]> 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
util: Introduce md_set_array_info()
[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 */
2432ce9b 105#define JOURNAL_DISK __cpu_to_le32(0x2000000) /* Device marked as Journaling Drive */
cdddbdbc 106 __u32 status; /* 0xF0 - 0xF3 */
1011e834 107 __u32 owner_cfg_num; /* which config 0,1,2... owns this disk */
5551b113
CA		 108 __u32 total_blocks_hi; /* 0xF4 - 0xF5 total blocks hi */
109#define IMSM_DISK_FILLERS 3
110 __u32 filler[IMSM_DISK_FILLERS]; /* 0xF5 - 0x107 MPB_DISK_FILLERS for future expansion */
cdddbdbc
DW		 111};
112
3b451610
AK		 113/* map selector for map managment
114 */
238c0a71
AK		 115#define MAP_0 0
116#define MAP_1 1
117#define MAP_X -1
3b451610 118
cdddbdbc
DW		 119/* RAID map configuration infos. */
120struct imsm_map {
5551b113
CA		 121 __u32 pba_of_lba0_lo; /* start address of partition */
122 __u32 blocks_per_member_lo;/* blocks per member */
123 __u32 num_data_stripes_lo; /* number of data stripes */
cdddbdbc
DW		 124 __u16 blocks_per_strip;
125 __u8 map_state; /* Normal, Uninitialized, Degraded, Failed */
126#define IMSM_T_STATE_NORMAL 0
127#define IMSM_T_STATE_UNINITIALIZED 1
e3bba0e0
DW		 128#define IMSM_T_STATE_DEGRADED 2
129#define IMSM_T_STATE_FAILED 3
cdddbdbc
DW		 130 __u8 raid_level;
131#define IMSM_T_RAID0 0
132#define IMSM_T_RAID1 1
133#define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */
134 __u8 num_members; /* number of member disks */
fe7ed8cb
DW		 135 __u8 num_domains; /* number of parity domains */
136 __u8 failed_disk_num; /* valid only when state is degraded */
252d23c0 137 __u8 ddf;
5551b113
CA		 138 __u32 pba_of_lba0_hi;
139 __u32 blocks_per_member_hi;
140 __u32 num_data_stripes_hi;
141 __u32 filler[4]; /* expansion area */
7eef0453 142#define IMSM_ORD_REBUILD (1 << 24)
cdddbdbc 143 __u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members],
7eef0453
DW		 144 * top byte contains some flags
145 */
cdddbdbc
DW		 146} __attribute__ ((packed));
147
148struct imsm_vol {
f8f603f1 149 __u32 curr_migr_unit;
fe7ed8cb 150 __u32 checkpoint_id; /* id to access curr_migr_unit */
cdddbdbc 151 __u8 migr_state; /* Normal or Migrating */
e3bba0e0
DW		 152#define MIGR_INIT 0
153#define MIGR_REBUILD 1
154#define MIGR_VERIFY 2 /* analagous to echo check > sync_action */
155#define MIGR_GEN_MIGR 3
156#define MIGR_STATE_CHANGE 4
1484e727 157#define MIGR_REPAIR 5
cdddbdbc 158 __u8 migr_type; /* Initializing, Rebuilding, ... */
2432ce9b
AP		 159#define RAIDVOL_CLEAN 0
160#define RAIDVOL_DIRTY 1
161#define RAIDVOL_DSRECORD_VALID 2
cdddbdbc 162 __u8 dirty;
fe7ed8cb
DW		 163 __u8 fs_state; /* fast-sync state for CnG (0xff == disabled) */
164 __u16 verify_errors; /* number of mismatches */
165 __u16 bad_blocks; /* number of bad blocks during verify */
166 __u32 filler[4];
cdddbdbc
DW		 167 struct imsm_map map[1];
168 /* here comes another one if migr_state */
169} __attribute__ ((packed));
170
171struct imsm_dev {
fe7ed8cb 172 __u8 volume[MAX_RAID_SERIAL_LEN];
cdddbdbc
DW		 173 __u32 size_low;
174 __u32 size_high;
fe7ed8cb
DW		 175#define DEV_BOOTABLE __cpu_to_le32(0x01)
176#define DEV_BOOT_DEVICE __cpu_to_le32(0x02)
177#define DEV_READ_COALESCING __cpu_to_le32(0x04)
178#define DEV_WRITE_COALESCING __cpu_to_le32(0x08)
179#define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10)
180#define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20)
181#define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40)
182#define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80)
183#define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100)
184#define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200)
185#define DEV_CLONE_N_GO __cpu_to_le32(0x400)
186#define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800)
187#define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000)
cdddbdbc
DW		 188 __u32 status; /* Persistent RaidDev status */
189 __u32 reserved_blocks; /* Reserved blocks at beginning of volume */
fe7ed8cb
DW		 190 __u8 migr_priority;
191 __u8 num_sub_vols;
192 __u8 tid;
193 __u8 cng_master_disk;
194 __u16 cache_policy;
195 __u8 cng_state;
196 __u8 cng_sub_state;
2432ce9b
AP		 197 __u16 my_vol_raid_dev_num; /* Used in Unique volume Id for this RaidDev */
198
199 /* NVM_EN */
200 __u8 nv_cache_mode;
201 __u8 nv_cache_flags;
202
203 /* Unique Volume Id of the NvCache Volume associated with this volume */
204 __u32 nvc_vol_orig_family_num;
205 __u16 nvc_vol_raid_dev_num;
206
207#define RWH_OFF 0
208#define RWH_DISTRIBUTED 1
209#define RWH_JOURNALING_DRIVE 2
210 __u8 rwh_policy; /* Raid Write Hole Policy */
211 __u8 jd_serial[MAX_RAID_SERIAL_LEN]; /* Journal Drive serial number */
212 __u8 filler1;
213
214#define IMSM_DEV_FILLERS 3
cdddbdbc
DW		 215 __u32 filler[IMSM_DEV_FILLERS];
216 struct imsm_vol vol;
217} __attribute__ ((packed));
218
219struct imsm_super {
220 __u8 sig[MAX_SIGNATURE_LENGTH]; /* 0x00 - 0x1F */
221 __u32 check_sum; /* 0x20 - 0x23 MPB Checksum */
222 __u32 mpb_size; /* 0x24 - 0x27 Size of MPB */
223 __u32 family_num; /* 0x28 - 0x2B Checksum from first time this config was written */
224 __u32 generation_num; /* 0x2C - 0x2F Incremented each time this array's MPB is written */
604b746f
JD		 225 __u32 error_log_size; /* 0x30 - 0x33 in bytes */
226 __u32 attributes; /* 0x34 - 0x37 */
cdddbdbc
DW		 227 __u8 num_disks; /* 0x38 Number of configured disks */
228 __u8 num_raid_devs; /* 0x39 Number of configured volumes */
604b746f
JD		 229 __u8 error_log_pos; /* 0x3A */
230 __u8 fill[1]; /* 0x3B */
231 __u32 cache_size; /* 0x3c - 0x40 in mb */
232 __u32 orig_family_num; /* 0x40 - 0x43 original family num */
233 __u32 pwr_cycle_count; /* 0x44 - 0x47 simulated power cycle count for array */
234 __u32 bbm_log_size; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */
235#define IMSM_FILLERS 35
236 __u32 filler[IMSM_FILLERS]; /* 0x4C - 0xD7 RAID_MPB_FILLERS */
cdddbdbc
DW		 237 struct imsm_disk disk[1]; /* 0xD8 diskTbl[numDisks] */
238 /* here comes imsm_dev[num_raid_devs] */
604b746f 239 /* here comes BBM logs */
cdddbdbc
DW		 240} __attribute__ ((packed));
241
604b746f 242#define BBM_LOG_MAX_ENTRIES 254
8d67477f
TM		 243#define BBM_LOG_MAX_LBA_ENTRY_VAL 256 /* Represents 256 LBAs */
244#define BBM_LOG_SIGNATURE 0xabadb10c
245
246struct bbm_log_block_addr {
247 __u16 w1;
248 __u32 dw1;
249} __attribute__ ((__packed__));
604b746f
JD		 250
251struct bbm_log_entry {
8d67477f
TM		 252 __u8 marked_count; /* Number of blocks marked - 1 */
253 __u8 disk_ordinal; /* Disk entry within the imsm_super */
254 struct bbm_log_block_addr defective_block_start;
604b746f
JD		 255} __attribute__ ((__packed__));
256
257struct bbm_log {
258 __u32 signature; /* 0xABADB10C */
259 __u32 entry_count;
8d67477f 260 struct bbm_log_entry marked_block_entries[BBM_LOG_MAX_ENTRIES];
604b746f
JD		 261} __attribute__ ((__packed__));
262
cdddbdbc
DW		 263#ifndef MDASSEMBLE
264static char *map_state_str[] = { "normal", "uninitialized", "degraded", "failed" };
265#endif
266
8e59f3d8
AK		 267#define RAID_DISK_RESERVED_BLOCKS_IMSM_HI 2209
268
269#define GEN_MIGR_AREA_SIZE 2048 /* General Migration Copy Area size in blocks */
270
de44e46f
PB		 271#define MIGR_REC_BUF_SECTORS 1 /* size of migr_record i/o buffer in sectors */
272#define MIGR_REC_SECTOR_POSITION 1 /* migr_record position offset on disk,
273 * MIGR_REC_BUF_SECTORS <= MIGR_REC_SECTOR_POS
17a4eaf9
AK		 274 */
275
8e59f3d8
AK		 276#define UNIT_SRC_NORMAL 0 /* Source data for curr_migr_unit must
277 * be recovered using srcMap */
278#define UNIT_SRC_IN_CP_AREA 1 /* Source data for curr_migr_unit has
279 * already been migrated and must
280 * be recovered from checkpoint area */
2432ce9b
AP		 281
282#define PPL_ENTRY_SPACE (128 * 1024) /* Size of the PPL, without the header */
283
8e59f3d8
AK		 284struct migr_record {
285 __u32 rec_status; /* Status used to determine how to restart
286 * migration in case it aborts
287 * in some fashion */
288 __u32 curr_migr_unit; /* 0..numMigrUnits-1 */
289 __u32 family_num; /* Family number of MPB
290 * containing the RaidDev
291 * that is migrating */
292 __u32 ascending_migr; /* True if migrating in increasing
293 * order of lbas */
294 __u32 blocks_per_unit; /* Num disk blocks per unit of operation */
295 __u32 dest_depth_per_unit; /* Num member blocks each destMap
296 * member disk
297 * advances per unit-of-operation */
298 __u32 ckpt_area_pba; /* Pba of first block of ckpt copy area */
299 __u32 dest_1st_member_lba; /* First member lba on first
300 * stripe of destination */
301 __u32 num_migr_units; /* Total num migration units-of-op */
302 __u32 post_migr_vol_cap; /* Size of volume after
303 * migration completes */
304 __u32 post_migr_vol_cap_hi; /* Expansion space for LBA64 */
305 __u32 ckpt_read_disk_num; /* Which member disk in destSubMap[0] the
306 * migration ckpt record was read from
307 * (for recovered migrations) */
308} __attribute__ ((__packed__));
309
ec50f7b6
LM		 310struct md_list {
311 /* usage marker:
312 * 1: load metadata
313 * 2: metadata does not match
314 * 4: already checked
315 */
316 int used;
317 char *devname;
318 int found;
319 int container;
320 dev_t st_rdev;
321 struct md_list *next;
322};
323
e7b84f9d 324#define pr_vrb(fmt, arg...) (void) (verbose && pr_err(fmt, ##arg))
ec50f7b6 325
1484e727
DW		 326static __u8 migr_type(struct imsm_dev *dev)
327{
328 if (dev->vol.migr_type == MIGR_VERIFY &&
329 dev->status & DEV_VERIFY_AND_FIX)
330 return MIGR_REPAIR;
331 else
332 return dev->vol.migr_type;
333}
334
335static void set_migr_type(struct imsm_dev *dev, __u8 migr_type)
336{
337 /* for compatibility with older oroms convert MIGR_REPAIR, into
338 * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status
339 */
340 if (migr_type == MIGR_REPAIR) {
341 dev->vol.migr_type = MIGR_VERIFY;
342 dev->status |= DEV_VERIFY_AND_FIX;
343 } else {
344 dev->vol.migr_type = migr_type;
345 dev->status &= ~DEV_VERIFY_AND_FIX;
346 }
347}
348
f36a9ecd 349static unsigned int sector_count(__u32 bytes, unsigned int sector_size)
cdddbdbc 350{
f36a9ecd 351 return ROUND_UP(bytes, sector_size) / sector_size;
87eb16df 352}
cdddbdbc 353
f36a9ecd
PB		 354static unsigned int mpb_sectors(struct imsm_super *mpb,
355 unsigned int sector_size)
87eb16df 356{
f36a9ecd 357 return sector_count(__le32_to_cpu(mpb->mpb_size), sector_size);
cdddbdbc
DW		 358}
359
ba2de7ba
DW		 360struct intel_dev {
361 struct imsm_dev *dev;
362 struct intel_dev *next;
f21e18ca 363 unsigned index;
ba2de7ba
DW		 364};
365
88654014
LM		 366struct intel_hba {
367 enum sys_dev_type type;
368 char *path;
369 char *pci_id;
370 struct intel_hba *next;
371};
372
1a64be56
LM		 373enum action {
374 DISK_REMOVE = 1,
375 DISK_ADD
376};
cdddbdbc
DW		 377/* internal representation of IMSM metadata */
378struct intel_super {
379 union {
949c47a0
DW		 380 void *buf; /* O_DIRECT buffer for reading/writing metadata */
381 struct imsm_super *anchor; /* immovable parameters */
cdddbdbc 382 };
8e59f3d8
AK		 383 union {
384 void *migr_rec_buf; /* buffer for I/O operations */
385 struct migr_record *migr_rec; /* migration record */
386 };
51d83f5d
AK		 387 int clean_migration_record_by_mdmon; /* when reshape is switched to next
388 array, it indicates that mdmon is allowed to clean migration
389 record */
949c47a0 390 size_t len; /* size of the 'buf' allocation */
bbab0940 391 size_t extra_space; /* extra space in 'buf' that is not used yet */
4d7b1503
DW		 392 void *next_buf; /* for realloc'ing buf from the manager */
393 size_t next_len;
c2c087e6 394 int updates_pending; /* count of pending updates for mdmon */
bf5a934a 395 int current_vol; /* index of raid device undergoing creation */
5551b113 396 unsigned long long create_offset; /* common start for 'current_vol' */
148acb7b 397 __u32 random; /* random data for seeding new family numbers */
ba2de7ba 398 struct intel_dev *devlist;
fa7bb6f8 399 unsigned int sector_size; /* sector size of used member drives */
cdddbdbc
DW		 400 struct dl {
401 struct dl *next;
402 int index;
403 __u8 serial[MAX_RAID_SERIAL_LEN];
404 int major, minor;
405 char *devname;
b9f594fe 406 struct imsm_disk disk;
cdddbdbc 407 int fd;
0dcecb2e
DW		 408 int extent_cnt;
409 struct extent *e; /* for determining freespace @ create */
efb30e7f 410 int raiddisk; /* slot to fill in autolayout */
1a64be56 411 enum action action;
ca0748fa 412 } *disks, *current_disk;
1a64be56
LM		 413 struct dl *disk_mgmt_list; /* list of disks to add/remove while mdmon
414 active */
47ee5a45 415 struct dl *missing; /* disks removed while we weren't looking */
43dad3d6 416 struct bbm_log *bbm_log;
88654014 417 struct intel_hba *hba; /* device path of the raid controller for this metadata */
88c32bb1 418 const struct imsm_orom *orom; /* platform firmware support */
a2b97981 419 struct intel_super *next; /* (temp) list for disambiguating family_num */
928f1424 420 struct md_bb bb; /* memory for get_bad_blocks call */
a2b97981
DW		 421};
422
423struct intel_disk {
424 struct imsm_disk disk;
425 #define IMSM_UNKNOWN_OWNER (-1)
426 int owner;
427 struct intel_disk *next;
cdddbdbc
DW		 428};
429
c2c087e6
DW		 430struct extent {
431 unsigned long long start, size;
432};
433
694575e7
KW		 434/* definitions of reshape process types */
435enum imsm_reshape_type {
436 CH_TAKEOVER,
b5347799 437 CH_MIGRATION,
7abc9871 438 CH_ARRAY_SIZE,
694575e7
KW		 439};
440
88758e9d
DW		 441/* definition of messages passed to imsm_process_update */
442enum imsm_update_type {
443 update_activate_spare,
8273f55e 444 update_create_array,
33414a01 445 update_kill_array,
aa534678 446 update_rename_array,
1a64be56 447 update_add_remove_disk,
78b10e66 448 update_reshape_container_disks,
48c5303a 449 update_reshape_migration,
2d40f3a1
AK		 450 update_takeover,
451 update_general_migration_checkpoint,
f3871fdc 452 update_size_change,
bbab0940 453 update_prealloc_badblocks_mem,
e6e9dd3f 454 update_rwh_policy,
88758e9d
DW		 455};
456
457struct imsm_update_activate_spare {
458 enum imsm_update_type type;
d23fe947 459 struct dl *dl;
88758e9d
DW		 460 int slot;
461 int array;
462 struct imsm_update_activate_spare *next;
463};
464
78b10e66 465struct geo_params {
4dd2df09 466 char devnm[32];
78b10e66 467 char *dev_name;
d04f65f4 468 unsigned long long size;
78b10e66
N		 469 int level;
470 int layout;
471 int chunksize;
472 int raid_disks;
473};
474
bb025c2f
KW		 475enum takeover_direction {
476 R10_TO_R0,
477 R0_TO_R10
478};
479struct imsm_update_takeover {
480 enum imsm_update_type type;
481 int subarray;
482 enum takeover_direction direction;
483};
78b10e66
N		 484
485struct imsm_update_reshape {
486 enum imsm_update_type type;
487 int old_raid_disks;
488 int new_raid_disks;
48c5303a
PC		 489
490 int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */
491};
492
493struct imsm_update_reshape_migration {
494 enum imsm_update_type type;
495 int old_raid_disks;
496 int new_raid_disks;
497 /* fields for array migration changes
498 */
499 int subdev;
500 int new_level;
501 int new_layout;
4bba0439 502 int new_chunksize;
48c5303a 503
d195167d 504 int new_disks[1]; /* new_raid_disks - old_raid_disks makedev number */
78b10e66
N		 505};
506
f3871fdc
AK		 507struct imsm_update_size_change {
508 enum imsm_update_type type;
509 int subdev;
510 long long new_size;
511};
512
2d40f3a1
AK		 513struct imsm_update_general_migration_checkpoint {
514 enum imsm_update_type type;
515 __u32 curr_migr_unit;
516};
517
54c2c1ea
DW		 518struct disk_info {
519 __u8 serial[MAX_RAID_SERIAL_LEN];
520};
521
8273f55e
DW		 522struct imsm_update_create_array {
523 enum imsm_update_type type;
8273f55e 524 int dev_idx;
6a3e913e 525 struct imsm_dev dev;
8273f55e
DW		 526};
527
33414a01
DW		 528struct imsm_update_kill_array {
529 enum imsm_update_type type;
530 int dev_idx;
531};
532
aa534678
DW		 533struct imsm_update_rename_array {
534 enum imsm_update_type type;
535 __u8 name[MAX_RAID_SERIAL_LEN];
536 int dev_idx;
537};
538
1a64be56 539struct imsm_update_add_remove_disk {
43dad3d6
DW		 540 enum imsm_update_type type;
541};
542
bbab0940
TM		 543struct imsm_update_prealloc_bb_mem {
544 enum imsm_update_type type;
545};
546
e6e9dd3f
AP		 547struct imsm_update_rwh_policy {
548 enum imsm_update_type type;
549 int new_policy;
550 int dev_idx;
551};
552
88654014
LM		 553static const char *_sys_dev_type[] = {
554 [SYS_DEV_UNKNOWN] = "Unknown",
555 [SYS_DEV_SAS] = "SAS",
614902f6 556 [SYS_DEV_SATA] = "SATA",
60f0f54d
PB		 557 [SYS_DEV_NVME] = "NVMe",
558 [SYS_DEV_VMD] = "VMD"
88654014
LM		 559};
560
561const char *get_sys_dev_type(enum sys_dev_type type)
562{
563 if (type >= SYS_DEV_MAX)
564 type = SYS_DEV_UNKNOWN;
565
566 return _sys_dev_type[type];
567}
568
569static struct intel_hba * alloc_intel_hba(struct sys_dev *device)
570{
503975b9
N		 571 struct intel_hba *result = xmalloc(sizeof(*result));
572
573 result->type = device->type;
574 result->path = xstrdup(device->path);
575 result->next = NULL;
576 if (result->path && (result->pci_id = strrchr(result->path, '/')) != NULL)
577 result->pci_id++;
578
88654014
LM		 579 return result;
580}
581
582static struct intel_hba * find_intel_hba(struct intel_hba *hba, struct sys_dev *device)
583{
594dc1b8
JS		 584 struct intel_hba *result;
585
88654014
LM		 586 for (result = hba; result; result = result->next) {
587 if (result->type == device->type && strcmp(result->path, device->path) == 0)
588 break;
589 }
590 return result;
591}
592
b4cf4cba 593static int attach_hba_to_super(struct intel_super *super, struct sys_dev *device)
88654014
LM		 594{
595 struct intel_hba *hba;
596
597 /* check if disk attached to Intel HBA */
598 hba = find_intel_hba(super->hba, device);
599 if (hba != NULL)
600 return 1;
601 /* Check if HBA is already attached to super */
602 if (super->hba == NULL) {
603 super->hba = alloc_intel_hba(device);
604 return 1;
6b781d33
AP		 605 }
606
607 hba = super->hba;
608 /* Intel metadata allows for all disks attached to the same type HBA.
614902f6 609 * Do not support HBA types mixing
6b781d33
AP		 610 */
611 if (device->type != hba->type)
88654014 612 return 2;
6b781d33
AP		 613
614 /* Multiple same type HBAs can be used if they share the same OROM */
615 const struct imsm_orom *device_orom = get_orom_by_device_id(device->dev_id);
616
617 if (device_orom != super->orom)
618 return 2;
619
620 while (hba->next)
621 hba = hba->next;
622
623 hba->next = alloc_intel_hba(device);
624 return 1;
88654014
LM		 625}
626
627static struct sys_dev* find_disk_attached_hba(int fd, const char *devname)
628{
9bc4ae77 629 struct sys_dev *list, *elem;
88654014
LM		 630 char *disk_path;
631
632 if ((list = find_intel_devices()) == NULL)
633 return 0;
634
635 if (fd < 0)
636 disk_path = (char *) devname;
637 else
638 disk_path = diskfd_to_devpath(fd);
639
9bc4ae77 640 if (!disk_path)
88654014 641 return 0;
88654014 642
9bc4ae77
N		 643 for (elem = list; elem; elem = elem->next)
644 if (path_attached_to_hba(disk_path, elem->path))
88654014 645 return elem;
9bc4ae77 646
88654014
LM		 647 if (disk_path != devname)
648 free(disk_path);
88654014
LM		 649
650 return NULL;
651}
652
d424212e
N		 653static int find_intel_hba_capability(int fd, struct intel_super *super,
654 char *devname);
f2f5c343 655
cdddbdbc
DW		 656static struct supertype *match_metadata_desc_imsm(char *arg)
657{
658 struct supertype *st;
659
660 if (strcmp(arg, "imsm") != 0 &&
661 strcmp(arg, "default") != 0
662 )
663 return NULL;
664
503975b9 665 st = xcalloc(1, sizeof(*st));
cdddbdbc
DW		 666 st->ss = &super_imsm;
667 st->max_devs = IMSM_MAX_DEVICES;
668 st->minor_version = 0;
669 st->sb = NULL;
670 return st;
671}
672
0e600426 673#ifndef MDASSEMBLE
cdddbdbc
DW		 674static __u8 *get_imsm_version(struct imsm_super *mpb)
675{
676 return &mpb->sig[MPB_SIG_LEN];
677}
9e2d750d 678#endif
cdddbdbc 679
949c47a0
DW		 680/* retrieve a disk directly from the anchor when the anchor is known to be
681 * up-to-date, currently only at load time
682 */
683static struct imsm_disk *__get_imsm_disk(struct imsm_super *mpb, __u8 index)
cdddbdbc 684{
949c47a0 685 if (index >= mpb->num_disks)
cdddbdbc
DW		 686 return NULL;
687 return &mpb->disk[index];
688}
689
95d07a2c
LM		 690/* retrieve the disk description based on a index of the disk
691 * in the sub-array
692 */
693static struct dl *get_imsm_dl_disk(struct intel_super *super, __u8 index)
949c47a0 694{
b9f594fe
DW		 695 struct dl *d;
696
697 for (d = super->disks; d; d = d->next)
698 if (d->index == index)
95d07a2c
LM		 699 return d;
700
701 return NULL;
702}
703/* retrieve a disk from the parsed metadata */
704static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index)
705{
706 struct dl *dl;
707
708 dl = get_imsm_dl_disk(super, index);
709 if (dl)
710 return &dl->disk;
711
b9f594fe 712 return NULL;
949c47a0
DW		 713}
714
715/* generate a checksum directly from the anchor when the anchor is known to be
716 * up-to-date, currently only at load or write_super after coalescing
717 */
718static __u32 __gen_imsm_checksum(struct imsm_super *mpb)
cdddbdbc
DW		 719{
720 __u32 end = mpb->mpb_size / sizeof(end);
721 __u32 *p = (__u32 *) mpb;
722 __u32 sum = 0;
723
5d500228
N		 724 while (end--) {
725 sum += __le32_to_cpu(*p);
97f734fd
N		 726 p++;
727 }
cdddbdbc 728
5d500228 729 return sum - __le32_to_cpu(mpb->check_sum);
cdddbdbc
DW		 730}
731
a965f303
DW		 732static size_t sizeof_imsm_map(struct imsm_map *map)
733{
734 return sizeof(struct imsm_map) + sizeof(__u32) * (map->num_members - 1);
735}
736
737struct imsm_map *get_imsm_map(struct imsm_dev *dev, int second_map)
cdddbdbc 738{
5e7b0330
AK		 739 /* A device can have 2 maps if it is in the middle of a migration.
740 * If second_map is:
238c0a71
AK		 741 * MAP_0 - we return the first map
742 * MAP_1 - we return the second map if it exists, else NULL
743 * MAP_X - we return the second map if it exists, else the first
5e7b0330 744 */
a965f303 745 struct imsm_map *map = &dev->vol.map[0];
9535fc47 746 struct imsm_map *map2 = NULL;
a965f303 747
9535fc47
AK		 748 if (dev->vol.migr_state)
749 map2 = (void *)map + sizeof_imsm_map(map);
a965f303 750
9535fc47 751 switch (second_map) {
3b451610 752 case MAP_0:
9535fc47 753 break;
3b451610 754 case MAP_1:
9535fc47
AK		 755 map = map2;
756 break;
238c0a71 757 case MAP_X:
9535fc47
AK		 758 if (map2)
759 map = map2;
760 break;
9535fc47
AK		 761 default:
762 map = NULL;
763 }
764 return map;
5e7b0330 765
a965f303 766}
cdddbdbc 767
3393c6af
DW		 768/* return the size of the device.
769 * migr_state increases the returned size if map[0] were to be duplicated
770 */
771static size_t sizeof_imsm_dev(struct imsm_dev *dev, int migr_state)
a965f303
DW		 772{
773 size_t size = sizeof(*dev) - sizeof(struct imsm_map) +
238c0a71 774 sizeof_imsm_map(get_imsm_map(dev, MAP_0));
cdddbdbc
DW		 775
776 /* migrating means an additional map */
a965f303 777 if (dev->vol.migr_state)
238c0a71 778 size += sizeof_imsm_map(get_imsm_map(dev, MAP_1));
3393c6af 779 else if (migr_state)
238c0a71 780 size += sizeof_imsm_map(get_imsm_map(dev, MAP_0));
cdddbdbc
DW		 781
782 return size;
783}
784
54c2c1ea
DW		 785#ifndef MDASSEMBLE
786/* retrieve disk serial number list from a metadata update */
787static struct disk_info *get_disk_info(struct imsm_update_create_array *update)
788{
789 void *u = update;
790 struct disk_info *inf;
791
792 inf = u + sizeof(*update) - sizeof(struct imsm_dev) +
793 sizeof_imsm_dev(&update->dev, 0);
794
795 return inf;
796}
797#endif
798
949c47a0 799static struct imsm_dev *__get_imsm_dev(struct imsm_super *mpb, __u8 index)
cdddbdbc
DW		 800{
801 int offset;
802 int i;
803 void *_mpb = mpb;
804
949c47a0 805 if (index >= mpb->num_raid_devs)
cdddbdbc
DW		 806 return NULL;
807
808 /* devices start after all disks */
809 offset = ((void *) &mpb->disk[mpb->num_disks]) - _mpb;
810
811 for (i = 0; i <= index; i++)
812 if (i == index)
813 return _mpb + offset;
814 else
3393c6af 815 offset += sizeof_imsm_dev(_mpb + offset, 0);
cdddbdbc
DW		 816
817 return NULL;
818}
819
949c47a0
DW		 820static struct imsm_dev *get_imsm_dev(struct intel_super *super, __u8 index)
821{
ba2de7ba
DW		 822 struct intel_dev *dv;
823
949c47a0
DW		 824 if (index >= super->anchor->num_raid_devs)
825 return NULL;
ba2de7ba
DW		 826 for (dv = super->devlist; dv; dv = dv->next)
827 if (dv->index == index)
828 return dv->dev;
829 return NULL;
949c47a0
DW		 830}
831
8d67477f
TM		 832static inline unsigned long long __le48_to_cpu(const struct bbm_log_block_addr
833 *addr)
834{
835 return ((((__u64)__le32_to_cpu(addr->dw1)) << 16) |
836 __le16_to_cpu(addr->w1));
837}
838
839static inline struct bbm_log_block_addr __cpu_to_le48(unsigned long long sec)
840{
841 struct bbm_log_block_addr addr;
842
843 addr.w1 = __cpu_to_le16((__u16)(sec & 0xffff));
844 addr.dw1 = __cpu_to_le32((__u32)(sec >> 16) & 0xffffffff);
845 return addr;
846}
847
848#ifndef MDASSEMBLE
849/* get size of the bbm log */
850static __u32 get_imsm_bbm_log_size(struct bbm_log *log)
851{
852 if (!log || log->entry_count == 0)
853 return 0;
854
855 return sizeof(log->signature) +
856 sizeof(log->entry_count) +
857 log->entry_count * sizeof(struct bbm_log_entry);
858}
6f50473f
TM		 859
860/* check if bad block is not partially stored in bbm log */
861static int is_stored_in_bbm(struct bbm_log *log, const __u8 idx, const unsigned
862 long long sector, const int length, __u32 *pos)
863{
864 __u32 i;
865
866 for (i = *pos; i < log->entry_count; i++) {
867 struct bbm_log_entry *entry = &log->marked_block_entries[i];
868 unsigned long long bb_start;
869 unsigned long long bb_end;
870
871 bb_start = __le48_to_cpu(&entry->defective_block_start);
872 bb_end = bb_start + (entry->marked_count + 1);
873
874 if ((entry->disk_ordinal == idx) && (bb_start >= sector) &&
875 (bb_end <= sector + length)) {
876 *pos = i;
877 return 1;
878 }
879 }
880 return 0;
881}
882
883/* record new bad block in bbm log */
884static int record_new_badblock(struct bbm_log *log, const __u8 idx, unsigned
885 long long sector, int length)
886{
887 int new_bb = 0;
888 __u32 pos = 0;
889 struct bbm_log_entry *entry = NULL;
890
891 while (is_stored_in_bbm(log, idx, sector, length, &pos)) {
892 struct bbm_log_entry *e = &log->marked_block_entries[pos];
893
894 if ((e->marked_count + 1 == BBM_LOG_MAX_LBA_ENTRY_VAL) &&
895 (__le48_to_cpu(&e->defective_block_start) == sector)) {
896 sector += BBM_LOG_MAX_LBA_ENTRY_VAL;
897 length -= BBM_LOG_MAX_LBA_ENTRY_VAL;
898 pos = pos + 1;
899 continue;
900 }
901 entry = e;
902 break;
903 }
904
905 if (entry) {
906 int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length :
907 BBM_LOG_MAX_LBA_ENTRY_VAL;
908 entry->defective_block_start = __cpu_to_le48(sector);
909 entry->marked_count = cnt - 1;
910 if (cnt == length)
911 return 1;
912 sector += cnt;
913 length -= cnt;
914 }
915
916 new_bb = ROUND_UP(length, BBM_LOG_MAX_LBA_ENTRY_VAL) /
917 BBM_LOG_MAX_LBA_ENTRY_VAL;
918 if (log->entry_count + new_bb > BBM_LOG_MAX_ENTRIES)
919 return 0;
920
921 while (length > 0) {
922 int cnt = (length <= BBM_LOG_MAX_LBA_ENTRY_VAL) ? length :
923 BBM_LOG_MAX_LBA_ENTRY_VAL;
924 struct bbm_log_entry *entry =
925 &log->marked_block_entries[log->entry_count];
926
927 entry->defective_block_start = __cpu_to_le48(sector);
928 entry->marked_count = cnt - 1;
929 entry->disk_ordinal = idx;
930
931 sector += cnt;
932 length -= cnt;
933
934 log->entry_count++;
935 }
936
937 return new_bb;
938}
c07a5a4f 939
4c9e8c1e
TM		 940/* clear all bad blocks for given disk */
941static void clear_disk_badblocks(struct bbm_log *log, const __u8 idx)
942{
943 __u32 i = 0;
944
945 while (i < log->entry_count) {
946 struct bbm_log_entry *entries = log->marked_block_entries;
947
948 if (entries[i].disk_ordinal == idx) {
949 if (i < log->entry_count - 1)
950 entries[i] = entries[log->entry_count - 1];
951 log->entry_count--;
952 } else {
953 i++;
954 }
955 }
956}
957
c07a5a4f
TM		 958/* clear given bad block */
959static int clear_badblock(struct bbm_log *log, const __u8 idx, const unsigned
960 long long sector, const int length) {
961 __u32 i = 0;
962
963 while (i < log->entry_count) {
964 struct bbm_log_entry *entries = log->marked_block_entries;
965
966 if ((entries[i].disk_ordinal == idx) &&
967 (__le48_to_cpu(&entries[i].defective_block_start) ==
968 sector) && (entries[i].marked_count + 1 == length)) {
969 if (i < log->entry_count - 1)
970 entries[i] = entries[log->entry_count - 1];
971 log->entry_count--;
972 break;
973 }
974 i++;
975 }
976
977 return 1;
978}
8d67477f
TM		 979#endif /* MDASSEMBLE */
980
981/* allocate and load BBM log from metadata */
982static int load_bbm_log(struct intel_super *super)
983{
984 struct imsm_super *mpb = super->anchor;
985 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
986
987 super->bbm_log = xcalloc(1, sizeof(struct bbm_log));
988 if (!super->bbm_log)
989 return 1;
990
991 if (bbm_log_size) {
992 struct bbm_log *log = (void *)mpb +
993 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
994
995 __u32 entry_count;
996
997 if (bbm_log_size < sizeof(log->signature) +
998 sizeof(log->entry_count))
999 return 2;
1000
1001 entry_count = __le32_to_cpu(log->entry_count);
1002 if ((__le32_to_cpu(log->signature) != BBM_LOG_SIGNATURE) ||
1003 (entry_count > BBM_LOG_MAX_ENTRIES))
1004 return 3;
1005
1006 if (bbm_log_size !=
1007 sizeof(log->signature) + sizeof(log->entry_count) +
1008 entry_count * sizeof(struct bbm_log_entry))
1009 return 4;
1010
1011 memcpy(super->bbm_log, log, bbm_log_size);
1012 } else {
1013 super->bbm_log->signature = __cpu_to_le32(BBM_LOG_SIGNATURE);
1014 super->bbm_log->entry_count = 0;
1015 }
1016
1017 return 0;
1018}
1019
b12796be
TM		 1020/* checks if bad block is within volume boundaries */
1021static int is_bad_block_in_volume(const struct bbm_log_entry *entry,
1022 const unsigned long long start_sector,
1023 const unsigned long long size)
1024{
1025 unsigned long long bb_start;
1026 unsigned long long bb_end;
1027
1028 bb_start = __le48_to_cpu(&entry->defective_block_start);
1029 bb_end = bb_start + (entry->marked_count + 1);
1030
1031 if (((bb_start >= start_sector) && (bb_start < start_sector + size)) ||
1032 ((bb_end >= start_sector) && (bb_end <= start_sector + size)))
1033 return 1;
1034
1035 return 0;
1036}
1037
1038/* get list of bad blocks on a drive for a volume */
1039static void get_volume_badblocks(const struct bbm_log *log, const __u8 idx,
1040 const unsigned long long start_sector,
1041 const unsigned long long size,
1042 struct md_bb *bbs)
1043{
1044 __u32 count = 0;
1045 __u32 i;
1046
1047 for (i = 0; i < log->entry_count; i++) {
1048 const struct bbm_log_entry *ent =
1049 &log->marked_block_entries[i];
1050 struct md_bb_entry *bb;
1051
1052 if ((ent->disk_ordinal == idx) &&
1053 is_bad_block_in_volume(ent, start_sector, size)) {
1054
1055 if (!bbs->entries) {
1056 bbs->entries = xmalloc(BBM_LOG_MAX_ENTRIES *
1057 sizeof(*bb));
1058 if (!bbs->entries)
1059 break;
1060 }
1061
1062 bb = &bbs->entries[count++];
1063 bb->sector = __le48_to_cpu(&ent->defective_block_start);
1064 bb->length = ent->marked_count + 1;
1065 }
1066 }
1067 bbs->count = count;
1068}
1069
98130f40
AK		 1070/*
1071 * for second_map:
238c0a71
AK		 1072 * == MAP_0 get first map
1073 * == MAP_1 get second map
1074 * == MAP_X than get map according to the current migr_state
98130f40
AK		 1075 */
1076static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev,
1077 int slot,
1078 int second_map)
7eef0453
DW		 1079{
1080 struct imsm_map *map;
1081
5e7b0330 1082 map = get_imsm_map(dev, second_map);
7eef0453 1083
ff077194
DW		 1084 /* top byte identifies disk under rebuild */
1085 return __le32_to_cpu(map->disk_ord_tbl[slot]);
1086}
1087
1088#define ord_to_idx(ord) (((ord) << 8) >> 8)
98130f40 1089static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot, int second_map)
ff077194 1090{
98130f40 1091 __u32 ord = get_imsm_ord_tbl_ent(dev, slot, second_map);
ff077194
DW		 1092
1093 return ord_to_idx(ord);
7eef0453
DW		 1094}
1095
be73972f
DW		 1096static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord)
1097{
1098 map->disk_ord_tbl[slot] = __cpu_to_le32(ord);
1099}
1100
f21e18ca 1101static int get_imsm_disk_slot(struct imsm_map *map, unsigned idx)
620b1713
DW		 1102{
1103 int slot;
1104 __u32 ord;
1105
1106 for (slot = 0; slot < map->num_members; slot++) {
1107 ord = __le32_to_cpu(map->disk_ord_tbl[slot]);
1108 if (ord_to_idx(ord) == idx)
1109 return slot;
1110 }
1111
1112 return -1;
1113}
1114
cdddbdbc
DW		 1115static int get_imsm_raid_level(struct imsm_map *map)
1116{
1117 if (map->raid_level == 1) {
1118 if (map->num_members == 2)
1119 return 1;
1120 else
1121 return 10;
1122 }
1123
1124 return map->raid_level;
1125}
1126
c2c087e6
DW		 1127static int cmp_extent(const void *av, const void *bv)
1128{
1129 const struct extent *a = av;
1130 const struct extent *b = bv;
1131 if (a->start < b->start)
1132 return -1;
1133 if (a->start > b->start)
1134 return 1;
1135 return 0;
1136}
1137
0dcecb2e 1138static int count_memberships(struct dl *dl, struct intel_super *super)
c2c087e6 1139{
c2c087e6 1140 int memberships = 0;
620b1713 1141 int i;
c2c087e6 1142
949c47a0
DW		 1143 for (i = 0; i < super->anchor->num_raid_devs; i++) {
1144 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 1145 struct imsm_map *map = get_imsm_map(dev, MAP_0);
c2c087e6 1146
620b1713
DW		 1147 if (get_imsm_disk_slot(map, dl->index) >= 0)
1148 memberships++;
c2c087e6 1149 }
0dcecb2e
DW		 1150
1151 return memberships;
1152}
1153
b81221b7
CA		 1154static __u32 imsm_min_reserved_sectors(struct intel_super *super);
1155
5551b113
CA		 1156static int split_ull(unsigned long long n, __u32 *lo, __u32 *hi)
1157{
1158 if (lo == 0 || hi == 0)
1159 return 1;
1160 *lo = __le32_to_cpu((unsigned)n);
1161 *hi = __le32_to_cpu((unsigned)(n >> 32));
1162 return 0;
1163}
1164
1165static unsigned long long join_u32(__u32 lo, __u32 hi)
1166{
1167 return (unsigned long long)__le32_to_cpu(lo) |
1168 (((unsigned long long)__le32_to_cpu(hi)) << 32);
1169}
1170
1171static unsigned long long total_blocks(struct imsm_disk *disk)
1172{
1173 if (disk == NULL)
1174 return 0;
1175 return join_u32(disk->total_blocks_lo, disk->total_blocks_hi);
1176}
1177
1178static unsigned long long pba_of_lba0(struct imsm_map *map)
1179{
1180 if (map == NULL)
1181 return 0;
1182 return join_u32(map->pba_of_lba0_lo, map->pba_of_lba0_hi);
1183}
1184
1185static unsigned long long blocks_per_member(struct imsm_map *map)
1186{
1187 if (map == NULL)
1188 return 0;
1189 return join_u32(map->blocks_per_member_lo, map->blocks_per_member_hi);
1190}
1191
1192static unsigned long long num_data_stripes(struct imsm_map *map)
1193{
1194 if (map == NULL)
1195 return 0;
1196 return join_u32(map->num_data_stripes_lo, map->num_data_stripes_hi);
1197}
1198
1199static void set_total_blocks(struct imsm_disk *disk, unsigned long long n)
1200{
1201 split_ull(n, &disk->total_blocks_lo, &disk->total_blocks_hi);
1202}
1203
1204static void set_pba_of_lba0(struct imsm_map *map, unsigned long long n)
1205{
1206 split_ull(n, &map->pba_of_lba0_lo, &map->pba_of_lba0_hi);
1207}
1208
1209static void set_blocks_per_member(struct imsm_map *map, unsigned long long n)
1210{
1211 split_ull(n, &map->blocks_per_member_lo, &map->blocks_per_member_hi);
1212}
1213
1214static void set_num_data_stripes(struct imsm_map *map, unsigned long long n)
1215{
1216 split_ull(n, &map->num_data_stripes_lo, &map->num_data_stripes_hi);
1217}
1218
0dcecb2e
DW		 1219static struct extent *get_extents(struct intel_super *super, struct dl *dl)
1220{
1221 /* find a list of used extents on the given physical device */
1222 struct extent *rv, *e;
620b1713 1223 int i;
0dcecb2e 1224 int memberships = count_memberships(dl, super);
b276dd33
DW		 1225 __u32 reservation;
1226
1227 /* trim the reserved area for spares, so they can join any array
1228 * regardless of whether the OROM has assigned sectors from the
1229 * IMSM_RESERVED_SECTORS region
1230 */
1231 if (dl->index == -1)
b81221b7 1232 reservation = imsm_min_reserved_sectors(super);
b276dd33
DW		 1233 else
1234 reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
0dcecb2e 1235
503975b9 1236 rv = xcalloc(sizeof(struct extent), (memberships + 1));
c2c087e6
DW		 1237 e = rv;
1238
949c47a0
DW		 1239 for (i = 0; i < super->anchor->num_raid_devs; i++) {
1240 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 1241 struct imsm_map *map = get_imsm_map(dev, MAP_0);
c2c087e6 1242
620b1713 1243 if (get_imsm_disk_slot(map, dl->index) >= 0) {
5551b113
CA		 1244 e->start = pba_of_lba0(map);
1245 e->size = blocks_per_member(map);
620b1713 1246 e++;
c2c087e6
DW		 1247 }
1248 }
1249 qsort(rv, memberships, sizeof(*rv), cmp_extent);
1250
1011e834 1251 /* determine the start of the metadata
14e8215b
DW		 1252 * when no raid devices are defined use the default
1253 * ...otherwise allow the metadata to truncate the value
1254 * as is the case with older versions of imsm
1255 */
1256 if (memberships) {
1257 struct extent *last = &rv[memberships - 1];
5551b113 1258 unsigned long long remainder;
14e8215b 1259
5551b113 1260 remainder = total_blocks(&dl->disk) - (last->start + last->size);
dda5855f
DW		 1261 /* round down to 1k block to satisfy precision of the kernel
1262 * 'size' interface
1263 */
1264 remainder &= ~1UL;
1265 /* make sure remainder is still sane */
f21e18ca 1266 if (remainder < (unsigned)ROUND_UP(super->len, 512) >> 9)
dda5855f 1267 remainder = ROUND_UP(super->len, 512) >> 9;
14e8215b
DW		 1268 if (reservation > remainder)
1269 reservation = remainder;
1270 }
5551b113 1271 e->start = total_blocks(&dl->disk) - reservation;
c2c087e6
DW		 1272 e->size = 0;
1273 return rv;
1274}
1275
14e8215b
DW		 1276/* try to determine how much space is reserved for metadata from
1277 * the last get_extents() entry, otherwise fallback to the
1278 * default
1279 */
1280static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl)
1281{
1282 struct extent *e;
1283 int i;
1284 __u32 rv;
1285
1286 /* for spares just return a minimal reservation which will grow
1287 * once the spare is picked up by an array
1288 */
1289 if (dl->index == -1)
1290 return MPB_SECTOR_CNT;
1291
1292 e = get_extents(super, dl);
1293 if (!e)
1294 return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
1295
1296 /* scroll to last entry */
1297 for (i = 0; e[i].size; i++)
1298 continue;
1299
5551b113 1300 rv = total_blocks(&dl->disk) - e[i].start;
14e8215b
DW		 1301
1302 free(e);
1303
1304 return rv;
1305}
1306
25ed7e59
DW		 1307static int is_spare(struct imsm_disk *disk)
1308{
1309 return (disk->status & SPARE_DISK) == SPARE_DISK;
1310}
1311
1312static int is_configured(struct imsm_disk *disk)
1313{
1314 return (disk->status & CONFIGURED_DISK) == CONFIGURED_DISK;
1315}
1316
1317static int is_failed(struct imsm_disk *disk)
1318{
1319 return (disk->status & FAILED_DISK) == FAILED_DISK;
1320}
1321
2432ce9b
AP		 1322static int is_journal(struct imsm_disk *disk)
1323{
1324 return (disk->status & JOURNAL_DISK) == JOURNAL_DISK;
1325}
1326
b81221b7
CA		 1327/* try to determine how much space is reserved for metadata from
1328 * the last get_extents() entry on the smallest active disk,
1329 * otherwise fallback to the default
1330 */
1331static __u32 imsm_min_reserved_sectors(struct intel_super *super)
1332{
1333 struct extent *e;
1334 int i;
5551b113
CA		 1335 unsigned long long min_active;
1336 __u32 remainder;
b81221b7
CA		 1337 __u32 rv = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
1338 struct dl *dl, *dl_min = NULL;
1339
1340 if (!super)
1341 return rv;
1342
1343 min_active = 0;
1344 for (dl = super->disks; dl; dl = dl->next) {
1345 if (dl->index < 0)
1346 continue;
5551b113
CA		 1347 unsigned long long blocks = total_blocks(&dl->disk);
1348 if (blocks < min_active || min_active == 0) {
b81221b7 1349 dl_min = dl;
5551b113 1350 min_active = blocks;
b81221b7
CA		 1351 }
1352 }
1353 if (!dl_min)
1354 return rv;
1355
1356 /* find last lba used by subarrays on the smallest active disk */
1357 e = get_extents(super, dl_min);
1358 if (!e)
1359 return rv;
1360 for (i = 0; e[i].size; i++)
1361 continue;
1362
1363 remainder = min_active - e[i].start;
1364 free(e);
1365
1366 /* to give priority to recovery we should not require full
1367 IMSM_RESERVED_SECTORS from the spare */
1368 rv = MPB_SECTOR_CNT + NUM_BLOCKS_DIRTY_STRIPE_REGION;
1369
1370 /* if real reservation is smaller use that value */
1371 return (remainder < rv) ? remainder : rv;
1372}
1373
80e7f8c3
AC		 1374/* Return minimum size of a spare that can be used in this array*/
1375static unsigned long long min_acceptable_spare_size_imsm(struct supertype *st)
1376{
1377 struct intel_super *super = st->sb;
1378 struct dl *dl;
1379 struct extent *e;
1380 int i;
1381 unsigned long long rv = 0;
1382
1383 if (!super)
1384 return rv;
1385 /* find first active disk in array */
1386 dl = super->disks;
1387 while (dl && (is_failed(&dl->disk) || dl->index == -1))
1388 dl = dl->next;
1389 if (!dl)
1390 return rv;
1391 /* find last lba used by subarrays */
1392 e = get_extents(super, dl);
1393 if (!e)
1394 return rv;
1395 for (i = 0; e[i].size; i++)
1396 continue;
1397 if (i > 0)
1398 rv = e[i-1].start + e[i-1].size;
1399 free(e);
b81221b7 1400
80e7f8c3 1401 /* add the amount of space needed for metadata */
b81221b7
CA		 1402 rv = rv + imsm_min_reserved_sectors(super);
1403
80e7f8c3
AC		 1404 return rv * 512;
1405}
1406
d1e02575
AK		 1407static int is_gen_migration(struct imsm_dev *dev);
1408
f36a9ecd
PB		 1409#define IMSM_4K_DIV 8
1410
1799c9e8 1411#ifndef MDASSEMBLE
c47b0ff6
AK		 1412static __u64 blocks_per_migr_unit(struct intel_super *super,
1413 struct imsm_dev *dev);
1e5c6983 1414
c47b0ff6
AK		 1415static void print_imsm_dev(struct intel_super *super,
1416 struct imsm_dev *dev,
1417 char *uuid,
1418 int disk_idx)
cdddbdbc
DW		 1419{
1420 __u64 sz;
0d80bb2f 1421 int slot, i;
238c0a71
AK		 1422 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1423 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
b10b37b8 1424 __u32 ord;
cdddbdbc
DW		 1425
1426 printf("\n");
1e7bc0ed 1427 printf("[%.16s]:\n", dev->volume);
44470971 1428 printf(" UUID : %s\n", uuid);
dd8bcb3b
AK		 1429 printf(" RAID Level : %d", get_imsm_raid_level(map));
1430 if (map2)
1431 printf(" <-- %d", get_imsm_raid_level(map2));
1432 printf("\n");
1433 printf(" Members : %d", map->num_members);
1434 if (map2)
1435 printf(" <-- %d", map2->num_members);
1436 printf("\n");
0d80bb2f
DW		 1437 printf(" Slots : [");
1438 for (i = 0; i < map->num_members; i++) {
238c0a71 1439 ord = get_imsm_ord_tbl_ent(dev, i, MAP_0);
0d80bb2f
DW		 1440 printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U");
1441 }
dd8bcb3b
AK		 1442 printf("]");
1443 if (map2) {
1444 printf(" <-- [");
1445 for (i = 0; i < map2->num_members; i++) {
238c0a71 1446 ord = get_imsm_ord_tbl_ent(dev, i, MAP_1);
dd8bcb3b
AK		 1447 printf("%s", ord & IMSM_ORD_REBUILD ? "_" : "U");
1448 }
1449 printf("]");
1450 }
1451 printf("\n");
7095bccb
AK		 1452 printf(" Failed disk : ");
1453 if (map->failed_disk_num == 0xff)
1454 printf("none");
1455 else
1456 printf("%i", map->failed_disk_num);
1457 printf("\n");
620b1713
DW		 1458 slot = get_imsm_disk_slot(map, disk_idx);
1459 if (slot >= 0) {
238c0a71 1460 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
b10b37b8
DW		 1461 printf(" This Slot : %d%s\n", slot,
1462 ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : "");
1463 } else
cdddbdbc
DW		 1464 printf(" This Slot : ?\n");
1465 sz = __le32_to_cpu(dev->size_high);
1466 sz <<= 32;
1467 sz += __le32_to_cpu(dev->size_low);
1468 printf(" Array Size : %llu%s\n", (unsigned long long)sz,
1469 human_size(sz * 512));
5551b113 1470 sz = blocks_per_member(map);
cdddbdbc
DW		 1471 printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz,
1472 human_size(sz * 512));
5551b113
CA		 1473 printf(" Sector Offset : %llu\n",
1474 pba_of_lba0(map));
1475 printf(" Num Stripes : %llu\n",
1476 num_data_stripes(map));
dd8bcb3b 1477 printf(" Chunk Size : %u KiB",
cdddbdbc 1478 __le16_to_cpu(map->blocks_per_strip) / 2);
dd8bcb3b
AK		 1479 if (map2)
1480 printf(" <-- %u KiB",
1481 __le16_to_cpu(map2->blocks_per_strip) / 2);
1482 printf("\n");
cdddbdbc 1483 printf(" Reserved : %d\n", __le32_to_cpu(dev->reserved_blocks));
8655a7b1 1484 printf(" Migrate State : ");
1484e727
DW		 1485 if (dev->vol.migr_state) {
1486 if (migr_type(dev) == MIGR_INIT)
8655a7b1 1487 printf("initialize\n");
1484e727 1488 else if (migr_type(dev) == MIGR_REBUILD)
8655a7b1 1489 printf("rebuild\n");
1484e727 1490 else if (migr_type(dev) == MIGR_VERIFY)
8655a7b1 1491 printf("check\n");
1484e727 1492 else if (migr_type(dev) == MIGR_GEN_MIGR)
8655a7b1 1493 printf("general migration\n");
1484e727 1494 else if (migr_type(dev) == MIGR_STATE_CHANGE)
8655a7b1 1495 printf("state change\n");
1484e727 1496 else if (migr_type(dev) == MIGR_REPAIR)
8655a7b1 1497 printf("repair\n");
1484e727 1498 else
8655a7b1
DW		 1499 printf("<unknown:%d>\n", migr_type(dev));
1500 } else
1501 printf("idle\n");
3393c6af
DW		 1502 printf(" Map State : %s", map_state_str[map->map_state]);
1503 if (dev->vol.migr_state) {
238c0a71 1504 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983 1505
b10b37b8 1506 printf(" <-- %s", map_state_str[map->map_state]);
464d40e8
LD		 1507 printf("\n Checkpoint : %u ",
1508 __le32_to_cpu(dev->vol.curr_migr_unit));
089f9d79 1509 if (is_gen_migration(dev) && (slot > 1 || slot < 0))
464d40e8
LD		 1510 printf("(N/A)");
1511 else
1512 printf("(%llu)", (unsigned long long)
1513 blocks_per_migr_unit(super, dev));
3393c6af
DW		 1514 }
1515 printf("\n");
2432ce9b
AP		 1516 printf(" Dirty State : %s\n", (dev->vol.dirty & RAIDVOL_DIRTY) ?
1517 "dirty" : "clean");
1518 printf(" RWH Policy : ");
1519 if (dev->rwh_policy == RWH_OFF)
1520 printf("off\n");
1521 else if (dev->rwh_policy == RWH_DISTRIBUTED)
1522 printf("PPL distributed\n");
1523 else if (dev->rwh_policy == RWH_JOURNALING_DRIVE)
1524 printf("PPL journaling drive\n");
1525 else
1526 printf("<unknown:%d>\n", dev->rwh_policy);
cdddbdbc
DW		 1527}
1528
ef5c214e
MK		 1529static void print_imsm_disk(struct imsm_disk *disk,
1530 int index,
1531 __u32 reserved,
1532 unsigned int sector_size) {
1f24f035 1533 char str[MAX_RAID_SERIAL_LEN + 1];
cdddbdbc
DW		 1534 __u64 sz;
1535
0ec1f4e8 1536 if (index < -1 || !disk)
e9d82038
DW		 1537 return;
1538
cdddbdbc 1539 printf("\n");
1f24f035 1540 snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial);
0ec1f4e8
DW		 1541 if (index >= 0)
1542 printf(" Disk%02d Serial : %s\n", index, str);
1543 else
1544 printf(" Disk Serial : %s\n", str);
2432ce9b
AP		 1545 printf(" State :%s%s%s%s\n", is_spare(disk) ? " spare" : "",
1546 is_configured(disk) ? " active" : "",
1547 is_failed(disk) ? " failed" : "",
1548 is_journal(disk) ? " journal" : "");
cdddbdbc 1549 printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id));
5551b113 1550 sz = total_blocks(disk) - reserved;
ef5c214e
MK		 1551 printf(" Usable Size : %llu%s\n",
1552 (unsigned long long)sz * 512 / sector_size,
cdddbdbc
DW		 1553 human_size(sz * 512));
1554}
1555
de44e46f
PB		 1556void convert_to_4k_imsm_migr_rec(struct intel_super *super)
1557{
1558 struct migr_record *migr_rec = super->migr_rec;
1559
1560 migr_rec->blocks_per_unit /= IMSM_4K_DIV;
1561 migr_rec->ckpt_area_pba /= IMSM_4K_DIV;
1562 migr_rec->dest_1st_member_lba /= IMSM_4K_DIV;
1563 migr_rec->dest_depth_per_unit /= IMSM_4K_DIV;
1564 split_ull((join_u32(migr_rec->post_migr_vol_cap,
1565 migr_rec->post_migr_vol_cap_hi) / IMSM_4K_DIV),
1566 &migr_rec->post_migr_vol_cap, &migr_rec->post_migr_vol_cap_hi);
1567}
1568
f36a9ecd
PB		 1569void convert_to_4k_imsm_disk(struct imsm_disk *disk)
1570{
1571 set_total_blocks(disk, (total_blocks(disk)/IMSM_4K_DIV));
1572}
1573
1574void convert_to_4k(struct intel_super *super)
1575{
1576 struct imsm_super *mpb = super->anchor;
1577 struct imsm_disk *disk;
1578 int i;
e4467bc7 1579 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
f36a9ecd
PB		 1580
1581 for (i = 0; i < mpb->num_disks ; i++) {
1582 disk = __get_imsm_disk(mpb, i);
1583 /* disk */
1584 convert_to_4k_imsm_disk(disk);
1585 }
1586 for (i = 0; i < mpb->num_raid_devs; i++) {
1587 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
1588 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1589 /* dev */
1590 split_ull((join_u32(dev->size_low, dev->size_high)/IMSM_4K_DIV),
1591 &dev->size_low, &dev->size_high);
1592 dev->vol.curr_migr_unit /= IMSM_4K_DIV;
1593
1594 /* map0 */
1595 set_blocks_per_member(map, blocks_per_member(map)/IMSM_4K_DIV);
1596 map->blocks_per_strip /= IMSM_4K_DIV;
1597 set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV);
1598
1599 if (dev->vol.migr_state) {
1600 /* map1 */
1601 map = get_imsm_map(dev, MAP_1);
1602 set_blocks_per_member(map,
1603 blocks_per_member(map)/IMSM_4K_DIV);
1604 map->blocks_per_strip /= IMSM_4K_DIV;
1605 set_pba_of_lba0(map, pba_of_lba0(map)/IMSM_4K_DIV);
1606 }
1607 }
e4467bc7
TM		 1608 if (bbm_log_size) {
1609 struct bbm_log *log = (void *)mpb +
1610 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
1611 __u32 i;
1612
1613 for (i = 0; i < log->entry_count; i++) {
1614 struct bbm_log_entry *entry =
1615 &log->marked_block_entries[i];
1616
1617 __u8 count = entry->marked_count + 1;
1618 unsigned long long sector =
1619 __le48_to_cpu(&entry->defective_block_start);
1620
1621 entry->defective_block_start =
1622 __cpu_to_le48(sector/IMSM_4K_DIV);
1623 entry->marked_count = max(count/IMSM_4K_DIV, 1) - 1;
1624 }
1625 }
f36a9ecd
PB		 1626
1627 mpb->check_sum = __gen_imsm_checksum(mpb);
1628}
1629
520e69e2
AK		 1630void examine_migr_rec_imsm(struct intel_super *super)
1631{
1632 struct migr_record *migr_rec = super->migr_rec;
1633 struct imsm_super *mpb = super->anchor;
1634 int i;
1635
1636 for (i = 0; i < mpb->num_raid_devs; i++) {
1637 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
3136abe5 1638 struct imsm_map *map;
b4ab44d8 1639 int slot = -1;
3136abe5 1640
520e69e2
AK		 1641 if (is_gen_migration(dev) == 0)
1642 continue;
1643
1644 printf("\nMigration Record Information:");
3136abe5 1645
44bfe6df
AK		 1646 /* first map under migration */
1647 map = get_imsm_map(dev, MAP_0);
3136abe5
AK		 1648 if (map)
1649 slot = get_imsm_disk_slot(map, super->disks->index);
089f9d79 1650 if (map == NULL || slot > 1 || slot < 0) {
520e69e2
AK		 1651 printf(" Empty\n ");
1652 printf("Examine one of first two disks in array\n");
1653 break;
1654 }
1655 printf("\n Status : ");
1656 if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL)
1657 printf("Normal\n");
1658 else
1659 printf("Contains Data\n");
1660 printf(" Current Unit : %u\n",
1661 __le32_to_cpu(migr_rec->curr_migr_unit));
1662 printf(" Family : %u\n",
1663 __le32_to_cpu(migr_rec->family_num));
1664 printf(" Ascending : %u\n",
1665 __le32_to_cpu(migr_rec->ascending_migr));
1666 printf(" Blocks Per Unit : %u\n",
1667 __le32_to_cpu(migr_rec->blocks_per_unit));
1668 printf(" Dest. Depth Per Unit : %u\n",
1669 __le32_to_cpu(migr_rec->dest_depth_per_unit));
1670 printf(" Checkpoint Area pba : %u\n",
1671 __le32_to_cpu(migr_rec->ckpt_area_pba));
1672 printf(" First member lba : %u\n",
1673 __le32_to_cpu(migr_rec->dest_1st_member_lba));
1674 printf(" Total Number of Units : %u\n",
1675 __le32_to_cpu(migr_rec->num_migr_units));
1676 printf(" Size of volume : %u\n",
1677 __le32_to_cpu(migr_rec->post_migr_vol_cap));
1678 printf(" Expansion space for LBA64 : %u\n",
1679 __le32_to_cpu(migr_rec->post_migr_vol_cap_hi));
1680 printf(" Record was read from : %u\n",
1681 __le32_to_cpu(migr_rec->ckpt_read_disk_num));
1682
1683 break;
1684 }
1685}
9e2d750d 1686#endif /* MDASSEMBLE */
f36a9ecd 1687
de44e46f
PB		 1688void convert_from_4k_imsm_migr_rec(struct intel_super *super)
1689{
1690 struct migr_record *migr_rec = super->migr_rec;
1691
1692 migr_rec->blocks_per_unit *= IMSM_4K_DIV;
1693 migr_rec->ckpt_area_pba *= IMSM_4K_DIV;
1694 migr_rec->dest_1st_member_lba *= IMSM_4K_DIV;
1695 migr_rec->dest_depth_per_unit *= IMSM_4K_DIV;
1696 split_ull((join_u32(migr_rec->post_migr_vol_cap,
1697 migr_rec->post_migr_vol_cap_hi) * IMSM_4K_DIV),
1698 &migr_rec->post_migr_vol_cap,
1699 &migr_rec->post_migr_vol_cap_hi);
1700}
1701
f36a9ecd
PB		 1702void convert_from_4k(struct intel_super *super)
1703{
1704 struct imsm_super *mpb = super->anchor;
1705 struct imsm_disk *disk;
1706 int i;
e4467bc7 1707 __u32 bbm_log_size = __le32_to_cpu(mpb->bbm_log_size);
f36a9ecd
PB		 1708
1709 for (i = 0; i < mpb->num_disks ; i++) {
1710 disk = __get_imsm_disk(mpb, i);
1711 /* disk */
1712 set_total_blocks(disk, (total_blocks(disk)*IMSM_4K_DIV));
1713 }
1714
1715 for (i = 0; i < mpb->num_raid_devs; i++) {
1716 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
1717 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1718 /* dev */
1719 split_ull((join_u32(dev->size_low, dev->size_high)*IMSM_4K_DIV),
1720 &dev->size_low, &dev->size_high);
1721 dev->vol.curr_migr_unit *= IMSM_4K_DIV;
1722
1723 /* map0 */
1724 set_blocks_per_member(map, blocks_per_member(map)*IMSM_4K_DIV);
1725 map->blocks_per_strip *= IMSM_4K_DIV;
1726 set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV);
1727
1728 if (dev->vol.migr_state) {
1729 /* map1 */
1730 map = get_imsm_map(dev, MAP_1);
1731 set_blocks_per_member(map,
1732 blocks_per_member(map)*IMSM_4K_DIV);
1733 map->blocks_per_strip *= IMSM_4K_DIV;
1734 set_pba_of_lba0(map, pba_of_lba0(map)*IMSM_4K_DIV);
1735 }
1736 }
e4467bc7
TM		 1737 if (bbm_log_size) {
1738 struct bbm_log *log = (void *)mpb +
1739 __le32_to_cpu(mpb->mpb_size) - bbm_log_size;
1740 __u32 i;
1741
1742 for (i = 0; i < log->entry_count; i++) {
1743 struct bbm_log_entry *entry =
1744 &log->marked_block_entries[i];
1745
1746 __u8 count = entry->marked_count + 1;
1747 unsigned long long sector =
1748 __le48_to_cpu(&entry->defective_block_start);
1749
1750 entry->defective_block_start =
1751 __cpu_to_le48(sector*IMSM_4K_DIV);
1752 entry->marked_count = count*IMSM_4K_DIV - 1;
1753 }
1754 }
f36a9ecd
PB		 1755
1756 mpb->check_sum = __gen_imsm_checksum(mpb);
1757}
1758
19482bcc
AK		 1759/*******************************************************************************
1760 * function: imsm_check_attributes
1761 * Description: Function checks if features represented by attributes flags
1011e834 1762 * are supported by mdadm.
