]> 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
Make get_component_size() work with named array.
[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
c2c087e6 91#define IMSM_RESERVED_SECTORS 4096
b81221b7 92#define NUM_BLOCKS_DIRTY_STRIPE_REGION 2056
979d38be 93#define SECT_PER_MB_SHIFT 11
f36a9ecd 94#define MAX_SECTOR_SIZE 4096
cdddbdbc
DW		 95
96/* Disk configuration info. */
97#define IMSM_MAX_DEVICES 255
98struct imsm_disk {
99 __u8 serial[MAX_RAID_SERIAL_LEN];/* 0xD8 - 0xE7 ascii serial number */
5551b113 100 __u32 total_blocks_lo; /* 0xE8 - 0xEB total blocks lo */
cdddbdbc 101 __u32 scsi_id; /* 0xEC - 0xEF scsi ID */
f2f27e63
DW		 102#define SPARE_DISK __cpu_to_le32(0x01) /* Spare */
103#define CONFIGURED_DISK __cpu_to_le32(0x02) /* Member of some RaidDev */
104#define FAILED_DISK __cpu_to_le32(0x04) /* Permanent failure */
cdddbdbc 105 __u32 status; /* 0xF0 - 0xF3 */
1011e834 106 __u32 owner_cfg_num; /* which config 0,1,2... owns this disk */
5551b113
CA		 107 __u32 total_blocks_hi; /* 0xF4 - 0xF5 total blocks hi */
108#define IMSM_DISK_FILLERS 3
109 __u32 filler[IMSM_DISK_FILLERS]; /* 0xF5 - 0x107 MPB_DISK_FILLERS for future expansion */
cdddbdbc
DW		 110};
111
3b451610
AK		 112/* map selector for map managment
113 */
238c0a71
AK		 114#define MAP_0 0
115#define MAP_1 1
116#define MAP_X -1
3b451610 117
cdddbdbc
DW		 118/* RAID map configuration infos. */
119struct imsm_map {
5551b113
CA		 120 __u32 pba_of_lba0_lo; /* start address of partition */
121 __u32 blocks_per_member_lo;/* blocks per member */
122 __u32 num_data_stripes_lo; /* number of data stripes */
cdddbdbc
DW		 123 __u16 blocks_per_strip;
124 __u8 map_state; /* Normal, Uninitialized, Degraded, Failed */
125#define IMSM_T_STATE_NORMAL 0
126#define IMSM_T_STATE_UNINITIALIZED 1
e3bba0e0
DW		 127#define IMSM_T_STATE_DEGRADED 2
128#define IMSM_T_STATE_FAILED 3
cdddbdbc
DW		 129 __u8 raid_level;
130#define IMSM_T_RAID0 0
131#define IMSM_T_RAID1 1
132#define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */
133 __u8 num_members; /* number of member disks */
fe7ed8cb
DW		 134 __u8 num_domains; /* number of parity domains */
135 __u8 failed_disk_num; /* valid only when state is degraded */
252d23c0 136 __u8 ddf;
5551b113
CA		 137 __u32 pba_of_lba0_hi;
138 __u32 blocks_per_member_hi;
139 __u32 num_data_stripes_hi;
140 __u32 filler[4]; /* expansion area */
7eef0453 141#define IMSM_ORD_REBUILD (1 << 24)
cdddbdbc 142 __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members],
7eef0453
DW		 143 * top byte contains some flags
144 */
cdddbdbc
DW		 145} __attribute__ ((packed));
146
147struct imsm_vol {
f8f603f1 148 __u32 curr_migr_unit;
fe7ed8cb 149 __u32 checkpoint_id; /* id to access curr_migr_unit */
cdddbdbc 150 __u8 migr_state; /* Normal or Migrating */
e3bba0e0
DW		 151#define MIGR_INIT 0
152#define MIGR_REBUILD 1
153#define MIGR_VERIFY 2 /* analagous to echo check > sync_action */
154#define MIGR_GEN_MIGR 3
155#define MIGR_STATE_CHANGE 4
1484e727 156#define MIGR_REPAIR 5
cdddbdbc
DW		 157 __u8 migr_type; /* Initializing, Rebuilding, ... */
158 __u8 dirty;
fe7ed8cb
DW		 159 __u8 fs_state; /* fast-sync state for CnG (0xff == disabled) */
160 __u16 verify_errors; /* number of mismatches */
161 __u16 bad_blocks; /* number of bad blocks during verify */
162 __u32 filler[4];
cdddbdbc
DW		 163 struct imsm_map map[1];
164 /* here comes another one if migr_state */
165} __attribute__ ((packed));
166
167struct imsm_dev {
fe7ed8cb 168 __u8 volume[MAX_RAID_SERIAL_LEN];
cdddbdbc
DW		 169 __u32 size_low;
170 __u32 size_high;
fe7ed8cb
DW		 171#define DEV_BOOTABLE __cpu_to_le32(0x01)
172#define DEV_BOOT_DEVICE __cpu_to_le32(0x02)
173#define DEV_READ_COALESCING __cpu_to_le32(0x04)
174#define DEV_WRITE_COALESCING __cpu_to_le32(0x08)
175#define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10)
176#define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20)
177#define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40)
178#define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80)
179#define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100)
180#define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200)
181#define DEV_CLONE_N_GO __cpu_to_le32(0x400)
182#define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800)
183#define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000)
cdddbdbc
DW		 184 __u32 status; /* Persistent RaidDev status */
185 __u32 reserved_blocks; /* Reserved blocks at beginning of volume */
fe7ed8cb
DW		 186 __u8 migr_priority;
187 __u8 num_sub_vols;
188 __u8 tid;
189 __u8 cng_master_disk;
190 __u16 cache_policy;
191 __u8 cng_state;
192 __u8 cng_sub_state;
193#define IMSM_DEV_FILLERS 10
cdddbdbc
DW		 194 __u32 filler[IMSM_DEV_FILLERS];
195 struct imsm_vol vol;
196} __attribute__ ((packed));
197
198struct imsm_super {
199 __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */
200 __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */
201 __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */
202 __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */
203 __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */
604b746f
JD		 204 __u32 error_log_size; /* 0x30 - 0x33 in bytes */
205 __u32 attributes; /* 0x34 - 0x37 */
cdddbdbc
DW		 206 __u8 num_disks; /* 0x38 Number of configured disks */
207 __u8 num_raid_devs; /* 0x39 Number of configured volumes */
604b746f
JD		 208 __u8 error_log_pos; /* 0x3A */
209 __u8 fill[1]; /* 0x3B */
210 __u32 cache_size; /* 0x3c - 0x40 in mb */
211 __u32 orig_family_num; /* 0x40 - 0x43 original family num */
212 __u32 pwr_cycle_count; /* 0x44 - 0x47 simulated power cycle count for array */
213 __u32 bbm_log_size; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */
214#define IMSM_FILLERS 35
215 __u32 filler[IMSM_FILLERS]; /* 0x4C - 0xD7 RAID_MPB_FILLERS */
cdddbdbc
DW		 216 struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */
217 /* here comes imsm_dev[num_raid_devs] */
604b746f 218 /* here comes BBM logs */
cdddbdbc
DW		 219} __attribute__ ((packed));
220
604b746f 221#define BBM_LOG_MAX_ENTRIES 254
8d67477f
TM		 222#define BBM_LOG_MAX_LBA_ENTRY_VAL 256 /* Represents 256 LBAs */
223#define BBM_LOG_SIGNATURE 0xabadb10c
224
225struct bbm_log_block_addr {
226 __u16 w1;
227 __u32 dw1;
228} __attribute__ ((__packed__));
604b746f
JD		 229
230struct bbm_log_entry {
8d67477f
TM		 231 __u8 marked_count; /* Number of blocks marked - 1 */
232 __u8 disk_ordinal; /* Disk entry within the imsm_super */
233 struct bbm_log_block_addr defective_block_start;
604b746f
JD		 234} __attribute__ ((__packed__));
235
236struct bbm_log {
237 __u32 signature; /* 0xABADB10C */
238 __u32 entry_count;
8d67477f 239 struct bbm_log_entry marked_block_entries[BBM_LOG_MAX_ENTRIES];
604b746f
JD		 240} __attribute__ ((__packed__));
241
cdddbdbc
DW		 242#ifndef MDASSEMBLE
243static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" };
244#endif
245
8e59f3d8
AK		 246#define RAID_DISK_RESERVED_BLOCKS_IMSM_HI 2209
247
248#define GEN_MIGR_AREA_SIZE 2048 /* General Migration Copy Area size in blocks */
249
de44e46f
PB		 250#define MIGR_REC_BUF_SECTORS 1 /* size of migr_record i/o buffer in sectors */
251#define MIGR_REC_SECTOR_POSITION 1 /* migr_record position offset on disk,
252 * MIGR_REC_BUF_SECTORS <= MIGR_REC_SECTOR_POS
17a4eaf9
AK		 253 */
254
8e59f3d8
AK		 255#define UNIT_SRC_NORMAL 0 /* Source data for curr_migr_unit must
256 * be recovered using srcMap */
257#define UNIT_SRC_IN_CP_AREA 1 /* Source data for curr_migr_unit has
258 * already been migrated and must
259 * be recovered from checkpoint area */
260struct migr_record {
261 __u32 rec_status; /* Status used to determine how to restart
262 * migration in case it aborts
263 * in some fashion */
264 __u32 curr_migr_unit; /* 0..numMigrUnits-1 */
265 __u32 family_num; /* Family number of MPB
266 * containing the RaidDev
267 * that is migrating */
268 __u32 ascending_migr; /* True if migrating in increasing
269 * order of lbas */
270 __u32 blocks_per_unit; /* Num disk blocks per unit of operation */
271 __u32 dest_depth_per_unit; /* Num member blocks each destMap
272 * member disk
273 * advances per unit-of-operation */
274 __u32 ckpt_area_pba; /* Pba of first block of ckpt copy area */
275 __u32 dest_1st_member_lba; /* First member lba on first
276 * stripe of destination */
277 __u32 num_migr_units; /* Total num migration units-of-op */
278 __u32 post_migr_vol_cap; /* Size of volume after
279 * migration completes */
280 __u32 post_migr_vol_cap_hi; /* Expansion space for LBA64 */
281 __u32 ckpt_read_disk_num; /* Which member disk in destSubMap[0] the
282 * migration ckpt record was read from
283 * (for recovered migrations) */
284} __attribute__ ((__packed__));
285
ec50f7b6
LM		 286struct md_list {
287 /* usage marker:
288 * 1: load metadata
289 * 2: metadata does not match
290 * 4: already checked
291 */
292 int used;
293 char *devname;
294 int found;
295 int container;
296 dev_t st_rdev;
297 struct md_list *next;
298};
299
e7b84f9d 300#define pr_vrb(fmt, arg...) (void) (verbose && pr_err(fmt, ##arg))
ec50f7b6 301
1484e727
DW		 302static __u8 migr_type(struct imsm_dev *dev)
303{
304 if (dev->vol.migr_type == MIGR_VERIFY &&
305 dev->status & DEV_VERIFY_AND_FIX)
306 return MIGR_REPAIR;
307 else
308 return dev->vol.migr_type;
309}
310
311static void set_migr_type(struct imsm_dev *dev, __u8 migr_type)
312{
313 /* for compatibility with older oroms convert MIGR_REPAIR, into
314 * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status
315 */
316 if (migr_type == MIGR_REPAIR) {
317 dev->vol.migr_type = MIGR_VERIFY;
318 dev->status |= DEV_VERIFY_AND_FIX;
319 } else {
320 dev->vol.migr_type = migr_type;
321 dev->status &= ~DEV_VERIFY_AND_FIX;
322 }
323}
324
f36a9ecd 325static unsigned int sector_count(__u32 bytes, unsigned int sector_size)
cdddbdbc 326{
f36a9ecd 327 return ROUND_UP(bytes, sector_size) / sector_size;
87eb16df 328}
cdddbdbc 329
f36a9ecd
PB		 330static unsigned int mpb_sectors(struct imsm_super *mpb,
331 unsigned int sector_size)
87eb16df 332{
f36a9ecd 333 return sector_count(__le32_to_cpu(mpb->mpb_size), sector_size);
cdddbdbc
DW		 334}
335
ba2de7ba
DW		 336struct intel_dev {
337 struct imsm_dev *dev;
338 struct intel_dev *next;
f21e18ca 339 unsigned index;
ba2de7ba
DW		 340};
341
88654014
LM		 342struct intel_hba {
343 enum sys_dev_type type;
344 char *path;
345 char *pci_id;
346 struct intel_hba *next;
347};
348
1a64be56
LM		 349enum action {
350 DISK_REMOVE = 1,
351 DISK_ADD
352};
cdddbdbc
DW		 353/* internal representation of IMSM metadata */
354struct intel_super {
355 union {
949c47a0
DW		 356 void *buf; /* O_DIRECT buffer for reading/writing metadata */
357 struct imsm_super *anchor; /* immovable parameters */
cdddbdbc 358 };
8e59f3d8
AK		 359 union {
360 void *migr_rec_buf; /* buffer for I/O operations */
361 struct migr_record *migr_rec; /* migration record */
362 };
51d83f5d
AK		 363 int clean_migration_record_by_mdmon; /* when reshape is switched to next
364 array, it indicates that mdmon is allowed to clean migration
365 record */
949c47a0 366 size_t len; /* size of the 'buf' allocation */
bbab0940 367 size_t extra_space; /* extra space in 'buf' that is not used yet */
4d7b1503
DW		 368 void *next_buf; /* for realloc'ing buf from the manager */
369 size_t next_len;
c2c087e6 370 int updates_pending; /* count of pending updates for mdmon */
bf5a934a 371 int current_vol; /* index of raid device undergoing creation */
5551b113 372 unsigned long long create_offset; /* common start for 'current_vol' */
148acb7b 373 __u32 random; /* random data for seeding new family numbers */
ba2de7ba 374 struct intel_dev *devlist;
fa7bb6f8 375 unsigned int sector_size; /* sector size of used member drives */
cdddbdbc
DW		 376 struct dl {
377 struct dl *next;
378 int index;
379 __u8 serial[MAX_RAID_SERIAL_LEN];
380 int major, minor;
381 char *devname;
b9f594fe 382 struct imsm_disk disk;
cdddbdbc 383 int fd;
0dcecb2e
DW		 384 int extent_cnt;
385 struct extent *e; /* for determining freespace @ create */
efb30e7f 386 int raiddisk; /* slot to fill in autolayout */
1a64be56 387 enum action action;
ca0748fa 388 } *disks, *current_disk;
1a64be56
LM		 389 struct dl *disk_mgmt_list; /* list of disks to add/remove while mdmon
390 active */
47ee5a45 391 struct dl *missing; /* disks removed while we weren't looking */
43dad3d6 392 struct bbm_log *bbm_log;
88654014 393 struct intel_hba *hba; /* device path of the raid controller for this metadata */
88c32bb1 394 const struct imsm_orom *orom; /* platform firmware support */
a2b97981 395 struct intel_super *next; /* (temp) list for disambiguating family_num */
928f1424 396 struct md_bb bb; /* memory for get_bad_blocks call */
a2b97981
DW		 397};
398
399struct intel_disk {
400 struct imsm_disk disk;
401 #define IMSM_UNKNOWN_OWNER (-1)
402 int owner;
403 struct intel_disk *next;
cdddbdbc
DW		 404};
405
c2c087e6
DW		 406struct extent {
407 unsigned long long start, size;
408};
409
694575e7
KW		 410/* definitions of reshape process types */
411enum imsm_reshape_type {
412 CH_TAKEOVER,
b5347799 413 CH_MIGRATION,
7abc9871 414 CH_ARRAY_SIZE,
694575e7
KW		 415};
416
88758e9d
DW		 417/* definition of messages passed to imsm_process_update */
418enum imsm_update_type {
419 update_activate_spare,
8273f55e 420 update_create_array,
33414a01 421 update_kill_array,
aa534678 422 update_rename_array,
1a64be56 423 update_add_remove_disk,
78b10e66 424 update_reshape_container_disks,
48c5303a 425 update_reshape_migration,
2d40f3a1
AK		 426 update_takeover,
427 update_general_migration_checkpoint,
f3871fdc 428 update_size_change,
bbab0940 429 update_prealloc_badblocks_mem,
88758e9d
DW		 430};
431
432struct imsm_update_activate_spare {
433 enum imsm_update_type type;
d23fe947 434 struct dl *dl;
88758e9d
DW		 435 int slot;
436 int array;
437 struct imsm_update_activate_spare *next;
438};
439
78b10e66 440struct geo_params {
4dd2df09 441 char devnm[32];
78b10e66 442 char *dev_name;
d04f65f4 443 unsigned long long size;
78b10e66
N		 444 int level;
445 int layout;
446 int chunksize;
447 int raid_disks;
448};
449
bb025c2f
KW		 450enum takeover_direction {
451 R10_TO_R0,
452 R0_TO_R10
453};
454struct imsm_update_takeover {
455 enum imsm_update_type type;
456 int subarray;
457 enum takeover_direction direction;
458};
78b10e66
N		 459
460struct imsm_update_reshape {
461 enum imsm_update_type type;
462 int old_raid_disks;
463 int new_raid_disks;
48c5303a
PC		 464
465 int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */
466};
467
468struct imsm_update_reshape_migration {
469 enum imsm_update_type type;
470 int old_raid_disks;
471 int new_raid_disks;
472 /* fields for array migration changes
473 */
474 int subdev;
475 int new_level;
476 int new_layout;
4bba0439 477 int new_chunksize;
48c5303a 478
d195167d 479 int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */
78b10e66
N		 480};
481
f3871fdc
AK		 482struct imsm_update_size_change {
483 enum imsm_update_type type;
484 int subdev;
485 long long new_size;
486};
487
2d40f3a1
AK		 488struct imsm_update_general_migration_checkpoint {
489 enum imsm_update_type type;
490 __u32 curr_migr_unit;
491};
492
54c2c1ea
DW		 493struct disk_info {
494 __u8 serial[MAX_RAID_SERIAL_LEN];
495};
496
8273f55e
DW		 497struct imsm_update_create_array {
498 enum imsm_update_type type;
8273f55e 499 int dev_idx;
6a3e913e 500 struct imsm_dev dev;
8273f55e
DW		 501};
502
33414a01
DW		 503struct imsm_update_kill_array {
504 enum imsm_update_type type;
505 int dev_idx;
506};
507
aa534678
DW		 508struct imsm_update_rename_array {
509 enum imsm_update_type type;
510 __u8 name[MAX_RAID_SERIAL_LEN];
511 int dev_idx;
512};
513
1a64be56 514struct imsm_update_add_remove_disk {
43dad3d6
DW		 515 enum imsm_update_type type;
516};
517
bbab0940
TM		 518struct imsm_update_prealloc_bb_mem {
519 enum imsm_update_type type;
520};
521
88654014
LM		 522static const char *_sys_dev_type[] = {
523 [SYS_DEV_UNKNOWN] = "Unknown",
524 [SYS_DEV_SAS] = "SAS",
614902f6 525 [SYS_DEV_SATA] = "SATA",
60f0f54d
PB		 526 [SYS_DEV_NVME] = "NVMe",
527 [SYS_DEV_VMD] = "VMD"
88654014
LM		 528};
529
530const char *get_sys_dev_type(enum sys_dev_type type)
531{
532 if (type >= SYS_DEV_MAX)
533 type = SYS_DEV_UNKNOWN;
534
535 return _sys_dev_type[type];
536}
537
538static struct intel_hba * alloc_intel_hba(struct sys_dev *device)
539{
503975b9
N		 540 struct intel_hba *result = xmalloc(sizeof(*result));
541
542 result->type = device->type;
543 result->path = xstrdup(device->path);
544 result->next = NULL;
545 if (result->path && (result->pci_id = strrchr(result->path, '/')) != NULL)
546 result->pci_id++;
547
88654014
LM		 548 return result;
549}
550
551static struct intel_hba * find_intel_hba(struct intel_hba *hba, struct sys_dev *device)
552{
594dc1b8
JS		 553 struct intel_hba *result;
554
88654014
LM		 555 for (result = hba; result; result = result->next) {
556 if (result->type == device->type && strcmp(result->path, device->path) == 0)
557 break;
558 }
559 return result;
560}
561
b4cf4cba 562static int attach_hba_to_super(struct intel_super *super, struct sys_dev *device)
88654014
LM		 563{
564 struct intel_hba *hba;
565
566 /* check if disk attached to Intel HBA */
567 hba = find_intel_hba(super->hba, device);
568 if (hba != NULL)
569 return 1;
570 /* Check if HBA is already attached to super */
571 if (super->hba == NULL) {
572 super->hba = alloc_intel_hba(device);
573 return 1;
6b781d33
AP		 574 }
575
576 hba = super->hba;
577 /* Intel metadata allows for all disks attached to the same type HBA.
614902f6 578 * Do not support HBA types mixing
6b781d33
AP		 579 */
580 if (device->type != hba->type)
88654014 581 return 2;
6b781d33
AP		 582
583 /* Multiple same type HBAs can be used if they share the same OROM */
584 const struct imsm_orom *device_orom = get_orom_by_device_id(device->dev_id);
585
586 if (device_orom != super->orom)
587 return 2;
588
589 while (hba->next)
590 hba = hba->next;
591
592 hba->next = alloc_intel_hba(device);
593 return 1;
88654014
LM		 594}
595
596static struct sys_dev* find_disk_attached_hba(int fd, const char *devname)
597{
9bc4ae77 598 struct sys_dev *list, *elem;
88654014
LM		 599 char *disk_path;
600
601 if ((list = find_intel_devices()) == NULL)
602 return 0;
603
604 if (fd < 0)
605 disk_path = (char *) devname;
606 else
607 disk_path = diskfd_to_devpath(fd);
608
9bc4ae77 609 if (!disk_path)
88654014 610 return 0;
88654014 611
9bc4ae77
N		 612 for (elem = list; elem; elem = elem->next)
613 if (path_attached_to_hba(disk_path, elem->path))
88654014 614 return elem;
9bc4ae77 615
88654014
LM		 616 if (disk_path != devname)
617 free(disk_path);
88654014
LM		 618
619 return NULL;
620}
621
d424212e
N		 622static int find_intel_hba_capability(int fd, struct intel_super *super,
623 char *devname);
f2f5c343 624
cdddbdbc
DW		 625static struct supertype *match_metadata_desc_imsm(char *arg)
626{
627 struct supertype *st;
628
629 if (strcmp(arg, "imsm") != 0 &&
630 strcmp(arg, "default") != 0
631 )
632 return NULL;
633
503975b9 634 st = xcalloc(1, sizeof(*st));
cdddbdbc
DW		 635 st->ss = &super_imsm;
636 st->max_devs = IMSM_MAX_DEVICES;
637 st->minor_version = 0;
638 st->sb = NULL;
639 return st;
640}
641
0e600426 642#ifndef MDASSEMBLE
cdddbdbc
DW		 643static __u8 *get_imsm_version(struct imsm_super *mpb)
644{
645 return &mpb->sig[MPB_SIG_LEN];
646}
9e2d750d 647#endif
cdddbdbc 648
949c47a0
DW		 649/* retrieve a disk directly from the anchor when the anchor is known to be
650 * up-to-date, currently only at load time
651 */
652static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index)
cdddbdbc 653{
949c47a0 654 if (index >= mpb->num_disks)
cdddbdbc
DW		 655 return NULL;
656 return &mpb->disk[index];
657}
658
95d07a2c
LM		 659/* retrieve the disk description based on a index of the disk
660 * in the sub-array
661 */
662static struct dl *get_imsm_dl_disk(struct intel_super *super, __u8 index)
949c47a0 663{
b9f594fe
DW		 664 struct dl *d;
665
666 for (d = super->disks; d; d = d->next)
667 if (d->index == index)
95d07a2c
LM		 668 return d;
669
670 return NULL;
671}
672/* retrieve a disk from the parsed metadata */
673static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index)
674{
675 struct dl *dl;
676
677 dl = get_imsm_dl_disk(super, index);
678 if (dl)
679 return &dl->disk;
680
b9f594fe 681 return NULL;
949c47a0
DW		 682}
683
684/* generate a checksum directly from the anchor when the anchor is known to be
685 * up-to-date, currently only at load or write_super after coalescing
686 */
687static __u32 __gen_imsm_checksum(struct imsm_super *mpb)
cdddbdbc
DW		 688{
689 __u32 end = mpb->mpb_size / sizeof(end);
690 __u32 *p = (__u32 *) mpb;
691 __u32 sum = 0;
692
5d500228
N		 693 while (end--) {
694 sum += __le32_to_cpu(*p);
97f734fd
N		 695 p++;
696 }
cdddbdbc 697
5d500228 698 return sum - __le32_to_cpu(mpb->check_sum);
cdddbdbc
DW		 699}
700
a965f303
DW		 701static size_t sizeof_imsm_map(struct imsm_map *map)
702{
703 return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1);
704}
705
706struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map)
cdddbdbc 707{
5e7b0330
AK		 708 /* A device can have 2 maps if it is in the middle of a migration.
709 * If second_map is:
238c0a71
AK		 710 * MAP_0 - we return the first map
711 * MAP_1 - we return the second map if it exists, else NULL
712 * MAP_X - we return the second map if it exists, else the first
5e7b0330 713 */
a965f303 714 struct imsm_map *map = &dev->vol.map[0];
9535fc47 715 struct imsm_map *map2 = NULL;
a965f303 716
9535fc47
AK		 717 if (dev->vol.migr_state)
718 map2 = (void *)map + sizeof_imsm_map(map);
a965f303 719
9535fc47 720 switch (second_map) {
3b451610 721 case MAP_0:
9535fc47 722 break;
3b451610 723 case MAP_1:
9535fc47
AK		 724 map = map2;
725 break;
238c0a71 726 case MAP_X:
9535fc47
AK		 727 if (map2)
728 map = map2;
729 break;
9535fc47
AK		 730 default:
731 map = NULL;
732 }
733 return map;
5e7b0330 734
a965f303 735}
cdddbdbc 736
3393c6af
DW		 737/* return the size of the device.
738 * migr_state increases the returned size if map[0] were to be duplicated
739 */
740static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state)
a965f303
DW		 741{
742 size_t size = sizeof(*dev) - sizeof(struct imsm_map) +
238c0a71 743 sizeof_imsm_map(get_imsm_map(dev, MAP_0));
cdddbdbc
DW		 744
745 /* migrating means an additional map */
a965f303 746 if (dev->vol.migr_state)
238c0a71 747 size += sizeof_imsm_map(get_imsm_map(dev, MAP_1));
3393c6af 748 else if (migr_state)
238c0a71 749 size += sizeof_imsm_map(get_imsm_map(dev, MAP_0));
cdddbdbc
DW		 750
751 return size;
752}
753
54c2c1ea
DW		 754#ifndef MDASSEMBLE
755/* retrieve disk serial number list from a metadata update */
756static struct disk_info *get_disk_info(struct imsm_update_create_array *update)
757{
758 void *u = update;
759 struct disk_info *inf;
760
761 inf = u + sizeof(*update) - sizeof(struct imsm_dev) +
762 sizeof_imsm_dev(&update->dev, 0);
763
764 return inf;
765}
766#endif
767
949c47a0 768static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index)
cdddbdbc
DW		 769{
770 int offset;
771 int i;
772 void *_mpb = mpb;
773
949c47a0 774 if (index >= mpb->num_raid_devs)
cdddbdbc
DW		 775 return NULL;
776
777 /* devices start after all disks */
778 offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb;
779
780 for (i = 0; i <= index; i++)
781 if (i == index)
782 return _mpb + offset;
783 else
3393c6af 784 offset += sizeof_imsm_dev(_mpb + offset, 0);
cdddbdbc
DW		 785
786 return NULL;
787}
788
949c47a0
DW		 789static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index)
790{
ba2de7ba
DW		 791 struct intel_dev *dv;
792
949c47a0
DW		 793 if (index >= super->anchor->num_raid_devs)
794 return NULL;
ba2de7ba
DW		 795 for (dv = super->devlist; dv; dv = dv->next)
796 if (dv->index == index)
797 return dv->dev;
798 return NULL;
949c47a0
DW		 799}
800
8d67477f
TM		 801static inline unsigned long long __le48_to_cpu(const struct bbm_log_block_addr
802 *addr)
803{
804 return ((((__u64)__le32_to_cpu(addr->dw1)) << 16) |
805 __le16_to_cpu(addr->w1));
806}
807
808static inline struct bbm_log_block_addr __cpu_to_le48(unsigned long long sec)
809{
810 struct bbm_log_block_addr addr;
811
812 addr.w1 = __cpu_to_le16((__u16)(sec & 0xffff));
813 addr.dw1 = __cpu_to_le32((__u32)(sec >> 16) & 0xffffffff);
814 return addr;
815}
816
817#ifndef MDASSEMBLE
818/* get size of the bbm log */
819static __u32 get_imsm_bbm_log_size(struct bbm_log *log)
820{
821 if (!log || log->entry_count == 0)
822 return 0;
823
824 return sizeof(log->signature) +
825 sizeof(log->entry_count) +
826 log->entry_count * sizeof(struct bbm_log_entry);
827}
6f50473f
TM		 828
829/* check if bad block is not partially stored in bbm log */
830static int is_stored_in_bbm(struct bbm_log *log, const __u8 idx, const unsigned
831 long long sector, const int length, __u32 *pos)
832{
833 __u32 i;
834
835 for (i = *pos; i < log->entry_count; i++) {
836 struct bbm_log_entry *entry = &log->marked_block_entries[i];
837 unsigned long long bb_start;
838 unsigned long long bb_end;
839
840 bb_start = __le48_to_cpu(&entry->defective_block_start);
841 bb_end = bb_start + (entry->marked_count + 1);
842
843 if ((entry->disk_ordinal == idx) && (bb_start >= sector) &&
844 (bb_end <= sector + length)) {
845 *pos = i;
846 return 1;
847 }
848 }
849 return 0;
850}
851
852/* record new bad block in bbm log */
853static int record_new_badblock(struct bbm_log *log, const __u8 idx, unsigned
854 long long sector, int length)
855{
856 int new_bb = 0;
857 __u32 pos = 0;
858 struct bbm_log_entry *entry = NULL;
859
860 while (is_stored_in_bbm(log, idx, sector, length, &pos)) {
861 struct bbm_log_entry *e = &log->marked_block_entries[pos];
862
863 if ((e->marked_count + 1 == BBM_LOG_MAX_LBA_ENTRY_VAL) &&
864 (__le48_to_cpu(&e->defective_block_start) == sector)) {
865 sector += BBM_LOG_MAX_LBA_ENTRY_VAL;
866 length -= BBM_LOG_MAX_LBA_ENTRY_VAL;
867 pos = pos + 1;
868 continue;
869 }
870 entry = e;
871 break;
872 }
873
874 if (entry) {
875 int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length :
876 BBM_LOG_MAX_LBA_ENTRY_VAL;
877 entry->defective_block_start = __cpu_to_le48(sector);
878 entry->marked_count = cnt - 1;
879 if (cnt == length)
880 return 1;
881 sector += cnt;
882 length -= cnt;
883 }
884
885 new_bb = ROUND_UP(length, BBM_LOG_MAX_LBA_ENTRY_VAL) /
886 BBM_LOG_MAX_LBA_ENTRY_VAL;
887 if (log->entry_count + new_bb > BBM_LOG_MAX_ENTRIES)
888 return 0;
889
890 while (length > 0) {
891 int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length :
892 BBM_LOG_MAX_LBA_ENTRY_VAL;
893 struct bbm_log_entry *entry =
894 &log->marked_block_entries[log->entry_count];
895
896 entry->defective_block_start = __cpu_to_le48(sector);
897 entry->marked_count = cnt - 1;
898 entry->disk_ordinal = idx;
899
900 sector += cnt;
901 length -= cnt;
902
903 log->entry_count++;
904 }
905
906 return new_bb;
907}
c07a5a4f 908
4c9e8c1e
TM		 909/* clear all bad blocks for given disk */
910static void clear_disk_badblocks(struct bbm_log *log, const __u8 idx)
911{
912 __u32 i = 0;
913
914 while (i < log->entry_count) {
915 struct bbm_log_entry *entries = log->marked_block_entries;
916
917 if (entries[i].disk_ordinal == idx) {
918 if (i < log->entry_count - 1)
919 entries[i] = entries[log->entry_count - 1];
920 log->entry_count--;
921 } else {
922 i++;
923 }
924 }
925}
926
c07a5a4f
TM		 927/* clear given bad block */
928static int clear_badblock(struct bbm_log *log, const __u8 idx, const unsigned
929 long long sector, const int length) {
930 __u32 i = 0;
931
932 while (i < log->entry_count) {
933 struct bbm_log_entry *entries = log->marked_block_entries;
934
935 if ((entries[i].disk_ordinal == idx) &&
936 (__le48_to_cpu(&entries[i].defective_block_start) ==
937 sector) && (entries[i].marked_count + 1 == length)) {
938 if (i < log->entry_count - 1)
939 entries[i] = entries[log->entry_count - 1];
940 log->entry_count--;
941 break;
942 }
943 i++;
944 }
945
946 return 1;
947}
8d67477f
TM		 948#endif /* MDASSEMBLE */
949
950/* allocate and load BBM log from metadata */
951static int load_bbm_log(struct intel_super *super)
952{
953 struct imsm_super *mpb = super->anchor;
954 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
955
956 super->bbm_log = xcalloc(1, sizeof(struct bbm_log));
957 if (!super->bbm_log)
958 return 1;
959
960 if (bbm_log_size) {
961 struct bbm_log *log = (void *)mpb +
962 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
963
964 __u32 entry_count;
965
966 if (bbm_log_size < sizeof(log->signature) +
967 sizeof(log->entry_count))
968 return 2;
969
970 entry_count = __le32_to_cpu(log->entry_count);
971 if ((__le32_to_cpu(log->signature) != BBM_LOG_SIGNATURE) ||
972 (entry_count > BBM_LOG_MAX_ENTRIES))
973 return 3;
974
975 if (bbm_log_size !=
976 sizeof(log->signature) + sizeof(log->entry_count) +
977 entry_count * sizeof(struct bbm_log_entry))
978 return 4;
979
980 memcpy(super->bbm_log, log, bbm_log_size);
981 } else {
982 super->bbm_log->signature = __cpu_to_le32(BBM_LOG_SIGNATURE);
983 super->bbm_log->entry_count = 0;
984 }
985
986 return 0;
987}
988
b12796be
TM		 989/* checks if bad block is within volume boundaries */
990static int is_bad_block_in_volume(const struct bbm_log_entry *entry,
991 const unsigned long long start_sector,
992 const unsigned long long size)
993{
994 unsigned long long bb_start;
995 unsigned long long bb_end;
996
997 bb_start = __le48_to_cpu(&entry->defective_block_start);
998 bb_end = bb_start + (entry->marked_count + 1);
999
1000 if (((bb_start >= start_sector) && (bb_start < start_sector + size)) ||
1001 ((bb_end >= start_sector) && (bb_end <= start_sector + size)))
1002 return 1;
1003
1004 return 0;
1005}
1006
1007/* get list of bad blocks on a drive for a volume */
1008static void get_volume_badblocks(const struct bbm_log *log, const __u8 idx,
1009 const unsigned long long start_sector,
1010 const unsigned long long size,
1011 struct md_bb *bbs)
1012{
1013 __u32 count = 0;
1014 __u32 i;
1015
1016 for (i = 0; i < log->entry_count; i++) {
1017 const struct bbm_log_entry *ent =
1018 &log->marked_block_entries[i];
1019 struct md_bb_entry *bb;
1020
1021 if ((ent->disk_ordinal == idx) &&
1022 is_bad_block_in_volume(ent, start_sector, size)) {
1023
1024 if (!bbs->entries) {
1025 bbs->entries = xmalloc(BBM_LOG_MAX_ENTRIES *
1026 sizeof(*bb));
1027 if (!bbs->entries)
1028 break;
1029 }
1030
1031 bb = &bbs->entries[count++];
1032 bb->sector = __le48_to_cpu(&ent->defective_block_start);
1033 bb->length = ent->marked_count + 1;
1034 }
1035 }
1036 bbs->count = count;
1037}
1038
98130f40
AK		 1039/*
1040 * for second_map:
238c0a71
AK		 1041 * == MAP_0 get first map
1042 * == MAP_1 get second map
1043 * == MAP_X than get map according to the current migr_state
98130f40
AK		 1044 */
1045static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev,
1046 int slot,
1047 int second_map)
7eef0453
DW		 1048{
1049 struct imsm_map *map;
1050
5e7b0330 1051 map = get_imsm_map(dev, second_map);
7eef0453 1052
ff077194
DW		 1053 /* top byte identifies disk under rebuild */
1054 return __le32_to_cpu(map->disk_ord_tbl[slot]);
1055}
1056
1057#define ord_to_idx(ord) (((ord) << 8) >> 8)
98130f40 1058static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot, int second_map)
ff077194 1059{
98130f40 1060 __u32 ord = get_imsm_ord_tbl_ent(dev, slot, second_map);
ff077194
DW		 1061
1062 return ord_to_idx(ord);
7eef0453
DW		 1063}
1064
be73972f
DW		 1065static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord)
1066{
1067 map->disk_ord_tbl[slot] = __cpu_to_le32(ord);
1068}
1069
f21e18ca 1070static int get_imsm_disk_slot(struct imsm_map *map, unsigned idx)
620b1713
DW		 1071{
1072 int slot;
1073 __u32 ord;
1074
1075 for (slot = 0; slot < map->num_members; slot++) {
1076 ord = __le32_to_cpu(map->disk_ord_tbl[slot]);
1077 if (ord_to_idx(ord) == idx)
1078 return slot;
1079 }
1080
1081 return -1;
1082}
1083
cdddbdbc
DW		 1084static int get_imsm_raid_level(struct imsm_map *map)
1085{
1086 if (map->raid_level == 1) {
1087 if (map->num_members == 2)
1088 return 1;
1089 else
1090 return 10;
1091 }
1092
1093 return map->raid_level;
1094}
1095
c2c087e6
DW		 1096static int cmp_extent(const void *av, const void *bv)
1097{
1098 const struct extent *a = av;
1099 const struct extent *b = bv;
1100 if (a->start < b->start)
1101 return -1;
1102 if (a->start > b->start)
1103 return 1;
1104 return 0;
1105}
1106
0dcecb2e 1107static int count_memberships(struct dl *dl, struct intel_super *super)
c2c087e6 1108{
c2c087e6 1109 int memberships = 0;
620b1713 1110 int i;
c2c087e6 1111
949c47a0
DW		 1112 for (i = 0; i < super->anchor->num_raid_devs; i++) {
1113 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 1114 struct imsm_map *map = get_imsm_map(dev, MAP_0);
c2c087e6 1115
620b1713
DW		 1116 if (get_imsm_disk_slot(map, dl->index) >= 0)
1117 memberships++;
c2c087e6 1118 }
0dcecb2e
DW		 1119
1120 return memberships;
1121}
1122
b81221b7
CA		 1123static __u32 imsm_min_reserved_sectors(struct intel_super *super);
1124
5551b113
CA		 1125static int split_ull(unsigned long long n, __u32 *lo, __u32 *hi)
1126{
1127 if (lo == 0 || hi == 0)
1128 return 1;
1129 *lo = __le32_to_cpu((unsigned)n);
1130 *hi = __le32_to_cpu((unsigned)(n >> 32));
1131 return 0;
1132}
1133
1134static unsigned long long join_u32(__u32 lo, __u32 hi)
1135{
1136 return (unsigned long long)__le32_to_cpu(lo) |
1137 (((unsigned long long)__le32_to_cpu(hi)) << 32);
1138}
1139
1140static unsigned long long total_blocks(struct imsm_disk *disk)
1141{
1142 if (disk == NULL)
1143 return 0;
1144 return join_u32(disk->total_blocks_lo, disk->total_blocks_hi);
1145}
1146
1147static unsigned long long pba_of_lba0(struct imsm_map *map)
1148{
1149 if (map == NULL)
1150 return 0;
1151 return join_u32(map->pba_of_lba0_lo, map->pba_of_lba0_hi);
1152}
1153
1154static unsigned long long blocks_per_member(struct imsm_map *map)
1155{
1156 if (map == NULL)
1157 return 0;
1158 return join_u32(map->blocks_per_member_lo, map->blocks_per_member_hi);
1159}
1160
1161static unsigned long long num_data_stripes(struct imsm_map *map)
1162{
1163 if (map == NULL)
1164 return 0;
1165 return join_u32(map->num_data_stripes_lo, map->num_data_stripes_hi);
1166}
1167
1168static void set_total_blocks(struct imsm_disk *disk, unsigned long long n)
1169{
1170 split_ull(n, &disk->total_blocks_lo, &disk->total_blocks_hi);
1171}
1172
1173static void set_pba_of_lba0(struct imsm_map *map, unsigned long long n)
1174{
1175 split_ull(n, &map->pba_of_lba0_lo, &map->pba_of_lba0_hi);
1176}
1177
1178static void set_blocks_per_member(struct imsm_map *map, unsigned long long n)
1179{
1180 split_ull(n, &map->blocks_per_member_lo, &map->blocks_per_member_hi);
1181}
1182
1183static void set_num_data_stripes(struct imsm_map *map, unsigned long long n)
1184{
1185 split_ull(n, &map->num_data_stripes_lo, &map->num_data_stripes_hi);
1186}
1187
0dcecb2e
DW		 1188static struct extent *get_extents(struct intel_super *super, struct dl *dl)
1189{
1190 /* find a list of used extents on the given physical device */
1191 struct extent *rv, *e;
620b1713 1192 int i;
0dcecb2e 1193 int memberships = count_memberships(dl, super);
b276dd33
DW		 1194 __u32 reservation;
1195
1196 /* trim the reserved area for spares, so they can join any array
1197 * regardless of whether the OROM has assigned sectors from the
1198 * IMSM_RESERVED_SECTORS region
1199 */
1200 if (dl->index == -1)
b81221b7 1201 reservation = imsm_min_reserved_sectors(super);
b276dd33
DW		 1202 else
1203 reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
0dcecb2e 1204
503975b9 1205 rv = xcalloc(sizeof(struct extent), (memberships + 1));
c2c087e6
DW		 1206 e = rv;
1207
949c47a0
DW		 1208 for (i = 0; i < super->anchor->num_raid_devs; i++) {
1209 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 1210 struct imsm_map *map = get_imsm_map(dev, MAP_0);
c2c087e6 1211
620b1713 1212 if (get_imsm_disk_slot(map, dl->index) >= 0) {
5551b113
CA		 1213 e->start = pba_of_lba0(map);
1214 e->size = blocks_per_member(map);
620b1713 1215 e++;
c2c087e6
DW		 1216 }
1217 }
1218 qsort(rv, memberships, sizeof(*rv), cmp_extent);
1219
1011e834 1220 /* determine the start of the metadata
14e8215b
DW		 1221 * when no raid devices are defined use the default
1222 * ...otherwise allow the metadata to truncate the value
1223 * as is the case with older versions of imsm
1224 */
1225 if (memberships) {
1226 struct extent *last = &rv[memberships - 1];
5551b113 1227 unsigned long long remainder;
14e8215b 1228
5551b113 1229 remainder = total_blocks(&dl->disk) - (last->start + last->size);
dda5855f
DW		 1230 /* round down to 1k block to satisfy precision of the kernel
1231 * 'size' interface
1232 */
1233 remainder &= ~1UL;
1234 /* make sure remainder is still sane */
f21e18ca 1235 if (remainder < (unsigned)ROUND_UP(super->len, 512) >> 9)
dda5855f 1236 remainder = ROUND_UP(super->len, 512) >> 9;
14e8215b
DW		 1237 if (reservation > remainder)
1238 reservation = remainder;
1239 }
5551b113 1240 e->start = total_blocks(&dl->disk) - reservation;
c2c087e6
DW		 1241 e->size = 0;
1242 return rv;
1243}
1244
14e8215b
DW		 1245/* try to determine how much space is reserved for metadata from
1246 * the last get_extents() entry, otherwise fallback to the
1247 * default
1248 */
1249static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl)
1250{
1251 struct extent *e;
1252 int i;
1253 __u32 rv;
1254
1255 /* for spares just return a minimal reservation which will grow
1256 * once the spare is picked up by an array
1257 */
1258 if (dl->index == -1)
1259 return MPB_SECTOR_CNT;
1260
1261 e = get_extents(super, dl);
1262 if (!e)
1263 return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
1264
1265 /* scroll to last entry */
1266 for (i = 0; e[i].size; i++)
1267 continue;
1268
5551b113 1269 rv = total_blocks(&dl->disk) - e[i].start;
14e8215b
DW		 1270
1271 free(e);
1272
1273 return rv;
1274}
1275
25ed7e59
DW		 1276static int is_spare(struct imsm_disk *disk)
1277{
1278 return (disk->status & SPARE_DISK) == SPARE_DISK;
1279}
1280
1281static int is_configured(struct imsm_disk *disk)
1282{
1283 return (disk->status & CONFIGURED_DISK) == CONFIGURED_DISK;
1284}
1285
1286static int is_failed(struct imsm_disk *disk)
1287{
1288 return (disk->status & FAILED_DISK) == FAILED_DISK;
1289}
1290
b81221b7
CA		 1291/* try to determine how much space is reserved for metadata from
1292 * the last get_extents() entry on the smallest active disk,
1293 * otherwise fallback to the default
1294 */
1295static __u32 imsm_min_reserved_sectors(struct intel_super *super)
1296{
1297 struct extent *e;
1298 int i;
5551b113
CA		 1299 unsigned long long min_active;
1300 __u32 remainder;
b81221b7
CA		 1301 __u32 rv = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
1302 struct dl *dl, *dl_min = NULL;
1303
1304 if (!super)
1305 return rv;
1306
1307 min_active = 0;
1308 for (dl = super->disks; dl; dl = dl->next) {
1309 if (dl->index < 0)
1310 continue;
5551b113
CA		 1311 unsigned long long blocks = total_blocks(&dl->disk);
1312 if (blocks < min_active || min_active == 0) {
b81221b7 1313 dl_min = dl;
5551b113 1314 min_active = blocks;
b81221b7
CA		 1315 }
1316 }
1317 if (!dl_min)
1318 return rv;
1319
1320 /* find last lba used by subarrays on the smallest active disk */
1321 e = get_extents(super, dl_min);
1322 if (!e)
1323 return rv;
1324 for (i = 0; e[i].size; i++)
1325 continue;
1326
1327 remainder = min_active - e[i].start;
1328 free(e);
1329
1330 /* to give priority to recovery we should not require full
1331 IMSM_RESERVED_SECTORS from the spare */
1332 rv = MPB_SECTOR_CNT + NUM_BLOCKS_DIRTY_STRIPE_REGION;
1333
1334 /* if real reservation is smaller use that value */
1335 return (remainder < rv) ? remainder : rv;
1336}
1337
80e7f8c3
AC		 1338/* Return minimum size of a spare that can be used in this array*/
1339static unsigned long long min_acceptable_spare_size_imsm(struct supertype *st)
1340{
1341 struct intel_super *super = st->sb;
1342 struct dl *dl;
1343 struct extent *e;
1344 int i;
1345 unsigned long long rv = 0;
1346
1347 if (!super)
1348 return rv;
1349 /* find first active disk in array */
1350 dl = super->disks;
1351 while (dl && (is_failed(&dl->disk) || dl->index == -1))
1352 dl = dl->next;
1353 if (!dl)
1354 return rv;
1355 /* find last lba used by subarrays */
1356 e = get_extents(super, dl);
1357 if (!e)
1358 return rv;
1359 for (i = 0; e[i].size; i++)
1360 continue;
1361 if (i > 0)
1362 rv = e[i-1].start + e[i-1].size;
1363 free(e);
b81221b7 1364
80e7f8c3 1365 /* add the amount of space needed for metadata */
b81221b7
CA		 1366 rv = rv + imsm_min_reserved_sectors(super);
1367
80e7f8c3
AC		 1368 return rv * 512;
1369}
1370
d1e02575
AK		 1371static int is_gen_migration(struct imsm_dev *dev);
1372
f36a9ecd
PB		 1373#define IMSM_4K_DIV 8
1374
1799c9e8 1375#ifndef MDASSEMBLE
c47b0ff6
AK		 1376static __u64 blocks_per_migr_unit(struct intel_super *super,
1377 struct imsm_dev *dev);
1e5c6983 1378
c47b0ff6
AK		 1379static void print_imsm_dev(struct intel_super *super,
1380 struct imsm_dev *dev,
1381 char *uuid,
1382 int disk_idx)
cdddbdbc
DW		 1383{
1384 __u64 sz;
0d80bb2f 1385 int slot, i;
238c0a71
AK		 1386 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1387 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
b10b37b8 1388 __u32 ord;
cdddbdbc
DW		 1389
1390 printf("\n");
1e7bc0ed 1391 printf("[%.16s]:\n", dev->volume);
44470971 1392 printf(" UUID : %s\n", uuid);
dd8bcb3b
AK		 1393 printf(" RAID Level : %d", get_imsm_raid_level(map));
1394 if (map2)
1395 printf(" <-- %d", get_imsm_raid_level(map2));
1396 printf("\n");
1397 printf(" Members : %d", map->num_members);
1398 if (map2)
1399 printf(" <-- %d", map2->num_members);
1400 printf("\n");
0d80bb2f
DW		 1401 printf(" Slots : [");
1402 for (i = 0; i < map->num_members; i++) {
238c0a71 1403 ord = get_imsm_ord_tbl_ent(dev, i, MAP_0);
0d80bb2f
DW		 1404 printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U");
1405 }
dd8bcb3b
AK		 1406 printf("]");
1407 if (map2) {
1408 printf(" <-- [");
1409 for (i = 0; i < map2->num_members; i++) {
238c0a71 1410 ord = get_imsm_ord_tbl_ent(dev, i, MAP_1);
dd8bcb3b
AK		 1411 printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U");
1412 }
1413 printf("]");
1414 }
1415 printf("\n");
7095bccb
AK		 1416 printf(" Failed disk : ");
1417 if (map->failed_disk_num == 0xff)
1418 printf("none");
1419 else
1420 printf("%i", map->failed_disk_num);
1421 printf("\n");
620b1713
DW		 1422 slot = get_imsm_disk_slot(map, disk_idx);
1423 if (slot >= 0) {
238c0a71 1424 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
b10b37b8
DW		 1425 printf(" This Slot : %d%s\n", slot,
1426 ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : "");
1427 } else
cdddbdbc
DW		 1428 printf(" This Slot : ?\n");
1429 sz = __le32_to_cpu(dev->size_high);
1430 sz <<= 32;
1431 sz += __le32_to_cpu(dev->size_low);
1432 printf(" Array Size : %llu%s\n", (unsigned long long)sz,
1433 human_size(sz * 512));
5551b113 1434 sz = blocks_per_member(map);
cdddbdbc
DW		 1435 printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz,
1436 human_size(sz * 512));
5551b113
CA		 1437 printf(" Sector Offset : %llu\n",
1438 pba_of_lba0(map));
1439 printf(" Num Stripes : %llu\n",
1440 num_data_stripes(map));
dd8bcb3b 1441 printf(" Chunk Size : %u KiB",
cdddbdbc 1442 __le16_to_cpu(map->blocks_per_strip) / 2);
dd8bcb3b
AK		 1443 if (map2)
1444 printf(" <-- %u KiB",
1445 __le16_to_cpu(map2->blocks_per_strip) / 2);
1446 printf("\n");
cdddbdbc 1447 printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks));
8655a7b1 1448 printf(" Migrate State : ");
1484e727
DW		 1449 if (dev->vol.migr_state) {
1450 if (migr_type(dev) == MIGR_INIT)
8655a7b1 1451 printf("initialize\n");
1484e727 1452 else if (migr_type(dev) == MIGR_REBUILD)
8655a7b1 1453 printf("rebuild\n");
1484e727 1454 else if (migr_type(dev) == MIGR_VERIFY)
8655a7b1 1455 printf("check\n");
1484e727 1456 else if (migr_type(dev) == MIGR_GEN_MIGR)
8655a7b1 1457 printf("general migration\n");
1484e727 1458 else if (migr_type(dev) == MIGR_STATE_CHANGE)
8655a7b1 1459 printf("state change\n");
1484e727 1460 else if (migr_type(dev) == MIGR_REPAIR)
8655a7b1 1461 printf("repair\n");
1484e727 1462 else
8655a7b1
DW		 1463 printf("<unknown:%d>\n", migr_type(dev));
1464 } else
1465 printf("idle\n");
3393c6af
DW		 1466 printf(" Map State : %s", map_state_str[map->map_state]);
1467 if (dev->vol.migr_state) {
238c0a71 1468 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983 1469
b10b37b8 1470 printf(" <-- %s", map_state_str[map->map_state]);
464d40e8
LD		 1471 printf("\n Checkpoint : %u ",
1472 __le32_to_cpu(dev->vol.curr_migr_unit));
089f9d79 1473 if (is_gen_migration(dev) && (slot > 1 || slot < 0))
464d40e8
LD		 1474 printf("(N/A)");
1475 else
1476 printf("(%llu)", (unsigned long long)
1477 blocks_per_migr_unit(super, dev));
3393c6af
DW		 1478 }
1479 printf("\n");
cdddbdbc 1480 printf(" Dirty State : %s\n", dev->vol.dirty ? "dirty" : "clean");
cdddbdbc
DW		 1481}
1482
0ec1f4e8 1483static void print_imsm_disk(struct imsm_disk *disk, int index, __u32 reserved)
cdddbdbc 1484{
1f24f035 1485 char str[MAX_RAID_SERIAL_LEN + 1];
cdddbdbc
DW		 1486 __u64 sz;
1487
0ec1f4e8 1488 if (index < -1 || !disk)
e9d82038
DW		 1489 return;
1490
cdddbdbc 1491 printf("\n");
1f24f035 1492 snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial);
0ec1f4e8
DW		 1493 if (index >= 0)
1494 printf(" Disk%02d Serial : %s\n", index, str);
1495 else
1496 printf(" Disk Serial : %s\n", str);
25ed7e59
DW		 1497 printf(" State :%s%s%s\n", is_spare(disk) ? " spare" : "",
1498 is_configured(disk) ? " active" : "",
1499 is_failed(disk) ? " failed" : "");
cdddbdbc 1500 printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id));
5551b113 1501 sz = total_blocks(disk) - reserved;
cdddbdbc
DW		 1502 printf(" Usable Size : %llu%s\n", (unsigned long long)sz,
1503 human_size(sz * 512));
1504}
1505
de44e46f
PB		 1506void convert_to_4k_imsm_migr_rec(struct intel_super *super)
1507{
1508 struct migr_record *migr_rec = super->migr_rec;
1509
1510 migr_rec->blocks_per_unit /= IMSM_4K_DIV;
1511 migr_rec->ckpt_area_pba /= IMSM_4K_DIV;
1512 migr_rec->dest_1st_member_lba /= IMSM_4K_DIV;
1513 migr_rec->dest_depth_per_unit /= IMSM_4K_DIV;
1514 split_ull((join_u32(migr_rec->post_migr_vol_cap,
1515 migr_rec->post_migr_vol_cap_hi) / IMSM_4K_DIV),
1516 &migr_rec->post_migr_vol_cap, &migr_rec->post_migr_vol_cap_hi);
1517}
1518
f36a9ecd
PB		 1519void convert_to_4k_imsm_disk(struct imsm_disk *disk)
1520{
1521 set_total_blocks(disk, (total_blocks(disk)/IMSM_4K_DIV));
1522}
1523
1524void convert_to_4k(struct intel_super *super)
1525{
1526 struct imsm_super *mpb = super->anchor;
1527 struct imsm_disk *disk;
1528 int i;
e4467bc7 1529 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
f36a9ecd
PB		 1530
1531 for (i = 0; i < mpb->num_disks ; i++) {
1532 disk = __get_imsm_disk(mpb, i);
1533 /* disk */
1534 convert_to_4k_imsm_disk(disk);
1535 }
1536 for (i = 0; i < mpb->num_raid_devs; i++) {
1537 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
1538 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1539 /* dev */
1540 split_ull((join_u32(dev->size_low, dev->size_high)/IMSM_4K_DIV),
1541 &dev->size_low, &dev->size_high);
1542 dev->vol.curr_migr_unit /= IMSM_4K_DIV;
1543
1544 /* map0 */
1545 set_blocks_per_member(map, blocks_per_member(map)/IMSM_4K_DIV);
1546 map->blocks_per_strip /= IMSM_4K_DIV;
1547 set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV);
1548
1549 if (dev->vol.migr_state) {
1550 /* map1 */
1551 map = get_imsm_map(dev, MAP_1);
1552 set_blocks_per_member(map,
1553 blocks_per_member(map)/IMSM_4K_DIV);
1554 map->blocks_per_strip /= IMSM_4K_DIV;
1555 set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV);
1556 }
1557 }
e4467bc7
TM		 1558 if (bbm_log_size) {
1559 struct bbm_log *log = (void *)mpb +
1560 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
1561 __u32 i;
1562
1563 for (i = 0; i < log->entry_count; i++) {
1564 struct bbm_log_entry *entry =
1565 &log->marked_block_entries[i];
1566
1567 __u8 count = entry->marked_count + 1;
1568 unsigned long long sector =
1569 __le48_to_cpu(&entry->defective_block_start);
1570
1571 entry->defective_block_start =
1572 __cpu_to_le48(sector/IMSM_4K_DIV);
1573 entry->marked_count = max(count/IMSM_4K_DIV, 1) - 1;
1574 }
1575 }
f36a9ecd
PB		 1576
1577 mpb->check_sum = __gen_imsm_checksum(mpb);
1578}
1579
520e69e2
AK		 1580void examine_migr_rec_imsm(struct intel_super *super)
1581{
1582 struct migr_record *migr_rec = super->migr_rec;
1583 struct imsm_super *mpb = super->anchor;
1584 int i;
1585
1586 for (i = 0; i < mpb->num_raid_devs; i++) {
1587 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
3136abe5 1588 struct imsm_map *map;
b4ab44d8 1589 int slot = -1;
3136abe5 1590
520e69e2
AK		 1591 if (is_gen_migration(dev) == 0)
1592 continue;
1593
1594 printf("\nMigration Record Information:");
3136abe5 1595
44bfe6df
AK		 1596 /* first map under migration */
1597 map = get_imsm_map(dev, MAP_0);
3136abe5
AK		 1598 if (map)
1599 slot = get_imsm_disk_slot(map, super->disks->index);
089f9d79 1600 if (map == NULL || slot > 1 || slot < 0) {
520e69e2
AK		 1601 printf(" Empty\n ");
1602 printf("Examine one of first two disks in array\n");
1603 break;
1604 }
1605 printf("\n Status : ");
1606 if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL)
1607 printf("Normal\n");
1608 else
1609 printf("Contains Data\n");
1610 printf(" Current Unit : %u\n",
1611 __le32_to_cpu(migr_rec->curr_migr_unit));
1612 printf(" Family : %u\n",
1613 __le32_to_cpu(migr_rec->family_num));
1614 printf(" Ascending : %u\n",
1615 __le32_to_cpu(migr_rec->ascending_migr));
1616 printf(" Blocks Per Unit : %u\n",
1617 __le32_to_cpu(migr_rec->blocks_per_unit));
1618 printf(" Dest. Depth Per Unit : %u\n",
1619 __le32_to_cpu(migr_rec->dest_depth_per_unit));
1620 printf(" Checkpoint Area pba : %u\n",
1621 __le32_to_cpu(migr_rec->ckpt_area_pba));
1622 printf(" First member lba : %u\n",
1623 __le32_to_cpu(migr_rec->dest_1st_member_lba));
1624 printf(" Total Number of Units : %u\n",
1625 __le32_to_cpu(migr_rec->num_migr_units));
1626 printf(" Size of volume : %u\n",
1627 __le32_to_cpu(migr_rec->post_migr_vol_cap));
1628 printf(" Expansion space for LBA64 : %u\n",
1629 __le32_to_cpu(migr_rec->post_migr_vol_cap_hi));
1630 printf(" Record was read from : %u\n",
1631 __le32_to_cpu(migr_rec->ckpt_read_disk_num));
1632
1633 break;
1634 }
1635}
9e2d750d 1636#endif /* MDASSEMBLE */
f36a9ecd 1637
de44e46f
PB		 1638void convert_from_4k_imsm_migr_rec(struct intel_super *super)
1639{
1640 struct migr_record *migr_rec = super->migr_rec;
1641
1642 migr_rec->blocks_per_unit *= IMSM_4K_DIV;
1643 migr_rec->ckpt_area_pba *= IMSM_4K_DIV;
1644 migr_rec->dest_1st_member_lba *= IMSM_4K_DIV;
1645 migr_rec->dest_depth_per_unit *= IMSM_4K_DIV;
1646 split_ull((join_u32(migr_rec->post_migr_vol_cap,
1647 migr_rec->post_migr_vol_cap_hi) * IMSM_4K_DIV),
1648 &migr_rec->post_migr_vol_cap,
1649 &migr_rec->post_migr_vol_cap_hi);
1650}
1651
f36a9ecd
PB		 1652void convert_from_4k(struct intel_super *super)
1653{
1654 struct imsm_super *mpb = super->anchor;
1655 struct imsm_disk *disk;
1656 int i;
e4467bc7 1657 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
f36a9ecd
PB		 1658
1659 for (i = 0; i < mpb->num_disks ; i++) {
1660 disk = __get_imsm_disk(mpb, i);
1661 /* disk */
1662 set_total_blocks(disk, (total_blocks(disk)*IMSM_4K_DIV));
1663 }
1664
1665 for (i = 0; i < mpb->num_raid_devs; i++) {
1666 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
1667 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1668 /* dev */
1669 split_ull((join_u32(dev->size_low, dev->size_high)*IMSM_4K_DIV),
1670 &dev->size_low, &dev->size_high);
1671 dev->vol.curr_migr_unit *= IMSM_4K_DIV;
1672
1673 /* map0 */
1674 set_blocks_per_member(map, blocks_per_member(map)*IMSM_4K_DIV);
1675 map->blocks_per_strip *= IMSM_4K_DIV;
1676 set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV);
1677
1678 if (dev->vol.migr_state) {
1679 /* map1 */
1680 map = get_imsm_map(dev, MAP_1);
1681 set_blocks_per_member(map,
1682 blocks_per_member(map)*IMSM_4K_DIV);
1683 map->blocks_per_strip *= IMSM_4K_DIV;
1684 set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV);
1685 }
1686 }
e4467bc7
TM		 1687 if (bbm_log_size) {
1688 struct bbm_log *log = (void *)mpb +
1689 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
1690 __u32 i;
1691
1692 for (i = 0; i < log->entry_count; i++) {
1693 struct bbm_log_entry *entry =
1694 &log->marked_block_entries[i];
1695
1696 __u8 count = entry->marked_count + 1;
1697 unsigned long long sector =
1698 __le48_to_cpu(&entry->defective_block_start);
1699
1700 entry->defective_block_start =
1701 __cpu_to_le48(sector*IMSM_4K_DIV);
1702 entry->marked_count = count*IMSM_4K_DIV - 1;
1703 }
1704 }
f36a9ecd
PB		 1705
1706 mpb->check_sum = __gen_imsm_checksum(mpb);
1707}
1708
19482bcc
AK		 1709/*******************************************************************************
1710 * function: imsm_check_attributes
1711 * Description: Function checks if features represented by attributes flags
1011e834 1712 * are supported by mdadm.
