drivers/md/md-cluster.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  * Copyright (C) 2015, SUSE
   4  */
   5
   6
   7 #include <linux/module.h>
   8 #include <linux/kthread.h>
   9 #include <linux/dlm.h>
  10 #include <linux/sched.h>
  11 #include <linux/raid/md_p.h>
  12 #include "md.h"
  13 #include "md-bitmap.h"
  14 #include "md-cluster.h"
  15
  16 #define LVB_SIZE        64
  17 #define NEW_DEV_TIMEOUT 5000
  18
  19 struct dlm_lock_resource {
  20         dlm_lockspace_t *ls;
  21         struct dlm_lksb lksb;
  22         char *name; /* lock name. */
  23         uint32_t flags; /* flags to pass to dlm_lock() */
  24         wait_queue_head_t sync_locking; /* wait queue for synchronized locking */
  25         bool sync_locking_done;
  26         void (*bast)(void *arg, int mode); /* blocking AST function pointer*/
  27         struct mddev *mddev; /* pointing back to mddev. */
  28         int mode;
  29 };
  30
  31 struct resync_info {
  32         __le64 lo;
  33         __le64 hi;
  34 };
  35
  36 /* md_cluster_info flags */
  37 #define         MD_CLUSTER_WAITING_FOR_NEWDISK          1
  38 #define         MD_CLUSTER_SUSPEND_READ_BALANCING       2
  39 #define         MD_CLUSTER_BEGIN_JOIN_CLUSTER           3
  40
  41 /* Lock the send communication. This is done through
  42  * bit manipulation as opposed to a mutex in order to
  43  * accomodate lock and hold. See next comment.
  44  */
  45 #define         MD_CLUSTER_SEND_LOCK                    4
  46 /* If cluster operations (such as adding a disk) must lock the
  47  * communication channel, so as to perform extra operations
  48  * (update metadata) and no other operation is allowed on the
  49  * MD. Token needs to be locked and held until the operation
  50  * completes witha md_update_sb(), which would eventually release
  51  * the lock.
  52  */
  53 #define         MD_CLUSTER_SEND_LOCKED_ALREADY          5
  54 /* We should receive message after node joined cluster and
  55  * set up all the related infos such as bitmap and personality */
  56 #define         MD_CLUSTER_ALREADY_IN_CLUSTER           6
  57 #define         MD_CLUSTER_PENDING_RECV_EVENT           7
  58 #define         MD_CLUSTER_HOLDING_MUTEX_FOR_RECVD              8
  59
  60 struct md_cluster_info {
  61         struct mddev *mddev; /* the md device which md_cluster_info belongs to */
  62         /* dlm lock space and resources for clustered raid. */
  63         dlm_lockspace_t *lockspace;
  64         int slot_number;
  65         struct completion completion;
  66         struct mutex recv_mutex;
  67         struct dlm_lock_resource *bitmap_lockres;
  68         struct dlm_lock_resource **other_bitmap_lockres;
  69         struct dlm_lock_resource *resync_lockres;
  70         struct list_head suspend_list;
  71
  72         spinlock_t suspend_lock;
  73         /* record the region which write should be suspended */
  74         sector_t suspend_lo;
  75         sector_t suspend_hi;
  76         int suspend_from; /* the slot which broadcast suspend_lo/hi */
  77
  78         struct md_thread *recovery_thread;
  79         unsigned long recovery_map;
  80         /* communication loc resources */
  81         struct dlm_lock_resource *ack_lockres;
  82         struct dlm_lock_resource *message_lockres;
  83         struct dlm_lock_resource *token_lockres;
  84         struct dlm_lock_resource *no_new_dev_lockres;
  85         struct md_thread *recv_thread;
  86         struct completion newdisk_completion;
  87         wait_queue_head_t wait;
  88         unsigned long state;
  89         /* record the region in RESYNCING message */
  90         sector_t sync_low;
  91         sector_t sync_hi;
  92 };
  93
  94 enum msg_type {
  95         METADATA_UPDATED = 0,
  96         RESYNCING,
  97         NEWDISK,
  98         REMOVE,
  99         RE_ADD,
 100         BITMAP_NEEDS_SYNC,
 101         CHANGE_CAPACITY,
 102         BITMAP_RESIZE,
 103 };
 104
 105 struct cluster_msg {
 106         __le32 type;
 107         __le32 slot;
 108         /* TODO: Unionize this for smaller footprint */
 109         __le64 low;
 110         __le64 high;
 111         char uuid[16];
 112         __le32 raid_slot;
 113 };
 114
 115 static void sync_ast(void *arg)
 116 {
 117         struct dlm_lock_resource *res;
 118
 119         res = arg;
 120         res->sync_locking_done = true;
 121         wake_up(&res->sync_locking);
 122 }
 123
 124 static int dlm_lock_sync(struct dlm_lock_resource *res, int mode)
 125 {
 126         int ret = 0;
 127
 128         ret = dlm_lock(res->ls, mode, &res->lksb,
 129                         res->flags, res->name, strlen(res->name),
 130                         0, sync_ast, res, res->bast);
 131         if (ret)
 132                 return ret;
 133         wait_event(res->sync_locking, res->sync_locking_done);
 134         res->sync_locking_done = false;
 135         if (res->lksb.sb_status == 0)
 136                 res->mode = mode;
 137         return res->lksb.sb_status;
 138 }
 139
 140 static int dlm_unlock_sync(struct dlm_lock_resource *res)
 141 {
 142         return dlm_lock_sync(res, DLM_LOCK_NL);
 143 }
 144
 145 /*
 146  * An variation of dlm_lock_sync, which make lock request could
 147  * be interrupted
 148  */
 149 static int dlm_lock_sync_interruptible(struct dlm_lock_resource *res, int mode,
 150                                        struct mddev *mddev)
 151 {
 152         int ret = 0;
 153
 154         ret = dlm_lock(res->ls, mode, &res->lksb,
 155                         res->flags, res->name, strlen(res->name),
 156                         0, sync_ast, res, res->bast);
 157         if (ret)
 158                 return ret;
 159
 160         wait_event(res->sync_locking, res->sync_locking_done
 161                                       || kthread_should_stop()
 162                                       || test_bit(MD_CLOSING, &mddev->flags));
 163         if (!res->sync_locking_done) {
 164                 /*
 165                  * the convert queue contains the lock request when request is
 166                  * interrupted, and sync_ast could still be run, so need to
 167                  * cancel the request and reset completion
 168                  */
 169                 ret = dlm_unlock(res->ls, res->lksb.sb_lkid, DLM_LKF_CANCEL,
 170                         &res->lksb, res);
 171                 res->sync_locking_done = false;
 172                 if (unlikely(ret != 0))
 173                         pr_info("failed to cancel previous lock request "
 174                                  "%s return %d\n", res->name, ret);
 175                 return -EPERM;
 176         } else
 177                 res->sync_locking_done = false;
 178         if (res->lksb.sb_status == 0)
 179                 res->mode = mode;
 180         return res->lksb.sb_status;
 181 }
 182
 183 static struct dlm_lock_resource *lockres_init(struct mddev *mddev,
 184                 char *name, void (*bastfn)(void *arg, int mode), int with_lvb)
 185 {
 186         struct dlm_lock_resource *res = NULL;
 187         int ret, namelen;
 188         struct md_cluster_info *cinfo = mddev->cluster_info;
 189
 190         res = kzalloc(sizeof(struct dlm_lock_resource), GFP_KERNEL);
 191         if (!res)
 192                 return NULL;
 193         init_waitqueue_head(&res->sync_locking);
 194         res->sync_locking_done = false;
 195         res->ls = cinfo->lockspace;
 196         res->mddev = mddev;
 197         res->mode = DLM_LOCK_IV;
 198         namelen = strlen(name);
 199         res->name = kzalloc(namelen + 1, GFP_KERNEL);
 200         if (!res->name) {
 201                 pr_err("md-cluster: Unable to allocate resource name for resource %s\n", name);
 202                 goto out_err;
 203         }
 204         strlcpy(res->name, name, namelen + 1);
 205         if (with_lvb) {
 206                 res->lksb.sb_lvbptr = kzalloc(LVB_SIZE, GFP_KERNEL);
 207                 if (!res->lksb.sb_lvbptr) {
 208                         pr_err("md-cluster: Unable to allocate LVB for resource %s\n", name);
 209                         goto out_err;
 210                 }
 211                 res->flags = DLM_LKF_VALBLK;
 212         }
 213
