]>
Commit | Line | Data |
---|---|---|
ccd979bd MF |
1 | /* -*- mode: c; c-basic-offset: 8; -*- |
2 | * vim: noexpandtab sw=8 ts=8 sts=0: | |
3 | * | |
4 | * dlmglue.c | |
5 | * | |
6 | * Code which implements an OCFS2 specific interface to our DLM. | |
7 | * | |
8 | * Copyright (C) 2003, 2004 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; either | |
13 | * version 2 of the License, or (at your option) any later version. | |
14 | * | |
15 | * This program is distributed in the hope that it will be useful, | |
16 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
17 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
18 | * General Public License for more details. | |
19 | * | |
20 | * You should have received a copy of the GNU General Public | |
21 | * License along with this program; if not, write to the | |
22 | * Free Software Foundation, Inc., 59 Temple Place - Suite 330, | |
23 | * Boston, MA 021110-1307, USA. | |
24 | */ | |
25 | ||
26 | #include <linux/types.h> | |
27 | #include <linux/slab.h> | |
28 | #include <linux/highmem.h> | |
29 | #include <linux/mm.h> | |
30 | #include <linux/smp_lock.h> | |
31 | #include <linux/crc32.h> | |
32 | #include <linux/kthread.h> | |
33 | #include <linux/pagemap.h> | |
34 | #include <linux/debugfs.h> | |
35 | #include <linux/seq_file.h> | |
36 | ||
37 | #include <cluster/heartbeat.h> | |
38 | #include <cluster/nodemanager.h> | |
39 | #include <cluster/tcp.h> | |
40 | ||
41 | #include <dlm/dlmapi.h> | |
42 | ||
43 | #define MLOG_MASK_PREFIX ML_DLM_GLUE | |
44 | #include <cluster/masklog.h> | |
45 | ||
46 | #include "ocfs2.h" | |
47 | ||
48 | #include "alloc.h" | |
d680efe9 | 49 | #include "dcache.h" |
ccd979bd MF |
50 | #include "dlmglue.h" |
51 | #include "extent_map.h" | |
52 | #include "heartbeat.h" | |
53 | #include "inode.h" | |
54 | #include "journal.h" | |
55 | #include "slot_map.h" | |
56 | #include "super.h" | |
57 | #include "uptodate.h" | |
58 | #include "vote.h" | |
59 | ||
60 | #include "buffer_head_io.h" | |
61 | ||
62 | struct ocfs2_mask_waiter { | |
63 | struct list_head mw_item; | |
64 | int mw_status; | |
65 | struct completion mw_complete; | |
66 | unsigned long mw_mask; | |
67 | unsigned long mw_goal; | |
68 | }; | |
69 | ||
ccd979bd MF |
70 | static void ocfs2_inode_bast_func(void *opaque, |
71 | int level); | |
d680efe9 MF |
72 | static void ocfs2_dentry_bast_func(void *opaque, |
73 | int level); | |
ccd979bd MF |
74 | static void ocfs2_super_bast_func(void *opaque, |
75 | int level); | |
ccd979bd MF |
76 | static void ocfs2_rename_bast_func(void *opaque, |
77 | int level); | |
78 | ||
d680efe9 MF |
79 | /* |
80 | * Return value from ocfs2_convert_worker_t functions. | |
81 | * | |
82 | * These control the precise actions of ocfs2_generic_unblock_lock() | |
83 | * and ocfs2_process_blocked_lock() | |
84 | * | |
85 | */ | |
86 | enum ocfs2_unblock_action { | |
87 | UNBLOCK_CONTINUE = 0, /* Continue downconvert */ | |
88 | UNBLOCK_CONTINUE_POST = 1, /* Continue downconvert, fire | |
89 | * ->post_unlock callback */ | |
90 | UNBLOCK_STOP_POST = 2, /* Do not downconvert, fire | |
91 | * ->post_unlock() callback. */ | |
92 | }; | |
93 | ||
94 | struct ocfs2_unblock_ctl { | |
95 | int requeue; | |
96 | enum ocfs2_unblock_action unblock_action; | |
97 | }; | |
98 | ||
ccd979bd | 99 | static int ocfs2_unblock_meta(struct ocfs2_lock_res *lockres, |
d680efe9 | 100 | struct ocfs2_unblock_ctl *ctl); |
ccd979bd | 101 | static int ocfs2_unblock_data(struct ocfs2_lock_res *lockres, |
d680efe9 | 102 | struct ocfs2_unblock_ctl *ctl); |
ccd979bd | 103 | static int ocfs2_unblock_inode_lock(struct ocfs2_lock_res *lockres, |
d680efe9 MF |
104 | struct ocfs2_unblock_ctl *ctl); |
105 | static int ocfs2_unblock_dentry_lock(struct ocfs2_lock_res *lockres, | |
106 | struct ocfs2_unblock_ctl *ctl); | |
ccd979bd | 107 | static int ocfs2_unblock_osb_lock(struct ocfs2_lock_res *lockres, |
d680efe9 MF |
108 | struct ocfs2_unblock_ctl *ctl); |
109 | ||
110 | static void ocfs2_dentry_post_unlock(struct ocfs2_super *osb, | |
111 | struct ocfs2_lock_res *lockres); | |
ccd979bd | 112 | |
f625c979 MF |
113 | /* |
114 | * OCFS2 Lock Resource Operations | |
115 | * | |
116 | * These fine tune the behavior of the generic dlmglue locking infrastructure. | |
117 | */ | |
ccd979bd | 118 | struct ocfs2_lock_res_ops { |
ccd979bd | 119 | void (*bast)(void *, int); |
d680efe9 MF |
120 | int (*unblock)(struct ocfs2_lock_res *, struct ocfs2_unblock_ctl *); |
121 | void (*post_unlock)(struct ocfs2_super *, struct ocfs2_lock_res *); | |
f625c979 MF |
122 | |
123 | /* | |
124 | * LOCK_TYPE_* flags which describe the specific requirements | |
125 | * of a lock type. Descriptions of each individual flag follow. | |
126 | */ | |
127 | int flags; | |
ccd979bd MF |
128 | }; |
129 | ||
f625c979 MF |
130 | /* |
131 | * Some locks want to "refresh" potentially stale data when a | |
132 | * meaningful (PRMODE or EXMODE) lock level is first obtained. If this | |
133 | * flag is set, the OCFS2_LOCK_NEEDS_REFRESH flag will be set on the | |
134 | * individual lockres l_flags member from the ast function. It is | |
135 | * expected that the locking wrapper will clear the | |
136 | * OCFS2_LOCK_NEEDS_REFRESH flag when done. | |
137 | */ | |
138 | #define LOCK_TYPE_REQUIRES_REFRESH 0x1 | |
139 | ||
d680efe9 MF |
140 | typedef int (ocfs2_convert_worker_t)(struct ocfs2_lock_res *, int); |
141 | static int ocfs2_generic_unblock_lock(struct ocfs2_super *osb, | |
142 | struct ocfs2_lock_res *lockres, | |
143 | struct ocfs2_unblock_ctl *ctl, | |
144 | ocfs2_convert_worker_t *worker); | |
145 | ||
ccd979bd | 146 | static struct ocfs2_lock_res_ops ocfs2_inode_rw_lops = { |
ccd979bd | 147 | .bast = ocfs2_inode_bast_func, |
ccd979bd | 148 | .unblock = ocfs2_unblock_inode_lock, |
f625c979 | 149 | .flags = 0, |
ccd979bd MF |
150 | }; |
151 | ||
152 | static struct ocfs2_lock_res_ops ocfs2_inode_meta_lops = { | |
ccd979bd | 153 | .bast = ocfs2_inode_bast_func, |
ccd979bd | 154 | .unblock = ocfs2_unblock_meta, |
f625c979 | 155 | .flags = LOCK_TYPE_REQUIRES_REFRESH, |
ccd979bd MF |
156 | }; |
157 | ||
ccd979bd | 158 | static struct ocfs2_lock_res_ops ocfs2_inode_data_lops = { |
ccd979bd | 159 | .bast = ocfs2_inode_bast_func, |
ccd979bd | 160 | .unblock = ocfs2_unblock_data, |
f625c979 | 161 | .flags = 0, |
ccd979bd MF |
162 | }; |
163 | ||
164 | static struct ocfs2_lock_res_ops ocfs2_super_lops = { | |
ccd979bd | 165 | .bast = ocfs2_super_bast_func, |
ccd979bd | 166 | .unblock = ocfs2_unblock_osb_lock, |
f625c979 | 167 | .flags = LOCK_TYPE_REQUIRES_REFRESH, |
ccd979bd MF |
168 | }; |
169 | ||
170 | static struct ocfs2_lock_res_ops ocfs2_rename_lops = { | |
ccd979bd | 171 | .bast = ocfs2_rename_bast_func, |
ccd979bd | 172 | .unblock = ocfs2_unblock_osb_lock, |
f625c979 | 173 | .flags = 0, |
ccd979bd MF |
174 | }; |
175 | ||
d680efe9 | 176 | static struct ocfs2_lock_res_ops ocfs2_dentry_lops = { |
d680efe9 | 177 | .bast = ocfs2_dentry_bast_func, |
d680efe9 MF |
178 | .unblock = ocfs2_unblock_dentry_lock, |
179 | .post_unlock = ocfs2_dentry_post_unlock, | |
f625c979 | 180 | .flags = 0, |
d680efe9 MF |
181 | }; |
182 | ||
ccd979bd MF |
183 | static inline int ocfs2_is_inode_lock(struct ocfs2_lock_res *lockres) |
184 | { | |
185 | return lockres->l_type == OCFS2_LOCK_TYPE_META || | |
186 | lockres->l_type == OCFS2_LOCK_TYPE_DATA || | |
187 | lockres->l_type == OCFS2_LOCK_TYPE_RW; | |
188 | } | |
189 | ||
190 | static inline int ocfs2_is_super_lock(struct ocfs2_lock_res *lockres) | |
191 | { | |
192 | return lockres->l_type == OCFS2_LOCK_TYPE_SUPER; | |
193 | } | |
194 | ||
195 | static inline int ocfs2_is_rename_lock(struct ocfs2_lock_res *lockres) | |
196 | { | |
197 | return lockres->l_type == OCFS2_LOCK_TYPE_RENAME; | |
198 | } | |
199 | ||
200 | static inline struct ocfs2_super *ocfs2_lock_res_super(struct ocfs2_lock_res *lockres) | |
201 | { | |
202 | BUG_ON(!ocfs2_is_super_lock(lockres) | |
203 | && !ocfs2_is_rename_lock(lockres)); | |
204 | ||
205 | return (struct ocfs2_super *) lockres->l_priv; | |
206 | } | |
207 | ||
208 | static inline struct inode *ocfs2_lock_res_inode(struct ocfs2_lock_res *lockres) | |
209 | { | |
210 | BUG_ON(!ocfs2_is_inode_lock(lockres)); | |
211 | ||
212 | return (struct inode *) lockres->l_priv; | |
213 | } | |
214 | ||
d680efe9 MF |
215 | static inline struct ocfs2_dentry_lock *ocfs2_lock_res_dl(struct ocfs2_lock_res *lockres) |
216 | { | |
217 | BUG_ON(lockres->l_type != OCFS2_LOCK_TYPE_DENTRY); | |
218 | ||
219 | return (struct ocfs2_dentry_lock *)lockres->l_priv; | |
220 | } | |
221 | ||
ccd979bd MF |
222 | static int ocfs2_lock_create(struct ocfs2_super *osb, |
223 | struct ocfs2_lock_res *lockres, | |
224 | int level, | |
225 | int dlm_flags); | |
226 | static inline int ocfs2_may_continue_on_blocked_lock(struct ocfs2_lock_res *lockres, | |
227 | int wanted); | |
228 | static void ocfs2_cluster_unlock(struct ocfs2_super *osb, | |
229 | struct ocfs2_lock_res *lockres, | |
230 | int level); | |
231 | static inline void ocfs2_generic_handle_downconvert_action(struct ocfs2_lock_res *lockres); | |
232 | static inline void ocfs2_generic_handle_convert_action(struct ocfs2_lock_res *lockres); | |
233 | static inline void ocfs2_generic_handle_attach_action(struct ocfs2_lock_res *lockres); | |
234 | static int ocfs2_generic_handle_bast(struct ocfs2_lock_res *lockres, int level); | |
235 | static void ocfs2_schedule_blocked_lock(struct ocfs2_super *osb, | |
236 | struct ocfs2_lock_res *lockres); | |
237 | static inline void ocfs2_recover_from_dlm_error(struct ocfs2_lock_res *lockres, | |
238 | int convert); | |
239 | #define ocfs2_log_dlm_error(_func, _stat, _lockres) do { \ | |
240 | mlog(ML_ERROR, "Dlm error \"%s\" while calling %s on " \ | |
241 | "resource %s: %s\n", dlm_errname(_stat), _func, \ | |
242 | _lockres->l_name, dlm_errmsg(_stat)); \ | |
243 | } while (0) | |
244 | static void ocfs2_vote_on_unlock(struct ocfs2_super *osb, | |
245 | struct ocfs2_lock_res *lockres); | |
246 | static int ocfs2_meta_lock_update(struct inode *inode, | |
247 | struct buffer_head **bh); | |
248 | static void ocfs2_drop_osb_locks(struct ocfs2_super *osb); | |
249 | static inline int ocfs2_highest_compat_lock_level(int level); | |
250 | static inline int ocfs2_can_downconvert_meta_lock(struct inode *inode, | |
251 | struct ocfs2_lock_res *lockres, | |
252 | int new_level); | |
253 | ||
ccd979bd MF |
254 | static void ocfs2_build_lock_name(enum ocfs2_lock_type type, |
255 | u64 blkno, | |
256 | u32 generation, | |
257 | char *name) | |
258 | { | |
259 | int len; | |
260 | ||
261 | mlog_entry_void(); | |
262 | ||
263 | BUG_ON(type >= OCFS2_NUM_LOCK_TYPES); | |
264 | ||
b0697053 MF |
265 | len = snprintf(name, OCFS2_LOCK_ID_MAX_LEN, "%c%s%016llx%08x", |
266 | ocfs2_lock_type_char(type), OCFS2_LOCK_ID_PAD, | |
267 | (long long)blkno, generation); | |
ccd979bd MF |
268 | |
269 | BUG_ON(len != (OCFS2_LOCK_ID_MAX_LEN - 1)); | |
270 | ||
271 | mlog(0, "built lock resource with name: %s\n", name); | |
272 | ||
273 | mlog_exit_void(); | |
274 | } | |
275 | ||
34af946a | 276 | static DEFINE_SPINLOCK(ocfs2_dlm_tracking_lock); |
ccd979bd MF |
277 | |
278 | static void ocfs2_add_lockres_tracking(struct ocfs2_lock_res *res, | |
279 | struct ocfs2_dlm_debug *dlm_debug) | |
280 | { | |
281 | mlog(0, "Add tracking for lockres %s\n", res->l_name); | |
282 | ||
283 | spin_lock(&ocfs2_dlm_tracking_lock); | |
284 | list_add(&res->l_debug_list, &dlm_debug->d_lockres_tracking); | |
285 | spin_unlock(&ocfs2_dlm_tracking_lock); | |
286 | } | |
287 | ||
288 | static void ocfs2_remove_lockres_tracking(struct ocfs2_lock_res *res) | |
289 | { | |
290 | spin_lock(&ocfs2_dlm_tracking_lock); | |
291 | if (!list_empty(&res->l_debug_list)) | |
292 | list_del_init(&res->l_debug_list); | |
293 | spin_unlock(&ocfs2_dlm_tracking_lock); | |
294 | } | |
295 | ||
296 | static void ocfs2_lock_res_init_common(struct ocfs2_super *osb, | |
297 | struct ocfs2_lock_res *res, | |
298 | enum ocfs2_lock_type type, | |
ccd979bd MF |
299 | struct ocfs2_lock_res_ops *ops, |
300 | void *priv) | |
301 | { | |
ccd979bd MF |
302 | res->l_type = type; |
303 | res->l_ops = ops; | |
304 | res->l_priv = priv; | |
305 | ||
306 | res->l_level = LKM_IVMODE; | |
307 | res->l_requested = LKM_IVMODE; | |
308 | res->l_blocking = LKM_IVMODE; | |
309 | res->l_action = OCFS2_AST_INVALID; | |
310 | res->l_unlock_action = OCFS2_UNLOCK_INVALID; | |
311 | ||
312 | res->l_flags = OCFS2_LOCK_INITIALIZED; | |
313 | ||
314 | ocfs2_add_lockres_tracking(res, osb->osb_dlm_debug); | |
315 | } | |
316 | ||
317 | void ocfs2_lock_res_init_once(struct ocfs2_lock_res *res) | |
318 | { | |
319 | /* This also clears out the lock status block */ | |
320 | memset(res, 0, sizeof(struct ocfs2_lock_res)); | |
321 | spin_lock_init(&res->l_lock); | |
322 | init_waitqueue_head(&res->l_event); | |
323 | INIT_LIST_HEAD(&res->l_blocked_list); | |
324 | INIT_LIST_HEAD(&res->l_mask_waiters); | |
325 | } | |
326 | ||
327 | void ocfs2_inode_lock_res_init(struct ocfs2_lock_res *res, | |
328 | enum ocfs2_lock_type type, | |
24c19ef4 | 329 | unsigned int generation, |
ccd979bd MF |
330 | struct inode *inode) |
331 | { | |
332 | struct ocfs2_lock_res_ops *ops; | |
333 | ||
334 | switch(type) { | |
335 | case OCFS2_LOCK_TYPE_RW: | |
336 | ops = &ocfs2_inode_rw_lops; | |
337 | break; | |
338 | case OCFS2_LOCK_TYPE_META: | |
339 | ops = &ocfs2_inode_meta_lops; | |
340 | break; | |
341 | case OCFS2_LOCK_TYPE_DATA: | |
342 | ops = &ocfs2_inode_data_lops; | |
343 | break; | |
344 | default: | |
345 | mlog_bug_on_msg(1, "type: %d\n", type); | |
346 | ops = NULL; /* thanks, gcc */ | |
347 | break; | |
348 | }; | |
349 | ||
d680efe9 | 350 | ocfs2_build_lock_name(type, OCFS2_I(inode)->ip_blkno, |
24c19ef4 | 351 | generation, res->l_name); |
d680efe9 MF |
352 | ocfs2_lock_res_init_common(OCFS2_SB(inode->i_sb), res, type, ops, inode); |
353 | } | |
354 | ||
355 | static __u64 ocfs2_get_dentry_lock_ino(struct ocfs2_lock_res *lockres) | |
356 | { | |
357 | __be64 inode_blkno_be; | |
358 | ||
359 | memcpy(&inode_blkno_be, &lockres->l_name[OCFS2_DENTRY_LOCK_INO_START], | |
360 | sizeof(__be64)); | |
361 | ||
362 | return be64_to_cpu(inode_blkno_be); | |
363 | } | |
