Loading...
1/* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
3 *
4 * Copyright (C) 2004, 2005, 2012 Oracle. All rights reserved.
5 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public
8 * License as published by the Free Software Foundation; either
9 * version 2 of the License, or (at your option) any later version.
10 *
11 * This program is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * General Public License for more details.
15 *
16 * You should have received a copy of the GNU General Public
17 * License along with this program; if not, write to the
18 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
19 * Boston, MA 021110-1307, USA.
20 */
21
22#include <linux/slab.h>
23#include <linux/kernel.h>
24#include <linux/module.h>
25#include <linux/configfs.h>
26
27#include "tcp.h"
28#include "nodemanager.h"
29#include "heartbeat.h"
30#include "masklog.h"
31
32/* for now we operate under the assertion that there can be only one
33 * cluster active at a time. Changing this will require trickling
34 * cluster references throughout where nodes are looked up */
35struct r2nm_cluster *r2nm_single_cluster;
36
37char *r2nm_fence_method_desc[R2NM_FENCE_METHODS] = {
38 "reset", /* R2NM_FENCE_RESET */
39 "panic", /* R2NM_FENCE_PANIC */
40};
41
42struct r2nm_node *r2nm_get_node_by_num(u8 node_num)
43{
44 struct r2nm_node *node = NULL;
45
46 if (node_num >= R2NM_MAX_NODES || r2nm_single_cluster == NULL)
47 goto out;
48
49 read_lock(&r2nm_single_cluster->cl_nodes_lock);
50 node = r2nm_single_cluster->cl_nodes[node_num];
51 if (node)
52 config_item_get(&node->nd_item);
53 read_unlock(&r2nm_single_cluster->cl_nodes_lock);
54out:
55 return node;
56}
57EXPORT_SYMBOL_GPL(r2nm_get_node_by_num);
58
59int r2nm_configured_node_map(unsigned long *map, unsigned bytes)
60{
61 struct r2nm_cluster *cluster = r2nm_single_cluster;
62
63 BUG_ON(bytes < (sizeof(cluster->cl_nodes_bitmap)));
64
65 if (cluster == NULL)
66 return -EINVAL;
67
68 read_lock(&cluster->cl_nodes_lock);
69 memcpy(map, cluster->cl_nodes_bitmap, sizeof(cluster->cl_nodes_bitmap));
70 read_unlock(&cluster->cl_nodes_lock);
71
72 return 0;
73}
74EXPORT_SYMBOL_GPL(r2nm_configured_node_map);
75
76static struct r2nm_node *r2nm_node_ip_tree_lookup(struct r2nm_cluster *cluster,
77 __be32 ip_needle,
78 struct rb_node ***ret_p,
79 struct rb_node **ret_parent)
80{
81 struct rb_node **p = &cluster->cl_node_ip_tree.rb_node;
82 struct rb_node *parent = NULL;
83 struct r2nm_node *node, *ret = NULL;
84
85 while (*p) {
86 int cmp;
87
88 parent = *p;
89 node = rb_entry(parent, struct r2nm_node, nd_ip_node);
90
91 cmp = memcmp(&ip_needle, &node->nd_ipv4_address,
92 sizeof(ip_needle));
93 if (cmp < 0)
94 p = &(*p)->rb_left;
95 else if (cmp > 0)
96 p = &(*p)->rb_right;
97 else {
98 ret = node;
99 break;
100 }
101 }
102
103 if (ret_p != NULL)
104 *ret_p = p;
105 if (ret_parent != NULL)
106 *ret_parent = parent;
107
108 return ret;
109}
110
111struct r2nm_node *r2nm_get_node_by_ip(__be32 addr)
112{
113 struct r2nm_node *node = NULL;
114 struct r2nm_cluster *cluster = r2nm_single_cluster;
115
116 if (cluster == NULL)
117 goto out;
118
119 read_lock(&cluster->cl_nodes_lock);
120 node = r2nm_node_ip_tree_lookup(cluster, addr, NULL, NULL);
121 if (node)
122 config_item_get(&node->nd_item);
123 read_unlock(&cluster->cl_nodes_lock);
124
125out:
126 return node;
127}
128EXPORT_SYMBOL_GPL(r2nm_get_node_by_ip);
129
130void r2nm_node_put(struct r2nm_node *node)
131{
132 config_item_put(&node->nd_item);
133}
134EXPORT_SYMBOL_GPL(r2nm_node_put);
135
136void r2nm_node_get(struct r2nm_node *node)
137{
138 config_item_get(&node->nd_item);
139}
140EXPORT_SYMBOL_GPL(r2nm_node_get);
141
142u8 r2nm_this_node(void)
143{
144 u8 node_num = R2NM_MAX_NODES;
145
146 if (r2nm_single_cluster && r2nm_single_cluster->cl_has_local)
147 node_num = r2nm_single_cluster->cl_local_node;
148
149 return node_num;
150}
151EXPORT_SYMBOL_GPL(r2nm_this_node);
152
153/* node configfs bits */
154
155static struct r2nm_cluster *to_r2nm_cluster(struct config_item *item)
156{
157 return item ?
