Commit | Line | Data |
---|---|---|
8adf0536 JB |
1 | /* -*- 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> | |
6427a727 JB |
21 | #include <linux/fs.h> |
22 | #include <linux/miscdevice.h> | |
23 | #include <linux/mutex.h> | |
24 | #include <linux/reboot.h> | |
462c7e6a | 25 | #include <asm/uaccess.h> |
8adf0536 JB |
26 | |
27 | #include "stackglue.h" | |
28 | ||
29 | ||
6427a727 JB |
30 | /* |
31 | * The control protocol starts with a handshake. Until the handshake | |
32 | * is complete, the control device will fail all write(2)s. | |
33 | * | |
34 | * The handshake is simple. First, the client reads until EOF. Each line | |
35 | * of output is a supported protocol tag. All protocol tags are a single | |
36 | * character followed by a two hex digit version number. Currently the | |
37 | * only things supported is T01, for "Text-base version 0x01". Next, the | |
de870ef0 JB |
38 | * client writes the version they would like to use, including the newline. |
39 | * Thus, the protocol tag is 'T01\n'. If the version tag written is | |
40 | * unknown, -EINVAL is returned. Once the negotiation is complete, the | |
41 | * client can start sending messages. | |
42 | * | |
d4b95eef | 43 | * The T01 protocol has three messages. First is the "SETN" message. |
3cfd4ab6 JB |
44 | * It has the following syntax: |
45 | * | |
46 | * SETN<space><8-char-hex-nodenum><newline> | |
47 | * | |
48 | * This is 14 characters. | |
49 | * | |
50 | * The "SETN" message must be the first message following the protocol. | |
51 | * It tells ocfs2_control the local node number. | |
52 | * | |
d4b95eef JB |
53 | * Next comes the "SETV" message. It has the following syntax: |
54 | * | |
55 | * SETV<space><2-char-hex-major><space><2-char-hex-minor><newline> | |
56 | * | |
57 | * This is 11 characters. | |
58 | * | |
59 | * The "SETV" message sets the filesystem locking protocol version as | |
60 | * negotiated by the client. The client negotiates based on the maximum | |
61 | * version advertised in /sys/fs/ocfs2/max_locking_protocol. The major | |
62 | * number from the "SETV" message must match | |
63 | * user_stack.sp_proto->lp_max_version.pv_major, and the minor number | |
64 | * must be less than or equal to ...->lp_max_version.pv_minor. | |
65 | * | |
66 | * Once this information has been set, mounts will be allowed. From this | |
67 | * point on, the "DOWN" message can be sent for node down notification. | |
68 | * It has the following syntax: | |
de870ef0 JB |
69 | * |
70 | * DOWN<space><32-char-cap-hex-uuid><space><8-char-hex-nodenum><newline> | |
71 | * | |
72 | * eg: | |
73 | * | |
74 | * DOWN 632A924FDD844190BDA93C0DF6B94899 00000001\n | |
75 | * | |
76 | * This is 47 characters. | |
6427a727 JB |
77 | */ |
78 | ||
462c7e6a JB |
79 | /* |
80 | * Whether or not the client has done the handshake. | |
81 | * For now, we have just one protocol version. | |
82 | */ | |
83 | #define OCFS2_CONTROL_PROTO "T01\n" | |
