]> 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
Remove Spare drives line from details 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
fbc42556
JR		 99/*
100 * Internal Write-intent bitmap is stored in the same area where PPL.
101 * Both features are mutually exclusive, so it is not an issue.
102 * The first 8KiB of the area are reserved and shall not be used.
103 */
104#define IMSM_BITMAP_AREA_RESERVED_SIZE 8192
105
106#define IMSM_BITMAP_HEADER_OFFSET (IMSM_BITMAP_AREA_RESERVED_SIZE)
107#define IMSM_BITMAP_HEADER_SIZE MAX_SECTOR_SIZE
108
109#define IMSM_BITMAP_START_OFFSET (IMSM_BITMAP_HEADER_OFFSET + IMSM_BITMAP_HEADER_SIZE)
110#define IMSM_BITMAP_AREA_SIZE (MULTIPLE_PPL_AREA_SIZE_IMSM - IMSM_BITMAP_START_OFFSET)
111#define IMSM_BITMAP_AND_HEADER_SIZE (IMSM_BITMAP_AREA_SIZE + IMSM_BITMAP_HEADER_SIZE)
112
113#define IMSM_DEFAULT_BITMAP_CHUNKSIZE (64 * 1024 * 1024)
114#define IMSM_DEFAULT_BITMAP_DAEMON_SLEEP 5
115
761e3bd9
N		 116/*
117 * This macro let's us ensure that no-one accidentally
118 * changes the size of a struct
119 */
120#define ASSERT_SIZE(_struct, size) \
121static inline void __assert_size_##_struct(void) \
122{ \
123 switch (0) { \
124 case 0: break; \
125 case (sizeof(struct _struct) == size): break; \
126 } \
127}
128
cdddbdbc
DW		 129/* Disk configuration info. */
130#define IMSM_MAX_DEVICES 255
131struct imsm_disk {
132 __u8 serial[MAX_RAID_SERIAL_LEN];/* 0xD8 - 0xE7 ascii serial number */
5551b113 133 __u32 total_blocks_lo; /* 0xE8 - 0xEB total blocks lo */
cdddbdbc 134 __u32 scsi_id; /* 0xEC - 0xEF scsi ID */
f2f27e63
DW		 135#define SPARE_DISK __cpu_to_le32(0x01) /* Spare */
136#define CONFIGURED_DISK __cpu_to_le32(0x02) /* Member of some RaidDev */
137#define FAILED_DISK __cpu_to_le32(0x04) /* Permanent failure */
2432ce9b 138#define JOURNAL_DISK __cpu_to_le32(0x2000000) /* Device marked as Journaling Drive */
cdddbdbc 139 __u32 status; /* 0xF0 - 0xF3 */
1011e834 140 __u32 owner_cfg_num; /* which config 0,1,2... owns this disk */
5551b113
CA		 141 __u32 total_blocks_hi; /* 0xF4 - 0xF5 total blocks hi */
142#define IMSM_DISK_FILLERS 3
143 __u32 filler[IMSM_DISK_FILLERS]; /* 0xF5 - 0x107 MPB_DISK_FILLERS for future expansion */
cdddbdbc 144};
761e3bd9 145ASSERT_SIZE(imsm_disk, 48)
cdddbdbc 146
3b451610
AK		 147/* map selector for map managment
148 */
238c0a71
AK		 149#define MAP_0 0
150#define MAP_1 1
151#define MAP_X -1
3b451610 152
cdddbdbc
DW		 153/* RAID map configuration infos. */
154struct imsm_map {
5551b113
CA		 155 __u32 pba_of_lba0_lo; /* start address of partition */
156 __u32 blocks_per_member_lo;/* blocks per member */
157 __u32 num_data_stripes_lo; /* number of data stripes */
cdddbdbc
DW		 158 __u16 blocks_per_strip;
159 __u8 map_state; /* Normal, Uninitialized, Degraded, Failed */
160#define IMSM_T_STATE_NORMAL 0
161#define IMSM_T_STATE_UNINITIALIZED 1
e3bba0e0
DW		 162#define IMSM_T_STATE_DEGRADED 2
163#define IMSM_T_STATE_FAILED 3
cdddbdbc
DW		 164 __u8 raid_level;
165#define IMSM_T_RAID0 0
166#define IMSM_T_RAID1 1
167#define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */
168 __u8 num_members; /* number of member disks */
fe7ed8cb
DW		 169 __u8 num_domains; /* number of parity domains */
170 __u8 failed_disk_num; /* valid only when state is degraded */
252d23c0 171 __u8 ddf;
5551b113
CA		 172 __u32 pba_of_lba0_hi;
173 __u32 blocks_per_member_hi;
174 __u32 num_data_stripes_hi;
175 __u32 filler[4]; /* expansion area */
7eef0453 176#define IMSM_ORD_REBUILD (1 << 24)
cdddbdbc 177 __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members],
7eef0453
DW		 178 * top byte contains some flags
179 */
761e3bd9
N		 180};
181ASSERT_SIZE(imsm_map, 52)
cdddbdbc
DW		 182
183struct imsm_vol {
4036e7ee 184 __u32 curr_migr_unit_lo;
fe7ed8cb 185 __u32 checkpoint_id; /* id to access curr_migr_unit */
cdddbdbc 186 __u8 migr_state; /* Normal or Migrating */
e3bba0e0
DW		 187#define MIGR_INIT 0
188#define MIGR_REBUILD 1
189#define MIGR_VERIFY 2 /* analagous to echo check > sync_action */
190#define MIGR_GEN_MIGR 3
191#define MIGR_STATE_CHANGE 4
1484e727 192#define MIGR_REPAIR 5
cdddbdbc 193 __u8 migr_type; /* Initializing, Rebuilding, ... */
2432ce9b
AP		 194#define RAIDVOL_CLEAN 0
195#define RAIDVOL_DIRTY 1
196#define RAIDVOL_DSRECORD_VALID 2
cdddbdbc 197 __u8 dirty;
fe7ed8cb
DW		 198 __u8 fs_state; /* fast-sync state for CnG (0xff == disabled) */
199 __u16 verify_errors; /* number of mismatches */
200 __u16 bad_blocks; /* number of bad blocks during verify */
4036e7ee
MT		 201 __u32 curr_migr_unit_hi;
202 __u32 filler[3];
cdddbdbc
DW		 203 struct imsm_map map[1];
204 /* here comes another one if migr_state */
761e3bd9
N		 205};
206ASSERT_SIZE(imsm_vol, 84)
cdddbdbc
DW		 207
208struct imsm_dev {
fe7ed8cb 209 __u8 volume[MAX_RAID_SERIAL_LEN];
cdddbdbc
DW		 210 __u32 size_low;
211 __u32 size_high;
fe7ed8cb
DW		 212#define DEV_BOOTABLE __cpu_to_le32(0x01)
213#define DEV_BOOT_DEVICE __cpu_to_le32(0x02)
214#define DEV_READ_COALESCING __cpu_to_le32(0x04)
215#define DEV_WRITE_COALESCING __cpu_to_le32(0x08)
216#define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10)
217#define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20)
218#define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40)
219#define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80)
220#define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100)
221#define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200)
222#define DEV_CLONE_N_GO __cpu_to_le32(0x400)
223#define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800)
224#define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000)
cdddbdbc
DW		 225 __u32 status; /* Persistent RaidDev status */
226 __u32 reserved_blocks; /* Reserved blocks at beginning of volume */
fe7ed8cb
DW		 227 __u8 migr_priority;
228 __u8 num_sub_vols;
229 __u8 tid;
230 __u8 cng_master_disk;
231 __u16 cache_policy;
232 __u8 cng_state;
233 __u8 cng_sub_state;
2432ce9b
AP		 234 __u16 my_vol_raid_dev_num; /* Used in Unique volume Id for this RaidDev */
235
236 /* NVM_EN */
237 __u8 nv_cache_mode;
238 __u8 nv_cache_flags;
239
240 /* Unique Volume Id of the NvCache Volume associated with this volume */
241 __u32 nvc_vol_orig_family_num;
242 __u16 nvc_vol_raid_dev_num;
243
244#define RWH_OFF 0
245#define RWH_DISTRIBUTED 1
246#define RWH_JOURNALING_DRIVE 2
c2462068
PB		 247#define RWH_MULTIPLE_DISTRIBUTED 3
248#define RWH_MULTIPLE_PPLS_JOURNALING_DRIVE 4
249#define RWH_MULTIPLE_OFF 5
fbc42556 250#define RWH_BITMAP 6
2432ce9b
AP		 251 __u8 rwh_policy; /* Raid Write Hole Policy */
252 __u8 jd_serial[MAX_RAID_SERIAL_LEN]; /* Journal Drive serial number */
253 __u8 filler1;
254
255#define IMSM_DEV_FILLERS 3
cdddbdbc
DW		 256 __u32 filler[IMSM_DEV_FILLERS];
257 struct imsm_vol vol;
761e3bd9
N		 258};
259ASSERT_SIZE(imsm_dev, 164)
cdddbdbc
DW		 260
261struct imsm_super {
262 __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */
263 __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */
264 __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */
265 __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */
266 __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */
604b746f
JD		 267 __u32 error_log_size; /* 0x30 - 0x33 in bytes */
268 __u32 attributes; /* 0x34 - 0x37 */
cdddbdbc
DW		 269 __u8 num_disks; /* 0x38 Number of configured disks */
270 __u8 num_raid_devs; /* 0x39 Number of configured volumes */
604b746f
JD		 271 __u8 error_log_pos; /* 0x3A */
272 __u8 fill[1]; /* 0x3B */
273 __u32 cache_size; /* 0x3c - 0x40 in mb */
274 __u32 orig_family_num; /* 0x40 - 0x43 original family num */
275 __u32 pwr_cycle_count; /* 0x44 - 0x47 simulated power cycle count for array */
276 __u32 bbm_log_size; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */
2a24dc1b
PB		 277 __u16 num_raid_devs_created; /* 0x4C - 0x4D Used for generating unique
278 * volume IDs for raid_dev created in this array
279 * (starts at 1)
280 */
281 __u16 filler1; /* 0x4E - 0x4F */
e48aed3c
AP		 282 __u64 creation_time; /* 0x50 - 0x57 Array creation time */
283#define IMSM_FILLERS 32
284 __u32 filler[IMSM_FILLERS]; /* 0x58 - 0xD7 RAID_MPB_FILLERS */
cdddbdbc
DW		 285 struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */
286 /* here comes imsm_dev[num_raid_devs] */
604b746f 287 /* here comes BBM logs */
761e3bd9
N		 288};
289ASSERT_SIZE(imsm_super, 264)
cdddbdbc 290
604b746f 291#define BBM_LOG_MAX_ENTRIES 254
8d67477f
TM		 292#define BBM_LOG_MAX_LBA_ENTRY_VAL 256 /* Represents 256 LBAs */
293#define BBM_LOG_SIGNATURE 0xabadb10c
294
295struct bbm_log_block_addr {
296 __u16 w1;
297 __u32 dw1;
298} __attribute__ ((__packed__));
604b746f
JD		 299
300struct bbm_log_entry {
8d67477f
TM		 301 __u8 marked_count; /* Number of blocks marked - 1 */
302 __u8 disk_ordinal; /* Disk entry within the imsm_super */
303 struct bbm_log_block_addr defective_block_start;
604b746f
JD		 304} __attribute__ ((__packed__));
305
306struct bbm_log {
307 __u32 signature; /* 0xABADB10C */
308 __u32 entry_count;
8d67477f 309 struct bbm_log_entry marked_block_entries[BBM_LOG_MAX_ENTRIES];
761e3bd9
N		 310};
311ASSERT_SIZE(bbm_log, 2040)
604b746f 312
cdddbdbc 313static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" };
cdddbdbc 314
b53bfba6
TM		 315#define BLOCKS_PER_KB (1024/512)
316
8e59f3d8
AK		 317#define RAID_DISK_RESERVED_BLOCKS_IMSM_HI 2209
318
319#define GEN_MIGR_AREA_SIZE 2048 /* General Migration Copy Area size in blocks */
320
de44e46f
PB		 321#define MIGR_REC_BUF_SECTORS 1 /* size of migr_record i/o buffer in sectors */
322#define MIGR_REC_SECTOR_POSITION 1 /* migr_record position offset on disk,
323 * MIGR_REC_BUF_SECTORS <= MIGR_REC_SECTOR_POS
17a4eaf9
AK		 324 */
325
8e59f3d8
AK		 326#define UNIT_SRC_NORMAL 0 /* Source data for curr_migr_unit must
327 * be recovered using srcMap */
328#define UNIT_SRC_IN_CP_AREA 1 /* Source data for curr_migr_unit has
329 * already been migrated and must
330 * be recovered from checkpoint area */
2432ce9b 331
c2462068 332#define PPL_ENTRY_SPACE (128 * 1024) /* Size of single PPL, without the header */
2432ce9b 333
8e59f3d8
AK		 334struct migr_record {
335 __u32 rec_status; /* Status used to determine how to restart
336 * migration in case it aborts
337 * in some fashion */
9f421827 338 __u32 curr_migr_unit_lo; /* 0..numMigrUnits-1 */
8e59f3d8
AK		 339 __u32 family_num; /* Family number of MPB
340 * containing the RaidDev
341 * that is migrating */
342 __u32 ascending_migr; /* True if migrating in increasing
343 * order of lbas */
344 __u32 blocks_per_unit; /* Num disk blocks per unit of operation */
345 __u32 dest_depth_per_unit; /* Num member blocks each destMap
346 * member disk
347 * advances per unit-of-operation */
9f421827
PB		 348 __u32 ckpt_area_pba_lo; /* Pba of first block of ckpt copy area */
349 __u32 dest_1st_member_lba_lo; /* First member lba on first
350 * stripe of destination */
351 __u32 num_migr_units_lo; /* Total num migration units-of-op */
8e59f3d8
AK		 352 __u32 post_migr_vol_cap; /* Size of volume after
353 * migration completes */
354 __u32 post_migr_vol_cap_hi; /* Expansion space for LBA64 */
355 __u32 ckpt_read_disk_num; /* Which member disk in destSubMap[0] the
356 * migration ckpt record was read from
357 * (for recovered migrations) */
9f421827
PB		 358 __u32 curr_migr_unit_hi; /* 0..numMigrUnits-1 high order 32 bits */
359 __u32 ckpt_area_pba_hi; /* Pba of first block of ckpt copy area
360 * high order 32 bits */
361 __u32 dest_1st_member_lba_hi; /* First member lba on first stripe of
362 * destination - high order 32 bits */
363 __u32 num_migr_units_hi; /* Total num migration units-of-op
364 * high order 32 bits */
4036e7ee 365 __u32 filler[16];
761e3bd9 366};
4036e7ee 367ASSERT_SIZE(migr_record, 128)
8e59f3d8 368
ec50f7b6
LM		 369struct md_list {
370 /* usage marker:
371 * 1: load metadata
372 * 2: metadata does not match
373 * 4: already checked
374 */
375 int used;
376 char *devname;
377 int found;
378 int container;
379 dev_t st_rdev;
380 struct md_list *next;
381};
382
e7b84f9d 383#define pr_vrb(fmt, arg...) (void) (verbose && pr_err(fmt, ##arg))
ec50f7b6 384
1484e727
DW		 385static __u8 migr_type(struct imsm_dev *dev)
386{
387 if (dev->vol.migr_type == MIGR_VERIFY &&
388 dev->status & DEV_VERIFY_AND_FIX)
389 return MIGR_REPAIR;
390 else
391 return dev->vol.migr_type;
392}
393
394static void set_migr_type(struct imsm_dev *dev, __u8 migr_type)
395{
396 /* for compatibility with older oroms convert MIGR_REPAIR, into
397 * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status
398 */
399 if (migr_type == MIGR_REPAIR) {
400 dev->vol.migr_type = MIGR_VERIFY;
401 dev->status |= DEV_VERIFY_AND_FIX;
402 } else {
403 dev->vol.migr_type = migr_type;
404 dev->status &= ~DEV_VERIFY_AND_FIX;
405 }
406}
407
f36a9ecd 408static unsigned int sector_count(__u32 bytes, unsigned int sector_size)
cdddbdbc 409{
f36a9ecd 410 return ROUND_UP(bytes, sector_size) / sector_size;
87eb16df 411}
cdddbdbc 412
f36a9ecd
PB		 413static unsigned int mpb_sectors(struct imsm_super *mpb,
414 unsigned int sector_size)
87eb16df 415{
f36a9ecd 416 return sector_count(__le32_to_cpu(mpb->mpb_size), sector_size);
cdddbdbc
DW		 417}
418
ba2de7ba
DW		 419struct intel_dev {
420 struct imsm_dev *dev;
421 struct intel_dev *next;
f21e18ca 422 unsigned index;
ba2de7ba
DW		 423};
424
88654014
LM		 425struct intel_hba {
426 enum sys_dev_type type;
427 char *path;
428 char *pci_id;
429 struct intel_hba *next;
430};
431
1a64be56
LM		 432enum action {
433 DISK_REMOVE = 1,
434 DISK_ADD
435};
cdddbdbc
DW		 436/* internal representation of IMSM metadata */
437struct intel_super {
438 union {
949c47a0
DW		 439 void *buf; /* O_DIRECT buffer for reading/writing metadata */
440 struct imsm_super *anchor; /* immovable parameters */
cdddbdbc 441 };
8e59f3d8
AK		 442 union {
443 void *migr_rec_buf; /* buffer for I/O operations */
444 struct migr_record *migr_rec; /* migration record */
445 };
51d83f5d
AK		 446 int clean_migration_record_by_mdmon; /* when reshape is switched to next
447 array, it indicates that mdmon is allowed to clean migration
448 record */
949c47a0 449 size_t len; /* size of the 'buf' allocation */
bbab0940 450 size_t extra_space; /* extra space in 'buf' that is not used yet */
4d7b1503
DW		 451 void *next_buf; /* for realloc'ing buf from the manager */
452 size_t next_len;
c2c087e6 453 int updates_pending; /* count of pending updates for mdmon */
bf5a934a 454 int current_vol; /* index of raid device undergoing creation */
5551b113 455 unsigned long long create_offset; /* common start for 'current_vol' */
148acb7b 456 __u32 random; /* random data for seeding new family numbers */
ba2de7ba 457 struct intel_dev *devlist;
fa7bb6f8 458 unsigned int sector_size; /* sector size of used member drives */
cdddbdbc
DW		 459 struct dl {
460 struct dl *next;
461 int index;
462 __u8 serial[MAX_RAID_SERIAL_LEN];
463 int major, minor;
464 char *devname;
b9f594fe 465 struct imsm_disk disk;
cdddbdbc 466 int fd;
0dcecb2e
DW		 467 int extent_cnt;
468 struct extent *e; /* for determining freespace @ create */
efb30e7f 469 int raiddisk; /* slot to fill in autolayout */
1a64be56 470 enum action action;
ca0748fa 471 } *disks, *current_disk;
1a64be56
LM		 472 struct dl *disk_mgmt_list; /* list of disks to add/remove while mdmon
473 active */
47ee5a45 474 struct dl *missing; /* disks removed while we weren't looking */
43dad3d6 475 struct bbm_log *bbm_log;
88654014 476 struct intel_hba *hba; /* device path of the raid controller for this metadata */
88c32bb1 477 const struct imsm_orom *orom; /* platform firmware support */
a2b97981 478 struct intel_super *next; /* (temp) list for disambiguating family_num */
928f1424 479 struct md_bb bb; /* memory for get_bad_blocks call */
a2b97981
DW		 480};
481
482struct intel_disk {
483 struct imsm_disk disk;
484 #define IMSM_UNKNOWN_OWNER (-1)
485 int owner;
486 struct intel_disk *next;
cdddbdbc
DW		 487};
488
c2c087e6
DW		 489struct extent {
490 unsigned long long start, size;
491};
492
694575e7
KW		 493/* definitions of reshape process types */
494enum imsm_reshape_type {
495 CH_TAKEOVER,
b5347799 496 CH_MIGRATION,
7abc9871 497 CH_ARRAY_SIZE,
694575e7
KW		 498};
499
88758e9d
DW		 500/* definition of messages passed to imsm_process_update */
501enum imsm_update_type {
502 update_activate_spare,
8273f55e 503 update_create_array,
33414a01 504 update_kill_array,
aa534678 505 update_rename_array,
1a64be56 506 update_add_remove_disk,
78b10e66 507 update_reshape_container_disks,
48c5303a 508 update_reshape_migration,
2d40f3a1
AK		 509 update_takeover,
510 update_general_migration_checkpoint,
f3871fdc 511 update_size_change,
bbab0940 512 update_prealloc_badblocks_mem,
e6e9dd3f 513 update_rwh_policy,
88758e9d
DW		 514};
515
516struct imsm_update_activate_spare {
517 enum imsm_update_type type;
d23fe947 518 struct dl *dl;
88758e9d
DW		 519 int slot;
520 int array;
521 struct imsm_update_activate_spare *next;
522};
523
78b10e66 524struct geo_params {
4dd2df09 525 char devnm[32];
78b10e66 526 char *dev_name;
d04f65f4 527 unsigned long long size;
78b10e66
N		 528 int level;
529 int layout;
530 int chunksize;
531 int raid_disks;
532};
533
bb025c2f
KW		 534enum takeover_direction {
535 R10_TO_R0,
536 R0_TO_R10
537};
538struct imsm_update_takeover {
539 enum imsm_update_type type;
540 int subarray;
541 enum takeover_direction direction;
542};
78b10e66
N		 543
544struct imsm_update_reshape {
545 enum imsm_update_type type;
546 int old_raid_disks;
547 int new_raid_disks;
48c5303a
PC		 548
549 int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */
550};
551
552struct imsm_update_reshape_migration {
553 enum imsm_update_type type;
554 int old_raid_disks;
555 int new_raid_disks;
556 /* fields for array migration changes
557 */
558 int subdev;
559 int new_level;
560 int new_layout;
4bba0439 561 int new_chunksize;
48c5303a 562
d195167d 563 int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */
78b10e66
N		 564};
565
f3871fdc
AK		 566struct imsm_update_size_change {
567 enum imsm_update_type type;
568 int subdev;
569 long long new_size;
570};
571
2d40f3a1
AK		 572struct imsm_update_general_migration_checkpoint {
573 enum imsm_update_type type;
4036e7ee 574 __u64 curr_migr_unit;
2d40f3a1
AK		 575};
576
54c2c1ea
DW		 577struct disk_info {
578 __u8 serial[MAX_RAID_SERIAL_LEN];
579};
580
8273f55e
DW		 581struct imsm_update_create_array {
582 enum imsm_update_type type;
8273f55e 583 int dev_idx;
6a3e913e 584 struct imsm_dev dev;
8273f55e
DW		 585};
586
33414a01
DW		 587struct imsm_update_kill_array {
588 enum imsm_update_type type;
589 int dev_idx;
590};
591
aa534678
DW		 592struct imsm_update_rename_array {
593 enum imsm_update_type type;
594 __u8 name[MAX_RAID_SERIAL_LEN];
595 int dev_idx;
596};
597
1a64be56 598struct imsm_update_add_remove_disk {
43dad3d6
DW		 599 enum imsm_update_type type;
600};
601
bbab0940
TM		 602struct imsm_update_prealloc_bb_mem {
603 enum imsm_update_type type;
604};
605
e6e9dd3f
AP		 606struct imsm_update_rwh_policy {
607 enum imsm_update_type type;
608 int new_policy;
609 int dev_idx;
610};
611
88654014
LM		 612static const char *_sys_dev_type[] = {
613 [SYS_DEV_UNKNOWN] = "Unknown",
614 [SYS_DEV_SAS] = "SAS",
614902f6 615 [SYS_DEV_SATA] = "SATA",
60f0f54d
PB		 616 [SYS_DEV_NVME] = "NVMe",
617 [SYS_DEV_VMD] = "VMD"
88654014
LM		 618};
619
620const char *get_sys_dev_type(enum sys_dev_type type)
621{
622 if (type >= SYS_DEV_MAX)
623 type = SYS_DEV_UNKNOWN;
624
625 return _sys_dev_type[type];
626}
627
628static struct intel_hba * alloc_intel_hba(struct sys_dev *device)
629{
503975b9
N		 630 struct intel_hba *result = xmalloc(sizeof(*result));
631
632 result->type = device->type;
633 result->path = xstrdup(device->path);
634 result->next = NULL;
635 if (result->path && (result->pci_id = strrchr(result->path, '/')) != NULL)
636 result->pci_id++;
637
88654014
LM		 638 return result;
639}
640
641static struct intel_hba * find_intel_hba(struct intel_hba *hba, struct sys_dev *device)
642{
594dc1b8
JS		 643 struct intel_hba *result;
644
88654014
LM		 645 for (result = hba; result; result = result->next) {
646 if (result->type == device->type && strcmp(result->path, device->path) == 0)
647 break;
648 }
649 return result;
650}
651
b4cf4cba 652static int attach_hba_to_super(struct intel_super *super, struct sys_dev *device)
88654014
LM		 653{
654 struct intel_hba *hba;
655
656 /* check if disk attached to Intel HBA */
657 hba = find_intel_hba(super->hba, device);
658 if (hba != NULL)
659 return 1;
660 /* Check if HBA is already attached to super */
661 if (super->hba == NULL) {
662 super->hba = alloc_intel_hba(device);
663 return 1;
6b781d33
AP		 664 }
665
666 hba = super->hba;
667 /* Intel metadata allows for all disks attached to the same type HBA.
614902f6 668 * Do not support HBA types mixing
6b781d33
AP		 669 */
670 if (device->type != hba->type)
88654014 671 return 2;
6b781d33
AP		 672
673 /* Multiple same type HBAs can be used if they share the same OROM */
674 const struct imsm_orom *device_orom = get_orom_by_device_id(device->dev_id);
675
676 if (device_orom != super->orom)
677 return 2;
678
679 while (hba->next)
680 hba = hba->next;
681
682 hba->next = alloc_intel_hba(device);
683 return 1;
88654014
LM		 684}
685
686static struct sys_dev* find_disk_attached_hba(int fd, const char *devname)
687{
9bc4ae77 688 struct sys_dev *list, *elem;
88654014
LM		 689 char *disk_path;
690
691 if ((list = find_intel_devices()) == NULL)
692 return 0;
693
694 if (fd < 0)
695 disk_path = (char *) devname;
696 else
7c798f87 697 disk_path = diskfd_to_devpath(fd, 1, NULL);
88654014 698
9bc4ae77 699 if (!disk_path)
88654014 700 return 0;
88654014 701
9bc4ae77
N		 702 for (elem = list; elem; elem = elem->next)
703 if (path_attached_to_hba(disk_path, elem->path))
88654014 704 return elem;
9bc4ae77 705
88654014
LM		 706 if (disk_path != devname)
707 free(disk_path);
88654014
LM		 708
709 return NULL;
710}
711
d424212e
N		 712static int find_intel_hba_capability(int fd, struct intel_super *super,
713 char *devname);
f2f5c343 714
cdddbdbc
DW		 715static struct supertype *match_metadata_desc_imsm(char *arg)
716{
717 struct supertype *st;
718
719 if (strcmp(arg, "imsm") != 0 &&
720 strcmp(arg, "default") != 0
721 )
722 return NULL;
723
503975b9 724 st = xcalloc(1, sizeof(*st));
cdddbdbc
DW		 725 st->ss = &super_imsm;
726 st->max_devs = IMSM_MAX_DEVICES;
727 st->minor_version = 0;
728 st->sb = NULL;
729 return st;
730}
731
cdddbdbc
DW		 732static __u8 *get_imsm_version(struct imsm_super *mpb)
733{
734 return &mpb->sig[MPB_SIG_LEN];
735}
736
949c47a0
DW		 737/* retrieve a disk directly from the anchor when the anchor is known to be
738 * up-to-date, currently only at load time
739 */
740static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index)
cdddbdbc 741{
949c47a0 742 if (index >= mpb->num_disks)
cdddbdbc
DW		 743 return NULL;
744 return &mpb->disk[index];
745}
746
95d07a2c
LM		 747/* retrieve the disk description based on a index of the disk
748 * in the sub-array
749 */
750static struct dl *get_imsm_dl_disk(struct intel_super *super, __u8 index)
949c47a0 751{
b9f594fe
DW		 752 struct dl *d;
753
754 for (d = super->disks; d; d = d->next)
755 if (d->index == index)
95d07a2c
LM		 756 return d;
757
758 return NULL;
759}
760/* retrieve a disk from the parsed metadata */
761static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index)
762{
763 struct dl *dl;
764
765 dl = get_imsm_dl_disk(super, index);
766 if (dl)
767 return &dl->disk;
768
b9f594fe 769 return NULL;
949c47a0
DW		 770}
771
772/* generate a checksum directly from the anchor when the anchor is known to be
773 * up-to-date, currently only at load or write_super after coalescing
774 */
775static __u32 __gen_imsm_checksum(struct imsm_super *mpb)
cdddbdbc
DW		 776{
777 __u32 end = mpb->mpb_size / sizeof(end);
778 __u32 *p = (__u32 *) mpb;
779 __u32 sum = 0;
780
5d500228
N		 781 while (end--) {
782 sum += __le32_to_cpu(*p);
97f734fd
N		 783 p++;
784 }
cdddbdbc 785
5d500228 786 return sum - __le32_to_cpu(mpb->check_sum);
cdddbdbc
DW		 787}
788
a965f303
DW		 789static size_t sizeof_imsm_map(struct imsm_map *map)
790{
791 return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1);
792}
793
794struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map)
cdddbdbc 795{
5e7b0330
AK		 796 /* A device can have 2 maps if it is in the middle of a migration.
797 * If second_map is:
238c0a71
AK		 798 * MAP_0 - we return the first map
799 * MAP_1 - we return the second map if it exists, else NULL
800 * MAP_X - we return the second map if it exists, else the first
5e7b0330 801 */
a965f303 802 struct imsm_map *map = &dev->vol.map[0];
9535fc47 803 struct imsm_map *map2 = NULL;
a965f303 804
9535fc47
AK		 805 if (dev->vol.migr_state)
806 map2 = (void *)map + sizeof_imsm_map(map);
a965f303 807
9535fc47 808 switch (second_map) {
3b451610 809 case MAP_0:
9535fc47 810 break;
3b451610 811 case MAP_1:
9535fc47
AK		 812 map = map2;
813 break;
238c0a71 814 case MAP_X:
9535fc47
AK		 815 if (map2)
816 map = map2;
817 break;
9535fc47
AK		 818 default:
819 map = NULL;
820 }
821 return map;
5e7b0330 822
a965f303 823}
cdddbdbc 824
3393c6af
DW		 825/* return the size of the device.
826 * migr_state increases the returned size if map[0] were to be duplicated
827 */
828static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state)
a965f303
DW		 829{
830 size_t size = sizeof(*dev) - sizeof(struct imsm_map) +
238c0a71 831 sizeof_imsm_map(get_imsm_map(dev, MAP_0));
cdddbdbc
DW		 832
833 /* migrating means an additional map */
a965f303 834 if (dev->vol.migr_state)
238c0a71 835 size += sizeof_imsm_map(get_imsm_map(dev, MAP_1));
3393c6af 836 else if (migr_state)
238c0a71 837 size += sizeof_imsm_map(get_imsm_map(dev, MAP_0));
cdddbdbc
DW		 838
839 return size;
840}
841
54c2c1ea
DW		 842/* retrieve disk serial number list from a metadata update */
843static struct disk_info *get_disk_info(struct imsm_update_create_array *update)
844{
845 void *u = update;
846 struct disk_info *inf;
847
848 inf = u + sizeof(*update) - sizeof(struct imsm_dev) +
849 sizeof_imsm_dev(&update->dev, 0);
850
851 return inf;
852}
54c2c1ea 853
949c47a0 854static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index)
cdddbdbc
DW		 855{
856 int offset;
857 int i;
858 void *_mpb = mpb;
859
949c47a0 860 if (index >= mpb->num_raid_devs)
cdddbdbc
DW		 861 return NULL;
862
863 /* devices start after all disks */
864 offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb;
865
866 for (i = 0; i <= index; i++)
867 if (i == index)
868 return _mpb + offset;
869 else
3393c6af 870 offset += sizeof_imsm_dev(_mpb + offset, 0);
cdddbdbc
DW		 871
872 return NULL;
873}
874
949c47a0
DW		 875static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index)
876{
ba2de7ba
DW		 877 struct intel_dev *dv;
878
949c47a0
DW		 879 if (index >= super->anchor->num_raid_devs)
880 return NULL;
ba2de7ba
DW		 881 for (dv = super->devlist; dv; dv = dv->next)
882 if (dv->index == index)
883 return dv->dev;
884 return NULL;
949c47a0
DW		 885}
886
8d67477f
TM		 887static inline unsigned long long __le48_to_cpu(const struct bbm_log_block_addr
888 *addr)
889{
890 return ((((__u64)__le32_to_cpu(addr->dw1)) << 16) |
891 __le16_to_cpu(addr->w1));
892}
893
894static inline struct bbm_log_block_addr __cpu_to_le48(unsigned long long sec)
895{
896 struct bbm_log_block_addr addr;
897
898 addr.w1 = __cpu_to_le16((__u16)(sec & 0xffff));
899 addr.dw1 = __cpu_to_le32((__u32)(sec >> 16) & 0xffffffff);
900 return addr;
901}
902
8d67477f
TM		 903/* get size of the bbm log */
904static __u32 get_imsm_bbm_log_size(struct bbm_log *log)
905{
906 if (!log || log->entry_count == 0)
907 return 0;
908
909 return sizeof(log->signature) +
910 sizeof(log->entry_count) +
911 log->entry_count * sizeof(struct bbm_log_entry);
912}
6f50473f
TM		 913
914/* check if bad block is not partially stored in bbm log */
915static int is_stored_in_bbm(struct bbm_log *log, const __u8 idx, const unsigned
916 long long sector, const int length, __u32 *pos)
917{
918 __u32 i;
919
920 for (i = *pos; i < log->entry_count; i++) {
921 struct bbm_log_entry *entry = &log->marked_block_entries[i];
922 unsigned long long bb_start;
923 unsigned long long bb_end;
924
925 bb_start = __le48_to_cpu(&entry->defective_block_start);
926 bb_end = bb_start + (entry->marked_count + 1);
927
928 if ((entry->disk_ordinal == idx) && (bb_start >= sector) &&
929 (bb_end <= sector + length)) {
930 *pos = i;
931 return 1;
932 }
933 }
934 return 0;
935}
936
937/* record new bad block in bbm log */
938static int record_new_badblock(struct bbm_log *log, const __u8 idx, unsigned
939 long long sector, int length)
940{
941 int new_bb = 0;
942 __u32 pos = 0;
943 struct bbm_log_entry *entry = NULL;
944
945 while (is_stored_in_bbm(log, idx, sector, length, &pos)) {
946 struct bbm_log_entry *e = &log->marked_block_entries[pos];
947
948 if ((e->marked_count + 1 == BBM_LOG_MAX_LBA_ENTRY_VAL) &&
949 (__le48_to_cpu(&e->defective_block_start) == sector)) {
950 sector += BBM_LOG_MAX_LBA_ENTRY_VAL;
951 length -= BBM_LOG_MAX_LBA_ENTRY_VAL;
952 pos = pos + 1;
953 continue;
954 }
955 entry = e;
956 break;
957 }
958
959 if (entry) {
960 int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length :
961 BBM_LOG_MAX_LBA_ENTRY_VAL;
962 entry->defective_block_start = __cpu_to_le48(sector);
963 entry->marked_count = cnt - 1;
964 if (cnt == length)
965 return 1;
966 sector += cnt;
967 length -= cnt;
968 }
969
970 new_bb = ROUND_UP(length, BBM_LOG_MAX_LBA_ENTRY_VAL) /
971 BBM_LOG_MAX_LBA_ENTRY_VAL;
972 if (log->entry_count + new_bb > BBM_LOG_MAX_ENTRIES)
973 return 0;
974
975 while (length > 0) {
976 int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length :
977 BBM_LOG_MAX_LBA_ENTRY_VAL;
978 struct bbm_log_entry *entry =
979 &log->marked_block_entries[log->entry_count];
980
981 entry->defective_block_start = __cpu_to_le48(sector);
982 entry->marked_count = cnt - 1;
983 entry->disk_ordinal = idx;
984
985 sector += cnt;
986 length -= cnt;
987
988 log->entry_count++;
989 }
990
991 return new_bb;
992}
c07a5a4f 993
4c9e8c1e
TM		 994/* clear all bad blocks for given disk */
995static void clear_disk_badblocks(struct bbm_log *log, const __u8 idx)
996{
997 __u32 i = 0;
998
999 while (i < log->entry_count) {
1000 struct bbm_log_entry *entries = log->marked_block_entries;
1001
1002 if (entries[i].disk_ordinal == idx) {
1003 if (i < log->entry_count - 1)
1004 entries[i] = entries[log->entry_count - 1];
1005 log->entry_count--;
1006 } else {
1007 i++;
1008 }
1009 }
1010}
1011
c07a5a4f
TM		 1012/* clear given bad block */
1013static int clear_badblock(struct bbm_log *log, const __u8 idx, const unsigned
1014 long long sector, const int length) {
1015 __u32 i = 0;
1016
1017 while (i < log->entry_count) {
1018 struct bbm_log_entry *entries = log->marked_block_entries;
1019
1020 if ((entries[i].disk_ordinal == idx) &&
1021 (__le48_to_cpu(&entries[i].defective_block_start) ==
1022 sector) && (entries[i].marked_count + 1 == length)) {
1023 if (i < log->entry_count - 1)
1024 entries[i] = entries[log->entry_count - 1];
1025 log->entry_count--;
1026 break;
1027 }
1028 i++;
1029 }
1030
1031 return 1;
1032}
8d67477f
TM		 1033
1034/* allocate and load BBM log from metadata */
1035static int load_bbm_log(struct intel_super *super)
1036{
1037 struct imsm_super *mpb = super->anchor;
1038 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
1039
1040 super->bbm_log = xcalloc(1, sizeof(struct bbm_log));
1041 if (!super->bbm_log)
1042 return 1;
1043
1044 if (bbm_log_size) {
1045 struct bbm_log *log = (void *)mpb +
1046 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
1047
1048 __u32 entry_count;
1049
1050 if (bbm_log_size < sizeof(log->signature) +
1051 sizeof(log->entry_count))
1052 return 2;
1053
1054 entry_count = __le32_to_cpu(log->entry_count);
1055 if ((__le32_to_cpu(log->signature) != BBM_LOG_SIGNATURE) ||
1056 (entry_count > BBM_LOG_MAX_ENTRIES))
1057 return 3;
1058
1059 if (bbm_log_size !=
1060 sizeof(log->signature) + sizeof(log->entry_count) +
1061 entry_count * sizeof(struct bbm_log_entry))
1062 return 4;
1063
1064 memcpy(super->bbm_log, log, bbm_log_size);
1065 } else {
1066 super->bbm_log->signature = __cpu_to_le32(BBM_LOG_SIGNATURE);
1067 super->bbm_log->entry_count = 0;
1068 }
1069
1070 return 0;
1071}
1072
b12796be
TM		 1073/* checks if bad block is within volume boundaries */
1074static int is_bad_block_in_volume(const struct bbm_log_entry *entry,
1075 const unsigned long long start_sector,
1076 const unsigned long long size)
1077{
1078 unsigned long long bb_start;
1079 unsigned long long bb_end;
1080
1081 bb_start = __le48_to_cpu(&entry->defective_block_start);
1082 bb_end = bb_start + (entry->marked_count + 1);
1083
1084 if (((bb_start >= start_sector) && (bb_start < start_sector + size)) ||
1085 ((bb_end >= start_sector) && (bb_end <= start_sector + size)))
1086 return 1;
1087
1088 return 0;
1089}
1090
1091/* get list of bad blocks on a drive for a volume */
1092static void get_volume_badblocks(const struct bbm_log *log, const __u8 idx,
1093 const unsigned long long start_sector,
1094 const unsigned long long size,
1095 struct md_bb *bbs)
1096{
1097 __u32 count = 0;
1098 __u32 i;
1099
1100 for (i = 0; i < log->entry_count; i++) {
1101 const struct bbm_log_entry *ent =
1102 &log->marked_block_entries[i];
1103 struct md_bb_entry *bb;
1104
1105 if ((ent->disk_ordinal == idx) &&
1106 is_bad_block_in_volume(ent, start_sector, size)) {
1107
1108 if (!bbs->entries) {
1109 bbs->entries = xmalloc(BBM_LOG_MAX_ENTRIES *
1110 sizeof(*bb));
1111 if (!bbs->entries)
1112 break;
1113 }
1114
1115 bb = &bbs->entries[count++];
1116 bb->sector = __le48_to_cpu(&ent->defective_block_start);
1117 bb->length = ent->marked_count + 1;
1118 }
1119 }
1120 bbs->count = count;
1121}
1122
98130f40
AK		 1123/*
1124 * for second_map:
238c0a71
AK		 1125 * == MAP_0 get first map
1126 * == MAP_1 get second map
1127 * == MAP_X than get map according to the current migr_state
98130f40
AK		 1128 */
1129static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev,
1130 int slot,
1131 int second_map)
7eef0453
DW		 1132{
1133 struct imsm_map *map;
1134
5e7b0330 1135 map = get_imsm_map(dev, second_map);
7eef0453 1136
ff077194
DW		 1137 /* top byte identifies disk under rebuild */
1138 return __le32_to_cpu(map->disk_ord_tbl[slot]);
1139}
1140
1141#define ord_to_idx(ord) (((ord) << 8) >> 8)
98130f40 1142static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot, int second_map)
ff077194 1143{
98130f40 1144 __u32 ord = get_imsm_ord_tbl_ent(dev, slot, second_map);
ff077194
DW		 1145
1146 return ord_to_idx(ord);
7eef0453
DW		 1147}
1148
be73972f
DW		 1149static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord)
1150{
1151 map->disk_ord_tbl[slot] = __cpu_to_le32(ord);
1152}
1153
f21e18ca 1154static int get_imsm_disk_slot(struct imsm_map *map, unsigned idx)
620b1713
DW		 1155{
1156 int slot;
1157 __u32 ord;
1158
1159 for (slot = 0; slot < map->num_members; slot++) {
1160 ord = __le32_to_cpu(map->disk_ord_tbl[slot]);
1161 if (ord_to_idx(ord) == idx)
1162 return slot;
1163 }
1164
1165 return -1;
1166}
1167
cdddbdbc
DW		 1168static int get_imsm_raid_level(struct imsm_map *map)
1169{
1170 if (map->raid_level == 1) {
1171 if (map->num_members == 2)
1172 return 1;
1173 else
1174 return 10;
1175 }
1176
1177 return map->raid_level;
1178}
1179
c2c087e6
DW		 1180static int cmp_extent(const void *av, const void *bv)
1181{
1182 const struct extent *a = av;
1183 const struct extent *b = bv;
1184 if (a->start < b->start)
1185 return -1;
1186 if (a->start > b->start)
1187 return 1;
1188 return 0;
1189}
1190
0dcecb2e 1191static int count_memberships(struct dl *dl, struct intel_super *super)
c2c087e6 1192{
c2c087e6 1193 int memberships = 0;
620b1713 1194 int i;
c2c087e6 1195
949c47a0
DW		 1196 for (i = 0; i < super->anchor->num_raid_devs; i++) {
1197 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 1198 struct imsm_map *map = get_imsm_map(dev, MAP_0);
c2c087e6 1199
620b1713
DW		 1200 if (get_imsm_disk_slot(map, dl->index) >= 0)
1201 memberships++;
c2c087e6 1202 }
0dcecb2e
DW		 1203
1204 return memberships;
1205}
1206
b81221b7
CA		 1207static __u32 imsm_min_reserved_sectors(struct intel_super *super);
1208
486720e0 1209static int split_ull(unsigned long long n, void *lo, void *hi)
5551b113
CA		 1210{
1211 if (lo == 0 || hi == 0)
1212 return 1;
486720e0
JS		 1213 __put_unaligned32(__cpu_to_le32((__u32)n), lo);
1214 __put_unaligned32(__cpu_to_le32((n >> 32)), hi);
5551b113
CA		 1215 return 0;
1216}
1217
1218static unsigned long long join_u32(__u32 lo, __u32 hi)
1219{
1220 return (unsigned long long)__le32_to_cpu(lo) |
1221 (((unsigned long long)__le32_to_cpu(hi)) << 32);
1222}
1223
1224static unsigned long long total_blocks(struct imsm_disk *disk)
1225{
1226 if (disk == NULL)
1227 return 0;
1228 return join_u32(disk->total_blocks_lo, disk->total_blocks_hi);
1229}
1230
1231static unsigned long long pba_of_lba0(struct imsm_map *map)
1232{
1233 if (map == NULL)
1234 return 0;
1235 return join_u32(map->pba_of_lba0_lo, map->pba_of_lba0_hi);
1236}
1237
1238static unsigned long long blocks_per_member(struct imsm_map *map)
1239{
1240 if (map == NULL)
1241 return 0;
1242 return join_u32(map->blocks_per_member_lo, map->blocks_per_member_hi);
1243}
1244
1245static unsigned long long num_data_stripes(struct imsm_map *map)
1246{
1247 if (map == NULL)
1248 return 0;
1249 return join_u32(map->num_data_stripes_lo, map->num_data_stripes_hi);
1250}
1251
4036e7ee
MT		 1252static unsigned long long vol_curr_migr_unit(struct imsm_dev *dev)
1253{
1254 if (dev == NULL)
1255 return 0;
1256
1257 return join_u32(dev->vol.curr_migr_unit_lo, dev->vol.curr_migr_unit_hi);
1258}
1259
fcc2c9da
MD		 1260static unsigned long long imsm_dev_size(struct imsm_dev *dev)
1261{
1262 if (dev == NULL)
1263 return 0;
1264 return join_u32(dev->size_low, dev->size_high);
1265}
1266
9f421827
PB		 1267static unsigned long long migr_chkp_area_pba(struct migr_record *migr_rec)
1268{
1269 if (migr_rec == NULL)
1270 return 0;
1271 return join_u32(migr_rec->ckpt_area_pba_lo,
1272 migr_rec->ckpt_area_pba_hi);
1273}
1274
1275static unsigned long long current_migr_unit(struct migr_record *migr_rec)
1276{
1277 if (migr_rec == NULL)
1278 return 0;
1279 return join_u32(migr_rec->curr_migr_unit_lo,
1280 migr_rec->curr_migr_unit_hi);
1281}
1282
1283static unsigned long long migr_dest_1st_member_lba(struct migr_record *migr_rec)
1284{
1285 if (migr_rec == NULL)
1286 return 0;
1287 return join_u32(migr_rec->dest_1st_member_lba_lo,
1288 migr_rec->dest_1st_member_lba_hi);
1289}
1290
1291static unsigned long long get_num_migr_units(struct migr_record *migr_rec)
1292{
1293 if (migr_rec == NULL)
1294 return 0;
1295 return join_u32(migr_rec->num_migr_units_lo,
1296 migr_rec->num_migr_units_hi);
1297}
1298
5551b113
CA		 1299static void set_total_blocks(struct imsm_disk *disk, unsigned long long n)
1300{
1301 split_ull(n, &disk->total_blocks_lo, &disk->total_blocks_hi);
1302}
1303
1304static void set_pba_of_lba0(struct imsm_map *map, unsigned long long n)
1305{
1306 split_ull(n, &map->pba_of_lba0_lo, &map->pba_of_lba0_hi);
1307}
1308
1309static void set_blocks_per_member(struct imsm_map *map, unsigned long long n)
1310{
1311 split_ull(n, &map->blocks_per_member_lo, &map->blocks_per_member_hi);
1312}
1313
1314static void set_num_data_stripes(struct imsm_map *map, unsigned long long n)
1315{
1316 split_ull(n, &map->num_data_stripes_lo, &map->num_data_stripes_hi);
1317}
1318
4036e7ee
MT		 1319static void set_vol_curr_migr_unit(struct imsm_dev *dev, unsigned long long n)
1320{
1321 if (dev == NULL)
1322 return;
1323
1324 split_ull(n, &dev->vol.curr_migr_unit_lo, &dev->vol.curr_migr_unit_hi);
1325}
1326
fcc2c9da
MD		 1327static void set_imsm_dev_size(struct imsm_dev *dev, unsigned long long n)
1328{
1329 split_ull(n, &dev->size_low, &dev->size_high);
1330}
1331
9f421827
PB		 1332static void set_migr_chkp_area_pba(struct migr_record *migr_rec,
1333 unsigned long long n)
1334{
1335 split_ull(n, &migr_rec->ckpt_area_pba_lo, &migr_rec->ckpt_area_pba_hi);
1336}
1337
1338static void set_current_migr_unit(struct migr_record *migr_rec,
1339 unsigned long long n)
1340{
1341 split_ull(n, &migr_rec->curr_migr_unit_lo,
1342 &migr_rec->curr_migr_unit_hi);
1343}
1344
1345static void set_migr_dest_1st_member_lba(struct migr_record *migr_rec,
1346 unsigned long long n)
1347{
1348 split_ull(n, &migr_rec->dest_1st_member_lba_lo,
1349 &migr_rec->dest_1st_member_lba_hi);
1350}
1351
1352static void set_num_migr_units(struct migr_record *migr_rec,
1353 unsigned long long n)
1354{
1355 split_ull(n, &migr_rec->num_migr_units_lo,
1356 &migr_rec->num_migr_units_hi);
1357}
1358
44490938
MD		 1359static unsigned long long per_dev_array_size(struct imsm_map *map)
1360{
1361 unsigned long long array_size = 0;
1362
1363 if (map == NULL)
1364 return array_size;
1365
1366 array_size = num_data_stripes(map) * map->blocks_per_strip;
1367 if (get_imsm_raid_level(map) == 1 || get_imsm_raid_level(map) == 10)
1368 array_size *= 2;
1369
1370 return array_size;
1371}
1372
05501181
PB		 1373static struct extent *get_extents(struct intel_super *super, struct dl *dl,
1374 int get_minimal_reservation)
0dcecb2e
DW		 1375{
1376 /* find a list of used extents on the given physical device */
1377 struct extent *rv, *e;
620b1713 1378 int i;
0dcecb2e 1379 int memberships = count_memberships(dl, super);
b276dd33
DW		 1380 __u32 reservation;
1381
1382 /* trim the reserved area for spares, so they can join any array
1383 * regardless of whether the OROM has assigned sectors from the
1384 * IMSM_RESERVED_SECTORS region
1385 */
05501181 1386 if (dl->index == -1 || get_minimal_reservation)
b81221b7 1387 reservation = imsm_min_reserved_sectors(super);
b276dd33
DW		 1388 else
1389 reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
0dcecb2e 1390
503975b9 1391 rv = xcalloc(sizeof(struct extent), (memberships + 1));
c2c087e6
DW		 1392 e = rv;
1393
949c47a0
DW		 1394 for (i = 0; i < super->anchor->num_raid_devs; i++) {
1395 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 1396 struct imsm_map *map = get_imsm_map(dev, MAP_0);
c2c087e6 1397
620b1713 1398 if (get_imsm_disk_slot(map, dl->index) >= 0) {
5551b113 1399 e->start = pba_of_lba0(map);
44490938 1400 e->size = per_dev_array_size(map);
620b1713 1401 e++;
c2c087e6
DW		 1402 }
1403 }
1404 qsort(rv, memberships, sizeof(*rv), cmp_extent);
1405
1011e834 1406 /* determine the start of the metadata
14e8215b
DW		 1407 * when no raid devices are defined use the default
1408 * ...otherwise allow the metadata to truncate the value
1409 * as is the case with older versions of imsm
1410 */
1411 if (memberships) {
1412 struct extent *last = &rv[memberships - 1];
5551b113 1413 unsigned long long remainder;
14e8215b 1414
5551b113 1415 remainder = total_blocks(&dl->disk) - (last->start + last->size);
dda5855f
DW		 1416 /* round down to 1k block to satisfy precision of the kernel
1417 * 'size' interface
1418 */
1419 remainder &= ~1UL;
1420 /* make sure remainder is still sane */
f21e18ca 1421 if (remainder < (unsigned)ROUND_UP(super->len, 512) >> 9)
dda5855f 1422 remainder = ROUND_UP(super->len, 512) >> 9;
14e8215b
DW		 1423 if (reservation > remainder)
1424 reservation = remainder;
1425 }
5551b113 1426 e->start = total_blocks(&dl->disk) - reservation;
c2c087e6
DW		 1427 e->size = 0;
1428 return rv;
1429}
1430
14e8215b
DW		 1431/* try to determine how much space is reserved for metadata from
1432 * the last get_extents() entry, otherwise fallback to the
1433 * default
1434 */
1435static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl)
1436{
1437 struct extent *e;
1438 int i;
1439 __u32 rv;
1440
1441 /* for spares just return a minimal reservation which will grow
1442 * once the spare is picked up by an array
1443 */
1444 if (dl->index == -1)
1445 return MPB_SECTOR_CNT;
1446
05501181 1447 e = get_extents(super, dl, 0);
14e8215b
DW		 1448 if (!e)
1449 return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
1450
1451 /* scroll to last entry */
1452 for (i = 0; e[i].size; i++)
1453 continue;
1454
5551b113 1455 rv = total_blocks(&dl->disk) - e[i].start;
14e8215b
DW		 1456
1457 free(e);
1458
1459 return rv;
1460}
1461
25ed7e59
DW		 1462static int is_spare(struct imsm_disk *disk)
1463{
1464 return (disk->status & SPARE_DISK) == SPARE_DISK;
1465}
1466
1467static int is_configured(struct imsm_disk *disk)
1468{
1469 return (disk->status & CONFIGURED_DISK) == CONFIGURED_DISK;
1470}
1471
1472static int is_failed(struct imsm_disk *disk)
1473{
1474 return (disk->status & FAILED_DISK) == FAILED_DISK;
1475}
1476
2432ce9b
AP		 1477static int is_journal(struct imsm_disk *disk)
1478{
1479 return (disk->status & JOURNAL_DISK) == JOURNAL_DISK;
1480}
1481
b53bfba6
TM		 1482/* round array size down to closest MB and ensure it splits evenly
1483 * between members
1484 */
1485static unsigned long long round_size_to_mb(unsigned long long size, unsigned int
1486 disk_count)
1487{
1488 size /= disk_count;
1489 size = (size >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT;
1490 size *= disk_count;
1491
1492 return size;
1493}
1494
8b9cd157
MK		 1495static int able_to_resync(int raid_level, int missing_disks)
1496{
1497 int max_missing_disks = 0;
1498
1499 switch (raid_level) {
1500 case 10:
1501 max_missing_disks = 1;
1502 break;
1503 default:
1504 max_missing_disks = 0;
1505 }
1506 return missing_disks <= max_missing_disks;
1507}
1508
b81221b7
CA		 1509/* try to determine how much space is reserved for metadata from
1510 * the last get_extents() entry on the smallest active disk,
1511 * otherwise fallback to the default
1512 */
1513static __u32 imsm_min_reserved_sectors(struct intel_super *super)
1514{
1515 struct extent *e;
1516 int i;
5551b113
CA		 1517 unsigned long long min_active;
1518 __u32 remainder;
b81221b7
CA		 1519 __u32 rv = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
1520 struct dl *dl, *dl_min = NULL;
1521
1522 if (!super)
1523 return rv;
1524
1525 min_active = 0;
1526 for (dl = super->disks; dl; dl = dl->next) {
1527 if (dl->index < 0)
1528 continue;
5551b113
CA		 1529 unsigned long long blocks = total_blocks(&dl->disk);
1530 if (blocks < min_active || min_active == 0) {
b81221b7 1531 dl_min = dl;
5551b113 1532 min_active = blocks;
b81221b7
CA		 1533 }
1534 }
1535 if (!dl_min)
1536 return rv;
1537
1538 /* find last lba used by subarrays on the smallest active disk */
05501181 1539 e = get_extents(super, dl_min, 0);
b81221b7
CA		 1540 if (!e)
1541 return rv;
1542 for (i = 0; e[i].size; i++)
1543 continue;
1544
1545 remainder = min_active - e[i].start;
1546 free(e);
1547
1548 /* to give priority to recovery we should not require full
1549 IMSM_RESERVED_SECTORS from the spare */
1550 rv = MPB_SECTOR_CNT + NUM_BLOCKS_DIRTY_STRIPE_REGION;
1551
1552 /* if real reservation is smaller use that value */
1553 return (remainder < rv) ? remainder : rv;
1554}
1555
fbfdcb06
AO		 1556/*
1557 * Return minimum size of a spare and sector size
1558 * that can be used in this array
1559 */
1560int get_spare_criteria_imsm(struct supertype *st, struct spare_criteria *c)
80e7f8c3
AC		 1561{
1562 struct intel_super *super = st->sb;
1563 struct dl *dl;
1564 struct extent *e;
1565 int i;
fbfdcb06
AO		 1566 unsigned long long size = 0;
1567
1568 c->min_size = 0;
4b57ecf6 1569 c->sector_size = 0;
80e7f8c3
AC		 1570
1571 if (!super)
fbfdcb06 1572 return -EINVAL;
80e7f8c3
AC		 1573 /* find first active disk in array */
1574 dl = super->disks;
1575 while (dl && (is_failed(&dl->disk) || dl->index == -1))
1576 dl = dl->next;
1577 if (!dl)
fbfdcb06 1578 return -EINVAL;
80e7f8c3 1579 /* find last lba used by subarrays */
05501181 1580 e = get_extents(super, dl, 0);
80e7f8c3 1581 if (!e)
fbfdcb06 1582 return -EINVAL;
80e7f8c3
AC		 1583 for (i = 0; e[i].size; i++)
1584 continue;
1585 if (i > 0)
fbfdcb06 1586 size = e[i-1].start + e[i-1].size;
80e7f8c3 1587 free(e);
b81221b7 1588
80e7f8c3 1589 /* add the amount of space needed for metadata */
fbfdcb06
AO		 1590 size += imsm_min_reserved_sectors(super);
1591
1592 c->min_size = size * 512;
4b57ecf6 1593 c->sector_size = super->sector_size;
b81221b7 1594
fbfdcb06 1595 return 0;
80e7f8c3
AC		 1596}
1597
d1e02575
AK		 1598static int is_gen_migration(struct imsm_dev *dev);
1599
f36a9ecd
PB		 1600#define IMSM_4K_DIV 8
1601
c47b0ff6
AK		 1602static __u64 blocks_per_migr_unit(struct intel_super *super,
1603 struct imsm_dev *dev);
1e5c6983 1604
c47b0ff6
AK		 1605static void print_imsm_dev(struct intel_super *super,
1606 struct imsm_dev *dev,
1607 char *uuid,
1608 int disk_idx)
cdddbdbc
DW		 1609{
1610 __u64 sz;
0d80bb2f 1611 int slot, i;
238c0a71
AK		 1612 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1613 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
b10b37b8 1614 __u32 ord;
cdddbdbc
DW		 1615
1616 printf("\n");
1e7bc0ed 1617 printf("[%.16s]:\n", dev->volume);
ba1b3bc8 1618 printf(" Subarray : %d\n", super->current_vol);
44470971 1619 printf(" UUID : %s\n", uuid);
dd8bcb3b
AK		 1620 printf(" RAID Level : %d", get_imsm_raid_level(map));
1621 if (map2)
1622 printf(" <-- %d", get_imsm_raid_level(map2));
1623 printf("\n");
1624 printf(" Members : %d", map->num_members);
1625 if (map2)
1626 printf(" <-- %d", map2->num_members);
1627 printf("\n");
0d80bb2f
DW		 1628 printf(" Slots : [");
1629 for (i = 0; i < map->num_members; i++) {
238c0a71 1630 ord = get_imsm_ord_tbl_ent(dev, i, MAP_0);
0d80bb2f
DW		 1631 printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U");
1632 }
dd8bcb3b
AK		 1633 printf("]");
1634 if (map2) {
1635 printf(" <-- [");
1636 for (i = 0; i < map2->num_members; i++) {
238c0a71 1637 ord = get_imsm_ord_tbl_ent(dev, i, MAP_1);
dd8bcb3b
AK		 1638 printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U");
1639 }
1640 printf("]");
1641 }
1642 printf("\n");
7095bccb
AK		 1643 printf(" Failed disk : ");
1644 if (map->failed_disk_num == 0xff)
1645 printf("none");
1646 else
1647 printf("%i", map->failed_disk_num);
1648 printf("\n");
620b1713
DW		 1649 slot = get_imsm_disk_slot(map, disk_idx);
1650 if (slot >= 0) {
238c0a71 1651 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
b10b37b8
DW		 1652 printf(" This Slot : %d%s\n", slot,
1653 ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : "");
1654 } else
cdddbdbc 1655 printf(" This Slot : ?\n");
84918897 1656 printf(" Sector Size : %u\n", super->sector_size);
fcc2c9da 1657 sz = imsm_dev_size(dev);
84918897
MK		 1658 printf(" Array Size : %llu%s\n",
1659 (unsigned long long)sz * 512 / super->sector_size,
cdddbdbc 1660 human_size(sz * 512));
5551b113 1661 sz = blocks_per_member(map);
84918897
MK		 1662 printf(" Per Dev Size : %llu%s\n",
1663 (unsigned long long)sz * 512 / super->sector_size,
cdddbdbc 1664 human_size(sz * 512));
5551b113 1665 printf(" Sector Offset : %llu\n",
7d8935cb 1666 pba_of_lba0(map) * 512 / super->sector_size);
5551b113
CA		 1667 printf(" Num Stripes : %llu\n",
1668 num_data_stripes(map));
dd8bcb3b 1669 printf(" Chunk Size : %u KiB",
cdddbdbc 1670 __le16_to_cpu(map->blocks_per_strip) / 2);
dd8bcb3b
AK		 1671 if (map2)
1672 printf(" <-- %u KiB",
1673 __le16_to_cpu(map2->blocks_per_strip) / 2);
1674 printf("\n");
cdddbdbc 1675 printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks));
8655a7b1 1676 printf(" Migrate State : ");
1484e727
DW		 1677 if (dev->vol.migr_state) {
1678 if (migr_type(dev) == MIGR_INIT)
8655a7b1 1679 printf("initialize\n");
1484e727 1680 else if (migr_type(dev) == MIGR_REBUILD)
8655a7b1 1681 printf("rebuild\n");
1484e727 1682 else if (migr_type(dev) == MIGR_VERIFY)
8655a7b1 1683 printf("check\n");
1484e727 1684 else if (migr_type(dev) == MIGR_GEN_MIGR)
8655a7b1 1685 printf("general migration\n");
1484e727 1686 else if (migr_type(dev) == MIGR_STATE_CHANGE)
8655a7b1 1687 printf("state change\n");
1484e727 1688 else if (migr_type(dev) == MIGR_REPAIR)
8655a7b1 1689 printf("repair\n");
1484e727 1690 else
8655a7b1
DW		 1691 printf("<unknown:%d>\n", migr_type(dev));
1692 } else
1693 printf("idle\n");
3393c6af
DW		 1694 printf(" Map State : %s", map_state_str[map->map_state]);
1695 if (dev->vol.migr_state) {
238c0a71 1696 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983 1697
b10b37b8 1698 printf(" <-- %s", map_state_str[map->map_state]);
4036e7ee 1699 printf("\n Checkpoint : %llu ", vol_curr_migr_unit(dev));
089f9d79 1700 if (is_gen_migration(dev) && (slot > 1 || slot < 0))
464d40e8
LD		 1701 printf("(N/A)");
1702 else
1703 printf("(%llu)", (unsigned long long)
1704 blocks_per_migr_unit(super, dev));
3393c6af
DW		 1705 }
1706 printf("\n");
2432ce9b
AP		 1707 printf(" Dirty State : %s\n", (dev->vol.dirty & RAIDVOL_DIRTY) ?
