fs/ocfs2/cluster/nodemanager.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /* -*- mode: c; c-basic-offset: 8; -*-
   3  * vim: noexpandtab sw=8 ts=8 sts=0:
   4  *
   5  * Copyright (C) 2004, 2005 Oracle.  All rights reserved.
   6  */
   7
   8 #include <linux/slab.h>
   9 #include <linux/kernel.h>
  10 #include <linux/module.h>
  11 #include <linux/configfs.h>
  12
  13 #include "tcp.h"
  14 #include "nodemanager.h"
  15 #include "heartbeat.h"
  16 #include "masklog.h"
  17 #include "sys.h"
  18
  19 /* for now we operate under the assertion that there can be only one
  20  * cluster active at a time.  Changing this will require trickling
  21  * cluster references throughout where nodes are looked up */
  22 struct o2nm_cluster *o2nm_single_cluster = NULL;
  23
  24 static const char *o2nm_fence_method_desc[O2NM_FENCE_METHODS] = {
  25         "reset",        /* O2NM_FENCE_RESET */
  26         "panic",        /* O2NM_FENCE_PANIC */
  27 };
  28
  29 static inline void o2nm_lock_subsystem(void);
  30 static inline void o2nm_unlock_subsystem(void);
  31
  32 struct o2nm_node *o2nm_get_node_by_num(u8 node_num)
  33 {
  34         struct o2nm_node *node = NULL;
  35
  36         if (node_num >= O2NM_MAX_NODES || o2nm_single_cluster == NULL)
  37                 goto out;
  38
  39         read_lock(&o2nm_single_cluster->cl_nodes_lock);
  40         node = o2nm_single_cluster->cl_nodes[node_num];
  41         if (node)
  42                 config_item_get(&node->nd_item);
  43         read_unlock(&o2nm_single_cluster->cl_nodes_lock);
  44 out:
  45         return node;
  46 }
  47 EXPORT_SYMBOL_GPL(o2nm_get_node_by_num);
  48
  49 int o2nm_configured_node_map(unsigned long *map, unsigned bytes)
  50 {
  51         struct o2nm_cluster *cluster = o2nm_single_cluster;
  52
  53         BUG_ON(bytes < (sizeof(cluster->cl_nodes_bitmap)));
  54
  55         if (cluster == NULL)
  56                 return -EINVAL;
  57
  58         read_lock(&cluster->cl_nodes_lock);
  59         memcpy(map, cluster->cl_nodes_bitmap, sizeof(cluster->cl_nodes_bitmap));
  60         read_unlock(&cluster->cl_nodes_lock);
  61
  62         return 0;
  63 }
  64 EXPORT_SYMBOL_GPL(o2nm_configured_node_map);
  65
  66 static struct o2nm_node *o2nm_node_ip_tree_lookup(struct o2nm_cluster *cluster,
  67                                                   __be32 ip_needle,
  68                                                   struct rb_node ***ret_p,
  69                                                   struct rb_node **ret_parent)
  70 {
  71         struct rb_node **p = &cluster->cl_node_ip_tree.rb_node;
  72         struct rb_node *parent = NULL;
  73         struct o2nm_node *node, *ret = NULL;
  74
  75         while (*p) {
  76                 int cmp;
  77
  78                 parent = *p;
  79                 node = rb_entry(parent, struct o2nm_node, nd_ip_node);
  80
  81                 cmp = memcmp(&ip_needle, &node->nd_ipv4_address,
  82                                 sizeof(ip_needle));
  83                 if (cmp < 0)
  84                         p = &(*p)->rb_left;
  85                 else if (cmp > 0)
  86                         p = &(*p)->rb_right;
  87                 else {
  88                         ret = node;
  89                         break;
  90                 }
  91         }
  92
  93         if (ret_p != NULL)
  94                 *ret_p = p;
  95         if (ret_parent != NULL)
  96                 *ret_parent = parent;
  97
  98         return ret;
  99 }
 100
 101 struct o2nm_node *o2nm_get_node_by_ip(__be32 addr)
 102 {
 103         struct o2nm_node *node = NULL;