84 | #define OCFS2_CONTROL_PROTO_LEN 4 | |
3cfd4ab6 JB |
85 | |
86 | /* Handshake states */ | |
462c7e6a JB |
87 | #define OCFS2_CONTROL_HANDSHAKE_INVALID (0) |
88 | #define OCFS2_CONTROL_HANDSHAKE_READ (1) | |
3cfd4ab6 JB |
89 | #define OCFS2_CONTROL_HANDSHAKE_PROTOCOL (2) |
90 | #define OCFS2_CONTROL_HANDSHAKE_VALID (3) | |
91 | ||
92 | /* Messages */ | |
93 | #define OCFS2_CONTROL_MESSAGE_OP_LEN 4 | |
94 | #define OCFS2_CONTROL_MESSAGE_SETNODE_OP "SETN" | |
95 | #define OCFS2_CONTROL_MESSAGE_SETNODE_TOTAL_LEN 14 | |
d4b95eef JB |
96 | #define OCFS2_CONTROL_MESSAGE_SETVERSION_OP "SETV" |
97 | #define OCFS2_CONTROL_MESSAGE_SETVERSION_TOTAL_LEN 11 | |
3cfd4ab6 | 98 | #define OCFS2_CONTROL_MESSAGE_DOWN_OP "DOWN" |
de870ef0 JB |
99 | #define OCFS2_CONTROL_MESSAGE_DOWN_TOTAL_LEN 47 |
100 | #define OCFS2_TEXT_UUID_LEN 32 | |
d4b95eef | 101 | #define OCFS2_CONTROL_MESSAGE_VERNUM_LEN 2 |
de870ef0 | 102 | #define OCFS2_CONTROL_MESSAGE_NODENUM_LEN 8 |
462c7e6a | 103 | |
6427a727 JB |
104 | /* |
105 | * ocfs2_live_connection is refcounted because the filesystem and | |
106 | * miscdevice sides can detach in different order. Let's just be safe. | |
107 | */ | |
108 | struct ocfs2_live_connection { | |
109 | struct list_head oc_list; | |
110 | struct ocfs2_cluster_connection *oc_conn; | |
111 | }; | |
112 | ||
462c7e6a JB |
113 | struct ocfs2_control_private { |
114 | struct list_head op_list; | |
115 | int op_state; | |
3cfd4ab6 | 116 | int op_this_node; |
d4b95eef | 117 | struct ocfs2_protocol_version op_proto; |
3cfd4ab6 JB |
118 | }; |
119 | ||
120 | /* SETN<space><8-char-hex-nodenum><newline> */ | |
121 | struct ocfs2_control_message_setn { | |
122 | char tag[OCFS2_CONTROL_MESSAGE_OP_LEN]; | |
123 | char space; | |
124 | char nodestr[OCFS2_CONTROL_MESSAGE_NODENUM_LEN]; | |
125 | char newline; | |
126 | }; | |
127 | ||
d4b95eef JB |
128 | /* SETV<space><2-char-hex-major><space><2-char-hex-minor><newline> */ |
129 | struct ocfs2_control_message_setv { | |
130 | char tag[OCFS2_CONTROL_MESSAGE_OP_LEN]; | |
131 | char space1; | |
132 | char major[OCFS2_CONTROL_MESSAGE_VERNUM_LEN]; | |
133 | char space2; | |
134 | char minor[OCFS2_CONTROL_MESSAGE_VERNUM_LEN]; | |
135 | char newline; | |
136 | }; | |
137 | ||
3cfd4ab6 JB |
138 | /* DOWN<space><32-char-cap-hex-uuid><space><8-char-hex-nodenum><newline> */ |
139 | struct ocfs2_control_message_down { | |
140 | char tag[OCFS2_CONTROL_MESSAGE_OP_LEN]; | |
141 | char space1; | |
142 | char uuid[OCFS2_TEXT_UUID_LEN]; | |
143 | char space2; | |
144 | char nodestr[OCFS2_CONTROL_MESSAGE_NODENUM_LEN]; | |
145 | char newline; | |
146 | }; | |
147 | ||
148 | union ocfs2_control_message { | |
149 | char tag[OCFS2_CONTROL_MESSAGE_OP_LEN]; | |
150 | struct ocfs2_control_message_setn u_setn; | |
d4b95eef | 151 | struct ocfs2_control_message_setv u_setv; |
3cfd4ab6 | 152 | struct ocfs2_control_message_down u_down; |
462c7e6a JB |
153 | }; |
154 | ||
6427a727 | 155 | static atomic_t ocfs2_control_opened; |