158 container_of(to_config_group(item), struct r2nm_cluster,
159 cl_group)
160 : NULL;
161}
162
163static struct r2nm_node *to_r2nm_node(struct config_item *item)
164{
165 return item ? container_of(item, struct r2nm_node, nd_item) : NULL;
166}
167
168static void r2nm_node_release(struct config_item *item)
169{
170 struct r2nm_node *node = to_r2nm_node(item);
171 kfree(node);
172}
173
174static ssize_t r2nm_node_num_read(struct r2nm_node *node, char *page)
175{
176 return sprintf(page, "%d\n", node->nd_num);
177}
178
179static struct r2nm_cluster *to_r2nm_cluster_from_node(struct r2nm_node *node)
180{
181 /* through the first node_set .parent
182 * mycluster/nodes/mynode == r2nm_cluster->r2nm_node_group->r2nm_node */
183 return to_r2nm_cluster(node->nd_item.ci_parent->ci_parent);
184}
185
186enum {
187 R2NM_NODE_ATTR_NUM = 0,
188 R2NM_NODE_ATTR_PORT,
189 R2NM_NODE_ATTR_ADDRESS,
190 R2NM_NODE_ATTR_LOCAL,
191};
192
193static ssize_t r2nm_node_num_write(struct r2nm_node *node, const char *page,
194 size_t count)
195{
196 struct r2nm_cluster *cluster = to_r2nm_cluster_from_node(node);
197 unsigned long tmp;
198 char *p = (char *)page;
199 int err;
200
201 err = kstrtoul(p, 10, &tmp);
202 if (err)
203 return err;
204
205 if (tmp >= R2NM_MAX_NODES)
206 return -ERANGE;
207
208 /* once we're in the cl_nodes tree networking can look us up by
209 * node number and try to use our address and port attributes
210 * to connect to this node.. make sure that they've been set
211 * before writing the node attribute? */
212 if (!test_bit(R2NM_NODE_ATTR_ADDRESS, &node->nd_set_attributes) ||
213 !test_bit(R2NM_NODE_ATTR_PORT, &node->nd_set_attributes))
214 return -EINVAL; /* XXX */
215
216 write_lock(&cluster->cl_nodes_lock);
217 if (cluster->cl_nodes[tmp])
218 p = NULL;
219 else {
220 cluster->cl_nodes[tmp] = node;
221 node->nd_num = tmp;
222 set_bit(tmp, cluster->cl_nodes_bitmap);
223 }
224 write_unlock(&cluster->cl_nodes_lock);
225 if (p == NULL)
226 return -EEXIST;
227
228 return count;
229}
230static ssize_t r2nm_node_ipv4_port_read(struct r2nm_node *node, char *page)
231{
232 return sprintf(page, "%u\n", ntohs(node->nd_ipv4_port));
233}
234
235static ssize_t r2nm_node_ipv4_port_write(struct r2nm_node *node,
236 const char *page, size_t count)
237{
238 unsigned long tmp;
239 char *p = (char *)page;
240 int err;
241
242 err = kstrtoul(p, 10, &tmp);
243 if (err)
244 return err;
245
246 if (tmp == 0)
247 return -EINVAL;
248 if (tmp >= (u16)-1)
249 return -ERANGE;
250
251 node->nd_ipv4_port = htons(tmp);
252
253 return count;
254}
255
256static ssize_t r2nm_node_ipv4_address_read(struct r2nm_node *node, char *page)
257{
258 return sprintf(page, "%pI4\n", &node->nd_ipv4_address);
259}
260
261static ssize_t r2nm_node_ipv4_address_write(struct r2nm_node *node,
262 const char *page,
263 size_t count)
264{
265 struct r2nm_cluster *cluster = to_r2nm_cluster_from_node(node);
266 int ret, i;
267 struct rb_node **p, *parent;
268 unsigned int octets[4];
269 __be32 ipv4_addr = 0;
270
271 ret = sscanf(page, "%3u.%3u.%3u.%3u", &octets[3], &octets[2],
272 &octets[1], &octets[0]);
273 if (ret != 4)
274 return -EINVAL;
275
276 for (i = 0; i < ARRAY_SIZE(octets); i++) {
277 if (octets[i] > 255)
278 return -ERANGE;
279 be32_add_cpu(&ipv4_addr, octets[i] << (i * 8));
280 }
281
282 ret = 0;
283 write_lock(&cluster->cl_nodes_lock);