1708 "dirty" : "clean");
1709 printf(" RWH Policy : ");
c2462068 1710 if (dev->rwh_policy == RWH_OFF || dev->rwh_policy == RWH_MULTIPLE_OFF)
2432ce9b
AP		 1711 printf("off\n");
1712 else if (dev->rwh_policy == RWH_DISTRIBUTED)
1713 printf("PPL distributed\n");
1714 else if (dev->rwh_policy == RWH_JOURNALING_DRIVE)
1715 printf("PPL journaling drive\n");
c2462068
PB		 1716 else if (dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
1717 printf("Multiple distributed PPLs\n");
1718 else if (dev->rwh_policy == RWH_MULTIPLE_PPLS_JOURNALING_DRIVE)
1719 printf("Multiple PPLs on journaling drive\n");
fbc42556
JR		 1720 else if (dev->rwh_policy == RWH_BITMAP)
1721 printf("Write-intent bitmap\n");
2432ce9b
AP		 1722 else
1723 printf("<unknown:%d>\n", dev->rwh_policy);
ba1b3bc8
AP		 1724
1725 printf(" Volume ID : %u\n", dev->my_vol_raid_dev_num);
cdddbdbc
DW		 1726}
1727
ef5c214e
MK		 1728static void print_imsm_disk(struct imsm_disk *disk,
1729 int index,
1730 __u32 reserved,
1731 unsigned int sector_size) {
1f24f035 1732 char str[MAX_RAID_SERIAL_LEN + 1];
cdddbdbc
DW		 1733 __u64 sz;
1734
0ec1f4e8 1735 if (index < -1 || !disk)
e9d82038
DW		 1736 return;
1737
cdddbdbc 1738 printf("\n");
1f24f035 1739 snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial);
0ec1f4e8
DW		 1740 if (index >= 0)
1741 printf(" Disk%02d Serial : %s\n", index, str);
1742 else
1743 printf(" Disk Serial : %s\n", str);
2432ce9b
AP		 1744 printf(" State :%s%s%s%s\n", is_spare(disk) ? " spare" : "",
1745 is_configured(disk) ? " active" : "",
1746 is_failed(disk) ? " failed" : "",
1747 is_journal(disk) ? " journal" : "");
cdddbdbc 1748 printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id));
5551b113 1749 sz = total_blocks(disk) - reserved;
ef5c214e
MK		 1750 printf(" Usable Size : %llu%s\n",
1751 (unsigned long long)sz * 512 / sector_size,
cdddbdbc
DW		 1752 human_size(sz * 512));
1753}
1754
de44e46f
PB		 1755void convert_to_4k_imsm_migr_rec(struct intel_super *super)
1756{
1757 struct migr_record *migr_rec = super->migr_rec;
1758
1759 migr_rec->blocks_per_unit /= IMSM_4K_DIV;
de44e46f
PB		 1760 migr_rec->dest_depth_per_unit /= IMSM_4K_DIV;
1761 split_ull((join_u32(migr_rec->post_migr_vol_cap,
1762 migr_rec->post_migr_vol_cap_hi) / IMSM_4K_DIV),
1763 &migr_rec->post_migr_vol_cap, &migr_rec->post_migr_vol_cap_hi);
9f421827
PB		 1764 set_migr_chkp_area_pba(migr_rec,
1765 migr_chkp_area_pba(migr_rec) / IMSM_4K_DIV);
1766 set_migr_dest_1st_member_lba(migr_rec,
1767 migr_dest_1st_member_lba(migr_rec) / IMSM_4K_DIV);
de44e46f
PB		 1768}
1769
f36a9ecd
PB		 1770void convert_to_4k_imsm_disk(struct imsm_disk *disk)
1771{
1772 set_total_blocks(disk, (total_blocks(disk)/IMSM_4K_DIV));
1773}
1774
1775void convert_to_4k(struct intel_super *super)
1776{
1777 struct imsm_super *mpb = super->anchor;
1778 struct imsm_disk *disk;
1779 int i;
e4467bc7 1780 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
f36a9ecd
PB		 1781
1782 for (i = 0; i < mpb->num_disks ; i++) {
1783 disk = __get_imsm_disk(mpb, i);
1784 /* disk */
1785 convert_to_4k_imsm_disk(disk);
1786 }
1787 for (i = 0; i < mpb->num_raid_devs; i++) {
1788 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
1789 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1790 /* dev */
fcc2c9da 1791 set_imsm_dev_size(dev, imsm_dev_size(dev)/IMSM_4K_DIV);
4036e7ee
MT		 1792 set_vol_curr_migr_unit(dev,
1793 vol_curr_migr_unit(dev) / IMSM_4K_DIV);
f36a9ecd
PB		 1794
1795 /* map0 */
1796 set_blocks_per_member(map, blocks_per_member(map)/IMSM_4K_DIV);
1797 map->blocks_per_strip /= IMSM_4K_DIV;
1798 set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV);
1799
1800 if (dev->vol.migr_state) {
1801 /* map1 */
1802 map = get_imsm_map(dev, MAP_1);
1803 set_blocks_per_member(map,
1804 blocks_per_member(map)/IMSM_4K_DIV);
1805 map->blocks_per_strip /= IMSM_4K_DIV;
1806 set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV);
1807 }
1808 }
e4467bc7
TM		 1809 if (bbm_log_size) {
1810 struct bbm_log *log = (void *)mpb +
1811 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
1812 __u32 i;
1813
1814 for (i = 0; i < log->entry_count; i++) {
1815 struct bbm_log_entry *entry =
1816 &log->marked_block_entries[i];
1817
1818 __u8 count = entry->marked_count + 1;
1819 unsigned long long sector =
1820 __le48_to_cpu(&entry->defective_block_start);
1821
1822 entry->defective_block_start =
1823 __cpu_to_le48(sector/IMSM_4K_DIV);
1824 entry->marked_count = max(count/IMSM_4K_DIV, 1) - 1;
1825 }
1826 }
f36a9ecd
PB		 1827
1828 mpb->check_sum = __gen_imsm_checksum(mpb);
1829}
1830
520e69e2
AK		 1831void examine_migr_rec_imsm(struct intel_super *super)
1832{
1833 struct migr_record *migr_rec = super->migr_rec;
1834 struct imsm_super *mpb = super->anchor;
1835 int i;
1836
1837 for (i = 0; i < mpb->num_raid_devs; i++) {
1838 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
3136abe5 1839 struct imsm_map *map;
b4ab44d8 1840 int slot = -1;
3136abe5 1841
520e69e2
AK		 1842 if (is_gen_migration(dev) == 0)
1843 continue;
1844
1845 printf("\nMigration Record Information:");
3136abe5 1846
44bfe6df
AK		 1847 /* first map under migration */
1848 map = get_imsm_map(dev, MAP_0);
3136abe5
AK		 1849 if (map)
1850 slot = get_imsm_disk_slot(map, super->disks->index);
089f9d79 1851 if (map == NULL || slot > 1 || slot < 0) {
520e69e2
AK		 1852 printf(" Empty\n ");
1853 printf("Examine one of first two disks in array\n");
1854 break;
1855 }
1856 printf("\n Status : ");
1857 if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL)
1858 printf("Normal\n");
1859 else
1860 printf("Contains Data\n");
9f421827
PB		 1861 printf(" Current Unit : %llu\n",
1862 current_migr_unit(migr_rec));
520e69e2
AK		 1863 printf(" Family : %u\n",
1864 __le32_to_cpu(migr_rec->family_num));
1865 printf(" Ascending : %u\n",
1866 __le32_to_cpu(migr_rec->ascending_migr));
1867 printf(" Blocks Per Unit : %u\n",
1868 __le32_to_cpu(migr_rec->blocks_per_unit));
1869 printf(" Dest. Depth Per Unit : %u\n",
1870 __le32_to_cpu(migr_rec->dest_depth_per_unit));
9f421827
PB		 1871 printf(" Checkpoint Area pba : %llu\n",
1872 migr_chkp_area_pba(migr_rec));
1873 printf(" First member lba : %llu\n",
1874 migr_dest_1st_member_lba(migr_rec));
1875 printf(" Total Number of Units : %llu\n",
1876 get_num_migr_units(migr_rec));
1877 printf(" Size of volume : %llu\n",
1878 join_u32(migr_rec->post_migr_vol_cap,
1879 migr_rec->post_migr_vol_cap_hi));
520e69e2
AK		 1880 printf(" Record was read from : %u\n",
1881 __le32_to_cpu(migr_rec->ckpt_read_disk_num));
1882
1883 break;
1884 }
1885}
f36a9ecd 1886
de44e46f
PB		 1887void convert_from_4k_imsm_migr_rec(struct intel_super *super)
1888{
1889 struct migr_record *migr_rec = super->migr_rec;
1890
1891 migr_rec->blocks_per_unit *= IMSM_4K_DIV;
de44e46f
PB		 1892 migr_rec->dest_depth_per_unit *= IMSM_4K_DIV;
1893 split_ull((join_u32(migr_rec->post_migr_vol_cap,
1894 migr_rec->post_migr_vol_cap_hi) * IMSM_4K_DIV),
1895 &migr_rec->post_migr_vol_cap,
1896 &migr_rec->post_migr_vol_cap_hi);
9f421827
PB		 1897 set_migr_chkp_area_pba(migr_rec,
1898 migr_chkp_area_pba(migr_rec) * IMSM_4K_DIV);
1899 set_migr_dest_1st_member_lba(migr_rec,
1900 migr_dest_1st_member_lba(migr_rec) * IMSM_4K_DIV);
de44e46f
PB		 1901}
1902
f36a9ecd
PB		 1903void convert_from_4k(struct intel_super *super)
1904{
1905 struct imsm_super *mpb = super->anchor;
1906 struct imsm_disk *disk;
1907 int i;
e4467bc7 1908 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
f36a9ecd
PB		 1909
1910 for (i = 0; i < mpb->num_disks ; i++) {
1911 disk = __get_imsm_disk(mpb, i);
1912 /* disk */
1913 set_total_blocks(disk, (total_blocks(disk)*IMSM_4K_DIV));
1914 }
1915
1916 for (i = 0; i < mpb->num_raid_devs; i++) {
1917 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
1918 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1919 /* dev */
fcc2c9da 1920 set_imsm_dev_size(dev, imsm_dev_size(dev)*IMSM_4K_DIV);
4036e7ee
MT		 1921 set_vol_curr_migr_unit(dev,
1922 vol_curr_migr_unit(dev) * IMSM_4K_DIV);
f36a9ecd
PB		 1923
1924 /* map0 */
1925 set_blocks_per_member(map, blocks_per_member(map)*IMSM_4K_DIV);
1926 map->blocks_per_strip *= IMSM_4K_DIV;
1927 set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV);
1928
1929 if (dev->vol.migr_state) {
1930 /* map1 */
1931 map = get_imsm_map(dev, MAP_1);
1932 set_blocks_per_member(map,
1933 blocks_per_member(map)*IMSM_4K_DIV);
1934 map->blocks_per_strip *= IMSM_4K_DIV;
1935 set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV);
1936 }
1937 }
e4467bc7
TM		 1938 if (bbm_log_size) {
1939 struct bbm_log *log = (void *)mpb +
1940 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
1941 __u32 i;
1942
1943 for (i = 0; i < log->entry_count; i++) {
1944 struct bbm_log_entry *entry =
1945 &log->marked_block_entries[i];
1946
1947 __u8 count = entry->marked_count + 1;
1948 unsigned long long sector =
1949 __le48_to_cpu(&entry->defective_block_start);
1950
1951 entry->defective_block_start =
1952 __cpu_to_le48(sector*IMSM_4K_DIV);
1953 entry->marked_count = count*IMSM_4K_DIV - 1;
1954 }
1955 }
f36a9ecd
PB		 1956
1957 mpb->check_sum = __gen_imsm_checksum(mpb);
1958}
1959
19482bcc
AK		 1960/*******************************************************************************
1961 * function: imsm_check_attributes
1962 * Description: Function checks if features represented by attributes flags
1011e834 1963 * are supported by mdadm.
19482bcc
AK		 1964 * Parameters:
1965 * attributes - Attributes read from metadata
1966 * Returns:
1011e834
N		 1967 * 0 - passed attributes contains unsupported features flags
1968 * 1 - all features are supported
19482bcc
AK		 1969 ******************************************************************************/
1970static int imsm_check_attributes(__u32 attributes)
1971{
1972 int ret_val = 1;
418f9b36
N		 1973 __u32 not_supported = MPB_ATTRIB_SUPPORTED^0xffffffff;
1974
1975 not_supported &= ~MPB_ATTRIB_IGNORED;
19482bcc
AK		 1976
1977 not_supported &= attributes;
1978 if (not_supported) {
e7b84f9d 1979 pr_err("(IMSM): Unsupported attributes : %x\n",
418f9b36 1980 (unsigned)__le32_to_cpu(not_supported));
19482bcc
AK		 1981 if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) {
1982 dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY \n");
1983 not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY;
1984 }
1985 if (not_supported & MPB_ATTRIB_2TB) {
1986 dprintf("\t\tMPB_ATTRIB_2TB\n");
1987 not_supported ^= MPB_ATTRIB_2TB;
1988 }
1989 if (not_supported & MPB_ATTRIB_RAID0) {
1990 dprintf("\t\tMPB_ATTRIB_RAID0\n");
1991 not_supported ^= MPB_ATTRIB_RAID0;
1992 }
1993 if (not_supported & MPB_ATTRIB_RAID1) {
1994 dprintf("\t\tMPB_ATTRIB_RAID1\n");
1995 not_supported ^= MPB_ATTRIB_RAID1;
1996 }
1997 if (not_supported & MPB_ATTRIB_RAID10) {
1998 dprintf("\t\tMPB_ATTRIB_RAID10\n");
1999 not_supported ^= MPB_ATTRIB_RAID10;
2000 }
2001 if (not_supported & MPB_ATTRIB_RAID1E) {
2002 dprintf("\t\tMPB_ATTRIB_RAID1E\n");
2003 not_supported ^= MPB_ATTRIB_RAID1E;
2004 }
2005 if (not_supported & MPB_ATTRIB_RAID5) {
2006 dprintf("\t\tMPB_ATTRIB_RAID5\n");
2007 not_supported ^= MPB_ATTRIB_RAID5;
2008 }
2009 if (not_supported & MPB_ATTRIB_RAIDCNG) {
2010 dprintf("\t\tMPB_ATTRIB_RAIDCNG\n");
2011 not_supported ^= MPB_ATTRIB_RAIDCNG;
2012 }
2013 if (not_supported & MPB_ATTRIB_BBM) {
2014 dprintf("\t\tMPB_ATTRIB_BBM\n");
2015 not_supported ^= MPB_ATTRIB_BBM;
2016 }
2017 if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) {
2018 dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY (== MPB_ATTRIB_LEGACY)\n");
2019 not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY;
2020 }
2021 if (not_supported & MPB_ATTRIB_EXP_STRIPE_SIZE) {
2022 dprintf("\t\tMPB_ATTRIB_EXP_STRIP_SIZE\n");
2023 not_supported ^= MPB_ATTRIB_EXP_STRIPE_SIZE;
2024 }
2025 if (not_supported & MPB_ATTRIB_2TB_DISK) {
2026 dprintf("\t\tMPB_ATTRIB_2TB_DISK\n");
2027 not_supported ^= MPB_ATTRIB_2TB_DISK;
2028 }
2029 if (not_supported & MPB_ATTRIB_NEVER_USE2) {
2030 dprintf("\t\tMPB_ATTRIB_NEVER_USE2\n");
2031 not_supported ^= MPB_ATTRIB_NEVER_USE2;
2032 }
2033 if (not_supported & MPB_ATTRIB_NEVER_USE) {
2034 dprintf("\t\tMPB_ATTRIB_NEVER_USE\n");
2035 not_supported ^= MPB_ATTRIB_NEVER_USE;
2036 }
2037
2038 if (not_supported)
1ade5cc1 2039 dprintf("(IMSM): Unknown attributes : %x\n", not_supported);
19482bcc
AK		 2040
2041 ret_val = 0;
2042 }
2043
2044 return ret_val;
2045}
2046
a5d85af7 2047static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map);
44470971 2048
cdddbdbc
DW		 2049static void examine_super_imsm(struct supertype *st, char *homehost)
2050{
2051 struct intel_super *super = st->sb;
949c47a0 2052 struct imsm_super *mpb = super->anchor;
cdddbdbc
DW		 2053 char str[MAX_SIGNATURE_LENGTH];
2054 int i;
27fd6274
DW		 2055 struct mdinfo info;
2056 char nbuf[64];
cdddbdbc 2057 __u32 sum;
14e8215b 2058 __u32 reserved = imsm_reserved_sectors(super, super->disks);
94827db3 2059 struct dl *dl;
e48aed3c 2060 time_t creation_time;
27fd6274 2061
618f4e6d
XN		 2062 strncpy(str, (char *)mpb->sig, MPB_SIG_LEN);
2063 str[MPB_SIG_LEN-1] = '\0';
cdddbdbc 2064 printf(" Magic : %s\n", str);
cdddbdbc 2065 printf(" Version : %s\n", get_imsm_version(mpb));
148acb7b 2066 printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num));
cdddbdbc
DW		 2067 printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num));
2068 printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num));
e48aed3c
AP		 2069 creation_time = __le64_to_cpu(mpb->creation_time);
2070 printf(" Creation Time : %.24s\n",
2071 creation_time ? ctime(&creation_time) : "Unknown");
19482bcc
AK		 2072 printf(" Attributes : ");
2073 if (imsm_check_attributes(mpb->attributes))
2074 printf("All supported\n");
2075 else
2076 printf("not supported\n");
a5d85af7 2077 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2078 fname_from_uuid(st, &info, nbuf, ':');
27fd6274 2079 printf(" UUID : %s\n", nbuf + 5);
cdddbdbc
DW		 2080 sum = __le32_to_cpu(mpb->check_sum);
2081 printf(" Checksum : %08x %s\n", sum,
949c47a0 2082 __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect");
f36a9ecd 2083 printf(" MPB Sectors : %d\n", mpb_sectors(mpb, super->sector_size));
cdddbdbc
DW		 2084 printf(" Disks : %d\n", mpb->num_disks);
2085 printf(" RAID Devices : %d\n", mpb->num_raid_devs);
ef5c214e
MK		 2086 print_imsm_disk(__get_imsm_disk(mpb, super->disks->index),
2087 super->disks->index, reserved, super->sector_size);
8d67477f 2088 if (get_imsm_bbm_log_size(super->bbm_log)) {
604b746f
JD		 2089 struct bbm_log *log = super->bbm_log;
2090
2091 printf("\n");
2092 printf("Bad Block Management Log:\n");
2093 printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size));
2094 printf(" Signature : %x\n", __le32_to_cpu(log->signature));
2095 printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count));
604b746f 2096 }
44470971
DW		 2097 for (i = 0; i < mpb->num_raid_devs; i++) {
2098 struct mdinfo info;
2099 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
2100
2101 super->current_vol = i;
a5d85af7 2102 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2103 fname_from_uuid(st, &info, nbuf, ':');
c47b0ff6 2104 print_imsm_dev(super, dev, nbuf + 5, super->disks->index);
44470971 2105 }
cdddbdbc
DW		 2106 for (i = 0; i < mpb->num_disks; i++) {
2107 if (i == super->disks->index)
2108 continue;
ef5c214e
MK		 2109 print_imsm_disk(__get_imsm_disk(mpb, i), i, reserved,
2110 super->sector_size);
cdddbdbc 2111 }
94827db3 2112
0ec1f4e8
DW		 2113 for (dl = super->disks; dl; dl = dl->next)
2114 if (dl->index == -1)
ef5c214e
MK		 2115 print_imsm_disk(&dl->disk, -1, reserved,
2116 super->sector_size);
520e69e2
AK		 2117
2118 examine_migr_rec_imsm(super);
cdddbdbc
DW		 2119}
2120
061f2c6a 2121static void brief_examine_super_imsm(struct supertype *st, int verbose)
cdddbdbc 2122{
27fd6274 2123 /* We just write a generic IMSM ARRAY entry */
ff54de6e
N		 2124 struct mdinfo info;
2125 char nbuf[64];
1e7bc0ed 2126 struct intel_super *super = st->sb;
1e7bc0ed 2127
0d5a423f
DW		 2128 if (!super->anchor->num_raid_devs) {
2129 printf("ARRAY metadata=imsm\n");
1e7bc0ed 2130 return;
0d5a423f 2131 }
ff54de6e 2132
a5d85af7 2133 getinfo_super_imsm(st, &info, NULL);
4737ae25
N		 2134 fname_from_uuid(st, &info, nbuf, ':');
2135 printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5);
2136}
2137
2138static void brief_examine_subarrays_imsm(struct supertype *st, int verbose)
2139{
2140 /* We just write a generic IMSM ARRAY entry */
2141 struct mdinfo info;
2142 char nbuf[64];
2143 char nbuf1[64];
2144 struct intel_super *super = st->sb;
2145 int i;
2146
2147 if (!super->anchor->num_raid_devs)
2148 return;
2149
a5d85af7 2150 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2151 fname_from_uuid(st, &info, nbuf, ':');
1e7bc0ed
DW		 2152 for (i = 0; i < super->anchor->num_raid_devs; i++) {
2153 struct imsm_dev *dev = get_imsm_dev(super, i);
2154
2155 super->current_vol = i;
a5d85af7 2156 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2157 fname_from_uuid(st, &info, nbuf1, ':');
1124b3cf 2158 printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n",
cf8de691 2159 dev->volume, nbuf + 5, i, nbuf1 + 5);
1e7bc0ed 2160 }
cdddbdbc
DW		 2161}
2162
9d84c8ea
DW		 2163static void export_examine_super_imsm(struct supertype *st)
2164{
2165 struct intel_super *super = st->sb;
2166 struct imsm_super *mpb = super->anchor;
2167 struct mdinfo info;
2168 char nbuf[64];
2169
a5d85af7 2170 getinfo_super_imsm(st, &info, NULL);
9d84c8ea
DW		 2171 fname_from_uuid(st, &info, nbuf, ':');
2172 printf("MD_METADATA=imsm\n");
2173 printf("MD_LEVEL=container\n");
2174 printf("MD_UUID=%s\n", nbuf+5);
2175 printf("MD_DEVICES=%u\n", mpb->num_disks);
e48aed3c 2176 printf("MD_CREATION_TIME=%llu\n", __le64_to_cpu(mpb->creation_time));
9d84c8ea
DW		 2177}
2178
b771faef
BK		 2179static void detail_super_imsm(struct supertype *st, char *homehost,
2180 char *subarray)
cdddbdbc 2181{
3ebe00a1
DW		 2182 struct mdinfo info;
2183 char nbuf[64];
b771faef
BK		 2184 struct intel_super *super = st->sb;
2185 int temp_vol = super->current_vol;
2186
2187 if (subarray)
2188 super->current_vol = strtoul(subarray, NULL, 10);
3ebe00a1 2189
a5d85af7 2190 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2191 fname_from_uuid(st, &info, nbuf, ':');
65884368 2192 printf("\n UUID : %s\n", nbuf + 5);
b771faef
BK		 2193
2194 super->current_vol = temp_vol;
cdddbdbc
DW		 2195}
2196
b771faef 2197static void brief_detail_super_imsm(struct supertype *st, char *subarray)
cdddbdbc 2198{
ff54de6e
N		 2199 struct mdinfo info;
2200 char nbuf[64];
b771faef
BK		 2201 struct intel_super *super = st->sb;
2202 int temp_vol = super->current_vol;
2203
2204 if (subarray)
2205 super->current_vol = strtoul(subarray, NULL, 10);
2206
a5d85af7 2207 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2208 fname_from_uuid(st, &info, nbuf, ':');
ff54de6e 2209 printf(" UUID=%s", nbuf + 5);
b771faef
BK		 2210
2211 super->current_vol = temp_vol;
cdddbdbc 2212}
d665cc31 2213
6da53c0e
BK		 2214static int imsm_read_serial(int fd, char *devname, __u8 *serial,
2215 size_t serial_buf_len);
d665cc31
DW		 2216static void fd2devname(int fd, char *name);
2217
120dc887 2218static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose)
d665cc31 2219{
120dc887
LM		 2220 /* dump an unsorted list of devices attached to AHCI Intel storage
2221 * controller, as well as non-connected ports
d665cc31
DW		 2222 */
2223 int hba_len = strlen(hba_path) + 1;
2224 struct dirent *ent;
2225 DIR *dir;
2226 char *path = NULL;
2227 int err = 0;
2228 unsigned long port_mask = (1 << port_count) - 1;
2229
f21e18ca 2230 if (port_count > (int)sizeof(port_mask) * 8) {
ba728be7 2231 if (verbose > 0)
e7b84f9d 2232 pr_err("port_count %d out of range\n", port_count);
d665cc31
DW		 2233 return 2;
2234 }
2235
2236 /* scroll through /sys/dev/block looking for devices attached to
2237 * this hba
2238 */
2239 dir = opendir("/sys/dev/block");
1a6dd6b9
PB		 2240 if (!dir)
2241 return 1;
2242
2243 for (ent = readdir(dir); ent; ent = readdir(dir)) {
d665cc31
DW		 2244 int fd;
2245 char model[64];
2246 char vendor[64];
2247 char buf[1024];
2248 int major, minor;
fcebeb77 2249 char device[PATH_MAX];
d665cc31
DW		 2250 char *c;
2251 int port;
2252 int type;
2253
2254 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
2255 continue;
7c798f87 2256 path = devt_to_devpath(makedev(major, minor), 1, NULL);
d665cc31
DW		 2257 if (!path)
2258 continue;
2259 if (!path_attached_to_hba(path, hba_path)) {
2260 free(path);
2261 path = NULL;
2262 continue;
2263 }
2264
fcebeb77
MT		 2265 /* retrieve the scsi device */
2266 if (!devt_to_devpath(makedev(major, minor), 1, device)) {
ba728be7 2267 if (verbose > 0)
fcebeb77 2268 pr_err("failed to get device\n");
d665cc31
DW		 2269 err = 2;
2270 break;
2271 }
fcebeb77 2272 if (devpath_to_char(device, "type", buf, sizeof(buf), 0)) {
d665cc31 2273 err = 2;
d665cc31
DW		 2274 break;
2275 }
2276 type = strtoul(buf, NULL, 10);
2277
2278 /* if it's not a disk print the vendor and model */
2279 if (!(type == 0 || type == 7 || type == 14)) {
2280 vendor[0] = '\0';
2281 model[0] = '\0';
fcebeb77
MT		 2282
2283 if (devpath_to_char(device, "vendor", buf,
2284 sizeof(buf), 0) == 0) {
d665cc31
DW		 2285 strncpy(vendor, buf, sizeof(vendor));
2286 vendor[sizeof(vendor) - 1] = '\0';
2287 c = (char *) &vendor[sizeof(vendor) - 1];
2288 while (isspace(*c) || *c == '\0')
2289 *c-- = '\0';
2290
2291 }
fcebeb77
MT		 2292
2293 if (devpath_to_char(device, "model", buf,
2294 sizeof(buf), 0) == 0) {
d665cc31
DW		 2295 strncpy(model, buf, sizeof(model));
2296 model[sizeof(model) - 1] = '\0';
2297 c = (char *) &model[sizeof(model) - 1];
2298 while (isspace(*c) || *c == '\0')
2299 *c-- = '\0';
2300 }
2301
2302 if (vendor[0] && model[0])
2303 sprintf(buf, "%.64s %.64s", vendor, model);
2304 else
2305 switch (type) { /* numbers from hald/linux/device.c */
2306 case 1: sprintf(buf, "tape"); break;
2307 case 2: sprintf(buf, "printer"); break;
2308 case 3: sprintf(buf, "processor"); break;
2309 case 4:
2310 case 5: sprintf(buf, "cdrom"); break;
2311 case 6: sprintf(buf, "scanner"); break;
2312 case 8: sprintf(buf, "media_changer"); break;
2313 case 9: sprintf(buf, "comm"); break;
2314 case 12: sprintf(buf, "raid"); break;
2315 default: sprintf(buf, "unknown");
2316 }
2317 } else
2318 buf[0] = '\0';
d665cc31
DW		 2319
2320 /* chop device path to 'host%d' and calculate the port number */
2321 c = strchr(&path[hba_len], '/');
4e5e717d 2322 if (!c) {
ba728be7 2323 if (verbose > 0)
e7b84f9d 2324 pr_err("%s - invalid path name\n", path + hba_len);
4e5e717d
AW		 2325 err = 2;
2326 break;
2327 }
d665cc31 2328 *c = '\0';
0858eccf
AP		 2329 if ((sscanf(&path[hba_len], "ata%d", &port) == 1) ||
2330 ((sscanf(&path[hba_len], "host%d", &port) == 1)))
d665cc31
DW		 2331 port -= host_base;
2332 else {
ba728be7 2333 if (verbose > 0) {
d665cc31 2334 *c = '/'; /* repair the full string */
e7b84f9d 2335 pr_err("failed to determine port number for %s\n",
d665cc31
DW		 2336 path);
2337 }
2338 err = 2;
2339 break;
2340 }
2341
2342 /* mark this port as used */
2343 port_mask &= ~(1 << port);
2344
2345 /* print out the device information */
2346 if (buf[0]) {
2347 printf(" Port%d : - non-disk device (%s) -\n", port, buf);
2348 continue;
2349 }
2350
2351 fd = dev_open(ent->d_name, O_RDONLY);
2352 if (fd < 0)
2353 printf(" Port%d : - disk info unavailable -\n", port);
2354 else {
2355 fd2devname(fd, buf);
2356 printf(" Port%d : %s", port, buf);
6da53c0e
BK		 2357 if (imsm_read_serial(fd, NULL, (__u8 *)buf,
2358 sizeof(buf)) == 0)
2359 printf(" (%s)\n", buf);
d665cc31 2360 else
664d5325 2361 printf(" ()\n");
4dab422a 2362 close(fd);
d665cc31 2363 }
d665cc31
DW		 2364 free(path);
2365 path = NULL;
2366 }
2367 if (path)
2368 free(path);
2369 if (dir)
2370 closedir(dir);
2371 if (err == 0) {
2372 int i;
2373
2374 for (i = 0; i < port_count; i++)
2375 if (port_mask & (1 << i))
2376 printf(" Port%d : - no device attached -\n", i);
2377 }
2378
2379 return err;
2380}
2381
6da53c0e 2382static int print_nvme_info(struct sys_dev *hba)
60f0f54d
PB		 2383{
2384 struct dirent *ent;
2385 DIR *dir;
60f0f54d 2386
6da53c0e 2387 dir = opendir("/sys/block/");
b9135011 2388 if (!dir)
b5eece69 2389 return 1;
b9135011
JS		 2390
2391 for (ent = readdir(dir); ent; ent = readdir(dir)) {
8662f92d
MT		 2392 char ns_path[PATH_MAX];
2393 char cntrl_path[PATH_MAX];
2394 char buf[PATH_MAX];
2395 int fd = -1;
60f0f54d 2396
8662f92d
MT		 2397 if (!strstr(ent->d_name, "nvme"))
2398 goto skip;
d835518b 2399
8662f92d
MT		 2400 fd = open_dev(ent->d_name);
2401 if (fd < 0)
2402 goto skip;
d835518b 2403
8662f92d
MT		 2404 if (!diskfd_to_devpath(fd, 0, ns_path) ||
2405 !diskfd_to_devpath(fd, 1, cntrl_path))
2406 goto skip;
2407
2408 if (!path_attached_to_hba(cntrl_path, hba->path))
2409 goto skip;
2410
2411 if (!imsm_is_nvme_namespace_supported(fd, 0))
2412 goto skip;
2413
2414 fd2devname(fd, buf);
2415 if (hba->type == SYS_DEV_VMD)
2416 printf(" NVMe under VMD : %s", buf);
2417 else if (hba->type == SYS_DEV_NVME)
2418 printf(" NVMe Device : %s", buf);
2419
2420 if (!imsm_read_serial(fd, NULL, (__u8 *)buf,
2421 sizeof(buf)))
2422 printf(" (%s)\n", buf);
2423 else
2424 printf("()\n");
2425
2426skip:
2427 if (fd > -1)
d835518b 2428 close(fd);
60f0f54d
PB		 2429 }
2430
b9135011 2431 closedir(dir);
b5eece69 2432 return 0;
60f0f54d
PB		 2433}
2434
120dc887
LM		 2435static void print_found_intel_controllers(struct sys_dev *elem)
2436{
2437 for (; elem; elem = elem->next) {
e7b84f9d 2438 pr_err("found Intel(R) ");
120dc887
LM		 2439 if (elem->type == SYS_DEV_SATA)
2440 fprintf(stderr, "SATA ");
155cbb4c
LM		 2441 else if (elem->type == SYS_DEV_SAS)
2442 fprintf(stderr, "SAS ");
0858eccf
AP		 2443 else if (elem->type == SYS_DEV_NVME)
2444 fprintf(stderr, "NVMe ");
60f0f54d
PB		 2445
2446 if (elem->type == SYS_DEV_VMD)
2447 fprintf(stderr, "VMD domain");
2448 else
2449 fprintf(stderr, "RAID controller");
2450
120dc887
LM		 2451 if (elem->pci_id)
2452 fprintf(stderr, " at %s", elem->pci_id);
2453 fprintf(stderr, ".\n");
2454 }
2455 fflush(stderr);
2456}
2457
120dc887
LM		 2458static int ahci_get_port_count(const char *hba_path, int *port_count)
2459{
2460 struct dirent *ent;
2461 DIR *dir;
2462 int host_base = -1;
2463
2464 *port_count = 0;
2465 if ((dir = opendir(hba_path)) == NULL)
2466 return -1;
2467
2468 for (ent = readdir(dir); ent; ent = readdir(dir)) {
2469 int host;
2470
0858eccf
AP		 2471 if ((sscanf(ent->d_name, "ata%d", &host) != 1) &&
2472 ((sscanf(ent->d_name, "host%d", &host) != 1)))
120dc887
LM		 2473 continue;
2474 if (*port_count == 0)
2475 host_base = host;
2476 else if (host < host_base)
2477 host_base = host;
2478
2479 if (host + 1 > *port_count + host_base)
2480 *port_count = host + 1 - host_base;
2481 }
2482 closedir(dir);
2483 return host_base;
2484}
2485
a891a3c2
LM		 2486static void print_imsm_capability(const struct imsm_orom *orom)
2487{
0858eccf
AP		 2488 printf(" Platform : Intel(R) ");
2489 if (orom->capabilities == 0 && orom->driver_features == 0)
2490 printf("Matrix Storage Manager\n");
ab0c6bb9
AP		 2491 else if (imsm_orom_is_enterprise(orom) && orom->major_ver >= 6)
2492 printf("Virtual RAID on CPU\n");
0858eccf
AP		 2493 else
2494 printf("Rapid Storage Technology%s\n",
2495 imsm_orom_is_enterprise(orom) ? " enterprise" : "");
2496 if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build)
2497 printf(" Version : %d.%d.%d.%d\n", orom->major_ver,
2498 orom->minor_ver, orom->hotfix_ver, orom->build);
a891a3c2
LM		 2499 printf(" RAID Levels :%s%s%s%s%s\n",
2500 imsm_orom_has_raid0(orom) ? " raid0" : "",
2501 imsm_orom_has_raid1(orom) ? " raid1" : "",
2502 imsm_orom_has_raid1e(orom) ? " raid1e" : "",
2503 imsm_orom_has_raid10(orom) ? " raid10" : "",
2504 imsm_orom_has_raid5(orom) ? " raid5" : "");
2505 printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
2506 imsm_orom_has_chunk(orom, 2) ? " 2k" : "",
2507 imsm_orom_has_chunk(orom, 4) ? " 4k" : "",
2508 imsm_orom_has_chunk(orom, 8) ? " 8k" : "",
2509 imsm_orom_has_chunk(orom, 16) ? " 16k" : "",
2510 imsm_orom_has_chunk(orom, 32) ? " 32k" : "",
2511 imsm_orom_has_chunk(orom, 64) ? " 64k" : "",
2512 imsm_orom_has_chunk(orom, 128) ? " 128k" : "",
2513 imsm_orom_has_chunk(orom, 256) ? " 256k" : "",
2514 imsm_orom_has_chunk(orom, 512) ? " 512k" : "",
2515 imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "",
2516 imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "",
2517 imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "",
2518 imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "",
2519 imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "",
2520 imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "",
2521 imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : "");
29cd0821
CA		 2522 printf(" 2TB volumes :%s supported\n",
2523 (orom->attr & IMSM_OROM_ATTR_2TB)?"":" not");
2524 printf(" 2TB disks :%s supported\n",
2525 (orom->attr & IMSM_OROM_ATTR_2TB_DISK)?"":" not");
0e7f69a8 2526 printf(" Max Disks : %d\n", orom->tds);
0858eccf
AP		 2527 printf(" Max Volumes : %d per array, %d per %s\n",
2528 orom->vpa, orom->vphba,
2529 imsm_orom_is_nvme(orom) ? "platform" : "controller");
a891a3c2
LM		 2530 return;
2531}
2532
e50cf220
MN		 2533static void print_imsm_capability_export(const struct imsm_orom *orom)
2534{
2535 printf("MD_FIRMWARE_TYPE=imsm\n");
0858eccf
AP		 2536 if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build)
2537 printf("IMSM_VERSION=%d.%d.%d.%d\n", orom->major_ver, orom->minor_ver,
2538 orom->hotfix_ver, orom->build);
e50cf220
MN		 2539 printf("IMSM_SUPPORTED_RAID_LEVELS=%s%s%s%s%s\n",
2540 imsm_orom_has_raid0(orom) ? "raid0 " : "",
2541 imsm_orom_has_raid1(orom) ? "raid1 " : "",
2542 imsm_orom_has_raid1e(orom) ? "raid1e " : "",
2543 imsm_orom_has_raid5(orom) ? "raid10 " : "",
2544 imsm_orom_has_raid10(orom) ? "raid5 " : "");
2545 printf("IMSM_SUPPORTED_CHUNK_SIZES=%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
2546 imsm_orom_has_chunk(orom, 2) ? "2k " : "",
2547 imsm_orom_has_chunk(orom, 4) ? "4k " : "",
2548 imsm_orom_has_chunk(orom, 8) ? "8k " : "",
2549 imsm_orom_has_chunk(orom, 16) ? "16k " : "",
2550 imsm_orom_has_chunk(orom, 32) ? "32k " : "",
2551 imsm_orom_has_chunk(orom, 64) ? "64k " : "",
2552 imsm_orom_has_chunk(orom, 128) ? "128k " : "",
2553 imsm_orom_has_chunk(orom, 256) ? "256k " : "",
2554 imsm_orom_has_chunk(orom, 512) ? "512k " : "",
2555 imsm_orom_has_chunk(orom, 1024*1) ? "1M " : "",
2556 imsm_orom_has_chunk(orom, 1024*2) ? "2M " : "",
2557 imsm_orom_has_chunk(orom, 1024*4) ? "4M " : "",
2558 imsm_orom_has_chunk(orom, 1024*8) ? "8M " : "",
2559 imsm_orom_has_chunk(orom, 1024*16) ? "16M " : "",
2560 imsm_orom_has_chunk(orom, 1024*32) ? "32M " : "",
2561 imsm_orom_has_chunk(orom, 1024*64) ? "64M " : "");
2562 printf("IMSM_2TB_VOLUMES=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB) ? "yes" : "no");
2563 printf("IMSM_2TB_DISKS=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB_DISK) ? "yes" : "no");
2564 printf("IMSM_MAX_DISKS=%d\n",orom->tds);
2565 printf("IMSM_MAX_VOLUMES_PER_ARRAY=%d\n",orom->vpa);
2566 printf("IMSM_MAX_VOLUMES_PER_CONTROLLER=%d\n",orom->vphba);
2567}
2568
9eafa1de 2569static int detail_platform_imsm(int verbose, int enumerate_only, char *controller_path)
d665cc31
DW		 2570{
2571 /* There are two components to imsm platform support, the ahci SATA
2572 * controller and the option-rom. To find the SATA controller we
2573 * simply look in /sys/bus/pci/drivers/ahci to see if an ahci
2574 * controller with the Intel vendor id is present. This approach
2575 * allows mdadm to leverage the kernel's ahci detection logic, with the
2576 * caveat that if ahci.ko is not loaded mdadm will not be able to
2577 * detect platform raid capabilities. The option-rom resides in a
2578 * platform "Adapter ROM". We scan for its signature to retrieve the
2579 * platform capabilities. If raid support is disabled in the BIOS the
2580 * option-rom capability structure will not be available.
2581 */
d665cc31 2582 struct sys_dev *list, *hba;
d665cc31
DW		 2583 int host_base = 0;
2584 int port_count = 0;
9eafa1de 2585 int result=1;
d665cc31 2586
5615172f 2587 if (enumerate_only) {
a891a3c2 2588 if (check_env("IMSM_NO_PLATFORM"))
5615172f 2589 return 0;
a891a3c2
LM		 2590 list = find_intel_devices();
2591 if (!list)
2592 return 2;
2593 for (hba = list; hba; hba = hba->next) {
6b781d33
AP		 2594 if (find_imsm_capability(hba)) {
2595 result = 0;
a891a3c2
LM		 2596 break;
2597 }
9eafa1de 2598 else
6b781d33 2599 result = 2;
a891a3c2 2600 }
a891a3c2 2601 return result;
5615172f
DW		 2602 }
2603
155cbb4c
LM		 2604 list = find_intel_devices();
2605 if (!list) {
ba728be7 2606 if (verbose > 0)
7a862a02 2607 pr_err("no active Intel(R) RAID controller found.\n");
d665cc31 2608 return 2;
ba728be7 2609 } else if (verbose > 0)
155cbb4c 2610 print_found_intel_controllers(list);
d665cc31 2611
a891a3c2 2612 for (hba = list; hba; hba = hba->next) {
0858eccf 2613 if (controller_path && (compare_paths(hba->path, controller_path) != 0))
9eafa1de 2614 continue;
0858eccf 2615 if (!find_imsm_capability(hba)) {
60f0f54d 2616 char buf[PATH_MAX];
e7b84f9d 2617 pr_err("imsm capabilities not found for controller: %s (type %s)\n",
60f0f54d
PB		 2618 hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path,
2619 get_sys_dev_type(hba->type));
0858eccf
AP		 2620 continue;
2621 }
2622 result = 0;
2623 }
2624
2625 if (controller_path && result == 1) {
2626 pr_err("no active Intel(R) RAID controller found under %s\n",
2627 controller_path);
2628 return result;
2629 }
2630
5e1d6128 2631 const struct orom_entry *entry;
0858eccf 2632
5e1d6128 2633 for (entry = orom_entries; entry; entry = entry->next) {
60f0f54d 2634 if (entry->type == SYS_DEV_VMD) {
07cb1e57 2635 print_imsm_capability(&entry->orom);
32716c51
PB		 2636 printf(" 3rd party NVMe :%s supported\n",
2637 imsm_orom_has_tpv_support(&entry->orom)?"":" not");
60f0f54d
PB		 2638 for (hba = list; hba; hba = hba->next) {
2639 if (hba->type == SYS_DEV_VMD) {
2640 char buf[PATH_MAX];
60f0f54d
PB		 2641 printf(" I/O Controller : %s (%s)\n",
2642 vmd_domain_to_controller(hba, buf), get_sys_dev_type(hba->type));
6da53c0e 2643 if (print_nvme_info(hba)) {
b5eece69
PB		 2644 if (verbose > 0)
2645 pr_err("failed to get devices attached to VMD domain.\n");
2646 result |= 2;
2647 }
60f0f54d
PB		 2648 }
2649 }
07cb1e57 2650 printf("\n");
60f0f54d
PB		 2651 continue;
2652 }
0858eccf 2653
60f0f54d
PB		 2654 print_imsm_capability(&entry->orom);
2655 if (entry->type == SYS_DEV_NVME) {
0858eccf
AP		 2656 for (hba = list; hba; hba = hba->next) {
2657 if (hba->type == SYS_DEV_NVME)
6da53c0e 2658 print_nvme_info(hba);
0858eccf 2659 }
60f0f54d 2660 printf("\n");
0858eccf
AP		 2661 continue;
2662 }
2663
2664 struct devid_list *devid;
5e1d6128 2665 for (devid = entry->devid_list; devid; devid = devid->next) {
0858eccf
AP		 2666 hba = device_by_id(devid->devid);
2667 if (!hba)
2668 continue;
2669
9eafa1de
MN		 2670 printf(" I/O Controller : %s (%s)\n",
2671 hba->path, get_sys_dev_type(hba->type));
2672 if (hba->type == SYS_DEV_SATA) {
2673 host_base = ahci_get_port_count(hba->path, &port_count);
2674 if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) {
2675 if (verbose > 0)
7a862a02 2676 pr_err("failed to enumerate ports on SATA controller at %s.\n", hba->pci_id);
9eafa1de
MN		 2677 result |= 2;
2678 }
120dc887
LM		 2679 }
2680 }
0858eccf 2681 printf("\n");
d665cc31 2682 }
155cbb4c 2683
120dc887 2684 return result;
d665cc31 2685}
e50cf220 2686
9eafa1de 2687static int export_detail_platform_imsm(int verbose, char *controller_path)
e50cf220 2688{
e50cf220
MN		 2689 struct sys_dev *list, *hba;
2690 int result=1;
2691
2692 list = find_intel_devices();
2693 if (!list) {
2694 if (verbose > 0)
2695 pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_INTEL_DEVICES\n");
2696 result = 2;
e50cf220
MN		 2697 return result;
2698 }
2699
2700 for (hba = list; hba; hba = hba->next) {
9eafa1de
MN		 2701 if (controller_path && (compare_paths(hba->path,controller_path) != 0))
2702 continue;
60f0f54d
PB		 2703 if (!find_imsm_capability(hba) && verbose > 0) {
2704 char buf[PATH_MAX];
2705 pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_IMSM_CAPABLE_DEVICE_UNDER_%s\n",
2706 hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path);
2707 }
0858eccf 2708 else
e50cf220 2709 result = 0;
e50cf220
MN		 2710 }
2711
5e1d6128 2712 const struct orom_entry *entry;
0858eccf 2713
60f0f54d
PB		 2714 for (entry = orom_entries; entry; entry = entry->next) {
2715 if (entry->type == SYS_DEV_VMD) {
2716 for (hba = list; hba; hba = hba->next)
2717 print_imsm_capability_export(&entry->orom);
2718 continue;
2719 }
5e1d6128 2720 print_imsm_capability_export(&entry->orom);
60f0f54d 2721 }
0858eccf 2722
e50cf220
MN		 2723 return result;
2724}
2725
cdddbdbc
DW		 2726static int match_home_imsm(struct supertype *st, char *homehost)
2727{
5115ca67
DW		 2728 /* the imsm metadata format does not specify any host
2729 * identification information. We return -1 since we can never
2730 * confirm nor deny whether a given array is "meant" for this
148acb7b 2731 * host. We rely on compare_super and the 'family_num' fields to
5115ca67
DW		 2732 * exclude member disks that do not belong, and we rely on
2733 * mdadm.conf to specify the arrays that should be assembled.
2734 * Auto-assembly may still pick up "foreign" arrays.
2735 */
cdddbdbc 2736
9362c1c8 2737 return -1;
cdddbdbc
DW		 2738}
2739
2740static void uuid_from_super_imsm(struct supertype *st, int uuid[4])
2741{
51006d85
N		 2742 /* The uuid returned here is used for:
2743 * uuid to put into bitmap file (Create, Grow)
2744 * uuid for backup header when saving critical section (Grow)
2745 * comparing uuids when re-adding a device into an array
2746 * In these cases the uuid required is that of the data-array,
2747 * not the device-set.
2748 * uuid to recognise same set when adding a missing device back
2749 * to an array. This is a uuid for the device-set.
1011e834 2750 *
51006d85
N		 2751 * For each of these we can make do with a truncated
2752 * or hashed uuid rather than the original, as long as
2753 * everyone agrees.
2754 * In each case the uuid required is that of the data-array,
2755 * not the device-set.
43dad3d6 2756 */
51006d85
N		 2757 /* imsm does not track uuid's so we synthesis one using sha1 on
2758 * - The signature (Which is constant for all imsm array, but no matter)
148acb7b 2759 * - the orig_family_num of the container
51006d85
N		 2760 * - the index number of the volume
2761 * - the 'serial' number of the volume.
2762 * Hopefully these are all constant.
2763 */
2764 struct intel_super *super = st->sb;
43dad3d6 2765
51006d85
N		 2766 char buf[20];
2767 struct sha1_ctx ctx;
2768 struct imsm_dev *dev = NULL;
148acb7b 2769 __u32 family_num;
51006d85 2770
148acb7b
DW		 2771 /* some mdadm versions failed to set ->orig_family_num, in which
2772 * case fall back to ->family_num. orig_family_num will be
2773 * fixed up with the first metadata update.
2774 */
2775 family_num = super->anchor->orig_family_num;
2776 if (family_num == 0)
2777 family_num = super->anchor->family_num;
51006d85 2778 sha1_init_ctx(&ctx);
92bd8f8d 2779 sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx);
148acb7b 2780 sha1_process_bytes(&family_num, sizeof(__u32), &ctx);
51006d85
N		 2781 if (super->current_vol >= 0)
2782 dev = get_imsm_dev(super, super->current_vol);
2783 if (dev) {
2784 __u32 vol = super->current_vol;
2785 sha1_process_bytes(&vol, sizeof(vol), &ctx);
2786 sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx);
2787 }
2788 sha1_finish_ctx(&ctx, buf);
2789 memcpy(uuid, buf, 4*4);
cdddbdbc
DW		 2790}
2791
0d481d37 2792#if 0
4f5bc454
DW		 2793static void
2794get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p)
cdddbdbc 2795{
cdddbdbc
DW		 2796 __u8 *v = get_imsm_version(mpb);
2797 __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH;
2798 char major[] = { 0, 0, 0 };
2799 char minor[] = { 0 ,0, 0 };
2800 char patch[] = { 0, 0, 0 };
2801 char *ver_parse[] = { major, minor, patch };
2802 int i, j;
2803
2804 i = j = 0;
2805 while (*v != '\0' && v < end) {
2806 if (*v != '.' && j < 2)
2807 ver_parse[i][j++] = *v;
2808 else {
2809 i++;
2810 j = 0;
2811 }
2812 v++;
2813 }
2814
4f5bc454
DW		 2815 *m = strtol(minor, NULL, 0);
2816 *p = strtol(patch, NULL, 0);
2817}
0d481d37 2818#endif
4f5bc454 2819
1e5c6983
DW		 2820static __u32 migr_strip_blocks_resync(struct imsm_dev *dev)
2821{
2822 /* migr_strip_size when repairing or initializing parity */
238c0a71 2823 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2824 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2825
2826 switch (get_imsm_raid_level(map)) {
2827 case 5:
2828 case 10:
2829 return chunk;
2830 default:
2831 return 128*1024 >> 9;
2832 }
2833}
2834
2835static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev)
2836{
2837 /* migr_strip_size when rebuilding a degraded disk, no idea why
2838 * this is different than migr_strip_size_resync(), but it's good
2839 * to be compatible
2840 */
238c0a71 2841 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2842 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2843
2844 switch (get_imsm_raid_level(map)) {
2845 case 1:
2846 case 10:
2847 if (map->num_members % map->num_domains == 0)
2848 return 128*1024 >> 9;
2849 else
2850 return chunk;
2851 case 5:
2852 return max((__u32) 64*1024 >> 9, chunk);
2853 default:
2854 return 128*1024 >> 9;
2855 }
2856}
2857
2858static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev)
2859{
238c0a71
AK		 2860 struct imsm_map *lo = get_imsm_map(dev, MAP_0);
2861 struct imsm_map *hi = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2862 __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip);
2863 __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip);
2864
2865 return max((__u32) 1, hi_chunk / lo_chunk);
2866}
2867
2868static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev)
2869{
238c0a71 2870 struct imsm_map *lo = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2871 int level = get_imsm_raid_level(lo);
2872
2873 if (level == 1 || level == 10) {
238c0a71 2874 struct imsm_map *hi = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2875
2876 return hi->num_domains;
2877 } else
2878 return num_stripes_per_unit_resync(dev);
2879}
2880
9529d343 2881static __u8 imsm_num_data_members(struct imsm_map *map)
1e5c6983
DW		 2882{
2883 /* named 'imsm_' because raid0, raid1 and raid10
2884 * counter-intuitively have the same number of data disks
2885 */
1e5c6983
DW		 2886 switch (get_imsm_raid_level(map)) {
2887 case 0:
36fd8ccc
AK		 2888 return map->num_members;
2889 break;
1e5c6983
DW		 2890 case 1:
2891 case 10:
36fd8ccc 2892 return map->num_members/2;
1e5c6983
DW		 2893 case 5:
2894 return map->num_members - 1;
2895 default:
1ade5cc1 2896 dprintf("unsupported raid level\n");
1e5c6983
DW		 2897 return 0;
2898 }
2899}
2900
44490938
MD		 2901static unsigned long long calc_component_size(struct imsm_map *map,
2902 struct imsm_dev *dev)
2903{
2904 unsigned long long component_size;
2905 unsigned long long dev_size = imsm_dev_size(dev);
a4f7290c 2906 long long calc_dev_size = 0;
44490938
MD		 2907 unsigned int member_disks = imsm_num_data_members(map);
2908
2909 if (member_disks == 0)
2910 return 0;
2911
2912 component_size = per_dev_array_size(map);
2913 calc_dev_size = component_size * member_disks;
2914
2915 /* Component size is rounded to 1MB so difference between size from
2916 * metadata and size calculated from num_data_stripes equals up to
2917 * 2048 blocks per each device. If the difference is higher it means
2918 * that array size was expanded and num_data_stripes was not updated.
2919 */
a4f7290c 2920 if (llabs(calc_dev_size - (long long)dev_size) >
44490938
MD		 2921 (1 << SECT_PER_MB_SHIFT) * member_disks) {
2922 component_size = dev_size / member_disks;
2923 dprintf("Invalid num_data_stripes in metadata; expected=%llu, found=%llu\n",
2924 component_size / map->blocks_per_strip,
2925 num_data_stripes(map));
2926 }
2927
2928 return component_size;
2929}
2930
1e5c6983
DW		 2931static __u32 parity_segment_depth(struct imsm_dev *dev)
2932{
238c0a71 2933 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2934 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2935
2936 switch(get_imsm_raid_level(map)) {
2937 case 1:
2938 case 10:
2939 return chunk * map->num_domains;
2940 case 5:
2941 return chunk * map->num_members;
2942 default:
2943 return chunk;
2944 }
2945}
2946
2947static __u32 map_migr_block(struct imsm_dev *dev, __u32 block)
2948{
238c0a71 2949 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2950 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2951 __u32 strip = block / chunk;
2952
2953 switch (get_imsm_raid_level(map)) {
2954 case 1:
2955 case 10: {
2956 __u32 vol_strip = (strip * map->num_domains) + 1;
2957 __u32 vol_stripe = vol_strip / map->num_members;
2958
2959 return vol_stripe * chunk + block % chunk;
2960 } case 5: {
2961 __u32 stripe = strip / (map->num_members - 1);
2962
2963 return stripe * chunk + block % chunk;
2964 }
2965 default:
2966 return 0;
2967 }
2968}
2969
c47b0ff6
AK		 2970static __u64 blocks_per_migr_unit(struct intel_super *super,
2971 struct imsm_dev *dev)
1e5c6983
DW		 2972{
2973 /* calculate the conversion factor between per member 'blocks'
2974 * (md/{resync,rebuild}_start) and imsm migration units, return
2975 * 0 for the 'not migrating' and 'unsupported migration' cases
2976 */
2977 if (!dev->vol.migr_state)
2978 return 0;
2979
2980 switch (migr_type(dev)) {
c47b0ff6
AK		 2981 case MIGR_GEN_MIGR: {
2982 struct migr_record *migr_rec = super->migr_rec;
2983 return __le32_to_cpu(migr_rec->blocks_per_unit);
2984 }
1e5c6983
DW		 2985 case MIGR_VERIFY:
2986 case MIGR_REPAIR:
2987 case MIGR_INIT: {
238c0a71 2988 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2989 __u32 stripes_per_unit;
2990 __u32 blocks_per_unit;
2991 __u32 parity_depth;
2992 __u32 migr_chunk;
2993 __u32 block_map;
2994 __u32 block_rel;
2995 __u32 segment;
2996 __u32 stripe;
2997 __u8 disks;
2998
2999 /* yes, this is really the translation of migr_units to
3000 * per-member blocks in the 'resync' case
3001 */
3002 stripes_per_unit = num_stripes_per_unit_resync(dev);
3003 migr_chunk = migr_strip_blocks_resync(dev);
9529d343 3004 disks = imsm_num_data_members(map);
1e5c6983 3005 blocks_per_unit = stripes_per_unit * migr_chunk * disks;
7b1ab482 3006 stripe = __le16_to_cpu(map->blocks_per_strip) * disks;
1e5c6983
DW		 3007 segment = blocks_per_unit / stripe;
3008 block_rel = blocks_per_unit - segment * stripe;
3009 parity_depth = parity_segment_depth(dev);
3010 block_map = map_migr_block(dev, block_rel);
3011 return block_map + parity_depth * segment;
3012 }
3013 case MIGR_REBUILD: {
3014 __u32 stripes_per_unit;
3015 __u32 migr_chunk;
3016
3017 stripes_per_unit = num_stripes_per_unit_rebuild(dev);
3018 migr_chunk = migr_strip_blocks_rebuild(dev);
3019 return migr_chunk * stripes_per_unit;
3020 }
1e5c6983
DW		 3021 case MIGR_STATE_CHANGE:
3022 default:
3023 return 0;
3024 }
3025}
3026
c2c087e6
DW		 3027static int imsm_level_to_layout(int level)
3028{
3029 switch (level) {
3030 case 0:
3031 case 1:
3032 return 0;
3033 case 5:
3034 case 6:
a380c027 3035 return ALGORITHM_LEFT_ASYMMETRIC;
c2c087e6 3036 case 10:
c92a2527 3037 return 0x102;
c2c087e6 3038 }
a18a888e 3039 return UnSet;
c2c087e6
DW		 3040}
3041
8e59f3d8
AK		 3042/*******************************************************************************
3043 * Function: read_imsm_migr_rec
3044 * Description: Function reads imsm migration record from last sector of disk
3045 * Parameters:
3046 * fd : disk descriptor
3047 * super : metadata info
3048 * Returns:
3049 * 0 : success,
3050 * -1 : fail
3051 ******************************************************************************/
3052static int read_imsm_migr_rec(int fd, struct intel_super *super)
3053{
3054 int ret_val = -1;
de44e46f 3055 unsigned int sector_size = super->sector_size;
8e59f3d8
AK		 3056 unsigned long long dsize;
3057
3058 get_dev_size(fd, NULL, &dsize);
de44e46f
PB		 3059 if (lseek64(fd, dsize - (sector_size*MIGR_REC_SECTOR_POSITION),
3060 SEEK_SET) < 0) {
e7b84f9d
N		 3061 pr_err("Cannot seek to anchor block: %s\n",
3062 strerror(errno));
8e59f3d8
AK		 3063 goto out;
3064 }
466070ad 3065 if ((unsigned int)read(fd, super->migr_rec_buf,
de44e46f
PB		 3066 MIGR_REC_BUF_SECTORS*sector_size) !=
3067 MIGR_REC_BUF_SECTORS*sector_size) {
e7b84f9d
N		 3068 pr_err("Cannot read migr record block: %s\n",
3069 strerror(errno));
8e59f3d8
AK		 3070 goto out;
3071 }
3072 ret_val = 0;
de44e46f
PB		 3073 if (sector_size == 4096)
3074 convert_from_4k_imsm_migr_rec(super);
8e59f3d8
AK		 3075
3076out:
3077 return ret_val;
3078}
3079
3136abe5
AK		 3080static struct imsm_dev *imsm_get_device_during_migration(
3081 struct intel_super *super)
3082{
3083
3084 struct intel_dev *dv;
3085
3086 for (dv = super->devlist; dv; dv = dv->next) {
3087 if (is_gen_migration(dv->dev))
3088 return dv->dev;
3089 }
3090 return NULL;
3091}
3092
8e59f3d8
AK		 3093/*******************************************************************************
3094 * Function: load_imsm_migr_rec
3095 * Description: Function reads imsm migration record (it is stored at the last
3096 * sector of disk)
3097 * Parameters:
3098 * super : imsm internal array info
8e59f3d8
AK		 3099 * Returns:
3100 * 0 : success
3101 * -1 : fail
4c965cc9 3102 * -2 : no migration in progress
8e59f3d8 3103 ******************************************************************************/
2f86fda3 3104static int load_imsm_migr_rec(struct intel_super *super)
8e59f3d8 3105{
594dc1b8 3106 struct dl *dl;
8e59f3d8
AK		 3107 char nm[30];
3108 int retval = -1;
3109 int fd = -1;
3136abe5 3110 struct imsm_dev *dev;
594dc1b8 3111 struct imsm_map *map;
b4ab44d8 3112 int slot = -1;
2f86fda3 3113 int keep_fd = 1;
3136abe5
AK		 3114
3115 /* find map under migration */
3116 dev = imsm_get_device_during_migration(super);
3117 /* nothing to load,no migration in progress?
3118 */
3119 if (dev == NULL)
4c965cc9 3120 return -2;
8e59f3d8 3121
2f86fda3
MT		 3122 map = get_imsm_map(dev, MAP_0);
3123 if (!map)
3124 return -1;
3136abe5 3125
2f86fda3
MT		 3126 for (dl = super->disks; dl; dl = dl->next) {
3127 /* skip spare and failed disks
3128 */
3129 if (dl->index < 0)
3130 continue;
3131 /* read only from one of the first two slots
3132 */
3133 slot = get_imsm_disk_slot(map, dl->index);
3134 if (slot > 1 || slot < 0)
3135 continue;
3136
3137 if (dl->fd < 0) {
8e59f3d8
AK		 3138 sprintf(nm, "%d:%d", dl->major, dl->minor);
3139 fd = dev_open(nm, O_RDONLY);
2f86fda3
MT		 3140 if (fd >= 0) {
3141 keep_fd = 0;
8e59f3d8 3142 break;
2f86fda3
MT		 3143 }
3144 } else {
3145 fd = dl->fd;
3146 break;
8e59f3d8
AK		 3147 }
3148 }
2f86fda3 3149
8e59f3d8 3150 if (fd < 0)
2f86fda3 3151 return retval;
8e59f3d8 3152 retval = read_imsm_migr_rec(fd, super);
2f86fda3 3153 if (!keep_fd)
8e59f3d8 3154 close(fd);
2f86fda3 3155
8e59f3d8
AK		 3156 return retval;
3157}
3158
c17608ea
AK		 3159/*******************************************************************************
3160 * function: imsm_create_metadata_checkpoint_update
3161 * Description: It creates update for checkpoint change.
3162 * Parameters:
3163 * super : imsm internal array info
3164 * u : pointer to prepared update
3165 * Returns:
3166 * Uptate length.
