
8273f55e
DW		 9822 break;
9823 }
9824 case update_create_array: {
9825 /* someone wants to create a new array, we need to be aware of
9826 * a few races/collisions:
9827 * 1/ 'Create' called by two separate instances of mdadm
9828 * 2/ 'Create' versus 'activate_spare': mdadm has chosen
9829 * devices that have since been assimilated via
9830 * activate_spare.
9831 * In the event this update can not be carried out mdadm will
9832 * (FIX ME) notice that its update did not take hold.
9833 */
9834 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 9835 struct intel_dev *dv;
8273f55e
DW		 9836 struct imsm_dev *dev;
9837 struct imsm_map *map, *new_map;
9838 unsigned long long start, end;
9839 unsigned long long new_start, new_end;
9840 int i;
54c2c1ea
DW		 9841 struct disk_info *inf;
9842 struct dl *dl;
8273f55e
DW		 9843
9844 /* handle racing creates: first come first serve */
9845 if (u->dev_idx < mpb->num_raid_devs) {
1ade5cc1 9846 dprintf("subarray %d already defined\n", u->dev_idx);
ba2de7ba 9847 goto create_error;
8273f55e
DW		 9848 }
9849
9850 /* check update is next in sequence */
9851 if (u->dev_idx != mpb->num_raid_devs) {
1ade5cc1
N		 9852 dprintf("can not create array %d expected index %d\n",
9853 u->dev_idx, mpb->num_raid_devs);
ba2de7ba 9854 goto create_error;
8273f55e
DW		 9855 }
9856
238c0a71 9857 new_map = get_imsm_map(&u->dev, MAP_0);
5551b113 9858 new_start = pba_of_lba0(new_map);
44490938 9859 new_end = new_start + per_dev_array_size(new_map);
54c2c1ea 9860 inf = get_disk_info(u);
8273f55e
DW		 9861
9862 /* handle activate_spare versus create race:
9863 * check to make sure that overlapping arrays do not include
9864 * overalpping disks
9865 */
9866 for (i = 0; i < mpb->num_raid_devs; i++) {
949c47a0 9867 dev = get_imsm_dev(super, i);
238c0a71 9868 map = get_imsm_map(dev, MAP_0);
5551b113 9869 start = pba_of_lba0(map);
44490938 9870 end = start + per_dev_array_size(map);
8273f55e
DW		 9871 if ((new_start >= start && new_start <= end) ||
9872 (start >= new_start && start <= new_end))
54c2c1ea
DW		 9873 /* overlap */;
9874 else
9875 continue;
9876
9877 if (disks_overlap(super, i, u)) {
1ade5cc1 9878 dprintf("arrays overlap\n");
ba2de7ba 9879 goto create_error;
8273f55e
DW		 9880 }
9881 }
8273f55e 9882
949c47a0
DW		 9883 /* check that prepare update was successful */
9884 if (!update->space) {
1ade5cc1 9885 dprintf("prepare update failed\n");
ba2de7ba 9886 goto create_error;
949c47a0
DW		 9887 }
9888
54c2c1ea
DW		 9889 /* check that all disks are still active before committing
9890 * changes. FIXME: could we instead handle this by creating a
9891 * degraded array? That's probably not what the user expects,
9892 * so better to drop this update on the floor.
9893 */
9894 for (i = 0; i < new_map->num_members; i++) {
9895 dl = serial_to_dl(inf[i].serial, super);
9896 if (!dl) {
1ade5cc1 9897 dprintf("disk disappeared\n");
ba2de7ba 9898 goto create_error;
54c2c1ea 9899 }
949c47a0
DW		 9900 }
9901
8273f55e 9902 super->updates_pending++;
54c2c1ea
DW		 9903
9904 /* convert spares to members and fixup ord_tbl */
9905 for (i = 0; i < new_map->num_members; i++) {
9906 dl = serial_to_dl(inf[i].serial, super);
9907 if (dl->index == -1) {
9908 dl->index = mpb->num_disks;
9909 mpb->num_disks++;
9910 dl->disk.status |= CONFIGURED_DISK;
9911 dl->disk.status &= ~SPARE_DISK;
9912 }
9913 set_imsm_ord_tbl_ent(new_map, i, dl->index);
9914 }
9915
ba2de7ba
DW		 9916 dv = update->space;
9917 dev = dv->dev;
949c47a0
DW		 9918 update->space = NULL;
9919 imsm_copy_dev(dev, &u->dev);
ba2de7ba
DW		 9920 dv->index = u->dev_idx;
9921 dv->next = super->devlist;
9922 super->devlist = dv;
8273f55e 9923 mpb->num_raid_devs++;
8273f55e 9924
4d1313e9 9925 imsm_update_version_info(super);
8273f55e 9926 break;
ba2de7ba
DW		 9927 create_error:
9928 /* mdmon knows how to release update->space, but not
9929 * ((struct intel_dev *) update->space)->dev
9930 */
9931 if (update->space) {
9932 dv = update->space;
9933 free(dv->dev);
9934 }
8273f55e 9935 break;
e8319a19 9936 }
33414a01
DW		 9937 case update_kill_array: {
9938 struct imsm_update_kill_array *u = (void *) update->buf;
9939 int victim = u->dev_idx;
9940 struct active_array *a;
9941 struct intel_dev **dp;
9942 struct imsm_dev *dev;
9943
9944 /* sanity check that we are not affecting the uuid of
9945 * active arrays, or deleting an active array
9946 *
9947 * FIXME when immutable ids are available, but note that
9948 * we'll also need to fixup the invalidated/active
9949 * subarray indexes in mdstat
9950 */
9951 for (a = st->arrays; a; a = a->next)
9952 if (a->info.container_member >= victim)
9953 break;
9954 /* by definition if mdmon is running at least one array
9955 * is active in the container, so checking
9956 * mpb->num_raid_devs is just extra paranoia
9957 */
9958 dev = get_imsm_dev(super, victim);
9959 if (a || !dev || mpb->num_raid_devs == 1) {
9960 dprintf("failed to delete subarray-%d\n", victim);
9961 break;
9962 }
9963
9964 for (dp = &super->devlist; *dp;)
f21e18ca 9965 if ((*dp)->index == (unsigned)super->current_vol) {
33414a01
DW		 9966 *dp = (*dp)->next;
9967 } else {
f21e18ca 9968 if ((*dp)->index > (unsigned)victim)
33414a01
DW		 9969 (*dp)->index--;
9970 dp = &(*dp)->next;
9971 }
9972 mpb->num_raid_devs--;
9973 super->updates_pending++;
9974 break;
9975 }
aa534678
DW		 9976 case update_rename_array: {
9977 struct imsm_update_rename_array *u = (void *) update->buf;
9978 char name[MAX_RAID_SERIAL_LEN+1];
9979 int target = u->dev_idx;
9980 struct active_array *a;
9981 struct imsm_dev *dev;
9982
9983 /* sanity check that we are not affecting the uuid of
9984 * an active array
9985 */
40659392 9986 memset(name, 0, sizeof(name));
aa534678
DW		 9987 snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name);
9988 name[MAX_RAID_SERIAL_LEN] = '\0';
9989 for (a = st->arrays; a; a = a->next)
9990 if (a->info.container_member == target)
9991 break;
9992 dev = get_imsm_dev(super, u->dev_idx);
9993 if (a || !dev || !check_name(super, name, 1)) {
9994 dprintf("failed to rename subarray-%d\n", target);
9995 break;
9996 }
9997
40659392 9998 memcpy(dev->volume, name, MAX_RAID_SERIAL_LEN);
aa534678
DW		 9999 super->updates_pending++;
10000 break;
10001 }
1a64be56 10002 case update_add_remove_disk: {
43dad3d6 10003 /* we may be able to repair some arrays if disks are
095b8088 10004 * being added, check the status of add_remove_disk
1a64be56
LM		 10005 * if discs has been added.
10006 */
10007 if (add_remove_disk_update(super)) {
43dad3d6 10008 struct active_array *a;
072b727f
DW		 10009
10010 super->updates_pending++;
1a64be56 10011 for (a = st->arrays; a; a = a->next)
43dad3d6
DW		 10012 a->check_degraded = 1;
10013 }
43dad3d6 10014 break;
e8319a19 10015 }
bbab0940
TM		 10016 case update_prealloc_badblocks_mem:
10017 break;
e6e9dd3f
AP		 10018 case update_rwh_policy: {
10019 struct imsm_update_rwh_policy *u = (void *)update->buf;
10020 int target = u->dev_idx;
10021 struct imsm_dev *dev = get_imsm_dev(super, target);
10022 if (!dev) {
10023 dprintf("could not find subarray-%d\n", target);
10024 break;
10025 }
10026
10027 if (dev->rwh_policy != u->new_policy) {
10028 dev->rwh_policy = u->new_policy;
10029 super->updates_pending++;
10030 }
10031 break;
10032 }
1a64be56 10033 default:
ebf3be99 10034 pr_err("error: unsupported process update type:(type: %d)\n", type);
1a64be56 10035 }
e8319a19 10036}
88758e9d 10037
bc0b9d34
PC		 10038static struct mdinfo *get_spares_for_grow(struct supertype *st);
10039
5fe6f031
N		 10040static int imsm_prepare_update(struct supertype *st,
10041 struct metadata_update *update)
8273f55e 10042{
949c47a0 10043 /**
4d7b1503
DW		 10044 * Allocate space to hold new disk entries, raid-device entries or a new
10045 * mpb if necessary. The manager synchronously waits for updates to
10046 * complete in the monitor, so new mpb buffers allocated here can be
10047 * integrated by the monitor thread without worrying about live pointers
10048 * in the manager thread.
8273f55e 10049 */
095b8088 10050 enum imsm_update_type type;
4d7b1503 10051 struct intel_super *super = st->sb;
f36a9ecd 10052 unsigned int sector_size = super->sector_size;
4d7b1503
DW		 10053 struct imsm_super *mpb = super->anchor;
10054 size_t buf_len;
10055 size_t len = 0;
949c47a0 10056
095b8088
N		 10057 if (update->len < (int)sizeof(type))
10058 return 0;
10059
10060 type = *(enum imsm_update_type *) update->buf;
10061
949c47a0 10062 switch (type) {
0ec5d470 10063 case update_general_migration_checkpoint:
095b8088
N		 10064 if (update->len < (int)sizeof(struct imsm_update_general_migration_checkpoint))
10065 return 0;
1ade5cc1 10066 dprintf("called for update_general_migration_checkpoint\n");
0ec5d470 10067 break;
abedf5fc
KW		 10068 case update_takeover: {
10069 struct imsm_update_takeover *u = (void *)update->buf;
095b8088
N		 10070 if (update->len < (int)sizeof(*u))
10071 return 0;
abedf5fc
KW		 10072 if (u->direction == R0_TO_R10) {
10073 void **tail = (void **)&update->space_list;
10074 struct imsm_dev *dev = get_imsm_dev(super, u->subarray);
238c0a71 10075 struct imsm_map *map = get_imsm_map(dev, MAP_0);
abedf5fc
KW		 10076 int num_members = map->num_members;
10077 void *space;
10078 int size, i;
abedf5fc
KW		 10079 /* allocate memory for added disks */
10080 for (i = 0; i < num_members; i++) {
10081 size = sizeof(struct dl);
503975b9 10082 space = xmalloc(size);
abedf5fc
KW		 10083 *tail = space;
10084 tail = space;
10085 *tail = NULL;
10086 }
10087 /* allocate memory for new device */
10088 size = sizeof_imsm_dev(super->devlist->dev, 0) +
10089 (num_members * sizeof(__u32));
503975b9
N		 10090 space = xmalloc(size);
10091 *tail = space;
10092 tail = space;
10093 *tail = NULL;
10094 len = disks_to_mpb_size(num_members * 2);
abedf5fc
KW		 10095 }
10096
10097 break;
10098 }
78b10e66 10099 case update_reshape_container_disks: {
d195167d
AK		 10100 /* Every raid device in the container is about to
10101 * gain some more devices, and we will enter a
10102 * reconfiguration.
10103 * So each 'imsm_map' will be bigger, and the imsm_vol
10104 * will now hold 2 of them.
10105 * Thus we need new 'struct imsm_dev' allocations sized
10106 * as sizeof_imsm_dev but with more devices in both maps.