284 if (r2nm_node_ip_tree_lookup(cluster, ipv4_addr, &p, &parent))
285 ret = -EEXIST;
286 else {
287 rb_link_node(&node->nd_ip_node, parent, p);
288 rb_insert_color(&node->nd_ip_node, &cluster->cl_node_ip_tree);
289 }
290 write_unlock(&cluster->cl_nodes_lock);
291 if (ret)
292 return ret;
293
294 memcpy(&node->nd_ipv4_address, &ipv4_addr, sizeof(ipv4_addr));
295
296 return count;
297}
298
299static ssize_t r2nm_node_local_read(struct r2nm_node *node, char *page)
300{
301 return sprintf(page, "%d\n", node->nd_local);
302}
303
304static ssize_t r2nm_node_local_write(struct r2nm_node *node, const char *page,
305 size_t count)
306{
307 struct r2nm_cluster *cluster = to_r2nm_cluster_from_node(node);
308 unsigned long tmp;
309 char *p = (char *)page;
310 ssize_t ret;
311 int err;
312
313 err = kstrtoul(p, 10, &tmp);
314 if (err)
315 return err;
316
317 tmp = !!tmp; /* boolean of whether this node wants to be local */
318
319 /* setting local turns on networking rx for now so we require having
320 * set everything else first */
321 if (!test_bit(R2NM_NODE_ATTR_ADDRESS, &node->nd_set_attributes) ||
322 !test_bit(R2NM_NODE_ATTR_NUM, &node->nd_set_attributes) ||
323 !test_bit(R2NM_NODE_ATTR_PORT, &node->nd_set_attributes))
324 return -EINVAL; /* XXX */
325
326 /* the only failure case is trying to set a new local node
327 * when a different one is already set */
328 if (tmp && tmp == cluster->cl_has_local &&
329 cluster->cl_local_node != node->nd_num)
330 return -EBUSY;
331
332 /* bring up the rx thread if we're setting the new local node. */
333 if (tmp && !cluster->cl_has_local) {
334 ret = r2net_start_listening(node);
335 if (ret)
336 return ret;
337 }
338
339 if (!tmp && cluster->cl_has_local &&
340 cluster->cl_local_node == node->nd_num) {
341 r2net_stop_listening(node);
342 cluster->cl_local_node = R2NM_INVALID_NODE_NUM;
343 }
344
345 node->nd_local = tmp;
346 if (node->nd_local) {
347 cluster->cl_has_local = tmp;
348 cluster->cl_local_node = node->nd_num;
349 }
350
351 return count;
352}
353
354struct r2nm_node_attribute {
355 struct configfs_attribute attr;
356 ssize_t (*show)(struct r2nm_node *, char *);
357 ssize_t (*store)(struct r2nm_node *, const char *, size_t);
358};
359
360static struct r2nm_node_attribute r2nm_node_attr_num = {
361 .attr = { .ca_owner = THIS_MODULE,
362 .ca_name = "num",
363 .ca_mode = S_IRUGO | S_IWUSR },
364 .show = r2nm_node_num_read,
365 .store = r2nm_node_num_write,
366};
367
368static struct r2nm_node_attribute r2nm_node_attr_ipv4_port = {
369 .attr = { .ca_owner = THIS_MODULE,
370 .ca_name = "ipv4_port",
371 .ca_mode = S_IRUGO | S_IWUSR },
372 .show = r2nm_node_ipv4_port_read,
373 .store = r2nm_node_ipv4_port_write,
374};
375
376static struct r2nm_node_attribute r2nm_node_attr_ipv4_address = {
377 .attr = { .ca_owner = THIS_MODULE,
378 .ca_name = "ipv4_address",
379 .ca_mode = S_IRUGO | S_IWUSR },
380 .show = r2nm_node_ipv4_address_read,
381 .store = r2nm_node_ipv4_address_write,
382};
383
384static struct r2nm_node_attribute r2nm_node_attr_local = {
385 .attr = { .ca_owner = THIS_MODULE,
386 .ca_name = "local",
387 .ca_mode = S_IRUGO | S_IWUSR },
388 .show = r2nm_node_local_read,
389 .store = r2nm_node_local_write,
390};
391
392static struct configfs_attribute *r2nm_node_attrs[] = {
393 [R2NM_NODE_ATTR_NUM] = &r2nm_node_attr_num.attr,