3167 * If length is equal to 0, input pointer u contains no update
3168 ******************************************************************************/
3169static int imsm_create_metadata_checkpoint_update(
3170 struct intel_super *super,
3171 struct imsm_update_general_migration_checkpoint **u)
3172{
3173
3174 int update_memory_size = 0;
3175
1ade5cc1 3176 dprintf("(enter)\n");
c17608ea
AK		 3177
3178 if (u == NULL)
3179 return 0;
3180 *u = NULL;
3181
3182 /* size of all update data without anchor */
3183 update_memory_size =
3184 sizeof(struct imsm_update_general_migration_checkpoint);
3185
503975b9 3186 *u = xcalloc(1, update_memory_size);
c17608ea 3187 if (*u == NULL) {
1ade5cc1 3188 dprintf("error: cannot get memory\n");
c17608ea
AK		 3189 return 0;
3190 }
3191 (*u)->type = update_general_migration_checkpoint;
9f421827 3192 (*u)->curr_migr_unit = current_migr_unit(super->migr_rec);
83b3de77 3193 dprintf("prepared for %llu\n", (unsigned long long)(*u)->curr_migr_unit);
c17608ea
AK		 3194
3195 return update_memory_size;
3196}
3197
c17608ea
AK		 3198static void imsm_update_metadata_locally(struct supertype *st,
3199 void *buf, int len);
3200
687629c2
AK		 3201/*******************************************************************************
3202 * Function: write_imsm_migr_rec
3203 * Description: Function writes imsm migration record
3204 * (at the last sector of disk)
3205 * Parameters:
3206 * super : imsm internal array info
3207 * Returns:
3208 * 0 : success
3209 * -1 : if fail
3210 ******************************************************************************/
3211static int write_imsm_migr_rec(struct supertype *st)
3212{
3213 struct intel_super *super = st->sb;
de44e46f 3214 unsigned int sector_size = super->sector_size;
687629c2 3215 unsigned long long dsize;
687629c2
AK		 3216 int retval = -1;
3217 struct dl *sd;
c17608ea
AK		 3218 int len;
3219 struct imsm_update_general_migration_checkpoint *u;
3136abe5 3220 struct imsm_dev *dev;
594dc1b8 3221 struct imsm_map *map;
3136abe5
AK		 3222
3223 /* find map under migration */
3224 dev = imsm_get_device_during_migration(super);
3225 /* if no migration, write buffer anyway to clear migr_record
3226 * on disk based on first available device
3227 */
3228 if (dev == NULL)
3229 dev = get_imsm_dev(super, super->current_vol < 0 ? 0 :
3230 super->current_vol);
3231
44bfe6df 3232 map = get_imsm_map(dev, MAP_0);
687629c2 3233
de44e46f
PB		 3234 if (sector_size == 4096)
3235 convert_to_4k_imsm_migr_rec(super);
687629c2 3236 for (sd = super->disks ; sd ; sd = sd->next) {
b4ab44d8 3237 int slot = -1;
3136abe5
AK		 3238
3239 /* skip failed and spare devices */
3240 if (sd->index < 0)
3241 continue;
687629c2 3242 /* write to 2 first slots only */
3136abe5
AK		 3243 if (map)
3244 slot = get_imsm_disk_slot(map, sd->index);
089f9d79 3245 if (map == NULL || slot > 1 || slot < 0)
687629c2 3246 continue;
3136abe5 3247
2f86fda3
MT		 3248 get_dev_size(sd->fd, NULL, &dsize);
3249 if (lseek64(sd->fd, dsize - (MIGR_REC_SECTOR_POSITION *
3250 sector_size),
de44e46f 3251 SEEK_SET) < 0) {
e7b84f9d
N		 3252 pr_err("Cannot seek to anchor block: %s\n",
3253 strerror(errno));
687629c2
AK		 3254 goto out;
3255 }
2f86fda3 3256 if ((unsigned int)write(sd->fd, super->migr_rec_buf,
de44e46f
PB		 3257 MIGR_REC_BUF_SECTORS*sector_size) !=
3258 MIGR_REC_BUF_SECTORS*sector_size) {
e7b84f9d
N		 3259 pr_err("Cannot write migr record block: %s\n",
3260 strerror(errno));
687629c2
AK		 3261 goto out;
3262 }
687629c2 3263 }
de44e46f
PB		 3264 if (sector_size == 4096)
3265 convert_from_4k_imsm_migr_rec(super);
c17608ea
AK		 3266 /* update checkpoint information in metadata */
3267 len = imsm_create_metadata_checkpoint_update(super, &u);
c17608ea
AK		 3268 if (len <= 0) {
3269 dprintf("imsm: Cannot prepare update\n");
3270 goto out;
3271 }
3272 /* update metadata locally */
3273 imsm_update_metadata_locally(st, u, len);
3274 /* and possibly remotely */
3275 if (st->update_tail) {
3276 append_metadata_update(st, u, len);
3277 /* during reshape we do all work inside metadata handler
3278 * manage_reshape(), so metadata update has to be triggered
3279 * insida it
3280 */
3281 flush_metadata_updates(st);
3282 st->update_tail = &st->updates;
3283 } else
3284 free(u);
687629c2
AK		 3285
3286 retval = 0;
3287 out:
687629c2
AK		 3288 return retval;
3289}
3290
e2962bfc
AK		 3291/* spare/missing disks activations are not allowe when
3292 * array/container performs reshape operation, because
3293 * all arrays in container works on the same disks set
3294 */
3295int imsm_reshape_blocks_arrays_changes(struct intel_super *super)
3296{
3297 int rv = 0;
3298 struct intel_dev *i_dev;
3299 struct imsm_dev *dev;
3300
3301 /* check whole container
3302 */
3303 for (i_dev = super->devlist; i_dev; i_dev = i_dev->next) {
3304 dev = i_dev->dev;
3ad25638 3305 if (is_gen_migration(dev)) {
e2962bfc
AK		 3306 /* No repair during any migration in container
3307 */
3308 rv = 1;
3309 break;
3310 }
3311 }
3312 return rv;
3313}
3e684231 3314static unsigned long long imsm_component_size_alignment_check(int level,
c41e00b2 3315 int chunk_size,
f36a9ecd 3316 unsigned int sector_size,
c41e00b2
AK		 3317 unsigned long long component_size)
3318{
3e684231 3319 unsigned int component_size_alignment;
c41e00b2 3320
3e684231 3321 /* check component size alignment
c41e00b2 3322 */
3e684231 3323 component_size_alignment = component_size % (chunk_size/sector_size);
c41e00b2 3324
3e684231 3325 dprintf("(Level: %i, chunk_size = %i, component_size = %llu), component_size_alignment = %u\n",
c41e00b2 3326 level, chunk_size, component_size,
3e684231 3327 component_size_alignment);
c41e00b2 3328
3e684231
MZ		 3329 if (component_size_alignment && (level != 1) && (level != UnSet)) {
3330 dprintf("imsm: reported component size aligned from %llu ",
c41e00b2 3331 component_size);
3e684231 3332 component_size -= component_size_alignment;
1ade5cc1 3333 dprintf_cont("to %llu (%i).\n",
3e684231 3334 component_size, component_size_alignment);
c41e00b2
AK		 3335 }
3336
3337 return component_size;
3338}
e2962bfc 3339
fbc42556
JR		 3340/*******************************************************************************
3341 * Function: get_bitmap_header_sector
3342 * Description: Returns the sector where the bitmap header is placed.
3343 * Parameters:
3344 * st : supertype information
3345 * dev_idx : index of the device with bitmap
3346 *
3347 * Returns:
3348 * The sector where the bitmap header is placed
3349 ******************************************************************************/
3350static unsigned long long get_bitmap_header_sector(struct intel_super *super,
3351 int dev_idx)
3352{
3353 struct imsm_dev *dev = get_imsm_dev(super, dev_idx);
3354 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3355
3356 if (!super->sector_size) {
3357 dprintf("sector size is not set\n");
3358 return 0;
3359 }
3360
3361 return pba_of_lba0(map) + calc_component_size(map, dev) +
3362 (IMSM_BITMAP_HEADER_OFFSET / super->sector_size);
3363}
3364
3365/*******************************************************************************
3366 * Function: get_bitmap_sector
3367 * Description: Returns the sector where the bitmap is placed.
3368 * Parameters:
3369 * st : supertype information
3370 * dev_idx : index of the device with bitmap
3371 *
3372 * Returns:
3373 * The sector where the bitmap is placed
3374 ******************************************************************************/
3375static unsigned long long get_bitmap_sector(struct intel_super *super,
3376 int dev_idx)
3377{
3378 if (!super->sector_size) {
3379 dprintf("sector size is not set\n");
3380 return 0;
3381 }
3382
3383 return get_bitmap_header_sector(super, dev_idx) +
3384 (IMSM_BITMAP_HEADER_SIZE / super->sector_size);
3385}
3386
2432ce9b
AP		 3387static unsigned long long get_ppl_sector(struct intel_super *super, int dev_idx)
3388{
3389 struct imsm_dev *dev = get_imsm_dev(super, dev_idx);
3390 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3391
3392 return pba_of_lba0(map) +
3393 (num_data_stripes(map) * map->blocks_per_strip);
3394}
3395
a5d85af7 3396static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap)
bf5a934a
DW		 3397{
3398 struct intel_super *super = st->sb;
c47b0ff6 3399 struct migr_record *migr_rec = super->migr_rec;
949c47a0 3400 struct imsm_dev *dev = get_imsm_dev(super, super->current_vol);
238c0a71
AK		 3401 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3402 struct imsm_map *prev_map = get_imsm_map(dev, MAP_1);
b335e593 3403 struct imsm_map *map_to_analyse = map;
efb30e7f 3404 struct dl *dl;
a5d85af7 3405 int map_disks = info->array.raid_disks;
bf5a934a 3406
95eeceeb 3407 memset(info, 0, sizeof(*info));
b335e593
AK		 3408 if (prev_map)
3409 map_to_analyse = prev_map;
3410
ca0748fa 3411 dl = super->current_disk;
9894ec0d 3412
bf5a934a 3413 info->container_member = super->current_vol;
cd0430a1 3414 info->array.raid_disks = map->num_members;
b335e593 3415 info->array.level = get_imsm_raid_level(map_to_analyse);
bf5a934a
DW		 3416 info->array.layout = imsm_level_to_layout(info->array.level);
3417 info->array.md_minor = -1;
3418 info->array.ctime = 0;
3419 info->array.utime = 0;
b335e593
AK		 3420 info->array.chunk_size =
3421 __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9;
2432ce9b 3422 info->array.state = !(dev->vol.dirty & RAIDVOL_DIRTY);
fcc2c9da 3423 info->custom_array_size = imsm_dev_size(dev);
3ad25638
AK		 3424 info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb);
3425
3f510843 3426 if (is_gen_migration(dev)) {
3f83228a 3427 info->reshape_active = 1;
b335e593
AK		 3428 info->new_level = get_imsm_raid_level(map);
3429 info->new_layout = imsm_level_to_layout(info->new_level);
3430 info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9;
3f83228a 3431 info->delta_disks = map->num_members - prev_map->num_members;
493f5dd6
N		 3432 if (info->delta_disks) {
3433 /* this needs to be applied to every array
3434 * in the container.
3435 */
81219e70 3436 info->reshape_active = CONTAINER_RESHAPE;
493f5dd6 3437 }
3f83228a
N		 3438 /* We shape information that we give to md might have to be
3439 * modify to cope with md's requirement for reshaping arrays.
3440 * For example, when reshaping a RAID0, md requires it to be
3441 * presented as a degraded RAID4.
3442 * Also if a RAID0 is migrating to a RAID5 we need to specify
3443 * the array as already being RAID5, but the 'before' layout
3444 * is a RAID4-like layout.
3445 */
3446 switch (info->array.level) {
3447 case 0:
3448 switch(info->new_level) {
3449 case 0:
3450 /* conversion is happening as RAID4 */
3451 info->array.level = 4;
3452 info->array.raid_disks += 1;
3453 break;
3454 case 5:
3455 /* conversion is happening as RAID5 */
3456 info->array.level = 5;
3457 info->array.layout = ALGORITHM_PARITY_N;
3f83228a
N		 3458 info->delta_disks -= 1;
3459 break;
3460 default:
3461 /* FIXME error message */
3462 info->array.level = UnSet;
3463 break;
3464 }
3465 break;
3466 }
b335e593
AK		 3467 } else {
3468 info->new_level = UnSet;
3469 info->new_layout = UnSet;
3470 info->new_chunk = info->array.chunk_size;
3f83228a 3471 info->delta_disks = 0;
b335e593 3472 }
ca0748fa 3473
efb30e7f
DW		 3474 if (dl) {
3475 info->disk.major = dl->major;
3476 info->disk.minor = dl->minor;
ca0748fa 3477 info->disk.number = dl->index;
656b6b5a
N		 3478 info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse,
3479 dl->index);
efb30e7f 3480 }
bf5a934a 3481
5551b113 3482 info->data_offset = pba_of_lba0(map_to_analyse);
44490938 3483 info->component_size = calc_component_size(map, dev);
3e684231 3484 info->component_size = imsm_component_size_alignment_check(
c41e00b2
AK		 3485 info->array.level,
3486 info->array.chunk_size,
f36a9ecd 3487 super->sector_size,
c41e00b2 3488 info->component_size);
5e46202e 3489 info->bb.supported = 1;
139dae11 3490
301406c9 3491 memset(info->uuid, 0, sizeof(info->uuid));
921d9e16 3492 info->recovery_start = MaxSector;
bf5a934a 3493
c2462068
PB		 3494 if (info->array.level == 5 &&
3495 (dev->rwh_policy == RWH_DISTRIBUTED ||
3496 dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)) {
2432ce9b
AP		 3497 info->consistency_policy = CONSISTENCY_POLICY_PPL;
3498 info->ppl_sector = get_ppl_sector(super, super->current_vol);
c2462068
PB		 3499 if (dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
3500 info->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9;
3501 else
3502 info->ppl_size = (PPL_HEADER_SIZE + PPL_ENTRY_SPACE)
3503 >> 9;
2432ce9b
AP		 3504 } else if (info->array.level <= 0) {
3505 info->consistency_policy = CONSISTENCY_POLICY_NONE;
3506 } else {
fbc42556
JR		 3507 if (dev->rwh_policy == RWH_BITMAP) {
3508 info->bitmap_offset = get_bitmap_sector(super, super->current_vol);
3509 info->consistency_policy = CONSISTENCY_POLICY_BITMAP;
3510 } else {
3511 info->consistency_policy = CONSISTENCY_POLICY_RESYNC;
3512 }
2432ce9b
AP		 3513 }
3514
d2e6d5d6 3515 info->reshape_progress = 0;
b6796ce1 3516 info->resync_start = MaxSector;
b9172665 3517 if ((map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED ||
2432ce9b 3518 !(info->array.state & 1)) &&
b9172665 3519 imsm_reshape_blocks_arrays_changes(super) == 0) {
301406c9 3520 info->resync_start = 0;
b6796ce1
AK		 3521 }
3522 if (dev->vol.migr_state) {
1e5c6983
DW		 3523 switch (migr_type(dev)) {
3524 case MIGR_REPAIR:
3525 case MIGR_INIT: {
c47b0ff6
AK		 3526 __u64 blocks_per_unit = blocks_per_migr_unit(super,
3527 dev);
4036e7ee 3528 __u64 units = vol_curr_migr_unit(dev);
1e5c6983
DW		 3529
3530 info->resync_start = blocks_per_unit * units;
3531 break;
3532 }
d2e6d5d6 3533 case MIGR_GEN_MIGR: {
c47b0ff6
AK		 3534 __u64 blocks_per_unit = blocks_per_migr_unit(super,
3535 dev);
9f421827 3536 __u64 units = current_migr_unit(migr_rec);
04fa9523 3537 int used_disks;
d2e6d5d6 3538
befb629b
AK		 3539 if (__le32_to_cpu(migr_rec->ascending_migr) &&
3540 (units <
9f421827 3541 (get_num_migr_units(migr_rec)-1)) &&
befb629b
AK		 3542 (super->migr_rec->rec_status ==
3543 __cpu_to_le32(UNIT_SRC_IN_CP_AREA)))
3544 units++;
3545
d2e6d5d6 3546 info->reshape_progress = blocks_per_unit * units;
6289d1e0 3547
7a862a02 3548 dprintf("IMSM: General Migration checkpoint : %llu (%llu) -> read reshape progress : %llu\n",
19986c72
MB		 3549 (unsigned long long)units,
3550 (unsigned long long)blocks_per_unit,
3551 info->reshape_progress);
75156c46 3552
9529d343 3553 used_disks = imsm_num_data_members(prev_map);
75156c46 3554 if (used_disks > 0) {
895ffd99 3555 info->custom_array_size = per_dev_array_size(map) *
75156c46 3556 used_disks;
75156c46 3557 }
d2e6d5d6 3558 }
1e5c6983
DW		 3559 case MIGR_VERIFY:
3560 /* we could emulate the checkpointing of
3561 * 'sync_action=check' migrations, but for now
3562 * we just immediately complete them
3563 */
3564 case MIGR_REBUILD:
3565 /* this is handled by container_content_imsm() */
1e5c6983
DW		 3566 case MIGR_STATE_CHANGE:
3567 /* FIXME handle other migrations */
3568 default:
3569 /* we are not dirty, so... */
3570 info->resync_start = MaxSector;
3571 }
b6796ce1 3572 }
301406c9
DW		 3573
3574 strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN);
3575 info->name[MAX_RAID_SERIAL_LEN] = 0;
bf5a934a 3576
f35f2525
N		 3577 info->array.major_version = -1;
3578 info->array.minor_version = -2;
4dd2df09 3579 sprintf(info->text_version, "/%s/%d", st->container_devnm, info->container_member);
a67dd8cc 3580 info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */
51006d85 3581 uuid_from_super_imsm(st, info->uuid);
a5d85af7
N		 3582
3583 if (dmap) {
3584 int i, j;
3585 for (i=0; i<map_disks; i++) {
3586 dmap[i] = 0;
3587 if (i < info->array.raid_disks) {
3588 struct imsm_disk *dsk;
238c0a71 3589 j = get_imsm_disk_idx(dev, i, MAP_X);
a5d85af7
N		 3590 dsk = get_imsm_disk(super, j);
3591 if (dsk && (dsk->status & CONFIGURED_DISK))
3592 dmap[i] = 1;
3593 }
3594 }
3595 }
81ac8b4d 3596}
bf5a934a 3597
3b451610
AK		 3598static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev,
3599 int failed, int look_in_map);
3600
3601static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev,
3602 int look_in_map);
3603
3604static void manage_second_map(struct intel_super *super, struct imsm_dev *dev)
3605{
3606 if (is_gen_migration(dev)) {
3607 int failed;
3608 __u8 map_state;
3609 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
3610
3611 failed = imsm_count_failed(super, dev, MAP_1);
238c0a71 3612 map_state = imsm_check_degraded(super, dev, failed, MAP_1);
3b451610
AK		 3613 if (map2->map_state != map_state) {
3614 map2->map_state = map_state;
3615 super->updates_pending++;
3616 }
3617 }
3618}
97b4d0e9
DW		 3619
3620static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index)
3621{
3622 struct dl *d;
3623
3624 for (d = super->missing; d; d = d->next)
3625 if (d->index == index)
3626 return &d->disk;
3627 return NULL;
3628}
3629
a5d85af7 3630static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map)
4f5bc454
DW		 3631{
3632 struct intel_super *super = st->sb;
4f5bc454 3633 struct imsm_disk *disk;
a5d85af7 3634 int map_disks = info->array.raid_disks;
ab3cb6b3
N		 3635 int max_enough = -1;
3636 int i;
3637 struct imsm_super *mpb;
4f5bc454 3638
bf5a934a 3639 if (super->current_vol >= 0) {
a5d85af7 3640 getinfo_super_imsm_volume(st, info, map);
bf5a934a
DW		 3641 return;
3642 }
95eeceeb 3643 memset(info, 0, sizeof(*info));
d23fe947
DW		 3644
3645 /* Set raid_disks to zero so that Assemble will always pull in valid
3646 * spares
3647 */
3648 info->array.raid_disks = 0;
cdddbdbc
DW		 3649 info->array.level = LEVEL_CONTAINER;
3650 info->array.layout = 0;
3651 info->array.md_minor = -1;
1011e834 3652 info->array.ctime = 0; /* N/A for imsm */
cdddbdbc
DW		 3653 info->array.utime = 0;
3654 info->array.chunk_size = 0;
3655
3656 info->disk.major = 0;
3657 info->disk.minor = 0;
cdddbdbc 3658 info->disk.raid_disk = -1;
c2c087e6 3659 info->reshape_active = 0;
f35f2525
N		 3660 info->array.major_version = -1;
3661 info->array.minor_version = -2;
c2c087e6 3662 strcpy(info->text_version, "imsm");
a67dd8cc 3663 info->safe_mode_delay = 0;
c2c087e6
DW		 3664 info->disk.number = -1;
3665 info->disk.state = 0;
c5afc314 3666 info->name[0] = 0;
921d9e16 3667 info->recovery_start = MaxSector;
3ad25638 3668 info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb);
5e46202e 3669 info->bb.supported = 1;
c2c087e6 3670
97b4d0e9 3671 /* do we have the all the insync disks that we expect? */
ab3cb6b3 3672 mpb = super->anchor;
b7d81a38 3673 info->events = __le32_to_cpu(mpb->generation_num);
97b4d0e9 3674
ab3cb6b3
N		 3675 for (i = 0; i < mpb->num_raid_devs; i++) {
3676 struct imsm_dev *dev = get_imsm_dev(super, i);
3677 int failed, enough, j, missing = 0;
3678 struct imsm_map *map;
3679 __u8 state;
97b4d0e9 3680
3b451610
AK		 3681 failed = imsm_count_failed(super, dev, MAP_0);
3682 state = imsm_check_degraded(super, dev, failed, MAP_0);
238c0a71 3683 map = get_imsm_map(dev, MAP_0);
ab3cb6b3
N		 3684
3685 /* any newly missing disks?
3686 * (catches single-degraded vs double-degraded)
3687 */
3688 for (j = 0; j < map->num_members; j++) {
238c0a71 3689 __u32 ord = get_imsm_ord_tbl_ent(dev, j, MAP_0);
ab3cb6b3
N		 3690 __u32 idx = ord_to_idx(ord);
3691
20dc76d1
MT		 3692 if (super->disks && super->disks->index == (int)idx)
3693 info->disk.raid_disk = j;
3694
ab3cb6b3
N		 3695 if (!(ord & IMSM_ORD_REBUILD) &&
3696 get_imsm_missing(super, idx)) {
3697 missing = 1;
3698 break;
3699 }
97b4d0e9 3700 }
ab3cb6b3
N		 3701
3702 if (state == IMSM_T_STATE_FAILED)
3703 enough = -1;
3704 else if (state == IMSM_T_STATE_DEGRADED &&
3705 (state != map->map_state || missing))
3706 enough = 0;
3707 else /* we're normal, or already degraded */
3708 enough = 1;
d2bde6d3
AK		 3709 if (is_gen_migration(dev) && missing) {
3710 /* during general migration we need all disks
3711 * that process is running on.
3712 * No new missing disk is allowed.
3713 */
3714 max_enough = -1;
3715 enough = -1;
3716 /* no more checks necessary
3717 */
3718 break;
3719 }
ab3cb6b3
N		 3720 /* in the missing/failed disk case check to see
3721 * if at least one array is runnable
3722 */
3723 max_enough = max(max_enough, enough);
3724 }
1ade5cc1 3725 dprintf("enough: %d\n", max_enough);
ab3cb6b3 3726 info->container_enough = max_enough;
97b4d0e9 3727
4a04ec6c 3728 if (super->disks) {
14e8215b
DW		 3729 __u32 reserved = imsm_reserved_sectors(super, super->disks);
3730
b9f594fe 3731 disk = &super->disks->disk;
5551b113 3732 info->data_offset = total_blocks(&super->disks->disk) - reserved;
14e8215b 3733 info->component_size = reserved;
25ed7e59 3734 info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0;
df474657
DW		 3735 /* we don't change info->disk.raid_disk here because
3736 * this state will be finalized in mdmon after we have
3737 * found the 'most fresh' version of the metadata
3738 */
25ed7e59 3739 info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0;
2432ce9b
AP		 3740 info->disk.state |= (is_spare(disk) || is_journal(disk)) ?
3741 0 : (1 << MD_DISK_SYNC);
cdddbdbc 3742 }
a575e2a7
DW		 3743
3744 /* only call uuid_from_super_imsm when this disk is part of a populated container,
3745 * ->compare_super may have updated the 'num_raid_devs' field for spares
3746 */
3747 if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs)
36ba7d48 3748 uuid_from_super_imsm(st, info->uuid);
22e263f6
AC		 3749 else
3750 memcpy(info->uuid, uuid_zero, sizeof(uuid_zero));
a5d85af7
N		 3751
3752 /* I don't know how to compute 'map' on imsm, so use safe default */
3753 if (map) {
3754 int i;
3755 for (i = 0; i < map_disks; i++)
3756 map[i] = 1;
3757 }
3758
cdddbdbc
DW		 3759}
3760
5c4cd5da
AC		 3761/* allocates memory and fills disk in mdinfo structure
3762 * for each disk in array */
3763struct mdinfo *getinfo_super_disks_imsm(struct supertype *st)
3764{
594dc1b8 3765 struct mdinfo *mddev;
5c4cd5da
AC		 3766 struct intel_super *super = st->sb;
3767 struct imsm_disk *disk;
3768 int count = 0;
3769 struct dl *dl;
3770 if (!super || !super->disks)
3771 return NULL;
3772 dl = super->disks;
503975b9 3773 mddev = xcalloc(1, sizeof(*mddev));
5c4cd5da
AC		 3774 while (dl) {
3775 struct mdinfo *tmp;
3776 disk = &dl->disk;
503975b9 3777 tmp = xcalloc(1, sizeof(*tmp));
5c4cd5da
AC		 3778 if (mddev->devs)
3779 tmp->next = mddev->devs;
3780 mddev->devs = tmp;
3781 tmp->disk.number = count++;
3782 tmp->disk.major = dl->major;
3783 tmp->disk.minor = dl->minor;
3784 tmp->disk.state = is_configured(disk) ?
3785 (1 << MD_DISK_ACTIVE) : 0;
3786 tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0;
3787 tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC);
3788 tmp->disk.raid_disk = -1;
3789 dl = dl->next;
3790 }
3791 return mddev;
3792}
3793
cdddbdbc
DW		 3794static int update_super_imsm(struct supertype *st, struct mdinfo *info,
3795 char *update, char *devname, int verbose,
3796 int uuid_set, char *homehost)
3797{
f352c545
DW		 3798 /* For 'assemble' and 'force' we need to return non-zero if any
3799 * change was made. For others, the return value is ignored.
3800 * Update options are:
3801 * force-one : This device looks a bit old but needs to be included,
3802 * update age info appropriately.
3803 * assemble: clear any 'faulty' flag to allow this device to
3804 * be assembled.
3805 * force-array: Array is degraded but being forced, mark it clean
3806 * if that will be needed to assemble it.
3807 *
3808 * newdev: not used ????
3809 * grow: Array has gained a new device - this is currently for
3810 * linear only
3811 * resync: mark as dirty so a resync will happen.
3812 * name: update the name - preserving the homehost
6e46bf34 3813 * uuid: Change the uuid of the array to match watch is given
f352c545
DW		 3814 *
3815 * Following are not relevant for this imsm:
3816 * sparc2.2 : update from old dodgey metadata
3817 * super-minor: change the preferred_minor number
3818 * summaries: update redundant counters.
f352c545
DW		 3819 * homehost: update the recorded homehost
3820 * _reshape_progress: record new reshape_progress position.
3821 */
6e46bf34
DW		 3822 int rv = 1;
3823 struct intel_super *super = st->sb;
3824 struct imsm_super *mpb;
f352c545 3825
6e46bf34
DW		 3826 /* we can only update container info */
3827 if (!super || super->current_vol >= 0 || !super->anchor)
3828 return 1;
3829
3830 mpb = super->anchor;
3831
81a5b4f5
N		 3832 if (strcmp(update, "uuid") == 0) {
3833 /* We take this to mean that the family_num should be updated.
3834 * However that is much smaller than the uuid so we cannot really
3835 * allow an explicit uuid to be given. And it is hard to reliably
3836 * know if one was.
3837 * So if !uuid_set we know the current uuid is random and just used
3838 * the first 'int' and copy it to the other 3 positions.
3839 * Otherwise we require the 4 'int's to be the same as would be the
3840 * case if we are using a random uuid. So an explicit uuid will be
3841 * accepted as long as all for ints are the same... which shouldn't hurt
6e46bf34 3842 */
81a5b4f5
N		 3843 if (!uuid_set) {
3844 info->uuid[1] = info->uuid[2] = info->uuid[3] = info->uuid[0];
6e46bf34 3845 rv = 0;
81a5b4f5
N		 3846 } else {
3847 if (info->uuid[0] != info->uuid[1] ||
3848 info->uuid[1] != info->uuid[2] ||
3849 info->uuid[2] != info->uuid[3])
3850 rv = -1;
3851 else
3852 rv = 0;
6e46bf34 3853 }
81a5b4f5
N		 3854 if (rv == 0)
3855 mpb->orig_family_num = info->uuid[0];
6e46bf34
DW		 3856 } else if (strcmp(update, "assemble") == 0)
3857 rv = 0;
3858 else
1e2b2765 3859 rv = -1;
f352c545 3860
6e46bf34
DW		 3861 /* successful update? recompute checksum */
3862 if (rv == 0)
3863 mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb));
f352c545
DW		 3864
3865 return rv;
cdddbdbc
DW		 3866}
3867
c2c087e6 3868static size_t disks_to_mpb_size(int disks)
cdddbdbc 3869{
c2c087e6 3870 size_t size;
cdddbdbc 3871
c2c087e6
DW		 3872 size = sizeof(struct imsm_super);
3873 size += (disks - 1) * sizeof(struct imsm_disk);
3874 size += 2 * sizeof(struct imsm_dev);
3875 /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */
3876 size += (4 - 2) * sizeof(struct imsm_map);
3877 /* 4 possible disk_ord_tbl's */
3878 size += 4 * (disks - 1) * sizeof(__u32);
bbab0940
TM		 3879 /* maximum bbm log */
3880 size += sizeof(struct bbm_log);
c2c087e6
DW		 3881
3882 return size;
3883}
3884
387fcd59
N		 3885static __u64 avail_size_imsm(struct supertype *st, __u64 devsize,
3886 unsigned long long data_offset)
c2c087e6
DW		 3887{
3888 if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS))
3889 return 0;
3890
3891 return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS);
cdddbdbc
DW		 3892}
3893
ba2de7ba
DW		 3894static void free_devlist(struct intel_super *super)
3895{
3896 struct intel_dev *dv;
3897
3898 while (super->devlist) {
3899 dv = super->devlist->next;
3900 free(super->devlist->dev);
3901 free(super->devlist);
3902 super->devlist = dv;
3903 }
3904}
3905
3906static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src)
3907{
3908 memcpy(dest, src, sizeof_imsm_dev(src, 0));
3909}
3910
c7b8547c
MT		 3911static int compare_super_imsm(struct supertype *st, struct supertype *tst,
3912 int verbose)
cdddbdbc
DW		 3913{
3914 /*
3915 * return:
3916 * 0 same, or first was empty, and second was copied
3917 * 1 second had wrong number
3918 * 2 wrong uuid
3919 * 3 wrong other info
3920 */
3921 struct intel_super *first = st->sb;
3922 struct intel_super *sec = tst->sb;
3923
5d500228
N		 3924 if (!first) {
3925 st->sb = tst->sb;
3926 tst->sb = NULL;
3927 return 0;
3928 }
8603ea6f
LM		 3929 /* in platform dependent environment test if the disks
3930 * use the same Intel hba
cb8f6859 3931 * If not on Intel hba at all, allow anything.
8603ea6f 3932 */
6b781d33
AP		 3933 if (!check_env("IMSM_NO_PLATFORM") && first->hba && sec->hba) {
3934 if (first->hba->type != sec->hba->type) {
c7b8547c
MT		 3935 if (verbose)
3936 pr_err("HBAs of devices do not match %s != %s\n",
3937 get_sys_dev_type(first->hba->type),
3938 get_sys_dev_type(sec->hba->type));
6b781d33
AP		 3939 return 3;
3940 }
c7b8547c 3941
6b781d33 3942 if (first->orom != sec->orom) {
c7b8547c
MT		 3943 if (verbose)
3944 pr_err("HBAs of devices do not match %s != %s\n",
3945 first->hba->pci_id, sec->hba->pci_id);
8603ea6f
LM		 3946 return 3;
3947 }
c7b8547c 3948
8603ea6f 3949 }
cdddbdbc 3950
d23fe947
DW		 3951 /* if an anchor does not have num_raid_devs set then it is a free
3952 * floating spare
3953 */
3954 if (first->anchor->num_raid_devs > 0 &&
3955 sec->anchor->num_raid_devs > 0) {
a2b97981
DW		 3956 /* Determine if these disks might ever have been
3957 * related. Further disambiguation can only take place
3958 * in load_super_imsm_all
3959 */
3960 __u32 first_family = first->anchor->orig_family_num;
3961 __u32 sec_family = sec->anchor->orig_family_num;
3962
f796af5d
DW		 3963 if (memcmp(first->anchor->sig, sec->anchor->sig,
3964 MAX_SIGNATURE_LENGTH) != 0)
3965 return 3;
3966
a2b97981
DW		 3967 if (first_family == 0)
3968 first_family = first->anchor->family_num;
3969 if (sec_family == 0)
3970 sec_family = sec->anchor->family_num;
3971
3972 if (first_family != sec_family)
d23fe947 3973 return 3;
f796af5d 3974
d23fe947 3975 }
cdddbdbc 3976
3e372e5a
DW		 3977 /* if 'first' is a spare promote it to a populated mpb with sec's
3978 * family number
3979 */
3980 if (first->anchor->num_raid_devs == 0 &&
3981 sec->anchor->num_raid_devs > 0) {
78d30f94 3982 int i;
ba2de7ba
DW		 3983 struct intel_dev *dv;
3984 struct imsm_dev *dev;
78d30f94
DW		 3985
3986 /* we need to copy raid device info from sec if an allocation
3987 * fails here we don't associate the spare
3988 */
3989 for (i = 0; i < sec->anchor->num_raid_devs; i++) {
503975b9
N		 3990 dv = xmalloc(sizeof(*dv));
3991 dev = xmalloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1));
ba2de7ba
DW		 3992 dv->dev = dev;
3993 dv->index = i;
3994 dv->next = first->devlist;
3995 first->devlist = dv;
78d30f94 3996 }
709743c5 3997 if (i < sec->anchor->num_raid_devs) {
ba2de7ba
DW		 3998 /* allocation failure */
3999 free_devlist(first);
e12b3daa 4000 pr_err("imsm: failed to associate spare\n");
ba2de7ba 4001 return 3;
78d30f94 4002 }
3e372e5a 4003 first->anchor->num_raid_devs = sec->anchor->num_raid_devs;
148acb7b 4004 first->anchor->orig_family_num = sec->anchor->orig_family_num;
3e372e5a 4005 first->anchor->family_num = sec->anchor->family_num;
ac6449be 4006 memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH);
709743c5
DW		 4007 for (i = 0; i < sec->anchor->num_raid_devs; i++)
4008 imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i));
3e372e5a
DW		 4009 }
4010
cdddbdbc
DW		 4011 return 0;
4012}
4013
0030e8d6
DW		 4014static void fd2devname(int fd, char *name)
4015{
0030e8d6 4016 char *nm;
0030e8d6 4017
7c798f87
MT		 4018 nm = fd2kname(fd);
4019 if (!nm)
0030e8d6 4020 return;
9587c373 4021
7c798f87 4022 snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm);
0030e8d6
DW		 4023}
4024
21e9380b
AP		 4025static int nvme_get_serial(int fd, void *buf, size_t buf_len)
4026{
fcebeb77 4027 char path[PATH_MAX];
21e9380b
AP		 4028 char *name = fd2kname(fd);
4029
4030 if (!name)
4031 return 1;
4032
4033 if (strncmp(name, "nvme", 4) != 0)
4034 return 1;
4035
fcebeb77
MT		 4036 if (!diskfd_to_devpath(fd, 1, path))
4037 return 1;
21e9380b 4038
fcebeb77 4039 return devpath_to_char(path, "serial", buf, buf_len, 0);
21e9380b
AP		 4040}
4041
cdddbdbc
DW		 4042extern int scsi_get_serial(int fd, void *buf, size_t buf_len);
4043
4044static int imsm_read_serial(int fd, char *devname,
6da53c0e 4045 __u8 *serial, size_t serial_buf_len)
cdddbdbc 4046{
21e9380b 4047 char buf[50];
cdddbdbc 4048 int rv;
6da53c0e 4049 size_t len;
316e2bf4
DW		 4050 char *dest;
4051 char *src;
21e9380b
AP		 4052 unsigned int i;
4053
4054 memset(buf, 0, sizeof(buf));
cdddbdbc 4055
21e9380b 4056 rv = nvme_get_serial(fd, buf, sizeof(buf));
cdddbdbc 4057
21e9380b
AP		 4058 if (rv)
4059 rv = scsi_get_serial(fd, buf, sizeof(buf));
f9ba0ff1 4060
40ebbb9c 4061 if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) {
f9ba0ff1
DW		 4062 memset(serial, 0, MAX_RAID_SERIAL_LEN);
4063 fd2devname(fd, (char *) serial);
0030e8d6
DW		 4064 return 0;
4065 }
4066
cdddbdbc
DW		 4067 if (rv != 0) {
4068 if (devname)
e7b84f9d
N		 4069 pr_err("Failed to retrieve serial for %s\n",
4070 devname);
cdddbdbc
DW		 4071 return rv;
4072 }
4073
316e2bf4
DW		 4074 /* trim all whitespace and non-printable characters and convert
4075 * ':' to ';'
4076 */
21e9380b
AP		 4077 for (i = 0, dest = buf; i < sizeof(buf) && buf[i]; i++) {
4078 src = &buf[i];
316e2bf4
DW		 4079 if (*src > 0x20) {
4080 /* ':' is reserved for use in placeholder serial
4081 * numbers for missing disks
4082 */
4083 if (*src == ':')
4084 *dest++ = ';';
4085 else
4086 *dest++ = *src;
4087 }
4088 }
21e9380b
AP		 4089 len = dest - buf;
4090 dest = buf;
316e2bf4 4091
6da53c0e
BK		 4092 if (len > serial_buf_len) {
4093 /* truncate leading characters */
4094 dest += len - serial_buf_len;
4095 len = serial_buf_len;
316e2bf4 4096 }
5c3db629 4097
6da53c0e 4098 memset(serial, 0, serial_buf_len);
316e2bf4 4099 memcpy(serial, dest, len);
cdddbdbc
DW		 4100
4101 return 0;
4102}
4103
1f24f035
DW		 4104static int serialcmp(__u8 *s1, __u8 *s2)
4105{
4106 return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN);
4107}
4108
4109static void serialcpy(__u8 *dest, __u8 *src)
4110{
4111 strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN);
4112}
4113
54c2c1ea
DW		 4114static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super)
4115{
4116 struct dl *dl;
4117
4118 for (dl = super->disks; dl; dl = dl->next)
4119 if (serialcmp(dl->serial, serial) == 0)
4120 break;
4121
4122 return dl;
4123}
4124
a2b97981
DW		 4125static struct imsm_disk *
4126__serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx)
4127{
4128 int i;
4129
4130 for (i = 0; i < mpb->num_disks; i++) {
4131 struct imsm_disk *disk = __get_imsm_disk(mpb, i);
4132
4133 if (serialcmp(disk->serial, serial) == 0) {
4134 if (idx)
4135 *idx = i;
4136 return disk;
4137 }
4138 }
4139
4140 return NULL;
4141}
4142
cdddbdbc
DW		 4143static int
4144load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd)
4145{
a2b97981 4146 struct imsm_disk *disk;
cdddbdbc
DW		 4147 struct dl *dl;
4148 struct stat stb;
cdddbdbc 4149 int rv;
a2b97981 4150 char name[40];
d23fe947
DW		 4151 __u8 serial[MAX_RAID_SERIAL_LEN];
4152
6da53c0e 4153 rv = imsm_read_serial(fd, devname, serial, MAX_RAID_SERIAL_LEN);
d23fe947
DW		 4154
4155 if (rv != 0)
4156 return 2;
4157
503975b9 4158 dl = xcalloc(1, sizeof(*dl));
cdddbdbc 4159
a2b97981
DW		 4160 fstat(fd, &stb);
4161 dl->major = major(stb.st_rdev);
4162 dl->minor = minor(stb.st_rdev);
4163 dl->next = super->disks;
4164 dl->fd = keep_fd ? fd : -1;
4165 assert(super->disks == NULL);
4166 super->disks = dl;
4167 serialcpy(dl->serial, serial);
4168 dl->index = -2;
4169 dl->e = NULL;
4170 fd2devname(fd, name);
4171 if (devname)
503975b9 4172 dl->devname = xstrdup(devname);
a2b97981 4173 else
503975b9 4174 dl->devname = xstrdup(name);
cdddbdbc 4175
d23fe947 4176 /* look up this disk's index in the current anchor */
a2b97981
DW		 4177 disk = __serial_to_disk(dl->serial, super->anchor, &dl->index);
4178 if (disk) {
4179 dl->disk = *disk;
4180 /* only set index on disks that are a member of a
4181 * populated contianer, i.e. one with raid_devs
4182 */
4183 if (is_failed(&dl->disk))
3f6efecc 4184 dl->index = -2;
2432ce9b 4185 else if (is_spare(&dl->disk) || is_journal(&dl->disk))
a2b97981 4186 dl->index = -1;
3f6efecc
DW		 4187 }
4188
949c47a0
DW		 4189 return 0;
4190}
4191
0c046afd
DW		 4192/* When migrating map0 contains the 'destination' state while map1
4193 * contains the current state. When not migrating map0 contains the
4194 * current state. This routine assumes that map[0].map_state is set to
4195 * the current array state before being called.
4196 *
4197 * Migration is indicated by one of the following states
4198 * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed)
e3bba0e0 4199 * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal
0c046afd 4200 * map1state=unitialized)
1484e727 4201 * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal
0c046afd 4202 * map1state=normal)
e3bba0e0 4203 * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal
0c046afd 4204 * map1state=degraded)
8e59f3d8
AK		 4205 * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal
4206 * map1state=normal)
0c046afd 4207 */
8e59f3d8
AK		 4208static void migrate(struct imsm_dev *dev, struct intel_super *super,
4209 __u8 to_state, int migr_type)
3393c6af 4210{
0c046afd 4211 struct imsm_map *dest;
238c0a71 4212 struct imsm_map *src = get_imsm_map(dev, MAP_0);
3393c6af 4213
0c046afd 4214 dev->vol.migr_state = 1;
1484e727 4215 set_migr_type(dev, migr_type);
4036e7ee 4216 set_vol_curr_migr_unit(dev, 0);
238c0a71 4217 dest = get_imsm_map(dev, MAP_1);
0c046afd 4218
0556e1a2 4219 /* duplicate and then set the target end state in map[0] */
3393c6af 4220 memcpy(dest, src, sizeof_imsm_map(src));
fb12a745 4221 if (migr_type == MIGR_GEN_MIGR) {
0556e1a2
DW		 4222 __u32 ord;
4223 int i;
4224
4225 for (i = 0; i < src->num_members; i++) {
4226 ord = __le32_to_cpu(src->disk_ord_tbl[i]);
4227 set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord));
4228 }
4229 }
4230
8e59f3d8
AK		 4231 if (migr_type == MIGR_GEN_MIGR)
4232 /* Clear migration record */
4233 memset(super->migr_rec, 0, sizeof(struct migr_record));
4234
0c046afd 4235 src->map_state = to_state;
949c47a0 4236}
f8f603f1 4237
809da78e
AK		 4238static void end_migration(struct imsm_dev *dev, struct intel_super *super,
4239 __u8 map_state)
f8f603f1 4240{
238c0a71
AK		 4241 struct imsm_map *map = get_imsm_map(dev, MAP_0);
4242 struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state == 0 ?
4243 MAP_0 : MAP_1);
28bce06f 4244 int i, j;
0556e1a2
DW		 4245
4246 /* merge any IMSM_ORD_REBUILD bits that were not successfully
4247 * completed in the last migration.
4248 *
28bce06f 4249 * FIXME add support for raid-level-migration
0556e1a2 4250 */
089f9d79
JS		 4251 if (map_state != map->map_state && (is_gen_migration(dev) == 0) &&
4252 prev->map_state != IMSM_T_STATE_UNINITIALIZED) {
809da78e
AK		 4253 /* when final map state is other than expected
4254 * merge maps (not for migration)
4255 */
4256 int failed;
4257
4258 for (i = 0; i < prev->num_members; i++)
4259 for (j = 0; j < map->num_members; j++)
4260 /* during online capacity expansion
4261 * disks position can be changed
4262 * if takeover is used
4263 */
4264 if (ord_to_idx(map->disk_ord_tbl[j]) ==
4265 ord_to_idx(prev->disk_ord_tbl[i])) {
4266 map->disk_ord_tbl[j] |=
4267 prev->disk_ord_tbl[i];
4268 break;
4269 }
4270 failed = imsm_count_failed(super, dev, MAP_0);
4271 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
4272 }
f8f603f1
DW		 4273
4274 dev->vol.migr_state = 0;
ea672ee1 4275 set_migr_type(dev, 0);
4036e7ee 4276 set_vol_curr_migr_unit(dev, 0);
f8f603f1
DW		 4277 map->map_state = map_state;
4278}
949c47a0
DW		 4279
4280static int parse_raid_devices(struct intel_super *super)
4281{
4282 int i;
4283 struct imsm_dev *dev_new;
4d7b1503 4284 size_t len, len_migr;
401d313b 4285 size_t max_len = 0;
4d7b1503
DW		 4286 size_t space_needed = 0;
4287 struct imsm_super *mpb = super->anchor;
949c47a0
DW		 4288
4289 for (i = 0; i < super->anchor->num_raid_devs; i++) {
4290 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
ba2de7ba 4291 struct intel_dev *dv;
949c47a0 4292
4d7b1503
DW		 4293 len = sizeof_imsm_dev(dev_iter, 0);
4294 len_migr = sizeof_imsm_dev(dev_iter, 1);
4295 if (len_migr > len)
4296 space_needed += len_migr - len;
ca9de185 4297
503975b9 4298 dv = xmalloc(sizeof(*dv));
401d313b
AK		 4299 if (max_len < len_migr)
4300 max_len = len_migr;
4301 if (max_len > len_migr)
4302 space_needed += max_len - len_migr;
503975b9 4303 dev_new = xmalloc(max_len);
949c47a0 4304 imsm_copy_dev(dev_new, dev_iter);
ba2de7ba
DW		 4305 dv->dev = dev_new;
4306 dv->index = i;
4307 dv->next = super->devlist;
4308 super->devlist = dv;
949c47a0 4309 }
cdddbdbc 4310
4d7b1503
DW		 4311 /* ensure that super->buf is large enough when all raid devices
4312 * are migrating
4313 */
4314 if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) {
4315 void *buf;
4316
f36a9ecd
PB		 4317 len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed,
4318 super->sector_size);
4319 if (posix_memalign(&buf, MAX_SECTOR_SIZE, len) != 0)
4d7b1503
DW		 4320 return 1;
4321
1f45a8ad
DW		 4322 memcpy(buf, super->buf, super->len);
4323 memset(buf + super->len, 0, len - super->len);
4d7b1503
DW		 4324 free(super->buf);
4325 super->buf = buf;
4326 super->len = len;
4327 }
ca9de185 4328
bbab0940
TM		 4329 super->extra_space += space_needed;
4330
cdddbdbc
DW		 4331 return 0;
4332}
4333
e2f41b2c
AK		 4334/*******************************************************************************
4335 * Function: check_mpb_migr_compatibility
4336 * Description: Function checks for unsupported migration features:
4337 * - migration optimization area (pba_of_lba0)
4338 * - descending reshape (ascending_migr)
4339 * Parameters:
4340 * super : imsm metadata information
4341 * Returns:
4342 * 0 : migration is compatible
4343 * -1 : migration is not compatible
4344 ******************************************************************************/
4345int check_mpb_migr_compatibility(struct intel_super *super)
4346{
4347 struct imsm_map *map0, *map1;
4348 struct migr_record *migr_rec = super->migr_rec;
4349 int i;
4350
4351 for (i = 0; i < super->anchor->num_raid_devs; i++) {
4352 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
4353
4354 if (dev_iter &&
4355 dev_iter->vol.migr_state == 1 &&
4356 dev_iter->vol.migr_type == MIGR_GEN_MIGR) {
4357 /* This device is migrating */
238c0a71
AK		 4358 map0 = get_imsm_map(dev_iter, MAP_0);
4359 map1 = get_imsm_map(dev_iter, MAP_1);
5551b113 4360 if (pba_of_lba0(map0) != pba_of_lba0(map1))
e2f41b2c
AK		 4361 /* migration optimization area was used */
4362 return -1;
fc54fe7a
JS		 4363 if (migr_rec->ascending_migr == 0 &&
4364 migr_rec->dest_depth_per_unit > 0)
e2f41b2c
AK		 4365 /* descending reshape not supported yet */
4366 return -1;
4367 }
4368 }
4369 return 0;
4370}
4371
d23fe947 4372static void __free_imsm(struct intel_super *super, int free_disks);
9ca2c81c 4373
cdddbdbc 4374/* load_imsm_mpb - read matrix metadata
f2f5c343 4375 * allocates super->mpb to be freed by free_imsm
cdddbdbc
DW		 4376 */
4377static int load_imsm_mpb(int fd, struct intel_super *super, char *devname)
4378{
4379 unsigned long long dsize;
cdddbdbc 4380 unsigned long long sectors;
f36a9ecd 4381 unsigned int sector_size = super->sector_size;
cdddbdbc 4382 struct stat;
6416d527 4383 struct imsm_super *anchor;
cdddbdbc
DW		 4384 __u32 check_sum;
4385
cdddbdbc 4386 get_dev_size(fd, NULL, &dsize);
f36a9ecd 4387 if (dsize < 2*sector_size) {
64436f06 4388 if (devname)
e7b84f9d
N		 4389 pr_err("%s: device to small for imsm\n",
4390 devname);
64436f06
N		 4391 return 1;
4392 }
cdddbdbc 4393
f36a9ecd 4394 if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0) {
cdddbdbc 4395 if (devname)
e7b84f9d
N		 4396 pr_err("Cannot seek to anchor block on %s: %s\n",
4397 devname, strerror(errno));
cdddbdbc
DW		 4398 return 1;
4399 }
4400
f36a9ecd 4401 if (posix_memalign((void **)&anchor, sector_size, sector_size) != 0) {
ad97895e 4402 if (devname)
7a862a02 4403 pr_err("Failed to allocate imsm anchor buffer on %s\n", devname);
ad97895e
DW		 4404 return 1;
4405 }
466070ad 4406 if ((unsigned int)read(fd, anchor, sector_size) != sector_size) {
cdddbdbc 4407 if (devname)
e7b84f9d
N		 4408 pr_err("Cannot read anchor block on %s: %s\n",
4409 devname, strerror(errno));
6416d527 4410 free(anchor);
cdddbdbc
DW		 4411 return 1;
4412 }
4413
6416d527 4414 if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) {
cdddbdbc 4415 if (devname)
e7b84f9d 4416 pr_err("no IMSM anchor on %s\n", devname);
6416d527 4417 free(anchor);
cdddbdbc
DW		 4418 return 2;
4419 }
4420
d23fe947 4421 __free_imsm(super, 0);
f2f5c343
LM		 4422 /* reload capability and hba */
4423
4424 /* capability and hba must be updated with new super allocation */
d424212e 4425 find_intel_hba_capability(fd, super, devname);
f36a9ecd
PB		 4426 super->len = ROUND_UP(anchor->mpb_size, sector_size);
4427 if (posix_memalign(&super->buf, MAX_SECTOR_SIZE, super->len) != 0) {
cdddbdbc 4428 if (devname)
e7b84f9d
N		 4429 pr_err("unable to allocate %zu byte mpb buffer\n",
4430 super->len);
6416d527 4431 free(anchor);
cdddbdbc
DW		 4432 return 2;
4433 }
f36a9ecd 4434 memcpy(super->buf, anchor, sector_size);
cdddbdbc 4435
f36a9ecd 4436 sectors = mpb_sectors(anchor, sector_size) - 1;
6416d527 4437 free(anchor);
8e59f3d8 4438
85337573
AO		 4439 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
4440 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) {
1ade5cc1 4441 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8
AK		 4442 free(super->buf);
4443 return 2;
4444 }
51d83f5d 4445 super->clean_migration_record_by_mdmon = 0;
8e59f3d8 4446
949c47a0 4447 if (!sectors) {
ecf45690
DW		 4448 check_sum = __gen_imsm_checksum(super->anchor);
4449 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
4450 if (devname)
e7b84f9d
N		 4451 pr_err("IMSM checksum %x != %x on %s\n",
4452 check_sum,
4453 __le32_to_cpu(super->anchor->check_sum),
4454 devname);
ecf45690
DW		 4455 return 2;
4456 }
4457
a2b97981 4458 return 0;
949c47a0 4459 }
cdddbdbc
DW		 4460
4461 /* read the extended mpb */
f36a9ecd 4462 if (lseek64(fd, dsize - (sector_size * (2 + sectors)), SEEK_SET) < 0) {
cdddbdbc 4463 if (devname)
e7b84f9d
N		 4464 pr_err("Cannot seek to extended mpb on %s: %s\n",
4465 devname, strerror(errno));
cdddbdbc
DW		 4466 return 1;
4467 }
4468
f36a9ecd
PB		 4469 if ((unsigned int)read(fd, super->buf + sector_size,
4470 super->len - sector_size) != super->len - sector_size) {
cdddbdbc 4471 if (devname)
e7b84f9d
N		 4472 pr_err("Cannot read extended mpb on %s: %s\n",
4473 devname, strerror(errno));
cdddbdbc
DW		 4474 return 2;
4475 }
4476
949c47a0
DW		 4477 check_sum = __gen_imsm_checksum(super->anchor);
4478 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
cdddbdbc 4479 if (devname)
e7b84f9d
N		 4480 pr_err("IMSM checksum %x != %x on %s\n",
4481 check_sum, __le32_to_cpu(super->anchor->check_sum),
4482 devname);
db575f3b 4483 return 3;
cdddbdbc
DW		 4484 }
4485
a2b97981
DW		 4486 return 0;
4487}
4488
8e59f3d8
AK		 4489static int read_imsm_migr_rec(int fd, struct intel_super *super);
4490
97f81ee2
CA		 4491/* clears hi bits in metadata if MPB_ATTRIB_2TB_DISK not set */
4492static void clear_hi(struct intel_super *super)
4493{
4494 struct imsm_super *mpb = super->anchor;
4495 int i, n;
4496 if (mpb->attributes & MPB_ATTRIB_2TB_DISK)
4497 return;
4498 for (i = 0; i < mpb->num_disks; ++i) {
4499 struct imsm_disk *disk = &mpb->disk[i];
4500 disk->total_blocks_hi = 0;
4501 }
4502 for (i = 0; i < mpb->num_raid_devs; ++i) {
4503 struct imsm_dev *dev = get_imsm_dev(super, i);
4504 if (!dev)
4505 return;
4506 for (n = 0; n < 2; ++n) {
4507 struct imsm_map *map = get_imsm_map(dev, n);
4508 if (!map)
4509 continue;
4510 map->pba_of_lba0_hi = 0;
4511 map->blocks_per_member_hi = 0;
4512 map->num_data_stripes_hi = 0;
4513 }
4514 }
4515}
4516
a2b97981
DW		 4517static int
4518load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd)
4519{
4520 int err;
4521
4522 err = load_imsm_mpb(fd, super, devname);
4523 if (err)
4524 return err;
f36a9ecd
PB		 4525 if (super->sector_size == 4096)
4526 convert_from_4k(super);
a2b97981
DW		 4527 err = load_imsm_disk(fd, super, devname, keep_fd);
4528 if (err)
4529 return err;
4530 err = parse_raid_devices(super);
8d67477f
TM		 4531 if (err)
4532 return err;
4533 err = load_bbm_log(super);
97f81ee2 4534 clear_hi(super);
a2b97981 4535 return err;
cdddbdbc
DW		 4536}
4537
ae6aad82
DW		 4538static void __free_imsm_disk(struct dl *d)
4539{
4540 if (d->fd >= 0)
4541 close(d->fd);
4542 if (d->devname)
4543 free(d->devname);
0dcecb2e
DW		 4544 if (d->e)
4545 free(d->e);
ae6aad82
DW		 4546 free(d);
4547
4548}
1a64be56 4549
cdddbdbc
DW		 4550static void free_imsm_disks(struct intel_super *super)
4551{
47ee5a45 4552 struct dl *d;
cdddbdbc 4553
47ee5a45
DW		 4554 while (super->disks) {
4555 d = super->disks;
cdddbdbc 4556 super->disks = d->next;
ae6aad82 4557 __free_imsm_disk(d);
cdddbdbc 4558 }
cb82edca
AK		 4559 while (super->disk_mgmt_list) {
4560 d = super->disk_mgmt_list;
4561 super->disk_mgmt_list = d->next;
4562 __free_imsm_disk(d);
4563 }
47ee5a45
DW		 4564 while (super->missing) {
4565 d = super->missing;
4566 super->missing = d->next;
4567 __free_imsm_disk(d);
4568 }
4569
cdddbdbc
DW		 4570}
4571
9ca2c81c 4572/* free all the pieces hanging off of a super pointer */
d23fe947 4573static void __free_imsm(struct intel_super *super, int free_disks)
cdddbdbc 4574{
88654014
LM		 4575 struct intel_hba *elem, *next;
4576
9ca2c81c 4577 if (super->buf) {
949c47a0 4578 free(super->buf);
9ca2c81c
DW		 4579 super->buf = NULL;
4580 }
f2f5c343
LM		 4581 /* unlink capability description */
4582 super->orom = NULL;
8e59f3d8
AK		 4583 if (super->migr_rec_buf) {
4584 free(super->migr_rec_buf);
4585 super->migr_rec_buf = NULL;
4586 }
d23fe947
DW		 4587 if (free_disks)
4588 free_imsm_disks(super);
ba2de7ba 4589 free_devlist(super);
88654014
LM		 4590 elem = super->hba;
4591 while (elem) {
4592 if (elem->path)
4593 free((void *)elem->path);
4594 next = elem->next;
4595 free(elem);
4596 elem = next;
88c32bb1 4597 }
8d67477f
TM		 4598 if (super->bbm_log)
4599 free(super->bbm_log);
88654014 4600 super->hba = NULL;
cdddbdbc
DW		 4601}
4602
9ca2c81c
DW		 4603static void free_imsm(struct intel_super *super)
4604{
d23fe947 4605 __free_imsm(super, 1);
928f1424 4606 free(super->bb.entries);
9ca2c81c
DW		 4607 free(super);
4608}
cdddbdbc
DW		 4609
4610static void free_super_imsm(struct supertype *st)
4611{
4612 struct intel_super *super = st->sb;
4613
4614 if (!super)
4615 return;
4616
4617 free_imsm(super);
4618 st->sb = NULL;
4619}
4620
49133e57 4621static struct intel_super *alloc_super(void)
c2c087e6 4622{
503975b9 4623 struct intel_super *super = xcalloc(1, sizeof(*super));
c2c087e6 4624
503975b9
N		 4625 super->current_vol = -1;
4626 super->create_offset = ~((unsigned long long) 0);
928f1424
TM		 4627
4628 super->bb.entries = xmalloc(BBM_LOG_MAX_ENTRIES *
4629 sizeof(struct md_bb_entry));
4630 if (!super->bb.entries) {
4631 free(super);
4632 return NULL;
4633 }
4634
c2c087e6
DW		 4635 return super;
4636}
4637
f0f5a016
LM		 4638/*
4639 * find and allocate hba and OROM/EFI based on valid fd of RAID component device
4640 */
d424212e 4641static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname)
f0f5a016
LM		 4642{
4643 struct sys_dev *hba_name;
4644 int rv = 0;
4645
3a30e28e
MT		 4646 if (fd >= 0 && test_partition(fd)) {
4647 pr_err("imsm: %s is a partition, cannot be used in IMSM\n",
4648 devname);
4649 return 1;
4650 }
089f9d79 4651 if (fd < 0 || check_env("IMSM_NO_PLATFORM")) {
f2f5c343 4652 super->orom = NULL;
f0f5a016
LM		 4653 super->hba = NULL;
4654 return 0;
4655 }
4656 hba_name = find_disk_attached_hba(fd, NULL);
4657 if (!hba_name) {
d424212e 4658 if (devname)
e7b84f9d
N		 4659 pr_err("%s is not attached to Intel(R) RAID controller.\n",
4660 devname);
f0f5a016
LM		 4661 return 1;
4662 }
4663 rv = attach_hba_to_super(super, hba_name);
4664 if (rv == 2) {
d424212e
N		 4665 if (devname) {
4666 struct intel_hba *hba = super->hba;
f0f5a016 4667
60f0f54d
PB		 4668 pr_err("%s is attached to Intel(R) %s %s (%s),\n"
4669 " but the container is assigned to Intel(R) %s %s (",
d424212e 4670 devname,
614902f6 4671 get_sys_dev_type(hba_name->type),
60f0f54d 4672 hba_name->type == SYS_DEV_VMD ? "domain" : "RAID controller",
f0f5a016 4673 hba_name->pci_id ? : "Err!",
60f0f54d
PB		 4674 get_sys_dev_type(super->hba->type),
4675 hba->type == SYS_DEV_VMD ? "domain" : "RAID controller");
f0f5a016 4676
f0f5a016
LM		 4677 while (hba) {
4678 fprintf(stderr, "%s", hba->pci_id ? : "Err!");
4679 if (hba->next)
4680 fprintf(stderr, ", ");
4681 hba = hba->next;
4682 }
6b781d33 4683 fprintf(stderr, ").\n"
cca67208 4684 " Mixing devices attached to different controllers is not allowed.\n");
f0f5a016 4685 }
f0f5a016
LM		 4686 return 2;
4687 }
6b781d33 4688 super->orom = find_imsm_capability(hba_name);
f2f5c343
LM		 4689 if (!super->orom)
4690 return 3;
614902f6 4691
f0f5a016
LM		 4692 return 0;
4693}
4694
47ee5a45
DW		 4695/* find_missing - helper routine for load_super_imsm_all that identifies
4696 * disks that have disappeared from the system. This routine relies on
4697 * the mpb being uptodate, which it is at load time.
4698 */
4699static int find_missing(struct intel_super *super)
4700{
4701 int i;
4702 struct imsm_super *mpb = super->anchor;
4703 struct dl *dl;
4704 struct imsm_disk *disk;
47ee5a45
DW		 4705
4706 for (i = 0; i < mpb->num_disks; i++) {
4707 disk = __get_imsm_disk(mpb, i);
54c2c1ea 4708 dl = serial_to_dl(disk->serial, super);
47ee5a45
DW		 4709 if (dl)
4710 continue;
47ee5a45 4711
503975b9 4712 dl = xmalloc(sizeof(*dl));
47ee5a45
DW		 4713 dl->major = 0;
4714 dl->minor = 0;
4715 dl->fd = -1;
503975b9 4716 dl->devname = xstrdup("missing");
47ee5a45
DW		 4717 dl->index = i;
4718 serialcpy(dl->serial, disk->serial);
4719 dl->disk = *disk;
689c9bf3 4720 dl->e = NULL;
47ee5a45
DW		 4721 dl->next = super->missing;
4722 super->missing = dl;
4723 }
4724
4725 return 0;
4726}
4727
a2b97981
DW		 4728static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list)
4729{
4730 struct intel_disk *idisk = disk_list;
4731
4732 while (idisk) {
4733 if (serialcmp(idisk->disk.serial, serial) == 0)
4734 break;
4735 idisk = idisk->next;
4736 }
4737
4738 return idisk;
4739}
4740
4741static int __prep_thunderdome(struct intel_super **table, int tbl_size,
4742 struct intel_super *super,
4743 struct intel_disk **disk_list)
4744{
4745 struct imsm_disk *d = &super->disks->disk;
4746 struct imsm_super *mpb = super->anchor;
4747 int i, j;
4748
4749 for (i = 0; i < tbl_size; i++) {
4750 struct imsm_super *tbl_mpb = table[i]->anchor;
4751 struct imsm_disk *tbl_d = &table[i]->disks->disk;
4752
4753 if (tbl_mpb->family_num == mpb->family_num) {
4754 if (tbl_mpb->check_sum == mpb->check_sum) {
1ade5cc1
N		 4755 dprintf("mpb from %d:%d matches %d:%d\n",
4756 super->disks->major,
a2b97981
DW		 4757 super->disks->minor,
4758 table[i]->disks->major,
4759 table[i]->disks->minor);
4760 break;
4761 }
4762
4763 if (((is_configured(d) && !is_configured(tbl_d)) ||
4764 is_configured(d) == is_configured(tbl_d)) &&
4765 tbl_mpb->generation_num < mpb->generation_num) {
4766 /* current version of the mpb is a
4767 * better candidate than the one in
4768 * super_table, but copy over "cross
4769 * generational" status
4770 */
4771 struct intel_disk *idisk;
4772
1ade5cc1
N		 4773 dprintf("mpb from %d:%d replaces %d:%d\n",
4774 super->disks->major,
a2b97981
DW		 4775 super->disks->minor,
4776 table[i]->disks->major,
4777 table[i]->disks->minor);
4778
4779 idisk = disk_list_get(tbl_d->serial, *disk_list);
4780 if (idisk && is_failed(&idisk->disk))
4781 tbl_d->status |= FAILED_DISK;
4782 break;
4783 } else {
4784 struct intel_disk *idisk;
4785 struct imsm_disk *disk;
4786
4787 /* tbl_mpb is more up to date, but copy
4788 * over cross generational status before
4789 * returning
4790 */
4791 disk = __serial_to_disk(d->serial, mpb, NULL);
4792 if (disk && is_failed(disk))
4793 d->status |= FAILED_DISK;
4794
4795 idisk = disk_list_get(d->serial, *disk_list);
4796 if (idisk) {
4797 idisk->owner = i;
4798 if (disk && is_configured(disk))
4799 idisk->disk.status |= CONFIGURED_DISK;
4800 }
4801
1ade5cc1
N		 4802 dprintf("mpb from %d:%d prefer %d:%d\n",
4803 super->disks->major,
a2b97981
DW		 4804 super->disks->minor,
4805 table[i]->disks->major,
4806 table[i]->disks->minor);
4807
4808 return tbl_size;
4809 }
4810 }
4811 }
4812
4813 if (i >= tbl_size)
4814 table[tbl_size++] = super;
4815 else
4816 table[i] = super;
4817
4818 /* update/extend the merged list of imsm_disk records */
4819 for (j = 0; j < mpb->num_disks; j++) {
4820 struct imsm_disk *disk = __get_imsm_disk(mpb, j);
4821 struct intel_disk *idisk;
4822
4823 idisk = disk_list_get(disk->serial, *disk_list);
4824 if (idisk) {
4825 idisk->disk.status |= disk->status;
4826 if (is_configured(&idisk->disk) ||
4827 is_failed(&idisk->disk))
4828 idisk->disk.status &= ~(SPARE_DISK);
4829 } else {
503975b9 4830 idisk = xcalloc(1, sizeof(*idisk));
a2b97981
DW		 4831 idisk->owner = IMSM_UNKNOWN_OWNER;
4832 idisk->disk = *disk;
4833 idisk->next = *disk_list;
4834 *disk_list = idisk;
4835 }
4836
4837 if (serialcmp(idisk->disk.serial, d->serial) == 0)
4838 idisk->owner = i;
4839 }
4840
4841 return tbl_size;
4842}
4843
4844static struct intel_super *
4845validate_members(struct intel_super *super, struct intel_disk *disk_list,
4846 const int owner)
4847{
4848 struct imsm_super *mpb = super->anchor;
4849 int ok_count = 0;
4850 int i;
4851
4852 for (i = 0; i < mpb->num_disks; i++) {
4853 struct imsm_disk *disk = __get_imsm_disk(mpb, i);
4854 struct intel_disk *idisk;
4855
4856 idisk = disk_list_get(disk->serial, disk_list);
4857 if (idisk) {
4858 if (idisk->owner == owner ||
4859 idisk->owner == IMSM_UNKNOWN_OWNER)
4860 ok_count++;
4861 else
1ade5cc1
N		 4862 dprintf("'%.16s' owner %d != %d\n",
4863 disk->serial, idisk->owner,
a2b97981
DW		 4864 owner);
4865 } else {
1ade5cc1
N		 4866 dprintf("unknown disk %x [%d]: %.16s\n",
4867 __le32_to_cpu(mpb->family_num), i,
a2b97981
DW		 4868 disk->serial);
4869 break;
4870 }
4871 }
4872
4873 if (ok_count == mpb->num_disks)
4874 return super;
4875 return NULL;
4876}
4877
4878static void show_conflicts(__u32 family_num, struct intel_super *super_list)
4879{
4880 struct intel_super *s;
4881
4882 for (s = super_list; s; s = s->next) {
4883 if (family_num != s->anchor->family_num)
4884 continue;
e12b3daa 4885 pr_err("Conflict, offlining family %#x on '%s'\n",
a2b97981
DW		 4886 __le32_to_cpu(family_num), s->disks->devname);
4887 }
4888}
4889
4890static struct intel_super *
4891imsm_thunderdome(struct intel_super **super_list, int len)
4892{
4893 struct intel_super *super_table[len];
4894 struct intel_disk *disk_list = NULL;
4895 struct intel_super *champion, *spare;
4896 struct intel_super *s, **del;
4897 int tbl_size = 0;
4898 int conflict;
4899 int i;
4900
4901 memset(super_table, 0, sizeof(super_table));
4902 for (s = *super_list; s; s = s->next)
4903 tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list);
4904
4905 for (i = 0; i < tbl_size; i++) {
4906 struct imsm_disk *d;
4907 struct intel_disk *idisk;
4908 struct imsm_super *mpb = super_table[i]->anchor;
4909
4910 s = super_table[i];
4911 d = &s->disks->disk;
4912
4913 /* 'd' must appear in merged disk list for its
4914 * configuration to be valid
4915 */
4916 idisk = disk_list_get(d->serial, disk_list);
4917 if (idisk && idisk->owner == i)
4918 s = validate_members(s, disk_list, i);
4919 else
4920 s = NULL;
4921
4922 if (!s)
1ade5cc1
N		 4923 dprintf("marking family: %#x from %d:%d offline\n",
4924 mpb->family_num,
a2b97981
DW		 4925 super_table[i]->disks->major,
4926 super_table[i]->disks->minor);
4927 super_table[i] = s;
4928 }
4929
4930 /* This is where the mdadm implementation differs from the Windows
4931 * driver which has no strict concept of a container. We can only
4932 * assemble one family from a container, so when returning a prodigal
4933 * array member to this system the code will not be able to disambiguate
4934 * the container contents that should be assembled ("foreign" versus
4935 * "local"). It requires user intervention to set the orig_family_num
4936 * to a new value to establish a new container. The Windows driver in
4937 * this situation fixes up the volume name in place and manages the
4938 * foreign array as an independent entity.
