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> | |
ccd979bd MF |
30 | #include <linux/kthread.h> |
31 | #include <linux/pagemap.h> | |
32 | #include <linux/debugfs.h> | |
33 | #include <linux/seq_file.h> | |
8ddb7b00 | 34 | #include <linux/time.h> |
ccd979bd | 35 | |
ccd979bd MF |
36 | #define MLOG_MASK_PREFIX ML_DLM_GLUE |
37 | #include <cluster/masklog.h> | |
38 | ||
39 | #include "ocfs2.h" | |
d24fbcda | 40 | #include "ocfs2_lockingver.h" |
ccd979bd MF |
41 | |
42 | #include "alloc.h" | |
d680efe9 | 43 | #include "dcache.h" |
ccd979bd MF |
44 | #include "dlmglue.h" |
45 | #include "extent_map.h" | |
7f1a37e3 | 46 | #include "file.h" |
ccd979bd MF |
47 | #include "heartbeat.h" |
48 | #include "inode.h" | |
49 | #include "journal.h" | |
24ef1815 | 50 | #include "stackglue.h" |
ccd979bd MF |
51 | #include "slot_map.h" |
52 | #include "super.h" | |
53 | #include "uptodate.h" | |
ccd979bd MF |
54 | |
55 | #include "buffer_head_io.h" | |
56 | ||
57 | struct ocfs2_mask_waiter { | |
58 | struct list_head mw_item; | |
59 | int mw_status; | |
60 | struct completion mw_complete; | |
61 | unsigned long mw_mask; | |
62 | unsigned long mw_goal; | |
8ddb7b00 SM |
63 | #ifdef CONFIG_OCFS2_FS_STATS |
64 | unsigned long long mw_lock_start; | |
65 | #endif | |
ccd979bd MF |
66 | }; |
67 | ||
54a7e755 MF |
68 | static struct ocfs2_super *ocfs2_get_dentry_osb(struct ocfs2_lock_res *lockres); |
69 | static struct ocfs2_super *ocfs2_get_inode_osb(struct ocfs2_lock_res *lockres); | |
cf8e06f1 | 70 | static struct ocfs2_super *ocfs2_get_file_osb(struct ocfs2_lock_res *lockres); |
ccd979bd | 71 | |
d680efe9 | 72 | /* |
cc567d89 | 73 | * Return value from ->downconvert_worker functions. |
d680efe9 | 74 | * |
b5e500e2 | 75 | * These control the precise actions of ocfs2_unblock_lock() |
d680efe9 MF |
76 | * and ocfs2_process_blocked_lock() |
77 | * | |
78 | */ | |
79 | enum ocfs2_unblock_action { | |
80 | UNBLOCK_CONTINUE = 0, /* Continue downconvert */ | |
81 | UNBLOCK_CONTINUE_POST = 1, /* Continue downconvert, fire | |
82 | * ->post_unlock callback */ | |
83 | UNBLOCK_STOP_POST = 2, /* Do not downconvert, fire | |
84 | * ->post_unlock() callback. */ | |
85 | }; | |
86 | ||
87 | struct ocfs2_unblock_ctl { | |
88 | int requeue; | |
89 | enum ocfs2_unblock_action unblock_action; | |
90 | }; | |
91 | ||
810d5aeb MF |
92 | static int ocfs2_check_meta_downconvert(struct ocfs2_lock_res *lockres, |
93 | int new_level); | |
94 | static void ocfs2_set_meta_lvb(struct ocfs2_lock_res *lockres); | |
95 | ||
cc567d89 MF |
96 | static int ocfs2_data_convert_worker(struct ocfs2_lock_res *lockres, |
97 | int blocking); | |
98 | ||
cc567d89 MF |
99 | static int ocfs2_dentry_convert_worker(struct ocfs2_lock_res *lockres, |
100 | int blocking); | |
d680efe9 MF |
101 | |
102 | static void ocfs2_dentry_post_unlock(struct ocfs2_super *osb, | |
103 | struct ocfs2_lock_res *lockres); | |
ccd979bd | 104 | |
6cb129f5 AB |
105 | |
106 | #define mlog_meta_lvb(__level, __lockres) ocfs2_dump_meta_lvb_info(__level, __PRETTY_FUNCTION__, __LINE__, __lockres) | |
107 | ||
108 | /* This aids in debugging situations where a bad LVB might be involved. */ | |
109 | static void ocfs2_dump_meta_lvb_info(u64 level, | |
110 | const char *function, | |
111 | unsigned int line, | |
112 | struct ocfs2_lock_res *lockres) | |
113 | { | |
8f2c9c1b JB |
114 | struct ocfs2_meta_lvb *lvb = |
115 | (struct ocfs2_meta_lvb *)ocfs2_dlm_lvb(&lockres->l_lksb); | |
6cb129f5 AB |
116 | |
117 | mlog(level, "LVB information for %s (called from %s:%u):\n", | |
118 | lockres->l_name, function, line); | |
119 | mlog(level, "version: %u, clusters: %u, generation: 0x%x\n", | |
120 | lvb->lvb_version, be32_to_cpu(lvb->lvb_iclusters), | |
121 | be32_to_cpu(lvb->lvb_igeneration)); | |
122 | mlog(level, "size: %llu, uid %u, gid %u, mode 0x%x\n", | |
123 | (unsigned long long)be64_to_cpu(lvb->lvb_isize), | |
124 | be32_to_cpu(lvb->lvb_iuid), be32_to_cpu(lvb->lvb_igid), | |
125 | be16_to_cpu(lvb->lvb_imode)); | |
126 | mlog(level, "nlink %u, atime_packed 0x%llx, ctime_packed 0x%llx, " | |
127 | "mtime_packed 0x%llx iattr 0x%x\n", be16_to_cpu(lvb->lvb_inlink), | |
128 | (long long)be64_to_cpu(lvb->lvb_iatime_packed), | |
129 | (long long)be64_to_cpu(lvb->lvb_ictime_packed), | |
130 | (long long)be64_to_cpu(lvb->lvb_imtime_packed), | |
131 | be32_to_cpu(lvb->lvb_iattr)); | |
132 | } | |
133 | ||
134 | ||
f625c979 MF |
135 | /* |
136 | * OCFS2 Lock Resource Operations | |
137 | * | |
138 | * These fine tune the behavior of the generic dlmglue locking infrastructure. | |
0d5dc6c2 MF |
139 | * |
140 | * The most basic of lock types can point ->l_priv to their respective | |
141 | * struct ocfs2_super and allow the default actions to manage things. | |
142 | * | |
143 | * Right now, each lock type also needs to implement an init function, | |
144 | * and trivial lock/unlock wrappers. ocfs2_simple_drop_lockres() | |
145 | * should be called when the lock is no longer needed (i.e., object | |
146 | * destruction time). | |
f625c979 | 147 | */ |
ccd979bd | 148 | struct ocfs2_lock_res_ops { |
54a7e755 MF |
149 | /* |
150 | * Translate an ocfs2_lock_res * into an ocfs2_super *. Define | |
151 | * this callback if ->l_priv is not an ocfs2_super pointer | |
152 | */ | |
153 | struct ocfs2_super * (*get_osb)(struct ocfs2_lock_res *); | |
b5e500e2 | 154 | |
0d5dc6c2 | 155 | /* |
34d024f8 MF |
156 | * Optionally called in the downconvert thread after a |
157 | * successful downconvert. The lockres will not be referenced | |
158 | * after this callback is called, so it is safe to free | |
159 | * memory, etc. | |
0d5dc6c2 MF |
160 | * |
161 | * The exact semantics of when this is called are controlled | |
162 | * by ->downconvert_worker() | |
163 | */ | |
d680efe9 | 164 | void (*post_unlock)(struct ocfs2_super *, struct ocfs2_lock_res *); |
f625c979 | 165 | |
16d5b956 MF |
166 | /* |
167 | * Allow a lock type to add checks to determine whether it is | |
168 | * safe to downconvert a lock. Return 0 to re-queue the | |
169 | * downconvert at a later time, nonzero to continue. | |
170 | * | |
171 | * For most locks, the default checks that there are no | |
172 | * incompatible holders are sufficient. | |
173 | * | |
174 | * Called with the lockres spinlock held. | |
175 | */ | |
176 | int (*check_downconvert)(struct ocfs2_lock_res *, int); | |
177 | ||
5ef0d4ea MF |
178 | /* |
179 | * Allows a lock type to populate the lock value block. This | |
180 | * is called on downconvert, and when we drop a lock. | |
181 | * | |
182 | * Locks that want to use this should set LOCK_TYPE_USES_LVB | |
183 | * in the flags field. | |
184 | * | |
185 | * Called with the lockres spinlock held. | |
186 | */ | |
187 | void (*set_lvb)(struct ocfs2_lock_res *); | |
188 | ||
cc567d89 MF |
189 | /* |
190 | * Called from the downconvert thread when it is determined | |
191 | * that a lock will be downconverted. This is called without | |
192 | * any locks held so the function can do work that might | |
193 | * schedule (syncing out data, etc). | |
194 | * | |
195 | * This should return any one of the ocfs2_unblock_action | |
196 | * values, depending on what it wants the thread to do. | |
197 | */ | |
198 | int (*downconvert_worker)(struct ocfs2_lock_res *, int); | |
199 | ||
f625c979 MF |
200 | /* |
201 | * LOCK_TYPE_* flags which describe the specific requirements | |
202 | * of a lock type. Descriptions of each individual flag follow. | |
203 | */ | |
204 | int flags; | |
ccd979bd MF |
205 | }; |
206 | ||
f625c979 MF |
207 | /* |
208 | * Some locks want to "refresh" potentially stale data when a | |
209 | * meaningful (PRMODE or EXMODE) lock level is first obtained. If this | |
210 | * flag is set, the OCFS2_LOCK_NEEDS_REFRESH flag will be set on the | |
211 | * individual lockres l_flags member from the ast function. It is | |
212 | * expected that the locking wrapper will clear the | |
213 | * OCFS2_LOCK_NEEDS_REFRESH flag when done. | |
214 | */ | |
215 | #define LOCK_TYPE_REQUIRES_REFRESH 0x1 | |
216 | ||
b80fc012 | 217 | /* |
5ef0d4ea MF |
218 | * Indicate that a lock type makes use of the lock value block. The |
219 | * ->set_lvb lock type callback must be defined. | |
b80fc012 MF |
220 | */ |
221 | #define LOCK_TYPE_USES_LVB 0x2 | |
222 | ||
ccd979bd | 223 | static struct ocfs2_lock_res_ops ocfs2_inode_rw_lops = { |
54a7e755 | 224 | .get_osb = ocfs2_get_inode_osb, |
f625c979 | 225 | .flags = 0, |
ccd979bd MF |
226 | }; |
227 | ||
e63aecb6 | 228 | static struct ocfs2_lock_res_ops ocfs2_inode_inode_lops = { |
54a7e755 | 229 | .get_osb = ocfs2_get_inode_osb, |
810d5aeb MF |
230 | .check_downconvert = ocfs2_check_meta_downconvert, |
231 | .set_lvb = ocfs2_set_meta_lvb, | |
f1f54068 | 232 | .downconvert_worker = ocfs2_data_convert_worker, |
b80fc012 | 233 | .flags = LOCK_TYPE_REQUIRES_REFRESH|LOCK_TYPE_USES_LVB, |
ccd979bd MF |
234 | }; |
235 | ||
ccd979bd | 236 | static struct ocfs2_lock_res_ops ocfs2_super_lops = { |
f625c979 | 237 | .flags = LOCK_TYPE_REQUIRES_REFRESH, |
ccd979bd MF |
238 | }; |
239 | ||
240 | static struct ocfs2_lock_res_ops ocfs2_rename_lops = { | |
f625c979 | 241 | .flags = 0, |
ccd979bd MF |
242 | }; |
243 | ||
d680efe9 | 244 | static struct ocfs2_lock_res_ops ocfs2_dentry_lops = { |
54a7e755 | 245 | .get_osb = ocfs2_get_dentry_osb, |
d680efe9 | 246 | .post_unlock = ocfs2_dentry_post_unlock, |
cc567d89 | 247 | .downconvert_worker = ocfs2_dentry_convert_worker, |
f625c979 | 248 | .flags = 0, |
d680efe9 MF |
249 | }; |
250 | ||
50008630 TY |
251 | static struct ocfs2_lock_res_ops ocfs2_inode_open_lops = { |
252 | .get_osb = ocfs2_get_inode_osb, | |
253 | .flags = 0, | |
254 | }; | |
255 | ||
cf8e06f1 MF |
256 | static struct ocfs2_lock_res_ops ocfs2_flock_lops = { |
257 | .get_osb = ocfs2_get_file_osb, | |
258 | .flags = 0, | |
259 | }; | |
260 | ||
ccd979bd MF |
261 | static inline int ocfs2_is_inode_lock(struct ocfs2_lock_res *lockres) |
262 | { | |
263 | return lockres->l_type == OCFS2_LOCK_TYPE_META || | |
50008630 TY |
264 | lockres->l_type == OCFS2_LOCK_TYPE_RW || |
265 | lockres->l_type == OCFS2_LOCK_TYPE_OPEN; | |
ccd979bd MF |
266 | } |
267 | ||
ccd979bd MF |
268 | static inline struct inode *ocfs2_lock_res_inode(struct ocfs2_lock_res *lockres) |
269 | { | |
270 | BUG_ON(!ocfs2_is_inode_lock(lockres)); | |
271 | ||
272 | return (struct inode *) lockres->l_priv; | |
273 | } | |
274 | ||
d680efe9 MF |
275 | static inline struct ocfs2_dentry_lock *ocfs2_lock_res_dl(struct ocfs2_lock_res *lockres) |
276 | { | |
277 | BUG_ON(lockres->l_type != OCFS2_LOCK_TYPE_DENTRY); | |
278 | ||
279 | return (struct ocfs2_dentry_lock *)lockres->l_priv; | |
280 | } | |
281 | ||
54a7e755 MF |
282 | static inline struct ocfs2_super *ocfs2_get_lockres_osb(struct ocfs2_lock_res *lockres) |
283 | { | |
284 | if (lockres->l_ops->get_osb) | |
285 | return lockres->l_ops->get_osb(lockres); | |
286 | ||
287 | return (struct ocfs2_super *)lockres->l_priv; | |
288 | } | |
289 | ||
ccd979bd MF |
290 | static int ocfs2_lock_create(struct ocfs2_super *osb, |
291 | struct ocfs2_lock_res *lockres, | |
292 | int level, | |
bd3e7610 | 293 | u32 dlm_flags); |
ccd979bd MF |
294 | static inline int ocfs2_may_continue_on_blocked_lock(struct ocfs2_lock_res *lockres, |
295 | int wanted); | |
296 | static void ocfs2_cluster_unlock(struct ocfs2_super *osb, | |
297 | struct ocfs2_lock_res *lockres, | |
298 | int level); | |
299 | static inline void ocfs2_generic_handle_downconvert_action(struct ocfs2_lock_res *lockres); | |
300 | static inline void ocfs2_generic_handle_convert_action(struct ocfs2_lock_res *lockres); | |
301 | static inline void ocfs2_generic_handle_attach_action(struct ocfs2_lock_res *lockres); | |
302 | static int ocfs2_generic_handle_bast(struct ocfs2_lock_res *lockres, int level); | |
303 | static void ocfs2_schedule_blocked_lock(struct ocfs2_super *osb, | |
304 | struct ocfs2_lock_res *lockres); | |
305 | static inline void ocfs2_recover_from_dlm_error(struct ocfs2_lock_res *lockres, | |
306 | int convert); | |
7431cd7e JB |
307 | #define ocfs2_log_dlm_error(_func, _err, _lockres) do { \ |
308 | mlog(ML_ERROR, "DLM error %d while calling %s on resource %s\n", \ | |
309 | _err, _func, _lockres->l_name); \ | |
ccd979bd | 310 | } while (0) |
34d024f8 MF |
311 | static int ocfs2_downconvert_thread(void *arg); |
312 | static void ocfs2_downconvert_on_unlock(struct ocfs2_super *osb, | |
313 | struct ocfs2_lock_res *lockres); | |
e63aecb6 | 314 | static int ocfs2_inode_lock_update(struct inode *inode, |
ccd979bd MF |
315 | struct buffer_head **bh); |
316 | static void ocfs2_drop_osb_locks(struct ocfs2_super *osb); | |
317 | static inline int ocfs2_highest_compat_lock_level(int level); | |
de551246 JB |
318 | static unsigned int ocfs2_prepare_downconvert(struct ocfs2_lock_res *lockres, |
319 | int new_level); | |
cf8e06f1 MF |
320 | static int ocfs2_downconvert_lock(struct ocfs2_super *osb, |
321 | struct ocfs2_lock_res *lockres, | |
322 | int new_level, | |
de551246 JB |
323 | int lvb, |
324 | unsigned int generation); | |
cf8e06f1 MF |
325 | static int ocfs2_prepare_cancel_convert(struct ocfs2_super *osb, |
326 | struct ocfs2_lock_res *lockres); | |
327 | static int ocfs2_cancel_convert(struct ocfs2_super *osb, | |
328 | struct ocfs2_lock_res *lockres); | |
329 | ||
ccd979bd | 330 | |
ccd979bd MF |
331 | static void ocfs2_build_lock_name(enum ocfs2_lock_type type, |
332 | u64 blkno, | |
333 | u32 generation, | |
334 | char *name) | |
335 | { | |
336 | int len; | |
337 | ||
338 | mlog_entry_void(); | |
339 | ||
340 | BUG_ON(type >= OCFS2_NUM_LOCK_TYPES); | |
341 | ||
b0697053 MF |
342 | len = snprintf(name, OCFS2_LOCK_ID_MAX_LEN, "%c%s%016llx%08x", |
343 | ocfs2_lock_type_char(type), OCFS2_LOCK_ID_PAD, | |
344 | (long long)blkno, generation); | |
ccd979bd MF |
345 | |
346 | BUG_ON(len != (OCFS2_LOCK_ID_MAX_LEN - 1)); | |
347 | ||
348 | mlog(0, "built lock resource with name: %s\n", name); | |
349 | ||
350 | mlog_exit_void(); | |
351 | } | |
352 | ||
34af946a | 353 | static DEFINE_SPINLOCK(ocfs2_dlm_tracking_lock); |
ccd979bd MF |
354 | |
355 | static void ocfs2_add_lockres_tracking(struct ocfs2_lock_res *res, | |
356 | struct ocfs2_dlm_debug *dlm_debug) | |
357 | { | |
358 | mlog(0, "Add tracking for lockres %s\n", res->l_name); | |
359 | ||
360 | spin_lock(&ocfs2_dlm_tracking_lock); | |
361 | list_add(&res->l_debug_list, &dlm_debug->d_lockres_tracking); | |
362 | spin_unlock(&ocfs2_dlm_tracking_lock); | |
363 | } | |
364 | ||
365 | static void ocfs2_remove_lockres_tracking(struct ocfs2_lock_res *res) | |
366 | { | |
367 | spin_lock(&ocfs2_dlm_tracking_lock); | |
368 | if (!list_empty(&res->l_debug_list)) | |
369 | list_del_init(&res->l_debug_list); | |
370 | spin_unlock(&ocfs2_dlm_tracking_lock); | |
371 | } | |
372 | ||
8ddb7b00 SM |
373 | #ifdef CONFIG_OCFS2_FS_STATS |
374 | static void ocfs2_init_lock_stats(struct ocfs2_lock_res *res) | |
375 | { | |
376 | res->l_lock_num_prmode = 0; | |
377 | res->l_lock_num_prmode_failed = 0; | |
378 | res->l_lock_total_prmode = 0; | |
379 | res->l_lock_max_prmode = 0; | |
380 | res->l_lock_num_exmode = 0; | |
381 | res->l_lock_num_exmode_failed = 0; | |
382 | res->l_lock_total_exmode = 0; | |
383 | res->l_lock_max_exmode = 0; | |
384 | res->l_lock_refresh = 0; | |
385 | } | |
386 | ||
387 | static void ocfs2_update_lock_stats(struct ocfs2_lock_res *res, int level, | |
388 | struct ocfs2_mask_waiter *mw, int ret) | |
389 | { | |
390 | unsigned long long *num, *sum; | |
391 | unsigned int *max, *failed; | |
392 | struct timespec ts = current_kernel_time(); | |
393 | unsigned long long time = timespec_to_ns(&ts) - mw->mw_lock_start; | |
394 | ||
395 | if (level == LKM_PRMODE) { | |
396 | num = &res->l_lock_num_prmode; | |
397 | sum = &res->l_lock_total_prmode; | |
398 | max = &res->l_lock_max_prmode; | |
399 | failed = &res->l_lock_num_prmode_failed; | |
400 | } else if (level == LKM_EXMODE) { | |
401 | num = &res->l_lock_num_exmode; | |
402 | sum = &res->l_lock_total_exmode; | |
403 | max = &res->l_lock_max_exmode; | |
404 | failed = &res->l_lock_num_exmode_failed; | |
405 | } else | |
406 | return; | |
407 | ||
408 | (*num)++; | |
409 | (*sum) += time; | |
410 | if (time > *max) | |
411 | *max = time; | |
412 | if (ret) | |
413 | (*failed)++; | |
414 | } | |
415 | ||
416 | static inline void ocfs2_track_lock_refresh(struct ocfs2_lock_res *lockres) | |
417 | { | |
418 | lockres->l_lock_refresh++; | |
419 | } | |
420 | ||
421 | static inline void ocfs2_init_start_time(struct ocfs2_mask_waiter *mw) | |
422 | { | |
423 | struct timespec ts = current_kernel_time(); | |
424 | mw->mw_lock_start = timespec_to_ns(&ts); | |
425 | } | |
426 | #else | |
427 | static inline void ocfs2_init_lock_stats(struct ocfs2_lock_res *res) | |
428 | { | |
429 | } | |
430 | static inline void ocfs2_update_lock_stats(struct ocfs2_lock_res *res, | |
431 | int level, struct ocfs2_mask_waiter *mw, int ret) | |
432 | { | |
433 | } | |
434 | static inline void ocfs2_track_lock_refresh(struct ocfs2_lock_res *lockres) | |
435 | { | |
436 | } | |
437 | static inline void ocfs2_init_start_time(struct ocfs2_mask_waiter *mw) | |
438 | { | |
439 | } | |
440 | #endif | |
441 | ||
ccd979bd MF |
442 | static void ocfs2_lock_res_init_common(struct ocfs2_super *osb, |
443 | struct ocfs2_lock_res *res, | |
444 | enum ocfs2_lock_type type, | |
ccd979bd MF |
445 | struct ocfs2_lock_res_ops *ops, |
446 | void *priv) | |
447 | { | |
ccd979bd MF |
448 | res->l_type = type; |