19482bcc
AK		 1763 * Parameters:
1764 * attributes - Attributes read from metadata
1765 * Returns:
1011e834
N		 1766 * 0 - passed attributes contains unsupported features flags
1767 * 1 - all features are supported
19482bcc
AK		 1768 ******************************************************************************/
1769static int imsm_check_attributes(__u32 attributes)
1770{
1771 int ret_val = 1;
418f9b36
N		 1772 __u32 not_supported = MPB_ATTRIB_SUPPORTED^0xffffffff;
1773
1774 not_supported &= ~MPB_ATTRIB_IGNORED;
19482bcc
AK		 1775
1776 not_supported &= attributes;
1777 if (not_supported) {
e7b84f9d 1778 pr_err("(IMSM): Unsupported attributes : %x\n",
418f9b36 1779 (unsigned)__le32_to_cpu(not_supported));
19482bcc
AK		 1780 if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) {
1781 dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY \n");
1782 not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY;
1783 }
1784 if (not_supported & MPB_ATTRIB_2TB) {
1785 dprintf("\t\tMPB_ATTRIB_2TB\n");
1786 not_supported ^= MPB_ATTRIB_2TB;
1787 }
1788 if (not_supported & MPB_ATTRIB_RAID0) {
1789 dprintf("\t\tMPB_ATTRIB_RAID0\n");
1790 not_supported ^= MPB_ATTRIB_RAID0;
1791 }
1792 if (not_supported & MPB_ATTRIB_RAID1) {
1793 dprintf("\t\tMPB_ATTRIB_RAID1\n");
1794 not_supported ^= MPB_ATTRIB_RAID1;
1795 }
1796 if (not_supported & MPB_ATTRIB_RAID10) {
1797 dprintf("\t\tMPB_ATTRIB_RAID10\n");
1798 not_supported ^= MPB_ATTRIB_RAID10;
1799 }
1800 if (not_supported & MPB_ATTRIB_RAID1E) {
1801 dprintf("\t\tMPB_ATTRIB_RAID1E\n");
1802 not_supported ^= MPB_ATTRIB_RAID1E;
1803 }
1804 if (not_supported & MPB_ATTRIB_RAID5) {
1805 dprintf("\t\tMPB_ATTRIB_RAID5\n");
1806 not_supported ^= MPB_ATTRIB_RAID5;
1807 }
1808 if (not_supported & MPB_ATTRIB_RAIDCNG) {
1809 dprintf("\t\tMPB_ATTRIB_RAIDCNG\n");
1810 not_supported ^= MPB_ATTRIB_RAIDCNG;
1811 }
1812 if (not_supported & MPB_ATTRIB_BBM) {
1813 dprintf("\t\tMPB_ATTRIB_BBM\n");
1814 not_supported ^= MPB_ATTRIB_BBM;
1815 }
1816 if (not_supported & MPB_ATTRIB_CHECKSUM_VERIFY) {
1817 dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY (== MPB_ATTRIB_LEGACY)\n");
1818 not_supported ^= MPB_ATTRIB_CHECKSUM_VERIFY;
1819 }
1820 if (not_supported & MPB_ATTRIB_EXP_STRIPE_SIZE) {
1821 dprintf("\t\tMPB_ATTRIB_EXP_STRIP_SIZE\n");
1822 not_supported ^= MPB_ATTRIB_EXP_STRIPE_SIZE;
1823 }
1824 if (not_supported & MPB_ATTRIB_2TB_DISK) {
1825 dprintf("\t\tMPB_ATTRIB_2TB_DISK\n");
1826 not_supported ^= MPB_ATTRIB_2TB_DISK;
1827 }
1828 if (not_supported & MPB_ATTRIB_NEVER_USE2) {
1829 dprintf("\t\tMPB_ATTRIB_NEVER_USE2\n");
1830 not_supported ^= MPB_ATTRIB_NEVER_USE2;
1831 }
1832 if (not_supported & MPB_ATTRIB_NEVER_USE) {
1833 dprintf("\t\tMPB_ATTRIB_NEVER_USE\n");
1834 not_supported ^= MPB_ATTRIB_NEVER_USE;
1835 }
1836
1837 if (not_supported)
1ade5cc1 1838 dprintf("(IMSM): Unknown attributes : %x\n", not_supported);
19482bcc
AK		 1839
1840 ret_val = 0;
1841 }
1842
1843 return ret_val;
1844}
1845
9e2d750d 1846#ifndef MDASSEMBLE
a5d85af7 1847static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map);
44470971 1848
cdddbdbc
DW		 1849static void examine_super_imsm(struct supertype *st, char *homehost)
1850{
1851 struct intel_super *super = st->sb;
949c47a0 1852 struct imsm_super *mpb = super->anchor;
cdddbdbc
DW		 1853 char str[MAX_SIGNATURE_LENGTH];
1854 int i;
27fd6274
DW		 1855 struct mdinfo info;
1856 char nbuf[64];
cdddbdbc 1857 __u32 sum;
14e8215b 1858 __u32 reserved = imsm_reserved_sectors(super, super->disks);
94827db3 1859 struct dl *dl;
27fd6274 1860
618f4e6d
XN		 1861 strncpy(str, (char *)mpb->sig, MPB_SIG_LEN);
1862 str[MPB_SIG_LEN-1] = '\0';
cdddbdbc
DW		 1863 printf(" Magic : %s\n", str);
1864 snprintf(str, strlen(MPB_VERSION_RAID0), "%s", get_imsm_version(mpb));
1865 printf(" Version : %s\n", get_imsm_version(mpb));
148acb7b 1866 printf(" Orig Family : %08x\n", __le32_to_cpu(mpb->orig_family_num));
cdddbdbc
DW		 1867 printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num));
1868 printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num));
19482bcc
AK		 1869 printf(" Attributes : ");
1870 if (imsm_check_attributes(mpb->attributes))
1871 printf("All supported\n");
1872 else
1873 printf("not supported\n");
a5d85af7 1874 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 1875 fname_from_uuid(st, &info, nbuf, ':');
27fd6274 1876 printf(" UUID : %s\n", nbuf + 5);
cdddbdbc
DW		 1877 sum = __le32_to_cpu(mpb->check_sum);
1878 printf(" Checksum : %08x %s\n", sum,
949c47a0 1879 __gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect");
f36a9ecd 1880 printf(" MPB Sectors : %d\n", mpb_sectors(mpb, super->sector_size));
cdddbdbc
DW		 1881 printf(" Disks : %d\n", mpb->num_disks);
1882 printf(" RAID Devices : %d\n", mpb->num_raid_devs);
ef5c214e
MK		 1883 print_imsm_disk(__get_imsm_disk(mpb, super->disks->index),
1884 super->disks->index, reserved, super->sector_size);
8d67477f 1885 if (get_imsm_bbm_log_size(super->bbm_log)) {
604b746f
JD		 1886 struct bbm_log *log = super->bbm_log;
1887
1888 printf("\n");
1889 printf("Bad Block Management Log:\n");
1890 printf(" Log Size : %d\n", __le32_to_cpu(mpb->bbm_log_size));
1891 printf(" Signature : %x\n", __le32_to_cpu(log->signature));
1892 printf(" Entry Count : %d\n", __le32_to_cpu(log->entry_count));
604b746f 1893 }
44470971
DW		 1894 for (i = 0; i < mpb->num_raid_devs; i++) {
1895 struct mdinfo info;
1896 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
1897
1898 super->current_vol = i;
a5d85af7 1899 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 1900 fname_from_uuid(st, &info, nbuf, ':');
c47b0ff6 1901 print_imsm_dev(super, dev, nbuf + 5, super->disks->index);
44470971 1902 }
cdddbdbc
DW		 1903 for (i = 0; i < mpb->num_disks; i++) {
1904 if (i == super->disks->index)
1905 continue;
ef5c214e
MK		 1906 print_imsm_disk(__get_imsm_disk(mpb, i), i, reserved,
1907 super->sector_size);
cdddbdbc 1908 }
94827db3 1909
0ec1f4e8
DW		 1910 for (dl = super->disks; dl; dl = dl->next)
1911 if (dl->index == -1)
ef5c214e
MK		 1912 print_imsm_disk(&dl->disk, -1, reserved,
1913 super->sector_size);
520e69e2
AK		 1914
1915 examine_migr_rec_imsm(super);
cdddbdbc
DW		 1916}
1917
061f2c6a 1918static void brief_examine_super_imsm(struct supertype *st, int verbose)
cdddbdbc 1919{
27fd6274 1920 /* We just write a generic IMSM ARRAY entry */
ff54de6e
N		 1921 struct mdinfo info;
1922 char nbuf[64];
1e7bc0ed 1923 struct intel_super *super = st->sb;
1e7bc0ed 1924
0d5a423f
DW		 1925 if (!super->anchor->num_raid_devs) {
1926 printf("ARRAY metadata=imsm\n");
1e7bc0ed 1927 return;
0d5a423f 1928 }
ff54de6e 1929
a5d85af7 1930 getinfo_super_imsm(st, &info, NULL);
4737ae25
N		 1931 fname_from_uuid(st, &info, nbuf, ':');
1932 printf("ARRAY metadata=imsm UUID=%s\n", nbuf + 5);
1933}
1934
1935static void brief_examine_subarrays_imsm(struct supertype *st, int verbose)
1936{
1937 /* We just write a generic IMSM ARRAY entry */
1938 struct mdinfo info;
1939 char nbuf[64];
1940 char nbuf1[64];
1941 struct intel_super *super = st->sb;
1942 int i;
1943
1944 if (!super->anchor->num_raid_devs)
1945 return;
1946
a5d85af7 1947 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 1948 fname_from_uuid(st, &info, nbuf, ':');
1e7bc0ed
DW		 1949 for (i = 0; i < super->anchor->num_raid_devs; i++) {
1950 struct imsm_dev *dev = get_imsm_dev(super, i);
1951
1952 super->current_vol = i;
a5d85af7 1953 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 1954 fname_from_uuid(st, &info, nbuf1, ':');
1124b3cf 1955 printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n",
cf8de691 1956 dev->volume, nbuf + 5, i, nbuf1 + 5);
1e7bc0ed 1957 }
cdddbdbc
DW		 1958}
1959
9d84c8ea
DW		 1960static void export_examine_super_imsm(struct supertype *st)
1961{
1962 struct intel_super *super = st->sb;
1963 struct imsm_super *mpb = super->anchor;
1964 struct mdinfo info;
1965 char nbuf[64];
1966
a5d85af7 1967 getinfo_super_imsm(st, &info, NULL);
9d84c8ea
DW		 1968 fname_from_uuid(st, &info, nbuf, ':');
1969 printf("MD_METADATA=imsm\n");
1970 printf("MD_LEVEL=container\n");
1971 printf("MD_UUID=%s\n", nbuf+5);
1972 printf("MD_DEVICES=%u\n", mpb->num_disks);
1973}
1974
74db60b0
N		 1975static int copy_metadata_imsm(struct supertype *st, int from, int to)
1976{
f36a9ecd 1977 /* The second last sector of the device contains
74db60b0
N		 1978 * the "struct imsm_super" metadata.
1979 * This contains mpb_size which is the size in bytes of the
1980 * extended metadata. This is located immediately before
1981 * the imsm_super.
1982 * We want to read all that, plus the last sector which
1983 * may contain a migration record, and write it all
1984 * to the target.
1985 */
1986 void *buf;
1987 unsigned long long dsize, offset;
1988 int sectors;
1989 struct imsm_super *sb;
f36a9ecd
PB		 1990 struct intel_super *super = st->sb;
1991 unsigned int sector_size = super->sector_size;
1992 unsigned int written = 0;
74db60b0 1993
de44e46f 1994 if (posix_memalign(&buf, MAX_SECTOR_SIZE, MAX_SECTOR_SIZE) != 0)
74db60b0
N		 1995 return 1;
1996
1997 if (!get_dev_size(from, NULL, &dsize))
1998 goto err;
1999
f36a9ecd 2000 if (lseek64(from, dsize-(2*sector_size), 0) < 0)
74db60b0 2001 goto err;
466070ad 2002 if ((unsigned int)read(from, buf, sector_size) != sector_size)
74db60b0
N		 2003 goto err;
2004 sb = buf;
2005 if (strncmp((char*)sb->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0)
2006 goto err;
2007
f36a9ecd
PB		 2008 sectors = mpb_sectors(sb, sector_size) + 2;
2009 offset = dsize - sectors * sector_size;
74db60b0
N		 2010 if (lseek64(from, offset, 0) < 0 ||
2011 lseek64(to, offset, 0) < 0)
2012 goto err;
f36a9ecd
PB		 2013 while (written < sectors * sector_size) {
2014 int n = sectors*sector_size - written;
74db60b0
N		 2015 if (n > 4096)
2016 n = 4096;
2017 if (read(from, buf, n) != n)
2018 goto err;
2019 if (write(to, buf, n) != n)
2020 goto err;
2021 written += n;
2022 }
2023 free(buf);
2024 return 0;
2025err:
2026 free(buf);
2027 return 1;
2028}
2029
cdddbdbc
DW		 2030static void detail_super_imsm(struct supertype *st, char *homehost)
2031{
3ebe00a1
DW		 2032 struct mdinfo info;
2033 char nbuf[64];
2034
a5d85af7 2035 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2036 fname_from_uuid(st, &info, nbuf, ':');
65884368 2037 printf("\n UUID : %s\n", nbuf + 5);
cdddbdbc
DW		 2038}
2039
2040static void brief_detail_super_imsm(struct supertype *st)
2041{
ff54de6e
N		 2042 struct mdinfo info;
2043 char nbuf[64];
a5d85af7 2044 getinfo_super_imsm(st, &info, NULL);
ae2bfd4e 2045 fname_from_uuid(st, &info, nbuf, ':');
ff54de6e 2046 printf(" UUID=%s", nbuf + 5);
cdddbdbc 2047}
d665cc31
DW		 2048
2049static int imsm_read_serial(int fd, char *devname, __u8 *serial);
2050static void fd2devname(int fd, char *name);
2051
120dc887 2052static int ahci_enumerate_ports(const char *hba_path, int port_count, int host_base, int verbose)
d665cc31 2053{
120dc887
LM		 2054 /* dump an unsorted list of devices attached to AHCI Intel storage
2055 * controller, as well as non-connected ports
d665cc31
DW		 2056 */
2057 int hba_len = strlen(hba_path) + 1;
2058 struct dirent *ent;
2059 DIR *dir;
2060 char *path = NULL;
2061 int err = 0;
2062 unsigned long port_mask = (1 << port_count) - 1;
2063
f21e18ca 2064 if (port_count > (int)sizeof(port_mask) * 8) {
ba728be7 2065 if (verbose > 0)
e7b84f9d 2066 pr_err("port_count %d out of range\n", port_count);
d665cc31
DW		 2067 return 2;
2068 }
2069
2070 /* scroll through /sys/dev/block looking for devices attached to
2071 * this hba
2072 */
2073 dir = opendir("/sys/dev/block");
1a6dd6b9
PB		 2074 if (!dir)
2075 return 1;
2076
2077 for (ent = readdir(dir); ent; ent = readdir(dir)) {
d665cc31
DW		 2078 int fd;
2079 char model[64];
2080 char vendor[64];
2081 char buf[1024];
2082 int major, minor;
2083 char *device;
2084 char *c;
2085 int port;
2086 int type;
2087
2088 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
2089 continue;
2090 path = devt_to_devpath(makedev(major, minor));
2091 if (!path)
2092 continue;
2093 if (!path_attached_to_hba(path, hba_path)) {
2094 free(path);
2095 path = NULL;
2096 continue;
2097 }
2098
2099 /* retrieve the scsi device type */
2100 if (asprintf(&device, "/sys/dev/block/%d:%d/device/xxxxxxx", major, minor) < 0) {
ba728be7 2101 if (verbose > 0)
e7b84f9d 2102 pr_err("failed to allocate 'device'\n");
d665cc31
DW		 2103 err = 2;
2104 break;
2105 }
2106 sprintf(device, "/sys/dev/block/%d:%d/device/type", major, minor);
193b6c0b 2107 if (load_sys(device, buf, sizeof(buf)) != 0) {
ba728be7 2108 if (verbose > 0)
e7b84f9d 2109 pr_err("failed to read device type for %s\n",
d665cc31
DW		 2110 path);
2111 err = 2;
2112 free(device);
2113 break;
2114 }
2115 type = strtoul(buf, NULL, 10);
2116
2117 /* if it's not a disk print the vendor and model */
2118 if (!(type == 0 || type == 7 || type == 14)) {
2119 vendor[0] = '\0';
2120 model[0] = '\0';
2121 sprintf(device, "/sys/dev/block/%d:%d/device/vendor", major, minor);
193b6c0b 2122 if (load_sys(device, buf, sizeof(buf)) == 0) {
d665cc31
DW		 2123 strncpy(vendor, buf, sizeof(vendor));
2124 vendor[sizeof(vendor) - 1] = '\0';
2125 c = (char *) &vendor[sizeof(vendor) - 1];
2126 while (isspace(*c) || *c == '\0')
2127 *c-- = '\0';
2128
2129 }
2130 sprintf(device, "/sys/dev/block/%d:%d/device/model", major, minor);
193b6c0b 2131 if (load_sys(device, buf, sizeof(buf)) == 0) {
d665cc31
DW		 2132 strncpy(model, buf, sizeof(model));
2133 model[sizeof(model) - 1] = '\0';
2134 c = (char *) &model[sizeof(model) - 1];
2135 while (isspace(*c) || *c == '\0')
2136 *c-- = '\0';
2137 }
2138
2139 if (vendor[0] && model[0])
2140 sprintf(buf, "%.64s %.64s", vendor, model);
2141 else
2142 switch (type) { /* numbers from hald/linux/device.c */
2143 case 1: sprintf(buf, "tape"); break;
2144 case 2: sprintf(buf, "printer"); break;
2145 case 3: sprintf(buf, "processor"); break;
2146 case 4:
2147 case 5: sprintf(buf, "cdrom"); break;
2148 case 6: sprintf(buf, "scanner"); break;
2149 case 8: sprintf(buf, "media_changer"); break;
2150 case 9: sprintf(buf, "comm"); break;
2151 case 12: sprintf(buf, "raid"); break;
2152 default: sprintf(buf, "unknown");
2153 }
2154 } else
2155 buf[0] = '\0';
2156 free(device);
2157
2158 /* chop device path to 'host%d' and calculate the port number */
2159 c = strchr(&path[hba_len], '/');
4e5e717d 2160 if (!c) {
ba728be7 2161 if (verbose > 0)
e7b84f9d 2162 pr_err("%s - invalid path name\n", path + hba_len);
4e5e717d
AW		 2163 err = 2;
2164 break;
2165 }
d665cc31 2166 *c = '\0';
0858eccf
AP		 2167 if ((sscanf(&path[hba_len], "ata%d", &port) == 1) ||
2168 ((sscanf(&path[hba_len], "host%d", &port) == 1)))
d665cc31
DW		 2169 port -= host_base;
2170 else {
ba728be7 2171 if (verbose > 0) {
d665cc31 2172 *c = '/'; /* repair the full string */
e7b84f9d 2173 pr_err("failed to determine port number for %s\n",
d665cc31
DW		 2174 path);
2175 }
2176 err = 2;
2177 break;
2178 }
2179
2180 /* mark this port as used */
2181 port_mask &= ~(1 << port);
2182
2183 /* print out the device information */
2184 if (buf[0]) {
2185 printf(" Port%d : - non-disk device (%s) -\n", port, buf);
2186 continue;
2187 }
2188
2189 fd = dev_open(ent->d_name, O_RDONLY);
2190 if (fd < 0)
2191 printf(" Port%d : - disk info unavailable -\n", port);
2192 else {
2193 fd2devname(fd, buf);
2194 printf(" Port%d : %s", port, buf);
2195 if (imsm_read_serial(fd, NULL, (__u8 *) buf) == 0)
664d5325 2196 printf(" (%.*s)\n", MAX_RAID_SERIAL_LEN, buf);
d665cc31 2197 else
664d5325 2198 printf(" ()\n");
4dab422a 2199 close(fd);
d665cc31 2200 }
d665cc31
DW		 2201 free(path);
2202 path = NULL;
2203 }
2204 if (path)
2205 free(path);
2206 if (dir)
2207 closedir(dir);
2208 if (err == 0) {
2209 int i;
2210
2211 for (i = 0; i < port_count; i++)
2212 if (port_mask & (1 << i))
2213 printf(" Port%d : - no device attached -\n", i);
2214 }
2215
2216 return err;
2217}
2218
b5eece69 2219static int print_vmd_attached_devs(struct sys_dev *hba)
60f0f54d
PB		 2220{
2221 struct dirent *ent;
2222 DIR *dir;
2223 char path[292];
2224 char link[256];
2225 char *c, *rp;
2226
2227 if (hba->type != SYS_DEV_VMD)
b5eece69 2228 return 1;
60f0f54d
PB		 2229
2230 /* scroll through /sys/dev/block looking for devices attached to
2231 * this hba
2232 */
2233 dir = opendir("/sys/bus/pci/drivers/nvme");
b9135011 2234 if (!dir)
b5eece69 2235 return 1;
b9135011
JS		 2236
2237 for (ent = readdir(dir); ent; ent = readdir(dir)) {
60f0f54d
PB		 2238 int n;
2239
2240 /* is 'ent' a device? check that the 'subsystem' link exists and
2241 * that its target matches 'bus'
2242 */
2243 sprintf(path, "/sys/bus/pci/drivers/nvme/%s/subsystem",
2244 ent->d_name);
2245 n = readlink(path, link, sizeof(link));
2246 if (n < 0 || n >= (int)sizeof(link))
2247 continue;
2248 link[n] = '\0';
2249 c = strrchr(link, '/');
2250 if (!c)
2251 continue;
2252 if (strncmp("pci", c+1, strlen("pci")) != 0)
2253 continue;
2254
2255 sprintf(path, "/sys/bus/pci/drivers/nvme/%s", ent->d_name);
60f0f54d
PB		 2256
2257 rp = realpath(path, NULL);
2258 if (!rp)
2259 continue;
2260
2261 if (path_attached_to_hba(rp, hba->path)) {
2262 printf(" NVMe under VMD : %s\n", rp);
2263 }
2264 free(rp);
2265 }
2266
b9135011 2267 closedir(dir);
b5eece69 2268 return 0;
60f0f54d
PB		 2269}
2270
120dc887
LM		 2271static void print_found_intel_controllers(struct sys_dev *elem)
2272{
2273 for (; elem; elem = elem->next) {
e7b84f9d 2274 pr_err("found Intel(R) ");
120dc887
LM		 2275 if (elem->type == SYS_DEV_SATA)
2276 fprintf(stderr, "SATA ");
155cbb4c
LM		 2277 else if (elem->type == SYS_DEV_SAS)
2278 fprintf(stderr, "SAS ");
0858eccf
AP		 2279 else if (elem->type == SYS_DEV_NVME)
2280 fprintf(stderr, "NVMe ");
60f0f54d
PB		 2281
2282 if (elem->type == SYS_DEV_VMD)
2283 fprintf(stderr, "VMD domain");
2284 else
2285 fprintf(stderr, "RAID controller");
2286
120dc887
LM		 2287 if (elem->pci_id)
2288 fprintf(stderr, " at %s", elem->pci_id);
2289 fprintf(stderr, ".\n");
2290 }
2291 fflush(stderr);
2292}
2293
120dc887
LM		 2294static int ahci_get_port_count(const char *hba_path, int *port_count)
2295{
2296 struct dirent *ent;
2297 DIR *dir;
2298 int host_base = -1;
2299
2300 *port_count = 0;
2301 if ((dir = opendir(hba_path)) == NULL)
2302 return -1;
2303
2304 for (ent = readdir(dir); ent; ent = readdir(dir)) {
2305 int host;
2306
0858eccf
AP		 2307 if ((sscanf(ent->d_name, "ata%d", &host) != 1) &&
2308 ((sscanf(ent->d_name, "host%d", &host) != 1)))
120dc887
LM		 2309 continue;
2310 if (*port_count == 0)
2311 host_base = host;
2312 else if (host < host_base)
2313 host_base = host;
2314
2315 if (host + 1 > *port_count + host_base)
2316 *port_count = host + 1 - host_base;
2317 }
2318 closedir(dir);
2319 return host_base;
2320}
2321
a891a3c2
LM		 2322static void print_imsm_capability(const struct imsm_orom *orom)
2323{
0858eccf
AP		 2324 printf(" Platform : Intel(R) ");
2325 if (orom->capabilities == 0 && orom->driver_features == 0)
2326 printf("Matrix Storage Manager\n");
2327 else
2328 printf("Rapid Storage Technology%s\n",
2329 imsm_orom_is_enterprise(orom) ? " enterprise" : "");
2330 if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build)
2331 printf(" Version : %d.%d.%d.%d\n", orom->major_ver,
2332 orom->minor_ver, orom->hotfix_ver, orom->build);
a891a3c2
LM		 2333 printf(" RAID Levels :%s%s%s%s%s\n",
2334 imsm_orom_has_raid0(orom) ? " raid0" : "",
2335 imsm_orom_has_raid1(orom) ? " raid1" : "",
2336 imsm_orom_has_raid1e(orom) ? " raid1e" : "",
2337 imsm_orom_has_raid10(orom) ? " raid10" : "",
2338 imsm_orom_has_raid5(orom) ? " raid5" : "");
2339 printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
2340 imsm_orom_has_chunk(orom, 2) ? " 2k" : "",
2341 imsm_orom_has_chunk(orom, 4) ? " 4k" : "",
2342 imsm_orom_has_chunk(orom, 8) ? " 8k" : "",
2343 imsm_orom_has_chunk(orom, 16) ? " 16k" : "",
2344 imsm_orom_has_chunk(orom, 32) ? " 32k" : "",
2345 imsm_orom_has_chunk(orom, 64) ? " 64k" : "",
2346 imsm_orom_has_chunk(orom, 128) ? " 128k" : "",
2347 imsm_orom_has_chunk(orom, 256) ? " 256k" : "",
2348 imsm_orom_has_chunk(orom, 512) ? " 512k" : "",
2349 imsm_orom_has_chunk(orom, 1024*1) ? " 1M" : "",
2350 imsm_orom_has_chunk(orom, 1024*2) ? " 2M" : "",
2351 imsm_orom_has_chunk(orom, 1024*4) ? " 4M" : "",
2352 imsm_orom_has_chunk(orom, 1024*8) ? " 8M" : "",
2353 imsm_orom_has_chunk(orom, 1024*16) ? " 16M" : "",
2354 imsm_orom_has_chunk(orom, 1024*32) ? " 32M" : "",
2355 imsm_orom_has_chunk(orom, 1024*64) ? " 64M" : "");
29cd0821
CA		 2356 printf(" 2TB volumes :%s supported\n",
2357 (orom->attr & IMSM_OROM_ATTR_2TB)?"":" not");
2358 printf(" 2TB disks :%s supported\n",
2359 (orom->attr & IMSM_OROM_ATTR_2TB_DISK)?"":" not");
0e7f69a8 2360 printf(" Max Disks : %d\n", orom->tds);
0858eccf
AP		 2361 printf(" Max Volumes : %d per array, %d per %s\n",
2362 orom->vpa, orom->vphba,
2363 imsm_orom_is_nvme(orom) ? "platform" : "controller");
a891a3c2
LM		 2364 return;
2365}
2366
e50cf220
MN		 2367static void print_imsm_capability_export(const struct imsm_orom *orom)
2368{
2369 printf("MD_FIRMWARE_TYPE=imsm\n");
0858eccf
AP		 2370 if (orom->major_ver || orom->minor_ver || orom->hotfix_ver || orom->build)
2371 printf("IMSM_VERSION=%d.%d.%d.%d\n", orom->major_ver, orom->minor_ver,
2372 orom->hotfix_ver, orom->build);
e50cf220
MN		 2373 printf("IMSM_SUPPORTED_RAID_LEVELS=%s%s%s%s%s\n",
2374 imsm_orom_has_raid0(orom) ? "raid0 " : "",
2375 imsm_orom_has_raid1(orom) ? "raid1 " : "",
2376 imsm_orom_has_raid1e(orom) ? "raid1e " : "",
2377 imsm_orom_has_raid5(orom) ? "raid10 " : "",
2378 imsm_orom_has_raid10(orom) ? "raid5 " : "");
2379 printf("IMSM_SUPPORTED_CHUNK_SIZES=%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
2380 imsm_orom_has_chunk(orom, 2) ? "2k " : "",
2381 imsm_orom_has_chunk(orom, 4) ? "4k " : "",
2382 imsm_orom_has_chunk(orom, 8) ? "8k " : "",
2383 imsm_orom_has_chunk(orom, 16) ? "16k " : "",
2384 imsm_orom_has_chunk(orom, 32) ? "32k " : "",
2385 imsm_orom_has_chunk(orom, 64) ? "64k " : "",
2386 imsm_orom_has_chunk(orom, 128) ? "128k " : "",
2387 imsm_orom_has_chunk(orom, 256) ? "256k " : "",
2388 imsm_orom_has_chunk(orom, 512) ? "512k " : "",
2389 imsm_orom_has_chunk(orom, 1024*1) ? "1M " : "",
2390 imsm_orom_has_chunk(orom, 1024*2) ? "2M " : "",
2391 imsm_orom_has_chunk(orom, 1024*4) ? "4M " : "",
2392 imsm_orom_has_chunk(orom, 1024*8) ? "8M " : "",
2393 imsm_orom_has_chunk(orom, 1024*16) ? "16M " : "",
2394 imsm_orom_has_chunk(orom, 1024*32) ? "32M " : "",
2395 imsm_orom_has_chunk(orom, 1024*64) ? "64M " : "");
2396 printf("IMSM_2TB_VOLUMES=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB) ? "yes" : "no");
2397 printf("IMSM_2TB_DISKS=%s\n",(orom->attr & IMSM_OROM_ATTR_2TB_DISK) ? "yes" : "no");
2398 printf("IMSM_MAX_DISKS=%d\n",orom->tds);
2399 printf("IMSM_MAX_VOLUMES_PER_ARRAY=%d\n",orom->vpa);
2400 printf("IMSM_MAX_VOLUMES_PER_CONTROLLER=%d\n",orom->vphba);
2401}
2402
9eafa1de 2403static int detail_platform_imsm(int verbose, int enumerate_only, char *controller_path)
d665cc31
DW		 2404{
2405 /* There are two components to imsm platform support, the ahci SATA
2406 * controller and the option-rom. To find the SATA controller we
2407 * simply look in /sys/bus/pci/drivers/ahci to see if an ahci
2408 * controller with the Intel vendor id is present. This approach
2409 * allows mdadm to leverage the kernel's ahci detection logic, with the
2410 * caveat that if ahci.ko is not loaded mdadm will not be able to
2411 * detect platform raid capabilities. The option-rom resides in a
2412 * platform "Adapter ROM". We scan for its signature to retrieve the
2413 * platform capabilities. If raid support is disabled in the BIOS the
2414 * option-rom capability structure will not be available.
2415 */
d665cc31 2416 struct sys_dev *list, *hba;
d665cc31
DW		 2417 int host_base = 0;
2418 int port_count = 0;
9eafa1de 2419 int result=1;
d665cc31 2420
5615172f 2421 if (enumerate_only) {
a891a3c2 2422 if (check_env("IMSM_NO_PLATFORM"))
5615172f 2423 return 0;
a891a3c2
LM		 2424 list = find_intel_devices();
2425 if (!list)
2426 return 2;
2427 for (hba = list; hba; hba = hba->next) {
6b781d33
AP		 2428 if (find_imsm_capability(hba)) {
2429 result = 0;
a891a3c2
LM		 2430 break;
2431 }
9eafa1de 2432 else
6b781d33 2433 result = 2;
a891a3c2 2434 }
a891a3c2 2435 return result;
5615172f
DW		 2436 }
2437
155cbb4c
LM		 2438 list = find_intel_devices();
2439 if (!list) {
ba728be7 2440 if (verbose > 0)
7a862a02 2441 pr_err("no active Intel(R) RAID controller found.\n");
d665cc31 2442 return 2;
ba728be7 2443 } else if (verbose > 0)
155cbb4c 2444 print_found_intel_controllers(list);
d665cc31 2445
a891a3c2 2446 for (hba = list; hba; hba = hba->next) {
0858eccf 2447 if (controller_path && (compare_paths(hba->path, controller_path) != 0))
9eafa1de 2448 continue;
0858eccf 2449 if (!find_imsm_capability(hba)) {
60f0f54d 2450 char buf[PATH_MAX];
e7b84f9d 2451 pr_err("imsm capabilities not found for controller: %s (type %s)\n",
60f0f54d
PB		 2452 hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path,
2453 get_sys_dev_type(hba->type));
0858eccf
AP		 2454 continue;
2455 }
2456 result = 0;
2457 }
2458
2459 if (controller_path && result == 1) {
2460 pr_err("no active Intel(R) RAID controller found under %s\n",
2461 controller_path);
2462 return result;
2463 }
2464
5e1d6128 2465 const struct orom_entry *entry;
0858eccf 2466
5e1d6128 2467 for (entry = orom_entries; entry; entry = entry->next) {
60f0f54d 2468 if (entry->type == SYS_DEV_VMD) {
07cb1e57 2469 print_imsm_capability(&entry->orom);
32716c51
PB		 2470 printf(" 3rd party NVMe :%s supported\n",
2471 imsm_orom_has_tpv_support(&entry->orom)?"":" not");
60f0f54d
PB		 2472 for (hba = list; hba; hba = hba->next) {
2473 if (hba->type == SYS_DEV_VMD) {
2474 char buf[PATH_MAX];
60f0f54d
PB		 2475 printf(" I/O Controller : %s (%s)\n",
2476 vmd_domain_to_controller(hba, buf), get_sys_dev_type(hba->type));
b5eece69
PB		 2477 if (print_vmd_attached_devs(hba)) {
2478 if (verbose > 0)
2479 pr_err("failed to get devices attached to VMD domain.\n");
2480 result |= 2;
2481 }
60f0f54d
PB		 2482 }
2483 }
07cb1e57 2484 printf("\n");
60f0f54d
PB		 2485 continue;
2486 }
0858eccf 2487
60f0f54d
PB		 2488 print_imsm_capability(&entry->orom);
2489 if (entry->type == SYS_DEV_NVME) {
0858eccf
AP		 2490 for (hba = list; hba; hba = hba->next) {
2491 if (hba->type == SYS_DEV_NVME)
2492 printf(" NVMe Device : %s\n", hba->path);
2493 }
60f0f54d 2494 printf("\n");
0858eccf
AP		 2495 continue;
2496 }
2497
2498 struct devid_list *devid;
5e1d6128 2499 for (devid = entry->devid_list; devid; devid = devid->next) {
0858eccf
AP		 2500 hba = device_by_id(devid->devid);
2501 if (!hba)
2502 continue;
2503
9eafa1de
MN		 2504 printf(" I/O Controller : %s (%s)\n",
2505 hba->path, get_sys_dev_type(hba->type));
2506 if (hba->type == SYS_DEV_SATA) {
2507 host_base = ahci_get_port_count(hba->path, &port_count);
2508 if (ahci_enumerate_ports(hba->path, port_count, host_base, verbose)) {
2509 if (verbose > 0)
7a862a02 2510 pr_err("failed to enumerate ports on SATA controller at %s.\n", hba->pci_id);
9eafa1de
MN		 2511 result |= 2;
2512 }
120dc887
LM		 2513 }
2514 }
0858eccf 2515 printf("\n");
d665cc31 2516 }
155cbb4c 2517
120dc887 2518 return result;
d665cc31 2519}
e50cf220 2520
9eafa1de 2521static int export_detail_platform_imsm(int verbose, char *controller_path)
e50cf220 2522{
e50cf220
MN		 2523 struct sys_dev *list, *hba;
2524 int result=1;
2525
2526 list = find_intel_devices();
2527 if (!list) {
2528 if (verbose > 0)
2529 pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_INTEL_DEVICES\n");
2530 result = 2;
e50cf220
MN		 2531 return result;
2532 }
2533
2534 for (hba = list; hba; hba = hba->next) {
9eafa1de
MN		 2535 if (controller_path && (compare_paths(hba->path,controller_path) != 0))
2536 continue;
60f0f54d
PB		 2537 if (!find_imsm_capability(hba) && verbose > 0) {
2538 char buf[PATH_MAX];
2539 pr_err("IMSM_DETAIL_PLATFORM_ERROR=NO_IMSM_CAPABLE_DEVICE_UNDER_%s\n",
2540 hba->type == SYS_DEV_VMD ? vmd_domain_to_controller(hba, buf) : hba->path);
2541 }
0858eccf 2542 else
e50cf220 2543 result = 0;
e50cf220
MN		 2544 }
2545
5e1d6128 2546 const struct orom_entry *entry;
0858eccf 2547
60f0f54d
PB		 2548 for (entry = orom_entries; entry; entry = entry->next) {
2549 if (entry->type == SYS_DEV_VMD) {
2550 for (hba = list; hba; hba = hba->next)
2551 print_imsm_capability_export(&entry->orom);
2552 continue;
2553 }
5e1d6128 2554 print_imsm_capability_export(&entry->orom);
60f0f54d 2555 }
0858eccf 2556
e50cf220
MN		 2557 return result;
2558}
2559
cdddbdbc
DW		 2560#endif
2561
2562static int match_home_imsm(struct supertype *st, char *homehost)
2563{
5115ca67
DW		 2564 /* the imsm metadata format does not specify any host
2565 * identification information. We return -1 since we can never
2566 * confirm nor deny whether a given array is "meant" for this
148acb7b 2567 * host. We rely on compare_super and the 'family_num' fields to
5115ca67
DW		 2568 * exclude member disks that do not belong, and we rely on
2569 * mdadm.conf to specify the arrays that should be assembled.
2570 * Auto-assembly may still pick up "foreign" arrays.
2571 */
cdddbdbc 2572
9362c1c8 2573 return -1;
cdddbdbc
DW		 2574}
2575
2576static void uuid_from_super_imsm(struct supertype *st, int uuid[4])
2577{
51006d85
N		 2578 /* The uuid returned here is used for:
2579 * uuid to put into bitmap file (Create, Grow)
2580 * uuid for backup header when saving critical section (Grow)
2581 * comparing uuids when re-adding a device into an array
2582 * In these cases the uuid required is that of the data-array,
2583 * not the device-set.
2584 * uuid to recognise same set when adding a missing device back
2585 * to an array. This is a uuid for the device-set.
1011e834 2586 *
51006d85
N		 2587 * For each of these we can make do with a truncated
2588 * or hashed uuid rather than the original, as long as
2589 * everyone agrees.
2590 * In each case the uuid required is that of the data-array,
2591 * not the device-set.
43dad3d6 2592 */
51006d85
N		 2593 /* imsm does not track uuid's so we synthesis one using sha1 on
2594 * - The signature (Which is constant for all imsm array, but no matter)
148acb7b 2595 * - the orig_family_num of the container
51006d85
N		 2596 * - the index number of the volume
2597 * - the 'serial' number of the volume.
2598 * Hopefully these are all constant.
2599 */
2600 struct intel_super *super = st->sb;
43dad3d6 2601
51006d85
N		 2602 char buf[20];
2603 struct sha1_ctx ctx;
2604 struct imsm_dev *dev = NULL;
148acb7b 2605 __u32 family_num;
51006d85 2606
148acb7b
DW		 2607 /* some mdadm versions failed to set ->orig_family_num, in which
2608 * case fall back to ->family_num. orig_family_num will be
2609 * fixed up with the first metadata update.
2610 */
2611 family_num = super->anchor->orig_family_num;
2612 if (family_num == 0)
2613 family_num = super->anchor->family_num;
51006d85 2614 sha1_init_ctx(&ctx);
92bd8f8d 2615 sha1_process_bytes(super->anchor->sig, MPB_SIG_LEN, &ctx);
148acb7b 2616 sha1_process_bytes(&family_num, sizeof(__u32), &ctx);
51006d85
N		 2617 if (super->current_vol >= 0)
2618 dev = get_imsm_dev(super, super->current_vol);
2619 if (dev) {
2620 __u32 vol = super->current_vol;
2621 sha1_process_bytes(&vol, sizeof(vol), &ctx);
2622 sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx);
2623 }
2624 sha1_finish_ctx(&ctx, buf);
2625 memcpy(uuid, buf, 4*4);
cdddbdbc
DW		 2626}
2627
0d481d37 2628#if 0
4f5bc454
DW		 2629static void
2630get_imsm_numerical_version(struct imsm_super *mpb, int *m, int *p)
cdddbdbc 2631{
cdddbdbc
DW		 2632 __u8 *v = get_imsm_version(mpb);
2633 __u8 *end = mpb->sig + MAX_SIGNATURE_LENGTH;
2634 char major[] = { 0, 0, 0 };
2635 char minor[] = { 0 ,0, 0 };
2636 char patch[] = { 0, 0, 0 };
2637 char *ver_parse[] = { major, minor, patch };
2638 int i, j;
2639
2640 i = j = 0;
2641 while (*v != '\0' && v < end) {
2642 if (*v != '.' && j < 2)
2643 ver_parse[i][j++] = *v;
2644 else {
2645 i++;
2646 j = 0;
2647 }
2648 v++;
2649 }
2650
4f5bc454
DW		 2651 *m = strtol(minor, NULL, 0);
2652 *p = strtol(patch, NULL, 0);
2653}
0d481d37 2654#endif
4f5bc454 2655
1e5c6983
DW		 2656static __u32 migr_strip_blocks_resync(struct imsm_dev *dev)
2657{
2658 /* migr_strip_size when repairing or initializing parity */
238c0a71 2659 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2660 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2661
2662 switch (get_imsm_raid_level(map)) {
2663 case 5:
2664 case 10:
2665 return chunk;
2666 default:
2667 return 128*1024 >> 9;
2668 }
2669}
2670
2671static __u32 migr_strip_blocks_rebuild(struct imsm_dev *dev)
2672{
2673 /* migr_strip_size when rebuilding a degraded disk, no idea why
2674 * this is different than migr_strip_size_resync(), but it's good
2675 * to be compatible
2676 */
238c0a71 2677 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2678 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2679
2680 switch (get_imsm_raid_level(map)) {
2681 case 1:
2682 case 10:
2683 if (map->num_members % map->num_domains == 0)
2684 return 128*1024 >> 9;
2685 else
2686 return chunk;
2687 case 5:
2688 return max((__u32) 64*1024 >> 9, chunk);
2689 default:
2690 return 128*1024 >> 9;
2691 }
2692}
2693
2694static __u32 num_stripes_per_unit_resync(struct imsm_dev *dev)
2695{
238c0a71
AK		 2696 struct imsm_map *lo = get_imsm_map(dev, MAP_0);
2697 struct imsm_map *hi = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2698 __u32 lo_chunk = __le32_to_cpu(lo->blocks_per_strip);
2699 __u32 hi_chunk = __le32_to_cpu(hi->blocks_per_strip);
2700
2701 return max((__u32) 1, hi_chunk / lo_chunk);
2702}
2703
2704static __u32 num_stripes_per_unit_rebuild(struct imsm_dev *dev)
2705{
238c0a71 2706 struct imsm_map *lo = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2707 int level = get_imsm_raid_level(lo);
2708
2709 if (level == 1 || level == 10) {
238c0a71 2710 struct imsm_map *hi = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2711
2712 return hi->num_domains;
2713 } else
2714 return num_stripes_per_unit_resync(dev);
2715}
2716
98130f40 2717static __u8 imsm_num_data_members(struct imsm_dev *dev, int second_map)
1e5c6983
DW		 2718{
2719 /* named 'imsm_' because raid0, raid1 and raid10
2720 * counter-intuitively have the same number of data disks
2721 */
98130f40 2722 struct imsm_map *map = get_imsm_map(dev, second_map);
1e5c6983
DW		 2723
2724 switch (get_imsm_raid_level(map)) {
2725 case 0:
36fd8ccc
AK		 2726 return map->num_members;
2727 break;
1e5c6983
DW		 2728 case 1:
2729 case 10:
36fd8ccc 2730 return map->num_members/2;
1e5c6983
DW		 2731 case 5:
2732 return map->num_members - 1;
2733 default:
1ade5cc1 2734 dprintf("unsupported raid level\n");
1e5c6983
DW		 2735 return 0;
2736 }
2737}
2738
2739static __u32 parity_segment_depth(struct imsm_dev *dev)
2740{
238c0a71 2741 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2742 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2743
2744 switch(get_imsm_raid_level(map)) {
2745 case 1:
2746 case 10:
2747 return chunk * map->num_domains;
2748 case 5:
2749 return chunk * map->num_members;
2750 default:
2751 return chunk;
2752 }
2753}
2754
2755static __u32 map_migr_block(struct imsm_dev *dev, __u32 block)
2756{
238c0a71 2757 struct imsm_map *map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 2758 __u32 chunk = __le32_to_cpu(map->blocks_per_strip);
2759 __u32 strip = block / chunk;
2760
2761 switch (get_imsm_raid_level(map)) {
2762 case 1:
2763 case 10: {
2764 __u32 vol_strip = (strip * map->num_domains) + 1;
2765 __u32 vol_stripe = vol_strip / map->num_members;
2766
2767 return vol_stripe * chunk + block % chunk;
2768 } case 5: {
2769 __u32 stripe = strip / (map->num_members - 1);
2770
2771 return stripe * chunk + block % chunk;
2772 }
2773 default:
2774 return 0;
2775 }
2776}
2777
c47b0ff6
AK		 2778static __u64 blocks_per_migr_unit(struct intel_super *super,
2779 struct imsm_dev *dev)
1e5c6983
DW		 2780{
2781 /* calculate the conversion factor between per member 'blocks'
2782 * (md/{resync,rebuild}_start) and imsm migration units, return
2783 * 0 for the 'not migrating' and 'unsupported migration' cases
2784 */
2785 if (!dev->vol.migr_state)
2786 return 0;
2787
2788 switch (migr_type(dev)) {
c47b0ff6
AK		 2789 case MIGR_GEN_MIGR: {
2790 struct migr_record *migr_rec = super->migr_rec;
2791 return __le32_to_cpu(migr_rec->blocks_per_unit);
2792 }
1e5c6983
DW		 2793 case MIGR_VERIFY:
2794 case MIGR_REPAIR:
2795 case MIGR_INIT: {
238c0a71 2796 struct imsm_map *map = get_imsm_map(dev, MAP_0);
1e5c6983
DW		 2797 __u32 stripes_per_unit;
2798 __u32 blocks_per_unit;
2799 __u32 parity_depth;
2800 __u32 migr_chunk;
2801 __u32 block_map;
2802 __u32 block_rel;
2803 __u32 segment;
2804 __u32 stripe;
2805 __u8 disks;
2806
2807 /* yes, this is really the translation of migr_units to
2808 * per-member blocks in the 'resync' case
2809 */
2810 stripes_per_unit = num_stripes_per_unit_resync(dev);
2811 migr_chunk = migr_strip_blocks_resync(dev);
238c0a71 2812 disks = imsm_num_data_members(dev, MAP_0);
1e5c6983 2813 blocks_per_unit = stripes_per_unit * migr_chunk * disks;
7b1ab482 2814 stripe = __le16_to_cpu(map->blocks_per_strip) * disks;
1e5c6983
DW		 2815 segment = blocks_per_unit / stripe;
2816 block_rel = blocks_per_unit - segment * stripe;
2817 parity_depth = parity_segment_depth(dev);
2818 block_map = map_migr_block(dev, block_rel);
2819 return block_map + parity_depth * segment;
2820 }
2821 case MIGR_REBUILD: {
2822 __u32 stripes_per_unit;
2823 __u32 migr_chunk;
2824
2825 stripes_per_unit = num_stripes_per_unit_rebuild(dev);
2826 migr_chunk = migr_strip_blocks_rebuild(dev);
2827 return migr_chunk * stripes_per_unit;
2828 }
1e5c6983
DW		 2829 case MIGR_STATE_CHANGE:
2830 default:
2831 return 0;
2832 }
2833}
2834
c2c087e6
DW		 2835static int imsm_level_to_layout(int level)
2836{
2837 switch (level) {
2838 case 0:
2839 case 1:
2840 return 0;
2841 case 5:
2842 case 6:
a380c027 2843 return ALGORITHM_LEFT_ASYMMETRIC;
c2c087e6 2844 case 10:
c92a2527 2845 return 0x102;
c2c087e6 2846 }
a18a888e 2847 return UnSet;
c2c087e6
DW		 2848}
2849
8e59f3d8
AK		 2850/*******************************************************************************
2851 * Function: read_imsm_migr_rec
2852 * Description: Function reads imsm migration record from last sector of disk
2853 * Parameters:
2854 * fd : disk descriptor
2855 * super : metadata info
2856 * Returns:
2857 * 0 : success,
2858 * -1 : fail
2859 ******************************************************************************/
2860static int read_imsm_migr_rec(int fd, struct intel_super *super)
2861{
2862 int ret_val = -1;
de44e46f 2863 unsigned int sector_size = super->sector_size;
8e59f3d8
AK		 2864 unsigned long long dsize;
2865
2866 get_dev_size(fd, NULL, &dsize);
de44e46f
PB		 2867 if (lseek64(fd, dsize - (sector_size*MIGR_REC_SECTOR_POSITION),
2868 SEEK_SET) < 0) {
e7b84f9d
N		 2869 pr_err("Cannot seek to anchor block: %s\n",
2870 strerror(errno));
8e59f3d8
AK		 2871 goto out;
2872 }
466070ad 2873 if ((unsigned int)read(fd, super->migr_rec_buf,
de44e46f
PB		 2874 MIGR_REC_BUF_SECTORS*sector_size) !=
2875 MIGR_REC_BUF_SECTORS*sector_size) {
e7b84f9d
N		 2876 pr_err("Cannot read migr record block: %s\n",
2877 strerror(errno));
8e59f3d8
AK		 2878 goto out;
2879 }
2880 ret_val = 0;
de44e46f
PB		 2881 if (sector_size == 4096)
2882 convert_from_4k_imsm_migr_rec(super);
8e59f3d8
AK		 2883
2884out:
2885 return ret_val;
2886}
2887
3136abe5
AK		 2888static struct imsm_dev *imsm_get_device_during_migration(
2889 struct intel_super *super)
2890{
2891
2892 struct intel_dev *dv;
2893
2894 for (dv = super->devlist; dv; dv = dv->next) {
2895 if (is_gen_migration(dv->dev))
2896 return dv->dev;
2897 }
2898 return NULL;
2899}
2900
8e59f3d8
AK		 2901/*******************************************************************************
2902 * Function: load_imsm_migr_rec
2903 * Description: Function reads imsm migration record (it is stored at the last
2904 * sector of disk)
2905 * Parameters:
2906 * super : imsm internal array info
2907 * info : general array info
2908 * Returns:
2909 * 0 : success
2910 * -1 : fail
4c965cc9 2911 * -2 : no migration in progress
8e59f3d8
AK		 2912 ******************************************************************************/
2913static int load_imsm_migr_rec(struct intel_super *super, struct mdinfo *info)
2914{
2915 struct mdinfo *sd;
594dc1b8 2916 struct dl *dl;
8e59f3d8
AK		 2917 char nm[30];
2918 int retval = -1;
2919 int fd = -1;
3136abe5 2920 struct imsm_dev *dev;
594dc1b8 2921 struct imsm_map *map;
b4ab44d8 2922 int slot = -1;
3136abe5
AK		 2923
2924 /* find map under migration */
2925 dev = imsm_get_device_during_migration(super);
2926 /* nothing to load,no migration in progress?
2927 */
2928 if (dev == NULL)
4c965cc9 2929 return -2;
8e59f3d8
AK		 2930
2931 if (info) {
2932 for (sd = info->devs ; sd ; sd = sd->next) {
2933 /* read only from one of the first two slots */
12fe93e9
TM		 2934 if ((sd->disk.raid_disk < 0) ||
2935 (sd->disk.raid_disk > 1))
8e59f3d8 2936 continue;
3136abe5 2937
8e59f3d8
AK		 2938 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
2939 fd = dev_open(nm, O_RDONLY);
2940 if (fd >= 0)
2941 break;
2942 }
2943 }
2944 if (fd < 0) {
12fe93e9 2945 map = get_imsm_map(dev, MAP_0);
8e59f3d8 2946 for (dl = super->disks; dl; dl = dl->next) {
3136abe5
AK		 2947 /* skip spare and failed disks
2948 */
2949 if (dl->index < 0)
2950 continue;
8e59f3d8 2951 /* read only from one of the first two slots */
3136abe5
AK		 2952 if (map)
2953 slot = get_imsm_disk_slot(map, dl->index);
089f9d79 2954 if (map == NULL || slot > 1 || slot < 0)
8e59f3d8
AK		 2955 continue;
2956 sprintf(nm, "%d:%d", dl->major, dl->minor);
2957 fd = dev_open(nm, O_RDONLY);
2958 if (fd >= 0)
2959 break;
2960 }
2961 }
2962 if (fd < 0)
2963 goto out;
2964 retval = read_imsm_migr_rec(fd, super);
2965
2966out:
2967 if (fd >= 0)
2968 close(fd);
2969 return retval;
2970}
2971
9e2d750d 2972#ifndef MDASSEMBLE
c17608ea
AK		 2973/*******************************************************************************
2974 * function: imsm_create_metadata_checkpoint_update
2975 * Description: It creates update for checkpoint change.
2976 * Parameters:
2977 * super : imsm internal array info
2978 * u : pointer to prepared update
2979 * Returns:
2980 * Uptate length.