4939 */
4940 s = NULL;
4941 spare = NULL;
4942 conflict = 0;
4943 for (i = 0; i < tbl_size; i++) {
4944 struct intel_super *tbl_ent = super_table[i];
4945 int is_spare = 0;
4946
4947 if (!tbl_ent)
4948 continue;
4949
4950 if (tbl_ent->anchor->num_raid_devs == 0) {
4951 spare = tbl_ent;
4952 is_spare = 1;
4953 }
4954
4955 if (s && !is_spare) {
4956 show_conflicts(tbl_ent->anchor->family_num, *super_list);
4957 conflict++;
4958 } else if (!s && !is_spare)
4959 s = tbl_ent;
4960 }
4961
4962 if (!s)
4963 s = spare;
4964 if (!s) {
4965 champion = NULL;
4966 goto out;
4967 }
4968 champion = s;
4969
4970 if (conflict)
7a862a02 4971 pr_err("Chose family %#x on '%s', assemble conflicts to new container with '--update=uuid'\n",
a2b97981
DW		 4972 __le32_to_cpu(s->anchor->family_num), s->disks->devname);
4973
4974 /* collect all dl's onto 'champion', and update them to
4975 * champion's version of the status
4976 */
4977 for (s = *super_list; s; s = s->next) {
4978 struct imsm_super *mpb = champion->anchor;
4979 struct dl *dl = s->disks;
4980
4981 if (s == champion)
4982 continue;
4983
5d7b407a
CA		 4984 mpb->attributes |= s->anchor->attributes & MPB_ATTRIB_2TB_DISK;
4985
a2b97981
DW		 4986 for (i = 0; i < mpb->num_disks; i++) {
4987 struct imsm_disk *disk;
4988
4989 disk = __serial_to_disk(dl->serial, mpb, &dl->index);
4990 if (disk) {
4991 dl->disk = *disk;
4992 /* only set index on disks that are a member of
4993 * a populated contianer, i.e. one with
4994 * raid_devs
4995 */
4996 if (is_failed(&dl->disk))
4997 dl->index = -2;
4998 else if (is_spare(&dl->disk))
4999 dl->index = -1;
5000 break;
5001 }
5002 }
5003
5004 if (i >= mpb->num_disks) {
5005 struct intel_disk *idisk;
5006
5007 idisk = disk_list_get(dl->serial, disk_list);
ecf408e9 5008 if (idisk && is_spare(&idisk->disk) &&
a2b97981
DW		 5009 !is_failed(&idisk->disk) && !is_configured(&idisk->disk))
5010 dl->index = -1;
5011 else {
5012 dl->index = -2;
5013 continue;
5014 }
5015 }
5016
5017 dl->next = champion->disks;
5018 champion->disks = dl;
5019 s->disks = NULL;
5020 }
5021
5022 /* delete 'champion' from super_list */
5023 for (del = super_list; *del; ) {
5024 if (*del == champion) {
5025 *del = (*del)->next;
5026 break;
5027 } else
5028 del = &(*del)->next;
5029 }
5030 champion->next = NULL;
5031
5032 out:
5033 while (disk_list) {
5034 struct intel_disk *idisk = disk_list;
5035
5036 disk_list = disk_list->next;
5037 free(idisk);
5038 }
5039
5040 return champion;
5041}
5042
9587c373
LM		 5043static int
5044get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd);
4dd2df09 5045static int get_super_block(struct intel_super **super_list, char *devnm, char *devname,
9587c373 5046 int major, int minor, int keep_fd);
ec50f7b6
LM		 5047static int
5048get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list,
5049 int *max, int keep_fd);
5050
cdddbdbc 5051static int load_super_imsm_all(struct supertype *st, int fd, void **sbp,
ec50f7b6
LM		 5052 char *devname, struct md_list *devlist,
5053 int keep_fd)
cdddbdbc 5054{
a2b97981
DW		 5055 struct intel_super *super_list = NULL;
5056 struct intel_super *super = NULL;
a2b97981 5057 int err = 0;
9587c373 5058 int i = 0;
dab4a513 5059
9587c373
LM		 5060 if (fd >= 0)
5061 /* 'fd' is an opened container */
5062 err = get_sra_super_block(fd, &super_list, devname, &i, keep_fd);
5063 else
ec50f7b6
LM		 5064 /* get super block from devlist devices */
5065 err = get_devlist_super_block(devlist, &super_list, &i, keep_fd);
9587c373 5066 if (err)
1602d52c 5067 goto error;
a2b97981
DW		 5068 /* all mpbs enter, maybe one leaves */
5069 super = imsm_thunderdome(&super_list, i);
5070 if (!super) {
5071 err = 1;
5072 goto error;
cdddbdbc
DW		 5073 }
5074
47ee5a45
DW		 5075 if (find_missing(super) != 0) {
5076 free_imsm(super);
a2b97981
DW		 5077 err = 2;
5078 goto error;
47ee5a45 5079 }
8e59f3d8
AK		 5080
5081 /* load migration record */
2f86fda3 5082 err = load_imsm_migr_rec(super);
4c965cc9
AK		 5083 if (err == -1) {
5084 /* migration is in progress,
5085 * but migr_rec cannot be loaded,
5086 */
8e59f3d8
AK		 5087 err = 4;
5088 goto error;
5089 }
e2f41b2c
AK		 5090
5091 /* Check migration compatibility */
089f9d79 5092 if (err == 0 && check_mpb_migr_compatibility(super) != 0) {
e7b84f9d 5093 pr_err("Unsupported migration detected");
e2f41b2c
AK		 5094 if (devname)
5095 fprintf(stderr, " on %s\n", devname);
5096 else
5097 fprintf(stderr, " (IMSM).\n");
5098
5099 err = 5;
5100 goto error;
5101 }
5102
a2b97981
DW		 5103 err = 0;
5104
5105 error:
5106 while (super_list) {
5107 struct intel_super *s = super_list;
5108
5109 super_list = super_list->next;
5110 free_imsm(s);
5111 }
9587c373 5112
a2b97981
DW		 5113 if (err)
5114 return err;
f7e7067b 5115
cdddbdbc 5116 *sbp = super;
9587c373 5117 if (fd >= 0)
4dd2df09 5118 strcpy(st->container_devnm, fd2devnm(fd));
9587c373 5119 else
4dd2df09 5120 st->container_devnm[0] = 0;
a2b97981 5121 if (err == 0 && st->ss == NULL) {
bf5a934a 5122 st->ss = &super_imsm;
cdddbdbc
DW		 5123 st->minor_version = 0;
5124 st->max_devs = IMSM_MAX_DEVICES;
5125 }
cdddbdbc
DW		 5126 return 0;
5127}
2b959fbf 5128
ec50f7b6
LM		 5129static int
5130get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list,
5131 int *max, int keep_fd)
5132{
5133 struct md_list *tmpdev;
5134 int err = 0;
5135 int i = 0;
9587c373 5136
ec50f7b6
LM		 5137 for (i = 0, tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
5138 if (tmpdev->used != 1)
5139 continue;
5140 if (tmpdev->container == 1) {
ca9de185 5141 int lmax = 0;
ec50f7b6
LM		 5142 int fd = dev_open(tmpdev->devname, O_RDONLY|O_EXCL);
5143 if (fd < 0) {
e7b84f9d 5144 pr_err("cannot open device %s: %s\n",
ec50f7b6
LM		 5145 tmpdev->devname, strerror(errno));
5146 err = 8;
5147 goto error;
5148 }
5149 err = get_sra_super_block(fd, super_list,
5150 tmpdev->devname, &lmax,
5151 keep_fd);
5152 i += lmax;
5153 close(fd);
5154 if (err) {
5155 err = 7;
5156 goto error;
5157 }
5158 } else {
5159 int major = major(tmpdev->st_rdev);
5160 int minor = minor(tmpdev->st_rdev);
5161 err = get_super_block(super_list,
4dd2df09 5162 NULL,
ec50f7b6
LM		 5163 tmpdev->devname,
5164 major, minor,
5165 keep_fd);
5166 i++;
5167 if (err) {
5168 err = 6;
5169 goto error;
5170 }
5171 }
5172 }
5173 error:
5174 *max = i;
5175 return err;
5176}
9587c373 5177
4dd2df09 5178static int get_super_block(struct intel_super **super_list, char *devnm, char *devname,
9587c373
LM		 5179 int major, int minor, int keep_fd)
5180{
594dc1b8 5181 struct intel_super *s;
9587c373
LM		 5182 char nm[32];
5183 int dfd = -1;
9587c373
LM		 5184 int err = 0;
5185 int retry;
5186
5187 s = alloc_super();
5188 if (!s) {
5189 err = 1;
5190 goto error;
5191 }
5192
5193 sprintf(nm, "%d:%d", major, minor);
5194 dfd = dev_open(nm, O_RDWR);
5195 if (dfd < 0) {
5196 err = 2;
5197 goto error;
5198 }
5199
aec01630
JS		 5200 if (!get_dev_sector_size(dfd, NULL, &s->sector_size)) {
5201 err = 2;
5202 goto error;
5203 }
cb8f6859 5204 find_intel_hba_capability(dfd, s, devname);
9587c373
LM		 5205 err = load_and_parse_mpb(dfd, s, NULL, keep_fd);
5206
5207 /* retry the load if we might have raced against mdmon */
4dd2df09 5208 if (err == 3 && devnm && mdmon_running(devnm))
9587c373
LM		 5209 for (retry = 0; retry < 3; retry++) {
5210 usleep(3000);
5211 err = load_and_parse_mpb(dfd, s, NULL, keep_fd);
5212 if (err != 3)
5213 break;
5214 }
5215 error:
5216 if (!err) {
5217 s->next = *super_list;
5218 *super_list = s;
5219 } else {
5220 if (s)
8d67477f 5221 free_imsm(s);
36614e95 5222 if (dfd >= 0)
9587c373
LM		 5223 close(dfd);
5224 }
089f9d79 5225 if (dfd >= 0 && !keep_fd)
9587c373
LM		 5226 close(dfd);
5227 return err;
5228
5229}
5230
5231static int
5232get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd)
5233{
5234 struct mdinfo *sra;
4dd2df09 5235 char *devnm;
9587c373
LM		 5236 struct mdinfo *sd;
5237 int err = 0;
5238 int i = 0;
4dd2df09 5239 sra = sysfs_read(fd, NULL, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE);
9587c373
LM		 5240 if (!sra)
5241 return 1;
5242
5243 if (sra->array.major_version != -1 ||
5244 sra->array.minor_version != -2 ||
5245 strcmp(sra->text_version, "imsm") != 0) {
5246 err = 1;
5247 goto error;
5248 }
5249 /* load all mpbs */
4dd2df09 5250 devnm = fd2devnm(fd);
9587c373 5251 for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) {
4dd2df09 5252 if (get_super_block(super_list, devnm, devname,
9587c373
LM		 5253 sd->disk.major, sd->disk.minor, keep_fd) != 0) {
5254 err = 7;
5255 goto error;
5256 }
5257 }
5258 error:
5259 sysfs_free(sra);
5260 *max = i;
5261 return err;
5262}
5263
2b959fbf
N		 5264static int load_container_imsm(struct supertype *st, int fd, char *devname)
5265{
ec50f7b6 5266 return load_super_imsm_all(st, fd, &st->sb, devname, NULL, 1);
2b959fbf 5267}
cdddbdbc
DW		 5268
5269static int load_super_imsm(struct supertype *st, int fd, char *devname)
5270{
5271 struct intel_super *super;
5272 int rv;
8a3544f8 5273 int retry;
cdddbdbc 5274
357ac106 5275 if (test_partition(fd))
691c6ee1
N		 5276 /* IMSM not allowed on partitions */
5277 return 1;
5278
37424f13
DW		 5279 free_super_imsm(st);
5280
49133e57 5281 super = alloc_super();
aec01630
JS		 5282 if (!get_dev_sector_size(fd, NULL, &super->sector_size))
5283 return 1;
8d67477f
TM		 5284 if (!super)
5285 return 1;
ea2bc72b
LM		 5286 /* Load hba and capabilities if they exist.
5287 * But do not preclude loading metadata in case capabilities or hba are
5288 * non-compliant and ignore_hw_compat is set.
5289 */
d424212e 5290 rv = find_intel_hba_capability(fd, super, devname);
f2f5c343 5291 /* no orom/efi or non-intel hba of the disk */
089f9d79 5292 if (rv != 0 && st->ignore_hw_compat == 0) {
f2f5c343 5293 if (devname)
e7b84f9d 5294 pr_err("No OROM/EFI properties for %s\n", devname);
f2f5c343
LM		 5295 free_imsm(super);
5296 return 2;
5297 }
a2b97981 5298 rv = load_and_parse_mpb(fd, super, devname, 0);
cdddbdbc 5299
8a3544f8
AP		 5300 /* retry the load if we might have raced against mdmon */
5301 if (rv == 3) {
f96b1302
AP		 5302 struct mdstat_ent *mdstat = NULL;
5303 char *name = fd2kname(fd);
5304
5305 if (name)
5306 mdstat = mdstat_by_component(name);
8a3544f8
AP		 5307
5308 if (mdstat && mdmon_running(mdstat->devnm) && getpid() != mdmon_pid(mdstat->devnm)) {
5309 for (retry = 0; retry < 3; retry++) {
5310 usleep(3000);
5311 rv = load_and_parse_mpb(fd, super, devname, 0);
5312 if (rv != 3)
5313 break;
5314 }
5315 }
5316
5317 free_mdstat(mdstat);
5318 }
5319
cdddbdbc
DW		 5320 if (rv) {
5321 if (devname)
7a862a02 5322 pr_err("Failed to load all information sections on %s\n", devname);
cdddbdbc
DW		 5323 free_imsm(super);
5324 return rv;
5325 }
5326
5327 st->sb = super;
5328 if (st->ss == NULL) {
5329 st->ss = &super_imsm;
5330 st->minor_version = 0;
5331 st->max_devs = IMSM_MAX_DEVICES;
5332 }
8e59f3d8
AK		 5333
5334 /* load migration record */
2f86fda3 5335 if (load_imsm_migr_rec(super) == 0) {
2e062e82
AK		 5336 /* Check for unsupported migration features */
5337 if (check_mpb_migr_compatibility(super) != 0) {
e7b84f9d 5338 pr_err("Unsupported migration detected");
2e062e82
AK		 5339 if (devname)
5340 fprintf(stderr, " on %s\n", devname);
5341 else
5342 fprintf(stderr, " (IMSM).\n");
5343 return 3;
5344 }
e2f41b2c
AK		 5345 }
5346
cdddbdbc
DW		 5347 return 0;
5348}
5349
ef6ffade
DW		 5350static __u16 info_to_blocks_per_strip(mdu_array_info_t *info)
5351{
5352 if (info->level == 1)
5353 return 128;
5354 return info->chunk_size >> 9;
5355}
5356
5551b113
CA		 5357static unsigned long long info_to_blocks_per_member(mdu_array_info_t *info,
5358 unsigned long long size)
fcfd9599 5359{
4025c288 5360 if (info->level == 1)
5551b113 5361 return size * 2;
4025c288 5362 else
5551b113 5363 return (size * 2) & ~(info_to_blocks_per_strip(info) - 1);
fcfd9599
DW		 5364}
5365
4d1313e9
DW		 5366static void imsm_update_version_info(struct intel_super *super)
5367{
5368 /* update the version and attributes */
5369 struct imsm_super *mpb = super->anchor;
5370 char *version;
5371 struct imsm_dev *dev;
5372 struct imsm_map *map;
5373 int i;
5374
5375 for (i = 0; i < mpb->num_raid_devs; i++) {
5376 dev = get_imsm_dev(super, i);
238c0a71 5377 map = get_imsm_map(dev, MAP_0);
4d1313e9
DW		 5378 if (__le32_to_cpu(dev->size_high) > 0)
5379 mpb->attributes |= MPB_ATTRIB_2TB;
5380
5381 /* FIXME detect when an array spans a port multiplier */
5382 #if 0
5383 mpb->attributes |= MPB_ATTRIB_PM;
5384 #endif
5385
5386 if (mpb->num_raid_devs > 1 ||
5387 mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) {
5388 version = MPB_VERSION_ATTRIBS;
5389 switch (get_imsm_raid_level(map)) {
5390 case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break;
5391 case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break;
5392 case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break;
5393 case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break;
5394 }
5395 } else {
5396 if (map->num_members >= 5)
5397 version = MPB_VERSION_5OR6_DISK_ARRAY;
5398 else if (dev->status == DEV_CLONE_N_GO)
5399 version = MPB_VERSION_CNG;
5400 else if (get_imsm_raid_level(map) == 5)
5401 version = MPB_VERSION_RAID5;
5402 else if (map->num_members >= 3)
5403 version = MPB_VERSION_3OR4_DISK_ARRAY;
5404 else if (get_imsm_raid_level(map) == 1)
5405 version = MPB_VERSION_RAID1;
5406 else
5407 version = MPB_VERSION_RAID0;
5408 }
5409 strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version);
5410 }
5411}
5412
aa534678
DW		 5413static int check_name(struct intel_super *super, char *name, int quiet)
5414{
5415 struct imsm_super *mpb = super->anchor;
5416 char *reason = NULL;
9bd99a90
RS		 5417 char *start = name;
5418 size_t len = strlen(name);
aa534678
DW		 5419 int i;
5420
9bd99a90
RS		 5421 if (len > 0) {
5422 while (isspace(start[len - 1]))
5423 start[--len] = 0;
5424 while (*start && isspace(*start))
5425 ++start, --len;
5426 memmove(name, start, len + 1);
5427 }
5428
5429 if (len > MAX_RAID_SERIAL_LEN)
aa534678 5430 reason = "must be 16 characters or less";
9bd99a90
RS		 5431 else if (len == 0)
5432 reason = "must be a non-empty string";
aa534678
DW		 5433
5434 for (i = 0; i < mpb->num_raid_devs; i++) {
5435 struct imsm_dev *dev = get_imsm_dev(super, i);
5436
5437 if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) {
5438 reason = "already exists";
5439 break;
5440 }
5441 }
5442
5443 if (reason && !quiet)
e7b84f9d 5444 pr_err("imsm volume name %s\n", reason);
aa534678
DW		 5445
5446 return !reason;
5447}
5448
8b353278 5449static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info,
5308f117 5450 struct shape *s, char *name,
83cd1e97
N		 5451 char *homehost, int *uuid,
5452 long long data_offset)
cdddbdbc 5453{
c2c087e6
DW		 5454 /* We are creating a volume inside a pre-existing container.
5455 * so st->sb is already set.
5456 */
5457 struct intel_super *super = st->sb;
f36a9ecd 5458 unsigned int sector_size = super->sector_size;
949c47a0 5459 struct imsm_super *mpb = super->anchor;
ba2de7ba 5460 struct intel_dev *dv;
c2c087e6
DW		 5461 struct imsm_dev *dev;
5462 struct imsm_vol *vol;
5463 struct imsm_map *map;
5464 int idx = mpb->num_raid_devs;
5465 int i;
760365f9 5466 int namelen;
c2c087e6 5467 unsigned long long array_blocks;
2c092cad 5468 size_t size_old, size_new;
5551b113 5469 unsigned long long num_data_stripes;
b53bfba6
TM		 5470 unsigned int data_disks;
5471 unsigned long long size_per_member;
cdddbdbc 5472
88c32bb1 5473 if (super->orom && mpb->num_raid_devs >= super->orom->vpa) {
7a862a02 5474 pr_err("This imsm-container already has the maximum of %d volumes\n", super->orom->vpa);
c2c087e6
DW		 5475 return 0;
5476 }
5477
2c092cad
DW		 5478 /* ensure the mpb is large enough for the new data */
5479 size_old = __le32_to_cpu(mpb->mpb_size);
5480 size_new = disks_to_mpb_size(info->nr_disks);
5481 if (size_new > size_old) {
5482 void *mpb_new;
f36a9ecd 5483 size_t size_round = ROUND_UP(size_new, sector_size);
2c092cad 5484
f36a9ecd 5485 if (posix_memalign(&mpb_new, sector_size, size_round) != 0) {
e7b84f9d 5486 pr_err("could not allocate new mpb\n");
2c092cad
DW		 5487 return 0;
5488 }
85337573
AO		 5489 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
5490 MIGR_REC_BUF_SECTORS*
5491 MAX_SECTOR_SIZE) != 0) {
1ade5cc1 5492 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8
AK		 5493 free(super->buf);
5494 free(super);
ea944c8f 5495 free(mpb_new);
8e59f3d8
AK		 5496 return 0;
5497 }
2c092cad
DW		 5498 memcpy(mpb_new, mpb, size_old);
5499 free(mpb);
5500 mpb = mpb_new;
949c47a0 5501 super->anchor = mpb_new;
2c092cad
DW		 5502 mpb->mpb_size = __cpu_to_le32(size_new);
5503 memset(mpb_new + size_old, 0, size_round - size_old);
bbab0940 5504 super->len = size_round;
2c092cad 5505 }
bf5a934a 5506 super->current_vol = idx;
3960e579
DW		 5507
5508 /* handle 'failed_disks' by either:
5509 * a) create dummy disk entries in the table if this the first
5510 * volume in the array. We add them here as this is the only
5511 * opportunity to add them. add_to_super_imsm_volume()
5512 * handles the non-failed disks and continues incrementing
5513 * mpb->num_disks.
5514 * b) validate that 'failed_disks' matches the current number
5515 * of missing disks if the container is populated
d23fe947 5516 */
3960e579 5517 if (super->current_vol == 0) {
d23fe947 5518 mpb->num_disks = 0;
3960e579
DW		 5519 for (i = 0; i < info->failed_disks; i++) {
5520 struct imsm_disk *disk;
5521
5522 mpb->num_disks++;
5523 disk = __get_imsm_disk(mpb, i);
5524 disk->status = CONFIGURED_DISK | FAILED_DISK;
5525 disk->scsi_id = __cpu_to_le32(~(__u32)0);
5526 snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN,
5b975129 5527 "missing:%d", (__u8)i);
3960e579
DW		 5528 }
5529 find_missing(super);
5530 } else {
5531 int missing = 0;
5532 struct dl *d;
5533
5534 for (d = super->missing; d; d = d->next)
5535 missing++;
5536 if (info->failed_disks > missing) {
e7b84f9d 5537 pr_err("unable to add 'missing' disk to container\n");
3960e579
DW		 5538 return 0;
5539 }
5540 }
5a038140 5541
aa534678
DW		 5542 if (!check_name(super, name, 0))
5543 return 0;
503975b9
N		 5544 dv = xmalloc(sizeof(*dv));
5545 dev = xcalloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1));
760365f9
JS		 5546 /*
5547 * Explicitly allow truncating to not confuse gcc's
5548 * -Werror=stringop-truncation
5549 */
5550 namelen = min((int) strlen(name), MAX_RAID_SERIAL_LEN);
5551 memcpy(dev->volume, name, namelen);
e03640bd 5552 array_blocks = calc_array_size(info->level, info->raid_disks,
03bcbc65 5553 info->layout, info->chunk_size,
b53bfba6
TM		 5554 s->size * BLOCKS_PER_KB);
5555 data_disks = get_data_disks(info->level, info->layout,
5556 info->raid_disks);
5557 array_blocks = round_size_to_mb(array_blocks, data_disks);
5558 size_per_member = array_blocks / data_disks;
979d38be 5559
fcc2c9da 5560 set_imsm_dev_size(dev, array_blocks);
1a2487c2 5561 dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING);
c2c087e6
DW		 5562 vol = &dev->vol;
5563 vol->migr_state = 0;
1484e727 5564 set_migr_type(dev, MIGR_INIT);
3960e579 5565 vol->dirty = !info->state;
4036e7ee 5566 set_vol_curr_migr_unit(dev, 0);
238c0a71 5567 map = get_imsm_map(dev, MAP_0);
5551b113 5568 set_pba_of_lba0(map, super->create_offset);
ef6ffade 5569 map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info));
0556e1a2 5570 map->failed_disk_num = ~0;
bf4442ab 5571 if (info->level > 0)
fffaf1ff
N		 5572 map->map_state = (info->state ? IMSM_T_STATE_NORMAL
5573 : IMSM_T_STATE_UNINITIALIZED);
bf4442ab
AK		 5574 else
5575 map->map_state = info->failed_disks ? IMSM_T_STATE_FAILED :
5576 IMSM_T_STATE_NORMAL;
252d23c0 5577 map->ddf = 1;
ef6ffade
DW		 5578
5579 if (info->level == 1 && info->raid_disks > 2) {
38950822
AW		 5580 free(dev);
5581 free(dv);
7a862a02 5582 pr_err("imsm does not support more than 2 disksin a raid1 volume\n");
ef6ffade
DW		 5583 return 0;
5584 }
81062a36
DW		 5585
5586 map->raid_level = info->level;
4d1313e9 5587 if (info->level == 10) {
c2c087e6 5588 map->raid_level = 1;
4d1313e9 5589 map->num_domains = info->raid_disks / 2;
81062a36
DW		 5590 } else if (info->level == 1)
5591 map->num_domains = info->raid_disks;
5592 else
ff596308 5593 map->num_domains = 1;
81062a36 5594
5551b113 5595 /* info->size is only int so use the 'size' parameter instead */
b53bfba6 5596 num_data_stripes = size_per_member / info_to_blocks_per_strip(info);
5551b113
CA		 5597 num_data_stripes /= map->num_domains;
5598 set_num_data_stripes(map, num_data_stripes);
ef6ffade 5599
44490938
MD		 5600 size_per_member += NUM_BLOCKS_DIRTY_STRIPE_REGION;
5601 set_blocks_per_member(map, info_to_blocks_per_member(info,
5602 size_per_member /
5603 BLOCKS_PER_KB));
5604
c2c087e6
DW		 5605 map->num_members = info->raid_disks;
5606 for (i = 0; i < map->num_members; i++) {
5607 /* initialized in add_to_super */
4eb26970 5608 set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD);
c2c087e6 5609 }
949c47a0 5610 mpb->num_raid_devs++;
2a24dc1b
PB		 5611 mpb->num_raid_devs_created++;
5612 dev->my_vol_raid_dev_num = mpb->num_raid_devs_created;
ba2de7ba 5613
b7580566 5614 if (s->consistency_policy <= CONSISTENCY_POLICY_RESYNC) {
c2462068 5615 dev->rwh_policy = RWH_MULTIPLE_OFF;
2432ce9b 5616 } else if (s->consistency_policy == CONSISTENCY_POLICY_PPL) {
c2462068 5617 dev->rwh_policy = RWH_MULTIPLE_DISTRIBUTED;
2432ce9b
AP		 5618 } else {
5619 free(dev);
5620 free(dv);
5621 pr_err("imsm does not support consistency policy %s\n",
5622 map_num(consistency_policies, s->consistency_policy));
5623 return 0;
5624 }
5625
ba2de7ba
DW		 5626 dv->dev = dev;
5627 dv->index = super->current_vol;
5628 dv->next = super->devlist;
5629 super->devlist = dv;
c2c087e6 5630
4d1313e9
DW		 5631 imsm_update_version_info(super);
5632
c2c087e6 5633 return 1;
cdddbdbc
DW		 5634}
5635
bf5a934a 5636static int init_super_imsm(struct supertype *st, mdu_array_info_t *info,
5308f117 5637 struct shape *s, char *name,
83cd1e97
N		 5638 char *homehost, int *uuid,
5639 unsigned long long data_offset)
bf5a934a
DW		 5640{
5641 /* This is primarily called by Create when creating a new array.
5642 * We will then get add_to_super called for each component, and then
5643 * write_init_super called to write it out to each device.
5644 * For IMSM, Create can create on fresh devices or on a pre-existing
5645 * array.
5646 * To create on a pre-existing array a different method will be called.
5647 * This one is just for fresh drives.
5648 */
5649 struct intel_super *super;
5650 struct imsm_super *mpb;
5651 size_t mpb_size;
4d1313e9 5652 char *version;
bf5a934a 5653
83cd1e97 5654 if (data_offset != INVALID_SECTORS) {
ed503f89 5655 pr_err("data-offset not supported by imsm\n");
83cd1e97
N		 5656 return 0;
5657 }
5658
bf5a934a 5659 if (st->sb)
5308f117 5660 return init_super_imsm_volume(st, info, s, name, homehost, uuid,
83cd1e97 5661 data_offset);
e683ca88
DW		 5662
5663 if (info)
5664 mpb_size = disks_to_mpb_size(info->nr_disks);
5665 else
f36a9ecd 5666 mpb_size = MAX_SECTOR_SIZE;
bf5a934a 5667
49133e57 5668 super = alloc_super();
f36a9ecd
PB		 5669 if (super &&
5670 posix_memalign(&super->buf, MAX_SECTOR_SIZE, mpb_size) != 0) {
8d67477f 5671 free_imsm(super);
e683ca88
DW		 5672 super = NULL;
5673 }
5674 if (!super) {
1ade5cc1 5675 pr_err("could not allocate superblock\n");
bf5a934a
DW		 5676 return 0;
5677 }
de44e46f
PB		 5678 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
5679 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) {
1ade5cc1 5680 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8 5681 free(super->buf);
8d67477f 5682 free_imsm(super);
8e59f3d8
AK		 5683 return 0;
5684 }
e683ca88 5685 memset(super->buf, 0, mpb_size);
ef649044 5686 mpb = super->buf;
e683ca88
DW		 5687 mpb->mpb_size = __cpu_to_le32(mpb_size);
5688 st->sb = super;
5689
5690 if (info == NULL) {
5691 /* zeroing superblock */
5692 return 0;
5693 }
bf5a934a 5694
4d1313e9
DW		 5695 mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
5696
5697 version = (char *) mpb->sig;
5698 strcpy(version, MPB_SIGNATURE);
5699 version += strlen(MPB_SIGNATURE);
5700 strcpy(version, MPB_VERSION_RAID0);
bf5a934a 5701
bf5a934a
DW		 5702 return 1;
5703}
5704
f2cc4f7d
AO		 5705static int drive_validate_sector_size(struct intel_super *super, struct dl *dl)
5706{
5707 unsigned int member_sector_size;
5708
5709 if (dl->fd < 0) {
5710 pr_err("Invalid file descriptor for %s\n", dl->devname);
5711 return 0;
5712 }
5713
5714 if (!get_dev_sector_size(dl->fd, dl->devname, &member_sector_size))
5715 return 0;
5716 if (member_sector_size != super->sector_size)
5717 return 0;
5718 return 1;
5719}
5720
f20c3968 5721static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk,
bf5a934a
DW		 5722 int fd, char *devname)
5723{
5724 struct intel_super *super = st->sb;
d23fe947 5725 struct imsm_super *mpb = super->anchor;
3960e579 5726 struct imsm_disk *_disk;
bf5a934a
DW		 5727 struct imsm_dev *dev;
5728 struct imsm_map *map;
3960e579 5729 struct dl *dl, *df;
4eb26970 5730 int slot;
bf5a934a 5731
949c47a0 5732 dev = get_imsm_dev(super, super->current_vol);
238c0a71 5733 map = get_imsm_map(dev, MAP_0);
bf5a934a 5734
208933a7 5735 if (! (dk->state & (1<<MD_DISK_SYNC))) {
e7b84f9d 5736 pr_err("%s: Cannot add spare devices to IMSM volume\n",
208933a7
N		 5737 devname);
5738 return 1;
5739 }
5740
efb30e7f
DW		 5741 if (fd == -1) {
5742 /* we're doing autolayout so grab the pre-marked (in
5743 * validate_geometry) raid_disk
5744 */
5745 for (dl = super->disks; dl; dl = dl->next)
5746 if (dl->raiddisk == dk->raid_disk)
5747 break;
5748 } else {
5749 for (dl = super->disks; dl ; dl = dl->next)
5750 if (dl->major == dk->major &&
5751 dl->minor == dk->minor)
5752 break;
5753 }
d23fe947 5754
208933a7 5755 if (!dl) {
e7b84f9d 5756 pr_err("%s is not a member of the same container\n", devname);
f20c3968 5757 return 1;
208933a7 5758 }
bf5a934a 5759
59632db9
MZ		 5760 if (mpb->num_disks == 0)
5761 if (!get_dev_sector_size(dl->fd, dl->devname,
5762 &super->sector_size))
5763 return 1;
5764
f2cc4f7d
AO		 5765 if (!drive_validate_sector_size(super, dl)) {
5766 pr_err("Combining drives of different sector size in one volume is not allowed\n");
5767 return 1;
5768 }
5769
d23fe947
DW		 5770 /* add a pristine spare to the metadata */
5771 if (dl->index < 0) {
5772 dl->index = super->anchor->num_disks;
5773 super->anchor->num_disks++;
5774 }
4eb26970
DW		 5775 /* Check the device has not already been added */
5776 slot = get_imsm_disk_slot(map, dl->index);
5777 if (slot >= 0 &&
238c0a71 5778 (get_imsm_ord_tbl_ent(dev, slot, MAP_X) & IMSM_ORD_REBUILD) == 0) {
e7b84f9d 5779 pr_err("%s has been included in this array twice\n",
4eb26970
DW		 5780 devname);
5781 return 1;
5782 }
656b6b5a 5783 set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index);
ee5aad5a 5784 dl->disk.status = CONFIGURED_DISK;
d23fe947 5785
3960e579
DW		 5786 /* update size of 'missing' disks to be at least as large as the
5787 * largest acitve member (we only have dummy missing disks when
5788 * creating the first volume)
5789 */
5790 if (super->current_vol == 0) {
5791 for (df = super->missing; df; df = df->next) {
5551b113
CA		 5792 if (total_blocks(&dl->disk) > total_blocks(&df->disk))
5793 set_total_blocks(&df->disk, total_blocks(&dl->disk));
3960e579
DW		 5794 _disk = __get_imsm_disk(mpb, df->index);
5795 *_disk = df->disk;
5796 }
5797 }
5798
5799 /* refresh unset/failed slots to point to valid 'missing' entries */
5800 for (df = super->missing; df; df = df->next)
5801 for (slot = 0; slot < mpb->num_disks; slot++) {
238c0a71 5802 __u32 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
3960e579
DW		 5803
5804 if ((ord & IMSM_ORD_REBUILD) == 0)
5805 continue;
5806 set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD);
1ace8403 5807 if (is_gen_migration(dev)) {
238c0a71
AK		 5808 struct imsm_map *map2 = get_imsm_map(dev,
5809 MAP_1);
0a108d63 5810 int slot2 = get_imsm_disk_slot(map2, df->index);
089f9d79 5811 if (slot2 < map2->num_members && slot2 >= 0) {
1ace8403 5812 __u32 ord2 = get_imsm_ord_tbl_ent(dev,
238c0a71
AK		 5813 slot2,
5814 MAP_1);
1ace8403
AK		 5815 if ((unsigned)df->index ==
5816 ord_to_idx(ord2))
5817 set_imsm_ord_tbl_ent(map2,
0a108d63 5818 slot2,
1ace8403
AK		 5819 df->index |
5820 IMSM_ORD_REBUILD);
5821 }
5822 }
3960e579
DW		 5823 dprintf("set slot:%d to missing disk:%d\n", slot, df->index);
5824 break;
5825 }
5826
d23fe947
DW		 5827 /* if we are creating the first raid device update the family number */
5828 if (super->current_vol == 0) {
5829 __u32 sum;
5830 struct imsm_dev *_dev = __get_imsm_dev(mpb, 0);
d23fe947 5831
3960e579 5832 _disk = __get_imsm_disk(mpb, dl->index);
791b666a 5833 if (!_dev || !_disk) {
e7b84f9d 5834 pr_err("BUG mpb setup error\n");
791b666a
AW		 5835 return 1;
5836 }
d23fe947
DW		 5837 *_dev = *dev;
5838 *_disk = dl->disk;
148acb7b
DW		 5839 sum = random32();
5840 sum += __gen_imsm_checksum(mpb);
d23fe947 5841 mpb->family_num = __cpu_to_le32(sum);
148acb7b 5842 mpb->orig_family_num = mpb->family_num;
e48aed3c 5843 mpb->creation_time = __cpu_to_le64((__u64)time(NULL));
d23fe947 5844 }
ca0748fa 5845 super->current_disk = dl;
f20c3968 5846 return 0;
bf5a934a
DW		 5847}
5848
a8619d23
AK		 5849/* mark_spare()
5850 * Function marks disk as spare and restores disk serial
5851 * in case it was previously marked as failed by takeover operation
5852 * reruns:
5853 * -1 : critical error
5854 * 0 : disk is marked as spare but serial is not set
5855 * 1 : success
5856 */
5857int mark_spare(struct dl *disk)
5858{
5859 __u8 serial[MAX_RAID_SERIAL_LEN];
5860 int ret_val = -1;
5861
5862 if (!disk)
5863 return ret_val;
5864
5865 ret_val = 0;
6da53c0e 5866 if (!imsm_read_serial(disk->fd, NULL, serial, MAX_RAID_SERIAL_LEN)) {
a8619d23
AK		 5867 /* Restore disk serial number, because takeover marks disk
5868 * as failed and adds to serial ':0' before it becomes
5869 * a spare disk.
5870 */
5871 serialcpy(disk->serial, serial);
5872 serialcpy(disk->disk.serial, serial);
5873 ret_val = 1;
5874 }
5875 disk->disk.status = SPARE_DISK;
5876 disk->index = -1;
5877
5878 return ret_val;
5879}
88654014 5880
12724c01
TM		 5881
5882static int write_super_imsm_spare(struct intel_super *super, struct dl *d);
5883
f20c3968 5884static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk,
72ca9bcf
N		 5885 int fd, char *devname,
5886 unsigned long long data_offset)
cdddbdbc 5887{
c2c087e6 5888 struct intel_super *super = st->sb;
c2c087e6
DW		 5889 struct dl *dd;
5890 unsigned long long size;
fa7bb6f8 5891 unsigned int member_sector_size;
f2f27e63 5892 __u32 id;
c2c087e6
DW		 5893 int rv;
5894 struct stat stb;
5895
88654014
LM		 5896 /* If we are on an RAID enabled platform check that the disk is
5897 * attached to the raid controller.
5898 * We do not need to test disks attachment for container based additions,
5899 * they shall be already tested when container was created/assembled.
88c32bb1 5900 */
d424212e 5901 rv = find_intel_hba_capability(fd, super, devname);
f2f5c343 5902 /* no orom/efi or non-intel hba of the disk */
f0f5a016
LM		 5903 if (rv != 0) {
5904 dprintf("capability: %p fd: %d ret: %d\n",
5905 super->orom, fd, rv);
5906 return 1;
88c32bb1
DW		 5907 }
5908
f20c3968
DW		 5909 if (super->current_vol >= 0)
5910 return add_to_super_imsm_volume(st, dk, fd, devname);
bf5a934a 5911
c2c087e6 5912 fstat(fd, &stb);
503975b9 5913 dd = xcalloc(sizeof(*dd), 1);
c2c087e6
DW		 5914 dd->major = major(stb.st_rdev);
5915 dd->minor = minor(stb.st_rdev);
503975b9 5916 dd->devname = devname ? xstrdup(devname) : NULL;
c2c087e6 5917 dd->fd = fd;
689c9bf3 5918 dd->e = NULL;
1a64be56 5919 dd->action = DISK_ADD;
6da53c0e 5920 rv = imsm_read_serial(fd, devname, dd->serial, MAX_RAID_SERIAL_LEN);
32ba9157 5921 if (rv) {
e7b84f9d 5922 pr_err("failed to retrieve scsi serial, aborting\n");
20bee0f8
PB		 5923 if (dd->devname)
5924 free(dd->devname);
949c47a0 5925 free(dd);
0030e8d6 5926 abort();
c2c087e6 5927 }
7c798f87 5928
20bee0f8
PB		 5929 if (super->hba && ((super->hba->type == SYS_DEV_NVME) ||
5930 (super->hba->type == SYS_DEV_VMD))) {
5931 int i;
7c798f87
MT		 5932 char cntrl_path[PATH_MAX];
5933 char *cntrl_name;
5934 char pci_dev_path[PATH_MAX];
20bee0f8 5935
7c798f87
MT		 5936 if (!diskfd_to_devpath(fd, 2, pci_dev_path) ||
5937 !diskfd_to_devpath(fd, 1, cntrl_path)) {
8662f92d 5938 pr_err("failed to get dev paths, aborting\n");
20bee0f8 5939
a8f3cfd5
MT		 5940 if (dd->devname)
5941 free(dd->devname);
5942 free(dd);
5943 return 1;
5944 }
5945
7c798f87
MT		 5946 cntrl_name = basename(cntrl_path);
5947 if (is_multipath_nvme(fd))
5948 pr_err("%s controller supports Multi-Path I/O, Intel (R) VROC does not support multipathing\n",
5949 cntrl_name);
5950
5951 if (devpath_to_vendor(pci_dev_path) == 0x8086) {
20bee0f8
PB		 5952 /*
5953 * If Intel's NVMe drive has serial ended with
5954 * "-A","-B","-1" or "-2" it means that this is "x8"
5955 * device (double drive on single PCIe card).
5956 * User should be warned about potential data loss.
5957 */
5958 for (i = MAX_RAID_SERIAL_LEN-1; i > 0; i--) {
5959 /* Skip empty character at the end */
5960 if (dd->serial[i] == 0)
5961 continue;
5962
5963 if (((dd->serial[i] == 'A') ||
5964 (dd->serial[i] == 'B') ||
5965 (dd->serial[i] == '1') ||
5966 (dd->serial[i] == '2')) &&
5967 (dd->serial[i-1] == '-'))
5968 pr_err("\tThe action you are about to take may put your data at risk.\n"
5969 "\tPlease note that x8 devices may consist of two separate x4 devices "
5970 "located on a single PCIe port.\n"
5971 "\tRAID 0 is the only supported configuration for this type of x8 device.\n");
5972 break;
5973 }
32716c51
PB		 5974 } else if (super->hba->type == SYS_DEV_VMD && super->orom &&
5975 !imsm_orom_has_tpv_support(super->orom)) {
5976 pr_err("\tPlatform configuration does not support non-Intel NVMe drives.\n"
8b751247 5977 "\tPlease refer to Intel(R) RSTe/VROC user guide.\n");
32716c51
PB		 5978 free(dd->devname);
5979 free(dd);
5980 return 1;
20bee0f8
PB		 5981 }
5982 }
c2c087e6 5983
c2c087e6 5984 get_dev_size(fd, NULL, &size);
aec01630
JS		 5985 if (!get_dev_sector_size(fd, NULL, &member_sector_size))
5986 return 1;
fa7bb6f8
PB		 5987
5988 if (super->sector_size == 0) {
5989 /* this a first device, so sector_size is not set yet */
5990 super->sector_size = member_sector_size;
fa7bb6f8
PB		 5991 }
5992
71e5411e 5993 /* clear migr_rec when adding disk to container */
85337573
AO		 5994 memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
5995 if (lseek64(fd, size - MIGR_REC_SECTOR_POSITION*member_sector_size,
de44e46f 5996 SEEK_SET) >= 0) {
466070ad 5997 if ((unsigned int)write(fd, super->migr_rec_buf,
85337573
AO		 5998 MIGR_REC_BUF_SECTORS*member_sector_size) !=
5999 MIGR_REC_BUF_SECTORS*member_sector_size)
71e5411e
PB		 6000 perror("Write migr_rec failed");
6001 }
6002
c2c087e6 6003 size /= 512;
1f24f035 6004 serialcpy(dd->disk.serial, dd->serial);
5551b113
CA		 6005 set_total_blocks(&dd->disk, size);
6006 if (__le32_to_cpu(dd->disk.total_blocks_hi) > 0) {
6007 struct imsm_super *mpb = super->anchor;
6008 mpb->attributes |= MPB_ATTRIB_2TB_DISK;
6009 }
a8619d23 6010 mark_spare(dd);
c2c087e6 6011 if (sysfs_disk_to_scsi_id(fd, &id) == 0)
b9f594fe 6012 dd->disk.scsi_id = __cpu_to_le32(id);
c2c087e6 6013 else
b9f594fe 6014 dd->disk.scsi_id = __cpu_to_le32(0);
43dad3d6
DW		 6015
6016 if (st->update_tail) {
1a64be56
LM		 6017 dd->next = super->disk_mgmt_list;
6018 super->disk_mgmt_list = dd;
43dad3d6 6019 } else {
12724c01
TM		 6020 /* this is called outside of mdmon
6021 * write initial spare metadata
6022 * mdmon will overwrite it.
6023 */
43dad3d6
DW		 6024 dd->next = super->disks;
6025 super->disks = dd;
12724c01 6026 write_super_imsm_spare(super, dd);
43dad3d6 6027 }
f20c3968
DW		 6028
6029 return 0;
cdddbdbc
DW		 6030}
6031
1a64be56
LM		 6032static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk)
6033{
6034 struct intel_super *super = st->sb;
6035 struct dl *dd;
6036
6037 /* remove from super works only in mdmon - for communication
6038 * manager - monitor. Check if communication memory buffer
6039 * is prepared.
6040 */
6041 if (!st->update_tail) {
1ade5cc1 6042 pr_err("shall be used in mdmon context only\n");
1a64be56
LM		 6043 return 1;
6044 }
503975b9 6045 dd = xcalloc(1, sizeof(*dd));
1a64be56
LM		 6046 dd->major = dk->major;
6047 dd->minor = dk->minor;
1a64be56 6048 dd->fd = -1;
a8619d23 6049 mark_spare(dd);
1a64be56
LM		 6050 dd->action = DISK_REMOVE;
6051
6052 dd->next = super->disk_mgmt_list;
6053 super->disk_mgmt_list = dd;
6054
1a64be56
LM		 6055 return 0;
6056}
6057
f796af5d
DW		 6058static int store_imsm_mpb(int fd, struct imsm_super *mpb);
6059
6060static union {
f36a9ecd 6061 char buf[MAX_SECTOR_SIZE];
f796af5d 6062 struct imsm_super anchor;
f36a9ecd 6063} spare_record __attribute__ ((aligned(MAX_SECTOR_SIZE)));
c2c087e6 6064
12724c01
TM		 6065
6066static int write_super_imsm_spare(struct intel_super *super, struct dl *d)
d23fe947 6067{
d23fe947 6068 struct imsm_super *mpb = super->anchor;
f796af5d 6069 struct imsm_super *spare = &spare_record.anchor;
d23fe947 6070 __u32 sum;
12724c01
TM		 6071
6072 if (d->index != -1)
6073 return 1;
d23fe947 6074
68641cdb
JS		 6075 spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super));
6076 spare->generation_num = __cpu_to_le32(1UL);
f796af5d 6077 spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
68641cdb
JS		 6078 spare->num_disks = 1;
6079 spare->num_raid_devs = 0;
6080 spare->cache_size = mpb->cache_size;
6081 spare->pwr_cycle_count = __cpu_to_le32(1);
f796af5d
DW		 6082
6083 snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH,
6084 MPB_SIGNATURE MPB_VERSION_RAID0);
d23fe947 6085
12724c01
TM		 6086 spare->disk[0] = d->disk;
6087 if (__le32_to_cpu(d->disk.total_blocks_hi) > 0)
6088 spare->attributes |= MPB_ATTRIB_2TB_DISK;
6089
6090 if (super->sector_size == 4096)
6091 convert_to_4k_imsm_disk(&spare->disk[0]);
d23fe947 6092
12724c01
TM		 6093 sum = __gen_imsm_checksum(spare);
6094 spare->family_num = __cpu_to_le32(sum);
6095 spare->orig_family_num = 0;
6096 sum = __gen_imsm_checksum(spare);
6097 spare->check_sum = __cpu_to_le32(sum);
027c374f 6098
12724c01
TM		 6099 if (store_imsm_mpb(d->fd, spare)) {
6100 pr_err("failed for device %d:%d %s\n",
6101 d->major, d->minor, strerror(errno));
6102 return 1;
6103 }
6104
6105 return 0;
6106}
6107/* spare records have their own family number and do not have any defined raid
6108 * devices
6109 */
6110static int write_super_imsm_spares(struct intel_super *super, int doclose)
6111{
6112 struct dl *d;
f36a9ecd 6113
12724c01
TM		 6114 for (d = super->disks; d; d = d->next) {
6115 if (d->index != -1)
6116 continue;
d23fe947 6117
12724c01 6118 if (write_super_imsm_spare(super, d))
e74255d9 6119 return 1;
12724c01 6120
d23fe947
DW		 6121 if (doclose) {
6122 close(d->fd);
6123 d->fd = -1;
6124 }
6125 }
6126
e74255d9 6127 return 0;
d23fe947
DW		 6128}
6129
36988a3d 6130static int write_super_imsm(struct supertype *st, int doclose)
cdddbdbc 6131{
36988a3d 6132 struct intel_super *super = st->sb;
f36a9ecd 6133 unsigned int sector_size = super->sector_size;
949c47a0 6134 struct imsm_super *mpb = super->anchor;
c2c087e6
DW		 6135 struct dl *d;
6136 __u32 generation;
6137 __u32 sum;
d23fe947 6138 int spares = 0;
949c47a0 6139 int i;
a48ac0a8 6140 __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk);
36988a3d 6141 int num_disks = 0;
146c6260 6142 int clear_migration_record = 1;
bbab0940 6143 __u32 bbm_log_size;
cdddbdbc 6144
c2c087e6
DW		 6145 /* 'generation' is incremented everytime the metadata is written */
6146 generation = __le32_to_cpu(mpb->generation_num);
6147 generation++;
6148 mpb->generation_num = __cpu_to_le32(generation);
6149
148acb7b
DW		 6150 /* fix up cases where previous mdadm releases failed to set
6151 * orig_family_num
6152 */
6153 if (mpb->orig_family_num == 0)
6154 mpb->orig_family_num = mpb->family_num;
6155
d23fe947 6156 for (d = super->disks; d; d = d->next) {
8796fdc4 6157 if (d->index == -1)
d23fe947 6158 spares++;
36988a3d 6159 else {
d23fe947 6160 mpb->disk[d->index] = d->disk;
36988a3d
AK		 6161 num_disks++;
6162 }
d23fe947 6163 }
36988a3d 6164 for (d = super->missing; d; d = d->next) {
47ee5a45 6165 mpb->disk[d->index] = d->disk;
36988a3d
AK		 6166 num_disks++;
6167 }
6168 mpb->num_disks = num_disks;
6169 mpb_size += sizeof(struct imsm_disk) * mpb->num_disks;
b9f594fe 6170
949c47a0
DW		 6171 for (i = 0; i < mpb->num_raid_devs; i++) {
6172 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
36988a3d
AK		 6173 struct imsm_dev *dev2 = get_imsm_dev(super, i);
6174 if (dev && dev2) {
6175 imsm_copy_dev(dev, dev2);
6176 mpb_size += sizeof_imsm_dev(dev, 0);
6177 }
146c6260
AK		 6178 if (is_gen_migration(dev2))
6179 clear_migration_record = 0;
949c47a0 6180 }
bbab0940
TM		 6181
6182 bbm_log_size = get_imsm_bbm_log_size(super->bbm_log);
6183
6184 if (bbm_log_size) {
6185 memcpy((void *)mpb + mpb_size, super->bbm_log, bbm_log_size);
6186 mpb->attributes |= MPB_ATTRIB_BBM;
6187 } else
6188 mpb->attributes &= ~MPB_ATTRIB_BBM;
6189
6190 super->anchor->bbm_log_size = __cpu_to_le32(bbm_log_size);
6191 mpb_size += bbm_log_size;
a48ac0a8 6192 mpb->mpb_size = __cpu_to_le32(mpb_size);
949c47a0 6193
bbab0940
TM		 6194#ifdef DEBUG
6195 assert(super->len == 0 || mpb_size <= super->len);
6196#endif
6197
c2c087e6 6198 /* recalculate checksum */
949c47a0 6199 sum = __gen_imsm_checksum(mpb);
c2c087e6
DW		 6200 mpb->check_sum = __cpu_to_le32(sum);
6201
51d83f5d
AK		 6202 if (super->clean_migration_record_by_mdmon) {
6203 clear_migration_record = 1;
6204 super->clean_migration_record_by_mdmon = 0;
6205 }
146c6260 6206 if (clear_migration_record)
de44e46f 6207 memset(super->migr_rec_buf, 0,
85337573 6208 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
146c6260 6209
f36a9ecd
PB		 6210 if (sector_size == 4096)
6211 convert_to_4k(super);
6212
d23fe947 6213 /* write the mpb for disks that compose raid devices */
c2c087e6 6214 for (d = super->disks; d ; d = d->next) {
86c54047 6215 if (d->index < 0 || is_failed(&d->disk))
d23fe947 6216 continue;
30602f53 6217
146c6260
AK		 6218 if (clear_migration_record) {
6219 unsigned long long dsize;
6220
6221 get_dev_size(d->fd, NULL, &dsize);
de44e46f
PB		 6222 if (lseek64(d->fd, dsize - sector_size,
6223 SEEK_SET) >= 0) {
466070ad
PB		 6224 if ((unsigned int)write(d->fd,
6225 super->migr_rec_buf,
de44e46f
PB		 6226 MIGR_REC_BUF_SECTORS*sector_size) !=
6227 MIGR_REC_BUF_SECTORS*sector_size)
9e2d750d 6228 perror("Write migr_rec failed");
146c6260
AK		 6229 }
6230 }
51d83f5d
AK		 6231
6232 if (store_imsm_mpb(d->fd, mpb))
6233 fprintf(stderr,
1ade5cc1
N		 6234 "failed for device %d:%d (fd: %d)%s\n",
6235 d->major, d->minor,
51d83f5d
AK		 6236 d->fd, strerror(errno));
6237
c2c087e6
DW		 6238 if (doclose) {
6239 close(d->fd);
6240 d->fd = -1;
6241 }
6242 }
6243
d23fe947
DW		 6244 if (spares)
6245 return write_super_imsm_spares(super, doclose);
6246
e74255d9 6247 return 0;
c2c087e6
DW		 6248}
6249
9b1fb677 6250static int create_array(struct supertype *st, int dev_idx)
43dad3d6
DW		 6251{
6252 size_t len;
6253 struct imsm_update_create_array *u;
6254 struct intel_super *super = st->sb;
9b1fb677 6255 struct imsm_dev *dev = get_imsm_dev(super, dev_idx);
238c0a71 6256 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW		 6257 struct disk_info *inf;
6258 struct imsm_disk *disk;
6259 int i;
43dad3d6 6260
54c2c1ea
DW		 6261 len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) +
6262 sizeof(*inf) * map->num_members;
503975b9 6263 u = xmalloc(len);
43dad3d6 6264 u->type = update_create_array;
9b1fb677 6265 u->dev_idx = dev_idx;
43dad3d6 6266 imsm_copy_dev(&u->dev, dev);
54c2c1ea
DW		 6267 inf = get_disk_info(u);
6268 for (i = 0; i < map->num_members; i++) {
238c0a71 6269 int idx = get_imsm_disk_idx(dev, i, MAP_X);
9b1fb677 6270
54c2c1ea 6271 disk = get_imsm_disk(super, idx);
1ca5c8e0
N		 6272 if (!disk)
6273 disk = get_imsm_missing(super, idx);
54c2c1ea
DW		 6274 serialcpy(inf[i].serial, disk->serial);
6275 }
43dad3d6
DW		 6276 append_metadata_update(st, u, len);
6277
6278 return 0;
6279}
6280
1a64be56 6281static int mgmt_disk(struct supertype *st)
43dad3d6
DW		 6282{
6283 struct intel_super *super = st->sb;
6284 size_t len;
1a64be56 6285 struct imsm_update_add_remove_disk *u;
43dad3d6 6286
1a64be56 6287 if (!super->disk_mgmt_list)
43dad3d6
DW		 6288 return 0;
6289
6290 len = sizeof(*u);
503975b9 6291 u = xmalloc(len);
1a64be56 6292 u->type = update_add_remove_disk;
43dad3d6
DW		 6293 append_metadata_update(st, u, len);
6294
6295 return 0;
6296}
2432ce9b
AP		 6297
6298__u32 crc32c_le(__u32 crc, unsigned char const *p, size_t len);
6299
e397cefe
AP		 6300static int write_ppl_header(unsigned long long ppl_sector, int fd, void *buf)
6301{
6302 struct ppl_header *ppl_hdr = buf;
6303 int ret;
6304
6305 ppl_hdr->checksum = __cpu_to_le32(~crc32c_le(~0, buf, PPL_HEADER_SIZE));
6306
6307 if (lseek64(fd, ppl_sector * 512, SEEK_SET) < 0) {
6308 ret = -errno;
6309 perror("Failed to seek to PPL header location");
6310 return ret;
6311 }
6312
6313 if (write(fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) {
6314 ret = -errno;
6315 perror("Write PPL header failed");
6316 return ret;
6317 }
6318
6319 fsync(fd);
6320
6321 return 0;
6322}
6323
2432ce9b
AP		 6324static int write_init_ppl_imsm(struct supertype *st, struct mdinfo *info, int fd)
6325{
6326 struct intel_super *super = st->sb;
6327 void *buf;
6328 struct ppl_header *ppl_hdr;
6329 int ret;
6330
b2514242
PB		 6331 /* first clear entire ppl space */
6332 ret = zero_disk_range(fd, info->ppl_sector, info->ppl_size);
6333 if (ret)
6334 return ret;
6335
6336 ret = posix_memalign(&buf, MAX_SECTOR_SIZE, PPL_HEADER_SIZE);
2432ce9b
AP		 6337 if (ret) {
6338 pr_err("Failed to allocate PPL header buffer\n");
e397cefe 6339 return -ret;
2432ce9b
AP		 6340 }
6341
6342 memset(buf, 0, PPL_HEADER_SIZE);
6343 ppl_hdr = buf;
6344 memset(ppl_hdr->reserved, 0xff, PPL_HDR_RESERVED);
6345 ppl_hdr->signature = __cpu_to_le32(super->anchor->orig_family_num);
b23d0750
AP		 6346
6347 if (info->mismatch_cnt) {
6348 /*
6349 * We are overwriting an invalid ppl. Make one entry with wrong
6350 * checksum to prevent the kernel from skipping resync.
6351 */
6352 ppl_hdr->entries_count = __cpu_to_le32(1);
6353 ppl_hdr->entries[0].checksum = ~0;
6354 }
6355
e397cefe 6356 ret = write_ppl_header(info->ppl_sector, fd, buf);
2432ce9b
AP		 6357
6358 free(buf);
6359 return ret;
6360}
6361
e397cefe
AP		 6362static int is_rebuilding(struct imsm_dev *dev);
6363
2432ce9b
AP		 6364static int validate_ppl_imsm(struct supertype *st, struct mdinfo *info,
6365 struct mdinfo *disk)
6366{
6367 struct intel_super *super = st->sb;
6368 struct dl *d;
e397cefe 6369 void *buf_orig, *buf, *buf_prev = NULL;
2432ce9b 6370 int ret = 0;
e397cefe 6371 struct ppl_header *ppl_hdr = NULL;
2432ce9b
AP		 6372 __u32 crc;
6373 struct imsm_dev *dev;
2432ce9b 6374 __u32 idx;
44b6b876
PB		 6375 unsigned int i;
6376 unsigned long long ppl_offset = 0;
6377 unsigned long long prev_gen_num = 0;
2432ce9b
AP		 6378
6379 if (disk->disk.raid_disk < 0)
6380 return 0;
6381
2432ce9b 6382 dev = get_imsm_dev(super, info->container_member);
2fc0fc63 6383 idx = get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_0);
2432ce9b
AP		 6384 d = get_imsm_dl_disk(super, idx);
6385
6386 if (!d || d->index < 0 || is_failed(&d->disk))
e397cefe
AP		 6387 return 0;
6388
6389 if (posix_memalign(&buf_orig, MAX_SECTOR_SIZE, PPL_HEADER_SIZE * 2)) {
6390 pr_err("Failed to allocate PPL header buffer\n");
6391 return -1;
6392 }
6393 buf = buf_orig;
2432ce9b 6394
44b6b876
PB		 6395 ret = 1;
6396 while (ppl_offset < MULTIPLE_PPL_AREA_SIZE_IMSM) {
e397cefe
AP		 6397 void *tmp;
6398
44b6b876 6399 dprintf("Checking potential PPL at offset: %llu\n", ppl_offset);
2432ce9b 6400
44b6b876
PB		 6401 if (lseek64(d->fd, info->ppl_sector * 512 + ppl_offset,
6402 SEEK_SET) < 0) {
6403 perror("Failed to seek to PPL header location");
6404 ret = -1;
e397cefe 6405 break;
44b6b876 6406 }
2432ce9b 6407
44b6b876
PB		 6408 if (read(d->fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) {
6409 perror("Read PPL header failed");
6410 ret = -1;
e397cefe 6411 break;
44b6b876 6412 }
2432ce9b 6413
44b6b876 6414 ppl_hdr = buf;
2432ce9b 6415
44b6b876
PB		 6416 crc = __le32_to_cpu(ppl_hdr->checksum);
6417 ppl_hdr->checksum = 0;
6418
6419 if (crc != ~crc32c_le(~0, buf, PPL_HEADER_SIZE)) {
6420 dprintf("Wrong PPL header checksum on %s\n",
6421 d->devname);
e397cefe 6422 break;
44b6b876
PB		 6423 }
6424
6425 if (prev_gen_num > __le64_to_cpu(ppl_hdr->generation)) {
6426 /* previous was newest, it was already checked */
e397cefe 6427 break;
44b6b876
PB		 6428 }
6429
6430 if ((__le32_to_cpu(ppl_hdr->signature) !=
6431 super->anchor->orig_family_num)) {
6432 dprintf("Wrong PPL header signature on %s\n",
6433 d->devname);
6434 ret = 1;
e397cefe 6435 break;
44b6b876
PB		 6436 }
6437
6438 ret = 0;
6439 prev_gen_num = __le64_to_cpu(ppl_hdr->generation);
2432ce9b 6440
44b6b876
PB		 6441 ppl_offset += PPL_HEADER_SIZE;
6442 for (i = 0; i < __le32_to_cpu(ppl_hdr->entries_count); i++)
6443 ppl_offset +=
6444 __le32_to_cpu(ppl_hdr->entries[i].pp_size);
e397cefe
AP		 6445
6446 if (!buf_prev)
6447 buf_prev = buf + PPL_HEADER_SIZE;
6448 tmp = buf_prev;
6449 buf_prev = buf;
6450 buf = tmp;
2432ce9b
AP		 6451 }
6452
e397cefe
AP		 6453 if (buf_prev) {
6454 buf = buf_prev;
6455 ppl_hdr = buf_prev;
6456 }
2432ce9b 6457
54148aba
PB		 6458 /*
6459 * Update metadata to use mutliple PPLs area (1MB).