394 [R2NM_NODE_ATTR_PORT] = &r2nm_node_attr_ipv4_port.attr,
395 [R2NM_NODE_ATTR_ADDRESS] = &r2nm_node_attr_ipv4_address.attr,
396 [R2NM_NODE_ATTR_LOCAL] = &r2nm_node_attr_local.attr,
397 NULL,
398};
399
400static int r2nm_attr_index(struct configfs_attribute *attr)
401{
402 int i;
403 for (i = 0; i < ARRAY_SIZE(r2nm_node_attrs); i++) {
404 if (attr == r2nm_node_attrs[i])
405 return i;
406 }
407 BUG();
408 return 0;
409}
410
411static ssize_t r2nm_node_show(struct config_item *item,
412 struct configfs_attribute *attr,
413 char *page)
414{
415 struct r2nm_node *node = to_r2nm_node(item);
416 struct r2nm_node_attribute *r2nm_node_attr =
417 container_of(attr, struct r2nm_node_attribute, attr);
418 ssize_t ret = 0;
419
420 if (r2nm_node_attr->show)
421 ret = r2nm_node_attr->show(node, page);
422 return ret;
423}
424
425static ssize_t r2nm_node_store(struct config_item *item,
426 struct configfs_attribute *attr,
427 const char *page, size_t count)
428{
429 struct r2nm_node *node = to_r2nm_node(item);
430 struct r2nm_node_attribute *r2nm_node_attr =
431 container_of(attr, struct r2nm_node_attribute, attr);
432 ssize_t ret;
433 int attr_index = r2nm_attr_index(attr);
434
435 if (r2nm_node_attr->store == NULL) {
436 ret = -EINVAL;
437 goto out;
438 }
439
440 if (test_bit(attr_index, &node->nd_set_attributes))
441 return -EBUSY;
442
443 ret = r2nm_node_attr->store(node, page, count);
444 if (ret < count)
445 goto out;
446
447 set_bit(attr_index, &node->nd_set_attributes);
448out:
449 return ret;
450}
451
452static struct configfs_item_operations r2nm_node_item_ops = {
453 .release = r2nm_node_release,
454 .show_attribute = r2nm_node_show,
455 .store_attribute = r2nm_node_store,
456};
457
458static struct config_item_type r2nm_node_type = {
459 .ct_item_ops = &r2nm_node_item_ops,
460 .ct_attrs = r2nm_node_attrs,
461 .ct_owner = THIS_MODULE,
462};
463
464/* node set */
465
466struct r2nm_node_group {
467 struct config_group ns_group;
468 /* some stuff? */
469};
470
471#if 0
472static struct r2nm_node_group *to_r2nm_node_group(struct config_group *group)
473{
474 return group ?
475 container_of(group, struct r2nm_node_group, ns_group)
476 : NULL;
477}
478#endif
479
480struct r2nm_cluster_attribute {
481 struct configfs_attribute attr;
482 ssize_t (*show)(struct r2nm_cluster *, char *);
483 ssize_t (*store)(struct r2nm_cluster *, const char *, size_t);
484};
485
486static ssize_t r2nm_cluster_attr_write(const char *page, ssize_t count,
487 unsigned int *val)
488{
489 unsigned long tmp;
490 char *p = (char *)page;
491 int err;
492
493 err = kstrtoul(p, 10, &tmp);
494 if (err)
495 return err;
496
497 if (tmp == 0)
498 return -EINVAL;
499 if (tmp >= (u32)-1)
500 return -ERANGE;
501
502 *val = tmp;
503
504 return count;
505}
506
507static ssize_t r2nm_cluster_attr_idle_timeout_ms_read(
508 struct r2nm_cluster *cluster, char *page)
509{
510 return sprintf(page, "%u\n", cluster->cl_idle_timeout_ms);
511}
512
513static ssize_t r2nm_cluster_attr_idle_timeout_ms_write(
514 struct r2nm_cluster *cluster, const char *page, size_t count)
515{
516 ssize_t ret;
517 unsigned int val = 0;
518
519 ret = r2nm_cluster_attr_write(page, count, &val);
520
521 if (ret > 0) {
522 if (cluster->cl_idle_timeout_ms != val
523 && r2net_num_connected_peers()) {
524 mlog(ML_NOTICE,
525 "r2net: cannot change idle timeout after "
526 "the first peer has agreed to it."
