]>
Commit | Line | Data |
---|---|---|
8df08c89 MF |
1 | /* -*- mode: c; c-basic-offset: 8; -*- |
2 | * vim: noexpandtab sw=8 ts=8 sts=0: | |
3 | * | |
4 | * dlmfs.c | |
5 | * | |
6 | * Code which implements the kernel side of a minimal userspace | |
7 | * interface to our DLM. This file handles the virtual file system | |
8 | * used for communication with userspace. Credit should go to ramfs, | |
9 | * which was a template for the fs side of this module. | |
10 | * | |
11 | * Copyright (C) 2003, 2004 Oracle. All rights reserved. | |
12 | * | |
13 | * This program is free software; you can redistribute it and/or | |
14 | * modify it under the terms of the GNU General Public | |
15 | * License as published by the Free Software Foundation; either | |
16 | * version 2 of the License, or (at your option) any later version. | |
17 | * | |
18 | * This program is distributed in the hope that it will be useful, | |
19 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
20 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
21 | * General Public License for more details. | |
22 | * | |
23 | * You should have received a copy of the GNU General Public | |
24 | * License along with this program; if not, write to the | |
25 | * Free Software Foundation, Inc., 59 Temple Place - Suite 330, | |
26 | * Boston, MA 021110-1307, USA. | |
27 | */ | |
28 | ||
29 | /* Simple VFS hooks based on: */ | |
30 | /* | |
31 | * Resizable simple ram filesystem for Linux. | |
32 | * | |
33 | * Copyright (C) 2000 Linus Torvalds. | |
34 | * 2000 Transmeta Corp. | |
35 | */ | |
36 | ||
37 | #include <linux/module.h> | |
38 | #include <linux/fs.h> | |
39 | #include <linux/pagemap.h> | |
40 | #include <linux/types.h> | |
41 | #include <linux/slab.h> | |
42 | #include <linux/highmem.h> | |
43 | #include <linux/init.h> | |
44 | #include <linux/string.h> | |
8df08c89 | 45 | #include <linux/backing-dev.h> |
65b6f340 | 46 | #include <linux/poll.h> |
8df08c89 MF |
47 | |
48 | #include <asm/uaccess.h> | |
49 | ||
50 | ||
51 | #include "cluster/nodemanager.h" | |
52 | #include "cluster/heartbeat.h" | |
53 | #include "cluster/tcp.h" | |
54 | ||
34a9dd7e | 55 | #include "dlm/dlmapi.h" |
8df08c89 MF |
56 | |
57 | #include "userdlm.h" | |
58 | ||
59 | #include "dlmfsver.h" | |
60 | ||
61 | #define MLOG_MASK_PREFIX ML_DLMFS | |
62 | #include "cluster/masklog.h" | |
63 | ||
d24fbcda JB |
64 | #include "ocfs2_lockingver.h" |
65 | ||
ee9b6d61 | 66 | static const struct super_operations dlmfs_ops; |
00977a59 | 67 | static const struct file_operations dlmfs_file_operations; |
92e1d5be AV |
68 | static const struct inode_operations dlmfs_dir_inode_operations; |
69 | static const struct inode_operations dlmfs_root_inode_operations; | |
70 | static const struct inode_operations dlmfs_file_inode_operations; | |
e18b890b | 71 | static struct kmem_cache *dlmfs_inode_cache; |
8df08c89 MF |
72 | |
73 | struct workqueue_struct *user_dlm_worker; | |
74 | ||
d24fbcda JB |
75 | /* |
76 | * This is the userdlmfs locking protocol version. | |
77 | * | |