6460 * This is done once for all RAID members
6461 */
6462 if (info->consistency_policy == CONSISTENCY_POLICY_PPL &&
6463 info->ppl_size != (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9)) {
6464 char subarray[20];
6465 struct mdinfo *member_dev;
6466
6467 sprintf(subarray, "%d", info->container_member);
6468
6469 if (mdmon_running(st->container_devnm))
6470 st->update_tail = &st->updates;
6471
6472 if (st->ss->update_subarray(st, subarray, "ppl", NULL)) {
6473 pr_err("Failed to update subarray %s\n",
6474 subarray);
6475 } else {
6476 if (st->update_tail)
6477 flush_metadata_updates(st);
6478 else
6479 st->ss->sync_metadata(st);
6480 info->ppl_size = (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9);
6481 for (member_dev = info->devs; member_dev;
6482 member_dev = member_dev->next)
6483 member_dev->ppl_size =
6484 (MULTIPLE_PPL_AREA_SIZE_IMSM >> 9);
6485 }
6486 }
6487
b23d0750 6488 if (ret == 1) {
2fc0fc63
AP		 6489 struct imsm_map *map = get_imsm_map(dev, MAP_X);
6490
50b9c10d
PB		 6491 if (map->map_state == IMSM_T_STATE_UNINITIALIZED ||
6492 (map->map_state == IMSM_T_STATE_NORMAL &&
2ec9d182 6493 !(dev->vol.dirty & RAIDVOL_DIRTY)) ||
e397cefe 6494 (is_rebuilding(dev) &&
4036e7ee 6495 vol_curr_migr_unit(dev) == 0 &&
2ec9d182 6496 get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_1) != idx))
b23d0750
AP		 6497 ret = st->ss->write_init_ppl(st, info, d->fd);
6498 else
6499 info->mismatch_cnt++;
e397cefe
AP		 6500 } else if (ret == 0 &&
6501 ppl_hdr->entries_count == 0 &&
6502 is_rebuilding(dev) &&
6503 info->resync_start == 0) {
6504 /*
6505 * The header has no entries - add a single empty entry and
6506 * rewrite the header to prevent the kernel from going into
6507 * resync after an interrupted rebuild.
6508 */
6509 ppl_hdr->entries_count = __cpu_to_le32(1);
6510 ret = write_ppl_header(info->ppl_sector, d->fd, buf);
b23d0750 6511 }
2432ce9b 6512
e397cefe
AP		 6513 free(buf_orig);
6514
2432ce9b
AP		 6515 return ret;
6516}
6517
2432ce9b
AP		 6518static int write_init_ppl_imsm_all(struct supertype *st, struct mdinfo *info)
6519{
6520 struct intel_super *super = st->sb;
6521 struct dl *d;
6522 int ret = 0;
6523
6524 if (info->consistency_policy != CONSISTENCY_POLICY_PPL ||
6525 info->array.level != 5)
6526 return 0;
6527
6528 for (d = super->disks; d ; d = d->next) {
6529 if (d->index < 0 || is_failed(&d->disk))
6530 continue;
6531
6532 ret = st->ss->write_init_ppl(st, info, d->fd);
6533 if (ret)
6534 break;
6535 }
6536
6537 return ret;
6538}
43dad3d6 6539
fbc42556
JR		 6540/*******************************************************************************
6541 * Function: write_init_bitmap_imsm_vol
6542 * Description: Write a bitmap header and prepares the area for the bitmap.
6543 * Parameters:
6544 * st : supertype information
6545 * vol_idx : the volume index to use
6546 *
6547 * Returns:
6548 * 0 : success
6549 * -1 : fail
6550 ******************************************************************************/
6551static int write_init_bitmap_imsm_vol(struct supertype *st, int vol_idx)
6552{
6553 struct intel_super *super = st->sb;
6554 int prev_current_vol = super->current_vol;
6555 struct dl *d;
6556 int ret = 0;
6557
6558 super->current_vol = vol_idx;
6559 for (d = super->disks; d; d = d->next) {
6560 if (d->index < 0 || is_failed(&d->disk))
6561 continue;
6562 ret = st->ss->write_bitmap(st, d->fd, NoUpdate);
6563 if (ret)
6564 break;
6565 }
6566 super->current_vol = prev_current_vol;
6567 return ret;
6568}
6569
6570/*******************************************************************************
6571 * Function: write_init_bitmap_imsm_all
6572 * Description: Write a bitmap header and prepares the area for the bitmap.
6573 * Operation is executed for volumes with CONSISTENCY_POLICY_BITMAP.
6574 * Parameters:
6575 * st : supertype information
6576 * info : info about the volume where the bitmap should be written
6577 * vol_idx : the volume index to use
6578 *
6579 * Returns:
6580 * 0 : success
6581 * -1 : fail
6582 ******************************************************************************/
6583static int write_init_bitmap_imsm_all(struct supertype *st, struct mdinfo *info,
6584 int vol_idx)
6585{
6586 int ret = 0;
6587
6588 if (info && (info->consistency_policy == CONSISTENCY_POLICY_BITMAP))
6589 ret = write_init_bitmap_imsm_vol(st, vol_idx);
6590
6591 return ret;
6592}
6593
c2c087e6
DW		 6594static int write_init_super_imsm(struct supertype *st)
6595{
9b1fb677
DW		 6596 struct intel_super *super = st->sb;
6597 int current_vol = super->current_vol;
2432ce9b
AP		 6598 int rv = 0;
6599 struct mdinfo info;
6600
6601 getinfo_super_imsm(st, &info, NULL);
9b1fb677
DW		 6602
6603 /* we are done with current_vol reset it to point st at the container */
6604 super->current_vol = -1;
6605
8273f55e 6606 if (st->update_tail) {
43dad3d6
DW		 6607 /* queue the recently created array / added disk
6608 * as a metadata update */
8273f55e 6609
43dad3d6 6610 /* determine if we are creating a volume or adding a disk */
9b1fb677 6611 if (current_vol < 0) {
1a64be56
LM		 6612 /* in the mgmt (add/remove) disk case we are running
6613 * in mdmon context, so don't close fd's
43dad3d6 6614 */
2432ce9b
AP		 6615 rv = mgmt_disk(st);
6616 } else {
fbc42556 6617 /* adding the second volume to the array */
2432ce9b 6618 rv = write_init_ppl_imsm_all(st, &info);
fbc42556
JR		 6619 if (!rv)
6620 rv = write_init_bitmap_imsm_all(st, &info, current_vol);
2432ce9b
AP		 6621 if (!rv)
6622 rv = create_array(st, current_vol);
6623 }
d682f344
N		 6624 } else {
6625 struct dl *d;
6626 for (d = super->disks; d; d = d->next)
ba728be7 6627 Kill(d->devname, NULL, 0, -1, 1);
fbc42556 6628 if (current_vol >= 0) {
2432ce9b 6629 rv = write_init_ppl_imsm_all(st, &info);
fbc42556
JR		 6630 if (!rv)
6631 rv = write_init_bitmap_imsm_all(st, &info, current_vol);
6632 }
6633
2432ce9b
AP		 6634 if (!rv)
6635 rv = write_super_imsm(st, 1);
d682f344 6636 }
2432ce9b
AP		 6637
6638 return rv;
cdddbdbc
DW		 6639}
6640
e683ca88 6641static int store_super_imsm(struct supertype *st, int fd)
cdddbdbc 6642{
e683ca88
DW		 6643 struct intel_super *super = st->sb;
6644 struct imsm_super *mpb = super ? super->anchor : NULL;
551c80c1 6645
e683ca88 6646 if (!mpb)
ad97895e
DW		 6647 return 1;
6648
f36a9ecd
PB		 6649 if (super->sector_size == 4096)
6650 convert_to_4k(super);
e683ca88 6651 return store_imsm_mpb(fd, mpb);
cdddbdbc
DW		 6652}
6653
cdddbdbc 6654static int validate_geometry_imsm_container(struct supertype *st, int level,
1f5d54a0 6655 int raiddisks,
af4348dd
N		 6656 unsigned long long data_offset,
6657 char *dev,
2c514b71
NB		 6658 unsigned long long *freesize,
6659 int verbose)
cdddbdbc 6660{
c2c087e6
DW		 6661 int fd;
6662 unsigned long long ldsize;
8662f92d 6663 struct intel_super *super = NULL;
f2f5c343 6664 int rv = 0;
cdddbdbc 6665
c2c087e6
DW		 6666 if (level != LEVEL_CONTAINER)
6667 return 0;
6668 if (!dev)
6669 return 1;
6670
dca80fcd 6671 fd = dev_open(dev, O_RDONLY|O_EXCL);
c2c087e6 6672 if (fd < 0) {
ba728be7 6673 if (verbose > 0)
e7b84f9d 6674 pr_err("imsm: Cannot open %s: %s\n",
2c514b71 6675 dev, strerror(errno));
c2c087e6
DW		 6676 return 0;
6677 }
8662f92d
MT		 6678 if (!get_dev_size(fd, dev, &ldsize))
6679 goto exit;
f2f5c343
LM		 6680
6681 /* capabilities retrieve could be possible
6682 * note that there is no fd for the disks in array.
6683 */
6684 super = alloc_super();
8662f92d
MT		 6685 if (!super)
6686 goto exit;
6687
6688 if (!get_dev_sector_size(fd, NULL, &super->sector_size))
6689 goto exit;
fa7bb6f8 6690
ba728be7 6691 rv = find_intel_hba_capability(fd, super, verbose > 0 ? dev : NULL);
f2f5c343
LM		 6692 if (rv != 0) {
6693#if DEBUG
6694 char str[256];
6695 fd2devname(fd, str);
1ade5cc1 6696 dprintf("fd: %d %s orom: %p rv: %d raiddisk: %d\n",
f2f5c343
LM		 6697 fd, str, super->orom, rv, raiddisks);
6698#endif
6699 /* no orom/efi or non-intel hba of the disk */
8662f92d
MT		 6700 rv = 0;
6701 goto exit;
f2f5c343 6702 }
9126b9a8
CA		 6703 if (super->orom) {
6704 if (raiddisks > super->orom->tds) {
6705 if (verbose)
7a862a02 6706 pr_err("%d exceeds maximum number of platform supported disks: %d\n",
9126b9a8 6707 raiddisks, super->orom->tds);
8662f92d 6708 goto exit;
9126b9a8
CA		 6709 }
6710 if ((super->orom->attr & IMSM_OROM_ATTR_2TB_DISK) == 0 &&
6711 (ldsize >> 9) >> 32 > 0) {
6712 if (verbose)
e7b84f9d 6713 pr_err("%s exceeds maximum platform supported size\n", dev);
8662f92d
MT		 6714 goto exit;
6715 }
6716
6717 if (super->hba->type == SYS_DEV_VMD ||
6718 super->hba->type == SYS_DEV_NVME) {
6719 if (!imsm_is_nvme_namespace_supported(fd, 1)) {
6720 if (verbose)
6721 pr_err("NVMe namespace %s is not supported by IMSM\n",
6722 basename(dev));
6723 goto exit;
6724 }
9126b9a8 6725 }
f2f5c343 6726 }
1f5d54a0
MT		 6727 if (freesize)
6728 *freesize = avail_size_imsm(st, ldsize >> 9, data_offset);
8662f92d
MT		 6729 rv = 1;
6730exit:
6731 if (super)
6732 free_imsm(super);
6733 close(fd);
c2c087e6 6734
8662f92d 6735 return rv;
cdddbdbc
DW		 6736}
6737
0dcecb2e
DW		 6738static unsigned long long find_size(struct extent *e, int *idx, int num_extents)
6739{
6740 const unsigned long long base_start = e[*idx].start;
6741 unsigned long long end = base_start + e[*idx].size;
6742 int i;
6743
6744 if (base_start == end)
6745 return 0;
6746
6747 *idx = *idx + 1;
6748 for (i = *idx; i < num_extents; i++) {
6749 /* extend overlapping extents */
6750 if (e[i].start >= base_start &&
6751 e[i].start <= end) {
6752 if (e[i].size == 0)
6753 return 0;
6754 if (e[i].start + e[i].size > end)
6755 end = e[i].start + e[i].size;
6756 } else if (e[i].start > end) {
6757 *idx = i;
6758 break;
6759 }
6760 }
6761
6762 return end - base_start;
6763}
6764
6765static unsigned long long merge_extents(struct intel_super *super, int sum_extents)
6766{
6767 /* build a composite disk with all known extents and generate a new
6768 * 'maxsize' given the "all disks in an array must share a common start
6769 * offset" constraint
6770 */
503975b9 6771 struct extent *e = xcalloc(sum_extents, sizeof(*e));
0dcecb2e
DW		 6772 struct dl *dl;
6773 int i, j;
6774 int start_extent;
6775 unsigned long long pos;
b9d77223 6776 unsigned long long start = 0;
0dcecb2e
DW		 6777 unsigned long long maxsize;
6778 unsigned long reserve;
6779
0dcecb2e
DW		 6780 /* coalesce and sort all extents. also, check to see if we need to
6781 * reserve space between member arrays
6782 */
6783 j = 0;
6784 for (dl = super->disks; dl; dl = dl->next) {
6785 if (!dl->e)
6786 continue;
6787 for (i = 0; i < dl->extent_cnt; i++)
6788 e[j++] = dl->e[i];
6789 }
6790 qsort(e, sum_extents, sizeof(*e), cmp_extent);
6791
6792 /* merge extents */
6793 i = 0;
6794 j = 0;
6795 while (i < sum_extents) {
6796 e[j].start = e[i].start;
6797 e[j].size = find_size(e, &i, sum_extents);
6798 j++;
6799 if (e[j-1].size == 0)
6800 break;
6801 }
6802
6803 pos = 0;
6804 maxsize = 0;
6805 start_extent = 0;
6806 i = 0;
6807 do {
6808 unsigned long long esize;
6809
6810 esize = e[i].start - pos;
6811 if (esize >= maxsize) {
6812 maxsize = esize;
6813 start = pos;
6814 start_extent = i;
6815 }
6816 pos = e[i].start + e[i].size;
6817 i++;
6818 } while (e[i-1].size);
6819 free(e);
6820
a7dd165b
DW		 6821 if (maxsize == 0)
6822 return 0;
6823
6824 /* FIXME assumes volume at offset 0 is the first volume in a
6825 * container
6826 */
0dcecb2e
DW		 6827 if (start_extent > 0)
6828 reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */
6829 else
6830 reserve = 0;
6831
6832 if (maxsize < reserve)
a7dd165b 6833 return 0;
0dcecb2e 6834
5551b113 6835 super->create_offset = ~((unsigned long long) 0);
0dcecb2e 6836 if (start + reserve > super->create_offset)
a7dd165b 6837 return 0; /* start overflows create_offset */
0dcecb2e
DW		 6838 super->create_offset = start + reserve;
6839
6840 return maxsize - reserve;
6841}
6842
88c32bb1
DW		 6843static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks)
6844{
6845 if (level < 0 || level == 6 || level == 4)
6846 return 0;
6847
6848 /* if we have an orom prevent invalid raid levels */
6849 if (orom)
6850 switch (level) {
6851 case 0: return imsm_orom_has_raid0(orom);
6852 case 1:
6853 if (raiddisks > 2)
6854 return imsm_orom_has_raid1e(orom);
1c556e92
DW		 6855 return imsm_orom_has_raid1(orom) && raiddisks == 2;
6856 case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4;
6857 case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2;
88c32bb1
DW		 6858 }
6859 else
6860 return 1; /* not on an Intel RAID platform so anything goes */
6861
6862 return 0;
6863}
6864
ca9de185
LM		 6865static int
6866active_arrays_by_format(char *name, char* hba, struct md_list **devlist,
6867 int dpa, int verbose)
6868{
6869 struct mdstat_ent *mdstat = mdstat_read(0, 0);
594dc1b8 6870 struct mdstat_ent *memb;
ca9de185
LM		 6871 int count = 0;
6872 int num = 0;
594dc1b8 6873 struct md_list *dv;
ca9de185
LM		 6874 int found;
6875
6876 for (memb = mdstat ; memb ; memb = memb->next) {
6877 if (memb->metadata_version &&
fc54fe7a 6878 (strncmp(memb->metadata_version, "external:", 9) == 0) &&
ca9de185
LM		 6879 (strcmp(&memb->metadata_version[9], name) == 0) &&
6880 !is_subarray(memb->metadata_version+9) &&
6881 memb->members) {
6882 struct dev_member *dev = memb->members;
6883 int fd = -1;
6884 while(dev && (fd < 0)) {
503975b9
N		 6885 char *path = xmalloc(strlen(dev->name) + strlen("/dev/") + 1);
6886 num = sprintf(path, "%s%s", "/dev/", dev->name);
6887 if (num > 0)
6888 fd = open(path, O_RDONLY, 0);
089f9d79 6889 if (num <= 0 || fd < 0) {
676e87a8 6890 pr_vrb("Cannot open %s: %s\n",
503975b9 6891 dev->name, strerror(errno));
ca9de185 6892 }
503975b9 6893 free(path);
ca9de185
LM		 6894 dev = dev->next;
6895 }
6896 found = 0;
089f9d79 6897 if (fd >= 0 && disk_attached_to_hba(fd, hba)) {
ca9de185
LM		 6898 struct mdstat_ent *vol;
6899 for (vol = mdstat ; vol ; vol = vol->next) {
089f9d79 6900 if (vol->active > 0 &&
ca9de185 6901 vol->metadata_version &&
9581efb1 6902 is_container_member(vol, memb->devnm)) {
ca9de185
LM		 6903 found++;
6904 count++;
6905 }
6906 }
6907 if (*devlist && (found < dpa)) {
503975b9 6908 dv = xcalloc(1, sizeof(*dv));
9581efb1
N		 6909 dv->devname = xmalloc(strlen(memb->devnm) + strlen("/dev/") + 1);
6910 sprintf(dv->devname, "%s%s", "/dev/", memb->devnm);
503975b9
N		 6911 dv->found = found;
6912 dv->used = 0;
6913 dv->next = *devlist;
6914 *devlist = dv;
ca9de185
LM		 6915 }
6916 }
6917 if (fd >= 0)
6918 close(fd);
6919 }
6920 }
6921 free_mdstat(mdstat);
6922 return count;
6923}
6924
6925#ifdef DEBUG_LOOP
6926static struct md_list*
6927get_loop_devices(void)
6928{
6929 int i;
6930 struct md_list *devlist = NULL;
594dc1b8 6931 struct md_list *dv;
ca9de185
LM		 6932
6933 for(i = 0; i < 12; i++) {
503975b9
N		 6934 dv = xcalloc(1, sizeof(*dv));
6935 dv->devname = xmalloc(40);
ca9de185
LM		 6936 sprintf(dv->devname, "/dev/loop%d", i);
6937 dv->next = devlist;
6938 devlist = dv;
6939 }
6940 return devlist;
6941}
6942#endif
6943
6944static struct md_list*
6945get_devices(const char *hba_path)
6946{
6947 struct md_list *devlist = NULL;
594dc1b8 6948 struct md_list *dv;
ca9de185
LM		 6949 struct dirent *ent;
6950 DIR *dir;
6951 int err = 0;
6952
6953#if DEBUG_LOOP
6954 devlist = get_loop_devices();
6955 return devlist;
6956#endif
6957 /* scroll through /sys/dev/block looking for devices attached to
6958 * this hba
6959 */
6960 dir = opendir("/sys/dev/block");
6961 for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) {
6962 int fd;
6963 char buf[1024];
6964 int major, minor;
6965 char *path = NULL;
6966 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
6967 continue;
7c798f87 6968 path = devt_to_devpath(makedev(major, minor), 1, NULL);
ca9de185
LM		 6969 if (!path)
6970 continue;
6971 if (!path_attached_to_hba(path, hba_path)) {
6972 free(path);
6973 path = NULL;
6974 continue;
6975 }
6976 free(path);
6977 path = NULL;
6978 fd = dev_open(ent->d_name, O_RDONLY);
6979 if (fd >= 0) {
6980 fd2devname(fd, buf);
6981 close(fd);
6982 } else {
e7b84f9d 6983 pr_err("cannot open device: %s\n",
ca9de185
LM		 6984 ent->d_name);
6985 continue;
6986 }
6987
503975b9
N		 6988 dv = xcalloc(1, sizeof(*dv));
6989 dv->devname = xstrdup(buf);
ca9de185
LM		 6990 dv->next = devlist;
6991 devlist = dv;
6992 }
6993 if (err) {
6994 while(devlist) {
6995 dv = devlist;
6996 devlist = devlist->next;
6997 free(dv->devname);
6998 free(dv);
6999 }
7000 }
562aa102 7001 closedir(dir);
ca9de185
LM		 7002 return devlist;
7003}
7004
7005static int
7006count_volumes_list(struct md_list *devlist, char *homehost,
7007 int verbose, int *found)
7008{
7009 struct md_list *tmpdev;
7010 int count = 0;
594dc1b8 7011 struct supertype *st;
ca9de185
LM		 7012
7013 /* first walk the list of devices to find a consistent set
7014 * that match the criterea, if that is possible.
7015 * We flag the ones we like with 'used'.
7016 */
7017 *found = 0;
7018 st = match_metadata_desc_imsm("imsm");
7019 if (st == NULL) {
676e87a8 7020 pr_vrb("cannot allocate memory for imsm supertype\n");
ca9de185
LM		 7021 return 0;
7022 }
7023
7024 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
7025 char *devname = tmpdev->devname;
0a6bff09 7026 dev_t rdev;
ca9de185
LM		 7027 struct supertype *tst;
7028 int dfd;
7029 if (tmpdev->used > 1)
7030 continue;
7031 tst = dup_super(st);
7032 if (tst == NULL) {
676e87a8 7033 pr_vrb("cannot allocate memory for imsm supertype\n");
ca9de185
LM		 7034 goto err_1;
7035 }
7036 tmpdev->container = 0;
7037 dfd = dev_open(devname, O_RDONLY|O_EXCL);
7038 if (dfd < 0) {
1ade5cc1 7039 dprintf("cannot open device %s: %s\n",
ca9de185
LM		 7040 devname, strerror(errno));
7041 tmpdev->used = 2;
0a6bff09 7042 } else if (!fstat_is_blkdev(dfd, devname, &rdev)) {
ca9de185
LM		 7043 tmpdev->used = 2;
7044 } else if (must_be_container(dfd)) {
7045 struct supertype *cst;
7046 cst = super_by_fd(dfd, NULL);
7047 if (cst == NULL) {
1ade5cc1 7048 dprintf("cannot recognize container type %s\n",
ca9de185
LM		 7049 devname);
7050 tmpdev->used = 2;
7051 } else if (tst->ss != st->ss) {
1ade5cc1 7052 dprintf("non-imsm container - ignore it: %s\n",
ca9de185
LM		 7053 devname);
7054 tmpdev->used = 2;
7055 } else if (!tst->ss->load_container ||
7056 tst->ss->load_container(tst, dfd, NULL))
7057 tmpdev->used = 2;
7058 else {
7059 tmpdev->container = 1;
7060 }
7061 if (cst)
7062 cst->ss->free_super(cst);
7063 } else {
0a6bff09 7064 tmpdev->st_rdev = rdev;
ca9de185 7065 if (tst->ss->load_super(tst,dfd, NULL)) {
1ade5cc1 7066 dprintf("no RAID superblock on %s\n",
ca9de185
LM		 7067 devname);
7068 tmpdev->used = 2;
7069 } else if (tst->ss->compare_super == NULL) {
1ade5cc1 7070 dprintf("Cannot assemble %s metadata on %s\n",
ca9de185
LM		 7071 tst->ss->name, devname);
7072 tmpdev->used = 2;
7073 }
7074 }
7075 if (dfd >= 0)
7076 close(dfd);
7077 if (tmpdev->used == 2 || tmpdev->used == 4) {
7078 /* Ignore unrecognised devices during auto-assembly */
7079 goto loop;
7080 }
7081 else {
7082 struct mdinfo info;
7083 tst->ss->getinfo_super(tst, &info, NULL);
7084
7085 if (st->minor_version == -1)
7086 st->minor_version = tst->minor_version;
7087
7088 if (memcmp(info.uuid, uuid_zero,
7089 sizeof(int[4])) == 0) {
7090 /* this is a floating spare. It cannot define
7091 * an array unless there are no more arrays of
7092 * this type to be found. It can be included
7093 * in an array of this type though.
7094 */
7095 tmpdev->used = 3;
7096 goto loop;
7097 }
7098
7099 if (st->ss != tst->ss ||
7100 st->minor_version != tst->minor_version ||
c7b8547c 7101 st->ss->compare_super(st, tst, 1) != 0) {
ca9de185
LM		 7102 /* Some mismatch. If exactly one array matches this host,
7103 * we can resolve on that one.
7104 * Or, if we are auto assembling, we just ignore the second
7105 * for now.
7106 */
1ade5cc1 7107 dprintf("superblock on %s doesn't match others - assembly aborted\n",
ca9de185
LM		 7108 devname);
7109 goto loop;
7110 }
7111 tmpdev->used = 1;
7112 *found = 1;
7113 dprintf("found: devname: %s\n", devname);
7114 }
7115 loop:
7116 if (tst)
7117 tst->ss->free_super(tst);
7118 }
7119 if (*found != 0) {
7120 int err;
7121 if ((err = load_super_imsm_all(st, -1, &st->sb, NULL, devlist, 0)) == 0) {
7122 struct mdinfo *iter, *head = st->ss->container_content(st, NULL);
7123 for (iter = head; iter; iter = iter->next) {
7124 dprintf("content->text_version: %s vol\n",
7125 iter->text_version);
7126 if (iter->array.state & (1<<MD_SB_BLOCK_VOLUME)) {
7127 /* do not assemble arrays with unsupported
7128 configurations */
1ade5cc1 7129 dprintf("Cannot activate member %s.\n",
ca9de185
LM		 7130 iter->text_version);
7131 } else
7132 count++;
7133 }
7134 sysfs_free(head);
7135
7136 } else {
1ade5cc1 7137 dprintf("No valid super block on device list: err: %d %p\n",
ca9de185
LM		 7138 err, st->sb);
7139 }
7140 } else {
1ade5cc1 7141 dprintf("no more devices to examine\n");
ca9de185
LM		 7142 }
7143
7144 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
089f9d79 7145 if (tmpdev->used == 1 && tmpdev->found) {
ca9de185
LM		 7146 if (count) {
7147 if (count < tmpdev->found)
7148 count = 0;
7149 else
7150 count -= tmpdev->found;
7151 }
7152 }
7153 if (tmpdev->used == 1)
7154 tmpdev->used = 4;
7155 }
7156 err_1:
7157 if (st)
7158 st->ss->free_super(st);
7159 return count;
7160}
7161
d3c11416
AO		 7162static int __count_volumes(char *hba_path, int dpa, int verbose,
7163 int cmp_hba_path)
ca9de185 7164{
72a45777 7165 struct sys_dev *idev, *intel_devices = find_intel_devices();
ca9de185 7166 int count = 0;
72a45777
PB		 7167 const struct orom_entry *entry;
7168 struct devid_list *dv, *devid_list;
ca9de185 7169
d3c11416 7170 if (!hba_path)
ca9de185
LM		 7171 return 0;
7172
72a45777 7173 for (idev = intel_devices; idev; idev = idev->next) {
d3c11416
AO		 7174 if (strstr(idev->path, hba_path))
7175 break;
72a45777
PB		 7176 }
7177
7178 if (!idev || !idev->dev_id)
ca9de185 7179 return 0;
72a45777
PB		 7180
7181 entry = get_orom_entry_by_device_id(idev->dev_id);
7182
7183 if (!entry || !entry->devid_list)
7184 return 0;
7185
7186 devid_list = entry->devid_list;
7187 for (dv = devid_list; dv; dv = dv->next) {
594dc1b8 7188 struct md_list *devlist;
d3c11416
AO		 7189 struct sys_dev *device = NULL;
7190 char *hpath;
72a45777
PB		 7191 int found = 0;
7192
d3c11416
AO		 7193 if (cmp_hba_path)
7194 device = device_by_id_and_path(dv->devid, hba_path);
7195 else
7196 device = device_by_id(dv->devid);
7197
72a45777 7198 if (device)
d3c11416 7199 hpath = device->path;
72a45777
PB		 7200 else
7201 return 0;
7202
d3c11416 7203 devlist = get_devices(hpath);
72a45777
PB		 7204 /* if no intel devices return zero volumes */
7205 if (devlist == NULL)
7206 return 0;
7207
d3c11416
AO		 7208 count += active_arrays_by_format("imsm", hpath, &devlist, dpa,
7209 verbose);
7210 dprintf("path: %s active arrays: %d\n", hpath, count);
72a45777
PB		 7211 if (devlist == NULL)
7212 return 0;
7213 do {
7214 found = 0;
7215 count += count_volumes_list(devlist,
7216 NULL,
7217 verbose,
7218 &found);
7219 dprintf("found %d count: %d\n", found, count);
7220 } while (found);
7221
d3c11416 7222 dprintf("path: %s total number of volumes: %d\n", hpath, count);
72a45777
PB		 7223
7224 while (devlist) {
7225 struct md_list *dv = devlist;
7226 devlist = devlist->next;
7227 free(dv->devname);
7228 free(dv);
7229 }
ca9de185
LM		 7230 }
7231 return count;
7232}
7233
d3c11416
AO		 7234static int count_volumes(struct intel_hba *hba, int dpa, int verbose)
7235{
7236 if (!hba)
7237 return 0;
7238 if (hba->type == SYS_DEV_VMD) {
7239 struct sys_dev *dev;
7240 int count = 0;
7241
7242 for (dev = find_intel_devices(); dev; dev = dev->next) {
7243 if (dev->type == SYS_DEV_VMD)
7244 count += __count_volumes(dev->path, dpa,
7245 verbose, 1);
7246 }
7247 return count;
7248 }
7249 return __count_volumes(hba->path, dpa, verbose, 0);
7250}
7251
cd9d1ac7
DW		 7252static int imsm_default_chunk(const struct imsm_orom *orom)
7253{
7254 /* up to 512 if the plaform supports it, otherwise the platform max.
7255 * 128 if no platform detected
7256 */
7257 int fs = max(7, orom ? fls(orom->sss) : 0);
7258
7259 return min(512, (1 << fs));
7260}
73408129 7261
6592ce37
DW		 7262static int
7263validate_geometry_imsm_orom(struct intel_super *super, int level, int layout,
2cc699af 7264 int raiddisks, int *chunk, unsigned long long size, int verbose)
6592ce37 7265{
660260d0
DW		 7266 /* check/set platform and metadata limits/defaults */
7267 if (super->orom && raiddisks > super->orom->dpa) {
676e87a8 7268 pr_vrb("platform supports a maximum of %d disks per array\n",
660260d0 7269 super->orom->dpa);
73408129
LM		 7270 return 0;
7271 }
7272
5d500228 7273 /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */
660260d0 7274 if (!is_raid_level_supported(super->orom, level, raiddisks)) {
676e87a8 7275 pr_vrb("platform does not support raid%d with %d disk%s\n",
6592ce37
DW		 7276 level, raiddisks, raiddisks > 1 ? "s" : "");
7277 return 0;
7278 }
cd9d1ac7 7279
7ccc4cc4 7280 if (*chunk == 0 || *chunk == UnSet)
cd9d1ac7
DW		 7281 *chunk = imsm_default_chunk(super->orom);
7282
7ccc4cc4 7283 if (super->orom && !imsm_orom_has_chunk(super->orom, *chunk)) {
676e87a8 7284 pr_vrb("platform does not support a chunk size of: %d\n", *chunk);
cd9d1ac7 7285 return 0;
6592ce37 7286 }
cd9d1ac7 7287
6592ce37
DW		 7288 if (layout != imsm_level_to_layout(level)) {
7289 if (level == 5)
676e87a8 7290 pr_vrb("imsm raid 5 only supports the left-asymmetric layout\n");
6592ce37 7291 else if (level == 10)
676e87a8 7292 pr_vrb("imsm raid 10 only supports the n2 layout\n");
6592ce37 7293 else
676e87a8 7294 pr_vrb("imsm unknown layout %#x for this raid level %d\n",
6592ce37
DW		 7295 layout, level);
7296 return 0;
7297 }
2cc699af 7298
7ccc4cc4 7299 if (super->orom && (super->orom->attr & IMSM_OROM_ATTR_2TB) == 0 &&
2cc699af 7300 (calc_array_size(level, raiddisks, layout, *chunk, size) >> 32) > 0) {
676e87a8 7301 pr_vrb("platform does not support a volume size over 2TB\n");
2cc699af
CA		 7302 return 0;
7303 }
614902f6 7304
6592ce37
DW		 7305 return 1;
7306}
7307
1011e834 7308/* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd
c2c087e6
DW		 7309 * FIX ME add ahci details
7310 */
8b353278 7311static int validate_geometry_imsm_volume(struct supertype *st, int level,
c21e737b 7312 int layout, int raiddisks, int *chunk,
af4348dd
N		 7313 unsigned long long size,
7314 unsigned long long data_offset,
7315 char *dev,
2c514b71
NB		 7316 unsigned long long *freesize,
7317 int verbose)
cdddbdbc 7318{
9e04ac1c 7319 dev_t rdev;
c2c087e6 7320 struct intel_super *super = st->sb;
b2916f25 7321 struct imsm_super *mpb;
c2c087e6
DW		 7322 struct dl *dl;
7323 unsigned long long pos = 0;
7324 unsigned long long maxsize;
7325 struct extent *e;
7326 int i;
cdddbdbc 7327
88c32bb1
DW		 7328 /* We must have the container info already read in. */
7329 if (!super)
c2c087e6
DW		 7330 return 0;
7331
b2916f25
JS		 7332 mpb = super->anchor;
7333
2cc699af 7334 if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, size, verbose)) {
3e684231 7335 pr_err("RAID geometry validation failed. Cannot proceed with the action(s).\n");
c2c087e6 7336 return 0;
d54559f0 7337 }
c2c087e6
DW		 7338 if (!dev) {
7339 /* General test: make sure there is space for
2da8544a
DW		 7340 * 'raiddisks' device extents of size 'size' at a given
7341 * offset
c2c087e6 7342 */
e46273eb 7343 unsigned long long minsize = size;
b7528a20 7344 unsigned long long start_offset = MaxSector;
c2c087e6
DW		 7345 int dcnt = 0;
7346 if (minsize == 0)
7347 minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
7348 for (dl = super->disks; dl ; dl = dl->next) {
7349 int found = 0;
7350
bf5a934a 7351 pos = 0;
c2c087e6 7352 i = 0;
05501181 7353 e = get_extents(super, dl, 0);
c2c087e6
DW		 7354 if (!e) continue;
7355 do {
7356 unsigned long long esize;
7357 esize = e[i].start - pos;
7358 if (esize >= minsize)
7359 found = 1;
b7528a20 7360 if (found && start_offset == MaxSector) {
2da8544a
DW		 7361 start_offset = pos;
7362 break;
7363 } else if (found && pos != start_offset) {
7364 found = 0;
7365 break;
7366 }
c2c087e6
DW		 7367 pos = e[i].start + e[i].size;
7368 i++;
7369 } while (e[i-1].size);
7370 if (found)
7371 dcnt++;
7372 free(e);
7373 }
7374 if (dcnt < raiddisks) {
2c514b71 7375 if (verbose)
7a862a02 7376 pr_err("imsm: Not enough devices with space for this array (%d < %d)\n",
2c514b71 7377 dcnt, raiddisks);
c2c087e6
DW		 7378 return 0;
7379 }
7380 return 1;
7381 }
0dcecb2e 7382
c2c087e6 7383 /* This device must be a member of the set */
9e04ac1c 7384 if (!stat_is_blkdev(dev, &rdev))
c2c087e6
DW		 7385 return 0;
7386 for (dl = super->disks ; dl ; dl = dl->next) {
9e04ac1c
ZL		 7387 if (dl->major == (int)major(rdev) &&
7388 dl->minor == (int)minor(rdev))
c2c087e6
DW		 7389 break;
7390 }
7391 if (!dl) {
2c514b71 7392 if (verbose)
7a862a02 7393 pr_err("%s is not in the same imsm set\n", dev);
c2c087e6 7394 return 0;
a20d2ba5
DW		 7395 } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) {
7396 /* If a volume is present then the current creation attempt
7397 * cannot incorporate new spares because the orom may not
7398 * understand this configuration (all member disks must be
7399 * members of each array in the container).
7400 */
7a862a02
N		 7401 pr_err("%s is a spare and a volume is already defined for this container\n", dev);
7402 pr_err("The option-rom requires all member disks to be a member of all volumes\n");
a20d2ba5 7403 return 0;
5fe62b94
WD		 7404 } else if (super->orom && mpb->num_raid_devs > 0 &&
7405 mpb->num_disks != raiddisks) {
7a862a02 7406 pr_err("The option-rom requires all member disks to be a member of all volumes\n");
5fe62b94 7407 return 0;
c2c087e6 7408 }
0dcecb2e
DW		 7409
7410 /* retrieve the largest free space block */
05501181 7411 e = get_extents(super, dl, 0);
c2c087e6
DW		 7412 maxsize = 0;
7413 i = 0;
0dcecb2e
DW		 7414 if (e) {
7415 do {
7416 unsigned long long esize;
7417
7418 esize = e[i].start - pos;
7419 if (esize >= maxsize)
7420 maxsize = esize;
7421 pos = e[i].start + e[i].size;
7422 i++;
7423 } while (e[i-1].size);
7424 dl->e = e;
7425 dl->extent_cnt = i;
7426 } else {
7427 if (verbose)
e7b84f9d 7428 pr_err("unable to determine free space for: %s\n",
0dcecb2e
DW		 7429 dev);
7430 return 0;
7431 }
7432 if (maxsize < size) {
7433 if (verbose)
e7b84f9d 7434 pr_err("%s not enough space (%llu < %llu)\n",
0dcecb2e
DW		 7435 dev, maxsize, size);
7436 return 0;
7437 }
7438
7439 /* count total number of extents for merge */
7440 i = 0;
7441 for (dl = super->disks; dl; dl = dl->next)
7442 if (dl->e)
7443 i += dl->extent_cnt;
7444
7445 maxsize = merge_extents(super, i);
3baa56ab 7446
1a1ced1e
KS		 7447 if (mpb->num_raid_devs > 0 && size && size != maxsize)
7448 pr_err("attempting to create a second volume with size less then remaining space.\n");
3baa56ab 7449
a7dd165b 7450 if (maxsize < size || maxsize == 0) {
b3071342
LD		 7451 if (verbose) {
7452 if (maxsize == 0)
7a862a02 7453 pr_err("no free space left on device. Aborting...\n");
b3071342 7454 else
7a862a02 7455 pr_err("not enough space to create volume of given size (%llu < %llu). Aborting...\n",
b3071342
LD		 7456 maxsize, size);
7457 }
0dcecb2e 7458 return 0;
0dcecb2e
DW		 7459 }
7460
c2c087e6
DW		 7461 *freesize = maxsize;
7462
ca9de185 7463 if (super->orom) {
72a45777 7464 int count = count_volumes(super->hba,
ca9de185
LM		 7465 super->orom->dpa, verbose);
7466 if (super->orom->vphba <= count) {
676e87a8 7467 pr_vrb("platform does not support more than %d raid volumes.\n",
ca9de185
LM		 7468 super->orom->vphba);
7469 return 0;
7470 }
7471 }
c2c087e6 7472 return 1;
cdddbdbc
DW		 7473}
7474
13bcac90 7475static int imsm_get_free_size(struct supertype *st, int raiddisks,
efb30e7f
DW		 7476 unsigned long long size, int chunk,
7477 unsigned long long *freesize)
7478{
7479 struct intel_super *super = st->sb;
7480 struct imsm_super *mpb = super->anchor;
7481 struct dl *dl;
7482 int i;
7483 int extent_cnt;
7484 struct extent *e;
7485 unsigned long long maxsize;
7486 unsigned long long minsize;
7487 int cnt;
7488 int used;
7489
7490 /* find the largest common start free region of the possible disks */
7491 used = 0;
7492 extent_cnt = 0;
7493 cnt = 0;
7494 for (dl = super->disks; dl; dl = dl->next) {
7495 dl->raiddisk = -1;
7496
7497 if (dl->index >= 0)
7498 used++;
7499
7500 /* don't activate new spares if we are orom constrained
7501 * and there is already a volume active in the container
7502 */
7503 if (super->orom && dl->index < 0 && mpb->num_raid_devs)
7504 continue;
7505
05501181 7506 e = get_extents(super, dl, 0);
efb30e7f
DW		 7507 if (!e)
7508 continue;
7509 for (i = 1; e[i-1].size; i++)
7510 ;
7511 dl->e = e;
7512 dl->extent_cnt = i;
7513 extent_cnt += i;
7514 cnt++;
7515 }
7516
7517 maxsize = merge_extents(super, extent_cnt);
7518 minsize = size;
7519 if (size == 0)
612e59d8
CA		 7520 /* chunk is in K */
7521 minsize = chunk * 2;
efb30e7f
DW		 7522
7523 if (cnt < raiddisks ||
7524 (super->orom && used && used != raiddisks) ||
a7dd165b
DW		 7525 maxsize < minsize ||
7526 maxsize == 0) {
e7b84f9d 7527 pr_err("not enough devices with space to create array.\n");
efb30e7f
DW		 7528 return 0; /* No enough free spaces large enough */
7529 }
7530
7531 if (size == 0) {
7532 size = maxsize;
7533 if (chunk) {
612e59d8
CA		 7534 size /= 2 * chunk;
7535 size *= 2 * chunk;
efb30e7f 7536 }
f878b242
LM		 7537 maxsize = size;
7538 }
1a1ced1e
KS		 7539 if (mpb->num_raid_devs > 0 && size && size != maxsize)
7540 pr_err("attempting to create a second volume with size less then remaining space.\n");
efb30e7f
DW		 7541 cnt = 0;
7542 for (dl = super->disks; dl; dl = dl->next)
7543 if (dl->e)
7544 dl->raiddisk = cnt++;
7545
7546 *freesize = size;
7547
13bcac90
AK		 7548 dprintf("imsm: imsm_get_free_size() returns : %llu\n", size);
7549
efb30e7f
DW		 7550 return 1;
7551}
7552
13bcac90
AK		 7553static int reserve_space(struct supertype *st, int raiddisks,
7554 unsigned long long size, int chunk,
7555 unsigned long long *freesize)
7556{
7557 struct intel_super *super = st->sb;
7558 struct dl *dl;
7559 int cnt;
7560 int rv = 0;
7561
7562 rv = imsm_get_free_size(st, raiddisks, size, chunk, freesize);
7563 if (rv) {
7564 cnt = 0;
7565 for (dl = super->disks; dl; dl = dl->next)
7566 if (dl->e)
7567 dl->raiddisk = cnt++;
7568 rv = 1;
7569 }
7570
7571 return rv;
7572}
7573
bf5a934a 7574static int validate_geometry_imsm(struct supertype *st, int level, int layout,
c21e737b 7575 int raiddisks, int *chunk, unsigned long long size,
af4348dd 7576 unsigned long long data_offset,
bf5a934a 7577 char *dev, unsigned long long *freesize,
5308f117 7578 int consistency_policy, int verbose)
bf5a934a
DW		 7579{
7580 int fd, cfd;
7581 struct mdinfo *sra;
20cbe8d2 7582 int is_member = 0;
bf5a934a 7583
d54559f0
LM		 7584 /* load capability
7585 * if given unused devices create a container
bf5a934a
DW		 7586 * if given given devices in a container create a member volume
7587 */
1f5d54a0 7588 if (level == LEVEL_CONTAINER)
bf5a934a 7589 /* Must be a fresh device to add to a container */
1f5d54a0
MT		 7590 return validate_geometry_imsm_container(st, level, raiddisks,
7591 data_offset, dev,
7592 freesize, verbose);
9587c373 7593
06a6101c
BK		 7594 /*
7595 * Size is given in sectors.
7596 */
7597 if (size && (size < 2048)) {
22dc741f 7598 pr_err("Given size must be greater than 1M.\n");
54865c30
RS		 7599 /* Depends on algorithm in Create.c :
7600 * if container was given (dev == NULL) return -1,
7601 * if block device was given ( dev != NULL) return 0.
7602 */
7603 return dev ? -1 : 0;
7604 }
7605
8592f29d 7606 if (!dev) {
e91a3bad 7607 if (st->sb) {
ca9de185 7608 struct intel_super *super = st->sb;
e91a3bad 7609 if (!validate_geometry_imsm_orom(st->sb, level, layout,
2cc699af 7610 raiddisks, chunk, size,
e91a3bad
LM		 7611 verbose))
7612 return 0;
efb30e7f
DW		 7613 /* we are being asked to automatically layout a
7614 * new volume based on the current contents of
7615 * the container. If the the parameters can be
7616 * satisfied reserve_space will record the disks,
7617 * start offset, and size of the volume to be
7618 * created. add_to_super and getinfo_super
7619 * detect when autolayout is in progress.
7620 */
ca9de185
LM		 7621 /* assuming that freesize is always given when array is
7622 created */
7623 if (super->orom && freesize) {
7624 int count;
72a45777 7625 count = count_volumes(super->hba,
ca9de185
LM		 7626 super->orom->dpa, verbose);
7627 if (super->orom->vphba <= count) {
676e87a8 7628 pr_vrb("platform does not support more than %d raid volumes.\n",
ca9de185
LM		 7629 super->orom->vphba);
7630 return 0;
7631 }
7632 }
e91a3bad
LM		 7633 if (freesize)
7634 return reserve_space(st, raiddisks, size,
7ccc4cc4 7635 *chunk, freesize);
8592f29d
N		 7636 }
7637 return 1;
7638 }
bf5a934a
DW		 7639 if (st->sb) {
7640 /* creating in a given container */
7641 return validate_geometry_imsm_volume(st, level, layout,
7642 raiddisks, chunk, size,
af4348dd 7643 data_offset,
bf5a934a
DW		 7644 dev, freesize, verbose);
7645 }
7646
bf5a934a
DW		 7647 /* This device needs to be a device in an 'imsm' container */
7648 fd = open(dev, O_RDONLY|O_EXCL, 0);
7649 if (fd >= 0) {
7650 if (verbose)
e7b84f9d
N		 7651 pr_err("Cannot create this array on device %s\n",
7652 dev);
bf5a934a
DW		 7653 close(fd);
7654 return 0;
7655 }
7656 if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) {
7657 if (verbose)
e7b84f9d 7658 pr_err("Cannot open %s: %s\n",
bf5a934a
DW		 7659 dev, strerror(errno));
7660 return 0;
7661 }
7662 /* Well, it is in use by someone, maybe an 'imsm' container. */
7663 cfd = open_container(fd);
20cbe8d2 7664 close(fd);
bf5a934a 7665 if (cfd < 0) {
bf5a934a 7666 if (verbose)
e7b84f9d 7667 pr_err("Cannot use %s: It is busy\n",
bf5a934a
DW		 7668 dev);
7669 return 0;
7670 }
4dd2df09 7671 sra = sysfs_read(cfd, NULL, GET_VERSION);
bf5a934a 7672 if (sra && sra->array.major_version == -1 &&
20cbe8d2
AW		 7673 strcmp(sra->text_version, "imsm") == 0)
7674 is_member = 1;
7675 sysfs_free(sra);
7676 if (is_member) {
bf5a934a
DW		 7677 /* This is a member of a imsm container. Load the container
7678 * and try to create a volume
7679 */
7680 struct intel_super *super;
7681
ec50f7b6 7682 if (load_super_imsm_all(st, cfd, (void **) &super, NULL, NULL, 1) == 0) {
bf5a934a 7683 st->sb = super;
4dd2df09 7684 strcpy(st->container_devnm, fd2devnm(cfd));
bf5a934a
DW		 7685 close(cfd);
7686 return validate_geometry_imsm_volume(st, level, layout,
7687 raiddisks, chunk,
af4348dd 7688 size, data_offset, dev,
ecbd9e81
N		 7689 freesize, 1)
7690 ? 1 : -1;
bf5a934a 7691 }
20cbe8d2 7692 }
bf5a934a 7693
20cbe8d2 7694 if (verbose)
e7b84f9d 7695 pr_err("failed container membership check\n");
20cbe8d2
AW		 7696
7697 close(cfd);
7698 return 0;
bf5a934a 7699}
0bd16cf2 7700
30f58b22 7701static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk)
0bd16cf2
DJ		 7702{
7703 struct intel_super *super = st->sb;
7704
30f58b22
DW		 7705 if (level && *level == UnSet)
7706 *level = LEVEL_CONTAINER;
7707
7708 if (level && layout && *layout == UnSet)
7709 *layout = imsm_level_to_layout(*level);
0bd16cf2 7710
cd9d1ac7
DW		 7711 if (chunk && (*chunk == UnSet || *chunk == 0))
7712 *chunk = imsm_default_chunk(super->orom);
0bd16cf2
DJ		 7713}
7714
33414a01
DW		 7715static void handle_missing(struct intel_super *super, struct imsm_dev *dev);
7716
3364781b 7717static int kill_subarray_imsm(struct supertype *st, char *subarray_id)
33414a01 7718{
3364781b 7719 /* remove the subarray currently referenced by subarray_id */
33414a01
DW		 7720 __u8 i;
7721 struct intel_dev **dp;
7722 struct intel_super *super = st->sb;
3364781b 7723 __u8 current_vol = strtoul(subarray_id, NULL, 10);
33414a01
DW		 7724 struct imsm_super *mpb = super->anchor;
7725
3364781b 7726 if (mpb->num_raid_devs == 0)
33414a01 7727 return 2;
33414a01
DW		 7728
7729 /* block deletions that would change the uuid of active subarrays
7730 *
7731 * FIXME when immutable ids are available, but note that we'll
7732 * also need to fixup the invalidated/active subarray indexes in
7733 * mdstat
7734 */
7735 for (i = 0; i < mpb->num_raid_devs; i++) {
7736 char subarray[4];
7737
7738 if (i < current_vol)
7739 continue;
7740 sprintf(subarray, "%u", i);
4dd2df09 7741 if (is_subarray_active(subarray, st->devnm)) {
e7b84f9d
N		 7742 pr_err("deleting subarray-%d would change the UUID of active subarray-%d, aborting\n",
7743 current_vol, i);
33414a01
DW		 7744
7745 return 2;
7746 }
7747 }
7748
7749 if (st->update_tail) {
503975b9 7750 struct imsm_update_kill_array *u = xmalloc(sizeof(*u));
33414a01 7751
33414a01
DW		 7752 u->type = update_kill_array;
7753 u->dev_idx = current_vol;
7754 append_metadata_update(st, u, sizeof(*u));
7755
7756 return 0;
7757 }
7758
7759 for (dp = &super->devlist; *dp;)
7760 if ((*dp)->index == current_vol) {
7761 *dp = (*dp)->next;
7762 } else {
7763 handle_missing(super, (*dp)->dev);
7764 if ((*dp)->index > current_vol)
7765 (*dp)->index--;
7766 dp = &(*dp)->next;
7767 }
7768
7769 /* no more raid devices, all active components are now spares,
7770 * but of course failed are still failed
7771 */
7772 if (--mpb->num_raid_devs == 0) {
7773 struct dl *d;
7774
7775 for (d = super->disks; d; d = d->next)
a8619d23
AK		 7776 if (d->index > -2)
7777 mark_spare(d);
33414a01
DW		 7778 }
7779
7780 super->updates_pending++;
7781
7782 return 0;
7783}
aa534678 7784
19ad203e
JR		 7785static int get_rwh_policy_from_update(char *update)
7786{
7787 if (strcmp(update, "ppl") == 0)
7788 return RWH_MULTIPLE_DISTRIBUTED;
7789 else if (strcmp(update, "no-ppl") == 0)
7790 return RWH_MULTIPLE_OFF;
7791 else if (strcmp(update, "bitmap") == 0)
7792 return RWH_BITMAP;
7793 else if (strcmp(update, "no-bitmap") == 0)
7794 return RWH_OFF;
7795 return -1;
7796}
7797
a951a4f7 7798static int update_subarray_imsm(struct supertype *st, char *subarray,
fa56eddb 7799 char *update, struct mddev_ident *ident)
aa534678
DW		 7800{
7801 /* update the subarray currently referenced by ->current_vol */
7802 struct intel_super *super = st->sb;
7803 struct imsm_super *mpb = super->anchor;
7804
aa534678
DW		 7805 if (strcmp(update, "name") == 0) {
7806 char *name = ident->name;
a951a4f7
N		 7807 char *ep;
7808 int vol;
aa534678 7809
4dd2df09 7810 if (is_subarray_active(subarray, st->devnm)) {
e7b84f9d 7811 pr_err("Unable to update name of active subarray\n");
aa534678
DW		 7812 return 2;
7813 }
7814
7815 if (!check_name(super, name, 0))
7816 return 2;
7817
a951a4f7
N		 7818 vol = strtoul(subarray, &ep, 10);
7819 if (*ep != '\0' || vol >= super->anchor->num_raid_devs)
7820 return 2;
7821
aa534678 7822 if (st->update_tail) {
503975b9 7823 struct imsm_update_rename_array *u = xmalloc(sizeof(*u));
aa534678 7824
aa534678 7825 u->type = update_rename_array;
a951a4f7 7826 u->dev_idx = vol;
618f4e6d
XN		 7827 strncpy((char *) u->name, name, MAX_RAID_SERIAL_LEN);
7828 u->name[MAX_RAID_SERIAL_LEN-1] = '\0';
aa534678
DW		 7829 append_metadata_update(st, u, sizeof(*u));
7830 } else {
7831 struct imsm_dev *dev;
ebad3af2 7832 int i, namelen;
aa534678 7833
a951a4f7 7834 dev = get_imsm_dev(super, vol);
ebad3af2
JS		 7835 memset(dev->volume, '\0', MAX_RAID_SERIAL_LEN);
7836 namelen = min((int)strlen(name), MAX_RAID_SERIAL_LEN);
7837 memcpy(dev->volume, name, namelen);
aa534678
DW		 7838 for (i = 0; i < mpb->num_raid_devs; i++) {
7839 dev = get_imsm_dev(super, i);
7840 handle_missing(super, dev);
7841 }
7842 super->updates_pending++;
7843 }
19ad203e 7844 } else if (get_rwh_policy_from_update(update) != -1) {
e6e9dd3f
AP		 7845 int new_policy;
7846 char *ep;
7847 int vol = strtoul(subarray, &ep, 10);
7848
7849 if (*ep != '\0' || vol >= super->anchor->num_raid_devs)
7850 return 2;
7851
19ad203e 7852 new_policy = get_rwh_policy_from_update(update);
e6e9dd3f
AP		 7853
7854 if (st->update_tail) {
7855 struct imsm_update_rwh_policy *u = xmalloc(sizeof(*u));
7856
7857 u->type = update_rwh_policy;
7858 u->dev_idx = vol;
7859 u->new_policy = new_policy;
7860 append_metadata_update(st, u, sizeof(*u));
7861 } else {
7862 struct imsm_dev *dev;
7863
7864 dev = get_imsm_dev(super, vol);
7865 dev->rwh_policy = new_policy;
7866 super->updates_pending++;
7867 }
19ad203e
JR		 7868 if (new_policy == RWH_BITMAP)
7869 return write_init_bitmap_imsm_vol(st, vol);
aa534678
DW		 7870 } else
7871 return 2;
7872
7873 return 0;
7874}
bf5a934a 7875
28bce06f
AK		 7876static int is_gen_migration(struct imsm_dev *dev)
7877{
7534230b
AK		 7878 if (dev == NULL)
7879 return 0;
7880
28bce06f
AK		 7881 if (!dev->vol.migr_state)
7882 return 0;
7883
7884 if (migr_type(dev) == MIGR_GEN_MIGR)
7885 return 1;
7886
7887 return 0;
7888}
7889
1e5c6983
DW		 7890static int is_rebuilding(struct imsm_dev *dev)
7891{
7892 struct imsm_map *migr_map;
7893
7894 if (!dev->vol.migr_state)
7895 return 0;
7896
7897 if (migr_type(dev) != MIGR_REBUILD)
7898 return 0;
7899
238c0a71 7900 migr_map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 7901
7902 if (migr_map->map_state == IMSM_T_STATE_DEGRADED)
7903 return 1;
7904 else
7905 return 0;
7906}
7907
6ce1fbf1
AK		 7908static int is_initializing(struct imsm_dev *dev)
7909{
7910 struct imsm_map *migr_map;
7911
7912 if (!dev->vol.migr_state)
7913 return 0;
7914
7915 if (migr_type(dev) != MIGR_INIT)
7916 return 0;
7917
238c0a71 7918 migr_map = get_imsm_map(dev, MAP_1);
6ce1fbf1
AK		 7919
7920 if (migr_map->map_state == IMSM_T_STATE_UNINITIALIZED)
7921 return 1;
7922
7923 return 0;
6ce1fbf1
AK		 7924}
7925
c47b0ff6
AK		 7926static void update_recovery_start(struct intel_super *super,
7927 struct imsm_dev *dev,
7928 struct mdinfo *array)
1e5c6983
DW		 7929{
7930 struct mdinfo *rebuild = NULL;
7931 struct mdinfo *d;
7932 __u32 units;
7933
7934 if (!is_rebuilding(dev))
7935 return;
7936
7937 /* Find the rebuild target, but punt on the dual rebuild case */
7938 for (d = array->devs; d; d = d->next)
7939 if (d->recovery_start == 0) {
7940 if (rebuild)
7941 return;
7942 rebuild = d;
7943 }
7944
4363fd80
DW		 7945 if (!rebuild) {
7946 /* (?) none of the disks are marked with
7947 * IMSM_ORD_REBUILD, so assume they are missing and the
7948 * disk_ord_tbl was not correctly updated
7949 */
1ade5cc1 7950 dprintf("failed to locate out-of-sync disk\n");
4363fd80
DW		 7951 return;
7952 }
7953
4036e7ee 7954 units = vol_curr_migr_unit(dev);
c47b0ff6 7955 rebuild->recovery_start = units * blocks_per_migr_unit(super, dev);
1e5c6983
DW		 7956}
7957
276d77db 7958static int recover_backup_imsm(struct supertype *st, struct mdinfo *info);
1e5c6983 7959
00bbdbda 7960static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray)
cdddbdbc 7961{
4f5bc454
DW		 7962 /* Given a container loaded by load_super_imsm_all,
7963 * extract information about all the arrays into
7964 * an mdinfo tree.
00bbdbda 7965 * If 'subarray' is given, just extract info about that array.
4f5bc454
DW		 7966 *
7967 * For each imsm_dev create an mdinfo, fill it in,
7968 * then look for matching devices in super->disks
7969 * and create appropriate device mdinfo.
7970 */
7971 struct intel_super *super = st->sb;
949c47a0 7972 struct imsm_super *mpb = super->anchor;
4f5bc454 7973 struct mdinfo *rest = NULL;
00bbdbda 7974 unsigned int i;
81219e70 7975 int sb_errors = 0;
abef11a3
AK		 7976 struct dl *d;
7977 int spare_disks = 0;
b6180160 7978 int current_vol = super->current_vol;
cdddbdbc 7979
19482bcc
AK		 7980 /* do not assemble arrays when not all attributes are supported */
7981 if (imsm_check_attributes(mpb->attributes) == 0) {
81219e70 7982 sb_errors = 1;
7a862a02 7983 pr_err("Unsupported attributes in IMSM metadata.Arrays activation is blocked.\n");
19482bcc
AK		 7984 }
7985
abef11a3
AK		 7986 /* count spare devices, not used in maps
7987 */
7988 for (d = super->disks; d; d = d->next)
7989 if (d->index == -1)
7990 spare_disks++;
7991
4f5bc454 7992 for (i = 0; i < mpb->num_raid_devs; i++) {
00bbdbda
N		 7993 struct imsm_dev *dev;
7994 struct imsm_map *map;
86e3692b 7995 struct imsm_map *map2;
4f5bc454 7996 struct mdinfo *this;
a6482415 7997 int slot;
a6482415 7998 int chunk;
00bbdbda 7999 char *ep;
8b9cd157 8000 int level;
00bbdbda
N		 8001
8002 if (subarray &&
8003 (i != strtoul(subarray, &ep, 10) || *ep != '\0'))
8004 continue;
8005
8006 dev = get_imsm_dev(super, i);
238c0a71
AK		 8007 map = get_imsm_map(dev, MAP_0);
8008 map2 = get_imsm_map(dev, MAP_1);
8b9cd157 8009 level = get_imsm_raid_level(map);
4f5bc454 8010
1ce0101c
DW		 8011 /* do not publish arrays that are in the middle of an
8012 * unsupported migration
8013 */
8014 if (dev->vol.migr_state &&
28bce06f 8015 (migr_type(dev) == MIGR_STATE_CHANGE)) {
7a862a02 8016 pr_err("cannot assemble volume '%.16s': unsupported migration in progress\n",
1ce0101c
DW		 8017 dev->volume);
8018 continue;
8019 }
2db86302
LM		 8020 /* do not publish arrays that are not support by controller's
8021 * OROM/EFI
8022 */
1ce0101c 8023
503975b9 8024 this = xmalloc(sizeof(*this));
4f5bc454 8025
301406c9 8026 super->current_vol = i;
a5d85af7 8027 getinfo_super_imsm_volume(st, this, NULL);
9894ec0d 8028 this->next = rest;
a6482415 8029 chunk = __le16_to_cpu(map->blocks_per_strip) >> 1;
81219e70
LM		 8030 /* mdadm does not support all metadata features- set the bit in all arrays state */
8031 if (!validate_geometry_imsm_orom(super,
8b9cd157
MK		 8032 level, /* RAID level */
8033 imsm_level_to_layout(level),
81219e70 8034 map->num_members, /* raid disks */
fcc2c9da 8035 &chunk, imsm_dev_size(dev),
81219e70 8036 1 /* verbose */)) {
7a862a02 8037 pr_err("IMSM RAID geometry validation failed. Array %s activation is blocked.\n",
81219e70
LM		 8038 dev->volume);
8039 this->array.state |=
8040 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
8041 (1<<MD_SB_BLOCK_VOLUME);
8042 }
81219e70
LM		 8043
8044 /* if array has bad blocks, set suitable bit in all arrays state */
8045 if (sb_errors)
8046 this->array.state |=
8047 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
8048 (1<<MD_SB_BLOCK_VOLUME);
8049
4f5bc454 8050 for (slot = 0 ; slot < map->num_members; slot++) {
1e5c6983 8051 unsigned long long recovery_start;
4f5bc454
DW		 8052 struct mdinfo *info_d;
8053 struct dl *d;
8054 int idx;
9a1608e5 8055 int skip;
7eef0453 8056 __u32 ord;
8b9cd157 8057 int missing = 0;
4f5bc454 8058
9a1608e5 8059 skip = 0;
238c0a71
AK		 8060 idx = get_imsm_disk_idx(dev, slot, MAP_0);
8061 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
4f5bc454
DW		 8062 for (d = super->disks; d ; d = d->next)
8063 if (d->index == idx)
0fbd635c 8064 break;
4f5bc454 8065
1e5c6983 8066 recovery_start = MaxSector;
4f5bc454 8067 if (d == NULL)
9a1608e5 8068 skip = 1;
25ed7e59 8069 if (d && is_failed(&d->disk))
9a1608e5 8070 skip = 1;
8b9cd157 8071 if (!skip && (ord & IMSM_ORD_REBUILD))
1e5c6983 8072 recovery_start = 0;
1e93d0d1
BK		 8073 if (!(ord & IMSM_ORD_REBUILD))
8074 this->array.working_disks++;
1011e834 8075 /*
9a1608e5 8076 * if we skip some disks the array will be assmebled degraded;
1e5c6983
DW		 8077 * reset resync start to avoid a dirty-degraded
8078 * situation when performing the intial sync
9a1608e5 8079 */
8b9cd157
MK		 8080 if (skip)
8081 missing++;
8082
8083 if (!(dev->vol.dirty & RAIDVOL_DIRTY)) {
8084 if ((!able_to_resync(level, missing) ||
8085 recovery_start == 0))
8086 this->resync_start = MaxSector;
8087 } else {
8088 /*
8089 * FIXME handle dirty degraded
8090 */
8091 }
8092
9a1608e5
DW		 8093 if (skip)
8094 continue;
4f5bc454 8095
503975b9 8096 info_d = xcalloc(1, sizeof(*info_d));
4f5bc454
DW		 8097 info_d->next = this->devs;
8098 this->devs = info_d;
8099
4f5bc454
DW		 8100 info_d->disk.number = d->index;
8101 info_d->disk.major = d->major;
8102 info_d->disk.minor = d->minor;
8103 info_d->disk.raid_disk = slot;
1e5c6983 8104 info_d->recovery_start = recovery_start;
86e3692b
AK		 8105 if (map2) {
8106 if (slot < map2->num_members)
8107 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK		 8108 else
8109 this->array.spare_disks++;
86e3692b
AK		 8110 } else {
8111 if (slot < map->num_members)
8112 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK		 8113 else
8114 this->array.spare_disks++;
86e3692b 8115 }
4f5bc454
DW		 8116
8117 info_d->events = __le32_to_cpu(mpb->generation_num);
5551b113 8118 info_d->data_offset = pba_of_lba0(map);
44490938 8119 info_d->component_size = calc_component_size(map, dev);
06fb291a
PB		 8120
8121 if (map->raid_level == 5) {
2432ce9b
AP		 8122 info_d->ppl_sector = this->ppl_sector;
8123 info_d->ppl_size = this->ppl_size;
98e96bdb
AP		 8124 if (this->consistency_policy == CONSISTENCY_POLICY_PPL &&
8125 recovery_start == 0)
8126 this->resync_start = 0;
06fb291a 8127 }
b12796be 8128
5e46202e 8129 info_d->bb.supported = 1;
b12796be
TM		 8130 get_volume_badblocks(super->bbm_log, ord_to_idx(ord),
8131 info_d->data_offset,
8132 info_d->component_size,
8133 &info_d->bb);
4f5bc454 8134 }
1e5c6983 8135 /* now that the disk list is up-to-date fixup recovery_start */
c47b0ff6 8136 update_recovery_start(super, dev, this);
abef11a3 8137 this->array.spare_disks += spare_disks;
276d77db
AK		 8138
8139 /* check for reshape */
8140 if (this->reshape_active == 1)
8141 recover_backup_imsm(st, this);
9a1608e5 8142 rest = this;
4f5bc454
DW		 8143 }
8144
b6180160 8145 super->current_vol = current_vol;
4f5bc454 8146 return rest;
cdddbdbc
DW		 8147}
8148
3b451610
AK		 8149static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev,
8150 int failed, int look_in_map)
c2a1e7da 8151{
3b451610
AK		 8152 struct imsm_map *map;
8153
8154 map = get_imsm_map(dev, look_in_map);
c2a1e7da
DW		 8155
8156 if (!failed)
1011e834 8157 return map->map_state == IMSM_T_STATE_UNINITIALIZED ?
