fs/dlm/midcomms.c

   1 // SPDX-License-Identifier: GPL-2.0-only
   2 /******************************************************************************
   3 *******************************************************************************
   4 **
   5 **  Copyright (C) Sistina Software, Inc.  1997-2003  All rights reserved.
   6 **  Copyright (C) 2004-2021 Red Hat, Inc.  All rights reserved.
   7 **
   8 **
   9 *******************************************************************************
  10 ******************************************************************************/
  11
  12 /*
  13  * midcomms.c
  14  *
  15  * This is the appallingly named "mid-level" comms layer. It takes care about
  16  * deliver an on application layer "reliable" communication above the used
  17  * lowcomms transport layer.
  18  *
  19  * How it works:
  20  *
  21  * Each nodes keeps track of all send DLM messages in send_queue with a sequence
  22  * number. The receive will send an DLM_ACK message back for every DLM message
  23  * received at the other side. If a reconnect happens in lowcomms we will send
  24  * all unacknowledged dlm messages again. The receiving side might drop any already
  25  * received message by comparing sequence numbers.
  26  *
  27  * How version detection works:
  28  *
  29  * Due the fact that dlm has pre-configured node addresses on every side
  30  * it is in it's nature that every side connects at starts to transmit
  31  * dlm messages which ends in a race. However DLM_RCOM_NAMES, DLM_RCOM_STATUS
  32  * and their replies are the first messages which are exchanges. Due backwards
  33  * compatibility these messages are not covered by the midcomms re-transmission
  34  * layer. These messages have their own re-transmission handling in the dlm
  35  * application layer. The version field of every node will be set on these RCOM
  36  * messages as soon as they arrived and the node isn't yet part of the nodes
  37  * hash. There exists also logic to detect version mismatched if something weird
  38  * going on or the first messages isn't an expected one.
  39  *
  40  * Termination:
  41  *
  42  * The midcomms layer does a 4 way handshake for termination on DLM protocol
  43  * like TCP supports it with half-closed socket support. SCTP doesn't support
  44  * half-closed socket, so we do it on DLM layer. Also socket shutdown() can be
  45  * interrupted by .e.g. tcp reset itself. Additional there exists the othercon
  46  * paradigm in lowcomms which cannot be easily without breaking backwards
  47  * compatibility. A node cannot send anything to another node when a DLM_FIN
  48  * message was send. There exists additional logic to print a warning if
  49  * DLM wants to do it. There exists a state handling like RFC 793 but reduced
  50  * to termination only. The event "member removal event" describes the cluster
  51  * manager removed the node from internal lists, at this point DLM does not
  52  * send any message to the other node. There exists two cases:
  53  *
  54  * 1. The cluster member was removed and we received a FIN
  55  * OR
  56  * 2. We received a FIN but the member was not removed yet
  57  *
  58  * One of these cases will do the CLOSE_WAIT to LAST_ACK change.
  59  *
  60  *
  61  *                              +---------+
  62  *                              | CLOSED  |
  63  *                              +---------+
  64  *                                   | add member/receive RCOM version
  65  *                                   |            detection msg
  66  *                                   V
  67  *                              +---------+
  68  *                              |  ESTAB  |
  69  *                              +---------+
  70  *                       CLOSE    |     |    rcv FIN
  71  *                      -------   |     |    -------
  72  * +---------+          snd FIN  /       \   snd ACK          +---------+
  73  * |  FIN    |<-----------------           ------------------>|  CLOSE  |
  74  * | WAIT-1  |------------------                              |   WAIT  |
  75  * +---------+          rcv FIN  \                            +---------+
  76  * | rcv ACK of FIN   -------   |                            CLOSE  | member
  77  * | --------------   snd ACK   |                           ------- | removal
  78  * V        x                   V                           snd FIN V event
  79  * +---------+                  +---------+                   +---------+
  80  * |FINWAIT-2|                  | CLOSING |                   | LAST-ACK|
  81  * +---------+                  +---------+                   +---------+
  82  * |                rcv ACK of FIN |                 rcv ACK of FIN |
  83  * |  rcv FIN       -------------- |                 -------------- |
  84  * |  -------              x       V                        x       V
  85  *  \ snd ACK                 +---------+                   +---------+
  86  *   ------------------------>| CLOSED  |                   | CLOSED  |
  87  *                            +---------+                   +---------+
  88  *
  89  * NOTE: any state can interrupted by midcomms_close() and state will be
  90  * switched to CLOSED in case of fencing. There exists also some timeout
  91  * handling when we receive the version detection RCOM messages which is
  92  * made by observation.
  93  *
  94  * Future improvements:
  95  *
  96  * There exists some known issues/improvements of the dlm handling. Some
  97  * of them should be done in a next major dlm version bump which makes
  98  * it incompatible with previous versions.
  99  *
 100  * Unaligned memory access:
 101  *
 102  * There exists cases when the dlm message buffer length is not aligned
 103  * to 8 byte. However seems nobody detected any problem with it. This
 104  * can be fixed in the next major version bump of dlm.
 105  *
 106  * Version detection:
 107  *
 108  * The version detection and how it's done is related to backwards
 109  * compatibility. There exists better ways to make a better handling.
 110  * However this should be changed in the next major version bump of dlm.
 111  *
 112  * Tail Size checking:
 113  *
 114  * There exists a message tail payload in e.g. DLM_MSG however we don't
 115  * check it against the message length yet regarding to the receive buffer
 116  * length. That need to be validated.
 117  *
 118  * Fencing bad nodes:
 119  *
 120  * At timeout places or weird sequence number behaviours we should send
 121  * a fencing request to the cluster manager.
 122  */
 123
 124 /* Debug switch to enable a 5 seconds sleep waiting of a termination.
 125  * This can be useful to test fencing while termination is running.
 126  * This requires a setup with only gfs2 as dlm user, so that the
 127  * last umount will terminate the connection.
 128  *
 129  * However it became useful to test, while the 5 seconds block in umount
 130  * just press the reset button. In a lot of dropping the termination
 131  * process can could take several seconds.
 132  */
 133 #define DLM_DEBUG_FENCE_TERMINATION     0
 134
 135 #include <trace/events/dlm.h>
 136 #include <net/tcp.h>
 137
 138 #include "dlm_internal.h"
 139 #include "lockspace.h"
 140 #include "lowcomms.h"
 141 #include "config.h"
 142 #include "memory.h"
 143 #include "lock.h"
 144 #include "util.h"
 145 #include "midcomms.h"
 146
 147 /* init value for sequence numbers for testing purpose only e.g. overflows */
 148 #define DLM_SEQ_INIT            0
 149 /* 3 minutes wait to sync ending of dlm */
 150 #define DLM_SHUTDOWN_TIMEOUT    msecs_to_jiffies(3 * 60 * 1000)
 151 #define DLM_VERSION_NOT_SET     0
 152
 153 struct midcomms_node {
 154         int nodeid;
 155         uint32_t version;
 156         uint32_t seq_send;
 157         uint32_t seq_next;
 158         /* These queues are unbound because we cannot drop any message in dlm.
 159          * We could send a fence signal for a specific node to the cluster
 160          * manager if queues hits some maximum value, however this handling
 161          * not supported yet.