364 | ||
365 | void ocfs2_dentry_lock_res_init(struct ocfs2_dentry_lock *dl, | |
366 | u64 parent, struct inode *inode) | |
367 | { | |
368 | int len; | |
369 | u64 inode_blkno = OCFS2_I(inode)->ip_blkno; | |
370 | __be64 inode_blkno_be = cpu_to_be64(inode_blkno); | |
371 | struct ocfs2_lock_res *lockres = &dl->dl_lockres; | |
372 | ||
373 | ocfs2_lock_res_init_once(lockres); | |
374 | ||
375 | /* | |
376 | * Unfortunately, the standard lock naming scheme won't work | |
377 | * here because we have two 16 byte values to use. Instead, | |
378 | * we'll stuff the inode number as a binary value. We still | |
379 | * want error prints to show something without garbling the | |
380 | * display, so drop a null byte in there before the inode | |
381 | * number. A future version of OCFS2 will likely use all | |
382 | * binary lock names. The stringified names have been a | |
383 | * tremendous aid in debugging, but now that the debugfs | |
384 | * interface exists, we can mangle things there if need be. | |
385 | * | |
386 | * NOTE: We also drop the standard "pad" value (the total lock | |
387 | * name size stays the same though - the last part is all | |
388 | * zeros due to the memset in ocfs2_lock_res_init_once() | |
389 | */ | |
390 | len = snprintf(lockres->l_name, OCFS2_DENTRY_LOCK_INO_START, | |
391 | "%c%016llx", | |
392 | ocfs2_lock_type_char(OCFS2_LOCK_TYPE_DENTRY), | |
393 | (long long)parent); | |
394 | ||
395 | BUG_ON(len != (OCFS2_DENTRY_LOCK_INO_START - 1)); | |
396 | ||
397 | memcpy(&lockres->l_name[OCFS2_DENTRY_LOCK_INO_START], &inode_blkno_be, | |
398 | sizeof(__be64)); | |
399 | ||
400 | ocfs2_lock_res_init_common(OCFS2_SB(inode->i_sb), lockres, | |
401 | OCFS2_LOCK_TYPE_DENTRY, &ocfs2_dentry_lops, | |
402 | dl); | |
ccd979bd MF |
403 | } |
404 | ||
405 | static void ocfs2_super_lock_res_init(struct ocfs2_lock_res *res, | |
406 | struct ocfs2_super *osb) | |
407 | { | |
408 | /* Superblock lockres doesn't come from a slab so we call init | |
409 | * once on it manually. */ | |
410 | ocfs2_lock_res_init_once(res); | |
d680efe9 MF |
411 | ocfs2_build_lock_name(OCFS2_LOCK_TYPE_SUPER, OCFS2_SUPER_BLOCK_BLKNO, |
412 | 0, res->l_name); | |
ccd979bd | 413 | ocfs2_lock_res_init_common(osb, res, OCFS2_LOCK_TYPE_SUPER, |
ccd979bd MF |
414 | &ocfs2_super_lops, osb); |
415 | } | |
416 | ||
417 | static void ocfs2_rename_lock_res_init(struct ocfs2_lock_res *res, | |
418 | struct ocfs2_super *osb) | |
419 | { | |
420 | /* Rename lockres doesn't come from a slab so we call init | |
421 | * once on it manually. */ | |
422 | ocfs2_lock_res_init_once(res); | |
d680efe9 MF |
423 | ocfs2_build_lock_name(OCFS2_LOCK_TYPE_RENAME, 0, 0, res->l_name); |
424 | ocfs2_lock_res_init_common(osb, res, OCFS2_LOCK_TYPE_RENAME, | |
ccd979bd MF |
425 | &ocfs2_rename_lops, osb); |
426 | } | |
427 | ||
428 | void ocfs2_lock_res_free(struct ocfs2_lock_res *res) | |
429 | { | |
430 | mlog_entry_void(); | |
431 | ||
432 | if (!(res->l_flags & OCFS2_LOCK_INITIALIZED)) | |
433 | return; | |
434 | ||
435 | ocfs2_remove_lockres_tracking(res); | |
436 | ||
437 | mlog_bug_on_msg(!list_empty(&res->l_blocked_list), | |
438 | "Lockres %s is on the blocked list\n", | |
439 | res->l_name); | |
440 | mlog_bug_on_msg(!list_empty(&res->l_mask_waiters), | |
441 | "Lockres %s has mask waiters pending\n", | |
442 | res->l_name); | |
443 | mlog_bug_on_msg(spin_is_locked(&res->l_lock), | |
444 | "Lockres %s is locked\n", | |
445 | res->l_name); | |
446 | mlog_bug_on_msg(res->l_ro_holders, | |
447 | "Lockres %s has %u ro holders\n", | |
448 | res->l_name, res->l_ro_holders); | |
449 | mlog_bug_on_msg(res->l_ex_holders, | |
450 | "Lockres %s has %u ex holders\n", | |
451 | res->l_name, res->l_ex_holders); | |
452 | ||
453 | /* Need to clear out the lock status block for the dlm */ | |
454 | memset(&res->l_lksb, 0, sizeof(res->l_lksb)); | |
455 | ||
456 | res->l_flags = 0UL; | |
457 | mlog_exit_void(); | |
458 | } | |
459 | ||
460 | static inline void ocfs2_inc_holders(struct ocfs2_lock_res *lockres, | |
461 | int level) | |
462 | { | |
463 | mlog_entry_void(); | |
464 | ||
465 | BUG_ON(!lockres); | |
466 | ||
467 | switch(level) { | |
468 | case LKM_EXMODE: | |
469 | lockres->l_ex_holders++; | |
470 | break; | |
471 | case LKM_PRMODE: | |
472 | lockres->l_ro_holders++; | |
473 | break; | |
474 | default: | |
475 | BUG(); | |
476 | } | |
477 | ||
478 | mlog_exit_void(); | |
479 | } | |
480 | ||
481 | static inline void ocfs2_dec_holders(struct ocfs2_lock_res *lockres, | |
482 | int level) | |
483 | { | |
484 | mlog_entry_void(); | |
485 | ||
486 | BUG_ON(!lockres); | |
487 | ||
488 | switch(level) { | |
489 | case LKM_EXMODE: | |
490 | BUG_ON(!lockres->l_ex_holders); | |
491 | lockres->l_ex_holders--; | |
492 | break; | |
493 | case LKM_PRMODE: | |
494 | BUG_ON(!lockres->l_ro_holders); | |
495 | lockres->l_ro_holders--; | |
496 | break; | |
497 | default: | |
498 | BUG(); | |
499 | } | |
500 | mlog_exit_void(); | |
501 | } | |
502 | ||
503 | /* WARNING: This function lives in a world where the only three lock | |
504 | * levels are EX, PR, and NL. It *will* have to be adjusted when more | |
505 | * lock types are added. */ | |
506 | static inline int ocfs2_highest_compat_lock_level(int level) | |
507 | { | |
508 | int new_level = LKM_EXMODE; | |
509 | ||
510 | if (level == LKM_EXMODE) | |
511 | new_level = LKM_NLMODE; | |
512 | else if (level == LKM_PRMODE) | |
513 | new_level = LKM_PRMODE; | |
514 | return new_level; | |
515 | } | |
516 | ||
517 | static void lockres_set_flags(struct ocfs2_lock_res *lockres, | |
518 | unsigned long newflags) | |
519 | { | |
520 | struct list_head *pos, *tmp; | |
521 | struct ocfs2_mask_waiter *mw; | |
522 | ||
523 | assert_spin_locked(&lockres->l_lock); | |
524 | ||
525 | lockres->l_flags = newflags; | |
526 | ||
527 | list_for_each_safe(pos, tmp, &lockres->l_mask_waiters) { | |
528 | mw = list_entry(pos, struct ocfs2_mask_waiter, mw_item); | |
529 | if ((lockres->l_flags & mw->mw_mask) != mw->mw_goal) | |
530 | continue; | |
531 | ||
532 | list_del_init(&mw->mw_item); | |
533 | mw->mw_status = 0; | |
534 | complete(&mw->mw_complete); | |
535 | } | |
536 | } | |
537 | static void lockres_or_flags(struct ocfs2_lock_res *lockres, unsigned long or) | |
538 | { | |
539 | lockres_set_flags(lockres, lockres->l_flags | or); | |
540 | } | |
541 | static void lockres_clear_flags(struct ocfs2_lock_res *lockres, | |
542 | unsigned long clear) | |
543 | { | |
544 | lockres_set_flags(lockres, lockres->l_flags & ~clear); | |
545 | } | |
546 | ||
547 | static inline void ocfs2_generic_handle_downconvert_action(struct ocfs2_lock_res *lockres) | |
548 | { | |
549 | mlog_entry_void(); | |
550 | ||
551 | BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BUSY)); | |
552 | BUG_ON(!(lockres->l_flags & OCFS2_LOCK_ATTACHED)); | |
553 | BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BLOCKED)); | |
554 | BUG_ON(lockres->l_blocking <= LKM_NLMODE); | |
555 | ||
556 | lockres->l_level = lockres->l_requested; | |
557 | if (lockres->l_level <= | |
558 | ocfs2_highest_compat_lock_level(lockres->l_blocking)) { | |
559 | lockres->l_blocking = LKM_NLMODE; | |
560 | lockres_clear_flags(lockres, OCFS2_LOCK_BLOCKED); | |
561 | } | |
562 | lockres_clear_flags(lockres, OCFS2_LOCK_BUSY); | |
563 | ||
564 | mlog_exit_void(); | |
565 | } | |
566 | ||
567 | static inline void ocfs2_generic_handle_convert_action(struct ocfs2_lock_res *lockres) | |
568 | { | |
569 | mlog_entry_void(); | |
570 | ||
571 | BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BUSY)); | |
572 | BUG_ON(!(lockres->l_flags & OCFS2_LOCK_ATTACHED)); | |
573 | ||
574 | /* Convert from RO to EX doesn't really need anything as our | |
575 | * information is already up to data. Convert from NL to | |
576 | * *anything* however should mark ourselves as needing an | |
577 | * update */ | |
f625c979 MF |
578 | if (lockres->l_level == LKM_NLMODE && |
579 | lockres->l_ops->flags & LOCK_TYPE_REQUIRES_REFRESH) | |
ccd979bd MF |
580 | lockres_or_flags(lockres, OCFS2_LOCK_NEEDS_REFRESH); |
581 | ||
582 | lockres->l_level = lockres->l_requested; | |
583 | lockres_clear_flags(lockres, OCFS2_LOCK_BUSY); | |
584 | ||
585 | mlog_exit_void(); | |
586 | } | |
587 | ||
588 | static inline void ocfs2_generic_handle_attach_action(struct ocfs2_lock_res *lockres) | |
589 | { | |
590 | mlog_entry_void(); | |
591 | ||
592 | BUG_ON((!lockres->l_flags & OCFS2_LOCK_BUSY)); | |
593 | BUG_ON(lockres->l_flags & OCFS2_LOCK_ATTACHED); | |
594 | ||
595 | if (lockres->l_requested > LKM_NLMODE && | |
f625c979 MF |
596 | !(lockres->l_flags & OCFS2_LOCK_LOCAL) && |
597 | lockres->l_ops->flags & LOCK_TYPE_REQUIRES_REFRESH) | |
ccd979bd MF |
598 | lockres_or_flags(lockres, OCFS2_LOCK_NEEDS_REFRESH); |
599 | ||
600 | lockres->l_level = lockres->l_requested; | |
601 | lockres_or_flags(lockres, OCFS2_LOCK_ATTACHED); | |
602 | lockres_clear_flags(lockres, OCFS2_LOCK_BUSY); | |
603 | ||
604 | mlog_exit_void(); | |
605 | } | |
606 | ||
ccd979bd MF |
607 | static int ocfs2_generic_handle_bast(struct ocfs2_lock_res *lockres, |
608 | int level) | |
609 | { | |
610 | int needs_downconvert = 0; | |
611 | mlog_entry_void(); | |
612 | ||
613 | assert_spin_locked(&lockres->l_lock); | |
614 | ||
615 | lockres_or_flags(lockres, OCFS2_LOCK_BLOCKED); | |
616 | ||
617 | if (level > lockres->l_blocking) { | |
618 | /* only schedule a downconvert if we haven't already scheduled | |
619 | * one that goes low enough to satisfy the level we're | |
620 | * blocking. this also catches the case where we get | |
621 | * duplicate BASTs */ | |
622 | if (ocfs2_highest_compat_lock_level(level) < | |
623 | ocfs2_highest_compat_lock_level(lockres->l_blocking)) | |
624 | needs_downconvert = 1; | |
625 | ||
626 | lockres->l_blocking = level; | |
627 | } | |
628 | ||
629 | mlog_exit(needs_downconvert); | |
630 | return needs_downconvert; | |
631 | } | |
632 | ||
633 | static void ocfs2_generic_bast_func(struct ocfs2_super *osb, | |
634 | struct ocfs2_lock_res *lockres, | |
635 | int level) | |
636 | { | |
637 | int needs_downconvert; | |
638 | unsigned long flags; | |
639 | ||
640 | mlog_entry_void(); | |
641 | ||
642 | BUG_ON(level <= LKM_NLMODE); | |
643 | ||
644 | spin_lock_irqsave(&lockres->l_lock, flags); | |
645 | needs_downconvert = ocfs2_generic_handle_bast(lockres, level); | |
646 | if (needs_downconvert) | |
647 | ocfs2_schedule_blocked_lock(osb, lockres); | |
648 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
649 | ||
d680efe9 MF |
650 | wake_up(&lockres->l_event); |
651 | ||
ccd979bd MF |
652 | ocfs2_kick_vote_thread(osb); |
653 | ||
ccd979bd MF |
654 | mlog_exit_void(); |
655 | } | |
656 | ||
657 | static void ocfs2_inode_bast_func(void *opaque, int level) | |
658 | { | |
659 | struct ocfs2_lock_res *lockres = opaque; | |
660 | struct inode *inode; | |
661 | struct ocfs2_super *osb; | |
662 | ||
663 | mlog_entry_void(); | |
664 | ||
665 | BUG_ON(!ocfs2_is_inode_lock(lockres)); | |
666 | ||
667 | inode = ocfs2_lock_res_inode(lockres); | |
668 | osb = OCFS2_SB(inode->i_sb); | |
669 | ||
b0697053 MF |
670 | mlog(0, "BAST fired for inode %llu, blocking %d, level %d type %s\n", |
671 | (unsigned long long)OCFS2_I(inode)->ip_blkno, level, | |
672 | lockres->l_level, ocfs2_lock_type_string(lockres->l_type)); | |
ccd979bd MF |
673 | |
674 | ocfs2_generic_bast_func(osb, lockres, level); | |
675 | ||
676 | mlog_exit_void(); | |
677 | } | |
678 | ||
e92d57df | 679 | static void ocfs2_locking_ast(void *opaque) |
ccd979bd | 680 | { |
e92d57df | 681 | struct ocfs2_lock_res *lockres = opaque; |
ccd979bd MF |
682 | struct dlm_lockstatus *lksb = &lockres->l_lksb; |
683 | unsigned long flags; | |
684 | ||
685 | spin_lock_irqsave(&lockres->l_lock, flags); | |
686 | ||
687 | if (lksb->status != DLM_NORMAL) { | |
688 | mlog(ML_ERROR, "lockres %s: lksb status value of %u!\n", | |
689 | lockres->l_name, lksb->status); | |
690 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
691 | return; | |
692 | } | |
693 | ||
694 | switch(lockres->l_action) { | |
695 | case OCFS2_AST_ATTACH: | |
696 | ocfs2_generic_handle_attach_action(lockres); | |
e92d57df | 697 | lockres_clear_flags(lockres, OCFS2_LOCK_LOCAL); |
ccd979bd MF |
698 | break; |
699 | case OCFS2_AST_CONVERT: | |
700 | ocfs2_generic_handle_convert_action(lockres); | |
701 | break; | |
702 | case OCFS2_AST_DOWNCONVERT: | |
703 | ocfs2_generic_handle_downconvert_action(lockres); | |
704 | break; | |
705 | default: | |
e92d57df MF |
706 | mlog(ML_ERROR, "lockres %s: ast fired with invalid action: %u " |
707 | "lockres flags = 0x%lx, unlock action: %u\n", | |
708 | lockres->l_name, lockres->l_action, lockres->l_flags, | |
709 | lockres->l_unlock_action); | |
ccd979bd MF |
710 | BUG(); |
711 | } | |
712 | ||
ccd979bd MF |
713 | /* set it to something invalid so if we get called again we |
714 | * can catch it. */ | |
715 | lockres->l_action = OCFS2_AST_INVALID; | |
ccd979bd MF |
716 | |
717 | wake_up(&lockres->l_event); | |
d680efe9 | 718 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
ccd979bd MF |
719 | } |
720 | ||
ccd979bd MF |
721 | static void ocfs2_super_bast_func(void *opaque, |
722 | int level) | |
723 | { | |
724 | struct ocfs2_lock_res *lockres = opaque; | |
725 | struct ocfs2_super *osb; | |
726 | ||
727 | mlog_entry_void(); | |
728 | mlog(0, "Superblock BAST fired\n"); | |
729 | ||
730 | BUG_ON(!ocfs2_is_super_lock(lockres)); | |
731 | osb = ocfs2_lock_res_super(lockres); | |
732 | ocfs2_generic_bast_func(osb, lockres, level); | |
733 | ||
734 | mlog_exit_void(); | |
735 | } | |
736 | ||
ccd979bd MF |
737 | static void ocfs2_rename_bast_func(void *opaque, |
738 | int level) | |
739 | { | |
740 | struct ocfs2_lock_res *lockres = opaque; | |
741 | struct ocfs2_super *osb; | |
742 | ||
743 | mlog_entry_void(); | |
744 | ||
745 | mlog(0, "Rename BAST fired\n"); | |
746 | ||
747 | BUG_ON(!ocfs2_is_rename_lock(lockres)); | |
748 | ||
749 | osb = ocfs2_lock_res_super(lockres); | |
750 | ocfs2_generic_bast_func(osb, lockres, level); | |
751 | ||
752 | mlog_exit_void(); | |
753 | } | |
754 | ||
d680efe9 MF |
755 | static void ocfs2_dentry_bast_func(void *opaque, int level) |
756 | { | |
757 | struct ocfs2_lock_res *lockres = opaque; | |
