Merge branch 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux...
[pandora-kernel.git] / fs / ocfs2 / cluster / nodemanager.c
1 /* -*- mode: c; c-basic-offset: 8; -*-
2  * vim: noexpandtab sw=8 ts=8 sts=0:
3  *
4  * Copyright (C) 2004, 2005 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/kernel.h>
23 #include <linux/module.h>
24 #include <linux/sysctl.h>
25 #include <linux/configfs.h>
26
27 #include "tcp.h"
28 #include "nodemanager.h"
29 #include "heartbeat.h"
30 #include "masklog.h"
31 #include "sys.h"
32 #include "ver.h"
33
34 /* for now we operate under the assertion that there can be only one
35  * cluster active at a time.  Changing this will require trickling
36  * cluster references throughout where nodes are looked up */
37 struct o2nm_cluster *o2nm_single_cluster = NULL;
38
39 #define OCFS2_MAX_HB_CTL_PATH 256
40 static char ocfs2_hb_ctl_path[OCFS2_MAX_HB_CTL_PATH] = "/sbin/ocfs2_hb_ctl";
41
42 static ctl_table ocfs2_nm_table[] = {
43         {
44                 .ctl_name       = 1,
45                 .procname       = "hb_ctl_path",
46                 .data           = ocfs2_hb_ctl_path,
47                 .maxlen         = OCFS2_MAX_HB_CTL_PATH,
48                 .mode           = 0644,
49                 .proc_handler   = &proc_dostring,
50                 .strategy       = &sysctl_string,
51         },
52         { .ctl_name = 0 }
53 };
54
55 static ctl_table ocfs2_mod_table[] = {
56         {
57                 .ctl_name       = FS_OCFS2_NM,
58                 .procname       = "nm",
59                 .data           = NULL,
60                 .maxlen         = 0,
61                 .mode           = 0555,
62                 .child          = ocfs2_nm_table
63         },
64         { .ctl_name = 0}
65 };
66
67 static ctl_table ocfs2_kern_table[] = {
68         {
69                 .ctl_name       = FS_OCFS2,
70                 .procname       = "ocfs2",
71                 .data           = NULL,
72                 .maxlen         = 0,
73                 .mode           = 0555,
74                 .child          = ocfs2_mod_table
75         },
76         { .ctl_name = 0}
77 };
78
79 static ctl_table ocfs2_root_table[] = {
80         {
81                 .ctl_name       = CTL_FS,
82                 .procname       = "fs",
83                 .data           = NULL,
84                 .maxlen         = 0,
85                 .mode           = 0555,
86                 .child          = ocfs2_kern_table
87         },
88         { .ctl_name = 0 }
89 };
90
91 static struct ctl_table_header *ocfs2_table_header = NULL;
92
93 const char *o2nm_get_hb_ctl_path(void)
94 {
95         return ocfs2_hb_ctl_path;
96 }
97 EXPORT_SYMBOL_GPL(o2nm_get_hb_ctl_path);
98
99 struct o2nm_node *o2nm_get_node_by_num(u8 node_num)
100 {
101         struct o2nm_node *node = NULL;
102
103         if (node_num >= O2NM_MAX_NODES || o2nm_single_cluster == NULL)
104                 goto out;
105
106         read_lock(&o2nm_single_cluster->cl_nodes_lock);
107         node = o2nm_single_cluster->cl_nodes[node_num];
108         if (node)
109                 config_item_get(&node->nd_item);
110         read_unlock(&o2nm_single_cluster->cl_nodes_lock);
111 out:
112         return node;
113 }
114 EXPORT_SYMBOL_GPL(o2nm_get_node_by_num);
115
116 int o2nm_configured_node_map(unsigned long *map, unsigned bytes)
117 {
118         struct o2nm_cluster *cluster = o2nm_single_cluster;
119
120         BUG_ON(bytes < (sizeof(cluster->cl_nodes_bitmap)));
121
122         if (cluster == NULL)
123                 return -EINVAL;
124
125         read_lock(&cluster->cl_nodes_lock);
126         memcpy(map, cluster->cl_nodes_bitmap, sizeof(cluster->cl_nodes_bitmap));
127         read_unlock(&cluster->cl_nodes_lock);
128
129         return 0;
130 }
131 EXPORT_SYMBOL_GPL(o2nm_configured_node_map);
132
133 static struct o2nm_node *o2nm_node_ip_tree_lookup(struct o2nm_cluster *cluster,
134                                                   __be32 ip_needle,
135                                                   struct rb_node ***ret_p,
136                                                   struct rb_node **ret_parent)
137 {
138         struct rb_node **p = &cluster->cl_node_ip_tree.rb_node;
139         struct rb_node *parent = NULL;
140         struct o2nm_node *node, *ret = NULL;
141
142         while (*p) {
143                 int cmp;
144
145                 parent = *p;