 162          */
 163         struct list_head send_queue;
 164         spinlock_t send_queue_lock;
 165         atomic_t send_queue_cnt;
 166 #define DLM_NODE_FLAG_CLOSE     1
 167 #define DLM_NODE_FLAG_STOP_TX   2
 168 #define DLM_NODE_FLAG_STOP_RX   3
 169 #define DLM_NODE_ULP_DELIVERED  4
 170         unsigned long flags;
 171         wait_queue_head_t shutdown_wait;
 172
 173         /* dlm tcp termination state */
 174 #define DLM_CLOSED      1
 175 #define DLM_ESTABLISHED 2
 176 #define DLM_FIN_WAIT1   3
 177 #define DLM_FIN_WAIT2   4
 178 #define DLM_CLOSE_WAIT  5
 179 #define DLM_LAST_ACK    6
 180 #define DLM_CLOSING     7
 181         int state;
 182         spinlock_t state_lock;
 183
 184         /* counts how many lockspaces are using this node
 185          * this refcount is necessary to determine if the
 186          * node wants to disconnect.
 187          */
 188         int users;
 189
 190         /* not protected by srcu, node_hash lifetime */
 191         void *debugfs;
 192
 193         struct hlist_node hlist;
 194         struct rcu_head rcu;
 195 };
 196
 197 struct dlm_mhandle {
 198         const union dlm_packet *inner_p;
 199         struct midcomms_node *node;
 200         struct dlm_opts *opts;
 201         struct dlm_msg *msg;
 202         bool committed;
 203         uint32_t seq;
 204
 205         void (*ack_rcv)(struct midcomms_node *node);
 206
 207         /* get_mhandle/commit srcu idx exchange */
 208         int idx;
 209
 210         struct list_head list;
 211         struct rcu_head rcu;
 212 };
 213
 214 static struct hlist_head node_hash[CONN_HASH_SIZE];
 215 static DEFINE_SPINLOCK(nodes_lock);
 216 DEFINE_STATIC_SRCU(nodes_srcu);
 217
 218 /* This mutex prevents that midcomms_close() is running while
 219  * stop() or remove(). As I experienced invalid memory access
 220  * behaviours when DLM_DEBUG_FENCE_TERMINATION is enabled and
 221  * resetting machines. I will end in some double deletion in nodes
 222  * datastructure.
 223  */
 224 static DEFINE_MUTEX(close_lock);
 225
 226 struct kmem_cache *dlm_midcomms_cache_create(void)
 227 {
 228         return kmem_cache_create("dlm_mhandle", sizeof(struct dlm_mhandle),
 229                                  0, 0, NULL);
 230 }
 231
 232 static inline const char *dlm_state_str(int state)
 233 {
 234         switch (state) {
 235         case DLM_CLOSED:
 236                 return "CLOSED";
 237         case DLM_ESTABLISHED:
 238                 return "ESTABLISHED";
 239         case DLM_FIN_WAIT1:
 240                 return "FIN_WAIT1";
 241         case DLM_FIN_WAIT2:
 242                 return "FIN_WAIT2";
 243         case DLM_CLOSE_WAIT:
 244                 return "CLOSE_WAIT";
 245         case DLM_LAST_ACK:
 246                 return "LAST_ACK";
 247         case DLM_CLOSING:
 248                 return "CLOSING";
 249         default:
 250                 return "UNKNOWN";
 251         }
 252 }
 253
 254 const char *dlm_midcomms_state(struct midcomms_node *node)
 255 {
 256         return dlm_state_str(node->state);
 257 }
 258
 259 unsigned long dlm_midcomms_flags(struct midcomms_node *node)
 260 {
 261         return node->flags;
 262 }
 263
 264 int dlm_midcomms_send_queue_cnt(struct midcomms_node *node)
 265 {
 266         return atomic_read(&node->send_queue_cnt);
 267 }
 268
 269 uint32_t dlm_midcomms_version(struct midcomms_node *node)
 270 {
 271         return node->version;
 272 }
 273
 274 static struct midcomms_node *__find_node(int nodeid, int r)
 275 {
 276         struct midcomms_node *node;
 277
 278         hlist_for_each_entry_rcu(node, &node_hash[r], hlist) {
 279                 if (node->nodeid == nodeid)
 280                         return node;
 281         }
 282
 283         return NULL;
 284 }
 285
 286 static void dlm_mhandle_release(struct rcu_head *rcu)
 287 {
 288         struct dlm_mhandle *mh = container_of(rcu, struct dlm_mhandle, rcu);
 289
 290         dlm_lowcomms_put_msg(mh->msg);
 291         dlm_free_mhandle(mh);
 292 }
 293
 294 static void dlm_mhandle_delete(struct midcomms_node *node,
 295                                struct dlm_mhandle *mh)
 296 {
 297         list_del_rcu(&mh->list);
 298         atomic_dec(&node->send_queue_cnt);
 299         call_rcu(&mh->rcu, dlm_mhandle_release);
 300 }
 301
 302 static void dlm_send_queue_flush(struct midcomms_node *node)
 303 {
 304         struct dlm_mhandle *mh;
 305
 306         pr_debug("flush midcomms send queue of node %d\n", node->nodeid);
 307
 308         rcu_read_lock();
 309         spin_lock_bh(&node->send_queue_lock);
 310         list_for_each_entry_rcu(mh, &node->send_queue, list) {
 311                 dlm_mhandle_delete(node, mh);
 312         }
 313         spin_unlock_bh(&node->send_queue_lock);
 314         rcu_read_unlock();
 315 }
 316
 317 static void midcomms_node_reset(struct midcomms_node *node)
 318 {
 319         pr_debug("reset node %d\n", node->nodeid);
 320
 321         node->seq_next = DLM_SEQ_INIT;
 322         node->seq_send = DLM_SEQ_INIT;
 323         node->version = DLM_VERSION_NOT_SET;
 324         node->flags = 0;
 325
 326         dlm_send_queue_flush(node);
 327         node->state = DLM_CLOSED;
 328         wake_up(&node->shutdown_wait);
 329 }
 330
 331 static struct midcomms_node *nodeid2node(int nodeid, gfp_t alloc)
 332 {
 333         struct midcomms_node *node, *tmp;
 334         int r = nodeid_hash(nodeid);
 335
 336         node = __find_node(nodeid, r);
 337         if (node || !alloc)
 338                 return node;
 339
 340         node = kmalloc(sizeof(*node), alloc);
 341         if (!node)
 342                 return NULL;
 343
 344         node->nodeid = nodeid;
 345         spin_lock_init(&node->state_lock);
 346         spin_lock_init(&node->send_queue_lock);
 347         atomic_set(&node->send_queue_cnt, 0);
 348         INIT_LIST_HEAD(&node->send_queue);
 349         init_waitqueue_head(&node->shutdown_wait);
 350         node->users = 0;
 351         midcomms_node_reset(node);
 352
 353         spin_lock(&nodes_lock);
 354         /* check again if there was somebody else
 355          * earlier here to add the node
 356          */
 357         tmp = __find_node(nodeid, r);
 358         if (tmp) {
 359                 spin_unlock(&nodes_lock);
 360                 kfree(node);
 361                 return tmp;
 362         }
 363
 364         hlist_add_head_rcu(&node->hlist, &node_hash[r]);
 365         spin_unlock(&nodes_lock);
 366
 367         node->debugfs = dlm_create_debug_comms_file(nodeid, node);
 368         return node;
 369 }
 370
 371 static int dlm_send_ack(int nodeid, uint32_t seq)
 372 {
 373         int mb_len = sizeof(struct dlm_header);
 374         struct dlm_header *m_header;
 375         struct dlm_msg *msg;
 376         char *ppc;
 377
 378         msg = dlm_lowcomms_new_msg(nodeid, mb_len, GFP_NOFS, &ppc,
 379                                    NULL, NULL);
 380         if (!msg)
 381                 return -ENOMEM;
 382
 383         m_header = (struct dlm_header *)ppc;
 384
 385         m_header->h_version = cpu_to_le32(DLM_HEADER_MAJOR | DLM_HEADER_MINOR);
 386         m_header->h_nodeid = cpu_to_le32(dlm_our_nodeid());
 387         m_header->h_length = cpu_to_le16(mb_len);
 388         m_header->h_cmd = DLM_ACK;
 389         m_header->u.h_seq = cpu_to_le32(seq);
 390
 391         dlm_lowcomms_commit_msg(msg);