449 | res->l_ops = ops; | |
450 | res->l_priv = priv; | |
451 | ||
bd3e7610 JB |
452 | res->l_level = DLM_LOCK_IV; |
453 | res->l_requested = DLM_LOCK_IV; | |
454 | res->l_blocking = DLM_LOCK_IV; | |
ccd979bd MF |
455 | res->l_action = OCFS2_AST_INVALID; |
456 | res->l_unlock_action = OCFS2_UNLOCK_INVALID; | |
457 | ||
458 | res->l_flags = OCFS2_LOCK_INITIALIZED; | |
459 | ||
460 | ocfs2_add_lockres_tracking(res, osb->osb_dlm_debug); | |
8ddb7b00 SM |
461 | |
462 | ocfs2_init_lock_stats(res); | |
ccd979bd MF |
463 | } |
464 | ||
465 | void ocfs2_lock_res_init_once(struct ocfs2_lock_res *res) | |
466 | { | |
467 | /* This also clears out the lock status block */ | |
468 | memset(res, 0, sizeof(struct ocfs2_lock_res)); | |
469 | spin_lock_init(&res->l_lock); | |
470 | init_waitqueue_head(&res->l_event); | |
471 | INIT_LIST_HEAD(&res->l_blocked_list); | |
472 | INIT_LIST_HEAD(&res->l_mask_waiters); | |
473 | } | |
474 | ||
475 | void ocfs2_inode_lock_res_init(struct ocfs2_lock_res *res, | |
476 | enum ocfs2_lock_type type, | |
24c19ef4 | 477 | unsigned int generation, |
ccd979bd MF |
478 | struct inode *inode) |
479 | { | |
480 | struct ocfs2_lock_res_ops *ops; | |
481 | ||
482 | switch(type) { | |
483 | case OCFS2_LOCK_TYPE_RW: | |
484 | ops = &ocfs2_inode_rw_lops; | |
485 | break; | |
486 | case OCFS2_LOCK_TYPE_META: | |
e63aecb6 | 487 | ops = &ocfs2_inode_inode_lops; |
ccd979bd | 488 | break; |
50008630 TY |
489 | case OCFS2_LOCK_TYPE_OPEN: |
490 | ops = &ocfs2_inode_open_lops; | |
491 | break; | |
ccd979bd MF |
492 | default: |
493 | mlog_bug_on_msg(1, "type: %d\n", type); | |
494 | ops = NULL; /* thanks, gcc */ | |
495 | break; | |
496 | }; | |
497 | ||
d680efe9 | 498 | ocfs2_build_lock_name(type, OCFS2_I(inode)->ip_blkno, |
24c19ef4 | 499 | generation, res->l_name); |
d680efe9 MF |
500 | ocfs2_lock_res_init_common(OCFS2_SB(inode->i_sb), res, type, ops, inode); |
501 | } | |
502 | ||
54a7e755 MF |
503 | static struct ocfs2_super *ocfs2_get_inode_osb(struct ocfs2_lock_res *lockres) |
504 | { | |
505 | struct inode *inode = ocfs2_lock_res_inode(lockres); | |
506 | ||
507 | return OCFS2_SB(inode->i_sb); | |
508 | } | |
509 | ||
cf8e06f1 MF |
510 | static struct ocfs2_super *ocfs2_get_file_osb(struct ocfs2_lock_res *lockres) |
511 | { | |
512 | struct ocfs2_file_private *fp = lockres->l_priv; | |
513 | ||
514 | return OCFS2_SB(fp->fp_file->f_mapping->host->i_sb); | |
515 | } | |
516 | ||
d680efe9 MF |
517 | static __u64 ocfs2_get_dentry_lock_ino(struct ocfs2_lock_res *lockres) |
518 | { | |
519 | __be64 inode_blkno_be; | |
520 | ||
521 | memcpy(&inode_blkno_be, &lockres->l_name[OCFS2_DENTRY_LOCK_INO_START], | |
522 | sizeof(__be64)); | |
523 | ||
524 | return be64_to_cpu(inode_blkno_be); | |
525 | } | |
526 | ||
54a7e755 MF |
527 | static struct ocfs2_super *ocfs2_get_dentry_osb(struct ocfs2_lock_res *lockres) |
528 | { | |
529 | struct ocfs2_dentry_lock *dl = lockres->l_priv; | |
530 | ||
531 | return OCFS2_SB(dl->dl_inode->i_sb); | |
532 | } | |
533 | ||
d680efe9 MF |
534 | void ocfs2_dentry_lock_res_init(struct ocfs2_dentry_lock *dl, |
535 | u64 parent, struct inode *inode) | |
536 | { | |
537 | int len; | |
538 | u64 inode_blkno = OCFS2_I(inode)->ip_blkno; | |
539 | __be64 inode_blkno_be = cpu_to_be64(inode_blkno); | |
540 | struct ocfs2_lock_res *lockres = &dl->dl_lockres; | |
541 | ||
542 | ocfs2_lock_res_init_once(lockres); | |
543 | ||
544 | /* | |
545 | * Unfortunately, the standard lock naming scheme won't work | |
546 | * here because we have two 16 byte values to use. Instead, | |
547 | * we'll stuff the inode number as a binary value. We still | |
548 | * want error prints to show something without garbling the | |
549 | * display, so drop a null byte in there before the inode | |
550 | * number. A future version of OCFS2 will likely use all | |
551 | * binary lock names. The stringified names have been a | |
552 | * tremendous aid in debugging, but now that the debugfs | |
553 | * interface exists, we can mangle things there if need be. | |
554 | * | |
555 | * NOTE: We also drop the standard "pad" value (the total lock | |
556 | * name size stays the same though - the last part is all | |
557 | * zeros due to the memset in ocfs2_lock_res_init_once() | |
558 | */ | |
559 | len = snprintf(lockres->l_name, OCFS2_DENTRY_LOCK_INO_START, | |
560 | "%c%016llx", | |
561 | ocfs2_lock_type_char(OCFS2_LOCK_TYPE_DENTRY), | |
562 | (long long)parent); | |
563 | ||
564 | BUG_ON(len != (OCFS2_DENTRY_LOCK_INO_START - 1)); | |
565 | ||
566 | memcpy(&lockres->l_name[OCFS2_DENTRY_LOCK_INO_START], &inode_blkno_be, | |
567 | sizeof(__be64)); | |
568 | ||
569 | ocfs2_lock_res_init_common(OCFS2_SB(inode->i_sb), lockres, | |
570 | OCFS2_LOCK_TYPE_DENTRY, &ocfs2_dentry_lops, | |
571 | dl); | |
ccd979bd MF |
572 | } |
573 | ||
574 | static void ocfs2_super_lock_res_init(struct ocfs2_lock_res *res, | |
575 | struct ocfs2_super *osb) | |
576 | { | |
577 | /* Superblock lockres doesn't come from a slab so we call init | |
578 | * once on it manually. */ | |
579 | ocfs2_lock_res_init_once(res); | |
d680efe9 MF |
580 | ocfs2_build_lock_name(OCFS2_LOCK_TYPE_SUPER, OCFS2_SUPER_BLOCK_BLKNO, |
581 | 0, res->l_name); | |
ccd979bd | 582 | ocfs2_lock_res_init_common(osb, res, OCFS2_LOCK_TYPE_SUPER, |
ccd979bd MF |
583 | &ocfs2_super_lops, osb); |
584 | } | |
585 | ||
586 | static void ocfs2_rename_lock_res_init(struct ocfs2_lock_res *res, | |
587 | struct ocfs2_super *osb) | |
588 | { | |
589 | /* Rename lockres doesn't come from a slab so we call init | |
590 | * once on it manually. */ | |
591 | ocfs2_lock_res_init_once(res); | |
d680efe9 MF |
592 | ocfs2_build_lock_name(OCFS2_LOCK_TYPE_RENAME, 0, 0, res->l_name); |
593 | ocfs2_lock_res_init_common(osb, res, OCFS2_LOCK_TYPE_RENAME, | |
ccd979bd MF |
594 | &ocfs2_rename_lops, osb); |
595 | } | |
596 | ||
cf8e06f1 MF |
597 | void ocfs2_file_lock_res_init(struct ocfs2_lock_res *lockres, |
598 | struct ocfs2_file_private *fp) | |
599 | { | |
600 | struct inode *inode = fp->fp_file->f_mapping->host; | |
601 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
602 | ||
603 | ocfs2_lock_res_init_once(lockres); | |
604 | ocfs2_build_lock_name(OCFS2_LOCK_TYPE_FLOCK, oi->ip_blkno, | |
605 | inode->i_generation, lockres->l_name); | |
606 | ocfs2_lock_res_init_common(OCFS2_SB(inode->i_sb), lockres, | |
607 | OCFS2_LOCK_TYPE_FLOCK, &ocfs2_flock_lops, | |
608 | fp); | |
609 | lockres->l_flags |= OCFS2_LOCK_NOCACHE; | |
610 | } | |
611 | ||
ccd979bd MF |
612 | void ocfs2_lock_res_free(struct ocfs2_lock_res *res) |
613 | { | |
614 | mlog_entry_void(); | |
615 | ||
616 | if (!(res->l_flags & OCFS2_LOCK_INITIALIZED)) | |
617 | return; | |
618 | ||
619 | ocfs2_remove_lockres_tracking(res); | |
620 | ||
621 | mlog_bug_on_msg(!list_empty(&res->l_blocked_list), | |
622 | "Lockres %s is on the blocked list\n", | |
623 | res->l_name); | |
624 | mlog_bug_on_msg(!list_empty(&res->l_mask_waiters), | |
625 | "Lockres %s has mask waiters pending\n", | |
626 | res->l_name); | |
627 | mlog_bug_on_msg(spin_is_locked(&res->l_lock), | |
628 | "Lockres %s is locked\n", | |
629 | res->l_name); | |
630 | mlog_bug_on_msg(res->l_ro_holders, | |
631 | "Lockres %s has %u ro holders\n", | |
632 | res->l_name, res->l_ro_holders); | |
633 | mlog_bug_on_msg(res->l_ex_holders, | |
634 | "Lockres %s has %u ex holders\n", | |
635 | res->l_name, res->l_ex_holders); | |
636 | ||
637 | /* Need to clear out the lock status block for the dlm */ | |
638 | memset(&res->l_lksb, 0, sizeof(res->l_lksb)); | |
639 | ||
640 | res->l_flags = 0UL; | |
641 | mlog_exit_void(); | |
642 | } | |
643 | ||
644 | static inline void ocfs2_inc_holders(struct ocfs2_lock_res *lockres, | |
645 | int level) | |
646 | { | |
647 | mlog_entry_void(); | |
648 | ||
649 | BUG_ON(!lockres); | |
650 | ||
651 | switch(level) { | |
bd3e7610 | 652 | case DLM_LOCK_EX: |
ccd979bd MF |
653 | lockres->l_ex_holders++; |
654 | break; | |
bd3e7610 | 655 | case DLM_LOCK_PR: |
ccd979bd MF |
656 | lockres->l_ro_holders++; |
657 | break; | |
658 | default: | |
659 | BUG(); | |
660 | } | |
661 | ||
662 | mlog_exit_void(); | |
663 | } | |
664 | ||
665 | static inline void ocfs2_dec_holders(struct ocfs2_lock_res *lockres, | |
666 | int level) | |
667 | { | |
668 | mlog_entry_void(); | |
669 | ||
670 | BUG_ON(!lockres); | |
671 | ||
672 | switch(level) { | |
bd3e7610 | 673 | case DLM_LOCK_EX: |
ccd979bd MF |
674 | BUG_ON(!lockres->l_ex_holders); |
675 | lockres->l_ex_holders--; | |
676 | break; | |
bd3e7610 | 677 | case DLM_LOCK_PR: |
ccd979bd MF |
678 | BUG_ON(!lockres->l_ro_holders); |
679 | lockres->l_ro_holders--; | |
680 | break; | |
681 | default: | |
682 | BUG(); | |
683 | } | |
684 | mlog_exit_void(); | |
685 | } | |
686 | ||
687 | /* WARNING: This function lives in a world where the only three lock | |
688 | * levels are EX, PR, and NL. It *will* have to be adjusted when more | |
689 | * lock types are added. */ | |
690 | static inline int ocfs2_highest_compat_lock_level(int level) | |
691 | { | |
bd3e7610 | 692 | int new_level = DLM_LOCK_EX; |
ccd979bd | 693 | |
bd3e7610 JB |
694 | if (level == DLM_LOCK_EX) |
695 | new_level = DLM_LOCK_NL; | |
696 | else if (level == DLM_LOCK_PR) | |
697 | new_level = DLM_LOCK_PR; | |
ccd979bd MF |
698 | return new_level; |
699 | } | |
700 | ||
701 | static void lockres_set_flags(struct ocfs2_lock_res *lockres, | |
702 | unsigned long newflags) | |
703 | { | |
800deef3 | 704 | struct ocfs2_mask_waiter *mw, *tmp; |
ccd979bd MF |
705 | |
706 | assert_spin_locked(&lockres->l_lock); | |
707 | ||
708 | lockres->l_flags = newflags; | |
709 | ||
800deef3 | 710 | list_for_each_entry_safe(mw, tmp, &lockres->l_mask_waiters, mw_item) { |
ccd979bd MF |
711 | if ((lockres->l_flags & mw->mw_mask) != mw->mw_goal) |
712 | continue; | |
713 | ||
714 | list_del_init(&mw->mw_item); | |
715 | mw->mw_status = 0; | |
716 | complete(&mw->mw_complete); | |
717 | } | |
718 | } | |
719 | static void lockres_or_flags(struct ocfs2_lock_res *lockres, unsigned long or) | |
720 | { | |
721 | lockres_set_flags(lockres, lockres->l_flags | or); | |
722 | } | |
723 | static void lockres_clear_flags(struct ocfs2_lock_res *lockres, | |
724 | unsigned long clear) | |
725 | { | |
726 | lockres_set_flags(lockres, lockres->l_flags & ~clear); | |
727 | } | |
728 | ||
729 | static inline void ocfs2_generic_handle_downconvert_action(struct ocfs2_lock_res *lockres) | |
730 | { | |
731 | mlog_entry_void(); | |
732 | ||
733 | BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BUSY)); | |
734 | BUG_ON(!(lockres->l_flags & OCFS2_LOCK_ATTACHED)); | |
735 | BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BLOCKED)); | |
bd3e7610 | 736 | BUG_ON(lockres->l_blocking <= DLM_LOCK_NL); |
ccd979bd MF |
737 | |
738 | lockres->l_level = lockres->l_requested; | |
739 | if (lockres->l_level <= | |
740 | ocfs2_highest_compat_lock_level(lockres->l_blocking)) { | |
bd3e7610 | 741 | lockres->l_blocking = DLM_LOCK_NL; |
ccd979bd MF |
742 | lockres_clear_flags(lockres, OCFS2_LOCK_BLOCKED); |
743 | } | |
744 | lockres_clear_flags(lockres, OCFS2_LOCK_BUSY); | |
745 | ||
746 | mlog_exit_void(); | |
747 | } | |
748 | ||
749 | static inline void ocfs2_generic_handle_convert_action(struct ocfs2_lock_res *lockres) | |
750 | { | |
751 | mlog_entry_void(); | |
752 | ||
753 | BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BUSY)); | |
754 | BUG_ON(!(lockres->l_flags & OCFS2_LOCK_ATTACHED)); | |
755 | ||
756 | /* Convert from RO to EX doesn't really need anything as our | |
757 | * information is already up to data. Convert from NL to | |
758 | * *anything* however should mark ourselves as needing an | |
759 | * update */ | |
bd3e7610 | 760 | if (lockres->l_level == DLM_LOCK_NL && |
f625c979 | 761 | lockres->l_ops->flags & LOCK_TYPE_REQUIRES_REFRESH) |
ccd979bd MF |
762 | lockres_or_flags(lockres, OCFS2_LOCK_NEEDS_REFRESH); |
763 | ||
764 | lockres->l_level = lockres->l_requested; | |
765 | lockres_clear_flags(lockres, OCFS2_LOCK_BUSY); | |
766 | ||
767 | mlog_exit_void(); | |
768 | } | |
769 | ||
770 | static inline void ocfs2_generic_handle_attach_action(struct ocfs2_lock_res *lockres) | |
771 | { | |
772 | mlog_entry_void(); | |
773 | ||
3cf0c507 | 774 | BUG_ON((!(lockres->l_flags & OCFS2_LOCK_BUSY))); |
ccd979bd MF |
775 | BUG_ON(lockres->l_flags & OCFS2_LOCK_ATTACHED); |
776 | ||
bd3e7610 | 777 | if (lockres->l_requested > DLM_LOCK_NL && |
f625c979 MF |
778 | !(lockres->l_flags & OCFS2_LOCK_LOCAL) && |
779 | lockres->l_ops->flags & LOCK_TYPE_REQUIRES_REFRESH) | |
ccd979bd MF |
780 | lockres_or_flags(lockres, OCFS2_LOCK_NEEDS_REFRESH); |
781 | ||
782 | lockres->l_level = lockres->l_requested; | |
783 | lockres_or_flags(lockres, OCFS2_LOCK_ATTACHED); | |
784 | lockres_clear_flags(lockres, OCFS2_LOCK_BUSY); | |
785 | ||
786 | mlog_exit_void(); | |
787 | } | |
788 | ||
ccd979bd MF |
789 | static int ocfs2_generic_handle_bast(struct ocfs2_lock_res *lockres, |
790 | int level) | |
791 | { | |
792 | int needs_downconvert = 0; | |
793 | mlog_entry_void(); | |
794 | ||
795 | assert_spin_locked(&lockres->l_lock); | |
796 | ||
797 | lockres_or_flags(lockres, OCFS2_LOCK_BLOCKED); | |
798 | ||
799 | if (level > lockres->l_blocking) { | |
800 | /* only schedule a downconvert if we haven't already scheduled | |
801 | * one that goes low enough to satisfy the level we're | |
802 | * blocking. this also catches the case where we get | |
803 | * duplicate BASTs */ | |
804 | if (ocfs2_highest_compat_lock_level(level) < | |
805 | ocfs2_highest_compat_lock_level(lockres->l_blocking)) | |
806 | needs_downconvert = 1; | |
807 | ||
808 | lockres->l_blocking = level; | |
809 | } | |
810 | ||
811 | mlog_exit(needs_downconvert); | |
812 | return needs_downconvert; | |
813 | } | |
814 | ||
de551246 JB |
815 | /* |
816 | * OCFS2_LOCK_PENDING and l_pending_gen. | |
817 | * | |
818 | * Why does OCFS2_LOCK_PENDING exist? To close a race between setting | |
819 | * OCFS2_LOCK_BUSY and calling ocfs2_dlm_lock(). See ocfs2_unblock_lock() | |
820 | * for more details on the race. | |
821 | * | |
822 | * OCFS2_LOCK_PENDING closes the race quite nicely. However, it introduces | |
823 | * a race on itself. In o2dlm, we can get the ast before ocfs2_dlm_lock() | |
824 | * returns. The ast clears OCFS2_LOCK_BUSY, and must therefore clear | |
825 | * OCFS2_LOCK_PENDING at the same time. When ocfs2_dlm_lock() returns, | |
826 | * the caller is going to try to clear PENDING again. If nothing else is | |
827 | * happening, __lockres_clear_pending() sees PENDING is unset and does | |
828 | * nothing. | |
829 | * | |
830 | * But what if another path (eg downconvert thread) has just started a | |
831 | * new locking action? The other path has re-set PENDING. Our path | |
832 | * cannot clear PENDING, because that will re-open the original race | |
833 | * window. | |
834 | * | |
835 | * [Example] | |
836 | * | |
837 | * ocfs2_meta_lock() | |
838 | * ocfs2_cluster_lock() | |
839 | * set BUSY | |
840 | * set PENDING | |
841 | * drop l_lock | |
842 | * ocfs2_dlm_lock() | |
843 | * ocfs2_locking_ast() ocfs2_downconvert_thread() | |
844 | * clear PENDING ocfs2_unblock_lock() | |
845 | * take_l_lock | |
846 | * !BUSY | |
847 | * ocfs2_prepare_downconvert() | |
848 | * set BUSY | |
849 | * set PENDING | |
850 | * drop l_lock | |
851 | * take l_lock | |
852 | * clear PENDING | |
853 | * drop l_lock | |
854 | * <window> | |
855 | * ocfs2_dlm_lock() | |
856 | * | |
857 | * So as you can see, we now have a window where l_lock is not held, | |
858 | * PENDING is not set, and ocfs2_dlm_lock() has not been called. | |
859 | * | |
860 | * The core problem is that ocfs2_cluster_lock() has cleared the PENDING | |
861 | * set by ocfs2_prepare_downconvert(). That wasn't nice. | |
862 | * | |
863 | * To solve this we introduce l_pending_gen. A call to | |
864 | * lockres_clear_pending() will only do so when it is passed a generation | |
865 | * number that matches the lockres. lockres_set_pending() will return the | |
866 | * current generation number. When ocfs2_cluster_lock() goes to clear | |
867 | * PENDING, it passes the generation it got from set_pending(). In our | |
868 | * example above, the generation numbers will *not* match. Thus, | |
869 | * ocfs2_cluster_lock() will not clear the PENDING set by | |
870 | * ocfs2_prepare_downconvert(). | |
871 | */ | |
872 | ||
873 | /* Unlocked version for ocfs2_locking_ast() */ | |
874 | static void __lockres_clear_pending(struct ocfs2_lock_res *lockres, | |
875 | unsigned int generation, | |
876 | struct ocfs2_super *osb) | |
877 | { | |
878 | assert_spin_locked(&lockres->l_lock); | |
879 | ||
880 | /* | |
881 | * The ast and locking functions can race us here. The winner | |
882 | * will clear pending, the loser will not. | |
883 | */ | |
884 | if (!(lockres->l_flags & OCFS2_LOCK_PENDING) || | |
885 | (lockres->l_pending_gen != generation)) | |
886 | return; | |
887 | ||
888 | lockres_clear_flags(lockres, OCFS2_LOCK_PENDING); | |
889 | lockres->l_pending_gen++; | |
890 | ||
891 | /* | |
892 | * The downconvert thread may have skipped us because we | |
893 | * were PENDING. Wake it up. | |
894 | */ | |
895 | if (lockres->l_flags & OCFS2_LOCK_BLOCKED) | |
896 | ocfs2_wake_downconvert_thread(osb); | |
897 | } | |
898 | ||
899 | /* Locked version for callers of ocfs2_dlm_lock() */ | |
900 | static void lockres_clear_pending(struct ocfs2_lock_res *lockres, | |
901 | unsigned int generation, | |
902 | struct ocfs2_super *osb) | |
903 | { | |
904 | unsigned long flags; | |