78 | * See fs/ocfs2/dlmglue.c for more details on locking versions. | |
79 | */ | |
80 | static const struct dlm_protocol_version user_locking_protocol = { | |
81 | .pv_major = OCFS2_LOCKING_PROTOCOL_MAJOR, | |
82 | .pv_minor = OCFS2_LOCKING_PROTOCOL_MINOR, | |
83 | }; | |
84 | ||
14a437c2 JB |
85 | |
86 | /* | |
87 | * These are the ABI capabilities of dlmfs. | |
88 | * | |
89 | * Over time, dlmfs has added some features that were not part of the | |
90 | * initial ABI. Unfortunately, some of these features are not detectable | |
91 | * via standard usage. For example, Linux's default poll always returns | |
92 | * POLLIN, so there is no way for a caller of poll(2) to know when dlmfs | |
93 | * added poll support. Instead, we provide this list of new capabilities. | |
94 | * | |
95 | * Capabilities is a read-only attribute. We do it as a module parameter | |
96 | * so we can discover it whether dlmfs is built in, loaded, or even not | |
97 | * loaded. | |
98 | * | |
99 | * The ABI features are local to this machine's dlmfs mount. This is | |
100 | * distinct from the locking protocol, which is concerned with inter-node | |
101 | * interaction. | |
65b6f340 JB |
102 | * |
103 | * Capabilities: | |
104 | * - bast : POLLIN against the file descriptor of a held lock | |
105 | * signifies a bast fired on the lock. | |
14a437c2 | 106 | */ |
65b6f340 | 107 | #define DLMFS_CAPABILITIES "bast" |
14a437c2 JB |
108 | extern int param_set_dlmfs_capabilities(const char *val, |
109 | struct kernel_param *kp) | |
110 | { | |
111 | printk(KERN_ERR "%s: readonly parameter\n", kp->name); | |
112 | return -EINVAL; | |
113 | } | |
114 | static int param_get_dlmfs_capabilities(char *buffer, | |
115 | struct kernel_param *kp) | |
116 | { | |
117 | return strlcpy(buffer, DLMFS_CAPABILITIES, | |
118 | strlen(DLMFS_CAPABILITIES) + 1); | |
119 | } | |
120 | module_param_call(capabilities, param_set_dlmfs_capabilities, | |
121 | param_get_dlmfs_capabilities, NULL, 0444); | |
122 | MODULE_PARM_DESC(capabilities, DLMFS_CAPABILITIES); | |
123 | ||
124 | ||
8df08c89 MF |
125 | /* |
126 | * decodes a set of open flags into a valid lock level and a set of flags. | |
127 | * returns < 0 if we have invalid flags | |
128 | * flags which mean something to us: | |
129 | * O_RDONLY -> PRMODE level | |
130 | * O_WRONLY -> EXMODE level | |
131 | * | |
132 | * O_NONBLOCK -> LKM_NOQUEUE | |
133 | */ | |
134 | static int dlmfs_decode_open_flags(int open_flags, | |
135 | int *level, | |
136 | int *flags) | |
137 | { | |
138 | if (open_flags & (O_WRONLY|O_RDWR)) | |
139 | *level = LKM_EXMODE; | |
140 | else | |
141 | *level = LKM_PRMODE; | |
142 | ||
143 | *flags = 0; | |
144 | if (open_flags & O_NONBLOCK) | |
145 | *flags |= LKM_NOQUEUE; | |
146 | ||
147 | return 0; | |
148 | } | |
149 | ||
150 | static int dlmfs_file_open(struct inode *inode, | |
151 | struct file *file) | |
152 | { | |
153 | int status, level, flags; | |
154 | struct dlmfs_filp_private *fp = NULL; | |
155 | struct dlmfs_inode_private *ip; | |
156 | ||
157 | if (S_ISDIR(inode->i_mode)) | |
158 | BUG(); | |
159 | ||
