blob: 2768eb1da2b865f4afba3c35087da7c477c87309 [file] [log] [blame]
Joel Becker8adf0532007-11-28 14:38:40 -08001/* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
3 *
4 * stack_user.c
5 *
6 * Code which interfaces ocfs2 with fs/dlm and a userspace stack.
7 *
8 * Copyright (C) 2007 Oracle. All rights reserved.
9 *
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public
12 * License as published by the Free Software Foundation, version 2.
13 *
14 * This program is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * General Public License for more details.
18 */
19
20#include <linux/module.h>
Joel Becker6427a722008-02-18 19:23:28 -080021#include <linux/fs.h>
22#include <linux/miscdevice.h>
23#include <linux/mutex.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090024#include <linux/slab.h>
Joel Becker6427a722008-02-18 19:23:28 -080025#include <linux/reboot.h>
Goldwyn Rodriguesc994c2e2014-01-21 15:48:32 -080026#include <linux/sched.h>
Joel Becker462c7e62008-02-18 19:40:12 -080027#include <asm/uaccess.h>
Joel Becker8adf0532007-11-28 14:38:40 -080028
29#include "stackglue.h"
30
Mark Fasheh53da4932008-07-21 14:29:16 -070031#include <linux/dlm_plock.h>
Joel Becker8adf0532007-11-28 14:38:40 -080032
Joel Becker6427a722008-02-18 19:23:28 -080033/*
34 * The control protocol starts with a handshake. Until the handshake
35 * is complete, the control device will fail all write(2)s.
36 *
37 * The handshake is simple. First, the client reads until EOF. Each line
38 * of output is a supported protocol tag. All protocol tags are a single
39 * character followed by a two hex digit version number. Currently the
40 * only things supported is T01, for "Text-base version 0x01". Next, the
Joel Beckerde870ef2008-02-18 17:07:09 -080041 * client writes the version they would like to use, including the newline.
42 * Thus, the protocol tag is 'T01\n'. If the version tag written is
43 * unknown, -EINVAL is returned. Once the negotiation is complete, the
44 * client can start sending messages.
45 *
Joel Beckerd4b95ee2008-02-20 15:39:44 -080046 * The T01 protocol has three messages. First is the "SETN" message.
Joel Becker3cfd4ab2008-02-20 14:44:34 -080047 * It has the following syntax:
48 *
49 * SETN<space><8-char-hex-nodenum><newline>
50 *
51 * This is 14 characters.
52 *
53 * The "SETN" message must be the first message following the protocol.
54 * It tells ocfs2_control the local node number.
55 *
Joel Beckerd4b95ee2008-02-20 15:39:44 -080056 * Next comes the "SETV" message. It has the following syntax:
57 *
58 * SETV<space><2-char-hex-major><space><2-char-hex-minor><newline>
59 *
60 * This is 11 characters.
61 *
62 * The "SETV" message sets the filesystem locking protocol version as
63 * negotiated by the client. The client negotiates based on the maximum
64 * version advertised in /sys/fs/ocfs2/max_locking_protocol. The major
65 * number from the "SETV" message must match
Joel Beckere603cfb2010-01-29 16:06:29 -080066 * ocfs2_user_plugin.sp_max_proto.pv_major, and the minor number
67 * must be less than or equal to ...sp_max_version.pv_minor.
Joel Beckerd4b95ee2008-02-20 15:39:44 -080068 *
69 * Once this information has been set, mounts will be allowed. From this
70 * point on, the "DOWN" message can be sent for node down notification.
71 * It has the following syntax:
Joel Beckerde870ef2008-02-18 17:07:09 -080072 *
73 * DOWN<space><32-char-cap-hex-uuid><space><8-char-hex-nodenum><newline>
74 *
75 * eg:
76 *
77 * DOWN 632A924FDD844190BDA93C0DF6B94899 00000001\n
78 *
79 * This is 47 characters.
Joel Becker6427a722008-02-18 19:23:28 -080080 */
81
82/*
Joel Becker462c7e62008-02-18 19:40:12 -080083 * Whether or not the client has done the handshake.
84 * For now, we have just one protocol version.
85 */
86#define OCFS2_CONTROL_PROTO "T01\n"
87#define OCFS2_CONTROL_PROTO_LEN 4
Joel Becker3cfd4ab2008-02-20 14:44:34 -080088
89/* Handshake states */
Joel Becker462c7e62008-02-18 19:40:12 -080090#define OCFS2_CONTROL_HANDSHAKE_INVALID (0)
91#define OCFS2_CONTROL_HANDSHAKE_READ (1)
Joel Becker3cfd4ab2008-02-20 14:44:34 -080092#define OCFS2_CONTROL_HANDSHAKE_PROTOCOL (2)
93#define OCFS2_CONTROL_HANDSHAKE_VALID (3)
94
95/* Messages */
96#define OCFS2_CONTROL_MESSAGE_OP_LEN 4
97#define OCFS2_CONTROL_MESSAGE_SETNODE_OP "SETN"
98#define OCFS2_CONTROL_MESSAGE_SETNODE_TOTAL_LEN 14
Joel Beckerd4b95ee2008-02-20 15:39:44 -080099#define OCFS2_CONTROL_MESSAGE_SETVERSION_OP "SETV"
100#define OCFS2_CONTROL_MESSAGE_SETVERSION_TOTAL_LEN 11
Joel Becker3cfd4ab2008-02-20 14:44:34 -0800101#define OCFS2_CONTROL_MESSAGE_DOWN_OP "DOWN"
Joel Beckerde870ef2008-02-18 17:07:09 -0800102#define OCFS2_CONTROL_MESSAGE_DOWN_TOTAL_LEN 47
103#define OCFS2_TEXT_UUID_LEN 32
Joel Beckerd4b95ee2008-02-20 15:39:44 -0800104#define OCFS2_CONTROL_MESSAGE_VERNUM_LEN 2
Joel Beckerde870ef2008-02-18 17:07:09 -0800105#define OCFS2_CONTROL_MESSAGE_NODENUM_LEN 8
Goldwyn Rodrigues41503632014-01-21 15:48:25 -0800106#define VERSION_LOCK "version_lock"
Joel Becker462c7e62008-02-18 19:40:12 -0800107
Goldwyn Rodrigues3e834152014-01-21 15:48:24 -0800108enum ocfs2_connection_type {
109 WITH_CONTROLD,
110 NO_CONTROLD
111};
112
Joel Becker462c7e62008-02-18 19:40:12 -0800113/*
Joel Becker6427a722008-02-18 19:23:28 -0800114 * ocfs2_live_connection is refcounted because the filesystem and
115 * miscdevice sides can detach in different order. Let's just be safe.