10107 */
10108 struct imsm_update_reshape *u = (void *)update->buf;
10109 struct intel_dev *dl;
10110 void **space_tail = (void**)&update->space_list;
10111
095b8088
N		 10112 if (update->len < (int)sizeof(*u))
10113 return 0;
10114
1ade5cc1 10115 dprintf("for update_reshape\n");
d195167d
AK		 10116
10117 for (dl = super->devlist; dl; dl = dl->next) {
10118 int size = sizeof_imsm_dev(dl->dev, 1);
10119 void *s;
d677e0b8
AK		 10120 if (u->new_raid_disks > u->old_raid_disks)
10121 size += sizeof(__u32)*2*
10122 (u->new_raid_disks - u->old_raid_disks);
503975b9 10123 s = xmalloc(size);
d195167d
AK		 10124 *space_tail = s;
10125 space_tail = s;
10126 *space_tail = NULL;
10127 }
10128
10129 len = disks_to_mpb_size(u->new_raid_disks);
10130 dprintf("New anchor length is %llu\n", (unsigned long long)len);
78b10e66
N		 10131 break;
10132 }
48c5303a 10133 case update_reshape_migration: {
bc0b9d34
PC		 10134 /* for migration level 0->5 we need to add disks
10135 * so the same as for container operation we will copy
10136 * device to the bigger location.
10137 * in memory prepared device and new disk area are prepared
10138 * for usage in process update
10139 */
10140 struct imsm_update_reshape_migration *u = (void *)update->buf;
10141 struct intel_dev *id;
10142 void **space_tail = (void **)&update->space_list;
10143 int size;
10144 void *s;
10145 int current_level = -1;
10146
095b8088
N		 10147 if (update->len < (int)sizeof(*u))
10148 return 0;
10149
1ade5cc1 10150 dprintf("for update_reshape\n");
bc0b9d34
PC		 10151
10152 /* add space for bigger array in update
10153 */
10154 for (id = super->devlist; id; id = id->next) {
10155 if (id->index == (unsigned)u->subdev) {
10156 size = sizeof_imsm_dev(id->dev, 1);
10157 if (u->new_raid_disks > u->old_raid_disks)
10158 size += sizeof(__u32)*2*
10159 (u->new_raid_disks - u->old_raid_disks);
503975b9 10160 s = xmalloc(size);
bc0b9d34
PC		 10161 *space_tail = s;
10162 space_tail = s;
10163 *space_tail = NULL;
10164 break;
10165 }
10166 }
10167 if (update->space_list == NULL)
10168 break;
10169
10170 /* add space for disk in update
10171 */
10172 size = sizeof(struct dl);
503975b9 10173 s = xmalloc(size);
bc0b9d34
PC		 10174 *space_tail = s;
10175 space_tail = s;
10176 *space_tail = NULL;
10177
10178 /* add spare device to update
10179 */
10180 for (id = super->devlist ; id; id = id->next)
10181 if (id->index == (unsigned)u->subdev) {
10182 struct imsm_dev *dev;
10183 struct imsm_map *map;
10184
10185 dev = get_imsm_dev(super, u->subdev);
238c0a71 10186 map = get_imsm_map(dev, MAP_0);
bc0b9d34
PC		 10187 current_level = map->raid_level;
10188 break;
10189 }
089f9d79 10190 if (u->new_level == 5 && u->new_level != current_level) {
bc0b9d34
PC		 10191 struct mdinfo *spares;
10192
10193 spares = get_spares_for_grow(st);
10194 if (spares) {
10195 struct dl *dl;
10196 struct mdinfo *dev;
10197
10198 dev = spares->devs;
10199 if (dev) {
10200 u->new_disks[0] =
10201 makedev(dev->disk.major,
10202 dev->disk.minor);
10203 dl = get_disk_super(super,
10204 dev->disk.major,
10205 dev->disk.minor);
10206 dl->index = u->old_raid_disks;
10207 dev = dev->next;
10208 }
10209 sysfs_free(spares);
10210 }
10211 }
10212 len = disks_to_mpb_size(u->new_raid_disks);
10213 dprintf("New anchor length is %llu\n", (unsigned long long)len);
48c5303a
PC		 10214 break;
10215 }
f3871fdc 10216 case update_size_change: {
095b8088
N		 10217 if (update->len < (int)sizeof(struct imsm_update_size_change))
10218 return 0;
10219 break;
10220 }
10221 case update_activate_spare: {
10222 if (update->len < (int)sizeof(struct imsm_update_activate_spare))
10223 return 0;
f3871fdc
AK		 10224 break;
10225 }
949c47a0
DW		 10226 case update_create_array: {
10227 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 10228 struct intel_dev *dv;
54c2c1ea 10229 struct imsm_dev *dev = &u->dev;
238c0a71 10230 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW		 10231 struct dl *dl;
10232 struct disk_info *inf;
10233 int i;
10234 int activate = 0;
949c47a0 10235
095b8088
N		 10236 if (update->len < (int)sizeof(*u))
10237 return 0;
10238
54c2c1ea
DW		 10239 inf = get_disk_info(u);
10240 len = sizeof_imsm_dev(dev, 1);
ba2de7ba 10241 /* allocate a new super->devlist entry */
503975b9
N		 10242 dv = xmalloc(sizeof(*dv));
10243 dv->dev = xmalloc(len);
10244 update->space = dv;
949c47a0 10245
54c2c1ea
DW		 10246 /* count how many spares will be converted to members */
10247 for (i = 0; i < map->num_members; i++) {
10248 dl = serial_to_dl(inf[i].serial, super);
10249 if (!dl) {
10250 /* hmm maybe it failed?, nothing we can do about
10251 * it here
10252 */
10253 continue;
10254 }
10255 if (count_memberships(dl, super) == 0)
10256 activate++;
10257 }
10258 len += activate * sizeof(struct imsm_disk);
949c47a0 10259 break;
095b8088
N		 10260 }
10261 case update_kill_array: {
10262 if (update->len < (int)sizeof(struct imsm_update_kill_array))
10263 return 0;
949c47a0
DW		 10264 break;
10265 }
095b8088
N		 10266 case update_rename_array: {
10267 if (update->len < (int)sizeof(struct imsm_update_rename_array))
10268 return 0;
10269 break;
10270 }
10271 case update_add_remove_disk:
10272 /* no update->len needed */
10273 break;
bbab0940
TM		 10274 case update_prealloc_badblocks_mem:
10275 super->extra_space += sizeof(struct bbm_log) -
10276 get_imsm_bbm_log_size(super->bbm_log);
10277 break;
e6e9dd3f
AP		 10278 case update_rwh_policy: {
10279 if (update->len < (int)sizeof(struct imsm_update_rwh_policy))
10280 return 0;
10281 break;
10282 }
095b8088
N		 10283 default:
10284 return 0;
949c47a0 10285 }
8273f55e 10286
4d7b1503
DW		 10287 /* check if we need a larger metadata buffer */
10288 if (super->next_buf)
10289 buf_len = super->next_len;
10290 else
10291 buf_len = super->len;
10292
bbab0940 10293 if (__le32_to_cpu(mpb->mpb_size) + super->extra_space + len > buf_len) {
4d7b1503
DW		 10294 /* ok we need a larger buf than what is currently allocated
10295 * if this allocation fails process_update will notice that
10296 * ->next_len is set and ->next_buf is NULL
10297 */
bbab0940
TM		 10298 buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) +
10299 super->extra_space + len, sector_size);
4d7b1503
DW		 10300 if (super->next_buf)
10301 free(super->next_buf);
10302
10303 super->next_len = buf_len;
f36a9ecd 10304 if (posix_memalign(&super->next_buf, sector_size, buf_len) == 0)
1f45a8ad
DW		 10305 memset(super->next_buf, 0, buf_len);
10306 else
4d7b1503
DW		 10307 super->next_buf = NULL;
10308 }
5fe6f031 10309 return 1;
8273f55e
DW		 10310}
10311
ae6aad82 10312/* must be called while manager is quiesced */
f21e18ca 10313static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index)
ae6aad82
DW		 10314{
10315 struct imsm_super *mpb = super->anchor;
ae6aad82
DW		 10316 struct dl *iter;
10317 struct imsm_dev *dev;
10318 struct imsm_map *map;
4c9e8c1e 10319 unsigned int i, j, num_members;
fb12a745 10320 __u32 ord, ord_map0;
4c9e8c1e 10321 struct bbm_log *log = super->bbm_log;
ae6aad82 10322
1ade5cc1 10323 dprintf("deleting device[%d] from imsm_super\n", index);
ae6aad82
DW		 10324
10325 /* shift all indexes down one */
10326 for (iter = super->disks; iter; iter = iter->next)
f21e18ca 10327 if (iter->index > (int)index)
ae6aad82 10328 iter->index--;
47ee5a45 10329 for (iter = super->missing; iter; iter = iter->next)
f21e18ca 10330 if (iter->index > (int)index)
47ee5a45 10331 iter->index--;
ae6aad82
DW		 10332
10333 for (i = 0; i < mpb->num_raid_devs; i++) {
10334 dev = get_imsm_dev(super, i);
238c0a71 10335 map = get_imsm_map(dev, MAP_0);
24565c9a
DW		 10336 num_members = map->num_members;
10337 for (j = 0; j < num_members; j++) {
10338 /* update ord entries being careful not to propagate
10339 * ord-flags to the first map
10340 */
238c0a71 10341 ord = get_imsm_ord_tbl_ent(dev, j, MAP_X);
fb12a745 10342 ord_map0 = get_imsm_ord_tbl_ent(dev, j, MAP_0);
ae6aad82 10343
24565c9a
DW		 10344 if (ord_to_idx(ord) <= index)
10345 continue;
ae6aad82 10346
238c0a71 10347 map = get_imsm_map(dev, MAP_0);
fb12a745 10348 set_imsm_ord_tbl_ent(map, j, ord_map0 - 1);
238c0a71 10349 map = get_imsm_map(dev, MAP_1);
24565c9a
DW		 10350 if (map)
10351 set_imsm_ord_tbl_ent(map, j, ord - 1);
ae6aad82
DW		 10352 }
10353 }
10354
4c9e8c1e
TM		 10355 for (i = 0; i < log->entry_count; i++) {
10356 struct bbm_log_entry *entry = &log->marked_block_entries[i];
10357
10358 if (entry->disk_ordinal <= index)
10359 continue;
10360 entry->disk_ordinal--;
10361 }
10362
ae6aad82
DW		 10363 mpb->num_disks--;
10364 super->updates_pending++;
24565c9a
DW		 10365 if (*dlp) {
10366 struct dl *dl = *dlp;
10367
10368 *dlp = (*dlp)->next;
10369 __free_imsm_disk(dl);
10370 }
ae6aad82 10371}
9a717282 10372
9a717282
AK		 10373static int imsm_get_allowed_degradation(int level, int raid_disks,
10374 struct intel_super *super,
10375 struct imsm_dev *dev)
10376{
10377 switch (level) {
bf5cf7c7 10378 case 1:
9a717282
AK		 10379 case 10:{
10380 int ret_val = 0;
10381 struct imsm_map *map;
10382 int i;
10383
10384 ret_val = raid_disks/2;
10385 /* check map if all disks pairs not failed
10386 * in both maps
10387 */
238c0a71 10388 map = get_imsm_map(dev, MAP_0);
9a717282
AK		 10389 for (i = 0; i < ret_val; i++) {
10390 int degradation = 0;
10391 if (get_imsm_disk(super, i) == NULL)
10392 degradation++;
10393 if (get_imsm_disk(super, i + 1) == NULL)
10394 degradation++;
10395 if (degradation == 2)
10396 return 0;
10397 }
238c0a71 10398 map = get_imsm_map(dev, MAP_1);
9a717282
AK		 10399 /* if there is no second map
10400 * result can be returned
10401 */
10402 if (map == NULL)
10403 return ret_val;
10404 /* check degradation in second map
10405 */
10406 for (i = 0; i < ret_val; i++) {
10407 int degradation = 0;
10408 if (get_imsm_disk(super, i) == NULL)
10409 degradation++;
10410 if (get_imsm_disk(super, i + 1) == NULL)
10411 degradation++;
10412 if (degradation == 2)
10413 return 0;
10414 }
10415 return ret_val;
10416 }
10417 case 5:
10418 return 1;
10419 case 6:
10420 return 2;
10421 default:
10422 return 0;
10423 }
10424}
10425
d31ad643
PB		 10426/*******************************************************************************
10427 * Function: validate_container_imsm
10428 * Description: This routine validates container after assemble,
10429 * eg. if devices in container are under the same controller.