905 | ||
906 | spin_lock_irqsave(&lockres->l_lock, flags); | |
907 | __lockres_clear_pending(lockres, generation, osb); | |
908 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
909 | } | |
910 | ||
911 | static unsigned int lockres_set_pending(struct ocfs2_lock_res *lockres) | |
912 | { | |
913 | assert_spin_locked(&lockres->l_lock); | |
914 | BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BUSY)); | |
915 | ||
916 | lockres_or_flags(lockres, OCFS2_LOCK_PENDING); | |
917 | ||
918 | return lockres->l_pending_gen; | |
919 | } | |
920 | ||
921 | ||
aa2623ad | 922 | static void ocfs2_blocking_ast(void *opaque, int level) |
ccd979bd | 923 | { |
aa2623ad MF |
924 | struct ocfs2_lock_res *lockres = opaque; |
925 | struct ocfs2_super *osb = ocfs2_get_lockres_osb(lockres); | |
ccd979bd MF |
926 | int needs_downconvert; |
927 | unsigned long flags; | |
928 | ||
bd3e7610 | 929 | BUG_ON(level <= DLM_LOCK_NL); |
ccd979bd | 930 | |
aa2623ad MF |
931 | mlog(0, "BAST fired for lockres %s, blocking %d, level %d type %s\n", |
932 | lockres->l_name, level, lockres->l_level, | |
933 | ocfs2_lock_type_string(lockres->l_type)); | |
934 | ||
cf8e06f1 MF |
935 | /* |
936 | * We can skip the bast for locks which don't enable caching - | |
937 | * they'll be dropped at the earliest possible time anyway. | |
938 | */ | |
939 | if (lockres->l_flags & OCFS2_LOCK_NOCACHE) | |
940 | return; | |
941 | ||
ccd979bd MF |
942 | spin_lock_irqsave(&lockres->l_lock, flags); |
943 | needs_downconvert = ocfs2_generic_handle_bast(lockres, level); | |
944 | if (needs_downconvert) | |
945 | ocfs2_schedule_blocked_lock(osb, lockres); | |
946 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
947 | ||
d680efe9 MF |
948 | wake_up(&lockres->l_event); |
949 | ||
34d024f8 | 950 | ocfs2_wake_downconvert_thread(osb); |
ccd979bd MF |
951 | } |
952 | ||
e92d57df | 953 | static void ocfs2_locking_ast(void *opaque) |
ccd979bd | 954 | { |
e92d57df | 955 | struct ocfs2_lock_res *lockres = opaque; |
de551246 | 956 | struct ocfs2_super *osb = ocfs2_get_lockres_osb(lockres); |
ccd979bd | 957 | unsigned long flags; |
1693a5c0 | 958 | int status; |
ccd979bd MF |
959 | |
960 | spin_lock_irqsave(&lockres->l_lock, flags); | |
961 | ||
1693a5c0 DT |
962 | status = ocfs2_dlm_lock_status(&lockres->l_lksb); |
963 | ||
964 | if (status == -EAGAIN) { | |
965 | lockres_clear_flags(lockres, OCFS2_LOCK_BUSY); | |
966 | goto out; | |
967 | } | |
968 | ||
969 | if (status) { | |
8f2c9c1b | 970 | mlog(ML_ERROR, "lockres %s: lksb status value of %d!\n", |
1693a5c0 | 971 | lockres->l_name, status); |
ccd979bd MF |
972 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
973 | return; | |
974 | } | |
975 | ||
976 | switch(lockres->l_action) { | |
977 | case OCFS2_AST_ATTACH: | |
978 | ocfs2_generic_handle_attach_action(lockres); | |
e92d57df | 979 | lockres_clear_flags(lockres, OCFS2_LOCK_LOCAL); |
ccd979bd MF |
980 | break; |
981 | case OCFS2_AST_CONVERT: | |
982 | ocfs2_generic_handle_convert_action(lockres); | |
983 | break; | |
984 | case OCFS2_AST_DOWNCONVERT: | |
985 | ocfs2_generic_handle_downconvert_action(lockres); | |
986 | break; | |
987 | default: | |
e92d57df MF |
988 | mlog(ML_ERROR, "lockres %s: ast fired with invalid action: %u " |
989 | "lockres flags = 0x%lx, unlock action: %u\n", | |
990 | lockres->l_name, lockres->l_action, lockres->l_flags, | |
991 | lockres->l_unlock_action); | |
ccd979bd MF |
992 | BUG(); |
993 | } | |
1693a5c0 | 994 | out: |
ccd979bd MF |
995 | /* set it to something invalid so if we get called again we |
996 | * can catch it. */ | |
997 | lockres->l_action = OCFS2_AST_INVALID; | |
ccd979bd | 998 | |
de551246 JB |
999 | /* Did we try to cancel this lock? Clear that state */ |
1000 | if (lockres->l_unlock_action == OCFS2_UNLOCK_CANCEL_CONVERT) | |
1001 | lockres->l_unlock_action = OCFS2_UNLOCK_INVALID; | |
1002 | ||
1003 | /* | |
1004 | * We may have beaten the locking functions here. We certainly | |
1005 | * know that dlm_lock() has been called :-) | |
1006 | * Because we can't have two lock calls in flight at once, we | |
1007 | * can use lockres->l_pending_gen. | |
1008 | */ | |
1009 | __lockres_clear_pending(lockres, lockres->l_pending_gen, osb); | |
1010 | ||
ccd979bd | 1011 | wake_up(&lockres->l_event); |
d680efe9 | 1012 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
ccd979bd MF |
1013 | } |
1014 | ||
ccd979bd MF |
1015 | static inline void ocfs2_recover_from_dlm_error(struct ocfs2_lock_res *lockres, |
1016 | int convert) | |
1017 | { | |
1018 | unsigned long flags; | |
1019 | ||
1020 | mlog_entry_void(); | |
1021 | spin_lock_irqsave(&lockres->l_lock, flags); | |
1022 | lockres_clear_flags(lockres, OCFS2_LOCK_BUSY); | |
1023 | if (convert) | |
1024 | lockres->l_action = OCFS2_AST_INVALID; | |
1025 | else | |
1026 | lockres->l_unlock_action = OCFS2_UNLOCK_INVALID; | |
1027 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
1028 | ||
1029 | wake_up(&lockres->l_event); | |
1030 | mlog_exit_void(); | |
1031 | } | |
1032 | ||
1033 | /* Note: If we detect another process working on the lock (i.e., | |
1034 | * OCFS2_LOCK_BUSY), we'll bail out returning 0. It's up to the caller | |
1035 | * to do the right thing in that case. | |
1036 | */ | |
1037 | static int ocfs2_lock_create(struct ocfs2_super *osb, | |
1038 | struct ocfs2_lock_res *lockres, | |
1039 | int level, | |
bd3e7610 | 1040 | u32 dlm_flags) |
ccd979bd MF |
1041 | { |
1042 | int ret = 0; | |
ccd979bd | 1043 | unsigned long flags; |
de551246 | 1044 | unsigned int gen; |
ccd979bd MF |
1045 | |
1046 | mlog_entry_void(); | |
1047 | ||
bd3e7610 | 1048 | mlog(0, "lock %s, level = %d, flags = %u\n", lockres->l_name, level, |
ccd979bd MF |
1049 | dlm_flags); |
1050 | ||
1051 | spin_lock_irqsave(&lockres->l_lock, flags); | |
1052 | if ((lockres->l_flags & OCFS2_LOCK_ATTACHED) || | |
1053 | (lockres->l_flags & OCFS2_LOCK_BUSY)) { | |
1054 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
1055 | goto bail; | |
1056 | } | |
1057 | ||
1058 | lockres->l_action = OCFS2_AST_ATTACH; | |
1059 | lockres->l_requested = level; | |
1060 | lockres_or_flags(lockres, OCFS2_LOCK_BUSY); | |
de551246 | 1061 | gen = lockres_set_pending(lockres); |
ccd979bd MF |
1062 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
1063 | ||
4670c46d | 1064 | ret = ocfs2_dlm_lock(osb->cconn, |
7431cd7e JB |
1065 | level, |
1066 | &lockres->l_lksb, | |
1067 | dlm_flags, | |
1068 | lockres->l_name, | |
1069 | OCFS2_LOCK_ID_MAX_LEN - 1, | |
1070 | lockres); | |
de551246 | 1071 | lockres_clear_pending(lockres, gen, osb); |
7431cd7e JB |
1072 | if (ret) { |
1073 | ocfs2_log_dlm_error("ocfs2_dlm_lock", ret, lockres); | |
ccd979bd MF |
1074 | ocfs2_recover_from_dlm_error(lockres, 1); |
1075 | } | |
1076 | ||
7431cd7e | 1077 | mlog(0, "lock %s, return from ocfs2_dlm_lock\n", lockres->l_name); |
ccd979bd MF |
1078 | |
1079 | bail: | |
1080 | mlog_exit(ret); | |
1081 | return ret; | |
1082 | } | |
1083 | ||
1084 | static inline int ocfs2_check_wait_flag(struct ocfs2_lock_res *lockres, | |
1085 | int flag) | |
1086 | { | |
1087 | unsigned long flags; | |
1088 | int ret; | |
1089 | ||
1090 | spin_lock_irqsave(&lockres->l_lock, flags); | |
1091 | ret = lockres->l_flags & flag; | |
1092 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
1093 | ||
1094 | return ret; | |
1095 | } | |
1096 | ||
1097 | static inline void ocfs2_wait_on_busy_lock(struct ocfs2_lock_res *lockres) | |
1098 | ||
1099 | { | |
1100 | wait_event(lockres->l_event, | |
1101 | !ocfs2_check_wait_flag(lockres, OCFS2_LOCK_BUSY)); | |
1102 | } | |
1103 | ||
1104 | static inline void ocfs2_wait_on_refreshing_lock(struct ocfs2_lock_res *lockres) | |
1105 | ||
1106 | { | |
1107 | wait_event(lockres->l_event, | |
1108 | !ocfs2_check_wait_flag(lockres, OCFS2_LOCK_REFRESHING)); | |
1109 | } | |
1110 | ||
1111 | /* predict what lock level we'll be dropping down to on behalf | |
1112 | * of another node, and return true if the currently wanted | |
1113 | * level will be compatible with it. */ | |
1114 | static inline int ocfs2_may_continue_on_blocked_lock(struct ocfs2_lock_res *lockres, | |
1115 | int wanted) | |
1116 | { | |
1117 | BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BLOCKED)); | |
1118 | ||
1119 | return wanted <= ocfs2_highest_compat_lock_level(lockres->l_blocking); | |
1120 | } | |
1121 | ||
1122 | static void ocfs2_init_mask_waiter(struct ocfs2_mask_waiter *mw) | |
1123 | { | |
1124 | INIT_LIST_HEAD(&mw->mw_item); | |
1125 | init_completion(&mw->mw_complete); | |
8ddb7b00 | 1126 | ocfs2_init_start_time(mw); |
ccd979bd MF |
1127 | } |
1128 | ||
1129 | static int ocfs2_wait_for_mask(struct ocfs2_mask_waiter *mw) | |
1130 | { | |
1131 | wait_for_completion(&mw->mw_complete); | |
1132 | /* Re-arm the completion in case we want to wait on it again */ | |
1133 | INIT_COMPLETION(mw->mw_complete); | |
1134 | return mw->mw_status; | |
1135 | } | |
1136 | ||
1137 | static void lockres_add_mask_waiter(struct ocfs2_lock_res *lockres, | |
1138 | struct ocfs2_mask_waiter *mw, | |
1139 | unsigned long mask, | |
1140 | unsigned long goal) | |
1141 | { | |
1142 | BUG_ON(!list_empty(&mw->mw_item)); | |
1143 | ||
1144 | assert_spin_locked(&lockres->l_lock); | |
1145 | ||
1146 | list_add_tail(&mw->mw_item, &lockres->l_mask_waiters); | |
1147 | mw->mw_mask = mask; | |
1148 | mw->mw_goal = goal; | |
1149 | } | |
1150 | ||
1151 | /* returns 0 if the mw that was removed was already satisfied, -EBUSY | |
1152 | * if the mask still hadn't reached its goal */ | |
1153 | static int lockres_remove_mask_waiter(struct ocfs2_lock_res *lockres, | |
1154 | struct ocfs2_mask_waiter *mw) | |
1155 | { | |
1156 | unsigned long flags; | |
1157 | int ret = 0; | |
1158 | ||
1159 | spin_lock_irqsave(&lockres->l_lock, flags); | |
1160 | if (!list_empty(&mw->mw_item)) { | |
1161 | if ((lockres->l_flags & mw->mw_mask) != mw->mw_goal) | |
1162 | ret = -EBUSY; | |
1163 | ||
1164 | list_del_init(&mw->mw_item); | |
1165 | init_completion(&mw->mw_complete); | |
1166 | } | |
1167 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
1168 | ||
1169 | return ret; | |
1170 | ||
1171 | } | |
1172 | ||
cf8e06f1 MF |
1173 | static int ocfs2_wait_for_mask_interruptible(struct ocfs2_mask_waiter *mw, |
1174 | struct ocfs2_lock_res *lockres) | |
1175 | { | |
1176 | int ret; | |
1177 | ||
1178 | ret = wait_for_completion_interruptible(&mw->mw_complete); | |
1179 | if (ret) | |
1180 | lockres_remove_mask_waiter(lockres, mw); | |
1181 | else | |
1182 | ret = mw->mw_status; | |
1183 | /* Re-arm the completion in case we want to wait on it again */ | |
1184 | INIT_COMPLETION(mw->mw_complete); | |
1185 | return ret; | |
1186 | } | |
1187 | ||
ccd979bd MF |
1188 | static int ocfs2_cluster_lock(struct ocfs2_super *osb, |
1189 | struct ocfs2_lock_res *lockres, | |
1190 | int level, | |
bd3e7610 | 1191 | u32 lkm_flags, |
ccd979bd MF |
1192 | int arg_flags) |
1193 | { | |
1194 | struct ocfs2_mask_waiter mw; | |
ccd979bd MF |
1195 | int wait, catch_signals = !(osb->s_mount_opt & OCFS2_MOUNT_NOINTR); |
1196 | int ret = 0; /* gcc doesn't realize wait = 1 guarantees ret is set */ | |
1197 | unsigned long flags; | |
de551246 | 1198 | unsigned int gen; |
1693a5c0 | 1199 | int noqueue_attempted = 0; |
ccd979bd MF |
1200 | |
1201 | mlog_entry_void(); | |
1202 | ||
1203 | ocfs2_init_mask_waiter(&mw); | |
1204 | ||
b80fc012 | 1205 | if (lockres->l_ops->flags & LOCK_TYPE_USES_LVB) |
bd3e7610 | 1206 | lkm_flags |= DLM_LKF_VALBLK; |
b80fc012 | 1207 | |
ccd979bd MF |
1208 | again: |
1209 | wait = 0; | |
1210 | ||
1211 | if (catch_signals && signal_pending(current)) { | |
1212 | ret = -ERESTARTSYS; | |
1213 | goto out; | |
1214 | } | |
1215 | ||
1216 | spin_lock_irqsave(&lockres->l_lock, flags); | |
1217 | ||
1218 | mlog_bug_on_msg(lockres->l_flags & OCFS2_LOCK_FREEING, | |
1219 | "Cluster lock called on freeing lockres %s! flags " | |
1220 | "0x%lx\n", lockres->l_name, lockres->l_flags); | |
1221 | ||
1222 | /* We only compare against the currently granted level | |
1223 | * here. If the lock is blocked waiting on a downconvert, | |
1224 | * we'll get caught below. */ | |
1225 | if (lockres->l_flags & OCFS2_LOCK_BUSY && | |
1226 | level > lockres->l_level) { | |
1227 | /* is someone sitting in dlm_lock? If so, wait on | |
1228 | * them. */ | |
1229 | lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_BUSY, 0); | |
1230 | wait = 1; | |
1231 | goto unlock; | |
1232 | } | |
1233 | ||
ccd979bd MF |
1234 | if (lockres->l_flags & OCFS2_LOCK_BLOCKED && |
1235 | !ocfs2_may_continue_on_blocked_lock(lockres, level)) { | |
1236 | /* is the lock is currently blocked on behalf of | |
1237 | * another node */ | |
1238 | lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_BLOCKED, 0); | |
1239 | wait = 1; | |
1240 | goto unlock; | |
1241 | } | |
1242 | ||
1243 | if (level > lockres->l_level) { | |
1693a5c0 DT |
1244 | if (noqueue_attempted > 0) { |
1245 | ret = -EAGAIN; | |
1246 | goto unlock; | |
1247 | } | |
1248 | if (lkm_flags & DLM_LKF_NOQUEUE) | |
1249 | noqueue_attempted = 1; | |
1250 | ||
ccd979bd MF |
1251 | if (lockres->l_action != OCFS2_AST_INVALID) |
1252 | mlog(ML_ERROR, "lockres %s has action %u pending\n", | |
1253 | lockres->l_name, lockres->l_action); | |
1254 | ||
019d1b22 MF |
1255 | if (!(lockres->l_flags & OCFS2_LOCK_ATTACHED)) { |
1256 | lockres->l_action = OCFS2_AST_ATTACH; | |
bd3e7610 | 1257 | lkm_flags &= ~DLM_LKF_CONVERT; |
019d1b22 MF |
1258 | } else { |
1259 | lockres->l_action = OCFS2_AST_CONVERT; | |
bd3e7610 | 1260 | lkm_flags |= DLM_LKF_CONVERT; |
019d1b22 MF |
1261 | } |
1262 | ||
ccd979bd MF |
1263 | lockres->l_requested = level; |
1264 | lockres_or_flags(lockres, OCFS2_LOCK_BUSY); | |
de551246 | 1265 | gen = lockres_set_pending(lockres); |
ccd979bd MF |
1266 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
1267 | ||
bd3e7610 JB |
1268 | BUG_ON(level == DLM_LOCK_IV); |
1269 | BUG_ON(level == DLM_LOCK_NL); | |
ccd979bd MF |
1270 | |
1271 | mlog(0, "lock %s, convert from %d to level = %d\n", | |
1272 | lockres->l_name, lockres->l_level, level); | |
1273 | ||
1274 | /* call dlm_lock to upgrade lock now */ | |
4670c46d | 1275 | ret = ocfs2_dlm_lock(osb->cconn, |
7431cd7e JB |
1276 | level, |
1277 | &lockres->l_lksb, | |
1278 | lkm_flags, | |
1279 | lockres->l_name, | |
1280 | OCFS2_LOCK_ID_MAX_LEN - 1, | |
1281 | lockres); | |
de551246 | 1282 | lockres_clear_pending(lockres, gen, osb); |
7431cd7e JB |
1283 | if (ret) { |
1284 | if (!(lkm_flags & DLM_LKF_NOQUEUE) || | |
1285 | (ret != -EAGAIN)) { | |
24ef1815 | 1286 | ocfs2_log_dlm_error("ocfs2_dlm_lock", |
7431cd7e | 1287 | ret, lockres); |
ccd979bd MF |
1288 | } |
1289 | ocfs2_recover_from_dlm_error(lockres, 1); | |
1290 | goto out; | |
1291 | } | |
1292 | ||
24ef1815 | 1293 | mlog(0, "lock %s, successfull return from ocfs2_dlm_lock\n", |
ccd979bd MF |
1294 | lockres->l_name); |
1295 | ||
1296 | /* At this point we've gone inside the dlm and need to | |
1297 | * complete our work regardless. */ | |
1298 | catch_signals = 0; | |
1299 | ||
1300 | /* wait for busy to clear and carry on */ | |
1301 | goto again; | |
1302 | } | |
1303 | ||
1304 | /* Ok, if we get here then we're good to go. */ | |
1305 | ocfs2_inc_holders(lockres, level); | |
1306 | ||
1307 | ret = 0; | |
1308 | unlock: | |
1309 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
1310 | out: | |
1311 | /* | |
1312 | * This is helping work around a lock inversion between the page lock | |
1313 | * and dlm locks. One path holds the page lock while calling aops | |
1314 | * which block acquiring dlm locks. The voting thread holds dlm | |
1315 | * locks while acquiring page locks while down converting data locks. | |
1316 | * This block is helping an aop path notice the inversion and back | |
1317 | * off to unlock its page lock before trying the dlm lock again. | |
1318 | */ | |
1319 | if (wait && arg_flags & OCFS2_LOCK_NONBLOCK && | |
1320 | mw.mw_mask & (OCFS2_LOCK_BUSY|OCFS2_LOCK_BLOCKED)) { | |
1321 | wait = 0; | |
1322 | if (lockres_remove_mask_waiter(lockres, &mw)) | |
1323 | ret = -EAGAIN; | |
1324 | else | |
1325 | goto again; | |
1326 | } | |
1327 | if (wait) { | |
1328 | ret = ocfs2_wait_for_mask(&mw); | |
1329 | if (ret == 0) | |
1330 | goto again; | |
1331 | mlog_errno(ret); | |
1332 | } | |
8ddb7b00 | 1333 | ocfs2_update_lock_stats(lockres, level, &mw, ret); |
ccd979bd MF |
1334 | |
1335 | mlog_exit(ret); | |
1336 | return ret; | |
1337 | } | |
1338 | ||
1339 | static void ocfs2_cluster_unlock(struct ocfs2_super *osb, | |
1340 | struct ocfs2_lock_res *lockres, | |
1341 | int level) | |
1342 | { | |
1343 | unsigned long flags; | |
1344 | ||
1345 | mlog_entry_void(); | |
1346 | spin_lock_irqsave(&lockres->l_lock, flags); | |
1347 | ocfs2_dec_holders(lockres, level); | |
34d024f8 | 1348 | ocfs2_downconvert_on_unlock(osb, lockres); |
ccd979bd MF |
1349 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
1350 | mlog_exit_void(); | |
1351 | } | |
1352 | ||
da66116e AB |
1353 | static int ocfs2_create_new_lock(struct ocfs2_super *osb, |
1354 | struct ocfs2_lock_res *lockres, | |
1355 | int ex, | |
1356 | int local) | |
ccd979bd | 1357 | { |
bd3e7610 | 1358 | int level = ex ? DLM_LOCK_EX : DLM_LOCK_PR; |
ccd979bd | 1359 | unsigned long flags; |
bd3e7610 | 1360 | u32 lkm_flags = local ? DLM_LKF_LOCAL : 0; |
ccd979bd MF |
1361 | |
1362 | spin_lock_irqsave(&lockres->l_lock, flags); | |
1363 | BUG_ON(lockres->l_flags & OCFS2_LOCK_ATTACHED); | |
1364 | lockres_or_flags(lockres, OCFS2_LOCK_LOCAL); | |