 214         if (bastfn)
 215                 res->bast = bastfn;
 216
 217         res->flags |= DLM_LKF_EXPEDITE;
 218
 219         ret = dlm_lock_sync(res, DLM_LOCK_NL);
 220         if (ret) {
 221                 pr_err("md-cluster: Unable to lock NL on new lock resource %s\n", name);
 222                 goto out_err;
 223         }
 224         res->flags &= ~DLM_LKF_EXPEDITE;
 225         res->flags |= DLM_LKF_CONVERT;
 226
 227         return res;
 228 out_err:
 229         kfree(res->lksb.sb_lvbptr);
 230         kfree(res->name);
 231         kfree(res);
 232         return NULL;
 233 }
 234
 235 static void lockres_free(struct dlm_lock_resource *res)
 236 {
 237         int ret = 0;
 238
 239         if (!res)
 240                 return;
 241
 242         /*
 243          * use FORCEUNLOCK flag, so we can unlock even the lock is on the
 244          * waiting or convert queue
 245          */
 246         ret = dlm_unlock(res->ls, res->lksb.sb_lkid, DLM_LKF_FORCEUNLOCK,
 247                 &res->lksb, res);
 248         if (unlikely(ret != 0))
 249                 pr_err("failed to unlock %s return %d\n", res->name, ret);
 250         else
 251                 wait_event(res->sync_locking, res->sync_locking_done);
 252
 253         kfree(res->name);
 254         kfree(res->lksb.sb_lvbptr);
 255         kfree(res);
 256 }
 257
 258 static void add_resync_info(struct dlm_lock_resource *lockres,
 259                             sector_t lo, sector_t hi)
 260 {
 261         struct resync_info *ri;
 262
 263         ri = (struct resync_info *)lockres->lksb.sb_lvbptr;
 264         ri->lo = cpu_to_le64(lo);
 265         ri->hi = cpu_to_le64(hi);
 266 }
 267
 268 static int read_resync_info(struct mddev *mddev,
 269                             struct dlm_lock_resource *lockres)
 270 {
 271         struct resync_info ri;
 272         struct md_cluster_info *cinfo = mddev->cluster_info;
 273         int ret = 0;
 274
 275         dlm_lock_sync(lockres, DLM_LOCK_CR);
 276         memcpy(&ri, lockres->lksb.sb_lvbptr, sizeof(struct resync_info));
 277         if (le64_to_cpu(ri.hi) > 0) {
 278                 cinfo->suspend_hi = le64_to_cpu(ri.hi);
 279                 cinfo->suspend_lo = le64_to_cpu(ri.lo);
 280                 ret = 1;
 281         }
 282         dlm_unlock_sync(lockres);
 283         return ret;
 284 }
 285
 286 static void recover_bitmaps(struct md_thread *thread)
 287 {
 288         struct mddev *mddev = thread->mddev;
 289         struct md_cluster_info *cinfo = mddev->cluster_info;
 290         struct dlm_lock_resource *bm_lockres;
 291         char str[64];
 292         int slot, ret;
 293         sector_t lo, hi;
 294
 295         while (cinfo->recovery_map) {
 296                 slot = fls64((u64)cinfo->recovery_map) - 1;
 297
 298                 snprintf(str, 64, "bitmap%04d", slot);
 299                 bm_lockres = lockres_init(mddev, str, NULL, 1);
 300                 if (!bm_lockres) {
 301                         pr_err("md-cluster: Cannot initialize bitmaps\n");
 302                         goto clear_bit;
 303                 }
 304
 305                 ret = dlm_lock_sync_interruptible(bm_lockres, DLM_LOCK_PW, mddev);
 306                 if (ret) {
 307                         pr_err("md-cluster: Could not DLM lock %s: %d\n",
 308                                         str, ret);
 309                         goto clear_bit;
 310                 }
 311                 ret = md_bitmap_copy_from_slot(mddev, slot, &lo, &hi, true);
 312                 if (ret) {
 313                         pr_err("md-cluster: Could not copy data from bitmap %d\n", slot);
 314                         goto clear_bit;
 315                 }
 316
 317                 /* Clear suspend_area associated with the bitmap */
 318                 spin_lock_irq(&cinfo->suspend_lock);
 319                 cinfo->suspend_hi = 0;
 320                 cinfo->suspend_lo = 0;
 321                 cinfo->suspend_from = -1;
 322                 spin_unlock_irq(&cinfo->suspend_lock);
 323
 324                 /* Kick off a reshape if needed */
 325                 if (test_bit(MD_RESYNCING_REMOTE, &mddev->recovery) &&
 326                     test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) &&
 327                     mddev->reshape_position != MaxSector)
 328                         md_wakeup_thread(mddev->sync_thread);
 329
 330                 if (hi > 0) {
 331                         if (lo < mddev->recovery_cp)
 332                                 mddev->recovery_cp = lo;
 333                         /* wake up thread to continue resync in case resync
 334                          * is not finished */
 335                         if (mddev->recovery_cp != MaxSector) {
 336                                 /*
 337                                  * clear the REMOTE flag since we will launch
 338                                  * resync thread in current node.
 339                                  */
 340                                 clear_bit(MD_RESYNCING_REMOTE,
 341                                           &mddev->recovery);
 342                                 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
 343                                 md_wakeup_thread(mddev->thread);
 344                         }
 345                 }
 346 clear_bit:
 347                 lockres_free(bm_lockres);
 348                 clear_bit(slot, &cinfo->recovery_map);
 349         }
 350 }
 351
 352 static void recover_prep(void *arg)
 353 {
 354         struct mddev *mddev = arg;
 355         struct md_cluster_info *cinfo = mddev->cluster_info;
 356         set_bit(MD_CLUSTER_SUSPEND_READ_BALANCING, &cinfo->state);
 357 }
 358
 359 static void __recover_slot(struct mddev *mddev, int slot)
 360 {
 361         struct md_cluster_info *cinfo = mddev->cluster_info;
 362
 363         set_bit(slot, &cinfo->recovery_map);
 364         if (!cinfo->recovery_thread) {
 365                 cinfo->recovery_thread = md_register_thread(recover_bitmaps,
 366                                 mddev, "recover");
 367                 if (!cinfo->recovery_thread) {
 368                         pr_warn("md-cluster: Could not create recovery thread\n");
 369                         return;
 370                 }
 371         }
 372         md_wakeup_thread(cinfo->recovery_thread);
 373 }
 374
 375 static void recover_slot(void *arg, struct dlm_slot *slot)
 376 {
 377         struct mddev *mddev = arg;
 378         struct md_cluster_info *cinfo = mddev->cluster_info;
 379
 380         pr_info("md-cluster: %s Node %d/%d down. My slot: %d. Initiating recovery.\n",
 381                         mddev->bitmap_info.cluster_name,
 382                         slot->nodeid, slot->slot,
 383                         cinfo->slot_number);
 384         /* deduct one since dlm slot starts from one while the num of
 385          * cluster-md begins with 0 */
 386         __recover_slot(mddev, slot->slot - 1);
 387 }
 388
 389 static void recover_done(void *arg, struct dlm_slot *slots,
 390                 int num_slots, int our_slot,
 391                 uint32_t generation)
 392 {
 393         struct mddev *mddev = arg;
 394         struct md_cluster_info *cinfo = mddev->cluster_info;
 395
 396         cinfo->slot_number = our_slot;
 397         /* completion is only need to be complete when node join cluster,
 398          * it doesn't need to run during another node's failure */
 399         if (test_bit(MD_CLUSTER_BEGIN_JOIN_CLUSTER, &cinfo->state)) {
 400                 complete(&cinfo->completion);
 401                 clear_bit(MD_CLUSTER_BEGIN_JOIN_CLUSTER, &cinfo->state);
 402         }
 403         clear_bit(MD_CLUSTER_SUSPEND_READ_BALANCING, &cinfo->state);
 404 }
 405
 406 /* the ops is called when node join the cluster, and do lock recovery
 407  * if node failure occurs */
 408 static const struct dlm_lockspace_ops md_ls_ops = {
 409         .recover_prep = recover_prep,
 410         .recover_slot = recover_slot,
 411         .recover_done = recover_done,
 412 };
 413
 414 /*
 415  * The BAST function for the ack lock resource
 416  * This function wakes up the receive thread in
 417  * order to receive and process the message.