 104         struct o2nm_cluster *cluster = o2nm_single_cluster;
 105
 106         if (cluster == NULL)
 107                 goto out;
 108
 109         read_lock(&cluster->cl_nodes_lock);
 110         node = o2nm_node_ip_tree_lookup(cluster, addr, NULL, NULL);
 111         if (node)
 112                 config_item_get(&node->nd_item);
 113         read_unlock(&cluster->cl_nodes_lock);
 114
 115 out:
 116         return node;
 117 }
 118 EXPORT_SYMBOL_GPL(o2nm_get_node_by_ip);
 119
 120 void o2nm_node_put(struct o2nm_node *node)
 121 {
 122         config_item_put(&node->nd_item);
 123 }
 124 EXPORT_SYMBOL_GPL(o2nm_node_put);
 125
 126 void o2nm_node_get(struct o2nm_node *node)
 127 {
 128         config_item_get(&node->nd_item);
 129 }
 130 EXPORT_SYMBOL_GPL(o2nm_node_get);
 131
 132 u8 o2nm_this_node(void)
 133 {
 134         u8 node_num = O2NM_MAX_NODES;
 135
 136         if (o2nm_single_cluster && o2nm_single_cluster->cl_has_local)
 137                 node_num = o2nm_single_cluster->cl_local_node;
 138
 139         return node_num;
 140 }
 141 EXPORT_SYMBOL_GPL(o2nm_this_node);
 142
 143 /* node configfs bits */
 144
 145 static struct o2nm_cluster *to_o2nm_cluster(struct config_item *item)
 146 {
 147         return item ?
 148                 container_of(to_config_group(item), struct o2nm_cluster,
 149                              cl_group)
 150                 : NULL;
 151 }
 152
 153 static struct o2nm_node *to_o2nm_node(struct config_item *item)
 154 {
 155         return item ? container_of(item, struct o2nm_node, nd_item) : NULL;
 156 }
 157
 158 static void o2nm_node_release(struct config_item *item)
 159 {
 160         struct o2nm_node *node = to_o2nm_node(item);
 161         kfree(node);
 162 }
 163
 164 static ssize_t o2nm_node_num_show(struct config_item *item, char *page)
 165 {
 166         return sprintf(page, "%d\n", to_o2nm_node(item)->nd_num);
 167 }
 168
 169 static struct o2nm_cluster *to_o2nm_cluster_from_node(struct o2nm_node *node)
 170 {
 171         /* through the first node_set .parent
 172          * mycluster/nodes/mynode == o2nm_cluster->o2nm_node_group->o2nm_node */
 173         if (node->nd_item.ci_parent)
 174                 return to_o2nm_cluster(node->nd_item.ci_parent->ci_parent);
 175         else
 176                 return NULL;
 177 }
 178
 179 enum {
 180         O2NM_NODE_ATTR_NUM = 0,
 181         O2NM_NODE_ATTR_PORT,
 182         O2NM_NODE_ATTR_ADDRESS,
 183 };
 184
 185 static ssize_t o2nm_node_num_store(struct config_item *item, const char *page,
 186                                    size_t count)
 187 {
 188         struct o2nm_node *node = to_o2nm_node(item);
 189         struct o2nm_cluster *cluster;
 190         unsigned long tmp;
 191         char *p = (char *)page;
 192         int ret = 0;
 193
 194         tmp = simple_strtoul(p, &p, 0);
 195         if (!p || (*p && (*p != '\n')))
 196                 return -EINVAL;
 197
 198         if (tmp >= O2NM_MAX_NODES)
 199                 return -ERANGE;
 200
 201         /* once we're in the cl_nodes tree networking can look us up by
 202          * node number and try to use our address and port attributes
 203          * to connect to this node.. make sure that they've been set
 204          * before writing the node attribute? */
 205         if (!test_bit(O2NM_NODE_ATTR_ADDRESS, &node->nd_set_attributes) ||
 206             !test_bit(O2NM_NODE_ATTR_PORT, &node->nd_set_attributes))
 207                 return -EINVAL; /* XXX */
 208
 209         o2nm_lock_subsystem();
 210         cluster = to_o2nm_cluster_from_node(node);
 211         if (!cluster) {
 212                 o2nm_unlock_subsystem();