146                 node = rb_entry(parent, struct o2nm_node, nd_ip_node);
147
148                 cmp = memcmp(&ip_needle, &node->nd_ipv4_address,
149                                 sizeof(ip_needle));
150                 if (cmp < 0)
151                         p = &(*p)->rb_left;
152                 else if (cmp > 0)
153                         p = &(*p)->rb_right;
154                 else {
155                         ret = node;
156                         break;
157                 }
158         }
159
160         if (ret_p != NULL)
161                 *ret_p = p;
162         if (ret_parent != NULL)
163                 *ret_parent = parent;
164
165         return ret;
166 }
167
168 struct o2nm_node *o2nm_get_node_by_ip(__be32 addr)
169 {
170         struct o2nm_node *node = NULL;
171         struct o2nm_cluster *cluster = o2nm_single_cluster;
172
173         if (cluster == NULL)
174                 goto out;
175
176         read_lock(&cluster->cl_nodes_lock);
177         node = o2nm_node_ip_tree_lookup(cluster, addr, NULL, NULL);
178         if (node)
179                 config_item_get(&node->nd_item);
180         read_unlock(&cluster->cl_nodes_lock);
181
182 out:
183         return node;
184 }
185 EXPORT_SYMBOL_GPL(o2nm_get_node_by_ip);
186
187 void o2nm_node_put(struct o2nm_node *node)
188 {
189         config_item_put(&node->nd_item);
190 }
191 EXPORT_SYMBOL_GPL(o2nm_node_put);
192
193 void o2nm_node_get(struct o2nm_node *node)
194 {
195         config_item_get(&node->nd_item);
196 }
197 EXPORT_SYMBOL_GPL(o2nm_node_get);
198
199 u8 o2nm_this_node(void)
200 {
201         u8 node_num = O2NM_MAX_NODES;
202
203         if (o2nm_single_cluster && o2nm_single_cluster->cl_has_local)
204                 node_num = o2nm_single_cluster->cl_local_node;
205
206         return node_num;
207 }
208 EXPORT_SYMBOL_GPL(o2nm_this_node);
209
210 /* node configfs bits */
211
212 static struct o2nm_cluster *to_o2nm_cluster(struct config_item *item)
213 {
214         return item ?
215                 container_of(to_config_group(item), struct o2nm_cluster,
216                              cl_group)
217                 : NULL;
218 }
219
220 static struct o2nm_node *to_o2nm_node(struct config_item *item)
221 {
222         return item ? container_of(item, struct o2nm_node, nd_item) : NULL;
223 }
224
225 static void o2nm_node_release(struct config_item *item)
226 {
227         struct o2nm_node *node = to_o2nm_node(item);
228         kfree(node);
229 }
230
231 static ssize_t o2nm_node_num_read(struct o2nm_node *node, char *page)
232 {
233         return sprintf(page, "%d\n", node->nd_num);
234 }
235
236 static struct o2nm_cluster *to_o2nm_cluster_from_node(struct o2nm_node *node)
237 {
238         /* through the first node_set .parent
239          * mycluster/nodes/mynode == o2nm_cluster->o2nm_node_group->o2nm_node */
240         return to_o2nm_cluster(node->nd_item.ci_parent->ci_parent);
241 }
242
243 enum {
244         O2NM_NODE_ATTR_NUM = 0,
245         O2NM_NODE_ATTR_PORT,
246         O2NM_NODE_ATTR_ADDRESS,
247         O2NM_NODE_ATTR_LOCAL,
248 };
249
250 static ssize_t o2nm_node_num_write(struct o2nm_node *node, const char *page,
251                                    size_t count)
252 {
253         struct o2nm_cluster *cluster = to_o2nm_cluster_from_node(node);
254         unsigned long tmp;
255         char *p = (char *)page;
256
257         tmp = simple_strtoul(p, &p, 0);
258         if (!p || (*p && (*p != '\n')))
259                 return -EINVAL;
260
261         if (tmp >= O2NM_MAX_NODES)
262                 return -ERANGE;
263
264         /* once we're in the cl_nodes tree networking can look us up by
265          * node number and try to use our address and port attributes
266          * to connect to this node.. make sure that they've been set
267          * before writing the node attribute? */
268         if (!test_bit(O2NM_NODE_ATTR_ADDRESS, &node->nd_set_attributes) ||
269             !test_bit(O2NM_NODE_ATTR_PORT, &node->nd_set_attributes))
270                 return -EINVAL; /* XXX */
271
272         write_lock(&cluster->cl_nodes_lock);
273         if (cluster->cl_nodes[tmp])
274                 p = NULL;
275         else  {
276                 cluster->cl_nodes[tmp] = node;
277                 node->nd_num = tmp;
278                 set_bit(tmp, cluster->cl_nodes_bitmap);
279         }
280         write_unlock(&cluster->cl_nodes_lock);
281         if (p == NULL)