1365 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
1366 | ||
24c19ef4 | 1367 | return ocfs2_lock_create(osb, lockres, level, lkm_flags); |
ccd979bd MF |
1368 | } |
1369 | ||
1370 | /* Grants us an EX lock on the data and metadata resources, skipping | |
1371 | * the normal cluster directory lookup. Use this ONLY on newly created | |
1372 | * inodes which other nodes can't possibly see, and which haven't been | |
1373 | * hashed in the inode hash yet. This can give us a good performance | |
1374 | * increase as it'll skip the network broadcast normally associated | |
1375 | * with creating a new lock resource. */ | |
1376 | int ocfs2_create_new_inode_locks(struct inode *inode) | |
1377 | { | |
1378 | int ret; | |
d680efe9 | 1379 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
ccd979bd MF |
1380 | |
1381 | BUG_ON(!inode); | |
1382 | BUG_ON(!ocfs2_inode_is_new(inode)); | |
1383 | ||
1384 | mlog_entry_void(); | |
1385 | ||
b0697053 | 1386 | mlog(0, "Inode %llu\n", (unsigned long long)OCFS2_I(inode)->ip_blkno); |
ccd979bd MF |
1387 | |
1388 | /* NOTE: That we don't increment any of the holder counts, nor | |
1389 | * do we add anything to a journal handle. Since this is | |
1390 | * supposed to be a new inode which the cluster doesn't know | |
1391 | * about yet, there is no need to. As far as the LVB handling | |
1392 | * is concerned, this is basically like acquiring an EX lock | |
1393 | * on a resource which has an invalid one -- we'll set it | |
1394 | * valid when we release the EX. */ | |
1395 | ||
24c19ef4 | 1396 | ret = ocfs2_create_new_lock(osb, &OCFS2_I(inode)->ip_rw_lockres, 1, 1); |
ccd979bd MF |
1397 | if (ret) { |
1398 | mlog_errno(ret); | |
1399 | goto bail; | |
1400 | } | |
1401 | ||
24c19ef4 | 1402 | /* |
bd3e7610 | 1403 | * We don't want to use DLM_LKF_LOCAL on a meta data lock as they |
24c19ef4 MF |
1404 | * don't use a generation in their lock names. |
1405 | */ | |
e63aecb6 | 1406 | ret = ocfs2_create_new_lock(osb, &OCFS2_I(inode)->ip_inode_lockres, 1, 0); |
ccd979bd MF |
1407 | if (ret) { |
1408 | mlog_errno(ret); | |
1409 | goto bail; | |
1410 | } | |
1411 | ||
50008630 TY |
1412 | ret = ocfs2_create_new_lock(osb, &OCFS2_I(inode)->ip_open_lockres, 0, 0); |
1413 | if (ret) { | |
1414 | mlog_errno(ret); | |
1415 | goto bail; | |
1416 | } | |
1417 | ||
ccd979bd MF |
1418 | bail: |
1419 | mlog_exit(ret); | |
1420 | return ret; | |
1421 | } | |
1422 | ||
1423 | int ocfs2_rw_lock(struct inode *inode, int write) | |
1424 | { | |
1425 | int status, level; | |
1426 | struct ocfs2_lock_res *lockres; | |
c271c5c2 | 1427 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
ccd979bd MF |
1428 | |
1429 | BUG_ON(!inode); | |
1430 | ||
1431 | mlog_entry_void(); | |
1432 | ||
b0697053 MF |
1433 | mlog(0, "inode %llu take %s RW lock\n", |
1434 | (unsigned long long)OCFS2_I(inode)->ip_blkno, | |
ccd979bd MF |
1435 | write ? "EXMODE" : "PRMODE"); |
1436 | ||
c271c5c2 SM |
1437 | if (ocfs2_mount_local(osb)) |
1438 | return 0; | |
1439 | ||
ccd979bd MF |
1440 | lockres = &OCFS2_I(inode)->ip_rw_lockres; |
1441 | ||
bd3e7610 | 1442 | level = write ? DLM_LOCK_EX : DLM_LOCK_PR; |
ccd979bd MF |
1443 | |
1444 | status = ocfs2_cluster_lock(OCFS2_SB(inode->i_sb), lockres, level, 0, | |
1445 | 0); | |
1446 | if (status < 0) | |
1447 | mlog_errno(status); | |
1448 | ||
1449 | mlog_exit(status); | |
1450 | return status; | |
1451 | } | |
1452 | ||
1453 | void ocfs2_rw_unlock(struct inode *inode, int write) | |
1454 | { | |
bd3e7610 | 1455 | int level = write ? DLM_LOCK_EX : DLM_LOCK_PR; |
ccd979bd | 1456 | struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_rw_lockres; |
c271c5c2 | 1457 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
ccd979bd MF |
1458 | |
1459 | mlog_entry_void(); | |
1460 | ||
b0697053 MF |
1461 | mlog(0, "inode %llu drop %s RW lock\n", |
1462 | (unsigned long long)OCFS2_I(inode)->ip_blkno, | |
ccd979bd MF |
1463 | write ? "EXMODE" : "PRMODE"); |
1464 | ||
c271c5c2 SM |
1465 | if (!ocfs2_mount_local(osb)) |
1466 | ocfs2_cluster_unlock(OCFS2_SB(inode->i_sb), lockres, level); | |
ccd979bd MF |
1467 | |
1468 | mlog_exit_void(); | |
1469 | } | |
1470 | ||
50008630 TY |
1471 | /* |
1472 | * ocfs2_open_lock always get PR mode lock. | |
1473 | */ | |
1474 | int ocfs2_open_lock(struct inode *inode) | |
1475 | { | |
1476 | int status = 0; | |
1477 | struct ocfs2_lock_res *lockres; | |
1478 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
1479 | ||
1480 | BUG_ON(!inode); | |
1481 | ||
1482 | mlog_entry_void(); | |
1483 | ||
1484 | mlog(0, "inode %llu take PRMODE open lock\n", | |
1485 | (unsigned long long)OCFS2_I(inode)->ip_blkno); | |
1486 | ||
1487 | if (ocfs2_mount_local(osb)) | |
1488 | goto out; | |
1489 | ||
1490 | lockres = &OCFS2_I(inode)->ip_open_lockres; | |
1491 | ||
1492 | status = ocfs2_cluster_lock(OCFS2_SB(inode->i_sb), lockres, | |
bd3e7610 | 1493 | DLM_LOCK_PR, 0, 0); |
50008630 TY |
1494 | if (status < 0) |
1495 | mlog_errno(status); | |
1496 | ||
1497 | out: | |
1498 | mlog_exit(status); | |
1499 | return status; | |
1500 | } | |
1501 | ||
1502 | int ocfs2_try_open_lock(struct inode *inode, int write) | |
1503 | { | |
1504 | int status = 0, level; | |
1505 | struct ocfs2_lock_res *lockres; | |
1506 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
1507 | ||
1508 | BUG_ON(!inode); | |
1509 | ||
1510 | mlog_entry_void(); | |
1511 | ||
1512 | mlog(0, "inode %llu try to take %s open lock\n", | |
1513 | (unsigned long long)OCFS2_I(inode)->ip_blkno, | |
1514 | write ? "EXMODE" : "PRMODE"); | |
1515 | ||
1516 | if (ocfs2_mount_local(osb)) | |
1517 | goto out; | |
1518 | ||
1519 | lockres = &OCFS2_I(inode)->ip_open_lockres; | |
1520 | ||
bd3e7610 | 1521 | level = write ? DLM_LOCK_EX : DLM_LOCK_PR; |
50008630 TY |
1522 | |
1523 | /* | |
1524 | * The file system may already holding a PRMODE/EXMODE open lock. | |
bd3e7610 | 1525 | * Since we pass DLM_LKF_NOQUEUE, the request won't block waiting on |
50008630 TY |
1526 | * other nodes and the -EAGAIN will indicate to the caller that |
1527 | * this inode is still in use. | |
1528 | */ | |
1529 | status = ocfs2_cluster_lock(OCFS2_SB(inode->i_sb), lockres, | |
bd3e7610 | 1530 | level, DLM_LKF_NOQUEUE, 0); |
50008630 TY |
1531 | |
1532 | out: | |
1533 | mlog_exit(status); | |
1534 | return status; | |
1535 | } | |
1536 | ||
1537 | /* | |
1538 | * ocfs2_open_unlock unlock PR and EX mode open locks. | |
1539 | */ | |
1540 | void ocfs2_open_unlock(struct inode *inode) | |
1541 | { | |
1542 | struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_open_lockres; | |
1543 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
1544 | ||
1545 | mlog_entry_void(); | |
1546 | ||
1547 | mlog(0, "inode %llu drop open lock\n", | |
1548 | (unsigned long long)OCFS2_I(inode)->ip_blkno); | |
1549 | ||
1550 | if (ocfs2_mount_local(osb)) | |
1551 | goto out; | |
1552 | ||
1553 | if(lockres->l_ro_holders) | |
1554 | ocfs2_cluster_unlock(OCFS2_SB(inode->i_sb), lockres, | |
bd3e7610 | 1555 | DLM_LOCK_PR); |
50008630 TY |
1556 | if(lockres->l_ex_holders) |
1557 | ocfs2_cluster_unlock(OCFS2_SB(inode->i_sb), lockres, | |
bd3e7610 | 1558 | DLM_LOCK_EX); |
50008630 TY |
1559 | |
1560 | out: | |
1561 | mlog_exit_void(); | |
1562 | } | |
1563 | ||
cf8e06f1 MF |
1564 | static int ocfs2_flock_handle_signal(struct ocfs2_lock_res *lockres, |
1565 | int level) | |
1566 | { | |
1567 | int ret; | |
1568 | struct ocfs2_super *osb = ocfs2_get_lockres_osb(lockres); | |
1569 | unsigned long flags; | |
1570 | struct ocfs2_mask_waiter mw; | |
1571 | ||
1572 | ocfs2_init_mask_waiter(&mw); | |
1573 | ||
1574 | retry_cancel: | |
1575 | spin_lock_irqsave(&lockres->l_lock, flags); | |
1576 | if (lockres->l_flags & OCFS2_LOCK_BUSY) { | |
1577 | ret = ocfs2_prepare_cancel_convert(osb, lockres); | |
1578 | if (ret) { | |
1579 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
1580 | ret = ocfs2_cancel_convert(osb, lockres); | |
1581 | if (ret < 0) { | |
1582 | mlog_errno(ret); | |
1583 | goto out; | |
1584 | } | |
1585 | goto retry_cancel; | |
1586 | } | |
1587 | lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_BUSY, 0); | |
1588 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
1589 | ||
1590 | ocfs2_wait_for_mask(&mw); | |
1591 | goto retry_cancel; | |
1592 | } | |
1593 | ||
1594 | ret = -ERESTARTSYS; | |
1595 | /* | |
1596 | * We may still have gotten the lock, in which case there's no | |
1597 | * point to restarting the syscall. | |
1598 | */ | |
1599 | if (lockres->l_level == level) | |
1600 | ret = 0; | |
1601 | ||
1602 | mlog(0, "Cancel returning %d. flags: 0x%lx, level: %d, act: %d\n", ret, | |
1603 | lockres->l_flags, lockres->l_level, lockres->l_action); | |
1604 | ||
1605 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
1606 | ||
1607 | out: | |
1608 | return ret; | |
1609 | } | |
1610 | ||
1611 | /* | |
1612 | * ocfs2_file_lock() and ocfs2_file_unlock() map to a single pair of | |
1613 | * flock() calls. The locking approach this requires is sufficiently | |
1614 | * different from all other cluster lock types that we implement a | |
1615 | * seperate path to the "low-level" dlm calls. In particular: | |
1616 | * | |
1617 | * - No optimization of lock levels is done - we take at exactly | |
1618 | * what's been requested. | |
1619 | * | |
1620 | * - No lock caching is employed. We immediately downconvert to | |
1621 | * no-lock at unlock time. This also means flock locks never go on | |
1622 | * the blocking list). | |
1623 | * | |
1624 | * - Since userspace can trivially deadlock itself with flock, we make | |
1625 | * sure to allow cancellation of a misbehaving applications flock() | |
1626 | * request. | |
1627 | * | |
1628 | * - Access to any flock lockres doesn't require concurrency, so we | |
1629 | * can simplify the code by requiring the caller to guarantee | |
1630 | * serialization of dlmglue flock calls. | |
1631 | */ | |
1632 | int ocfs2_file_lock(struct file *file, int ex, int trylock) | |
1633 | { | |
e988cf1c MF |
1634 | int ret, level = ex ? DLM_LOCK_EX : DLM_LOCK_PR; |
1635 | unsigned int lkm_flags = trylock ? DLM_LKF_NOQUEUE : 0; | |
cf8e06f1 MF |
1636 | unsigned long flags; |
1637 | struct ocfs2_file_private *fp = file->private_data; | |
1638 | struct ocfs2_lock_res *lockres = &fp->fp_flock; | |
1639 | struct ocfs2_super *osb = OCFS2_SB(file->f_mapping->host->i_sb); | |
1640 | struct ocfs2_mask_waiter mw; | |
1641 | ||
1642 | ocfs2_init_mask_waiter(&mw); | |
1643 | ||
1644 | if ((lockres->l_flags & OCFS2_LOCK_BUSY) || | |
bd3e7610 | 1645 | (lockres->l_level > DLM_LOCK_NL)) { |
cf8e06f1 MF |
1646 | mlog(ML_ERROR, |
1647 | "File lock \"%s\" has busy or locked state: flags: 0x%lx, " | |
1648 | "level: %u\n", lockres->l_name, lockres->l_flags, | |
1649 | lockres->l_level); | |
1650 | return -EINVAL; | |
1651 | } | |
1652 | ||
1653 | spin_lock_irqsave(&lockres->l_lock, flags); | |
1654 | if (!(lockres->l_flags & OCFS2_LOCK_ATTACHED)) { | |
1655 | lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_BUSY, 0); | |
1656 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
1657 | ||
1658 | /* | |
1659 | * Get the lock at NLMODE to start - that way we | |
1660 | * can cancel the upconvert request if need be. | |
1661 | */ | |
e988cf1c | 1662 | ret = ocfs2_lock_create(osb, lockres, DLM_LOCK_NL, 0); |
cf8e06f1 MF |
1663 | if (ret < 0) { |
1664 | mlog_errno(ret); | |
1665 | goto out; | |
1666 | } | |
1667 | ||
1668 | ret = ocfs2_wait_for_mask(&mw); | |
1669 | if (ret) { | |
1670 | mlog_errno(ret); | |
1671 | goto out; | |
1672 | } | |
1673 | spin_lock_irqsave(&lockres->l_lock, flags); | |
1674 | } | |
1675 | ||
1676 | lockres->l_action = OCFS2_AST_CONVERT; | |
e988cf1c | 1677 | lkm_flags |= DLM_LKF_CONVERT; |
cf8e06f1 MF |
1678 | lockres->l_requested = level; |
1679 | lockres_or_flags(lockres, OCFS2_LOCK_BUSY); | |
1680 | ||
1681 | lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_BUSY, 0); | |
1682 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
1683 | ||
4670c46d | 1684 | ret = ocfs2_dlm_lock(osb->cconn, level, &lockres->l_lksb, lkm_flags, |
24ef1815 JB |
1685 | lockres->l_name, OCFS2_LOCK_ID_MAX_LEN - 1, |
1686 | lockres); | |
7431cd7e JB |
1687 | if (ret) { |
1688 | if (!trylock || (ret != -EAGAIN)) { | |
24ef1815 | 1689 | ocfs2_log_dlm_error("ocfs2_dlm_lock", ret, lockres); |
cf8e06f1 MF |
1690 | ret = -EINVAL; |
1691 | } | |
1692 | ||
1693 | ocfs2_recover_from_dlm_error(lockres, 1); | |
1694 | lockres_remove_mask_waiter(lockres, &mw); | |
1695 | goto out; | |
1696 | } | |
1697 | ||
1698 | ret = ocfs2_wait_for_mask_interruptible(&mw, lockres); | |
1699 | if (ret == -ERESTARTSYS) { | |
1700 | /* | |
1701 | * Userspace can cause deadlock itself with | |
1702 | * flock(). Current behavior locally is to allow the | |
1703 | * deadlock, but abort the system call if a signal is | |
1704 | * received. We follow this example, otherwise a | |
1705 | * poorly written program could sit in kernel until | |
1706 | * reboot. | |
1707 | * | |
1708 | * Handling this is a bit more complicated for Ocfs2 | |
1709 | * though. We can't exit this function with an | |
1710 | * outstanding lock request, so a cancel convert is | |
1711 | * required. We intentionally overwrite 'ret' - if the | |
1712 | * cancel fails and the lock was granted, it's easier | |
1713 | * to just bubble sucess back up to the user. | |
1714 | */ | |
1715 | ret = ocfs2_flock_handle_signal(lockres, level); | |
1693a5c0 DT |
1716 | } else if (!ret && (level > lockres->l_level)) { |
1717 | /* Trylock failed asynchronously */ | |
1718 | BUG_ON(!trylock); | |
1719 | ret = -EAGAIN; | |
cf8e06f1 MF |
1720 | } |
1721 | ||
1722 | out: | |
1723 | ||
1724 | mlog(0, "Lock: \"%s\" ex: %d, trylock: %d, returns: %d\n", | |
1725 | lockres->l_name, ex, trylock, ret); | |
1726 | return ret; | |
1727 | } | |
1728 | ||
1729 | void ocfs2_file_unlock(struct file *file) | |
1730 | { | |
1731 | int ret; | |
de551246 | 1732 | unsigned int gen; |
cf8e06f1 MF |
1733 | unsigned long flags; |
1734 | struct ocfs2_file_private *fp = file->private_data; | |
1735 | struct ocfs2_lock_res *lockres = &fp->fp_flock; | |
1736 | struct ocfs2_super *osb = OCFS2_SB(file->f_mapping->host->i_sb); | |
1737 | struct ocfs2_mask_waiter mw; | |
1738 | ||
1739 | ocfs2_init_mask_waiter(&mw); | |
1740 | ||
1741 | if (!(lockres->l_flags & OCFS2_LOCK_ATTACHED)) | |
1742 | return; | |
1743 | ||
e988cf1c | 1744 | if (lockres->l_level == DLM_LOCK_NL) |
cf8e06f1 MF |
1745 | return; |
1746 | ||
1747 | mlog(0, "Unlock: \"%s\" flags: 0x%lx, level: %d, act: %d\n", | |
1748 | lockres->l_name, lockres->l_flags, lockres->l_level, | |
1749 | lockres->l_action); | |
1750 | ||
1751 | spin_lock_irqsave(&lockres->l_lock, flags); | |
1752 | /* | |
1753 | * Fake a blocking ast for the downconvert code. | |
1754 | */ | |
1755 | lockres_or_flags(lockres, OCFS2_LOCK_BLOCKED); | |
bd3e7610 | 1756 | lockres->l_blocking = DLM_LOCK_EX; |
cf8e06f1 | 1757 | |
e988cf1c | 1758 | gen = ocfs2_prepare_downconvert(lockres, DLM_LOCK_NL); |
cf8e06f1 MF |
1759 | lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_BUSY, 0); |
1760 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
1761 | ||
e988cf1c | 1762 | ret = ocfs2_downconvert_lock(osb, lockres, DLM_LOCK_NL, 0, gen); |
cf8e06f1 MF |
1763 | if (ret) { |
1764 | mlog_errno(ret); | |
1765 | return; | |
1766 | } | |
1767 | ||
1768 | ret = ocfs2_wait_for_mask(&mw); | |
1769 | if (ret) | |
1770 | mlog_errno(ret); | |
1771 | } | |
1772 | ||
34d024f8 MF |
1773 | static void ocfs2_downconvert_on_unlock(struct ocfs2_super *osb, |
1774 | struct ocfs2_lock_res *lockres) | |
ccd979bd MF |
1775 | { |
1776 | int kick = 0; | |
1777 | ||
1778 | mlog_entry_void(); | |
1779 | ||
1780 | /* If we know that another node is waiting on our lock, kick | |
34d024f8 | 1781 | * the downconvert thread * pre-emptively when we reach a release |
ccd979bd MF |
1782 | * condition. */ |
1783 | if (lockres->l_flags & OCFS2_LOCK_BLOCKED) { | |
1784 | switch(lockres->l_blocking) { | |
bd3e7610 | 1785 | case DLM_LOCK_EX: |
ccd979bd MF |
1786 | if (!lockres->l_ex_holders && !lockres->l_ro_holders) |
1787 | kick = 1; | |
1788 | break; | |
bd3e7610 | 1789 | case DLM_LOCK_PR: |
ccd979bd MF |
1790 | if (!lockres->l_ex_holders) |
1791 | kick = 1; | |
1792 | break; | |
1793 | default: | |
1794 | BUG(); | |
1795 | } | |
1796 | } | |
1797 | ||
1798 | if (kick) | |
34d024f8 | 1799 | ocfs2_wake_downconvert_thread(osb); |
ccd979bd MF |
1800 | |
1801 | mlog_exit_void(); | |
1802 | } | |
1803 | ||
ccd979bd MF |
1804 | #define OCFS2_SEC_BITS 34 |
1805 | #define OCFS2_SEC_SHIFT (64 - 34) | |
1806 | #define OCFS2_NSEC_MASK ((1ULL << OCFS2_SEC_SHIFT) - 1) | |
1807 | ||
1808 | /* LVB only has room for 64 bits of time here so we pack it for | |
1809 | * now. */ | |
1810 | static u64 ocfs2_pack_timespec(struct timespec *spec) | |
1811 | { | |
1812 | u64 res; | |
1813 | u64 sec = spec->tv_sec; | |
1814 | u32 nsec = spec->tv_nsec; | |
1815 | ||
1816 | res = (sec << OCFS2_SEC_SHIFT) | (nsec & OCFS2_NSEC_MASK); | |
1817 | ||
1818 | return res; | |
1819 | } | |
1820 | ||
1821 | /* Call this with the lockres locked. I am reasonably sure we don't | |
1822 | * need ip_lock in this function as anyone who would be changing those | |
e63aecb6 | 1823 | * values is supposed to be blocked in ocfs2_inode_lock right now. */ |
ccd979bd MF |
1824 | static void __ocfs2_stuff_meta_lvb(struct inode *inode) |
1825 | { | |
1826 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
e63aecb6 | 1827 | struct ocfs2_lock_res *lockres = &oi->ip_inode_lockres; |