19482bcc
AK		 1713 * Parameters:
1714 * attributes - Attributes read from metadata
1715 * Returns:
1011e834
N		 1716 * 0 - passed attributes contains unsupported features flags
1717 * 1 - all features are supported
19482bcc
AK		 1718 ******************************************************************************/
1719static int imsm_check_attributes(__u32 attributes)
1720{
1721 int ret_val = 1;
418f9b36
N		 1722 __u32 not_supported = MPB_ATTRIB_SUPPORTED^0xffffffff;
1723
1724 not_supported &= ~MPB_ATTRIB_IGNORED;
19482bcc
AK		 1725
1726 not_supported &= attributes;
1727 if (not_supported) {
e7b84f9d 1728 pr_err("(IMSM): Unsupported attributes : %x\n",
418f9b36 1729 (unsigned)__le32_to_cpu(not_supported));
19482bcc
AK		 1730 if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) {
1731 dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY \n");
1732 not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY;
1733 }
1734 if (not_supported & MPB_ATTRIB_2TB) {
1735 dprintf("\t\tMPB_ATTRIB_2TB\n");
1736 not_supported ^= MPB_ATTRIB_2TB;
1737 }
1738 if (not_supported & MPB_ATTRIB_RAID0) {
1739 dprintf("\t\tMPB_ATTRIB_RAID0\n");
1740 not_supported ^= MPB_ATTRIB_RAID0;
1741 }
1742 if (not_supported & MPB_ATTRIB_RAID1) {
1743 dprintf("\t\tMPB_ATTRIB_RAID1\n");
1744 not_supported ^= MPB_ATTRIB_RAID1;
1745 }
1746 if (not_supported & MPB_ATTRIB_RAID10) {
1747 dprintf("\t\tMPB_ATTRIB_RAID10\n");
1748 not_supported ^= MPB_ATTRIB_RAID10;
1749 }
1750 if (not_supported & MPB_ATTRIB_RAID1E) {
1751 dprintf("\t\tMPB_ATTRIB_RAID1E\n");
1752 not_supported ^= MPB_ATTRIB_RAID1E;
1753 }
1754 if (not_supported & MPB_ATTRIB_RAID5) {
1755 dprintf("\t\tMPB_ATTRIB_RAID5\n");
1756 not_supported ^= MPB_ATTRIB_RAID5;
1757 }
1758 if (not_supported & MPB_ATTRIB_RAIDCNG) {
1759 dprintf("\t\tMPB_ATTRIB_RAIDCNG\n");
1760 not_supported ^= MPB_ATTRIB_RAIDCNG;
1761 }
1762 if (not_supported & MPB_ATTRIB_BBM) {
1763 dprintf("\t\tMPB_ATTRIB_BBM\n");
1764 not_supported ^= MPB_ATTRIB_BBM;
1765 }
1766 if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) {
1767 dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY (== MPB_ATTRIB_LEGACY)\n");
1768 not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY;
1769 }
1770 if (not_supported & MPB_ATTRIB_EXP_STRIPE_SIZE) {
1771 dprintf("\t\tMPB_ATTRIB_EXP_STRIP_SIZE\n");
1772 not_supported ^= MPB_ATTRIB_EXP_STRIPE_SIZE;
1773 }
1774 if (not_supported & MPB_ATTRIB_2TB_DISK) {
1775 dprintf("\t\tMPB_ATTRIB_2TB_DISK\n");
1776 not_supported ^= MPB_ATTRIB_2TB_DISK;
1777 }
1778 if (not_supported & MPB_ATTRIB_NEVER_USE2) {
1779 dprintf("\t\tMPB_ATTRIB_NEVER_USE2\n");
1780 not_supported ^= MPB_ATTRIB_NEVER_USE2;
1781 }
1782 if (not_supported & MPB_ATTRIB_NEVER_USE) {
1783 dprintf("\t\tMPB_ATTRIB_NEVER_USE\n");
1784 not_supported ^= MPB_ATTRIB_NEVER_USE;
1785 }
1786
1787 if (not_supported)
1ade5cc1 1788 dprintf("(IMSM): Unknown attributes : %x\n", not_supported);
19482bcc
AK		 1789
1790 ret_val = 0;
1791 }
1792
1793 return ret_val;
1794}
1795
9e2d750d 1796#ifndef MDASSEMBLE
a5d85af7 1797static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map);
44470971 1798
cdddbdbc
DW		 1799static void examine_super_imsm(struct supertype *st, char *homehost)
1800{
1801 struct intel_super *super = st->sb;
949c47a0 1802 struct imsm_super *mpb = super->anchor;
cdddbdbc
DW		 1803 char str[MAX_SIGNATURE_LENGTH];
1804 int i;
27fd6274
DW		 1805 struct mdinfo info;
1806 char nbuf[64];
cdddbdbc 1807 __u32 sum;
14e8215b 1808 __u32 reserved = imsm_reserved_sectors(super, super->disks);
94827db3 1809 struct dl *dl;
27fd6274 1810
cdddbdbc
DW		 1811 snprintf(str, MPB_SIG_LEN, "%s", mpb->sig);
1812 printf(" Magic : %s\n", str);
1813 snprintf(str, strlen(MPB_VERSION_RAID0), "%s", get_imsm_version(mpb));
1814 printf(" Version : %s\n", get_imsm_version(mpb));
148acb7b 1815 printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num));
cdddbdbc
DW		 1816 printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num));
1817 printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num));
19482bcc
AK		 1818 printf(" Attributes : ");
1819 if (imsm_check_attributes(mpb->attributes))
1820 printf("All supported\n");
1821 else
1822 printf("not supported\n");
a5d85af7 1823 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 1824 fname_from_uuid(st, &info, nbuf, ':');
27fd6274 1825 printf(" UUID : %s\n", nbuf + 5);
cdddbdbc
DW		 1826 sum = __le32_to_cpu(mpb->check_sum);
1827 printf(" Checksum : %08x %s\n", sum,
949c47a0 1828 __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect");
f36a9ecd 1829 printf(" MPB Sectors : %d\n", mpb_sectors(mpb, super->sector_size));
cdddbdbc
DW		 1830 printf(" Disks : %d\n", mpb->num_disks);
1831 printf(" RAID Devices : %d\n", mpb->num_raid_devs);
0ec1f4e8 1832 print_imsm_disk(__get_imsm_disk(mpb, super->disks->index), super->disks->index, reserved);
8d67477f 1833 if (get_imsm_bbm_log_size(super->bbm_log)) {
604b746f
JD		 1834 struct bbm_log *log = super->bbm_log;
1835
1836 printf("\n");
1837 printf("Bad Block Management Log:\n");
1838 printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size));
1839 printf(" Signature : %x\n", __le32_to_cpu(log->signature));
1840 printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count));
604b746f 1841 }
44470971
DW		 1842 for (i = 0; i < mpb->num_raid_devs; i++) {
1843 struct mdinfo info;
1844 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
1845
1846 super->current_vol = i;
a5d85af7 1847 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 1848 fname_from_uuid(st, &info, nbuf, ':');
c47b0ff6 1849 print_imsm_dev(super, dev, nbuf + 5, super->disks->index);
44470971 1850 }
cdddbdbc
DW		 1851 for (i = 0; i < mpb->num_disks; i++) {
1852 if (i == super->disks->index)
1853 continue;
0ec1f4e8 1854 print_imsm_disk(__get_imsm_disk(mpb, i), i, reserved);
cdddbdbc 1855 }
94827db3 1856
0ec1f4e8
DW		 1857 for (dl = super->disks; dl; dl = dl->next)
1858 if (dl->index == -1)
1859 print_imsm_disk(&dl->disk, -1, reserved);
520e69e2
AK		 1860
1861 examine_migr_rec_imsm(super);
cdddbdbc
DW		 1862}
1863
061f2c6a 1864static void brief_examine_super_imsm(struct supertype *st, int verbose)
cdddbdbc 1865{
27fd6274 1866 /* We just write a generic IMSM ARRAY entry */
ff54de6e
N		 1867 struct mdinfo info;
1868 char nbuf[64];
1e7bc0ed 1869 struct intel_super *super = st->sb;
1e7bc0ed 1870
0d5a423f
DW		 1871 if (!super->anchor->num_raid_devs) {
1872 printf("ARRAY metadata=imsm\n");
1e7bc0ed 1873 return;
0d5a423f 1874 }
ff54de6e 1875
a5d85af7 1876 getinfo_super_imsm(st, &info, NULL);
4737ae25
N		 1877 fname_from_uuid(st, &info, nbuf, ':');
1878 printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5);
1879}
1880
1881static void brief_examine_subarrays_imsm(struct supertype *st, int verbose)
1882{
1883 /* We just write a generic IMSM ARRAY entry */
1884 struct mdinfo info;
1885 char nbuf[64];
1886 char nbuf1[64];
1887 struct intel_super *super = st->sb;
1888 int i;
1889
1890 if (!super->anchor->num_raid_devs)
1891 return;
1892
a5d85af7 1893 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 1894 fname_from_uuid(st, &info, nbuf, ':');
1e7bc0ed
DW		 1895 for (i = 0; i < super->anchor->num_raid_devs; i++) {
1896 struct imsm_dev *dev = get_imsm_dev(super, i);
1897
1898 super->current_vol = i;
a5d85af7 1899 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 1900 fname_from_uuid(st, &info, nbuf1, ':');
1124b3cf 1901 printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n",
cf8de691 1902 dev->volume, nbuf + 5, i, nbuf1 + 5);
1e7bc0ed 1903 }
cdddbdbc
DW		 1904}
1905
9d84c8ea
DW		 1906static void export_examine_super_imsm(struct supertype *st)
1907{
1908 struct intel_super *super = st->sb;
1909 struct imsm_super *mpb = super->anchor;
1910 struct mdinfo info;
1911 char nbuf[64];
1912
a5d85af7 1913 getinfo_super_imsm(st, &info, NULL);
9d84c8ea
DW		 1914 fname_from_uuid(st, &info, nbuf, ':');
1915 printf("MD_METADATA=imsm\n");
1916 printf("MD_LEVEL=container\n");
1917 printf("MD_UUID=%s\n", nbuf+5);
1918 printf("MD_DEVICES=%u\n", mpb->num_disks);
1919}
1920
74db60b0
N		 1921static int copy_metadata_imsm(struct supertype *st, int from, int to)
1922{
f36a9ecd 1923 /* The second last sector of the device contains
74db60b0
N		 1924 * the "struct imsm_super" metadata.
1925 * This contains mpb_size which is the size in bytes of the
1926 * extended metadata. This is located immediately before
1927 * the imsm_super.
1928 * We want to read all that, plus the last sector which
1929 * may contain a migration record, and write it all
1930 * to the target.
1931 */
1932 void *buf;
1933 unsigned long long dsize, offset;
1934 int sectors;
1935 struct imsm_super *sb;
f36a9ecd
PB		 1936 struct intel_super *super = st->sb;
1937 unsigned int sector_size = super->sector_size;
1938 unsigned int written = 0;
74db60b0 1939
de44e46f 1940 if (posix_memalign(&buf, MAX_SECTOR_SIZE, MAX_SECTOR_SIZE) != 0)
74db60b0
N		 1941 return 1;
1942
1943 if (!get_dev_size(from, NULL, &dsize))
1944 goto err;
1945
f36a9ecd 1946 if (lseek64(from, dsize-(2*sector_size), 0) < 0)
74db60b0 1947 goto err;
f36a9ecd 1948 if (read(from, buf, sector_size) != sector_size)
74db60b0
N		 1949 goto err;
1950 sb = buf;
1951 if (strncmp((char*)sb->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0)
1952 goto err;
1953
f36a9ecd
PB		 1954 sectors = mpb_sectors(sb, sector_size) + 2;
1955 offset = dsize - sectors * sector_size;
74db60b0
N		 1956 if (lseek64(from, offset, 0) < 0 ||
1957 lseek64(to, offset, 0) < 0)
1958 goto err;
f36a9ecd
PB		 1959 while (written < sectors * sector_size) {
1960 int n = sectors*sector_size - written;
74db60b0
N		 1961 if (n > 4096)
1962 n = 4096;
1963 if (read(from, buf, n) != n)
1964 goto err;
1965 if (write(to, buf, n) != n)
1966 goto err;
1967 written += n;
1968 }
1969 free(buf);
1970 return 0;
1971err:
1972 free(buf);
1973 return 1;
1974}
1975
cdddbdbc
DW		 1976static void detail_super_imsm(struct supertype *st, char *homehost)
1977{
3ebe00a1
DW		 1978 struct mdinfo info;
1979 char nbuf[64];
1980
a5d85af7 1981 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 1982 fname_from_uuid(st, &info, nbuf, ':');
3ebe00a1 1983 printf("\n UUID : %s\n", nbuf + 5);
cdddbdbc
DW		 1984}
1985
1986static void brief_detail_super_imsm(struct supertype *st)
1987{
ff54de6e
N		 1988 struct mdinfo info;
1989 char nbuf[64];
a5d85af7 1990 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 1991 fname_from_uuid(st, &info, nbuf, ':');
ff54de6e 1992 printf(" UUID=%s", nbuf + 5);
cdddbdbc 1993}
d665cc31
DW		 1994
1995static int imsm_read_serial(int fd, char *devname, __u8 *serial);
1996static void fd2devname(int fd, char *name);
1997
120dc887 1998static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose)
d665cc31 1999{
120dc887
LM		 2000 /* dump an unsorted list of devices attached to AHCI Intel storage
2001 * controller, as well as non-connected ports
d665cc31
DW		 2002 */
2003 int hba_len = strlen(hba_path) + 1;
2004 struct dirent *ent;
2005 DIR *dir;
2006 char *path = NULL;
2007 int err = 0;
2008 unsigned long port_mask = (1 << port_count) - 1;
2009
f21e18ca 2010 if (port_count > (int)sizeof(port_mask) * 8) {
ba728be7 2011 if (verbose > 0)
e7b84f9d 2012 pr_err("port_count %d out of range\n", port_count);
d665cc31
DW		 2013 return 2;
2014 }
2015
2016 /* scroll through /sys/dev/block looking for devices attached to
2017 * this hba
2018 */
2019 dir = opendir("/sys/dev/block");
1a6dd6b9
PB		 2020 if (!dir)
2021 return 1;
2022
2023 for (ent = readdir(dir); ent; ent = readdir(dir)) {
d665cc31
DW		 2024 int fd;
2025 char model[64];
2026 char vendor[64];
2027 char buf[1024];
2028 int major, minor;
2029 char *device;
2030 char *c;
2031 int port;
2032 int type;
2033
2034 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
2035 continue;
2036 path = devt_to_devpath(makedev(major, minor));
2037 if (!path)
2038 continue;
2039 if (!path_attached_to_hba(path, hba_path)) {
2040 free(path);
2041 path = NULL;
2042 continue;
2043 }
2044
2045 /* retrieve the scsi device type */
2046 if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) {
ba728be7 2047 if (verbose > 0)
e7b84f9d 2048 pr_err("failed to allocate 'device'\n");
d665cc31
DW		 2049 err = 2;
2050 break;
2051 }
2052 sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor);
193b6c0b 2053 if (load_sys(device, buf, sizeof(buf)) != 0) {
ba728be7 2054 if (verbose > 0)
e7b84f9d 2055 pr_err("failed to read device type for %s\n",
d665cc31
DW		 2056 path);
2057 err = 2;
2058 free(device);
2059 break;
2060 }
2061 type = strtoul(buf, NULL, 10);
2062
2063 /* if it's not a disk print the vendor and model */
2064 if (!(type == 0 || type == 7 || type == 14)) {
2065 vendor[0] = '\0';
2066 model[0] = '\0';
2067 sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor);
193b6c0b 2068 if (load_sys(device, buf, sizeof(buf)) == 0) {
d665cc31
DW		 2069 strncpy(vendor, buf, sizeof(vendor));
2070 vendor[sizeof(vendor) - 1] = '\0';
2071 c = (char *) &vendor[sizeof(vendor) - 1];
2072 while (isspace(*c) || *c == '\0')
2073 *c-- = '\0';
2074
2075 }
2076 sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor);
193b6c0b 2077 if (load_sys(device, buf, sizeof(buf)) == 0) {
d665cc31
DW		 2078 strncpy(model, buf, sizeof(model));
2079 model[sizeof(model) - 1] = '\0';
2080 c = (char *) &model[sizeof(model) - 1];
2081 while (isspace(*c) || *c == '\0')
2082 *c-- = '\0';
2083 }
2084
2085 if (vendor[0] && model[0])
2086 sprintf(buf, "%.64s %.64s", vendor, model);
2087 else
2088 switch (type) { /* numbers from hald/linux/device.c */
2089 case 1: sprintf(buf, "tape"); break;
2090 case 2: sprintf(buf, "printer"); break;
2091 case 3: sprintf(buf, "processor"); break;
2092 case 4:
2093 case 5: sprintf(buf, "cdrom"); break;
2094 case 6: sprintf(buf, "scanner"); break;
2095 case 8: sprintf(buf, "media_changer"); break;
2096 case 9: sprintf(buf, "comm"); break;
2097 case 12: sprintf(buf, "raid"); break;
2098 default: sprintf(buf, "unknown");
2099 }
2100 } else
2101 buf[0] = '\0';
2102 free(device);
2103
2104 /* chop device path to 'host%d' and calculate the port number */
2105 c = strchr(&path[hba_len], '/');
4e5e717d 2106 if (!c) {
ba728be7 2107 if (verbose > 0)
e7b84f9d 2108 pr_err("%s - invalid path name\n", path + hba_len);
4e5e717d
AW		 2109 err = 2;
2110 break;
2111 }
d665cc31 2112 *c = '\0';
0858eccf
AP		 2113 if ((sscanf(&path[hba_len], "ata%d", &port) == 1) ||
2114 ((sscanf(&path[hba_len], "host%d", &port) == 1)))
d665cc31
DW		 2115 port -= host_base;
2116 else {
ba728be7 2117 if (verbose > 0) {
d665cc31 2118 *c = '/'; /* repair the full string */
e7b84f9d 2119 pr_err("failed to determine port number for %s\n",
d665cc31
DW		 2120 path);
2121 }
2122 err = 2;
2123 break;
2124 }
2125
2126 /* mark this port as used */
2127 port_mask &= ~(1 << port);
2128
2129 /* print out the device information */
2130 if (buf[0]) {
2131 printf(" Port%d : - non-disk device (%s) -\n", port, buf);
2132 continue;
2133 }
2134
2135 fd = dev_open(ent->d_name, O_RDONLY);
2136 if (fd < 0)
2137 printf(" Port%d : - disk info unavailable -\n", port);
2138 else {
2139 fd2devname(fd, buf);
2140 printf(" Port%d : %s", port, buf);
2141 if (imsm_read_serial(fd, NULL, (__u8 *) buf) == 0)
664d5325 2142 printf(" (%.*s)\n", MAX_RAID_SERIAL_LEN, buf);
d665cc31 2143 else
664d5325 2144 printf(" ()\n");
4dab422a 2145 close(fd);
d665cc31 2146 }
d665cc31
DW		 2147 free(path);
2148 path = NULL;
2149 }
2150 if (path)
2151 free(path);
2152 if (dir)
2153 closedir(dir);
2154 if (err == 0) {
2155 int i;
2156
2157 for (i = 0; i < port_count; i++)
2158 if (port_mask & (1 << i))
2159 printf(" Port%d : - no device attached -\n", i);
2160 }
2161
2162 return err;
2163}
2164
b5eece69 2165static int print_vmd_attached_devs(struct sys_dev *hba)
60f0f54d
PB		 2166{
2167 struct dirent *ent;
2168 DIR *dir;
2169 char path[292];
2170 char link[256];
2171 char *c, *rp;
2172
2173 if (hba->type != SYS_DEV_VMD)
b5eece69 2174 return 1;
60f0f54d
PB		 2175
2176 /* scroll through /sys/dev/block looking for devices attached to
2177 * this hba
2178 */
2179 dir = opendir("/sys/bus/pci/drivers/nvme");
b9135011 2180 if (!dir)
b5eece69 2181 return 1;
b9135011
JS		 2182
2183 for (ent = readdir(dir); ent; ent = readdir(dir)) {
60f0f54d
PB		 2184 int n;
2185
2186 /* is 'ent' a device? check that the 'subsystem' link exists and
2187 * that its target matches 'bus'
2188 */
2189 sprintf(path, "/sys/bus/pci/drivers/nvme/%s/subsystem",
2190 ent->d_name);
2191 n = readlink(path, link, sizeof(link));
2192 if (n < 0 || n >= (int)sizeof(link))
2193 continue;
2194 link[n] = '\0';
2195 c = strrchr(link, '/');
2196 if (!c)
2197 continue;
2198 if (strncmp("pci", c+1, strlen("pci")) != 0)
2199 continue;
2200
2201 sprintf(path, "/sys/bus/pci/drivers/nvme/%s", ent->d_name);
60f0f54d
PB		 2202
2203 rp = realpath(path, NULL);
2204 if (!rp)
2205 continue;
2206
2207 if (path_attached_to_hba(rp, hba->path)) {
2208 printf(" NVMe under VMD : %s\n", rp);
2209 }
2210 free(rp);
2211 }
2212
b9135011 2213 closedir(dir);
b5eece69 2214 return 0;
60f0f54d
PB		 2215}
2216
120dc887
LM		 2217static void print_found_intel_controllers(struct sys_dev *elem)
2218{
2219 for (; elem; elem = elem->next) {
e7b84f9d 2220 pr_err("found Intel(R) ");
120dc887
LM		 2221 if (elem->type == SYS_DEV_SATA)
2222 fprintf(stderr, "SATA ");
155cbb4c
LM		 2223 else if (elem->type == SYS_DEV_SAS)
2224 fprintf(stderr, "SAS ");
0858eccf
AP		 2225 else if (elem->type == SYS_DEV_NVME)
2226 fprintf(stderr, "NVMe ");
60f0f54d
PB		 2227
2228 if (elem->type == SYS_DEV_VMD)
2229 fprintf(stderr, "VMD domain");
2230 else
2231 fprintf(stderr, "RAID controller");
2232
120dc887
LM		 2233 if (elem->pci_id)
2234 fprintf(stderr, " at %s", elem->pci_id);
2235 fprintf(stderr, ".\n");
2236 }
2237 fflush(stderr);
2238}
2239
120dc887
LM		 2240static int ahci_get_port_count(const char *hba_path, int *port_count)
2241{
2242 struct dirent *ent;
2243 DIR *dir;
2244 int host_base = -1;
2245
2246 *port_count = 0;
2247 if ((dir = opendir(hba_path)) == NULL)
2248 return -1;
2249
2250 for (ent = readdir(dir); ent; ent = readdir(dir)) {
2251 int host;
2252
0858eccf
AP		 2253 if ((sscanf(ent->d_name, "ata%d", &host) != 1) &&
2254 ((sscanf(ent->d_name, "host%d", &host) != 1)))
120dc887
LM		 2255 continue;
2256 if (*port_count == 0)
2257 host_base = host;
2258 else if (host < host_base)
2259 host_base = host;
2260
2261 if (host + 1 > *port_count + host_base)
2262 *port_count = host + 1 - host_base;
2263 }
2264 closedir(dir);
2265 return host_base;
2266}
2267
a891a3c2
LM		 2268static void print_imsm_capability(const struct imsm_orom *orom)
2269{
0858eccf
AP		 2270 printf(" Platform : Intel(R) ");
2271 if (orom->capabilities == 0 && orom->driver_features == 0)
2272 printf("Matrix Storage Manager\n");
2273 else
2274 printf("Rapid Storage Technology%s\n",
2275 imsm_orom_is_enterprise(orom) ? " enterprise" : "");
2276 if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build)
2277 printf(" Version : %d.%d.%d.%d\n", orom->major_ver,
2278 orom->minor_ver, orom->hotfix_ver, orom->build);
a891a3c2
LM		 2279 printf(" RAID Levels :%s%s%s%s%s\n",
2280 imsm_orom_has_raid0(orom) ? " raid0" : "",
2281 imsm_orom_has_raid1(orom) ? " raid1" : "",
2282 imsm_orom_has_raid1e(orom) ? " raid1e" : "",
2283 imsm_orom_has_raid10(orom) ? " raid10" : "",
2284 imsm_orom_has_raid5(orom) ? " raid5" : "");
2285 printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
2286 imsm_orom_has_chunk(orom, 2) ? " 2k" : "",
2287 imsm_orom_has_chunk(orom, 4) ? " 4k" : "",
2288 imsm_orom_has_chunk(orom, 8) ? " 8k" : "",
2289 imsm_orom_has_chunk(orom, 16) ? " 16k" : "",
2290 imsm_orom_has_chunk(orom, 32) ? " 32k" : "",
2291 imsm_orom_has_chunk(orom, 64) ? " 64k" : "",
2292 imsm_orom_has_chunk(orom, 128) ? " 128k" : "",
2293 imsm_orom_has_chunk(orom, 256) ? " 256k" : "",
2294 imsm_orom_has_chunk(orom, 512) ? " 512k" : "",
2295 imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "",
2296 imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "",
2297 imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "",
2298 imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "",
2299 imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "",
2300 imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "",
2301 imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : "");
29cd0821
CA		 2302 printf(" 2TB volumes :%s supported\n",
2303 (orom->attr & IMSM_OROM_ATTR_2TB)?"":" not");
2304 printf(" 2TB disks :%s supported\n",
2305 (orom->attr & IMSM_OROM_ATTR_2TB_DISK)?"":" not");
0e7f69a8 2306 printf(" Max Disks : %d\n", orom->tds);
0858eccf
AP		 2307 printf(" Max Volumes : %d per array, %d per %s\n",
2308 orom->vpa, orom->vphba,
2309 imsm_orom_is_nvme(orom) ? "platform" : "controller");
a891a3c2
LM		 2310 return;
2311}
2312
e50cf220
MN		 2313static void print_imsm_capability_export(const struct imsm_orom *orom)
2314{
2315 printf("MD_FIRMWARE_TYPE=imsm\n");
0858eccf
AP		 2316 if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build)
2317 printf("IMSM_VERSION=%d.%d.%d.%d\n", orom->major_ver, orom->minor_ver,
2318 orom->hotfix_ver, orom->build);
e50cf220
MN		 2319 printf("IMSM_SUPPORTED_RAID_LEVELS=%s%s%s%s%s\n",
2320 imsm_orom_has_raid0(orom) ? "raid0 " : "",
2321 imsm_orom_has_raid1(orom) ? "raid1 " : "",
2322 imsm_orom_has_raid1e(orom) ? "raid1e " : "",
2323 imsm_orom_has_raid5(orom) ? "raid10 " : "",
2324 imsm_orom_has_raid10(orom) ? "raid5 " : "");
2325 printf("IMSM_SUPPORTED_CHUNK_SIZES=%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
2326 imsm_orom_has_chunk(orom, 2) ? "2k " : "",
2327 imsm_orom_has_chunk(orom, 4) ? "4k " : "",
2328 imsm_orom_has_chunk(orom, 8) ? "8k " : "",
2329 imsm_orom_has_chunk(orom, 16) ? "16k " : "",
2330 imsm_orom_has_chunk(orom, 32) ? "32k " : "",
2331 imsm_orom_has_chunk(orom, 64) ? "64k " : "",
2332 imsm_orom_has_chunk(orom, 128) ? "128k " : "",
2333 imsm_orom_has_chunk(orom, 256) ? "256k " : "",
2334 imsm_orom_has_chunk(orom, 512) ? "512k " : "",
2335 imsm_orom_has_chunk(orom, 1024*1) ? "1M " : "",
2336 imsm_orom_has_chunk(orom, 1024*2) ? "2M " : "",
2337 imsm_orom_has_chunk(orom, 1024*4) ? "4M " : "",
2338 imsm_orom_has_chunk(orom, 1024*8) ? "8M " : "",
2339 imsm_orom_has_chunk(orom, 1024*16) ? "16M " : "",
2340 imsm_orom_has_chunk(orom, 1024*32) ? "32M " : "",
2341 imsm_orom_has_chunk(orom, 1024*64) ? "64M " : "");
2342 printf("IMSM_2TB_VOLUMES=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB) ? "yes" : "no");
2343 printf("IMSM_2TB_DISKS=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB_DISK) ? "yes" : "no");
2344 printf("IMSM_MAX_DISKS=%d\n",orom->tds);
2345 printf("IMSM_MAX_VOLUMES_PER_ARRAY=%d\n",orom->vpa);
2346 printf("IMSM_MAX_VOLUMES_PER_CONTROLLER=%d\n",orom->vphba);
2347}
2348
9eafa1de 2349static int detail_platform_imsm(int verbose, int enumerate_only, char *controller_path)
d665cc31
DW		 2350{
2351 /* There are two components to imsm platform support, the ahci SATA
2352 * controller and the option-rom. To find the SATA controller we
2353 * simply look in /sys/bus/pci/drivers/ahci to see if an ahci
2354 * controller with the Intel vendor id is present. This approach
2355 * allows mdadm to leverage the kernel's ahci detection logic, with the
2356 * caveat that if ahci.ko is not loaded mdadm will not be able to
2357 * detect platform raid capabilities. The option-rom resides in a
2358 * platform "Adapter ROM". We scan for its signature to retrieve the
2359 * platform capabilities. If raid support is disabled in the BIOS the
2360 * option-rom capability structure will not be available.
2361 */
d665cc31 2362 struct sys_dev *list, *hba;
d665cc31
DW		 2363 int host_base = 0;
2364 int port_count = 0;
9eafa1de 2365 int result=1;
d665cc31 2366
5615172f 2367 if (enumerate_only) {
a891a3c2 2368 if (check_env("IMSM_NO_PLATFORM"))
5615172f 2369 return 0;
a891a3c2
LM		 2370 list = find_intel_devices();
2371 if (!list)
2372 return 2;
2373 for (hba = list; hba; hba = hba->next) {
6b781d33
AP		 2374 if (find_imsm_capability(hba)) {
2375 result = 0;
a891a3c2
LM		 2376 break;
2377 }
9eafa1de 2378 else
6b781d33 2379 result = 2;
a891a3c2 2380 }
a891a3c2 2381 return result;
5615172f
DW		 2382 }
2383
155cbb4c
LM		 2384 list = find_intel_devices();
2385 if (!list) {
ba728be7 2386 if (verbose > 0)
7a862a02 2387 pr_err("no active Intel(R) RAID controller found.\n");
d665cc31 2388 return 2;
ba728be7 2389 } else if (verbose > 0)
155cbb4c 2390 print_found_intel_controllers(list);
d665cc31 2391
a891a3c2 2392 for (hba = list; hba; hba = hba->next) {
0858eccf 2393 if (controller_path && (compare_paths(hba->path, controller_path) != 0))
9eafa1de 2394 continue;
0858eccf 2395 if (!find_imsm_capability(hba)) {
60f0f54d 2396 char buf[PATH_MAX];
e7b84f9d 2397 pr_err("imsm capabilities not found for controller: %s (type %s)\n",
60f0f54d
PB		 2398 hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path,
2399 get_sys_dev_type(hba->type));
0858eccf
AP		 2400 continue;
2401 }
2402 result = 0;
2403 }
2404
2405 if (controller_path && result == 1) {
2406 pr_err("no active Intel(R) RAID controller found under %s\n",
2407 controller_path);
2408 return result;
2409 }
2410
5e1d6128 2411 const struct orom_entry *entry;
0858eccf 2412
5e1d6128 2413 for (entry = orom_entries; entry; entry = entry->next) {
60f0f54d 2414 if (entry->type == SYS_DEV_VMD) {
07cb1e57 2415 print_imsm_capability(&entry->orom);
32716c51
PB		 2416 printf(" 3rd party NVMe :%s supported\n",
2417 imsm_orom_has_tpv_support(&entry->orom)?"":" not");
60f0f54d
PB		 2418 for (hba = list; hba; hba = hba->next) {
2419 if (hba->type == SYS_DEV_VMD) {
2420 char buf[PATH_MAX];
60f0f54d
PB		 2421 printf(" I/O Controller : %s (%s)\n",
2422 vmd_domain_to_controller(hba, buf), get_sys_dev_type(hba->type));
b5eece69
PB		 2423 if (print_vmd_attached_devs(hba)) {
2424 if (verbose > 0)
2425 pr_err("failed to get devices attached to VMD domain.\n");
2426 result |= 2;
2427 }
60f0f54d
PB		 2428 }
2429 }
07cb1e57 2430 printf("\n");
60f0f54d
PB		 2431 continue;
2432 }
0858eccf 2433
60f0f54d
PB		 2434 print_imsm_capability(&entry->orom);
2435 if (entry->type == SYS_DEV_NVME) {
0858eccf
AP		 2436 for (hba = list; hba; hba = hba->next) {
2437 if (hba->type == SYS_DEV_NVME)
2438 printf(" NVMe Device : %s\n", hba->path);
2439 }
60f0f54d 2440 printf("\n");
0858eccf
AP		 2441 continue;
2442 }
2443
2444 struct devid_list *devid;
5e1d6128 2445 for (devid = entry->devid_list; devid; devid = devid->next) {
0858eccf
AP		 2446 hba = device_by_id(devid->devid);
2447 if (!hba)
2448 continue;
2449
9eafa1de
MN		 2450 printf(" I/O Controller : %s (%s)\n",
2451 hba->path, get_sys_dev_type(hba->type));
2452 if (hba->type == SYS_DEV_SATA) {
2453 host_base = ahci_get_port_count(hba->path, &port_count);
2454 if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) {
2455 if (verbose > 0)
7a862a02 2456 pr_err("failed to enumerate ports on SATA controller at %s.\n", hba->pci_id);
9eafa1de
MN		 2457 result |= 2;
2458 }
120dc887
LM		 2459 }
2460 }
0858eccf 2461 printf("\n");
d665cc31 2462 }
155cbb4c 2463
120dc887 2464 return result;
d665cc31 2465}
e50cf220 2466
9eafa1de 2467static int export_detail_platform_imsm(int verbose, char *controller_path)
e50cf220 2468{
e50cf220
MN		 2469 struct sys_dev *list, *hba;
2470 int result=1;
2471
2472 list = find_intel_devices();
2473 if (!list) {
2474 if (verbose > 0)
2475 pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_INTEL_DEVICES\n");
2476 result = 2;
e50cf220
MN		 2477 return result;
2478 }
2479
2480 for (hba = list; hba; hba = hba->next) {
9eafa1de
MN		 2481 if (controller_path && (compare_paths(hba->path,controller_path) != 0))
2482 continue;
60f0f54d
PB		 2483 if (!find_imsm_capability(hba) && verbose > 0) {
2484 char buf[PATH_MAX];
2485 pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_IMSM_CAPABLE_DEVICE_UNDER_%s\n",
2486 hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path);
2487 }
0858eccf 2488 else
e50cf220 2489 result = 0;
e50cf220
MN		 2490 }
2491
5e1d6128 2492 const struct orom_entry *entry;
0858eccf 2493
60f0f54d
PB		 2494 for (entry = orom_entries; entry; entry = entry->next) {
2495 if (entry->type == SYS_DEV_VMD) {
2496 for (hba = list; hba; hba = hba->next)
2497 print_imsm_capability_export(&entry->orom);
2498 continue;
2499 }
5e1d6128 2500 print_imsm_capability_export(&entry->orom);
60f0f54d 2501 }
0858eccf 2502
e50cf220
MN		 2503 return result;
2504}
2505
cdddbdbc
DW		 2506#endif
2507
2508static int match_home_imsm(struct supertype *st, char *homehost)
2509{
5115ca67
DW		 2510 /* the imsm metadata format does not specify any host
2511 * identification information. We return -1 since we can never
2512 * confirm nor deny whether a given array is "meant" for this
148acb7b 2513 * host. We rely on compare_super and the 'family_num' fields to
5115ca67
DW		 2514 * exclude member disks that do not belong, and we rely on
2515 * mdadm.conf to specify the arrays that should be assembled.
2516 * Auto-assembly may still pick up "foreign" arrays.
2517 */
cdddbdbc 2518
9362c1c8 2519 return -1;
cdddbdbc
DW		 2520}
2521
2522static void uuid_from_super_imsm(struct supertype *st, int uuid[4])
2523{
51006d85
N		 2524 /* The uuid returned here is used for:
2525 * uuid to put into bitmap file (Create, Grow)
2526 * uuid for backup header when saving critical section (Grow)
2527 * comparing uuids when re-adding a device into an array
2528 * In these cases the uuid required is that of the data-array,
2529 * not the device-set.
2530 * uuid to recognise same set when adding a missing device back
2531 * to an array. This is a uuid for the device-set.
1011e834 2532 *
51006d85
N		 2533 * For each of these we can make do with a truncated
2534 * or hashed uuid rather than the original, as long as
2535 * everyone agrees.
2536 * In each case the uuid required is that of the data-array,
2537 * not the device-set.
43dad3d6 2538 */
51006d85
N		 2539 /* imsm does not track uuid's so we synthesis one using sha1 on
2540 * - The signature (Which is constant for all imsm array, but no matter)
148acb7b 2541 * - the orig_family_num of the container
51006d85
N		 2542 * - the index number of the volume
2543 * - the 'serial' number of the volume.
2544 * Hopefully these are all constant.
2545 */
2546 struct intel_super *super = st->sb;
43dad3d6 2547
51006d85
N		 2548 char buf[20];
2549 struct sha1_ctx ctx;
2550 struct imsm_dev *dev = NULL;
148acb7b 2551 __u32 family_num;
51006d85 2552
148acb7b
DW		 2553 /* some mdadm versions failed to set ->orig_family_num, in which
2554 * case fall back to ->family_num. orig_family_num will be
2555 * fixed up with the first metadata update.
2556 */
2557 family_num = super->anchor->orig_family_num;
2558 if (family_num == 0)
2559 family_num = super->anchor->family_num;
51006d85 2560 sha1_init_ctx(&ctx);
92bd8f8d 2561 sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx);
148acb7b 2562 sha1_process_bytes(&family_num, sizeof(__u32), &ctx);
51006d85
N		 2563 if (super->current_vol >= 0)
2564 dev = get_imsm_dev(super, super->current_vol);
2565 if (dev) {
2566 __u32 vol = super->current_vol;
2567 sha1_process_bytes(&vol, sizeof(vol), &ctx);
2568 sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx);
2569 }
2570 sha1_finish_ctx(&ctx, buf);
2571 memcpy(uuid, buf, 4*4);
cdddbdbc
DW		 2572}
2573
0d481d37 2574#if 0
4f5bc454
DW		 2575static void
2576get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p)
cdddbdbc 2577{
cdddbdbc
DW		 2578 __u8 *v = get_imsm_version(mpb);
2579 __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH;
2580 char major[] = { 0, 0, 0 };
2581 char minor[] = { 0 ,0, 0 };
2582 char patch[] = { 0, 0, 0 };
2583 char *ver_parse[] = { major, minor, patch };
2584 int i, j;
2585
2586 i = j = 0;
2587 while (*v != '\0' && v < end) {
2588 if (*v != '.' && j < 2)
2589 ver_parse[i][j++] = *v;
2590 else {
2591 i++;
2592 j = 0;
2593 }
2594 v++;
2595 }
2596
4f5bc454
DW		 2597 *m = strtol(minor, NULL, 0);
2598 *p = strtol(patch, NULL, 0);
2599}
0d481d37 2600#endif
4f5bc454 2601
1e5c6983
DW		 2602static __u32 migr_strip_blocks_resync(struct imsm_dev *dev)
2603{
2604 /* migr_strip_size when repairing or initializing parity */
238c0a71 2605 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2606 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2607
2608 switch (get_imsm_raid_level(map)) {
2609 case 5:
2610 case 10:
2611 return chunk;
2612 default:
2613 return 128*1024 >> 9;
2614 }
2615}
2616
2617static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev)
2618{
2619 /* migr_strip_size when rebuilding a degraded disk, no idea why
2620 * this is different than migr_strip_size_resync(), but it's good
2621 * to be compatible
2622 */
238c0a71 2623 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2624 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2625
2626 switch (get_imsm_raid_level(map)) {
2627 case 1:
2628 case 10:
2629 if (map->num_members % map->num_domains == 0)
2630 return 128*1024 >> 9;
2631 else
2632 return chunk;
2633 case 5:
2634 return max((__u32) 64*1024 >> 9, chunk);
2635 default:
2636 return 128*1024 >> 9;
2637 }
2638}
2639
2640static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev)
2641{
238c0a71
AK		 2642 struct imsm_map *lo = get_imsm_map(dev, MAP_0);
2643 struct imsm_map *hi = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2644 __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip);
2645 __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip);
2646
2647 return max((__u32) 1, hi_chunk / lo_chunk);
2648}
2649
2650static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev)
2651{
238c0a71 2652 struct imsm_map *lo = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2653 int level = get_imsm_raid_level(lo);
2654
2655 if (level == 1 || level == 10) {
238c0a71 2656 struct imsm_map *hi = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2657
2658 return hi->num_domains;
2659 } else
2660 return num_stripes_per_unit_resync(dev);
2661}
2662
98130f40 2663static __u8 imsm_num_data_members(struct imsm_dev *dev, int second_map)
1e5c6983
DW		 2664{
2665 /* named 'imsm_' because raid0, raid1 and raid10
2666 * counter-intuitively have the same number of data disks
2667 */
98130f40 2668 struct imsm_map *map = get_imsm_map(dev, second_map);
1e5c6983
DW		 2669
2670 switch (get_imsm_raid_level(map)) {
2671 case 0:
36fd8ccc
AK		 2672 return map->num_members;
2673 break;
1e5c6983
DW		 2674 case 1:
2675 case 10:
36fd8ccc 2676 return map->num_members/2;
1e5c6983
DW		 2677 case 5:
2678 return map->num_members - 1;
2679 default:
1ade5cc1 2680 dprintf("unsupported raid level\n");
1e5c6983
DW		 2681 return 0;
2682 }
2683}
2684
2685static __u32 parity_segment_depth(struct imsm_dev *dev)
2686{
238c0a71 2687 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2688 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2689
2690 switch(get_imsm_raid_level(map)) {
2691 case 1:
2692 case 10:
2693 return chunk * map->num_domains;
2694 case 5:
2695 return chunk * map->num_members;
2696 default:
2697 return chunk;
2698 }
2699}
2700
2701static __u32 map_migr_block(struct imsm_dev *dev, __u32 block)
2702{
238c0a71 2703 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2704 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2705 __u32 strip = block / chunk;
2706
2707 switch (get_imsm_raid_level(map)) {
2708 case 1:
2709 case 10: {
2710 __u32 vol_strip = (strip * map->num_domains) + 1;
2711 __u32 vol_stripe = vol_strip / map->num_members;
2712
2713 return vol_stripe * chunk + block % chunk;
2714 } case 5: {
2715 __u32 stripe = strip / (map->num_members - 1);
2716
2717 return stripe * chunk + block % chunk;
2718 }
2719 default:
2720 return 0;
2721 }
2722}
2723
c47b0ff6
AK		 2724static __u64 blocks_per_migr_unit(struct intel_super *super,
2725 struct imsm_dev *dev)
1e5c6983
DW		 2726{
2727 /* calculate the conversion factor between per member 'blocks'
2728 * (md/{resync,rebuild}_start) and imsm migration units, return
2729 * 0 for the 'not migrating' and 'unsupported migration' cases
2730 */
2731 if (!dev->vol.migr_state)
2732 return 0;
2733
2734 switch (migr_type(dev)) {
c47b0ff6
AK		 2735 case MIGR_GEN_MIGR: {
2736 struct migr_record *migr_rec = super->migr_rec;
2737 return __le32_to_cpu(migr_rec->blocks_per_unit);
2738 }
1e5c6983
DW		 2739 case MIGR_VERIFY:
2740 case MIGR_REPAIR:
2741 case MIGR_INIT: {
238c0a71 2742 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2743 __u32 stripes_per_unit;
2744 __u32 blocks_per_unit;
2745 __u32 parity_depth;
2746 __u32 migr_chunk;
2747 __u32 block_map;
2748 __u32 block_rel;
2749 __u32 segment;
2750 __u32 stripe;
2751 __u8 disks;
2752
2753 /* yes, this is really the translation of migr_units to
2754 * per-member blocks in the 'resync' case
2755 */
2756 stripes_per_unit = num_stripes_per_unit_resync(dev);
2757 migr_chunk = migr_strip_blocks_resync(dev);
238c0a71 2758 disks = imsm_num_data_members(dev, MAP_0);
1e5c6983 2759 blocks_per_unit = stripes_per_unit * migr_chunk * disks;
7b1ab482 2760 stripe = __le16_to_cpu(map->blocks_per_strip) * disks;
1e5c6983
DW		 2761 segment = blocks_per_unit / stripe;
2762 block_rel = blocks_per_unit - segment * stripe;
2763 parity_depth = parity_segment_depth(dev);
2764 block_map = map_migr_block(dev, block_rel);
2765 return block_map + parity_depth * segment;
2766 }
2767 case MIGR_REBUILD: {
2768 __u32 stripes_per_unit;
2769 __u32 migr_chunk;
2770
2771 stripes_per_unit = num_stripes_per_unit_rebuild(dev);
2772 migr_chunk = migr_strip_blocks_rebuild(dev);
2773 return migr_chunk * stripes_per_unit;
2774 }
1e5c6983
DW		 2775 case MIGR_STATE_CHANGE:
2776 default:
2777 return 0;
2778 }
2779}
2780
c2c087e6
DW		 2781static int imsm_level_to_layout(int level)
2782{
2783 switch (level) {
2784 case 0:
2785 case 1:
2786 return 0;
2787 case 5:
2788 case 6:
a380c027 2789 return ALGORITHM_LEFT_ASYMMETRIC;
c2c087e6 2790 case 10:
c92a2527 2791 return 0x102;
c2c087e6 2792 }
a18a888e 2793 return UnSet;
c2c087e6
DW		 2794}
2795
8e59f3d8
AK		 2796/*******************************************************************************
2797 * Function: read_imsm_migr_rec
2798 * Description: Function reads imsm migration record from last sector of disk
2799 * Parameters:
2800 * fd : disk descriptor
2801 * super : metadata info
2802 * Returns:
2803 * 0 : success,
2804 * -1 : fail
2805 ******************************************************************************/
2806static int read_imsm_migr_rec(int fd, struct intel_super *super)
2807{
2808 int ret_val = -1;
de44e46f 2809 unsigned int sector_size = super->sector_size;
8e59f3d8
AK		 2810 unsigned long long dsize;
2811
2812 get_dev_size(fd, NULL, &dsize);
de44e46f
PB		 2813 if (lseek64(fd, dsize - (sector_size*MIGR_REC_SECTOR_POSITION),
2814 SEEK_SET) < 0) {
e7b84f9d
N		 2815 pr_err("Cannot seek to anchor block: %s\n",
2816 strerror(errno));
8e59f3d8
AK		 2817 goto out;
2818 }
de44e46f
PB		 2819 if (read(fd, super->migr_rec_buf,
2820 MIGR_REC_BUF_SECTORS*sector_size) !=
2821 MIGR_REC_BUF_SECTORS*sector_size) {
e7b84f9d
N		 2822 pr_err("Cannot read migr record block: %s\n",
2823 strerror(errno));
8e59f3d8
AK		 2824 goto out;
2825 }
2826 ret_val = 0;
de44e46f
PB		 2827 if (sector_size == 4096)
2828 convert_from_4k_imsm_migr_rec(super);
8e59f3d8
AK		 2829
2830out:
2831 return ret_val;
2832}
2833
3136abe5
AK		 2834static struct imsm_dev *imsm_get_device_during_migration(
2835 struct intel_super *super)
2836{
2837
2838 struct intel_dev *dv;
2839
2840 for (dv = super->devlist; dv; dv = dv->next) {
2841 if (is_gen_migration(dv->dev))
2842 return dv->dev;
2843 }
2844 return NULL;
2845}
2846
8e59f3d8
AK		 2847/*******************************************************************************
2848 * Function: load_imsm_migr_rec
2849 * Description: Function reads imsm migration record (it is stored at the last
2850 * sector of disk)
2851 * Parameters:
2852 * super : imsm internal array info
2853 * info : general array info
2854 * Returns:
2855 * 0 : success
2856 * -1 : fail
4c965cc9 2857 * -2 : no migration in progress
8e59f3d8
AK		 2858 ******************************************************************************/
2859static int load_imsm_migr_rec(struct intel_super *super, struct mdinfo *info)
2860{
2861 struct mdinfo *sd;
594dc1b8 2862 struct dl *dl;
8e59f3d8
AK		 2863 char nm[30];
2864 int retval = -1;
2865 int fd = -1;
3136abe5 2866 struct imsm_dev *dev;
594dc1b8 2867 struct imsm_map *map;
b4ab44d8 2868 int slot = -1;
3136abe5
AK		 2869
2870 /* find map under migration */
2871 dev = imsm_get_device_during_migration(super);
2872 /* nothing to load,no migration in progress?
2873 */
2874 if (dev == NULL)
4c965cc9 2875 return -2;
8e59f3d8
AK		 2876
2877 if (info) {
2878 for (sd = info->devs ; sd ; sd = sd->next) {
2879 /* read only from one of the first two slots */
12fe93e9
TM		 2880 if ((sd->disk.raid_disk < 0) ||
2881 (sd->disk.raid_disk > 1))
8e59f3d8 2882 continue;
3136abe5 2883
8e59f3d8
AK		 2884 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
2885 fd = dev_open(nm, O_RDONLY);
2886 if (fd >= 0)
2887 break;
2888 }
2889 }
2890 if (fd < 0) {
12fe93e9 2891 map = get_imsm_map(dev, MAP_0);
8e59f3d8 2892 for (dl = super->disks; dl; dl = dl->next) {
3136abe5
AK		 2893 /* skip spare and failed disks
2894 */
2895 if (dl->index < 0)
2896 continue;
8e59f3d8 2897 /* read only from one of the first two slots */
3136abe5
AK		 2898 if (map)
2899 slot = get_imsm_disk_slot(map, dl->index);
089f9d79 2900 if (map == NULL || slot > 1 || slot < 0)
8e59f3d8
AK		 2901 continue;
2902 sprintf(nm, "%d:%d", dl->major, dl->minor);
2903 fd = dev_open(nm, O_RDONLY);
2904 if (fd >= 0)
2905 break;
2906 }
2907 }
2908 if (fd < 0)
2909 goto out;
2910 retval = read_imsm_migr_rec(fd, super);
2911
2912out:
2913 if (fd >= 0)
2914 close(fd);
2915 return retval;
2916}
2917
9e2d750d 2918#ifndef MDASSEMBLE
c17608ea
AK		 2919/*******************************************************************************
2920 * function: imsm_create_metadata_checkpoint_update
2921 * Description: It creates update for checkpoint change.
2922 * Parameters:
2923 * super : imsm internal array info
2924 * u : pointer to prepared update
2925 * Returns:
2926 * Uptate length.