3393c6af 8158 IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL;
c2a1e7da
DW		 8159
8160 switch (get_imsm_raid_level(map)) {
8161 case 0:
8162 return IMSM_T_STATE_FAILED;
8163 break;
8164 case 1:
8165 if (failed < map->num_members)
8166 return IMSM_T_STATE_DEGRADED;
8167 else
8168 return IMSM_T_STATE_FAILED;
8169 break;
8170 case 10:
8171 {
8172 /**
c92a2527
DW		 8173 * check to see if any mirrors have failed, otherwise we
8174 * are degraded. Even numbered slots are mirrored on
8175 * slot+1
c2a1e7da 8176 */
c2a1e7da 8177 int i;
d9b420a5
N		 8178 /* gcc -Os complains that this is unused */
8179 int insync = insync;
c2a1e7da
DW		 8180
8181 for (i = 0; i < map->num_members; i++) {
238c0a71 8182 __u32 ord = get_imsm_ord_tbl_ent(dev, i, MAP_X);
c92a2527
DW		 8183 int idx = ord_to_idx(ord);
8184 struct imsm_disk *disk;
c2a1e7da 8185
c92a2527 8186 /* reset the potential in-sync count on even-numbered
1011e834 8187 * slots. num_copies is always 2 for imsm raid10
c92a2527
DW		 8188 */
8189 if ((i & 1) == 0)
8190 insync = 2;
c2a1e7da 8191
c92a2527 8192 disk = get_imsm_disk(super, idx);
25ed7e59 8193 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
c92a2527 8194 insync--;
c2a1e7da 8195
c92a2527
DW		 8196 /* no in-sync disks left in this mirror the
8197 * array has failed
8198 */
8199 if (insync == 0)
8200 return IMSM_T_STATE_FAILED;
c2a1e7da
DW		 8201 }
8202
8203 return IMSM_T_STATE_DEGRADED;
8204 }
8205 case 5:
8206 if (failed < 2)
8207 return IMSM_T_STATE_DEGRADED;
8208 else
8209 return IMSM_T_STATE_FAILED;
8210 break;
8211 default:
8212 break;
8213 }
8214
8215 return map->map_state;
8216}
8217
3b451610
AK		 8218static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev,
8219 int look_in_map)
c2a1e7da
DW		 8220{
8221 int i;
8222 int failed = 0;
8223 struct imsm_disk *disk;
d5985138
AK		 8224 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8225 struct imsm_map *prev = get_imsm_map(dev, MAP_1);
68fe4598 8226 struct imsm_map *map_for_loop;
0556e1a2
DW		 8227 __u32 ord;
8228 int idx;
d5985138 8229 int idx_1;
c2a1e7da 8230
0556e1a2
DW		 8231 /* at the beginning of migration we set IMSM_ORD_REBUILD on
8232 * disks that are being rebuilt. New failures are recorded to
8233 * map[0]. So we look through all the disks we started with and
8234 * see if any failures are still present, or if any new ones
8235 * have arrived
0556e1a2 8236 */
d5985138
AK		 8237 map_for_loop = map;
8238 if (prev && (map->num_members < prev->num_members))
8239 map_for_loop = prev;
68fe4598
LD		 8240
8241 for (i = 0; i < map_for_loop->num_members; i++) {
d5985138 8242 idx_1 = -255;
238c0a71
AK		 8243 /* when MAP_X is passed both maps failures are counted
8244 */
d5985138 8245 if (prev &&
089f9d79
JS		 8246 (look_in_map == MAP_1 || look_in_map == MAP_X) &&
8247 i < prev->num_members) {
d5985138
AK		 8248 ord = __le32_to_cpu(prev->disk_ord_tbl[i]);
8249 idx_1 = ord_to_idx(ord);
c2a1e7da 8250
d5985138
AK		 8251 disk = get_imsm_disk(super, idx_1);
8252 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
8253 failed++;
8254 }
089f9d79
JS		 8255 if ((look_in_map == MAP_0 || look_in_map == MAP_X) &&
8256 i < map->num_members) {
d5985138
AK		 8257 ord = __le32_to_cpu(map->disk_ord_tbl[i]);
8258 idx = ord_to_idx(ord);
8259
8260 if (idx != idx_1) {
8261 disk = get_imsm_disk(super, idx);
8262 if (!disk || is_failed(disk) ||
8263 ord & IMSM_ORD_REBUILD)
8264 failed++;
8265 }
8266 }
c2a1e7da
DW		 8267 }
8268
8269 return failed;
845dea95
NB		 8270}
8271
97b4d0e9
DW		 8272static int imsm_open_new(struct supertype *c, struct active_array *a,
8273 char *inst)
8274{
8275 struct intel_super *super = c->sb;
8276 struct imsm_super *mpb = super->anchor;
bbab0940 8277 struct imsm_update_prealloc_bb_mem u;
9587c373 8278
97b4d0e9 8279 if (atoi(inst) >= mpb->num_raid_devs) {
1ade5cc1 8280 pr_err("subarry index %d, out of range\n", atoi(inst));
97b4d0e9
DW		 8281 return -ENODEV;
8282 }
8283
8284 dprintf("imsm: open_new %s\n", inst);
8285 a->info.container_member = atoi(inst);
bbab0940
TM		 8286
8287 u.type = update_prealloc_badblocks_mem;
8288 imsm_update_metadata_locally(c, &u, sizeof(u));
8289
97b4d0e9
DW		 8290 return 0;
8291}
8292
0c046afd
DW		 8293static int is_resyncing(struct imsm_dev *dev)
8294{
8295 struct imsm_map *migr_map;
8296
8297 if (!dev->vol.migr_state)
8298 return 0;
8299
1484e727
DW		 8300 if (migr_type(dev) == MIGR_INIT ||
8301 migr_type(dev) == MIGR_REPAIR)
0c046afd
DW		 8302 return 1;
8303
4c9bc37b
AK		 8304 if (migr_type(dev) == MIGR_GEN_MIGR)
8305 return 0;
8306
238c0a71 8307 migr_map = get_imsm_map(dev, MAP_1);
0c046afd 8308
089f9d79
JS		 8309 if (migr_map->map_state == IMSM_T_STATE_NORMAL &&
8310 dev->vol.migr_type != MIGR_GEN_MIGR)
0c046afd
DW		 8311 return 1;
8312 else
8313 return 0;
8314}
8315
0556e1a2 8316/* return true if we recorded new information */
4c9e8c1e
TM		 8317static int mark_failure(struct intel_super *super,
8318 struct imsm_dev *dev, struct imsm_disk *disk, int idx)
47ee5a45 8319{
0556e1a2
DW		 8320 __u32 ord;
8321 int slot;
8322 struct imsm_map *map;
86c54047
DW		 8323 char buf[MAX_RAID_SERIAL_LEN+3];
8324 unsigned int len, shift = 0;
0556e1a2
DW		 8325
8326 /* new failures are always set in map[0] */
238c0a71 8327 map = get_imsm_map(dev, MAP_0);
0556e1a2
DW		 8328
8329 slot = get_imsm_disk_slot(map, idx);
8330 if (slot < 0)
8331 return 0;
8332
8333 ord = __le32_to_cpu(map->disk_ord_tbl[slot]);
25ed7e59 8334 if (is_failed(disk) && (ord & IMSM_ORD_REBUILD))
0556e1a2
DW		 8335 return 0;
8336
7d0c5e24
LD		 8337 memcpy(buf, disk->serial, MAX_RAID_SERIAL_LEN);
8338 buf[MAX_RAID_SERIAL_LEN] = '\000';
8339 strcat(buf, ":0");
86c54047
DW		 8340 if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN)
8341 shift = len - MAX_RAID_SERIAL_LEN + 1;
167d8bb8 8342 memcpy(disk->serial, &buf[shift], len + 1 - shift);
86c54047 8343
f2f27e63 8344 disk->status |= FAILED_DISK;
0556e1a2 8345 set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD);
17788994
AK		 8346 /* mark failures in second map if second map exists and this disk
8347 * in this slot.
8348 * This is valid for migration, initialization and rebuild
8349 */
8350 if (dev->vol.migr_state) {
238c0a71 8351 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
0a108d63
AK		 8352 int slot2 = get_imsm_disk_slot(map2, idx);
8353
089f9d79 8354 if (slot2 < map2->num_members && slot2 >= 0)
0a108d63 8355 set_imsm_ord_tbl_ent(map2, slot2,
1ace8403
AK		 8356 idx | IMSM_ORD_REBUILD);
8357 }
d7a1fda2
MT		 8358 if (map->failed_disk_num == 0xff ||
8359 (!is_rebuilding(dev) && map->failed_disk_num > slot))
0556e1a2 8360 map->failed_disk_num = slot;
4c9e8c1e
TM		 8361
8362 clear_disk_badblocks(super->bbm_log, ord_to_idx(ord));
8363
0556e1a2
DW		 8364 return 1;
8365}
8366
4c9e8c1e
TM		 8367static void mark_missing(struct intel_super *super,
8368 struct imsm_dev *dev, struct imsm_disk *disk, int idx)
0556e1a2 8369{
4c9e8c1e 8370 mark_failure(super, dev, disk, idx);
0556e1a2
DW		 8371
8372 if (disk->scsi_id == __cpu_to_le32(~(__u32)0))
8373 return;
8374
47ee5a45
DW		 8375 disk->scsi_id = __cpu_to_le32(~(__u32)0);
8376 memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1);
8377}
8378
33414a01
DW		 8379static void handle_missing(struct intel_super *super, struct imsm_dev *dev)
8380{
33414a01 8381 struct dl *dl;
33414a01
DW		 8382
8383 if (!super->missing)
8384 return;
33414a01 8385
79b68f1b
PC		 8386 /* When orom adds replacement for missing disk it does
8387 * not remove entry of missing disk, but just updates map with
8388 * new added disk. So it is not enough just to test if there is
8389 * any missing disk, we have to look if there are any failed disks
8390 * in map to stop migration */
8391
33414a01 8392 dprintf("imsm: mark missing\n");
3d59f0c0
AK		 8393 /* end process for initialization and rebuild only
8394 */
8395 if (is_gen_migration(dev) == 0) {
fb12a745 8396 int failed = imsm_count_failed(super, dev, MAP_0);
3d59f0c0 8397
fb12a745
TM		 8398 if (failed) {
8399 __u8 map_state;
8400 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8401 struct imsm_map *map1;
8402 int i, ord, ord_map1;
8403 int rebuilt = 1;
3d59f0c0 8404
fb12a745
TM		 8405 for (i = 0; i < map->num_members; i++) {
8406 ord = get_imsm_ord_tbl_ent(dev, i, MAP_0);
8407 if (!(ord & IMSM_ORD_REBUILD))
8408 continue;
8409
8410 map1 = get_imsm_map(dev, MAP_1);
8411 if (!map1)
8412 continue;
8413
8414 ord_map1 = __le32_to_cpu(map1->disk_ord_tbl[i]);
8415 if (ord_map1 & IMSM_ORD_REBUILD)
8416 rebuilt = 0;
8417 }
8418
8419 if (rebuilt) {
8420 map_state = imsm_check_degraded(super, dev,
8421 failed, MAP_0);
8422 end_migration(dev, super, map_state);
8423 }
8424 }
3d59f0c0 8425 }
33414a01 8426 for (dl = super->missing; dl; dl = dl->next)
4c9e8c1e 8427 mark_missing(super, dev, &dl->disk, dl->index);
33414a01
DW		 8428 super->updates_pending++;
8429}
8430
f3871fdc
AK		 8431static unsigned long long imsm_set_array_size(struct imsm_dev *dev,
8432 long long new_size)
70bdf0dc 8433{
70bdf0dc 8434 unsigned long long array_blocks;
9529d343
MD		 8435 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8436 int used_disks = imsm_num_data_members(map);
70bdf0dc
AK		 8437
8438 if (used_disks == 0) {
8439 /* when problems occures
8440 * return current array_blocks value
8441 */
fcc2c9da 8442 array_blocks = imsm_dev_size(dev);
70bdf0dc
AK		 8443
8444 return array_blocks;
8445 }
8446
8447 /* set array size in metadata
8448 */
9529d343 8449 if (new_size <= 0)
f3871fdc
AK		 8450 /* OLCE size change is caused by added disks
8451 */
44490938 8452 array_blocks = per_dev_array_size(map) * used_disks;
9529d343 8453 else
f3871fdc
AK		 8454 /* Online Volume Size Change
8455 * Using available free space
8456 */
8457 array_blocks = new_size;
70bdf0dc 8458
b53bfba6 8459 array_blocks = round_size_to_mb(array_blocks, used_disks);
fcc2c9da 8460 set_imsm_dev_size(dev, array_blocks);
70bdf0dc
AK		 8461
8462 return array_blocks;
8463}
8464
28bce06f
AK		 8465static void imsm_set_disk(struct active_array *a, int n, int state);
8466
0e2d1a4e
AK		 8467static void imsm_progress_container_reshape(struct intel_super *super)
8468{
8469 /* if no device has a migr_state, but some device has a
8470 * different number of members than the previous device, start
8471 * changing the number of devices in this device to match
8472 * previous.
8473 */
8474 struct imsm_super *mpb = super->anchor;
8475 int prev_disks = -1;
8476 int i;
1dfaa380 8477 int copy_map_size;
0e2d1a4e
AK		 8478
8479 for (i = 0; i < mpb->num_raid_devs; i++) {
8480 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 8481 struct imsm_map *map = get_imsm_map(dev, MAP_0);
0e2d1a4e
AK		 8482 struct imsm_map *map2;
8483 int prev_num_members;
0e2d1a4e
AK		 8484
8485 if (dev->vol.migr_state)
8486 return;
8487
8488 if (prev_disks == -1)
8489 prev_disks = map->num_members;
8490 if (prev_disks == map->num_members)
8491 continue;
8492
8493 /* OK, this array needs to enter reshape mode.
8494 * i.e it needs a migr_state
8495 */
8496
1dfaa380 8497 copy_map_size = sizeof_imsm_map(map);
0e2d1a4e
AK		 8498 prev_num_members = map->num_members;
8499 map->num_members = prev_disks;
8500 dev->vol.migr_state = 1;
4036e7ee 8501 set_vol_curr_migr_unit(dev, 0);
ea672ee1 8502 set_migr_type(dev, MIGR_GEN_MIGR);
0e2d1a4e
AK		 8503 for (i = prev_num_members;
8504 i < map->num_members; i++)
8505 set_imsm_ord_tbl_ent(map, i, i);
238c0a71 8506 map2 = get_imsm_map(dev, MAP_1);
0e2d1a4e 8507 /* Copy the current map */
1dfaa380 8508 memcpy(map2, map, copy_map_size);
0e2d1a4e
AK		 8509 map2->num_members = prev_num_members;
8510
f3871fdc 8511 imsm_set_array_size(dev, -1);
51d83f5d 8512 super->clean_migration_record_by_mdmon = 1;
0e2d1a4e
AK		 8513 super->updates_pending++;
8514 }
8515}
8516
aad6f216 8517/* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild
0c046afd
DW		 8518 * states are handled in imsm_set_disk() with one exception, when a
8519 * resync is stopped due to a new failure this routine will set the
8520 * 'degraded' state for the array.
8521 */
01f157d7 8522static int imsm_set_array_state(struct active_array *a, int consistent)
a862209d
DW		 8523{
8524 int inst = a->info.container_member;
8525 struct intel_super *super = a->container->sb;
949c47a0 8526 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8527 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3b451610
AK		 8528 int failed = imsm_count_failed(super, dev, MAP_0);
8529 __u8 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
1e5c6983 8530 __u32 blocks_per_unit;
a862209d 8531
1af97990
AK		 8532 if (dev->vol.migr_state &&
8533 dev->vol.migr_type == MIGR_GEN_MIGR) {
8534 /* array state change is blocked due to reshape action
aad6f216
N		 8535 * We might need to
8536 * - abort the reshape (if last_checkpoint is 0 and action!= reshape)
8537 * - finish the reshape (if last_checkpoint is big and action != reshape)
4036e7ee 8538 * - update vol_curr_migr_unit
1af97990 8539 */
aad6f216 8540 if (a->curr_action == reshape) {
4036e7ee 8541 /* still reshaping, maybe update vol_curr_migr_unit */
633b5610 8542 goto mark_checkpoint;
aad6f216
N		 8543 } else {
8544 if (a->last_checkpoint == 0 && a->prev_action == reshape) {
8545 /* for some reason we aborted the reshape.
b66e591b
AK		 8546 *
8547 * disable automatic metadata rollback
8548 * user action is required to recover process
aad6f216 8549 */
b66e591b 8550 if (0) {
238c0a71
AK		 8551 struct imsm_map *map2 =
8552 get_imsm_map(dev, MAP_1);
8553 dev->vol.migr_state = 0;
8554 set_migr_type(dev, 0);
4036e7ee 8555 set_vol_curr_migr_unit(dev, 0);
238c0a71
AK		 8556 memcpy(map, map2,
8557 sizeof_imsm_map(map2));
8558 super->updates_pending++;
b66e591b 8559 }
aad6f216
N		 8560 }
8561 if (a->last_checkpoint >= a->info.component_size) {
8562 unsigned long long array_blocks;
8563 int used_disks;
e154ced3 8564 struct mdinfo *mdi;
aad6f216 8565
9529d343 8566 used_disks = imsm_num_data_members(map);
d55adef9
AK		 8567 if (used_disks > 0) {
8568 array_blocks =
44490938 8569 per_dev_array_size(map) *
d55adef9 8570 used_disks;
b53bfba6
TM		 8571 array_blocks =
8572 round_size_to_mb(array_blocks,
8573 used_disks);
d55adef9
AK		 8574 a->info.custom_array_size = array_blocks;
8575 /* encourage manager to update array
8576 * size
8577 */
e154ced3 8578
d55adef9 8579 a->check_reshape = 1;
633b5610 8580 }
e154ced3
AK		 8581 /* finalize online capacity expansion/reshape */
8582 for (mdi = a->info.devs; mdi; mdi = mdi->next)
8583 imsm_set_disk(a,
8584 mdi->disk.raid_disk,
8585 mdi->curr_state);
8586
0e2d1a4e 8587 imsm_progress_container_reshape(super);
e154ced3 8588 }
aad6f216 8589 }
1af97990
AK		 8590 }
8591
47ee5a45 8592 /* before we activate this array handle any missing disks */
33414a01
DW		 8593 if (consistent == 2)
8594 handle_missing(super, dev);
1e5c6983 8595
0c046afd 8596 if (consistent == 2 &&
b7941fd6 8597 (!is_resync_complete(&a->info) ||
0c046afd
DW		 8598 map_state != IMSM_T_STATE_NORMAL ||
8599 dev->vol.migr_state))
01f157d7 8600 consistent = 0;
272906ef 8601
b7941fd6 8602 if (is_resync_complete(&a->info)) {
0c046afd 8603 /* complete intialization / resync,
0556e1a2
DW		 8604 * recovery and interrupted recovery is completed in
8605 * ->set_disk
0c046afd
DW		 8606 */
8607 if (is_resyncing(dev)) {
8608 dprintf("imsm: mark resync done\n");
809da78e 8609 end_migration(dev, super, map_state);
115c3803 8610 super->updates_pending++;
484240d8 8611 a->last_checkpoint = 0;
115c3803 8612 }
b9172665
AK		 8613 } else if ((!is_resyncing(dev) && !failed) &&
8614 (imsm_reshape_blocks_arrays_changes(super) == 0)) {
0c046afd 8615 /* mark the start of the init process if nothing is failed */
b7941fd6 8616 dprintf("imsm: mark resync start\n");
1484e727 8617 if (map->map_state == IMSM_T_STATE_UNINITIALIZED)
8e59f3d8 8618 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT);
1484e727 8619 else
8e59f3d8 8620 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR);
3393c6af 8621 super->updates_pending++;
115c3803 8622 }
a862209d 8623
633b5610 8624mark_checkpoint:
5b83bacf
AK		 8625 /* skip checkpointing for general migration,
8626 * it is controlled in mdadm
8627 */
8628 if (is_gen_migration(dev))
8629 goto skip_mark_checkpoint;
8630
4036e7ee
MT		 8631 /* check if we can update vol_curr_migr_unit from resync_start,
8632 * recovery_start
8633 */
c47b0ff6 8634 blocks_per_unit = blocks_per_migr_unit(super, dev);
4f0a7acc 8635 if (blocks_per_unit) {
4036e7ee
MT		 8636 set_vol_curr_migr_unit(dev,
8637 a->last_checkpoint / blocks_per_unit);
8638 dprintf("imsm: mark checkpoint (%llu)\n",
8639 vol_curr_migr_unit(dev));
8640 super->updates_pending++;
1e5c6983 8641 }
f8f603f1 8642
5b83bacf 8643skip_mark_checkpoint:
3393c6af 8644 /* mark dirty / clean */
2432ce9b
AP		 8645 if (((dev->vol.dirty & RAIDVOL_DIRTY) && consistent) ||
8646 (!(dev->vol.dirty & RAIDVOL_DIRTY) && !consistent)) {
b7941fd6 8647 dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty");
2432ce9b
AP		 8648 if (consistent) {
8649 dev->vol.dirty = RAIDVOL_CLEAN;
8650 } else {
8651 dev->vol.dirty = RAIDVOL_DIRTY;
c2462068
PB		 8652 if (dev->rwh_policy == RWH_DISTRIBUTED ||
8653 dev->rwh_policy == RWH_MULTIPLE_DISTRIBUTED)
2432ce9b
AP		 8654 dev->vol.dirty |= RAIDVOL_DSRECORD_VALID;
8655 }
a862209d
DW		 8656 super->updates_pending++;
8657 }
28bce06f 8658
01f157d7 8659 return consistent;
a862209d
DW		 8660}
8661
6f50473f
TM		 8662static int imsm_disk_slot_to_ord(struct active_array *a, int slot)
8663{
8664 int inst = a->info.container_member;
8665 struct intel_super *super = a->container->sb;
8666 struct imsm_dev *dev = get_imsm_dev(super, inst);
8667 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8668
8669 if (slot > map->num_members) {
8670 pr_err("imsm: imsm_disk_slot_to_ord %d out of range 0..%d\n",
8671 slot, map->num_members - 1);
8672 return -1;
8673 }
8674
8675 if (slot < 0)
8676 return -1;
8677
8678 return get_imsm_ord_tbl_ent(dev, slot, MAP_0);
8679}
8680
8d45d196 8681static void imsm_set_disk(struct active_array *a, int n, int state)
845dea95 8682{
8d45d196
DW		 8683 int inst = a->info.container_member;
8684 struct intel_super *super = a->container->sb;
949c47a0 8685 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8686 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8d45d196 8687 struct imsm_disk *disk;
7ce05701
LD		 8688 struct mdinfo *mdi;
8689 int recovery_not_finished = 0;
0c046afd 8690 int failed;
6f50473f 8691 int ord;
0c046afd 8692 __u8 map_state;
fb12a745
TM		 8693 int rebuild_done = 0;
8694 int i;
8d45d196 8695
fb12a745 8696 ord = get_imsm_ord_tbl_ent(dev, n, MAP_X);
6f50473f 8697 if (ord < 0)
8d45d196
DW		 8698 return;
8699
4e6e574a 8700 dprintf("imsm: set_disk %d:%x\n", n, state);
b10b37b8 8701 disk = get_imsm_disk(super, ord_to_idx(ord));
8d45d196 8702
5802a811 8703 /* check for new failures */
ae7d61e3 8704 if (disk && (state & DS_FAULTY)) {
4c9e8c1e 8705 if (mark_failure(super, dev, disk, ord_to_idx(ord)))
0556e1a2 8706 super->updates_pending++;
8d45d196 8707 }
47ee5a45 8708
19859edc 8709 /* check if in_sync */
0556e1a2 8710 if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) {
238c0a71 8711 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
b10b37b8
DW		 8712
8713 set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord));
fb12a745 8714 rebuild_done = 1;
19859edc
DW		 8715 super->updates_pending++;
8716 }
8d45d196 8717
3b451610
AK		 8718 failed = imsm_count_failed(super, dev, MAP_0);
8719 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
5802a811 8720
0c046afd 8721 /* check if recovery complete, newly degraded, or failed */
94002678
AK		 8722 dprintf("imsm: Detected transition to state ");
8723 switch (map_state) {
8724 case IMSM_T_STATE_NORMAL: /* transition to normal state */
8725 dprintf("normal: ");
8726 if (is_rebuilding(dev)) {
1ade5cc1 8727 dprintf_cont("while rebuilding");
7ce05701
LD		 8728 /* check if recovery is really finished */
8729 for (mdi = a->info.devs; mdi ; mdi = mdi->next)
8730 if (mdi->recovery_start != MaxSector) {
8731 recovery_not_finished = 1;
8732 break;
8733 }
8734 if (recovery_not_finished) {
1ade5cc1
N		 8735 dprintf_cont("\n");
8736 dprintf("Rebuild has not finished yet, state not changed");
7ce05701
LD		 8737 if (a->last_checkpoint < mdi->recovery_start) {
8738 a->last_checkpoint = mdi->recovery_start;
8739 super->updates_pending++;
8740 }
8741 break;
8742 }
94002678 8743 end_migration(dev, super, map_state);
94002678
AK		 8744 map->failed_disk_num = ~0;
8745 super->updates_pending++;
8746 a->last_checkpoint = 0;
8747 break;
8748 }
8749 if (is_gen_migration(dev)) {
1ade5cc1 8750 dprintf_cont("while general migration");
bf2f0071 8751 if (a->last_checkpoint >= a->info.component_size)
809da78e 8752 end_migration(dev, super, map_state);
94002678
AK		 8753 else
8754 map->map_state = map_state;
28bce06f 8755 map->failed_disk_num = ~0;
94002678 8756 super->updates_pending++;
bf2f0071 8757 break;
94002678
AK		 8758 }
8759 break;
8760 case IMSM_T_STATE_DEGRADED: /* transition to degraded state */
1ade5cc1 8761 dprintf_cont("degraded: ");
089f9d79 8762 if (map->map_state != map_state && !dev->vol.migr_state) {
1ade5cc1 8763 dprintf_cont("mark degraded");
94002678
AK		 8764 map->map_state = map_state;
8765 super->updates_pending++;
8766 a->last_checkpoint = 0;
8767 break;
8768 }
8769 if (is_rebuilding(dev)) {
d7a1fda2 8770 dprintf_cont("while rebuilding ");
a4e96fd8
MT		 8771 if (state & DS_FAULTY) {
8772 dprintf_cont("removing failed drive ");
d7a1fda2
MT		 8773 if (n == map->failed_disk_num) {
8774 dprintf_cont("end migration");
8775 end_migration(dev, super, map_state);
a4e96fd8 8776 a->last_checkpoint = 0;
d7a1fda2 8777 } else {
a4e96fd8 8778 dprintf_cont("fail detected during rebuild, changing map state");
d7a1fda2
MT		 8779 map->map_state = map_state;
8780 }
94002678 8781 super->updates_pending++;
fb12a745
TM		 8782 }
8783
a4e96fd8
MT		 8784 if (!rebuild_done)
8785 break;
8786
fb12a745
TM		 8787 /* check if recovery is really finished */
8788 for (mdi = a->info.devs; mdi ; mdi = mdi->next)
8789 if (mdi->recovery_start != MaxSector) {
8790 recovery_not_finished = 1;
8791 break;
8792 }
8793 if (recovery_not_finished) {
8794 dprintf_cont("\n");
a4e96fd8 8795 dprintf_cont("Rebuild has not finished yet");
fb12a745
TM		 8796 if (a->last_checkpoint < mdi->recovery_start) {
8797 a->last_checkpoint =
8798 mdi->recovery_start;
8799 super->updates_pending++;
8800 }
8801 break;
94002678 8802 }
fb12a745
TM		 8803
8804 dprintf_cont(" Rebuild done, still degraded");
a4e96fd8
MT		 8805 end_migration(dev, super, map_state);
8806 a->last_checkpoint = 0;
8807 super->updates_pending++;
fb12a745
TM		 8808
8809 for (i = 0; i < map->num_members; i++) {
8810 int idx = get_imsm_ord_tbl_ent(dev, i, MAP_0);
8811
8812 if (idx & IMSM_ORD_REBUILD)
8813 map->failed_disk_num = i;
8814 }
8815 super->updates_pending++;
94002678
AK		 8816 break;
8817 }
8818 if (is_gen_migration(dev)) {
1ade5cc1 8819 dprintf_cont("while general migration");
bf2f0071 8820 if (a->last_checkpoint >= a->info.component_size)
809da78e 8821 end_migration(dev, super, map_state);
94002678
AK		 8822 else {
8823 map->map_state = map_state;
3b451610 8824 manage_second_map(super, dev);
94002678
AK		 8825 }
8826 super->updates_pending++;
bf2f0071 8827 break;
28bce06f 8828 }
6ce1fbf1 8829 if (is_initializing(dev)) {
1ade5cc1 8830 dprintf_cont("while initialization.");
6ce1fbf1
AK		 8831 map->map_state = map_state;
8832 super->updates_pending++;
8833 break;
8834 }
94002678
AK		 8835 break;
8836 case IMSM_T_STATE_FAILED: /* transition to failed state */
1ade5cc1 8837 dprintf_cont("failed: ");
94002678 8838 if (is_gen_migration(dev)) {
1ade5cc1 8839 dprintf_cont("while general migration");
94002678
AK		 8840 map->map_state = map_state;
8841 super->updates_pending++;
8842 break;
8843 }
8844 if (map->map_state != map_state) {
1ade5cc1 8845 dprintf_cont("mark failed");
94002678
AK		 8846 end_migration(dev, super, map_state);
8847 super->updates_pending++;
8848 a->last_checkpoint = 0;
8849 break;
8850 }
8851 break;
8852 default:
1ade5cc1 8853 dprintf_cont("state %i\n", map_state);
5802a811 8854 }
1ade5cc1 8855 dprintf_cont("\n");
845dea95
NB		 8856}
8857
f796af5d 8858static int store_imsm_mpb(int fd, struct imsm_super *mpb)
c2a1e7da 8859{
f796af5d 8860 void *buf = mpb;
c2a1e7da
DW		 8861 __u32 mpb_size = __le32_to_cpu(mpb->mpb_size);
8862 unsigned long long dsize;
8863 unsigned long long sectors;
f36a9ecd 8864 unsigned int sector_size;
c2a1e7da 8865
aec01630
JS		 8866 if (!get_dev_sector_size(fd, NULL, &sector_size))
8867 return 1;
c2a1e7da
DW		 8868 get_dev_size(fd, NULL, &dsize);
8869
f36a9ecd 8870 if (mpb_size > sector_size) {
272f648f 8871 /* -1 to account for anchor */
f36a9ecd 8872 sectors = mpb_sectors(mpb, sector_size) - 1;
c2a1e7da 8873
272f648f 8874 /* write the extended mpb to the sectors preceeding the anchor */
f36a9ecd
PB		 8875 if (lseek64(fd, dsize - (sector_size * (2 + sectors)),
8876 SEEK_SET) < 0)
272f648f 8877 return 1;
c2a1e7da 8878
f36a9ecd
PB		 8879 if ((unsigned long long)write(fd, buf + sector_size,
8880 sector_size * sectors) != sector_size * sectors)
272f648f
DW		 8881 return 1;
8882 }
c2a1e7da 8883
272f648f 8884 /* first block is stored on second to last sector of the disk */
f36a9ecd 8885 if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0)
c2a1e7da
DW		 8886 return 1;
8887
466070ad 8888 if ((unsigned int)write(fd, buf, sector_size) != sector_size)
c2a1e7da
DW		 8889 return 1;
8890
c2a1e7da
DW		 8891 return 0;
8892}
8893
2e735d19 8894static void imsm_sync_metadata(struct supertype *container)
845dea95 8895{
2e735d19 8896 struct intel_super *super = container->sb;
c2a1e7da 8897
1a64be56 8898 dprintf("sync metadata: %d\n", super->updates_pending);
c2a1e7da
DW		 8899 if (!super->updates_pending)
8900 return;
8901
36988a3d 8902 write_super_imsm(container, 0);
c2a1e7da
DW		 8903
8904 super->updates_pending = 0;
845dea95
NB		 8905}
8906
272906ef
DW		 8907static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a)
8908{
8909 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8910 int i = get_imsm_disk_idx(dev, idx, MAP_X);
272906ef
DW		 8911 struct dl *dl;
8912
8913 for (dl = super->disks; dl; dl = dl->next)
8914 if (dl->index == i)
8915 break;
8916
25ed7e59 8917 if (dl && is_failed(&dl->disk))
272906ef
DW		 8918 dl = NULL;
8919
8920 if (dl)
1ade5cc1 8921 dprintf("found %x:%x\n", dl->major, dl->minor);
272906ef
DW		 8922
8923 return dl;
8924}
8925
a20d2ba5 8926static struct dl *imsm_add_spare(struct intel_super *super, int slot,
8ba77d32
AK		 8927 struct active_array *a, int activate_new,
8928 struct mdinfo *additional_test_list)
272906ef
DW		 8929{
8930 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8931 int idx = get_imsm_disk_idx(dev, slot, MAP_X);
a20d2ba5
DW		 8932 struct imsm_super *mpb = super->anchor;
8933 struct imsm_map *map;
272906ef
DW		 8934 unsigned long long pos;
8935 struct mdinfo *d;
8936 struct extent *ex;
a20d2ba5 8937 int i, j;
272906ef 8938 int found;
569cc43f
DW		 8939 __u32 array_start = 0;
8940 __u32 array_end = 0;
272906ef 8941 struct dl *dl;
6c932028 8942 struct mdinfo *test_list;
272906ef
DW		 8943
8944 for (dl = super->disks; dl; dl = dl->next) {
8945 /* If in this array, skip */
8946 for (d = a->info.devs ; d ; d = d->next)
e553d2a4
DW		 8947 if (d->state_fd >= 0 &&
8948 d->disk.major == dl->major &&
272906ef 8949 d->disk.minor == dl->minor) {
8ba77d32
AK		 8950 dprintf("%x:%x already in array\n",
8951 dl->major, dl->minor);
272906ef
DW		 8952 break;
8953 }
8954 if (d)
8955 continue;
6c932028
AK		 8956 test_list = additional_test_list;
8957 while (test_list) {
8958 if (test_list->disk.major == dl->major &&
8959 test_list->disk.minor == dl->minor) {
8ba77d32
AK		 8960 dprintf("%x:%x already in additional test list\n",
8961 dl->major, dl->minor);
8962 break;
8963 }
6c932028 8964 test_list = test_list->next;
8ba77d32 8965 }
6c932028 8966 if (test_list)
8ba77d32 8967 continue;
272906ef 8968
e553d2a4 8969 /* skip in use or failed drives */
25ed7e59 8970 if (is_failed(&dl->disk) || idx == dl->index ||
df474657
DW		 8971 dl->index == -2) {
8972 dprintf("%x:%x status (failed: %d index: %d)\n",
25ed7e59 8973 dl->major, dl->minor, is_failed(&dl->disk), idx);
9a1608e5
DW		 8974 continue;
8975 }
8976
a20d2ba5
DW		 8977 /* skip pure spares when we are looking for partially
8978 * assimilated drives
8979 */
8980 if (dl->index == -1 && !activate_new)
8981 continue;
8982
f2cc4f7d
AO		 8983 if (!drive_validate_sector_size(super, dl))
8984 continue;
8985
272906ef 8986 /* Does this unused device have the requisite free space?
a20d2ba5 8987 * It needs to be able to cover all member volumes
272906ef 8988 */
05501181 8989 ex = get_extents(super, dl, 1);
272906ef
DW		 8990 if (!ex) {
8991 dprintf("cannot get extents\n");
8992 continue;
8993 }
a20d2ba5
DW		 8994 for (i = 0; i < mpb->num_raid_devs; i++) {
8995 dev = get_imsm_dev(super, i);
238c0a71 8996 map = get_imsm_map(dev, MAP_0);
272906ef 8997
a20d2ba5
DW		 8998 /* check if this disk is already a member of
8999 * this array
272906ef 9000 */
620b1713 9001 if (get_imsm_disk_slot(map, dl->index) >= 0)
a20d2ba5
DW		 9002 continue;
9003
9004 found = 0;
9005 j = 0;
9006 pos = 0;
5551b113 9007 array_start = pba_of_lba0(map);
329c8278 9008 array_end = array_start +
44490938 9009 per_dev_array_size(map) - 1;
a20d2ba5
DW		 9010
9011 do {
9012 /* check that we can start at pba_of_lba0 with
44490938 9013 * num_data_stripes*blocks_per_stripe of space
a20d2ba5 9014 */
329c8278 9015 if (array_start >= pos && array_end < ex[j].start) {
a20d2ba5
DW		 9016 found = 1;
9017 break;
9018 }
9019 pos = ex[j].start + ex[j].size;
9020 j++;
9021 } while (ex[j-1].size);
9022
9023 if (!found)
272906ef 9024 break;
a20d2ba5 9025 }
272906ef
DW		 9026
9027 free(ex);
a20d2ba5 9028 if (i < mpb->num_raid_devs) {
329c8278
DW		 9029 dprintf("%x:%x does not have %u to %u available\n",
9030 dl->major, dl->minor, array_start, array_end);
272906ef
DW		 9031 /* No room */
9032 continue;
a20d2ba5
DW		 9033 }
9034 return dl;
272906ef
DW		 9035 }
9036
9037 return dl;
9038}
9039
95d07a2c
LM		 9040static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed)
9041{
9042 struct imsm_dev *dev2;
9043 struct imsm_map *map;
9044 struct dl *idisk;
9045 int slot;
9046 int idx;
9047 __u8 state;
9048
9049 dev2 = get_imsm_dev(cont->sb, dev_idx);
9050 if (dev2) {
238c0a71 9051 state = imsm_check_degraded(cont->sb, dev2, failed, MAP_0);
95d07a2c 9052 if (state == IMSM_T_STATE_FAILED) {
238c0a71 9053 map = get_imsm_map(dev2, MAP_0);
95d07a2c
LM		 9054 if (!map)
9055 return 1;
9056 for (slot = 0; slot < map->num_members; slot++) {
9057 /*
9058 * Check if failed disks are deleted from intel
9059 * disk list or are marked to be deleted
9060 */
238c0a71 9061 idx = get_imsm_disk_idx(dev2, slot, MAP_X);
95d07a2c
LM		 9062 idisk = get_imsm_dl_disk(cont->sb, idx);
9063 /*
9064 * Do not rebuild the array if failed disks
9065 * from failed sub-array are not removed from
9066 * container.
9067 */
9068 if (idisk &&
9069 is_failed(&idisk->disk) &&
9070 (idisk->action != DISK_REMOVE))
9071 return 0;
9072 }
9073 }
9074 }
9075 return 1;
9076}
9077
88758e9d
DW		 9078static struct mdinfo *imsm_activate_spare(struct active_array *a,
9079 struct metadata_update **updates)
9080{
9081 /**
d23fe947
DW		 9082 * Find a device with unused free space and use it to replace a
9083 * failed/vacant region in an array. We replace failed regions one a
9084 * array at a time. The result is that a new spare disk will be added
9085 * to the first failed array and after the monitor has finished
9086 * propagating failures the remainder will be consumed.
88758e9d 9087 *
d23fe947
DW		 9088 * FIXME add a capability for mdmon to request spares from another
9089 * container.
88758e9d
DW		 9090 */
9091
9092 struct intel_super *super = a->container->sb;
88758e9d 9093 int inst = a->info.container_member;
949c47a0 9094 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 9095 struct imsm_map *map = get_imsm_map(dev, MAP_0);
88758e9d
DW		 9096 int failed = a->info.array.raid_disks;
9097 struct mdinfo *rv = NULL;
9098 struct mdinfo *d;
9099 struct mdinfo *di;
9100 struct metadata_update *mu;
9101 struct dl *dl;
9102 struct imsm_update_activate_spare *u;
9103 int num_spares = 0;
9104 int i;
95d07a2c 9105 int allowed;
88758e9d
DW		 9106
9107 for (d = a->info.devs ; d ; d = d->next) {
9108 if ((d->curr_state & DS_FAULTY) &&
9109 d->state_fd >= 0)
9110 /* wait for Removal to happen */
9111 return NULL;
9112 if (d->state_fd >= 0)
9113 failed--;
9114 }
9115
9116 dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n",
9117 inst, failed, a->info.array.raid_disks, a->info.array.level);
1af97990 9118
e2962bfc
AK		 9119 if (imsm_reshape_blocks_arrays_changes(super))
9120 return NULL;
1af97990 9121
fc8ca064
AK		 9122 /* Cannot activate another spare if rebuild is in progress already
9123 */
9124 if (is_rebuilding(dev)) {
7a862a02 9125 dprintf("imsm: No spare activation allowed. Rebuild in progress already.\n");
fc8ca064
AK		 9126 return NULL;
9127 }
9128
89c67882
AK		 9129 if (a->info.array.level == 4)
9130 /* No repair for takeovered array
9131 * imsm doesn't support raid4
9132 */
9133 return NULL;
9134
3b451610
AK		 9135 if (imsm_check_degraded(super, dev, failed, MAP_0) !=
9136 IMSM_T_STATE_DEGRADED)
88758e9d
DW		 9137 return NULL;
9138
83ca7d45
AP		 9139 if (get_imsm_map(dev, MAP_0)->map_state == IMSM_T_STATE_UNINITIALIZED) {
9140 dprintf("imsm: No spare activation allowed. Volume is not initialized.\n");
9141 return NULL;
9142 }
9143
95d07a2c
LM		 9144 /*
9145 * If there are any failed disks check state of the other volume.
9146 * Block rebuild if the another one is failed until failed disks
9147 * are removed from container.
9148 */
9149 if (failed) {
7a862a02 9150 dprintf("found failed disks in %.*s, check if there anotherfailed sub-array.\n",
c4acd1e5 9151 MAX_RAID_SERIAL_LEN, dev->volume);
95d07a2c
LM		 9152 /* check if states of the other volumes allow for rebuild */
9153 for (i = 0; i < super->anchor->num_raid_devs; i++) {
9154 if (i != inst) {
9155 allowed = imsm_rebuild_allowed(a->container,
9156 i, failed);
9157 if (!allowed)
9158 return NULL;
9159 }
9160 }
9161 }
9162
88758e9d 9163 /* For each slot, if it is not working, find a spare */
88758e9d
DW		 9164 for (i = 0; i < a->info.array.raid_disks; i++) {
9165 for (d = a->info.devs ; d ; d = d->next)
9166 if (d->disk.raid_disk == i)
9167 break;
9168 dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0);
9169 if (d && (d->state_fd >= 0))
9170 continue;
9171
272906ef 9172 /*
a20d2ba5
DW		 9173 * OK, this device needs recovery. Try to re-add the
9174 * previous occupant of this slot, if this fails see if
9175 * we can continue the assimilation of a spare that was
9176 * partially assimilated, finally try to activate a new
9177 * spare.
272906ef
DW		 9178 */
9179 dl = imsm_readd(super, i, a);
9180 if (!dl)
b303fe21 9181 dl = imsm_add_spare(super, i, a, 0, rv);
a20d2ba5 9182 if (!dl)
b303fe21 9183 dl = imsm_add_spare(super, i, a, 1, rv);
272906ef
DW		 9184 if (!dl)
9185 continue;
1011e834 9186
272906ef 9187 /* found a usable disk with enough space */
503975b9 9188 di = xcalloc(1, sizeof(*di));
272906ef
DW		 9189
9190 /* dl->index will be -1 in the case we are activating a
9191 * pristine spare. imsm_process_update() will create a
9192 * new index in this case. Once a disk is found to be
9193 * failed in all member arrays it is kicked from the
9194 * metadata
9195 */
9196 di->disk.number = dl->index;
d23fe947 9197
272906ef
DW		 9198 /* (ab)use di->devs to store a pointer to the device
9199 * we chose
9200 */
9201 di->devs = (struct mdinfo *) dl;
9202
9203 di->disk.raid_disk = i;
9204 di->disk.major = dl->major;
9205 di->disk.minor = dl->minor;
9206 di->disk.state = 0;
d23534e4 9207 di->recovery_start = 0;
5551b113 9208 di->data_offset = pba_of_lba0(map);
272906ef
DW		 9209 di->component_size = a->info.component_size;
9210 di->container_member = inst;
5e46202e 9211 di->bb.supported = 1;
2c8890e9 9212 if (a->info.consistency_policy == CONSISTENCY_POLICY_PPL) {
2432ce9b 9213 di->ppl_sector = get_ppl_sector(super, inst);
c2462068 9214 di->ppl_size = MULTIPLE_PPL_AREA_SIZE_IMSM >> 9;
2432ce9b 9215 }
148acb7b 9216 super->random = random32();
272906ef
DW		 9217 di->next = rv;
9218 rv = di;
9219 num_spares++;
9220 dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor,
9221 i, di->data_offset);
88758e9d
DW		 9222 }
9223
9224 if (!rv)
9225 /* No spares found */
9226 return rv;
9227 /* Now 'rv' has a list of devices to return.
9228 * Create a metadata_update record to update the
9229 * disk_ord_tbl for the array
9230 */
503975b9 9231 mu = xmalloc(sizeof(*mu));
1011e834 9232 mu->buf = xcalloc(num_spares,
503975b9 9233 sizeof(struct imsm_update_activate_spare));
88758e9d 9234 mu->space = NULL;
cb23f1f4 9235 mu->space_list = NULL;
88758e9d
DW		 9236 mu->len = sizeof(struct imsm_update_activate_spare) * num_spares;
9237 mu->next = *updates;
9238 u = (struct imsm_update_activate_spare *) mu->buf;
9239
9240 for (di = rv ; di ; di = di->next) {
9241 u->type = update_activate_spare;
d23fe947
DW		 9242 u->dl = (struct dl *) di->devs;
9243 di->devs = NULL;
88758e9d
DW		 9244 u->slot = di->disk.raid_disk;
9245 u->array = inst;
9246 u->next = u + 1;
9247 u++;
9248 }
9249 (u-1)->next = NULL;
9250 *updates = mu;
9251
9252 return rv;
9253}
9254
54c2c1ea 9255static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u)
8273f55e 9256{
54c2c1ea 9257 struct imsm_dev *dev = get_imsm_dev(super, idx);
238c0a71
AK		 9258 struct imsm_map *map = get_imsm_map(dev, MAP_0);
9259 struct imsm_map *new_map = get_imsm_map(&u->dev, MAP_0);
54c2c1ea
DW		 9260 struct disk_info *inf = get_disk_info(u);
9261 struct imsm_disk *disk;
8273f55e
DW		 9262 int i;
9263 int j;
8273f55e 9264
54c2c1ea 9265 for (i = 0; i < map->num_members; i++) {
238c0a71 9266 disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, MAP_X));
54c2c1ea
DW		 9267 for (j = 0; j < new_map->num_members; j++)
9268 if (serialcmp(disk->serial, inf[j].serial) == 0)
8273f55e
DW		 9269 return 1;
9270 }
9271
9272 return 0;
9273}
9274
1a64be56
LM		 9275static struct dl *get_disk_super(struct intel_super *super, int major, int minor)
9276{
594dc1b8
JS		 9277 struct dl *dl;
9278
1a64be56 9279 for (dl = super->disks; dl; dl = dl->next)
089f9d79 9280 if (dl->major == major && dl->minor == minor)
1a64be56
LM		 9281 return dl;
9282 return NULL;
9283}
9284
9285static int remove_disk_super(struct intel_super *super, int major, int minor)
9286{
594dc1b8 9287 struct dl *prev;
1a64be56
LM		 9288 struct dl *dl;
9289
9290 prev = NULL;
9291 for (dl = super->disks; dl; dl = dl->next) {
089f9d79 9292 if (dl->major == major && dl->minor == minor) {
1a64be56
LM		 9293 /* remove */
9294 if (prev)
9295 prev->next = dl->next;
9296 else
9297 super->disks = dl->next;
9298 dl->next = NULL;
9299 __free_imsm_disk(dl);
1ade5cc1 9300 dprintf("removed %x:%x\n", major, minor);
1a64be56
LM		 9301 break;
9302 }
9303 prev = dl;
9304 }
9305 return 0;
9306}
9307
f21e18ca 9308static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index);
ae6aad82 9309
1a64be56
LM		 9310static int add_remove_disk_update(struct intel_super *super)
9311{
9312 int check_degraded = 0;
594dc1b8
JS		 9313 struct dl *disk;
9314
1a64be56
LM		 9315 /* add/remove some spares to/from the metadata/contrainer */
9316 while (super->disk_mgmt_list) {
9317 struct dl *disk_cfg;
9318
9319 disk_cfg = super->disk_mgmt_list;
9320 super->disk_mgmt_list = disk_cfg->next;
9321 disk_cfg->next = NULL;
9322
9323 if (disk_cfg->action == DISK_ADD) {
9324 disk_cfg->next = super->disks;
9325 super->disks = disk_cfg;
9326 check_degraded = 1;
1ade5cc1
N		 9327 dprintf("added %x:%x\n",
9328 disk_cfg->major, disk_cfg->minor);
1a64be56
LM		 9329 } else if (disk_cfg->action == DISK_REMOVE) {
9330 dprintf("Disk remove action processed: %x.%x\n",
9331 disk_cfg->major, disk_cfg->minor);
9332 disk = get_disk_super(super,
9333 disk_cfg->major,
9334 disk_cfg->minor);
9335 if (disk) {
9336 /* store action status */
9337 disk->action = DISK_REMOVE;
9338 /* remove spare disks only */
9339 if (disk->index == -1) {
9340 remove_disk_super(super,
9341 disk_cfg->major,
9342 disk_cfg->minor);
91c97c54
MT		 9343 } else {
9344 disk_cfg->fd = disk->fd;
9345 disk->fd = -1;
1a64be56
LM		 9346 }
9347 }
9348 /* release allocate disk structure */
9349 __free_imsm_disk(disk_cfg);
9350 }
9351 }
9352 return check_degraded;
9353}
9354
a29911da
PC		 9355static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u,
9356 struct intel_super *super,
9357 void ***space_list)
9358{
9359 struct intel_dev *id;
9360 void **tofree = NULL;
9361 int ret_val = 0;
9362
1ade5cc1 9363 dprintf("(enter)\n");
089f9d79 9364 if (u->subdev < 0 || u->subdev > 1) {
a29911da
PC		 9365 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
9366 return ret_val;
9367 }
089f9d79 9368 if (space_list == NULL || *space_list == NULL) {
a29911da
PC		 9369 dprintf("imsm: Error: Memory is not allocated\n");
9370 return ret_val;
9371 }
9372
9373 for (id = super->devlist ; id; id = id->next) {
9374 if (id->index == (unsigned)u->subdev) {
9375 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
9376 struct imsm_map *map;
9377 struct imsm_dev *new_dev =
9378 (struct imsm_dev *)*space_list;
238c0a71 9379 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
a29911da
PC		 9380 int to_state;
9381 struct dl *new_disk;
9382
9383 if (new_dev == NULL)
9384 return ret_val;
9385 *space_list = **space_list;
9386 memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0));
238c0a71 9387 map = get_imsm_map(new_dev, MAP_0);
a29911da
PC		 9388 if (migr_map) {
9389 dprintf("imsm: Error: migration in progress");
9390 return ret_val;
9391 }
9392
9393 to_state = map->map_state;
9394 if ((u->new_level == 5) && (map->raid_level == 0)) {
9395 map->num_members++;
9396 /* this should not happen */
9397 if (u->new_disks[0] < 0) {
9398 map->failed_disk_num =
9399 map->num_members - 1;
9400 to_state = IMSM_T_STATE_DEGRADED;
9401 } else
9402 to_state = IMSM_T_STATE_NORMAL;
9403 }
8e59f3d8 9404 migrate(new_dev, super, to_state, MIGR_GEN_MIGR);
a29911da
PC		 9405 if (u->new_level > -1)
9406 map->raid_level = u->new_level;
238c0a71 9407 migr_map = get_imsm_map(new_dev, MAP_1);
a29911da
PC		 9408 if ((u->new_level == 5) &&
9409 (migr_map->raid_level == 0)) {
9410 int ord = map->num_members - 1;
9411 migr_map->num_members--;
9412 if (u->new_disks[0] < 0)
9413 ord |= IMSM_ORD_REBUILD;
9414 set_imsm_ord_tbl_ent(map,
9415 map->num_members - 1,
9416 ord);
9417 }
9418 id->dev = new_dev;
9419 tofree = (void **)dev;
9420
4bba0439
PC		 9421 /* update chunk size
9422 */
06fb291a
PB		 9423 if (u->new_chunksize > 0) {
9424 unsigned long long num_data_stripes;
9529d343
MD		 9425 struct imsm_map *dest_map =
9426 get_imsm_map(dev, MAP_0);
06fb291a 9427 int used_disks =
9529d343 9428 imsm_num_data_members(dest_map);
06fb291a
PB		 9429
9430 if (used_disks == 0)
9431 return ret_val;
9432
4bba0439
PC		 9433 map->blocks_per_strip =
9434 __cpu_to_le16(u->new_chunksize * 2);
06fb291a 9435 num_data_stripes =
fcc2c9da 9436 imsm_dev_size(dev) / used_disks;
06fb291a
PB		 9437 num_data_stripes /= map->blocks_per_strip;
9438 num_data_stripes /= map->num_domains;
9439 set_num_data_stripes(map, num_data_stripes);
9440 }
4bba0439 9441
44490938
MD		 9442 /* ensure blocks_per_member has valid value
9443 */
9444 set_blocks_per_member(map,
9445 per_dev_array_size(map) +
9446 NUM_BLOCKS_DIRTY_STRIPE_REGION);
9447
a29911da
PC		 9448 /* add disk
9449 */
089f9d79
JS		 9450 if (u->new_level != 5 || migr_map->raid_level != 0 ||
9451 migr_map->raid_level == map->raid_level)
a29911da
PC		 9452 goto skip_disk_add;
9453
9454 if (u->new_disks[0] >= 0) {
9455 /* use passes spare
9456 */
9457 new_disk = get_disk_super(super,
9458 major(u->new_disks[0]),
9459 minor(u->new_disks[0]));
7a862a02 9460 dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n",
a29911da
PC		 9461 major(u->new_disks[0]),
9462 minor(u->new_disks[0]),
9463 new_disk, new_disk->index);
9464 if (new_disk == NULL)
9465 goto error_disk_add;
9466
9467 new_disk->index = map->num_members - 1;
9468 /* slot to fill in autolayout
9469 */
9470 new_disk->raiddisk = new_disk->index;
9471 new_disk->disk.status |= CONFIGURED_DISK;
9472 new_disk->disk.status &= ~SPARE_DISK;
9473 } else
9474 goto error_disk_add;
9475
9476skip_disk_add:
9477 *tofree = *space_list;
9478 /* calculate new size
9479 */
f3871fdc 9480 imsm_set_array_size(new_dev, -1);
a29911da
PC		 9481
9482 ret_val = 1;
9483 }
9484 }
9485
9486 if (tofree)
9487 *space_list = tofree;
9488 return ret_val;
9489
9490error_disk_add:
9491 dprintf("Error: imsm: Cannot find disk.\n");
9492 return ret_val;
9493}
9494
f3871fdc
AK		 9495static int apply_size_change_update(struct imsm_update_size_change *u,
9496 struct intel_super *super)
9497{
9498 struct intel_dev *id;
9499 int ret_val = 0;
9500
1ade5cc1 9501 dprintf("(enter)\n");
089f9d79 9502 if (u->subdev < 0 || u->subdev > 1) {
f3871fdc
AK		 9503 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
9504 return ret_val;
9505 }
9506
9507 for (id = super->devlist ; id; id = id->next) {
9508 if (id->index == (unsigned)u->subdev) {
9509 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
9510 struct imsm_map *map = get_imsm_map(dev, MAP_0);
9529d343 9511 int used_disks = imsm_num_data_members(map);
f3871fdc 9512 unsigned long long blocks_per_member;
06fb291a 9513 unsigned long long num_data_stripes;
44490938
MD		 9514 unsigned long long new_size_per_disk;
9515
9516 if (used_disks == 0)
9517 return 0;
f3871fdc
AK		 9518
9519 /* calculate new size
9520 */
44490938
MD		 9521 new_size_per_disk = u->new_size / used_disks;
9522 blocks_per_member = new_size_per_disk +
9523 NUM_BLOCKS_DIRTY_STRIPE_REGION;
9524 num_data_stripes = new_size_per_disk /
06fb291a
PB		 9525 map->blocks_per_strip;
9526 num_data_stripes /= map->num_domains;
9527 dprintf("(size: %llu, blocks per member: %llu, num_data_stipes: %llu)\n",
44490938 9528 u->new_size, new_size_per_disk,
06fb291a 9529 num_data_stripes);
f3871fdc 9530 set_blocks_per_member(map, blocks_per_member);
06fb291a 9531 set_num_data_stripes(map, num_data_stripes);
f3871fdc
AK		 9532 imsm_set_array_size(dev, u->new_size);
9533
9534 ret_val = 1;
9535 break;
9536 }
9537 }
9538
9539 return ret_val;
9540}
9541
69d40de4
JR		 9542static int prepare_spare_to_activate(struct supertype *st,
9543 struct imsm_update_activate_spare *u)
9544{
9545 struct intel_super *super = st->sb;
9546 int prev_current_vol = super->current_vol;
9547 struct active_array *a;
9548 int ret = 1;
9549
9550 for (a = st->arrays; a; a = a->next)
9551 /*
9552 * Additional initialization (adding bitmap header, filling
9553 * the bitmap area with '1's to force initial rebuild for a whole
9554 * data-area) is required when adding the spare to the volume
9555 * with write-intent bitmap.
9556 */
9557 if (a->info.container_member == u->array &&
9558 a->info.consistency_policy == CONSISTENCY_POLICY_BITMAP) {
9559 struct dl *dl;
9560
9561 for (dl = super->disks; dl; dl = dl->next)
9562 if (dl == u->dl)
9563 break;
9564 if (!dl)
9565 break;
9566
9567 super->current_vol = u->array;
9568 if (st->ss->write_bitmap(st, dl->fd, NoUpdate))
9569 ret = 0;
9570 super->current_vol = prev_current_vol;
9571 }
9572 return ret;
9573}
9574
061d7da3 9575static int apply_update_activate_spare(struct imsm_update_activate_spare *u,
ca9de185 9576 struct intel_super *super,
061d7da3
LO		 9577 struct active_array *active_array)
9578{
9579 struct imsm_super *mpb = super->anchor;
9580 struct imsm_dev *dev = get_imsm_dev(super, u->array);
238c0a71 9581 struct imsm_map *map = get_imsm_map(dev, MAP_0);
061d7da3
LO		 9582 struct imsm_map *migr_map;
9583 struct active_array *a;
9584 struct imsm_disk *disk;
9585 __u8 to_state;
9586 struct dl *dl;
9587 unsigned int found;
9588 int failed;
5961eeec 9589 int victim;
061d7da3 9590 int i;
5961eeec 9591 int second_map_created = 0;
061d7da3 9592
5961eeec 9593 for (; u; u = u->next) {
238c0a71 9594 victim = get_imsm_disk_idx(dev, u->slot, MAP_X);
061d7da3 9595
5961eeec 9596 if (victim < 0)
9597 return 0;
061d7da3 9598
5961eeec 9599 for (dl = super->disks; dl; dl = dl->next)
9600 if (dl == u->dl)
9601 break;
061d7da3 9602
5961eeec 9603 if (!dl) {
7a862a02 9604 pr_err("error: imsm_activate_spare passed an unknown disk (index: %d)\n",
5961eeec 9605 u->dl->index);
9606 return 0;
9607 }
061d7da3 9608
5961eeec 9609 /* count failures (excluding rebuilds and the victim)
9610 * to determine map[0] state
9611 */
9612 failed = 0;
9613 for (i = 0; i < map->num_members; i++) {
9614 if (i == u->slot)
9615 continue;
9616 disk = get_imsm_disk(super,
238c0a71 9617 get_imsm_disk_idx(dev, i, MAP_X));
5961eeec 9618 if (!disk || is_failed(disk))
9619 failed++;
9620 }
061d7da3 9621
5961eeec 9622 /* adding a pristine spare, assign a new index */
9623 if (dl->index < 0) {
9624 dl->index = super->anchor->num_disks;
9625 super->anchor->num_disks++;
9626 }
9627 disk = &dl->disk;
9628 disk->status |= CONFIGURED_DISK;
9629 disk->status &= ~SPARE_DISK;
9630
9631 /* mark rebuild */
238c0a71 9632 to_state = imsm_check_degraded(super, dev, failed, MAP_0);
5961eeec 9633 if (!second_map_created) {
9634 second_map_created = 1;
9635 map->map_state = IMSM_T_STATE_DEGRADED;
9636 migrate(dev, super, to_state, MIGR_REBUILD);
9637 } else
9638 map->map_state = to_state;
238c0a71 9639 migr_map = get_imsm_map(dev, MAP_1);
5961eeec 9640 set_imsm_ord_tbl_ent(map, u->slot, dl->index);
9641 set_imsm_ord_tbl_ent(migr_map, u->slot,
9642 dl->index | IMSM_ORD_REBUILD);
9643
9644 /* update the family_num to mark a new container
9645 * generation, being careful to record the existing
9646 * family_num in orig_family_num to clean up after
9647 * earlier mdadm versions that neglected to set it.