758 | struct ocfs2_dentry_lock *dl = lockres->l_priv; | |
759 | struct ocfs2_super *osb = OCFS2_SB(dl->dl_inode->i_sb); | |
760 | ||
761 | mlog(0, "Dentry bast: level: %d, name: %s\n", level, | |
762 | lockres->l_name); | |
763 | ||
764 | ocfs2_generic_bast_func(osb, lockres, level); | |
765 | } | |
766 | ||
ccd979bd MF |
767 | static inline void ocfs2_recover_from_dlm_error(struct ocfs2_lock_res *lockres, |
768 | int convert) | |
769 | { | |
770 | unsigned long flags; | |
771 | ||
772 | mlog_entry_void(); | |
773 | spin_lock_irqsave(&lockres->l_lock, flags); | |
774 | lockres_clear_flags(lockres, OCFS2_LOCK_BUSY); | |
775 | if (convert) | |
776 | lockres->l_action = OCFS2_AST_INVALID; | |
777 | else | |
778 | lockres->l_unlock_action = OCFS2_UNLOCK_INVALID; | |
779 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
780 | ||
781 | wake_up(&lockres->l_event); | |
782 | mlog_exit_void(); | |
783 | } | |
784 | ||
785 | /* Note: If we detect another process working on the lock (i.e., | |
786 | * OCFS2_LOCK_BUSY), we'll bail out returning 0. It's up to the caller | |
787 | * to do the right thing in that case. | |
788 | */ | |
789 | static int ocfs2_lock_create(struct ocfs2_super *osb, | |
790 | struct ocfs2_lock_res *lockres, | |
791 | int level, | |
792 | int dlm_flags) | |
793 | { | |
794 | int ret = 0; | |
795 | enum dlm_status status; | |
796 | unsigned long flags; | |
797 | ||
798 | mlog_entry_void(); | |
799 | ||
800 | mlog(0, "lock %s, level = %d, flags = %d\n", lockres->l_name, level, | |
801 | dlm_flags); | |
802 | ||
803 | spin_lock_irqsave(&lockres->l_lock, flags); | |
804 | if ((lockres->l_flags & OCFS2_LOCK_ATTACHED) || | |
805 | (lockres->l_flags & OCFS2_LOCK_BUSY)) { | |
806 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
807 | goto bail; | |
808 | } | |
809 | ||
810 | lockres->l_action = OCFS2_AST_ATTACH; | |
811 | lockres->l_requested = level; | |
812 | lockres_or_flags(lockres, OCFS2_LOCK_BUSY); | |
813 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
814 | ||
815 | status = dlmlock(osb->dlm, | |
816 | level, | |
817 | &lockres->l_lksb, | |
818 | dlm_flags, | |
819 | lockres->l_name, | |
f0681062 | 820 | OCFS2_LOCK_ID_MAX_LEN - 1, |
e92d57df | 821 | ocfs2_locking_ast, |
ccd979bd MF |
822 | lockres, |
823 | lockres->l_ops->bast); | |
824 | if (status != DLM_NORMAL) { | |
825 | ocfs2_log_dlm_error("dlmlock", status, lockres); | |
826 | ret = -EINVAL; | |
827 | ocfs2_recover_from_dlm_error(lockres, 1); | |
828 | } | |
829 | ||
830 | mlog(0, "lock %s, successfull return from dlmlock\n", lockres->l_name); | |
831 | ||
832 | bail: | |
833 | mlog_exit(ret); | |
834 | return ret; | |
835 | } | |
836 | ||
837 | static inline int ocfs2_check_wait_flag(struct ocfs2_lock_res *lockres, | |
838 | int flag) | |
839 | { | |
840 | unsigned long flags; | |
841 | int ret; | |
842 | ||
843 | spin_lock_irqsave(&lockres->l_lock, flags); | |
844 | ret = lockres->l_flags & flag; | |
845 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
846 | ||
847 | return ret; | |
848 | } | |
849 | ||
850 | static inline void ocfs2_wait_on_busy_lock(struct ocfs2_lock_res *lockres) | |
851 | ||
852 | { | |
853 | wait_event(lockres->l_event, | |
854 | !ocfs2_check_wait_flag(lockres, OCFS2_LOCK_BUSY)); | |
855 | } | |
856 | ||
857 | static inline void ocfs2_wait_on_refreshing_lock(struct ocfs2_lock_res *lockres) | |
858 | ||
859 | { | |
860 | wait_event(lockres->l_event, | |
861 | !ocfs2_check_wait_flag(lockres, OCFS2_LOCK_REFRESHING)); | |
862 | } | |
863 | ||
864 | /* predict what lock level we'll be dropping down to on behalf | |
865 | * of another node, and return true if the currently wanted | |
866 | * level will be compatible with it. */ | |
867 | static inline int ocfs2_may_continue_on_blocked_lock(struct ocfs2_lock_res *lockres, | |
868 | int wanted) | |
869 | { | |
870 | BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BLOCKED)); | |
871 | ||
872 | return wanted <= ocfs2_highest_compat_lock_level(lockres->l_blocking); | |
873 | } | |
874 | ||
875 | static void ocfs2_init_mask_waiter(struct ocfs2_mask_waiter *mw) | |
876 | { | |
877 | INIT_LIST_HEAD(&mw->mw_item); | |
878 | init_completion(&mw->mw_complete); | |
879 | } | |
880 | ||
881 | static int ocfs2_wait_for_mask(struct ocfs2_mask_waiter *mw) | |
882 | { | |
883 | wait_for_completion(&mw->mw_complete); | |
884 | /* Re-arm the completion in case we want to wait on it again */ | |
885 | INIT_COMPLETION(mw->mw_complete); | |
886 | return mw->mw_status; | |
887 | } | |
888 | ||
889 | static void lockres_add_mask_waiter(struct ocfs2_lock_res *lockres, | |
890 | struct ocfs2_mask_waiter *mw, | |
891 | unsigned long mask, | |
892 | unsigned long goal) | |
893 | { | |
894 | BUG_ON(!list_empty(&mw->mw_item)); | |
895 | ||
896 | assert_spin_locked(&lockres->l_lock); | |
897 | ||
898 | list_add_tail(&mw->mw_item, &lockres->l_mask_waiters); | |
899 | mw->mw_mask = mask; | |
900 | mw->mw_goal = goal; | |
901 | } | |
902 | ||
903 | /* returns 0 if the mw that was removed was already satisfied, -EBUSY | |
904 | * if the mask still hadn't reached its goal */ | |
905 | static int lockres_remove_mask_waiter(struct ocfs2_lock_res *lockres, | |
906 | struct ocfs2_mask_waiter *mw) | |
907 | { | |
908 | unsigned long flags; | |
909 | int ret = 0; | |
910 | ||
911 | spin_lock_irqsave(&lockres->l_lock, flags); | |
912 | if (!list_empty(&mw->mw_item)) { | |
913 | if ((lockres->l_flags & mw->mw_mask) != mw->mw_goal) | |
914 | ret = -EBUSY; | |
915 | ||
916 | list_del_init(&mw->mw_item); | |
917 | init_completion(&mw->mw_complete); | |
918 | } | |
919 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
920 | ||
921 | return ret; | |
922 | ||
923 | } | |
924 | ||
925 | static int ocfs2_cluster_lock(struct ocfs2_super *osb, | |
926 | struct ocfs2_lock_res *lockres, | |
927 | int level, | |
928 | int lkm_flags, | |
929 | int arg_flags) | |
930 | { | |
931 | struct ocfs2_mask_waiter mw; | |
932 | enum dlm_status status; | |
933 | int wait, catch_signals = !(osb->s_mount_opt & OCFS2_MOUNT_NOINTR); | |
934 | int ret = 0; /* gcc doesn't realize wait = 1 guarantees ret is set */ | |
935 | unsigned long flags; | |
936 | ||
937 | mlog_entry_void(); | |
938 | ||
939 | ocfs2_init_mask_waiter(&mw); | |
940 | ||
941 | again: | |
942 | wait = 0; | |
943 | ||
944 | if (catch_signals && signal_pending(current)) { | |
945 | ret = -ERESTARTSYS; | |
946 | goto out; | |
947 | } | |
948 | ||
949 | spin_lock_irqsave(&lockres->l_lock, flags); | |
950 | ||
951 | mlog_bug_on_msg(lockres->l_flags & OCFS2_LOCK_FREEING, | |
952 | "Cluster lock called on freeing lockres %s! flags " | |
953 | "0x%lx\n", lockres->l_name, lockres->l_flags); | |
954 | ||
955 | /* We only compare against the currently granted level | |
956 | * here. If the lock is blocked waiting on a downconvert, | |
957 | * we'll get caught below. */ | |
958 | if (lockres->l_flags & OCFS2_LOCK_BUSY && | |
959 | level > lockres->l_level) { | |
960 | /* is someone sitting in dlm_lock? If so, wait on | |
961 | * them. */ | |
962 | lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_BUSY, 0); | |
963 | wait = 1; | |
964 | goto unlock; | |
965 | } | |
966 | ||
967 | if (!(lockres->l_flags & OCFS2_LOCK_ATTACHED)) { | |
968 | /* lock has not been created yet. */ | |
969 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
970 | ||
971 | ret = ocfs2_lock_create(osb, lockres, LKM_NLMODE, 0); | |
972 | if (ret < 0) { | |
973 | mlog_errno(ret); | |
974 | goto out; | |
975 | } | |
976 | goto again; | |
977 | } | |
978 | ||
979 | if (lockres->l_flags & OCFS2_LOCK_BLOCKED && | |
980 | !ocfs2_may_continue_on_blocked_lock(lockres, level)) { | |
981 | /* is the lock is currently blocked on behalf of | |
982 | * another node */ | |
983 | lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_BLOCKED, 0); | |
984 | wait = 1; | |
985 | goto unlock; | |
986 | } | |
987 | ||
988 | if (level > lockres->l_level) { | |
989 | if (lockres->l_action != OCFS2_AST_INVALID) | |
990 | mlog(ML_ERROR, "lockres %s has action %u pending\n", | |
991 | lockres->l_name, lockres->l_action); | |
992 | ||
993 | lockres->l_action = OCFS2_AST_CONVERT; | |
994 | lockres->l_requested = level; | |
995 | lockres_or_flags(lockres, OCFS2_LOCK_BUSY); | |
996 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
997 | ||
998 | BUG_ON(level == LKM_IVMODE); | |
999 | BUG_ON(level == LKM_NLMODE); | |
1000 | ||
1001 | mlog(0, "lock %s, convert from %d to level = %d\n", | |
1002 | lockres->l_name, lockres->l_level, level); | |
1003 | ||
1004 | /* call dlm_lock to upgrade lock now */ | |
1005 | status = dlmlock(osb->dlm, | |
1006 | level, | |
1007 | &lockres->l_lksb, | |
1008 | lkm_flags|LKM_CONVERT|LKM_VALBLK, | |
1009 | lockres->l_name, | |
f0681062 | 1010 | OCFS2_LOCK_ID_MAX_LEN - 1, |
e92d57df | 1011 | ocfs2_locking_ast, |
ccd979bd MF |
1012 | lockres, |
1013 | lockres->l_ops->bast); | |
1014 | if (status != DLM_NORMAL) { | |
1015 | if ((lkm_flags & LKM_NOQUEUE) && | |
1016 | (status == DLM_NOTQUEUED)) | |
1017 | ret = -EAGAIN; | |
1018 | else { | |
1019 | ocfs2_log_dlm_error("dlmlock", status, | |
1020 | lockres); | |
1021 | ret = -EINVAL; | |
1022 | } | |
1023 | ocfs2_recover_from_dlm_error(lockres, 1); | |
1024 | goto out; | |
1025 | } | |
1026 | ||
1027 | mlog(0, "lock %s, successfull return from dlmlock\n", | |
1028 | lockres->l_name); | |
1029 | ||
1030 | /* At this point we've gone inside the dlm and need to | |
1031 | * complete our work regardless. */ | |
1032 | catch_signals = 0; | |
1033 | ||
1034 | /* wait for busy to clear and carry on */ | |
1035 | goto again; | |
1036 | } | |
1037 | ||
1038 | /* Ok, if we get here then we're good to go. */ | |
1039 | ocfs2_inc_holders(lockres, level); | |
1040 | ||
1041 | ret = 0; | |
1042 | unlock: | |
1043 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
1044 | out: | |
1045 | /* | |
1046 | * This is helping work around a lock inversion between the page lock | |
1047 | * and dlm locks. One path holds the page lock while calling aops | |
1048 | * which block acquiring dlm locks. The voting thread holds dlm | |
1049 | * locks while acquiring page locks while down converting data locks. | |
1050 | * This block is helping an aop path notice the inversion and back | |
1051 | * off to unlock its page lock before trying the dlm lock again. | |
1052 | */ | |
1053 | if (wait && arg_flags & OCFS2_LOCK_NONBLOCK && | |
1054 | mw.mw_mask & (OCFS2_LOCK_BUSY|OCFS2_LOCK_BLOCKED)) { | |
1055 | wait = 0; | |
1056 | if (lockres_remove_mask_waiter(lockres, &mw)) | |
1057 | ret = -EAGAIN; | |
1058 | else | |
1059 | goto again; | |
1060 | } | |
1061 | if (wait) { | |
1062 | ret = ocfs2_wait_for_mask(&mw); | |
1063 | if (ret == 0) | |
1064 | goto again; | |
1065 | mlog_errno(ret); | |
1066 | } | |
1067 | ||
1068 | mlog_exit(ret); | |
1069 | return ret; | |
1070 | } | |
1071 | ||
1072 | static void ocfs2_cluster_unlock(struct ocfs2_super *osb, | |
1073 | struct ocfs2_lock_res *lockres, | |
1074 | int level) | |
1075 | { | |
1076 | unsigned long flags; | |
1077 | ||
1078 | mlog_entry_void(); | |
1079 | spin_lock_irqsave(&lockres->l_lock, flags); | |
1080 | ocfs2_dec_holders(lockres, level); | |
1081 | ocfs2_vote_on_unlock(osb, lockres); | |
1082 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
1083 | mlog_exit_void(); | |
1084 | } | |
1085 | ||
d680efe9 MF |
1086 | int ocfs2_create_new_lock(struct ocfs2_super *osb, |
1087 | struct ocfs2_lock_res *lockres, | |
24c19ef4 MF |
1088 | int ex, |
1089 | int local) | |
ccd979bd | 1090 | { |
d680efe9 | 1091 | int level = ex ? LKM_EXMODE : LKM_PRMODE; |
ccd979bd | 1092 | unsigned long flags; |
24c19ef4 | 1093 | int lkm_flags = local ? LKM_LOCAL : 0; |
ccd979bd MF |
1094 | |
1095 | spin_lock_irqsave(&lockres->l_lock, flags); | |
1096 | BUG_ON(lockres->l_flags & OCFS2_LOCK_ATTACHED); | |
1097 | lockres_or_flags(lockres, OCFS2_LOCK_LOCAL); | |
1098 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
1099 | ||
24c19ef4 | 1100 | return ocfs2_lock_create(osb, lockres, level, lkm_flags); |
ccd979bd MF |
1101 | } |
1102 | ||
1103 | /* Grants us an EX lock on the data and metadata resources, skipping | |
1104 | * the normal cluster directory lookup. Use this ONLY on newly created | |
1105 | * inodes which other nodes can't possibly see, and which haven't been | |
1106 | * hashed in the inode hash yet. This can give us a good performance | |
1107 | * increase as it'll skip the network broadcast normally associated | |
1108 | * with creating a new lock resource. */ | |
1109 | int ocfs2_create_new_inode_locks(struct inode *inode) | |
1110 | { | |
1111 | int ret; | |
d680efe9 | 1112 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
ccd979bd MF |
1113 | |
1114 | BUG_ON(!inode); | |
1115 | BUG_ON(!ocfs2_inode_is_new(inode)); | |
1116 | ||
1117 | mlog_entry_void(); | |
1118 | ||
b0697053 | 1119 | mlog(0, "Inode %llu\n", (unsigned long long)OCFS2_I(inode)->ip_blkno); |
ccd979bd MF |
1120 | |
1121 | /* NOTE: That we don't increment any of the holder counts, nor | |
1122 | * do we add anything to a journal handle. Since this is | |
1123 | * supposed to be a new inode which the cluster doesn't know | |
1124 | * about yet, there is no need to. As far as the LVB handling | |
1125 | * is concerned, this is basically like acquiring an EX lock | |
1126 | * on a resource which has an invalid one -- we'll set it | |
1127 | * valid when we release the EX. */ | |
1128 | ||
24c19ef4 | 1129 | ret = ocfs2_create_new_lock(osb, &OCFS2_I(inode)->ip_rw_lockres, 1, 1); |
ccd979bd MF |
1130 | if (ret) { |
1131 | mlog_errno(ret); | |
1132 | goto bail; | |
1133 | } | |
1134 | ||
24c19ef4 MF |
1135 | /* |
1136 | * We don't want to use LKM_LOCAL on a meta data lock as they | |
1137 | * don't use a generation in their lock names. | |
1138 | */ | |
1139 | ret = ocfs2_create_new_lock(osb, &OCFS2_I(inode)->ip_meta_lockres, 1, 0); | |
ccd979bd MF |
1140 | if (ret) { |
1141 | mlog_errno(ret); | |
1142 | goto bail; | |
1143 | } | |
1144 | ||
24c19ef4 | 1145 | ret = ocfs2_create_new_lock(osb, &OCFS2_I(inode)->ip_data_lockres, 1, 1); |
ccd979bd MF |
1146 | if (ret) { |
1147 | mlog_errno(ret); | |
1148 | goto bail; | |
1149 | } | |
1150 | ||
1151 | bail: | |
1152 | mlog_exit(ret); | |
1153 | return ret; | |
1154 | } | |