282                 return -EEXIST;
283
284         return count;
285 }
286 static ssize_t o2nm_node_ipv4_port_read(struct o2nm_node *node, char *page)
287 {
288         return sprintf(page, "%u\n", ntohs(node->nd_ipv4_port));
289 }
290
291 static ssize_t o2nm_node_ipv4_port_write(struct o2nm_node *node,
292                                          const char *page, size_t count)
293 {
294         unsigned long tmp;
295         char *p = (char *)page;
296
297         tmp = simple_strtoul(p, &p, 0);
298         if (!p || (*p && (*p != '\n')))
299                 return -EINVAL;
300
301         if (tmp == 0)
302                 return -EINVAL;
303         if (tmp >= (u16)-1)
304                 return -ERANGE;
305
306         node->nd_ipv4_port = htons(tmp);
307
308         return count;
309 }
310
311 static ssize_t o2nm_node_ipv4_address_read(struct o2nm_node *node, char *page)
312 {
313         return sprintf(page, "%u.%u.%u.%u\n", NIPQUAD(node->nd_ipv4_address));
314 }
315
316 static ssize_t o2nm_node_ipv4_address_write(struct o2nm_node *node,
317                                             const char *page,
318                                             size_t count)
319 {
320         struct o2nm_cluster *cluster = to_o2nm_cluster_from_node(node);
321         int ret, i;
322         struct rb_node **p, *parent;
323         unsigned int octets[4];
324         __be32 ipv4_addr = 0;
325
326         ret = sscanf(page, "%3u.%3u.%3u.%3u", &octets[3], &octets[2],
327                      &octets[1], &octets[0]);
328         if (ret != 4)
329                 return -EINVAL;
330
331         for (i = 0; i < ARRAY_SIZE(octets); i++) {
332                 if (octets[i] > 255)
333                         return -ERANGE;
334                 be32_add_cpu(&ipv4_addr, octets[i] << (i * 8));
335         }
336
337         ret = 0;
338         write_lock(&cluster->cl_nodes_lock);
339         if (o2nm_node_ip_tree_lookup(cluster, ipv4_addr, &p, &parent))
340                 ret = -EEXIST;
341         else {
342                 rb_link_node(&node->nd_ip_node, parent, p);
343                 rb_insert_color(&node->nd_ip_node, &cluster->cl_node_ip_tree);
344         }
345         write_unlock(&cluster->cl_nodes_lock);
346         if (ret)
347                 return ret;
348
349         memcpy(&node->nd_ipv4_address, &ipv4_addr, sizeof(ipv4_addr));
350
351         return count;
352 }
353
354 static ssize_t o2nm_node_local_read(struct o2nm_node *node, char *page)
355 {
356         return sprintf(page, "%d\n", node->nd_local);
357 }
358
359 static ssize_t o2nm_node_local_write(struct o2nm_node *node, const char *page,
360                                      size_t count)
361 {
362         struct o2nm_cluster *cluster = to_o2nm_cluster_from_node(node);
363         unsigned long tmp;
364         char *p = (char *)page;
365         ssize_t ret;
366
367         tmp = simple_strtoul(p, &p, 0);
368         if (!p || (*p && (*p != '\n')))
369                 return -EINVAL;
370
371         tmp = !!tmp; /* boolean of whether this node wants to be local */
372
373         /* setting local turns on networking rx for now so we require having
374          * set everything else first */
375         if (!test_bit(O2NM_NODE_ATTR_ADDRESS, &node->nd_set_attributes) ||
376             !test_bit(O2NM_NODE_ATTR_NUM, &node->nd_set_attributes) ||
377             !test_bit(O2NM_NODE_ATTR_PORT, &node->nd_set_attributes))
378                 return -EINVAL; /* XXX */
379
380         /* the only failure case is trying to set a new local node
381          * when a different one is already set */
382         if (tmp && tmp == cluster->cl_has_local &&
383             cluster->cl_local_node != node->nd_num)
384                 return -EBUSY;
385
386         /* bring up the rx thread if we're setting the new local node. */
387         if (tmp && !cluster->cl_has_local) {
388                 ret = o2net_start_listening(node);
389                 if (ret)
390                         return ret;
391         }
392
393         if (!tmp && cluster->cl_has_local &&
394             cluster->cl_local_node == node->nd_num) {
395                 o2net_stop_listening(node);
396                 cluster->cl_local_node = O2NM_INVALID_NODE_NUM;
397         }
398
399         node->nd_local = tmp;
400         if (node->nd_local) {
401                 cluster->cl_has_local = tmp;