116 */
117struct ocfs2_live_connection {
118 struct list_head oc_list;
119 struct ocfs2_cluster_connection *oc_conn;
Goldwyn Rodrigues3e834152014-01-21 15:48:24 -0800120 enum ocfs2_connection_type oc_type;
Goldwyn Rodrigues66e188f2014-01-21 15:48:22 -0800121 atomic_t oc_this_node;
122 int oc_our_slot;
Goldwyn Rodrigues41503632014-01-21 15:48:25 -0800123 struct dlm_lksb oc_version_lksb;
124 char oc_lvb[DLM_LVB_LEN];
125 struct completion oc_sync_wait;
Goldwyn Rodriguesc994c2e2014-01-21 15:48:32 -0800126 wait_queue_head_t oc_wait;
Joel Becker6427a722008-02-18 19:23:28 -0800127};
128
Joel Becker462c7e62008-02-18 19:40:12 -0800129struct ocfs2_control_private {
130 struct list_head op_list;
131 int op_state;
Joel Becker3cfd4ab2008-02-20 14:44:34 -0800132 int op_this_node;
Joel Beckerd4b95ee2008-02-20 15:39:44 -0800133 struct ocfs2_protocol_version op_proto;
Joel Becker3cfd4ab2008-02-20 14:44:34 -0800134};
135
136/* SETN<space><8-char-hex-nodenum><newline> */
137struct ocfs2_control_message_setn {
138 char tag[OCFS2_CONTROL_MESSAGE_OP_LEN];
139 char space;
140 char nodestr[OCFS2_CONTROL_MESSAGE_NODENUM_LEN];
141 char newline;
142};
143
Joel Beckerd4b95ee2008-02-20 15:39:44 -0800144/* SETV<space><2-char-hex-major><space><2-char-hex-minor><newline> */
145struct ocfs2_control_message_setv {
146 char tag[OCFS2_CONTROL_MESSAGE_OP_LEN];
147 char space1;
148 char major[OCFS2_CONTROL_MESSAGE_VERNUM_LEN];
149 char space2;
150 char minor[OCFS2_CONTROL_MESSAGE_VERNUM_LEN];
151 char newline;
152};
153
Joel Becker3cfd4ab2008-02-20 14:44:34 -0800154/* DOWN<space><32-char-cap-hex-uuid><space><8-char-hex-nodenum><newline> */
155struct ocfs2_control_message_down {
156 char tag[OCFS2_CONTROL_MESSAGE_OP_LEN];
157 char space1;
158 char uuid[OCFS2_TEXT_UUID_LEN];
159 char space2;
160 char nodestr[OCFS2_CONTROL_MESSAGE_NODENUM_LEN];
161 char newline;
162};
163
164union ocfs2_control_message {
165 char tag[OCFS2_CONTROL_MESSAGE_OP_LEN];
166 struct ocfs2_control_message_setn u_setn;
Joel Beckerd4b95ee2008-02-20 15:39:44 -0800167 struct ocfs2_control_message_setv u_setv;
Joel Becker3cfd4ab2008-02-20 14:44:34 -0800168 struct ocfs2_control_message_down u_down;
Joel Becker462c7e62008-02-18 19:40:12 -0800169};
170
Joel Beckera12630b2008-05-09 18:49:29 -0700171static struct ocfs2_stack_plugin ocfs2_user_plugin;
David Teiglandcf4d8d72008-02-20 14:29:27 -0800172
Joel Becker6427a722008-02-18 19:23:28 -0800173static atomic_t ocfs2_control_opened;
Joel Becker3cfd4ab2008-02-20 14:44:34 -0800174static int ocfs2_control_this_node = -1;
Joel Beckerd4b95ee2008-02-20 15:39:44 -0800175static struct ocfs2_protocol_version running_proto;
Joel Becker6427a722008-02-18 19:23:28 -0800176
177static LIST_HEAD(ocfs2_live_connection_list);
Joel Becker462c7e62008-02-18 19:40:12 -0800178static LIST_HEAD(ocfs2_control_private_list);
Joel Becker6427a722008-02-18 19:23:28 -0800179static DEFINE_MUTEX(ocfs2_control_lock);
180
Joel Becker462c7e62008-02-18 19:40:12 -0800181static inline void ocfs2_control_set_handshake_state(struct file *file,
182 int state)
183{
184 struct ocfs2_control_private *p = file->private_data;
185 p->op_state = state;
186}
187
188static inline int ocfs2_control_get_handshake_state(struct file *file)
189{
190 struct ocfs2_control_private *p = file->private_data;
191 return p->op_state;
192}
193
Joel Becker6427a722008-02-18 19:23:28 -0800194static struct ocfs2_live_connection *ocfs2_connection_find(const char *name)
195{
196 size_t len = strlen(name);
197 struct ocfs2_live_connection *c;
198
199 BUG_ON(!mutex_is_locked(&ocfs2_control_lock));
200
201 list_for_each_entry(c, &ocfs2_live_connection_list, oc_list) {
202 if ((c->oc_conn->cc_namelen == len) &&
203 !strncmp(c->oc_conn->cc_name, name, len))
204 return c;
205 }
206
dann frazier226291a2010-11-18 15:03:09 -0700207 return NULL;
Joel Becker6427a722008-02-18 19:23:28 -0800208}
209
210/*
211 * ocfs2_live_connection structures are created underneath the ocfs2
212 * mount path. Since the VFS prevents multiple calls to
213 * fill_super(), we can't get dupes here.
214 */
Goldwyn Rodrigues24aa3382014-01-21 15:48:23 -0800215static int ocfs2_live_connection_attach(struct ocfs2_cluster_connection *conn,
216 struct ocfs2_live_connection *c)
Joel Becker6427a722008-02-18 19:23:28 -0800217{
218 int rc = 0;
Joel Becker6427a722008-02-18 19:23:28 -0800219
220 mutex_lock(&ocfs2_control_lock);
221 c->oc_conn = conn;
222
Goldwyn Rodriguesc994c2e2014-01-21 15:48:32 -0800223 if ((c->oc_type == NO_CONTROLD) || atomic_read(&ocfs2_control_opened))
Joel Becker6427a722008-02-18 19:23:28 -0800224 list_add(&c->oc_list, &ocfs2_live_connection_list);
225 else {
226 printk(KERN_ERR
227 "ocfs2: Userspace control daemon is not present\n");
228 rc = -ESRCH;
229 }
230
231 mutex_unlock(&ocfs2_control_lock);
Joel Becker6427a722008-02-18 19:23:28 -0800232 return rc;
233}
234
235/*
236 * This function disconnects the cluster connection from ocfs2_control.
237 * Afterwards, userspace can't affect the cluster connection.