ccd979bd MF |
1828 | struct ocfs2_meta_lvb *lvb; |
1829 | ||
1830 | mlog_entry_void(); | |
1831 | ||
8f2c9c1b | 1832 | lvb = (struct ocfs2_meta_lvb *)ocfs2_dlm_lvb(&lockres->l_lksb); |
ccd979bd | 1833 | |
24c19ef4 MF |
1834 | /* |
1835 | * Invalidate the LVB of a deleted inode - this way other | |
1836 | * nodes are forced to go to disk and discover the new inode | |
1837 | * status. | |
1838 | */ | |
1839 | if (oi->ip_flags & OCFS2_INODE_DELETED) { | |
1840 | lvb->lvb_version = 0; | |
1841 | goto out; | |
1842 | } | |
1843 | ||
4d3b83f7 | 1844 | lvb->lvb_version = OCFS2_LVB_VERSION; |
ccd979bd MF |
1845 | lvb->lvb_isize = cpu_to_be64(i_size_read(inode)); |
1846 | lvb->lvb_iclusters = cpu_to_be32(oi->ip_clusters); | |
1847 | lvb->lvb_iuid = cpu_to_be32(inode->i_uid); | |
1848 | lvb->lvb_igid = cpu_to_be32(inode->i_gid); | |
1849 | lvb->lvb_imode = cpu_to_be16(inode->i_mode); | |
1850 | lvb->lvb_inlink = cpu_to_be16(inode->i_nlink); | |
1851 | lvb->lvb_iatime_packed = | |
1852 | cpu_to_be64(ocfs2_pack_timespec(&inode->i_atime)); | |
1853 | lvb->lvb_ictime_packed = | |
1854 | cpu_to_be64(ocfs2_pack_timespec(&inode->i_ctime)); | |
1855 | lvb->lvb_imtime_packed = | |
1856 | cpu_to_be64(ocfs2_pack_timespec(&inode->i_mtime)); | |
ca4d147e | 1857 | lvb->lvb_iattr = cpu_to_be32(oi->ip_attr); |
15b1e36b | 1858 | lvb->lvb_idynfeatures = cpu_to_be16(oi->ip_dyn_features); |
f9e2d82e | 1859 | lvb->lvb_igeneration = cpu_to_be32(inode->i_generation); |
ccd979bd | 1860 | |
24c19ef4 | 1861 | out: |
ccd979bd MF |
1862 | mlog_meta_lvb(0, lockres); |
1863 | ||
1864 | mlog_exit_void(); | |
1865 | } | |
1866 | ||
1867 | static void ocfs2_unpack_timespec(struct timespec *spec, | |
1868 | u64 packed_time) | |
1869 | { | |
1870 | spec->tv_sec = packed_time >> OCFS2_SEC_SHIFT; | |
1871 | spec->tv_nsec = packed_time & OCFS2_NSEC_MASK; | |
1872 | } | |
1873 | ||
1874 | static void ocfs2_refresh_inode_from_lvb(struct inode *inode) | |
1875 | { | |
1876 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
e63aecb6 | 1877 | struct ocfs2_lock_res *lockres = &oi->ip_inode_lockres; |
ccd979bd MF |
1878 | struct ocfs2_meta_lvb *lvb; |
1879 | ||
1880 | mlog_entry_void(); | |
1881 | ||
1882 | mlog_meta_lvb(0, lockres); | |
1883 | ||
8f2c9c1b | 1884 | lvb = (struct ocfs2_meta_lvb *)ocfs2_dlm_lvb(&lockres->l_lksb); |
ccd979bd MF |
1885 | |
1886 | /* We're safe here without the lockres lock... */ | |
1887 | spin_lock(&oi->ip_lock); | |
1888 | oi->ip_clusters = be32_to_cpu(lvb->lvb_iclusters); | |
1889 | i_size_write(inode, be64_to_cpu(lvb->lvb_isize)); | |
1890 | ||
ca4d147e | 1891 | oi->ip_attr = be32_to_cpu(lvb->lvb_iattr); |
15b1e36b | 1892 | oi->ip_dyn_features = be16_to_cpu(lvb->lvb_idynfeatures); |
ca4d147e HP |
1893 | ocfs2_set_inode_flags(inode); |
1894 | ||
ccd979bd MF |
1895 | /* fast-symlinks are a special case */ |
1896 | if (S_ISLNK(inode->i_mode) && !oi->ip_clusters) | |
1897 | inode->i_blocks = 0; | |
1898 | else | |
8110b073 | 1899 | inode->i_blocks = ocfs2_inode_sector_count(inode); |
ccd979bd MF |
1900 | |
1901 | inode->i_uid = be32_to_cpu(lvb->lvb_iuid); | |
1902 | inode->i_gid = be32_to_cpu(lvb->lvb_igid); | |
1903 | inode->i_mode = be16_to_cpu(lvb->lvb_imode); | |
1904 | inode->i_nlink = be16_to_cpu(lvb->lvb_inlink); | |
1905 | ocfs2_unpack_timespec(&inode->i_atime, | |
1906 | be64_to_cpu(lvb->lvb_iatime_packed)); | |
1907 | ocfs2_unpack_timespec(&inode->i_mtime, | |
1908 | be64_to_cpu(lvb->lvb_imtime_packed)); | |
1909 | ocfs2_unpack_timespec(&inode->i_ctime, | |
1910 | be64_to_cpu(lvb->lvb_ictime_packed)); | |
1911 | spin_unlock(&oi->ip_lock); | |
1912 | ||
1913 | mlog_exit_void(); | |
1914 | } | |
1915 | ||
f9e2d82e MF |
1916 | static inline int ocfs2_meta_lvb_is_trustable(struct inode *inode, |
1917 | struct ocfs2_lock_res *lockres) | |
ccd979bd | 1918 | { |
8f2c9c1b JB |
1919 | struct ocfs2_meta_lvb *lvb = |
1920 | (struct ocfs2_meta_lvb *)ocfs2_dlm_lvb(&lockres->l_lksb); | |
ccd979bd | 1921 | |
f9e2d82e MF |
1922 | if (lvb->lvb_version == OCFS2_LVB_VERSION |
1923 | && be32_to_cpu(lvb->lvb_igeneration) == inode->i_generation) | |
ccd979bd MF |
1924 | return 1; |
1925 | return 0; | |
1926 | } | |
1927 | ||
1928 | /* Determine whether a lock resource needs to be refreshed, and | |
1929 | * arbitrate who gets to refresh it. | |
1930 | * | |
1931 | * 0 means no refresh needed. | |
1932 | * | |
1933 | * > 0 means you need to refresh this and you MUST call | |
1934 | * ocfs2_complete_lock_res_refresh afterwards. */ | |
1935 | static int ocfs2_should_refresh_lock_res(struct ocfs2_lock_res *lockres) | |
1936 | { | |
1937 | unsigned long flags; | |
1938 | int status = 0; | |
1939 | ||
1940 | mlog_entry_void(); | |
1941 | ||
1942 | refresh_check: | |
1943 | spin_lock_irqsave(&lockres->l_lock, flags); | |
1944 | if (!(lockres->l_flags & OCFS2_LOCK_NEEDS_REFRESH)) { | |
1945 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
1946 | goto bail; | |
1947 | } | |
1948 | ||
1949 | if (lockres->l_flags & OCFS2_LOCK_REFRESHING) { | |
1950 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
1951 | ||
1952 | ocfs2_wait_on_refreshing_lock(lockres); | |
1953 | goto refresh_check; | |
1954 | } | |
1955 | ||
1956 | /* Ok, I'll be the one to refresh this lock. */ | |
1957 | lockres_or_flags(lockres, OCFS2_LOCK_REFRESHING); | |
1958 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
1959 | ||
1960 | status = 1; | |
1961 | bail: | |
1962 | mlog_exit(status); | |
1963 | return status; | |
1964 | } | |
1965 | ||
1966 | /* If status is non zero, I'll mark it as not being in refresh | |
1967 | * anymroe, but i won't clear the needs refresh flag. */ | |
1968 | static inline void ocfs2_complete_lock_res_refresh(struct ocfs2_lock_res *lockres, | |
1969 | int status) | |
1970 | { | |
1971 | unsigned long flags; | |
1972 | mlog_entry_void(); | |
1973 | ||
1974 | spin_lock_irqsave(&lockres->l_lock, flags); | |
1975 | lockres_clear_flags(lockres, OCFS2_LOCK_REFRESHING); | |
1976 | if (!status) | |
1977 | lockres_clear_flags(lockres, OCFS2_LOCK_NEEDS_REFRESH); | |
1978 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
1979 | ||
1980 | wake_up(&lockres->l_event); | |
1981 | ||
1982 | mlog_exit_void(); | |
1983 | } | |
1984 | ||
1985 | /* may or may not return a bh if it went to disk. */ | |
e63aecb6 | 1986 | static int ocfs2_inode_lock_update(struct inode *inode, |
ccd979bd MF |
1987 | struct buffer_head **bh) |
1988 | { | |
1989 | int status = 0; | |
1990 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
e63aecb6 | 1991 | struct ocfs2_lock_res *lockres = &oi->ip_inode_lockres; |
ccd979bd | 1992 | struct ocfs2_dinode *fe; |
c271c5c2 | 1993 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
ccd979bd MF |
1994 | |
1995 | mlog_entry_void(); | |
1996 | ||
be9e986b MF |
1997 | if (ocfs2_mount_local(osb)) |
1998 | goto bail; | |
1999 | ||
ccd979bd MF |
2000 | spin_lock(&oi->ip_lock); |
2001 | if (oi->ip_flags & OCFS2_INODE_DELETED) { | |
b0697053 | 2002 | mlog(0, "Orphaned inode %llu was deleted while we " |
ccd979bd | 2003 | "were waiting on a lock. ip_flags = 0x%x\n", |
b0697053 | 2004 | (unsigned long long)oi->ip_blkno, oi->ip_flags); |
ccd979bd MF |
2005 | spin_unlock(&oi->ip_lock); |
2006 | status = -ENOENT; | |
2007 | goto bail; | |
2008 | } | |
2009 | spin_unlock(&oi->ip_lock); | |
2010 | ||
be9e986b MF |
2011 | if (!ocfs2_should_refresh_lock_res(lockres)) |
2012 | goto bail; | |
ccd979bd MF |
2013 | |
2014 | /* This will discard any caching information we might have had | |
2015 | * for the inode metadata. */ | |
2016 | ocfs2_metadata_cache_purge(inode); | |
2017 | ||
83418978 MF |
2018 | ocfs2_extent_map_trunc(inode, 0); |
2019 | ||
be9e986b | 2020 | if (ocfs2_meta_lvb_is_trustable(inode, lockres)) { |
b0697053 MF |
2021 | mlog(0, "Trusting LVB on inode %llu\n", |
2022 | (unsigned long long)oi->ip_blkno); | |
ccd979bd MF |
2023 | ocfs2_refresh_inode_from_lvb(inode); |
2024 | } else { | |
2025 | /* Boo, we have to go to disk. */ | |
2026 | /* read bh, cast, ocfs2_refresh_inode */ | |
0fcaa56a | 2027 | status = ocfs2_read_block(inode, oi->ip_blkno, bh); |
ccd979bd MF |
2028 | if (status < 0) { |
2029 | mlog_errno(status); | |
2030 | goto bail_refresh; | |
2031 | } | |
2032 | fe = (struct ocfs2_dinode *) (*bh)->b_data; | |
2033 | ||
2034 | /* This is a good chance to make sure we're not | |
2035 | * locking an invalid object. | |
2036 | * | |
2037 | * We bug on a stale inode here because we checked | |
2038 | * above whether it was wiped from disk. The wiping | |
2039 | * node provides a guarantee that we receive that | |
2040 | * message and can mark the inode before dropping any | |
2041 | * locks associated with it. */ | |
2042 | if (!OCFS2_IS_VALID_DINODE(fe)) { | |
2043 | OCFS2_RO_ON_INVALID_DINODE(inode->i_sb, fe); | |
2044 | status = -EIO; | |
2045 | goto bail_refresh; | |
2046 | } | |
2047 | mlog_bug_on_msg(inode->i_generation != | |
2048 | le32_to_cpu(fe->i_generation), | |
b0697053 | 2049 | "Invalid dinode %llu disk generation: %u " |
ccd979bd | 2050 | "inode->i_generation: %u\n", |
b0697053 MF |
2051 | (unsigned long long)oi->ip_blkno, |
2052 | le32_to_cpu(fe->i_generation), | |
ccd979bd MF |
2053 | inode->i_generation); |
2054 | mlog_bug_on_msg(le64_to_cpu(fe->i_dtime) || | |
2055 | !(fe->i_flags & cpu_to_le32(OCFS2_VALID_FL)), | |
b0697053 MF |
2056 | "Stale dinode %llu dtime: %llu flags: 0x%x\n", |
2057 | (unsigned long long)oi->ip_blkno, | |
2058 | (unsigned long long)le64_to_cpu(fe->i_dtime), | |
ccd979bd MF |
2059 | le32_to_cpu(fe->i_flags)); |
2060 | ||
2061 | ocfs2_refresh_inode(inode, fe); | |
8ddb7b00 | 2062 | ocfs2_track_lock_refresh(lockres); |
ccd979bd MF |
2063 | } |
2064 | ||
2065 | status = 0; | |
2066 | bail_refresh: | |
be9e986b | 2067 | ocfs2_complete_lock_res_refresh(lockres, status); |
ccd979bd MF |
2068 | bail: |
2069 | mlog_exit(status); | |
2070 | return status; | |
2071 | } | |
2072 | ||
2073 | static int ocfs2_assign_bh(struct inode *inode, | |
2074 | struct buffer_head **ret_bh, | |
2075 | struct buffer_head *passed_bh) | |
2076 | { | |
2077 | int status; | |
2078 | ||
2079 | if (passed_bh) { | |
2080 | /* Ok, the update went to disk for us, use the | |
2081 | * returned bh. */ | |
2082 | *ret_bh = passed_bh; | |
2083 | get_bh(*ret_bh); | |
2084 | ||
2085 | return 0; | |
2086 | } | |
2087 | ||
0fcaa56a | 2088 | status = ocfs2_read_block(inode, OCFS2_I(inode)->ip_blkno, ret_bh); |
ccd979bd MF |
2089 | if (status < 0) |
2090 | mlog_errno(status); | |
2091 | ||
2092 | return status; | |
2093 | } | |
2094 | ||
2095 | /* | |
2096 | * returns < 0 error if the callback will never be called, otherwise | |
2097 | * the result of the lock will be communicated via the callback. | |
2098 | */ | |
e63aecb6 | 2099 | int ocfs2_inode_lock_full(struct inode *inode, |
ccd979bd MF |
2100 | struct buffer_head **ret_bh, |
2101 | int ex, | |
2102 | int arg_flags) | |
2103 | { | |
bd3e7610 JB |
2104 | int status, level, acquired; |
2105 | u32 dlm_flags; | |
c271c5c2 | 2106 | struct ocfs2_lock_res *lockres = NULL; |
ccd979bd MF |
2107 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
2108 | struct buffer_head *local_bh = NULL; | |
2109 | ||
2110 | BUG_ON(!inode); | |
2111 | ||
2112 | mlog_entry_void(); | |
2113 | ||
b0697053 MF |
2114 | mlog(0, "inode %llu, take %s META lock\n", |
2115 | (unsigned long long)OCFS2_I(inode)->ip_blkno, | |
ccd979bd MF |
2116 | ex ? "EXMODE" : "PRMODE"); |
2117 | ||
2118 | status = 0; | |
2119 | acquired = 0; | |
2120 | /* We'll allow faking a readonly metadata lock for | |
2121 | * rodevices. */ | |
2122 | if (ocfs2_is_hard_readonly(osb)) { | |
2123 | if (ex) | |
2124 | status = -EROFS; | |
2125 | goto bail; | |
2126 | } | |
2127 | ||
c271c5c2 SM |
2128 | if (ocfs2_mount_local(osb)) |
2129 | goto local; | |
2130 | ||
ccd979bd | 2131 | if (!(arg_flags & OCFS2_META_LOCK_RECOVERY)) |
553abd04 | 2132 | ocfs2_wait_for_recovery(osb); |
ccd979bd | 2133 | |
e63aecb6 | 2134 | lockres = &OCFS2_I(inode)->ip_inode_lockres; |
bd3e7610 | 2135 | level = ex ? DLM_LOCK_EX : DLM_LOCK_PR; |
ccd979bd MF |
2136 | dlm_flags = 0; |
2137 | if (arg_flags & OCFS2_META_LOCK_NOQUEUE) | |
bd3e7610 | 2138 | dlm_flags |= DLM_LKF_NOQUEUE; |
ccd979bd MF |
2139 | |
2140 | status = ocfs2_cluster_lock(osb, lockres, level, dlm_flags, arg_flags); | |
2141 | if (status < 0) { | |
2142 | if (status != -EAGAIN && status != -EIOCBRETRY) | |
2143 | mlog_errno(status); | |
2144 | goto bail; | |
2145 | } | |
2146 | ||
2147 | /* Notify the error cleanup path to drop the cluster lock. */ | |
2148 | acquired = 1; | |
2149 | ||
2150 | /* We wait twice because a node may have died while we were in | |
2151 | * the lower dlm layers. The second time though, we've | |
2152 | * committed to owning this lock so we don't allow signals to | |
2153 | * abort the operation. */ | |
2154 | if (!(arg_flags & OCFS2_META_LOCK_RECOVERY)) | |
553abd04 | 2155 | ocfs2_wait_for_recovery(osb); |
ccd979bd | 2156 | |
c271c5c2 | 2157 | local: |
24c19ef4 MF |
2158 | /* |
2159 | * We only see this flag if we're being called from | |
2160 | * ocfs2_read_locked_inode(). It means we're locking an inode | |
2161 | * which hasn't been populated yet, so clear the refresh flag | |
2162 | * and let the caller handle it. | |
2163 | */ | |
2164 | if (inode->i_state & I_NEW) { | |
2165 | status = 0; | |
c271c5c2 SM |
2166 | if (lockres) |
2167 | ocfs2_complete_lock_res_refresh(lockres, 0); | |
24c19ef4 MF |
2168 | goto bail; |
2169 | } | |
2170 | ||
ccd979bd | 2171 | /* This is fun. The caller may want a bh back, or it may |
e63aecb6 | 2172 | * not. ocfs2_inode_lock_update definitely wants one in, but |
ccd979bd MF |
2173 | * may or may not read one, depending on what's in the |
2174 | * LVB. The result of all of this is that we've *only* gone to | |
2175 | * disk if we have to, so the complexity is worthwhile. */ | |
e63aecb6 | 2176 | status = ocfs2_inode_lock_update(inode, &local_bh); |
ccd979bd MF |
2177 | if (status < 0) { |
2178 | if (status != -ENOENT) | |
2179 | mlog_errno(status); | |
2180 | goto bail; | |
2181 | } | |
2182 | ||
2183 | if (ret_bh) { | |
2184 | status = ocfs2_assign_bh(inode, ret_bh, local_bh); | |
2185 | if (status < 0) { | |
2186 | mlog_errno(status); | |
2187 | goto bail; | |
2188 | } | |
2189 | } | |
2190 | ||
ccd979bd MF |
2191 | bail: |
2192 | if (status < 0) { | |
2193 | if (ret_bh && (*ret_bh)) { | |
2194 | brelse(*ret_bh); | |
2195 | *ret_bh = NULL; | |
2196 | } | |
2197 | if (acquired) | |
e63aecb6 | 2198 | ocfs2_inode_unlock(inode, ex); |
ccd979bd MF |
2199 | } |
2200 | ||
2201 | if (local_bh) | |
2202 | brelse(local_bh); | |
2203 | ||
2204 | mlog_exit(status); | |
2205 | return status; | |
2206 | } | |
2207 | ||
2208 | /* | |
34d024f8 MF |
2209 | * This is working around a lock inversion between tasks acquiring DLM |
2210 | * locks while holding a page lock and the downconvert thread which | |
2211 | * blocks dlm lock acquiry while acquiring page locks. | |
ccd979bd MF |
2212 | * |
2213 | * ** These _with_page variantes are only intended to be called from aop | |
2214 | * methods that hold page locks and return a very specific *positive* error | |
2215 | * code that aop methods pass up to the VFS -- test for errors with != 0. ** | |
2216 | * | |
34d024f8 MF |
2217 | * The DLM is called such that it returns -EAGAIN if it would have |
2218 | * blocked waiting for the downconvert thread. In that case we unlock | |
2219 | * our page so the downconvert thread can make progress. Once we've | |
2220 | * done this we have to return AOP_TRUNCATED_PAGE so the aop method | |
2221 | * that called us can bubble that back up into the VFS who will then | |
2222 | * immediately retry the aop call. | |
ccd979bd MF |
2223 | * |
2224 | * We do a blocking lock and immediate unlock before returning, though, so that | |
2225 | * the lock has a great chance of being cached on this node by the time the VFS | |
2226 | * calls back to retry the aop. This has a potential to livelock as nodes | |
2227 | * ping locks back and forth, but that's a risk we're willing to take to avoid | |
2228 | * the lock inversion simply. | |
2229 | */ | |
e63aecb6 | 2230 | int ocfs2_inode_lock_with_page(struct inode *inode, |
ccd979bd MF |
2231 | struct buffer_head **ret_bh, |
2232 | int ex, | |
2233 | struct page *page) | |
2234 | { | |
2235 | int ret; | |
2236 | ||
e63aecb6 | 2237 | ret = ocfs2_inode_lock_full(inode, ret_bh, ex, OCFS2_LOCK_NONBLOCK); |
ccd979bd MF |
2238 | if (ret == -EAGAIN) { |
2239 | unlock_page(page); | |
e63aecb6 MF |
2240 | if (ocfs2_inode_lock(inode, ret_bh, ex) == 0) |
2241 | ocfs2_inode_unlock(inode, ex); | |
ccd979bd MF |
2242 | ret = AOP_TRUNCATED_PAGE; |
2243 | } | |
2244 | ||
2245 | return ret; | |
2246 | } | |
2247 | ||
e63aecb6 | 2248 | int ocfs2_inode_lock_atime(struct inode *inode, |
7f1a37e3 TY |
2249 | struct vfsmount *vfsmnt, |
2250 | int *level) | |
2251 | { | |
2252 | int ret; | |
2253 | ||
2254 | mlog_entry_void(); | |
e63aecb6 | 2255 | ret = ocfs2_inode_lock(inode, NULL, 0); |
7f1a37e3 TY |
2256 | if (ret < 0) { |
2257 | mlog_errno(ret); | |
2258 | return ret; | |
2259 | } | |
2260 | ||
2261 | /* | |
2262 | * If we should update atime, we will get EX lock, | |
2263 | * otherwise we just get PR lock. | |
2264 | */ | |
2265 | if (ocfs2_should_update_atime(inode, vfsmnt)) { | |
2266 | struct buffer_head *bh = NULL; | |
2267 | ||
e63aecb6 MF |
2268 | ocfs2_inode_unlock(inode, 0); |
2269 | ret = ocfs2_inode_lock(inode, &bh, 1); | |
7f1a37e3 TY |