2981 * If length is equal to 0, input pointer u contains no update
2982 ******************************************************************************/
2983static int imsm_create_metadata_checkpoint_update(
2984 struct intel_super *super,
2985 struct imsm_update_general_migration_checkpoint **u)
2986{
2987
2988 int update_memory_size = 0;
2989
1ade5cc1 2990 dprintf("(enter)\n");
c17608ea
AK		 2991
2992 if (u == NULL)
2993 return 0;
2994 *u = NULL;
2995
2996 /* size of all update data without anchor */
2997 update_memory_size =
2998 sizeof(struct imsm_update_general_migration_checkpoint);
2999
503975b9 3000 *u = xcalloc(1, update_memory_size);
c17608ea 3001 if (*u == NULL) {
1ade5cc1 3002 dprintf("error: cannot get memory\n");
c17608ea
AK		 3003 return 0;
3004 }
3005 (*u)->type = update_general_migration_checkpoint;
3006 (*u)->curr_migr_unit = __le32_to_cpu(super->migr_rec->curr_migr_unit);
1ade5cc1 3007 dprintf("prepared for %u\n", (*u)->curr_migr_unit);
c17608ea
AK		 3008
3009 return update_memory_size;
3010}
3011
c17608ea
AK		 3012static void imsm_update_metadata_locally(struct supertype *st,
3013 void *buf, int len);
3014
687629c2
AK		 3015/*******************************************************************************
3016 * Function: write_imsm_migr_rec
3017 * Description: Function writes imsm migration record
3018 * (at the last sector of disk)
3019 * Parameters:
3020 * super : imsm internal array info
3021 * Returns:
3022 * 0 : success
3023 * -1 : if fail
3024 ******************************************************************************/
3025static int write_imsm_migr_rec(struct supertype *st)
3026{
3027 struct intel_super *super = st->sb;
de44e46f 3028 unsigned int sector_size = super->sector_size;
687629c2
AK		 3029 unsigned long long dsize;
3030 char nm[30];
3031 int fd = -1;
3032 int retval = -1;
3033 struct dl *sd;
c17608ea
AK		 3034 int len;
3035 struct imsm_update_general_migration_checkpoint *u;
3136abe5 3036 struct imsm_dev *dev;
594dc1b8 3037 struct imsm_map *map;
3136abe5
AK		 3038
3039 /* find map under migration */
3040 dev = imsm_get_device_during_migration(super);
3041 /* if no migration, write buffer anyway to clear migr_record
3042 * on disk based on first available device
3043 */
3044 if (dev == NULL)
3045 dev = get_imsm_dev(super, super->current_vol < 0 ? 0 :
3046 super->current_vol);
3047
44bfe6df 3048 map = get_imsm_map(dev, MAP_0);
687629c2 3049
de44e46f
PB		 3050 if (sector_size == 4096)
3051 convert_to_4k_imsm_migr_rec(super);
687629c2 3052 for (sd = super->disks ; sd ; sd = sd->next) {
b4ab44d8 3053 int slot = -1;
3136abe5
AK		 3054
3055 /* skip failed and spare devices */
3056 if (sd->index < 0)
3057 continue;
687629c2 3058 /* write to 2 first slots only */
3136abe5
AK		 3059 if (map)
3060 slot = get_imsm_disk_slot(map, sd->index);
089f9d79 3061 if (map == NULL || slot > 1 || slot < 0)
687629c2 3062 continue;
3136abe5 3063
687629c2
AK		 3064 sprintf(nm, "%d:%d", sd->major, sd->minor);
3065 fd = dev_open(nm, O_RDWR);
3066 if (fd < 0)
3067 continue;
3068 get_dev_size(fd, NULL, &dsize);
de44e46f
PB		 3069 if (lseek64(fd, dsize - (MIGR_REC_SECTOR_POSITION*sector_size),
3070 SEEK_SET) < 0) {
e7b84f9d
N		 3071 pr_err("Cannot seek to anchor block: %s\n",
3072 strerror(errno));
687629c2
AK		 3073 goto out;
3074 }
466070ad 3075 if ((unsigned int)write(fd, super->migr_rec_buf,
de44e46f
PB		 3076 MIGR_REC_BUF_SECTORS*sector_size) !=
3077 MIGR_REC_BUF_SECTORS*sector_size) {
e7b84f9d
N		 3078 pr_err("Cannot write migr record block: %s\n",
3079 strerror(errno));
687629c2
AK		 3080 goto out;
3081 }
3082 close(fd);
3083 fd = -1;
3084 }
de44e46f
PB		 3085 if (sector_size == 4096)
3086 convert_from_4k_imsm_migr_rec(super);
c17608ea
AK		 3087 /* update checkpoint information in metadata */
3088 len = imsm_create_metadata_checkpoint_update(super, &u);
c17608ea
AK		 3089 if (len <= 0) {
3090 dprintf("imsm: Cannot prepare update\n");
3091 goto out;
3092 }
3093 /* update metadata locally */
3094 imsm_update_metadata_locally(st, u, len);
3095 /* and possibly remotely */
3096 if (st->update_tail) {
3097 append_metadata_update(st, u, len);
3098 /* during reshape we do all work inside metadata handler
3099 * manage_reshape(), so metadata update has to be triggered
3100 * insida it
3101 */
3102 flush_metadata_updates(st);
3103 st->update_tail = &st->updates;
3104 } else
3105 free(u);
687629c2
AK		 3106
3107 retval = 0;
3108 out:
3109 if (fd >= 0)
3110 close(fd);
3111 return retval;
3112}
9e2d750d 3113#endif /* MDASSEMBLE */
687629c2 3114
e2962bfc
AK		 3115/* spare/missing disks activations are not allowe when
3116 * array/container performs reshape operation, because
3117 * all arrays in container works on the same disks set
3118 */
3119int imsm_reshape_blocks_arrays_changes(struct intel_super *super)
3120{
3121 int rv = 0;
3122 struct intel_dev *i_dev;
3123 struct imsm_dev *dev;
3124
3125 /* check whole container
3126 */
3127 for (i_dev = super->devlist; i_dev; i_dev = i_dev->next) {
3128 dev = i_dev->dev;
3ad25638 3129 if (is_gen_migration(dev)) {
e2962bfc
AK		 3130 /* No repair during any migration in container
3131 */
3132 rv = 1;
3133 break;
3134 }
3135 }
3136 return rv;
3137}
c41e00b2
AK		 3138static unsigned long long imsm_component_size_aligment_check(int level,
3139 int chunk_size,
f36a9ecd 3140 unsigned int sector_size,
c41e00b2
AK		 3141 unsigned long long component_size)
3142{
3143 unsigned int component_size_alligment;
3144
3145 /* check component size aligment
3146 */
f36a9ecd 3147 component_size_alligment = component_size % (chunk_size/sector_size);
c41e00b2 3148
1ade5cc1 3149 dprintf("(Level: %i, chunk_size = %i, component_size = %llu), component_size_alligment = %u\n",
c41e00b2
AK		 3150 level, chunk_size, component_size,
3151 component_size_alligment);
3152
3153 if (component_size_alligment && (level != 1) && (level != UnSet)) {
3154 dprintf("imsm: reported component size alligned from %llu ",
3155 component_size);
3156 component_size -= component_size_alligment;
1ade5cc1 3157 dprintf_cont("to %llu (%i).\n",
c41e00b2
AK		 3158 component_size, component_size_alligment);
3159 }
3160
3161 return component_size;
3162}
e2962bfc 3163
2432ce9b
AP		 3164static unsigned long long get_ppl_sector(struct intel_super *super, int dev_idx)
3165{
3166 struct imsm_dev *dev = get_imsm_dev(super, dev_idx);
3167 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3168
3169 return pba_of_lba0(map) +
3170 (num_data_stripes(map) * map->blocks_per_strip);
3171}
3172
a5d85af7 3173static void getinfo_super_imsm_volume(struct supertype *st, struct mdinfo *info, char *dmap)
bf5a934a
DW		 3174{
3175 struct intel_super *super = st->sb;
c47b0ff6 3176 struct migr_record *migr_rec = super->migr_rec;
949c47a0 3177 struct imsm_dev *dev = get_imsm_dev(super, super->current_vol);
238c0a71
AK		 3178 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3179 struct imsm_map *prev_map = get_imsm_map(dev, MAP_1);
b335e593 3180 struct imsm_map *map_to_analyse = map;
efb30e7f 3181 struct dl *dl;
a5d85af7 3182 int map_disks = info->array.raid_disks;
bf5a934a 3183
95eeceeb 3184 memset(info, 0, sizeof(*info));
b335e593
AK		 3185 if (prev_map)
3186 map_to_analyse = prev_map;
3187
ca0748fa 3188 dl = super->current_disk;
9894ec0d 3189
bf5a934a 3190 info->container_member = super->current_vol;
cd0430a1 3191 info->array.raid_disks = map->num_members;
b335e593 3192 info->array.level = get_imsm_raid_level(map_to_analyse);
bf5a934a
DW		 3193 info->array.layout = imsm_level_to_layout(info->array.level);
3194 info->array.md_minor = -1;
3195 info->array.ctime = 0;
3196 info->array.utime = 0;
b335e593
AK		 3197 info->array.chunk_size =
3198 __le16_to_cpu(map_to_analyse->blocks_per_strip) << 9;
2432ce9b 3199 info->array.state = !(dev->vol.dirty & RAIDVOL_DIRTY);
da9b4a62
DW		 3200 info->custom_array_size = __le32_to_cpu(dev->size_high);
3201 info->custom_array_size <<= 32;
3202 info->custom_array_size |= __le32_to_cpu(dev->size_low);
3ad25638
AK		 3203 info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb);
3204
3f510843 3205 if (is_gen_migration(dev)) {
3f83228a 3206 info->reshape_active = 1;
b335e593
AK		 3207 info->new_level = get_imsm_raid_level(map);
3208 info->new_layout = imsm_level_to_layout(info->new_level);
3209 info->new_chunk = __le16_to_cpu(map->blocks_per_strip) << 9;
3f83228a 3210 info->delta_disks = map->num_members - prev_map->num_members;
493f5dd6
N		 3211 if (info->delta_disks) {
3212 /* this needs to be applied to every array
3213 * in the container.
3214 */
81219e70 3215 info->reshape_active = CONTAINER_RESHAPE;
493f5dd6 3216 }
3f83228a
N		 3217 /* We shape information that we give to md might have to be
3218 * modify to cope with md's requirement for reshaping arrays.
3219 * For example, when reshaping a RAID0, md requires it to be
3220 * presented as a degraded RAID4.
3221 * Also if a RAID0 is migrating to a RAID5 we need to specify
3222 * the array as already being RAID5, but the 'before' layout
3223 * is a RAID4-like layout.
3224 */
3225 switch (info->array.level) {
3226 case 0:
3227 switch(info->new_level) {
3228 case 0:
3229 /* conversion is happening as RAID4 */
3230 info->array.level = 4;
3231 info->array.raid_disks += 1;
3232 break;
3233 case 5:
3234 /* conversion is happening as RAID5 */
3235 info->array.level = 5;
3236 info->array.layout = ALGORITHM_PARITY_N;
3f83228a
N		 3237 info->delta_disks -= 1;
3238 break;
3239 default:
3240 /* FIXME error message */
3241 info->array.level = UnSet;
3242 break;
3243 }
3244 break;
3245 }
b335e593
AK		 3246 } else {
3247 info->new_level = UnSet;
3248 info->new_layout = UnSet;
3249 info->new_chunk = info->array.chunk_size;
3f83228a 3250 info->delta_disks = 0;
b335e593 3251 }
ca0748fa 3252
efb30e7f
DW		 3253 if (dl) {
3254 info->disk.major = dl->major;
3255 info->disk.minor = dl->minor;
ca0748fa 3256 info->disk.number = dl->index;
656b6b5a
N		 3257 info->disk.raid_disk = get_imsm_disk_slot(map_to_analyse,
3258 dl->index);
efb30e7f 3259 }
bf5a934a 3260
5551b113 3261 info->data_offset = pba_of_lba0(map_to_analyse);
06fb291a
PB		 3262
3263 if (info->array.level == 5) {
3264 info->component_size = num_data_stripes(map_to_analyse) *
3265 map_to_analyse->blocks_per_strip;
3266 } else {
3267 info->component_size = blocks_per_member(map_to_analyse);
3268 }
139dae11 3269
c41e00b2
AK		 3270 info->component_size = imsm_component_size_aligment_check(
3271 info->array.level,
3272 info->array.chunk_size,
f36a9ecd 3273 super->sector_size,
c41e00b2 3274 info->component_size);
5e46202e 3275 info->bb.supported = 1;
139dae11 3276
301406c9 3277 memset(info->uuid, 0, sizeof(info->uuid));
921d9e16 3278 info->recovery_start = MaxSector;
bf5a934a 3279
2432ce9b
AP		 3280 if (info->array.level == 5 && dev->rwh_policy == RWH_DISTRIBUTED) {
3281 info->consistency_policy = CONSISTENCY_POLICY_PPL;
3282 info->ppl_sector = get_ppl_sector(super, super->current_vol);
3283 info->ppl_size = (PPL_HEADER_SIZE + PPL_ENTRY_SPACE) >> 9;
3284 } else if (info->array.level <= 0) {
3285 info->consistency_policy = CONSISTENCY_POLICY_NONE;
3286 } else {
3287 info->consistency_policy = CONSISTENCY_POLICY_RESYNC;
3288 }
3289
d2e6d5d6 3290 info->reshape_progress = 0;
b6796ce1 3291 info->resync_start = MaxSector;
b9172665 3292 if ((map_to_analyse->map_state == IMSM_T_STATE_UNINITIALIZED ||
2432ce9b 3293 !(info->array.state & 1)) &&
b9172665 3294 imsm_reshape_blocks_arrays_changes(super) == 0) {
301406c9 3295 info->resync_start = 0;
b6796ce1
AK		 3296 }
3297 if (dev->vol.migr_state) {
1e5c6983
DW		 3298 switch (migr_type(dev)) {
3299 case MIGR_REPAIR:
3300 case MIGR_INIT: {
c47b0ff6
AK		 3301 __u64 blocks_per_unit = blocks_per_migr_unit(super,
3302 dev);
1e5c6983
DW		 3303 __u64 units = __le32_to_cpu(dev->vol.curr_migr_unit);
3304
3305 info->resync_start = blocks_per_unit * units;
3306 break;
3307 }
d2e6d5d6 3308 case MIGR_GEN_MIGR: {
c47b0ff6
AK		 3309 __u64 blocks_per_unit = blocks_per_migr_unit(super,
3310 dev);
3311 __u64 units = __le32_to_cpu(migr_rec->curr_migr_unit);
04fa9523
AK		 3312 unsigned long long array_blocks;
3313 int used_disks;
d2e6d5d6 3314
befb629b
AK		 3315 if (__le32_to_cpu(migr_rec->ascending_migr) &&
3316 (units <
3317 (__le32_to_cpu(migr_rec->num_migr_units)-1)) &&
3318 (super->migr_rec->rec_status ==
3319 __cpu_to_le32(UNIT_SRC_IN_CP_AREA)))
3320 units++;
3321
d2e6d5d6 3322 info->reshape_progress = blocks_per_unit * units;
6289d1e0 3323
7a862a02 3324 dprintf("IMSM: General Migration checkpoint : %llu (%llu) -> read reshape progress : %llu\n",
19986c72
MB		 3325 (unsigned long long)units,
3326 (unsigned long long)blocks_per_unit,
3327 info->reshape_progress);
75156c46 3328
238c0a71 3329 used_disks = imsm_num_data_members(dev, MAP_1);
75156c46 3330 if (used_disks > 0) {
5551b113 3331 array_blocks = blocks_per_member(map) *
75156c46
AK		 3332 used_disks;
3333 /* round array size down to closest MB
3334 */
3335 info->custom_array_size = (array_blocks
3336 >> SECT_PER_MB_SHIFT)
3337 << SECT_PER_MB_SHIFT;
3338 }
d2e6d5d6 3339 }
1e5c6983
DW		 3340 case MIGR_VERIFY:
3341 /* we could emulate the checkpointing of
3342 * 'sync_action=check' migrations, but for now
3343 * we just immediately complete them
3344 */
3345 case MIGR_REBUILD:
3346 /* this is handled by container_content_imsm() */
1e5c6983
DW		 3347 case MIGR_STATE_CHANGE:
3348 /* FIXME handle other migrations */
3349 default:
3350 /* we are not dirty, so... */
3351 info->resync_start = MaxSector;
3352 }
b6796ce1 3353 }
301406c9
DW		 3354
3355 strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN);
3356 info->name[MAX_RAID_SERIAL_LEN] = 0;
bf5a934a 3357
f35f2525
N		 3358 info->array.major_version = -1;
3359 info->array.minor_version = -2;
4dd2df09 3360 sprintf(info->text_version, "/%s/%d", st->container_devnm, info->container_member);
a67dd8cc 3361 info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */
51006d85 3362 uuid_from_super_imsm(st, info->uuid);
a5d85af7
N		 3363
3364 if (dmap) {
3365 int i, j;
3366 for (i=0; i<map_disks; i++) {
3367 dmap[i] = 0;
3368 if (i < info->array.raid_disks) {
3369 struct imsm_disk *dsk;
238c0a71 3370 j = get_imsm_disk_idx(dev, i, MAP_X);
a5d85af7
N		 3371 dsk = get_imsm_disk(super, j);
3372 if (dsk && (dsk->status & CONFIGURED_DISK))
3373 dmap[i] = 1;
3374 }
3375 }
3376 }
81ac8b4d 3377}
bf5a934a 3378
3b451610
AK		 3379static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev,
3380 int failed, int look_in_map);
3381
3382static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev,
3383 int look_in_map);
3384
b4ab44d8 3385#ifndef MDASSEMBLE
3b451610
AK		 3386static void manage_second_map(struct intel_super *super, struct imsm_dev *dev)
3387{
3388 if (is_gen_migration(dev)) {
3389 int failed;
3390 __u8 map_state;
3391 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
3392
3393 failed = imsm_count_failed(super, dev, MAP_1);
238c0a71 3394 map_state = imsm_check_degraded(super, dev, failed, MAP_1);
3b451610
AK		 3395 if (map2->map_state != map_state) {
3396 map2->map_state = map_state;
3397 super->updates_pending++;
3398 }
3399 }
3400}
b4ab44d8 3401#endif
97b4d0e9
DW		 3402
3403static struct imsm_disk *get_imsm_missing(struct intel_super *super, __u8 index)
3404{
3405 struct dl *d;
3406
3407 for (d = super->missing; d; d = d->next)
3408 if (d->index == index)
3409 return &d->disk;
3410 return NULL;
3411}
3412
a5d85af7 3413static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info, char *map)
4f5bc454
DW		 3414{
3415 struct intel_super *super = st->sb;
4f5bc454 3416 struct imsm_disk *disk;
a5d85af7 3417 int map_disks = info->array.raid_disks;
ab3cb6b3
N		 3418 int max_enough = -1;
3419 int i;
3420 struct imsm_super *mpb;
4f5bc454 3421
bf5a934a 3422 if (super->current_vol >= 0) {
a5d85af7 3423 getinfo_super_imsm_volume(st, info, map);
bf5a934a
DW		 3424 return;
3425 }
95eeceeb 3426 memset(info, 0, sizeof(*info));
d23fe947
DW		 3427
3428 /* Set raid_disks to zero so that Assemble will always pull in valid
3429 * spares
3430 */
3431 info->array.raid_disks = 0;
cdddbdbc
DW		 3432 info->array.level = LEVEL_CONTAINER;
3433 info->array.layout = 0;
3434 info->array.md_minor = -1;
1011e834 3435 info->array.ctime = 0; /* N/A for imsm */
cdddbdbc
DW		 3436 info->array.utime = 0;
3437 info->array.chunk_size = 0;
3438
3439 info->disk.major = 0;
3440 info->disk.minor = 0;
cdddbdbc 3441 info->disk.raid_disk = -1;
c2c087e6 3442 info->reshape_active = 0;
f35f2525
N		 3443 info->array.major_version = -1;
3444 info->array.minor_version = -2;
c2c087e6 3445 strcpy(info->text_version, "imsm");
a67dd8cc 3446 info->safe_mode_delay = 0;
c2c087e6
DW		 3447 info->disk.number = -1;
3448 info->disk.state = 0;
c5afc314 3449 info->name[0] = 0;
921d9e16 3450 info->recovery_start = MaxSector;
3ad25638 3451 info->recovery_blocked = imsm_reshape_blocks_arrays_changes(st->sb);
5e46202e 3452 info->bb.supported = 1;
c2c087e6 3453
97b4d0e9 3454 /* do we have the all the insync disks that we expect? */
ab3cb6b3 3455 mpb = super->anchor;
b7d81a38 3456 info->events = __le32_to_cpu(mpb->generation_num);
97b4d0e9 3457
ab3cb6b3
N		 3458 for (i = 0; i < mpb->num_raid_devs; i++) {
3459 struct imsm_dev *dev = get_imsm_dev(super, i);
3460 int failed, enough, j, missing = 0;
3461 struct imsm_map *map;
3462 __u8 state;
97b4d0e9 3463
3b451610
AK		 3464 failed = imsm_count_failed(super, dev, MAP_0);
3465 state = imsm_check_degraded(super, dev, failed, MAP_0);
238c0a71 3466 map = get_imsm_map(dev, MAP_0);
ab3cb6b3
N		 3467
3468 /* any newly missing disks?
3469 * (catches single-degraded vs double-degraded)
3470 */
3471 for (j = 0; j < map->num_members; j++) {
238c0a71 3472 __u32 ord = get_imsm_ord_tbl_ent(dev, j, MAP_0);
ab3cb6b3
N		 3473 __u32 idx = ord_to_idx(ord);
3474
3475 if (!(ord & IMSM_ORD_REBUILD) &&
3476 get_imsm_missing(super, idx)) {
3477 missing = 1;
3478 break;
3479 }
97b4d0e9 3480 }
ab3cb6b3
N		 3481
3482 if (state == IMSM_T_STATE_FAILED)
3483 enough = -1;
3484 else if (state == IMSM_T_STATE_DEGRADED &&
3485 (state != map->map_state || missing))
3486 enough = 0;
3487 else /* we're normal, or already degraded */
3488 enough = 1;
d2bde6d3
AK		 3489 if (is_gen_migration(dev) && missing) {
3490 /* during general migration we need all disks
3491 * that process is running on.
3492 * No new missing disk is allowed.
3493 */
3494 max_enough = -1;
3495 enough = -1;
3496 /* no more checks necessary
3497 */
3498 break;
3499 }
ab3cb6b3
N		 3500 /* in the missing/failed disk case check to see
3501 * if at least one array is runnable
3502 */
3503 max_enough = max(max_enough, enough);
3504 }
1ade5cc1 3505 dprintf("enough: %d\n", max_enough);
ab3cb6b3 3506 info->container_enough = max_enough;
97b4d0e9 3507
4a04ec6c 3508 if (super->disks) {
14e8215b
DW		 3509 __u32 reserved = imsm_reserved_sectors(super, super->disks);
3510
b9f594fe 3511 disk = &super->disks->disk;
5551b113 3512 info->data_offset = total_blocks(&super->disks->disk) - reserved;
14e8215b 3513 info->component_size = reserved;
25ed7e59 3514 info->disk.state = is_configured(disk) ? (1 << MD_DISK_ACTIVE) : 0;
df474657
DW		 3515 /* we don't change info->disk.raid_disk here because
3516 * this state will be finalized in mdmon after we have
3517 * found the 'most fresh' version of the metadata
3518 */
25ed7e59 3519 info->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0;
2432ce9b
AP		 3520 info->disk.state |= (is_spare(disk) || is_journal(disk)) ?
3521 0 : (1 << MD_DISK_SYNC);
cdddbdbc 3522 }
a575e2a7
DW		 3523
3524 /* only call uuid_from_super_imsm when this disk is part of a populated container,
3525 * ->compare_super may have updated the 'num_raid_devs' field for spares
3526 */
3527 if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs)
36ba7d48 3528 uuid_from_super_imsm(st, info->uuid);
22e263f6
AC		 3529 else
3530 memcpy(info->uuid, uuid_zero, sizeof(uuid_zero));
a5d85af7
N		 3531
3532 /* I don't know how to compute 'map' on imsm, so use safe default */
3533 if (map) {
3534 int i;
3535 for (i = 0; i < map_disks; i++)
3536 map[i] = 1;
3537 }
3538
cdddbdbc
DW		 3539}
3540
5c4cd5da
AC		 3541/* allocates memory and fills disk in mdinfo structure
3542 * for each disk in array */
3543struct mdinfo *getinfo_super_disks_imsm(struct supertype *st)
3544{
594dc1b8 3545 struct mdinfo *mddev;
5c4cd5da
AC		 3546 struct intel_super *super = st->sb;
3547 struct imsm_disk *disk;
3548 int count = 0;
3549 struct dl *dl;
3550 if (!super || !super->disks)
3551 return NULL;
3552 dl = super->disks;
503975b9 3553 mddev = xcalloc(1, sizeof(*mddev));
5c4cd5da
AC		 3554 while (dl) {
3555 struct mdinfo *tmp;
3556 disk = &dl->disk;
503975b9 3557 tmp = xcalloc(1, sizeof(*tmp));
5c4cd5da
AC		 3558 if (mddev->devs)
3559 tmp->next = mddev->devs;
3560 mddev->devs = tmp;
3561 tmp->disk.number = count++;
3562 tmp->disk.major = dl->major;
3563 tmp->disk.minor = dl->minor;
3564 tmp->disk.state = is_configured(disk) ?
3565 (1 << MD_DISK_ACTIVE) : 0;
3566 tmp->disk.state |= is_failed(disk) ? (1 << MD_DISK_FAULTY) : 0;
3567 tmp->disk.state |= is_spare(disk) ? 0 : (1 << MD_DISK_SYNC);
3568 tmp->disk.raid_disk = -1;
3569 dl = dl->next;
3570 }
3571 return mddev;
3572}
3573
cdddbdbc
DW		 3574static int update_super_imsm(struct supertype *st, struct mdinfo *info,
3575 char *update, char *devname, int verbose,
3576 int uuid_set, char *homehost)
3577{
f352c545
DW		 3578 /* For 'assemble' and 'force' we need to return non-zero if any
3579 * change was made. For others, the return value is ignored.
3580 * Update options are:
3581 * force-one : This device looks a bit old but needs to be included,
3582 * update age info appropriately.
3583 * assemble: clear any 'faulty' flag to allow this device to
3584 * be assembled.
3585 * force-array: Array is degraded but being forced, mark it clean
3586 * if that will be needed to assemble it.
3587 *
3588 * newdev: not used ????
3589 * grow: Array has gained a new device - this is currently for
3590 * linear only
3591 * resync: mark as dirty so a resync will happen.
3592 * name: update the name - preserving the homehost
6e46bf34 3593 * uuid: Change the uuid of the array to match watch is given
f352c545
DW		 3594 *
3595 * Following are not relevant for this imsm:
3596 * sparc2.2 : update from old dodgey metadata
3597 * super-minor: change the preferred_minor number
3598 * summaries: update redundant counters.
f352c545
DW		 3599 * homehost: update the recorded homehost
3600 * _reshape_progress: record new reshape_progress position.
3601 */
6e46bf34
DW		 3602 int rv = 1;
3603 struct intel_super *super = st->sb;
3604 struct imsm_super *mpb;
f352c545 3605
6e46bf34
DW		 3606 /* we can only update container info */
3607 if (!super || super->current_vol >= 0 || !super->anchor)
3608 return 1;
3609
3610 mpb = super->anchor;
3611
81a5b4f5
N		 3612 if (strcmp(update, "uuid") == 0) {
3613 /* We take this to mean that the family_num should be updated.
3614 * However that is much smaller than the uuid so we cannot really
3615 * allow an explicit uuid to be given. And it is hard to reliably
3616 * know if one was.
3617 * So if !uuid_set we know the current uuid is random and just used
3618 * the first 'int' and copy it to the other 3 positions.
3619 * Otherwise we require the 4 'int's to be the same as would be the
3620 * case if we are using a random uuid. So an explicit uuid will be
3621 * accepted as long as all for ints are the same... which shouldn't hurt
6e46bf34 3622 */
81a5b4f5
N		 3623 if (!uuid_set) {
3624 info->uuid[1] = info->uuid[2] = info->uuid[3] = info->uuid[0];
6e46bf34 3625 rv = 0;
81a5b4f5
N		 3626 } else {
3627 if (info->uuid[0] != info->uuid[1] ||
3628 info->uuid[1] != info->uuid[2] ||
3629 info->uuid[2] != info->uuid[3])
3630 rv = -1;
3631 else
3632 rv = 0;
6e46bf34 3633 }
81a5b4f5
N		 3634 if (rv == 0)
3635 mpb->orig_family_num = info->uuid[0];
6e46bf34
DW		 3636 } else if (strcmp(update, "assemble") == 0)
3637 rv = 0;
3638 else
1e2b2765 3639 rv = -1;
f352c545 3640
6e46bf34
DW		 3641 /* successful update? recompute checksum */
3642 if (rv == 0)
3643 mpb->check_sum = __le32_to_cpu(__gen_imsm_checksum(mpb));
f352c545
DW		 3644
3645 return rv;
cdddbdbc
DW		 3646}
3647
c2c087e6 3648static size_t disks_to_mpb_size(int disks)
cdddbdbc 3649{
c2c087e6 3650 size_t size;
cdddbdbc 3651
c2c087e6
DW		 3652 size = sizeof(struct imsm_super);
3653 size += (disks - 1) * sizeof(struct imsm_disk);
3654 size += 2 * sizeof(struct imsm_dev);
3655 /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */
3656 size += (4 - 2) * sizeof(struct imsm_map);
3657 /* 4 possible disk_ord_tbl's */
3658 size += 4 * (disks - 1) * sizeof(__u32);
bbab0940
TM		 3659 /* maximum bbm log */
3660 size += sizeof(struct bbm_log);
c2c087e6
DW		 3661
3662 return size;
3663}
3664
387fcd59
N		 3665static __u64 avail_size_imsm(struct supertype *st, __u64 devsize,
3666 unsigned long long data_offset)
c2c087e6
DW		 3667{
3668 if (devsize < (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS))
3669 return 0;
3670
3671 return devsize - (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS);
cdddbdbc
DW		 3672}
3673
ba2de7ba
DW		 3674static void free_devlist(struct intel_super *super)
3675{
3676 struct intel_dev *dv;
3677
3678 while (super->devlist) {
3679 dv = super->devlist->next;
3680 free(super->devlist->dev);
3681 free(super->devlist);
3682 super->devlist = dv;
3683 }
3684}
3685
3686static void imsm_copy_dev(struct imsm_dev *dest, struct imsm_dev *src)
3687{
3688 memcpy(dest, src, sizeof_imsm_dev(src, 0));
3689}
3690
cdddbdbc
DW		 3691static int compare_super_imsm(struct supertype *st, struct supertype *tst)
3692{
3693 /*
3694 * return:
3695 * 0 same, or first was empty, and second was copied
3696 * 1 second had wrong number
3697 * 2 wrong uuid
3698 * 3 wrong other info
3699 */
3700 struct intel_super *first = st->sb;
3701 struct intel_super *sec = tst->sb;
3702
5d500228
N		 3703 if (!first) {
3704 st->sb = tst->sb;
3705 tst->sb = NULL;
3706 return 0;
3707 }
8603ea6f
LM		 3708 /* in platform dependent environment test if the disks
3709 * use the same Intel hba
cb8f6859 3710 * If not on Intel hba at all, allow anything.
8603ea6f 3711 */
6b781d33
AP		 3712 if (!check_env("IMSM_NO_PLATFORM") && first->hba && sec->hba) {
3713 if (first->hba->type != sec->hba->type) {
8603ea6f 3714 fprintf(stderr,
6b781d33
AP		 3715 "HBAs of devices do not match %s != %s\n",
3716 get_sys_dev_type(first->hba->type),
3717 get_sys_dev_type(sec->hba->type));
3718 return 3;
3719 }
3720 if (first->orom != sec->orom) {
3721 fprintf(stderr,
3722 "HBAs of devices do not match %s != %s\n",
3723 first->hba->pci_id, sec->hba->pci_id);
8603ea6f
LM		 3724 return 3;
3725 }
3726 }
cdddbdbc 3727
d23fe947
DW		 3728 /* if an anchor does not have num_raid_devs set then it is a free
3729 * floating spare
3730 */
3731 if (first->anchor->num_raid_devs > 0 &&
3732 sec->anchor->num_raid_devs > 0) {
a2b97981
DW		 3733 /* Determine if these disks might ever have been
3734 * related. Further disambiguation can only take place
3735 * in load_super_imsm_all
3736 */
3737 __u32 first_family = first->anchor->orig_family_num;
3738 __u32 sec_family = sec->anchor->orig_family_num;
3739
f796af5d
DW		 3740 if (memcmp(first->anchor->sig, sec->anchor->sig,
3741 MAX_SIGNATURE_LENGTH) != 0)
3742 return 3;
3743
a2b97981
DW		 3744 if (first_family == 0)
3745 first_family = first->anchor->family_num;
3746 if (sec_family == 0)
3747 sec_family = sec->anchor->family_num;
3748
3749 if (first_family != sec_family)
d23fe947 3750 return 3;
f796af5d 3751
d23fe947 3752 }
cdddbdbc 3753
3e372e5a
DW		 3754 /* if 'first' is a spare promote it to a populated mpb with sec's
3755 * family number
3756 */
3757 if (first->anchor->num_raid_devs == 0 &&
3758 sec->anchor->num_raid_devs > 0) {
78d30f94 3759 int i;
ba2de7ba
DW		 3760 struct intel_dev *dv;
3761 struct imsm_dev *dev;
78d30f94
DW		 3762
3763 /* we need to copy raid device info from sec if an allocation
3764 * fails here we don't associate the spare
3765 */
3766 for (i = 0; i < sec->anchor->num_raid_devs; i++) {
503975b9
N		 3767 dv = xmalloc(sizeof(*dv));
3768 dev = xmalloc(sizeof_imsm_dev(get_imsm_dev(sec, i), 1));
ba2de7ba
DW		 3769 dv->dev = dev;
3770 dv->index = i;
3771 dv->next = first->devlist;
3772 first->devlist = dv;
78d30f94 3773 }
709743c5 3774 if (i < sec->anchor->num_raid_devs) {
ba2de7ba
DW		 3775 /* allocation failure */
3776 free_devlist(first);
e12b3daa 3777 pr_err("imsm: failed to associate spare\n");
ba2de7ba 3778 return 3;
78d30f94 3779 }
3e372e5a 3780 first->anchor->num_raid_devs = sec->anchor->num_raid_devs;
148acb7b 3781 first->anchor->orig_family_num = sec->anchor->orig_family_num;
3e372e5a 3782 first->anchor->family_num = sec->anchor->family_num;
ac6449be 3783 memcpy(first->anchor->sig, sec->anchor->sig, MAX_SIGNATURE_LENGTH);
709743c5
DW		 3784 for (i = 0; i < sec->anchor->num_raid_devs; i++)
3785 imsm_copy_dev(get_imsm_dev(first, i), get_imsm_dev(sec, i));
3e372e5a
DW		 3786 }
3787
cdddbdbc
DW		 3788 return 0;
3789}
3790
0030e8d6
DW		 3791static void fd2devname(int fd, char *name)
3792{
3793 struct stat st;
3794 char path[256];
33a6535d 3795 char dname[PATH_MAX];
0030e8d6
DW		 3796 char *nm;
3797 int rv;
3798
3799 name[0] = '\0';
3800 if (fstat(fd, &st) != 0)
3801 return;
3802 sprintf(path, "/sys/dev/block/%d:%d",
3803 major(st.st_rdev), minor(st.st_rdev));
3804
9cf014ec 3805 rv = readlink(path, dname, sizeof(dname)-1);
0030e8d6
DW		 3806 if (rv <= 0)
3807 return;
9587c373 3808
0030e8d6
DW		 3809 dname[rv] = '\0';
3810 nm = strrchr(dname, '/');
7897de29
JS		 3811 if (nm) {
3812 nm++;
3813 snprintf(name, MAX_RAID_SERIAL_LEN, "/dev/%s", nm);
3814 }
0030e8d6
DW		 3815}
3816
21e9380b
AP		 3817static int nvme_get_serial(int fd, void *buf, size_t buf_len)
3818{
3819 char path[60];
3820 char *name = fd2kname(fd);
3821
3822 if (!name)
3823 return 1;
3824
3825 if (strncmp(name, "nvme", 4) != 0)
3826 return 1;
3827
3828 snprintf(path, sizeof(path) - 1, "/sys/block/%s/device/serial", name);
3829
3830 return load_sys(path, buf, buf_len);
3831}
3832
cdddbdbc
DW		 3833extern int scsi_get_serial(int fd, void *buf, size_t buf_len);
3834
3835static int imsm_read_serial(int fd, char *devname,
3836 __u8 serial[MAX_RAID_SERIAL_LEN])
3837{
21e9380b 3838 char buf[50];
cdddbdbc 3839 int rv;
1f24f035 3840 int len;
316e2bf4
DW		 3841 char *dest;
3842 char *src;
21e9380b
AP		 3843 unsigned int i;
3844
3845 memset(buf, 0, sizeof(buf));
cdddbdbc 3846
21e9380b 3847 rv = nvme_get_serial(fd, buf, sizeof(buf));
cdddbdbc 3848
21e9380b
AP		 3849 if (rv)
3850 rv = scsi_get_serial(fd, buf, sizeof(buf));
f9ba0ff1 3851
40ebbb9c 3852 if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) {
f9ba0ff1
DW		 3853 memset(serial, 0, MAX_RAID_SERIAL_LEN);
3854 fd2devname(fd, (char *) serial);
0030e8d6
DW		 3855 return 0;
3856 }
3857
cdddbdbc
DW		 3858 if (rv != 0) {
3859 if (devname)
e7b84f9d
N		 3860 pr_err("Failed to retrieve serial for %s\n",
3861 devname);
cdddbdbc
DW		 3862 return rv;
3863 }
3864
316e2bf4
DW		 3865 /* trim all whitespace and non-printable characters and convert
3866 * ':' to ';'
3867 */
21e9380b
AP		 3868 for (i = 0, dest = buf; i < sizeof(buf) && buf[i]; i++) {
3869 src = &buf[i];
316e2bf4
DW		 3870 if (*src > 0x20) {
3871 /* ':' is reserved for use in placeholder serial
3872 * numbers for missing disks
3873 */
3874 if (*src == ':')
3875 *dest++ = ';';
3876 else
3877 *dest++ = *src;
3878 }
3879 }
21e9380b
AP		 3880 len = dest - buf;
3881 dest = buf;
316e2bf4
DW		 3882
3883 /* truncate leading characters */
3884 if (len > MAX_RAID_SERIAL_LEN) {
3885 dest += len - MAX_RAID_SERIAL_LEN;
1f24f035 3886 len = MAX_RAID_SERIAL_LEN;
316e2bf4 3887 }
5c3db629 3888
5c3db629 3889 memset(serial, 0, MAX_RAID_SERIAL_LEN);
316e2bf4 3890 memcpy(serial, dest, len);
cdddbdbc
DW		 3891
3892 return 0;
3893}
3894
1f24f035
DW		 3895static int serialcmp(__u8 *s1, __u8 *s2)
3896{
3897 return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN);
3898}
3899
3900static void serialcpy(__u8 *dest, __u8 *src)
3901{
3902 strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN);
3903}
3904
54c2c1ea
DW		 3905static struct dl *serial_to_dl(__u8 *serial, struct intel_super *super)
3906{
3907 struct dl *dl;
3908
3909 for (dl = super->disks; dl; dl = dl->next)
3910 if (serialcmp(dl->serial, serial) == 0)
3911 break;
3912
3913 return dl;
3914}
3915
a2b97981
DW		 3916static struct imsm_disk *
3917__serial_to_disk(__u8 *serial, struct imsm_super *mpb, int *idx)
3918{
3919 int i;
3920
3921 for (i = 0; i < mpb->num_disks; i++) {
3922 struct imsm_disk *disk = __get_imsm_disk(mpb, i);
3923
3924 if (serialcmp(disk->serial, serial) == 0) {
3925 if (idx)
3926 *idx = i;
3927 return disk;
3928 }
3929 }
3930
3931 return NULL;
3932}
3933
cdddbdbc
DW		 3934static int
3935load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd)
3936{
a2b97981 3937 struct imsm_disk *disk;
cdddbdbc
DW		 3938 struct dl *dl;
3939 struct stat stb;
cdddbdbc 3940 int rv;
a2b97981 3941 char name[40];
d23fe947
DW		 3942 __u8 serial[MAX_RAID_SERIAL_LEN];
3943
3944 rv = imsm_read_serial(fd, devname, serial);
3945
3946 if (rv != 0)
3947 return 2;
3948
503975b9 3949 dl = xcalloc(1, sizeof(*dl));
cdddbdbc 3950
a2b97981
DW		 3951 fstat(fd, &stb);
3952 dl->major = major(stb.st_rdev);
3953 dl->minor = minor(stb.st_rdev);
3954 dl->next = super->disks;
3955 dl->fd = keep_fd ? fd : -1;
3956 assert(super->disks == NULL);
3957 super->disks = dl;
3958 serialcpy(dl->serial, serial);
3959 dl->index = -2;
3960 dl->e = NULL;
3961 fd2devname(fd, name);
3962 if (devname)
503975b9 3963 dl->devname = xstrdup(devname);
a2b97981 3964 else
503975b9 3965 dl->devname = xstrdup(name);
cdddbdbc 3966
d23fe947 3967 /* look up this disk's index in the current anchor */
a2b97981
DW		 3968 disk = __serial_to_disk(dl->serial, super->anchor, &dl->index);
3969 if (disk) {
3970 dl->disk = *disk;
3971 /* only set index on disks that are a member of a
3972 * populated contianer, i.e. one with raid_devs
3973 */
3974 if (is_failed(&dl->disk))
3f6efecc 3975 dl->index = -2;
2432ce9b 3976 else if (is_spare(&dl->disk) || is_journal(&dl->disk))
a2b97981 3977 dl->index = -1;
3f6efecc
DW		 3978 }
3979
949c47a0
DW		 3980 return 0;
3981}
3982
0e600426 3983#ifndef MDASSEMBLE
0c046afd
DW		 3984/* When migrating map0 contains the 'destination' state while map1
3985 * contains the current state. When not migrating map0 contains the
3986 * current state. This routine assumes that map[0].map_state is set to
3987 * the current array state before being called.
3988 *
3989 * Migration is indicated by one of the following states
3990 * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed)
e3bba0e0 3991 * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal
0c046afd 3992 * map1state=unitialized)
1484e727 3993 * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal
0c046afd 3994 * map1state=normal)
e3bba0e0 3995 * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal
0c046afd 3996 * map1state=degraded)
8e59f3d8
AK		 3997 * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal
3998 * map1state=normal)
0c046afd 3999 */
8e59f3d8
AK		 4000static void migrate(struct imsm_dev *dev, struct intel_super *super,
4001 __u8 to_state, int migr_type)
3393c6af 4002{
0c046afd 4003 struct imsm_map *dest;
238c0a71 4004 struct imsm_map *src = get_imsm_map(dev, MAP_0);
3393c6af 4005
0c046afd 4006 dev->vol.migr_state = 1;
1484e727 4007 set_migr_type(dev, migr_type);
f8f603f1 4008 dev->vol.curr_migr_unit = 0;
238c0a71 4009 dest = get_imsm_map(dev, MAP_1);
0c046afd 4010
0556e1a2 4011 /* duplicate and then set the target end state in map[0] */
3393c6af 4012 memcpy(dest, src, sizeof_imsm_map(src));
089f9d79 4013 if (migr_type == MIGR_REBUILD || migr_type == MIGR_GEN_MIGR) {
0556e1a2
DW		 4014 __u32 ord;
4015 int i;
4016
4017 for (i = 0; i < src->num_members; i++) {
4018 ord = __le32_to_cpu(src->disk_ord_tbl[i]);
4019 set_imsm_ord_tbl_ent(src, i, ord_to_idx(ord));
4020 }
4021 }
4022
8e59f3d8
AK		 4023 if (migr_type == MIGR_GEN_MIGR)
4024 /* Clear migration record */
4025 memset(super->migr_rec, 0, sizeof(struct migr_record));
4026
0c046afd 4027 src->map_state = to_state;
949c47a0 4028}
f8f603f1 4029
809da78e
AK		 4030static void end_migration(struct imsm_dev *dev, struct intel_super *super,
4031 __u8 map_state)
f8f603f1 4032{
238c0a71
AK		 4033 struct imsm_map *map = get_imsm_map(dev, MAP_0);
4034 struct imsm_map *prev = get_imsm_map(dev, dev->vol.migr_state == 0 ?
4035 MAP_0 : MAP_1);
28bce06f 4036 int i, j;
0556e1a2
DW		 4037
4038 /* merge any IMSM_ORD_REBUILD bits that were not successfully
4039 * completed in the last migration.
4040 *
28bce06f 4041 * FIXME add support for raid-level-migration
0556e1a2 4042 */
089f9d79
JS		 4043 if (map_state != map->map_state && (is_gen_migration(dev) == 0) &&
4044 prev->map_state != IMSM_T_STATE_UNINITIALIZED) {
809da78e
AK		 4045 /* when final map state is other than expected
4046 * merge maps (not for migration)
4047 */
4048 int failed;
4049
4050 for (i = 0; i < prev->num_members; i++)
4051 for (j = 0; j < map->num_members; j++)
4052 /* during online capacity expansion
4053 * disks position can be changed
4054 * if takeover is used
4055 */
4056 if (ord_to_idx(map->disk_ord_tbl[j]) ==
4057 ord_to_idx(prev->disk_ord_tbl[i])) {
4058 map->disk_ord_tbl[j] |=
4059 prev->disk_ord_tbl[i];
4060 break;
4061 }
4062 failed = imsm_count_failed(super, dev, MAP_0);
4063 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
4064 }
f8f603f1
DW		 4065
4066 dev->vol.migr_state = 0;
ea672ee1 4067 set_migr_type(dev, 0);
f8f603f1
DW		 4068 dev->vol.curr_migr_unit = 0;
4069 map->map_state = map_state;
4070}
0e600426 4071#endif
949c47a0
DW		 4072
4073static int parse_raid_devices(struct intel_super *super)
4074{
4075 int i;
4076 struct imsm_dev *dev_new;
4d7b1503 4077 size_t len, len_migr;
401d313b 4078 size_t max_len = 0;
4d7b1503
DW		 4079 size_t space_needed = 0;
4080 struct imsm_super *mpb = super->anchor;
949c47a0
DW		 4081
4082 for (i = 0; i < super->anchor->num_raid_devs; i++) {
4083 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
ba2de7ba 4084 struct intel_dev *dv;
949c47a0 4085
4d7b1503
DW		 4086 len = sizeof_imsm_dev(dev_iter, 0);
4087 len_migr = sizeof_imsm_dev(dev_iter, 1);
4088 if (len_migr > len)
4089 space_needed += len_migr - len;
ca9de185 4090
503975b9 4091 dv = xmalloc(sizeof(*dv));
401d313b
AK		 4092 if (max_len < len_migr)
4093 max_len = len_migr;
4094 if (max_len > len_migr)
4095 space_needed += max_len - len_migr;
503975b9 4096 dev_new = xmalloc(max_len);
949c47a0 4097 imsm_copy_dev(dev_new, dev_iter);
ba2de7ba
DW		 4098 dv->dev = dev_new;
4099 dv->index = i;
4100 dv->next = super->devlist;
4101 super->devlist = dv;
949c47a0 4102 }
cdddbdbc 4103
4d7b1503
DW		 4104 /* ensure that super->buf is large enough when all raid devices
4105 * are migrating
4106 */
4107 if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) {
4108 void *buf;
4109
f36a9ecd
PB		 4110 len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed,
4111 super->sector_size);
4112 if (posix_memalign(&buf, MAX_SECTOR_SIZE, len) != 0)
4d7b1503
DW		 4113 return 1;
4114
1f45a8ad
DW		 4115 memcpy(buf, super->buf, super->len);
4116 memset(buf + super->len, 0, len - super->len);
4d7b1503
DW		 4117 free(super->buf);
4118 super->buf = buf;
4119 super->len = len;
4120 }
ca9de185 4121
bbab0940
TM		 4122 super->extra_space += space_needed;
4123
cdddbdbc
DW		 4124 return 0;
4125}
4126
e2f41b2c
AK		 4127/*******************************************************************************
4128 * Function: check_mpb_migr_compatibility
4129 * Description: Function checks for unsupported migration features:
4130 * - migration optimization area (pba_of_lba0)
4131 * - descending reshape (ascending_migr)
4132 * Parameters:
4133 * super : imsm metadata information
4134 * Returns:
4135 * 0 : migration is compatible
4136 * -1 : migration is not compatible
4137 ******************************************************************************/
4138int check_mpb_migr_compatibility(struct intel_super *super)
4139{
4140 struct imsm_map *map0, *map1;
4141 struct migr_record *migr_rec = super->migr_rec;
4142 int i;
4143
4144 for (i = 0; i < super->anchor->num_raid_devs; i++) {
4145 struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
4146
4147 if (dev_iter &&
4148 dev_iter->vol.migr_state == 1 &&
4149 dev_iter->vol.migr_type == MIGR_GEN_MIGR) {
4150 /* This device is migrating */
238c0a71
AK		 4151 map0 = get_imsm_map(dev_iter, MAP_0);
4152 map1 = get_imsm_map(dev_iter, MAP_1);
5551b113 4153 if (pba_of_lba0(map0) != pba_of_lba0(map1))
e2f41b2c
AK		 4154 /* migration optimization area was used */
4155 return -1;
4156 if (migr_rec->ascending_migr == 0
4157 && migr_rec->dest_depth_per_unit > 0)
4158 /* descending reshape not supported yet */
4159 return -1;
4160 }
4161 }
4162 return 0;
4163}
4164
d23fe947 4165static void __free_imsm(struct intel_super *super, int free_disks);
9ca2c81c 4166
cdddbdbc 4167/* load_imsm_mpb - read matrix metadata
f2f5c343 4168 * allocates super->mpb to be freed by free_imsm
cdddbdbc
DW		 4169 */
4170static int load_imsm_mpb(int fd, struct intel_super *super, char *devname)
4171{
4172 unsigned long long dsize;
cdddbdbc 4173 unsigned long long sectors;
f36a9ecd 4174 unsigned int sector_size = super->sector_size;
cdddbdbc 4175 struct stat;
6416d527 4176 struct imsm_super *anchor;
cdddbdbc
DW		 4177 __u32 check_sum;
4178
cdddbdbc 4179 get_dev_size(fd, NULL, &dsize);
f36a9ecd 4180 if (dsize < 2*sector_size) {
64436f06 4181 if (devname)
e7b84f9d
N		 4182 pr_err("%s: device to small for imsm\n",
4183 devname);
64436f06
N		 4184 return 1;
4185 }
cdddbdbc 4186
f36a9ecd 4187 if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0) {
cdddbdbc 4188 if (devname)
e7b84f9d
N		 4189 pr_err("Cannot seek to anchor block on %s: %s\n",
4190 devname, strerror(errno));
cdddbdbc
DW		 4191 return 1;
4192 }
4193
f36a9ecd 4194 if (posix_memalign((void **)&anchor, sector_size, sector_size) != 0) {
ad97895e 4195 if (devname)
7a862a02 4196 pr_err("Failed to allocate imsm anchor buffer on %s\n", devname);
ad97895e
DW		 4197 return 1;
4198 }
466070ad 4199 if ((unsigned int)read(fd, anchor, sector_size) != sector_size) {
cdddbdbc 4200 if (devname)
e7b84f9d
N		 4201 pr_err("Cannot read anchor block on %s: %s\n",
4202 devname, strerror(errno));
6416d527 4203 free(anchor);
cdddbdbc
DW		 4204 return 1;
4205 }
4206
6416d527 4207 if (strncmp((char *) anchor->sig, MPB_SIGNATURE, MPB_SIG_LEN) != 0) {
cdddbdbc 4208 if (devname)
e7b84f9d 4209 pr_err("no IMSM anchor on %s\n", devname);
6416d527 4210 free(anchor);
cdddbdbc
DW		 4211 return 2;
4212 }
4213
d23fe947 4214 __free_imsm(super, 0);
f2f5c343
LM		 4215 /* reload capability and hba */
4216
4217 /* capability and hba must be updated with new super allocation */
d424212e 4218 find_intel_hba_capability(fd, super, devname);
f36a9ecd
PB		 4219 super->len = ROUND_UP(anchor->mpb_size, sector_size);
4220 if (posix_memalign(&super->buf, MAX_SECTOR_SIZE, super->len) != 0) {
cdddbdbc 4221 if (devname)
e7b84f9d
N		 4222 pr_err("unable to allocate %zu byte mpb buffer\n",
4223 super->len);
6416d527 4224 free(anchor);
cdddbdbc
DW		 4225 return 2;
4226 }
f36a9ecd 4227 memcpy(super->buf, anchor, sector_size);
cdddbdbc 4228
f36a9ecd 4229 sectors = mpb_sectors(anchor, sector_size) - 1;
6416d527 4230 free(anchor);
8e59f3d8 4231
de44e46f
PB		 4232 if (posix_memalign(&super->migr_rec_buf, sector_size,
4233 MIGR_REC_BUF_SECTORS*sector_size) != 0) {
1ade5cc1 4234 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8
AK		 4235 free(super->buf);
4236 return 2;
4237 }
51d83f5d 4238 super->clean_migration_record_by_mdmon = 0;
8e59f3d8 4239
949c47a0 4240 if (!sectors) {
ecf45690
DW		 4241 check_sum = __gen_imsm_checksum(super->anchor);
4242 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
4243 if (devname)
e7b84f9d
N		 4244 pr_err("IMSM checksum %x != %x on %s\n",
4245 check_sum,
4246 __le32_to_cpu(super->anchor->check_sum),
4247 devname);
ecf45690
DW		 4248 return 2;
4249 }
4250
a2b97981 4251 return 0;
949c47a0 4252 }
cdddbdbc
DW		 4253
4254 /* read the extended mpb */
f36a9ecd 4255 if (lseek64(fd, dsize - (sector_size * (2 + sectors)), SEEK_SET) < 0) {
cdddbdbc 4256 if (devname)
e7b84f9d
N		 4257 pr_err("Cannot seek to extended mpb on %s: %s\n",
4258 devname, strerror(errno));
cdddbdbc
DW		 4259 return 1;
4260 }
4261
f36a9ecd
PB		 4262 if ((unsigned int)read(fd, super->buf + sector_size,
4263 super->len - sector_size) != super->len - sector_size) {
cdddbdbc 4264 if (devname)
e7b84f9d
N		 4265 pr_err("Cannot read extended mpb on %s: %s\n",
4266 devname, strerror(errno));
cdddbdbc
DW		 4267 return 2;
4268 }
4269
949c47a0
DW		 4270 check_sum = __gen_imsm_checksum(super->anchor);
4271 if (check_sum != __le32_to_cpu(super->anchor->check_sum)) {
cdddbdbc 4272 if (devname)
e7b84f9d
N		 4273 pr_err("IMSM checksum %x != %x on %s\n",
4274 check_sum, __le32_to_cpu(super->anchor->check_sum),
4275 devname);
db575f3b 4276 return 3;
cdddbdbc
DW		 4277 }
4278
a2b97981
DW		 4279 return 0;
4280}
4281
8e59f3d8
AK		 4282static int read_imsm_migr_rec(int fd, struct intel_super *super);
4283
97f81ee2
CA		 4284/* clears hi bits in metadata if MPB_ATTRIB_2TB_DISK not set */
4285static void clear_hi(struct intel_super *super)
4286{
4287 struct imsm_super *mpb = super->anchor;
4288 int i, n;
4289 if (mpb->attributes & MPB_ATTRIB_2TB_DISK)
4290 return;
4291 for (i = 0; i < mpb->num_disks; ++i) {
4292 struct imsm_disk *disk = &mpb->disk[i];
4293 disk->total_blocks_hi = 0;
4294 }
4295 for (i = 0; i < mpb->num_raid_devs; ++i) {
4296 struct imsm_dev *dev = get_imsm_dev(super, i);
4297 if (!dev)
4298 return;
4299 for (n = 0; n < 2; ++n) {
4300 struct imsm_map *map = get_imsm_map(dev, n);
4301 if (!map)
4302 continue;
4303 map->pba_of_lba0_hi = 0;
4304 map->blocks_per_member_hi = 0;
4305 map->num_data_stripes_hi = 0;
4306 }
4307 }
4308}
4309
a2b97981
DW		 4310static int
4311load_and_parse_mpb(int fd, struct intel_super *super, char *devname, int keep_fd)
4312{
4313 int err;
4314
4315 err = load_imsm_mpb(fd, super, devname);
4316 if (err)
4317 return err;
f36a9ecd
PB		 4318 if (super->sector_size == 4096)
4319 convert_from_4k(super);
a2b97981
DW		 4320 err = load_imsm_disk(fd, super, devname, keep_fd);
4321 if (err)
4322 return err;
4323 err = parse_raid_devices(super);
8d67477f
TM		 4324 if (err)
4325 return err;
4326 err = load_bbm_log(super);
97f81ee2 4327 clear_hi(super);
a2b97981 4328 return err;
cdddbdbc
DW		 4329}
4330
ae6aad82
DW		 4331static void __free_imsm_disk(struct dl *d)
4332{
4333 if (d->fd >= 0)
4334 close(d->fd);
4335 if (d->devname)
4336 free(d->devname);
0dcecb2e
DW		 4337 if (d->e)
4338 free(d->e);
ae6aad82
DW		 4339 free(d);
4340
4341}
1a64be56 4342
cdddbdbc
DW		 4343static void free_imsm_disks(struct intel_super *super)
4344{
47ee5a45 4345 struct dl *d;
cdddbdbc 4346
47ee5a45
DW		 4347 while (super->disks) {
4348 d = super->disks;
cdddbdbc 4349 super->disks = d->next;
ae6aad82 4350 __free_imsm_disk(d);
cdddbdbc 4351 }
cb82edca
AK		 4352 while (super->disk_mgmt_list) {
4353 d = super->disk_mgmt_list;
4354 super->disk_mgmt_list = d->next;
4355 __free_imsm_disk(d);
4356 }
47ee5a45
DW		 4357 while (super->missing) {
4358 d = super->missing;
4359 super->missing = d->next;
4360 __free_imsm_disk(d);
4361 }
4362
cdddbdbc
DW		 4363}
4364
9ca2c81c 4365/* free all the pieces hanging off of a super pointer */
d23fe947 4366static void __free_imsm(struct intel_super *super, int free_disks)
cdddbdbc 4367{
88654014
LM		 4368 struct intel_hba *elem, *next;
4369
9ca2c81c 4370 if (super->buf) {
949c47a0 4371 free(super->buf);
9ca2c81c
DW		 4372 super->buf = NULL;
4373 }
f2f5c343
LM		 4374 /* unlink capability description */
4375 super->orom = NULL;
8e59f3d8
AK		 4376 if (super->migr_rec_buf) {
4377 free(super->migr_rec_buf);
4378 super->migr_rec_buf = NULL;
4379 }
d23fe947
DW		 4380 if (free_disks)
4381 free_imsm_disks(super);
ba2de7ba 4382 free_devlist(super);
88654014
LM		 4383 elem = super->hba;
4384 while (elem) {
4385 if (elem->path)
4386 free((void *)elem->path);
4387 next = elem->next;
4388 free(elem);
4389 elem = next;
88c32bb1 4390 }
8d67477f
TM		 4391 if (super->bbm_log)
4392 free(super->bbm_log);
88654014 4393 super->hba = NULL;
cdddbdbc
DW		 4394}
4395
9ca2c81c
DW		 4396static void free_imsm(struct intel_super *super)
4397{
d23fe947 4398 __free_imsm(super, 1);
928f1424 4399 free(super->bb.entries);
9ca2c81c
DW		 4400 free(super);
4401}
cdddbdbc
DW		 4402
4403static void free_super_imsm(struct supertype *st)
4404{
4405 struct intel_super *super = st->sb;
4406
4407 if (!super)
4408 return;
4409
4410 free_imsm(super);
4411 st->sb = NULL;
4412}
4413
49133e57 4414static struct intel_super *alloc_super(void)
c2c087e6 4415{
503975b9 4416 struct intel_super *super = xcalloc(1, sizeof(*super));
c2c087e6 4417
503975b9
N		 4418 super->current_vol = -1;
4419 super->create_offset = ~((unsigned long long) 0);
928f1424
TM		 4420
4421 super->bb.entries = xmalloc(BBM_LOG_MAX_ENTRIES *
4422 sizeof(struct md_bb_entry));
4423 if (!super->bb.entries) {
4424 free(super);
4425 return NULL;
4426 }
4427
c2c087e6
DW		 4428 return super;
4429}
4430
f0f5a016
LM		 4431/*
4432 * find and allocate hba and OROM/EFI based on valid fd of RAID component device
4433 */
d424212e 4434static int find_intel_hba_capability(int fd, struct intel_super *super, char *devname)
f0f5a016
LM		 4435{
4436 struct sys_dev *hba_name;
4437 int rv = 0;
4438
089f9d79 4439 if (fd < 0 || check_env("IMSM_NO_PLATFORM")) {
f2f5c343 4440 super->orom = NULL;
f0f5a016
LM		 4441 super->hba = NULL;
4442 return 0;
4443 }
4444 hba_name = find_disk_attached_hba(fd, NULL);
4445 if (!hba_name) {
d424212e 4446 if (devname)
e7b84f9d
N		 4447 pr_err("%s is not attached to Intel(R) RAID controller.\n",
4448 devname);
f0f5a016
LM		 4449 return 1;
4450 }
4451 rv = attach_hba_to_super(super, hba_name);
4452 if (rv == 2) {
d424212e
N		 4453 if (devname) {
4454 struct intel_hba *hba = super->hba;
f0f5a016 4455
60f0f54d
PB		 4456 pr_err("%s is attached to Intel(R) %s %s (%s),\n"
4457 " but the container is assigned to Intel(R) %s %s (",
d424212e 4458 devname,
614902f6 4459 get_sys_dev_type(hba_name->type),
60f0f54d 4460 hba_name->type == SYS_DEV_VMD ? "domain" : "RAID controller",
f0f5a016 4461 hba_name->pci_id ? : "Err!",
60f0f54d
PB		 4462 get_sys_dev_type(super->hba->type),
4463 hba->type == SYS_DEV_VMD ? "domain" : "RAID controller");
f0f5a016 4464
f0f5a016
LM		 4465 while (hba) {
4466 fprintf(stderr, "%s", hba->pci_id ? : "Err!");
4467 if (hba->next)
4468 fprintf(stderr, ", ");
4469 hba = hba->next;
4470 }
6b781d33 4471 fprintf(stderr, ").\n"
60f0f54d
PB		 4472 " Mixing devices attached to different %s is not allowed.\n",
4473 hba_name->type == SYS_DEV_VMD ? "VMD domains" : "controllers");
f0f5a016 4474 }
f0f5a016
LM		 4475 return 2;
4476 }
6b781d33 4477 super->orom = find_imsm_capability(hba_name);
f2f5c343
LM		 4478 if (!super->orom)
4479 return 3;
614902f6 4480
f0f5a016
LM		 4481 return 0;
4482}
4483
47ee5a45
DW		 4484/* find_missing - helper routine for load_super_imsm_all that identifies
4485 * disks that have disappeared from the system. This routine relies on
4486 * the mpb being uptodate, which it is at load time.