 213                 return -EINVAL;
 214         }
 215
 216         write_lock(&cluster->cl_nodes_lock);
 217         if (cluster->cl_nodes[tmp])
 218                 ret = -EEXIST;
 219         else if (test_and_set_bit(O2NM_NODE_ATTR_NUM,
 220                         &node->nd_set_attributes))
 221                 ret = -EBUSY;
 222         else  {
 223                 cluster->cl_nodes[tmp] = node;
 224                 node->nd_num = tmp;
 225                 set_bit(tmp, cluster->cl_nodes_bitmap);
 226         }
 227         write_unlock(&cluster->cl_nodes_lock);
 228         o2nm_unlock_subsystem();
 229
 230         if (ret)
 231                 return ret;
 232
 233         return count;
 234 }
 235 static ssize_t o2nm_node_ipv4_port_show(struct config_item *item, char *page)
 236 {
 237         return sprintf(page, "%u\n", ntohs(to_o2nm_node(item)->nd_ipv4_port));
 238 }
 239
 240 static ssize_t o2nm_node_ipv4_port_store(struct config_item *item,
 241                                          const char *page, size_t count)
 242 {
 243         struct o2nm_node *node = to_o2nm_node(item);
 244         unsigned long tmp;
 245         char *p = (char *)page;
 246
 247         tmp = simple_strtoul(p, &p, 0);
 248         if (!p || (*p && (*p != '\n')))
 249                 return -EINVAL;
 250
 251         if (tmp == 0)
 252                 return -EINVAL;
 253         if (tmp >= (u16)-1)
 254                 return -ERANGE;
 255
 256         if (test_and_set_bit(O2NM_NODE_ATTR_PORT, &node->nd_set_attributes))
 257                 return -EBUSY;
 258         node->nd_ipv4_port = htons(tmp);
 259
 260         return count;
 261 }
 262
 263 static ssize_t o2nm_node_ipv4_address_show(struct config_item *item, char *page)
 264 {
 265         return sprintf(page, "%pI4\n", &to_o2nm_node(item)->nd_ipv4_address);
 266 }
 267
 268 static ssize_t o2nm_node_ipv4_address_store(struct config_item *item,
 269                                             const char *page,
 270                                             size_t count)
 271 {
 272         struct o2nm_node *node = to_o2nm_node(item);
 273         struct o2nm_cluster *cluster;
 274         int ret, i;
 275         struct rb_node **p, *parent;
 276         unsigned int octets[4];
 277         __be32 ipv4_addr = 0;
 278
 279         ret = sscanf(page, "%3u.%3u.%3u.%3u", &octets[3], &octets[2],
 280                      &octets[1], &octets[0]);
 281         if (ret != 4)
 282                 return -EINVAL;
 283
 284         for (i = 0; i < ARRAY_SIZE(octets); i++) {
 285                 if (octets[i] > 255)
 286                         return -ERANGE;
 287                 be32_add_cpu(&ipv4_addr, octets[i] << (i * 8));
 288         }
 289
 290         o2nm_lock_subsystem();
 291         cluster = to_o2nm_cluster_from_node(node);
 292         if (!cluster) {
 293                 o2nm_unlock_subsystem();
 294                 return -EINVAL;
 295         }
 296
 297         ret = 0;
 298         write_lock(&cluster->cl_nodes_lock);
 299         if (o2nm_node_ip_tree_lookup(cluster, ipv4_addr, &p, &parent))
 300                 ret = -EEXIST;
 301         else if (test_and_set_bit(O2NM_NODE_ATTR_ADDRESS,
 302                         &node->nd_set_attributes))
 303                 ret = -EBUSY;
 304         else {
 305                 rb_link_node(&node->nd_ip_node, parent, p);
 306                 rb_insert_color(&node->nd_ip_node, &cluster->cl_node_ip_tree);
 307         }
 308         write_unlock(&cluster->cl_nodes_lock);
 309         o2nm_unlock_subsystem();
 310
 311         if (ret)
 312                 return ret;
 313