2927 * If length is equal to 0, input pointer u contains no update
2928 ******************************************************************************/
2929static int imsm_create_metadata_checkpoint_update(
2930 struct intel_super *super,
2931 struct imsm_update_general_migration_checkpoint **u)
2932{
2933
2934 int update_memory_size = 0;
2935
1ade5cc1 2936 dprintf("(enter)\n");
c17608ea
AK		 2937
2938 if (u == NULL)
2939 return 0;
2940 *u = NULL;
2941
2942 /* size of all update data without anchor */
2943 update_memory_size =
2944 sizeof(struct imsm_update_general_migration_checkpoint);
2945
503975b9 2946 *u = xcalloc(1, update_memory_size);
c17608ea 2947 if (*u == NULL) {
1ade5cc1 2948 dprintf("error: cannot get memory\n");
c17608ea
AK		 2949 return 0;
2950 }
2951 (*u)->type = update_general_migration_checkpoint;
2952 (*u)->curr_migr_unit = __le32_to_cpu(super->migr_rec->curr_migr_unit);
1ade5cc1 2953 dprintf("prepared for %u\n", (*u)->curr_migr_unit);
c17608ea
AK		 2954
2955 return update_memory_size;
2956}
2957
c17608ea
AK		 2958static void imsm_update_metadata_locally(struct supertype *st,
2959 void *buf, int len);
2960
687629c2
AK		 2961/*******************************************************************************
2962 * Function: write_imsm_migr_rec
2963 * Description: Function writes imsm migration record
2964 * (at the last sector of disk)
2965 * Parameters:
2966 * super : imsm internal array info
2967 * Returns:
2968 * 0 : success
2969 * -1 : if fail
2970 ******************************************************************************/
2971static int write_imsm_migr_rec(struct supertype *st)
2972{
2973 struct intel_super *super = st->sb;
de44e46f 2974 unsigned int sector_size = super->sector_size;
687629c2
AK		 2975 unsigned long long dsize;
2976 char nm[30];
2977 int fd = -1;
2978 int retval = -1;
2979 struct dl *sd;
c17608ea
AK		 2980 int len;
2981 struct imsm_update_general_migration_checkpoint *u;
3136abe5 2982 struct imsm_dev *dev;
594dc1b8 2983 struct imsm_map *map;
3136abe5
AK		 2984
2985 /* find map under migration */
2986 dev = imsm_get_device_during_migration(super);
2987 /* if no migration, write buffer anyway to clear migr_record
2988 * on disk based on first available device
2989 */
2990 if (dev == NULL)
2991 dev = get_imsm_dev(super, super->current_vol < 0 ? 0 :
2992 super->current_vol);
2993
44bfe6df 2994 map = get_imsm_map(dev, MAP_0);
687629c2 2995
de44e46f
PB		 2996 if (sector_size == 4096)
2997 convert_to_4k_imsm_migr_rec(super);
687629c2 2998 for (sd = super->disks ; sd ; sd = sd->next) {
b4ab44d8 2999 int slot = -1;
3136abe5
AK		 3000
3001 /* skip failed and spare devices */
3002 if (sd->index < 0)
3003 continue;
687629c2 3004 /* write to 2 first slots only */
3136abe5
AK		 3005 if (map)
3006 slot = get_imsm_disk_slot(map, sd->index);
089f9d79 3007 if (map == NULL || slot > 1 || slot < 0)
687629c2 3008 continue;
3136abe5 3009
687629c2
AK		 3010 sprintf(nm, "%d:%d", sd->major, sd->minor);
3011 fd = dev_open(nm, O_RDWR);
3012 if (fd < 0)
3013 continue;
3014 get_dev_size(fd, NULL, &dsize);
de44e46f
PB		 3015 if (lseek64(fd, dsize - (MIGR_REC_SECTOR_POSITION*sector_size),
3016 SEEK_SET) < 0) {
e7b84f9d
N		 3017 pr_err("Cannot seek to anchor block: %s\n",
3018 strerror(errno));
687629c2
AK		 3019 goto out;
3020 }
de44e46f
PB		 3021 if (write(fd, super->migr_rec_buf,
3022 MIGR_REC_BUF_SECTORS*sector_size) !=
3023 MIGR_REC_BUF_SECTORS*sector_size) {
e7b84f9d
N		 3024 pr_err("Cannot write migr record block: %s\n",
3025 strerror(errno));
687629c2
AK		 3026 goto out;
3027 }
3028 close(fd);
3029 fd = -1;
3030 }
de44e46f
PB		 3031 if (sector_size == 4096)
3032 convert_from_4k_imsm_migr_rec(super);
c17608ea
AK		 3033 /* update checkpoint information in metadata */
3034 len = imsm_create_metadata_checkpoint_update(super, &u);
c17608ea
AK		 3035 if (len <= 0) {
3036 dprintf("imsm: Cannot prepare update\n");
3037 goto out;
3038 }
3039 /* update metadata locally */
3040 imsm_update_metadata_locally(st, u, len);
3041 /* and possibly remotely */
3042 if (st->update_tail) {
3043 append_metadata_update(st, u, len);
3044 /* during reshape we do all work inside metadata handler
3045 * manage_reshape(), so metadata update has to be triggered
3046 * insida it
3047 */
3048 flush_metadata_updates(st);
3049 st->update_tail = &st->updates;
3050 } else
3051 free(u);
687629c2
AK		 3052
3053 retval = 0;
3054 out:
3055 if (fd >= 0)
3056 close(fd);
3057 return retval;
3058}
9e2d750d 3059#endif /* MDASSEMBLE */
687629c2 3060
e2962bfc
AK		 3061/* spare/missing disks activations are not allowe when
3062 * array/container performs reshape operation, because
3063 * all arrays in container works on the same disks set
3064 */
3065int imsm_reshape_blocks_arrays_changes(struct intel_super *super)
3066{
3067 int rv = 0;
3068 struct intel_dev *i_dev;
3069 struct imsm_dev *dev;
3070
3071 /* check whole container
3072 */
3073 for (i_dev = super->devlist; i_dev; i_dev = i_dev->next) {
3074 dev = i_dev->dev;
3ad25638 3075 if (is_gen_migration(dev)) {
e2962bfc
AK		 3076 /* No repair during any migration in container
3077 */
3078 rv = 1;
3079 break;
3080 }
3081 }
3082 return rv;
3083}
c41e00b2
AK		 3084static unsigned long long imsm_component_size_aligment_check(int level,
3085 int chunk_size,
f36a9ecd 3086 unsigned int sector_size,
c41e00b2
AK		 3087 unsigned long long component_size)
3088{
3089 unsigned int component_size_alligment;
3090
3091 /* check component size aligment
3092 */
f36a9ecd 3093 component_size_alligment = component_size % (chunk_size/sector_size);
c41e00b2 3094
1ade5cc1 3095 dprintf("(Level: %i, chunk_size = %i, component_size = %llu), component_size_alligment = %u\n",
c41e00b2
AK		 3096 level, chunk_size, component_size,
3097 component_size_alligment);
3098
3099 if (component_size_alligment && (level != 1) && (level != UnSet)) {
3100 dprintf("imsm: reported component size alligned from %llu ",
3101 component_size);
3102 component_size -= component_size_alligment;
1ade5cc1 3103 dprintf_cont("to %llu (%i).\n",
c41e00b2
AK		 3104 component_size, component_size_alligment);
3105 }
3106
3107 return component_size;
3108}
e2962bfc 3109
a5d85af7 3110static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap)
bf5a934a
DW		 3111{
3112 struct intel_super *super = st->sb;
c47b0ff6 3113 struct migr_record *migr_rec = super->migr_rec;
949c47a0 3114 struct imsm_dev *dev = get_imsm_dev(super, super->current_vol);
238c0a71
AK		 3115 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3116 struct imsm_map *prev_map = get_imsm_map(dev, MAP_1);
b335e593 3117 struct imsm_map *map_to_analyse = map;
efb30e7f 3118 struct dl *dl;
a5d85af7 3119 int map_disks = info->array.raid_disks;
bf5a934a 3120
95eeceeb 3121 memset(info, 0, sizeof(*info));
b335e593
AK		 3122 if (prev_map)
3123 map_to_analyse = prev_map;
3124
ca0748fa 3125 dl = super->current_disk;
9894ec0d 3126
bf5a934a 3127 info->container_member = super->current_vol;
cd0430a1 3128 info->array.raid_disks = map->num_members;
b335e593 3129 info->array.level = get_imsm_raid_level(map_to_analyse);
bf5a934a
DW		 3130 info->array.layout = imsm_level_to_layout(info->array.level);
3131 info->array.md_minor = -1;
3132 info->array.ctime = 0;
3133 info->array.utime = 0;
b335e593
AK		 3134 info->array.chunk_size =
3135 __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9;
301406c9 3136 info->array.state = !dev->vol.dirty;
da9b4a62
DW		 3137 info->custom_array_size = __le32_to_cpu(dev->size_high);
3138 info->custom_array_size <<= 32;
3139 info->custom_array_size |= __le32_to_cpu(dev->size_low);
3ad25638
AK		 3140 info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb);
3141
3f510843 3142 if (is_gen_migration(dev)) {
3f83228a 3143 info->reshape_active = 1;
b335e593
AK		 3144 info->new_level = get_imsm_raid_level(map);
3145 info->new_layout = imsm_level_to_layout(info->new_level);
3146 info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9;
3f83228a 3147 info->delta_disks = map->num_members - prev_map->num_members;
493f5dd6
N		 3148 if (info->delta_disks) {
3149 /* this needs to be applied to every array
3150 * in the container.
3151 */
81219e70 3152 info->reshape_active = CONTAINER_RESHAPE;
493f5dd6 3153 }
3f83228a
N		 3154 /* We shape information that we give to md might have to be
3155 * modify to cope with md's requirement for reshaping arrays.
3156 * For example, when reshaping a RAID0, md requires it to be
3157 * presented as a degraded RAID4.
3158 * Also if a RAID0 is migrating to a RAID5 we need to specify
3159 * the array as already being RAID5, but the 'before' layout
3160 * is a RAID4-like layout.
3161 */
3162 switch (info->array.level) {
3163 case 0:
3164 switch(info->new_level) {
3165 case 0:
3166 /* conversion is happening as RAID4 */
3167 info->array.level = 4;
3168 info->array.raid_disks += 1;
3169 break;
3170 case 5:
3171 /* conversion is happening as RAID5 */
3172 info->array.level = 5;
3173 info->array.layout = ALGORITHM_PARITY_N;
3f83228a
N		 3174 info->delta_disks -= 1;
3175 break;
3176 default:
3177 /* FIXME error message */
3178 info->array.level = UnSet;
3179 break;
3180 }
3181 break;
3182 }
b335e593
AK		 3183 } else {
3184 info->new_level = UnSet;
3185 info->new_layout = UnSet;
3186 info->new_chunk = info->array.chunk_size;
3f83228a 3187 info->delta_disks = 0;
b335e593 3188 }
ca0748fa 3189
efb30e7f
DW		 3190 if (dl) {
3191 info->disk.major = dl->major;
3192 info->disk.minor = dl->minor;
ca0748fa 3193 info->disk.number = dl->index;
656b6b5a
N		 3194 info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse,
3195 dl->index);
efb30e7f 3196 }
bf5a934a 3197
5551b113 3198 info->data_offset = pba_of_lba0(map_to_analyse);
06fb291a
PB		 3199
3200 if (info->array.level == 5) {
3201 info->component_size = num_data_stripes(map_to_analyse) *
3202 map_to_analyse->blocks_per_strip;
3203 } else {
3204 info->component_size = blocks_per_member(map_to_analyse);
3205 }
139dae11 3206
c41e00b2
AK		 3207 info->component_size = imsm_component_size_aligment_check(
3208 info->array.level,
3209 info->array.chunk_size,
f36a9ecd 3210 super->sector_size,
c41e00b2 3211 info->component_size);
b12796be 3212 info->bb.supported = 0;
139dae11 3213
301406c9 3214 memset(info->uuid, 0, sizeof(info->uuid));
921d9e16 3215 info->recovery_start = MaxSector;
bf5a934a 3216
d2e6d5d6 3217 info->reshape_progress = 0;
b6796ce1 3218 info->resync_start = MaxSector;
b9172665
AK		 3219 if ((map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED ||
3220 dev->vol.dirty) &&
3221 imsm_reshape_blocks_arrays_changes(super) == 0) {
301406c9 3222 info->resync_start = 0;
b6796ce1
AK		 3223 }
3224 if (dev->vol.migr_state) {
1e5c6983
DW		 3225 switch (migr_type(dev)) {
3226 case MIGR_REPAIR:
3227 case MIGR_INIT: {
c47b0ff6
AK		 3228 __u64 blocks_per_unit = blocks_per_migr_unit(super,
3229 dev);
1e5c6983
DW		 3230 __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit);
3231
3232 info->resync_start = blocks_per_unit * units;
3233 break;
3234 }
d2e6d5d6 3235 case MIGR_GEN_MIGR: {
c47b0ff6
AK		 3236 __u64 blocks_per_unit = blocks_per_migr_unit(super,
3237 dev);
3238 __u64 units = __le32_to_cpu(migr_rec->curr_migr_unit);
04fa9523
AK		 3239 unsigned long long array_blocks;
3240 int used_disks;
d2e6d5d6 3241
befb629b
AK		 3242 if (__le32_to_cpu(migr_rec->ascending_migr) &&
3243 (units <
3244 (__le32_to_cpu(migr_rec->num_migr_units)-1)) &&
3245 (super->migr_rec->rec_status ==
3246 __cpu_to_le32(UNIT_SRC_IN_CP_AREA)))
3247 units++;
3248
d2e6d5d6 3249 info->reshape_progress = blocks_per_unit * units;
6289d1e0 3250
7a862a02 3251 dprintf("IMSM: General Migration checkpoint : %llu (%llu) -> read reshape progress : %llu\n",
19986c72
MB		 3252 (unsigned long long)units,
3253 (unsigned long long)blocks_per_unit,
3254 info->reshape_progress);
75156c46 3255
238c0a71 3256 used_disks = imsm_num_data_members(dev, MAP_1);
75156c46 3257 if (used_disks > 0) {
5551b113 3258 array_blocks = blocks_per_member(map) *
75156c46
AK		 3259 used_disks;
3260 /* round array size down to closest MB
3261 */
3262 info->custom_array_size = (array_blocks
3263 >> SECT_PER_MB_SHIFT)
3264 << SECT_PER_MB_SHIFT;
3265 }
d2e6d5d6 3266 }
1e5c6983
DW		 3267 case MIGR_VERIFY:
3268 /* we could emulate the checkpointing of
3269 * 'sync_action=check' migrations, but for now
3270 * we just immediately complete them
3271 */
3272 case MIGR_REBUILD:
3273 /* this is handled by container_content_imsm() */
1e5c6983
DW		 3274 case MIGR_STATE_CHANGE:
3275 /* FIXME handle other migrations */
3276 default:
3277 /* we are not dirty, so... */
3278 info->resync_start = MaxSector;
3279 }
b6796ce1 3280 }
301406c9
DW		 3281
3282 strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN);
3283 info->name[MAX_RAID_SERIAL_LEN] = 0;
bf5a934a 3284
f35f2525
N		 3285 info->array.major_version = -1;
3286 info->array.minor_version = -2;
4dd2df09 3287 sprintf(info->text_version, "/%s/%d", st->container_devnm, info->container_member);
a67dd8cc 3288 info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */
51006d85 3289 uuid_from_super_imsm(st, info->uuid);
a5d85af7
N		 3290
3291 if (dmap) {
3292 int i, j;
3293 for (i=0; i<map_disks; i++) {
3294 dmap[i] = 0;
3295 if (i < info->array.raid_disks) {
3296 struct imsm_disk *dsk;
238c0a71 3297 j = get_imsm_disk_idx(dev, i, MAP_X);
a5d85af7
N		 3298 dsk = get_imsm_disk(super, j);
3299 if (dsk && (dsk->status & CONFIGURED_DISK))
3300 dmap[i] = 1;
3301 }
3302 }
3303 }
81ac8b4d 3304}
bf5a934a 3305
3b451610
AK		 3306static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev,
3307 int failed, int look_in_map);
3308
3309static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev,
3310 int look_in_map);
3311
b4ab44d8 3312#ifndef MDASSEMBLE
3b451610
AK		 3313static void manage_second_map(struct intel_super *super, struct imsm_dev *dev)
3314{
3315 if (is_gen_migration(dev)) {
3316 int failed;
3317 __u8 map_state;
3318 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
3319
3320 failed = imsm_count_failed(super, dev, MAP_1);
238c0a71 3321 map_state = imsm_check_degraded(super, dev, failed, MAP_1);
3b451610
AK		 3322 if (map2->map_state != map_state) {
3323 map2->map_state = map_state;
3324 super->updates_pending++;
3325 }
3326 }
3327}
b4ab44d8 3328#endif
97b4d0e9
DW		 3329
3330static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index)
3331{
3332 struct dl *d;
3333
3334 for (d = super->missing; d; d = d->next)
3335 if (d->index == index)
3336 return &d->disk;
3337 return NULL;
3338}
3339
a5d85af7 3340static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map)
4f5bc454
DW		 3341{
3342 struct intel_super *super = st->sb;
4f5bc454 3343 struct imsm_disk *disk;
a5d85af7 3344 int map_disks = info->array.raid_disks;
ab3cb6b3
N		 3345 int max_enough = -1;
3346 int i;
3347 struct imsm_super *mpb;
4f5bc454 3348
bf5a934a 3349 if (super->current_vol >= 0) {
a5d85af7 3350 getinfo_super_imsm_volume(st, info, map);
bf5a934a
DW		 3351 return;
3352 }
95eeceeb 3353 memset(info, 0, sizeof(*info));
d23fe947
DW		 3354
3355 /* Set raid_disks to zero so that Assemble will always pull in valid
3356 * spares
3357 */
3358 info->array.raid_disks = 0;
cdddbdbc
DW		 3359 info->array.level = LEVEL_CONTAINER;
3360 info->array.layout = 0;
3361 info->array.md_minor = -1;
1011e834 3362 info->array.ctime = 0; /* N/A for imsm */
cdddbdbc
DW		 3363 info->array.utime = 0;
3364 info->array.chunk_size = 0;
3365
3366 info->disk.major = 0;
3367 info->disk.minor = 0;
cdddbdbc 3368 info->disk.raid_disk = -1;
c2c087e6 3369 info->reshape_active = 0;
f35f2525
N		 3370 info->array.major_version = -1;
3371 info->array.minor_version = -2;
c2c087e6 3372 strcpy(info->text_version, "imsm");
a67dd8cc 3373 info->safe_mode_delay = 0;
c2c087e6
DW		 3374 info->disk.number = -1;
3375 info->disk.state = 0;
c5afc314 3376 info->name[0] = 0;
921d9e16 3377 info->recovery_start = MaxSector;
3ad25638 3378 info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb);
b12796be 3379 info->bb.supported = 0;
c2c087e6 3380
97b4d0e9 3381 /* do we have the all the insync disks that we expect? */
ab3cb6b3 3382 mpb = super->anchor;
b7d81a38 3383 info->events = __le32_to_cpu(mpb->generation_num);
97b4d0e9 3384
ab3cb6b3
N		 3385 for (i = 0; i < mpb->num_raid_devs; i++) {
3386 struct imsm_dev *dev = get_imsm_dev(super, i);
3387 int failed, enough, j, missing = 0;
3388 struct imsm_map *map;
3389 __u8 state;
97b4d0e9 3390
3b451610
AK		 3391 failed = imsm_count_failed(super, dev, MAP_0);
3392 state = imsm_check_degraded(super, dev, failed, MAP_0);
238c0a71 3393 map = get_imsm_map(dev, MAP_0);
ab3cb6b3
N		 3394
3395 /* any newly missing disks?
3396 * (catches single-degraded vs double-degraded)
3397 */
3398 for (j = 0; j < map->num_members; j++) {
238c0a71 3399 __u32 ord = get_imsm_ord_tbl_ent(dev, j, MAP_0);
ab3cb6b3
N		 3400 __u32 idx = ord_to_idx(ord);
3401
3402 if (!(ord & IMSM_ORD_REBUILD) &&
3403 get_imsm_missing(super, idx)) {
3404 missing = 1;
3405 break;
3406 }
97b4d0e9 3407 }
ab3cb6b3
N		 3408
3409 if (state == IMSM_T_STATE_FAILED)
3410 enough = -1;
3411 else if (state == IMSM_T_STATE_DEGRADED &&
3412 (state != map->map_state || missing))
3413 enough = 0;
3414 else /* we're normal, or already degraded */
3415 enough = 1;
d2bde6d3
AK		 3416 if (is_gen_migration(dev) && missing) {
3417 /* during general migration we need all disks
3418 * that process is running on.
3419 * No new missing disk is allowed.
3420 */
3421 max_enough = -1;
3422 enough = -1;
3423 /* no more checks necessary
3424 */
3425 break;
3426 }
ab3cb6b3
N		 3427 /* in the missing/failed disk case check to see
3428 * if at least one array is runnable
3429 */
3430 max_enough = max(max_enough, enough);
3431 }
1ade5cc1 3432 dprintf("enough: %d\n", max_enough);
ab3cb6b3 3433 info->container_enough = max_enough;
97b4d0e9 3434
4a04ec6c 3435 if (super->disks) {
14e8215b
DW		 3436 __u32 reserved = imsm_reserved_sectors(super, super->disks);
3437
b9f594fe 3438 disk = &super->disks->disk;
5551b113 3439 info->data_offset = total_blocks(&super->disks->disk) - reserved;
14e8215b 3440 info->component_size = reserved;
25ed7e59 3441 info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0;
df474657
DW		 3442 /* we don't change info->disk.raid_disk here because
3443 * this state will be finalized in mdmon after we have
3444 * found the 'most fresh' version of the metadata
3445 */
25ed7e59
DW		 3446 info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0;
3447 info->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC);
cdddbdbc 3448 }
a575e2a7
DW		 3449
3450 /* only call uuid_from_super_imsm when this disk is part of a populated container,
3451 * ->compare_super may have updated the 'num_raid_devs' field for spares
3452 */
3453 if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs)
36ba7d48 3454 uuid_from_super_imsm(st, info->uuid);
22e263f6
AC		 3455 else
3456 memcpy(info->uuid, uuid_zero, sizeof(uuid_zero));
a5d85af7
N		 3457
3458 /* I don't know how to compute 'map' on imsm, so use safe default */
3459 if (map) {
3460 int i;
3461 for (i = 0; i < map_disks; i++)
3462 map[i] = 1;
3463 }
3464
cdddbdbc
DW		 3465}
3466
5c4cd5da
AC		 3467/* allocates memory and fills disk in mdinfo structure
3468 * for each disk in array */
3469struct mdinfo *getinfo_super_disks_imsm(struct supertype *st)
3470{
594dc1b8 3471 struct mdinfo *mddev;
5c4cd5da
AC		 3472 struct intel_super *super = st->sb;
3473 struct imsm_disk *disk;
3474 int count = 0;
3475 struct dl *dl;
3476 if (!super || !super->disks)
3477 return NULL;
3478 dl = super->disks;
503975b9 3479 mddev = xcalloc(1, sizeof(*mddev));
5c4cd5da
AC		 3480 while (dl) {
3481 struct mdinfo *tmp;
3482 disk = &dl->disk;
503975b9 3483 tmp = xcalloc(1, sizeof(*tmp));
5c4cd5da
AC		 3484 if (mddev->devs)
3485 tmp->next = mddev->devs;
3486 mddev->devs = tmp;
3487 tmp->disk.number = count++;
3488 tmp->disk.major = dl->major;
3489 tmp->disk.minor = dl->minor;
3490 tmp->disk.state = is_configured(disk) ?
3491 (1 << MD_DISK_ACTIVE) : 0;
3492 tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0;
3493 tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC);
3494 tmp->disk.raid_disk = -1;
3495 dl = dl->next;
3496 }
3497 return mddev;
3498}
3499
cdddbdbc
DW		 3500static int update_super_imsm(struct supertype *st, struct mdinfo *info,
3501 char *update, char *devname, int verbose,
3502 int uuid_set, char *homehost)
3503{
f352c545
DW		 3504 /* For 'assemble' and 'force' we need to return non-zero if any
3505 * change was made. For others, the return value is ignored.
3506 * Update options are:
3507 * force-one : This device looks a bit old but needs to be included,
3508 * update age info appropriately.
3509 * assemble: clear any 'faulty' flag to allow this device to
3510 * be assembled.
3511 * force-array: Array is degraded but being forced, mark it clean
3512 * if that will be needed to assemble it.
3513 *
3514 * newdev: not used ????
3515 * grow: Array has gained a new device - this is currently for
3516 * linear only
3517 * resync: mark as dirty so a resync will happen.
3518 * name: update the name - preserving the homehost
6e46bf34 3519 * uuid: Change the uuid of the array to match watch is given
f352c545
DW		 3520 *
3521 * Following are not relevant for this imsm:
3522 * sparc2.2 : update from old dodgey metadata
3523 * super-minor: change the preferred_minor number
3524 * summaries: update redundant counters.
f352c545
DW		 3525 * homehost: update the recorded homehost
3526 * _reshape_progress: record new reshape_progress position.
3527 */
6e46bf34
DW		 3528 int rv = 1;
3529 struct intel_super *super = st->sb;
3530 struct imsm_super *mpb;
f352c545 3531
6e46bf34
DW		 3532 /* we can only update container info */
3533 if (!super || super->current_vol >= 0 || !super->anchor)
3534 return 1;
3535
3536 mpb = super->anchor;
3537
81a5b4f5
N		 3538 if (strcmp(update, "uuid") == 0) {
3539 /* We take this to mean that the family_num should be updated.
3540 * However that is much smaller than the uuid so we cannot really
3541 * allow an explicit uuid to be given. And it is hard to reliably
3542 * know if one was.
3543 * So if !uuid_set we know the current uuid is random and just used
3544 * the first 'int' and copy it to the other 3 positions.
3545 * Otherwise we require the 4 'int's to be the same as would be the
3546 * case if we are using a random uuid. So an explicit uuid will be
3547 * accepted as long as all for ints are the same... which shouldn't hurt
6e46bf34 3548 */
81a5b4f5
N		 3549 if (!uuid_set) {
3550 info->uuid[1] = info->uuid[2] = info->uuid[3] = info->uuid[0];
6e46bf34 3551 rv = 0;
81a5b4f5
N		 3552 } else {
3553 if (info->uuid[0] != info->uuid[1] ||
3554 info->uuid[1] != info->uuid[2] ||
3555 info->uuid[2] != info->uuid[3])
3556 rv = -1;
3557 else
3558 rv = 0;
6e46bf34 3559 }
81a5b4f5
N		 3560 if (rv == 0)
3561 mpb->orig_family_num = info->uuid[0];
6e46bf34
DW		 3562 } else if (strcmp(update, "assemble") == 0)
3563 rv = 0;
3564 else
1e2b2765 3565 rv = -1;
f352c545 3566
6e46bf34
DW		 3567 /* successful update? recompute checksum */
3568 if (rv == 0)
3569 mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb));
f352c545
DW		 3570
3571 return rv;
cdddbdbc
DW		 3572}
3573
c2c087e6 3574static size_t disks_to_mpb_size(int disks)
cdddbdbc 3575{
c2c087e6 3576 size_t size;
cdddbdbc 3577
c2c087e6
DW		 3578 size = sizeof(struct imsm_super);
3579 size += (disks - 1) * sizeof(struct imsm_disk);
3580 size += 2 * sizeof(struct imsm_dev);
3581 /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */
3582 size += (4 - 2) * sizeof(struct imsm_map);
3583 /* 4 possible disk_ord_tbl's */
3584 size += 4 * (disks - 1) * sizeof(__u32);
bbab0940
TM		 3585 /* maximum bbm log */
3586 size += sizeof(struct bbm_log);
c2c087e6
DW		 3587
3588 return size;
3589}
3590
387fcd59
N		 3591static __u64 avail_size_imsm(struct supertype *st, __u64 devsize,
3592 unsigned long long data_offset)
c2c087e6
DW		 3593{
3594 if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS))
3595 return 0;
3596
3597 return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS);
cdddbdbc
DW		 3598}
3599
ba2de7ba
DW		 3600static void free_devlist(struct intel_super *super)
3601{
3602 struct intel_dev *dv;
3603
3604 while (super->devlist) {
3605 dv = super->devlist->next;
3606 free(super->devlist->dev);
3607 free(super->devlist);
3608 super->devlist = dv;
3609 }
3610}
3611
3612static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src)
3613{
3614 memcpy(dest, src, sizeof_imsm_dev(src, 0));
3615}
3616
cdddbdbc
DW		 3617static int compare_super_imsm(struct supertype *st, struct supertype *tst)
3618{
3619 /*
3620 * return:
3621 * 0 same, or first was empty, and second was copied
3622 * 1 second had wrong number
3623 * 2 wrong uuid
3624 * 3 wrong other info
3625 */
3626 struct intel_super *first = st->sb;
3627 struct intel_super *sec = tst->sb;
3628
5d500228
N		 3629 if (!first) {
3630 st->sb = tst->sb;
3631 tst->sb = NULL;
3632 return 0;
3633 }
8603ea6f
LM		 3634 /* in platform dependent environment test if the disks
3635 * use the same Intel hba
cb8f6859 3636 * If not on Intel hba at all, allow anything.
8603ea6f 3637 */
6b781d33
AP		 3638 if (!check_env("IMSM_NO_PLATFORM") && first->hba && sec->hba) {
3639 if (first->hba->type != sec->hba->type) {
8603ea6f 3640 fprintf(stderr,
6b781d33
AP		 3641 "HBAs of devices do not match %s != %s\n",
3642 get_sys_dev_type(first->hba->type),
3643 get_sys_dev_type(sec->hba->type));
3644 return 3;
3645 }
3646 if (first->orom != sec->orom) {
3647 fprintf(stderr,
3648 "HBAs of devices do not match %s != %s\n",
3649 first->hba->pci_id, sec->hba->pci_id);
8603ea6f
LM		 3650 return 3;
3651 }
3652 }
cdddbdbc 3653
d23fe947
DW		 3654 /* if an anchor does not have num_raid_devs set then it is a free
3655 * floating spare
3656 */
3657 if (first->anchor->num_raid_devs > 0 &&
3658 sec->anchor->num_raid_devs > 0) {
a2b97981
DW		 3659 /* Determine if these disks might ever have been
3660 * related. Further disambiguation can only take place
3661 * in load_super_imsm_all
3662 */
3663 __u32 first_family = first->anchor->orig_family_num;
3664 __u32 sec_family = sec->anchor->orig_family_num;
3665
f796af5d
DW		 3666 if (memcmp(first->anchor->sig, sec->anchor->sig,
3667 MAX_SIGNATURE_LENGTH) != 0)
3668 return 3;
3669
a2b97981
DW		 3670 if (first_family == 0)
3671 first_family = first->anchor->family_num;
3672 if (sec_family == 0)
3673 sec_family = sec->anchor->family_num;
3674
3675 if (first_family != sec_family)
d23fe947 3676 return 3;
f796af5d 3677
d23fe947 3678 }
cdddbdbc 3679
3e372e5a
DW		 3680 /* if 'first' is a spare promote it to a populated mpb with sec's
3681 * family number
3682 */
3683 if (first->anchor->num_raid_devs == 0 &&
3684 sec->anchor->num_raid_devs > 0) {
78d30f94 3685 int i;
ba2de7ba
DW		 3686 struct intel_dev *dv;
3687 struct imsm_dev *dev;
78d30f94
DW		 3688
3689 /* we need to copy raid device info from sec if an allocation
3690 * fails here we don't associate the spare
3691 */
3692 for (i = 0; i < sec->anchor->num_raid_devs; i++) {
503975b9
N		 3693 dv = xmalloc(sizeof(*dv));
3694 dev = xmalloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1));
ba2de7ba
DW		 3695 dv->dev = dev;
3696 dv->index = i;
3697 dv->next = first->devlist;
3698 first->devlist = dv;
78d30f94 3699 }
709743c5 3700 if (i < sec->anchor->num_raid_devs) {
ba2de7ba
DW		 3701 /* allocation failure */
3702 free_devlist(first);
e12b3daa 3703 pr_err("imsm: failed to associate spare\n");
ba2de7ba 3704 return 3;
78d30f94 3705 }
3e372e5a 3706 first->anchor->num_raid_devs = sec->anchor->num_raid_devs;
148acb7b 3707 first->anchor->orig_family_num = sec->anchor->orig_family_num;
3e372e5a 3708 first->anchor->family_num = sec->anchor->family_num;
ac6449be 3709 memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH);
709743c5
DW		 3710 for (i = 0; i < sec->anchor->num_raid_devs; i++)
3711 imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i));
3e372e5a
DW		 3712 }
3713
cdddbdbc
DW		 3714 return 0;
3715}
3716
0030e8d6
DW		 3717static void fd2devname(int fd, char *name)
3718{
3719 struct stat st;
3720 char path[256];
33a6535d 3721 char dname[PATH_MAX];
0030e8d6
DW		 3722 char *nm;
3723 int rv;
3724
3725 name[0] = '\0';
3726 if (fstat(fd, &st) != 0)
3727 return;
3728 sprintf(path, "/sys/dev/block/%d:%d",
3729 major(st.st_rdev), minor(st.st_rdev));
3730
9cf014ec 3731 rv = readlink(path, dname, sizeof(dname)-1);
0030e8d6
DW		 3732 if (rv <= 0)
3733 return;
9587c373 3734
0030e8d6
DW		 3735 dname[rv] = '\0';
3736 nm = strrchr(dname, '/');
7897de29
JS		 3737 if (nm) {
3738 nm++;
3739 snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm);
3740 }
0030e8d6
DW		 3741}
3742
21e9380b
AP		 3743static int nvme_get_serial(int fd, void *buf, size_t buf_len)
3744{
3745 char path[60];
3746 char *name = fd2kname(fd);
3747
3748 if (!name)
3749 return 1;
3750
3751 if (strncmp(name, "nvme", 4) != 0)
3752 return 1;
3753
3754 snprintf(path, sizeof(path) - 1, "/sys/block/%s/device/serial", name);
3755
3756 return load_sys(path, buf, buf_len);
3757}
3758
cdddbdbc
DW		 3759extern int scsi_get_serial(int fd, void *buf, size_t buf_len);
3760
3761static int imsm_read_serial(int fd, char *devname,
3762 __u8 serial[MAX_RAID_SERIAL_LEN])
3763{
21e9380b 3764 char buf[50];
cdddbdbc 3765 int rv;
1f24f035 3766 int len;
316e2bf4
DW		 3767 char *dest;
3768 char *src;
21e9380b
AP		 3769 unsigned int i;
3770
3771 memset(buf, 0, sizeof(buf));
cdddbdbc 3772
21e9380b 3773 rv = nvme_get_serial(fd, buf, sizeof(buf));
cdddbdbc 3774
21e9380b
AP		 3775 if (rv)
3776 rv = scsi_get_serial(fd, buf, sizeof(buf));
f9ba0ff1 3777
40ebbb9c 3778 if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) {
f9ba0ff1
DW		 3779 memset(serial, 0, MAX_RAID_SERIAL_LEN);
3780 fd2devname(fd, (char *) serial);
0030e8d6
DW		 3781 return 0;
3782 }
3783
cdddbdbc
DW		 3784 if (rv != 0) {
3785 if (devname)
e7b84f9d
N		 3786 pr_err("Failed to retrieve serial for %s\n",
3787 devname);
cdddbdbc
DW		 3788 return rv;
3789 }
3790
316e2bf4
DW		 3791 /* trim all whitespace and non-printable characters and convert
3792 * ':' to ';'
3793 */
21e9380b
AP		 3794 for (i = 0, dest = buf; i < sizeof(buf) && buf[i]; i++) {
3795 src = &buf[i];
316e2bf4
DW		 3796 if (*src > 0x20) {
3797 /* ':' is reserved for use in placeholder serial
3798 * numbers for missing disks
3799 */
3800 if (*src == ':')
3801 *dest++ = ';';
3802 else
3803 *dest++ = *src;
3804 }
3805 }
21e9380b
AP		 3806 len = dest - buf;
3807 dest = buf;
316e2bf4
DW		 3808
3809 /* truncate leading characters */
3810 if (len > MAX_RAID_SERIAL_LEN) {
3811 dest += len - MAX_RAID_SERIAL_LEN;
1f24f035 3812 len = MAX_RAID_SERIAL_LEN;
316e2bf4 3813 }
5c3db629 3814
5c3db629 3815 memset(serial, 0, MAX_RAID_SERIAL_LEN);
316e2bf4 3816 memcpy(serial, dest, len);
cdddbdbc
DW		 3817
3818 return 0;
3819}
3820
1f24f035
DW		 3821static int serialcmp(__u8 *s1, __u8 *s2)
3822{
3823 return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN);
3824}
3825
3826static void serialcpy(__u8 *dest, __u8 *src)
3827{
3828 strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN);
3829}
3830
54c2c1ea
DW		 3831static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super)
3832{
3833 struct dl *dl;
3834
3835 for (dl = super->disks; dl; dl = dl->next)
3836 if (serialcmp(dl->serial, serial) == 0)
3837 break;
3838
3839 return dl;
3840}
3841
a2b97981
DW		 3842static struct imsm_disk *
3843__serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx)
3844{
3845 int i;
3846
3847 for (i = 0; i < mpb->num_disks; i++) {
3848 struct imsm_disk *disk = __get_imsm_disk(mpb, i);
3849
3850 if (serialcmp(disk->serial, serial) == 0) {
3851 if (idx)
3852 *idx = i;
3853 return disk;
3854 }
3855 }
3856
3857 return NULL;
3858}
3859
cdddbdbc
DW		 3860static int
3861load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd)
3862{
a2b97981 3863 struct imsm_disk *disk;
cdddbdbc
DW		 3864 struct dl *dl;
3865 struct stat stb;
cdddbdbc 3866 int rv;
a2b97981 3867 char name[40];
d23fe947
DW		 3868 __u8 serial[MAX_RAID_SERIAL_LEN];
3869
3870 rv = imsm_read_serial(fd, devname, serial);
3871
3872 if (rv != 0)
3873 return 2;
3874
503975b9 3875 dl = xcalloc(1, sizeof(*dl));
cdddbdbc 3876
a2b97981
DW		 3877 fstat(fd, &stb);
3878 dl->major = major(stb.st_rdev);
3879 dl->minor = minor(stb.st_rdev);
3880 dl->next = super->disks;
3881 dl->fd = keep_fd ? fd : -1;
3882 assert(super->disks == NULL);
3883 super->disks = dl;
3884 serialcpy(dl->serial, serial);
3885 dl->index = -2;
3886 dl->e = NULL;
3887 fd2devname(fd, name);
3888 if (devname)
503975b9 3889 dl->devname = xstrdup(devname);
a2b97981 3890 else
503975b9 3891 dl->devname = xstrdup(name);
cdddbdbc 3892
d23fe947 3893 /* look up this disk's index in the current anchor */
a2b97981
DW		 3894 disk = __serial_to_disk(dl->serial, super->anchor, &dl->index);
3895 if (disk) {
3896 dl->disk = *disk;
3897 /* only set index on disks that are a member of a
3898 * populated contianer, i.e. one with raid_devs
3899 */
3900 if (is_failed(&dl->disk))
3f6efecc 3901 dl->index = -2;
a2b97981
DW		 3902 else if (is_spare(&dl->disk))
3903 dl->index = -1;
3f6efecc
DW		 3904 }
3905
949c47a0
DW		 3906 return 0;
3907}
3908
0e600426 3909#ifndef MDASSEMBLE
0c046afd
DW		 3910/* When migrating map0 contains the 'destination' state while map1
3911 * contains the current state. When not migrating map0 contains the
3912 * current state. This routine assumes that map[0].map_state is set to
3913 * the current array state before being called.
3914 *
3915 * Migration is indicated by one of the following states
3916 * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed)
e3bba0e0 3917 * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal
0c046afd 3918 * map1state=unitialized)
1484e727 3919 * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal
0c046afd 3920 * map1state=normal)
e3bba0e0 3921 * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal
0c046afd 3922 * map1state=degraded)
8e59f3d8
AK		 3923 * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal
3924 * map1state=normal)
0c046afd 3925 */
8e59f3d8
AK		 3926static void migrate(struct imsm_dev *dev, struct intel_super *super,
3927 __u8 to_state, int migr_type)
3393c6af 3928{
0c046afd 3929 struct imsm_map *dest;
238c0a71 3930 struct imsm_map *src = get_imsm_map(dev, MAP_0);
3393c6af 3931
0c046afd 3932 dev->vol.migr_state = 1;
1484e727 3933 set_migr_type(dev, migr_type);
f8f603f1 3934 dev->vol.curr_migr_unit = 0;
238c0a71 3935 dest = get_imsm_map(dev, MAP_1);
0c046afd 3936
0556e1a2 3937 /* duplicate and then set the target end state in map[0] */
3393c6af 3938 memcpy(dest, src, sizeof_imsm_map(src));
089f9d79 3939 if (migr_type == MIGR_REBUILD || migr_type == MIGR_GEN_MIGR) {
0556e1a2
DW		 3940 __u32 ord;
3941 int i;
3942
3943 for (i = 0; i < src->num_members; i++) {
3944 ord = __le32_to_cpu(src->disk_ord_tbl[i]);
3945 set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord));
3946 }
3947 }
3948
8e59f3d8
AK		 3949 if (migr_type == MIGR_GEN_MIGR)
3950 /* Clear migration record */
3951 memset(super->migr_rec, 0, sizeof(struct migr_record));
3952
0c046afd 3953 src->map_state = to_state;
949c47a0 3954}
f8f603f1 3955
809da78e
AK		 3956static void end_migration(struct imsm_dev *dev, struct intel_super *super,
3957 __u8 map_state)
f8f603f1 3958{
238c0a71
AK		 3959 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3960 struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state == 0 ?
3961 MAP_0 : MAP_1);
28bce06f 3962 int i, j;
0556e1a2
DW		 3963
3964 /* merge any IMSM_ORD_REBUILD bits that were not successfully
3965 * completed in the last migration.
3966 *
28bce06f 3967 * FIXME add support for raid-level-migration
0556e1a2 3968 */
089f9d79
JS		 3969 if (map_state != map->map_state && (is_gen_migration(dev) == 0) &&
3970 prev->map_state != IMSM_T_STATE_UNINITIALIZED) {
809da78e
AK		 3971 /* when final map state is other than expected
3972 * merge maps (not for migration)
3973 */
3974 int failed;
3975
3976 for (i = 0; i < prev->num_members; i++)
3977 for (j = 0; j < map->num_members; j++)
3978 /* during online capacity expansion
3979 * disks position can be changed
3980 * if takeover is used
3981 */
3982 if (ord_to_idx(map->disk_ord_tbl[j]) ==
3983 ord_to_idx(prev->disk_ord_tbl[i])) {
3984 map->disk_ord_tbl[j] |=
3985 prev->disk_ord_tbl[i];
3986 break;
3987 }
3988 failed = imsm_count_failed(super, dev, MAP_0);
3989 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
3990 }
f8f603f1
DW		 3991
3992 dev->vol.migr_state = 0;
ea672ee1 3993 set_migr_type(dev, 0);
f8f603f1
DW		 3994 dev->vol.curr_migr_unit = 0;
3995 map->map_state = map_state;
3996}
0e600426 3997#endif
949c47a0
DW		 3998
3999static int parse_raid_devices(struct intel_super *super)
4000{
4001 int i;
4002 struct imsm_dev *dev_new;
4d7b1503 4003 size_t len, len_migr;
401d313b 4004 size_t max_len = 0;
4d7b1503
DW		 4005 size_t space_needed = 0;
4006 struct imsm_super *mpb = super->anchor;
949c47a0
DW		 4007
4008 for (i = 0; i < super->anchor->num_raid_devs; i++) {
4009 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
ba2de7ba 4010 struct intel_dev *dv;
949c47a0 4011
4d7b1503
DW		 4012 len = sizeof_imsm_dev(dev_iter, 0);
4013 len_migr = sizeof_imsm_dev(dev_iter, 1);
4014 if (len_migr > len)
4015 space_needed += len_migr - len;
ca9de185 4016
503975b9 4017 dv = xmalloc(sizeof(*dv));
401d313b
AK		 4018 if (max_len < len_migr)
4019 max_len = len_migr;
4020 if (max_len > len_migr)
4021 space_needed += max_len - len_migr;
503975b9 4022 dev_new = xmalloc(max_len);
949c47a0 4023 imsm_copy_dev(dev_new, dev_iter);
ba2de7ba
DW		 4024 dv->dev = dev_new;
4025 dv->index = i;
4026 dv->next = super->devlist;
4027 super->devlist = dv;
949c47a0 4028 }
cdddbdbc 4029
4d7b1503
DW		 4030 /* ensure that super->buf is large enough when all raid devices
4031 * are migrating
4032 */
4033 if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) {
4034 void *buf;
4035
f36a9ecd
PB		 4036 len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed,
4037 super->sector_size);
4038 if (posix_memalign(&buf, MAX_SECTOR_SIZE, len) != 0)
4d7b1503
DW		 4039 return 1;
4040
1f45a8ad
DW		 4041 memcpy(buf, super->buf, super->len);
4042 memset(buf + super->len, 0, len - super->len);
4d7b1503
DW		 4043 free(super->buf);
4044 super->buf = buf;
4045 super->len = len;
4046 }
ca9de185 4047
bbab0940
TM		 4048 super->extra_space += space_needed;
4049
cdddbdbc
DW		 4050 return 0;
4051}
4052
e2f41b2c
AK		 4053/*******************************************************************************
4054 * Function: check_mpb_migr_compatibility
4055 * Description: Function checks for unsupported migration features:
4056 * - migration optimization area (pba_of_lba0)
4057 * - descending reshape (ascending_migr)
4058 * Parameters:
4059 * super : imsm metadata information
4060 * Returns:
4061 * 0 : migration is compatible
4062 * -1 : migration is not compatible
4063 ******************************************************************************/
4064int check_mpb_migr_compatibility(struct intel_super *super)
4065{
4066 struct imsm_map *map0, *map1;
4067 struct migr_record *migr_rec = super->migr_rec;
4068 int i;
4069
4070 for (i = 0; i < super->anchor->num_raid_devs; i++) {
4071 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
4072
4073 if (dev_iter &&
4074 dev_iter->vol.migr_state == 1 &&
4075 dev_iter->vol.migr_type == MIGR_GEN_MIGR) {
4076 /* This device is migrating */
238c0a71
AK		 4077 map0 = get_imsm_map(dev_iter, MAP_0);
4078 map1 = get_imsm_map(dev_iter, MAP_1);
5551b113 4079 if (pba_of_lba0(map0) != pba_of_lba0(map1))
e2f41b2c
AK		 4080 /* migration optimization area was used */
4081 return -1;
4082 if (migr_rec->ascending_migr == 0
4083 && migr_rec->dest_depth_per_unit > 0)
4084 /* descending reshape not supported yet */
4085 return -1;
4086 }
4087 }
4088 return 0;
4089}
4090
d23fe947 4091static void __free_imsm(struct intel_super *super, int free_disks);
9ca2c81c 4092
cdddbdbc 4093/* load_imsm_mpb - read matrix metadata
f2f5c343 4094 * allocates super->mpb to be freed by free_imsm
cdddbdbc
DW		 4095 */
4096static int load_imsm_mpb(int fd, struct intel_super *super, char *devname)
4097{
4098 unsigned long long dsize;
cdddbdbc 4099 unsigned long long sectors;
f36a9ecd 4100 unsigned int sector_size = super->sector_size;
cdddbdbc 4101 struct stat;
6416d527 4102 struct imsm_super *anchor;
cdddbdbc
DW		 4103 __u32 check_sum;
4104
cdddbdbc 4105 get_dev_size(fd, NULL, &dsize);
f36a9ecd 4106 if (dsize < 2*sector_size) {
64436f06 4107 if (devname)
e7b84f9d
N		 4108 pr_err("%s: device to small for imsm\n",
4109 devname);
64436f06
N		 4110 return 1;
4111 }
cdddbdbc 4112
f36a9ecd 4113 if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0) {
cdddbdbc 4114 if (devname)
e7b84f9d
N		 4115 pr_err("Cannot seek to anchor block on %s: %s\n",
4116 devname, strerror(errno));
cdddbdbc
DW		 4117 return 1;
4118 }
4119
f36a9ecd 4120 if (posix_memalign((void **)&anchor, sector_size, sector_size) != 0) {
ad97895e 4121 if (devname)
7a862a02 4122 pr_err("Failed to allocate imsm anchor buffer on %s\n", devname);
ad97895e
DW		 4123 return 1;
4124 }
f36a9ecd 4125 if (read(fd, anchor, sector_size) != sector_size) {
cdddbdbc 4126 if (devname)
e7b84f9d
N		 4127 pr_err("Cannot read anchor block on %s: %s\n",
4128 devname, strerror(errno));
6416d527 4129 free(anchor);
cdddbdbc
DW		 4130 return 1;
4131 }
4132
6416d527 4133 if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) {
cdddbdbc 4134 if (devname)
e7b84f9d 4135 pr_err("no IMSM anchor on %s\n", devname);
6416d527 4136 free(anchor);
cdddbdbc
DW		 4137 return 2;
4138 }
4139
d23fe947 4140 __free_imsm(super, 0);
f2f5c343
LM		 4141 /* reload capability and hba */
4142
4143 /* capability and hba must be updated with new super allocation */
d424212e 4144 find_intel_hba_capability(fd, super, devname);
f36a9ecd
PB		 4145 super->len = ROUND_UP(anchor->mpb_size, sector_size);
4146 if (posix_memalign(&super->buf, MAX_SECTOR_SIZE, super->len) != 0) {
cdddbdbc 4147 if (devname)
e7b84f9d
N		 4148 pr_err("unable to allocate %zu byte mpb buffer\n",
4149 super->len);
6416d527 4150 free(anchor);
cdddbdbc
DW		 4151 return 2;
4152 }
f36a9ecd 4153 memcpy(super->buf, anchor, sector_size);
cdddbdbc 4154
f36a9ecd 4155 sectors = mpb_sectors(anchor, sector_size) - 1;
6416d527 4156 free(anchor);
8e59f3d8 4157
de44e46f
PB		 4158 if (posix_memalign(&super->migr_rec_buf, sector_size,
4159 MIGR_REC_BUF_SECTORS*sector_size) != 0) {
1ade5cc1 4160 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8
AK		 4161 free(super->buf);
4162 return 2;
4163 }
51d83f5d 4164 super->clean_migration_record_by_mdmon = 0;
8e59f3d8 4165
949c47a0 4166 if (!sectors) {
ecf45690
DW		 4167 check_sum = __gen_imsm_checksum(super->anchor);
4168 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
4169 if (devname)
e7b84f9d
N		 4170 pr_err("IMSM checksum %x != %x on %s\n",
4171 check_sum,
4172 __le32_to_cpu(super->anchor->check_sum),
4173 devname);
ecf45690
DW		 4174 return 2;
4175 }
4176
a2b97981 4177 return 0;
949c47a0 4178 }
cdddbdbc
DW		 4179
4180 /* read the extended mpb */
f36a9ecd 4181 if (lseek64(fd, dsize - (sector_size * (2 + sectors)), SEEK_SET) < 0) {
cdddbdbc 4182 if (devname)
e7b84f9d
N		 4183 pr_err("Cannot seek to extended mpb on %s: %s\n",
4184 devname, strerror(errno));
cdddbdbc
DW		 4185 return 1;
4186 }
4187
f36a9ecd
PB		 4188 if ((unsigned int)read(fd, super->buf + sector_size,
4189 super->len - sector_size) != super->len - sector_size) {
cdddbdbc 4190 if (devname)
e7b84f9d
N		 4191 pr_err("Cannot read extended mpb on %s: %s\n",
4192 devname, strerror(errno));
cdddbdbc
DW		 4193 return 2;
4194 }
4195
949c47a0
DW		 4196 check_sum = __gen_imsm_checksum(super->anchor);
4197 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
cdddbdbc 4198 if (devname)
e7b84f9d
N		 4199 pr_err("IMSM checksum %x != %x on %s\n",
4200 check_sum, __le32_to_cpu(super->anchor->check_sum),
4201 devname);
db575f3b 4202 return 3;
cdddbdbc
DW		 4203 }
4204
a2b97981
DW		 4205 return 0;
4206}
4207
8e59f3d8
AK		 4208static int read_imsm_migr_rec(int fd, struct intel_super *super);
4209
97f81ee2
CA		 4210/* clears hi bits in metadata if MPB_ATTRIB_2TB_DISK not set */
4211static void clear_hi(struct intel_super *super)
4212{
4213 struct imsm_super *mpb = super->anchor;
4214 int i, n;
4215 if (mpb->attributes & MPB_ATTRIB_2TB_DISK)
4216 return;
4217 for (i = 0; i < mpb->num_disks; ++i) {
4218 struct imsm_disk *disk = &mpb->disk[i];
4219 disk->total_blocks_hi = 0;
4220 }
4221 for (i = 0; i < mpb->num_raid_devs; ++i) {
4222 struct imsm_dev *dev = get_imsm_dev(super, i);
4223 if (!dev)
4224 return;
4225 for (n = 0; n < 2; ++n) {
4226 struct imsm_map *map = get_imsm_map(dev, n);
4227 if (!map)
4228 continue;
4229 map->pba_of_lba0_hi = 0;
4230 map->blocks_per_member_hi = 0;
4231 map->num_data_stripes_hi = 0;
4232 }
4233 }
4234}
4235
a2b97981
DW		 4236static int
4237load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd)
4238{
4239 int err;
4240
4241 err = load_imsm_mpb(fd, super, devname);
4242 if (err)
4243 return err;
f36a9ecd
PB		 4244 if (super->sector_size == 4096)
4245 convert_from_4k(super);
a2b97981
DW		 4246 err = load_imsm_disk(fd, super, devname, keep_fd);
4247 if (err)
4248 return err;
4249 err = parse_raid_devices(super);
8d67477f
TM		 4250 if (err)
4251 return err;
4252 err = load_bbm_log(super);
97f81ee2 4253 clear_hi(super);
a2b97981 4254 return err;
cdddbdbc
DW		 4255}
4256
ae6aad82
DW		 4257static void __free_imsm_disk(struct dl *d)
4258{
4259 if (d->fd >= 0)
4260 close(d->fd);
4261 if (d->devname)
4262 free(d->devname);
0dcecb2e
DW		 4263 if (d->e)
4264 free(d->e);
ae6aad82
DW		 4265 free(d);
4266
4267}
1a64be56 4268
cdddbdbc
DW		 4269static void free_imsm_disks(struct intel_super *super)
4270{
47ee5a45 4271 struct dl *d;
cdddbdbc 4272
47ee5a45
DW		 4273 while (super->disks) {
4274 d = super->disks;
cdddbdbc 4275 super->disks = d->next;
ae6aad82 4276 __free_imsm_disk(d);
cdddbdbc 4277 }
cb82edca
AK		 4278 while (super->disk_mgmt_list) {
4279 d = super->disk_mgmt_list;
4280 super->disk_mgmt_list = d->next;
4281 __free_imsm_disk(d);
4282 }
47ee5a45
DW		 4283 while (super->missing) {
4284 d = super->missing;
4285 super->missing = d->next;
4286 __free_imsm_disk(d);
4287 }
4288
cdddbdbc
DW		 4289}
4290
9ca2c81c 4291/* free all the pieces hanging off of a super pointer */
d23fe947 4292static void __free_imsm(struct intel_super *super, int free_disks)
cdddbdbc 4293{
88654014
LM		 4294 struct intel_hba *elem, *next;
4295
9ca2c81c 4296 if (super->buf) {
949c47a0 4297 free(super->buf);
9ca2c81c
DW		 4298 super->buf = NULL;
4299 }
f2f5c343
LM		 4300 /* unlink capability description */
4301 super->orom = NULL;
8e59f3d8
AK		 4302 if (super->migr_rec_buf) {
4303 free(super->migr_rec_buf);
4304 super->migr_rec_buf = NULL;
4305 }
d23fe947
DW		 4306 if (free_disks)
4307 free_imsm_disks(super);
ba2de7ba 4308 free_devlist(super);
88654014
LM		 4309 elem = super->hba;
4310 while (elem) {
4311 if (elem->path)
4312 free((void *)elem->path);
4313 next = elem->next;
4314 free(elem);
4315 elem = next;
88c32bb1 4316 }
8d67477f
TM		 4317 if (super->bbm_log)
4318 free(super->bbm_log);
88654014 4319 super->hba = NULL;
cdddbdbc
DW		 4320}
4321
9ca2c81c
DW		 4322static void free_imsm(struct intel_super *super)
4323{
d23fe947 4324 __free_imsm(super, 1);
928f1424 4325 free(super->bb.entries);
9ca2c81c
DW		 4326 free(super);
4327}
cdddbdbc
DW		 4328
4329static void free_super_imsm(struct supertype *st)
4330{
4331 struct intel_super *super = st->sb;
4332
4333 if (!super)
4334 return;
4335
4336 free_imsm(super);
4337 st->sb = NULL;
4338}
4339
49133e57 4340static struct intel_super *alloc_super(void)
c2c087e6 4341{
503975b9 4342 struct intel_super *super = xcalloc(1, sizeof(*super));
c2c087e6 4343
503975b9
N		 4344 super->current_vol = -1;
4345 super->create_offset = ~((unsigned long long) 0);
928f1424
TM		 4346
4347 super->bb.entries = xmalloc(BBM_LOG_MAX_ENTRIES *
4348 sizeof(struct md_bb_entry));
4349 if (!super->bb.entries) {
4350 free(super);
4351 return NULL;
4352 }
4353
c2c087e6
DW		 4354 return super;
4355}
4356
f0f5a016
LM		 4357/*
4358 * find and allocate hba and OROM/EFI based on valid fd of RAID component device
4359 */
d424212e 4360static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname)
f0f5a016
LM		 4361{
4362 struct sys_dev *hba_name;
4363 int rv = 0;
4364
089f9d79 4365 if (fd < 0 || check_env("IMSM_NO_PLATFORM")) {
f2f5c343 4366 super->orom = NULL;
f0f5a016
LM		 4367 super->hba = NULL;
4368 return 0;
4369 }
4370 hba_name = find_disk_attached_hba(fd, NULL);
4371 if (!hba_name) {
d424212e 4372 if (devname)
e7b84f9d
N		 4373 pr_err("%s is not attached to Intel(R) RAID controller.\n",
4374 devname);
f0f5a016
LM		 4375 return 1;
4376 }
4377 rv = attach_hba_to_super(super, hba_name);
4378 if (rv == 2) {
d424212e
N		 4379 if (devname) {
4380 struct intel_hba *hba = super->hba;
f0f5a016 4381
60f0f54d
PB		 4382 pr_err("%s is attached to Intel(R) %s %s (%s),\n"
4383 " but the container is assigned to Intel(R) %s %s (",
d424212e 4384 devname,
614902f6 4385 get_sys_dev_type(hba_name->type),
60f0f54d 4386 hba_name->type == SYS_DEV_VMD ? "domain" : "RAID controller",
f0f5a016 4387 hba_name->pci_id ? : "Err!",
60f0f54d
PB		 4388 get_sys_dev_type(super->hba->type),
4389 hba->type == SYS_DEV_VMD ? "domain" : "RAID controller");
f0f5a016 4390
f0f5a016
LM		 4391 while (hba) {
4392 fprintf(stderr, "%s", hba->pci_id ? : "Err!");
4393 if (hba->next)
4394 fprintf(stderr, ", ");
4395 hba = hba->next;
4396 }
6b781d33 4397 fprintf(stderr, ").\n"
60f0f54d
PB		 4398 " Mixing devices attached to different %s is not allowed.\n",
4399 hba_name->type == SYS_DEV_VMD ? "VMD domains" : "controllers");
f0f5a016 4400 }
f0f5a016
LM		 4401 return 2;
4402 }
6b781d33 4403 super->orom = find_imsm_capability(hba_name);
f2f5c343
LM		 4404 if (!super->orom)
4405 return 3;
614902f6 4406
f0f5a016
LM		 4407 return 0;
4408}
4409
47ee5a45
DW		 4410/* find_missing - helper routine for load_super_imsm_all that identifies
4411 * disks that have disappeared from the system. This routine relies on
4412 * the mpb being uptodate, which it is at load time.