160 | mlog(0, "open called on inode %lu, flags 0x%x\n", inode->i_ino, | |
161 | file->f_flags); | |
162 | ||
163 | status = dlmfs_decode_open_flags(file->f_flags, &level, &flags); | |
164 | if (status < 0) | |
165 | goto bail; | |
166 | ||
167 | /* We don't want to honor O_APPEND at read/write time as it | |
168 | * doesn't make sense for LVB writes. */ | |
169 | file->f_flags &= ~O_APPEND; | |
170 | ||
ad8100e0 | 171 | fp = kmalloc(sizeof(*fp), GFP_NOFS); |
8df08c89 MF |
172 | if (!fp) { |
173 | status = -ENOMEM; | |
174 | goto bail; | |
175 | } | |
176 | fp->fp_lock_level = level; | |
177 | ||
178 | ip = DLMFS_I(inode); | |
179 | ||
180 | status = user_dlm_cluster_lock(&ip->ip_lockres, level, flags); | |
181 | if (status < 0) { | |
182 | /* this is a strange error to return here but I want | |
183 | * to be able userspace to be able to distinguish a | |
184 | * valid lock request from one that simply couldn't be | |
185 | * granted. */ | |
186 | if (flags & LKM_NOQUEUE && status == -EAGAIN) | |
187 | status = -ETXTBSY; | |
188 | kfree(fp); | |
189 | goto bail; | |
190 | } | |
191 | ||
192 | file->private_data = fp; | |
193 | bail: | |
194 | return status; | |
195 | } | |
196 | ||
197 | static int dlmfs_file_release(struct inode *inode, | |
198 | struct file *file) | |
199 | { | |
200 | int level, status; | |
201 | struct dlmfs_inode_private *ip = DLMFS_I(inode); | |
202 | struct dlmfs_filp_private *fp = | |
203 | (struct dlmfs_filp_private *) file->private_data; | |
204 | ||
205 | if (S_ISDIR(inode->i_mode)) | |
206 | BUG(); | |
207 | ||
208 | mlog(0, "close called on inode %lu\n", inode->i_ino); | |
209 | ||
210 | status = 0; | |
211 | if (fp) { | |
212 | level = fp->fp_lock_level; | |
213 | if (level != LKM_IVMODE) | |
214 | user_dlm_cluster_unlock(&ip->ip_lockres, level); | |
215 | ||
216 | kfree(fp); | |
217 | file->private_data = NULL; | |
218 | } | |
219 | ||
220 | return 0; | |
221 | } | |
222 | ||
e8fce482 JB |
223 | /* |
224 | * We do ->setattr() just to override size changes. Our size is the size | |
225 | * of the LVB and nothing else. | |
226 | */ | |
227 | static int dlmfs_file_setattr(struct dentry *dentry, struct iattr *attr) | |
228 | { | |
229 | int error; | |
230 | struct inode *inode = dentry->d_inode; | |
231 | ||
232 | attr->ia_valid &= ~ATTR_SIZE; | |
233 | error = inode_change_ok(inode, attr); | |
234 | if (!error) | |
235 | error = inode_setattr(inode, attr); | |
236 | ||
237 | return error; | |
238 | } | |
239 | ||
65b6f340 JB |
240 | static unsigned int dlmfs_file_poll(struct file *file, poll_table *wait) |
241 | { | |
242 | int event = 0; | |
243 | struct inode *inode = file->f_path.dentry->d_inode; | |
244 | struct dlmfs_inode_private *ip = DLMFS_I(inode); | |
245 | ||
246 | poll_wait(file, &ip->ip_lockres.l_event, wait); | |
247 | ||
248 | spin_lock(&ip->ip_lockres.l_lock); | |
249 | if (ip->ip_lockres.l_flags & USER_LOCK_BLOCKED) | |
250 | event = POLLIN | POLLRDNORM; | |
251 | spin_unlock(&ip->ip_lockres.l_lock); | |
252 | ||
253 | return event; | |
254 | } | |
255 | ||
8df08c89 MF |