 314         memcpy(&node->nd_ipv4_address, &ipv4_addr, sizeof(ipv4_addr));
 315
 316         return count;
 317 }
 318
 319 static ssize_t o2nm_node_local_show(struct config_item *item, char *page)
 320 {
 321         return sprintf(page, "%d\n", to_o2nm_node(item)->nd_local);
 322 }
 323
 324 static ssize_t o2nm_node_local_store(struct config_item *item, const char *page,
 325                                      size_t count)
 326 {
 327         struct o2nm_node *node = to_o2nm_node(item);
 328         struct o2nm_cluster *cluster;
 329         unsigned long tmp;
 330         char *p = (char *)page;
 331         ssize_t ret;
 332
 333         tmp = simple_strtoul(p, &p, 0);
 334         if (!p || (*p && (*p != '\n')))
 335                 return -EINVAL;
 336
 337         tmp = !!tmp; /* boolean of whether this node wants to be local */
 338
 339         /* setting local turns on networking rx for now so we require having
 340          * set everything else first */
 341         if (!test_bit(O2NM_NODE_ATTR_ADDRESS, &node->nd_set_attributes) ||
 342             !test_bit(O2NM_NODE_ATTR_NUM, &node->nd_set_attributes) ||
 343             !test_bit(O2NM_NODE_ATTR_PORT, &node->nd_set_attributes))
 344                 return -EINVAL; /* XXX */
 345
 346         o2nm_lock_subsystem();
 347         cluster = to_o2nm_cluster_from_node(node);
 348         if (!cluster) {
 349                 ret = -EINVAL;
 350                 goto out;
 351         }
 352
 353         /* the only failure case is trying to set a new local node
 354          * when a different one is already set */
 355         if (tmp && tmp == cluster->cl_has_local &&
 356             cluster->cl_local_node != node->nd_num) {
 357                 ret = -EBUSY;
 358                 goto out;
 359         }
 360
 361         /* bring up the rx thread if we're setting the new local node. */
 362         if (tmp && !cluster->cl_has_local) {
 363                 ret = o2net_start_listening(node);
 364                 if (ret)
 365                         goto out;
 366         }
 367
 368         if (!tmp && cluster->cl_has_local &&
 369             cluster->cl_local_node == node->nd_num) {
 370                 o2net_stop_listening(node);
 371                 cluster->cl_local_node = O2NM_INVALID_NODE_NUM;
 372         }
 373
 374         node->nd_local = tmp;
 375         if (node->nd_local) {
 376                 cluster->cl_has_local = tmp;
 377                 cluster->cl_local_node = node->nd_num;
 378         }
 379
 380         ret = count;
 381
 382 out:
 383         o2nm_unlock_subsystem();
 384         return ret;
 385 }
 386
 387 CONFIGFS_ATTR(o2nm_node_, num);
 388 CONFIGFS_ATTR(o2nm_node_, ipv4_port);
 389 CONFIGFS_ATTR(o2nm_node_, ipv4_address);
 390 CONFIGFS_ATTR(o2nm_node_, local);
 391
 392 static struct configfs_attribute *o2nm_node_attrs[] = {
 393         &o2nm_node_attr_num,
 394         &o2nm_node_attr_ipv4_port,
 395         &o2nm_node_attr_ipv4_address,
 396         &o2nm_node_attr_local,
 397         NULL,
 398 };
 399
 400 static struct configfs_item_operations o2nm_node_item_ops = {
 401         .release                = o2nm_node_release,
 402 };
 403
 404 static const struct config_item_type o2nm_node_type = {
 405         .ct_item_ops    = &o2nm_node_item_ops,
 406         .ct_attrs       = o2nm_node_attrs,
 407         .ct_owner       = THIS_MODULE,
 408 };
 409
 410 /* node set */
 411
 412 struct o2nm_node_group {
 413         struct config_group ns_group;
 414         /* some stuff? */
 415 };
 416
 417 #if 0
 418 static struct o2nm_node_group *to_o2nm_node_group(struct config_group *group)
 419 {
 420         return group ?