527 " %d connected peers\n",
528 r2net_num_connected_peers());
529 ret = -EINVAL;
530 } else if (val <= cluster->cl_keepalive_delay_ms) {
531 mlog(ML_NOTICE, "r2net: idle timeout must be larger "
532 "than keepalive delay\n");
533 ret = -EINVAL;
534 } else {
535 cluster->cl_idle_timeout_ms = val;
536 }
537 }
538
539 return ret;
540}
541
542static ssize_t r2nm_cluster_attr_keepalive_delay_ms_read(
543 struct r2nm_cluster *cluster, char *page)
544{
545 return sprintf(page, "%u\n", cluster->cl_keepalive_delay_ms);
546}
547
548static ssize_t r2nm_cluster_attr_keepalive_delay_ms_write(
549 struct r2nm_cluster *cluster, const char *page, size_t count)
550{
551 ssize_t ret;
552 unsigned int val = 0;
553
554 ret = r2nm_cluster_attr_write(page, count, &val);
555
556 if (ret > 0) {
557 if (cluster->cl_keepalive_delay_ms != val
558 && r2net_num_connected_peers()) {
559 mlog(ML_NOTICE,
560 "r2net: cannot change keepalive delay after"
561 " the first peer has agreed to it."
562 " %d connected peers\n",
563 r2net_num_connected_peers());
564 ret = -EINVAL;
565 } else if (val >= cluster->cl_idle_timeout_ms) {
566 mlog(ML_NOTICE, "r2net: keepalive delay must be "
567 "smaller than idle timeout\n");
568 ret = -EINVAL;
569 } else {
570 cluster->cl_keepalive_delay_ms = val;
571 }
572 }
573
574 return ret;
575}
576
577static ssize_t r2nm_cluster_attr_reconnect_delay_ms_read(
578 struct r2nm_cluster *cluster, char *page)
579{
580 return sprintf(page, "%u\n", cluster->cl_reconnect_delay_ms);
581}
582
583static ssize_t r2nm_cluster_attr_reconnect_delay_ms_write(
584 struct r2nm_cluster *cluster, const char *page, size_t count)
585{
586 return r2nm_cluster_attr_write(page, count,
587 &cluster->cl_reconnect_delay_ms);
588}
589
590static ssize_t r2nm_cluster_attr_fence_method_read(
591 struct r2nm_cluster *cluster, char *page)
592{
593 ssize_t ret = 0;
594
595 if (cluster)
596 ret = sprintf(page, "%s\n",
597 r2nm_fence_method_desc[cluster->cl_fence_method]);
598 return ret;
599}
600
601static ssize_t r2nm_cluster_attr_fence_method_write(
602 struct r2nm_cluster *cluster, const char *page, size_t count)
603{
604 unsigned int i;
605
606 if (page[count - 1] != '\n')
607 goto bail;
608
609 for (i = 0; i < R2NM_FENCE_METHODS; ++i) {
610 if (count != strlen(r2nm_fence_method_desc[i]) + 1)
611 continue;
612 if (strncasecmp(page, r2nm_fence_method_desc[i], count - 1))
613 continue;
614 if (cluster->cl_fence_method != i) {
615 printk(KERN_INFO "ramster: Changing fence method to %s\n",
616 r2nm_fence_method_desc[i]);
617 cluster->cl_fence_method = i;
618 }
619 return count;
620 }
621
622bail:
623 return -EINVAL;
624}
625
626static struct r2nm_cluster_attribute r2nm_cluster_attr_idle_timeout_ms = {
627 .attr = { .ca_owner = THIS_MODULE,
628 .ca_name = "idle_timeout_ms",
629 .ca_mode = S_IRUGO | S_IWUSR },
630 .show = r2nm_cluster_attr_idle_timeout_ms_read,
631 .store = r2nm_cluster_attr_idle_timeout_ms_write,
632};
633
634static struct r2nm_cluster_attribute r2nm_cluster_attr_keepalive_delay_ms = {
635 .attr = { .ca_owner = THIS_MODULE,
636 .ca_name = "keepalive_delay_ms",
637 .ca_mode = S_IRUGO | S_IWUSR },
638 .show = r2nm_cluster_attr_keepalive_delay_ms_read,
639 .store = r2nm_cluster_attr_keepalive_delay_ms_write,