9648 */
9649 if (mpb->orig_family_num == 0)
9650 mpb->orig_family_num = mpb->family_num;
9651 mpb->family_num += super->random;
9652
9653 /* count arrays using the victim in the metadata */
9654 found = 0;
9655 for (a = active_array; a ; a = a->next) {
9656 dev = get_imsm_dev(super, a->info.container_member);
238c0a71 9657 map = get_imsm_map(dev, MAP_0);
061d7da3 9658
5961eeec 9659 if (get_imsm_disk_slot(map, victim) >= 0)
9660 found++;
9661 }
061d7da3 9662
5961eeec 9663 /* delete the victim if it is no longer being
9664 * utilized anywhere
061d7da3 9665 */
5961eeec 9666 if (!found) {
9667 struct dl **dlp;
061d7da3 9668
5961eeec 9669 /* We know that 'manager' isn't touching anything,
9670 * so it is safe to delete
9671 */
9672 for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next)
061d7da3
LO		 9673 if ((*dlp)->index == victim)
9674 break;
5961eeec 9675
9676 /* victim may be on the missing list */
9677 if (!*dlp)
9678 for (dlp = &super->missing; *dlp;
9679 dlp = &(*dlp)->next)
9680 if ((*dlp)->index == victim)
9681 break;
9682 imsm_delete(super, dlp, victim);
9683 }
061d7da3
LO		 9684 }
9685
9686 return 1;
9687}
a29911da 9688
2e5dc010
N		 9689static int apply_reshape_container_disks_update(struct imsm_update_reshape *u,
9690 struct intel_super *super,
9691 void ***space_list)
9692{
9693 struct dl *new_disk;
9694 struct intel_dev *id;
9695 int i;
9696 int delta_disks = u->new_raid_disks - u->old_raid_disks;
ee4beede 9697 int disk_count = u->old_raid_disks;
2e5dc010
N		 9698 void **tofree = NULL;
9699 int devices_to_reshape = 1;
9700 struct imsm_super *mpb = super->anchor;
9701 int ret_val = 0;
d098291a 9702 unsigned int dev_id;
2e5dc010 9703
1ade5cc1 9704 dprintf("(enter)\n");
2e5dc010
N		 9705
9706 /* enable spares to use in array */
9707 for (i = 0; i < delta_disks; i++) {
9708 new_disk = get_disk_super(super,
9709 major(u->new_disks[i]),
9710 minor(u->new_disks[i]));
7a862a02 9711 dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n",
2e5dc010
N		 9712 major(u->new_disks[i]), minor(u->new_disks[i]),
9713 new_disk, new_disk->index);
089f9d79
JS		 9714 if (new_disk == NULL ||
9715 (new_disk->index >= 0 &&
9716 new_disk->index < u->old_raid_disks))
2e5dc010 9717 goto update_reshape_exit;
ee4beede 9718 new_disk->index = disk_count++;
2e5dc010
N		 9719 /* slot to fill in autolayout
9720 */
9721 new_disk->raiddisk = new_disk->index;
9722 new_disk->disk.status |=
9723 CONFIGURED_DISK;
9724 new_disk->disk.status &= ~SPARE_DISK;
9725 }
9726
ed7333bd
AK		 9727 dprintf("imsm: volume set mpb->num_raid_devs = %i\n",
9728 mpb->num_raid_devs);
2e5dc010
N		 9729 /* manage changes in volume
9730 */
d098291a 9731 for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) {
2e5dc010
N		 9732 void **sp = *space_list;
9733 struct imsm_dev *newdev;
9734 struct imsm_map *newmap, *oldmap;
9735
d098291a
AK		 9736 for (id = super->devlist ; id; id = id->next) {
9737 if (id->index == dev_id)
9738 break;
9739 }
9740 if (id == NULL)
9741 break;
2e5dc010
N		 9742 if (!sp)
9743 continue;
9744 *space_list = *sp;
9745 newdev = (void*)sp;
9746 /* Copy the dev, but not (all of) the map */
9747 memcpy(newdev, id->dev, sizeof(*newdev));
238c0a71
AK		 9748 oldmap = get_imsm_map(id->dev, MAP_0);
9749 newmap = get_imsm_map(newdev, MAP_0);
2e5dc010
N		 9750 /* Copy the current map */
9751 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
9752 /* update one device only
9753 */
9754 if (devices_to_reshape) {
ed7333bd
AK		 9755 dprintf("imsm: modifying subdev: %i\n",
9756 id->index);
2e5dc010
N		 9757 devices_to_reshape--;
9758 newdev->vol.migr_state = 1;
4036e7ee 9759 set_vol_curr_migr_unit(newdev, 0);
ea672ee1 9760 set_migr_type(newdev, MIGR_GEN_MIGR);
2e5dc010
N		 9761 newmap->num_members = u->new_raid_disks;
9762 for (i = 0; i < delta_disks; i++) {
9763 set_imsm_ord_tbl_ent(newmap,
9764 u->old_raid_disks + i,
9765 u->old_raid_disks + i);
9766 }
9767 /* New map is correct, now need to save old map
9768 */
238c0a71 9769 newmap = get_imsm_map(newdev, MAP_1);
2e5dc010
N		 9770 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
9771
f3871fdc 9772 imsm_set_array_size(newdev, -1);
2e5dc010
N		 9773 }
9774
9775 sp = (void **)id->dev;
9776 id->dev = newdev;
9777 *sp = tofree;
9778 tofree = sp;
8e59f3d8
AK		 9779
9780 /* Clear migration record */
9781 memset(super->migr_rec, 0, sizeof(struct migr_record));
2e5dc010 9782 }
819bc634
AK		 9783 if (tofree)
9784 *space_list = tofree;
2e5dc010
N		 9785 ret_val = 1;
9786
9787update_reshape_exit:
9788
9789 return ret_val;
9790}
9791
bb025c2f 9792static int apply_takeover_update(struct imsm_update_takeover *u,
8ca6df95
KW		 9793 struct intel_super *super,
9794 void ***space_list)
bb025c2f
KW		 9795{
9796 struct imsm_dev *dev = NULL;
8ca6df95
KW		 9797 struct intel_dev *dv;
9798 struct imsm_dev *dev_new;
bb025c2f
KW		 9799 struct imsm_map *map;
9800 struct dl *dm, *du;
8ca6df95 9801 int i;
bb025c2f
KW		 9802
9803 for (dv = super->devlist; dv; dv = dv->next)
9804 if (dv->index == (unsigned int)u->subarray) {
9805 dev = dv->dev;
9806 break;
9807 }
9808
9809 if (dev == NULL)
9810 return 0;
9811
238c0a71 9812 map = get_imsm_map(dev, MAP_0);
bb025c2f
KW		 9813
9814 if (u->direction == R10_TO_R0) {
06fb291a
PB		 9815 unsigned long long num_data_stripes;
9816
43d5ec18 9817 /* Number of failed disks must be half of initial disk number */
3b451610
AK		 9818 if (imsm_count_failed(super, dev, MAP_0) !=
9819 (map->num_members / 2))
43d5ec18
KW		 9820 return 0;
9821
bb025c2f
KW		 9822 /* iterate through devices to mark removed disks as spare */
9823 for (dm = super->disks; dm; dm = dm->next) {
9824 if (dm->disk.status & FAILED_DISK) {
9825 int idx = dm->index;
9826 /* update indexes on the disk list */
9827/* FIXME this loop-with-the-loop looks wrong, I'm not convinced
9828 the index values will end up being correct.... NB */
9829 for (du = super->disks; du; du = du->next)
9830 if (du->index > idx)
9831 du->index--;
9832 /* mark as spare disk */
a8619d23 9833 mark_spare(dm);
bb025c2f
KW		 9834 }
9835 }
bb025c2f
KW		 9836 /* update map */
9837 map->num_members = map->num_members / 2;
9838 map->map_state = IMSM_T_STATE_NORMAL;
9839 map->num_domains = 1;
9840 map->raid_level = 0;
9841 map->failed_disk_num = -1;
4a353e6e
RS		 9842 num_data_stripes = imsm_dev_size(dev) / 2;
9843 num_data_stripes /= map->blocks_per_strip;
9844 set_num_data_stripes(map, num_data_stripes);
bb025c2f
KW		 9845 }
9846
8ca6df95
KW		 9847 if (u->direction == R0_TO_R10) {
9848 void **space;
4a353e6e
RS		 9849 unsigned long long num_data_stripes;
9850
8ca6df95
KW		 9851 /* update slots in current disk list */
9852 for (dm = super->disks; dm; dm = dm->next) {
9853 if (dm->index >= 0)
9854 dm->index *= 2;
9855 }
9856 /* create new *missing* disks */
9857 for (i = 0; i < map->num_members; i++) {
9858 space = *space_list;
9859 if (!space)
9860 continue;
9861 *space_list = *space;
9862 du = (void *)space;
9863 memcpy(du, super->disks, sizeof(*du));
8ca6df95
KW		 9864 du->fd = -1;
9865 du->minor = 0;
9866 du->major = 0;
9867 du->index = (i * 2) + 1;
9868 sprintf((char *)du->disk.serial,
9869 " MISSING_%d", du->index);
9870 sprintf((char *)du->serial,
9871 "MISSING_%d", du->index);
9872 du->next = super->missing;
9873 super->missing = du;
9874 }
9875 /* create new dev and map */
9876 space = *space_list;
9877 if (!space)
9878 return 0;
9879 *space_list = *space;
9880 dev_new = (void *)space;
9881 memcpy(dev_new, dev, sizeof(*dev));
9882 /* update new map */
238c0a71 9883 map = get_imsm_map(dev_new, MAP_0);
8ca6df95 9884 map->num_members = map->num_members * 2;
1a2487c2 9885 map->map_state = IMSM_T_STATE_DEGRADED;
8ca6df95
KW		 9886 map->num_domains = 2;
9887 map->raid_level = 1;
4a353e6e
RS		 9888 num_data_stripes = imsm_dev_size(dev) / 2;
9889 num_data_stripes /= map->blocks_per_strip;
9890 num_data_stripes /= map->num_domains;
9891 set_num_data_stripes(map, num_data_stripes);
9892
8ca6df95
KW		 9893 /* replace dev<->dev_new */
9894 dv->dev = dev_new;
9895 }
bb025c2f
KW		 9896 /* update disk order table */
9897 for (du = super->disks; du; du = du->next)
9898 if (du->index >= 0)
9899 set_imsm_ord_tbl_ent(map, du->index, du->index);
8ca6df95 9900 for (du = super->missing; du; du = du->next)
1a2487c2
KW		 9901 if (du->index >= 0) {
9902 set_imsm_ord_tbl_ent(map, du->index, du->index);
4c9e8c1e 9903 mark_missing(super, dv->dev, &du->disk, du->index);
1a2487c2 9904 }
bb025c2f
KW		 9905
9906 return 1;
9907}
9908
e8319a19
DW		 9909static void imsm_process_update(struct supertype *st,
9910 struct metadata_update *update)
9911{
9912 /**
9913 * crack open the metadata_update envelope to find the update record
9914 * update can be one of:
d195167d
AK		 9915 * update_reshape_container_disks - all the arrays in the container
9916 * are being reshaped to have more devices. We need to mark
9917 * the arrays for general migration and convert selected spares
9918 * into active devices.
9919 * update_activate_spare - a spare device has replaced a failed
1011e834
N		 9920 * device in an array, update the disk_ord_tbl. If this disk is
9921 * present in all member arrays then also clear the SPARE_DISK
9922 * flag
d195167d
AK		 9923 * update_create_array
9924 * update_kill_array
9925 * update_rename_array
9926 * update_add_remove_disk
e8319a19
DW		 9927 */
9928 struct intel_super *super = st->sb;
4d7b1503 9929 struct imsm_super *mpb;
e8319a19
DW		 9930 enum imsm_update_type type = *(enum imsm_update_type *) update->buf;
9931
4d7b1503
DW		 9932 /* update requires a larger buf but the allocation failed */
9933 if (super->next_len && !super->next_buf) {
9934 super->next_len = 0;
9935 return;
9936 }
9937
9938 if (super->next_buf) {
9939 memcpy(super->next_buf, super->buf, super->len);
9940 free(super->buf);
9941 super->len = super->next_len;
9942 super->buf = super->next_buf;
9943
9944 super->next_len = 0;
9945 super->next_buf = NULL;
9946 }
9947
9948 mpb = super->anchor;
9949
e8319a19 9950 switch (type) {
0ec5d470
AK		 9951 case update_general_migration_checkpoint: {
9952 struct intel_dev *id;
9953 struct imsm_update_general_migration_checkpoint *u =
9954 (void *)update->buf;
9955
1ade5cc1 9956 dprintf("called for update_general_migration_checkpoint\n");
0ec5d470
AK		 9957
9958 /* find device under general migration */
9959 for (id = super->devlist ; id; id = id->next) {
9960 if (is_gen_migration(id->dev)) {
4036e7ee
MT		 9961 set_vol_curr_migr_unit(id->dev,
9962 u->curr_migr_unit);
0ec5d470
AK		 9963 super->updates_pending++;
9964 }
9965 }
9966 break;
9967 }
bb025c2f
KW		 9968 case update_takeover: {
9969 struct imsm_update_takeover *u = (void *)update->buf;
1a2487c2
KW		 9970 if (apply_takeover_update(u, super, &update->space_list)) {
9971 imsm_update_version_info(super);
bb025c2f 9972 super->updates_pending++;
1a2487c2 9973 }
bb025c2f
KW		 9974 break;
9975 }
9976
78b10e66 9977 case update_reshape_container_disks: {
d195167d 9978 struct imsm_update_reshape *u = (void *)update->buf;
2e5dc010
N		 9979 if (apply_reshape_container_disks_update(
9980 u, super, &update->space_list))
9981 super->updates_pending++;
78b10e66
N		 9982 break;
9983 }
48c5303a 9984 case update_reshape_migration: {
a29911da
PC		 9985 struct imsm_update_reshape_migration *u = (void *)update->buf;
9986 if (apply_reshape_migration_update(
9987 u, super, &update->space_list))
9988 super->updates_pending++;
48c5303a
PC		 9989 break;
9990 }
f3871fdc
AK		 9991 case update_size_change: {
9992 struct imsm_update_size_change *u = (void *)update->buf;
9993 if (apply_size_change_update(u, super))
9994 super->updates_pending++;
9995 break;
9996 }
e8319a19 9997 case update_activate_spare: {
1011e834 9998 struct imsm_update_activate_spare *u = (void *) update->buf;
69d40de4
JR		 9999
10000 if (prepare_spare_to_activate(st, u) &&
10001 apply_update_activate_spare(u, super, st->arrays))
061d7da3 10002 super->updates_pending++;
8273f55e
DW		 10003 break;
10004 }
10005 case update_create_array: {
10006 /* someone wants to create a new array, we need to be aware of
10007 * a few races/collisions:
10008 * 1/ 'Create' called by two separate instances of mdadm
10009 * 2/ 'Create' versus 'activate_spare': mdadm has chosen
10010 * devices that have since been assimilated via
10011 * activate_spare.
10012 * In the event this update can not be carried out mdadm will
10013 * (FIX ME) notice that its update did not take hold.
10014 */
10015 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 10016 struct intel_dev *dv;
8273f55e
DW		 10017 struct imsm_dev *dev;
10018 struct imsm_map *map, *new_map;
10019 unsigned long long start, end;
10020 unsigned long long new_start, new_end;
10021 int i;
54c2c1ea
DW		 10022 struct disk_info *inf;
10023 struct dl *dl;
8273f55e
DW		 10024
10025 /* handle racing creates: first come first serve */
10026 if (u->dev_idx < mpb->num_raid_devs) {
1ade5cc1 10027 dprintf("subarray %d already defined\n", u->dev_idx);
ba2de7ba 10028 goto create_error;
8273f55e
DW		 10029 }
10030
10031 /* check update is next in sequence */
10032 if (u->dev_idx != mpb->num_raid_devs) {
1ade5cc1
N		 10033 dprintf("can not create array %d expected index %d\n",
10034 u->dev_idx, mpb->num_raid_devs);
ba2de7ba 10035 goto create_error;
8273f55e
DW		 10036 }
10037
238c0a71 10038 new_map = get_imsm_map(&u->dev, MAP_0);
5551b113 10039 new_start = pba_of_lba0(new_map);
44490938 10040 new_end = new_start + per_dev_array_size(new_map);
54c2c1ea 10041 inf = get_disk_info(u);
8273f55e
DW		 10042
10043 /* handle activate_spare versus create race:
10044 * check to make sure that overlapping arrays do not include
10045 * overalpping disks
10046 */
10047 for (i = 0; i < mpb->num_raid_devs; i++) {
949c47a0 10048 dev = get_imsm_dev(super, i);
238c0a71 10049 map = get_imsm_map(dev, MAP_0);
5551b113 10050 start = pba_of_lba0(map);
44490938 10051 end = start + per_dev_array_size(map);
8273f55e
DW		 10052 if ((new_start >= start && new_start <= end) ||
10053 (start >= new_start && start <= new_end))
54c2c1ea
DW		 10054 /* overlap */;
10055 else
10056 continue;
10057
10058 if (disks_overlap(super, i, u)) {
1ade5cc1 10059 dprintf("arrays overlap\n");
ba2de7ba 10060 goto create_error;
8273f55e
DW		 10061 }
10062 }
8273f55e 10063
949c47a0
DW		 10064 /* check that prepare update was successful */
10065 if (!update->space) {
1ade5cc1 10066 dprintf("prepare update failed\n");
ba2de7ba 10067 goto create_error;
949c47a0
DW		 10068 }
10069
54c2c1ea
DW		 10070 /* check that all disks are still active before committing
10071 * changes. FIXME: could we instead handle this by creating a
10072 * degraded array? That's probably not what the user expects,
10073 * so better to drop this update on the floor.
10074 */
10075 for (i = 0; i < new_map->num_members; i++) {
10076 dl = serial_to_dl(inf[i].serial, super);
10077 if (!dl) {
1ade5cc1 10078 dprintf("disk disappeared\n");
ba2de7ba 10079 goto create_error;
54c2c1ea 10080 }
949c47a0
DW		 10081 }
10082
8273f55e 10083 super->updates_pending++;
54c2c1ea
DW		 10084
10085 /* convert spares to members and fixup ord_tbl */
10086 for (i = 0; i < new_map->num_members; i++) {
10087 dl = serial_to_dl(inf[i].serial, super);
10088 if (dl->index == -1) {
10089 dl->index = mpb->num_disks;
10090 mpb->num_disks++;
10091 dl->disk.status |= CONFIGURED_DISK;
10092 dl->disk.status &= ~SPARE_DISK;
10093 }
10094 set_imsm_ord_tbl_ent(new_map, i, dl->index);
10095 }
10096
ba2de7ba
DW		 10097 dv = update->space;
10098 dev = dv->dev;
949c47a0
DW		 10099 update->space = NULL;
10100 imsm_copy_dev(dev, &u->dev);
ba2de7ba
DW		 10101 dv->index = u->dev_idx;
10102 dv->next = super->devlist;
10103 super->devlist = dv;
8273f55e 10104 mpb->num_raid_devs++;
8273f55e 10105
4d1313e9 10106 imsm_update_version_info(super);
8273f55e 10107 break;
ba2de7ba
DW		 10108 create_error:
10109 /* mdmon knows how to release update->space, but not
10110 * ((struct intel_dev *) update->space)->dev
10111 */
10112 if (update->space) {
10113 dv = update->space;
10114 free(dv->dev);
10115 }
8273f55e 10116 break;
e8319a19 10117 }
33414a01
DW		 10118 case update_kill_array: {
10119 struct imsm_update_kill_array *u = (void *) update->buf;
10120 int victim = u->dev_idx;
10121 struct active_array *a;
10122 struct intel_dev **dp;
10123 struct imsm_dev *dev;
10124
10125 /* sanity check that we are not affecting the uuid of
10126 * active arrays, or deleting an active array
10127 *
10128 * FIXME when immutable ids are available, but note that
10129 * we'll also need to fixup the invalidated/active
10130 * subarray indexes in mdstat
10131 */
10132 for (a = st->arrays; a; a = a->next)
10133 if (a->info.container_member >= victim)
10134 break;
10135 /* by definition if mdmon is running at least one array
10136 * is active in the container, so checking
10137 * mpb->num_raid_devs is just extra paranoia
10138 */
10139 dev = get_imsm_dev(super, victim);
10140 if (a || !dev || mpb->num_raid_devs == 1) {
10141 dprintf("failed to delete subarray-%d\n", victim);
10142 break;
10143 }
10144
10145 for (dp = &super->devlist; *dp;)
f21e18ca 10146 if ((*dp)->index == (unsigned)super->current_vol) {
33414a01
DW		 10147 *dp = (*dp)->next;
10148 } else {
f21e18ca 10149 if ((*dp)->index > (unsigned)victim)
33414a01
DW		 10150 (*dp)->index--;
10151 dp = &(*dp)->next;
10152 }
10153 mpb->num_raid_devs--;
10154 super->updates_pending++;
10155 break;
10156 }
aa534678
DW		 10157 case update_rename_array: {
10158 struct imsm_update_rename_array *u = (void *) update->buf;
10159 char name[MAX_RAID_SERIAL_LEN+1];
10160 int target = u->dev_idx;
10161 struct active_array *a;
10162 struct imsm_dev *dev;
10163
10164 /* sanity check that we are not affecting the uuid of
10165 * an active array
10166 */
40659392 10167 memset(name, 0, sizeof(name));
aa534678
DW		 10168 snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name);
10169 name[MAX_RAID_SERIAL_LEN] = '\0';
10170 for (a = st->arrays; a; a = a->next)
10171 if (a->info.container_member == target)
10172 break;
10173 dev = get_imsm_dev(super, u->dev_idx);
10174 if (a || !dev || !check_name(super, name, 1)) {
10175 dprintf("failed to rename subarray-%d\n", target);
10176 break;
10177 }
10178
40659392 10179 memcpy(dev->volume, name, MAX_RAID_SERIAL_LEN);
aa534678
DW		 10180 super->updates_pending++;
10181 break;
10182 }
1a64be56 10183 case update_add_remove_disk: {
43dad3d6 10184 /* we may be able to repair some arrays if disks are
095b8088 10185 * being added, check the status of add_remove_disk
1a64be56
LM		 10186 * if discs has been added.
10187 */
10188 if (add_remove_disk_update(super)) {
43dad3d6 10189 struct active_array *a;
072b727f
DW		 10190
10191 super->updates_pending++;
1a64be56 10192 for (a = st->arrays; a; a = a->next)
43dad3d6
DW		 10193 a->check_degraded = 1;
10194 }
43dad3d6 10195 break;
e8319a19 10196 }
bbab0940
TM		 10197 case update_prealloc_badblocks_mem:
10198 break;
e6e9dd3f
AP		 10199 case update_rwh_policy: {
10200 struct imsm_update_rwh_policy *u = (void *)update->buf;
10201 int target = u->dev_idx;
10202 struct imsm_dev *dev = get_imsm_dev(super, target);
10203 if (!dev) {
10204 dprintf("could not find subarray-%d\n", target);
10205 break;
10206 }
10207
10208 if (dev->rwh_policy != u->new_policy) {
10209 dev->rwh_policy = u->new_policy;
10210 super->updates_pending++;
10211 }
10212 break;
10213 }
1a64be56 10214 default:
ebf3be99 10215 pr_err("error: unsupported process update type:(type: %d)\n", type);
1a64be56 10216 }
e8319a19 10217}
88758e9d 10218
bc0b9d34
PC		 10219static struct mdinfo *get_spares_for_grow(struct supertype *st);
10220
5fe6f031
N		 10221static int imsm_prepare_update(struct supertype *st,
10222 struct metadata_update *update)
8273f55e 10223{
949c47a0 10224 /**
4d7b1503
DW		 10225 * Allocate space to hold new disk entries, raid-device entries or a new
10226 * mpb if necessary. The manager synchronously waits for updates to
10227 * complete in the monitor, so new mpb buffers allocated here can be
10228 * integrated by the monitor thread without worrying about live pointers
10229 * in the manager thread.
8273f55e 10230 */
095b8088 10231 enum imsm_update_type type;
4d7b1503 10232 struct intel_super *super = st->sb;
f36a9ecd 10233 unsigned int sector_size = super->sector_size;
4d7b1503
DW		 10234 struct imsm_super *mpb = super->anchor;
10235 size_t buf_len;
10236 size_t len = 0;
949c47a0 10237
095b8088
N		 10238 if (update->len < (int)sizeof(type))
10239 return 0;
10240
10241 type = *(enum imsm_update_type *) update->buf;
10242
949c47a0 10243 switch (type) {
0ec5d470 10244 case update_general_migration_checkpoint:
095b8088
N		 10245 if (update->len < (int)sizeof(struct imsm_update_general_migration_checkpoint))
10246 return 0;
1ade5cc1 10247 dprintf("called for update_general_migration_checkpoint\n");
0ec5d470 10248 break;
abedf5fc
KW		 10249 case update_takeover: {
10250 struct imsm_update_takeover *u = (void *)update->buf;
095b8088
N		 10251 if (update->len < (int)sizeof(*u))
10252 return 0;
abedf5fc
KW		 10253 if (u->direction == R0_TO_R10) {
10254 void **tail = (void **)&update->space_list;
10255 struct imsm_dev *dev = get_imsm_dev(super, u->subarray);
238c0a71 10256 struct imsm_map *map = get_imsm_map(dev, MAP_0);
abedf5fc
KW		 10257 int num_members = map->num_members;
10258 void *space;
10259 int size, i;
abedf5fc
KW		 10260 /* allocate memory for added disks */
10261 for (i = 0; i < num_members; i++) {
10262 size = sizeof(struct dl);
503975b9 10263 space = xmalloc(size);
abedf5fc
KW		 10264 *tail = space;
10265 tail = space;
10266 *tail = NULL;
10267 }
10268 /* allocate memory for new device */
10269 size = sizeof_imsm_dev(super->devlist->dev, 0) +
10270 (num_members * sizeof(__u32));
503975b9
N		 10271 space = xmalloc(size);
10272 *tail = space;
10273 tail = space;
10274 *tail = NULL;
10275 len = disks_to_mpb_size(num_members * 2);
abedf5fc
KW		 10276 }
10277
10278 break;
10279 }
78b10e66 10280 case update_reshape_container_disks: {
d195167d
AK		 10281 /* Every raid device in the container is about to
10282 * gain some more devices, and we will enter a
10283 * reconfiguration.
10284 * So each 'imsm_map' will be bigger, and the imsm_vol
10285 * will now hold 2 of them.
10286 * Thus we need new 'struct imsm_dev' allocations sized
10287 * as sizeof_imsm_dev but with more devices in both maps.
10288 */
10289 struct imsm_update_reshape *u = (void *)update->buf;
10290 struct intel_dev *dl;
10291 void **space_tail = (void**)&update->space_list;
10292
095b8088
N		 10293 if (update->len < (int)sizeof(*u))
10294 return 0;
10295
1ade5cc1 10296 dprintf("for update_reshape\n");
d195167d
AK		 10297
10298 for (dl = super->devlist; dl; dl = dl->next) {
10299 int size = sizeof_imsm_dev(dl->dev, 1);
10300 void *s;
d677e0b8
AK		 10301 if (u->new_raid_disks > u->old_raid_disks)
10302 size += sizeof(__u32)*2*
10303 (u->new_raid_disks - u->old_raid_disks);
503975b9 10304 s = xmalloc(size);
d195167d
AK		 10305 *space_tail = s;
10306 space_tail = s;
10307 *space_tail = NULL;
10308 }
10309
10310 len = disks_to_mpb_size(u->new_raid_disks);
10311 dprintf("New anchor length is %llu\n", (unsigned long long)len);
78b10e66
N		 10312 break;
10313 }
48c5303a 10314 case update_reshape_migration: {
bc0b9d34
PC		 10315 /* for migration level 0->5 we need to add disks
10316 * so the same as for container operation we will copy
10317 * device to the bigger location.
10318 * in memory prepared device and new disk area are prepared
10319 * for usage in process update
10320 */
10321 struct imsm_update_reshape_migration *u = (void *)update->buf;
10322 struct intel_dev *id;
10323 void **space_tail = (void **)&update->space_list;
10324 int size;
10325 void *s;
10326 int current_level = -1;
10327
095b8088
N		 10328 if (update->len < (int)sizeof(*u))
10329 return 0;
10330
1ade5cc1 10331 dprintf("for update_reshape\n");
bc0b9d34
PC		 10332
10333 /* add space for bigger array in update
10334 */
10335 for (id = super->devlist; id; id = id->next) {
10336 if (id->index == (unsigned)u->subdev) {
10337 size = sizeof_imsm_dev(id->dev, 1);
10338 if (u->new_raid_disks > u->old_raid_disks)
10339 size += sizeof(__u32)*2*
10340 (u->new_raid_disks - u->old_raid_disks);
503975b9 10341 s = xmalloc(size);
bc0b9d34
PC		 10342 *space_tail = s;
10343 space_tail = s;
10344 *space_tail = NULL;
10345 break;
10346 }
10347 }
10348 if (update->space_list == NULL)
10349 break;
10350
10351 /* add space for disk in update
10352 */
10353 size = sizeof(struct dl);
503975b9 10354 s = xmalloc(size);
bc0b9d34
PC		 10355 *space_tail = s;
10356 space_tail = s;
10357 *space_tail = NULL;
10358
10359 /* add spare device to update
10360 */
10361 for (id = super->devlist ; id; id = id->next)
10362 if (id->index == (unsigned)u->subdev) {
10363 struct imsm_dev *dev;
10364 struct imsm_map *map;
10365
10366 dev = get_imsm_dev(super, u->subdev);
238c0a71 10367 map = get_imsm_map(dev, MAP_0);
bc0b9d34
PC		 10368 current_level = map->raid_level;
10369 break;
10370 }
089f9d79 10371 if (u->new_level == 5 && u->new_level != current_level) {
bc0b9d34
PC		 10372 struct mdinfo *spares;
10373
10374 spares = get_spares_for_grow(st);
10375 if (spares) {
10376 struct dl *dl;
10377 struct mdinfo *dev;
10378
10379 dev = spares->devs;
10380 if (dev) {
10381 u->new_disks[0] =
10382 makedev(dev->disk.major,
10383 dev->disk.minor);
10384 dl = get_disk_super(super,
10385 dev->disk.major,
10386 dev->disk.minor);
10387 dl->index = u->old_raid_disks;
10388 dev = dev->next;
10389 }
10390 sysfs_free(spares);
10391 }
10392 }
10393 len = disks_to_mpb_size(u->new_raid_disks);
10394 dprintf("New anchor length is %llu\n", (unsigned long long)len);
48c5303a
PC		 10395 break;
10396 }
f3871fdc 10397 case update_size_change: {
095b8088
N		 10398 if (update->len < (int)sizeof(struct imsm_update_size_change))
10399 return 0;
10400 break;
10401 }
10402 case update_activate_spare: {
10403 if (update->len < (int)sizeof(struct imsm_update_activate_spare))
10404 return 0;
f3871fdc
AK		 10405 break;
10406 }
949c47a0
DW		 10407 case update_create_array: {
10408 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 10409 struct intel_dev *dv;
54c2c1ea 10410 struct imsm_dev *dev = &u->dev;
238c0a71 10411 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW		 10412 struct dl *dl;
10413 struct disk_info *inf;
10414 int i;
10415 int activate = 0;
949c47a0 10416
095b8088
N		 10417 if (update->len < (int)sizeof(*u))
10418 return 0;
10419
54c2c1ea
DW		 10420 inf = get_disk_info(u);
10421 len = sizeof_imsm_dev(dev, 1);
ba2de7ba 10422 /* allocate a new super->devlist entry */
503975b9
N		 10423 dv = xmalloc(sizeof(*dv));
10424 dv->dev = xmalloc(len);
10425 update->space = dv;
949c47a0 10426
54c2c1ea
DW		 10427 /* count how many spares will be converted to members */
10428 for (i = 0; i < map->num_members; i++) {
10429 dl = serial_to_dl(inf[i].serial, super);
10430 if (!dl) {
10431 /* hmm maybe it failed?, nothing we can do about
10432 * it here
10433 */
10434 continue;
10435 }
10436 if (count_memberships(dl, super) == 0)
10437 activate++;
10438 }
10439 len += activate * sizeof(struct imsm_disk);
949c47a0 10440 break;
095b8088
N		 10441 }
10442 case update_kill_array: {
10443 if (update->len < (int)sizeof(struct imsm_update_kill_array))
10444 return 0;
949c47a0
DW		 10445 break;
10446 }
095b8088
N		 10447 case update_rename_array: {
10448 if (update->len < (int)sizeof(struct imsm_update_rename_array))
10449 return 0;
10450 break;
10451 }
10452 case update_add_remove_disk:
10453 /* no update->len needed */
10454 break;
bbab0940
TM		 10455 case update_prealloc_badblocks_mem:
10456 super->extra_space += sizeof(struct bbm_log) -
10457 get_imsm_bbm_log_size(super->bbm_log);
10458 break;
e6e9dd3f
AP		 10459 case update_rwh_policy: {
10460 if (update->len < (int)sizeof(struct imsm_update_rwh_policy))
10461 return 0;
10462 break;
10463 }
095b8088
N		 10464 default:
10465 return 0;
949c47a0 10466 }
8273f55e 10467
4d7b1503
DW		 10468 /* check if we need a larger metadata buffer */
10469 if (super->next_buf)
10470 buf_len = super->next_len;
10471 else
10472 buf_len = super->len;
10473
bbab0940 10474 if (__le32_to_cpu(mpb->mpb_size) + super->extra_space + len > buf_len) {
4d7b1503
DW		 10475 /* ok we need a larger buf than what is currently allocated
10476 * if this allocation fails process_update will notice that
10477 * ->next_len is set and ->next_buf is NULL
10478 */
bbab0940
TM		 10479 buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) +
10480 super->extra_space + len, sector_size);
4d7b1503
DW		 10481 if (super->next_buf)
10482 free(super->next_buf);
10483
10484 super->next_len = buf_len;
f36a9ecd 10485 if (posix_memalign(&super->next_buf, sector_size, buf_len) == 0)
1f45a8ad
DW		 10486 memset(super->next_buf, 0, buf_len);
10487 else
4d7b1503
DW		 10488 super->next_buf = NULL;
10489 }
5fe6f031 10490 return 1;
8273f55e
DW		 10491}
10492
ae6aad82 10493/* must be called while manager is quiesced */
f21e18ca 10494static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index)
ae6aad82
DW		 10495{
10496 struct imsm_super *mpb = super->anchor;
ae6aad82
DW		 10497 struct dl *iter;
10498 struct imsm_dev *dev;
10499 struct imsm_map *map;
4c9e8c1e 10500 unsigned int i, j, num_members;
fb12a745 10501 __u32 ord, ord_map0;
4c9e8c1e 10502 struct bbm_log *log = super->bbm_log;
ae6aad82 10503
1ade5cc1 10504 dprintf("deleting device[%d] from imsm_super\n", index);
ae6aad82
DW		 10505
10506 /* shift all indexes down one */
10507 for (iter = super->disks; iter; iter = iter->next)
f21e18ca 10508 if (iter->index > (int)index)
ae6aad82 10509 iter->index--;
47ee5a45 10510 for (iter = super->missing; iter; iter = iter->next)
f21e18ca 10511 if (iter->index > (int)index)
47ee5a45 10512 iter->index--;
ae6aad82
DW		 10513
10514 for (i = 0; i < mpb->num_raid_devs; i++) {
10515 dev = get_imsm_dev(super, i);
238c0a71 10516 map = get_imsm_map(dev, MAP_0);
24565c9a
DW		 10517 num_members = map->num_members;
10518 for (j = 0; j < num_members; j++) {
10519 /* update ord entries being careful not to propagate
10520 * ord-flags to the first map
10521 */
238c0a71 10522 ord = get_imsm_ord_tbl_ent(dev, j, MAP_X);
fb12a745 10523 ord_map0 = get_imsm_ord_tbl_ent(dev, j, MAP_0);
ae6aad82 10524
24565c9a
DW		 10525 if (ord_to_idx(ord) <= index)
10526 continue;
ae6aad82 10527
238c0a71 10528 map = get_imsm_map(dev, MAP_0);
fb12a745 10529 set_imsm_ord_tbl_ent(map, j, ord_map0 - 1);
238c0a71 10530 map = get_imsm_map(dev, MAP_1);
24565c9a
DW		 10531 if (map)
10532 set_imsm_ord_tbl_ent(map, j, ord - 1);
ae6aad82
DW		 10533 }
10534 }
10535
4c9e8c1e
TM		 10536 for (i = 0; i < log->entry_count; i++) {
10537 struct bbm_log_entry *entry = &log->marked_block_entries[i];
10538
10539 if (entry->disk_ordinal <= index)
10540 continue;
10541 entry->disk_ordinal--;
10542 }
10543
ae6aad82
DW		 10544 mpb->num_disks--;
10545 super->updates_pending++;
24565c9a
DW		 10546 if (*dlp) {
10547 struct dl *dl = *dlp;
10548
10549 *dlp = (*dlp)->next;
10550 __free_imsm_disk(dl);
10551 }
ae6aad82 10552}
9a717282 10553
9a717282
AK		 10554static int imsm_get_allowed_degradation(int level, int raid_disks,
10555 struct intel_super *super,
10556 struct imsm_dev *dev)
10557{
10558 switch (level) {
bf5cf7c7 10559 case 1:
9a717282
AK		 10560 case 10:{
10561 int ret_val = 0;
10562 struct imsm_map *map;
10563 int i;
10564
10565 ret_val = raid_disks/2;
10566 /* check map if all disks pairs not failed
10567 * in both maps
10568 */
238c0a71 10569 map = get_imsm_map(dev, MAP_0);
9a717282
AK		 10570 for (i = 0; i < ret_val; i++) {
10571 int degradation = 0;
10572 if (get_imsm_disk(super, i) == NULL)
10573 degradation++;
10574 if (get_imsm_disk(super, i + 1) == NULL)
10575 degradation++;
10576 if (degradation == 2)
10577 return 0;
10578 }
238c0a71 10579 map = get_imsm_map(dev, MAP_1);
9a717282
AK		 10580 /* if there is no second map
10581 * result can be returned
10582 */
10583 if (map == NULL)
10584 return ret_val;
10585 /* check degradation in second map
10586 */
10587 for (i = 0; i < ret_val; i++) {
10588 int degradation = 0;
10589 if (get_imsm_disk(super, i) == NULL)
10590 degradation++;
10591 if (get_imsm_disk(super, i + 1) == NULL)
10592 degradation++;
10593 if (degradation == 2)
10594 return 0;
10595 }
10596 return ret_val;
10597 }
10598 case 5:
10599 return 1;
10600 case 6:
10601 return 2;
10602 default:
10603 return 0;
10604 }
10605}
10606
d31ad643
PB		 10607/*******************************************************************************
10608 * Function: validate_container_imsm
10609 * Description: This routine validates container after assemble,
10610 * eg. if devices in container are under the same controller.
10611 *
10612 * Parameters:
10613 * info : linked list with info about devices used in array
10614 * Returns:
10615 * 1 : HBA mismatch
10616 * 0 : Success
10617 ******************************************************************************/
10618int validate_container_imsm(struct mdinfo *info)
10619{
6b781d33
AP		 10620 if (check_env("IMSM_NO_PLATFORM"))
10621 return 0;
d31ad643 10622
6b781d33
AP		 10623 struct sys_dev *idev;
10624 struct sys_dev *hba = NULL;
10625 struct sys_dev *intel_devices = find_intel_devices();
10626 char *dev_path = devt_to_devpath(makedev(info->disk.major,
7c798f87 10627 info->disk.minor), 1, NULL);
6b781d33
AP		 10628
10629 for (idev = intel_devices; idev; idev = idev->next) {
10630 if (dev_path && strstr(dev_path, idev->path)) {
10631 hba = idev;
10632 break;
d31ad643 10633 }
6b781d33
AP		 10634 }
10635 if (dev_path)
d31ad643
PB		 10636 free(dev_path);
10637
6b781d33
AP		 10638 if (!hba) {
10639 pr_err("WARNING - Cannot detect HBA for device %s!\n",
10640 devid2kname(makedev(info->disk.major, info->disk.minor)));
10641 return 1;
10642 }
10643
10644 const struct imsm_orom *orom = get_orom_by_device_id(hba->dev_id);
10645 struct mdinfo *dev;
10646
10647 for (dev = info->next; dev; dev = dev->next) {
7c798f87
MT		 10648 dev_path = devt_to_devpath(makedev(dev->disk.major,
10649 dev->disk.minor), 1, NULL);
6b781d33
AP		 10650
10651 struct sys_dev *hba2 = NULL;
10652 for (idev = intel_devices; idev; idev = idev->next) {
10653 if (dev_path && strstr(dev_path, idev->path)) {
10654 hba2 = idev;
10655 break;
d31ad643
PB		 10656 }
10657 }
6b781d33
AP		 10658 if (dev_path)
10659 free(dev_path);
10660
10661 const struct imsm_orom *orom2 = hba2 == NULL ? NULL :
10662 get_orom_by_device_id(hba2->dev_id);
10663
10664 if (hba2 && hba->type != hba2->type) {
10665 pr_err("WARNING - HBAs of devices do not match %s != %s\n",
10666 get_sys_dev_type(hba->type), get_sys_dev_type(hba2->type));
10667 return 1;
10668 }
10669
07cb1e57 10670 if (orom != orom2) {
6b781d33
AP		 10671 pr_err("WARNING - IMSM container assembled with disks under different HBAs!\n"
10672 " This operation is not supported and can lead to data loss.\n");
10673 return 1;
10674 }
10675
10676 if (!orom) {
10677 pr_err("WARNING - IMSM container assembled with disks under HBAs without IMSM platform support!\n"
10678 " This operation is not supported and can lead to data loss.\n");
10679 return 1;
10680 }
d31ad643 10681 }
6b781d33 10682
d31ad643
PB		 10683 return 0;
10684}
32141c17 10685
6f50473f
TM		 10686/*******************************************************************************
10687* Function: imsm_record_badblock
10688* Description: This routine stores new bad block record in BBM log
10689*
10690* Parameters:
10691* a : array containing a bad block
10692* slot : disk number containing a bad block
10693* sector : bad block sector
10694* length : bad block sectors range
10695* Returns:
10696* 1 : Success
10697* 0 : Error
10698******************************************************************************/
10699static int imsm_record_badblock(struct active_array *a, int slot,
10700 unsigned long long sector, int length)
10701{
10702 struct intel_super *super = a->container->sb;
10703 int ord;
10704 int ret;
10705
10706 ord = imsm_disk_slot_to_ord(a, slot);
10707 if (ord < 0)
10708 return 0;
10709
10710 ret = record_new_badblock(super->bbm_log, ord_to_idx(ord), sector,
10711 length);
10712 if (ret)
10713 super->updates_pending++;
10714
10715 return ret;
10716}
c07a5a4f
TM		 10717/*******************************************************************************
10718* Function: imsm_clear_badblock
10719* Description: This routine clears bad block record from BBM log
10720*
10721* Parameters:
10722* a : array containing a bad block
10723* slot : disk number containing a bad block
10724* sector : bad block sector
10725* length : bad block sectors range
10726* Returns:
10727* 1 : Success
10728* 0 : Error
10729******************************************************************************/
10730static int imsm_clear_badblock(struct active_array *a, int slot,
10731 unsigned long long sector, int length)
10732{
10733 struct intel_super *super = a->container->sb;
10734 int ord;
10735 int ret;
10736
10737 ord = imsm_disk_slot_to_ord(a, slot);
10738 if (ord < 0)
10739 return 0;
10740
10741 ret = clear_badblock(super->bbm_log, ord_to_idx(ord), sector, length);
10742 if (ret)
10743 super->updates_pending++;
10744
10745 return ret;
10746}
928f1424
TM		 10747/*******************************************************************************
10748* Function: imsm_get_badblocks
10749* Description: This routine get list of bad blocks for an array
10750*
10751* Parameters:
10752* a : array
10753* slot : disk number
10754* Returns:
10755* bb : structure containing bad blocks
10756* NULL : error
10757******************************************************************************/
10758static struct md_bb *imsm_get_badblocks(struct active_array *a, int slot)
10759{
10760 int inst = a->info.container_member;
10761 struct intel_super *super = a->container->sb;
10762 struct imsm_dev *dev = get_imsm_dev(super, inst);
10763 struct imsm_map *map = get_imsm_map(dev, MAP_0);
10764 int ord;
10765
10766 ord = imsm_disk_slot_to_ord(a, slot);
10767 if (ord < 0)
10768 return NULL;
10769
10770 get_volume_badblocks(super->bbm_log, ord_to_idx(ord), pba_of_lba0(map),
44490938 10771 per_dev_array_size(map), &super->bb);
928f1424
TM		 10772
10773 return &super->bb;
10774}
27156a57
TM		 10775/*******************************************************************************
10776* Function: examine_badblocks_imsm
10777* Description: Prints list of bad blocks on a disk to the standard output
10778*
10779* Parameters:
10780* st : metadata handler
10781* fd : open file descriptor for device
10782* devname : device name
10783* Returns:
10784* 0 : Success
10785* 1 : Error
10786******************************************************************************/
10787static int examine_badblocks_imsm(struct supertype *st, int fd, char *devname)
10788{
10789 struct intel_super *super = st->sb;
10790 struct bbm_log *log = super->bbm_log;
10791 struct dl *d = NULL;
10792 int any = 0;
10793
10794 for (d = super->disks; d ; d = d->next) {
10795 if (strcmp(d->devname, devname) == 0)
10796 break;
10797 }
10798
10799 if ((d == NULL) || (d->index < 0)) { /* serial mismatch probably */
10800 pr_err("%s doesn't appear to be part of a raid array\n",
10801 devname);
10802 return 1;
10803 }
10804
10805 if (log != NULL) {
10806 unsigned int i;
10807 struct bbm_log_entry *entry = &log->marked_block_entries[0];
10808
10809 for (i = 0; i < log->entry_count; i++) {
10810 if (entry[i].disk_ordinal == d->index) {
10811 unsigned long long sector = __le48_to_cpu(
10812 &entry[i].defective_block_start);
10813 int cnt = entry[i].marked_count + 1;
10814
10815 if (!any) {
10816 printf("Bad-blocks on %s:\n", devname);
10817 any = 1;
10818 }
10819
10820 printf("%20llu for %d sectors\n", sector, cnt);
10821 }
10822 }
10823 }
10824
10825 if (!any)
10826 printf("No bad-blocks list configured on %s\n", devname);
10827
10828 return 0;
10829}
687629c2
AK		 10830/*******************************************************************************
10831 * Function: init_migr_record_imsm
10832 * Description: Function inits imsm migration record
10833 * Parameters:
10834 * super : imsm internal array info
10835 * dev : device under migration
10836 * info : general array info to find the smallest device
10837 * Returns:
10838 * none
10839 ******************************************************************************/
10840void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev,
10841 struct mdinfo *info)
10842{
10843 struct intel_super *super = st->sb;
10844 struct migr_record *migr_rec = super->migr_rec;
10845 int new_data_disks;
10846 unsigned long long dsize, dev_sectors;
10847 long long unsigned min_dev_sectors = -1LLU;
238c0a71
AK		 10848 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
10849 struct imsm_map *map_src = get_imsm_map(dev, MAP_1);
687629c2 10850 unsigned long long num_migr_units;
3ef4403c 10851 unsigned long long array_blocks;
2f86fda3 10852 struct dl *dl_disk = NULL;
687629c2
AK		 10853
10854 memset(migr_rec, 0, sizeof(struct migr_record));
10855 migr_rec->family_num = __cpu_to_le32(super->anchor->family_num);
10856
10857 /* only ascending reshape supported now */
10858 migr_rec->ascending_migr = __cpu_to_le32(1);
10859
10860 migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE /
10861 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
e1742195
AK		 10862 migr_rec->dest_depth_per_unit *=
10863 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
9529d343 10864 new_data_disks = imsm_num_data_members(map_dest);
687629c2
AK		 10865 migr_rec->blocks_per_unit =
10866 __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks);
10867 migr_rec->dest_depth_per_unit =
10868 __cpu_to_le32(migr_rec->dest_depth_per_unit);
3ef4403c 10869 array_blocks = info->component_size * new_data_disks;
687629c2
AK		 10870 num_migr_units =
10871 array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit);
10872
10873 if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit))
10874 num_migr_units++;
9f421827 10875 set_num_migr_units(migr_rec, num_migr_units);
687629c2
AK		 10876
10877 migr_rec->post_migr_vol_cap = dev->size_low;
10878 migr_rec->post_migr_vol_cap_hi = dev->size_high;
10879
687629c2 10880 /* Find the smallest dev */
2f86fda3
MT		 10881 for (dl_disk = super->disks; dl_disk ; dl_disk = dl_disk->next) {
10882 /* ignore spares in container */
10883 if (dl_disk->index < 0)
687629c2 10884 continue;
2f86fda3 10885 get_dev_size(dl_disk->fd, NULL, &dsize);
687629c2
AK		 10886 dev_sectors = dsize / 512;
10887 if (dev_sectors < min_dev_sectors)
10888 min_dev_sectors = dev_sectors;
687629c2 10889 }
9f421827 10890 set_migr_chkp_area_pba(migr_rec, min_dev_sectors -
687629c2
AK		 10891 RAID_DISK_RESERVED_BLOCKS_IMSM_HI);
10892
10893 write_imsm_migr_rec(st);
10894
10895 return;
10896}
10897
10898/*******************************************************************************
10899 * Function: save_backup_imsm
10900 * Description: Function saves critical data stripes to Migration Copy Area
10901 * and updates the current migration unit status.
10902 * Use restore_stripes() to form a destination stripe,
10903 * and to write it to the Copy Area.
10904 * Parameters:
10905 * st : supertype information
aea93171 10906 * dev : imsm device that backup is saved for
687629c2
AK		 10907 * info : general array info
10908 * buf : input buffer
687629c2
AK		 10909 * length : length of data to backup (blocks_per_unit)
10910 * Returns:
10911 * 0 : success
10912 *, -1 : fail
10913 ******************************************************************************/
10914int save_backup_imsm(struct supertype *st,
10915 struct imsm_dev *dev,
10916 struct mdinfo *info,
10917 void *buf,
687629c2
AK		 10918 int length)
10919{
10920 int rv = -1;
10921 struct intel_super *super = st->sb;
594dc1b8
JS		 10922 unsigned long long *target_offsets;
10923 int *targets;
687629c2 10924 int i;
238c0a71 10925 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
687629c2 10926 int new_disks = map_dest->num_members;
ab724b98
AK		 10927 int dest_layout = 0;
10928 int dest_chunk;
d1877f69 10929 unsigned long long start;
9529d343 10930 int data_disks = imsm_num_data_members(map_dest);
687629c2 10931
503975b9 10932 targets = xmalloc(new_disks * sizeof(int));
687629c2 10933
2f86fda3
MT		 10934 for (i = 0; i < new_disks; i++) {
10935 struct dl *dl_disk = get_imsm_dl_disk(super, i);
10936
10937 targets[i] = dl_disk->fd;
10938 }
7e45b550 10939
503975b9 10940 target_offsets = xcalloc(new_disks, sizeof(unsigned long long));
687629c2 10941
d1877f69 10942 start = info->reshape_progress * 512;
687629c2 10943 for (i = 0; i < new_disks; i++) {
9f421827 10944 target_offsets[i] = migr_chkp_area_pba(super->migr_rec) * 512;
d1877f69
AK		 10945 /* move back copy area adderss, it will be moved forward
10946 * in restore_stripes() using start input variable
10947 */
10948 target_offsets[i] -= start/data_disks;
687629c2
AK		 10949 }
10950
68eb8bc6 10951 dest_layout = imsm_level_to_layout(map_dest->raid_level);
ab724b98
AK		 10952 dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512;
10953
687629c2
AK		 10954 if (restore_stripes(targets, /* list of dest devices */
10955 target_offsets, /* migration record offsets */
10956 new_disks,
ab724b98
AK		 10957 dest_chunk,
10958 map_dest->raid_level,
10959 dest_layout,
10960 -1, /* source backup file descriptor */
10961 0, /* input buf offset
10962 * always 0 buf is already offseted */
d1877f69 10963 start,
687629c2
AK		 10964 length,
10965 buf) != 0) {
e7b84f9d 10966 pr_err("Error restoring stripes\n");
687629c2
AK		 10967 goto abort;
10968 }
10969
10970 rv = 0;
10971
10972abort:
10973 if (targets) {
687629c2
AK		 10974 free(targets);
10975 }
10976 free(target_offsets);
10977
10978 return rv;
10979}
10980
10981/*******************************************************************************
10982 * Function: save_checkpoint_imsm
10983 * Description: Function called for current unit status update
10984 * in the migration record. It writes it to disk.
10985 * Parameters:
10986 * super : imsm internal array info
10987 * info : general array info
10988 * Returns:
10989 * 0: success
10990 * 1: failure
0228d92c
AK		 10991 * 2: failure, means no valid migration record
10992 * / no general migration in progress /
687629c2
AK		 10993 ******************************************************************************/
10994int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state)
10995{
10996 struct intel_super *super = st->sb;
f8b72ef5
AK		 10997 unsigned long long blocks_per_unit;
10998 unsigned long long curr_migr_unit;
10999
2f86fda3 11000 if (load_imsm_migr_rec(super) != 0) {
7a862a02 11001 dprintf("imsm: ERROR: Cannot read migration record for checkpoint save.\n");
2e062e82
AK		 11002 return 1;
11003 }
11004
f8b72ef5
AK		 11005 blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit);
11006 if (blocks_per_unit == 0) {
0228d92c
AK		 11007 dprintf("imsm: no migration in progress.\n");
11008 return 2;
687629c2 11009 }
f8b72ef5
AK		 11010 curr_migr_unit = info->reshape_progress / blocks_per_unit;
11011 /* check if array is alligned to copy area
11012 * if it is not alligned, add one to current migration unit value
11013 * this can happend on array reshape finish only
11014 */
11015 if (info->reshape_progress % blocks_per_unit)
11016 curr_migr_unit++;
687629c2 11017
9f421827 11018 set_current_migr_unit(super->migr_rec, curr_migr_unit);
687629c2 11019 super->migr_rec->rec_status = __cpu_to_le32(state);
9f421827
PB		 11020 set_migr_dest_1st_member_lba(super->migr_rec,
11021 super->migr_rec->dest_depth_per_unit * curr_migr_unit);
11022
687629c2 11023 if (write_imsm_migr_rec(st) < 0) {
7a862a02 11024 dprintf("imsm: Cannot write migration record outside backup area\n");
687629c2
AK		 11025 return 1;
11026 }
11027
11028 return 0;
11029}
11030
276d77db
AK		 11031/*******************************************************************************
11032 * Function: recover_backup_imsm
11033 * Description: Function recovers critical data from the Migration Copy Area
11034 * while assembling an array.
11035 * Parameters:
11036 * super : imsm internal array info
11037 * info : general array info
11038 * Returns:
11039 * 0 : success (or there is no data to recover)
11040 * 1 : fail
11041 ******************************************************************************/
11042int recover_backup_imsm(struct supertype *st, struct mdinfo *info)
11043{
11044 struct intel_super *super = st->sb;
11045 struct migr_record *migr_rec = super->migr_rec;
594dc1b8 11046 struct imsm_map *map_dest;
276d77db
AK		 11047 struct intel_dev *id = NULL;
11048 unsigned long long read_offset;
11049 unsigned long long write_offset;
11050 unsigned unit_len;
2f86fda3 11051 int new_disks, err;
276d77db
AK		 11052 char *buf = NULL;
11053 int retval = 1;
f36a9ecd 11054 unsigned int sector_size = super->sector_size;
4036e7ee
MT		 11055 unsigned long long curr_migr_unit = current_migr_unit(migr_rec);
11056 unsigned long long num_migr_units = get_num_migr_units(migr_rec);
276d77db 11057 char buffer[20];
6c3560c0 11058 int skipped_disks = 0;
2f86fda3 11059 struct dl *dl_disk;
276d77db
AK		 11060
11061 err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20);
11062 if (err < 1)
11063 return 1;
11064
11065 /* recover data only during assemblation */
11066 if (strncmp(buffer, "inactive", 8) != 0)
11067 return 0;
11068 /* no data to recover */
11069 if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL)
11070 return 0;
11071 if (curr_migr_unit >= num_migr_units)
11072 return 1;
11073
11074 /* find device during reshape */
11075 for (id = super->devlist; id; id = id->next)
11076 if (is_gen_migration(id->dev))
11077 break;
11078 if (id == NULL)
11079 return 1;
11080
238c0a71 11081 map_dest = get_imsm_map(id->dev, MAP_0);
276d77db
AK		 11082 new_disks = map_dest->num_members;
11083
9f421827 11084 read_offset = migr_chkp_area_pba(migr_rec) * 512;
276d77db 11085
9f421827 11086 write_offset = (migr_dest_1st_member_lba(migr_rec) +
5551b113 11087 pba_of_lba0(map_dest)) * 512;
276d77db
AK		 11088
11089 unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
f36a9ecd 11090 if (posix_memalign((void **)&buf, sector_size, unit_len) != 0)
276d77db 11091 goto abort;
276d77db 11092
2f86fda3
MT		 11093 for (dl_disk = super->disks; dl_disk; dl_disk = dl_disk->next) {
11094 if (dl_disk->index < 0)
11095 continue;
276d77db 11096
2f86fda3 11097 if (dl_disk->fd < 0) {
6c3560c0
AK		 11098 skipped_disks++;
11099 continue;
11100 }
2f86fda3 11101 if (lseek64(dl_disk->fd, read_offset, SEEK_SET) < 0) {
e7b84f9d
N		 11102 pr_err("Cannot seek to block: %s\n",
11103 strerror(errno));
137debce
AK		 11104 skipped_disks++;
11105 continue;
276d77db 11106 }
83b3de77 11107 if (read(dl_disk->fd, buf, unit_len) != (ssize_t)unit_len) {
e7b84f9d
N		 11108 pr_err("Cannot read copy area block: %s\n",
11109 strerror(errno));
137debce
AK		 11110 skipped_disks++;
11111 continue;
276d77db 11112 }
2f86fda3 11113 if (lseek64(dl_disk->fd, write_offset, SEEK_SET) < 0) {
e7b84f9d
N		 11114 pr_err("Cannot seek to block: %s\n",
11115 strerror(errno));
137debce
AK		 11116 skipped_disks++;
11117 continue;
276d77db 11118 }
83b3de77 11119 if (write(dl_disk->fd, buf, unit_len) != (ssize_t)unit_len) {
e7b84f9d
N		 11120 pr_err("Cannot restore block: %s\n",
11121 strerror(errno));
137debce
AK		 11122 skipped_disks++;
11123 continue;
276d77db
AK		 11124 }
11125 }
11126
137debce
AK		 11127 if (skipped_disks > imsm_get_allowed_degradation(info->new_level,
11128 new_disks,
11129 super,
11130 id->dev)) {
7a862a02 11131 pr_err("Cannot restore data from backup. Too many failed disks\n");
6c3560c0
AK		 11132 goto abort;
11133 }
11134
befb629b
AK		 11135 if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) {
11136 /* ignore error == 2, this can mean end of reshape here
11137 */
7a862a02 11138 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL) during restart\n");
befb629b 11139 } else
276d77db 11140 retval = 0;
276d77db
AK		 11141
11142abort:
276d77db
AK		 11143 free(buf);
11144 return retval;
11145}
11146
2cda7640
ML		 11147static char disk_by_path[] = "/dev/disk/by-path/";
11148
11149static const char *imsm_get_disk_controller_domain(const char *path)
11150{
2cda7640 11151 char disk_path[PATH_MAX];
96234762
LM		 11152 char *drv=NULL;
11153 struct stat st;
2cda7640 11154
6d8d290a 11155 strcpy(disk_path, disk_by_path);
96234762
LM		 11156 strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1);
11157 if (stat(disk_path, &st) == 0) {
11158 struct sys_dev* hba;
594dc1b8 11159 char *path;
96234762 11160
7c798f87 11161 path = devt_to_devpath(st.st_rdev, 1, NULL);
96234762
LM		 11162 if (path == NULL)
11163 return "unknown";
11164 hba = find_disk_attached_hba(-1, path);
11165 if (hba && hba->type == SYS_DEV_SAS)
11166 drv = "isci";
11167 else if (hba && hba->type == SYS_DEV_SATA)
11168 drv = "ahci";
c6839718
MT		 11169 else if (hba && hba->type == SYS_DEV_VMD)
11170 drv = "vmd";
11171 else if (hba && hba->type == SYS_DEV_NVME)
11172 drv = "nvme";
1011e834 11173 else
96234762
LM		 11174 drv = "unknown";
11175 dprintf("path: %s hba: %s attached: %s\n",
11176 path, (hba) ? hba->path : "NULL", drv);
11177 free(path);
2cda7640 11178 }
96234762 11179 return drv;
2cda7640
ML		 11180}
11181
4dd2df09 11182static char *imsm_find_array_devnm_by_subdev(int subdev, char *container)
78b10e66 11183{
4dd2df09 11184 static char devnm[32];
78b10e66
N		 11185 char subdev_name[20];
11186 struct mdstat_ent *mdstat;
11187
11188 sprintf(subdev_name, "%d", subdev);
11189 mdstat = mdstat_by_subdev(subdev_name, container);
11190 if (!mdstat)
4dd2df09 11191 return NULL;
78b10e66 11192
4dd2df09 11193 strcpy(devnm, mdstat->devnm);
78b10e66 11194 free_mdstat(mdstat);
4dd2df09 11195 return devnm;
78b10e66
N		 11196}
11197
11198static int imsm_reshape_is_allowed_on_container(struct supertype *st,
11199 struct geo_params *geo,
fbf3d202
AK		 11200 int *old_raid_disks,
11201 int direction)
78b10e66 11202{
694575e7
KW		 11203 /* currently we only support increasing the number of devices
11204 * for a container. This increases the number of device for each
11205 * member array. They must all be RAID0 or RAID5.
11206 */
78b10e66
N		 11207 int ret_val = 0;
11208 struct mdinfo *info, *member;
11209 int devices_that_can_grow = 0;
11210
7a862a02 11211 dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): st->devnm = (%s)\n", st->devnm);
78b10e66 11212
d04f65f4 11213 if (geo->size > 0 ||
78b10e66
N		 11214 geo->level != UnSet ||
11215 geo->layout != UnSet ||
11216 geo->chunksize != 0 ||
11217 geo->raid_disks == UnSet) {
7a862a02 11218 dprintf("imsm: Container operation is allowed for raid disks number change only.\n");
78b10e66
N		 11219 return ret_val;
11220 }
11221
fbf3d202 11222 if (direction == ROLLBACK_METADATA_CHANGES) {
7a862a02 11223 dprintf("imsm: Metadata changes rollback is not supported for container operation.\n");
fbf3d202
AK		 11224 return ret_val;
11225 }
11226
78b10e66
N		 11227 info = container_content_imsm(st, NULL);
11228 for (member = info; member; member = member->next) {
4dd2df09 11229 char *result;
78b10e66
N		 11230
11231 dprintf("imsm: checking device_num: %i\n",
11232 member->container_member);
11233
d7d205bd 11234 if (geo->raid_disks <= member->array.raid_disks) {
78b10e66
N		 11235 /* we work on container for Online Capacity Expansion
11236 * only so raid_disks has to grow
11237 */
7a862a02 11238 dprintf("imsm: for container operation raid disks increase is required\n");
78b10e66
N		 11239 break;
11240 }
11241
089f9d79 11242 if (info->array.level != 0 && info->array.level != 5) {
78b10e66
N		 11243 /* we cannot use this container with other raid level
11244 */
7a862a02 11245 dprintf("imsm: for container operation wrong raid level (%i) detected\n",
78b10e66
N		 11246 info->array.level);
11247 break;
11248 } else {
11249 /* check for platform support
11250 * for this raid level configuration
11251 */
11252 struct intel_super *super = st->sb;
11253 if (!is_raid_level_supported(super->orom,
11254 member->array.level,
11255 geo->raid_disks)) {
7a862a02 11256 dprintf("platform does not support raid%d with %d disk%s\n",
78b10e66
N		 11257 info->array.level,
11258 geo->raid_disks,
11259 geo->raid_disks > 1 ? "s" : "");
11260 break;
11261 }
2a4a08e7
AK		 11262 /* check if component size is aligned to chunk size
11263 */
11264 if (info->component_size %
11265 (info->array.chunk_size/512)) {
7a862a02 11266 dprintf("Component size is not aligned to chunk size\n");
2a4a08e7
AK		 11267 break;
11268 }
78b10e66
N		 11269 }
11270
11271 if (*old_raid_disks &&
11272 info->array.raid_disks != *old_raid_disks)
11273 break;
11274 *old_raid_disks = info->array.raid_disks;
11275
11276 /* All raid5 and raid0 volumes in container
11277 * have to be ready for Online Capacity Expansion
11278 * so they need to be assembled. We have already
11279 * checked that no recovery etc is happening.
11280 */
4dd2df09
N		 11281 result = imsm_find_array_devnm_by_subdev(member->container_member,
11282 st->container_devnm);
11283 if (result == NULL) {
78b10e66
N		 11284 dprintf("imsm: cannot find array\n");
11285 break;
11286 }
11287 devices_that_can_grow++;
11288 }
11289 sysfs_free(info);
11290 if (!member && devices_that_can_grow)
11291 ret_val = 1;
11292
11293 if (ret_val)
1ade5cc1 11294 dprintf("Container operation allowed\n");
78b10e66 11295 else
1ade5cc1 11296 dprintf("Error: %i\n", ret_val);
78b10e66
N		 11297
11298 return ret_val;
11299}
11300
11301/* Function: get_spares_for_grow
11302 * Description: Allocates memory and creates list of spare devices
1011e834 11303 * avaliable in container. Checks if spare drive size is acceptable.
78b10e66
N		 11304 * Parameters: Pointer to the supertype structure
11305 * Returns: Pointer to the list of spare devices (mdinfo structure) on success,
1011e834 11306 * NULL if fail
78b10e66
N		 11307 */
11308static struct mdinfo *get_spares_for_grow(struct supertype *st)
11309{
fbfdcb06
AO		 11310 struct spare_criteria sc;
11311
11312 get_spare_criteria_imsm(st, &sc);
11313 return container_choose_spares(st, &sc, NULL, NULL, NULL, 0);
78b10e66
N		 11314}
11315
11316/******************************************************************************
11317 * function: imsm_create_metadata_update_for_reshape
11318 * Function creates update for whole IMSM container.