2270 | if (ret < 0) { |
2271 | mlog_errno(ret); | |
2272 | return ret; | |
2273 | } | |
2274 | *level = 1; | |
2275 | if (ocfs2_should_update_atime(inode, vfsmnt)) | |
2276 | ocfs2_update_inode_atime(inode, bh); | |
2277 | if (bh) | |
2278 | brelse(bh); | |
2279 | } else | |
2280 | *level = 0; | |
2281 | ||
2282 | mlog_exit(ret); | |
2283 | return ret; | |
2284 | } | |
2285 | ||
e63aecb6 | 2286 | void ocfs2_inode_unlock(struct inode *inode, |
ccd979bd MF |
2287 | int ex) |
2288 | { | |
bd3e7610 | 2289 | int level = ex ? DLM_LOCK_EX : DLM_LOCK_PR; |
e63aecb6 | 2290 | struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_inode_lockres; |
c271c5c2 | 2291 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
ccd979bd MF |
2292 | |
2293 | mlog_entry_void(); | |
2294 | ||
b0697053 MF |
2295 | mlog(0, "inode %llu drop %s META lock\n", |
2296 | (unsigned long long)OCFS2_I(inode)->ip_blkno, | |
ccd979bd MF |
2297 | ex ? "EXMODE" : "PRMODE"); |
2298 | ||
c271c5c2 SM |
2299 | if (!ocfs2_is_hard_readonly(OCFS2_SB(inode->i_sb)) && |
2300 | !ocfs2_mount_local(osb)) | |
ccd979bd MF |
2301 | ocfs2_cluster_unlock(OCFS2_SB(inode->i_sb), lockres, level); |
2302 | ||
2303 | mlog_exit_void(); | |
2304 | } | |
2305 | ||
2306 | int ocfs2_super_lock(struct ocfs2_super *osb, | |
2307 | int ex) | |
2308 | { | |
c271c5c2 | 2309 | int status = 0; |
bd3e7610 | 2310 | int level = ex ? DLM_LOCK_EX : DLM_LOCK_PR; |
ccd979bd | 2311 | struct ocfs2_lock_res *lockres = &osb->osb_super_lockres; |
ccd979bd MF |
2312 | |
2313 | mlog_entry_void(); | |
2314 | ||
2315 | if (ocfs2_is_hard_readonly(osb)) | |
2316 | return -EROFS; | |
2317 | ||
c271c5c2 SM |
2318 | if (ocfs2_mount_local(osb)) |
2319 | goto bail; | |
2320 | ||
ccd979bd MF |
2321 | status = ocfs2_cluster_lock(osb, lockres, level, 0, 0); |
2322 | if (status < 0) { | |
2323 | mlog_errno(status); | |
2324 | goto bail; | |
2325 | } | |
2326 | ||
2327 | /* The super block lock path is really in the best position to | |
2328 | * know when resources covered by the lock need to be | |
2329 | * refreshed, so we do it here. Of course, making sense of | |
2330 | * everything is up to the caller :) */ | |
2331 | status = ocfs2_should_refresh_lock_res(lockres); | |
2332 | if (status < 0) { | |
2333 | mlog_errno(status); | |
2334 | goto bail; | |
2335 | } | |
2336 | if (status) { | |
8e8a4603 | 2337 | status = ocfs2_refresh_slot_info(osb); |
ccd979bd MF |
2338 | |
2339 | ocfs2_complete_lock_res_refresh(lockres, status); | |
2340 | ||
2341 | if (status < 0) | |
2342 | mlog_errno(status); | |
8ddb7b00 | 2343 | ocfs2_track_lock_refresh(lockres); |
ccd979bd MF |
2344 | } |
2345 | bail: | |
2346 | mlog_exit(status); | |
2347 | return status; | |
2348 | } | |
2349 | ||
2350 | void ocfs2_super_unlock(struct ocfs2_super *osb, | |
2351 | int ex) | |
2352 | { | |
bd3e7610 | 2353 | int level = ex ? DLM_LOCK_EX : DLM_LOCK_PR; |
ccd979bd MF |
2354 | struct ocfs2_lock_res *lockres = &osb->osb_super_lockres; |
2355 | ||
c271c5c2 SM |
2356 | if (!ocfs2_mount_local(osb)) |
2357 | ocfs2_cluster_unlock(osb, lockres, level); | |
ccd979bd MF |
2358 | } |
2359 | ||
2360 | int ocfs2_rename_lock(struct ocfs2_super *osb) | |
2361 | { | |
2362 | int status; | |
2363 | struct ocfs2_lock_res *lockres = &osb->osb_rename_lockres; | |
2364 | ||
2365 | if (ocfs2_is_hard_readonly(osb)) | |
2366 | return -EROFS; | |
2367 | ||
c271c5c2 SM |
2368 | if (ocfs2_mount_local(osb)) |
2369 | return 0; | |
2370 | ||
bd3e7610 | 2371 | status = ocfs2_cluster_lock(osb, lockres, DLM_LOCK_EX, 0, 0); |
ccd979bd MF |
2372 | if (status < 0) |
2373 | mlog_errno(status); | |
2374 | ||
2375 | return status; | |
2376 | } | |
2377 | ||
2378 | void ocfs2_rename_unlock(struct ocfs2_super *osb) | |
2379 | { | |
2380 | struct ocfs2_lock_res *lockres = &osb->osb_rename_lockres; | |
2381 | ||
c271c5c2 | 2382 | if (!ocfs2_mount_local(osb)) |
bd3e7610 | 2383 | ocfs2_cluster_unlock(osb, lockres, DLM_LOCK_EX); |
ccd979bd MF |
2384 | } |
2385 | ||
d680efe9 MF |
2386 | int ocfs2_dentry_lock(struct dentry *dentry, int ex) |
2387 | { | |
2388 | int ret; | |
bd3e7610 | 2389 | int level = ex ? DLM_LOCK_EX : DLM_LOCK_PR; |
d680efe9 MF |
2390 | struct ocfs2_dentry_lock *dl = dentry->d_fsdata; |
2391 | struct ocfs2_super *osb = OCFS2_SB(dentry->d_sb); | |
2392 | ||
2393 | BUG_ON(!dl); | |
2394 | ||
2395 | if (ocfs2_is_hard_readonly(osb)) | |
2396 | return -EROFS; | |
2397 | ||
c271c5c2 SM |
2398 | if (ocfs2_mount_local(osb)) |
2399 | return 0; | |
2400 | ||
d680efe9 MF |
2401 | ret = ocfs2_cluster_lock(osb, &dl->dl_lockres, level, 0, 0); |
2402 | if (ret < 0) | |
2403 | mlog_errno(ret); | |
2404 | ||
2405 | return ret; | |
2406 | } | |
2407 | ||
2408 | void ocfs2_dentry_unlock(struct dentry *dentry, int ex) | |
2409 | { | |
bd3e7610 | 2410 | int level = ex ? DLM_LOCK_EX : DLM_LOCK_PR; |
d680efe9 MF |
2411 | struct ocfs2_dentry_lock *dl = dentry->d_fsdata; |
2412 | struct ocfs2_super *osb = OCFS2_SB(dentry->d_sb); | |
2413 | ||
c271c5c2 SM |
2414 | if (!ocfs2_mount_local(osb)) |
2415 | ocfs2_cluster_unlock(osb, &dl->dl_lockres, level); | |
d680efe9 MF |
2416 | } |
2417 | ||
ccd979bd MF |
2418 | /* Reference counting of the dlm debug structure. We want this because |
2419 | * open references on the debug inodes can live on after a mount, so | |
2420 | * we can't rely on the ocfs2_super to always exist. */ | |
2421 | static void ocfs2_dlm_debug_free(struct kref *kref) | |
2422 | { | |
2423 | struct ocfs2_dlm_debug *dlm_debug; | |
2424 | ||
2425 | dlm_debug = container_of(kref, struct ocfs2_dlm_debug, d_refcnt); | |
2426 | ||
2427 | kfree(dlm_debug); | |
2428 | } | |
2429 | ||
2430 | void ocfs2_put_dlm_debug(struct ocfs2_dlm_debug *dlm_debug) | |
2431 | { | |
2432 | if (dlm_debug) | |
2433 | kref_put(&dlm_debug->d_refcnt, ocfs2_dlm_debug_free); | |
2434 | } | |
2435 | ||
2436 | static void ocfs2_get_dlm_debug(struct ocfs2_dlm_debug *debug) | |
2437 | { | |
2438 | kref_get(&debug->d_refcnt); | |
2439 | } | |
2440 | ||
2441 | struct ocfs2_dlm_debug *ocfs2_new_dlm_debug(void) | |
2442 | { | |
2443 | struct ocfs2_dlm_debug *dlm_debug; | |
2444 | ||
2445 | dlm_debug = kmalloc(sizeof(struct ocfs2_dlm_debug), GFP_KERNEL); | |
2446 | if (!dlm_debug) { | |
2447 | mlog_errno(-ENOMEM); | |
2448 | goto out; | |
2449 | } | |
2450 | ||
2451 | kref_init(&dlm_debug->d_refcnt); | |
2452 | INIT_LIST_HEAD(&dlm_debug->d_lockres_tracking); | |
2453 | dlm_debug->d_locking_state = NULL; | |
2454 | out: | |
2455 | return dlm_debug; | |
2456 | } | |
2457 | ||
2458 | /* Access to this is arbitrated for us via seq_file->sem. */ | |
2459 | struct ocfs2_dlm_seq_priv { | |
2460 | struct ocfs2_dlm_debug *p_dlm_debug; | |
2461 | struct ocfs2_lock_res p_iter_res; | |
2462 | struct ocfs2_lock_res p_tmp_res; | |
2463 | }; | |
2464 | ||
2465 | static struct ocfs2_lock_res *ocfs2_dlm_next_res(struct ocfs2_lock_res *start, | |
2466 | struct ocfs2_dlm_seq_priv *priv) | |
2467 | { | |
2468 | struct ocfs2_lock_res *iter, *ret = NULL; | |
2469 | struct ocfs2_dlm_debug *dlm_debug = priv->p_dlm_debug; | |
2470 | ||
2471 | assert_spin_locked(&ocfs2_dlm_tracking_lock); | |
2472 | ||
2473 | list_for_each_entry(iter, &start->l_debug_list, l_debug_list) { | |
2474 | /* discover the head of the list */ | |
2475 | if (&iter->l_debug_list == &dlm_debug->d_lockres_tracking) { | |
2476 | mlog(0, "End of list found, %p\n", ret); | |
2477 | break; | |
2478 | } | |
2479 | ||
2480 | /* We track our "dummy" iteration lockres' by a NULL | |
2481 | * l_ops field. */ | |
2482 | if (iter->l_ops != NULL) { | |
2483 | ret = iter; | |
2484 | break; | |
2485 | } | |
2486 | } | |
2487 | ||
2488 | return ret; | |
2489 | } | |
2490 | ||
2491 | static void *ocfs2_dlm_seq_start(struct seq_file *m, loff_t *pos) | |
2492 | { | |
2493 | struct ocfs2_dlm_seq_priv *priv = m->private; | |
2494 | struct ocfs2_lock_res *iter; | |
2495 | ||
2496 | spin_lock(&ocfs2_dlm_tracking_lock); | |
2497 | iter = ocfs2_dlm_next_res(&priv->p_iter_res, priv); | |
2498 | if (iter) { | |
2499 | /* Since lockres' have the lifetime of their container | |
2500 | * (which can be inodes, ocfs2_supers, etc) we want to | |
2501 | * copy this out to a temporary lockres while still | |
2502 | * under the spinlock. Obviously after this we can't | |
2503 | * trust any pointers on the copy returned, but that's | |
2504 | * ok as the information we want isn't typically held | |
2505 | * in them. */ | |
2506 | priv->p_tmp_res = *iter; | |
2507 | iter = &priv->p_tmp_res; | |
2508 | } | |
2509 | spin_unlock(&ocfs2_dlm_tracking_lock); | |
2510 | ||
2511 | return iter; | |
2512 | } | |
2513 | ||
2514 | static void ocfs2_dlm_seq_stop(struct seq_file *m, void *v) | |
2515 | { | |
2516 | } | |
2517 | ||
2518 | static void *ocfs2_dlm_seq_next(struct seq_file *m, void *v, loff_t *pos) | |
2519 | { | |
2520 | struct ocfs2_dlm_seq_priv *priv = m->private; | |
2521 | struct ocfs2_lock_res *iter = v; | |
2522 | struct ocfs2_lock_res *dummy = &priv->p_iter_res; | |
2523 | ||
2524 | spin_lock(&ocfs2_dlm_tracking_lock); | |
2525 | iter = ocfs2_dlm_next_res(iter, priv); | |
2526 | list_del_init(&dummy->l_debug_list); | |
2527 | if (iter) { | |
2528 | list_add(&dummy->l_debug_list, &iter->l_debug_list); | |
2529 | priv->p_tmp_res = *iter; | |
2530 | iter = &priv->p_tmp_res; | |
2531 | } | |
2532 | spin_unlock(&ocfs2_dlm_tracking_lock); | |
2533 | ||
2534 | return iter; | |
2535 | } | |
2536 | ||
2537 | /* So that debugfs.ocfs2 can determine which format is being used */ | |
8ddb7b00 | 2538 | #define OCFS2_DLM_DEBUG_STR_VERSION 2 |
ccd979bd MF |
2539 | static int ocfs2_dlm_seq_show(struct seq_file *m, void *v) |
2540 | { | |
2541 | int i; | |
2542 | char *lvb; | |
2543 | struct ocfs2_lock_res *lockres = v; | |
2544 | ||
2545 | if (!lockres) | |
2546 | return -EINVAL; | |
2547 | ||
d680efe9 MF |
2548 | seq_printf(m, "0x%x\t", OCFS2_DLM_DEBUG_STR_VERSION); |
2549 | ||
2550 | if (lockres->l_type == OCFS2_LOCK_TYPE_DENTRY) | |
2551 | seq_printf(m, "%.*s%08x\t", OCFS2_DENTRY_LOCK_INO_START - 1, | |
2552 | lockres->l_name, | |
2553 | (unsigned int)ocfs2_get_dentry_lock_ino(lockres)); | |
2554 | else | |
2555 | seq_printf(m, "%.*s\t", OCFS2_LOCK_ID_MAX_LEN, lockres->l_name); | |
2556 | ||
2557 | seq_printf(m, "%d\t" | |
ccd979bd MF |
2558 | "0x%lx\t" |
2559 | "0x%x\t" | |
2560 | "0x%x\t" | |
2561 | "%u\t" | |
2562 | "%u\t" | |
2563 | "%d\t" | |
2564 | "%d\t", | |
ccd979bd MF |
2565 | lockres->l_level, |
2566 | lockres->l_flags, | |
2567 | lockres->l_action, | |
2568 | lockres->l_unlock_action, | |
2569 | lockres->l_ro_holders, | |
2570 | lockres->l_ex_holders, | |
2571 | lockres->l_requested, | |
2572 | lockres->l_blocking); | |
2573 | ||
2574 | /* Dump the raw LVB */ | |
8f2c9c1b | 2575 | lvb = ocfs2_dlm_lvb(&lockres->l_lksb); |
ccd979bd MF |
2576 | for(i = 0; i < DLM_LVB_LEN; i++) |
2577 | seq_printf(m, "0x%x\t", lvb[i]); | |
2578 | ||
8ddb7b00 SM |
2579 | #ifdef CONFIG_OCFS2_FS_STATS |
2580 | # define lock_num_prmode(_l) (_l)->l_lock_num_prmode | |
2581 | # define lock_num_exmode(_l) (_l)->l_lock_num_exmode | |
2582 | # define lock_num_prmode_failed(_l) (_l)->l_lock_num_prmode_failed | |
2583 | # define lock_num_exmode_failed(_l) (_l)->l_lock_num_exmode_failed | |
2584 | # define lock_total_prmode(_l) (_l)->l_lock_total_prmode | |
2585 | # define lock_total_exmode(_l) (_l)->l_lock_total_exmode | |
2586 | # define lock_max_prmode(_l) (_l)->l_lock_max_prmode | |
2587 | # define lock_max_exmode(_l) (_l)->l_lock_max_exmode | |
2588 | # define lock_refresh(_l) (_l)->l_lock_refresh | |
2589 | #else | |
dd25e55e RD |
2590 | # define lock_num_prmode(_l) (0ULL) |
2591 | # define lock_num_exmode(_l) (0ULL) | |
8ddb7b00 SM |
2592 | # define lock_num_prmode_failed(_l) (0) |
2593 | # define lock_num_exmode_failed(_l) (0) | |
dd25e55e RD |
2594 | # define lock_total_prmode(_l) (0ULL) |
2595 | # define lock_total_exmode(_l) (0ULL) | |
8ddb7b00 SM |
2596 | # define lock_max_prmode(_l) (0) |
2597 | # define lock_max_exmode(_l) (0) | |
2598 | # define lock_refresh(_l) (0) | |
2599 | #endif | |
2600 | /* The following seq_print was added in version 2 of this output */ | |
2601 | seq_printf(m, "%llu\t" | |
2602 | "%llu\t" | |
2603 | "%u\t" | |
2604 | "%u\t" | |
2605 | "%llu\t" | |
2606 | "%llu\t" | |
2607 | "%u\t" | |
2608 | "%u\t" | |
2609 | "%u\t", | |
2610 | lock_num_prmode(lockres), | |
2611 | lock_num_exmode(lockres), | |
2612 | lock_num_prmode_failed(lockres), | |
2613 | lock_num_exmode_failed(lockres), | |
2614 | lock_total_prmode(lockres), | |
2615 | lock_total_exmode(lockres), | |
2616 | lock_max_prmode(lockres), | |
2617 | lock_max_exmode(lockres), | |
2618 | lock_refresh(lockres)); | |
2619 | ||
ccd979bd MF |
2620 | /* End the line */ |
2621 | seq_printf(m, "\n"); | |
2622 | return 0; | |
2623 | } | |
2624 | ||
90d99779 | 2625 | static const struct seq_operations ocfs2_dlm_seq_ops = { |
ccd979bd MF |
2626 | .start = ocfs2_dlm_seq_start, |
2627 | .stop = ocfs2_dlm_seq_stop, | |
2628 | .next = ocfs2_dlm_seq_next, | |
2629 | .show = ocfs2_dlm_seq_show, | |
2630 | }; | |
2631 | ||
2632 | static int ocfs2_dlm_debug_release(struct inode *inode, struct file *file) | |
2633 | { | |
2634 | struct seq_file *seq = (struct seq_file *) file->private_data; | |
2635 | struct ocfs2_dlm_seq_priv *priv = seq->private; | |
2636 | struct ocfs2_lock_res *res = &priv->p_iter_res; | |
2637 | ||
2638 | ocfs2_remove_lockres_tracking(res); | |
2639 | ocfs2_put_dlm_debug(priv->p_dlm_debug); | |
2640 | return seq_release_private(inode, file); | |
2641 | } | |
2642 | ||
2643 | static int ocfs2_dlm_debug_open(struct inode *inode, struct file *file) | |
2644 | { | |
2645 | int ret; | |
2646 | struct ocfs2_dlm_seq_priv *priv; | |
2647 | struct seq_file *seq; | |
2648 | struct ocfs2_super *osb; | |
2649 | ||
2650 | priv = kzalloc(sizeof(struct ocfs2_dlm_seq_priv), GFP_KERNEL); | |
2651 | if (!priv) { | |
2652 | ret = -ENOMEM; | |
2653 | mlog_errno(ret); | |
2654 | goto out; | |
2655 | } | |
8e18e294 | 2656 | osb = inode->i_private; |
ccd979bd MF |
2657 | ocfs2_get_dlm_debug(osb->osb_dlm_debug); |
2658 | priv->p_dlm_debug = osb->osb_dlm_debug; | |
2659 | INIT_LIST_HEAD(&priv->p_iter_res.l_debug_list); | |
2660 | ||
2661 | ret = seq_open(file, &ocfs2_dlm_seq_ops); | |
2662 | if (ret) { | |
2663 | kfree(priv); | |
2664 | mlog_errno(ret); | |
2665 | goto out; | |
2666 | } | |
2667 | ||
2668 | seq = (struct seq_file *) file->private_data; | |
2669 | seq->private = priv; | |
2670 | ||
2671 | ocfs2_add_lockres_tracking(&priv->p_iter_res, | |
2672 | priv->p_dlm_debug); | |
2673 | ||
2674 | out: | |
2675 | return ret; | |
2676 | } | |
2677 | ||
4b6f5d20 | 2678 | static const struct file_operations ocfs2_dlm_debug_fops = { |
ccd979bd MF |
2679 | .open = ocfs2_dlm_debug_open, |
2680 | .release = ocfs2_dlm_debug_release, | |
2681 | .read = seq_read, | |
2682 | .llseek = seq_lseek, | |
2683 | }; | |
2684 | ||
2685 | static int ocfs2_dlm_init_debug(struct ocfs2_super *osb) | |
2686 | { | |
2687 | int ret = 0; | |
2688 | struct ocfs2_dlm_debug *dlm_debug = osb->osb_dlm_debug; | |
2689 | ||
2690 | dlm_debug->d_locking_state = debugfs_create_file("locking_state", | |
2691 | S_IFREG|S_IRUSR, | |
2692 | osb->osb_debug_root, | |
2693 | osb, | |
2694 | &ocfs2_dlm_debug_fops); | |
2695 | if (!dlm_debug->d_locking_state) { | |
2696 | ret = -EINVAL; | |
2697 | mlog(ML_ERROR, | |
2698 | "Unable to create locking state debugfs file.\n"); | |
2699 | goto out; | |
2700 | } | |
2701 | ||
2702 | ocfs2_get_dlm_debug(dlm_debug); | |
2703 | out: | |
2704 | return ret; | |
2705 | } | |
2706 | ||
2707 | static void ocfs2_dlm_shutdown_debug(struct ocfs2_super *osb) | |
2708 | { | |
2709 | struct ocfs2_dlm_debug *dlm_debug = osb->osb_dlm_debug; | |
2710 | ||
2711 | if (dlm_debug) { | |
2712 | debugfs_remove(dlm_debug->d_locking_state); | |
2713 | ocfs2_put_dlm_debug(dlm_debug); | |
2714 | } | |
2715 | } | |
2716 | ||
2717 | int ocfs2_dlm_init(struct ocfs2_super *osb) | |
2718 | { | |
c271c5c2 | 2719 | int status = 0; |
4670c46d | 2720 | struct ocfs2_cluster_connection *conn = NULL; |
ccd979bd MF |
2721 | |
2722 | mlog_entry_void(); | |
2723 | ||
0abd6d18 MF |
2724 | if (ocfs2_mount_local(osb)) { |
2725 | osb->node_num = 0; | |
c271c5c2 | 2726 | goto local; |
0abd6d18 | 2727 | } |
c271c5c2 | 2728 | |
ccd979bd MF |
2729 | status = ocfs2_dlm_init_debug(osb); |
2730 | if (status < 0) { | |
2731 | mlog_errno(status); | |
2732 | goto bail; | |
2733 | } | |
2734 | ||
34d024f8 MF |
2735 | /* launch downconvert thread */ |
2736 | osb->dc_task = kthread_run(ocfs2_downconvert_thread, osb, "ocfs2dc"); | |
2737 | if (IS_ERR(osb->dc_task)) { | |
2738 | status = PTR_ERR(osb->dc_task); | |
2739 | osb->dc_task = NULL; | |
ccd979bd MF |
2740 | mlog_errno(status); |
2741 | goto bail; | |
2742 | } | |
2743 | ||
ccd979bd | 2744 | /* for now, uuid == domain */ |
9c6c877c JB |
2745 | status = ocfs2_cluster_connect(osb->osb_cluster_stack, |
2746 | osb->uuid_str, | |
4670c46d JB |
2747 | strlen(osb->uuid_str), |
2748 | ocfs2_do_node_down, osb, | |
2749 | &conn); | |
2750 | if (status) { | |
ccd979bd MF |
2751 | mlog_errno(status); |
2752 | goto bail; | |
2753 | } | |
2754 | ||
0abd6d18 MF |
2755 | status = ocfs2_cluster_this_node(&osb->node_num); |
2756 | if (status < 0) { | |
2757 | mlog_errno(status); | |
2758 | mlog(ML_ERROR, | |
2759 | "could not find this host's node number\n"); | |
286eaa95 | 2760 | ocfs2_cluster_disconnect(conn, 0); |
0abd6d18 MF |
2761 | goto bail; |
2762 | } | |
2763 | ||
c271c5c2 | 2764 | local: |
ccd979bd MF |