 392         dlm_lowcomms_put_msg(msg);
 393
 394         return 0;
 395 }
 396
 397 static int dlm_send_fin(struct midcomms_node *node,
 398                         void (*ack_rcv)(struct midcomms_node *node))
 399 {
 400         int mb_len = sizeof(struct dlm_header);
 401         struct dlm_header *m_header;
 402         struct dlm_mhandle *mh;
 403         char *ppc;
 404
 405         mh = dlm_midcomms_get_mhandle(node->nodeid, mb_len, GFP_NOFS, &ppc);
 406         if (!mh)
 407                 return -ENOMEM;
 408
 409         mh->ack_rcv = ack_rcv;
 410
 411         m_header = (struct dlm_header *)ppc;
 412
 413         m_header->h_version = cpu_to_le32(DLM_HEADER_MAJOR | DLM_HEADER_MINOR);
 414         m_header->h_nodeid = cpu_to_le32(dlm_our_nodeid());
 415         m_header->h_length = cpu_to_le16(mb_len);
 416         m_header->h_cmd = DLM_FIN;
 417
 418         pr_debug("sending fin msg to node %d\n", node->nodeid);
 419         dlm_midcomms_commit_mhandle(mh, NULL, 0);
 420         set_bit(DLM_NODE_FLAG_STOP_TX, &node->flags);
 421
 422         return 0;
 423 }
 424
 425 static void dlm_receive_ack(struct midcomms_node *node, uint32_t seq)
 426 {
 427         struct dlm_mhandle *mh;
 428
 429         rcu_read_lock();
 430         list_for_each_entry_rcu(mh, &node->send_queue, list) {
 431                 if (before(mh->seq, seq)) {
 432                         if (mh->ack_rcv)
 433                                 mh->ack_rcv(node);
 434                 } else {
 435                         /* send queue should be ordered */
 436                         break;
 437                 }
 438         }
 439
 440         spin_lock_bh(&node->send_queue_lock);
 441         list_for_each_entry_rcu(mh, &node->send_queue, list) {
 442                 if (before(mh->seq, seq)) {
 443                         dlm_mhandle_delete(node, mh);
 444                 } else {
 445                         /* send queue should be ordered */
 446                         break;
 447                 }
 448         }
 449         spin_unlock_bh(&node->send_queue_lock);
 450         rcu_read_unlock();
 451 }
 452
 453 static void dlm_pas_fin_ack_rcv(struct midcomms_node *node)
 454 {
 455         spin_lock(&node->state_lock);
 456         pr_debug("receive passive fin ack from node %d with state %s\n",
 457                  node->nodeid, dlm_state_str(node->state));
 458
 459         switch (node->state) {
 460         case DLM_LAST_ACK:
 461                 /* DLM_CLOSED */
 462                 midcomms_node_reset(node);
 463                 break;
 464         case DLM_CLOSED:
 465                 /* not valid but somehow we got what we want */
 466                 wake_up(&node->shutdown_wait);
 467                 break;
 468         default:
 469                 spin_unlock(&node->state_lock);
 470                 log_print("%s: unexpected state: %d\n",
 471                           __func__, node->state);
 472                 WARN_ON_ONCE(1);
 473                 return;
 474         }
 475         spin_unlock(&node->state_lock);
 476 }
 477
 478 static void dlm_receive_buffer_3_2_trace(uint32_t seq, union dlm_packet *p)
 479 {
 480         switch (p->header.h_cmd) {
 481         case DLM_MSG:
 482                 trace_dlm_recv_message(dlm_our_nodeid(), seq, &p->message);
 483                 break;
 484         case DLM_RCOM:
 485                 trace_dlm_recv_rcom(dlm_our_nodeid(), seq, &p->rcom);
 486                 break;
 487         default:
 488                 break;
 489         }
 490 }
 491
 492 static void dlm_midcomms_receive_buffer(union dlm_packet *p,
 493                                         struct midcomms_node *node,
 494                                         uint32_t seq)
 495 {
 496         if (seq == node->seq_next) {
 497                 node->seq_next++;
 498
 499                 switch (p->header.h_cmd) {
 500                 case DLM_FIN:
 501                         /* send ack before fin */
 502                         dlm_send_ack(node->nodeid, node->seq_next);
 503
 504                         spin_lock(&node->state_lock);
 505                         pr_debug("receive fin msg from node %d with state %s\n",
 506                                  node->nodeid, dlm_state_str(node->state));
 507
 508                         switch (node->state) {
 509                         case DLM_ESTABLISHED:
 510                                 node->state = DLM_CLOSE_WAIT;
 511                                 pr_debug("switch node %d to state %s\n",
 512                                          node->nodeid, dlm_state_str(node->state));
 513                                 /* passive shutdown DLM_LAST_ACK case 1
 514                                  * additional we check if the node is used by
 515                                  * cluster manager events at all.
 516                                  */
 517                                 if (node->users == 0) {
 518                                         node->state = DLM_LAST_ACK;
 519                                         pr_debug("switch node %d to state %s case 1\n",
 520                                                  node->nodeid, dlm_state_str(node->state));
 521                                         spin_unlock(&node->state_lock);
 522                                         goto send_fin;
 523                                 }
 524                                 break;
 525                         case DLM_FIN_WAIT1:
 526                                 node->state = DLM_CLOSING;
 527                                 pr_debug("switch node %d to state %s\n",
 528                                          node->nodeid, dlm_state_str(node->state));
 529                                 break;
 530                         case DLM_FIN_WAIT2:
 531                                 midcomms_node_reset(node);
 532                                 pr_debug("switch node %d to state %s\n",
 533                                          node->nodeid, dlm_state_str(node->state));
 534                                 wake_up(&node->shutdown_wait);
 535                                 break;
 536                         case DLM_LAST_ACK:
 537                                 /* probably remove_member caught it, do nothing */
 538                                 break;
 539                         default:
 540                                 spin_unlock(&node->state_lock);
 541                                 log_print("%s: unexpected state: %d\n",
 542                                           __func__, node->state);
 543                                 WARN_ON_ONCE(1);
 544                                 return;
 545                         }
 546                         spin_unlock(&node->state_lock);
 547
 548                         set_bit(DLM_NODE_FLAG_STOP_RX, &node->flags);
 549                         break;
 550                 default:
 551                         WARN_ON_ONCE(test_bit(DLM_NODE_FLAG_STOP_RX, &node->flags));
 552                         dlm_receive_buffer_3_2_trace(seq, p);
 553                         dlm_receive_buffer(p, node->nodeid);
 554                         set_bit(DLM_NODE_ULP_DELIVERED, &node->flags);
 555                         break;
 556                 }
 557         } else {
 558                 /* retry to ack message which we already have by sending back
 559                  * current node->seq_next number as ack.