10430 *
10431 * Parameters:
10432 * info : linked list with info about devices used in array
10433 * Returns:
10434 * 1 : HBA mismatch
10435 * 0 : Success
10436 ******************************************************************************/
10437int validate_container_imsm(struct mdinfo *info)
10438{
6b781d33
AP		 10439 if (check_env("IMSM_NO_PLATFORM"))
10440 return 0;
d31ad643 10441
6b781d33
AP		 10442 struct sys_dev *idev;
10443 struct sys_dev *hba = NULL;
10444 struct sys_dev *intel_devices = find_intel_devices();
10445 char *dev_path = devt_to_devpath(makedev(info->disk.major,
10446 info->disk.minor));
10447
10448 for (idev = intel_devices; idev; idev = idev->next) {
10449 if (dev_path && strstr(dev_path, idev->path)) {
10450 hba = idev;
10451 break;
d31ad643 10452 }
6b781d33
AP		 10453 }
10454 if (dev_path)
d31ad643
PB		 10455 free(dev_path);
10456
6b781d33
AP		 10457 if (!hba) {
10458 pr_err("WARNING - Cannot detect HBA for device %s!\n",
10459 devid2kname(makedev(info->disk.major, info->disk.minor)));
10460 return 1;
10461 }
10462
10463 const struct imsm_orom *orom = get_orom_by_device_id(hba->dev_id);
10464 struct mdinfo *dev;
10465
10466 for (dev = info->next; dev; dev = dev->next) {
10467 dev_path = devt_to_devpath(makedev(dev->disk.major, dev->disk.minor));
10468
10469 struct sys_dev *hba2 = NULL;
10470 for (idev = intel_devices; idev; idev = idev->next) {
10471 if (dev_path && strstr(dev_path, idev->path)) {
10472 hba2 = idev;
10473 break;
d31ad643
PB		 10474 }
10475 }
6b781d33
AP		 10476 if (dev_path)
10477 free(dev_path);
10478
10479 const struct imsm_orom *orom2 = hba2 == NULL ? NULL :
10480 get_orom_by_device_id(hba2->dev_id);
10481
10482 if (hba2 && hba->type != hba2->type) {
10483 pr_err("WARNING - HBAs of devices do not match %s != %s\n",
10484 get_sys_dev_type(hba->type), get_sys_dev_type(hba2->type));
10485 return 1;
10486 }
10487
07cb1e57 10488 if (orom != orom2) {
6b781d33
AP		 10489 pr_err("WARNING - IMSM container assembled with disks under different HBAs!\n"
10490 " This operation is not supported and can lead to data loss.\n");
10491 return 1;
10492 }
10493
10494 if (!orom) {
10495 pr_err("WARNING - IMSM container assembled with disks under HBAs without IMSM platform support!\n"
10496 " This operation is not supported and can lead to data loss.\n");
10497 return 1;
10498 }
d31ad643 10499 }
6b781d33 10500
d31ad643
PB		 10501 return 0;
10502}
32141c17 10503
6f50473f
TM		 10504/*******************************************************************************
10505* Function: imsm_record_badblock
10506* Description: This routine stores new bad block record in 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_record_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 = record_new_badblock(super->bbm_log, ord_to_idx(ord), sector,
10529 length);
10530 if (ret)
10531 super->updates_pending++;
10532
10533 return ret;
10534}
c07a5a4f
TM		 10535/*******************************************************************************
10536* Function: imsm_clear_badblock
10537* Description: This routine clears bad block record from BBM log
10538*
10539* Parameters:
10540* a : array containing a bad block
10541* slot : disk number containing a bad block
10542* sector : bad block sector
10543* length : bad block sectors range
10544* Returns:
10545* 1 : Success
10546* 0 : Error
10547******************************************************************************/
10548static int imsm_clear_badblock(struct active_array *a, int slot,
10549 unsigned long long sector, int length)
10550{
10551 struct intel_super *super = a->container->sb;
10552 int ord;
10553 int ret;
10554
10555 ord = imsm_disk_slot_to_ord(a, slot);
10556 if (ord < 0)
10557 return 0;
10558
10559 ret = clear_badblock(super->bbm_log, ord_to_idx(ord), sector, length);
10560 if (ret)
10561 super->updates_pending++;
10562
10563 return ret;
10564}
928f1424
TM		 10565/*******************************************************************************
10566* Function: imsm_get_badblocks
10567* Description: This routine get list of bad blocks for an array
10568*
10569* Parameters:
10570* a : array
10571* slot : disk number
10572* Returns:
10573* bb : structure containing bad blocks
10574* NULL : error
10575******************************************************************************/
10576static struct md_bb *imsm_get_badblocks(struct active_array *a, int slot)
10577{
10578 int inst = a->info.container_member;
10579 struct intel_super *super = a->container->sb;
10580 struct imsm_dev *dev = get_imsm_dev(super, inst);
10581 struct imsm_map *map = get_imsm_map(dev, MAP_0);
10582 int ord;
10583
10584 ord = imsm_disk_slot_to_ord(a, slot);
10585 if (ord < 0)
10586 return NULL;
10587
10588 get_volume_badblocks(super->bbm_log, ord_to_idx(ord), pba_of_lba0(map),
44490938 10589 per_dev_array_size(map), &super->bb);
928f1424
TM		 10590
10591 return &super->bb;
10592}
27156a57
TM		 10593/*******************************************************************************
10594* Function: examine_badblocks_imsm
10595* Description: Prints list of bad blocks on a disk to the standard output
10596*
10597* Parameters:
10598* st : metadata handler
10599* fd : open file descriptor for device
10600* devname : device name
10601* Returns:
10602* 0 : Success
10603* 1 : Error
10604******************************************************************************/
10605static int examine_badblocks_imsm(struct supertype *st, int fd, char *devname)
10606{
10607 struct intel_super *super = st->sb;
10608 struct bbm_log *log = super->bbm_log;
10609 struct dl *d = NULL;
10610 int any = 0;
10611
10612 for (d = super->disks; d ; d = d->next) {
10613 if (strcmp(d->devname, devname) == 0)
10614 break;
10615 }
10616
10617 if ((d == NULL) || (d->index < 0)) { /* serial mismatch probably */
10618 pr_err("%s doesn't appear to be part of a raid array\n",
10619 devname);
10620 return 1;
10621 }
10622
10623 if (log != NULL) {
10624 unsigned int i;
10625 struct bbm_log_entry *entry = &log->marked_block_entries[0];
10626
10627 for (i = 0; i < log->entry_count; i++) {
10628 if (entry[i].disk_ordinal == d->index) {
10629 unsigned long long sector = __le48_to_cpu(
10630 &entry[i].defective_block_start);
10631 int cnt = entry[i].marked_count + 1;
10632
10633 if (!any) {
10634 printf("Bad-blocks on %s:\n", devname);
10635 any = 1;
10636 }
10637
10638 printf("%20llu for %d sectors\n", sector, cnt);
10639 }
10640 }
10641 }
10642
10643 if (!any)
10644 printf("No bad-blocks list configured on %s\n", devname);
10645
10646 return 0;
10647}
687629c2
AK		 10648/*******************************************************************************
10649 * Function: init_migr_record_imsm
10650 * Description: Function inits imsm migration record
10651 * Parameters:
10652 * super : imsm internal array info
10653 * dev : device under migration
10654 * info : general array info to find the smallest device
10655 * Returns:
10656 * none
10657 ******************************************************************************/
10658void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev,
10659 struct mdinfo *info)
10660{
10661 struct intel_super *super = st->sb;
10662 struct migr_record *migr_rec = super->migr_rec;
10663 int new_data_disks;
10664 unsigned long long dsize, dev_sectors;
10665 long long unsigned min_dev_sectors = -1LLU;
238c0a71
AK		 10666 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
10667 struct imsm_map *map_src = get_imsm_map(dev, MAP_1);
687629c2 10668 unsigned long long num_migr_units;
3ef4403c 10669 unsigned long long array_blocks;
2f86fda3 10670 struct dl *dl_disk = NULL;
687629c2
AK		 10671
10672 memset(migr_rec, 0, sizeof(struct migr_record));
10673 migr_rec->family_num = __cpu_to_le32(super->anchor->family_num);
10674
10675 /* only ascending reshape supported now */
10676 migr_rec->ascending_migr = __cpu_to_le32(1);
10677
10678 migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE /
10679 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
e1742195
AK		 10680 migr_rec->dest_depth_per_unit *=
10681 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
9529d343 10682 new_data_disks = imsm_num_data_members(map_dest);
687629c2
AK		 10683 migr_rec->blocks_per_unit =
10684 __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks);
10685 migr_rec->dest_depth_per_unit =
10686 __cpu_to_le32(migr_rec->dest_depth_per_unit);
3ef4403c 10687 array_blocks = info->component_size * new_data_disks;
687629c2
AK		 10688 num_migr_units =
10689 array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit);
10690
10691 if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit))
10692 num_migr_units++;
9f421827 10693 set_num_migr_units(migr_rec, num_migr_units);
687629c2
AK		 10694
10695 migr_rec->post_migr_vol_cap = dev->size_low;
10696 migr_rec->post_migr_vol_cap_hi = dev->size_high;
10697
687629c2 10698 /* Find the smallest dev */
2f86fda3
MT		 10699 for (dl_disk = super->disks; dl_disk ; dl_disk = dl_disk->next) {
10700 /* ignore spares in container */
10701 if (dl_disk->index < 0)
687629c2 10702 continue;
2f86fda3 10703 get_dev_size(dl_disk->fd, NULL, &dsize);
687629c2
AK		 10704 dev_sectors = dsize / 512;
10705 if (dev_sectors < min_dev_sectors)
10706 min_dev_sectors = dev_sectors;
687629c2 10707 }
9f421827 10708 set_migr_chkp_area_pba(migr_rec, min_dev_sectors -
687629c2
AK		 10709 RAID_DISK_RESERVED_BLOCKS_IMSM_HI);
10710
10711 write_imsm_migr_rec(st);
10712
10713 return;
10714}
10715
10716/*******************************************************************************
10717 * Function: save_backup_imsm
10718 * Description: Function saves critical data stripes to Migration Copy Area
10719 * and updates the current migration unit status.
10720 * Use restore_stripes() to form a destination stripe,
10721 * and to write it to the Copy Area.
10722 * Parameters:
10723 * st : supertype information
aea93171 10724 * dev : imsm device that backup is saved for
687629c2
AK		 10725 * info : general array info
10726 * buf : input buffer
687629c2
AK		 10727 * length : length of data to backup (blocks_per_unit)
10728 * Returns:
10729 * 0 : success
10730 *, -1 : fail
10731 ******************************************************************************/
10732int save_backup_imsm(struct supertype *st,
10733 struct imsm_dev *dev,
10734 struct mdinfo *info,
10735 void *buf,
687629c2
AK		 10736 int length)
10737{
10738 int rv = -1;
10739 struct intel_super *super = st->sb;
594dc1b8
JS		 10740 unsigned long long *target_offsets;
10741 int *targets;
687629c2 10742 int i;
238c0a71 10743 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
687629c2 10744 int new_disks = map_dest->num_members;
ab724b98
AK		 10745 int dest_layout = 0;
10746 int dest_chunk;
d1877f69 10747 unsigned long long start;
9529d343 10748 int data_disks = imsm_num_data_members(map_dest);
687629c2 10749
503975b9 10750 targets = xmalloc(new_disks * sizeof(int));
687629c2 10751
2f86fda3
MT		 10752 for (i = 0; i < new_disks; i++) {
10753 struct dl *dl_disk = get_imsm_dl_disk(super, i);
10754
10755 targets[i] = dl_disk->fd;
10756 }
7e45b550 10757
503975b9 10758 target_offsets = xcalloc(new_disks, sizeof(unsigned long long));
687629c2 10759
d1877f69 10760 start = info->reshape_progress * 512;
687629c2 10761 for (i = 0; i < new_disks; i++) {
9f421827 10762 target_offsets[i] = migr_chkp_area_pba(super->migr_rec) * 512;
d1877f69
AK		 10763 /* move back copy area adderss, it will be moved forward
10764 * in restore_stripes() using start input variable
10765 */
10766 target_offsets[i] -= start/data_disks;
687629c2
AK		 10767 }
10768
68eb8bc6 10769 dest_layout = imsm_level_to_layout(map_dest->raid_level);
ab724b98
AK		 10770 dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512;
10771
687629c2
AK		 10772 if (restore_stripes(targets, /* list of dest devices */
10773 target_offsets, /* migration record offsets */
10774 new_disks,
ab724b98
AK		 10775 dest_chunk,
10776 map_dest->raid_level,
10777 dest_layout,
10778 -1, /* source backup file descriptor */
10779 0, /* input buf offset
10780 * always 0 buf is already offseted */
d1877f69 10781 start,
687629c2
AK		 10782 length,
10783 buf) != 0) {
e7b84f9d 10784 pr_err("Error restoring stripes\n");
687629c2
AK		 10785 goto abort;
10786 }
10787
10788 rv = 0;
10789
10790abort:
10791 if (targets) {
687629c2
AK		 10792 free(targets);
10793 }
10794 free(target_offsets);
10795
10796 return rv;
10797}
10798
10799/*******************************************************************************
10800 * Function: save_checkpoint_imsm
10801 * Description: Function called for current unit status update
10802 * in the migration record. It writes it to disk.