 418  */
 419 static void ack_bast(void *arg, int mode)
 420 {
 421         struct dlm_lock_resource *res = arg;
 422         struct md_cluster_info *cinfo = res->mddev->cluster_info;
 423
 424         if (mode == DLM_LOCK_EX) {
 425                 if (test_bit(MD_CLUSTER_ALREADY_IN_CLUSTER, &cinfo->state))
 426                         md_wakeup_thread(cinfo->recv_thread);
 427                 else
 428                         set_bit(MD_CLUSTER_PENDING_RECV_EVENT, &cinfo->state);
 429         }
 430 }
 431
 432 static void remove_suspend_info(struct mddev *mddev, int slot)
 433 {
 434         struct md_cluster_info *cinfo = mddev->cluster_info;
 435         mddev->pers->quiesce(mddev, 1);
 436         spin_lock_irq(&cinfo->suspend_lock);
 437         cinfo->suspend_hi = 0;
 438         cinfo->suspend_lo = 0;
 439         spin_unlock_irq(&cinfo->suspend_lock);
 440         mddev->pers->quiesce(mddev, 0);
 441 }
 442
 443 static void process_suspend_info(struct mddev *mddev,
 444                 int slot, sector_t lo, sector_t hi)
 445 {
 446         struct md_cluster_info *cinfo = mddev->cluster_info;
 447         struct mdp_superblock_1 *sb = NULL;
 448         struct md_rdev *rdev;
 449
 450         if (!hi) {
 451                 /*
 452                  * clear the REMOTE flag since resync or recovery is finished
 453                  * in remote node.
 454                  */
 455                 clear_bit(MD_RESYNCING_REMOTE, &mddev->recovery);
 456                 remove_suspend_info(mddev, slot);
 457                 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
 458                 md_wakeup_thread(mddev->thread);
 459                 return;
 460         }
 461
 462         rdev_for_each(rdev, mddev)
 463                 if (rdev->raid_disk > -1 && !test_bit(Faulty, &rdev->flags)) {
 464                         sb = page_address(rdev->sb_page);
 465                         break;
 466                 }
 467
 468         /*
 469          * The bitmaps are not same for different nodes
 470          * if RESYNCING is happening in one node, then
 471          * the node which received the RESYNCING message
 472          * probably will perform resync with the region
 473          * [lo, hi] again, so we could reduce resync time
 474          * a lot if we can ensure that the bitmaps among
 475          * different nodes are match up well.
 476          *
 477          * sync_low/hi is used to record the region which
 478          * arrived in the previous RESYNCING message,
 479          *
 480          * Call md_bitmap_sync_with_cluster to clear NEEDED_MASK
 481          * and set RESYNC_MASK since  resync thread is running
 482          * in another node, so we don't need to do the resync
 483          * again with the same section.
 484          *
 485          * Skip md_bitmap_sync_with_cluster in case reshape
 486          * happening, because reshaping region is small and
 487          * we don't want to trigger lots of WARN.
 488          */
 489         if (sb && !(le32_to_cpu(sb->feature_map) & MD_FEATURE_RESHAPE_ACTIVE))
 490                 md_bitmap_sync_with_cluster(mddev, cinfo->sync_low,
 491                                             cinfo->sync_hi, lo, hi);
 492         cinfo->sync_low = lo;
 493         cinfo->sync_hi = hi;
 494
 495         mddev->pers->quiesce(mddev, 1);
 496         spin_lock_irq(&cinfo->suspend_lock);
 497         cinfo->suspend_from = slot;
 498         cinfo->suspend_lo = lo;
 499         cinfo->suspend_hi = hi;
 500         spin_unlock_irq(&cinfo->suspend_lock);
 501         mddev->pers->quiesce(mddev, 0);
 502 }
 503
 504 static void process_add_new_disk(struct mddev *mddev, struct cluster_msg *cmsg)
 505 {
 506         char disk_uuid[64];
 507         struct md_cluster_info *cinfo = mddev->cluster_info;
 508         char event_name[] = "EVENT=ADD_DEVICE";
 509         char raid_slot[16];
 510         char *envp[] = {event_name, disk_uuid, raid_slot, NULL};
 511         int len;
 512
 513         len = snprintf(disk_uuid, 64, "DEVICE_UUID=");
 514         sprintf(disk_uuid + len, "%pU", cmsg->uuid);
 515         snprintf(raid_slot, 16, "RAID_DISK=%d", le32_to_cpu(cmsg->raid_slot));
 516         pr_info("%s:%d Sending kobject change with %s and %s\n", __func__, __LINE__, disk_uuid, raid_slot);
 517         init_completion(&cinfo->newdisk_completion);
 518         set_bit(MD_CLUSTER_WAITING_FOR_NEWDISK, &cinfo->state);
 519         kobject_uevent_env(&disk_to_dev(mddev->gendisk)->kobj, KOBJ_CHANGE, envp);
 520         wait_for_completion_timeout(&cinfo->newdisk_completion,
 521                         NEW_DEV_TIMEOUT);
 522         clear_bit(MD_CLUSTER_WAITING_FOR_NEWDISK, &cinfo->state);
 523 }
 524
 525
 526 static void process_metadata_update(struct mddev *mddev, struct cluster_msg *msg)
 527 {
 528         int got_lock = 0;
 529         struct md_cluster_info *cinfo = mddev->cluster_info;
 530         mddev->good_device_nr = le32_to_cpu(msg->raid_slot);
 531
 532         dlm_lock_sync(cinfo->no_new_dev_lockres, DLM_LOCK_CR);
 533         wait_event(mddev->thread->wqueue,
 534                    (got_lock = mddev_trylock(mddev)) ||
 535                     test_bit(MD_CLUSTER_HOLDING_MUTEX_FOR_RECVD, &cinfo->state));
 536         md_reload_sb(mddev, mddev->good_device_nr);
 537         if (got_lock)
 538                 mddev_unlock(mddev);
 539 }
 540
 541 static void process_remove_disk(struct mddev *mddev, struct cluster_msg *msg)
 542 {
 543         struct md_rdev *rdev;
 544
 545         rcu_read_lock();
 546         rdev = md_find_rdev_nr_rcu(mddev, le32_to_cpu(msg->raid_slot));
 547         if (rdev) {
 548                 set_bit(ClusterRemove, &rdev->flags);
 549                 set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
 550                 md_wakeup_thread(mddev->thread);
 551         }
 552         else
 553                 pr_warn("%s: %d Could not find disk(%d) to REMOVE\n",
 554                         __func__, __LINE__, le32_to_cpu(msg->raid_slot));
 555         rcu_read_unlock();
 556 }
 557
 558 static void process_readd_disk(struct mddev *mddev, struct cluster_msg *msg)
 559 {
 560         struct md_rdev *rdev;
 561
 562         rcu_read_lock();
 563         rdev = md_find_rdev_nr_rcu(mddev, le32_to_cpu(msg->raid_slot));
 564         if (rdev && test_bit(Faulty, &rdev->flags))
 565                 clear_bit(Faulty, &rdev->flags);
 566         else
 567                 pr_warn("%s: %d Could not find disk(%d) which is faulty",
 568                         __func__, __LINE__, le32_to_cpu(msg->raid_slot));
 569         rcu_read_unlock();
 570 }
 571
 572 static int process_recvd_msg(struct mddev *mddev, struct cluster_msg *msg)
 573 {
 574         int ret = 0;
 575
 576         if (WARN(mddev->cluster_info->slot_number - 1 == le32_to_cpu(msg->slot),
 577                 "node %d received it's own msg\n", le32_to_cpu(msg->slot)))
 578                 return -1;
 579         switch (le32_to_cpu(msg->type)) {
 580         case METADATA_UPDATED:
 581                 process_metadata_update(mddev, msg);
 582                 break;
 583         case CHANGE_CAPACITY:
 584                 set_capacity(mddev->gendisk, mddev->array_sectors);
 585                 revalidate_disk_size(mddev->gendisk, true);
 586                 break;
 587         case RESYNCING:
 588                 set_bit(MD_RESYNCING_REMOTE, &mddev->recovery);
 589                 process_suspend_info(mddev, le32_to_cpu(msg->slot),
 590                                      le64_to_cpu(msg->low),
 591                                      le64_to_cpu(msg->high));
 592                 break;
 593         case NEWDISK:
 594                 process_add_new_disk(mddev, msg);
 595                 break;
 596         case REMOVE:
 597                 process_remove_disk(mddev, msg);
 598                 break;
 599         case RE_ADD:
 600                 process_readd_disk(mddev, msg);
 601                 break;
 602         case BITMAP_NEEDS_SYNC:
 603                 __recover_slot(mddev, le32_to_cpu(msg->slot));
 604                 break;
 605         case BITMAP_RESIZE:
 606                 if (le64_to_cpu(msg->high) != mddev->pers->size(mddev, 0, 0))
 607                         ret = md_bitmap_resize(mddev->bitmap,
 608                                             le64_to_cpu(msg->high), 0, 0);
 609                 break;
 610         default:
 611                 ret = -1;
 612                 pr_warn("%s:%d Received unknown message from %d\n",
 613                         __func__, __LINE__, msg->slot);
 614         }
 615         return ret;
 616 }
 617
 618 /*
 619  * thread for receiving message
 620  */
 621 static void recv_daemon(struct md_thread *thread)
 622 {
 623         struct md_cluster_info *cinfo = thread->mddev->cluster_info;
 624         struct dlm_lock_resource *ack_lockres = cinfo->ack_lockres;
 625         struct dlm_lock_resource *message_lockres = cinfo->message_lockres;
 626         struct cluster_msg msg;
 627         int ret;
 628
 629         mutex_lock(&cinfo->recv_mutex);
 630         /*get CR on Message*/
 631         if (dlm_lock_sync(message_lockres, DLM_LOCK_CR)) {
 632                 pr_err("md/raid1:failed to get CR on MESSAGE\n");
 633                 mutex_unlock(&cinfo->recv_mutex);
 634                 return;
 635         }
 636
 637         /* read lvb and wake up thread to process this message_lockres */
 638         memcpy(&msg, message_lockres->lksb.sb_lvbptr, sizeof(struct cluster_msg));
 639         ret = process_recvd_msg(thread->mddev, &msg);
 640         if (ret)
 641                 goto out;
 642
 643         /*release CR on ack_lockres*/
 644         ret = dlm_unlock_sync(ack_lockres);
 645         if (unlikely(ret != 0))
 646                 pr_info("unlock ack failed return %d\n", ret);
 647         /*up-convert to PR on message_lockres*/
 648         ret = dlm_lock_sync(message_lockres, DLM_LOCK_PR);
 649         if (unlikely(ret != 0))
 650                 pr_info("lock PR on msg failed return %d\n", ret);
 651         /*get CR on ack_lockres again*/
 652         ret = dlm_lock_sync(ack_lockres, DLM_LOCK_CR);
 653         if (unlikely(ret != 0))
 654                 pr_info("lock CR on ack failed return %d\n", ret);
 655 out:
 656         /*release CR on message_lockres*/
 657         ret = dlm_unlock_sync(message_lockres);
 658         if (unlikely(ret != 0))
 659                 pr_info("unlock msg failed return %d\n", ret);
 660         mutex_unlock(&cinfo->recv_mutex);
 661 }
 662
 663 /* lock_token()
 664  * Takes the lock on the TOKEN lock resource so no other
 665  * node can communicate while the operation is underway.