4413 */
4414static int find_missing(struct intel_super *super)
4415{
4416 int i;
4417 struct imsm_super *mpb = super->anchor;
4418 struct dl *dl;
4419 struct imsm_disk *disk;
47ee5a45
DW		 4420
4421 for (i = 0; i < mpb->num_disks; i++) {
4422 disk = __get_imsm_disk(mpb, i);
54c2c1ea 4423 dl = serial_to_dl(disk->serial, super);
47ee5a45
DW		 4424 if (dl)
4425 continue;
47ee5a45 4426
503975b9 4427 dl = xmalloc(sizeof(*dl));
47ee5a45
DW		 4428 dl->major = 0;
4429 dl->minor = 0;
4430 dl->fd = -1;
503975b9 4431 dl->devname = xstrdup("missing");
47ee5a45
DW		 4432 dl->index = i;
4433 serialcpy(dl->serial, disk->serial);
4434 dl->disk = *disk;
689c9bf3 4435 dl->e = NULL;
47ee5a45
DW		 4436 dl->next = super->missing;
4437 super->missing = dl;
4438 }
4439
4440 return 0;
4441}
4442
3960e579 4443#ifndef MDASSEMBLE
a2b97981
DW		 4444static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list)
4445{
4446 struct intel_disk *idisk = disk_list;
4447
4448 while (idisk) {
4449 if (serialcmp(idisk->disk.serial, serial) == 0)
4450 break;
4451 idisk = idisk->next;
4452 }
4453
4454 return idisk;
4455}
4456
4457static int __prep_thunderdome(struct intel_super **table, int tbl_size,
4458 struct intel_super *super,
4459 struct intel_disk **disk_list)
4460{
4461 struct imsm_disk *d = &super->disks->disk;
4462 struct imsm_super *mpb = super->anchor;
4463 int i, j;
4464
4465 for (i = 0; i < tbl_size; i++) {
4466 struct imsm_super *tbl_mpb = table[i]->anchor;
4467 struct imsm_disk *tbl_d = &table[i]->disks->disk;
4468
4469 if (tbl_mpb->family_num == mpb->family_num) {
4470 if (tbl_mpb->check_sum == mpb->check_sum) {
1ade5cc1
N		 4471 dprintf("mpb from %d:%d matches %d:%d\n",
4472 super->disks->major,
a2b97981
DW		 4473 super->disks->minor,
4474 table[i]->disks->major,
4475 table[i]->disks->minor);
4476 break;
4477 }
4478
4479 if (((is_configured(d) && !is_configured(tbl_d)) ||
4480 is_configured(d) == is_configured(tbl_d)) &&
4481 tbl_mpb->generation_num < mpb->generation_num) {
4482 /* current version of the mpb is a
4483 * better candidate than the one in
4484 * super_table, but copy over "cross
4485 * generational" status
4486 */
4487 struct intel_disk *idisk;
4488
1ade5cc1
N		 4489 dprintf("mpb from %d:%d replaces %d:%d\n",
4490 super->disks->major,
a2b97981
DW		 4491 super->disks->minor,
4492 table[i]->disks->major,
4493 table[i]->disks->minor);
4494
4495 idisk = disk_list_get(tbl_d->serial, *disk_list);
4496 if (idisk && is_failed(&idisk->disk))
4497 tbl_d->status |= FAILED_DISK;
4498 break;
4499 } else {
4500 struct intel_disk *idisk;
4501 struct imsm_disk *disk;
4502
4503 /* tbl_mpb is more up to date, but copy
4504 * over cross generational status before
4505 * returning
4506 */
4507 disk = __serial_to_disk(d->serial, mpb, NULL);
4508 if (disk && is_failed(disk))
4509 d->status |= FAILED_DISK;
4510
4511 idisk = disk_list_get(d->serial, *disk_list);
4512 if (idisk) {
4513 idisk->owner = i;
4514 if (disk && is_configured(disk))
4515 idisk->disk.status |= CONFIGURED_DISK;
4516 }
4517
1ade5cc1
N		 4518 dprintf("mpb from %d:%d prefer %d:%d\n",
4519 super->disks->major,
a2b97981
DW		 4520 super->disks->minor,
4521 table[i]->disks->major,
4522 table[i]->disks->minor);
4523
4524 return tbl_size;
4525 }
4526 }
4527 }
4528
4529 if (i >= tbl_size)
4530 table[tbl_size++] = super;
4531 else
4532 table[i] = super;
4533
4534 /* update/extend the merged list of imsm_disk records */
4535 for (j = 0; j < mpb->num_disks; j++) {
4536 struct imsm_disk *disk = __get_imsm_disk(mpb, j);
4537 struct intel_disk *idisk;
4538
4539 idisk = disk_list_get(disk->serial, *disk_list);
4540 if (idisk) {
4541 idisk->disk.status |= disk->status;
4542 if (is_configured(&idisk->disk) ||
4543 is_failed(&idisk->disk))
4544 idisk->disk.status &= ~(SPARE_DISK);
4545 } else {
503975b9 4546 idisk = xcalloc(1, sizeof(*idisk));
a2b97981
DW		 4547 idisk->owner = IMSM_UNKNOWN_OWNER;
4548 idisk->disk = *disk;
4549 idisk->next = *disk_list;
4550 *disk_list = idisk;
4551 }
4552
4553 if (serialcmp(idisk->disk.serial, d->serial) == 0)
4554 idisk->owner = i;
4555 }
4556
4557 return tbl_size;
4558}
4559
4560static struct intel_super *
4561validate_members(struct intel_super *super, struct intel_disk *disk_list,
4562 const int owner)
4563{
4564 struct imsm_super *mpb = super->anchor;
4565 int ok_count = 0;
4566 int i;
4567
4568 for (i = 0; i < mpb->num_disks; i++) {
4569 struct imsm_disk *disk = __get_imsm_disk(mpb, i);
4570 struct intel_disk *idisk;
4571
4572 idisk = disk_list_get(disk->serial, disk_list);
4573 if (idisk) {
4574 if (idisk->owner == owner ||
4575 idisk->owner == IMSM_UNKNOWN_OWNER)
4576 ok_count++;
4577 else
1ade5cc1
N		 4578 dprintf("'%.16s' owner %d != %d\n",
4579 disk->serial, idisk->owner,
a2b97981
DW		 4580 owner);
4581 } else {
1ade5cc1
N		 4582 dprintf("unknown disk %x [%d]: %.16s\n",
4583 __le32_to_cpu(mpb->family_num), i,
a2b97981
DW		 4584 disk->serial);
4585 break;
4586 }
4587 }
4588
4589 if (ok_count == mpb->num_disks)
4590 return super;
4591 return NULL;
4592}
4593
4594static void show_conflicts(__u32 family_num, struct intel_super *super_list)
4595{
4596 struct intel_super *s;
4597
4598 for (s = super_list; s; s = s->next) {
4599 if (family_num != s->anchor->family_num)
4600 continue;
e12b3daa 4601 pr_err("Conflict, offlining family %#x on '%s'\n",
a2b97981
DW		 4602 __le32_to_cpu(family_num), s->disks->devname);
4603 }
4604}
4605
4606static struct intel_super *
4607imsm_thunderdome(struct intel_super **super_list, int len)
4608{
4609 struct intel_super *super_table[len];
4610 struct intel_disk *disk_list = NULL;
4611 struct intel_super *champion, *spare;
4612 struct intel_super *s, **del;
4613 int tbl_size = 0;
4614 int conflict;
4615 int i;
4616
4617 memset(super_table, 0, sizeof(super_table));
4618 for (s = *super_list; s; s = s->next)
4619 tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list);
4620
4621 for (i = 0; i < tbl_size; i++) {
4622 struct imsm_disk *d;
4623 struct intel_disk *idisk;
4624 struct imsm_super *mpb = super_table[i]->anchor;
4625
4626 s = super_table[i];
4627 d = &s->disks->disk;
4628
4629 /* 'd' must appear in merged disk list for its
4630 * configuration to be valid
4631 */
4632 idisk = disk_list_get(d->serial, disk_list);
4633 if (idisk && idisk->owner == i)
4634 s = validate_members(s, disk_list, i);
4635 else
4636 s = NULL;
4637
4638 if (!s)
1ade5cc1
N		 4639 dprintf("marking family: %#x from %d:%d offline\n",
4640 mpb->family_num,
a2b97981
DW		 4641 super_table[i]->disks->major,
4642 super_table[i]->disks->minor);
4643 super_table[i] = s;
4644 }
4645
4646 /* This is where the mdadm implementation differs from the Windows
4647 * driver which has no strict concept of a container. We can only
4648 * assemble one family from a container, so when returning a prodigal
4649 * array member to this system the code will not be able to disambiguate
4650 * the container contents that should be assembled ("foreign" versus
4651 * "local"). It requires user intervention to set the orig_family_num
4652 * to a new value to establish a new container. The Windows driver in
4653 * this situation fixes up the volume name in place and manages the
4654 * foreign array as an independent entity.
4655 */
4656 s = NULL;
4657 spare = NULL;
4658 conflict = 0;
4659 for (i = 0; i < tbl_size; i++) {
4660 struct intel_super *tbl_ent = super_table[i];
4661 int is_spare = 0;
4662
4663 if (!tbl_ent)
4664 continue;
4665
4666 if (tbl_ent->anchor->num_raid_devs == 0) {
4667 spare = tbl_ent;
4668 is_spare = 1;
4669 }
4670
4671 if (s && !is_spare) {
4672 show_conflicts(tbl_ent->anchor->family_num, *super_list);
4673 conflict++;
4674 } else if (!s && !is_spare)
4675 s = tbl_ent;
4676 }
4677
4678 if (!s)
4679 s = spare;
4680 if (!s) {
4681 champion = NULL;
4682 goto out;
4683 }
4684 champion = s;
4685
4686 if (conflict)
7a862a02 4687 pr_err("Chose family %#x on '%s', assemble conflicts to new container with '--update=uuid'\n",
a2b97981
DW		 4688 __le32_to_cpu(s->anchor->family_num), s->disks->devname);
4689
4690 /* collect all dl's onto 'champion', and update them to
4691 * champion's version of the status
4692 */
4693 for (s = *super_list; s; s = s->next) {
4694 struct imsm_super *mpb = champion->anchor;
4695 struct dl *dl = s->disks;
4696
4697 if (s == champion)
4698 continue;
4699
5d7b407a
CA		 4700 mpb->attributes |= s->anchor->attributes & MPB_ATTRIB_2TB_DISK;
4701
a2b97981
DW		 4702 for (i = 0; i < mpb->num_disks; i++) {
4703 struct imsm_disk *disk;
4704
4705 disk = __serial_to_disk(dl->serial, mpb, &dl->index);
4706 if (disk) {
4707 dl->disk = *disk;
4708 /* only set index on disks that are a member of
4709 * a populated contianer, i.e. one with
4710 * raid_devs
4711 */
4712 if (is_failed(&dl->disk))
4713 dl->index = -2;
4714 else if (is_spare(&dl->disk))
4715 dl->index = -1;
4716 break;
4717 }
4718 }
4719
4720 if (i >= mpb->num_disks) {
4721 struct intel_disk *idisk;
4722
4723 idisk = disk_list_get(dl->serial, disk_list);
ecf408e9 4724 if (idisk && is_spare(&idisk->disk) &&
a2b97981
DW		 4725 !is_failed(&idisk->disk) && !is_configured(&idisk->disk))
4726 dl->index = -1;
4727 else {
4728 dl->index = -2;
4729 continue;
4730 }
4731 }
4732
4733 dl->next = champion->disks;
4734 champion->disks = dl;
4735 s->disks = NULL;
4736 }
4737
4738 /* delete 'champion' from super_list */
4739 for (del = super_list; *del; ) {
4740 if (*del == champion) {
4741 *del = (*del)->next;
4742 break;
4743 } else
4744 del = &(*del)->next;
4745 }
4746 champion->next = NULL;
4747
4748 out:
4749 while (disk_list) {
4750 struct intel_disk *idisk = disk_list;
4751
4752 disk_list = disk_list->next;
4753 free(idisk);
4754 }
4755
4756 return champion;
4757}
4758
9587c373
LM		 4759static int
4760get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd);
4dd2df09 4761static int get_super_block(struct intel_super **super_list, char *devnm, char *devname,
9587c373 4762 int major, int minor, int keep_fd);
ec50f7b6
LM		 4763static int
4764get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list,
4765 int *max, int keep_fd);
4766
cdddbdbc 4767static int load_super_imsm_all(struct supertype *st, int fd, void **sbp,
ec50f7b6
LM		 4768 char *devname, struct md_list *devlist,
4769 int keep_fd)
cdddbdbc 4770{
a2b97981
DW		 4771 struct intel_super *super_list = NULL;
4772 struct intel_super *super = NULL;
a2b97981 4773 int err = 0;
9587c373 4774 int i = 0;
dab4a513 4775
9587c373
LM		 4776 if (fd >= 0)
4777 /* 'fd' is an opened container */
4778 err = get_sra_super_block(fd, &super_list, devname, &i, keep_fd);
4779 else
ec50f7b6
LM		 4780 /* get super block from devlist devices */
4781 err = get_devlist_super_block(devlist, &super_list, &i, keep_fd);
9587c373 4782 if (err)
1602d52c 4783 goto error;
a2b97981
DW		 4784 /* all mpbs enter, maybe one leaves */
4785 super = imsm_thunderdome(&super_list, i);
4786 if (!super) {
4787 err = 1;
4788 goto error;
cdddbdbc
DW		 4789 }
4790
47ee5a45
DW		 4791 if (find_missing(super) != 0) {
4792 free_imsm(super);
a2b97981
DW		 4793 err = 2;
4794 goto error;
47ee5a45 4795 }
8e59f3d8
AK		 4796
4797 /* load migration record */
4798 err = load_imsm_migr_rec(super, NULL);
4c965cc9
AK		 4799 if (err == -1) {
4800 /* migration is in progress,
4801 * but migr_rec cannot be loaded,
4802 */
8e59f3d8
AK		 4803 err = 4;
4804 goto error;
4805 }
e2f41b2c
AK		 4806
4807 /* Check migration compatibility */
089f9d79 4808 if (err == 0 && check_mpb_migr_compatibility(super) != 0) {
e7b84f9d 4809 pr_err("Unsupported migration detected");
e2f41b2c
AK		 4810 if (devname)
4811 fprintf(stderr, " on %s\n", devname);
4812 else
4813 fprintf(stderr, " (IMSM).\n");
4814
4815 err = 5;
4816 goto error;
4817 }
4818
a2b97981
DW		 4819 err = 0;
4820
4821 error:
4822 while (super_list) {
4823 struct intel_super *s = super_list;
4824
4825 super_list = super_list->next;
4826 free_imsm(s);
4827 }
9587c373 4828
a2b97981
DW		 4829 if (err)
4830 return err;
f7e7067b 4831
cdddbdbc 4832 *sbp = super;
9587c373 4833 if (fd >= 0)
4dd2df09 4834 strcpy(st->container_devnm, fd2devnm(fd));
9587c373 4835 else
4dd2df09 4836 st->container_devnm[0] = 0;
a2b97981 4837 if (err == 0 && st->ss == NULL) {
bf5a934a 4838 st->ss = &super_imsm;
cdddbdbc
DW		 4839 st->minor_version = 0;
4840 st->max_devs = IMSM_MAX_DEVICES;
4841 }
cdddbdbc
DW		 4842 return 0;
4843}
2b959fbf 4844
ec50f7b6
LM		 4845static int
4846get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list,
4847 int *max, int keep_fd)
4848{
4849 struct md_list *tmpdev;
4850 int err = 0;
4851 int i = 0;
9587c373 4852
ec50f7b6
LM		 4853 for (i = 0, tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
4854 if (tmpdev->used != 1)
4855 continue;
4856 if (tmpdev->container == 1) {
ca9de185 4857 int lmax = 0;
ec50f7b6
LM		 4858 int fd = dev_open(tmpdev->devname, O_RDONLY|O_EXCL);
4859 if (fd < 0) {
e7b84f9d 4860 pr_err("cannot open device %s: %s\n",
ec50f7b6
LM		 4861 tmpdev->devname, strerror(errno));
4862 err = 8;
4863 goto error;
4864 }
4865 err = get_sra_super_block(fd, super_list,
4866 tmpdev->devname, &lmax,
4867 keep_fd);
4868 i += lmax;
4869 close(fd);
4870 if (err) {
4871 err = 7;
4872 goto error;
4873 }
4874 } else {
4875 int major = major(tmpdev->st_rdev);
4876 int minor = minor(tmpdev->st_rdev);
4877 err = get_super_block(super_list,
4dd2df09 4878 NULL,
ec50f7b6
LM		 4879 tmpdev->devname,
4880 major, minor,
4881 keep_fd);
4882 i++;
4883 if (err) {
4884 err = 6;
4885 goto error;
4886 }
4887 }
4888 }
4889 error:
4890 *max = i;
4891 return err;
4892}
9587c373 4893
4dd2df09 4894static int get_super_block(struct intel_super **super_list, char *devnm, char *devname,
9587c373
LM		 4895 int major, int minor, int keep_fd)
4896{
594dc1b8 4897 struct intel_super *s;
9587c373
LM		 4898 char nm[32];
4899 int dfd = -1;
9587c373
LM		 4900 int err = 0;
4901 int retry;
4902
4903 s = alloc_super();
4904 if (!s) {
4905 err = 1;
4906 goto error;
4907 }
4908
4909 sprintf(nm, "%d:%d", major, minor);
4910 dfd = dev_open(nm, O_RDWR);
4911 if (dfd < 0) {
4912 err = 2;
4913 goto error;
4914 }
4915
fa7bb6f8 4916 get_dev_sector_size(dfd, NULL, &s->sector_size);
cb8f6859 4917 find_intel_hba_capability(dfd, s, devname);
9587c373
LM		 4918 err = load_and_parse_mpb(dfd, s, NULL, keep_fd);
4919
4920 /* retry the load if we might have raced against mdmon */
4dd2df09 4921 if (err == 3 && devnm && mdmon_running(devnm))
9587c373
LM		 4922 for (retry = 0; retry < 3; retry++) {
4923 usleep(3000);
4924 err = load_and_parse_mpb(dfd, s, NULL, keep_fd);
4925 if (err != 3)
4926 break;
4927 }
4928 error:
4929 if (!err) {
4930 s->next = *super_list;
4931 *super_list = s;
4932 } else {
4933 if (s)
8d67477f 4934 free_imsm(s);
36614e95 4935 if (dfd >= 0)
9587c373
LM		 4936 close(dfd);
4937 }
089f9d79 4938 if (dfd >= 0 && !keep_fd)
9587c373
LM		 4939 close(dfd);
4940 return err;
4941
4942}
4943
4944static int
4945get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd)
4946{
4947 struct mdinfo *sra;
4dd2df09 4948 char *devnm;
9587c373
LM		 4949 struct mdinfo *sd;
4950 int err = 0;
4951 int i = 0;
4dd2df09 4952 sra = sysfs_read(fd, NULL, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE);
9587c373
LM		 4953 if (!sra)
4954 return 1;
4955
4956 if (sra->array.major_version != -1 ||
4957 sra->array.minor_version != -2 ||
4958 strcmp(sra->text_version, "imsm") != 0) {
4959 err = 1;
4960 goto error;
4961 }
4962 /* load all mpbs */
4dd2df09 4963 devnm = fd2devnm(fd);
9587c373 4964 for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) {
4dd2df09 4965 if (get_super_block(super_list, devnm, devname,
9587c373
LM		 4966 sd->disk.major, sd->disk.minor, keep_fd) != 0) {
4967 err = 7;
4968 goto error;
4969 }
4970 }
4971 error:
4972 sysfs_free(sra);
4973 *max = i;
4974 return err;
4975}
4976
2b959fbf
N		 4977static int load_container_imsm(struct supertype *st, int fd, char *devname)
4978{
ec50f7b6 4979 return load_super_imsm_all(st, fd, &st->sb, devname, NULL, 1);
2b959fbf 4980}
cdddbdbc
DW		 4981#endif
4982
4983static int load_super_imsm(struct supertype *st, int fd, char *devname)
4984{
4985 struct intel_super *super;
4986 int rv;
8a3544f8 4987 int retry;
cdddbdbc 4988
357ac106 4989 if (test_partition(fd))
691c6ee1
N		 4990 /* IMSM not allowed on partitions */
4991 return 1;
4992
37424f13
DW		 4993 free_super_imsm(st);
4994
49133e57 4995 super = alloc_super();
fa7bb6f8 4996 get_dev_sector_size(fd, NULL, &super->sector_size);
8d67477f
TM		 4997 if (!super)
4998 return 1;
ea2bc72b
LM		 4999 /* Load hba and capabilities if they exist.
5000 * But do not preclude loading metadata in case capabilities or hba are
5001 * non-compliant and ignore_hw_compat is set.
5002 */
d424212e 5003 rv = find_intel_hba_capability(fd, super, devname);
f2f5c343 5004 /* no orom/efi or non-intel hba of the disk */
089f9d79 5005 if (rv != 0 && st->ignore_hw_compat == 0) {
f2f5c343 5006 if (devname)
e7b84f9d 5007 pr_err("No OROM/EFI properties for %s\n", devname);
f2f5c343
LM		 5008 free_imsm(super);
5009 return 2;
5010 }
a2b97981 5011 rv = load_and_parse_mpb(fd, super, devname, 0);
cdddbdbc 5012
8a3544f8
AP		 5013 /* retry the load if we might have raced against mdmon */
5014 if (rv == 3) {
f96b1302
AP		 5015 struct mdstat_ent *mdstat = NULL;
5016 char *name = fd2kname(fd);
5017
5018 if (name)
5019 mdstat = mdstat_by_component(name);
8a3544f8
AP		 5020
5021 if (mdstat && mdmon_running(mdstat->devnm) && getpid() != mdmon_pid(mdstat->devnm)) {
5022 for (retry = 0; retry < 3; retry++) {
5023 usleep(3000);
5024 rv = load_and_parse_mpb(fd, super, devname, 0);
5025 if (rv != 3)
5026 break;
5027 }
5028 }
5029
5030 free_mdstat(mdstat);
5031 }
5032
cdddbdbc
DW		 5033 if (rv) {
5034 if (devname)
7a862a02 5035 pr_err("Failed to load all information sections on %s\n", devname);
cdddbdbc
DW		 5036 free_imsm(super);
5037 return rv;
5038 }
5039
5040 st->sb = super;
5041 if (st->ss == NULL) {
5042 st->ss = &super_imsm;
5043 st->minor_version = 0;
5044 st->max_devs = IMSM_MAX_DEVICES;
5045 }
8e59f3d8
AK		 5046
5047 /* load migration record */
2e062e82
AK		 5048 if (load_imsm_migr_rec(super, NULL) == 0) {
5049 /* Check for unsupported migration features */
5050 if (check_mpb_migr_compatibility(super) != 0) {
e7b84f9d 5051 pr_err("Unsupported migration detected");
2e062e82
AK		 5052 if (devname)
5053 fprintf(stderr, " on %s\n", devname);
5054 else
5055 fprintf(stderr, " (IMSM).\n");
5056 return 3;
5057 }
e2f41b2c
AK		 5058 }
5059
cdddbdbc
DW		 5060 return 0;
5061}
5062
ef6ffade
DW		 5063static __u16 info_to_blocks_per_strip(mdu_array_info_t *info)
5064{
5065 if (info->level == 1)
5066 return 128;
5067 return info->chunk_size >> 9;
5068}
5069
5551b113
CA		 5070static unsigned long long info_to_blocks_per_member(mdu_array_info_t *info,
5071 unsigned long long size)
fcfd9599 5072{
4025c288 5073 if (info->level == 1)
5551b113 5074 return size * 2;
4025c288 5075 else
5551b113 5076 return (size * 2) & ~(info_to_blocks_per_strip(info) - 1);
fcfd9599
DW		 5077}
5078
4d1313e9
DW		 5079static void imsm_update_version_info(struct intel_super *super)
5080{
5081 /* update the version and attributes */
5082 struct imsm_super *mpb = super->anchor;
5083 char *version;
5084 struct imsm_dev *dev;
5085 struct imsm_map *map;
5086 int i;
5087
5088 for (i = 0; i < mpb->num_raid_devs; i++) {
5089 dev = get_imsm_dev(super, i);
238c0a71 5090 map = get_imsm_map(dev, MAP_0);
4d1313e9
DW		 5091 if (__le32_to_cpu(dev->size_high) > 0)
5092 mpb->attributes |= MPB_ATTRIB_2TB;
5093
5094 /* FIXME detect when an array spans a port multiplier */
5095 #if 0
5096 mpb->attributes |= MPB_ATTRIB_PM;
5097 #endif
5098
5099 if (mpb->num_raid_devs > 1 ||
5100 mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) {
5101 version = MPB_VERSION_ATTRIBS;
5102 switch (get_imsm_raid_level(map)) {
5103 case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break;
5104 case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break;
5105 case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break;
5106 case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break;
5107 }
5108 } else {
5109 if (map->num_members >= 5)
5110 version = MPB_VERSION_5OR6_DISK_ARRAY;
5111 else if (dev->status == DEV_CLONE_N_GO)
5112 version = MPB_VERSION_CNG;
5113 else if (get_imsm_raid_level(map) == 5)
5114 version = MPB_VERSION_RAID5;
5115 else if (map->num_members >= 3)
5116 version = MPB_VERSION_3OR4_DISK_ARRAY;
5117 else if (get_imsm_raid_level(map) == 1)
5118 version = MPB_VERSION_RAID1;
5119 else
5120 version = MPB_VERSION_RAID0;
5121 }
5122 strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version);
5123 }
5124}
5125
aa534678
DW		 5126static int check_name(struct intel_super *super, char *name, int quiet)
5127{
5128 struct imsm_super *mpb = super->anchor;
5129 char *reason = NULL;
5130 int i;
5131
5132 if (strlen(name) > MAX_RAID_SERIAL_LEN)
5133 reason = "must be 16 characters or less";
5134
5135 for (i = 0; i < mpb->num_raid_devs; i++) {
5136 struct imsm_dev *dev = get_imsm_dev(super, i);
5137
5138 if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) {
5139 reason = "already exists";
5140 break;
5141 }
5142 }
5143
5144 if (reason && !quiet)
e7b84f9d 5145 pr_err("imsm volume name %s\n", reason);
aa534678
DW		 5146
5147 return !reason;
5148}
5149
8b353278
DW		 5150static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info,
5151 unsigned long long size, char *name,
83cd1e97
N		 5152 char *homehost, int *uuid,
5153 long long data_offset)
cdddbdbc 5154{
c2c087e6
DW		 5155 /* We are creating a volume inside a pre-existing container.
5156 * so st->sb is already set.
5157 */
5158 struct intel_super *super = st->sb;
f36a9ecd 5159 unsigned int sector_size = super->sector_size;
949c47a0 5160 struct imsm_super *mpb = super->anchor;
ba2de7ba 5161 struct intel_dev *dv;
c2c087e6
DW		 5162 struct imsm_dev *dev;
5163 struct imsm_vol *vol;
5164 struct imsm_map *map;
5165 int idx = mpb->num_raid_devs;
5166 int i;
5167 unsigned long long array_blocks;
2c092cad 5168 size_t size_old, size_new;
5551b113 5169 unsigned long long num_data_stripes;
cdddbdbc 5170
88c32bb1 5171 if (super->orom && mpb->num_raid_devs >= super->orom->vpa) {
7a862a02 5172 pr_err("This imsm-container already has the maximum of %d volumes\n", super->orom->vpa);
c2c087e6
DW		 5173 return 0;
5174 }
5175
2c092cad
DW		 5176 /* ensure the mpb is large enough for the new data */
5177 size_old = __le32_to_cpu(mpb->mpb_size);
5178 size_new = disks_to_mpb_size(info->nr_disks);
5179 if (size_new > size_old) {
5180 void *mpb_new;
f36a9ecd 5181 size_t size_round = ROUND_UP(size_new, sector_size);
2c092cad 5182
f36a9ecd 5183 if (posix_memalign(&mpb_new, sector_size, size_round) != 0) {
e7b84f9d 5184 pr_err("could not allocate new mpb\n");
2c092cad
DW		 5185 return 0;
5186 }
de44e46f
PB		 5187 if (posix_memalign(&super->migr_rec_buf, sector_size,
5188 MIGR_REC_BUF_SECTORS*sector_size) != 0) {
1ade5cc1 5189 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8
AK		 5190 free(super->buf);
5191 free(super);
ea944c8f 5192 free(mpb_new);
8e59f3d8
AK		 5193 return 0;
5194 }
2c092cad
DW		 5195 memcpy(mpb_new, mpb, size_old);
5196 free(mpb);
5197 mpb = mpb_new;
949c47a0 5198 super->anchor = mpb_new;
2c092cad
DW		 5199 mpb->mpb_size = __cpu_to_le32(size_new);
5200 memset(mpb_new + size_old, 0, size_round - size_old);
bbab0940 5201 super->len = size_round;
2c092cad 5202 }
bf5a934a 5203 super->current_vol = idx;
3960e579
DW		 5204
5205 /* handle 'failed_disks' by either:
5206 * a) create dummy disk entries in the table if this the first
5207 * volume in the array. We add them here as this is the only
5208 * opportunity to add them. add_to_super_imsm_volume()
5209 * handles the non-failed disks and continues incrementing
5210 * mpb->num_disks.
5211 * b) validate that 'failed_disks' matches the current number
5212 * of missing disks if the container is populated
d23fe947 5213 */
3960e579 5214 if (super->current_vol == 0) {
d23fe947 5215 mpb->num_disks = 0;
3960e579
DW		 5216 for (i = 0; i < info->failed_disks; i++) {
5217 struct imsm_disk *disk;
5218
5219 mpb->num_disks++;
5220 disk = __get_imsm_disk(mpb, i);
5221 disk->status = CONFIGURED_DISK | FAILED_DISK;
5222 disk->scsi_id = __cpu_to_le32(~(__u32)0);
5223 snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN,
5224 "missing:%d", i);
5225 }
5226 find_missing(super);
5227 } else {
5228 int missing = 0;
5229 struct dl *d;
5230
5231 for (d = super->missing; d; d = d->next)
5232 missing++;
5233 if (info->failed_disks > missing) {
e7b84f9d 5234 pr_err("unable to add 'missing' disk to container\n");
3960e579
DW		 5235 return 0;
5236 }
5237 }
5a038140 5238
aa534678
DW		 5239 if (!check_name(super, name, 0))
5240 return 0;
503975b9
N		 5241 dv = xmalloc(sizeof(*dv));
5242 dev = xcalloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1));
c2c087e6 5243 strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN);
e03640bd 5244 array_blocks = calc_array_size(info->level, info->raid_disks,
03bcbc65 5245 info->layout, info->chunk_size,
5551b113 5246 size * 2);
979d38be
DW		 5247 /* round array size down to closest MB */
5248 array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT;
5249
c2c087e6
DW		 5250 dev->size_low = __cpu_to_le32((__u32) array_blocks);
5251 dev->size_high = __cpu_to_le32((__u32) (array_blocks >> 32));
1a2487c2 5252 dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING);
c2c087e6
DW		 5253 vol = &dev->vol;
5254 vol->migr_state = 0;
1484e727 5255 set_migr_type(dev, MIGR_INIT);
3960e579 5256 vol->dirty = !info->state;
f8f603f1 5257 vol->curr_migr_unit = 0;
238c0a71 5258 map = get_imsm_map(dev, MAP_0);
5551b113
CA		 5259 set_pba_of_lba0(map, super->create_offset);
5260 set_blocks_per_member(map, info_to_blocks_per_member(info, size));
ef6ffade 5261 map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info));
0556e1a2 5262 map->failed_disk_num = ~0;
bf4442ab 5263 if (info->level > 0)
fffaf1ff
N		 5264 map->map_state = (info->state ? IMSM_T_STATE_NORMAL
5265 : IMSM_T_STATE_UNINITIALIZED);
bf4442ab
AK		 5266 else
5267 map->map_state = info->failed_disks ? IMSM_T_STATE_FAILED :
5268 IMSM_T_STATE_NORMAL;
252d23c0 5269 map->ddf = 1;
ef6ffade
DW		 5270
5271 if (info->level == 1 && info->raid_disks > 2) {
38950822
AW		 5272 free(dev);
5273 free(dv);
7a862a02 5274 pr_err("imsm does not support more than 2 disksin a raid1 volume\n");
ef6ffade
DW		 5275 return 0;
5276 }
81062a36
DW		 5277
5278 map->raid_level = info->level;
4d1313e9 5279 if (info->level == 10) {
c2c087e6 5280 map->raid_level = 1;
4d1313e9 5281 map->num_domains = info->raid_disks / 2;
81062a36
DW		 5282 } else if (info->level == 1)
5283 map->num_domains = info->raid_disks;
5284 else
ff596308 5285 map->num_domains = 1;
81062a36 5286
5551b113
CA		 5287 /* info->size is only int so use the 'size' parameter instead */
5288 num_data_stripes = (size * 2) / info_to_blocks_per_strip(info);
5289 num_data_stripes /= map->num_domains;
5290 set_num_data_stripes(map, num_data_stripes);
ef6ffade 5291
c2c087e6
DW		 5292 map->num_members = info->raid_disks;
5293 for (i = 0; i < map->num_members; i++) {
5294 /* initialized in add_to_super */
4eb26970 5295 set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD);
c2c087e6 5296 }
949c47a0 5297 mpb->num_raid_devs++;
ba2de7ba
DW		 5298
5299 dv->dev = dev;
5300 dv->index = super->current_vol;
5301 dv->next = super->devlist;
5302 super->devlist = dv;
c2c087e6 5303
4d1313e9
DW		 5304 imsm_update_version_info(super);
5305
c2c087e6 5306 return 1;
cdddbdbc
DW		 5307}
5308
bf5a934a
DW		 5309static int init_super_imsm(struct supertype *st, mdu_array_info_t *info,
5310 unsigned long long size, char *name,
83cd1e97
N		 5311 char *homehost, int *uuid,
5312 unsigned long long data_offset)
bf5a934a
DW		 5313{
5314 /* This is primarily called by Create when creating a new array.
5315 * We will then get add_to_super called for each component, and then
5316 * write_init_super called to write it out to each device.
5317 * For IMSM, Create can create on fresh devices or on a pre-existing
5318 * array.
5319 * To create on a pre-existing array a different method will be called.
5320 * This one is just for fresh drives.
5321 */
5322 struct intel_super *super;
5323 struct imsm_super *mpb;
5324 size_t mpb_size;
4d1313e9 5325 char *version;
bf5a934a 5326
83cd1e97 5327 if (data_offset != INVALID_SECTORS) {
ed503f89 5328 pr_err("data-offset not supported by imsm\n");
83cd1e97
N		 5329 return 0;
5330 }
5331
bf5a934a 5332 if (st->sb)
83cd1e97
N		 5333 return init_super_imsm_volume(st, info, size, name, homehost, uuid,
5334 data_offset);
e683ca88
DW		 5335
5336 if (info)
5337 mpb_size = disks_to_mpb_size(info->nr_disks);
5338 else
f36a9ecd 5339 mpb_size = MAX_SECTOR_SIZE;
bf5a934a 5340
49133e57 5341 super = alloc_super();
f36a9ecd
PB		 5342 if (super &&
5343 posix_memalign(&super->buf, MAX_SECTOR_SIZE, mpb_size) != 0) {
8d67477f 5344 free_imsm(super);
e683ca88
DW		 5345 super = NULL;
5346 }
5347 if (!super) {
1ade5cc1 5348 pr_err("could not allocate superblock\n");
bf5a934a
DW		 5349 return 0;
5350 }
de44e46f
PB		 5351 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
5352 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) {
1ade5cc1 5353 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8 5354 free(super->buf);
8d67477f 5355 free_imsm(super);
8e59f3d8
AK		 5356 return 0;
5357 }
e683ca88 5358 memset(super->buf, 0, mpb_size);
ef649044 5359 mpb = super->buf;
e683ca88
DW		 5360 mpb->mpb_size = __cpu_to_le32(mpb_size);
5361 st->sb = super;
5362
5363 if (info == NULL) {
5364 /* zeroing superblock */
5365 return 0;
5366 }
bf5a934a 5367
4d1313e9
DW		 5368 mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
5369
5370 version = (char *) mpb->sig;
5371 strcpy(version, MPB_SIGNATURE);
5372 version += strlen(MPB_SIGNATURE);
5373 strcpy(version, MPB_VERSION_RAID0);
bf5a934a 5374
bf5a934a
DW		 5375 return 1;
5376}
5377
0e600426 5378#ifndef MDASSEMBLE
f20c3968 5379static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk,
bf5a934a
DW		 5380 int fd, char *devname)
5381{
5382 struct intel_super *super = st->sb;
d23fe947 5383 struct imsm_super *mpb = super->anchor;
3960e579 5384 struct imsm_disk *_disk;
bf5a934a
DW		 5385 struct imsm_dev *dev;
5386 struct imsm_map *map;
3960e579 5387 struct dl *dl, *df;
4eb26970 5388 int slot;
bf5a934a 5389
949c47a0 5390 dev = get_imsm_dev(super, super->current_vol);
238c0a71 5391 map = get_imsm_map(dev, MAP_0);
bf5a934a 5392
208933a7 5393 if (! (dk->state & (1<<MD_DISK_SYNC))) {
e7b84f9d 5394 pr_err("%s: Cannot add spare devices to IMSM volume\n",
208933a7
N		 5395 devname);
5396 return 1;
5397 }
5398
efb30e7f
DW		 5399 if (fd == -1) {
5400 /* we're doing autolayout so grab the pre-marked (in
5401 * validate_geometry) raid_disk
5402 */
5403 for (dl = super->disks; dl; dl = dl->next)
5404 if (dl->raiddisk == dk->raid_disk)
5405 break;
5406 } else {
5407 for (dl = super->disks; dl ; dl = dl->next)
5408 if (dl->major == dk->major &&
5409 dl->minor == dk->minor)
5410 break;
5411 }
d23fe947 5412
208933a7 5413 if (!dl) {
e7b84f9d 5414 pr_err("%s is not a member of the same container\n", devname);
f20c3968 5415 return 1;
208933a7 5416 }
bf5a934a 5417
d23fe947
DW		 5418 /* add a pristine spare to the metadata */
5419 if (dl->index < 0) {
5420 dl->index = super->anchor->num_disks;
5421 super->anchor->num_disks++;
5422 }
4eb26970
DW		 5423 /* Check the device has not already been added */
5424 slot = get_imsm_disk_slot(map, dl->index);
5425 if (slot >= 0 &&
238c0a71 5426 (get_imsm_ord_tbl_ent(dev, slot, MAP_X) & IMSM_ORD_REBUILD) == 0) {
e7b84f9d 5427 pr_err("%s has been included in this array twice\n",
4eb26970
DW		 5428 devname);
5429 return 1;
5430 }
656b6b5a 5431 set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index);
ee5aad5a 5432 dl->disk.status = CONFIGURED_DISK;
d23fe947 5433
3960e579
DW		 5434 /* update size of 'missing' disks to be at least as large as the
5435 * largest acitve member (we only have dummy missing disks when
5436 * creating the first volume)
5437 */
5438 if (super->current_vol == 0) {
5439 for (df = super->missing; df; df = df->next) {
5551b113
CA		 5440 if (total_blocks(&dl->disk) > total_blocks(&df->disk))
5441 set_total_blocks(&df->disk, total_blocks(&dl->disk));
3960e579
DW		 5442 _disk = __get_imsm_disk(mpb, df->index);
5443 *_disk = df->disk;
5444 }
5445 }
5446
5447 /* refresh unset/failed slots to point to valid 'missing' entries */
5448 for (df = super->missing; df; df = df->next)
5449 for (slot = 0; slot < mpb->num_disks; slot++) {
238c0a71 5450 __u32 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
3960e579
DW		 5451
5452 if ((ord & IMSM_ORD_REBUILD) == 0)
5453 continue;
5454 set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD);
1ace8403 5455 if (is_gen_migration(dev)) {
238c0a71
AK		 5456 struct imsm_map *map2 = get_imsm_map(dev,
5457 MAP_1);
0a108d63 5458 int slot2 = get_imsm_disk_slot(map2, df->index);
089f9d79 5459 if (slot2 < map2->num_members && slot2 >= 0) {
1ace8403 5460 __u32 ord2 = get_imsm_ord_tbl_ent(dev,
238c0a71
AK		 5461 slot2,
5462 MAP_1);
1ace8403
AK		 5463 if ((unsigned)df->index ==
5464 ord_to_idx(ord2))
5465 set_imsm_ord_tbl_ent(map2,
0a108d63 5466 slot2,
1ace8403
AK		 5467 df->index |
5468 IMSM_ORD_REBUILD);
5469 }
5470 }
3960e579
DW		 5471 dprintf("set slot:%d to missing disk:%d\n", slot, df->index);
5472 break;
5473 }
5474
d23fe947
DW		 5475 /* if we are creating the first raid device update the family number */
5476 if (super->current_vol == 0) {
5477 __u32 sum;
5478 struct imsm_dev *_dev = __get_imsm_dev(mpb, 0);
d23fe947 5479
3960e579 5480 _disk = __get_imsm_disk(mpb, dl->index);
791b666a 5481 if (!_dev || !_disk) {
e7b84f9d 5482 pr_err("BUG mpb setup error\n");
791b666a
AW		 5483 return 1;
5484 }
d23fe947
DW		 5485 *_dev = *dev;
5486 *_disk = dl->disk;
148acb7b
DW		 5487 sum = random32();
5488 sum += __gen_imsm_checksum(mpb);
d23fe947 5489 mpb->family_num = __cpu_to_le32(sum);
148acb7b 5490 mpb->orig_family_num = mpb->family_num;
d23fe947 5491 }
ca0748fa 5492 super->current_disk = dl;
f20c3968 5493 return 0;
bf5a934a
DW		 5494}
5495
a8619d23
AK		 5496/* mark_spare()
5497 * Function marks disk as spare and restores disk serial
5498 * in case it was previously marked as failed by takeover operation
5499 * reruns:
5500 * -1 : critical error
5501 * 0 : disk is marked as spare but serial is not set
5502 * 1 : success
5503 */
5504int mark_spare(struct dl *disk)
5505{
5506 __u8 serial[MAX_RAID_SERIAL_LEN];
5507 int ret_val = -1;
5508
5509 if (!disk)
5510 return ret_val;
5511
5512 ret_val = 0;
5513 if (!imsm_read_serial(disk->fd, NULL, serial)) {
5514 /* Restore disk serial number, because takeover marks disk
5515 * as failed and adds to serial ':0' before it becomes
5516 * a spare disk.
5517 */
5518 serialcpy(disk->serial, serial);
5519 serialcpy(disk->disk.serial, serial);
5520 ret_val = 1;
5521 }
5522 disk->disk.status = SPARE_DISK;
5523 disk->index = -1;
5524
5525 return ret_val;
5526}
88654014 5527
f20c3968 5528static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk,
72ca9bcf
N		 5529 int fd, char *devname,
5530 unsigned long long data_offset)
cdddbdbc 5531{
c2c087e6 5532 struct intel_super *super = st->sb;
c2c087e6
DW		 5533 struct dl *dd;
5534 unsigned long long size;
fa7bb6f8 5535 unsigned int member_sector_size;
f2f27e63 5536 __u32 id;
c2c087e6
DW		 5537 int rv;
5538 struct stat stb;
5539
88654014
LM		 5540 /* If we are on an RAID enabled platform check that the disk is
5541 * attached to the raid controller.
5542 * We do not need to test disks attachment for container based additions,
5543 * they shall be already tested when container was created/assembled.
88c32bb1 5544 */
d424212e 5545 rv = find_intel_hba_capability(fd, super, devname);
f2f5c343 5546 /* no orom/efi or non-intel hba of the disk */
f0f5a016
LM		 5547 if (rv != 0) {
5548 dprintf("capability: %p fd: %d ret: %d\n",
5549 super->orom, fd, rv);
5550 return 1;
88c32bb1
DW		 5551 }
5552
f20c3968
DW		 5553 if (super->current_vol >= 0)
5554 return add_to_super_imsm_volume(st, dk, fd, devname);
bf5a934a 5555
c2c087e6 5556 fstat(fd, &stb);
503975b9 5557 dd = xcalloc(sizeof(*dd), 1);
c2c087e6
DW		 5558 dd->major = major(stb.st_rdev);
5559 dd->minor = minor(stb.st_rdev);
503975b9 5560 dd->devname = devname ? xstrdup(devname) : NULL;
c2c087e6 5561 dd->fd = fd;
689c9bf3 5562 dd->e = NULL;
1a64be56 5563 dd->action = DISK_ADD;
c2c087e6 5564 rv = imsm_read_serial(fd, devname, dd->serial);
32ba9157 5565 if (rv) {
e7b84f9d 5566 pr_err("failed to retrieve scsi serial, aborting\n");
20bee0f8
PB		 5567 if (dd->devname)
5568 free(dd->devname);
949c47a0 5569 free(dd);
0030e8d6 5570 abort();
c2c087e6 5571 }
20bee0f8
PB		 5572 if (super->hba && ((super->hba->type == SYS_DEV_NVME) ||
5573 (super->hba->type == SYS_DEV_VMD))) {
5574 int i;
5575 char *devpath = diskfd_to_devpath(fd);
5576 char controller_path[PATH_MAX];
5577
5578 if (!devpath) {
5579 pr_err("failed to get devpath, aborting\n");
5580 if (dd->devname)
5581 free(dd->devname);
5582 free(dd);
5583 return 1;
5584 }
5585
5586 snprintf(controller_path, PATH_MAX-1, "%s/device", devpath);
5587 free(devpath);
5588
5589 if (devpath_to_vendor(controller_path) == 0x8086) {
5590 /*
5591 * If Intel's NVMe drive has serial ended with
5592 * "-A","-B","-1" or "-2" it means that this is "x8"
5593 * device (double drive on single PCIe card).
5594 * User should be warned about potential data loss.
5595 */
5596 for (i = MAX_RAID_SERIAL_LEN-1; i > 0; i--) {
5597 /* Skip empty character at the end */
5598 if (dd->serial[i] == 0)
5599 continue;
5600
5601 if (((dd->serial[i] == 'A') ||
5602 (dd->serial[i] == 'B') ||
5603 (dd->serial[i] == '1') ||
5604 (dd->serial[i] == '2')) &&
5605 (dd->serial[i-1] == '-'))
5606 pr_err("\tThe action you are about to take may put your data at risk.\n"
5607 "\tPlease note that x8 devices may consist of two separate x4 devices "
5608 "located on a single PCIe port.\n"
5609 "\tRAID 0 is the only supported configuration for this type of x8 device.\n");
5610 break;
5611 }
32716c51
PB		 5612 } else if (super->hba->type == SYS_DEV_VMD && super->orom &&
5613 !imsm_orom_has_tpv_support(super->orom)) {
5614 pr_err("\tPlatform configuration does not support non-Intel NVMe drives.\n"
5615 "\tPlease refer to Intel(R) RSTe user guide.\n");
5616 free(dd->devname);
5617 free(dd);
5618 return 1;
20bee0f8
PB		 5619 }
5620 }
c2c087e6 5621
c2c087e6 5622 get_dev_size(fd, NULL, &size);
fa7bb6f8
PB		 5623 get_dev_sector_size(fd, NULL, &member_sector_size);
5624
5625 if (super->sector_size == 0) {
5626 /* this a first device, so sector_size is not set yet */
5627 super->sector_size = member_sector_size;
5628 } else if (member_sector_size != super->sector_size) {
5629 pr_err("Mixing between different sector size is forbidden, aborting...\n");
5630 if (dd->devname)
5631 free(dd->devname);
5632 free(dd);
5633 return 1;
5634 }
5635
71e5411e 5636 /* clear migr_rec when adding disk to container */
de44e46f
PB		 5637 memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*super->sector_size);
5638 if (lseek64(fd, size - MIGR_REC_SECTOR_POSITION*super->sector_size,
5639 SEEK_SET) >= 0) {
71e5411e 5640 if (write(fd, super->migr_rec_buf,
de44e46f
PB		 5641 MIGR_REC_BUF_SECTORS*super->sector_size) !=
5642 MIGR_REC_BUF_SECTORS*super->sector_size)
71e5411e
PB		 5643 perror("Write migr_rec failed");
5644 }
5645
c2c087e6 5646 size /= 512;
1f24f035 5647 serialcpy(dd->disk.serial, dd->serial);
5551b113
CA		 5648 set_total_blocks(&dd->disk, size);
5649 if (__le32_to_cpu(dd->disk.total_blocks_hi) > 0) {
5650 struct imsm_super *mpb = super->anchor;
5651 mpb->attributes |= MPB_ATTRIB_2TB_DISK;
5652 }
a8619d23 5653 mark_spare(dd);
c2c087e6 5654 if (sysfs_disk_to_scsi_id(fd, &id) == 0)
b9f594fe 5655 dd->disk.scsi_id = __cpu_to_le32(id);
c2c087e6 5656 else
b9f594fe 5657 dd->disk.scsi_id = __cpu_to_le32(0);
43dad3d6
DW		 5658
5659 if (st->update_tail) {
1a64be56
LM		 5660 dd->next = super->disk_mgmt_list;
5661 super->disk_mgmt_list = dd;
43dad3d6
DW		 5662 } else {
5663 dd->next = super->disks;
5664 super->disks = dd;
ceaf0ee1 5665 super->updates_pending++;
43dad3d6 5666 }
f20c3968
DW		 5667
5668 return 0;
cdddbdbc
DW		 5669}
5670
1a64be56
LM		 5671static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk)
5672{
5673 struct intel_super *super = st->sb;
5674 struct dl *dd;
5675
5676 /* remove from super works only in mdmon - for communication
5677 * manager - monitor. Check if communication memory buffer
5678 * is prepared.