3cfd4ab6 | 156 | static int ocfs2_control_this_node = -1; |
d4b95eef | 157 | static struct ocfs2_protocol_version running_proto; |
6427a727 JB |
158 | |
159 | static LIST_HEAD(ocfs2_live_connection_list); | |
462c7e6a | 160 | static LIST_HEAD(ocfs2_control_private_list); |
6427a727 JB |
161 | static DEFINE_MUTEX(ocfs2_control_lock); |
162 | ||
462c7e6a JB |
163 | static inline void ocfs2_control_set_handshake_state(struct file *file, |
164 | int state) | |
165 | { | |
166 | struct ocfs2_control_private *p = file->private_data; | |
167 | p->op_state = state; | |
168 | } | |
169 | ||
170 | static inline int ocfs2_control_get_handshake_state(struct file *file) | |
171 | { | |
172 | struct ocfs2_control_private *p = file->private_data; | |
173 | return p->op_state; | |
174 | } | |
175 | ||
6427a727 JB |
176 | static struct ocfs2_live_connection *ocfs2_connection_find(const char *name) |
177 | { | |
178 | size_t len = strlen(name); | |
179 | struct ocfs2_live_connection *c; | |
180 | ||
181 | BUG_ON(!mutex_is_locked(&ocfs2_control_lock)); | |
182 | ||
183 | list_for_each_entry(c, &ocfs2_live_connection_list, oc_list) { | |
184 | if ((c->oc_conn->cc_namelen == len) && | |
185 | !strncmp(c->oc_conn->cc_name, name, len)) | |
186 | return c; | |
187 | } | |
188 | ||
189 | return c; | |
190 | } | |
191 | ||
192 | /* | |
193 | * ocfs2_live_connection structures are created underneath the ocfs2 | |
194 | * mount path. Since the VFS prevents multiple calls to | |
195 | * fill_super(), we can't get dupes here. | |
196 | */ | |
197 | static int ocfs2_live_connection_new(struct ocfs2_cluster_connection *conn, | |
198 | struct ocfs2_live_connection **c_ret) | |
199 | { | |
200 | int rc = 0; | |
201 | struct ocfs2_live_connection *c; | |
202 | ||
203 | c = kzalloc(sizeof(struct ocfs2_live_connection), GFP_KERNEL); | |
204 | if (!c) | |
205 | return -ENOMEM; | |
206 | ||
207 | mutex_lock(&ocfs2_control_lock); | |
208 | c->oc_conn = conn; | |
209 | ||
210 | if (atomic_read(&ocfs2_control_opened)) | |
211 | list_add(&c->oc_list, &ocfs2_live_connection_list); | |
212 | else { | |
213 | printk(KERN_ERR | |
214 | "ocfs2: Userspace control daemon is not present\n"); | |
215 | rc = -ESRCH; | |
216 | } | |
217 | ||
218 | mutex_unlock(&ocfs2_control_lock); | |
219 | ||
220 | if (!rc) | |
221 | *c_ret = c; | |
222 | else | |
223 | kfree(c); | |
224 | ||
225 | return rc; | |
226 | } | |
227 | ||
228 | /* | |
229 | * This function disconnects the cluster connection from ocfs2_control. | |
230 | * Afterwards, userspace can't affect the cluster connection. | |
231 | */ | |
232 | static void ocfs2_live_connection_drop(struct ocfs2_live_connection *c) | |
233 | { | |
234 | mutex_lock(&ocfs2_control_lock); | |
235 | list_del_init(&c->oc_list); | |
236 | c->oc_conn = NULL; | |
237 | mutex_unlock(&ocfs2_control_lock); | |
238 | ||
239 | kfree(c); | |
240 | } | |
241 | ||
3cfd4ab6 JB |