402                 cluster->cl_local_node = node->nd_num;
403         }
404
405         return count;
406 }
407
408 struct o2nm_node_attribute {
409         struct configfs_attribute attr;
410         ssize_t (*show)(struct o2nm_node *, char *);
411         ssize_t (*store)(struct o2nm_node *, const char *, size_t);
412 };
413
414 static struct o2nm_node_attribute o2nm_node_attr_num = {
415         .attr   = { .ca_owner = THIS_MODULE,
416                     .ca_name = "num",
417                     .ca_mode = S_IRUGO | S_IWUSR },
418         .show   = o2nm_node_num_read,
419         .store  = o2nm_node_num_write,
420 };
421
422 static struct o2nm_node_attribute o2nm_node_attr_ipv4_port = {
423         .attr   = { .ca_owner = THIS_MODULE,
424                     .ca_name = "ipv4_port",
425                     .ca_mode = S_IRUGO | S_IWUSR },
426         .show   = o2nm_node_ipv4_port_read,
427         .store  = o2nm_node_ipv4_port_write,
428 };
429
430 static struct o2nm_node_attribute o2nm_node_attr_ipv4_address = {
431         .attr   = { .ca_owner = THIS_MODULE,
432                     .ca_name = "ipv4_address",
433                     .ca_mode = S_IRUGO | S_IWUSR },
434         .show   = o2nm_node_ipv4_address_read,
435         .store  = o2nm_node_ipv4_address_write,
436 };
437
438 static struct o2nm_node_attribute o2nm_node_attr_local = {
439         .attr   = { .ca_owner = THIS_MODULE,
440                     .ca_name = "local",
441                     .ca_mode = S_IRUGO | S_IWUSR },
442         .show   = o2nm_node_local_read,
443         .store  = o2nm_node_local_write,
444 };
445
446 static struct configfs_attribute *o2nm_node_attrs[] = {
447         [O2NM_NODE_ATTR_NUM] = &o2nm_node_attr_num.attr,
448         [O2NM_NODE_ATTR_PORT] = &o2nm_node_attr_ipv4_port.attr,
449         [O2NM_NODE_ATTR_ADDRESS] = &o2nm_node_attr_ipv4_address.attr,
450         [O2NM_NODE_ATTR_LOCAL] = &o2nm_node_attr_local.attr,
451         NULL,
452 };
453
454 static int o2nm_attr_index(struct configfs_attribute *attr)
455 {
456         int i;
457         for (i = 0; i < ARRAY_SIZE(o2nm_node_attrs); i++) {
458                 if (attr == o2nm_node_attrs[i])
459                         return i;
460         }
461         BUG();
462         return 0;
463 }
464
465 static ssize_t o2nm_node_show(struct config_item *item,
466                               struct configfs_attribute *attr,
467                               char *page)
468 {
469         struct o2nm_node *node = to_o2nm_node(item);
470         struct o2nm_node_attribute *o2nm_node_attr =
471                 container_of(attr, struct o2nm_node_attribute, attr);
472         ssize_t ret = 0;
473
474         if (o2nm_node_attr->show)
475                 ret = o2nm_node_attr->show(node, page);
476         return ret;
477 }
478
479 static ssize_t o2nm_node_store(struct config_item *item,
480                                struct configfs_attribute *attr,
481                                const char *page, size_t count)
482 {
483         struct o2nm_node *node = to_o2nm_node(item);
484         struct o2nm_node_attribute *o2nm_node_attr =
485                 container_of(attr, struct o2nm_node_attribute, attr);
486         ssize_t ret;
487         int attr_index = o2nm_attr_index(attr);
488
489         if (o2nm_node_attr->store == NULL) {
490                 ret = -EINVAL;
491                 goto out;
492         }
493
494         if (test_bit(attr_index, &node->nd_set_attributes))
495                 return -EBUSY;
496
497         ret = o2nm_node_attr->store(node, page, count);
498         if (ret < count)
499                 goto out;
500
501         set_bit(attr_index, &node->nd_set_attributes);
502 out:
503         return ret;
504 }
505
506 static struct configfs_item_operations o2nm_node_item_ops = {
507         .release                = o2nm_node_release,
508         .show_attribute         = o2nm_node_show,
509         .store_attribute        = o2nm_node_store,
510 };
511
512 static struct config_item_type o2nm_node_type = {
513         .ct_item_ops    = &o2nm_node_item_ops,
514         .ct_attrs       = o2nm_node_attrs,
515         .ct_owner       = THIS_MODULE,
516 };
517
518 /* node set */
519
520 struct o2nm_node_group {
521         struct config_group ns_group;
522         /* some stuff? */
523 };
524
525 #if 0
526 static struct o2nm_node_group *to_o2nm_node_group(struct config_group *group)
527 {
528         return group ?