238 */
239static void ocfs2_live_connection_drop(struct ocfs2_live_connection *c)
240{
241 mutex_lock(&ocfs2_control_lock);
242 list_del_init(&c->oc_list);
243 c->oc_conn = NULL;
244 mutex_unlock(&ocfs2_control_lock);
245
246 kfree(c);
247}
248
Joel Becker3cfd4ab2008-02-20 14:44:34 -0800249static int ocfs2_control_cfu(void *target, size_t target_len,
250 const char __user *buf, size_t count)
Joel Becker462c7e62008-02-18 19:40:12 -0800251{
252 /* The T01 expects write(2) calls to have exactly one command */
Joel Becker3cfd4ab2008-02-20 14:44:34 -0800253 if ((count != target_len) ||
254 (count > sizeof(union ocfs2_control_message)))
Joel Becker462c7e62008-02-18 19:40:12 -0800255 return -EINVAL;
256
257 if (copy_from_user(target, buf, target_len))
258 return -EFAULT;
259
Joel Becker3cfd4ab2008-02-20 14:44:34 -0800260 return 0;
Joel Becker462c7e62008-02-18 19:40:12 -0800261}
262
Joel Becker3cfd4ab2008-02-20 14:44:34 -0800263static ssize_t ocfs2_control_validate_protocol(struct file *file,
264 const char __user *buf,
265 size_t count)
Joel Becker462c7e62008-02-18 19:40:12 -0800266{
267 ssize_t ret;
268 char kbuf[OCFS2_CONTROL_PROTO_LEN];
269
270 ret = ocfs2_control_cfu(kbuf, OCFS2_CONTROL_PROTO_LEN,
271 buf, count);
Joel Becker3cfd4ab2008-02-20 14:44:34 -0800272 if (ret)
Joel Becker462c7e62008-02-18 19:40:12 -0800273 return ret;
274
275 if (strncmp(kbuf, OCFS2_CONTROL_PROTO, OCFS2_CONTROL_PROTO_LEN))
276 return -EINVAL;
277
Joel Becker462c7e62008-02-18 19:40:12 -0800278 ocfs2_control_set_handshake_state(file,
Joel Becker3cfd4ab2008-02-20 14:44:34 -0800279 OCFS2_CONTROL_HANDSHAKE_PROTOCOL);
Joel Becker462c7e62008-02-18 19:40:12 -0800280
281 return count;
282}
283
Joel Beckerde870ef2008-02-18 17:07:09 -0800284static void ocfs2_control_send_down(const char *uuid,
285 int nodenum)
286{
287 struct ocfs2_live_connection *c;
288
289 mutex_lock(&ocfs2_control_lock);
290
291 c = ocfs2_connection_find(uuid);
292 if (c) {
293 BUG_ON(c->oc_conn == NULL);
294 c->oc_conn->cc_recovery_handler(nodenum,
295 c->oc_conn->cc_recovery_data);
296 }
297
298 mutex_unlock(&ocfs2_control_lock);
299}
300
Joel Becker3cfd4ab2008-02-20 14:44:34 -0800301/*
302 * Called whenever configuration elements are sent to /dev/ocfs2_control.
303 * If all configuration elements are present, try to set the global
Joel Beckerd4b95ee2008-02-20 15:39:44 -0800304 * values. If there is a problem, return an error. Skip any missing
305 * elements, and only bump ocfs2_control_opened when we have all elements
306 * and are successful.
Joel Becker3cfd4ab2008-02-20 14:44:34 -0800307 */
308static int ocfs2_control_install_private(struct file *file)
Joel Beckerde870ef2008-02-18 17:07:09 -0800309{
Joel Becker3cfd4ab2008-02-20 14:44:34 -0800310 int rc = 0;
311 int set_p = 1;
312 struct ocfs2_control_private *p = file->private_data;
313
314 BUG_ON(p->op_state != OCFS2_CONTROL_HANDSHAKE_PROTOCOL);
315
Joel Becker3cfd4ab2008-02-20 14:44:34 -0800316 mutex_lock(&ocfs2_control_lock);
Joel Beckerd4b95ee2008-02-20 15:39:44 -0800317
318 if (p->op_this_node < 0) {
319 set_p = 0;
320 } else if ((ocfs2_control_this_node >= 0) &&
321 (ocfs2_control_this_node != p->op_this_node)) {
Joel Becker3cfd4ab2008-02-20 14:44:34 -0800322 rc = -EINVAL;
Joel Beckerd4b95ee2008-02-20 15:39:44 -0800323 goto out_unlock;
324 }
325
326 if (!p->op_proto.pv_major) {
327 set_p = 0;
328 } else if (!list_empty(&ocfs2_live_connection_list) &&
329 ((running_proto.pv_major != p->op_proto.pv_major) ||
330 (running_proto.pv_minor != p->op_proto.pv_minor))) {
331 rc = -EINVAL;
332 goto out_unlock;
333 }
334
335 if (set_p) {
336 ocfs2_control_this_node = p->op_this_node;
337 running_proto.pv_major = p->op_proto.pv_major;
338 running_proto.pv_minor = p->op_proto.pv_minor;
339 }
340
341out_unlock:
Joel Becker3cfd4ab2008-02-20 14:44:34 -0800342 mutex_unlock(&ocfs2_control_lock);
343
344 if (!rc && set_p) {
345 /* We set the global values successfully */
346 atomic_inc(&ocfs2_control_opened);
347 ocfs2_control_set_handshake_state(file,
348 OCFS2_CONTROL_HANDSHAKE_VALID);
349 }
350
351 return rc;
352}
353
David Teiglandcf4d8d72008-02-20 14:29:27 -0800354static int ocfs2_control_get_this_node(void)
355{
356 int rc;
357
358 mutex_lock(&ocfs2_control_lock);
359 if (ocfs2_control_this_node < 0)
360 rc = -EINVAL;
361 else
362 rc = ocfs2_control_this_node;
363 mutex_unlock(&ocfs2_control_lock);
364
365 return rc;
366}
367
Joel Becker3cfd4ab2008-02-20 14:44:34 -0800368static int ocfs2_control_do_setnode_msg(struct file *file,
369 struct ocfs2_control_message_setn *msg)
370{
Joel Beckerde870ef2008-02-18 17:07:09 -0800371 long nodenum;
Joel Becker3cfd4ab2008-02-20 14:44:34 -0800372 char *ptr = NULL;
373 struct ocfs2_control_private *p = file->private_data;
Joel Beckerde870ef2008-02-18 17:07:09 -0800374
Joel Becker3cfd4ab2008-02-20 14:44:34 -0800375 if (ocfs2_control_get_handshake_state(file) !=
376 OCFS2_CONTROL_HANDSHAKE_PROTOCOL)
Joel Beckerde870ef2008-02-18 17:07:09 -0800377 return -EINVAL;
378
Joel Becker3cfd4ab2008-02-20 14:44:34 -0800379 if (strncmp(msg->tag, OCFS2_CONTROL_MESSAGE_SETNODE_OP,
380 OCFS2_CONTROL_MESSAGE_OP_LEN))
Joel Beckerde870ef2008-02-18 17:07:09 -0800381 return -EINVAL;
Joel Beckerde870ef2008-02-18 17:07:09 -0800382
Joel Becker3cfd4ab2008-02-20 14:44:34 -0800383 if ((msg->space != ' ') || (msg->newline != '\n'))
384 return -EINVAL;
385 msg->space = msg->newline = '\0';
386
387 nodenum = simple_strtol(msg->nodestr, &ptr, 16);
388 if (!ptr || *ptr)
389 return -EINVAL;
390
391 if ((nodenum == LONG_MIN) || (nodenum == LONG_MAX) ||
392 (nodenum > INT_MAX) || (nodenum < 0))
393 return -ERANGE;
394 p->op_this_node = nodenum;
395
396 return ocfs2_control_install_private(file);
397}
398
Joel Beckerd4b95ee2008-02-20 15:39:44 -0800399static int ocfs2_control_do_setversion_msg(struct file *file,
400 struct ocfs2_control_message_setv *msg)
401 {
402 long major, minor;
403 char *ptr = NULL;
404 struct ocfs2_control_private *p = file->private_data;
405 struct ocfs2_protocol_version *max =
Joel Beckere603cfb2010-01-29 16:06:29 -0800406 &ocfs2_user_plugin.sp_max_proto;
Joel Beckerd4b95ee2008-02-20 15:39:44 -0800407
408 if (ocfs2_control_get_handshake_state(file) !=
409 OCFS2_CONTROL_HANDSHAKE_PROTOCOL)
410 return -EINVAL;
411
412 if (strncmp(msg->tag, OCFS2_CONTROL_MESSAGE_SETVERSION_OP,
413 OCFS2_CONTROL_MESSAGE_OP_LEN))
414 return -EINVAL;
415
416 if ((msg->space1 != ' ') || (msg->space2 != ' ') ||
417 (msg->newline != '\n'))
418 return -EINVAL;
419 msg->space1 = msg->space2 = msg->newline = '\0';
420
421 major = simple_strtol(msg->major, &ptr, 16);
422 if (!ptr || *ptr)
423 return -EINVAL;
424 minor = simple_strtol(msg->minor, &ptr, 16);
425 if (!ptr || *ptr)
426 return -EINVAL;
427
428 /*
429 * The major must be between 1 and 255, inclusive. The minor
430 * must be between 0 and 255, inclusive. The version passed in
431 * must be within the maximum version supported by the filesystem.
