omap: gpio: fix incorrect matching of IRQ_TYPE_EDGE_BOTH
[linux-2.6/cjktty.git] / fs / ocfs2 / stack_user.c
blob5ae8812b28647975bdba323cd09997f6c6213d5c
1 /* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
4 * stack_user.c
6 * Code which interfaces ocfs2 with fs/dlm and a userspace stack.
8 * Copyright (C) 2007 Oracle. All rights reserved.
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.
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.
20 #include <linux/module.h>
21 #include <linux/fs.h>
22 #include <linux/miscdevice.h>
23 #include <linux/mutex.h>
24 #include <linux/smp_lock.h>
25 #include <linux/reboot.h>
26 #include <asm/uaccess.h>
28 #include "stackglue.h"
30 #include <linux/dlm_plock.h>
33 * The control protocol starts with a handshake. Until the handshake
34 * is complete, the control device will fail all write(2)s.
36 * The handshake is simple. First, the client reads until EOF. Each line
37 * of output is a supported protocol tag. All protocol tags are a single
38 * character followed by a two hex digit version number. Currently the
39 * only things supported is T01, for "Text-base version 0x01". Next, the
40 * client writes the version they would like to use, including the newline.
41 * Thus, the protocol tag is 'T01\n'. If the version tag written is
42 * unknown, -EINVAL is returned. Once the negotiation is complete, the
43 * client can start sending messages.
45 * The T01 protocol has three messages. First is the "SETN" message.
46 * It has the following syntax:
48 * SETN<space><8-char-hex-nodenum><newline>
50 * This is 14 characters.
52 * The "SETN" message must be the first message following the protocol.
53 * It tells ocfs2_control the local node number.
55 * Next comes the "SETV" message. It has the following syntax:
57 * SETV<space><2-char-hex-major><space><2-char-hex-minor><newline>
59 * This is 11 characters.
61 * The "SETV" message sets the filesystem locking protocol version as
62 * negotiated by the client. The client negotiates based on the maximum
63 * version advertised in /sys/fs/ocfs2/max_locking_protocol. The major
64 * number from the "SETV" message must match
65 * ocfs2_user_plugin.sp_max_proto.pv_major, and the minor number
66 * must be less than or equal to ...sp_max_version.pv_minor.
68 * Once this information has been set, mounts will be allowed. From this
69 * point on, the "DOWN" message can be sent for node down notification.
70 * It has the following syntax:
72 * DOWN<space><32-char-cap-hex-uuid><space><8-char-hex-nodenum><newline>
74 * eg:
76 * DOWN 632A924FDD844190BDA93C0DF6B94899 00000001\n
78 * This is 47 characters.
82 * Whether or not the client has done the handshake.
83 * For now, we have just one protocol version.
85 #define OCFS2_CONTROL_PROTO "T01\n"
86 #define OCFS2_CONTROL_PROTO_LEN 4
88 /* Handshake states */
89 #define OCFS2_CONTROL_HANDSHAKE_INVALID (0)
90 #define OCFS2_CONTROL_HANDSHAKE_READ (1)
91 #define OCFS2_CONTROL_HANDSHAKE_PROTOCOL (2)
92 #define OCFS2_CONTROL_HANDSHAKE_VALID (3)
94 /* Messages */
95 #define OCFS2_CONTROL_MESSAGE_OP_LEN 4
96 #define OCFS2_CONTROL_MESSAGE_SETNODE_OP "SETN"
97 #define OCFS2_CONTROL_MESSAGE_SETNODE_TOTAL_LEN 14
98 #define OCFS2_CONTROL_MESSAGE_SETVERSION_OP "SETV"
99 #define OCFS2_CONTROL_MESSAGE_SETVERSION_TOTAL_LEN 11
100 #define OCFS2_CONTROL_MESSAGE_DOWN_OP "DOWN"
101 #define OCFS2_CONTROL_MESSAGE_DOWN_TOTAL_LEN 47
102 #define OCFS2_TEXT_UUID_LEN 32
103 #define OCFS2_CONTROL_MESSAGE_VERNUM_LEN 2
104 #define OCFS2_CONTROL_MESSAGE_NODENUM_LEN 8
107 * ocfs2_live_connection is refcounted because the filesystem and
108 * miscdevice sides can detach in different order. Let's just be safe.