 421                 container_of(group, struct o2nm_node_group, ns_group)
 422                 : NULL;
 423 }
 424 #endif
 425
 426 static ssize_t o2nm_cluster_attr_write(const char *page, ssize_t count,
 427                                        unsigned int *val)
 428 {
 429         unsigned long tmp;
 430         char *p = (char *)page;
 431
 432         tmp = simple_strtoul(p, &p, 0);
 433         if (!p || (*p && (*p != '\n')))
 434                 return -EINVAL;
 435
 436         if (tmp == 0)
 437                 return -EINVAL;
 438         if (tmp >= (u32)-1)
 439                 return -ERANGE;
 440
 441         *val = tmp;
 442
 443         return count;
 444 }
 445
 446 static ssize_t o2nm_cluster_idle_timeout_ms_show(struct config_item *item,
 447         char *page)
 448 {
 449         return sprintf(page, "%u\n", to_o2nm_cluster(item)->cl_idle_timeout_ms);
 450 }
 451
 452 static ssize_t o2nm_cluster_idle_timeout_ms_store(struct config_item *item,
 453         const char *page, size_t count)
 454 {
 455         struct o2nm_cluster *cluster = to_o2nm_cluster(item);
 456         ssize_t ret;
 457         unsigned int val;
 458
 459         ret =  o2nm_cluster_attr_write(page, count, &val);
 460
 461         if (ret > 0) {
 462                 if (cluster->cl_idle_timeout_ms != val
 463                         && o2net_num_connected_peers()) {
 464                         mlog(ML_NOTICE,
 465                              "o2net: cannot change idle timeout after "
 466                              "the first peer has agreed to it."
 467                              "  %d connected peers\n",
 468                              o2net_num_connected_peers());
 469                         ret = -EINVAL;
 470                 } else if (val <= cluster->cl_keepalive_delay_ms) {
 471                         mlog(ML_NOTICE, "o2net: idle timeout must be larger "
 472                              "than keepalive delay\n");
 473                         ret = -EINVAL;
 474                 } else {
 475                         cluster->cl_idle_timeout_ms = val;
 476                 }
 477         }
 478
 479         return ret;
 480 }
 481
 482 static ssize_t o2nm_cluster_keepalive_delay_ms_show(
 483         struct config_item *item, char *page)
 484 {
 485         return sprintf(page, "%u\n",
 486                         to_o2nm_cluster(item)->cl_keepalive_delay_ms);
 487 }
 488
 489 static ssize_t o2nm_cluster_keepalive_delay_ms_store(
 490         struct config_item *item, const char *page, size_t count)
 491 {
 492         struct o2nm_cluster *cluster = to_o2nm_cluster(item);
 493         ssize_t ret;
 494         unsigned int val;
 495
 496         ret =  o2nm_cluster_attr_write(page, count, &val);
 497
 498         if (ret > 0) {
 499                 if (cluster->cl_keepalive_delay_ms != val
 500                     && o2net_num_connected_peers()) {
 501                         mlog(ML_NOTICE,
 502                              "o2net: cannot change keepalive delay after"
 503                              " the first peer has agreed to it."
 504                              "  %d connected peers\n",
 505                              o2net_num_connected_peers());
 506                         ret = -EINVAL;
 507                 } else if (val >= cluster->cl_idle_timeout_ms) {
 508                         mlog(ML_NOTICE, "o2net: keepalive delay must be "
 509                              "smaller than idle timeout\n");
 510                         ret = -EINVAL;
 511                 } else {
 512                         cluster->cl_keepalive_delay_ms = val;
 513                 }
 514         }
 515
 516         return ret;
 517 }
 518
 519 static ssize_t o2nm_cluster_reconnect_delay_ms_show(
 520         struct config_item *item, char *page)
 521 {
 522         return sprintf(page, "%u\n",
 523                         to_o2nm_cluster(item)->cl_reconnect_delay_ms);
 524 }
 525
 526 static ssize_t o2nm_cluster_reconnect_delay_ms_store(