5679 */
5680 if (!st->update_tail) {
1ade5cc1 5681 pr_err("shall be used in mdmon context only\n");
1a64be56
LM		 5682 return 1;
5683 }
503975b9 5684 dd = xcalloc(1, sizeof(*dd));
1a64be56
LM		 5685 dd->major = dk->major;
5686 dd->minor = dk->minor;
1a64be56 5687 dd->fd = -1;
a8619d23 5688 mark_spare(dd);
1a64be56
LM		 5689 dd->action = DISK_REMOVE;
5690
5691 dd->next = super->disk_mgmt_list;
5692 super->disk_mgmt_list = dd;
5693
1a64be56
LM		 5694 return 0;
5695}
5696
f796af5d
DW		 5697static int store_imsm_mpb(int fd, struct imsm_super *mpb);
5698
5699static union {
f36a9ecd 5700 char buf[MAX_SECTOR_SIZE];
f796af5d 5701 struct imsm_super anchor;
f36a9ecd 5702} spare_record __attribute__ ((aligned(MAX_SECTOR_SIZE)));
c2c087e6 5703
d23fe947
DW		 5704/* spare records have their own family number and do not have any defined raid
5705 * devices
5706 */
5707static int write_super_imsm_spares(struct intel_super *super, int doclose)
5708{
d23fe947 5709 struct imsm_super *mpb = super->anchor;
f796af5d 5710 struct imsm_super *spare = &spare_record.anchor;
d23fe947
DW		 5711 __u32 sum;
5712 struct dl *d;
5713
68641cdb
JS		 5714 spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super));
5715 spare->generation_num = __cpu_to_le32(1UL);
f796af5d 5716 spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
68641cdb
JS		 5717 spare->num_disks = 1;
5718 spare->num_raid_devs = 0;
5719 spare->cache_size = mpb->cache_size;
5720 spare->pwr_cycle_count = __cpu_to_le32(1);
f796af5d
DW		 5721
5722 snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH,
5723 MPB_SIGNATURE MPB_VERSION_RAID0);
d23fe947
DW		 5724
5725 for (d = super->disks; d; d = d->next) {
8796fdc4 5726 if (d->index != -1)
d23fe947
DW		 5727 continue;
5728
f796af5d 5729 spare->disk[0] = d->disk;
027c374f
CA		 5730 if (__le32_to_cpu(d->disk.total_blocks_hi) > 0)
5731 spare->attributes |= MPB_ATTRIB_2TB_DISK;
5732
f36a9ecd
PB		 5733 if (super->sector_size == 4096)
5734 convert_to_4k_imsm_disk(&spare->disk[0]);
5735
f796af5d
DW		 5736 sum = __gen_imsm_checksum(spare);
5737 spare->family_num = __cpu_to_le32(sum);
5738 spare->orig_family_num = 0;
5739 sum = __gen_imsm_checksum(spare);
5740 spare->check_sum = __cpu_to_le32(sum);
d23fe947 5741
f796af5d 5742 if (store_imsm_mpb(d->fd, spare)) {
1ade5cc1
N		 5743 pr_err("failed for device %d:%d %s\n",
5744 d->major, d->minor, strerror(errno));
e74255d9 5745 return 1;
d23fe947
DW		 5746 }
5747 if (doclose) {
5748 close(d->fd);
5749 d->fd = -1;
5750 }
5751 }
5752
e74255d9 5753 return 0;
d23fe947
DW		 5754}
5755
36988a3d 5756static int write_super_imsm(struct supertype *st, int doclose)
cdddbdbc 5757{
36988a3d 5758 struct intel_super *super = st->sb;
f36a9ecd 5759 unsigned int sector_size = super->sector_size;
949c47a0 5760 struct imsm_super *mpb = super->anchor;
c2c087e6
DW		 5761 struct dl *d;
5762 __u32 generation;
5763 __u32 sum;
d23fe947 5764 int spares = 0;
949c47a0 5765 int i;
a48ac0a8 5766 __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk);
36988a3d 5767 int num_disks = 0;
146c6260 5768 int clear_migration_record = 1;
bbab0940 5769 __u32 bbm_log_size;
cdddbdbc 5770
c2c087e6
DW		 5771 /* 'generation' is incremented everytime the metadata is written */
5772 generation = __le32_to_cpu(mpb->generation_num);
5773 generation++;
5774 mpb->generation_num = __cpu_to_le32(generation);
5775
148acb7b
DW		 5776 /* fix up cases where previous mdadm releases failed to set
5777 * orig_family_num
5778 */
5779 if (mpb->orig_family_num == 0)
5780 mpb->orig_family_num = mpb->family_num;
5781
d23fe947 5782 for (d = super->disks; d; d = d->next) {
8796fdc4 5783 if (d->index == -1)
d23fe947 5784 spares++;
36988a3d 5785 else {
d23fe947 5786 mpb->disk[d->index] = d->disk;
36988a3d
AK		 5787 num_disks++;
5788 }
d23fe947 5789 }
36988a3d 5790 for (d = super->missing; d; d = d->next) {
47ee5a45 5791 mpb->disk[d->index] = d->disk;
36988a3d
AK		 5792 num_disks++;
5793 }
5794 mpb->num_disks = num_disks;
5795 mpb_size += sizeof(struct imsm_disk) * mpb->num_disks;
b9f594fe 5796
949c47a0
DW		 5797 for (i = 0; i < mpb->num_raid_devs; i++) {
5798 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
36988a3d
AK		 5799 struct imsm_dev *dev2 = get_imsm_dev(super, i);
5800 if (dev && dev2) {
5801 imsm_copy_dev(dev, dev2);
5802 mpb_size += sizeof_imsm_dev(dev, 0);
5803 }
146c6260
AK		 5804 if (is_gen_migration(dev2))
5805 clear_migration_record = 0;
949c47a0 5806 }
bbab0940
TM		 5807
5808 bbm_log_size = get_imsm_bbm_log_size(super->bbm_log);
5809
5810 if (bbm_log_size) {
5811 memcpy((void *)mpb + mpb_size, super->bbm_log, bbm_log_size);
5812 mpb->attributes |= MPB_ATTRIB_BBM;
5813 } else
5814 mpb->attributes &= ~MPB_ATTRIB_BBM;
5815
5816 super->anchor->bbm_log_size = __cpu_to_le32(bbm_log_size);
5817 mpb_size += bbm_log_size;
a48ac0a8 5818 mpb->mpb_size = __cpu_to_le32(mpb_size);
949c47a0 5819
bbab0940
TM		 5820#ifdef DEBUG
5821 assert(super->len == 0 || mpb_size <= super->len);
5822#endif
5823
c2c087e6 5824 /* recalculate checksum */
949c47a0 5825 sum = __gen_imsm_checksum(mpb);
c2c087e6
DW		 5826 mpb->check_sum = __cpu_to_le32(sum);
5827
51d83f5d
AK		 5828 if (super->clean_migration_record_by_mdmon) {
5829 clear_migration_record = 1;
5830 super->clean_migration_record_by_mdmon = 0;
5831 }
146c6260 5832 if (clear_migration_record)
de44e46f
PB		 5833 memset(super->migr_rec_buf, 0,
5834 MIGR_REC_BUF_SECTORS*sector_size);
146c6260 5835
f36a9ecd
PB		 5836 if (sector_size == 4096)
5837 convert_to_4k(super);
5838
d23fe947 5839 /* write the mpb for disks that compose raid devices */
c2c087e6 5840 for (d = super->disks; d ; d = d->next) {
86c54047 5841 if (d->index < 0 || is_failed(&d->disk))
d23fe947 5842 continue;
30602f53 5843
146c6260
AK		 5844 if (clear_migration_record) {
5845 unsigned long long dsize;
5846
5847 get_dev_size(d->fd, NULL, &dsize);
de44e46f
PB		 5848 if (lseek64(d->fd, dsize - sector_size,
5849 SEEK_SET) >= 0) {
17a4eaf9 5850 if (write(d->fd, super->migr_rec_buf,
de44e46f
PB		 5851 MIGR_REC_BUF_SECTORS*sector_size) !=
5852 MIGR_REC_BUF_SECTORS*sector_size)
9e2d750d 5853 perror("Write migr_rec failed");
146c6260
AK		 5854 }
5855 }
51d83f5d
AK		 5856
5857 if (store_imsm_mpb(d->fd, mpb))
5858 fprintf(stderr,
1ade5cc1
N		 5859 "failed for device %d:%d (fd: %d)%s\n",
5860 d->major, d->minor,
51d83f5d
AK		 5861 d->fd, strerror(errno));
5862
c2c087e6
DW		 5863 if (doclose) {
5864 close(d->fd);
5865 d->fd = -1;
5866 }
5867 }
5868
d23fe947
DW		 5869 if (spares)
5870 return write_super_imsm_spares(super, doclose);
5871
e74255d9 5872 return 0;
c2c087e6
DW		 5873}
5874
9b1fb677 5875static int create_array(struct supertype *st, int dev_idx)
43dad3d6
DW		 5876{
5877 size_t len;
5878 struct imsm_update_create_array *u;
5879 struct intel_super *super = st->sb;
9b1fb677 5880 struct imsm_dev *dev = get_imsm_dev(super, dev_idx);
238c0a71 5881 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW		 5882 struct disk_info *inf;
5883 struct imsm_disk *disk;
5884 int i;
43dad3d6 5885
54c2c1ea
DW		 5886 len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) +
5887 sizeof(*inf) * map->num_members;
503975b9 5888 u = xmalloc(len);
43dad3d6 5889 u->type = update_create_array;
9b1fb677 5890 u->dev_idx = dev_idx;
43dad3d6 5891 imsm_copy_dev(&u->dev, dev);
54c2c1ea
DW		 5892 inf = get_disk_info(u);
5893 for (i = 0; i < map->num_members; i++) {
238c0a71 5894 int idx = get_imsm_disk_idx(dev, i, MAP_X);
9b1fb677 5895
54c2c1ea 5896 disk = get_imsm_disk(super, idx);
1ca5c8e0
N		 5897 if (!disk)
5898 disk = get_imsm_missing(super, idx);
54c2c1ea
DW		 5899 serialcpy(inf[i].serial, disk->serial);
5900 }
43dad3d6
DW		 5901 append_metadata_update(st, u, len);
5902
5903 return 0;
5904}
5905
1a64be56 5906static int mgmt_disk(struct supertype *st)
43dad3d6
DW		 5907{
5908 struct intel_super *super = st->sb;
5909 size_t len;
1a64be56 5910 struct imsm_update_add_remove_disk *u;
43dad3d6 5911
1a64be56 5912 if (!super->disk_mgmt_list)
43dad3d6
DW		 5913 return 0;
5914
5915 len = sizeof(*u);
503975b9 5916 u = xmalloc(len);
1a64be56 5917 u->type = update_add_remove_disk;
43dad3d6
DW		 5918 append_metadata_update(st, u, len);
5919
5920 return 0;
5921}
5922
c2c087e6
DW		 5923static int write_init_super_imsm(struct supertype *st)
5924{
9b1fb677
DW		 5925 struct intel_super *super = st->sb;
5926 int current_vol = super->current_vol;
5927
5928 /* we are done with current_vol reset it to point st at the container */
5929 super->current_vol = -1;
5930
8273f55e 5931 if (st->update_tail) {
43dad3d6
DW		 5932 /* queue the recently created array / added disk
5933 * as a metadata update */
43dad3d6 5934 int rv;
8273f55e 5935
43dad3d6 5936 /* determine if we are creating a volume or adding a disk */
9b1fb677 5937 if (current_vol < 0) {
1a64be56
LM		 5938 /* in the mgmt (add/remove) disk case we are running
5939 * in mdmon context, so don't close fd's
43dad3d6 5940 */
1a64be56 5941 return mgmt_disk(st);
43dad3d6 5942 } else
9b1fb677 5943 rv = create_array(st, current_vol);
8273f55e 5944
43dad3d6 5945 return rv;
d682f344
N		 5946 } else {
5947 struct dl *d;
5948 for (d = super->disks; d; d = d->next)
ba728be7 5949 Kill(d->devname, NULL, 0, -1, 1);
36988a3d 5950 return write_super_imsm(st, 1);
d682f344 5951 }
cdddbdbc 5952}
0e600426 5953#endif
cdddbdbc 5954
e683ca88 5955static int store_super_imsm(struct supertype *st, int fd)
cdddbdbc 5956{
e683ca88
DW		 5957 struct intel_super *super = st->sb;
5958 struct imsm_super *mpb = super ? super->anchor : NULL;
551c80c1 5959
e683ca88 5960 if (!mpb)
ad97895e
DW		 5961 return 1;
5962
1799c9e8 5963#ifndef MDASSEMBLE
f36a9ecd
PB		 5964 if (super->sector_size == 4096)
5965 convert_to_4k(super);
e683ca88 5966 return store_imsm_mpb(fd, mpb);
1799c9e8
N		 5967#else
5968 return 1;
5969#endif
cdddbdbc
DW		 5970}
5971
0e600426 5972#ifndef MDASSEMBLE
cdddbdbc
DW		 5973static int validate_geometry_imsm_container(struct supertype *st, int level,
5974 int layout, int raiddisks, int chunk,
af4348dd
N		 5975 unsigned long long size,
5976 unsigned long long data_offset,
5977 char *dev,
2c514b71
NB		 5978 unsigned long long *freesize,
5979 int verbose)
cdddbdbc 5980{
c2c087e6
DW		 5981 int fd;
5982 unsigned long long ldsize;
594dc1b8 5983 struct intel_super *super;
f2f5c343 5984 int rv = 0;
cdddbdbc 5985
c2c087e6
DW		 5986 if (level != LEVEL_CONTAINER)
5987 return 0;
5988 if (!dev)
5989 return 1;
5990
5991 fd = open(dev, O_RDONLY|O_EXCL, 0);
5992 if (fd < 0) {
ba728be7 5993 if (verbose > 0)
e7b84f9d 5994 pr_err("imsm: Cannot open %s: %s\n",
2c514b71 5995 dev, strerror(errno));
c2c087e6
DW		 5996 return 0;
5997 }
5998 if (!get_dev_size(fd, dev, &ldsize)) {
5999 close(fd);
6000 return 0;
6001 }
f2f5c343
LM		 6002
6003 /* capabilities retrieve could be possible
6004 * note that there is no fd for the disks in array.
6005 */
6006 super = alloc_super();
8d67477f
TM		 6007 if (!super) {
6008 close(fd);
6009 return 0;
6010 }
fa7bb6f8
PB		 6011 if (!get_dev_sector_size(fd, NULL, &super->sector_size)) {
6012 close(fd);
6013 free_imsm(super);
6014 return 0;
6015 }
6016
ba728be7 6017 rv = find_intel_hba_capability(fd, super, verbose > 0 ? dev : NULL);
f2f5c343
LM		 6018 if (rv != 0) {
6019#if DEBUG
6020 char str[256];
6021 fd2devname(fd, str);
1ade5cc1 6022 dprintf("fd: %d %s orom: %p rv: %d raiddisk: %d\n",
f2f5c343
LM		 6023 fd, str, super->orom, rv, raiddisks);
6024#endif
6025 /* no orom/efi or non-intel hba of the disk */
6026 close(fd);
6027 free_imsm(super);
6028 return 0;
6029 }
c2c087e6 6030 close(fd);
9126b9a8
CA		 6031 if (super->orom) {
6032 if (raiddisks > super->orom->tds) {
6033 if (verbose)
7a862a02 6034 pr_err("%d exceeds maximum number of platform supported disks: %d\n",
9126b9a8
CA		 6035 raiddisks, super->orom->tds);
6036 free_imsm(super);
6037 return 0;
6038 }
6039 if ((super->orom->attr & IMSM_OROM_ATTR_2TB_DISK) == 0 &&
6040 (ldsize >> 9) >> 32 > 0) {
6041 if (verbose)
e7b84f9d 6042 pr_err("%s exceeds maximum platform supported size\n", dev);
9126b9a8
CA		 6043 free_imsm(super);
6044 return 0;
6045 }
f2f5c343 6046 }
c2c087e6 6047
af4348dd 6048 *freesize = avail_size_imsm(st, ldsize >> 9, data_offset);
f2f5c343 6049 free_imsm(super);
c2c087e6
DW		 6050
6051 return 1;
cdddbdbc
DW		 6052}
6053
0dcecb2e
DW		 6054static unsigned long long find_size(struct extent *e, int *idx, int num_extents)
6055{
6056 const unsigned long long base_start = e[*idx].start;
6057 unsigned long long end = base_start + e[*idx].size;
6058 int i;
6059
6060 if (base_start == end)
6061 return 0;
6062
6063 *idx = *idx + 1;
6064 for (i = *idx; i < num_extents; i++) {
6065 /* extend overlapping extents */
6066 if (e[i].start >= base_start &&
6067 e[i].start <= end) {
6068 if (e[i].size == 0)
6069 return 0;
6070 if (e[i].start + e[i].size > end)
6071 end = e[i].start + e[i].size;
6072 } else if (e[i].start > end) {
6073 *idx = i;
6074 break;
6075 }
6076 }
6077
6078 return end - base_start;
6079}
6080
6081static unsigned long long merge_extents(struct intel_super *super, int sum_extents)
6082{
6083 /* build a composite disk with all known extents and generate a new
6084 * 'maxsize' given the "all disks in an array must share a common start
6085 * offset" constraint
6086 */
503975b9 6087 struct extent *e = xcalloc(sum_extents, sizeof(*e));
0dcecb2e
DW		 6088 struct dl *dl;
6089 int i, j;
6090 int start_extent;
6091 unsigned long long pos;
b9d77223 6092 unsigned long long start = 0;
0dcecb2e
DW		 6093 unsigned long long maxsize;
6094 unsigned long reserve;
6095
0dcecb2e
DW		 6096 /* coalesce and sort all extents. also, check to see if we need to
6097 * reserve space between member arrays
6098 */
6099 j = 0;
6100 for (dl = super->disks; dl; dl = dl->next) {
6101 if (!dl->e)
6102 continue;
6103 for (i = 0; i < dl->extent_cnt; i++)
6104 e[j++] = dl->e[i];
6105 }
6106 qsort(e, sum_extents, sizeof(*e), cmp_extent);
6107
6108 /* merge extents */
6109 i = 0;
6110 j = 0;
6111 while (i < sum_extents) {
6112 e[j].start = e[i].start;
6113 e[j].size = find_size(e, &i, sum_extents);
6114 j++;
6115 if (e[j-1].size == 0)
6116 break;
6117 }
6118
6119 pos = 0;
6120 maxsize = 0;
6121 start_extent = 0;
6122 i = 0;
6123 do {
6124 unsigned long long esize;
6125
6126 esize = e[i].start - pos;
6127 if (esize >= maxsize) {
6128 maxsize = esize;
6129 start = pos;
6130 start_extent = i;
6131 }
6132 pos = e[i].start + e[i].size;
6133 i++;
6134 } while (e[i-1].size);
6135 free(e);
6136
a7dd165b
DW		 6137 if (maxsize == 0)
6138 return 0;
6139
6140 /* FIXME assumes volume at offset 0 is the first volume in a
6141 * container
6142 */
0dcecb2e
DW		 6143 if (start_extent > 0)
6144 reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */
6145 else
6146 reserve = 0;
6147
6148 if (maxsize < reserve)
a7dd165b 6149 return 0;
0dcecb2e 6150
5551b113 6151 super->create_offset = ~((unsigned long long) 0);
0dcecb2e 6152 if (start + reserve > super->create_offset)
a7dd165b 6153 return 0; /* start overflows create_offset */
0dcecb2e
DW		 6154 super->create_offset = start + reserve;
6155
6156 return maxsize - reserve;
6157}
6158
88c32bb1
DW		 6159static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks)
6160{
6161 if (level < 0 || level == 6 || level == 4)
6162 return 0;
6163
6164 /* if we have an orom prevent invalid raid levels */
6165 if (orom)
6166 switch (level) {
6167 case 0: return imsm_orom_has_raid0(orom);
6168 case 1:
6169 if (raiddisks > 2)
6170 return imsm_orom_has_raid1e(orom);
1c556e92
DW		 6171 return imsm_orom_has_raid1(orom) && raiddisks == 2;
6172 case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4;
6173 case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2;
88c32bb1
DW		 6174 }
6175 else
6176 return 1; /* not on an Intel RAID platform so anything goes */
6177
6178 return 0;
6179}
6180
ca9de185
LM		 6181static int
6182active_arrays_by_format(char *name, char* hba, struct md_list **devlist,
6183 int dpa, int verbose)
6184{
6185 struct mdstat_ent *mdstat = mdstat_read(0, 0);
594dc1b8 6186 struct mdstat_ent *memb;
ca9de185
LM		 6187 int count = 0;
6188 int num = 0;
594dc1b8 6189 struct md_list *dv;
ca9de185
LM		 6190 int found;
6191
6192 for (memb = mdstat ; memb ; memb = memb->next) {
6193 if (memb->metadata_version &&
6194 (strncmp(memb->metadata_version, "external:", 9) == 0) &&
6195 (strcmp(&memb->metadata_version[9], name) == 0) &&
6196 !is_subarray(memb->metadata_version+9) &&
6197 memb->members) {
6198 struct dev_member *dev = memb->members;
6199 int fd = -1;
6200 while(dev && (fd < 0)) {
503975b9
N		 6201 char *path = xmalloc(strlen(dev->name) + strlen("/dev/") + 1);
6202 num = sprintf(path, "%s%s", "/dev/", dev->name);
6203 if (num > 0)
6204 fd = open(path, O_RDONLY, 0);
089f9d79 6205 if (num <= 0 || fd < 0) {
676e87a8 6206 pr_vrb("Cannot open %s: %s\n",
503975b9 6207 dev->name, strerror(errno));
ca9de185 6208 }
503975b9 6209 free(path);
ca9de185
LM		 6210 dev = dev->next;
6211 }
6212 found = 0;
089f9d79 6213 if (fd >= 0 && disk_attached_to_hba(fd, hba)) {
ca9de185
LM		 6214 struct mdstat_ent *vol;
6215 for (vol = mdstat ; vol ; vol = vol->next) {
089f9d79 6216 if (vol->active > 0 &&
ca9de185 6217 vol->metadata_version &&
9581efb1 6218 is_container_member(vol, memb->devnm)) {
ca9de185
LM		 6219 found++;
6220 count++;
6221 }
6222 }
6223 if (*devlist && (found < dpa)) {
503975b9 6224 dv = xcalloc(1, sizeof(*dv));
9581efb1
N		 6225 dv->devname = xmalloc(strlen(memb->devnm) + strlen("/dev/") + 1);
6226 sprintf(dv->devname, "%s%s", "/dev/", memb->devnm);
503975b9
N		 6227 dv->found = found;
6228 dv->used = 0;
6229 dv->next = *devlist;
6230 *devlist = dv;
ca9de185
LM		 6231 }
6232 }
6233 if (fd >= 0)
6234 close(fd);
6235 }
6236 }
6237 free_mdstat(mdstat);
6238 return count;
6239}
6240
6241#ifdef DEBUG_LOOP
6242static struct md_list*
6243get_loop_devices(void)
6244{
6245 int i;
6246 struct md_list *devlist = NULL;
594dc1b8 6247 struct md_list *dv;
ca9de185
LM		 6248
6249 for(i = 0; i < 12; i++) {
503975b9
N		 6250 dv = xcalloc(1, sizeof(*dv));
6251 dv->devname = xmalloc(40);
ca9de185
LM		 6252 sprintf(dv->devname, "/dev/loop%d", i);
6253 dv->next = devlist;
6254 devlist = dv;
6255 }
6256 return devlist;
6257}
6258#endif
6259
6260static struct md_list*
6261get_devices(const char *hba_path)
6262{
6263 struct md_list *devlist = NULL;
594dc1b8 6264 struct md_list *dv;
ca9de185
LM		 6265 struct dirent *ent;
6266 DIR *dir;
6267 int err = 0;
6268
6269#if DEBUG_LOOP
6270 devlist = get_loop_devices();
6271 return devlist;
6272#endif
6273 /* scroll through /sys/dev/block looking for devices attached to
6274 * this hba
6275 */
6276 dir = opendir("/sys/dev/block");
6277 for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) {
6278 int fd;
6279 char buf[1024];
6280 int major, minor;
6281 char *path = NULL;
6282 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
6283 continue;
6284 path = devt_to_devpath(makedev(major, minor));
6285 if (!path)
6286 continue;
6287 if (!path_attached_to_hba(path, hba_path)) {
6288 free(path);
6289 path = NULL;
6290 continue;
6291 }
6292 free(path);
6293 path = NULL;
6294 fd = dev_open(ent->d_name, O_RDONLY);
6295 if (fd >= 0) {
6296 fd2devname(fd, buf);
6297 close(fd);
6298 } else {
e7b84f9d 6299 pr_err("cannot open device: %s\n",
ca9de185
LM		 6300 ent->d_name);
6301 continue;
6302 }
6303
503975b9
N		 6304 dv = xcalloc(1, sizeof(*dv));
6305 dv->devname = xstrdup(buf);
ca9de185
LM		 6306 dv->next = devlist;
6307 devlist = dv;
6308 }
6309 if (err) {
6310 while(devlist) {
6311 dv = devlist;
6312 devlist = devlist->next;
6313 free(dv->devname);
6314 free(dv);
6315 }
6316 }
562aa102 6317 closedir(dir);
ca9de185
LM		 6318 return devlist;
6319}
6320
6321static int
6322count_volumes_list(struct md_list *devlist, char *homehost,
6323 int verbose, int *found)
6324{
6325 struct md_list *tmpdev;
6326 int count = 0;
594dc1b8 6327 struct supertype *st;
ca9de185
LM		 6328
6329 /* first walk the list of devices to find a consistent set
6330 * that match the criterea, if that is possible.
6331 * We flag the ones we like with 'used'.
6332 */
6333 *found = 0;
6334 st = match_metadata_desc_imsm("imsm");
6335 if (st == NULL) {
676e87a8 6336 pr_vrb("cannot allocate memory for imsm supertype\n");
ca9de185
LM		 6337 return 0;
6338 }
6339
6340 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
6341 char *devname = tmpdev->devname;
6342 struct stat stb;
6343 struct supertype *tst;
6344 int dfd;
6345 if (tmpdev->used > 1)
6346 continue;
6347 tst = dup_super(st);
6348 if (tst == NULL) {
676e87a8 6349 pr_vrb("cannot allocate memory for imsm supertype\n");
ca9de185
LM		 6350 goto err_1;
6351 }
6352 tmpdev->container = 0;
6353 dfd = dev_open(devname, O_RDONLY|O_EXCL);
6354 if (dfd < 0) {
1ade5cc1 6355 dprintf("cannot open device %s: %s\n",
ca9de185
LM		 6356 devname, strerror(errno));
6357 tmpdev->used = 2;
6358 } else if (fstat(dfd, &stb)< 0) {
6359 /* Impossible! */
1ade5cc1 6360 dprintf("fstat failed for %s: %s\n",
ca9de185
LM		 6361 devname, strerror(errno));
6362 tmpdev->used = 2;
6363 } else if ((stb.st_mode & S_IFMT) != S_IFBLK) {
1ade5cc1 6364 dprintf("%s is not a block device.\n",
ca9de185
LM		 6365 devname);
6366 tmpdev->used = 2;
6367 } else if (must_be_container(dfd)) {
6368 struct supertype *cst;
6369 cst = super_by_fd(dfd, NULL);
6370 if (cst == NULL) {
1ade5cc1 6371 dprintf("cannot recognize container type %s\n",
ca9de185
LM		 6372 devname);
6373 tmpdev->used = 2;
6374 } else if (tst->ss != st->ss) {
1ade5cc1 6375 dprintf("non-imsm container - ignore it: %s\n",
ca9de185
LM		 6376 devname);
6377 tmpdev->used = 2;
6378 } else if (!tst->ss->load_container ||
6379 tst->ss->load_container(tst, dfd, NULL))
6380 tmpdev->used = 2;
6381 else {
6382 tmpdev->container = 1;
6383 }
6384 if (cst)
6385 cst->ss->free_super(cst);
6386 } else {
6387 tmpdev->st_rdev = stb.st_rdev;
6388 if (tst->ss->load_super(tst,dfd, NULL)) {
1ade5cc1 6389 dprintf("no RAID superblock on %s\n",
ca9de185
LM		 6390 devname);
6391 tmpdev->used = 2;
6392 } else if (tst->ss->compare_super == NULL) {
1ade5cc1 6393 dprintf("Cannot assemble %s metadata on %s\n",
ca9de185
LM		 6394 tst->ss->name, devname);
6395 tmpdev->used = 2;
6396 }
6397 }
6398 if (dfd >= 0)
6399 close(dfd);
6400 if (tmpdev->used == 2 || tmpdev->used == 4) {
6401 /* Ignore unrecognised devices during auto-assembly */
6402 goto loop;
6403 }
6404 else {
6405 struct mdinfo info;
6406 tst->ss->getinfo_super(tst, &info, NULL);
6407
6408 if (st->minor_version == -1)
6409 st->minor_version = tst->minor_version;
6410
6411 if (memcmp(info.uuid, uuid_zero,
6412 sizeof(int[4])) == 0) {
6413 /* this is a floating spare. It cannot define
6414 * an array unless there are no more arrays of
6415 * this type to be found. It can be included
6416 * in an array of this type though.
6417 */
6418 tmpdev->used = 3;
6419 goto loop;
6420 }
6421
6422 if (st->ss != tst->ss ||
6423 st->minor_version != tst->minor_version ||
6424 st->ss->compare_super(st, tst) != 0) {
6425 /* Some mismatch. If exactly one array matches this host,
6426 * we can resolve on that one.
6427 * Or, if we are auto assembling, we just ignore the second
6428 * for now.
6429 */
1ade5cc1 6430 dprintf("superblock on %s doesn't match others - assembly aborted\n",
ca9de185
LM		 6431 devname);
6432 goto loop;
6433 }
6434 tmpdev->used = 1;
6435 *found = 1;
6436 dprintf("found: devname: %s\n", devname);
6437 }
6438 loop:
6439 if (tst)
6440 tst->ss->free_super(tst);
6441 }
6442 if (*found != 0) {
6443 int err;
6444 if ((err = load_super_imsm_all(st, -1, &st->sb, NULL, devlist, 0)) == 0) {
6445 struct mdinfo *iter, *head = st->ss->container_content(st, NULL);
6446 for (iter = head; iter; iter = iter->next) {
6447 dprintf("content->text_version: %s vol\n",
6448 iter->text_version);
6449 if (iter->array.state & (1<<MD_SB_BLOCK_VOLUME)) {
6450 /* do not assemble arrays with unsupported
6451 configurations */
1ade5cc1 6452 dprintf("Cannot activate member %s.\n",
ca9de185
LM		 6453 iter->text_version);
6454 } else
6455 count++;
6456 }
6457 sysfs_free(head);
6458
6459 } else {
1ade5cc1 6460 dprintf("No valid super block on device list: err: %d %p\n",
ca9de185
LM		 6461 err, st->sb);
6462 }
6463 } else {
1ade5cc1 6464 dprintf("no more devices to examine\n");
ca9de185
LM		 6465 }
6466
6467 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
089f9d79 6468 if (tmpdev->used == 1 && tmpdev->found) {
ca9de185
LM		 6469 if (count) {
6470 if (count < tmpdev->found)
6471 count = 0;
6472 else
6473 count -= tmpdev->found;
6474 }
6475 }
6476 if (tmpdev->used == 1)
6477 tmpdev->used = 4;
6478 }
6479 err_1:
6480 if (st)
6481 st->ss->free_super(st);
6482 return count;
6483}
6484
ca9de185 6485static int
72a45777 6486count_volumes(struct intel_hba *hba, int dpa, int verbose)
ca9de185 6487{
72a45777 6488 struct sys_dev *idev, *intel_devices = find_intel_devices();
ca9de185 6489 int count = 0;
72a45777
PB		 6490 const struct orom_entry *entry;
6491 struct devid_list *dv, *devid_list;
ca9de185 6492
72a45777 6493 if (!hba || !hba->path)
ca9de185
LM		 6494 return 0;
6495
72a45777
PB		 6496 for (idev = intel_devices; idev; idev = idev->next) {
6497 if (strstr(idev->path, hba->path))
6498 break;
6499 }
6500
6501 if (!idev || !idev->dev_id)
ca9de185 6502 return 0;
72a45777
PB		 6503
6504 entry = get_orom_entry_by_device_id(idev->dev_id);
6505
6506 if (!entry || !entry->devid_list)
6507 return 0;
6508
6509 devid_list = entry->devid_list;
6510 for (dv = devid_list; dv; dv = dv->next) {
594dc1b8 6511 struct md_list *devlist;
72a45777
PB		 6512 struct sys_dev *device = device_by_id(dv->devid);
6513 char *hba_path;
6514 int found = 0;
6515
6516 if (device)
6517 hba_path = device->path;
6518 else
6519 return 0;
6520
6521 devlist = get_devices(hba_path);
6522 /* if no intel devices return zero volumes */
6523 if (devlist == NULL)
6524 return 0;
6525
6526 count += active_arrays_by_format("imsm", hba_path, &devlist, dpa, verbose);
6527 dprintf("path: %s active arrays: %d\n", hba_path, count);
6528 if (devlist == NULL)
6529 return 0;
6530 do {
6531 found = 0;
6532 count += count_volumes_list(devlist,
6533 NULL,
6534 verbose,
6535 &found);
6536 dprintf("found %d count: %d\n", found, count);
6537 } while (found);
6538
6539 dprintf("path: %s total number of volumes: %d\n", hba_path, count);
6540
6541 while (devlist) {
6542 struct md_list *dv = devlist;
6543 devlist = devlist->next;
6544 free(dv->devname);
6545 free(dv);
6546 }
ca9de185
LM		 6547 }
6548 return count;
6549}
6550
cd9d1ac7
DW		 6551static int imsm_default_chunk(const struct imsm_orom *orom)
6552{
6553 /* up to 512 if the plaform supports it, otherwise the platform max.
6554 * 128 if no platform detected
6555 */
6556 int fs = max(7, orom ? fls(orom->sss) : 0);
6557
6558 return min(512, (1 << fs));
6559}
73408129 6560
6592ce37
DW		 6561static int
6562validate_geometry_imsm_orom(struct intel_super *super, int level, int layout,
2cc699af 6563 int raiddisks, int *chunk, unsigned long long size, int verbose)
6592ce37 6564{
660260d0
DW		 6565 /* check/set platform and metadata limits/defaults */
6566 if (super->orom && raiddisks > super->orom->dpa) {
676e87a8 6567 pr_vrb("platform supports a maximum of %d disks per array\n",
660260d0 6568 super->orom->dpa);
73408129
LM		 6569 return 0;
6570 }
6571
5d500228 6572 /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */
660260d0 6573 if (!is_raid_level_supported(super->orom, level, raiddisks)) {
676e87a8 6574 pr_vrb("platform does not support raid%d with %d disk%s\n",
6592ce37
DW		 6575 level, raiddisks, raiddisks > 1 ? "s" : "");
6576 return 0;
6577 }
cd9d1ac7 6578
7ccc4cc4 6579 if (*chunk == 0 || *chunk == UnSet)
cd9d1ac7
DW		 6580 *chunk = imsm_default_chunk(super->orom);
6581
7ccc4cc4 6582 if (super->orom && !imsm_orom_has_chunk(super->orom, *chunk)) {
676e87a8 6583 pr_vrb("platform does not support a chunk size of: %d\n", *chunk);
cd9d1ac7 6584 return 0;
6592ce37 6585 }
cd9d1ac7 6586
6592ce37
DW		 6587 if (layout != imsm_level_to_layout(level)) {
6588 if (level == 5)
676e87a8 6589 pr_vrb("imsm raid 5 only supports the left-asymmetric layout\n");
6592ce37 6590 else if (level == 10)
676e87a8 6591 pr_vrb("imsm raid 10 only supports the n2 layout\n");
6592ce37 6592 else
676e87a8 6593 pr_vrb("imsm unknown layout %#x for this raid level %d\n",
6592ce37
DW		 6594 layout, level);
6595 return 0;
6596 }
2cc699af 6597
7ccc4cc4 6598 if (super->orom && (super->orom->attr & IMSM_OROM_ATTR_2TB) == 0 &&
2cc699af 6599 (calc_array_size(level, raiddisks, layout, *chunk, size) >> 32) > 0) {
676e87a8 6600 pr_vrb("platform does not support a volume size over 2TB\n");
2cc699af
CA		 6601 return 0;
6602 }
614902f6 6603
6592ce37
DW		 6604 return 1;
6605}
6606
1011e834 6607/* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd
c2c087e6
DW		 6608 * FIX ME add ahci details
6609 */
8b353278 6610static int validate_geometry_imsm_volume(struct supertype *st, int level,
c21e737b 6611 int layout, int raiddisks, int *chunk,
af4348dd
N		 6612 unsigned long long size,
6613 unsigned long long data_offset,
6614 char *dev,
2c514b71
NB		 6615 unsigned long long *freesize,
6616 int verbose)
cdddbdbc 6617{
c2c087e6
DW		 6618 struct stat stb;
6619 struct intel_super *super = st->sb;
b2916f25 6620 struct imsm_super *mpb;
c2c087e6
DW		 6621 struct dl *dl;
6622 unsigned long long pos = 0;
6623 unsigned long long maxsize;
6624 struct extent *e;
6625 int i;
cdddbdbc 6626
88c32bb1
DW		 6627 /* We must have the container info already read in. */
6628 if (!super)
c2c087e6
DW		 6629 return 0;
6630
b2916f25
JS		 6631 mpb = super->anchor;
6632
2cc699af 6633 if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, size, verbose)) {
7a862a02 6634 pr_err("RAID gemetry validation failed. Cannot proceed with the action(s).\n");
c2c087e6 6635 return 0;
d54559f0 6636 }
c2c087e6
DW		 6637 if (!dev) {
6638 /* General test: make sure there is space for
2da8544a
DW		 6639 * 'raiddisks' device extents of size 'size' at a given
6640 * offset
c2c087e6 6641 */
e46273eb 6642 unsigned long long minsize = size;
b7528a20 6643 unsigned long long start_offset = MaxSector;
c2c087e6
DW		 6644 int dcnt = 0;
6645 if (minsize == 0)
6646 minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
6647 for (dl = super->disks; dl ; dl = dl->next) {
6648 int found = 0;
6649
bf5a934a 6650 pos = 0;
c2c087e6
DW		 6651 i = 0;
6652 e = get_extents(super, dl);
6653 if (!e) continue;
6654 do {
6655 unsigned long long esize;
6656 esize = e[i].start - pos;
6657 if (esize >= minsize)
6658 found = 1;
b7528a20 6659 if (found && start_offset == MaxSector) {
2da8544a
DW		 6660 start_offset = pos;
6661 break;
6662 } else if (found && pos != start_offset) {
6663 found = 0;
6664 break;
6665 }
c2c087e6
DW		 6666 pos = e[i].start + e[i].size;
6667 i++;
6668 } while (e[i-1].size);
6669 if (found)
6670 dcnt++;
6671 free(e);
6672 }
6673 if (dcnt < raiddisks) {
2c514b71 6674 if (verbose)
7a862a02 6675 pr_err("imsm: Not enough devices with space for this array (%d < %d)\n",
2c514b71 6676 dcnt, raiddisks);
c2c087e6
DW		 6677 return 0;
6678 }
6679 return 1;
6680 }
0dcecb2e 6681
c2c087e6
DW		 6682 /* This device must be a member of the set */
6683 if (stat(dev, &stb) < 0)
6684 return 0;
6685 if ((S_IFMT & stb.st_mode) != S_IFBLK)
6686 return 0;
6687 for (dl = super->disks ; dl ; dl = dl->next) {
f21e18ca
N		 6688 if (dl->major == (int)major(stb.st_rdev) &&
6689 dl->minor == (int)minor(stb.st_rdev))
c2c087e6
DW		 6690 break;
6691 }
6692 if (!dl) {
2c514b71 6693 if (verbose)
7a862a02 6694 pr_err("%s is not in the same imsm set\n", dev);
c2c087e6 6695 return 0;
a20d2ba5
DW		 6696 } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) {
6697 /* If a volume is present then the current creation attempt
6698 * cannot incorporate new spares because the orom may not
6699 * understand this configuration (all member disks must be
6700 * members of each array in the container).
6701 */
7a862a02
N		 6702 pr_err("%s is a spare and a volume is already defined for this container\n", dev);
6703 pr_err("The option-rom requires all member disks to be a member of all volumes\n");
a20d2ba5 6704 return 0;
5fe62b94
WD		 6705 } else if (super->orom && mpb->num_raid_devs > 0 &&
6706 mpb->num_disks != raiddisks) {
7a862a02 6707 pr_err("The option-rom requires all member disks to be a member of all volumes\n");
5fe62b94 6708 return 0;
c2c087e6 6709 }
0dcecb2e
DW		 6710
6711 /* retrieve the largest free space block */
c2c087e6
DW		 6712 e = get_extents(super, dl);
6713 maxsize = 0;
6714 i = 0;
0dcecb2e
DW		 6715 if (e) {
6716 do {
6717 unsigned long long esize;
6718
6719 esize = e[i].start - pos;
6720 if (esize >= maxsize)
6721 maxsize = esize;
6722 pos = e[i].start + e[i].size;
6723 i++;
6724 } while (e[i-1].size);
6725 dl->e = e;
6726 dl->extent_cnt = i;
6727 } else {
6728 if (verbose)
e7b84f9d 6729 pr_err("unable to determine free space for: %s\n",
0dcecb2e
DW		 6730 dev);
6731 return 0;
6732 }
6733 if (maxsize < size) {
6734 if (verbose)
e7b84f9d 6735 pr_err("%s not enough space (%llu < %llu)\n",
0dcecb2e
DW		 6736 dev, maxsize, size);
6737 return 0;
6738 }
6739
6740 /* count total number of extents for merge */
6741 i = 0;
6742 for (dl = super->disks; dl; dl = dl->next)
6743 if (dl->e)
6744 i += dl->extent_cnt;
6745
6746 maxsize = merge_extents(super, i);
3baa56ab
LO		 6747
6748 if (!check_env("IMSM_NO_PLATFORM") &&
6749 mpb->num_raid_devs > 0 && size && size != maxsize) {
7a862a02 6750 pr_err("attempting to create a second volume with size less then remaining space. Aborting...\n");
3baa56ab
LO		 6751 return 0;
6752 }
6753
a7dd165b 6754 if (maxsize < size || maxsize == 0) {
b3071342
LD		 6755 if (verbose) {
6756 if (maxsize == 0)
7a862a02 6757 pr_err("no free space left on device. Aborting...\n");
b3071342 6758 else
7a862a02 6759 pr_err("not enough space to create volume of given size (%llu < %llu). Aborting...\n",
b3071342
LD		 6760 maxsize, size);
6761 }
0dcecb2e 6762 return 0;
0dcecb2e
DW		 6763 }
6764
c2c087e6
DW		 6765 *freesize = maxsize;
6766
ca9de185 6767 if (super->orom) {
72a45777 6768 int count = count_volumes(super->hba,
ca9de185
LM		 6769 super->orom->dpa, verbose);
6770 if (super->orom->vphba <= count) {
676e87a8 6771 pr_vrb("platform does not support more than %d raid volumes.\n",
ca9de185
LM		 6772 super->orom->vphba);
6773 return 0;
6774 }
6775 }
c2c087e6 6776 return 1;
cdddbdbc
DW		 6777}
6778
13bcac90 6779static int imsm_get_free_size(struct supertype *st, int raiddisks,
efb30e7f
DW		 6780 unsigned long long size, int chunk,
6781 unsigned long long *freesize)
6782{
6783 struct intel_super *super = st->sb;
6784 struct imsm_super *mpb = super->anchor;
6785 struct dl *dl;
6786 int i;
6787 int extent_cnt;
6788 struct extent *e;
6789 unsigned long long maxsize;
6790 unsigned long long minsize;
6791 int cnt;
6792 int used;
6793
6794 /* find the largest common start free region of the possible disks */
6795 used = 0;
6796 extent_cnt = 0;
6797 cnt = 0;
6798 for (dl = super->disks; dl; dl = dl->next) {
6799 dl->raiddisk = -1;
6800
6801 if (dl->index >= 0)
6802 used++;
6803
6804 /* don't activate new spares if we are orom constrained
6805 * and there is already a volume active in the container
6806 */
6807 if (super->orom && dl->index < 0 && mpb->num_raid_devs)
6808 continue;
6809
6810 e = get_extents(super, dl);
6811 if (!e)
6812 continue;
6813 for (i = 1; e[i-1].size; i++)
6814 ;
6815 dl->e = e;
6816 dl->extent_cnt = i;
6817 extent_cnt += i;
6818 cnt++;
6819 }
6820
6821 maxsize = merge_extents(super, extent_cnt);
6822 minsize = size;
6823 if (size == 0)
612e59d8
CA		 6824 /* chunk is in K */
6825 minsize = chunk * 2;
efb30e7f
DW		 6826
6827 if (cnt < raiddisks ||
6828 (super->orom && used && used != raiddisks) ||
a7dd165b
DW		 6829 maxsize < minsize ||
6830 maxsize == 0) {
e7b84f9d 6831 pr_err("not enough devices with space to create array.\n");
efb30e7f
DW		 6832 return 0; /* No enough free spaces large enough */
6833 }
6834
6835 if (size == 0) {
6836 size = maxsize;
6837 if (chunk) {
612e59d8
CA		 6838 size /= 2 * chunk;
6839 size *= 2 * chunk;
efb30e7f 6840 }
f878b242
LM		 6841 maxsize = size;
6842 }
6843 if (!check_env("IMSM_NO_PLATFORM") &&
6844 mpb->num_raid_devs > 0 && size && size != maxsize) {
7a862a02 6845 pr_err("attempting to create a second volume with size less then remaining space. Aborting...\n");
f878b242 6846 return 0;
efb30e7f 6847 }
efb30e7f
DW		 6848 cnt = 0;
6849 for (dl = super->disks; dl; dl = dl->next)
6850 if (dl->e)
6851 dl->raiddisk = cnt++;
6852
6853 *freesize = size;
6854
13bcac90
AK		 6855 dprintf("imsm: imsm_get_free_size() returns : %llu\n", size);
6856
efb30e7f
DW		 6857 return 1;
6858}
6859
13bcac90
AK		 6860static int reserve_space(struct supertype *st, int raiddisks,
6861 unsigned long long size, int chunk,
6862 unsigned long long *freesize)
6863{
6864 struct intel_super *super = st->sb;
6865 struct dl *dl;
6866 int cnt;
6867 int rv = 0;
6868
6869 rv = imsm_get_free_size(st, raiddisks, size, chunk, freesize);
6870 if (rv) {
6871 cnt = 0;
6872 for (dl = super->disks; dl; dl = dl->next)
6873 if (dl->e)
6874 dl->raiddisk = cnt++;
6875 rv = 1;
6876 }
6877
6878 return rv;
6879}
6880
bf5a934a 6881static int validate_geometry_imsm(struct supertype *st, int level, int layout,
c21e737b 6882 int raiddisks, int *chunk, unsigned long long size,
af4348dd 6883 unsigned long long data_offset,
bf5a934a
DW		 6884 char *dev, unsigned long long *freesize,
6885 int verbose)
6886{
6887 int fd, cfd;
6888 struct mdinfo *sra;
20cbe8d2 6889 int is_member = 0;
bf5a934a 6890
d54559f0
LM		 6891 /* load capability
6892 * if given unused devices create a container
bf5a934a
DW		 6893 * if given given devices in a container create a member volume
6894 */
6895 if (level == LEVEL_CONTAINER) {
6896 /* Must be a fresh device to add to a container */
6897 return validate_geometry_imsm_container(st, level, layout,
c21e737b 6898 raiddisks,
7ccc4cc4 6899 *chunk,
af4348dd 6900 size, data_offset,
bf5a934a
DW		 6901 dev, freesize,
6902 verbose);
6903 }
9587c373 6904
8592f29d 6905 if (!dev) {
e91a3bad 6906 if (st->sb) {
ca9de185 6907 struct intel_super *super = st->sb;
e91a3bad 6908 if (!validate_geometry_imsm_orom(st->sb, level, layout,
2cc699af 6909 raiddisks, chunk, size,
e91a3bad
LM		 6910 verbose))
6911 return 0;
efb30e7f
DW		 6912 /* we are being asked to automatically layout a
6913 * new volume based on the current contents of
6914 * the container. If the the parameters can be
6915 * satisfied reserve_space will record the disks,
6916 * start offset, and size of the volume to be
6917 * created. add_to_super and getinfo_super
6918 * detect when autolayout is in progress.
6919 */
ca9de185
LM		 6920 /* assuming that freesize is always given when array is
6921 created */
6922 if (super->orom && freesize) {
6923 int count;
72a45777 6924 count = count_volumes(super->hba,
ca9de185
LM		 6925 super->orom->dpa, verbose);
6926 if (super->orom->vphba <= count) {
676e87a8 6927 pr_vrb("platform does not support more than %d raid volumes.\n",
ca9de185
LM		 6928 super->orom->vphba);
6929 return 0;
6930 }
6931 }
e91a3bad
LM		 6932 if (freesize)
6933 return reserve_space(st, raiddisks, size,
7ccc4cc4 6934 *chunk, freesize);
8592f29d
N		 6935 }
6936 return 1;
6937 }
bf5a934a
DW		 6938 if (st->sb) {
6939 /* creating in a given container */
6940 return validate_geometry_imsm_volume(st, level, layout,
6941 raiddisks, chunk, size,
af4348dd 6942 data_offset,
bf5a934a
DW		 6943 dev, freesize, verbose);
6944 }
6945
bf5a934a
DW		 6946 /* This device needs to be a device in an 'imsm' container */
6947 fd = open(dev, O_RDONLY|O_EXCL, 0);
6948 if (fd >= 0) {
6949 if (verbose)
e7b84f9d
N		 6950 pr_err("Cannot create this array on device %s\n",
6951 dev);
bf5a934a
DW		 6952 close(fd);
6953 return 0;
6954 }
6955 if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) {
6956 if (verbose)
e7b84f9d 6957 pr_err("Cannot open %s: %s\n",
bf5a934a
DW		 6958 dev, strerror(errno));
6959 return 0;
6960 }
6961 /* Well, it is in use by someone, maybe an 'imsm' container. */
6962 cfd = open_container(fd);
20cbe8d2 6963 close(fd);
bf5a934a 6964 if (cfd < 0) {
bf5a934a 6965 if (verbose)
e7b84f9d 6966 pr_err("Cannot use %s: It is busy\n",
bf5a934a
DW		 6967 dev);
6968 return 0;
6969 }
4dd2df09 6970 sra = sysfs_read(cfd, NULL, GET_VERSION);
bf5a934a 6971 if (sra && sra->array.major_version == -1 &&
20cbe8d2
AW		 6972 strcmp(sra->text_version, "imsm") == 0)
6973 is_member = 1;
6974 sysfs_free(sra);
6975 if (is_member) {
bf5a934a
DW		 6976 /* This is a member of a imsm container. Load the container
6977 * and try to create a volume
6978 */
6979 struct intel_super *super;
6980
ec50f7b6 6981 if (load_super_imsm_all(st, cfd, (void **) &super, NULL, NULL, 1) == 0) {
bf5a934a 6982 st->sb = super;
4dd2df09 6983 strcpy(st->container_devnm, fd2devnm(cfd));
bf5a934a
DW		 6984 close(cfd);
6985 return validate_geometry_imsm_volume(st, level, layout,
6986 raiddisks, chunk,
af4348dd 6987 size, data_offset, dev,
ecbd9e81
N		 6988 freesize, 1)
6989 ? 1 : -1;
bf5a934a 6990 }
20cbe8d2 6991 }
bf5a934a 6992
20cbe8d2 6993 if (verbose)
e7b84f9d 6994 pr_err("failed container membership check\n");
20cbe8d2
AW		 6995
6996 close(cfd);
6997 return 0;
bf5a934a 6998}
0bd16cf2 6999
30f58b22 7000static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk)
0bd16cf2
DJ		 7001{
7002 struct intel_super *super = st->sb;
7003
30f58b22
DW		 7004 if (level && *level == UnSet)
7005 *level = LEVEL_CONTAINER;
7006
7007 if (level && layout && *layout == UnSet)
7008 *layout = imsm_level_to_layout(*level);
0bd16cf2 7009
cd9d1ac7
DW		 7010 if (chunk && (*chunk == UnSet || *chunk == 0))
7011 *chunk = imsm_default_chunk(super->orom);
0bd16cf2
DJ		 7012}
7013
33414a01
DW		 7014static void handle_missing(struct intel_super *super, struct imsm_dev *dev);
7015
7016static int kill_subarray_imsm(struct supertype *st)
7017{
7018 /* remove the subarray currently referenced by ->current_vol */
7019 __u8 i;
7020 struct intel_dev **dp;
7021 struct intel_super *super = st->sb;
7022 __u8 current_vol = super->current_vol;
7023 struct imsm_super *mpb = super->anchor;
7024
7025 if (super->current_vol < 0)
7026 return 2;
7027 super->current_vol = -1; /* invalidate subarray cursor */
7028
7029 /* block deletions that would change the uuid of active subarrays
7030 *
7031 * FIXME when immutable ids are available, but note that we'll
7032 * also need to fixup the invalidated/active subarray indexes in
7033 * mdstat
7034 */
7035 for (i = 0; i < mpb->num_raid_devs; i++) {
7036 char subarray[4];
7037
7038 if (i < current_vol)
7039 continue;
7040 sprintf(subarray, "%u", i);
4dd2df09 7041 if (is_subarray_active(subarray, st->devnm)) {
e7b84f9d
N		 7042 pr_err("deleting subarray-%d would change the UUID of active subarray-%d, aborting\n",
7043 current_vol, i);
33414a01
DW		 7044
7045 return 2;
7046 }
7047 }
7048
7049 if (st->update_tail) {
503975b9 7050 struct imsm_update_kill_array *u = xmalloc(sizeof(*u));
33414a01 7051
33414a01
DW		 7052 u->type = update_kill_array;
7053 u->dev_idx = current_vol;
7054 append_metadata_update(st, u, sizeof(*u));
7055
7056 return 0;
7057 }
7058
7059 for (dp = &super->devlist; *dp;)
7060 if ((*dp)->index == current_vol) {
7061 *dp = (*dp)->next;
7062 } else {
7063 handle_missing(super, (*dp)->dev);
7064 if ((*dp)->index > current_vol)
7065 (*dp)->index--;
7066 dp = &(*dp)->next;
7067 }
7068
7069 /* no more raid devices, all active components are now spares,
7070 * but of course failed are still failed
7071 */
7072 if (--mpb->num_raid_devs == 0) {
7073 struct dl *d;
7074
7075 for (d = super->disks; d; d = d->next)
a8619d23
AK		 7076 if (d->index > -2)
7077 mark_spare(d);
33414a01
DW		 7078 }
7079
7080 super->updates_pending++;
7081
7082 return 0;
7083}
aa534678 7084
a951a4f7 7085static int update_subarray_imsm(struct supertype *st, char *subarray,
fa56eddb 7086 char *update, struct mddev_ident *ident)
aa534678
DW		 7087{
7088 /* update the subarray currently referenced by ->current_vol */
7089 struct intel_super *super = st->sb;
7090 struct imsm_super *mpb = super->anchor;
7091
aa534678
DW		 7092 if (strcmp(update, "name") == 0) {
7093 char *name = ident->name;
a951a4f7
N		 7094 char *ep;
7095 int vol;
aa534678 7096
4dd2df09 7097 if (is_subarray_active(subarray, st->devnm)) {
e7b84f9d 7098 pr_err("Unable to update name of active subarray\n");
aa534678
DW		 7099 return 2;
7100 }
7101
7102 if (!check_name(super, name, 0))
7103 return 2;
7104
a951a4f7
N		 7105 vol = strtoul(subarray, &ep, 10);
7106 if (*ep != '\0' || vol >= super->anchor->num_raid_devs)
7107 return 2;
7108
aa534678 7109 if (st->update_tail) {
503975b9 7110 struct imsm_update_rename_array *u = xmalloc(sizeof(*u));
aa534678 7111
aa534678 7112 u->type = update_rename_array;
a951a4f7 7113 u->dev_idx = vol;
aa534678
DW		 7114 snprintf((char *) u->name, MAX_RAID_SERIAL_LEN, "%s", name);
7115 append_metadata_update(st, u, sizeof(*u));
7116 } else {
7117 struct imsm_dev *dev;
7118 int i;
7119
a951a4f7 7120 dev = get_imsm_dev(super, vol);
aa534678
DW		 7121 snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name);
7122 for (i = 0; i < mpb->num_raid_devs; i++) {
7123 dev = get_imsm_dev(super, i);
7124 handle_missing(super, dev);
7125 }
7126 super->updates_pending++;
7127 }
7128 } else
7129 return 2;
7130
7131 return 0;
7132}
d1e02575 7133#endif /* MDASSEMBLE */
bf5a934a 7134
28bce06f
AK		 7135static int is_gen_migration(struct imsm_dev *dev)
7136{
7534230b
AK		 7137 if (dev == NULL)
7138 return 0;
7139
28bce06f
AK		 7140 if (!dev->vol.migr_state)
7141 return 0;
7142
7143 if (migr_type(dev) == MIGR_GEN_MIGR)
7144 return 1;
7145
7146 return 0;
7147}
7148
1e5c6983
DW		 7149static int is_rebuilding(struct imsm_dev *dev)
7150{
7151 struct imsm_map *migr_map;
7152
7153 if (!dev->vol.migr_state)
7154 return 0;
7155
7156 if (migr_type(dev) != MIGR_REBUILD)
7157 return 0;
7158
238c0a71 7159 migr_map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 7160
7161 if (migr_map->map_state == IMSM_T_STATE_DEGRADED)
7162 return 1;
7163 else
7164 return 0;
7165}
7166
b4ab44d8 7167#ifndef MDASSEMBLE
6ce1fbf1
AK		 7168static int is_initializing(struct imsm_dev *dev)
7169{
7170 struct imsm_map *migr_map;
7171
7172 if (!dev->vol.migr_state)
7173 return 0;
7174
7175 if (migr_type(dev) != MIGR_INIT)
7176 return 0;
7177
238c0a71 7178 migr_map = get_imsm_map(dev, MAP_1);
6ce1fbf1
AK		 7179
7180 if (migr_map->map_state == IMSM_T_STATE_UNINITIALIZED)
7181 return 1;
7182
7183 return 0;
6ce1fbf1 7184}
b4ab44d8 7185#endif
6ce1fbf1 7186
c47b0ff6
AK		 7187static void update_recovery_start(struct intel_super *super,
7188 struct imsm_dev *dev,
7189 struct mdinfo *array)
1e5c6983
DW		 7190{
7191 struct mdinfo *rebuild = NULL;
7192 struct mdinfo *d;
7193 __u32 units;
7194
7195 if (!is_rebuilding(dev))
7196 return;
7197
7198 /* Find the rebuild target, but punt on the dual rebuild case */
7199 for (d = array->devs; d; d = d->next)
7200 if (d->recovery_start == 0) {
7201 if (rebuild)
7202 return;
7203 rebuild = d;
7204 }
7205
4363fd80
DW		 7206 if (!rebuild) {
7207 /* (?) none of the disks are marked with
7208 * IMSM_ORD_REBUILD, so assume they are missing and the
7209 * disk_ord_tbl was not correctly updated
7210 */
1ade5cc1 7211 dprintf("failed to locate out-of-sync disk\n");
4363fd80
DW		 7212 return;
7213 }
7214
1e5c6983 7215 units = __le32_to_cpu(dev->vol.curr_migr_unit);
c47b0ff6 7216 rebuild->recovery_start = units * blocks_per_migr_unit(super, dev);
1e5c6983
DW		 7217}
7218
9e2d750d 7219#ifndef MDASSEMBLE
276d77db 7220static int recover_backup_imsm(struct supertype *st, struct mdinfo *info);
9e2d750d 7221#endif
1e5c6983 7222
00bbdbda 7223static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray)
cdddbdbc 7224{
4f5bc454
DW		 7225 /* Given a container loaded by load_super_imsm_all,
7226 * extract information about all the arrays into
7227 * an mdinfo tree.
00bbdbda 7228 * If 'subarray' is given, just extract info about that array.
4f5bc454
DW		 7229 *
7230 * For each imsm_dev create an mdinfo, fill it in,
7231 * then look for matching devices in super->disks
7232 * and create appropriate device mdinfo.