10803 * Parameters:
10804 * super : imsm internal array info
10805 * info : general array info
10806 * Returns:
10807 * 0: success
10808 * 1: failure
0228d92c
AK		 10809 * 2: failure, means no valid migration record
10810 * / no general migration in progress /
687629c2
AK		 10811 ******************************************************************************/
10812int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state)
10813{
10814 struct intel_super *super = st->sb;
f8b72ef5
AK		 10815 unsigned long long blocks_per_unit;
10816 unsigned long long curr_migr_unit;
10817
2f86fda3 10818 if (load_imsm_migr_rec(super) != 0) {
7a862a02 10819 dprintf("imsm: ERROR: Cannot read migration record for checkpoint save.\n");
2e062e82
AK		 10820 return 1;
10821 }
10822
f8b72ef5
AK		 10823 blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit);
10824 if (blocks_per_unit == 0) {
0228d92c
AK		 10825 dprintf("imsm: no migration in progress.\n");
10826 return 2;
687629c2 10827 }
f8b72ef5
AK		 10828 curr_migr_unit = info->reshape_progress / blocks_per_unit;
10829 /* check if array is alligned to copy area
10830 * if it is not alligned, add one to current migration unit value
10831 * this can happend on array reshape finish only
10832 */
10833 if (info->reshape_progress % blocks_per_unit)
10834 curr_migr_unit++;
687629c2 10835
9f421827 10836 set_current_migr_unit(super->migr_rec, curr_migr_unit);
687629c2 10837 super->migr_rec->rec_status = __cpu_to_le32(state);
9f421827
PB		 10838 set_migr_dest_1st_member_lba(super->migr_rec,
10839 super->migr_rec->dest_depth_per_unit * curr_migr_unit);
10840
687629c2 10841 if (write_imsm_migr_rec(st) < 0) {
7a862a02 10842 dprintf("imsm: Cannot write migration record outside backup area\n");
687629c2
AK		 10843 return 1;
10844 }
10845
10846 return 0;
10847}
10848
276d77db
AK		 10849/*******************************************************************************
10850 * Function: recover_backup_imsm
10851 * Description: Function recovers critical data from the Migration Copy Area
10852 * while assembling an array.
10853 * Parameters:
10854 * super : imsm internal array info
10855 * info : general array info
10856 * Returns:
10857 * 0 : success (or there is no data to recover)
10858 * 1 : fail
10859 ******************************************************************************/
10860int recover_backup_imsm(struct supertype *st, struct mdinfo *info)
10861{
10862 struct intel_super *super = st->sb;
10863 struct migr_record *migr_rec = super->migr_rec;
594dc1b8 10864 struct imsm_map *map_dest;
276d77db
AK		 10865 struct intel_dev *id = NULL;
10866 unsigned long long read_offset;
10867 unsigned long long write_offset;
10868 unsigned unit_len;
2f86fda3 10869 int new_disks, err;
276d77db
AK		 10870 char *buf = NULL;
10871 int retval = 1;
f36a9ecd 10872 unsigned int sector_size = super->sector_size;
9f421827
PB		 10873 unsigned long curr_migr_unit = current_migr_unit(migr_rec);
10874 unsigned long num_migr_units = get_num_migr_units(migr_rec);
276d77db 10875 char buffer[20];
6c3560c0 10876 int skipped_disks = 0;
2f86fda3 10877 struct dl *dl_disk;
276d77db
AK		 10878
10879 err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20);
10880 if (err < 1)
10881 return 1;
10882
10883 /* recover data only during assemblation */
10884 if (strncmp(buffer, "inactive", 8) != 0)
10885 return 0;
10886 /* no data to recover */
10887 if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL)
10888 return 0;
10889 if (curr_migr_unit >= num_migr_units)
10890 return 1;
10891
10892 /* find device during reshape */
10893 for (id = super->devlist; id; id = id->next)
10894 if (is_gen_migration(id->dev))
10895 break;
10896 if (id == NULL)
10897 return 1;
10898
238c0a71 10899 map_dest = get_imsm_map(id->dev, MAP_0);
276d77db
AK		 10900 new_disks = map_dest->num_members;
10901
9f421827 10902 read_offset = migr_chkp_area_pba(migr_rec) * 512;
276d77db 10903
9f421827 10904 write_offset = (migr_dest_1st_member_lba(migr_rec) +
5551b113 10905 pba_of_lba0(map_dest)) * 512;
276d77db
AK		 10906
10907 unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
f36a9ecd 10908 if (posix_memalign((void **)&buf, sector_size, unit_len) != 0)
276d77db 10909 goto abort;
276d77db 10910
2f86fda3
MT		 10911 for (dl_disk = super->disks; dl_disk; dl_disk = dl_disk->next) {
10912 if (dl_disk->index < 0)
10913 continue;
276d77db 10914
2f86fda3 10915 if (dl_disk->fd < 0) {
6c3560c0
AK		 10916 skipped_disks++;
10917 continue;
10918 }
2f86fda3 10919 if (lseek64(dl_disk->fd, read_offset, SEEK_SET) < 0) {
e7b84f9d
N		 10920 pr_err("Cannot seek to block: %s\n",
10921 strerror(errno));
137debce
AK		 10922 skipped_disks++;
10923 continue;
276d77db 10924 }
2f86fda3 10925 if (read(dl_disk->fd, buf, unit_len) != unit_len) {
e7b84f9d
N		 10926 pr_err("Cannot read copy area block: %s\n",
10927 strerror(errno));
137debce
AK		 10928 skipped_disks++;
10929 continue;
276d77db 10930 }
2f86fda3 10931 if (lseek64(dl_disk->fd, write_offset, SEEK_SET) < 0) {
e7b84f9d
N		 10932 pr_err("Cannot seek to block: %s\n",
10933 strerror(errno));
137debce
AK		 10934 skipped_disks++;
10935 continue;
276d77db 10936 }
2f86fda3 10937 if (write(dl_disk->fd, buf, unit_len) != unit_len) {
e7b84f9d
N		 10938 pr_err("Cannot restore block: %s\n",
10939 strerror(errno));
137debce
AK		 10940 skipped_disks++;
10941 continue;
276d77db
AK		 10942 }
10943 }
10944
137debce
AK		 10945 if (skipped_disks > imsm_get_allowed_degradation(info->new_level,
10946 new_disks,
10947 super,
10948 id->dev)) {
7a862a02 10949 pr_err("Cannot restore data from backup. Too many failed disks\n");
6c3560c0
AK		 10950 goto abort;
10951 }
10952
befb629b
AK		 10953 if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) {
10954 /* ignore error == 2, this can mean end of reshape here
10955 */
7a862a02 10956 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL) during restart\n");
befb629b 10957 } else
276d77db 10958 retval = 0;
276d77db
AK		 10959
10960abort:
276d77db
AK		 10961 free(buf);
10962 return retval;
10963}
10964
2cda7640
ML		 10965static char disk_by_path[] = "/dev/disk/by-path/";
10966
10967static const char *imsm_get_disk_controller_domain(const char *path)
10968{
2cda7640 10969 char disk_path[PATH_MAX];
96234762
LM		 10970 char *drv=NULL;
10971 struct stat st;
2cda7640 10972
6d8d290a 10973 strcpy(disk_path, disk_by_path);
96234762
LM		 10974 strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1);
10975 if (stat(disk_path, &st) == 0) {
10976 struct sys_dev* hba;
594dc1b8 10977 char *path;
96234762
LM		 10978
10979 path = devt_to_devpath(st.st_rdev);
10980 if (path == NULL)
10981 return "unknown";
10982 hba = find_disk_attached_hba(-1, path);
10983 if (hba && hba->type == SYS_DEV_SAS)
10984 drv = "isci";
10985 else if (hba && hba->type == SYS_DEV_SATA)
10986 drv = "ahci";
c6839718
MT		 10987 else if (hba && hba->type == SYS_DEV_VMD)
10988 drv = "vmd";
10989 else if (hba && hba->type == SYS_DEV_NVME)
10990 drv = "nvme";
1011e834 10991 else
96234762
LM		 10992 drv = "unknown";
10993 dprintf("path: %s hba: %s attached: %s\n",
10994 path, (hba) ? hba->path : "NULL", drv);
10995 free(path);
2cda7640 10996 }
96234762 10997 return drv;
2cda7640
ML		 10998}
10999
4dd2df09 11000static char *imsm_find_array_devnm_by_subdev(int subdev, char *container)
78b10e66 11001{
4dd2df09 11002 static char devnm[32];
78b10e66
N		 11003 char subdev_name[20];
11004 struct mdstat_ent *mdstat;
11005
11006 sprintf(subdev_name, "%d", subdev);
11007 mdstat = mdstat_by_subdev(subdev_name, container);
11008 if (!mdstat)
4dd2df09 11009 return NULL;
78b10e66 11010
4dd2df09 11011 strcpy(devnm, mdstat->devnm);
78b10e66 11012 free_mdstat(mdstat);
4dd2df09 11013 return devnm;
78b10e66
N		 11014}
11015
11016static int imsm_reshape_is_allowed_on_container(struct supertype *st,
11017 struct geo_params *geo,
fbf3d202
AK		 11018 int *old_raid_disks,
11019 int direction)
78b10e66 11020{
694575e7
KW		 11021 /* currently we only support increasing the number of devices
11022 * for a container. This increases the number of device for each
11023 * member array. They must all be RAID0 or RAID5.
11024 */
78b10e66
N		 11025 int ret_val = 0;
11026 struct mdinfo *info, *member;
11027 int devices_that_can_grow = 0;
11028
7a862a02 11029 dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): st->devnm = (%s)\n", st->devnm);
78b10e66 11030
d04f65f4 11031 if (geo->size > 0 ||
78b10e66
N		 11032 geo->level != UnSet ||
11033 geo->layout != UnSet ||
11034 geo->chunksize != 0 ||
11035 geo->raid_disks == UnSet) {
7a862a02 11036 dprintf("imsm: Container operation is allowed for raid disks number change only.\n");
78b10e66
N		 11037 return ret_val;
11038 }
11039
fbf3d202 11040 if (direction == ROLLBACK_METADATA_CHANGES) {
7a862a02 11041 dprintf("imsm: Metadata changes rollback is not supported for container operation.\n");
fbf3d202
AK		 11042 return ret_val;
11043 }
11044
78b10e66
N		 11045 info = container_content_imsm(st, NULL);
11046 for (member = info; member; member = member->next) {
4dd2df09 11047 char *result;
78b10e66
N		 11048
11049 dprintf("imsm: checking device_num: %i\n",
11050 member->container_member);
11051
d7d205bd 11052 if (geo->raid_disks <= member->array.raid_disks) {
78b10e66
N		 11053 /* we work on container for Online Capacity Expansion
11054 * only so raid_disks has to grow
11055 */
7a862a02 11056 dprintf("imsm: for container operation raid disks increase is required\n");
78b10e66
N		 11057 break;
11058 }
11059
089f9d79 11060 if (info->array.level != 0 && info->array.level != 5) {
78b10e66
N		 11061 /* we cannot use this container with other raid level
11062 */
7a862a02 11063 dprintf("imsm: for container operation wrong raid level (%i) detected\n",
78b10e66
N		 11064 info->array.level);
11065 break;
11066 } else {
11067 /* check for platform support
11068 * for this raid level configuration
11069 */
11070 struct intel_super *super = st->sb;
11071 if (!is_raid_level_supported(super->orom,
11072 member->array.level,
11073 geo->raid_disks)) {
7a862a02 11074 dprintf("platform does not support raid%d with %d disk%s\n",
78b10e66
N		 11075 info->array.level,
11076 geo->raid_disks,
11077 geo->raid_disks > 1 ? "s" : "");
11078 break;
11079 }
2a4a08e7
AK		 11080 /* check if component size is aligned to chunk size
11081 */
11082 if (info->component_size %
11083 (info->array.chunk_size/512)) {
7a862a02 11084 dprintf("Component size is not aligned to chunk size\n");
2a4a08e7
AK		 11085 break;
11086 }
78b10e66
N		 11087 }
11088
11089 if (*old_raid_disks &&
11090 info->array.raid_disks != *old_raid_disks)
11091 break;
11092 *old_raid_disks = info->array.raid_disks;
11093
11094 /* All raid5 and raid0 volumes in container
11095 * have to be ready for Online Capacity Expansion
11096 * so they need to be assembled. We have already
11097 * checked that no recovery etc is happening.