 527         struct config_item *item, const char *page, size_t count)
 528 {
 529         return o2nm_cluster_attr_write(page, count,
 530                                &to_o2nm_cluster(item)->cl_reconnect_delay_ms);
 531 }
 532
 533 static ssize_t o2nm_cluster_fence_method_show(
 534         struct config_item *item, char *page)
 535 {
 536         struct o2nm_cluster *cluster = to_o2nm_cluster(item);
 537         ssize_t ret = 0;
 538
 539         if (cluster)
 540                 ret = sprintf(page, "%s\n",
 541                               o2nm_fence_method_desc[cluster->cl_fence_method]);
 542         return ret;
 543 }
 544
 545 static ssize_t o2nm_cluster_fence_method_store(
 546         struct config_item *item, const char *page, size_t count)
 547 {
 548         unsigned int i;
 549
 550         if (page[count - 1] != '\n')
 551                 goto bail;
 552
 553         for (i = 0; i < O2NM_FENCE_METHODS; ++i) {
 554                 if (count != strlen(o2nm_fence_method_desc[i]) + 1)
 555                         continue;
 556                 if (strncasecmp(page, o2nm_fence_method_desc[i], count - 1))
 557                         continue;
 558                 if (to_o2nm_cluster(item)->cl_fence_method != i) {
 559                         printk(KERN_INFO "ocfs2: Changing fence method to %s\n",
 560                                o2nm_fence_method_desc[i]);
 561                         to_o2nm_cluster(item)->cl_fence_method = i;
 562                 }
 563                 return count;
 564         }
 565
 566 bail:
 567         return -EINVAL;
 568 }
 569
 570 CONFIGFS_ATTR(o2nm_cluster_, idle_timeout_ms);
 571 CONFIGFS_ATTR(o2nm_cluster_, keepalive_delay_ms);
 572 CONFIGFS_ATTR(o2nm_cluster_, reconnect_delay_ms);
 573 CONFIGFS_ATTR(o2nm_cluster_, fence_method);
 574
 575 static struct configfs_attribute *o2nm_cluster_attrs[] = {
 576         &o2nm_cluster_attr_idle_timeout_ms,
 577         &o2nm_cluster_attr_keepalive_delay_ms,
 578         &o2nm_cluster_attr_reconnect_delay_ms,
 579         &o2nm_cluster_attr_fence_method,
 580         NULL,
 581 };
 582
 583 static struct config_item *o2nm_node_group_make_item(struct config_group *group,
 584                                                      const char *name)
 585 {
 586         struct o2nm_node *node = NULL;
 587
 588         if (strlen(name) > O2NM_MAX_NAME_LEN)
 589                 return ERR_PTR(-ENAMETOOLONG);
 590
 591         node = kzalloc(sizeof(struct o2nm_node), GFP_KERNEL);
 592         if (node == NULL)
 593                 return ERR_PTR(-ENOMEM);
 594
 595         strcpy(node->nd_name, name); /* use item.ci_namebuf instead? */
 596         config_item_init_type_name(&node->nd_item, name, &o2nm_node_type);
 597         spin_lock_init(&node->nd_lock);
 598
 599         mlog(ML_CLUSTER, "o2nm: Registering node %s\n", name);
 600
 601         return &node->nd_item;
 602 }
 603
 604 static void o2nm_node_group_drop_item(struct config_group *group,
 605                                       struct config_item *item)
 606 {
 607         struct o2nm_node *node = to_o2nm_node(item);
 608         struct o2nm_cluster *cluster = to_o2nm_cluster(group->cg_item.ci_parent);
 609
 610         if (cluster->cl_nodes[node->nd_num] == node) {
 611                 o2net_disconnect_node(node);
 612
 613                 if (cluster->cl_has_local &&
 614                     (cluster->cl_local_node == node->nd_num)) {
 615                         cluster->cl_has_local = 0;
 616                         cluster->cl_local_node = O2NM_INVALID_NODE_NUM;
 617                         o2net_stop_listening(node);
 618                 }
 619         }
 620
 621         /* XXX call into net to stop this node from trading messages */
 622
 623         write_lock(&cluster->cl_nodes_lock);
 624
 625         /* XXX sloppy */
 626         if (node->nd_ipv4_address)
 627                 rb_erase(&node->nd_ip_node, &cluster->cl_node_ip_tree);
 628
 629         /* nd_num might be 0 if the node number hasn't been set.. */