640};
641
642static struct r2nm_cluster_attribute r2nm_cluster_attr_reconnect_delay_ms = {
643 .attr = { .ca_owner = THIS_MODULE,
644 .ca_name = "reconnect_delay_ms",
645 .ca_mode = S_IRUGO | S_IWUSR },
646 .show = r2nm_cluster_attr_reconnect_delay_ms_read,
647 .store = r2nm_cluster_attr_reconnect_delay_ms_write,
648};
649
650static struct r2nm_cluster_attribute r2nm_cluster_attr_fence_method = {
651 .attr = { .ca_owner = THIS_MODULE,
652 .ca_name = "fence_method",
653 .ca_mode = S_IRUGO | S_IWUSR },
654 .show = r2nm_cluster_attr_fence_method_read,
655 .store = r2nm_cluster_attr_fence_method_write,
656};
657
658static struct configfs_attribute *r2nm_cluster_attrs[] = {
659 &r2nm_cluster_attr_idle_timeout_ms.attr,
660 &r2nm_cluster_attr_keepalive_delay_ms.attr,
661 &r2nm_cluster_attr_reconnect_delay_ms.attr,
662 &r2nm_cluster_attr_fence_method.attr,
663 NULL,
664};
665static ssize_t r2nm_cluster_show(struct config_item *item,
666 struct configfs_attribute *attr,
667 char *page)
668{
669 struct r2nm_cluster *cluster = to_r2nm_cluster(item);
670 struct r2nm_cluster_attribute *r2nm_cluster_attr =
671 container_of(attr, struct r2nm_cluster_attribute, attr);
672 ssize_t ret = 0;
673
674 if (r2nm_cluster_attr->show)
675 ret = r2nm_cluster_attr->show(cluster, page);
676 return ret;
677}
678
679static ssize_t r2nm_cluster_store(struct config_item *item,
680 struct configfs_attribute *attr,
681 const char *page, size_t count)
682{
683 struct r2nm_cluster *cluster = to_r2nm_cluster(item);
684 struct r2nm_cluster_attribute *r2nm_cluster_attr =
685 container_of(attr, struct r2nm_cluster_attribute, attr);
686 ssize_t ret;
687
688 if (r2nm_cluster_attr->store == NULL) {
689 ret = -EINVAL;
690 goto out;
691 }
692
693 ret = r2nm_cluster_attr->store(cluster, page, count);
694 if (ret < count)
695 goto out;
696out:
697 return ret;
698}
699
700static struct config_item *r2nm_node_group_make_item(struct config_group *group,
701 const char *name)
702{
703 struct r2nm_node *node = NULL;
704
705 if (strlen(name) > R2NM_MAX_NAME_LEN)
706 return ERR_PTR(-ENAMETOOLONG);
707
708 node = kzalloc(sizeof(struct r2nm_node), GFP_KERNEL);
709 if (node == NULL)
710 return ERR_PTR(-ENOMEM);
711
712 strcpy(node->nd_name, name); /* use item.ci_namebuf instead? */
713 config_item_init_type_name(&node->nd_item, name, &r2nm_node_type);
714 spin_lock_init(&node->nd_lock);
715
716 mlog(ML_CLUSTER, "r2nm: Registering node %s\n", name);
717
718 return &node->nd_item;
719}
720
721static void r2nm_node_group_drop_item(struct config_group *group,
722 struct config_item *item)
723{
724 struct r2nm_node *node = to_r2nm_node(item);
725 struct r2nm_cluster *cluster =
726 to_r2nm_cluster(group->cg_item.ci_parent);
727
728 r2net_disconnect_node(node);
729
730 if (cluster->cl_has_local &&
731 (cluster->cl_local_node == node->nd_num)) {
732 cluster->cl_has_local = 0;
733 cluster->cl_local_node = R2NM_INVALID_NODE_NUM;
734 r2net_stop_listening(node);
735 }
736
737 /* XXX call into net to stop this node from trading messages */
738
739 write_lock(&cluster->cl_nodes_lock);
740
741 /* XXX sloppy */
742 if (node->nd_ipv4_address)
743 rb_erase(&node->nd_ip_node, &cluster->cl_node_ip_tree);
744
745 /* nd_num might be 0 if the node number hasn't been set.. */