 666  */
 667 static int lock_token(struct md_cluster_info *cinfo)
 668 {
 669         int error;
 670
 671         error = dlm_lock_sync(cinfo->token_lockres, DLM_LOCK_EX);
 672         if (error) {
 673                 pr_err("md-cluster(%s:%d): failed to get EX on TOKEN (%d)\n",
 674                                 __func__, __LINE__, error);
 675         } else {
 676                 /* Lock the receive sequence */
 677                 mutex_lock(&cinfo->recv_mutex);
 678         }
 679         return error;
 680 }
 681
 682 /* lock_comm()
 683  * Sets the MD_CLUSTER_SEND_LOCK bit to lock the send channel.
 684  */
 685 static int lock_comm(struct md_cluster_info *cinfo, bool mddev_locked)
 686 {
 687         int rv, set_bit = 0;
 688         struct mddev *mddev = cinfo->mddev;
 689
 690         /*
 691          * If resync thread run after raid1d thread, then process_metadata_update
 692          * could not continue if raid1d held reconfig_mutex (and raid1d is blocked
 693          * since another node already got EX on Token and waitting the EX of Ack),
 694          * so let resync wake up thread in case flag is set.
 695          */
 696         if (mddev_locked && !test_bit(MD_CLUSTER_HOLDING_MUTEX_FOR_RECVD,
 697                                       &cinfo->state)) {
 698                 rv = test_and_set_bit_lock(MD_CLUSTER_HOLDING_MUTEX_FOR_RECVD,
 699                                               &cinfo->state);
 700                 WARN_ON_ONCE(rv);
 701                 md_wakeup_thread(mddev->thread);
 702                 set_bit = 1;
 703         }
 704
 705         wait_event(cinfo->wait,
 706                    !test_and_set_bit(MD_CLUSTER_SEND_LOCK, &cinfo->state));
 707         rv = lock_token(cinfo);
 708         if (set_bit)
 709                 clear_bit_unlock(MD_CLUSTER_HOLDING_MUTEX_FOR_RECVD, &cinfo->state);
 710         return rv;
 711 }
 712
 713 static void unlock_comm(struct md_cluster_info *cinfo)
 714 {
 715         WARN_ON(cinfo->token_lockres->mode != DLM_LOCK_EX);
 716         mutex_unlock(&cinfo->recv_mutex);
 717         dlm_unlock_sync(cinfo->token_lockres);
 718         clear_bit(MD_CLUSTER_SEND_LOCK, &cinfo->state);
 719         wake_up(&cinfo->wait);
 720 }
 721
 722 /* __sendmsg()
 723  * This function performs the actual sending of the message. This function is
 724  * usually called after performing the encompassing operation
 725  * The function:
 726  * 1. Grabs the message lockresource in EX mode
 727  * 2. Copies the message to the message LVB
 728  * 3. Downconverts message lockresource to CW
 729  * 4. Upconverts ack lock resource from CR to EX. This forces the BAST on other nodes
 730  *    and the other nodes read the message. The thread will wait here until all other
 731  *    nodes have released ack lock resource.
 732  * 5. Downconvert ack lockresource to CR
 733  */
 734 static int __sendmsg(struct md_cluster_info *cinfo, struct cluster_msg *cmsg)
 735 {
 736         int error;
 737         int slot = cinfo->slot_number - 1;
 738
 739         cmsg->slot = cpu_to_le32(slot);
 740         /*get EX on Message*/
 741         error = dlm_lock_sync(cinfo->message_lockres, DLM_LOCK_EX);
 742         if (error) {
 743                 pr_err("md-cluster: failed to get EX on MESSAGE (%d)\n", error);
 744                 goto failed_message;
 745         }
 746
 747         memcpy(cinfo->message_lockres->lksb.sb_lvbptr, (void *)cmsg,
 748                         sizeof(struct cluster_msg));
 749         /*down-convert EX to CW on Message*/
 750         error = dlm_lock_sync(cinfo->message_lockres, DLM_LOCK_CW);
 751         if (error) {
 752                 pr_err("md-cluster: failed to convert EX to CW on MESSAGE(%d)\n",
 753                                 error);
 754                 goto failed_ack;
 755         }
 756
 757         /*up-convert CR to EX on Ack*/
 758         error = dlm_lock_sync(cinfo->ack_lockres, DLM_LOCK_EX);
 759         if (error) {
 760                 pr_err("md-cluster: failed to convert CR to EX on ACK(%d)\n",
 761                                 error);
 762                 goto failed_ack;
 763         }
 764
 765         /*down-convert EX to CR on Ack*/
 766         error = dlm_lock_sync(cinfo->ack_lockres, DLM_LOCK_CR);
 767         if (error) {
 768                 pr_err("md-cluster: failed to convert EX to CR on ACK(%d)\n",
 769                                 error);
 770                 goto failed_ack;
 771         }
 772
 773 failed_ack:
 774         error = dlm_unlock_sync(cinfo->message_lockres);
 775         if (unlikely(error != 0)) {
 776                 pr_err("md-cluster: failed convert to NL on MESSAGE(%d)\n",
 777                         error);
 778                 /* in case the message can't be released due to some reason */
 779                 goto failed_ack;
 780         }
 781 failed_message:
 782         return error;
 783 }
 784
 785 static int sendmsg(struct md_cluster_info *cinfo, struct cluster_msg *cmsg,
 786                    bool mddev_locked)
 787 {
 788         int ret;
 789
 790         ret = lock_comm(cinfo, mddev_locked);
 791         if (!ret) {
 792                 ret = __sendmsg(cinfo, cmsg);
 793                 unlock_comm(cinfo);
 794         }
 795         return ret;
 796 }
 797
 798 static int gather_all_resync_info(struct mddev *mddev, int total_slots)
 799 {
 800         struct md_cluster_info *cinfo = mddev->cluster_info;
 801         int i, ret = 0;
 802         struct dlm_lock_resource *bm_lockres;
 803         char str[64];
 804         sector_t lo, hi;
 805
 806
 807         for (i = 0; i < total_slots; i++) {
 808                 memset(str, '\0', 64);
 809                 snprintf(str, 64, "bitmap%04d", i);
 810                 bm_lockres = lockres_init(mddev, str, NULL, 1);
 811                 if (!bm_lockres)
 812                         return -ENOMEM;
 813                 if (i == (cinfo->slot_number - 1)) {
 814                         lockres_free(bm_lockres);
 815                         continue;
 816                 }
 817
 818                 bm_lockres->flags |= DLM_LKF_NOQUEUE;
 819                 ret = dlm_lock_sync(bm_lockres, DLM_LOCK_PW);
 820                 if (ret == -EAGAIN) {
 821                         if (read_resync_info(mddev, bm_lockres)) {
 822                                 pr_info("%s:%d Resync[%llu..%llu] in progress on %d\n",
 823                                                 __func__, __LINE__,
 824                                         (unsigned long long) cinfo->suspend_lo,
 825                                         (unsigned long long) cinfo->suspend_hi,
 826                                         i);
 827                                 cinfo->suspend_from = i;
 828                         }
 829                         ret = 0;
 830                         lockres_free(bm_lockres);
 831                         continue;
 832                 }
 833                 if (ret) {
 834                         lockres_free(bm_lockres);
 835                         goto out;