 560                  */
 561                 if (seq < node->seq_next)
 562                         dlm_send_ack(node->nodeid, node->seq_next);
 563
 564                 log_print_ratelimited("ignore dlm msg because seq mismatch, seq: %u, expected: %u, nodeid: %d",
 565                                       seq, node->seq_next, node->nodeid);
 566         }
 567
 568         return;
 569
 570 send_fin:
 571         set_bit(DLM_NODE_FLAG_STOP_RX, &node->flags);
 572         dlm_send_fin(node, dlm_pas_fin_ack_rcv);
 573 }
 574
 575 static struct midcomms_node *
 576 dlm_midcomms_recv_node_lookup(int nodeid, const union dlm_packet *p,
 577                               uint16_t msglen, int (*cb)(struct midcomms_node *node))
 578 {
 579         struct midcomms_node *node = NULL;
 580         gfp_t allocation = 0;
 581         int ret;
 582
 583         switch (p->header.h_cmd) {
 584         case DLM_RCOM:
 585                 if (msglen < sizeof(struct dlm_rcom)) {
 586                         log_print("rcom msg too small: %u, will skip this message from node %d",
 587                                   msglen, nodeid);
 588                         return NULL;
 589                 }
 590
 591                 switch (p->rcom.rc_type) {
 592                 case cpu_to_le32(DLM_RCOM_NAMES):
 593                         fallthrough;
 594                 case cpu_to_le32(DLM_RCOM_NAMES_REPLY):
 595                         fallthrough;
 596                 case cpu_to_le32(DLM_RCOM_STATUS):
 597                         fallthrough;
 598                 case cpu_to_le32(DLM_RCOM_STATUS_REPLY):
 599                         node = nodeid2node(nodeid, 0);
 600                         if (node) {
 601                                 spin_lock(&node->state_lock);
 602                                 if (node->state != DLM_ESTABLISHED)
 603                                         pr_debug("receive begin RCOM msg from node %d with state %s\n",
 604                                                  node->nodeid, dlm_state_str(node->state));
 605
 606                                 switch (node->state) {
 607                                 case DLM_CLOSED:
 608                                         node->state = DLM_ESTABLISHED;
 609                                         pr_debug("switch node %d to state %s\n",
 610                                                  node->nodeid, dlm_state_str(node->state));
 611                                         break;
 612                                 case DLM_ESTABLISHED:
 613                                         break;
 614                                 default:
 615                                         /* some invalid state passive shutdown
 616                                          * was failed, we try to reset and
 617                                          * hope it will go on.
 618                                          */
 619                                         log_print("reset node %d because shutdown stuck",
 620                                                   node->nodeid);
 621
 622                                         midcomms_node_reset(node);
 623                                         node->state = DLM_ESTABLISHED;
 624                                         break;
 625                                 }
 626                                 spin_unlock(&node->state_lock);
 627                         }
 628
 629                         allocation = GFP_NOFS;
 630                         break;
 631                 default:
 632                         break;
 633                 }
 634
 635                 break;
 636         default:
 637                 break;
 638         }
 639
 640         node = nodeid2node(nodeid, allocation);
 641         if (!node) {
 642                 switch (p->header.h_cmd) {
 643                 case DLM_OPTS:
 644                         if (msglen < sizeof(struct dlm_opts)) {
 645                                 log_print("opts msg too small: %u, will skip this message from node %d",
 646                                           msglen, nodeid);
 647                                 return NULL;
 648                         }
 649
 650                         log_print_ratelimited("received dlm opts message nextcmd %d from node %d in an invalid sequence",
 651                                               p->opts.o_nextcmd, nodeid);
 652                         break;
 653                 default:
 654                         log_print_ratelimited("received dlm message cmd %d from node %d in an invalid sequence",
 655                                               p->header.h_cmd, nodeid);
 656                         break;
 657                 }
 658
 659                 return NULL;
 660         }
 661
 662         ret = cb(node);
 663         if (ret < 0)
 664                 return NULL;
 665
 666         return node;
 667 }
 668
 669 static int dlm_midcomms_version_check_3_2(struct midcomms_node *node)
 670 {
 671         switch (node->version) {
 672         case DLM_VERSION_NOT_SET:
 673                 node->version = DLM_VERSION_3_2;
 674                 log_print("version 0x%08x for node %d detected", DLM_VERSION_3_2,
 675                           node->nodeid);
 676                 break;
 677         case DLM_VERSION_3_2:
 678                 break;
 679         default:
 680                 log_print_ratelimited("version mismatch detected, assumed 0x%08x but node %d has 0x%08x",
 681                                       DLM_VERSION_3_2, node->nodeid, node->version);
 682                 return -1;
 683         }
 684
 685         return 0;
 686 }
 687
 688 static int dlm_opts_check_msglen(union dlm_packet *p, uint16_t msglen, int nodeid)
 689 {
 690         int len = msglen;
 691
 692         /* we only trust outer header msglen because
 693          * it's checked against receive buffer length.
 694          */
 695         if (len < sizeof(struct dlm_opts))
 696                 return -1;
 697         len -= sizeof(struct dlm_opts);
 698
 699         if (len < le16_to_cpu(p->opts.o_optlen))
 700                 return -1;
 701         len -= le16_to_cpu(p->opts.o_optlen);
 702
 703         switch (p->opts.o_nextcmd) {
 704         case DLM_FIN:
 705                 if (len < sizeof(struct dlm_header)) {
 706                         log_print("fin too small: %d, will skip this message from node %d",
 707                                   len, nodeid);
 708                         return -1;
 709                 }
 710
 711                 break;
 712         case DLM_MSG:
 713                 if (len < sizeof(struct dlm_message)) {
 714                         log_print("msg too small: %d, will skip this message from node %d",
 715                                   msglen, nodeid);
 716                         return -1;
 717                 }
 718
 719                 break;
 720         case DLM_RCOM:
 721                 if (len < sizeof(struct dlm_rcom)) {
 722                         log_print("rcom msg too small: %d, will skip this message from node %d",
 723                                   len, nodeid);
 724                         return -1;
 725                 }
 726
 727                 break;
 728         default:
 729                 log_print("unsupported o_nextcmd received: %u, will skip this message from node %d",
 730                           p->opts.o_nextcmd, nodeid);
 731                 return -1;
 732         }
 733
 734         return 0;
 735 }
 736
 737 static void dlm_midcomms_receive_buffer_3_2(union dlm_packet *p, int nodeid)
 738 {
 739         uint16_t msglen = le16_to_cpu(p->header.h_length);
 740         struct midcomms_node *node;
 741         uint32_t seq;
 742         int ret, idx;
 743
 744         idx = srcu_read_lock(&nodes_srcu);
 745         node = dlm_midcomms_recv_node_lookup(nodeid, p, msglen,
 746                                              dlm_midcomms_version_check_3_2);
 747         if (!node)
 748                 goto out;
 749
 750         switch (p->header.h_cmd) {
 751         case DLM_RCOM:
 752                 /* these rcom message we use to determine version.
 753                  * they have their own retransmission handling and
 754                  * are the first messages of dlm.
 755                  *
 756                  * length already checked.