242 | static int ocfs2_control_cfu(void *target, size_t target_len, |
243 | const char __user *buf, size_t count) | |
462c7e6a JB |
244 | { |
245 | /* The T01 expects write(2) calls to have exactly one command */ | |
3cfd4ab6 JB |
246 | if ((count != target_len) || |
247 | (count > sizeof(union ocfs2_control_message))) | |
462c7e6a JB |
248 | return -EINVAL; |
249 | ||
250 | if (copy_from_user(target, buf, target_len)) | |
251 | return -EFAULT; | |
252 | ||
3cfd4ab6 | 253 | return 0; |
462c7e6a JB |
254 | } |
255 | ||
3cfd4ab6 JB |
256 | static ssize_t ocfs2_control_validate_protocol(struct file *file, |
257 | const char __user *buf, | |
258 | size_t count) | |
462c7e6a JB |
259 | { |
260 | ssize_t ret; | |
261 | char kbuf[OCFS2_CONTROL_PROTO_LEN]; | |
262 | ||
263 | ret = ocfs2_control_cfu(kbuf, OCFS2_CONTROL_PROTO_LEN, | |
264 | buf, count); | |
3cfd4ab6 | 265 | if (ret) |
462c7e6a JB |
266 | return ret; |
267 | ||
268 | if (strncmp(kbuf, OCFS2_CONTROL_PROTO, OCFS2_CONTROL_PROTO_LEN)) | |
269 | return -EINVAL; | |
270 | ||
462c7e6a | 271 | ocfs2_control_set_handshake_state(file, |
3cfd4ab6 | 272 | OCFS2_CONTROL_HANDSHAKE_PROTOCOL); |
462c7e6a JB |
273 | |
274 | return count; | |
275 | } | |
276 | ||
de870ef0 JB |
277 | static void ocfs2_control_send_down(const char *uuid, |
278 | int nodenum) | |
279 | { | |
280 | struct ocfs2_live_connection *c; | |
281 | ||
282 | mutex_lock(&ocfs2_control_lock); | |
283 | ||
284 | c = ocfs2_connection_find(uuid); | |
285 | if (c) { | |
286 | BUG_ON(c->oc_conn == NULL); | |
287 | c->oc_conn->cc_recovery_handler(nodenum, | |
288 | c->oc_conn->cc_recovery_data); | |
289 | } | |
290 | ||
291 | mutex_unlock(&ocfs2_control_lock); | |
292 | } | |
293 | ||
3cfd4ab6 JB |
294 | /* |
295 | * Called whenever configuration elements are sent to /dev/ocfs2_control. | |
296 | * If all configuration elements are present, try to set the global | |
d4b95eef JB |
297 | * values. If there is a problem, return an error. Skip any missing |
298 | * elements, and only bump ocfs2_control_opened when we have all elements | |
299 | * and are successful. | |
3cfd4ab6 JB |
300 | */ |
301 | static int ocfs2_control_install_private(struct file *file) | |
302 | { | |
303 | int rc = 0; | |
304 | int set_p = 1; | |
305 | struct ocfs2_control_private *p = file->private_data; | |
de870ef0 | 306 | |
3cfd4ab6 JB |
307 | BUG_ON(p->op_state != OCFS2_CONTROL_HANDSHAKE_PROTOCOL); |
308 | ||
d4b95eef JB |
309 | mutex_lock(&ocfs2_control_lock); |
310 | ||
311 | if (p->op_this_node < 0) { | |
3cfd4ab6 | 312 | set_p = 0; |
d4b95eef JB |
313 | } else if ((ocfs2_control_this_node >= 0) && |
314 | (ocfs2_control_this_node != p->op_this_node)) { | |
315 | rc = -EINVAL; | |
316 | goto out_unlock; | |
317 | } | |
3cfd4ab6 | 318 | |
d4b95eef JB |
319 | if (!p->op_proto.pv_major) { |
320 | set_p = 0; | |
321 | } else if (!list_empty(&ocfs2_live_connection_list) && | |
322 | ((running_proto.pv_major != p->op_proto.pv_major) || | |