11319 *
11320 ******************************************************************************/
11321static int imsm_create_metadata_update_for_reshape(
11322 struct supertype *st,
11323 struct geo_params *geo,
11324 int old_raid_disks,
11325 struct imsm_update_reshape **updatep)
11326{
11327 struct intel_super *super = st->sb;
11328 struct imsm_super *mpb = super->anchor;
594dc1b8
JS		 11329 int update_memory_size;
11330 struct imsm_update_reshape *u;
11331 struct mdinfo *spares;
78b10e66 11332 int i;
594dc1b8 11333 int delta_disks;
bbd24d86 11334 struct mdinfo *dev;
78b10e66 11335
1ade5cc1 11336 dprintf("(enter) raid_disks = %i\n", geo->raid_disks);
78b10e66
N		 11337
11338 delta_disks = geo->raid_disks - old_raid_disks;
11339
11340 /* size of all update data without anchor */
11341 update_memory_size = sizeof(struct imsm_update_reshape);
11342
11343 /* now add space for spare disks that we need to add. */
11344 update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1);
11345
503975b9 11346 u = xcalloc(1, update_memory_size);
78b10e66
N		 11347 u->type = update_reshape_container_disks;
11348 u->old_raid_disks = old_raid_disks;
11349 u->new_raid_disks = geo->raid_disks;
11350
11351 /* now get spare disks list
11352 */
11353 spares = get_spares_for_grow(st);
11354
d7be7d87 11355 if (spares == NULL || delta_disks > spares->array.spare_disks) {
7a862a02 11356 pr_err("imsm: ERROR: Cannot get spare devices for %s.\n", geo->dev_name);
e4c72d1d 11357 i = -1;
78b10e66
N		 11358 goto abort;
11359 }
11360
11361 /* we have got spares
11362 * update disk list in imsm_disk list table in anchor
11363 */
11364 dprintf("imsm: %i spares are available.\n\n",
11365 spares->array.spare_disks);
11366
bbd24d86 11367 dev = spares->devs;
78b10e66 11368 for (i = 0; i < delta_disks; i++) {
78b10e66
N		 11369 struct dl *dl;
11370
bbd24d86
AK		 11371 if (dev == NULL)
11372 break;
78b10e66
N		 11373 u->new_disks[i] = makedev(dev->disk.major,
11374 dev->disk.minor);
11375 dl = get_disk_super(super, dev->disk.major, dev->disk.minor);
ee4beede
AK		 11376 dl->index = mpb->num_disks;
11377 mpb->num_disks++;
bbd24d86 11378 dev = dev->next;
78b10e66 11379 }
78b10e66
N		 11380
11381abort:
11382 /* free spares
11383 */
11384 sysfs_free(spares);
11385
d677e0b8 11386 dprintf("imsm: reshape update preparation :");
78b10e66 11387 if (i == delta_disks) {
1ade5cc1 11388 dprintf_cont(" OK\n");
78b10e66
N		 11389 *updatep = u;
11390 return update_memory_size;
11391 }
11392 free(u);
1ade5cc1 11393 dprintf_cont(" Error\n");
78b10e66
N		 11394
11395 return 0;
11396}
11397
f3871fdc
AK		 11398/******************************************************************************
11399 * function: imsm_create_metadata_update_for_size_change()
11400 * Creates update for IMSM array for array size change.
11401 *
11402 ******************************************************************************/
11403static int imsm_create_metadata_update_for_size_change(
11404 struct supertype *st,
11405 struct geo_params *geo,
11406 struct imsm_update_size_change **updatep)
11407{
11408 struct intel_super *super = st->sb;
594dc1b8
JS		 11409 int update_memory_size;
11410 struct imsm_update_size_change *u;
f3871fdc 11411
1ade5cc1 11412 dprintf("(enter) New size = %llu\n", geo->size);
f3871fdc
AK		 11413
11414 /* size of all update data without anchor */
11415 update_memory_size = sizeof(struct imsm_update_size_change);
11416
503975b9 11417 u = xcalloc(1, update_memory_size);
f3871fdc
AK		 11418 u->type = update_size_change;
11419 u->subdev = super->current_vol;
11420 u->new_size = geo->size;
11421
11422 dprintf("imsm: reshape update preparation : OK\n");
11423 *updatep = u;
11424
11425 return update_memory_size;
11426}
11427
48c5303a
PC		 11428/******************************************************************************
11429 * function: imsm_create_metadata_update_for_migration()
11430 * Creates update for IMSM array.
11431 *
11432 ******************************************************************************/
11433static int imsm_create_metadata_update_for_migration(
11434 struct supertype *st,
11435 struct geo_params *geo,
11436 struct imsm_update_reshape_migration **updatep)
11437{
11438 struct intel_super *super = st->sb;
594dc1b8
JS		 11439 int update_memory_size;
11440 struct imsm_update_reshape_migration *u;
48c5303a
PC		 11441 struct imsm_dev *dev;
11442 int previous_level = -1;
11443
1ade5cc1 11444 dprintf("(enter) New Level = %i\n", geo->level);
48c5303a
PC		 11445
11446 /* size of all update data without anchor */
11447 update_memory_size = sizeof(struct imsm_update_reshape_migration);
11448
503975b9 11449 u = xcalloc(1, update_memory_size);
48c5303a
PC		 11450 u->type = update_reshape_migration;
11451 u->subdev = super->current_vol;
11452 u->new_level = geo->level;
11453 u->new_layout = geo->layout;
11454 u->new_raid_disks = u->old_raid_disks = geo->raid_disks;
11455 u->new_disks[0] = -1;
4bba0439 11456 u->new_chunksize = -1;
48c5303a
PC		 11457
11458 dev = get_imsm_dev(super, u->subdev);
11459 if (dev) {
11460 struct imsm_map *map;
11461
238c0a71 11462 map = get_imsm_map(dev, MAP_0);
4bba0439
PC		 11463 if (map) {
11464 int current_chunk_size =
11465 __le16_to_cpu(map->blocks_per_strip) / 2;
11466
11467 if (geo->chunksize != current_chunk_size) {
11468 u->new_chunksize = geo->chunksize / 1024;
7a862a02 11469 dprintf("imsm: chunk size change from %i to %i\n",
4bba0439
PC		 11470 current_chunk_size, u->new_chunksize);
11471 }
48c5303a 11472 previous_level = map->raid_level;
4bba0439 11473 }
48c5303a 11474 }
089f9d79 11475 if (geo->level == 5 && previous_level == 0) {
48c5303a
PC		 11476 struct mdinfo *spares = NULL;
11477
11478 u->new_raid_disks++;
11479 spares = get_spares_for_grow(st);
089f9d79 11480 if (spares == NULL || spares->array.spare_disks < 1) {
48c5303a
PC		 11481 free(u);
11482 sysfs_free(spares);
11483 update_memory_size = 0;
565cc99e 11484 pr_err("cannot get spare device for requested migration\n");
48c5303a
PC		 11485 return 0;
11486 }
11487 sysfs_free(spares);
11488 }
11489 dprintf("imsm: reshape update preparation : OK\n");
11490 *updatep = u;
11491
11492 return update_memory_size;
11493}
11494
8dd70bce
AK		 11495static void imsm_update_metadata_locally(struct supertype *st,
11496 void *buf, int len)
11497{
11498 struct metadata_update mu;
11499
11500 mu.buf = buf;
11501 mu.len = len;
11502 mu.space = NULL;
11503 mu.space_list = NULL;
11504 mu.next = NULL;
5fe6f031
N		 11505 if (imsm_prepare_update(st, &mu))
11506 imsm_process_update(st, &mu);
8dd70bce
AK		 11507
11508 while (mu.space_list) {
11509 void **space = mu.space_list;
11510 mu.space_list = *space;
11511 free(space);
11512 }
11513}
78b10e66 11514
471bceb6 11515/***************************************************************************
694575e7 11516* Function: imsm_analyze_change
471bceb6 11517* Description: Function analyze change for single volume
1011e834 11518* and validate if transition is supported
fbf3d202
AK		 11519* Parameters: Geometry parameters, supertype structure,
11520* metadata change direction (apply/rollback)
694575e7 11521* Returns: Operation type code on success, -1 if fail
471bceb6
KW		 11522****************************************************************************/
11523enum imsm_reshape_type imsm_analyze_change(struct supertype *st,
fbf3d202
AK		 11524 struct geo_params *geo,
11525 int direction)
694575e7 11526{
471bceb6
KW		 11527 struct mdinfo info;
11528 int change = -1;
11529 int check_devs = 0;
c21e737b 11530 int chunk;
67a2db32
AK		 11531 /* number of added/removed disks in operation result */
11532 int devNumChange = 0;
11533 /* imsm compatible layout value for array geometry verification */
11534 int imsm_layout = -1;
7abc9871
AK		 11535 int data_disks;
11536 struct imsm_dev *dev;
9529d343 11537 struct imsm_map *map;
7abc9871 11538 struct intel_super *super;
d04f65f4 11539 unsigned long long current_size;
65d38cca 11540 unsigned long long free_size;
d04f65f4 11541 unsigned long long max_size;
65d38cca 11542 int rv;
471bceb6
KW		 11543
11544 getinfo_super_imsm_volume(st, &info, NULL);
089f9d79
JS		 11545 if (geo->level != info.array.level && geo->level >= 0 &&
11546 geo->level != UnSet) {
471bceb6
KW		 11547 switch (info.array.level) {
11548 case 0:
11549 if (geo->level == 5) {
b5347799 11550 change = CH_MIGRATION;
e13ce846 11551 if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) {
7a862a02 11552 pr_err("Error. Requested Layout not supported (left-asymmetric layout is supported only)!\n");
e13ce846
AK		 11553 change = -1;
11554 goto analyse_change_exit;
11555 }
67a2db32 11556 imsm_layout = geo->layout;
471bceb6 11557 check_devs = 1;
e91a3bad
LM		 11558 devNumChange = 1; /* parity disk added */
11559 } else if (geo->level == 10) {
471bceb6
KW		 11560 change = CH_TAKEOVER;
11561 check_devs = 1;
e91a3bad 11562 devNumChange = 2; /* two mirrors added */
67a2db32 11563 imsm_layout = 0x102; /* imsm supported layout */
471bceb6 11564 }
dfe77a9e
KW		 11565 break;
11566 case 1:
471bceb6
KW		 11567 case 10:
11568 if (geo->level == 0) {
11569 change = CH_TAKEOVER;
11570 check_devs = 1;
e91a3bad 11571 devNumChange = -(geo->raid_disks/2);
67a2db32 11572 imsm_layout = 0; /* imsm raid0 layout */
471bceb6
KW		 11573 }
11574 break;
11575 }
11576 if (change == -1) {
7a862a02 11577 pr_err("Error. Level Migration from %d to %d not supported!\n",
e7b84f9d 11578 info.array.level, geo->level);
471bceb6
KW		 11579 goto analyse_change_exit;
11580 }
11581 } else
11582 geo->level = info.array.level;
11583
089f9d79
JS		 11584 if (geo->layout != info.array.layout &&
11585 (geo->layout != UnSet && geo->layout != -1)) {
b5347799 11586 change = CH_MIGRATION;
089f9d79
JS		 11587 if (info.array.layout == 0 && info.array.level == 5 &&
11588 geo->layout == 5) {
471bceb6 11589 /* reshape 5 -> 4 */
089f9d79
JS		 11590 } else if (info.array.layout == 5 && info.array.level == 5 &&
11591 geo->layout == 0) {
471bceb6
KW		 11592 /* reshape 4 -> 5 */
11593 geo->layout = 0;
11594 geo->level = 5;
11595 } else {
7a862a02 11596 pr_err("Error. Layout Migration from %d to %d not supported!\n",
e7b84f9d 11597 info.array.layout, geo->layout);
471bceb6
KW		 11598 change = -1;
11599 goto analyse_change_exit;
11600 }
67a2db32 11601 } else {
471bceb6 11602 geo->layout = info.array.layout;
67a2db32
AK		 11603 if (imsm_layout == -1)
11604 imsm_layout = info.array.layout;
11605 }
471bceb6 11606
089f9d79
JS		 11607 if (geo->chunksize > 0 && geo->chunksize != UnSet &&
11608 geo->chunksize != info.array.chunk_size) {
2d2b0eb7
MD		 11609 if (info.array.level == 10) {
11610 pr_err("Error. Chunk size change for RAID 10 is not supported.\n");
11611 change = -1;
11612 goto analyse_change_exit;
1e9b2c3f
PB		 11613 } else if (info.component_size % (geo->chunksize/512)) {
11614 pr_err("New chunk size (%dK) does not evenly divide device size (%lluk). Aborting...\n",
11615 geo->chunksize/1024, info.component_size/2);
11616 change = -1;
11617 goto analyse_change_exit;
2d2b0eb7 11618 }
b5347799 11619 change = CH_MIGRATION;
2d2b0eb7 11620 } else {
471bceb6 11621 geo->chunksize = info.array.chunk_size;
2d2b0eb7 11622 }
471bceb6 11623
c21e737b 11624 chunk = geo->chunksize / 1024;
7abc9871
AK		 11625
11626 super = st->sb;
11627 dev = get_imsm_dev(super, super->current_vol);
9529d343
MD		 11628 map = get_imsm_map(dev, MAP_0);
11629 data_disks = imsm_num_data_members(map);
c41e00b2 11630 /* compute current size per disk member
7abc9871 11631 */
c41e00b2
AK		 11632 current_size = info.custom_array_size / data_disks;
11633
089f9d79 11634 if (geo->size > 0 && geo->size != MAX_SIZE) {
c41e00b2
AK		 11635 /* align component size
11636 */
3e684231 11637 geo->size = imsm_component_size_alignment_check(
c41e00b2 11638 get_imsm_raid_level(dev->vol.map),
f36a9ecd 11639 chunk * 1024, super->sector_size,
c41e00b2 11640 geo->size * 2);
65d0b4ce 11641 if (geo->size == 0) {
7a862a02 11642 pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is 0).\n",
65d0b4ce
LD		 11643 current_size);
11644 goto analyse_change_exit;
11645 }
c41e00b2 11646 }
7abc9871 11647
089f9d79 11648 if (current_size != geo->size && geo->size > 0) {
7abc9871 11649 if (change != -1) {
7a862a02 11650 pr_err("Error. Size change should be the only one at a time.\n");
7abc9871
AK		 11651 change = -1;
11652 goto analyse_change_exit;
11653 }
11654 if ((super->current_vol + 1) != super->anchor->num_raid_devs) {
7a862a02 11655 pr_err("Error. The last volume in container can be expanded only (%i/%s).\n",
4dd2df09 11656 super->current_vol, st->devnm);
7abc9871
AK		 11657 goto analyse_change_exit;
11658 }
65d38cca
LD		 11659 /* check the maximum available size
11660 */
11661 rv = imsm_get_free_size(st, dev->vol.map->num_members,
11662 0, chunk, &free_size);
11663 if (rv == 0)
11664 /* Cannot find maximum available space
11665 */
11666 max_size = 0;
11667 else {
11668 max_size = free_size + current_size;
11669 /* align component size
11670 */
3e684231 11671 max_size = imsm_component_size_alignment_check(
65d38cca 11672 get_imsm_raid_level(dev->vol.map),
f36a9ecd 11673 chunk * 1024, super->sector_size,
65d38cca
LD		 11674 max_size);
11675 }
d04f65f4 11676 if (geo->size == MAX_SIZE) {
b130333f
AK		 11677 /* requested size change to the maximum available size
11678 */
65d38cca 11679 if (max_size == 0) {
7a862a02 11680 pr_err("Error. Cannot find maximum available space.\n");
b130333f
AK		 11681 change = -1;
11682 goto analyse_change_exit;
65d38cca
LD		 11683 } else
11684 geo->size = max_size;
c41e00b2 11685 }
b130333f 11686
681b7ae2 11687 if (direction == ROLLBACK_METADATA_CHANGES) {
fbf3d202
AK		 11688 /* accept size for rollback only
11689 */
11690 } else {
11691 /* round size due to metadata compatibility
11692 */
11693 geo->size = (geo->size >> SECT_PER_MB_SHIFT)
11694 << SECT_PER_MB_SHIFT;
11695 dprintf("Prepare update for size change to %llu\n",
11696 geo->size );
11697 if (current_size >= geo->size) {
7a862a02 11698 pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is %llu).\n",
e7b84f9d 11699 current_size, geo->size);
fbf3d202
AK		 11700 goto analyse_change_exit;
11701 }
65d38cca 11702 if (max_size && geo->size > max_size) {
7a862a02 11703 pr_err("Error. Requested size is larger than maximum available size (maximum available size is %llu, requested size /rounded/ is %llu).\n",
e7b84f9d 11704 max_size, geo->size);
65d38cca
LD		 11705 goto analyse_change_exit;
11706 }
7abc9871
AK		 11707 }
11708 geo->size *= data_disks;
11709 geo->raid_disks = dev->vol.map->num_members;
11710 change = CH_ARRAY_SIZE;
11711 }
471bceb6
KW		 11712 if (!validate_geometry_imsm(st,
11713 geo->level,
67a2db32 11714 imsm_layout,
e91a3bad 11715 geo->raid_disks + devNumChange,
c21e737b 11716 &chunk,
af4348dd 11717 geo->size, INVALID_SECTORS,
5308f117 11718 0, 0, info.consistency_policy, 1))
471bceb6
KW		 11719 change = -1;
11720
11721 if (check_devs) {
11722 struct intel_super *super = st->sb;
11723 struct imsm_super *mpb = super->anchor;
11724
11725 if (mpb->num_raid_devs > 1) {
7a862a02 11726 pr_err("Error. Cannot perform operation on %s- for this operation it MUST be single array in container\n",
e7b84f9d 11727 geo->dev_name);
471bceb6
KW		 11728 change = -1;
11729 }
11730 }
11731
11732analyse_change_exit:
089f9d79
JS		 11733 if (direction == ROLLBACK_METADATA_CHANGES &&
11734 (change == CH_MIGRATION || change == CH_TAKEOVER)) {
7a862a02 11735 dprintf("imsm: Metadata changes rollback is not supported for migration and takeover operations.\n");
fbf3d202
AK		 11736 change = -1;
11737 }
471bceb6 11738 return change;
694575e7
KW		 11739}
11740
bb025c2f
KW		 11741int imsm_takeover(struct supertype *st, struct geo_params *geo)
11742{
11743 struct intel_super *super = st->sb;
11744 struct imsm_update_takeover *u;
11745
503975b9 11746 u = xmalloc(sizeof(struct imsm_update_takeover));
bb025c2f
KW		 11747
11748 u->type = update_takeover;
11749 u->subarray = super->current_vol;
11750
11751 /* 10->0 transition */
11752 if (geo->level == 0)
11753 u->direction = R10_TO_R0;
11754
0529c688
KW		 11755 /* 0->10 transition */
11756 if (geo->level == 10)
11757 u->direction = R0_TO_R10;
11758
bb025c2f
KW		 11759 /* update metadata locally */
11760 imsm_update_metadata_locally(st, u,
11761 sizeof(struct imsm_update_takeover));
11762 /* and possibly remotely */
11763 if (st->update_tail)
11764 append_metadata_update(st, u,
11765 sizeof(struct imsm_update_takeover));
11766 else
11767 free(u);
11768
11769 return 0;
11770}
11771
895ffd99
MT		 11772/* Flush size update if size calculated by num_data_stripes is higher than
11773 * imsm_dev_size to eliminate differences during reshape.
11774 * Mdmon will recalculate them correctly.
11775 * If subarray index is not set then check whole container.
11776 * Returns:
11777 * 0 - no error occurred
11778 * 1 - error detected
11779 */
11780static int imsm_fix_size_mismatch(struct supertype *st, int subarray_index)
11781{
11782 struct intel_super *super = st->sb;
11783 int tmp = super->current_vol;
11784 int ret_val = 1;
11785 int i;
11786
11787 for (i = 0; i < super->anchor->num_raid_devs; i++) {
11788 if (subarray_index >= 0 && i != subarray_index)
11789 continue;
11790 super->current_vol = i;
11791 struct imsm_dev *dev = get_imsm_dev(super, super->current_vol);
11792 struct imsm_map *map = get_imsm_map(dev, MAP_0);
11793 unsigned int disc_count = imsm_num_data_members(map);
11794 struct geo_params geo;
11795 struct imsm_update_size_change *update;
11796 unsigned long long calc_size = per_dev_array_size(map) * disc_count;
11797 unsigned long long d_size = imsm_dev_size(dev);
11798 int u_size;
11799
11800 if (calc_size == d_size || dev->vol.migr_type == MIGR_GEN_MIGR)
11801 continue;
11802
ff904202
MT		 11803 /* There is a difference, confirm that imsm_dev_size is
11804 * smaller and push update.
895ffd99 11805 */
ff904202
MT		 11806 if (d_size > calc_size) {
11807 pr_err("imsm: dev size of subarray %d is incorrect\n",
11808 i);
895ffd99
MT		 11809 goto exit;
11810 }
11811 memset(&geo, 0, sizeof(struct geo_params));
11812 geo.size = d_size;
11813 u_size = imsm_create_metadata_update_for_size_change(st, &geo,
11814 &update);
11815 if (u_size < 1) {
11816 dprintf("imsm: Cannot prepare size change update\n");
11817 goto exit;
11818 }
11819 imsm_update_metadata_locally(st, update, u_size);
11820 if (st->update_tail) {
11821 append_metadata_update(st, update, u_size);
11822 flush_metadata_updates(st);
11823 st->update_tail = &st->updates;
11824 } else {
11825 imsm_sync_metadata(st);
11826 }
11827 }
11828 ret_val = 0;
11829exit:
11830 super->current_vol = tmp;
11831 return ret_val;
11832}
11833
d04f65f4
N		 11834static int imsm_reshape_super(struct supertype *st, unsigned long long size,
11835 int level,
78b10e66 11836 int layout, int chunksize, int raid_disks,
41784c88 11837 int delta_disks, char *backup, char *dev,
016e00f5 11838 int direction, int verbose)
78b10e66 11839{
78b10e66
N		 11840 int ret_val = 1;
11841 struct geo_params geo;
11842
1ade5cc1 11843 dprintf("(enter)\n");
78b10e66 11844
71204a50 11845 memset(&geo, 0, sizeof(struct geo_params));
78b10e66
N		 11846
11847 geo.dev_name = dev;
4dd2df09 11848 strcpy(geo.devnm, st->devnm);
78b10e66
N		 11849 geo.size = size;
11850 geo.level = level;
11851 geo.layout = layout;
11852 geo.chunksize = chunksize;
11853 geo.raid_disks = raid_disks;
41784c88
AK		 11854 if (delta_disks != UnSet)
11855 geo.raid_disks += delta_disks;
78b10e66 11856
1ade5cc1
N		 11857 dprintf("for level : %i\n", geo.level);
11858 dprintf("for raid_disks : %i\n", geo.raid_disks);
78b10e66 11859
4dd2df09 11860 if (strcmp(st->container_devnm, st->devnm) == 0) {
694575e7
KW		 11861 /* On container level we can only increase number of devices. */
11862 dprintf("imsm: info: Container operation\n");
78b10e66 11863 int old_raid_disks = 0;
6dc0be30 11864
78b10e66 11865 if (imsm_reshape_is_allowed_on_container(
fbf3d202 11866 st, &geo, &old_raid_disks, direction)) {
78b10e66
N		 11867 struct imsm_update_reshape *u = NULL;
11868 int len;
11869
895ffd99
MT		 11870 if (imsm_fix_size_mismatch(st, -1)) {
11871 dprintf("imsm: Cannot fix size mismatch\n");
11872 goto exit_imsm_reshape_super;
11873 }
11874
78b10e66
N		 11875 len = imsm_create_metadata_update_for_reshape(
11876 st, &geo, old_raid_disks, &u);
11877
ed08d51c
AK		 11878 if (len <= 0) {
11879 dprintf("imsm: Cannot prepare update\n");
11880 goto exit_imsm_reshape_super;
11881 }
11882
8dd70bce
AK		 11883 ret_val = 0;
11884 /* update metadata locally */
11885 imsm_update_metadata_locally(st, u, len);
11886 /* and possibly remotely */
11887 if (st->update_tail)
11888 append_metadata_update(st, u, len);
11889 else
ed08d51c 11890 free(u);
8dd70bce 11891
694575e7 11892 } else {
7a862a02 11893 pr_err("(imsm) Operation is not allowed on this container\n");
694575e7
KW		 11894 }
11895 } else {
11896 /* On volume level we support following operations
471bceb6
KW		 11897 * - takeover: raid10 -> raid0; raid0 -> raid10
11898 * - chunk size migration
11899 * - migration: raid5 -> raid0; raid0 -> raid5
11900 */
11901 struct intel_super *super = st->sb;
11902 struct intel_dev *dev = super->devlist;
4dd2df09 11903 int change;
694575e7 11904 dprintf("imsm: info: Volume operation\n");
471bceb6
KW		 11905 /* find requested device */
11906 while (dev) {
1011e834 11907 char *devnm =
4dd2df09
N		 11908 imsm_find_array_devnm_by_subdev(
11909 dev->index, st->container_devnm);
11910 if (devnm && strcmp(devnm, geo.devnm) == 0)
471bceb6
KW		 11911 break;
11912 dev = dev->next;
11913 }
11914 if (dev == NULL) {
4dd2df09
N		 11915 pr_err("Cannot find %s (%s) subarray\n",
11916 geo.dev_name, geo.devnm);
471bceb6
KW		 11917 goto exit_imsm_reshape_super;
11918 }
11919 super->current_vol = dev->index;
fbf3d202 11920 change = imsm_analyze_change(st, &geo, direction);
694575e7 11921 switch (change) {
471bceb6 11922 case CH_TAKEOVER:
bb025c2f 11923 ret_val = imsm_takeover(st, &geo);
694575e7 11924 break;
48c5303a
PC		 11925 case CH_MIGRATION: {
11926 struct imsm_update_reshape_migration *u = NULL;
11927 int len =
11928 imsm_create_metadata_update_for_migration(
11929 st, &geo, &u);
11930 if (len < 1) {
7a862a02 11931 dprintf("imsm: Cannot prepare update\n");
48c5303a
PC		 11932 break;
11933 }
471bceb6 11934 ret_val = 0;
48c5303a
PC		 11935 /* update metadata locally */
11936 imsm_update_metadata_locally(st, u, len);
11937 /* and possibly remotely */
11938 if (st->update_tail)
11939 append_metadata_update(st, u, len);
11940 else
11941 free(u);
11942 }
11943 break;
7abc9871 11944 case CH_ARRAY_SIZE: {
f3871fdc
AK		 11945 struct imsm_update_size_change *u = NULL;
11946 int len =
11947 imsm_create_metadata_update_for_size_change(
11948 st, &geo, &u);
11949 if (len < 1) {
7a862a02 11950 dprintf("imsm: Cannot prepare update\n");
f3871fdc
AK		 11951 break;
11952 }
11953 ret_val = 0;
11954 /* update metadata locally */
11955 imsm_update_metadata_locally(st, u, len);
11956 /* and possibly remotely */
11957 if (st->update_tail)
11958 append_metadata_update(st, u, len);
11959 else
11960 free(u);
7abc9871
AK		 11961 }
11962 break;
471bceb6
KW		 11963 default:
11964 ret_val = 1;
694575e7 11965 }
694575e7 11966 }
78b10e66 11967
ed08d51c 11968exit_imsm_reshape_super:
78b10e66
N		 11969 dprintf("imsm: reshape_super Exit code = %i\n", ret_val);
11970 return ret_val;
11971}
2cda7640 11972
0febb20c
AO		 11973#define COMPLETED_OK 0
11974#define COMPLETED_NONE 1
11975#define COMPLETED_DELAYED 2
11976
11977static int read_completed(int fd, unsigned long long *val)
11978{
11979 int ret;
11980 char buf[50];
11981
11982 ret = sysfs_fd_get_str(fd, buf, 50);
11983 if (ret < 0)
11984 return ret;
11985
11986 ret = COMPLETED_OK;
11987 if (strncmp(buf, "none", 4) == 0) {
11988 ret = COMPLETED_NONE;
11989 } else if (strncmp(buf, "delayed", 7) == 0) {
11990 ret = COMPLETED_DELAYED;
11991 } else {
11992 char *ep;
11993 *val = strtoull(buf, &ep, 0);
11994 if (ep == buf || (*ep != 0 && *ep != '\n' && *ep != ' '))
11995 ret = -1;
11996 }
11997 return ret;
11998}
11999
eee67a47
AK		 12000/*******************************************************************************
12001 * Function: wait_for_reshape_imsm
12002 * Description: Function writes new sync_max value and waits until
12003 * reshape process reach new position
12004 * Parameters:
12005 * sra : general array info
eee67a47
AK		 12006 * ndata : number of disks in new array's layout
12007 * Returns:
12008 * 0 : success,
12009 * 1 : there is no reshape in progress,
12010 * -1 : fail
12011 ******************************************************************************/
ae9f01f8 12012int wait_for_reshape_imsm(struct mdinfo *sra, int ndata)
eee67a47 12013{
85ca499c 12014 int fd = sysfs_get_fd(sra, NULL, "sync_completed");
df2647fa 12015 int retry = 3;
eee67a47 12016 unsigned long long completed;
ae9f01f8
AK		 12017 /* to_complete : new sync_max position */
12018 unsigned long long to_complete = sra->reshape_progress;
12019 unsigned long long position_to_set = to_complete / ndata;
eee67a47 12020
ae9f01f8 12021 if (fd < 0) {
1ade5cc1 12022 dprintf("cannot open reshape_position\n");
eee67a47 12023 return 1;
ae9f01f8 12024 }
eee67a47 12025
df2647fa
PB		 12026 do {
12027 if (sysfs_fd_get_ll(fd, &completed) < 0) {
12028 if (!retry) {
12029 dprintf("cannot read reshape_position (no reshape in progres)\n");
12030 close(fd);
12031 return 1;
12032 }
12033 usleep(30000);
12034 } else
12035 break;
12036 } while (retry--);
eee67a47 12037
85ca499c 12038 if (completed > position_to_set) {
1ade5cc1 12039 dprintf("wrong next position to set %llu (%llu)\n",
85ca499c 12040 to_complete, position_to_set);
ae9f01f8
AK		 12041 close(fd);
12042 return -1;
12043 }
12044 dprintf("Position set: %llu\n", position_to_set);
12045 if (sysfs_set_num(sra, NULL, "sync_max",
12046 position_to_set) != 0) {
1ade5cc1 12047 dprintf("cannot set reshape position to %llu\n",
ae9f01f8
AK		 12048 position_to_set);
12049 close(fd);
12050 return -1;
eee67a47
AK		 12051 }
12052
eee67a47 12053 do {
0febb20c 12054 int rc;
eee67a47 12055 char action[20];
5ff3a780 12056 int timeout = 3000;
0febb20c 12057
5ff3a780 12058 sysfs_wait(fd, &timeout);
a47e44fb
AK		 12059 if (sysfs_get_str(sra, NULL, "sync_action",
12060 action, 20) > 0 &&
d7d3809a 12061 strncmp(action, "reshape", 7) != 0) {
b2be2b62
AO		 12062 if (strncmp(action, "idle", 4) == 0)
12063 break;
d7d3809a
AP		 12064 close(fd);
12065 return -1;
12066 }
0febb20c
AO		 12067
12068 rc = read_completed(fd, &completed);
12069 if (rc < 0) {
1ade5cc1 12070 dprintf("cannot read reshape_position (in loop)\n");
eee67a47
AK		 12071 close(fd);
12072 return 1;
0febb20c
AO		 12073 } else if (rc == COMPLETED_NONE)
12074 break;
85ca499c 12075 } while (completed < position_to_set);
b2be2b62 12076
eee67a47
AK		 12077 close(fd);
12078 return 0;
eee67a47
AK		 12079}
12080
b915c95f
AK		 12081/*******************************************************************************
12082 * Function: check_degradation_change
12083 * Description: Check that array hasn't become failed.
12084 * Parameters:
12085 * info : for sysfs access
12086 * sources : source disks descriptors
12087 * degraded: previous degradation level
12088 * Returns:
12089 * degradation level
12090 ******************************************************************************/
12091int check_degradation_change(struct mdinfo *info,
12092 int *sources,
12093 int degraded)
12094{
12095 unsigned long long new_degraded;
e1993023
LD		 12096 int rv;
12097
12098 rv = sysfs_get_ll(info, NULL, "degraded", &new_degraded);
089f9d79 12099 if (rv == -1 || (new_degraded != (unsigned long long)degraded)) {
b915c95f
AK		 12100 /* check each device to ensure it is still working */
12101 struct mdinfo *sd;
12102 new_degraded = 0;
12103 for (sd = info->devs ; sd ; sd = sd->next) {
12104 if (sd->disk.state & (1<<MD_DISK_FAULTY))
12105 continue;
12106 if (sd->disk.state & (1<<MD_DISK_SYNC)) {
cf52eff5
TM		 12107 char sbuf[100];
12108
b915c95f 12109 if (sysfs_get_str(info,
cf52eff5 12110 sd, "state", sbuf, sizeof(sbuf)) < 0 ||
b915c95f
AK		 12111 strstr(sbuf, "faulty") ||
12112 strstr(sbuf, "in_sync") == NULL) {
12113 /* this device is dead */
12114 sd->disk.state = (1<<MD_DISK_FAULTY);
12115 if (sd->disk.raid_disk >= 0 &&
12116 sources[sd->disk.raid_disk] >= 0) {
12117 close(sources[
12118 sd->disk.raid_disk]);
12119 sources[sd->disk.raid_disk] =
12120 -1;
12121 }
12122 new_degraded++;
12123 }
12124 }
12125 }
12126 }
12127
12128 return new_degraded;
12129}
12130
10f22854
AK		 12131/*******************************************************************************
12132 * Function: imsm_manage_reshape
12133 * Description: Function finds array under reshape and it manages reshape
12134 * process. It creates stripes backups (if required) and sets
942e1cdb 12135 * checkpoints.
10f22854
AK		 12136 * Parameters:
12137 * afd : Backup handle (nattive) - not used
12138 * sra : general array info
12139 * reshape : reshape parameters - not used
12140 * st : supertype structure
12141 * blocks : size of critical section [blocks]
12142 * fds : table of source device descriptor
12143 * offsets : start of array (offest per devices)
12144 * dests : not used
12145 * destfd : table of destination device descriptor
12146 * destoffsets : table of destination offsets (per device)
12147 * Returns:
12148 * 1 : success, reshape is done
12149 * 0 : fail
12150 ******************************************************************************/
999b4972
N		 12151static int imsm_manage_reshape(
12152 int afd, struct mdinfo *sra, struct reshape *reshape,
10f22854 12153 struct supertype *st, unsigned long backup_blocks,
999b4972
N		 12154 int *fds, unsigned long long *offsets,
12155 int dests, int *destfd, unsigned long long *destoffsets)
12156{
10f22854
AK		 12157 int ret_val = 0;
12158 struct intel_super *super = st->sb;
594dc1b8 12159 struct intel_dev *dv;
de44e46f 12160 unsigned int sector_size = super->sector_size;
10f22854 12161 struct imsm_dev *dev = NULL;
9529d343 12162 struct imsm_map *map_src, *map_dest;
10f22854
AK		 12163 int migr_vol_qan = 0;
12164 int ndata, odata; /* [bytes] */
12165 int chunk; /* [bytes] */
12166 struct migr_record *migr_rec;
12167 char *buf = NULL;
12168 unsigned int buf_size; /* [bytes] */
12169 unsigned long long max_position; /* array size [bytes] */
12170 unsigned long long next_step; /* [blocks]/[bytes] */
12171 unsigned long long old_data_stripe_length;
10f22854
AK		 12172 unsigned long long start_src; /* [bytes] */
12173 unsigned long long start; /* [bytes] */
12174 unsigned long long start_buf_shift; /* [bytes] */
b915c95f 12175 int degraded = 0;
ab724b98 12176 int source_layout = 0;
895ffd99 12177 int subarray_index = -1;
10f22854 12178
79a16a9b
JS		 12179 if (!sra)
12180 return ret_val;
12181
12182 if (!fds || !offsets)
10f22854
AK		 12183 goto abort;
12184
12185 /* Find volume during the reshape */
12186 for (dv = super->devlist; dv; dv = dv->next) {
fc54fe7a
JS		 12187 if (dv->dev->vol.migr_type == MIGR_GEN_MIGR &&
12188 dv->dev->vol.migr_state == 1) {
10f22854
AK		 12189 dev = dv->dev;
12190 migr_vol_qan++;
895ffd99 12191 subarray_index = dv->index;
10f22854
AK		 12192 }
12193 }
12194 /* Only one volume can migrate at the same time */
12195 if (migr_vol_qan != 1) {
676e87a8 12196 pr_err("%s", migr_vol_qan ?
10f22854
AK		 12197 "Number of migrating volumes greater than 1\n" :
12198 "There is no volume during migrationg\n");
12199 goto abort;
12200 }
12201
9529d343 12202 map_dest = get_imsm_map(dev, MAP_0);
238c0a71 12203 map_src = get_imsm_map(dev, MAP_1);
10f22854
AK		 12204 if (map_src == NULL)
12205 goto abort;
10f22854 12206
9529d343
MD		 12207 ndata = imsm_num_data_members(map_dest);
12208 odata = imsm_num_data_members(map_src);
10f22854 12209
7b1ab482 12210 chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512;
10f22854
AK		 12211 old_data_stripe_length = odata * chunk;
12212
12213 migr_rec = super->migr_rec;
12214
10f22854
AK		 12215 /* initialize migration record for start condition */
12216 if (sra->reshape_progress == 0)
12217 init_migr_record_imsm(st, dev, sra);
b2c59438
AK		 12218 else {
12219 if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) {
7a862a02 12220 dprintf("imsm: cannot restart migration when data are present in copy area.\n");
b2c59438
AK		 12221 goto abort;
12222 }
6a75c8ca
AK		 12223 /* Save checkpoint to update migration record for current
12224 * reshape position (in md). It can be farther than current
12225 * reshape position in metadata.
12226 */
12227 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
12228 /* ignore error == 2, this can mean end of reshape here
12229 */
7a862a02 12230 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL, initial save)\n");
6a75c8ca
AK		 12231 goto abort;
12232 }
b2c59438 12233 }
10f22854
AK		 12234
12235 /* size for data */
12236 buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512;
12237 /* extend buffer size for parity disk */
12238 buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
3e684231 12239 /* add space for stripe alignment */
10f22854 12240 buf_size += old_data_stripe_length;
de44e46f
PB		 12241 if (posix_memalign((void **)&buf, MAX_SECTOR_SIZE, buf_size)) {
12242 dprintf("imsm: Cannot allocate checkpoint buffer\n");
10f22854
AK		 12243 goto abort;
12244 }
12245
3ef4403c 12246 max_position = sra->component_size * ndata;
68eb8bc6 12247 source_layout = imsm_level_to_layout(map_src->raid_level);
10f22854 12248
9f421827
PB		 12249 while (current_migr_unit(migr_rec) <
12250 get_num_migr_units(migr_rec)) {
10f22854
AK		 12251 /* current reshape position [blocks] */
12252 unsigned long long current_position =
12253 __le32_to_cpu(migr_rec->blocks_per_unit)
9f421827 12254 * current_migr_unit(migr_rec);
10f22854
AK		 12255 unsigned long long border;
12256
b915c95f
AK		 12257 /* Check that array hasn't become failed.
12258 */
12259 degraded = check_degradation_change(sra, fds, degraded);
12260 if (degraded > 1) {
7a862a02 12261 dprintf("imsm: Abort reshape due to degradation level (%i)\n", degraded);
b915c95f
AK		 12262 goto abort;
12263 }
12264
10f22854
AK		 12265 next_step = __le32_to_cpu(migr_rec->blocks_per_unit);
12266
12267 if ((current_position + next_step) > max_position)
12268 next_step = max_position - current_position;
12269
92144abf 12270 start = current_position * 512;
10f22854 12271
942e1cdb 12272 /* align reading start to old geometry */
10f22854
AK		 12273 start_buf_shift = start % old_data_stripe_length;
12274 start_src = start - start_buf_shift;
12275
12276 border = (start_src / odata) - (start / ndata);
12277 border /= 512;
12278 if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) {
12279 /* save critical stripes to buf
12280 * start - start address of current unit
12281 * to backup [bytes]
12282 * start_src - start address of current unit
12283 * to backup alligned to source array
12284 * [bytes]
12285 */
594dc1b8 12286 unsigned long long next_step_filler;
10f22854
AK		 12287 unsigned long long copy_length = next_step * 512;
12288
12289 /* allign copy area length to stripe in old geometry */
12290 next_step_filler = ((copy_length + start_buf_shift)
12291 % old_data_stripe_length);
12292 if (next_step_filler)
12293 next_step_filler = (old_data_stripe_length
12294 - next_step_filler);
7a862a02 12295 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		 12296 start, start_src, copy_length,
12297 start_buf_shift, next_step_filler);
12298
12299 if (save_stripes(fds, offsets, map_src->num_members,
ab724b98
AK		 12300 chunk, map_src->raid_level,
12301 source_layout, 0, NULL, start_src,
10f22854
AK		 12302 copy_length +
12303 next_step_filler + start_buf_shift,
12304 buf)) {
7a862a02 12305 dprintf("imsm: Cannot save stripes to buffer\n");
10f22854
AK		 12306 goto abort;
12307 }
12308 /* Convert data to destination format and store it
12309 * in backup general migration area
12310 */
12311 if (save_backup_imsm(st, dev, sra,
aea93171 12312 buf + start_buf_shift, copy_length)) {
7a862a02 12313 dprintf("imsm: Cannot save stripes to target devices\n");
10f22854
AK		 12314 goto abort;
12315 }
12316 if (save_checkpoint_imsm(st, sra,
12317 UNIT_SRC_IN_CP_AREA)) {
7a862a02 12318 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_IN_CP_AREA)\n");
10f22854
AK		 12319 goto abort;
12320 }
8016a6d4
AK		 12321 } else {
12322 /* set next step to use whole border area */
12323 border /= next_step;
12324 if (border > 1)
12325 next_step *= border;
10f22854
AK		 12326 }
12327 /* When data backed up, checkpoint stored,
12328 * kick the kernel to reshape unit of data
12329 */
12330 next_step = next_step + sra->reshape_progress;
8016a6d4
AK		 12331 /* limit next step to array max position */
12332 if (next_step > max_position)
12333 next_step = max_position;
10f22854
AK		 12334 sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress);
12335 sysfs_set_num(sra, NULL, "suspend_hi", next_step);
ae9f01f8 12336 sra->reshape_progress = next_step;
10f22854
AK		 12337
12338 /* wait until reshape finish */
c85338c6 12339 if (wait_for_reshape_imsm(sra, ndata)) {
c47b0ff6
AK		 12340 dprintf("wait_for_reshape_imsm returned error!\n");
12341 goto abort;
12342 }
84d11e6c
N		 12343 if (sigterm)
12344 goto abort;
10f22854 12345
0228d92c
AK		 12346 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
12347 /* ignore error == 2, this can mean end of reshape here
12348 */
7a862a02 12349 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL)\n");
10f22854
AK		 12350 goto abort;
12351 }
12352
12353 }
12354
71e5411e
PB		 12355 /* clear migr_rec on disks after successful migration */
12356 struct dl *d;
12357
85337573 12358 memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE);
71e5411e
PB		 12359 for (d = super->disks; d; d = d->next) {
12360 if (d->index < 0 || is_failed(&d->disk))
12361 continue;
12362 unsigned long long dsize;
12363
12364 get_dev_size(d->fd, NULL, &dsize);
de44e46f 12365 if (lseek64(d->fd, dsize - MIGR_REC_SECTOR_POSITION*sector_size,
71e5411e 12366 SEEK_SET) >= 0) {
466070ad 12367 if ((unsigned int)write(d->fd, super->migr_rec_buf,
de44e46f
PB		 12368 MIGR_REC_BUF_SECTORS*sector_size) !=
12369 MIGR_REC_BUF_SECTORS*sector_size)
71e5411e
PB		 12370 perror("Write migr_rec failed");
12371 }
12372 }
12373
10f22854
AK		 12374 /* return '1' if done */
12375 ret_val = 1;
895ffd99
MT		 12376
12377 /* After the reshape eliminate size mismatch in metadata.
12378 * Don't update md/component_size here, volume hasn't
12379 * to take whole space. It is allowed by kernel.
12380 * md/component_size will be set propoperly after next assembly.
12381 */
12382 imsm_fix_size_mismatch(st, subarray_index);
12383
10f22854
AK		 12384abort:
12385 free(buf);
942e1cdb
N		 12386 /* See Grow.c: abort_reshape() for further explanation */
12387 sysfs_set_num(sra, NULL, "suspend_lo", 0x7FFFFFFFFFFFFFFFULL);
12388 sysfs_set_num(sra, NULL, "suspend_hi", 0);
12389 sysfs_set_num(sra, NULL, "suspend_lo", 0);
10f22854
AK		 12390
12391 return ret_val;
999b4972 12392}
0c21b485 12393
fbc42556
JR		 12394/*******************************************************************************
12395 * Function: calculate_bitmap_min_chunksize
12396 * Description: Calculates the minimal valid bitmap chunk size
12397 * Parameters:
12398 * max_bits : indicate how many bits can be used for the bitmap
12399 * data_area_size : the size of the data area covered by the bitmap
12400 *
12401 * Returns:
12402 * The bitmap chunk size
12403 ******************************************************************************/
12404static unsigned long long
12405calculate_bitmap_min_chunksize(unsigned long long max_bits,
12406 unsigned long long data_area_size)
12407{
12408 unsigned long long min_chunk =
12409 4096; /* sub-page chunks don't work yet.. */
12410 unsigned long long bits = data_area_size / min_chunk + 1;
12411
12412 while (bits > max_bits) {
12413 min_chunk *= 2;
12414 bits = (bits + 1) / 2;
12415 }
12416 return min_chunk;
12417}
12418
12419/*******************************************************************************
12420 * Function: calculate_bitmap_chunksize
12421 * Description: Calculates the bitmap chunk size for the given device
12422 * Parameters:
12423 * st : supertype information
12424 * dev : device for the bitmap
12425 *
12426 * Returns:
12427 * The bitmap chunk size
12428 ******************************************************************************/
12429static unsigned long long calculate_bitmap_chunksize(struct supertype *st,
12430 struct imsm_dev *dev)
12431{
12432 struct intel_super *super = st->sb;
12433 unsigned long long min_chunksize;
12434 unsigned long long result = IMSM_DEFAULT_BITMAP_CHUNKSIZE;
12435 size_t dev_size = imsm_dev_size(dev);
12436
12437 min_chunksize = calculate_bitmap_min_chunksize(
12438 IMSM_BITMAP_AREA_SIZE * super->sector_size, dev_size);
12439
12440 if (result < min_chunksize)
12441 result = min_chunksize;
12442
12443 return result;
12444}
12445
12446/*******************************************************************************
12447 * Function: init_bitmap_header
12448 * Description: Initialize the bitmap header structure
12449 * Parameters:
12450 * st : supertype information
12451 * bms : bitmap header struct to initialize
12452 * dev : device for the bitmap
12453 *
12454 * Returns:
12455 * 0 : success
12456 * -1 : fail
12457 ******************************************************************************/
12458static int init_bitmap_header(struct supertype *st, struct bitmap_super_s *bms,
12459 struct imsm_dev *dev)
12460{
12461 int vol_uuid[4];
12462
12463 if (!bms || !dev)
12464 return -1;
12465
12466 bms->magic = __cpu_to_le32(BITMAP_MAGIC);
12467 bms->version = __cpu_to_le32(BITMAP_MAJOR_HI);
12468 bms->daemon_sleep = __cpu_to_le32(IMSM_DEFAULT_BITMAP_DAEMON_SLEEP);
12469 bms->sync_size = __cpu_to_le64(IMSM_BITMAP_AREA_SIZE);
12470 bms->write_behind = __cpu_to_le32(0);
12471
12472 uuid_from_super_imsm(st, vol_uuid);
12473 memcpy(bms->uuid, vol_uuid, 16);
12474
12475 bms->chunksize = calculate_bitmap_chunksize(st, dev);
12476
12477 return 0;
12478}
12479
12480/*******************************************************************************
12481 * Function: validate_internal_bitmap_for_drive
12482 * Description: Verify if the bitmap header for a given drive.
12483 * Parameters:
12484 * st : supertype information
12485 * offset : The offset from the beginning of the drive where to look for
12486 * the bitmap header.
12487 * d : the drive info
12488 *
12489 * Returns:
12490 * 0 : success
12491 * -1 : fail
12492 ******************************************************************************/
12493static int validate_internal_bitmap_for_drive(struct supertype *st,
12494 unsigned long long offset,
12495 struct dl *d)
12496{
12497 struct intel_super *super = st->sb;
12498 int ret = -1;
12499 int vol_uuid[4];
12500 bitmap_super_t *bms;
12501 int fd;
12502
12503 if (!d)
12504 return -1;
12505
12506 void *read_buf;
12507
12508 if (posix_memalign(&read_buf, MAX_SECTOR_SIZE, IMSM_BITMAP_HEADER_SIZE))
12509 return -1;
12510
12511 fd = d->fd;
12512 if (fd < 0) {
12513 fd = open(d->devname, O_RDONLY, 0);
12514 if (fd < 0) {
12515 dprintf("cannot open the device %s\n", d->devname);
12516 goto abort;
12517 }
12518 }
12519
12520 if (lseek64(fd, offset * super->sector_size, SEEK_SET) < 0)
12521 goto abort;
12522 if (read(fd, read_buf, IMSM_BITMAP_HEADER_SIZE) !=
12523 IMSM_BITMAP_HEADER_SIZE)
12524 goto abort;
12525
12526 uuid_from_super_imsm(st, vol_uuid);
12527
12528 bms = read_buf;
12529 if ((bms->magic != __cpu_to_le32(BITMAP_MAGIC)) ||
12530 (bms->version != __cpu_to_le32(BITMAP_MAJOR_HI)) ||
12531 (!same_uuid((int *)bms->uuid, vol_uuid, st->ss->swapuuid))) {
12532 dprintf("wrong bitmap header detected\n");
12533 goto abort;
12534 }
12535
12536 ret = 0;
12537abort:
12538 if ((d->fd < 0) && (fd >= 0))
12539 close(fd);
12540 if (read_buf)
12541 free(read_buf);
12542
12543 return ret;
12544}
12545
12546/*******************************************************************************
12547 * Function: validate_internal_bitmap_imsm
12548 * Description: Verify if the bitmap header is in place and with proper data.
12549 * Parameters:
12550 * st : supertype information
12551 *
12552 * Returns:
12553 * 0 : success or device w/o RWH_BITMAP
12554 * -1 : fail
12555 ******************************************************************************/
12556static int validate_internal_bitmap_imsm(struct supertype *st)
12557{
12558 struct intel_super *super = st->sb;
12559 struct imsm_dev *dev = get_imsm_dev(super, super->current_vol);
12560 unsigned long long offset;
12561 struct dl *d;
12562
12563 if (!dev)
12564 return -1;
12565
12566 if (dev->rwh_policy != RWH_BITMAP)
12567 return 0;
12568
12569 offset = get_bitmap_header_sector(super, super->current_vol);
12570 for (d = super->disks; d; d = d->next) {
12571 if (d->index < 0 || is_failed(&d->disk))
12572 continue;
12573
12574 if (validate_internal_bitmap_for_drive(st, offset, d)) {
12575 pr_err("imsm: bitmap validation failed\n");
12576 return -1;
12577 }
12578 }
12579 return 0;
12580}
12581
12582/*******************************************************************************
12583 * Function: add_internal_bitmap_imsm
12584 * Description: Mark the volume to use the bitmap and updates the chunk size value.
12585 * Parameters:
12586 * st : supertype information
12587 * chunkp : bitmap chunk size
12588 * delay : not used for imsm
12589 * write_behind : not used for imsm
12590 * size : not used for imsm
12591 * may_change : not used for imsm
12592 * amajor : not used for imsm
12593 *
12594 * Returns:
12595 * 0 : success
12596 * -1 : fail
12597 ******************************************************************************/
12598static int add_internal_bitmap_imsm(struct supertype *st, int *chunkp,
12599 int delay, int write_behind,
12600 unsigned long long size, int may_change,
12601 int amajor)
12602{
12603 struct intel_super *super = st->sb;
12604 int vol_idx = super->current_vol;
12605 struct imsm_dev *dev;
12606
12607 if (!super->devlist || vol_idx == -1 || !chunkp)
12608 return -1;
12609
12610 dev = get_imsm_dev(super, vol_idx);
12611
12612 if (!dev) {
12613 dprintf("cannot find the device for volume index %d\n",
12614 vol_idx);
12615 return -1;
12616 }
12617 dev->rwh_policy = RWH_BITMAP;
12618
12619 *chunkp = calculate_bitmap_chunksize(st, dev);
12620
12621 return 0;
12622}
12623
12624/*******************************************************************************
12625 * Function: locate_bitmap_imsm
12626 * Description: Seek 'fd' to start of write-intent-bitmap.
12627 * Parameters:
12628 * st : supertype information
12629 * fd : file descriptor for the device
12630 * node_num : not used for imsm
12631 *
12632 * Returns:
12633 * 0 : success
12634 * -1 : fail
12635 ******************************************************************************/
12636static int locate_bitmap_imsm(struct supertype *st, int fd, int node_num)
12637{
12638 struct intel_super *super = st->sb;
12639 unsigned long long offset;
12640 int vol_idx = super->current_vol;
12641
12642 if (!super->devlist || vol_idx == -1)
12643 return -1;
12644
12645 offset = get_bitmap_header_sector(super, super->current_vol);
12646 dprintf("bitmap header offset is %llu\n", offset);
12647
12648 lseek64(fd, offset << 9, 0);
12649
12650 return 0;
12651}
12652
12653/*******************************************************************************
12654 * Function: write_init_bitmap_imsm
12655 * Description: Write a bitmap header and prepares the area for the bitmap.
12656 * Parameters:
12657 * st : supertype information
12658 * fd : file descriptor for the device
12659 * update : not used for imsm
12660 *
12661 * Returns:
12662 * 0 : success
12663 * -1 : fail
12664 ******************************************************************************/
12665static int write_init_bitmap_imsm(struct supertype *st, int fd,
12666 enum bitmap_update update)
12667{
12668 struct intel_super *super = st->sb;
12669 int vol_idx = super->current_vol;
12670 int ret = 0;
12671 unsigned long long offset;
12672 bitmap_super_t bms = { 0 };
12673 size_t written = 0;
12674 size_t to_write;
12675 ssize_t rv_num;
12676 void *buf;
12677
12678 if (!super->devlist || !super->sector_size || vol_idx == -1)
12679 return -1;
12680
12681 struct imsm_dev *dev = get_imsm_dev(super, vol_idx);
12682
12683 /* first clear the space for bitmap header */
12684 unsigned long long bitmap_area_start =
12685 get_bitmap_header_sector(super, vol_idx);
12686
12687 dprintf("zeroing area start (%llu) and size (%u)\n", bitmap_area_start,
12688 IMSM_BITMAP_AND_HEADER_SIZE / super->sector_size);
12689 if (zero_disk_range(fd, bitmap_area_start,
12690 IMSM_BITMAP_HEADER_SIZE / super->sector_size)) {
12691 pr_err("imsm: cannot zeroing the space for the bitmap\n");
12692 return -1;
12693 }
12694
12695 /* The bitmap area should be filled with "1"s to perform initial
12696 * synchronization.
12697 */
12698 if (posix_memalign(&buf, MAX_SECTOR_SIZE, MAX_SECTOR_SIZE))
12699 return -1;
12700 memset(buf, 0xFF, MAX_SECTOR_SIZE);
12701 offset = get_bitmap_sector(super, vol_idx);
12702 lseek64(fd, offset << 9, 0);
12703 while (written < IMSM_BITMAP_AREA_SIZE) {
12704 to_write = IMSM_BITMAP_AREA_SIZE - written;
12705 if (to_write > MAX_SECTOR_SIZE)
12706 to_write = MAX_SECTOR_SIZE;
12707 rv_num = write(fd, buf, MAX_SECTOR_SIZE);
12708 if (rv_num != MAX_SECTOR_SIZE) {
12709 ret = -1;
12710 dprintf("cannot initialize bitmap area\n");
12711 goto abort;
12712 }
12713 written += rv_num;
12714 }
12715
12716 /* write a bitmap header */
12717 init_bitmap_header(st, &bms, dev);
12718 memset(buf, 0, MAX_SECTOR_SIZE);
12719 memcpy(buf, &bms, sizeof(bitmap_super_t));
12720 if (locate_bitmap_imsm(st, fd, 0)) {
12721 ret = -1;
12722 dprintf("cannot locate the bitmap\n");
12723 goto abort;
12724 }
12725 if (write(fd, buf, MAX_SECTOR_SIZE) != MAX_SECTOR_SIZE) {
12726 ret = -1;
12727 dprintf("cannot write the bitmap header\n");
12728 goto abort;
12729 }
12730 fsync(fd);
12731
12732abort:
12733 free(buf);
12734
12735 return ret;
12736}
12737
12738/*******************************************************************************
12739 * Function: is_vol_to_setup_bitmap
12740 * Description: Checks if a bitmap should be activated on the dev.
12741 * Parameters:
12742 * info : info about the volume to setup the bitmap
12743 * dev : the device to check against bitmap creation
12744 *
12745 * Returns:
12746 * 0 : bitmap should be set up on the device
12747 * -1 : otherwise
12748 ******************************************************************************/
12749static int is_vol_to_setup_bitmap(struct mdinfo *info, struct imsm_dev *dev)
12750{
12751 if (!dev || !info)
12752 return -1;
12753
12754 if ((strcmp((char *)dev->volume, info->name) == 0) &&
12755 (dev->rwh_policy == RWH_BITMAP))
12756 return -1;
12757
12758 return 0;
12759}
12760
12761/*******************************************************************************
12762 * Function: set_bitmap_sysfs
12763 * Description: Set the sysfs atributes of a given volume to activate the bitmap.
12764 * Parameters:
12765 * info : info about the volume where the bitmap should be setup
12766 * chunksize : bitmap chunk size
12767 * location : location of the bitmap
12768 *
12769 * Returns:
12770 * 0 : success
12771 * -1 : fail
12772 ******************************************************************************/
12773static int set_bitmap_sysfs(struct mdinfo *info, unsigned long long chunksize,
12774 char *location)
12775{
12776 /* The bitmap/metadata is set to external to allow changing of value for
12777 * bitmap/location. When external is used, the kernel will treat an offset
12778 * related to the device's first lba (in opposition to the "internal" case
12779 * when this value is related to the beginning of the superblock).
12780 */
12781 if (sysfs_set_str(info, NULL, "bitmap/metadata", "external")) {
12782 dprintf("failed to set bitmap/metadata\n");
12783 return -1;
12784 }
12785
12786 /* It can only be changed when no bitmap is active.
12787 * Should be bigger than 512 and must be power of 2.
12788 * It is expecting the value in bytes.
12789 */
12790 if (sysfs_set_num(info, NULL, "bitmap/chunksize",
12791 __cpu_to_le32(chunksize))) {
12792 dprintf("failed to set bitmap/chunksize\n");
12793 return -1;
12794 }
12795
12796 /* It is expecting the value in sectors. */
12797 if (sysfs_set_num(info, NULL, "bitmap/space",
12798 __cpu_to_le64(IMSM_BITMAP_AREA_SIZE))) {
12799 dprintf("failed to set bitmap/space\n");
12800 return -1;
12801 }
12802
12803 /* Determines the delay between the bitmap updates.
12804 * It is expecting the value in seconds.
12805 */
12806 if (sysfs_set_num(info, NULL, "bitmap/time_base",
12807 __cpu_to_le64(IMSM_DEFAULT_BITMAP_DAEMON_SLEEP))) {
12808 dprintf("failed to set bitmap/time_base\n");
12809 return -1;
12810 }
12811
12812 /* It is expecting the value in sectors with a sign at the beginning. */
12813 if (sysfs_set_str(info, NULL, "bitmap/location", location)) {
12814 dprintf("failed to set bitmap/location\n");
12815 return -1;
12816 }
12817
12818 return 0;
12819}
12820
12821/*******************************************************************************
12822 * Function: set_bitmap_imsm
12823 * Description: Setup the bitmap for the given volume
12824 * Parameters:
12825 * st : supertype information
12826 * info : info about the volume where the bitmap should be setup
12827 *
12828 * Returns:
12829 * 0 : success
12830 * -1 : fail
12831 ******************************************************************************/
12832static int set_bitmap_imsm(struct supertype *st, struct mdinfo *info)
12833{
12834 struct intel_super *super = st->sb;
12835 int prev_current_vol = super->current_vol;
12836 struct imsm_dev *dev;
12837 int ret = -1;
12838 char location[16] = "";
12839 unsigned long long chunksize;
12840 struct intel_dev *dev_it;
12841
12842 for (dev_it = super->devlist; dev_it; dev_it = dev_it->next) {
12843 super->current_vol = dev_it->index;
12844 dev = get_imsm_dev(super, super->current_vol);
12845
12846 if (is_vol_to_setup_bitmap(info, dev)) {
12847 if (validate_internal_bitmap_imsm(st)) {
12848 dprintf("bitmap header validation failed\n");
12849 goto abort;
12850 }
12851
12852 chunksize = calculate_bitmap_chunksize(st, dev);
12853 dprintf("chunk size is %llu\n", chunksize);
12854
12855 snprintf(location, sizeof(location), "+%llu",
12856 get_bitmap_sector(super, super->current_vol));
12857 dprintf("bitmap offset is %s\n", location);
12858
12859 if (set_bitmap_sysfs(info, chunksize, location)) {
12860 dprintf("cannot setup the bitmap\n");
12861 goto abort;
12862 }
12863 }
12864 }
12865 ret = 0;
12866abort:
12867 super->current_vol = prev_current_vol;
12868 return ret;
12869}
12870
cdddbdbc 12871struct superswitch super_imsm = {
cdddbdbc
DW		 12872 .examine_super = examine_super_imsm,
12873 .brief_examine_super = brief_examine_super_imsm,
4737ae25 12874 .brief_examine_subarrays = brief_examine_subarrays_imsm,
9d84c8ea 12875 .export_examine_super = export_examine_super_imsm,
cdddbdbc
DW		 12876 .detail_super = detail_super_imsm,
12877 .brief_detail_super = brief_detail_super_imsm,
bf5a934a 12878 .write_init_super = write_init_super_imsm,
0e600426
N		 12879 .validate_geometry = validate_geometry_imsm,
12880 .add_to_super = add_to_super_imsm,
1a64be56 12881 .remove_from_super = remove_from_super_imsm,
d665cc31 12882 .detail_platform = detail_platform_imsm,
e50cf220 12883 .export_detail_platform = export_detail_platform_imsm,
33414a01 12884 .kill_subarray = kill_subarray_imsm,
aa534678 12885 .update_subarray = update_subarray_imsm,
2b959fbf 12886 .load_container = load_container_imsm,
71204a50
N		 12887 .default_geometry = default_geometry_imsm,
12888 .get_disk_controller_domain = imsm_get_disk_controller_domain,
12889 .reshape_super = imsm_reshape_super,
12890 .manage_reshape = imsm_manage_reshape,
9e2d750d 12891 .recover_backup = recover_backup_imsm,
27156a57 12892 .examine_badblocks = examine_badblocks_imsm,
cdddbdbc
DW		 12893 .match_home = match_home_imsm,
12894 .uuid_from_super= uuid_from_super_imsm,
12895 .getinfo_super = getinfo_super_imsm,
5c4cd5da 12896 .getinfo_super_disks = getinfo_super_disks_imsm,
cdddbdbc
DW		 12897 .update_super = update_super_imsm,
12898
12899 .avail_size = avail_size_imsm,
fbfdcb06 12900 .get_spare_criteria = get_spare_criteria_imsm,
cdddbdbc
DW		 12901
12902 .compare_super = compare_super_imsm,
12903
12904 .load_super = load_super_imsm,
bf5a934a 12905 .init_super = init_super_imsm,
e683ca88 12906 .store_super = store_super_imsm,
cdddbdbc
DW		 12907 .free_super = free_super_imsm,
12908 .match_metadata_desc = match_metadata_desc_imsm,
bf5a934a 12909 .container_content = container_content_imsm,
0c21b485 12910 .validate_container = validate_container_imsm,
cdddbdbc 12911
fbc42556
JR		 12912 .add_internal_bitmap = add_internal_bitmap_imsm,
12913 .locate_bitmap = locate_bitmap_imsm,
12914 .write_bitmap = write_init_bitmap_imsm,
12915 .set_bitmap = set_bitmap_imsm,
12916
2432ce9b
AP		 12917 .write_init_ppl = write_init_ppl_imsm,
12918 .validate_ppl = validate_ppl_imsm,
12919
cdddbdbc 12920 .external = 1,
4cce4069 12921 .name = "imsm",
845dea95
NB		 12922
12923/* for mdmon */
12924 .open_new = imsm_open_new,
ed9d66aa 12925 .set_array_state= imsm_set_array_state,
845dea95
NB		 12926 .set_disk = imsm_set_disk,
12927 .sync_metadata = imsm_sync_metadata,
88758e9d 12928 .activate_spare = imsm_activate_spare,
e8319a19 12929 .process_update = imsm_process_update,
8273f55e 12930 .prepare_update = imsm_prepare_update,
6f50473f 12931 .record_bad_block = imsm_record_badblock,
c07a5a4f 12932 .clear_bad_block = imsm_clear_badblock,
928f1424 12933 .get_bad_blocks = imsm_get_badblocks,
cdddbdbc 12934};
Unnamed repository; edit this file 'description' to name the repository.