 757                  */
 758                 switch (p->rcom.rc_type) {
 759                 case cpu_to_le32(DLM_RCOM_NAMES):
 760                         fallthrough;
 761                 case cpu_to_le32(DLM_RCOM_NAMES_REPLY):
 762                         fallthrough;
 763                 case cpu_to_le32(DLM_RCOM_STATUS):
 764                         fallthrough;
 765                 case cpu_to_le32(DLM_RCOM_STATUS_REPLY):
 766                         break;
 767                 default:
 768                         log_print("unsupported rcom type received: %u, will skip this message from node %d",
 769                                   le32_to_cpu(p->rcom.rc_type), nodeid);
 770                         goto out;
 771                 }
 772
 773                 WARN_ON_ONCE(test_bit(DLM_NODE_FLAG_STOP_RX, &node->flags));
 774                 dlm_receive_buffer(p, nodeid);
 775                 break;
 776         case DLM_OPTS:
 777                 seq = le32_to_cpu(p->header.u.h_seq);
 778
 779                 ret = dlm_opts_check_msglen(p, msglen, nodeid);
 780                 if (ret < 0) {
 781                         log_print("opts msg too small: %u, will skip this message from node %d",
 782                                   msglen, nodeid);
 783                         goto out;
 784                 }
 785
 786                 p = (union dlm_packet *)((unsigned char *)p->opts.o_opts +
 787                                          le16_to_cpu(p->opts.o_optlen));
 788
 789                 /* recheck inner msglen just if it's not garbage */
 790                 msglen = le16_to_cpu(p->header.h_length);
 791                 switch (p->header.h_cmd) {
 792                 case DLM_RCOM:
 793                         if (msglen < sizeof(struct dlm_rcom)) {
 794                                 log_print("inner rcom msg too small: %u, will skip this message from node %d",
 795                                           msglen, nodeid);
 796                                 goto out;
 797                         }
 798
 799                         break;
 800                 case DLM_MSG:
 801                         if (msglen < sizeof(struct dlm_message)) {
 802                                 log_print("inner msg too small: %u, will skip this message from node %d",
 803                                           msglen, nodeid);
 804                                 goto out;
 805                         }
 806
 807                         break;
 808                 case DLM_FIN:
 809                         if (msglen < sizeof(struct dlm_header)) {
 810                                 log_print("inner fin too small: %u, will skip this message from node %d",
 811                                           msglen, nodeid);
 812                                 goto out;
 813                         }
 814
 815                         break;
 816                 default:
 817                         log_print("unsupported inner h_cmd received: %u, will skip this message from node %d",
 818                                   msglen, nodeid);
 819                         goto out;
 820                 }
 821
 822                 dlm_midcomms_receive_buffer(p, node, seq);
 823                 break;
 824         case DLM_ACK:
 825                 seq = le32_to_cpu(p->header.u.h_seq);
 826                 dlm_receive_ack(node, seq);
 827                 break;
 828         default:
 829                 log_print("unsupported h_cmd received: %u, will skip this message from node %d",
 830                           p->header.h_cmd, nodeid);
 831                 break;
 832         }
 833
 834 out:
 835         srcu_read_unlock(&nodes_srcu, idx);
 836 }
 837
 838 static int dlm_midcomms_version_check_3_1(struct midcomms_node *node)
 839 {
 840         switch (node->version) {
 841         case DLM_VERSION_NOT_SET:
 842                 node->version = DLM_VERSION_3_1;
 843                 log_print("version 0x%08x for node %d detected", DLM_VERSION_3_1,
 844                           node->nodeid);
 845                 break;
 846         case DLM_VERSION_3_1:
 847                 break;
 848         default:
 849                 log_print_ratelimited("version mismatch detected, assumed 0x%08x but node %d has 0x%08x",
 850                                       DLM_VERSION_3_1, node->nodeid, node->version);
 851                 return -1;
 852         }
 853
 854         return 0;
 855 }
 856
 857 static void dlm_midcomms_receive_buffer_3_1(union dlm_packet *p, int nodeid)
 858 {
 859         uint16_t msglen = le16_to_cpu(p->header.h_length);
 860         struct midcomms_node *node;
 861         int idx;
 862
 863         idx = srcu_read_lock(&nodes_srcu);
 864         node = dlm_midcomms_recv_node_lookup(nodeid, p, msglen,
 865                                              dlm_midcomms_version_check_3_1);
 866         if (!node) {
 867                 srcu_read_unlock(&nodes_srcu, idx);
 868                 return;
 869         }
 870         srcu_read_unlock(&nodes_srcu, idx);
 871
 872         switch (p->header.h_cmd) {
 873         case DLM_RCOM:
 874                 /* length already checked */
 875                 break;
 876         case DLM_MSG:
 877                 if (msglen < sizeof(struct dlm_message)) {
 878                         log_print("msg too small: %u, will skip this message from node %d",
 879                                   msglen, nodeid);
 880                         return;
 881                 }
 882
 883                 break;
 884         default:
 885                 log_print("unsupported h_cmd received: %u, will skip this message from node %d",
 886                           p->header.h_cmd, nodeid);
 887                 return;
 888         }
 889
 890         dlm_receive_buffer(p, nodeid);
 891 }
 892
 893 int dlm_validate_incoming_buffer(int nodeid, unsigned char *buf, int len)
 894 {
 895         const unsigned char *ptr = buf;
 896         const struct dlm_header *hd;
 897         uint16_t msglen;
 898         int ret = 0;
 899
 900         while (len >= sizeof(struct dlm_header)) {
 901                 hd = (struct dlm_header *)ptr;
 902
 903                 /* no message should be more than DLM_MAX_SOCKET_BUFSIZE or
 904                  * less than dlm_header size.
 905                  *
 906                  * Some messages does not have a 8 byte length boundary yet
 907                  * which can occur in a unaligned memory access of some dlm
 908                  * messages. However this problem need to be fixed at the
 909                  * sending side, for now it seems nobody run into architecture
 910                  * related issues yet but it slows down some processing.
 911                  * Fixing this issue should be scheduled in future by doing
 912                  * the next major version bump.
 913                  */
 914                 msglen = le16_to_cpu(hd->h_length);
 915                 if (msglen > DLM_MAX_SOCKET_BUFSIZE ||
 916                     msglen < sizeof(struct dlm_header)) {
 917                         log_print("received invalid length header: %u from node %d, will abort message parsing",
 918                                   msglen, nodeid);
 919                         return -EBADMSG;
 920                 }
 921
 922                 /* caller will take care that leftover
 923                  * will be parsed next call with more data
 924                  */
 925                 if (msglen > len)
 926                         break;
 927
 928                 ret += msglen;
 929                 len -= msglen;
 930                 ptr += msglen;
 931         }
 932
 933         return ret;
 934 }
 935
 936 /*
 937  * Called from the low-level comms layer to process a buffer of
 938  * commands.