11098 */
4dd2df09
N		 11099 result = imsm_find_array_devnm_by_subdev(member->container_member,
11100 st->container_devnm);
11101 if (result == NULL) {
78b10e66
N		 11102 dprintf("imsm: cannot find array\n");
11103 break;
11104 }
11105 devices_that_can_grow++;
11106 }
11107 sysfs_free(info);
11108 if (!member && devices_that_can_grow)
11109 ret_val = 1;
11110
11111 if (ret_val)
1ade5cc1 11112 dprintf("Container operation allowed\n");
78b10e66 11113 else
1ade5cc1 11114 dprintf("Error: %i\n", ret_val);
78b10e66
N		 11115
11116 return ret_val;
11117}
11118
11119/* Function: get_spares_for_grow
11120 * Description: Allocates memory and creates list of spare devices
1011e834 11121 * avaliable in container. Checks if spare drive size is acceptable.
78b10e66
N		 11122 * Parameters: Pointer to the supertype structure
11123 * Returns: Pointer to the list of spare devices (mdinfo structure) on success,
1011e834 11124 * NULL if fail
78b10e66
N		 11125 */
11126static struct mdinfo *get_spares_for_grow(struct supertype *st)
11127{
fbfdcb06
AO		 11128 struct spare_criteria sc;
11129
11130 get_spare_criteria_imsm(st, &sc);
11131 return container_choose_spares(st, &sc, NULL, NULL, NULL, 0);
78b10e66
N		 11132}
11133
11134/******************************************************************************
11135 * function: imsm_create_metadata_update_for_reshape
11136 * Function creates update for whole IMSM container.
11137 *
11138 ******************************************************************************/
11139static int imsm_create_metadata_update_for_reshape(
11140 struct supertype *st,
11141 struct geo_params *geo,
11142 int old_raid_disks,
11143 struct imsm_update_reshape **updatep)
11144{
11145 struct intel_super *super = st->sb;
11146 struct imsm_super *mpb = super->anchor;
594dc1b8
JS		 11147 int update_memory_size;
11148 struct imsm_update_reshape *u;
11149 struct mdinfo *spares;
78b10e66 11150 int i;
594dc1b8 11151 int delta_disks;
bbd24d86 11152 struct mdinfo *dev;
78b10e66 11153
1ade5cc1 11154 dprintf("(enter) raid_disks = %i\n", geo->raid_disks);
78b10e66
N		 11155
11156 delta_disks = geo->raid_disks - old_raid_disks;
11157
11158 /* size of all update data without anchor */
11159 update_memory_size = sizeof(struct imsm_update_reshape);
11160
11161 /* now add space for spare disks that we need to add. */
11162 update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1);
11163
503975b9 11164 u = xcalloc(1, update_memory_size);
78b10e66
N		 11165 u->type = update_reshape_container_disks;
11166 u->old_raid_disks = old_raid_disks;
11167 u->new_raid_disks = geo->raid_disks;
11168
11169 /* now get spare disks list
11170 */
11171 spares = get_spares_for_grow(st);
11172
d7be7d87 11173 if (spares == NULL || delta_disks > spares->array.spare_disks) {
7a862a02 11174 pr_err("imsm: ERROR: Cannot get spare devices for %s.\n", geo->dev_name);
e4c72d1d 11175 i = -1;
78b10e66
N		 11176 goto abort;
11177 }
11178
11179 /* we have got spares
11180 * update disk list in imsm_disk list table in anchor
11181 */
11182 dprintf("imsm: %i spares are available.\n\n",
11183 spares->array.spare_disks);
11184
bbd24d86 11185 dev = spares->devs;
78b10e66 11186 for (i = 0; i < delta_disks; i++) {
78b10e66
N		 11187 struct dl *dl;
11188
bbd24d86
AK		 11189 if (dev == NULL)
11190 break;
78b10e66
N		 11191 u->new_disks[i] = makedev(dev->disk.major,
11192 dev->disk.minor);
11193 dl = get_disk_super(super, dev->disk.major, dev->disk.minor);
ee4beede
AK		 11194 dl->index = mpb->num_disks;
11195 mpb->num_disks++;
bbd24d86 11196 dev = dev->next;
78b10e66 11197 }
78b10e66
N		 11198
11199abort:
11200 /* free spares
11201 */
11202 sysfs_free(spares);
11203
d677e0b8 11204 dprintf("imsm: reshape update preparation :");
78b10e66 11205 if (i == delta_disks) {
1ade5cc1 11206 dprintf_cont(" OK\n");
78b10e66
N		 11207 *updatep = u;
11208 return update_memory_size;
11209 }
11210 free(u);
1ade5cc1 11211 dprintf_cont(" Error\n");
78b10e66
N		 11212
11213 return 0;
11214}
11215
f3871fdc
AK		 11216/******************************************************************************
11217 * function: imsm_create_metadata_update_for_size_change()
11218 * Creates update for IMSM array for array size change.
11219 *
11220 ******************************************************************************/
11221static int imsm_create_metadata_update_for_size_change(
11222 struct supertype *st,
11223 struct geo_params *geo,
11224 struct imsm_update_size_change **updatep)
11225{
11226 struct intel_super *super = st->sb;
594dc1b8
JS		 11227 int update_memory_size;
11228 struct imsm_update_size_change *u;
f3871fdc 11229
1ade5cc1 11230 dprintf("(enter) New size = %llu\n", geo->size);
f3871fdc
AK		 11231
11232 /* size of all update data without anchor */
11233 update_memory_size = sizeof(struct imsm_update_size_change);
11234
503975b9 11235 u = xcalloc(1, update_memory_size);
f3871fdc
AK		 11236 u->type = update_size_change;
11237 u->subdev = super->current_vol;
11238 u->new_size = geo->size;
11239
11240 dprintf("imsm: reshape update preparation : OK\n");
11241 *updatep = u;
11242
11243 return update_memory_size;
11244}
11245
48c5303a
PC		 11246/******************************************************************************
11247 * function: imsm_create_metadata_update_for_migration()
11248 * Creates update for IMSM array.
11249 *
11250 ******************************************************************************/
11251static int imsm_create_metadata_update_for_migration(
11252 struct supertype *st,
11253 struct geo_params *geo,
11254 struct imsm_update_reshape_migration **updatep)
11255{
11256 struct intel_super *super = st->sb;
594dc1b8
JS		 11257 int update_memory_size;
11258 struct imsm_update_reshape_migration *u;
48c5303a
PC		 11259 struct imsm_dev *dev;
11260 int previous_level = -1;
11261
1ade5cc1 11262 dprintf("(enter) New Level = %i\n", geo->level);
48c5303a
PC		 11263
11264 /* size of all update data without anchor */
11265 update_memory_size = sizeof(struct imsm_update_reshape_migration);
11266
503975b9 11267 u = xcalloc(1, update_memory_size);
48c5303a
PC		 11268 u->type = update_reshape_migration;
11269 u->subdev = super->current_vol;
11270 u->new_level = geo->level;
11271 u->new_layout = geo->layout;
11272 u->new_raid_disks = u->old_raid_disks = geo->raid_disks;
11273 u->new_disks[0] = -1;
4bba0439 11274 u->new_chunksize = -1;
48c5303a
PC		 11275
11276 dev = get_imsm_dev(super, u->subdev);
11277 if (dev) {
11278 struct imsm_map *map;
11279
238c0a71 11280 map = get_imsm_map(dev, MAP_0);
4bba0439
PC		 11281 if (map) {
11282 int current_chunk_size =
11283 __le16_to_cpu(map->blocks_per_strip) / 2;
11284
11285 if (geo->chunksize != current_chunk_size) {
11286 u->new_chunksize = geo->chunksize / 1024;
7a862a02 11287 dprintf("imsm: chunk size change from %i to %i\n",
4bba0439
PC		 11288 current_chunk_size, u->new_chunksize);
11289 }
48c5303a 11290 previous_level = map->raid_level;
4bba0439 11291 }
48c5303a 11292 }
089f9d79 11293 if (geo->level == 5 && previous_level == 0) {
48c5303a
PC		 11294 struct mdinfo *spares = NULL;
11295
11296 u->new_raid_disks++;
11297 spares = get_spares_for_grow(st);
089f9d79 11298 if (spares == NULL || spares->array.spare_disks < 1) {
48c5303a
PC		 11299 free(u);
11300 sysfs_free(spares);
11301 update_memory_size = 0;
565cc99e 11302 pr_err("cannot get spare device for requested migration\n");
48c5303a
PC		 11303 return 0;
11304 }
11305 sysfs_free(spares);
11306 }
11307 dprintf("imsm: reshape update preparation : OK\n");
11308 *updatep = u;
11309
11310 return update_memory_size;
11311}
11312
8dd70bce
AK		 11313static void imsm_update_metadata_locally(struct supertype *st,
11314 void *buf, int len)
11315{
11316 struct metadata_update mu;
11317
11318 mu.buf = buf;
11319 mu.len = len;
11320 mu.space = NULL;
11321 mu.space_list = NULL;
11322 mu.next = NULL;
5fe6f031
N		 11323 if (imsm_prepare_update(st, &mu))
11324 imsm_process_update(st, &mu);
8dd70bce
AK		 11325
11326 while (mu.space_list) {
11327 void **space = mu.space_list;
11328 mu.space_list = *space;
11329 free(space);
11330 }
11331}
78b10e66 11332
471bceb6 11333/***************************************************************************
694575e7 11334* Function: imsm_analyze_change
471bceb6 11335* Description: Function analyze change for single volume
1011e834 11336* and validate if transition is supported
fbf3d202
AK		 11337* Parameters: Geometry parameters, supertype structure,
11338* metadata change direction (apply/rollback)
694575e7 11339* Returns: Operation type code on success, -1 if fail
471bceb6
KW		 11340****************************************************************************/
11341enum imsm_reshape_type imsm_analyze_change(struct supertype *st,
fbf3d202
AK		 11342 struct geo_params *geo,
11343 int direction)
694575e7 11344{
471bceb6
KW		 11345 struct mdinfo info;
11346 int change = -1;
11347 int check_devs = 0;
c21e737b 11348 int chunk;
67a2db32
AK		 11349 /* number of added/removed disks in operation result */
11350 int devNumChange = 0;
11351 /* imsm compatible layout value for array geometry verification */
11352 int imsm_layout = -1;
7abc9871
AK		 11353 int data_disks;
11354 struct imsm_dev *dev;
9529d343 11355 struct imsm_map *map;
7abc9871 11356 struct intel_super *super;
d04f65f4 11357 unsigned long long current_size;
65d38cca 11358 unsigned long long free_size;
d04f65f4 11359 unsigned long long max_size;
65d38cca 11360 int rv;
471bceb6
KW		 11361
11362 getinfo_super_imsm_volume(st, &info, NULL);
089f9d79
JS		 11363 if (geo->level != info.array.level && geo->level >= 0 &&