4487 */
4488static int find_missing(struct intel_super *super)
4489{
4490 int i;
4491 struct imsm_super *mpb = super->anchor;
4492 struct dl *dl;
4493 struct imsm_disk *disk;
47ee5a45
DW		 4494
4495 for (i = 0; i < mpb->num_disks; i++) {
4496 disk = __get_imsm_disk(mpb, i);
54c2c1ea 4497 dl = serial_to_dl(disk->serial, super);
47ee5a45
DW		 4498 if (dl)
4499 continue;
47ee5a45 4500
503975b9 4501 dl = xmalloc(sizeof(*dl));
47ee5a45
DW		 4502 dl->major = 0;
4503 dl->minor = 0;
4504 dl->fd = -1;
503975b9 4505 dl->devname = xstrdup("missing");
47ee5a45
DW		 4506 dl->index = i;
4507 serialcpy(dl->serial, disk->serial);
4508 dl->disk = *disk;
689c9bf3 4509 dl->e = NULL;
47ee5a45
DW		 4510 dl->next = super->missing;
4511 super->missing = dl;
4512 }
4513
4514 return 0;
4515}
4516
3960e579 4517#ifndef MDASSEMBLE
a2b97981
DW		 4518static struct intel_disk *disk_list_get(__u8 *serial, struct intel_disk *disk_list)
4519{
4520 struct intel_disk *idisk = disk_list;
4521
4522 while (idisk) {
4523 if (serialcmp(idisk->disk.serial, serial) == 0)
4524 break;
4525 idisk = idisk->next;
4526 }
4527
4528 return idisk;
4529}
4530
4531static int __prep_thunderdome(struct intel_super **table, int tbl_size,
4532 struct intel_super *super,
4533 struct intel_disk **disk_list)
4534{
4535 struct imsm_disk *d = &super->disks->disk;
4536 struct imsm_super *mpb = super->anchor;
4537 int i, j;
4538
4539 for (i = 0; i < tbl_size; i++) {
4540 struct imsm_super *tbl_mpb = table[i]->anchor;
4541 struct imsm_disk *tbl_d = &table[i]->disks->disk;
4542
4543 if (tbl_mpb->family_num == mpb->family_num) {
4544 if (tbl_mpb->check_sum == mpb->check_sum) {
1ade5cc1
N		 4545 dprintf("mpb from %d:%d matches %d:%d\n",
4546 super->disks->major,
a2b97981
DW		 4547 super->disks->minor,
4548 table[i]->disks->major,
4549 table[i]->disks->minor);
4550 break;
4551 }
4552
4553 if (((is_configured(d) && !is_configured(tbl_d)) ||
4554 is_configured(d) == is_configured(tbl_d)) &&
4555 tbl_mpb->generation_num < mpb->generation_num) {
4556 /* current version of the mpb is a
4557 * better candidate than the one in
4558 * super_table, but copy over "cross
4559 * generational" status
4560 */
4561 struct intel_disk *idisk;
4562
1ade5cc1
N		 4563 dprintf("mpb from %d:%d replaces %d:%d\n",
4564 super->disks->major,
a2b97981
DW		 4565 super->disks->minor,
4566 table[i]->disks->major,
4567 table[i]->disks->minor);
4568
4569 idisk = disk_list_get(tbl_d->serial, *disk_list);
4570 if (idisk && is_failed(&idisk->disk))
4571 tbl_d->status |= FAILED_DISK;
4572 break;
4573 } else {
4574 struct intel_disk *idisk;
4575 struct imsm_disk *disk;
4576
4577 /* tbl_mpb is more up to date, but copy
4578 * over cross generational status before
4579 * returning
4580 */
4581 disk = __serial_to_disk(d->serial, mpb, NULL);
4582 if (disk && is_failed(disk))
4583 d->status |= FAILED_DISK;
4584
4585 idisk = disk_list_get(d->serial, *disk_list);
4586 if (idisk) {
4587 idisk->owner = i;
4588 if (disk && is_configured(disk))
4589 idisk->disk.status |= CONFIGURED_DISK;
4590 }
4591
1ade5cc1
N		 4592 dprintf("mpb from %d:%d prefer %d:%d\n",
4593 super->disks->major,
a2b97981
DW		 4594 super->disks->minor,
4595 table[i]->disks->major,
4596 table[i]->disks->minor);
4597
4598 return tbl_size;
4599 }
4600 }
4601 }
4602
4603 if (i >= tbl_size)
4604 table[tbl_size++] = super;
4605 else
4606 table[i] = super;
4607
4608 /* update/extend the merged list of imsm_disk records */
4609 for (j = 0; j < mpb->num_disks; j++) {
4610 struct imsm_disk *disk = __get_imsm_disk(mpb, j);
4611 struct intel_disk *idisk;
4612
4613 idisk = disk_list_get(disk->serial, *disk_list);
4614 if (idisk) {
4615 idisk->disk.status |= disk->status;
4616 if (is_configured(&idisk->disk) ||
4617 is_failed(&idisk->disk))
4618 idisk->disk.status &= ~(SPARE_DISK);
4619 } else {
503975b9 4620 idisk = xcalloc(1, sizeof(*idisk));
a2b97981
DW		 4621 idisk->owner = IMSM_UNKNOWN_OWNER;
4622 idisk->disk = *disk;
4623 idisk->next = *disk_list;
4624 *disk_list = idisk;
4625 }
4626
4627 if (serialcmp(idisk->disk.serial, d->serial) == 0)
4628 idisk->owner = i;
4629 }
4630
4631 return tbl_size;
4632}
4633
4634static struct intel_super *
4635validate_members(struct intel_super *super, struct intel_disk *disk_list,
4636 const int owner)
4637{
4638 struct imsm_super *mpb = super->anchor;
4639 int ok_count = 0;
4640 int i;
4641
4642 for (i = 0; i < mpb->num_disks; i++) {
4643 struct imsm_disk *disk = __get_imsm_disk(mpb, i);
4644 struct intel_disk *idisk;
4645
4646 idisk = disk_list_get(disk->serial, disk_list);
4647 if (idisk) {
4648 if (idisk->owner == owner ||
4649 idisk->owner == IMSM_UNKNOWN_OWNER)
4650 ok_count++;
4651 else
1ade5cc1
N		 4652 dprintf("'%.16s' owner %d != %d\n",
4653 disk->serial, idisk->owner,
a2b97981
DW		 4654 owner);
4655 } else {
1ade5cc1
N		 4656 dprintf("unknown disk %x [%d]: %.16s\n",
4657 __le32_to_cpu(mpb->family_num), i,
a2b97981
DW		 4658 disk->serial);
4659 break;
4660 }
4661 }
4662
4663 if (ok_count == mpb->num_disks)
4664 return super;
4665 return NULL;
4666}
4667
4668static void show_conflicts(__u32 family_num, struct intel_super *super_list)
4669{
4670 struct intel_super *s;
4671
4672 for (s = super_list; s; s = s->next) {
4673 if (family_num != s->anchor->family_num)
4674 continue;
e12b3daa 4675 pr_err("Conflict, offlining family %#x on '%s'\n",
a2b97981
DW		 4676 __le32_to_cpu(family_num), s->disks->devname);
4677 }
4678}
4679
4680static struct intel_super *
4681imsm_thunderdome(struct intel_super **super_list, int len)
4682{
4683 struct intel_super *super_table[len];
4684 struct intel_disk *disk_list = NULL;
4685 struct intel_super *champion, *spare;
4686 struct intel_super *s, **del;
4687 int tbl_size = 0;
4688 int conflict;
4689 int i;
4690
4691 memset(super_table, 0, sizeof(super_table));
4692 for (s = *super_list; s; s = s->next)
4693 tbl_size = __prep_thunderdome(super_table, tbl_size, s, &disk_list);
4694
4695 for (i = 0; i < tbl_size; i++) {
4696 struct imsm_disk *d;
4697 struct intel_disk *idisk;
4698 struct imsm_super *mpb = super_table[i]->anchor;
4699
4700 s = super_table[i];
4701 d = &s->disks->disk;
4702
4703 /* 'd' must appear in merged disk list for its
4704 * configuration to be valid
4705 */
4706 idisk = disk_list_get(d->serial, disk_list);
4707 if (idisk && idisk->owner == i)
4708 s = validate_members(s, disk_list, i);
4709 else
4710 s = NULL;
4711
4712 if (!s)
1ade5cc1
N		 4713 dprintf("marking family: %#x from %d:%d offline\n",
4714 mpb->family_num,
a2b97981
DW		 4715 super_table[i]->disks->major,
4716 super_table[i]->disks->minor);
4717 super_table[i] = s;
4718 }
4719
4720 /* This is where the mdadm implementation differs from the Windows
4721 * driver which has no strict concept of a container. We can only
4722 * assemble one family from a container, so when returning a prodigal
4723 * array member to this system the code will not be able to disambiguate
4724 * the container contents that should be assembled ("foreign" versus
4725 * "local"). It requires user intervention to set the orig_family_num
4726 * to a new value to establish a new container. The Windows driver in
4727 * this situation fixes up the volume name in place and manages the
4728 * foreign array as an independent entity.
4729 */
4730 s = NULL;
4731 spare = NULL;
4732 conflict = 0;
4733 for (i = 0; i < tbl_size; i++) {
4734 struct intel_super *tbl_ent = super_table[i];
4735 int is_spare = 0;
4736
4737 if (!tbl_ent)
4738 continue;
4739
4740 if (tbl_ent->anchor->num_raid_devs == 0) {
4741 spare = tbl_ent;
4742 is_spare = 1;
4743 }
4744
4745 if (s && !is_spare) {
4746 show_conflicts(tbl_ent->anchor->family_num, *super_list);
4747 conflict++;
4748 } else if (!s && !is_spare)
4749 s = tbl_ent;
4750 }
4751
4752 if (!s)
4753 s = spare;
4754 if (!s) {
4755 champion = NULL;
4756 goto out;
4757 }
4758 champion = s;
4759
4760 if (conflict)
7a862a02 4761 pr_err("Chose family %#x on '%s', assemble conflicts to new container with '--update=uuid'\n",
a2b97981
DW		 4762 __le32_to_cpu(s->anchor->family_num), s->disks->devname);
4763
4764 /* collect all dl's onto 'champion', and update them to
4765 * champion's version of the status
4766 */
4767 for (s = *super_list; s; s = s->next) {
4768 struct imsm_super *mpb = champion->anchor;
4769 struct dl *dl = s->disks;
4770
4771 if (s == champion)
4772 continue;
4773
5d7b407a
CA		 4774 mpb->attributes |= s->anchor->attributes & MPB_ATTRIB_2TB_DISK;
4775
a2b97981
DW		 4776 for (i = 0; i < mpb->num_disks; i++) {
4777 struct imsm_disk *disk;
4778
4779 disk = __serial_to_disk(dl->serial, mpb, &dl->index);
4780 if (disk) {
4781 dl->disk = *disk;
4782 /* only set index on disks that are a member of
4783 * a populated contianer, i.e. one with
4784 * raid_devs
4785 */
4786 if (is_failed(&dl->disk))
4787 dl->index = -2;
4788 else if (is_spare(&dl->disk))
4789 dl->index = -1;
4790 break;
4791 }
4792 }
4793
4794 if (i >= mpb->num_disks) {
4795 struct intel_disk *idisk;
4796
4797 idisk = disk_list_get(dl->serial, disk_list);
ecf408e9 4798 if (idisk && is_spare(&idisk->disk) &&
a2b97981
DW		 4799 !is_failed(&idisk->disk) && !is_configured(&idisk->disk))
4800 dl->index = -1;
4801 else {
4802 dl->index = -2;
4803 continue;
4804 }
4805 }
4806
4807 dl->next = champion->disks;
4808 champion->disks = dl;
4809 s->disks = NULL;
4810 }
4811
4812 /* delete 'champion' from super_list */
4813 for (del = super_list; *del; ) {
4814 if (*del == champion) {
4815 *del = (*del)->next;
4816 break;
4817 } else
4818 del = &(*del)->next;
4819 }
4820 champion->next = NULL;
4821
4822 out:
4823 while (disk_list) {
4824 struct intel_disk *idisk = disk_list;
4825
4826 disk_list = disk_list->next;
4827 free(idisk);
4828 }
4829
4830 return champion;
4831}
4832
9587c373
LM		 4833static int
4834get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd);
4dd2df09 4835static int get_super_block(struct intel_super **super_list, char *devnm, char *devname,
9587c373 4836 int major, int minor, int keep_fd);
ec50f7b6
LM		 4837static int
4838get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list,
4839 int *max, int keep_fd);
4840
cdddbdbc 4841static int load_super_imsm_all(struct supertype *st, int fd, void **sbp,
ec50f7b6
LM		 4842 char *devname, struct md_list *devlist,
4843 int keep_fd)
cdddbdbc 4844{
a2b97981
DW		 4845 struct intel_super *super_list = NULL;
4846 struct intel_super *super = NULL;
a2b97981 4847 int err = 0;
9587c373 4848 int i = 0;
dab4a513 4849
9587c373
LM		 4850 if (fd >= 0)
4851 /* 'fd' is an opened container */
4852 err = get_sra_super_block(fd, &super_list, devname, &i, keep_fd);
4853 else
ec50f7b6
LM		 4854 /* get super block from devlist devices */
4855 err = get_devlist_super_block(devlist, &super_list, &i, keep_fd);
9587c373 4856 if (err)
1602d52c 4857 goto error;
a2b97981
DW		 4858 /* all mpbs enter, maybe one leaves */
4859 super = imsm_thunderdome(&super_list, i);
4860 if (!super) {
4861 err = 1;
4862 goto error;
cdddbdbc
DW		 4863 }
4864
47ee5a45
DW		 4865 if (find_missing(super) != 0) {
4866 free_imsm(super);
a2b97981
DW		 4867 err = 2;
4868 goto error;
47ee5a45 4869 }
8e59f3d8
AK		 4870
4871 /* load migration record */
4872 err = load_imsm_migr_rec(super, NULL);
4c965cc9
AK		 4873 if (err == -1) {
4874 /* migration is in progress,
4875 * but migr_rec cannot be loaded,
4876 */
8e59f3d8
AK		 4877 err = 4;
4878 goto error;
4879 }
e2f41b2c
AK		 4880
4881 /* Check migration compatibility */
089f9d79 4882 if (err == 0 && check_mpb_migr_compatibility(super) != 0) {
e7b84f9d 4883 pr_err("Unsupported migration detected");
e2f41b2c
AK		 4884 if (devname)
4885 fprintf(stderr, " on %s\n", devname);
4886 else
4887 fprintf(stderr, " (IMSM).\n");
4888
4889 err = 5;
4890 goto error;
4891 }
4892
a2b97981
DW		 4893 err = 0;
4894
4895 error:
4896 while (super_list) {
4897 struct intel_super *s = super_list;
4898
4899 super_list = super_list->next;
4900 free_imsm(s);
4901 }
9587c373 4902
a2b97981
DW		 4903 if (err)
4904 return err;
f7e7067b 4905
cdddbdbc 4906 *sbp = super;
9587c373 4907 if (fd >= 0)
4dd2df09 4908 strcpy(st->container_devnm, fd2devnm(fd));
9587c373 4909 else
4dd2df09 4910 st->container_devnm[0] = 0;
a2b97981 4911 if (err == 0 && st->ss == NULL) {
bf5a934a 4912 st->ss = &super_imsm;
cdddbdbc
DW		 4913 st->minor_version = 0;
4914 st->max_devs = IMSM_MAX_DEVICES;
4915 }
cdddbdbc
DW		 4916 return 0;
4917}
2b959fbf 4918
ec50f7b6
LM		 4919static int
4920get_devlist_super_block(struct md_list *devlist, struct intel_super **super_list,
4921 int *max, int keep_fd)
4922{
4923 struct md_list *tmpdev;
4924 int err = 0;
4925 int i = 0;
9587c373 4926
ec50f7b6
LM		 4927 for (i = 0, tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
4928 if (tmpdev->used != 1)
4929 continue;
4930 if (tmpdev->container == 1) {
ca9de185 4931 int lmax = 0;
ec50f7b6
LM		 4932 int fd = dev_open(tmpdev->devname, O_RDONLY|O_EXCL);
4933 if (fd < 0) {
e7b84f9d 4934 pr_err("cannot open device %s: %s\n",
ec50f7b6
LM		 4935 tmpdev->devname, strerror(errno));
4936 err = 8;
4937 goto error;
4938 }
4939 err = get_sra_super_block(fd, super_list,
4940 tmpdev->devname, &lmax,
4941 keep_fd);
4942 i += lmax;
4943 close(fd);
4944 if (err) {
4945 err = 7;
4946 goto error;
4947 }
4948 } else {
4949 int major = major(tmpdev->st_rdev);
4950 int minor = minor(tmpdev->st_rdev);
4951 err = get_super_block(super_list,
4dd2df09 4952 NULL,
ec50f7b6
LM		 4953 tmpdev->devname,
4954 major, minor,
4955 keep_fd);
4956 i++;
4957 if (err) {
4958 err = 6;
4959 goto error;
4960 }
4961 }
4962 }
4963 error:
4964 *max = i;
4965 return err;
4966}
9587c373 4967
4dd2df09 4968static int get_super_block(struct intel_super **super_list, char *devnm, char *devname,
9587c373
LM		 4969 int major, int minor, int keep_fd)
4970{
594dc1b8 4971 struct intel_super *s;
9587c373
LM		 4972 char nm[32];
4973 int dfd = -1;
9587c373
LM		 4974 int err = 0;
4975 int retry;
4976
4977 s = alloc_super();
4978 if (!s) {
4979 err = 1;
4980 goto error;
4981 }
4982
4983 sprintf(nm, "%d:%d", major, minor);
4984 dfd = dev_open(nm, O_RDWR);
4985 if (dfd < 0) {
4986 err = 2;
4987 goto error;
4988 }
4989
fa7bb6f8 4990 get_dev_sector_size(dfd, NULL, &s->sector_size);
cb8f6859 4991 find_intel_hba_capability(dfd, s, devname);
9587c373
LM		 4992 err = load_and_parse_mpb(dfd, s, NULL, keep_fd);
4993
4994 /* retry the load if we might have raced against mdmon */
4dd2df09 4995 if (err == 3 && devnm && mdmon_running(devnm))
9587c373
LM		 4996 for (retry = 0; retry < 3; retry++) {
4997 usleep(3000);
4998 err = load_and_parse_mpb(dfd, s, NULL, keep_fd);
4999 if (err != 3)
5000 break;
5001 }
5002 error:
5003 if (!err) {
5004 s->next = *super_list;
5005 *super_list = s;
5006 } else {
5007 if (s)
8d67477f 5008 free_imsm(s);
36614e95 5009 if (dfd >= 0)
9587c373
LM		 5010 close(dfd);
5011 }
089f9d79 5012 if (dfd >= 0 && !keep_fd)
9587c373
LM		 5013 close(dfd);
5014 return err;
5015
5016}
5017
5018static int
5019get_sra_super_block(int fd, struct intel_super **super_list, char *devname, int *max, int keep_fd)
5020{
5021 struct mdinfo *sra;
4dd2df09 5022 char *devnm;
9587c373
LM		 5023 struct mdinfo *sd;
5024 int err = 0;
5025 int i = 0;
4dd2df09 5026 sra = sysfs_read(fd, NULL, GET_LEVEL|GET_VERSION|GET_DEVS|GET_STATE);
9587c373
LM		 5027 if (!sra)
5028 return 1;
5029
5030 if (sra->array.major_version != -1 ||
5031 sra->array.minor_version != -2 ||
5032 strcmp(sra->text_version, "imsm") != 0) {
5033 err = 1;
5034 goto error;
5035 }
5036 /* load all mpbs */
4dd2df09 5037 devnm = fd2devnm(fd);
9587c373 5038 for (sd = sra->devs, i = 0; sd; sd = sd->next, i++) {
4dd2df09 5039 if (get_super_block(super_list, devnm, devname,
9587c373
LM		 5040 sd->disk.major, sd->disk.minor, keep_fd) != 0) {
5041 err = 7;
5042 goto error;
5043 }
5044 }
5045 error:
5046 sysfs_free(sra);
5047 *max = i;
5048 return err;
5049}
5050
2b959fbf
N		 5051static int load_container_imsm(struct supertype *st, int fd, char *devname)
5052{
ec50f7b6 5053 return load_super_imsm_all(st, fd, &st->sb, devname, NULL, 1);
2b959fbf 5054}
cdddbdbc
DW		 5055#endif
5056
5057static int load_super_imsm(struct supertype *st, int fd, char *devname)
5058{
5059 struct intel_super *super;
5060 int rv;
8a3544f8 5061 int retry;
cdddbdbc 5062
357ac106 5063 if (test_partition(fd))
691c6ee1
N		 5064 /* IMSM not allowed on partitions */
5065 return 1;
5066
37424f13
DW		 5067 free_super_imsm(st);
5068
49133e57 5069 super = alloc_super();
fa7bb6f8 5070 get_dev_sector_size(fd, NULL, &super->sector_size);
8d67477f
TM		 5071 if (!super)
5072 return 1;
ea2bc72b
LM		 5073 /* Load hba and capabilities if they exist.
5074 * But do not preclude loading metadata in case capabilities or hba are
5075 * non-compliant and ignore_hw_compat is set.
5076 */
d424212e 5077 rv = find_intel_hba_capability(fd, super, devname);
f2f5c343 5078 /* no orom/efi or non-intel hba of the disk */
089f9d79 5079 if (rv != 0 && st->ignore_hw_compat == 0) {
f2f5c343 5080 if (devname)
e7b84f9d 5081 pr_err("No OROM/EFI properties for %s\n", devname);
f2f5c343
LM		 5082 free_imsm(super);
5083 return 2;
5084 }
a2b97981 5085 rv = load_and_parse_mpb(fd, super, devname, 0);
cdddbdbc 5086
8a3544f8
AP		 5087 /* retry the load if we might have raced against mdmon */
5088 if (rv == 3) {
f96b1302
AP		 5089 struct mdstat_ent *mdstat = NULL;
5090 char *name = fd2kname(fd);
5091
5092 if (name)
5093 mdstat = mdstat_by_component(name);
8a3544f8
AP		 5094
5095 if (mdstat && mdmon_running(mdstat->devnm) && getpid() != mdmon_pid(mdstat->devnm)) {
5096 for (retry = 0; retry < 3; retry++) {
5097 usleep(3000);
5098 rv = load_and_parse_mpb(fd, super, devname, 0);
5099 if (rv != 3)
5100 break;
5101 }
5102 }
5103
5104 free_mdstat(mdstat);
5105 }
5106
cdddbdbc
DW		 5107 if (rv) {
5108 if (devname)
7a862a02 5109 pr_err("Failed to load all information sections on %s\n", devname);
cdddbdbc
DW		 5110 free_imsm(super);
5111 return rv;
5112 }
5113
5114 st->sb = super;
5115 if (st->ss == NULL) {
5116 st->ss = &super_imsm;
5117 st->minor_version = 0;
5118 st->max_devs = IMSM_MAX_DEVICES;
5119 }
8e59f3d8
AK		 5120
5121 /* load migration record */
2e062e82
AK		 5122 if (load_imsm_migr_rec(super, NULL) == 0) {
5123 /* Check for unsupported migration features */
5124 if (check_mpb_migr_compatibility(super) != 0) {
e7b84f9d 5125 pr_err("Unsupported migration detected");
2e062e82
AK		 5126 if (devname)
5127 fprintf(stderr, " on %s\n", devname);
5128 else
5129 fprintf(stderr, " (IMSM).\n");
5130 return 3;
5131 }
e2f41b2c
AK		 5132 }
5133
cdddbdbc
DW		 5134 return 0;
5135}
5136
ef6ffade
DW		 5137static __u16 info_to_blocks_per_strip(mdu_array_info_t *info)
5138{
5139 if (info->level == 1)
5140 return 128;
5141 return info->chunk_size >> 9;
5142}
5143
5551b113
CA		 5144static unsigned long long info_to_blocks_per_member(mdu_array_info_t *info,
5145 unsigned long long size)
fcfd9599 5146{
4025c288 5147 if (info->level == 1)
5551b113 5148 return size * 2;
4025c288 5149 else
5551b113 5150 return (size * 2) & ~(info_to_blocks_per_strip(info) - 1);
fcfd9599
DW		 5151}
5152
4d1313e9
DW		 5153static void imsm_update_version_info(struct intel_super *super)
5154{
5155 /* update the version and attributes */
5156 struct imsm_super *mpb = super->anchor;
5157 char *version;
5158 struct imsm_dev *dev;
5159 struct imsm_map *map;
5160 int i;
5161
5162 for (i = 0; i < mpb->num_raid_devs; i++) {
5163 dev = get_imsm_dev(super, i);
238c0a71 5164 map = get_imsm_map(dev, MAP_0);
4d1313e9
DW		 5165 if (__le32_to_cpu(dev->size_high) > 0)
5166 mpb->attributes |= MPB_ATTRIB_2TB;
5167
5168 /* FIXME detect when an array spans a port multiplier */
5169 #if 0
5170 mpb->attributes |= MPB_ATTRIB_PM;
5171 #endif
5172
5173 if (mpb->num_raid_devs > 1 ||
5174 mpb->attributes != MPB_ATTRIB_CHECKSUM_VERIFY) {
5175 version = MPB_VERSION_ATTRIBS;
5176 switch (get_imsm_raid_level(map)) {
5177 case 0: mpb->attributes |= MPB_ATTRIB_RAID0; break;
5178 case 1: mpb->attributes |= MPB_ATTRIB_RAID1; break;
5179 case 10: mpb->attributes |= MPB_ATTRIB_RAID10; break;
5180 case 5: mpb->attributes |= MPB_ATTRIB_RAID5; break;
5181 }
5182 } else {
5183 if (map->num_members >= 5)
5184 version = MPB_VERSION_5OR6_DISK_ARRAY;
5185 else if (dev->status == DEV_CLONE_N_GO)
5186 version = MPB_VERSION_CNG;
5187 else if (get_imsm_raid_level(map) == 5)
5188 version = MPB_VERSION_RAID5;
5189 else if (map->num_members >= 3)
5190 version = MPB_VERSION_3OR4_DISK_ARRAY;
5191 else if (get_imsm_raid_level(map) == 1)
5192 version = MPB_VERSION_RAID1;
5193 else
5194 version = MPB_VERSION_RAID0;
5195 }
5196 strcpy(((char *) mpb->sig) + strlen(MPB_SIGNATURE), version);
5197 }
5198}
5199
aa534678
DW		 5200static int check_name(struct intel_super *super, char *name, int quiet)
5201{
5202 struct imsm_super *mpb = super->anchor;
5203 char *reason = NULL;
5204 int i;
5205
5206 if (strlen(name) > MAX_RAID_SERIAL_LEN)
5207 reason = "must be 16 characters or less";
5208
5209 for (i = 0; i < mpb->num_raid_devs; i++) {
5210 struct imsm_dev *dev = get_imsm_dev(super, i);
5211
5212 if (strncmp((char *) dev->volume, name, MAX_RAID_SERIAL_LEN) == 0) {
5213 reason = "already exists";
5214 break;
5215 }
5216 }
5217
5218 if (reason && !quiet)
e7b84f9d 5219 pr_err("imsm volume name %s\n", reason);
aa534678
DW		 5220
5221 return !reason;
5222}
5223
8b353278 5224static int init_super_imsm_volume(struct supertype *st, mdu_array_info_t *info,
5308f117 5225 struct shape *s, char *name,
83cd1e97
N		 5226 char *homehost, int *uuid,
5227 long long data_offset)
cdddbdbc 5228{
c2c087e6
DW		 5229 /* We are creating a volume inside a pre-existing container.
5230 * so st->sb is already set.
5231 */
5232 struct intel_super *super = st->sb;
f36a9ecd 5233 unsigned int sector_size = super->sector_size;
949c47a0 5234 struct imsm_super *mpb = super->anchor;
ba2de7ba 5235 struct intel_dev *dv;
c2c087e6
DW		 5236 struct imsm_dev *dev;
5237 struct imsm_vol *vol;
5238 struct imsm_map *map;
5239 int idx = mpb->num_raid_devs;
5240 int i;
5241 unsigned long long array_blocks;
2c092cad 5242 size_t size_old, size_new;
5551b113 5243 unsigned long long num_data_stripes;
cdddbdbc 5244
88c32bb1 5245 if (super->orom && mpb->num_raid_devs >= super->orom->vpa) {
7a862a02 5246 pr_err("This imsm-container already has the maximum of %d volumes\n", super->orom->vpa);
c2c087e6
DW		 5247 return 0;
5248 }
5249
2c092cad
DW		 5250 /* ensure the mpb is large enough for the new data */
5251 size_old = __le32_to_cpu(mpb->mpb_size);
5252 size_new = disks_to_mpb_size(info->nr_disks);
5253 if (size_new > size_old) {
5254 void *mpb_new;
f36a9ecd 5255 size_t size_round = ROUND_UP(size_new, sector_size);
2c092cad 5256
f36a9ecd 5257 if (posix_memalign(&mpb_new, sector_size, size_round) != 0) {
e7b84f9d 5258 pr_err("could not allocate new mpb\n");
2c092cad
DW		 5259 return 0;
5260 }
de44e46f
PB		 5261 if (posix_memalign(&super->migr_rec_buf, sector_size,
5262 MIGR_REC_BUF_SECTORS*sector_size) != 0) {
1ade5cc1 5263 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8
AK		 5264 free(super->buf);
5265 free(super);
ea944c8f 5266 free(mpb_new);
8e59f3d8
AK		 5267 return 0;
5268 }
2c092cad
DW		 5269 memcpy(mpb_new, mpb, size_old);
5270 free(mpb);
5271 mpb = mpb_new;
949c47a0 5272 super->anchor = mpb_new;
2c092cad
DW		 5273 mpb->mpb_size = __cpu_to_le32(size_new);
5274 memset(mpb_new + size_old, 0, size_round - size_old);
bbab0940 5275 super->len = size_round;
2c092cad 5276 }
bf5a934a 5277 super->current_vol = idx;
3960e579
DW		 5278
5279 /* handle 'failed_disks' by either:
5280 * a) create dummy disk entries in the table if this the first
5281 * volume in the array. We add them here as this is the only
5282 * opportunity to add them. add_to_super_imsm_volume()
5283 * handles the non-failed disks and continues incrementing
5284 * mpb->num_disks.
5285 * b) validate that 'failed_disks' matches the current number
5286 * of missing disks if the container is populated
d23fe947 5287 */
3960e579 5288 if (super->current_vol == 0) {
d23fe947 5289 mpb->num_disks = 0;
3960e579
DW		 5290 for (i = 0; i < info->failed_disks; i++) {
5291 struct imsm_disk *disk;
5292
5293 mpb->num_disks++;
5294 disk = __get_imsm_disk(mpb, i);
5295 disk->status = CONFIGURED_DISK | FAILED_DISK;
5296 disk->scsi_id = __cpu_to_le32(~(__u32)0);
5297 snprintf((char *) disk->serial, MAX_RAID_SERIAL_LEN,
5b975129 5298 "missing:%d", (__u8)i);
3960e579
DW		 5299 }
5300 find_missing(super);
5301 } else {
5302 int missing = 0;
5303 struct dl *d;
5304
5305 for (d = super->missing; d; d = d->next)
5306 missing++;
5307 if (info->failed_disks > missing) {
e7b84f9d 5308 pr_err("unable to add 'missing' disk to container\n");
3960e579
DW		 5309 return 0;
5310 }
5311 }
5a038140 5312
aa534678
DW		 5313 if (!check_name(super, name, 0))
5314 return 0;
503975b9
N		 5315 dv = xmalloc(sizeof(*dv));
5316 dev = xcalloc(1, sizeof(*dev) + sizeof(__u32) * (info->raid_disks - 1));
c2c087e6 5317 strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN);
e03640bd 5318 array_blocks = calc_array_size(info->level, info->raid_disks,
03bcbc65 5319 info->layout, info->chunk_size,
5308f117 5320 s->size * 2);
979d38be
DW		 5321 /* round array size down to closest MB */
5322 array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT;
5323
c2c087e6
DW		 5324 dev->size_low = __cpu_to_le32((__u32) array_blocks);
5325 dev->size_high = __cpu_to_le32((__u32) (array_blocks >> 32));
1a2487c2 5326 dev->status = (DEV_READ_COALESCING | DEV_WRITE_COALESCING);
c2c087e6
DW		 5327 vol = &dev->vol;
5328 vol->migr_state = 0;
1484e727 5329 set_migr_type(dev, MIGR_INIT);
3960e579 5330 vol->dirty = !info->state;
f8f603f1 5331 vol->curr_migr_unit = 0;
238c0a71 5332 map = get_imsm_map(dev, MAP_0);
5551b113 5333 set_pba_of_lba0(map, super->create_offset);
5308f117 5334 set_blocks_per_member(map, info_to_blocks_per_member(info, s->size));
ef6ffade 5335 map->blocks_per_strip = __cpu_to_le16(info_to_blocks_per_strip(info));
0556e1a2 5336 map->failed_disk_num = ~0;
bf4442ab 5337 if (info->level > 0)
fffaf1ff
N		 5338 map->map_state = (info->state ? IMSM_T_STATE_NORMAL
5339 : IMSM_T_STATE_UNINITIALIZED);
bf4442ab
AK		 5340 else
5341 map->map_state = info->failed_disks ? IMSM_T_STATE_FAILED :
5342 IMSM_T_STATE_NORMAL;
252d23c0 5343 map->ddf = 1;
ef6ffade
DW		 5344
5345 if (info->level == 1 && info->raid_disks > 2) {
38950822
AW		 5346 free(dev);
5347 free(dv);
7a862a02 5348 pr_err("imsm does not support more than 2 disksin a raid1 volume\n");
ef6ffade
DW		 5349 return 0;
5350 }
81062a36
DW		 5351
5352 map->raid_level = info->level;
4d1313e9 5353 if (info->level == 10) {
c2c087e6 5354 map->raid_level = 1;
4d1313e9 5355 map->num_domains = info->raid_disks / 2;
81062a36
DW		 5356 } else if (info->level == 1)
5357 map->num_domains = info->raid_disks;
5358 else
ff596308 5359 map->num_domains = 1;
81062a36 5360
5551b113 5361 /* info->size is only int so use the 'size' parameter instead */
5308f117 5362 num_data_stripes = (s->size * 2) / info_to_blocks_per_strip(info);
5551b113
CA		 5363 num_data_stripes /= map->num_domains;
5364 set_num_data_stripes(map, num_data_stripes);
ef6ffade 5365
c2c087e6
DW		 5366 map->num_members = info->raid_disks;
5367 for (i = 0; i < map->num_members; i++) {
5368 /* initialized in add_to_super */
4eb26970 5369 set_imsm_ord_tbl_ent(map, i, IMSM_ORD_REBUILD);
c2c087e6 5370 }
949c47a0 5371 mpb->num_raid_devs++;
ba2de7ba 5372
2432ce9b
AP		 5373 if (s->consistency_policy == UnSet ||
5374 s->consistency_policy == CONSISTENCY_POLICY_RESYNC ||
5375 s->consistency_policy == CONSISTENCY_POLICY_NONE) {
5376 dev->rwh_policy = RWH_OFF;
5377 } else if (s->consistency_policy == CONSISTENCY_POLICY_PPL) {
5378 dev->rwh_policy = RWH_DISTRIBUTED;
5379 } else {
5380 free(dev);
5381 free(dv);
5382 pr_err("imsm does not support consistency policy %s\n",
5383 map_num(consistency_policies, s->consistency_policy));
5384 return 0;
5385 }
5386
ba2de7ba
DW		 5387 dv->dev = dev;
5388 dv->index = super->current_vol;
5389 dv->next = super->devlist;
5390 super->devlist = dv;
c2c087e6 5391
4d1313e9
DW		 5392 imsm_update_version_info(super);
5393
c2c087e6 5394 return 1;
cdddbdbc
DW		 5395}
5396
bf5a934a 5397static int init_super_imsm(struct supertype *st, mdu_array_info_t *info,
5308f117 5398 struct shape *s, char *name,
83cd1e97
N		 5399 char *homehost, int *uuid,
5400 unsigned long long data_offset)
bf5a934a
DW		 5401{
5402 /* This is primarily called by Create when creating a new array.
5403 * We will then get add_to_super called for each component, and then
5404 * write_init_super called to write it out to each device.
5405 * For IMSM, Create can create on fresh devices or on a pre-existing
5406 * array.
5407 * To create on a pre-existing array a different method will be called.
5408 * This one is just for fresh drives.
5409 */
5410 struct intel_super *super;
5411 struct imsm_super *mpb;
5412 size_t mpb_size;
4d1313e9 5413 char *version;
bf5a934a 5414
83cd1e97 5415 if (data_offset != INVALID_SECTORS) {
ed503f89 5416 pr_err("data-offset not supported by imsm\n");
83cd1e97
N		 5417 return 0;
5418 }
5419
bf5a934a 5420 if (st->sb)
5308f117 5421 return init_super_imsm_volume(st, info, s, name, homehost, uuid,
83cd1e97 5422 data_offset);
e683ca88
DW		 5423
5424 if (info)
5425 mpb_size = disks_to_mpb_size(info->nr_disks);
5426 else
f36a9ecd 5427 mpb_size = MAX_SECTOR_SIZE;
bf5a934a 5428
49133e57 5429 super = alloc_super();
f36a9ecd
PB		 5430 if (super &&
5431 posix_memalign(&super->buf, MAX_SECTOR_SIZE, mpb_size) != 0) {
8d67477f 5432 free_imsm(super);
e683ca88
DW		 5433 super = NULL;
5434 }
5435 if (!super) {
1ade5cc1 5436 pr_err("could not allocate superblock\n");
bf5a934a
DW		 5437 return 0;
5438 }
de44e46f
PB		 5439 if (posix_memalign(&super->migr_rec_buf, MAX_SECTOR_SIZE,
5440 MIGR_REC_BUF_SECTORS*MAX_SECTOR_SIZE) != 0) {
1ade5cc1 5441 pr_err("could not allocate migr_rec buffer\n");
8e59f3d8 5442 free(super->buf);
8d67477f 5443 free_imsm(super);
8e59f3d8
AK		 5444 return 0;
5445 }
e683ca88 5446 memset(super->buf, 0, mpb_size);
ef649044 5447 mpb = super->buf;
e683ca88
DW		 5448 mpb->mpb_size = __cpu_to_le32(mpb_size);
5449 st->sb = super;
5450
5451 if (info == NULL) {
5452 /* zeroing superblock */
5453 return 0;
5454 }
bf5a934a 5455
4d1313e9
DW		 5456 mpb->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
5457
5458 version = (char *) mpb->sig;
5459 strcpy(version, MPB_SIGNATURE);
5460 version += strlen(MPB_SIGNATURE);
5461 strcpy(version, MPB_VERSION_RAID0);
bf5a934a 5462
bf5a934a
DW		 5463 return 1;
5464}
5465
0e600426 5466#ifndef MDASSEMBLE
f20c3968 5467static int add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk,
bf5a934a
DW		 5468 int fd, char *devname)
5469{
5470 struct intel_super *super = st->sb;
d23fe947 5471 struct imsm_super *mpb = super->anchor;
3960e579 5472 struct imsm_disk *_disk;
bf5a934a
DW		 5473 struct imsm_dev *dev;
5474 struct imsm_map *map;
3960e579 5475 struct dl *dl, *df;
4eb26970 5476 int slot;
bf5a934a 5477
949c47a0 5478 dev = get_imsm_dev(super, super->current_vol);
238c0a71 5479 map = get_imsm_map(dev, MAP_0);
bf5a934a 5480
208933a7 5481 if (! (dk->state & (1<<MD_DISK_SYNC))) {
e7b84f9d 5482 pr_err("%s: Cannot add spare devices to IMSM volume\n",
208933a7
N		 5483 devname);
5484 return 1;
5485 }
5486
efb30e7f
DW		 5487 if (fd == -1) {
5488 /* we're doing autolayout so grab the pre-marked (in
5489 * validate_geometry) raid_disk
5490 */
5491 for (dl = super->disks; dl; dl = dl->next)
5492 if (dl->raiddisk == dk->raid_disk)
5493 break;
5494 } else {
5495 for (dl = super->disks; dl ; dl = dl->next)
5496 if (dl->major == dk->major &&
5497 dl->minor == dk->minor)
5498 break;
5499 }
d23fe947 5500
208933a7 5501 if (!dl) {
e7b84f9d 5502 pr_err("%s is not a member of the same container\n", devname);
f20c3968 5503 return 1;
208933a7 5504 }
bf5a934a 5505
d23fe947
DW		 5506 /* add a pristine spare to the metadata */
5507 if (dl->index < 0) {
5508 dl->index = super->anchor->num_disks;
5509 super->anchor->num_disks++;
5510 }
4eb26970
DW		 5511 /* Check the device has not already been added */
5512 slot = get_imsm_disk_slot(map, dl->index);
5513 if (slot >= 0 &&
238c0a71 5514 (get_imsm_ord_tbl_ent(dev, slot, MAP_X) & IMSM_ORD_REBUILD) == 0) {
e7b84f9d 5515 pr_err("%s has been included in this array twice\n",
4eb26970
DW		 5516 devname);
5517 return 1;
5518 }
656b6b5a 5519 set_imsm_ord_tbl_ent(map, dk->raid_disk, dl->index);
ee5aad5a 5520 dl->disk.status = CONFIGURED_DISK;
d23fe947 5521
3960e579
DW		 5522 /* update size of 'missing' disks to be at least as large as the
5523 * largest acitve member (we only have dummy missing disks when
5524 * creating the first volume)
5525 */
5526 if (super->current_vol == 0) {
5527 for (df = super->missing; df; df = df->next) {
5551b113
CA		 5528 if (total_blocks(&dl->disk) > total_blocks(&df->disk))
5529 set_total_blocks(&df->disk, total_blocks(&dl->disk));
3960e579
DW		 5530 _disk = __get_imsm_disk(mpb, df->index);
5531 *_disk = df->disk;
5532 }
5533 }
5534
5535 /* refresh unset/failed slots to point to valid 'missing' entries */
5536 for (df = super->missing; df; df = df->next)
5537 for (slot = 0; slot < mpb->num_disks; slot++) {
238c0a71 5538 __u32 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
3960e579
DW		 5539
5540 if ((ord & IMSM_ORD_REBUILD) == 0)
5541 continue;
5542 set_imsm_ord_tbl_ent(map, slot, df->index | IMSM_ORD_REBUILD);
1ace8403 5543 if (is_gen_migration(dev)) {
238c0a71
AK		 5544 struct imsm_map *map2 = get_imsm_map(dev,
5545 MAP_1);
0a108d63 5546 int slot2 = get_imsm_disk_slot(map2, df->index);
089f9d79 5547 if (slot2 < map2->num_members && slot2 >= 0) {
1ace8403 5548 __u32 ord2 = get_imsm_ord_tbl_ent(dev,
238c0a71
AK		 5549 slot2,
5550 MAP_1);
1ace8403
AK		 5551 if ((unsigned)df->index ==
5552 ord_to_idx(ord2))
5553 set_imsm_ord_tbl_ent(map2,
0a108d63 5554 slot2,
1ace8403
AK		 5555 df->index |
5556 IMSM_ORD_REBUILD);
5557 }
5558 }
3960e579
DW		 5559 dprintf("set slot:%d to missing disk:%d\n", slot, df->index);
5560 break;
5561 }
5562
d23fe947
DW		 5563 /* if we are creating the first raid device update the family number */
5564 if (super->current_vol == 0) {
5565 __u32 sum;
5566 struct imsm_dev *_dev = __get_imsm_dev(mpb, 0);
d23fe947 5567
3960e579 5568 _disk = __get_imsm_disk(mpb, dl->index);
791b666a 5569 if (!_dev || !_disk) {
e7b84f9d 5570 pr_err("BUG mpb setup error\n");
791b666a
AW		 5571 return 1;
5572 }
d23fe947
DW		 5573 *_dev = *dev;
5574 *_disk = dl->disk;
148acb7b
DW		 5575 sum = random32();
5576 sum += __gen_imsm_checksum(mpb);
d23fe947 5577 mpb->family_num = __cpu_to_le32(sum);
148acb7b 5578 mpb->orig_family_num = mpb->family_num;
d23fe947 5579 }
ca0748fa 5580 super->current_disk = dl;
f20c3968 5581 return 0;
bf5a934a
DW		 5582}
5583
a8619d23
AK		 5584/* mark_spare()
5585 * Function marks disk as spare and restores disk serial
5586 * in case it was previously marked as failed by takeover operation
5587 * reruns:
5588 * -1 : critical error
5589 * 0 : disk is marked as spare but serial is not set
5590 * 1 : success
5591 */
5592int mark_spare(struct dl *disk)
5593{
5594 __u8 serial[MAX_RAID_SERIAL_LEN];
5595 int ret_val = -1;
5596
5597 if (!disk)
5598 return ret_val;
5599
5600 ret_val = 0;
5601 if (!imsm_read_serial(disk->fd, NULL, serial)) {
5602 /* Restore disk serial number, because takeover marks disk
5603 * as failed and adds to serial ':0' before it becomes
5604 * a spare disk.
5605 */
5606 serialcpy(disk->serial, serial);
5607 serialcpy(disk->disk.serial, serial);
5608 ret_val = 1;
5609 }
5610 disk->disk.status = SPARE_DISK;
5611 disk->index = -1;
5612
5613 return ret_val;
5614}
88654014 5615
f20c3968 5616static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk,
72ca9bcf
N		 5617 int fd, char *devname,
5618 unsigned long long data_offset)
cdddbdbc 5619{
c2c087e6 5620 struct intel_super *super = st->sb;
c2c087e6
DW		 5621 struct dl *dd;
5622 unsigned long long size;
fa7bb6f8 5623 unsigned int member_sector_size;
f2f27e63 5624 __u32 id;
c2c087e6
DW		 5625 int rv;
5626 struct stat stb;
5627
88654014
LM		 5628 /* If we are on an RAID enabled platform check that the disk is
5629 * attached to the raid controller.
5630 * We do not need to test disks attachment for container based additions,
5631 * they shall be already tested when container was created/assembled.
88c32bb1 5632 */
d424212e 5633 rv = find_intel_hba_capability(fd, super, devname);
f2f5c343 5634 /* no orom/efi or non-intel hba of the disk */
f0f5a016
LM		 5635 if (rv != 0) {
5636 dprintf("capability: %p fd: %d ret: %d\n",
5637 super->orom, fd, rv);
5638 return 1;
88c32bb1
DW		 5639 }
5640
f20c3968
DW		 5641 if (super->current_vol >= 0)
5642 return add_to_super_imsm_volume(st, dk, fd, devname);
bf5a934a 5643
c2c087e6 5644 fstat(fd, &stb);
503975b9 5645 dd = xcalloc(sizeof(*dd), 1);
c2c087e6
DW		 5646 dd->major = major(stb.st_rdev);
5647 dd->minor = minor(stb.st_rdev);
503975b9 5648 dd->devname = devname ? xstrdup(devname) : NULL;
c2c087e6 5649 dd->fd = fd;
689c9bf3 5650 dd->e = NULL;
1a64be56 5651 dd->action = DISK_ADD;
c2c087e6 5652 rv = imsm_read_serial(fd, devname, dd->serial);
32ba9157 5653 if (rv) {
e7b84f9d 5654 pr_err("failed to retrieve scsi serial, aborting\n");
20bee0f8
PB		 5655 if (dd->devname)
5656 free(dd->devname);
949c47a0 5657 free(dd);
0030e8d6 5658 abort();
c2c087e6 5659 }
20bee0f8
PB		 5660 if (super->hba && ((super->hba->type == SYS_DEV_NVME) ||
5661 (super->hba->type == SYS_DEV_VMD))) {
5662 int i;
5663 char *devpath = diskfd_to_devpath(fd);
5664 char controller_path[PATH_MAX];
5665
5666 if (!devpath) {
5667 pr_err("failed to get devpath, aborting\n");
5668 if (dd->devname)
5669 free(dd->devname);
5670 free(dd);
5671 return 1;
5672 }
5673
5674 snprintf(controller_path, PATH_MAX-1, "%s/device", devpath);
5675 free(devpath);
5676
5677 if (devpath_to_vendor(controller_path) == 0x8086) {
5678 /*
5679 * If Intel's NVMe drive has serial ended with
5680 * "-A","-B","-1" or "-2" it means that this is "x8"
5681 * device (double drive on single PCIe card).
5682 * User should be warned about potential data loss.
5683 */
5684 for (i = MAX_RAID_SERIAL_LEN-1; i > 0; i--) {
5685 /* Skip empty character at the end */
5686 if (dd->serial[i] == 0)
5687 continue;
5688
5689 if (((dd->serial[i] == 'A') ||
5690 (dd->serial[i] == 'B') ||
5691 (dd->serial[i] == '1') ||
5692 (dd->serial[i] == '2')) &&
5693 (dd->serial[i-1] == '-'))
5694 pr_err("\tThe action you are about to take may put your data at risk.\n"
5695 "\tPlease note that x8 devices may consist of two separate x4 devices "
5696 "located on a single PCIe port.\n"
5697 "\tRAID 0 is the only supported configuration for this type of x8 device.\n");
5698 break;
5699 }
32716c51
PB		 5700 } else if (super->hba->type == SYS_DEV_VMD && super->orom &&
5701 !imsm_orom_has_tpv_support(super->orom)) {
5702 pr_err("\tPlatform configuration does not support non-Intel NVMe drives.\n"
5703 "\tPlease refer to Intel(R) RSTe user guide.\n");
5704 free(dd->devname);
5705 free(dd);
5706 return 1;
20bee0f8
PB		 5707 }
5708 }
c2c087e6 5709
c2c087e6 5710 get_dev_size(fd, NULL, &size);
fa7bb6f8
PB		 5711 get_dev_sector_size(fd, NULL, &member_sector_size);
5712
5713 if (super->sector_size == 0) {
5714 /* this a first device, so sector_size is not set yet */
5715 super->sector_size = member_sector_size;
5716 } else if (member_sector_size != super->sector_size) {
5717 pr_err("Mixing between different sector size is forbidden, aborting...\n");
5718 if (dd->devname)
5719 free(dd->devname);
5720 free(dd);
5721 return 1;
5722 }
5723
71e5411e 5724 /* clear migr_rec when adding disk to container */
de44e46f
PB		 5725 memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*super->sector_size);
5726 if (lseek64(fd, size - MIGR_REC_SECTOR_POSITION*super->sector_size,
5727 SEEK_SET) >= 0) {
466070ad 5728 if ((unsigned int)write(fd, super->migr_rec_buf,
de44e46f
PB		 5729 MIGR_REC_BUF_SECTORS*super->sector_size) !=
5730 MIGR_REC_BUF_SECTORS*super->sector_size)
71e5411e
PB		 5731 perror("Write migr_rec failed");
5732 }
5733
c2c087e6 5734 size /= 512;
1f24f035 5735 serialcpy(dd->disk.serial, dd->serial);
5551b113
CA		 5736 set_total_blocks(&dd->disk, size);
5737 if (__le32_to_cpu(dd->disk.total_blocks_hi) > 0) {
5738 struct imsm_super *mpb = super->anchor;
5739 mpb->attributes |= MPB_ATTRIB_2TB_DISK;
5740 }
a8619d23 5741 mark_spare(dd);
c2c087e6 5742 if (sysfs_disk_to_scsi_id(fd, &id) == 0)
b9f594fe 5743 dd->disk.scsi_id = __cpu_to_le32(id);
c2c087e6 5744 else
b9f594fe 5745 dd->disk.scsi_id = __cpu_to_le32(0);
43dad3d6
DW		 5746
5747 if (st->update_tail) {
1a64be56
LM		 5748 dd->next = super->disk_mgmt_list;
5749 super->disk_mgmt_list = dd;
43dad3d6
DW		 5750 } else {
5751 dd->next = super->disks;
5752 super->disks = dd;
ceaf0ee1 5753 super->updates_pending++;
43dad3d6 5754 }
f20c3968
DW		 5755
5756 return 0;
cdddbdbc
DW		 5757}
5758
1a64be56
LM		 5759static int remove_from_super_imsm(struct supertype *st, mdu_disk_info_t *dk)
5760{
5761 struct intel_super *super = st->sb;
5762 struct dl *dd;
5763
5764 /* remove from super works only in mdmon - for communication
5765 * manager - monitor. Check if communication memory buffer
5766 * is prepared.