 939  */
 940 int dlm_process_incoming_buffer(int nodeid, unsigned char *buf, int len)
 941 {
 942         const unsigned char *ptr = buf;
 943         const struct dlm_header *hd;
 944         uint16_t msglen;
 945         int ret = 0;
 946
 947         while (len >= sizeof(struct dlm_header)) {
 948                 hd = (struct dlm_header *)ptr;
 949
 950                 msglen = le16_to_cpu(hd->h_length);
 951                 if (msglen > len)
 952                         break;
 953
 954                 switch (hd->h_version) {
 955                 case cpu_to_le32(DLM_VERSION_3_1):
 956                         dlm_midcomms_receive_buffer_3_1((union dlm_packet *)ptr, nodeid);
 957                         break;
 958                 case cpu_to_le32(DLM_VERSION_3_2):
 959                         dlm_midcomms_receive_buffer_3_2((union dlm_packet *)ptr, nodeid);
 960                         break;
 961                 default:
 962                         log_print("received invalid version header: %u from node %d, will skip this message",
 963                                   le32_to_cpu(hd->h_version), nodeid);
 964                         break;
 965                 }
 966
 967                 ret += msglen;
 968                 len -= msglen;
 969                 ptr += msglen;
 970         }
 971
 972         return ret;
 973 }
 974
 975 void dlm_midcomms_receive_done(int nodeid)
 976 {
 977         struct midcomms_node *node;
 978         int idx;
 979
 980         idx = srcu_read_lock(&nodes_srcu);
 981         node = nodeid2node(nodeid, 0);
 982         if (!node) {
 983                 srcu_read_unlock(&nodes_srcu, idx);
 984                 return;
 985         }
 986
 987         /* old protocol, we do nothing */
 988         switch (node->version) {
 989         case DLM_VERSION_3_2:
 990                 break;
 991         default:
 992                 srcu_read_unlock(&nodes_srcu, idx);
 993                 return;
 994         }
 995
 996         /* do nothing if we didn't delivered stateful to ulp */
 997         if (!test_and_clear_bit(DLM_NODE_ULP_DELIVERED,
 998                                 &node->flags)) {
 999                 srcu_read_unlock(&nodes_srcu, idx);
1000                 return;
1001         }
1002
1003         spin_lock(&node->state_lock);
1004         /* we only ack if state is ESTABLISHED */
1005         switch (node->state) {
1006         case DLM_ESTABLISHED:
1007                 spin_unlock(&node->state_lock);
1008                 dlm_send_ack(node->nodeid, node->seq_next);
1009                 break;
1010         default:
1011                 spin_unlock(&node->state_lock);
1012                 /* do nothing FIN has it's own ack send */
1013                 break;
1014         }
1015         srcu_read_unlock(&nodes_srcu, idx);
1016 }
1017
1018 void dlm_midcomms_unack_msg_resend(int nodeid)
1019 {
1020         struct midcomms_node *node;
1021         struct dlm_mhandle *mh;
1022         int idx, ret;
1023
1024         idx = srcu_read_lock(&nodes_srcu);
1025         node = nodeid2node(nodeid, 0);
1026         if (!node) {
1027                 srcu_read_unlock(&nodes_srcu, idx);
1028                 return;
1029         }
1030
1031         /* old protocol, we don't support to retransmit on failure */
1032         switch (node->version) {
1033         case DLM_VERSION_3_2:
1034                 break;
1035         default:
1036                 srcu_read_unlock(&nodes_srcu, idx);
1037                 return;
1038         }
1039
1040         rcu_read_lock();
1041         list_for_each_entry_rcu(mh, &node->send_queue, list) {
1042                 if (!mh->committed)
1043                         continue;
1044
1045                 ret = dlm_lowcomms_resend_msg(mh->msg);
1046                 if (!ret)
1047                         log_print_ratelimited("retransmit dlm msg, seq %u, nodeid %d",
1048                                               mh->seq, node->nodeid);
1049         }
1050         rcu_read_unlock();
1051         srcu_read_unlock(&nodes_srcu, idx);
1052 }
1053
1054 static void dlm_fill_opts_header(struct dlm_opts *opts, uint16_t inner_len,
1055                                  uint32_t seq)
1056 {
1057         opts->o_header.h_cmd = DLM_OPTS;
1058         opts->o_header.h_version = cpu_to_le32(DLM_HEADER_MAJOR | DLM_HEADER_MINOR);
1059         opts->o_header.h_nodeid = cpu_to_le32(dlm_our_nodeid());
1060         opts->o_header.h_length = cpu_to_le16(DLM_MIDCOMMS_OPT_LEN + inner_len);
1061         opts->o_header.u.h_seq = cpu_to_le32(seq);
1062 }
1063
1064 static void midcomms_new_msg_cb(void *data)
1065 {
1066         struct dlm_mhandle *mh = data;
1067
1068         atomic_inc(&mh->node->send_queue_cnt);
1069
1070         spin_lock_bh(&mh->node->send_queue_lock);
1071         list_add_tail_rcu(&mh->list, &mh->node->send_queue);
1072         spin_unlock_bh(&mh->node->send_queue_lock);
1073
1074         mh->seq = mh->node->seq_send++;
1075 }
1076
1077 static struct dlm_msg *dlm_midcomms_get_msg_3_2(struct dlm_mhandle *mh, int nodeid,
1078                                                 int len, gfp_t allocation, char **ppc)
1079 {
1080         struct dlm_opts *opts;
1081         struct dlm_msg *msg;
1082
1083         msg = dlm_lowcomms_new_msg(nodeid, len + DLM_MIDCOMMS_OPT_LEN,
1084                                    allocation, ppc, midcomms_new_msg_cb, mh);
1085         if (!msg)
1086                 return NULL;
1087
1088         opts = (struct dlm_opts *)*ppc;
1089         mh->opts = opts;
1090
1091         /* add possible options here */
1092         dlm_fill_opts_header(opts, len, mh->seq);
1093
1094         *ppc += sizeof(*opts);
1095         mh->inner_p = (const union dlm_packet *)*ppc;
1096         return msg;
1097 }
1098
1099 /* avoid false positive for nodes_srcu, unlock happens in
1100  * dlm_midcomms_commit_mhandle which is a must call if success
1101  */
1102 #ifndef __CHECKER__
1103 struct dlm_mhandle *dlm_midcomms_get_mhandle(int nodeid, int len,
1104                                              gfp_t allocation, char **ppc)
1105 {
1106         struct midcomms_node *node;
1107         struct dlm_mhandle *mh;
1108         struct dlm_msg *msg;
1109         int idx;
1110
1111         idx = srcu_read_lock(&nodes_srcu);
1112         node = nodeid2node(nodeid, 0);
1113         if (!node) {
1114                 WARN_ON_ONCE(1);
1115                 goto err;
1116         }
1117
1118         /* this is a bug, however we going on and hope it will be resolved */
1119         WARN_ON_ONCE(test_bit(DLM_NODE_FLAG_STOP_TX, &node->flags));
1120
1121         mh = dlm_allocate_mhandle(allocation);
1122         if (!mh)
1123                 goto err;
1124
1125         mh->committed = false;
1126         mh->ack_rcv = NULL;
1127         mh->idx = idx;
1128         mh->node = node;
1129
1130         switch (node->version) {
1131         case DLM_VERSION_3_1:
1132                 msg = dlm_lowcomms_new_msg(nodeid, len, allocation, ppc,
1133                                            NULL, NULL);
1134                 if (!msg) {
1135                         dlm_free_mhandle(mh);
1136                         goto err;
1137                 }
1138
1139                 break;
1140         case DLM_VERSION_3_2:
1141                 msg = dlm_midcomms_get_msg_3_2(mh, nodeid, len, allocation,
1142                                                ppc);
1143                 if (!msg) {
1144                         dlm_free_mhandle(mh);
1145                         goto err;
1146                 }
1147
1148                 break;
1149         default:
1150                 dlm_free_mhandle(mh);
1151                 WARN_ON_ONCE(1);
1152                 goto err;
1153         }
1154
1155         mh->msg = msg;
1156
1157         /* keep in mind that is a must to call
1158          * dlm_midcomms_commit_msg() which releases
1159          * nodes_srcu using mh->idx which is assumed
1160          * here that the application will call it.