1155 | ||
1156 | int ocfs2_rw_lock(struct inode *inode, int write) | |
1157 | { | |
1158 | int status, level; | |
1159 | struct ocfs2_lock_res *lockres; | |
1160 | ||
1161 | BUG_ON(!inode); | |
1162 | ||
1163 | mlog_entry_void(); | |
1164 | ||
b0697053 MF |
1165 | mlog(0, "inode %llu take %s RW lock\n", |
1166 | (unsigned long long)OCFS2_I(inode)->ip_blkno, | |
ccd979bd MF |
1167 | write ? "EXMODE" : "PRMODE"); |
1168 | ||
1169 | lockres = &OCFS2_I(inode)->ip_rw_lockres; | |
1170 | ||
1171 | level = write ? LKM_EXMODE : LKM_PRMODE; | |
1172 | ||
1173 | status = ocfs2_cluster_lock(OCFS2_SB(inode->i_sb), lockres, level, 0, | |
1174 | 0); | |
1175 | if (status < 0) | |
1176 | mlog_errno(status); | |
1177 | ||
1178 | mlog_exit(status); | |
1179 | return status; | |
1180 | } | |
1181 | ||
1182 | void ocfs2_rw_unlock(struct inode *inode, int write) | |
1183 | { | |
1184 | int level = write ? LKM_EXMODE : LKM_PRMODE; | |
1185 | struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_rw_lockres; | |
1186 | ||
1187 | mlog_entry_void(); | |
1188 | ||
b0697053 MF |
1189 | mlog(0, "inode %llu drop %s RW lock\n", |
1190 | (unsigned long long)OCFS2_I(inode)->ip_blkno, | |
ccd979bd MF |
1191 | write ? "EXMODE" : "PRMODE"); |
1192 | ||
1193 | ocfs2_cluster_unlock(OCFS2_SB(inode->i_sb), lockres, level); | |
1194 | ||
1195 | mlog_exit_void(); | |
1196 | } | |
1197 | ||
1198 | int ocfs2_data_lock_full(struct inode *inode, | |
1199 | int write, | |
1200 | int arg_flags) | |
1201 | { | |
1202 | int status = 0, level; | |
1203 | struct ocfs2_lock_res *lockres; | |
1204 | ||
1205 | BUG_ON(!inode); | |
1206 | ||
1207 | mlog_entry_void(); | |
1208 | ||
b0697053 MF |
1209 | mlog(0, "inode %llu take %s DATA lock\n", |
1210 | (unsigned long long)OCFS2_I(inode)->ip_blkno, | |
ccd979bd MF |
1211 | write ? "EXMODE" : "PRMODE"); |
1212 | ||
1213 | /* We'll allow faking a readonly data lock for | |
1214 | * rodevices. */ | |
1215 | if (ocfs2_is_hard_readonly(OCFS2_SB(inode->i_sb))) { | |
1216 | if (write) { | |
1217 | status = -EROFS; | |
1218 | mlog_errno(status); | |
1219 | } | |
1220 | goto out; | |
1221 | } | |
1222 | ||
1223 | lockres = &OCFS2_I(inode)->ip_data_lockres; | |
1224 | ||
1225 | level = write ? LKM_EXMODE : LKM_PRMODE; | |
1226 | ||
1227 | status = ocfs2_cluster_lock(OCFS2_SB(inode->i_sb), lockres, level, | |
1228 | 0, arg_flags); | |
1229 | if (status < 0 && status != -EAGAIN) | |
1230 | mlog_errno(status); | |
1231 | ||
1232 | out: | |
1233 | mlog_exit(status); | |
1234 | return status; | |
1235 | } | |
1236 | ||
1237 | /* see ocfs2_meta_lock_with_page() */ | |
1238 | int ocfs2_data_lock_with_page(struct inode *inode, | |
1239 | int write, | |
1240 | struct page *page) | |
1241 | { | |
1242 | int ret; | |
1243 | ||
1244 | ret = ocfs2_data_lock_full(inode, write, OCFS2_LOCK_NONBLOCK); | |
1245 | if (ret == -EAGAIN) { | |
1246 | unlock_page(page); | |
1247 | if (ocfs2_data_lock(inode, write) == 0) | |
1248 | ocfs2_data_unlock(inode, write); | |
1249 | ret = AOP_TRUNCATED_PAGE; | |
1250 | } | |
1251 | ||
1252 | return ret; | |
1253 | } | |
1254 | ||
1255 | static void ocfs2_vote_on_unlock(struct ocfs2_super *osb, | |
1256 | struct ocfs2_lock_res *lockres) | |
1257 | { | |
1258 | int kick = 0; | |
1259 | ||
1260 | mlog_entry_void(); | |
1261 | ||
1262 | /* If we know that another node is waiting on our lock, kick | |
1263 | * the vote thread * pre-emptively when we reach a release | |
1264 | * condition. */ | |
1265 | if (lockres->l_flags & OCFS2_LOCK_BLOCKED) { | |
1266 | switch(lockres->l_blocking) { | |
1267 | case LKM_EXMODE: | |
1268 | if (!lockres->l_ex_holders && !lockres->l_ro_holders) | |
1269 | kick = 1; | |
1270 | break; | |
1271 | case LKM_PRMODE: | |
1272 | if (!lockres->l_ex_holders) | |
1273 | kick = 1; | |
1274 | break; | |
1275 | default: | |
1276 | BUG(); | |
1277 | } | |
1278 | } | |
1279 | ||
1280 | if (kick) | |
1281 | ocfs2_kick_vote_thread(osb); | |
1282 | ||
1283 | mlog_exit_void(); | |
1284 | } | |
1285 | ||
1286 | void ocfs2_data_unlock(struct inode *inode, | |
1287 | int write) | |
1288 | { | |
1289 | int level = write ? LKM_EXMODE : LKM_PRMODE; | |
1290 | struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_data_lockres; | |
1291 | ||
1292 | mlog_entry_void(); | |
1293 | ||
b0697053 MF |
1294 | mlog(0, "inode %llu drop %s DATA lock\n", |
1295 | (unsigned long long)OCFS2_I(inode)->ip_blkno, | |
ccd979bd MF |
1296 | write ? "EXMODE" : "PRMODE"); |
1297 | ||
1298 | if (!ocfs2_is_hard_readonly(OCFS2_SB(inode->i_sb))) | |
1299 | ocfs2_cluster_unlock(OCFS2_SB(inode->i_sb), lockres, level); | |
1300 | ||
1301 | mlog_exit_void(); | |
1302 | } | |
1303 | ||
1304 | #define OCFS2_SEC_BITS 34 | |
1305 | #define OCFS2_SEC_SHIFT (64 - 34) | |
1306 | #define OCFS2_NSEC_MASK ((1ULL << OCFS2_SEC_SHIFT) - 1) | |
1307 | ||
1308 | /* LVB only has room for 64 bits of time here so we pack it for | |
1309 | * now. */ | |
1310 | static u64 ocfs2_pack_timespec(struct timespec *spec) | |
1311 | { | |
1312 | u64 res; | |
1313 | u64 sec = spec->tv_sec; | |
1314 | u32 nsec = spec->tv_nsec; | |
1315 | ||
1316 | res = (sec << OCFS2_SEC_SHIFT) | (nsec & OCFS2_NSEC_MASK); | |
1317 | ||
1318 | return res; | |
1319 | } | |
1320 | ||
1321 | /* Call this with the lockres locked. I am reasonably sure we don't | |
1322 | * need ip_lock in this function as anyone who would be changing those | |
1323 | * values is supposed to be blocked in ocfs2_meta_lock right now. */ | |
1324 | static void __ocfs2_stuff_meta_lvb(struct inode *inode) | |
1325 | { | |
1326 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
1327 | struct ocfs2_lock_res *lockres = &oi->ip_meta_lockres; | |
1328 | struct ocfs2_meta_lvb *lvb; | |
1329 | ||
1330 | mlog_entry_void(); | |
1331 | ||
1332 | lvb = (struct ocfs2_meta_lvb *) lockres->l_lksb.lvb; | |
1333 | ||
24c19ef4 MF |
1334 | /* |
1335 | * Invalidate the LVB of a deleted inode - this way other | |
1336 | * nodes are forced to go to disk and discover the new inode | |
1337 | * status. | |
1338 | */ | |
1339 | if (oi->ip_flags & OCFS2_INODE_DELETED) { | |
1340 | lvb->lvb_version = 0; | |
1341 | goto out; | |
1342 | } | |
1343 | ||
4d3b83f7 | 1344 | lvb->lvb_version = OCFS2_LVB_VERSION; |
ccd979bd MF |
1345 | lvb->lvb_isize = cpu_to_be64(i_size_read(inode)); |
1346 | lvb->lvb_iclusters = cpu_to_be32(oi->ip_clusters); | |
1347 | lvb->lvb_iuid = cpu_to_be32(inode->i_uid); | |
1348 | lvb->lvb_igid = cpu_to_be32(inode->i_gid); | |
1349 | lvb->lvb_imode = cpu_to_be16(inode->i_mode); | |
1350 | lvb->lvb_inlink = cpu_to_be16(inode->i_nlink); | |
1351 | lvb->lvb_iatime_packed = | |
1352 | cpu_to_be64(ocfs2_pack_timespec(&inode->i_atime)); | |
1353 | lvb->lvb_ictime_packed = | |
1354 | cpu_to_be64(ocfs2_pack_timespec(&inode->i_ctime)); | |
1355 | lvb->lvb_imtime_packed = | |
1356 | cpu_to_be64(ocfs2_pack_timespec(&inode->i_mtime)); | |
ca4d147e | 1357 | lvb->lvb_iattr = cpu_to_be32(oi->ip_attr); |
f9e2d82e | 1358 | lvb->lvb_igeneration = cpu_to_be32(inode->i_generation); |
ccd979bd | 1359 | |
24c19ef4 | 1360 | out: |
ccd979bd MF |
1361 | mlog_meta_lvb(0, lockres); |
1362 | ||
1363 | mlog_exit_void(); | |
1364 | } | |
1365 | ||
1366 | static void ocfs2_unpack_timespec(struct timespec *spec, | |
1367 | u64 packed_time) | |
1368 | { | |
1369 | spec->tv_sec = packed_time >> OCFS2_SEC_SHIFT; | |
1370 | spec->tv_nsec = packed_time & OCFS2_NSEC_MASK; | |
1371 | } | |
1372 | ||
1373 | static void ocfs2_refresh_inode_from_lvb(struct inode *inode) | |
1374 | { | |
1375 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
1376 | struct ocfs2_lock_res *lockres = &oi->ip_meta_lockres; | |
1377 | struct ocfs2_meta_lvb *lvb; | |
1378 | ||
1379 | mlog_entry_void(); | |
1380 | ||
1381 | mlog_meta_lvb(0, lockres); | |
1382 | ||
1383 | lvb = (struct ocfs2_meta_lvb *) lockres->l_lksb.lvb; | |
1384 | ||
1385 | /* We're safe here without the lockres lock... */ | |
1386 | spin_lock(&oi->ip_lock); | |
1387 | oi->ip_clusters = be32_to_cpu(lvb->lvb_iclusters); | |
1388 | i_size_write(inode, be64_to_cpu(lvb->lvb_isize)); | |
1389 | ||
ca4d147e HP |
1390 | oi->ip_attr = be32_to_cpu(lvb->lvb_iattr); |
1391 | ocfs2_set_inode_flags(inode); | |
1392 | ||
ccd979bd MF |
1393 | /* fast-symlinks are a special case */ |
1394 | if (S_ISLNK(inode->i_mode) && !oi->ip_clusters) | |
1395 | inode->i_blocks = 0; | |
1396 | else | |
1397 | inode->i_blocks = | |
1398 | ocfs2_align_bytes_to_sectors(i_size_read(inode)); | |
1399 | ||
1400 | inode->i_uid = be32_to_cpu(lvb->lvb_iuid); | |
1401 | inode->i_gid = be32_to_cpu(lvb->lvb_igid); | |
1402 | inode->i_mode = be16_to_cpu(lvb->lvb_imode); | |
1403 | inode->i_nlink = be16_to_cpu(lvb->lvb_inlink); | |
1404 | ocfs2_unpack_timespec(&inode->i_atime, | |
1405 | be64_to_cpu(lvb->lvb_iatime_packed)); | |
1406 | ocfs2_unpack_timespec(&inode->i_mtime, | |
1407 | be64_to_cpu(lvb->lvb_imtime_packed)); | |
1408 | ocfs2_unpack_timespec(&inode->i_ctime, | |
1409 | be64_to_cpu(lvb->lvb_ictime_packed)); | |
1410 | spin_unlock(&oi->ip_lock); | |
1411 | ||
1412 | mlog_exit_void(); | |
1413 | } | |
1414 | ||
f9e2d82e MF |
1415 | static inline int ocfs2_meta_lvb_is_trustable(struct inode *inode, |
1416 | struct ocfs2_lock_res *lockres) | |
ccd979bd MF |
1417 | { |
1418 | struct ocfs2_meta_lvb *lvb = (struct ocfs2_meta_lvb *) lockres->l_lksb.lvb; | |
1419 | ||
f9e2d82e MF |
1420 | if (lvb->lvb_version == OCFS2_LVB_VERSION |
1421 | && be32_to_cpu(lvb->lvb_igeneration) == inode->i_generation) | |
ccd979bd MF |
1422 | return 1; |
1423 | return 0; | |
1424 | } | |
1425 | ||
1426 | /* Determine whether a lock resource needs to be refreshed, and | |
1427 | * arbitrate who gets to refresh it. | |
1428 | * | |
1429 | * 0 means no refresh needed. | |
1430 | * | |
1431 | * > 0 means you need to refresh this and you MUST call | |
1432 | * ocfs2_complete_lock_res_refresh afterwards. */ | |
1433 | static int ocfs2_should_refresh_lock_res(struct ocfs2_lock_res *lockres) | |
1434 | { | |
1435 | unsigned long flags; | |
1436 | int status = 0; | |
1437 | ||
1438 | mlog_entry_void(); | |
1439 | ||
1440 | refresh_check: | |
1441 | spin_lock_irqsave(&lockres->l_lock, flags); | |
1442 | if (!(lockres->l_flags & OCFS2_LOCK_NEEDS_REFRESH)) { | |
1443 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
1444 | goto bail; | |
1445 | } | |
1446 | ||
1447 | if (lockres->l_flags & OCFS2_LOCK_REFRESHING) { | |
1448 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
1449 | ||
1450 | ocfs2_wait_on_refreshing_lock(lockres); | |
1451 | goto refresh_check; | |
1452 | } | |
1453 | ||
1454 | /* Ok, I'll be the one to refresh this lock. */ | |
1455 | lockres_or_flags(lockres, OCFS2_LOCK_REFRESHING); | |
1456 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
1457 | ||
1458 | status = 1; | |
1459 | bail: | |
1460 | mlog_exit(status); | |
1461 | return status; | |
1462 | } | |
1463 | ||
1464 | /* If status is non zero, I'll mark it as not being in refresh | |
1465 | * anymroe, but i won't clear the needs refresh flag. */ | |
1466 | static inline void ocfs2_complete_lock_res_refresh(struct ocfs2_lock_res *lockres, | |
1467 | int status) | |
1468 | { | |
1469 | unsigned long flags; | |
1470 | mlog_entry_void(); | |
1471 | ||
1472 | spin_lock_irqsave(&lockres->l_lock, flags); | |
1473 | lockres_clear_flags(lockres, OCFS2_LOCK_REFRESHING); | |
1474 | if (!status) | |
1475 | lockres_clear_flags(lockres, OCFS2_LOCK_NEEDS_REFRESH); | |
1476 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
1477 | ||
1478 | wake_up(&lockres->l_event); | |
1479 | ||
1480 | mlog_exit_void(); | |
1481 | } | |
1482 | ||
1483 | /* may or may not return a bh if it went to disk. */ | |
1484 | static int ocfs2_meta_lock_update(struct inode *inode, | |
1485 | struct buffer_head **bh) | |
1486 | { | |
1487 | int status = 0; | |
1488 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
1489 | struct ocfs2_lock_res *lockres; | |
1490 | struct ocfs2_dinode *fe; | |
1491 | ||
1492 | mlog_entry_void(); | |
1493 | ||
1494 | spin_lock(&oi->ip_lock); | |
1495 | if (oi->ip_flags & OCFS2_INODE_DELETED) { | |
b0697053 | 1496 | mlog(0, "Orphaned inode %llu was deleted while we " |
ccd979bd | 1497 | "were waiting on a lock. ip_flags = 0x%x\n", |
b0697053 | 1498 | (unsigned long long)oi->ip_blkno, oi->ip_flags); |
ccd979bd MF |
1499 | spin_unlock(&oi->ip_lock); |
1500 | status = -ENOENT; | |
1501 | goto bail; | |
1502 | } | |
1503 | spin_unlock(&oi->ip_lock); | |
1504 | ||
1505 | lockres = &oi->ip_meta_lockres; | |
1506 | ||
1507 | if (!ocfs2_should_refresh_lock_res(lockres)) | |
1508 | goto bail; | |
1509 | ||
1510 | /* This will discard any caching information we might have had | |
1511 | * for the inode metadata. */ | |
1512 | ocfs2_metadata_cache_purge(inode); | |
1513 | ||
1514 | /* will do nothing for inode types that don't use the extent | |
1515 | * map (directories, bitmap files, etc) */ | |
1516 | ocfs2_extent_map_trunc(inode, 0); | |
1517 | ||
f9e2d82e | 1518 | if (ocfs2_meta_lvb_is_trustable(inode, lockres)) { |
b0697053 MF |
1519 | mlog(0, "Trusting LVB on inode %llu\n", |
1520 | (unsigned long long)oi->ip_blkno); | |
ccd979bd MF |
1521 | ocfs2_refresh_inode_from_lvb(inode); |
1522 | } else { | |
1523 | /* Boo, we have to go to disk. */ | |
1524 | /* read bh, cast, ocfs2_refresh_inode */ | |
1525 | status = ocfs2_read_block(OCFS2_SB(inode->i_sb), oi->ip_blkno, | |
1526 | bh, OCFS2_BH_CACHED, inode); | |
1527 | if (status < 0) { | |
1528 | mlog_errno(status); | |
1529 | goto bail_refresh; | |
1530 | } | |
1531 | fe = (struct ocfs2_dinode *) (*bh)->b_data; | |
1532 | ||
1533 | /* This is a good chance to make sure we're not | |
1534 | * locking an invalid object. | |
1535 | * | |
1536 | * We bug on a stale inode here because we checked | |
1537 | * above whether it was wiped from disk. The wiping | |
1538 | * node provides a guarantee that we receive that | |
1539 | * message and can mark the inode before dropping any | |
1540 | * locks associated with it. */ | |
1541 | if (!OCFS2_IS_VALID_DINODE(fe)) { | |
1542 | OCFS2_RO_ON_INVALID_DINODE(inode->i_sb, fe); | |
1543 | status = -EIO; | |
1544 | goto bail_refresh; | |
1545 | } | |
1546 | mlog_bug_on_msg(inode->i_generation != | |
1547 | le32_to_cpu(fe->i_generation), | |