529                 container_of(group, struct o2nm_node_group, ns_group)
530                 : NULL;
531 }
532 #endif
533
534 struct o2nm_cluster_attribute {
535         struct configfs_attribute attr;
536         ssize_t (*show)(struct o2nm_cluster *, char *);
537         ssize_t (*store)(struct o2nm_cluster *, const char *, size_t);
538 };
539
540 static ssize_t o2nm_cluster_attr_write(const char *page, ssize_t count,
541                                        unsigned int *val)
542 {
543         unsigned long tmp;
544         char *p = (char *)page;
545
546         tmp = simple_strtoul(p, &p, 0);
547         if (!p || (*p && (*p != '\n')))
548                 return -EINVAL;
549
550         if (tmp == 0)
551                 return -EINVAL;
552         if (tmp >= (u32)-1)
553                 return -ERANGE;
554
555         *val = tmp;
556
557         return count;
558 }
559
560 static ssize_t o2nm_cluster_attr_idle_timeout_ms_read(
561         struct o2nm_cluster *cluster, char *page)
562 {
563         return sprintf(page, "%u\n", cluster->cl_idle_timeout_ms);
564 }
565
566 static ssize_t o2nm_cluster_attr_idle_timeout_ms_write(
567         struct o2nm_cluster *cluster, const char *page, size_t count)
568 {
569         ssize_t ret;
570         unsigned int val;
571
572         ret =  o2nm_cluster_attr_write(page, count, &val);
573
574         if (ret > 0) {
575                 if (cluster->cl_idle_timeout_ms != val
576                         && o2net_num_connected_peers()) {
577                         mlog(ML_NOTICE,
578                              "o2net: cannot change idle timeout after "
579                              "the first peer has agreed to it."
580                              "  %d connected peers\n",
581                              o2net_num_connected_peers());
582                         ret = -EINVAL;
583                 } else if (val <= cluster->cl_keepalive_delay_ms) {
584                         mlog(ML_NOTICE, "o2net: idle timeout must be larger "
585                              "than keepalive delay\n");
586                         ret = -EINVAL;
587                 } else {
588                         cluster->cl_idle_timeout_ms = val;
589                 }
590         }
591
592         return ret;
593 }
594
595 static ssize_t o2nm_cluster_attr_keepalive_delay_ms_read(
596         struct o2nm_cluster *cluster, char *page)
597 {
598         return sprintf(page, "%u\n", cluster->cl_keepalive_delay_ms);
599 }
600
601 static ssize_t o2nm_cluster_attr_keepalive_delay_ms_write(
602         struct o2nm_cluster *cluster, const char *page, size_t count)
603 {
604         ssize_t ret;
605         unsigned int val;
606
607         ret =  o2nm_cluster_attr_write(page, count, &val);
608
609         if (ret > 0) {
610                 if (cluster->cl_keepalive_delay_ms != val
611                     && o2net_num_connected_peers()) {
612                         mlog(ML_NOTICE,
613                              "o2net: cannot change keepalive delay after"
614                              " the first peer has agreed to it."