432 */
433 if ((major == LONG_MIN) || (major == LONG_MAX) ||
434 (major > (u8)-1) || (major < 1))
435 return -ERANGE;
436 if ((minor == LONG_MIN) || (minor == LONG_MAX) ||
437 (minor > (u8)-1) || (minor < 0))
438 return -ERANGE;
439 if ((major != max->pv_major) ||
440 (minor > max->pv_minor))
441 return -EINVAL;
442
443 p->op_proto.pv_major = major;
444 p->op_proto.pv_minor = minor;
445
446 return ocfs2_control_install_private(file);
447}
448
Joel Becker3cfd4ab2008-02-20 14:44:34 -0800449static int ocfs2_control_do_down_msg(struct file *file,
450 struct ocfs2_control_message_down *msg)
451{
452 long nodenum;
453 char *p = NULL;
454
455 if (ocfs2_control_get_handshake_state(file) !=
456 OCFS2_CONTROL_HANDSHAKE_VALID)
457 return -EINVAL;
458
459 if (strncmp(msg->tag, OCFS2_CONTROL_MESSAGE_DOWN_OP,
460 OCFS2_CONTROL_MESSAGE_OP_LEN))
461 return -EINVAL;
462
463 if ((msg->space1 != ' ') || (msg->space2 != ' ') ||
464 (msg->newline != '\n'))
465 return -EINVAL;
466 msg->space1 = msg->space2 = msg->newline = '\0';
467
468 nodenum = simple_strtol(msg->nodestr, &p, 16);
Joel Beckerde870ef2008-02-18 17:07:09 -0800469 if (!p || *p)
470 return -EINVAL;
471
472 if ((nodenum == LONG_MIN) || (nodenum == LONG_MAX) ||
473 (nodenum > INT_MAX) || (nodenum < 0))
474 return -ERANGE;
475
Joel Becker3cfd4ab2008-02-20 14:44:34 -0800476 ocfs2_control_send_down(msg->uuid, nodenum);
Joel Beckerde870ef2008-02-18 17:07:09 -0800477
Joel Becker3cfd4ab2008-02-20 14:44:34 -0800478 return 0;
479}
480
481static ssize_t ocfs2_control_message(struct file *file,
482 const char __user *buf,
483 size_t count)
484{
485 ssize_t ret;
486 union ocfs2_control_message msg;
487
488 /* Try to catch padding issues */
489 WARN_ON(offsetof(struct ocfs2_control_message_down, uuid) !=
490 (sizeof(msg.u_down.tag) + sizeof(msg.u_down.space1)));
491
492 memset(&msg, 0, sizeof(union ocfs2_control_message));
493 ret = ocfs2_control_cfu(&msg, count, buf, count);
494 if (ret)
495 goto out;
496
497 if ((count == OCFS2_CONTROL_MESSAGE_SETNODE_TOTAL_LEN) &&
498 !strncmp(msg.tag, OCFS2_CONTROL_MESSAGE_SETNODE_OP,
499 OCFS2_CONTROL_MESSAGE_OP_LEN))
500 ret = ocfs2_control_do_setnode_msg(file, &msg.u_setn);
Joel Beckerd4b95ee2008-02-20 15:39:44 -0800501 else if ((count == OCFS2_CONTROL_MESSAGE_SETVERSION_TOTAL_LEN) &&
502 !strncmp(msg.tag, OCFS2_CONTROL_MESSAGE_SETVERSION_OP,
503 OCFS2_CONTROL_MESSAGE_OP_LEN))
504 ret = ocfs2_control_do_setversion_msg(file, &msg.u_setv);
Joel Becker3cfd4ab2008-02-20 14:44:34 -0800505 else if ((count == OCFS2_CONTROL_MESSAGE_DOWN_TOTAL_LEN) &&
506 !strncmp(msg.tag, OCFS2_CONTROL_MESSAGE_DOWN_OP,
507 OCFS2_CONTROL_MESSAGE_OP_LEN))
508 ret = ocfs2_control_do_down_msg(file, &msg.u_down);
509 else
510 ret = -EINVAL;
511
512out:
513 return ret ? ret : count;
Joel Beckerde870ef2008-02-18 17:07:09 -0800514}
Joel Becker6427a722008-02-18 19:23:28 -0800515
516static ssize_t ocfs2_control_write(struct file *file,
517 const char __user *buf,
518 size_t count,
519 loff_t *ppos)
Joel Becker8adf0532007-11-28 14:38:40 -0800520{
Joel Becker462c7e62008-02-18 19:40:12 -0800521 ssize_t ret;
522
523 switch (ocfs2_control_get_handshake_state(file)) {
524 case OCFS2_CONTROL_HANDSHAKE_INVALID:
525 ret = -EINVAL;
526 break;
527
528 case OCFS2_CONTROL_HANDSHAKE_READ:
Joel Becker3cfd4ab2008-02-20 14:44:34 -0800529 ret = ocfs2_control_validate_protocol(file, buf,
530 count);
Joel Becker462c7e62008-02-18 19:40:12 -0800531 break;
532
Joel Becker3cfd4ab2008-02-20 14:44:34 -0800533 case OCFS2_CONTROL_HANDSHAKE_PROTOCOL:
Joel Becker462c7e62008-02-18 19:40:12 -0800534 case OCFS2_CONTROL_HANDSHAKE_VALID:
Joel Beckerde870ef2008-02-18 17:07:09 -0800535 ret = ocfs2_control_message(file, buf, count);
Joel Becker462c7e62008-02-18 19:40:12 -0800536 break;
537
538 default:
539 BUG();
540 ret = -EIO;
541 break;
542 }
543
544 return ret;
Joel Becker8adf0532007-11-28 14:38:40 -0800545}
546
Joel Becker462c7e62008-02-18 19:40:12 -0800547/*
548 * This is a naive version. If we ever have a new protocol, we'll expand
549 * it. Probably using seq_file.