 836                 }
 837
 838                 /* Read the disk bitmap sb and check if it needs recovery */
 839                 ret = md_bitmap_copy_from_slot(mddev, i, &lo, &hi, false);
 840                 if (ret) {
 841                         pr_warn("md-cluster: Could not gather bitmaps from slot %d", i);
 842                         lockres_free(bm_lockres);
 843                         continue;
 844                 }
 845                 if ((hi > 0) && (lo < mddev->recovery_cp)) {
 846                         set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
 847                         mddev->recovery_cp = lo;
 848                         md_check_recovery(mddev);
 849                 }
 850
 851                 lockres_free(bm_lockres);
 852         }
 853 out:
 854         return ret;
 855 }
 856
 857 static int join(struct mddev *mddev, int nodes)
 858 {
 859         struct md_cluster_info *cinfo;
 860         int ret, ops_rv;
 861         char str[64];
 862
 863         cinfo = kzalloc(sizeof(struct md_cluster_info), GFP_KERNEL);
 864         if (!cinfo)
 865                 return -ENOMEM;
 866
 867         INIT_LIST_HEAD(&cinfo->suspend_list);
 868         spin_lock_init(&cinfo->suspend_lock);
 869         init_completion(&cinfo->completion);
 870         set_bit(MD_CLUSTER_BEGIN_JOIN_CLUSTER, &cinfo->state);
 871         init_waitqueue_head(&cinfo->wait);
 872         mutex_init(&cinfo->recv_mutex);
 873
 874         mddev->cluster_info = cinfo;
 875         cinfo->mddev = mddev;
 876
 877         memset(str, 0, 64);
 878         sprintf(str, "%pU", mddev->uuid);
 879         ret = dlm_new_lockspace(str, mddev->bitmap_info.cluster_name,
 880                                 DLM_LSFL_FS, LVB_SIZE,
 881                                 &md_ls_ops, mddev, &ops_rv, &cinfo->lockspace);
 882         if (ret)
 883                 goto err;
 884         wait_for_completion(&cinfo->completion);
 885         if (nodes < cinfo->slot_number) {
 886                 pr_err("md-cluster: Slot allotted(%d) is greater than available slots(%d).",
 887                         cinfo->slot_number, nodes);
 888                 ret = -ERANGE;
 889                 goto err;
 890         }
 891         /* Initiate the communication resources */
 892         ret = -ENOMEM;
 893         cinfo->recv_thread = md_register_thread(recv_daemon, mddev, "cluster_recv");
 894         if (!cinfo->recv_thread) {
 895                 pr_err("md-cluster: cannot allocate memory for recv_thread!\n");
 896                 goto err;
 897         }
 898         cinfo->message_lockres = lockres_init(mddev, "message", NULL, 1);
 899         if (!cinfo->message_lockres)
 900                 goto err;
 901         cinfo->token_lockres = lockres_init(mddev, "token", NULL, 0);
 902         if (!cinfo->token_lockres)
 903                 goto err;
 904         cinfo->no_new_dev_lockres = lockres_init(mddev, "no-new-dev", NULL, 0);
 905         if (!cinfo->no_new_dev_lockres)
 906                 goto err;
 907
 908         ret = dlm_lock_sync(cinfo->token_lockres, DLM_LOCK_EX);
 909         if (ret) {
 910                 ret = -EAGAIN;
 911                 pr_err("md-cluster: can't join cluster to avoid lock issue\n");
 912                 goto err;
 913         }
 914         cinfo->ack_lockres = lockres_init(mddev, "ack", ack_bast, 0);
 915         if (!cinfo->ack_lockres) {
 916                 ret = -ENOMEM;
 917                 goto err;
 918         }
 919         /* get sync CR lock on ACK. */
 920         if (dlm_lock_sync(cinfo->ack_lockres, DLM_LOCK_CR))
 921                 pr_err("md-cluster: failed to get a sync CR lock on ACK!(%d)\n",
 922                                 ret);
 923         dlm_unlock_sync(cinfo->token_lockres);
 924         /* get sync CR lock on no-new-dev. */
 925         if (dlm_lock_sync(cinfo->no_new_dev_lockres, DLM_LOCK_CR))
 926                 pr_err("md-cluster: failed to get a sync CR lock on no-new-dev!(%d)\n", ret);
 927
 928
 929         pr_info("md-cluster: Joined cluster %s slot %d\n", str, cinfo->slot_number);
 930         snprintf(str, 64, "bitmap%04d", cinfo->slot_number - 1);
 931         cinfo->bitmap_lockres = lockres_init(mddev, str, NULL, 1);
 932         if (!cinfo->bitmap_lockres) {
 933                 ret = -ENOMEM;
 934                 goto err;
 935         }
 936         if (dlm_lock_sync(cinfo->bitmap_lockres, DLM_LOCK_PW)) {
 937                 pr_err("Failed to get bitmap lock\n");
 938                 ret = -EINVAL;
 939                 goto err;
 940         }
 941
 942         cinfo->resync_lockres = lockres_init(mddev, "resync", NULL, 0);
 943         if (!cinfo->resync_lockres) {
 944                 ret = -ENOMEM;
 945                 goto err;
 946         }
 947
 948         return 0;
 949 err:
 950         set_bit(MD_CLUSTER_HOLDING_MUTEX_FOR_RECVD, &cinfo->state);
 951         md_unregister_thread(&cinfo->recovery_thread);
 952         md_unregister_thread(&cinfo->recv_thread);
 953         lockres_free(cinfo->message_lockres);
 954         lockres_free(cinfo->token_lockres);
 955         lockres_free(cinfo->ack_lockres);
 956         lockres_free(cinfo->no_new_dev_lockres);
 957         lockres_free(cinfo->resync_lockres);
 958         lockres_free(cinfo->bitmap_lockres);
 959         if (cinfo->lockspace)
 960                 dlm_release_lockspace(cinfo->lockspace, 2);
 961         mddev->cluster_info = NULL;
 962         kfree(cinfo);
 963         return ret;
 964 }
 965
 966 static void load_bitmaps(struct mddev *mddev, int total_slots)
 967 {
 968         struct md_cluster_info *cinfo = mddev->cluster_info;
 969
 970         /* load all the node's bitmap info for resync */
 971         if (gather_all_resync_info(mddev, total_slots))
 972                 pr_err("md-cluster: failed to gather all resyn infos\n");
 973         set_bit(MD_CLUSTER_ALREADY_IN_CLUSTER, &cinfo->state);
 974         /* wake up recv thread in case something need to be handled */
 975         if (test_and_clear_bit(MD_CLUSTER_PENDING_RECV_EVENT, &cinfo->state))
 976                 md_wakeup_thread(cinfo->recv_thread);
 977 }
 978
 979 static void resync_bitmap(struct mddev *mddev)
 980 {
 981         struct md_cluster_info *cinfo = mddev->cluster_info;
 982         struct cluster_msg cmsg = {0};
 983         int err;
 984
 985         cmsg.type = cpu_to_le32(BITMAP_NEEDS_SYNC);
 986         err = sendmsg(cinfo, &cmsg, 1);
 987         if (err)
 988                 pr_err("%s:%d: failed to send BITMAP_NEEDS_SYNC message (%d)\n",
 989                         __func__, __LINE__, err);
 990 }
 991
 992 static void unlock_all_bitmaps(struct mddev *mddev);
 993 static int leave(struct mddev *mddev)
 994 {
 995         struct md_cluster_info *cinfo = mddev->cluster_info;
 996
 997         if (!cinfo)
 998                 return 0;
 999
1000         /*
1001          * BITMAP_NEEDS_SYNC message should be sent when node
1002          * is leaving the cluster with dirty bitmap, also we
1003          * can only deliver it when dlm connection is available.