615                              "  %d connected peers\n",
616                              o2net_num_connected_peers());
617                         ret = -EINVAL;
618                 } else if (val >= cluster->cl_idle_timeout_ms) {
619                         mlog(ML_NOTICE, "o2net: keepalive delay must be "
620                              "smaller than idle timeout\n");
621                         ret = -EINVAL;
622                 } else {
623                         cluster->cl_keepalive_delay_ms = val;
624                 }
625         }
626
627         return ret;
628 }
629
630 static ssize_t o2nm_cluster_attr_reconnect_delay_ms_read(
631         struct o2nm_cluster *cluster, char *page)
632 {
633         return sprintf(page, "%u\n", cluster->cl_reconnect_delay_ms);
634 }
635
636 static ssize_t o2nm_cluster_attr_reconnect_delay_ms_write(
637         struct o2nm_cluster *cluster, const char *page, size_t count)
638 {
639         return o2nm_cluster_attr_write(page, count,
640                                        &cluster->cl_reconnect_delay_ms);
641 }
642 static struct o2nm_cluster_attribute o2nm_cluster_attr_idle_timeout_ms = {
643         .attr   = { .ca_owner = THIS_MODULE,
644                     .ca_name = "idle_timeout_ms",
645                     .ca_mode = S_IRUGO | S_IWUSR },
646         .show   = o2nm_cluster_attr_idle_timeout_ms_read,
647         .store  = o2nm_cluster_attr_idle_timeout_ms_write,
648 };
649
650 static struct o2nm_cluster_attribute o2nm_cluster_attr_keepalive_delay_ms = {
651         .attr   = { .ca_owner = THIS_MODULE,
652                     .ca_name = "keepalive_delay_ms",
653                     .ca_mode = S_IRUGO | S_IWUSR },
654         .show   = o2nm_cluster_attr_keepalive_delay_ms_read,
655         .store  = o2nm_cluster_attr_keepalive_delay_ms_write,
656 };
657
658 static struct o2nm_cluster_attribute o2nm_cluster_attr_reconnect_delay_ms = {
659         .attr   = { .ca_owner = THIS_MODULE,
660                     .ca_name = "reconnect_delay_ms",
661                     .ca_mode = S_IRUGO | S_IWUSR },
662         .show   = o2nm_cluster_attr_reconnect_delay_ms_read,
663         .store  = o2nm_cluster_attr_reconnect_delay_ms_write,
664 };
665
666 static struct configfs_attribute *o2nm_cluster_attrs[] = {
667         &o2nm_cluster_attr_idle_timeout_ms.attr,
668         &o2nm_cluster_attr_keepalive_delay_ms.attr,
669         &o2nm_cluster_attr_reconnect_delay_ms.attr,
670         NULL,
671 };
672 static ssize_t o2nm_cluster_show(struct config_item *item,
673                                  struct configfs_attribute *attr,
674                                  char *page)
675 {
676         struct o2nm_cluster *cluster = to_o2nm_cluster(item);
677         struct o2nm_cluster_attribute *o2nm_cluster_attr =
678                 container_of(attr, struct o2nm_cluster_attribute, attr);
679         ssize_t ret = 0;
680
681         if (o2nm_cluster_attr->show)
682                 ret = o2nm_cluster_attr->show(cluster, page);
683         return ret;
684 }
685
686 static ssize_t o2nm_cluster_store(struct config_item *item,
687                                   struct configfs_attribute *attr,
688                                   const char *page, size_t count)
689 {
690         struct o2nm_cluster *cluster = to_o2nm_cluster(item);
691         struct o2nm_cluster_attribute *o2nm_cluster_attr =
692                 container_of(attr, struct o2nm_cluster_attribute, attr);
693         ssize_t ret;
694
695         if (o2nm_cluster_attr->store == NULL) {
696                 ret = -EINVAL;
697                 goto out;
698         }
699
700         ret = o2nm_cluster_attr->store(cluster, page, count);
701         if (ret < count)
702                 goto out;
703 out:
704         return ret;
705 }
706
707 static struct config_item *o2nm_node_group_make_item(struct config_group *group,
708                                                      const char *name)
709 {
710         struct o2nm_node *node = NULL;
711         struct config_item *ret = NULL;
712
713         if (strlen(name) > O2NM_MAX_NAME_LEN)
714                 goto out; /* ENAMETOOLONG */
715
716         node = kzalloc(sizeof(struct o2nm_node), GFP_KERNEL);
717         if (node == NULL)
718                 goto out; /* ENOMEM */
719
720         strcpy(node->nd_name, name); /* use item.ci_namebuf instead? */
721         config_item_init_type_name(&node->nd_item, name, &o2nm_node_type);
722         spin_lock_init(&node->nd_lock);
723
724         ret = &node->nd_item;
725
726 out:
727         if (ret == NULL)
728                 kfree(node);
729
730         return ret;
731 }