550 */
Joel Becker6427a722008-02-18 19:23:28 -0800551static ssize_t ocfs2_control_read(struct file *file,
552 char __user *buf,
553 size_t count,
554 loff_t *ppos)
555{
Akinobu Mita7600c722008-06-09 16:34:23 -0700556 ssize_t ret;
Joel Becker462c7e62008-02-18 19:40:12 -0800557
Akinobu Mita7600c722008-06-09 16:34:23 -0700558 ret = simple_read_from_buffer(buf, count, ppos,
559 OCFS2_CONTROL_PROTO, OCFS2_CONTROL_PROTO_LEN);
Joel Becker462c7e62008-02-18 19:40:12 -0800560
561 /* Have we read the whole protocol list? */
Akinobu Mita7600c722008-06-09 16:34:23 -0700562 if (ret > 0 && *ppos >= OCFS2_CONTROL_PROTO_LEN)
Joel Becker462c7e62008-02-18 19:40:12 -0800563 ocfs2_control_set_handshake_state(file,
564 OCFS2_CONTROL_HANDSHAKE_READ);
565
Akinobu Mita7600c722008-06-09 16:34:23 -0700566 return ret;
Joel Becker6427a722008-02-18 19:23:28 -0800567}
568
569static int ocfs2_control_release(struct inode *inode, struct file *file)
570{
Joel Becker462c7e62008-02-18 19:40:12 -0800571 struct ocfs2_control_private *p = file->private_data;
572
573 mutex_lock(&ocfs2_control_lock);
574
575 if (ocfs2_control_get_handshake_state(file) !=
576 OCFS2_CONTROL_HANDSHAKE_VALID)
577 goto out;
578
Joel Becker6427a722008-02-18 19:23:28 -0800579 if (atomic_dec_and_test(&ocfs2_control_opened)) {
Joel Becker6427a722008-02-18 19:23:28 -0800580 if (!list_empty(&ocfs2_live_connection_list)) {
581 /* XXX: Do bad things! */
582 printk(KERN_ERR
583 "ocfs2: Unexpected release of ocfs2_control!\n"
584 " Loss of cluster connection requires "
585 "an emergency restart!\n");
586 emergency_restart();
587 }
Joel Beckerd4b95ee2008-02-20 15:39:44 -0800588 /*
589 * Last valid close clears the node number and resets
590 * the locking protocol version
591 */
Joel Becker3cfd4ab2008-02-20 14:44:34 -0800592 ocfs2_control_this_node = -1;
Joel Beckerd4b95ee2008-02-20 15:39:44 -0800593 running_proto.pv_major = 0;
Andrew Morton98acbf62014-10-09 15:24:52 -0700594 running_proto.pv_minor = 0;
Joel Becker6427a722008-02-18 19:23:28 -0800595 }
596
Joel Becker462c7e62008-02-18 19:40:12 -0800597out:
598 list_del_init(&p->op_list);
599 file->private_data = NULL;
600
601 mutex_unlock(&ocfs2_control_lock);
602
603 kfree(p);
604
Joel Becker6427a722008-02-18 19:23:28 -0800605 return 0;
606}
607
608static int ocfs2_control_open(struct inode *inode, struct file *file)
609{
Joel Becker462c7e62008-02-18 19:40:12 -0800610 struct ocfs2_control_private *p;
611
612 p = kzalloc(sizeof(struct ocfs2_control_private), GFP_KERNEL);
613 if (!p)
614 return -ENOMEM;
Joel Becker3cfd4ab2008-02-20 14:44:34 -0800615 p->op_this_node = -1;
Joel Becker462c7e62008-02-18 19:40:12 -0800616
617 mutex_lock(&ocfs2_control_lock);
618 file->private_data = p;
619 list_add(&p->op_list, &ocfs2_control_private_list);
620 mutex_unlock(&ocfs2_control_lock);
Joel Becker6427a722008-02-18 19:23:28 -0800621
622 return 0;
623}
624
625static const struct file_operations ocfs2_control_fops = {
626 .open = ocfs2_control_open,
627 .release = ocfs2_control_release,
628 .read = ocfs2_control_read,
629 .write = ocfs2_control_write,
630 .owner = THIS_MODULE,
Arnd Bergmann6038f372010-08-15 18:52:59 +0200631 .llseek = default_llseek,
Joel Becker6427a722008-02-18 19:23:28 -0800632};
633
Adrian Bunk4d8755b2008-04-21 11:49:26 +0300634static struct miscdevice ocfs2_control_device = {
Joel Becker6427a722008-02-18 19:23:28 -0800635 .minor = MISC_DYNAMIC_MINOR,
636 .name = "ocfs2_control",
637 .fops = &ocfs2_control_fops,
638};
639
640static int ocfs2_control_init(void)
641{
642 int rc;
643
644 atomic_set(&ocfs2_control_opened, 0);
645
646 rc = misc_register(&ocfs2_control_device);
647 if (rc)
648 printk(KERN_ERR
649 "ocfs2: Unable to register ocfs2_control device "
650 "(errno %d)\n",
651 -rc);
652
653 return rc;
654}
655
656static void ocfs2_control_exit(void)
657{
658 int rc;
659
660 rc = misc_deregister(&ocfs2_control_device);
661 if (rc)
662 printk(KERN_ERR
663 "ocfs2: Unable to deregister ocfs2_control device "
664 "(errno %d)\n",
665 -rc);
666}
667
David Teiglandcf4d8d72008-02-20 14:29:27 -0800668static void fsdlm_lock_ast_wrapper(void *astarg)
669{
Joel Beckerc0e41332010-01-29 14:46:44 -0800670 struct ocfs2_dlm_lksb *lksb = astarg;
Joel Beckera796d282010-01-28 19:22:39 -0800671 int status = lksb->lksb_fsdlm.sb_status;
David Teiglandcf4d8d72008-02-20 14:29:27 -0800672
David Teiglandcf4d8d72008-02-20 14:29:27 -0800673 /*
674 * For now we're punting on the issue of other non-standard errors
675 * where we can't tell if the unlock_ast or lock_ast should be called.
676 * The main "other error" that's possible is EINVAL which means the
677 * function was called with invalid args, which shouldn't be possible
678 * since the caller here is under our control. Other non-standard
679 * errors probably fall into the same category, or otherwise are fatal
680 * which means we can't carry on anyway.