1004          *
1005          * Also, we should send BITMAP_NEEDS_SYNC message in
1006          * case reshaping is interrupted.
1007          */
1008         if ((cinfo->slot_number > 0 && mddev->recovery_cp != MaxSector) ||
1009             (mddev->reshape_position != MaxSector &&
1010              test_bit(MD_CLOSING, &mddev->flags)))
1011                 resync_bitmap(mddev);
1012
1013         set_bit(MD_CLUSTER_HOLDING_MUTEX_FOR_RECVD, &cinfo->state);
1014         md_unregister_thread(&cinfo->recovery_thread);
1015         md_unregister_thread(&cinfo->recv_thread);
1016         lockres_free(cinfo->message_lockres);
1017         lockres_free(cinfo->token_lockres);
1018         lockres_free(cinfo->ack_lockres);
1019         lockres_free(cinfo->no_new_dev_lockres);
1020         lockres_free(cinfo->resync_lockres);
1021         lockres_free(cinfo->bitmap_lockres);
1022         unlock_all_bitmaps(mddev);
1023         dlm_release_lockspace(cinfo->lockspace, 2);
1024         kfree(cinfo);
1025         return 0;
1026 }
1027
1028 /* slot_number(): Returns the MD slot number to use
1029  * DLM starts the slot numbers from 1, wheras cluster-md
1030  * wants the number to be from zero, so we deduct one
1031  */
1032 static int slot_number(struct mddev *mddev)
1033 {
1034         struct md_cluster_info *cinfo = mddev->cluster_info;
1035
1036         return cinfo->slot_number - 1;
1037 }
1038
1039 /*
1040  * Check if the communication is already locked, else lock the communication
1041  * channel.
1042  * If it is already locked, token is in EX mode, and hence lock_token()
1043  * should not be called.
1044  */
1045 static int metadata_update_start(struct mddev *mddev)
1046 {
1047         struct md_cluster_info *cinfo = mddev->cluster_info;
1048         int ret;
1049
1050         /*
1051          * metadata_update_start is always called with the protection of
1052          * reconfig_mutex, so set WAITING_FOR_TOKEN here.
1053          */
1054         ret = test_and_set_bit_lock(MD_CLUSTER_HOLDING_MUTEX_FOR_RECVD,
1055                                     &cinfo->state);
1056         WARN_ON_ONCE(ret);
1057         md_wakeup_thread(mddev->thread);
1058
1059         wait_event(cinfo->wait,
1060                    !test_and_set_bit(MD_CLUSTER_SEND_LOCK, &cinfo->state) ||
1061                    test_and_clear_bit(MD_CLUSTER_SEND_LOCKED_ALREADY, &cinfo->state));
1062
1063         /* If token is already locked, return 0 */
1064         if (cinfo->token_lockres->mode == DLM_LOCK_EX) {
1065                 clear_bit_unlock(MD_CLUSTER_HOLDING_MUTEX_FOR_RECVD, &cinfo->state);
1066                 return 0;
1067         }
1068
1069         ret = lock_token(cinfo);
1070         clear_bit_unlock(MD_CLUSTER_HOLDING_MUTEX_FOR_RECVD, &cinfo->state);
1071         return ret;
1072 }
1073
1074 static int metadata_update_finish(struct mddev *mddev)
1075 {
1076         struct md_cluster_info *cinfo = mddev->cluster_info;
1077         struct cluster_msg cmsg;
1078         struct md_rdev *rdev;
1079         int ret = 0;
1080         int raid_slot = -1;
1081
1082         memset(&cmsg, 0, sizeof(cmsg));
1083         cmsg.type = cpu_to_le32(METADATA_UPDATED);
1084         /* Pick up a good active device number to send.
1085          */
1086         rdev_for_each(rdev, mddev)
1087                 if (rdev->raid_disk > -1 && !test_bit(Faulty, &rdev->flags)) {
1088                         raid_slot = rdev->desc_nr;
1089                         break;
1090                 }
1091         if (raid_slot >= 0) {
1092                 cmsg.raid_slot = cpu_to_le32(raid_slot);
1093                 ret = __sendmsg(cinfo, &cmsg);
1094         } else
1095                 pr_warn("md-cluster: No good device id found to send\n");
1096         clear_bit(MD_CLUSTER_SEND_LOCKED_ALREADY, &cinfo->state);
1097         unlock_comm(cinfo);
1098         return ret;
1099 }
1100
1101 static void metadata_update_cancel(struct mddev *mddev)
1102 {
1103         struct md_cluster_info *cinfo = mddev->cluster_info;
1104         clear_bit(MD_CLUSTER_SEND_LOCKED_ALREADY, &cinfo->state);
1105         unlock_comm(cinfo);
1106 }
1107
1108 static int update_bitmap_size(struct mddev *mddev, sector_t size)
1109 {
1110         struct md_cluster_info *cinfo = mddev->cluster_info;
1111         struct cluster_msg cmsg = {0};
1112         int ret;
1113
1114         cmsg.type = cpu_to_le32(BITMAP_RESIZE);
1115         cmsg.high = cpu_to_le64(size);
1116         ret = sendmsg(cinfo, &cmsg, 0);
1117         if (ret)
1118                 pr_err("%s:%d: failed to send BITMAP_RESIZE message (%d)\n",
1119                         __func__, __LINE__, ret);
1120         return ret;
1121 }
1122
1123 static int resize_bitmaps(struct mddev *mddev, sector_t newsize, sector_t oldsize)
1124 {
1125         struct bitmap_counts *counts;
1126         char str[64];
1127         struct dlm_lock_resource *bm_lockres;
1128         struct bitmap *bitmap = mddev->bitmap;
1129         unsigned long my_pages = bitmap->counts.pages;
1130         int i, rv;
1131
1132         /*
1133          * We need to ensure all the nodes can grow to a larger
1134          * bitmap size before make the reshaping.
1135          */
1136         rv = update_bitmap_size(mddev, newsize);
1137         if (rv)
1138                 return rv;
1139
1140         for (i = 0; i < mddev->bitmap_info.nodes; i++) {
1141                 if (i == md_cluster_ops->slot_number(mddev))
1142                         continue;
1143
1144                 bitmap = get_bitmap_from_slot(mddev, i);
1145                 if (IS_ERR(bitmap)) {
1146                         pr_err("can't get bitmap from slot %d\n", i);
1147                         bitmap = NULL;
1148                         goto out;
1149                 }
1150                 counts = &bitmap->counts;
1151
1152                 /*
1153                  * If we can hold the bitmap lock of one node then
1154                  * the slot is not occupied, update the pages.
1155                  */
1156                 snprintf(str, 64, "bitmap%04d", i);
1157                 bm_lockres = lockres_init(mddev, str, NULL, 1);
1158                 if (!bm_lockres) {
1159                         pr_err("Cannot initialize %s lock\n", str);
1160                         goto out;
1161                 }
1162                 bm_lockres->flags |= DLM_LKF_NOQUEUE;
1163                 rv = dlm_lock_sync(bm_lockres, DLM_LOCK_PW);
1164                 if (!rv)
1165                         counts->pages = my_pages;
1166                 lockres_free(bm_lockres);
1167
1168                 if (my_pages != counts->pages)
1169                         /*
1170                          * Let's revert the bitmap size if one node
1171                          * can't resize bitmap
1172                          */
1173                         goto out;
1174                 md_bitmap_free(bitmap);
1175         }
1176
1177         return 0;
1178 out:
1179         md_bitmap_free(bitmap);
1180         update_bitmap_size(mddev, oldsize);
1181         return -1;
1182 }
1183
1184 /*
1185  * return 0 if all the bitmaps have the same sync_size
1186  */
1187 static int cluster_check_sync_size(struct mddev *mddev)
1188 {
1189         int i, rv;
1190         bitmap_super_t *sb;
1191         unsigned long my_sync_size, sync_size = 0;
1192         int node_num = mddev->bitmap_info.nodes;
1193         int current_slot = md_cluster_ops->slot_number(mddev);
1194         struct bitmap *bitmap = mddev->bitmap;
1195         char str[64];
1196         struct dlm_lock_resource *bm_lockres;
1197
1198         sb = kmap_atomic(bitmap->storage.sb_page);
1199         my_sync_size = sb->sync_size;
1200         kunmap_atomic(sb);
1201
1202         for (i = 0; i < node_num; i++) {
1203                 if (i == current_slot)
1204                         continue;
1205
1206                 bitmap = get_bitmap_from_slot(mddev, i);
1207                 if (IS_ERR(bitmap)) {
1208                         pr_err("can't get bitmap from slot %d\n", i);
1209                         return -1;
1210                 }
1211
1212                 /*
1213                  * If we can hold the bitmap lock of one node then
1214                  * the slot is not occupied, update the sb.