 630         if (cluster->cl_nodes[node->nd_num] == node) {
 631                 cluster->cl_nodes[node->nd_num] = NULL;
 632                 clear_bit(node->nd_num, cluster->cl_nodes_bitmap);
 633         }
 634         write_unlock(&cluster->cl_nodes_lock);
 635
 636         mlog(ML_CLUSTER, "o2nm: Unregistered node %s\n",
 637              config_item_name(&node->nd_item));
 638
 639         config_item_put(item);
 640 }
 641
 642 static struct configfs_group_operations o2nm_node_group_group_ops = {
 643         .make_item      = o2nm_node_group_make_item,
 644         .drop_item      = o2nm_node_group_drop_item,
 645 };
 646
 647 static const struct config_item_type o2nm_node_group_type = {
 648         .ct_group_ops   = &o2nm_node_group_group_ops,
 649         .ct_owner       = THIS_MODULE,
 650 };
 651
 652 /* cluster */
 653
 654 static void o2nm_cluster_release(struct config_item *item)
 655 {
 656         struct o2nm_cluster *cluster = to_o2nm_cluster(item);
 657
 658         kfree(cluster);
 659 }
 660
 661 static struct configfs_item_operations o2nm_cluster_item_ops = {
 662         .release        = o2nm_cluster_release,
 663 };
 664
 665 static const struct config_item_type o2nm_cluster_type = {
 666         .ct_item_ops    = &o2nm_cluster_item_ops,
 667         .ct_attrs       = o2nm_cluster_attrs,
 668         .ct_owner       = THIS_MODULE,
 669 };
 670
 671 /* cluster set */
 672
 673 struct o2nm_cluster_group {
 674         struct configfs_subsystem cs_subsys;
 675         /* some stuff? */
 676 };
 677
 678 #if 0
 679 static struct o2nm_cluster_group *to_o2nm_cluster_group(struct config_group *group)
 680 {
 681         return group ?
 682                 container_of(to_configfs_subsystem(group), struct o2nm_cluster_group, cs_subsys)
 683                : NULL;
 684 }
 685 #endif
 686
 687 static struct config_group *o2nm_cluster_group_make_group(struct config_group *group,
 688                                                           const char *name)
 689 {
 690         struct o2nm_cluster *cluster = NULL;
 691         struct o2nm_node_group *ns = NULL;
 692         struct config_group *o2hb_group = NULL, *ret = NULL;
 693
 694         /* this runs under the parent dir's i_mutex; there can be only
 695          * one caller in here at a time */
 696         if (o2nm_single_cluster)
 697                 return ERR_PTR(-ENOSPC);
 698
 699         cluster = kzalloc(sizeof(struct o2nm_cluster), GFP_KERNEL);
 700         ns = kzalloc(sizeof(struct o2nm_node_group), GFP_KERNEL);
 701         o2hb_group = o2hb_alloc_hb_set();
 702         if (cluster == NULL || ns == NULL || o2hb_group == NULL)
 703                 goto out;
 704
 705         config_group_init_type_name(&cluster->cl_group, name,
 706                                     &o2nm_cluster_type);
 707         configfs_add_default_group(&ns->ns_group, &cluster->cl_group);
 708
 709         config_group_init_type_name(&ns->ns_group, "node",
 710                                     &o2nm_node_group_type);
 711         configfs_add_default_group(o2hb_group, &cluster->cl_group);
 712
 713         rwlock_init(&cluster->cl_nodes_lock);
 714         cluster->cl_node_ip_tree = RB_ROOT;
 715         cluster->cl_reconnect_delay_ms = O2NET_RECONNECT_DELAY_MS_DEFAULT;
 716         cluster->cl_idle_timeout_ms    = O2NET_IDLE_TIMEOUT_MS_DEFAULT;
 717         cluster->cl_keepalive_delay_ms = O2NET_KEEPALIVE_DELAY_MS_DEFAULT;
 718         cluster->cl_fence_method       = O2NM_FENCE_RESET;
 719
 720         ret = &cluster->cl_group;
 721         o2nm_single_cluster = cluster;
 722
 723 out:
 724         if (ret == NULL) {
 725                 kfree(cluster);
 726                 kfree(ns);
 727                 o2hb_free_hb_set(o2hb_group);
 728                 ret = ERR_PTR(-ENOMEM);
 729         }
 730
 731         return ret;
 732 }
 733
 734 static void o2nm_cluster_group_drop_item(struct config_group *group, struct config_item *item)
 735 {