323 | (running_proto.pv_minor != p->op_proto.pv_minor))) { | |
3cfd4ab6 | 324 | rc = -EINVAL; |
d4b95eef JB |
325 | goto out_unlock; |
326 | } | |
327 | ||
328 | if (set_p) { | |
329 | ocfs2_control_this_node = p->op_this_node; | |
330 | running_proto.pv_major = p->op_proto.pv_major; | |
331 | running_proto.pv_minor = p->op_proto.pv_minor; | |
332 | } | |
333 | ||
334 | out_unlock: | |
3cfd4ab6 JB |
335 | mutex_unlock(&ocfs2_control_lock); |
336 | ||
337 | if (!rc && set_p) { | |
338 | /* We set the global values successfully */ | |
339 | atomic_inc(&ocfs2_control_opened); | |
340 | ocfs2_control_set_handshake_state(file, | |
341 | OCFS2_CONTROL_HANDSHAKE_VALID); | |
342 | } | |
343 | ||
344 | return rc; | |
345 | } | |
346 | ||
347 | static int ocfs2_control_do_setnode_msg(struct file *file, | |
348 | struct ocfs2_control_message_setn *msg) | |
de870ef0 | 349 | { |
de870ef0 | 350 | long nodenum; |
3cfd4ab6 JB |
351 | char *ptr = NULL; |
352 | struct ocfs2_control_private *p = file->private_data; | |
de870ef0 | 353 | |
3cfd4ab6 JB |
354 | if (ocfs2_control_get_handshake_state(file) != |
355 | OCFS2_CONTROL_HANDSHAKE_PROTOCOL) | |
356 | return -EINVAL; | |
de870ef0 | 357 | |
3cfd4ab6 JB |
358 | if (strncmp(msg->tag, OCFS2_CONTROL_MESSAGE_SETNODE_OP, |
359 | OCFS2_CONTROL_MESSAGE_OP_LEN)) | |
360 | return -EINVAL; | |
361 | ||
362 | if ((msg->space != ' ') || (msg->newline != '\n')) | |
363 | return -EINVAL; | |
364 | msg->space = msg->newline = '\0'; | |
365 | ||
366 | nodenum = simple_strtol(msg->nodestr, &ptr, 16); | |
367 | if (!ptr || *ptr) | |
368 | return -EINVAL; | |
369 | ||
370 | if ((nodenum == LONG_MIN) || (nodenum == LONG_MAX) || | |
371 | (nodenum > INT_MAX) || (nodenum < 0)) | |
372 | return -ERANGE; | |
373 | p->op_this_node = nodenum; | |
374 | ||
375 | return ocfs2_control_install_private(file); | |
376 | } | |
377 | ||
d4b95eef JB |
378 | static int ocfs2_control_do_setversion_msg(struct file *file, |
379 | struct ocfs2_control_message_setv *msg) | |
380 | { | |
381 | long major, minor; | |
382 | char *ptr = NULL; | |
383 | struct ocfs2_control_private *p = file->private_data; | |
384 | struct ocfs2_protocol_version *max = | |
385 | &user_stack.sp_proto->lp_max_version; | |
386 | ||
387 | if (ocfs2_control_get_handshake_state(file) != | |
388 | OCFS2_CONTROL_HANDSHAKE_PROTOCOL) | |
389 | return -EINVAL; | |
390 | ||
391 | if (strncmp(msg->tag, OCFS2_CONTROL_MESSAGE_SETVERSION_OP, | |
392 | OCFS2_CONTROL_MESSAGE_OP_LEN)) | |
393 | return -EINVAL; | |
394 | ||
395 | if ((msg->space1 != ' ') || (msg->space2 != ' ') || | |
396 | (msg->newline != '\n')) | |
397 | return -EINVAL; | |
398 | msg->space1 = msg->space2 = msg->newline = '\0'; | |
399 | ||
400 | major = simple_strtol(msg->major, &ptr, 16); | |
401 | if (!ptr || *ptr) | |
402 | return -EINVAL; | |
403 | minor = simple_strtol(msg->minor, &ptr, 16); | |
404 | if (!ptr || *ptr) | |
405 | return -EINVAL; | |
406 | ||
407 | /* | |