110 struct ocfs2_live_connection {
111 struct list_head oc_list;
112 struct ocfs2_cluster_connection *oc_conn;
115 struct ocfs2_control_private {
116 struct list_head op_list;
117 int op_state;
118 int op_this_node;
119 struct ocfs2_protocol_version op_proto;
122 /* SETN<space><8-char-hex-nodenum><newline> */
123 struct ocfs2_control_message_setn {
124 char tag[OCFS2_CONTROL_MESSAGE_OP_LEN];
125 char space;
126 char nodestr[OCFS2_CONTROL_MESSAGE_NODENUM_LEN];
127 char newline;
130 /* SETV<space><2-char-hex-major><space><2-char-hex-minor><newline> */
131 struct ocfs2_control_message_setv {
132 char tag[OCFS2_CONTROL_MESSAGE_OP_LEN];
133 char space1;
134 char major[OCFS2_CONTROL_MESSAGE_VERNUM_LEN];
135 char space2;
136 char minor[OCFS2_CONTROL_MESSAGE_VERNUM_LEN];
137 char newline;
140 /* DOWN<space><32-char-cap-hex-uuid><space><8-char-hex-nodenum><newline> */
141 struct ocfs2_control_message_down {
142 char tag[OCFS2_CONTROL_MESSAGE_OP_LEN];
143 char space1;
144 char uuid[OCFS2_TEXT_UUID_LEN];
145 char space2;
146 char nodestr[OCFS2_CONTROL_MESSAGE_NODENUM_LEN];
147 char newline;
150 union ocfs2_control_message {
151 char tag[OCFS2_CONTROL_MESSAGE_OP_LEN];
152 struct ocfs2_control_message_setn u_setn;
153 struct ocfs2_control_message_setv u_setv;
154 struct ocfs2_control_message_down u_down;
157 static struct ocfs2_stack_plugin ocfs2_user_plugin;
159 static atomic_t ocfs2_control_opened;
160 static int ocfs2_control_this_node = -1;
161 static struct ocfs2_protocol_version running_proto;
163 static LIST_HEAD(ocfs2_live_connection_list);
164 static LIST_HEAD(ocfs2_control_private_list);
165 static DEFINE_MUTEX(ocfs2_control_lock);
167 static inline void ocfs2_control_set_handshake_state(struct file *file,
168 int state)
170 struct ocfs2_control_private *p = file->private_data;
171 p->op_state = state;
174 static inline int ocfs2_control_get_handshake_state(struct file *file)
176 struct ocfs2_control_private *p = file->private_data;
177 return p->op_state;
180 static struct ocfs2_live_connection *ocfs2_connection_find(const char *name)
182 size_t len = strlen(name);
183 struct ocfs2_live_connection *c;
185 BUG_ON(!mutex_is_locked(&ocfs2_control_lock));
187 list_for_each_entry(c, &ocfs2_live_connection_list, oc_list) {
188 if ((c->oc_conn->cc_namelen == len) &&
189 !strncmp(c->oc_conn->cc_name, name, len))
190 return c;
193 return c;
197 * ocfs2_live_connection structures are created underneath the ocfs2
198 * mount path. Since the VFS prevents multiple calls to
199 * fill_super(), we can't get dupes here.