b0697053 | 1548 | "Invalid dinode %llu disk generation: %u " |
ccd979bd | 1549 | "inode->i_generation: %u\n", |
b0697053 MF |
1550 | (unsigned long long)oi->ip_blkno, |
1551 | le32_to_cpu(fe->i_generation), | |
ccd979bd MF |
1552 | inode->i_generation); |
1553 | mlog_bug_on_msg(le64_to_cpu(fe->i_dtime) || | |
1554 | !(fe->i_flags & cpu_to_le32(OCFS2_VALID_FL)), | |
b0697053 MF |
1555 | "Stale dinode %llu dtime: %llu flags: 0x%x\n", |
1556 | (unsigned long long)oi->ip_blkno, | |
1557 | (unsigned long long)le64_to_cpu(fe->i_dtime), | |
ccd979bd MF |
1558 | le32_to_cpu(fe->i_flags)); |
1559 | ||
1560 | ocfs2_refresh_inode(inode, fe); | |
1561 | } | |
1562 | ||
1563 | status = 0; | |
1564 | bail_refresh: | |
1565 | ocfs2_complete_lock_res_refresh(lockres, status); | |
1566 | bail: | |
1567 | mlog_exit(status); | |
1568 | return status; | |
1569 | } | |
1570 | ||
1571 | static int ocfs2_assign_bh(struct inode *inode, | |
1572 | struct buffer_head **ret_bh, | |
1573 | struct buffer_head *passed_bh) | |
1574 | { | |
1575 | int status; | |
1576 | ||
1577 | if (passed_bh) { | |
1578 | /* Ok, the update went to disk for us, use the | |
1579 | * returned bh. */ | |
1580 | *ret_bh = passed_bh; | |
1581 | get_bh(*ret_bh); | |
1582 | ||
1583 | return 0; | |
1584 | } | |
1585 | ||
1586 | status = ocfs2_read_block(OCFS2_SB(inode->i_sb), | |
1587 | OCFS2_I(inode)->ip_blkno, | |
1588 | ret_bh, | |
1589 | OCFS2_BH_CACHED, | |
1590 | inode); | |
1591 | if (status < 0) | |
1592 | mlog_errno(status); | |
1593 | ||
1594 | return status; | |
1595 | } | |
1596 | ||
1597 | /* | |
1598 | * returns < 0 error if the callback will never be called, otherwise | |
1599 | * the result of the lock will be communicated via the callback. | |
1600 | */ | |
1601 | int ocfs2_meta_lock_full(struct inode *inode, | |
1602 | struct ocfs2_journal_handle *handle, | |
1603 | struct buffer_head **ret_bh, | |
1604 | int ex, | |
1605 | int arg_flags) | |
1606 | { | |
1607 | int status, level, dlm_flags, acquired; | |
1608 | struct ocfs2_lock_res *lockres; | |
1609 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
1610 | struct buffer_head *local_bh = NULL; | |
1611 | ||
1612 | BUG_ON(!inode); | |
1613 | ||
1614 | mlog_entry_void(); | |
1615 | ||
b0697053 MF |
1616 | mlog(0, "inode %llu, take %s META lock\n", |
1617 | (unsigned long long)OCFS2_I(inode)->ip_blkno, | |
ccd979bd MF |
1618 | ex ? "EXMODE" : "PRMODE"); |
1619 | ||
1620 | status = 0; | |
1621 | acquired = 0; | |
1622 | /* We'll allow faking a readonly metadata lock for | |
1623 | * rodevices. */ | |
1624 | if (ocfs2_is_hard_readonly(osb)) { | |
1625 | if (ex) | |
1626 | status = -EROFS; | |
1627 | goto bail; | |
1628 | } | |
1629 | ||
1630 | if (!(arg_flags & OCFS2_META_LOCK_RECOVERY)) | |
1631 | wait_event(osb->recovery_event, | |
1632 | ocfs2_node_map_is_empty(osb, &osb->recovery_map)); | |
1633 | ||
1634 | acquired = 0; | |
1635 | lockres = &OCFS2_I(inode)->ip_meta_lockres; | |
1636 | level = ex ? LKM_EXMODE : LKM_PRMODE; | |
1637 | dlm_flags = 0; | |
1638 | if (arg_flags & OCFS2_META_LOCK_NOQUEUE) | |
1639 | dlm_flags |= LKM_NOQUEUE; | |
1640 | ||
1641 | status = ocfs2_cluster_lock(osb, lockres, level, dlm_flags, arg_flags); | |
1642 | if (status < 0) { | |
1643 | if (status != -EAGAIN && status != -EIOCBRETRY) | |
1644 | mlog_errno(status); | |
1645 | goto bail; | |
1646 | } | |
1647 | ||
1648 | /* Notify the error cleanup path to drop the cluster lock. */ | |
1649 | acquired = 1; | |
1650 | ||
1651 | /* We wait twice because a node may have died while we were in | |
1652 | * the lower dlm layers. The second time though, we've | |
1653 | * committed to owning this lock so we don't allow signals to | |
1654 | * abort the operation. */ | |
1655 | if (!(arg_flags & OCFS2_META_LOCK_RECOVERY)) | |
1656 | wait_event(osb->recovery_event, | |
1657 | ocfs2_node_map_is_empty(osb, &osb->recovery_map)); | |
1658 | ||
24c19ef4 MF |
1659 | /* |
1660 | * We only see this flag if we're being called from | |
1661 | * ocfs2_read_locked_inode(). It means we're locking an inode | |
1662 | * which hasn't been populated yet, so clear the refresh flag | |
1663 | * and let the caller handle it. | |
1664 | */ | |
1665 | if (inode->i_state & I_NEW) { | |
1666 | status = 0; | |
1667 | ocfs2_complete_lock_res_refresh(lockres, 0); | |
1668 | goto bail; | |
1669 | } | |
1670 | ||
ccd979bd MF |
1671 | /* This is fun. The caller may want a bh back, or it may |
1672 | * not. ocfs2_meta_lock_update definitely wants one in, but | |
1673 | * may or may not read one, depending on what's in the | |
1674 | * LVB. The result of all of this is that we've *only* gone to | |
1675 | * disk if we have to, so the complexity is worthwhile. */ | |
1676 | status = ocfs2_meta_lock_update(inode, &local_bh); | |
1677 | if (status < 0) { | |
1678 | if (status != -ENOENT) | |
1679 | mlog_errno(status); | |
1680 | goto bail; | |
1681 | } | |
1682 | ||
1683 | if (ret_bh) { | |
1684 | status = ocfs2_assign_bh(inode, ret_bh, local_bh); | |
1685 | if (status < 0) { | |
1686 | mlog_errno(status); | |
1687 | goto bail; | |
1688 | } | |
1689 | } | |
1690 | ||
1691 | if (handle) { | |
1692 | status = ocfs2_handle_add_lock(handle, inode); | |
1693 | if (status < 0) | |
1694 | mlog_errno(status); | |
1695 | } | |
1696 | ||
1697 | bail: | |
1698 | if (status < 0) { | |
1699 | if (ret_bh && (*ret_bh)) { | |
1700 | brelse(*ret_bh); | |
1701 | *ret_bh = NULL; | |
1702 | } | |
1703 | if (acquired) | |
1704 | ocfs2_meta_unlock(inode, ex); | |
1705 | } | |
1706 | ||
1707 | if (local_bh) | |
1708 | brelse(local_bh); | |
1709 | ||
1710 | mlog_exit(status); | |
1711 | return status; | |
1712 | } | |
1713 | ||
1714 | /* | |
1715 | * This is working around a lock inversion between tasks acquiring DLM locks | |
1716 | * while holding a page lock and the vote thread which blocks dlm lock acquiry | |
1717 | * while acquiring page locks. | |
1718 | * | |
1719 | * ** These _with_page variantes are only intended to be called from aop | |
1720 | * methods that hold page locks and return a very specific *positive* error | |
1721 | * code that aop methods pass up to the VFS -- test for errors with != 0. ** | |
1722 | * | |
1723 | * The DLM is called such that it returns -EAGAIN if it would have blocked | |
1724 | * waiting for the vote thread. In that case we unlock our page so the vote | |
1725 | * thread can make progress. Once we've done this we have to return | |
1726 | * AOP_TRUNCATED_PAGE so the aop method that called us can bubble that back up | |
1727 | * into the VFS who will then immediately retry the aop call. | |
1728 | * | |
1729 | * We do a blocking lock and immediate unlock before returning, though, so that | |
1730 | * the lock has a great chance of being cached on this node by the time the VFS | |
1731 | * calls back to retry the aop. This has a potential to livelock as nodes | |
1732 | * ping locks back and forth, but that's a risk we're willing to take to avoid | |
1733 | * the lock inversion simply. | |
1734 | */ | |
1735 | int ocfs2_meta_lock_with_page(struct inode *inode, | |
1736 | struct ocfs2_journal_handle *handle, | |
1737 | struct buffer_head **ret_bh, | |
1738 | int ex, | |
1739 | struct page *page) | |
1740 | { | |
1741 | int ret; | |
1742 | ||
1743 | ret = ocfs2_meta_lock_full(inode, handle, ret_bh, ex, | |
1744 | OCFS2_LOCK_NONBLOCK); | |
1745 | if (ret == -EAGAIN) { | |
1746 | unlock_page(page); | |
1747 | if (ocfs2_meta_lock(inode, handle, ret_bh, ex) == 0) | |
1748 | ocfs2_meta_unlock(inode, ex); | |
1749 | ret = AOP_TRUNCATED_PAGE; | |
1750 | } | |
1751 | ||
1752 | return ret; | |
1753 | } | |
1754 | ||
1755 | void ocfs2_meta_unlock(struct inode *inode, | |
1756 | int ex) | |
1757 | { | |
1758 | int level = ex ? LKM_EXMODE : LKM_PRMODE; | |
1759 | struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_meta_lockres; | |
1760 | ||
1761 | mlog_entry_void(); | |
1762 | ||
b0697053 MF |
1763 | mlog(0, "inode %llu drop %s META lock\n", |
1764 | (unsigned long long)OCFS2_I(inode)->ip_blkno, | |
ccd979bd MF |
1765 | ex ? "EXMODE" : "PRMODE"); |
1766 | ||
1767 | if (!ocfs2_is_hard_readonly(OCFS2_SB(inode->i_sb))) | |
1768 | ocfs2_cluster_unlock(OCFS2_SB(inode->i_sb), lockres, level); | |
1769 | ||
1770 | mlog_exit_void(); | |
1771 | } | |
1772 | ||
1773 | int ocfs2_super_lock(struct ocfs2_super *osb, | |
1774 | int ex) | |
1775 | { | |
1776 | int status; | |
1777 | int level = ex ? LKM_EXMODE : LKM_PRMODE; | |
1778 | struct ocfs2_lock_res *lockres = &osb->osb_super_lockres; | |
1779 | struct buffer_head *bh; | |
1780 | struct ocfs2_slot_info *si = osb->slot_info; | |
1781 | ||
1782 | mlog_entry_void(); | |
1783 | ||
1784 | if (ocfs2_is_hard_readonly(osb)) | |
1785 | return -EROFS; | |
1786 | ||
1787 | status = ocfs2_cluster_lock(osb, lockres, level, 0, 0); | |
1788 | if (status < 0) { | |
1789 | mlog_errno(status); | |
1790 | goto bail; | |
1791 | } | |
1792 | ||
1793 | /* The super block lock path is really in the best position to | |
1794 | * know when resources covered by the lock need to be | |
1795 | * refreshed, so we do it here. Of course, making sense of | |
1796 | * everything is up to the caller :) */ | |
1797 | status = ocfs2_should_refresh_lock_res(lockres); | |
1798 | if (status < 0) { | |
1799 | mlog_errno(status); | |
1800 | goto bail; | |
1801 | } | |
1802 | if (status) { | |
1803 | bh = si->si_bh; | |
1804 | status = ocfs2_read_block(osb, bh->b_blocknr, &bh, 0, | |
1805 | si->si_inode); | |
1806 | if (status == 0) | |
1807 | ocfs2_update_slot_info(si); | |
1808 | ||
1809 | ocfs2_complete_lock_res_refresh(lockres, status); | |
1810 | ||
1811 | if (status < 0) | |
1812 | mlog_errno(status); | |
1813 | } | |
1814 | bail: | |
1815 | mlog_exit(status); | |
1816 | return status; | |
1817 | } | |
1818 | ||
1819 | void ocfs2_super_unlock(struct ocfs2_super *osb, | |
1820 | int ex) | |
1821 | { | |
1822 | int level = ex ? LKM_EXMODE : LKM_PRMODE; | |
1823 | struct ocfs2_lock_res *lockres = &osb->osb_super_lockres; | |
1824 | ||
1825 | ocfs2_cluster_unlock(osb, lockres, level); | |
1826 | } | |
1827 | ||
1828 | int ocfs2_rename_lock(struct ocfs2_super *osb) | |
1829 | { | |
1830 | int status; | |
1831 | struct ocfs2_lock_res *lockres = &osb->osb_rename_lockres; | |
1832 | ||
1833 | if (ocfs2_is_hard_readonly(osb)) | |
1834 | return -EROFS; | |
1835 | ||
1836 | status = ocfs2_cluster_lock(osb, lockres, LKM_EXMODE, 0, 0); | |
1837 | if (status < 0) | |
1838 | mlog_errno(status); | |
1839 | ||
1840 | return status; | |
1841 | } | |
1842 | ||
1843 | void ocfs2_rename_unlock(struct ocfs2_super *osb) | |
1844 | { | |
1845 | struct ocfs2_lock_res *lockres = &osb->osb_rename_lockres; | |
1846 | ||
1847 | ocfs2_cluster_unlock(osb, lockres, LKM_EXMODE); | |
1848 | } | |
1849 | ||
d680efe9 MF |
1850 | int ocfs2_dentry_lock(struct dentry *dentry, int ex) |
1851 | { | |
1852 | int ret; | |
1853 | int level = ex ? LKM_EXMODE : LKM_PRMODE; | |
1854 | struct ocfs2_dentry_lock *dl = dentry->d_fsdata; | |
1855 | struct ocfs2_super *osb = OCFS2_SB(dentry->d_sb); | |
1856 | ||
1857 | BUG_ON(!dl); | |
1858 | ||
1859 | if (ocfs2_is_hard_readonly(osb)) | |
1860 | return -EROFS; | |
1861 | ||
1862 | ret = ocfs2_cluster_lock(osb, &dl->dl_lockres, level, 0, 0); | |
1863 | if (ret < 0) | |
1864 | mlog_errno(ret); | |
1865 | ||
1866 | return ret; | |
1867 | } | |
1868 | ||
1869 | void ocfs2_dentry_unlock(struct dentry *dentry, int ex) | |
1870 | { | |
1871 | int level = ex ? LKM_EXMODE : LKM_PRMODE; | |
1872 | struct ocfs2_dentry_lock *dl = dentry->d_fsdata; | |
1873 | struct ocfs2_super *osb = OCFS2_SB(dentry->d_sb); | |
1874 | ||
1875 | ocfs2_cluster_unlock(osb, &dl->dl_lockres, level); | |
1876 | } | |
1877 | ||
ccd979bd MF |
1878 | /* Reference counting of the dlm debug structure. We want this because |
1879 | * open references on the debug inodes can live on after a mount, so | |
1880 | * we can't rely on the ocfs2_super to always exist. */ | |
1881 | static void ocfs2_dlm_debug_free(struct kref *kref) | |
1882 | { | |
1883 | struct ocfs2_dlm_debug *dlm_debug; | |
1884 | ||
1885 | dlm_debug = container_of(kref, struct ocfs2_dlm_debug, d_refcnt); | |
1886 | ||
1887 | kfree(dlm_debug); | |
1888 | } | |
1889 | ||
1890 | void ocfs2_put_dlm_debug(struct ocfs2_dlm_debug *dlm_debug) | |
1891 | { | |
1892 | if (dlm_debug) | |
1893 | kref_put(&dlm_debug->d_refcnt, ocfs2_dlm_debug_free); | |
1894 | } | |
1895 | ||
1896 | static void ocfs2_get_dlm_debug(struct ocfs2_dlm_debug *debug) | |
1897 | { | |
1898 | kref_get(&debug->d_refcnt); | |
1899 | } | |
1900 | ||
1901 | struct ocfs2_dlm_debug *ocfs2_new_dlm_debug(void) | |
1902 | { | |
1903 | struct ocfs2_dlm_debug *dlm_debug; | |
1904 | ||
1905 | dlm_debug = kmalloc(sizeof(struct ocfs2_dlm_debug), GFP_KERNEL); | |
1906 | if (!dlm_debug) { | |
1907 | mlog_errno(-ENOMEM); | |
1908 | goto out; | |
1909 | } | |
1910 | ||
1911 | kref_init(&dlm_debug->d_refcnt); | |
1912 | INIT_LIST_HEAD(&dlm_debug->d_lockres_tracking); | |
1913 | dlm_debug->d_locking_state = NULL; | |
1914 | out: | |
1915 | return dlm_debug; | |
1916 | } | |
1917 | ||
1918 | /* Access to this is arbitrated for us via seq_file->sem. */ | |
1919 | struct ocfs2_dlm_seq_priv { | |
1920 | struct ocfs2_dlm_debug *p_dlm_debug; | |
1921 | struct ocfs2_lock_res p_iter_res; | |
1922 | struct ocfs2_lock_res p_tmp_res; | |
1923 | }; | |
1924 | ||
1925 | static struct ocfs2_lock_res *ocfs2_dlm_next_res(struct ocfs2_lock_res *start, | |
1926 | struct ocfs2_dlm_seq_priv *priv) | |
1927 | { | |
1928 | struct ocfs2_lock_res *iter, *ret = NULL; | |
1929 | struct ocfs2_dlm_debug *dlm_debug = priv->p_dlm_debug; | |
1930 | ||
1931 | assert_spin_locked(&ocfs2_dlm_tracking_lock); | |
1932 | ||
1933 | list_for_each_entry(iter, &start->l_debug_list, l_debug_list) { | |
1934 | /* discover the head of the list */ | |
1935 | if (&iter->l_debug_list == &dlm_debug->d_lockres_tracking) { | |
1936 | mlog(0, "End of list found, %p\n", ret); | |
1937 | break; | |
1938 | } | |
1939 | ||
1940 | /* We track our "dummy" iteration lockres' by a NULL | |
1941 | * l_ops field. */ | |
1942 | if (iter->l_ops != NULL) { | |
1943 | ret = iter; | |
1944 | break; | |
1945 | } | |
1946 | } | |
1947 | ||
1948 | return ret; | |
1949 | } | |
1950 | ||