5767 */
5768 if (!st->update_tail) {
1ade5cc1 5769 pr_err("shall be used in mdmon context only\n");
1a64be56
LM		 5770 return 1;
5771 }
503975b9 5772 dd = xcalloc(1, sizeof(*dd));
1a64be56
LM		 5773 dd->major = dk->major;
5774 dd->minor = dk->minor;
1a64be56 5775 dd->fd = -1;
a8619d23 5776 mark_spare(dd);
1a64be56
LM		 5777 dd->action = DISK_REMOVE;
5778
5779 dd->next = super->disk_mgmt_list;
5780 super->disk_mgmt_list = dd;
5781
1a64be56
LM		 5782 return 0;
5783}
5784
f796af5d
DW		 5785static int store_imsm_mpb(int fd, struct imsm_super *mpb);
5786
5787static union {
f36a9ecd 5788 char buf[MAX_SECTOR_SIZE];
f796af5d 5789 struct imsm_super anchor;
f36a9ecd 5790} spare_record __attribute__ ((aligned(MAX_SECTOR_SIZE)));
c2c087e6 5791
d23fe947
DW		 5792/* spare records have their own family number and do not have any defined raid
5793 * devices
5794 */
5795static int write_super_imsm_spares(struct intel_super *super, int doclose)
5796{
d23fe947 5797 struct imsm_super *mpb = super->anchor;
f796af5d 5798 struct imsm_super *spare = &spare_record.anchor;
d23fe947
DW		 5799 __u32 sum;
5800 struct dl *d;
5801
68641cdb
JS		 5802 spare->mpb_size = __cpu_to_le32(sizeof(struct imsm_super));
5803 spare->generation_num = __cpu_to_le32(1UL);
f796af5d 5804 spare->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
68641cdb
JS		 5805 spare->num_disks = 1;
5806 spare->num_raid_devs = 0;
5807 spare->cache_size = mpb->cache_size;
5808 spare->pwr_cycle_count = __cpu_to_le32(1);
f796af5d
DW		 5809
5810 snprintf((char *) spare->sig, MAX_SIGNATURE_LENGTH,
5811 MPB_SIGNATURE MPB_VERSION_RAID0);
d23fe947
DW		 5812
5813 for (d = super->disks; d; d = d->next) {
8796fdc4 5814 if (d->index != -1)
d23fe947
DW		 5815 continue;
5816
f796af5d 5817 spare->disk[0] = d->disk;
027c374f
CA		 5818 if (__le32_to_cpu(d->disk.total_blocks_hi) > 0)
5819 spare->attributes |= MPB_ATTRIB_2TB_DISK;
5820
f36a9ecd
PB		 5821 if (super->sector_size == 4096)
5822 convert_to_4k_imsm_disk(&spare->disk[0]);
5823
f796af5d
DW		 5824 sum = __gen_imsm_checksum(spare);
5825 spare->family_num = __cpu_to_le32(sum);
5826 spare->orig_family_num = 0;
5827 sum = __gen_imsm_checksum(spare);
5828 spare->check_sum = __cpu_to_le32(sum);
d23fe947 5829
f796af5d 5830 if (store_imsm_mpb(d->fd, spare)) {
1ade5cc1
N		 5831 pr_err("failed for device %d:%d %s\n",
5832 d->major, d->minor, strerror(errno));
e74255d9 5833 return 1;
d23fe947
DW		 5834 }
5835 if (doclose) {
5836 close(d->fd);
5837 d->fd = -1;
5838 }
5839 }
5840
e74255d9 5841 return 0;
d23fe947
DW		 5842}
5843
36988a3d 5844static int write_super_imsm(struct supertype *st, int doclose)
cdddbdbc 5845{
36988a3d 5846 struct intel_super *super = st->sb;
f36a9ecd 5847 unsigned int sector_size = super->sector_size;
949c47a0 5848 struct imsm_super *mpb = super->anchor;
c2c087e6
DW		 5849 struct dl *d;
5850 __u32 generation;
5851 __u32 sum;
d23fe947 5852 int spares = 0;
949c47a0 5853 int i;
a48ac0a8 5854 __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk);
36988a3d 5855 int num_disks = 0;
146c6260 5856 int clear_migration_record = 1;
bbab0940 5857 __u32 bbm_log_size;
cdddbdbc 5858
c2c087e6
DW		 5859 /* 'generation' is incremented everytime the metadata is written */
5860 generation = __le32_to_cpu(mpb->generation_num);
5861 generation++;
5862 mpb->generation_num = __cpu_to_le32(generation);
5863
148acb7b
DW		 5864 /* fix up cases where previous mdadm releases failed to set
5865 * orig_family_num
5866 */
5867 if (mpb->orig_family_num == 0)
5868 mpb->orig_family_num = mpb->family_num;
5869
d23fe947 5870 for (d = super->disks; d; d = d->next) {
8796fdc4 5871 if (d->index == -1)
d23fe947 5872 spares++;
36988a3d 5873 else {
d23fe947 5874 mpb->disk[d->index] = d->disk;
36988a3d
AK		 5875 num_disks++;
5876 }
d23fe947 5877 }
36988a3d 5878 for (d = super->missing; d; d = d->next) {
47ee5a45 5879 mpb->disk[d->index] = d->disk;
36988a3d
AK		 5880 num_disks++;
5881 }
5882 mpb->num_disks = num_disks;
5883 mpb_size += sizeof(struct imsm_disk) * mpb->num_disks;
b9f594fe 5884
949c47a0
DW		 5885 for (i = 0; i < mpb->num_raid_devs; i++) {
5886 struct imsm_dev *dev = __get_imsm_dev(mpb, i);
36988a3d
AK		 5887 struct imsm_dev *dev2 = get_imsm_dev(super, i);
5888 if (dev && dev2) {
5889 imsm_copy_dev(dev, dev2);
5890 mpb_size += sizeof_imsm_dev(dev, 0);
5891 }
146c6260
AK		 5892 if (is_gen_migration(dev2))
5893 clear_migration_record = 0;
949c47a0 5894 }
bbab0940
TM		 5895
5896 bbm_log_size = get_imsm_bbm_log_size(super->bbm_log);
5897
5898 if (bbm_log_size) {
5899 memcpy((void *)mpb + mpb_size, super->bbm_log, bbm_log_size);
5900 mpb->attributes |= MPB_ATTRIB_BBM;
5901 } else
5902 mpb->attributes &= ~MPB_ATTRIB_BBM;
5903
5904 super->anchor->bbm_log_size = __cpu_to_le32(bbm_log_size);
5905 mpb_size += bbm_log_size;
a48ac0a8 5906 mpb->mpb_size = __cpu_to_le32(mpb_size);
949c47a0 5907
bbab0940
TM		 5908#ifdef DEBUG
5909 assert(super->len == 0 || mpb_size <= super->len);
5910#endif
5911
c2c087e6 5912 /* recalculate checksum */
949c47a0 5913 sum = __gen_imsm_checksum(mpb);
c2c087e6
DW		 5914 mpb->check_sum = __cpu_to_le32(sum);
5915
51d83f5d
AK		 5916 if (super->clean_migration_record_by_mdmon) {
5917 clear_migration_record = 1;
5918 super->clean_migration_record_by_mdmon = 0;
5919 }
146c6260 5920 if (clear_migration_record)
de44e46f
PB		 5921 memset(super->migr_rec_buf, 0,
5922 MIGR_REC_BUF_SECTORS*sector_size);
146c6260 5923
f36a9ecd
PB		 5924 if (sector_size == 4096)
5925 convert_to_4k(super);
5926
d23fe947 5927 /* write the mpb for disks that compose raid devices */
c2c087e6 5928 for (d = super->disks; d ; d = d->next) {
86c54047 5929 if (d->index < 0 || is_failed(&d->disk))
d23fe947 5930 continue;
30602f53 5931
146c6260
AK		 5932 if (clear_migration_record) {
5933 unsigned long long dsize;
5934
5935 get_dev_size(d->fd, NULL, &dsize);
de44e46f
PB		 5936 if (lseek64(d->fd, dsize - sector_size,
5937 SEEK_SET) >= 0) {
466070ad
PB		 5938 if ((unsigned int)write(d->fd,
5939 super->migr_rec_buf,
de44e46f
PB		 5940 MIGR_REC_BUF_SECTORS*sector_size) !=
5941 MIGR_REC_BUF_SECTORS*sector_size)
9e2d750d 5942 perror("Write migr_rec failed");
146c6260
AK		 5943 }
5944 }
51d83f5d
AK		 5945
5946 if (store_imsm_mpb(d->fd, mpb))
5947 fprintf(stderr,
1ade5cc1
N		 5948 "failed for device %d:%d (fd: %d)%s\n",
5949 d->major, d->minor,
51d83f5d
AK		 5950 d->fd, strerror(errno));
5951
c2c087e6
DW		 5952 if (doclose) {
5953 close(d->fd);
5954 d->fd = -1;
5955 }
5956 }
5957
d23fe947
DW		 5958 if (spares)
5959 return write_super_imsm_spares(super, doclose);
5960
e74255d9 5961 return 0;
c2c087e6
DW		 5962}
5963
9b1fb677 5964static int create_array(struct supertype *st, int dev_idx)
43dad3d6
DW		 5965{
5966 size_t len;
5967 struct imsm_update_create_array *u;
5968 struct intel_super *super = st->sb;
9b1fb677 5969 struct imsm_dev *dev = get_imsm_dev(super, dev_idx);
238c0a71 5970 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW		 5971 struct disk_info *inf;
5972 struct imsm_disk *disk;
5973 int i;
43dad3d6 5974
54c2c1ea
DW		 5975 len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0) +
5976 sizeof(*inf) * map->num_members;
503975b9 5977 u = xmalloc(len);
43dad3d6 5978 u->type = update_create_array;
9b1fb677 5979 u->dev_idx = dev_idx;
43dad3d6 5980 imsm_copy_dev(&u->dev, dev);
54c2c1ea
DW		 5981 inf = get_disk_info(u);
5982 for (i = 0; i < map->num_members; i++) {
238c0a71 5983 int idx = get_imsm_disk_idx(dev, i, MAP_X);
9b1fb677 5984
54c2c1ea 5985 disk = get_imsm_disk(super, idx);
1ca5c8e0
N		 5986 if (!disk)
5987 disk = get_imsm_missing(super, idx);
54c2c1ea
DW		 5988 serialcpy(inf[i].serial, disk->serial);
5989 }
43dad3d6
DW		 5990 append_metadata_update(st, u, len);
5991
5992 return 0;
5993}
5994
1a64be56 5995static int mgmt_disk(struct supertype *st)
43dad3d6
DW		 5996{
5997 struct intel_super *super = st->sb;
5998 size_t len;
1a64be56 5999 struct imsm_update_add_remove_disk *u;
43dad3d6 6000
1a64be56 6001 if (!super->disk_mgmt_list)
43dad3d6
DW		 6002 return 0;
6003
6004 len = sizeof(*u);
503975b9 6005 u = xmalloc(len);
1a64be56 6006 u->type = update_add_remove_disk;
43dad3d6
DW		 6007 append_metadata_update(st, u, len);
6008
6009 return 0;
6010}
2432ce9b
AP		 6011#endif
6012
6013__u32 crc32c_le(__u32 crc, unsigned char const *p, size_t len);
6014
6015static int write_init_ppl_imsm(struct supertype *st, struct mdinfo *info, int fd)
6016{
6017 struct intel_super *super = st->sb;
6018 void *buf;
6019 struct ppl_header *ppl_hdr;
6020 int ret;
6021
6022 ret = posix_memalign(&buf, 4096, PPL_HEADER_SIZE);
6023 if (ret) {
6024 pr_err("Failed to allocate PPL header buffer\n");
6025 return ret;
6026 }
6027
6028 memset(buf, 0, PPL_HEADER_SIZE);
6029 ppl_hdr = buf;
6030 memset(ppl_hdr->reserved, 0xff, PPL_HDR_RESERVED);
6031 ppl_hdr->signature = __cpu_to_le32(super->anchor->orig_family_num);
6032 ppl_hdr->checksum = __cpu_to_le32(~crc32c_le(~0, buf, PPL_HEADER_SIZE));
6033
6034 if (lseek64(fd, info->ppl_sector * 512, SEEK_SET) < 0) {
6035 ret = errno;
6036 perror("Failed to seek to PPL header location");
6037 }
6038
6039 if (!ret && write(fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) {
6040 ret = errno;
6041 perror("Write PPL header failed");
6042 }
6043
6044 if (!ret)
6045 fsync(fd);
6046
6047 free(buf);
6048 return ret;
6049}
6050
6051static int validate_ppl_imsm(struct supertype *st, struct mdinfo *info,
6052 struct mdinfo *disk)
6053{
6054 struct intel_super *super = st->sb;
6055 struct dl *d;
6056 void *buf;
6057 int ret = 0;
6058 struct ppl_header *ppl_hdr;
6059 __u32 crc;
6060 struct imsm_dev *dev;
6061 struct imsm_map *map;
6062 __u32 idx;
6063
6064 if (disk->disk.raid_disk < 0)
6065 return 0;
6066
6067 if (posix_memalign(&buf, 4096, PPL_HEADER_SIZE)) {
6068 pr_err("Failed to allocate PPL header buffer\n");
6069 return -1;
6070 }
6071
6072 dev = get_imsm_dev(super, info->container_member);
6073 map = get_imsm_map(dev, MAP_X);
6074 idx = get_imsm_disk_idx(dev, disk->disk.raid_disk, MAP_X);
6075 d = get_imsm_dl_disk(super, idx);
6076
6077 if (!d || d->index < 0 || is_failed(&d->disk))
6078 goto out;
6079
6080 if (lseek64(d->fd, info->ppl_sector * 512, SEEK_SET) < 0) {
6081 perror("Failed to seek to PPL header location");
6082 ret = -1;
6083 goto out;
6084 }
6085
6086 if (read(d->fd, buf, PPL_HEADER_SIZE) != PPL_HEADER_SIZE) {
6087 perror("Read PPL header failed");
6088 ret = -1;
6089 goto out;
6090 }
6091
6092 ppl_hdr = buf;
6093
6094 crc = __le32_to_cpu(ppl_hdr->checksum);
6095 ppl_hdr->checksum = 0;
6096
6097 if (crc != ~crc32c_le(~0, buf, PPL_HEADER_SIZE)) {
6098 dprintf("Wrong PPL header checksum on %s\n",
6099 d->devname);
6100 ret = 1;
6101 }
6102
6103 if (!ret && (__le32_to_cpu(ppl_hdr->signature) !=
6104 super->anchor->orig_family_num)) {
6105 dprintf("Wrong PPL header signature on %s\n",
6106 d->devname);
6107 ret = 1;
6108 }
6109
6110out:
6111 free(buf);
6112
6113 if (ret == 1 && map->map_state == IMSM_T_STATE_UNINITIALIZED)
6114 return st->ss->write_init_ppl(st, info, d->fd);
6115
6116 return ret;
6117}
6118
6119#ifndef MDASSEMBLE
6120
6121static int write_init_ppl_imsm_all(struct supertype *st, struct mdinfo *info)
6122{
6123 struct intel_super *super = st->sb;
6124 struct dl *d;
6125 int ret = 0;
6126
6127 if (info->consistency_policy != CONSISTENCY_POLICY_PPL ||
6128 info->array.level != 5)
6129 return 0;
6130
6131 for (d = super->disks; d ; d = d->next) {
6132 if (d->index < 0 || is_failed(&d->disk))
6133 continue;
6134
6135 ret = st->ss->write_init_ppl(st, info, d->fd);
6136 if (ret)
6137 break;
6138 }
6139
6140 return ret;
6141}
43dad3d6 6142
c2c087e6
DW		 6143static int write_init_super_imsm(struct supertype *st)
6144{
9b1fb677
DW		 6145 struct intel_super *super = st->sb;
6146 int current_vol = super->current_vol;
2432ce9b
AP		 6147 int rv = 0;
6148 struct mdinfo info;
6149
6150 getinfo_super_imsm(st, &info, NULL);
9b1fb677
DW		 6151
6152 /* we are done with current_vol reset it to point st at the container */
6153 super->current_vol = -1;
6154
8273f55e 6155 if (st->update_tail) {
43dad3d6
DW		 6156 /* queue the recently created array / added disk
6157 * as a metadata update */
8273f55e 6158
43dad3d6 6159 /* determine if we are creating a volume or adding a disk */
9b1fb677 6160 if (current_vol < 0) {
1a64be56
LM		 6161 /* in the mgmt (add/remove) disk case we are running
6162 * in mdmon context, so don't close fd's
43dad3d6 6163 */
2432ce9b
AP		 6164 rv = mgmt_disk(st);
6165 } else {
6166 rv = write_init_ppl_imsm_all(st, &info);
6167 if (!rv)
6168 rv = create_array(st, current_vol);
6169 }
d682f344
N		 6170 } else {
6171 struct dl *d;
6172 for (d = super->disks; d; d = d->next)
ba728be7 6173 Kill(d->devname, NULL, 0, -1, 1);
2432ce9b
AP		 6174 if (current_vol >= 0)
6175 rv = write_init_ppl_imsm_all(st, &info);
6176 if (!rv)
6177 rv = write_super_imsm(st, 1);
d682f344 6178 }
2432ce9b
AP		 6179
6180 return rv;
cdddbdbc 6181}
0e600426 6182#endif
cdddbdbc 6183
e683ca88 6184static int store_super_imsm(struct supertype *st, int fd)
cdddbdbc 6185{
e683ca88
DW		 6186 struct intel_super *super = st->sb;
6187 struct imsm_super *mpb = super ? super->anchor : NULL;
551c80c1 6188
e683ca88 6189 if (!mpb)
ad97895e
DW		 6190 return 1;
6191
1799c9e8 6192#ifndef MDASSEMBLE
f36a9ecd
PB		 6193 if (super->sector_size == 4096)
6194 convert_to_4k(super);
e683ca88 6195 return store_imsm_mpb(fd, mpb);
1799c9e8
N		 6196#else
6197 return 1;
6198#endif
cdddbdbc
DW		 6199}
6200
0e600426 6201#ifndef MDASSEMBLE
cdddbdbc
DW		 6202static int validate_geometry_imsm_container(struct supertype *st, int level,
6203 int layout, int raiddisks, int chunk,
af4348dd
N		 6204 unsigned long long size,
6205 unsigned long long data_offset,
6206 char *dev,
2c514b71
NB		 6207 unsigned long long *freesize,
6208 int verbose)
cdddbdbc 6209{
c2c087e6
DW		 6210 int fd;
6211 unsigned long long ldsize;
594dc1b8 6212 struct intel_super *super;
f2f5c343 6213 int rv = 0;
cdddbdbc 6214
c2c087e6
DW		 6215 if (level != LEVEL_CONTAINER)
6216 return 0;
6217 if (!dev)
6218 return 1;
6219
6220 fd = open(dev, O_RDONLY|O_EXCL, 0);
6221 if (fd < 0) {
ba728be7 6222 if (verbose > 0)
e7b84f9d 6223 pr_err("imsm: Cannot open %s: %s\n",
2c514b71 6224 dev, strerror(errno));
c2c087e6
DW		 6225 return 0;
6226 }
6227 if (!get_dev_size(fd, dev, &ldsize)) {
6228 close(fd);
6229 return 0;
6230 }
f2f5c343
LM		 6231
6232 /* capabilities retrieve could be possible
6233 * note that there is no fd for the disks in array.
6234 */
6235 super = alloc_super();
8d67477f
TM		 6236 if (!super) {
6237 close(fd);
6238 return 0;
6239 }
fa7bb6f8
PB		 6240 if (!get_dev_sector_size(fd, NULL, &super->sector_size)) {
6241 close(fd);
6242 free_imsm(super);
6243 return 0;
6244 }
6245
ba728be7 6246 rv = find_intel_hba_capability(fd, super, verbose > 0 ? dev : NULL);
f2f5c343
LM		 6247 if (rv != 0) {
6248#if DEBUG
6249 char str[256];
6250 fd2devname(fd, str);
1ade5cc1 6251 dprintf("fd: %d %s orom: %p rv: %d raiddisk: %d\n",
f2f5c343
LM		 6252 fd, str, super->orom, rv, raiddisks);
6253#endif
6254 /* no orom/efi or non-intel hba of the disk */
6255 close(fd);
6256 free_imsm(super);
6257 return 0;
6258 }
c2c087e6 6259 close(fd);
9126b9a8
CA		 6260 if (super->orom) {
6261 if (raiddisks > super->orom->tds) {
6262 if (verbose)
7a862a02 6263 pr_err("%d exceeds maximum number of platform supported disks: %d\n",
9126b9a8
CA		 6264 raiddisks, super->orom->tds);
6265 free_imsm(super);
6266 return 0;
6267 }
6268 if ((super->orom->attr & IMSM_OROM_ATTR_2TB_DISK) == 0 &&
6269 (ldsize >> 9) >> 32 > 0) {
6270 if (verbose)
e7b84f9d 6271 pr_err("%s exceeds maximum platform supported size\n", dev);
9126b9a8
CA		 6272 free_imsm(super);
6273 return 0;
6274 }
f2f5c343 6275 }
c2c087e6 6276
af4348dd 6277 *freesize = avail_size_imsm(st, ldsize >> 9, data_offset);
f2f5c343 6278 free_imsm(super);
c2c087e6
DW		 6279
6280 return 1;
cdddbdbc
DW		 6281}
6282
0dcecb2e
DW		 6283static unsigned long long find_size(struct extent *e, int *idx, int num_extents)
6284{
6285 const unsigned long long base_start = e[*idx].start;
6286 unsigned long long end = base_start + e[*idx].size;
6287 int i;
6288
6289 if (base_start == end)
6290 return 0;
6291
6292 *idx = *idx + 1;
6293 for (i = *idx; i < num_extents; i++) {
6294 /* extend overlapping extents */
6295 if (e[i].start >= base_start &&
6296 e[i].start <= end) {
6297 if (e[i].size == 0)
6298 return 0;
6299 if (e[i].start + e[i].size > end)
6300 end = e[i].start + e[i].size;
6301 } else if (e[i].start > end) {
6302 *idx = i;
6303 break;
6304 }
6305 }
6306
6307 return end - base_start;
6308}
6309
6310static unsigned long long merge_extents(struct intel_super *super, int sum_extents)
6311{
6312 /* build a composite disk with all known extents and generate a new
6313 * 'maxsize' given the "all disks in an array must share a common start
6314 * offset" constraint
6315 */
503975b9 6316 struct extent *e = xcalloc(sum_extents, sizeof(*e));
0dcecb2e
DW		 6317 struct dl *dl;
6318 int i, j;
6319 int start_extent;
6320 unsigned long long pos;
b9d77223 6321 unsigned long long start = 0;
0dcecb2e
DW		 6322 unsigned long long maxsize;
6323 unsigned long reserve;
6324
0dcecb2e
DW		 6325 /* coalesce and sort all extents. also, check to see if we need to
6326 * reserve space between member arrays
6327 */
6328 j = 0;
6329 for (dl = super->disks; dl; dl = dl->next) {
6330 if (!dl->e)
6331 continue;
6332 for (i = 0; i < dl->extent_cnt; i++)
6333 e[j++] = dl->e[i];
6334 }
6335 qsort(e, sum_extents, sizeof(*e), cmp_extent);
6336
6337 /* merge extents */
6338 i = 0;
6339 j = 0;
6340 while (i < sum_extents) {
6341 e[j].start = e[i].start;
6342 e[j].size = find_size(e, &i, sum_extents);
6343 j++;
6344 if (e[j-1].size == 0)
6345 break;
6346 }
6347
6348 pos = 0;
6349 maxsize = 0;
6350 start_extent = 0;
6351 i = 0;
6352 do {
6353 unsigned long long esize;
6354
6355 esize = e[i].start - pos;
6356 if (esize >= maxsize) {
6357 maxsize = esize;
6358 start = pos;
6359 start_extent = i;
6360 }
6361 pos = e[i].start + e[i].size;
6362 i++;
6363 } while (e[i-1].size);
6364 free(e);
6365
a7dd165b
DW		 6366 if (maxsize == 0)
6367 return 0;
6368
6369 /* FIXME assumes volume at offset 0 is the first volume in a
6370 * container
6371 */
0dcecb2e
DW		 6372 if (start_extent > 0)
6373 reserve = IMSM_RESERVED_SECTORS; /* gap between raid regions */
6374 else
6375 reserve = 0;
6376
6377 if (maxsize < reserve)
a7dd165b 6378 return 0;
0dcecb2e 6379
5551b113 6380 super->create_offset = ~((unsigned long long) 0);
0dcecb2e 6381 if (start + reserve > super->create_offset)
a7dd165b 6382 return 0; /* start overflows create_offset */
0dcecb2e
DW		 6383 super->create_offset = start + reserve;
6384
6385 return maxsize - reserve;
6386}
6387
88c32bb1
DW		 6388static int is_raid_level_supported(const struct imsm_orom *orom, int level, int raiddisks)
6389{
6390 if (level < 0 || level == 6 || level == 4)
6391 return 0;
6392
6393 /* if we have an orom prevent invalid raid levels */
6394 if (orom)
6395 switch (level) {
6396 case 0: return imsm_orom_has_raid0(orom);
6397 case 1:
6398 if (raiddisks > 2)
6399 return imsm_orom_has_raid1e(orom);
1c556e92
DW		 6400 return imsm_orom_has_raid1(orom) && raiddisks == 2;
6401 case 10: return imsm_orom_has_raid10(orom) && raiddisks == 4;
6402 case 5: return imsm_orom_has_raid5(orom) && raiddisks > 2;
88c32bb1
DW		 6403 }
6404 else
6405 return 1; /* not on an Intel RAID platform so anything goes */
6406
6407 return 0;
6408}
6409
ca9de185
LM		 6410static int
6411active_arrays_by_format(char *name, char* hba, struct md_list **devlist,
6412 int dpa, int verbose)
6413{
6414 struct mdstat_ent *mdstat = mdstat_read(0, 0);
594dc1b8 6415 struct mdstat_ent *memb;
ca9de185
LM		 6416 int count = 0;
6417 int num = 0;
594dc1b8 6418 struct md_list *dv;
ca9de185
LM		 6419 int found;
6420
6421 for (memb = mdstat ; memb ; memb = memb->next) {
6422 if (memb->metadata_version &&
6423 (strncmp(memb->metadata_version, "external:", 9) == 0) &&
6424 (strcmp(&memb->metadata_version[9], name) == 0) &&
6425 !is_subarray(memb->metadata_version+9) &&
6426 memb->members) {
6427 struct dev_member *dev = memb->members;
6428 int fd = -1;
6429 while(dev && (fd < 0)) {
503975b9
N		 6430 char *path = xmalloc(strlen(dev->name) + strlen("/dev/") + 1);
6431 num = sprintf(path, "%s%s", "/dev/", dev->name);
6432 if (num > 0)
6433 fd = open(path, O_RDONLY, 0);
089f9d79 6434 if (num <= 0 || fd < 0) {
676e87a8 6435 pr_vrb("Cannot open %s: %s\n",
503975b9 6436 dev->name, strerror(errno));
ca9de185 6437 }
503975b9 6438 free(path);
ca9de185
LM		 6439 dev = dev->next;
6440 }
6441 found = 0;
089f9d79 6442 if (fd >= 0 && disk_attached_to_hba(fd, hba)) {
ca9de185
LM		 6443 struct mdstat_ent *vol;
6444 for (vol = mdstat ; vol ; vol = vol->next) {
089f9d79 6445 if (vol->active > 0 &&
ca9de185 6446 vol->metadata_version &&
9581efb1 6447 is_container_member(vol, memb->devnm)) {
ca9de185
LM		 6448 found++;
6449 count++;
6450 }
6451 }
6452 if (*devlist && (found < dpa)) {
503975b9 6453 dv = xcalloc(1, sizeof(*dv));
9581efb1
N		 6454 dv->devname = xmalloc(strlen(memb->devnm) + strlen("/dev/") + 1);
6455 sprintf(dv->devname, "%s%s", "/dev/", memb->devnm);
503975b9
N		 6456 dv->found = found;
6457 dv->used = 0;
6458 dv->next = *devlist;
6459 *devlist = dv;
ca9de185
LM		 6460 }
6461 }
6462 if (fd >= 0)
6463 close(fd);
6464 }
6465 }
6466 free_mdstat(mdstat);
6467 return count;
6468}
6469
6470#ifdef DEBUG_LOOP
6471static struct md_list*
6472get_loop_devices(void)
6473{
6474 int i;
6475 struct md_list *devlist = NULL;
594dc1b8 6476 struct md_list *dv;
ca9de185
LM		 6477
6478 for(i = 0; i < 12; i++) {
503975b9
N		 6479 dv = xcalloc(1, sizeof(*dv));
6480 dv->devname = xmalloc(40);
ca9de185
LM		 6481 sprintf(dv->devname, "/dev/loop%d", i);
6482 dv->next = devlist;
6483 devlist = dv;
6484 }
6485 return devlist;
6486}
6487#endif
6488
6489static struct md_list*
6490get_devices(const char *hba_path)
6491{
6492 struct md_list *devlist = NULL;
594dc1b8 6493 struct md_list *dv;
ca9de185
LM		 6494 struct dirent *ent;
6495 DIR *dir;
6496 int err = 0;
6497
6498#if DEBUG_LOOP
6499 devlist = get_loop_devices();
6500 return devlist;
6501#endif
6502 /* scroll through /sys/dev/block looking for devices attached to
6503 * this hba
6504 */
6505 dir = opendir("/sys/dev/block");
6506 for (ent = dir ? readdir(dir) : NULL; ent; ent = readdir(dir)) {
6507 int fd;
6508 char buf[1024];
6509 int major, minor;
6510 char *path = NULL;
6511 if (sscanf(ent->d_name, "%d:%d", &major, &minor) != 2)
6512 continue;
6513 path = devt_to_devpath(makedev(major, minor));
6514 if (!path)
6515 continue;
6516 if (!path_attached_to_hba(path, hba_path)) {
6517 free(path);
6518 path = NULL;
6519 continue;
6520 }
6521 free(path);
6522 path = NULL;
6523 fd = dev_open(ent->d_name, O_RDONLY);
6524 if (fd >= 0) {
6525 fd2devname(fd, buf);
6526 close(fd);
6527 } else {
e7b84f9d 6528 pr_err("cannot open device: %s\n",
ca9de185
LM		 6529 ent->d_name);
6530 continue;
6531 }
6532
503975b9
N		 6533 dv = xcalloc(1, sizeof(*dv));
6534 dv->devname = xstrdup(buf);
ca9de185
LM		 6535 dv->next = devlist;
6536 devlist = dv;
6537 }
6538 if (err) {
6539 while(devlist) {
6540 dv = devlist;
6541 devlist = devlist->next;
6542 free(dv->devname);
6543 free(dv);
6544 }
6545 }
562aa102 6546 closedir(dir);
ca9de185
LM		 6547 return devlist;
6548}
6549
6550static int
6551count_volumes_list(struct md_list *devlist, char *homehost,
6552 int verbose, int *found)
6553{
6554 struct md_list *tmpdev;
6555 int count = 0;
594dc1b8 6556 struct supertype *st;
ca9de185
LM		 6557
6558 /* first walk the list of devices to find a consistent set
6559 * that match the criterea, if that is possible.
6560 * We flag the ones we like with 'used'.
6561 */
6562 *found = 0;
6563 st = match_metadata_desc_imsm("imsm");
6564 if (st == NULL) {
676e87a8 6565 pr_vrb("cannot allocate memory for imsm supertype\n");
ca9de185
LM		 6566 return 0;
6567 }
6568
6569 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
6570 char *devname = tmpdev->devname;
6571 struct stat stb;
6572 struct supertype *tst;
6573 int dfd;
6574 if (tmpdev->used > 1)
6575 continue;
6576 tst = dup_super(st);
6577 if (tst == NULL) {
676e87a8 6578 pr_vrb("cannot allocate memory for imsm supertype\n");
ca9de185
LM		 6579 goto err_1;
6580 }
6581 tmpdev->container = 0;
6582 dfd = dev_open(devname, O_RDONLY|O_EXCL);
6583 if (dfd < 0) {
1ade5cc1 6584 dprintf("cannot open device %s: %s\n",
ca9de185
LM		 6585 devname, strerror(errno));
6586 tmpdev->used = 2;
6587 } else if (fstat(dfd, &stb)< 0) {
6588 /* Impossible! */
1ade5cc1 6589 dprintf("fstat failed for %s: %s\n",
ca9de185
LM		 6590 devname, strerror(errno));
6591 tmpdev->used = 2;
6592 } else if ((stb.st_mode & S_IFMT) != S_IFBLK) {
1ade5cc1 6593 dprintf("%s is not a block device.\n",
ca9de185
LM		 6594 devname);
6595 tmpdev->used = 2;
6596 } else if (must_be_container(dfd)) {
6597 struct supertype *cst;
6598 cst = super_by_fd(dfd, NULL);
6599 if (cst == NULL) {
1ade5cc1 6600 dprintf("cannot recognize container type %s\n",
ca9de185
LM		 6601 devname);
6602 tmpdev->used = 2;
6603 } else if (tst->ss != st->ss) {
1ade5cc1 6604 dprintf("non-imsm container - ignore it: %s\n",
ca9de185
LM		 6605 devname);
6606 tmpdev->used = 2;
6607 } else if (!tst->ss->load_container ||
6608 tst->ss->load_container(tst, dfd, NULL))
6609 tmpdev->used = 2;
6610 else {
6611 tmpdev->container = 1;
6612 }
6613 if (cst)
6614 cst->ss->free_super(cst);
6615 } else {
6616 tmpdev->st_rdev = stb.st_rdev;
6617 if (tst->ss->load_super(tst,dfd, NULL)) {
1ade5cc1 6618 dprintf("no RAID superblock on %s\n",
ca9de185
LM		 6619 devname);
6620 tmpdev->used = 2;
6621 } else if (tst->ss->compare_super == NULL) {
1ade5cc1 6622 dprintf("Cannot assemble %s metadata on %s\n",
ca9de185
LM		 6623 tst->ss->name, devname);
6624 tmpdev->used = 2;
6625 }
6626 }
6627 if (dfd >= 0)
6628 close(dfd);
6629 if (tmpdev->used == 2 || tmpdev->used == 4) {
6630 /* Ignore unrecognised devices during auto-assembly */
6631 goto loop;
6632 }
6633 else {
6634 struct mdinfo info;
6635 tst->ss->getinfo_super(tst, &info, NULL);
6636
6637 if (st->minor_version == -1)
6638 st->minor_version = tst->minor_version;
6639
6640 if (memcmp(info.uuid, uuid_zero,
6641 sizeof(int[4])) == 0) {
6642 /* this is a floating spare. It cannot define
6643 * an array unless there are no more arrays of
6644 * this type to be found. It can be included
6645 * in an array of this type though.
6646 */
6647 tmpdev->used = 3;
6648 goto loop;
6649 }
6650
6651 if (st->ss != tst->ss ||
6652 st->minor_version != tst->minor_version ||
6653 st->ss->compare_super(st, tst) != 0) {
6654 /* Some mismatch. If exactly one array matches this host,
6655 * we can resolve on that one.
6656 * Or, if we are auto assembling, we just ignore the second
6657 * for now.
6658 */
1ade5cc1 6659 dprintf("superblock on %s doesn't match others - assembly aborted\n",
ca9de185
LM		 6660 devname);
6661 goto loop;
6662 }
6663 tmpdev->used = 1;
6664 *found = 1;
6665 dprintf("found: devname: %s\n", devname);
6666 }
6667 loop:
6668 if (tst)
6669 tst->ss->free_super(tst);
6670 }
6671 if (*found != 0) {
6672 int err;
6673 if ((err = load_super_imsm_all(st, -1, &st->sb, NULL, devlist, 0)) == 0) {
6674 struct mdinfo *iter, *head = st->ss->container_content(st, NULL);
6675 for (iter = head; iter; iter = iter->next) {
6676 dprintf("content->text_version: %s vol\n",
6677 iter->text_version);
6678 if (iter->array.state & (1<<MD_SB_BLOCK_VOLUME)) {
6679 /* do not assemble arrays with unsupported
6680 configurations */
1ade5cc1 6681 dprintf("Cannot activate member %s.\n",
ca9de185
LM		 6682 iter->text_version);
6683 } else
6684 count++;
6685 }
6686 sysfs_free(head);
6687
6688 } else {
1ade5cc1 6689 dprintf("No valid super block on device list: err: %d %p\n",
ca9de185
LM		 6690 err, st->sb);
6691 }
6692 } else {
1ade5cc1 6693 dprintf("no more devices to examine\n");
ca9de185
LM		 6694 }
6695
6696 for (tmpdev = devlist; tmpdev; tmpdev = tmpdev->next) {
089f9d79 6697 if (tmpdev->used == 1 && tmpdev->found) {
ca9de185
LM		 6698 if (count) {
6699 if (count < tmpdev->found)
6700 count = 0;
6701 else
6702 count -= tmpdev->found;
6703 }
6704 }
6705 if (tmpdev->used == 1)
6706 tmpdev->used = 4;
6707 }
6708 err_1:
6709 if (st)
6710 st->ss->free_super(st);
6711 return count;
6712}
6713
d3c11416
AO		 6714static int __count_volumes(char *hba_path, int dpa, int verbose,
6715 int cmp_hba_path)
ca9de185 6716{
72a45777 6717 struct sys_dev *idev, *intel_devices = find_intel_devices();
ca9de185 6718 int count = 0;
72a45777
PB		 6719 const struct orom_entry *entry;
6720 struct devid_list *dv, *devid_list;
ca9de185 6721
d3c11416 6722 if (!hba_path)
ca9de185
LM		 6723 return 0;
6724
72a45777 6725 for (idev = intel_devices; idev; idev = idev->next) {
d3c11416
AO		 6726 if (strstr(idev->path, hba_path))
6727 break;
72a45777
PB		 6728 }
6729
6730 if (!idev || !idev->dev_id)
ca9de185 6731 return 0;
72a45777
PB		 6732
6733 entry = get_orom_entry_by_device_id(idev->dev_id);
6734
6735 if (!entry || !entry->devid_list)
6736 return 0;
6737
6738 devid_list = entry->devid_list;
6739 for (dv = devid_list; dv; dv = dv->next) {
594dc1b8 6740 struct md_list *devlist;
d3c11416
AO		 6741 struct sys_dev *device = NULL;
6742 char *hpath;
72a45777
PB		 6743 int found = 0;
6744
d3c11416
AO		 6745 if (cmp_hba_path)
6746 device = device_by_id_and_path(dv->devid, hba_path);
6747 else
6748 device = device_by_id(dv->devid);
6749
72a45777 6750 if (device)
d3c11416 6751 hpath = device->path;
72a45777
PB		 6752 else
6753 return 0;
6754
d3c11416 6755 devlist = get_devices(hpath);
72a45777
PB		 6756 /* if no intel devices return zero volumes */
6757 if (devlist == NULL)
6758 return 0;
6759
d3c11416
AO		 6760 count += active_arrays_by_format("imsm", hpath, &devlist, dpa,
6761 verbose);
6762 dprintf("path: %s active arrays: %d\n", hpath, count);
72a45777
PB		 6763 if (devlist == NULL)
6764 return 0;
6765 do {
6766 found = 0;
6767 count += count_volumes_list(devlist,
6768 NULL,
6769 verbose,
6770 &found);
6771 dprintf("found %d count: %d\n", found, count);
6772 } while (found);
6773
d3c11416 6774 dprintf("path: %s total number of volumes: %d\n", hpath, count);
72a45777
PB		 6775
6776 while (devlist) {
6777 struct md_list *dv = devlist;
6778 devlist = devlist->next;
6779 free(dv->devname);
6780 free(dv);
6781 }
ca9de185
LM		 6782 }
6783 return count;
6784}
6785
d3c11416
AO		 6786static int count_volumes(struct intel_hba *hba, int dpa, int verbose)
6787{
6788 if (!hba)
6789 return 0;
6790 if (hba->type == SYS_DEV_VMD) {
6791 struct sys_dev *dev;
6792 int count = 0;
6793
6794 for (dev = find_intel_devices(); dev; dev = dev->next) {
6795 if (dev->type == SYS_DEV_VMD)
6796 count += __count_volumes(dev->path, dpa,
6797 verbose, 1);
6798 }
6799 return count;
6800 }
6801 return __count_volumes(hba->path, dpa, verbose, 0);
6802}
6803
cd9d1ac7
DW		 6804static int imsm_default_chunk(const struct imsm_orom *orom)
6805{
6806 /* up to 512 if the plaform supports it, otherwise the platform max.
6807 * 128 if no platform detected
6808 */
6809 int fs = max(7, orom ? fls(orom->sss) : 0);
6810
6811 return min(512, (1 << fs));
6812}
73408129 6813
6592ce37
DW		 6814static int
6815validate_geometry_imsm_orom(struct intel_super *super, int level, int layout,
2cc699af 6816 int raiddisks, int *chunk, unsigned long long size, int verbose)
6592ce37 6817{
660260d0
DW		 6818 /* check/set platform and metadata limits/defaults */
6819 if (super->orom && raiddisks > super->orom->dpa) {
676e87a8 6820 pr_vrb("platform supports a maximum of %d disks per array\n",
660260d0 6821 super->orom->dpa);
73408129
LM		 6822 return 0;
6823 }
6824
5d500228 6825 /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */
660260d0 6826 if (!is_raid_level_supported(super->orom, level, raiddisks)) {
676e87a8 6827 pr_vrb("platform does not support raid%d with %d disk%s\n",
6592ce37
DW		 6828 level, raiddisks, raiddisks > 1 ? "s" : "");
6829 return 0;
6830 }
cd9d1ac7 6831
7ccc4cc4 6832 if (*chunk == 0 || *chunk == UnSet)
cd9d1ac7
DW		 6833 *chunk = imsm_default_chunk(super->orom);
6834
7ccc4cc4 6835 if (super->orom && !imsm_orom_has_chunk(super->orom, *chunk)) {
676e87a8 6836 pr_vrb("platform does not support a chunk size of: %d\n", *chunk);
cd9d1ac7 6837 return 0;
6592ce37 6838 }
cd9d1ac7 6839
6592ce37
DW		 6840 if (layout != imsm_level_to_layout(level)) {
6841 if (level == 5)
676e87a8 6842 pr_vrb("imsm raid 5 only supports the left-asymmetric layout\n");
6592ce37 6843 else if (level == 10)
676e87a8 6844 pr_vrb("imsm raid 10 only supports the n2 layout\n");
6592ce37 6845 else
676e87a8 6846 pr_vrb("imsm unknown layout %#x for this raid level %d\n",
6592ce37
DW		 6847 layout, level);
6848 return 0;
6849 }
2cc699af 6850
7ccc4cc4 6851 if (super->orom && (super->orom->attr & IMSM_OROM_ATTR_2TB) == 0 &&
2cc699af 6852 (calc_array_size(level, raiddisks, layout, *chunk, size) >> 32) > 0) {
676e87a8 6853 pr_vrb("platform does not support a volume size over 2TB\n");
2cc699af
CA		 6854 return 0;
6855 }
614902f6 6856
6592ce37
DW		 6857 return 1;
6858}
6859
1011e834 6860/* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd
c2c087e6
DW		 6861 * FIX ME add ahci details
6862 */
8b353278 6863static int validate_geometry_imsm_volume(struct supertype *st, int level,
c21e737b 6864 int layout, int raiddisks, int *chunk,
af4348dd
N		 6865 unsigned long long size,
6866 unsigned long long data_offset,
6867 char *dev,
2c514b71
NB		 6868 unsigned long long *freesize,
6869 int verbose)
cdddbdbc 6870{
c2c087e6
DW		 6871 struct stat stb;
6872 struct intel_super *super = st->sb;
b2916f25 6873 struct imsm_super *mpb;
c2c087e6
DW		 6874 struct dl *dl;
6875 unsigned long long pos = 0;
6876 unsigned long long maxsize;
6877 struct extent *e;
6878 int i;
cdddbdbc 6879
88c32bb1
DW		 6880 /* We must have the container info already read in. */
6881 if (!super)
c2c087e6
DW		 6882 return 0;
6883
b2916f25
JS		 6884 mpb = super->anchor;
6885
2cc699af 6886 if (!validate_geometry_imsm_orom(super, level, layout, raiddisks, chunk, size, verbose)) {
7a862a02 6887 pr_err("RAID gemetry validation failed. Cannot proceed with the action(s).\n");
c2c087e6 6888 return 0;
d54559f0 6889 }
c2c087e6
DW		 6890 if (!dev) {
6891 /* General test: make sure there is space for
2da8544a
DW		 6892 * 'raiddisks' device extents of size 'size' at a given
6893 * offset
c2c087e6 6894 */
e46273eb 6895 unsigned long long minsize = size;
b7528a20 6896 unsigned long long start_offset = MaxSector;
c2c087e6
DW		 6897 int dcnt = 0;
6898 if (minsize == 0)
6899 minsize = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
6900 for (dl = super->disks; dl ; dl = dl->next) {
6901 int found = 0;
6902
bf5a934a 6903 pos = 0;
c2c087e6
DW		 6904 i = 0;
6905 e = get_extents(super, dl);
6906 if (!e) continue;
6907 do {
6908 unsigned long long esize;
6909 esize = e[i].start - pos;
6910 if (esize >= minsize)
6911 found = 1;
b7528a20 6912 if (found && start_offset == MaxSector) {
2da8544a
DW		 6913 start_offset = pos;
6914 break;
6915 } else if (found && pos != start_offset) {
6916 found = 0;
6917 break;
6918 }
c2c087e6
DW		 6919 pos = e[i].start + e[i].size;
6920 i++;
6921 } while (e[i-1].size);
6922 if (found)
6923 dcnt++;
6924 free(e);
6925 }
6926 if (dcnt < raiddisks) {
2c514b71 6927 if (verbose)
7a862a02 6928 pr_err("imsm: Not enough devices with space for this array (%d < %d)\n",
2c514b71 6929 dcnt, raiddisks);
c2c087e6
DW		 6930 return 0;
6931 }
6932 return 1;
6933 }
0dcecb2e 6934
c2c087e6
DW		 6935 /* This device must be a member of the set */
6936 if (stat(dev, &stb) < 0)
6937 return 0;
6938 if ((S_IFMT & stb.st_mode) != S_IFBLK)
6939 return 0;
6940 for (dl = super->disks ; dl ; dl = dl->next) {
f21e18ca
N		 6941 if (dl->major == (int)major(stb.st_rdev) &&
6942 dl->minor == (int)minor(stb.st_rdev))
c2c087e6
DW		 6943 break;
6944 }
6945 if (!dl) {
2c514b71 6946 if (verbose)
7a862a02 6947 pr_err("%s is not in the same imsm set\n", dev);
c2c087e6 6948 return 0;
a20d2ba5
DW		 6949 } else if (super->orom && dl->index < 0 && mpb->num_raid_devs) {
6950 /* If a volume is present then the current creation attempt
6951 * cannot incorporate new spares because the orom may not
6952 * understand this configuration (all member disks must be
6953 * members of each array in the container).
6954 */
7a862a02
N		 6955 pr_err("%s is a spare and a volume is already defined for this container\n", dev);
6956 pr_err("The option-rom requires all member disks to be a member of all volumes\n");
a20d2ba5 6957 return 0;
5fe62b94
WD		 6958 } else if (super->orom && mpb->num_raid_devs > 0 &&
6959 mpb->num_disks != raiddisks) {
7a862a02 6960 pr_err("The option-rom requires all member disks to be a member of all volumes\n");
5fe62b94 6961 return 0;
c2c087e6 6962 }
0dcecb2e
DW		 6963
6964 /* retrieve the largest free space block */
c2c087e6
DW		 6965 e = get_extents(super, dl);
6966 maxsize = 0;
6967 i = 0;
0dcecb2e
DW		 6968 if (e) {
6969 do {
6970 unsigned long long esize;
6971
6972 esize = e[i].start - pos;
6973 if (esize >= maxsize)
6974 maxsize = esize;
6975 pos = e[i].start + e[i].size;
6976 i++;
6977 } while (e[i-1].size);
6978 dl->e = e;
6979 dl->extent_cnt = i;
6980 } else {
6981 if (verbose)
e7b84f9d 6982 pr_err("unable to determine free space for: %s\n",
0dcecb2e
DW		 6983 dev);
6984 return 0;
6985 }
6986 if (maxsize < size) {
6987 if (verbose)
e7b84f9d 6988 pr_err("%s not enough space (%llu < %llu)\n",
0dcecb2e
DW		 6989 dev, maxsize, size);
6990 return 0;
6991 }
6992
6993 /* count total number of extents for merge */
6994 i = 0;
6995 for (dl = super->disks; dl; dl = dl->next)
6996 if (dl->e)
6997 i += dl->extent_cnt;
6998
6999 maxsize = merge_extents(super, i);
3baa56ab
LO		 7000
7001 if (!check_env("IMSM_NO_PLATFORM") &&
7002 mpb->num_raid_devs > 0 && size && size != maxsize) {
7a862a02 7003 pr_err("attempting to create a second volume with size less then remaining space. Aborting...\n");
3baa56ab
LO		 7004 return 0;
7005 }
7006
a7dd165b 7007 if (maxsize < size || maxsize == 0) {
b3071342
LD		 7008 if (verbose) {
7009 if (maxsize == 0)
7a862a02 7010 pr_err("no free space left on device. Aborting...\n");
b3071342 7011 else
7a862a02 7012 pr_err("not enough space to create volume of given size (%llu < %llu). Aborting...\n",
b3071342
LD		 7013 maxsize, size);
7014 }
0dcecb2e 7015 return 0;
0dcecb2e
DW		 7016 }
7017
c2c087e6
DW		 7018 *freesize = maxsize;
7019
ca9de185 7020 if (super->orom) {
72a45777 7021 int count = count_volumes(super->hba,
ca9de185
LM		 7022 super->orom->dpa, verbose);
7023 if (super->orom->vphba <= count) {
676e87a8 7024 pr_vrb("platform does not support more than %d raid volumes.\n",
ca9de185
LM		 7025 super->orom->vphba);
7026 return 0;
7027 }
7028 }
c2c087e6 7029 return 1;
cdddbdbc
DW		 7030}
7031
13bcac90 7032static int imsm_get_free_size(struct supertype *st, int raiddisks,
efb30e7f
DW		 7033 unsigned long long size, int chunk,
7034 unsigned long long *freesize)
7035{
7036 struct intel_super *super = st->sb;
7037 struct imsm_super *mpb = super->anchor;
7038 struct dl *dl;
7039 int i;
7040 int extent_cnt;
7041 struct extent *e;
7042 unsigned long long maxsize;
7043 unsigned long long minsize;
7044 int cnt;
7045 int used;
7046
7047 /* find the largest common start free region of the possible disks */
7048 used = 0;
7049 extent_cnt = 0;
7050 cnt = 0;
7051 for (dl = super->disks; dl; dl = dl->next) {
7052 dl->raiddisk = -1;
7053
7054 if (dl->index >= 0)
7055 used++;
7056
7057 /* don't activate new spares if we are orom constrained
7058 * and there is already a volume active in the container
7059 */
7060 if (super->orom && dl->index < 0 && mpb->num_raid_devs)
7061 continue;
7062
7063 e = get_extents(super, dl);
7064 if (!e)
7065 continue;
7066 for (i = 1; e[i-1].size; i++)
7067 ;
7068 dl->e = e;
7069 dl->extent_cnt = i;
7070 extent_cnt += i;
7071 cnt++;
7072 }
7073
7074 maxsize = merge_extents(super, extent_cnt);
7075 minsize = size;
7076 if (size == 0)
612e59d8
CA		 7077 /* chunk is in K */
7078 minsize = chunk * 2;
efb30e7f
DW		 7079
7080 if (cnt < raiddisks ||
7081 (super->orom && used && used != raiddisks) ||
a7dd165b
DW		 7082 maxsize < minsize ||
7083 maxsize == 0) {
e7b84f9d 7084 pr_err("not enough devices with space to create array.\n");
efb30e7f
DW		 7085 return 0; /* No enough free spaces large enough */
7086 }
7087
7088 if (size == 0) {
7089 size = maxsize;
7090 if (chunk) {
612e59d8
CA		 7091 size /= 2 * chunk;
7092 size *= 2 * chunk;
efb30e7f 7093 }
f878b242
LM		 7094 maxsize = size;
7095 }
7096 if (!check_env("IMSM_NO_PLATFORM") &&
7097 mpb->num_raid_devs > 0 && size && size != maxsize) {
7a862a02 7098 pr_err("attempting to create a second volume with size less then remaining space. Aborting...\n");
f878b242 7099 return 0;
efb30e7f 7100 }
efb30e7f
DW		 7101 cnt = 0;
7102 for (dl = super->disks; dl; dl = dl->next)
7103 if (dl->e)
7104 dl->raiddisk = cnt++;
7105
7106 *freesize = size;
7107
13bcac90
AK		 7108 dprintf("imsm: imsm_get_free_size() returns : %llu\n", size);
7109
efb30e7f
DW		 7110 return 1;
7111}
7112
13bcac90
AK		 7113static int reserve_space(struct supertype *st, int raiddisks,
7114 unsigned long long size, int chunk,
7115 unsigned long long *freesize)
7116{
7117 struct intel_super *super = st->sb;
7118 struct dl *dl;
7119 int cnt;
7120 int rv = 0;
7121
7122 rv = imsm_get_free_size(st, raiddisks, size, chunk, freesize);
7123 if (rv) {
7124 cnt = 0;
7125 for (dl = super->disks; dl; dl = dl->next)
7126 if (dl->e)
7127 dl->raiddisk = cnt++;
7128 rv = 1;
7129 }
7130
7131 return rv;
7132}
7133
bf5a934a 7134static int validate_geometry_imsm(struct supertype *st, int level, int layout,
c21e737b 7135 int raiddisks, int *chunk, unsigned long long size,
af4348dd 7136 unsigned long long data_offset,
bf5a934a 7137 char *dev, unsigned long long *freesize,
5308f117 7138 int consistency_policy, int verbose)
bf5a934a
DW		 7139{
7140 int fd, cfd;
7141 struct mdinfo *sra;
20cbe8d2 7142 int is_member = 0;
bf5a934a 7143
d54559f0
LM		 7144 /* load capability
7145 * if given unused devices create a container
bf5a934a
DW		 7146 * if given given devices in a container create a member volume
7147 */
7148 if (level == LEVEL_CONTAINER) {
7149 /* Must be a fresh device to add to a container */
7150 return validate_geometry_imsm_container(st, level, layout,
c21e737b 7151 raiddisks,
7ccc4cc4 7152 *chunk,
af4348dd 7153 size, data_offset,
bf5a934a
DW		 7154 dev, freesize,
7155 verbose);
7156 }
9587c373 7157
8592f29d 7158 if (!dev) {
e91a3bad 7159 if (st->sb) {
ca9de185 7160 struct intel_super *super = st->sb;
e91a3bad 7161 if (!validate_geometry_imsm_orom(st->sb, level, layout,
2cc699af 7162 raiddisks, chunk, size,
e91a3bad
LM		 7163 verbose))
7164 return 0;
efb30e7f
DW		 7165 /* we are being asked to automatically layout a
7166 * new volume based on the current contents of
7167 * the container. If the the parameters can be
7168 * satisfied reserve_space will record the disks,
7169 * start offset, and size of the volume to be
7170 * created. add_to_super and getinfo_super
7171 * detect when autolayout is in progress.
7172 */
ca9de185
LM		 7173 /* assuming that freesize is always given when array is
7174 created */
7175 if (super->orom && freesize) {
7176 int count;
72a45777 7177 count = count_volumes(super->hba,
ca9de185
LM		 7178 super->orom->dpa, verbose);
7179 if (super->orom->vphba <= count) {
676e87a8 7180 pr_vrb("platform does not support more than %d raid volumes.\n",
ca9de185
LM		 7181 super->orom->vphba);
7182 return 0;
7183 }
7184 }
e91a3bad
LM		 7185 if (freesize)
7186 return reserve_space(st, raiddisks, size,
7ccc4cc4 7187 *chunk, freesize);
8592f29d
N		 7188 }
7189 return 1;
7190 }
bf5a934a
DW		 7191 if (st->sb) {
7192 /* creating in a given container */
7193 return validate_geometry_imsm_volume(st, level, layout,
7194 raiddisks, chunk, size,
af4348dd 7195 data_offset,
bf5a934a
DW		 7196 dev, freesize, verbose);
7197 }
7198
bf5a934a
DW		 7199 /* This device needs to be a device in an 'imsm' container */
7200 fd = open(dev, O_RDONLY|O_EXCL, 0);
7201 if (fd >= 0) {
7202 if (verbose)
e7b84f9d
N		 7203 pr_err("Cannot create this array on device %s\n",
7204 dev);
bf5a934a
DW		 7205 close(fd);
7206 return 0;
7207 }
7208 if (errno != EBUSY || (fd = open(dev, O_RDONLY, 0)) < 0) {
7209 if (verbose)
e7b84f9d 7210 pr_err("Cannot open %s: %s\n",
bf5a934a
DW		 7211 dev, strerror(errno));
7212 return 0;
7213 }
7214 /* Well, it is in use by someone, maybe an 'imsm' container. */
7215 cfd = open_container(fd);
20cbe8d2 7216 close(fd);
bf5a934a 7217 if (cfd < 0) {
bf5a934a 7218 if (verbose)
e7b84f9d 7219 pr_err("Cannot use %s: It is busy\n",
bf5a934a
DW		 7220 dev);
7221 return 0;
7222 }
4dd2df09 7223 sra = sysfs_read(cfd, NULL, GET_VERSION);
bf5a934a 7224 if (sra && sra->array.major_version == -1 &&
20cbe8d2
AW		 7225 strcmp(sra->text_version, "imsm") == 0)
7226 is_member = 1;
7227 sysfs_free(sra);
7228 if (is_member) {
bf5a934a
DW		 7229 /* This is a member of a imsm container. Load the container
7230 * and try to create a volume
7231 */
7232 struct intel_super *super;
7233
ec50f7b6 7234 if (load_super_imsm_all(st, cfd, (void **) &super, NULL, NULL, 1) == 0) {
bf5a934a 7235 st->sb = super;
4dd2df09 7236 strcpy(st->container_devnm, fd2devnm(cfd));
bf5a934a
DW		 7237 close(cfd);
7238 return validate_geometry_imsm_volume(st, level, layout,
7239 raiddisks, chunk,
af4348dd 7240 size, data_offset, dev,
ecbd9e81
N		 7241 freesize, 1)
7242 ? 1 : -1;
bf5a934a 7243 }
20cbe8d2 7244 }
bf5a934a 7245
20cbe8d2 7246 if (verbose)
e7b84f9d 7247 pr_err("failed container membership check\n");
20cbe8d2
AW		 7248
7249 close(cfd);
7250 return 0;
bf5a934a 7251}
0bd16cf2 7252
30f58b22 7253static void default_geometry_imsm(struct supertype *st, int *level, int *layout, int *chunk)
0bd16cf2
DJ		 7254{
7255 struct intel_super *super = st->sb;
7256
30f58b22
DW		 7257 if (level && *level == UnSet)
7258 *level = LEVEL_CONTAINER;
7259
7260 if (level && layout && *layout == UnSet)
7261 *layout = imsm_level_to_layout(*level);
0bd16cf2 7262
cd9d1ac7
DW		 7263 if (chunk && (*chunk == UnSet || *chunk == 0))
7264 *chunk = imsm_default_chunk(super->orom);
0bd16cf2
DJ		 7265}
7266
33414a01
DW		 7267static void handle_missing(struct intel_super *super, struct imsm_dev *dev);
7268
7269static int kill_subarray_imsm(struct supertype *st)
7270{
7271 /* remove the subarray currently referenced by ->current_vol */
7272 __u8 i;
7273 struct intel_dev **dp;
7274 struct intel_super *super = st->sb;
7275 __u8 current_vol = super->current_vol;
7276 struct imsm_super *mpb = super->anchor;
7277
7278 if (super->current_vol < 0)
7279 return 2;
7280 super->current_vol = -1; /* invalidate subarray cursor */
7281
7282 /* block deletions that would change the uuid of active subarrays
7283 *
7284 * FIXME when immutable ids are available, but note that we'll
7285 * also need to fixup the invalidated/active subarray indexes in
7286 * mdstat
7287 */
7288 for (i = 0; i < mpb->num_raid_devs; i++) {
7289 char subarray[4];
7290
7291 if (i < current_vol)
7292 continue;
7293 sprintf(subarray, "%u", i);
4dd2df09 7294 if (is_subarray_active(subarray, st->devnm)) {
e7b84f9d
N		 7295 pr_err("deleting subarray-%d would change the UUID of active subarray-%d, aborting\n",
7296 current_vol, i);
33414a01
DW		 7297
7298 return 2;
7299 }
7300 }
7301
7302 if (st->update_tail) {
503975b9 7303 struct imsm_update_kill_array *u = xmalloc(sizeof(*u));
33414a01 7304
33414a01
DW		 7305 u->type = update_kill_array;
7306 u->dev_idx = current_vol;
7307 append_metadata_update(st, u, sizeof(*u));
7308
7309 return 0;
7310 }
7311
7312 for (dp = &super->devlist; *dp;)
7313 if ((*dp)->index == current_vol) {
7314 *dp = (*dp)->next;
7315 } else {
7316 handle_missing(super, (*dp)->dev);
7317 if ((*dp)->index > current_vol)
7318 (*dp)->index--;
7319 dp = &(*dp)->next;
7320 }
7321
7322 /* no more raid devices, all active components are now spares,
7323 * but of course failed are still failed
7324 */
7325 if (--mpb->num_raid_devs == 0) {
7326 struct dl *d;
7327
7328 for (d = super->disks; d; d = d->next)
a8619d23
AK		 7329 if (d->index > -2)
7330 mark_spare(d);
33414a01
DW		 7331 }
7332
7333 super->updates_pending++;
7334
7335 return 0;
7336}
aa534678 7337
a951a4f7 7338static int update_subarray_imsm(struct supertype *st, char *subarray,
fa56eddb 7339 char *update, struct mddev_ident *ident)
aa534678
DW		 7340{
7341 /* update the subarray currently referenced by ->current_vol */
7342 struct intel_super *super = st->sb;
7343 struct imsm_super *mpb = super->anchor;
7344
aa534678
DW		 7345 if (strcmp(update, "name") == 0) {
7346 char *name = ident->name;
a951a4f7
N		 7347 char *ep;
7348 int vol;
aa534678 7349
4dd2df09 7350 if (is_subarray_active(subarray, st->devnm)) {
e7b84f9d 7351 pr_err("Unable to update name of active subarray\n");
aa534678
DW		 7352 return 2;
7353 }
7354
7355 if (!check_name(super, name, 0))
7356 return 2;
7357
a951a4f7
N		 7358 vol = strtoul(subarray, &ep, 10);
7359 if (*ep != '\0' || vol >= super->anchor->num_raid_devs)
7360 return 2;
7361
aa534678 7362 if (st->update_tail) {
503975b9 7363 struct imsm_update_rename_array *u = xmalloc(sizeof(*u));
aa534678 7364
aa534678 7365 u->type = update_rename_array;
a951a4f7 7366 u->dev_idx = vol;
618f4e6d
XN		 7367 strncpy((char *) u->name, name, MAX_RAID_SERIAL_LEN);
7368 u->name[MAX_RAID_SERIAL_LEN-1] = '\0';
aa534678
DW		 7369 append_metadata_update(st, u, sizeof(*u));
7370 } else {
7371 struct imsm_dev *dev;
7372 int i;
7373
a951a4f7 7374 dev = get_imsm_dev(super, vol);
618f4e6d
XN		 7375 strncpy((char *) dev->volume, name, MAX_RAID_SERIAL_LEN);
7376 dev->volume[MAX_RAID_SERIAL_LEN-1] = '\0';
aa534678
DW		 7377 for (i = 0; i < mpb->num_raid_devs; i++) {
7378 dev = get_imsm_dev(super, i);
7379 handle_missing(super, dev);
7380 }
7381 super->updates_pending++;
7382 }
e6e9dd3f
AP		 7383 } else if (strcmp(update, "ppl") == 0 ||
7384 strcmp(update, "no-ppl") == 0) {
7385 int new_policy;
7386 char *ep;
7387 int vol = strtoul(subarray, &ep, 10);
7388
7389 if (*ep != '\0' || vol >= super->anchor->num_raid_devs)
7390 return 2;
7391
7392 if (strcmp(update, "ppl") == 0)
7393 new_policy = RWH_DISTRIBUTED;
7394 else
7395 new_policy = RWH_OFF;
7396
7397 if (st->update_tail) {
7398 struct imsm_update_rwh_policy *u = xmalloc(sizeof(*u));
7399
7400 u->type = update_rwh_policy;
7401 u->dev_idx = vol;
7402 u->new_policy = new_policy;
7403 append_metadata_update(st, u, sizeof(*u));
7404 } else {
7405 struct imsm_dev *dev;
7406
7407 dev = get_imsm_dev(super, vol);
7408 dev->rwh_policy = new_policy;
7409 super->updates_pending++;
7410 }
aa534678
DW		 7411 } else
7412 return 2;
7413
7414 return 0;
7415}
d1e02575 7416#endif /* MDASSEMBLE */
bf5a934a 7417
28bce06f
AK		 7418static int is_gen_migration(struct imsm_dev *dev)
7419{
7534230b
AK		 7420 if (dev == NULL)
7421 return 0;
7422
28bce06f
AK		 7423 if (!dev->vol.migr_state)
7424 return 0;
7425
7426 if (migr_type(dev) == MIGR_GEN_MIGR)
7427 return 1;
7428
7429 return 0;
7430}
7431
1e5c6983
DW		 7432static int is_rebuilding(struct imsm_dev *dev)
7433{
7434 struct imsm_map *migr_map;
7435
7436 if (!dev->vol.migr_state)
7437 return 0;
7438
7439 if (migr_type(dev) != MIGR_REBUILD)
7440 return 0;
7441
238c0a71 7442 migr_map = get_imsm_map(dev, MAP_1);
1e5c6983
DW		 7443
7444 if (migr_map->map_state == IMSM_T_STATE_DEGRADED)
7445 return 1;
7446 else
7447 return 0;
7448}
7449
b4ab44d8 7450#ifndef MDASSEMBLE
6ce1fbf1
AK		 7451static int is_initializing(struct imsm_dev *dev)
7452{
7453 struct imsm_map *migr_map;
7454
7455 if (!dev->vol.migr_state)
7456 return 0;
7457
7458 if (migr_type(dev) != MIGR_INIT)
7459 return 0;
7460
238c0a71 7461 migr_map = get_imsm_map(dev, MAP_1);
6ce1fbf1
AK		 7462
7463 if (migr_map->map_state == IMSM_T_STATE_UNINITIALIZED)
7464 return 1;
7465
7466 return 0;
6ce1fbf1 7467}
b4ab44d8 7468#endif
6ce1fbf1 7469
c47b0ff6
AK		 7470static void update_recovery_start(struct intel_super *super,
7471 struct imsm_dev *dev,
7472 struct mdinfo *array)
1e5c6983
DW		 7473{
7474 struct mdinfo *rebuild = NULL;
7475 struct mdinfo *d;
7476 __u32 units;
7477
7478 if (!is_rebuilding(dev))
7479 return;
7480
7481 /* Find the rebuild target, but punt on the dual rebuild case */
7482 for (d = array->devs; d; d = d->next)
7483 if (d->recovery_start == 0) {
7484 if (rebuild)
7485 return;
7486 rebuild = d;
7487 }
7488
4363fd80
DW		 7489 if (!rebuild) {
7490 /* (?) none of the disks are marked with
7491 * IMSM_ORD_REBUILD, so assume they are missing and the
7492 * disk_ord_tbl was not correctly updated
7493 */
1ade5cc1 7494 dprintf("failed to locate out-of-sync disk\n");
4363fd80
DW		 7495 return;
7496 }
7497
1e5c6983 7498 units = __le32_to_cpu(dev->vol.curr_migr_unit);
c47b0ff6 7499 rebuild->recovery_start = units * blocks_per_migr_unit(super, dev);
1e5c6983
DW		 7500}
7501
9e2d750d 7502#ifndef MDASSEMBLE
276d77db 7503static int recover_backup_imsm(struct supertype *st, struct mdinfo *info);
9e2d750d 7504#endif
1e5c6983 7505
00bbdbda 7506static struct mdinfo *container_content_imsm(struct supertype *st, char *subarray)
cdddbdbc 7507{
4f5bc454
DW		 7508 /* Given a container loaded by load_super_imsm_all,
7509 * extract information about all the arrays into
7510 * an mdinfo tree.
00bbdbda 7511 * If 'subarray' is given, just extract info about that array.
4f5bc454
DW		 7512 *
7513 * For each imsm_dev create an mdinfo, fill it in,
7514 * then look for matching devices in super->disks
7515 * and create appropriate device mdinfo.