201 static int ocfs2_live_connection_new(struct ocfs2_cluster_connection *conn,
202 struct ocfs2_live_connection **c_ret)
204 int rc = 0;
205 struct ocfs2_live_connection *c;
207 c = kzalloc(sizeof(struct ocfs2_live_connection), GFP_KERNEL);
208 if (!c)
209 return -ENOMEM;
211 mutex_lock(&ocfs2_control_lock);
212 c->oc_conn = conn;
214 if (atomic_read(&ocfs2_control_opened))
215 list_add(&c->oc_list, &ocfs2_live_connection_list);
216 else {
217 printk(KERN_ERR
218 "ocfs2: Userspace control daemon is not present\n");
219 rc = -ESRCH;
222 mutex_unlock(&ocfs2_control_lock);
224 if (!rc)
225 *c_ret = c;
226 else
227 kfree(c);
229 return rc;
233 * This function disconnects the cluster connection from ocfs2_control.
234 * Afterwards, userspace can't affect the cluster connection.
236 static void ocfs2_live_connection_drop(struct ocfs2_live_connection *c)
238 mutex_lock(&ocfs2_control_lock);
239 list_del_init(&c->oc_list);
240 c->oc_conn = NULL;
241 mutex_unlock(&ocfs2_control_lock);
243 kfree(c);
246 static int ocfs2_control_cfu(void *target, size_t target_len,
247 const char __user *buf, size_t count)
249 /* The T01 expects write(2) calls to have exactly one command */
250 if ((count != target_len) ||
251 (count > sizeof(union ocfs2_control_message)))
252 return -EINVAL;
254 if (copy_from_user(target, buf, target_len))
255 return -EFAULT;
257 return 0;
260 static ssize_t ocfs2_control_validate_protocol(struct file *file,
261 const char __user *buf,
262 size_t count)
264 ssize_t ret;
265 char kbuf[OCFS2_CONTROL_PROTO_LEN];
267 ret = ocfs2_control_cfu(kbuf, OCFS2_CONTROL_PROTO_LEN,
268 buf, count);
269 if (ret)
270 return ret;
272 if (strncmp(kbuf, OCFS2_CONTROL_PROTO, OCFS2_CONTROL_PROTO_LEN))
273 return -EINVAL;
275 ocfs2_control_set_handshake_state(file,
276 OCFS2_CONTROL_HANDSHAKE_PROTOCOL);
278 return count;
281 static void ocfs2_control_send_down(const char *uuid,
282 int nodenum)
284 struct ocfs2_live_connection *c;
286 mutex_lock(&ocfs2_control_lock);
288 c = ocfs2_connection_find(uuid);
289 if (c) {
290 BUG_ON(c->oc_conn == NULL);
291 c->oc_conn->cc_recovery_handler(nodenum,
292 c->oc_conn->cc_recovery_data);
295 mutex_unlock(&ocfs2_control_lock);
299 * Called whenever configuration elements are sent to /dev/ocfs2_control.
300 * If all configuration elements are present, try to set the global
301 * values. If there is a problem, return an error. Skip any missing
302 * elements, and only bump ocfs2_control_opened when we have all elements
303 * and are successful.