1215                  */
1216                 snprintf(str, 64, "bitmap%04d", i);
1217                 bm_lockres = lockres_init(mddev, str, NULL, 1);
1218                 if (!bm_lockres) {
1219                         pr_err("md-cluster: Cannot initialize %s\n", str);
1220                         md_bitmap_free(bitmap);
1221                         return -1;
1222                 }
1223                 bm_lockres->flags |= DLM_LKF_NOQUEUE;
1224                 rv = dlm_lock_sync(bm_lockres, DLM_LOCK_PW);
1225                 if (!rv)
1226                         md_bitmap_update_sb(bitmap);
1227                 lockres_free(bm_lockres);
1228
1229                 sb = kmap_atomic(bitmap->storage.sb_page);
1230                 if (sync_size == 0)
1231                         sync_size = sb->sync_size;
1232                 else if (sync_size != sb->sync_size) {
1233                         kunmap_atomic(sb);
1234                         md_bitmap_free(bitmap);
1235                         return -1;
1236                 }
1237                 kunmap_atomic(sb);
1238                 md_bitmap_free(bitmap);
1239         }
1240
1241         return (my_sync_size == sync_size) ? 0 : -1;
1242 }
1243
1244 /*
1245  * Update the size for cluster raid is a little more complex, we perform it
1246  * by the steps:
1247  * 1. hold token lock and update superblock in initiator node.
1248  * 2. send METADATA_UPDATED msg to other nodes.
1249  * 3. The initiator node continues to check each bitmap's sync_size, if all
1250  *    bitmaps have the same value of sync_size, then we can set capacity and
1251  *    let other nodes to perform it. If one node can't update sync_size
1252  *    accordingly, we need to revert to previous value.
1253  */
1254 static void update_size(struct mddev *mddev, sector_t old_dev_sectors)
1255 {
1256         struct md_cluster_info *cinfo = mddev->cluster_info;
1257         struct cluster_msg cmsg;
1258         struct md_rdev *rdev;
1259         int ret = 0;
1260         int raid_slot = -1;
1261
1262         md_update_sb(mddev, 1);
1263         if (lock_comm(cinfo, 1)) {
1264                 pr_err("%s: lock_comm failed\n", __func__);
1265                 return;
1266         }
1267
1268         memset(&cmsg, 0, sizeof(cmsg));
1269         cmsg.type = cpu_to_le32(METADATA_UPDATED);
1270         rdev_for_each(rdev, mddev)
1271                 if (rdev->raid_disk >= 0 && !test_bit(Faulty, &rdev->flags)) {
1272                         raid_slot = rdev->desc_nr;
1273                         break;
1274                 }
1275         if (raid_slot >= 0) {
1276                 cmsg.raid_slot = cpu_to_le32(raid_slot);
1277                 /*
1278                  * We can only change capiticy after all the nodes can do it,
1279                  * so need to wait after other nodes already received the msg
1280                  * and handled the change
1281                  */
1282                 ret = __sendmsg(cinfo, &cmsg);
1283                 if (ret) {
1284                         pr_err("%s:%d: failed to send METADATA_UPDATED msg\n",
1285                                __func__, __LINE__);
1286                         unlock_comm(cinfo);
1287                         return;
1288                 }
1289         } else {
1290                 pr_err("md-cluster: No good device id found to send\n");
1291                 unlock_comm(cinfo);
1292                 return;
1293         }
1294
1295         /*
1296          * check the sync_size from other node's bitmap, if sync_size
1297          * have already updated in other nodes as expected, send an
1298          * empty metadata msg to permit the change of capacity
1299          */
1300         if (cluster_check_sync_size(mddev) == 0) {
1301                 memset(&cmsg, 0, sizeof(cmsg));
1302                 cmsg.type = cpu_to_le32(CHANGE_CAPACITY);
1303                 ret = __sendmsg(cinfo, &cmsg);
1304                 if (ret)
1305                         pr_err("%s:%d: failed to send CHANGE_CAPACITY msg\n",
1306                                __func__, __LINE__);
1307                 set_capacity(mddev->gendisk, mddev->array_sectors);
1308                 revalidate_disk_size(mddev->gendisk, true);
1309         } else {
1310                 /* revert to previous sectors */
1311                 ret = mddev->pers->resize(mddev, old_dev_sectors);
1312                 if (!ret)
1313                         revalidate_disk_size(mddev->gendisk, true);
1314                 ret = __sendmsg(cinfo, &cmsg);
1315                 if (ret)
1316                         pr_err("%s:%d: failed to send METADATA_UPDATED msg\n",
1317                                __func__, __LINE__);
1318         }
1319         unlock_comm(cinfo);
1320 }
1321
1322 static int resync_start(struct mddev *mddev)
1323 {
1324         struct md_cluster_info *cinfo = mddev->cluster_info;
1325         return dlm_lock_sync_interruptible(cinfo->resync_lockres, DLM_LOCK_EX, mddev);
1326 }
1327
1328 static void resync_info_get(struct mddev *mddev, sector_t *lo, sector_t *hi)
1329 {
1330         struct md_cluster_info *cinfo = mddev->cluster_info;
1331
1332         spin_lock_irq(&cinfo->suspend_lock);
1333         *lo = cinfo->suspend_lo;
1334         *hi = cinfo->suspend_hi;
1335         spin_unlock_irq(&cinfo->suspend_lock);
1336 }
1337
1338 static int resync_info_update(struct mddev *mddev, sector_t lo, sector_t hi)
1339 {
1340         struct md_cluster_info *cinfo = mddev->cluster_info;
1341         struct resync_info ri;
1342         struct cluster_msg cmsg = {0};
1343
1344         /* do not send zero again, if we have sent before */
1345         if (hi == 0) {
1346                 memcpy(&ri, cinfo->bitmap_lockres->lksb.sb_lvbptr, sizeof(struct resync_info));
1347                 if (le64_to_cpu(ri.hi) == 0)
1348                         return 0;
1349         }
1350
1351         add_resync_info(cinfo->bitmap_lockres, lo, hi);
1352         /* Re-acquire the lock to refresh LVB */
1353         dlm_lock_sync(cinfo->bitmap_lockres, DLM_LOCK_PW);
1354         cmsg.type = cpu_to_le32(RESYNCING);
1355         cmsg.low = cpu_to_le64(lo);
1356         cmsg.high = cpu_to_le64(hi);
1357
1358         /*
1359          * mddev_lock is held if resync_info_update is called from
1360          * resync_finish (md_reap_sync_thread -> resync_finish)
1361          */
1362         if (lo == 0 && hi == 0)
1363                 return sendmsg(cinfo, &cmsg, 1);
1364         else
1365                 return sendmsg(cinfo, &cmsg, 0);
1366 }
1367
1368 static int resync_finish(struct mddev *mddev)
1369 {
1370         struct md_cluster_info *cinfo = mddev->cluster_info;
1371         int ret = 0;
1372
1373         clear_bit(MD_RESYNCING_REMOTE, &mddev->recovery);
1374
1375         /*
1376          * If resync thread is interrupted so we can't say resync is finished,
1377          * another node will launch resync thread to continue.