7516 */
7517 struct intel_super *super = st->sb;
949c47a0 7518 struct imsm_super *mpb = super->anchor;
4f5bc454 7519 struct mdinfo *rest = NULL;
00bbdbda 7520 unsigned int i;
81219e70 7521 int sb_errors = 0;
abef11a3
AK		 7522 struct dl *d;
7523 int spare_disks = 0;
cdddbdbc 7524
19482bcc
AK		 7525 /* do not assemble arrays when not all attributes are supported */
7526 if (imsm_check_attributes(mpb->attributes) == 0) {
81219e70 7527 sb_errors = 1;
7a862a02 7528 pr_err("Unsupported attributes in IMSM metadata.Arrays activation is blocked.\n");
19482bcc
AK		 7529 }
7530
abef11a3
AK		 7531 /* count spare devices, not used in maps
7532 */
7533 for (d = super->disks; d; d = d->next)
7534 if (d->index == -1)
7535 spare_disks++;
7536
4f5bc454 7537 for (i = 0; i < mpb->num_raid_devs; i++) {
00bbdbda
N		 7538 struct imsm_dev *dev;
7539 struct imsm_map *map;
86e3692b 7540 struct imsm_map *map2;
4f5bc454 7541 struct mdinfo *this;
a6482415
N		 7542 int slot;
7543#ifndef MDASSEMBLE
7544 int chunk;
7545#endif
00bbdbda
N		 7546 char *ep;
7547
7548 if (subarray &&
7549 (i != strtoul(subarray, &ep, 10) || *ep != '\0'))
7550 continue;
7551
7552 dev = get_imsm_dev(super, i);
238c0a71
AK		 7553 map = get_imsm_map(dev, MAP_0);
7554 map2 = get_imsm_map(dev, MAP_1);
4f5bc454 7555
1ce0101c
DW		 7556 /* do not publish arrays that are in the middle of an
7557 * unsupported migration
7558 */
7559 if (dev->vol.migr_state &&
28bce06f 7560 (migr_type(dev) == MIGR_STATE_CHANGE)) {
7a862a02 7561 pr_err("cannot assemble volume '%.16s': unsupported migration in progress\n",
1ce0101c
DW		 7562 dev->volume);
7563 continue;
7564 }
2db86302
LM		 7565 /* do not publish arrays that are not support by controller's
7566 * OROM/EFI
7567 */
1ce0101c 7568
503975b9 7569 this = xmalloc(sizeof(*this));
4f5bc454 7570
301406c9 7571 super->current_vol = i;
a5d85af7 7572 getinfo_super_imsm_volume(st, this, NULL);
9894ec0d 7573 this->next = rest;
81219e70 7574#ifndef MDASSEMBLE
a6482415 7575 chunk = __le16_to_cpu(map->blocks_per_strip) >> 1;
81219e70
LM		 7576 /* mdadm does not support all metadata features- set the bit in all arrays state */
7577 if (!validate_geometry_imsm_orom(super,
7578 get_imsm_raid_level(map), /* RAID level */
7579 imsm_level_to_layout(get_imsm_raid_level(map)),
7580 map->num_members, /* raid disks */
2cc699af 7581 &chunk, join_u32(dev->size_low, dev->size_high),
81219e70 7582 1 /* verbose */)) {
7a862a02 7583 pr_err("IMSM RAID geometry validation failed. Array %s activation is blocked.\n",
81219e70
LM		 7584 dev->volume);
7585 this->array.state |=
7586 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
7587 (1<<MD_SB_BLOCK_VOLUME);
7588 }
7589#endif
7590
7591 /* if array has bad blocks, set suitable bit in all arrays state */
7592 if (sb_errors)
7593 this->array.state |=
7594 (1<<MD_SB_BLOCK_CONTAINER_RESHAPE) |
7595 (1<<MD_SB_BLOCK_VOLUME);
7596
4f5bc454 7597 for (slot = 0 ; slot < map->num_members; slot++) {
1e5c6983 7598 unsigned long long recovery_start;
4f5bc454
DW		 7599 struct mdinfo *info_d;
7600 struct dl *d;
7601 int idx;
9a1608e5 7602 int skip;
7eef0453 7603 __u32 ord;
4f5bc454 7604
9a1608e5 7605 skip = 0;
238c0a71
AK		 7606 idx = get_imsm_disk_idx(dev, slot, MAP_0);
7607 ord = get_imsm_ord_tbl_ent(dev, slot, MAP_X);
4f5bc454
DW		 7608 for (d = super->disks; d ; d = d->next)
7609 if (d->index == idx)
0fbd635c 7610 break;
4f5bc454 7611
1e5c6983 7612 recovery_start = MaxSector;
4f5bc454 7613 if (d == NULL)
9a1608e5 7614 skip = 1;
25ed7e59 7615 if (d && is_failed(&d->disk))
9a1608e5 7616 skip = 1;
7eef0453 7617 if (ord & IMSM_ORD_REBUILD)
1e5c6983 7618 recovery_start = 0;
9a1608e5 7619
1011e834 7620 /*
9a1608e5 7621 * if we skip some disks the array will be assmebled degraded;
1e5c6983
DW		 7622 * reset resync start to avoid a dirty-degraded
7623 * situation when performing the intial sync
9a1608e5
DW		 7624 *
7625 * FIXME handle dirty degraded
7626 */
2432ce9b
AP		 7627 if ((skip || recovery_start == 0) &&
7628 !(dev->vol.dirty & RAIDVOL_DIRTY))
b7528a20 7629 this->resync_start = MaxSector;
9a1608e5
DW		 7630 if (skip)
7631 continue;
4f5bc454 7632
503975b9 7633 info_d = xcalloc(1, sizeof(*info_d));
4f5bc454
DW		 7634 info_d->next = this->devs;
7635 this->devs = info_d;
7636
4f5bc454
DW		 7637 info_d->disk.number = d->index;
7638 info_d->disk.major = d->major;
7639 info_d->disk.minor = d->minor;
7640 info_d->disk.raid_disk = slot;
1e5c6983 7641 info_d->recovery_start = recovery_start;
86e3692b
AK		 7642 if (map2) {
7643 if (slot < map2->num_members)
7644 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK		 7645 else
7646 this->array.spare_disks++;
86e3692b
AK		 7647 } else {
7648 if (slot < map->num_members)
7649 info_d->disk.state = (1 << MD_DISK_ACTIVE);
04c3c514
AK		 7650 else
7651 this->array.spare_disks++;
86e3692b 7652 }
1e5c6983
DW		 7653 if (info_d->recovery_start == MaxSector)
7654 this->array.working_disks++;
4f5bc454
DW		 7655
7656 info_d->events = __le32_to_cpu(mpb->generation_num);
5551b113 7657 info_d->data_offset = pba_of_lba0(map);
06fb291a
PB		 7658
7659 if (map->raid_level == 5) {
7660 info_d->component_size =
7661 num_data_stripes(map) *
7662 map->blocks_per_strip;
2432ce9b
AP		 7663 info_d->ppl_sector = this->ppl_sector;
7664 info_d->ppl_size = this->ppl_size;
06fb291a
PB		 7665 } else {
7666 info_d->component_size = blocks_per_member(map);
7667 }
2432ce9b 7668 info_d->consistency_policy = this->consistency_policy;
b12796be 7669
5e46202e 7670 info_d->bb.supported = 1;
b12796be
TM		 7671 get_volume_badblocks(super->bbm_log, ord_to_idx(ord),
7672 info_d->data_offset,
7673 info_d->component_size,
7674 &info_d->bb);
4f5bc454 7675 }
1e5c6983 7676 /* now that the disk list is up-to-date fixup recovery_start */
c47b0ff6 7677 update_recovery_start(super, dev, this);
abef11a3 7678 this->array.spare_disks += spare_disks;
276d77db 7679
9e2d750d 7680#ifndef MDASSEMBLE
276d77db
AK		 7681 /* check for reshape */
7682 if (this->reshape_active == 1)
7683 recover_backup_imsm(st, this);
9e2d750d 7684#endif
9a1608e5 7685 rest = this;
4f5bc454
DW		 7686 }
7687
7688 return rest;
cdddbdbc
DW		 7689}
7690
3b451610
AK		 7691static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev,
7692 int failed, int look_in_map)
c2a1e7da 7693{
3b451610
AK		 7694 struct imsm_map *map;
7695
7696 map = get_imsm_map(dev, look_in_map);
c2a1e7da
DW		 7697
7698 if (!failed)
1011e834 7699 return map->map_state == IMSM_T_STATE_UNINITIALIZED ?
3393c6af 7700 IMSM_T_STATE_UNINITIALIZED : IMSM_T_STATE_NORMAL;
c2a1e7da
DW		 7701
7702 switch (get_imsm_raid_level(map)) {
7703 case 0:
7704 return IMSM_T_STATE_FAILED;
7705 break;
7706 case 1:
7707 if (failed < map->num_members)
7708 return IMSM_T_STATE_DEGRADED;
7709 else
7710 return IMSM_T_STATE_FAILED;
7711 break;
7712 case 10:
7713 {
7714 /**
c92a2527
DW		 7715 * check to see if any mirrors have failed, otherwise we
7716 * are degraded. Even numbered slots are mirrored on
7717 * slot+1
c2a1e7da 7718 */
c2a1e7da 7719 int i;
d9b420a5
N		 7720 /* gcc -Os complains that this is unused */
7721 int insync = insync;
c2a1e7da
DW		 7722
7723 for (i = 0; i < map->num_members; i++) {
238c0a71 7724 __u32 ord = get_imsm_ord_tbl_ent(dev, i, MAP_X);
c92a2527
DW		 7725 int idx = ord_to_idx(ord);
7726 struct imsm_disk *disk;
c2a1e7da 7727
c92a2527 7728 /* reset the potential in-sync count on even-numbered
1011e834 7729 * slots. num_copies is always 2 for imsm raid10
c92a2527
DW		 7730 */
7731 if ((i & 1) == 0)
7732 insync = 2;
c2a1e7da 7733
c92a2527 7734 disk = get_imsm_disk(super, idx);
25ed7e59 7735 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
c92a2527 7736 insync--;
c2a1e7da 7737
c92a2527
DW		 7738 /* no in-sync disks left in this mirror the
7739 * array has failed
7740 */
7741 if (insync == 0)
7742 return IMSM_T_STATE_FAILED;
c2a1e7da
DW		 7743 }
7744
7745 return IMSM_T_STATE_DEGRADED;
7746 }
7747 case 5:
7748 if (failed < 2)
7749 return IMSM_T_STATE_DEGRADED;
7750 else
7751 return IMSM_T_STATE_FAILED;
7752 break;
7753 default:
7754 break;
7755 }
7756
7757 return map->map_state;
7758}
7759
3b451610
AK		 7760static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev,
7761 int look_in_map)
c2a1e7da
DW		 7762{
7763 int i;
7764 int failed = 0;
7765 struct imsm_disk *disk;
d5985138
AK		 7766 struct imsm_map *map = get_imsm_map(dev, MAP_0);
7767 struct imsm_map *prev = get_imsm_map(dev, MAP_1);
68fe4598 7768 struct imsm_map *map_for_loop;
0556e1a2
DW		 7769 __u32 ord;
7770 int idx;
d5985138 7771 int idx_1;
c2a1e7da 7772
0556e1a2
DW		 7773 /* at the beginning of migration we set IMSM_ORD_REBUILD on
7774 * disks that are being rebuilt. New failures are recorded to
7775 * map[0]. So we look through all the disks we started with and
7776 * see if any failures are still present, or if any new ones
7777 * have arrived
0556e1a2 7778 */
d5985138
AK		 7779 map_for_loop = map;
7780 if (prev && (map->num_members < prev->num_members))
7781 map_for_loop = prev;
68fe4598
LD		 7782
7783 for (i = 0; i < map_for_loop->num_members; i++) {
d5985138 7784 idx_1 = -255;
238c0a71
AK		 7785 /* when MAP_X is passed both maps failures are counted
7786 */
d5985138 7787 if (prev &&
089f9d79
JS		 7788 (look_in_map == MAP_1 || look_in_map == MAP_X) &&
7789 i < prev->num_members) {
d5985138
AK		 7790 ord = __le32_to_cpu(prev->disk_ord_tbl[i]);
7791 idx_1 = ord_to_idx(ord);
c2a1e7da 7792
d5985138
AK		 7793 disk = get_imsm_disk(super, idx_1);
7794 if (!disk || is_failed(disk) || ord & IMSM_ORD_REBUILD)
7795 failed++;
7796 }
089f9d79
JS		 7797 if ((look_in_map == MAP_0 || look_in_map == MAP_X) &&
7798 i < map->num_members) {
d5985138
AK		 7799 ord = __le32_to_cpu(map->disk_ord_tbl[i]);
7800 idx = ord_to_idx(ord);
7801
7802 if (idx != idx_1) {
7803 disk = get_imsm_disk(super, idx);
7804 if (!disk || is_failed(disk) ||
7805 ord & IMSM_ORD_REBUILD)
7806 failed++;
7807 }
7808 }
c2a1e7da
DW		 7809 }
7810
7811 return failed;
845dea95
NB		 7812}
7813
97b4d0e9
DW		 7814#ifndef MDASSEMBLE
7815static int imsm_open_new(struct supertype *c, struct active_array *a,
7816 char *inst)
7817{
7818 struct intel_super *super = c->sb;
7819 struct imsm_super *mpb = super->anchor;
bbab0940 7820 struct imsm_update_prealloc_bb_mem u;
9587c373 7821
97b4d0e9 7822 if (atoi(inst) >= mpb->num_raid_devs) {
1ade5cc1 7823 pr_err("subarry index %d, out of range\n", atoi(inst));
97b4d0e9
DW		 7824 return -ENODEV;
7825 }
7826
7827 dprintf("imsm: open_new %s\n", inst);
7828 a->info.container_member = atoi(inst);
bbab0940
TM		 7829
7830 u.type = update_prealloc_badblocks_mem;
7831 imsm_update_metadata_locally(c, &u, sizeof(u));
7832
97b4d0e9
DW		 7833 return 0;
7834}
7835
0c046afd
DW		 7836static int is_resyncing(struct imsm_dev *dev)
7837{
7838 struct imsm_map *migr_map;
7839
7840 if (!dev->vol.migr_state)
7841 return 0;
7842
1484e727
DW		 7843 if (migr_type(dev) == MIGR_INIT ||
7844 migr_type(dev) == MIGR_REPAIR)
0c046afd
DW		 7845 return 1;
7846
4c9bc37b
AK		 7847 if (migr_type(dev) == MIGR_GEN_MIGR)
7848 return 0;
7849
238c0a71 7850 migr_map = get_imsm_map(dev, MAP_1);
0c046afd 7851
089f9d79
JS		 7852 if (migr_map->map_state == IMSM_T_STATE_NORMAL &&
7853 dev->vol.migr_type != MIGR_GEN_MIGR)
0c046afd
DW		 7854 return 1;
7855 else
7856 return 0;
7857}
7858
0556e1a2 7859/* return true if we recorded new information */
4c9e8c1e
TM		 7860static int mark_failure(struct intel_super *super,
7861 struct imsm_dev *dev, struct imsm_disk *disk, int idx)
47ee5a45 7862{
0556e1a2
DW		 7863 __u32 ord;
7864 int slot;
7865 struct imsm_map *map;
86c54047
DW		 7866 char buf[MAX_RAID_SERIAL_LEN+3];
7867 unsigned int len, shift = 0;
0556e1a2
DW		 7868
7869 /* new failures are always set in map[0] */
238c0a71 7870 map = get_imsm_map(dev, MAP_0);
0556e1a2
DW		 7871
7872 slot = get_imsm_disk_slot(map, idx);
7873 if (slot < 0)
7874 return 0;
7875
7876 ord = __le32_to_cpu(map->disk_ord_tbl[slot]);
25ed7e59 7877 if (is_failed(disk) && (ord & IMSM_ORD_REBUILD))
0556e1a2
DW		 7878 return 0;
7879
7d0c5e24
LD		 7880 memcpy(buf, disk->serial, MAX_RAID_SERIAL_LEN);
7881 buf[MAX_RAID_SERIAL_LEN] = '\000';
7882 strcat(buf, ":0");
86c54047
DW		 7883 if ((len = strlen(buf)) >= MAX_RAID_SERIAL_LEN)
7884 shift = len - MAX_RAID_SERIAL_LEN + 1;
7885 strncpy((char *)disk->serial, &buf[shift], MAX_RAID_SERIAL_LEN);
7886
f2f27e63 7887 disk->status |= FAILED_DISK;
0556e1a2 7888 set_imsm_ord_tbl_ent(map, slot, idx | IMSM_ORD_REBUILD);
17788994
AK		 7889 /* mark failures in second map if second map exists and this disk
7890 * in this slot.
7891 * This is valid for migration, initialization and rebuild
7892 */
7893 if (dev->vol.migr_state) {
238c0a71 7894 struct imsm_map *map2 = get_imsm_map(dev, MAP_1);
0a108d63
AK		 7895 int slot2 = get_imsm_disk_slot(map2, idx);
7896
089f9d79 7897 if (slot2 < map2->num_members && slot2 >= 0)
0a108d63 7898 set_imsm_ord_tbl_ent(map2, slot2,
1ace8403
AK		 7899 idx | IMSM_ORD_REBUILD);
7900 }
f21e18ca 7901 if (map->failed_disk_num == 0xff)
0556e1a2 7902 map->failed_disk_num = slot;
4c9e8c1e
TM		 7903
7904 clear_disk_badblocks(super->bbm_log, ord_to_idx(ord));
7905
0556e1a2
DW		 7906 return 1;
7907}
7908
4c9e8c1e
TM		 7909static void mark_missing(struct intel_super *super,
7910 struct imsm_dev *dev, struct imsm_disk *disk, int idx)
0556e1a2 7911{
4c9e8c1e 7912 mark_failure(super, dev, disk, idx);
0556e1a2
DW		 7913
7914 if (disk->scsi_id == __cpu_to_le32(~(__u32)0))
7915 return;
7916
47ee5a45
DW		 7917 disk->scsi_id = __cpu_to_le32(~(__u32)0);
7918 memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1);
7919}
7920
33414a01
DW		 7921static void handle_missing(struct intel_super *super, struct imsm_dev *dev)
7922{
33414a01 7923 struct dl *dl;
33414a01
DW		 7924
7925 if (!super->missing)
7926 return;
33414a01 7927
79b68f1b
PC		 7928 /* When orom adds replacement for missing disk it does
7929 * not remove entry of missing disk, but just updates map with
7930 * new added disk. So it is not enough just to test if there is
7931 * any missing disk, we have to look if there are any failed disks
7932 * in map to stop migration */
7933
33414a01 7934 dprintf("imsm: mark missing\n");
3d59f0c0
AK		 7935 /* end process for initialization and rebuild only
7936 */
7937 if (is_gen_migration(dev) == 0) {
7938 __u8 map_state;
7939 int failed;
7940
7941 failed = imsm_count_failed(super, dev, MAP_0);
7942 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
7943
79b68f1b
PC		 7944 if (failed)
7945 end_migration(dev, super, map_state);
3d59f0c0 7946 }
33414a01 7947 for (dl = super->missing; dl; dl = dl->next)
4c9e8c1e 7948 mark_missing(super, dev, &dl->disk, dl->index);
33414a01
DW		 7949 super->updates_pending++;
7950}
7951
f3871fdc
AK		 7952static unsigned long long imsm_set_array_size(struct imsm_dev *dev,
7953 long long new_size)
70bdf0dc 7954{
238c0a71 7955 int used_disks = imsm_num_data_members(dev, MAP_0);
70bdf0dc
AK		 7956 unsigned long long array_blocks;
7957 struct imsm_map *map;
7958
7959 if (used_disks == 0) {
7960 /* when problems occures
7961 * return current array_blocks value
7962 */
7963 array_blocks = __le32_to_cpu(dev->size_high);
7964 array_blocks = array_blocks << 32;
7965 array_blocks += __le32_to_cpu(dev->size_low);
7966
7967 return array_blocks;
7968 }
7969
7970 /* set array size in metadata
7971 */
f3871fdc
AK		 7972 if (new_size <= 0) {
7973 /* OLCE size change is caused by added disks
7974 */
7975 map = get_imsm_map(dev, MAP_0);
7976 array_blocks = blocks_per_member(map) * used_disks;
7977 } else {
7978 /* Online Volume Size Change
7979 * Using available free space
7980 */
7981 array_blocks = new_size;
7982 }
70bdf0dc
AK		 7983
7984 /* round array size down to closest MB
7985 */
7986 array_blocks = (array_blocks >> SECT_PER_MB_SHIFT) << SECT_PER_MB_SHIFT;
7987 dev->size_low = __cpu_to_le32((__u32)array_blocks);
7988 dev->size_high = __cpu_to_le32((__u32)(array_blocks >> 32));
7989
7990 return array_blocks;
7991}
7992
28bce06f
AK		 7993static void imsm_set_disk(struct active_array *a, int n, int state);
7994
0e2d1a4e
AK		 7995static void imsm_progress_container_reshape(struct intel_super *super)
7996{
7997 /* if no device has a migr_state, but some device has a
7998 * different number of members than the previous device, start
7999 * changing the number of devices in this device to match
8000 * previous.
8001 */
8002 struct imsm_super *mpb = super->anchor;
8003 int prev_disks = -1;
8004 int i;
1dfaa380 8005 int copy_map_size;
0e2d1a4e
AK		 8006
8007 for (i = 0; i < mpb->num_raid_devs; i++) {
8008 struct imsm_dev *dev = get_imsm_dev(super, i);
238c0a71 8009 struct imsm_map *map = get_imsm_map(dev, MAP_0);
0e2d1a4e
AK		 8010 struct imsm_map *map2;
8011 int prev_num_members;
0e2d1a4e
AK		 8012
8013 if (dev->vol.migr_state)
8014 return;
8015
8016 if (prev_disks == -1)
8017 prev_disks = map->num_members;
8018 if (prev_disks == map->num_members)
8019 continue;
8020
8021 /* OK, this array needs to enter reshape mode.
8022 * i.e it needs a migr_state
8023 */
8024
1dfaa380 8025 copy_map_size = sizeof_imsm_map(map);
0e2d1a4e
AK		 8026 prev_num_members = map->num_members;
8027 map->num_members = prev_disks;
8028 dev->vol.migr_state = 1;
8029 dev->vol.curr_migr_unit = 0;
ea672ee1 8030 set_migr_type(dev, MIGR_GEN_MIGR);
0e2d1a4e
AK		 8031 for (i = prev_num_members;
8032 i < map->num_members; i++)
8033 set_imsm_ord_tbl_ent(map, i, i);
238c0a71 8034 map2 = get_imsm_map(dev, MAP_1);
0e2d1a4e 8035 /* Copy the current map */
1dfaa380 8036 memcpy(map2, map, copy_map_size);
0e2d1a4e
AK		 8037 map2->num_members = prev_num_members;
8038
f3871fdc 8039 imsm_set_array_size(dev, -1);
51d83f5d 8040 super->clean_migration_record_by_mdmon = 1;
0e2d1a4e
AK		 8041 super->updates_pending++;
8042 }
8043}
8044
aad6f216 8045/* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild
0c046afd
DW		 8046 * states are handled in imsm_set_disk() with one exception, when a
8047 * resync is stopped due to a new failure this routine will set the
8048 * 'degraded' state for the array.
8049 */
01f157d7 8050static int imsm_set_array_state(struct active_array *a, int consistent)
a862209d
DW		 8051{
8052 int inst = a->info.container_member;
8053 struct intel_super *super = a->container->sb;
949c47a0 8054 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8055 struct imsm_map *map = get_imsm_map(dev, MAP_0);
3b451610
AK		 8056 int failed = imsm_count_failed(super, dev, MAP_0);
8057 __u8 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
1e5c6983 8058 __u32 blocks_per_unit;
a862209d 8059
1af97990
AK		 8060 if (dev->vol.migr_state &&
8061 dev->vol.migr_type == MIGR_GEN_MIGR) {
8062 /* array state change is blocked due to reshape action
aad6f216
N		 8063 * We might need to
8064 * - abort the reshape (if last_checkpoint is 0 and action!= reshape)
8065 * - finish the reshape (if last_checkpoint is big and action != reshape)
8066 * - update curr_migr_unit
1af97990 8067 */
aad6f216
N		 8068 if (a->curr_action == reshape) {
8069 /* still reshaping, maybe update curr_migr_unit */
633b5610 8070 goto mark_checkpoint;
aad6f216
N		 8071 } else {
8072 if (a->last_checkpoint == 0 && a->prev_action == reshape) {
8073 /* for some reason we aborted the reshape.
b66e591b
AK		 8074 *
8075 * disable automatic metadata rollback
8076 * user action is required to recover process
aad6f216 8077 */
b66e591b 8078 if (0) {
238c0a71
AK		 8079 struct imsm_map *map2 =
8080 get_imsm_map(dev, MAP_1);
8081 dev->vol.migr_state = 0;
8082 set_migr_type(dev, 0);
8083 dev->vol.curr_migr_unit = 0;
8084 memcpy(map, map2,
8085 sizeof_imsm_map(map2));
8086 super->updates_pending++;
b66e591b 8087 }
aad6f216
N		 8088 }
8089 if (a->last_checkpoint >= a->info.component_size) {
8090 unsigned long long array_blocks;
8091 int used_disks;
e154ced3 8092 struct mdinfo *mdi;
aad6f216 8093
238c0a71 8094 used_disks = imsm_num_data_members(dev, MAP_0);
d55adef9
AK		 8095 if (used_disks > 0) {
8096 array_blocks =
5551b113 8097 blocks_per_member(map) *
d55adef9
AK		 8098 used_disks;
8099 /* round array size down to closest MB
8100 */
8101 array_blocks = (array_blocks
8102 >> SECT_PER_MB_SHIFT)
8103 << SECT_PER_MB_SHIFT;
d55adef9
AK		 8104 a->info.custom_array_size = array_blocks;
8105 /* encourage manager to update array
8106 * size
8107 */
e154ced3 8108
d55adef9 8109 a->check_reshape = 1;
633b5610 8110 }
e154ced3
AK		 8111 /* finalize online capacity expansion/reshape */
8112 for (mdi = a->info.devs; mdi; mdi = mdi->next)
8113 imsm_set_disk(a,
8114 mdi->disk.raid_disk,
8115 mdi->curr_state);
8116
0e2d1a4e 8117 imsm_progress_container_reshape(super);
e154ced3 8118 }
aad6f216 8119 }
1af97990
AK		 8120 }
8121
47ee5a45 8122 /* before we activate this array handle any missing disks */
33414a01
DW		 8123 if (consistent == 2)
8124 handle_missing(super, dev);
1e5c6983 8125
0c046afd 8126 if (consistent == 2 &&
b7941fd6 8127 (!is_resync_complete(&a->info) ||
0c046afd
DW		 8128 map_state != IMSM_T_STATE_NORMAL ||
8129 dev->vol.migr_state))
01f157d7 8130 consistent = 0;
272906ef 8131
b7941fd6 8132 if (is_resync_complete(&a->info)) {
0c046afd 8133 /* complete intialization / resync,
0556e1a2
DW		 8134 * recovery and interrupted recovery is completed in
8135 * ->set_disk
0c046afd
DW		 8136 */
8137 if (is_resyncing(dev)) {
8138 dprintf("imsm: mark resync done\n");
809da78e 8139 end_migration(dev, super, map_state);
115c3803 8140 super->updates_pending++;
484240d8 8141 a->last_checkpoint = 0;
115c3803 8142 }
b9172665
AK		 8143 } else if ((!is_resyncing(dev) && !failed) &&
8144 (imsm_reshape_blocks_arrays_changes(super) == 0)) {
0c046afd 8145 /* mark the start of the init process if nothing is failed */
b7941fd6 8146 dprintf("imsm: mark resync start\n");
1484e727 8147 if (map->map_state == IMSM_T_STATE_UNINITIALIZED)
8e59f3d8 8148 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_INIT);
1484e727 8149 else
8e59f3d8 8150 migrate(dev, super, IMSM_T_STATE_NORMAL, MIGR_REPAIR);
3393c6af 8151 super->updates_pending++;
115c3803 8152 }
a862209d 8153
633b5610 8154mark_checkpoint:
5b83bacf
AK		 8155 /* skip checkpointing for general migration,
8156 * it is controlled in mdadm
8157 */
8158 if (is_gen_migration(dev))
8159 goto skip_mark_checkpoint;
8160
1e5c6983 8161 /* check if we can update curr_migr_unit from resync_start, recovery_start */
c47b0ff6 8162 blocks_per_unit = blocks_per_migr_unit(super, dev);
4f0a7acc 8163 if (blocks_per_unit) {
1e5c6983
DW		 8164 __u32 units32;
8165 __u64 units;
8166
4f0a7acc 8167 units = a->last_checkpoint / blocks_per_unit;
1e5c6983
DW		 8168 units32 = units;
8169
8170 /* check that we did not overflow 32-bits, and that
8171 * curr_migr_unit needs updating
8172 */
8173 if (units32 == units &&
bfd80a56 8174 units32 != 0 &&
1e5c6983
DW		 8175 __le32_to_cpu(dev->vol.curr_migr_unit) != units32) {
8176 dprintf("imsm: mark checkpoint (%u)\n", units32);
8177 dev->vol.curr_migr_unit = __cpu_to_le32(units32);
8178 super->updates_pending++;
8179 }
8180 }
f8f603f1 8181
5b83bacf 8182skip_mark_checkpoint:
3393c6af 8183 /* mark dirty / clean */
2432ce9b
AP		 8184 if (((dev->vol.dirty & RAIDVOL_DIRTY) && consistent) ||
8185 (!(dev->vol.dirty & RAIDVOL_DIRTY) && !consistent)) {
b7941fd6 8186 dprintf("imsm: mark '%s'\n", consistent ? "clean" : "dirty");
2432ce9b
AP		 8187 if (consistent) {
8188 dev->vol.dirty = RAIDVOL_CLEAN;
8189 } else {
8190 dev->vol.dirty = RAIDVOL_DIRTY;
8191 if (dev->rwh_policy == RWH_DISTRIBUTED)
8192 dev->vol.dirty |= RAIDVOL_DSRECORD_VALID;
8193 }
a862209d
DW		 8194 super->updates_pending++;
8195 }
28bce06f 8196
01f157d7 8197 return consistent;
a862209d
DW		 8198}
8199
6f50473f
TM		 8200static int imsm_disk_slot_to_ord(struct active_array *a, int slot)
8201{
8202 int inst = a->info.container_member;
8203 struct intel_super *super = a->container->sb;
8204 struct imsm_dev *dev = get_imsm_dev(super, inst);
8205 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8206
8207 if (slot > map->num_members) {
8208 pr_err("imsm: imsm_disk_slot_to_ord %d out of range 0..%d\n",
8209 slot, map->num_members - 1);
8210 return -1;
8211 }
8212
8213 if (slot < 0)
8214 return -1;
8215
8216 return get_imsm_ord_tbl_ent(dev, slot, MAP_0);
8217}
8218
8d45d196 8219static void imsm_set_disk(struct active_array *a, int n, int state)
845dea95 8220{
8d45d196
DW		 8221 int inst = a->info.container_member;
8222 struct intel_super *super = a->container->sb;
949c47a0 8223 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8224 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8d45d196 8225 struct imsm_disk *disk;
7ce05701
LD		 8226 struct mdinfo *mdi;
8227 int recovery_not_finished = 0;
0c046afd 8228 int failed;
6f50473f 8229 int ord;
0c046afd 8230 __u8 map_state;
8d45d196 8231
6f50473f
TM		 8232 ord = imsm_disk_slot_to_ord(a, n);
8233 if (ord < 0)
8d45d196
DW		 8234 return;
8235
4e6e574a 8236 dprintf("imsm: set_disk %d:%x\n", n, state);
b10b37b8 8237 disk = get_imsm_disk(super, ord_to_idx(ord));
8d45d196 8238
5802a811 8239 /* check for new failures */
0556e1a2 8240 if (state & DS_FAULTY) {
4c9e8c1e 8241 if (mark_failure(super, dev, disk, ord_to_idx(ord)))
0556e1a2 8242 super->updates_pending++;
8d45d196 8243 }
47ee5a45 8244
19859edc 8245 /* check if in_sync */
0556e1a2 8246 if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD && is_rebuilding(dev)) {
238c0a71 8247 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
b10b37b8
DW		 8248
8249 set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord));
19859edc
DW		 8250 super->updates_pending++;
8251 }
8d45d196 8252
3b451610
AK		 8253 failed = imsm_count_failed(super, dev, MAP_0);
8254 map_state = imsm_check_degraded(super, dev, failed, MAP_0);
5802a811 8255
0c046afd 8256 /* check if recovery complete, newly degraded, or failed */
94002678
AK		 8257 dprintf("imsm: Detected transition to state ");
8258 switch (map_state) {
8259 case IMSM_T_STATE_NORMAL: /* transition to normal state */
8260 dprintf("normal: ");
8261 if (is_rebuilding(dev)) {
1ade5cc1 8262 dprintf_cont("while rebuilding");
7ce05701
LD		 8263 /* check if recovery is really finished */
8264 for (mdi = a->info.devs; mdi ; mdi = mdi->next)
8265 if (mdi->recovery_start != MaxSector) {
8266 recovery_not_finished = 1;
8267 break;
8268 }
8269 if (recovery_not_finished) {
1ade5cc1
N		 8270 dprintf_cont("\n");
8271 dprintf("Rebuild has not finished yet, state not changed");
7ce05701
LD		 8272 if (a->last_checkpoint < mdi->recovery_start) {
8273 a->last_checkpoint = mdi->recovery_start;
8274 super->updates_pending++;
8275 }
8276 break;
8277 }
94002678 8278 end_migration(dev, super, map_state);
238c0a71 8279 map = get_imsm_map(dev, MAP_0);
94002678
AK		 8280 map->failed_disk_num = ~0;
8281 super->updates_pending++;
8282 a->last_checkpoint = 0;
8283 break;
8284 }
8285 if (is_gen_migration(dev)) {
1ade5cc1 8286 dprintf_cont("while general migration");
bf2f0071 8287 if (a->last_checkpoint >= a->info.component_size)
809da78e 8288 end_migration(dev, super, map_state);
94002678
AK		 8289 else
8290 map->map_state = map_state;
238c0a71 8291 map = get_imsm_map(dev, MAP_0);
28bce06f 8292 map->failed_disk_num = ~0;
94002678 8293 super->updates_pending++;
bf2f0071 8294 break;
94002678
AK		 8295 }
8296 break;
8297 case IMSM_T_STATE_DEGRADED: /* transition to degraded state */
1ade5cc1 8298 dprintf_cont("degraded: ");
089f9d79 8299 if (map->map_state != map_state && !dev->vol.migr_state) {
1ade5cc1 8300 dprintf_cont("mark degraded");
94002678
AK		 8301 map->map_state = map_state;
8302 super->updates_pending++;
8303 a->last_checkpoint = 0;
8304 break;
8305 }
8306 if (is_rebuilding(dev)) {
1ade5cc1 8307 dprintf_cont("while rebuilding.");
94002678 8308 if (map->map_state != map_state) {
1ade5cc1 8309 dprintf_cont(" Map state change");
94002678
AK		 8310 end_migration(dev, super, map_state);
8311 super->updates_pending++;
8312 }
8313 break;
8314 }
8315 if (is_gen_migration(dev)) {
1ade5cc1 8316 dprintf_cont("while general migration");
bf2f0071 8317 if (a->last_checkpoint >= a->info.component_size)
809da78e 8318 end_migration(dev, super, map_state);
94002678
AK		 8319 else {
8320 map->map_state = map_state;
3b451610 8321 manage_second_map(super, dev);
94002678
AK		 8322 }
8323 super->updates_pending++;
bf2f0071 8324 break;
28bce06f 8325 }
6ce1fbf1 8326 if (is_initializing(dev)) {
1ade5cc1 8327 dprintf_cont("while initialization.");
6ce1fbf1
AK		 8328 map->map_state = map_state;
8329 super->updates_pending++;
8330 break;
8331 }
94002678
AK		 8332 break;
8333 case IMSM_T_STATE_FAILED: /* transition to failed state */
1ade5cc1 8334 dprintf_cont("failed: ");
94002678 8335 if (is_gen_migration(dev)) {
1ade5cc1 8336 dprintf_cont("while general migration");
94002678
AK		 8337 map->map_state = map_state;
8338 super->updates_pending++;
8339 break;
8340 }
8341 if (map->map_state != map_state) {
1ade5cc1 8342 dprintf_cont("mark failed");
94002678
AK		 8343 end_migration(dev, super, map_state);
8344 super->updates_pending++;
8345 a->last_checkpoint = 0;
8346 break;
8347 }
8348 break;
8349 default:
1ade5cc1 8350 dprintf_cont("state %i\n", map_state);
5802a811 8351 }
1ade5cc1 8352 dprintf_cont("\n");
845dea95
NB		 8353}
8354
f796af5d 8355static int store_imsm_mpb(int fd, struct imsm_super *mpb)
c2a1e7da 8356{
f796af5d 8357 void *buf = mpb;
c2a1e7da
DW		 8358 __u32 mpb_size = __le32_to_cpu(mpb->mpb_size);
8359 unsigned long long dsize;
8360 unsigned long long sectors;
f36a9ecd 8361 unsigned int sector_size;
c2a1e7da 8362
f36a9ecd 8363 get_dev_sector_size(fd, NULL, &sector_size);
c2a1e7da
DW		 8364 get_dev_size(fd, NULL, &dsize);
8365
f36a9ecd 8366 if (mpb_size > sector_size) {
272f648f 8367 /* -1 to account for anchor */
f36a9ecd 8368 sectors = mpb_sectors(mpb, sector_size) - 1;
c2a1e7da 8369
272f648f 8370 /* write the extended mpb to the sectors preceeding the anchor */
f36a9ecd
PB		 8371 if (lseek64(fd, dsize - (sector_size * (2 + sectors)),
8372 SEEK_SET) < 0)
272f648f 8373 return 1;
c2a1e7da 8374
f36a9ecd
PB		 8375 if ((unsigned long long)write(fd, buf + sector_size,
8376 sector_size * sectors) != sector_size * sectors)
272f648f
DW		 8377 return 1;
8378 }
c2a1e7da 8379
272f648f 8380 /* first block is stored on second to last sector of the disk */
f36a9ecd 8381 if (lseek64(fd, dsize - (sector_size * 2), SEEK_SET) < 0)
c2a1e7da
DW		 8382 return 1;
8383
466070ad 8384 if ((unsigned int)write(fd, buf, sector_size) != sector_size)
c2a1e7da
DW		 8385 return 1;
8386
c2a1e7da
DW		 8387 return 0;
8388}
8389
2e735d19 8390static void imsm_sync_metadata(struct supertype *container)
845dea95 8391{
2e735d19 8392 struct intel_super *super = container->sb;
c2a1e7da 8393
1a64be56 8394 dprintf("sync metadata: %d\n", super->updates_pending);
c2a1e7da
DW		 8395 if (!super->updates_pending)
8396 return;
8397
36988a3d 8398 write_super_imsm(container, 0);
c2a1e7da
DW		 8399
8400 super->updates_pending = 0;
845dea95
NB		 8401}
8402
272906ef
DW		 8403static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a)
8404{
8405 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8406 int i = get_imsm_disk_idx(dev, idx, MAP_X);
272906ef
DW		 8407 struct dl *dl;
8408
8409 for (dl = super->disks; dl; dl = dl->next)
8410 if (dl->index == i)
8411 break;
8412
25ed7e59 8413 if (dl && is_failed(&dl->disk))
272906ef
DW		 8414 dl = NULL;
8415
8416 if (dl)
1ade5cc1 8417 dprintf("found %x:%x\n", dl->major, dl->minor);
272906ef
DW		 8418
8419 return dl;
8420}
8421
a20d2ba5 8422static struct dl *imsm_add_spare(struct intel_super *super, int slot,
8ba77d32
AK		 8423 struct active_array *a, int activate_new,
8424 struct mdinfo *additional_test_list)
272906ef
DW		 8425{
8426 struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
238c0a71 8427 int idx = get_imsm_disk_idx(dev, slot, MAP_X);
a20d2ba5
DW		 8428 struct imsm_super *mpb = super->anchor;
8429 struct imsm_map *map;
272906ef
DW		 8430 unsigned long long pos;
8431 struct mdinfo *d;
8432 struct extent *ex;
a20d2ba5 8433 int i, j;
272906ef 8434 int found;
569cc43f
DW		 8435 __u32 array_start = 0;
8436 __u32 array_end = 0;
272906ef 8437 struct dl *dl;
6c932028 8438 struct mdinfo *test_list;
272906ef
DW		 8439
8440 for (dl = super->disks; dl; dl = dl->next) {
8441 /* If in this array, skip */
8442 for (d = a->info.devs ; d ; d = d->next)
e553d2a4
DW		 8443 if (d->state_fd >= 0 &&
8444 d->disk.major == dl->major &&
272906ef 8445 d->disk.minor == dl->minor) {
8ba77d32
AK		 8446 dprintf("%x:%x already in array\n",
8447 dl->major, dl->minor);
272906ef
DW		 8448 break;
8449 }
8450 if (d)
8451 continue;
6c932028
AK		 8452 test_list = additional_test_list;
8453 while (test_list) {
8454 if (test_list->disk.major == dl->major &&
8455 test_list->disk.minor == dl->minor) {
8ba77d32
AK		 8456 dprintf("%x:%x already in additional test list\n",
8457 dl->major, dl->minor);
8458 break;
8459 }
6c932028 8460 test_list = test_list->next;
8ba77d32 8461 }
6c932028 8462 if (test_list)
8ba77d32 8463 continue;
272906ef 8464
e553d2a4 8465 /* skip in use or failed drives */
25ed7e59 8466 if (is_failed(&dl->disk) || idx == dl->index ||
df474657
DW		 8467 dl->index == -2) {
8468 dprintf("%x:%x status (failed: %d index: %d)\n",
25ed7e59 8469 dl->major, dl->minor, is_failed(&dl->disk), idx);
9a1608e5
DW		 8470 continue;
8471 }
8472
a20d2ba5
DW		 8473 /* skip pure spares when we are looking for partially
8474 * assimilated drives
8475 */
8476 if (dl->index == -1 && !activate_new)
8477 continue;
8478
272906ef 8479 /* Does this unused device have the requisite free space?
a20d2ba5 8480 * It needs to be able to cover all member volumes
272906ef
DW		 8481 */
8482 ex = get_extents(super, dl);
8483 if (!ex) {
8484 dprintf("cannot get extents\n");
8485 continue;
8486 }
a20d2ba5
DW		 8487 for (i = 0; i < mpb->num_raid_devs; i++) {
8488 dev = get_imsm_dev(super, i);
238c0a71 8489 map = get_imsm_map(dev, MAP_0);
272906ef 8490
a20d2ba5
DW		 8491 /* check if this disk is already a member of
8492 * this array
272906ef 8493 */
620b1713 8494 if (get_imsm_disk_slot(map, dl->index) >= 0)
a20d2ba5
DW		 8495 continue;
8496
8497 found = 0;
8498 j = 0;
8499 pos = 0;
5551b113 8500 array_start = pba_of_lba0(map);
329c8278 8501 array_end = array_start +
5551b113 8502 blocks_per_member(map) - 1;
a20d2ba5
DW		 8503
8504 do {
8505 /* check that we can start at pba_of_lba0 with
8506 * blocks_per_member of space
8507 */
329c8278 8508 if (array_start >= pos && array_end < ex[j].start) {
a20d2ba5
DW		 8509 found = 1;
8510 break;
8511 }
8512 pos = ex[j].start + ex[j].size;
8513 j++;
8514 } while (ex[j-1].size);
8515
8516 if (!found)
272906ef 8517 break;
a20d2ba5 8518 }
272906ef
DW		 8519
8520 free(ex);
a20d2ba5 8521 if (i < mpb->num_raid_devs) {
329c8278
DW		 8522 dprintf("%x:%x does not have %u to %u available\n",
8523 dl->major, dl->minor, array_start, array_end);
272906ef
DW		 8524 /* No room */
8525 continue;
a20d2ba5
DW		 8526 }
8527 return dl;
272906ef
DW		 8528 }
8529
8530 return dl;
8531}
8532
95d07a2c
LM		 8533static int imsm_rebuild_allowed(struct supertype *cont, int dev_idx, int failed)
8534{
8535 struct imsm_dev *dev2;
8536 struct imsm_map *map;
8537 struct dl *idisk;
8538 int slot;
8539 int idx;
8540 __u8 state;
8541
8542 dev2 = get_imsm_dev(cont->sb, dev_idx);
8543 if (dev2) {
238c0a71 8544 state = imsm_check_degraded(cont->sb, dev2, failed, MAP_0);
95d07a2c 8545 if (state == IMSM_T_STATE_FAILED) {
238c0a71 8546 map = get_imsm_map(dev2, MAP_0);
95d07a2c
LM		 8547 if (!map)
8548 return 1;
8549 for (slot = 0; slot < map->num_members; slot++) {
8550 /*
8551 * Check if failed disks are deleted from intel
8552 * disk list or are marked to be deleted
8553 */
238c0a71 8554 idx = get_imsm_disk_idx(dev2, slot, MAP_X);
95d07a2c
LM		 8555 idisk = get_imsm_dl_disk(cont->sb, idx);
8556 /*
8557 * Do not rebuild the array if failed disks
8558 * from failed sub-array are not removed from
8559 * container.
8560 */
8561 if (idisk &&
8562 is_failed(&idisk->disk) &&
8563 (idisk->action != DISK_REMOVE))
8564 return 0;
8565 }
8566 }
8567 }
8568 return 1;
8569}
8570
88758e9d
DW		 8571static struct mdinfo *imsm_activate_spare(struct active_array *a,
8572 struct metadata_update **updates)
8573{
8574 /**
d23fe947
DW		 8575 * Find a device with unused free space and use it to replace a
8576 * failed/vacant region in an array. We replace failed regions one a
8577 * array at a time. The result is that a new spare disk will be added
8578 * to the first failed array and after the monitor has finished
8579 * propagating failures the remainder will be consumed.
88758e9d 8580 *
d23fe947
DW		 8581 * FIXME add a capability for mdmon to request spares from another
8582 * container.
88758e9d
DW		 8583 */
8584
8585 struct intel_super *super = a->container->sb;
88758e9d 8586 int inst = a->info.container_member;
949c47a0 8587 struct imsm_dev *dev = get_imsm_dev(super, inst);
238c0a71 8588 struct imsm_map *map = get_imsm_map(dev, MAP_0);
88758e9d
DW		 8589 int failed = a->info.array.raid_disks;
8590 struct mdinfo *rv = NULL;
8591 struct mdinfo *d;
8592 struct mdinfo *di;
8593 struct metadata_update *mu;
8594 struct dl *dl;
8595 struct imsm_update_activate_spare *u;
8596 int num_spares = 0;
8597 int i;
95d07a2c 8598 int allowed;
88758e9d
DW		 8599
8600 for (d = a->info.devs ; d ; d = d->next) {
8601 if ((d->curr_state & DS_FAULTY) &&
8602 d->state_fd >= 0)
8603 /* wait for Removal to happen */
8604 return NULL;
8605 if (d->state_fd >= 0)
8606 failed--;
8607 }
8608
8609 dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n",
8610 inst, failed, a->info.array.raid_disks, a->info.array.level);
1af97990 8611
e2962bfc
AK		 8612 if (imsm_reshape_blocks_arrays_changes(super))
8613 return NULL;
1af97990 8614
fc8ca064
AK		 8615 /* Cannot activate another spare if rebuild is in progress already
8616 */
8617 if (is_rebuilding(dev)) {
7a862a02 8618 dprintf("imsm: No spare activation allowed. Rebuild in progress already.\n");
fc8ca064
AK		 8619 return NULL;
8620 }
8621
89c67882
AK		 8622 if (a->info.array.level == 4)
8623 /* No repair for takeovered array
8624 * imsm doesn't support raid4
8625 */
8626 return NULL;
8627
3b451610
AK		 8628 if (imsm_check_degraded(super, dev, failed, MAP_0) !=
8629 IMSM_T_STATE_DEGRADED)
88758e9d
DW		 8630 return NULL;
8631
83ca7d45
AP		 8632 if (get_imsm_map(dev, MAP_0)->map_state == IMSM_T_STATE_UNINITIALIZED) {
8633 dprintf("imsm: No spare activation allowed. Volume is not initialized.\n");
8634 return NULL;
8635 }
8636
95d07a2c
LM		 8637 /*
8638 * If there are any failed disks check state of the other volume.
8639 * Block rebuild if the another one is failed until failed disks
8640 * are removed from container.
8641 */
8642 if (failed) {
7a862a02 8643 dprintf("found failed disks in %.*s, check if there anotherfailed sub-array.\n",
c4acd1e5 8644 MAX_RAID_SERIAL_LEN, dev->volume);
95d07a2c
LM		 8645 /* check if states of the other volumes allow for rebuild */
8646 for (i = 0; i < super->anchor->num_raid_devs; i++) {
8647 if (i != inst) {
8648 allowed = imsm_rebuild_allowed(a->container,
8649 i, failed);
8650 if (!allowed)
8651 return NULL;
8652 }
8653 }
8654 }
8655
88758e9d 8656 /* For each slot, if it is not working, find a spare */
88758e9d
DW		 8657 for (i = 0; i < a->info.array.raid_disks; i++) {
8658 for (d = a->info.devs ; d ; d = d->next)
8659 if (d->disk.raid_disk == i)
8660 break;
8661 dprintf("found %d: %p %x\n", i, d, d?d->curr_state:0);
8662 if (d && (d->state_fd >= 0))
8663 continue;
8664
272906ef 8665 /*
a20d2ba5
DW		 8666 * OK, this device needs recovery. Try to re-add the
8667 * previous occupant of this slot, if this fails see if
8668 * we can continue the assimilation of a spare that was
8669 * partially assimilated, finally try to activate a new
8670 * spare.
272906ef
DW		 8671 */
8672 dl = imsm_readd(super, i, a);
8673 if (!dl)
b303fe21 8674 dl = imsm_add_spare(super, i, a, 0, rv);
a20d2ba5 8675 if (!dl)
b303fe21 8676 dl = imsm_add_spare(super, i, a, 1, rv);
272906ef
DW		 8677 if (!dl)
8678 continue;
1011e834 8679
272906ef 8680 /* found a usable disk with enough space */
503975b9 8681 di = xcalloc(1, sizeof(*di));
272906ef
DW		 8682
8683 /* dl->index will be -1 in the case we are activating a
8684 * pristine spare. imsm_process_update() will create a
8685 * new index in this case. Once a disk is found to be
8686 * failed in all member arrays it is kicked from the
8687 * metadata
8688 */
8689 di->disk.number = dl->index;
d23fe947 8690
272906ef
DW		 8691 /* (ab)use di->devs to store a pointer to the device
8692 * we chose
8693 */
8694 di->devs = (struct mdinfo *) dl;
8695
8696 di->disk.raid_disk = i;
8697 di->disk.major = dl->major;
8698 di->disk.minor = dl->minor;
8699 di->disk.state = 0;
d23534e4 8700 di->recovery_start = 0;
5551b113 8701 di->data_offset = pba_of_lba0(map);
272906ef
DW		 8702 di->component_size = a->info.component_size;
8703 di->container_member = inst;
5e46202e 8704 di->bb.supported = 1;
2432ce9b
AP		 8705 if (dev->rwh_policy == RWH_DISTRIBUTED) {
8706 di->consistency_policy = CONSISTENCY_POLICY_PPL;
8707 di->ppl_sector = get_ppl_sector(super, inst);
8708 di->ppl_size = (PPL_HEADER_SIZE + PPL_ENTRY_SPACE) >> 9;
8709 }
148acb7b 8710 super->random = random32();
272906ef
DW		 8711 di->next = rv;
8712 rv = di;
8713 num_spares++;
8714 dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor,
8715 i, di->data_offset);
88758e9d
DW		 8716 }
8717
8718 if (!rv)
8719 /* No spares found */
8720 return rv;
8721 /* Now 'rv' has a list of devices to return.
8722 * Create a metadata_update record to update the
8723 * disk_ord_tbl for the array
8724 */
503975b9 8725 mu = xmalloc(sizeof(*mu));
1011e834 8726 mu->buf = xcalloc(num_spares,
503975b9 8727 sizeof(struct imsm_update_activate_spare));
88758e9d 8728 mu->space = NULL;
cb23f1f4 8729 mu->space_list = NULL;
88758e9d
DW		 8730 mu->len = sizeof(struct imsm_update_activate_spare) * num_spares;
8731 mu->next = *updates;
8732 u = (struct imsm_update_activate_spare *) mu->buf;
8733
8734 for (di = rv ; di ; di = di->next) {
8735 u->type = update_activate_spare;
d23fe947
DW		 8736 u->dl = (struct dl *) di->devs;
8737 di->devs = NULL;
88758e9d
DW		 8738 u->slot = di->disk.raid_disk;
8739 u->array = inst;
8740 u->next = u + 1;
8741 u++;
8742 }
8743 (u-1)->next = NULL;
8744 *updates = mu;
8745
8746 return rv;
8747}
8748
54c2c1ea 8749static int disks_overlap(struct intel_super *super, int idx, struct imsm_update_create_array *u)
8273f55e 8750{
54c2c1ea 8751 struct imsm_dev *dev = get_imsm_dev(super, idx);
238c0a71
AK		 8752 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8753 struct imsm_map *new_map = get_imsm_map(&u->dev, MAP_0);
54c2c1ea
DW		 8754 struct disk_info *inf = get_disk_info(u);
8755 struct imsm_disk *disk;
8273f55e
DW		 8756 int i;
8757 int j;
8273f55e 8758
54c2c1ea 8759 for (i = 0; i < map->num_members; i++) {
238c0a71 8760 disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i, MAP_X));
54c2c1ea
DW		 8761 for (j = 0; j < new_map->num_members; j++)
8762 if (serialcmp(disk->serial, inf[j].serial) == 0)
8273f55e
DW		 8763 return 1;
8764 }
8765
8766 return 0;
8767}
8768
1a64be56
LM		 8769static struct dl *get_disk_super(struct intel_super *super, int major, int minor)
8770{
594dc1b8
JS		 8771 struct dl *dl;
8772
1a64be56 8773 for (dl = super->disks; dl; dl = dl->next)
089f9d79 8774 if (dl->major == major && dl->minor == minor)
1a64be56
LM		 8775 return dl;
8776 return NULL;
8777}
8778
8779static int remove_disk_super(struct intel_super *super, int major, int minor)
8780{
594dc1b8 8781 struct dl *prev;
1a64be56
LM		 8782 struct dl *dl;
8783
8784 prev = NULL;
8785 for (dl = super->disks; dl; dl = dl->next) {
089f9d79 8786 if (dl->major == major && dl->minor == minor) {
1a64be56
LM		 8787 /* remove */
8788 if (prev)
8789 prev->next = dl->next;
8790 else
8791 super->disks = dl->next;
8792 dl->next = NULL;
8793 __free_imsm_disk(dl);
1ade5cc1 8794 dprintf("removed %x:%x\n", major, minor);
1a64be56
LM		 8795 break;
8796 }
8797 prev = dl;
8798 }
8799 return 0;
8800}
8801
f21e18ca 8802static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index);
ae6aad82 8803
1a64be56
LM		 8804static int add_remove_disk_update(struct intel_super *super)
8805{
8806 int check_degraded = 0;
594dc1b8
JS		 8807 struct dl *disk;
8808
1a64be56
LM		 8809 /* add/remove some spares to/from the metadata/contrainer */
8810 while (super->disk_mgmt_list) {
8811 struct dl *disk_cfg;
8812
8813 disk_cfg = super->disk_mgmt_list;
8814 super->disk_mgmt_list = disk_cfg->next;
8815 disk_cfg->next = NULL;
8816
8817 if (disk_cfg->action == DISK_ADD) {
8818 disk_cfg->next = super->disks;
8819 super->disks = disk_cfg;
8820 check_degraded = 1;
1ade5cc1
N		 8821 dprintf("added %x:%x\n",
8822 disk_cfg->major, disk_cfg->minor);
1a64be56
LM		 8823 } else if (disk_cfg->action == DISK_REMOVE) {
8824 dprintf("Disk remove action processed: %x.%x\n",
8825 disk_cfg->major, disk_cfg->minor);
8826 disk = get_disk_super(super,
8827 disk_cfg->major,
8828 disk_cfg->minor);
8829 if (disk) {
8830 /* store action status */
8831 disk->action = DISK_REMOVE;
8832 /* remove spare disks only */
8833 if (disk->index == -1) {
8834 remove_disk_super(super,
8835 disk_cfg->major,
8836 disk_cfg->minor);
8837 }
8838 }
8839 /* release allocate disk structure */
8840 __free_imsm_disk(disk_cfg);
8841 }
8842 }
8843 return check_degraded;
8844}
8845
a29911da
PC		 8846static int apply_reshape_migration_update(struct imsm_update_reshape_migration *u,
8847 struct intel_super *super,
8848 void ***space_list)
8849{
8850 struct intel_dev *id;
8851 void **tofree = NULL;
8852 int ret_val = 0;
8853
1ade5cc1 8854 dprintf("(enter)\n");
089f9d79 8855 if (u->subdev < 0 || u->subdev > 1) {
a29911da
PC		 8856 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
8857 return ret_val;
8858 }
089f9d79 8859 if (space_list == NULL || *space_list == NULL) {
a29911da
PC		 8860 dprintf("imsm: Error: Memory is not allocated\n");
8861 return ret_val;
8862 }
8863
8864 for (id = super->devlist ; id; id = id->next) {
8865 if (id->index == (unsigned)u->subdev) {
8866 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
8867 struct imsm_map *map;
8868 struct imsm_dev *new_dev =
8869 (struct imsm_dev *)*space_list;
238c0a71 8870 struct imsm_map *migr_map = get_imsm_map(dev, MAP_1);
a29911da
PC		 8871 int to_state;
8872 struct dl *new_disk;
8873
8874 if (new_dev == NULL)
8875 return ret_val;
8876 *space_list = **space_list;
8877 memcpy(new_dev, dev, sizeof_imsm_dev(dev, 0));
238c0a71 8878 map = get_imsm_map(new_dev, MAP_0);
a29911da
PC		 8879 if (migr_map) {
8880 dprintf("imsm: Error: migration in progress");
8881 return ret_val;
8882 }
8883
8884 to_state = map->map_state;
8885 if ((u->new_level == 5) && (map->raid_level == 0)) {
8886 map->num_members++;
8887 /* this should not happen */
8888 if (u->new_disks[0] < 0) {
8889 map->failed_disk_num =
8890 map->num_members - 1;
8891 to_state = IMSM_T_STATE_DEGRADED;
8892 } else
8893 to_state = IMSM_T_STATE_NORMAL;
8894 }
8e59f3d8 8895 migrate(new_dev, super, to_state, MIGR_GEN_MIGR);
a29911da
PC		 8896 if (u->new_level > -1)
8897 map->raid_level = u->new_level;
238c0a71 8898 migr_map = get_imsm_map(new_dev, MAP_1);
a29911da
PC		 8899 if ((u->new_level == 5) &&
8900 (migr_map->raid_level == 0)) {
8901 int ord = map->num_members - 1;
8902 migr_map->num_members--;
8903 if (u->new_disks[0] < 0)
8904 ord |= IMSM_ORD_REBUILD;
8905 set_imsm_ord_tbl_ent(map,
8906 map->num_members - 1,
8907 ord);
8908 }
8909 id->dev = new_dev;
8910 tofree = (void **)dev;
8911
4bba0439
PC		 8912 /* update chunk size
8913 */
06fb291a
PB		 8914 if (u->new_chunksize > 0) {
8915 unsigned long long num_data_stripes;
8916 int used_disks =
8917 imsm_num_data_members(dev, MAP_0);
8918
8919 if (used_disks == 0)
8920 return ret_val;
8921
4bba0439
PC		 8922 map->blocks_per_strip =
8923 __cpu_to_le16(u->new_chunksize * 2);
06fb291a
PB		 8924 num_data_stripes =
8925 (join_u32(dev->size_low, dev->size_high)
8926 / used_disks);
8927 num_data_stripes /= map->blocks_per_strip;
8928 num_data_stripes /= map->num_domains;
8929 set_num_data_stripes(map, num_data_stripes);
8930 }
4bba0439 8931
a29911da
PC		 8932 /* add disk
8933 */
089f9d79
JS		 8934 if (u->new_level != 5 || migr_map->raid_level != 0 ||
8935 migr_map->raid_level == map->raid_level)
a29911da
PC		 8936 goto skip_disk_add;
8937
8938 if (u->new_disks[0] >= 0) {
8939 /* use passes spare
8940 */
8941 new_disk = get_disk_super(super,
8942 major(u->new_disks[0]),
8943 minor(u->new_disks[0]));
7a862a02 8944 dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n",
a29911da
PC		 8945 major(u->new_disks[0]),
8946 minor(u->new_disks[0]),
8947 new_disk, new_disk->index);
8948 if (new_disk == NULL)
8949 goto error_disk_add;
8950
8951 new_disk->index = map->num_members - 1;
8952 /* slot to fill in autolayout
8953 */
8954 new_disk->raiddisk = new_disk->index;
8955 new_disk->disk.status |= CONFIGURED_DISK;
8956 new_disk->disk.status &= ~SPARE_DISK;
8957 } else
8958 goto error_disk_add;
8959
8960skip_disk_add:
8961 *tofree = *space_list;
8962 /* calculate new size
8963 */
f3871fdc 8964 imsm_set_array_size(new_dev, -1);
a29911da
PC		 8965
8966 ret_val = 1;
8967 }
8968 }
8969
8970 if (tofree)
8971 *space_list = tofree;
8972 return ret_val;
8973
8974error_disk_add:
8975 dprintf("Error: imsm: Cannot find disk.\n");
8976 return ret_val;
8977}
8978
f3871fdc
AK		 8979static int apply_size_change_update(struct imsm_update_size_change *u,
8980 struct intel_super *super)
8981{
8982 struct intel_dev *id;
8983 int ret_val = 0;
8984
1ade5cc1 8985 dprintf("(enter)\n");
089f9d79 8986 if (u->subdev < 0 || u->subdev > 1) {
f3871fdc
AK		 8987 dprintf("imsm: Error: Wrong subdev: %i\n", u->subdev);
8988 return ret_val;
8989 }
8990
8991 for (id = super->devlist ; id; id = id->next) {
8992 if (id->index == (unsigned)u->subdev) {
8993 struct imsm_dev *dev = get_imsm_dev(super, u->subdev);
8994 struct imsm_map *map = get_imsm_map(dev, MAP_0);
8995 int used_disks = imsm_num_data_members(dev, MAP_0);
8996 unsigned long long blocks_per_member;
06fb291a 8997 unsigned long long num_data_stripes;
f3871fdc
AK		 8998
8999 /* calculate new size
9000 */
9001 blocks_per_member = u->new_size / used_disks;
06fb291a
PB		 9002 num_data_stripes = blocks_per_member /
9003 map->blocks_per_strip;
9004 num_data_stripes /= map->num_domains;
9005 dprintf("(size: %llu, blocks per member: %llu, num_data_stipes: %llu)\n",
9006 u->new_size, blocks_per_member,
9007 num_data_stripes);
f3871fdc 9008 set_blocks_per_member(map, blocks_per_member);
06fb291a 9009 set_num_data_stripes(map, num_data_stripes);
f3871fdc
AK		 9010 imsm_set_array_size(dev, u->new_size);
9011
9012 ret_val = 1;
9013 break;
9014 }
9015 }
9016
9017 return ret_val;
9018}
9019
061d7da3 9020static int apply_update_activate_spare(struct imsm_update_activate_spare *u,
ca9de185 9021 struct intel_super *super,
061d7da3
LO		 9022 struct active_array *active_array)
9023{
9024 struct imsm_super *mpb = super->anchor;
9025 struct imsm_dev *dev = get_imsm_dev(super, u->array);
238c0a71 9026 struct imsm_map *map = get_imsm_map(dev, MAP_0);
061d7da3
LO		 9027 struct imsm_map *migr_map;
9028 struct active_array *a;
9029 struct imsm_disk *disk;
9030 __u8 to_state;
9031 struct dl *dl;
9032 unsigned int found;
9033 int failed;
5961eeec 9034 int victim;
061d7da3 9035 int i;
5961eeec 9036 int second_map_created = 0;
061d7da3 9037
5961eeec 9038 for (; u; u = u->next) {
238c0a71 9039 victim = get_imsm_disk_idx(dev, u->slot, MAP_X);
061d7da3 9040
5961eeec 9041 if (victim < 0)
9042 return 0;
061d7da3 9043
5961eeec 9044 for (dl = super->disks; dl; dl = dl->next)
9045 if (dl == u->dl)
9046 break;
061d7da3 9047
5961eeec 9048 if (!dl) {
7a862a02 9049 pr_err("error: imsm_activate_spare passed an unknown disk (index: %d)\n",
5961eeec 9050 u->dl->index);
9051 return 0;
9052 }
061d7da3 9053
5961eeec 9054 /* count failures (excluding rebuilds and the victim)
9055 * to determine map[0] state
9056 */
9057 failed = 0;
9058 for (i = 0; i < map->num_members; i++) {
9059 if (i == u->slot)
9060 continue;
9061 disk = get_imsm_disk(super,
238c0a71 9062 get_imsm_disk_idx(dev, i, MAP_X));
5961eeec 9063 if (!disk || is_failed(disk))
9064 failed++;
9065 }
061d7da3 9066
5961eeec 9067 /* adding a pristine spare, assign a new index */
9068 if (dl->index < 0) {
9069 dl->index = super->anchor->num_disks;
9070 super->anchor->num_disks++;
9071 }
9072 disk = &dl->disk;
9073 disk->status |= CONFIGURED_DISK;
9074 disk->status &= ~SPARE_DISK;
9075
9076 /* mark rebuild */
238c0a71 9077 to_state = imsm_check_degraded(super, dev, failed, MAP_0);
5961eeec 9078 if (!second_map_created) {
9079 second_map_created = 1;
9080 map->map_state = IMSM_T_STATE_DEGRADED;
9081 migrate(dev, super, to_state, MIGR_REBUILD);
9082 } else
9083 map->map_state = to_state;
238c0a71 9084 migr_map = get_imsm_map(dev, MAP_1);
5961eeec 9085 set_imsm_ord_tbl_ent(map, u->slot, dl->index);
9086 set_imsm_ord_tbl_ent(migr_map, u->slot,
9087 dl->index | IMSM_ORD_REBUILD);
9088
9089 /* update the family_num to mark a new container
9090 * generation, being careful to record the existing
9091 * family_num in orig_family_num to clean up after
9092 * earlier mdadm versions that neglected to set it.