408 | * The major must be between 1 and 255, inclusive. The minor | |
409 | * must be between 0 and 255, inclusive. The version passed in | |
410 | * must be within the maximum version supported by the filesystem. | |
411 | */ | |
412 | if ((major == LONG_MIN) || (major == LONG_MAX) || | |
413 | (major > (u8)-1) || (major < 1)) | |
414 | return -ERANGE; | |
415 | if ((minor == LONG_MIN) || (minor == LONG_MAX) || | |
416 | (minor > (u8)-1) || (minor < 0)) | |
417 | return -ERANGE; | |
418 | if ((major != max->pv_major) || | |
419 | (minor > max->pv_minor)) | |
420 | return -EINVAL; | |
421 | ||
422 | p->op_proto.pv_major = major; | |
423 | p->op_proto.pv_minor = minor; | |
424 | ||
425 | return ocfs2_control_install_private(file); | |
426 | } | |
427 | ||
3cfd4ab6 JB |
428 | static int ocfs2_control_do_down_msg(struct file *file, |
429 | struct ocfs2_control_message_down *msg) | |
430 | { | |
431 | long nodenum; | |
432 | char *p = NULL; | |
433 | ||
434 | if (ocfs2_control_get_handshake_state(file) != | |
435 | OCFS2_CONTROL_HANDSHAKE_VALID) | |
436 | return -EINVAL; | |
de870ef0 | 437 | |
3cfd4ab6 JB |
438 | if (strncmp(msg->tag, OCFS2_CONTROL_MESSAGE_DOWN_OP, |
439 | OCFS2_CONTROL_MESSAGE_OP_LEN)) | |
de870ef0 JB |
440 | return -EINVAL; |
441 | ||
3cfd4ab6 JB |
442 | if ((msg->space1 != ' ') || (msg->space2 != ' ') || |
443 | (msg->newline != '\n')) | |
de870ef0 | 444 | return -EINVAL; |
3cfd4ab6 | 445 | msg->space1 = msg->space2 = msg->newline = '\0'; |
de870ef0 | 446 | |
3cfd4ab6 | 447 | nodenum = simple_strtol(msg->nodestr, &p, 16); |
de870ef0 JB |
448 | if (!p || *p) |
449 | return -EINVAL; | |
450 | ||
451 | if ((nodenum == LONG_MIN) || (nodenum == LONG_MAX) || | |
452 | (nodenum > INT_MAX) || (nodenum < 0)) | |
453 | return -ERANGE; | |
454 | ||
3cfd4ab6 | 455 | ocfs2_control_send_down(msg->uuid, nodenum); |
de870ef0 | 456 | |
3cfd4ab6 JB |
457 | return 0; |
458 | } | |
459 | ||
460 | static ssize_t ocfs2_control_message(struct file *file, | |
461 | const char __user *buf, | |
462 | size_t count) | |
463 | { | |
464 | ssize_t ret; | |
465 | union ocfs2_control_message msg; | |
466 | ||
467 | /* Try to catch padding issues */ | |
468 | WARN_ON(offsetof(struct ocfs2_control_message_down, uuid) != | |
469 | (sizeof(msg.u_down.tag) + sizeof(msg.u_down.space1))); | |
470 | ||
471 | memset(&msg, 0, sizeof(union ocfs2_control_message)); | |
472 | ret = ocfs2_control_cfu(&msg, count, buf, count); | |
473 | if (ret) | |
474 | goto out; | |
475 | ||
476 | if ((count == OCFS2_CONTROL_MESSAGE_SETNODE_TOTAL_LEN) && | |
477 | !strncmp(msg.tag, OCFS2_CONTROL_MESSAGE_SETNODE_OP, | |
478 | OCFS2_CONTROL_MESSAGE_OP_LEN)) | |
479 | ret = ocfs2_control_do_setnode_msg(file, &msg.u_setn); | |
d4b95eef JB |
480 | else if ((count == OCFS2_CONTROL_MESSAGE_SETVERSION_TOTAL_LEN) && |
481 | !strncmp(msg.tag, OCFS2_CONTROL_MESSAGE_SETVERSION_OP, | |
482 | OCFS2_CONTROL_MESSAGE_OP_LEN)) | |
483 | ret = ocfs2_control_do_setversion_msg(file, &msg.u_setv); | |