11364 geo->level != UnSet) {
471bceb6
KW		 11365 switch (info.array.level) {
11366 case 0:
11367 if (geo->level == 5) {
b5347799 11368 change = CH_MIGRATION;
e13ce846 11369 if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) {
7a862a02 11370 pr_err("Error. Requested Layout not supported (left-asymmetric layout is supported only)!\n");
e13ce846
AK		 11371 change = -1;
11372 goto analyse_change_exit;
11373 }
67a2db32 11374 imsm_layout = geo->layout;
471bceb6 11375 check_devs = 1;
e91a3bad
LM		 11376 devNumChange = 1; /* parity disk added */
11377 } else if (geo->level == 10) {
471bceb6
KW		 11378 change = CH_TAKEOVER;
11379 check_devs = 1;
e91a3bad 11380 devNumChange = 2; /* two mirrors added */
67a2db32 11381 imsm_layout = 0x102; /* imsm supported layout */
471bceb6 11382 }
dfe77a9e
KW		 11383 break;
11384 case 1:
471bceb6
KW		 11385 case 10:
11386 if (geo->level == 0) {
11387 change = CH_TAKEOVER;
11388 check_devs = 1;
e91a3bad 11389 devNumChange = -(geo->raid_disks/2);
67a2db32 11390 imsm_layout = 0; /* imsm raid0 layout */
471bceb6
KW		 11391 }
11392 break;
11393 }
11394 if (change == -1) {
7a862a02 11395 pr_err("Error. Level Migration from %d to %d not supported!\n",
e7b84f9d 11396 info.array.level, geo->level);
471bceb6
KW		 11397 goto analyse_change_exit;
11398 }
11399 } else
11400 geo->level = info.array.level;
11401
089f9d79
JS		 11402 if (geo->layout != info.array.layout &&
11403 (geo->layout != UnSet && geo->layout != -1)) {
b5347799 11404 change = CH_MIGRATION;
089f9d79
JS		 11405 if (info.array.layout == 0 && info.array.level == 5 &&
11406 geo->layout == 5) {
471bceb6 11407 /* reshape 5 -> 4 */
089f9d79
JS		 11408 } else if (info.array.layout == 5 && info.array.level == 5 &&
11409 geo->layout == 0) {
471bceb6
KW		 11410 /* reshape 4 -> 5 */
11411 geo->layout = 0;
11412 geo->level = 5;
11413 } else {
7a862a02 11414 pr_err("Error. Layout Migration from %d to %d not supported!\n",
e7b84f9d 11415 info.array.layout, geo->layout);
471bceb6
KW		 11416 change = -1;
11417 goto analyse_change_exit;
11418 }
67a2db32 11419 } else {
471bceb6 11420 geo->layout = info.array.layout;
67a2db32
AK		 11421 if (imsm_layout == -1)
11422 imsm_layout = info.array.layout;
11423 }
471bceb6 11424
089f9d79
JS		 11425 if (geo->chunksize > 0 && geo->chunksize != UnSet &&
11426 geo->chunksize != info.array.chunk_size) {
2d2b0eb7
MD		 11427 if (info.array.level == 10) {
11428 pr_err("Error. Chunk size change for RAID 10 is not supported.\n");
11429 change = -1;
11430 goto analyse_change_exit;
1e9b2c3f
PB		 11431 } else if (info.component_size % (geo->chunksize/512)) {
11432 pr_err("New chunk size (%dK) does not evenly divide device size (%lluk). Aborting...\n",
11433 geo->chunksize/1024, info.component_size/2);
11434 change = -1;
11435 goto analyse_change_exit;
2d2b0eb7 11436 }
b5347799 11437 change = CH_MIGRATION;
2d2b0eb7 11438 } else {
471bceb6 11439 geo->chunksize = info.array.chunk_size;
2d2b0eb7 11440 }
471bceb6 11441
c21e737b 11442 chunk = geo->chunksize / 1024;
7abc9871
AK		 11443
11444 super = st->sb;
11445 dev = get_imsm_dev(super, super->current_vol);
9529d343
MD		 11446 map = get_imsm_map(dev, MAP_0);
11447 data_disks = imsm_num_data_members(map);
c41e00b2 11448 /* compute current size per disk member
7abc9871 11449 */
c41e00b2
AK		 11450 current_size = info.custom_array_size / data_disks;
11451
089f9d79 11452 if (geo->size > 0 && geo->size != MAX_SIZE) {
c41e00b2
AK		 11453 /* align component size
11454 */
3e684231 11455 geo->size = imsm_component_size_alignment_check(
c41e00b2 11456 get_imsm_raid_level(dev->vol.map),
f36a9ecd 11457 chunk * 1024, super->sector_size,
c41e00b2 11458 geo->size * 2);
65d0b4ce 11459 if (geo->size == 0) {
7a862a02 11460 pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is 0).\n",
65d0b4ce
LD		 11461 current_size);
11462 goto analyse_change_exit;
11463 }
c41e00b2 11464 }
7abc9871 11465
089f9d79 11466 if (current_size != geo->size && geo->size > 0) {
7abc9871 11467 if (change != -1) {
7a862a02 11468 pr_err("Error. Size change should be the only one at a time.\n");
7abc9871
AK		 11469 change = -1;
11470 goto analyse_change_exit;
11471 }
11472 if ((super->current_vol + 1) != super->anchor->num_raid_devs) {
7a862a02 11473 pr_err("Error. The last volume in container can be expanded only (%i/%s).\n",
4dd2df09 11474 super->current_vol, st->devnm);
7abc9871
AK		 11475 goto analyse_change_exit;
11476 }
65d38cca
LD		 11477 /* check the maximum available size
11478 */
11479 rv = imsm_get_free_size(st, dev->vol.map->num_members,
11480 0, chunk, &free_size);
11481 if (rv == 0)
11482 /* Cannot find maximum available space
11483 */
11484 max_size = 0;
11485 else {
11486 max_size = free_size + current_size;
11487 /* align component size
11488 */
3e684231 11489 max_size = imsm_component_size_alignment_check(
65d38cca 11490 get_imsm_raid_level(dev->vol.map),
f36a9ecd 11491 chunk * 1024, super->sector_size,
65d38cca
LD		 11492 max_size);
11493 }
d04f65f4 11494 if (geo->size == MAX_SIZE) {
b130333f
AK		 11495 /* requested size change to the maximum available size
11496 */
65d38cca 11497 if (max_size == 0) {
7a862a02 11498 pr_err("Error. Cannot find maximum available space.\n");
b130333f
AK		 11499 change = -1;
11500 goto analyse_change_exit;
65d38cca
LD		 11501 } else
11502 geo->size = max_size;
c41e00b2 11503 }
b130333f 11504
681b7ae2 11505 if (direction == ROLLBACK_METADATA_CHANGES) {
fbf3d202
AK		 11506 /* accept size for rollback only
11507 */
11508 } else {
11509 /* round size due to metadata compatibility
11510 */
11511 geo->size = (geo->size >> SECT_PER_MB_SHIFT)
11512 << SECT_PER_MB_SHIFT;
11513 dprintf("Prepare update for size change to %llu\n",
11514 geo->size );
11515 if (current_size >= geo->size) {
7a862a02 11516 pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is %llu).\n",
e7b84f9d 11517 current_size, geo->size);
fbf3d202
AK		 11518 goto analyse_change_exit;
11519 }
65d38cca 11520 if (max_size && geo->size > max_size) {
7a862a02 11521 pr_err("Error. Requested size is larger than maximum available size (maximum available size is %llu, requested size /rounded/ is %llu).\n",
e7b84f9d 11522 max_size, geo->size);
65d38cca
LD		 11523 goto analyse_change_exit;
11524 }
7abc9871
AK		 11525 }
11526 geo->size *= data_disks;
11527 geo->raid_disks = dev->vol.map->num_members;
11528 change = CH_ARRAY_SIZE;
11529 }
471bceb6
KW		 11530 if (!validate_geometry_imsm(st,
11531 geo->level,
67a2db32 11532 imsm_layout,
e91a3bad 11533 geo->raid_disks + devNumChange,
c21e737b 11534 &chunk,
af4348dd 11535 geo->size, INVALID_SECTORS,
5308f117 11536 0, 0, info.consistency_policy, 1))
471bceb6
KW		 11537 change = -1;
11538
11539 if (check_devs) {
11540 struct intel_super *super = st->sb;
11541 struct imsm_super *mpb = super->anchor;
11542
11543 if (mpb->num_raid_devs > 1) {
7a862a02 11544 pr_err("Error. Cannot perform operation on %s- for this operation it MUST be single array in container\n",
e7b84f9d 11545 geo->dev_name);
471bceb6
KW		 11546 change = -1;
11547 }
11548 }
11549
11550analyse_change_exit:
089f9d79
JS		 11551 if (direction == ROLLBACK_METADATA_CHANGES &&
11552 (change == CH_MIGRATION || change == CH_TAKEOVER)) {
7a862a02 11553 dprintf("imsm: Metadata changes rollback is not supported for migration and takeover operations.\n");
fbf3d202
AK		 11554 change = -1;
11555 }
471bceb6 11556 return change;
694575e7
KW		 11557}
11558
bb025c2f
KW		 11559int imsm_takeover(struct supertype *st, struct geo_params *geo)
11560{
11561 struct intel_super *super = st->sb;
11562 struct imsm_update_takeover *u;
11563
503975b9 11564 u = xmalloc(sizeof(struct imsm_update_takeover));
bb025c2f
KW		 11565
11566 u->type = update_takeover;
11567 u->subarray = super->current_vol;
11568
11569 /* 10->0 transition */
11570 if (geo->level == 0)
11571 u->direction = R10_TO_R0;
11572
0529c688
KW		 11573 /* 0->10 transition */
11574 if (geo->level == 10)
11575 u->direction = R0_TO_R10;
11576
bb025c2f
KW		 11577 /* update metadata locally */
11578 imsm_update_metadata_locally(st, u,
11579 sizeof(struct imsm_update_takeover));
11580 /* and possibly remotely */
11581 if (st->update_tail)
11582 append_metadata_update(st, u,
11583 sizeof(struct imsm_update_takeover));
11584 else
11585 free(u);
11586
11587 return 0;
11588}
11589
895ffd99
MT		 11590/* Flush size update if size calculated by num_data_stripes is higher than
11591 * imsm_dev_size to eliminate differences during reshape.
11592 * Mdmon will recalculate them correctly.
11593 * If subarray index is not set then check whole container.
11594 * Returns:
11595 * 0 - no error occurred
11596 * 1 - error detected
11597 */
11598static int imsm_fix_size_mismatch(struct supertype *st, int subarray_index)
11599{
11600 struct intel_super *super = st->sb;
11601 int tmp = super->current_vol;
11602 int ret_val = 1;
11603 int i;
11604
11605 for (i = 0; i < super->anchor->num_raid_devs; i++) {
11606 if (subarray_index >= 0 && i != subarray_index)
11607 continue;
11608 super->current_vol = i;
11609 struct imsm_dev *dev = get_imsm_dev(super, super->current_vol);
11610 struct imsm_map *map = get_imsm_map(dev, MAP_0);
11611 unsigned int disc_count = imsm_num_data_members(map);
11612 struct geo_params geo;
11613 struct imsm_update_size_change *update;
11614 unsigned long long calc_size = per_dev_array_size(map) * disc_count;
11615 unsigned long long d_size = imsm_dev_size(dev);
11616 int u_size;
11617
11618 if (calc_size == d_size || dev->vol.migr_type == MIGR_GEN_MIGR)
11619 continue;
11620
11621 /* There is a difference, verify that imsm_dev_size is
11622 * rounded correctly and push update.