9093 */
9094 if (mpb->orig_family_num == 0)
9095 mpb->orig_family_num = mpb->family_num;
9096 mpb->family_num += super->random;
9097
9098 /* count arrays using the victim in the metadata */
9099 found = 0;
9100 for (a = active_array; a ; a = a->next) {
9101 dev = get_imsm_dev(super, a->info.container_member);
238c0a71 9102 map = get_imsm_map(dev, MAP_0);
061d7da3 9103
5961eeec 9104 if (get_imsm_disk_slot(map, victim) >= 0)
9105 found++;
9106 }
061d7da3 9107
5961eeec 9108 /* delete the victim if it is no longer being
9109 * utilized anywhere
061d7da3 9110 */
5961eeec 9111 if (!found) {
9112 struct dl **dlp;
061d7da3 9113
5961eeec 9114 /* We know that 'manager' isn't touching anything,
9115 * so it is safe to delete
9116 */
9117 for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next)
061d7da3
LO		 9118 if ((*dlp)->index == victim)
9119 break;
5961eeec 9120
9121 /* victim may be on the missing list */
9122 if (!*dlp)
9123 for (dlp = &super->missing; *dlp;
9124 dlp = &(*dlp)->next)
9125 if ((*dlp)->index == victim)
9126 break;
9127 imsm_delete(super, dlp, victim);
9128 }
061d7da3
LO		 9129 }
9130
9131 return 1;
9132}
a29911da 9133
2e5dc010
N		 9134static int apply_reshape_container_disks_update(struct imsm_update_reshape *u,
9135 struct intel_super *super,
9136 void ***space_list)
9137{
9138 struct dl *new_disk;
9139 struct intel_dev *id;
9140 int i;
9141 int delta_disks = u->new_raid_disks - u->old_raid_disks;
ee4beede 9142 int disk_count = u->old_raid_disks;
2e5dc010
N		 9143 void **tofree = NULL;
9144 int devices_to_reshape = 1;
9145 struct imsm_super *mpb = super->anchor;
9146 int ret_val = 0;
d098291a 9147 unsigned int dev_id;
2e5dc010 9148
1ade5cc1 9149 dprintf("(enter)\n");
2e5dc010
N		 9150
9151 /* enable spares to use in array */
9152 for (i = 0; i < delta_disks; i++) {
9153 new_disk = get_disk_super(super,
9154 major(u->new_disks[i]),
9155 minor(u->new_disks[i]));
7a862a02 9156 dprintf("imsm: new disk for reshape is: %i:%i (%p, index = %i)\n",
2e5dc010
N		 9157 major(u->new_disks[i]), minor(u->new_disks[i]),
9158 new_disk, new_disk->index);
089f9d79
JS		 9159 if (new_disk == NULL ||
9160 (new_disk->index >= 0 &&
9161 new_disk->index < u->old_raid_disks))
2e5dc010 9162 goto update_reshape_exit;
ee4beede 9163 new_disk->index = disk_count++;
2e5dc010
N		 9164 /* slot to fill in autolayout
9165 */
9166 new_disk->raiddisk = new_disk->index;
9167 new_disk->disk.status |=
9168 CONFIGURED_DISK;
9169 new_disk->disk.status &= ~SPARE_DISK;
9170 }
9171
ed7333bd
AK		 9172 dprintf("imsm: volume set mpb->num_raid_devs = %i\n",
9173 mpb->num_raid_devs);
2e5dc010
N		 9174 /* manage changes in volume
9175 */
d098291a 9176 for (dev_id = 0; dev_id < mpb->num_raid_devs; dev_id++) {
2e5dc010
N		 9177 void **sp = *space_list;
9178 struct imsm_dev *newdev;
9179 struct imsm_map *newmap, *oldmap;
9180
d098291a
AK		 9181 for (id = super->devlist ; id; id = id->next) {
9182 if (id->index == dev_id)
9183 break;
9184 }
9185 if (id == NULL)
9186 break;
2e5dc010
N		 9187 if (!sp)
9188 continue;
9189 *space_list = *sp;
9190 newdev = (void*)sp;
9191 /* Copy the dev, but not (all of) the map */
9192 memcpy(newdev, id->dev, sizeof(*newdev));
238c0a71
AK		 9193 oldmap = get_imsm_map(id->dev, MAP_0);
9194 newmap = get_imsm_map(newdev, MAP_0);
2e5dc010
N		 9195 /* Copy the current map */
9196 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
9197 /* update one device only
9198 */
9199 if (devices_to_reshape) {
ed7333bd
AK		 9200 dprintf("imsm: modifying subdev: %i\n",
9201 id->index);
2e5dc010
N		 9202 devices_to_reshape--;
9203 newdev->vol.migr_state = 1;
9204 newdev->vol.curr_migr_unit = 0;
ea672ee1 9205 set_migr_type(newdev, MIGR_GEN_MIGR);
2e5dc010
N		 9206 newmap->num_members = u->new_raid_disks;
9207 for (i = 0; i < delta_disks; i++) {
9208 set_imsm_ord_tbl_ent(newmap,
9209 u->old_raid_disks + i,
9210 u->old_raid_disks + i);
9211 }
9212 /* New map is correct, now need to save old map
9213 */
238c0a71 9214 newmap = get_imsm_map(newdev, MAP_1);
2e5dc010
N		 9215 memcpy(newmap, oldmap, sizeof_imsm_map(oldmap));
9216
f3871fdc 9217 imsm_set_array_size(newdev, -1);
2e5dc010
N		 9218 }
9219
9220 sp = (void **)id->dev;
9221 id->dev = newdev;
9222 *sp = tofree;
9223 tofree = sp;
8e59f3d8
AK		 9224
9225 /* Clear migration record */
9226 memset(super->migr_rec, 0, sizeof(struct migr_record));
2e5dc010 9227 }
819bc634
AK		 9228 if (tofree)
9229 *space_list = tofree;
2e5dc010
N		 9230 ret_val = 1;
9231
9232update_reshape_exit:
9233
9234 return ret_val;
9235}
9236
bb025c2f 9237static int apply_takeover_update(struct imsm_update_takeover *u,
8ca6df95
KW		 9238 struct intel_super *super,
9239 void ***space_list)
bb025c2f
KW		 9240{
9241 struct imsm_dev *dev = NULL;
8ca6df95
KW		 9242 struct intel_dev *dv;
9243 struct imsm_dev *dev_new;
bb025c2f
KW		 9244 struct imsm_map *map;
9245 struct dl *dm, *du;
8ca6df95 9246 int i;
bb025c2f
KW		 9247
9248 for (dv = super->devlist; dv; dv = dv->next)
9249 if (dv->index == (unsigned int)u->subarray) {
9250 dev = dv->dev;
9251 break;
9252 }
9253
9254 if (dev == NULL)
9255 return 0;
9256
238c0a71 9257 map = get_imsm_map(dev, MAP_0);
bb025c2f
KW		 9258
9259 if (u->direction == R10_TO_R0) {
06fb291a
PB		 9260 unsigned long long num_data_stripes;
9261
9262 map->num_domains = 1;
9263 num_data_stripes = blocks_per_member(map);
9264 num_data_stripes /= map->blocks_per_strip;
9265 num_data_stripes /= map->num_domains;
9266 set_num_data_stripes(map, num_data_stripes);
9267
43d5ec18 9268 /* Number of failed disks must be half of initial disk number */
3b451610
AK		 9269 if (imsm_count_failed(super, dev, MAP_0) !=
9270 (map->num_members / 2))
43d5ec18
KW		 9271 return 0;
9272
bb025c2f
KW		 9273 /* iterate through devices to mark removed disks as spare */
9274 for (dm = super->disks; dm; dm = dm->next) {
9275 if (dm->disk.status & FAILED_DISK) {
9276 int idx = dm->index;
9277 /* update indexes on the disk list */
9278/* FIXME this loop-with-the-loop looks wrong, I'm not convinced
9279 the index values will end up being correct.... NB */
9280 for (du = super->disks; du; du = du->next)
9281 if (du->index > idx)
9282 du->index--;
9283 /* mark as spare disk */
a8619d23 9284 mark_spare(dm);
bb025c2f
KW		 9285 }
9286 }
bb025c2f
KW		 9287 /* update map */
9288 map->num_members = map->num_members / 2;
9289 map->map_state = IMSM_T_STATE_NORMAL;
9290 map->num_domains = 1;
9291 map->raid_level = 0;
9292 map->failed_disk_num = -1;
9293 }
9294
8ca6df95
KW		 9295 if (u->direction == R0_TO_R10) {
9296 void **space;
9297 /* update slots in current disk list */
9298 for (dm = super->disks; dm; dm = dm->next) {
9299 if (dm->index >= 0)
9300 dm->index *= 2;
9301 }
9302 /* create new *missing* disks */
9303 for (i = 0; i < map->num_members; i++) {
9304 space = *space_list;
9305 if (!space)
9306 continue;
9307 *space_list = *space;
9308 du = (void *)space;
9309 memcpy(du, super->disks, sizeof(*du));
8ca6df95
KW		 9310 du->fd = -1;
9311 du->minor = 0;
9312 du->major = 0;
9313 du->index = (i * 2) + 1;
9314 sprintf((char *)du->disk.serial,
9315 " MISSING_%d", du->index);
9316 sprintf((char *)du->serial,
9317 "MISSING_%d", du->index);
9318 du->next = super->missing;
9319 super->missing = du;
9320 }
9321 /* create new dev and map */
9322 space = *space_list;
9323 if (!space)
9324 return 0;
9325 *space_list = *space;
9326 dev_new = (void *)space;
9327 memcpy(dev_new, dev, sizeof(*dev));
9328 /* update new map */
238c0a71 9329 map = get_imsm_map(dev_new, MAP_0);
8ca6df95 9330 map->num_members = map->num_members * 2;
1a2487c2 9331 map->map_state = IMSM_T_STATE_DEGRADED;
8ca6df95
KW		 9332 map->num_domains = 2;
9333 map->raid_level = 1;
9334 /* replace dev<->dev_new */
9335 dv->dev = dev_new;
9336 }
bb025c2f
KW		 9337 /* update disk order table */
9338 for (du = super->disks; du; du = du->next)
9339 if (du->index >= 0)
9340 set_imsm_ord_tbl_ent(map, du->index, du->index);
8ca6df95 9341 for (du = super->missing; du; du = du->next)
1a2487c2
KW		 9342 if (du->index >= 0) {
9343 set_imsm_ord_tbl_ent(map, du->index, du->index);
4c9e8c1e 9344 mark_missing(super, dv->dev, &du->disk, du->index);
1a2487c2 9345 }
bb025c2f
KW		 9346
9347 return 1;
9348}
9349
e8319a19
DW		 9350static void imsm_process_update(struct supertype *st,
9351 struct metadata_update *update)
9352{
9353 /**
9354 * crack open the metadata_update envelope to find the update record
9355 * update can be one of:
d195167d
AK		 9356 * update_reshape_container_disks - all the arrays in the container
9357 * are being reshaped to have more devices. We need to mark
9358 * the arrays for general migration and convert selected spares
9359 * into active devices.
9360 * update_activate_spare - a spare device has replaced a failed
1011e834
N		 9361 * device in an array, update the disk_ord_tbl. If this disk is
9362 * present in all member arrays then also clear the SPARE_DISK
9363 * flag
d195167d
AK		 9364 * update_create_array
9365 * update_kill_array
9366 * update_rename_array
9367 * update_add_remove_disk
e8319a19
DW		 9368 */
9369 struct intel_super *super = st->sb;
4d7b1503 9370 struct imsm_super *mpb;
e8319a19
DW		 9371 enum imsm_update_type type = *(enum imsm_update_type *) update->buf;
9372
4d7b1503
DW		 9373 /* update requires a larger buf but the allocation failed */
9374 if (super->next_len && !super->next_buf) {
9375 super->next_len = 0;
9376 return;
9377 }
9378
9379 if (super->next_buf) {
9380 memcpy(super->next_buf, super->buf, super->len);
9381 free(super->buf);
9382 super->len = super->next_len;
9383 super->buf = super->next_buf;
9384
9385 super->next_len = 0;
9386 super->next_buf = NULL;
9387 }
9388
9389 mpb = super->anchor;
9390
e8319a19 9391 switch (type) {
0ec5d470
AK		 9392 case update_general_migration_checkpoint: {
9393 struct intel_dev *id;
9394 struct imsm_update_general_migration_checkpoint *u =
9395 (void *)update->buf;
9396
1ade5cc1 9397 dprintf("called for update_general_migration_checkpoint\n");
0ec5d470
AK		 9398
9399 /* find device under general migration */
9400 for (id = super->devlist ; id; id = id->next) {
9401 if (is_gen_migration(id->dev)) {
9402 id->dev->vol.curr_migr_unit =
9403 __cpu_to_le32(u->curr_migr_unit);
9404 super->updates_pending++;
9405 }
9406 }
9407 break;
9408 }
bb025c2f
KW		 9409 case update_takeover: {
9410 struct imsm_update_takeover *u = (void *)update->buf;
1a2487c2
KW		 9411 if (apply_takeover_update(u, super, &update->space_list)) {
9412 imsm_update_version_info(super);
bb025c2f 9413 super->updates_pending++;
1a2487c2 9414 }
bb025c2f
KW		 9415 break;
9416 }
9417
78b10e66 9418 case update_reshape_container_disks: {
d195167d 9419 struct imsm_update_reshape *u = (void *)update->buf;
2e5dc010
N		 9420 if (apply_reshape_container_disks_update(
9421 u, super, &update->space_list))
9422 super->updates_pending++;
78b10e66
N		 9423 break;
9424 }
48c5303a 9425 case update_reshape_migration: {
a29911da
PC		 9426 struct imsm_update_reshape_migration *u = (void *)update->buf;
9427 if (apply_reshape_migration_update(
9428 u, super, &update->space_list))
9429 super->updates_pending++;
48c5303a
PC		 9430 break;
9431 }
f3871fdc
AK		 9432 case update_size_change: {
9433 struct imsm_update_size_change *u = (void *)update->buf;
9434 if (apply_size_change_update(u, super))
9435 super->updates_pending++;
9436 break;
9437 }
e8319a19 9438 case update_activate_spare: {
1011e834 9439 struct imsm_update_activate_spare *u = (void *) update->buf;
061d7da3
LO		 9440 if (apply_update_activate_spare(u, super, st->arrays))
9441 super->updates_pending++;
8273f55e
DW		 9442 break;
9443 }
9444 case update_create_array: {
9445 /* someone wants to create a new array, we need to be aware of
9446 * a few races/collisions:
9447 * 1/ 'Create' called by two separate instances of mdadm
9448 * 2/ 'Create' versus 'activate_spare': mdadm has chosen
9449 * devices that have since been assimilated via
9450 * activate_spare.
9451 * In the event this update can not be carried out mdadm will
9452 * (FIX ME) notice that its update did not take hold.
9453 */
9454 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 9455 struct intel_dev *dv;
8273f55e
DW		 9456 struct imsm_dev *dev;
9457 struct imsm_map *map, *new_map;
9458 unsigned long long start, end;
9459 unsigned long long new_start, new_end;
9460 int i;
54c2c1ea
DW		 9461 struct disk_info *inf;
9462 struct dl *dl;
8273f55e
DW		 9463
9464 /* handle racing creates: first come first serve */
9465 if (u->dev_idx < mpb->num_raid_devs) {
1ade5cc1 9466 dprintf("subarray %d already defined\n", u->dev_idx);
ba2de7ba 9467 goto create_error;
8273f55e
DW		 9468 }
9469
9470 /* check update is next in sequence */
9471 if (u->dev_idx != mpb->num_raid_devs) {
1ade5cc1
N		 9472 dprintf("can not create array %d expected index %d\n",
9473 u->dev_idx, mpb->num_raid_devs);
ba2de7ba 9474 goto create_error;
8273f55e
DW		 9475 }
9476
238c0a71 9477 new_map = get_imsm_map(&u->dev, MAP_0);
5551b113
CA		 9478 new_start = pba_of_lba0(new_map);
9479 new_end = new_start + blocks_per_member(new_map);
54c2c1ea 9480 inf = get_disk_info(u);
8273f55e
DW		 9481
9482 /* handle activate_spare versus create race:
9483 * check to make sure that overlapping arrays do not include
9484 * overalpping disks
9485 */
9486 for (i = 0; i < mpb->num_raid_devs; i++) {
949c47a0 9487 dev = get_imsm_dev(super, i);
238c0a71 9488 map = get_imsm_map(dev, MAP_0);
5551b113
CA		 9489 start = pba_of_lba0(map);
9490 end = start + blocks_per_member(map);
8273f55e
DW		 9491 if ((new_start >= start && new_start <= end) ||
9492 (start >= new_start && start <= new_end))
54c2c1ea
DW		 9493 /* overlap */;
9494 else
9495 continue;
9496
9497 if (disks_overlap(super, i, u)) {
1ade5cc1 9498 dprintf("arrays overlap\n");
ba2de7ba 9499 goto create_error;
8273f55e
DW		 9500 }
9501 }
8273f55e 9502
949c47a0
DW		 9503 /* check that prepare update was successful */
9504 if (!update->space) {
1ade5cc1 9505 dprintf("prepare update failed\n");
ba2de7ba 9506 goto create_error;
949c47a0
DW		 9507 }
9508
54c2c1ea
DW		 9509 /* check that all disks are still active before committing
9510 * changes. FIXME: could we instead handle this by creating a
9511 * degraded array? That's probably not what the user expects,
9512 * so better to drop this update on the floor.
9513 */
9514 for (i = 0; i < new_map->num_members; i++) {
9515 dl = serial_to_dl(inf[i].serial, super);
9516 if (!dl) {
1ade5cc1 9517 dprintf("disk disappeared\n");
ba2de7ba 9518 goto create_error;
54c2c1ea 9519 }
949c47a0
DW		 9520 }
9521
8273f55e 9522 super->updates_pending++;
54c2c1ea
DW		 9523
9524 /* convert spares to members and fixup ord_tbl */
9525 for (i = 0; i < new_map->num_members; i++) {
9526 dl = serial_to_dl(inf[i].serial, super);
9527 if (dl->index == -1) {
9528 dl->index = mpb->num_disks;
9529 mpb->num_disks++;
9530 dl->disk.status |= CONFIGURED_DISK;
9531 dl->disk.status &= ~SPARE_DISK;
9532 }
9533 set_imsm_ord_tbl_ent(new_map, i, dl->index);
9534 }
9535
ba2de7ba
DW		 9536 dv = update->space;
9537 dev = dv->dev;
949c47a0
DW		 9538 update->space = NULL;
9539 imsm_copy_dev(dev, &u->dev);
ba2de7ba
DW		 9540 dv->index = u->dev_idx;
9541 dv->next = super->devlist;
9542 super->devlist = dv;
8273f55e 9543 mpb->num_raid_devs++;
8273f55e 9544
4d1313e9 9545 imsm_update_version_info(super);
8273f55e 9546 break;
ba2de7ba
DW		 9547 create_error:
9548 /* mdmon knows how to release update->space, but not
9549 * ((struct intel_dev *) update->space)->dev
9550 */
9551 if (update->space) {
9552 dv = update->space;
9553 free(dv->dev);
9554 }
8273f55e 9555 break;
e8319a19 9556 }
33414a01
DW		 9557 case update_kill_array: {
9558 struct imsm_update_kill_array *u = (void *) update->buf;
9559 int victim = u->dev_idx;
9560 struct active_array *a;
9561 struct intel_dev **dp;
9562 struct imsm_dev *dev;
9563
9564 /* sanity check that we are not affecting the uuid of
9565 * active arrays, or deleting an active array
9566 *
9567 * FIXME when immutable ids are available, but note that
9568 * we'll also need to fixup the invalidated/active
9569 * subarray indexes in mdstat
9570 */
9571 for (a = st->arrays; a; a = a->next)
9572 if (a->info.container_member >= victim)
9573 break;
9574 /* by definition if mdmon is running at least one array
9575 * is active in the container, so checking
9576 * mpb->num_raid_devs is just extra paranoia
9577 */
9578 dev = get_imsm_dev(super, victim);
9579 if (a || !dev || mpb->num_raid_devs == 1) {
9580 dprintf("failed to delete subarray-%d\n", victim);
9581 break;
9582 }
9583
9584 for (dp = &super->devlist; *dp;)
f21e18ca 9585 if ((*dp)->index == (unsigned)super->current_vol) {
33414a01
DW		 9586 *dp = (*dp)->next;
9587 } else {
f21e18ca 9588 if ((*dp)->index > (unsigned)victim)
33414a01
DW		 9589 (*dp)->index--;
9590 dp = &(*dp)->next;
9591 }
9592 mpb->num_raid_devs--;
9593 super->updates_pending++;
9594 break;
9595 }
aa534678
DW		 9596 case update_rename_array: {
9597 struct imsm_update_rename_array *u = (void *) update->buf;
9598 char name[MAX_RAID_SERIAL_LEN+1];
9599 int target = u->dev_idx;
9600 struct active_array *a;
9601 struct imsm_dev *dev;
9602
9603 /* sanity check that we are not affecting the uuid of
9604 * an active array
9605 */
9606 snprintf(name, MAX_RAID_SERIAL_LEN, "%s", (char *) u->name);
9607 name[MAX_RAID_SERIAL_LEN] = '\0';
9608 for (a = st->arrays; a; a = a->next)
9609 if (a->info.container_member == target)
9610 break;
9611 dev = get_imsm_dev(super, u->dev_idx);
9612 if (a || !dev || !check_name(super, name, 1)) {
9613 dprintf("failed to rename subarray-%d\n", target);
9614 break;
9615 }
9616
cdbe98cd 9617 snprintf((char *) dev->volume, MAX_RAID_SERIAL_LEN, "%s", name);
aa534678
DW		 9618 super->updates_pending++;
9619 break;
9620 }
1a64be56 9621 case update_add_remove_disk: {
43dad3d6 9622 /* we may be able to repair some arrays if disks are
095b8088 9623 * being added, check the status of add_remove_disk
1a64be56
LM		 9624 * if discs has been added.
9625 */
9626 if (add_remove_disk_update(super)) {
43dad3d6 9627 struct active_array *a;
072b727f
DW		 9628
9629 super->updates_pending++;
1a64be56 9630 for (a = st->arrays; a; a = a->next)
43dad3d6
DW		 9631 a->check_degraded = 1;
9632 }
43dad3d6 9633 break;
e8319a19 9634 }
bbab0940
TM		 9635 case update_prealloc_badblocks_mem:
9636 break;
e6e9dd3f
AP		 9637 case update_rwh_policy: {
9638 struct imsm_update_rwh_policy *u = (void *)update->buf;
9639 int target = u->dev_idx;
9640 struct imsm_dev *dev = get_imsm_dev(super, target);
9641 if (!dev) {
9642 dprintf("could not find subarray-%d\n", target);
9643 break;
9644 }
9645
9646 if (dev->rwh_policy != u->new_policy) {
9647 dev->rwh_policy = u->new_policy;
9648 super->updates_pending++;
9649 }
9650 break;
9651 }
1a64be56 9652 default:
7a862a02 9653 pr_err("error: unsuported process update type:(type: %d)\n", type);
1a64be56 9654 }
e8319a19 9655}
88758e9d 9656
bc0b9d34
PC		 9657static struct mdinfo *get_spares_for_grow(struct supertype *st);
9658
5fe6f031
N		 9659static int imsm_prepare_update(struct supertype *st,
9660 struct metadata_update *update)
8273f55e 9661{
949c47a0 9662 /**
4d7b1503
DW		 9663 * Allocate space to hold new disk entries, raid-device entries or a new
9664 * mpb if necessary. The manager synchronously waits for updates to
9665 * complete in the monitor, so new mpb buffers allocated here can be
9666 * integrated by the monitor thread without worrying about live pointers
9667 * in the manager thread.
8273f55e 9668 */
095b8088 9669 enum imsm_update_type type;
4d7b1503 9670 struct intel_super *super = st->sb;
f36a9ecd 9671 unsigned int sector_size = super->sector_size;
4d7b1503
DW		 9672 struct imsm_super *mpb = super->anchor;
9673 size_t buf_len;
9674 size_t len = 0;
949c47a0 9675
095b8088
N		 9676 if (update->len < (int)sizeof(type))
9677 return 0;
9678
9679 type = *(enum imsm_update_type *) update->buf;
9680
949c47a0 9681 switch (type) {
0ec5d470 9682 case update_general_migration_checkpoint:
095b8088
N		 9683 if (update->len < (int)sizeof(struct imsm_update_general_migration_checkpoint))
9684 return 0;
1ade5cc1 9685 dprintf("called for update_general_migration_checkpoint\n");
0ec5d470 9686 break;
abedf5fc
KW		 9687 case update_takeover: {
9688 struct imsm_update_takeover *u = (void *)update->buf;
095b8088
N		 9689 if (update->len < (int)sizeof(*u))
9690 return 0;
abedf5fc
KW		 9691 if (u->direction == R0_TO_R10) {
9692 void **tail = (void **)&update->space_list;
9693 struct imsm_dev *dev = get_imsm_dev(super, u->subarray);
238c0a71 9694 struct imsm_map *map = get_imsm_map(dev, MAP_0);
abedf5fc
KW		 9695 int num_members = map->num_members;
9696 void *space;
9697 int size, i;
abedf5fc
KW		 9698 /* allocate memory for added disks */
9699 for (i = 0; i < num_members; i++) {
9700 size = sizeof(struct dl);
503975b9 9701 space = xmalloc(size);
abedf5fc
KW		 9702 *tail = space;
9703 tail = space;
9704 *tail = NULL;
9705 }
9706 /* allocate memory for new device */
9707 size = sizeof_imsm_dev(super->devlist->dev, 0) +
9708 (num_members * sizeof(__u32));
503975b9
N		 9709 space = xmalloc(size);
9710 *tail = space;
9711 tail = space;
9712 *tail = NULL;
9713 len = disks_to_mpb_size(num_members * 2);
abedf5fc
KW		 9714 }
9715
9716 break;
9717 }
78b10e66 9718 case update_reshape_container_disks: {
d195167d
AK		 9719 /* Every raid device in the container is about to
9720 * gain some more devices, and we will enter a
9721 * reconfiguration.
9722 * So each 'imsm_map' will be bigger, and the imsm_vol
9723 * will now hold 2 of them.
9724 * Thus we need new 'struct imsm_dev' allocations sized
9725 * as sizeof_imsm_dev but with more devices in both maps.
9726 */
9727 struct imsm_update_reshape *u = (void *)update->buf;
9728 struct intel_dev *dl;
9729 void **space_tail = (void**)&update->space_list;
9730
095b8088
N		 9731 if (update->len < (int)sizeof(*u))
9732 return 0;
9733
1ade5cc1 9734 dprintf("for update_reshape\n");
d195167d
AK		 9735
9736 for (dl = super->devlist; dl; dl = dl->next) {
9737 int size = sizeof_imsm_dev(dl->dev, 1);
9738 void *s;
d677e0b8
AK		 9739 if (u->new_raid_disks > u->old_raid_disks)
9740 size += sizeof(__u32)*2*
9741 (u->new_raid_disks - u->old_raid_disks);
503975b9 9742 s = xmalloc(size);
d195167d
AK		 9743 *space_tail = s;
9744 space_tail = s;
9745 *space_tail = NULL;
9746 }
9747
9748 len = disks_to_mpb_size(u->new_raid_disks);
9749 dprintf("New anchor length is %llu\n", (unsigned long long)len);
78b10e66
N		 9750 break;
9751 }
48c5303a 9752 case update_reshape_migration: {
bc0b9d34
PC		 9753 /* for migration level 0->5 we need to add disks
9754 * so the same as for container operation we will copy
9755 * device to the bigger location.
9756 * in memory prepared device and new disk area are prepared
9757 * for usage in process update
9758 */
9759 struct imsm_update_reshape_migration *u = (void *)update->buf;
9760 struct intel_dev *id;
9761 void **space_tail = (void **)&update->space_list;
9762 int size;
9763 void *s;
9764 int current_level = -1;
9765
095b8088
N		 9766 if (update->len < (int)sizeof(*u))
9767 return 0;
9768
1ade5cc1 9769 dprintf("for update_reshape\n");
bc0b9d34
PC		 9770
9771 /* add space for bigger array in update
9772 */
9773 for (id = super->devlist; id; id = id->next) {
9774 if (id->index == (unsigned)u->subdev) {
9775 size = sizeof_imsm_dev(id->dev, 1);
9776 if (u->new_raid_disks > u->old_raid_disks)
9777 size += sizeof(__u32)*2*
9778 (u->new_raid_disks - u->old_raid_disks);
503975b9 9779 s = xmalloc(size);
bc0b9d34
PC		 9780 *space_tail = s;
9781 space_tail = s;
9782 *space_tail = NULL;
9783 break;
9784 }
9785 }
9786 if (update->space_list == NULL)
9787 break;
9788
9789 /* add space for disk in update
9790 */
9791 size = sizeof(struct dl);
503975b9 9792 s = xmalloc(size);
bc0b9d34
PC		 9793 *space_tail = s;
9794 space_tail = s;
9795 *space_tail = NULL;
9796
9797 /* add spare device to update
9798 */
9799 for (id = super->devlist ; id; id = id->next)
9800 if (id->index == (unsigned)u->subdev) {
9801 struct imsm_dev *dev;
9802 struct imsm_map *map;
9803
9804 dev = get_imsm_dev(super, u->subdev);
238c0a71 9805 map = get_imsm_map(dev, MAP_0);
bc0b9d34
PC		 9806 current_level = map->raid_level;
9807 break;
9808 }
089f9d79 9809 if (u->new_level == 5 && u->new_level != current_level) {
bc0b9d34
PC		 9810 struct mdinfo *spares;
9811
9812 spares = get_spares_for_grow(st);
9813 if (spares) {
9814 struct dl *dl;
9815 struct mdinfo *dev;
9816
9817 dev = spares->devs;
9818 if (dev) {
9819 u->new_disks[0] =
9820 makedev(dev->disk.major,
9821 dev->disk.minor);
9822 dl = get_disk_super(super,
9823 dev->disk.major,
9824 dev->disk.minor);
9825 dl->index = u->old_raid_disks;
9826 dev = dev->next;
9827 }
9828 sysfs_free(spares);
9829 }
9830 }
9831 len = disks_to_mpb_size(u->new_raid_disks);
9832 dprintf("New anchor length is %llu\n", (unsigned long long)len);
48c5303a
PC		 9833 break;
9834 }
f3871fdc 9835 case update_size_change: {
095b8088
N		 9836 if (update->len < (int)sizeof(struct imsm_update_size_change))
9837 return 0;
9838 break;
9839 }
9840 case update_activate_spare: {
9841 if (update->len < (int)sizeof(struct imsm_update_activate_spare))
9842 return 0;
f3871fdc
AK		 9843 break;
9844 }
949c47a0
DW		 9845 case update_create_array: {
9846 struct imsm_update_create_array *u = (void *) update->buf;
ba2de7ba 9847 struct intel_dev *dv;
54c2c1ea 9848 struct imsm_dev *dev = &u->dev;
238c0a71 9849 struct imsm_map *map = get_imsm_map(dev, MAP_0);
54c2c1ea
DW		 9850 struct dl *dl;
9851 struct disk_info *inf;
9852 int i;
9853 int activate = 0;
949c47a0 9854
095b8088
N		 9855 if (update->len < (int)sizeof(*u))
9856 return 0;
9857
54c2c1ea
DW		 9858 inf = get_disk_info(u);
9859 len = sizeof_imsm_dev(dev, 1);
ba2de7ba 9860 /* allocate a new super->devlist entry */
503975b9
N		 9861 dv = xmalloc(sizeof(*dv));
9862 dv->dev = xmalloc(len);
9863 update->space = dv;
949c47a0 9864
54c2c1ea
DW		 9865 /* count how many spares will be converted to members */
9866 for (i = 0; i < map->num_members; i++) {
9867 dl = serial_to_dl(inf[i].serial, super);
9868 if (!dl) {
9869 /* hmm maybe it failed?, nothing we can do about
9870 * it here
9871 */
9872 continue;
9873 }
9874 if (count_memberships(dl, super) == 0)
9875 activate++;
9876 }
9877 len += activate * sizeof(struct imsm_disk);
949c47a0 9878 break;
095b8088
N		 9879 }
9880 case update_kill_array: {
9881 if (update->len < (int)sizeof(struct imsm_update_kill_array))
9882 return 0;
949c47a0
DW		 9883 break;
9884 }
095b8088
N		 9885 case update_rename_array: {
9886 if (update->len < (int)sizeof(struct imsm_update_rename_array))
9887 return 0;
9888 break;
9889 }
9890 case update_add_remove_disk:
9891 /* no update->len needed */
9892 break;
bbab0940
TM		 9893 case update_prealloc_badblocks_mem:
9894 super->extra_space += sizeof(struct bbm_log) -
9895 get_imsm_bbm_log_size(super->bbm_log);
9896 break;
e6e9dd3f
AP		 9897 case update_rwh_policy: {
9898 if (update->len < (int)sizeof(struct imsm_update_rwh_policy))
9899 return 0;
9900 break;
9901 }
095b8088
N		 9902 default:
9903 return 0;
949c47a0 9904 }
8273f55e 9905
4d7b1503
DW		 9906 /* check if we need a larger metadata buffer */
9907 if (super->next_buf)
9908 buf_len = super->next_len;
9909 else
9910 buf_len = super->len;
9911
bbab0940 9912 if (__le32_to_cpu(mpb->mpb_size) + super->extra_space + len > buf_len) {
4d7b1503
DW		 9913 /* ok we need a larger buf than what is currently allocated
9914 * if this allocation fails process_update will notice that
9915 * ->next_len is set and ->next_buf is NULL
9916 */
bbab0940
TM		 9917 buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) +
9918 super->extra_space + len, sector_size);
4d7b1503
DW		 9919 if (super->next_buf)
9920 free(super->next_buf);
9921
9922 super->next_len = buf_len;
f36a9ecd 9923 if (posix_memalign(&super->next_buf, sector_size, buf_len) == 0)
1f45a8ad
DW		 9924 memset(super->next_buf, 0, buf_len);
9925 else
4d7b1503
DW		 9926 super->next_buf = NULL;
9927 }
5fe6f031 9928 return 1;
8273f55e
DW		 9929}
9930
ae6aad82 9931/* must be called while manager is quiesced */
f21e18ca 9932static void imsm_delete(struct intel_super *super, struct dl **dlp, unsigned index)
ae6aad82
DW		 9933{
9934 struct imsm_super *mpb = super->anchor;
ae6aad82
DW		 9935 struct dl *iter;
9936 struct imsm_dev *dev;
9937 struct imsm_map *map;
4c9e8c1e 9938 unsigned int i, j, num_members;
24565c9a 9939 __u32 ord;
4c9e8c1e 9940 struct bbm_log *log = super->bbm_log;
ae6aad82 9941
1ade5cc1 9942 dprintf("deleting device[%d] from imsm_super\n", index);
ae6aad82
DW		 9943
9944 /* shift all indexes down one */
9945 for (iter = super->disks; iter; iter = iter->next)
f21e18ca 9946 if (iter->index > (int)index)
ae6aad82 9947 iter->index--;
47ee5a45 9948 for (iter = super->missing; iter; iter = iter->next)
f21e18ca 9949 if (iter->index > (int)index)
47ee5a45 9950 iter->index--;
ae6aad82
DW		 9951
9952 for (i = 0; i < mpb->num_raid_devs; i++) {
9953 dev = get_imsm_dev(super, i);
238c0a71 9954 map = get_imsm_map(dev, MAP_0);
24565c9a
DW		 9955 num_members = map->num_members;
9956 for (j = 0; j < num_members; j++) {
9957 /* update ord entries being careful not to propagate
9958 * ord-flags to the first map
9959 */
238c0a71 9960 ord = get_imsm_ord_tbl_ent(dev, j, MAP_X);
ae6aad82 9961
24565c9a
DW		 9962 if (ord_to_idx(ord) <= index)
9963 continue;
ae6aad82 9964
238c0a71 9965 map = get_imsm_map(dev, MAP_0);
24565c9a 9966 set_imsm_ord_tbl_ent(map, j, ord_to_idx(ord - 1));
238c0a71 9967 map = get_imsm_map(dev, MAP_1);
24565c9a
DW		 9968 if (map)
9969 set_imsm_ord_tbl_ent(map, j, ord - 1);
ae6aad82
DW		 9970 }
9971 }
9972
4c9e8c1e
TM		 9973 for (i = 0; i < log->entry_count; i++) {
9974 struct bbm_log_entry *entry = &log->marked_block_entries[i];
9975
9976 if (entry->disk_ordinal <= index)
9977 continue;
9978 entry->disk_ordinal--;
9979 }
9980
ae6aad82
DW		 9981 mpb->num_disks--;
9982 super->updates_pending++;
24565c9a
DW		 9983 if (*dlp) {
9984 struct dl *dl = *dlp;
9985
9986 *dlp = (*dlp)->next;
9987 __free_imsm_disk(dl);
9988 }
ae6aad82 9989}
9e2d750d 9990#endif /* MDASSEMBLE */
9a717282
AK		 9991
9992static void close_targets(int *targets, int new_disks)
9993{
9994 int i;
9995
9996 if (!targets)
9997 return;
9998
9999 for (i = 0; i < new_disks; i++) {
10000 if (targets[i] >= 0) {
10001 close(targets[i]);
10002 targets[i] = -1;
10003 }
10004 }
10005}
10006
10007static int imsm_get_allowed_degradation(int level, int raid_disks,
10008 struct intel_super *super,
10009 struct imsm_dev *dev)
10010{
10011 switch (level) {
bf5cf7c7 10012 case 1:
9a717282
AK		 10013 case 10:{
10014 int ret_val = 0;
10015 struct imsm_map *map;
10016 int i;
10017
10018 ret_val = raid_disks/2;
10019 /* check map if all disks pairs not failed
10020 * in both maps
10021 */
238c0a71 10022 map = get_imsm_map(dev, MAP_0);
9a717282
AK		 10023 for (i = 0; i < ret_val; i++) {
10024 int degradation = 0;
10025 if (get_imsm_disk(super, i) == NULL)
10026 degradation++;
10027 if (get_imsm_disk(super, i + 1) == NULL)
10028 degradation++;
10029 if (degradation == 2)
10030 return 0;
10031 }
238c0a71 10032 map = get_imsm_map(dev, MAP_1);
9a717282
AK		 10033 /* if there is no second map
10034 * result can be returned
10035 */
10036 if (map == NULL)
10037 return ret_val;
10038 /* check degradation in second map
10039 */
10040 for (i = 0; i < ret_val; i++) {
10041 int degradation = 0;
10042 if (get_imsm_disk(super, i) == NULL)
10043 degradation++;
10044 if (get_imsm_disk(super, i + 1) == NULL)
10045 degradation++;
10046 if (degradation == 2)
10047 return 0;
10048 }
10049 return ret_val;
10050 }
10051 case 5:
10052 return 1;
10053 case 6:
10054 return 2;
10055 default:
10056 return 0;
10057 }
10058}
10059
687629c2
AK		 10060/*******************************************************************************
10061 * Function: open_backup_targets
10062 * Description: Function opens file descriptors for all devices given in
10063 * info->devs
10064 * Parameters:
10065 * info : general array info
10066 * raid_disks : number of disks
10067 * raid_fds : table of device's file descriptors
9a717282
AK		 10068 * super : intel super for raid10 degradation check
10069 * dev : intel device for raid10 degradation check
687629c2
AK		 10070 * Returns:
10071 * 0 : success
10072 * -1 : fail
10073 ******************************************************************************/
9a717282
AK		 10074int open_backup_targets(struct mdinfo *info, int raid_disks, int *raid_fds,
10075 struct intel_super *super, struct imsm_dev *dev)
687629c2
AK		 10076{
10077 struct mdinfo *sd;
f627f5ad 10078 int i;
9a717282 10079 int opened = 0;
f627f5ad
AK		 10080
10081 for (i = 0; i < raid_disks; i++)
10082 raid_fds[i] = -1;
687629c2
AK		 10083
10084 for (sd = info->devs ; sd ; sd = sd->next) {
10085 char *dn;
10086
10087 if (sd->disk.state & (1<<MD_DISK_FAULTY)) {
10088 dprintf("disk is faulty!!\n");
10089 continue;
10090 }
10091
089f9d79 10092 if (sd->disk.raid_disk >= raid_disks || sd->disk.raid_disk < 0)
687629c2
AK		 10093 continue;
10094
10095 dn = map_dev(sd->disk.major,
10096 sd->disk.minor, 1);
10097 raid_fds[sd->disk.raid_disk] = dev_open(dn, O_RDWR);
10098 if (raid_fds[sd->disk.raid_disk] < 0) {
e12b3daa 10099 pr_err("cannot open component\n");
9a717282 10100 continue;
687629c2 10101 }
9a717282
AK		 10102 opened++;
10103 }
10104 /* check if maximum array degradation level is not exceeded
10105 */
10106 if ((raid_disks - opened) >
089f9d79
JS		 10107 imsm_get_allowed_degradation(info->new_level, raid_disks,
10108 super, dev)) {
e12b3daa 10109 pr_err("Not enough disks can be opened.\n");
9a717282
AK		 10110 close_targets(raid_fds, raid_disks);
10111 return -2;
687629c2
AK		 10112 }
10113 return 0;
10114}
10115
d31ad643
PB		 10116/*******************************************************************************
10117 * Function: validate_container_imsm
10118 * Description: This routine validates container after assemble,
10119 * eg. if devices in container are under the same controller.
10120 *
10121 * Parameters:
10122 * info : linked list with info about devices used in array
10123 * Returns:
10124 * 1 : HBA mismatch
10125 * 0 : Success
10126 ******************************************************************************/
10127int validate_container_imsm(struct mdinfo *info)
10128{
6b781d33
AP		 10129 if (check_env("IMSM_NO_PLATFORM"))
10130 return 0;
d31ad643 10131
6b781d33
AP		 10132 struct sys_dev *idev;
10133 struct sys_dev *hba = NULL;
10134 struct sys_dev *intel_devices = find_intel_devices();
10135 char *dev_path = devt_to_devpath(makedev(info->disk.major,
10136 info->disk.minor));
10137
10138 for (idev = intel_devices; idev; idev = idev->next) {
10139 if (dev_path && strstr(dev_path, idev->path)) {
10140 hba = idev;
10141 break;
d31ad643 10142 }
6b781d33
AP		 10143 }
10144 if (dev_path)
d31ad643
PB		 10145 free(dev_path);
10146
6b781d33
AP		 10147 if (!hba) {
10148 pr_err("WARNING - Cannot detect HBA for device %s!\n",
10149 devid2kname(makedev(info->disk.major, info->disk.minor)));
10150 return 1;
10151 }
10152
10153 const struct imsm_orom *orom = get_orom_by_device_id(hba->dev_id);
10154 struct mdinfo *dev;
10155
10156 for (dev = info->next; dev; dev = dev->next) {
10157 dev_path = devt_to_devpath(makedev(dev->disk.major, dev->disk.minor));
10158
10159 struct sys_dev *hba2 = NULL;
10160 for (idev = intel_devices; idev; idev = idev->next) {
10161 if (dev_path && strstr(dev_path, idev->path)) {
10162 hba2 = idev;
10163 break;
d31ad643
PB		 10164 }
10165 }
6b781d33
AP		 10166 if (dev_path)
10167 free(dev_path);
10168
10169 const struct imsm_orom *orom2 = hba2 == NULL ? NULL :
10170 get_orom_by_device_id(hba2->dev_id);
10171
10172 if (hba2 && hba->type != hba2->type) {
10173 pr_err("WARNING - HBAs of devices do not match %s != %s\n",
10174 get_sys_dev_type(hba->type), get_sys_dev_type(hba2->type));
10175 return 1;
10176 }
10177
07cb1e57 10178 if (orom != orom2) {
6b781d33
AP		 10179 pr_err("WARNING - IMSM container assembled with disks under different HBAs!\n"
10180 " This operation is not supported and can lead to data loss.\n");
10181 return 1;
10182 }
10183
10184 if (!orom) {
10185 pr_err("WARNING - IMSM container assembled with disks under HBAs without IMSM platform support!\n"
10186 " This operation is not supported and can lead to data loss.\n");
10187 return 1;
10188 }
d31ad643 10189 }
6b781d33 10190
d31ad643
PB		 10191 return 0;
10192}
9e2d750d 10193#ifndef MDASSEMBLE
6f50473f
TM		 10194/*******************************************************************************
10195* Function: imsm_record_badblock
10196* Description: This routine stores new bad block record in BBM log
10197*
10198* Parameters:
10199* a : array containing a bad block
10200* slot : disk number containing a bad block
10201* sector : bad block sector
10202* length : bad block sectors range
10203* Returns:
10204* 1 : Success
10205* 0 : Error
10206******************************************************************************/
10207static int imsm_record_badblock(struct active_array *a, int slot,
10208 unsigned long long sector, int length)
10209{
10210 struct intel_super *super = a->container->sb;
10211 int ord;
10212 int ret;
10213
10214 ord = imsm_disk_slot_to_ord(a, slot);
10215 if (ord < 0)
10216 return 0;
10217
10218 ret = record_new_badblock(super->bbm_log, ord_to_idx(ord), sector,
10219 length);
10220 if (ret)
10221 super->updates_pending++;
10222
10223 return ret;
10224}
c07a5a4f
TM		 10225/*******************************************************************************
10226* Function: imsm_clear_badblock
10227* Description: This routine clears bad block record from BBM log
10228*
10229* Parameters:
10230* a : array containing a bad block
10231* slot : disk number containing a bad block
10232* sector : bad block sector
10233* length : bad block sectors range
10234* Returns:
10235* 1 : Success
10236* 0 : Error
10237******************************************************************************/
10238static int imsm_clear_badblock(struct active_array *a, int slot,
10239 unsigned long long sector, int length)
10240{
10241 struct intel_super *super = a->container->sb;
10242 int ord;
10243 int ret;
10244
10245 ord = imsm_disk_slot_to_ord(a, slot);
10246 if (ord < 0)
10247 return 0;
10248
10249 ret = clear_badblock(super->bbm_log, ord_to_idx(ord), sector, length);
10250 if (ret)
10251 super->updates_pending++;
10252
10253 return ret;
10254}
928f1424
TM		 10255/*******************************************************************************
10256* Function: imsm_get_badblocks
10257* Description: This routine get list of bad blocks for an array
10258*
10259* Parameters:
10260* a : array
10261* slot : disk number
10262* Returns:
10263* bb : structure containing bad blocks
10264* NULL : error
10265******************************************************************************/
10266static struct md_bb *imsm_get_badblocks(struct active_array *a, int slot)
10267{
10268 int inst = a->info.container_member;
10269 struct intel_super *super = a->container->sb;
10270 struct imsm_dev *dev = get_imsm_dev(super, inst);
10271 struct imsm_map *map = get_imsm_map(dev, MAP_0);
10272 int ord;
10273
10274 ord = imsm_disk_slot_to_ord(a, slot);
10275 if (ord < 0)
10276 return NULL;
10277
10278 get_volume_badblocks(super->bbm_log, ord_to_idx(ord), pba_of_lba0(map),
10279 blocks_per_member(map), &super->bb);
10280
10281 return &super->bb;
10282}
27156a57
TM		 10283/*******************************************************************************
10284* Function: examine_badblocks_imsm
10285* Description: Prints list of bad blocks on a disk to the standard output
10286*
10287* Parameters:
10288* st : metadata handler
10289* fd : open file descriptor for device
10290* devname : device name
10291* Returns:
10292* 0 : Success
10293* 1 : Error
10294******************************************************************************/
10295static int examine_badblocks_imsm(struct supertype *st, int fd, char *devname)
10296{
10297 struct intel_super *super = st->sb;
10298 struct bbm_log *log = super->bbm_log;
10299 struct dl *d = NULL;
10300 int any = 0;
10301
10302 for (d = super->disks; d ; d = d->next) {
10303 if (strcmp(d->devname, devname) == 0)
10304 break;
10305 }
10306
10307 if ((d == NULL) || (d->index < 0)) { /* serial mismatch probably */
10308 pr_err("%s doesn't appear to be part of a raid array\n",
10309 devname);
10310 return 1;
10311 }
10312
10313 if (log != NULL) {
10314 unsigned int i;
10315 struct bbm_log_entry *entry = &log->marked_block_entries[0];
10316
10317 for (i = 0; i < log->entry_count; i++) {
10318 if (entry[i].disk_ordinal == d->index) {
10319 unsigned long long sector = __le48_to_cpu(
10320 &entry[i].defective_block_start);
10321 int cnt = entry[i].marked_count + 1;
10322
10323 if (!any) {
10324 printf("Bad-blocks on %s:\n", devname);
10325 any = 1;
10326 }
10327
10328 printf("%20llu for %d sectors\n", sector, cnt);
10329 }
10330 }
10331 }
10332
10333 if (!any)
10334 printf("No bad-blocks list configured on %s\n", devname);
10335
10336 return 0;
10337}
687629c2
AK		 10338/*******************************************************************************
10339 * Function: init_migr_record_imsm
10340 * Description: Function inits imsm migration record
10341 * Parameters:
10342 * super : imsm internal array info
10343 * dev : device under migration
10344 * info : general array info to find the smallest device
10345 * Returns:
10346 * none
10347 ******************************************************************************/
10348void init_migr_record_imsm(struct supertype *st, struct imsm_dev *dev,
10349 struct mdinfo *info)
10350{
10351 struct intel_super *super = st->sb;
10352 struct migr_record *migr_rec = super->migr_rec;
10353 int new_data_disks;
10354 unsigned long long dsize, dev_sectors;
10355 long long unsigned min_dev_sectors = -1LLU;
10356 struct mdinfo *sd;
10357 char nm[30];
10358 int fd;
238c0a71
AK		 10359 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
10360 struct imsm_map *map_src = get_imsm_map(dev, MAP_1);
687629c2 10361 unsigned long long num_migr_units;
3ef4403c 10362 unsigned long long array_blocks;
687629c2
AK		 10363
10364 memset(migr_rec, 0, sizeof(struct migr_record));
10365 migr_rec->family_num = __cpu_to_le32(super->anchor->family_num);
10366
10367 /* only ascending reshape supported now */
10368 migr_rec->ascending_migr = __cpu_to_le32(1);
10369
10370 migr_rec->dest_depth_per_unit = GEN_MIGR_AREA_SIZE /
10371 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
e1742195
AK		 10372 migr_rec->dest_depth_per_unit *=
10373 max(map_dest->blocks_per_strip, map_src->blocks_per_strip);
238c0a71 10374 new_data_disks = imsm_num_data_members(dev, MAP_0);
687629c2
AK		 10375 migr_rec->blocks_per_unit =
10376 __cpu_to_le32(migr_rec->dest_depth_per_unit * new_data_disks);
10377 migr_rec->dest_depth_per_unit =
10378 __cpu_to_le32(migr_rec->dest_depth_per_unit);
3ef4403c 10379 array_blocks = info->component_size * new_data_disks;
687629c2
AK		 10380 num_migr_units =
10381 array_blocks / __le32_to_cpu(migr_rec->blocks_per_unit);
10382
10383 if (array_blocks % __le32_to_cpu(migr_rec->blocks_per_unit))
10384 num_migr_units++;
10385 migr_rec->num_migr_units = __cpu_to_le32(num_migr_units);
10386
10387 migr_rec->post_migr_vol_cap = dev->size_low;
10388 migr_rec->post_migr_vol_cap_hi = dev->size_high;
10389
687629c2
AK		 10390 /* Find the smallest dev */
10391 for (sd = info->devs ; sd ; sd = sd->next) {
10392 sprintf(nm, "%d:%d", sd->disk.major, sd->disk.minor);
10393 fd = dev_open(nm, O_RDONLY);
10394 if (fd < 0)
10395 continue;
10396 get_dev_size(fd, NULL, &dsize);
10397 dev_sectors = dsize / 512;
10398 if (dev_sectors < min_dev_sectors)
10399 min_dev_sectors = dev_sectors;
10400 close(fd);
10401 }
10402 migr_rec->ckpt_area_pba = __cpu_to_le32(min_dev_sectors -
10403 RAID_DISK_RESERVED_BLOCKS_IMSM_HI);
10404
10405 write_imsm_migr_rec(st);
10406
10407 return;
10408}
10409
10410/*******************************************************************************
10411 * Function: save_backup_imsm
10412 * Description: Function saves critical data stripes to Migration Copy Area
10413 * and updates the current migration unit status.
10414 * Use restore_stripes() to form a destination stripe,
10415 * and to write it to the Copy Area.
10416 * Parameters:
10417 * st : supertype information
aea93171 10418 * dev : imsm device that backup is saved for
687629c2
AK		 10419 * info : general array info
10420 * buf : input buffer
687629c2
AK		 10421 * length : length of data to backup (blocks_per_unit)
10422 * Returns:
10423 * 0 : success
10424 *, -1 : fail
10425 ******************************************************************************/
10426int save_backup_imsm(struct supertype *st,
10427 struct imsm_dev *dev,
10428 struct mdinfo *info,
10429 void *buf,
687629c2
AK		 10430 int length)
10431{
10432 int rv = -1;
10433 struct intel_super *super = st->sb;
594dc1b8
JS		 10434 unsigned long long *target_offsets;
10435 int *targets;
687629c2 10436 int i;
238c0a71 10437 struct imsm_map *map_dest = get_imsm_map(dev, MAP_0);
687629c2 10438 int new_disks = map_dest->num_members;
ab724b98
AK		 10439 int dest_layout = 0;
10440 int dest_chunk;
d1877f69 10441 unsigned long long start;
238c0a71 10442 int data_disks = imsm_num_data_members(dev, MAP_0);
687629c2 10443
503975b9 10444 targets = xmalloc(new_disks * sizeof(int));
687629c2 10445
7e45b550
AK		 10446 for (i = 0; i < new_disks; i++)
10447 targets[i] = -1;
10448
503975b9 10449 target_offsets = xcalloc(new_disks, sizeof(unsigned long long));
687629c2 10450
d1877f69 10451 start = info->reshape_progress * 512;
687629c2 10452 for (i = 0; i < new_disks; i++) {
687629c2
AK		 10453 target_offsets[i] = (unsigned long long)
10454 __le32_to_cpu(super->migr_rec->ckpt_area_pba) * 512;
d1877f69
AK		 10455 /* move back copy area adderss, it will be moved forward
10456 * in restore_stripes() using start input variable
10457 */
10458 target_offsets[i] -= start/data_disks;
687629c2
AK		 10459 }
10460
9a717282
AK		 10461 if (open_backup_targets(info, new_disks, targets,
10462 super, dev))
687629c2
AK		 10463 goto abort;
10464
68eb8bc6 10465 dest_layout = imsm_level_to_layout(map_dest->raid_level);
ab724b98
AK		 10466 dest_chunk = __le16_to_cpu(map_dest->blocks_per_strip) * 512;
10467
687629c2
AK		 10468 if (restore_stripes(targets, /* list of dest devices */
10469 target_offsets, /* migration record offsets */
10470 new_disks,
ab724b98
AK		 10471 dest_chunk,
10472 map_dest->raid_level,
10473 dest_layout,
10474 -1, /* source backup file descriptor */
10475 0, /* input buf offset
10476 * always 0 buf is already offseted */
d1877f69 10477 start,
687629c2
AK		 10478 length,
10479 buf) != 0) {
e7b84f9d 10480 pr_err("Error restoring stripes\n");
687629c2
AK		 10481 goto abort;
10482 }
10483
10484 rv = 0;
10485
10486abort:
10487 if (targets) {
9a717282 10488 close_targets(targets, new_disks);
687629c2
AK		 10489 free(targets);
10490 }
10491 free(target_offsets);
10492
10493 return rv;
10494}
10495
10496/*******************************************************************************
10497 * Function: save_checkpoint_imsm
10498 * Description: Function called for current unit status update
10499 * in the migration record. It writes it to disk.