681 */
682
683 if (status == -DLM_EUNLOCK || status == -DLM_ECANCEL)
Joel Becker110946c2010-01-29 15:46:23 -0800684 lksb->lksb_conn->cc_proto->lp_unlock_ast(lksb, 0);
David Teiglandcf4d8d72008-02-20 14:29:27 -0800685 else
Joel Becker110946c2010-01-29 15:46:23 -0800686 lksb->lksb_conn->cc_proto->lp_lock_ast(lksb);
David Teiglandcf4d8d72008-02-20 14:29:27 -0800687}
688
689static void fsdlm_blocking_ast_wrapper(void *astarg, int level)
690{
Joel Beckerc0e41332010-01-29 14:46:44 -0800691 struct ocfs2_dlm_lksb *lksb = astarg;
Joel Beckera796d282010-01-28 19:22:39 -0800692
Joel Becker110946c2010-01-29 15:46:23 -0800693 lksb->lksb_conn->cc_proto->lp_blocking_ast(lksb, level);
David Teiglandcf4d8d72008-02-20 14:29:27 -0800694}
695
696static int user_dlm_lock(struct ocfs2_cluster_connection *conn,
697 int mode,
Joel Beckerc0e41332010-01-29 14:46:44 -0800698 struct ocfs2_dlm_lksb *lksb,
David Teiglandcf4d8d72008-02-20 14:29:27 -0800699 u32 flags,
700 void *name,
Joel Beckera796d282010-01-28 19:22:39 -0800701 unsigned int namelen)
David Teiglandcf4d8d72008-02-20 14:29:27 -0800702{
703 int ret;
704
705 if (!lksb->lksb_fsdlm.sb_lvbptr)
706 lksb->lksb_fsdlm.sb_lvbptr = (char *)lksb +
707 sizeof(struct dlm_lksb);
708
709 ret = dlm_lock(conn->cc_lockspace, mode, &lksb->lksb_fsdlm,
710 flags|DLM_LKF_NODLCKWT, name, namelen, 0,
Joel Beckera796d282010-01-28 19:22:39 -0800711 fsdlm_lock_ast_wrapper, lksb,
David Teiglandcf4d8d72008-02-20 14:29:27 -0800712 fsdlm_blocking_ast_wrapper);
713 return ret;
714}
715
716static int user_dlm_unlock(struct ocfs2_cluster_connection *conn,
Joel Beckerc0e41332010-01-29 14:46:44 -0800717 struct ocfs2_dlm_lksb *lksb,
Joel Beckera796d282010-01-28 19:22:39 -0800718 u32 flags)
David Teiglandcf4d8d72008-02-20 14:29:27 -0800719{
720 int ret;
721
722 ret = dlm_unlock(conn->cc_lockspace, lksb->lksb_fsdlm.sb_lkid,
Joel Beckera796d282010-01-28 19:22:39 -0800723 flags, &lksb->lksb_fsdlm, lksb);
David Teiglandcf4d8d72008-02-20 14:29:27 -0800724 return ret;
725}
726
Joel Beckerc0e41332010-01-29 14:46:44 -0800727static int user_dlm_lock_status(struct ocfs2_dlm_lksb *lksb)
David Teiglandcf4d8d72008-02-20 14:29:27 -0800728{
729 return lksb->lksb_fsdlm.sb_status;
730}
731
Joel Beckerc0e41332010-01-29 14:46:44 -0800732static int user_dlm_lvb_valid(struct ocfs2_dlm_lksb *lksb)
Joel Becker1c520df2009-06-19 15:14:13 -0700733{
734 int invalid = lksb->lksb_fsdlm.sb_flags & DLM_SBF_VALNOTVALID;
735
736 return !invalid;
737}
738
Joel Beckerc0e41332010-01-29 14:46:44 -0800739static void *user_dlm_lvb(struct ocfs2_dlm_lksb *lksb)
David Teiglandcf4d8d72008-02-20 14:29:27 -0800740{
David Teigland66f502a2008-11-10 16:24:57 -0600741 if (!lksb->lksb_fsdlm.sb_lvbptr)
742 lksb->lksb_fsdlm.sb_lvbptr = (char *)lksb +
743 sizeof(struct dlm_lksb);
David Teiglandcf4d8d72008-02-20 14:29:27 -0800744 return (void *)(lksb->lksb_fsdlm.sb_lvbptr);
745}
746
Joel Beckerc0e41332010-01-29 14:46:44 -0800747static void user_dlm_dump_lksb(struct ocfs2_dlm_lksb *lksb)
David Teiglandcf4d8d72008-02-20 14:29:27 -0800748{
749}
750
Mark Fasheh53da4932008-07-21 14:29:16 -0700751static int user_plock(struct ocfs2_cluster_connection *conn,
752 u64 ino,
753 struct file *file,
754 int cmd,
755 struct file_lock *fl)
756{
757 /*
758 * This more or less just demuxes the plock request into any
759 * one of three dlm calls.
760 *
761 * Internally, fs/dlm will pass these to a misc device, which
762 * a userspace daemon will read and write to.
763 *
764 * For now, cancel requests (which happen internally only),
765 * are turned into unlocks. Most of this function taken from
766 * gfs2_lock.
767 */
768
769 if (cmd == F_CANCELLK) {
770 cmd = F_SETLK;
771 fl->fl_type = F_UNLCK;
772 }
773
774 if (IS_GETLK(cmd))
775 return dlm_posix_get(conn->cc_lockspace, ino, file, fl);
776 else if (fl->fl_type == F_UNLCK)
777 return dlm_posix_unlock(conn->cc_lockspace, ino, file, fl);
778 else
779 return dlm_posix_lock(conn->cc_lockspace, ino, file, cmd, fl);
780}
781
David Teiglandcf4d8d72008-02-20 14:29:27 -0800782/*
783 * Compare a requested locking protocol version against the current one.
784 *
785 * If the major numbers are different, they are incompatible.
786 * If the current minor is greater than the request, they are incompatible.
787 * If the current minor is less than or equal to the request, they are
788 * compatible, and the requester should run at the current minor version.
789 */
790static int fs_protocol_compare(struct ocfs2_protocol_version *existing,
791 struct ocfs2_protocol_version *request)
792{
793 if (existing->pv_major != request->pv_major)
794 return 1;
795
796 if (existing->pv_minor > request->pv_minor)
797 return 1;
798
799 if (existing->pv_minor < request->pv_minor)
800 request->pv_minor = existing->pv_minor;
801
802 return 0;
803}
804
Goldwyn Rodrigues41503632014-01-21 15:48:25 -0800805static void lvb_to_version(char *lvb, struct ocfs2_protocol_version *ver)
806{
807 struct ocfs2_protocol_version *pv =
808 (struct ocfs2_protocol_version *)lvb;
809 /*
810 * ocfs2_protocol_version has two u8 variables, so we don't
811 * need any endian conversion.
812 */
813 ver->pv_major = pv->pv_major;
814 ver->pv_minor = pv->pv_minor;
815}
816
817static void version_to_lvb(struct ocfs2_protocol_version *ver, char *lvb)
818{
819 struct ocfs2_protocol_version *pv =
820 (struct ocfs2_protocol_version *)lvb;
821 /*
822 * ocfs2_protocol_version has two u8 variables, so we don't
823 * need any endian conversion.