256 | static ssize_t dlmfs_file_read(struct file *filp, |
257 | char __user *buf, | |
258 | size_t count, | |
259 | loff_t *ppos) | |
260 | { | |
261 | int bytes_left; | |
262 | ssize_t readlen; | |
263 | char *lvb_buf; | |
d28c9174 | 264 | struct inode *inode = filp->f_path.dentry->d_inode; |
8df08c89 MF |
265 | |
266 | mlog(0, "inode %lu, count = %zu, *ppos = %llu\n", | |
267 | inode->i_ino, count, *ppos); | |
268 | ||
269 | if (*ppos >= i_size_read(inode)) | |
270 | return 0; | |
271 | ||
272 | if (!count) | |
273 | return 0; | |
274 | ||
275 | if (!access_ok(VERIFY_WRITE, buf, count)) | |
276 | return -EFAULT; | |
277 | ||
278 | /* don't read past the lvb */ | |
279 | if ((count + *ppos) > i_size_read(inode)) | |
280 | readlen = i_size_read(inode) - *ppos; | |
281 | else | |
282 | readlen = count - *ppos; | |
283 | ||
ad8100e0 | 284 | lvb_buf = kmalloc(readlen, GFP_NOFS); |
8df08c89 MF |
285 | if (!lvb_buf) |
286 | return -ENOMEM; | |
287 | ||
288 | user_dlm_read_lvb(inode, lvb_buf, readlen); | |
289 | bytes_left = __copy_to_user(buf, lvb_buf, readlen); | |
290 | readlen -= bytes_left; | |
291 | ||
292 | kfree(lvb_buf); | |
293 | ||
294 | *ppos = *ppos + readlen; | |
295 | ||
296 | mlog(0, "read %zd bytes\n", readlen); | |
297 | return readlen; | |
298 | } | |
299 | ||
300 | static ssize_t dlmfs_file_write(struct file *filp, | |
301 | const char __user *buf, | |
302 | size_t count, | |
303 | loff_t *ppos) | |
304 | { | |
305 | int bytes_left; | |
306 | ssize_t writelen; | |
307 | char *lvb_buf; | |
d28c9174 | 308 | struct inode *inode = filp->f_path.dentry->d_inode; |
8df08c89 MF |
309 | |
310 | mlog(0, "inode %lu, count = %zu, *ppos = %llu\n", | |
311 | inode->i_ino, count, *ppos); | |
312 | ||
313 | if (*ppos >= i_size_read(inode)) | |
314 | return -ENOSPC; | |
315 | ||
316 | if (!count) | |
317 | return 0; | |
318 | ||
319 | if (!access_ok(VERIFY_READ, buf, count)) | |
320 | return -EFAULT; | |
321 | ||
322 | /* don't write past the lvb */ | |
323 | if ((count + *ppos) > i_size_read(inode)) | |
324 | writelen = i_size_read(inode) - *ppos; | |
325 | else | |
326 | writelen = count - *ppos; | |
327 | ||
ad8100e0 | 328 | lvb_buf = kmalloc(writelen, GFP_NOFS); |
8df08c89 MF |
329 | if (!lvb_buf) |
330 | return -ENOMEM; | |
331 | ||
332 | bytes_left = copy_from_user(lvb_buf, buf, writelen); | |
333 | writelen -= bytes_left; | |
334 | if (writelen) | |
335 | user_dlm_write_lvb(inode, lvb_buf, writelen); | |
336 | ||
337 | kfree(lvb_buf); | |
338 | ||
339 | *ppos = *ppos + writelen; | |
340 | mlog(0, "wrote %zd bytes\n", writelen); | |
341 | return writelen; | |
342 | } | |
343 | ||
51cc5068 | 344 | static void dlmfs_init_once(void *foo) |
8df08c89 MF |
345 | { |
346 | struct dlmfs_inode_private *ip = | |
347 | (struct dlmfs_inode_private *) foo; | |
348 | ||
a35afb83 CL |
349 | ip->ip_dlm = NULL; |
350 | ip->ip_parent = NULL; | |
8df08c89 | 351 | |
a35afb83 | 352 | inode_init_once(&ip->ip_vfs_inode); |
8df08c89 MF |
353 | } |
354 | ||
355 | static struct inode *dlmfs_alloc_inode(struct super_block *sb) | |
356 | { | |
357 | struct dlmfs_inode_private *ip; | |