10500 * Parameters:
10501 * super : imsm internal array info
10502 * info : general array info
10503 * Returns:
10504 * 0: success
10505 * 1: failure
0228d92c
AK		 10506 * 2: failure, means no valid migration record
10507 * / no general migration in progress /
687629c2
AK		 10508 ******************************************************************************/
10509int save_checkpoint_imsm(struct supertype *st, struct mdinfo *info, int state)
10510{
10511 struct intel_super *super = st->sb;
f8b72ef5
AK		 10512 unsigned long long blocks_per_unit;
10513 unsigned long long curr_migr_unit;
10514
2e062e82 10515 if (load_imsm_migr_rec(super, info) != 0) {
7a862a02 10516 dprintf("imsm: ERROR: Cannot read migration record for checkpoint save.\n");
2e062e82
AK		 10517 return 1;
10518 }
10519
f8b72ef5
AK		 10520 blocks_per_unit = __le32_to_cpu(super->migr_rec->blocks_per_unit);
10521 if (blocks_per_unit == 0) {
0228d92c
AK		 10522 dprintf("imsm: no migration in progress.\n");
10523 return 2;
687629c2 10524 }
f8b72ef5
AK		 10525 curr_migr_unit = info->reshape_progress / blocks_per_unit;
10526 /* check if array is alligned to copy area
10527 * if it is not alligned, add one to current migration unit value
10528 * this can happend on array reshape finish only
10529 */
10530 if (info->reshape_progress % blocks_per_unit)
10531 curr_migr_unit++;
687629c2
AK		 10532
10533 super->migr_rec->curr_migr_unit =
f8b72ef5 10534 __cpu_to_le32(curr_migr_unit);
687629c2
AK		 10535 super->migr_rec->rec_status = __cpu_to_le32(state);
10536 super->migr_rec->dest_1st_member_lba =
f8b72ef5
AK		 10537 __cpu_to_le32(curr_migr_unit *
10538 __le32_to_cpu(super->migr_rec->dest_depth_per_unit));
687629c2 10539 if (write_imsm_migr_rec(st) < 0) {
7a862a02 10540 dprintf("imsm: Cannot write migration record outside backup area\n");
687629c2
AK		 10541 return 1;
10542 }
10543
10544 return 0;
10545}
10546
276d77db
AK		 10547/*******************************************************************************
10548 * Function: recover_backup_imsm
10549 * Description: Function recovers critical data from the Migration Copy Area
10550 * while assembling an array.
10551 * Parameters:
10552 * super : imsm internal array info
10553 * info : general array info
10554 * Returns:
10555 * 0 : success (or there is no data to recover)
10556 * 1 : fail
10557 ******************************************************************************/
10558int recover_backup_imsm(struct supertype *st, struct mdinfo *info)
10559{
10560 struct intel_super *super = st->sb;
10561 struct migr_record *migr_rec = super->migr_rec;
594dc1b8 10562 struct imsm_map *map_dest;
276d77db
AK		 10563 struct intel_dev *id = NULL;
10564 unsigned long long read_offset;
10565 unsigned long long write_offset;
10566 unsigned unit_len;
10567 int *targets = NULL;
10568 int new_disks, i, err;
10569 char *buf = NULL;
10570 int retval = 1;
f36a9ecd 10571 unsigned int sector_size = super->sector_size;
276d77db
AK		 10572 unsigned long curr_migr_unit = __le32_to_cpu(migr_rec->curr_migr_unit);
10573 unsigned long num_migr_units = __le32_to_cpu(migr_rec->num_migr_units);
276d77db 10574 char buffer[20];
6c3560c0 10575 int skipped_disks = 0;
276d77db
AK		 10576
10577 err = sysfs_get_str(info, NULL, "array_state", (char *)buffer, 20);
10578 if (err < 1)
10579 return 1;
10580
10581 /* recover data only during assemblation */
10582 if (strncmp(buffer, "inactive", 8) != 0)
10583 return 0;
10584 /* no data to recover */
10585 if (__le32_to_cpu(migr_rec->rec_status) == UNIT_SRC_NORMAL)
10586 return 0;
10587 if (curr_migr_unit >= num_migr_units)
10588 return 1;
10589
10590 /* find device during reshape */
10591 for (id = super->devlist; id; id = id->next)
10592 if (is_gen_migration(id->dev))
10593 break;
10594 if (id == NULL)
10595 return 1;
10596
238c0a71 10597 map_dest = get_imsm_map(id->dev, MAP_0);
276d77db
AK		 10598 new_disks = map_dest->num_members;
10599
10600 read_offset = (unsigned long long)
10601 __le32_to_cpu(migr_rec->ckpt_area_pba) * 512;
10602
10603 write_offset = ((unsigned long long)
10604 __le32_to_cpu(migr_rec->dest_1st_member_lba) +
5551b113 10605 pba_of_lba0(map_dest)) * 512;
276d77db
AK		 10606
10607 unit_len = __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
f36a9ecd 10608 if (posix_memalign((void **)&buf, sector_size, unit_len) != 0)
276d77db 10609 goto abort;
503975b9 10610 targets = xcalloc(new_disks, sizeof(int));
276d77db 10611
9a717282 10612 if (open_backup_targets(info, new_disks, targets, super, id->dev)) {
e7b84f9d 10613 pr_err("Cannot open some devices belonging to array.\n");
f627f5ad
AK		 10614 goto abort;
10615 }
276d77db
AK		 10616
10617 for (i = 0; i < new_disks; i++) {
6c3560c0
AK		 10618 if (targets[i] < 0) {
10619 skipped_disks++;
10620 continue;
10621 }
276d77db 10622 if (lseek64(targets[i], read_offset, SEEK_SET) < 0) {
e7b84f9d
N		 10623 pr_err("Cannot seek to block: %s\n",
10624 strerror(errno));
137debce
AK		 10625 skipped_disks++;
10626 continue;
276d77db 10627 }
9ec11d1a 10628 if ((unsigned)read(targets[i], buf, unit_len) != unit_len) {
e7b84f9d
N		 10629 pr_err("Cannot read copy area block: %s\n",
10630 strerror(errno));
137debce
AK		 10631 skipped_disks++;
10632 continue;
276d77db
AK		 10633 }
10634 if (lseek64(targets[i], write_offset, SEEK_SET) < 0) {
e7b84f9d
N		 10635 pr_err("Cannot seek to block: %s\n",
10636 strerror(errno));
137debce
AK		 10637 skipped_disks++;
10638 continue;
276d77db 10639 }
9ec11d1a 10640 if ((unsigned)write(targets[i], buf, unit_len) != unit_len) {
e7b84f9d
N		 10641 pr_err("Cannot restore block: %s\n",
10642 strerror(errno));
137debce
AK		 10643 skipped_disks++;
10644 continue;
276d77db
AK		 10645 }
10646 }
10647
137debce
AK		 10648 if (skipped_disks > imsm_get_allowed_degradation(info->new_level,
10649 new_disks,
10650 super,
10651 id->dev)) {
7a862a02 10652 pr_err("Cannot restore data from backup. Too many failed disks\n");
6c3560c0
AK		 10653 goto abort;
10654 }
10655
befb629b
AK		 10656 if (save_checkpoint_imsm(st, info, UNIT_SRC_NORMAL)) {
10657 /* ignore error == 2, this can mean end of reshape here
10658 */
7a862a02 10659 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL) during restart\n");
befb629b 10660 } else
276d77db 10661 retval = 0;
276d77db
AK		 10662
10663abort:
10664 if (targets) {
10665 for (i = 0; i < new_disks; i++)
10666 if (targets[i])
10667 close(targets[i]);
10668 free(targets);
10669 }
10670 free(buf);
10671 return retval;
10672}
10673
2cda7640
ML		 10674static char disk_by_path[] = "/dev/disk/by-path/";
10675
10676static const char *imsm_get_disk_controller_domain(const char *path)
10677{
2cda7640 10678 char disk_path[PATH_MAX];
96234762
LM		 10679 char *drv=NULL;
10680 struct stat st;
2cda7640 10681
6d8d290a 10682 strcpy(disk_path, disk_by_path);
96234762
LM		 10683 strncat(disk_path, path, PATH_MAX - strlen(disk_path) - 1);
10684 if (stat(disk_path, &st) == 0) {
10685 struct sys_dev* hba;
594dc1b8 10686 char *path;
96234762
LM		 10687
10688 path = devt_to_devpath(st.st_rdev);
10689 if (path == NULL)
10690 return "unknown";
10691 hba = find_disk_attached_hba(-1, path);
10692 if (hba && hba->type == SYS_DEV_SAS)
10693 drv = "isci";
10694 else if (hba && hba->type == SYS_DEV_SATA)
10695 drv = "ahci";
1011e834 10696 else
96234762
LM		 10697 drv = "unknown";
10698 dprintf("path: %s hba: %s attached: %s\n",
10699 path, (hba) ? hba->path : "NULL", drv);
10700 free(path);
2cda7640 10701 }
96234762 10702 return drv;
2cda7640
ML		 10703}
10704
4dd2df09 10705static char *imsm_find_array_devnm_by_subdev(int subdev, char *container)
78b10e66 10706{
4dd2df09 10707 static char devnm[32];
78b10e66
N		 10708 char subdev_name[20];
10709 struct mdstat_ent *mdstat;
10710
10711 sprintf(subdev_name, "%d", subdev);
10712 mdstat = mdstat_by_subdev(subdev_name, container);
10713 if (!mdstat)
4dd2df09 10714 return NULL;
78b10e66 10715
4dd2df09 10716 strcpy(devnm, mdstat->devnm);
78b10e66 10717 free_mdstat(mdstat);
4dd2df09 10718 return devnm;
78b10e66
N		 10719}
10720
10721static int imsm_reshape_is_allowed_on_container(struct supertype *st,
10722 struct geo_params *geo,
fbf3d202
AK		 10723 int *old_raid_disks,
10724 int direction)
78b10e66 10725{
694575e7
KW		 10726 /* currently we only support increasing the number of devices
10727 * for a container. This increases the number of device for each
10728 * member array. They must all be RAID0 or RAID5.
10729 */
78b10e66
N		 10730 int ret_val = 0;
10731 struct mdinfo *info, *member;
10732 int devices_that_can_grow = 0;
10733
7a862a02 10734 dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): st->devnm = (%s)\n", st->devnm);
78b10e66 10735
d04f65f4 10736 if (geo->size > 0 ||
78b10e66
N		 10737 geo->level != UnSet ||
10738 geo->layout != UnSet ||
10739 geo->chunksize != 0 ||
10740 geo->raid_disks == UnSet) {
7a862a02 10741 dprintf("imsm: Container operation is allowed for raid disks number change only.\n");
78b10e66
N		 10742 return ret_val;
10743 }
10744
fbf3d202 10745 if (direction == ROLLBACK_METADATA_CHANGES) {
7a862a02 10746 dprintf("imsm: Metadata changes rollback is not supported for container operation.\n");
fbf3d202
AK		 10747 return ret_val;
10748 }
10749
78b10e66
N		 10750 info = container_content_imsm(st, NULL);
10751 for (member = info; member; member = member->next) {
4dd2df09 10752 char *result;
78b10e66
N		 10753
10754 dprintf("imsm: checking device_num: %i\n",
10755 member->container_member);
10756
d7d205bd 10757 if (geo->raid_disks <= member->array.raid_disks) {
78b10e66
N		 10758 /* we work on container for Online Capacity Expansion
10759 * only so raid_disks has to grow
10760 */
7a862a02 10761 dprintf("imsm: for container operation raid disks increase is required\n");
78b10e66
N		 10762 break;
10763 }
10764
089f9d79 10765 if (info->array.level != 0 && info->array.level != 5) {
78b10e66
N		 10766 /* we cannot use this container with other raid level
10767 */
7a862a02 10768 dprintf("imsm: for container operation wrong raid level (%i) detected\n",
78b10e66
N		 10769 info->array.level);
10770 break;
10771 } else {
10772 /* check for platform support
10773 * for this raid level configuration
10774 */
10775 struct intel_super *super = st->sb;
10776 if (!is_raid_level_supported(super->orom,
10777 member->array.level,
10778 geo->raid_disks)) {
7a862a02 10779 dprintf("platform does not support raid%d with %d disk%s\n",
78b10e66
N		 10780 info->array.level,
10781 geo->raid_disks,
10782 geo->raid_disks > 1 ? "s" : "");
10783 break;
10784 }
2a4a08e7
AK		 10785 /* check if component size is aligned to chunk size
10786 */
10787 if (info->component_size %
10788 (info->array.chunk_size/512)) {
7a862a02 10789 dprintf("Component size is not aligned to chunk size\n");
2a4a08e7
AK		 10790 break;
10791 }
78b10e66
N		 10792 }
10793
10794 if (*old_raid_disks &&
10795 info->array.raid_disks != *old_raid_disks)
10796 break;
10797 *old_raid_disks = info->array.raid_disks;
10798
10799 /* All raid5 and raid0 volumes in container
10800 * have to be ready for Online Capacity Expansion
10801 * so they need to be assembled. We have already
10802 * checked that no recovery etc is happening.
10803 */
4dd2df09
N		 10804 result = imsm_find_array_devnm_by_subdev(member->container_member,
10805 st->container_devnm);
10806 if (result == NULL) {
78b10e66
N		 10807 dprintf("imsm: cannot find array\n");
10808 break;
10809 }
10810 devices_that_can_grow++;
10811 }
10812 sysfs_free(info);
10813 if (!member && devices_that_can_grow)
10814 ret_val = 1;
10815
10816 if (ret_val)
1ade5cc1 10817 dprintf("Container operation allowed\n");
78b10e66 10818 else
1ade5cc1 10819 dprintf("Error: %i\n", ret_val);
78b10e66
N		 10820
10821 return ret_val;
10822}
10823
10824/* Function: get_spares_for_grow
10825 * Description: Allocates memory and creates list of spare devices
1011e834 10826 * avaliable in container. Checks if spare drive size is acceptable.
78b10e66
N		 10827 * Parameters: Pointer to the supertype structure
10828 * Returns: Pointer to the list of spare devices (mdinfo structure) on success,
1011e834 10829 * NULL if fail
78b10e66
N		 10830 */
10831static struct mdinfo *get_spares_for_grow(struct supertype *st)
10832{
78b10e66 10833 unsigned long long min_size = min_acceptable_spare_size_imsm(st);
326727d9 10834 return container_choose_spares(st, min_size, NULL, NULL, NULL, 0);
78b10e66
N		 10835}
10836
10837/******************************************************************************
10838 * function: imsm_create_metadata_update_for_reshape
10839 * Function creates update for whole IMSM container.
10840 *
10841 ******************************************************************************/
10842static int imsm_create_metadata_update_for_reshape(
10843 struct supertype *st,
10844 struct geo_params *geo,
10845 int old_raid_disks,
10846 struct imsm_update_reshape **updatep)
10847{
10848 struct intel_super *super = st->sb;
10849 struct imsm_super *mpb = super->anchor;
594dc1b8
JS		 10850 int update_memory_size;
10851 struct imsm_update_reshape *u;
10852 struct mdinfo *spares;
78b10e66 10853 int i;
594dc1b8 10854 int delta_disks;
bbd24d86 10855 struct mdinfo *dev;
78b10e66 10856
1ade5cc1 10857 dprintf("(enter) raid_disks = %i\n", geo->raid_disks);
78b10e66
N		 10858
10859 delta_disks = geo->raid_disks - old_raid_disks;
10860
10861 /* size of all update data without anchor */
10862 update_memory_size = sizeof(struct imsm_update_reshape);
10863
10864 /* now add space for spare disks that we need to add. */
10865 update_memory_size += sizeof(u->new_disks[0]) * (delta_disks - 1);
10866
503975b9 10867 u = xcalloc(1, update_memory_size);
78b10e66
N		 10868 u->type = update_reshape_container_disks;
10869 u->old_raid_disks = old_raid_disks;
10870 u->new_raid_disks = geo->raid_disks;
10871
10872 /* now get spare disks list
10873 */
10874 spares = get_spares_for_grow(st);
10875
10876 if (spares == NULL
10877 || delta_disks > spares->array.spare_disks) {
7a862a02 10878 pr_err("imsm: ERROR: Cannot get spare devices for %s.\n", geo->dev_name);
e4c72d1d 10879 i = -1;
78b10e66
N		 10880 goto abort;
10881 }
10882
10883 /* we have got spares
10884 * update disk list in imsm_disk list table in anchor
10885 */
10886 dprintf("imsm: %i spares are available.\n\n",
10887 spares->array.spare_disks);
10888
bbd24d86 10889 dev = spares->devs;
78b10e66 10890 for (i = 0; i < delta_disks; i++) {
78b10e66
N		 10891 struct dl *dl;
10892
bbd24d86
AK		 10893 if (dev == NULL)
10894 break;
78b10e66
N		 10895 u->new_disks[i] = makedev(dev->disk.major,
10896 dev->disk.minor);
10897 dl = get_disk_super(super, dev->disk.major, dev->disk.minor);
ee4beede
AK		 10898 dl->index = mpb->num_disks;
10899 mpb->num_disks++;
bbd24d86 10900 dev = dev->next;
78b10e66 10901 }
78b10e66
N		 10902
10903abort:
10904 /* free spares
10905 */
10906 sysfs_free(spares);
10907
d677e0b8 10908 dprintf("imsm: reshape update preparation :");
78b10e66 10909 if (i == delta_disks) {
1ade5cc1 10910 dprintf_cont(" OK\n");
78b10e66
N		 10911 *updatep = u;
10912 return update_memory_size;
10913 }
10914 free(u);
1ade5cc1 10915 dprintf_cont(" Error\n");
78b10e66
N		 10916
10917 return 0;
10918}
10919
f3871fdc
AK		 10920/******************************************************************************
10921 * function: imsm_create_metadata_update_for_size_change()
10922 * Creates update for IMSM array for array size change.
10923 *
10924 ******************************************************************************/
10925static int imsm_create_metadata_update_for_size_change(
10926 struct supertype *st,
10927 struct geo_params *geo,
10928 struct imsm_update_size_change **updatep)
10929{
10930 struct intel_super *super = st->sb;
594dc1b8
JS		 10931 int update_memory_size;
10932 struct imsm_update_size_change *u;
f3871fdc 10933
1ade5cc1 10934 dprintf("(enter) New size = %llu\n", geo->size);
f3871fdc
AK		 10935
10936 /* size of all update data without anchor */
10937 update_memory_size = sizeof(struct imsm_update_size_change);
10938
503975b9 10939 u = xcalloc(1, update_memory_size);
f3871fdc
AK		 10940 u->type = update_size_change;
10941 u->subdev = super->current_vol;
10942 u->new_size = geo->size;
10943
10944 dprintf("imsm: reshape update preparation : OK\n");
10945 *updatep = u;
10946
10947 return update_memory_size;
10948}
10949
48c5303a
PC		 10950/******************************************************************************
10951 * function: imsm_create_metadata_update_for_migration()
10952 * Creates update for IMSM array.
10953 *
10954 ******************************************************************************/
10955static int imsm_create_metadata_update_for_migration(
10956 struct supertype *st,
10957 struct geo_params *geo,
10958 struct imsm_update_reshape_migration **updatep)
10959{
10960 struct intel_super *super = st->sb;
594dc1b8
JS		 10961 int update_memory_size;
10962 struct imsm_update_reshape_migration *u;
48c5303a
PC		 10963 struct imsm_dev *dev;
10964 int previous_level = -1;
10965
1ade5cc1 10966 dprintf("(enter) New Level = %i\n", geo->level);
48c5303a
PC		 10967
10968 /* size of all update data without anchor */
10969 update_memory_size = sizeof(struct imsm_update_reshape_migration);
10970
503975b9 10971 u = xcalloc(1, update_memory_size);
48c5303a
PC		 10972 u->type = update_reshape_migration;
10973 u->subdev = super->current_vol;
10974 u->new_level = geo->level;
10975 u->new_layout = geo->layout;
10976 u->new_raid_disks = u->old_raid_disks = geo->raid_disks;
10977 u->new_disks[0] = -1;
4bba0439 10978 u->new_chunksize = -1;
48c5303a
PC		 10979
10980 dev = get_imsm_dev(super, u->subdev);
10981 if (dev) {
10982 struct imsm_map *map;
10983
238c0a71 10984 map = get_imsm_map(dev, MAP_0);
4bba0439
PC		 10985 if (map) {
10986 int current_chunk_size =
10987 __le16_to_cpu(map->blocks_per_strip) / 2;
10988
10989 if (geo->chunksize != current_chunk_size) {
10990 u->new_chunksize = geo->chunksize / 1024;
7a862a02 10991 dprintf("imsm: chunk size change from %i to %i\n",
4bba0439
PC		 10992 current_chunk_size, u->new_chunksize);
10993 }
48c5303a 10994 previous_level = map->raid_level;
4bba0439 10995 }
48c5303a 10996 }
089f9d79 10997 if (geo->level == 5 && previous_level == 0) {
48c5303a
PC		 10998 struct mdinfo *spares = NULL;
10999
11000 u->new_raid_disks++;
11001 spares = get_spares_for_grow(st);
089f9d79 11002 if (spares == NULL || spares->array.spare_disks < 1) {
48c5303a
PC		 11003 free(u);
11004 sysfs_free(spares);
11005 update_memory_size = 0;
565cc99e 11006 pr_err("cannot get spare device for requested migration\n");
48c5303a
PC		 11007 return 0;
11008 }
11009 sysfs_free(spares);
11010 }
11011 dprintf("imsm: reshape update preparation : OK\n");
11012 *updatep = u;
11013
11014 return update_memory_size;
11015}
11016
8dd70bce
AK		 11017static void imsm_update_metadata_locally(struct supertype *st,
11018 void *buf, int len)
11019{
11020 struct metadata_update mu;
11021
11022 mu.buf = buf;
11023 mu.len = len;
11024 mu.space = NULL;
11025 mu.space_list = NULL;
11026 mu.next = NULL;
5fe6f031
N		 11027 if (imsm_prepare_update(st, &mu))
11028 imsm_process_update(st, &mu);
8dd70bce
AK		 11029
11030 while (mu.space_list) {
11031 void **space = mu.space_list;
11032 mu.space_list = *space;
11033 free(space);
11034 }
11035}
78b10e66 11036
471bceb6 11037/***************************************************************************
694575e7 11038* Function: imsm_analyze_change
471bceb6 11039* Description: Function analyze change for single volume
1011e834 11040* and validate if transition is supported
fbf3d202
AK		 11041* Parameters: Geometry parameters, supertype structure,
11042* metadata change direction (apply/rollback)
694575e7 11043* Returns: Operation type code on success, -1 if fail
471bceb6
KW		 11044****************************************************************************/
11045enum imsm_reshape_type imsm_analyze_change(struct supertype *st,
fbf3d202
AK		 11046 struct geo_params *geo,
11047 int direction)
694575e7 11048{
471bceb6
KW		 11049 struct mdinfo info;
11050 int change = -1;
11051 int check_devs = 0;
c21e737b 11052 int chunk;
67a2db32
AK		 11053 /* number of added/removed disks in operation result */
11054 int devNumChange = 0;
11055 /* imsm compatible layout value for array geometry verification */
11056 int imsm_layout = -1;
7abc9871
AK		 11057 int data_disks;
11058 struct imsm_dev *dev;
11059 struct intel_super *super;
d04f65f4 11060 unsigned long long current_size;
65d38cca 11061 unsigned long long free_size;
d04f65f4 11062 unsigned long long max_size;
65d38cca 11063 int rv;
471bceb6
KW		 11064
11065 getinfo_super_imsm_volume(st, &info, NULL);
089f9d79
JS		 11066 if (geo->level != info.array.level && geo->level >= 0 &&
11067 geo->level != UnSet) {
471bceb6
KW		 11068 switch (info.array.level) {
11069 case 0:
11070 if (geo->level == 5) {
b5347799 11071 change = CH_MIGRATION;
e13ce846 11072 if (geo->layout != ALGORITHM_LEFT_ASYMMETRIC) {
7a862a02 11073 pr_err("Error. Requested Layout not supported (left-asymmetric layout is supported only)!\n");
e13ce846
AK		 11074 change = -1;
11075 goto analyse_change_exit;
11076 }
67a2db32 11077 imsm_layout = geo->layout;
471bceb6 11078 check_devs = 1;
e91a3bad
LM		 11079 devNumChange = 1; /* parity disk added */
11080 } else if (geo->level == 10) {
471bceb6
KW		 11081 change = CH_TAKEOVER;
11082 check_devs = 1;
e91a3bad 11083 devNumChange = 2; /* two mirrors added */
67a2db32 11084 imsm_layout = 0x102; /* imsm supported layout */
471bceb6 11085 }
dfe77a9e
KW		 11086 break;
11087 case 1:
471bceb6
KW		 11088 case 10:
11089 if (geo->level == 0) {
11090 change = CH_TAKEOVER;
11091 check_devs = 1;
e91a3bad 11092 devNumChange = -(geo->raid_disks/2);
67a2db32 11093 imsm_layout = 0; /* imsm raid0 layout */
471bceb6
KW		 11094 }
11095 break;
11096 }
11097 if (change == -1) {
7a862a02 11098 pr_err("Error. Level Migration from %d to %d not supported!\n",
e7b84f9d 11099 info.array.level, geo->level);
471bceb6
KW		 11100 goto analyse_change_exit;
11101 }
11102 } else
11103 geo->level = info.array.level;
11104
089f9d79
JS		 11105 if (geo->layout != info.array.layout &&
11106 (geo->layout != UnSet && geo->layout != -1)) {
b5347799 11107 change = CH_MIGRATION;
089f9d79
JS		 11108 if (info.array.layout == 0 && info.array.level == 5 &&
11109 geo->layout == 5) {
471bceb6 11110 /* reshape 5 -> 4 */
089f9d79
JS		 11111 } else if (info.array.layout == 5 && info.array.level == 5 &&
11112 geo->layout == 0) {
471bceb6
KW		 11113 /* reshape 4 -> 5 */
11114 geo->layout = 0;
11115 geo->level = 5;
11116 } else {
7a862a02 11117 pr_err("Error. Layout Migration from %d to %d not supported!\n",
e7b84f9d 11118 info.array.layout, geo->layout);
471bceb6
KW		 11119 change = -1;
11120 goto analyse_change_exit;
11121 }
67a2db32 11122 } else {
471bceb6 11123 geo->layout = info.array.layout;
67a2db32
AK		 11124 if (imsm_layout == -1)
11125 imsm_layout = info.array.layout;
11126 }
471bceb6 11127
089f9d79
JS		 11128 if (geo->chunksize > 0 && geo->chunksize != UnSet &&
11129 geo->chunksize != info.array.chunk_size) {
2d2b0eb7
MD		 11130 if (info.array.level == 10) {
11131 pr_err("Error. Chunk size change for RAID 10 is not supported.\n");
11132 change = -1;
11133 goto analyse_change_exit;
1e9b2c3f
PB		 11134 } else if (info.component_size % (geo->chunksize/512)) {
11135 pr_err("New chunk size (%dK) does not evenly divide device size (%lluk). Aborting...\n",
11136 geo->chunksize/1024, info.component_size/2);
11137 change = -1;
11138 goto analyse_change_exit;
2d2b0eb7 11139 }
b5347799 11140 change = CH_MIGRATION;
2d2b0eb7 11141 } else {
471bceb6 11142 geo->chunksize = info.array.chunk_size;
2d2b0eb7 11143 }
471bceb6 11144
c21e737b 11145 chunk = geo->chunksize / 1024;
7abc9871
AK		 11146
11147 super = st->sb;
11148 dev = get_imsm_dev(super, super->current_vol);
11149 data_disks = imsm_num_data_members(dev , MAP_0);
c41e00b2 11150 /* compute current size per disk member
7abc9871 11151 */
c41e00b2
AK		 11152 current_size = info.custom_array_size / data_disks;
11153
089f9d79 11154 if (geo->size > 0 && geo->size != MAX_SIZE) {
c41e00b2
AK		 11155 /* align component size
11156 */
11157 geo->size = imsm_component_size_aligment_check(
11158 get_imsm_raid_level(dev->vol.map),
f36a9ecd 11159 chunk * 1024, super->sector_size,
c41e00b2 11160 geo->size * 2);
65d0b4ce 11161 if (geo->size == 0) {
7a862a02 11162 pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is 0).\n",
65d0b4ce
LD		 11163 current_size);
11164 goto analyse_change_exit;
11165 }
c41e00b2 11166 }
7abc9871 11167
089f9d79 11168 if (current_size != geo->size && geo->size > 0) {
7abc9871 11169 if (change != -1) {
7a862a02 11170 pr_err("Error. Size change should be the only one at a time.\n");
7abc9871
AK		 11171 change = -1;
11172 goto analyse_change_exit;
11173 }
11174 if ((super->current_vol + 1) != super->anchor->num_raid_devs) {
7a862a02 11175 pr_err("Error. The last volume in container can be expanded only (%i/%s).\n",
4dd2df09 11176 super->current_vol, st->devnm);
7abc9871
AK		 11177 goto analyse_change_exit;
11178 }
65d38cca
LD		 11179 /* check the maximum available size
11180 */
11181 rv = imsm_get_free_size(st, dev->vol.map->num_members,
11182 0, chunk, &free_size);
11183 if (rv == 0)
11184 /* Cannot find maximum available space
11185 */
11186 max_size = 0;
11187 else {
11188 max_size = free_size + current_size;
11189 /* align component size
11190 */
11191 max_size = imsm_component_size_aligment_check(
11192 get_imsm_raid_level(dev->vol.map),
f36a9ecd 11193 chunk * 1024, super->sector_size,
65d38cca
LD		 11194 max_size);
11195 }
d04f65f4 11196 if (geo->size == MAX_SIZE) {
b130333f
AK		 11197 /* requested size change to the maximum available size
11198 */
65d38cca 11199 if (max_size == 0) {
7a862a02 11200 pr_err("Error. Cannot find maximum available space.\n");
b130333f
AK		 11201 change = -1;
11202 goto analyse_change_exit;
65d38cca
LD		 11203 } else
11204 geo->size = max_size;
c41e00b2 11205 }
b130333f 11206
681b7ae2 11207 if (direction == ROLLBACK_METADATA_CHANGES) {
fbf3d202
AK		 11208 /* accept size for rollback only
11209 */
11210 } else {
11211 /* round size due to metadata compatibility
11212 */
11213 geo->size = (geo->size >> SECT_PER_MB_SHIFT)
11214 << SECT_PER_MB_SHIFT;
11215 dprintf("Prepare update for size change to %llu\n",
11216 geo->size );
11217 if (current_size >= geo->size) {
7a862a02 11218 pr_err("Error. Size expansion is supported only (current size is %llu, requested size /rounded/ is %llu).\n",
e7b84f9d 11219 current_size, geo->size);
fbf3d202
AK		 11220 goto analyse_change_exit;
11221 }
65d38cca 11222 if (max_size && geo->size > max_size) {
7a862a02 11223 pr_err("Error. Requested size is larger than maximum available size (maximum available size is %llu, requested size /rounded/ is %llu).\n",
e7b84f9d 11224 max_size, geo->size);
65d38cca
LD		 11225 goto analyse_change_exit;
11226 }
7abc9871
AK		 11227 }
11228 geo->size *= data_disks;
11229 geo->raid_disks = dev->vol.map->num_members;
11230 change = CH_ARRAY_SIZE;
11231 }
471bceb6
KW		 11232 if (!validate_geometry_imsm(st,
11233 geo->level,
67a2db32 11234 imsm_layout,
e91a3bad 11235 geo->raid_disks + devNumChange,
c21e737b 11236 &chunk,
af4348dd 11237 geo->size, INVALID_SECTORS,
5308f117 11238 0, 0, info.consistency_policy, 1))
471bceb6
KW		 11239 change = -1;
11240
11241 if (check_devs) {
11242 struct intel_super *super = st->sb;
11243 struct imsm_super *mpb = super->anchor;
11244
11245 if (mpb->num_raid_devs > 1) {
7a862a02 11246 pr_err("Error. Cannot perform operation on %s- for this operation it MUST be single array in container\n",
e7b84f9d 11247 geo->dev_name);
471bceb6
KW		 11248 change = -1;
11249 }
11250 }
11251
11252analyse_change_exit:
089f9d79
JS		 11253 if (direction == ROLLBACK_METADATA_CHANGES &&
11254 (change == CH_MIGRATION || change == CH_TAKEOVER)) {
7a862a02 11255 dprintf("imsm: Metadata changes rollback is not supported for migration and takeover operations.\n");
fbf3d202
AK		 11256 change = -1;
11257 }
471bceb6 11258 return change;
694575e7
KW		 11259}
11260
bb025c2f
KW		 11261int imsm_takeover(struct supertype *st, struct geo_params *geo)
11262{
11263 struct intel_super *super = st->sb;
11264 struct imsm_update_takeover *u;
11265
503975b9 11266 u = xmalloc(sizeof(struct imsm_update_takeover));
bb025c2f
KW		 11267
11268 u->type = update_takeover;
11269 u->subarray = super->current_vol;
11270
11271 /* 10->0 transition */
11272 if (geo->level == 0)
11273 u->direction = R10_TO_R0;
11274
0529c688
KW		 11275 /* 0->10 transition */
11276 if (geo->level == 10)
11277 u->direction = R0_TO_R10;
11278
bb025c2f
KW		 11279 /* update metadata locally */
11280 imsm_update_metadata_locally(st, u,
11281 sizeof(struct imsm_update_takeover));
11282 /* and possibly remotely */
11283 if (st->update_tail)
11284 append_metadata_update(st, u,
11285 sizeof(struct imsm_update_takeover));
11286 else
11287 free(u);
11288
11289 return 0;
11290}
11291
d04f65f4
N		 11292static int imsm_reshape_super(struct supertype *st, unsigned long long size,
11293 int level,
78b10e66 11294 int layout, int chunksize, int raid_disks,
41784c88 11295 int delta_disks, char *backup, char *dev,
016e00f5 11296 int direction, int verbose)
78b10e66 11297{
78b10e66
N		 11298 int ret_val = 1;
11299 struct geo_params geo;
11300
1ade5cc1 11301 dprintf("(enter)\n");
78b10e66 11302
71204a50 11303 memset(&geo, 0, sizeof(struct geo_params));
78b10e66
N		 11304
11305 geo.dev_name = dev;
4dd2df09 11306 strcpy(geo.devnm, st->devnm);
78b10e66
N		 11307 geo.size = size;
11308 geo.level = level;
11309 geo.layout = layout;
11310 geo.chunksize = chunksize;
11311 geo.raid_disks = raid_disks;
41784c88
AK		 11312 if (delta_disks != UnSet)
11313 geo.raid_disks += delta_disks;
78b10e66 11314
1ade5cc1
N		 11315 dprintf("for level : %i\n", geo.level);
11316 dprintf("for raid_disks : %i\n", geo.raid_disks);
78b10e66
N		 11317
11318 if (experimental() == 0)
11319 return ret_val;
11320
4dd2df09 11321 if (strcmp(st->container_devnm, st->devnm) == 0) {
694575e7
KW		 11322 /* On container level we can only increase number of devices. */
11323 dprintf("imsm: info: Container operation\n");
78b10e66 11324 int old_raid_disks = 0;
6dc0be30 11325
78b10e66 11326 if (imsm_reshape_is_allowed_on_container(
fbf3d202 11327 st, &geo, &old_raid_disks, direction)) {
78b10e66
N		 11328 struct imsm_update_reshape *u = NULL;
11329 int len;
11330
11331 len = imsm_create_metadata_update_for_reshape(
11332 st, &geo, old_raid_disks, &u);
11333
ed08d51c
AK		 11334 if (len <= 0) {
11335 dprintf("imsm: Cannot prepare update\n");
11336 goto exit_imsm_reshape_super;
11337 }
11338
8dd70bce
AK		 11339 ret_val = 0;
11340 /* update metadata locally */
11341 imsm_update_metadata_locally(st, u, len);
11342 /* and possibly remotely */
11343 if (st->update_tail)
11344 append_metadata_update(st, u, len);
11345 else
ed08d51c 11346 free(u);
8dd70bce 11347
694575e7 11348 } else {
7a862a02 11349 pr_err("(imsm) Operation is not allowed on this container\n");
694575e7
KW		 11350 }
11351 } else {
11352 /* On volume level we support following operations
471bceb6
KW		 11353 * - takeover: raid10 -> raid0; raid0 -> raid10
11354 * - chunk size migration
11355 * - migration: raid5 -> raid0; raid0 -> raid5
11356 */
11357 struct intel_super *super = st->sb;
11358 struct intel_dev *dev = super->devlist;
4dd2df09 11359 int change;
694575e7 11360 dprintf("imsm: info: Volume operation\n");
471bceb6
KW		 11361 /* find requested device */
11362 while (dev) {
1011e834 11363 char *devnm =
4dd2df09
N		 11364 imsm_find_array_devnm_by_subdev(
11365 dev->index, st->container_devnm);
11366 if (devnm && strcmp(devnm, geo.devnm) == 0)
471bceb6
KW		 11367 break;
11368 dev = dev->next;
11369 }
11370 if (dev == NULL) {
4dd2df09
N		 11371 pr_err("Cannot find %s (%s) subarray\n",
11372 geo.dev_name, geo.devnm);
471bceb6
KW		 11373 goto exit_imsm_reshape_super;
11374 }
11375 super->current_vol = dev->index;
fbf3d202 11376 change = imsm_analyze_change(st, &geo, direction);
694575e7 11377 switch (change) {
471bceb6 11378 case CH_TAKEOVER:
bb025c2f 11379 ret_val = imsm_takeover(st, &geo);
694575e7 11380 break;
48c5303a
PC		 11381 case CH_MIGRATION: {
11382 struct imsm_update_reshape_migration *u = NULL;
11383 int len =
11384 imsm_create_metadata_update_for_migration(
11385 st, &geo, &u);
11386 if (len < 1) {
7a862a02 11387 dprintf("imsm: Cannot prepare update\n");
48c5303a
PC		 11388 break;
11389 }
471bceb6 11390 ret_val = 0;
48c5303a
PC		 11391 /* update metadata locally */
11392 imsm_update_metadata_locally(st, u, len);
11393 /* and possibly remotely */
11394 if (st->update_tail)
11395 append_metadata_update(st, u, len);
11396 else
11397 free(u);
11398 }
11399 break;
7abc9871 11400 case CH_ARRAY_SIZE: {
f3871fdc
AK		 11401 struct imsm_update_size_change *u = NULL;
11402 int len =
11403 imsm_create_metadata_update_for_size_change(
11404 st, &geo, &u);
11405 if (len < 1) {
7a862a02 11406 dprintf("imsm: Cannot prepare update\n");
f3871fdc
AK		 11407 break;
11408 }
11409 ret_val = 0;
11410 /* update metadata locally */
11411 imsm_update_metadata_locally(st, u, len);
11412 /* and possibly remotely */
11413 if (st->update_tail)
11414 append_metadata_update(st, u, len);
11415 else
11416 free(u);
7abc9871
AK		 11417 }
11418 break;
471bceb6
KW		 11419 default:
11420 ret_val = 1;
694575e7 11421 }
694575e7 11422 }
78b10e66 11423
ed08d51c 11424exit_imsm_reshape_super:
78b10e66
N		 11425 dprintf("imsm: reshape_super Exit code = %i\n", ret_val);
11426 return ret_val;
11427}
2cda7640 11428
0febb20c
AO		 11429#define COMPLETED_OK 0
11430#define COMPLETED_NONE 1
11431#define COMPLETED_DELAYED 2
11432
11433static int read_completed(int fd, unsigned long long *val)
11434{
11435 int ret;
11436 char buf[50];
11437
11438 ret = sysfs_fd_get_str(fd, buf, 50);
11439 if (ret < 0)
11440 return ret;
11441
11442 ret = COMPLETED_OK;
11443 if (strncmp(buf, "none", 4) == 0) {
11444 ret = COMPLETED_NONE;
11445 } else if (strncmp(buf, "delayed", 7) == 0) {
11446 ret = COMPLETED_DELAYED;
11447 } else {
11448 char *ep;
11449 *val = strtoull(buf, &ep, 0);
11450 if (ep == buf || (*ep != 0 && *ep != '\n' && *ep != ' '))
11451 ret = -1;
11452 }
11453 return ret;
11454}
11455
eee67a47
AK		 11456/*******************************************************************************
11457 * Function: wait_for_reshape_imsm
11458 * Description: Function writes new sync_max value and waits until
11459 * reshape process reach new position
11460 * Parameters:
11461 * sra : general array info
eee67a47
AK		 11462 * ndata : number of disks in new array's layout
11463 * Returns:
11464 * 0 : success,
11465 * 1 : there is no reshape in progress,
11466 * -1 : fail
11467 ******************************************************************************/
ae9f01f8 11468int wait_for_reshape_imsm(struct mdinfo *sra, int ndata)
eee67a47 11469{
85ca499c 11470 int fd = sysfs_get_fd(sra, NULL, "sync_completed");
df2647fa 11471 int retry = 3;
eee67a47 11472 unsigned long long completed;
ae9f01f8
AK		 11473 /* to_complete : new sync_max position */
11474 unsigned long long to_complete = sra->reshape_progress;
11475 unsigned long long position_to_set = to_complete / ndata;
eee67a47 11476
ae9f01f8 11477 if (fd < 0) {
1ade5cc1 11478 dprintf("cannot open reshape_position\n");
eee67a47 11479 return 1;
ae9f01f8 11480 }
eee67a47 11481
df2647fa
PB		 11482 do {
11483 if (sysfs_fd_get_ll(fd, &completed) < 0) {
11484 if (!retry) {
11485 dprintf("cannot read reshape_position (no reshape in progres)\n");
11486 close(fd);
11487 return 1;
11488 }
11489 usleep(30000);
11490 } else
11491 break;
11492 } while (retry--);
eee67a47 11493
85ca499c 11494 if (completed > position_to_set) {
1ade5cc1 11495 dprintf("wrong next position to set %llu (%llu)\n",
85ca499c 11496 to_complete, position_to_set);
ae9f01f8
AK		 11497 close(fd);
11498 return -1;
11499 }
11500 dprintf("Position set: %llu\n", position_to_set);
11501 if (sysfs_set_num(sra, NULL, "sync_max",
11502 position_to_set) != 0) {
1ade5cc1 11503 dprintf("cannot set reshape position to %llu\n",
ae9f01f8
AK		 11504 position_to_set);
11505 close(fd);
11506 return -1;
eee67a47
AK		 11507 }
11508
eee67a47 11509 do {
0febb20c 11510 int rc;
eee67a47 11511 char action[20];
5ff3a780 11512 int timeout = 3000;
0febb20c 11513
5ff3a780 11514 sysfs_wait(fd, &timeout);
a47e44fb
AK		 11515 if (sysfs_get_str(sra, NULL, "sync_action",
11516 action, 20) > 0 &&
d7d3809a 11517 strncmp(action, "reshape", 7) != 0) {
b2be2b62
AO		 11518 if (strncmp(action, "idle", 4) == 0)
11519 break;
d7d3809a
AP		 11520 close(fd);
11521 return -1;
11522 }
0febb20c
AO		 11523
11524 rc = read_completed(fd, &completed);
11525 if (rc < 0) {
1ade5cc1 11526 dprintf("cannot read reshape_position (in loop)\n");
eee67a47
AK		 11527 close(fd);
11528 return 1;
0febb20c
AO		 11529 } else if (rc == COMPLETED_NONE)
11530 break;
85ca499c 11531 } while (completed < position_to_set);
b2be2b62 11532
eee67a47
AK		 11533 close(fd);
11534 return 0;
eee67a47
AK		 11535}
11536
b915c95f
AK		 11537/*******************************************************************************
11538 * Function: check_degradation_change
11539 * Description: Check that array hasn't become failed.
11540 * Parameters:
11541 * info : for sysfs access
11542 * sources : source disks descriptors
11543 * degraded: previous degradation level
11544 * Returns:
11545 * degradation level
11546 ******************************************************************************/
11547int check_degradation_change(struct mdinfo *info,
11548 int *sources,
11549 int degraded)
11550{
11551 unsigned long long new_degraded;
e1993023
LD		 11552 int rv;
11553
11554 rv = sysfs_get_ll(info, NULL, "degraded", &new_degraded);
089f9d79 11555 if (rv == -1 || (new_degraded != (unsigned long long)degraded)) {
b915c95f
AK		 11556 /* check each device to ensure it is still working */
11557 struct mdinfo *sd;
11558 new_degraded = 0;
11559 for (sd = info->devs ; sd ; sd = sd->next) {
11560 if (sd->disk.state & (1<<MD_DISK_FAULTY))
11561 continue;
11562 if (sd->disk.state & (1<<MD_DISK_SYNC)) {
cf52eff5
TM		 11563 char sbuf[100];
11564
b915c95f 11565 if (sysfs_get_str(info,
cf52eff5 11566 sd, "state", sbuf, sizeof(sbuf)) < 0 ||
b915c95f
AK		 11567 strstr(sbuf, "faulty") ||
11568 strstr(sbuf, "in_sync") == NULL) {
11569 /* this device is dead */
11570 sd->disk.state = (1<<MD_DISK_FAULTY);
11571 if (sd->disk.raid_disk >= 0 &&
11572 sources[sd->disk.raid_disk] >= 0) {
11573 close(sources[
11574 sd->disk.raid_disk]);
11575 sources[sd->disk.raid_disk] =
11576 -1;
11577 }
11578 new_degraded++;
11579 }
11580 }
11581 }
11582 }
11583
11584 return new_degraded;
11585}
11586
10f22854
AK		 11587/*******************************************************************************
11588 * Function: imsm_manage_reshape
11589 * Description: Function finds array under reshape and it manages reshape
11590 * process. It creates stripes backups (if required) and sets
942e1cdb 11591 * checkpoints.
10f22854
AK		 11592 * Parameters:
11593 * afd : Backup handle (nattive) - not used
11594 * sra : general array info
11595 * reshape : reshape parameters - not used
11596 * st : supertype structure
11597 * blocks : size of critical section [blocks]
11598 * fds : table of source device descriptor
11599 * offsets : start of array (offest per devices)
11600 * dests : not used
11601 * destfd : table of destination device descriptor
11602 * destoffsets : table of destination offsets (per device)
11603 * Returns:
11604 * 1 : success, reshape is done
11605 * 0 : fail
11606 ******************************************************************************/
999b4972
N		 11607static int imsm_manage_reshape(
11608 int afd, struct mdinfo *sra, struct reshape *reshape,
10f22854 11609 struct supertype *st, unsigned long backup_blocks,
999b4972
N		 11610 int *fds, unsigned long long *offsets,
11611 int dests, int *destfd, unsigned long long *destoffsets)
11612{
10f22854
AK		 11613 int ret_val = 0;
11614 struct intel_super *super = st->sb;
594dc1b8 11615 struct intel_dev *dv;
de44e46f 11616 unsigned int sector_size = super->sector_size;
10f22854 11617 struct imsm_dev *dev = NULL;
a6b6d984 11618 struct imsm_map *map_src;
10f22854
AK		 11619 int migr_vol_qan = 0;
11620 int ndata, odata; /* [bytes] */
11621 int chunk; /* [bytes] */
11622 struct migr_record *migr_rec;
11623 char *buf = NULL;
11624 unsigned int buf_size; /* [bytes] */
11625 unsigned long long max_position; /* array size [bytes] */
11626 unsigned long long next_step; /* [blocks]/[bytes] */
11627 unsigned long long old_data_stripe_length;
10f22854
AK		 11628 unsigned long long start_src; /* [bytes] */
11629 unsigned long long start; /* [bytes] */
11630 unsigned long long start_buf_shift; /* [bytes] */
b915c95f 11631 int degraded = 0;
ab724b98 11632 int source_layout = 0;
10f22854 11633
79a16a9b
JS		 11634 if (!sra)
11635 return ret_val;
11636
11637 if (!fds || !offsets)
10f22854
AK		 11638 goto abort;
11639
11640 /* Find volume during the reshape */
11641 for (dv = super->devlist; dv; dv = dv->next) {
11642 if (dv->dev->vol.migr_type == MIGR_GEN_MIGR
11643 && dv->dev->vol.migr_state == 1) {
11644 dev = dv->dev;
11645 migr_vol_qan++;
11646 }
11647 }
11648 /* Only one volume can migrate at the same time */
11649 if (migr_vol_qan != 1) {
676e87a8 11650 pr_err("%s", migr_vol_qan ?
10f22854
AK		 11651 "Number of migrating volumes greater than 1\n" :
11652 "There is no volume during migrationg\n");
11653 goto abort;
11654 }
11655
238c0a71 11656 map_src = get_imsm_map(dev, MAP_1);
10f22854
AK		 11657 if (map_src == NULL)
11658 goto abort;
10f22854 11659
238c0a71
AK		 11660 ndata = imsm_num_data_members(dev, MAP_0);
11661 odata = imsm_num_data_members(dev, MAP_1);
10f22854 11662
7b1ab482 11663 chunk = __le16_to_cpu(map_src->blocks_per_strip) * 512;
10f22854
AK		 11664 old_data_stripe_length = odata * chunk;
11665
11666 migr_rec = super->migr_rec;
11667
10f22854
AK		 11668 /* initialize migration record for start condition */
11669 if (sra->reshape_progress == 0)
11670 init_migr_record_imsm(st, dev, sra);
b2c59438
AK		 11671 else {
11672 if (__le32_to_cpu(migr_rec->rec_status) != UNIT_SRC_NORMAL) {
7a862a02 11673 dprintf("imsm: cannot restart migration when data are present in copy area.\n");
b2c59438
AK		 11674 goto abort;
11675 }
6a75c8ca
AK		 11676 /* Save checkpoint to update migration record for current
11677 * reshape position (in md). It can be farther than current
11678 * reshape position in metadata.
11679 */
11680 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
11681 /* ignore error == 2, this can mean end of reshape here
11682 */
7a862a02 11683 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL, initial save)\n");
6a75c8ca
AK		 11684 goto abort;
11685 }
b2c59438 11686 }
10f22854
AK		 11687
11688 /* size for data */
11689 buf_size = __le32_to_cpu(migr_rec->blocks_per_unit) * 512;
11690 /* extend buffer size for parity disk */
11691 buf_size += __le32_to_cpu(migr_rec->dest_depth_per_unit) * 512;
11692 /* add space for stripe aligment */
11693 buf_size += old_data_stripe_length;
de44e46f
PB		 11694 if (posix_memalign((void **)&buf, MAX_SECTOR_SIZE, buf_size)) {
11695 dprintf("imsm: Cannot allocate checkpoint buffer\n");
10f22854
AK		 11696 goto abort;
11697 }
11698
3ef4403c 11699 max_position = sra->component_size * ndata;
68eb8bc6 11700 source_layout = imsm_level_to_layout(map_src->raid_level);
10f22854
AK		 11701
11702 while (__le32_to_cpu(migr_rec->curr_migr_unit) <
11703 __le32_to_cpu(migr_rec->num_migr_units)) {
11704 /* current reshape position [blocks] */
11705 unsigned long long current_position =
11706 __le32_to_cpu(migr_rec->blocks_per_unit)
11707 * __le32_to_cpu(migr_rec->curr_migr_unit);
11708 unsigned long long border;
11709
b915c95f
AK		 11710 /* Check that array hasn't become failed.
11711 */
11712 degraded = check_degradation_change(sra, fds, degraded);
11713 if (degraded > 1) {
7a862a02 11714 dprintf("imsm: Abort reshape due to degradation level (%i)\n", degraded);
b915c95f
AK		 11715 goto abort;
11716 }
11717
10f22854
AK		 11718 next_step = __le32_to_cpu(migr_rec->blocks_per_unit);
11719
11720 if ((current_position + next_step) > max_position)
11721 next_step = max_position - current_position;
11722
92144abf 11723 start = current_position * 512;
10f22854 11724
942e1cdb 11725 /* align reading start to old geometry */
10f22854
AK		 11726 start_buf_shift = start % old_data_stripe_length;
11727 start_src = start - start_buf_shift;
11728
11729 border = (start_src / odata) - (start / ndata);
11730 border /= 512;
11731 if (border <= __le32_to_cpu(migr_rec->dest_depth_per_unit)) {
11732 /* save critical stripes to buf
11733 * start - start address of current unit
11734 * to backup [bytes]
11735 * start_src - start address of current unit
11736 * to backup alligned to source array
11737 * [bytes]
11738 */
594dc1b8 11739 unsigned long long next_step_filler;
10f22854
AK		 11740 unsigned long long copy_length = next_step * 512;
11741
11742 /* allign copy area length to stripe in old geometry */
11743 next_step_filler = ((copy_length + start_buf_shift)
11744 % old_data_stripe_length);
11745 if (next_step_filler)
11746 next_step_filler = (old_data_stripe_length
11747 - next_step_filler);
7a862a02 11748 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		 11749 start, start_src, copy_length,
11750 start_buf_shift, next_step_filler);
11751
11752 if (save_stripes(fds, offsets, map_src->num_members,
ab724b98
AK		 11753 chunk, map_src->raid_level,
11754 source_layout, 0, NULL, start_src,
10f22854
AK		 11755 copy_length +
11756 next_step_filler + start_buf_shift,
11757 buf)) {
7a862a02 11758 dprintf("imsm: Cannot save stripes to buffer\n");
10f22854
AK		 11759 goto abort;
11760 }
11761 /* Convert data to destination format and store it
11762 * in backup general migration area
11763 */
11764 if (save_backup_imsm(st, dev, sra,
aea93171 11765 buf + start_buf_shift, copy_length)) {
7a862a02 11766 dprintf("imsm: Cannot save stripes to target devices\n");
10f22854
AK		 11767 goto abort;
11768 }
11769 if (save_checkpoint_imsm(st, sra,
11770 UNIT_SRC_IN_CP_AREA)) {
7a862a02 11771 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_IN_CP_AREA)\n");
10f22854
AK		 11772 goto abort;
11773 }
8016a6d4
AK		 11774 } else {
11775 /* set next step to use whole border area */
11776 border /= next_step;
11777 if (border > 1)
11778 next_step *= border;
10f22854
AK		 11779 }
11780 /* When data backed up, checkpoint stored,
11781 * kick the kernel to reshape unit of data
11782 */
11783 next_step = next_step + sra->reshape_progress;
8016a6d4
AK		 11784 /* limit next step to array max position */
11785 if (next_step > max_position)
11786 next_step = max_position;
10f22854
AK		 11787 sysfs_set_num(sra, NULL, "suspend_lo", sra->reshape_progress);
11788 sysfs_set_num(sra, NULL, "suspend_hi", next_step);
ae9f01f8 11789 sra->reshape_progress = next_step;
10f22854
AK		 11790
11791 /* wait until reshape finish */
c85338c6 11792 if (wait_for_reshape_imsm(sra, ndata)) {
c47b0ff6
AK		 11793 dprintf("wait_for_reshape_imsm returned error!\n");
11794 goto abort;
11795 }
84d11e6c
N		 11796 if (sigterm)
11797 goto abort;
10f22854 11798
0228d92c
AK		 11799 if (save_checkpoint_imsm(st, sra, UNIT_SRC_NORMAL) == 1) {
11800 /* ignore error == 2, this can mean end of reshape here
11801 */
7a862a02 11802 dprintf("imsm: Cannot write checkpoint to migration record (UNIT_SRC_NORMAL)\n");
10f22854
AK		 11803 goto abort;
11804 }
11805
11806 }
11807
71e5411e
PB		 11808 /* clear migr_rec on disks after successful migration */
11809 struct dl *d;
11810
de44e46f 11811 memset(super->migr_rec_buf, 0, MIGR_REC_BUF_SECTORS*sector_size);
71e5411e
PB		 11812 for (d = super->disks; d; d = d->next) {
11813 if (d->index < 0 || is_failed(&d->disk))
11814 continue;
11815 unsigned long long dsize;
11816
11817 get_dev_size(d->fd, NULL, &dsize);
de44e46f 11818 if (lseek64(d->fd, dsize - MIGR_REC_SECTOR_POSITION*sector_size,
71e5411e 11819 SEEK_SET) >= 0) {
466070ad 11820 if ((unsigned int)write(d->fd, super->migr_rec_buf,
de44e46f
PB		 11821 MIGR_REC_BUF_SECTORS*sector_size) !=
11822 MIGR_REC_BUF_SECTORS*sector_size)
71e5411e
PB		 11823 perror("Write migr_rec failed");
11824 }
11825 }
11826
10f22854
AK		 11827 /* return '1' if done */
11828 ret_val = 1;
11829abort:
11830 free(buf);
942e1cdb
N		 11831 /* See Grow.c: abort_reshape() for further explanation */
11832 sysfs_set_num(sra, NULL, "suspend_lo", 0x7FFFFFFFFFFFFFFFULL);
11833 sysfs_set_num(sra, NULL, "suspend_hi", 0);
11834 sysfs_set_num(sra, NULL, "suspend_lo", 0);
10f22854
AK		 11835
11836 return ret_val;
999b4972 11837}
0c21b485 11838
71204a50 11839#endif /* MDASSEMBLE */
999b4972 11840
cdddbdbc
DW		 11841struct superswitch super_imsm = {
11842#ifndef MDASSEMBLE
11843 .examine_super = examine_super_imsm,
11844 .brief_examine_super = brief_examine_super_imsm,
4737ae25 11845 .brief_examine_subarrays = brief_examine_subarrays_imsm,
9d84c8ea 11846 .export_examine_super = export_examine_super_imsm,
cdddbdbc
DW		 11847 .detail_super = detail_super_imsm,
11848 .brief_detail_super = brief_detail_super_imsm,
bf5a934a 11849 .write_init_super = write_init_super_imsm,
0e600426
N		 11850 .validate_geometry = validate_geometry_imsm,
11851 .add_to_super = add_to_super_imsm,
1a64be56 11852 .remove_from_super = remove_from_super_imsm,
d665cc31 11853 .detail_platform = detail_platform_imsm,
e50cf220 11854 .export_detail_platform = export_detail_platform_imsm,
33414a01 11855 .kill_subarray = kill_subarray_imsm,
aa534678 11856 .update_subarray = update_subarray_imsm,
2b959fbf 11857 .load_container = load_container_imsm,
71204a50
N		 11858 .default_geometry = default_geometry_imsm,
11859 .get_disk_controller_domain = imsm_get_disk_controller_domain,
11860 .reshape_super = imsm_reshape_super,
11861 .manage_reshape = imsm_manage_reshape,
9e2d750d 11862 .recover_backup = recover_backup_imsm,
74db60b0 11863 .copy_metadata = copy_metadata_imsm,
27156a57 11864 .examine_badblocks = examine_badblocks_imsm,
cdddbdbc
DW		 11865#endif
11866 .match_home = match_home_imsm,
11867 .uuid_from_super= uuid_from_super_imsm,
11868 .getinfo_super = getinfo_super_imsm,
5c4cd5da 11869 .getinfo_super_disks = getinfo_super_disks_imsm,
cdddbdbc
DW		 11870 .update_super = update_super_imsm,
11871
11872 .avail_size = avail_size_imsm,
80e7f8c3 11873 .min_acceptable_spare_size = min_acceptable_spare_size_imsm,
cdddbdbc
DW		 11874
11875 .compare_super = compare_super_imsm,
11876
11877 .load_super = load_super_imsm,
bf5a934a 11878 .init_super = init_super_imsm,
e683ca88 11879 .store_super = store_super_imsm,
cdddbdbc
DW		 11880 .free_super = free_super_imsm,
11881 .match_metadata_desc = match_metadata_desc_imsm,
bf5a934a 11882 .container_content = container_content_imsm,
0c21b485 11883 .validate_container = validate_container_imsm,
cdddbdbc 11884
2432ce9b
AP		 11885 .write_init_ppl = write_init_ppl_imsm,
11886 .validate_ppl = validate_ppl_imsm,
11887
cdddbdbc 11888 .external = 1,
4cce4069 11889 .name = "imsm",
845dea95 11890
0e600426 11891#ifndef MDASSEMBLE
845dea95
NB		 11892/* for mdmon */
11893 .open_new = imsm_open_new,
ed9d66aa 11894 .set_array_state= imsm_set_array_state,
845dea95
NB		 11895 .set_disk = imsm_set_disk,
11896 .sync_metadata = imsm_sync_metadata,
88758e9d 11897 .activate_spare = imsm_activate_spare,
e8319a19 11898 .process_update = imsm_process_update,
8273f55e 11899 .prepare_update = imsm_prepare_update,
6f50473f 11900 .record_bad_block = imsm_record_badblock,
c07a5a4f 11901 .clear_bad_block = imsm_clear_badblock,
928f1424 11902 .get_bad_blocks = imsm_get_badblocks,
0e600426 11903#endif /* MDASSEMBLE */
cdddbdbc 11904};
Unnamed repository; edit this file 'description' to name the repository.