305 static int ocfs2_control_install_private(struct file *file)
307 int rc = 0;
308 int set_p = 1;
309 struct ocfs2_control_private *p = file->private_data;
311 BUG_ON(p->op_state != OCFS2_CONTROL_HANDSHAKE_PROTOCOL);
313 mutex_lock(&ocfs2_control_lock);
315 if (p->op_this_node < 0) {
316 set_p = 0;
317 } else if ((ocfs2_control_this_node >= 0) &&
318 (ocfs2_control_this_node != p->op_this_node)) {
319 rc = -EINVAL;
320 goto out_unlock;
323 if (!p->op_proto.pv_major) {
324 set_p = 0;
325 } else if (!list_empty(&ocfs2_live_connection_list) &&
326 ((running_proto.pv_major != p->op_proto.pv_major) ||
327 (running_proto.pv_minor != p->op_proto.pv_minor))) {
328 rc = -EINVAL;
329 goto out_unlock;
332 if (set_p) {
333 ocfs2_control_this_node = p->op_this_node;
334 running_proto.pv_major = p->op_proto.pv_major;
335 running_proto.pv_minor = p->op_proto.pv_minor;
338 out_unlock:
339 mutex_unlock(&ocfs2_control_lock);
341 if (!rc && set_p) {
342 /* We set the global values successfully */
343 atomic_inc(&ocfs2_control_opened);
344 ocfs2_control_set_handshake_state(file,
345 OCFS2_CONTROL_HANDSHAKE_VALID);
348 return rc;
351 static int ocfs2_control_get_this_node(void)
353 int rc;
355 mutex_lock(&ocfs2_control_lock);
356 if (ocfs2_control_this_node < 0)
357 rc = -EINVAL;
358 else
359 rc = ocfs2_control_this_node;
360 mutex_unlock(&ocfs2_control_lock);
362 return rc;
365 static int ocfs2_control_do_setnode_msg(struct file *file,
366 struct ocfs2_control_message_setn *msg)
368 long nodenum;
369 char *ptr = NULL;
370 struct ocfs2_control_private *p = file->private_data;
372 if (ocfs2_control_get_handshake_state(file) !=
373 OCFS2_CONTROL_HANDSHAKE_PROTOCOL)
374 return -EINVAL;
376 if (strncmp(msg->tag, OCFS2_CONTROL_MESSAGE_SETNODE_OP,
377 OCFS2_CONTROL_MESSAGE_OP_LEN))
378 return -EINVAL;
380 if ((msg->space != ' ') || (msg->newline != '\n'))
381 return -EINVAL;
382 msg->space = msg->newline = '\0';
384 nodenum = simple_strtol(msg->nodestr, &ptr, 16);
385 if (!ptr || *ptr)
386 return -EINVAL;
388 if ((nodenum == LONG_MIN) || (nodenum == LONG_MAX) ||
389 (nodenum > INT_MAX) || (nodenum < 0))
390 return -ERANGE;
391 p->op_this_node = nodenum;
393 return ocfs2_control_install_private(file);
396 static int ocfs2_control_do_setversion_msg(struct file *file,
397 struct ocfs2_control_message_setv *msg)
399 long major, minor;
400 char *ptr = NULL;
401 struct ocfs2_control_private *p = file->private_data;
402 struct ocfs2_protocol_version *max =
403 &ocfs2_user_plugin.sp_max_proto;
405 if (ocfs2_control_get_handshake_state(file) !=
406 OCFS2_CONTROL_HANDSHAKE_PROTOCOL)
407 return -EINVAL;
409 if (strncmp(msg->tag, OCFS2_CONTROL_MESSAGE_SETVERSION_OP,
410 OCFS2_CONTROL_MESSAGE_OP_LEN))
411 return -EINVAL;
413 if ((msg->space1 != ' ') || (msg->space2 != ' ') ||
414 (msg->newline != '\n'))
415 return -EINVAL;
416 msg->space1 = msg->space2 = msg->newline = '\0';
418 major = simple_strtol(msg->major, &ptr, 16);
419 if (!ptr || *ptr)
420 return -EINVAL;
421 minor = simple_strtol(msg->minor, &ptr, 16);
422 if (!ptr || *ptr)
423 return -EINVAL;
426 * The major must be between 1 and 255, inclusive. The minor
427 * must be between 0 and 255, inclusive. The version passed in
428 * must be within the maximum version supported by the filesystem.