732
733 static void o2nm_node_group_drop_item(struct config_group *group,
734                                       struct config_item *item)
735 {
736         struct o2nm_node *node = to_o2nm_node(item);
737         struct o2nm_cluster *cluster = to_o2nm_cluster(group->cg_item.ci_parent);
738
739         o2net_disconnect_node(node);
740
741         if (cluster->cl_has_local &&
742             (cluster->cl_local_node == node->nd_num)) {
743                 cluster->cl_has_local = 0;
744                 cluster->cl_local_node = O2NM_INVALID_NODE_NUM;
745                 o2net_stop_listening(node);
746         }
747
748         /* XXX call into net to stop this node from trading messages */
749
750         write_lock(&cluster->cl_nodes_lock);
751
752         /* XXX sloppy */
753         if (node->nd_ipv4_address)
754                 rb_erase(&node->nd_ip_node, &cluster->cl_node_ip_tree);
755
756         /* nd_num might be 0 if the node number hasn't been set.. */
757         if (cluster->cl_nodes[node->nd_num] == node) {
758                 cluster->cl_nodes[node->nd_num] = NULL;
759                 clear_bit(node->nd_num, cluster->cl_nodes_bitmap);
760         }
761         write_unlock(&cluster->cl_nodes_lock);
762
763         config_item_put(item);
764 }
765
766 static struct configfs_group_operations o2nm_node_group_group_ops = {
767         .make_item      = o2nm_node_group_make_item,
768         .drop_item      = o2nm_node_group_drop_item,
769 };
770
771 static struct config_item_type o2nm_node_group_type = {
772         .ct_group_ops   = &o2nm_node_group_group_ops,
773         .ct_owner       = THIS_MODULE,
774 };
775
776 /* cluster */
777
778 static void o2nm_cluster_release(struct config_item *item)
779 {
780         struct o2nm_cluster *cluster = to_o2nm_cluster(item);
781
782         kfree(cluster->cl_group.default_groups);
783         kfree(cluster);
784 }
785
786 static struct configfs_item_operations o2nm_cluster_item_ops = {
787         .release        = o2nm_cluster_release,
788         .show_attribute         = o2nm_cluster_show,
789         .store_attribute        = o2nm_cluster_store,
790 };
791
792 static struct config_item_type o2nm_cluster_type = {
793         .ct_item_ops    = &o2nm_cluster_item_ops,
794         .ct_attrs       = o2nm_cluster_attrs,
795         .ct_owner       = THIS_MODULE,
796 };
797
798 /* cluster set */
799
800 struct o2nm_cluster_group {
801         struct configfs_subsystem cs_subsys;
802         /* some stuff? */
803 };
804
805 #if 0
806 static struct o2nm_cluster_group *to_o2nm_cluster_group(struct config_group *group)
807 {
808         return group ?
809                 container_of(to_configfs_subsystem(group), struct o2nm_cluster_group, cs_subsys)
810                : NULL;
811 }
812 #endif
813
814 static struct config_group *o2nm_cluster_group_make_group(struct config_group *group,
815                                                           const char *name)
816 {
817         struct o2nm_cluster *cluster = NULL;
818         struct o2nm_node_group *ns = NULL;
819         struct config_group *o2hb_group = NULL, *ret = NULL;
820         void *defs = NULL;
821
822         /* this runs under the parent dir's i_mutex; there can be only
823          * one caller in here at a time */
824         if (o2nm_single_cluster)
825                 goto out; /* ENOSPC */
826
827         cluster = kzalloc(sizeof(struct o2nm_cluster), GFP_KERNEL);
828         ns = kzalloc(sizeof(struct o2nm_node_group), GFP_KERNEL);
829         defs = kcalloc(3, sizeof(struct config_group *), GFP_KERNEL);
830         o2hb_group = o2hb_alloc_hb_set();
831         if (cluster == NULL || ns == NULL || o2hb_group == NULL || defs == NULL)
832                 goto out;
833
834         config_group_init_type_name(&cluster->cl_group, name,
835                                     &o2nm_cluster_type);
836         config_group_init_type_name(&ns->ns_group, "node",
837                                     &o2nm_node_group_type);
838
839         cluster->cl_group.default_groups = defs;
840         cluster->cl_group.default_groups[0] = &ns->ns_group;
841         cluster->cl_group.default_groups[1] = o2hb_group;
842         cluster->cl_group.default_groups[2] = NULL;
843         rwlock_init(&cluster->cl_nodes_lock);
844         cluster->cl_node_ip_tree = RB_ROOT;
845         cluster->cl_reconnect_delay_ms = O2NET_RECONNECT_DELAY_MS_DEFAULT;
846         cluster->cl_idle_timeout_ms    = O2NET_IDLE_TIMEOUT_MS_DEFAULT;
847         cluster->cl_keepalive_delay_ms = O2NET_KEEPALIVE_DELAY_MS_DEFAULT;
848
849         ret = &cluster->cl_group;