2765 | ocfs2_super_lock_res_init(&osb->osb_super_lockres, osb); |
2766 | ocfs2_rename_lock_res_init(&osb->osb_rename_lockres, osb); | |
2767 | ||
4670c46d | 2768 | osb->cconn = conn; |
ccd979bd MF |
2769 | |
2770 | status = 0; | |
2771 | bail: | |
2772 | if (status < 0) { | |
2773 | ocfs2_dlm_shutdown_debug(osb); | |
34d024f8 MF |
2774 | if (osb->dc_task) |
2775 | kthread_stop(osb->dc_task); | |
ccd979bd MF |
2776 | } |
2777 | ||
2778 | mlog_exit(status); | |
2779 | return status; | |
2780 | } | |
2781 | ||
286eaa95 JB |
2782 | void ocfs2_dlm_shutdown(struct ocfs2_super *osb, |
2783 | int hangup_pending) | |
ccd979bd MF |
2784 | { |
2785 | mlog_entry_void(); | |
2786 | ||
ccd979bd MF |
2787 | ocfs2_drop_osb_locks(osb); |
2788 | ||
4670c46d JB |
2789 | /* |
2790 | * Now that we have dropped all locks and ocfs2_dismount_volume() | |
2791 | * has disabled recovery, the DLM won't be talking to us. It's | |
2792 | * safe to tear things down before disconnecting the cluster. | |
2793 | */ | |
2794 | ||
34d024f8 MF |
2795 | if (osb->dc_task) { |
2796 | kthread_stop(osb->dc_task); | |
2797 | osb->dc_task = NULL; | |
ccd979bd MF |
2798 | } |
2799 | ||
2800 | ocfs2_lock_res_free(&osb->osb_super_lockres); | |
2801 | ocfs2_lock_res_free(&osb->osb_rename_lockres); | |
2802 | ||
286eaa95 | 2803 | ocfs2_cluster_disconnect(osb->cconn, hangup_pending); |
4670c46d | 2804 | osb->cconn = NULL; |
ccd979bd MF |
2805 | |
2806 | ocfs2_dlm_shutdown_debug(osb); | |
2807 | ||
2808 | mlog_exit_void(); | |
2809 | } | |
2810 | ||
7431cd7e | 2811 | static void ocfs2_unlock_ast(void *opaque, int error) |
ccd979bd MF |
2812 | { |
2813 | struct ocfs2_lock_res *lockres = opaque; | |
2814 | unsigned long flags; | |
2815 | ||
2816 | mlog_entry_void(); | |
2817 | ||
2818 | mlog(0, "UNLOCK AST called on lock %s, action = %d\n", lockres->l_name, | |
2819 | lockres->l_unlock_action); | |
2820 | ||
2821 | spin_lock_irqsave(&lockres->l_lock, flags); | |
de551246 | 2822 | if (error) { |
7431cd7e JB |
2823 | mlog(ML_ERROR, "Dlm passes error %d for lock %s, " |
2824 | "unlock_action %d\n", error, lockres->l_name, | |
ccd979bd MF |
2825 | lockres->l_unlock_action); |
2826 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
2827 | return; | |
2828 | } | |
2829 | ||
2830 | switch(lockres->l_unlock_action) { | |
2831 | case OCFS2_UNLOCK_CANCEL_CONVERT: | |
2832 | mlog(0, "Cancel convert success for %s\n", lockres->l_name); | |
2833 | lockres->l_action = OCFS2_AST_INVALID; | |
2834 | break; | |
2835 | case OCFS2_UNLOCK_DROP_LOCK: | |
bd3e7610 | 2836 | lockres->l_level = DLM_LOCK_IV; |
ccd979bd MF |
2837 | break; |
2838 | default: | |
2839 | BUG(); | |
2840 | } | |
2841 | ||
2842 | lockres_clear_flags(lockres, OCFS2_LOCK_BUSY); | |
ccd979bd | 2843 | lockres->l_unlock_action = OCFS2_UNLOCK_INVALID; |
ccd979bd | 2844 | wake_up(&lockres->l_event); |
07f9eebc | 2845 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
ccd979bd MF |
2846 | |
2847 | mlog_exit_void(); | |
2848 | } | |
2849 | ||
ccd979bd | 2850 | static int ocfs2_drop_lock(struct ocfs2_super *osb, |
0d5dc6c2 | 2851 | struct ocfs2_lock_res *lockres) |
ccd979bd | 2852 | { |
7431cd7e | 2853 | int ret; |
ccd979bd | 2854 | unsigned long flags; |
bd3e7610 | 2855 | u32 lkm_flags = 0; |
ccd979bd MF |
2856 | |
2857 | /* We didn't get anywhere near actually using this lockres. */ | |
2858 | if (!(lockres->l_flags & OCFS2_LOCK_INITIALIZED)) | |
2859 | goto out; | |
2860 | ||
b80fc012 | 2861 | if (lockres->l_ops->flags & LOCK_TYPE_USES_LVB) |
bd3e7610 | 2862 | lkm_flags |= DLM_LKF_VALBLK; |
b80fc012 | 2863 | |
ccd979bd MF |
2864 | spin_lock_irqsave(&lockres->l_lock, flags); |
2865 | ||
2866 | mlog_bug_on_msg(!(lockres->l_flags & OCFS2_LOCK_FREEING), | |
2867 | "lockres %s, flags 0x%lx\n", | |
2868 | lockres->l_name, lockres->l_flags); | |
2869 | ||
2870 | while (lockres->l_flags & OCFS2_LOCK_BUSY) { | |
2871 | mlog(0, "waiting on busy lock \"%s\": flags = %lx, action = " | |
2872 | "%u, unlock_action = %u\n", | |
2873 | lockres->l_name, lockres->l_flags, lockres->l_action, | |
2874 | lockres->l_unlock_action); | |
2875 | ||
2876 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
2877 | ||
2878 | /* XXX: Today we just wait on any busy | |
2879 | * locks... Perhaps we need to cancel converts in the | |
2880 | * future? */ | |
2881 | ocfs2_wait_on_busy_lock(lockres); | |
2882 | ||
2883 | spin_lock_irqsave(&lockres->l_lock, flags); | |
2884 | } | |
2885 | ||
0d5dc6c2 MF |
2886 | if (lockres->l_ops->flags & LOCK_TYPE_USES_LVB) { |
2887 | if (lockres->l_flags & OCFS2_LOCK_ATTACHED && | |
bd3e7610 | 2888 | lockres->l_level == DLM_LOCK_EX && |
0d5dc6c2 MF |
2889 | !(lockres->l_flags & OCFS2_LOCK_NEEDS_REFRESH)) |
2890 | lockres->l_ops->set_lvb(lockres); | |
2891 | } | |
ccd979bd MF |
2892 | |
2893 | if (lockres->l_flags & OCFS2_LOCK_BUSY) | |
2894 | mlog(ML_ERROR, "destroying busy lock: \"%s\"\n", | |
2895 | lockres->l_name); | |
2896 | if (lockres->l_flags & OCFS2_LOCK_BLOCKED) | |
2897 | mlog(0, "destroying blocked lock: \"%s\"\n", lockres->l_name); | |
2898 | ||
2899 | if (!(lockres->l_flags & OCFS2_LOCK_ATTACHED)) { | |
2900 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
2901 | goto out; | |
2902 | } | |
2903 | ||
2904 | lockres_clear_flags(lockres, OCFS2_LOCK_ATTACHED); | |
2905 | ||
2906 | /* make sure we never get here while waiting for an ast to | |
2907 | * fire. */ | |
2908 | BUG_ON(lockres->l_action != OCFS2_AST_INVALID); | |
2909 | ||
2910 | /* is this necessary? */ | |
2911 | lockres_or_flags(lockres, OCFS2_LOCK_BUSY); | |
2912 | lockres->l_unlock_action = OCFS2_UNLOCK_DROP_LOCK; | |
2913 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
2914 | ||
2915 | mlog(0, "lock %s\n", lockres->l_name); | |
2916 | ||
4670c46d | 2917 | ret = ocfs2_dlm_unlock(osb->cconn, &lockres->l_lksb, lkm_flags, |
7431cd7e JB |
2918 | lockres); |
2919 | if (ret) { | |
2920 | ocfs2_log_dlm_error("ocfs2_dlm_unlock", ret, lockres); | |
ccd979bd | 2921 | mlog(ML_ERROR, "lockres flags: %lu\n", lockres->l_flags); |
cf0acdcd | 2922 | ocfs2_dlm_dump_lksb(&lockres->l_lksb); |
ccd979bd MF |
2923 | BUG(); |
2924 | } | |
24ef1815 | 2925 | mlog(0, "lock %s, successfull return from ocfs2_dlm_unlock\n", |
ccd979bd MF |
2926 | lockres->l_name); |
2927 | ||
2928 | ocfs2_wait_on_busy_lock(lockres); | |
2929 | out: | |
2930 | mlog_exit(0); | |
2931 | return 0; | |
2932 | } | |
2933 | ||
2934 | /* Mark the lockres as being dropped. It will no longer be | |
2935 | * queued if blocking, but we still may have to wait on it | |
34d024f8 | 2936 | * being dequeued from the downconvert thread before we can consider |
ccd979bd MF |
2937 | * it safe to drop. |
2938 | * | |
2939 | * You can *not* attempt to call cluster_lock on this lockres anymore. */ | |
2940 | void ocfs2_mark_lockres_freeing(struct ocfs2_lock_res *lockres) | |
2941 | { | |
2942 | int status; | |
2943 | struct ocfs2_mask_waiter mw; | |
2944 | unsigned long flags; | |
2945 | ||
2946 | ocfs2_init_mask_waiter(&mw); | |
2947 | ||
2948 | spin_lock_irqsave(&lockres->l_lock, flags); | |
2949 | lockres->l_flags |= OCFS2_LOCK_FREEING; | |
2950 | while (lockres->l_flags & OCFS2_LOCK_QUEUED) { | |
2951 | lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_QUEUED, 0); | |
2952 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
2953 | ||
2954 | mlog(0, "Waiting on lockres %s\n", lockres->l_name); | |
2955 | ||
2956 | status = ocfs2_wait_for_mask(&mw); | |
2957 | if (status) | |
2958 | mlog_errno(status); | |
2959 | ||
2960 | spin_lock_irqsave(&lockres->l_lock, flags); | |
2961 | } | |
2962 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
2963 | } | |
2964 | ||
d680efe9 MF |
2965 | void ocfs2_simple_drop_lockres(struct ocfs2_super *osb, |
2966 | struct ocfs2_lock_res *lockres) | |
ccd979bd | 2967 | { |
d680efe9 | 2968 | int ret; |
ccd979bd | 2969 | |
d680efe9 | 2970 | ocfs2_mark_lockres_freeing(lockres); |
0d5dc6c2 | 2971 | ret = ocfs2_drop_lock(osb, lockres); |
d680efe9 MF |
2972 | if (ret) |
2973 | mlog_errno(ret); | |
2974 | } | |
ccd979bd | 2975 | |
d680efe9 MF |
2976 | static void ocfs2_drop_osb_locks(struct ocfs2_super *osb) |
2977 | { | |
2978 | ocfs2_simple_drop_lockres(osb, &osb->osb_super_lockres); | |
2979 | ocfs2_simple_drop_lockres(osb, &osb->osb_rename_lockres); | |
ccd979bd MF |
2980 | } |
2981 | ||
ccd979bd MF |
2982 | int ocfs2_drop_inode_locks(struct inode *inode) |
2983 | { | |
2984 | int status, err; | |
ccd979bd MF |
2985 | |
2986 | mlog_entry_void(); | |
2987 | ||
2988 | /* No need to call ocfs2_mark_lockres_freeing here - | |
2989 | * ocfs2_clear_inode has done it for us. */ | |
2990 | ||
2991 | err = ocfs2_drop_lock(OCFS2_SB(inode->i_sb), | |
50008630 | 2992 | &OCFS2_I(inode)->ip_open_lockres); |
ccd979bd MF |
2993 | if (err < 0) |
2994 | mlog_errno(err); | |
2995 | ||
2996 | status = err; | |
2997 | ||
2998 | err = ocfs2_drop_lock(OCFS2_SB(inode->i_sb), | |
e63aecb6 | 2999 | &OCFS2_I(inode)->ip_inode_lockres); |
ccd979bd MF |
3000 | if (err < 0) |
3001 | mlog_errno(err); | |
3002 | if (err < 0 && !status) | |
3003 | status = err; | |
3004 | ||
3005 | err = ocfs2_drop_lock(OCFS2_SB(inode->i_sb), | |
0d5dc6c2 | 3006 | &OCFS2_I(inode)->ip_rw_lockres); |
ccd979bd MF |
3007 | if (err < 0) |
3008 | mlog_errno(err); | |
3009 | if (err < 0 && !status) | |
3010 | status = err; | |
3011 | ||
3012 | mlog_exit(status); | |
3013 | return status; | |
3014 | } | |
3015 | ||
de551246 JB |
3016 | static unsigned int ocfs2_prepare_downconvert(struct ocfs2_lock_res *lockres, |
3017 | int new_level) | |
ccd979bd MF |
3018 | { |
3019 | assert_spin_locked(&lockres->l_lock); | |
3020 | ||
bd3e7610 | 3021 | BUG_ON(lockres->l_blocking <= DLM_LOCK_NL); |
ccd979bd MF |
3022 | |
3023 | if (lockres->l_level <= new_level) { | |
bd3e7610 | 3024 | mlog(ML_ERROR, "lockres->l_level (%d) <= new_level (%d)\n", |
ccd979bd MF |
3025 | lockres->l_level, new_level); |
3026 | BUG(); | |
3027 | } | |
3028 | ||
3029 | mlog(0, "lock %s, new_level = %d, l_blocking = %d\n", | |
3030 | lockres->l_name, new_level, lockres->l_blocking); | |
3031 | ||
3032 | lockres->l_action = OCFS2_AST_DOWNCONVERT; | |
3033 | lockres->l_requested = new_level; | |
3034 | lockres_or_flags(lockres, OCFS2_LOCK_BUSY); | |
de551246 | 3035 | return lockres_set_pending(lockres); |
ccd979bd MF |
3036 | } |
3037 | ||
3038 | static int ocfs2_downconvert_lock(struct ocfs2_super *osb, | |
3039 | struct ocfs2_lock_res *lockres, | |
3040 | int new_level, | |
de551246 JB |
3041 | int lvb, |
3042 | unsigned int generation) | |
ccd979bd | 3043 | { |
bd3e7610 JB |
3044 | int ret; |
3045 | u32 dlm_flags = DLM_LKF_CONVERT; | |
ccd979bd MF |
3046 | |
3047 | mlog_entry_void(); | |
3048 | ||
3049 | if (lvb) | |
bd3e7610 | 3050 | dlm_flags |= DLM_LKF_VALBLK; |
ccd979bd | 3051 | |
4670c46d | 3052 | ret = ocfs2_dlm_lock(osb->cconn, |
7431cd7e JB |
3053 | new_level, |
3054 | &lockres->l_lksb, | |
3055 | dlm_flags, | |
3056 | lockres->l_name, | |
3057 | OCFS2_LOCK_ID_MAX_LEN - 1, | |
3058 | lockres); | |
de551246 | 3059 | lockres_clear_pending(lockres, generation, osb); |
7431cd7e JB |
3060 | if (ret) { |
3061 | ocfs2_log_dlm_error("ocfs2_dlm_lock", ret, lockres); | |
ccd979bd MF |
3062 | ocfs2_recover_from_dlm_error(lockres, 1); |
3063 | goto bail; | |
3064 | } | |
3065 | ||
3066 | ret = 0; | |
3067 | bail: | |
3068 | mlog_exit(ret); | |
3069 | return ret; | |
3070 | } | |
3071 | ||
24ef1815 | 3072 | /* returns 1 when the caller should unlock and call ocfs2_dlm_unlock */ |
ccd979bd MF |
3073 | static int ocfs2_prepare_cancel_convert(struct ocfs2_super *osb, |
3074 | struct ocfs2_lock_res *lockres) | |
3075 | { | |
3076 | assert_spin_locked(&lockres->l_lock); | |
3077 | ||
3078 | mlog_entry_void(); | |
3079 | mlog(0, "lock %s\n", lockres->l_name); | |
3080 | ||
3081 | if (lockres->l_unlock_action == OCFS2_UNLOCK_CANCEL_CONVERT) { | |
3082 | /* If we're already trying to cancel a lock conversion | |
3083 | * then just drop the spinlock and allow the caller to | |
3084 | * requeue this lock. */ | |
3085 | ||
3086 | mlog(0, "Lockres %s, skip convert\n", lockres->l_name); | |
3087 | return 0; | |
3088 | } | |
3089 | ||
3090 | /* were we in a convert when we got the bast fire? */ | |
3091 | BUG_ON(lockres->l_action != OCFS2_AST_CONVERT && | |
3092 | lockres->l_action != OCFS2_AST_DOWNCONVERT); | |
3093 | /* set things up for the unlockast to know to just | |
3094 | * clear out the ast_action and unset busy, etc. */ | |
3095 | lockres->l_unlock_action = OCFS2_UNLOCK_CANCEL_CONVERT; | |
3096 | ||
3097 | mlog_bug_on_msg(!(lockres->l_flags & OCFS2_LOCK_BUSY), | |
3098 | "lock %s, invalid flags: 0x%lx\n", | |
3099 | lockres->l_name, lockres->l_flags); | |
3100 | ||
3101 | return 1; | |
3102 | } | |
3103 | ||
3104 | static int ocfs2_cancel_convert(struct ocfs2_super *osb, | |
3105 | struct ocfs2_lock_res *lockres) | |
3106 | { | |
3107 | int ret; | |
ccd979bd MF |
3108 | |
3109 | mlog_entry_void(); | |
3110 | mlog(0, "lock %s\n", lockres->l_name); | |
3111 | ||
4670c46d | 3112 | ret = ocfs2_dlm_unlock(osb->cconn, &lockres->l_lksb, |
7431cd7e JB |
3113 | DLM_LKF_CANCEL, lockres); |
3114 | if (ret) { | |
3115 | ocfs2_log_dlm_error("ocfs2_dlm_unlock", ret, lockres); | |
ccd979bd MF |
3116 | ocfs2_recover_from_dlm_error(lockres, 0); |
3117 | } | |
3118 | ||
24ef1815 | 3119 | mlog(0, "lock %s return from ocfs2_dlm_unlock\n", lockres->l_name); |
ccd979bd | 3120 | |
ccd979bd MF |
3121 | mlog_exit(ret); |
3122 | return ret; | |
3123 | } | |
3124 | ||
b5e500e2 MF |
3125 | static int ocfs2_unblock_lock(struct ocfs2_super *osb, |
3126 | struct ocfs2_lock_res *lockres, | |
3127 | struct ocfs2_unblock_ctl *ctl) | |
ccd979bd MF |
3128 | { |
3129 | unsigned long flags; | |
3130 | int blocking; | |
3131 | int new_level; | |
3132 | int ret = 0; | |
5ef0d4ea | 3133 | int set_lvb = 0; |
de551246 | 3134 | unsigned int gen; |
ccd979bd MF |
3135 | |
3136 | mlog_entry_void(); | |
3137 | ||
3138 | spin_lock_irqsave(&lockres->l_lock, flags); | |
3139 | ||
3140 | BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BLOCKED)); | |
3141 | ||
3142 | recheck: | |
3143 | if (lockres->l_flags & OCFS2_LOCK_BUSY) { | |
de551246 JB |
3144 | /* XXX |
3145 | * This is a *big* race. The OCFS2_LOCK_PENDING flag | |
3146 | * exists entirely for one reason - another thread has set | |
3147 | * OCFS2_LOCK_BUSY, but has *NOT* yet called dlm_lock(). | |
3148 | * | |
3149 | * If we do ocfs2_cancel_convert() before the other thread | |
3150 | * calls dlm_lock(), our cancel will do nothing. We will | |
3151 | * get no ast, and we will have no way of knowing the | |
3152 | * cancel failed. Meanwhile, the other thread will call | |
3153 | * into dlm_lock() and wait...forever. | |
3154 | * | |
3155 | * Why forever? Because another node has asked for the | |
3156 | * lock first; that's why we're here in unblock_lock(). | |
3157 | * | |
3158 | * The solution is OCFS2_LOCK_PENDING. When PENDING is | |
3159 | * set, we just requeue the unblock. Only when the other | |
3160 | * thread has called dlm_lock() and cleared PENDING will | |
3161 | * we then cancel their request. | |
3162 | * | |
3163 | * All callers of dlm_lock() must set OCFS2_DLM_PENDING | |
3164 | * at the same time they set OCFS2_DLM_BUSY. They must | |
3165 | * clear OCFS2_DLM_PENDING after dlm_lock() returns. | |
3166 | */ | |
3167 | if (lockres->l_flags & OCFS2_LOCK_PENDING) | |
3168 | goto leave_requeue; | |
3169 | ||
d680efe9 | 3170 | ctl->requeue = 1; |
ccd979bd MF |
3171 | ret = ocfs2_prepare_cancel_convert(osb, lockres); |
3172 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
3173 | if (ret) { | |
3174 | ret = ocfs2_cancel_convert(osb, lockres); | |
3175 | if (ret < 0) | |
3176 | mlog_errno(ret); | |
3177 | } | |
3178 | goto leave; | |
3179 | } | |
3180 | ||
3181 | /* if we're blocking an exclusive and we have *any* holders, | |
3182 | * then requeue. */ | |
bd3e7610 | 3183 | if ((lockres->l_blocking == DLM_LOCK_EX) |
f7fbfdd1 MF |
3184 | && (lockres->l_ex_holders || lockres->l_ro_holders)) |
3185 | goto leave_requeue; | |
ccd979bd MF |
3186 | |
3187 | /* If it's a PR we're blocking, then only | |
3188 | * requeue if we've got any EX holders */ | |
bd3e7610 | 3189 | if (lockres->l_blocking == DLM_LOCK_PR && |
f7fbfdd1 MF |
3190 | lockres->l_ex_holders) |
3191 | goto leave_requeue; | |
3192 | ||
3193 | /* | |
3194 | * Can we get a lock in this state if the holder counts are | |
3195 | * zero? The meta data unblock code used to check this. | |
3196 | */ | |
3197 | if ((lockres->l_ops->flags & LOCK_TYPE_REQUIRES_REFRESH) | |
3198 | && (lockres->l_flags & OCFS2_LOCK_REFRESHING)) | |
3199 | goto leave_requeue; | |
ccd979bd | 3200 | |
16d5b956 MF |
3201 | new_level = ocfs2_highest_compat_lock_level(lockres->l_blocking); |
3202 | ||
3203 | if (lockres->l_ops->check_downconvert | |
3204 | && !lockres->l_ops->check_downconvert(lockres, new_level)) | |
3205 | goto leave_requeue; | |
3206 | ||
ccd979bd MF |
3207 | /* If we get here, then we know that there are no more |
3208 | * incompatible holders (and anyone asking for an incompatible | |
3209 | * lock is blocked). We can now downconvert the lock */ | |
cc567d89 | 3210 | if (!lockres->l_ops->downconvert_worker) |
ccd979bd MF |
3211 | goto downconvert; |
3212 | ||
3213 | /* Some lockres types want to do a bit of work before | |
3214 | * downconverting a lock. Allow that here. The worker function | |
3215 | * may sleep, so we save off a copy of what we're blocking as | |
3216 | * it may change while we're not holding the spin lock. */ | |