1161          */
1162         return mh;
1163
1164 err:
1165         srcu_read_unlock(&nodes_srcu, idx);
1166         return NULL;
1167 }
1168 #endif
1169
1170 static void dlm_midcomms_commit_msg_3_2_trace(const struct dlm_mhandle *mh,
1171                                               const void *name, int namelen)
1172 {
1173         switch (mh->inner_p->header.h_cmd) {
1174         case DLM_MSG:
1175                 trace_dlm_send_message(mh->node->nodeid, mh->seq,
1176                                        &mh->inner_p->message,
1177                                        name, namelen);
1178                 break;
1179         case DLM_RCOM:
1180                 trace_dlm_send_rcom(mh->node->nodeid, mh->seq,
1181                                     &mh->inner_p->rcom);
1182                 break;
1183         default:
1184                 /* nothing to trace */
1185                 break;
1186         }
1187 }
1188
1189 static void dlm_midcomms_commit_msg_3_2(struct dlm_mhandle *mh,
1190                                         const void *name, int namelen)
1191 {
1192         /* nexthdr chain for fast lookup */
1193         mh->opts->o_nextcmd = mh->inner_p->header.h_cmd;
1194         mh->committed = true;
1195         dlm_midcomms_commit_msg_3_2_trace(mh, name, namelen);
1196         dlm_lowcomms_commit_msg(mh->msg);
1197 }
1198
1199 /* avoid false positive for nodes_srcu, lock was happen in
1200  * dlm_midcomms_get_mhandle
1201  */
1202 #ifndef __CHECKER__
1203 void dlm_midcomms_commit_mhandle(struct dlm_mhandle *mh,
1204                                  const void *name, int namelen)
1205 {
1206
1207         switch (mh->node->version) {
1208         case DLM_VERSION_3_1:
1209                 srcu_read_unlock(&nodes_srcu, mh->idx);
1210
1211                 dlm_lowcomms_commit_msg(mh->msg);
1212                 dlm_lowcomms_put_msg(mh->msg);
1213                 /* mh is not part of rcu list in this case */
1214                 dlm_free_mhandle(mh);
1215                 break;
1216         case DLM_VERSION_3_2:
1217                 dlm_midcomms_commit_msg_3_2(mh, name, namelen);
1218                 srcu_read_unlock(&nodes_srcu, mh->idx);
1219                 break;
1220         default:
1221                 srcu_read_unlock(&nodes_srcu, mh->idx);
1222                 WARN_ON_ONCE(1);
1223                 break;
1224         }
1225 }
1226 #endif
1227
1228 int dlm_midcomms_start(void)
1229 {
1230         return dlm_lowcomms_start();
1231 }
1232
1233 void dlm_midcomms_stop(void)
1234 {
1235         dlm_lowcomms_stop();
1236 }
1237
1238 void dlm_midcomms_init(void)
1239 {
1240         int i;
1241
1242         for (i = 0; i < CONN_HASH_SIZE; i++)
1243                 INIT_HLIST_HEAD(&node_hash[i]);
1244
1245         dlm_lowcomms_init();
1246 }
1247
1248 void dlm_midcomms_exit(void)
1249 {
1250         dlm_lowcomms_exit();
1251 }
1252
1253 static void dlm_act_fin_ack_rcv(struct midcomms_node *node)
1254 {
1255         spin_lock(&node->state_lock);
1256         pr_debug("receive active fin ack from node %d with state %s\n",
1257                  node->nodeid, dlm_state_str(node->state));
1258
1259         switch (node->state) {
1260         case DLM_FIN_WAIT1:
1261                 node->state = DLM_FIN_WAIT2;
1262                 pr_debug("switch node %d to state %s\n",
1263                          node->nodeid, dlm_state_str(node->state));
1264                 break;
1265         case DLM_CLOSING:
1266                 midcomms_node_reset(node);
1267                 pr_debug("switch node %d to state %s\n",
1268                          node->nodeid, dlm_state_str(node->state));
1269                 wake_up(&node->shutdown_wait);
1270                 break;
1271         case DLM_CLOSED:
1272                 /* not valid but somehow we got what we want */
1273                 wake_up(&node->shutdown_wait);
1274                 break;
1275         default:
1276                 spin_unlock(&node->state_lock);
1277                 log_print("%s: unexpected state: %d\n",
1278                           __func__, node->state);
1279                 WARN_ON_ONCE(1);
1280                 return;
1281         }
1282         spin_unlock(&node->state_lock);
1283 }
1284
1285 void dlm_midcomms_add_member(int nodeid)
1286 {
1287         struct midcomms_node *node;
1288         int idx;
1289
1290         if (nodeid == dlm_our_nodeid())
1291                 return;
1292
1293         idx = srcu_read_lock(&nodes_srcu);
1294         node = nodeid2node(nodeid, GFP_NOFS);
1295         if (!node) {
1296                 srcu_read_unlock(&nodes_srcu, idx);
1297                 return;
1298         }
1299
1300         spin_lock(&node->state_lock);
1301         if (!node->users) {
1302                 pr_debug("receive add member from node %d with state %s\n",
1303                          node->nodeid, dlm_state_str(node->state));
1304                 switch (node->state) {
1305                 case DLM_ESTABLISHED:
1306                         break;
1307                 case DLM_CLOSED:
1308                         node->state = DLM_ESTABLISHED;
1309                         pr_debug("switch node %d to state %s\n",
1310                                  node->nodeid, dlm_state_str(node->state));
1311                         break;
1312                 default:
1313                         /* some invalid state passive shutdown
1314                          * was failed, we try to reset and
1315                          * hope it will go on.
1316                          */
1317                         log_print("reset node %d because shutdown stuck",
1318                                   node->nodeid);
1319
1320                         midcomms_node_reset(node);
1321                         node->state = DLM_ESTABLISHED;
1322                         break;
1323                 }
1324         }
1325
1326         node->users++;
1327         pr_debug("node %d users inc count %d\n", nodeid, node->users);
1328         spin_unlock(&node->state_lock);
1329
1330         srcu_read_unlock(&nodes_srcu, idx);
1331 }
1332
1333 void dlm_midcomms_remove_member(int nodeid)
1334 {
1335         struct midcomms_node *node;
1336         int idx;
1337
1338         if (nodeid == dlm_our_nodeid())
1339                 return;
1340
1341         idx = srcu_read_lock(&nodes_srcu);
1342         node = nodeid2node(nodeid, 0);
1343         if (!node) {
1344                 srcu_read_unlock(&nodes_srcu, idx);
1345                 return;
1346         }
1347
1348         spin_lock(&node->state_lock);
1349         node->users--;
1350         pr_debug("node %d users dec count %d\n", nodeid, node->users);
1351
1352         /* hitting users count to zero means the
1353          * other side is running dlm_midcomms_stop()
1354          * we meet us to have a clean disconnect.