11623 */
11624 if (d_size != round_size_to_mb(d_size, disc_count)) {
11625 dprintf("imsm: Size of volume %d is not rounded correctly\n",
11626 i);
11627 goto exit;
11628 }
11629 memset(&geo, 0, sizeof(struct geo_params));
11630 geo.size = d_size;
11631 u_size = imsm_create_metadata_update_for_size_change(st, &geo,
11632 &update);
11633 if (u_size < 1) {
11634 dprintf("imsm: Cannot prepare size change update\n");
11635 goto exit;
11636 }
11637 imsm_update_metadata_locally(st, update, u_size);
11638 if (st->update_tail) {
11639 append_metadata_update(st, update, u_size);
11640 flush_metadata_updates(st);
11641 st->update_tail = &st->updates;
11642 } else {
11643 imsm_sync_metadata(st);
11644 }
11645 }
11646 ret_val = 0;
11647exit:
11648 super->current_vol = tmp;
11649 return ret_val;
11650}
11651
d04f65f4
N		 11652static int imsm_reshape_super(struct supertype *st, unsigned long long size,
11653 int level,
78b10e66 11654 int layout, int chunksize, int raid_disks,
41784c88 11655 int delta_disks, char *backup, char *dev,
016e00f5 11656 int direction, int verbose)
78b10e66 11657{
78b10e66
N		 11658 int ret_val = 1;
11659 struct geo_params geo;
11660
1ade5cc1 11661 dprintf("(enter)\n");
78b10e66 11662
71204a50 11663 memset(&geo, 0, sizeof(struct geo_params));
78b10e66
N		 11664
11665 geo.dev_name = dev;
4dd2df09 11666 strcpy(geo.devnm, st->devnm);
78b10e66
N		 11667 geo.size = size;
11668 geo.level = level;
11669 geo.layout = layout;
11670 geo.chunksize = chunksize;
11671 geo.raid_disks = raid_disks;
41784c88
AK		 11672 if (delta_disks != UnSet)
11673 geo.raid_disks += delta_disks;
78b10e66 11674
1ade5cc1
N		 11675 dprintf("for level : %i\n", geo.level);
11676 dprintf("for raid_disks : %i\n", geo.raid_disks);
78b10e66 11677
4dd2df09 11678 if (strcmp(st->container_devnm, st->devnm) == 0) {
694575e7
KW		 11679 /* On container level we can only increase number of devices. */
11680 dprintf("imsm: info: Container operation\n");
78b10e66 11681 int old_raid_disks = 0;
6dc0be30 11682
78b10e66 11683 if (imsm_reshape_is_allowed_on_container(
fbf3d202 11684 st, &geo, &old_raid_disks, direction)) {
78b10e66
N		 11685 struct imsm_update_reshape *u = NULL;
11686 int len;
11687
895ffd99
MT		 11688 if (imsm_fix_size_mismatch(st, -1)) {
11689 dprintf("imsm: Cannot fix size mismatch\n");
11690 goto exit_imsm_reshape_super;
11691 }
11692
78b10e66
N		 11693 len = imsm_create_metadata_update_for_reshape(
11694 st, &geo, old_raid_disks, &u);
11695
ed08d51c
AK		 11696 if (len <= 0) {
11697 dprintf("imsm: Cannot prepare update\n");
11698 goto exit_imsm_reshape_super;
11699 }
11700
8dd70bce
AK		 11701 ret_val = 0;
11702 /* update metadata locally */
11703 imsm_update_metadata_locally(st, u, len);
11704 /* and possibly remotely */
11705 if (st->update_tail)
11706 append_metadata_update(st, u, len);
11707 else
ed08d51c 11708 free(u);
8dd70bce 11709
694575e7 11710 } else {
7a862a02 11711 pr_err("(imsm) Operation is not allowed on this container\n");
694575e7
KW		 11712 }
11713 } else {
11714 /* On volume level we support following operations
471bceb6
KW		 11715 * - takeover: raid10 -> raid0; raid0 -> raid10
11716 * - chunk size migration
11717 * - migration: raid5 -> raid0; raid0 -> raid5
11718 */
11719 struct intel_super *super = st->sb;
11720 struct intel_dev *dev = super->devlist;
4dd2df09 11721 int change;
694575e7 11722 dprintf("imsm: info: Volume operation\n");
471bceb6
KW		 11723 /* find requested device */
11724 while (dev) {
1011e834 11725 char *devnm =
4dd2df09
N		 11726 imsm_find_array_devnm_by_subdev(
11727 dev->index, st->container_devnm);
11728 if (devnm && strcmp(devnm, geo.devnm) == 0)
471bceb6
KW		 11729 break;
11730 dev = dev->next;
11731 }
11732 if (dev == NULL) {
4dd2df09
N		 11733 pr_err("Cannot find %s (%s) subarray\n",
11734 geo.dev_name, geo.devnm);
471bceb6
KW		 11735 goto exit_imsm_reshape_super;
11736 }
11737 super->current_vol = dev->index;
fbf3d202 11738 change = imsm_analyze_change(st, &geo, direction);
694575e7 11739 switch (change) {
471bceb6 11740 case CH_TAKEOVER:
bb025c2f 11741 ret_val = imsm_takeover(st, &geo);
694575e7 11742 break;
48c5303a
PC		 11743 case CH_MIGRATION: {
11744 struct imsm_update_reshape_migration *u = NULL;
11745 int len =
11746 imsm_create_metadata_update_for_migration(
11747 st, &geo, &u);
11748 if (len < 1) {
7a862a02 11749 dprintf("imsm: Cannot prepare update\n");
48c5303a
PC		 11750 break;
11751 }
471bceb6 11752 ret_val = 0;
48c5303a
PC		 11753 /* update metadata locally */
11754 imsm_update_metadata_locally(st, u, len);
11755 /* and possibly remotely */
11756 if (st->update_tail)
11757 append_metadata_update(st, u, len);
11758 else
11759 free(u);
11760 }
11761 break;
7abc9871 11762 case CH_ARRAY_SIZE: {
f3871fdc
AK		 11763 struct imsm_update_size_change *u = NULL;
11764 int len =
11765 imsm_create_metadata_update_for_size_change(
11766 st, &geo, &u);
11767 if (len < 1) {
7a862a02 11768 dprintf("imsm: Cannot prepare update\n");
f3871fdc
AK		 11769 break;
11770 }
11771 ret_val = 0;
11772 /* update metadata locally */
11773 imsm_update_metadata_locally(st, u, len);
11774 /* and possibly remotely */
11775 if (st->update_tail)
11776 append_metadata_update(st, u, len);
11777 else
11778 free(u);
7abc9871
AK		 11779 }
11780 break;
471bceb6
KW		 11781 default:
11782 ret_val = 1;
694575e7 11783 }
694575e7 11784 }
78b10e66 11785
ed08d51c 11786exit_imsm_reshape_super:
78b10e66
N		 11787 dprintf("imsm: reshape_super Exit code = %i\n", ret_val);
11788 return ret_val;
11789}
2cda7640 11790
0febb20c
AO		 11791#define COMPLETED_OK 0
11792#define COMPLETED_NONE 1
11793#define COMPLETED_DELAYED 2
11794
11795static int read_completed(int fd, unsigned long long *val)
11796{
11797 int ret;
11798 char buf[50];
11799
11800 ret = sysfs_fd_get_str(fd, buf, 50);
11801 if (ret < 0)
11802 return ret;
11803
11804 ret = COMPLETED_OK;
11805 if (strncmp(buf, "none", 4) == 0) {
11806 ret = COMPLETED_NONE;
11807 } else if (strncmp(buf, "delayed", 7) == 0) {
11808 ret = COMPLETED_DELAYED;
11809 } else {
11810 char *ep;
11811 *val = strtoull(buf, &ep, 0);
11812 if (ep == buf || (*ep != 0 && *ep != '\n' && *ep != ' '))
11813 ret = -1;
11814 }
11815 return ret;
11816}
11817
eee67a47
AK		 11818/*******************************************************************************
11819 * Function: wait_for_reshape_imsm
11820 * Description: Function writes new sync_max value and waits until
11821 * reshape process reach new position
11822 * Parameters:
11823 * sra : general array info
eee67a47
AK		 11824 * ndata : number of disks in new array's layout
11825 * Returns:
11826 * 0 : success,
11827 * 1 : there is no reshape in progress,
11828 * -1 : fail
11829 ******************************************************************************/
ae9f01f8 11830int wait_for_reshape_imsm(struct mdinfo *sra, int ndata)
eee67a47 11831{
85ca499c 11832 int fd = sysfs_get_fd(sra, NULL, "sync_completed");
df2647fa 11833 int retry = 3;
eee67a47 11834 unsigned long long completed;
ae9f01f8
AK		 11835 /* to_complete : new sync_max position */
11836 unsigned long long to_complete = sra->reshape_progress;
11837 unsigned long long position_to_set = to_complete / ndata;
eee67a47 11838
ae9f01f8 11839 if (fd < 0) {
1ade5cc1 11840 dprintf("cannot open reshape_position\n");
eee67a47 11841 return 1;
ae9f01f8 11842 }
eee67a47 11843
df2647fa
PB		 11844 do {
11845 if (sysfs_fd_get_ll(fd, &completed) < 0) {
11846 if (!retry) {
11847 dprintf("cannot read reshape_position (no reshape in progres)\n");
11848 close(fd);
11849 return 1;
11850 }
11851 usleep(30000);
11852 } else
11853 break;
11854 } while (retry--);
eee67a47 11855
85ca499c 11856 if (completed > position_to_set) {
1ade5cc1 11857 dprintf("wrong next position to set %llu (%llu)\n",
85ca499c 11858 to_complete, position_to_set);
ae9f01f8
AK		 11859 close(fd);
11860 return -1;
11861 }
11862 dprintf("Position set: %llu\n", position_to_set);
11863 if (sysfs_set_num(sra, NULL, "sync_max",
11864 position_to_set) != 0) {
1ade5cc1 11865 dprintf("cannot set reshape position to %llu\n",
ae9f01f8
AK		 11866 position_to_set);
11867 close(fd);
11868 return -1;
eee67a47
AK		 11869 }
11870
eee67a47 11871 do {
0febb20c 11872 int rc;
eee67a47 11873 char action[20];
5ff3a780 11874 int timeout = 3000;
0febb20c 11875
5ff3a780 11876 sysfs_wait(fd, &timeout);
a47e44fb
AK		 11877 if (sysfs_get_str(sra, NULL, "sync_action",
11878 action, 20) > 0 &&
d7d3809a 11879 strncmp(action, "reshape", 7) != 0) {
b2be2b62
AO		 11880 if (strncmp(action, "idle", 4) == 0)
11881 break;
d7d3809a
AP		 11882 close(fd);
11883 return -1;
11884 }
0febb20c
AO		 11885
11886 rc = read_completed(fd, &completed);
11887 if (rc < 0) {
1ade5cc1 11888 dprintf("cannot read reshape_position (in loop)\n");
eee67a47
AK		 11889 close(fd);
11890 return 1;
0febb20c
AO		 11891 } else if (rc == COMPLETED_NONE)
11892 break;
85ca499c 11893 } while (completed < position_to_set);
b2be2b62 11894
eee67a47
AK		 11895 close(fd);
11896 return 0;
eee67a47
AK		 11897}
11898
b915c95f
AK		 11899/*******************************************************************************
11900 * Function: check_degradation_change
11901 * Description: Check that array hasn't become failed.
11902 * Parameters:
11903 * info : for sysfs access
11904 * sources : source disks descriptors
11905 * degraded: previous degradation level
11906 * Returns:
11907 * degradation level
11908 ******************************************************************************/
11909int check_degradation_change(struct mdinfo *info,
11910 int *sources,
11911 int degraded)
11912{
11913 unsigned long long new_degraded;
e1993023
LD		 11914 int rv;
11915
11916 rv = sysfs_get_ll(info, NULL, "degraded", &new_degraded);
089f9d79 11917 if (rv == -1 || (new_degraded != (unsigned long long)degraded)) {
b915c95f
AK		 11918 /* check each device to ensure it is still working */
11919 struct mdinfo *sd;
11920 new_degraded = 0;
11921 for (sd = info->devs ; sd ; sd = sd->next) {
11922 if (sd->disk.state & (1<<MD_DISK_FAULTY))
11923 continue;
11924 if (sd->disk.state & (1<<MD_DISK_SYNC)) {
cf52eff5
TM		 11925 char sbuf[100];
11926
b915c95f 11927 if (sysfs_get_str(info,
cf52eff5 11928 sd, "state", sbuf, sizeof(sbuf)) < 0 ||
b915c95f
AK		 11929 strstr(sbuf, "faulty") ||
11930 strstr(sbuf, "in_sync") == NULL) {
11931 /* this device is dead */
11932 sd->disk.state = (1<<MD_DISK_FAULTY);
11933 if (sd->disk.raid_disk >= 0 &&
11934 sources[sd->disk.raid_disk] >= 0) {
11935 close(sources[
11936 sd->disk.raid_disk]);
11937 sources[sd->disk.raid_disk] =
11938 -1;
11939 }
11940 new_degraded++;
11941 }
11942 }
11943 }
11944 }
11945
11946 return new_degraded;
11947}
11948
10f22854
AK		 11949/*******************************************************************************
11950 * Function: imsm_manage_reshape
11951 * Description: Function finds array under reshape and it manages reshape
11952 * process. It creates stripes backups (if required) and sets
942e1cdb 11953 * checkpoints.
10f22854
AK		 11954 * Parameters:
11955 * afd : Backup handle (nattive) - not used
11956 * sra : general array info
11957 * reshape : reshape parameters - not used
11958 * st : supertype structure
11959 * blocks : size of critical section [blocks]
11960 * fds : table of source device descriptor
11961 * offsets : start of array (offest per devices)
11962 * dests : not used
11963 * destfd : table of destination device descriptor
11964 * destoffsets : table of destination offsets (per device)
11965 * Returns:
11966 * 1 : success, reshape is done
11967 * 0 : fail
11968 ******************************************************************************/
999b4972
N		 11969static int imsm_manage_reshape(
11970 int afd, struct mdinfo *sra, struct reshape *reshape,
10f22854 11971 struct supertype *st, unsigned long backup_blocks,
999b4972
N		 11972 int *fds, unsigned long long *offsets,
11973 int dests, int *destfd, unsigned long long *destoffsets)
11974{
10f22854
AK		 11975 int ret_val = 0;
11976 struct intel_super *super = st->sb;
594dc1b8 11977 struct intel_dev *dv;
de44e46f 11978 unsigned int sector_size = super->sector_size;
10f22854 11979 struct imsm_dev *dev = NULL;
9529d343 11980 struct imsm_map *map_src, *map_dest;
10f22854
AK		 11981 int migr_vol_qan = 0;
11982 int ndata, odata; /* [bytes] */
11983 int chunk; /* [bytes] */
11984 struct migr_record *migr_rec;
11985 char *buf = NULL;
11986 unsigned int buf_size; /* [bytes] */
11987 unsigned long long max_position; /* array size [bytes] */
11988 unsigned long long next_step; /* [blocks]/[bytes] */
11989 unsigned long long old_data_stripe_length;
10f22854
AK		 11990 unsigned long long start_src; /* [bytes] */
11991 unsigned long long start; /* [bytes] */
11992 unsigned long long start_buf_shift; /* [bytes] */
b915c95f 11993 int degraded = 0;
ab724b98 11994 int source_layout = 0;
895ffd99 11995 int subarray_index = -1;
10f22854 11996
79a16a9b
JS		 11997 if (!sra)
11998 return ret_val;
11999
12000 if (!fds || !offsets)
10f22854
AK		 12001 goto abort;
12002
12003 /* Find volume during the reshape */
12004 for (dv = super->devlist; dv; dv = dv->next) {
fc54fe7a
JS		 12005 if (dv->dev->vol.migr_type == MIGR_GEN_MIGR &&
12006 dv->dev->vol.migr_state == 1) {
10f22854
AK		 12007 dev = dv->dev;
12008 migr_vol_qan++;
895ffd99 12009 subarray_index = dv->index;
10f22854
AK		 12010 }
12011 }
12012 /* Only one volume can migrate at the same time */
12013 if (migr_vol_qan != 1) {
676e87a8 12014 pr_err("%s", migr_vol_qan ?