7233 */
7234 struct intel_super *super = st->sb;
949c47a0 7235 struct imsm_super *mpb = super->anchor;
4f5bc454 7236 struct mdinfo *rest = NULL;
00bbdbda 7237 unsigned int i;
81219e70 7238 int sb_errors = 0;
abef11a3
AK		 7239 struct dl *d;
7240 int spare_disks = 0;
cdddbdbc 7241
19482bcc
AK		 7242 /* do not assemble arrays when not all attributes are supported */
7243 if (imsm_check_attributes(mpb->attributes) == 0) {
81219e70 7244 sb_errors = 1;
7a862a02 7245 pr_err("Unsupported attributes in IMSM metadata.Arrays activation is blocked.\n");
19482bcc
AK		 7246 }
7247
abef11a3
AK		 7248 /* count spare devices, not used in maps
7249 */
7250 for (d = super->disks; d; d = d->next)
7251 if (d->index == -1)
7252 spare_disks++;
7253
4f5bc454 7254 for (i = 0; i < mpb->num_raid_devs; i++) {
00bbdbda
N		 7255 struct imsm_dev *dev;
7256 struct imsm_map *map;
86e3692b 7257 struct imsm_map *map2;
4f5bc454 7258 struct mdinfo *this;
a6482415
N		 7259 int slot;
7260#ifndef MDASSEMBLE
7261 int chunk;
7262#endif
00bbdbda
N		 7263 char *ep;
7264
7265 if (subarray &&
7266 (i != strtoul(subarray, &ep, 10) || *ep != '\0'))
7267 continue;
7268
7269 dev = get_imsm_dev(super, i);
238c0a71
AK		 7270 map = get_imsm_map(dev, MAP_0);
7271 map2 = get_imsm_map(dev, MAP_1);
4f5bc454 7272
1ce0101c
DW		 7273 /* do not publish arrays that are in the middle of an
7274 * unsupported migration
7275 */
7276 if (dev->vol.migr_state &&
28bce06f 7277 (migr_type(dev) == MIGR_STATE_CHANGE)) {
7a862a02 7278 pr_err("cannot assemble volume '%.16s': unsupported migration in progress\n",
1ce0101c
DW		 7279 dev->volume);
7280 continue;
7281 }
2db86302
LM		 7282 /* do not publish arrays that are not support by controller's
7283 * OROM/EFI
7284 */
1ce0101c 7285
503975b9 7286 this = xmalloc(sizeof(*this));
4f5bc454 7287
301406c9 7288 super->current_vol = i;
a5d85af7 7289 getinfo_super_imsm_volume(st, this, NULL);
9894ec0d 7290 this->next = rest;
81219e70 7291#ifndef MDASSEMBLE
a6482415 7292 chunk = __le16_to_cpu(map->blocks_per_strip) >> 1;
81219e70
LM		 7293 /* mdadm does not support all metadata features- set the bit in all arrays state */
7294 if (!validate_geometry_imsm_orom(super,
7295 get_imsm_raid_level(map), /* RAID level */
7296 imsm_level_to_layout(get_imsm_raid_level(map)),
7297 map->num_members, /* raid disks */
2cc699af 7298 &chunk, join_u32(dev->size_low, dev->size_high),
81219e70 7299 1 /* verbose */)) {
7a862a02 7300 pr_err("IMSM RAID geometry validation failed. Array %s activation is blocked.\n",
81219e70
LM		 7301 dev->volume);
7302 this->array.state |=
7303 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
7304 (1<<MD_SB_BLOCK_VOLUME);
7305 }
7306#endif
7307
7308 /* if array has bad blocks, set suitable bit in all arrays state */
7309 if (sb_errors)
7310 this->array.state |=
7311 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
7312 (1<<MD_SB_BLOCK_VOLUME);
7313
4f5bc454 7314 for (slot = 0 ; slot < map->num_members; slot++) {
1e5c6983 7315 unsigned long long recovery_start;
4f5bc454
DW		 7316 struct mdinfo *info_d;
7317 struct dl *d;
7318 int idx;
9a1608e5 7319 int skip;
7eef0453 7320 __u32 ord;
4f5bc454 7321
9a1608e5 7322 skip = 0;
238c0a71
AK		 7323 idx = get_imsm_disk_idx(dev, slot, MAP_0);
7324 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
4f5bc454
DW		 7325 for (d = super->disks; d ; d = d->next)
7326 if (d->index == idx)
0fbd635c 7327 break;
4f5bc454 7328
1e5c6983 7329 recovery_start = MaxSector;
4f5bc454 7330 if (d == NULL)
9a1608e5 7331 skip = 1;
25ed7e59 7332 if (d && is_failed(&d->disk))
9a1608e5 7333 skip = 1;
7eef0453 7334 if (ord & IMSM_ORD_REBUILD)
1e5c6983 7335 recovery_start = 0;
9a1608e5 7336
1011e834 7337 /*
9a1608e5 7338 * if we skip some disks the array will be assmebled degraded;
1e5c6983
DW		 7339 * reset resync start to avoid a dirty-degraded
7340 * situation when performing the intial sync
9a1608e5
DW		 7341 *
7342 * FIXME handle dirty degraded
7343 */
1e5c6983 7344 if ((skip || recovery_start == 0) && !dev->vol.dirty)
b7528a20 7345 this->resync_start = MaxSector;
9a1608e5
DW		 7346 if (skip)
7347 continue;
4f5bc454 7348
503975b9 7349 info_d = xcalloc(1, sizeof(*info_d));
4f5bc454
DW		 7350 info_d->next = this->devs;
7351 this->devs = info_d;
7352
4f5bc454
DW		 7353 info_d->disk.number = d->index;
7354 info_d->disk.major = d->major;
7355 info_d->disk.minor = d->minor;
7356 info_d->disk.raid_disk = slot;
1e5c6983 7357 info_d->recovery_start = recovery_start;
86e3692b
AK		 7358 if (map2) {
7359 if (slot < map2->num_members)
7360 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK		 7361 else
7362 this->array.spare_disks++;
86e3692b
AK		 7363 } else {
7364 if (slot < map->num_members)
7365 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK		 7366 else
7367 this->array.spare_disks++;
86e3692b 7368 }
1e5c6983
DW		 7369 if (info_d->recovery_start == MaxSector)
7370 this->array.working_disks++;
4f5bc454
DW		 7371
7372 info_d->events = __le32_to_cpu(mpb->generation_num);
5551b113 7373 info_d->data_offset = pba_of_lba0(map);
06fb291a
PB		 7374
7375 if (map->raid_level == 5) {
7376 info_d->component_size =
7377 num_data_stripes(map) *
7378 map->blocks_per_strip;
7379 } else {
7380 info_d->component_size = blocks_per_member(map);
7381 }
b12796be
TM		 7382
7383 info_d->bb.supported = 0;
7384 get_volume_badblocks(super->bbm_log, ord_to_idx(ord),
7385 info_d->data_offset,
7386 info_d->component_size,
7387 &info_d->bb);
4f5bc454 7388 }
1e5c6983 7389 /* now that the disk list is up-to-date fixup recovery_start */
c47b0ff6 7390 update_recovery_start(super, dev, this);
abef11a3 7391 this->array.spare_disks += spare_disks;
276d77db 7392
9e2d750d 7393#ifndef MDASSEMBLE
276d77db
AK		 7394 /* check for reshape */
7395 if (this->reshape_active == 1)
7396 recover_backup_imsm(st, this);
9e2d750d 7397#endif
9a1608e5 7398 rest = this;
4f5bc454
DW		 7399 }
7400
7401 return rest;
cdddbdbc
DW		 7402}
7403
3b451610
AK		 7404static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev,
7405 int failed, int look_in_map)
c2a1e7da 7406{
3b451610
AK		 7407 struct imsm_map *map;
7408
7409 map = get_imsm_map(dev, look_in_map);
c2a1e7da
DW		 7410
7411 if (!failed)
1011e834 7412 return map->map_state == IMSM_T_STATE_UNINITIALIZED ?
3393c6af 7413 IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL;
c2a1e7da
DW		 7414
7415 switch (get_imsm_raid_level(map)) {
7416 case 0:
7417 return IMSM_T_STATE_FAILED;
7418 break;
7419 case 1:
7420 if (failed < map->num_members)
7421 return IMSM_T_STATE_DEGRADED;
7422 else
7423 return IMSM_T_STATE_FAILED;
7424 break;
7425 case 10:
7426 {
7427 /**
c92a2527
DW		 7428 * check to see if any mirrors have failed, otherwise we
7429 * are degraded. Even numbered slots are mirrored on
7430 * slot+1
c2a1e7da 7431 */
c2a1e7da 7432 int i;
d9b420a5
N		 7433 /* gcc -Os complains that this is unused */
7434 int insync = insync;
c2a1e7da
DW		 7435
7436 for (i = 0; i < map->num_members; i++) {
238c0a71 7437 __u32 ord = get_imsm_ord_tbl_ent(dev, i, MAP_X);
c92a2527
DW		 7438 int idx = ord_to_idx(ord);
7439 struct imsm_disk *disk;
c2a1e7da 7440
c92a2527 7441 /* reset the potential in-sync count on even-numbered
1011e834 7442 * slots. num_copies is always 2 for imsm raid10
c92a2527
DW		 7443 */
7444 if ((i & 1) == 0)
7445 insync = 2;
c2a1e7da 7446
c92a2527 7447 disk = get_imsm_disk(super, idx);
25ed7e59 7448 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
c92a2527 7449 insync--;
c2a1e7da 7450
c92a2527
DW		 7451 /* no in-sync disks left in this mirror the
7452 * array has failed
7453 */
7454 if (insync == 0)
7455 return IMSM_T_STATE_FAILED;
c2a1e7da
DW		 7456 }
7457
7458 return IMSM_T_STATE_DEGRADED;
7459 }
7460 case 5:
7461 if (failed < 2)
7462 return IMSM_T_STATE_DEGRADED;
7463 else
7464 return IMSM_T_STATE_FAILED;
7465 break;
7466 default:
7467 break;
7468 }
7469
7470 return map->map_state;
7471}
7472
3b451610
AK		 7473static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev,
7474 int look_in_map)
c2a1e7da
DW		 7475{
7476 int i;
7477 int failed = 0;
7478 struct imsm_disk *disk;
d5985138
AK		 7479 struct imsm_map *map = get_imsm_map(dev, MAP_0);
7480 struct imsm_map *prev = get_imsm_map(dev, MAP_1);
68fe4598 7481 struct imsm_map *map_for_loop;
0556e1a2
DW		 7482 __u32 ord;
7483 int idx;
d5985138 7484 int idx_1;
c2a1e7da 7485
0556e1a2
DW		 7486 /* at the beginning of migration we set IMSM_ORD_REBUILD on
7487 * disks that are being rebuilt. New failures are recorded to
7488 * map[0]. So we look through all the disks we started with and
7489 * see if any failures are still present, or if any new ones
7490 * have arrived
0556e1a2 7491 */
d5985138
AK		 7492 map_for_loop = map;
7493 if (prev && (map->num_members < prev->num_members))
7494 map_for_loop = prev;
68fe4598
LD		 7495
7496 for (i = 0; i < map_for_loop->num_members; i++) {
d5985138 7497 idx_1 = -255;
238c0a71
AK		 7498 /* when MAP_X is passed both maps failures are counted
7499 */
d5985138 7500 if (prev &&
089f9d79
JS		 7501 (look_in_map == MAP_1 || look_in_map == MAP_X) &&
7502 i < prev->num_members) {
d5985138
AK		 7503 ord = __le32_to_cpu(prev->disk_ord_tbl[i]);
7504 idx_1 = ord_to_idx(ord);
c2a1e7da 7505
d5985138
AK		 7506 disk = get_imsm_disk(super, idx_1);
7507 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
7508 failed++;
7509 }
089f9d79
JS		 7510 if ((look_in_map == MAP_0 || look_in_map == MAP_X) &&
7511 i < map->num_members) {
d5985138
AK		 7512 ord = __le32_to_cpu(map->disk_ord_tbl[i]);
7513 idx = ord_to_idx(ord);
7514
7515 if (idx != idx_1) {
7516 disk = get_imsm_disk(super, idx);
7517 if (!disk || is_failed(disk) ||
7518 ord & IMSM_ORD_REBUILD)
7519 failed++;
7520 }
7521 }
c2a1e7da
DW		 7522 }
7523
7524 return failed;
845dea95
NB		 7525}
7526
97b4d0e9
DW		 7527#ifndef MDASSEMBLE
7528static int imsm_open_new(struct supertype *c, struct active_array *a,
7529 char *inst)
7530{
7531 struct intel_super *super = c->sb;
7532 struct imsm_super *mpb = super->anchor;
bbab0940 7533 struct imsm_update_prealloc_bb_mem u;
9587c373 7534
97b4d0e9 7535 if (atoi(inst) >= mpb->num_raid_devs) {
1ade5cc1 7536 pr_err("subarry index %d, out of range\n", atoi(inst));
97b4d0e9
DW		 7537 return -ENODEV;
7538 }
7539
7540 dprintf("imsm: open_new %s\n", inst);
7541 a->info.container_member = atoi(inst);
bbab0940
TM		 7542
7543 u.type = update_prealloc_badblocks_mem;
7544 imsm_update_metadata_locally(c, &u, sizeof(u));
7545
97b4d0e9
DW		 7546 return 0;
7547}
7548
0c046afd
DW		 7549static int is_resyncing(struct imsm_dev *dev)
7550{
7551 struct imsm_map *migr_map;
7552
7553 if (!dev->vol.migr_state)
7554 return 0;
7555
1484e727
DW		 7556 if (migr_type(dev) == MIGR_INIT ||
7557 migr_type(dev) == MIGR_REPAIR)
0c046afd
DW		 7558 return 1;
7559
4c9bc37b
AK		 7560 if (migr_type(dev) == MIGR_GEN_MIGR)
7561 return 0;
7562
238c0a71 7563 migr_map = get_imsm_map(dev, MAP_1);
0c046afd 7564
089f9d79
JS		 7565 if (migr_map->map_state == IMSM_T_STATE_NORMAL &&
7566 dev->vol.migr_type != MIGR_GEN_MIGR)
0c046afd
DW		 7567 return 1;
7568 else
7569 return 0;
7570}
7571
0556e1a2 7572/* return true if we recorded new information */
4c9e8c1e
TM		 7573static int mark_failure(struct intel_super *super,
7574 struct imsm_dev *dev, struct imsm_disk *disk, int idx)
47ee5a45 7575{
0556e1a2
DW		 7576 __u32 ord;
7577 int slot;
7578 struct imsm_map *map;
86c54047
DW		 7579 char buf[MAX_RAID_SERIAL_LEN+3];
7580 unsigned int len, shift = 0;
0556e1a2
DW		 7581
7582 /* new failures are always set in map[0] */
238c0a71 7583 map = get_imsm_map(dev, MAP_0);
0556e1a2
DW		 7584
7585 slot = get_imsm_disk_slot(map, idx);
7586 if (slot < 0)
7587 return 0;
7588
7589 ord = __le32_to_cpu(map->disk_ord_tbl[slot]);
25ed7e59 7590 if (is_failed(disk) && (ord & IMSM_ORD_REBUILD))
0556e1a2
DW		 7591 return 0;
7592
7d0c5e24
LD		 7593 memcpy(buf, disk->serial, MAX_RAID_SERIAL_LEN);
7594 buf[MAX_RAID_SERIAL_LEN] = '\000';
7595 strcat(buf, ":0");
86c54047
DW		 7596 if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN)
7597 shift = len - MAX_RAID_SERIAL_LEN + 1;
7598 strncpy((char *)disk->serial, &buf[shift], MAX_RAID_SERIAL_LEN);
7599
f2f27e63 7600 disk->status |= FAILED_DISK;
0556e1a2 7601 set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD);
17788994
AK		 7602 /* mark failures in second map if second map exists and this disk
7603 * in this slot.
7604 * This is valid for migration, initialization and rebuild
7605 */
7606 if (dev->vol.migr_state) {
238c0a71 7607 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
0a108d63
AK		 7608 int slot2 = get_imsm_disk_slot(map2, idx);
7609
089f9d79 7610 if (slot2 < map2->num_members && slot2 >= 0)
0a108d63 7611 set_imsm_ord_tbl_ent(map2, slot2,
1ace8403
AK		 7612 idx | IMSM_ORD_REBUILD);
7613 }
f21e18ca 7614 if (map->failed_disk_num == 0xff)
0556e1a2 7615 map->failed_disk_num = slot;
4c9e8c1e
TM		 7616
7617 clear_disk_badblocks(super->bbm_log, ord_to_idx(ord));
7618
0556e1a2
DW		 7619 return 1;
7620}
7621
4c9e8c1e
TM		 7622static void mark_missing(struct intel_super *super,
7623 struct imsm_dev *dev, struct imsm_disk *disk, int idx)
0556e1a2 7624{
4c9e8c1e 7625 mark_failure(super, dev, disk, idx);
0556e1a2
DW		 7626
7627 if (disk->scsi_id == __cpu_to_le32(~(__u32)0))
7628 return;
7629
47ee5a45
DW		 7630 disk->scsi_id = __cpu_to_le32(~(__u32)0);
7631 memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1);
7632}
7633
33414a01
DW		 7634static void handle_missing(struct intel_super *super, struct imsm_dev *dev)
7635{
33414a01 7636 struct dl *dl;
33414a01
DW		 7637
7638 if (!super->missing)
7639 return;
33414a01 7640
79b68f1b
PC		 7641 /* When orom adds replacement for missing disk it does
7642 * not remove entry of missing disk, but just updates map with
7643 * new added disk. So it is not enough just to test if there is
7644 * any missing disk, we have to look if there are any failed disks
7645 * in map to stop migration */
7646
33414a01 7647 dprintf("imsm: mark missing\n");
3d59f0c0
AK		 7648 /* end process for initialization and rebuild only
7649 */
7650 if (is_gen_migration(dev) == 0) {
7651 __u8 map_state;
7652 int failed;
7653
7654 failed = imsm_count_failed(super, dev, MAP_0);
7655 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
7656
79b68f1b
PC		 7657 if (failed)
7658 end_migration(dev, super, map_state);
3d59f0c0 7659 }
33414a01 7660 for (dl = super->missing; dl; dl = dl->next)
4c9e8c1e 7661 mark_missing(super, dev, &dl->disk, dl->index);
33414a01
DW		 7662 super->updates_pending++;
7663}
7664
f3871fdc
AK		 7665static unsigned long long imsm_set_array_size(struct imsm_dev *dev,
7666 long long new_size)
70bdf0dc 7667{
238c0a71 7668 int used_disks = imsm_num_data_members(dev, MAP_0);
70bdf0dc
AK		 7669 unsigned long long array_blocks;
7670 struct imsm_map *map;
7671
7672 if (used_disks == 0) {
7673 /* when problems occures
7674 * return current array_blocks value
7675 */
7676 array_blocks = __le32_to_cpu(dev->size_high);
7677 array_blocks = array_blocks << 32;
7678 array_blocks += __le32_to_cpu(dev->size_low);
7679
7680 return array_blocks;
7681 }
7682
7683 /* set array size in metadata
7684 */
f3871fdc
AK		 7685 if (new_size <= 0) {
7686 /* OLCE size change is caused by added disks
7687 */
7688 map = get_imsm_map(dev, MAP_0);
7689 array_blocks = blocks_per_member(map) * used_disks;
7690 } else {
7691 /* Online Volume Size Change
7692 * Using available free space
7693 */
7694 array_blocks = new_size;
7695 }
70bdf0dc
AK		 7696
7697 /* round array size down to closest MB
7698 */
7699 array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT;
7700 dev->size_low = __cpu_to_le32((__u32)array_blocks);
7701 dev->size_high = __cpu_to_le32((__u32)(array_blocks >> 32));
7702
7703 return array_blocks;
7704}
7705
28bce06f
AK		 7706static void imsm_set_disk(struct active_array *a, int n, int state);
7707
0e2d1a4e
AK		 7708static void imsm_progress_container_reshape(struct intel_super *super)
7709{
7710 /* if no device has a migr_state, but some device has a
7711 * different number of members than the previous device, start
7712 * changing the number of devices in this device to match
7713 * previous.
7714 */
7715 struct imsm_super *mpb = super->anchor;
7716 int prev_disks = -1;
7717 int i;
1dfaa380 7718 int copy_map_size;
0e2d1a4e
AK		 7719
7720 for (i = 0; i < mpb->num_raid_devs; i++) {
7721 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 7722 struct imsm_map *map = get_imsm_map(dev, MAP_0);
0e2d1a4e
AK		 7723 struct imsm_map *map2;
7724 int prev_num_members;
0e2d1a4e
AK		 7725
7726 if (dev->vol.migr_state)
7727 return;
7728
7729 if (prev_disks == -1)
7730 prev_disks = map->num_members;
7731 if (prev_disks == map->num_members)
7732 continue;
7733
7734 /* OK, this array needs to enter reshape mode.
7735 * i.e it needs a migr_state
7736 */
7737
1dfaa380 7738 copy_map_size = sizeof_imsm_map(map);
0e2d1a4e
AK		 7739 prev_num_members = map->num_members;
7740 map->num_members = prev_disks;
7741 dev->vol.migr_state = 1;
7742 dev->vol.curr_migr_unit = 0;
ea672ee1 7743 set_migr_type(dev, MIGR_GEN_MIGR);
0e2d1a4e
AK		 7744 for (i = prev_num_members;
7745 i < map->num_members; i++)
7746 set_imsm_ord_tbl_ent(map, i, i);
238c0a71 7747 map2 = get_imsm_map(dev, MAP_1);
0e2d1a4e 7748 /* Copy the current map */
1dfaa380 7749 memcpy(map2, map, copy_map_size);
0e2d1a4e
AK		 7750 map2->num_members = prev_num_members;
7751
f3871fdc 7752 imsm_set_array_size(dev, -1);
51d83f5d 7753 super->clean_migration_record_by_mdmon = 1;
0e2d1a4e
AK		 7754 super->updates_pending++;
7755 }
7756}
7757
aad6f216 7758/* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild
0c046afd
DW		 7759 * states are handled in imsm_set_disk() with one exception, when a
7760 * resync is stopped due to a new failure this routine will set the
7761 * 'degraded' state for the array.
7762 */
01f157d7 7763static int imsm_set_array_state(struct active_array *a, int consistent)
a862209d
DW		 7764{
7765 int inst = a->info.container_member;
7766 struct intel_super *super = a->container->sb;
949c47a0 7767 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 7768 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3b451610
AK		 7769 int failed = imsm_count_failed(super, dev, MAP_0);
7770 __u8 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
1e5c6983 7771 __u32 blocks_per_unit;
a862209d 7772
1af97990
AK		 7773 if (dev->vol.migr_state &&
7774 dev->vol.migr_type == MIGR_GEN_MIGR) {
7775 /* array state change is blocked due to reshape action
aad6f216
N		 7776 * We might need to
7777 * - abort the reshape (if last_checkpoint is 0 and action!= reshape)
7778 * - finish the reshape (if last_checkpoint is big and action != reshape)
7779 * - update curr_migr_unit
1af97990 7780 */
aad6f216
N		 7781 if (a->curr_action == reshape) {
7782 /* still reshaping, maybe update curr_migr_unit */
633b5610 7783 goto mark_checkpoint;
aad6f216
N		 7784 } else {
7785 if (a->last_checkpoint == 0 && a->prev_action == reshape) {
7786 /* for some reason we aborted the reshape.
b66e591b
AK		 7787 *
7788 * disable automatic metadata rollback
7789 * user action is required to recover process
aad6f216 7790 */
b66e591b 7791 if (0) {
238c0a71
AK		 7792 struct imsm_map *map2 =
7793 get_imsm_map(dev, MAP_1);
7794 dev->vol.migr_state = 0;
7795 set_migr_type(dev, 0);
7796 dev->vol.curr_migr_unit = 0;
7797 memcpy(map, map2,
7798 sizeof_imsm_map(map2));
7799 super->updates_pending++;
b66e591b 7800 }
aad6f216
N		 7801 }
7802 if (a->last_checkpoint >= a->info.component_size) {
7803 unsigned long long array_blocks;
7804 int used_disks;
e154ced3 7805 struct mdinfo *mdi;
aad6f216 7806
238c0a71 7807 used_disks = imsm_num_data_members(dev, MAP_0);
d55adef9
AK		 7808 if (used_disks > 0) {
7809 array_blocks =
5551b113 7810 blocks_per_member(map) *
d55adef9
AK		 7811 used_disks;
7812 /* round array size down to closest MB
7813 */
7814 array_blocks = (array_blocks
7815 >> SECT_PER_MB_SHIFT)
7816 << SECT_PER_MB_SHIFT;
d55adef9
AK		 7817 a->info.custom_array_size = array_blocks;
7818 /* encourage manager to update array
7819 * size
7820 */
e154ced3 7821
d55adef9 7822 a->check_reshape = 1;
633b5610 7823 }
e154ced3
AK		 7824 /* finalize online capacity expansion/reshape */
7825 for (mdi = a->info.devs; mdi; mdi = mdi->next)
7826 imsm_set_disk(a,
7827 mdi->disk.raid_disk,
7828 mdi->curr_state);
7829
0e2d1a4e 7830 imsm_progress_container_reshape(super);
e154ced3 7831 }
aad6f216 7832 }
1af97990
AK		 7833 }
7834
47ee5a45 7835 /* before we activate this array handle any missing disks */
33414a01
DW		 7836 if (consistent == 2)
7837 handle_missing(super, dev);
1e5c6983 7838
0c046afd 7839 if (consistent == 2 &&
b7941fd6 7840 (!is_resync_complete(&a->info) ||
0c046afd
DW		 7841 map_state != IMSM_T_STATE_NORMAL ||
7842 dev->vol.migr_state))
01f157d7 7843 consistent = 0;
272906ef 7844
b7941fd6 7845 if (is_resync_complete(&a->info)) {
0c046afd 7846 /* complete intialization / resync,
0556e1a2
DW		 7847 * recovery and interrupted recovery is completed in
7848 * ->set_disk
0c046afd
DW		 7849 */
7850 if (is_resyncing(dev)) {
7851 dprintf("imsm: mark resync done\n");
809da78e 7852 end_migration(dev, super, map_state);
115c3803 7853 super->updates_pending++;
484240d8 7854 a->last_checkpoint = 0;
115c3803 7855 }
b9172665
AK		 7856 } else if ((!is_resyncing(dev) && !failed) &&
7857 (imsm_reshape_blocks_arrays_changes(super) == 0)) {
0c046afd 7858 /* mark the start of the init process if nothing is failed */
b7941fd6 7859 dprintf("imsm: mark resync start\n");
1484e727 7860 if (map->map_state == IMSM_T_STATE_UNINITIALIZED)
8e59f3d8 7861 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT);
1484e727 7862 else
8e59f3d8 7863 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR);
3393c6af 7864 super->updates_pending++;
115c3803 7865 }
a862209d 7866
633b5610 7867mark_checkpoint:
5b83bacf
AK		 7868 /* skip checkpointing for general migration,
7869 * it is controlled in mdadm
7870 */
7871 if (is_gen_migration(dev))
7872 goto skip_mark_checkpoint;
7873
1e5c6983 7874 /* check if we can update curr_migr_unit from resync_start, recovery_start */
c47b0ff6 7875 blocks_per_unit = blocks_per_migr_unit(super, dev);
4f0a7acc 7876 if (blocks_per_unit) {
1e5c6983
DW		 7877 __u32 units32;
7878 __u64 units;
7879
4f0a7acc 7880 units = a->last_checkpoint / blocks_per_unit;
1e5c6983
DW		 7881 units32 = units;
7882
7883 /* check that we did not overflow 32-bits, and that
7884 * curr_migr_unit needs updating
7885 */
7886 if (units32 == units &&
bfd80a56 7887 units32 != 0 &&
1e5c6983
DW		 7888 __le32_to_cpu(dev->vol.curr_migr_unit) != units32) {
7889 dprintf("imsm: mark checkpoint (%u)\n", units32);
7890 dev->vol.curr_migr_unit = __cpu_to_le32(units32);
7891 super->updates_pending++;
7892 }
7893 }
f8f603f1 7894
5b83bacf 7895skip_mark_checkpoint:
3393c6af 7896 /* mark dirty / clean */
0c046afd 7897 if (dev->vol.dirty != !consistent) {
b7941fd6 7898 dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty");
0c046afd
DW		 7899 if (consistent)
7900 dev->vol.dirty = 0;
7901 else
7902 dev->vol.dirty = 1;
a862209d
DW		 7903 super->updates_pending++;
7904 }
28bce06f 7905
01f157d7 7906 return consistent;
a862209d
DW		 7907}
7908
6f50473f
TM		 7909static int imsm_disk_slot_to_ord(struct active_array *a, int slot)
7910{
7911 int inst = a->info.container_member;
7912 struct intel_super *super = a->container->sb;
7913 struct imsm_dev *dev = get_imsm_dev(super, inst);
7914 struct imsm_map *map = get_imsm_map(dev, MAP_0);
7915
7916 if (slot > map->num_members) {
7917 pr_err("imsm: imsm_disk_slot_to_ord %d out of range 0..%d\n",
7918 slot, map->num_members - 1);
7919 return -1;
7920 }
7921
7922 if (slot < 0)
7923 return -1;
7924
7925 return get_imsm_ord_tbl_ent(dev, slot, MAP_0);
7926}
7927
8d45d196 7928static void imsm_set_disk(struct active_array *a, int n, int state)
845dea95 7929{
8d45d196
DW		 7930 int inst = a->info.container_member;
7931 struct intel_super *super = a->container->sb;
949c47a0 7932 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 7933 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8d45d196 7934 struct imsm_disk *disk;
7ce05701
LD		 7935 struct mdinfo *mdi;
7936 int recovery_not_finished = 0;
0c046afd 7937 int failed;
6f50473f 7938 int ord;
0c046afd 7939 __u8 map_state;
8d45d196 7940
6f50473f
TM		 7941 ord = imsm_disk_slot_to_ord(a, n);
7942 if (ord < 0)
8d45d196
DW		 7943 return;
7944
4e6e574a 7945 dprintf("imsm: set_disk %d:%x\n", n, state);
b10b37b8 7946 disk = get_imsm_disk(super, ord_to_idx(ord));
8d45d196 7947
5802a811 7948 /* check for new failures */
0556e1a2 7949 if (state & DS_FAULTY) {
4c9e8c1e 7950 if (mark_failure(super, dev, disk, ord_to_idx(ord)))
0556e1a2 7951 super->updates_pending++;
8d45d196 7952 }
47ee5a45 7953
19859edc 7954 /* check if in_sync */
0556e1a2 7955 if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) {
238c0a71 7956 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
b10b37b8
DW		 7957
7958 set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord));
19859edc
DW		 7959 super->updates_pending++;
7960 }
8d45d196 7961
3b451610
AK		 7962 failed = imsm_count_failed(super, dev, MAP_0);
7963 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
5802a811 7964
0c046afd 7965 /* check if recovery complete, newly degraded, or failed */
94002678
AK		 7966 dprintf("imsm: Detected transition to state ");
7967 switch (map_state) {
7968 case IMSM_T_STATE_NORMAL: /* transition to normal state */
7969 dprintf("normal: ");
7970 if (is_rebuilding(dev)) {
1ade5cc1 7971 dprintf_cont("while rebuilding");
7ce05701
LD		 7972 /* check if recovery is really finished */
7973 for (mdi = a->info.devs; mdi ; mdi = mdi->next)
7974 if (mdi->recovery_start != MaxSector) {
7975 recovery_not_finished = 1;
7976 break;
7977 }
7978 if (recovery_not_finished) {
1ade5cc1
N		 7979 dprintf_cont("\n");
7980 dprintf("Rebuild has not finished yet, state not changed");
7ce05701
LD		 7981 if (a->last_checkpoint < mdi->recovery_start) {
7982 a->last_checkpoint = mdi->recovery_start;
7983 super->updates_pending++;
7984 }
7985 break;
7986 }
94002678 7987 end_migration(dev, super, map_state);
238c0a71 7988 map = get_imsm_map(dev, MAP_0);
94002678
AK		 7989 map->failed_disk_num = ~0;
7990 super->updates_pending++;
7991 a->last_checkpoint = 0;
7992 break;
7993 }
7994 if (is_gen_migration(dev)) {
1ade5cc1 7995 dprintf_cont("while general migration");
bf2f0071 7996 if (a->last_checkpoint >= a->info.component_size)
809da78e 7997 end_migration(dev, super, map_state);
94002678
AK		 7998 else
7999 map->map_state = map_state;
238c0a71 8000 map = get_imsm_map(dev, MAP_0);
28bce06f 8001 map->failed_disk_num = ~0;
94002678 8002 super->updates_pending++;
bf2f0071 8003 break;
94002678
AK		 8004 }
8005 break;
8006 case IMSM_T_STATE_DEGRADED: /* transition to degraded state */
1ade5cc1 8007 dprintf_cont("degraded: ");
089f9d79 8008 if (map->map_state != map_state && !dev->vol.migr_state) {
1ade5cc1 8009 dprintf_cont("mark degraded");
94002678
AK		 8010 map->map_state = map_state;
8011 super->updates_pending++;
8012 a->last_checkpoint = 0;
8013 break;
8014 }
8015 if (is_rebuilding(dev)) {
1ade5cc1 8016 dprintf_cont("while rebuilding.");
94002678 8017 if (map->map_state != map_state) {
1ade5cc1 8018 dprintf_cont(" Map state change");
94002678
AK		 8019 end_migration(dev, super, map_state);
8020 super->updates_pending++;
8021 }
8022 break;
8023 }
8024 if (is_gen_migration(dev)) {
1ade5cc1 8025 dprintf_cont("while general migration");
bf2f0071 8026 if (a->last_checkpoint >= a->info.component_size)
809da78e 8027 end_migration(dev, super, map_state);
94002678
AK		 8028 else {
8029 map->map_state = map_state;
3b451610 8030 manage_second_map(super, dev);
94002678
AK		 8031 }
8032 super->updates_pending++;
bf2f0071 8033 break;
28bce06f 8034 }
6ce1fbf1 8035 if (is_initializing(dev)) {
1ade5cc1 8036 dprintf_cont("while initialization.");
6ce1fbf1
AK		 8037 map->map_state = map_state;
8038 super->updates_pending++;
8039 break;
8040 }
94002678
AK		 8041 break;
8042 case IMSM_T_STATE_FAILED: /* transition to failed state */
1ade5cc1 8043 dprintf_cont("failed: ");
94002678 8044 if (is_gen_migration(dev)) {
1ade5cc1 8045 dprintf_cont("while general migration");
94002678
AK		 8046 map->map_state = map_state;
8047 super->updates_pending++;
8048 break;
8049 }
8050 if (map->map_state != map_state) {
1ade5cc1 8051 dprintf_cont("mark failed");
94002678
AK		 8052 end_migration(dev, super, map_state);
8053 super->updates_pending++;
8054 a->last_checkpoint = 0;
8055 break;
8056 }
8057 break;
8058 default:
1ade5cc1 8059 dprintf_cont("state %i\n", map_state);
5802a811 8060 }
1ade5cc1 8061 dprintf_cont("\n");
845dea95
NB		 8062}
8063
f796af5d 8064static int store_imsm_mpb(int fd, struct imsm_super *mpb)
c2a1e7da 8065{
f796af5d 8066 void *buf = mpb;
c2a1e7da
DW		 8067 __u32 mpb_size = __le32_to_cpu(mpb->mpb_size);
8068 unsigned long long dsize;
8069 unsigned long long sectors;
f36a9ecd 8070 unsigned int sector_size;
c2a1e7da 8071
f36a9ecd 8072 get_dev_sector_size(fd, NULL, &sector_size);
c2a1e7da
DW		 8073 get_dev_size(fd, NULL, &dsize);
8074
f36a9ecd 8075 if (mpb_size > sector_size) {
272f648f 8076 /* -1 to account for anchor */
f36a9ecd 8077 sectors = mpb_sectors(mpb, sector_size) - 1;
c2a1e7da 8078
272f648f 8079 /* write the extended mpb to the sectors preceeding the anchor */
f36a9ecd
PB		 8080 if (lseek64(fd, dsize - (sector_size * (2 + sectors)),
8081 SEEK_SET) < 0)
272f648f 8082 return 1;
c2a1e7da 8083
f36a9ecd
PB		 8084 if ((unsigned long long)write(fd, buf + sector_size,
8085 sector_size * sectors) != sector_size * sectors)
272f648f
DW		 8086 return 1;
8087 }
c2a1e7da 8088
272f648f 8089 /* first block is stored on second to last sector of the disk */
f36a9ecd 8090 if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0)
c2a1e7da
DW		 8091 return 1;
8092
f36a9ecd 8093 if (write(fd, buf, sector_size) != sector_size)
c2a1e7da
DW		 8094 return 1;
8095
c2a1e7da
DW		 8096 return 0;
8097}
8098
2e735d19 8099static void imsm_sync_metadata(struct supertype *container)
845dea95 8100{
2e735d19 8101 struct intel_super *super = container->sb;
c2a1e7da 8102
1a64be56 8103 dprintf("sync metadata: %d\n", super->updates_pending);
c2a1e7da
DW		 8104 if (!super->updates_pending)
8105 return;
8106
36988a3d 8107 write_super_imsm(container, 0);
c2a1e7da
DW		 8108
8109 super->updates_pending = 0;
845dea95
NB		 8110}
8111
272906ef
DW		 8112static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a)
8113{
8114 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8115 int i = get_imsm_disk_idx(dev, idx, MAP_X);
272906ef
DW		 8116 struct dl *dl;
8117
8118 for (dl = super->disks; dl; dl = dl->next)
8119 if (dl->index == i)
8120 break;
8121
25ed7e59 8122 if (dl && is_failed(&dl->disk))
272906ef
DW		 8123 dl = NULL;
8124
8125 if (dl)
1ade5cc1 8126 dprintf("found %x:%x\n", dl->major, dl->minor);
272906ef
DW		 8127
8128 return dl;
8129}
8130
a20d2ba5 8131static struct dl *imsm_add_spare(struct intel_super *super, int slot,
8ba77d32
AK		 8132 struct active_array *a, int activate_new,
8133 struct mdinfo *additional_test_list)
272906ef
DW		 8134{
8135 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8136 int idx = get_imsm_disk_idx(dev, slot, MAP_X);
a20d2ba5
DW		 8137 struct imsm_super *mpb = super->anchor;
8138 struct imsm_map *map;
272906ef
DW		 8139 unsigned long long pos;
8140 struct mdinfo *d;
8141 struct extent *ex;
a20d2ba5 8142 int i, j;
272906ef 8143 int found;
569cc43f
DW		 8144 __u32 array_start = 0;
8145 __u32 array_end = 0;
272906ef 8146 struct dl *dl;
6c932028 8147 struct mdinfo *test_list;
272906ef
DW		 8148
8149 for (dl = super->disks; dl; dl = dl->next) {
8150 /* If in this array, skip */
8151 for (d = a->info.devs ; d ; d = d->next)
e553d2a4
DW		 8152 if (d->state_fd >= 0 &&
8153 d->disk.major == dl->major &&
272906ef 8154 d->disk.minor == dl->minor) {
8ba77d32
AK		 8155 dprintf("%x:%x already in array\n",
8156 dl->major, dl->minor);
272906ef
DW		 8157 break;
8158 }
8159 if (d)
8160 continue;
6c932028
AK		 8161 test_list = additional_test_list;
8162 while (test_list) {
8163 if (test_list->disk.major == dl->major &&
8164 test_list->disk.minor == dl->minor) {
8ba77d32
AK		 8165 dprintf("%x:%x already in additional test list\n",
8166 dl->major, dl->minor);
8167 break;
8168 }
6c932028 8169 test_list = test_list->next;
8ba77d32 8170 }
6c932028 8171 if (test_list)
8ba77d32 8172 continue;
272906ef 8173
e553d2a4 8174 /* skip in use or failed drives */
25ed7e59 8175 if (is_failed(&dl->disk) || idx == dl->index ||
df474657
DW		 8176 dl->index == -2) {
8177 dprintf("%x:%x status (failed: %d index: %d)\n",
25ed7e59 8178 dl->major, dl->minor, is_failed(&dl->disk), idx);
9a1608e5
DW		 8179 continue;
8180 }
8181
a20d2ba5
DW		 8182 /* skip pure spares when we are looking for partially
8183 * assimilated drives
8184 */
8185 if (dl->index == -1 && !activate_new)
8186 continue;
8187
272906ef 8188 /* Does this unused device have the requisite free space?
a20d2ba5 8189 * It needs to be able to cover all member volumes
272906ef
DW		 8190 */
8191 ex = get_extents(super, dl);
8192 if (!ex) {
8193 dprintf("cannot get extents\n");
8194 continue;
8195 }
a20d2ba5
DW		 8196 for (i = 0; i < mpb->num_raid_devs; i++) {
8197 dev = get_imsm_dev(super, i);
238c0a71 8198 map = get_imsm_map(dev, MAP_0);
272906ef 8199
a20d2ba5
DW		 8200 /* check if this disk is already a member of
8201 * this array
272906ef 8202 */
620b1713 8203 if (get_imsm_disk_slot(map, dl->index) >= 0)
a20d2ba5
DW		 8204 continue;
8205
8206 found = 0;
8207 j = 0;
8208 pos = 0;
5551b113 8209 array_start = pba_of_lba0(map);
329c8278 8210 array_end = array_start +
5551b113 8211 blocks_per_member(map) - 1;
a20d2ba5
DW		 8212
8213 do {
8214 /* check that we can start at pba_of_lba0 with
8215 * blocks_per_member of space
8216 */
329c8278 8217 if (array_start >= pos && array_end < ex[j].start) {
a20d2ba5
DW		 8218 found = 1;
8219 break;
8220 }
8221 pos = ex[j].start + ex[j].size;
8222 j++;
8223 } while (ex[j-1].size);
8224
8225 if (!found)
272906ef 8226 break;
a20d2ba5 8227 }
272906ef
DW		 8228
8229 free(ex);
a20d2ba5 8230 if (i < mpb->num_raid_devs) {
329c8278
DW		 8231 dprintf("%x:%x does not have %u to %u available\n",
8232 dl->major, dl->minor, array_start, array_end);
272906ef
DW		 8233 /* No room */
8234 continue;
a20d2ba5
DW		 8235 }
8236 return dl;
272906ef
DW		 8237 }
8238
8239 return dl;
8240}
8241
95d07a2c
LM		 8242static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed)
8243{
8244 struct imsm_dev *dev2;
8245 struct imsm_map *map;
8246 struct dl *idisk;
8247 int slot;
8248 int idx;
8249 __u8 state;
8250
8251 dev2 = get_imsm_dev(cont->sb, dev_idx);
8252 if (dev2) {
238c0a71 8253 state = imsm_check_degraded(cont->sb, dev2, failed, MAP_0);
95d07a2c 8254 if (state == IMSM_T_STATE_FAILED) {
238c0a71 8255 map = get_imsm_map(dev2, MAP_0);
95d07a2c
LM		 8256 if (!map)
8257 return 1;
8258 for (slot = 0; slot < map->num_members; slot++) {
8259 /*
8260 * Check if failed disks are deleted from intel
8261 * disk list or are marked to be deleted
8262 */
238c0a71 8263 idx = get_imsm_disk_idx(dev2, slot, MAP_X);
95d07a2c
LM		 8264 idisk = get_imsm_dl_disk(cont->sb, idx);
8265 /*
8266 * Do not rebuild the array if failed disks
8267 * from failed sub-array are not removed from
8268 * container.
8269 */
8270 if (idisk &&
8271 is_failed(&idisk->disk) &&
8272 (idisk->action != DISK_REMOVE))
8273 return 0;
8274 }
8275 }
8276 }
8277 return 1;
8278}
8279
88758e9d
DW		 8280static struct mdinfo *imsm_activate_spare(struct active_array *a,
8281 struct metadata_update **updates)
8282{
8283 /**
d23fe947
DW		 8284 * Find a device with unused free space and use it to replace a
8285 * failed/vacant region in an array. We replace failed regions one a
8286 * array at a time. The result is that a new spare disk will be added
8287 * to the first failed array and after the monitor has finished
8288 * propagating failures the remainder will be consumed.
88758e9d 8289 *
d23fe947
DW		 8290 * FIXME add a capability for mdmon to request spares from another
8291 * container.
88758e9d
DW		 8292 */
8293
8294 struct intel_super *super = a->container->sb;
88758e9d 8295 int inst = a->info.container_member;
949c47a0 8296 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8297 struct imsm_map *map = get_imsm_map(dev, MAP_0);
88758e9d
DW		 8298 int failed = a->info.array.raid_disks;
8299 struct mdinfo *rv = NULL;
8300 struct mdinfo *d;
8301 struct mdinfo *di;
8302 struct metadata_update *mu;
8303 struct dl *dl;
8304 struct imsm_update_activate_spare *u;
8305 int num_spares = 0;
8306 int i;
95d07a2c 8307 int allowed;
88758e9d
DW		 8308
8309 for (d = a->info.devs ; d ; d = d->next) {
8310 if ((d->curr_state & DS_FAULTY) &&
8311 d->state_fd >= 0)
8312 /* wait for Removal to happen */
8313 return NULL;
8314 if (d->state_fd >= 0)
8315 failed--;
8316 }
8317
8318 dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n",
8319 inst, failed, a->info.array.raid_disks, a->info.array.level);
1af97990 8320
e2962bfc
AK		 8321 if (imsm_reshape_blocks_arrays_changes(super))
8322 return NULL;
1af97990 8323
fc8ca064
AK		 8324 /* Cannot activate another spare if rebuild is in progress already
8325 */
8326 if (is_rebuilding(dev)) {
7a862a02 8327 dprintf("imsm: No spare activation allowed. Rebuild in progress already.\n");
fc8ca064
AK		 8328 return NULL;
8329 }
8330
89c67882
AK		 8331 if (a->info.array.level == 4)
8332 /* No repair for takeovered array
8333 * imsm doesn't support raid4
8334 */
8335 return NULL;
8336
3b451610
AK		 8337 if (imsm_check_degraded(super, dev, failed, MAP_0) !=
8338 IMSM_T_STATE_DEGRADED)
88758e9d
DW		 8339 return NULL;
8340
83ca7d45
AP		 8341 if (get_imsm_map(dev, MAP_0)->map_state == IMSM_T_STATE_UNINITIALIZED) {
8342 dprintf("imsm: No spare activation allowed. Volume is not initialized.\n");
8343 return NULL;
8344 }
8345
95d07a2c
LM		 8346 /*
8347 * If there are any failed disks check state of the other volume.
8348 * Block rebuild if the another one is failed until failed disks
8349 * are removed from container.
8350 */
8351 if (failed) {
7a862a02 8352 dprintf("found failed disks in %.*s, check if there anotherfailed sub-array.\n",
c4acd1e5 8353 MAX_RAID_SERIAL_LEN, dev->volume);
95d07a2c
LM		 8354 /* check if states of the other volumes allow for rebuild */
8355 for (i = 0; i < super->anchor->num_raid_devs; i++) {
8356 if (i != inst) {
8357 allowed = imsm_rebuild_allowed(a->container,
8358 i, failed);
8359 if (!allowed)
8360 return NULL;
8361 }
8362 }
8363 }
8364
88758e9d 8365 /* For each slot, if it is not working, find a spare */
88758e9d
DW		 8366 for (i = 0; i < a->info.array.raid_disks; i++) {
8367 for (d = a->info.devs ; d ; d = d->next)
8368 if (d->disk.raid_disk == i)
8369 break;
8370 dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0);
8371 if (d && (d->state_fd >= 0))
8372 continue;
8373
272906ef 8374 /*
a20d2ba5
DW		 8375 * OK, this device needs recovery. Try to re-add the
8376 * previous occupant of this slot, if this fails see if
8377 * we can continue the assimilation of a spare that was
8378 * partially assimilated, finally try to activate a new
8379 * spare.
272906ef
DW		 8380 */
8381 dl = imsm_readd(super, i, a);
8382 if (!dl)
b303fe21 8383 dl = imsm_add_spare(super, i, a, 0, rv);
a20d2ba5 8384 if (!dl)
b303fe21 8385 dl = imsm_add_spare(super, i, a, 1, rv);
272906ef
DW		 8386 if (!dl)
8387 continue;
1011e834 8388
272906ef 8389 /* found a usable disk with enough space */
503975b9 8390 di = xcalloc(1, sizeof(*di));
272906ef
DW		 8391
8392 /* dl->index will be -1 in the case we are activating a
8393 * pristine spare. imsm_process_update() will create a
8394 * new index in this case. Once a disk is found to be
8395 * failed in all member arrays it is kicked from the
8396 * metadata
8397 */
8398 di->disk.number = dl->index;
d23fe947 8399
272906ef
DW		 8400 /* (ab)use di->devs to store a pointer to the device
8401 * we chose
8402 */
8403 di->devs = (struct mdinfo *) dl;
8404
8405 di->disk.raid_disk = i;
8406 di->disk.major = dl->major;
8407 di->disk.minor = dl->minor;
8408 di->disk.state = 0;
d23534e4 8409 di->recovery_start = 0;
5551b113 8410 di->data_offset = pba_of_lba0(map);
272906ef
DW		 8411 di->component_size = a->info.component_size;
8412 di->container_member = inst;
b12796be 8413 di->bb.supported = 0;
148acb7b 8414 super->random = random32();
272906ef
DW		 8415 di->next = rv;
8416 rv = di;
8417 num_spares++;
8418 dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor,
8419 i, di->data_offset);
88758e9d
DW		 8420 }
8421
8422 if (!rv)
8423 /* No spares found */
8424 return rv;
8425 /* Now 'rv' has a list of devices to return.
8426 * Create a metadata_update record to update the
8427 * disk_ord_tbl for the array
8428 */
503975b9 8429 mu = xmalloc(sizeof(*mu));
1011e834 8430 mu->buf = xcalloc(num_spares,
503975b9 8431 sizeof(struct imsm_update_activate_spare));
88758e9d 8432 mu->space = NULL;
cb23f1f4 8433 mu->space_list = NULL;
88758e9d
DW		 8434 mu->len = sizeof(struct imsm_update_activate_spare) * num_spares;
8435 mu->next = *updates;
8436 u = (struct imsm_update_activate_spare *) mu->buf;
8437
8438 for (di = rv ; di ; di = di->next) {
8439 u->type = update_activate_spare;
d23fe947
DW		 8440 u->dl = (struct dl *) di->devs;
8441 di->devs = NULL;
88758e9d
DW		 8442 u->slot = di->disk.raid_disk;
8443 u->array = inst;
8444 u->next = u + 1;
8445 u++;
8446 }
8447 (u-1)->next = NULL;
8448 *updates = mu;
8449
8450 return rv;
8451}
8452
54c2c1ea 8453static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u)
8273f55e 8454{
54c2c1ea 8455 struct imsm_dev *dev = get_imsm_dev(super, idx);
238c0a71
AK		 8456 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8457 struct imsm_map *new_map = get_imsm_map(&u->dev, MAP_0);
54c2c1ea
DW		 8458 struct disk_info *inf = get_disk_info(u);
8459 struct imsm_disk *disk;
8273f55e
DW		 8460 int i;
8461 int j;
8273f55e 8462
54c2c1ea 8463 for (i = 0; i < map->num_members; i++) {
238c0a71 8464 disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, MAP_X));
54c2c1ea
DW		 8465 for (j = 0; j < new_map->num_members; j++)
8466 if (serialcmp(disk->serial, inf[j].serial) == 0)
8273f55e
DW		 8467 return 1;
8468 }
8469
8470 return 0;
8471}
8472
1a64be56
LM		 8473static struct dl *get_disk_super(struct intel_super *super, int major, int minor)
8474{
594dc1b8
JS		 8475 struct dl *dl;
8476
1a64be56 8477 for (dl = super->disks; dl; dl = dl->next)
089f9d79 8478 if (dl->major == major && dl->minor == minor)
1a64be56
LM		 8479 return dl;
8480 return NULL;
8481}
8482
8483static int remove_disk_super(struct intel_super *super, int major, int minor)
8484{
594dc1b8 8485 struct dl *prev;
1a64be56
LM		 8486 struct dl *dl;
8487
8488 prev = NULL;
8489 for (dl = super->disks; dl; dl = dl->next) {
089f9d79 8490 if (dl->major == major && dl->minor == minor) {
1a64be56
LM		 8491 /* remove */
8492 if (prev)
8493 prev->next = dl->next;
8494 else
8495 super->disks = dl->next;
8496 dl->next = NULL;
8497 __free_imsm_disk(dl);
1ade5cc1 8498 dprintf("removed %x:%x\n", major, minor);
1a64be56
LM		 8499 break;
8500 }
8501 prev = dl;
8502 }
8503 return 0;
8504}
8505
f21e18ca 8506static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index);
ae6aad82 8507
1a64be56
LM		 8508static int add_remove_disk_update(struct intel_super *super)
8509{
8510 int check_degraded = 0;
594dc1b8
JS		 8511 struct dl *disk;
8512
1a64be56
LM		 8513 /* add/remove some spares to/from the metadata/contrainer */
8514 while (super->disk_mgmt_list) {
8515 struct dl *disk_cfg;
8516
8517 disk_cfg = super->disk_mgmt_list;
8518 super->disk_mgmt_list = disk_cfg->next;
8519 disk_cfg->next = NULL;
8520
8521 if (disk_cfg->action == DISK_ADD) {
8522 disk_cfg->next = super->disks;
8523 super->disks = disk_cfg;
8524 check_degraded = 1;
1ade5cc1
N		 8525 dprintf("added %x:%x\n",
8526 disk_cfg->major, disk_cfg->minor);
1a64be56
LM		 8527 } else if (disk_cfg->action == DISK_REMOVE) {
8528 dprintf("Disk remove action processed: %x.%x\n",
8529 disk_cfg->major, disk_cfg->minor);
8530 disk = get_disk_super(super,
8531 disk_cfg->major,
8532 disk_cfg->minor);
8533 if (disk) {
8534 /* store action status */
8535 disk->action = DISK_REMOVE;
8536 /* remove spare disks only */
8537 if (disk->index == -1) {
8538 remove_disk_super(super,
8539 disk_cfg->major,
8540 disk_cfg->minor);
8541 }
8542 }
8543 /* release allocate disk structure */
8544 __free_imsm_disk(disk_cfg);
8545 }
8546 }
8547 return check_degraded;
8548}
8549
a29911da
PC		 8550static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u,
8551 struct intel_super *super,
8552 void ***space_list)
8553{
8554 struct intel_dev *id;
8555 void **tofree = NULL;
8556 int ret_val = 0;
8557
1ade5cc1 8558 dprintf("(enter)\n");
089f9d79 8559 if (u->subdev < 0 || u->subdev > 1) {
a29911da
PC		 8560 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
8561 return ret_val;
8562 }
089f9d79 8563 if (space_list == NULL || *space_list == NULL) {
a29911da
PC		 8564 dprintf("imsm: Error: Memory is not allocated\n");
8565 return ret_val;
8566 }
8567
8568 for (id = super->devlist ; id; id = id->next) {
8569 if (id->index == (unsigned)u->subdev) {
8570 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
8571 struct imsm_map *map;
8572 struct imsm_dev *new_dev =
8573 (struct imsm_dev *)*space_list;
238c0a71 8574 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
a29911da
PC		 8575 int to_state;
8576 struct dl *new_disk;
8577
8578 if (new_dev == NULL)
8579 return ret_val;
8580 *space_list = **space_list;
8581 memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0));
238c0a71 8582 map = get_imsm_map(new_dev, MAP_0);
a29911da
PC		 8583 if (migr_map) {
8584 dprintf("imsm: Error: migration in progress");
8585 return ret_val;
8586 }
8587
8588 to_state = map->map_state;
8589 if ((u->new_level == 5) && (map->raid_level == 0)) {
8590 map->num_members++;
8591 /* this should not happen */
8592 if (u->new_disks[0] < 0) {
8593 map->failed_disk_num =
8594 map->num_members - 1;
8595 to_state = IMSM_T_STATE_DEGRADED;
8596 } else
8597 to_state = IMSM_T_STATE_NORMAL;
8598 }
8e59f3d8 8599 migrate(new_dev, super, to_state, MIGR_GEN_MIGR);
a29911da
PC		 8600 if (u->new_level > -1)
8601 map->raid_level = u->new_level;
238c0a71 8602 migr_map = get_imsm_map(new_dev, MAP_1);
a29911da
PC		 8603 if ((u->new_level == 5) &&
8604 (migr_map->raid_level == 0)) {
8605 int ord = map->num_members - 1;
8606 migr_map->num_members--;
8607 if (u->new_disks[0] < 0)
8608 ord |= IMSM_ORD_REBUILD;
8609 set_imsm_ord_tbl_ent(map,
8610 map->num_members - 1,
8611 ord);
8612 }
8613 id->dev = new_dev;
8614 tofree = (void **)dev;
8615
4bba0439
PC		 8616 /* update chunk size
8617 */
06fb291a
PB		 8618 if (u->new_chunksize > 0) {
8619 unsigned long long num_data_stripes;
8620 int used_disks =
8621 imsm_num_data_members(dev, MAP_0);
8622
8623 if (used_disks == 0)
8624 return ret_val;
8625
4bba0439
PC		 8626 map->blocks_per_strip =
8627 __cpu_to_le16(u->new_chunksize * 2);
06fb291a
PB		 8628 num_data_stripes =
8629 (join_u32(dev->size_low, dev->size_high)
8630 / used_disks);
8631 num_data_stripes /= map->blocks_per_strip;
8632 num_data_stripes /= map->num_domains;
8633 set_num_data_stripes(map, num_data_stripes);
8634 }
4bba0439 8635
a29911da
PC		 8636 /* add disk
8637 */
089f9d79
JS		 8638 if (u->new_level != 5 || migr_map->raid_level != 0 ||
8639 migr_map->raid_level == map->raid_level)
a29911da
PC		 8640 goto skip_disk_add;
8641
8642 if (u->new_disks[0] >= 0) {
8643 /* use passes spare
8644 */
8645 new_disk = get_disk_super(super,
8646 major(u->new_disks[0]),
8647 minor(u->new_disks[0]));
7a862a02 8648 dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n",
a29911da
PC		 8649 major(u->new_disks[0]),
8650 minor(u->new_disks[0]),
8651 new_disk, new_disk->index);
8652 if (new_disk == NULL)
8653 goto error_disk_add;
8654
8655 new_disk->index = map->num_members - 1;
8656 /* slot to fill in autolayout
8657 */
8658 new_disk->raiddisk = new_disk->index;
8659 new_disk->disk.status |= CONFIGURED_DISK;
8660 new_disk->disk.status &= ~SPARE_DISK;
8661 } else
8662 goto error_disk_add;
8663
8664skip_disk_add:
8665 *tofree = *space_list;
8666 /* calculate new size
8667 */
f3871fdc 8668 imsm_set_array_size(new_dev, -1);
a29911da
PC		 8669
8670 ret_val = 1;
8671 }
8672 }
8673
8674 if (tofree)
8675 *space_list = tofree;
8676 return ret_val;
8677
8678error_disk_add:
8679 dprintf("Error: imsm: Cannot find disk.\n");
8680 return ret_val;
8681}
8682
f3871fdc
AK		 8683static int apply_size_change_update(struct imsm_update_size_change *u,
8684 struct intel_super *super)
8685{
8686 struct intel_dev *id;
8687 int ret_val = 0;
8688
1ade5cc1 8689 dprintf("(enter)\n");
089f9d79 8690 if (u->subdev < 0 || u->subdev > 1) {
f3871fdc
AK		 8691 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
8692 return ret_val;
8693 }
8694
8695 for (id = super->devlist ; id; id = id->next) {
8696 if (id->index == (unsigned)u->subdev) {
8697 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
8698 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8699 int used_disks = imsm_num_data_members(dev, MAP_0);
8700 unsigned long long blocks_per_member;
06fb291a 8701 unsigned long long num_data_stripes;
f3871fdc
AK		 8702
8703 /* calculate new size
8704 */
8705 blocks_per_member = u->new_size / used_disks;
06fb291a
PB		 8706 num_data_stripes = blocks_per_member /
8707 map->blocks_per_strip;
8708 num_data_stripes /= map->num_domains;
8709 dprintf("(size: %llu, blocks per member: %llu, num_data_stipes: %llu)\n",
8710 u->new_size, blocks_per_member,
8711 num_data_stripes);
f3871fdc 8712 set_blocks_per_member(map, blocks_per_member);
06fb291a 8713 set_num_data_stripes(map, num_data_stripes);
f3871fdc
AK		 8714 imsm_set_array_size(dev, u->new_size);
8715
8716 ret_val = 1;
8717 break;
8718 }
8719 }
8720
8721 return ret_val;
8722}
8723
061d7da3 8724static int apply_update_activate_spare(struct imsm_update_activate_spare *u,
ca9de185 8725 struct intel_super *super,
061d7da3
LO		 8726 struct active_array *active_array)
8727{
8728 struct imsm_super *mpb = super->anchor;
8729 struct imsm_dev *dev = get_imsm_dev(super, u->array);
238c0a71 8730 struct imsm_map *map = get_imsm_map(dev, MAP_0);
061d7da3
LO		 8731 struct imsm_map *migr_map;
8732 struct active_array *a;
8733 struct imsm_disk *disk;
8734 __u8 to_state;
8735 struct dl *dl;
8736 unsigned int found;
8737 int failed;
5961eeec 8738 int victim;
061d7da3 8739 int i;
5961eeec 8740 int second_map_created = 0;
061d7da3 8741
5961eeec 8742 for (; u; u = u->next) {
238c0a71 8743 victim = get_imsm_disk_idx(dev, u->slot, MAP_X);
061d7da3 8744
5961eeec 8745 if (victim < 0)
8746 return 0;
061d7da3 8747
5961eeec 8748 for (dl = super->disks; dl; dl = dl->next)
8749 if (dl == u->dl)
8750 break;
061d7da3 8751
5961eeec 8752 if (!dl) {
7a862a02 8753 pr_err("error: imsm_activate_spare passed an unknown disk (index: %d)\n",
5961eeec 8754 u->dl->index);
8755 return 0;
8756 }
061d7da3 8757
5961eeec 8758 /* count failures (excluding rebuilds and the victim)
8759 * to determine map[0] state
8760 */
8761 failed = 0;
8762 for (i = 0; i < map->num_members; i++) {
8763 if (i == u->slot)
8764 continue;
8765 disk = get_imsm_disk(super,
238c0a71 8766 get_imsm_disk_idx(dev, i, MAP_X));
5961eeec 8767 if (!disk || is_failed(disk))
8768 failed++;
8769 }
061d7da3 8770
5961eeec 8771 /* adding a pristine spare, assign a new index */
8772 if (dl->index < 0) {
8773 dl->index = super->anchor->num_disks;
8774 super->anchor->num_disks++;
8775 }
8776 disk = &dl->disk;
8777 disk->status |= CONFIGURED_DISK;
8778 disk->status &= ~SPARE_DISK;
8779
8780 /* mark rebuild */
238c0a71 8781 to_state = imsm_check_degraded(super, dev, failed, MAP_0);
5961eeec 8782 if (!second_map_created) {
8783 second_map_created = 1;
8784 map->map_state = IMSM_T_STATE_DEGRADED;
8785 migrate(dev, super, to_state, MIGR_REBUILD);
8786 } else
8787 map->map_state = to_state;
238c0a71 8788 migr_map = get_imsm_map(dev, MAP_1);
5961eeec 8789 set_imsm_ord_tbl_ent(map, u->slot, dl->index);
8790 set_imsm_ord_tbl_ent(migr_map, u->slot,
8791 dl->index | IMSM_ORD_REBUILD);
8792
8793 /* update the family_num to mark a new container
8794 * generation, being careful to record the existing
8795 * family_num in orig_family_num to clean up after
8796 * earlier mdadm versions that neglected to set it.