3cfd4ab6 JB |
484 | else if ((count == OCFS2_CONTROL_MESSAGE_DOWN_TOTAL_LEN) && |
485 | !strncmp(msg.tag, OCFS2_CONTROL_MESSAGE_DOWN_OP, | |
486 | OCFS2_CONTROL_MESSAGE_OP_LEN)) | |
487 | ret = ocfs2_control_do_down_msg(file, &msg.u_down); | |
488 | else | |
489 | ret = -EINVAL; | |
490 | ||
491 | out: | |
492 | return ret ? ret : count; | |
de870ef0 | 493 | } |
6427a727 JB |
494 | |
495 | static ssize_t ocfs2_control_write(struct file *file, | |
496 | const char __user *buf, | |
497 | size_t count, | |
498 | loff_t *ppos) | |
8adf0536 | 499 | { |
462c7e6a JB |
500 | ssize_t ret; |
501 | ||
502 | switch (ocfs2_control_get_handshake_state(file)) { | |
503 | case OCFS2_CONTROL_HANDSHAKE_INVALID: | |
504 | ret = -EINVAL; | |
505 | break; | |
506 | ||
507 | case OCFS2_CONTROL_HANDSHAKE_READ: | |
3cfd4ab6 JB |
508 | ret = ocfs2_control_validate_protocol(file, buf, |
509 | count); | |
462c7e6a JB |
510 | break; |
511 | ||
3cfd4ab6 | 512 | case OCFS2_CONTROL_HANDSHAKE_PROTOCOL: |
462c7e6a | 513 | case OCFS2_CONTROL_HANDSHAKE_VALID: |
de870ef0 | 514 | ret = ocfs2_control_message(file, buf, count); |
462c7e6a JB |
515 | break; |
516 | ||
517 | default: | |
518 | BUG(); | |
519 | ret = -EIO; | |
520 | break; | |
521 | } | |
522 | ||
523 | return ret; | |
8adf0536 JB |
524 | } |
525 | ||
462c7e6a JB |
526 | /* |
527 | * This is a naive version. If we ever have a new protocol, we'll expand | |
528 | * it. Probably using seq_file. | |
529 | */ | |
6427a727 JB |
530 | static ssize_t ocfs2_control_read(struct file *file, |
531 | char __user *buf, | |
532 | size_t count, | |
533 | loff_t *ppos) | |
534 | { | |
462c7e6a JB |
535 | char *proto_string = OCFS2_CONTROL_PROTO; |
536 | size_t to_write = 0; | |
537 | ||
538 | if (*ppos >= OCFS2_CONTROL_PROTO_LEN) | |
539 | return 0; | |
540 | ||
541 | to_write = OCFS2_CONTROL_PROTO_LEN - *ppos; | |
542 | if (to_write > count) | |
543 | to_write = count; | |
544 | if (copy_to_user(buf, proto_string + *ppos, to_write)) | |
545 | return -EFAULT; | |
546 | ||
547 | *ppos += to_write; | |
548 | ||
549 | /* Have we read the whole protocol list? */ | |
550 | if (*ppos >= OCFS2_CONTROL_PROTO_LEN) | |
551 | ocfs2_control_set_handshake_state(file, | |
552 | OCFS2_CONTROL_HANDSHAKE_READ); | |
553 | ||
554 | return to_write; | |
6427a727 JB |
555 | } |
556 | ||
557 | static int ocfs2_control_release(struct inode *inode, struct file *file) | |
558 | { | |
462c7e6a JB |
559 | struct ocfs2_control_private *p = file->private_data; |
560 | ||
561 | mutex_lock(&ocfs2_control_lock); | |
562 | ||
563 | if (ocfs2_control_get_handshake_state(file) != | |
564 | OCFS2_CONTROL_HANDSHAKE_VALID) | |
565 | goto out; | |
566 | ||
6427a727 | 567 | if (atomic_dec_and_test(&ocfs2_control_opened)) { |
6427a727 JB |
568 | if (!list_empty(&ocfs2_live_connection_list)) { |
569 | /* XXX: Do bad things! */ | |
570 | printk(KERN_ERR | |
571 | "ocfs2: Unexpected release of ocfs2_control!\n" | |