430 if ((major == LONG_MIN) || (major == LONG_MAX) ||
431 (major > (u8)-1) || (major < 1))
432 return -ERANGE;
433 if ((minor == LONG_MIN) || (minor == LONG_MAX) ||
434 (minor > (u8)-1) || (minor < 0))
435 return -ERANGE;
436 if ((major != max->pv_major) ||
437 (minor > max->pv_minor))
438 return -EINVAL;
440 p->op_proto.pv_major = major;
441 p->op_proto.pv_minor = minor;
443 return ocfs2_control_install_private(file);
446 static int ocfs2_control_do_down_msg(struct file *file,
447 struct ocfs2_control_message_down *msg)
449 long nodenum;
450 char *p = NULL;
452 if (ocfs2_control_get_handshake_state(file) !=
453 OCFS2_CONTROL_HANDSHAKE_VALID)
454 return -EINVAL;
456 if (strncmp(msg->tag, OCFS2_CONTROL_MESSAGE_DOWN_OP,
457 OCFS2_CONTROL_MESSAGE_OP_LEN))
458 return -EINVAL;
460 if ((msg->space1 != ' ') || (msg->space2 != ' ') ||
461 (msg->newline != '\n'))
462 return -EINVAL;
463 msg->space1 = msg->space2 = msg->newline = '\0';
465 nodenum = simple_strtol(msg->nodestr, &p, 16);
466 if (!p || *p)
467 return -EINVAL;
469 if ((nodenum == LONG_MIN) || (nodenum == LONG_MAX) ||
470 (nodenum > INT_MAX) || (nodenum < 0))
471 return -ERANGE;
473 ocfs2_control_send_down(msg->uuid, nodenum);
475 return 0;
478 static ssize_t ocfs2_control_message(struct file *file,
479 const char __user *buf,
480 size_t count)
482 ssize_t ret;
483 union ocfs2_control_message msg;
485 /* Try to catch padding issues */
486 WARN_ON(offsetof(struct ocfs2_control_message_down, uuid) !=
487 (sizeof(msg.u_down.tag) + sizeof(msg.u_down.space1)));
489 memset(&msg, 0, sizeof(union ocfs2_control_message));
490 ret = ocfs2_control_cfu(&msg, count, buf, count);
491 if (ret)
492 goto out;
494 if ((count == OCFS2_CONTROL_MESSAGE_SETNODE_TOTAL_LEN) &&
495 !strncmp(msg.tag, OCFS2_CONTROL_MESSAGE_SETNODE_OP,
496 OCFS2_CONTROL_MESSAGE_OP_LEN))
497 ret = ocfs2_control_do_setnode_msg(file, &msg.u_setn);
498 else if ((count == OCFS2_CONTROL_MESSAGE_SETVERSION_TOTAL_LEN) &&
499 !strncmp(msg.tag, OCFS2_CONTROL_MESSAGE_SETVERSION_OP,
500 OCFS2_CONTROL_MESSAGE_OP_LEN))
501 ret = ocfs2_control_do_setversion_msg(file, &msg.u_setv);
502 else if ((count == OCFS2_CONTROL_MESSAGE_DOWN_TOTAL_LEN) &&
503 !strncmp(msg.tag, OCFS2_CONTROL_MESSAGE_DOWN_OP,
504 OCFS2_CONTROL_MESSAGE_OP_LEN))
505 ret = ocfs2_control_do_down_msg(file, &msg.u_down);
506 else
507 ret = -EINVAL;
509 out:
510 return ret ? ret : count;
513 static ssize_t ocfs2_control_write(struct file *file,
514 const char __user *buf,
515 size_t count,
516 loff_t *ppos)
518 ssize_t ret;
520 switch (ocfs2_control_get_handshake_state(file)) {
521 case OCFS2_CONTROL_HANDSHAKE_INVALID:
522 ret = -EINVAL;
523 break;
525 case OCFS2_CONTROL_HANDSHAKE_READ:
526 ret = ocfs2_control_validate_protocol(file, buf,
527 count);
528 break;
530 case OCFS2_CONTROL_HANDSHAKE_PROTOCOL:
531 case OCFS2_CONTROL_HANDSHAKE_VALID:
532 ret = ocfs2_control_message(file, buf, count);
533 break;
535 default:
536 BUG();
537 ret = -EIO;
538 break;
541 return ret;
545 * This is a naive version. If we ever have a new protocol, we'll expand
546 * it. Probably using seq_file.