746 if (cluster->cl_nodes[node->nd_num] == node) {
747 cluster->cl_nodes[node->nd_num] = NULL;
748 clear_bit(node->nd_num, cluster->cl_nodes_bitmap);
749 }
750 write_unlock(&cluster->cl_nodes_lock);
751
752 mlog(ML_CLUSTER, "r2nm: Unregistered node %s\n",
753 config_item_name(&node->nd_item));
754
755 config_item_put(item);
756}
757
758static struct configfs_group_operations r2nm_node_group_group_ops = {
759 .make_item = r2nm_node_group_make_item,
760 .drop_item = r2nm_node_group_drop_item,
761};
762
763static struct config_item_type r2nm_node_group_type = {
764 .ct_group_ops = &r2nm_node_group_group_ops,
765 .ct_owner = THIS_MODULE,
766};
767
768/* cluster */
769
770static void r2nm_cluster_release(struct config_item *item)
771{
772 struct r2nm_cluster *cluster = to_r2nm_cluster(item);
773
774 kfree(cluster->cl_group.default_groups);
775 kfree(cluster);
776}
777
778static struct configfs_item_operations r2nm_cluster_item_ops = {
779 .release = r2nm_cluster_release,
780 .show_attribute = r2nm_cluster_show,
781 .store_attribute = r2nm_cluster_store,
782};
783
784static struct config_item_type r2nm_cluster_type = {
785 .ct_item_ops = &r2nm_cluster_item_ops,
786 .ct_attrs = r2nm_cluster_attrs,
787 .ct_owner = THIS_MODULE,
788};
789
790/* cluster set */
791
792struct r2nm_cluster_group {
793 struct configfs_subsystem cs_subsys;
794 /* some stuff? */
795};
796
797#if 0
798static struct r2nm_cluster_group *
799to_r2nm_cluster_group(struct config_group *group)
800{
801 return group ?
802 container_of(to_configfs_subsystem(group),
803 struct r2nm_cluster_group, cs_subsys)
804 : NULL;
805}
806#endif
807
808static struct config_group *
809r2nm_cluster_group_make_group(struct config_group *group,
810 const char *name)
811{
812 struct r2nm_cluster *cluster = NULL;
813 struct r2nm_node_group *ns = NULL;
814 struct config_group *r2hb_group = NULL, *ret = NULL;
815 void *defs = NULL;
816
817 /* this runs under the parent dir's i_mutex; there can be only
818 * one caller in here at a time */
819 if (r2nm_single_cluster)
820 return ERR_PTR(-ENOSPC);
821
822 cluster = kzalloc(sizeof(struct r2nm_cluster), GFP_KERNEL);
823 ns = kzalloc(sizeof(struct r2nm_node_group), GFP_KERNEL);
824 defs = kcalloc(3, sizeof(struct config_group *), GFP_KERNEL);
825 r2hb_group = r2hb_alloc_hb_set();
826 if (cluster == NULL || ns == NULL || r2hb_group == NULL || defs == NULL)
827 goto out;
828
829 config_group_init_type_name(&cluster->cl_group, name,
830 &r2nm_cluster_type);
831 config_group_init_type_name(&ns->ns_group, "node",
832 &r2nm_node_group_type);
833
834 cluster->cl_group.default_groups = defs;
835 cluster->cl_group.default_groups[0] = &ns->ns_group;
836 cluster->cl_group.default_groups[1] = r2hb_group;
837 cluster->cl_group.default_groups[2] = NULL;
838 rwlock_init(&cluster->cl_nodes_lock);
839 cluster->cl_node_ip_tree = RB_ROOT;
840 cluster->cl_reconnect_delay_ms = R2NET_RECONNECT_DELAY_MS_DEFAULT;
841 cluster->cl_idle_timeout_ms = R2NET_IDLE_TIMEOUT_MS_DEFAULT;
842 cluster->cl_keepalive_delay_ms = R2NET_KEEPALIVE_DELAY_MS_DEFAULT;
843 cluster->cl_fence_method = R2NM_FENCE_RESET;
844
845 ret = &cluster->cl_group;
846 r2nm_single_cluster = cluster;
847
848out:
849 if (ret == NULL) {
850 kfree(cluster);