1355          */
1356         if (node->users == 0) {
1357                 pr_debug("receive remove member from node %d with state %s\n",
1358                          node->nodeid, dlm_state_str(node->state));
1359                 switch (node->state) {
1360                 case DLM_ESTABLISHED:
1361                         break;
1362                 case DLM_CLOSE_WAIT:
1363                         /* passive shutdown DLM_LAST_ACK case 2 */
1364                         node->state = DLM_LAST_ACK;
1365                         spin_unlock(&node->state_lock);
1366
1367                         pr_debug("switch node %d to state %s case 2\n",
1368                                  node->nodeid, dlm_state_str(node->state));
1369                         goto send_fin;
1370                 case DLM_LAST_ACK:
1371                         /* probably receive fin caught it, do nothing */
1372                         break;
1373                 case DLM_CLOSED:
1374                         /* already gone, do nothing */
1375                         break;
1376                 default:
1377                         log_print("%s: unexpected state: %d\n",
1378                                   __func__, node->state);
1379                         break;
1380                 }
1381         }
1382         spin_unlock(&node->state_lock);
1383
1384         srcu_read_unlock(&nodes_srcu, idx);
1385         return;
1386
1387 send_fin:
1388         set_bit(DLM_NODE_FLAG_STOP_RX, &node->flags);
1389         dlm_send_fin(node, dlm_pas_fin_ack_rcv);
1390         srcu_read_unlock(&nodes_srcu, idx);
1391 }
1392
1393 static void midcomms_node_release(struct rcu_head *rcu)
1394 {
1395         struct midcomms_node *node = container_of(rcu, struct midcomms_node, rcu);
1396
1397         WARN_ON_ONCE(atomic_read(&node->send_queue_cnt));
1398         kfree(node);
1399 }
1400
1401 static void midcomms_shutdown(struct midcomms_node *node)
1402 {
1403         int ret;
1404
1405         /* old protocol, we don't wait for pending operations */
1406         switch (node->version) {
1407         case DLM_VERSION_3_2:
1408                 break;
1409         default:
1410                 return;
1411         }
1412
1413         spin_lock(&node->state_lock);
1414         pr_debug("receive active shutdown for node %d with state %s\n",
1415                  node->nodeid, dlm_state_str(node->state));
1416         switch (node->state) {
1417         case DLM_ESTABLISHED:
1418                 node->state = DLM_FIN_WAIT1;
1419                 pr_debug("switch node %d to state %s case 2\n",
1420                          node->nodeid, dlm_state_str(node->state));
1421                 break;
1422         case DLM_CLOSED:
1423                 /* we have what we want */
1424                 spin_unlock(&node->state_lock);
1425                 return;
1426         default:
1427                 /* busy to enter DLM_FIN_WAIT1, wait until passive
1428                  * done in shutdown_wait to enter DLM_CLOSED.
1429                  */
1430                 break;
1431         }
1432         spin_unlock(&node->state_lock);
1433
1434         if (node->state == DLM_FIN_WAIT1) {
1435                 dlm_send_fin(node, dlm_act_fin_ack_rcv);
1436
1437                 if (DLM_DEBUG_FENCE_TERMINATION)
1438                         msleep(5000);
1439         }
1440
1441         /* wait for other side dlm + fin */
1442         ret = wait_event_timeout(node->shutdown_wait,
1443                                  node->state == DLM_CLOSED ||
1444                                  test_bit(DLM_NODE_FLAG_CLOSE, &node->flags),
1445                                  DLM_SHUTDOWN_TIMEOUT);
1446         if (!ret || test_bit(DLM_NODE_FLAG_CLOSE, &node->flags)) {
1447                 pr_debug("active shutdown timed out for node %d with state %s\n",
1448                          node->nodeid, dlm_state_str(node->state));
1449                 midcomms_node_reset(node);
1450                 dlm_lowcomms_shutdown_node(node->nodeid, true);
1451                 return;
1452         }
1453
1454         pr_debug("active shutdown done for node %d with state %s\n",
1455                  node->nodeid, dlm_state_str(node->state));
1456         dlm_lowcomms_shutdown_node(node->nodeid, false);
1457 }
1458
1459 void dlm_midcomms_shutdown(void)
1460 {
1461         struct midcomms_node *node;
1462         int i, idx;
1463
1464         dlm_lowcomms_shutdown();
1465
1466         mutex_lock(&close_lock);
1467         idx = srcu_read_lock(&nodes_srcu);
1468         for (i = 0; i < CONN_HASH_SIZE; i++) {
1469                 hlist_for_each_entry_rcu(node, &node_hash[i], hlist) {
1470                         midcomms_shutdown(node);
1471
1472                         dlm_delete_debug_comms_file(node->debugfs);
1473
1474                         spin_lock(&nodes_lock);
1475                         hlist_del_rcu(&node->hlist);
1476                         spin_unlock(&nodes_lock);
1477
1478                         call_srcu(&nodes_srcu, &node->rcu, midcomms_node_release);
1479                 }
1480         }
1481         srcu_read_unlock(&nodes_srcu, idx);
1482         mutex_unlock(&close_lock);
1483 }
1484
1485 int dlm_midcomms_close(int nodeid)
1486 {
1487         struct midcomms_node *node;
1488         int idx, ret;
1489
1490         if (nodeid == dlm_our_nodeid())
1491                 return 0;
1492
1493         dlm_stop_lockspaces_check();
1494
1495         idx = srcu_read_lock(&nodes_srcu);
1496         /* Abort pending close/remove operation */
1497         node = nodeid2node(nodeid, 0);
1498         if (node) {
1499                 /* let shutdown waiters leave */
1500                 set_bit(DLM_NODE_FLAG_CLOSE, &node->flags);
1501                 wake_up(&node->shutdown_wait);
1502         }
1503         srcu_read_unlock(&nodes_srcu, idx);
1504
1505         synchronize_srcu(&nodes_srcu);
1506
1507         idx = srcu_read_lock(&nodes_srcu);
1508         mutex_lock(&close_lock);
1509         node = nodeid2node(nodeid, 0);
1510         if (!node) {
1511                 mutex_unlock(&close_lock);
1512                 srcu_read_unlock(&nodes_srcu, idx);
1513                 return dlm_lowcomms_close(nodeid);
1514         }
1515
1516         ret = dlm_lowcomms_close(nodeid);
1517         spin_lock(&node->state_lock);
1518         midcomms_node_reset(node);
1519         spin_unlock(&node->state_lock);
1520         srcu_read_unlock(&nodes_srcu, idx);
1521         mutex_unlock(&close_lock);
1522
1523         return ret;
1524 }
1525
1526 /* debug functionality to send raw dlm msg from user space */
1527 struct dlm_rawmsg_data {
1528         struct midcomms_node *node;
1529         void *buf;
1530 };
1531
1532 static void midcomms_new_rawmsg_cb(void *data)
1533 {
1534         struct dlm_rawmsg_data *rd = data;
1535         struct dlm_header *h = rd->buf;
1536
1537         switch (h->h_version) {
1538         case cpu_to_le32(DLM_VERSION_3_1):
1539                 break;
1540         default:
1541                 switch (h->h_cmd) {
1542                 case DLM_OPTS:
1543                         if (!h->u.h_seq)
1544                                 h->u.h_seq = cpu_to_le32(rd->node->seq_send++);
1545                         break;
1546                 default:
1547                         break;
1548                 }
1549                 break;
1550         }
1551 }
1552
1553 int dlm_midcomms_rawmsg_send(struct midcomms_node *node, void *buf,
1554                              int buflen)
1555 {
1556         struct dlm_rawmsg_data rd;
1557         struct dlm_msg *msg;
1558         char *msgbuf;
1559
1560         rd.node = node;
1561         rd.buf = buf;
1562
1563         msg = dlm_lowcomms_new_msg(node->nodeid, buflen, GFP_NOFS,
1564                                    &msgbuf, midcomms_new_rawmsg_cb, &rd);
1565         if (!msg)
1566                 return -ENOMEM;
1567
1568         memcpy(msgbuf, buf, buflen);
1569         dlm_lowcomms_commit_msg(msg);
1570         return 0;
1571 }
1572