3217 | blocking = lockres->l_blocking; | |
3218 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
3219 | ||
cc567d89 | 3220 | ctl->unblock_action = lockres->l_ops->downconvert_worker(lockres, blocking); |
d680efe9 MF |
3221 | |
3222 | if (ctl->unblock_action == UNBLOCK_STOP_POST) | |
3223 | goto leave; | |
ccd979bd MF |
3224 | |
3225 | spin_lock_irqsave(&lockres->l_lock, flags); | |
3226 | if (blocking != lockres->l_blocking) { | |
3227 | /* If this changed underneath us, then we can't drop | |
3228 | * it just yet. */ | |
3229 | goto recheck; | |
3230 | } | |
3231 | ||
3232 | downconvert: | |
d680efe9 | 3233 | ctl->requeue = 0; |
ccd979bd | 3234 | |
5ef0d4ea | 3235 | if (lockres->l_ops->flags & LOCK_TYPE_USES_LVB) { |
bd3e7610 | 3236 | if (lockres->l_level == DLM_LOCK_EX) |
5ef0d4ea MF |
3237 | set_lvb = 1; |
3238 | ||
3239 | /* | |
3240 | * We only set the lvb if the lock has been fully | |
3241 | * refreshed - otherwise we risk setting stale | |
3242 | * data. Otherwise, there's no need to actually clear | |
3243 | * out the lvb here as it's value is still valid. | |
3244 | */ | |
3245 | if (set_lvb && !(lockres->l_flags & OCFS2_LOCK_NEEDS_REFRESH)) | |
3246 | lockres->l_ops->set_lvb(lockres); | |
3247 | } | |
3248 | ||
de551246 | 3249 | gen = ocfs2_prepare_downconvert(lockres, new_level); |
ccd979bd | 3250 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
de551246 JB |
3251 | ret = ocfs2_downconvert_lock(osb, lockres, new_level, set_lvb, |
3252 | gen); | |
3253 | ||
ccd979bd MF |
3254 | leave: |
3255 | mlog_exit(ret); | |
3256 | return ret; | |
f7fbfdd1 MF |
3257 | |
3258 | leave_requeue: | |
3259 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
3260 | ctl->requeue = 1; | |
3261 | ||
3262 | mlog_exit(0); | |
3263 | return 0; | |
ccd979bd MF |
3264 | } |
3265 | ||
d680efe9 MF |
3266 | static int ocfs2_data_convert_worker(struct ocfs2_lock_res *lockres, |
3267 | int blocking) | |
ccd979bd MF |
3268 | { |
3269 | struct inode *inode; | |
3270 | struct address_space *mapping; | |
3271 | ||
ccd979bd MF |
3272 | inode = ocfs2_lock_res_inode(lockres); |
3273 | mapping = inode->i_mapping; | |
3274 | ||
1044e401 | 3275 | if (!S_ISREG(inode->i_mode)) |
f1f54068 MF |
3276 | goto out; |
3277 | ||
7f4a2a97 MF |
3278 | /* |
3279 | * We need this before the filemap_fdatawrite() so that it can | |
3280 | * transfer the dirty bit from the PTE to the | |
3281 | * page. Unfortunately this means that even for EX->PR | |
3282 | * downconverts, we'll lose our mappings and have to build | |
3283 | * them up again. | |
3284 | */ | |
3285 | unmap_mapping_range(mapping, 0, 0, 0); | |
3286 | ||
ccd979bd | 3287 | if (filemap_fdatawrite(mapping)) { |
b0697053 MF |
3288 | mlog(ML_ERROR, "Could not sync inode %llu for downconvert!", |
3289 | (unsigned long long)OCFS2_I(inode)->ip_blkno); | |
ccd979bd MF |
3290 | } |
3291 | sync_mapping_buffers(mapping); | |
bd3e7610 | 3292 | if (blocking == DLM_LOCK_EX) { |
ccd979bd | 3293 | truncate_inode_pages(mapping, 0); |
ccd979bd MF |
3294 | } else { |
3295 | /* We only need to wait on the I/O if we're not also | |
3296 | * truncating pages because truncate_inode_pages waits | |
3297 | * for us above. We don't truncate pages if we're | |
3298 | * blocking anything < EXMODE because we want to keep | |
3299 | * them around in that case. */ | |
3300 | filemap_fdatawait(mapping); | |
3301 | } | |
3302 | ||
f1f54068 | 3303 | out: |
d680efe9 | 3304 | return UNBLOCK_CONTINUE; |
ccd979bd MF |
3305 | } |
3306 | ||
810d5aeb MF |
3307 | static int ocfs2_check_meta_downconvert(struct ocfs2_lock_res *lockres, |
3308 | int new_level) | |
3309 | { | |
3310 | struct inode *inode = ocfs2_lock_res_inode(lockres); | |
3311 | int checkpointed = ocfs2_inode_fully_checkpointed(inode); | |
3312 | ||
bd3e7610 JB |
3313 | BUG_ON(new_level != DLM_LOCK_NL && new_level != DLM_LOCK_PR); |
3314 | BUG_ON(lockres->l_level != DLM_LOCK_EX && !checkpointed); | |
810d5aeb MF |
3315 | |
3316 | if (checkpointed) | |
3317 | return 1; | |
3318 | ||
3319 | ocfs2_start_checkpoint(OCFS2_SB(inode->i_sb)); | |
3320 | return 0; | |
3321 | } | |
3322 | ||
3323 | static void ocfs2_set_meta_lvb(struct ocfs2_lock_res *lockres) | |
3324 | { | |
3325 | struct inode *inode = ocfs2_lock_res_inode(lockres); | |
3326 | ||
3327 | __ocfs2_stuff_meta_lvb(inode); | |
3328 | } | |
3329 | ||
d680efe9 MF |
3330 | /* |
3331 | * Does the final reference drop on our dentry lock. Right now this | |
34d024f8 | 3332 | * happens in the downconvert thread, but we could choose to simplify the |
d680efe9 MF |
3333 | * dlmglue API and push these off to the ocfs2_wq in the future. |
3334 | */ | |
3335 | static void ocfs2_dentry_post_unlock(struct ocfs2_super *osb, | |
3336 | struct ocfs2_lock_res *lockres) | |
3337 | { | |
3338 | struct ocfs2_dentry_lock *dl = ocfs2_lock_res_dl(lockres); | |
3339 | ocfs2_dentry_lock_put(osb, dl); | |
3340 | } | |
3341 | ||
3342 | /* | |
3343 | * d_delete() matching dentries before the lock downconvert. | |
3344 | * | |
3345 | * At this point, any process waiting to destroy the | |
3346 | * dentry_lock due to last ref count is stopped by the | |
3347 | * OCFS2_LOCK_QUEUED flag. | |
3348 | * | |
3349 | * We have two potential problems | |
3350 | * | |
3351 | * 1) If we do the last reference drop on our dentry_lock (via dput) | |
3352 | * we'll wind up in ocfs2_release_dentry_lock(), waiting on | |
3353 | * the downconvert to finish. Instead we take an elevated | |
3354 | * reference and push the drop until after we've completed our | |
3355 | * unblock processing. | |
3356 | * | |
3357 | * 2) There might be another process with a final reference, | |
3358 | * waiting on us to finish processing. If this is the case, we | |
3359 | * detect it and exit out - there's no more dentries anyway. | |
3360 | */ | |
3361 | static int ocfs2_dentry_convert_worker(struct ocfs2_lock_res *lockres, | |
3362 | int blocking) | |
3363 | { | |
3364 | struct ocfs2_dentry_lock *dl = ocfs2_lock_res_dl(lockres); | |
3365 | struct ocfs2_inode_info *oi = OCFS2_I(dl->dl_inode); | |
3366 | struct dentry *dentry; | |
3367 | unsigned long flags; | |
3368 | int extra_ref = 0; | |
3369 | ||
3370 | /* | |
3371 | * This node is blocking another node from getting a read | |
3372 | * lock. This happens when we've renamed within a | |
3373 | * directory. We've forced the other nodes to d_delete(), but | |
3374 | * we never actually dropped our lock because it's still | |
3375 | * valid. The downconvert code will retain a PR for this node, | |
3376 | * so there's no further work to do. | |
3377 | */ | |
bd3e7610 | 3378 | if (blocking == DLM_LOCK_PR) |
d680efe9 MF |
3379 | return UNBLOCK_CONTINUE; |
3380 | ||
3381 | /* | |
3382 | * Mark this inode as potentially orphaned. The code in | |
3383 | * ocfs2_delete_inode() will figure out whether it actually | |
3384 | * needs to be freed or not. | |
3385 | */ | |
3386 | spin_lock(&oi->ip_lock); | |
3387 | oi->ip_flags |= OCFS2_INODE_MAYBE_ORPHANED; | |
3388 | spin_unlock(&oi->ip_lock); | |
3389 | ||
3390 | /* | |
3391 | * Yuck. We need to make sure however that the check of | |
3392 | * OCFS2_LOCK_FREEING and the extra reference are atomic with | |
3393 | * respect to a reference decrement or the setting of that | |
3394 | * flag. | |
3395 | */ | |
3396 | spin_lock_irqsave(&lockres->l_lock, flags); | |
3397 | spin_lock(&dentry_attach_lock); | |
3398 | if (!(lockres->l_flags & OCFS2_LOCK_FREEING) | |
3399 | && dl->dl_count) { | |
3400 | dl->dl_count++; | |
3401 | extra_ref = 1; | |
3402 | } | |
3403 | spin_unlock(&dentry_attach_lock); | |
3404 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
3405 | ||
3406 | mlog(0, "extra_ref = %d\n", extra_ref); | |
3407 | ||
3408 | /* | |
3409 | * We have a process waiting on us in ocfs2_dentry_iput(), | |
3410 | * which means we can't have any more outstanding | |
3411 | * aliases. There's no need to do any more work. | |
3412 | */ | |
3413 | if (!extra_ref) | |
3414 | return UNBLOCK_CONTINUE; | |
3415 | ||
3416 | spin_lock(&dentry_attach_lock); | |
3417 | while (1) { | |
3418 | dentry = ocfs2_find_local_alias(dl->dl_inode, | |
3419 | dl->dl_parent_blkno, 1); | |
3420 | if (!dentry) | |
3421 | break; | |
3422 | spin_unlock(&dentry_attach_lock); | |
3423 | ||
3424 | mlog(0, "d_delete(%.*s);\n", dentry->d_name.len, | |
3425 | dentry->d_name.name); | |
3426 | ||
3427 | /* | |
3428 | * The following dcache calls may do an | |
3429 | * iput(). Normally we don't want that from the | |
3430 | * downconverting thread, but in this case it's ok | |
3431 | * because the requesting node already has an | |
3432 | * exclusive lock on the inode, so it can't be queued | |
3433 | * for a downconvert. | |
3434 | */ | |
3435 | d_delete(dentry); | |
3436 | dput(dentry); | |
3437 | ||
3438 | spin_lock(&dentry_attach_lock); | |
3439 | } | |
3440 | spin_unlock(&dentry_attach_lock); | |
3441 | ||
3442 | /* | |
3443 | * If we are the last holder of this dentry lock, there is no | |
3444 | * reason to downconvert so skip straight to the unlock. | |
3445 | */ | |
3446 | if (dl->dl_count == 1) | |
3447 | return UNBLOCK_STOP_POST; | |
3448 | ||
3449 | return UNBLOCK_CONTINUE_POST; | |
3450 | } | |
3451 | ||
4670c46d JB |
3452 | /* |
3453 | * This is the filesystem locking protocol. It provides the lock handling | |
3454 | * hooks for the underlying DLM. It has a maximum version number. | |
3455 | * The version number allows interoperability with systems running at | |
3456 | * the same major number and an equal or smaller minor number. | |
3457 | * | |
3458 | * Whenever the filesystem does new things with locks (adds or removes a | |
3459 | * lock, orders them differently, does different things underneath a lock), | |
3460 | * the version must be changed. The protocol is negotiated when joining | |
3461 | * the dlm domain. A node may join the domain if its major version is | |
3462 | * identical to all other nodes and its minor version is greater than | |
3463 | * or equal to all other nodes. When its minor version is greater than | |
3464 | * the other nodes, it will run at the minor version specified by the | |
3465 | * other nodes. | |
3466 | * | |
3467 | * If a locking change is made that will not be compatible with older | |
3468 | * versions, the major number must be increased and the minor version set | |
3469 | * to zero. If a change merely adds a behavior that can be disabled when | |
3470 | * speaking to older versions, the minor version must be increased. If a | |
3471 | * change adds a fully backwards compatible change (eg, LVB changes that | |
3472 | * are just ignored by older versions), the version does not need to be | |
3473 | * updated. | |
3474 | */ | |
24ef1815 | 3475 | static struct ocfs2_locking_protocol lproto = { |
4670c46d JB |
3476 | .lp_max_version = { |
3477 | .pv_major = OCFS2_LOCKING_PROTOCOL_MAJOR, | |
3478 | .pv_minor = OCFS2_LOCKING_PROTOCOL_MINOR, | |
3479 | }, | |
24ef1815 JB |
3480 | .lp_lock_ast = ocfs2_locking_ast, |
3481 | .lp_blocking_ast = ocfs2_blocking_ast, | |
3482 | .lp_unlock_ast = ocfs2_unlock_ast, | |
3483 | }; | |
3484 | ||
63e0c48a | 3485 | void ocfs2_set_locking_protocol(void) |
24ef1815 | 3486 | { |
63e0c48a | 3487 | ocfs2_stack_glue_set_locking_protocol(&lproto); |
24ef1815 JB |
3488 | } |
3489 | ||
24ef1815 | 3490 | |
00600056 AB |
3491 | static void ocfs2_process_blocked_lock(struct ocfs2_super *osb, |
3492 | struct ocfs2_lock_res *lockres) | |
ccd979bd MF |
3493 | { |
3494 | int status; | |
d680efe9 | 3495 | struct ocfs2_unblock_ctl ctl = {0, 0,}; |
ccd979bd MF |
3496 | unsigned long flags; |
3497 | ||
3498 | /* Our reference to the lockres in this function can be | |
3499 | * considered valid until we remove the OCFS2_LOCK_QUEUED | |
3500 | * flag. */ | |
3501 | ||
3502 | mlog_entry_void(); | |
3503 | ||
3504 | BUG_ON(!lockres); | |
3505 | BUG_ON(!lockres->l_ops); | |
ccd979bd MF |
3506 | |
3507 | mlog(0, "lockres %s blocked.\n", lockres->l_name); | |
3508 | ||
3509 | /* Detect whether a lock has been marked as going away while | |
34d024f8 | 3510 | * the downconvert thread was processing other things. A lock can |
ccd979bd MF |
3511 | * still be marked with OCFS2_LOCK_FREEING after this check, |
3512 | * but short circuiting here will still save us some | |
3513 | * performance. */ | |
3514 | spin_lock_irqsave(&lockres->l_lock, flags); | |
3515 | if (lockres->l_flags & OCFS2_LOCK_FREEING) | |
3516 | goto unqueue; | |
3517 | spin_unlock_irqrestore(&lockres->l_lock, flags); | |
3518 | ||
b5e500e2 | 3519 | status = ocfs2_unblock_lock(osb, lockres, &ctl); |
ccd979bd MF |
3520 | if (status < 0) |
3521 | mlog_errno(status); | |
3522 | ||
3523 | spin_lock_irqsave(&lockres->l_lock, flags); | |
3524 | unqueue: | |
d680efe9 | 3525 | if (lockres->l_flags & OCFS2_LOCK_FREEING || !ctl.requeue) { |
ccd979bd MF |
3526 | lockres_clear_flags(lockres, OCFS2_LOCK_QUEUED); |
3527 | } else | |
3528 | ocfs2_schedule_blocked_lock(osb, lockres); | |
3529 | ||
3530 | mlog(0, "lockres %s, requeue = %s.\n", lockres->l_name, | |
d680efe9 | 3531 | ctl.requeue ? "yes" : "no"); |
ccd979bd MF |
3532 | spin_unlock_irqrestore(&lockres->l_lock, flags); |
3533 | ||
d680efe9 MF |
3534 | if (ctl.unblock_action != UNBLOCK_CONTINUE |
3535 | && lockres->l_ops->post_unlock) | |
3536 | lockres->l_ops->post_unlock(osb, lockres); | |
3537 | ||
ccd979bd MF |
3538 | mlog_exit_void(); |
3539 | } | |
3540 | ||
3541 | static void ocfs2_schedule_blocked_lock(struct ocfs2_super *osb, | |
3542 | struct ocfs2_lock_res *lockres) | |
3543 | { | |
3544 | mlog_entry_void(); | |
3545 | ||
3546 | assert_spin_locked(&lockres->l_lock); | |
3547 | ||
3548 | if (lockres->l_flags & OCFS2_LOCK_FREEING) { | |
3549 | /* Do not schedule a lock for downconvert when it's on | |
3550 | * the way to destruction - any nodes wanting access | |
3551 | * to the resource will get it soon. */ | |
3552 | mlog(0, "Lockres %s won't be scheduled: flags 0x%lx\n", | |
3553 | lockres->l_name, lockres->l_flags); | |
3554 | return; | |
3555 | } | |
3556 | ||
3557 | lockres_or_flags(lockres, OCFS2_LOCK_QUEUED); | |
3558 | ||
34d024f8 | 3559 | spin_lock(&osb->dc_task_lock); |
ccd979bd MF |
3560 | if (list_empty(&lockres->l_blocked_list)) { |
3561 | list_add_tail(&lockres->l_blocked_list, | |
3562 | &osb->blocked_lock_list); | |
3563 | osb->blocked_lock_count++; | |
3564 | } | |
34d024f8 | 3565 | spin_unlock(&osb->dc_task_lock); |
ccd979bd MF |
3566 | |
3567 | mlog_exit_void(); | |
3568 | } | |
34d024f8 MF |
3569 | |
3570 | static void ocfs2_downconvert_thread_do_work(struct ocfs2_super *osb) | |
3571 | { | |
3572 | unsigned long processed; | |
3573 | struct ocfs2_lock_res *lockres; | |
3574 | ||
3575 | mlog_entry_void(); | |
3576 | ||
3577 | spin_lock(&osb->dc_task_lock); | |
3578 | /* grab this early so we know to try again if a state change and | |
3579 | * wake happens part-way through our work */ | |
3580 | osb->dc_work_sequence = osb->dc_wake_sequence; | |
3581 | ||
3582 | processed = osb->blocked_lock_count; | |
3583 | while (processed) { | |
3584 | BUG_ON(list_empty(&osb->blocked_lock_list)); | |
3585 | ||
3586 | lockres = list_entry(osb->blocked_lock_list.next, | |
3587 | struct ocfs2_lock_res, l_blocked_list); | |
3588 | list_del_init(&lockres->l_blocked_list); | |
3589 | osb->blocked_lock_count--; | |
3590 | spin_unlock(&osb->dc_task_lock); | |
3591 | ||
3592 | BUG_ON(!processed); | |
3593 | processed--; | |
3594 | ||
3595 | ocfs2_process_blocked_lock(osb, lockres); | |
3596 | ||
3597 | spin_lock(&osb->dc_task_lock); | |
3598 | } | |
3599 | spin_unlock(&osb->dc_task_lock); | |
3600 | ||
3601 | mlog_exit_void(); | |
3602 | } | |
3603 | ||
3604 | static int ocfs2_downconvert_thread_lists_empty(struct ocfs2_super *osb) | |
3605 | { | |
3606 | int empty = 0; | |
3607 | ||
3608 | spin_lock(&osb->dc_task_lock); | |
3609 | if (list_empty(&osb->blocked_lock_list)) | |
3610 | empty = 1; | |
3611 | ||
3612 | spin_unlock(&osb->dc_task_lock); | |
3613 | return empty; | |
3614 | } | |
3615 | ||
3616 | static int ocfs2_downconvert_thread_should_wake(struct ocfs2_super *osb) | |
3617 | { | |
3618 | int should_wake = 0; | |
3619 | ||
3620 | spin_lock(&osb->dc_task_lock); | |
3621 | if (osb->dc_work_sequence != osb->dc_wake_sequence) | |
3622 | should_wake = 1; | |
3623 | spin_unlock(&osb->dc_task_lock); | |
3624 | ||
3625 | return should_wake; | |
3626 | } | |
3627 | ||
200bfae3 | 3628 | static int ocfs2_downconvert_thread(void *arg) |
34d024f8 MF |
3629 | { |
3630 | int status = 0; | |
3631 | struct ocfs2_super *osb = arg; | |
3632 | ||
3633 | /* only quit once we've been asked to stop and there is no more | |
3634 | * work available */ | |
3635 | while (!(kthread_should_stop() && | |
3636 | ocfs2_downconvert_thread_lists_empty(osb))) { | |
3637 | ||
3638 | wait_event_interruptible(osb->dc_event, | |
3639 | ocfs2_downconvert_thread_should_wake(osb) || | |
3640 | kthread_should_stop()); | |
3641 | ||
3642 | mlog(0, "downconvert_thread: awoken\n"); | |
3643 | ||
3644 | ocfs2_downconvert_thread_do_work(osb); | |
3645 | } | |
3646 | ||
3647 | osb->dc_task = NULL; | |
3648 | return status; | |
3649 | } | |
3650 | ||
3651 | void ocfs2_wake_downconvert_thread(struct ocfs2_super *osb) | |
3652 | { | |
3653 | spin_lock(&osb->dc_task_lock); | |
3654 | /* make sure the voting thread gets a swipe at whatever changes | |
3655 | * the caller may have made to the voting state */ | |
3656 | osb->dc_wake_sequence++; | |
3657 | spin_unlock(&osb->dc_task_lock); | |
3658 | wake_up(&osb->dc_event); | |
3659 | } |