358 | ||
e6b4f8da | 359 | ip = kmem_cache_alloc(dlmfs_inode_cache, GFP_NOFS); |
8df08c89 MF |
360 | if (!ip) |
361 | return NULL; | |
362 | ||
363 | return &ip->ip_vfs_inode; | |
364 | } | |
365 | ||
366 | static void dlmfs_destroy_inode(struct inode *inode) | |
367 | { | |
368 | kmem_cache_free(dlmfs_inode_cache, DLMFS_I(inode)); | |
369 | } | |
370 | ||
371 | static void dlmfs_clear_inode(struct inode *inode) | |
372 | { | |
373 | int status; | |
374 | struct dlmfs_inode_private *ip; | |
375 | ||
376 | if (!inode) | |
377 | return; | |
378 | ||
379 | mlog(0, "inode %lu\n", inode->i_ino); | |
380 | ||
381 | ip = DLMFS_I(inode); | |
382 | ||
383 | if (S_ISREG(inode->i_mode)) { | |
384 | status = user_dlm_destroy_lock(&ip->ip_lockres); | |
385 | if (status < 0) | |
386 | mlog_errno(status); | |
387 | iput(ip->ip_parent); | |
388 | goto clear_fields; | |
389 | } | |
390 | ||
391 | mlog(0, "we're a directory, ip->ip_dlm = 0x%p\n", ip->ip_dlm); | |
392 | /* we must be a directory. If required, lets unregister the | |
393 | * dlm context now. */ | |
394 | if (ip->ip_dlm) | |
395 | user_dlm_unregister_context(ip->ip_dlm); | |
396 | clear_fields: | |
397 | ip->ip_parent = NULL; | |
398 | ip->ip_dlm = NULL; | |
399 | } | |
400 | ||
401 | static struct backing_dev_info dlmfs_backing_dev_info = { | |
d993831f | 402 | .name = "ocfs2-dlmfs", |
8df08c89 | 403 | .ra_pages = 0, /* No readahead */ |
e4ad08fe | 404 | .capabilities = BDI_CAP_NO_ACCT_AND_WRITEBACK, |
8df08c89 MF |
405 | }; |
406 | ||
407 | static struct inode *dlmfs_get_root_inode(struct super_block *sb) | |
408 | { | |
409 | struct inode *inode = new_inode(sb); | |
410 | int mode = S_IFDIR | 0755; | |
411 | struct dlmfs_inode_private *ip; | |
412 | ||
413 | if (inode) { | |
414 | ip = DLMFS_I(inode); | |
415 | ||
416 | inode->i_mode = mode; | |
b19c2a3b DH |
417 | inode->i_uid = current_fsuid(); |
418 | inode->i_gid = current_fsgid(); | |
8df08c89 MF |
419 | inode->i_mapping->backing_dev_info = &dlmfs_backing_dev_info; |
420 | inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME; | |
d8c76e6f | 421 | inc_nlink(inode); |
8df08c89 MF |
422 | |
423 | inode->i_fop = &simple_dir_operations; | |
424 | inode->i_op = &dlmfs_root_inode_operations; | |
425 | } | |
426 | ||
427 | return inode; | |
428 | } | |
429 | ||
430 | static struct inode *dlmfs_get_inode(struct inode *parent, | |
431 | struct dentry *dentry, | |
432 | int mode) | |
433 | { | |
434 | struct super_block *sb = parent->i_sb; | |
435 | struct inode * inode = new_inode(sb); | |
436 | struct dlmfs_inode_private *ip; | |
437 | ||
438 | if (!inode) | |
439 | return NULL; | |
440 | ||
441 | inode->i_mode = mode; | |
b19c2a3b DH |
442 | inode->i_uid = current_fsuid(); |
443 | inode->i_gid = current_fsgid(); | |
8df08c89 MF |
444 | inode->i_mapping->backing_dev_info = &dlmfs_backing_dev_info; |
445 | inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME; | |
446 | ||
447 | ip = DLMFS_I(inode); | |
448 | ip->ip_dlm = DLMFS_I(parent)->ip_dlm; | |
449 | ||
450 | switch (mode & S_IFMT) { | |