8797 */
8798 if (mpb->orig_family_num == 0)
8799 mpb->orig_family_num = mpb->family_num;
8800 mpb->family_num += super->random;
8801
8802 /* count arrays using the victim in the metadata */
8803 found = 0;
8804 for (a = active_array; a ; a = a->next) {
8805 dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8806 map = get_imsm_map(dev, MAP_0);
061d7da3 8807
5961eeec 8808 if (get_imsm_disk_slot(map, victim) >= 0)
8809 found++;
8810 }
061d7da3 8811
5961eeec 8812 /* delete the victim if it is no longer being
8813 * utilized anywhere
061d7da3 8814 */
5961eeec 8815 if (!found) {
8816 struct dl **dlp;
061d7da3 8817
5961eeec 8818 /* We know that 'manager' isn't touching anything,
8819 * so it is safe to delete
8820 */
8821 for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next)
061d7da3
LO		 8822 if ((*dlp)->index == victim)
8823 break;
5961eeec 8824
8825 /* victim may be on the missing list */
8826 if (!*dlp)
8827 for (dlp = &super->missing; *dlp;
8828 dlp = &(*dlp)->next)
8829 if ((*dlp)->index == victim)
8830 break;
8831 imsm_delete(super, dlp, victim);
8832 }
061d7da3
LO		 8833 }
8834
8835 return 1;
8836}
a29911da 8837
2e5dc010
N		 8838static int apply_reshape_container_disks_update(struct imsm_update_reshape *u,
8839 struct intel_super *super,
8840 void ***space_list)
8841{
8842 struct dl *new_disk;
8843 struct intel_dev *id;
8844 int i;
8845 int delta_disks = u->new_raid_disks - u->old_raid_disks;
ee4beede 8846 int disk_count = u->old_raid_disks;
2e5dc010
N		 8847 void **tofree = NULL;
8848 int devices_to_reshape = 1;
8849 struct imsm_super *mpb = super->anchor;
8850 int ret_val = 0;
d098291a 8851 unsigned int dev_id;
2e5dc010 8852
1ade5cc1 8853 dprintf("(enter)\n");
2e5dc010
N		 8854
8855 /* enable spares to use in array */
8856 for (i = 0; i < delta_disks; i++) {
8857 new_disk = get_disk_super(super,
8858 major(u->new_disks[i]),
8859 minor(u->new_disks[i]));
7a862a02 8860 dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n",
2e5dc010
N		 8861 major(u->new_disks[i]), minor(u->new_disks[i]),
8862 new_disk, new_disk->index);
089f9d79
JS		 8863 if (new_disk == NULL ||
8864 (new_disk->index >= 0 &&
8865 new_disk->index < u->old_raid_disks))
2e5dc010 8866 goto update_reshape_exit;
ee4beede 8867 new_disk->index = disk_count++;
2e5dc010
N		 8868 /* slot to fill in autolayout
8869 */
8870 new_disk->raiddisk = new_disk->index;
8871 new_disk->disk.status |=
8872 CONFIGURED_DISK;
8873 new_disk->disk.status &= ~SPARE_DISK;
8874 }
8875
ed7333bd
AK		 8876 dprintf("imsm: volume set mpb->num_raid_devs = %i\n",
8877 mpb->num_raid_devs);
2e5dc010
N		 8878 /* manage changes in volume
8879 */
d098291a 8880 for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) {
2e5dc010
N		 8881 void **sp = *space_list;
8882 struct imsm_dev *newdev;
8883 struct imsm_map *newmap, *oldmap;
8884
d098291a
AK		 8885 for (id = super->devlist ; id; id = id->next) {
8886 if (id->index == dev_id)
8887 break;
8888 }
8889 if (id == NULL)
8890 break;
2e5dc010
N		 8891 if (!sp)
8892 continue;
8893 *space_list = *sp;
8894 newdev = (void*)sp;
8895 /* Copy the dev, but not (all of) the map */
8896 memcpy(newdev, id->dev, sizeof(*newdev));
238c0a71
AK		 8897 oldmap = get_imsm_map(id->dev, MAP_0);
8898 newmap = get_imsm_map(newdev, MAP_0);
2e5dc010
N		 8899 /* Copy the current map */
8900 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
8901 /* update one device only
8902 */
8903 if (devices_to_reshape) {
ed7333bd
AK		 8904 dprintf("imsm: modifying subdev: %i\n",
8905 id->index);
2e5dc010
N		 8906 devices_to_reshape--;
8907 newdev->vol.migr_state = 1;
8908 newdev->vol.curr_migr_unit = 0;
ea672ee1 8909 set_migr_type(newdev, MIGR_GEN_MIGR);
2e5dc010
N		 8910 newmap->num_members = u->new_raid_disks;
8911 for (i = 0; i < delta_disks; i++) {
8912 set_imsm_ord_tbl_ent(newmap,
8913 u->old_raid_disks + i,
8914 u->old_raid_disks + i);
8915 }
8916 /* New map is correct, now need to save old map
8917 */
238c0a71 8918 newmap = get_imsm_map(newdev, MAP_1);
2e5dc010
N		 8919 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
8920
f3871fdc 8921 imsm_set_array_size(newdev, -1);
2e5dc010
N		 8922 }
8923
8924 sp = (void **)id->dev;
8925 id->dev = newdev;
8926 *sp = tofree;
8927 tofree = sp;
8e59f3d8
AK		 8928
8929 /* Clear migration record */
8930 memset(super->migr_rec, 0, sizeof(struct migr_record));
2e5dc010 8931 }
819bc634
AK		 8932 if (tofree)
8933 *space_list = tofree;
2e5dc010
N		 8934 ret_val = 1;
8935
8936update_reshape_exit:
8937
8938 return ret_val;
8939}
8940
bb025c2f 8941static int apply_takeover_update(struct imsm_update_takeover *u,
8ca6df95
KW		 8942 struct intel_super *super,
8943 void ***space_list)
bb025c2f
KW		 8944{
8945 struct imsm_dev *dev = NULL;
8ca6df95
KW		 8946 struct intel_dev *dv;
8947 struct imsm_dev *dev_new;
bb025c2f
KW		 8948 struct imsm_map *map;
8949 struct dl *dm, *du;
8ca6df95 8950 int i;
bb025c2f
KW		 8951
8952 for (dv = super->devlist; dv; dv = dv->next)
8953 if (dv->index == (unsigned int)u->subarray) {
8954 dev = dv->dev;
8955 break;
8956 }
8957
8958 if (dev == NULL)
8959 return 0;
8960
238c0a71 8961 map = get_imsm_map(dev, MAP_0);
bb025c2f
KW		 8962
8963 if (u->direction == R10_TO_R0) {
06fb291a
PB		 8964 unsigned long long num_data_stripes;
8965
8966 map->num_domains = 1;
8967 num_data_stripes = blocks_per_member(map);
8968 num_data_stripes /= map->blocks_per_strip;
8969 num_data_stripes /= map->num_domains;
8970 set_num_data_stripes(map, num_data_stripes);
8971
43d5ec18 8972 /* Number of failed disks must be half of initial disk number */
3b451610
AK		 8973 if (imsm_count_failed(super, dev, MAP_0) !=
8974 (map->num_members / 2))
43d5ec18
KW		 8975 return 0;
8976
bb025c2f
KW		 8977 /* iterate through devices to mark removed disks as spare */
8978 for (dm = super->disks; dm; dm = dm->next) {
8979 if (dm->disk.status & FAILED_DISK) {
8980 int idx = dm->index;
8981 /* update indexes on the disk list */
8982/* FIXME this loop-with-the-loop looks wrong, I'm not convinced
8983 the index values will end up being correct.... NB */
8984 for (du = super->disks; du; du = du->next)
8985 if (du->index > idx)
8986 du->index--;
8987 /* mark as spare disk */
a8619d23 8988 mark_spare(dm);
bb025c2f
KW		 8989 }
8990 }
bb025c2f
KW		 8991 /* update map */
8992 map->num_members = map->num_members / 2;
8993 map->map_state = IMSM_T_STATE_NORMAL;
8994 map->num_domains = 1;
8995 map->raid_level = 0;
8996 map->failed_disk_num = -1;
8997 }
8998
8ca6df95
KW		 8999 if (u->direction == R0_TO_R10) {
9000 void **space;
9001 /* update slots in current disk list */
9002 for (dm = super->disks; dm; dm = dm->next) {
9003 if (dm->index >= 0)
9004 dm->index *= 2;
9005 }
9006 /* create new *missing* disks */
9007 for (i = 0; i < map->num_members; i++) {
9008 space = *space_list;
9009 if (!space)
9010 continue;
9011 *space_list = *space;
9012 du = (void *)space;
9013 memcpy(du, super->disks, sizeof(*du));
8ca6df95
KW		 9014 du->fd = -1;
9015 du->minor = 0;
9016 du->major = 0;
9017 du->index = (i * 2) + 1;
9018 sprintf((char *)du->disk.serial,
9019 " MISSING_%d", du->index);
9020 sprintf((char *)du->serial,
9021 "MISSING_%d", du->index);
9022 du->next = super->missing;
9023 super->missing = du;
9024 }
9025 /* create new dev and map */
9026 space = *space_list;
9027 if (!space)
9028 return 0;
9029 *space_list = *space;
9030 dev_new = (void *)space;
9031 memcpy(dev_new, dev, sizeof(*dev));
9032 /* update new map */
238c0a71 9033 map = get_imsm_map(dev_new, MAP_0);
8ca6df95 9034 map->num_members = map->num_members * 2;
1a2487c2 9035 map->map_state = IMSM_T_STATE_DEGRADED;
8ca6df95
KW		 9036 map->num_domains = 2;
9037 map->raid_level = 1;
9038 /* replace dev<->dev_new */
9039 dv->dev = dev_new;
9040 }
bb025c2f
KW		 9041 /* update disk order table */
9042 for (du = super->disks; du; du = du->next)
9043 if (du->index >= 0)
9044 set_imsm_ord_tbl_ent(map, du->index, du->index);
8ca6df95 9045 for (du = super->missing; du; du = du->next)
1a2487c2
KW		 9046 if (du->index >= 0) {
9047 set_imsm_ord_tbl_ent(map, du->index, du->index);
4c9e8c1e 9048 mark_missing(super, dv->dev, &du->disk, du->index);
1a2487c2 9049 }
bb025c2f
KW		 9050
9051 return 1;
9052}
9053
e8319a19
DW		 9054static void imsm_process_update(struct supertype *st,
9055 struct metadata_update *update)
9056{
9057 /**
9058 * crack open the metadata_update envelope to find the update record
9059 * update can be one of:
d195167d
AK		 9060 * update_reshape_container_disks - all the arrays in the container
9061 * are being reshaped to have more devices. We need to mark
9062 * the arrays for general migration and convert selected spares
9063 * into active devices.
9064 * update_activate_spare - a spare device has replaced a failed
1011e834
N		 9065 * device in an array, update the disk_ord_tbl. If this disk is
9066 * present in all member arrays then also clear the SPARE_DISK
9067 * flag
d195167d
AK		 9068 * update_create_array
9069 * update_kill_array
9070 * update_rename_array
9071 * update_add_remove_disk
e8319a19
DW		 9072 */
9073 struct intel_super *super = st->sb;
4d7b1503 9074 struct imsm_super *mpb;
e8319a19
DW		 9075 enum imsm_update_type type = *(enum imsm_update_type *) update->buf;
9076
4d7b1503
DW		 9077 /* update requires a larger buf but the allocation failed */
9078 if (super->next_len && !super->next_buf) {
9079 super->next_len = 0;
9080 return;
9081 }
9082
9083 if (super->next_buf) {
9084 memcpy(super->next_buf, super->buf, super->len);
9085 free(super->buf);
9086 super->len = super->next_len;
9087 super->buf = super->next_buf;
9088
9089 super->next_len = 0;
9090 super->next_buf = NULL;
9091 }
9092
9093 mpb = super->anchor;
9094
e8319a19 9095 switch (type) {
0ec5d470
AK		 9096 case update_general_migration_checkpoint: {
9097 struct intel_dev *id;
9098 struct imsm_update_general_migration_checkpoint *u =
9099 (void *)update->buf;
9100
1ade5cc1 9101 dprintf("called for update_general_migration_checkpoint\n");
0ec5d470
AK		 9102
9103 /* find device under general migration */
9104 for (id = super->devlist ; id; id = id->next) {
9105 if (is_gen_migration(id->dev)) {
9106 id->dev->vol.curr_migr_unit =
9107 __cpu_to_le32(u->curr_migr_unit);
9108 super->updates_pending++;
9109 }
9110 }
9111 break;
9112 }
bb025c2f
KW		 9113 case update_takeover: {
9114 struct imsm_update_takeover *u = (void *)update->buf;
1a2487c2
KW		 9115 if (apply_takeover_update(u, super, &update->space_list)) {
9116 imsm_update_version_info(super);
bb025c2f 9117 super->updates_pending++;
1a2487c2 9118 }
bb025c2f
KW		 9119 break;
9120 }
9121
78b10e66 9122 case update_reshape_container_disks: {
d195167d 9123 struct imsm_update_reshape *u = (void *)update->buf;
2e5dc010
N		 9124 if (apply_reshape_container_disks_update(
9125 u, super, &update->space_list))
9126 super->updates_pending++;
78b10e66
N		 9127 break;
9128 }
48c5303a 9129 case update_reshape_migration: {
a29911da
PC		 9130 struct imsm_update_reshape_migration *u = (void *)update->buf;
9131 if (apply_reshape_migration_update(
9132 u, super, &update->space_list))
9133 super->updates_pending++;
48c5303a
PC		 9134 break;
9135 }
f3871fdc
AK		 9136 case update_size_change: {
9137 struct imsm_update_size_change *u = (void *)update->buf;
9138 if (apply_size_change_update(u, super))
9139 super->updates_pending++;
9140 break;
9141 }
e8319a19 9142 case update_activate_spare: {
1011e834 9143 struct imsm_update_activate_spare *u = (void *) update->buf;
061d7da3
LO		 9144 if (apply_update_activate_spare(u, super, st->arrays))
9145 super->updates_pending++;
8273f55e
DW		 9146 break;
9147 }
9148 case update_create_array: {
9149 /* someone wants to create a new array, we need to be aware of
9150 * a few races/collisions:
9151 * 1/ 'Create' called by two separate instances of mdadm
9152 * 2/ 'Create' versus 'activate_spare': mdadm has chosen
9153 * devices that have since been assimilated via
9154 * activate_spare.
9155 * In the event this update can not be carried out mdadm will
9156 * (FIX ME) notice that its update did not take hold.
9157 */
9158 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 9159 struct intel_dev *dv;
8273f55e
DW		 9160 struct imsm_dev *dev;
9161 struct imsm_map *map, *new_map;
9162 unsigned long long start, end;
9163 unsigned long long new_start, new_end;
9164 int i;
54c2c1ea
DW		 9165 struct disk_info *inf;
9166 struct dl *dl;
8273f55e
DW		 9167
9168 /* handle racing creates: first come first serve */
9169 if (u->dev_idx < mpb->num_raid_devs) {
1ade5cc1 9170 dprintf("subarray %d already defined\n", u->dev_idx);
ba2de7ba 9171 goto create_error;
8273f55e
DW		 9172 }
9173
9174 /* check update is next in sequence */
9175 if (u->dev_idx != mpb->num_raid_devs) {
1ade5cc1
N		 9176 dprintf("can not create array %d expected index %d\n",
9177 u->dev_idx, mpb->num_raid_devs);
ba2de7ba 9178 goto create_error;
8273f55e
DW		 9179 }
9180
238c0a71 9181 new_map = get_imsm_map(&u->dev, MAP_0);
5551b113
CA		 9182 new_start = pba_of_lba0(new_map);
9183 new_end = new_start + blocks_per_member(new_map);
54c2c1ea 9184 inf = get_disk_info(u);
8273f55e
DW		 9185
9186 /* handle activate_spare versus create race:
9187 * check to make sure that overlapping arrays do not include
9188 * overalpping disks
9189 */
9190 for (i = 0; i < mpb->num_raid_devs; i++) {
949c47a0 9191 dev = get_imsm_dev(super, i);
238c0a71 9192 map = get_imsm_map(dev, MAP_0);
5551b113
CA		 9193 start = pba_of_lba0(map);
9194 end = start + blocks_per_member(map);
8273f55e
DW		 9195 if ((new_start >= start && new_start <= end) ||
9196 (start >= new_start && start <= new_end))
54c2c1ea
DW		 9197 /* overlap */;
9198 else
9199 continue;
9200
9201 if (disks_overlap(super, i, u)) {
1ade5cc1 9202 dprintf("arrays overlap\n");
ba2de7ba 9203 goto create_error;
8273f55e
DW		 9204 }
9205 }
8273f55e 9206
949c47a0
DW		 9207 /* check that prepare update was successful */
9208 if (!update->space) {
1ade5cc1 9209 dprintf("prepare update failed\n");
ba2de7ba 9210 goto create_error;
949c47a0
DW		 9211 }
9212
54c2c1ea
DW		 9213 /* check that all disks are still active before committing
9214 * changes. FIXME: could we instead handle this by creating a
9215 * degraded array? That's probably not what the user expects,
9216 * so better to drop this update on the floor.
9217 */
9218 for (i = 0; i < new_map->num_members; i++) {
9219 dl = serial_to_dl(inf[i].serial, super);
9220 if (!dl) {
1ade5cc1 9221 dprintf("disk disappeared\n");
ba2de7ba 9222 goto create_error;
54c2c1ea 9223 }
949c47a0
DW		 9224 }
9225
8273f55e 9226 super->updates_pending++;
54c2c1ea
DW		 9227
9228 /* convert spares to members and fixup ord_tbl */
9229 for (i = 0; i < new_map->num_members; i++) {
9230 dl = serial_to_dl(inf[i].serial, super);
9231 if (dl->index == -1) {
9232 dl->index = mpb->num_disks;
9233 mpb->num_disks++;
9234 dl->disk.status |= CONFIGURED_DISK;
9235 dl->disk.status &= ~SPARE_DISK;
9236 }
9237 set_imsm_ord_tbl_ent(new_map, i, dl->index);
9238 }
9239
ba2de7ba
DW		 9240 dv = update->space;
9241 dev = dv->dev;
949c47a0
DW		 9242 update->space = NULL;
9243 imsm_copy_dev(dev, &u->dev);
ba2de7ba
DW		 9244 dv->index = u->dev_idx;
9245 dv->next = super->devlist;
9246 super->devlist = dv;
8273f55e 9247 mpb->num_raid_devs++;
8273f55e 9248
4d1313e9 9249 imsm_update_version_info(super);
8273f55e 9250 break;
ba2de7ba
DW		 9251 create_error:
9252 /* mdmon knows how to release update->space, but not
9253 * ((struct intel_dev *) update->space)->dev
9254 */
9255 if (update->space) {
9256 dv = update->space;
9257 free(dv->dev);
9258 }
8273f55e 9259 break;
e8319a19 9260 }
33414a01
DW		 9261 case update_kill_array: {
9262 struct imsm_update_kill_array *u = (void *) update->buf;
9263 int victim = u->dev_idx;
9264 struct active_array *a;
9265 struct intel_dev **dp;
9266 struct imsm_dev *dev;
9267
9268 /* sanity check that we are not affecting the uuid of
9269 * active arrays, or deleting an active array
9270 *
9271 * FIXME when immutable ids are available, but note that
9272 * we'll also need to fixup the invalidated/active
9273 * subarray indexes in mdstat
9274 */
9275 for (a = st->arrays; a; a = a->next)
9276 if (a->info.container_member >= victim)
9277 break;
9278 /* by definition if mdmon is running at least one array
9279 * is active in the container, so checking
9280 * mpb->num_raid_devs is just extra paranoia
9281 */
9282 dev = get_imsm_dev(super, victim);
9283 if (a || !dev || mpb->num_raid_devs == 1) {
9284 dprintf("failed to delete subarray-%d\n", victim);
9285 break;
9286 }
9287
9288 for (dp = &super->devlist; *dp;)
f21e18ca 9289 if ((*dp)->index == (unsigned)super->current_vol) {
33414a01
DW		 9290 *dp = (*dp)->next;
9291 } else {
f21e18ca 9292 if ((*dp)->index > (unsigned)victim)
33414a01
DW		 9293 (*dp)->index--;
9294 dp = &(*dp)->next;
9295 }
9296 mpb->num_raid_devs--;
9297 super->updates_pending++;
9298 break;
9299 }
aa534678
DW		 9300 case update_rename_array: {
9301 struct imsm_update_rename_array *u = (void *) update->buf;
9302 char name[MAX_RAID_SERIAL_LEN+1];
9303 int target = u->dev_idx;
9304 struct active_array *a;
9305 struct imsm_dev *dev;
9306
9307 /* sanity check that we are not affecting the uuid of
9308 * an active array
9309 */
9310 snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name);
9311 name[MAX_RAID_SERIAL_LEN] = '\0';
9312 for (a = st->arrays; a; a = a->next)
9313 if (a->info.container_member == target)
9314 break;
9315 dev = get_imsm_dev(super, u->dev_idx);
9316 if (a || !dev || !check_name(super, name, 1)) {
9317 dprintf("failed to rename subarray-%d\n", target);
9318 break;
9319 }
9320
cdbe98cd 9321 snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name);
aa534678
DW		 9322 super->updates_pending++;
9323 break;
9324 }
1a64be56 9325 case update_add_remove_disk: {
43dad3d6 9326 /* we may be able to repair some arrays if disks are
095b8088 9327 * being added, check the status of add_remove_disk
1a64be56
LM		 9328 * if discs has been added.
9329 */
9330 if (add_remove_disk_update(super)) {
43dad3d6 9331 struct active_array *a;
072b727f
DW		 9332
9333 super->updates_pending++;
1a64be56 9334 for (a = st->arrays; a; a = a->next)
43dad3d6
DW		 9335 a->check_degraded = 1;
9336 }
43dad3d6 9337 break;
e8319a19 9338 }
bbab0940
TM		 9339 case update_prealloc_badblocks_mem:
9340 break;
1a64be56 9341 default:
7a862a02 9342 pr_err("error: unsuported process update type:(type: %d)\n", type);
1a64be56 9343 }
e8319a19 9344}
88758e9d 9345
bc0b9d34
PC		 9346static struct mdinfo *get_spares_for_grow(struct supertype *st);
9347
5fe6f031
N		 9348static int imsm_prepare_update(struct supertype *st,
9349 struct metadata_update *update)
8273f55e 9350{
949c47a0 9351 /**
4d7b1503
DW		 9352 * Allocate space to hold new disk entries, raid-device entries or a new
9353 * mpb if necessary. The manager synchronously waits for updates to
9354 * complete in the monitor, so new mpb buffers allocated here can be
9355 * integrated by the monitor thread without worrying about live pointers
9356 * in the manager thread.
8273f55e 9357 */
095b8088 9358 enum imsm_update_type type;
4d7b1503 9359 struct intel_super *super = st->sb;
f36a9ecd 9360 unsigned int sector_size = super->sector_size;
4d7b1503
DW		 9361 struct imsm_super *mpb = super->anchor;
9362 size_t buf_len;
9363 size_t len = 0;
949c47a0 9364
095b8088
N		 9365 if (update->len < (int)sizeof(type))
9366 return 0;
9367
9368 type = *(enum imsm_update_type *) update->buf;
9369
949c47a0 9370 switch (type) {
0ec5d470 9371 case update_general_migration_checkpoint:
095b8088
N		 9372 if (update->len < (int)sizeof(struct imsm_update_general_migration_checkpoint))
9373 return 0;
1ade5cc1 9374 dprintf("called for update_general_migration_checkpoint\n");
0ec5d470 9375 break;
abedf5fc
KW		 9376 case update_takeover: {
9377 struct imsm_update_takeover *u = (void *)update->buf;
095b8088
N		 9378 if (update->len < (int)sizeof(*u))
9379 return 0;
abedf5fc
KW		 9380 if (u->direction == R0_TO_R10) {
9381 void **tail = (void **)&update->space_list;
9382 struct imsm_dev *dev = get_imsm_dev(super, u->subarray);
238c0a71 9383 struct imsm_map *map = get_imsm_map(dev, MAP_0);
abedf5fc
KW		 9384 int num_members = map->num_members;
9385 void *space;
9386 int size, i;
abedf5fc
KW		 9387 /* allocate memory for added disks */
9388 for (i = 0; i < num_members; i++) {
9389 size = sizeof(struct dl);
503975b9 9390 space = xmalloc(size);
abedf5fc
KW		 9391 *tail = space;
9392 tail = space;
9393 *tail = NULL;
9394 }
9395 /* allocate memory for new device */
9396 size = sizeof_imsm_dev(super->devlist->dev, 0) +
9397 (num_members * sizeof(__u32));
503975b9
N		 9398 space = xmalloc(size);
9399 *tail = space;
9400 tail = space;
9401 *tail = NULL;
9402 len = disks_to_mpb_size(num_members * 2);
abedf5fc
KW		 9403 }
9404
9405 break;
9406 }
78b10e66 9407 case update_reshape_container_disks: {
d195167d
AK		 9408 /* Every raid device in the container is about to
9409 * gain some more devices, and we will enter a
9410 * reconfiguration.
9411 * So each 'imsm_map' will be bigger, and the imsm_vol
9412 * will now hold 2 of them.
9413 * Thus we need new 'struct imsm_dev' allocations sized
9414 * as sizeof_imsm_dev but with more devices in both maps.
9415 */
9416 struct imsm_update_reshape *u = (void *)update->buf;
9417 struct intel_dev *dl;
9418 void **space_tail = (void**)&update->space_list;
9419
095b8088
N		 9420 if (update->len < (int)sizeof(*u))
9421 return 0;
9422
1ade5cc1 9423 dprintf("for update_reshape\n");
d195167d
AK		 9424
9425 for (dl = super->devlist; dl; dl = dl->next) {
9426 int size = sizeof_imsm_dev(dl->dev, 1);
9427 void *s;
d677e0b8
AK		 9428 if (u->new_raid_disks > u->old_raid_disks)
9429 size += sizeof(__u32)*2*
9430 (u->new_raid_disks - u->old_raid_disks);
503975b9 9431 s = xmalloc(size);
d195167d
AK		 9432 *space_tail = s;
9433 space_tail = s;
9434 *space_tail = NULL;
9435 }
9436
9437 len = disks_to_mpb_size(u->new_raid_disks);
9438 dprintf("New anchor length is %llu\n", (unsigned long long)len);
78b10e66
N		 9439 break;
9440 }
48c5303a 9441 case update_reshape_migration: {
bc0b9d34
PC		 9442 /* for migration level 0->5 we need to add disks
9443 * so the same as for container operation we will copy
9444 * device to the bigger location.
9445 * in memory prepared device and new disk area are prepared
9446 * for usage in process update
9447 */
9448 struct imsm_update_reshape_migration *u = (void *)update->buf;
9449 struct intel_dev *id;
9450 void **space_tail = (void **)&update->space_list;
9451 int size;
9452 void *s;
9453 int current_level = -1;
9454
095b8088
N		 9455 if (update->len < (int)sizeof(*u))
9456 return 0;
9457
1ade5cc1 9458 dprintf("for update_reshape\n");
bc0b9d34
PC		 9459
9460 /* add space for bigger array in update
9461 */
9462 for (id = super->devlist; id; id = id->next) {
9463 if (id->index == (unsigned)u->subdev) {
9464 size = sizeof_imsm_dev(id->dev, 1);
9465 if (u->new_raid_disks > u->old_raid_disks)
9466 size += sizeof(__u32)*2*
9467 (u->new_raid_disks - u->old_raid_disks);
503975b9 9468 s = xmalloc(size);
bc0b9d34
PC		 9469 *space_tail = s;
9470 space_tail = s;
9471 *space_tail = NULL;
9472 break;
9473 }
9474 }
9475 if (update->space_list == NULL)
9476 break;
9477
9478 /* add space for disk in update
9479 */
9480 size = sizeof(struct dl);
503975b9 9481 s = xmalloc(size);
bc0b9d34
PC		 9482 *space_tail = s;
9483 space_tail = s;
9484 *space_tail = NULL;
9485
9486 /* add spare device to update
9487 */
9488 for (id = super->devlist ; id; id = id->next)
9489 if (id->index == (unsigned)u->subdev) {
9490 struct imsm_dev *dev;
9491 struct imsm_map *map;
9492
9493 dev = get_imsm_dev(super, u->subdev);
238c0a71 9494 map = get_imsm_map(dev, MAP_0);
bc0b9d34
PC		 9495 current_level = map->raid_level;
9496 break;
9497 }
089f9d79 9498 if (u->new_level == 5 && u->new_level != current_level) {
bc0b9d34
PC		 9499 struct mdinfo *spares;
9500
9501 spares = get_spares_for_grow(st);
9502 if (spares) {
9503 struct dl *dl;
9504 struct mdinfo *dev;
9505
9506 dev = spares->devs;
9507 if (dev) {
9508 u->new_disks[0] =
9509 makedev(dev->disk.major,
9510 dev->disk.minor);
9511 dl = get_disk_super(super,
9512 dev->disk.major,
9513 dev->disk.minor);
9514 dl->index = u->old_raid_disks;
9515 dev = dev->next;
9516 }
9517 sysfs_free(spares);
9518 }
9519 }
9520 len = disks_to_mpb_size(u->new_raid_disks);
9521 dprintf("New anchor length is %llu\n", (unsigned long long)len);
48c5303a
PC		 9522 break;
9523 }
f3871fdc 9524 case update_size_change: {
095b8088
N		 9525 if (update->len < (int)sizeof(struct imsm_update_size_change))
9526 return 0;
9527 break;
9528 }
9529 case update_activate_spare: {
9530 if (update->len < (int)sizeof(struct imsm_update_activate_spare))
9531 return 0;
f3871fdc
AK		 9532 break;
9533 }
949c47a0
DW		 9534 case update_create_array: {
9535 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 9536 struct intel_dev *dv;
54c2c1ea 9537 struct imsm_dev *dev = &u->dev;
238c0a71 9538 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW		 9539 struct dl *dl;
9540 struct disk_info *inf;
9541 int i;
9542 int activate = 0;
949c47a0 9543
095b8088
N		 9544 if (update->len < (int)sizeof(*u))
9545 return 0;
9546
54c2c1ea
DW		 9547 inf = get_disk_info(u);
9548 len = sizeof_imsm_dev(dev, 1);
ba2de7ba 9549 /* allocate a new super->devlist entry */
503975b9
N		 9550 dv = xmalloc(sizeof(*dv));
9551 dv->dev = xmalloc(len);
9552 update->space = dv;
949c47a0 9553
54c2c1ea
DW		 9554 /* count how many spares will be converted to members */
9555 for (i = 0; i < map->num_members; i++) {
9556 dl = serial_to_dl(inf[i].serial, super);
9557 if (!dl) {
9558 /* hmm maybe it failed?, nothing we can do about
9559 * it here
9560 */
9561 continue;
9562 }
9563 if (count_memberships(dl, super) == 0)
9564 activate++;
9565 }
9566 len += activate * sizeof(struct imsm_disk);
949c47a0 9567 break;
095b8088
N		 9568 }
9569 case update_kill_array: {
9570 if (update->len < (int)sizeof(struct imsm_update_kill_array))
9571 return 0;
949c47a0
DW		 9572 break;
9573 }
095b8088
N		 9574 case update_rename_array: {
9575 if (update->len < (int)sizeof(struct imsm_update_rename_array))
9576 return 0;
9577 break;
9578 }
9579 case update_add_remove_disk:
9580 /* no update->len needed */
9581 break;
bbab0940
TM		 9582 case update_prealloc_badblocks_mem:
9583 super->extra_space += sizeof(struct bbm_log) -
9584 get_imsm_bbm_log_size(super->bbm_log);
9585 break;
095b8088
N		 9586 default:
9587 return 0;
949c47a0 9588 }
8273f55e 9589
4d7b1503
DW		 9590 /* check if we need a larger metadata buffer */
9591 if (super->next_buf)
9592 buf_len = super->next_len;
9593 else
9594 buf_len = super->len;
9595
bbab0940 9596 if (__le32_to_cpu(mpb->mpb_size) + super->extra_space + len > buf_len) {
4d7b1503
DW		 9597 /* ok we need a larger buf than what is currently allocated
9598 * if this allocation fails process_update will notice that
9599 * ->next_len is set and ->next_buf is NULL
9600 */
bbab0940
TM		 9601 buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) +
9602 super->extra_space + len, sector_size);
4d7b1503
DW		 9603 if (super->next_buf)
9604 free(super->next_buf);
9605
9606 super->next_len = buf_len;
f36a9ecd 9607 if (posix_memalign(&super->next_buf, sector_size, buf_len) == 0)
1f45a8ad
DW		 9608 memset(super->next_buf, 0, buf_len);
9609 else
4d7b1503
DW		 9610 super->next_buf = NULL;
9611 }
5fe6f031 9612 return 1;
8273f55e
DW		 9613}
9614
ae6aad82 9615/* must be called while manager is quiesced */
f21e18ca 9616static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index)
ae6aad82
DW		 9617{
9618 struct imsm_super *mpb = super->anchor;
ae6aad82
DW		 9619 struct dl *iter;
9620 struct imsm_dev *dev;
9621 struct imsm_map *map;
4c9e8c1e 9622 unsigned int i, j, num_members;
24565c9a 9623 __u32 ord;
4c9e8c1e 9624 struct bbm_log *log = super->bbm_log;
ae6aad82 9625
1ade5cc1 9626 dprintf("deleting device[%d] from imsm_super\n", index);
ae6aad82
DW		 9627
9628 /* shift all indexes down one */
9629 for (iter = super->disks; iter; iter = iter->next)
f21e18ca 9630 if (iter->index > (int)index)
ae6aad82 9631 iter->index--;
47ee5a45 9632 for (iter = super->missing; iter; iter = iter->next)
f21e18ca 9633 if (iter->index > (int)index)
47ee5a45 9634 iter->index--;
ae6aad82
DW		 9635
9636 for (i = 0; i < mpb->num_raid_devs; i++) {
9637 dev = get_imsm_dev(super, i);
238c0a71 9638 map = get_imsm_map(dev, MAP_0);
24565c9a
DW		 9639 num_members = map->num_members;
9640 for (j = 0; j < num_members; j++) {
9641 /* update ord entries being careful not to propagate
9642 * ord-flags to the first map
9643 */
238c0a71 9644 ord = get_imsm_ord_tbl_ent(dev, j, MAP_X);
ae6aad82 9645
24565c9a
DW		 9646 if (ord_to_idx(ord) <= index)
9647 continue;
ae6aad82 9648
238c0a71 9649 map = get_imsm_map(dev, MAP_0);
24565c9a 9650 set_imsm_ord_tbl_ent(map, j, ord_to_idx(ord - 1));
238c0a71 9651 map = get_imsm_map(dev, MAP_1);
24565c9a
DW		 9652 if (map)
9653 set_imsm_ord_tbl_ent(map, j, ord - 1);
ae6aad82
DW		 9654 }
9655 }
9656
4c9e8c1e
TM		 9657 for (i = 0; i < log->entry_count; i++) {
9658 struct bbm_log_entry *entry = &log->marked_block_entries[i];
9659
9660 if (entry->disk_ordinal <= index)
9661 continue;
9662 entry->disk_ordinal--;
9663 }
9664
ae6aad82
DW		 9665 mpb->num_disks--;
9666 super->updates_pending++;
24565c9a
DW		 9667 if (*dlp) {
9668 struct dl *dl = *dlp;
9669
9670 *dlp = (*dlp)->next;
9671 __free_imsm_disk(dl);
9672 }
ae6aad82 9673}
9e2d750d 9674#endif /* MDASSEMBLE */
9a717282
AK		 9675
9676static void close_targets(int *targets, int new_disks)
9677{
9678 int i;
9679
9680 if (!targets)
9681 return;
9682
9683 for (i = 0; i < new_disks; i++) {
9684 if (targets[i] >= 0) {
9685 close(targets[i]);
9686 targets[i] = -1;
9687 }
9688 }
9689}
9690
9691static int imsm_get_allowed_degradation(int level, int raid_disks,
9692 struct intel_super *super,
9693 struct imsm_dev *dev)
9694{
9695 switch (level) {
bf5cf7c7 9696 case 1:
9a717282
AK		 9697 case 10:{
9698 int ret_val = 0;
9699 struct imsm_map *map;
9700 int i;
9701
9702 ret_val = raid_disks/2;
9703 /* check map if all disks pairs not failed
9704 * in both maps
9705 */
238c0a71 9706 map = get_imsm_map(dev, MAP_0);
9a717282
AK		 9707 for (i = 0; i < ret_val; i++) {
9708 int degradation = 0;
9709 if (get_imsm_disk(super, i) == NULL)
9710 degradation++;
9711 if (get_imsm_disk(super, i + 1) == NULL)
9712 degradation++;
9713 if (degradation == 2)
9714 return 0;
9715 }
238c0a71 9716 map = get_imsm_map(dev, MAP_1);
9a717282
AK		 9717 /* if there is no second map
9718 * result can be returned
9719 */
9720 if (map == NULL)
9721 return ret_val;
9722 /* check degradation in second map
9723 */
9724 for (i = 0; i < ret_val; i++) {
9725 int degradation = 0;
9726 if (get_imsm_disk(super, i) == NULL)
9727 degradation++;
9728 if (get_imsm_disk(super, i + 1) == NULL)
9729 degradation++;
9730 if (degradation == 2)
9731 return 0;
9732 }
9733 return ret_val;
9734 }
9735 case 5:
9736 return 1;
9737 case 6:
9738 return 2;
9739 default:
9740 return 0;
9741 }
9742}
9743
687629c2
AK		 9744/*******************************************************************************
9745 * Function: open_backup_targets
9746 * Description: Function opens file descriptors for all devices given in
9747 * info->devs
9748 * Parameters:
9749 * info : general array info
9750 * raid_disks : number of disks
9751 * raid_fds : table of device's file descriptors
9a717282
AK		 9752 * super : intel super for raid10 degradation check
9753 * dev : intel device for raid10 degradation check
687629c2
AK		 9754 * Returns:
9755 * 0 : success
9756 * -1 : fail
9757 ******************************************************************************/
9a717282
AK		 9758int open_backup_targets(struct mdinfo *info, int raid_disks, int *raid_fds,
9759 struct intel_super *super, struct imsm_dev *dev)
687629c2
AK		 9760{
9761 struct mdinfo *sd;
f627f5ad 9762 int i;
9a717282 9763 int opened = 0;
f627f5ad
AK		 9764
9765 for (i = 0; i < raid_disks; i++)
9766 raid_fds[i] = -1;
687629c2
AK		 9767
9768 for (sd = info->devs ; sd ; sd = sd->next) {
9769 char *dn;
9770
9771 if (sd->disk.state & (1<<MD_DISK_FAULTY)) {
9772 dprintf("disk is faulty!!\n");
9773 continue;
9774 }
9775
089f9d79 9776 if (sd->disk.raid_disk >= raid_disks || sd->disk.raid_disk < 0)
687629c2
AK		 9777 continue;
9778
9779 dn = map_dev(sd->disk.major,
9780 sd->disk.minor, 1);
9781 raid_fds[sd->disk.raid_disk] = dev_open(dn, O_RDWR);
9782 if (raid_fds[sd->disk.raid_disk] < 0) {
e12b3daa 9783 pr_err("cannot open component\n");
9a717282 9784 continue;
687629c2 9785 }
9a717282
AK		 9786 opened++;
9787 }
9788 /* check if maximum array degradation level is not exceeded
9789 */
9790 if ((raid_disks - opened) >
089f9d79
JS		 9791 imsm_get_allowed_degradation(info->new_level, raid_disks,
9792 super, dev)) {
e12b3daa 9793 pr_err("Not enough disks can be opened.\n");
9a717282
AK		 9794 close_targets(raid_fds, raid_disks);
9795 return -2;
687629c2
AK		 9796 }
9797 return 0;
9798}
9799
d31ad643
PB		 9800/*******************************************************************************
9801 * Function: validate_container_imsm
9802 * Description: This routine validates container after assemble,
9803 * eg. if devices in container are under the same controller.
9804 *
9805 * Parameters:
9806 * info : linked list with info about devices used in array
9807 * Returns:
9808 * 1 : HBA mismatch
9809 * 0 : Success
9810 ******************************************************************************/
9811int validate_container_imsm(struct mdinfo *info)
9812{
6b781d33
AP		 9813 if (check_env("IMSM_NO_PLATFORM"))
9814 return 0;
d31ad643 9815
6b781d33
AP		 9816 struct sys_dev *idev;
9817 struct sys_dev *hba = NULL;
9818 struct sys_dev *intel_devices = find_intel_devices();
9819 char *dev_path = devt_to_devpath(makedev(info->disk.major,
9820 info->disk.minor));
9821
9822 for (idev = intel_devices; idev; idev = idev->next) {
9823 if (dev_path && strstr(dev_path, idev->path)) {
9824 hba = idev;
9825 break;
d31ad643 9826 }
6b781d33
AP		 9827 }
9828 if (dev_path)
d31ad643
PB		 9829 free(dev_path);
9830
6b781d33
AP		 9831 if (!hba) {
9832 pr_err("WARNING - Cannot detect HBA for device %s!\n",
9833 devid2kname(makedev(info->disk.major, info->disk.minor)));
9834 return 1;
9835 }
9836
9837 const struct imsm_orom *orom = get_orom_by_device_id(hba->dev_id);
9838 struct mdinfo *dev;
9839
9840 for (dev = info->next; dev; dev = dev->next) {
9841 dev_path = devt_to_devpath(makedev(dev->disk.major, dev->disk.minor));
9842
9843 struct sys_dev *hba2 = NULL;
9844 for (idev = intel_devices; idev; idev = idev->next) {
9845 if (dev_path && strstr(dev_path, idev->path)) {
9846 hba2 = idev;
9847 break;
d31ad643
PB		 9848 }
9849 }
6b781d33
AP		 9850 if (dev_path)
9851 free(dev_path);
9852
9853 const struct imsm_orom *orom2 = hba2 == NULL ? NULL :
9854 get_orom_by_device_id(hba2->dev_id);
9855
9856 if (hba2 && hba->type != hba2->type) {
9857 pr_err("WARNING - HBAs of devices do not match %s != %s\n",
9858 get_sys_dev_type(hba->type), get_sys_dev_type(hba2->type));
9859 return 1;
9860 }
9861
07cb1e57 9862 if (orom != orom2) {
6b781d33
AP		 9863 pr_err("WARNING - IMSM container assembled with disks under different HBAs!\n"
9864 " This operation is not supported and can lead to data loss.\n");
9865 return 1;
9866 }
9867
9868 if (!orom) {
9869 pr_err("WARNING - IMSM container assembled with disks under HBAs without IMSM platform support!\n"
9870 " This operation is not supported and can lead to data loss.\n");
9871 return 1;
9872 }
d31ad643 9873 }
6b781d33 9874
d31ad643
PB		 9875 return 0;
9876}
9e2d750d 9877#ifndef MDASSEMBLE
6f50473f
TM		 9878/*******************************************************************************
9879* Function: imsm_record_badblock
9880* Description: This routine stores new bad block record in BBM log
9881*
9882* Parameters:
9883* a : array containing a bad block
9884* slot : disk number containing a bad block
9885* sector : bad block sector
9886* length : bad block sectors range
9887* Returns:
9888* 1 : Success
9889* 0 : Error
9890******************************************************************************/
9891static int imsm_record_badblock(struct active_array *a, int slot,
9892 unsigned long long sector, int length)
9893{
9894 struct intel_super *super = a->container->sb;
9895 int ord;
9896 int ret;
9897
9898 ord = imsm_disk_slot_to_ord(a, slot);
9899 if (ord < 0)
9900 return 0;
9901
9902 ret = record_new_badblock(super->bbm_log, ord_to_idx(ord), sector,
9903 length);
9904 if (ret)
9905 super->updates_pending++;
9906
9907 return ret;
9908}
c07a5a4f
TM		 9909/*******************************************************************************
9910* Function: imsm_clear_badblock
9911* Description: This routine clears bad block record from BBM log
9912*
9913* Parameters:
9914* a : array containing a bad block
9915* slot : disk number containing a bad block
9916* sector : bad block sector
9917* length : bad block sectors range
9918* Returns:
9919* 1 : Success
9920* 0 : Error
9921******************************************************************************/
9922static int imsm_clear_badblock(struct active_array *a, int slot,
9923 unsigned long long sector, int length)
9924{
9925 struct intel_super *super = a->container->sb;
9926 int ord;
9927 int ret;
9928
9929 ord = imsm_disk_slot_to_ord(a, slot);
9930 if (ord < 0)
9931 return 0;
9932
9933 ret = clear_badblock(super->bbm_log, ord_to_idx(ord), sector, length);
9934 if (ret)
9935 super->updates_pending++;
9936
9937 return ret;
9938}
928f1424
TM		 9939/*******************************************************************************
9940* Function: imsm_get_badblocks
9941* Description: This routine get list of bad blocks for an array
9942*
9943* Parameters:
9944* a : array
9945* slot : disk number
9946* Returns:
9947* bb : structure containing bad blocks
9948* NULL : error
9949******************************************************************************/
9950static struct md_bb *imsm_get_badblocks(struct active_array *a, int slot)
9951{
9952 int inst = a->info.container_member;
9953 struct intel_super *super = a->container->sb;
9954 struct imsm_dev *dev = get_imsm_dev(super, inst);
9955 struct imsm_map *map = get_imsm_map(dev, MAP_0);
9956 int ord;
9957
9958 ord = imsm_disk_slot_to_ord(a, slot);
9959 if (ord < 0)
9960 return NULL;
9961
9962 get_volume_badblocks(super->bbm_log, ord_to_idx(ord), pba_of_lba0(map),
9963 blocks_per_member(map), &super->bb);
9964
9965 return &super->bb;
9966}
27156a57
TM		 9967/*******************************************************************************
9968* Function: examine_badblocks_imsm
9969* Description: Prints list of bad blocks on a disk to the standard output
9970*
9971* Parameters:
9972* st : metadata handler
9973* fd : open file descriptor for device
9974* devname : device name
9975* Returns:
9976* 0 : Success
9977* 1 : Error
9978******************************************************************************/
9979static int examine_badblocks_imsm(struct supertype *st, int fd, char *devname)
9980{
9981 struct intel_super *super = st->sb;
9982 struct bbm_log *log = super->bbm_log;
9983 struct dl *d = NULL;
9984 int any = 0;
9985
9986 for (d = super->disks; d ; d = d->next) {
9987 if (strcmp(d->devname, devname) == 0)
9988 break;
9989 }
9990
9991 if ((d == NULL) || (d->index < 0)) { /* serial mismatch probably */
9992 pr_err("%s doesn't appear to be part of a raid array\n",
9993 devname);
9994 return 1;
9995 }
9996
9997 if (log != NULL) {
9998 unsigned int i;
9999 struct bbm_log_entry *entry = &log->marked_block_entries[0];
10000
10001 for (i = 0; i < log->entry_count; i++) {
10002 if (entry[i].disk_ordinal == d->index) {
10003 unsigned long long sector = __le48_to_cpu(
10004 &entry[i].defective_block_start);
10005 int cnt = entry[i].marked_count + 1;
10006
10007 if (!any) {
10008 printf("Bad-blocks on %s:\n", devname);
10009 any = 1;
10010 }
10011
10012 printf("%20llu for %d sectors\n", sector, cnt);
10013 }
10014 }
10015 }
10016
10017 if (!any)
10018 printf("No bad-blocks list configured on %s\n", devname);
10019
10020 return 0;
10021}
687629c2
AK		 10022/*******************************************************************************
10023 * Function: init_migr_record_imsm
10024 * Description: Function inits imsm migration record
10025 * Parameters:
10026 * super : imsm internal array info
10027 * dev : device under migration
10028 * info : general array info to find the smallest device
10029 * Returns:
10030 * none
10031 ******************************************************************************/
10032void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev,
10033 struct mdinfo *info)
10034{
10035 struct intel_super *super = st->sb;
10036 struct migr_record *migr_rec = super->migr_rec;
10037 int new_data_disks;
10038 unsigned long long dsize, dev_sectors;
10039 long long unsigned min_dev_sectors = -1LLU;
10040 struct mdinfo *sd;
10041 char nm[30];
10042 int fd;
238c0a71
AK		 10043 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
10044 struct imsm_map *map_src = get_imsm_map(dev, MAP_1);
687629c2 10045 unsigned long long num_migr_units;
3ef4403c 10046 unsigned long long array_blocks;
687629c2
AK		 10047
10048 memset(migr_rec, 0, sizeof(struct migr_record));
10049 migr_rec->family_num = __cpu_to_le32(super->anchor->family_num);
10050
10051 /* only ascending reshape supported now */
10052 migr_rec->ascending_migr = __cpu_to_le32(1);
10053
10054 migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE /
10055 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
e1742195
AK		 10056 migr_rec->dest_depth_per_unit *=
10057 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
238c0a71 10058 new_data_disks = imsm_num_data_members(dev, MAP_0);
687629c2
AK		 10059 migr_rec->blocks_per_unit =
10060 __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks);
10061 migr_rec->dest_depth_per_unit =
10062 __cpu_to_le32(migr_rec->dest_depth_per_unit);
3ef4403c 10063 array_blocks = info->component_size * new_data_disks;
687629c2
AK		 10064 num_migr_units =
10065 array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit);
10066
10067 if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit))
10068 num_migr_units++;
10069 migr_rec->num_migr_units = __cpu_to_le32(num_migr_units);
10070
10071 migr_rec->post_migr_vol_cap = dev->size_low;
10072 migr_rec->post_migr_vol_cap_hi = dev->size_high;
10073
687629c2
AK		 10074 /* Find the smallest dev */
10075 for (sd = info->devs ; sd ; sd = sd->next) {
10076 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
10077 fd = dev_open(nm, O_RDONLY);
10078 if (fd < 0)
10079 continue;
10080 get_dev_size(fd, NULL, &dsize);
10081 dev_sectors = dsize / 512;
10082 if (dev_sectors < min_dev_sectors)
10083 min_dev_sectors = dev_sectors;
10084 close(fd);
10085 }
10086 migr_rec->ckpt_area_pba = __cpu_to_le32(min_dev_sectors -
10087 RAID_DISK_RESERVED_BLOCKS_IMSM_HI);
10088
10089 write_imsm_migr_rec(st);
10090
10091 return;
10092}
10093
10094/*******************************************************************************
10095 * Function: save_backup_imsm
10096 * Description: Function saves critical data stripes to Migration Copy Area
10097 * and updates the current migration unit status.
10098 * Use restore_stripes() to form a destination stripe,
10099 * and to write it to the Copy Area.
10100 * Parameters:
10101 * st : supertype information
aea93171 10102 * dev : imsm device that backup is saved for
687629c2
AK		 10103 * info : general array info
10104 * buf : input buffer
687629c2
AK		 10105 * length : length of data to backup (blocks_per_unit)
10106 * Returns:
10107 * 0 : success
10108 *, -1 : fail
10109 ******************************************************************************/
10110int save_backup_imsm(struct supertype *st,
10111 struct imsm_dev *dev,
10112 struct mdinfo *info,
10113 void *buf,
687629c2
AK		 10114 int length)
10115{
10116 int rv = -1;
10117 struct intel_super *super = st->sb;
594dc1b8
JS		 10118 unsigned long long *target_offsets;
10119 int *targets;
687629c2 10120 int i;
238c0a71 10121 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
687629c2 10122 int new_disks = map_dest->num_members;
ab724b98
AK		 10123 int dest_layout = 0;
10124 int dest_chunk;
d1877f69 10125 unsigned long long start;
238c0a71 10126 int data_disks = imsm_num_data_members(dev, MAP_0);
687629c2 10127
503975b9 10128 targets = xmalloc(new_disks * sizeof(int));
687629c2 10129
7e45b550
AK		 10130 for (i = 0; i < new_disks; i++)
10131 targets[i] = -1;
10132
503975b9 10133 target_offsets = xcalloc(new_disks, sizeof(unsigned long long));
687629c2 10134
d1877f69 10135 start = info->reshape_progress * 512;
687629c2 10136 for (i = 0; i < new_disks; i++) {
687629c2
AK		 10137 target_offsets[i] = (unsigned long long)
10138 __le32_to_cpu(super->migr_rec->ckpt_area_pba) * 512;
d1877f69
AK		 10139 /* move back copy area adderss, it will be moved forward
10140 * in restore_stripes() using start input variable
10141 */
10142 target_offsets[i] -= start/data_disks;
687629c2
AK		 10143 }
10144
9a717282
AK		 10145 if (open_backup_targets(info, new_disks, targets,
10146 super, dev))
687629c2
AK		 10147 goto abort;
10148
68eb8bc6 10149 dest_layout = imsm_level_to_layout(map_dest->raid_level);
ab724b98
AK		 10150 dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512;
10151
687629c2
AK		 10152 if (restore_stripes(targets, /* list of dest devices */
10153 target_offsets, /* migration record offsets */
10154 new_disks,
ab724b98
AK		 10155 dest_chunk,
10156 map_dest->raid_level,
10157 dest_layout,
10158 -1, /* source backup file descriptor */
10159 0, /* input buf offset
10160 * always 0 buf is already offseted */
d1877f69 10161 start,
687629c2
AK		 10162 length,
10163 buf) != 0) {
e7b84f9d 10164 pr_err("Error restoring stripes\n");
687629c2
AK		 10165 goto abort;
10166 }
10167
10168 rv = 0;
10169
10170abort:
10171 if (targets) {
9a717282 10172 close_targets(targets, new_disks);
687629c2
AK		 10173 free(targets);
10174 }
10175 free(target_offsets);
10176
10177 return rv;
10178}
10179
10180/*******************************************************************************
10181 * Function: save_checkpoint_imsm
10182 * Description: Function called for current unit status update
10183 * in the migration record. It writes it to disk.
10184 * Parameters:
10185 * super : imsm internal array info
10186 * info : general array info
10187 * Returns:
10188 * 0: success
10189 * 1: failure
0228d92c
AK		 10190 * 2: failure, means no valid migration record
10191 * / no general migration in progress /
687629c2
AK		 10192 ******************************************************************************/
10193int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state)
10194{
10195 struct intel_super *super = st->sb;
f8b72ef5
AK		 10196 unsigned long long blocks_per_unit;
10197 unsigned long long curr_migr_unit;
10198
2e062e82 10199 if (load_imsm_migr_rec(super, info) != 0) {
7a862a02 10200 dprintf("imsm: ERROR: Cannot read migration record for checkpoint save.\n");
2e062e82
AK		 10201 return 1;
10202 }
10203
f8b72ef5
AK		 10204 blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit);
10205 if (blocks_per_unit == 0) {
0228d92c
AK		 10206 dprintf("imsm: no migration in progress.\n");
10207 return 2;
687629c2 10208 }
f8b72ef5
AK		 10209 curr_migr_unit = info->reshape_progress / blocks_per_unit;
10210 /* check if array is alligned to copy area
10211 * if it is not alligned, add one to current migration unit value
10212 * this can happend on array reshape finish only
10213 */
10214 if (info->reshape_progress % blocks_per_unit)
10215 curr_migr_unit++;
687629c2
AK		 10216
10217 super->migr_rec->curr_migr_unit =
f8b72ef5 10218 __cpu_to_le32(curr_migr_unit);
687629c2
AK		 10219 super->migr_rec->rec_status = __cpu_to_le32(state);
10220 super->migr_rec->dest_1st_member_lba =
f8b72ef5
AK		 10221 __cpu_to_le32(curr_migr_unit *
10222 __le32_to_cpu(super->migr_rec->dest_depth_per_unit));
687629c2 10223 if (write_imsm_migr_rec(st) < 0) {
7a862a02 10224 dprintf("imsm: Cannot write migration record outside backup area\n");
687629c2
AK		 10225 return 1;
10226 }
10227
10228 return 0;
10229}
10230
276d77db
AK		 10231/*******************************************************************************
10232 * Function: recover_backup_imsm
10233 * Description: Function recovers critical data from the Migration Copy Area
10234 * while assembling an array.