1378          */
1379         if (!test_bit(MD_CLOSING, &mddev->flags))
1380                 ret = resync_info_update(mddev, 0, 0);
1381         dlm_unlock_sync(cinfo->resync_lockres);
1382         return ret;
1383 }
1384
1385 static int area_resyncing(struct mddev *mddev, int direction,
1386                 sector_t lo, sector_t hi)
1387 {
1388         struct md_cluster_info *cinfo = mddev->cluster_info;
1389         int ret = 0;
1390
1391         if ((direction == READ) &&
1392                 test_bit(MD_CLUSTER_SUSPEND_READ_BALANCING, &cinfo->state))
1393                 return 1;
1394
1395         spin_lock_irq(&cinfo->suspend_lock);
1396         if (hi > cinfo->suspend_lo && lo < cinfo->suspend_hi)
1397                 ret = 1;
1398         spin_unlock_irq(&cinfo->suspend_lock);
1399         return ret;
1400 }
1401
1402 /* add_new_disk() - initiates a disk add
1403  * However, if this fails before writing md_update_sb(),
1404  * add_new_disk_cancel() must be called to release token lock
1405  */
1406 static int add_new_disk(struct mddev *mddev, struct md_rdev *rdev)
1407 {
1408         struct md_cluster_info *cinfo = mddev->cluster_info;
1409         struct cluster_msg cmsg;
1410         int ret = 0;
1411         struct mdp_superblock_1 *sb = page_address(rdev->sb_page);
1412         char *uuid = sb->device_uuid;
1413
1414         memset(&cmsg, 0, sizeof(cmsg));
1415         cmsg.type = cpu_to_le32(NEWDISK);
1416         memcpy(cmsg.uuid, uuid, 16);
1417         cmsg.raid_slot = cpu_to_le32(rdev->desc_nr);
1418         if (lock_comm(cinfo, 1))
1419                 return -EAGAIN;
1420         ret = __sendmsg(cinfo, &cmsg);
1421         if (ret) {
1422                 unlock_comm(cinfo);
1423                 return ret;
1424         }
1425         cinfo->no_new_dev_lockres->flags |= DLM_LKF_NOQUEUE;
1426         ret = dlm_lock_sync(cinfo->no_new_dev_lockres, DLM_LOCK_EX);
1427         cinfo->no_new_dev_lockres->flags &= ~DLM_LKF_NOQUEUE;
1428         /* Some node does not "see" the device */
1429         if (ret == -EAGAIN)
1430                 ret = -ENOENT;
1431         if (ret)
1432                 unlock_comm(cinfo);
1433         else {
1434                 dlm_lock_sync(cinfo->no_new_dev_lockres, DLM_LOCK_CR);
1435                 /* Since MD_CHANGE_DEVS will be set in add_bound_rdev which
1436                  * will run soon after add_new_disk, the below path will be
1437                  * invoked:
1438                  *   md_wakeup_thread(mddev->thread)
1439                  *      -> conf->thread (raid1d)
1440                  *      -> md_check_recovery -> md_update_sb
1441                  *      -> metadata_update_start/finish
1442                  * MD_CLUSTER_SEND_LOCKED_ALREADY will be cleared eventually.
1443                  *
1444                  * For other failure cases, metadata_update_cancel and
1445                  * add_new_disk_cancel also clear below bit as well.
1446                  * */
1447                 set_bit(MD_CLUSTER_SEND_LOCKED_ALREADY, &cinfo->state);
1448                 wake_up(&cinfo->wait);
1449         }
1450         return ret;
1451 }
1452
1453 static void add_new_disk_cancel(struct mddev *mddev)
1454 {
1455         struct md_cluster_info *cinfo = mddev->cluster_info;
1456         clear_bit(MD_CLUSTER_SEND_LOCKED_ALREADY, &cinfo->state);
1457         unlock_comm(cinfo);
1458 }
1459
1460 static int new_disk_ack(struct mddev *mddev, bool ack)
1461 {
1462         struct md_cluster_info *cinfo = mddev->cluster_info;
1463
1464         if (!test_bit(MD_CLUSTER_WAITING_FOR_NEWDISK, &cinfo->state)) {
1465                 pr_warn("md-cluster(%s): Spurious cluster confirmation\n", mdname(mddev));
1466                 return -EINVAL;
1467         }
1468
1469         if (ack)
1470                 dlm_unlock_sync(cinfo->no_new_dev_lockres);
1471         complete(&cinfo->newdisk_completion);
1472         return 0;
1473 }
1474
1475 static int remove_disk(struct mddev *mddev, struct md_rdev *rdev)
1476 {
1477         struct cluster_msg cmsg = {0};
1478         struct md_cluster_info *cinfo = mddev->cluster_info;
1479         cmsg.type = cpu_to_le32(REMOVE);
1480         cmsg.raid_slot = cpu_to_le32(rdev->desc_nr);
1481         return sendmsg(cinfo, &cmsg, 1);
1482 }
1483
1484 static int lock_all_bitmaps(struct mddev *mddev)
1485 {
1486         int slot, my_slot, ret, held = 1, i = 0;
1487         char str[64];
1488         struct md_cluster_info *cinfo = mddev->cluster_info;
1489
1490         cinfo->other_bitmap_lockres =
1491                 kcalloc(mddev->bitmap_info.nodes - 1,
1492                         sizeof(struct dlm_lock_resource *), GFP_KERNEL);
1493         if (!cinfo->other_bitmap_lockres) {
1494                 pr_err("md: can't alloc mem for other bitmap locks\n");
1495                 return 0;
1496         }
1497
1498         my_slot = slot_number(mddev);
1499         for (slot = 0; slot < mddev->bitmap_info.nodes; slot++) {
1500                 if (slot == my_slot)
1501                         continue;
1502
1503                 memset(str, '\0', 64);
1504                 snprintf(str, 64, "bitmap%04d", slot);
1505                 cinfo->other_bitmap_lockres[i] = lockres_init(mddev, str, NULL, 1);
1506                 if (!cinfo->other_bitmap_lockres[i])
1507                         return -ENOMEM;
1508
1509                 cinfo->other_bitmap_lockres[i]->flags |= DLM_LKF_NOQUEUE;
1510                 ret = dlm_lock_sync(cinfo->other_bitmap_lockres[i], DLM_LOCK_PW);
1511                 if (ret)
1512                         held = -1;
1513                 i++;
1514         }
1515
1516         return held;
1517 }
1518
1519 static void unlock_all_bitmaps(struct mddev *mddev)
1520 {
1521         struct md_cluster_info *cinfo = mddev->cluster_info;
1522         int i;
1523
1524         /* release other node's bitmap lock if they are existed */
1525         if (cinfo->other_bitmap_lockres) {
1526                 for (i = 0; i < mddev->bitmap_info.nodes - 1; i++) {
1527                         if (cinfo->other_bitmap_lockres[i]) {
1528                                 lockres_free(cinfo->other_bitmap_lockres[i]);
1529                         }
1530                 }
1531                 kfree(cinfo->other_bitmap_lockres);
1532                 cinfo->other_bitmap_lockres = NULL;
1533         }
1534 }
1535
1536 static int gather_bitmaps(struct md_rdev *rdev)
1537 {
1538         int sn, err;
1539         sector_t lo, hi;
1540         struct cluster_msg cmsg = {0};
1541         struct mddev *mddev = rdev->mddev;
1542         struct md_cluster_info *cinfo = mddev->cluster_info;
1543
1544         cmsg.type = cpu_to_le32(RE_ADD);
1545         cmsg.raid_slot = cpu_to_le32(rdev->desc_nr);
1546         err = sendmsg(cinfo, &cmsg, 1);
1547         if (err)
1548                 goto out;
1549
1550         for (sn = 0; sn < mddev->bitmap_info.nodes; sn++) {
1551                 if (sn == (cinfo->slot_number - 1))
1552                         continue;
1553                 err = md_bitmap_copy_from_slot(mddev, sn, &lo, &hi, false);
1554                 if (err) {
1555                         pr_warn("md-cluster: Could not gather bitmaps from slot %d", sn);
1556                         goto out;
1557                 }
1558                 if ((hi > 0) && (lo < mddev->recovery_cp))
1559                         mddev->recovery_cp = lo;
1560         }
1561 out:
1562         return err;
1563 }
1564
1565 static struct md_cluster_operations cluster_ops = {
1566         .join   = join,
1567         .leave  = leave,
1568         .slot_number = slot_number,
1569         .resync_start = resync_start,
1570         .resync_finish = resync_finish,
1571         .resync_info_update = resync_info_update,
1572         .resync_info_get = resync_info_get,
1573         .metadata_update_start = metadata_update_start,
1574         .metadata_update_finish = metadata_update_finish,
1575         .metadata_update_cancel = metadata_update_cancel,
1576         .area_resyncing = area_resyncing,
1577         .add_new_disk = add_new_disk,
1578         .add_new_disk_cancel = add_new_disk_cancel,
1579         .new_disk_ack = new_disk_ack,
1580         .remove_disk = remove_disk,
1581         .load_bitmaps = load_bitmaps,
1582         .gather_bitmaps = gather_bitmaps,
1583         .resize_bitmaps = resize_bitmaps,
1584         .lock_all_bitmaps = lock_all_bitmaps,
1585         .unlock_all_bitmaps = unlock_all_bitmaps,
1586         .update_size = update_size,
1587 };
1588
1589 static int __init cluster_init(void)
1590 {
1591         pr_warn("md-cluster: support raid1 and raid10 (limited support)\n");
1592         pr_info("Registering Cluster MD functions\n");
1593         register_md_cluster_operations(&cluster_ops, THIS_MODULE);
1594         return 0;
1595 }
1596
1597 static void cluster_exit(void)
1598 {
1599         unregister_md_cluster_operations();
1600 }
1601
1602 module_init(cluster_init);
1603 module_exit(cluster_exit);
1604 MODULE_AUTHOR("SUSE");
1605 MODULE_LICENSE("GPL");
1606 MODULE_DESCRIPTION("Clustering support for MD");