1951 | static void *ocfs2_dlm_seq_start(struct seq_file *m, loff_t *pos) | |
1952 | { | |
1953 | struct ocfs2_dlm_seq_priv *priv = m->private; | |
1954 | struct ocfs2_lock_res *iter; | |
1955 | ||
1956 | spin_lock(&ocfs2_dlm_tracking_lock); | |
1957 | iter = ocfs2_dlm_next_res(&priv->p_iter_res, priv); | |
1958 | if (iter) { | |
1959 | /* Since lockres' have the lifetime of their container | |
1960 | * (which can be inodes, ocfs2_supers, etc) we want to | |
1961 | * copy this out to a temporary lockres while still | |
1962 | * under the spinlock. Obviously after this we can't | |
1963 | * trust any pointers on the copy returned, but that's | |
1964 | * ok as the information we want isn't typically held | |
1965 | * in them. */ | |
1966 | priv->p_tmp_res = *iter; | |
1967 | iter = &priv->p_tmp_res; | |
1968 | } | |
1969 | spin_unlock(&ocfs2_dlm_tracking_lock); | |
1970 | ||
1971 | return iter; | |
1972 | } | |
1973 | ||
1974 | static void ocfs2_dlm_seq_stop(struct seq_file *m, void *v) | |
1975 | { | |
1976 | } | |
1977 | ||
1978 | static void *ocfs2_dlm_seq_next(struct seq_file *m, void *v, loff_t *pos) | |
1979 | { | |
1980 | struct ocfs2_dlm_seq_priv *priv = m->private; | |
1981 | struct ocfs2_lock_res *iter = v; | |
1982 | struct ocfs2_lock_res *dummy = &priv->p_iter_res; | |
1983 | ||
1984 | spin_lock(&ocfs2_dlm_tracking_lock); | |
1985 | iter = ocfs2_dlm_next_res(iter, priv); | |
1986 | list_del_init(&dummy->l_debug_list); | |
1987 | if (iter) { | |
1988 | list_add(&dummy->l_debug_list, &iter->l_debug_list); | |
1989 | priv->p_tmp_res = *iter; | |
1990 | iter = &priv->p_tmp_res; | |
1991 | } | |
1992 | spin_unlock(&ocfs2_dlm_tracking_lock); | |
1993 | ||
1994 | return iter; | |
1995 | } | |
1996 | ||
1997 | /* So that debugfs.ocfs2 can determine which format is being used */ | |
1998 | #define OCFS2_DLM_DEBUG_STR_VERSION 1 | |
1999 | static int ocfs2_dlm_seq_show(struct seq_file *m, void *v) | |
2000 | { | |
2001 | int i; | |
2002 | char *lvb; | |
2003 | struct ocfs2_lock_res *lockres = v; | |
2004 | ||
2005 | if (!lockres) | |
2006 | return -EINVAL; | |
2007 | ||
d680efe9 MF |
2008 | seq_printf(m, "0x%x\t", OCFS2_DLM_DEBUG_STR_VERSION); |
2009 | ||
2010 | if (lockres->l_type == OCFS2_LOCK_TYPE_DENTRY) | |
2011 | seq_printf(m, "%.*s%08x\t", OCFS2_DENTRY_LOCK_INO_START - 1, | |
2012 | lockres->l_name, | |
2013 | (unsigned int)ocfs2_get_dentry_lock_ino(lockres)); | |
2014 | else | |
2015 | seq_printf(m, "%.*s\t", OCFS2_LOCK_ID_MAX_LEN, lockres->l_name); | |
2016 | ||
2017 | seq_printf(m, "%d\t" | |
ccd979bd MF |
2018 | "0x%lx\t" |
2019 | "0x%x\t" | |
2020 | "0x%x\t" | |
2021 | "%u\t" | |
2022 | "%u\t" | |
2023 | "%d\t" | |
2024 | "%d\t", | |
ccd979bd MF |
2025 | lockres->l_level, |
2026 | lockres->l_flags, | |
2027 | lockres->l_action, | |
2028 | lockres->l_unlock_action, | |
2029 | lockres->l_ro_holders, | |
2030 | lockres->l_ex_holders, | |
2031 | lockres->l_requested, | |
2032 | lockres->l_blocking); | |
2033 | ||
2034 | /* Dump the raw LVB */ | |
2035 | lvb = lockres->l_lksb.lvb; | |
2036 | for(i = 0; i < DLM_LVB_LEN; i++) | |
2037 | seq_printf(m, "0x%x\t", lvb[i]); | |
2038 | ||
2039 | /* End the line */ | |
2040 | seq_printf(m, "\n"); | |
2041 | return 0; | |
2042 | } | |
2043 | ||
2044 | static struct seq_operations ocfs2_dlm_seq_ops = { | |
2045 | .start = ocfs2_dlm_seq_start, | |
2046 | .stop = ocfs2_dlm_seq_stop, | |
2047 | .next = ocfs2_dlm_seq_next, | |
2048 | .show = ocfs2_dlm_seq_show, | |
2049 | }; | |
2050 | ||
2051 | static int ocfs2_dlm_debug_release(struct inode *inode, struct file *file) | |
2052 | { | |
2053 | struct seq_file *seq = (struct seq_file *) file->private_data; | |
2054 | struct ocfs2_dlm_seq_priv *priv = seq->private; | |
2055 | struct ocfs2_lock_res *res = &priv->p_iter_res; | |
2056 | ||
2057 | ocfs2_remove_lockres_tracking(res); | |
2058 | ocfs2_put_dlm_debug(priv->p_dlm_debug); | |
2059 | return seq_release_private(inode, file); | |
2060 | } | |
2061 | ||
2062 | static int ocfs2_dlm_debug_open(struct inode *inode, struct file *file) | |
2063 | { | |
2064 | int ret; | |
2065 | struct ocfs2_dlm_seq_priv *priv; | |
2066 | struct seq_file *seq; | |
2067 | struct ocfs2_super *osb; | |
2068 | ||
2069 | priv = kzalloc(sizeof(struct ocfs2_dlm_seq_priv), GFP_KERNEL); | |
2070 | if (!priv) { | |
2071 | ret = -ENOMEM; | |
2072 | mlog_errno(ret); | |
2073 | goto out; | |
2074 | } | |
2075 | osb = (struct ocfs2_super *) inode->u.generic_ip; | |
2076 | ocfs2_get_dlm_debug(osb->osb_dlm_debug); | |
2077 | priv->p_dlm_debug = osb->osb_dlm_debug; | |
2078 | INIT_LIST_HEAD(&priv->p_iter_res.l_debug_list); | |
2079 | ||
2080 | ret = seq_open(file, &ocfs2_dlm_seq_ops); | |
2081 | if (ret) { | |
2082 | kfree(priv); | |
2083 | mlog_errno(ret); | |
2084 | goto out; | |
2085 | } | |
2086 | ||
2087 | seq = (struct seq_file *) file->private_data; | |
2088 | seq->private = priv; | |
2089 | ||
2090 | ocfs2_add_lockres_tracking(&priv->p_iter_res, | |
2091 | priv->p_dlm_debug); | |
2092 | ||
2093 | out: | |
2094 | return ret; | |
2095 | } | |
2096 | ||
4b6f5d20 | 2097 | static const struct file_operations ocfs2_dlm_debug_fops = { |
ccd979bd MF |
2098 | .open = ocfs2_dlm_debug_open, |
2099 | .release = ocfs2_dlm_debug_release, | |
2100 | .read = seq_read, | |
2101 | .llseek = seq_lseek, | |
2102 | }; | |
2103 | ||
2104 | static int ocfs2_dlm_init_debug(struct ocfs2_super *osb) | |
2105 | { | |
2106 | int ret = 0; | |
2107 | struct ocfs2_dlm_debug *dlm_debug = osb->osb_dlm_debug; | |
2108 | ||
2109 | dlm_debug->d_locking_state = debugfs_create_file("locking_state", | |
2110 | S_IFREG|S_IRUSR, | |
2111 | osb->osb_debug_root, | |
2112 | osb, | |
2113 | &ocfs2_dlm_debug_fops); | |
2114 | if (!dlm_debug->d_locking_state) { | |
2115 | ret = -EINVAL; | |
2116 | mlog(ML_ERROR, | |
2117 | "Unable to create locking state debugfs file.\n"); | |
2118 | goto out; | |
2119 | } | |
2120 | ||
2121 | ocfs2_get_dlm_debug(dlm_debug); | |
2122 | out: | |
2123 | return ret; | |
2124 | } | |
2125 | ||
2126 | static void ocfs2_dlm_shutdown_debug(struct ocfs2_super *osb) | |
2127 | { | |
2128 | struct ocfs2_dlm_debug *dlm_debug = osb->osb_dlm_debug; | |
2129 | ||
2130 | if (dlm_debug) { | |
2131 | debugfs_remove(dlm_debug->d_locking_state); | |
2132 | ocfs2_put_dlm_debug(dlm_debug); | |
2133 | } | |
2134 | } | |
2135 | ||
2136 | int ocfs2_dlm_init(struct ocfs2_super *osb) | |
2137 | { | |
2138 | int status; | |
2139 | u32 dlm_key; | |
2140 | struct dlm_ctxt *dlm; | |
2141 | ||
2142 | mlog_entry_void(); | |
2143 | ||
2144 | status = ocfs2_dlm_init_debug(osb); | |
2145 | if (status < 0) { | |
2146 | mlog_errno(status); | |
2147 | goto bail; | |
2148 | } | |
2149 | ||
2150 | /* launch vote thread */ | |
78427043 | 2151 | osb->vote_task = kthread_run(ocfs2_vote_thread, osb, "ocfs2vote"); |
ccd979bd MF |
2152 | if (IS_ERR(osb->vote_task)) { |
2153 | status = PTR_ERR(osb->vote_task); | |
2154 | osb->vote_task = NULL; | |
2155 | mlog_errno(status); | |
2156 | goto bail; | |
2157 | } | |
2158 | ||
2159 | /* used by the dlm code to make message headers unique, each | |
2160 | * node in this domain must agree on this. */ | |
2161 | dlm_key = crc32_le(0, osb->uuid_str, strlen(osb->uuid_str)); | |
2162 | ||
2163 | /* for now, uuid == domain */ | |
2164 | dlm = dlm_register_domain(osb->uuid_str, dlm_key); | |
2165 | if (IS_ERR(dlm)) { | |
2166 | status = PTR_ERR(dlm); | |
2167 | mlog_errno(status); | |
2168 | goto bail; | |
2169 | } | |
2170 | ||
2171 | ocfs2_super_lock_res_init(&osb->osb_super_lockres, osb); | |
2172 | ocfs2_rename_lock_res_init(&osb->osb_rename_lockres, osb); | |
2173 | ||
2174 | dlm_register_eviction_cb(dlm, &osb->osb_eviction_cb); | |
2175 | ||
2176 | osb->dlm = dlm; | |
2177 | ||
2178 | status = 0; | |
2179 | bail: | |
2180 | if (status < 0) { | |
2181 | ocfs2_dlm_shutdown_debug(osb); | |
2182 | if (osb->vote_task) | |
2183 | kthread_stop(osb->vote_task); | |
2184 | } | |
2185 | ||
2186 | mlog_exit(status); | |
2187 | return status; | |
2188 | } | |
2189 | ||
2190 | void ocfs2_dlm_shutdown(struct ocfs2_super *osb) | |
2191 | { | |
2192 | mlog_entry_void(); | |
2193 | ||
2194 | dlm_unregister_eviction_cb(&osb->osb_eviction_cb); | |
2195 | ||
2196 | ocfs2_drop_osb_locks(osb); | |
2197 | ||
2198 | if (osb->vote_task) { | |
2199 | kthread_stop(osb->vote_task); | |
2200 | osb->vote_task = NULL; | |
2201 | } | |
2202 | ||
2203 | ocfs2_lock_res_free(&osb->osb_super_lockres); | |
2204 | ocfs2_lock_res_free(&osb->osb_rename_lockres); | |
2205 | ||
2206 | dlm_unregister_domain(osb->dlm); | |
2207 | osb->dlm = NULL; | |
2208 | ||
2209 | ocfs2_dlm_shutdown_debug(osb); | |
2210 | ||
2211 | mlog_exit_void(); | |
2212 | } | |
2213 | ||
2a45f2d1 | 2214 | static void ocfs2_unlock_ast(void *opaque, enum dlm_status status) |
ccd979bd MF |
2215 | { |
2216 | struct ocfs2_lock_res *lockres = opaque; | |
2217 | unsigned long flags; | |
2218 | ||
2219 | mlog_entry_void(); | |
2220 | ||
2221 | mlog(0, "UNLOCK AST called on lock %s, action = %d\n", lockres->l_name, | |
2222 | lockres->l_unlock_action); | |
2223 | ||
2224 | spin_lock_irqsave(&lockres->l_lock, flags); | |
2225 | /* We tried to cancel a convert request, but it was already | |
2226 | * granted. All we want to do here is clear our unlock | |
2227 | * state. The wake_up call done at the bottom is redundant | |
2228 | * (ocfs2_prepare_cancel_convert doesn't sleep on this) but doesn't | |
2229 | * hurt anything anyway */ | |
2230 | if (status == DLM_CANCELGRANT && | |
2231 | lockres->l_unlock_action == OCFS2_UNLOCK_CANCEL_CONVERT) { | |
2232 | mlog(0, "Got cancelgrant for %s\n", lockres->l_name); | |
2233 | ||
2234 | /* We don't clear the busy flag in this case as it | |
2235 | * should have been cleared by the ast which the dlm | |
2236 | * has called. */ | |
2237 | goto complete_unlock; | |
2238 | } | |
2239 | ||
2240 | if (status != DLM_NORMAL) { | |
2241 | mlog(ML_ERROR, "Dlm passes status %d for lock %s, " | |
2242 | "unlock_action %d\n", status, lockres->l_name, | |
2243 | lockres->l_unlock_action); | |
2244 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
2245 | return; | |
2246 | } | |
2247 | ||
2248 | switch(lockres->l_unlock_action) { | |
2249 | case OCFS2_UNLOCK_CANCEL_CONVERT: | |
2250 | mlog(0, "Cancel convert success for %s\n", lockres->l_name); | |
2251 | lockres->l_action = OCFS2_AST_INVALID; | |
2252 | break; | |
2253 | case OCFS2_UNLOCK_DROP_LOCK: | |
2254 | lockres->l_level = LKM_IVMODE; | |
2255 | break; | |
2256 | default: | |
2257 | BUG(); | |
2258 | } | |
2259 | ||
2260 | lockres_clear_flags(lockres, OCFS2_LOCK_BUSY); | |
2261 | complete_unlock: | |
2262 | lockres->l_unlock_action = OCFS2_UNLOCK_INVALID; | |
2263 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
2264 | ||
2265 | wake_up(&lockres->l_event); | |
2266 | ||
2267 | mlog_exit_void(); | |
2268 | } | |
2269 | ||
2270 | typedef void (ocfs2_pre_drop_cb_t)(struct ocfs2_lock_res *, void *); | |
2271 | ||
2272 | struct drop_lock_cb { | |
2273 | ocfs2_pre_drop_cb_t *drop_func; | |
2274 | void *drop_data; | |
2275 | }; | |
2276 | ||
2277 | static int ocfs2_drop_lock(struct ocfs2_super *osb, | |
2278 | struct ocfs2_lock_res *lockres, | |
2279 | struct drop_lock_cb *dcb) | |
2280 | { | |
2281 | enum dlm_status status; | |
2282 | unsigned long flags; | |
2283 | ||
2284 | /* We didn't get anywhere near actually using this lockres. */ | |
2285 | if (!(lockres->l_flags & OCFS2_LOCK_INITIALIZED)) | |
2286 | goto out; | |
2287 | ||
2288 | spin_lock_irqsave(&lockres->l_lock, flags); | |
2289 | ||
2290 | mlog_bug_on_msg(!(lockres->l_flags & OCFS2_LOCK_FREEING), | |
2291 | "lockres %s, flags 0x%lx\n", | |
2292 | lockres->l_name, lockres->l_flags); | |
2293 | ||
2294 | while (lockres->l_flags & OCFS2_LOCK_BUSY) { | |
2295 | mlog(0, "waiting on busy lock \"%s\": flags = %lx, action = " | |
2296 | "%u, unlock_action = %u\n", | |
2297 | lockres->l_name, lockres->l_flags, lockres->l_action, | |
2298 | lockres->l_unlock_action); | |
2299 | ||
2300 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
2301 | ||
2302 | /* XXX: Today we just wait on any busy | |
2303 | * locks... Perhaps we need to cancel converts in the | |
2304 | * future? */ | |
2305 | ocfs2_wait_on_busy_lock(lockres); | |
2306 | ||
2307 | spin_lock_irqsave(&lockres->l_lock, flags); | |
2308 | } | |
2309 | ||
2310 | if (dcb) | |
2311 | dcb->drop_func(lockres, dcb->drop_data); | |
2312 | ||
2313 | if (lockres->l_flags & OCFS2_LOCK_BUSY) | |
2314 | mlog(ML_ERROR, "destroying busy lock: \"%s\"\n", | |
2315 | lockres->l_name); | |
2316 | if (lockres->l_flags & OCFS2_LOCK_BLOCKED) | |
2317 | mlog(0, "destroying blocked lock: \"%s\"\n", lockres->l_name); | |
2318 | ||
2319 | if (!(lockres->l_flags & OCFS2_LOCK_ATTACHED)) { | |
2320 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
2321 | goto out; | |
2322 | } | |
2323 | ||
2324 | lockres_clear_flags(lockres, OCFS2_LOCK_ATTACHED); | |
2325 | ||
2326 | /* make sure we never get here while waiting for an ast to | |
2327 | * fire. */ | |
2328 | BUG_ON(lockres->l_action != OCFS2_AST_INVALID); | |
2329 | ||
2330 | /* is this necessary? */ | |
2331 | lockres_or_flags(lockres, OCFS2_LOCK_BUSY); | |
2332 | lockres->l_unlock_action = OCFS2_UNLOCK_DROP_LOCK; | |
2333 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
2334 | ||
2335 | mlog(0, "lock %s\n", lockres->l_name); | |
2336 | ||
2337 | status = dlmunlock(osb->dlm, &lockres->l_lksb, LKM_VALBLK, | |
2a45f2d1 | 2338 | ocfs2_unlock_ast, lockres); |
ccd979bd MF |
2339 | if (status != DLM_NORMAL) { |
2340 | ocfs2_log_dlm_error("dlmunlock", status, lockres); | |
2341 | mlog(ML_ERROR, "lockres flags: %lu\n", lockres->l_flags); | |
2342 | dlm_print_one_lock(lockres->l_lksb.lockid); | |
2343 | BUG(); | |
2344 | } | |
2345 | mlog(0, "lock %s, successfull return from dlmunlock\n", | |
2346 | lockres->l_name); | |
2347 | ||
2348 | ocfs2_wait_on_busy_lock(lockres); | |
2349 | out: | |
2350 | mlog_exit(0); | |
2351 | return 0; | |
2352 | } | |
2353 | ||
2354 | /* Mark the lockres as being dropped. It will no longer be | |
2355 | * queued if blocking, but we still may have to wait on it | |
2356 | * being dequeued from the vote thread before we can consider | |
2357 | * it safe to drop. | |
2358 | * | |