548 static ssize_t ocfs2_control_read(struct file *file,
549 char __user *buf,
550 size_t count,
551 loff_t *ppos)
553 ssize_t ret;
555 ret = simple_read_from_buffer(buf, count, ppos,
556 OCFS2_CONTROL_PROTO, OCFS2_CONTROL_PROTO_LEN);
558 /* Have we read the whole protocol list? */
559 if (ret > 0 && *ppos >= OCFS2_CONTROL_PROTO_LEN)
560 ocfs2_control_set_handshake_state(file,
561 OCFS2_CONTROL_HANDSHAKE_READ);
563 return ret;
566 static int ocfs2_control_release(struct inode *inode, struct file *file)
568 struct ocfs2_control_private *p = file->private_data;
570 mutex_lock(&ocfs2_control_lock);
572 if (ocfs2_control_get_handshake_state(file) !=
573 OCFS2_CONTROL_HANDSHAKE_VALID)
574 goto out;
576 if (atomic_dec_and_test(&ocfs2_control_opened)) {
577 if (!list_empty(&ocfs2_live_connection_list)) {
578 /* XXX: Do bad things! */
579 printk(KERN_ERR
580 "ocfs2: Unexpected release of ocfs2_control!\n"
581 " Loss of cluster connection requires "
582 "an emergency restart!\n");
583 emergency_restart();
586 * Last valid close clears the node number and resets
587 * the locking protocol version
589 ocfs2_control_this_node = -1;
590 running_proto.pv_major = 0;
591 running_proto.pv_major = 0;
594 out:
595 list_del_init(&p->op_list);
596 file->private_data = NULL;
598 mutex_unlock(&ocfs2_control_lock);
600 kfree(p);
602 return 0;
605 static int ocfs2_control_open(struct inode *inode, struct file *file)
607 struct ocfs2_control_private *p;
609 p = kzalloc(sizeof(struct ocfs2_control_private), GFP_KERNEL);
610 if (!p)
611 return -ENOMEM;
612 p->op_this_node = -1;
614 lock_kernel();
615 mutex_lock(&ocfs2_control_lock);
616 file->private_data = p;
617 list_add(&p->op_list, &ocfs2_control_private_list);
618 mutex_unlock(&ocfs2_control_lock);
619 unlock_kernel();
621 return 0;
624 static const struct file_operations ocfs2_control_fops = {
625 .open = ocfs2_control_open,
626 .release = ocfs2_control_release,
627 .read = ocfs2_control_read,
628 .write = ocfs2_control_write,
629 .owner = THIS_MODULE,
632 static struct miscdevice ocfs2_control_device = {
633 .minor = MISC_DYNAMIC_MINOR,
634 .name = "ocfs2_control",
635 .fops = &ocfs2_control_fops,
638 static int ocfs2_control_init(void)
640 int rc;
642 atomic_set(&ocfs2_control_opened, 0);
644 rc = misc_register(&ocfs2_control_device);
645 if (rc)
646 printk(KERN_ERR
647 "ocfs2: Unable to register ocfs2_control device "
648 "(errno %d)\n",
649 -rc);
651 return rc;
654 static void ocfs2_control_exit(void)
656 int rc;
658 rc = misc_deregister(&ocfs2_control_device);
659 if (rc)
660 printk(KERN_ERR
661 "ocfs2: Unable to deregister ocfs2_control device "
662 "(errno %d)\n",
663 -rc);
666 static void fsdlm_lock_ast_wrapper(void *astarg)
668 struct ocfs2_dlm_lksb *lksb = astarg;
669 int status = lksb->lksb_fsdlm.sb_status;
672 * For now we're punting on the issue of other non-standard errors
673 * where we can't tell if the unlock_ast or lock_ast should be called.