850         o2nm_single_cluster = cluster;
851
852 out:
853         if (ret == NULL) {
854                 kfree(cluster);
855                 kfree(ns);
856                 o2hb_free_hb_set(o2hb_group);
857                 kfree(defs);
858         }
859
860         return ret;
861 }
862
863 static void o2nm_cluster_group_drop_item(struct config_group *group, struct config_item *item)
864 {
865         struct o2nm_cluster *cluster = to_o2nm_cluster(item);
866         int i;
867         struct config_item *killme;
868
869         BUG_ON(o2nm_single_cluster != cluster);
870         o2nm_single_cluster = NULL;
871
872         for (i = 0; cluster->cl_group.default_groups[i]; i++) {
873                 killme = &cluster->cl_group.default_groups[i]->cg_item;
874                 cluster->cl_group.default_groups[i] = NULL;
875                 config_item_put(killme);
876         }
877
878         config_item_put(item);
879 }
880
881 static struct configfs_group_operations o2nm_cluster_group_group_ops = {
882         .make_group     = o2nm_cluster_group_make_group,
883         .drop_item      = o2nm_cluster_group_drop_item,
884 };
885
886 static struct config_item_type o2nm_cluster_group_type = {
887         .ct_group_ops   = &o2nm_cluster_group_group_ops,
888         .ct_owner       = THIS_MODULE,
889 };
890
891 static struct o2nm_cluster_group o2nm_cluster_group = {
892         .cs_subsys = {
893                 .su_group = {
894                         .cg_item = {
895                                 .ci_namebuf = "cluster",
896                                 .ci_type = &o2nm_cluster_group_type,
897                         },
898                 },
899         },
900 };
901
902 int o2nm_depend_item(struct config_item *item)
903 {
904         return configfs_depend_item(&o2nm_cluster_group.cs_subsys, item);
905 }
906
907 void o2nm_undepend_item(struct config_item *item)
908 {
909         configfs_undepend_item(&o2nm_cluster_group.cs_subsys, item);
910 }
911
912 int o2nm_depend_this_node(void)
913 {
914         int ret = 0;
915         struct o2nm_node *local_node;
916
917         local_node = o2nm_get_node_by_num(o2nm_this_node());
918         if (!local_node) {
919                 ret = -EINVAL;
920                 goto out;
921         }
922
923         ret = o2nm_depend_item(&local_node->nd_item);
924         o2nm_node_put(local_node);
925
926 out:
927         return ret;
928 }
929
930 void o2nm_undepend_this_node(void)
931 {
932         struct o2nm_node *local_node;
933
934         local_node = o2nm_get_node_by_num(o2nm_this_node());
935         BUG_ON(!local_node);
936
937         o2nm_undepend_item(&local_node->nd_item);
938         o2nm_node_put(local_node);
939 }
940
941
942 static void __exit exit_o2nm(void)
943 {
944         if (ocfs2_table_header)
945                 unregister_sysctl_table(ocfs2_table_header);
946
947         /* XXX sync with hb callbacks and shut down hb? */
948         o2net_unregister_hb_callbacks();
949         configfs_unregister_subsystem(&o2nm_cluster_group.cs_subsys);
950         o2cb_sys_shutdown();
951
952         o2net_exit();
953 }
954
955 static int __init init_o2nm(void)
956 {
957         int ret = -1;
958
959         cluster_print_version();
960
961         o2hb_init();
962
963         ret = o2net_init();
964         if (ret)
965                 goto out;
966
967         ocfs2_table_header = register_sysctl_table(ocfs2_root_table);
968         if (!ocfs2_table_header) {
969                 printk(KERN_ERR "nodemanager: unable to register sysctl\n");
970                 ret = -ENOMEM; /* or something. */
971                 goto out_o2net;
972         }
973
974         ret = o2net_register_hb_callbacks();
975         if (ret)
976                 goto out_sysctl;
977
978         config_group_init(&o2nm_cluster_group.cs_subsys.su_group);
979         mutex_init(&o2nm_cluster_group.cs_subsys.su_mutex);
980         ret = configfs_register_subsystem(&o2nm_cluster_group.cs_subsys);
981         if (ret) {
982                 printk(KERN_ERR "nodemanager: Registration returned %d\n", ret);
983                 goto out_callbacks;
984         }
985
986         ret = o2cb_sys_init();
987         if (!ret)
988                 goto out;
989
990         configfs_unregister_subsystem(&o2nm_cluster_group.cs_subsys);
991 out_callbacks:
992         o2net_unregister_hb_callbacks();
993 out_sysctl:
994         unregister_sysctl_table(ocfs2_table_header);
995 out_o2net:
996         o2net_exit();
997 out:
998         return ret;
999 }
1000
1001 MODULE_AUTHOR("Oracle");
1002 MODULE_LICENSE("GPL");
1003
1004 module_init(init_o2nm)
1005 module_exit(exit_o2nm)