572 | " Loss of cluster connection requires " | |
573 | "an emergency restart!\n"); | |
574 | emergency_restart(); | |
575 | } | |
d4b95eef JB |
576 | /* |
577 | * Last valid close clears the node number and resets | |
578 | * the locking protocol version | |
579 | */ | |
3cfd4ab6 | 580 | ocfs2_control_this_node = -1; |
d4b95eef JB |
581 | running_proto.pv_major = 0; |
582 | running_proto.pv_major = 0; | |
6427a727 JB |
583 | } |
584 | ||
462c7e6a JB |
585 | out: |
586 | list_del_init(&p->op_list); | |
587 | file->private_data = NULL; | |
588 | ||
589 | mutex_unlock(&ocfs2_control_lock); | |
590 | ||
591 | kfree(p); | |
592 | ||
6427a727 JB |
593 | return 0; |
594 | } | |
595 | ||
596 | static int ocfs2_control_open(struct inode *inode, struct file *file) | |
597 | { | |
462c7e6a JB |
598 | struct ocfs2_control_private *p; |
599 | ||
600 | p = kzalloc(sizeof(struct ocfs2_control_private), GFP_KERNEL); | |
601 | if (!p) | |
602 | return -ENOMEM; | |
3cfd4ab6 | 603 | p->op_this_node = -1; |
462c7e6a JB |
604 | |
605 | mutex_lock(&ocfs2_control_lock); | |
606 | file->private_data = p; | |
607 | list_add(&p->op_list, &ocfs2_control_private_list); | |
608 | mutex_unlock(&ocfs2_control_lock); | |
6427a727 JB |
609 | |
610 | return 0; | |
611 | } | |
612 | ||
613 | static const struct file_operations ocfs2_control_fops = { | |
614 | .open = ocfs2_control_open, | |
615 | .release = ocfs2_control_release, | |
616 | .read = ocfs2_control_read, | |
617 | .write = ocfs2_control_write, | |
618 | .owner = THIS_MODULE, | |
619 | }; | |
620 | ||
621 | struct miscdevice ocfs2_control_device = { | |
622 | .minor = MISC_DYNAMIC_MINOR, | |
623 | .name = "ocfs2_control", | |
624 | .fops = &ocfs2_control_fops, | |
625 | }; | |
626 | ||
627 | static int ocfs2_control_init(void) | |
628 | { | |
629 | int rc; | |
630 | ||
631 | atomic_set(&ocfs2_control_opened, 0); | |
632 | ||
633 | rc = misc_register(&ocfs2_control_device); | |
634 | if (rc) | |
635 | printk(KERN_ERR | |
636 | "ocfs2: Unable to register ocfs2_control device " | |
637 | "(errno %d)\n", | |
638 | -rc); | |
639 | ||
640 | return rc; | |
641 | } | |
642 | ||
643 | static void ocfs2_control_exit(void) | |
644 | { | |
645 | int rc; | |
646 | ||
647 | rc = misc_deregister(&ocfs2_control_device); | |
648 | if (rc) | |
649 | printk(KERN_ERR | |
650 | "ocfs2: Unable to deregister ocfs2_control device " | |
651 | "(errno %d)\n", | |
652 | -rc); | |
653 | } | |
654 | ||
655 | static int __init user_stack_init(void) | |
656 | { | |
657 | return ocfs2_control_init(); | |
658 | } | |
659 | ||
8adf0536 JB |
660 | static void __exit user_stack_exit(void) |
661 | { | |
6427a727 | 662 | ocfs2_control_exit(); |
8adf0536 JB |
663 | } |
664 | ||
665 | MODULE_AUTHOR("Oracle"); | |
666 | MODULE_DESCRIPTION("ocfs2 driver for userspace cluster stacks"); | |
667 | MODULE_LICENSE("GPL"); | |
668 | module_init(user_stack_init); | |
669 | module_exit(user_stack_exit); |