824 */
825 pv->pv_major = ver->pv_major;
826 pv->pv_minor = ver->pv_minor;
827}
828
829static void sync_wait_cb(void *arg)
830{
831 struct ocfs2_cluster_connection *conn = arg;
832 struct ocfs2_live_connection *lc = conn->cc_private;
833 complete(&lc->oc_sync_wait);
834}
835
836static int sync_unlock(struct ocfs2_cluster_connection *conn,
837 struct dlm_lksb *lksb, char *name)
838{
839 int error;
840 struct ocfs2_live_connection *lc = conn->cc_private;
841
842 error = dlm_unlock(conn->cc_lockspace, lksb->sb_lkid, 0, lksb, conn);
843 if (error) {
844 printk(KERN_ERR "%s lkid %x error %d\n",
845 name, lksb->sb_lkid, error);
846 return error;
847 }
848
849 wait_for_completion(&lc->oc_sync_wait);
850
851 if (lksb->sb_status != -DLM_EUNLOCK) {
852 printk(KERN_ERR "%s lkid %x status %d\n",
853 name, lksb->sb_lkid, lksb->sb_status);
854 return -1;
855 }
856 return 0;
857}
858
859static int sync_lock(struct ocfs2_cluster_connection *conn,
860 int mode, uint32_t flags,
861 struct dlm_lksb *lksb, char *name)
862{
863 int error, status;
864 struct ocfs2_live_connection *lc = conn->cc_private;
865
866 error = dlm_lock(conn->cc_lockspace, mode, lksb, flags,
867 name, strlen(name),
868 0, sync_wait_cb, conn, NULL);
869 if (error) {
870 printk(KERN_ERR "%s lkid %x flags %x mode %d error %d\n",
871 name, lksb->sb_lkid, flags, mode, error);
872 return error;
873 }
874
875 wait_for_completion(&lc->oc_sync_wait);
876
877 status = lksb->sb_status;
878
879 if (status && status != -EAGAIN) {
880 printk(KERN_ERR "%s lkid %x flags %x mode %d status %d\n",
881 name, lksb->sb_lkid, flags, mode, status);
882 }
883
884 return status;
885}
886
887
888static int version_lock(struct ocfs2_cluster_connection *conn, int mode,
889 int flags)
890{
891 struct ocfs2_live_connection *lc = conn->cc_private;
892 return sync_lock(conn, mode, flags,
893 &lc->oc_version_lksb, VERSION_LOCK);
894}
895
896static int version_unlock(struct ocfs2_cluster_connection *conn)
897{
898 struct ocfs2_live_connection *lc = conn->cc_private;
899 return sync_unlock(conn, &lc->oc_version_lksb, VERSION_LOCK);
900}
901
Goldwyn Rodriguesc994c2e2014-01-21 15:48:32 -0800902/* get_protocol_version()
903 *
904 * To exchange ocfs2 versioning, we use the LVB of the version dlm lock.
905 * The algorithm is:
906 * 1. Attempt to take the lock in EX mode (non-blocking).
907 * 2. If successful (which means it is the first mount), write the
908 * version number and downconvert to PR lock.
909 * 3. If unsuccessful (returns -EAGAIN), read the version from the LVB after
910 * taking the PR lock.
911 */
912
913static int get_protocol_version(struct ocfs2_cluster_connection *conn)
914{
915 int ret;
916 struct ocfs2_live_connection *lc = conn->cc_private;
917 struct ocfs2_protocol_version pv;
918
919 running_proto.pv_major =
920 ocfs2_user_plugin.sp_max_proto.pv_major;
921 running_proto.pv_minor =
922 ocfs2_user_plugin.sp_max_proto.pv_minor;
923
924 lc->oc_version_lksb.sb_lvbptr = lc->oc_lvb;
925 ret = version_lock(conn, DLM_LOCK_EX,
926 DLM_LKF_VALBLK|DLM_LKF_NOQUEUE);
927 if (!ret) {
928 conn->cc_version.pv_major = running_proto.pv_major;
929 conn->cc_version.pv_minor = running_proto.pv_minor;
930 version_to_lvb(&running_proto, lc->oc_lvb);
931 version_lock(conn, DLM_LOCK_PR, DLM_LKF_CONVERT|DLM_LKF_VALBLK);
932 } else if (ret == -EAGAIN) {
933 ret = version_lock(conn, DLM_LOCK_PR, DLM_LKF_VALBLK);
934 if (ret)
935 goto out;
936 lvb_to_version(lc->oc_lvb, &pv);
937
938 if ((pv.pv_major != running_proto.pv_major) ||
939 (pv.pv_minor > running_proto.pv_minor)) {
940 ret = -EINVAL;
941 goto out;
942 }
943
944 conn->cc_version.pv_major = pv.pv_major;
945 conn->cc_version.pv_minor = pv.pv_minor;
946 }
947out:
948 return ret;
949}
950
Goldwyn Rodrigues66e188f2014-01-21 15:48:22 -0800951static void user_recover_prep(void *arg)
952{
953}
954
955static void user_recover_slot(void *arg, struct dlm_slot *slot)
956{
957 struct ocfs2_cluster_connection *conn = arg;
958 printk(KERN_INFO "ocfs2: Node %d/%d down. Initiating recovery.\n",
959 slot->nodeid, slot->slot);
960 conn->cc_recovery_handler(slot->nodeid, conn->cc_recovery_data);
961
962}
963
964static void user_recover_done(void *arg, struct dlm_slot *slots,
965 int num_slots, int our_slot,
966 uint32_t generation)
967{
968 struct ocfs2_cluster_connection *conn = arg;
969 struct ocfs2_live_connection *lc = conn->cc_private;
970 int i;
971
972 for (i = 0; i < num_slots; i++)
973 if (slots[i].slot == our_slot) {
974 atomic_set(&lc->oc_this_node, slots[i].nodeid);
975 break;
976 }
977
978 lc->oc_our_slot = our_slot;
Goldwyn Rodriguesc994c2e2014-01-21 15:48:32 -0800979 wake_up(&lc->oc_wait);
Goldwyn Rodrigues66e188f2014-01-21 15:48:22 -0800980}
981
Wei Yongjun16eac4b2014-01-21 15:48:37 -0800982static const struct dlm_lockspace_ops ocfs2_ls_ops = {
Goldwyn Rodrigues66e188f2014-01-21 15:48:22 -0800983 .recover_prep = user_recover_prep,
984 .recover_slot = user_recover_slot,
985 .recover_done = user_recover_done,
986};
987
Goldwyn Rodriguesc994c2e2014-01-21 15:48:32 -0800988static int user_cluster_disconnect(struct ocfs2_cluster_connection *conn)
989{
990 version_unlock(conn);
991 dlm_release_lockspace(conn->cc_lockspace, 2);
992 conn->cc_lockspace = NULL;
993 ocfs2_live_connection_drop(conn->cc_private);
994 conn->cc_private = NULL;
995 return 0;
996}
997
David Teiglandcf4d8d72008-02-20 14:29:27 -0800998static int user_cluster_connect(struct ocfs2_cluster_connection *conn)
999{
1000 dlm_lockspace_t *fsdlm;
Goldwyn Rodrigues24aa3382014-01-21 15:48:23 -08001001 struct ocfs2_live_connection *lc;
Goldwyn Rodriguesc994c2e2014-01-21 15:48:32 -08001002 int rc, ops_rv;
David Teiglandcf4d8d72008-02-20 14:29:27 -08001003
1004 BUG_ON(conn == NULL);
1005
Goldwyn Rodrigues24aa3382014-01-21 15:48:23 -08001006 lc = kzalloc(sizeof(struct ocfs2_live_connection), GFP_KERNEL);
Markus Elfring43ee9ca2015-04-14 15:42:59 -07001007 if (!lc)
1008 return -ENOMEM;
Goldwyn Rodrigues24aa3382014-01-21 15:48:23 -08001009
Goldwyn Rodriguesc994c2e2014-01-21 15:48:32 -08001010 init_waitqueue_head(&lc->oc_wait);
1011 init_completion(&lc->oc_sync_wait);
1012 atomic_set(&lc->oc_this_node, 0);
1013 conn->cc_private = lc;
1014 lc->oc_type = NO_CONTROLD;
1015
1016 rc = dlm_new_lockspace(conn->cc_name, conn->cc_cluster_name,
1017 DLM_LSFL_FS, DLM_LVB_LEN,
1018 &ocfs2_ls_ops, conn, &ops_rv, &fsdlm);
1019 if (rc)
1020 goto out;
1021
1022 if (ops_rv == -EOPNOTSUPP) {
1023 lc->oc_type = WITH_CONTROLD;
1024 printk(KERN_NOTICE "ocfs2: You seem to be using an older "
1025 "version of dlm_controld and/or ocfs2-tools."