451 | default: | |
452 | /* for now we don't support anything other than | |
453 | * directories and regular files. */ | |
454 | BUG(); | |
455 | break; | |
456 | case S_IFREG: | |
457 | inode->i_op = &dlmfs_file_inode_operations; | |
458 | inode->i_fop = &dlmfs_file_operations; | |
459 | ||
460 | i_size_write(inode, DLM_LVB_LEN); | |
461 | ||
462 | user_dlm_lock_res_init(&ip->ip_lockres, dentry); | |
463 | ||
464 | /* released at clear_inode time, this insures that we | |
465 | * get to drop the dlm reference on each lock *before* | |
466 | * we call the unregister code for releasing parent | |
467 | * directories. */ | |
468 | ip->ip_parent = igrab(parent); | |
469 | BUG_ON(!ip->ip_parent); | |
470 | break; | |
471 | case S_IFDIR: | |
472 | inode->i_op = &dlmfs_dir_inode_operations; | |
473 | inode->i_fop = &simple_dir_operations; | |
474 | ||
475 | /* directory inodes start off with i_nlink == | |
476 | * 2 (for "." entry) */ | |
d8c76e6f | 477 | inc_nlink(inode); |
8df08c89 MF |
478 | break; |
479 | } | |
480 | ||
481 | if (parent->i_mode & S_ISGID) { | |
482 | inode->i_gid = parent->i_gid; | |
483 | if (S_ISDIR(mode)) | |
484 | inode->i_mode |= S_ISGID; | |
485 | } | |
486 | ||
487 | return inode; | |
488 | } | |
489 | ||
490 | /* | |
491 | * File creation. Allocate an inode, and we're done.. | |
492 | */ | |
493 | /* SMP-safe */ | |
494 | static int dlmfs_mkdir(struct inode * dir, | |
495 | struct dentry * dentry, | |
496 | int mode) | |
497 | { | |
498 | int status; | |
499 | struct inode *inode = NULL; | |
500 | struct qstr *domain = &dentry->d_name; | |
501 | struct dlmfs_inode_private *ip; | |
502 | struct dlm_ctxt *dlm; | |
d24fbcda | 503 | struct dlm_protocol_version proto = user_locking_protocol; |
8df08c89 MF |
504 | |
505 | mlog(0, "mkdir %.*s\n", domain->len, domain->name); | |
506 | ||
507 | /* verify that we have a proper domain */ | |
508 | if (domain->len >= O2NM_MAX_NAME_LEN) { | |
509 | status = -EINVAL; | |
510 | mlog(ML_ERROR, "invalid domain name for directory.\n"); | |
511 | goto bail; | |
512 | } | |
513 | ||
514 | inode = dlmfs_get_inode(dir, dentry, mode | S_IFDIR); | |
515 | if (!inode) { | |
516 | status = -ENOMEM; | |
517 | mlog_errno(status); | |
518 | goto bail; | |
519 | } | |
520 | ||
521 | ip = DLMFS_I(inode); | |
522 | ||
d24fbcda | 523 | dlm = user_dlm_register_context(domain, &proto); |
8df08c89 MF |
524 | if (IS_ERR(dlm)) { |
525 | status = PTR_ERR(dlm); | |
526 | mlog(ML_ERROR, "Error %d could not register domain \"%.*s\"\n", | |
527 | status, domain->len, domain->name); | |
528 | goto bail; | |
529 | } | |
530 | ip->ip_dlm = dlm; | |
531 | ||
d8c76e6f | 532 | inc_nlink(dir); |
8df08c89 MF |
533 | d_instantiate(dentry, inode); |
534 | dget(dentry); /* Extra count - pin the dentry in core */ | |
535 | ||
536 | status = 0; | |
537 | bail: | |
538 | if (status < 0) | |
539 | iput(inode); | |
540 | return status; | |
541 | } | |
542 | ||
543 | static int dlmfs_create(struct inode *dir, | |
544 | struct dentry *dentry, | |
545 | int mode, | |
546 | struct nameidata *nd) | |
547 | { | |