10235 * Parameters:
10236 * super : imsm internal array info
10237 * info : general array info
10238 * Returns:
10239 * 0 : success (or there is no data to recover)
10240 * 1 : fail
10241 ******************************************************************************/
10242int recover_backup_imsm(struct supertype *st, struct mdinfo *info)
10243{
10244 struct intel_super *super = st->sb;
10245 struct migr_record *migr_rec = super->migr_rec;
594dc1b8 10246 struct imsm_map *map_dest;
276d77db
AK		 10247 struct intel_dev *id = NULL;
10248 unsigned long long read_offset;
10249 unsigned long long write_offset;
10250 unsigned unit_len;
10251 int *targets = NULL;
10252 int new_disks, i, err;
10253 char *buf = NULL;
10254 int retval = 1;
f36a9ecd 10255 unsigned int sector_size = super->sector_size;
276d77db
AK		 10256 unsigned long curr_migr_unit = __le32_to_cpu(migr_rec->curr_migr_unit);
10257 unsigned long num_migr_units = __le32_to_cpu(migr_rec->num_migr_units);
276d77db 10258 char buffer[20];
6c3560c0 10259 int skipped_disks = 0;
276d77db
AK		 10260
10261 err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20);
10262 if (err < 1)
10263 return 1;
10264
10265 /* recover data only during assemblation */
10266 if (strncmp(buffer, "inactive", 8) != 0)
10267 return 0;
10268 /* no data to recover */
10269 if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL)
10270 return 0;
10271 if (curr_migr_unit >= num_migr_units)
10272 return 1;
10273
10274 /* find device during reshape */
10275 for (id = super->devlist; id; id = id->next)
10276 if (is_gen_migration(id->dev))
10277 break;
10278 if (id == NULL)
10279 return 1;
10280
238c0a71 10281 map_dest = get_imsm_map(id->dev, MAP_0);
276d77db
AK		 10282 new_disks = map_dest->num_members;
10283
10284 read_offset = (unsigned long long)
10285 __le32_to_cpu(migr_rec->ckpt_area_pba) * 512;
10286
10287 write_offset = ((unsigned long long)
10288 __le32_to_cpu(migr_rec->dest_1st_member_lba) +
5551b113 10289 pba_of_lba0(map_dest)) * 512;
276d77db
AK		 10290
10291 unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
f36a9ecd 10292 if (posix_memalign((void **)&buf, sector_size, unit_len) != 0)
276d77db 10293 goto abort;
503975b9 10294 targets = xcalloc(new_disks, sizeof(int));
276d77db 10295
9a717282 10296 if (open_backup_targets(info, new_disks, targets, super, id->dev)) {
e7b84f9d 10297 pr_err("Cannot open some devices belonging to array.\n");
f627f5ad
AK		 10298 goto abort;
10299 }
276d77db
AK		 10300
10301 for (i = 0; i < new_disks; i++) {
6c3560c0
AK		 10302 if (targets[i] < 0) {
10303 skipped_disks++;
10304 continue;
10305 }
276d77db 10306 if (lseek64(targets[i], read_offset, SEEK_SET) < 0) {
e7b84f9d
N		 10307 pr_err("Cannot seek to block: %s\n",
10308 strerror(errno));
137debce
AK		 10309 skipped_disks++;
10310 continue;
276d77db 10311 }
9ec11d1a 10312 if ((unsigned)read(targets[i], buf, unit_len) != unit_len) {
e7b84f9d
N		 10313 pr_err("Cannot read copy area block: %s\n",
10314 strerror(errno));
137debce
AK		 10315 skipped_disks++;
10316 continue;
276d77db
AK		 10317 }
10318 if (lseek64(targets[i], write_offset, SEEK_SET) < 0) {
e7b84f9d
N		 10319 pr_err("Cannot seek to block: %s\n",
10320 strerror(errno));
137debce
AK		 10321 skipped_disks++;
10322 continue;
276d77db 10323 }
9ec11d1a 10324 if ((unsigned)write(targets[i], buf, unit_len) != unit_len) {
e7b84f9d
N		 10325 pr_err("Cannot restore block: %s\n",
10326 strerror(errno));
137debce
AK		 10327 skipped_disks++;
10328 continue;
276d77db
AK		 10329 }
10330 }
10331
137debce
AK		 10332 if (skipped_disks > imsm_get_allowed_degradation(info->new_level,
10333 new_disks,
10334 super,
10335 id->dev)) {
7a862a02 10336 pr_err("Cannot restore data from backup. Too many failed disks\n");
6c3560c0
AK		 10337 goto abort;
10338 }
10339
befb629b
AK		 10340 if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) {
10341 /* ignore error == 2, this can mean end of reshape here
10342 */
7a862a02 10343 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL) during restart\n");
befb629b 10344 } else
276d77db 10345 retval = 0;
276d77db
AK		 10346
10347abort:
10348 if (targets) {
10349 for (i = 0; i < new_disks; i++)
10350 if (targets[i])
10351 close(targets[i]);
10352 free(targets);
10353 }
10354 free(buf);
10355 return retval;
10356}
10357
2cda7640
ML		 10358static char disk_by_path[] = "/dev/disk/by-path/";
10359
10360static const char *imsm_get_disk_controller_domain(const char *path)
10361{
2cda7640 10362 char disk_path[PATH_MAX];
96234762
LM		 10363 char *drv=NULL;
10364 struct stat st;
2cda7640 10365
6d8d290a 10366 strcpy(disk_path, disk_by_path);
96234762
LM		 10367 strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1);
10368 if (stat(disk_path, &st) == 0) {
10369 struct sys_dev* hba;
594dc1b8 10370 char *path;
96234762
LM		 10371
10372 path = devt_to_devpath(st.st_rdev);
10373 if (path == NULL)
10374 return "unknown";
10375 hba = find_disk_attached_hba(-1, path);
10376 if (hba && hba->type == SYS_DEV_SAS)
10377 drv = "isci";
10378 else if (hba && hba->type == SYS_DEV_SATA)
10379 drv = "ahci";
1011e834 10380 else
96234762
LM		 10381 drv = "unknown";
10382 dprintf("path: %s hba: %s attached: %s\n",
10383 path, (hba) ? hba->path : "NULL", drv);
10384 free(path);
2cda7640 10385 }
96234762 10386 return drv;
2cda7640
ML		 10387}
10388
4dd2df09 10389static char *imsm_find_array_devnm_by_subdev(int subdev, char *container)
78b10e66 10390{
4dd2df09 10391 static char devnm[32];
78b10e66
N		 10392 char subdev_name[20];
10393 struct mdstat_ent *mdstat;
10394
10395 sprintf(subdev_name, "%d", subdev);
10396 mdstat = mdstat_by_subdev(subdev_name, container);
10397 if (!mdstat)
4dd2df09 10398 return NULL;
78b10e66 10399
4dd2df09 10400 strcpy(devnm, mdstat->devnm);
78b10e66 10401 free_mdstat(mdstat);
4dd2df09 10402 return devnm;
78b10e66
N		 10403}
10404
10405static int imsm_reshape_is_allowed_on_container(struct supertype *st,
10406 struct geo_params *geo,
fbf3d202
AK		 10407 int *old_raid_disks,
10408 int direction)
78b10e66 10409{
694575e7
KW		 10410 /* currently we only support increasing the number of devices
10411 * for a container. This increases the number of device for each
10412 * member array. They must all be RAID0 or RAID5.
10413 */
78b10e66
N		 10414 int ret_val = 0;
10415 struct mdinfo *info, *member;
10416 int devices_that_can_grow = 0;
10417
7a862a02 10418 dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): st->devnm = (%s)\n", st->devnm);
78b10e66 10419
d04f65f4 10420 if (geo->size > 0 ||
78b10e66
N		 10421 geo->level != UnSet ||
10422 geo->layout != UnSet ||
10423 geo->chunksize != 0 ||
10424 geo->raid_disks == UnSet) {
7a862a02 10425 dprintf("imsm: Container operation is allowed for raid disks number change only.\n");
78b10e66
N		 10426 return ret_val;
10427 }
10428
fbf3d202 10429 if (direction == ROLLBACK_METADATA_CHANGES) {
7a862a02 10430 dprintf("imsm: Metadata changes rollback is not supported for container operation.\n");
fbf3d202
AK		 10431 return ret_val;
10432 }
10433
78b10e66
N		 10434 info = container_content_imsm(st, NULL);
10435 for (member = info; member; member = member->next) {
4dd2df09 10436 char *result;
78b10e66
N		 10437
10438 dprintf("imsm: checking device_num: %i\n",
10439 member->container_member);
10440
d7d205bd 10441 if (geo->raid_disks <= member->array.raid_disks) {
78b10e66
N		 10442 /* we work on container for Online Capacity Expansion
10443 * only so raid_disks has to grow
10444 */
7a862a02 10445 dprintf("imsm: for container operation raid disks increase is required\n");
78b10e66
N		 10446 break;
10447 }
10448
089f9d79 10449 if (info->array.level != 0 && info->array.level != 5) {
78b10e66
N		 10450 /* we cannot use this container with other raid level
10451 */
7a862a02 10452 dprintf("imsm: for container operation wrong raid level (%i) detected\n",
78b10e66
N		 10453 info->array.level);
10454 break;
10455 } else {
10456 /* check for platform support
10457 * for this raid level configuration
10458 */
10459 struct intel_super *super = st->sb;
10460 if (!is_raid_level_supported(super->orom,
10461 member->array.level,
10462 geo->raid_disks)) {
7a862a02 10463 dprintf("platform does not support raid%d with %d disk%s\n",
78b10e66
N		 10464 info->array.level,
10465 geo->raid_disks,
10466 geo->raid_disks > 1 ? "s" : "");
10467 break;
10468 }
2a4a08e7
AK		 10469 /* check if component size is aligned to chunk size
10470 */
10471 if (info->component_size %
10472 (info->array.chunk_size/512)) {
7a862a02 10473 dprintf("Component size is not aligned to chunk size\n");
2a4a08e7
AK		 10474 break;
10475 }
78b10e66
N		 10476 }
10477
10478 if (*old_raid_disks &&
10479 info->array.raid_disks != *old_raid_disks)
10480 break;
10481 *old_raid_disks = info->array.raid_disks;
10482
10483 /* All raid5 and raid0 volumes in container
10484 * have to be ready for Online Capacity Expansion
10485 * so they need to be assembled. We have already
10486 * checked that no recovery etc is happening.
10487 */
4dd2df09
N		 10488 result = imsm_find_array_devnm_by_subdev(member->container_member,
10489 st->container_devnm);
10490 if (result == NULL) {
78b10e66
N		 10491 dprintf("imsm: cannot find array\n");
10492 break;
10493 }
10494 devices_that_can_grow++;
10495 }
10496 sysfs_free(info);
10497 if (!member && devices_that_can_grow)
10498 ret_val = 1;
10499
10500 if (ret_val)
1ade5cc1 10501 dprintf("Container operation allowed\n");
78b10e66 10502 else
1ade5cc1 10503 dprintf("Error: %i\n", ret_val);
78b10e66
N		 10504
10505 return ret_val;
10506}
10507
10508/* Function: get_spares_for_grow
10509 * Description: Allocates memory and creates list of spare devices
1011e834 10510 * avaliable in container. Checks if spare drive size is acceptable.
78b10e66
N		 10511 * Parameters: Pointer to the supertype structure
10512 * Returns: Pointer to the list of spare devices (mdinfo structure) on success,
1011e834 10513 * NULL if fail
78b10e66
N		 10514 */
10515static struct mdinfo *get_spares_for_grow(struct supertype *st)
10516{
78b10e66 10517 unsigned long long min_size = min_acceptable_spare_size_imsm(st);
326727d9 10518 return container_choose_spares(st, min_size, NULL, NULL, NULL, 0);
78b10e66
N		 10519}
10520
10521/******************************************************************************
10522 * function: imsm_create_metadata_update_for_reshape
10523 * Function creates update for whole IMSM container.
10524 *
10525 ******************************************************************************/
10526static int imsm_create_metadata_update_for_reshape(
10527 struct supertype *st,
10528 struct geo_params *geo,
10529 int old_raid_disks,
10530 struct imsm_update_reshape **updatep)
10531{
10532 struct intel_super *super = st->sb;
10533 struct imsm_super *mpb = super->anchor;
594dc1b8
JS		 10534 int update_memory_size;
10535 struct imsm_update_reshape *u;
10536 struct mdinfo *spares;
78b10e66 10537 int i;
594dc1b8 10538 int delta_disks;
bbd24d86 10539 struct mdinfo *dev;
78b10e66 10540
1ade5cc1 10541 dprintf("(enter) raid_disks = %i\n", geo->raid_disks);
78b10e66
N		 10542
10543 delta_disks = geo->raid_disks - old_raid_disks;
10544
10545 /* size of all update data without anchor */
10546 update_memory_size = sizeof(struct imsm_update_reshape);
10547
10548 /* now add space for spare disks that we need to add. */
10549 update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1);
10550
503975b9 10551 u = xcalloc(1, update_memory_size);
78b10e66
N		 10552 u->type = update_reshape_container_disks;
10553 u->old_raid_disks = old_raid_disks;
10554 u->new_raid_disks = geo->raid_disks;
10555
10556 /* now get spare disks list
10557 */
10558 spares = get_spares_for_grow(st);
10559
10560 if (spares == NULL
10561 || delta_disks > spares->array.spare_disks) {
7a862a02 10562 pr_err("imsm: ERROR: Cannot get spare devices for %s.\n", geo->dev_name);
e4c72d1d 10563 i = -1;
78b10e66
N		 10564 goto abort;
10565 }
10566
10567 /* we have got spares
10568 * update disk list in imsm_disk list table in anchor
10569 */
10570 dprintf("imsm: %i spares are available.\n\n",
10571 spares->array.spare_disks);
10572
bbd24d86 10573 dev = spares->devs;
78b10e66 10574 for (i = 0; i < delta_disks; i++) {
78b10e66
N		 10575 struct dl *dl;
10576
bbd24d86
AK		 10577 if (dev == NULL)
10578 break;
78b10e66
N		 10579 u->new_disks[i] = makedev(dev->disk.major,
10580 dev->disk.minor);
10581 dl = get_disk_super(super, dev->disk.major, dev->disk.minor);
ee4beede
AK		 10582 dl->index = mpb->num_disks;
10583 mpb->num_disks++;
bbd24d86 10584 dev = dev->next;
78b10e66 10585 }
78b10e66
N		 10586
10587abort:
10588 /* free spares
10589 */
10590 sysfs_free(spares);
10591
d677e0b8 10592 dprintf("imsm: reshape update preparation :");
78b10e66 10593 if (i == delta_disks) {
1ade5cc1 10594 dprintf_cont(" OK\n");
78b10e66
N		 10595 *updatep = u;
10596 return update_memory_size;
10597 }
10598 free(u);
1ade5cc1 10599 dprintf_cont(" Error\n");
78b10e66
N		 10600
10601 return 0;
10602}
10603
f3871fdc
AK		 10604/******************************************************************************
10605 * function: imsm_create_metadata_update_for_size_change()
10606 * Creates update for IMSM array for array size change.
10607 *
10608 ******************************************************************************/
10609static int imsm_create_metadata_update_for_size_change(
10610 struct supertype *st,
10611 struct geo_params *geo,
10612 struct imsm_update_size_change **updatep)
10613{
10614 struct intel_super *super = st->sb;
594dc1b8
JS		 10615 int update_memory_size;
10616 struct imsm_update_size_change *u;
f3871fdc 10617
1ade5cc1 10618 dprintf("(enter) New size = %llu\n", geo->size);
f3871fdc
AK		 10619
10620 /* size of all update data without anchor */
10621 update_memory_size = sizeof(struct imsm_update_size_change);
10622
503975b9 10623 u = xcalloc(1, update_memory_size);
f3871fdc
AK		 10624 u->type = update_size_change;
10625 u->subdev = super->current_vol;
10626 u->new_size = geo->size;
10627
10628 dprintf("imsm: reshape update preparation : OK\n");
10629 *updatep = u;
10630
10631 return update_memory_size;
10632}
10633
48c5303a
PC		 10634/******************************************************************************
10635 * function: imsm_create_metadata_update_for_migration()
10636 * Creates update for IMSM array.
10637 *
10638 ******************************************************************************/
10639static int imsm_create_metadata_update_for_migration(
10640 struct supertype *st,
10641 struct geo_params *geo,
10642 struct imsm_update_reshape_migration **updatep)
10643{
10644 struct intel_super *super = st->sb;
594dc1b8
JS		 10645 int update_memory_size;
10646 struct imsm_update_reshape_migration *u;
48c5303a
PC		 10647 struct imsm_dev *dev;
10648 int previous_level = -1;
10649
1ade5cc1 10650 dprintf("(enter) New Level = %i\n", geo->level);
48c5303a
PC		 10651
10652 /* size of all update data without anchor */
10653 update_memory_size = sizeof(struct imsm_update_reshape_migration);
10654
503975b9 10655 u = xcalloc(1, update_memory_size);
48c5303a
PC		 10656 u->type = update_reshape_migration;
10657 u->subdev = super->current_vol;
10658 u->new_level = geo->level;
10659 u->new_layout = geo->layout;
10660 u->new_raid_disks = u->old_raid_disks = geo->raid_disks;
10661 u->new_disks[0] = -1;
4bba0439 10662 u->new_chunksize = -1;
48c5303a
PC		 10663
10664 dev = get_imsm_dev(super, u->subdev);
10665 if (dev) {
10666 struct imsm_map *map;
10667
238c0a71 10668 map = get_imsm_map(dev, MAP_0);
4bba0439
PC		 10669 if (map) {
10670 int current_chunk_size =
10671 __le16_to_cpu(map->blocks_per_strip) / 2;
10672
10673 if (geo->chunksize != current_chunk_size) {
10674 u->new_chunksize = geo->chunksize / 1024;
7a862a02 10675 dprintf("imsm: chunk size change from %i to %i\n",
4bba0439
PC		 10676 current_chunk_size, u->new_chunksize);
10677 }
48c5303a 10678 previous_level = map->raid_level;
4bba0439 10679 }
48c5303a 10680 }
089f9d79 10681 if (geo->level == 5 && previous_level == 0) {
48c5303a
PC		 10682 struct mdinfo *spares = NULL;
10683
10684 u->new_raid_disks++;
10685 spares = get_spares_for_grow(st);
089f9d79 10686 if (spares == NULL || spares->array.spare_disks < 1) {
48c5303a
PC		 10687 free(u);
10688 sysfs_free(spares);
10689 update_memory_size = 0;
7a862a02 10690 dprintf("error: cannot get spare device for requested migration");
48c5303a
PC		 10691 return 0;
10692 }
10693 sysfs_free(spares);
10694 }
10695 dprintf("imsm: reshape update preparation : OK\n");
10696 *updatep = u;
10697
10698 return update_memory_size;
10699}
10700
8dd70bce
AK		 10701static void imsm_update_metadata_locally(struct supertype *st,
10702 void *buf, int len)
10703{
10704 struct metadata_update mu;
10705
10706 mu.buf = buf;
10707 mu.len = len;
10708 mu.space = NULL;
10709 mu.space_list = NULL;
10710 mu.next = NULL;
5fe6f031
N		 10711 if (imsm_prepare_update(st, &mu))
10712 imsm_process_update(st, &mu);
8dd70bce
AK		 10713
10714 while (mu.space_list) {
10715 void **space = mu.space_list;
10716 mu.space_list = *space;
10717 free(space);
10718 }
10719}
78b10e66 10720
471bceb6 10721/***************************************************************************
694575e7 10722* Function: imsm_analyze_change
471bceb6 10723* Description: Function analyze change for single volume
1011e834 10724* and validate if transition is supported
fbf3d202
AK		 10725* Parameters: Geometry parameters, supertype structure,
10726* metadata change direction (apply/rollback)
694575e7 10727* Returns: Operation type code on success, -1 if fail
471bceb6
KW		 10728****************************************************************************/
10729enum imsm_reshape_type imsm_analyze_change(struct supertype *st,
fbf3d202
AK		 10730 struct geo_params *geo,
10731 int direction)
694575e7 10732{
471bceb6
KW		 10733 struct mdinfo info;
10734 int change = -1;
10735 int check_devs = 0;
c21e737b 10736 int chunk;
67a2db32
AK		 10737 /* number of added/removed disks in operation result */
10738 int devNumChange = 0;
10739 /* imsm compatible layout value for array geometry verification */
10740 int imsm_layout = -1;
7abc9871
AK		 10741 int data_disks;
10742 struct imsm_dev *dev;
10743 struct intel_super *super;
d04f65f4 10744 unsigned long long current_size;
65d38cca 10745 unsigned long long free_size;
d04f65f4 10746 unsigned long long max_size;
65d38cca 10747 int rv;
471bceb6
KW		 10748
10749 getinfo_super_imsm_volume(st, &info, NULL);
089f9d79
JS		 10750 if (geo->level != info.array.level && geo->level >= 0 &&
10751 geo->level != UnSet) {
471bceb6
KW		 10752 switch (info.array.level) {
10753 case 0:
10754 if (geo->level == 5) {
b5347799 10755 change = CH_MIGRATION;
e13ce846 10756 if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) {
7a862a02 10757 pr_err("Error. Requested Layout not supported (left-asymmetric layout is supported only)!\n");
e13ce846
AK		 10758 change = -1;
10759 goto analyse_change_exit;
10760 }
67a2db32 10761 imsm_layout = geo->layout;
471bceb6 10762 check_devs = 1;
e91a3bad
LM		 10763 devNumChange = 1; /* parity disk added */
10764 } else if (geo->level == 10) {
471bceb6
KW		 10765 change = CH_TAKEOVER;
10766 check_devs = 1;
e91a3bad 10767 devNumChange = 2; /* two mirrors added */
67a2db32 10768 imsm_layout = 0x102; /* imsm supported layout */
471bceb6 10769 }
dfe77a9e
KW		 10770 break;
10771 case 1:
471bceb6
KW		 10772 case 10:
10773 if (geo->level == 0) {
10774 change = CH_TAKEOVER;
10775 check_devs = 1;
e91a3bad 10776 devNumChange = -(geo->raid_disks/2);
67a2db32 10777 imsm_layout = 0; /* imsm raid0 layout */
471bceb6
KW		 10778 }
10779 break;
10780 }
10781 if (change == -1) {
7a862a02 10782 pr_err("Error. Level Migration from %d to %d not supported!\n",
e7b84f9d 10783 info.array.level, geo->level);
471bceb6
KW		 10784 goto analyse_change_exit;
10785 }
10786 } else
10787 geo->level = info.array.level;
10788
089f9d79
JS		 10789 if (geo->layout != info.array.layout &&
10790 (geo->layout != UnSet && geo->layout != -1)) {
b5347799 10791 change = CH_MIGRATION;
089f9d79
JS		 10792 if (info.array.layout == 0 && info.array.level == 5 &&
10793 geo->layout == 5) {
471bceb6 10794 /* reshape 5 -> 4 */
089f9d79
JS		 10795 } else if (info.array.layout == 5 && info.array.level == 5 &&
10796 geo->layout == 0) {
471bceb6
KW		 10797 /* reshape 4 -> 5 */
10798 geo->layout = 0;
10799 geo->level = 5;
10800 } else {
7a862a02 10801 pr_err("Error. Layout Migration from %d to %d not supported!\n",
e7b84f9d 10802 info.array.layout, geo->layout);
471bceb6
KW		 10803 change = -1;
10804 goto analyse_change_exit;
10805 }
67a2db32 10806 } else {
471bceb6 10807 geo->layout = info.array.layout;
67a2db32
AK		 10808 if (imsm_layout == -1)
10809 imsm_layout = info.array.layout;
10810 }
471bceb6 10811
089f9d79
JS		 10812 if (geo->chunksize > 0 && geo->chunksize != UnSet &&
10813 geo->chunksize != info.array.chunk_size) {
2d2b0eb7
MD		 10814 if (info.array.level == 10) {
10815 pr_err("Error. Chunk size change for RAID 10 is not supported.\n");
10816 change = -1;
10817 goto analyse_change_exit;
10818 }
b5347799 10819 change = CH_MIGRATION;
2d2b0eb7 10820 } else {
471bceb6 10821 geo->chunksize = info.array.chunk_size;
2d2b0eb7 10822 }
471bceb6 10823
c21e737b 10824 chunk = geo->chunksize / 1024;
7abc9871
AK		 10825
10826 super = st->sb;
10827 dev = get_imsm_dev(super, super->current_vol);
10828 data_disks = imsm_num_data_members(dev , MAP_0);
c41e00b2 10829 /* compute current size per disk member
7abc9871 10830 */
c41e00b2
AK		 10831 current_size = info.custom_array_size / data_disks;
10832
089f9d79 10833 if (geo->size > 0 && geo->size != MAX_SIZE) {
c41e00b2
AK		 10834 /* align component size
10835 */
10836 geo->size = imsm_component_size_aligment_check(
10837 get_imsm_raid_level(dev->vol.map),
f36a9ecd 10838 chunk * 1024, super->sector_size,
c41e00b2 10839 geo->size * 2);
65d0b4ce 10840 if (geo->size == 0) {
7a862a02 10841 pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is 0).\n",
65d0b4ce
LD		 10842 current_size);
10843 goto analyse_change_exit;
10844 }
c41e00b2 10845 }
7abc9871 10846
089f9d79 10847 if (current_size != geo->size && geo->size > 0) {
7abc9871 10848 if (change != -1) {
7a862a02 10849 pr_err("Error. Size change should be the only one at a time.\n");
7abc9871
AK		 10850 change = -1;
10851 goto analyse_change_exit;
10852 }
10853 if ((super->current_vol + 1) != super->anchor->num_raid_devs) {
7a862a02 10854 pr_err("Error. The last volume in container can be expanded only (%i/%s).\n",
4dd2df09 10855 super->current_vol, st->devnm);
7abc9871
AK		 10856 goto analyse_change_exit;
10857 }
65d38cca
LD		 10858 /* check the maximum available size
10859 */
10860 rv = imsm_get_free_size(st, dev->vol.map->num_members,
10861 0, chunk, &free_size);
10862 if (rv == 0)
10863 /* Cannot find maximum available space
10864 */
10865 max_size = 0;
10866 else {
10867 max_size = free_size + current_size;
10868 /* align component size
10869 */
10870 max_size = imsm_component_size_aligment_check(
10871 get_imsm_raid_level(dev->vol.map),
f36a9ecd 10872 chunk * 1024, super->sector_size,
65d38cca
LD		 10873 max_size);
10874 }
d04f65f4 10875 if (geo->size == MAX_SIZE) {
b130333f
AK		 10876 /* requested size change to the maximum available size
10877 */
65d38cca 10878 if (max_size == 0) {
7a862a02 10879 pr_err("Error. Cannot find maximum available space.\n");
b130333f
AK		 10880 change = -1;
10881 goto analyse_change_exit;
65d38cca
LD		 10882 } else
10883 geo->size = max_size;
c41e00b2 10884 }
b130333f 10885
681b7ae2 10886 if (direction == ROLLBACK_METADATA_CHANGES) {
fbf3d202
AK		 10887 /* accept size for rollback only
10888 */
10889 } else {
10890 /* round size due to metadata compatibility
10891 */
10892 geo->size = (geo->size >> SECT_PER_MB_SHIFT)
10893 << SECT_PER_MB_SHIFT;
10894 dprintf("Prepare update for size change to %llu\n",
10895 geo->size );
10896 if (current_size >= geo->size) {
7a862a02 10897 pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is %llu).\n",
e7b84f9d 10898 current_size, geo->size);
fbf3d202
AK		 10899 goto analyse_change_exit;
10900 }
65d38cca 10901 if (max_size && geo->size > max_size) {
7a862a02 10902 pr_err("Error. Requested size is larger than maximum available size (maximum available size is %llu, requested size /rounded/ is %llu).\n",
e7b84f9d 10903 max_size, geo->size);
65d38cca
LD		 10904 goto analyse_change_exit;
10905 }
7abc9871
AK		 10906 }
10907 geo->size *= data_disks;
10908 geo->raid_disks = dev->vol.map->num_members;
10909 change = CH_ARRAY_SIZE;
10910 }
471bceb6
KW		 10911 if (!validate_geometry_imsm(st,
10912 geo->level,
67a2db32 10913 imsm_layout,
e91a3bad 10914 geo->raid_disks + devNumChange,
c21e737b 10915 &chunk,
af4348dd 10916 geo->size, INVALID_SECTORS,
471bceb6
KW		 10917 0, 0, 1))
10918 change = -1;
10919
10920 if (check_devs) {
10921 struct intel_super *super = st->sb;
10922 struct imsm_super *mpb = super->anchor;
10923
10924 if (mpb->num_raid_devs > 1) {
7a862a02 10925 pr_err("Error. Cannot perform operation on %s- for this operation it MUST be single array in container\n",
e7b84f9d 10926 geo->dev_name);
471bceb6
KW		 10927 change = -1;
10928 }
10929 }
10930
10931analyse_change_exit:
089f9d79
JS		 10932 if (direction == ROLLBACK_METADATA_CHANGES &&
10933 (change == CH_MIGRATION || change == CH_TAKEOVER)) {
7a862a02 10934 dprintf("imsm: Metadata changes rollback is not supported for migration and takeover operations.\n");
fbf3d202
AK		 10935 change = -1;
10936 }
471bceb6 10937 return change;
694575e7
KW		 10938}
10939
bb025c2f
KW		 10940int imsm_takeover(struct supertype *st, struct geo_params *geo)
10941{
10942 struct intel_super *super = st->sb;
10943 struct imsm_update_takeover *u;
10944
503975b9 10945 u = xmalloc(sizeof(struct imsm_update_takeover));
bb025c2f
KW		 10946
10947 u->type = update_takeover;
10948 u->subarray = super->current_vol;
10949
10950 /* 10->0 transition */
10951 if (geo->level == 0)
10952 u->direction = R10_TO_R0;
10953
0529c688
KW		 10954 /* 0->10 transition */
10955 if (geo->level == 10)
10956 u->direction = R0_TO_R10;
10957
bb025c2f
KW		 10958 /* update metadata locally */
10959 imsm_update_metadata_locally(st, u,
10960 sizeof(struct imsm_update_takeover));
10961 /* and possibly remotely */
10962 if (st->update_tail)
10963 append_metadata_update(st, u,
10964 sizeof(struct imsm_update_takeover));
10965 else
10966 free(u);
10967
10968 return 0;
10969}
10970
d04f65f4
N		 10971static int imsm_reshape_super(struct supertype *st, unsigned long long size,
10972 int level,
78b10e66 10973 int layout, int chunksize, int raid_disks,
41784c88 10974 int delta_disks, char *backup, char *dev,
016e00f5 10975 int direction, int verbose)
78b10e66 10976{
78b10e66
N		 10977 int ret_val = 1;
10978 struct geo_params geo;
10979
1ade5cc1 10980 dprintf("(enter)\n");
78b10e66 10981
71204a50 10982 memset(&geo, 0, sizeof(struct geo_params));
78b10e66
N		 10983
10984 geo.dev_name = dev;
4dd2df09 10985 strcpy(geo.devnm, st->devnm);
78b10e66
N		 10986 geo.size = size;
10987 geo.level = level;
10988 geo.layout = layout;
10989 geo.chunksize = chunksize;
10990 geo.raid_disks = raid_disks;
41784c88
AK		 10991 if (delta_disks != UnSet)
10992 geo.raid_disks += delta_disks;
78b10e66 10993
1ade5cc1
N		 10994 dprintf("for level : %i\n", geo.level);
10995 dprintf("for raid_disks : %i\n", geo.raid_disks);
78b10e66
N		 10996
10997 if (experimental() == 0)
10998 return ret_val;
10999
4dd2df09 11000 if (strcmp(st->container_devnm, st->devnm) == 0) {
694575e7
KW		 11001 /* On container level we can only increase number of devices. */
11002 dprintf("imsm: info: Container operation\n");
78b10e66 11003 int old_raid_disks = 0;
6dc0be30 11004
78b10e66 11005 if (imsm_reshape_is_allowed_on_container(
fbf3d202 11006 st, &geo, &old_raid_disks, direction)) {
78b10e66
N		 11007 struct imsm_update_reshape *u = NULL;
11008 int len;
11009
11010 len = imsm_create_metadata_update_for_reshape(
11011 st, &geo, old_raid_disks, &u);
11012
ed08d51c
AK		 11013 if (len <= 0) {
11014 dprintf("imsm: Cannot prepare update\n");
11015 goto exit_imsm_reshape_super;
11016 }
11017
8dd70bce
AK		 11018 ret_val = 0;
11019 /* update metadata locally */
11020 imsm_update_metadata_locally(st, u, len);
11021 /* and possibly remotely */
11022 if (st->update_tail)
11023 append_metadata_update(st, u, len);
11024 else
ed08d51c 11025 free(u);
8dd70bce 11026
694575e7 11027 } else {
7a862a02 11028 pr_err("(imsm) Operation is not allowed on this container\n");
694575e7
KW		 11029 }
11030 } else {
11031 /* On volume level we support following operations
471bceb6
KW		 11032 * - takeover: raid10 -> raid0; raid0 -> raid10
11033 * - chunk size migration
11034 * - migration: raid5 -> raid0; raid0 -> raid5
11035 */
11036 struct intel_super *super = st->sb;
11037 struct intel_dev *dev = super->devlist;
4dd2df09 11038 int change;
694575e7 11039 dprintf("imsm: info: Volume operation\n");
471bceb6
KW		 11040 /* find requested device */
11041 while (dev) {
1011e834 11042 char *devnm =
4dd2df09
N		 11043 imsm_find_array_devnm_by_subdev(
11044 dev->index, st->container_devnm);
11045 if (devnm && strcmp(devnm, geo.devnm) == 0)
471bceb6
KW		 11046 break;
11047 dev = dev->next;
11048 }
11049 if (dev == NULL) {
4dd2df09
N		 11050 pr_err("Cannot find %s (%s) subarray\n",
11051 geo.dev_name, geo.devnm);
471bceb6
KW		 11052 goto exit_imsm_reshape_super;
11053 }
11054 super->current_vol = dev->index;
fbf3d202 11055 change = imsm_analyze_change(st, &geo, direction);
694575e7 11056 switch (change) {
471bceb6 11057 case CH_TAKEOVER:
bb025c2f 11058 ret_val = imsm_takeover(st, &geo);
694575e7 11059 break;
48c5303a
PC		 11060 case CH_MIGRATION: {
11061 struct imsm_update_reshape_migration *u = NULL;
11062 int len =
11063 imsm_create_metadata_update_for_migration(
11064 st, &geo, &u);
11065 if (len < 1) {
7a862a02 11066 dprintf("imsm: Cannot prepare update\n");
48c5303a
PC		 11067 break;
11068 }
471bceb6 11069 ret_val = 0;
48c5303a
PC		 11070 /* update metadata locally */
11071 imsm_update_metadata_locally(st, u, len);
11072 /* and possibly remotely */
11073 if (st->update_tail)
11074 append_metadata_update(st, u, len);
11075 else
11076 free(u);
11077 }
11078 break;
7abc9871 11079 case CH_ARRAY_SIZE: {
f3871fdc
AK		 11080 struct imsm_update_size_change *u = NULL;
11081 int len =
11082 imsm_create_metadata_update_for_size_change(
11083 st, &geo, &u);
11084 if (len < 1) {
7a862a02 11085 dprintf("imsm: Cannot prepare update\n");
f3871fdc
AK		 11086 break;
11087 }
11088 ret_val = 0;
11089 /* update metadata locally */
11090 imsm_update_metadata_locally(st, u, len);
11091 /* and possibly remotely */
11092 if (st->update_tail)
11093 append_metadata_update(st, u, len);
11094 else
11095 free(u);
7abc9871
AK		 11096 }
11097 break;
471bceb6
KW		 11098 default:
11099 ret_val = 1;
694575e7 11100 }
694575e7 11101 }
78b10e66 11102
ed08d51c 11103exit_imsm_reshape_super:
78b10e66
N		 11104 dprintf("imsm: reshape_super Exit code = %i\n", ret_val);
11105 return ret_val;
11106}
2cda7640 11107
0febb20c
AO		 11108#define COMPLETED_OK 0
11109#define COMPLETED_NONE 1
11110#define COMPLETED_DELAYED 2
11111
11112static int read_completed(int fd, unsigned long long *val)
11113{
11114 int ret;
11115 char buf[50];
11116
11117 ret = sysfs_fd_get_str(fd, buf, 50);
11118 if (ret < 0)
11119 return ret;
11120
11121 ret = COMPLETED_OK;
11122 if (strncmp(buf, "none", 4) == 0) {
11123 ret = COMPLETED_NONE;
11124 } else if (strncmp(buf, "delayed", 7) == 0) {
11125 ret = COMPLETED_DELAYED;
11126 } else {
11127 char *ep;
11128 *val = strtoull(buf, &ep, 0);
11129 if (ep == buf || (*ep != 0 && *ep != '\n' && *ep != ' '))
11130 ret = -1;
11131 }
11132 return ret;
11133}
11134
eee67a47
AK		 11135/*******************************************************************************
11136 * Function: wait_for_reshape_imsm
11137 * Description: Function writes new sync_max value and waits until
11138 * reshape process reach new position
11139 * Parameters:
11140 * sra : general array info
eee67a47
AK		 11141 * ndata : number of disks in new array's layout
11142 * Returns:
11143 * 0 : success,
11144 * 1 : there is no reshape in progress,
11145 * -1 : fail
11146 ******************************************************************************/
ae9f01f8 11147int wait_for_reshape_imsm(struct mdinfo *sra, int ndata)
eee67a47 11148{
85ca499c 11149 int fd = sysfs_get_fd(sra, NULL, "sync_completed");
df2647fa 11150 int retry = 3;
eee67a47 11151 unsigned long long completed;
ae9f01f8
AK		 11152 /* to_complete : new sync_max position */
11153 unsigned long long to_complete = sra->reshape_progress;
11154 unsigned long long position_to_set = to_complete / ndata;
eee67a47 11155
ae9f01f8 11156 if (fd < 0) {
1ade5cc1 11157 dprintf("cannot open reshape_position\n");
eee67a47 11158 return 1;
ae9f01f8 11159 }
eee67a47 11160
df2647fa
PB		 11161 do {
11162 if (sysfs_fd_get_ll(fd, &completed) < 0) {
11163 if (!retry) {
11164 dprintf("cannot read reshape_position (no reshape in progres)\n");
11165 close(fd);
11166 return 1;
11167 }
11168 usleep(30000);
11169 } else
11170 break;
11171 } while (retry--);
eee67a47 11172
85ca499c 11173 if (completed > position_to_set) {
1ade5cc1 11174 dprintf("wrong next position to set %llu (%llu)\n",
85ca499c 11175 to_complete, position_to_set);
ae9f01f8
AK		 11176 close(fd);
11177 return -1;
11178 }
11179 dprintf("Position set: %llu\n", position_to_set);
11180 if (sysfs_set_num(sra, NULL, "sync_max",
11181 position_to_set) != 0) {
1ade5cc1 11182 dprintf("cannot set reshape position to %llu\n",
ae9f01f8
AK		 11183 position_to_set);
11184 close(fd);
11185 return -1;
eee67a47
AK		 11186 }
11187
eee67a47 11188 do {
0febb20c 11189 int rc;
eee67a47 11190 char action[20];
5ff3a780 11191 int timeout = 3000;
0febb20c 11192
5ff3a780 11193 sysfs_wait(fd, &timeout);
a47e44fb
AK		 11194 if (sysfs_get_str(sra, NULL, "sync_action",
11195 action, 20) > 0 &&
d7d3809a 11196 strncmp(action, "reshape", 7) != 0) {
b2be2b62
AO		 11197 if (strncmp(action, "idle", 4) == 0)
11198 break;
d7d3809a
AP		 11199 close(fd);
11200 return -1;
11201 }
0febb20c
AO		 11202
11203 rc = read_completed(fd, &completed);
11204 if (rc < 0) {
1ade5cc1 11205 dprintf("cannot read reshape_position (in loop)\n");
eee67a47
AK		 11206 close(fd);
11207 return 1;
0febb20c
AO		 11208 } else if (rc == COMPLETED_NONE)
11209 break;
85ca499c 11210 } while (completed < position_to_set);
b2be2b62 11211
eee67a47
AK		 11212 close(fd);
11213 return 0;
eee67a47
AK		 11214}
11215
b915c95f
AK		 11216/*******************************************************************************
11217 * Function: check_degradation_change
11218 * Description: Check that array hasn't become failed.
11219 * Parameters:
11220 * info : for sysfs access
11221 * sources : source disks descriptors
11222 * degraded: previous degradation level
11223 * Returns:
11224 * degradation level
11225 ******************************************************************************/
11226int check_degradation_change(struct mdinfo *info,
11227 int *sources,
11228 int degraded)
11229{
11230 unsigned long long new_degraded;
e1993023
LD		 11231 int rv;
11232
11233 rv = sysfs_get_ll(info, NULL, "degraded", &new_degraded);
089f9d79 11234 if (rv == -1 || (new_degraded != (unsigned long long)degraded)) {
b915c95f
AK		 11235 /* check each device to ensure it is still working */
11236 struct mdinfo *sd;
11237 new_degraded = 0;
11238 for (sd = info->devs ; sd ; sd = sd->next) {
11239 if (sd->disk.state & (1<<MD_DISK_FAULTY))
11240 continue;
11241 if (sd->disk.state & (1<<MD_DISK_SYNC)) {
cf52eff5
TM		 11242 char sbuf[100];
11243
b915c95f 11244 if (sysfs_get_str(info,
cf52eff5 11245 sd, "state", sbuf, sizeof(sbuf)) < 0 ||
b915c95f
AK		 11246 strstr(sbuf, "faulty") ||
11247 strstr(sbuf, "in_sync") == NULL) {
11248 /* this device is dead */
11249 sd->disk.state = (1<<MD_DISK_FAULTY);
11250 if (sd->disk.raid_disk >= 0 &&
11251 sources[sd->disk.raid_disk] >= 0) {
11252 close(sources[
11253 sd->disk.raid_disk]);
11254 sources[sd->disk.raid_disk] =
11255 -1;
11256 }
11257 new_degraded++;
11258 }
11259 }
11260 }
11261 }
11262
11263 return new_degraded;
11264}
11265
10f22854
AK		 11266/*******************************************************************************
11267 * Function: imsm_manage_reshape
11268 * Description: Function finds array under reshape and it manages reshape
11269 * process. It creates stripes backups (if required) and sets
942e1cdb 11270 * checkpoints.
10f22854
AK		 11271 * Parameters:
11272 * afd : Backup handle (nattive) - not used
11273 * sra : general array info
11274 * reshape : reshape parameters - not used
11275 * st : supertype structure
11276 * blocks : size of critical section [blocks]
11277 * fds : table of source device descriptor
11278 * offsets : start of array (offest per devices)
11279 * dests : not used
11280 * destfd : table of destination device descriptor
11281 * destoffsets : table of destination offsets (per device)
11282 * Returns:
11283 * 1 : success, reshape is done
11284 * 0 : fail
11285 ******************************************************************************/
999b4972
N		 11286static int imsm_manage_reshape(
11287 int afd, struct mdinfo *sra, struct reshape *reshape,
10f22854 11288 struct supertype *st, unsigned long backup_blocks,
999b4972
N		 11289 int *fds, unsigned long long *offsets,
11290 int dests, int *destfd, unsigned long long *destoffsets)
11291{
10f22854
AK		 11292 int ret_val = 0;
11293 struct intel_super *super = st->sb;
594dc1b8 11294 struct intel_dev *dv;
de44e46f 11295 unsigned int sector_size = super->sector_size;
10f22854 11296 struct imsm_dev *dev = NULL;
a6b6d984 11297 struct imsm_map *map_src;
10f22854
AK		 11298 int migr_vol_qan = 0;
11299 int ndata, odata; /* [bytes] */
11300 int chunk; /* [bytes] */
11301 struct migr_record *migr_rec;
11302 char *buf = NULL;
11303 unsigned int buf_size; /* [bytes] */
11304 unsigned long long max_position; /* array size [bytes] */
11305 unsigned long long next_step; /* [blocks]/[bytes] */
11306 unsigned long long old_data_stripe_length;
10f22854
AK		 11307 unsigned long long start_src; /* [bytes] */
11308 unsigned long long start; /* [bytes] */
11309 unsigned long long start_buf_shift; /* [bytes] */
b915c95f 11310 int degraded = 0;
ab724b98 11311 int source_layout = 0;
10f22854 11312
79a16a9b
JS		 11313 if (!sra)
11314 return ret_val;
11315
11316 if (!fds || !offsets)
10f22854
AK		 11317 goto abort;
11318
11319 /* Find volume during the reshape */
11320 for (dv = super->devlist; dv; dv = dv->next) {
11321 if (dv->dev->vol.migr_type == MIGR_GEN_MIGR
11322 && dv->dev->vol.migr_state == 1) {
11323 dev = dv->dev;
11324 migr_vol_qan++;
11325 }
11326 }
11327 /* Only one volume can migrate at the same time */
11328 if (migr_vol_qan != 1) {
676e87a8 11329 pr_err("%s", migr_vol_qan ?
10f22854
AK		 11330 "Number of migrating volumes greater than 1\n" :
11331 "There is no volume during migrationg\n");
11332 goto abort;
11333 }
11334
238c0a71 11335 map_src = get_imsm_map(dev, MAP_1);
10f22854
AK		 11336 if (map_src == NULL)
11337 goto abort;
10f22854 11338
238c0a71
AK		 11339 ndata = imsm_num_data_members(dev, MAP_0);
11340 odata = imsm_num_data_members(dev, MAP_1);
10f22854 11341
7b1ab482 11342 chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512;
10f22854
AK		 11343 old_data_stripe_length = odata * chunk;
11344
11345 migr_rec = super->migr_rec;
11346
10f22854
AK		 11347 /* initialize migration record for start condition */
11348 if (sra->reshape_progress == 0)
11349 init_migr_record_imsm(st, dev, sra);
b2c59438
AK		 11350 else {
11351 if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) {
7a862a02 11352 dprintf("imsm: cannot restart migration when data are present in copy area.\n");
b2c59438
AK		 11353 goto abort;
11354 }
6a75c8ca
AK		 11355 /* Save checkpoint to update migration record for current
11356 * reshape position (in md). It can be farther than current
11357 * reshape position in metadata.
11358 */
11359 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
11360 /* ignore error == 2, this can mean end of reshape here
11361 */
7a862a02 11362 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL, initial save)\n");
6a75c8ca
AK		 11363 goto abort;
11364 }
b2c59438 11365 }
10f22854
AK		 11366
11367 /* size for data */
11368 buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512;
11369 /* extend buffer size for parity disk */
11370 buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
11371 /* add space for stripe aligment */
11372 buf_size += old_data_stripe_length;
de44e46f
PB		 11373 if (posix_memalign((void **)&buf, MAX_SECTOR_SIZE, buf_size)) {
11374 dprintf("imsm: Cannot allocate checkpoint buffer\n");
10f22854
AK		 11375 goto abort;
11376 }
11377
3ef4403c 11378 max_position = sra->component_size * ndata;
68eb8bc6 11379 source_layout = imsm_level_to_layout(map_src->raid_level);
10f22854
AK		 11380
11381 while (__le32_to_cpu(migr_rec->curr_migr_unit) <
11382 __le32_to_cpu(migr_rec->num_migr_units)) {
11383 /* current reshape position [blocks] */
11384 unsigned long long current_position =
11385 __le32_to_cpu(migr_rec->blocks_per_unit)
11386 * __le32_to_cpu(migr_rec->curr_migr_unit);
11387 unsigned long long border;
11388
b915c95f
AK		 11389 /* Check that array hasn't become failed.
11390 */
11391 degraded = check_degradation_change(sra, fds, degraded);
11392 if (degraded > 1) {
7a862a02 11393 dprintf("imsm: Abort reshape due to degradation level (%i)\n", degraded);
b915c95f
AK		 11394 goto abort;
11395 }
11396
10f22854
AK		 11397 next_step = __le32_to_cpu(migr_rec->blocks_per_unit);
11398
11399 if ((current_position + next_step) > max_position)
11400 next_step = max_position - current_position;
11401
92144abf 11402 start = current_position * 512;
10f22854 11403
942e1cdb 11404 /* align reading start to old geometry */
10f22854
AK		 11405 start_buf_shift = start % old_data_stripe_length;
11406 start_src = start - start_buf_shift;
11407
11408 border = (start_src / odata) - (start / ndata);
11409 border /= 512;
11410 if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) {
11411 /* save critical stripes to buf
11412 * start - start address of current unit
11413 * to backup [bytes]
11414 * start_src - start address of current unit
11415 * to backup alligned to source array
11416 * [bytes]
11417 */
594dc1b8 11418 unsigned long long next_step_filler;
10f22854
AK		 11419 unsigned long long copy_length = next_step * 512;
11420
11421 /* allign copy area length to stripe in old geometry */
11422 next_step_filler = ((copy_length + start_buf_shift)
11423 % old_data_stripe_length);
11424 if (next_step_filler)
11425 next_step_filler = (old_data_stripe_length
11426 - next_step_filler);
7a862a02 11427 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		 11428 start, start_src, copy_length,
11429 start_buf_shift, next_step_filler);
11430
11431 if (save_stripes(fds, offsets, map_src->num_members,
ab724b98
AK		 11432 chunk, map_src->raid_level,
11433 source_layout, 0, NULL, start_src,
10f22854
AK		 11434 copy_length +
11435 next_step_filler + start_buf_shift,
11436 buf)) {
7a862a02 11437 dprintf("imsm: Cannot save stripes to buffer\n");
10f22854
AK		 11438 goto abort;
11439 }
11440 /* Convert data to destination format and store it
11441 * in backup general migration area
11442 */
11443 if (save_backup_imsm(st, dev, sra,
aea93171 11444 buf + start_buf_shift, copy_length)) {
7a862a02 11445 dprintf("imsm: Cannot save stripes to target devices\n");
10f22854
AK		 11446 goto abort;
11447 }
11448 if (save_checkpoint_imsm(st, sra,
11449 UNIT_SRC_IN_CP_AREA)) {
7a862a02 11450 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_IN_CP_AREA)\n");
10f22854
AK		 11451 goto abort;
11452 }
8016a6d4
AK		 11453 } else {
11454 /* set next step to use whole border area */
11455 border /= next_step;
11456 if (border > 1)
11457 next_step *= border;
10f22854
AK		 11458 }
11459 /* When data backed up, checkpoint stored,
11460 * kick the kernel to reshape unit of data
11461 */
11462 next_step = next_step + sra->reshape_progress;
8016a6d4
AK		 11463 /* limit next step to array max position */
11464 if (next_step > max_position)
11465 next_step = max_position;
10f22854
AK		 11466 sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress);
11467 sysfs_set_num(sra, NULL, "suspend_hi", next_step);
ae9f01f8 11468 sra->reshape_progress = next_step;
10f22854
AK		 11469
11470 /* wait until reshape finish */
c85338c6 11471 if (wait_for_reshape_imsm(sra, ndata)) {
c47b0ff6
AK		 11472 dprintf("wait_for_reshape_imsm returned error!\n");
11473 goto abort;
11474 }
84d11e6c
N		 11475 if (sigterm)
11476 goto abort;
10f22854 11477
0228d92c
AK		 11478 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
11479 /* ignore error == 2, this can mean end of reshape here
11480 */
7a862a02 11481 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL)\n");
10f22854
AK		 11482 goto abort;
11483 }
11484
11485 }
11486
71e5411e
PB		 11487 /* clear migr_rec on disks after successful migration */
11488 struct dl *d;
11489
de44e46f 11490 memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*sector_size);
71e5411e
PB		 11491 for (d = super->disks; d; d = d->next) {
11492 if (d->index < 0 || is_failed(&d->disk))
11493 continue;
11494 unsigned long long dsize;
11495
11496 get_dev_size(d->fd, NULL, &dsize);
de44e46f 11497 if (lseek64(d->fd, dsize - MIGR_REC_SECTOR_POSITION*sector_size,
71e5411e
PB		 11498 SEEK_SET) >= 0) {
11499 if (write(d->fd, super->migr_rec_buf,
de44e46f
PB		 11500 MIGR_REC_BUF_SECTORS*sector_size) !=
11501 MIGR_REC_BUF_SECTORS*sector_size)
71e5411e
PB		 11502 perror("Write migr_rec failed");
11503 }
11504 }
11505
10f22854
AK		 11506 /* return '1' if done */
11507 ret_val = 1;
11508abort:
11509 free(buf);
942e1cdb
N		 11510 /* See Grow.c: abort_reshape() for further explanation */
11511 sysfs_set_num(sra, NULL, "suspend_lo", 0x7FFFFFFFFFFFFFFFULL);
11512 sysfs_set_num(sra, NULL, "suspend_hi", 0);
11513 sysfs_set_num(sra, NULL, "suspend_lo", 0);
10f22854
AK		 11514
11515 return ret_val;
999b4972 11516}
0c21b485 11517
71204a50 11518#endif /* MDASSEMBLE */
999b4972 11519
cdddbdbc
DW		 11520struct superswitch super_imsm = {
11521#ifndef MDASSEMBLE
11522 .examine_super = examine_super_imsm,
11523 .brief_examine_super = brief_examine_super_imsm,
4737ae25 11524 .brief_examine_subarrays = brief_examine_subarrays_imsm,
9d84c8ea 11525 .export_examine_super = export_examine_super_imsm,
cdddbdbc
DW		 11526 .detail_super = detail_super_imsm,
11527 .brief_detail_super = brief_detail_super_imsm,
bf5a934a 11528 .write_init_super = write_init_super_imsm,
0e600426
N		 11529 .validate_geometry = validate_geometry_imsm,
11530 .add_to_super = add_to_super_imsm,
1a64be56 11531 .remove_from_super = remove_from_super_imsm,
d665cc31 11532 .detail_platform = detail_platform_imsm,
e50cf220 11533 .export_detail_platform = export_detail_platform_imsm,
33414a01 11534 .kill_subarray = kill_subarray_imsm,
aa534678 11535 .update_subarray = update_subarray_imsm,
2b959fbf 11536 .load_container = load_container_imsm,
71204a50
N		 11537 .default_geometry = default_geometry_imsm,
11538 .get_disk_controller_domain = imsm_get_disk_controller_domain,
11539 .reshape_super = imsm_reshape_super,
11540 .manage_reshape = imsm_manage_reshape,
9e2d750d 11541 .recover_backup = recover_backup_imsm,
74db60b0 11542 .copy_metadata = copy_metadata_imsm,
27156a57 11543 .examine_badblocks = examine_badblocks_imsm,
cdddbdbc
DW		 11544#endif
11545 .match_home = match_home_imsm,
11546 .uuid_from_super= uuid_from_super_imsm,
11547 .getinfo_super = getinfo_super_imsm,
5c4cd5da 11548 .getinfo_super_disks = getinfo_super_disks_imsm,
cdddbdbc
DW		 11549 .update_super = update_super_imsm,
11550
11551 .avail_size = avail_size_imsm,
80e7f8c3 11552 .min_acceptable_spare_size = min_acceptable_spare_size_imsm,
cdddbdbc
DW		 11553
11554 .compare_super = compare_super_imsm,
11555
11556 .load_super = load_super_imsm,
bf5a934a 11557 .init_super = init_super_imsm,
e683ca88 11558 .store_super = store_super_imsm,
cdddbdbc
DW		 11559 .free_super = free_super_imsm,
11560 .match_metadata_desc = match_metadata_desc_imsm,
bf5a934a 11561 .container_content = container_content_imsm,
0c21b485 11562 .validate_container = validate_container_imsm,
cdddbdbc 11563
cdddbdbc 11564 .external = 1,
4cce4069 11565 .name = "imsm",
845dea95 11566
0e600426 11567#ifndef MDASSEMBLE
845dea95
NB		 11568/* for mdmon */
11569 .open_new = imsm_open_new,
ed9d66aa 11570 .set_array_state= imsm_set_array_state,
845dea95
NB		 11571 .set_disk = imsm_set_disk,
11572 .sync_metadata = imsm_sync_metadata,
88758e9d 11573 .activate_spare = imsm_activate_spare,
e8319a19 11574 .process_update = imsm_process_update,
8273f55e 11575 .prepare_update = imsm_prepare_update,
6f50473f 11576 .record_bad_block = imsm_record_badblock,
c07a5a4f 11577 .clear_bad_block = imsm_clear_badblock,
928f1424 11578 .get_bad_blocks = imsm_get_badblocks,
0e600426 11579#endif /* MDASSEMBLE */
cdddbdbc 11580};
Unnamed repository; edit this file 'description' to name the repository.