1026 " Please consider upgrading.\n");
1027 } else if (ops_rv) {
1028 rc = ops_rv;
1029 goto out;
1030 }
1031 conn->cc_lockspace = fsdlm;
1032
Goldwyn Rodrigues24aa3382014-01-21 15:48:23 -08001033 rc = ocfs2_live_connection_attach(conn, lc);
David Teiglandcf4d8d72008-02-20 14:29:27 -08001034 if (rc)
1035 goto out;
1036
Goldwyn Rodriguesc994c2e2014-01-21 15:48:32 -08001037 if (lc->oc_type == NO_CONTROLD) {
1038 rc = get_protocol_version(conn);
1039 if (rc) {
1040 printk(KERN_ERR "ocfs2: Could not determine"
1041 " locking version\n");
1042 user_cluster_disconnect(conn);
1043 goto out;
1044 }
1045 wait_event(lc->oc_wait, (atomic_read(&lc->oc_this_node) > 0));
1046 }
1047
David Teiglandcf4d8d72008-02-20 14:29:27 -08001048 /*
1049 * running_proto must have been set before we allowed any mounts
1050 * to proceed.
1051 */
1052 if (fs_protocol_compare(&running_proto, &conn->cc_version)) {
1053 printk(KERN_ERR
1054 "Unable to mount with fs locking protocol version "
Goldwyn Rodriguesc994c2e2014-01-21 15:48:32 -08001055 "%u.%u because negotiated protocol is %u.%u\n",
David Teiglandcf4d8d72008-02-20 14:29:27 -08001056 conn->cc_version.pv_major, conn->cc_version.pv_minor,
1057 running_proto.pv_major, running_proto.pv_minor);
1058 rc = -EPROTO;
Goldwyn Rodrigues24aa3382014-01-21 15:48:23 -08001059 ocfs2_live_connection_drop(lc);
1060 lc = NULL;
David Teiglandcf4d8d72008-02-20 14:29:27 -08001061 }
1062
David Teiglandcf4d8d72008-02-20 14:29:27 -08001063out:
Markus Elfringfd90d4df2015-04-14 15:42:42 -07001064 if (rc)
Goldwyn Rodrigues24aa3382014-01-21 15:48:23 -08001065 kfree(lc);
David Teiglandcf4d8d72008-02-20 14:29:27 -08001066 return rc;
1067}
1068
David Teiglandcf4d8d72008-02-20 14:29:27 -08001069
Goldwyn Rodrigues3e834152014-01-21 15:48:24 -08001070static int user_cluster_this_node(struct ocfs2_cluster_connection *conn,
1071 unsigned int *this_node)
David Teiglandcf4d8d72008-02-20 14:29:27 -08001072{
1073 int rc;
Goldwyn Rodrigues3e834152014-01-21 15:48:24 -08001074 struct ocfs2_live_connection *lc = conn->cc_private;
David Teiglandcf4d8d72008-02-20 14:29:27 -08001075
Goldwyn Rodrigues3e834152014-01-21 15:48:24 -08001076 if (lc->oc_type == WITH_CONTROLD)
1077 rc = ocfs2_control_get_this_node();
Goldwyn Rodriguesc994c2e2014-01-21 15:48:32 -08001078 else if (lc->oc_type == NO_CONTROLD)
1079 rc = atomic_read(&lc->oc_this_node);
Goldwyn Rodrigues3e834152014-01-21 15:48:24 -08001080 else
1081 rc = -EINVAL;
Goldwyn Rodriguesc994c2e2014-01-21 15:48:32 -08001082
David Teiglandcf4d8d72008-02-20 14:29:27 -08001083 if (rc < 0)
1084 return rc;
1085
1086 *this_node = rc;
1087 return 0;
1088}
1089
Joel Beckera12630b2008-05-09 18:49:29 -07001090static struct ocfs2_stack_operations ocfs2_user_plugin_ops = {
David Teiglandcf4d8d72008-02-20 14:29:27 -08001091 .connect = user_cluster_connect,
1092 .disconnect = user_cluster_disconnect,
1093 .this_node = user_cluster_this_node,
1094 .dlm_lock = user_dlm_lock,
1095 .dlm_unlock = user_dlm_unlock,
1096 .lock_status = user_dlm_lock_status,
Joel Becker1c520df2009-06-19 15:14:13 -07001097 .lvb_valid = user_dlm_lvb_valid,
David Teiglandcf4d8d72008-02-20 14:29:27 -08001098 .lock_lvb = user_dlm_lvb,
Mark Fasheh53da4932008-07-21 14:29:16 -07001099 .plock = user_plock,
David Teiglandcf4d8d72008-02-20 14:29:27 -08001100 .dump_lksb = user_dlm_dump_lksb,
1101};
1102
Joel Beckera12630b2008-05-09 18:49:29 -07001103static struct ocfs2_stack_plugin ocfs2_user_plugin = {
David Teiglandcf4d8d72008-02-20 14:29:27 -08001104 .sp_name = "user",
Joel Beckera12630b2008-05-09 18:49:29 -07001105 .sp_ops = &ocfs2_user_plugin_ops,
David Teiglandcf4d8d72008-02-20 14:29:27 -08001106 .sp_owner = THIS_MODULE,
1107};
1108
1109
Joel Beckera12630b2008-05-09 18:49:29 -07001110static int __init ocfs2_user_plugin_init(void)
Joel Becker6427a722008-02-18 19:23:28 -08001111{
David Teiglandcf4d8d72008-02-20 14:29:27 -08001112 int rc;
1113
1114 rc = ocfs2_control_init();
1115 if (!rc) {
Joel Beckera12630b2008-05-09 18:49:29 -07001116 rc = ocfs2_stack_glue_register(&ocfs2_user_plugin);
David Teiglandcf4d8d72008-02-20 14:29:27 -08001117 if (rc)
1118 ocfs2_control_exit();
1119 }
1120
1121 return rc;
Joel Becker6427a722008-02-18 19:23:28 -08001122}
1123
Joel Beckera12630b2008-05-09 18:49:29 -07001124static void __exit ocfs2_user_plugin_exit(void)
Joel Becker8adf0532007-11-28 14:38:40 -08001125{
Joel Beckera12630b2008-05-09 18:49:29 -07001126 ocfs2_stack_glue_unregister(&ocfs2_user_plugin);
Joel Becker6427a722008-02-18 19:23:28 -08001127 ocfs2_control_exit();
Joel Becker8adf0532007-11-28 14:38:40 -08001128}
1129
1130MODULE_AUTHOR("Oracle");
1131MODULE_DESCRIPTION("ocfs2 driver for userspace cluster stacks");
1132MODULE_LICENSE("GPL");
Joel Beckera12630b2008-05-09 18:49:29 -07001133module_init(ocfs2_user_plugin_init);
1134module_exit(ocfs2_user_plugin_exit);