674 * The main "other error" that's possible is EINVAL which means the
675 * function was called with invalid args, which shouldn't be possible
676 * since the caller here is under our control. Other non-standard
677 * errors probably fall into the same category, or otherwise are fatal
678 * which means we can't carry on anyway.
681 if (status == -DLM_EUNLOCK || status == -DLM_ECANCEL)
682 lksb->lksb_conn->cc_proto->lp_unlock_ast(lksb, 0);
683 else
684 lksb->lksb_conn->cc_proto->lp_lock_ast(lksb);
687 static void fsdlm_blocking_ast_wrapper(void *astarg, int level)
689 struct ocfs2_dlm_lksb *lksb = astarg;
691 lksb->lksb_conn->cc_proto->lp_blocking_ast(lksb, level);
694 static int user_dlm_lock(struct ocfs2_cluster_connection *conn,
695 int mode,
696 struct ocfs2_dlm_lksb *lksb,
697 u32 flags,
698 void *name,
699 unsigned int namelen)
701 int ret;
703 if (!lksb->lksb_fsdlm.sb_lvbptr)
704 lksb->lksb_fsdlm.sb_lvbptr = (char *)lksb +
705 sizeof(struct dlm_lksb);
707 ret = dlm_lock(conn->cc_lockspace, mode, &lksb->lksb_fsdlm,
708 flags|DLM_LKF_NODLCKWT, name, namelen, 0,
709 fsdlm_lock_ast_wrapper, lksb,
710 fsdlm_blocking_ast_wrapper);
711 return ret;
714 static int user_dlm_unlock(struct ocfs2_cluster_connection *conn,
715 struct ocfs2_dlm_lksb *lksb,
716 u32 flags)
718 int ret;
720 ret = dlm_unlock(conn->cc_lockspace, lksb->lksb_fsdlm.sb_lkid,
721 flags, &lksb->lksb_fsdlm, lksb);
722 return ret;
725 static int user_dlm_lock_status(struct ocfs2_dlm_lksb *lksb)
727 return lksb->lksb_fsdlm.sb_status;
730 static int user_dlm_lvb_valid(struct ocfs2_dlm_lksb *lksb)
732 int invalid = lksb->lksb_fsdlm.sb_flags & DLM_SBF_VALNOTVALID;
734 return !invalid;
737 static void *user_dlm_lvb(struct ocfs2_dlm_lksb *lksb)
739 if (!lksb->lksb_fsdlm.sb_lvbptr)
740 lksb->lksb_fsdlm.sb_lvbptr = (char *)lksb +
741 sizeof(struct dlm_lksb);
742 return (void *)(lksb->lksb_fsdlm.sb_lvbptr);
745 static void user_dlm_dump_lksb(struct ocfs2_dlm_lksb *lksb)
749 static int user_plock(struct ocfs2_cluster_connection *conn,
750 u64 ino,
751 struct file *file,
752 int cmd,
753 struct file_lock *fl)
756 * This more or less just demuxes the plock request into any
757 * one of three dlm calls.
759 * Internally, fs/dlm will pass these to a misc device, which
760 * a userspace daemon will read and write to.
762 * For now, cancel requests (which happen internally only),
763 * are turned into unlocks. Most of this function taken from
764 * gfs2_lock.
767 if (cmd == F_CANCELLK) {
768 cmd = F_SETLK;
769 fl->fl_type = F_UNLCK;
772 if (IS_GETLK(cmd))
773 return dlm_posix_get(conn->cc_lockspace, ino, file, fl);
774 else if (fl->fl_type == F_UNLCK)
775 return dlm_posix_unlock(conn->cc_lockspace, ino, file, fl);
776 else
777 return dlm_posix_lock(conn->cc_lockspace, ino, file, cmd, fl);
781 * Compare a requested locking protocol version against the current one.
783 * If the major numbers are different, they are incompatible.
784 * If the current minor is greater than the request, they are incompatible.