548 | int status = 0; | |
549 | struct inode *inode; | |
550 | struct qstr *name = &dentry->d_name; | |
551 | ||
552 | mlog(0, "create %.*s\n", name->len, name->name); | |
553 | ||
554 | /* verify name is valid and doesn't contain any dlm reserved | |
555 | * characters */ | |
556 | if (name->len >= USER_DLM_LOCK_ID_MAX_LEN || | |
557 | name->name[0] == '$') { | |
558 | status = -EINVAL; | |
559 | mlog(ML_ERROR, "invalid lock name, %.*s\n", name->len, | |
560 | name->name); | |
561 | goto bail; | |
562 | } | |
563 | ||
564 | inode = dlmfs_get_inode(dir, dentry, mode | S_IFREG); | |
565 | if (!inode) { | |
566 | status = -ENOMEM; | |
567 | mlog_errno(status); | |
568 | goto bail; | |
569 | } | |
570 | ||
571 | d_instantiate(dentry, inode); | |
572 | dget(dentry); /* Extra count - pin the dentry in core */ | |
573 | bail: | |
574 | return status; | |
575 | } | |
576 | ||
577 | static int dlmfs_unlink(struct inode *dir, | |
578 | struct dentry *dentry) | |
579 | { | |
580 | int status; | |
581 | struct inode *inode = dentry->d_inode; | |
582 | ||
583 | mlog(0, "unlink inode %lu\n", inode->i_ino); | |
584 | ||
585 | /* if there are no current holders, or none that are waiting | |
586 | * to acquire a lock, this basically destroys our lockres. */ | |
587 | status = user_dlm_destroy_lock(&DLMFS_I(inode)->ip_lockres); | |
588 | if (status < 0) { | |
589 | mlog(ML_ERROR, "unlink %.*s, error %d from destroy\n", | |
590 | dentry->d_name.len, dentry->d_name.name, status); | |
591 | goto bail; | |
592 | } | |
593 | status = simple_unlink(dir, dentry); | |
594 | bail: | |
595 | return status; | |
596 | } | |
597 | ||
598 | static int dlmfs_fill_super(struct super_block * sb, | |
599 | void * data, | |
600 | int silent) | |
601 | { | |
602 | struct inode * inode; | |
603 | struct dentry * root; | |
604 | ||
605 | sb->s_maxbytes = MAX_LFS_FILESIZE; | |
606 | sb->s_blocksize = PAGE_CACHE_SIZE; | |
607 | sb->s_blocksize_bits = PAGE_CACHE_SHIFT; | |
608 | sb->s_magic = DLMFS_MAGIC; | |
609 | sb->s_op = &dlmfs_ops; | |
610 | inode = dlmfs_get_root_inode(sb); | |
611 | if (!inode) | |
612 | return -ENOMEM; | |
613 | ||
614 | root = d_alloc_root(inode); | |
615 | if (!root) { | |
616 | iput(inode); | |
617 | return -ENOMEM; | |
618 | } | |
619 | sb->s_root = root; | |
620 | return 0; | |
621 | } | |
622 | ||
00977a59 | 623 | static const struct file_operations dlmfs_file_operations = { |
8df08c89 MF |
624 | .open = dlmfs_file_open, |
625 | .release = dlmfs_file_release, | |
65b6f340 | 626 | .poll = dlmfs_file_poll, |
8df08c89 MF |
627 | .read = dlmfs_file_read, |
628 | .write = dlmfs_file_write, | |
629 | }; | |
630 | ||
92e1d5be | 631 | static const struct inode_operations dlmfs_dir_inode_operations = { |
8df08c89 MF |
632 | .create = dlmfs_create, |
633 | .lookup = simple_lookup, | |
634 | .unlink = dlmfs_unlink, | |
635 | }; | |
636 | ||
637 | /* this way we can restrict mkdir to only the toplevel of the fs. */ | |
92e1d5be | 638 | static const struct inode_operations dlmfs_root_inode_operations = { |
8df08c89 MF |
639 | .lookup = simple_lookup, |