2359 | * You can *not* attempt to call cluster_lock on this lockres anymore. */ | |
2360 | void ocfs2_mark_lockres_freeing(struct ocfs2_lock_res *lockres) | |
2361 | { | |
2362 | int status; | |
2363 | struct ocfs2_mask_waiter mw; | |
2364 | unsigned long flags; | |
2365 | ||
2366 | ocfs2_init_mask_waiter(&mw); | |
2367 | ||
2368 | spin_lock_irqsave(&lockres->l_lock, flags); | |
2369 | lockres->l_flags |= OCFS2_LOCK_FREEING; | |
2370 | while (lockres->l_flags & OCFS2_LOCK_QUEUED) { | |
2371 | lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_QUEUED, 0); | |
2372 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
2373 | ||
2374 | mlog(0, "Waiting on lockres %s\n", lockres->l_name); | |
2375 | ||
2376 | status = ocfs2_wait_for_mask(&mw); | |
2377 | if (status) | |
2378 | mlog_errno(status); | |
2379 | ||
2380 | spin_lock_irqsave(&lockres->l_lock, flags); | |
2381 | } | |
2382 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
2383 | } | |
2384 | ||
d680efe9 MF |
2385 | void ocfs2_simple_drop_lockres(struct ocfs2_super *osb, |
2386 | struct ocfs2_lock_res *lockres) | |
ccd979bd | 2387 | { |
d680efe9 | 2388 | int ret; |
ccd979bd | 2389 | |
d680efe9 MF |
2390 | ocfs2_mark_lockres_freeing(lockres); |
2391 | ret = ocfs2_drop_lock(osb, lockres, NULL); | |
2392 | if (ret) | |
2393 | mlog_errno(ret); | |
2394 | } | |
ccd979bd | 2395 | |
d680efe9 MF |
2396 | static void ocfs2_drop_osb_locks(struct ocfs2_super *osb) |
2397 | { | |
2398 | ocfs2_simple_drop_lockres(osb, &osb->osb_super_lockres); | |
2399 | ocfs2_simple_drop_lockres(osb, &osb->osb_rename_lockres); | |
ccd979bd MF |
2400 | } |
2401 | ||
2402 | static void ocfs2_meta_pre_drop(struct ocfs2_lock_res *lockres, void *data) | |
2403 | { | |
2404 | struct inode *inode = data; | |
2405 | ||
2406 | /* the metadata lock requires a bit more work as we have an | |
2407 | * LVB to worry about. */ | |
2408 | if (lockres->l_flags & OCFS2_LOCK_ATTACHED && | |
2409 | lockres->l_level == LKM_EXMODE && | |
2410 | !(lockres->l_flags & OCFS2_LOCK_NEEDS_REFRESH)) | |
2411 | __ocfs2_stuff_meta_lvb(inode); | |
2412 | } | |
2413 | ||
2414 | int ocfs2_drop_inode_locks(struct inode *inode) | |
2415 | { | |
2416 | int status, err; | |
2417 | struct drop_lock_cb meta_dcb = { ocfs2_meta_pre_drop, inode, }; | |
2418 | ||
2419 | mlog_entry_void(); | |
2420 | ||
2421 | /* No need to call ocfs2_mark_lockres_freeing here - | |
2422 | * ocfs2_clear_inode has done it for us. */ | |
2423 | ||
2424 | err = ocfs2_drop_lock(OCFS2_SB(inode->i_sb), | |
2425 | &OCFS2_I(inode)->ip_data_lockres, | |
2426 | NULL); | |
2427 | if (err < 0) | |
2428 | mlog_errno(err); | |
2429 | ||
2430 | status = err; | |
2431 | ||
2432 | err = ocfs2_drop_lock(OCFS2_SB(inode->i_sb), | |
2433 | &OCFS2_I(inode)->ip_meta_lockres, | |
2434 | &meta_dcb); | |
2435 | if (err < 0) | |
2436 | mlog_errno(err); | |
2437 | if (err < 0 && !status) | |
2438 | status = err; | |
2439 | ||
2440 | err = ocfs2_drop_lock(OCFS2_SB(inode->i_sb), | |
2441 | &OCFS2_I(inode)->ip_rw_lockres, | |
2442 | NULL); | |
2443 | if (err < 0) | |
2444 | mlog_errno(err); | |
2445 | if (err < 0 && !status) | |
2446 | status = err; | |
2447 | ||
2448 | mlog_exit(status); | |
2449 | return status; | |
2450 | } | |
2451 | ||
2452 | static void ocfs2_prepare_downconvert(struct ocfs2_lock_res *lockres, | |
2453 | int new_level) | |
2454 | { | |
2455 | assert_spin_locked(&lockres->l_lock); | |
2456 | ||
2457 | BUG_ON(lockres->l_blocking <= LKM_NLMODE); | |
2458 | ||
2459 | if (lockres->l_level <= new_level) { | |
2460 | mlog(ML_ERROR, "lockres->l_level (%u) <= new_level (%u)\n", | |
2461 | lockres->l_level, new_level); | |
2462 | BUG(); | |
2463 | } | |
2464 | ||
2465 | mlog(0, "lock %s, new_level = %d, l_blocking = %d\n", | |
2466 | lockres->l_name, new_level, lockres->l_blocking); | |
2467 | ||
2468 | lockres->l_action = OCFS2_AST_DOWNCONVERT; | |
2469 | lockres->l_requested = new_level; | |
2470 | lockres_or_flags(lockres, OCFS2_LOCK_BUSY); | |
2471 | } | |
2472 | ||
2473 | static int ocfs2_downconvert_lock(struct ocfs2_super *osb, | |
2474 | struct ocfs2_lock_res *lockres, | |
2475 | int new_level, | |
2476 | int lvb) | |
2477 | { | |
2478 | int ret, dlm_flags = LKM_CONVERT; | |
2479 | enum dlm_status status; | |
2480 | ||
2481 | mlog_entry_void(); | |
2482 | ||
2483 | if (lvb) | |
2484 | dlm_flags |= LKM_VALBLK; | |
2485 | ||
2486 | status = dlmlock(osb->dlm, | |
2487 | new_level, | |
2488 | &lockres->l_lksb, | |
2489 | dlm_flags, | |
2490 | lockres->l_name, | |
f0681062 | 2491 | OCFS2_LOCK_ID_MAX_LEN - 1, |
e92d57df | 2492 | ocfs2_locking_ast, |
ccd979bd MF |
2493 | lockres, |
2494 | lockres->l_ops->bast); | |
2495 | if (status != DLM_NORMAL) { | |
2496 | ocfs2_log_dlm_error("dlmlock", status, lockres); | |
2497 | ret = -EINVAL; | |
2498 | ocfs2_recover_from_dlm_error(lockres, 1); | |
2499 | goto bail; | |
2500 | } | |
2501 | ||
2502 | ret = 0; | |
2503 | bail: | |
2504 | mlog_exit(ret); | |
2505 | return ret; | |
2506 | } | |
2507 | ||
2508 | /* returns 1 when the caller should unlock and call dlmunlock */ | |
2509 | static int ocfs2_prepare_cancel_convert(struct ocfs2_super *osb, | |
2510 | struct ocfs2_lock_res *lockres) | |
2511 | { | |
2512 | assert_spin_locked(&lockres->l_lock); | |
2513 | ||
2514 | mlog_entry_void(); | |
2515 | mlog(0, "lock %s\n", lockres->l_name); | |
2516 | ||
2517 | if (lockres->l_unlock_action == OCFS2_UNLOCK_CANCEL_CONVERT) { | |
2518 | /* If we're already trying to cancel a lock conversion | |
2519 | * then just drop the spinlock and allow the caller to | |
2520 | * requeue this lock. */ | |
2521 | ||
2522 | mlog(0, "Lockres %s, skip convert\n", lockres->l_name); | |
2523 | return 0; | |
2524 | } | |
2525 | ||
2526 | /* were we in a convert when we got the bast fire? */ | |
2527 | BUG_ON(lockres->l_action != OCFS2_AST_CONVERT && | |
2528 | lockres->l_action != OCFS2_AST_DOWNCONVERT); | |
2529 | /* set things up for the unlockast to know to just | |
2530 | * clear out the ast_action and unset busy, etc. */ | |
2531 | lockres->l_unlock_action = OCFS2_UNLOCK_CANCEL_CONVERT; | |
2532 | ||
2533 | mlog_bug_on_msg(!(lockres->l_flags & OCFS2_LOCK_BUSY), | |
2534 | "lock %s, invalid flags: 0x%lx\n", | |
2535 | lockres->l_name, lockres->l_flags); | |
2536 | ||
2537 | return 1; | |
2538 | } | |
2539 | ||
2540 | static int ocfs2_cancel_convert(struct ocfs2_super *osb, | |
2541 | struct ocfs2_lock_res *lockres) | |
2542 | { | |
2543 | int ret; | |
2544 | enum dlm_status status; | |
2545 | ||
2546 | mlog_entry_void(); | |
2547 | mlog(0, "lock %s\n", lockres->l_name); | |
2548 | ||
2549 | ret = 0; | |
2550 | status = dlmunlock(osb->dlm, | |
2551 | &lockres->l_lksb, | |
2552 | LKM_CANCEL, | |
2a45f2d1 | 2553 | ocfs2_unlock_ast, |
ccd979bd MF |
2554 | lockres); |
2555 | if (status != DLM_NORMAL) { | |
2556 | ocfs2_log_dlm_error("dlmunlock", status, lockres); | |
2557 | ret = -EINVAL; | |
2558 | ocfs2_recover_from_dlm_error(lockres, 0); | |
2559 | } | |
2560 | ||
2561 | mlog(0, "lock %s return from dlmunlock\n", lockres->l_name); | |
2562 | ||
2563 | mlog_exit(ret); | |
2564 | return ret; | |
2565 | } | |
2566 | ||
2567 | static inline int ocfs2_can_downconvert_meta_lock(struct inode *inode, | |
2568 | struct ocfs2_lock_res *lockres, | |
2569 | int new_level) | |
2570 | { | |
2571 | int ret; | |
2572 | ||
2573 | mlog_entry_void(); | |
2574 | ||
2575 | BUG_ON(new_level != LKM_NLMODE && new_level != LKM_PRMODE); | |
2576 | ||
2577 | if (lockres->l_flags & OCFS2_LOCK_REFRESHING) { | |
2578 | ret = 0; | |
2579 | mlog(0, "lockres %s currently being refreshed -- backing " | |
2580 | "off!\n", lockres->l_name); | |
2581 | } else if (new_level == LKM_PRMODE) | |
2582 | ret = !lockres->l_ex_holders && | |
2583 | ocfs2_inode_fully_checkpointed(inode); | |
2584 | else /* Must be NLMODE we're converting to. */ | |
2585 | ret = !lockres->l_ro_holders && !lockres->l_ex_holders && | |
2586 | ocfs2_inode_fully_checkpointed(inode); | |
2587 | ||
2588 | mlog_exit(ret); | |
2589 | return ret; | |
2590 | } | |
2591 | ||
2592 | static int ocfs2_do_unblock_meta(struct inode *inode, | |
2593 | int *requeue) | |
2594 | { | |
2595 | int new_level; | |
2596 | int set_lvb = 0; | |
2597 | int ret = 0; | |
2598 | struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_meta_lockres; | |
2599 | unsigned long flags; | |
2600 | ||
2601 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
2602 | ||
2603 | mlog_entry_void(); | |
2604 | ||
2605 | spin_lock_irqsave(&lockres->l_lock, flags); | |
2606 | ||
2607 | BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BLOCKED)); | |
2608 | ||
2609 | mlog(0, "l_level=%d, l_blocking=%d\n", lockres->l_level, | |
2610 | lockres->l_blocking); | |
2611 | ||
2612 | BUG_ON(lockres->l_level != LKM_EXMODE && | |
2613 | lockres->l_level != LKM_PRMODE); | |
2614 | ||
2615 | if (lockres->l_flags & OCFS2_LOCK_BUSY) { | |
2616 | *requeue = 1; | |
2617 | ret = ocfs2_prepare_cancel_convert(osb, lockres); | |
2618 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
2619 | if (ret) { | |
2620 | ret = ocfs2_cancel_convert(osb, lockres); | |
2621 | if (ret < 0) | |
2622 | mlog_errno(ret); | |
2623 | } | |
2624 | goto leave; | |
2625 | } | |
2626 | ||
2627 | new_level = ocfs2_highest_compat_lock_level(lockres->l_blocking); | |
2628 | ||
2629 | mlog(0, "l_level=%d, l_blocking=%d, new_level=%d\n", | |
2630 | lockres->l_level, lockres->l_blocking, new_level); | |
2631 | ||
2632 | if (ocfs2_can_downconvert_meta_lock(inode, lockres, new_level)) { | |
2633 | if (lockres->l_level == LKM_EXMODE) | |
2634 | set_lvb = 1; | |
2635 | ||
2636 | /* If the lock hasn't been refreshed yet (rare), then | |
2637 | * our memory inode values are old and we skip | |
2638 | * stuffing the lvb. There's no need to actually clear | |
2639 | * out the lvb here as it's value is still valid. */ | |
2640 | if (!(lockres->l_flags & OCFS2_LOCK_NEEDS_REFRESH)) { | |
2641 | if (set_lvb) | |
2642 | __ocfs2_stuff_meta_lvb(inode); | |
2643 | } else | |
2644 | mlog(0, "lockres %s: downconverting stale lock!\n", | |
2645 | lockres->l_name); | |
2646 | ||
2647 | mlog(0, "calling ocfs2_downconvert_lock with l_level=%d, " | |
2648 | "l_blocking=%d, new_level=%d\n", | |
2649 | lockres->l_level, lockres->l_blocking, new_level); | |
2650 | ||
2651 | ocfs2_prepare_downconvert(lockres, new_level); | |
2652 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
2653 | ret = ocfs2_downconvert_lock(osb, lockres, new_level, set_lvb); | |
2654 | goto leave; | |
2655 | } | |
2656 | if (!ocfs2_inode_fully_checkpointed(inode)) | |
2657 | ocfs2_start_checkpoint(osb); | |
2658 | ||
2659 | *requeue = 1; | |
2660 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
2661 | ret = 0; | |
2662 | leave: | |
2663 | mlog_exit(ret); | |
2664 | return ret; | |
2665 | } | |
2666 | ||
2667 | static int ocfs2_generic_unblock_lock(struct ocfs2_super *osb, | |
2668 | struct ocfs2_lock_res *lockres, | |
d680efe9 | 2669 | struct ocfs2_unblock_ctl *ctl, |
ccd979bd MF |
2670 | ocfs2_convert_worker_t *worker) |
2671 | { | |
2672 | unsigned long flags; | |
2673 | int blocking; | |
2674 | int new_level; | |
2675 | int ret = 0; | |
2676 | ||
2677 | mlog_entry_void(); | |
2678 | ||
2679 | spin_lock_irqsave(&lockres->l_lock, flags); | |
2680 | ||
2681 | BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BLOCKED)); | |
2682 | ||
2683 | recheck: | |
2684 | if (lockres->l_flags & OCFS2_LOCK_BUSY) { | |
d680efe9 | 2685 | ctl->requeue = 1; |
ccd979bd MF |
2686 | ret = ocfs2_prepare_cancel_convert(osb, lockres); |
2687 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
2688 | if (ret) { | |
2689 | ret = ocfs2_cancel_convert(osb, lockres); | |
2690 | if (ret < 0) | |
2691 | mlog_errno(ret); | |
2692 | } | |
2693 | goto leave; | |
2694 | } | |
2695 | ||
2696 | /* if we're blocking an exclusive and we have *any* holders, | |
2697 | * then requeue. */ | |
2698 | if ((lockres->l_blocking == LKM_EXMODE) | |
2699 | && (lockres->l_ex_holders || lockres->l_ro_holders)) { | |
2700 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
d680efe9 | 2701 | ctl->requeue = 1; |
ccd979bd MF |
2702 | ret = 0; |
2703 | goto leave; | |
2704 | } | |
2705 | ||
2706 | /* If it's a PR we're blocking, then only | |
2707 | * requeue if we've got any EX holders */ | |
2708 | if (lockres->l_blocking == LKM_PRMODE && | |
2709 | lockres->l_ex_holders) { | |
2710 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
d680efe9 | 2711 | ctl->requeue = 1; |
ccd979bd MF |
2712 | ret = 0; |
2713 | goto leave; | |
2714 | } | |
2715 | ||
2716 | /* If we get here, then we know that there are no more | |
2717 | * incompatible holders (and anyone asking for an incompatible | |
2718 | * lock is blocked). We can now downconvert the lock */ | |
2719 | if (!worker) | |
2720 | goto downconvert; | |
2721 | ||
2722 | /* Some lockres types want to do a bit of work before | |
2723 | * downconverting a lock. Allow that here. The worker function | |
2724 | * may sleep, so we save off a copy of what we're blocking as | |
2725 | * it may change while we're not holding the spin lock. */ | |
2726 | blocking = lockres->l_blocking; | |
2727 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
2728 | ||
d680efe9 MF |
2729 | ctl->unblock_action = worker(lockres, blocking); |
2730 | ||
2731 | if (ctl->unblock_action == UNBLOCK_STOP_POST) | |
2732 | goto leave; | |
ccd979bd MF |
2733 | |
2734 | spin_lock_irqsave(&lockres->l_lock, flags); | |
2735 | if (blocking != lockres->l_blocking) { | |
2736 | /* If this changed underneath us, then we can't drop | |
2737 | * it just yet. */ | |
2738 | goto recheck; | |
2739 | } | |
2740 | ||
2741 | downconvert: | |
d680efe9 | 2742 | ctl->requeue = 0; |
ccd979bd MF |
2743 | new_level = ocfs2_highest_compat_lock_level(lockres->l_blocking); |
2744 | ||
2745 | ocfs2_prepare_downconvert(lockres, new_level); | |
2746 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
2747 | ret = ocfs2_downconvert_lock(osb, lockres, new_level, 0); | |
2748 | leave: | |
2749 | mlog_exit(ret); | |
2750 | return ret; | |
2751 | } | |
2752 | ||
d680efe9 MF |
2753 | static int ocfs2_data_convert_worker(struct ocfs2_lock_res *lockres, |
2754 | int blocking) | |
ccd979bd MF |
2755 | { |
2756 | struct inode *inode; | |
2757 | struct address_space *mapping; | |