 736         struct o2nm_cluster *cluster = to_o2nm_cluster(item);
 737
 738         BUG_ON(o2nm_single_cluster != cluster);
 739         o2nm_single_cluster = NULL;
 740
 741         configfs_remove_default_groups(&cluster->cl_group);
 742         config_item_put(item);
 743 }
 744
 745 static struct configfs_group_operations o2nm_cluster_group_group_ops = {
 746         .make_group     = o2nm_cluster_group_make_group,
 747         .drop_item      = o2nm_cluster_group_drop_item,
 748 };
 749
 750 static const struct config_item_type o2nm_cluster_group_type = {
 751         .ct_group_ops   = &o2nm_cluster_group_group_ops,
 752         .ct_owner       = THIS_MODULE,
 753 };
 754
 755 static struct o2nm_cluster_group o2nm_cluster_group = {
 756         .cs_subsys = {
 757                 .su_group = {
 758                         .cg_item = {
 759                                 .ci_namebuf = "cluster",
 760                                 .ci_type = &o2nm_cluster_group_type,
 761                         },
 762                 },
 763         },
 764 };
 765
 766 static inline void o2nm_lock_subsystem(void)
 767 {
 768         mutex_lock(&o2nm_cluster_group.cs_subsys.su_mutex);
 769 }
 770
 771 static inline void o2nm_unlock_subsystem(void)
 772 {
 773         mutex_unlock(&o2nm_cluster_group.cs_subsys.su_mutex);
 774 }
 775
 776 int o2nm_depend_item(struct config_item *item)
 777 {
 778         return configfs_depend_item(&o2nm_cluster_group.cs_subsys, item);
 779 }
 780
 781 void o2nm_undepend_item(struct config_item *item)
 782 {
 783         configfs_undepend_item(item);
 784 }
 785
 786 int o2nm_depend_this_node(void)
 787 {
 788         int ret = 0;
 789         struct o2nm_node *local_node;
 790
 791         local_node = o2nm_get_node_by_num(o2nm_this_node());
 792         if (!local_node) {
 793                 ret = -EINVAL;
 794                 goto out;
 795         }
 796
 797         ret = o2nm_depend_item(&local_node->nd_item);
 798         o2nm_node_put(local_node);
 799
 800 out:
 801         return ret;
 802 }
 803
 804 void o2nm_undepend_this_node(void)
 805 {
 806         struct o2nm_node *local_node;
 807
 808         local_node = o2nm_get_node_by_num(o2nm_this_node());
 809         BUG_ON(!local_node);
 810
 811         o2nm_undepend_item(&local_node->nd_item);
 812         o2nm_node_put(local_node);
 813 }
 814
 815
 816 static void __exit exit_o2nm(void)
 817 {
 818         /* XXX sync with hb callbacks and shut down hb? */
 819         o2net_unregister_hb_callbacks();
 820         configfs_unregister_subsystem(&o2nm_cluster_group.cs_subsys);
 821         o2cb_sys_shutdown();
 822
 823         o2net_exit();
 824         o2hb_exit();
 825 }
 826
 827 static int __init init_o2nm(void)
 828 {
 829         int ret = -1;
 830
 831         ret = o2hb_init();
 832         if (ret)
 833                 goto out;
 834
 835         ret = o2net_init();
 836         if (ret)
 837                 goto out_o2hb;
 838
 839         ret = o2net_register_hb_callbacks();
 840         if (ret)
 841                 goto out_o2net;
 842
 843         config_group_init(&o2nm_cluster_group.cs_subsys.su_group);
 844         mutex_init(&o2nm_cluster_group.cs_subsys.su_mutex);
 845         ret = configfs_register_subsystem(&o2nm_cluster_group.cs_subsys);
 846         if (ret) {
 847                 printk(KERN_ERR "nodemanager: Registration returned %d\n", ret);
 848                 goto out_callbacks;
 849         }
 850
 851         ret = o2cb_sys_init();
 852         if (!ret)
 853                 goto out;
 854
 855         configfs_unregister_subsystem(&o2nm_cluster_group.cs_subsys);
 856 out_callbacks:
 857         o2net_unregister_hb_callbacks();
 858 out_o2net:
 859         o2net_exit();
 860 out_o2hb:
 861         o2hb_exit();
 862 out:
 863         return ret;
 864 }
 865
 866 MODULE_AUTHOR("Oracle");
 867 MODULE_LICENSE("GPL");
 868 MODULE_DESCRIPTION("OCFS2 cluster management");
 869
 870 module_init(init_o2nm)
 871 module_exit(exit_o2nm)