640 | .mkdir = dlmfs_mkdir, | |
641 | .rmdir = simple_rmdir, | |
642 | }; | |
643 | ||
ee9b6d61 | 644 | static const struct super_operations dlmfs_ops = { |
8df08c89 MF |
645 | .statfs = simple_statfs, |
646 | .alloc_inode = dlmfs_alloc_inode, | |
647 | .destroy_inode = dlmfs_destroy_inode, | |
648 | .clear_inode = dlmfs_clear_inode, | |
649 | .drop_inode = generic_delete_inode, | |
650 | }; | |
651 | ||
92e1d5be | 652 | static const struct inode_operations dlmfs_file_inode_operations = { |
8df08c89 | 653 | .getattr = simple_getattr, |
e8fce482 | 654 | .setattr = dlmfs_file_setattr, |
8df08c89 MF |
655 | }; |
656 | ||
454e2398 DH |
657 | static int dlmfs_get_sb(struct file_system_type *fs_type, |
658 | int flags, const char *dev_name, void *data, struct vfsmount *mnt) | |
8df08c89 | 659 | { |
454e2398 | 660 | return get_sb_nodev(fs_type, flags, data, dlmfs_fill_super, mnt); |
8df08c89 MF |
661 | } |
662 | ||
663 | static struct file_system_type dlmfs_fs_type = { | |
664 | .owner = THIS_MODULE, | |
665 | .name = "ocfs2_dlmfs", | |
666 | .get_sb = dlmfs_get_sb, | |
667 | .kill_sb = kill_litter_super, | |
668 | }; | |
669 | ||
670 | static int __init init_dlmfs_fs(void) | |
671 | { | |
672 | int status; | |
673 | int cleanup_inode = 0, cleanup_worker = 0; | |
674 | ||
675 | dlmfs_print_version(); | |
676 | ||
e0bf68dd PZ |
677 | status = bdi_init(&dlmfs_backing_dev_info); |
678 | if (status) | |
679 | return status; | |
680 | ||
8df08c89 MF |
681 | dlmfs_inode_cache = kmem_cache_create("dlmfs_inode_cache", |
682 | sizeof(struct dlmfs_inode_private), | |
fffb60f9 PJ |
683 | 0, (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT| |
684 | SLAB_MEM_SPREAD), | |
20c2df83 | 685 | dlmfs_init_once); |
07d9a395 CL |
686 | if (!dlmfs_inode_cache) { |
687 | status = -ENOMEM; | |
e0bf68dd | 688 | goto bail; |
07d9a395 | 689 | } |
8df08c89 MF |
690 | cleanup_inode = 1; |
691 | ||
692 | user_dlm_worker = create_singlethread_workqueue("user_dlm"); | |
693 | if (!user_dlm_worker) { | |
694 | status = -ENOMEM; | |
695 | goto bail; | |
696 | } | |
697 | cleanup_worker = 1; | |
698 | ||
699 | status = register_filesystem(&dlmfs_fs_type); | |
700 | bail: | |
701 | if (status) { | |
702 | if (cleanup_inode) | |
703 | kmem_cache_destroy(dlmfs_inode_cache); | |
704 | if (cleanup_worker) | |
705 | destroy_workqueue(user_dlm_worker); | |
e0bf68dd | 706 | bdi_destroy(&dlmfs_backing_dev_info); |
8df08c89 MF |
707 | } else |
708 | printk("OCFS2 User DLM kernel interface loaded\n"); | |
709 | return status; | |
710 | } | |
711 | ||
712 | static void __exit exit_dlmfs_fs(void) | |
713 | { | |
714 | unregister_filesystem(&dlmfs_fs_type); | |
715 | ||
716 | flush_workqueue(user_dlm_worker); | |
717 | destroy_workqueue(user_dlm_worker); | |
718 | ||
1a1d92c1 | 719 | kmem_cache_destroy(dlmfs_inode_cache); |
e0bf68dd PZ |
720 | |
721 | bdi_destroy(&dlmfs_backing_dev_info); | |
8df08c89 MF |
722 | } |
723 | ||
724 | MODULE_AUTHOR("Oracle"); | |
725 | MODULE_LICENSE("GPL"); | |
726 | ||
727 | module_init(init_dlmfs_fs) | |
728 | module_exit(exit_dlmfs_fs) |