851 kfree(ns);
852 r2hb_free_hb_set(r2hb_group);
853 kfree(defs);
854 ret = ERR_PTR(-ENOMEM);
855 }
856
857 return ret;
858}
859
860static void r2nm_cluster_group_drop_item(struct config_group *group,
861 struct config_item *item)
862{
863 struct r2nm_cluster *cluster = to_r2nm_cluster(item);
864 int i;
865 struct config_item *killme;
866
867 BUG_ON(r2nm_single_cluster != cluster);
868 r2nm_single_cluster = NULL;
869
870 for (i = 0; cluster->cl_group.default_groups[i]; i++) {
871 killme = &cluster->cl_group.default_groups[i]->cg_item;
872 cluster->cl_group.default_groups[i] = NULL;
873 config_item_put(killme);
874 }
875
876 config_item_put(item);
877}
878
879static struct configfs_group_operations r2nm_cluster_group_group_ops = {
880 .make_group = r2nm_cluster_group_make_group,
881 .drop_item = r2nm_cluster_group_drop_item,
882};
883
884static struct config_item_type r2nm_cluster_group_type = {
885 .ct_group_ops = &r2nm_cluster_group_group_ops,
886 .ct_owner = THIS_MODULE,
887};
888
889static struct r2nm_cluster_group r2nm_cluster_group = {
890 .cs_subsys = {
891 .su_group = {
892 .cg_item = {
893 .ci_namebuf = "cluster",
894 .ci_type = &r2nm_cluster_group_type,
895 },
896 },
897 },
898};
899
900int r2nm_depend_item(struct config_item *item)
901{
902 return configfs_depend_item(&r2nm_cluster_group.cs_subsys, item);
903}
904
905void r2nm_undepend_item(struct config_item *item)
906{
907 configfs_undepend_item(&r2nm_cluster_group.cs_subsys, item);
908}
909
910int r2nm_depend_this_node(void)
911{
912 int ret = 0;
913 struct r2nm_node *local_node;
914
915 local_node = r2nm_get_node_by_num(r2nm_this_node());
916 if (!local_node) {
917 ret = -EINVAL;
918 goto out;
919 }
920
921 ret = r2nm_depend_item(&local_node->nd_item);
922 r2nm_node_put(local_node);
923
924out:
925 return ret;
926}
927
928void r2nm_undepend_this_node(void)
929{
930 struct r2nm_node *local_node;
931
932 local_node = r2nm_get_node_by_num(r2nm_this_node());
933 BUG_ON(!local_node);
934
935 r2nm_undepend_item(&local_node->nd_item);
936 r2nm_node_put(local_node);
937}
938
939
940static void __exit exit_r2nm(void)
941{
942 /* XXX sync with hb callbacks and shut down hb? */
943 r2net_unregister_hb_callbacks();
944 configfs_unregister_subsystem(&r2nm_cluster_group.cs_subsys);
945
946 r2net_exit();
947 r2hb_exit();
948}
949
950static int __init init_r2nm(void)
951{
952 int ret = -1;
953
954 ret = r2hb_init();
955 if (ret)
956 goto out;
957
958 ret = r2net_init();
959 if (ret)
960 goto out_r2hb;
961
962 ret = r2net_register_hb_callbacks();
963 if (ret)
964 goto out_r2net;
965
966 config_group_init(&r2nm_cluster_group.cs_subsys.su_group);
967 mutex_init(&r2nm_cluster_group.cs_subsys.su_mutex);
968 ret = configfs_register_subsystem(&r2nm_cluster_group.cs_subsys);
969 if (ret) {
970 printk(KERN_ERR "nodemanager: Registration returned %d\n", ret);
971 goto out_callbacks;
972 }
973
974 if (!ret)
975 goto out;
976
977 configfs_unregister_subsystem(&r2nm_cluster_group.cs_subsys);
978out_callbacks:
979 r2net_unregister_hb_callbacks();
980out_r2net:
981 r2net_exit();
982out_r2hb:
983 r2hb_exit();
984out:
985 return ret;
986}
987
988MODULE_AUTHOR("Oracle");
989MODULE_LICENSE("GPL");
990
991module_init(init_r2nm)
992module_exit(exit_r2nm)