2758 | ||
ccd979bd MF |
2759 | inode = ocfs2_lock_res_inode(lockres); |
2760 | mapping = inode->i_mapping; | |
2761 | ||
2762 | if (filemap_fdatawrite(mapping)) { | |
b0697053 MF |
2763 | mlog(ML_ERROR, "Could not sync inode %llu for downconvert!", |
2764 | (unsigned long long)OCFS2_I(inode)->ip_blkno); | |
ccd979bd MF |
2765 | } |
2766 | sync_mapping_buffers(mapping); | |
2767 | if (blocking == LKM_EXMODE) { | |
2768 | truncate_inode_pages(mapping, 0); | |
2769 | unmap_mapping_range(mapping, 0, 0, 0); | |
2770 | } else { | |
2771 | /* We only need to wait on the I/O if we're not also | |
2772 | * truncating pages because truncate_inode_pages waits | |
2773 | * for us above. We don't truncate pages if we're | |
2774 | * blocking anything < EXMODE because we want to keep | |
2775 | * them around in that case. */ | |
2776 | filemap_fdatawait(mapping); | |
2777 | } | |
2778 | ||
d680efe9 | 2779 | return UNBLOCK_CONTINUE; |
ccd979bd MF |
2780 | } |
2781 | ||
2782 | int ocfs2_unblock_data(struct ocfs2_lock_res *lockres, | |
d680efe9 | 2783 | struct ocfs2_unblock_ctl *ctl) |
ccd979bd MF |
2784 | { |
2785 | int status; | |
2786 | struct inode *inode; | |
2787 | struct ocfs2_super *osb; | |
2788 | ||
2789 | mlog_entry_void(); | |
2790 | ||
2791 | inode = ocfs2_lock_res_inode(lockres); | |
2792 | osb = OCFS2_SB(inode->i_sb); | |
2793 | ||
b0697053 MF |
2794 | mlog(0, "unblock inode %llu\n", |
2795 | (unsigned long long)OCFS2_I(inode)->ip_blkno); | |
ccd979bd | 2796 | |
d680efe9 | 2797 | status = ocfs2_generic_unblock_lock(osb, lockres, ctl, |
ccd979bd MF |
2798 | ocfs2_data_convert_worker); |
2799 | if (status < 0) | |
2800 | mlog_errno(status); | |
2801 | ||
b0697053 | 2802 | mlog(0, "inode %llu, requeue = %d\n", |
d680efe9 | 2803 | (unsigned long long)OCFS2_I(inode)->ip_blkno, ctl->requeue); |
ccd979bd MF |
2804 | |
2805 | mlog_exit(status); | |
2806 | return status; | |
2807 | } | |
2808 | ||
2809 | static int ocfs2_unblock_inode_lock(struct ocfs2_lock_res *lockres, | |
d680efe9 | 2810 | struct ocfs2_unblock_ctl *ctl) |
ccd979bd MF |
2811 | { |
2812 | int status; | |
2813 | struct inode *inode; | |
2814 | ||
2815 | mlog_entry_void(); | |
2816 | ||
2817 | mlog(0, "Unblock lockres %s\n", lockres->l_name); | |
2818 | ||
2819 | inode = ocfs2_lock_res_inode(lockres); | |
2820 | ||
2821 | status = ocfs2_generic_unblock_lock(OCFS2_SB(inode->i_sb), | |
d680efe9 | 2822 | lockres, ctl, NULL); |
ccd979bd MF |
2823 | if (status < 0) |
2824 | mlog_errno(status); | |
2825 | ||
2826 | mlog_exit(status); | |
2827 | return status; | |
2828 | } | |
2829 | ||
d680efe9 MF |
2830 | static int ocfs2_unblock_meta(struct ocfs2_lock_res *lockres, |
2831 | struct ocfs2_unblock_ctl *ctl) | |
ccd979bd MF |
2832 | { |
2833 | int status; | |
2834 | struct inode *inode; | |
2835 | ||
2836 | mlog_entry_void(); | |
2837 | ||
2838 | inode = ocfs2_lock_res_inode(lockres); | |
2839 | ||
b0697053 MF |
2840 | mlog(0, "unblock inode %llu\n", |
2841 | (unsigned long long)OCFS2_I(inode)->ip_blkno); | |
ccd979bd | 2842 | |
d680efe9 | 2843 | status = ocfs2_do_unblock_meta(inode, &ctl->requeue); |
ccd979bd MF |
2844 | if (status < 0) |
2845 | mlog_errno(status); | |
2846 | ||
b0697053 | 2847 | mlog(0, "inode %llu, requeue = %d\n", |
d680efe9 | 2848 | (unsigned long long)OCFS2_I(inode)->ip_blkno, ctl->requeue); |
ccd979bd MF |
2849 | |
2850 | mlog_exit(status); | |
2851 | return status; | |
2852 | } | |
2853 | ||
d680efe9 MF |
2854 | /* |
2855 | * Does the final reference drop on our dentry lock. Right now this | |
2856 | * happens in the vote thread, but we could choose to simplify the | |
2857 | * dlmglue API and push these off to the ocfs2_wq in the future. | |
2858 | */ | |
2859 | static void ocfs2_dentry_post_unlock(struct ocfs2_super *osb, | |
2860 | struct ocfs2_lock_res *lockres) | |
2861 | { | |
2862 | struct ocfs2_dentry_lock *dl = ocfs2_lock_res_dl(lockres); | |
2863 | ocfs2_dentry_lock_put(osb, dl); | |
2864 | } | |
2865 | ||
2866 | /* | |
2867 | * d_delete() matching dentries before the lock downconvert. | |
2868 | * | |
2869 | * At this point, any process waiting to destroy the | |
2870 | * dentry_lock due to last ref count is stopped by the | |
2871 | * OCFS2_LOCK_QUEUED flag. | |
2872 | * | |
2873 | * We have two potential problems | |
2874 | * | |
2875 | * 1) If we do the last reference drop on our dentry_lock (via dput) | |
2876 | * we'll wind up in ocfs2_release_dentry_lock(), waiting on | |
2877 | * the downconvert to finish. Instead we take an elevated | |
2878 | * reference and push the drop until after we've completed our | |
2879 | * unblock processing. | |
2880 | * | |
2881 | * 2) There might be another process with a final reference, | |
2882 | * waiting on us to finish processing. If this is the case, we | |
2883 | * detect it and exit out - there's no more dentries anyway. | |
2884 | */ | |
2885 | static int ocfs2_dentry_convert_worker(struct ocfs2_lock_res *lockres, | |
2886 | int blocking) | |
2887 | { | |
2888 | struct ocfs2_dentry_lock *dl = ocfs2_lock_res_dl(lockres); | |
2889 | struct ocfs2_inode_info *oi = OCFS2_I(dl->dl_inode); | |
2890 | struct dentry *dentry; | |
2891 | unsigned long flags; | |
2892 | int extra_ref = 0; | |
2893 | ||
2894 | /* | |
2895 | * This node is blocking another node from getting a read | |
2896 | * lock. This happens when we've renamed within a | |
2897 | * directory. We've forced the other nodes to d_delete(), but | |
2898 | * we never actually dropped our lock because it's still | |
2899 | * valid. The downconvert code will retain a PR for this node, | |
2900 | * so there's no further work to do. | |
2901 | */ | |
2902 | if (blocking == LKM_PRMODE) | |
2903 | return UNBLOCK_CONTINUE; | |
2904 | ||
2905 | /* | |
2906 | * Mark this inode as potentially orphaned. The code in | |
2907 | * ocfs2_delete_inode() will figure out whether it actually | |
2908 | * needs to be freed or not. | |
2909 | */ | |
2910 | spin_lock(&oi->ip_lock); | |
2911 | oi->ip_flags |= OCFS2_INODE_MAYBE_ORPHANED; | |
2912 | spin_unlock(&oi->ip_lock); | |
2913 | ||
2914 | /* | |
2915 | * Yuck. We need to make sure however that the check of | |
2916 | * OCFS2_LOCK_FREEING and the extra reference are atomic with | |
2917 | * respect to a reference decrement or the setting of that | |
2918 | * flag. | |
2919 | */ | |
2920 | spin_lock_irqsave(&lockres->l_lock, flags); | |
2921 | spin_lock(&dentry_attach_lock); | |
2922 | if (!(lockres->l_flags & OCFS2_LOCK_FREEING) | |
2923 | && dl->dl_count) { | |
2924 | dl->dl_count++; | |
2925 | extra_ref = 1; | |
2926 | } | |
2927 | spin_unlock(&dentry_attach_lock); | |
2928 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
2929 | ||
2930 | mlog(0, "extra_ref = %d\n", extra_ref); | |
2931 | ||
2932 | /* | |
2933 | * We have a process waiting on us in ocfs2_dentry_iput(), | |
2934 | * which means we can't have any more outstanding | |
2935 | * aliases. There's no need to do any more work. | |
2936 | */ | |
2937 | if (!extra_ref) | |
2938 | return UNBLOCK_CONTINUE; | |
2939 | ||
2940 | spin_lock(&dentry_attach_lock); | |
2941 | while (1) { | |
2942 | dentry = ocfs2_find_local_alias(dl->dl_inode, | |
2943 | dl->dl_parent_blkno, 1); | |
2944 | if (!dentry) | |
2945 | break; | |
2946 | spin_unlock(&dentry_attach_lock); | |
2947 | ||
2948 | mlog(0, "d_delete(%.*s);\n", dentry->d_name.len, | |
2949 | dentry->d_name.name); | |
2950 | ||
2951 | /* | |
2952 | * The following dcache calls may do an | |
2953 | * iput(). Normally we don't want that from the | |
2954 | * downconverting thread, but in this case it's ok | |
2955 | * because the requesting node already has an | |
2956 | * exclusive lock on the inode, so it can't be queued | |
2957 | * for a downconvert. | |
2958 | */ | |
2959 | d_delete(dentry); | |
2960 | dput(dentry); | |
2961 | ||
2962 | spin_lock(&dentry_attach_lock); | |
2963 | } | |
2964 | spin_unlock(&dentry_attach_lock); | |
2965 | ||
2966 | /* | |
2967 | * If we are the last holder of this dentry lock, there is no | |
2968 | * reason to downconvert so skip straight to the unlock. | |
2969 | */ | |
2970 | if (dl->dl_count == 1) | |
2971 | return UNBLOCK_STOP_POST; | |
2972 | ||
2973 | return UNBLOCK_CONTINUE_POST; | |
2974 | } | |
2975 | ||
2976 | static int ocfs2_unblock_dentry_lock(struct ocfs2_lock_res *lockres, | |
2977 | struct ocfs2_unblock_ctl *ctl) | |
2978 | { | |
2979 | int ret; | |
2980 | struct ocfs2_dentry_lock *dl = ocfs2_lock_res_dl(lockres); | |
2981 | struct ocfs2_super *osb = OCFS2_SB(dl->dl_inode->i_sb); | |
2982 | ||
2983 | mlog(0, "unblock dentry lock: %llu\n", | |
2984 | (unsigned long long)OCFS2_I(dl->dl_inode)->ip_blkno); | |
2985 | ||
2986 | ret = ocfs2_generic_unblock_lock(osb, | |
2987 | lockres, | |
2988 | ctl, | |
2989 | ocfs2_dentry_convert_worker); | |
2990 | if (ret < 0) | |
2991 | mlog_errno(ret); | |
2992 | ||
2993 | mlog(0, "requeue = %d, post = %d\n", ctl->requeue, ctl->unblock_action); | |
2994 | ||
2995 | return ret; | |
2996 | } | |
2997 | ||
ccd979bd MF |
2998 | /* Generic unblock function for any lockres whose private data is an |
2999 | * ocfs2_super pointer. */ | |
3000 | static int ocfs2_unblock_osb_lock(struct ocfs2_lock_res *lockres, | |
d680efe9 | 3001 | struct ocfs2_unblock_ctl *ctl) |
ccd979bd MF |
3002 | { |
3003 | int status; | |
3004 | struct ocfs2_super *osb; | |
3005 | ||
3006 | mlog_entry_void(); | |
3007 | ||
3008 | mlog(0, "Unblock lockres %s\n", lockres->l_name); | |
3009 | ||
3010 | osb = ocfs2_lock_res_super(lockres); | |
3011 | ||
3012 | status = ocfs2_generic_unblock_lock(osb, | |
3013 | lockres, | |
d680efe9 | 3014 | ctl, |
ccd979bd MF |
3015 | NULL); |
3016 | if (status < 0) | |
3017 | mlog_errno(status); | |
3018 | ||
3019 | mlog_exit(status); | |
3020 | return status; | |
3021 | } | |
3022 | ||
3023 | void ocfs2_process_blocked_lock(struct ocfs2_super *osb, | |
3024 | struct ocfs2_lock_res *lockres) | |
3025 | { | |
3026 | int status; | |
d680efe9 | 3027 | struct ocfs2_unblock_ctl ctl = {0, 0,}; |
ccd979bd MF |
3028 | unsigned long flags; |
3029 | ||
3030 | /* Our reference to the lockres in this function can be | |
3031 | * considered valid until we remove the OCFS2_LOCK_QUEUED | |
3032 | * flag. */ | |
3033 | ||
3034 | mlog_entry_void(); | |
3035 | ||
3036 | BUG_ON(!lockres); | |
3037 | BUG_ON(!lockres->l_ops); | |
3038 | BUG_ON(!lockres->l_ops->unblock); | |
3039 | ||
3040 | mlog(0, "lockres %s blocked.\n", lockres->l_name); | |
3041 | ||
3042 | /* Detect whether a lock has been marked as going away while | |
3043 | * the vote thread was processing other things. A lock can | |
3044 | * still be marked with OCFS2_LOCK_FREEING after this check, | |
3045 | * but short circuiting here will still save us some | |
3046 | * performance. */ | |
3047 | spin_lock_irqsave(&lockres->l_lock, flags); | |
3048 | if (lockres->l_flags & OCFS2_LOCK_FREEING) | |
3049 | goto unqueue; | |
3050 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
3051 | ||
d680efe9 | 3052 | status = lockres->l_ops->unblock(lockres, &ctl); |
ccd979bd MF |
3053 | if (status < 0) |
3054 | mlog_errno(status); | |
3055 | ||
3056 | spin_lock_irqsave(&lockres->l_lock, flags); | |
3057 | unqueue: | |
d680efe9 | 3058 | if (lockres->l_flags & OCFS2_LOCK_FREEING || !ctl.requeue) { |
ccd979bd MF |
3059 | lockres_clear_flags(lockres, OCFS2_LOCK_QUEUED); |
3060 | } else | |
3061 | ocfs2_schedule_blocked_lock(osb, lockres); | |
3062 | ||
3063 | mlog(0, "lockres %s, requeue = %s.\n", lockres->l_name, | |
d680efe9 | 3064 | ctl.requeue ? "yes" : "no"); |
ccd979bd MF |
3065 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
3066 | ||
d680efe9 MF |
3067 | if (ctl.unblock_action != UNBLOCK_CONTINUE |
3068 | && lockres->l_ops->post_unlock) | |
3069 | lockres->l_ops->post_unlock(osb, lockres); | |
3070 | ||
ccd979bd MF |
3071 | mlog_exit_void(); |
3072 | } | |
3073 | ||
3074 | static void ocfs2_schedule_blocked_lock(struct ocfs2_super *osb, | |
3075 | struct ocfs2_lock_res *lockres) | |
3076 | { | |
3077 | mlog_entry_void(); | |
3078 | ||
3079 | assert_spin_locked(&lockres->l_lock); | |
3080 | ||
3081 | if (lockres->l_flags & OCFS2_LOCK_FREEING) { | |
3082 | /* Do not schedule a lock for downconvert when it's on | |
3083 | * the way to destruction - any nodes wanting access | |
3084 | * to the resource will get it soon. */ | |
3085 | mlog(0, "Lockres %s won't be scheduled: flags 0x%lx\n", | |
3086 | lockres->l_name, lockres->l_flags); | |
3087 | return; | |
3088 | } | |
3089 | ||
3090 | lockres_or_flags(lockres, OCFS2_LOCK_QUEUED); | |
3091 | ||
3092 | spin_lock(&osb->vote_task_lock); | |
3093 | if (list_empty(&lockres->l_blocked_list)) { | |
3094 | list_add_tail(&lockres->l_blocked_list, | |
3095 | &osb->blocked_lock_list); | |
3096 | osb->blocked_lock_count++; | |
3097 | } | |
3098 | spin_unlock(&osb->vote_task_lock); | |
3099 | ||
3100 | mlog_exit_void(); | |
3101 | } | |
3102 | ||
3103 | /* This aids in debugging situations where a bad LVB might be involved. */ | |
3104 | void ocfs2_dump_meta_lvb_info(u64 level, | |
3105 | const char *function, | |
3106 | unsigned int line, | |
3107 | struct ocfs2_lock_res *lockres) | |
3108 | { | |
3109 | struct ocfs2_meta_lvb *lvb = (struct ocfs2_meta_lvb *) lockres->l_lksb.lvb; | |
3110 | ||
3111 | mlog(level, "LVB information for %s (called from %s:%u):\n", | |
3112 | lockres->l_name, function, line); | |
f9e2d82e MF |
3113 | mlog(level, "version: %u, clusters: %u, generation: 0x%x\n", |
3114 | lvb->lvb_version, be32_to_cpu(lvb->lvb_iclusters), | |
3115 | be32_to_cpu(lvb->lvb_igeneration)); | |
b0697053 MF |
3116 | mlog(level, "size: %llu, uid %u, gid %u, mode 0x%x\n", |
3117 | (unsigned long long)be64_to_cpu(lvb->lvb_isize), | |
3118 | be32_to_cpu(lvb->lvb_iuid), be32_to_cpu(lvb->lvb_igid), | |
3119 | be16_to_cpu(lvb->lvb_imode)); | |
3120 | mlog(level, "nlink %u, atime_packed 0x%llx, ctime_packed 0x%llx, " | |
ca4d147e | 3121 | "mtime_packed 0x%llx iattr 0x%x\n", be16_to_cpu(lvb->lvb_inlink), |
b0697053 MF |
3122 | (long long)be64_to_cpu(lvb->lvb_iatime_packed), |
3123 | (long long)be64_to_cpu(lvb->lvb_ictime_packed), | |
ca4d147e HP |
3124 | (long long)be64_to_cpu(lvb->lvb_imtime_packed), |
3125 | be32_to_cpu(lvb->lvb_iattr)); | |
ccd979bd | 3126 | } |