10f22854
AK		 12015 "Number of migrating volumes greater than 1\n" :
12016 "There is no volume during migrationg\n");
12017 goto abort;
12018 }
12019
9529d343 12020 map_dest = get_imsm_map(dev, MAP_0);
238c0a71 12021 map_src = get_imsm_map(dev, MAP_1);
10f22854
AK		 12022 if (map_src == NULL)
12023 goto abort;
10f22854 12024
9529d343
MD		 12025 ndata = imsm_num_data_members(map_dest);
12026 odata = imsm_num_data_members(map_src);
10f22854 12027
7b1ab482 12028 chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512;
10f22854
AK		 12029 old_data_stripe_length = odata * chunk;
12030
12031 migr_rec = super->migr_rec;
12032
10f22854
AK		 12033 /* initialize migration record for start condition */
12034 if (sra->reshape_progress == 0)
12035 init_migr_record_imsm(st, dev, sra);
b2c59438
AK		 12036 else {
12037 if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) {
7a862a02 12038 dprintf("imsm: cannot restart migration when data are present in copy area.\n");
b2c59438
AK		 12039 goto abort;
12040 }
6a75c8ca
AK		 12041 /* Save checkpoint to update migration record for current
12042 * reshape position (in md). It can be farther than current
12043 * reshape position in metadata.
12044 */
12045 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
12046 /* ignore error == 2, this can mean end of reshape here
12047 */
7a862a02 12048 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL, initial save)\n");
6a75c8ca
AK		 12049 goto abort;
12050 }
b2c59438 12051 }
10f22854
AK		 12052
12053 /* size for data */
12054 buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512;
12055 /* extend buffer size for parity disk */
12056 buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
3e684231 12057 /* add space for stripe alignment */
10f22854 12058 buf_size += old_data_stripe_length;
de44e46f
PB		 12059 if (posix_memalign((void **)&buf, MAX_SECTOR_SIZE, buf_size)) {
12060 dprintf("imsm: Cannot allocate checkpoint buffer\n");
10f22854
AK		 12061 goto abort;
12062 }
12063
3ef4403c 12064 max_position = sra->component_size * ndata;
68eb8bc6 12065 source_layout = imsm_level_to_layout(map_src->raid_level);
10f22854 12066
9f421827
PB		 12067 while (current_migr_unit(migr_rec) <
12068 get_num_migr_units(migr_rec)) {
10f22854
AK		 12069 /* current reshape position [blocks] */
12070 unsigned long long current_position =
12071 __le32_to_cpu(migr_rec->blocks_per_unit)
9f421827 12072 * current_migr_unit(migr_rec);
10f22854
AK		 12073 unsigned long long border;
12074
b915c95f
AK		 12075 /* Check that array hasn't become failed.
12076 */
12077 degraded = check_degradation_change(sra, fds, degraded);
12078 if (degraded > 1) {
7a862a02 12079 dprintf("imsm: Abort reshape due to degradation level (%i)\n", degraded);
b915c95f
AK		 12080 goto abort;
12081 }
12082
10f22854
AK		 12083 next_step = __le32_to_cpu(migr_rec->blocks_per_unit);
12084
12085 if ((current_position + next_step) > max_position)
12086 next_step = max_position - current_position;
12087
92144abf 12088 start = current_position * 512;
10f22854 12089
942e1cdb 12090 /* align reading start to old geometry */
10f22854
AK		 12091 start_buf_shift = start % old_data_stripe_length;
12092 start_src = start - start_buf_shift;
12093
12094 border = (start_src / odata) - (start / ndata);
12095 border /= 512;
12096 if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) {
12097 /* save critical stripes to buf
12098 * start - start address of current unit
12099 * to backup [bytes]
12100 * start_src - start address of current unit
12101 * to backup alligned to source array
12102 * [bytes]
12103 */
594dc1b8 12104 unsigned long long next_step_filler;
10f22854
AK		 12105 unsigned long long copy_length = next_step * 512;
12106
12107 /* allign copy area length to stripe in old geometry */
12108 next_step_filler = ((copy_length + start_buf_shift)
12109 % old_data_stripe_length);
12110 if (next_step_filler)
12111 next_step_filler = (old_data_stripe_length
12112 - next_step_filler);
7a862a02 12113 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		 12114 start, start_src, copy_length,
12115 start_buf_shift, next_step_filler);
12116
12117 if (save_stripes(fds, offsets, map_src->num_members,
ab724b98
AK		 12118 chunk, map_src->raid_level,
12119 source_layout, 0, NULL, start_src,
10f22854
AK		 12120 copy_length +
12121 next_step_filler + start_buf_shift,
12122 buf)) {
7a862a02 12123 dprintf("imsm: Cannot save stripes to buffer\n");
10f22854
AK		 12124 goto abort;
12125 }
12126 /* Convert data to destination format and store it
12127 * in backup general migration area
12128 */
12129 if (save_backup_imsm(st, dev, sra,
aea93171 12130 buf + start_buf_shift, copy_length)) {
7a862a02 12131 dprintf("imsm: Cannot save stripes to target devices\n");
10f22854
AK		 12132 goto abort;
12133 }
12134 if (save_checkpoint_imsm(st, sra,
12135 UNIT_SRC_IN_CP_AREA)) {
7a862a02 12136 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_IN_CP_AREA)\n");
10f22854
AK		 12137 goto abort;
12138 }
8016a6d4
AK		 12139 } else {
12140 /* set next step to use whole border area */
12141 border /= next_step;
12142 if (border > 1)
12143 next_step *= border;
10f22854
AK		 12144 }
12145 /* When data backed up, checkpoint stored,
12146 * kick the kernel to reshape unit of data
12147 */
12148 next_step = next_step + sra->reshape_progress;
8016a6d4
AK		 12149 /* limit next step to array max position */
12150 if (next_step > max_position)
12151 next_step = max_position;
10f22854
AK		 12152 sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress);
12153 sysfs_set_num(sra, NULL, "suspend_hi", next_step);
ae9f01f8 12154 sra->reshape_progress = next_step;
10f22854
AK		 12155
12156 /* wait until reshape finish */
c85338c6 12157 if (wait_for_reshape_imsm(sra, ndata)) {
c47b0ff6
AK		 12158 dprintf("wait_for_reshape_imsm returned error!\n");
12159 goto abort;
12160 }
84d11e6c
N		 12161 if (sigterm)
12162 goto abort;
10f22854 12163
0228d92c
AK		 12164 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
12165 /* ignore error == 2, this can mean end of reshape here
12166 */
7a862a02 12167 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL)\n");
10f22854
AK		 12168 goto abort;
12169 }
12170
12171 }
12172
71e5411e
PB		 12173 /* clear migr_rec on disks after successful migration */
12174 struct dl *d;
12175
85337573 12176 memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
71e5411e
PB		 12177 for (d = super->disks; d; d = d->next) {
12178 if (d->index < 0 || is_failed(&d->disk))
12179 continue;
12180 unsigned long long dsize;
12181
12182 get_dev_size(d->fd, NULL, &dsize);
de44e46f 12183 if (lseek64(d->fd, dsize - MIGR_REC_SECTOR_POSITION*sector_size,
71e5411e 12184 SEEK_SET) >= 0) {
466070ad 12185 if ((unsigned int)write(d->fd, super->migr_rec_buf,
de44e46f
PB		 12186 MIGR_REC_BUF_SECTORS*sector_size) !=
12187 MIGR_REC_BUF_SECTORS*sector_size)
71e5411e
PB		 12188 perror("Write migr_rec failed");
12189 }
12190 }
12191
10f22854
AK		 12192 /* return '1' if done */
12193 ret_val = 1;
895ffd99
MT		 12194
12195 /* After the reshape eliminate size mismatch in metadata.
12196 * Don't update md/component_size here, volume hasn't
12197 * to take whole space. It is allowed by kernel.
12198 * md/component_size will be set propoperly after next assembly.
12199 */
12200 imsm_fix_size_mismatch(st, subarray_index);
12201
10f22854
AK		 12202abort:
12203 free(buf);
942e1cdb
N		 12204 /* See Grow.c: abort_reshape() for further explanation */
12205 sysfs_set_num(sra, NULL, "suspend_lo", 0x7FFFFFFFFFFFFFFFULL);
12206 sysfs_set_num(sra, NULL, "suspend_hi", 0);
12207 sysfs_set_num(sra, NULL, "suspend_lo", 0);
10f22854
AK		 12208
12209 return ret_val;
999b4972 12210}
0c21b485 12211
cdddbdbc 12212struct superswitch super_imsm = {
cdddbdbc
DW		 12213 .examine_super = examine_super_imsm,
12214 .brief_examine_super = brief_examine_super_imsm,
4737ae25 12215 .brief_examine_subarrays = brief_examine_subarrays_imsm,
9d84c8ea 12216 .export_examine_super = export_examine_super_imsm,
cdddbdbc
DW		 12217 .detail_super = detail_super_imsm,
12218 .brief_detail_super = brief_detail_super_imsm,
bf5a934a 12219 .write_init_super = write_init_super_imsm,
0e600426
N		 12220 .validate_geometry = validate_geometry_imsm,
12221 .add_to_super = add_to_super_imsm,
1a64be56 12222 .remove_from_super = remove_from_super_imsm,
d665cc31 12223 .detail_platform = detail_platform_imsm,
e50cf220 12224 .export_detail_platform = export_detail_platform_imsm,
33414a01 12225 .kill_subarray = kill_subarray_imsm,
aa534678 12226 .update_subarray = update_subarray_imsm,
2b959fbf 12227 .load_container = load_container_imsm,
71204a50
N		 12228 .default_geometry = default_geometry_imsm,
12229 .get_disk_controller_domain = imsm_get_disk_controller_domain,
12230 .reshape_super = imsm_reshape_super,
12231 .manage_reshape = imsm_manage_reshape,
9e2d750d 12232 .recover_backup = recover_backup_imsm,
27156a57 12233 .examine_badblocks = examine_badblocks_imsm,
cdddbdbc
DW		 12234 .match_home = match_home_imsm,
12235 .uuid_from_super= uuid_from_super_imsm,
12236 .getinfo_super = getinfo_super_imsm,
5c4cd5da 12237 .getinfo_super_disks = getinfo_super_disks_imsm,
cdddbdbc
DW		 12238 .update_super = update_super_imsm,
12239
12240 .avail_size = avail_size_imsm,
fbfdcb06 12241 .get_spare_criteria = get_spare_criteria_imsm,
cdddbdbc
DW		 12242
12243 .compare_super = compare_super_imsm,
12244
12245 .load_super = load_super_imsm,
bf5a934a 12246 .init_super = init_super_imsm,
e683ca88 12247 .store_super = store_super_imsm,
cdddbdbc
DW		 12248 .free_super = free_super_imsm,
12249 .match_metadata_desc = match_metadata_desc_imsm,
bf5a934a 12250 .container_content = container_content_imsm,
0c21b485 12251 .validate_container = validate_container_imsm,
cdddbdbc 12252
2432ce9b
AP		 12253 .write_init_ppl = write_init_ppl_imsm,
12254 .validate_ppl = validate_ppl_imsm,
12255
cdddbdbc 12256 .external = 1,
4cce4069 12257 .name = "imsm",
845dea95
NB		 12258
12259/* for mdmon */
12260 .open_new = imsm_open_new,
ed9d66aa 12261 .set_array_state= imsm_set_array_state,
845dea95
NB		 12262 .set_disk = imsm_set_disk,
12263 .sync_metadata = imsm_sync_metadata,
88758e9d 12264 .activate_spare = imsm_activate_spare,
e8319a19 12265 .process_update = imsm_process_update,
8273f55e 12266 .prepare_update = imsm_prepare_update,
6f50473f 12267 .record_bad_block = imsm_record_badblock,
c07a5a4f 12268 .clear_bad_block = imsm_clear_badblock,
928f1424 12269 .get_bad_blocks = imsm_get_badblocks,
cdddbdbc 12270};
Unnamed repository; edit this file 'description' to name the repository.