785 * If the current minor is less than or equal to the request, they are
786 * compatible, and the requester should run at the current minor version.
788 static int fs_protocol_compare(struct ocfs2_protocol_version *existing,
789 struct ocfs2_protocol_version *request)
791 if (existing->pv_major != request->pv_major)
792 return 1;
794 if (existing->pv_minor > request->pv_minor)
795 return 1;
797 if (existing->pv_minor < request->pv_minor)
798 request->pv_minor = existing->pv_minor;
800 return 0;
803 static int user_cluster_connect(struct ocfs2_cluster_connection *conn)
805 dlm_lockspace_t *fsdlm;
806 struct ocfs2_live_connection *uninitialized_var(control);
807 int rc = 0;
809 BUG_ON(conn == NULL);
811 rc = ocfs2_live_connection_new(conn, &control);
812 if (rc)
813 goto out;
816 * running_proto must have been set before we allowed any mounts
817 * to proceed.
819 if (fs_protocol_compare(&running_proto, &conn->cc_version)) {
820 printk(KERN_ERR
821 "Unable to mount with fs locking protocol version "
822 "%u.%u because the userspace control daemon has "
823 "negotiated %u.%u\n",
824 conn->cc_version.pv_major, conn->cc_version.pv_minor,
825 running_proto.pv_major, running_proto.pv_minor);
826 rc = -EPROTO;
827 ocfs2_live_connection_drop(control);
828 goto out;
831 rc = dlm_new_lockspace(conn->cc_name, strlen(conn->cc_name),
832 &fsdlm, DLM_LSFL_FS, DLM_LVB_LEN);
833 if (rc) {
834 ocfs2_live_connection_drop(control);
835 goto out;
838 conn->cc_private = control;
839 conn->cc_lockspace = fsdlm;
840 out:
841 return rc;
844 static int user_cluster_disconnect(struct ocfs2_cluster_connection *conn)
846 dlm_release_lockspace(conn->cc_lockspace, 2);
847 conn->cc_lockspace = NULL;
848 ocfs2_live_connection_drop(conn->cc_private);
849 conn->cc_private = NULL;
850 return 0;
853 static int user_cluster_this_node(unsigned int *this_node)
855 int rc;
857 rc = ocfs2_control_get_this_node();
858 if (rc < 0)
859 return rc;
861 *this_node = rc;
862 return 0;
865 static struct ocfs2_stack_operations ocfs2_user_plugin_ops = {
866 .connect = user_cluster_connect,
867 .disconnect = user_cluster_disconnect,
868 .this_node = user_cluster_this_node,
869 .dlm_lock = user_dlm_lock,
870 .dlm_unlock = user_dlm_unlock,
871 .lock_status = user_dlm_lock_status,
872 .lvb_valid = user_dlm_lvb_valid,
873 .lock_lvb = user_dlm_lvb,
874 .plock = user_plock,
875 .dump_lksb = user_dlm_dump_lksb,
878 static struct ocfs2_stack_plugin ocfs2_user_plugin = {
879 .sp_name = "user",
880 .sp_ops = &ocfs2_user_plugin_ops,
881 .sp_owner = THIS_MODULE,
885 static int __init ocfs2_user_plugin_init(void)
887 int rc;
889 rc = ocfs2_control_init();
890 if (!rc) {
891 rc = ocfs2_stack_glue_register(&ocfs2_user_plugin);
892 if (rc)
893 ocfs2_control_exit();
896 return rc;
899 static void __exit ocfs2_user_plugin_exit(void)
901 ocfs2_stack_glue_unregister(&ocfs2_user_plugin);
902 ocfs2_control_exit();
905 MODULE_AUTHOR("Oracle");
906 MODULE_DESCRIPTION("ocfs2 driver for userspace cluster stacks");
907 MODULE_LICENSE("GPL");
908 module_init(ocfs2_user_plugin_init);
909 module_exit(ocfs2_user_plugin_exit);